OMAP3: PM: CPUidle: only scale voltage for deeper C-states
[linux-ginger.git] / net / ipv6 / addrconf.c
blob1fd0a3d775d26767dec15c78f599dc96b1c3c08d
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 read_lock(&dev_base_lock);
485 for_each_netdev(net, dev) {
486 rcu_read_lock();
487 idev = __in6_dev_get(dev);
488 if (idev) {
489 int changed = (!idev->cnf.forwarding) ^ (!newf);
490 idev->cnf.forwarding = newf;
491 if (changed)
492 dev_forward_change(idev);
494 rcu_read_unlock();
496 read_unlock(&dev_base_lock);
499 static int addrconf_fixup_forwarding(struct ctl_table *table, int *p, int old)
501 struct net *net;
503 net = (struct net *)table->extra2;
504 if (p == &net->ipv6.devconf_dflt->forwarding)
505 return 0;
507 if (!rtnl_trylock())
508 return restart_syscall();
510 if (p == &net->ipv6.devconf_all->forwarding) {
511 __s32 newf = net->ipv6.devconf_all->forwarding;
512 net->ipv6.devconf_dflt->forwarding = newf;
513 addrconf_forward_change(net, newf);
514 } else if ((!*p) ^ (!old))
515 dev_forward_change((struct inet6_dev *)table->extra1);
516 rtnl_unlock();
518 if (*p)
519 rt6_purge_dflt_routers(net);
520 return 1;
522 #endif
524 /* Nobody refers to this ifaddr, destroy it */
526 void inet6_ifa_finish_destroy(struct inet6_ifaddr *ifp)
528 WARN_ON(ifp->if_next != NULL);
529 WARN_ON(ifp->lst_next != NULL);
531 #ifdef NET_REFCNT_DEBUG
532 printk(KERN_DEBUG "inet6_ifa_finish_destroy\n");
533 #endif
535 in6_dev_put(ifp->idev);
537 if (del_timer(&ifp->timer))
538 printk("Timer is still running, when freeing ifa=%p\n", ifp);
540 if (!ifp->dead) {
541 printk("Freeing alive inet6 address %p\n", ifp);
542 return;
544 dst_release(&ifp->rt->u.dst);
546 kfree(ifp);
549 static void
550 ipv6_link_dev_addr(struct inet6_dev *idev, struct inet6_ifaddr *ifp)
552 struct inet6_ifaddr *ifa, **ifap;
553 int ifp_scope = ipv6_addr_src_scope(&ifp->addr);
556 * Each device address list is sorted in order of scope -
557 * global before linklocal.
559 for (ifap = &idev->addr_list; (ifa = *ifap) != NULL;
560 ifap = &ifa->if_next) {
561 if (ifp_scope >= ipv6_addr_src_scope(&ifa->addr))
562 break;
565 ifp->if_next = *ifap;
566 *ifap = ifp;
570 * Hash function taken from net_alias.c
572 static u8 ipv6_addr_hash(const struct in6_addr *addr)
574 __u32 word;
577 * We perform the hash function over the last 64 bits of the address
578 * This will include the IEEE address token on links that support it.
581 word = (__force u32)(addr->s6_addr32[2] ^ addr->s6_addr32[3]);
582 word ^= (word >> 16);
583 word ^= (word >> 8);
585 return ((word ^ (word >> 4)) & 0x0f);
588 /* On success it returns ifp with increased reference count */
590 static struct inet6_ifaddr *
591 ipv6_add_addr(struct inet6_dev *idev, const struct in6_addr *addr, int pfxlen,
592 int scope, u32 flags)
594 struct inet6_ifaddr *ifa = NULL;
595 struct rt6_info *rt;
596 int hash;
597 int err = 0;
598 int addr_type = ipv6_addr_type(addr);
600 if (addr_type == IPV6_ADDR_ANY ||
601 addr_type & IPV6_ADDR_MULTICAST ||
602 (!(idev->dev->flags & IFF_LOOPBACK) &&
603 addr_type & IPV6_ADDR_LOOPBACK))
604 return ERR_PTR(-EADDRNOTAVAIL);
606 rcu_read_lock_bh();
607 if (idev->dead) {
608 err = -ENODEV; /*XXX*/
609 goto out2;
612 if (idev->cnf.disable_ipv6) {
613 err = -EACCES;
614 goto out2;
617 write_lock(&addrconf_hash_lock);
619 /* Ignore adding duplicate addresses on an interface */
620 if (ipv6_chk_same_addr(dev_net(idev->dev), addr, idev->dev)) {
621 ADBG(("ipv6_add_addr: already assigned\n"));
622 err = -EEXIST;
623 goto out;
626 ifa = kzalloc(sizeof(struct inet6_ifaddr), GFP_ATOMIC);
628 if (ifa == NULL) {
629 ADBG(("ipv6_add_addr: malloc failed\n"));
630 err = -ENOBUFS;
631 goto out;
634 rt = addrconf_dst_alloc(idev, addr, 0);
635 if (IS_ERR(rt)) {
636 err = PTR_ERR(rt);
637 goto out;
640 ipv6_addr_copy(&ifa->addr, addr);
642 spin_lock_init(&ifa->lock);
643 init_timer(&ifa->timer);
644 ifa->timer.data = (unsigned long) ifa;
645 ifa->scope = scope;
646 ifa->prefix_len = pfxlen;
647 ifa->flags = flags | IFA_F_TENTATIVE;
648 ifa->cstamp = ifa->tstamp = jiffies;
650 ifa->rt = rt;
653 * part one of RFC 4429, section 3.3
654 * We should not configure an address as
655 * optimistic if we do not yet know the link
656 * layer address of our nexhop router
659 if (rt->rt6i_nexthop == NULL)
660 ifa->flags &= ~IFA_F_OPTIMISTIC;
662 ifa->idev = idev;
663 in6_dev_hold(idev);
664 /* For caller */
665 in6_ifa_hold(ifa);
667 /* Add to big hash table */
668 hash = ipv6_addr_hash(addr);
670 ifa->lst_next = inet6_addr_lst[hash];
671 inet6_addr_lst[hash] = ifa;
672 in6_ifa_hold(ifa);
673 write_unlock(&addrconf_hash_lock);
675 write_lock(&idev->lock);
676 /* Add to inet6_dev unicast addr list. */
677 ipv6_link_dev_addr(idev, ifa);
679 #ifdef CONFIG_IPV6_PRIVACY
680 if (ifa->flags&IFA_F_TEMPORARY) {
681 ifa->tmp_next = idev->tempaddr_list;
682 idev->tempaddr_list = ifa;
683 in6_ifa_hold(ifa);
685 #endif
687 in6_ifa_hold(ifa);
688 write_unlock(&idev->lock);
689 out2:
690 rcu_read_unlock_bh();
692 if (likely(err == 0))
693 atomic_notifier_call_chain(&inet6addr_chain, NETDEV_UP, ifa);
694 else {
695 kfree(ifa);
696 ifa = ERR_PTR(err);
699 return ifa;
700 out:
701 write_unlock(&addrconf_hash_lock);
702 goto out2;
705 /* This function wants to get referenced ifp and releases it before return */
707 static void ipv6_del_addr(struct inet6_ifaddr *ifp)
709 struct inet6_ifaddr *ifa, **ifap;
710 struct inet6_dev *idev = ifp->idev;
711 int hash;
712 int deleted = 0, onlink = 0;
713 unsigned long expires = jiffies;
715 hash = ipv6_addr_hash(&ifp->addr);
717 ifp->dead = 1;
719 write_lock_bh(&addrconf_hash_lock);
720 for (ifap = &inet6_addr_lst[hash]; (ifa=*ifap) != NULL;
721 ifap = &ifa->lst_next) {
722 if (ifa == ifp) {
723 *ifap = ifa->lst_next;
724 __in6_ifa_put(ifp);
725 ifa->lst_next = NULL;
726 break;
729 write_unlock_bh(&addrconf_hash_lock);
731 write_lock_bh(&idev->lock);
732 #ifdef CONFIG_IPV6_PRIVACY
733 if (ifp->flags&IFA_F_TEMPORARY) {
734 for (ifap = &idev->tempaddr_list; (ifa=*ifap) != NULL;
735 ifap = &ifa->tmp_next) {
736 if (ifa == ifp) {
737 *ifap = ifa->tmp_next;
738 if (ifp->ifpub) {
739 in6_ifa_put(ifp->ifpub);
740 ifp->ifpub = NULL;
742 __in6_ifa_put(ifp);
743 ifa->tmp_next = NULL;
744 break;
748 #endif
750 for (ifap = &idev->addr_list; (ifa=*ifap) != NULL;) {
751 if (ifa == ifp) {
752 *ifap = ifa->if_next;
753 __in6_ifa_put(ifp);
754 ifa->if_next = NULL;
755 if (!(ifp->flags & IFA_F_PERMANENT) || onlink > 0)
756 break;
757 deleted = 1;
758 continue;
759 } else if (ifp->flags & IFA_F_PERMANENT) {
760 if (ipv6_prefix_equal(&ifa->addr, &ifp->addr,
761 ifp->prefix_len)) {
762 if (ifa->flags & IFA_F_PERMANENT) {
763 onlink = 1;
764 if (deleted)
765 break;
766 } else {
767 unsigned long lifetime;
769 if (!onlink)
770 onlink = -1;
772 spin_lock(&ifa->lock);
774 lifetime = addrconf_timeout_fixup(ifa->valid_lft, HZ);
776 * Note: Because this address is
777 * not permanent, lifetime <
778 * LONG_MAX / HZ here.
780 if (time_before(expires,
781 ifa->tstamp + lifetime * HZ))
782 expires = ifa->tstamp + lifetime * HZ;
783 spin_unlock(&ifa->lock);
787 ifap = &ifa->if_next;
789 write_unlock_bh(&idev->lock);
791 addrconf_del_timer(ifp);
793 ipv6_ifa_notify(RTM_DELADDR, ifp);
795 atomic_notifier_call_chain(&inet6addr_chain, NETDEV_DOWN, ifp);
798 * Purge or update corresponding prefix
800 * 1) we don't purge prefix here if address was not permanent.
801 * prefix is managed by its own lifetime.
802 * 2) if there're no addresses, delete prefix.
803 * 3) if there're still other permanent address(es),
804 * corresponding prefix is still permanent.
805 * 4) otherwise, update prefix lifetime to the
806 * longest valid lifetime among the corresponding
807 * addresses on the device.
808 * Note: subsequent RA will update lifetime.
810 * --yoshfuji
812 if ((ifp->flags & IFA_F_PERMANENT) && onlink < 1) {
813 struct in6_addr prefix;
814 struct rt6_info *rt;
815 struct net *net = dev_net(ifp->idev->dev);
816 ipv6_addr_prefix(&prefix, &ifp->addr, ifp->prefix_len);
817 rt = rt6_lookup(net, &prefix, NULL, ifp->idev->dev->ifindex, 1);
819 if (rt && addrconf_is_prefix_route(rt)) {
820 if (onlink == 0) {
821 ip6_del_rt(rt);
822 rt = NULL;
823 } else if (!(rt->rt6i_flags & RTF_EXPIRES)) {
824 rt->rt6i_expires = expires;
825 rt->rt6i_flags |= RTF_EXPIRES;
828 dst_release(&rt->u.dst);
831 in6_ifa_put(ifp);
834 #ifdef CONFIG_IPV6_PRIVACY
835 static int ipv6_create_tempaddr(struct inet6_ifaddr *ifp, struct inet6_ifaddr *ift)
837 struct inet6_dev *idev = ifp->idev;
838 struct in6_addr addr, *tmpaddr;
839 unsigned long tmp_prefered_lft, tmp_valid_lft, tmp_cstamp, tmp_tstamp;
840 unsigned long regen_advance;
841 int tmp_plen;
842 int ret = 0;
843 int max_addresses;
844 u32 addr_flags;
846 write_lock(&idev->lock);
847 if (ift) {
848 spin_lock_bh(&ift->lock);
849 memcpy(&addr.s6_addr[8], &ift->addr.s6_addr[8], 8);
850 spin_unlock_bh(&ift->lock);
851 tmpaddr = &addr;
852 } else {
853 tmpaddr = NULL;
855 retry:
856 in6_dev_hold(idev);
857 if (idev->cnf.use_tempaddr <= 0) {
858 write_unlock(&idev->lock);
859 printk(KERN_INFO
860 "ipv6_create_tempaddr(): use_tempaddr is disabled.\n");
861 in6_dev_put(idev);
862 ret = -1;
863 goto out;
865 spin_lock_bh(&ifp->lock);
866 if (ifp->regen_count++ >= idev->cnf.regen_max_retry) {
867 idev->cnf.use_tempaddr = -1; /*XXX*/
868 spin_unlock_bh(&ifp->lock);
869 write_unlock(&idev->lock);
870 printk(KERN_WARNING
871 "ipv6_create_tempaddr(): regeneration time exceeded. disabled temporary address support.\n");
872 in6_dev_put(idev);
873 ret = -1;
874 goto out;
876 in6_ifa_hold(ifp);
877 memcpy(addr.s6_addr, ifp->addr.s6_addr, 8);
878 if (__ipv6_try_regen_rndid(idev, tmpaddr) < 0) {
879 spin_unlock_bh(&ifp->lock);
880 write_unlock(&idev->lock);
881 printk(KERN_WARNING
882 "ipv6_create_tempaddr(): regeneration of randomized interface id failed.\n");
883 in6_ifa_put(ifp);
884 in6_dev_put(idev);
885 ret = -1;
886 goto out;
888 memcpy(&addr.s6_addr[8], idev->rndid, 8);
889 tmp_valid_lft = min_t(__u32,
890 ifp->valid_lft,
891 idev->cnf.temp_valid_lft);
892 tmp_prefered_lft = min_t(__u32,
893 ifp->prefered_lft,
894 idev->cnf.temp_prefered_lft - desync_factor / HZ);
895 tmp_plen = ifp->prefix_len;
896 max_addresses = idev->cnf.max_addresses;
897 tmp_cstamp = ifp->cstamp;
898 tmp_tstamp = ifp->tstamp;
899 spin_unlock_bh(&ifp->lock);
901 regen_advance = idev->cnf.regen_max_retry *
902 idev->cnf.dad_transmits *
903 idev->nd_parms->retrans_time / HZ;
904 write_unlock(&idev->lock);
906 /* A temporary address is created only if this calculated Preferred
907 * Lifetime is greater than REGEN_ADVANCE time units. In particular,
908 * an implementation must not create a temporary address with a zero
909 * Preferred Lifetime.
911 if (tmp_prefered_lft <= regen_advance) {
912 in6_ifa_put(ifp);
913 in6_dev_put(idev);
914 ret = -1;
915 goto out;
918 addr_flags = IFA_F_TEMPORARY;
919 /* set in addrconf_prefix_rcv() */
920 if (ifp->flags & IFA_F_OPTIMISTIC)
921 addr_flags |= IFA_F_OPTIMISTIC;
923 ift = !max_addresses ||
924 ipv6_count_addresses(idev) < max_addresses ?
925 ipv6_add_addr(idev, &addr, tmp_plen,
926 ipv6_addr_type(&addr)&IPV6_ADDR_SCOPE_MASK,
927 addr_flags) : NULL;
928 if (!ift || IS_ERR(ift)) {
929 in6_ifa_put(ifp);
930 in6_dev_put(idev);
931 printk(KERN_INFO
932 "ipv6_create_tempaddr(): retry temporary address regeneration.\n");
933 tmpaddr = &addr;
934 write_lock(&idev->lock);
935 goto retry;
938 spin_lock_bh(&ift->lock);
939 ift->ifpub = ifp;
940 ift->valid_lft = tmp_valid_lft;
941 ift->prefered_lft = tmp_prefered_lft;
942 ift->cstamp = tmp_cstamp;
943 ift->tstamp = tmp_tstamp;
944 spin_unlock_bh(&ift->lock);
946 addrconf_dad_start(ift, 0);
947 in6_ifa_put(ift);
948 in6_dev_put(idev);
949 out:
950 return ret;
952 #endif
955 * Choose an appropriate source address (RFC3484)
957 enum {
958 IPV6_SADDR_RULE_INIT = 0,
959 IPV6_SADDR_RULE_LOCAL,
960 IPV6_SADDR_RULE_SCOPE,
961 IPV6_SADDR_RULE_PREFERRED,
962 #ifdef CONFIG_IPV6_MIP6
963 IPV6_SADDR_RULE_HOA,
964 #endif
965 IPV6_SADDR_RULE_OIF,
966 IPV6_SADDR_RULE_LABEL,
967 #ifdef CONFIG_IPV6_PRIVACY
968 IPV6_SADDR_RULE_PRIVACY,
969 #endif
970 IPV6_SADDR_RULE_ORCHID,
971 IPV6_SADDR_RULE_PREFIX,
972 IPV6_SADDR_RULE_MAX
975 struct ipv6_saddr_score {
976 int rule;
977 int addr_type;
978 struct inet6_ifaddr *ifa;
979 DECLARE_BITMAP(scorebits, IPV6_SADDR_RULE_MAX);
980 int scopedist;
981 int matchlen;
984 struct ipv6_saddr_dst {
985 const struct in6_addr *addr;
986 int ifindex;
987 int scope;
988 int label;
989 unsigned int prefs;
992 static inline int ipv6_saddr_preferred(int type)
994 if (type & (IPV6_ADDR_MAPPED|IPV6_ADDR_COMPATv4|
995 IPV6_ADDR_LOOPBACK|IPV6_ADDR_RESERVED))
996 return 1;
997 return 0;
1000 static int ipv6_get_saddr_eval(struct net *net,
1001 struct ipv6_saddr_score *score,
1002 struct ipv6_saddr_dst *dst,
1003 int i)
1005 int ret;
1007 if (i <= score->rule) {
1008 switch (i) {
1009 case IPV6_SADDR_RULE_SCOPE:
1010 ret = score->scopedist;
1011 break;
1012 case IPV6_SADDR_RULE_PREFIX:
1013 ret = score->matchlen;
1014 break;
1015 default:
1016 ret = !!test_bit(i, score->scorebits);
1018 goto out;
1021 switch (i) {
1022 case IPV6_SADDR_RULE_INIT:
1023 /* Rule 0: remember if hiscore is not ready yet */
1024 ret = !!score->ifa;
1025 break;
1026 case IPV6_SADDR_RULE_LOCAL:
1027 /* Rule 1: Prefer same address */
1028 ret = ipv6_addr_equal(&score->ifa->addr, dst->addr);
1029 break;
1030 case IPV6_SADDR_RULE_SCOPE:
1031 /* Rule 2: Prefer appropriate scope
1033 * ret
1035 * -1 | d 15
1036 * ---+--+-+---> scope
1038 * | d is scope of the destination.
1039 * B-d | \
1040 * | \ <- smaller scope is better if
1041 * B-15 | \ if scope is enough for destinaion.
1042 * | ret = B - scope (-1 <= scope >= d <= 15).
1043 * d-C-1 | /
1044 * |/ <- greater is better
1045 * -C / if scope is not enough for destination.
1046 * /| ret = scope - C (-1 <= d < scope <= 15).
1048 * d - C - 1 < B -15 (for all -1 <= d <= 15).
1049 * C > d + 14 - B >= 15 + 14 - B = 29 - B.
1050 * Assume B = 0 and we get C > 29.
1052 ret = __ipv6_addr_src_scope(score->addr_type);
1053 if (ret >= dst->scope)
1054 ret = -ret;
1055 else
1056 ret -= 128; /* 30 is enough */
1057 score->scopedist = ret;
1058 break;
1059 case IPV6_SADDR_RULE_PREFERRED:
1060 /* Rule 3: Avoid deprecated and optimistic addresses */
1061 ret = ipv6_saddr_preferred(score->addr_type) ||
1062 !(score->ifa->flags & (IFA_F_DEPRECATED|IFA_F_OPTIMISTIC));
1063 break;
1064 #ifdef CONFIG_IPV6_MIP6
1065 case IPV6_SADDR_RULE_HOA:
1067 /* Rule 4: Prefer home address */
1068 int prefhome = !(dst->prefs & IPV6_PREFER_SRC_COA);
1069 ret = !(score->ifa->flags & IFA_F_HOMEADDRESS) ^ prefhome;
1070 break;
1072 #endif
1073 case IPV6_SADDR_RULE_OIF:
1074 /* Rule 5: Prefer outgoing interface */
1075 ret = (!dst->ifindex ||
1076 dst->ifindex == score->ifa->idev->dev->ifindex);
1077 break;
1078 case IPV6_SADDR_RULE_LABEL:
1079 /* Rule 6: Prefer matching label */
1080 ret = ipv6_addr_label(net,
1081 &score->ifa->addr, score->addr_type,
1082 score->ifa->idev->dev->ifindex) == dst->label;
1083 break;
1084 #ifdef CONFIG_IPV6_PRIVACY
1085 case IPV6_SADDR_RULE_PRIVACY:
1087 /* Rule 7: Prefer public address
1088 * Note: prefer temprary address if use_tempaddr >= 2
1090 int preftmp = dst->prefs & (IPV6_PREFER_SRC_PUBLIC|IPV6_PREFER_SRC_TMP) ?
1091 !!(dst->prefs & IPV6_PREFER_SRC_TMP) :
1092 score->ifa->idev->cnf.use_tempaddr >= 2;
1093 ret = (!(score->ifa->flags & IFA_F_TEMPORARY)) ^ preftmp;
1094 break;
1096 #endif
1097 case IPV6_SADDR_RULE_ORCHID:
1098 /* Rule 8-: Prefer ORCHID vs ORCHID or
1099 * non-ORCHID vs non-ORCHID
1101 ret = !(ipv6_addr_orchid(&score->ifa->addr) ^
1102 ipv6_addr_orchid(dst->addr));
1103 break;
1104 case IPV6_SADDR_RULE_PREFIX:
1105 /* Rule 8: Use longest matching prefix */
1106 score->matchlen = ret = ipv6_addr_diff(&score->ifa->addr,
1107 dst->addr);
1108 break;
1109 default:
1110 ret = 0;
1113 if (ret)
1114 __set_bit(i, score->scorebits);
1115 score->rule = i;
1116 out:
1117 return ret;
1120 int ipv6_dev_get_saddr(struct net *net, struct net_device *dst_dev,
1121 const struct in6_addr *daddr, unsigned int prefs,
1122 struct in6_addr *saddr)
1124 struct ipv6_saddr_score scores[2],
1125 *score = &scores[0], *hiscore = &scores[1];
1126 struct ipv6_saddr_dst dst;
1127 struct net_device *dev;
1128 int dst_type;
1130 dst_type = __ipv6_addr_type(daddr);
1131 dst.addr = daddr;
1132 dst.ifindex = dst_dev ? dst_dev->ifindex : 0;
1133 dst.scope = __ipv6_addr_src_scope(dst_type);
1134 dst.label = ipv6_addr_label(net, daddr, dst_type, dst.ifindex);
1135 dst.prefs = prefs;
1137 hiscore->rule = -1;
1138 hiscore->ifa = NULL;
1140 read_lock(&dev_base_lock);
1141 rcu_read_lock();
1143 for_each_netdev(net, dev) {
1144 struct inet6_dev *idev;
1146 /* Candidate Source Address (section 4)
1147 * - multicast and link-local destination address,
1148 * the set of candidate source address MUST only
1149 * include addresses assigned to interfaces
1150 * belonging to the same link as the outgoing
1151 * interface.
1152 * (- For site-local destination addresses, the
1153 * set of candidate source addresses MUST only
1154 * include addresses assigned to interfaces
1155 * belonging to the same site as the outgoing
1156 * interface.)
1158 if (((dst_type & IPV6_ADDR_MULTICAST) ||
1159 dst.scope <= IPV6_ADDR_SCOPE_LINKLOCAL) &&
1160 dst.ifindex && dev->ifindex != dst.ifindex)
1161 continue;
1163 idev = __in6_dev_get(dev);
1164 if (!idev)
1165 continue;
1167 read_lock_bh(&idev->lock);
1168 for (score->ifa = idev->addr_list; score->ifa; score->ifa = score->ifa->if_next) {
1169 int i;
1172 * - Tentative Address (RFC2462 section 5.4)
1173 * - A tentative address is not considered
1174 * "assigned to an interface" in the traditional
1175 * sense, unless it is also flagged as optimistic.
1176 * - Candidate Source Address (section 4)
1177 * - In any case, anycast addresses, multicast
1178 * addresses, and the unspecified address MUST
1179 * NOT be included in a candidate set.
1181 if ((score->ifa->flags & IFA_F_TENTATIVE) &&
1182 (!(score->ifa->flags & IFA_F_OPTIMISTIC)))
1183 continue;
1185 score->addr_type = __ipv6_addr_type(&score->ifa->addr);
1187 if (unlikely(score->addr_type == IPV6_ADDR_ANY ||
1188 score->addr_type & IPV6_ADDR_MULTICAST)) {
1189 LIMIT_NETDEBUG(KERN_DEBUG
1190 "ADDRCONF: unspecified / multicast address "
1191 "assigned as unicast address on %s",
1192 dev->name);
1193 continue;
1196 score->rule = -1;
1197 bitmap_zero(score->scorebits, IPV6_SADDR_RULE_MAX);
1199 for (i = 0; i < IPV6_SADDR_RULE_MAX; i++) {
1200 int minihiscore, miniscore;
1202 minihiscore = ipv6_get_saddr_eval(net, hiscore, &dst, i);
1203 miniscore = ipv6_get_saddr_eval(net, score, &dst, i);
1205 if (minihiscore > miniscore) {
1206 if (i == IPV6_SADDR_RULE_SCOPE &&
1207 score->scopedist > 0) {
1209 * special case:
1210 * each remaining entry
1211 * has too small (not enough)
1212 * scope, because ifa entries
1213 * are sorted by their scope
1214 * values.
1216 goto try_nextdev;
1218 break;
1219 } else if (minihiscore < miniscore) {
1220 if (hiscore->ifa)
1221 in6_ifa_put(hiscore->ifa);
1223 in6_ifa_hold(score->ifa);
1225 swap(hiscore, score);
1227 /* restore our iterator */
1228 score->ifa = hiscore->ifa;
1230 break;
1234 try_nextdev:
1235 read_unlock_bh(&idev->lock);
1237 rcu_read_unlock();
1238 read_unlock(&dev_base_lock);
1240 if (!hiscore->ifa)
1241 return -EADDRNOTAVAIL;
1243 ipv6_addr_copy(saddr, &hiscore->ifa->addr);
1244 in6_ifa_put(hiscore->ifa);
1245 return 0;
1248 EXPORT_SYMBOL(ipv6_dev_get_saddr);
1250 int ipv6_get_lladdr(struct net_device *dev, struct in6_addr *addr,
1251 unsigned char banned_flags)
1253 struct inet6_dev *idev;
1254 int err = -EADDRNOTAVAIL;
1256 rcu_read_lock();
1257 if ((idev = __in6_dev_get(dev)) != NULL) {
1258 struct inet6_ifaddr *ifp;
1260 read_lock_bh(&idev->lock);
1261 for (ifp=idev->addr_list; ifp; ifp=ifp->if_next) {
1262 if (ifp->scope == IFA_LINK && !(ifp->flags & banned_flags)) {
1263 ipv6_addr_copy(addr, &ifp->addr);
1264 err = 0;
1265 break;
1268 read_unlock_bh(&idev->lock);
1270 rcu_read_unlock();
1271 return err;
1274 static int ipv6_count_addresses(struct inet6_dev *idev)
1276 int cnt = 0;
1277 struct inet6_ifaddr *ifp;
1279 read_lock_bh(&idev->lock);
1280 for (ifp=idev->addr_list; ifp; ifp=ifp->if_next)
1281 cnt++;
1282 read_unlock_bh(&idev->lock);
1283 return cnt;
1286 int ipv6_chk_addr(struct net *net, struct in6_addr *addr,
1287 struct net_device *dev, int strict)
1289 struct inet6_ifaddr * ifp;
1290 u8 hash = ipv6_addr_hash(addr);
1292 read_lock_bh(&addrconf_hash_lock);
1293 for(ifp = inet6_addr_lst[hash]; ifp; ifp=ifp->lst_next) {
1294 if (!net_eq(dev_net(ifp->idev->dev), net))
1295 continue;
1296 if (ipv6_addr_equal(&ifp->addr, addr) &&
1297 !(ifp->flags&IFA_F_TENTATIVE)) {
1298 if (dev == NULL || ifp->idev->dev == dev ||
1299 !(ifp->scope&(IFA_LINK|IFA_HOST) || strict))
1300 break;
1303 read_unlock_bh(&addrconf_hash_lock);
1304 return ifp != NULL;
1306 EXPORT_SYMBOL(ipv6_chk_addr);
1308 static
1309 int ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
1310 struct net_device *dev)
1312 struct inet6_ifaddr * ifp;
1313 u8 hash = ipv6_addr_hash(addr);
1315 for(ifp = inet6_addr_lst[hash]; ifp; ifp=ifp->lst_next) {
1316 if (!net_eq(dev_net(ifp->idev->dev), net))
1317 continue;
1318 if (ipv6_addr_equal(&ifp->addr, addr)) {
1319 if (dev == NULL || ifp->idev->dev == dev)
1320 break;
1323 return ifp != NULL;
1326 int ipv6_chk_prefix(struct in6_addr *addr, struct net_device *dev)
1328 struct inet6_dev *idev;
1329 struct inet6_ifaddr *ifa;
1330 int onlink;
1332 onlink = 0;
1333 rcu_read_lock();
1334 idev = __in6_dev_get(dev);
1335 if (idev) {
1336 read_lock_bh(&idev->lock);
1337 for (ifa = idev->addr_list; ifa; ifa = ifa->if_next) {
1338 onlink = ipv6_prefix_equal(addr, &ifa->addr,
1339 ifa->prefix_len);
1340 if (onlink)
1341 break;
1343 read_unlock_bh(&idev->lock);
1345 rcu_read_unlock();
1346 return onlink;
1349 EXPORT_SYMBOL(ipv6_chk_prefix);
1351 struct inet6_ifaddr *ipv6_get_ifaddr(struct net *net, const struct in6_addr *addr,
1352 struct net_device *dev, int strict)
1354 struct inet6_ifaddr * ifp;
1355 u8 hash = ipv6_addr_hash(addr);
1357 read_lock_bh(&addrconf_hash_lock);
1358 for(ifp = inet6_addr_lst[hash]; ifp; ifp=ifp->lst_next) {
1359 if (!net_eq(dev_net(ifp->idev->dev), net))
1360 continue;
1361 if (ipv6_addr_equal(&ifp->addr, addr)) {
1362 if (dev == NULL || ifp->idev->dev == dev ||
1363 !(ifp->scope&(IFA_LINK|IFA_HOST) || strict)) {
1364 in6_ifa_hold(ifp);
1365 break;
1369 read_unlock_bh(&addrconf_hash_lock);
1371 return ifp;
1374 /* Gets referenced address, destroys ifaddr */
1376 static void addrconf_dad_stop(struct inet6_ifaddr *ifp, int dad_failed)
1378 if (ifp->flags&IFA_F_PERMANENT) {
1379 spin_lock_bh(&ifp->lock);
1380 addrconf_del_timer(ifp);
1381 ifp->flags |= IFA_F_TENTATIVE;
1382 if (dad_failed)
1383 ifp->flags |= IFA_F_DADFAILED;
1384 spin_unlock_bh(&ifp->lock);
1385 in6_ifa_put(ifp);
1386 #ifdef CONFIG_IPV6_PRIVACY
1387 } else if (ifp->flags&IFA_F_TEMPORARY) {
1388 struct inet6_ifaddr *ifpub;
1389 spin_lock_bh(&ifp->lock);
1390 ifpub = ifp->ifpub;
1391 if (ifpub) {
1392 in6_ifa_hold(ifpub);
1393 spin_unlock_bh(&ifp->lock);
1394 ipv6_create_tempaddr(ifpub, ifp);
1395 in6_ifa_put(ifpub);
1396 } else {
1397 spin_unlock_bh(&ifp->lock);
1399 ipv6_del_addr(ifp);
1400 #endif
1401 } else
1402 ipv6_del_addr(ifp);
1405 void addrconf_dad_failure(struct inet6_ifaddr *ifp)
1407 struct inet6_dev *idev = ifp->idev;
1409 if (net_ratelimit())
1410 printk(KERN_INFO "%s: IPv6 duplicate address %pI6c detected!\n",
1411 ifp->idev->dev->name, &ifp->addr);
1413 if (idev->cnf.accept_dad > 1 && !idev->cnf.disable_ipv6) {
1414 struct in6_addr addr;
1416 addr.s6_addr32[0] = htonl(0xfe800000);
1417 addr.s6_addr32[1] = 0;
1419 if (!ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) &&
1420 ipv6_addr_equal(&ifp->addr, &addr)) {
1421 /* DAD failed for link-local based on MAC address */
1422 idev->cnf.disable_ipv6 = 1;
1424 printk(KERN_INFO "%s: IPv6 being disabled!\n",
1425 ifp->idev->dev->name);
1429 addrconf_dad_stop(ifp, 1);
1432 /* Join to solicited addr multicast group. */
1434 void addrconf_join_solict(struct net_device *dev, struct in6_addr *addr)
1436 struct in6_addr maddr;
1438 if (dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1439 return;
1441 addrconf_addr_solict_mult(addr, &maddr);
1442 ipv6_dev_mc_inc(dev, &maddr);
1445 void addrconf_leave_solict(struct inet6_dev *idev, struct in6_addr *addr)
1447 struct in6_addr maddr;
1449 if (idev->dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1450 return;
1452 addrconf_addr_solict_mult(addr, &maddr);
1453 __ipv6_dev_mc_dec(idev, &maddr);
1456 static void addrconf_join_anycast(struct inet6_ifaddr *ifp)
1458 struct in6_addr addr;
1459 ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1460 if (ipv6_addr_any(&addr))
1461 return;
1462 ipv6_dev_ac_inc(ifp->idev->dev, &addr);
1465 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp)
1467 struct in6_addr addr;
1468 ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1469 if (ipv6_addr_any(&addr))
1470 return;
1471 __ipv6_dev_ac_dec(ifp->idev, &addr);
1474 static int addrconf_ifid_eui48(u8 *eui, struct net_device *dev)
1476 if (dev->addr_len != ETH_ALEN)
1477 return -1;
1478 memcpy(eui, dev->dev_addr, 3);
1479 memcpy(eui + 5, dev->dev_addr + 3, 3);
1482 * The zSeries OSA network cards can be shared among various
1483 * OS instances, but the OSA cards have only one MAC address.
1484 * This leads to duplicate address conflicts in conjunction
1485 * with IPv6 if more than one instance uses the same card.
1487 * The driver for these cards can deliver a unique 16-bit
1488 * identifier for each instance sharing the same card. It is
1489 * placed instead of 0xFFFE in the interface identifier. The
1490 * "u" bit of the interface identifier is not inverted in this
1491 * case. Hence the resulting interface identifier has local
1492 * scope according to RFC2373.
1494 if (dev->dev_id) {
1495 eui[3] = (dev->dev_id >> 8) & 0xFF;
1496 eui[4] = dev->dev_id & 0xFF;
1497 } else {
1498 eui[3] = 0xFF;
1499 eui[4] = 0xFE;
1500 eui[0] ^= 2;
1502 return 0;
1505 static int addrconf_ifid_arcnet(u8 *eui, struct net_device *dev)
1507 /* XXX: inherit EUI-64 from other interface -- yoshfuji */
1508 if (dev->addr_len != ARCNET_ALEN)
1509 return -1;
1510 memset(eui, 0, 7);
1511 eui[7] = *(u8*)dev->dev_addr;
1512 return 0;
1515 static int addrconf_ifid_infiniband(u8 *eui, struct net_device *dev)
1517 if (dev->addr_len != INFINIBAND_ALEN)
1518 return -1;
1519 memcpy(eui, dev->dev_addr + 12, 8);
1520 eui[0] |= 2;
1521 return 0;
1524 int __ipv6_isatap_ifid(u8 *eui, __be32 addr)
1526 if (addr == 0)
1527 return -1;
1528 eui[0] = (ipv4_is_zeronet(addr) || ipv4_is_private_10(addr) ||
1529 ipv4_is_loopback(addr) || ipv4_is_linklocal_169(addr) ||
1530 ipv4_is_private_172(addr) || ipv4_is_test_192(addr) ||
1531 ipv4_is_anycast_6to4(addr) || ipv4_is_private_192(addr) ||
1532 ipv4_is_test_198(addr) || ipv4_is_multicast(addr) ||
1533 ipv4_is_lbcast(addr)) ? 0x00 : 0x02;
1534 eui[1] = 0;
1535 eui[2] = 0x5E;
1536 eui[3] = 0xFE;
1537 memcpy(eui + 4, &addr, 4);
1538 return 0;
1540 EXPORT_SYMBOL(__ipv6_isatap_ifid);
1542 static int addrconf_ifid_sit(u8 *eui, struct net_device *dev)
1544 if (dev->priv_flags & IFF_ISATAP)
1545 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
1546 return -1;
1549 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev)
1551 switch (dev->type) {
1552 case ARPHRD_ETHER:
1553 case ARPHRD_FDDI:
1554 case ARPHRD_IEEE802_TR:
1555 return addrconf_ifid_eui48(eui, dev);
1556 case ARPHRD_ARCNET:
1557 return addrconf_ifid_arcnet(eui, dev);
1558 case ARPHRD_INFINIBAND:
1559 return addrconf_ifid_infiniband(eui, dev);
1560 case ARPHRD_SIT:
1561 return addrconf_ifid_sit(eui, dev);
1563 return -1;
1566 static int ipv6_inherit_eui64(u8 *eui, struct inet6_dev *idev)
1568 int err = -1;
1569 struct inet6_ifaddr *ifp;
1571 read_lock_bh(&idev->lock);
1572 for (ifp=idev->addr_list; ifp; ifp=ifp->if_next) {
1573 if (ifp->scope == IFA_LINK && !(ifp->flags&IFA_F_TENTATIVE)) {
1574 memcpy(eui, ifp->addr.s6_addr+8, 8);
1575 err = 0;
1576 break;
1579 read_unlock_bh(&idev->lock);
1580 return err;
1583 #ifdef CONFIG_IPV6_PRIVACY
1584 /* (re)generation of randomized interface identifier (RFC 3041 3.2, 3.5) */
1585 static int __ipv6_regen_rndid(struct inet6_dev *idev)
1587 regen:
1588 get_random_bytes(idev->rndid, sizeof(idev->rndid));
1589 idev->rndid[0] &= ~0x02;
1592 * <draft-ietf-ipngwg-temp-addresses-v2-00.txt>:
1593 * check if generated address is not inappropriate
1595 * - Reserved subnet anycast (RFC 2526)
1596 * 11111101 11....11 1xxxxxxx
1597 * - ISATAP (RFC4214) 6.1
1598 * 00-00-5E-FE-xx-xx-xx-xx
1599 * - value 0
1600 * - XXX: already assigned to an address on the device
1602 if (idev->rndid[0] == 0xfd &&
1603 (idev->rndid[1]&idev->rndid[2]&idev->rndid[3]&idev->rndid[4]&idev->rndid[5]&idev->rndid[6]) == 0xff &&
1604 (idev->rndid[7]&0x80))
1605 goto regen;
1606 if ((idev->rndid[0]|idev->rndid[1]) == 0) {
1607 if (idev->rndid[2] == 0x5e && idev->rndid[3] == 0xfe)
1608 goto regen;
1609 if ((idev->rndid[2]|idev->rndid[3]|idev->rndid[4]|idev->rndid[5]|idev->rndid[6]|idev->rndid[7]) == 0x00)
1610 goto regen;
1613 return 0;
1616 static void ipv6_regen_rndid(unsigned long data)
1618 struct inet6_dev *idev = (struct inet6_dev *) data;
1619 unsigned long expires;
1621 rcu_read_lock_bh();
1622 write_lock_bh(&idev->lock);
1624 if (idev->dead)
1625 goto out;
1627 if (__ipv6_regen_rndid(idev) < 0)
1628 goto out;
1630 expires = jiffies +
1631 idev->cnf.temp_prefered_lft * HZ -
1632 idev->cnf.regen_max_retry * idev->cnf.dad_transmits * idev->nd_parms->retrans_time - desync_factor;
1633 if (time_before(expires, jiffies)) {
1634 printk(KERN_WARNING
1635 "ipv6_regen_rndid(): too short regeneration interval; timer disabled for %s.\n",
1636 idev->dev->name);
1637 goto out;
1640 if (!mod_timer(&idev->regen_timer, expires))
1641 in6_dev_hold(idev);
1643 out:
1644 write_unlock_bh(&idev->lock);
1645 rcu_read_unlock_bh();
1646 in6_dev_put(idev);
1649 static int __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr) {
1650 int ret = 0;
1652 if (tmpaddr && memcmp(idev->rndid, &tmpaddr->s6_addr[8], 8) == 0)
1653 ret = __ipv6_regen_rndid(idev);
1654 return ret;
1656 #endif
1659 * Add prefix route.
1662 static void
1663 addrconf_prefix_route(struct in6_addr *pfx, int plen, struct net_device *dev,
1664 unsigned long expires, u32 flags)
1666 struct fib6_config cfg = {
1667 .fc_table = RT6_TABLE_PREFIX,
1668 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1669 .fc_ifindex = dev->ifindex,
1670 .fc_expires = expires,
1671 .fc_dst_len = plen,
1672 .fc_flags = RTF_UP | flags,
1673 .fc_nlinfo.nl_net = dev_net(dev),
1674 .fc_protocol = RTPROT_KERNEL,
1677 ipv6_addr_copy(&cfg.fc_dst, pfx);
1679 /* Prevent useless cloning on PtP SIT.
1680 This thing is done here expecting that the whole
1681 class of non-broadcast devices need not cloning.
1683 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
1684 if (dev->type == ARPHRD_SIT && (dev->flags & IFF_POINTOPOINT))
1685 cfg.fc_flags |= RTF_NONEXTHOP;
1686 #endif
1688 ip6_route_add(&cfg);
1691 /* Create "default" multicast route to the interface */
1693 static void addrconf_add_mroute(struct net_device *dev)
1695 struct fib6_config cfg = {
1696 .fc_table = RT6_TABLE_LOCAL,
1697 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1698 .fc_ifindex = dev->ifindex,
1699 .fc_dst_len = 8,
1700 .fc_flags = RTF_UP,
1701 .fc_nlinfo.nl_net = dev_net(dev),
1704 ipv6_addr_set(&cfg.fc_dst, htonl(0xFF000000), 0, 0, 0);
1706 ip6_route_add(&cfg);
1709 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
1710 static void sit_route_add(struct net_device *dev)
1712 struct fib6_config cfg = {
1713 .fc_table = RT6_TABLE_MAIN,
1714 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1715 .fc_ifindex = dev->ifindex,
1716 .fc_dst_len = 96,
1717 .fc_flags = RTF_UP | RTF_NONEXTHOP,
1718 .fc_nlinfo.nl_net = dev_net(dev),
1721 /* prefix length - 96 bits "::d.d.d.d" */
1722 ip6_route_add(&cfg);
1724 #endif
1726 static void addrconf_add_lroute(struct net_device *dev)
1728 struct in6_addr addr;
1730 ipv6_addr_set(&addr, htonl(0xFE800000), 0, 0, 0);
1731 addrconf_prefix_route(&addr, 64, dev, 0, 0);
1734 static struct inet6_dev *addrconf_add_dev(struct net_device *dev)
1736 struct inet6_dev *idev;
1738 ASSERT_RTNL();
1740 if ((idev = ipv6_find_idev(dev)) == NULL)
1741 return NULL;
1743 /* Add default multicast route */
1744 addrconf_add_mroute(dev);
1746 /* Add link local route */
1747 addrconf_add_lroute(dev);
1748 return idev;
1751 void addrconf_prefix_rcv(struct net_device *dev, u8 *opt, int len)
1753 struct prefix_info *pinfo;
1754 __u32 valid_lft;
1755 __u32 prefered_lft;
1756 int addr_type;
1757 struct inet6_dev *in6_dev;
1758 struct net *net = dev_net(dev);
1760 pinfo = (struct prefix_info *) opt;
1762 if (len < sizeof(struct prefix_info)) {
1763 ADBG(("addrconf: prefix option too short\n"));
1764 return;
1768 * Validation checks ([ADDRCONF], page 19)
1771 addr_type = ipv6_addr_type(&pinfo->prefix);
1773 if (addr_type & (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL))
1774 return;
1776 valid_lft = ntohl(pinfo->valid);
1777 prefered_lft = ntohl(pinfo->prefered);
1779 if (prefered_lft > valid_lft) {
1780 if (net_ratelimit())
1781 printk(KERN_WARNING "addrconf: prefix option has invalid lifetime\n");
1782 return;
1785 in6_dev = in6_dev_get(dev);
1787 if (in6_dev == NULL) {
1788 if (net_ratelimit())
1789 printk(KERN_DEBUG "addrconf: device %s not configured\n", dev->name);
1790 return;
1794 * Two things going on here:
1795 * 1) Add routes for on-link prefixes
1796 * 2) Configure prefixes with the auto flag set
1799 if (pinfo->onlink) {
1800 struct rt6_info *rt;
1801 unsigned long rt_expires;
1803 /* Avoid arithmetic overflow. Really, we could
1804 * save rt_expires in seconds, likely valid_lft,
1805 * but it would require division in fib gc, that it
1806 * not good.
1808 if (HZ > USER_HZ)
1809 rt_expires = addrconf_timeout_fixup(valid_lft, HZ);
1810 else
1811 rt_expires = addrconf_timeout_fixup(valid_lft, USER_HZ);
1813 if (addrconf_finite_timeout(rt_expires))
1814 rt_expires *= HZ;
1816 rt = rt6_lookup(net, &pinfo->prefix, NULL,
1817 dev->ifindex, 1);
1819 if (rt && addrconf_is_prefix_route(rt)) {
1820 /* Autoconf prefix route */
1821 if (valid_lft == 0) {
1822 ip6_del_rt(rt);
1823 rt = NULL;
1824 } else if (addrconf_finite_timeout(rt_expires)) {
1825 /* not infinity */
1826 rt->rt6i_expires = jiffies + rt_expires;
1827 rt->rt6i_flags |= RTF_EXPIRES;
1828 } else {
1829 rt->rt6i_flags &= ~RTF_EXPIRES;
1830 rt->rt6i_expires = 0;
1832 } else if (valid_lft) {
1833 clock_t expires = 0;
1834 int flags = RTF_ADDRCONF | RTF_PREFIX_RT;
1835 if (addrconf_finite_timeout(rt_expires)) {
1836 /* not infinity */
1837 flags |= RTF_EXPIRES;
1838 expires = jiffies_to_clock_t(rt_expires);
1840 addrconf_prefix_route(&pinfo->prefix, pinfo->prefix_len,
1841 dev, expires, flags);
1843 if (rt)
1844 dst_release(&rt->u.dst);
1847 /* Try to figure out our local address for this prefix */
1849 if (pinfo->autoconf && in6_dev->cnf.autoconf) {
1850 struct inet6_ifaddr * ifp;
1851 struct in6_addr addr;
1852 int create = 0, update_lft = 0;
1854 if (pinfo->prefix_len == 64) {
1855 memcpy(&addr, &pinfo->prefix, 8);
1856 if (ipv6_generate_eui64(addr.s6_addr + 8, dev) &&
1857 ipv6_inherit_eui64(addr.s6_addr + 8, in6_dev)) {
1858 in6_dev_put(in6_dev);
1859 return;
1861 goto ok;
1863 if (net_ratelimit())
1864 printk(KERN_DEBUG "IPv6 addrconf: prefix with wrong length %d\n",
1865 pinfo->prefix_len);
1866 in6_dev_put(in6_dev);
1867 return;
1871 ifp = ipv6_get_ifaddr(net, &addr, dev, 1);
1873 if (ifp == NULL && valid_lft) {
1874 int max_addresses = in6_dev->cnf.max_addresses;
1875 u32 addr_flags = 0;
1877 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1878 if (in6_dev->cnf.optimistic_dad &&
1879 !net->ipv6.devconf_all->forwarding)
1880 addr_flags = IFA_F_OPTIMISTIC;
1881 #endif
1883 /* Do not allow to create too much of autoconfigured
1884 * addresses; this would be too easy way to crash kernel.
1886 if (!max_addresses ||
1887 ipv6_count_addresses(in6_dev) < max_addresses)
1888 ifp = ipv6_add_addr(in6_dev, &addr, pinfo->prefix_len,
1889 addr_type&IPV6_ADDR_SCOPE_MASK,
1890 addr_flags);
1892 if (!ifp || IS_ERR(ifp)) {
1893 in6_dev_put(in6_dev);
1894 return;
1897 update_lft = create = 1;
1898 ifp->cstamp = jiffies;
1899 addrconf_dad_start(ifp, RTF_ADDRCONF|RTF_PREFIX_RT);
1902 if (ifp) {
1903 int flags;
1904 unsigned long now;
1905 #ifdef CONFIG_IPV6_PRIVACY
1906 struct inet6_ifaddr *ift;
1907 #endif
1908 u32 stored_lft;
1910 /* update lifetime (RFC2462 5.5.3 e) */
1911 spin_lock(&ifp->lock);
1912 now = jiffies;
1913 if (ifp->valid_lft > (now - ifp->tstamp) / HZ)
1914 stored_lft = ifp->valid_lft - (now - ifp->tstamp) / HZ;
1915 else
1916 stored_lft = 0;
1917 if (!update_lft && stored_lft) {
1918 if (valid_lft > MIN_VALID_LIFETIME ||
1919 valid_lft > stored_lft)
1920 update_lft = 1;
1921 else if (stored_lft <= MIN_VALID_LIFETIME) {
1922 /* valid_lft <= stored_lft is always true */
1924 * RFC 4862 Section 5.5.3e:
1925 * "Note that the preferred lifetime of
1926 * the corresponding address is always
1927 * reset to the Preferred Lifetime in
1928 * the received Prefix Information
1929 * option, regardless of whether the
1930 * valid lifetime is also reset or
1931 * ignored."
1933 * So if the preferred lifetime in
1934 * this advertisement is different
1935 * than what we have stored, but the
1936 * valid lifetime is invalid, just
1937 * reset prefered_lft.
1939 * We must set the valid lifetime
1940 * to the stored lifetime since we'll
1941 * be updating the timestamp below,
1942 * else we'll set it back to the
1943 * minumum.
1945 if (prefered_lft != ifp->prefered_lft) {
1946 valid_lft = stored_lft;
1947 update_lft = 1;
1949 } else {
1950 valid_lft = MIN_VALID_LIFETIME;
1951 if (valid_lft < prefered_lft)
1952 prefered_lft = valid_lft;
1953 update_lft = 1;
1957 if (update_lft) {
1958 ifp->valid_lft = valid_lft;
1959 ifp->prefered_lft = prefered_lft;
1960 ifp->tstamp = now;
1961 flags = ifp->flags;
1962 ifp->flags &= ~IFA_F_DEPRECATED;
1963 spin_unlock(&ifp->lock);
1965 if (!(flags&IFA_F_TENTATIVE))
1966 ipv6_ifa_notify(0, ifp);
1967 } else
1968 spin_unlock(&ifp->lock);
1970 #ifdef CONFIG_IPV6_PRIVACY
1971 read_lock_bh(&in6_dev->lock);
1972 /* update all temporary addresses in the list */
1973 for (ift=in6_dev->tempaddr_list; ift; ift=ift->tmp_next) {
1975 * When adjusting the lifetimes of an existing
1976 * temporary address, only lower the lifetimes.
1977 * Implementations must not increase the
1978 * lifetimes of an existing temporary address
1979 * when processing a Prefix Information Option.
1981 if (ifp != ift->ifpub)
1982 continue;
1984 spin_lock(&ift->lock);
1985 flags = ift->flags;
1986 if (ift->valid_lft > valid_lft &&
1987 ift->valid_lft - valid_lft > (jiffies - ift->tstamp) / HZ)
1988 ift->valid_lft = valid_lft + (jiffies - ift->tstamp) / HZ;
1989 if (ift->prefered_lft > prefered_lft &&
1990 ift->prefered_lft - prefered_lft > (jiffies - ift->tstamp) / HZ)
1991 ift->prefered_lft = prefered_lft + (jiffies - ift->tstamp) / HZ;
1992 spin_unlock(&ift->lock);
1993 if (!(flags&IFA_F_TENTATIVE))
1994 ipv6_ifa_notify(0, ift);
1997 if (create && in6_dev->cnf.use_tempaddr > 0) {
1999 * When a new public address is created as described in [ADDRCONF],
2000 * also create a new temporary address.
2002 read_unlock_bh(&in6_dev->lock);
2003 ipv6_create_tempaddr(ifp, NULL);
2004 } else {
2005 read_unlock_bh(&in6_dev->lock);
2007 #endif
2008 in6_ifa_put(ifp);
2009 addrconf_verify(0);
2012 inet6_prefix_notify(RTM_NEWPREFIX, in6_dev, pinfo);
2013 in6_dev_put(in6_dev);
2017 * Set destination address.
2018 * Special case for SIT interfaces where we create a new "virtual"
2019 * device.
2021 int addrconf_set_dstaddr(struct net *net, void __user *arg)
2023 struct in6_ifreq ireq;
2024 struct net_device *dev;
2025 int err = -EINVAL;
2027 rtnl_lock();
2029 err = -EFAULT;
2030 if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2031 goto err_exit;
2033 dev = __dev_get_by_index(net, ireq.ifr6_ifindex);
2035 err = -ENODEV;
2036 if (dev == NULL)
2037 goto err_exit;
2039 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2040 if (dev->type == ARPHRD_SIT) {
2041 const struct net_device_ops *ops = dev->netdev_ops;
2042 struct ifreq ifr;
2043 struct ip_tunnel_parm p;
2045 err = -EADDRNOTAVAIL;
2046 if (!(ipv6_addr_type(&ireq.ifr6_addr) & IPV6_ADDR_COMPATv4))
2047 goto err_exit;
2049 memset(&p, 0, sizeof(p));
2050 p.iph.daddr = ireq.ifr6_addr.s6_addr32[3];
2051 p.iph.saddr = 0;
2052 p.iph.version = 4;
2053 p.iph.ihl = 5;
2054 p.iph.protocol = IPPROTO_IPV6;
2055 p.iph.ttl = 64;
2056 ifr.ifr_ifru.ifru_data = (__force void __user *)&p;
2058 if (ops->ndo_do_ioctl) {
2059 mm_segment_t oldfs = get_fs();
2061 set_fs(KERNEL_DS);
2062 err = ops->ndo_do_ioctl(dev, &ifr, SIOCADDTUNNEL);
2063 set_fs(oldfs);
2064 } else
2065 err = -EOPNOTSUPP;
2067 if (err == 0) {
2068 err = -ENOBUFS;
2069 dev = __dev_get_by_name(net, p.name);
2070 if (!dev)
2071 goto err_exit;
2072 err = dev_open(dev);
2075 #endif
2077 err_exit:
2078 rtnl_unlock();
2079 return err;
2083 * Manual configuration of address on an interface
2085 static int inet6_addr_add(struct net *net, int ifindex, struct in6_addr *pfx,
2086 unsigned int plen, __u8 ifa_flags, __u32 prefered_lft,
2087 __u32 valid_lft)
2089 struct inet6_ifaddr *ifp;
2090 struct inet6_dev *idev;
2091 struct net_device *dev;
2092 int scope;
2093 u32 flags;
2094 clock_t expires;
2095 unsigned long timeout;
2097 ASSERT_RTNL();
2099 if (plen > 128)
2100 return -EINVAL;
2102 /* check the lifetime */
2103 if (!valid_lft || prefered_lft > valid_lft)
2104 return -EINVAL;
2106 dev = __dev_get_by_index(net, ifindex);
2107 if (!dev)
2108 return -ENODEV;
2110 if ((idev = addrconf_add_dev(dev)) == NULL)
2111 return -ENOBUFS;
2113 scope = ipv6_addr_scope(pfx);
2115 timeout = addrconf_timeout_fixup(valid_lft, HZ);
2116 if (addrconf_finite_timeout(timeout)) {
2117 expires = jiffies_to_clock_t(timeout * HZ);
2118 valid_lft = timeout;
2119 flags = RTF_EXPIRES;
2120 } else {
2121 expires = 0;
2122 flags = 0;
2123 ifa_flags |= IFA_F_PERMANENT;
2126 timeout = addrconf_timeout_fixup(prefered_lft, HZ);
2127 if (addrconf_finite_timeout(timeout)) {
2128 if (timeout == 0)
2129 ifa_flags |= IFA_F_DEPRECATED;
2130 prefered_lft = timeout;
2133 ifp = ipv6_add_addr(idev, pfx, plen, scope, ifa_flags);
2135 if (!IS_ERR(ifp)) {
2136 spin_lock_bh(&ifp->lock);
2137 ifp->valid_lft = valid_lft;
2138 ifp->prefered_lft = prefered_lft;
2139 ifp->tstamp = jiffies;
2140 spin_unlock_bh(&ifp->lock);
2142 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, dev,
2143 expires, flags);
2145 * Note that section 3.1 of RFC 4429 indicates
2146 * that the Optimistic flag should not be set for
2147 * manually configured addresses
2149 addrconf_dad_start(ifp, 0);
2150 in6_ifa_put(ifp);
2151 addrconf_verify(0);
2152 return 0;
2155 return PTR_ERR(ifp);
2158 static int inet6_addr_del(struct net *net, int ifindex, struct in6_addr *pfx,
2159 unsigned int plen)
2161 struct inet6_ifaddr *ifp;
2162 struct inet6_dev *idev;
2163 struct net_device *dev;
2165 if (plen > 128)
2166 return -EINVAL;
2168 dev = __dev_get_by_index(net, ifindex);
2169 if (!dev)
2170 return -ENODEV;
2172 if ((idev = __in6_dev_get(dev)) == NULL)
2173 return -ENXIO;
2175 read_lock_bh(&idev->lock);
2176 for (ifp = idev->addr_list; ifp; ifp=ifp->if_next) {
2177 if (ifp->prefix_len == plen &&
2178 ipv6_addr_equal(pfx, &ifp->addr)) {
2179 in6_ifa_hold(ifp);
2180 read_unlock_bh(&idev->lock);
2182 ipv6_del_addr(ifp);
2184 /* If the last address is deleted administratively,
2185 disable IPv6 on this interface.
2187 if (idev->addr_list == NULL)
2188 addrconf_ifdown(idev->dev, 1);
2189 return 0;
2192 read_unlock_bh(&idev->lock);
2193 return -EADDRNOTAVAIL;
2197 int addrconf_add_ifaddr(struct net *net, void __user *arg)
2199 struct in6_ifreq ireq;
2200 int err;
2202 if (!capable(CAP_NET_ADMIN))
2203 return -EPERM;
2205 if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2206 return -EFAULT;
2208 rtnl_lock();
2209 err = inet6_addr_add(net, ireq.ifr6_ifindex, &ireq.ifr6_addr,
2210 ireq.ifr6_prefixlen, IFA_F_PERMANENT,
2211 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
2212 rtnl_unlock();
2213 return err;
2216 int addrconf_del_ifaddr(struct net *net, void __user *arg)
2218 struct in6_ifreq ireq;
2219 int err;
2221 if (!capable(CAP_NET_ADMIN))
2222 return -EPERM;
2224 if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2225 return -EFAULT;
2227 rtnl_lock();
2228 err = inet6_addr_del(net, ireq.ifr6_ifindex, &ireq.ifr6_addr,
2229 ireq.ifr6_prefixlen);
2230 rtnl_unlock();
2231 return err;
2234 static void add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
2235 int plen, int scope)
2237 struct inet6_ifaddr *ifp;
2239 ifp = ipv6_add_addr(idev, addr, plen, scope, IFA_F_PERMANENT);
2240 if (!IS_ERR(ifp)) {
2241 spin_lock_bh(&ifp->lock);
2242 ifp->flags &= ~IFA_F_TENTATIVE;
2243 spin_unlock_bh(&ifp->lock);
2244 ipv6_ifa_notify(RTM_NEWADDR, ifp);
2245 in6_ifa_put(ifp);
2249 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2250 static void sit_add_v4_addrs(struct inet6_dev *idev)
2252 struct in6_addr addr;
2253 struct net_device *dev;
2254 struct net *net = dev_net(idev->dev);
2255 int scope;
2257 ASSERT_RTNL();
2259 memset(&addr, 0, sizeof(struct in6_addr));
2260 memcpy(&addr.s6_addr32[3], idev->dev->dev_addr, 4);
2262 if (idev->dev->flags&IFF_POINTOPOINT) {
2263 addr.s6_addr32[0] = htonl(0xfe800000);
2264 scope = IFA_LINK;
2265 } else {
2266 scope = IPV6_ADDR_COMPATv4;
2269 if (addr.s6_addr32[3]) {
2270 add_addr(idev, &addr, 128, scope);
2271 return;
2274 for_each_netdev(net, dev) {
2275 struct in_device * in_dev = __in_dev_get_rtnl(dev);
2276 if (in_dev && (dev->flags & IFF_UP)) {
2277 struct in_ifaddr * ifa;
2279 int flag = scope;
2281 for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) {
2282 int plen;
2284 addr.s6_addr32[3] = ifa->ifa_local;
2286 if (ifa->ifa_scope == RT_SCOPE_LINK)
2287 continue;
2288 if (ifa->ifa_scope >= RT_SCOPE_HOST) {
2289 if (idev->dev->flags&IFF_POINTOPOINT)
2290 continue;
2291 flag |= IFA_HOST;
2293 if (idev->dev->flags&IFF_POINTOPOINT)
2294 plen = 64;
2295 else
2296 plen = 96;
2298 add_addr(idev, &addr, plen, flag);
2303 #endif
2305 static void init_loopback(struct net_device *dev)
2307 struct inet6_dev *idev;
2309 /* ::1 */
2311 ASSERT_RTNL();
2313 if ((idev = ipv6_find_idev(dev)) == NULL) {
2314 printk(KERN_DEBUG "init loopback: add_dev failed\n");
2315 return;
2318 add_addr(idev, &in6addr_loopback, 128, IFA_HOST);
2321 static void addrconf_add_linklocal(struct inet6_dev *idev, struct in6_addr *addr)
2323 struct inet6_ifaddr * ifp;
2324 u32 addr_flags = IFA_F_PERMANENT;
2326 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2327 if (idev->cnf.optimistic_dad &&
2328 !dev_net(idev->dev)->ipv6.devconf_all->forwarding)
2329 addr_flags |= IFA_F_OPTIMISTIC;
2330 #endif
2333 ifp = ipv6_add_addr(idev, addr, 64, IFA_LINK, addr_flags);
2334 if (!IS_ERR(ifp)) {
2335 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, idev->dev, 0, 0);
2336 addrconf_dad_start(ifp, 0);
2337 in6_ifa_put(ifp);
2341 static void addrconf_dev_config(struct net_device *dev)
2343 struct in6_addr addr;
2344 struct inet6_dev * idev;
2346 ASSERT_RTNL();
2348 if ((dev->type != ARPHRD_ETHER) &&
2349 (dev->type != ARPHRD_FDDI) &&
2350 (dev->type != ARPHRD_IEEE802_TR) &&
2351 (dev->type != ARPHRD_ARCNET) &&
2352 (dev->type != ARPHRD_INFINIBAND)) {
2353 /* Alas, we support only Ethernet autoconfiguration. */
2354 return;
2357 idev = addrconf_add_dev(dev);
2358 if (idev == NULL)
2359 return;
2361 memset(&addr, 0, sizeof(struct in6_addr));
2362 addr.s6_addr32[0] = htonl(0xFE800000);
2364 if (ipv6_generate_eui64(addr.s6_addr + 8, dev) == 0)
2365 addrconf_add_linklocal(idev, &addr);
2368 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2369 static void addrconf_sit_config(struct net_device *dev)
2371 struct inet6_dev *idev;
2373 ASSERT_RTNL();
2376 * Configure the tunnel with one of our IPv4
2377 * addresses... we should configure all of
2378 * our v4 addrs in the tunnel
2381 if ((idev = ipv6_find_idev(dev)) == NULL) {
2382 printk(KERN_DEBUG "init sit: add_dev failed\n");
2383 return;
2386 if (dev->priv_flags & IFF_ISATAP) {
2387 struct in6_addr addr;
2389 ipv6_addr_set(&addr, htonl(0xFE800000), 0, 0, 0);
2390 addrconf_prefix_route(&addr, 64, dev, 0, 0);
2391 if (!ipv6_generate_eui64(addr.s6_addr + 8, dev))
2392 addrconf_add_linklocal(idev, &addr);
2393 return;
2396 sit_add_v4_addrs(idev);
2398 if (dev->flags&IFF_POINTOPOINT) {
2399 addrconf_add_mroute(dev);
2400 addrconf_add_lroute(dev);
2401 } else
2402 sit_route_add(dev);
2404 #endif
2406 static inline int
2407 ipv6_inherit_linklocal(struct inet6_dev *idev, struct net_device *link_dev)
2409 struct in6_addr lladdr;
2411 if (!ipv6_get_lladdr(link_dev, &lladdr, IFA_F_TENTATIVE)) {
2412 addrconf_add_linklocal(idev, &lladdr);
2413 return 0;
2415 return -1;
2418 static void ip6_tnl_add_linklocal(struct inet6_dev *idev)
2420 struct net_device *link_dev;
2421 struct net *net = dev_net(idev->dev);
2423 /* first try to inherit the link-local address from the link device */
2424 if (idev->dev->iflink &&
2425 (link_dev = __dev_get_by_index(net, idev->dev->iflink))) {
2426 if (!ipv6_inherit_linklocal(idev, link_dev))
2427 return;
2429 /* then try to inherit it from any device */
2430 for_each_netdev(net, link_dev) {
2431 if (!ipv6_inherit_linklocal(idev, link_dev))
2432 return;
2434 printk(KERN_DEBUG "init ip6-ip6: add_linklocal failed\n");
2438 * Autoconfigure tunnel with a link-local address so routing protocols,
2439 * DHCPv6, MLD etc. can be run over the virtual link
2442 static void addrconf_ip6_tnl_config(struct net_device *dev)
2444 struct inet6_dev *idev;
2446 ASSERT_RTNL();
2448 if ((idev = addrconf_add_dev(dev)) == NULL) {
2449 printk(KERN_DEBUG "init ip6-ip6: add_dev failed\n");
2450 return;
2452 ip6_tnl_add_linklocal(idev);
2455 static int addrconf_notify(struct notifier_block *this, unsigned long event,
2456 void * data)
2458 struct net_device *dev = (struct net_device *) data;
2459 struct inet6_dev *idev = __in6_dev_get(dev);
2460 int run_pending = 0;
2461 int err;
2463 switch(event) {
2464 case NETDEV_REGISTER:
2465 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2466 idev = ipv6_add_dev(dev);
2467 if (!idev)
2468 return notifier_from_errno(-ENOMEM);
2470 break;
2471 case NETDEV_UP:
2472 case NETDEV_CHANGE:
2473 if (dev->flags & IFF_SLAVE)
2474 break;
2476 if (event == NETDEV_UP) {
2477 if (!addrconf_qdisc_ok(dev)) {
2478 /* device is not ready yet. */
2479 printk(KERN_INFO
2480 "ADDRCONF(NETDEV_UP): %s: "
2481 "link is not ready\n",
2482 dev->name);
2483 break;
2486 if (!idev && dev->mtu >= IPV6_MIN_MTU)
2487 idev = ipv6_add_dev(dev);
2489 if (idev) {
2490 idev->if_flags |= IF_READY;
2491 run_pending = 1;
2493 } else {
2494 if (!addrconf_qdisc_ok(dev)) {
2495 /* device is still not ready. */
2496 break;
2499 if (idev) {
2500 if (idev->if_flags & IF_READY) {
2501 /* device is already configured. */
2502 break;
2504 idev->if_flags |= IF_READY;
2507 printk(KERN_INFO
2508 "ADDRCONF(NETDEV_CHANGE): %s: "
2509 "link becomes ready\n",
2510 dev->name);
2512 run_pending = 1;
2515 switch(dev->type) {
2516 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2517 case ARPHRD_SIT:
2518 addrconf_sit_config(dev);
2519 break;
2520 #endif
2521 case ARPHRD_TUNNEL6:
2522 addrconf_ip6_tnl_config(dev);
2523 break;
2524 case ARPHRD_LOOPBACK:
2525 init_loopback(dev);
2526 break;
2528 default:
2529 addrconf_dev_config(dev);
2530 break;
2532 if (idev) {
2533 if (run_pending)
2534 addrconf_dad_run(idev);
2536 /* If the MTU changed during the interface down, when the
2537 interface up, the changed MTU must be reflected in the
2538 idev as well as routers.
2540 if (idev->cnf.mtu6 != dev->mtu && dev->mtu >= IPV6_MIN_MTU) {
2541 rt6_mtu_change(dev, dev->mtu);
2542 idev->cnf.mtu6 = dev->mtu;
2544 idev->tstamp = jiffies;
2545 inet6_ifinfo_notify(RTM_NEWLINK, idev);
2546 /* If the changed mtu during down is lower than IPV6_MIN_MTU
2547 stop IPv6 on this interface.
2549 if (dev->mtu < IPV6_MIN_MTU)
2550 addrconf_ifdown(dev, event != NETDEV_DOWN);
2552 break;
2554 case NETDEV_CHANGEMTU:
2555 if (idev && dev->mtu >= IPV6_MIN_MTU) {
2556 rt6_mtu_change(dev, dev->mtu);
2557 idev->cnf.mtu6 = dev->mtu;
2558 break;
2561 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2562 idev = ipv6_add_dev(dev);
2563 if (idev)
2564 break;
2567 /* MTU falled under IPV6_MIN_MTU. Stop IPv6 on this interface. */
2569 case NETDEV_DOWN:
2570 case NETDEV_UNREGISTER:
2572 * Remove all addresses from this interface.
2574 addrconf_ifdown(dev, event != NETDEV_DOWN);
2575 break;
2577 case NETDEV_CHANGENAME:
2578 if (idev) {
2579 snmp6_unregister_dev(idev);
2580 addrconf_sysctl_unregister(idev);
2581 addrconf_sysctl_register(idev);
2582 err = snmp6_register_dev(idev);
2583 if (err)
2584 return notifier_from_errno(err);
2586 break;
2587 case NETDEV_BONDING_OLDTYPE:
2588 case NETDEV_BONDING_NEWTYPE:
2589 addrconf_bonding_change(dev, event);
2590 break;
2593 return NOTIFY_OK;
2597 * addrconf module should be notified of a device going up
2599 static struct notifier_block ipv6_dev_notf = {
2600 .notifier_call = addrconf_notify,
2601 .priority = 0
2604 static void addrconf_bonding_change(struct net_device *dev, unsigned long event)
2606 struct inet6_dev *idev;
2607 ASSERT_RTNL();
2609 idev = __in6_dev_get(dev);
2611 if (event == NETDEV_BONDING_NEWTYPE)
2612 ipv6_mc_remap(idev);
2613 else if (event == NETDEV_BONDING_OLDTYPE)
2614 ipv6_mc_unmap(idev);
2617 static int addrconf_ifdown(struct net_device *dev, int how)
2619 struct inet6_dev *idev;
2620 struct inet6_ifaddr *ifa, **bifa;
2621 struct net *net = dev_net(dev);
2622 int i;
2624 ASSERT_RTNL();
2626 rt6_ifdown(net, dev);
2627 neigh_ifdown(&nd_tbl, dev);
2629 idev = __in6_dev_get(dev);
2630 if (idev == NULL)
2631 return -ENODEV;
2633 /* Step 1: remove reference to ipv6 device from parent device.
2634 Do not dev_put!
2636 if (how) {
2637 idev->dead = 1;
2639 /* protected by rtnl_lock */
2640 rcu_assign_pointer(dev->ip6_ptr, NULL);
2642 /* Step 1.5: remove snmp6 entry */
2643 snmp6_unregister_dev(idev);
2647 /* Step 2: clear hash table */
2648 for (i=0; i<IN6_ADDR_HSIZE; i++) {
2649 bifa = &inet6_addr_lst[i];
2651 write_lock_bh(&addrconf_hash_lock);
2652 while ((ifa = *bifa) != NULL) {
2653 if (ifa->idev == idev) {
2654 *bifa = ifa->lst_next;
2655 ifa->lst_next = NULL;
2656 addrconf_del_timer(ifa);
2657 in6_ifa_put(ifa);
2658 continue;
2660 bifa = &ifa->lst_next;
2662 write_unlock_bh(&addrconf_hash_lock);
2665 write_lock_bh(&idev->lock);
2667 /* Step 3: clear flags for stateless addrconf */
2668 if (!how)
2669 idev->if_flags &= ~(IF_RS_SENT|IF_RA_RCVD|IF_READY);
2671 /* Step 4: clear address list */
2672 #ifdef CONFIG_IPV6_PRIVACY
2673 if (how && del_timer(&idev->regen_timer))
2674 in6_dev_put(idev);
2676 /* clear tempaddr list */
2677 while ((ifa = idev->tempaddr_list) != NULL) {
2678 idev->tempaddr_list = ifa->tmp_next;
2679 ifa->tmp_next = NULL;
2680 ifa->dead = 1;
2681 write_unlock_bh(&idev->lock);
2682 spin_lock_bh(&ifa->lock);
2684 if (ifa->ifpub) {
2685 in6_ifa_put(ifa->ifpub);
2686 ifa->ifpub = NULL;
2688 spin_unlock_bh(&ifa->lock);
2689 in6_ifa_put(ifa);
2690 write_lock_bh(&idev->lock);
2692 #endif
2693 while ((ifa = idev->addr_list) != NULL) {
2694 idev->addr_list = ifa->if_next;
2695 ifa->if_next = NULL;
2696 ifa->dead = 1;
2697 addrconf_del_timer(ifa);
2698 write_unlock_bh(&idev->lock);
2700 __ipv6_ifa_notify(RTM_DELADDR, ifa);
2701 atomic_notifier_call_chain(&inet6addr_chain, NETDEV_DOWN, ifa);
2702 in6_ifa_put(ifa);
2704 write_lock_bh(&idev->lock);
2706 write_unlock_bh(&idev->lock);
2708 /* Step 5: Discard multicast list */
2710 if (how)
2711 ipv6_mc_destroy_dev(idev);
2712 else
2713 ipv6_mc_down(idev);
2715 idev->tstamp = jiffies;
2717 /* Shot the device (if unregistered) */
2719 if (how) {
2720 addrconf_sysctl_unregister(idev);
2721 neigh_parms_release(&nd_tbl, idev->nd_parms);
2722 neigh_ifdown(&nd_tbl, dev);
2723 in6_dev_put(idev);
2725 return 0;
2728 static void addrconf_rs_timer(unsigned long data)
2730 struct inet6_ifaddr *ifp = (struct inet6_ifaddr *) data;
2732 if (ifp->idev->cnf.forwarding)
2733 goto out;
2735 if (ifp->idev->if_flags & IF_RA_RCVD) {
2737 * Announcement received after solicitation
2738 * was sent
2740 goto out;
2743 spin_lock(&ifp->lock);
2744 if (ifp->probes++ < ifp->idev->cnf.rtr_solicits) {
2745 /* The wait after the last probe can be shorter */
2746 addrconf_mod_timer(ifp, AC_RS,
2747 (ifp->probes == ifp->idev->cnf.rtr_solicits) ?
2748 ifp->idev->cnf.rtr_solicit_delay :
2749 ifp->idev->cnf.rtr_solicit_interval);
2750 spin_unlock(&ifp->lock);
2752 ndisc_send_rs(ifp->idev->dev, &ifp->addr, &in6addr_linklocal_allrouters);
2753 } else {
2754 spin_unlock(&ifp->lock);
2756 * Note: we do not support deprecated "all on-link"
2757 * assumption any longer.
2759 printk(KERN_DEBUG "%s: no IPv6 routers present\n",
2760 ifp->idev->dev->name);
2763 out:
2764 in6_ifa_put(ifp);
2768 * Duplicate Address Detection
2770 static void addrconf_dad_kick(struct inet6_ifaddr *ifp)
2772 unsigned long rand_num;
2773 struct inet6_dev *idev = ifp->idev;
2775 if (ifp->flags & IFA_F_OPTIMISTIC)
2776 rand_num = 0;
2777 else
2778 rand_num = net_random() % (idev->cnf.rtr_solicit_delay ? : 1);
2780 ifp->probes = idev->cnf.dad_transmits;
2781 addrconf_mod_timer(ifp, AC_DAD, rand_num);
2784 static void addrconf_dad_start(struct inet6_ifaddr *ifp, u32 flags)
2786 struct inet6_dev *idev = ifp->idev;
2787 struct net_device *dev = idev->dev;
2789 addrconf_join_solict(dev, &ifp->addr);
2791 net_srandom(ifp->addr.s6_addr32[3]);
2793 read_lock_bh(&idev->lock);
2794 if (ifp->dead)
2795 goto out;
2796 spin_lock_bh(&ifp->lock);
2798 if (dev->flags&(IFF_NOARP|IFF_LOOPBACK) ||
2799 idev->cnf.accept_dad < 1 ||
2800 !(ifp->flags&IFA_F_TENTATIVE) ||
2801 ifp->flags & IFA_F_NODAD) {
2802 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
2803 spin_unlock_bh(&ifp->lock);
2804 read_unlock_bh(&idev->lock);
2806 addrconf_dad_completed(ifp);
2807 return;
2810 if (!(idev->if_flags & IF_READY)) {
2811 spin_unlock_bh(&ifp->lock);
2812 read_unlock_bh(&idev->lock);
2814 * If the device is not ready:
2815 * - keep it tentative if it is a permanent address.
2816 * - otherwise, kill it.
2818 in6_ifa_hold(ifp);
2819 addrconf_dad_stop(ifp, 0);
2820 return;
2824 * Optimistic nodes can start receiving
2825 * Frames right away
2827 if(ifp->flags & IFA_F_OPTIMISTIC)
2828 ip6_ins_rt(ifp->rt);
2830 addrconf_dad_kick(ifp);
2831 spin_unlock_bh(&ifp->lock);
2832 out:
2833 read_unlock_bh(&idev->lock);
2836 static void addrconf_dad_timer(unsigned long data)
2838 struct inet6_ifaddr *ifp = (struct inet6_ifaddr *) data;
2839 struct inet6_dev *idev = ifp->idev;
2840 struct in6_addr mcaddr;
2842 read_lock_bh(&idev->lock);
2843 if (idev->dead) {
2844 read_unlock_bh(&idev->lock);
2845 goto out;
2847 spin_lock_bh(&ifp->lock);
2848 if (ifp->probes == 0) {
2850 * DAD was successful
2853 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
2854 spin_unlock_bh(&ifp->lock);
2855 read_unlock_bh(&idev->lock);
2857 addrconf_dad_completed(ifp);
2859 goto out;
2862 ifp->probes--;
2863 addrconf_mod_timer(ifp, AC_DAD, ifp->idev->nd_parms->retrans_time);
2864 spin_unlock_bh(&ifp->lock);
2865 read_unlock_bh(&idev->lock);
2867 /* send a neighbour solicitation for our addr */
2868 addrconf_addr_solict_mult(&ifp->addr, &mcaddr);
2869 ndisc_send_ns(ifp->idev->dev, NULL, &ifp->addr, &mcaddr, &in6addr_any);
2870 out:
2871 in6_ifa_put(ifp);
2874 static void addrconf_dad_completed(struct inet6_ifaddr *ifp)
2876 struct net_device * dev = ifp->idev->dev;
2879 * Configure the address for reception. Now it is valid.
2882 ipv6_ifa_notify(RTM_NEWADDR, ifp);
2884 /* If added prefix is link local and forwarding is off,
2885 start sending router solicitations.
2888 if (ifp->idev->cnf.forwarding == 0 &&
2889 ifp->idev->cnf.rtr_solicits > 0 &&
2890 (dev->flags&IFF_LOOPBACK) == 0 &&
2891 (ipv6_addr_type(&ifp->addr) & IPV6_ADDR_LINKLOCAL)) {
2893 * If a host as already performed a random delay
2894 * [...] as part of DAD [...] there is no need
2895 * to delay again before sending the first RS
2897 ndisc_send_rs(ifp->idev->dev, &ifp->addr, &in6addr_linklocal_allrouters);
2899 spin_lock_bh(&ifp->lock);
2900 ifp->probes = 1;
2901 ifp->idev->if_flags |= IF_RS_SENT;
2902 addrconf_mod_timer(ifp, AC_RS, ifp->idev->cnf.rtr_solicit_interval);
2903 spin_unlock_bh(&ifp->lock);
2907 static void addrconf_dad_run(struct inet6_dev *idev) {
2908 struct inet6_ifaddr *ifp;
2910 read_lock_bh(&idev->lock);
2911 for (ifp = idev->addr_list; ifp; ifp = ifp->if_next) {
2912 spin_lock_bh(&ifp->lock);
2913 if (!(ifp->flags & IFA_F_TENTATIVE)) {
2914 spin_unlock_bh(&ifp->lock);
2915 continue;
2917 spin_unlock_bh(&ifp->lock);
2918 addrconf_dad_kick(ifp);
2920 read_unlock_bh(&idev->lock);
2923 #ifdef CONFIG_PROC_FS
2924 struct if6_iter_state {
2925 struct seq_net_private p;
2926 int bucket;
2929 static struct inet6_ifaddr *if6_get_first(struct seq_file *seq)
2931 struct inet6_ifaddr *ifa = NULL;
2932 struct if6_iter_state *state = seq->private;
2933 struct net *net = seq_file_net(seq);
2935 for (state->bucket = 0; state->bucket < IN6_ADDR_HSIZE; ++state->bucket) {
2936 ifa = inet6_addr_lst[state->bucket];
2938 while (ifa && !net_eq(dev_net(ifa->idev->dev), net))
2939 ifa = ifa->lst_next;
2940 if (ifa)
2941 break;
2943 return ifa;
2946 static struct inet6_ifaddr *if6_get_next(struct seq_file *seq, struct inet6_ifaddr *ifa)
2948 struct if6_iter_state *state = seq->private;
2949 struct net *net = seq_file_net(seq);
2951 ifa = ifa->lst_next;
2952 try_again:
2953 if (ifa) {
2954 if (!net_eq(dev_net(ifa->idev->dev), net)) {
2955 ifa = ifa->lst_next;
2956 goto try_again;
2960 if (!ifa && ++state->bucket < IN6_ADDR_HSIZE) {
2961 ifa = inet6_addr_lst[state->bucket];
2962 goto try_again;
2965 return ifa;
2968 static struct inet6_ifaddr *if6_get_idx(struct seq_file *seq, loff_t pos)
2970 struct inet6_ifaddr *ifa = if6_get_first(seq);
2972 if (ifa)
2973 while(pos && (ifa = if6_get_next(seq, ifa)) != NULL)
2974 --pos;
2975 return pos ? NULL : ifa;
2978 static void *if6_seq_start(struct seq_file *seq, loff_t *pos)
2979 __acquires(addrconf_hash_lock)
2981 read_lock_bh(&addrconf_hash_lock);
2982 return if6_get_idx(seq, *pos);
2985 static void *if6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
2987 struct inet6_ifaddr *ifa;
2989 ifa = if6_get_next(seq, v);
2990 ++*pos;
2991 return ifa;
2994 static void if6_seq_stop(struct seq_file *seq, void *v)
2995 __releases(addrconf_hash_lock)
2997 read_unlock_bh(&addrconf_hash_lock);
3000 static int if6_seq_show(struct seq_file *seq, void *v)
3002 struct inet6_ifaddr *ifp = (struct inet6_ifaddr *)v;
3003 seq_printf(seq, "%pi6 %02x %02x %02x %02x %8s\n",
3004 &ifp->addr,
3005 ifp->idev->dev->ifindex,
3006 ifp->prefix_len,
3007 ifp->scope,
3008 ifp->flags,
3009 ifp->idev->dev->name);
3010 return 0;
3013 static const struct seq_operations if6_seq_ops = {
3014 .start = if6_seq_start,
3015 .next = if6_seq_next,
3016 .show = if6_seq_show,
3017 .stop = if6_seq_stop,
3020 static int if6_seq_open(struct inode *inode, struct file *file)
3022 return seq_open_net(inode, file, &if6_seq_ops,
3023 sizeof(struct if6_iter_state));
3026 static const struct file_operations if6_fops = {
3027 .owner = THIS_MODULE,
3028 .open = if6_seq_open,
3029 .read = seq_read,
3030 .llseek = seq_lseek,
3031 .release = seq_release_net,
3034 static int if6_proc_net_init(struct net *net)
3036 if (!proc_net_fops_create(net, "if_inet6", S_IRUGO, &if6_fops))
3037 return -ENOMEM;
3038 return 0;
3041 static void if6_proc_net_exit(struct net *net)
3043 proc_net_remove(net, "if_inet6");
3046 static struct pernet_operations if6_proc_net_ops = {
3047 .init = if6_proc_net_init,
3048 .exit = if6_proc_net_exit,
3051 int __init if6_proc_init(void)
3053 return register_pernet_subsys(&if6_proc_net_ops);
3056 void if6_proc_exit(void)
3058 unregister_pernet_subsys(&if6_proc_net_ops);
3060 #endif /* CONFIG_PROC_FS */
3062 #if defined(CONFIG_IPV6_MIP6) || defined(CONFIG_IPV6_MIP6_MODULE)
3063 /* Check if address is a home address configured on any interface. */
3064 int ipv6_chk_home_addr(struct net *net, struct in6_addr *addr)
3066 int ret = 0;
3067 struct inet6_ifaddr * ifp;
3068 u8 hash = ipv6_addr_hash(addr);
3069 read_lock_bh(&addrconf_hash_lock);
3070 for (ifp = inet6_addr_lst[hash]; ifp; ifp = ifp->lst_next) {
3071 if (!net_eq(dev_net(ifp->idev->dev), net))
3072 continue;
3073 if (ipv6_addr_equal(&ifp->addr, addr) &&
3074 (ifp->flags & IFA_F_HOMEADDRESS)) {
3075 ret = 1;
3076 break;
3079 read_unlock_bh(&addrconf_hash_lock);
3080 return ret;
3082 #endif
3085 * Periodic address status verification
3088 static void addrconf_verify(unsigned long foo)
3090 struct inet6_ifaddr *ifp;
3091 unsigned long now, next;
3092 int i;
3094 spin_lock_bh(&addrconf_verify_lock);
3095 now = jiffies;
3096 next = now + ADDR_CHECK_FREQUENCY;
3098 del_timer(&addr_chk_timer);
3100 for (i=0; i < IN6_ADDR_HSIZE; i++) {
3102 restart:
3103 read_lock(&addrconf_hash_lock);
3104 for (ifp=inet6_addr_lst[i]; ifp; ifp=ifp->lst_next) {
3105 unsigned long age;
3106 #ifdef CONFIG_IPV6_PRIVACY
3107 unsigned long regen_advance;
3108 #endif
3110 if (ifp->flags & IFA_F_PERMANENT)
3111 continue;
3113 spin_lock(&ifp->lock);
3114 age = (now - ifp->tstamp) / HZ;
3116 #ifdef CONFIG_IPV6_PRIVACY
3117 regen_advance = ifp->idev->cnf.regen_max_retry *
3118 ifp->idev->cnf.dad_transmits *
3119 ifp->idev->nd_parms->retrans_time / HZ;
3120 #endif
3122 if (ifp->valid_lft != INFINITY_LIFE_TIME &&
3123 age >= ifp->valid_lft) {
3124 spin_unlock(&ifp->lock);
3125 in6_ifa_hold(ifp);
3126 read_unlock(&addrconf_hash_lock);
3127 ipv6_del_addr(ifp);
3128 goto restart;
3129 } else if (ifp->prefered_lft == INFINITY_LIFE_TIME) {
3130 spin_unlock(&ifp->lock);
3131 continue;
3132 } else if (age >= ifp->prefered_lft) {
3133 /* jiffies - ifp->tstamp > age >= ifp->prefered_lft */
3134 int deprecate = 0;
3136 if (!(ifp->flags&IFA_F_DEPRECATED)) {
3137 deprecate = 1;
3138 ifp->flags |= IFA_F_DEPRECATED;
3141 if (time_before(ifp->tstamp + ifp->valid_lft * HZ, next))
3142 next = ifp->tstamp + ifp->valid_lft * HZ;
3144 spin_unlock(&ifp->lock);
3146 if (deprecate) {
3147 in6_ifa_hold(ifp);
3148 read_unlock(&addrconf_hash_lock);
3150 ipv6_ifa_notify(0, ifp);
3151 in6_ifa_put(ifp);
3152 goto restart;
3154 #ifdef CONFIG_IPV6_PRIVACY
3155 } else if ((ifp->flags&IFA_F_TEMPORARY) &&
3156 !(ifp->flags&IFA_F_TENTATIVE)) {
3157 if (age >= ifp->prefered_lft - regen_advance) {
3158 struct inet6_ifaddr *ifpub = ifp->ifpub;
3159 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3160 next = ifp->tstamp + ifp->prefered_lft * HZ;
3161 if (!ifp->regen_count && ifpub) {
3162 ifp->regen_count++;
3163 in6_ifa_hold(ifp);
3164 in6_ifa_hold(ifpub);
3165 spin_unlock(&ifp->lock);
3166 read_unlock(&addrconf_hash_lock);
3167 spin_lock(&ifpub->lock);
3168 ifpub->regen_count = 0;
3169 spin_unlock(&ifpub->lock);
3170 ipv6_create_tempaddr(ifpub, ifp);
3171 in6_ifa_put(ifpub);
3172 in6_ifa_put(ifp);
3173 goto restart;
3175 } else if (time_before(ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ, next))
3176 next = ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ;
3177 spin_unlock(&ifp->lock);
3178 #endif
3179 } else {
3180 /* ifp->prefered_lft <= ifp->valid_lft */
3181 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3182 next = ifp->tstamp + ifp->prefered_lft * HZ;
3183 spin_unlock(&ifp->lock);
3186 read_unlock(&addrconf_hash_lock);
3189 addr_chk_timer.expires = time_before(next, jiffies + HZ) ? jiffies + HZ : next;
3190 add_timer(&addr_chk_timer);
3191 spin_unlock_bh(&addrconf_verify_lock);
3194 static struct in6_addr *extract_addr(struct nlattr *addr, struct nlattr *local)
3196 struct in6_addr *pfx = NULL;
3198 if (addr)
3199 pfx = nla_data(addr);
3201 if (local) {
3202 if (pfx && nla_memcmp(local, pfx, sizeof(*pfx)))
3203 pfx = NULL;
3204 else
3205 pfx = nla_data(local);
3208 return pfx;
3211 static const struct nla_policy ifa_ipv6_policy[IFA_MAX+1] = {
3212 [IFA_ADDRESS] = { .len = sizeof(struct in6_addr) },
3213 [IFA_LOCAL] = { .len = sizeof(struct in6_addr) },
3214 [IFA_CACHEINFO] = { .len = sizeof(struct ifa_cacheinfo) },
3217 static int
3218 inet6_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
3220 struct net *net = sock_net(skb->sk);
3221 struct ifaddrmsg *ifm;
3222 struct nlattr *tb[IFA_MAX+1];
3223 struct in6_addr *pfx;
3224 int err;
3226 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3227 if (err < 0)
3228 return err;
3230 ifm = nlmsg_data(nlh);
3231 pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3232 if (pfx == NULL)
3233 return -EINVAL;
3235 return inet6_addr_del(net, ifm->ifa_index, pfx, ifm->ifa_prefixlen);
3238 static int inet6_addr_modify(struct inet6_ifaddr *ifp, u8 ifa_flags,
3239 u32 prefered_lft, u32 valid_lft)
3241 u32 flags;
3242 clock_t expires;
3243 unsigned long timeout;
3245 if (!valid_lft || (prefered_lft > valid_lft))
3246 return -EINVAL;
3248 timeout = addrconf_timeout_fixup(valid_lft, HZ);
3249 if (addrconf_finite_timeout(timeout)) {
3250 expires = jiffies_to_clock_t(timeout * HZ);
3251 valid_lft = timeout;
3252 flags = RTF_EXPIRES;
3253 } else {
3254 expires = 0;
3255 flags = 0;
3256 ifa_flags |= IFA_F_PERMANENT;
3259 timeout = addrconf_timeout_fixup(prefered_lft, HZ);
3260 if (addrconf_finite_timeout(timeout)) {
3261 if (timeout == 0)
3262 ifa_flags |= IFA_F_DEPRECATED;
3263 prefered_lft = timeout;
3266 spin_lock_bh(&ifp->lock);
3267 ifp->flags = (ifp->flags & ~(IFA_F_DEPRECATED | IFA_F_PERMANENT | IFA_F_NODAD | IFA_F_HOMEADDRESS)) | ifa_flags;
3268 ifp->tstamp = jiffies;
3269 ifp->valid_lft = valid_lft;
3270 ifp->prefered_lft = prefered_lft;
3272 spin_unlock_bh(&ifp->lock);
3273 if (!(ifp->flags&IFA_F_TENTATIVE))
3274 ipv6_ifa_notify(0, ifp);
3276 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, ifp->idev->dev,
3277 expires, flags);
3278 addrconf_verify(0);
3280 return 0;
3283 static int
3284 inet6_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
3286 struct net *net = sock_net(skb->sk);
3287 struct ifaddrmsg *ifm;
3288 struct nlattr *tb[IFA_MAX+1];
3289 struct in6_addr *pfx;
3290 struct inet6_ifaddr *ifa;
3291 struct net_device *dev;
3292 u32 valid_lft = INFINITY_LIFE_TIME, preferred_lft = INFINITY_LIFE_TIME;
3293 u8 ifa_flags;
3294 int err;
3296 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3297 if (err < 0)
3298 return err;
3300 ifm = nlmsg_data(nlh);
3301 pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3302 if (pfx == NULL)
3303 return -EINVAL;
3305 if (tb[IFA_CACHEINFO]) {
3306 struct ifa_cacheinfo *ci;
3308 ci = nla_data(tb[IFA_CACHEINFO]);
3309 valid_lft = ci->ifa_valid;
3310 preferred_lft = ci->ifa_prefered;
3311 } else {
3312 preferred_lft = INFINITY_LIFE_TIME;
3313 valid_lft = INFINITY_LIFE_TIME;
3316 dev = __dev_get_by_index(net, ifm->ifa_index);
3317 if (dev == NULL)
3318 return -ENODEV;
3320 /* We ignore other flags so far. */
3321 ifa_flags = ifm->ifa_flags & (IFA_F_NODAD | IFA_F_HOMEADDRESS);
3323 ifa = ipv6_get_ifaddr(net, pfx, dev, 1);
3324 if (ifa == NULL) {
3326 * It would be best to check for !NLM_F_CREATE here but
3327 * userspace alreay relies on not having to provide this.
3329 return inet6_addr_add(net, ifm->ifa_index, pfx,
3330 ifm->ifa_prefixlen, ifa_flags,
3331 preferred_lft, valid_lft);
3334 if (nlh->nlmsg_flags & NLM_F_EXCL ||
3335 !(nlh->nlmsg_flags & NLM_F_REPLACE))
3336 err = -EEXIST;
3337 else
3338 err = inet6_addr_modify(ifa, ifa_flags, preferred_lft, valid_lft);
3340 in6_ifa_put(ifa);
3342 return err;
3345 static void put_ifaddrmsg(struct nlmsghdr *nlh, u8 prefixlen, u8 flags,
3346 u8 scope, int ifindex)
3348 struct ifaddrmsg *ifm;
3350 ifm = nlmsg_data(nlh);
3351 ifm->ifa_family = AF_INET6;
3352 ifm->ifa_prefixlen = prefixlen;
3353 ifm->ifa_flags = flags;
3354 ifm->ifa_scope = scope;
3355 ifm->ifa_index = ifindex;
3358 static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
3359 unsigned long tstamp, u32 preferred, u32 valid)
3361 struct ifa_cacheinfo ci;
3363 ci.cstamp = (u32)(TIME_DELTA(cstamp, INITIAL_JIFFIES) / HZ * 100
3364 + TIME_DELTA(cstamp, INITIAL_JIFFIES) % HZ * 100 / HZ);
3365 ci.tstamp = (u32)(TIME_DELTA(tstamp, INITIAL_JIFFIES) / HZ * 100
3366 + TIME_DELTA(tstamp, INITIAL_JIFFIES) % HZ * 100 / HZ);
3367 ci.ifa_prefered = preferred;
3368 ci.ifa_valid = valid;
3370 return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
3373 static inline int rt_scope(int ifa_scope)
3375 if (ifa_scope & IFA_HOST)
3376 return RT_SCOPE_HOST;
3377 else if (ifa_scope & IFA_LINK)
3378 return RT_SCOPE_LINK;
3379 else if (ifa_scope & IFA_SITE)
3380 return RT_SCOPE_SITE;
3381 else
3382 return RT_SCOPE_UNIVERSE;
3385 static inline int inet6_ifaddr_msgsize(void)
3387 return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
3388 + nla_total_size(16) /* IFA_ADDRESS */
3389 + nla_total_size(sizeof(struct ifa_cacheinfo));
3392 static int inet6_fill_ifaddr(struct sk_buff *skb, struct inet6_ifaddr *ifa,
3393 u32 pid, u32 seq, int event, unsigned int flags)
3395 struct nlmsghdr *nlh;
3396 u32 preferred, valid;
3398 nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3399 if (nlh == NULL)
3400 return -EMSGSIZE;
3402 put_ifaddrmsg(nlh, ifa->prefix_len, ifa->flags, rt_scope(ifa->scope),
3403 ifa->idev->dev->ifindex);
3405 if (!(ifa->flags&IFA_F_PERMANENT)) {
3406 preferred = ifa->prefered_lft;
3407 valid = ifa->valid_lft;
3408 if (preferred != INFINITY_LIFE_TIME) {
3409 long tval = (jiffies - ifa->tstamp)/HZ;
3410 if (preferred > tval)
3411 preferred -= tval;
3412 else
3413 preferred = 0;
3414 if (valid != INFINITY_LIFE_TIME)
3415 valid -= tval;
3417 } else {
3418 preferred = INFINITY_LIFE_TIME;
3419 valid = INFINITY_LIFE_TIME;
3422 if (nla_put(skb, IFA_ADDRESS, 16, &ifa->addr) < 0 ||
3423 put_cacheinfo(skb, ifa->cstamp, ifa->tstamp, preferred, valid) < 0) {
3424 nlmsg_cancel(skb, nlh);
3425 return -EMSGSIZE;
3428 return nlmsg_end(skb, nlh);
3431 static int inet6_fill_ifmcaddr(struct sk_buff *skb, struct ifmcaddr6 *ifmca,
3432 u32 pid, u32 seq, int event, u16 flags)
3434 struct nlmsghdr *nlh;
3435 u8 scope = RT_SCOPE_UNIVERSE;
3436 int ifindex = ifmca->idev->dev->ifindex;
3438 if (ipv6_addr_scope(&ifmca->mca_addr) & IFA_SITE)
3439 scope = RT_SCOPE_SITE;
3441 nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3442 if (nlh == NULL)
3443 return -EMSGSIZE;
3445 put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3446 if (nla_put(skb, IFA_MULTICAST, 16, &ifmca->mca_addr) < 0 ||
3447 put_cacheinfo(skb, ifmca->mca_cstamp, ifmca->mca_tstamp,
3448 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3449 nlmsg_cancel(skb, nlh);
3450 return -EMSGSIZE;
3453 return nlmsg_end(skb, nlh);
3456 static int inet6_fill_ifacaddr(struct sk_buff *skb, struct ifacaddr6 *ifaca,
3457 u32 pid, u32 seq, int event, unsigned int flags)
3459 struct nlmsghdr *nlh;
3460 u8 scope = RT_SCOPE_UNIVERSE;
3461 int ifindex = ifaca->aca_idev->dev->ifindex;
3463 if (ipv6_addr_scope(&ifaca->aca_addr) & IFA_SITE)
3464 scope = RT_SCOPE_SITE;
3466 nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3467 if (nlh == NULL)
3468 return -EMSGSIZE;
3470 put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3471 if (nla_put(skb, IFA_ANYCAST, 16, &ifaca->aca_addr) < 0 ||
3472 put_cacheinfo(skb, ifaca->aca_cstamp, ifaca->aca_tstamp,
3473 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3474 nlmsg_cancel(skb, nlh);
3475 return -EMSGSIZE;
3478 return nlmsg_end(skb, nlh);
3481 enum addr_type_t
3483 UNICAST_ADDR,
3484 MULTICAST_ADDR,
3485 ANYCAST_ADDR,
3488 static int inet6_dump_addr(struct sk_buff *skb, struct netlink_callback *cb,
3489 enum addr_type_t type)
3491 int idx, ip_idx;
3492 int s_idx, s_ip_idx;
3493 int err = 1;
3494 struct net_device *dev;
3495 struct inet6_dev *idev = NULL;
3496 struct inet6_ifaddr *ifa;
3497 struct ifmcaddr6 *ifmca;
3498 struct ifacaddr6 *ifaca;
3499 struct net *net = sock_net(skb->sk);
3501 s_idx = cb->args[0];
3502 s_ip_idx = ip_idx = cb->args[1];
3504 idx = 0;
3505 for_each_netdev(net, dev) {
3506 if (idx < s_idx)
3507 goto cont;
3508 if (idx > s_idx)
3509 s_ip_idx = 0;
3510 ip_idx = 0;
3511 if ((idev = in6_dev_get(dev)) == NULL)
3512 goto cont;
3513 read_lock_bh(&idev->lock);
3514 switch (type) {
3515 case UNICAST_ADDR:
3516 /* unicast address incl. temp addr */
3517 for (ifa = idev->addr_list; ifa;
3518 ifa = ifa->if_next, ip_idx++) {
3519 if (ip_idx < s_ip_idx)
3520 continue;
3521 err = inet6_fill_ifaddr(skb, ifa,
3522 NETLINK_CB(cb->skb).pid,
3523 cb->nlh->nlmsg_seq,
3524 RTM_NEWADDR,
3525 NLM_F_MULTI);
3527 break;
3528 case MULTICAST_ADDR:
3529 /* multicast address */
3530 for (ifmca = idev->mc_list; ifmca;
3531 ifmca = ifmca->next, ip_idx++) {
3532 if (ip_idx < s_ip_idx)
3533 continue;
3534 err = inet6_fill_ifmcaddr(skb, ifmca,
3535 NETLINK_CB(cb->skb).pid,
3536 cb->nlh->nlmsg_seq,
3537 RTM_GETMULTICAST,
3538 NLM_F_MULTI);
3540 break;
3541 case ANYCAST_ADDR:
3542 /* anycast address */
3543 for (ifaca = idev->ac_list; ifaca;
3544 ifaca = ifaca->aca_next, ip_idx++) {
3545 if (ip_idx < s_ip_idx)
3546 continue;
3547 err = inet6_fill_ifacaddr(skb, ifaca,
3548 NETLINK_CB(cb->skb).pid,
3549 cb->nlh->nlmsg_seq,
3550 RTM_GETANYCAST,
3551 NLM_F_MULTI);
3553 break;
3554 default:
3555 break;
3557 read_unlock_bh(&idev->lock);
3558 in6_dev_put(idev);
3560 if (err <= 0)
3561 break;
3562 cont:
3563 idx++;
3565 cb->args[0] = idx;
3566 cb->args[1] = ip_idx;
3567 return skb->len;
3570 static int inet6_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
3572 enum addr_type_t type = UNICAST_ADDR;
3574 return inet6_dump_addr(skb, cb, type);
3577 static int inet6_dump_ifmcaddr(struct sk_buff *skb, struct netlink_callback *cb)
3579 enum addr_type_t type = MULTICAST_ADDR;
3581 return inet6_dump_addr(skb, cb, type);
3585 static int inet6_dump_ifacaddr(struct sk_buff *skb, struct netlink_callback *cb)
3587 enum addr_type_t type = ANYCAST_ADDR;
3589 return inet6_dump_addr(skb, cb, type);
3592 static int inet6_rtm_getaddr(struct sk_buff *in_skb, struct nlmsghdr* nlh,
3593 void *arg)
3595 struct net *net = sock_net(in_skb->sk);
3596 struct ifaddrmsg *ifm;
3597 struct nlattr *tb[IFA_MAX+1];
3598 struct in6_addr *addr = NULL;
3599 struct net_device *dev = NULL;
3600 struct inet6_ifaddr *ifa;
3601 struct sk_buff *skb;
3602 int err;
3604 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3605 if (err < 0)
3606 goto errout;
3608 addr = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3609 if (addr == NULL) {
3610 err = -EINVAL;
3611 goto errout;
3614 ifm = nlmsg_data(nlh);
3615 if (ifm->ifa_index)
3616 dev = __dev_get_by_index(net, ifm->ifa_index);
3618 if ((ifa = ipv6_get_ifaddr(net, addr, dev, 1)) == NULL) {
3619 err = -EADDRNOTAVAIL;
3620 goto errout;
3623 if ((skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_KERNEL)) == NULL) {
3624 err = -ENOBUFS;
3625 goto errout_ifa;
3628 err = inet6_fill_ifaddr(skb, ifa, NETLINK_CB(in_skb).pid,
3629 nlh->nlmsg_seq, RTM_NEWADDR, 0);
3630 if (err < 0) {
3631 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
3632 WARN_ON(err == -EMSGSIZE);
3633 kfree_skb(skb);
3634 goto errout_ifa;
3636 err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).pid);
3637 errout_ifa:
3638 in6_ifa_put(ifa);
3639 errout:
3640 return err;
3643 static void inet6_ifa_notify(int event, struct inet6_ifaddr *ifa)
3645 struct sk_buff *skb;
3646 struct net *net = dev_net(ifa->idev->dev);
3647 int err = -ENOBUFS;
3649 skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_ATOMIC);
3650 if (skb == NULL)
3651 goto errout;
3653 err = inet6_fill_ifaddr(skb, ifa, 0, 0, event, 0);
3654 if (err < 0) {
3655 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
3656 WARN_ON(err == -EMSGSIZE);
3657 kfree_skb(skb);
3658 goto errout;
3660 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
3661 return;
3662 errout:
3663 if (err < 0)
3664 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
3667 static inline void ipv6_store_devconf(struct ipv6_devconf *cnf,
3668 __s32 *array, int bytes)
3670 BUG_ON(bytes < (DEVCONF_MAX * 4));
3672 memset(array, 0, bytes);
3673 array[DEVCONF_FORWARDING] = cnf->forwarding;
3674 array[DEVCONF_HOPLIMIT] = cnf->hop_limit;
3675 array[DEVCONF_MTU6] = cnf->mtu6;
3676 array[DEVCONF_ACCEPT_RA] = cnf->accept_ra;
3677 array[DEVCONF_ACCEPT_REDIRECTS] = cnf->accept_redirects;
3678 array[DEVCONF_AUTOCONF] = cnf->autoconf;
3679 array[DEVCONF_DAD_TRANSMITS] = cnf->dad_transmits;
3680 array[DEVCONF_RTR_SOLICITS] = cnf->rtr_solicits;
3681 array[DEVCONF_RTR_SOLICIT_INTERVAL] = cnf->rtr_solicit_interval;
3682 array[DEVCONF_RTR_SOLICIT_DELAY] = cnf->rtr_solicit_delay;
3683 array[DEVCONF_FORCE_MLD_VERSION] = cnf->force_mld_version;
3684 #ifdef CONFIG_IPV6_PRIVACY
3685 array[DEVCONF_USE_TEMPADDR] = cnf->use_tempaddr;
3686 array[DEVCONF_TEMP_VALID_LFT] = cnf->temp_valid_lft;
3687 array[DEVCONF_TEMP_PREFERED_LFT] = cnf->temp_prefered_lft;
3688 array[DEVCONF_REGEN_MAX_RETRY] = cnf->regen_max_retry;
3689 array[DEVCONF_MAX_DESYNC_FACTOR] = cnf->max_desync_factor;
3690 #endif
3691 array[DEVCONF_MAX_ADDRESSES] = cnf->max_addresses;
3692 array[DEVCONF_ACCEPT_RA_DEFRTR] = cnf->accept_ra_defrtr;
3693 array[DEVCONF_ACCEPT_RA_PINFO] = cnf->accept_ra_pinfo;
3694 #ifdef CONFIG_IPV6_ROUTER_PREF
3695 array[DEVCONF_ACCEPT_RA_RTR_PREF] = cnf->accept_ra_rtr_pref;
3696 array[DEVCONF_RTR_PROBE_INTERVAL] = cnf->rtr_probe_interval;
3697 #ifdef CONFIG_IPV6_ROUTE_INFO
3698 array[DEVCONF_ACCEPT_RA_RT_INFO_MAX_PLEN] = cnf->accept_ra_rt_info_max_plen;
3699 #endif
3700 #endif
3701 array[DEVCONF_PROXY_NDP] = cnf->proxy_ndp;
3702 array[DEVCONF_ACCEPT_SOURCE_ROUTE] = cnf->accept_source_route;
3703 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
3704 array[DEVCONF_OPTIMISTIC_DAD] = cnf->optimistic_dad;
3705 #endif
3706 #ifdef CONFIG_IPV6_MROUTE
3707 array[DEVCONF_MC_FORWARDING] = cnf->mc_forwarding;
3708 #endif
3709 array[DEVCONF_DISABLE_IPV6] = cnf->disable_ipv6;
3710 array[DEVCONF_ACCEPT_DAD] = cnf->accept_dad;
3713 static inline size_t inet6_if_nlmsg_size(void)
3715 return NLMSG_ALIGN(sizeof(struct ifinfomsg))
3716 + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
3717 + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
3718 + nla_total_size(4) /* IFLA_MTU */
3719 + nla_total_size(4) /* IFLA_LINK */
3720 + nla_total_size( /* IFLA_PROTINFO */
3721 nla_total_size(4) /* IFLA_INET6_FLAGS */
3722 + nla_total_size(sizeof(struct ifla_cacheinfo))
3723 + nla_total_size(DEVCONF_MAX * 4) /* IFLA_INET6_CONF */
3724 + nla_total_size(IPSTATS_MIB_MAX * 8) /* IFLA_INET6_STATS */
3725 + nla_total_size(ICMP6_MIB_MAX * 8) /* IFLA_INET6_ICMP6STATS */
3729 static inline void __snmp6_fill_stats(u64 *stats, void **mib, int items,
3730 int bytes)
3732 int i;
3733 int pad = bytes - sizeof(u64) * items;
3734 BUG_ON(pad < 0);
3736 /* Use put_unaligned() because stats may not be aligned for u64. */
3737 put_unaligned(items, &stats[0]);
3738 for (i = 1; i < items; i++)
3739 put_unaligned(snmp_fold_field(mib, i), &stats[i]);
3741 memset(&stats[items], 0, pad);
3744 static void snmp6_fill_stats(u64 *stats, struct inet6_dev *idev, int attrtype,
3745 int bytes)
3747 switch(attrtype) {
3748 case IFLA_INET6_STATS:
3749 __snmp6_fill_stats(stats, (void **)idev->stats.ipv6, IPSTATS_MIB_MAX, bytes);
3750 break;
3751 case IFLA_INET6_ICMP6STATS:
3752 __snmp6_fill_stats(stats, (void **)idev->stats.icmpv6, ICMP6_MIB_MAX, bytes);
3753 break;
3757 static int inet6_fill_ifinfo(struct sk_buff *skb, struct inet6_dev *idev,
3758 u32 pid, u32 seq, int event, unsigned int flags)
3760 struct net_device *dev = idev->dev;
3761 struct nlattr *nla;
3762 struct ifinfomsg *hdr;
3763 struct nlmsghdr *nlh;
3764 void *protoinfo;
3765 struct ifla_cacheinfo ci;
3767 nlh = nlmsg_put(skb, pid, seq, event, sizeof(*hdr), flags);
3768 if (nlh == NULL)
3769 return -EMSGSIZE;
3771 hdr = nlmsg_data(nlh);
3772 hdr->ifi_family = AF_INET6;
3773 hdr->__ifi_pad = 0;
3774 hdr->ifi_type = dev->type;
3775 hdr->ifi_index = dev->ifindex;
3776 hdr->ifi_flags = dev_get_flags(dev);
3777 hdr->ifi_change = 0;
3779 NLA_PUT_STRING(skb, IFLA_IFNAME, dev->name);
3781 if (dev->addr_len)
3782 NLA_PUT(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr);
3784 NLA_PUT_U32(skb, IFLA_MTU, dev->mtu);
3785 if (dev->ifindex != dev->iflink)
3786 NLA_PUT_U32(skb, IFLA_LINK, dev->iflink);
3788 protoinfo = nla_nest_start(skb, IFLA_PROTINFO);
3789 if (protoinfo == NULL)
3790 goto nla_put_failure;
3792 NLA_PUT_U32(skb, IFLA_INET6_FLAGS, idev->if_flags);
3794 ci.max_reasm_len = IPV6_MAXPLEN;
3795 ci.tstamp = (__u32)(TIME_DELTA(idev->tstamp, INITIAL_JIFFIES) / HZ * 100
3796 + TIME_DELTA(idev->tstamp, INITIAL_JIFFIES) % HZ * 100 / HZ);
3797 ci.reachable_time = idev->nd_parms->reachable_time;
3798 ci.retrans_time = idev->nd_parms->retrans_time;
3799 NLA_PUT(skb, IFLA_INET6_CACHEINFO, sizeof(ci), &ci);
3801 nla = nla_reserve(skb, IFLA_INET6_CONF, DEVCONF_MAX * sizeof(s32));
3802 if (nla == NULL)
3803 goto nla_put_failure;
3804 ipv6_store_devconf(&idev->cnf, nla_data(nla), nla_len(nla));
3806 /* XXX - MC not implemented */
3808 nla = nla_reserve(skb, IFLA_INET6_STATS, IPSTATS_MIB_MAX * sizeof(u64));
3809 if (nla == NULL)
3810 goto nla_put_failure;
3811 snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_STATS, nla_len(nla));
3813 nla = nla_reserve(skb, IFLA_INET6_ICMP6STATS, ICMP6_MIB_MAX * sizeof(u64));
3814 if (nla == NULL)
3815 goto nla_put_failure;
3816 snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_ICMP6STATS, nla_len(nla));
3818 nla_nest_end(skb, protoinfo);
3819 return nlmsg_end(skb, nlh);
3821 nla_put_failure:
3822 nlmsg_cancel(skb, nlh);
3823 return -EMSGSIZE;
3826 static int inet6_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
3828 struct net *net = sock_net(skb->sk);
3829 int idx, err;
3830 int s_idx = cb->args[0];
3831 struct net_device *dev;
3832 struct inet6_dev *idev;
3834 read_lock(&dev_base_lock);
3835 idx = 0;
3836 for_each_netdev(net, dev) {
3837 if (idx < s_idx)
3838 goto cont;
3839 if ((idev = in6_dev_get(dev)) == NULL)
3840 goto cont;
3841 err = inet6_fill_ifinfo(skb, idev, NETLINK_CB(cb->skb).pid,
3842 cb->nlh->nlmsg_seq, RTM_NEWLINK, NLM_F_MULTI);
3843 in6_dev_put(idev);
3844 if (err <= 0)
3845 break;
3846 cont:
3847 idx++;
3849 read_unlock(&dev_base_lock);
3850 cb->args[0] = idx;
3852 return skb->len;
3855 void inet6_ifinfo_notify(int event, struct inet6_dev *idev)
3857 struct sk_buff *skb;
3858 struct net *net = dev_net(idev->dev);
3859 int err = -ENOBUFS;
3861 skb = nlmsg_new(inet6_if_nlmsg_size(), GFP_ATOMIC);
3862 if (skb == NULL)
3863 goto errout;
3865 err = inet6_fill_ifinfo(skb, idev, 0, 0, event, 0);
3866 if (err < 0) {
3867 /* -EMSGSIZE implies BUG in inet6_if_nlmsg_size() */
3868 WARN_ON(err == -EMSGSIZE);
3869 kfree_skb(skb);
3870 goto errout;
3872 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
3873 return;
3874 errout:
3875 if (err < 0)
3876 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
3879 static inline size_t inet6_prefix_nlmsg_size(void)
3881 return NLMSG_ALIGN(sizeof(struct prefixmsg))
3882 + nla_total_size(sizeof(struct in6_addr))
3883 + nla_total_size(sizeof(struct prefix_cacheinfo));
3886 static int inet6_fill_prefix(struct sk_buff *skb, struct inet6_dev *idev,
3887 struct prefix_info *pinfo, u32 pid, u32 seq,
3888 int event, unsigned int flags)
3890 struct prefixmsg *pmsg;
3891 struct nlmsghdr *nlh;
3892 struct prefix_cacheinfo ci;
3894 nlh = nlmsg_put(skb, pid, seq, event, sizeof(*pmsg), flags);
3895 if (nlh == NULL)
3896 return -EMSGSIZE;
3898 pmsg = nlmsg_data(nlh);
3899 pmsg->prefix_family = AF_INET6;
3900 pmsg->prefix_pad1 = 0;
3901 pmsg->prefix_pad2 = 0;
3902 pmsg->prefix_ifindex = idev->dev->ifindex;
3903 pmsg->prefix_len = pinfo->prefix_len;
3904 pmsg->prefix_type = pinfo->type;
3905 pmsg->prefix_pad3 = 0;
3906 pmsg->prefix_flags = 0;
3907 if (pinfo->onlink)
3908 pmsg->prefix_flags |= IF_PREFIX_ONLINK;
3909 if (pinfo->autoconf)
3910 pmsg->prefix_flags |= IF_PREFIX_AUTOCONF;
3912 NLA_PUT(skb, PREFIX_ADDRESS, sizeof(pinfo->prefix), &pinfo->prefix);
3914 ci.preferred_time = ntohl(pinfo->prefered);
3915 ci.valid_time = ntohl(pinfo->valid);
3916 NLA_PUT(skb, PREFIX_CACHEINFO, sizeof(ci), &ci);
3918 return nlmsg_end(skb, nlh);
3920 nla_put_failure:
3921 nlmsg_cancel(skb, nlh);
3922 return -EMSGSIZE;
3925 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
3926 struct prefix_info *pinfo)
3928 struct sk_buff *skb;
3929 struct net *net = dev_net(idev->dev);
3930 int err = -ENOBUFS;
3932 skb = nlmsg_new(inet6_prefix_nlmsg_size(), GFP_ATOMIC);
3933 if (skb == NULL)
3934 goto errout;
3936 err = inet6_fill_prefix(skb, idev, pinfo, 0, 0, event, 0);
3937 if (err < 0) {
3938 /* -EMSGSIZE implies BUG in inet6_prefix_nlmsg_size() */
3939 WARN_ON(err == -EMSGSIZE);
3940 kfree_skb(skb);
3941 goto errout;
3943 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_PREFIX, NULL, GFP_ATOMIC);
3944 return;
3945 errout:
3946 if (err < 0)
3947 rtnl_set_sk_err(net, RTNLGRP_IPV6_PREFIX, err);
3950 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
3952 inet6_ifa_notify(event ? : RTM_NEWADDR, ifp);
3954 switch (event) {
3955 case RTM_NEWADDR:
3957 * If the address was optimistic
3958 * we inserted the route at the start of
3959 * our DAD process, so we don't need
3960 * to do it again
3962 if (!(ifp->rt->rt6i_node))
3963 ip6_ins_rt(ifp->rt);
3964 if (ifp->idev->cnf.forwarding)
3965 addrconf_join_anycast(ifp);
3966 break;
3967 case RTM_DELADDR:
3968 if (ifp->idev->cnf.forwarding)
3969 addrconf_leave_anycast(ifp);
3970 addrconf_leave_solict(ifp->idev, &ifp->addr);
3971 dst_hold(&ifp->rt->u.dst);
3972 if (ip6_del_rt(ifp->rt))
3973 dst_free(&ifp->rt->u.dst);
3974 break;
3978 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
3980 rcu_read_lock_bh();
3981 if (likely(ifp->idev->dead == 0))
3982 __ipv6_ifa_notify(event, ifp);
3983 rcu_read_unlock_bh();
3986 #ifdef CONFIG_SYSCTL
3988 static
3989 int addrconf_sysctl_forward(ctl_table *ctl, int write,
3990 void __user *buffer, size_t *lenp, loff_t *ppos)
3992 int *valp = ctl->data;
3993 int val = *valp;
3994 int ret;
3996 ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
3998 if (write)
3999 ret = addrconf_fixup_forwarding(ctl, valp, val);
4000 return ret;
4003 static int addrconf_sysctl_forward_strategy(ctl_table *table,
4004 void __user *oldval,
4005 size_t __user *oldlenp,
4006 void __user *newval, size_t newlen)
4008 int *valp = table->data;
4009 int val = *valp;
4010 int new;
4012 if (!newval || !newlen)
4013 return 0;
4014 if (newlen != sizeof(int))
4015 return -EINVAL;
4016 if (get_user(new, (int __user *)newval))
4017 return -EFAULT;
4018 if (new == *valp)
4019 return 0;
4020 if (oldval && oldlenp) {
4021 size_t len;
4022 if (get_user(len, oldlenp))
4023 return -EFAULT;
4024 if (len) {
4025 if (len > table->maxlen)
4026 len = table->maxlen;
4027 if (copy_to_user(oldval, valp, len))
4028 return -EFAULT;
4029 if (put_user(len, oldlenp))
4030 return -EFAULT;
4034 *valp = new;
4035 return addrconf_fixup_forwarding(table, valp, val);
4038 static void dev_disable_change(struct inet6_dev *idev)
4040 if (!idev || !idev->dev)
4041 return;
4043 if (idev->cnf.disable_ipv6)
4044 addrconf_notify(NULL, NETDEV_DOWN, idev->dev);
4045 else
4046 addrconf_notify(NULL, NETDEV_UP, idev->dev);
4049 static void addrconf_disable_change(struct net *net, __s32 newf)
4051 struct net_device *dev;
4052 struct inet6_dev *idev;
4054 read_lock(&dev_base_lock);
4055 for_each_netdev(net, dev) {
4056 rcu_read_lock();
4057 idev = __in6_dev_get(dev);
4058 if (idev) {
4059 int changed = (!idev->cnf.disable_ipv6) ^ (!newf);
4060 idev->cnf.disable_ipv6 = newf;
4061 if (changed)
4062 dev_disable_change(idev);
4064 rcu_read_unlock();
4066 read_unlock(&dev_base_lock);
4069 static int addrconf_disable_ipv6(struct ctl_table *table, int *p, int old)
4071 struct net *net;
4073 net = (struct net *)table->extra2;
4075 if (p == &net->ipv6.devconf_dflt->disable_ipv6)
4076 return 0;
4078 if (!rtnl_trylock())
4079 return restart_syscall();
4081 if (p == &net->ipv6.devconf_all->disable_ipv6) {
4082 __s32 newf = net->ipv6.devconf_all->disable_ipv6;
4083 net->ipv6.devconf_dflt->disable_ipv6 = newf;
4084 addrconf_disable_change(net, newf);
4085 } else if ((!*p) ^ (!old))
4086 dev_disable_change((struct inet6_dev *)table->extra1);
4088 rtnl_unlock();
4089 return 0;
4092 static
4093 int addrconf_sysctl_disable(ctl_table *ctl, int write,
4094 void __user *buffer, size_t *lenp, loff_t *ppos)
4096 int *valp = ctl->data;
4097 int val = *valp;
4098 int ret;
4100 ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
4102 if (write)
4103 ret = addrconf_disable_ipv6(ctl, valp, val);
4104 return ret;
4107 static struct addrconf_sysctl_table
4109 struct ctl_table_header *sysctl_header;
4110 ctl_table addrconf_vars[DEVCONF_MAX+1];
4111 char *dev_name;
4112 } addrconf_sysctl __read_mostly = {
4113 .sysctl_header = NULL,
4114 .addrconf_vars = {
4116 .ctl_name = NET_IPV6_FORWARDING,
4117 .procname = "forwarding",
4118 .data = &ipv6_devconf.forwarding,
4119 .maxlen = sizeof(int),
4120 .mode = 0644,
4121 .proc_handler = addrconf_sysctl_forward,
4122 .strategy = addrconf_sysctl_forward_strategy,
4125 .ctl_name = NET_IPV6_HOP_LIMIT,
4126 .procname = "hop_limit",
4127 .data = &ipv6_devconf.hop_limit,
4128 .maxlen = sizeof(int),
4129 .mode = 0644,
4130 .proc_handler = proc_dointvec,
4133 .ctl_name = NET_IPV6_MTU,
4134 .procname = "mtu",
4135 .data = &ipv6_devconf.mtu6,
4136 .maxlen = sizeof(int),
4137 .mode = 0644,
4138 .proc_handler = proc_dointvec,
4141 .ctl_name = NET_IPV6_ACCEPT_RA,
4142 .procname = "accept_ra",
4143 .data = &ipv6_devconf.accept_ra,
4144 .maxlen = sizeof(int),
4145 .mode = 0644,
4146 .proc_handler = proc_dointvec,
4149 .ctl_name = NET_IPV6_ACCEPT_REDIRECTS,
4150 .procname = "accept_redirects",
4151 .data = &ipv6_devconf.accept_redirects,
4152 .maxlen = sizeof(int),
4153 .mode = 0644,
4154 .proc_handler = proc_dointvec,
4157 .ctl_name = NET_IPV6_AUTOCONF,
4158 .procname = "autoconf",
4159 .data = &ipv6_devconf.autoconf,
4160 .maxlen = sizeof(int),
4161 .mode = 0644,
4162 .proc_handler = proc_dointvec,
4165 .ctl_name = NET_IPV6_DAD_TRANSMITS,
4166 .procname = "dad_transmits",
4167 .data = &ipv6_devconf.dad_transmits,
4168 .maxlen = sizeof(int),
4169 .mode = 0644,
4170 .proc_handler = proc_dointvec,
4173 .ctl_name = NET_IPV6_RTR_SOLICITS,
4174 .procname = "router_solicitations",
4175 .data = &ipv6_devconf.rtr_solicits,
4176 .maxlen = sizeof(int),
4177 .mode = 0644,
4178 .proc_handler = proc_dointvec,
4181 .ctl_name = NET_IPV6_RTR_SOLICIT_INTERVAL,
4182 .procname = "router_solicitation_interval",
4183 .data = &ipv6_devconf.rtr_solicit_interval,
4184 .maxlen = sizeof(int),
4185 .mode = 0644,
4186 .proc_handler = proc_dointvec_jiffies,
4187 .strategy = sysctl_jiffies,
4190 .ctl_name = NET_IPV6_RTR_SOLICIT_DELAY,
4191 .procname = "router_solicitation_delay",
4192 .data = &ipv6_devconf.rtr_solicit_delay,
4193 .maxlen = sizeof(int),
4194 .mode = 0644,
4195 .proc_handler = proc_dointvec_jiffies,
4196 .strategy = sysctl_jiffies,
4199 .ctl_name = NET_IPV6_FORCE_MLD_VERSION,
4200 .procname = "force_mld_version",
4201 .data = &ipv6_devconf.force_mld_version,
4202 .maxlen = sizeof(int),
4203 .mode = 0644,
4204 .proc_handler = proc_dointvec,
4206 #ifdef CONFIG_IPV6_PRIVACY
4208 .ctl_name = NET_IPV6_USE_TEMPADDR,
4209 .procname = "use_tempaddr",
4210 .data = &ipv6_devconf.use_tempaddr,
4211 .maxlen = sizeof(int),
4212 .mode = 0644,
4213 .proc_handler = proc_dointvec,
4216 .ctl_name = NET_IPV6_TEMP_VALID_LFT,
4217 .procname = "temp_valid_lft",
4218 .data = &ipv6_devconf.temp_valid_lft,
4219 .maxlen = sizeof(int),
4220 .mode = 0644,
4221 .proc_handler = proc_dointvec,
4224 .ctl_name = NET_IPV6_TEMP_PREFERED_LFT,
4225 .procname = "temp_prefered_lft",
4226 .data = &ipv6_devconf.temp_prefered_lft,
4227 .maxlen = sizeof(int),
4228 .mode = 0644,
4229 .proc_handler = proc_dointvec,
4232 .ctl_name = NET_IPV6_REGEN_MAX_RETRY,
4233 .procname = "regen_max_retry",
4234 .data = &ipv6_devconf.regen_max_retry,
4235 .maxlen = sizeof(int),
4236 .mode = 0644,
4237 .proc_handler = proc_dointvec,
4240 .ctl_name = NET_IPV6_MAX_DESYNC_FACTOR,
4241 .procname = "max_desync_factor",
4242 .data = &ipv6_devconf.max_desync_factor,
4243 .maxlen = sizeof(int),
4244 .mode = 0644,
4245 .proc_handler = proc_dointvec,
4247 #endif
4249 .ctl_name = NET_IPV6_MAX_ADDRESSES,
4250 .procname = "max_addresses",
4251 .data = &ipv6_devconf.max_addresses,
4252 .maxlen = sizeof(int),
4253 .mode = 0644,
4254 .proc_handler = proc_dointvec,
4257 .ctl_name = NET_IPV6_ACCEPT_RA_DEFRTR,
4258 .procname = "accept_ra_defrtr",
4259 .data = &ipv6_devconf.accept_ra_defrtr,
4260 .maxlen = sizeof(int),
4261 .mode = 0644,
4262 .proc_handler = proc_dointvec,
4265 .ctl_name = NET_IPV6_ACCEPT_RA_PINFO,
4266 .procname = "accept_ra_pinfo",
4267 .data = &ipv6_devconf.accept_ra_pinfo,
4268 .maxlen = sizeof(int),
4269 .mode = 0644,
4270 .proc_handler = proc_dointvec,
4272 #ifdef CONFIG_IPV6_ROUTER_PREF
4274 .ctl_name = NET_IPV6_ACCEPT_RA_RTR_PREF,
4275 .procname = "accept_ra_rtr_pref",
4276 .data = &ipv6_devconf.accept_ra_rtr_pref,
4277 .maxlen = sizeof(int),
4278 .mode = 0644,
4279 .proc_handler = proc_dointvec,
4282 .ctl_name = NET_IPV6_RTR_PROBE_INTERVAL,
4283 .procname = "router_probe_interval",
4284 .data = &ipv6_devconf.rtr_probe_interval,
4285 .maxlen = sizeof(int),
4286 .mode = 0644,
4287 .proc_handler = proc_dointvec_jiffies,
4288 .strategy = sysctl_jiffies,
4290 #ifdef CONFIG_IPV6_ROUTE_INFO
4292 .ctl_name = NET_IPV6_ACCEPT_RA_RT_INFO_MAX_PLEN,
4293 .procname = "accept_ra_rt_info_max_plen",
4294 .data = &ipv6_devconf.accept_ra_rt_info_max_plen,
4295 .maxlen = sizeof(int),
4296 .mode = 0644,
4297 .proc_handler = proc_dointvec,
4299 #endif
4300 #endif
4302 .ctl_name = NET_IPV6_PROXY_NDP,
4303 .procname = "proxy_ndp",
4304 .data = &ipv6_devconf.proxy_ndp,
4305 .maxlen = sizeof(int),
4306 .mode = 0644,
4307 .proc_handler = proc_dointvec,
4310 .ctl_name = NET_IPV6_ACCEPT_SOURCE_ROUTE,
4311 .procname = "accept_source_route",
4312 .data = &ipv6_devconf.accept_source_route,
4313 .maxlen = sizeof(int),
4314 .mode = 0644,
4315 .proc_handler = proc_dointvec,
4317 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
4319 .ctl_name = CTL_UNNUMBERED,
4320 .procname = "optimistic_dad",
4321 .data = &ipv6_devconf.optimistic_dad,
4322 .maxlen = sizeof(int),
4323 .mode = 0644,
4324 .proc_handler = proc_dointvec,
4327 #endif
4328 #ifdef CONFIG_IPV6_MROUTE
4330 .ctl_name = CTL_UNNUMBERED,
4331 .procname = "mc_forwarding",
4332 .data = &ipv6_devconf.mc_forwarding,
4333 .maxlen = sizeof(int),
4334 .mode = 0444,
4335 .proc_handler = proc_dointvec,
4337 #endif
4339 .ctl_name = CTL_UNNUMBERED,
4340 .procname = "disable_ipv6",
4341 .data = &ipv6_devconf.disable_ipv6,
4342 .maxlen = sizeof(int),
4343 .mode = 0644,
4344 .proc_handler = addrconf_sysctl_disable,
4345 .strategy = sysctl_intvec,
4348 .ctl_name = CTL_UNNUMBERED,
4349 .procname = "accept_dad",
4350 .data = &ipv6_devconf.accept_dad,
4351 .maxlen = sizeof(int),
4352 .mode = 0644,
4353 .proc_handler = proc_dointvec,
4356 .ctl_name = 0, /* sentinel */
4361 static int __addrconf_sysctl_register(struct net *net, char *dev_name,
4362 int ctl_name, struct inet6_dev *idev, struct ipv6_devconf *p)
4364 int i;
4365 struct addrconf_sysctl_table *t;
4367 #define ADDRCONF_CTL_PATH_DEV 3
4369 struct ctl_path addrconf_ctl_path[] = {
4370 { .procname = "net", .ctl_name = CTL_NET, },
4371 { .procname = "ipv6", .ctl_name = NET_IPV6, },
4372 { .procname = "conf", .ctl_name = NET_IPV6_CONF, },
4373 { /* to be set */ },
4374 { },
4378 t = kmemdup(&addrconf_sysctl, sizeof(*t), GFP_KERNEL);
4379 if (t == NULL)
4380 goto out;
4382 for (i=0; t->addrconf_vars[i].data; i++) {
4383 t->addrconf_vars[i].data += (char*)p - (char*)&ipv6_devconf;
4384 t->addrconf_vars[i].extra1 = idev; /* embedded; no ref */
4385 t->addrconf_vars[i].extra2 = net;
4389 * Make a copy of dev_name, because '.procname' is regarded as const
4390 * by sysctl and we wouldn't want anyone to change it under our feet
4391 * (see SIOCSIFNAME).
4393 t->dev_name = kstrdup(dev_name, GFP_KERNEL);
4394 if (!t->dev_name)
4395 goto free;
4397 addrconf_ctl_path[ADDRCONF_CTL_PATH_DEV].procname = t->dev_name;
4398 addrconf_ctl_path[ADDRCONF_CTL_PATH_DEV].ctl_name = ctl_name;
4400 t->sysctl_header = register_net_sysctl_table(net, addrconf_ctl_path,
4401 t->addrconf_vars);
4402 if (t->sysctl_header == NULL)
4403 goto free_procname;
4405 p->sysctl = t;
4406 return 0;
4408 free_procname:
4409 kfree(t->dev_name);
4410 free:
4411 kfree(t);
4412 out:
4413 return -ENOBUFS;
4416 static void __addrconf_sysctl_unregister(struct ipv6_devconf *p)
4418 struct addrconf_sysctl_table *t;
4420 if (p->sysctl == NULL)
4421 return;
4423 t = p->sysctl;
4424 p->sysctl = NULL;
4425 unregister_sysctl_table(t->sysctl_header);
4426 kfree(t->dev_name);
4427 kfree(t);
4430 static void addrconf_sysctl_register(struct inet6_dev *idev)
4432 neigh_sysctl_register(idev->dev, idev->nd_parms, NET_IPV6,
4433 NET_IPV6_NEIGH, "ipv6",
4434 &ndisc_ifinfo_sysctl_change,
4435 ndisc_ifinfo_sysctl_strategy);
4436 __addrconf_sysctl_register(dev_net(idev->dev), idev->dev->name,
4437 idev->dev->ifindex, idev, &idev->cnf);
4440 static void addrconf_sysctl_unregister(struct inet6_dev *idev)
4442 __addrconf_sysctl_unregister(&idev->cnf);
4443 neigh_sysctl_unregister(idev->nd_parms);
4447 #endif
4449 static int addrconf_init_net(struct net *net)
4451 int err;
4452 struct ipv6_devconf *all, *dflt;
4454 err = -ENOMEM;
4455 all = &ipv6_devconf;
4456 dflt = &ipv6_devconf_dflt;
4458 if (net != &init_net) {
4459 all = kmemdup(all, sizeof(ipv6_devconf), GFP_KERNEL);
4460 if (all == NULL)
4461 goto err_alloc_all;
4463 dflt = kmemdup(dflt, sizeof(ipv6_devconf_dflt), GFP_KERNEL);
4464 if (dflt == NULL)
4465 goto err_alloc_dflt;
4466 } else {
4467 /* these will be inherited by all namespaces */
4468 dflt->autoconf = ipv6_defaults.autoconf;
4469 dflt->disable_ipv6 = ipv6_defaults.disable_ipv6;
4472 net->ipv6.devconf_all = all;
4473 net->ipv6.devconf_dflt = dflt;
4475 #ifdef CONFIG_SYSCTL
4476 err = __addrconf_sysctl_register(net, "all", NET_PROTO_CONF_ALL,
4477 NULL, all);
4478 if (err < 0)
4479 goto err_reg_all;
4481 err = __addrconf_sysctl_register(net, "default", NET_PROTO_CONF_DEFAULT,
4482 NULL, dflt);
4483 if (err < 0)
4484 goto err_reg_dflt;
4485 #endif
4486 return 0;
4488 #ifdef CONFIG_SYSCTL
4489 err_reg_dflt:
4490 __addrconf_sysctl_unregister(all);
4491 err_reg_all:
4492 kfree(dflt);
4493 #endif
4494 err_alloc_dflt:
4495 kfree(all);
4496 err_alloc_all:
4497 return err;
4500 static void addrconf_exit_net(struct net *net)
4502 #ifdef CONFIG_SYSCTL
4503 __addrconf_sysctl_unregister(net->ipv6.devconf_dflt);
4504 __addrconf_sysctl_unregister(net->ipv6.devconf_all);
4505 #endif
4506 if (net != &init_net) {
4507 kfree(net->ipv6.devconf_dflt);
4508 kfree(net->ipv6.devconf_all);
4512 static struct pernet_operations addrconf_ops = {
4513 .init = addrconf_init_net,
4514 .exit = addrconf_exit_net,
4518 * Device notifier
4521 int register_inet6addr_notifier(struct notifier_block *nb)
4523 return atomic_notifier_chain_register(&inet6addr_chain, nb);
4526 EXPORT_SYMBOL(register_inet6addr_notifier);
4528 int unregister_inet6addr_notifier(struct notifier_block *nb)
4530 return atomic_notifier_chain_unregister(&inet6addr_chain,nb);
4533 EXPORT_SYMBOL(unregister_inet6addr_notifier);
4536 * Init / cleanup code
4539 int __init addrconf_init(void)
4541 int err;
4543 if ((err = ipv6_addr_label_init()) < 0) {
4544 printk(KERN_CRIT "IPv6 Addrconf: cannot initialize default policy table: %d.\n",
4545 err);
4546 return err;
4549 register_pernet_subsys(&addrconf_ops);
4551 /* The addrconf netdev notifier requires that loopback_dev
4552 * has it's ipv6 private information allocated and setup
4553 * before it can bring up and give link-local addresses
4554 * to other devices which are up.
4556 * Unfortunately, loopback_dev is not necessarily the first
4557 * entry in the global dev_base list of net devices. In fact,
4558 * it is likely to be the very last entry on that list.
4559 * So this causes the notifier registry below to try and
4560 * give link-local addresses to all devices besides loopback_dev
4561 * first, then loopback_dev, which cases all the non-loopback_dev
4562 * devices to fail to get a link-local address.
4564 * So, as a temporary fix, allocate the ipv6 structure for
4565 * loopback_dev first by hand.
4566 * Longer term, all of the dependencies ipv6 has upon the loopback
4567 * device and it being up should be removed.
4569 rtnl_lock();
4570 if (!ipv6_add_dev(init_net.loopback_dev))
4571 err = -ENOMEM;
4572 rtnl_unlock();
4573 if (err)
4574 goto errlo;
4576 register_netdevice_notifier(&ipv6_dev_notf);
4578 addrconf_verify(0);
4580 err = __rtnl_register(PF_INET6, RTM_GETLINK, NULL, inet6_dump_ifinfo);
4581 if (err < 0)
4582 goto errout;
4584 /* Only the first call to __rtnl_register can fail */
4585 __rtnl_register(PF_INET6, RTM_NEWADDR, inet6_rtm_newaddr, NULL);
4586 __rtnl_register(PF_INET6, RTM_DELADDR, inet6_rtm_deladdr, NULL);
4587 __rtnl_register(PF_INET6, RTM_GETADDR, inet6_rtm_getaddr, inet6_dump_ifaddr);
4588 __rtnl_register(PF_INET6, RTM_GETMULTICAST, NULL, inet6_dump_ifmcaddr);
4589 __rtnl_register(PF_INET6, RTM_GETANYCAST, NULL, inet6_dump_ifacaddr);
4591 ipv6_addr_label_rtnl_register();
4593 return 0;
4594 errout:
4595 unregister_netdevice_notifier(&ipv6_dev_notf);
4596 errlo:
4597 unregister_pernet_subsys(&addrconf_ops);
4599 return err;
4602 void addrconf_cleanup(void)
4604 struct inet6_ifaddr *ifa;
4605 struct net_device *dev;
4606 int i;
4608 unregister_netdevice_notifier(&ipv6_dev_notf);
4609 unregister_pernet_subsys(&addrconf_ops);
4611 rtnl_lock();
4613 /* clean dev list */
4614 for_each_netdev(&init_net, dev) {
4615 if (__in6_dev_get(dev) == NULL)
4616 continue;
4617 addrconf_ifdown(dev, 1);
4619 addrconf_ifdown(init_net.loopback_dev, 2);
4622 * Check hash table.
4624 write_lock_bh(&addrconf_hash_lock);
4625 for (i=0; i < IN6_ADDR_HSIZE; i++) {
4626 for (ifa=inet6_addr_lst[i]; ifa; ) {
4627 struct inet6_ifaddr *bifa;
4629 bifa = ifa;
4630 ifa = ifa->lst_next;
4631 printk(KERN_DEBUG "bug: IPv6 address leakage detected: ifa=%p\n", bifa);
4632 /* Do not free it; something is wrong.
4633 Now we can investigate it with debugger.
4637 write_unlock_bh(&addrconf_hash_lock);
4639 del_timer(&addr_chk_timer);
4640 rtnl_unlock();