x86, mce, AMD: Fix leaving freed data in a list
[linux-2.6/linux-acpi-2.6/ibm-acpi-2.6.git] / net / ipv6 / addrconf.c
blob143791da062c5118656f826033bb673435ab09f2
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/kernel.h>
44 #include <linux/socket.h>
45 #include <linux/sockios.h>
46 #include <linux/net.h>
47 #include <linux/in6.h>
48 #include <linux/netdevice.h>
49 #include <linux/if_addr.h>
50 #include <linux/if_arp.h>
51 #include <linux/if_arcnet.h>
52 #include <linux/if_infiniband.h>
53 #include <linux/route.h>
54 #include <linux/inetdevice.h>
55 #include <linux/init.h>
56 #ifdef CONFIG_SYSCTL
57 #include <linux/sysctl.h>
58 #endif
59 #include <linux/capability.h>
60 #include <linux/delay.h>
61 #include <linux/notifier.h>
62 #include <linux/string.h>
64 #include <net/net_namespace.h>
65 #include <net/sock.h>
66 #include <net/snmp.h>
68 #include <net/ipv6.h>
69 #include <net/protocol.h>
70 #include <net/ndisc.h>
71 #include <net/ip6_route.h>
72 #include <net/addrconf.h>
73 #include <net/tcp.h>
74 #include <net/ip.h>
75 #include <net/netlink.h>
76 #include <net/pkt_sched.h>
77 #include <linux/if_tunnel.h>
78 #include <linux/rtnetlink.h>
80 #ifdef CONFIG_IPV6_PRIVACY
81 #include <linux/random.h>
82 #endif
84 #include <asm/uaccess.h>
85 #include <asm/unaligned.h>
87 #include <linux/proc_fs.h>
88 #include <linux/seq_file.h>
90 /* Set to 3 to get tracing... */
91 #define ACONF_DEBUG 2
93 #if ACONF_DEBUG >= 3
94 #define ADBG(x) printk x
95 #else
96 #define ADBG(x)
97 #endif
99 #define INFINITY_LIFE_TIME 0xFFFFFFFF
100 #define TIME_DELTA(a,b) ((unsigned long)((long)(a) - (long)(b)))
102 #ifdef CONFIG_SYSCTL
103 static void addrconf_sysctl_register(struct inet6_dev *idev);
104 static void addrconf_sysctl_unregister(struct inet6_dev *idev);
105 #else
106 static inline void addrconf_sysctl_register(struct inet6_dev *idev)
110 static inline void addrconf_sysctl_unregister(struct inet6_dev *idev)
113 #endif
115 #ifdef CONFIG_IPV6_PRIVACY
116 static int __ipv6_regen_rndid(struct inet6_dev *idev);
117 static int __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr);
118 static void ipv6_regen_rndid(unsigned long data);
120 static int desync_factor = MAX_DESYNC_FACTOR * HZ;
121 #endif
123 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev);
124 static int ipv6_count_addresses(struct inet6_dev *idev);
127 * Configured unicast address hash table
129 static struct inet6_ifaddr *inet6_addr_lst[IN6_ADDR_HSIZE];
130 static DEFINE_RWLOCK(addrconf_hash_lock);
132 static void addrconf_verify(unsigned long);
134 static DEFINE_TIMER(addr_chk_timer, addrconf_verify, 0, 0);
135 static DEFINE_SPINLOCK(addrconf_verify_lock);
137 static void addrconf_join_anycast(struct inet6_ifaddr *ifp);
138 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp);
140 static void addrconf_bonding_change(struct net_device *dev,
141 unsigned long event);
142 static int addrconf_ifdown(struct net_device *dev, int how);
144 static void addrconf_dad_start(struct inet6_ifaddr *ifp, u32 flags);
145 static void addrconf_dad_timer(unsigned long data);
146 static void addrconf_dad_completed(struct inet6_ifaddr *ifp);
147 static void addrconf_dad_run(struct inet6_dev *idev);
148 static void addrconf_rs_timer(unsigned long data);
149 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
150 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
152 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
153 struct prefix_info *pinfo);
154 static int ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
155 struct net_device *dev);
157 static ATOMIC_NOTIFIER_HEAD(inet6addr_chain);
159 static struct ipv6_devconf ipv6_devconf __read_mostly = {
160 .forwarding = 0,
161 .hop_limit = IPV6_DEFAULT_HOPLIMIT,
162 .mtu6 = IPV6_MIN_MTU,
163 .accept_ra = 1,
164 .accept_redirects = 1,
165 .autoconf = 1,
166 .force_mld_version = 0,
167 .dad_transmits = 1,
168 .rtr_solicits = MAX_RTR_SOLICITATIONS,
169 .rtr_solicit_interval = RTR_SOLICITATION_INTERVAL,
170 .rtr_solicit_delay = MAX_RTR_SOLICITATION_DELAY,
171 #ifdef CONFIG_IPV6_PRIVACY
172 .use_tempaddr = 0,
173 .temp_valid_lft = TEMP_VALID_LIFETIME,
174 .temp_prefered_lft = TEMP_PREFERRED_LIFETIME,
175 .regen_max_retry = REGEN_MAX_RETRY,
176 .max_desync_factor = MAX_DESYNC_FACTOR,
177 #endif
178 .max_addresses = IPV6_MAX_ADDRESSES,
179 .accept_ra_defrtr = 1,
180 .accept_ra_pinfo = 1,
181 #ifdef CONFIG_IPV6_ROUTER_PREF
182 .accept_ra_rtr_pref = 1,
183 .rtr_probe_interval = 60 * HZ,
184 #ifdef CONFIG_IPV6_ROUTE_INFO
185 .accept_ra_rt_info_max_plen = 0,
186 #endif
187 #endif
188 .proxy_ndp = 0,
189 .accept_source_route = 0, /* we do not accept RH0 by default. */
190 .disable_ipv6 = 0,
191 .accept_dad = 1,
194 static struct ipv6_devconf ipv6_devconf_dflt __read_mostly = {
195 .forwarding = 0,
196 .hop_limit = IPV6_DEFAULT_HOPLIMIT,
197 .mtu6 = IPV6_MIN_MTU,
198 .accept_ra = 1,
199 .accept_redirects = 1,
200 .autoconf = 1,
201 .dad_transmits = 1,
202 .rtr_solicits = MAX_RTR_SOLICITATIONS,
203 .rtr_solicit_interval = RTR_SOLICITATION_INTERVAL,
204 .rtr_solicit_delay = MAX_RTR_SOLICITATION_DELAY,
205 #ifdef CONFIG_IPV6_PRIVACY
206 .use_tempaddr = 0,
207 .temp_valid_lft = TEMP_VALID_LIFETIME,
208 .temp_prefered_lft = TEMP_PREFERRED_LIFETIME,
209 .regen_max_retry = REGEN_MAX_RETRY,
210 .max_desync_factor = MAX_DESYNC_FACTOR,
211 #endif
212 .max_addresses = IPV6_MAX_ADDRESSES,
213 .accept_ra_defrtr = 1,
214 .accept_ra_pinfo = 1,
215 #ifdef CONFIG_IPV6_ROUTER_PREF
216 .accept_ra_rtr_pref = 1,
217 .rtr_probe_interval = 60 * HZ,
218 #ifdef CONFIG_IPV6_ROUTE_INFO
219 .accept_ra_rt_info_max_plen = 0,
220 #endif
221 #endif
222 .proxy_ndp = 0,
223 .accept_source_route = 0, /* we do not accept RH0 by default. */
224 .disable_ipv6 = 0,
225 .accept_dad = 1,
228 /* IPv6 Wildcard Address and Loopback Address defined by RFC2553 */
229 const struct in6_addr in6addr_any = IN6ADDR_ANY_INIT;
230 const struct in6_addr in6addr_loopback = IN6ADDR_LOOPBACK_INIT;
231 const struct in6_addr in6addr_linklocal_allnodes = IN6ADDR_LINKLOCAL_ALLNODES_INIT;
232 const struct in6_addr in6addr_linklocal_allrouters = IN6ADDR_LINKLOCAL_ALLROUTERS_INIT;
234 /* Check if a valid qdisc is available */
235 static inline bool addrconf_qdisc_ok(const struct net_device *dev)
237 return !qdisc_tx_is_noop(dev);
240 /* Check if a route is valid prefix route */
241 static inline int addrconf_is_prefix_route(const struct rt6_info *rt)
243 return ((rt->rt6i_flags & (RTF_GATEWAY | RTF_DEFAULT)) == 0);
246 static void addrconf_del_timer(struct inet6_ifaddr *ifp)
248 if (del_timer(&ifp->timer))
249 __in6_ifa_put(ifp);
252 enum addrconf_timer_t
254 AC_NONE,
255 AC_DAD,
256 AC_RS,
259 static void addrconf_mod_timer(struct inet6_ifaddr *ifp,
260 enum addrconf_timer_t what,
261 unsigned long when)
263 if (!del_timer(&ifp->timer))
264 in6_ifa_hold(ifp);
266 switch (what) {
267 case AC_DAD:
268 ifp->timer.function = addrconf_dad_timer;
269 break;
270 case AC_RS:
271 ifp->timer.function = addrconf_rs_timer;
272 break;
273 default:;
275 ifp->timer.expires = jiffies + when;
276 add_timer(&ifp->timer);
279 static int snmp6_alloc_dev(struct inet6_dev *idev)
281 if (snmp_mib_init((void **)idev->stats.ipv6,
282 sizeof(struct ipstats_mib)) < 0)
283 goto err_ip;
284 if (snmp_mib_init((void **)idev->stats.icmpv6,
285 sizeof(struct icmpv6_mib)) < 0)
286 goto err_icmp;
287 if (snmp_mib_init((void **)idev->stats.icmpv6msg,
288 sizeof(struct icmpv6msg_mib)) < 0)
289 goto err_icmpmsg;
291 return 0;
293 err_icmpmsg:
294 snmp_mib_free((void **)idev->stats.icmpv6);
295 err_icmp:
296 snmp_mib_free((void **)idev->stats.ipv6);
297 err_ip:
298 return -ENOMEM;
301 static void snmp6_free_dev(struct inet6_dev *idev)
303 snmp_mib_free((void **)idev->stats.icmpv6msg);
304 snmp_mib_free((void **)idev->stats.icmpv6);
305 snmp_mib_free((void **)idev->stats.ipv6);
308 /* Nobody refers to this device, we may destroy it. */
310 static void in6_dev_finish_destroy_rcu(struct rcu_head *head)
312 struct inet6_dev *idev = container_of(head, struct inet6_dev, rcu);
313 kfree(idev);
316 void in6_dev_finish_destroy(struct inet6_dev *idev)
318 struct net_device *dev = idev->dev;
320 WARN_ON(idev->addr_list != NULL);
321 WARN_ON(idev->mc_list != NULL);
323 #ifdef NET_REFCNT_DEBUG
324 printk(KERN_DEBUG "in6_dev_finish_destroy: %s\n", dev ? dev->name : "NIL");
325 #endif
326 dev_put(dev);
327 if (!idev->dead) {
328 printk("Freeing alive inet6 device %p\n", idev);
329 return;
331 snmp6_free_dev(idev);
332 call_rcu(&idev->rcu, in6_dev_finish_destroy_rcu);
335 EXPORT_SYMBOL(in6_dev_finish_destroy);
337 static struct inet6_dev * ipv6_add_dev(struct net_device *dev)
339 struct inet6_dev *ndev;
341 ASSERT_RTNL();
343 if (dev->mtu < IPV6_MIN_MTU)
344 return NULL;
346 ndev = kzalloc(sizeof(struct inet6_dev), GFP_KERNEL);
348 if (ndev == NULL)
349 return NULL;
351 rwlock_init(&ndev->lock);
352 ndev->dev = dev;
353 memcpy(&ndev->cnf, dev_net(dev)->ipv6.devconf_dflt, sizeof(ndev->cnf));
354 ndev->cnf.mtu6 = dev->mtu;
355 ndev->cnf.sysctl = NULL;
356 ndev->nd_parms = neigh_parms_alloc(dev, &nd_tbl);
357 if (ndev->nd_parms == NULL) {
358 kfree(ndev);
359 return NULL;
361 if (ndev->cnf.forwarding)
362 dev_disable_lro(dev);
363 /* We refer to the device */
364 dev_hold(dev);
366 if (snmp6_alloc_dev(ndev) < 0) {
367 ADBG((KERN_WARNING
368 "%s(): cannot allocate memory for statistics; dev=%s.\n",
369 __func__, dev->name));
370 neigh_parms_release(&nd_tbl, ndev->nd_parms);
371 ndev->dead = 1;
372 in6_dev_finish_destroy(ndev);
373 return NULL;
376 if (snmp6_register_dev(ndev) < 0) {
377 ADBG((KERN_WARNING
378 "%s(): cannot create /proc/net/dev_snmp6/%s\n",
379 __func__, dev->name));
380 neigh_parms_release(&nd_tbl, ndev->nd_parms);
381 ndev->dead = 1;
382 in6_dev_finish_destroy(ndev);
383 return NULL;
386 /* One reference from device. We must do this before
387 * we invoke __ipv6_regen_rndid().
389 in6_dev_hold(ndev);
391 if (dev->flags & (IFF_NOARP | IFF_LOOPBACK))
392 ndev->cnf.accept_dad = -1;
394 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
395 if (dev->type == ARPHRD_SIT && (dev->priv_flags & IFF_ISATAP)) {
396 printk(KERN_INFO
397 "%s: Disabled Multicast RS\n",
398 dev->name);
399 ndev->cnf.rtr_solicits = 0;
401 #endif
403 #ifdef CONFIG_IPV6_PRIVACY
404 setup_timer(&ndev->regen_timer, ipv6_regen_rndid, (unsigned long)ndev);
405 if ((dev->flags&IFF_LOOPBACK) ||
406 dev->type == ARPHRD_TUNNEL ||
407 dev->type == ARPHRD_TUNNEL6 ||
408 dev->type == ARPHRD_SIT ||
409 dev->type == ARPHRD_NONE) {
410 printk(KERN_INFO
411 "%s: Disabled Privacy Extensions\n",
412 dev->name);
413 ndev->cnf.use_tempaddr = -1;
414 } else {
415 in6_dev_hold(ndev);
416 ipv6_regen_rndid((unsigned long) ndev);
418 #endif
420 if (netif_running(dev) && addrconf_qdisc_ok(dev))
421 ndev->if_flags |= IF_READY;
423 ipv6_mc_init_dev(ndev);
424 ndev->tstamp = jiffies;
425 addrconf_sysctl_register(ndev);
426 /* protected by rtnl_lock */
427 rcu_assign_pointer(dev->ip6_ptr, ndev);
429 /* Join all-node multicast group */
430 ipv6_dev_mc_inc(dev, &in6addr_linklocal_allnodes);
432 return ndev;
435 static struct inet6_dev * ipv6_find_idev(struct net_device *dev)
437 struct inet6_dev *idev;
439 ASSERT_RTNL();
441 if ((idev = __in6_dev_get(dev)) == NULL) {
442 if ((idev = ipv6_add_dev(dev)) == NULL)
443 return NULL;
446 if (dev->flags&IFF_UP)
447 ipv6_mc_up(idev);
448 return idev;
451 #ifdef CONFIG_SYSCTL
452 static void dev_forward_change(struct inet6_dev *idev)
454 struct net_device *dev;
455 struct inet6_ifaddr *ifa;
457 if (!idev)
458 return;
459 dev = idev->dev;
460 if (idev->cnf.forwarding)
461 dev_disable_lro(dev);
462 if (dev && (dev->flags & IFF_MULTICAST)) {
463 if (idev->cnf.forwarding)
464 ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
465 else
466 ipv6_dev_mc_dec(dev, &in6addr_linklocal_allrouters);
468 for (ifa=idev->addr_list; ifa; ifa=ifa->if_next) {
469 if (ifa->flags&IFA_F_TENTATIVE)
470 continue;
471 if (idev->cnf.forwarding)
472 addrconf_join_anycast(ifa);
473 else
474 addrconf_leave_anycast(ifa);
479 static void addrconf_forward_change(struct net *net, __s32 newf)
481 struct net_device *dev;
482 struct inet6_dev *idev;
484 rcu_read_lock();
485 for_each_netdev_rcu(net, dev) {
486 idev = __in6_dev_get(dev);
487 if (idev) {
488 int changed = (!idev->cnf.forwarding) ^ (!newf);
489 idev->cnf.forwarding = newf;
490 if (changed)
491 dev_forward_change(idev);
494 rcu_read_unlock();
497 static int addrconf_fixup_forwarding(struct ctl_table *table, int *p, int old)
499 struct net *net;
501 net = (struct net *)table->extra2;
502 if (p == &net->ipv6.devconf_dflt->forwarding)
503 return 0;
505 if (!rtnl_trylock()) {
506 /* Restore the original values before restarting */
507 *p = old;
508 return restart_syscall();
511 if (p == &net->ipv6.devconf_all->forwarding) {
512 __s32 newf = net->ipv6.devconf_all->forwarding;
513 net->ipv6.devconf_dflt->forwarding = newf;
514 addrconf_forward_change(net, newf);
515 } else if ((!*p) ^ (!old))
516 dev_forward_change((struct inet6_dev *)table->extra1);
517 rtnl_unlock();
519 if (*p)
520 rt6_purge_dflt_routers(net);
521 return 1;
523 #endif
525 /* Nobody refers to this ifaddr, destroy it */
527 void inet6_ifa_finish_destroy(struct inet6_ifaddr *ifp)
529 WARN_ON(ifp->if_next != NULL);
530 WARN_ON(ifp->lst_next != NULL);
532 #ifdef NET_REFCNT_DEBUG
533 printk(KERN_DEBUG "inet6_ifa_finish_destroy\n");
534 #endif
536 in6_dev_put(ifp->idev);
538 if (del_timer(&ifp->timer))
539 printk("Timer is still running, when freeing ifa=%p\n", ifp);
541 if (!ifp->dead) {
542 printk("Freeing alive inet6 address %p\n", ifp);
543 return;
545 dst_release(&ifp->rt->u.dst);
547 kfree(ifp);
550 static void
551 ipv6_link_dev_addr(struct inet6_dev *idev, struct inet6_ifaddr *ifp)
553 struct inet6_ifaddr *ifa, **ifap;
554 int ifp_scope = ipv6_addr_src_scope(&ifp->addr);
557 * Each device address list is sorted in order of scope -
558 * global before linklocal.
560 for (ifap = &idev->addr_list; (ifa = *ifap) != NULL;
561 ifap = &ifa->if_next) {
562 if (ifp_scope >= ipv6_addr_src_scope(&ifa->addr))
563 break;
566 ifp->if_next = *ifap;
567 *ifap = ifp;
571 * Hash function taken from net_alias.c
573 static u8 ipv6_addr_hash(const struct in6_addr *addr)
575 __u32 word;
578 * We perform the hash function over the last 64 bits of the address
579 * This will include the IEEE address token on links that support it.
582 word = (__force u32)(addr->s6_addr32[2] ^ addr->s6_addr32[3]);
583 word ^= (word >> 16);
584 word ^= (word >> 8);
586 return ((word ^ (word >> 4)) & 0x0f);
589 /* On success it returns ifp with increased reference count */
591 static struct inet6_ifaddr *
592 ipv6_add_addr(struct inet6_dev *idev, const struct in6_addr *addr, int pfxlen,
593 int scope, u32 flags)
595 struct inet6_ifaddr *ifa = NULL;
596 struct rt6_info *rt;
597 int hash;
598 int err = 0;
599 int addr_type = ipv6_addr_type(addr);
601 if (addr_type == IPV6_ADDR_ANY ||
602 addr_type & IPV6_ADDR_MULTICAST ||
603 (!(idev->dev->flags & IFF_LOOPBACK) &&
604 addr_type & IPV6_ADDR_LOOPBACK))
605 return ERR_PTR(-EADDRNOTAVAIL);
607 rcu_read_lock_bh();
608 if (idev->dead) {
609 err = -ENODEV; /*XXX*/
610 goto out2;
613 if (idev->cnf.disable_ipv6) {
614 err = -EACCES;
615 goto out2;
618 write_lock(&addrconf_hash_lock);
620 /* Ignore adding duplicate addresses on an interface */
621 if (ipv6_chk_same_addr(dev_net(idev->dev), addr, idev->dev)) {
622 ADBG(("ipv6_add_addr: already assigned\n"));
623 err = -EEXIST;
624 goto out;
627 ifa = kzalloc(sizeof(struct inet6_ifaddr), GFP_ATOMIC);
629 if (ifa == NULL) {
630 ADBG(("ipv6_add_addr: malloc failed\n"));
631 err = -ENOBUFS;
632 goto out;
635 rt = addrconf_dst_alloc(idev, addr, 0);
636 if (IS_ERR(rt)) {
637 err = PTR_ERR(rt);
638 goto out;
641 ipv6_addr_copy(&ifa->addr, addr);
643 spin_lock_init(&ifa->lock);
644 init_timer(&ifa->timer);
645 ifa->timer.data = (unsigned long) ifa;
646 ifa->scope = scope;
647 ifa->prefix_len = pfxlen;
648 ifa->flags = flags | IFA_F_TENTATIVE;
649 ifa->cstamp = ifa->tstamp = jiffies;
651 ifa->rt = rt;
654 * part one of RFC 4429, section 3.3
655 * We should not configure an address as
656 * optimistic if we do not yet know the link
657 * layer address of our nexhop router
660 if (rt->rt6i_nexthop == NULL)
661 ifa->flags &= ~IFA_F_OPTIMISTIC;
663 ifa->idev = idev;
664 in6_dev_hold(idev);
665 /* For caller */
666 in6_ifa_hold(ifa);
668 /* Add to big hash table */
669 hash = ipv6_addr_hash(addr);
671 ifa->lst_next = inet6_addr_lst[hash];
672 inet6_addr_lst[hash] = ifa;
673 in6_ifa_hold(ifa);
674 write_unlock(&addrconf_hash_lock);
676 write_lock(&idev->lock);
677 /* Add to inet6_dev unicast addr list. */
678 ipv6_link_dev_addr(idev, ifa);
680 #ifdef CONFIG_IPV6_PRIVACY
681 if (ifa->flags&IFA_F_TEMPORARY) {
682 ifa->tmp_next = idev->tempaddr_list;
683 idev->tempaddr_list = ifa;
684 in6_ifa_hold(ifa);
686 #endif
688 in6_ifa_hold(ifa);
689 write_unlock(&idev->lock);
690 out2:
691 rcu_read_unlock_bh();
693 if (likely(err == 0))
694 atomic_notifier_call_chain(&inet6addr_chain, NETDEV_UP, ifa);
695 else {
696 kfree(ifa);
697 ifa = ERR_PTR(err);
700 return ifa;
701 out:
702 write_unlock(&addrconf_hash_lock);
703 goto out2;
706 /* This function wants to get referenced ifp and releases it before return */
708 static void ipv6_del_addr(struct inet6_ifaddr *ifp)
710 struct inet6_ifaddr *ifa, **ifap;
711 struct inet6_dev *idev = ifp->idev;
712 int hash;
713 int deleted = 0, onlink = 0;
714 unsigned long expires = jiffies;
716 hash = ipv6_addr_hash(&ifp->addr);
718 ifp->dead = 1;
720 write_lock_bh(&addrconf_hash_lock);
721 for (ifap = &inet6_addr_lst[hash]; (ifa=*ifap) != NULL;
722 ifap = &ifa->lst_next) {
723 if (ifa == ifp) {
724 *ifap = ifa->lst_next;
725 __in6_ifa_put(ifp);
726 ifa->lst_next = NULL;
727 break;
730 write_unlock_bh(&addrconf_hash_lock);
732 write_lock_bh(&idev->lock);
733 #ifdef CONFIG_IPV6_PRIVACY
734 if (ifp->flags&IFA_F_TEMPORARY) {
735 for (ifap = &idev->tempaddr_list; (ifa=*ifap) != NULL;
736 ifap = &ifa->tmp_next) {
737 if (ifa == ifp) {
738 *ifap = ifa->tmp_next;
739 if (ifp->ifpub) {
740 in6_ifa_put(ifp->ifpub);
741 ifp->ifpub = NULL;
743 __in6_ifa_put(ifp);
744 ifa->tmp_next = NULL;
745 break;
749 #endif
751 for (ifap = &idev->addr_list; (ifa=*ifap) != NULL;) {
752 if (ifa == ifp) {
753 *ifap = ifa->if_next;
754 __in6_ifa_put(ifp);
755 ifa->if_next = NULL;
756 if (!(ifp->flags & IFA_F_PERMANENT) || onlink > 0)
757 break;
758 deleted = 1;
759 continue;
760 } else if (ifp->flags & IFA_F_PERMANENT) {
761 if (ipv6_prefix_equal(&ifa->addr, &ifp->addr,
762 ifp->prefix_len)) {
763 if (ifa->flags & IFA_F_PERMANENT) {
764 onlink = 1;
765 if (deleted)
766 break;
767 } else {
768 unsigned long lifetime;
770 if (!onlink)
771 onlink = -1;
773 spin_lock(&ifa->lock);
775 lifetime = addrconf_timeout_fixup(ifa->valid_lft, HZ);
777 * Note: Because this address is
778 * not permanent, lifetime <
779 * LONG_MAX / HZ here.
781 if (time_before(expires,
782 ifa->tstamp + lifetime * HZ))
783 expires = ifa->tstamp + lifetime * HZ;
784 spin_unlock(&ifa->lock);
788 ifap = &ifa->if_next;
790 write_unlock_bh(&idev->lock);
792 addrconf_del_timer(ifp);
794 ipv6_ifa_notify(RTM_DELADDR, ifp);
796 atomic_notifier_call_chain(&inet6addr_chain, NETDEV_DOWN, ifp);
799 * Purge or update corresponding prefix
801 * 1) we don't purge prefix here if address was not permanent.
802 * prefix is managed by its own lifetime.
803 * 2) if there're no addresses, delete prefix.
804 * 3) if there're still other permanent address(es),
805 * corresponding prefix is still permanent.
806 * 4) otherwise, update prefix lifetime to the
807 * longest valid lifetime among the corresponding
808 * addresses on the device.
809 * Note: subsequent RA will update lifetime.
811 * --yoshfuji
813 if ((ifp->flags & IFA_F_PERMANENT) && onlink < 1) {
814 struct in6_addr prefix;
815 struct rt6_info *rt;
816 struct net *net = dev_net(ifp->idev->dev);
817 ipv6_addr_prefix(&prefix, &ifp->addr, ifp->prefix_len);
818 rt = rt6_lookup(net, &prefix, NULL, ifp->idev->dev->ifindex, 1);
820 if (rt && addrconf_is_prefix_route(rt)) {
821 if (onlink == 0) {
822 ip6_del_rt(rt);
823 rt = NULL;
824 } else if (!(rt->rt6i_flags & RTF_EXPIRES)) {
825 rt->rt6i_expires = expires;
826 rt->rt6i_flags |= RTF_EXPIRES;
829 dst_release(&rt->u.dst);
832 in6_ifa_put(ifp);
835 #ifdef CONFIG_IPV6_PRIVACY
836 static int ipv6_create_tempaddr(struct inet6_ifaddr *ifp, struct inet6_ifaddr *ift)
838 struct inet6_dev *idev = ifp->idev;
839 struct in6_addr addr, *tmpaddr;
840 unsigned long tmp_prefered_lft, tmp_valid_lft, tmp_cstamp, tmp_tstamp;
841 unsigned long regen_advance;
842 int tmp_plen;
843 int ret = 0;
844 int max_addresses;
845 u32 addr_flags;
847 write_lock(&idev->lock);
848 if (ift) {
849 spin_lock_bh(&ift->lock);
850 memcpy(&addr.s6_addr[8], &ift->addr.s6_addr[8], 8);
851 spin_unlock_bh(&ift->lock);
852 tmpaddr = &addr;
853 } else {
854 tmpaddr = NULL;
856 retry:
857 in6_dev_hold(idev);
858 if (idev->cnf.use_tempaddr <= 0) {
859 write_unlock(&idev->lock);
860 printk(KERN_INFO
861 "ipv6_create_tempaddr(): use_tempaddr is disabled.\n");
862 in6_dev_put(idev);
863 ret = -1;
864 goto out;
866 spin_lock_bh(&ifp->lock);
867 if (ifp->regen_count++ >= idev->cnf.regen_max_retry) {
868 idev->cnf.use_tempaddr = -1; /*XXX*/
869 spin_unlock_bh(&ifp->lock);
870 write_unlock(&idev->lock);
871 printk(KERN_WARNING
872 "ipv6_create_tempaddr(): regeneration time exceeded. disabled temporary address support.\n");
873 in6_dev_put(idev);
874 ret = -1;
875 goto out;
877 in6_ifa_hold(ifp);
878 memcpy(addr.s6_addr, ifp->addr.s6_addr, 8);
879 if (__ipv6_try_regen_rndid(idev, tmpaddr) < 0) {
880 spin_unlock_bh(&ifp->lock);
881 write_unlock(&idev->lock);
882 printk(KERN_WARNING
883 "ipv6_create_tempaddr(): regeneration of randomized interface id failed.\n");
884 in6_ifa_put(ifp);
885 in6_dev_put(idev);
886 ret = -1;
887 goto out;
889 memcpy(&addr.s6_addr[8], idev->rndid, 8);
890 tmp_valid_lft = min_t(__u32,
891 ifp->valid_lft,
892 idev->cnf.temp_valid_lft);
893 tmp_prefered_lft = min_t(__u32,
894 ifp->prefered_lft,
895 idev->cnf.temp_prefered_lft - desync_factor / HZ);
896 tmp_plen = ifp->prefix_len;
897 max_addresses = idev->cnf.max_addresses;
898 tmp_cstamp = ifp->cstamp;
899 tmp_tstamp = ifp->tstamp;
900 spin_unlock_bh(&ifp->lock);
902 regen_advance = idev->cnf.regen_max_retry *
903 idev->cnf.dad_transmits *
904 idev->nd_parms->retrans_time / HZ;
905 write_unlock(&idev->lock);
907 /* A temporary address is created only if this calculated Preferred
908 * Lifetime is greater than REGEN_ADVANCE time units. In particular,
909 * an implementation must not create a temporary address with a zero
910 * Preferred Lifetime.
912 if (tmp_prefered_lft <= regen_advance) {
913 in6_ifa_put(ifp);
914 in6_dev_put(idev);
915 ret = -1;
916 goto out;
919 addr_flags = IFA_F_TEMPORARY;
920 /* set in addrconf_prefix_rcv() */
921 if (ifp->flags & IFA_F_OPTIMISTIC)
922 addr_flags |= IFA_F_OPTIMISTIC;
924 ift = !max_addresses ||
925 ipv6_count_addresses(idev) < max_addresses ?
926 ipv6_add_addr(idev, &addr, tmp_plen,
927 ipv6_addr_type(&addr)&IPV6_ADDR_SCOPE_MASK,
928 addr_flags) : NULL;
929 if (!ift || IS_ERR(ift)) {
930 in6_ifa_put(ifp);
931 in6_dev_put(idev);
932 printk(KERN_INFO
933 "ipv6_create_tempaddr(): retry temporary address regeneration.\n");
934 tmpaddr = &addr;
935 write_lock(&idev->lock);
936 goto retry;
939 spin_lock_bh(&ift->lock);
940 ift->ifpub = ifp;
941 ift->valid_lft = tmp_valid_lft;
942 ift->prefered_lft = tmp_prefered_lft;
943 ift->cstamp = tmp_cstamp;
944 ift->tstamp = tmp_tstamp;
945 spin_unlock_bh(&ift->lock);
947 addrconf_dad_start(ift, 0);
948 in6_ifa_put(ift);
949 in6_dev_put(idev);
950 out:
951 return ret;
953 #endif
956 * Choose an appropriate source address (RFC3484)
958 enum {
959 IPV6_SADDR_RULE_INIT = 0,
960 IPV6_SADDR_RULE_LOCAL,
961 IPV6_SADDR_RULE_SCOPE,
962 IPV6_SADDR_RULE_PREFERRED,
963 #ifdef CONFIG_IPV6_MIP6
964 IPV6_SADDR_RULE_HOA,
965 #endif
966 IPV6_SADDR_RULE_OIF,
967 IPV6_SADDR_RULE_LABEL,
968 #ifdef CONFIG_IPV6_PRIVACY
969 IPV6_SADDR_RULE_PRIVACY,
970 #endif
971 IPV6_SADDR_RULE_ORCHID,
972 IPV6_SADDR_RULE_PREFIX,
973 IPV6_SADDR_RULE_MAX
976 struct ipv6_saddr_score {
977 int rule;
978 int addr_type;
979 struct inet6_ifaddr *ifa;
980 DECLARE_BITMAP(scorebits, IPV6_SADDR_RULE_MAX);
981 int scopedist;
982 int matchlen;
985 struct ipv6_saddr_dst {
986 const struct in6_addr *addr;
987 int ifindex;
988 int scope;
989 int label;
990 unsigned int prefs;
993 static inline int ipv6_saddr_preferred(int type)
995 if (type & (IPV6_ADDR_MAPPED|IPV6_ADDR_COMPATv4|
996 IPV6_ADDR_LOOPBACK|IPV6_ADDR_RESERVED))
997 return 1;
998 return 0;
1001 static int ipv6_get_saddr_eval(struct net *net,
1002 struct ipv6_saddr_score *score,
1003 struct ipv6_saddr_dst *dst,
1004 int i)
1006 int ret;
1008 if (i <= score->rule) {
1009 switch (i) {
1010 case IPV6_SADDR_RULE_SCOPE:
1011 ret = score->scopedist;
1012 break;
1013 case IPV6_SADDR_RULE_PREFIX:
1014 ret = score->matchlen;
1015 break;
1016 default:
1017 ret = !!test_bit(i, score->scorebits);
1019 goto out;
1022 switch (i) {
1023 case IPV6_SADDR_RULE_INIT:
1024 /* Rule 0: remember if hiscore is not ready yet */
1025 ret = !!score->ifa;
1026 break;
1027 case IPV6_SADDR_RULE_LOCAL:
1028 /* Rule 1: Prefer same address */
1029 ret = ipv6_addr_equal(&score->ifa->addr, dst->addr);
1030 break;
1031 case IPV6_SADDR_RULE_SCOPE:
1032 /* Rule 2: Prefer appropriate scope
1034 * ret
1036 * -1 | d 15
1037 * ---+--+-+---> scope
1039 * | d is scope of the destination.
1040 * B-d | \
1041 * | \ <- smaller scope is better if
1042 * B-15 | \ if scope is enough for destinaion.
1043 * | ret = B - scope (-1 <= scope >= d <= 15).
1044 * d-C-1 | /
1045 * |/ <- greater is better
1046 * -C / if scope is not enough for destination.
1047 * /| ret = scope - C (-1 <= d < scope <= 15).
1049 * d - C - 1 < B -15 (for all -1 <= d <= 15).
1050 * C > d + 14 - B >= 15 + 14 - B = 29 - B.
1051 * Assume B = 0 and we get C > 29.
1053 ret = __ipv6_addr_src_scope(score->addr_type);
1054 if (ret >= dst->scope)
1055 ret = -ret;
1056 else
1057 ret -= 128; /* 30 is enough */
1058 score->scopedist = ret;
1059 break;
1060 case IPV6_SADDR_RULE_PREFERRED:
1061 /* Rule 3: Avoid deprecated and optimistic addresses */
1062 ret = ipv6_saddr_preferred(score->addr_type) ||
1063 !(score->ifa->flags & (IFA_F_DEPRECATED|IFA_F_OPTIMISTIC));
1064 break;
1065 #ifdef CONFIG_IPV6_MIP6
1066 case IPV6_SADDR_RULE_HOA:
1068 /* Rule 4: Prefer home address */
1069 int prefhome = !(dst->prefs & IPV6_PREFER_SRC_COA);
1070 ret = !(score->ifa->flags & IFA_F_HOMEADDRESS) ^ prefhome;
1071 break;
1073 #endif
1074 case IPV6_SADDR_RULE_OIF:
1075 /* Rule 5: Prefer outgoing interface */
1076 ret = (!dst->ifindex ||
1077 dst->ifindex == score->ifa->idev->dev->ifindex);
1078 break;
1079 case IPV6_SADDR_RULE_LABEL:
1080 /* Rule 6: Prefer matching label */
1081 ret = ipv6_addr_label(net,
1082 &score->ifa->addr, score->addr_type,
1083 score->ifa->idev->dev->ifindex) == dst->label;
1084 break;
1085 #ifdef CONFIG_IPV6_PRIVACY
1086 case IPV6_SADDR_RULE_PRIVACY:
1088 /* Rule 7: Prefer public address
1089 * Note: prefer temprary address if use_tempaddr >= 2
1091 int preftmp = dst->prefs & (IPV6_PREFER_SRC_PUBLIC|IPV6_PREFER_SRC_TMP) ?
1092 !!(dst->prefs & IPV6_PREFER_SRC_TMP) :
1093 score->ifa->idev->cnf.use_tempaddr >= 2;
1094 ret = (!(score->ifa->flags & IFA_F_TEMPORARY)) ^ preftmp;
1095 break;
1097 #endif
1098 case IPV6_SADDR_RULE_ORCHID:
1099 /* Rule 8-: Prefer ORCHID vs ORCHID or
1100 * non-ORCHID vs non-ORCHID
1102 ret = !(ipv6_addr_orchid(&score->ifa->addr) ^
1103 ipv6_addr_orchid(dst->addr));
1104 break;
1105 case IPV6_SADDR_RULE_PREFIX:
1106 /* Rule 8: Use longest matching prefix */
1107 score->matchlen = ret = ipv6_addr_diff(&score->ifa->addr,
1108 dst->addr);
1109 break;
1110 default:
1111 ret = 0;
1114 if (ret)
1115 __set_bit(i, score->scorebits);
1116 score->rule = i;
1117 out:
1118 return ret;
1121 int ipv6_dev_get_saddr(struct net *net, struct net_device *dst_dev,
1122 const struct in6_addr *daddr, unsigned int prefs,
1123 struct in6_addr *saddr)
1125 struct ipv6_saddr_score scores[2],
1126 *score = &scores[0], *hiscore = &scores[1];
1127 struct ipv6_saddr_dst dst;
1128 struct net_device *dev;
1129 int dst_type;
1131 dst_type = __ipv6_addr_type(daddr);
1132 dst.addr = daddr;
1133 dst.ifindex = dst_dev ? dst_dev->ifindex : 0;
1134 dst.scope = __ipv6_addr_src_scope(dst_type);
1135 dst.label = ipv6_addr_label(net, daddr, dst_type, dst.ifindex);
1136 dst.prefs = prefs;
1138 hiscore->rule = -1;
1139 hiscore->ifa = NULL;
1141 rcu_read_lock();
1143 for_each_netdev_rcu(net, dev) {
1144 struct inet6_dev *idev;
1146 /* Candidate Source Address (section 4)
1147 * - multicast and link-local destination address,
1148 * the set of candidate source address MUST only
1149 * include addresses assigned to interfaces
1150 * belonging to the same link as the outgoing
1151 * interface.
1152 * (- For site-local destination addresses, the
1153 * set of candidate source addresses MUST only
1154 * include addresses assigned to interfaces
1155 * belonging to the same site as the outgoing
1156 * interface.)
1158 if (((dst_type & IPV6_ADDR_MULTICAST) ||
1159 dst.scope <= IPV6_ADDR_SCOPE_LINKLOCAL) &&
1160 dst.ifindex && dev->ifindex != dst.ifindex)
1161 continue;
1163 idev = __in6_dev_get(dev);
1164 if (!idev)
1165 continue;
1167 read_lock_bh(&idev->lock);
1168 for (score->ifa = idev->addr_list; score->ifa; score->ifa = score->ifa->if_next) {
1169 int i;
1172 * - Tentative Address (RFC2462 section 5.4)
1173 * - A tentative address is not considered
1174 * "assigned to an interface" in the traditional
1175 * sense, unless it is also flagged as optimistic.
1176 * - Candidate Source Address (section 4)
1177 * - In any case, anycast addresses, multicast
1178 * addresses, and the unspecified address MUST
1179 * NOT be included in a candidate set.
1181 if ((score->ifa->flags & IFA_F_TENTATIVE) &&
1182 (!(score->ifa->flags & IFA_F_OPTIMISTIC)))
1183 continue;
1185 score->addr_type = __ipv6_addr_type(&score->ifa->addr);
1187 if (unlikely(score->addr_type == IPV6_ADDR_ANY ||
1188 score->addr_type & IPV6_ADDR_MULTICAST)) {
1189 LIMIT_NETDEBUG(KERN_DEBUG
1190 "ADDRCONF: unspecified / multicast address "
1191 "assigned as unicast address on %s",
1192 dev->name);
1193 continue;
1196 score->rule = -1;
1197 bitmap_zero(score->scorebits, IPV6_SADDR_RULE_MAX);
1199 for (i = 0; i < IPV6_SADDR_RULE_MAX; i++) {
1200 int minihiscore, miniscore;
1202 minihiscore = ipv6_get_saddr_eval(net, hiscore, &dst, i);
1203 miniscore = ipv6_get_saddr_eval(net, score, &dst, i);
1205 if (minihiscore > miniscore) {
1206 if (i == IPV6_SADDR_RULE_SCOPE &&
1207 score->scopedist > 0) {
1209 * special case:
1210 * each remaining entry
1211 * has too small (not enough)
1212 * scope, because ifa entries
1213 * are sorted by their scope
1214 * values.
1216 goto try_nextdev;
1218 break;
1219 } else if (minihiscore < miniscore) {
1220 if (hiscore->ifa)
1221 in6_ifa_put(hiscore->ifa);
1223 in6_ifa_hold(score->ifa);
1225 swap(hiscore, score);
1227 /* restore our iterator */
1228 score->ifa = hiscore->ifa;
1230 break;
1234 try_nextdev:
1235 read_unlock_bh(&idev->lock);
1237 rcu_read_unlock();
1239 if (!hiscore->ifa)
1240 return -EADDRNOTAVAIL;
1242 ipv6_addr_copy(saddr, &hiscore->ifa->addr);
1243 in6_ifa_put(hiscore->ifa);
1244 return 0;
1247 EXPORT_SYMBOL(ipv6_dev_get_saddr);
1249 int ipv6_get_lladdr(struct net_device *dev, struct in6_addr *addr,
1250 unsigned char banned_flags)
1252 struct inet6_dev *idev;
1253 int err = -EADDRNOTAVAIL;
1255 rcu_read_lock();
1256 if ((idev = __in6_dev_get(dev)) != NULL) {
1257 struct inet6_ifaddr *ifp;
1259 read_lock_bh(&idev->lock);
1260 for (ifp=idev->addr_list; ifp; ifp=ifp->if_next) {
1261 if (ifp->scope == IFA_LINK && !(ifp->flags & banned_flags)) {
1262 ipv6_addr_copy(addr, &ifp->addr);
1263 err = 0;
1264 break;
1267 read_unlock_bh(&idev->lock);
1269 rcu_read_unlock();
1270 return err;
1273 static int ipv6_count_addresses(struct inet6_dev *idev)
1275 int cnt = 0;
1276 struct inet6_ifaddr *ifp;
1278 read_lock_bh(&idev->lock);
1279 for (ifp=idev->addr_list; ifp; ifp=ifp->if_next)
1280 cnt++;
1281 read_unlock_bh(&idev->lock);
1282 return cnt;
1285 int ipv6_chk_addr(struct net *net, struct in6_addr *addr,
1286 struct net_device *dev, int strict)
1288 struct inet6_ifaddr * ifp;
1289 u8 hash = ipv6_addr_hash(addr);
1291 read_lock_bh(&addrconf_hash_lock);
1292 for(ifp = inet6_addr_lst[hash]; ifp; ifp=ifp->lst_next) {
1293 if (!net_eq(dev_net(ifp->idev->dev), net))
1294 continue;
1295 if (ipv6_addr_equal(&ifp->addr, addr) &&
1296 !(ifp->flags&IFA_F_TENTATIVE)) {
1297 if (dev == NULL || ifp->idev->dev == dev ||
1298 !(ifp->scope&(IFA_LINK|IFA_HOST) || strict))
1299 break;
1302 read_unlock_bh(&addrconf_hash_lock);
1303 return ifp != NULL;
1305 EXPORT_SYMBOL(ipv6_chk_addr);
1307 static
1308 int ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
1309 struct net_device *dev)
1311 struct inet6_ifaddr * ifp;
1312 u8 hash = ipv6_addr_hash(addr);
1314 for(ifp = inet6_addr_lst[hash]; ifp; ifp=ifp->lst_next) {
1315 if (!net_eq(dev_net(ifp->idev->dev), net))
1316 continue;
1317 if (ipv6_addr_equal(&ifp->addr, addr)) {
1318 if (dev == NULL || ifp->idev->dev == dev)
1319 break;
1322 return ifp != NULL;
1325 int ipv6_chk_prefix(struct in6_addr *addr, struct net_device *dev)
1327 struct inet6_dev *idev;
1328 struct inet6_ifaddr *ifa;
1329 int onlink;
1331 onlink = 0;
1332 rcu_read_lock();
1333 idev = __in6_dev_get(dev);
1334 if (idev) {
1335 read_lock_bh(&idev->lock);
1336 for (ifa = idev->addr_list; ifa; ifa = ifa->if_next) {
1337 onlink = ipv6_prefix_equal(addr, &ifa->addr,
1338 ifa->prefix_len);
1339 if (onlink)
1340 break;
1342 read_unlock_bh(&idev->lock);
1344 rcu_read_unlock();
1345 return onlink;
1348 EXPORT_SYMBOL(ipv6_chk_prefix);
1350 struct inet6_ifaddr *ipv6_get_ifaddr(struct net *net, const struct in6_addr *addr,
1351 struct net_device *dev, int strict)
1353 struct inet6_ifaddr * ifp;
1354 u8 hash = ipv6_addr_hash(addr);
1356 read_lock_bh(&addrconf_hash_lock);
1357 for(ifp = inet6_addr_lst[hash]; ifp; ifp=ifp->lst_next) {
1358 if (!net_eq(dev_net(ifp->idev->dev), net))
1359 continue;
1360 if (ipv6_addr_equal(&ifp->addr, addr)) {
1361 if (dev == NULL || ifp->idev->dev == dev ||
1362 !(ifp->scope&(IFA_LINK|IFA_HOST) || strict)) {
1363 in6_ifa_hold(ifp);
1364 break;
1368 read_unlock_bh(&addrconf_hash_lock);
1370 return ifp;
1373 /* Gets referenced address, destroys ifaddr */
1375 static void addrconf_dad_stop(struct inet6_ifaddr *ifp, int dad_failed)
1377 if (ifp->flags&IFA_F_PERMANENT) {
1378 spin_lock_bh(&ifp->lock);
1379 addrconf_del_timer(ifp);
1380 ifp->flags |= IFA_F_TENTATIVE;
1381 if (dad_failed)
1382 ifp->flags |= IFA_F_DADFAILED;
1383 spin_unlock_bh(&ifp->lock);
1384 in6_ifa_put(ifp);
1385 #ifdef CONFIG_IPV6_PRIVACY
1386 } else if (ifp->flags&IFA_F_TEMPORARY) {
1387 struct inet6_ifaddr *ifpub;
1388 spin_lock_bh(&ifp->lock);
1389 ifpub = ifp->ifpub;
1390 if (ifpub) {
1391 in6_ifa_hold(ifpub);
1392 spin_unlock_bh(&ifp->lock);
1393 ipv6_create_tempaddr(ifpub, ifp);
1394 in6_ifa_put(ifpub);
1395 } else {
1396 spin_unlock_bh(&ifp->lock);
1398 ipv6_del_addr(ifp);
1399 #endif
1400 } else
1401 ipv6_del_addr(ifp);
1404 void addrconf_dad_failure(struct inet6_ifaddr *ifp)
1406 struct inet6_dev *idev = ifp->idev;
1408 if (net_ratelimit())
1409 printk(KERN_INFO "%s: IPv6 duplicate address %pI6c detected!\n",
1410 ifp->idev->dev->name, &ifp->addr);
1412 if (idev->cnf.accept_dad > 1 && !idev->cnf.disable_ipv6) {
1413 struct in6_addr addr;
1415 addr.s6_addr32[0] = htonl(0xfe800000);
1416 addr.s6_addr32[1] = 0;
1418 if (!ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) &&
1419 ipv6_addr_equal(&ifp->addr, &addr)) {
1420 /* DAD failed for link-local based on MAC address */
1421 idev->cnf.disable_ipv6 = 1;
1423 printk(KERN_INFO "%s: IPv6 being disabled!\n",
1424 ifp->idev->dev->name);
1428 addrconf_dad_stop(ifp, 1);
1431 /* Join to solicited addr multicast group. */
1433 void addrconf_join_solict(struct net_device *dev, struct in6_addr *addr)
1435 struct in6_addr maddr;
1437 if (dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1438 return;
1440 addrconf_addr_solict_mult(addr, &maddr);
1441 ipv6_dev_mc_inc(dev, &maddr);
1444 void addrconf_leave_solict(struct inet6_dev *idev, struct in6_addr *addr)
1446 struct in6_addr maddr;
1448 if (idev->dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1449 return;
1451 addrconf_addr_solict_mult(addr, &maddr);
1452 __ipv6_dev_mc_dec(idev, &maddr);
1455 static void addrconf_join_anycast(struct inet6_ifaddr *ifp)
1457 struct in6_addr addr;
1458 ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1459 if (ipv6_addr_any(&addr))
1460 return;
1461 ipv6_dev_ac_inc(ifp->idev->dev, &addr);
1464 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp)
1466 struct in6_addr addr;
1467 ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1468 if (ipv6_addr_any(&addr))
1469 return;
1470 __ipv6_dev_ac_dec(ifp->idev, &addr);
1473 static int addrconf_ifid_eui48(u8 *eui, struct net_device *dev)
1475 if (dev->addr_len != ETH_ALEN)
1476 return -1;
1477 memcpy(eui, dev->dev_addr, 3);
1478 memcpy(eui + 5, dev->dev_addr + 3, 3);
1481 * The zSeries OSA network cards can be shared among various
1482 * OS instances, but the OSA cards have only one MAC address.
1483 * This leads to duplicate address conflicts in conjunction
1484 * with IPv6 if more than one instance uses the same card.
1486 * The driver for these cards can deliver a unique 16-bit
1487 * identifier for each instance sharing the same card. It is
1488 * placed instead of 0xFFFE in the interface identifier. The
1489 * "u" bit of the interface identifier is not inverted in this
1490 * case. Hence the resulting interface identifier has local
1491 * scope according to RFC2373.
1493 if (dev->dev_id) {
1494 eui[3] = (dev->dev_id >> 8) & 0xFF;
1495 eui[4] = dev->dev_id & 0xFF;
1496 } else {
1497 eui[3] = 0xFF;
1498 eui[4] = 0xFE;
1499 eui[0] ^= 2;
1501 return 0;
1504 static int addrconf_ifid_arcnet(u8 *eui, struct net_device *dev)
1506 /* XXX: inherit EUI-64 from other interface -- yoshfuji */
1507 if (dev->addr_len != ARCNET_ALEN)
1508 return -1;
1509 memset(eui, 0, 7);
1510 eui[7] = *(u8*)dev->dev_addr;
1511 return 0;
1514 static int addrconf_ifid_infiniband(u8 *eui, struct net_device *dev)
1516 if (dev->addr_len != INFINIBAND_ALEN)
1517 return -1;
1518 memcpy(eui, dev->dev_addr + 12, 8);
1519 eui[0] |= 2;
1520 return 0;
1523 int __ipv6_isatap_ifid(u8 *eui, __be32 addr)
1525 if (addr == 0)
1526 return -1;
1527 eui[0] = (ipv4_is_zeronet(addr) || ipv4_is_private_10(addr) ||
1528 ipv4_is_loopback(addr) || ipv4_is_linklocal_169(addr) ||
1529 ipv4_is_private_172(addr) || ipv4_is_test_192(addr) ||
1530 ipv4_is_anycast_6to4(addr) || ipv4_is_private_192(addr) ||
1531 ipv4_is_test_198(addr) || ipv4_is_multicast(addr) ||
1532 ipv4_is_lbcast(addr)) ? 0x00 : 0x02;
1533 eui[1] = 0;
1534 eui[2] = 0x5E;
1535 eui[3] = 0xFE;
1536 memcpy(eui + 4, &addr, 4);
1537 return 0;
1539 EXPORT_SYMBOL(__ipv6_isatap_ifid);
1541 static int addrconf_ifid_sit(u8 *eui, struct net_device *dev)
1543 if (dev->priv_flags & IFF_ISATAP)
1544 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
1545 return -1;
1548 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev)
1550 switch (dev->type) {
1551 case ARPHRD_ETHER:
1552 case ARPHRD_FDDI:
1553 case ARPHRD_IEEE802_TR:
1554 return addrconf_ifid_eui48(eui, dev);
1555 case ARPHRD_ARCNET:
1556 return addrconf_ifid_arcnet(eui, dev);
1557 case ARPHRD_INFINIBAND:
1558 return addrconf_ifid_infiniband(eui, dev);
1559 case ARPHRD_SIT:
1560 return addrconf_ifid_sit(eui, dev);
1562 return -1;
1565 static int ipv6_inherit_eui64(u8 *eui, struct inet6_dev *idev)
1567 int err = -1;
1568 struct inet6_ifaddr *ifp;
1570 read_lock_bh(&idev->lock);
1571 for (ifp=idev->addr_list; ifp; ifp=ifp->if_next) {
1572 if (ifp->scope == IFA_LINK && !(ifp->flags&IFA_F_TENTATIVE)) {
1573 memcpy(eui, ifp->addr.s6_addr+8, 8);
1574 err = 0;
1575 break;
1578 read_unlock_bh(&idev->lock);
1579 return err;
1582 #ifdef CONFIG_IPV6_PRIVACY
1583 /* (re)generation of randomized interface identifier (RFC 3041 3.2, 3.5) */
1584 static int __ipv6_regen_rndid(struct inet6_dev *idev)
1586 regen:
1587 get_random_bytes(idev->rndid, sizeof(idev->rndid));
1588 idev->rndid[0] &= ~0x02;
1591 * <draft-ietf-ipngwg-temp-addresses-v2-00.txt>:
1592 * check if generated address is not inappropriate
1594 * - Reserved subnet anycast (RFC 2526)
1595 * 11111101 11....11 1xxxxxxx
1596 * - ISATAP (RFC4214) 6.1
1597 * 00-00-5E-FE-xx-xx-xx-xx
1598 * - value 0
1599 * - XXX: already assigned to an address on the device
1601 if (idev->rndid[0] == 0xfd &&
1602 (idev->rndid[1]&idev->rndid[2]&idev->rndid[3]&idev->rndid[4]&idev->rndid[5]&idev->rndid[6]) == 0xff &&
1603 (idev->rndid[7]&0x80))
1604 goto regen;
1605 if ((idev->rndid[0]|idev->rndid[1]) == 0) {
1606 if (idev->rndid[2] == 0x5e && idev->rndid[3] == 0xfe)
1607 goto regen;
1608 if ((idev->rndid[2]|idev->rndid[3]|idev->rndid[4]|idev->rndid[5]|idev->rndid[6]|idev->rndid[7]) == 0x00)
1609 goto regen;
1612 return 0;
1615 static void ipv6_regen_rndid(unsigned long data)
1617 struct inet6_dev *idev = (struct inet6_dev *) data;
1618 unsigned long expires;
1620 rcu_read_lock_bh();
1621 write_lock_bh(&idev->lock);
1623 if (idev->dead)
1624 goto out;
1626 if (__ipv6_regen_rndid(idev) < 0)
1627 goto out;
1629 expires = jiffies +
1630 idev->cnf.temp_prefered_lft * HZ -
1631 idev->cnf.regen_max_retry * idev->cnf.dad_transmits * idev->nd_parms->retrans_time - desync_factor;
1632 if (time_before(expires, jiffies)) {
1633 printk(KERN_WARNING
1634 "ipv6_regen_rndid(): too short regeneration interval; timer disabled for %s.\n",
1635 idev->dev->name);
1636 goto out;
1639 if (!mod_timer(&idev->regen_timer, expires))
1640 in6_dev_hold(idev);
1642 out:
1643 write_unlock_bh(&idev->lock);
1644 rcu_read_unlock_bh();
1645 in6_dev_put(idev);
1648 static int __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr) {
1649 int ret = 0;
1651 if (tmpaddr && memcmp(idev->rndid, &tmpaddr->s6_addr[8], 8) == 0)
1652 ret = __ipv6_regen_rndid(idev);
1653 return ret;
1655 #endif
1658 * Add prefix route.
1661 static void
1662 addrconf_prefix_route(struct in6_addr *pfx, int plen, struct net_device *dev,
1663 unsigned long expires, u32 flags)
1665 struct fib6_config cfg = {
1666 .fc_table = RT6_TABLE_PREFIX,
1667 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1668 .fc_ifindex = dev->ifindex,
1669 .fc_expires = expires,
1670 .fc_dst_len = plen,
1671 .fc_flags = RTF_UP | flags,
1672 .fc_nlinfo.nl_net = dev_net(dev),
1673 .fc_protocol = RTPROT_KERNEL,
1676 ipv6_addr_copy(&cfg.fc_dst, pfx);
1678 /* Prevent useless cloning on PtP SIT.
1679 This thing is done here expecting that the whole
1680 class of non-broadcast devices need not cloning.
1682 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
1683 if (dev->type == ARPHRD_SIT && (dev->flags & IFF_POINTOPOINT))
1684 cfg.fc_flags |= RTF_NONEXTHOP;
1685 #endif
1687 ip6_route_add(&cfg);
1690 /* Create "default" multicast route to the interface */
1692 static void addrconf_add_mroute(struct net_device *dev)
1694 struct fib6_config cfg = {
1695 .fc_table = RT6_TABLE_LOCAL,
1696 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1697 .fc_ifindex = dev->ifindex,
1698 .fc_dst_len = 8,
1699 .fc_flags = RTF_UP,
1700 .fc_nlinfo.nl_net = dev_net(dev),
1703 ipv6_addr_set(&cfg.fc_dst, htonl(0xFF000000), 0, 0, 0);
1705 ip6_route_add(&cfg);
1708 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
1709 static void sit_route_add(struct net_device *dev)
1711 struct fib6_config cfg = {
1712 .fc_table = RT6_TABLE_MAIN,
1713 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1714 .fc_ifindex = dev->ifindex,
1715 .fc_dst_len = 96,
1716 .fc_flags = RTF_UP | RTF_NONEXTHOP,
1717 .fc_nlinfo.nl_net = dev_net(dev),
1720 /* prefix length - 96 bits "::d.d.d.d" */
1721 ip6_route_add(&cfg);
1723 #endif
1725 static void addrconf_add_lroute(struct net_device *dev)
1727 struct in6_addr addr;
1729 ipv6_addr_set(&addr, htonl(0xFE800000), 0, 0, 0);
1730 addrconf_prefix_route(&addr, 64, dev, 0, 0);
1733 static struct inet6_dev *addrconf_add_dev(struct net_device *dev)
1735 struct inet6_dev *idev;
1737 ASSERT_RTNL();
1739 if ((idev = ipv6_find_idev(dev)) == NULL)
1740 return NULL;
1742 /* Add default multicast route */
1743 addrconf_add_mroute(dev);
1745 /* Add link local route */
1746 addrconf_add_lroute(dev);
1747 return idev;
1750 void addrconf_prefix_rcv(struct net_device *dev, u8 *opt, int len)
1752 struct prefix_info *pinfo;
1753 __u32 valid_lft;
1754 __u32 prefered_lft;
1755 int addr_type;
1756 struct inet6_dev *in6_dev;
1757 struct net *net = dev_net(dev);
1759 pinfo = (struct prefix_info *) opt;
1761 if (len < sizeof(struct prefix_info)) {
1762 ADBG(("addrconf: prefix option too short\n"));
1763 return;
1767 * Validation checks ([ADDRCONF], page 19)
1770 addr_type = ipv6_addr_type(&pinfo->prefix);
1772 if (addr_type & (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL))
1773 return;
1775 valid_lft = ntohl(pinfo->valid);
1776 prefered_lft = ntohl(pinfo->prefered);
1778 if (prefered_lft > valid_lft) {
1779 if (net_ratelimit())
1780 printk(KERN_WARNING "addrconf: prefix option has invalid lifetime\n");
1781 return;
1784 in6_dev = in6_dev_get(dev);
1786 if (in6_dev == NULL) {
1787 if (net_ratelimit())
1788 printk(KERN_DEBUG "addrconf: device %s not configured\n", dev->name);
1789 return;
1793 * Two things going on here:
1794 * 1) Add routes for on-link prefixes
1795 * 2) Configure prefixes with the auto flag set
1798 if (pinfo->onlink) {
1799 struct rt6_info *rt;
1800 unsigned long rt_expires;
1802 /* Avoid arithmetic overflow. Really, we could
1803 * save rt_expires in seconds, likely valid_lft,
1804 * but it would require division in fib gc, that it
1805 * not good.
1807 if (HZ > USER_HZ)
1808 rt_expires = addrconf_timeout_fixup(valid_lft, HZ);
1809 else
1810 rt_expires = addrconf_timeout_fixup(valid_lft, USER_HZ);
1812 if (addrconf_finite_timeout(rt_expires))
1813 rt_expires *= HZ;
1815 rt = rt6_lookup(net, &pinfo->prefix, NULL,
1816 dev->ifindex, 1);
1818 if (rt && addrconf_is_prefix_route(rt)) {
1819 /* Autoconf prefix route */
1820 if (valid_lft == 0) {
1821 ip6_del_rt(rt);
1822 rt = NULL;
1823 } else if (addrconf_finite_timeout(rt_expires)) {
1824 /* not infinity */
1825 rt->rt6i_expires = jiffies + rt_expires;
1826 rt->rt6i_flags |= RTF_EXPIRES;
1827 } else {
1828 rt->rt6i_flags &= ~RTF_EXPIRES;
1829 rt->rt6i_expires = 0;
1831 } else if (valid_lft) {
1832 clock_t expires = 0;
1833 int flags = RTF_ADDRCONF | RTF_PREFIX_RT;
1834 if (addrconf_finite_timeout(rt_expires)) {
1835 /* not infinity */
1836 flags |= RTF_EXPIRES;
1837 expires = jiffies_to_clock_t(rt_expires);
1839 addrconf_prefix_route(&pinfo->prefix, pinfo->prefix_len,
1840 dev, expires, flags);
1842 if (rt)
1843 dst_release(&rt->u.dst);
1846 /* Try to figure out our local address for this prefix */
1848 if (pinfo->autoconf && in6_dev->cnf.autoconf) {
1849 struct inet6_ifaddr * ifp;
1850 struct in6_addr addr;
1851 int create = 0, update_lft = 0;
1853 if (pinfo->prefix_len == 64) {
1854 memcpy(&addr, &pinfo->prefix, 8);
1855 if (ipv6_generate_eui64(addr.s6_addr + 8, dev) &&
1856 ipv6_inherit_eui64(addr.s6_addr + 8, in6_dev)) {
1857 in6_dev_put(in6_dev);
1858 return;
1860 goto ok;
1862 if (net_ratelimit())
1863 printk(KERN_DEBUG "IPv6 addrconf: prefix with wrong length %d\n",
1864 pinfo->prefix_len);
1865 in6_dev_put(in6_dev);
1866 return;
1870 ifp = ipv6_get_ifaddr(net, &addr, dev, 1);
1872 if (ifp == NULL && valid_lft) {
1873 int max_addresses = in6_dev->cnf.max_addresses;
1874 u32 addr_flags = 0;
1876 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1877 if (in6_dev->cnf.optimistic_dad &&
1878 !net->ipv6.devconf_all->forwarding)
1879 addr_flags = IFA_F_OPTIMISTIC;
1880 #endif
1882 /* Do not allow to create too much of autoconfigured
1883 * addresses; this would be too easy way to crash kernel.
1885 if (!max_addresses ||
1886 ipv6_count_addresses(in6_dev) < max_addresses)
1887 ifp = ipv6_add_addr(in6_dev, &addr, pinfo->prefix_len,
1888 addr_type&IPV6_ADDR_SCOPE_MASK,
1889 addr_flags);
1891 if (!ifp || IS_ERR(ifp)) {
1892 in6_dev_put(in6_dev);
1893 return;
1896 update_lft = create = 1;
1897 ifp->cstamp = jiffies;
1898 addrconf_dad_start(ifp, RTF_ADDRCONF|RTF_PREFIX_RT);
1901 if (ifp) {
1902 int flags;
1903 unsigned long now;
1904 #ifdef CONFIG_IPV6_PRIVACY
1905 struct inet6_ifaddr *ift;
1906 #endif
1907 u32 stored_lft;
1909 /* update lifetime (RFC2462 5.5.3 e) */
1910 spin_lock(&ifp->lock);
1911 now = jiffies;
1912 if (ifp->valid_lft > (now - ifp->tstamp) / HZ)
1913 stored_lft = ifp->valid_lft - (now - ifp->tstamp) / HZ;
1914 else
1915 stored_lft = 0;
1916 if (!update_lft && stored_lft) {
1917 if (valid_lft > MIN_VALID_LIFETIME ||
1918 valid_lft > stored_lft)
1919 update_lft = 1;
1920 else if (stored_lft <= MIN_VALID_LIFETIME) {
1921 /* valid_lft <= stored_lft is always true */
1923 * RFC 4862 Section 5.5.3e:
1924 * "Note that the preferred lifetime of
1925 * the corresponding address is always
1926 * reset to the Preferred Lifetime in
1927 * the received Prefix Information
1928 * option, regardless of whether the
1929 * valid lifetime is also reset or
1930 * ignored."
1932 * So if the preferred lifetime in
1933 * this advertisement is different
1934 * than what we have stored, but the
1935 * valid lifetime is invalid, just
1936 * reset prefered_lft.
1938 * We must set the valid lifetime
1939 * to the stored lifetime since we'll
1940 * be updating the timestamp below,
1941 * else we'll set it back to the
1942 * minumum.
1944 if (prefered_lft != ifp->prefered_lft) {
1945 valid_lft = stored_lft;
1946 update_lft = 1;
1948 } else {
1949 valid_lft = MIN_VALID_LIFETIME;
1950 if (valid_lft < prefered_lft)
1951 prefered_lft = valid_lft;
1952 update_lft = 1;
1956 if (update_lft) {
1957 ifp->valid_lft = valid_lft;
1958 ifp->prefered_lft = prefered_lft;
1959 ifp->tstamp = now;
1960 flags = ifp->flags;
1961 ifp->flags &= ~IFA_F_DEPRECATED;
1962 spin_unlock(&ifp->lock);
1964 if (!(flags&IFA_F_TENTATIVE))
1965 ipv6_ifa_notify(0, ifp);
1966 } else
1967 spin_unlock(&ifp->lock);
1969 #ifdef CONFIG_IPV6_PRIVACY
1970 read_lock_bh(&in6_dev->lock);
1971 /* update all temporary addresses in the list */
1972 for (ift=in6_dev->tempaddr_list; ift; ift=ift->tmp_next) {
1974 * When adjusting the lifetimes of an existing
1975 * temporary address, only lower the lifetimes.
1976 * Implementations must not increase the
1977 * lifetimes of an existing temporary address
1978 * when processing a Prefix Information Option.
1980 if (ifp != ift->ifpub)
1981 continue;
1983 spin_lock(&ift->lock);
1984 flags = ift->flags;
1985 if (ift->valid_lft > valid_lft &&
1986 ift->valid_lft - valid_lft > (jiffies - ift->tstamp) / HZ)
1987 ift->valid_lft = valid_lft + (jiffies - ift->tstamp) / HZ;
1988 if (ift->prefered_lft > prefered_lft &&
1989 ift->prefered_lft - prefered_lft > (jiffies - ift->tstamp) / HZ)
1990 ift->prefered_lft = prefered_lft + (jiffies - ift->tstamp) / HZ;
1991 spin_unlock(&ift->lock);
1992 if (!(flags&IFA_F_TENTATIVE))
1993 ipv6_ifa_notify(0, ift);
1996 if (create && in6_dev->cnf.use_tempaddr > 0) {
1998 * When a new public address is created as described in [ADDRCONF],
1999 * also create a new temporary address.
2001 read_unlock_bh(&in6_dev->lock);
2002 ipv6_create_tempaddr(ifp, NULL);
2003 } else {
2004 read_unlock_bh(&in6_dev->lock);
2006 #endif
2007 in6_ifa_put(ifp);
2008 addrconf_verify(0);
2011 inet6_prefix_notify(RTM_NEWPREFIX, in6_dev, pinfo);
2012 in6_dev_put(in6_dev);
2016 * Set destination address.
2017 * Special case for SIT interfaces where we create a new "virtual"
2018 * device.
2020 int addrconf_set_dstaddr(struct net *net, void __user *arg)
2022 struct in6_ifreq ireq;
2023 struct net_device *dev;
2024 int err = -EINVAL;
2026 rtnl_lock();
2028 err = -EFAULT;
2029 if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2030 goto err_exit;
2032 dev = __dev_get_by_index(net, ireq.ifr6_ifindex);
2034 err = -ENODEV;
2035 if (dev == NULL)
2036 goto err_exit;
2038 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2039 if (dev->type == ARPHRD_SIT) {
2040 const struct net_device_ops *ops = dev->netdev_ops;
2041 struct ifreq ifr;
2042 struct ip_tunnel_parm p;
2044 err = -EADDRNOTAVAIL;
2045 if (!(ipv6_addr_type(&ireq.ifr6_addr) & IPV6_ADDR_COMPATv4))
2046 goto err_exit;
2048 memset(&p, 0, sizeof(p));
2049 p.iph.daddr = ireq.ifr6_addr.s6_addr32[3];
2050 p.iph.saddr = 0;
2051 p.iph.version = 4;
2052 p.iph.ihl = 5;
2053 p.iph.protocol = IPPROTO_IPV6;
2054 p.iph.ttl = 64;
2055 ifr.ifr_ifru.ifru_data = (__force void __user *)&p;
2057 if (ops->ndo_do_ioctl) {
2058 mm_segment_t oldfs = get_fs();
2060 set_fs(KERNEL_DS);
2061 err = ops->ndo_do_ioctl(dev, &ifr, SIOCADDTUNNEL);
2062 set_fs(oldfs);
2063 } else
2064 err = -EOPNOTSUPP;
2066 if (err == 0) {
2067 err = -ENOBUFS;
2068 dev = __dev_get_by_name(net, p.name);
2069 if (!dev)
2070 goto err_exit;
2071 err = dev_open(dev);
2074 #endif
2076 err_exit:
2077 rtnl_unlock();
2078 return err;
2082 * Manual configuration of address on an interface
2084 static int inet6_addr_add(struct net *net, int ifindex, struct in6_addr *pfx,
2085 unsigned int plen, __u8 ifa_flags, __u32 prefered_lft,
2086 __u32 valid_lft)
2088 struct inet6_ifaddr *ifp;
2089 struct inet6_dev *idev;
2090 struct net_device *dev;
2091 int scope;
2092 u32 flags;
2093 clock_t expires;
2094 unsigned long timeout;
2096 ASSERT_RTNL();
2098 if (plen > 128)
2099 return -EINVAL;
2101 /* check the lifetime */
2102 if (!valid_lft || prefered_lft > valid_lft)
2103 return -EINVAL;
2105 dev = __dev_get_by_index(net, ifindex);
2106 if (!dev)
2107 return -ENODEV;
2109 if ((idev = addrconf_add_dev(dev)) == NULL)
2110 return -ENOBUFS;
2112 scope = ipv6_addr_scope(pfx);
2114 timeout = addrconf_timeout_fixup(valid_lft, HZ);
2115 if (addrconf_finite_timeout(timeout)) {
2116 expires = jiffies_to_clock_t(timeout * HZ);
2117 valid_lft = timeout;
2118 flags = RTF_EXPIRES;
2119 } else {
2120 expires = 0;
2121 flags = 0;
2122 ifa_flags |= IFA_F_PERMANENT;
2125 timeout = addrconf_timeout_fixup(prefered_lft, HZ);
2126 if (addrconf_finite_timeout(timeout)) {
2127 if (timeout == 0)
2128 ifa_flags |= IFA_F_DEPRECATED;
2129 prefered_lft = timeout;
2132 ifp = ipv6_add_addr(idev, pfx, plen, scope, ifa_flags);
2134 if (!IS_ERR(ifp)) {
2135 spin_lock_bh(&ifp->lock);
2136 ifp->valid_lft = valid_lft;
2137 ifp->prefered_lft = prefered_lft;
2138 ifp->tstamp = jiffies;
2139 spin_unlock_bh(&ifp->lock);
2141 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, dev,
2142 expires, flags);
2144 * Note that section 3.1 of RFC 4429 indicates
2145 * that the Optimistic flag should not be set for
2146 * manually configured addresses
2148 addrconf_dad_start(ifp, 0);
2149 in6_ifa_put(ifp);
2150 addrconf_verify(0);
2151 return 0;
2154 return PTR_ERR(ifp);
2157 static int inet6_addr_del(struct net *net, int ifindex, struct in6_addr *pfx,
2158 unsigned int plen)
2160 struct inet6_ifaddr *ifp;
2161 struct inet6_dev *idev;
2162 struct net_device *dev;
2164 if (plen > 128)
2165 return -EINVAL;
2167 dev = __dev_get_by_index(net, ifindex);
2168 if (!dev)
2169 return -ENODEV;
2171 if ((idev = __in6_dev_get(dev)) == NULL)
2172 return -ENXIO;
2174 read_lock_bh(&idev->lock);
2175 for (ifp = idev->addr_list; ifp; ifp=ifp->if_next) {
2176 if (ifp->prefix_len == plen &&
2177 ipv6_addr_equal(pfx, &ifp->addr)) {
2178 in6_ifa_hold(ifp);
2179 read_unlock_bh(&idev->lock);
2181 ipv6_del_addr(ifp);
2183 /* If the last address is deleted administratively,
2184 disable IPv6 on this interface.
2186 if (idev->addr_list == NULL)
2187 addrconf_ifdown(idev->dev, 1);
2188 return 0;
2191 read_unlock_bh(&idev->lock);
2192 return -EADDRNOTAVAIL;
2196 int addrconf_add_ifaddr(struct net *net, void __user *arg)
2198 struct in6_ifreq ireq;
2199 int err;
2201 if (!capable(CAP_NET_ADMIN))
2202 return -EPERM;
2204 if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2205 return -EFAULT;
2207 rtnl_lock();
2208 err = inet6_addr_add(net, ireq.ifr6_ifindex, &ireq.ifr6_addr,
2209 ireq.ifr6_prefixlen, IFA_F_PERMANENT,
2210 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
2211 rtnl_unlock();
2212 return err;
2215 int addrconf_del_ifaddr(struct net *net, void __user *arg)
2217 struct in6_ifreq ireq;
2218 int err;
2220 if (!capable(CAP_NET_ADMIN))
2221 return -EPERM;
2223 if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2224 return -EFAULT;
2226 rtnl_lock();
2227 err = inet6_addr_del(net, ireq.ifr6_ifindex, &ireq.ifr6_addr,
2228 ireq.ifr6_prefixlen);
2229 rtnl_unlock();
2230 return err;
2233 static void add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
2234 int plen, int scope)
2236 struct inet6_ifaddr *ifp;
2238 ifp = ipv6_add_addr(idev, addr, plen, scope, IFA_F_PERMANENT);
2239 if (!IS_ERR(ifp)) {
2240 spin_lock_bh(&ifp->lock);
2241 ifp->flags &= ~IFA_F_TENTATIVE;
2242 spin_unlock_bh(&ifp->lock);
2243 ipv6_ifa_notify(RTM_NEWADDR, ifp);
2244 in6_ifa_put(ifp);
2248 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2249 static void sit_add_v4_addrs(struct inet6_dev *idev)
2251 struct in6_addr addr;
2252 struct net_device *dev;
2253 struct net *net = dev_net(idev->dev);
2254 int scope;
2256 ASSERT_RTNL();
2258 memset(&addr, 0, sizeof(struct in6_addr));
2259 memcpy(&addr.s6_addr32[3], idev->dev->dev_addr, 4);
2261 if (idev->dev->flags&IFF_POINTOPOINT) {
2262 addr.s6_addr32[0] = htonl(0xfe800000);
2263 scope = IFA_LINK;
2264 } else {
2265 scope = IPV6_ADDR_COMPATv4;
2268 if (addr.s6_addr32[3]) {
2269 add_addr(idev, &addr, 128, scope);
2270 return;
2273 for_each_netdev(net, dev) {
2274 struct in_device * in_dev = __in_dev_get_rtnl(dev);
2275 if (in_dev && (dev->flags & IFF_UP)) {
2276 struct in_ifaddr * ifa;
2278 int flag = scope;
2280 for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) {
2281 int plen;
2283 addr.s6_addr32[3] = ifa->ifa_local;
2285 if (ifa->ifa_scope == RT_SCOPE_LINK)
2286 continue;
2287 if (ifa->ifa_scope >= RT_SCOPE_HOST) {
2288 if (idev->dev->flags&IFF_POINTOPOINT)
2289 continue;
2290 flag |= IFA_HOST;
2292 if (idev->dev->flags&IFF_POINTOPOINT)
2293 plen = 64;
2294 else
2295 plen = 96;
2297 add_addr(idev, &addr, plen, flag);
2302 #endif
2304 static void init_loopback(struct net_device *dev)
2306 struct inet6_dev *idev;
2308 /* ::1 */
2310 ASSERT_RTNL();
2312 if ((idev = ipv6_find_idev(dev)) == NULL) {
2313 printk(KERN_DEBUG "init loopback: add_dev failed\n");
2314 return;
2317 add_addr(idev, &in6addr_loopback, 128, IFA_HOST);
2320 static void addrconf_add_linklocal(struct inet6_dev *idev, struct in6_addr *addr)
2322 struct inet6_ifaddr * ifp;
2323 u32 addr_flags = IFA_F_PERMANENT;
2325 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2326 if (idev->cnf.optimistic_dad &&
2327 !dev_net(idev->dev)->ipv6.devconf_all->forwarding)
2328 addr_flags |= IFA_F_OPTIMISTIC;
2329 #endif
2332 ifp = ipv6_add_addr(idev, addr, 64, IFA_LINK, addr_flags);
2333 if (!IS_ERR(ifp)) {
2334 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, idev->dev, 0, 0);
2335 addrconf_dad_start(ifp, 0);
2336 in6_ifa_put(ifp);
2340 static void addrconf_dev_config(struct net_device *dev)
2342 struct in6_addr addr;
2343 struct inet6_dev * idev;
2345 ASSERT_RTNL();
2347 if ((dev->type != ARPHRD_ETHER) &&
2348 (dev->type != ARPHRD_FDDI) &&
2349 (dev->type != ARPHRD_IEEE802_TR) &&
2350 (dev->type != ARPHRD_ARCNET) &&
2351 (dev->type != ARPHRD_INFINIBAND)) {
2352 /* Alas, we support only Ethernet autoconfiguration. */
2353 return;
2356 idev = addrconf_add_dev(dev);
2357 if (idev == NULL)
2358 return;
2360 memset(&addr, 0, sizeof(struct in6_addr));
2361 addr.s6_addr32[0] = htonl(0xFE800000);
2363 if (ipv6_generate_eui64(addr.s6_addr + 8, dev) == 0)
2364 addrconf_add_linklocal(idev, &addr);
2367 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2368 static void addrconf_sit_config(struct net_device *dev)
2370 struct inet6_dev *idev;
2372 ASSERT_RTNL();
2375 * Configure the tunnel with one of our IPv4
2376 * addresses... we should configure all of
2377 * our v4 addrs in the tunnel
2380 if ((idev = ipv6_find_idev(dev)) == NULL) {
2381 printk(KERN_DEBUG "init sit: add_dev failed\n");
2382 return;
2385 if (dev->priv_flags & IFF_ISATAP) {
2386 struct in6_addr addr;
2388 ipv6_addr_set(&addr, htonl(0xFE800000), 0, 0, 0);
2389 addrconf_prefix_route(&addr, 64, dev, 0, 0);
2390 if (!ipv6_generate_eui64(addr.s6_addr + 8, dev))
2391 addrconf_add_linklocal(idev, &addr);
2392 return;
2395 sit_add_v4_addrs(idev);
2397 if (dev->flags&IFF_POINTOPOINT) {
2398 addrconf_add_mroute(dev);
2399 addrconf_add_lroute(dev);
2400 } else
2401 sit_route_add(dev);
2403 #endif
2405 static inline int
2406 ipv6_inherit_linklocal(struct inet6_dev *idev, struct net_device *link_dev)
2408 struct in6_addr lladdr;
2410 if (!ipv6_get_lladdr(link_dev, &lladdr, IFA_F_TENTATIVE)) {
2411 addrconf_add_linklocal(idev, &lladdr);
2412 return 0;
2414 return -1;
2417 static void ip6_tnl_add_linklocal(struct inet6_dev *idev)
2419 struct net_device *link_dev;
2420 struct net *net = dev_net(idev->dev);
2422 /* first try to inherit the link-local address from the link device */
2423 if (idev->dev->iflink &&
2424 (link_dev = __dev_get_by_index(net, idev->dev->iflink))) {
2425 if (!ipv6_inherit_linklocal(idev, link_dev))
2426 return;
2428 /* then try to inherit it from any device */
2429 for_each_netdev(net, link_dev) {
2430 if (!ipv6_inherit_linklocal(idev, link_dev))
2431 return;
2433 printk(KERN_DEBUG "init ip6-ip6: add_linklocal failed\n");
2437 * Autoconfigure tunnel with a link-local address so routing protocols,
2438 * DHCPv6, MLD etc. can be run over the virtual link
2441 static void addrconf_ip6_tnl_config(struct net_device *dev)
2443 struct inet6_dev *idev;
2445 ASSERT_RTNL();
2447 if ((idev = addrconf_add_dev(dev)) == NULL) {
2448 printk(KERN_DEBUG "init ip6-ip6: add_dev failed\n");
2449 return;
2451 ip6_tnl_add_linklocal(idev);
2454 static int addrconf_notify(struct notifier_block *this, unsigned long event,
2455 void * data)
2457 struct net_device *dev = (struct net_device *) data;
2458 struct inet6_dev *idev = __in6_dev_get(dev);
2459 int run_pending = 0;
2460 int err;
2462 switch(event) {
2463 case NETDEV_REGISTER:
2464 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2465 idev = ipv6_add_dev(dev);
2466 if (!idev)
2467 return notifier_from_errno(-ENOMEM);
2469 break;
2470 case NETDEV_UP:
2471 case NETDEV_CHANGE:
2472 if (dev->flags & IFF_SLAVE)
2473 break;
2475 if (event == NETDEV_UP) {
2476 if (!addrconf_qdisc_ok(dev)) {
2477 /* device is not ready yet. */
2478 printk(KERN_INFO
2479 "ADDRCONF(NETDEV_UP): %s: "
2480 "link is not ready\n",
2481 dev->name);
2482 break;
2485 if (!idev && dev->mtu >= IPV6_MIN_MTU)
2486 idev = ipv6_add_dev(dev);
2488 if (idev) {
2489 idev->if_flags |= IF_READY;
2490 run_pending = 1;
2492 } else {
2493 if (!addrconf_qdisc_ok(dev)) {
2494 /* device is still not ready. */
2495 break;
2498 if (idev) {
2499 if (idev->if_flags & IF_READY) {
2500 /* device is already configured. */
2501 break;
2503 idev->if_flags |= IF_READY;
2506 printk(KERN_INFO
2507 "ADDRCONF(NETDEV_CHANGE): %s: "
2508 "link becomes ready\n",
2509 dev->name);
2511 run_pending = 1;
2514 switch(dev->type) {
2515 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2516 case ARPHRD_SIT:
2517 addrconf_sit_config(dev);
2518 break;
2519 #endif
2520 case ARPHRD_TUNNEL6:
2521 addrconf_ip6_tnl_config(dev);
2522 break;
2523 case ARPHRD_LOOPBACK:
2524 init_loopback(dev);
2525 break;
2527 default:
2528 addrconf_dev_config(dev);
2529 break;
2531 if (idev) {
2532 if (run_pending)
2533 addrconf_dad_run(idev);
2535 /* If the MTU changed during the interface down, when the
2536 interface up, the changed MTU must be reflected in the
2537 idev as well as routers.
2539 if (idev->cnf.mtu6 != dev->mtu && dev->mtu >= IPV6_MIN_MTU) {
2540 rt6_mtu_change(dev, dev->mtu);
2541 idev->cnf.mtu6 = dev->mtu;
2543 idev->tstamp = jiffies;
2544 inet6_ifinfo_notify(RTM_NEWLINK, idev);
2545 /* If the changed mtu during down is lower than IPV6_MIN_MTU
2546 stop IPv6 on this interface.
2548 if (dev->mtu < IPV6_MIN_MTU)
2549 addrconf_ifdown(dev, event != NETDEV_DOWN);
2551 break;
2553 case NETDEV_CHANGEMTU:
2554 if (idev && dev->mtu >= IPV6_MIN_MTU) {
2555 rt6_mtu_change(dev, dev->mtu);
2556 idev->cnf.mtu6 = dev->mtu;
2557 break;
2560 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2561 idev = ipv6_add_dev(dev);
2562 if (idev)
2563 break;
2566 /* MTU falled under IPV6_MIN_MTU. Stop IPv6 on this interface. */
2568 case NETDEV_DOWN:
2569 case NETDEV_UNREGISTER:
2571 * Remove all addresses from this interface.
2573 addrconf_ifdown(dev, event != NETDEV_DOWN);
2574 break;
2576 case NETDEV_CHANGENAME:
2577 if (idev) {
2578 snmp6_unregister_dev(idev);
2579 addrconf_sysctl_unregister(idev);
2580 addrconf_sysctl_register(idev);
2581 err = snmp6_register_dev(idev);
2582 if (err)
2583 return notifier_from_errno(err);
2585 break;
2586 case NETDEV_BONDING_OLDTYPE:
2587 case NETDEV_BONDING_NEWTYPE:
2588 addrconf_bonding_change(dev, event);
2589 break;
2592 return NOTIFY_OK;
2596 * addrconf module should be notified of a device going up
2598 static struct notifier_block ipv6_dev_notf = {
2599 .notifier_call = addrconf_notify,
2600 .priority = 0
2603 static void addrconf_bonding_change(struct net_device *dev, unsigned long event)
2605 struct inet6_dev *idev;
2606 ASSERT_RTNL();
2608 idev = __in6_dev_get(dev);
2610 if (event == NETDEV_BONDING_NEWTYPE)
2611 ipv6_mc_remap(idev);
2612 else if (event == NETDEV_BONDING_OLDTYPE)
2613 ipv6_mc_unmap(idev);
2616 static int addrconf_ifdown(struct net_device *dev, int how)
2618 struct inet6_dev *idev;
2619 struct inet6_ifaddr *ifa, **bifa;
2620 struct net *net = dev_net(dev);
2621 int i;
2623 ASSERT_RTNL();
2625 rt6_ifdown(net, dev);
2626 neigh_ifdown(&nd_tbl, dev);
2628 idev = __in6_dev_get(dev);
2629 if (idev == NULL)
2630 return -ENODEV;
2632 /* Step 1: remove reference to ipv6 device from parent device.
2633 Do not dev_put!
2635 if (how) {
2636 idev->dead = 1;
2638 /* protected by rtnl_lock */
2639 rcu_assign_pointer(dev->ip6_ptr, NULL);
2641 /* Step 1.5: remove snmp6 entry */
2642 snmp6_unregister_dev(idev);
2646 /* Step 2: clear hash table */
2647 for (i=0; i<IN6_ADDR_HSIZE; i++) {
2648 bifa = &inet6_addr_lst[i];
2650 write_lock_bh(&addrconf_hash_lock);
2651 while ((ifa = *bifa) != NULL) {
2652 if (ifa->idev == idev) {
2653 *bifa = ifa->lst_next;
2654 ifa->lst_next = NULL;
2655 addrconf_del_timer(ifa);
2656 in6_ifa_put(ifa);
2657 continue;
2659 bifa = &ifa->lst_next;
2661 write_unlock_bh(&addrconf_hash_lock);
2664 write_lock_bh(&idev->lock);
2666 /* Step 3: clear flags for stateless addrconf */
2667 if (!how)
2668 idev->if_flags &= ~(IF_RS_SENT|IF_RA_RCVD|IF_READY);
2670 /* Step 4: clear address list */
2671 #ifdef CONFIG_IPV6_PRIVACY
2672 if (how && del_timer(&idev->regen_timer))
2673 in6_dev_put(idev);
2675 /* clear tempaddr list */
2676 while ((ifa = idev->tempaddr_list) != NULL) {
2677 idev->tempaddr_list = ifa->tmp_next;
2678 ifa->tmp_next = NULL;
2679 ifa->dead = 1;
2680 write_unlock_bh(&idev->lock);
2681 spin_lock_bh(&ifa->lock);
2683 if (ifa->ifpub) {
2684 in6_ifa_put(ifa->ifpub);
2685 ifa->ifpub = NULL;
2687 spin_unlock_bh(&ifa->lock);
2688 in6_ifa_put(ifa);
2689 write_lock_bh(&idev->lock);
2691 #endif
2692 while ((ifa = idev->addr_list) != NULL) {
2693 idev->addr_list = ifa->if_next;
2694 ifa->if_next = NULL;
2695 ifa->dead = 1;
2696 addrconf_del_timer(ifa);
2697 write_unlock_bh(&idev->lock);
2699 __ipv6_ifa_notify(RTM_DELADDR, ifa);
2700 atomic_notifier_call_chain(&inet6addr_chain, NETDEV_DOWN, ifa);
2701 in6_ifa_put(ifa);
2703 write_lock_bh(&idev->lock);
2705 write_unlock_bh(&idev->lock);
2707 /* Step 5: Discard multicast list */
2709 if (how)
2710 ipv6_mc_destroy_dev(idev);
2711 else
2712 ipv6_mc_down(idev);
2714 idev->tstamp = jiffies;
2716 /* Shot the device (if unregistered) */
2718 if (how) {
2719 addrconf_sysctl_unregister(idev);
2720 neigh_parms_release(&nd_tbl, idev->nd_parms);
2721 neigh_ifdown(&nd_tbl, dev);
2722 in6_dev_put(idev);
2724 return 0;
2727 static void addrconf_rs_timer(unsigned long data)
2729 struct inet6_ifaddr *ifp = (struct inet6_ifaddr *) data;
2731 if (ifp->idev->cnf.forwarding)
2732 goto out;
2734 if (ifp->idev->if_flags & IF_RA_RCVD) {
2736 * Announcement received after solicitation
2737 * was sent
2739 goto out;
2742 spin_lock(&ifp->lock);
2743 if (ifp->probes++ < ifp->idev->cnf.rtr_solicits) {
2744 /* The wait after the last probe can be shorter */
2745 addrconf_mod_timer(ifp, AC_RS,
2746 (ifp->probes == ifp->idev->cnf.rtr_solicits) ?
2747 ifp->idev->cnf.rtr_solicit_delay :
2748 ifp->idev->cnf.rtr_solicit_interval);
2749 spin_unlock(&ifp->lock);
2751 ndisc_send_rs(ifp->idev->dev, &ifp->addr, &in6addr_linklocal_allrouters);
2752 } else {
2753 spin_unlock(&ifp->lock);
2755 * Note: we do not support deprecated "all on-link"
2756 * assumption any longer.
2758 printk(KERN_DEBUG "%s: no IPv6 routers present\n",
2759 ifp->idev->dev->name);
2762 out:
2763 in6_ifa_put(ifp);
2767 * Duplicate Address Detection
2769 static void addrconf_dad_kick(struct inet6_ifaddr *ifp)
2771 unsigned long rand_num;
2772 struct inet6_dev *idev = ifp->idev;
2774 if (ifp->flags & IFA_F_OPTIMISTIC)
2775 rand_num = 0;
2776 else
2777 rand_num = net_random() % (idev->cnf.rtr_solicit_delay ? : 1);
2779 ifp->probes = idev->cnf.dad_transmits;
2780 addrconf_mod_timer(ifp, AC_DAD, rand_num);
2783 static void addrconf_dad_start(struct inet6_ifaddr *ifp, u32 flags)
2785 struct inet6_dev *idev = ifp->idev;
2786 struct net_device *dev = idev->dev;
2788 addrconf_join_solict(dev, &ifp->addr);
2790 net_srandom(ifp->addr.s6_addr32[3]);
2792 read_lock_bh(&idev->lock);
2793 if (ifp->dead)
2794 goto out;
2795 spin_lock_bh(&ifp->lock);
2797 if (dev->flags&(IFF_NOARP|IFF_LOOPBACK) ||
2798 idev->cnf.accept_dad < 1 ||
2799 !(ifp->flags&IFA_F_TENTATIVE) ||
2800 ifp->flags & IFA_F_NODAD) {
2801 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
2802 spin_unlock_bh(&ifp->lock);
2803 read_unlock_bh(&idev->lock);
2805 addrconf_dad_completed(ifp);
2806 return;
2809 if (!(idev->if_flags & IF_READY)) {
2810 spin_unlock_bh(&ifp->lock);
2811 read_unlock_bh(&idev->lock);
2813 * If the device is not ready:
2814 * - keep it tentative if it is a permanent address.
2815 * - otherwise, kill it.
2817 in6_ifa_hold(ifp);
2818 addrconf_dad_stop(ifp, 0);
2819 return;
2823 * Optimistic nodes can start receiving
2824 * Frames right away
2826 if(ifp->flags & IFA_F_OPTIMISTIC)
2827 ip6_ins_rt(ifp->rt);
2829 addrconf_dad_kick(ifp);
2830 spin_unlock_bh(&ifp->lock);
2831 out:
2832 read_unlock_bh(&idev->lock);
2835 static void addrconf_dad_timer(unsigned long data)
2837 struct inet6_ifaddr *ifp = (struct inet6_ifaddr *) data;
2838 struct inet6_dev *idev = ifp->idev;
2839 struct in6_addr mcaddr;
2841 read_lock_bh(&idev->lock);
2842 if (idev->dead) {
2843 read_unlock_bh(&idev->lock);
2844 goto out;
2846 spin_lock_bh(&ifp->lock);
2847 if (ifp->probes == 0) {
2849 * DAD was successful
2852 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
2853 spin_unlock_bh(&ifp->lock);
2854 read_unlock_bh(&idev->lock);
2856 addrconf_dad_completed(ifp);
2858 goto out;
2861 ifp->probes--;
2862 addrconf_mod_timer(ifp, AC_DAD, ifp->idev->nd_parms->retrans_time);
2863 spin_unlock_bh(&ifp->lock);
2864 read_unlock_bh(&idev->lock);
2866 /* send a neighbour solicitation for our addr */
2867 addrconf_addr_solict_mult(&ifp->addr, &mcaddr);
2868 ndisc_send_ns(ifp->idev->dev, NULL, &ifp->addr, &mcaddr, &in6addr_any);
2869 out:
2870 in6_ifa_put(ifp);
2873 static void addrconf_dad_completed(struct inet6_ifaddr *ifp)
2875 struct net_device * dev = ifp->idev->dev;
2878 * Configure the address for reception. Now it is valid.
2881 ipv6_ifa_notify(RTM_NEWADDR, ifp);
2883 /* If added prefix is link local and forwarding is off,
2884 start sending router solicitations.
2887 if (ifp->idev->cnf.forwarding == 0 &&
2888 ifp->idev->cnf.rtr_solicits > 0 &&
2889 (dev->flags&IFF_LOOPBACK) == 0 &&
2890 (ipv6_addr_type(&ifp->addr) & IPV6_ADDR_LINKLOCAL)) {
2892 * If a host as already performed a random delay
2893 * [...] as part of DAD [...] there is no need
2894 * to delay again before sending the first RS
2896 ndisc_send_rs(ifp->idev->dev, &ifp->addr, &in6addr_linklocal_allrouters);
2898 spin_lock_bh(&ifp->lock);
2899 ifp->probes = 1;
2900 ifp->idev->if_flags |= IF_RS_SENT;
2901 addrconf_mod_timer(ifp, AC_RS, ifp->idev->cnf.rtr_solicit_interval);
2902 spin_unlock_bh(&ifp->lock);
2906 static void addrconf_dad_run(struct inet6_dev *idev) {
2907 struct inet6_ifaddr *ifp;
2909 read_lock_bh(&idev->lock);
2910 for (ifp = idev->addr_list; ifp; ifp = ifp->if_next) {
2911 spin_lock_bh(&ifp->lock);
2912 if (!(ifp->flags & IFA_F_TENTATIVE)) {
2913 spin_unlock_bh(&ifp->lock);
2914 continue;
2916 spin_unlock_bh(&ifp->lock);
2917 addrconf_dad_kick(ifp);
2919 read_unlock_bh(&idev->lock);
2922 #ifdef CONFIG_PROC_FS
2923 struct if6_iter_state {
2924 struct seq_net_private p;
2925 int bucket;
2928 static struct inet6_ifaddr *if6_get_first(struct seq_file *seq)
2930 struct inet6_ifaddr *ifa = NULL;
2931 struct if6_iter_state *state = seq->private;
2932 struct net *net = seq_file_net(seq);
2934 for (state->bucket = 0; state->bucket < IN6_ADDR_HSIZE; ++state->bucket) {
2935 ifa = inet6_addr_lst[state->bucket];
2937 while (ifa && !net_eq(dev_net(ifa->idev->dev), net))
2938 ifa = ifa->lst_next;
2939 if (ifa)
2940 break;
2942 return ifa;
2945 static struct inet6_ifaddr *if6_get_next(struct seq_file *seq, struct inet6_ifaddr *ifa)
2947 struct if6_iter_state *state = seq->private;
2948 struct net *net = seq_file_net(seq);
2950 ifa = ifa->lst_next;
2951 try_again:
2952 if (ifa) {
2953 if (!net_eq(dev_net(ifa->idev->dev), net)) {
2954 ifa = ifa->lst_next;
2955 goto try_again;
2959 if (!ifa && ++state->bucket < IN6_ADDR_HSIZE) {
2960 ifa = inet6_addr_lst[state->bucket];
2961 goto try_again;
2964 return ifa;
2967 static struct inet6_ifaddr *if6_get_idx(struct seq_file *seq, loff_t pos)
2969 struct inet6_ifaddr *ifa = if6_get_first(seq);
2971 if (ifa)
2972 while(pos && (ifa = if6_get_next(seq, ifa)) != NULL)
2973 --pos;
2974 return pos ? NULL : ifa;
2977 static void *if6_seq_start(struct seq_file *seq, loff_t *pos)
2978 __acquires(addrconf_hash_lock)
2980 read_lock_bh(&addrconf_hash_lock);
2981 return if6_get_idx(seq, *pos);
2984 static void *if6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
2986 struct inet6_ifaddr *ifa;
2988 ifa = if6_get_next(seq, v);
2989 ++*pos;
2990 return ifa;
2993 static void if6_seq_stop(struct seq_file *seq, void *v)
2994 __releases(addrconf_hash_lock)
2996 read_unlock_bh(&addrconf_hash_lock);
2999 static int if6_seq_show(struct seq_file *seq, void *v)
3001 struct inet6_ifaddr *ifp = (struct inet6_ifaddr *)v;
3002 seq_printf(seq, "%pi6 %02x %02x %02x %02x %8s\n",
3003 &ifp->addr,
3004 ifp->idev->dev->ifindex,
3005 ifp->prefix_len,
3006 ifp->scope,
3007 ifp->flags,
3008 ifp->idev->dev->name);
3009 return 0;
3012 static const struct seq_operations if6_seq_ops = {
3013 .start = if6_seq_start,
3014 .next = if6_seq_next,
3015 .show = if6_seq_show,
3016 .stop = if6_seq_stop,
3019 static int if6_seq_open(struct inode *inode, struct file *file)
3021 return seq_open_net(inode, file, &if6_seq_ops,
3022 sizeof(struct if6_iter_state));
3025 static const struct file_operations if6_fops = {
3026 .owner = THIS_MODULE,
3027 .open = if6_seq_open,
3028 .read = seq_read,
3029 .llseek = seq_lseek,
3030 .release = seq_release_net,
3033 static int if6_proc_net_init(struct net *net)
3035 if (!proc_net_fops_create(net, "if_inet6", S_IRUGO, &if6_fops))
3036 return -ENOMEM;
3037 return 0;
3040 static void if6_proc_net_exit(struct net *net)
3042 proc_net_remove(net, "if_inet6");
3045 static struct pernet_operations if6_proc_net_ops = {
3046 .init = if6_proc_net_init,
3047 .exit = if6_proc_net_exit,
3050 int __init if6_proc_init(void)
3052 return register_pernet_subsys(&if6_proc_net_ops);
3055 void if6_proc_exit(void)
3057 unregister_pernet_subsys(&if6_proc_net_ops);
3059 #endif /* CONFIG_PROC_FS */
3061 #if defined(CONFIG_IPV6_MIP6) || defined(CONFIG_IPV6_MIP6_MODULE)
3062 /* Check if address is a home address configured on any interface. */
3063 int ipv6_chk_home_addr(struct net *net, struct in6_addr *addr)
3065 int ret = 0;
3066 struct inet6_ifaddr * ifp;
3067 u8 hash = ipv6_addr_hash(addr);
3068 read_lock_bh(&addrconf_hash_lock);
3069 for (ifp = inet6_addr_lst[hash]; ifp; ifp = ifp->lst_next) {
3070 if (!net_eq(dev_net(ifp->idev->dev), net))
3071 continue;
3072 if (ipv6_addr_equal(&ifp->addr, addr) &&
3073 (ifp->flags & IFA_F_HOMEADDRESS)) {
3074 ret = 1;
3075 break;
3078 read_unlock_bh(&addrconf_hash_lock);
3079 return ret;
3081 #endif
3084 * Periodic address status verification
3087 static void addrconf_verify(unsigned long foo)
3089 struct inet6_ifaddr *ifp;
3090 unsigned long now, next;
3091 int i;
3093 spin_lock_bh(&addrconf_verify_lock);
3094 now = jiffies;
3095 next = now + ADDR_CHECK_FREQUENCY;
3097 del_timer(&addr_chk_timer);
3099 for (i=0; i < IN6_ADDR_HSIZE; i++) {
3101 restart:
3102 read_lock(&addrconf_hash_lock);
3103 for (ifp=inet6_addr_lst[i]; ifp; ifp=ifp->lst_next) {
3104 unsigned long age;
3105 #ifdef CONFIG_IPV6_PRIVACY
3106 unsigned long regen_advance;
3107 #endif
3109 if (ifp->flags & IFA_F_PERMANENT)
3110 continue;
3112 spin_lock(&ifp->lock);
3113 age = (now - ifp->tstamp) / HZ;
3115 #ifdef CONFIG_IPV6_PRIVACY
3116 regen_advance = ifp->idev->cnf.regen_max_retry *
3117 ifp->idev->cnf.dad_transmits *
3118 ifp->idev->nd_parms->retrans_time / HZ;
3119 #endif
3121 if (ifp->valid_lft != INFINITY_LIFE_TIME &&
3122 age >= ifp->valid_lft) {
3123 spin_unlock(&ifp->lock);
3124 in6_ifa_hold(ifp);
3125 read_unlock(&addrconf_hash_lock);
3126 ipv6_del_addr(ifp);
3127 goto restart;
3128 } else if (ifp->prefered_lft == INFINITY_LIFE_TIME) {
3129 spin_unlock(&ifp->lock);
3130 continue;
3131 } else if (age >= ifp->prefered_lft) {
3132 /* jiffies - ifp->tstamp > age >= ifp->prefered_lft */
3133 int deprecate = 0;
3135 if (!(ifp->flags&IFA_F_DEPRECATED)) {
3136 deprecate = 1;
3137 ifp->flags |= IFA_F_DEPRECATED;
3140 if (time_before(ifp->tstamp + ifp->valid_lft * HZ, next))
3141 next = ifp->tstamp + ifp->valid_lft * HZ;
3143 spin_unlock(&ifp->lock);
3145 if (deprecate) {
3146 in6_ifa_hold(ifp);
3147 read_unlock(&addrconf_hash_lock);
3149 ipv6_ifa_notify(0, ifp);
3150 in6_ifa_put(ifp);
3151 goto restart;
3153 #ifdef CONFIG_IPV6_PRIVACY
3154 } else if ((ifp->flags&IFA_F_TEMPORARY) &&
3155 !(ifp->flags&IFA_F_TENTATIVE)) {
3156 if (age >= ifp->prefered_lft - regen_advance) {
3157 struct inet6_ifaddr *ifpub = ifp->ifpub;
3158 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3159 next = ifp->tstamp + ifp->prefered_lft * HZ;
3160 if (!ifp->regen_count && ifpub) {
3161 ifp->regen_count++;
3162 in6_ifa_hold(ifp);
3163 in6_ifa_hold(ifpub);
3164 spin_unlock(&ifp->lock);
3165 read_unlock(&addrconf_hash_lock);
3166 spin_lock(&ifpub->lock);
3167 ifpub->regen_count = 0;
3168 spin_unlock(&ifpub->lock);
3169 ipv6_create_tempaddr(ifpub, ifp);
3170 in6_ifa_put(ifpub);
3171 in6_ifa_put(ifp);
3172 goto restart;
3174 } else if (time_before(ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ, next))
3175 next = ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ;
3176 spin_unlock(&ifp->lock);
3177 #endif
3178 } else {
3179 /* ifp->prefered_lft <= ifp->valid_lft */
3180 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3181 next = ifp->tstamp + ifp->prefered_lft * HZ;
3182 spin_unlock(&ifp->lock);
3185 read_unlock(&addrconf_hash_lock);
3188 addr_chk_timer.expires = time_before(next, jiffies + HZ) ? jiffies + HZ : next;
3189 add_timer(&addr_chk_timer);
3190 spin_unlock_bh(&addrconf_verify_lock);
3193 static struct in6_addr *extract_addr(struct nlattr *addr, struct nlattr *local)
3195 struct in6_addr *pfx = NULL;
3197 if (addr)
3198 pfx = nla_data(addr);
3200 if (local) {
3201 if (pfx && nla_memcmp(local, pfx, sizeof(*pfx)))
3202 pfx = NULL;
3203 else
3204 pfx = nla_data(local);
3207 return pfx;
3210 static const struct nla_policy ifa_ipv6_policy[IFA_MAX+1] = {
3211 [IFA_ADDRESS] = { .len = sizeof(struct in6_addr) },
3212 [IFA_LOCAL] = { .len = sizeof(struct in6_addr) },
3213 [IFA_CACHEINFO] = { .len = sizeof(struct ifa_cacheinfo) },
3216 static int
3217 inet6_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
3219 struct net *net = sock_net(skb->sk);
3220 struct ifaddrmsg *ifm;
3221 struct nlattr *tb[IFA_MAX+1];
3222 struct in6_addr *pfx;
3223 int err;
3225 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3226 if (err < 0)
3227 return err;
3229 ifm = nlmsg_data(nlh);
3230 pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3231 if (pfx == NULL)
3232 return -EINVAL;
3234 return inet6_addr_del(net, ifm->ifa_index, pfx, ifm->ifa_prefixlen);
3237 static int inet6_addr_modify(struct inet6_ifaddr *ifp, u8 ifa_flags,
3238 u32 prefered_lft, u32 valid_lft)
3240 u32 flags;
3241 clock_t expires;
3242 unsigned long timeout;
3244 if (!valid_lft || (prefered_lft > valid_lft))
3245 return -EINVAL;
3247 timeout = addrconf_timeout_fixup(valid_lft, HZ);
3248 if (addrconf_finite_timeout(timeout)) {
3249 expires = jiffies_to_clock_t(timeout * HZ);
3250 valid_lft = timeout;
3251 flags = RTF_EXPIRES;
3252 } else {
3253 expires = 0;
3254 flags = 0;
3255 ifa_flags |= IFA_F_PERMANENT;
3258 timeout = addrconf_timeout_fixup(prefered_lft, HZ);
3259 if (addrconf_finite_timeout(timeout)) {
3260 if (timeout == 0)
3261 ifa_flags |= IFA_F_DEPRECATED;
3262 prefered_lft = timeout;
3265 spin_lock_bh(&ifp->lock);
3266 ifp->flags = (ifp->flags & ~(IFA_F_DEPRECATED | IFA_F_PERMANENT | IFA_F_NODAD | IFA_F_HOMEADDRESS)) | ifa_flags;
3267 ifp->tstamp = jiffies;
3268 ifp->valid_lft = valid_lft;
3269 ifp->prefered_lft = prefered_lft;
3271 spin_unlock_bh(&ifp->lock);
3272 if (!(ifp->flags&IFA_F_TENTATIVE))
3273 ipv6_ifa_notify(0, ifp);
3275 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, ifp->idev->dev,
3276 expires, flags);
3277 addrconf_verify(0);
3279 return 0;
3282 static int
3283 inet6_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
3285 struct net *net = sock_net(skb->sk);
3286 struct ifaddrmsg *ifm;
3287 struct nlattr *tb[IFA_MAX+1];
3288 struct in6_addr *pfx;
3289 struct inet6_ifaddr *ifa;
3290 struct net_device *dev;
3291 u32 valid_lft = INFINITY_LIFE_TIME, preferred_lft = INFINITY_LIFE_TIME;
3292 u8 ifa_flags;
3293 int err;
3295 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3296 if (err < 0)
3297 return err;
3299 ifm = nlmsg_data(nlh);
3300 pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3301 if (pfx == NULL)
3302 return -EINVAL;
3304 if (tb[IFA_CACHEINFO]) {
3305 struct ifa_cacheinfo *ci;
3307 ci = nla_data(tb[IFA_CACHEINFO]);
3308 valid_lft = ci->ifa_valid;
3309 preferred_lft = ci->ifa_prefered;
3310 } else {
3311 preferred_lft = INFINITY_LIFE_TIME;
3312 valid_lft = INFINITY_LIFE_TIME;
3315 dev = __dev_get_by_index(net, ifm->ifa_index);
3316 if (dev == NULL)
3317 return -ENODEV;
3319 /* We ignore other flags so far. */
3320 ifa_flags = ifm->ifa_flags & (IFA_F_NODAD | IFA_F_HOMEADDRESS);
3322 ifa = ipv6_get_ifaddr(net, pfx, dev, 1);
3323 if (ifa == NULL) {
3325 * It would be best to check for !NLM_F_CREATE here but
3326 * userspace alreay relies on not having to provide this.
3328 return inet6_addr_add(net, ifm->ifa_index, pfx,
3329 ifm->ifa_prefixlen, ifa_flags,
3330 preferred_lft, valid_lft);
3333 if (nlh->nlmsg_flags & NLM_F_EXCL ||
3334 !(nlh->nlmsg_flags & NLM_F_REPLACE))
3335 err = -EEXIST;
3336 else
3337 err = inet6_addr_modify(ifa, ifa_flags, preferred_lft, valid_lft);
3339 in6_ifa_put(ifa);
3341 return err;
3344 static void put_ifaddrmsg(struct nlmsghdr *nlh, u8 prefixlen, u8 flags,
3345 u8 scope, int ifindex)
3347 struct ifaddrmsg *ifm;
3349 ifm = nlmsg_data(nlh);
3350 ifm->ifa_family = AF_INET6;
3351 ifm->ifa_prefixlen = prefixlen;
3352 ifm->ifa_flags = flags;
3353 ifm->ifa_scope = scope;
3354 ifm->ifa_index = ifindex;
3357 static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
3358 unsigned long tstamp, u32 preferred, u32 valid)
3360 struct ifa_cacheinfo ci;
3362 ci.cstamp = (u32)(TIME_DELTA(cstamp, INITIAL_JIFFIES) / HZ * 100
3363 + TIME_DELTA(cstamp, INITIAL_JIFFIES) % HZ * 100 / HZ);
3364 ci.tstamp = (u32)(TIME_DELTA(tstamp, INITIAL_JIFFIES) / HZ * 100
3365 + TIME_DELTA(tstamp, INITIAL_JIFFIES) % HZ * 100 / HZ);
3366 ci.ifa_prefered = preferred;
3367 ci.ifa_valid = valid;
3369 return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
3372 static inline int rt_scope(int ifa_scope)
3374 if (ifa_scope & IFA_HOST)
3375 return RT_SCOPE_HOST;
3376 else if (ifa_scope & IFA_LINK)
3377 return RT_SCOPE_LINK;
3378 else if (ifa_scope & IFA_SITE)
3379 return RT_SCOPE_SITE;
3380 else
3381 return RT_SCOPE_UNIVERSE;
3384 static inline int inet6_ifaddr_msgsize(void)
3386 return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
3387 + nla_total_size(16) /* IFA_ADDRESS */
3388 + nla_total_size(sizeof(struct ifa_cacheinfo));
3391 static int inet6_fill_ifaddr(struct sk_buff *skb, struct inet6_ifaddr *ifa,
3392 u32 pid, u32 seq, int event, unsigned int flags)
3394 struct nlmsghdr *nlh;
3395 u32 preferred, valid;
3397 nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3398 if (nlh == NULL)
3399 return -EMSGSIZE;
3401 put_ifaddrmsg(nlh, ifa->prefix_len, ifa->flags, rt_scope(ifa->scope),
3402 ifa->idev->dev->ifindex);
3404 if (!(ifa->flags&IFA_F_PERMANENT)) {
3405 preferred = ifa->prefered_lft;
3406 valid = ifa->valid_lft;
3407 if (preferred != INFINITY_LIFE_TIME) {
3408 long tval = (jiffies - ifa->tstamp)/HZ;
3409 if (preferred > tval)
3410 preferred -= tval;
3411 else
3412 preferred = 0;
3413 if (valid != INFINITY_LIFE_TIME)
3414 valid -= tval;
3416 } else {
3417 preferred = INFINITY_LIFE_TIME;
3418 valid = INFINITY_LIFE_TIME;
3421 if (nla_put(skb, IFA_ADDRESS, 16, &ifa->addr) < 0 ||
3422 put_cacheinfo(skb, ifa->cstamp, ifa->tstamp, preferred, valid) < 0) {
3423 nlmsg_cancel(skb, nlh);
3424 return -EMSGSIZE;
3427 return nlmsg_end(skb, nlh);
3430 static int inet6_fill_ifmcaddr(struct sk_buff *skb, struct ifmcaddr6 *ifmca,
3431 u32 pid, u32 seq, int event, u16 flags)
3433 struct nlmsghdr *nlh;
3434 u8 scope = RT_SCOPE_UNIVERSE;
3435 int ifindex = ifmca->idev->dev->ifindex;
3437 if (ipv6_addr_scope(&ifmca->mca_addr) & IFA_SITE)
3438 scope = RT_SCOPE_SITE;
3440 nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3441 if (nlh == NULL)
3442 return -EMSGSIZE;
3444 put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3445 if (nla_put(skb, IFA_MULTICAST, 16, &ifmca->mca_addr) < 0 ||
3446 put_cacheinfo(skb, ifmca->mca_cstamp, ifmca->mca_tstamp,
3447 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3448 nlmsg_cancel(skb, nlh);
3449 return -EMSGSIZE;
3452 return nlmsg_end(skb, nlh);
3455 static int inet6_fill_ifacaddr(struct sk_buff *skb, struct ifacaddr6 *ifaca,
3456 u32 pid, u32 seq, int event, unsigned int flags)
3458 struct nlmsghdr *nlh;
3459 u8 scope = RT_SCOPE_UNIVERSE;
3460 int ifindex = ifaca->aca_idev->dev->ifindex;
3462 if (ipv6_addr_scope(&ifaca->aca_addr) & IFA_SITE)
3463 scope = RT_SCOPE_SITE;
3465 nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3466 if (nlh == NULL)
3467 return -EMSGSIZE;
3469 put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3470 if (nla_put(skb, IFA_ANYCAST, 16, &ifaca->aca_addr) < 0 ||
3471 put_cacheinfo(skb, ifaca->aca_cstamp, ifaca->aca_tstamp,
3472 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3473 nlmsg_cancel(skb, nlh);
3474 return -EMSGSIZE;
3477 return nlmsg_end(skb, nlh);
3480 enum addr_type_t
3482 UNICAST_ADDR,
3483 MULTICAST_ADDR,
3484 ANYCAST_ADDR,
3487 /* called with rcu_read_lock() */
3488 static int in6_dump_addrs(struct inet6_dev *idev, struct sk_buff *skb,
3489 struct netlink_callback *cb, enum addr_type_t type,
3490 int s_ip_idx, int *p_ip_idx)
3492 struct inet6_ifaddr *ifa;
3493 struct ifmcaddr6 *ifmca;
3494 struct ifacaddr6 *ifaca;
3495 int err = 1;
3496 int ip_idx = *p_ip_idx;
3498 read_lock_bh(&idev->lock);
3499 switch (type) {
3500 case UNICAST_ADDR:
3501 /* unicast address incl. temp addr */
3502 for (ifa = idev->addr_list; ifa;
3503 ifa = ifa->if_next, ip_idx++) {
3504 if (ip_idx < s_ip_idx)
3505 continue;
3506 err = inet6_fill_ifaddr(skb, ifa,
3507 NETLINK_CB(cb->skb).pid,
3508 cb->nlh->nlmsg_seq,
3509 RTM_NEWADDR,
3510 NLM_F_MULTI);
3511 if (err <= 0)
3512 break;
3514 break;
3515 case MULTICAST_ADDR:
3516 /* multicast address */
3517 for (ifmca = idev->mc_list; ifmca;
3518 ifmca = ifmca->next, ip_idx++) {
3519 if (ip_idx < s_ip_idx)
3520 continue;
3521 err = inet6_fill_ifmcaddr(skb, ifmca,
3522 NETLINK_CB(cb->skb).pid,
3523 cb->nlh->nlmsg_seq,
3524 RTM_GETMULTICAST,
3525 NLM_F_MULTI);
3526 if (err <= 0)
3527 break;
3529 break;
3530 case ANYCAST_ADDR:
3531 /* anycast address */
3532 for (ifaca = idev->ac_list; ifaca;
3533 ifaca = ifaca->aca_next, ip_idx++) {
3534 if (ip_idx < s_ip_idx)
3535 continue;
3536 err = inet6_fill_ifacaddr(skb, ifaca,
3537 NETLINK_CB(cb->skb).pid,
3538 cb->nlh->nlmsg_seq,
3539 RTM_GETANYCAST,
3540 NLM_F_MULTI);
3541 if (err <= 0)
3542 break;
3544 break;
3545 default:
3546 break;
3548 read_unlock_bh(&idev->lock);
3549 *p_ip_idx = ip_idx;
3550 return err;
3553 static int inet6_dump_addr(struct sk_buff *skb, struct netlink_callback *cb,
3554 enum addr_type_t type)
3556 struct net *net = sock_net(skb->sk);
3557 int h, s_h;
3558 int idx, ip_idx;
3559 int s_idx, s_ip_idx;
3560 struct net_device *dev;
3561 struct inet6_dev *idev;
3562 struct hlist_head *head;
3563 struct hlist_node *node;
3565 s_h = cb->args[0];
3566 s_idx = idx = cb->args[1];
3567 s_ip_idx = ip_idx = cb->args[2];
3569 rcu_read_lock();
3570 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
3571 idx = 0;
3572 head = &net->dev_index_head[h];
3573 hlist_for_each_entry_rcu(dev, node, head, index_hlist) {
3574 if (idx < s_idx)
3575 goto cont;
3576 if (idx > s_idx)
3577 s_ip_idx = 0;
3578 ip_idx = 0;
3579 if ((idev = __in6_dev_get(dev)) == NULL)
3580 goto cont;
3582 if (in6_dump_addrs(idev, skb, cb, type,
3583 s_ip_idx, &ip_idx) <= 0)
3584 goto done;
3585 cont:
3586 idx++;
3589 done:
3590 rcu_read_unlock();
3591 cb->args[0] = h;
3592 cb->args[1] = idx;
3593 cb->args[2] = ip_idx;
3595 return skb->len;
3598 static int inet6_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
3600 enum addr_type_t type = UNICAST_ADDR;
3602 return inet6_dump_addr(skb, cb, type);
3605 static int inet6_dump_ifmcaddr(struct sk_buff *skb, struct netlink_callback *cb)
3607 enum addr_type_t type = MULTICAST_ADDR;
3609 return inet6_dump_addr(skb, cb, type);
3613 static int inet6_dump_ifacaddr(struct sk_buff *skb, struct netlink_callback *cb)
3615 enum addr_type_t type = ANYCAST_ADDR;
3617 return inet6_dump_addr(skb, cb, type);
3620 static int inet6_rtm_getaddr(struct sk_buff *in_skb, struct nlmsghdr* nlh,
3621 void *arg)
3623 struct net *net = sock_net(in_skb->sk);
3624 struct ifaddrmsg *ifm;
3625 struct nlattr *tb[IFA_MAX+1];
3626 struct in6_addr *addr = NULL;
3627 struct net_device *dev = NULL;
3628 struct inet6_ifaddr *ifa;
3629 struct sk_buff *skb;
3630 int err;
3632 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3633 if (err < 0)
3634 goto errout;
3636 addr = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3637 if (addr == NULL) {
3638 err = -EINVAL;
3639 goto errout;
3642 ifm = nlmsg_data(nlh);
3643 if (ifm->ifa_index)
3644 dev = __dev_get_by_index(net, ifm->ifa_index);
3646 if ((ifa = ipv6_get_ifaddr(net, addr, dev, 1)) == NULL) {
3647 err = -EADDRNOTAVAIL;
3648 goto errout;
3651 if ((skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_KERNEL)) == NULL) {
3652 err = -ENOBUFS;
3653 goto errout_ifa;
3656 err = inet6_fill_ifaddr(skb, ifa, NETLINK_CB(in_skb).pid,
3657 nlh->nlmsg_seq, RTM_NEWADDR, 0);
3658 if (err < 0) {
3659 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
3660 WARN_ON(err == -EMSGSIZE);
3661 kfree_skb(skb);
3662 goto errout_ifa;
3664 err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).pid);
3665 errout_ifa:
3666 in6_ifa_put(ifa);
3667 errout:
3668 return err;
3671 static void inet6_ifa_notify(int event, struct inet6_ifaddr *ifa)
3673 struct sk_buff *skb;
3674 struct net *net = dev_net(ifa->idev->dev);
3675 int err = -ENOBUFS;
3677 skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_ATOMIC);
3678 if (skb == NULL)
3679 goto errout;
3681 err = inet6_fill_ifaddr(skb, ifa, 0, 0, event, 0);
3682 if (err < 0) {
3683 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
3684 WARN_ON(err == -EMSGSIZE);
3685 kfree_skb(skb);
3686 goto errout;
3688 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
3689 return;
3690 errout:
3691 if (err < 0)
3692 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
3695 static inline void ipv6_store_devconf(struct ipv6_devconf *cnf,
3696 __s32 *array, int bytes)
3698 BUG_ON(bytes < (DEVCONF_MAX * 4));
3700 memset(array, 0, bytes);
3701 array[DEVCONF_FORWARDING] = cnf->forwarding;
3702 array[DEVCONF_HOPLIMIT] = cnf->hop_limit;
3703 array[DEVCONF_MTU6] = cnf->mtu6;
3704 array[DEVCONF_ACCEPT_RA] = cnf->accept_ra;
3705 array[DEVCONF_ACCEPT_REDIRECTS] = cnf->accept_redirects;
3706 array[DEVCONF_AUTOCONF] = cnf->autoconf;
3707 array[DEVCONF_DAD_TRANSMITS] = cnf->dad_transmits;
3708 array[DEVCONF_RTR_SOLICITS] = cnf->rtr_solicits;
3709 array[DEVCONF_RTR_SOLICIT_INTERVAL] = cnf->rtr_solicit_interval;
3710 array[DEVCONF_RTR_SOLICIT_DELAY] = cnf->rtr_solicit_delay;
3711 array[DEVCONF_FORCE_MLD_VERSION] = cnf->force_mld_version;
3712 #ifdef CONFIG_IPV6_PRIVACY
3713 array[DEVCONF_USE_TEMPADDR] = cnf->use_tempaddr;
3714 array[DEVCONF_TEMP_VALID_LFT] = cnf->temp_valid_lft;
3715 array[DEVCONF_TEMP_PREFERED_LFT] = cnf->temp_prefered_lft;
3716 array[DEVCONF_REGEN_MAX_RETRY] = cnf->regen_max_retry;
3717 array[DEVCONF_MAX_DESYNC_FACTOR] = cnf->max_desync_factor;
3718 #endif
3719 array[DEVCONF_MAX_ADDRESSES] = cnf->max_addresses;
3720 array[DEVCONF_ACCEPT_RA_DEFRTR] = cnf->accept_ra_defrtr;
3721 array[DEVCONF_ACCEPT_RA_PINFO] = cnf->accept_ra_pinfo;
3722 #ifdef CONFIG_IPV6_ROUTER_PREF
3723 array[DEVCONF_ACCEPT_RA_RTR_PREF] = cnf->accept_ra_rtr_pref;
3724 array[DEVCONF_RTR_PROBE_INTERVAL] = cnf->rtr_probe_interval;
3725 #ifdef CONFIG_IPV6_ROUTE_INFO
3726 array[DEVCONF_ACCEPT_RA_RT_INFO_MAX_PLEN] = cnf->accept_ra_rt_info_max_plen;
3727 #endif
3728 #endif
3729 array[DEVCONF_PROXY_NDP] = cnf->proxy_ndp;
3730 array[DEVCONF_ACCEPT_SOURCE_ROUTE] = cnf->accept_source_route;
3731 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
3732 array[DEVCONF_OPTIMISTIC_DAD] = cnf->optimistic_dad;
3733 #endif
3734 #ifdef CONFIG_IPV6_MROUTE
3735 array[DEVCONF_MC_FORWARDING] = cnf->mc_forwarding;
3736 #endif
3737 array[DEVCONF_DISABLE_IPV6] = cnf->disable_ipv6;
3738 array[DEVCONF_ACCEPT_DAD] = cnf->accept_dad;
3739 array[DEVCONF_FORCE_TLLAO] = cnf->force_tllao;
3742 static inline size_t inet6_if_nlmsg_size(void)
3744 return NLMSG_ALIGN(sizeof(struct ifinfomsg))
3745 + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
3746 + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
3747 + nla_total_size(4) /* IFLA_MTU */
3748 + nla_total_size(4) /* IFLA_LINK */
3749 + nla_total_size( /* IFLA_PROTINFO */
3750 nla_total_size(4) /* IFLA_INET6_FLAGS */
3751 + nla_total_size(sizeof(struct ifla_cacheinfo))
3752 + nla_total_size(DEVCONF_MAX * 4) /* IFLA_INET6_CONF */
3753 + nla_total_size(IPSTATS_MIB_MAX * 8) /* IFLA_INET6_STATS */
3754 + nla_total_size(ICMP6_MIB_MAX * 8) /* IFLA_INET6_ICMP6STATS */
3758 static inline void __snmp6_fill_stats(u64 *stats, void **mib, int items,
3759 int bytes)
3761 int i;
3762 int pad = bytes - sizeof(u64) * items;
3763 BUG_ON(pad < 0);
3765 /* Use put_unaligned() because stats may not be aligned for u64. */
3766 put_unaligned(items, &stats[0]);
3767 for (i = 1; i < items; i++)
3768 put_unaligned(snmp_fold_field(mib, i), &stats[i]);
3770 memset(&stats[items], 0, pad);
3773 static void snmp6_fill_stats(u64 *stats, struct inet6_dev *idev, int attrtype,
3774 int bytes)
3776 switch(attrtype) {
3777 case IFLA_INET6_STATS:
3778 __snmp6_fill_stats(stats, (void **)idev->stats.ipv6, IPSTATS_MIB_MAX, bytes);
3779 break;
3780 case IFLA_INET6_ICMP6STATS:
3781 __snmp6_fill_stats(stats, (void **)idev->stats.icmpv6, ICMP6_MIB_MAX, bytes);
3782 break;
3786 static int inet6_fill_ifinfo(struct sk_buff *skb, struct inet6_dev *idev,
3787 u32 pid, u32 seq, int event, unsigned int flags)
3789 struct net_device *dev = idev->dev;
3790 struct nlattr *nla;
3791 struct ifinfomsg *hdr;
3792 struct nlmsghdr *nlh;
3793 void *protoinfo;
3794 struct ifla_cacheinfo ci;
3796 nlh = nlmsg_put(skb, pid, seq, event, sizeof(*hdr), flags);
3797 if (nlh == NULL)
3798 return -EMSGSIZE;
3800 hdr = nlmsg_data(nlh);
3801 hdr->ifi_family = AF_INET6;
3802 hdr->__ifi_pad = 0;
3803 hdr->ifi_type = dev->type;
3804 hdr->ifi_index = dev->ifindex;
3805 hdr->ifi_flags = dev_get_flags(dev);
3806 hdr->ifi_change = 0;
3808 NLA_PUT_STRING(skb, IFLA_IFNAME, dev->name);
3810 if (dev->addr_len)
3811 NLA_PUT(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr);
3813 NLA_PUT_U32(skb, IFLA_MTU, dev->mtu);
3814 if (dev->ifindex != dev->iflink)
3815 NLA_PUT_U32(skb, IFLA_LINK, dev->iflink);
3817 protoinfo = nla_nest_start(skb, IFLA_PROTINFO);
3818 if (protoinfo == NULL)
3819 goto nla_put_failure;
3821 NLA_PUT_U32(skb, IFLA_INET6_FLAGS, idev->if_flags);
3823 ci.max_reasm_len = IPV6_MAXPLEN;
3824 ci.tstamp = (__u32)(TIME_DELTA(idev->tstamp, INITIAL_JIFFIES) / HZ * 100
3825 + TIME_DELTA(idev->tstamp, INITIAL_JIFFIES) % HZ * 100 / HZ);
3826 ci.reachable_time = idev->nd_parms->reachable_time;
3827 ci.retrans_time = idev->nd_parms->retrans_time;
3828 NLA_PUT(skb, IFLA_INET6_CACHEINFO, sizeof(ci), &ci);
3830 nla = nla_reserve(skb, IFLA_INET6_CONF, DEVCONF_MAX * sizeof(s32));
3831 if (nla == NULL)
3832 goto nla_put_failure;
3833 ipv6_store_devconf(&idev->cnf, nla_data(nla), nla_len(nla));
3835 /* XXX - MC not implemented */
3837 nla = nla_reserve(skb, IFLA_INET6_STATS, IPSTATS_MIB_MAX * sizeof(u64));
3838 if (nla == NULL)
3839 goto nla_put_failure;
3840 snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_STATS, nla_len(nla));
3842 nla = nla_reserve(skb, IFLA_INET6_ICMP6STATS, ICMP6_MIB_MAX * sizeof(u64));
3843 if (nla == NULL)
3844 goto nla_put_failure;
3845 snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_ICMP6STATS, nla_len(nla));
3847 nla_nest_end(skb, protoinfo);
3848 return nlmsg_end(skb, nlh);
3850 nla_put_failure:
3851 nlmsg_cancel(skb, nlh);
3852 return -EMSGSIZE;
3855 static int inet6_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
3857 struct net *net = sock_net(skb->sk);
3858 int h, s_h;
3859 int idx = 0, s_idx;
3860 struct net_device *dev;
3861 struct inet6_dev *idev;
3862 struct hlist_head *head;
3863 struct hlist_node *node;
3865 s_h = cb->args[0];
3866 s_idx = cb->args[1];
3868 rcu_read_lock();
3869 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
3870 idx = 0;
3871 head = &net->dev_index_head[h];
3872 hlist_for_each_entry_rcu(dev, node, head, index_hlist) {
3873 if (idx < s_idx)
3874 goto cont;
3875 idev = __in6_dev_get(dev);
3876 if (!idev)
3877 goto cont;
3878 if (inet6_fill_ifinfo(skb, idev,
3879 NETLINK_CB(cb->skb).pid,
3880 cb->nlh->nlmsg_seq,
3881 RTM_NEWLINK, NLM_F_MULTI) <= 0)
3882 goto out;
3883 cont:
3884 idx++;
3887 out:
3888 rcu_read_unlock();
3889 cb->args[1] = idx;
3890 cb->args[0] = h;
3892 return skb->len;
3895 void inet6_ifinfo_notify(int event, struct inet6_dev *idev)
3897 struct sk_buff *skb;
3898 struct net *net = dev_net(idev->dev);
3899 int err = -ENOBUFS;
3901 skb = nlmsg_new(inet6_if_nlmsg_size(), GFP_ATOMIC);
3902 if (skb == NULL)
3903 goto errout;
3905 err = inet6_fill_ifinfo(skb, idev, 0, 0, event, 0);
3906 if (err < 0) {
3907 /* -EMSGSIZE implies BUG in inet6_if_nlmsg_size() */
3908 WARN_ON(err == -EMSGSIZE);
3909 kfree_skb(skb);
3910 goto errout;
3912 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
3913 return;
3914 errout:
3915 if (err < 0)
3916 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
3919 static inline size_t inet6_prefix_nlmsg_size(void)
3921 return NLMSG_ALIGN(sizeof(struct prefixmsg))
3922 + nla_total_size(sizeof(struct in6_addr))
3923 + nla_total_size(sizeof(struct prefix_cacheinfo));
3926 static int inet6_fill_prefix(struct sk_buff *skb, struct inet6_dev *idev,
3927 struct prefix_info *pinfo, u32 pid, u32 seq,
3928 int event, unsigned int flags)
3930 struct prefixmsg *pmsg;
3931 struct nlmsghdr *nlh;
3932 struct prefix_cacheinfo ci;
3934 nlh = nlmsg_put(skb, pid, seq, event, sizeof(*pmsg), flags);
3935 if (nlh == NULL)
3936 return -EMSGSIZE;
3938 pmsg = nlmsg_data(nlh);
3939 pmsg->prefix_family = AF_INET6;
3940 pmsg->prefix_pad1 = 0;
3941 pmsg->prefix_pad2 = 0;
3942 pmsg->prefix_ifindex = idev->dev->ifindex;
3943 pmsg->prefix_len = pinfo->prefix_len;
3944 pmsg->prefix_type = pinfo->type;
3945 pmsg->prefix_pad3 = 0;
3946 pmsg->prefix_flags = 0;
3947 if (pinfo->onlink)
3948 pmsg->prefix_flags |= IF_PREFIX_ONLINK;
3949 if (pinfo->autoconf)
3950 pmsg->prefix_flags |= IF_PREFIX_AUTOCONF;
3952 NLA_PUT(skb, PREFIX_ADDRESS, sizeof(pinfo->prefix), &pinfo->prefix);
3954 ci.preferred_time = ntohl(pinfo->prefered);
3955 ci.valid_time = ntohl(pinfo->valid);
3956 NLA_PUT(skb, PREFIX_CACHEINFO, sizeof(ci), &ci);
3958 return nlmsg_end(skb, nlh);
3960 nla_put_failure:
3961 nlmsg_cancel(skb, nlh);
3962 return -EMSGSIZE;
3965 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
3966 struct prefix_info *pinfo)
3968 struct sk_buff *skb;
3969 struct net *net = dev_net(idev->dev);
3970 int err = -ENOBUFS;
3972 skb = nlmsg_new(inet6_prefix_nlmsg_size(), GFP_ATOMIC);
3973 if (skb == NULL)
3974 goto errout;
3976 err = inet6_fill_prefix(skb, idev, pinfo, 0, 0, event, 0);
3977 if (err < 0) {
3978 /* -EMSGSIZE implies BUG in inet6_prefix_nlmsg_size() */
3979 WARN_ON(err == -EMSGSIZE);
3980 kfree_skb(skb);
3981 goto errout;
3983 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_PREFIX, NULL, GFP_ATOMIC);
3984 return;
3985 errout:
3986 if (err < 0)
3987 rtnl_set_sk_err(net, RTNLGRP_IPV6_PREFIX, err);
3990 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
3992 inet6_ifa_notify(event ? : RTM_NEWADDR, ifp);
3994 switch (event) {
3995 case RTM_NEWADDR:
3997 * If the address was optimistic
3998 * we inserted the route at the start of
3999 * our DAD process, so we don't need
4000 * to do it again
4002 if (!(ifp->rt->rt6i_node))
4003 ip6_ins_rt(ifp->rt);
4004 if (ifp->idev->cnf.forwarding)
4005 addrconf_join_anycast(ifp);
4006 break;
4007 case RTM_DELADDR:
4008 if (ifp->idev->cnf.forwarding)
4009 addrconf_leave_anycast(ifp);
4010 addrconf_leave_solict(ifp->idev, &ifp->addr);
4011 dst_hold(&ifp->rt->u.dst);
4012 if (ip6_del_rt(ifp->rt))
4013 dst_free(&ifp->rt->u.dst);
4014 break;
4018 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
4020 rcu_read_lock_bh();
4021 if (likely(ifp->idev->dead == 0))
4022 __ipv6_ifa_notify(event, ifp);
4023 rcu_read_unlock_bh();
4026 #ifdef CONFIG_SYSCTL
4028 static
4029 int addrconf_sysctl_forward(ctl_table *ctl, int write,
4030 void __user *buffer, size_t *lenp, loff_t *ppos)
4032 int *valp = ctl->data;
4033 int val = *valp;
4034 loff_t pos = *ppos;
4035 int ret;
4037 ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
4039 if (write)
4040 ret = addrconf_fixup_forwarding(ctl, valp, val);
4041 if (ret)
4042 *ppos = pos;
4043 return ret;
4046 static void dev_disable_change(struct inet6_dev *idev)
4048 if (!idev || !idev->dev)
4049 return;
4051 if (idev->cnf.disable_ipv6)
4052 addrconf_notify(NULL, NETDEV_DOWN, idev->dev);
4053 else
4054 addrconf_notify(NULL, NETDEV_UP, idev->dev);
4057 static void addrconf_disable_change(struct net *net, __s32 newf)
4059 struct net_device *dev;
4060 struct inet6_dev *idev;
4062 rcu_read_lock();
4063 for_each_netdev_rcu(net, dev) {
4064 idev = __in6_dev_get(dev);
4065 if (idev) {
4066 int changed = (!idev->cnf.disable_ipv6) ^ (!newf);
4067 idev->cnf.disable_ipv6 = newf;
4068 if (changed)
4069 dev_disable_change(idev);
4072 rcu_read_unlock();
4075 static int addrconf_disable_ipv6(struct ctl_table *table, int *p, int old)
4077 struct net *net;
4079 net = (struct net *)table->extra2;
4081 if (p == &net->ipv6.devconf_dflt->disable_ipv6)
4082 return 0;
4084 if (!rtnl_trylock()) {
4085 /* Restore the original values before restarting */
4086 *p = old;
4087 return restart_syscall();
4090 if (p == &net->ipv6.devconf_all->disable_ipv6) {
4091 __s32 newf = net->ipv6.devconf_all->disable_ipv6;
4092 net->ipv6.devconf_dflt->disable_ipv6 = newf;
4093 addrconf_disable_change(net, newf);
4094 } else if ((!*p) ^ (!old))
4095 dev_disable_change((struct inet6_dev *)table->extra1);
4097 rtnl_unlock();
4098 return 0;
4101 static
4102 int addrconf_sysctl_disable(ctl_table *ctl, int write,
4103 void __user *buffer, size_t *lenp, loff_t *ppos)
4105 int *valp = ctl->data;
4106 int val = *valp;
4107 loff_t pos = *ppos;
4108 int ret;
4110 ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
4112 if (write)
4113 ret = addrconf_disable_ipv6(ctl, valp, val);
4114 if (ret)
4115 *ppos = pos;
4116 return ret;
4119 static struct addrconf_sysctl_table
4121 struct ctl_table_header *sysctl_header;
4122 ctl_table addrconf_vars[DEVCONF_MAX+1];
4123 char *dev_name;
4124 } addrconf_sysctl __read_mostly = {
4125 .sysctl_header = NULL,
4126 .addrconf_vars = {
4128 .procname = "forwarding",
4129 .data = &ipv6_devconf.forwarding,
4130 .maxlen = sizeof(int),
4131 .mode = 0644,
4132 .proc_handler = addrconf_sysctl_forward,
4135 .procname = "hop_limit",
4136 .data = &ipv6_devconf.hop_limit,
4137 .maxlen = sizeof(int),
4138 .mode = 0644,
4139 .proc_handler = proc_dointvec,
4142 .procname = "mtu",
4143 .data = &ipv6_devconf.mtu6,
4144 .maxlen = sizeof(int),
4145 .mode = 0644,
4146 .proc_handler = proc_dointvec,
4149 .procname = "accept_ra",
4150 .data = &ipv6_devconf.accept_ra,
4151 .maxlen = sizeof(int),
4152 .mode = 0644,
4153 .proc_handler = proc_dointvec,
4156 .procname = "accept_redirects",
4157 .data = &ipv6_devconf.accept_redirects,
4158 .maxlen = sizeof(int),
4159 .mode = 0644,
4160 .proc_handler = proc_dointvec,
4163 .procname = "autoconf",
4164 .data = &ipv6_devconf.autoconf,
4165 .maxlen = sizeof(int),
4166 .mode = 0644,
4167 .proc_handler = proc_dointvec,
4170 .procname = "dad_transmits",
4171 .data = &ipv6_devconf.dad_transmits,
4172 .maxlen = sizeof(int),
4173 .mode = 0644,
4174 .proc_handler = proc_dointvec,
4177 .procname = "router_solicitations",
4178 .data = &ipv6_devconf.rtr_solicits,
4179 .maxlen = sizeof(int),
4180 .mode = 0644,
4181 .proc_handler = proc_dointvec,
4184 .procname = "router_solicitation_interval",
4185 .data = &ipv6_devconf.rtr_solicit_interval,
4186 .maxlen = sizeof(int),
4187 .mode = 0644,
4188 .proc_handler = proc_dointvec_jiffies,
4191 .procname = "router_solicitation_delay",
4192 .data = &ipv6_devconf.rtr_solicit_delay,
4193 .maxlen = sizeof(int),
4194 .mode = 0644,
4195 .proc_handler = proc_dointvec_jiffies,
4198 .procname = "force_mld_version",
4199 .data = &ipv6_devconf.force_mld_version,
4200 .maxlen = sizeof(int),
4201 .mode = 0644,
4202 .proc_handler = proc_dointvec,
4204 #ifdef CONFIG_IPV6_PRIVACY
4206 .procname = "use_tempaddr",
4207 .data = &ipv6_devconf.use_tempaddr,
4208 .maxlen = sizeof(int),
4209 .mode = 0644,
4210 .proc_handler = proc_dointvec,
4213 .procname = "temp_valid_lft",
4214 .data = &ipv6_devconf.temp_valid_lft,
4215 .maxlen = sizeof(int),
4216 .mode = 0644,
4217 .proc_handler = proc_dointvec,
4220 .procname = "temp_prefered_lft",
4221 .data = &ipv6_devconf.temp_prefered_lft,
4222 .maxlen = sizeof(int),
4223 .mode = 0644,
4224 .proc_handler = proc_dointvec,
4227 .procname = "regen_max_retry",
4228 .data = &ipv6_devconf.regen_max_retry,
4229 .maxlen = sizeof(int),
4230 .mode = 0644,
4231 .proc_handler = proc_dointvec,
4234 .procname = "max_desync_factor",
4235 .data = &ipv6_devconf.max_desync_factor,
4236 .maxlen = sizeof(int),
4237 .mode = 0644,
4238 .proc_handler = proc_dointvec,
4240 #endif
4242 .procname = "max_addresses",
4243 .data = &ipv6_devconf.max_addresses,
4244 .maxlen = sizeof(int),
4245 .mode = 0644,
4246 .proc_handler = proc_dointvec,
4249 .procname = "accept_ra_defrtr",
4250 .data = &ipv6_devconf.accept_ra_defrtr,
4251 .maxlen = sizeof(int),
4252 .mode = 0644,
4253 .proc_handler = proc_dointvec,
4256 .procname = "accept_ra_pinfo",
4257 .data = &ipv6_devconf.accept_ra_pinfo,
4258 .maxlen = sizeof(int),
4259 .mode = 0644,
4260 .proc_handler = proc_dointvec,
4262 #ifdef CONFIG_IPV6_ROUTER_PREF
4264 .procname = "accept_ra_rtr_pref",
4265 .data = &ipv6_devconf.accept_ra_rtr_pref,
4266 .maxlen = sizeof(int),
4267 .mode = 0644,
4268 .proc_handler = proc_dointvec,
4271 .procname = "router_probe_interval",
4272 .data = &ipv6_devconf.rtr_probe_interval,
4273 .maxlen = sizeof(int),
4274 .mode = 0644,
4275 .proc_handler = proc_dointvec_jiffies,
4277 #ifdef CONFIG_IPV6_ROUTE_INFO
4279 .procname = "accept_ra_rt_info_max_plen",
4280 .data = &ipv6_devconf.accept_ra_rt_info_max_plen,
4281 .maxlen = sizeof(int),
4282 .mode = 0644,
4283 .proc_handler = proc_dointvec,
4285 #endif
4286 #endif
4288 .procname = "proxy_ndp",
4289 .data = &ipv6_devconf.proxy_ndp,
4290 .maxlen = sizeof(int),
4291 .mode = 0644,
4292 .proc_handler = proc_dointvec,
4295 .procname = "accept_source_route",
4296 .data = &ipv6_devconf.accept_source_route,
4297 .maxlen = sizeof(int),
4298 .mode = 0644,
4299 .proc_handler = proc_dointvec,
4301 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
4303 .procname = "optimistic_dad",
4304 .data = &ipv6_devconf.optimistic_dad,
4305 .maxlen = sizeof(int),
4306 .mode = 0644,
4307 .proc_handler = proc_dointvec,
4310 #endif
4311 #ifdef CONFIG_IPV6_MROUTE
4313 .procname = "mc_forwarding",
4314 .data = &ipv6_devconf.mc_forwarding,
4315 .maxlen = sizeof(int),
4316 .mode = 0444,
4317 .proc_handler = proc_dointvec,
4319 #endif
4321 .procname = "disable_ipv6",
4322 .data = &ipv6_devconf.disable_ipv6,
4323 .maxlen = sizeof(int),
4324 .mode = 0644,
4325 .proc_handler = addrconf_sysctl_disable,
4328 .procname = "accept_dad",
4329 .data = &ipv6_devconf.accept_dad,
4330 .maxlen = sizeof(int),
4331 .mode = 0644,
4332 .proc_handler = proc_dointvec,
4335 .procname = "force_tllao",
4336 .data = &ipv6_devconf.force_tllao,
4337 .maxlen = sizeof(int),
4338 .mode = 0644,
4339 .proc_handler = proc_dointvec
4342 /* sentinel */
4347 static int __addrconf_sysctl_register(struct net *net, char *dev_name,
4348 struct inet6_dev *idev, struct ipv6_devconf *p)
4350 int i;
4351 struct addrconf_sysctl_table *t;
4353 #define ADDRCONF_CTL_PATH_DEV 3
4355 struct ctl_path addrconf_ctl_path[] = {
4356 { .procname = "net", },
4357 { .procname = "ipv6", },
4358 { .procname = "conf", },
4359 { /* to be set */ },
4360 { },
4364 t = kmemdup(&addrconf_sysctl, sizeof(*t), GFP_KERNEL);
4365 if (t == NULL)
4366 goto out;
4368 for (i=0; t->addrconf_vars[i].data; i++) {
4369 t->addrconf_vars[i].data += (char*)p - (char*)&ipv6_devconf;
4370 t->addrconf_vars[i].extra1 = idev; /* embedded; no ref */
4371 t->addrconf_vars[i].extra2 = net;
4375 * Make a copy of dev_name, because '.procname' is regarded as const
4376 * by sysctl and we wouldn't want anyone to change it under our feet
4377 * (see SIOCSIFNAME).
4379 t->dev_name = kstrdup(dev_name, GFP_KERNEL);
4380 if (!t->dev_name)
4381 goto free;
4383 addrconf_ctl_path[ADDRCONF_CTL_PATH_DEV].procname = t->dev_name;
4385 t->sysctl_header = register_net_sysctl_table(net, addrconf_ctl_path,
4386 t->addrconf_vars);
4387 if (t->sysctl_header == NULL)
4388 goto free_procname;
4390 p->sysctl = t;
4391 return 0;
4393 free_procname:
4394 kfree(t->dev_name);
4395 free:
4396 kfree(t);
4397 out:
4398 return -ENOBUFS;
4401 static void __addrconf_sysctl_unregister(struct ipv6_devconf *p)
4403 struct addrconf_sysctl_table *t;
4405 if (p->sysctl == NULL)
4406 return;
4408 t = p->sysctl;
4409 p->sysctl = NULL;
4410 unregister_sysctl_table(t->sysctl_header);
4411 kfree(t->dev_name);
4412 kfree(t);
4415 static void addrconf_sysctl_register(struct inet6_dev *idev)
4417 neigh_sysctl_register(idev->dev, idev->nd_parms, NET_IPV6,
4418 NET_IPV6_NEIGH, "ipv6",
4419 &ndisc_ifinfo_sysctl_change);
4420 __addrconf_sysctl_register(dev_net(idev->dev), idev->dev->name,
4421 idev, &idev->cnf);
4424 static void addrconf_sysctl_unregister(struct inet6_dev *idev)
4426 __addrconf_sysctl_unregister(&idev->cnf);
4427 neigh_sysctl_unregister(idev->nd_parms);
4431 #endif
4433 static int addrconf_init_net(struct net *net)
4435 int err;
4436 struct ipv6_devconf *all, *dflt;
4438 err = -ENOMEM;
4439 all = &ipv6_devconf;
4440 dflt = &ipv6_devconf_dflt;
4442 if (!net_eq(net, &init_net)) {
4443 all = kmemdup(all, sizeof(ipv6_devconf), GFP_KERNEL);
4444 if (all == NULL)
4445 goto err_alloc_all;
4447 dflt = kmemdup(dflt, sizeof(ipv6_devconf_dflt), GFP_KERNEL);
4448 if (dflt == NULL)
4449 goto err_alloc_dflt;
4450 } else {
4451 /* these will be inherited by all namespaces */
4452 dflt->autoconf = ipv6_defaults.autoconf;
4453 dflt->disable_ipv6 = ipv6_defaults.disable_ipv6;
4456 net->ipv6.devconf_all = all;
4457 net->ipv6.devconf_dflt = dflt;
4459 #ifdef CONFIG_SYSCTL
4460 err = __addrconf_sysctl_register(net, "all", NULL, all);
4461 if (err < 0)
4462 goto err_reg_all;
4464 err = __addrconf_sysctl_register(net, "default", NULL, dflt);
4465 if (err < 0)
4466 goto err_reg_dflt;
4467 #endif
4468 return 0;
4470 #ifdef CONFIG_SYSCTL
4471 err_reg_dflt:
4472 __addrconf_sysctl_unregister(all);
4473 err_reg_all:
4474 kfree(dflt);
4475 #endif
4476 err_alloc_dflt:
4477 kfree(all);
4478 err_alloc_all:
4479 return err;
4482 static void addrconf_exit_net(struct net *net)
4484 #ifdef CONFIG_SYSCTL
4485 __addrconf_sysctl_unregister(net->ipv6.devconf_dflt);
4486 __addrconf_sysctl_unregister(net->ipv6.devconf_all);
4487 #endif
4488 if (!net_eq(net, &init_net)) {
4489 kfree(net->ipv6.devconf_dflt);
4490 kfree(net->ipv6.devconf_all);
4494 static struct pernet_operations addrconf_ops = {
4495 .init = addrconf_init_net,
4496 .exit = addrconf_exit_net,
4500 * Device notifier
4503 int register_inet6addr_notifier(struct notifier_block *nb)
4505 return atomic_notifier_chain_register(&inet6addr_chain, nb);
4508 EXPORT_SYMBOL(register_inet6addr_notifier);
4510 int unregister_inet6addr_notifier(struct notifier_block *nb)
4512 return atomic_notifier_chain_unregister(&inet6addr_chain,nb);
4515 EXPORT_SYMBOL(unregister_inet6addr_notifier);
4518 * Init / cleanup code
4521 int __init addrconf_init(void)
4523 int err;
4525 if ((err = ipv6_addr_label_init()) < 0) {
4526 printk(KERN_CRIT "IPv6 Addrconf: cannot initialize default policy table: %d.\n",
4527 err);
4528 return err;
4531 register_pernet_subsys(&addrconf_ops);
4533 /* The addrconf netdev notifier requires that loopback_dev
4534 * has it's ipv6 private information allocated and setup
4535 * before it can bring up and give link-local addresses
4536 * to other devices which are up.
4538 * Unfortunately, loopback_dev is not necessarily the first
4539 * entry in the global dev_base list of net devices. In fact,
4540 * it is likely to be the very last entry on that list.
4541 * So this causes the notifier registry below to try and
4542 * give link-local addresses to all devices besides loopback_dev
4543 * first, then loopback_dev, which cases all the non-loopback_dev
4544 * devices to fail to get a link-local address.
4546 * So, as a temporary fix, allocate the ipv6 structure for
4547 * loopback_dev first by hand.
4548 * Longer term, all of the dependencies ipv6 has upon the loopback
4549 * device and it being up should be removed.
4551 rtnl_lock();
4552 if (!ipv6_add_dev(init_net.loopback_dev))
4553 err = -ENOMEM;
4554 rtnl_unlock();
4555 if (err)
4556 goto errlo;
4558 register_netdevice_notifier(&ipv6_dev_notf);
4560 addrconf_verify(0);
4562 err = __rtnl_register(PF_INET6, RTM_GETLINK, NULL, inet6_dump_ifinfo);
4563 if (err < 0)
4564 goto errout;
4566 /* Only the first call to __rtnl_register can fail */
4567 __rtnl_register(PF_INET6, RTM_NEWADDR, inet6_rtm_newaddr, NULL);
4568 __rtnl_register(PF_INET6, RTM_DELADDR, inet6_rtm_deladdr, NULL);
4569 __rtnl_register(PF_INET6, RTM_GETADDR, inet6_rtm_getaddr, inet6_dump_ifaddr);
4570 __rtnl_register(PF_INET6, RTM_GETMULTICAST, NULL, inet6_dump_ifmcaddr);
4571 __rtnl_register(PF_INET6, RTM_GETANYCAST, NULL, inet6_dump_ifacaddr);
4573 ipv6_addr_label_rtnl_register();
4575 return 0;
4576 errout:
4577 unregister_netdevice_notifier(&ipv6_dev_notf);
4578 errlo:
4579 unregister_pernet_subsys(&addrconf_ops);
4581 return err;
4584 void addrconf_cleanup(void)
4586 struct inet6_ifaddr *ifa;
4587 struct net_device *dev;
4588 int i;
4590 unregister_netdevice_notifier(&ipv6_dev_notf);
4591 unregister_pernet_subsys(&addrconf_ops);
4593 rtnl_lock();
4595 /* clean dev list */
4596 for_each_netdev(&init_net, dev) {
4597 if (__in6_dev_get(dev) == NULL)
4598 continue;
4599 addrconf_ifdown(dev, 1);
4601 addrconf_ifdown(init_net.loopback_dev, 2);
4604 * Check hash table.
4606 write_lock_bh(&addrconf_hash_lock);
4607 for (i=0; i < IN6_ADDR_HSIZE; i++) {
4608 for (ifa=inet6_addr_lst[i]; ifa; ) {
4609 struct inet6_ifaddr *bifa;
4611 bifa = ifa;
4612 ifa = ifa->lst_next;
4613 printk(KERN_DEBUG "bug: IPv6 address leakage detected: ifa=%p\n", bifa);
4614 /* Do not free it; something is wrong.
4615 Now we can investigate it with debugger.
4619 write_unlock_bh(&addrconf_hash_lock);
4621 del_timer(&addr_chk_timer);
4622 rtnl_unlock();