be2net: Add be_rx_polls counter
[linux-2.6/linux-acpi-2.6/ibm-acpi-2.6.git] / net / ipv6 / addrconf.c
blobb1ce8fc62049e9bf4c6b97294930c012f0585fb1
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 return restart_syscall();
508 if (p == &net->ipv6.devconf_all->forwarding) {
509 __s32 newf = net->ipv6.devconf_all->forwarding;
510 net->ipv6.devconf_dflt->forwarding = newf;
511 addrconf_forward_change(net, newf);
512 } else if ((!*p) ^ (!old))
513 dev_forward_change((struct inet6_dev *)table->extra1);
514 rtnl_unlock();
516 if (*p)
517 rt6_purge_dflt_routers(net);
518 return 1;
520 #endif
522 /* Nobody refers to this ifaddr, destroy it */
524 void inet6_ifa_finish_destroy(struct inet6_ifaddr *ifp)
526 WARN_ON(ifp->if_next != NULL);
527 WARN_ON(ifp->lst_next != NULL);
529 #ifdef NET_REFCNT_DEBUG
530 printk(KERN_DEBUG "inet6_ifa_finish_destroy\n");
531 #endif
533 in6_dev_put(ifp->idev);
535 if (del_timer(&ifp->timer))
536 printk("Timer is still running, when freeing ifa=%p\n", ifp);
538 if (!ifp->dead) {
539 printk("Freeing alive inet6 address %p\n", ifp);
540 return;
542 dst_release(&ifp->rt->u.dst);
544 kfree(ifp);
547 static void
548 ipv6_link_dev_addr(struct inet6_dev *idev, struct inet6_ifaddr *ifp)
550 struct inet6_ifaddr *ifa, **ifap;
551 int ifp_scope = ipv6_addr_src_scope(&ifp->addr);
554 * Each device address list is sorted in order of scope -
555 * global before linklocal.
557 for (ifap = &idev->addr_list; (ifa = *ifap) != NULL;
558 ifap = &ifa->if_next) {
559 if (ifp_scope >= ipv6_addr_src_scope(&ifa->addr))
560 break;
563 ifp->if_next = *ifap;
564 *ifap = ifp;
568 * Hash function taken from net_alias.c
570 static u8 ipv6_addr_hash(const struct in6_addr *addr)
572 __u32 word;
575 * We perform the hash function over the last 64 bits of the address
576 * This will include the IEEE address token on links that support it.
579 word = (__force u32)(addr->s6_addr32[2] ^ addr->s6_addr32[3]);
580 word ^= (word >> 16);
581 word ^= (word >> 8);
583 return ((word ^ (word >> 4)) & 0x0f);
586 /* On success it returns ifp with increased reference count */
588 static struct inet6_ifaddr *
589 ipv6_add_addr(struct inet6_dev *idev, const struct in6_addr *addr, int pfxlen,
590 int scope, u32 flags)
592 struct inet6_ifaddr *ifa = NULL;
593 struct rt6_info *rt;
594 int hash;
595 int err = 0;
596 int addr_type = ipv6_addr_type(addr);
598 if (addr_type == IPV6_ADDR_ANY ||
599 addr_type & IPV6_ADDR_MULTICAST ||
600 (!(idev->dev->flags & IFF_LOOPBACK) &&
601 addr_type & IPV6_ADDR_LOOPBACK))
602 return ERR_PTR(-EADDRNOTAVAIL);
604 rcu_read_lock_bh();
605 if (idev->dead) {
606 err = -ENODEV; /*XXX*/
607 goto out2;
610 if (idev->cnf.disable_ipv6) {
611 err = -EACCES;
612 goto out2;
615 write_lock(&addrconf_hash_lock);
617 /* Ignore adding duplicate addresses on an interface */
618 if (ipv6_chk_same_addr(dev_net(idev->dev), addr, idev->dev)) {
619 ADBG(("ipv6_add_addr: already assigned\n"));
620 err = -EEXIST;
621 goto out;
624 ifa = kzalloc(sizeof(struct inet6_ifaddr), GFP_ATOMIC);
626 if (ifa == NULL) {
627 ADBG(("ipv6_add_addr: malloc failed\n"));
628 err = -ENOBUFS;
629 goto out;
632 rt = addrconf_dst_alloc(idev, addr, 0);
633 if (IS_ERR(rt)) {
634 err = PTR_ERR(rt);
635 goto out;
638 ipv6_addr_copy(&ifa->addr, addr);
640 spin_lock_init(&ifa->lock);
641 init_timer(&ifa->timer);
642 ifa->timer.data = (unsigned long) ifa;
643 ifa->scope = scope;
644 ifa->prefix_len = pfxlen;
645 ifa->flags = flags | IFA_F_TENTATIVE;
646 ifa->cstamp = ifa->tstamp = jiffies;
648 ifa->rt = rt;
651 * part one of RFC 4429, section 3.3
652 * We should not configure an address as
653 * optimistic if we do not yet know the link
654 * layer address of our nexhop router
657 if (rt->rt6i_nexthop == NULL)
658 ifa->flags &= ~IFA_F_OPTIMISTIC;
660 ifa->idev = idev;
661 in6_dev_hold(idev);
662 /* For caller */
663 in6_ifa_hold(ifa);
665 /* Add to big hash table */
666 hash = ipv6_addr_hash(addr);
668 ifa->lst_next = inet6_addr_lst[hash];
669 inet6_addr_lst[hash] = ifa;
670 in6_ifa_hold(ifa);
671 write_unlock(&addrconf_hash_lock);
673 write_lock(&idev->lock);
674 /* Add to inet6_dev unicast addr list. */
675 ipv6_link_dev_addr(idev, ifa);
677 #ifdef CONFIG_IPV6_PRIVACY
678 if (ifa->flags&IFA_F_TEMPORARY) {
679 ifa->tmp_next = idev->tempaddr_list;
680 idev->tempaddr_list = ifa;
681 in6_ifa_hold(ifa);
683 #endif
685 in6_ifa_hold(ifa);
686 write_unlock(&idev->lock);
687 out2:
688 rcu_read_unlock_bh();
690 if (likely(err == 0))
691 atomic_notifier_call_chain(&inet6addr_chain, NETDEV_UP, ifa);
692 else {
693 kfree(ifa);
694 ifa = ERR_PTR(err);
697 return ifa;
698 out:
699 write_unlock(&addrconf_hash_lock);
700 goto out2;
703 /* This function wants to get referenced ifp and releases it before return */
705 static void ipv6_del_addr(struct inet6_ifaddr *ifp)
707 struct inet6_ifaddr *ifa, **ifap;
708 struct inet6_dev *idev = ifp->idev;
709 int hash;
710 int deleted = 0, onlink = 0;
711 unsigned long expires = jiffies;
713 hash = ipv6_addr_hash(&ifp->addr);
715 ifp->dead = 1;
717 write_lock_bh(&addrconf_hash_lock);
718 for (ifap = &inet6_addr_lst[hash]; (ifa=*ifap) != NULL;
719 ifap = &ifa->lst_next) {
720 if (ifa == ifp) {
721 *ifap = ifa->lst_next;
722 __in6_ifa_put(ifp);
723 ifa->lst_next = NULL;
724 break;
727 write_unlock_bh(&addrconf_hash_lock);
729 write_lock_bh(&idev->lock);
730 #ifdef CONFIG_IPV6_PRIVACY
731 if (ifp->flags&IFA_F_TEMPORARY) {
732 for (ifap = &idev->tempaddr_list; (ifa=*ifap) != NULL;
733 ifap = &ifa->tmp_next) {
734 if (ifa == ifp) {
735 *ifap = ifa->tmp_next;
736 if (ifp->ifpub) {
737 in6_ifa_put(ifp->ifpub);
738 ifp->ifpub = NULL;
740 __in6_ifa_put(ifp);
741 ifa->tmp_next = NULL;
742 break;
746 #endif
748 for (ifap = &idev->addr_list; (ifa=*ifap) != NULL;) {
749 if (ifa == ifp) {
750 *ifap = ifa->if_next;
751 __in6_ifa_put(ifp);
752 ifa->if_next = NULL;
753 if (!(ifp->flags & IFA_F_PERMANENT) || onlink > 0)
754 break;
755 deleted = 1;
756 continue;
757 } else if (ifp->flags & IFA_F_PERMANENT) {
758 if (ipv6_prefix_equal(&ifa->addr, &ifp->addr,
759 ifp->prefix_len)) {
760 if (ifa->flags & IFA_F_PERMANENT) {
761 onlink = 1;
762 if (deleted)
763 break;
764 } else {
765 unsigned long lifetime;
767 if (!onlink)
768 onlink = -1;
770 spin_lock(&ifa->lock);
772 lifetime = addrconf_timeout_fixup(ifa->valid_lft, HZ);
774 * Note: Because this address is
775 * not permanent, lifetime <
776 * LONG_MAX / HZ here.
778 if (time_before(expires,
779 ifa->tstamp + lifetime * HZ))
780 expires = ifa->tstamp + lifetime * HZ;
781 spin_unlock(&ifa->lock);
785 ifap = &ifa->if_next;
787 write_unlock_bh(&idev->lock);
789 addrconf_del_timer(ifp);
791 ipv6_ifa_notify(RTM_DELADDR, ifp);
793 atomic_notifier_call_chain(&inet6addr_chain, NETDEV_DOWN, ifp);
796 * Purge or update corresponding prefix
798 * 1) we don't purge prefix here if address was not permanent.
799 * prefix is managed by its own lifetime.
800 * 2) if there're no addresses, delete prefix.
801 * 3) if there're still other permanent address(es),
802 * corresponding prefix is still permanent.
803 * 4) otherwise, update prefix lifetime to the
804 * longest valid lifetime among the corresponding
805 * addresses on the device.
806 * Note: subsequent RA will update lifetime.
808 * --yoshfuji
810 if ((ifp->flags & IFA_F_PERMANENT) && onlink < 1) {
811 struct in6_addr prefix;
812 struct rt6_info *rt;
813 struct net *net = dev_net(ifp->idev->dev);
814 ipv6_addr_prefix(&prefix, &ifp->addr, ifp->prefix_len);
815 rt = rt6_lookup(net, &prefix, NULL, ifp->idev->dev->ifindex, 1);
817 if (rt && addrconf_is_prefix_route(rt)) {
818 if (onlink == 0) {
819 ip6_del_rt(rt);
820 rt = NULL;
821 } else if (!(rt->rt6i_flags & RTF_EXPIRES)) {
822 rt->rt6i_expires = expires;
823 rt->rt6i_flags |= RTF_EXPIRES;
826 dst_release(&rt->u.dst);
829 in6_ifa_put(ifp);
832 #ifdef CONFIG_IPV6_PRIVACY
833 static int ipv6_create_tempaddr(struct inet6_ifaddr *ifp, struct inet6_ifaddr *ift)
835 struct inet6_dev *idev = ifp->idev;
836 struct in6_addr addr, *tmpaddr;
837 unsigned long tmp_prefered_lft, tmp_valid_lft, tmp_cstamp, tmp_tstamp;
838 unsigned long regen_advance;
839 int tmp_plen;
840 int ret = 0;
841 int max_addresses;
842 u32 addr_flags;
844 write_lock(&idev->lock);
845 if (ift) {
846 spin_lock_bh(&ift->lock);
847 memcpy(&addr.s6_addr[8], &ift->addr.s6_addr[8], 8);
848 spin_unlock_bh(&ift->lock);
849 tmpaddr = &addr;
850 } else {
851 tmpaddr = NULL;
853 retry:
854 in6_dev_hold(idev);
855 if (idev->cnf.use_tempaddr <= 0) {
856 write_unlock(&idev->lock);
857 printk(KERN_INFO
858 "ipv6_create_tempaddr(): use_tempaddr is disabled.\n");
859 in6_dev_put(idev);
860 ret = -1;
861 goto out;
863 spin_lock_bh(&ifp->lock);
864 if (ifp->regen_count++ >= idev->cnf.regen_max_retry) {
865 idev->cnf.use_tempaddr = -1; /*XXX*/
866 spin_unlock_bh(&ifp->lock);
867 write_unlock(&idev->lock);
868 printk(KERN_WARNING
869 "ipv6_create_tempaddr(): regeneration time exceeded. disabled temporary address support.\n");
870 in6_dev_put(idev);
871 ret = -1;
872 goto out;
874 in6_ifa_hold(ifp);
875 memcpy(addr.s6_addr, ifp->addr.s6_addr, 8);
876 if (__ipv6_try_regen_rndid(idev, tmpaddr) < 0) {
877 spin_unlock_bh(&ifp->lock);
878 write_unlock(&idev->lock);
879 printk(KERN_WARNING
880 "ipv6_create_tempaddr(): regeneration of randomized interface id failed.\n");
881 in6_ifa_put(ifp);
882 in6_dev_put(idev);
883 ret = -1;
884 goto out;
886 memcpy(&addr.s6_addr[8], idev->rndid, 8);
887 tmp_valid_lft = min_t(__u32,
888 ifp->valid_lft,
889 idev->cnf.temp_valid_lft);
890 tmp_prefered_lft = min_t(__u32,
891 ifp->prefered_lft,
892 idev->cnf.temp_prefered_lft - desync_factor / HZ);
893 tmp_plen = ifp->prefix_len;
894 max_addresses = idev->cnf.max_addresses;
895 tmp_cstamp = ifp->cstamp;
896 tmp_tstamp = ifp->tstamp;
897 spin_unlock_bh(&ifp->lock);
899 regen_advance = idev->cnf.regen_max_retry *
900 idev->cnf.dad_transmits *
901 idev->nd_parms->retrans_time / HZ;
902 write_unlock(&idev->lock);
904 /* A temporary address is created only if this calculated Preferred
905 * Lifetime is greater than REGEN_ADVANCE time units. In particular,
906 * an implementation must not create a temporary address with a zero
907 * Preferred Lifetime.
909 if (tmp_prefered_lft <= regen_advance) {
910 in6_ifa_put(ifp);
911 in6_dev_put(idev);
912 ret = -1;
913 goto out;
916 addr_flags = IFA_F_TEMPORARY;
917 /* set in addrconf_prefix_rcv() */
918 if (ifp->flags & IFA_F_OPTIMISTIC)
919 addr_flags |= IFA_F_OPTIMISTIC;
921 ift = !max_addresses ||
922 ipv6_count_addresses(idev) < max_addresses ?
923 ipv6_add_addr(idev, &addr, tmp_plen,
924 ipv6_addr_type(&addr)&IPV6_ADDR_SCOPE_MASK,
925 addr_flags) : NULL;
926 if (!ift || IS_ERR(ift)) {
927 in6_ifa_put(ifp);
928 in6_dev_put(idev);
929 printk(KERN_INFO
930 "ipv6_create_tempaddr(): retry temporary address regeneration.\n");
931 tmpaddr = &addr;
932 write_lock(&idev->lock);
933 goto retry;
936 spin_lock_bh(&ift->lock);
937 ift->ifpub = ifp;
938 ift->valid_lft = tmp_valid_lft;
939 ift->prefered_lft = tmp_prefered_lft;
940 ift->cstamp = tmp_cstamp;
941 ift->tstamp = tmp_tstamp;
942 spin_unlock_bh(&ift->lock);
944 addrconf_dad_start(ift, 0);
945 in6_ifa_put(ift);
946 in6_dev_put(idev);
947 out:
948 return ret;
950 #endif
953 * Choose an appropriate source address (RFC3484)
955 enum {
956 IPV6_SADDR_RULE_INIT = 0,
957 IPV6_SADDR_RULE_LOCAL,
958 IPV6_SADDR_RULE_SCOPE,
959 IPV6_SADDR_RULE_PREFERRED,
960 #ifdef CONFIG_IPV6_MIP6
961 IPV6_SADDR_RULE_HOA,
962 #endif
963 IPV6_SADDR_RULE_OIF,
964 IPV6_SADDR_RULE_LABEL,
965 #ifdef CONFIG_IPV6_PRIVACY
966 IPV6_SADDR_RULE_PRIVACY,
967 #endif
968 IPV6_SADDR_RULE_ORCHID,
969 IPV6_SADDR_RULE_PREFIX,
970 IPV6_SADDR_RULE_MAX
973 struct ipv6_saddr_score {
974 int rule;
975 int addr_type;
976 struct inet6_ifaddr *ifa;
977 DECLARE_BITMAP(scorebits, IPV6_SADDR_RULE_MAX);
978 int scopedist;
979 int matchlen;
982 struct ipv6_saddr_dst {
983 const struct in6_addr *addr;
984 int ifindex;
985 int scope;
986 int label;
987 unsigned int prefs;
990 static inline int ipv6_saddr_preferred(int type)
992 if (type & (IPV6_ADDR_MAPPED|IPV6_ADDR_COMPATv4|
993 IPV6_ADDR_LOOPBACK|IPV6_ADDR_RESERVED))
994 return 1;
995 return 0;
998 static int ipv6_get_saddr_eval(struct net *net,
999 struct ipv6_saddr_score *score,
1000 struct ipv6_saddr_dst *dst,
1001 int i)
1003 int ret;
1005 if (i <= score->rule) {
1006 switch (i) {
1007 case IPV6_SADDR_RULE_SCOPE:
1008 ret = score->scopedist;
1009 break;
1010 case IPV6_SADDR_RULE_PREFIX:
1011 ret = score->matchlen;
1012 break;
1013 default:
1014 ret = !!test_bit(i, score->scorebits);
1016 goto out;
1019 switch (i) {
1020 case IPV6_SADDR_RULE_INIT:
1021 /* Rule 0: remember if hiscore is not ready yet */
1022 ret = !!score->ifa;
1023 break;
1024 case IPV6_SADDR_RULE_LOCAL:
1025 /* Rule 1: Prefer same address */
1026 ret = ipv6_addr_equal(&score->ifa->addr, dst->addr);
1027 break;
1028 case IPV6_SADDR_RULE_SCOPE:
1029 /* Rule 2: Prefer appropriate scope
1031 * ret
1033 * -1 | d 15
1034 * ---+--+-+---> scope
1036 * | d is scope of the destination.
1037 * B-d | \
1038 * | \ <- smaller scope is better if
1039 * B-15 | \ if scope is enough for destinaion.
1040 * | ret = B - scope (-1 <= scope >= d <= 15).
1041 * d-C-1 | /
1042 * |/ <- greater is better
1043 * -C / if scope is not enough for destination.
1044 * /| ret = scope - C (-1 <= d < scope <= 15).
1046 * d - C - 1 < B -15 (for all -1 <= d <= 15).
1047 * C > d + 14 - B >= 15 + 14 - B = 29 - B.
1048 * Assume B = 0 and we get C > 29.
1050 ret = __ipv6_addr_src_scope(score->addr_type);
1051 if (ret >= dst->scope)
1052 ret = -ret;
1053 else
1054 ret -= 128; /* 30 is enough */
1055 score->scopedist = ret;
1056 break;
1057 case IPV6_SADDR_RULE_PREFERRED:
1058 /* Rule 3: Avoid deprecated and optimistic addresses */
1059 ret = ipv6_saddr_preferred(score->addr_type) ||
1060 !(score->ifa->flags & (IFA_F_DEPRECATED|IFA_F_OPTIMISTIC));
1061 break;
1062 #ifdef CONFIG_IPV6_MIP6
1063 case IPV6_SADDR_RULE_HOA:
1065 /* Rule 4: Prefer home address */
1066 int prefhome = !(dst->prefs & IPV6_PREFER_SRC_COA);
1067 ret = !(score->ifa->flags & IFA_F_HOMEADDRESS) ^ prefhome;
1068 break;
1070 #endif
1071 case IPV6_SADDR_RULE_OIF:
1072 /* Rule 5: Prefer outgoing interface */
1073 ret = (!dst->ifindex ||
1074 dst->ifindex == score->ifa->idev->dev->ifindex);
1075 break;
1076 case IPV6_SADDR_RULE_LABEL:
1077 /* Rule 6: Prefer matching label */
1078 ret = ipv6_addr_label(net,
1079 &score->ifa->addr, score->addr_type,
1080 score->ifa->idev->dev->ifindex) == dst->label;
1081 break;
1082 #ifdef CONFIG_IPV6_PRIVACY
1083 case IPV6_SADDR_RULE_PRIVACY:
1085 /* Rule 7: Prefer public address
1086 * Note: prefer temprary address if use_tempaddr >= 2
1088 int preftmp = dst->prefs & (IPV6_PREFER_SRC_PUBLIC|IPV6_PREFER_SRC_TMP) ?
1089 !!(dst->prefs & IPV6_PREFER_SRC_TMP) :
1090 score->ifa->idev->cnf.use_tempaddr >= 2;
1091 ret = (!(score->ifa->flags & IFA_F_TEMPORARY)) ^ preftmp;
1092 break;
1094 #endif
1095 case IPV6_SADDR_RULE_ORCHID:
1096 /* Rule 8-: Prefer ORCHID vs ORCHID or
1097 * non-ORCHID vs non-ORCHID
1099 ret = !(ipv6_addr_orchid(&score->ifa->addr) ^
1100 ipv6_addr_orchid(dst->addr));
1101 break;
1102 case IPV6_SADDR_RULE_PREFIX:
1103 /* Rule 8: Use longest matching prefix */
1104 score->matchlen = ret = ipv6_addr_diff(&score->ifa->addr,
1105 dst->addr);
1106 break;
1107 default:
1108 ret = 0;
1111 if (ret)
1112 __set_bit(i, score->scorebits);
1113 score->rule = i;
1114 out:
1115 return ret;
1118 int ipv6_dev_get_saddr(struct net *net, struct net_device *dst_dev,
1119 const struct in6_addr *daddr, unsigned int prefs,
1120 struct in6_addr *saddr)
1122 struct ipv6_saddr_score scores[2],
1123 *score = &scores[0], *hiscore = &scores[1];
1124 struct ipv6_saddr_dst dst;
1125 struct net_device *dev;
1126 int dst_type;
1128 dst_type = __ipv6_addr_type(daddr);
1129 dst.addr = daddr;
1130 dst.ifindex = dst_dev ? dst_dev->ifindex : 0;
1131 dst.scope = __ipv6_addr_src_scope(dst_type);
1132 dst.label = ipv6_addr_label(net, daddr, dst_type, dst.ifindex);
1133 dst.prefs = prefs;
1135 hiscore->rule = -1;
1136 hiscore->ifa = NULL;
1138 rcu_read_lock();
1140 for_each_netdev_rcu(net, dev) {
1141 struct inet6_dev *idev;
1143 /* Candidate Source Address (section 4)
1144 * - multicast and link-local destination address,
1145 * the set of candidate source address MUST only
1146 * include addresses assigned to interfaces
1147 * belonging to the same link as the outgoing
1148 * interface.
1149 * (- For site-local destination addresses, the
1150 * set of candidate source addresses MUST only
1151 * include addresses assigned to interfaces
1152 * belonging to the same site as the outgoing
1153 * interface.)
1155 if (((dst_type & IPV6_ADDR_MULTICAST) ||
1156 dst.scope <= IPV6_ADDR_SCOPE_LINKLOCAL) &&
1157 dst.ifindex && dev->ifindex != dst.ifindex)
1158 continue;
1160 idev = __in6_dev_get(dev);
1161 if (!idev)
1162 continue;
1164 read_lock_bh(&idev->lock);
1165 for (score->ifa = idev->addr_list; score->ifa; score->ifa = score->ifa->if_next) {
1166 int i;
1169 * - Tentative Address (RFC2462 section 5.4)
1170 * - A tentative address is not considered
1171 * "assigned to an interface" in the traditional
1172 * sense, unless it is also flagged as optimistic.
1173 * - Candidate Source Address (section 4)
1174 * - In any case, anycast addresses, multicast
1175 * addresses, and the unspecified address MUST
1176 * NOT be included in a candidate set.
1178 if ((score->ifa->flags & IFA_F_TENTATIVE) &&
1179 (!(score->ifa->flags & IFA_F_OPTIMISTIC)))
1180 continue;
1182 score->addr_type = __ipv6_addr_type(&score->ifa->addr);
1184 if (unlikely(score->addr_type == IPV6_ADDR_ANY ||
1185 score->addr_type & IPV6_ADDR_MULTICAST)) {
1186 LIMIT_NETDEBUG(KERN_DEBUG
1187 "ADDRCONF: unspecified / multicast address "
1188 "assigned as unicast address on %s",
1189 dev->name);
1190 continue;
1193 score->rule = -1;
1194 bitmap_zero(score->scorebits, IPV6_SADDR_RULE_MAX);
1196 for (i = 0; i < IPV6_SADDR_RULE_MAX; i++) {
1197 int minihiscore, miniscore;
1199 minihiscore = ipv6_get_saddr_eval(net, hiscore, &dst, i);
1200 miniscore = ipv6_get_saddr_eval(net, score, &dst, i);
1202 if (minihiscore > miniscore) {
1203 if (i == IPV6_SADDR_RULE_SCOPE &&
1204 score->scopedist > 0) {
1206 * special case:
1207 * each remaining entry
1208 * has too small (not enough)
1209 * scope, because ifa entries
1210 * are sorted by their scope
1211 * values.
1213 goto try_nextdev;
1215 break;
1216 } else if (minihiscore < miniscore) {
1217 if (hiscore->ifa)
1218 in6_ifa_put(hiscore->ifa);
1220 in6_ifa_hold(score->ifa);
1222 swap(hiscore, score);
1224 /* restore our iterator */
1225 score->ifa = hiscore->ifa;
1227 break;
1231 try_nextdev:
1232 read_unlock_bh(&idev->lock);
1234 rcu_read_unlock();
1236 if (!hiscore->ifa)
1237 return -EADDRNOTAVAIL;
1239 ipv6_addr_copy(saddr, &hiscore->ifa->addr);
1240 in6_ifa_put(hiscore->ifa);
1241 return 0;
1244 EXPORT_SYMBOL(ipv6_dev_get_saddr);
1246 int ipv6_get_lladdr(struct net_device *dev, struct in6_addr *addr,
1247 unsigned char banned_flags)
1249 struct inet6_dev *idev;
1250 int err = -EADDRNOTAVAIL;
1252 rcu_read_lock();
1253 if ((idev = __in6_dev_get(dev)) != NULL) {
1254 struct inet6_ifaddr *ifp;
1256 read_lock_bh(&idev->lock);
1257 for (ifp=idev->addr_list; ifp; ifp=ifp->if_next) {
1258 if (ifp->scope == IFA_LINK && !(ifp->flags & banned_flags)) {
1259 ipv6_addr_copy(addr, &ifp->addr);
1260 err = 0;
1261 break;
1264 read_unlock_bh(&idev->lock);
1266 rcu_read_unlock();
1267 return err;
1270 static int ipv6_count_addresses(struct inet6_dev *idev)
1272 int cnt = 0;
1273 struct inet6_ifaddr *ifp;
1275 read_lock_bh(&idev->lock);
1276 for (ifp=idev->addr_list; ifp; ifp=ifp->if_next)
1277 cnt++;
1278 read_unlock_bh(&idev->lock);
1279 return cnt;
1282 int ipv6_chk_addr(struct net *net, struct in6_addr *addr,
1283 struct net_device *dev, int strict)
1285 struct inet6_ifaddr * ifp;
1286 u8 hash = ipv6_addr_hash(addr);
1288 read_lock_bh(&addrconf_hash_lock);
1289 for(ifp = inet6_addr_lst[hash]; ifp; ifp=ifp->lst_next) {
1290 if (!net_eq(dev_net(ifp->idev->dev), net))
1291 continue;
1292 if (ipv6_addr_equal(&ifp->addr, addr) &&
1293 !(ifp->flags&IFA_F_TENTATIVE)) {
1294 if (dev == NULL || ifp->idev->dev == dev ||
1295 !(ifp->scope&(IFA_LINK|IFA_HOST) || strict))
1296 break;
1299 read_unlock_bh(&addrconf_hash_lock);
1300 return ifp != NULL;
1302 EXPORT_SYMBOL(ipv6_chk_addr);
1304 static
1305 int ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
1306 struct net_device *dev)
1308 struct inet6_ifaddr * ifp;
1309 u8 hash = ipv6_addr_hash(addr);
1311 for(ifp = inet6_addr_lst[hash]; ifp; ifp=ifp->lst_next) {
1312 if (!net_eq(dev_net(ifp->idev->dev), net))
1313 continue;
1314 if (ipv6_addr_equal(&ifp->addr, addr)) {
1315 if (dev == NULL || ifp->idev->dev == dev)
1316 break;
1319 return ifp != NULL;
1322 int ipv6_chk_prefix(struct in6_addr *addr, struct net_device *dev)
1324 struct inet6_dev *idev;
1325 struct inet6_ifaddr *ifa;
1326 int onlink;
1328 onlink = 0;
1329 rcu_read_lock();
1330 idev = __in6_dev_get(dev);
1331 if (idev) {
1332 read_lock_bh(&idev->lock);
1333 for (ifa = idev->addr_list; ifa; ifa = ifa->if_next) {
1334 onlink = ipv6_prefix_equal(addr, &ifa->addr,
1335 ifa->prefix_len);
1336 if (onlink)
1337 break;
1339 read_unlock_bh(&idev->lock);
1341 rcu_read_unlock();
1342 return onlink;
1345 EXPORT_SYMBOL(ipv6_chk_prefix);
1347 struct inet6_ifaddr *ipv6_get_ifaddr(struct net *net, const struct in6_addr *addr,
1348 struct net_device *dev, int strict)
1350 struct inet6_ifaddr * ifp;
1351 u8 hash = ipv6_addr_hash(addr);
1353 read_lock_bh(&addrconf_hash_lock);
1354 for(ifp = inet6_addr_lst[hash]; ifp; ifp=ifp->lst_next) {
1355 if (!net_eq(dev_net(ifp->idev->dev), net))
1356 continue;
1357 if (ipv6_addr_equal(&ifp->addr, addr)) {
1358 if (dev == NULL || ifp->idev->dev == dev ||
1359 !(ifp->scope&(IFA_LINK|IFA_HOST) || strict)) {
1360 in6_ifa_hold(ifp);
1361 break;
1365 read_unlock_bh(&addrconf_hash_lock);
1367 return ifp;
1370 /* Gets referenced address, destroys ifaddr */
1372 static void addrconf_dad_stop(struct inet6_ifaddr *ifp, int dad_failed)
1374 if (ifp->flags&IFA_F_PERMANENT) {
1375 spin_lock_bh(&ifp->lock);
1376 addrconf_del_timer(ifp);
1377 ifp->flags |= IFA_F_TENTATIVE;
1378 if (dad_failed)
1379 ifp->flags |= IFA_F_DADFAILED;
1380 spin_unlock_bh(&ifp->lock);
1381 in6_ifa_put(ifp);
1382 #ifdef CONFIG_IPV6_PRIVACY
1383 } else if (ifp->flags&IFA_F_TEMPORARY) {
1384 struct inet6_ifaddr *ifpub;
1385 spin_lock_bh(&ifp->lock);
1386 ifpub = ifp->ifpub;
1387 if (ifpub) {
1388 in6_ifa_hold(ifpub);
1389 spin_unlock_bh(&ifp->lock);
1390 ipv6_create_tempaddr(ifpub, ifp);
1391 in6_ifa_put(ifpub);
1392 } else {
1393 spin_unlock_bh(&ifp->lock);
1395 ipv6_del_addr(ifp);
1396 #endif
1397 } else
1398 ipv6_del_addr(ifp);
1401 void addrconf_dad_failure(struct inet6_ifaddr *ifp)
1403 struct inet6_dev *idev = ifp->idev;
1405 if (net_ratelimit())
1406 printk(KERN_INFO "%s: IPv6 duplicate address %pI6c detected!\n",
1407 ifp->idev->dev->name, &ifp->addr);
1409 if (idev->cnf.accept_dad > 1 && !idev->cnf.disable_ipv6) {
1410 struct in6_addr addr;
1412 addr.s6_addr32[0] = htonl(0xfe800000);
1413 addr.s6_addr32[1] = 0;
1415 if (!ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) &&
1416 ipv6_addr_equal(&ifp->addr, &addr)) {
1417 /* DAD failed for link-local based on MAC address */
1418 idev->cnf.disable_ipv6 = 1;
1420 printk(KERN_INFO "%s: IPv6 being disabled!\n",
1421 ifp->idev->dev->name);
1425 addrconf_dad_stop(ifp, 1);
1428 /* Join to solicited addr multicast group. */
1430 void addrconf_join_solict(struct net_device *dev, struct in6_addr *addr)
1432 struct in6_addr maddr;
1434 if (dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1435 return;
1437 addrconf_addr_solict_mult(addr, &maddr);
1438 ipv6_dev_mc_inc(dev, &maddr);
1441 void addrconf_leave_solict(struct inet6_dev *idev, struct in6_addr *addr)
1443 struct in6_addr maddr;
1445 if (idev->dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1446 return;
1448 addrconf_addr_solict_mult(addr, &maddr);
1449 __ipv6_dev_mc_dec(idev, &maddr);
1452 static void addrconf_join_anycast(struct inet6_ifaddr *ifp)
1454 struct in6_addr addr;
1455 ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1456 if (ipv6_addr_any(&addr))
1457 return;
1458 ipv6_dev_ac_inc(ifp->idev->dev, &addr);
1461 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp)
1463 struct in6_addr addr;
1464 ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1465 if (ipv6_addr_any(&addr))
1466 return;
1467 __ipv6_dev_ac_dec(ifp->idev, &addr);
1470 static int addrconf_ifid_eui48(u8 *eui, struct net_device *dev)
1472 if (dev->addr_len != ETH_ALEN)
1473 return -1;
1474 memcpy(eui, dev->dev_addr, 3);
1475 memcpy(eui + 5, dev->dev_addr + 3, 3);
1478 * The zSeries OSA network cards can be shared among various
1479 * OS instances, but the OSA cards have only one MAC address.
1480 * This leads to duplicate address conflicts in conjunction
1481 * with IPv6 if more than one instance uses the same card.
1483 * The driver for these cards can deliver a unique 16-bit
1484 * identifier for each instance sharing the same card. It is
1485 * placed instead of 0xFFFE in the interface identifier. The
1486 * "u" bit of the interface identifier is not inverted in this
1487 * case. Hence the resulting interface identifier has local
1488 * scope according to RFC2373.
1490 if (dev->dev_id) {
1491 eui[3] = (dev->dev_id >> 8) & 0xFF;
1492 eui[4] = dev->dev_id & 0xFF;
1493 } else {
1494 eui[3] = 0xFF;
1495 eui[4] = 0xFE;
1496 eui[0] ^= 2;
1498 return 0;
1501 static int addrconf_ifid_arcnet(u8 *eui, struct net_device *dev)
1503 /* XXX: inherit EUI-64 from other interface -- yoshfuji */
1504 if (dev->addr_len != ARCNET_ALEN)
1505 return -1;
1506 memset(eui, 0, 7);
1507 eui[7] = *(u8*)dev->dev_addr;
1508 return 0;
1511 static int addrconf_ifid_infiniband(u8 *eui, struct net_device *dev)
1513 if (dev->addr_len != INFINIBAND_ALEN)
1514 return -1;
1515 memcpy(eui, dev->dev_addr + 12, 8);
1516 eui[0] |= 2;
1517 return 0;
1520 int __ipv6_isatap_ifid(u8 *eui, __be32 addr)
1522 if (addr == 0)
1523 return -1;
1524 eui[0] = (ipv4_is_zeronet(addr) || ipv4_is_private_10(addr) ||
1525 ipv4_is_loopback(addr) || ipv4_is_linklocal_169(addr) ||
1526 ipv4_is_private_172(addr) || ipv4_is_test_192(addr) ||
1527 ipv4_is_anycast_6to4(addr) || ipv4_is_private_192(addr) ||
1528 ipv4_is_test_198(addr) || ipv4_is_multicast(addr) ||
1529 ipv4_is_lbcast(addr)) ? 0x00 : 0x02;
1530 eui[1] = 0;
1531 eui[2] = 0x5E;
1532 eui[3] = 0xFE;
1533 memcpy(eui + 4, &addr, 4);
1534 return 0;
1536 EXPORT_SYMBOL(__ipv6_isatap_ifid);
1538 static int addrconf_ifid_sit(u8 *eui, struct net_device *dev)
1540 if (dev->priv_flags & IFF_ISATAP)
1541 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
1542 return -1;
1545 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev)
1547 switch (dev->type) {
1548 case ARPHRD_ETHER:
1549 case ARPHRD_FDDI:
1550 case ARPHRD_IEEE802_TR:
1551 return addrconf_ifid_eui48(eui, dev);
1552 case ARPHRD_ARCNET:
1553 return addrconf_ifid_arcnet(eui, dev);
1554 case ARPHRD_INFINIBAND:
1555 return addrconf_ifid_infiniband(eui, dev);
1556 case ARPHRD_SIT:
1557 return addrconf_ifid_sit(eui, dev);
1559 return -1;
1562 static int ipv6_inherit_eui64(u8 *eui, struct inet6_dev *idev)
1564 int err = -1;
1565 struct inet6_ifaddr *ifp;
1567 read_lock_bh(&idev->lock);
1568 for (ifp=idev->addr_list; ifp; ifp=ifp->if_next) {
1569 if (ifp->scope == IFA_LINK && !(ifp->flags&IFA_F_TENTATIVE)) {
1570 memcpy(eui, ifp->addr.s6_addr+8, 8);
1571 err = 0;
1572 break;
1575 read_unlock_bh(&idev->lock);
1576 return err;
1579 #ifdef CONFIG_IPV6_PRIVACY
1580 /* (re)generation of randomized interface identifier (RFC 3041 3.2, 3.5) */
1581 static int __ipv6_regen_rndid(struct inet6_dev *idev)
1583 regen:
1584 get_random_bytes(idev->rndid, sizeof(idev->rndid));
1585 idev->rndid[0] &= ~0x02;
1588 * <draft-ietf-ipngwg-temp-addresses-v2-00.txt>:
1589 * check if generated address is not inappropriate
1591 * - Reserved subnet anycast (RFC 2526)
1592 * 11111101 11....11 1xxxxxxx
1593 * - ISATAP (RFC4214) 6.1
1594 * 00-00-5E-FE-xx-xx-xx-xx
1595 * - value 0
1596 * - XXX: already assigned to an address on the device
1598 if (idev->rndid[0] == 0xfd &&
1599 (idev->rndid[1]&idev->rndid[2]&idev->rndid[3]&idev->rndid[4]&idev->rndid[5]&idev->rndid[6]) == 0xff &&
1600 (idev->rndid[7]&0x80))
1601 goto regen;
1602 if ((idev->rndid[0]|idev->rndid[1]) == 0) {
1603 if (idev->rndid[2] == 0x5e && idev->rndid[3] == 0xfe)
1604 goto regen;
1605 if ((idev->rndid[2]|idev->rndid[3]|idev->rndid[4]|idev->rndid[5]|idev->rndid[6]|idev->rndid[7]) == 0x00)
1606 goto regen;
1609 return 0;
1612 static void ipv6_regen_rndid(unsigned long data)
1614 struct inet6_dev *idev = (struct inet6_dev *) data;
1615 unsigned long expires;
1617 rcu_read_lock_bh();
1618 write_lock_bh(&idev->lock);
1620 if (idev->dead)
1621 goto out;
1623 if (__ipv6_regen_rndid(idev) < 0)
1624 goto out;
1626 expires = jiffies +
1627 idev->cnf.temp_prefered_lft * HZ -
1628 idev->cnf.regen_max_retry * idev->cnf.dad_transmits * idev->nd_parms->retrans_time - desync_factor;
1629 if (time_before(expires, jiffies)) {
1630 printk(KERN_WARNING
1631 "ipv6_regen_rndid(): too short regeneration interval; timer disabled for %s.\n",
1632 idev->dev->name);
1633 goto out;
1636 if (!mod_timer(&idev->regen_timer, expires))
1637 in6_dev_hold(idev);
1639 out:
1640 write_unlock_bh(&idev->lock);
1641 rcu_read_unlock_bh();
1642 in6_dev_put(idev);
1645 static int __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr) {
1646 int ret = 0;
1648 if (tmpaddr && memcmp(idev->rndid, &tmpaddr->s6_addr[8], 8) == 0)
1649 ret = __ipv6_regen_rndid(idev);
1650 return ret;
1652 #endif
1655 * Add prefix route.
1658 static void
1659 addrconf_prefix_route(struct in6_addr *pfx, int plen, struct net_device *dev,
1660 unsigned long expires, u32 flags)
1662 struct fib6_config cfg = {
1663 .fc_table = RT6_TABLE_PREFIX,
1664 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1665 .fc_ifindex = dev->ifindex,
1666 .fc_expires = expires,
1667 .fc_dst_len = plen,
1668 .fc_flags = RTF_UP | flags,
1669 .fc_nlinfo.nl_net = dev_net(dev),
1670 .fc_protocol = RTPROT_KERNEL,
1673 ipv6_addr_copy(&cfg.fc_dst, pfx);
1675 /* Prevent useless cloning on PtP SIT.
1676 This thing is done here expecting that the whole
1677 class of non-broadcast devices need not cloning.
1679 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
1680 if (dev->type == ARPHRD_SIT && (dev->flags & IFF_POINTOPOINT))
1681 cfg.fc_flags |= RTF_NONEXTHOP;
1682 #endif
1684 ip6_route_add(&cfg);
1687 /* Create "default" multicast route to the interface */
1689 static void addrconf_add_mroute(struct net_device *dev)
1691 struct fib6_config cfg = {
1692 .fc_table = RT6_TABLE_LOCAL,
1693 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1694 .fc_ifindex = dev->ifindex,
1695 .fc_dst_len = 8,
1696 .fc_flags = RTF_UP,
1697 .fc_nlinfo.nl_net = dev_net(dev),
1700 ipv6_addr_set(&cfg.fc_dst, htonl(0xFF000000), 0, 0, 0);
1702 ip6_route_add(&cfg);
1705 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
1706 static void sit_route_add(struct net_device *dev)
1708 struct fib6_config cfg = {
1709 .fc_table = RT6_TABLE_MAIN,
1710 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1711 .fc_ifindex = dev->ifindex,
1712 .fc_dst_len = 96,
1713 .fc_flags = RTF_UP | RTF_NONEXTHOP,
1714 .fc_nlinfo.nl_net = dev_net(dev),
1717 /* prefix length - 96 bits "::d.d.d.d" */
1718 ip6_route_add(&cfg);
1720 #endif
1722 static void addrconf_add_lroute(struct net_device *dev)
1724 struct in6_addr addr;
1726 ipv6_addr_set(&addr, htonl(0xFE800000), 0, 0, 0);
1727 addrconf_prefix_route(&addr, 64, dev, 0, 0);
1730 static struct inet6_dev *addrconf_add_dev(struct net_device *dev)
1732 struct inet6_dev *idev;
1734 ASSERT_RTNL();
1736 if ((idev = ipv6_find_idev(dev)) == NULL)
1737 return NULL;
1739 /* Add default multicast route */
1740 addrconf_add_mroute(dev);
1742 /* Add link local route */
1743 addrconf_add_lroute(dev);
1744 return idev;
1747 void addrconf_prefix_rcv(struct net_device *dev, u8 *opt, int len)
1749 struct prefix_info *pinfo;
1750 __u32 valid_lft;
1751 __u32 prefered_lft;
1752 int addr_type;
1753 struct inet6_dev *in6_dev;
1754 struct net *net = dev_net(dev);
1756 pinfo = (struct prefix_info *) opt;
1758 if (len < sizeof(struct prefix_info)) {
1759 ADBG(("addrconf: prefix option too short\n"));
1760 return;
1764 * Validation checks ([ADDRCONF], page 19)
1767 addr_type = ipv6_addr_type(&pinfo->prefix);
1769 if (addr_type & (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL))
1770 return;
1772 valid_lft = ntohl(pinfo->valid);
1773 prefered_lft = ntohl(pinfo->prefered);
1775 if (prefered_lft > valid_lft) {
1776 if (net_ratelimit())
1777 printk(KERN_WARNING "addrconf: prefix option has invalid lifetime\n");
1778 return;
1781 in6_dev = in6_dev_get(dev);
1783 if (in6_dev == NULL) {
1784 if (net_ratelimit())
1785 printk(KERN_DEBUG "addrconf: device %s not configured\n", dev->name);
1786 return;
1790 * Two things going on here:
1791 * 1) Add routes for on-link prefixes
1792 * 2) Configure prefixes with the auto flag set
1795 if (pinfo->onlink) {
1796 struct rt6_info *rt;
1797 unsigned long rt_expires;
1799 /* Avoid arithmetic overflow. Really, we could
1800 * save rt_expires in seconds, likely valid_lft,
1801 * but it would require division in fib gc, that it
1802 * not good.
1804 if (HZ > USER_HZ)
1805 rt_expires = addrconf_timeout_fixup(valid_lft, HZ);
1806 else
1807 rt_expires = addrconf_timeout_fixup(valid_lft, USER_HZ);
1809 if (addrconf_finite_timeout(rt_expires))
1810 rt_expires *= HZ;
1812 rt = rt6_lookup(net, &pinfo->prefix, NULL,
1813 dev->ifindex, 1);
1815 if (rt && addrconf_is_prefix_route(rt)) {
1816 /* Autoconf prefix route */
1817 if (valid_lft == 0) {
1818 ip6_del_rt(rt);
1819 rt = NULL;
1820 } else if (addrconf_finite_timeout(rt_expires)) {
1821 /* not infinity */
1822 rt->rt6i_expires = jiffies + rt_expires;
1823 rt->rt6i_flags |= RTF_EXPIRES;
1824 } else {
1825 rt->rt6i_flags &= ~RTF_EXPIRES;
1826 rt->rt6i_expires = 0;
1828 } else if (valid_lft) {
1829 clock_t expires = 0;
1830 int flags = RTF_ADDRCONF | RTF_PREFIX_RT;
1831 if (addrconf_finite_timeout(rt_expires)) {
1832 /* not infinity */
1833 flags |= RTF_EXPIRES;
1834 expires = jiffies_to_clock_t(rt_expires);
1836 addrconf_prefix_route(&pinfo->prefix, pinfo->prefix_len,
1837 dev, expires, flags);
1839 if (rt)
1840 dst_release(&rt->u.dst);
1843 /* Try to figure out our local address for this prefix */
1845 if (pinfo->autoconf && in6_dev->cnf.autoconf) {
1846 struct inet6_ifaddr * ifp;
1847 struct in6_addr addr;
1848 int create = 0, update_lft = 0;
1850 if (pinfo->prefix_len == 64) {
1851 memcpy(&addr, &pinfo->prefix, 8);
1852 if (ipv6_generate_eui64(addr.s6_addr + 8, dev) &&
1853 ipv6_inherit_eui64(addr.s6_addr + 8, in6_dev)) {
1854 in6_dev_put(in6_dev);
1855 return;
1857 goto ok;
1859 if (net_ratelimit())
1860 printk(KERN_DEBUG "IPv6 addrconf: prefix with wrong length %d\n",
1861 pinfo->prefix_len);
1862 in6_dev_put(in6_dev);
1863 return;
1867 ifp = ipv6_get_ifaddr(net, &addr, dev, 1);
1869 if (ifp == NULL && valid_lft) {
1870 int max_addresses = in6_dev->cnf.max_addresses;
1871 u32 addr_flags = 0;
1873 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1874 if (in6_dev->cnf.optimistic_dad &&
1875 !net->ipv6.devconf_all->forwarding)
1876 addr_flags = IFA_F_OPTIMISTIC;
1877 #endif
1879 /* Do not allow to create too much of autoconfigured
1880 * addresses; this would be too easy way to crash kernel.
1882 if (!max_addresses ||
1883 ipv6_count_addresses(in6_dev) < max_addresses)
1884 ifp = ipv6_add_addr(in6_dev, &addr, pinfo->prefix_len,
1885 addr_type&IPV6_ADDR_SCOPE_MASK,
1886 addr_flags);
1888 if (!ifp || IS_ERR(ifp)) {
1889 in6_dev_put(in6_dev);
1890 return;
1893 update_lft = create = 1;
1894 ifp->cstamp = jiffies;
1895 addrconf_dad_start(ifp, RTF_ADDRCONF|RTF_PREFIX_RT);
1898 if (ifp) {
1899 int flags;
1900 unsigned long now;
1901 #ifdef CONFIG_IPV6_PRIVACY
1902 struct inet6_ifaddr *ift;
1903 #endif
1904 u32 stored_lft;
1906 /* update lifetime (RFC2462 5.5.3 e) */
1907 spin_lock(&ifp->lock);
1908 now = jiffies;
1909 if (ifp->valid_lft > (now - ifp->tstamp) / HZ)
1910 stored_lft = ifp->valid_lft - (now - ifp->tstamp) / HZ;
1911 else
1912 stored_lft = 0;
1913 if (!update_lft && stored_lft) {
1914 if (valid_lft > MIN_VALID_LIFETIME ||
1915 valid_lft > stored_lft)
1916 update_lft = 1;
1917 else if (stored_lft <= MIN_VALID_LIFETIME) {
1918 /* valid_lft <= stored_lft is always true */
1920 * RFC 4862 Section 5.5.3e:
1921 * "Note that the preferred lifetime of
1922 * the corresponding address is always
1923 * reset to the Preferred Lifetime in
1924 * the received Prefix Information
1925 * option, regardless of whether the
1926 * valid lifetime is also reset or
1927 * ignored."
1929 * So if the preferred lifetime in
1930 * this advertisement is different
1931 * than what we have stored, but the
1932 * valid lifetime is invalid, just
1933 * reset prefered_lft.
1935 * We must set the valid lifetime
1936 * to the stored lifetime since we'll
1937 * be updating the timestamp below,
1938 * else we'll set it back to the
1939 * minumum.
1941 if (prefered_lft != ifp->prefered_lft) {
1942 valid_lft = stored_lft;
1943 update_lft = 1;
1945 } else {
1946 valid_lft = MIN_VALID_LIFETIME;
1947 if (valid_lft < prefered_lft)
1948 prefered_lft = valid_lft;
1949 update_lft = 1;
1953 if (update_lft) {
1954 ifp->valid_lft = valid_lft;
1955 ifp->prefered_lft = prefered_lft;
1956 ifp->tstamp = now;
1957 flags = ifp->flags;
1958 ifp->flags &= ~IFA_F_DEPRECATED;
1959 spin_unlock(&ifp->lock);
1961 if (!(flags&IFA_F_TENTATIVE))
1962 ipv6_ifa_notify(0, ifp);
1963 } else
1964 spin_unlock(&ifp->lock);
1966 #ifdef CONFIG_IPV6_PRIVACY
1967 read_lock_bh(&in6_dev->lock);
1968 /* update all temporary addresses in the list */
1969 for (ift=in6_dev->tempaddr_list; ift; ift=ift->tmp_next) {
1971 * When adjusting the lifetimes of an existing
1972 * temporary address, only lower the lifetimes.
1973 * Implementations must not increase the
1974 * lifetimes of an existing temporary address
1975 * when processing a Prefix Information Option.
1977 if (ifp != ift->ifpub)
1978 continue;
1980 spin_lock(&ift->lock);
1981 flags = ift->flags;
1982 if (ift->valid_lft > valid_lft &&
1983 ift->valid_lft - valid_lft > (jiffies - ift->tstamp) / HZ)
1984 ift->valid_lft = valid_lft + (jiffies - ift->tstamp) / HZ;
1985 if (ift->prefered_lft > prefered_lft &&
1986 ift->prefered_lft - prefered_lft > (jiffies - ift->tstamp) / HZ)
1987 ift->prefered_lft = prefered_lft + (jiffies - ift->tstamp) / HZ;
1988 spin_unlock(&ift->lock);
1989 if (!(flags&IFA_F_TENTATIVE))
1990 ipv6_ifa_notify(0, ift);
1993 if (create && in6_dev->cnf.use_tempaddr > 0) {
1995 * When a new public address is created as described in [ADDRCONF],
1996 * also create a new temporary address.
1998 read_unlock_bh(&in6_dev->lock);
1999 ipv6_create_tempaddr(ifp, NULL);
2000 } else {
2001 read_unlock_bh(&in6_dev->lock);
2003 #endif
2004 in6_ifa_put(ifp);
2005 addrconf_verify(0);
2008 inet6_prefix_notify(RTM_NEWPREFIX, in6_dev, pinfo);
2009 in6_dev_put(in6_dev);
2013 * Set destination address.
2014 * Special case for SIT interfaces where we create a new "virtual"
2015 * device.
2017 int addrconf_set_dstaddr(struct net *net, void __user *arg)
2019 struct in6_ifreq ireq;
2020 struct net_device *dev;
2021 int err = -EINVAL;
2023 rtnl_lock();
2025 err = -EFAULT;
2026 if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2027 goto err_exit;
2029 dev = __dev_get_by_index(net, ireq.ifr6_ifindex);
2031 err = -ENODEV;
2032 if (dev == NULL)
2033 goto err_exit;
2035 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2036 if (dev->type == ARPHRD_SIT) {
2037 const struct net_device_ops *ops = dev->netdev_ops;
2038 struct ifreq ifr;
2039 struct ip_tunnel_parm p;
2041 err = -EADDRNOTAVAIL;
2042 if (!(ipv6_addr_type(&ireq.ifr6_addr) & IPV6_ADDR_COMPATv4))
2043 goto err_exit;
2045 memset(&p, 0, sizeof(p));
2046 p.iph.daddr = ireq.ifr6_addr.s6_addr32[3];
2047 p.iph.saddr = 0;
2048 p.iph.version = 4;
2049 p.iph.ihl = 5;
2050 p.iph.protocol = IPPROTO_IPV6;
2051 p.iph.ttl = 64;
2052 ifr.ifr_ifru.ifru_data = (__force void __user *)&p;
2054 if (ops->ndo_do_ioctl) {
2055 mm_segment_t oldfs = get_fs();
2057 set_fs(KERNEL_DS);
2058 err = ops->ndo_do_ioctl(dev, &ifr, SIOCADDTUNNEL);
2059 set_fs(oldfs);
2060 } else
2061 err = -EOPNOTSUPP;
2063 if (err == 0) {
2064 err = -ENOBUFS;
2065 dev = __dev_get_by_name(net, p.name);
2066 if (!dev)
2067 goto err_exit;
2068 err = dev_open(dev);
2071 #endif
2073 err_exit:
2074 rtnl_unlock();
2075 return err;
2079 * Manual configuration of address on an interface
2081 static int inet6_addr_add(struct net *net, int ifindex, struct in6_addr *pfx,
2082 unsigned int plen, __u8 ifa_flags, __u32 prefered_lft,
2083 __u32 valid_lft)
2085 struct inet6_ifaddr *ifp;
2086 struct inet6_dev *idev;
2087 struct net_device *dev;
2088 int scope;
2089 u32 flags;
2090 clock_t expires;
2091 unsigned long timeout;
2093 ASSERT_RTNL();
2095 if (plen > 128)
2096 return -EINVAL;
2098 /* check the lifetime */
2099 if (!valid_lft || prefered_lft > valid_lft)
2100 return -EINVAL;
2102 dev = __dev_get_by_index(net, ifindex);
2103 if (!dev)
2104 return -ENODEV;
2106 if ((idev = addrconf_add_dev(dev)) == NULL)
2107 return -ENOBUFS;
2109 scope = ipv6_addr_scope(pfx);
2111 timeout = addrconf_timeout_fixup(valid_lft, HZ);
2112 if (addrconf_finite_timeout(timeout)) {
2113 expires = jiffies_to_clock_t(timeout * HZ);
2114 valid_lft = timeout;
2115 flags = RTF_EXPIRES;
2116 } else {
2117 expires = 0;
2118 flags = 0;
2119 ifa_flags |= IFA_F_PERMANENT;
2122 timeout = addrconf_timeout_fixup(prefered_lft, HZ);
2123 if (addrconf_finite_timeout(timeout)) {
2124 if (timeout == 0)
2125 ifa_flags |= IFA_F_DEPRECATED;
2126 prefered_lft = timeout;
2129 ifp = ipv6_add_addr(idev, pfx, plen, scope, ifa_flags);
2131 if (!IS_ERR(ifp)) {
2132 spin_lock_bh(&ifp->lock);
2133 ifp->valid_lft = valid_lft;
2134 ifp->prefered_lft = prefered_lft;
2135 ifp->tstamp = jiffies;
2136 spin_unlock_bh(&ifp->lock);
2138 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, dev,
2139 expires, flags);
2141 * Note that section 3.1 of RFC 4429 indicates
2142 * that the Optimistic flag should not be set for
2143 * manually configured addresses
2145 addrconf_dad_start(ifp, 0);
2146 in6_ifa_put(ifp);
2147 addrconf_verify(0);
2148 return 0;
2151 return PTR_ERR(ifp);
2154 static int inet6_addr_del(struct net *net, int ifindex, struct in6_addr *pfx,
2155 unsigned int plen)
2157 struct inet6_ifaddr *ifp;
2158 struct inet6_dev *idev;
2159 struct net_device *dev;
2161 if (plen > 128)
2162 return -EINVAL;
2164 dev = __dev_get_by_index(net, ifindex);
2165 if (!dev)
2166 return -ENODEV;
2168 if ((idev = __in6_dev_get(dev)) == NULL)
2169 return -ENXIO;
2171 read_lock_bh(&idev->lock);
2172 for (ifp = idev->addr_list; ifp; ifp=ifp->if_next) {
2173 if (ifp->prefix_len == plen &&
2174 ipv6_addr_equal(pfx, &ifp->addr)) {
2175 in6_ifa_hold(ifp);
2176 read_unlock_bh(&idev->lock);
2178 ipv6_del_addr(ifp);
2180 /* If the last address is deleted administratively,
2181 disable IPv6 on this interface.
2183 if (idev->addr_list == NULL)
2184 addrconf_ifdown(idev->dev, 1);
2185 return 0;
2188 read_unlock_bh(&idev->lock);
2189 return -EADDRNOTAVAIL;
2193 int addrconf_add_ifaddr(struct net *net, void __user *arg)
2195 struct in6_ifreq ireq;
2196 int err;
2198 if (!capable(CAP_NET_ADMIN))
2199 return -EPERM;
2201 if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2202 return -EFAULT;
2204 rtnl_lock();
2205 err = inet6_addr_add(net, ireq.ifr6_ifindex, &ireq.ifr6_addr,
2206 ireq.ifr6_prefixlen, IFA_F_PERMANENT,
2207 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
2208 rtnl_unlock();
2209 return err;
2212 int addrconf_del_ifaddr(struct net *net, void __user *arg)
2214 struct in6_ifreq ireq;
2215 int err;
2217 if (!capable(CAP_NET_ADMIN))
2218 return -EPERM;
2220 if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2221 return -EFAULT;
2223 rtnl_lock();
2224 err = inet6_addr_del(net, ireq.ifr6_ifindex, &ireq.ifr6_addr,
2225 ireq.ifr6_prefixlen);
2226 rtnl_unlock();
2227 return err;
2230 static void add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
2231 int plen, int scope)
2233 struct inet6_ifaddr *ifp;
2235 ifp = ipv6_add_addr(idev, addr, plen, scope, IFA_F_PERMANENT);
2236 if (!IS_ERR(ifp)) {
2237 spin_lock_bh(&ifp->lock);
2238 ifp->flags &= ~IFA_F_TENTATIVE;
2239 spin_unlock_bh(&ifp->lock);
2240 ipv6_ifa_notify(RTM_NEWADDR, ifp);
2241 in6_ifa_put(ifp);
2245 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2246 static void sit_add_v4_addrs(struct inet6_dev *idev)
2248 struct in6_addr addr;
2249 struct net_device *dev;
2250 struct net *net = dev_net(idev->dev);
2251 int scope;
2253 ASSERT_RTNL();
2255 memset(&addr, 0, sizeof(struct in6_addr));
2256 memcpy(&addr.s6_addr32[3], idev->dev->dev_addr, 4);
2258 if (idev->dev->flags&IFF_POINTOPOINT) {
2259 addr.s6_addr32[0] = htonl(0xfe800000);
2260 scope = IFA_LINK;
2261 } else {
2262 scope = IPV6_ADDR_COMPATv4;
2265 if (addr.s6_addr32[3]) {
2266 add_addr(idev, &addr, 128, scope);
2267 return;
2270 for_each_netdev(net, dev) {
2271 struct in_device * in_dev = __in_dev_get_rtnl(dev);
2272 if (in_dev && (dev->flags & IFF_UP)) {
2273 struct in_ifaddr * ifa;
2275 int flag = scope;
2277 for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) {
2278 int plen;
2280 addr.s6_addr32[3] = ifa->ifa_local;
2282 if (ifa->ifa_scope == RT_SCOPE_LINK)
2283 continue;
2284 if (ifa->ifa_scope >= RT_SCOPE_HOST) {
2285 if (idev->dev->flags&IFF_POINTOPOINT)
2286 continue;
2287 flag |= IFA_HOST;
2289 if (idev->dev->flags&IFF_POINTOPOINT)
2290 plen = 64;
2291 else
2292 plen = 96;
2294 add_addr(idev, &addr, plen, flag);
2299 #endif
2301 static void init_loopback(struct net_device *dev)
2303 struct inet6_dev *idev;
2305 /* ::1 */
2307 ASSERT_RTNL();
2309 if ((idev = ipv6_find_idev(dev)) == NULL) {
2310 printk(KERN_DEBUG "init loopback: add_dev failed\n");
2311 return;
2314 add_addr(idev, &in6addr_loopback, 128, IFA_HOST);
2317 static void addrconf_add_linklocal(struct inet6_dev *idev, struct in6_addr *addr)
2319 struct inet6_ifaddr * ifp;
2320 u32 addr_flags = IFA_F_PERMANENT;
2322 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2323 if (idev->cnf.optimistic_dad &&
2324 !dev_net(idev->dev)->ipv6.devconf_all->forwarding)
2325 addr_flags |= IFA_F_OPTIMISTIC;
2326 #endif
2329 ifp = ipv6_add_addr(idev, addr, 64, IFA_LINK, addr_flags);
2330 if (!IS_ERR(ifp)) {
2331 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, idev->dev, 0, 0);
2332 addrconf_dad_start(ifp, 0);
2333 in6_ifa_put(ifp);
2337 static void addrconf_dev_config(struct net_device *dev)
2339 struct in6_addr addr;
2340 struct inet6_dev * idev;
2342 ASSERT_RTNL();
2344 if ((dev->type != ARPHRD_ETHER) &&
2345 (dev->type != ARPHRD_FDDI) &&
2346 (dev->type != ARPHRD_IEEE802_TR) &&
2347 (dev->type != ARPHRD_ARCNET) &&
2348 (dev->type != ARPHRD_INFINIBAND)) {
2349 /* Alas, we support only Ethernet autoconfiguration. */
2350 return;
2353 idev = addrconf_add_dev(dev);
2354 if (idev == NULL)
2355 return;
2357 memset(&addr, 0, sizeof(struct in6_addr));
2358 addr.s6_addr32[0] = htonl(0xFE800000);
2360 if (ipv6_generate_eui64(addr.s6_addr + 8, dev) == 0)
2361 addrconf_add_linklocal(idev, &addr);
2364 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2365 static void addrconf_sit_config(struct net_device *dev)
2367 struct inet6_dev *idev;
2369 ASSERT_RTNL();
2372 * Configure the tunnel with one of our IPv4
2373 * addresses... we should configure all of
2374 * our v4 addrs in the tunnel
2377 if ((idev = ipv6_find_idev(dev)) == NULL) {
2378 printk(KERN_DEBUG "init sit: add_dev failed\n");
2379 return;
2382 if (dev->priv_flags & IFF_ISATAP) {
2383 struct in6_addr addr;
2385 ipv6_addr_set(&addr, htonl(0xFE800000), 0, 0, 0);
2386 addrconf_prefix_route(&addr, 64, dev, 0, 0);
2387 if (!ipv6_generate_eui64(addr.s6_addr + 8, dev))
2388 addrconf_add_linklocal(idev, &addr);
2389 return;
2392 sit_add_v4_addrs(idev);
2394 if (dev->flags&IFF_POINTOPOINT) {
2395 addrconf_add_mroute(dev);
2396 addrconf_add_lroute(dev);
2397 } else
2398 sit_route_add(dev);
2400 #endif
2402 static inline int
2403 ipv6_inherit_linklocal(struct inet6_dev *idev, struct net_device *link_dev)
2405 struct in6_addr lladdr;
2407 if (!ipv6_get_lladdr(link_dev, &lladdr, IFA_F_TENTATIVE)) {
2408 addrconf_add_linklocal(idev, &lladdr);
2409 return 0;
2411 return -1;
2414 static void ip6_tnl_add_linklocal(struct inet6_dev *idev)
2416 struct net_device *link_dev;
2417 struct net *net = dev_net(idev->dev);
2419 /* first try to inherit the link-local address from the link device */
2420 if (idev->dev->iflink &&
2421 (link_dev = __dev_get_by_index(net, idev->dev->iflink))) {
2422 if (!ipv6_inherit_linklocal(idev, link_dev))
2423 return;
2425 /* then try to inherit it from any device */
2426 for_each_netdev(net, link_dev) {
2427 if (!ipv6_inherit_linklocal(idev, link_dev))
2428 return;
2430 printk(KERN_DEBUG "init ip6-ip6: add_linklocal failed\n");
2434 * Autoconfigure tunnel with a link-local address so routing protocols,
2435 * DHCPv6, MLD etc. can be run over the virtual link
2438 static void addrconf_ip6_tnl_config(struct net_device *dev)
2440 struct inet6_dev *idev;
2442 ASSERT_RTNL();
2444 if ((idev = addrconf_add_dev(dev)) == NULL) {
2445 printk(KERN_DEBUG "init ip6-ip6: add_dev failed\n");
2446 return;
2448 ip6_tnl_add_linklocal(idev);
2451 static int addrconf_notify(struct notifier_block *this, unsigned long event,
2452 void * data)
2454 struct net_device *dev = (struct net_device *) data;
2455 struct inet6_dev *idev = __in6_dev_get(dev);
2456 int run_pending = 0;
2457 int err;
2459 switch(event) {
2460 case NETDEV_REGISTER:
2461 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2462 idev = ipv6_add_dev(dev);
2463 if (!idev)
2464 return notifier_from_errno(-ENOMEM);
2466 break;
2467 case NETDEV_UP:
2468 case NETDEV_CHANGE:
2469 if (dev->flags & IFF_SLAVE)
2470 break;
2472 if (event == NETDEV_UP) {
2473 if (!addrconf_qdisc_ok(dev)) {
2474 /* device is not ready yet. */
2475 printk(KERN_INFO
2476 "ADDRCONF(NETDEV_UP): %s: "
2477 "link is not ready\n",
2478 dev->name);
2479 break;
2482 if (!idev && dev->mtu >= IPV6_MIN_MTU)
2483 idev = ipv6_add_dev(dev);
2485 if (idev) {
2486 idev->if_flags |= IF_READY;
2487 run_pending = 1;
2489 } else {
2490 if (!addrconf_qdisc_ok(dev)) {
2491 /* device is still not ready. */
2492 break;
2495 if (idev) {
2496 if (idev->if_flags & IF_READY) {
2497 /* device is already configured. */
2498 break;
2500 idev->if_flags |= IF_READY;
2503 printk(KERN_INFO
2504 "ADDRCONF(NETDEV_CHANGE): %s: "
2505 "link becomes ready\n",
2506 dev->name);
2508 run_pending = 1;
2511 switch(dev->type) {
2512 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2513 case ARPHRD_SIT:
2514 addrconf_sit_config(dev);
2515 break;
2516 #endif
2517 case ARPHRD_TUNNEL6:
2518 addrconf_ip6_tnl_config(dev);
2519 break;
2520 case ARPHRD_LOOPBACK:
2521 init_loopback(dev);
2522 break;
2524 default:
2525 addrconf_dev_config(dev);
2526 break;
2528 if (idev) {
2529 if (run_pending)
2530 addrconf_dad_run(idev);
2532 /* If the MTU changed during the interface down, when the
2533 interface up, the changed MTU must be reflected in the
2534 idev as well as routers.
2536 if (idev->cnf.mtu6 != dev->mtu && dev->mtu >= IPV6_MIN_MTU) {
2537 rt6_mtu_change(dev, dev->mtu);
2538 idev->cnf.mtu6 = dev->mtu;
2540 idev->tstamp = jiffies;
2541 inet6_ifinfo_notify(RTM_NEWLINK, idev);
2542 /* If the changed mtu during down is lower than IPV6_MIN_MTU
2543 stop IPv6 on this interface.
2545 if (dev->mtu < IPV6_MIN_MTU)
2546 addrconf_ifdown(dev, event != NETDEV_DOWN);
2548 break;
2550 case NETDEV_CHANGEMTU:
2551 if (idev && dev->mtu >= IPV6_MIN_MTU) {
2552 rt6_mtu_change(dev, dev->mtu);
2553 idev->cnf.mtu6 = dev->mtu;
2554 break;
2557 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2558 idev = ipv6_add_dev(dev);
2559 if (idev)
2560 break;
2563 /* MTU falled under IPV6_MIN_MTU. Stop IPv6 on this interface. */
2565 case NETDEV_DOWN:
2566 case NETDEV_UNREGISTER:
2568 * Remove all addresses from this interface.
2570 addrconf_ifdown(dev, event != NETDEV_DOWN);
2571 break;
2573 case NETDEV_CHANGENAME:
2574 if (idev) {
2575 snmp6_unregister_dev(idev);
2576 addrconf_sysctl_unregister(idev);
2577 addrconf_sysctl_register(idev);
2578 err = snmp6_register_dev(idev);
2579 if (err)
2580 return notifier_from_errno(err);
2582 break;
2583 case NETDEV_BONDING_OLDTYPE:
2584 case NETDEV_BONDING_NEWTYPE:
2585 addrconf_bonding_change(dev, event);
2586 break;
2589 return NOTIFY_OK;
2593 * addrconf module should be notified of a device going up
2595 static struct notifier_block ipv6_dev_notf = {
2596 .notifier_call = addrconf_notify,
2597 .priority = 0
2600 static void addrconf_bonding_change(struct net_device *dev, unsigned long event)
2602 struct inet6_dev *idev;
2603 ASSERT_RTNL();
2605 idev = __in6_dev_get(dev);
2607 if (event == NETDEV_BONDING_NEWTYPE)
2608 ipv6_mc_remap(idev);
2609 else if (event == NETDEV_BONDING_OLDTYPE)
2610 ipv6_mc_unmap(idev);
2613 static int addrconf_ifdown(struct net_device *dev, int how)
2615 struct inet6_dev *idev;
2616 struct inet6_ifaddr *ifa, **bifa;
2617 struct net *net = dev_net(dev);
2618 int i;
2620 ASSERT_RTNL();
2622 rt6_ifdown(net, dev);
2623 neigh_ifdown(&nd_tbl, dev);
2625 idev = __in6_dev_get(dev);
2626 if (idev == NULL)
2627 return -ENODEV;
2629 /* Step 1: remove reference to ipv6 device from parent device.
2630 Do not dev_put!
2632 if (how) {
2633 idev->dead = 1;
2635 /* protected by rtnl_lock */
2636 rcu_assign_pointer(dev->ip6_ptr, NULL);
2638 /* Step 1.5: remove snmp6 entry */
2639 snmp6_unregister_dev(idev);
2643 /* Step 2: clear hash table */
2644 for (i=0; i<IN6_ADDR_HSIZE; i++) {
2645 bifa = &inet6_addr_lst[i];
2647 write_lock_bh(&addrconf_hash_lock);
2648 while ((ifa = *bifa) != NULL) {
2649 if (ifa->idev == idev) {
2650 *bifa = ifa->lst_next;
2651 ifa->lst_next = NULL;
2652 addrconf_del_timer(ifa);
2653 in6_ifa_put(ifa);
2654 continue;
2656 bifa = &ifa->lst_next;
2658 write_unlock_bh(&addrconf_hash_lock);
2661 write_lock_bh(&idev->lock);
2663 /* Step 3: clear flags for stateless addrconf */
2664 if (!how)
2665 idev->if_flags &= ~(IF_RS_SENT|IF_RA_RCVD|IF_READY);
2667 /* Step 4: clear address list */
2668 #ifdef CONFIG_IPV6_PRIVACY
2669 if (how && del_timer(&idev->regen_timer))
2670 in6_dev_put(idev);
2672 /* clear tempaddr list */
2673 while ((ifa = idev->tempaddr_list) != NULL) {
2674 idev->tempaddr_list = ifa->tmp_next;
2675 ifa->tmp_next = NULL;
2676 ifa->dead = 1;
2677 write_unlock_bh(&idev->lock);
2678 spin_lock_bh(&ifa->lock);
2680 if (ifa->ifpub) {
2681 in6_ifa_put(ifa->ifpub);
2682 ifa->ifpub = NULL;
2684 spin_unlock_bh(&ifa->lock);
2685 in6_ifa_put(ifa);
2686 write_lock_bh(&idev->lock);
2688 #endif
2689 while ((ifa = idev->addr_list) != NULL) {
2690 idev->addr_list = ifa->if_next;
2691 ifa->if_next = NULL;
2692 ifa->dead = 1;
2693 addrconf_del_timer(ifa);
2694 write_unlock_bh(&idev->lock);
2696 __ipv6_ifa_notify(RTM_DELADDR, ifa);
2697 atomic_notifier_call_chain(&inet6addr_chain, NETDEV_DOWN, ifa);
2698 in6_ifa_put(ifa);
2700 write_lock_bh(&idev->lock);
2702 write_unlock_bh(&idev->lock);
2704 /* Step 5: Discard multicast list */
2706 if (how)
2707 ipv6_mc_destroy_dev(idev);
2708 else
2709 ipv6_mc_down(idev);
2711 idev->tstamp = jiffies;
2713 /* Shot the device (if unregistered) */
2715 if (how) {
2716 addrconf_sysctl_unregister(idev);
2717 neigh_parms_release(&nd_tbl, idev->nd_parms);
2718 neigh_ifdown(&nd_tbl, dev);
2719 in6_dev_put(idev);
2721 return 0;
2724 static void addrconf_rs_timer(unsigned long data)
2726 struct inet6_ifaddr *ifp = (struct inet6_ifaddr *) data;
2728 if (ifp->idev->cnf.forwarding)
2729 goto out;
2731 if (ifp->idev->if_flags & IF_RA_RCVD) {
2733 * Announcement received after solicitation
2734 * was sent
2736 goto out;
2739 spin_lock(&ifp->lock);
2740 if (ifp->probes++ < ifp->idev->cnf.rtr_solicits) {
2741 /* The wait after the last probe can be shorter */
2742 addrconf_mod_timer(ifp, AC_RS,
2743 (ifp->probes == ifp->idev->cnf.rtr_solicits) ?
2744 ifp->idev->cnf.rtr_solicit_delay :
2745 ifp->idev->cnf.rtr_solicit_interval);
2746 spin_unlock(&ifp->lock);
2748 ndisc_send_rs(ifp->idev->dev, &ifp->addr, &in6addr_linklocal_allrouters);
2749 } else {
2750 spin_unlock(&ifp->lock);
2752 * Note: we do not support deprecated "all on-link"
2753 * assumption any longer.
2755 printk(KERN_DEBUG "%s: no IPv6 routers present\n",
2756 ifp->idev->dev->name);
2759 out:
2760 in6_ifa_put(ifp);
2764 * Duplicate Address Detection
2766 static void addrconf_dad_kick(struct inet6_ifaddr *ifp)
2768 unsigned long rand_num;
2769 struct inet6_dev *idev = ifp->idev;
2771 if (ifp->flags & IFA_F_OPTIMISTIC)
2772 rand_num = 0;
2773 else
2774 rand_num = net_random() % (idev->cnf.rtr_solicit_delay ? : 1);
2776 ifp->probes = idev->cnf.dad_transmits;
2777 addrconf_mod_timer(ifp, AC_DAD, rand_num);
2780 static void addrconf_dad_start(struct inet6_ifaddr *ifp, u32 flags)
2782 struct inet6_dev *idev = ifp->idev;
2783 struct net_device *dev = idev->dev;
2785 addrconf_join_solict(dev, &ifp->addr);
2787 net_srandom(ifp->addr.s6_addr32[3]);
2789 read_lock_bh(&idev->lock);
2790 if (ifp->dead)
2791 goto out;
2792 spin_lock_bh(&ifp->lock);
2794 if (dev->flags&(IFF_NOARP|IFF_LOOPBACK) ||
2795 idev->cnf.accept_dad < 1 ||
2796 !(ifp->flags&IFA_F_TENTATIVE) ||
2797 ifp->flags & IFA_F_NODAD) {
2798 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
2799 spin_unlock_bh(&ifp->lock);
2800 read_unlock_bh(&idev->lock);
2802 addrconf_dad_completed(ifp);
2803 return;
2806 if (!(idev->if_flags & IF_READY)) {
2807 spin_unlock_bh(&ifp->lock);
2808 read_unlock_bh(&idev->lock);
2810 * If the device is not ready:
2811 * - keep it tentative if it is a permanent address.
2812 * - otherwise, kill it.
2814 in6_ifa_hold(ifp);
2815 addrconf_dad_stop(ifp, 0);
2816 return;
2820 * Optimistic nodes can start receiving
2821 * Frames right away
2823 if(ifp->flags & IFA_F_OPTIMISTIC)
2824 ip6_ins_rt(ifp->rt);
2826 addrconf_dad_kick(ifp);
2827 spin_unlock_bh(&ifp->lock);
2828 out:
2829 read_unlock_bh(&idev->lock);
2832 static void addrconf_dad_timer(unsigned long data)
2834 struct inet6_ifaddr *ifp = (struct inet6_ifaddr *) data;
2835 struct inet6_dev *idev = ifp->idev;
2836 struct in6_addr mcaddr;
2838 read_lock_bh(&idev->lock);
2839 if (idev->dead) {
2840 read_unlock_bh(&idev->lock);
2841 goto out;
2843 spin_lock_bh(&ifp->lock);
2844 if (ifp->probes == 0) {
2846 * DAD was successful
2849 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
2850 spin_unlock_bh(&ifp->lock);
2851 read_unlock_bh(&idev->lock);
2853 addrconf_dad_completed(ifp);
2855 goto out;
2858 ifp->probes--;
2859 addrconf_mod_timer(ifp, AC_DAD, ifp->idev->nd_parms->retrans_time);
2860 spin_unlock_bh(&ifp->lock);
2861 read_unlock_bh(&idev->lock);
2863 /* send a neighbour solicitation for our addr */
2864 addrconf_addr_solict_mult(&ifp->addr, &mcaddr);
2865 ndisc_send_ns(ifp->idev->dev, NULL, &ifp->addr, &mcaddr, &in6addr_any);
2866 out:
2867 in6_ifa_put(ifp);
2870 static void addrconf_dad_completed(struct inet6_ifaddr *ifp)
2872 struct net_device * dev = ifp->idev->dev;
2875 * Configure the address for reception. Now it is valid.
2878 ipv6_ifa_notify(RTM_NEWADDR, ifp);
2880 /* If added prefix is link local and forwarding is off,
2881 start sending router solicitations.
2884 if (ifp->idev->cnf.forwarding == 0 &&
2885 ifp->idev->cnf.rtr_solicits > 0 &&
2886 (dev->flags&IFF_LOOPBACK) == 0 &&
2887 (ipv6_addr_type(&ifp->addr) & IPV6_ADDR_LINKLOCAL)) {
2889 * If a host as already performed a random delay
2890 * [...] as part of DAD [...] there is no need
2891 * to delay again before sending the first RS
2893 ndisc_send_rs(ifp->idev->dev, &ifp->addr, &in6addr_linklocal_allrouters);
2895 spin_lock_bh(&ifp->lock);
2896 ifp->probes = 1;
2897 ifp->idev->if_flags |= IF_RS_SENT;
2898 addrconf_mod_timer(ifp, AC_RS, ifp->idev->cnf.rtr_solicit_interval);
2899 spin_unlock_bh(&ifp->lock);
2903 static void addrconf_dad_run(struct inet6_dev *idev) {
2904 struct inet6_ifaddr *ifp;
2906 read_lock_bh(&idev->lock);
2907 for (ifp = idev->addr_list; ifp; ifp = ifp->if_next) {
2908 spin_lock_bh(&ifp->lock);
2909 if (!(ifp->flags & IFA_F_TENTATIVE)) {
2910 spin_unlock_bh(&ifp->lock);
2911 continue;
2913 spin_unlock_bh(&ifp->lock);
2914 addrconf_dad_kick(ifp);
2916 read_unlock_bh(&idev->lock);
2919 #ifdef CONFIG_PROC_FS
2920 struct if6_iter_state {
2921 struct seq_net_private p;
2922 int bucket;
2925 static struct inet6_ifaddr *if6_get_first(struct seq_file *seq)
2927 struct inet6_ifaddr *ifa = NULL;
2928 struct if6_iter_state *state = seq->private;
2929 struct net *net = seq_file_net(seq);
2931 for (state->bucket = 0; state->bucket < IN6_ADDR_HSIZE; ++state->bucket) {
2932 ifa = inet6_addr_lst[state->bucket];
2934 while (ifa && !net_eq(dev_net(ifa->idev->dev), net))
2935 ifa = ifa->lst_next;
2936 if (ifa)
2937 break;
2939 return ifa;
2942 static struct inet6_ifaddr *if6_get_next(struct seq_file *seq, struct inet6_ifaddr *ifa)
2944 struct if6_iter_state *state = seq->private;
2945 struct net *net = seq_file_net(seq);
2947 ifa = ifa->lst_next;
2948 try_again:
2949 if (ifa) {
2950 if (!net_eq(dev_net(ifa->idev->dev), net)) {
2951 ifa = ifa->lst_next;
2952 goto try_again;
2956 if (!ifa && ++state->bucket < IN6_ADDR_HSIZE) {
2957 ifa = inet6_addr_lst[state->bucket];
2958 goto try_again;
2961 return ifa;
2964 static struct inet6_ifaddr *if6_get_idx(struct seq_file *seq, loff_t pos)
2966 struct inet6_ifaddr *ifa = if6_get_first(seq);
2968 if (ifa)
2969 while(pos && (ifa = if6_get_next(seq, ifa)) != NULL)
2970 --pos;
2971 return pos ? NULL : ifa;
2974 static void *if6_seq_start(struct seq_file *seq, loff_t *pos)
2975 __acquires(addrconf_hash_lock)
2977 read_lock_bh(&addrconf_hash_lock);
2978 return if6_get_idx(seq, *pos);
2981 static void *if6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
2983 struct inet6_ifaddr *ifa;
2985 ifa = if6_get_next(seq, v);
2986 ++*pos;
2987 return ifa;
2990 static void if6_seq_stop(struct seq_file *seq, void *v)
2991 __releases(addrconf_hash_lock)
2993 read_unlock_bh(&addrconf_hash_lock);
2996 static int if6_seq_show(struct seq_file *seq, void *v)
2998 struct inet6_ifaddr *ifp = (struct inet6_ifaddr *)v;
2999 seq_printf(seq, "%pi6 %02x %02x %02x %02x %8s\n",
3000 &ifp->addr,
3001 ifp->idev->dev->ifindex,
3002 ifp->prefix_len,
3003 ifp->scope,
3004 ifp->flags,
3005 ifp->idev->dev->name);
3006 return 0;
3009 static const struct seq_operations if6_seq_ops = {
3010 .start = if6_seq_start,
3011 .next = if6_seq_next,
3012 .show = if6_seq_show,
3013 .stop = if6_seq_stop,
3016 static int if6_seq_open(struct inode *inode, struct file *file)
3018 return seq_open_net(inode, file, &if6_seq_ops,
3019 sizeof(struct if6_iter_state));
3022 static const struct file_operations if6_fops = {
3023 .owner = THIS_MODULE,
3024 .open = if6_seq_open,
3025 .read = seq_read,
3026 .llseek = seq_lseek,
3027 .release = seq_release_net,
3030 static int if6_proc_net_init(struct net *net)
3032 if (!proc_net_fops_create(net, "if_inet6", S_IRUGO, &if6_fops))
3033 return -ENOMEM;
3034 return 0;
3037 static void if6_proc_net_exit(struct net *net)
3039 proc_net_remove(net, "if_inet6");
3042 static struct pernet_operations if6_proc_net_ops = {
3043 .init = if6_proc_net_init,
3044 .exit = if6_proc_net_exit,
3047 int __init if6_proc_init(void)
3049 return register_pernet_subsys(&if6_proc_net_ops);
3052 void if6_proc_exit(void)
3054 unregister_pernet_subsys(&if6_proc_net_ops);
3056 #endif /* CONFIG_PROC_FS */
3058 #if defined(CONFIG_IPV6_MIP6) || defined(CONFIG_IPV6_MIP6_MODULE)
3059 /* Check if address is a home address configured on any interface. */
3060 int ipv6_chk_home_addr(struct net *net, struct in6_addr *addr)
3062 int ret = 0;
3063 struct inet6_ifaddr * ifp;
3064 u8 hash = ipv6_addr_hash(addr);
3065 read_lock_bh(&addrconf_hash_lock);
3066 for (ifp = inet6_addr_lst[hash]; ifp; ifp = ifp->lst_next) {
3067 if (!net_eq(dev_net(ifp->idev->dev), net))
3068 continue;
3069 if (ipv6_addr_equal(&ifp->addr, addr) &&
3070 (ifp->flags & IFA_F_HOMEADDRESS)) {
3071 ret = 1;
3072 break;
3075 read_unlock_bh(&addrconf_hash_lock);
3076 return ret;
3078 #endif
3081 * Periodic address status verification
3084 static void addrconf_verify(unsigned long foo)
3086 struct inet6_ifaddr *ifp;
3087 unsigned long now, next;
3088 int i;
3090 spin_lock_bh(&addrconf_verify_lock);
3091 now = jiffies;
3092 next = now + ADDR_CHECK_FREQUENCY;
3094 del_timer(&addr_chk_timer);
3096 for (i=0; i < IN6_ADDR_HSIZE; i++) {
3098 restart:
3099 read_lock(&addrconf_hash_lock);
3100 for (ifp=inet6_addr_lst[i]; ifp; ifp=ifp->lst_next) {
3101 unsigned long age;
3102 #ifdef CONFIG_IPV6_PRIVACY
3103 unsigned long regen_advance;
3104 #endif
3106 if (ifp->flags & IFA_F_PERMANENT)
3107 continue;
3109 spin_lock(&ifp->lock);
3110 age = (now - ifp->tstamp) / HZ;
3112 #ifdef CONFIG_IPV6_PRIVACY
3113 regen_advance = ifp->idev->cnf.regen_max_retry *
3114 ifp->idev->cnf.dad_transmits *
3115 ifp->idev->nd_parms->retrans_time / HZ;
3116 #endif
3118 if (ifp->valid_lft != INFINITY_LIFE_TIME &&
3119 age >= ifp->valid_lft) {
3120 spin_unlock(&ifp->lock);
3121 in6_ifa_hold(ifp);
3122 read_unlock(&addrconf_hash_lock);
3123 ipv6_del_addr(ifp);
3124 goto restart;
3125 } else if (ifp->prefered_lft == INFINITY_LIFE_TIME) {
3126 spin_unlock(&ifp->lock);
3127 continue;
3128 } else if (age >= ifp->prefered_lft) {
3129 /* jiffies - ifp->tstamp > age >= ifp->prefered_lft */
3130 int deprecate = 0;
3132 if (!(ifp->flags&IFA_F_DEPRECATED)) {
3133 deprecate = 1;
3134 ifp->flags |= IFA_F_DEPRECATED;
3137 if (time_before(ifp->tstamp + ifp->valid_lft * HZ, next))
3138 next = ifp->tstamp + ifp->valid_lft * HZ;
3140 spin_unlock(&ifp->lock);
3142 if (deprecate) {
3143 in6_ifa_hold(ifp);
3144 read_unlock(&addrconf_hash_lock);
3146 ipv6_ifa_notify(0, ifp);
3147 in6_ifa_put(ifp);
3148 goto restart;
3150 #ifdef CONFIG_IPV6_PRIVACY
3151 } else if ((ifp->flags&IFA_F_TEMPORARY) &&
3152 !(ifp->flags&IFA_F_TENTATIVE)) {
3153 if (age >= ifp->prefered_lft - regen_advance) {
3154 struct inet6_ifaddr *ifpub = ifp->ifpub;
3155 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3156 next = ifp->tstamp + ifp->prefered_lft * HZ;
3157 if (!ifp->regen_count && ifpub) {
3158 ifp->regen_count++;
3159 in6_ifa_hold(ifp);
3160 in6_ifa_hold(ifpub);
3161 spin_unlock(&ifp->lock);
3162 read_unlock(&addrconf_hash_lock);
3163 spin_lock(&ifpub->lock);
3164 ifpub->regen_count = 0;
3165 spin_unlock(&ifpub->lock);
3166 ipv6_create_tempaddr(ifpub, ifp);
3167 in6_ifa_put(ifpub);
3168 in6_ifa_put(ifp);
3169 goto restart;
3171 } else if (time_before(ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ, next))
3172 next = ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ;
3173 spin_unlock(&ifp->lock);
3174 #endif
3175 } else {
3176 /* ifp->prefered_lft <= ifp->valid_lft */
3177 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3178 next = ifp->tstamp + ifp->prefered_lft * HZ;
3179 spin_unlock(&ifp->lock);
3182 read_unlock(&addrconf_hash_lock);
3185 addr_chk_timer.expires = time_before(next, jiffies + HZ) ? jiffies + HZ : next;
3186 add_timer(&addr_chk_timer);
3187 spin_unlock_bh(&addrconf_verify_lock);
3190 static struct in6_addr *extract_addr(struct nlattr *addr, struct nlattr *local)
3192 struct in6_addr *pfx = NULL;
3194 if (addr)
3195 pfx = nla_data(addr);
3197 if (local) {
3198 if (pfx && nla_memcmp(local, pfx, sizeof(*pfx)))
3199 pfx = NULL;
3200 else
3201 pfx = nla_data(local);
3204 return pfx;
3207 static const struct nla_policy ifa_ipv6_policy[IFA_MAX+1] = {
3208 [IFA_ADDRESS] = { .len = sizeof(struct in6_addr) },
3209 [IFA_LOCAL] = { .len = sizeof(struct in6_addr) },
3210 [IFA_CACHEINFO] = { .len = sizeof(struct ifa_cacheinfo) },
3213 static int
3214 inet6_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
3216 struct net *net = sock_net(skb->sk);
3217 struct ifaddrmsg *ifm;
3218 struct nlattr *tb[IFA_MAX+1];
3219 struct in6_addr *pfx;
3220 int err;
3222 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3223 if (err < 0)
3224 return err;
3226 ifm = nlmsg_data(nlh);
3227 pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3228 if (pfx == NULL)
3229 return -EINVAL;
3231 return inet6_addr_del(net, ifm->ifa_index, pfx, ifm->ifa_prefixlen);
3234 static int inet6_addr_modify(struct inet6_ifaddr *ifp, u8 ifa_flags,
3235 u32 prefered_lft, u32 valid_lft)
3237 u32 flags;
3238 clock_t expires;
3239 unsigned long timeout;
3241 if (!valid_lft || (prefered_lft > valid_lft))
3242 return -EINVAL;
3244 timeout = addrconf_timeout_fixup(valid_lft, HZ);
3245 if (addrconf_finite_timeout(timeout)) {
3246 expires = jiffies_to_clock_t(timeout * HZ);
3247 valid_lft = timeout;
3248 flags = RTF_EXPIRES;
3249 } else {
3250 expires = 0;
3251 flags = 0;
3252 ifa_flags |= IFA_F_PERMANENT;
3255 timeout = addrconf_timeout_fixup(prefered_lft, HZ);
3256 if (addrconf_finite_timeout(timeout)) {
3257 if (timeout == 0)
3258 ifa_flags |= IFA_F_DEPRECATED;
3259 prefered_lft = timeout;
3262 spin_lock_bh(&ifp->lock);
3263 ifp->flags = (ifp->flags & ~(IFA_F_DEPRECATED | IFA_F_PERMANENT | IFA_F_NODAD | IFA_F_HOMEADDRESS)) | ifa_flags;
3264 ifp->tstamp = jiffies;
3265 ifp->valid_lft = valid_lft;
3266 ifp->prefered_lft = prefered_lft;
3268 spin_unlock_bh(&ifp->lock);
3269 if (!(ifp->flags&IFA_F_TENTATIVE))
3270 ipv6_ifa_notify(0, ifp);
3272 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, ifp->idev->dev,
3273 expires, flags);
3274 addrconf_verify(0);
3276 return 0;
3279 static int
3280 inet6_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
3282 struct net *net = sock_net(skb->sk);
3283 struct ifaddrmsg *ifm;
3284 struct nlattr *tb[IFA_MAX+1];
3285 struct in6_addr *pfx;
3286 struct inet6_ifaddr *ifa;
3287 struct net_device *dev;
3288 u32 valid_lft = INFINITY_LIFE_TIME, preferred_lft = INFINITY_LIFE_TIME;
3289 u8 ifa_flags;
3290 int err;
3292 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3293 if (err < 0)
3294 return err;
3296 ifm = nlmsg_data(nlh);
3297 pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3298 if (pfx == NULL)
3299 return -EINVAL;
3301 if (tb[IFA_CACHEINFO]) {
3302 struct ifa_cacheinfo *ci;
3304 ci = nla_data(tb[IFA_CACHEINFO]);
3305 valid_lft = ci->ifa_valid;
3306 preferred_lft = ci->ifa_prefered;
3307 } else {
3308 preferred_lft = INFINITY_LIFE_TIME;
3309 valid_lft = INFINITY_LIFE_TIME;
3312 dev = __dev_get_by_index(net, ifm->ifa_index);
3313 if (dev == NULL)
3314 return -ENODEV;
3316 /* We ignore other flags so far. */
3317 ifa_flags = ifm->ifa_flags & (IFA_F_NODAD | IFA_F_HOMEADDRESS);
3319 ifa = ipv6_get_ifaddr(net, pfx, dev, 1);
3320 if (ifa == NULL) {
3322 * It would be best to check for !NLM_F_CREATE here but
3323 * userspace alreay relies on not having to provide this.
3325 return inet6_addr_add(net, ifm->ifa_index, pfx,
3326 ifm->ifa_prefixlen, ifa_flags,
3327 preferred_lft, valid_lft);
3330 if (nlh->nlmsg_flags & NLM_F_EXCL ||
3331 !(nlh->nlmsg_flags & NLM_F_REPLACE))
3332 err = -EEXIST;
3333 else
3334 err = inet6_addr_modify(ifa, ifa_flags, preferred_lft, valid_lft);
3336 in6_ifa_put(ifa);
3338 return err;
3341 static void put_ifaddrmsg(struct nlmsghdr *nlh, u8 prefixlen, u8 flags,
3342 u8 scope, int ifindex)
3344 struct ifaddrmsg *ifm;
3346 ifm = nlmsg_data(nlh);
3347 ifm->ifa_family = AF_INET6;
3348 ifm->ifa_prefixlen = prefixlen;
3349 ifm->ifa_flags = flags;
3350 ifm->ifa_scope = scope;
3351 ifm->ifa_index = ifindex;
3354 static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
3355 unsigned long tstamp, u32 preferred, u32 valid)
3357 struct ifa_cacheinfo ci;
3359 ci.cstamp = (u32)(TIME_DELTA(cstamp, INITIAL_JIFFIES) / HZ * 100
3360 + TIME_DELTA(cstamp, INITIAL_JIFFIES) % HZ * 100 / HZ);
3361 ci.tstamp = (u32)(TIME_DELTA(tstamp, INITIAL_JIFFIES) / HZ * 100
3362 + TIME_DELTA(tstamp, INITIAL_JIFFIES) % HZ * 100 / HZ);
3363 ci.ifa_prefered = preferred;
3364 ci.ifa_valid = valid;
3366 return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
3369 static inline int rt_scope(int ifa_scope)
3371 if (ifa_scope & IFA_HOST)
3372 return RT_SCOPE_HOST;
3373 else if (ifa_scope & IFA_LINK)
3374 return RT_SCOPE_LINK;
3375 else if (ifa_scope & IFA_SITE)
3376 return RT_SCOPE_SITE;
3377 else
3378 return RT_SCOPE_UNIVERSE;
3381 static inline int inet6_ifaddr_msgsize(void)
3383 return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
3384 + nla_total_size(16) /* IFA_ADDRESS */
3385 + nla_total_size(sizeof(struct ifa_cacheinfo));
3388 static int inet6_fill_ifaddr(struct sk_buff *skb, struct inet6_ifaddr *ifa,
3389 u32 pid, u32 seq, int event, unsigned int flags)
3391 struct nlmsghdr *nlh;
3392 u32 preferred, valid;
3394 nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3395 if (nlh == NULL)
3396 return -EMSGSIZE;
3398 put_ifaddrmsg(nlh, ifa->prefix_len, ifa->flags, rt_scope(ifa->scope),
3399 ifa->idev->dev->ifindex);
3401 if (!(ifa->flags&IFA_F_PERMANENT)) {
3402 preferred = ifa->prefered_lft;
3403 valid = ifa->valid_lft;
3404 if (preferred != INFINITY_LIFE_TIME) {
3405 long tval = (jiffies - ifa->tstamp)/HZ;
3406 if (preferred > tval)
3407 preferred -= tval;
3408 else
3409 preferred = 0;
3410 if (valid != INFINITY_LIFE_TIME)
3411 valid -= tval;
3413 } else {
3414 preferred = INFINITY_LIFE_TIME;
3415 valid = INFINITY_LIFE_TIME;
3418 if (nla_put(skb, IFA_ADDRESS, 16, &ifa->addr) < 0 ||
3419 put_cacheinfo(skb, ifa->cstamp, ifa->tstamp, preferred, valid) < 0) {
3420 nlmsg_cancel(skb, nlh);
3421 return -EMSGSIZE;
3424 return nlmsg_end(skb, nlh);
3427 static int inet6_fill_ifmcaddr(struct sk_buff *skb, struct ifmcaddr6 *ifmca,
3428 u32 pid, u32 seq, int event, u16 flags)
3430 struct nlmsghdr *nlh;
3431 u8 scope = RT_SCOPE_UNIVERSE;
3432 int ifindex = ifmca->idev->dev->ifindex;
3434 if (ipv6_addr_scope(&ifmca->mca_addr) & IFA_SITE)
3435 scope = RT_SCOPE_SITE;
3437 nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3438 if (nlh == NULL)
3439 return -EMSGSIZE;
3441 put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3442 if (nla_put(skb, IFA_MULTICAST, 16, &ifmca->mca_addr) < 0 ||
3443 put_cacheinfo(skb, ifmca->mca_cstamp, ifmca->mca_tstamp,
3444 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3445 nlmsg_cancel(skb, nlh);
3446 return -EMSGSIZE;
3449 return nlmsg_end(skb, nlh);
3452 static int inet6_fill_ifacaddr(struct sk_buff *skb, struct ifacaddr6 *ifaca,
3453 u32 pid, u32 seq, int event, unsigned int flags)
3455 struct nlmsghdr *nlh;
3456 u8 scope = RT_SCOPE_UNIVERSE;
3457 int ifindex = ifaca->aca_idev->dev->ifindex;
3459 if (ipv6_addr_scope(&ifaca->aca_addr) & IFA_SITE)
3460 scope = RT_SCOPE_SITE;
3462 nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3463 if (nlh == NULL)
3464 return -EMSGSIZE;
3466 put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3467 if (nla_put(skb, IFA_ANYCAST, 16, &ifaca->aca_addr) < 0 ||
3468 put_cacheinfo(skb, ifaca->aca_cstamp, ifaca->aca_tstamp,
3469 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3470 nlmsg_cancel(skb, nlh);
3471 return -EMSGSIZE;
3474 return nlmsg_end(skb, nlh);
3477 enum addr_type_t
3479 UNICAST_ADDR,
3480 MULTICAST_ADDR,
3481 ANYCAST_ADDR,
3484 /* called with rcu_read_lock() */
3485 static int in6_dump_addrs(struct inet6_dev *idev, struct sk_buff *skb,
3486 struct netlink_callback *cb, enum addr_type_t type,
3487 int s_ip_idx, int *p_ip_idx)
3489 struct inet6_ifaddr *ifa;
3490 struct ifmcaddr6 *ifmca;
3491 struct ifacaddr6 *ifaca;
3492 int err = 1;
3493 int ip_idx = *p_ip_idx;
3495 read_lock_bh(&idev->lock);
3496 switch (type) {
3497 case UNICAST_ADDR:
3498 /* unicast address incl. temp addr */
3499 for (ifa = idev->addr_list; ifa;
3500 ifa = ifa->if_next, ip_idx++) {
3501 if (ip_idx < s_ip_idx)
3502 continue;
3503 err = inet6_fill_ifaddr(skb, ifa,
3504 NETLINK_CB(cb->skb).pid,
3505 cb->nlh->nlmsg_seq,
3506 RTM_NEWADDR,
3507 NLM_F_MULTI);
3508 if (err <= 0)
3509 break;
3511 break;
3512 case MULTICAST_ADDR:
3513 /* multicast address */
3514 for (ifmca = idev->mc_list; ifmca;
3515 ifmca = ifmca->next, ip_idx++) {
3516 if (ip_idx < s_ip_idx)
3517 continue;
3518 err = inet6_fill_ifmcaddr(skb, ifmca,
3519 NETLINK_CB(cb->skb).pid,
3520 cb->nlh->nlmsg_seq,
3521 RTM_GETMULTICAST,
3522 NLM_F_MULTI);
3523 if (err <= 0)
3524 break;
3526 break;
3527 case ANYCAST_ADDR:
3528 /* anycast address */
3529 for (ifaca = idev->ac_list; ifaca;
3530 ifaca = ifaca->aca_next, ip_idx++) {
3531 if (ip_idx < s_ip_idx)
3532 continue;
3533 err = inet6_fill_ifacaddr(skb, ifaca,
3534 NETLINK_CB(cb->skb).pid,
3535 cb->nlh->nlmsg_seq,
3536 RTM_GETANYCAST,
3537 NLM_F_MULTI);
3538 if (err <= 0)
3539 break;
3541 break;
3542 default:
3543 break;
3545 read_unlock_bh(&idev->lock);
3546 *p_ip_idx = ip_idx;
3547 return err;
3550 static int inet6_dump_addr(struct sk_buff *skb, struct netlink_callback *cb,
3551 enum addr_type_t type)
3553 struct net *net = sock_net(skb->sk);
3554 int h, s_h;
3555 int idx, ip_idx;
3556 int s_idx, s_ip_idx;
3557 struct net_device *dev;
3558 struct inet6_dev *idev;
3559 struct hlist_head *head;
3560 struct hlist_node *node;
3562 s_h = cb->args[0];
3563 s_idx = idx = cb->args[1];
3564 s_ip_idx = ip_idx = cb->args[2];
3566 rcu_read_lock();
3567 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
3568 idx = 0;
3569 head = &net->dev_index_head[h];
3570 hlist_for_each_entry_rcu(dev, node, head, index_hlist) {
3571 if (idx < s_idx)
3572 goto cont;
3573 if (idx > s_idx)
3574 s_ip_idx = 0;
3575 ip_idx = 0;
3576 if ((idev = __in6_dev_get(dev)) == NULL)
3577 goto cont;
3579 if (in6_dump_addrs(idev, skb, cb, type,
3580 s_ip_idx, &ip_idx) <= 0)
3581 goto done;
3582 cont:
3583 idx++;
3586 done:
3587 rcu_read_unlock();
3588 cb->args[0] = h;
3589 cb->args[1] = idx;
3590 cb->args[2] = ip_idx;
3592 return skb->len;
3595 static int inet6_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
3597 enum addr_type_t type = UNICAST_ADDR;
3599 return inet6_dump_addr(skb, cb, type);
3602 static int inet6_dump_ifmcaddr(struct sk_buff *skb, struct netlink_callback *cb)
3604 enum addr_type_t type = MULTICAST_ADDR;
3606 return inet6_dump_addr(skb, cb, type);
3610 static int inet6_dump_ifacaddr(struct sk_buff *skb, struct netlink_callback *cb)
3612 enum addr_type_t type = ANYCAST_ADDR;
3614 return inet6_dump_addr(skb, cb, type);
3617 static int inet6_rtm_getaddr(struct sk_buff *in_skb, struct nlmsghdr* nlh,
3618 void *arg)
3620 struct net *net = sock_net(in_skb->sk);
3621 struct ifaddrmsg *ifm;
3622 struct nlattr *tb[IFA_MAX+1];
3623 struct in6_addr *addr = NULL;
3624 struct net_device *dev = NULL;
3625 struct inet6_ifaddr *ifa;
3626 struct sk_buff *skb;
3627 int err;
3629 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3630 if (err < 0)
3631 goto errout;
3633 addr = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3634 if (addr == NULL) {
3635 err = -EINVAL;
3636 goto errout;
3639 ifm = nlmsg_data(nlh);
3640 if (ifm->ifa_index)
3641 dev = __dev_get_by_index(net, ifm->ifa_index);
3643 if ((ifa = ipv6_get_ifaddr(net, addr, dev, 1)) == NULL) {
3644 err = -EADDRNOTAVAIL;
3645 goto errout;
3648 if ((skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_KERNEL)) == NULL) {
3649 err = -ENOBUFS;
3650 goto errout_ifa;
3653 err = inet6_fill_ifaddr(skb, ifa, NETLINK_CB(in_skb).pid,
3654 nlh->nlmsg_seq, RTM_NEWADDR, 0);
3655 if (err < 0) {
3656 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
3657 WARN_ON(err == -EMSGSIZE);
3658 kfree_skb(skb);
3659 goto errout_ifa;
3661 err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).pid);
3662 errout_ifa:
3663 in6_ifa_put(ifa);
3664 errout:
3665 return err;
3668 static void inet6_ifa_notify(int event, struct inet6_ifaddr *ifa)
3670 struct sk_buff *skb;
3671 struct net *net = dev_net(ifa->idev->dev);
3672 int err = -ENOBUFS;
3674 skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_ATOMIC);
3675 if (skb == NULL)
3676 goto errout;
3678 err = inet6_fill_ifaddr(skb, ifa, 0, 0, event, 0);
3679 if (err < 0) {
3680 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
3681 WARN_ON(err == -EMSGSIZE);
3682 kfree_skb(skb);
3683 goto errout;
3685 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
3686 return;
3687 errout:
3688 if (err < 0)
3689 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
3692 static inline void ipv6_store_devconf(struct ipv6_devconf *cnf,
3693 __s32 *array, int bytes)
3695 BUG_ON(bytes < (DEVCONF_MAX * 4));
3697 memset(array, 0, bytes);
3698 array[DEVCONF_FORWARDING] = cnf->forwarding;
3699 array[DEVCONF_HOPLIMIT] = cnf->hop_limit;
3700 array[DEVCONF_MTU6] = cnf->mtu6;
3701 array[DEVCONF_ACCEPT_RA] = cnf->accept_ra;
3702 array[DEVCONF_ACCEPT_REDIRECTS] = cnf->accept_redirects;
3703 array[DEVCONF_AUTOCONF] = cnf->autoconf;
3704 array[DEVCONF_DAD_TRANSMITS] = cnf->dad_transmits;
3705 array[DEVCONF_RTR_SOLICITS] = cnf->rtr_solicits;
3706 array[DEVCONF_RTR_SOLICIT_INTERVAL] = cnf->rtr_solicit_interval;
3707 array[DEVCONF_RTR_SOLICIT_DELAY] = cnf->rtr_solicit_delay;
3708 array[DEVCONF_FORCE_MLD_VERSION] = cnf->force_mld_version;
3709 #ifdef CONFIG_IPV6_PRIVACY
3710 array[DEVCONF_USE_TEMPADDR] = cnf->use_tempaddr;
3711 array[DEVCONF_TEMP_VALID_LFT] = cnf->temp_valid_lft;
3712 array[DEVCONF_TEMP_PREFERED_LFT] = cnf->temp_prefered_lft;
3713 array[DEVCONF_REGEN_MAX_RETRY] = cnf->regen_max_retry;
3714 array[DEVCONF_MAX_DESYNC_FACTOR] = cnf->max_desync_factor;
3715 #endif
3716 array[DEVCONF_MAX_ADDRESSES] = cnf->max_addresses;
3717 array[DEVCONF_ACCEPT_RA_DEFRTR] = cnf->accept_ra_defrtr;
3718 array[DEVCONF_ACCEPT_RA_PINFO] = cnf->accept_ra_pinfo;
3719 #ifdef CONFIG_IPV6_ROUTER_PREF
3720 array[DEVCONF_ACCEPT_RA_RTR_PREF] = cnf->accept_ra_rtr_pref;
3721 array[DEVCONF_RTR_PROBE_INTERVAL] = cnf->rtr_probe_interval;
3722 #ifdef CONFIG_IPV6_ROUTE_INFO
3723 array[DEVCONF_ACCEPT_RA_RT_INFO_MAX_PLEN] = cnf->accept_ra_rt_info_max_plen;
3724 #endif
3725 #endif
3726 array[DEVCONF_PROXY_NDP] = cnf->proxy_ndp;
3727 array[DEVCONF_ACCEPT_SOURCE_ROUTE] = cnf->accept_source_route;
3728 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
3729 array[DEVCONF_OPTIMISTIC_DAD] = cnf->optimistic_dad;
3730 #endif
3731 #ifdef CONFIG_IPV6_MROUTE
3732 array[DEVCONF_MC_FORWARDING] = cnf->mc_forwarding;
3733 #endif
3734 array[DEVCONF_DISABLE_IPV6] = cnf->disable_ipv6;
3735 array[DEVCONF_ACCEPT_DAD] = cnf->accept_dad;
3736 array[DEVCONF_FORCE_TLLAO] = cnf->force_tllao;
3739 static inline size_t inet6_if_nlmsg_size(void)
3741 return NLMSG_ALIGN(sizeof(struct ifinfomsg))
3742 + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
3743 + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
3744 + nla_total_size(4) /* IFLA_MTU */
3745 + nla_total_size(4) /* IFLA_LINK */
3746 + nla_total_size( /* IFLA_PROTINFO */
3747 nla_total_size(4) /* IFLA_INET6_FLAGS */
3748 + nla_total_size(sizeof(struct ifla_cacheinfo))
3749 + nla_total_size(DEVCONF_MAX * 4) /* IFLA_INET6_CONF */
3750 + nla_total_size(IPSTATS_MIB_MAX * 8) /* IFLA_INET6_STATS */
3751 + nla_total_size(ICMP6_MIB_MAX * 8) /* IFLA_INET6_ICMP6STATS */
3755 static inline void __snmp6_fill_stats(u64 *stats, void **mib, int items,
3756 int bytes)
3758 int i;
3759 int pad = bytes - sizeof(u64) * items;
3760 BUG_ON(pad < 0);
3762 /* Use put_unaligned() because stats may not be aligned for u64. */
3763 put_unaligned(items, &stats[0]);
3764 for (i = 1; i < items; i++)
3765 put_unaligned(snmp_fold_field(mib, i), &stats[i]);
3767 memset(&stats[items], 0, pad);
3770 static void snmp6_fill_stats(u64 *stats, struct inet6_dev *idev, int attrtype,
3771 int bytes)
3773 switch(attrtype) {
3774 case IFLA_INET6_STATS:
3775 __snmp6_fill_stats(stats, (void **)idev->stats.ipv6, IPSTATS_MIB_MAX, bytes);
3776 break;
3777 case IFLA_INET6_ICMP6STATS:
3778 __snmp6_fill_stats(stats, (void **)idev->stats.icmpv6, ICMP6_MIB_MAX, bytes);
3779 break;
3783 static int inet6_fill_ifinfo(struct sk_buff *skb, struct inet6_dev *idev,
3784 u32 pid, u32 seq, int event, unsigned int flags)
3786 struct net_device *dev = idev->dev;
3787 struct nlattr *nla;
3788 struct ifinfomsg *hdr;
3789 struct nlmsghdr *nlh;
3790 void *protoinfo;
3791 struct ifla_cacheinfo ci;
3793 nlh = nlmsg_put(skb, pid, seq, event, sizeof(*hdr), flags);
3794 if (nlh == NULL)
3795 return -EMSGSIZE;
3797 hdr = nlmsg_data(nlh);
3798 hdr->ifi_family = AF_INET6;
3799 hdr->__ifi_pad = 0;
3800 hdr->ifi_type = dev->type;
3801 hdr->ifi_index = dev->ifindex;
3802 hdr->ifi_flags = dev_get_flags(dev);
3803 hdr->ifi_change = 0;
3805 NLA_PUT_STRING(skb, IFLA_IFNAME, dev->name);
3807 if (dev->addr_len)
3808 NLA_PUT(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr);
3810 NLA_PUT_U32(skb, IFLA_MTU, dev->mtu);
3811 if (dev->ifindex != dev->iflink)
3812 NLA_PUT_U32(skb, IFLA_LINK, dev->iflink);
3814 protoinfo = nla_nest_start(skb, IFLA_PROTINFO);
3815 if (protoinfo == NULL)
3816 goto nla_put_failure;
3818 NLA_PUT_U32(skb, IFLA_INET6_FLAGS, idev->if_flags);
3820 ci.max_reasm_len = IPV6_MAXPLEN;
3821 ci.tstamp = (__u32)(TIME_DELTA(idev->tstamp, INITIAL_JIFFIES) / HZ * 100
3822 + TIME_DELTA(idev->tstamp, INITIAL_JIFFIES) % HZ * 100 / HZ);
3823 ci.reachable_time = idev->nd_parms->reachable_time;
3824 ci.retrans_time = idev->nd_parms->retrans_time;
3825 NLA_PUT(skb, IFLA_INET6_CACHEINFO, sizeof(ci), &ci);
3827 nla = nla_reserve(skb, IFLA_INET6_CONF, DEVCONF_MAX * sizeof(s32));
3828 if (nla == NULL)
3829 goto nla_put_failure;
3830 ipv6_store_devconf(&idev->cnf, nla_data(nla), nla_len(nla));
3832 /* XXX - MC not implemented */
3834 nla = nla_reserve(skb, IFLA_INET6_STATS, IPSTATS_MIB_MAX * sizeof(u64));
3835 if (nla == NULL)
3836 goto nla_put_failure;
3837 snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_STATS, nla_len(nla));
3839 nla = nla_reserve(skb, IFLA_INET6_ICMP6STATS, ICMP6_MIB_MAX * sizeof(u64));
3840 if (nla == NULL)
3841 goto nla_put_failure;
3842 snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_ICMP6STATS, nla_len(nla));
3844 nla_nest_end(skb, protoinfo);
3845 return nlmsg_end(skb, nlh);
3847 nla_put_failure:
3848 nlmsg_cancel(skb, nlh);
3849 return -EMSGSIZE;
3852 static int inet6_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
3854 struct net *net = sock_net(skb->sk);
3855 int h, s_h;
3856 int idx = 0, s_idx;
3857 struct net_device *dev;
3858 struct inet6_dev *idev;
3859 struct hlist_head *head;
3860 struct hlist_node *node;
3862 s_h = cb->args[0];
3863 s_idx = cb->args[1];
3865 rcu_read_lock();
3866 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
3867 idx = 0;
3868 head = &net->dev_index_head[h];
3869 hlist_for_each_entry_rcu(dev, node, head, index_hlist) {
3870 if (idx < s_idx)
3871 goto cont;
3872 idev = __in6_dev_get(dev);
3873 if (!idev)
3874 goto cont;
3875 if (inet6_fill_ifinfo(skb, idev,
3876 NETLINK_CB(cb->skb).pid,
3877 cb->nlh->nlmsg_seq,
3878 RTM_NEWLINK, NLM_F_MULTI) <= 0)
3879 goto out;
3880 cont:
3881 idx++;
3884 out:
3885 rcu_read_unlock();
3886 cb->args[1] = idx;
3887 cb->args[0] = h;
3889 return skb->len;
3892 void inet6_ifinfo_notify(int event, struct inet6_dev *idev)
3894 struct sk_buff *skb;
3895 struct net *net = dev_net(idev->dev);
3896 int err = -ENOBUFS;
3898 skb = nlmsg_new(inet6_if_nlmsg_size(), GFP_ATOMIC);
3899 if (skb == NULL)
3900 goto errout;
3902 err = inet6_fill_ifinfo(skb, idev, 0, 0, event, 0);
3903 if (err < 0) {
3904 /* -EMSGSIZE implies BUG in inet6_if_nlmsg_size() */
3905 WARN_ON(err == -EMSGSIZE);
3906 kfree_skb(skb);
3907 goto errout;
3909 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
3910 return;
3911 errout:
3912 if (err < 0)
3913 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
3916 static inline size_t inet6_prefix_nlmsg_size(void)
3918 return NLMSG_ALIGN(sizeof(struct prefixmsg))
3919 + nla_total_size(sizeof(struct in6_addr))
3920 + nla_total_size(sizeof(struct prefix_cacheinfo));
3923 static int inet6_fill_prefix(struct sk_buff *skb, struct inet6_dev *idev,
3924 struct prefix_info *pinfo, u32 pid, u32 seq,
3925 int event, unsigned int flags)
3927 struct prefixmsg *pmsg;
3928 struct nlmsghdr *nlh;
3929 struct prefix_cacheinfo ci;
3931 nlh = nlmsg_put(skb, pid, seq, event, sizeof(*pmsg), flags);
3932 if (nlh == NULL)
3933 return -EMSGSIZE;
3935 pmsg = nlmsg_data(nlh);
3936 pmsg->prefix_family = AF_INET6;
3937 pmsg->prefix_pad1 = 0;
3938 pmsg->prefix_pad2 = 0;
3939 pmsg->prefix_ifindex = idev->dev->ifindex;
3940 pmsg->prefix_len = pinfo->prefix_len;
3941 pmsg->prefix_type = pinfo->type;
3942 pmsg->prefix_pad3 = 0;
3943 pmsg->prefix_flags = 0;
3944 if (pinfo->onlink)
3945 pmsg->prefix_flags |= IF_PREFIX_ONLINK;
3946 if (pinfo->autoconf)
3947 pmsg->prefix_flags |= IF_PREFIX_AUTOCONF;
3949 NLA_PUT(skb, PREFIX_ADDRESS, sizeof(pinfo->prefix), &pinfo->prefix);
3951 ci.preferred_time = ntohl(pinfo->prefered);
3952 ci.valid_time = ntohl(pinfo->valid);
3953 NLA_PUT(skb, PREFIX_CACHEINFO, sizeof(ci), &ci);
3955 return nlmsg_end(skb, nlh);
3957 nla_put_failure:
3958 nlmsg_cancel(skb, nlh);
3959 return -EMSGSIZE;
3962 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
3963 struct prefix_info *pinfo)
3965 struct sk_buff *skb;
3966 struct net *net = dev_net(idev->dev);
3967 int err = -ENOBUFS;
3969 skb = nlmsg_new(inet6_prefix_nlmsg_size(), GFP_ATOMIC);
3970 if (skb == NULL)
3971 goto errout;
3973 err = inet6_fill_prefix(skb, idev, pinfo, 0, 0, event, 0);
3974 if (err < 0) {
3975 /* -EMSGSIZE implies BUG in inet6_prefix_nlmsg_size() */
3976 WARN_ON(err == -EMSGSIZE);
3977 kfree_skb(skb);
3978 goto errout;
3980 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_PREFIX, NULL, GFP_ATOMIC);
3981 return;
3982 errout:
3983 if (err < 0)
3984 rtnl_set_sk_err(net, RTNLGRP_IPV6_PREFIX, err);
3987 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
3989 inet6_ifa_notify(event ? : RTM_NEWADDR, ifp);
3991 switch (event) {
3992 case RTM_NEWADDR:
3994 * If the address was optimistic
3995 * we inserted the route at the start of
3996 * our DAD process, so we don't need
3997 * to do it again
3999 if (!(ifp->rt->rt6i_node))
4000 ip6_ins_rt(ifp->rt);
4001 if (ifp->idev->cnf.forwarding)
4002 addrconf_join_anycast(ifp);
4003 break;
4004 case RTM_DELADDR:
4005 if (ifp->idev->cnf.forwarding)
4006 addrconf_leave_anycast(ifp);
4007 addrconf_leave_solict(ifp->idev, &ifp->addr);
4008 dst_hold(&ifp->rt->u.dst);
4009 if (ip6_del_rt(ifp->rt))
4010 dst_free(&ifp->rt->u.dst);
4011 break;
4015 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
4017 rcu_read_lock_bh();
4018 if (likely(ifp->idev->dead == 0))
4019 __ipv6_ifa_notify(event, ifp);
4020 rcu_read_unlock_bh();
4023 #ifdef CONFIG_SYSCTL
4025 static
4026 int addrconf_sysctl_forward(ctl_table *ctl, int write,
4027 void __user *buffer, size_t *lenp, loff_t *ppos)
4029 int *valp = ctl->data;
4030 int val = *valp;
4031 int ret;
4033 ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
4035 if (write)
4036 ret = addrconf_fixup_forwarding(ctl, valp, val);
4037 return ret;
4040 static int addrconf_sysctl_forward_strategy(ctl_table *table,
4041 void __user *oldval,
4042 size_t __user *oldlenp,
4043 void __user *newval, size_t newlen)
4045 int *valp = table->data;
4046 int val = *valp;
4047 int new;
4049 if (!newval || !newlen)
4050 return 0;
4051 if (newlen != sizeof(int))
4052 return -EINVAL;
4053 if (get_user(new, (int __user *)newval))
4054 return -EFAULT;
4055 if (new == *valp)
4056 return 0;
4057 if (oldval && oldlenp) {
4058 size_t len;
4059 if (get_user(len, oldlenp))
4060 return -EFAULT;
4061 if (len) {
4062 if (len > table->maxlen)
4063 len = table->maxlen;
4064 if (copy_to_user(oldval, valp, len))
4065 return -EFAULT;
4066 if (put_user(len, oldlenp))
4067 return -EFAULT;
4071 *valp = new;
4072 return addrconf_fixup_forwarding(table, valp, val);
4075 static void dev_disable_change(struct inet6_dev *idev)
4077 if (!idev || !idev->dev)
4078 return;
4080 if (idev->cnf.disable_ipv6)
4081 addrconf_notify(NULL, NETDEV_DOWN, idev->dev);
4082 else
4083 addrconf_notify(NULL, NETDEV_UP, idev->dev);
4086 static void addrconf_disable_change(struct net *net, __s32 newf)
4088 struct net_device *dev;
4089 struct inet6_dev *idev;
4091 rcu_read_lock();
4092 for_each_netdev_rcu(net, dev) {
4093 idev = __in6_dev_get(dev);
4094 if (idev) {
4095 int changed = (!idev->cnf.disable_ipv6) ^ (!newf);
4096 idev->cnf.disable_ipv6 = newf;
4097 if (changed)
4098 dev_disable_change(idev);
4101 rcu_read_unlock();
4104 static int addrconf_disable_ipv6(struct ctl_table *table, int *p, int old)
4106 struct net *net;
4108 net = (struct net *)table->extra2;
4110 if (p == &net->ipv6.devconf_dflt->disable_ipv6)
4111 return 0;
4113 if (!rtnl_trylock())
4114 return restart_syscall();
4116 if (p == &net->ipv6.devconf_all->disable_ipv6) {
4117 __s32 newf = net->ipv6.devconf_all->disable_ipv6;
4118 net->ipv6.devconf_dflt->disable_ipv6 = newf;
4119 addrconf_disable_change(net, newf);
4120 } else if ((!*p) ^ (!old))
4121 dev_disable_change((struct inet6_dev *)table->extra1);
4123 rtnl_unlock();
4124 return 0;
4127 static
4128 int addrconf_sysctl_disable(ctl_table *ctl, int write,
4129 void __user *buffer, size_t *lenp, loff_t *ppos)
4131 int *valp = ctl->data;
4132 int val = *valp;
4133 int ret;
4135 ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
4137 if (write)
4138 ret = addrconf_disable_ipv6(ctl, valp, val);
4139 return ret;
4142 static struct addrconf_sysctl_table
4144 struct ctl_table_header *sysctl_header;
4145 ctl_table addrconf_vars[DEVCONF_MAX+1];
4146 char *dev_name;
4147 } addrconf_sysctl __read_mostly = {
4148 .sysctl_header = NULL,
4149 .addrconf_vars = {
4151 .ctl_name = NET_IPV6_FORWARDING,
4152 .procname = "forwarding",
4153 .data = &ipv6_devconf.forwarding,
4154 .maxlen = sizeof(int),
4155 .mode = 0644,
4156 .proc_handler = addrconf_sysctl_forward,
4157 .strategy = addrconf_sysctl_forward_strategy,
4160 .ctl_name = NET_IPV6_HOP_LIMIT,
4161 .procname = "hop_limit",
4162 .data = &ipv6_devconf.hop_limit,
4163 .maxlen = sizeof(int),
4164 .mode = 0644,
4165 .proc_handler = proc_dointvec,
4168 .ctl_name = NET_IPV6_MTU,
4169 .procname = "mtu",
4170 .data = &ipv6_devconf.mtu6,
4171 .maxlen = sizeof(int),
4172 .mode = 0644,
4173 .proc_handler = proc_dointvec,
4176 .ctl_name = NET_IPV6_ACCEPT_RA,
4177 .procname = "accept_ra",
4178 .data = &ipv6_devconf.accept_ra,
4179 .maxlen = sizeof(int),
4180 .mode = 0644,
4181 .proc_handler = proc_dointvec,
4184 .ctl_name = NET_IPV6_ACCEPT_REDIRECTS,
4185 .procname = "accept_redirects",
4186 .data = &ipv6_devconf.accept_redirects,
4187 .maxlen = sizeof(int),
4188 .mode = 0644,
4189 .proc_handler = proc_dointvec,
4192 .ctl_name = NET_IPV6_AUTOCONF,
4193 .procname = "autoconf",
4194 .data = &ipv6_devconf.autoconf,
4195 .maxlen = sizeof(int),
4196 .mode = 0644,
4197 .proc_handler = proc_dointvec,
4200 .ctl_name = NET_IPV6_DAD_TRANSMITS,
4201 .procname = "dad_transmits",
4202 .data = &ipv6_devconf.dad_transmits,
4203 .maxlen = sizeof(int),
4204 .mode = 0644,
4205 .proc_handler = proc_dointvec,
4208 .ctl_name = NET_IPV6_RTR_SOLICITS,
4209 .procname = "router_solicitations",
4210 .data = &ipv6_devconf.rtr_solicits,
4211 .maxlen = sizeof(int),
4212 .mode = 0644,
4213 .proc_handler = proc_dointvec,
4216 .ctl_name = NET_IPV6_RTR_SOLICIT_INTERVAL,
4217 .procname = "router_solicitation_interval",
4218 .data = &ipv6_devconf.rtr_solicit_interval,
4219 .maxlen = sizeof(int),
4220 .mode = 0644,
4221 .proc_handler = proc_dointvec_jiffies,
4222 .strategy = sysctl_jiffies,
4225 .ctl_name = NET_IPV6_RTR_SOLICIT_DELAY,
4226 .procname = "router_solicitation_delay",
4227 .data = &ipv6_devconf.rtr_solicit_delay,
4228 .maxlen = sizeof(int),
4229 .mode = 0644,
4230 .proc_handler = proc_dointvec_jiffies,
4231 .strategy = sysctl_jiffies,
4234 .ctl_name = NET_IPV6_FORCE_MLD_VERSION,
4235 .procname = "force_mld_version",
4236 .data = &ipv6_devconf.force_mld_version,
4237 .maxlen = sizeof(int),
4238 .mode = 0644,
4239 .proc_handler = proc_dointvec,
4241 #ifdef CONFIG_IPV6_PRIVACY
4243 .ctl_name = NET_IPV6_USE_TEMPADDR,
4244 .procname = "use_tempaddr",
4245 .data = &ipv6_devconf.use_tempaddr,
4246 .maxlen = sizeof(int),
4247 .mode = 0644,
4248 .proc_handler = proc_dointvec,
4251 .ctl_name = NET_IPV6_TEMP_VALID_LFT,
4252 .procname = "temp_valid_lft",
4253 .data = &ipv6_devconf.temp_valid_lft,
4254 .maxlen = sizeof(int),
4255 .mode = 0644,
4256 .proc_handler = proc_dointvec,
4259 .ctl_name = NET_IPV6_TEMP_PREFERED_LFT,
4260 .procname = "temp_prefered_lft",
4261 .data = &ipv6_devconf.temp_prefered_lft,
4262 .maxlen = sizeof(int),
4263 .mode = 0644,
4264 .proc_handler = proc_dointvec,
4267 .ctl_name = NET_IPV6_REGEN_MAX_RETRY,
4268 .procname = "regen_max_retry",
4269 .data = &ipv6_devconf.regen_max_retry,
4270 .maxlen = sizeof(int),
4271 .mode = 0644,
4272 .proc_handler = proc_dointvec,
4275 .ctl_name = NET_IPV6_MAX_DESYNC_FACTOR,
4276 .procname = "max_desync_factor",
4277 .data = &ipv6_devconf.max_desync_factor,
4278 .maxlen = sizeof(int),
4279 .mode = 0644,
4280 .proc_handler = proc_dointvec,
4282 #endif
4284 .ctl_name = NET_IPV6_MAX_ADDRESSES,
4285 .procname = "max_addresses",
4286 .data = &ipv6_devconf.max_addresses,
4287 .maxlen = sizeof(int),
4288 .mode = 0644,
4289 .proc_handler = proc_dointvec,
4292 .ctl_name = NET_IPV6_ACCEPT_RA_DEFRTR,
4293 .procname = "accept_ra_defrtr",
4294 .data = &ipv6_devconf.accept_ra_defrtr,
4295 .maxlen = sizeof(int),
4296 .mode = 0644,
4297 .proc_handler = proc_dointvec,
4300 .ctl_name = NET_IPV6_ACCEPT_RA_PINFO,
4301 .procname = "accept_ra_pinfo",
4302 .data = &ipv6_devconf.accept_ra_pinfo,
4303 .maxlen = sizeof(int),
4304 .mode = 0644,
4305 .proc_handler = proc_dointvec,
4307 #ifdef CONFIG_IPV6_ROUTER_PREF
4309 .ctl_name = NET_IPV6_ACCEPT_RA_RTR_PREF,
4310 .procname = "accept_ra_rtr_pref",
4311 .data = &ipv6_devconf.accept_ra_rtr_pref,
4312 .maxlen = sizeof(int),
4313 .mode = 0644,
4314 .proc_handler = proc_dointvec,
4317 .ctl_name = NET_IPV6_RTR_PROBE_INTERVAL,
4318 .procname = "router_probe_interval",
4319 .data = &ipv6_devconf.rtr_probe_interval,
4320 .maxlen = sizeof(int),
4321 .mode = 0644,
4322 .proc_handler = proc_dointvec_jiffies,
4323 .strategy = sysctl_jiffies,
4325 #ifdef CONFIG_IPV6_ROUTE_INFO
4327 .ctl_name = NET_IPV6_ACCEPT_RA_RT_INFO_MAX_PLEN,
4328 .procname = "accept_ra_rt_info_max_plen",
4329 .data = &ipv6_devconf.accept_ra_rt_info_max_plen,
4330 .maxlen = sizeof(int),
4331 .mode = 0644,
4332 .proc_handler = proc_dointvec,
4334 #endif
4335 #endif
4337 .ctl_name = NET_IPV6_PROXY_NDP,
4338 .procname = "proxy_ndp",
4339 .data = &ipv6_devconf.proxy_ndp,
4340 .maxlen = sizeof(int),
4341 .mode = 0644,
4342 .proc_handler = proc_dointvec,
4345 .ctl_name = NET_IPV6_ACCEPT_SOURCE_ROUTE,
4346 .procname = "accept_source_route",
4347 .data = &ipv6_devconf.accept_source_route,
4348 .maxlen = sizeof(int),
4349 .mode = 0644,
4350 .proc_handler = proc_dointvec,
4352 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
4354 .ctl_name = CTL_UNNUMBERED,
4355 .procname = "optimistic_dad",
4356 .data = &ipv6_devconf.optimistic_dad,
4357 .maxlen = sizeof(int),
4358 .mode = 0644,
4359 .proc_handler = proc_dointvec,
4362 #endif
4363 #ifdef CONFIG_IPV6_MROUTE
4365 .ctl_name = CTL_UNNUMBERED,
4366 .procname = "mc_forwarding",
4367 .data = &ipv6_devconf.mc_forwarding,
4368 .maxlen = sizeof(int),
4369 .mode = 0444,
4370 .proc_handler = proc_dointvec,
4372 #endif
4374 .ctl_name = CTL_UNNUMBERED,
4375 .procname = "disable_ipv6",
4376 .data = &ipv6_devconf.disable_ipv6,
4377 .maxlen = sizeof(int),
4378 .mode = 0644,
4379 .proc_handler = addrconf_sysctl_disable,
4380 .strategy = sysctl_intvec,
4383 .ctl_name = CTL_UNNUMBERED,
4384 .procname = "accept_dad",
4385 .data = &ipv6_devconf.accept_dad,
4386 .maxlen = sizeof(int),
4387 .mode = 0644,
4388 .proc_handler = proc_dointvec,
4391 .ctl_name = CTL_UNNUMBERED,
4392 .procname = "force_tllao",
4393 .data = &ipv6_devconf.force_tllao,
4394 .maxlen = sizeof(int),
4395 .mode = 0644,
4396 .proc_handler = proc_dointvec
4399 .ctl_name = 0, /* sentinel */
4404 static int __addrconf_sysctl_register(struct net *net, char *dev_name,
4405 int ctl_name, struct inet6_dev *idev, struct ipv6_devconf *p)
4407 int i;
4408 struct addrconf_sysctl_table *t;
4410 #define ADDRCONF_CTL_PATH_DEV 3
4412 struct ctl_path addrconf_ctl_path[] = {
4413 { .procname = "net", .ctl_name = CTL_NET, },
4414 { .procname = "ipv6", .ctl_name = NET_IPV6, },
4415 { .procname = "conf", .ctl_name = NET_IPV6_CONF, },
4416 { /* to be set */ },
4417 { },
4421 t = kmemdup(&addrconf_sysctl, sizeof(*t), GFP_KERNEL);
4422 if (t == NULL)
4423 goto out;
4425 for (i=0; t->addrconf_vars[i].data; i++) {
4426 t->addrconf_vars[i].data += (char*)p - (char*)&ipv6_devconf;
4427 t->addrconf_vars[i].extra1 = idev; /* embedded; no ref */
4428 t->addrconf_vars[i].extra2 = net;
4432 * Make a copy of dev_name, because '.procname' is regarded as const
4433 * by sysctl and we wouldn't want anyone to change it under our feet
4434 * (see SIOCSIFNAME).
4436 t->dev_name = kstrdup(dev_name, GFP_KERNEL);
4437 if (!t->dev_name)
4438 goto free;
4440 addrconf_ctl_path[ADDRCONF_CTL_PATH_DEV].procname = t->dev_name;
4441 addrconf_ctl_path[ADDRCONF_CTL_PATH_DEV].ctl_name = ctl_name;
4443 t->sysctl_header = register_net_sysctl_table(net, addrconf_ctl_path,
4444 t->addrconf_vars);
4445 if (t->sysctl_header == NULL)
4446 goto free_procname;
4448 p->sysctl = t;
4449 return 0;
4451 free_procname:
4452 kfree(t->dev_name);
4453 free:
4454 kfree(t);
4455 out:
4456 return -ENOBUFS;
4459 static void __addrconf_sysctl_unregister(struct ipv6_devconf *p)
4461 struct addrconf_sysctl_table *t;
4463 if (p->sysctl == NULL)
4464 return;
4466 t = p->sysctl;
4467 p->sysctl = NULL;
4468 unregister_sysctl_table(t->sysctl_header);
4469 kfree(t->dev_name);
4470 kfree(t);
4473 static void addrconf_sysctl_register(struct inet6_dev *idev)
4475 neigh_sysctl_register(idev->dev, idev->nd_parms, NET_IPV6,
4476 NET_IPV6_NEIGH, "ipv6",
4477 &ndisc_ifinfo_sysctl_change,
4478 ndisc_ifinfo_sysctl_strategy);
4479 __addrconf_sysctl_register(dev_net(idev->dev), idev->dev->name,
4480 idev->dev->ifindex, idev, &idev->cnf);
4483 static void addrconf_sysctl_unregister(struct inet6_dev *idev)
4485 __addrconf_sysctl_unregister(&idev->cnf);
4486 neigh_sysctl_unregister(idev->nd_parms);
4490 #endif
4492 static int addrconf_init_net(struct net *net)
4494 int err;
4495 struct ipv6_devconf *all, *dflt;
4497 err = -ENOMEM;
4498 all = &ipv6_devconf;
4499 dflt = &ipv6_devconf_dflt;
4501 if (!net_eq(net, &init_net)) {
4502 all = kmemdup(all, sizeof(ipv6_devconf), GFP_KERNEL);
4503 if (all == NULL)
4504 goto err_alloc_all;
4506 dflt = kmemdup(dflt, sizeof(ipv6_devconf_dflt), GFP_KERNEL);
4507 if (dflt == NULL)
4508 goto err_alloc_dflt;
4509 } else {
4510 /* these will be inherited by all namespaces */
4511 dflt->autoconf = ipv6_defaults.autoconf;
4512 dflt->disable_ipv6 = ipv6_defaults.disable_ipv6;
4515 net->ipv6.devconf_all = all;
4516 net->ipv6.devconf_dflt = dflt;
4518 #ifdef CONFIG_SYSCTL
4519 err = __addrconf_sysctl_register(net, "all", NET_PROTO_CONF_ALL,
4520 NULL, all);
4521 if (err < 0)
4522 goto err_reg_all;
4524 err = __addrconf_sysctl_register(net, "default", NET_PROTO_CONF_DEFAULT,
4525 NULL, dflt);
4526 if (err < 0)
4527 goto err_reg_dflt;
4528 #endif
4529 return 0;
4531 #ifdef CONFIG_SYSCTL
4532 err_reg_dflt:
4533 __addrconf_sysctl_unregister(all);
4534 err_reg_all:
4535 kfree(dflt);
4536 #endif
4537 err_alloc_dflt:
4538 kfree(all);
4539 err_alloc_all:
4540 return err;
4543 static void addrconf_exit_net(struct net *net)
4545 #ifdef CONFIG_SYSCTL
4546 __addrconf_sysctl_unregister(net->ipv6.devconf_dflt);
4547 __addrconf_sysctl_unregister(net->ipv6.devconf_all);
4548 #endif
4549 if (!net_eq(net, &init_net)) {
4550 kfree(net->ipv6.devconf_dflt);
4551 kfree(net->ipv6.devconf_all);
4555 static struct pernet_operations addrconf_ops = {
4556 .init = addrconf_init_net,
4557 .exit = addrconf_exit_net,
4561 * Device notifier
4564 int register_inet6addr_notifier(struct notifier_block *nb)
4566 return atomic_notifier_chain_register(&inet6addr_chain, nb);
4569 EXPORT_SYMBOL(register_inet6addr_notifier);
4571 int unregister_inet6addr_notifier(struct notifier_block *nb)
4573 return atomic_notifier_chain_unregister(&inet6addr_chain,nb);
4576 EXPORT_SYMBOL(unregister_inet6addr_notifier);
4579 * Init / cleanup code
4582 int __init addrconf_init(void)
4584 int err;
4586 if ((err = ipv6_addr_label_init()) < 0) {
4587 printk(KERN_CRIT "IPv6 Addrconf: cannot initialize default policy table: %d.\n",
4588 err);
4589 return err;
4592 register_pernet_subsys(&addrconf_ops);
4594 /* The addrconf netdev notifier requires that loopback_dev
4595 * has it's ipv6 private information allocated and setup
4596 * before it can bring up and give link-local addresses
4597 * to other devices which are up.
4599 * Unfortunately, loopback_dev is not necessarily the first
4600 * entry in the global dev_base list of net devices. In fact,
4601 * it is likely to be the very last entry on that list.
4602 * So this causes the notifier registry below to try and
4603 * give link-local addresses to all devices besides loopback_dev
4604 * first, then loopback_dev, which cases all the non-loopback_dev
4605 * devices to fail to get a link-local address.
4607 * So, as a temporary fix, allocate the ipv6 structure for
4608 * loopback_dev first by hand.
4609 * Longer term, all of the dependencies ipv6 has upon the loopback
4610 * device and it being up should be removed.
4612 rtnl_lock();
4613 if (!ipv6_add_dev(init_net.loopback_dev))
4614 err = -ENOMEM;
4615 rtnl_unlock();
4616 if (err)
4617 goto errlo;
4619 register_netdevice_notifier(&ipv6_dev_notf);
4621 addrconf_verify(0);
4623 err = __rtnl_register(PF_INET6, RTM_GETLINK, NULL, inet6_dump_ifinfo);
4624 if (err < 0)
4625 goto errout;
4627 /* Only the first call to __rtnl_register can fail */
4628 __rtnl_register(PF_INET6, RTM_NEWADDR, inet6_rtm_newaddr, NULL);
4629 __rtnl_register(PF_INET6, RTM_DELADDR, inet6_rtm_deladdr, NULL);
4630 __rtnl_register(PF_INET6, RTM_GETADDR, inet6_rtm_getaddr, inet6_dump_ifaddr);
4631 __rtnl_register(PF_INET6, RTM_GETMULTICAST, NULL, inet6_dump_ifmcaddr);
4632 __rtnl_register(PF_INET6, RTM_GETANYCAST, NULL, inet6_dump_ifacaddr);
4634 ipv6_addr_label_rtnl_register();
4636 return 0;
4637 errout:
4638 unregister_netdevice_notifier(&ipv6_dev_notf);
4639 errlo:
4640 unregister_pernet_subsys(&addrconf_ops);
4642 return err;
4645 void addrconf_cleanup(void)
4647 struct inet6_ifaddr *ifa;
4648 struct net_device *dev;
4649 int i;
4651 unregister_netdevice_notifier(&ipv6_dev_notf);
4652 unregister_pernet_subsys(&addrconf_ops);
4654 rtnl_lock();
4656 /* clean dev list */
4657 for_each_netdev(&init_net, dev) {
4658 if (__in6_dev_get(dev) == NULL)
4659 continue;
4660 addrconf_ifdown(dev, 1);
4662 addrconf_ifdown(init_net.loopback_dev, 2);
4665 * Check hash table.
4667 write_lock_bh(&addrconf_hash_lock);
4668 for (i=0; i < IN6_ADDR_HSIZE; i++) {
4669 for (ifa=inet6_addr_lst[i]; ifa; ) {
4670 struct inet6_ifaddr *bifa;
4672 bifa = ifa;
4673 ifa = ifa->lst_next;
4674 printk(KERN_DEBUG "bug: IPv6 address leakage detected: ifa=%p\n", bifa);
4675 /* Do not free it; something is wrong.
4676 Now we can investigate it with debugger.
4680 write_unlock_bh(&addrconf_hash_lock);
4682 del_timer(&addr_chk_timer);
4683 rtnl_unlock();