USB: simplify hub_restart() logic
[linux-2.6/mini2440.git] / net / ipv6 / addrconf.c
blob9f4fcce6379b88d4a6e0b0f388257b8907b4aec3
1 /*
2 * IPv6 Address [auto]configuration
3 * Linux INET6 implementation
5 * Authors:
6 * Pedro Roque <roque@di.fc.ul.pt>
7 * Alexey Kuznetsov <kuznet@ms2.inr.ac.ru>
9 * This program is free software; you can redistribute it and/or
10 * modify it under the terms of the GNU General Public License
11 * as published by the Free Software Foundation; either version
12 * 2 of the License, or (at your option) any later version.
16 * Changes:
18 * Janos Farkas : delete timer on ifdown
19 * <chexum@bankinf.banki.hu>
20 * Andi Kleen : kill double kfree on module
21 * unload.
22 * Maciej W. Rozycki : FDDI support
23 * sekiya@USAGI : Don't send too many RS
24 * packets.
25 * yoshfuji@USAGI : Fixed interval between DAD
26 * packets.
27 * YOSHIFUJI Hideaki @USAGI : improved accuracy of
28 * address validation timer.
29 * YOSHIFUJI Hideaki @USAGI : Privacy Extensions (RFC3041)
30 * support.
31 * Yuji SEKIYA @USAGI : Don't assign a same IPv6
32 * address on a same interface.
33 * YOSHIFUJI Hideaki @USAGI : ARCnet support
34 * YOSHIFUJI Hideaki @USAGI : convert /proc/net/if_inet6 to
35 * seq_file.
36 * YOSHIFUJI Hideaki @USAGI : improved source address
37 * selection; consider scope,
38 * status etc.
41 #include <linux/errno.h>
42 #include <linux/types.h>
43 #include <linux/socket.h>
44 #include <linux/sockios.h>
45 #include <linux/net.h>
46 #include <linux/in6.h>
47 #include <linux/netdevice.h>
48 #include <linux/if_addr.h>
49 #include <linux/if_arp.h>
50 #include <linux/if_arcnet.h>
51 #include <linux/if_infiniband.h>
52 #include <linux/route.h>
53 #include <linux/inetdevice.h>
54 #include <linux/init.h>
55 #ifdef CONFIG_SYSCTL
56 #include <linux/sysctl.h>
57 #endif
58 #include <linux/capability.h>
59 #include <linux/delay.h>
60 #include <linux/notifier.h>
61 #include <linux/string.h>
63 #include <net/net_namespace.h>
64 #include <net/sock.h>
65 #include <net/snmp.h>
67 #include <net/ipv6.h>
68 #include <net/protocol.h>
69 #include <net/ndisc.h>
70 #include <net/ip6_route.h>
71 #include <net/addrconf.h>
72 #include <net/tcp.h>
73 #include <net/ip.h>
74 #include <net/netlink.h>
75 #include <net/pkt_sched.h>
76 #include <linux/if_tunnel.h>
77 #include <linux/rtnetlink.h>
79 #ifdef CONFIG_IPV6_PRIVACY
80 #include <linux/random.h>
81 #endif
83 #include <asm/uaccess.h>
84 #include <asm/unaligned.h>
86 #include <linux/proc_fs.h>
87 #include <linux/seq_file.h>
89 /* Set to 3 to get tracing... */
90 #define ACONF_DEBUG 2
92 #if ACONF_DEBUG >= 3
93 #define ADBG(x) printk x
94 #else
95 #define ADBG(x)
96 #endif
98 #define INFINITY_LIFE_TIME 0xFFFFFFFF
99 #define TIME_DELTA(a,b) ((unsigned long)((long)(a) - (long)(b)))
101 #ifdef CONFIG_SYSCTL
102 static void addrconf_sysctl_register(struct inet6_dev *idev);
103 static void addrconf_sysctl_unregister(struct inet6_dev *idev);
104 #else
105 static inline void addrconf_sysctl_register(struct inet6_dev *idev)
109 static inline void addrconf_sysctl_unregister(struct inet6_dev *idev)
112 #endif
114 #ifdef CONFIG_IPV6_PRIVACY
115 static int __ipv6_regen_rndid(struct inet6_dev *idev);
116 static int __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr);
117 static void ipv6_regen_rndid(unsigned long data);
119 static int desync_factor = MAX_DESYNC_FACTOR * HZ;
120 #endif
122 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev);
123 static int ipv6_count_addresses(struct inet6_dev *idev);
126 * Configured unicast address hash table
128 static struct inet6_ifaddr *inet6_addr_lst[IN6_ADDR_HSIZE];
129 static DEFINE_RWLOCK(addrconf_hash_lock);
131 static void addrconf_verify(unsigned long);
133 static DEFINE_TIMER(addr_chk_timer, addrconf_verify, 0, 0);
134 static DEFINE_SPINLOCK(addrconf_verify_lock);
136 static void addrconf_join_anycast(struct inet6_ifaddr *ifp);
137 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp);
139 static int addrconf_ifdown(struct net_device *dev, int how);
141 static void addrconf_dad_start(struct inet6_ifaddr *ifp, u32 flags);
142 static void addrconf_dad_timer(unsigned long data);
143 static void addrconf_dad_completed(struct inet6_ifaddr *ifp);
144 static void addrconf_dad_run(struct inet6_dev *idev);
145 static void addrconf_rs_timer(unsigned long data);
146 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
147 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
149 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
150 struct prefix_info *pinfo);
151 static int ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
152 struct net_device *dev);
154 static ATOMIC_NOTIFIER_HEAD(inet6addr_chain);
156 struct ipv6_devconf ipv6_devconf __read_mostly = {
157 .forwarding = 0,
158 .hop_limit = IPV6_DEFAULT_HOPLIMIT,
159 .mtu6 = IPV6_MIN_MTU,
160 .accept_ra = 1,
161 .accept_redirects = 1,
162 .autoconf = 1,
163 .force_mld_version = 0,
164 .dad_transmits = 1,
165 .rtr_solicits = MAX_RTR_SOLICITATIONS,
166 .rtr_solicit_interval = RTR_SOLICITATION_INTERVAL,
167 .rtr_solicit_delay = MAX_RTR_SOLICITATION_DELAY,
168 #ifdef CONFIG_IPV6_PRIVACY
169 .use_tempaddr = 0,
170 .temp_valid_lft = TEMP_VALID_LIFETIME,
171 .temp_prefered_lft = TEMP_PREFERRED_LIFETIME,
172 .regen_max_retry = REGEN_MAX_RETRY,
173 .max_desync_factor = MAX_DESYNC_FACTOR,
174 #endif
175 .max_addresses = IPV6_MAX_ADDRESSES,
176 .accept_ra_defrtr = 1,
177 .accept_ra_pinfo = 1,
178 #ifdef CONFIG_IPV6_ROUTER_PREF
179 .accept_ra_rtr_pref = 1,
180 .rtr_probe_interval = 60 * HZ,
181 #ifdef CONFIG_IPV6_ROUTE_INFO
182 .accept_ra_rt_info_max_plen = 0,
183 #endif
184 #endif
185 .proxy_ndp = 0,
186 .accept_source_route = 0, /* we do not accept RH0 by default. */
187 .disable_ipv6 = 0,
188 .accept_dad = 1,
191 static struct ipv6_devconf ipv6_devconf_dflt __read_mostly = {
192 .forwarding = 0,
193 .hop_limit = IPV6_DEFAULT_HOPLIMIT,
194 .mtu6 = IPV6_MIN_MTU,
195 .accept_ra = 1,
196 .accept_redirects = 1,
197 .autoconf = 1,
198 .dad_transmits = 1,
199 .rtr_solicits = MAX_RTR_SOLICITATIONS,
200 .rtr_solicit_interval = RTR_SOLICITATION_INTERVAL,
201 .rtr_solicit_delay = MAX_RTR_SOLICITATION_DELAY,
202 #ifdef CONFIG_IPV6_PRIVACY
203 .use_tempaddr = 0,
204 .temp_valid_lft = TEMP_VALID_LIFETIME,
205 .temp_prefered_lft = TEMP_PREFERRED_LIFETIME,
206 .regen_max_retry = REGEN_MAX_RETRY,
207 .max_desync_factor = MAX_DESYNC_FACTOR,
208 #endif
209 .max_addresses = IPV6_MAX_ADDRESSES,
210 .accept_ra_defrtr = 1,
211 .accept_ra_pinfo = 1,
212 #ifdef CONFIG_IPV6_ROUTER_PREF
213 .accept_ra_rtr_pref = 1,
214 .rtr_probe_interval = 60 * HZ,
215 #ifdef CONFIG_IPV6_ROUTE_INFO
216 .accept_ra_rt_info_max_plen = 0,
217 #endif
218 #endif
219 .proxy_ndp = 0,
220 .accept_source_route = 0, /* we do not accept RH0 by default. */
221 .disable_ipv6 = 0,
222 .accept_dad = 1,
225 /* IPv6 Wildcard Address and Loopback Address defined by RFC2553 */
226 const struct in6_addr in6addr_any = IN6ADDR_ANY_INIT;
227 const struct in6_addr in6addr_loopback = IN6ADDR_LOOPBACK_INIT;
228 const struct in6_addr in6addr_linklocal_allnodes = IN6ADDR_LINKLOCAL_ALLNODES_INIT;
229 const struct in6_addr in6addr_linklocal_allrouters = IN6ADDR_LINKLOCAL_ALLROUTERS_INIT;
231 /* Check if a valid qdisc is available */
232 static inline bool addrconf_qdisc_ok(const struct net_device *dev)
234 return !qdisc_tx_is_noop(dev);
237 /* Check if a route is valid prefix route */
238 static inline int addrconf_is_prefix_route(const struct rt6_info *rt)
240 return ((rt->rt6i_flags & (RTF_GATEWAY | RTF_DEFAULT)) == 0);
243 static void addrconf_del_timer(struct inet6_ifaddr *ifp)
245 if (del_timer(&ifp->timer))
246 __in6_ifa_put(ifp);
249 enum addrconf_timer_t
251 AC_NONE,
252 AC_DAD,
253 AC_RS,
256 static void addrconf_mod_timer(struct inet6_ifaddr *ifp,
257 enum addrconf_timer_t what,
258 unsigned long when)
260 if (!del_timer(&ifp->timer))
261 in6_ifa_hold(ifp);
263 switch (what) {
264 case AC_DAD:
265 ifp->timer.function = addrconf_dad_timer;
266 break;
267 case AC_RS:
268 ifp->timer.function = addrconf_rs_timer;
269 break;
270 default:;
272 ifp->timer.expires = jiffies + when;
273 add_timer(&ifp->timer);
276 static int snmp6_alloc_dev(struct inet6_dev *idev)
278 if (snmp_mib_init((void **)idev->stats.ipv6,
279 sizeof(struct ipstats_mib)) < 0)
280 goto err_ip;
281 if (snmp_mib_init((void **)idev->stats.icmpv6,
282 sizeof(struct icmpv6_mib)) < 0)
283 goto err_icmp;
284 if (snmp_mib_init((void **)idev->stats.icmpv6msg,
285 sizeof(struct icmpv6msg_mib)) < 0)
286 goto err_icmpmsg;
288 return 0;
290 err_icmpmsg:
291 snmp_mib_free((void **)idev->stats.icmpv6);
292 err_icmp:
293 snmp_mib_free((void **)idev->stats.ipv6);
294 err_ip:
295 return -ENOMEM;
298 static void snmp6_free_dev(struct inet6_dev *idev)
300 snmp_mib_free((void **)idev->stats.icmpv6msg);
301 snmp_mib_free((void **)idev->stats.icmpv6);
302 snmp_mib_free((void **)idev->stats.ipv6);
305 /* Nobody refers to this device, we may destroy it. */
307 static void in6_dev_finish_destroy_rcu(struct rcu_head *head)
309 struct inet6_dev *idev = container_of(head, struct inet6_dev, rcu);
310 kfree(idev);
313 void in6_dev_finish_destroy(struct inet6_dev *idev)
315 struct net_device *dev = idev->dev;
316 BUG_TRAP(idev->addr_list==NULL);
317 BUG_TRAP(idev->mc_list==NULL);
318 #ifdef NET_REFCNT_DEBUG
319 printk(KERN_DEBUG "in6_dev_finish_destroy: %s\n", dev ? dev->name : "NIL");
320 #endif
321 dev_put(dev);
322 if (!idev->dead) {
323 printk("Freeing alive inet6 device %p\n", idev);
324 return;
326 snmp6_free_dev(idev);
327 call_rcu(&idev->rcu, in6_dev_finish_destroy_rcu);
330 EXPORT_SYMBOL(in6_dev_finish_destroy);
332 static struct inet6_dev * ipv6_add_dev(struct net_device *dev)
334 struct inet6_dev *ndev;
336 ASSERT_RTNL();
338 if (dev->mtu < IPV6_MIN_MTU)
339 return NULL;
341 ndev = kzalloc(sizeof(struct inet6_dev), GFP_KERNEL);
343 if (ndev == NULL)
344 return NULL;
346 rwlock_init(&ndev->lock);
347 ndev->dev = dev;
348 memcpy(&ndev->cnf, dev_net(dev)->ipv6.devconf_dflt, sizeof(ndev->cnf));
349 ndev->cnf.mtu6 = dev->mtu;
350 ndev->cnf.sysctl = NULL;
351 ndev->nd_parms = neigh_parms_alloc(dev, &nd_tbl);
352 if (ndev->nd_parms == NULL) {
353 kfree(ndev);
354 return NULL;
356 if (ndev->cnf.forwarding)
357 dev_disable_lro(dev);
358 /* We refer to the device */
359 dev_hold(dev);
361 if (snmp6_alloc_dev(ndev) < 0) {
362 ADBG((KERN_WARNING
363 "%s(): cannot allocate memory for statistics; dev=%s.\n",
364 __func__, dev->name));
365 neigh_parms_release(&nd_tbl, ndev->nd_parms);
366 ndev->dead = 1;
367 in6_dev_finish_destroy(ndev);
368 return NULL;
371 if (snmp6_register_dev(ndev) < 0) {
372 ADBG((KERN_WARNING
373 "%s(): cannot create /proc/net/dev_snmp6/%s\n",
374 __func__, dev->name));
375 neigh_parms_release(&nd_tbl, ndev->nd_parms);
376 ndev->dead = 1;
377 in6_dev_finish_destroy(ndev);
378 return NULL;
381 /* One reference from device. We must do this before
382 * we invoke __ipv6_regen_rndid().
384 in6_dev_hold(ndev);
386 if (dev->flags & (IFF_NOARP | IFF_LOOPBACK))
387 ndev->cnf.accept_dad = -1;
389 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
390 if (dev->type == ARPHRD_SIT && (dev->priv_flags & IFF_ISATAP)) {
391 printk(KERN_INFO
392 "%s: Disabled Multicast RS\n",
393 dev->name);
394 ndev->cnf.rtr_solicits = 0;
396 #endif
398 #ifdef CONFIG_IPV6_PRIVACY
399 setup_timer(&ndev->regen_timer, ipv6_regen_rndid, (unsigned long)ndev);
400 if ((dev->flags&IFF_LOOPBACK) ||
401 dev->type == ARPHRD_TUNNEL ||
402 dev->type == ARPHRD_TUNNEL6 ||
403 dev->type == ARPHRD_SIT ||
404 dev->type == ARPHRD_NONE) {
405 printk(KERN_INFO
406 "%s: Disabled Privacy Extensions\n",
407 dev->name);
408 ndev->cnf.use_tempaddr = -1;
409 } else {
410 in6_dev_hold(ndev);
411 ipv6_regen_rndid((unsigned long) ndev);
413 #endif
415 if (netif_running(dev) && addrconf_qdisc_ok(dev))
416 ndev->if_flags |= IF_READY;
418 ipv6_mc_init_dev(ndev);
419 ndev->tstamp = jiffies;
420 addrconf_sysctl_register(ndev);
421 /* protected by rtnl_lock */
422 rcu_assign_pointer(dev->ip6_ptr, ndev);
424 /* Join all-node multicast group */
425 ipv6_dev_mc_inc(dev, &in6addr_linklocal_allnodes);
427 return ndev;
430 static struct inet6_dev * ipv6_find_idev(struct net_device *dev)
432 struct inet6_dev *idev;
434 ASSERT_RTNL();
436 if ((idev = __in6_dev_get(dev)) == NULL) {
437 if ((idev = ipv6_add_dev(dev)) == NULL)
438 return NULL;
441 if (dev->flags&IFF_UP)
442 ipv6_mc_up(idev);
443 return idev;
446 #ifdef CONFIG_SYSCTL
447 static void dev_forward_change(struct inet6_dev *idev)
449 struct net_device *dev;
450 struct inet6_ifaddr *ifa;
452 if (!idev)
453 return;
454 dev = idev->dev;
455 if (idev->cnf.forwarding)
456 dev_disable_lro(dev);
457 if (dev && (dev->flags & IFF_MULTICAST)) {
458 if (idev->cnf.forwarding)
459 ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
460 else
461 ipv6_dev_mc_dec(dev, &in6addr_linklocal_allrouters);
463 for (ifa=idev->addr_list; ifa; ifa=ifa->if_next) {
464 if (ifa->flags&IFA_F_TENTATIVE)
465 continue;
466 if (idev->cnf.forwarding)
467 addrconf_join_anycast(ifa);
468 else
469 addrconf_leave_anycast(ifa);
474 static void addrconf_forward_change(struct net *net, __s32 newf)
476 struct net_device *dev;
477 struct inet6_dev *idev;
479 read_lock(&dev_base_lock);
480 for_each_netdev(net, dev) {
481 rcu_read_lock();
482 idev = __in6_dev_get(dev);
483 if (idev) {
484 int changed = (!idev->cnf.forwarding) ^ (!newf);
485 idev->cnf.forwarding = newf;
486 if (changed)
487 dev_forward_change(idev);
489 rcu_read_unlock();
491 read_unlock(&dev_base_lock);
494 static void addrconf_fixup_forwarding(struct ctl_table *table, int *p, int old)
496 struct net *net;
498 net = (struct net *)table->extra2;
499 if (p == &net->ipv6.devconf_dflt->forwarding)
500 return;
502 rtnl_lock();
503 if (p == &net->ipv6.devconf_all->forwarding) {
504 __s32 newf = net->ipv6.devconf_all->forwarding;
505 net->ipv6.devconf_dflt->forwarding = newf;
506 addrconf_forward_change(net, newf);
507 } else if ((!*p) ^ (!old))
508 dev_forward_change((struct inet6_dev *)table->extra1);
509 rtnl_unlock();
511 if (*p)
512 rt6_purge_dflt_routers(net);
514 #endif
516 /* Nobody refers to this ifaddr, destroy it */
518 void inet6_ifa_finish_destroy(struct inet6_ifaddr *ifp)
520 BUG_TRAP(ifp->if_next==NULL);
521 BUG_TRAP(ifp->lst_next==NULL);
522 #ifdef NET_REFCNT_DEBUG
523 printk(KERN_DEBUG "inet6_ifa_finish_destroy\n");
524 #endif
526 in6_dev_put(ifp->idev);
528 if (del_timer(&ifp->timer))
529 printk("Timer is still running, when freeing ifa=%p\n", ifp);
531 if (!ifp->dead) {
532 printk("Freeing alive inet6 address %p\n", ifp);
533 return;
535 dst_release(&ifp->rt->u.dst);
537 kfree(ifp);
540 static void
541 ipv6_link_dev_addr(struct inet6_dev *idev, struct inet6_ifaddr *ifp)
543 struct inet6_ifaddr *ifa, **ifap;
544 int ifp_scope = ipv6_addr_src_scope(&ifp->addr);
547 * Each device address list is sorted in order of scope -
548 * global before linklocal.
550 for (ifap = &idev->addr_list; (ifa = *ifap) != NULL;
551 ifap = &ifa->if_next) {
552 if (ifp_scope >= ipv6_addr_src_scope(&ifa->addr))
553 break;
556 ifp->if_next = *ifap;
557 *ifap = ifp;
561 * Hash function taken from net_alias.c
563 static u8 ipv6_addr_hash(const struct in6_addr *addr)
565 __u32 word;
568 * We perform the hash function over the last 64 bits of the address
569 * This will include the IEEE address token on links that support it.
572 word = (__force u32)(addr->s6_addr32[2] ^ addr->s6_addr32[3]);
573 word ^= (word >> 16);
574 word ^= (word >> 8);
576 return ((word ^ (word >> 4)) & 0x0f);
579 /* On success it returns ifp with increased reference count */
581 static struct inet6_ifaddr *
582 ipv6_add_addr(struct inet6_dev *idev, const struct in6_addr *addr, int pfxlen,
583 int scope, u32 flags)
585 struct inet6_ifaddr *ifa = NULL;
586 struct rt6_info *rt;
587 int hash;
588 int err = 0;
589 int addr_type = ipv6_addr_type(addr);
591 if (addr_type == IPV6_ADDR_ANY ||
592 addr_type & IPV6_ADDR_MULTICAST ||
593 (!(idev->dev->flags & IFF_LOOPBACK) &&
594 addr_type & IPV6_ADDR_LOOPBACK))
595 return ERR_PTR(-EADDRNOTAVAIL);
597 rcu_read_lock_bh();
598 if (idev->dead) {
599 err = -ENODEV; /*XXX*/
600 goto out2;
603 write_lock(&addrconf_hash_lock);
605 /* Ignore adding duplicate addresses on an interface */
606 if (ipv6_chk_same_addr(dev_net(idev->dev), addr, idev->dev)) {
607 ADBG(("ipv6_add_addr: already assigned\n"));
608 err = -EEXIST;
609 goto out;
612 ifa = kzalloc(sizeof(struct inet6_ifaddr), GFP_ATOMIC);
614 if (ifa == NULL) {
615 ADBG(("ipv6_add_addr: malloc failed\n"));
616 err = -ENOBUFS;
617 goto out;
620 rt = addrconf_dst_alloc(idev, addr, 0);
621 if (IS_ERR(rt)) {
622 err = PTR_ERR(rt);
623 goto out;
626 ipv6_addr_copy(&ifa->addr, addr);
628 spin_lock_init(&ifa->lock);
629 init_timer(&ifa->timer);
630 ifa->timer.data = (unsigned long) ifa;
631 ifa->scope = scope;
632 ifa->prefix_len = pfxlen;
633 ifa->flags = flags | IFA_F_TENTATIVE;
634 ifa->cstamp = ifa->tstamp = jiffies;
636 ifa->rt = rt;
639 * part one of RFC 4429, section 3.3
640 * We should not configure an address as
641 * optimistic if we do not yet know the link
642 * layer address of our nexhop router
645 if (rt->rt6i_nexthop == NULL)
646 ifa->flags &= ~IFA_F_OPTIMISTIC;
648 ifa->idev = idev;
649 in6_dev_hold(idev);
650 /* For caller */
651 in6_ifa_hold(ifa);
653 /* Add to big hash table */
654 hash = ipv6_addr_hash(addr);
656 ifa->lst_next = inet6_addr_lst[hash];
657 inet6_addr_lst[hash] = ifa;
658 in6_ifa_hold(ifa);
659 write_unlock(&addrconf_hash_lock);
661 write_lock(&idev->lock);
662 /* Add to inet6_dev unicast addr list. */
663 ipv6_link_dev_addr(idev, ifa);
665 #ifdef CONFIG_IPV6_PRIVACY
666 if (ifa->flags&IFA_F_TEMPORARY) {
667 ifa->tmp_next = idev->tempaddr_list;
668 idev->tempaddr_list = ifa;
669 in6_ifa_hold(ifa);
671 #endif
673 in6_ifa_hold(ifa);
674 write_unlock(&idev->lock);
675 out2:
676 rcu_read_unlock_bh();
678 if (likely(err == 0))
679 atomic_notifier_call_chain(&inet6addr_chain, NETDEV_UP, ifa);
680 else {
681 kfree(ifa);
682 ifa = ERR_PTR(err);
685 return ifa;
686 out:
687 write_unlock(&addrconf_hash_lock);
688 goto out2;
691 /* This function wants to get referenced ifp and releases it before return */
693 static void ipv6_del_addr(struct inet6_ifaddr *ifp)
695 struct inet6_ifaddr *ifa, **ifap;
696 struct inet6_dev *idev = ifp->idev;
697 int hash;
698 int deleted = 0, onlink = 0;
699 unsigned long expires = jiffies;
701 hash = ipv6_addr_hash(&ifp->addr);
703 ifp->dead = 1;
705 write_lock_bh(&addrconf_hash_lock);
706 for (ifap = &inet6_addr_lst[hash]; (ifa=*ifap) != NULL;
707 ifap = &ifa->lst_next) {
708 if (ifa == ifp) {
709 *ifap = ifa->lst_next;
710 __in6_ifa_put(ifp);
711 ifa->lst_next = NULL;
712 break;
715 write_unlock_bh(&addrconf_hash_lock);
717 write_lock_bh(&idev->lock);
718 #ifdef CONFIG_IPV6_PRIVACY
719 if (ifp->flags&IFA_F_TEMPORARY) {
720 for (ifap = &idev->tempaddr_list; (ifa=*ifap) != NULL;
721 ifap = &ifa->tmp_next) {
722 if (ifa == ifp) {
723 *ifap = ifa->tmp_next;
724 if (ifp->ifpub) {
725 in6_ifa_put(ifp->ifpub);
726 ifp->ifpub = NULL;
728 __in6_ifa_put(ifp);
729 ifa->tmp_next = NULL;
730 break;
734 #endif
736 for (ifap = &idev->addr_list; (ifa=*ifap) != NULL;) {
737 if (ifa == ifp) {
738 *ifap = ifa->if_next;
739 __in6_ifa_put(ifp);
740 ifa->if_next = NULL;
741 if (!(ifp->flags & IFA_F_PERMANENT) || onlink > 0)
742 break;
743 deleted = 1;
744 continue;
745 } else if (ifp->flags & IFA_F_PERMANENT) {
746 if (ipv6_prefix_equal(&ifa->addr, &ifp->addr,
747 ifp->prefix_len)) {
748 if (ifa->flags & IFA_F_PERMANENT) {
749 onlink = 1;
750 if (deleted)
751 break;
752 } else {
753 unsigned long lifetime;
755 if (!onlink)
756 onlink = -1;
758 spin_lock(&ifa->lock);
760 lifetime = addrconf_timeout_fixup(ifa->valid_lft, HZ);
762 * Note: Because this address is
763 * not permanent, lifetime <
764 * LONG_MAX / HZ here.
766 if (time_before(expires,
767 ifa->tstamp + lifetime * HZ))
768 expires = ifa->tstamp + lifetime * HZ;
769 spin_unlock(&ifa->lock);
773 ifap = &ifa->if_next;
775 write_unlock_bh(&idev->lock);
777 addrconf_del_timer(ifp);
779 ipv6_ifa_notify(RTM_DELADDR, ifp);
781 atomic_notifier_call_chain(&inet6addr_chain, NETDEV_DOWN, ifp);
784 * Purge or update corresponding prefix
786 * 1) we don't purge prefix here if address was not permanent.
787 * prefix is managed by its own lifetime.
788 * 2) if there're no addresses, delete prefix.
789 * 3) if there're still other permanent address(es),
790 * corresponding prefix is still permanent.
791 * 4) otherwise, update prefix lifetime to the
792 * longest valid lifetime among the corresponding
793 * addresses on the device.
794 * Note: subsequent RA will update lifetime.
796 * --yoshfuji
798 if ((ifp->flags & IFA_F_PERMANENT) && onlink < 1) {
799 struct in6_addr prefix;
800 struct rt6_info *rt;
801 struct net *net = dev_net(ifp->idev->dev);
802 ipv6_addr_prefix(&prefix, &ifp->addr, ifp->prefix_len);
803 rt = rt6_lookup(net, &prefix, NULL, ifp->idev->dev->ifindex, 1);
805 if (rt && addrconf_is_prefix_route(rt)) {
806 if (onlink == 0) {
807 ip6_del_rt(rt);
808 rt = NULL;
809 } else if (!(rt->rt6i_flags & RTF_EXPIRES)) {
810 rt->rt6i_expires = expires;
811 rt->rt6i_flags |= RTF_EXPIRES;
814 dst_release(&rt->u.dst);
817 in6_ifa_put(ifp);
820 #ifdef CONFIG_IPV6_PRIVACY
821 static int ipv6_create_tempaddr(struct inet6_ifaddr *ifp, struct inet6_ifaddr *ift)
823 struct inet6_dev *idev = ifp->idev;
824 struct in6_addr addr, *tmpaddr;
825 unsigned long tmp_prefered_lft, tmp_valid_lft, tmp_cstamp, tmp_tstamp;
826 unsigned long regen_advance;
827 int tmp_plen;
828 int ret = 0;
829 int max_addresses;
830 u32 addr_flags;
832 write_lock(&idev->lock);
833 if (ift) {
834 spin_lock_bh(&ift->lock);
835 memcpy(&addr.s6_addr[8], &ift->addr.s6_addr[8], 8);
836 spin_unlock_bh(&ift->lock);
837 tmpaddr = &addr;
838 } else {
839 tmpaddr = NULL;
841 retry:
842 in6_dev_hold(idev);
843 if (idev->cnf.use_tempaddr <= 0) {
844 write_unlock(&idev->lock);
845 printk(KERN_INFO
846 "ipv6_create_tempaddr(): use_tempaddr is disabled.\n");
847 in6_dev_put(idev);
848 ret = -1;
849 goto out;
851 spin_lock_bh(&ifp->lock);
852 if (ifp->regen_count++ >= idev->cnf.regen_max_retry) {
853 idev->cnf.use_tempaddr = -1; /*XXX*/
854 spin_unlock_bh(&ifp->lock);
855 write_unlock(&idev->lock);
856 printk(KERN_WARNING
857 "ipv6_create_tempaddr(): regeneration time exceeded. disabled temporary address support.\n");
858 in6_dev_put(idev);
859 ret = -1;
860 goto out;
862 in6_ifa_hold(ifp);
863 memcpy(addr.s6_addr, ifp->addr.s6_addr, 8);
864 if (__ipv6_try_regen_rndid(idev, tmpaddr) < 0) {
865 spin_unlock_bh(&ifp->lock);
866 write_unlock(&idev->lock);
867 printk(KERN_WARNING
868 "ipv6_create_tempaddr(): regeneration of randomized interface id failed.\n");
869 in6_ifa_put(ifp);
870 in6_dev_put(idev);
871 ret = -1;
872 goto out;
874 memcpy(&addr.s6_addr[8], idev->rndid, 8);
875 tmp_valid_lft = min_t(__u32,
876 ifp->valid_lft,
877 idev->cnf.temp_valid_lft);
878 tmp_prefered_lft = min_t(__u32,
879 ifp->prefered_lft,
880 idev->cnf.temp_prefered_lft - desync_factor / HZ);
881 tmp_plen = ifp->prefix_len;
882 max_addresses = idev->cnf.max_addresses;
883 tmp_cstamp = ifp->cstamp;
884 tmp_tstamp = ifp->tstamp;
885 spin_unlock_bh(&ifp->lock);
887 regen_advance = idev->cnf.regen_max_retry *
888 idev->cnf.dad_transmits *
889 idev->nd_parms->retrans_time / HZ;
890 write_unlock(&idev->lock);
892 /* A temporary address is created only if this calculated Preferred
893 * Lifetime is greater than REGEN_ADVANCE time units. In particular,
894 * an implementation must not create a temporary address with a zero
895 * Preferred Lifetime.
897 if (tmp_prefered_lft <= regen_advance) {
898 in6_ifa_put(ifp);
899 in6_dev_put(idev);
900 ret = -1;
901 goto out;
904 addr_flags = IFA_F_TEMPORARY;
905 /* set in addrconf_prefix_rcv() */
906 if (ifp->flags & IFA_F_OPTIMISTIC)
907 addr_flags |= IFA_F_OPTIMISTIC;
909 ift = !max_addresses ||
910 ipv6_count_addresses(idev) < max_addresses ?
911 ipv6_add_addr(idev, &addr, tmp_plen,
912 ipv6_addr_type(&addr)&IPV6_ADDR_SCOPE_MASK,
913 addr_flags) : NULL;
914 if (!ift || IS_ERR(ift)) {
915 in6_ifa_put(ifp);
916 in6_dev_put(idev);
917 printk(KERN_INFO
918 "ipv6_create_tempaddr(): retry temporary address regeneration.\n");
919 tmpaddr = &addr;
920 write_lock(&idev->lock);
921 goto retry;
924 spin_lock_bh(&ift->lock);
925 ift->ifpub = ifp;
926 ift->valid_lft = tmp_valid_lft;
927 ift->prefered_lft = tmp_prefered_lft;
928 ift->cstamp = tmp_cstamp;
929 ift->tstamp = tmp_tstamp;
930 spin_unlock_bh(&ift->lock);
932 addrconf_dad_start(ift, 0);
933 in6_ifa_put(ift);
934 in6_dev_put(idev);
935 out:
936 return ret;
938 #endif
941 * Choose an appropriate source address (RFC3484)
943 enum {
944 IPV6_SADDR_RULE_INIT = 0,
945 IPV6_SADDR_RULE_LOCAL,
946 IPV6_SADDR_RULE_SCOPE,
947 IPV6_SADDR_RULE_PREFERRED,
948 #ifdef CONFIG_IPV6_MIP6
949 IPV6_SADDR_RULE_HOA,
950 #endif
951 IPV6_SADDR_RULE_OIF,
952 IPV6_SADDR_RULE_LABEL,
953 #ifdef CONFIG_IPV6_PRIVACY
954 IPV6_SADDR_RULE_PRIVACY,
955 #endif
956 IPV6_SADDR_RULE_ORCHID,
957 IPV6_SADDR_RULE_PREFIX,
958 IPV6_SADDR_RULE_MAX
961 struct ipv6_saddr_score {
962 int rule;
963 int addr_type;
964 struct inet6_ifaddr *ifa;
965 DECLARE_BITMAP(scorebits, IPV6_SADDR_RULE_MAX);
966 int scopedist;
967 int matchlen;
970 struct ipv6_saddr_dst {
971 const struct in6_addr *addr;
972 int ifindex;
973 int scope;
974 int label;
975 unsigned int prefs;
978 static inline int ipv6_saddr_preferred(int type)
980 if (type & (IPV6_ADDR_MAPPED|IPV6_ADDR_COMPATv4|
981 IPV6_ADDR_LOOPBACK|IPV6_ADDR_RESERVED))
982 return 1;
983 return 0;
986 static int ipv6_get_saddr_eval(struct net *net,
987 struct ipv6_saddr_score *score,
988 struct ipv6_saddr_dst *dst,
989 int i)
991 int ret;
993 if (i <= score->rule) {
994 switch (i) {
995 case IPV6_SADDR_RULE_SCOPE:
996 ret = score->scopedist;
997 break;
998 case IPV6_SADDR_RULE_PREFIX:
999 ret = score->matchlen;
1000 break;
1001 default:
1002 ret = !!test_bit(i, score->scorebits);
1004 goto out;
1007 switch (i) {
1008 case IPV6_SADDR_RULE_INIT:
1009 /* Rule 0: remember if hiscore is not ready yet */
1010 ret = !!score->ifa;
1011 break;
1012 case IPV6_SADDR_RULE_LOCAL:
1013 /* Rule 1: Prefer same address */
1014 ret = ipv6_addr_equal(&score->ifa->addr, dst->addr);
1015 break;
1016 case IPV6_SADDR_RULE_SCOPE:
1017 /* Rule 2: Prefer appropriate scope
1019 * ret
1021 * -1 | d 15
1022 * ---+--+-+---> scope
1024 * | d is scope of the destination.
1025 * B-d | \
1026 * | \ <- smaller scope is better if
1027 * B-15 | \ if scope is enough for destinaion.
1028 * | ret = B - scope (-1 <= scope >= d <= 15).
1029 * d-C-1 | /
1030 * |/ <- greater is better
1031 * -C / if scope is not enough for destination.
1032 * /| ret = scope - C (-1 <= d < scope <= 15).
1034 * d - C - 1 < B -15 (for all -1 <= d <= 15).
1035 * C > d + 14 - B >= 15 + 14 - B = 29 - B.
1036 * Assume B = 0 and we get C > 29.
1038 ret = __ipv6_addr_src_scope(score->addr_type);
1039 if (ret >= dst->scope)
1040 ret = -ret;
1041 else
1042 ret -= 128; /* 30 is enough */
1043 score->scopedist = ret;
1044 break;
1045 case IPV6_SADDR_RULE_PREFERRED:
1046 /* Rule 3: Avoid deprecated and optimistic addresses */
1047 ret = ipv6_saddr_preferred(score->addr_type) ||
1048 !(score->ifa->flags & (IFA_F_DEPRECATED|IFA_F_OPTIMISTIC));
1049 break;
1050 #ifdef CONFIG_IPV6_MIP6
1051 case IPV6_SADDR_RULE_HOA:
1053 /* Rule 4: Prefer home address */
1054 int prefhome = !(dst->prefs & IPV6_PREFER_SRC_COA);
1055 ret = !(score->ifa->flags & IFA_F_HOMEADDRESS) ^ prefhome;
1056 break;
1058 #endif
1059 case IPV6_SADDR_RULE_OIF:
1060 /* Rule 5: Prefer outgoing interface */
1061 ret = (!dst->ifindex ||
1062 dst->ifindex == score->ifa->idev->dev->ifindex);
1063 break;
1064 case IPV6_SADDR_RULE_LABEL:
1065 /* Rule 6: Prefer matching label */
1066 ret = ipv6_addr_label(net,
1067 &score->ifa->addr, score->addr_type,
1068 score->ifa->idev->dev->ifindex) == dst->label;
1069 break;
1070 #ifdef CONFIG_IPV6_PRIVACY
1071 case IPV6_SADDR_RULE_PRIVACY:
1073 /* Rule 7: Prefer public address
1074 * Note: prefer temprary address if use_tempaddr >= 2
1076 int preftmp = dst->prefs & (IPV6_PREFER_SRC_PUBLIC|IPV6_PREFER_SRC_TMP) ?
1077 !!(dst->prefs & IPV6_PREFER_SRC_TMP) :
1078 score->ifa->idev->cnf.use_tempaddr >= 2;
1079 ret = (!(score->ifa->flags & IFA_F_TEMPORARY)) ^ preftmp;
1080 break;
1082 #endif
1083 case IPV6_SADDR_RULE_ORCHID:
1084 /* Rule 8-: Prefer ORCHID vs ORCHID or
1085 * non-ORCHID vs non-ORCHID
1087 ret = !(ipv6_addr_orchid(&score->ifa->addr) ^
1088 ipv6_addr_orchid(dst->addr));
1089 break;
1090 case IPV6_SADDR_RULE_PREFIX:
1091 /* Rule 8: Use longest matching prefix */
1092 score->matchlen = ret = ipv6_addr_diff(&score->ifa->addr,
1093 dst->addr);
1094 break;
1095 default:
1096 ret = 0;
1099 if (ret)
1100 __set_bit(i, score->scorebits);
1101 score->rule = i;
1102 out:
1103 return ret;
1106 int ipv6_dev_get_saddr(struct net_device *dst_dev,
1107 const struct in6_addr *daddr, unsigned int prefs,
1108 struct in6_addr *saddr)
1110 struct ipv6_saddr_score scores[2],
1111 *score = &scores[0], *hiscore = &scores[1];
1112 struct net *net = dev_net(dst_dev);
1113 struct ipv6_saddr_dst dst;
1114 struct net_device *dev;
1115 int dst_type;
1117 dst_type = __ipv6_addr_type(daddr);
1118 dst.addr = daddr;
1119 dst.ifindex = dst_dev ? dst_dev->ifindex : 0;
1120 dst.scope = __ipv6_addr_src_scope(dst_type);
1121 dst.label = ipv6_addr_label(net, daddr, dst_type, dst.ifindex);
1122 dst.prefs = prefs;
1124 hiscore->rule = -1;
1125 hiscore->ifa = NULL;
1127 read_lock(&dev_base_lock);
1128 rcu_read_lock();
1130 for_each_netdev(net, dev) {
1131 struct inet6_dev *idev;
1133 /* Candidate Source Address (section 4)
1134 * - multicast and link-local destination address,
1135 * the set of candidate source address MUST only
1136 * include addresses assigned to interfaces
1137 * belonging to the same link as the outgoing
1138 * interface.
1139 * (- For site-local destination addresses, the
1140 * set of candidate source addresses MUST only
1141 * include addresses assigned to interfaces
1142 * belonging to the same site as the outgoing
1143 * interface.)
1145 if (((dst_type & IPV6_ADDR_MULTICAST) ||
1146 dst.scope <= IPV6_ADDR_SCOPE_LINKLOCAL) &&
1147 dst.ifindex && dev->ifindex != dst.ifindex)
1148 continue;
1150 idev = __in6_dev_get(dev);
1151 if (!idev)
1152 continue;
1154 read_lock_bh(&idev->lock);
1155 for (score->ifa = idev->addr_list; score->ifa; score->ifa = score->ifa->if_next) {
1156 int i;
1159 * - Tentative Address (RFC2462 section 5.4)
1160 * - A tentative address is not considered
1161 * "assigned to an interface" in the traditional
1162 * sense, unless it is also flagged as optimistic.
1163 * - Candidate Source Address (section 4)
1164 * - In any case, anycast addresses, multicast
1165 * addresses, and the unspecified address MUST
1166 * NOT be included in a candidate set.
1168 if ((score->ifa->flags & IFA_F_TENTATIVE) &&
1169 (!(score->ifa->flags & IFA_F_OPTIMISTIC)))
1170 continue;
1172 score->addr_type = __ipv6_addr_type(&score->ifa->addr);
1174 if (unlikely(score->addr_type == IPV6_ADDR_ANY ||
1175 score->addr_type & IPV6_ADDR_MULTICAST)) {
1176 LIMIT_NETDEBUG(KERN_DEBUG
1177 "ADDRCONF: unspecified / multicast address "
1178 "assigned as unicast address on %s",
1179 dev->name);
1180 continue;
1183 score->rule = -1;
1184 bitmap_zero(score->scorebits, IPV6_SADDR_RULE_MAX);
1186 for (i = 0; i < IPV6_SADDR_RULE_MAX; i++) {
1187 int minihiscore, miniscore;
1189 minihiscore = ipv6_get_saddr_eval(net, hiscore, &dst, i);
1190 miniscore = ipv6_get_saddr_eval(net, score, &dst, i);
1192 if (minihiscore > miniscore) {
1193 if (i == IPV6_SADDR_RULE_SCOPE &&
1194 score->scopedist > 0) {
1196 * special case:
1197 * each remaining entry
1198 * has too small (not enough)
1199 * scope, because ifa entries
1200 * are sorted by their scope
1201 * values.
1203 goto try_nextdev;
1205 break;
1206 } else if (minihiscore < miniscore) {
1207 struct ipv6_saddr_score *tmp;
1209 if (hiscore->ifa)
1210 in6_ifa_put(hiscore->ifa);
1212 in6_ifa_hold(score->ifa);
1214 tmp = hiscore;
1215 hiscore = score;
1216 score = tmp;
1218 /* restore our iterator */
1219 score->ifa = hiscore->ifa;
1221 break;
1225 try_nextdev:
1226 read_unlock_bh(&idev->lock);
1228 rcu_read_unlock();
1229 read_unlock(&dev_base_lock);
1231 if (!hiscore->ifa)
1232 return -EADDRNOTAVAIL;
1234 ipv6_addr_copy(saddr, &hiscore->ifa->addr);
1235 in6_ifa_put(hiscore->ifa);
1236 return 0;
1239 EXPORT_SYMBOL(ipv6_dev_get_saddr);
1241 int ipv6_get_lladdr(struct net_device *dev, struct in6_addr *addr,
1242 unsigned char banned_flags)
1244 struct inet6_dev *idev;
1245 int err = -EADDRNOTAVAIL;
1247 rcu_read_lock();
1248 if ((idev = __in6_dev_get(dev)) != NULL) {
1249 struct inet6_ifaddr *ifp;
1251 read_lock_bh(&idev->lock);
1252 for (ifp=idev->addr_list; ifp; ifp=ifp->if_next) {
1253 if (ifp->scope == IFA_LINK && !(ifp->flags & banned_flags)) {
1254 ipv6_addr_copy(addr, &ifp->addr);
1255 err = 0;
1256 break;
1259 read_unlock_bh(&idev->lock);
1261 rcu_read_unlock();
1262 return err;
1265 static int ipv6_count_addresses(struct inet6_dev *idev)
1267 int cnt = 0;
1268 struct inet6_ifaddr *ifp;
1270 read_lock_bh(&idev->lock);
1271 for (ifp=idev->addr_list; ifp; ifp=ifp->if_next)
1272 cnt++;
1273 read_unlock_bh(&idev->lock);
1274 return cnt;
1277 int ipv6_chk_addr(struct net *net, struct in6_addr *addr,
1278 struct net_device *dev, int strict)
1280 struct inet6_ifaddr * ifp;
1281 u8 hash = ipv6_addr_hash(addr);
1283 read_lock_bh(&addrconf_hash_lock);
1284 for(ifp = inet6_addr_lst[hash]; ifp; ifp=ifp->lst_next) {
1285 if (!net_eq(dev_net(ifp->idev->dev), net))
1286 continue;
1287 if (ipv6_addr_equal(&ifp->addr, addr) &&
1288 !(ifp->flags&IFA_F_TENTATIVE)) {
1289 if (dev == NULL || ifp->idev->dev == dev ||
1290 !(ifp->scope&(IFA_LINK|IFA_HOST) || strict))
1291 break;
1294 read_unlock_bh(&addrconf_hash_lock);
1295 return ifp != NULL;
1297 EXPORT_SYMBOL(ipv6_chk_addr);
1299 static
1300 int ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
1301 struct net_device *dev)
1303 struct inet6_ifaddr * ifp;
1304 u8 hash = ipv6_addr_hash(addr);
1306 for(ifp = inet6_addr_lst[hash]; ifp; ifp=ifp->lst_next) {
1307 if (!net_eq(dev_net(ifp->idev->dev), net))
1308 continue;
1309 if (ipv6_addr_equal(&ifp->addr, addr)) {
1310 if (dev == NULL || ifp->idev->dev == dev)
1311 break;
1314 return ifp != NULL;
1317 int ipv6_chk_prefix(struct in6_addr *addr, struct net_device *dev)
1319 struct inet6_dev *idev;
1320 struct inet6_ifaddr *ifa;
1321 int onlink;
1323 onlink = 0;
1324 rcu_read_lock();
1325 idev = __in6_dev_get(dev);
1326 if (idev) {
1327 read_lock_bh(&idev->lock);
1328 for (ifa = idev->addr_list; ifa; ifa = ifa->if_next) {
1329 onlink = ipv6_prefix_equal(addr, &ifa->addr,
1330 ifa->prefix_len);
1331 if (onlink)
1332 break;
1334 read_unlock_bh(&idev->lock);
1336 rcu_read_unlock();
1337 return onlink;
1340 EXPORT_SYMBOL(ipv6_chk_prefix);
1342 struct inet6_ifaddr *ipv6_get_ifaddr(struct net *net, const struct in6_addr *addr,
1343 struct net_device *dev, int strict)
1345 struct inet6_ifaddr * ifp;
1346 u8 hash = ipv6_addr_hash(addr);
1348 read_lock_bh(&addrconf_hash_lock);
1349 for(ifp = inet6_addr_lst[hash]; ifp; ifp=ifp->lst_next) {
1350 if (!net_eq(dev_net(ifp->idev->dev), net))
1351 continue;
1352 if (ipv6_addr_equal(&ifp->addr, addr)) {
1353 if (dev == NULL || ifp->idev->dev == dev ||
1354 !(ifp->scope&(IFA_LINK|IFA_HOST) || strict)) {
1355 in6_ifa_hold(ifp);
1356 break;
1360 read_unlock_bh(&addrconf_hash_lock);
1362 return ifp;
1365 int ipv6_rcv_saddr_equal(const struct sock *sk, const struct sock *sk2)
1367 const struct in6_addr *sk_rcv_saddr6 = &inet6_sk(sk)->rcv_saddr;
1368 const struct in6_addr *sk2_rcv_saddr6 = inet6_rcv_saddr(sk2);
1369 __be32 sk_rcv_saddr = inet_sk(sk)->rcv_saddr;
1370 __be32 sk2_rcv_saddr = inet_rcv_saddr(sk2);
1371 int sk_ipv6only = ipv6_only_sock(sk);
1372 int sk2_ipv6only = inet_v6_ipv6only(sk2);
1373 int addr_type = ipv6_addr_type(sk_rcv_saddr6);
1374 int addr_type2 = sk2_rcv_saddr6 ? ipv6_addr_type(sk2_rcv_saddr6) : IPV6_ADDR_MAPPED;
1376 if (!sk2_rcv_saddr && !sk_ipv6only)
1377 return 1;
1379 if (addr_type2 == IPV6_ADDR_ANY &&
1380 !(sk2_ipv6only && addr_type == IPV6_ADDR_MAPPED))
1381 return 1;
1383 if (addr_type == IPV6_ADDR_ANY &&
1384 !(sk_ipv6only && addr_type2 == IPV6_ADDR_MAPPED))
1385 return 1;
1387 if (sk2_rcv_saddr6 &&
1388 ipv6_addr_equal(sk_rcv_saddr6, sk2_rcv_saddr6))
1389 return 1;
1391 if (addr_type == IPV6_ADDR_MAPPED &&
1392 !sk2_ipv6only &&
1393 (!sk2_rcv_saddr || !sk_rcv_saddr || sk_rcv_saddr == sk2_rcv_saddr))
1394 return 1;
1396 return 0;
1399 /* Gets referenced address, destroys ifaddr */
1401 static void addrconf_dad_stop(struct inet6_ifaddr *ifp)
1403 if (ifp->flags&IFA_F_PERMANENT) {
1404 spin_lock_bh(&ifp->lock);
1405 addrconf_del_timer(ifp);
1406 ifp->flags |= IFA_F_TENTATIVE;
1407 spin_unlock_bh(&ifp->lock);
1408 in6_ifa_put(ifp);
1409 #ifdef CONFIG_IPV6_PRIVACY
1410 } else if (ifp->flags&IFA_F_TEMPORARY) {
1411 struct inet6_ifaddr *ifpub;
1412 spin_lock_bh(&ifp->lock);
1413 ifpub = ifp->ifpub;
1414 if (ifpub) {
1415 in6_ifa_hold(ifpub);
1416 spin_unlock_bh(&ifp->lock);
1417 ipv6_create_tempaddr(ifpub, ifp);
1418 in6_ifa_put(ifpub);
1419 } else {
1420 spin_unlock_bh(&ifp->lock);
1422 ipv6_del_addr(ifp);
1423 #endif
1424 } else
1425 ipv6_del_addr(ifp);
1428 void addrconf_dad_failure(struct inet6_ifaddr *ifp)
1430 struct inet6_dev *idev = ifp->idev;
1431 if (idev->cnf.accept_dad > 1 && !idev->cnf.disable_ipv6) {
1432 struct in6_addr addr;
1434 addr.s6_addr32[0] = htonl(0xfe800000);
1435 addr.s6_addr32[1] = 0;
1437 if (!ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) &&
1438 ipv6_addr_equal(&ifp->addr, &addr)) {
1439 /* DAD failed for link-local based on MAC address */
1440 idev->cnf.disable_ipv6 = 1;
1444 if (net_ratelimit())
1445 printk(KERN_INFO "%s: duplicate address detected!\n", ifp->idev->dev->name);
1446 addrconf_dad_stop(ifp);
1449 /* Join to solicited addr multicast group. */
1451 void addrconf_join_solict(struct net_device *dev, struct in6_addr *addr)
1453 struct in6_addr maddr;
1455 if (dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1456 return;
1458 addrconf_addr_solict_mult(addr, &maddr);
1459 ipv6_dev_mc_inc(dev, &maddr);
1462 void addrconf_leave_solict(struct inet6_dev *idev, struct in6_addr *addr)
1464 struct in6_addr maddr;
1466 if (idev->dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1467 return;
1469 addrconf_addr_solict_mult(addr, &maddr);
1470 __ipv6_dev_mc_dec(idev, &maddr);
1473 static void addrconf_join_anycast(struct inet6_ifaddr *ifp)
1475 struct in6_addr addr;
1476 ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1477 if (ipv6_addr_any(&addr))
1478 return;
1479 ipv6_dev_ac_inc(ifp->idev->dev, &addr);
1482 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp)
1484 struct in6_addr addr;
1485 ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1486 if (ipv6_addr_any(&addr))
1487 return;
1488 __ipv6_dev_ac_dec(ifp->idev, &addr);
1491 static int addrconf_ifid_eui48(u8 *eui, struct net_device *dev)
1493 if (dev->addr_len != ETH_ALEN)
1494 return -1;
1495 memcpy(eui, dev->dev_addr, 3);
1496 memcpy(eui + 5, dev->dev_addr + 3, 3);
1499 * The zSeries OSA network cards can be shared among various
1500 * OS instances, but the OSA cards have only one MAC address.
1501 * This leads to duplicate address conflicts in conjunction
1502 * with IPv6 if more than one instance uses the same card.
1504 * The driver for these cards can deliver a unique 16-bit
1505 * identifier for each instance sharing the same card. It is
1506 * placed instead of 0xFFFE in the interface identifier. The
1507 * "u" bit of the interface identifier is not inverted in this
1508 * case. Hence the resulting interface identifier has local
1509 * scope according to RFC2373.
1511 if (dev->dev_id) {
1512 eui[3] = (dev->dev_id >> 8) & 0xFF;
1513 eui[4] = dev->dev_id & 0xFF;
1514 } else {
1515 eui[3] = 0xFF;
1516 eui[4] = 0xFE;
1517 eui[0] ^= 2;
1519 return 0;
1522 static int addrconf_ifid_arcnet(u8 *eui, struct net_device *dev)
1524 /* XXX: inherit EUI-64 from other interface -- yoshfuji */
1525 if (dev->addr_len != ARCNET_ALEN)
1526 return -1;
1527 memset(eui, 0, 7);
1528 eui[7] = *(u8*)dev->dev_addr;
1529 return 0;
1532 static int addrconf_ifid_infiniband(u8 *eui, struct net_device *dev)
1534 if (dev->addr_len != INFINIBAND_ALEN)
1535 return -1;
1536 memcpy(eui, dev->dev_addr + 12, 8);
1537 eui[0] |= 2;
1538 return 0;
1541 int __ipv6_isatap_ifid(u8 *eui, __be32 addr)
1543 eui[0] = (ipv4_is_zeronet(addr) || ipv4_is_private_10(addr) ||
1544 ipv4_is_loopback(addr) || ipv4_is_linklocal_169(addr) ||
1545 ipv4_is_private_172(addr) || ipv4_is_test_192(addr) ||
1546 ipv4_is_anycast_6to4(addr) || ipv4_is_private_192(addr) ||
1547 ipv4_is_test_198(addr) || ipv4_is_multicast(addr) ||
1548 ipv4_is_lbcast(addr)) ? 0x00 : 0x02;
1549 eui[1] = 0;
1550 eui[2] = 0x5E;
1551 eui[3] = 0xFE;
1552 memcpy(eui + 4, &addr, 4);
1553 return 0;
1555 EXPORT_SYMBOL(__ipv6_isatap_ifid);
1557 static int addrconf_ifid_sit(u8 *eui, struct net_device *dev)
1559 if (dev->priv_flags & IFF_ISATAP)
1560 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
1561 return -1;
1564 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev)
1566 switch (dev->type) {
1567 case ARPHRD_ETHER:
1568 case ARPHRD_FDDI:
1569 case ARPHRD_IEEE802_TR:
1570 return addrconf_ifid_eui48(eui, dev);
1571 case ARPHRD_ARCNET:
1572 return addrconf_ifid_arcnet(eui, dev);
1573 case ARPHRD_INFINIBAND:
1574 return addrconf_ifid_infiniband(eui, dev);
1575 case ARPHRD_SIT:
1576 return addrconf_ifid_sit(eui, dev);
1578 return -1;
1581 static int ipv6_inherit_eui64(u8 *eui, struct inet6_dev *idev)
1583 int err = -1;
1584 struct inet6_ifaddr *ifp;
1586 read_lock_bh(&idev->lock);
1587 for (ifp=idev->addr_list; ifp; ifp=ifp->if_next) {
1588 if (ifp->scope == IFA_LINK && !(ifp->flags&IFA_F_TENTATIVE)) {
1589 memcpy(eui, ifp->addr.s6_addr+8, 8);
1590 err = 0;
1591 break;
1594 read_unlock_bh(&idev->lock);
1595 return err;
1598 #ifdef CONFIG_IPV6_PRIVACY
1599 /* (re)generation of randomized interface identifier (RFC 3041 3.2, 3.5) */
1600 static int __ipv6_regen_rndid(struct inet6_dev *idev)
1602 regen:
1603 get_random_bytes(idev->rndid, sizeof(idev->rndid));
1604 idev->rndid[0] &= ~0x02;
1607 * <draft-ietf-ipngwg-temp-addresses-v2-00.txt>:
1608 * check if generated address is not inappropriate
1610 * - Reserved subnet anycast (RFC 2526)
1611 * 11111101 11....11 1xxxxxxx
1612 * - ISATAP (RFC4214) 6.1
1613 * 00-00-5E-FE-xx-xx-xx-xx
1614 * - value 0
1615 * - XXX: already assigned to an address on the device
1617 if (idev->rndid[0] == 0xfd &&
1618 (idev->rndid[1]&idev->rndid[2]&idev->rndid[3]&idev->rndid[4]&idev->rndid[5]&idev->rndid[6]) == 0xff &&
1619 (idev->rndid[7]&0x80))
1620 goto regen;
1621 if ((idev->rndid[0]|idev->rndid[1]) == 0) {
1622 if (idev->rndid[2] == 0x5e && idev->rndid[3] == 0xfe)
1623 goto regen;
1624 if ((idev->rndid[2]|idev->rndid[3]|idev->rndid[4]|idev->rndid[5]|idev->rndid[6]|idev->rndid[7]) == 0x00)
1625 goto regen;
1628 return 0;
1631 static void ipv6_regen_rndid(unsigned long data)
1633 struct inet6_dev *idev = (struct inet6_dev *) data;
1634 unsigned long expires;
1636 rcu_read_lock_bh();
1637 write_lock_bh(&idev->lock);
1639 if (idev->dead)
1640 goto out;
1642 if (__ipv6_regen_rndid(idev) < 0)
1643 goto out;
1645 expires = jiffies +
1646 idev->cnf.temp_prefered_lft * HZ -
1647 idev->cnf.regen_max_retry * idev->cnf.dad_transmits * idev->nd_parms->retrans_time - desync_factor;
1648 if (time_before(expires, jiffies)) {
1649 printk(KERN_WARNING
1650 "ipv6_regen_rndid(): too short regeneration interval; timer disabled for %s.\n",
1651 idev->dev->name);
1652 goto out;
1655 if (!mod_timer(&idev->regen_timer, expires))
1656 in6_dev_hold(idev);
1658 out:
1659 write_unlock_bh(&idev->lock);
1660 rcu_read_unlock_bh();
1661 in6_dev_put(idev);
1664 static int __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr) {
1665 int ret = 0;
1667 if (tmpaddr && memcmp(idev->rndid, &tmpaddr->s6_addr[8], 8) == 0)
1668 ret = __ipv6_regen_rndid(idev);
1669 return ret;
1671 #endif
1674 * Add prefix route.
1677 static void
1678 addrconf_prefix_route(struct in6_addr *pfx, int plen, struct net_device *dev,
1679 unsigned long expires, u32 flags)
1681 struct fib6_config cfg = {
1682 .fc_table = RT6_TABLE_PREFIX,
1683 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1684 .fc_ifindex = dev->ifindex,
1685 .fc_expires = expires,
1686 .fc_dst_len = plen,
1687 .fc_flags = RTF_UP | flags,
1688 .fc_nlinfo.nl_net = dev_net(dev),
1691 ipv6_addr_copy(&cfg.fc_dst, pfx);
1693 /* Prevent useless cloning on PtP SIT.
1694 This thing is done here expecting that the whole
1695 class of non-broadcast devices need not cloning.
1697 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
1698 if (dev->type == ARPHRD_SIT && (dev->flags & IFF_POINTOPOINT))
1699 cfg.fc_flags |= RTF_NONEXTHOP;
1700 #endif
1702 ip6_route_add(&cfg);
1705 /* Create "default" multicast route to the interface */
1707 static void addrconf_add_mroute(struct net_device *dev)
1709 struct fib6_config cfg = {
1710 .fc_table = RT6_TABLE_LOCAL,
1711 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1712 .fc_ifindex = dev->ifindex,
1713 .fc_dst_len = 8,
1714 .fc_flags = RTF_UP,
1715 .fc_nlinfo.nl_net = dev_net(dev),
1718 ipv6_addr_set(&cfg.fc_dst, htonl(0xFF000000), 0, 0, 0);
1720 ip6_route_add(&cfg);
1723 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
1724 static void sit_route_add(struct net_device *dev)
1726 struct fib6_config cfg = {
1727 .fc_table = RT6_TABLE_MAIN,
1728 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1729 .fc_ifindex = dev->ifindex,
1730 .fc_dst_len = 96,
1731 .fc_flags = RTF_UP | RTF_NONEXTHOP,
1732 .fc_nlinfo.nl_net = dev_net(dev),
1735 /* prefix length - 96 bits "::d.d.d.d" */
1736 ip6_route_add(&cfg);
1738 #endif
1740 static void addrconf_add_lroute(struct net_device *dev)
1742 struct in6_addr addr;
1744 ipv6_addr_set(&addr, htonl(0xFE800000), 0, 0, 0);
1745 addrconf_prefix_route(&addr, 64, dev, 0, 0);
1748 static struct inet6_dev *addrconf_add_dev(struct net_device *dev)
1750 struct inet6_dev *idev;
1752 ASSERT_RTNL();
1754 if ((idev = ipv6_find_idev(dev)) == NULL)
1755 return NULL;
1757 /* Add default multicast route */
1758 addrconf_add_mroute(dev);
1760 /* Add link local route */
1761 addrconf_add_lroute(dev);
1762 return idev;
1765 void addrconf_prefix_rcv(struct net_device *dev, u8 *opt, int len)
1767 struct prefix_info *pinfo;
1768 __u32 valid_lft;
1769 __u32 prefered_lft;
1770 int addr_type;
1771 struct inet6_dev *in6_dev;
1773 pinfo = (struct prefix_info *) opt;
1775 if (len < sizeof(struct prefix_info)) {
1776 ADBG(("addrconf: prefix option too short\n"));
1777 return;
1781 * Validation checks ([ADDRCONF], page 19)
1784 addr_type = ipv6_addr_type(&pinfo->prefix);
1786 if (addr_type & (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL))
1787 return;
1789 valid_lft = ntohl(pinfo->valid);
1790 prefered_lft = ntohl(pinfo->prefered);
1792 if (prefered_lft > valid_lft) {
1793 if (net_ratelimit())
1794 printk(KERN_WARNING "addrconf: prefix option has invalid lifetime\n");
1795 return;
1798 in6_dev = in6_dev_get(dev);
1800 if (in6_dev == NULL) {
1801 if (net_ratelimit())
1802 printk(KERN_DEBUG "addrconf: device %s not configured\n", dev->name);
1803 return;
1807 * Two things going on here:
1808 * 1) Add routes for on-link prefixes
1809 * 2) Configure prefixes with the auto flag set
1812 if (pinfo->onlink) {
1813 struct rt6_info *rt;
1814 unsigned long rt_expires;
1816 /* Avoid arithmetic overflow. Really, we could
1817 * save rt_expires in seconds, likely valid_lft,
1818 * but it would require division in fib gc, that it
1819 * not good.
1821 if (HZ > USER_HZ)
1822 rt_expires = addrconf_timeout_fixup(valid_lft, HZ);
1823 else
1824 rt_expires = addrconf_timeout_fixup(valid_lft, USER_HZ);
1826 if (addrconf_finite_timeout(rt_expires))
1827 rt_expires *= HZ;
1829 rt = rt6_lookup(dev_net(dev), &pinfo->prefix, NULL,
1830 dev->ifindex, 1);
1832 if (rt && addrconf_is_prefix_route(rt)) {
1833 /* Autoconf prefix route */
1834 if (valid_lft == 0) {
1835 ip6_del_rt(rt);
1836 rt = NULL;
1837 } else if (addrconf_finite_timeout(rt_expires)) {
1838 /* not infinity */
1839 rt->rt6i_expires = jiffies + rt_expires;
1840 rt->rt6i_flags |= RTF_EXPIRES;
1841 } else {
1842 rt->rt6i_flags &= ~RTF_EXPIRES;
1843 rt->rt6i_expires = 0;
1845 } else if (valid_lft) {
1846 clock_t expires = 0;
1847 int flags = RTF_ADDRCONF | RTF_PREFIX_RT;
1848 if (addrconf_finite_timeout(rt_expires)) {
1849 /* not infinity */
1850 flags |= RTF_EXPIRES;
1851 expires = jiffies_to_clock_t(rt_expires);
1853 addrconf_prefix_route(&pinfo->prefix, pinfo->prefix_len,
1854 dev, expires, flags);
1856 if (rt)
1857 dst_release(&rt->u.dst);
1860 /* Try to figure out our local address for this prefix */
1862 if (pinfo->autoconf && in6_dev->cnf.autoconf) {
1863 struct inet6_ifaddr * ifp;
1864 struct in6_addr addr;
1865 int create = 0, update_lft = 0;
1866 struct net *net = dev_net(dev);
1868 if (pinfo->prefix_len == 64) {
1869 memcpy(&addr, &pinfo->prefix, 8);
1870 if (ipv6_generate_eui64(addr.s6_addr + 8, dev) &&
1871 ipv6_inherit_eui64(addr.s6_addr + 8, in6_dev)) {
1872 in6_dev_put(in6_dev);
1873 return;
1875 goto ok;
1877 if (net_ratelimit())
1878 printk(KERN_DEBUG "IPv6 addrconf: prefix with wrong length %d\n",
1879 pinfo->prefix_len);
1880 in6_dev_put(in6_dev);
1881 return;
1885 ifp = ipv6_get_ifaddr(net, &addr, dev, 1);
1887 if (ifp == NULL && valid_lft) {
1888 int max_addresses = in6_dev->cnf.max_addresses;
1889 u32 addr_flags = 0;
1891 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1892 if (in6_dev->cnf.optimistic_dad &&
1893 !net->ipv6.devconf_all->forwarding)
1894 addr_flags = IFA_F_OPTIMISTIC;
1895 #endif
1897 /* Do not allow to create too much of autoconfigured
1898 * addresses; this would be too easy way to crash kernel.
1900 if (!max_addresses ||
1901 ipv6_count_addresses(in6_dev) < max_addresses)
1902 ifp = ipv6_add_addr(in6_dev, &addr, pinfo->prefix_len,
1903 addr_type&IPV6_ADDR_SCOPE_MASK,
1904 addr_flags);
1906 if (!ifp || IS_ERR(ifp)) {
1907 in6_dev_put(in6_dev);
1908 return;
1911 update_lft = create = 1;
1912 ifp->cstamp = jiffies;
1913 addrconf_dad_start(ifp, RTF_ADDRCONF|RTF_PREFIX_RT);
1916 if (ifp) {
1917 int flags;
1918 unsigned long now;
1919 #ifdef CONFIG_IPV6_PRIVACY
1920 struct inet6_ifaddr *ift;
1921 #endif
1922 u32 stored_lft;
1924 /* update lifetime (RFC2462 5.5.3 e) */
1925 spin_lock(&ifp->lock);
1926 now = jiffies;
1927 if (ifp->valid_lft > (now - ifp->tstamp) / HZ)
1928 stored_lft = ifp->valid_lft - (now - ifp->tstamp) / HZ;
1929 else
1930 stored_lft = 0;
1931 if (!update_lft && stored_lft) {
1932 if (valid_lft > MIN_VALID_LIFETIME ||
1933 valid_lft > stored_lft)
1934 update_lft = 1;
1935 else if (stored_lft <= MIN_VALID_LIFETIME) {
1936 /* valid_lft <= stored_lft is always true */
1937 /* XXX: IPsec */
1938 update_lft = 0;
1939 } else {
1940 valid_lft = MIN_VALID_LIFETIME;
1941 if (valid_lft < prefered_lft)
1942 prefered_lft = valid_lft;
1943 update_lft = 1;
1947 if (update_lft) {
1948 ifp->valid_lft = valid_lft;
1949 ifp->prefered_lft = prefered_lft;
1950 ifp->tstamp = now;
1951 flags = ifp->flags;
1952 ifp->flags &= ~IFA_F_DEPRECATED;
1953 spin_unlock(&ifp->lock);
1955 if (!(flags&IFA_F_TENTATIVE))
1956 ipv6_ifa_notify(0, ifp);
1957 } else
1958 spin_unlock(&ifp->lock);
1960 #ifdef CONFIG_IPV6_PRIVACY
1961 read_lock_bh(&in6_dev->lock);
1962 /* update all temporary addresses in the list */
1963 for (ift=in6_dev->tempaddr_list; ift; ift=ift->tmp_next) {
1965 * When adjusting the lifetimes of an existing
1966 * temporary address, only lower the lifetimes.
1967 * Implementations must not increase the
1968 * lifetimes of an existing temporary address
1969 * when processing a Prefix Information Option.
1971 if (ifp != ift->ifpub)
1972 continue;
1974 spin_lock(&ift->lock);
1975 flags = ift->flags;
1976 if (ift->valid_lft > valid_lft &&
1977 ift->valid_lft - valid_lft > (jiffies - ift->tstamp) / HZ)
1978 ift->valid_lft = valid_lft + (jiffies - ift->tstamp) / HZ;
1979 if (ift->prefered_lft > prefered_lft &&
1980 ift->prefered_lft - prefered_lft > (jiffies - ift->tstamp) / HZ)
1981 ift->prefered_lft = prefered_lft + (jiffies - ift->tstamp) / HZ;
1982 spin_unlock(&ift->lock);
1983 if (!(flags&IFA_F_TENTATIVE))
1984 ipv6_ifa_notify(0, ift);
1987 if (create && in6_dev->cnf.use_tempaddr > 0) {
1989 * When a new public address is created as described in [ADDRCONF],
1990 * also create a new temporary address.
1992 read_unlock_bh(&in6_dev->lock);
1993 ipv6_create_tempaddr(ifp, NULL);
1994 } else {
1995 read_unlock_bh(&in6_dev->lock);
1997 #endif
1998 in6_ifa_put(ifp);
1999 addrconf_verify(0);
2002 inet6_prefix_notify(RTM_NEWPREFIX, in6_dev, pinfo);
2003 in6_dev_put(in6_dev);
2007 * Set destination address.
2008 * Special case for SIT interfaces where we create a new "virtual"
2009 * device.
2011 int addrconf_set_dstaddr(struct net *net, void __user *arg)
2013 struct in6_ifreq ireq;
2014 struct net_device *dev;
2015 int err = -EINVAL;
2017 rtnl_lock();
2019 err = -EFAULT;
2020 if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2021 goto err_exit;
2023 dev = __dev_get_by_index(net, ireq.ifr6_ifindex);
2025 err = -ENODEV;
2026 if (dev == NULL)
2027 goto err_exit;
2029 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2030 if (dev->type == ARPHRD_SIT) {
2031 struct ifreq ifr;
2032 mm_segment_t oldfs;
2033 struct ip_tunnel_parm p;
2035 err = -EADDRNOTAVAIL;
2036 if (!(ipv6_addr_type(&ireq.ifr6_addr) & IPV6_ADDR_COMPATv4))
2037 goto err_exit;
2039 memset(&p, 0, sizeof(p));
2040 p.iph.daddr = ireq.ifr6_addr.s6_addr32[3];
2041 p.iph.saddr = 0;
2042 p.iph.version = 4;
2043 p.iph.ihl = 5;
2044 p.iph.protocol = IPPROTO_IPV6;
2045 p.iph.ttl = 64;
2046 ifr.ifr_ifru.ifru_data = (__force void __user *)&p;
2048 oldfs = get_fs(); set_fs(KERNEL_DS);
2049 err = dev->do_ioctl(dev, &ifr, SIOCADDTUNNEL);
2050 set_fs(oldfs);
2052 if (err == 0) {
2053 err = -ENOBUFS;
2054 dev = __dev_get_by_name(net, p.name);
2055 if (!dev)
2056 goto err_exit;
2057 err = dev_open(dev);
2060 #endif
2062 err_exit:
2063 rtnl_unlock();
2064 return err;
2068 * Manual configuration of address on an interface
2070 static int inet6_addr_add(struct net *net, int ifindex, struct in6_addr *pfx,
2071 unsigned int plen, __u8 ifa_flags, __u32 prefered_lft,
2072 __u32 valid_lft)
2074 struct inet6_ifaddr *ifp;
2075 struct inet6_dev *idev;
2076 struct net_device *dev;
2077 int scope;
2078 u32 flags;
2079 clock_t expires;
2080 unsigned long timeout;
2082 ASSERT_RTNL();
2084 if (plen > 128)
2085 return -EINVAL;
2087 /* check the lifetime */
2088 if (!valid_lft || prefered_lft > valid_lft)
2089 return -EINVAL;
2091 dev = __dev_get_by_index(net, ifindex);
2092 if (!dev)
2093 return -ENODEV;
2095 if ((idev = addrconf_add_dev(dev)) == NULL)
2096 return -ENOBUFS;
2098 scope = ipv6_addr_scope(pfx);
2100 timeout = addrconf_timeout_fixup(valid_lft, HZ);
2101 if (addrconf_finite_timeout(timeout)) {
2102 expires = jiffies_to_clock_t(timeout * HZ);
2103 valid_lft = timeout;
2104 flags = RTF_EXPIRES;
2105 } else {
2106 expires = 0;
2107 flags = 0;
2108 ifa_flags |= IFA_F_PERMANENT;
2111 timeout = addrconf_timeout_fixup(prefered_lft, HZ);
2112 if (addrconf_finite_timeout(timeout)) {
2113 if (timeout == 0)
2114 ifa_flags |= IFA_F_DEPRECATED;
2115 prefered_lft = timeout;
2118 ifp = ipv6_add_addr(idev, pfx, plen, scope, ifa_flags);
2120 if (!IS_ERR(ifp)) {
2121 spin_lock_bh(&ifp->lock);
2122 ifp->valid_lft = valid_lft;
2123 ifp->prefered_lft = prefered_lft;
2124 ifp->tstamp = jiffies;
2125 spin_unlock_bh(&ifp->lock);
2127 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, dev,
2128 expires, flags);
2130 * Note that section 3.1 of RFC 4429 indicates
2131 * that the Optimistic flag should not be set for
2132 * manually configured addresses
2134 addrconf_dad_start(ifp, 0);
2135 in6_ifa_put(ifp);
2136 addrconf_verify(0);
2137 return 0;
2140 return PTR_ERR(ifp);
2143 static int inet6_addr_del(struct net *net, int ifindex, struct in6_addr *pfx,
2144 unsigned int plen)
2146 struct inet6_ifaddr *ifp;
2147 struct inet6_dev *idev;
2148 struct net_device *dev;
2150 if (plen > 128)
2151 return -EINVAL;
2153 dev = __dev_get_by_index(net, ifindex);
2154 if (!dev)
2155 return -ENODEV;
2157 if ((idev = __in6_dev_get(dev)) == NULL)
2158 return -ENXIO;
2160 read_lock_bh(&idev->lock);
2161 for (ifp = idev->addr_list; ifp; ifp=ifp->if_next) {
2162 if (ifp->prefix_len == plen &&
2163 ipv6_addr_equal(pfx, &ifp->addr)) {
2164 in6_ifa_hold(ifp);
2165 read_unlock_bh(&idev->lock);
2167 ipv6_del_addr(ifp);
2169 /* If the last address is deleted administratively,
2170 disable IPv6 on this interface.
2172 if (idev->addr_list == NULL)
2173 addrconf_ifdown(idev->dev, 1);
2174 return 0;
2177 read_unlock_bh(&idev->lock);
2178 return -EADDRNOTAVAIL;
2182 int addrconf_add_ifaddr(struct net *net, void __user *arg)
2184 struct in6_ifreq ireq;
2185 int err;
2187 if (!capable(CAP_NET_ADMIN))
2188 return -EPERM;
2190 if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2191 return -EFAULT;
2193 rtnl_lock();
2194 err = inet6_addr_add(net, ireq.ifr6_ifindex, &ireq.ifr6_addr,
2195 ireq.ifr6_prefixlen, IFA_F_PERMANENT,
2196 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
2197 rtnl_unlock();
2198 return err;
2201 int addrconf_del_ifaddr(struct net *net, void __user *arg)
2203 struct in6_ifreq ireq;
2204 int err;
2206 if (!capable(CAP_NET_ADMIN))
2207 return -EPERM;
2209 if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2210 return -EFAULT;
2212 rtnl_lock();
2213 err = inet6_addr_del(net, ireq.ifr6_ifindex, &ireq.ifr6_addr,
2214 ireq.ifr6_prefixlen);
2215 rtnl_unlock();
2216 return err;
2219 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2220 static void sit_add_v4_addrs(struct inet6_dev *idev)
2222 struct inet6_ifaddr * ifp;
2223 struct in6_addr addr;
2224 struct net_device *dev;
2225 struct net *net = dev_net(idev->dev);
2226 int scope;
2228 ASSERT_RTNL();
2230 memset(&addr, 0, sizeof(struct in6_addr));
2231 memcpy(&addr.s6_addr32[3], idev->dev->dev_addr, 4);
2233 if (idev->dev->flags&IFF_POINTOPOINT) {
2234 addr.s6_addr32[0] = htonl(0xfe800000);
2235 scope = IFA_LINK;
2236 } else {
2237 scope = IPV6_ADDR_COMPATv4;
2240 if (addr.s6_addr32[3]) {
2241 ifp = ipv6_add_addr(idev, &addr, 128, scope, IFA_F_PERMANENT);
2242 if (!IS_ERR(ifp)) {
2243 spin_lock_bh(&ifp->lock);
2244 ifp->flags &= ~IFA_F_TENTATIVE;
2245 spin_unlock_bh(&ifp->lock);
2246 ipv6_ifa_notify(RTM_NEWADDR, ifp);
2247 in6_ifa_put(ifp);
2249 return;
2252 for_each_netdev(net, dev) {
2253 struct in_device * in_dev = __in_dev_get_rtnl(dev);
2254 if (in_dev && (dev->flags & IFF_UP)) {
2255 struct in_ifaddr * ifa;
2257 int flag = scope;
2259 for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) {
2260 int plen;
2262 addr.s6_addr32[3] = ifa->ifa_local;
2264 if (ifa->ifa_scope == RT_SCOPE_LINK)
2265 continue;
2266 if (ifa->ifa_scope >= RT_SCOPE_HOST) {
2267 if (idev->dev->flags&IFF_POINTOPOINT)
2268 continue;
2269 flag |= IFA_HOST;
2271 if (idev->dev->flags&IFF_POINTOPOINT)
2272 plen = 64;
2273 else
2274 plen = 96;
2276 ifp = ipv6_add_addr(idev, &addr, plen, flag,
2277 IFA_F_PERMANENT);
2278 if (!IS_ERR(ifp)) {
2279 spin_lock_bh(&ifp->lock);
2280 ifp->flags &= ~IFA_F_TENTATIVE;
2281 spin_unlock_bh(&ifp->lock);
2282 ipv6_ifa_notify(RTM_NEWADDR, ifp);
2283 in6_ifa_put(ifp);
2289 #endif
2291 static void init_loopback(struct net_device *dev)
2293 struct inet6_dev *idev;
2294 struct inet6_ifaddr * ifp;
2296 /* ::1 */
2298 ASSERT_RTNL();
2300 if ((idev = ipv6_find_idev(dev)) == NULL) {
2301 printk(KERN_DEBUG "init loopback: add_dev failed\n");
2302 return;
2305 ifp = ipv6_add_addr(idev, &in6addr_loopback, 128, IFA_HOST, IFA_F_PERMANENT);
2306 if (!IS_ERR(ifp)) {
2307 spin_lock_bh(&ifp->lock);
2308 ifp->flags &= ~IFA_F_TENTATIVE;
2309 spin_unlock_bh(&ifp->lock);
2310 ipv6_ifa_notify(RTM_NEWADDR, ifp);
2311 in6_ifa_put(ifp);
2315 static void addrconf_add_linklocal(struct inet6_dev *idev, struct in6_addr *addr)
2317 struct inet6_ifaddr * ifp;
2318 u32 addr_flags = IFA_F_PERMANENT;
2320 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2321 if (idev->cnf.optimistic_dad &&
2322 !dev_net(idev->dev)->ipv6.devconf_all->forwarding)
2323 addr_flags |= IFA_F_OPTIMISTIC;
2324 #endif
2327 ifp = ipv6_add_addr(idev, addr, 64, IFA_LINK, addr_flags);
2328 if (!IS_ERR(ifp)) {
2329 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, idev->dev, 0, 0);
2330 addrconf_dad_start(ifp, 0);
2331 in6_ifa_put(ifp);
2335 static void addrconf_dev_config(struct net_device *dev)
2337 struct in6_addr addr;
2338 struct inet6_dev * idev;
2340 ASSERT_RTNL();
2342 if ((dev->type != ARPHRD_ETHER) &&
2343 (dev->type != ARPHRD_FDDI) &&
2344 (dev->type != ARPHRD_IEEE802_TR) &&
2345 (dev->type != ARPHRD_ARCNET) &&
2346 (dev->type != ARPHRD_INFINIBAND)) {
2347 /* Alas, we support only Ethernet autoconfiguration. */
2348 return;
2351 idev = addrconf_add_dev(dev);
2352 if (idev == NULL)
2353 return;
2355 memset(&addr, 0, sizeof(struct in6_addr));
2356 addr.s6_addr32[0] = htonl(0xFE800000);
2358 if (ipv6_generate_eui64(addr.s6_addr + 8, dev) == 0)
2359 addrconf_add_linklocal(idev, &addr);
2362 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2363 static void addrconf_sit_config(struct net_device *dev)
2365 struct inet6_dev *idev;
2367 ASSERT_RTNL();
2370 * Configure the tunnel with one of our IPv4
2371 * addresses... we should configure all of
2372 * our v4 addrs in the tunnel
2375 if ((idev = ipv6_find_idev(dev)) == NULL) {
2376 printk(KERN_DEBUG "init sit: add_dev failed\n");
2377 return;
2380 if (dev->priv_flags & IFF_ISATAP) {
2381 struct in6_addr addr;
2383 ipv6_addr_set(&addr, htonl(0xFE800000), 0, 0, 0);
2384 addrconf_prefix_route(&addr, 64, dev, 0, 0);
2385 if (!ipv6_generate_eui64(addr.s6_addr + 8, dev))
2386 addrconf_add_linklocal(idev, &addr);
2387 return;
2390 sit_add_v4_addrs(idev);
2392 if (dev->flags&IFF_POINTOPOINT) {
2393 addrconf_add_mroute(dev);
2394 addrconf_add_lroute(dev);
2395 } else
2396 sit_route_add(dev);
2398 #endif
2400 static inline int
2401 ipv6_inherit_linklocal(struct inet6_dev *idev, struct net_device *link_dev)
2403 struct in6_addr lladdr;
2405 if (!ipv6_get_lladdr(link_dev, &lladdr, IFA_F_TENTATIVE)) {
2406 addrconf_add_linklocal(idev, &lladdr);
2407 return 0;
2409 return -1;
2412 static void ip6_tnl_add_linklocal(struct inet6_dev *idev)
2414 struct net_device *link_dev;
2415 struct net *net = dev_net(idev->dev);
2417 /* first try to inherit the link-local address from the link device */
2418 if (idev->dev->iflink &&
2419 (link_dev = __dev_get_by_index(net, idev->dev->iflink))) {
2420 if (!ipv6_inherit_linklocal(idev, link_dev))
2421 return;
2423 /* then try to inherit it from any device */
2424 for_each_netdev(net, link_dev) {
2425 if (!ipv6_inherit_linklocal(idev, link_dev))
2426 return;
2428 printk(KERN_DEBUG "init ip6-ip6: add_linklocal failed\n");
2432 * Autoconfigure tunnel with a link-local address so routing protocols,
2433 * DHCPv6, MLD etc. can be run over the virtual link
2436 static void addrconf_ip6_tnl_config(struct net_device *dev)
2438 struct inet6_dev *idev;
2440 ASSERT_RTNL();
2442 if ((idev = addrconf_add_dev(dev)) == NULL) {
2443 printk(KERN_DEBUG "init ip6-ip6: add_dev failed\n");
2444 return;
2446 ip6_tnl_add_linklocal(idev);
2449 static int addrconf_notify(struct notifier_block *this, unsigned long event,
2450 void * data)
2452 struct net_device *dev = (struct net_device *) data;
2453 struct inet6_dev *idev = __in6_dev_get(dev);
2454 int run_pending = 0;
2455 int err;
2457 switch(event) {
2458 case NETDEV_REGISTER:
2459 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2460 idev = ipv6_add_dev(dev);
2461 if (!idev)
2462 return notifier_from_errno(-ENOMEM);
2464 break;
2465 case NETDEV_UP:
2466 case NETDEV_CHANGE:
2467 if (dev->flags & IFF_SLAVE)
2468 break;
2470 if (event == NETDEV_UP) {
2471 if (!addrconf_qdisc_ok(dev)) {
2472 /* device is not ready yet. */
2473 printk(KERN_INFO
2474 "ADDRCONF(NETDEV_UP): %s: "
2475 "link is not ready\n",
2476 dev->name);
2477 break;
2480 if (!idev && dev->mtu >= IPV6_MIN_MTU)
2481 idev = ipv6_add_dev(dev);
2483 if (idev)
2484 idev->if_flags |= IF_READY;
2485 } else {
2486 if (!addrconf_qdisc_ok(dev)) {
2487 /* device is still not ready. */
2488 break;
2491 if (idev) {
2492 if (idev->if_flags & IF_READY) {
2493 /* device is already configured. */
2494 break;
2496 idev->if_flags |= IF_READY;
2499 printk(KERN_INFO
2500 "ADDRCONF(NETDEV_CHANGE): %s: "
2501 "link becomes ready\n",
2502 dev->name);
2504 run_pending = 1;
2507 switch(dev->type) {
2508 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2509 case ARPHRD_SIT:
2510 addrconf_sit_config(dev);
2511 break;
2512 #endif
2513 case ARPHRD_TUNNEL6:
2514 addrconf_ip6_tnl_config(dev);
2515 break;
2516 case ARPHRD_LOOPBACK:
2517 init_loopback(dev);
2518 break;
2520 default:
2521 addrconf_dev_config(dev);
2522 break;
2524 if (idev) {
2525 if (run_pending)
2526 addrconf_dad_run(idev);
2528 /* If the MTU changed during the interface down, when the
2529 interface up, the changed MTU must be reflected in the
2530 idev as well as routers.
2532 if (idev->cnf.mtu6 != dev->mtu && dev->mtu >= IPV6_MIN_MTU) {
2533 rt6_mtu_change(dev, dev->mtu);
2534 idev->cnf.mtu6 = dev->mtu;
2536 idev->tstamp = jiffies;
2537 inet6_ifinfo_notify(RTM_NEWLINK, idev);
2538 /* If the changed mtu during down is lower than IPV6_MIN_MTU
2539 stop IPv6 on this interface.
2541 if (dev->mtu < IPV6_MIN_MTU)
2542 addrconf_ifdown(dev, event != NETDEV_DOWN);
2544 break;
2546 case NETDEV_CHANGEMTU:
2547 if (idev && dev->mtu >= IPV6_MIN_MTU) {
2548 rt6_mtu_change(dev, dev->mtu);
2549 idev->cnf.mtu6 = dev->mtu;
2550 break;
2553 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2554 idev = ipv6_add_dev(dev);
2555 if (idev)
2556 break;
2559 /* MTU falled under IPV6_MIN_MTU. Stop IPv6 on this interface. */
2561 case NETDEV_DOWN:
2562 case NETDEV_UNREGISTER:
2564 * Remove all addresses from this interface.
2566 addrconf_ifdown(dev, event != NETDEV_DOWN);
2567 break;
2569 case NETDEV_CHANGENAME:
2570 if (idev) {
2571 snmp6_unregister_dev(idev);
2572 addrconf_sysctl_unregister(idev);
2573 addrconf_sysctl_register(idev);
2574 err = snmp6_register_dev(idev);
2575 if (err)
2576 return notifier_from_errno(err);
2578 break;
2581 return NOTIFY_OK;
2585 * addrconf module should be notified of a device going up
2587 static struct notifier_block ipv6_dev_notf = {
2588 .notifier_call = addrconf_notify,
2589 .priority = 0
2592 static int addrconf_ifdown(struct net_device *dev, int how)
2594 struct inet6_dev *idev;
2595 struct inet6_ifaddr *ifa, **bifa;
2596 struct net *net = dev_net(dev);
2597 int i;
2599 ASSERT_RTNL();
2601 if ((dev->flags & IFF_LOOPBACK) && how == 1)
2602 how = 0;
2604 rt6_ifdown(net, dev);
2605 neigh_ifdown(&nd_tbl, dev);
2607 idev = __in6_dev_get(dev);
2608 if (idev == NULL)
2609 return -ENODEV;
2611 /* Step 1: remove reference to ipv6 device from parent device.
2612 Do not dev_put!
2614 if (how) {
2615 idev->dead = 1;
2617 /* protected by rtnl_lock */
2618 rcu_assign_pointer(dev->ip6_ptr, NULL);
2620 /* Step 1.5: remove snmp6 entry */
2621 snmp6_unregister_dev(idev);
2625 /* Step 2: clear hash table */
2626 for (i=0; i<IN6_ADDR_HSIZE; i++) {
2627 bifa = &inet6_addr_lst[i];
2629 write_lock_bh(&addrconf_hash_lock);
2630 while ((ifa = *bifa) != NULL) {
2631 if (ifa->idev == idev) {
2632 *bifa = ifa->lst_next;
2633 ifa->lst_next = NULL;
2634 addrconf_del_timer(ifa);
2635 in6_ifa_put(ifa);
2636 continue;
2638 bifa = &ifa->lst_next;
2640 write_unlock_bh(&addrconf_hash_lock);
2643 write_lock_bh(&idev->lock);
2645 /* Step 3: clear flags for stateless addrconf */
2646 if (!how)
2647 idev->if_flags &= ~(IF_RS_SENT|IF_RA_RCVD|IF_READY);
2649 /* Step 4: clear address list */
2650 #ifdef CONFIG_IPV6_PRIVACY
2651 if (how && del_timer(&idev->regen_timer))
2652 in6_dev_put(idev);
2654 /* clear tempaddr list */
2655 while ((ifa = idev->tempaddr_list) != NULL) {
2656 idev->tempaddr_list = ifa->tmp_next;
2657 ifa->tmp_next = NULL;
2658 ifa->dead = 1;
2659 write_unlock_bh(&idev->lock);
2660 spin_lock_bh(&ifa->lock);
2662 if (ifa->ifpub) {
2663 in6_ifa_put(ifa->ifpub);
2664 ifa->ifpub = NULL;
2666 spin_unlock_bh(&ifa->lock);
2667 in6_ifa_put(ifa);
2668 write_lock_bh(&idev->lock);
2670 #endif
2671 while ((ifa = idev->addr_list) != NULL) {
2672 idev->addr_list = ifa->if_next;
2673 ifa->if_next = NULL;
2674 ifa->dead = 1;
2675 addrconf_del_timer(ifa);
2676 write_unlock_bh(&idev->lock);
2678 __ipv6_ifa_notify(RTM_DELADDR, ifa);
2679 atomic_notifier_call_chain(&inet6addr_chain, NETDEV_DOWN, ifa);
2680 in6_ifa_put(ifa);
2682 write_lock_bh(&idev->lock);
2684 write_unlock_bh(&idev->lock);
2686 /* Step 5: Discard multicast list */
2688 if (how)
2689 ipv6_mc_destroy_dev(idev);
2690 else
2691 ipv6_mc_down(idev);
2693 idev->tstamp = jiffies;
2695 /* Shot the device (if unregistered) */
2697 if (how) {
2698 addrconf_sysctl_unregister(idev);
2699 neigh_parms_release(&nd_tbl, idev->nd_parms);
2700 neigh_ifdown(&nd_tbl, dev);
2701 in6_dev_put(idev);
2703 return 0;
2706 static void addrconf_rs_timer(unsigned long data)
2708 struct inet6_ifaddr *ifp = (struct inet6_ifaddr *) data;
2710 if (ifp->idev->cnf.forwarding)
2711 goto out;
2713 if (ifp->idev->if_flags & IF_RA_RCVD) {
2715 * Announcement received after solicitation
2716 * was sent
2718 goto out;
2721 spin_lock(&ifp->lock);
2722 if (ifp->probes++ < ifp->idev->cnf.rtr_solicits) {
2723 /* The wait after the last probe can be shorter */
2724 addrconf_mod_timer(ifp, AC_RS,
2725 (ifp->probes == ifp->idev->cnf.rtr_solicits) ?
2726 ifp->idev->cnf.rtr_solicit_delay :
2727 ifp->idev->cnf.rtr_solicit_interval);
2728 spin_unlock(&ifp->lock);
2730 ndisc_send_rs(ifp->idev->dev, &ifp->addr, &in6addr_linklocal_allrouters);
2731 } else {
2732 spin_unlock(&ifp->lock);
2734 * Note: we do not support deprecated "all on-link"
2735 * assumption any longer.
2737 printk(KERN_DEBUG "%s: no IPv6 routers present\n",
2738 ifp->idev->dev->name);
2741 out:
2742 in6_ifa_put(ifp);
2746 * Duplicate Address Detection
2748 static void addrconf_dad_kick(struct inet6_ifaddr *ifp)
2750 unsigned long rand_num;
2751 struct inet6_dev *idev = ifp->idev;
2753 if (ifp->flags & IFA_F_OPTIMISTIC)
2754 rand_num = 0;
2755 else
2756 rand_num = net_random() % (idev->cnf.rtr_solicit_delay ? : 1);
2758 ifp->probes = idev->cnf.dad_transmits;
2759 addrconf_mod_timer(ifp, AC_DAD, rand_num);
2762 static void addrconf_dad_start(struct inet6_ifaddr *ifp, u32 flags)
2764 struct inet6_dev *idev = ifp->idev;
2765 struct net_device *dev = idev->dev;
2767 addrconf_join_solict(dev, &ifp->addr);
2769 net_srandom(ifp->addr.s6_addr32[3]);
2771 read_lock_bh(&idev->lock);
2772 if (ifp->dead)
2773 goto out;
2774 spin_lock_bh(&ifp->lock);
2776 if (dev->flags&(IFF_NOARP|IFF_LOOPBACK) ||
2777 idev->cnf.accept_dad < 1 ||
2778 !(ifp->flags&IFA_F_TENTATIVE) ||
2779 ifp->flags & IFA_F_NODAD) {
2780 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC);
2781 spin_unlock_bh(&ifp->lock);
2782 read_unlock_bh(&idev->lock);
2784 addrconf_dad_completed(ifp);
2785 return;
2788 if (!(idev->if_flags & IF_READY)) {
2789 spin_unlock_bh(&ifp->lock);
2790 read_unlock_bh(&idev->lock);
2792 * If the defice is not ready:
2793 * - keep it tentative if it is a permanent address.
2794 * - otherwise, kill it.
2796 in6_ifa_hold(ifp);
2797 addrconf_dad_stop(ifp);
2798 return;
2802 * Optimistic nodes can start receiving
2803 * Frames right away
2805 if(ifp->flags & IFA_F_OPTIMISTIC)
2806 ip6_ins_rt(ifp->rt);
2808 addrconf_dad_kick(ifp);
2809 spin_unlock_bh(&ifp->lock);
2810 out:
2811 read_unlock_bh(&idev->lock);
2814 static void addrconf_dad_timer(unsigned long data)
2816 struct inet6_ifaddr *ifp = (struct inet6_ifaddr *) data;
2817 struct inet6_dev *idev = ifp->idev;
2818 struct in6_addr mcaddr;
2820 read_lock_bh(&idev->lock);
2821 if (idev->dead) {
2822 read_unlock_bh(&idev->lock);
2823 goto out;
2825 if (idev->cnf.accept_dad > 1 && idev->cnf.disable_ipv6) {
2826 read_unlock_bh(&idev->lock);
2827 addrconf_dad_failure(ifp);
2828 return;
2830 spin_lock_bh(&ifp->lock);
2831 if (ifp->probes == 0) {
2833 * DAD was successful
2836 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC);
2837 spin_unlock_bh(&ifp->lock);
2838 read_unlock_bh(&idev->lock);
2840 addrconf_dad_completed(ifp);
2842 goto out;
2845 ifp->probes--;
2846 addrconf_mod_timer(ifp, AC_DAD, ifp->idev->nd_parms->retrans_time);
2847 spin_unlock_bh(&ifp->lock);
2848 read_unlock_bh(&idev->lock);
2850 /* send a neighbour solicitation for our addr */
2851 addrconf_addr_solict_mult(&ifp->addr, &mcaddr);
2852 ndisc_send_ns(ifp->idev->dev, NULL, &ifp->addr, &mcaddr, &in6addr_any);
2853 out:
2854 in6_ifa_put(ifp);
2857 static void addrconf_dad_completed(struct inet6_ifaddr *ifp)
2859 struct net_device * dev = ifp->idev->dev;
2862 * Configure the address for reception. Now it is valid.
2865 ipv6_ifa_notify(RTM_NEWADDR, ifp);
2867 /* If added prefix is link local and forwarding is off,
2868 start sending router solicitations.
2871 if (ifp->idev->cnf.forwarding == 0 &&
2872 ifp->idev->cnf.rtr_solicits > 0 &&
2873 (dev->flags&IFF_LOOPBACK) == 0 &&
2874 (ipv6_addr_type(&ifp->addr) & IPV6_ADDR_LINKLOCAL)) {
2876 * If a host as already performed a random delay
2877 * [...] as part of DAD [...] there is no need
2878 * to delay again before sending the first RS
2880 ndisc_send_rs(ifp->idev->dev, &ifp->addr, &in6addr_linklocal_allrouters);
2882 spin_lock_bh(&ifp->lock);
2883 ifp->probes = 1;
2884 ifp->idev->if_flags |= IF_RS_SENT;
2885 addrconf_mod_timer(ifp, AC_RS, ifp->idev->cnf.rtr_solicit_interval);
2886 spin_unlock_bh(&ifp->lock);
2890 static void addrconf_dad_run(struct inet6_dev *idev) {
2891 struct inet6_ifaddr *ifp;
2893 read_lock_bh(&idev->lock);
2894 for (ifp = idev->addr_list; ifp; ifp = ifp->if_next) {
2895 spin_lock_bh(&ifp->lock);
2896 if (!(ifp->flags & IFA_F_TENTATIVE)) {
2897 spin_unlock_bh(&ifp->lock);
2898 continue;
2900 spin_unlock_bh(&ifp->lock);
2901 addrconf_dad_kick(ifp);
2903 read_unlock_bh(&idev->lock);
2906 #ifdef CONFIG_PROC_FS
2907 struct if6_iter_state {
2908 struct seq_net_private p;
2909 int bucket;
2912 static struct inet6_ifaddr *if6_get_first(struct seq_file *seq)
2914 struct inet6_ifaddr *ifa = NULL;
2915 struct if6_iter_state *state = seq->private;
2916 struct net *net = seq_file_net(seq);
2918 for (state->bucket = 0; state->bucket < IN6_ADDR_HSIZE; ++state->bucket) {
2919 ifa = inet6_addr_lst[state->bucket];
2921 while (ifa && !net_eq(dev_net(ifa->idev->dev), net))
2922 ifa = ifa->lst_next;
2923 if (ifa)
2924 break;
2926 return ifa;
2929 static struct inet6_ifaddr *if6_get_next(struct seq_file *seq, struct inet6_ifaddr *ifa)
2931 struct if6_iter_state *state = seq->private;
2932 struct net *net = seq_file_net(seq);
2934 ifa = ifa->lst_next;
2935 try_again:
2936 if (ifa) {
2937 if (!net_eq(dev_net(ifa->idev->dev), net)) {
2938 ifa = ifa->lst_next;
2939 goto try_again;
2943 if (!ifa && ++state->bucket < IN6_ADDR_HSIZE) {
2944 ifa = inet6_addr_lst[state->bucket];
2945 goto try_again;
2948 return ifa;
2951 static struct inet6_ifaddr *if6_get_idx(struct seq_file *seq, loff_t pos)
2953 struct inet6_ifaddr *ifa = if6_get_first(seq);
2955 if (ifa)
2956 while(pos && (ifa = if6_get_next(seq, ifa)) != NULL)
2957 --pos;
2958 return pos ? NULL : ifa;
2961 static void *if6_seq_start(struct seq_file *seq, loff_t *pos)
2962 __acquires(addrconf_hash_lock)
2964 read_lock_bh(&addrconf_hash_lock);
2965 return if6_get_idx(seq, *pos);
2968 static void *if6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
2970 struct inet6_ifaddr *ifa;
2972 ifa = if6_get_next(seq, v);
2973 ++*pos;
2974 return ifa;
2977 static void if6_seq_stop(struct seq_file *seq, void *v)
2978 __releases(addrconf_hash_lock)
2980 read_unlock_bh(&addrconf_hash_lock);
2983 static int if6_seq_show(struct seq_file *seq, void *v)
2985 struct inet6_ifaddr *ifp = (struct inet6_ifaddr *)v;
2986 seq_printf(seq,
2987 NIP6_SEQFMT " %02x %02x %02x %02x %8s\n",
2988 NIP6(ifp->addr),
2989 ifp->idev->dev->ifindex,
2990 ifp->prefix_len,
2991 ifp->scope,
2992 ifp->flags,
2993 ifp->idev->dev->name);
2994 return 0;
2997 static const struct seq_operations if6_seq_ops = {
2998 .start = if6_seq_start,
2999 .next = if6_seq_next,
3000 .show = if6_seq_show,
3001 .stop = if6_seq_stop,
3004 static int if6_seq_open(struct inode *inode, struct file *file)
3006 return seq_open_net(inode, file, &if6_seq_ops,
3007 sizeof(struct if6_iter_state));
3010 static const struct file_operations if6_fops = {
3011 .owner = THIS_MODULE,
3012 .open = if6_seq_open,
3013 .read = seq_read,
3014 .llseek = seq_lseek,
3015 .release = seq_release_net,
3018 static int if6_proc_net_init(struct net *net)
3020 if (!proc_net_fops_create(net, "if_inet6", S_IRUGO, &if6_fops))
3021 return -ENOMEM;
3022 return 0;
3025 static void if6_proc_net_exit(struct net *net)
3027 proc_net_remove(net, "if_inet6");
3030 static struct pernet_operations if6_proc_net_ops = {
3031 .init = if6_proc_net_init,
3032 .exit = if6_proc_net_exit,
3035 int __init if6_proc_init(void)
3037 return register_pernet_subsys(&if6_proc_net_ops);
3040 void if6_proc_exit(void)
3042 unregister_pernet_subsys(&if6_proc_net_ops);
3044 #endif /* CONFIG_PROC_FS */
3046 #if defined(CONFIG_IPV6_MIP6) || defined(CONFIG_IPV6_MIP6_MODULE)
3047 /* Check if address is a home address configured on any interface. */
3048 int ipv6_chk_home_addr(struct net *net, struct in6_addr *addr)
3050 int ret = 0;
3051 struct inet6_ifaddr * ifp;
3052 u8 hash = ipv6_addr_hash(addr);
3053 read_lock_bh(&addrconf_hash_lock);
3054 for (ifp = inet6_addr_lst[hash]; ifp; ifp = ifp->lst_next) {
3055 if (!net_eq(dev_net(ifp->idev->dev), net))
3056 continue;
3057 if (ipv6_addr_equal(&ifp->addr, addr) &&
3058 (ifp->flags & IFA_F_HOMEADDRESS)) {
3059 ret = 1;
3060 break;
3063 read_unlock_bh(&addrconf_hash_lock);
3064 return ret;
3066 #endif
3069 * Periodic address status verification
3072 static void addrconf_verify(unsigned long foo)
3074 struct inet6_ifaddr *ifp;
3075 unsigned long now, next;
3076 int i;
3078 spin_lock_bh(&addrconf_verify_lock);
3079 now = jiffies;
3080 next = now + ADDR_CHECK_FREQUENCY;
3082 del_timer(&addr_chk_timer);
3084 for (i=0; i < IN6_ADDR_HSIZE; i++) {
3086 restart:
3087 read_lock(&addrconf_hash_lock);
3088 for (ifp=inet6_addr_lst[i]; ifp; ifp=ifp->lst_next) {
3089 unsigned long age;
3090 #ifdef CONFIG_IPV6_PRIVACY
3091 unsigned long regen_advance;
3092 #endif
3094 if (ifp->flags & IFA_F_PERMANENT)
3095 continue;
3097 spin_lock(&ifp->lock);
3098 age = (now - ifp->tstamp) / HZ;
3100 #ifdef CONFIG_IPV6_PRIVACY
3101 regen_advance = ifp->idev->cnf.regen_max_retry *
3102 ifp->idev->cnf.dad_transmits *
3103 ifp->idev->nd_parms->retrans_time / HZ;
3104 #endif
3106 if (ifp->valid_lft != INFINITY_LIFE_TIME &&
3107 age >= ifp->valid_lft) {
3108 spin_unlock(&ifp->lock);
3109 in6_ifa_hold(ifp);
3110 read_unlock(&addrconf_hash_lock);
3111 ipv6_del_addr(ifp);
3112 goto restart;
3113 } else if (ifp->prefered_lft == INFINITY_LIFE_TIME) {
3114 spin_unlock(&ifp->lock);
3115 continue;
3116 } else if (age >= ifp->prefered_lft) {
3117 /* jiffies - ifp->tsamp > age >= ifp->prefered_lft */
3118 int deprecate = 0;
3120 if (!(ifp->flags&IFA_F_DEPRECATED)) {
3121 deprecate = 1;
3122 ifp->flags |= IFA_F_DEPRECATED;
3125 if (time_before(ifp->tstamp + ifp->valid_lft * HZ, next))
3126 next = ifp->tstamp + ifp->valid_lft * HZ;
3128 spin_unlock(&ifp->lock);
3130 if (deprecate) {
3131 in6_ifa_hold(ifp);
3132 read_unlock(&addrconf_hash_lock);
3134 ipv6_ifa_notify(0, ifp);
3135 in6_ifa_put(ifp);
3136 goto restart;
3138 #ifdef CONFIG_IPV6_PRIVACY
3139 } else if ((ifp->flags&IFA_F_TEMPORARY) &&
3140 !(ifp->flags&IFA_F_TENTATIVE)) {
3141 if (age >= ifp->prefered_lft - regen_advance) {
3142 struct inet6_ifaddr *ifpub = ifp->ifpub;
3143 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3144 next = ifp->tstamp + ifp->prefered_lft * HZ;
3145 if (!ifp->regen_count && ifpub) {
3146 ifp->regen_count++;
3147 in6_ifa_hold(ifp);
3148 in6_ifa_hold(ifpub);
3149 spin_unlock(&ifp->lock);
3150 read_unlock(&addrconf_hash_lock);
3151 spin_lock(&ifpub->lock);
3152 ifpub->regen_count = 0;
3153 spin_unlock(&ifpub->lock);
3154 ipv6_create_tempaddr(ifpub, ifp);
3155 in6_ifa_put(ifpub);
3156 in6_ifa_put(ifp);
3157 goto restart;
3159 } else if (time_before(ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ, next))
3160 next = ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ;
3161 spin_unlock(&ifp->lock);
3162 #endif
3163 } else {
3164 /* ifp->prefered_lft <= ifp->valid_lft */
3165 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3166 next = ifp->tstamp + ifp->prefered_lft * HZ;
3167 spin_unlock(&ifp->lock);
3170 read_unlock(&addrconf_hash_lock);
3173 addr_chk_timer.expires = time_before(next, jiffies + HZ) ? jiffies + HZ : next;
3174 add_timer(&addr_chk_timer);
3175 spin_unlock_bh(&addrconf_verify_lock);
3178 static struct in6_addr *extract_addr(struct nlattr *addr, struct nlattr *local)
3180 struct in6_addr *pfx = NULL;
3182 if (addr)
3183 pfx = nla_data(addr);
3185 if (local) {
3186 if (pfx && nla_memcmp(local, pfx, sizeof(*pfx)))
3187 pfx = NULL;
3188 else
3189 pfx = nla_data(local);
3192 return pfx;
3195 static const struct nla_policy ifa_ipv6_policy[IFA_MAX+1] = {
3196 [IFA_ADDRESS] = { .len = sizeof(struct in6_addr) },
3197 [IFA_LOCAL] = { .len = sizeof(struct in6_addr) },
3198 [IFA_CACHEINFO] = { .len = sizeof(struct ifa_cacheinfo) },
3201 static int
3202 inet6_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
3204 struct net *net = sock_net(skb->sk);
3205 struct ifaddrmsg *ifm;
3206 struct nlattr *tb[IFA_MAX+1];
3207 struct in6_addr *pfx;
3208 int err;
3210 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3211 if (err < 0)
3212 return err;
3214 ifm = nlmsg_data(nlh);
3215 pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3216 if (pfx == NULL)
3217 return -EINVAL;
3219 return inet6_addr_del(net, ifm->ifa_index, pfx, ifm->ifa_prefixlen);
3222 static int inet6_addr_modify(struct inet6_ifaddr *ifp, u8 ifa_flags,
3223 u32 prefered_lft, u32 valid_lft)
3225 u32 flags;
3226 clock_t expires;
3227 unsigned long timeout;
3229 if (!valid_lft || (prefered_lft > valid_lft))
3230 return -EINVAL;
3232 timeout = addrconf_timeout_fixup(valid_lft, HZ);
3233 if (addrconf_finite_timeout(timeout)) {
3234 expires = jiffies_to_clock_t(timeout * HZ);
3235 valid_lft = timeout;
3236 flags = RTF_EXPIRES;
3237 } else {
3238 expires = 0;
3239 flags = 0;
3240 ifa_flags |= IFA_F_PERMANENT;
3243 timeout = addrconf_timeout_fixup(prefered_lft, HZ);
3244 if (addrconf_finite_timeout(timeout)) {
3245 if (timeout == 0)
3246 ifa_flags |= IFA_F_DEPRECATED;
3247 prefered_lft = timeout;
3250 spin_lock_bh(&ifp->lock);
3251 ifp->flags = (ifp->flags & ~(IFA_F_DEPRECATED | IFA_F_PERMANENT | IFA_F_NODAD | IFA_F_HOMEADDRESS)) | ifa_flags;
3252 ifp->tstamp = jiffies;
3253 ifp->valid_lft = valid_lft;
3254 ifp->prefered_lft = prefered_lft;
3256 spin_unlock_bh(&ifp->lock);
3257 if (!(ifp->flags&IFA_F_TENTATIVE))
3258 ipv6_ifa_notify(0, ifp);
3260 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, ifp->idev->dev,
3261 expires, flags);
3262 addrconf_verify(0);
3264 return 0;
3267 static int
3268 inet6_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
3270 struct net *net = sock_net(skb->sk);
3271 struct ifaddrmsg *ifm;
3272 struct nlattr *tb[IFA_MAX+1];
3273 struct in6_addr *pfx;
3274 struct inet6_ifaddr *ifa;
3275 struct net_device *dev;
3276 u32 valid_lft = INFINITY_LIFE_TIME, preferred_lft = INFINITY_LIFE_TIME;
3277 u8 ifa_flags;
3278 int err;
3280 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3281 if (err < 0)
3282 return err;
3284 ifm = nlmsg_data(nlh);
3285 pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3286 if (pfx == NULL)
3287 return -EINVAL;
3289 if (tb[IFA_CACHEINFO]) {
3290 struct ifa_cacheinfo *ci;
3292 ci = nla_data(tb[IFA_CACHEINFO]);
3293 valid_lft = ci->ifa_valid;
3294 preferred_lft = ci->ifa_prefered;
3295 } else {
3296 preferred_lft = INFINITY_LIFE_TIME;
3297 valid_lft = INFINITY_LIFE_TIME;
3300 dev = __dev_get_by_index(net, ifm->ifa_index);
3301 if (dev == NULL)
3302 return -ENODEV;
3304 /* We ignore other flags so far. */
3305 ifa_flags = ifm->ifa_flags & (IFA_F_NODAD | IFA_F_HOMEADDRESS);
3307 ifa = ipv6_get_ifaddr(net, pfx, dev, 1);
3308 if (ifa == NULL) {
3310 * It would be best to check for !NLM_F_CREATE here but
3311 * userspace alreay relies on not having to provide this.
3313 return inet6_addr_add(net, ifm->ifa_index, pfx,
3314 ifm->ifa_prefixlen, ifa_flags,
3315 preferred_lft, valid_lft);
3318 if (nlh->nlmsg_flags & NLM_F_EXCL ||
3319 !(nlh->nlmsg_flags & NLM_F_REPLACE))
3320 err = -EEXIST;
3321 else
3322 err = inet6_addr_modify(ifa, ifa_flags, preferred_lft, valid_lft);
3324 in6_ifa_put(ifa);
3326 return err;
3329 static void put_ifaddrmsg(struct nlmsghdr *nlh, u8 prefixlen, u8 flags,
3330 u8 scope, int ifindex)
3332 struct ifaddrmsg *ifm;
3334 ifm = nlmsg_data(nlh);
3335 ifm->ifa_family = AF_INET6;
3336 ifm->ifa_prefixlen = prefixlen;
3337 ifm->ifa_flags = flags;
3338 ifm->ifa_scope = scope;
3339 ifm->ifa_index = ifindex;
3342 static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
3343 unsigned long tstamp, u32 preferred, u32 valid)
3345 struct ifa_cacheinfo ci;
3347 ci.cstamp = (u32)(TIME_DELTA(cstamp, INITIAL_JIFFIES) / HZ * 100
3348 + TIME_DELTA(cstamp, INITIAL_JIFFIES) % HZ * 100 / HZ);
3349 ci.tstamp = (u32)(TIME_DELTA(tstamp, INITIAL_JIFFIES) / HZ * 100
3350 + TIME_DELTA(tstamp, INITIAL_JIFFIES) % HZ * 100 / HZ);
3351 ci.ifa_prefered = preferred;
3352 ci.ifa_valid = valid;
3354 return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
3357 static inline int rt_scope(int ifa_scope)
3359 if (ifa_scope & IFA_HOST)
3360 return RT_SCOPE_HOST;
3361 else if (ifa_scope & IFA_LINK)
3362 return RT_SCOPE_LINK;
3363 else if (ifa_scope & IFA_SITE)
3364 return RT_SCOPE_SITE;
3365 else
3366 return RT_SCOPE_UNIVERSE;
3369 static inline int inet6_ifaddr_msgsize(void)
3371 return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
3372 + nla_total_size(16) /* IFA_ADDRESS */
3373 + nla_total_size(sizeof(struct ifa_cacheinfo));
3376 static int inet6_fill_ifaddr(struct sk_buff *skb, struct inet6_ifaddr *ifa,
3377 u32 pid, u32 seq, int event, unsigned int flags)
3379 struct nlmsghdr *nlh;
3380 u32 preferred, valid;
3382 nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3383 if (nlh == NULL)
3384 return -EMSGSIZE;
3386 put_ifaddrmsg(nlh, ifa->prefix_len, ifa->flags, rt_scope(ifa->scope),
3387 ifa->idev->dev->ifindex);
3389 if (!(ifa->flags&IFA_F_PERMANENT)) {
3390 preferred = ifa->prefered_lft;
3391 valid = ifa->valid_lft;
3392 if (preferred != INFINITY_LIFE_TIME) {
3393 long tval = (jiffies - ifa->tstamp)/HZ;
3394 preferred -= tval;
3395 if (valid != INFINITY_LIFE_TIME)
3396 valid -= tval;
3398 } else {
3399 preferred = INFINITY_LIFE_TIME;
3400 valid = INFINITY_LIFE_TIME;
3403 if (nla_put(skb, IFA_ADDRESS, 16, &ifa->addr) < 0 ||
3404 put_cacheinfo(skb, ifa->cstamp, ifa->tstamp, preferred, valid) < 0) {
3405 nlmsg_cancel(skb, nlh);
3406 return -EMSGSIZE;
3409 return nlmsg_end(skb, nlh);
3412 static int inet6_fill_ifmcaddr(struct sk_buff *skb, struct ifmcaddr6 *ifmca,
3413 u32 pid, u32 seq, int event, u16 flags)
3415 struct nlmsghdr *nlh;
3416 u8 scope = RT_SCOPE_UNIVERSE;
3417 int ifindex = ifmca->idev->dev->ifindex;
3419 if (ipv6_addr_scope(&ifmca->mca_addr) & IFA_SITE)
3420 scope = RT_SCOPE_SITE;
3422 nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3423 if (nlh == NULL)
3424 return -EMSGSIZE;
3426 put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3427 if (nla_put(skb, IFA_MULTICAST, 16, &ifmca->mca_addr) < 0 ||
3428 put_cacheinfo(skb, ifmca->mca_cstamp, ifmca->mca_tstamp,
3429 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3430 nlmsg_cancel(skb, nlh);
3431 return -EMSGSIZE;
3434 return nlmsg_end(skb, nlh);
3437 static int inet6_fill_ifacaddr(struct sk_buff *skb, struct ifacaddr6 *ifaca,
3438 u32 pid, u32 seq, int event, unsigned int flags)
3440 struct nlmsghdr *nlh;
3441 u8 scope = RT_SCOPE_UNIVERSE;
3442 int ifindex = ifaca->aca_idev->dev->ifindex;
3444 if (ipv6_addr_scope(&ifaca->aca_addr) & IFA_SITE)
3445 scope = RT_SCOPE_SITE;
3447 nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3448 if (nlh == NULL)
3449 return -EMSGSIZE;
3451 put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3452 if (nla_put(skb, IFA_ANYCAST, 16, &ifaca->aca_addr) < 0 ||
3453 put_cacheinfo(skb, ifaca->aca_cstamp, ifaca->aca_tstamp,
3454 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3455 nlmsg_cancel(skb, nlh);
3456 return -EMSGSIZE;
3459 return nlmsg_end(skb, nlh);
3462 enum addr_type_t
3464 UNICAST_ADDR,
3465 MULTICAST_ADDR,
3466 ANYCAST_ADDR,
3469 static int inet6_dump_addr(struct sk_buff *skb, struct netlink_callback *cb,
3470 enum addr_type_t type)
3472 int idx, ip_idx;
3473 int s_idx, s_ip_idx;
3474 int err = 1;
3475 struct net_device *dev;
3476 struct inet6_dev *idev = NULL;
3477 struct inet6_ifaddr *ifa;
3478 struct ifmcaddr6 *ifmca;
3479 struct ifacaddr6 *ifaca;
3480 struct net *net = sock_net(skb->sk);
3482 s_idx = cb->args[0];
3483 s_ip_idx = ip_idx = cb->args[1];
3485 idx = 0;
3486 for_each_netdev(net, dev) {
3487 if (idx < s_idx)
3488 goto cont;
3489 if (idx > s_idx)
3490 s_ip_idx = 0;
3491 ip_idx = 0;
3492 if ((idev = in6_dev_get(dev)) == NULL)
3493 goto cont;
3494 read_lock_bh(&idev->lock);
3495 switch (type) {
3496 case UNICAST_ADDR:
3497 /* unicast address incl. temp addr */
3498 for (ifa = idev->addr_list; ifa;
3499 ifa = ifa->if_next, ip_idx++) {
3500 if (ip_idx < s_ip_idx)
3501 continue;
3502 err = inet6_fill_ifaddr(skb, ifa,
3503 NETLINK_CB(cb->skb).pid,
3504 cb->nlh->nlmsg_seq,
3505 RTM_NEWADDR,
3506 NLM_F_MULTI);
3508 break;
3509 case MULTICAST_ADDR:
3510 /* multicast address */
3511 for (ifmca = idev->mc_list; ifmca;
3512 ifmca = ifmca->next, ip_idx++) {
3513 if (ip_idx < s_ip_idx)
3514 continue;
3515 err = inet6_fill_ifmcaddr(skb, ifmca,
3516 NETLINK_CB(cb->skb).pid,
3517 cb->nlh->nlmsg_seq,
3518 RTM_GETMULTICAST,
3519 NLM_F_MULTI);
3521 break;
3522 case ANYCAST_ADDR:
3523 /* anycast address */
3524 for (ifaca = idev->ac_list; ifaca;
3525 ifaca = ifaca->aca_next, ip_idx++) {
3526 if (ip_idx < s_ip_idx)
3527 continue;
3528 err = inet6_fill_ifacaddr(skb, ifaca,
3529 NETLINK_CB(cb->skb).pid,
3530 cb->nlh->nlmsg_seq,
3531 RTM_GETANYCAST,
3532 NLM_F_MULTI);
3534 break;
3535 default:
3536 break;
3538 read_unlock_bh(&idev->lock);
3539 in6_dev_put(idev);
3541 if (err <= 0)
3542 break;
3543 cont:
3544 idx++;
3546 cb->args[0] = idx;
3547 cb->args[1] = ip_idx;
3548 return skb->len;
3551 static int inet6_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
3553 enum addr_type_t type = UNICAST_ADDR;
3555 return inet6_dump_addr(skb, cb, type);
3558 static int inet6_dump_ifmcaddr(struct sk_buff *skb, struct netlink_callback *cb)
3560 enum addr_type_t type = MULTICAST_ADDR;
3562 return inet6_dump_addr(skb, cb, type);
3566 static int inet6_dump_ifacaddr(struct sk_buff *skb, struct netlink_callback *cb)
3568 enum addr_type_t type = ANYCAST_ADDR;
3570 return inet6_dump_addr(skb, cb, type);
3573 static int inet6_rtm_getaddr(struct sk_buff *in_skb, struct nlmsghdr* nlh,
3574 void *arg)
3576 struct net *net = sock_net(in_skb->sk);
3577 struct ifaddrmsg *ifm;
3578 struct nlattr *tb[IFA_MAX+1];
3579 struct in6_addr *addr = NULL;
3580 struct net_device *dev = NULL;
3581 struct inet6_ifaddr *ifa;
3582 struct sk_buff *skb;
3583 int err;
3585 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3586 if (err < 0)
3587 goto errout;
3589 addr = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3590 if (addr == NULL) {
3591 err = -EINVAL;
3592 goto errout;
3595 ifm = nlmsg_data(nlh);
3596 if (ifm->ifa_index)
3597 dev = __dev_get_by_index(net, ifm->ifa_index);
3599 if ((ifa = ipv6_get_ifaddr(net, addr, dev, 1)) == NULL) {
3600 err = -EADDRNOTAVAIL;
3601 goto errout;
3604 if ((skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_KERNEL)) == NULL) {
3605 err = -ENOBUFS;
3606 goto errout_ifa;
3609 err = inet6_fill_ifaddr(skb, ifa, NETLINK_CB(in_skb).pid,
3610 nlh->nlmsg_seq, RTM_NEWADDR, 0);
3611 if (err < 0) {
3612 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
3613 WARN_ON(err == -EMSGSIZE);
3614 kfree_skb(skb);
3615 goto errout_ifa;
3617 err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).pid);
3618 errout_ifa:
3619 in6_ifa_put(ifa);
3620 errout:
3621 return err;
3624 static void inet6_ifa_notify(int event, struct inet6_ifaddr *ifa)
3626 struct sk_buff *skb;
3627 struct net *net = dev_net(ifa->idev->dev);
3628 int err = -ENOBUFS;
3630 skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_ATOMIC);
3631 if (skb == NULL)
3632 goto errout;
3634 err = inet6_fill_ifaddr(skb, ifa, 0, 0, event, 0);
3635 if (err < 0) {
3636 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
3637 WARN_ON(err == -EMSGSIZE);
3638 kfree_skb(skb);
3639 goto errout;
3641 err = rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
3642 errout:
3643 if (err < 0)
3644 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
3647 static inline void ipv6_store_devconf(struct ipv6_devconf *cnf,
3648 __s32 *array, int bytes)
3650 BUG_ON(bytes < (DEVCONF_MAX * 4));
3652 memset(array, 0, bytes);
3653 array[DEVCONF_FORWARDING] = cnf->forwarding;
3654 array[DEVCONF_HOPLIMIT] = cnf->hop_limit;
3655 array[DEVCONF_MTU6] = cnf->mtu6;
3656 array[DEVCONF_ACCEPT_RA] = cnf->accept_ra;
3657 array[DEVCONF_ACCEPT_REDIRECTS] = cnf->accept_redirects;
3658 array[DEVCONF_AUTOCONF] = cnf->autoconf;
3659 array[DEVCONF_DAD_TRANSMITS] = cnf->dad_transmits;
3660 array[DEVCONF_RTR_SOLICITS] = cnf->rtr_solicits;
3661 array[DEVCONF_RTR_SOLICIT_INTERVAL] = cnf->rtr_solicit_interval;
3662 array[DEVCONF_RTR_SOLICIT_DELAY] = cnf->rtr_solicit_delay;
3663 array[DEVCONF_FORCE_MLD_VERSION] = cnf->force_mld_version;
3664 #ifdef CONFIG_IPV6_PRIVACY
3665 array[DEVCONF_USE_TEMPADDR] = cnf->use_tempaddr;
3666 array[DEVCONF_TEMP_VALID_LFT] = cnf->temp_valid_lft;
3667 array[DEVCONF_TEMP_PREFERED_LFT] = cnf->temp_prefered_lft;
3668 array[DEVCONF_REGEN_MAX_RETRY] = cnf->regen_max_retry;
3669 array[DEVCONF_MAX_DESYNC_FACTOR] = cnf->max_desync_factor;
3670 #endif
3671 array[DEVCONF_MAX_ADDRESSES] = cnf->max_addresses;
3672 array[DEVCONF_ACCEPT_RA_DEFRTR] = cnf->accept_ra_defrtr;
3673 array[DEVCONF_ACCEPT_RA_PINFO] = cnf->accept_ra_pinfo;
3674 #ifdef CONFIG_IPV6_ROUTER_PREF
3675 array[DEVCONF_ACCEPT_RA_RTR_PREF] = cnf->accept_ra_rtr_pref;
3676 array[DEVCONF_RTR_PROBE_INTERVAL] = cnf->rtr_probe_interval;
3677 #ifdef CONFIG_IPV6_ROUTE_INFO
3678 array[DEVCONF_ACCEPT_RA_RT_INFO_MAX_PLEN] = cnf->accept_ra_rt_info_max_plen;
3679 #endif
3680 #endif
3681 array[DEVCONF_PROXY_NDP] = cnf->proxy_ndp;
3682 array[DEVCONF_ACCEPT_SOURCE_ROUTE] = cnf->accept_source_route;
3683 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
3684 array[DEVCONF_OPTIMISTIC_DAD] = cnf->optimistic_dad;
3685 #endif
3686 #ifdef CONFIG_IPV6_MROUTE
3687 array[DEVCONF_MC_FORWARDING] = cnf->mc_forwarding;
3688 #endif
3689 array[DEVCONF_DISABLE_IPV6] = cnf->disable_ipv6;
3690 array[DEVCONF_ACCEPT_DAD] = cnf->accept_dad;
3693 static inline size_t inet6_if_nlmsg_size(void)
3695 return NLMSG_ALIGN(sizeof(struct ifinfomsg))
3696 + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
3697 + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
3698 + nla_total_size(4) /* IFLA_MTU */
3699 + nla_total_size(4) /* IFLA_LINK */
3700 + nla_total_size( /* IFLA_PROTINFO */
3701 nla_total_size(4) /* IFLA_INET6_FLAGS */
3702 + nla_total_size(sizeof(struct ifla_cacheinfo))
3703 + nla_total_size(DEVCONF_MAX * 4) /* IFLA_INET6_CONF */
3704 + nla_total_size(IPSTATS_MIB_MAX * 8) /* IFLA_INET6_STATS */
3705 + nla_total_size(ICMP6_MIB_MAX * 8) /* IFLA_INET6_ICMP6STATS */
3709 static inline void __snmp6_fill_stats(u64 *stats, void **mib, int items,
3710 int bytes)
3712 int i;
3713 int pad = bytes - sizeof(u64) * items;
3714 BUG_ON(pad < 0);
3716 /* Use put_unaligned() because stats may not be aligned for u64. */
3717 put_unaligned(items, &stats[0]);
3718 for (i = 1; i < items; i++)
3719 put_unaligned(snmp_fold_field(mib, i), &stats[i]);
3721 memset(&stats[items], 0, pad);
3724 static void snmp6_fill_stats(u64 *stats, struct inet6_dev *idev, int attrtype,
3725 int bytes)
3727 switch(attrtype) {
3728 case IFLA_INET6_STATS:
3729 __snmp6_fill_stats(stats, (void **)idev->stats.ipv6, IPSTATS_MIB_MAX, bytes);
3730 break;
3731 case IFLA_INET6_ICMP6STATS:
3732 __snmp6_fill_stats(stats, (void **)idev->stats.icmpv6, ICMP6_MIB_MAX, bytes);
3733 break;
3737 static int inet6_fill_ifinfo(struct sk_buff *skb, struct inet6_dev *idev,
3738 u32 pid, u32 seq, int event, unsigned int flags)
3740 struct net_device *dev = idev->dev;
3741 struct nlattr *nla;
3742 struct ifinfomsg *hdr;
3743 struct nlmsghdr *nlh;
3744 void *protoinfo;
3745 struct ifla_cacheinfo ci;
3747 nlh = nlmsg_put(skb, pid, seq, event, sizeof(*hdr), flags);
3748 if (nlh == NULL)
3749 return -EMSGSIZE;
3751 hdr = nlmsg_data(nlh);
3752 hdr->ifi_family = AF_INET6;
3753 hdr->__ifi_pad = 0;
3754 hdr->ifi_type = dev->type;
3755 hdr->ifi_index = dev->ifindex;
3756 hdr->ifi_flags = dev_get_flags(dev);
3757 hdr->ifi_change = 0;
3759 NLA_PUT_STRING(skb, IFLA_IFNAME, dev->name);
3761 if (dev->addr_len)
3762 NLA_PUT(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr);
3764 NLA_PUT_U32(skb, IFLA_MTU, dev->mtu);
3765 if (dev->ifindex != dev->iflink)
3766 NLA_PUT_U32(skb, IFLA_LINK, dev->iflink);
3768 protoinfo = nla_nest_start(skb, IFLA_PROTINFO);
3769 if (protoinfo == NULL)
3770 goto nla_put_failure;
3772 NLA_PUT_U32(skb, IFLA_INET6_FLAGS, idev->if_flags);
3774 ci.max_reasm_len = IPV6_MAXPLEN;
3775 ci.tstamp = (__u32)(TIME_DELTA(idev->tstamp, INITIAL_JIFFIES) / HZ * 100
3776 + TIME_DELTA(idev->tstamp, INITIAL_JIFFIES) % HZ * 100 / HZ);
3777 ci.reachable_time = idev->nd_parms->reachable_time;
3778 ci.retrans_time = idev->nd_parms->retrans_time;
3779 NLA_PUT(skb, IFLA_INET6_CACHEINFO, sizeof(ci), &ci);
3781 nla = nla_reserve(skb, IFLA_INET6_CONF, DEVCONF_MAX * sizeof(s32));
3782 if (nla == NULL)
3783 goto nla_put_failure;
3784 ipv6_store_devconf(&idev->cnf, nla_data(nla), nla_len(nla));
3786 /* XXX - MC not implemented */
3788 nla = nla_reserve(skb, IFLA_INET6_STATS, IPSTATS_MIB_MAX * sizeof(u64));
3789 if (nla == NULL)
3790 goto nla_put_failure;
3791 snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_STATS, nla_len(nla));
3793 nla = nla_reserve(skb, IFLA_INET6_ICMP6STATS, ICMP6_MIB_MAX * sizeof(u64));
3794 if (nla == NULL)
3795 goto nla_put_failure;
3796 snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_ICMP6STATS, nla_len(nla));
3798 nla_nest_end(skb, protoinfo);
3799 return nlmsg_end(skb, nlh);
3801 nla_put_failure:
3802 nlmsg_cancel(skb, nlh);
3803 return -EMSGSIZE;
3806 static int inet6_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
3808 struct net *net = sock_net(skb->sk);
3809 int idx, err;
3810 int s_idx = cb->args[0];
3811 struct net_device *dev;
3812 struct inet6_dev *idev;
3814 read_lock(&dev_base_lock);
3815 idx = 0;
3816 for_each_netdev(net, dev) {
3817 if (idx < s_idx)
3818 goto cont;
3819 if ((idev = in6_dev_get(dev)) == NULL)
3820 goto cont;
3821 err = inet6_fill_ifinfo(skb, idev, NETLINK_CB(cb->skb).pid,
3822 cb->nlh->nlmsg_seq, RTM_NEWLINK, NLM_F_MULTI);
3823 in6_dev_put(idev);
3824 if (err <= 0)
3825 break;
3826 cont:
3827 idx++;
3829 read_unlock(&dev_base_lock);
3830 cb->args[0] = idx;
3832 return skb->len;
3835 void inet6_ifinfo_notify(int event, struct inet6_dev *idev)
3837 struct sk_buff *skb;
3838 struct net *net = dev_net(idev->dev);
3839 int err = -ENOBUFS;
3841 skb = nlmsg_new(inet6_if_nlmsg_size(), GFP_ATOMIC);
3842 if (skb == NULL)
3843 goto errout;
3845 err = inet6_fill_ifinfo(skb, idev, 0, 0, event, 0);
3846 if (err < 0) {
3847 /* -EMSGSIZE implies BUG in inet6_if_nlmsg_size() */
3848 WARN_ON(err == -EMSGSIZE);
3849 kfree_skb(skb);
3850 goto errout;
3852 err = rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
3853 errout:
3854 if (err < 0)
3855 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
3858 static inline size_t inet6_prefix_nlmsg_size(void)
3860 return NLMSG_ALIGN(sizeof(struct prefixmsg))
3861 + nla_total_size(sizeof(struct in6_addr))
3862 + nla_total_size(sizeof(struct prefix_cacheinfo));
3865 static int inet6_fill_prefix(struct sk_buff *skb, struct inet6_dev *idev,
3866 struct prefix_info *pinfo, u32 pid, u32 seq,
3867 int event, unsigned int flags)
3869 struct prefixmsg *pmsg;
3870 struct nlmsghdr *nlh;
3871 struct prefix_cacheinfo ci;
3873 nlh = nlmsg_put(skb, pid, seq, event, sizeof(*pmsg), flags);
3874 if (nlh == NULL)
3875 return -EMSGSIZE;
3877 pmsg = nlmsg_data(nlh);
3878 pmsg->prefix_family = AF_INET6;
3879 pmsg->prefix_pad1 = 0;
3880 pmsg->prefix_pad2 = 0;
3881 pmsg->prefix_ifindex = idev->dev->ifindex;
3882 pmsg->prefix_len = pinfo->prefix_len;
3883 pmsg->prefix_type = pinfo->type;
3884 pmsg->prefix_pad3 = 0;
3885 pmsg->prefix_flags = 0;
3886 if (pinfo->onlink)
3887 pmsg->prefix_flags |= IF_PREFIX_ONLINK;
3888 if (pinfo->autoconf)
3889 pmsg->prefix_flags |= IF_PREFIX_AUTOCONF;
3891 NLA_PUT(skb, PREFIX_ADDRESS, sizeof(pinfo->prefix), &pinfo->prefix);
3893 ci.preferred_time = ntohl(pinfo->prefered);
3894 ci.valid_time = ntohl(pinfo->valid);
3895 NLA_PUT(skb, PREFIX_CACHEINFO, sizeof(ci), &ci);
3897 return nlmsg_end(skb, nlh);
3899 nla_put_failure:
3900 nlmsg_cancel(skb, nlh);
3901 return -EMSGSIZE;
3904 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
3905 struct prefix_info *pinfo)
3907 struct sk_buff *skb;
3908 struct net *net = dev_net(idev->dev);
3909 int err = -ENOBUFS;
3911 skb = nlmsg_new(inet6_prefix_nlmsg_size(), GFP_ATOMIC);
3912 if (skb == NULL)
3913 goto errout;
3915 err = inet6_fill_prefix(skb, idev, pinfo, 0, 0, event, 0);
3916 if (err < 0) {
3917 /* -EMSGSIZE implies BUG in inet6_prefix_nlmsg_size() */
3918 WARN_ON(err == -EMSGSIZE);
3919 kfree_skb(skb);
3920 goto errout;
3922 err = rtnl_notify(skb, net, 0, RTNLGRP_IPV6_PREFIX, NULL, GFP_ATOMIC);
3923 errout:
3924 if (err < 0)
3925 rtnl_set_sk_err(net, RTNLGRP_IPV6_PREFIX, err);
3928 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
3930 inet6_ifa_notify(event ? : RTM_NEWADDR, ifp);
3932 switch (event) {
3933 case RTM_NEWADDR:
3935 * If the address was optimistic
3936 * we inserted the route at the start of
3937 * our DAD process, so we don't need
3938 * to do it again
3940 if (!(ifp->rt->rt6i_node))
3941 ip6_ins_rt(ifp->rt);
3942 if (ifp->idev->cnf.forwarding)
3943 addrconf_join_anycast(ifp);
3944 break;
3945 case RTM_DELADDR:
3946 if (ifp->idev->cnf.forwarding)
3947 addrconf_leave_anycast(ifp);
3948 addrconf_leave_solict(ifp->idev, &ifp->addr);
3949 dst_hold(&ifp->rt->u.dst);
3950 if (ip6_del_rt(ifp->rt))
3951 dst_free(&ifp->rt->u.dst);
3952 break;
3956 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
3958 rcu_read_lock_bh();
3959 if (likely(ifp->idev->dead == 0))
3960 __ipv6_ifa_notify(event, ifp);
3961 rcu_read_unlock_bh();
3964 #ifdef CONFIG_SYSCTL
3966 static
3967 int addrconf_sysctl_forward(ctl_table *ctl, int write, struct file * filp,
3968 void __user *buffer, size_t *lenp, loff_t *ppos)
3970 int *valp = ctl->data;
3971 int val = *valp;
3972 int ret;
3974 ret = proc_dointvec(ctl, write, filp, buffer, lenp, ppos);
3976 if (write)
3977 addrconf_fixup_forwarding(ctl, valp, val);
3978 return ret;
3981 static int addrconf_sysctl_forward_strategy(ctl_table *table,
3982 int __user *name, int nlen,
3983 void __user *oldval,
3984 size_t __user *oldlenp,
3985 void __user *newval, size_t newlen)
3987 int *valp = table->data;
3988 int val = *valp;
3989 int new;
3991 if (!newval || !newlen)
3992 return 0;
3993 if (newlen != sizeof(int))
3994 return -EINVAL;
3995 if (get_user(new, (int __user *)newval))
3996 return -EFAULT;
3997 if (new == *valp)
3998 return 0;
3999 if (oldval && oldlenp) {
4000 size_t len;
4001 if (get_user(len, oldlenp))
4002 return -EFAULT;
4003 if (len) {
4004 if (len > table->maxlen)
4005 len = table->maxlen;
4006 if (copy_to_user(oldval, valp, len))
4007 return -EFAULT;
4008 if (put_user(len, oldlenp))
4009 return -EFAULT;
4013 *valp = new;
4014 addrconf_fixup_forwarding(table, valp, val);
4015 return 1;
4018 static struct addrconf_sysctl_table
4020 struct ctl_table_header *sysctl_header;
4021 ctl_table addrconf_vars[DEVCONF_MAX+1];
4022 char *dev_name;
4023 } addrconf_sysctl __read_mostly = {
4024 .sysctl_header = NULL,
4025 .addrconf_vars = {
4027 .ctl_name = NET_IPV6_FORWARDING,
4028 .procname = "forwarding",
4029 .data = &ipv6_devconf.forwarding,
4030 .maxlen = sizeof(int),
4031 .mode = 0644,
4032 .proc_handler = &addrconf_sysctl_forward,
4033 .strategy = &addrconf_sysctl_forward_strategy,
4036 .ctl_name = NET_IPV6_HOP_LIMIT,
4037 .procname = "hop_limit",
4038 .data = &ipv6_devconf.hop_limit,
4039 .maxlen = sizeof(int),
4040 .mode = 0644,
4041 .proc_handler = proc_dointvec,
4044 .ctl_name = NET_IPV6_MTU,
4045 .procname = "mtu",
4046 .data = &ipv6_devconf.mtu6,
4047 .maxlen = sizeof(int),
4048 .mode = 0644,
4049 .proc_handler = &proc_dointvec,
4052 .ctl_name = NET_IPV6_ACCEPT_RA,
4053 .procname = "accept_ra",
4054 .data = &ipv6_devconf.accept_ra,
4055 .maxlen = sizeof(int),
4056 .mode = 0644,
4057 .proc_handler = &proc_dointvec,
4060 .ctl_name = NET_IPV6_ACCEPT_REDIRECTS,
4061 .procname = "accept_redirects",
4062 .data = &ipv6_devconf.accept_redirects,
4063 .maxlen = sizeof(int),
4064 .mode = 0644,
4065 .proc_handler = &proc_dointvec,
4068 .ctl_name = NET_IPV6_AUTOCONF,
4069 .procname = "autoconf",
4070 .data = &ipv6_devconf.autoconf,
4071 .maxlen = sizeof(int),
4072 .mode = 0644,
4073 .proc_handler = &proc_dointvec,
4076 .ctl_name = NET_IPV6_DAD_TRANSMITS,
4077 .procname = "dad_transmits",
4078 .data = &ipv6_devconf.dad_transmits,
4079 .maxlen = sizeof(int),
4080 .mode = 0644,
4081 .proc_handler = &proc_dointvec,
4084 .ctl_name = NET_IPV6_RTR_SOLICITS,
4085 .procname = "router_solicitations",
4086 .data = &ipv6_devconf.rtr_solicits,
4087 .maxlen = sizeof(int),
4088 .mode = 0644,
4089 .proc_handler = &proc_dointvec,
4092 .ctl_name = NET_IPV6_RTR_SOLICIT_INTERVAL,
4093 .procname = "router_solicitation_interval",
4094 .data = &ipv6_devconf.rtr_solicit_interval,
4095 .maxlen = sizeof(int),
4096 .mode = 0644,
4097 .proc_handler = &proc_dointvec_jiffies,
4098 .strategy = &sysctl_jiffies,
4101 .ctl_name = NET_IPV6_RTR_SOLICIT_DELAY,
4102 .procname = "router_solicitation_delay",
4103 .data = &ipv6_devconf.rtr_solicit_delay,
4104 .maxlen = sizeof(int),
4105 .mode = 0644,
4106 .proc_handler = &proc_dointvec_jiffies,
4107 .strategy = &sysctl_jiffies,
4110 .ctl_name = NET_IPV6_FORCE_MLD_VERSION,
4111 .procname = "force_mld_version",
4112 .data = &ipv6_devconf.force_mld_version,
4113 .maxlen = sizeof(int),
4114 .mode = 0644,
4115 .proc_handler = &proc_dointvec,
4117 #ifdef CONFIG_IPV6_PRIVACY
4119 .ctl_name = NET_IPV6_USE_TEMPADDR,
4120 .procname = "use_tempaddr",
4121 .data = &ipv6_devconf.use_tempaddr,
4122 .maxlen = sizeof(int),
4123 .mode = 0644,
4124 .proc_handler = &proc_dointvec,
4127 .ctl_name = NET_IPV6_TEMP_VALID_LFT,
4128 .procname = "temp_valid_lft",
4129 .data = &ipv6_devconf.temp_valid_lft,
4130 .maxlen = sizeof(int),
4131 .mode = 0644,
4132 .proc_handler = &proc_dointvec,
4135 .ctl_name = NET_IPV6_TEMP_PREFERED_LFT,
4136 .procname = "temp_prefered_lft",
4137 .data = &ipv6_devconf.temp_prefered_lft,
4138 .maxlen = sizeof(int),
4139 .mode = 0644,
4140 .proc_handler = &proc_dointvec,
4143 .ctl_name = NET_IPV6_REGEN_MAX_RETRY,
4144 .procname = "regen_max_retry",
4145 .data = &ipv6_devconf.regen_max_retry,
4146 .maxlen = sizeof(int),
4147 .mode = 0644,
4148 .proc_handler = &proc_dointvec,
4151 .ctl_name = NET_IPV6_MAX_DESYNC_FACTOR,
4152 .procname = "max_desync_factor",
4153 .data = &ipv6_devconf.max_desync_factor,
4154 .maxlen = sizeof(int),
4155 .mode = 0644,
4156 .proc_handler = &proc_dointvec,
4158 #endif
4160 .ctl_name = NET_IPV6_MAX_ADDRESSES,
4161 .procname = "max_addresses",
4162 .data = &ipv6_devconf.max_addresses,
4163 .maxlen = sizeof(int),
4164 .mode = 0644,
4165 .proc_handler = &proc_dointvec,
4168 .ctl_name = NET_IPV6_ACCEPT_RA_DEFRTR,
4169 .procname = "accept_ra_defrtr",
4170 .data = &ipv6_devconf.accept_ra_defrtr,
4171 .maxlen = sizeof(int),
4172 .mode = 0644,
4173 .proc_handler = &proc_dointvec,
4176 .ctl_name = NET_IPV6_ACCEPT_RA_PINFO,
4177 .procname = "accept_ra_pinfo",
4178 .data = &ipv6_devconf.accept_ra_pinfo,
4179 .maxlen = sizeof(int),
4180 .mode = 0644,
4181 .proc_handler = &proc_dointvec,
4183 #ifdef CONFIG_IPV6_ROUTER_PREF
4185 .ctl_name = NET_IPV6_ACCEPT_RA_RTR_PREF,
4186 .procname = "accept_ra_rtr_pref",
4187 .data = &ipv6_devconf.accept_ra_rtr_pref,
4188 .maxlen = sizeof(int),
4189 .mode = 0644,
4190 .proc_handler = &proc_dointvec,
4193 .ctl_name = NET_IPV6_RTR_PROBE_INTERVAL,
4194 .procname = "router_probe_interval",
4195 .data = &ipv6_devconf.rtr_probe_interval,
4196 .maxlen = sizeof(int),
4197 .mode = 0644,
4198 .proc_handler = &proc_dointvec_jiffies,
4199 .strategy = &sysctl_jiffies,
4201 #ifdef CONFIG_IPV6_ROUTE_INFO
4203 .ctl_name = NET_IPV6_ACCEPT_RA_RT_INFO_MAX_PLEN,
4204 .procname = "accept_ra_rt_info_max_plen",
4205 .data = &ipv6_devconf.accept_ra_rt_info_max_plen,
4206 .maxlen = sizeof(int),
4207 .mode = 0644,
4208 .proc_handler = &proc_dointvec,
4210 #endif
4211 #endif
4213 .ctl_name = NET_IPV6_PROXY_NDP,
4214 .procname = "proxy_ndp",
4215 .data = &ipv6_devconf.proxy_ndp,
4216 .maxlen = sizeof(int),
4217 .mode = 0644,
4218 .proc_handler = &proc_dointvec,
4221 .ctl_name = NET_IPV6_ACCEPT_SOURCE_ROUTE,
4222 .procname = "accept_source_route",
4223 .data = &ipv6_devconf.accept_source_route,
4224 .maxlen = sizeof(int),
4225 .mode = 0644,
4226 .proc_handler = &proc_dointvec,
4228 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
4230 .ctl_name = CTL_UNNUMBERED,
4231 .procname = "optimistic_dad",
4232 .data = &ipv6_devconf.optimistic_dad,
4233 .maxlen = sizeof(int),
4234 .mode = 0644,
4235 .proc_handler = &proc_dointvec,
4238 #endif
4239 #ifdef CONFIG_IPV6_MROUTE
4241 .ctl_name = CTL_UNNUMBERED,
4242 .procname = "mc_forwarding",
4243 .data = &ipv6_devconf.mc_forwarding,
4244 .maxlen = sizeof(int),
4245 .mode = 0644,
4246 .proc_handler = &proc_dointvec,
4248 #endif
4250 .ctl_name = CTL_UNNUMBERED,
4251 .procname = "disable_ipv6",
4252 .data = &ipv6_devconf.disable_ipv6,
4253 .maxlen = sizeof(int),
4254 .mode = 0644,
4255 .proc_handler = &proc_dointvec,
4258 .ctl_name = CTL_UNNUMBERED,
4259 .procname = "accept_dad",
4260 .data = &ipv6_devconf.accept_dad,
4261 .maxlen = sizeof(int),
4262 .mode = 0644,
4263 .proc_handler = &proc_dointvec,
4266 .ctl_name = 0, /* sentinel */
4271 static int __addrconf_sysctl_register(struct net *net, char *dev_name,
4272 int ctl_name, struct inet6_dev *idev, struct ipv6_devconf *p)
4274 int i;
4275 struct addrconf_sysctl_table *t;
4277 #define ADDRCONF_CTL_PATH_DEV 3
4279 struct ctl_path addrconf_ctl_path[] = {
4280 { .procname = "net", .ctl_name = CTL_NET, },
4281 { .procname = "ipv6", .ctl_name = NET_IPV6, },
4282 { .procname = "conf", .ctl_name = NET_IPV6_CONF, },
4283 { /* to be set */ },
4284 { },
4288 t = kmemdup(&addrconf_sysctl, sizeof(*t), GFP_KERNEL);
4289 if (t == NULL)
4290 goto out;
4292 for (i=0; t->addrconf_vars[i].data; i++) {
4293 t->addrconf_vars[i].data += (char*)p - (char*)&ipv6_devconf;
4294 t->addrconf_vars[i].extra1 = idev; /* embedded; no ref */
4295 t->addrconf_vars[i].extra2 = net;
4299 * Make a copy of dev_name, because '.procname' is regarded as const
4300 * by sysctl and we wouldn't want anyone to change it under our feet
4301 * (see SIOCSIFNAME).
4303 t->dev_name = kstrdup(dev_name, GFP_KERNEL);
4304 if (!t->dev_name)
4305 goto free;
4307 addrconf_ctl_path[ADDRCONF_CTL_PATH_DEV].procname = t->dev_name;
4308 addrconf_ctl_path[ADDRCONF_CTL_PATH_DEV].ctl_name = ctl_name;
4310 t->sysctl_header = register_net_sysctl_table(net, addrconf_ctl_path,
4311 t->addrconf_vars);
4312 if (t->sysctl_header == NULL)
4313 goto free_procname;
4315 p->sysctl = t;
4316 return 0;
4318 free_procname:
4319 kfree(t->dev_name);
4320 free:
4321 kfree(t);
4322 out:
4323 return -ENOBUFS;
4326 static void __addrconf_sysctl_unregister(struct ipv6_devconf *p)
4328 struct addrconf_sysctl_table *t;
4330 if (p->sysctl == NULL)
4331 return;
4333 t = p->sysctl;
4334 p->sysctl = NULL;
4335 unregister_sysctl_table(t->sysctl_header);
4336 kfree(t->dev_name);
4337 kfree(t);
4340 static void addrconf_sysctl_register(struct inet6_dev *idev)
4342 neigh_sysctl_register(idev->dev, idev->nd_parms, NET_IPV6,
4343 NET_IPV6_NEIGH, "ipv6",
4344 &ndisc_ifinfo_sysctl_change,
4345 ndisc_ifinfo_sysctl_strategy);
4346 __addrconf_sysctl_register(dev_net(idev->dev), idev->dev->name,
4347 idev->dev->ifindex, idev, &idev->cnf);
4350 static void addrconf_sysctl_unregister(struct inet6_dev *idev)
4352 __addrconf_sysctl_unregister(&idev->cnf);
4353 neigh_sysctl_unregister(idev->nd_parms);
4357 #endif
4359 static int addrconf_init_net(struct net *net)
4361 int err;
4362 struct ipv6_devconf *all, *dflt;
4364 err = -ENOMEM;
4365 all = &ipv6_devconf;
4366 dflt = &ipv6_devconf_dflt;
4368 if (net != &init_net) {
4369 all = kmemdup(all, sizeof(ipv6_devconf), GFP_KERNEL);
4370 if (all == NULL)
4371 goto err_alloc_all;
4373 dflt = kmemdup(dflt, sizeof(ipv6_devconf_dflt), GFP_KERNEL);
4374 if (dflt == NULL)
4375 goto err_alloc_dflt;
4378 net->ipv6.devconf_all = all;
4379 net->ipv6.devconf_dflt = dflt;
4381 #ifdef CONFIG_SYSCTL
4382 err = __addrconf_sysctl_register(net, "all", NET_PROTO_CONF_ALL,
4383 NULL, all);
4384 if (err < 0)
4385 goto err_reg_all;
4387 err = __addrconf_sysctl_register(net, "default", NET_PROTO_CONF_DEFAULT,
4388 NULL, dflt);
4389 if (err < 0)
4390 goto err_reg_dflt;
4391 #endif
4392 return 0;
4394 #ifdef CONFIG_SYSCTL
4395 err_reg_dflt:
4396 __addrconf_sysctl_unregister(all);
4397 err_reg_all:
4398 kfree(dflt);
4399 #endif
4400 err_alloc_dflt:
4401 kfree(all);
4402 err_alloc_all:
4403 return err;
4406 static void addrconf_exit_net(struct net *net)
4408 #ifdef CONFIG_SYSCTL
4409 __addrconf_sysctl_unregister(net->ipv6.devconf_dflt);
4410 __addrconf_sysctl_unregister(net->ipv6.devconf_all);
4411 #endif
4412 if (net != &init_net) {
4413 kfree(net->ipv6.devconf_dflt);
4414 kfree(net->ipv6.devconf_all);
4418 static struct pernet_operations addrconf_ops = {
4419 .init = addrconf_init_net,
4420 .exit = addrconf_exit_net,
4424 * Device notifier
4427 int register_inet6addr_notifier(struct notifier_block *nb)
4429 return atomic_notifier_chain_register(&inet6addr_chain, nb);
4432 EXPORT_SYMBOL(register_inet6addr_notifier);
4434 int unregister_inet6addr_notifier(struct notifier_block *nb)
4436 return atomic_notifier_chain_unregister(&inet6addr_chain,nb);
4439 EXPORT_SYMBOL(unregister_inet6addr_notifier);
4441 static void addrconf_net_exit(struct net *net)
4443 struct net_device *dev;
4445 rtnl_lock();
4446 /* clean dev list */
4447 for_each_netdev(net, dev) {
4448 if (__in6_dev_get(dev) == NULL)
4449 continue;
4450 addrconf_ifdown(dev, 1);
4452 addrconf_ifdown(net->loopback_dev, 2);
4453 rtnl_unlock();
4456 static struct pernet_operations addrconf_net_ops = {
4457 .exit = addrconf_net_exit,
4461 * Init / cleanup code
4464 int __init addrconf_init(void)
4466 int err;
4468 if ((err = ipv6_addr_label_init()) < 0) {
4469 printk(KERN_CRIT "IPv6 Addrconf: cannot initialize default policy table: %d.\n",
4470 err);
4471 return err;
4474 register_pernet_subsys(&addrconf_ops);
4476 /* The addrconf netdev notifier requires that loopback_dev
4477 * has it's ipv6 private information allocated and setup
4478 * before it can bring up and give link-local addresses
4479 * to other devices which are up.
4481 * Unfortunately, loopback_dev is not necessarily the first
4482 * entry in the global dev_base list of net devices. In fact,
4483 * it is likely to be the very last entry on that list.
4484 * So this causes the notifier registry below to try and
4485 * give link-local addresses to all devices besides loopback_dev
4486 * first, then loopback_dev, which cases all the non-loopback_dev
4487 * devices to fail to get a link-local address.
4489 * So, as a temporary fix, allocate the ipv6 structure for
4490 * loopback_dev first by hand.
4491 * Longer term, all of the dependencies ipv6 has upon the loopback
4492 * device and it being up should be removed.
4494 rtnl_lock();
4495 if (!ipv6_add_dev(init_net.loopback_dev))
4496 err = -ENOMEM;
4497 rtnl_unlock();
4498 if (err)
4499 goto errlo;
4501 err = register_pernet_device(&addrconf_net_ops);
4502 if (err)
4503 return err;
4505 register_netdevice_notifier(&ipv6_dev_notf);
4507 addrconf_verify(0);
4509 err = __rtnl_register(PF_INET6, RTM_GETLINK, NULL, inet6_dump_ifinfo);
4510 if (err < 0)
4511 goto errout;
4513 /* Only the first call to __rtnl_register can fail */
4514 __rtnl_register(PF_INET6, RTM_NEWADDR, inet6_rtm_newaddr, NULL);
4515 __rtnl_register(PF_INET6, RTM_DELADDR, inet6_rtm_deladdr, NULL);
4516 __rtnl_register(PF_INET6, RTM_GETADDR, inet6_rtm_getaddr, inet6_dump_ifaddr);
4517 __rtnl_register(PF_INET6, RTM_GETMULTICAST, NULL, inet6_dump_ifmcaddr);
4518 __rtnl_register(PF_INET6, RTM_GETANYCAST, NULL, inet6_dump_ifacaddr);
4520 ipv6_addr_label_rtnl_register();
4522 return 0;
4523 errout:
4524 unregister_netdevice_notifier(&ipv6_dev_notf);
4525 errlo:
4526 unregister_pernet_subsys(&addrconf_ops);
4528 return err;
4531 void addrconf_cleanup(void)
4533 struct inet6_ifaddr *ifa;
4534 int i;
4536 unregister_netdevice_notifier(&ipv6_dev_notf);
4537 unregister_pernet_device(&addrconf_net_ops);
4539 unregister_pernet_subsys(&addrconf_ops);
4541 rtnl_lock();
4544 * Check hash table.
4546 write_lock_bh(&addrconf_hash_lock);
4547 for (i=0; i < IN6_ADDR_HSIZE; i++) {
4548 for (ifa=inet6_addr_lst[i]; ifa; ) {
4549 struct inet6_ifaddr *bifa;
4551 bifa = ifa;
4552 ifa = ifa->lst_next;
4553 printk(KERN_DEBUG "bug: IPv6 address leakage detected: ifa=%p\n", bifa);
4554 /* Do not free it; something is wrong.
4555 Now we can investigate it with debugger.
4559 write_unlock_bh(&addrconf_hash_lock);
4561 del_timer(&addr_chk_timer);
4562 rtnl_unlock();
4564 unregister_pernet_subsys(&addrconf_net_ops);