nfsd41: encode replay sequence from the slot values
[linux-2.6/linux-acpi-2.6/ibm-acpi-2.6.git] / net / ipv6 / addrconf.c
blob8c1e86afbbf5208c4eeadee48ce407dbf3b0016f
1 /*
2 * IPv6 Address [auto]configuration
3 * Linux INET6 implementation
5 * Authors:
6 * Pedro Roque <roque@di.fc.ul.pt>
7 * Alexey Kuznetsov <kuznet@ms2.inr.ac.ru>
9 * This program is free software; you can redistribute it and/or
10 * modify it under the terms of the GNU General Public License
11 * as published by the Free Software Foundation; either version
12 * 2 of the License, or (at your option) any later version.
16 * Changes:
18 * Janos Farkas : delete timer on ifdown
19 * <chexum@bankinf.banki.hu>
20 * Andi Kleen : kill double kfree on module
21 * unload.
22 * Maciej W. Rozycki : FDDI support
23 * sekiya@USAGI : Don't send too many RS
24 * packets.
25 * yoshfuji@USAGI : Fixed interval between DAD
26 * packets.
27 * YOSHIFUJI Hideaki @USAGI : improved accuracy of
28 * address validation timer.
29 * YOSHIFUJI Hideaki @USAGI : Privacy Extensions (RFC3041)
30 * support.
31 * Yuji SEKIYA @USAGI : Don't assign a same IPv6
32 * address on a same interface.
33 * YOSHIFUJI Hideaki @USAGI : ARCnet support
34 * YOSHIFUJI Hideaki @USAGI : convert /proc/net/if_inet6 to
35 * seq_file.
36 * YOSHIFUJI Hideaki @USAGI : improved source address
37 * selection; consider scope,
38 * status etc.
41 #include <linux/errno.h>
42 #include <linux/types.h>
43 #include <linux/kernel.h>
44 #include <linux/socket.h>
45 #include <linux/sockios.h>
46 #include <linux/net.h>
47 #include <linux/in6.h>
48 #include <linux/netdevice.h>
49 #include <linux/if_addr.h>
50 #include <linux/if_arp.h>
51 #include <linux/if_arcnet.h>
52 #include <linux/if_infiniband.h>
53 #include <linux/route.h>
54 #include <linux/inetdevice.h>
55 #include <linux/init.h>
56 #ifdef CONFIG_SYSCTL
57 #include <linux/sysctl.h>
58 #endif
59 #include <linux/capability.h>
60 #include <linux/delay.h>
61 #include <linux/notifier.h>
62 #include <linux/string.h>
64 #include <net/net_namespace.h>
65 #include <net/sock.h>
66 #include <net/snmp.h>
68 #include <net/ipv6.h>
69 #include <net/protocol.h>
70 #include <net/ndisc.h>
71 #include <net/ip6_route.h>
72 #include <net/addrconf.h>
73 #include <net/tcp.h>
74 #include <net/ip.h>
75 #include <net/netlink.h>
76 #include <net/pkt_sched.h>
77 #include <linux/if_tunnel.h>
78 #include <linux/rtnetlink.h>
80 #ifdef CONFIG_IPV6_PRIVACY
81 #include <linux/random.h>
82 #endif
84 #include <asm/uaccess.h>
85 #include <asm/unaligned.h>
87 #include <linux/proc_fs.h>
88 #include <linux/seq_file.h>
90 /* Set to 3 to get tracing... */
91 #define ACONF_DEBUG 2
93 #if ACONF_DEBUG >= 3
94 #define ADBG(x) printk x
95 #else
96 #define ADBG(x)
97 #endif
99 #define INFINITY_LIFE_TIME 0xFFFFFFFF
100 #define TIME_DELTA(a,b) ((unsigned long)((long)(a) - (long)(b)))
102 #ifdef CONFIG_SYSCTL
103 static void addrconf_sysctl_register(struct inet6_dev *idev);
104 static void addrconf_sysctl_unregister(struct inet6_dev *idev);
105 #else
106 static inline void addrconf_sysctl_register(struct inet6_dev *idev)
110 static inline void addrconf_sysctl_unregister(struct inet6_dev *idev)
113 #endif
115 #ifdef CONFIG_IPV6_PRIVACY
116 static int __ipv6_regen_rndid(struct inet6_dev *idev);
117 static int __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr);
118 static void ipv6_regen_rndid(unsigned long data);
120 static int desync_factor = MAX_DESYNC_FACTOR * HZ;
121 #endif
123 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev);
124 static int ipv6_count_addresses(struct inet6_dev *idev);
127 * Configured unicast address hash table
129 static struct inet6_ifaddr *inet6_addr_lst[IN6_ADDR_HSIZE];
130 static DEFINE_RWLOCK(addrconf_hash_lock);
132 static void addrconf_verify(unsigned long);
134 static DEFINE_TIMER(addr_chk_timer, addrconf_verify, 0, 0);
135 static DEFINE_SPINLOCK(addrconf_verify_lock);
137 static void addrconf_join_anycast(struct inet6_ifaddr *ifp);
138 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp);
140 static int addrconf_ifdown(struct net_device *dev, int how);
142 static void addrconf_dad_start(struct inet6_ifaddr *ifp, u32 flags);
143 static void addrconf_dad_timer(unsigned long data);
144 static void addrconf_dad_completed(struct inet6_ifaddr *ifp);
145 static void addrconf_dad_run(struct inet6_dev *idev);
146 static void addrconf_rs_timer(unsigned long data);
147 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
148 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
150 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
151 struct prefix_info *pinfo);
152 static int ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
153 struct net_device *dev);
155 static ATOMIC_NOTIFIER_HEAD(inet6addr_chain);
157 static struct ipv6_devconf ipv6_devconf __read_mostly = {
158 .forwarding = 0,
159 .hop_limit = IPV6_DEFAULT_HOPLIMIT,
160 .mtu6 = IPV6_MIN_MTU,
161 .accept_ra = 1,
162 .accept_redirects = 1,
163 .autoconf = 1,
164 .force_mld_version = 0,
165 .dad_transmits = 1,
166 .rtr_solicits = MAX_RTR_SOLICITATIONS,
167 .rtr_solicit_interval = RTR_SOLICITATION_INTERVAL,
168 .rtr_solicit_delay = MAX_RTR_SOLICITATION_DELAY,
169 #ifdef CONFIG_IPV6_PRIVACY
170 .use_tempaddr = 0,
171 .temp_valid_lft = TEMP_VALID_LIFETIME,
172 .temp_prefered_lft = TEMP_PREFERRED_LIFETIME,
173 .regen_max_retry = REGEN_MAX_RETRY,
174 .max_desync_factor = MAX_DESYNC_FACTOR,
175 #endif
176 .max_addresses = IPV6_MAX_ADDRESSES,
177 .accept_ra_defrtr = 1,
178 .accept_ra_pinfo = 1,
179 #ifdef CONFIG_IPV6_ROUTER_PREF
180 .accept_ra_rtr_pref = 1,
181 .rtr_probe_interval = 60 * HZ,
182 #ifdef CONFIG_IPV6_ROUTE_INFO
183 .accept_ra_rt_info_max_plen = 0,
184 #endif
185 #endif
186 .proxy_ndp = 0,
187 .accept_source_route = 0, /* we do not accept RH0 by default. */
188 .disable_ipv6 = 0,
189 .accept_dad = 1,
192 static struct ipv6_devconf ipv6_devconf_dflt __read_mostly = {
193 .forwarding = 0,
194 .hop_limit = IPV6_DEFAULT_HOPLIMIT,
195 .mtu6 = IPV6_MIN_MTU,
196 .accept_ra = 1,
197 .accept_redirects = 1,
198 .autoconf = 1,
199 .dad_transmits = 1,
200 .rtr_solicits = MAX_RTR_SOLICITATIONS,
201 .rtr_solicit_interval = RTR_SOLICITATION_INTERVAL,
202 .rtr_solicit_delay = MAX_RTR_SOLICITATION_DELAY,
203 #ifdef CONFIG_IPV6_PRIVACY
204 .use_tempaddr = 0,
205 .temp_valid_lft = TEMP_VALID_LIFETIME,
206 .temp_prefered_lft = TEMP_PREFERRED_LIFETIME,
207 .regen_max_retry = REGEN_MAX_RETRY,
208 .max_desync_factor = MAX_DESYNC_FACTOR,
209 #endif
210 .max_addresses = IPV6_MAX_ADDRESSES,
211 .accept_ra_defrtr = 1,
212 .accept_ra_pinfo = 1,
213 #ifdef CONFIG_IPV6_ROUTER_PREF
214 .accept_ra_rtr_pref = 1,
215 .rtr_probe_interval = 60 * HZ,
216 #ifdef CONFIG_IPV6_ROUTE_INFO
217 .accept_ra_rt_info_max_plen = 0,
218 #endif
219 #endif
220 .proxy_ndp = 0,
221 .accept_source_route = 0, /* we do not accept RH0 by default. */
222 .disable_ipv6 = 0,
223 .accept_dad = 1,
226 /* IPv6 Wildcard Address and Loopback Address defined by RFC2553 */
227 const struct in6_addr in6addr_any = IN6ADDR_ANY_INIT;
228 const struct in6_addr in6addr_loopback = IN6ADDR_LOOPBACK_INIT;
229 const struct in6_addr in6addr_linklocal_allnodes = IN6ADDR_LINKLOCAL_ALLNODES_INIT;
230 const struct in6_addr in6addr_linklocal_allrouters = IN6ADDR_LINKLOCAL_ALLROUTERS_INIT;
232 /* Check if a valid qdisc is available */
233 static inline bool addrconf_qdisc_ok(const struct net_device *dev)
235 return !qdisc_tx_is_noop(dev);
238 /* Check if a route is valid prefix route */
239 static inline int addrconf_is_prefix_route(const struct rt6_info *rt)
241 return ((rt->rt6i_flags & (RTF_GATEWAY | RTF_DEFAULT)) == 0);
244 static void addrconf_del_timer(struct inet6_ifaddr *ifp)
246 if (del_timer(&ifp->timer))
247 __in6_ifa_put(ifp);
250 enum addrconf_timer_t
252 AC_NONE,
253 AC_DAD,
254 AC_RS,
257 static void addrconf_mod_timer(struct inet6_ifaddr *ifp,
258 enum addrconf_timer_t what,
259 unsigned long when)
261 if (!del_timer(&ifp->timer))
262 in6_ifa_hold(ifp);
264 switch (what) {
265 case AC_DAD:
266 ifp->timer.function = addrconf_dad_timer;
267 break;
268 case AC_RS:
269 ifp->timer.function = addrconf_rs_timer;
270 break;
271 default:;
273 ifp->timer.expires = jiffies + when;
274 add_timer(&ifp->timer);
277 static int snmp6_alloc_dev(struct inet6_dev *idev)
279 if (snmp_mib_init((void **)idev->stats.ipv6,
280 sizeof(struct ipstats_mib)) < 0)
281 goto err_ip;
282 if (snmp_mib_init((void **)idev->stats.icmpv6,
283 sizeof(struct icmpv6_mib)) < 0)
284 goto err_icmp;
285 if (snmp_mib_init((void **)idev->stats.icmpv6msg,
286 sizeof(struct icmpv6msg_mib)) < 0)
287 goto err_icmpmsg;
289 return 0;
291 err_icmpmsg:
292 snmp_mib_free((void **)idev->stats.icmpv6);
293 err_icmp:
294 snmp_mib_free((void **)idev->stats.ipv6);
295 err_ip:
296 return -ENOMEM;
299 static void snmp6_free_dev(struct inet6_dev *idev)
301 snmp_mib_free((void **)idev->stats.icmpv6msg);
302 snmp_mib_free((void **)idev->stats.icmpv6);
303 snmp_mib_free((void **)idev->stats.ipv6);
306 /* Nobody refers to this device, we may destroy it. */
308 static void in6_dev_finish_destroy_rcu(struct rcu_head *head)
310 struct inet6_dev *idev = container_of(head, struct inet6_dev, rcu);
311 kfree(idev);
314 void in6_dev_finish_destroy(struct inet6_dev *idev)
316 struct net_device *dev = idev->dev;
318 WARN_ON(idev->addr_list != NULL);
319 WARN_ON(idev->mc_list != NULL);
321 #ifdef NET_REFCNT_DEBUG
322 printk(KERN_DEBUG "in6_dev_finish_destroy: %s\n", dev ? dev->name : "NIL");
323 #endif
324 dev_put(dev);
325 if (!idev->dead) {
326 printk("Freeing alive inet6 device %p\n", idev);
327 return;
329 snmp6_free_dev(idev);
330 call_rcu(&idev->rcu, in6_dev_finish_destroy_rcu);
333 EXPORT_SYMBOL(in6_dev_finish_destroy);
335 static struct inet6_dev * ipv6_add_dev(struct net_device *dev)
337 struct inet6_dev *ndev;
339 ASSERT_RTNL();
341 if (dev->mtu < IPV6_MIN_MTU)
342 return NULL;
344 ndev = kzalloc(sizeof(struct inet6_dev), GFP_KERNEL);
346 if (ndev == NULL)
347 return NULL;
349 rwlock_init(&ndev->lock);
350 ndev->dev = dev;
351 memcpy(&ndev->cnf, dev_net(dev)->ipv6.devconf_dflt, sizeof(ndev->cnf));
352 ndev->cnf.mtu6 = dev->mtu;
353 ndev->cnf.sysctl = NULL;
354 ndev->nd_parms = neigh_parms_alloc(dev, &nd_tbl);
355 if (ndev->nd_parms == NULL) {
356 kfree(ndev);
357 return NULL;
359 if (ndev->cnf.forwarding)
360 dev_disable_lro(dev);
361 /* We refer to the device */
362 dev_hold(dev);
364 if (snmp6_alloc_dev(ndev) < 0) {
365 ADBG((KERN_WARNING
366 "%s(): cannot allocate memory for statistics; dev=%s.\n",
367 __func__, dev->name));
368 neigh_parms_release(&nd_tbl, ndev->nd_parms);
369 ndev->dead = 1;
370 in6_dev_finish_destroy(ndev);
371 return NULL;
374 if (snmp6_register_dev(ndev) < 0) {
375 ADBG((KERN_WARNING
376 "%s(): cannot create /proc/net/dev_snmp6/%s\n",
377 __func__, dev->name));
378 neigh_parms_release(&nd_tbl, ndev->nd_parms);
379 ndev->dead = 1;
380 in6_dev_finish_destroy(ndev);
381 return NULL;
384 /* One reference from device. We must do this before
385 * we invoke __ipv6_regen_rndid().
387 in6_dev_hold(ndev);
389 if (dev->flags & (IFF_NOARP | IFF_LOOPBACK))
390 ndev->cnf.accept_dad = -1;
392 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
393 if (dev->type == ARPHRD_SIT && (dev->priv_flags & IFF_ISATAP)) {
394 printk(KERN_INFO
395 "%s: Disabled Multicast RS\n",
396 dev->name);
397 ndev->cnf.rtr_solicits = 0;
399 #endif
401 #ifdef CONFIG_IPV6_PRIVACY
402 setup_timer(&ndev->regen_timer, ipv6_regen_rndid, (unsigned long)ndev);
403 if ((dev->flags&IFF_LOOPBACK) ||
404 dev->type == ARPHRD_TUNNEL ||
405 dev->type == ARPHRD_TUNNEL6 ||
406 dev->type == ARPHRD_SIT ||
407 dev->type == ARPHRD_NONE) {
408 printk(KERN_INFO
409 "%s: Disabled Privacy Extensions\n",
410 dev->name);
411 ndev->cnf.use_tempaddr = -1;
412 } else {
413 in6_dev_hold(ndev);
414 ipv6_regen_rndid((unsigned long) ndev);
416 #endif
418 if (netif_running(dev) && addrconf_qdisc_ok(dev))
419 ndev->if_flags |= IF_READY;
421 ipv6_mc_init_dev(ndev);
422 ndev->tstamp = jiffies;
423 addrconf_sysctl_register(ndev);
424 /* protected by rtnl_lock */
425 rcu_assign_pointer(dev->ip6_ptr, ndev);
427 /* Join all-node multicast group */
428 ipv6_dev_mc_inc(dev, &in6addr_linklocal_allnodes);
430 return ndev;
433 static struct inet6_dev * ipv6_find_idev(struct net_device *dev)
435 struct inet6_dev *idev;
437 ASSERT_RTNL();
439 if ((idev = __in6_dev_get(dev)) == NULL) {
440 if ((idev = ipv6_add_dev(dev)) == NULL)
441 return NULL;
444 if (dev->flags&IFF_UP)
445 ipv6_mc_up(idev);
446 return idev;
449 #ifdef CONFIG_SYSCTL
450 static void dev_forward_change(struct inet6_dev *idev)
452 struct net_device *dev;
453 struct inet6_ifaddr *ifa;
455 if (!idev)
456 return;
457 dev = idev->dev;
458 if (idev->cnf.forwarding)
459 dev_disable_lro(dev);
460 if (dev && (dev->flags & IFF_MULTICAST)) {
461 if (idev->cnf.forwarding)
462 ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
463 else
464 ipv6_dev_mc_dec(dev, &in6addr_linklocal_allrouters);
466 for (ifa=idev->addr_list; ifa; ifa=ifa->if_next) {
467 if (ifa->flags&IFA_F_TENTATIVE)
468 continue;
469 if (idev->cnf.forwarding)
470 addrconf_join_anycast(ifa);
471 else
472 addrconf_leave_anycast(ifa);
477 static void addrconf_forward_change(struct net *net, __s32 newf)
479 struct net_device *dev;
480 struct inet6_dev *idev;
482 read_lock(&dev_base_lock);
483 for_each_netdev(net, dev) {
484 rcu_read_lock();
485 idev = __in6_dev_get(dev);
486 if (idev) {
487 int changed = (!idev->cnf.forwarding) ^ (!newf);
488 idev->cnf.forwarding = newf;
489 if (changed)
490 dev_forward_change(idev);
492 rcu_read_unlock();
494 read_unlock(&dev_base_lock);
497 static int addrconf_fixup_forwarding(struct ctl_table *table, int *p, int old)
499 struct net *net;
501 net = (struct net *)table->extra2;
502 if (p == &net->ipv6.devconf_dflt->forwarding)
503 return 0;
505 if (!rtnl_trylock())
506 return restart_syscall();
508 if (p == &net->ipv6.devconf_all->forwarding) {
509 __s32 newf = net->ipv6.devconf_all->forwarding;
510 net->ipv6.devconf_dflt->forwarding = newf;
511 addrconf_forward_change(net, newf);
512 } else if ((!*p) ^ (!old))
513 dev_forward_change((struct inet6_dev *)table->extra1);
514 rtnl_unlock();
516 if (*p)
517 rt6_purge_dflt_routers(net);
518 return 1;
520 #endif
522 /* Nobody refers to this ifaddr, destroy it */
524 void inet6_ifa_finish_destroy(struct inet6_ifaddr *ifp)
526 WARN_ON(ifp->if_next != NULL);
527 WARN_ON(ifp->lst_next != NULL);
529 #ifdef NET_REFCNT_DEBUG
530 printk(KERN_DEBUG "inet6_ifa_finish_destroy\n");
531 #endif
533 in6_dev_put(ifp->idev);
535 if (del_timer(&ifp->timer))
536 printk("Timer is still running, when freeing ifa=%p\n", ifp);
538 if (!ifp->dead) {
539 printk("Freeing alive inet6 address %p\n", ifp);
540 return;
542 dst_release(&ifp->rt->u.dst);
544 kfree(ifp);
547 static void
548 ipv6_link_dev_addr(struct inet6_dev *idev, struct inet6_ifaddr *ifp)
550 struct inet6_ifaddr *ifa, **ifap;
551 int ifp_scope = ipv6_addr_src_scope(&ifp->addr);
554 * Each device address list is sorted in order of scope -
555 * global before linklocal.
557 for (ifap = &idev->addr_list; (ifa = *ifap) != NULL;
558 ifap = &ifa->if_next) {
559 if (ifp_scope >= ipv6_addr_src_scope(&ifa->addr))
560 break;
563 ifp->if_next = *ifap;
564 *ifap = ifp;
568 * Hash function taken from net_alias.c
570 static u8 ipv6_addr_hash(const struct in6_addr *addr)
572 __u32 word;
575 * We perform the hash function over the last 64 bits of the address
576 * This will include the IEEE address token on links that support it.
579 word = (__force u32)(addr->s6_addr32[2] ^ addr->s6_addr32[3]);
580 word ^= (word >> 16);
581 word ^= (word >> 8);
583 return ((word ^ (word >> 4)) & 0x0f);
586 /* On success it returns ifp with increased reference count */
588 static struct inet6_ifaddr *
589 ipv6_add_addr(struct inet6_dev *idev, const struct in6_addr *addr, int pfxlen,
590 int scope, u32 flags)
592 struct inet6_ifaddr *ifa = NULL;
593 struct rt6_info *rt;
594 int hash;
595 int err = 0;
596 int addr_type = ipv6_addr_type(addr);
598 if (addr_type == IPV6_ADDR_ANY ||
599 addr_type & IPV6_ADDR_MULTICAST ||
600 (!(idev->dev->flags & IFF_LOOPBACK) &&
601 addr_type & IPV6_ADDR_LOOPBACK))
602 return ERR_PTR(-EADDRNOTAVAIL);
604 rcu_read_lock_bh();
605 if (idev->dead) {
606 err = -ENODEV; /*XXX*/
607 goto out2;
610 if (idev->cnf.disable_ipv6) {
611 err = -EACCES;
612 goto out2;
615 write_lock(&addrconf_hash_lock);
617 /* Ignore adding duplicate addresses on an interface */
618 if (ipv6_chk_same_addr(dev_net(idev->dev), addr, idev->dev)) {
619 ADBG(("ipv6_add_addr: already assigned\n"));
620 err = -EEXIST;
621 goto out;
624 ifa = kzalloc(sizeof(struct inet6_ifaddr), GFP_ATOMIC);
626 if (ifa == NULL) {
627 ADBG(("ipv6_add_addr: malloc failed\n"));
628 err = -ENOBUFS;
629 goto out;
632 rt = addrconf_dst_alloc(idev, addr, 0);
633 if (IS_ERR(rt)) {
634 err = PTR_ERR(rt);
635 goto out;
638 ipv6_addr_copy(&ifa->addr, addr);
640 spin_lock_init(&ifa->lock);
641 init_timer(&ifa->timer);
642 ifa->timer.data = (unsigned long) ifa;
643 ifa->scope = scope;
644 ifa->prefix_len = pfxlen;
645 ifa->flags = flags | IFA_F_TENTATIVE;
646 ifa->cstamp = ifa->tstamp = jiffies;
648 ifa->rt = rt;
651 * part one of RFC 4429, section 3.3
652 * We should not configure an address as
653 * optimistic if we do not yet know the link
654 * layer address of our nexhop router
657 if (rt->rt6i_nexthop == NULL)
658 ifa->flags &= ~IFA_F_OPTIMISTIC;
660 ifa->idev = idev;
661 in6_dev_hold(idev);
662 /* For caller */
663 in6_ifa_hold(ifa);
665 /* Add to big hash table */
666 hash = ipv6_addr_hash(addr);
668 ifa->lst_next = inet6_addr_lst[hash];
669 inet6_addr_lst[hash] = ifa;
670 in6_ifa_hold(ifa);
671 write_unlock(&addrconf_hash_lock);
673 write_lock(&idev->lock);
674 /* Add to inet6_dev unicast addr list. */
675 ipv6_link_dev_addr(idev, ifa);
677 #ifdef CONFIG_IPV6_PRIVACY
678 if (ifa->flags&IFA_F_TEMPORARY) {
679 ifa->tmp_next = idev->tempaddr_list;
680 idev->tempaddr_list = ifa;
681 in6_ifa_hold(ifa);
683 #endif
685 in6_ifa_hold(ifa);
686 write_unlock(&idev->lock);
687 out2:
688 rcu_read_unlock_bh();
690 if (likely(err == 0))
691 atomic_notifier_call_chain(&inet6addr_chain, NETDEV_UP, ifa);
692 else {
693 kfree(ifa);
694 ifa = ERR_PTR(err);
697 return ifa;
698 out:
699 write_unlock(&addrconf_hash_lock);
700 goto out2;
703 /* This function wants to get referenced ifp and releases it before return */
705 static void ipv6_del_addr(struct inet6_ifaddr *ifp)
707 struct inet6_ifaddr *ifa, **ifap;
708 struct inet6_dev *idev = ifp->idev;
709 int hash;
710 int deleted = 0, onlink = 0;
711 unsigned long expires = jiffies;
713 hash = ipv6_addr_hash(&ifp->addr);
715 ifp->dead = 1;
717 write_lock_bh(&addrconf_hash_lock);
718 for (ifap = &inet6_addr_lst[hash]; (ifa=*ifap) != NULL;
719 ifap = &ifa->lst_next) {
720 if (ifa == ifp) {
721 *ifap = ifa->lst_next;
722 __in6_ifa_put(ifp);
723 ifa->lst_next = NULL;
724 break;
727 write_unlock_bh(&addrconf_hash_lock);
729 write_lock_bh(&idev->lock);
730 #ifdef CONFIG_IPV6_PRIVACY
731 if (ifp->flags&IFA_F_TEMPORARY) {
732 for (ifap = &idev->tempaddr_list; (ifa=*ifap) != NULL;
733 ifap = &ifa->tmp_next) {
734 if (ifa == ifp) {
735 *ifap = ifa->tmp_next;
736 if (ifp->ifpub) {
737 in6_ifa_put(ifp->ifpub);
738 ifp->ifpub = NULL;
740 __in6_ifa_put(ifp);
741 ifa->tmp_next = NULL;
742 break;
746 #endif
748 for (ifap = &idev->addr_list; (ifa=*ifap) != NULL;) {
749 if (ifa == ifp) {
750 *ifap = ifa->if_next;
751 __in6_ifa_put(ifp);
752 ifa->if_next = NULL;
753 if (!(ifp->flags & IFA_F_PERMANENT) || onlink > 0)
754 break;
755 deleted = 1;
756 continue;
757 } else if (ifp->flags & IFA_F_PERMANENT) {
758 if (ipv6_prefix_equal(&ifa->addr, &ifp->addr,
759 ifp->prefix_len)) {
760 if (ifa->flags & IFA_F_PERMANENT) {
761 onlink = 1;
762 if (deleted)
763 break;
764 } else {
765 unsigned long lifetime;
767 if (!onlink)
768 onlink = -1;
770 spin_lock(&ifa->lock);
772 lifetime = addrconf_timeout_fixup(ifa->valid_lft, HZ);
774 * Note: Because this address is
775 * not permanent, lifetime <
776 * LONG_MAX / HZ here.
778 if (time_before(expires,
779 ifa->tstamp + lifetime * HZ))
780 expires = ifa->tstamp + lifetime * HZ;
781 spin_unlock(&ifa->lock);
785 ifap = &ifa->if_next;
787 write_unlock_bh(&idev->lock);
789 addrconf_del_timer(ifp);
791 ipv6_ifa_notify(RTM_DELADDR, ifp);
793 atomic_notifier_call_chain(&inet6addr_chain, NETDEV_DOWN, ifp);
796 * Purge or update corresponding prefix
798 * 1) we don't purge prefix here if address was not permanent.
799 * prefix is managed by its own lifetime.
800 * 2) if there're no addresses, delete prefix.
801 * 3) if there're still other permanent address(es),
802 * corresponding prefix is still permanent.
803 * 4) otherwise, update prefix lifetime to the
804 * longest valid lifetime among the corresponding
805 * addresses on the device.
806 * Note: subsequent RA will update lifetime.
808 * --yoshfuji
810 if ((ifp->flags & IFA_F_PERMANENT) && onlink < 1) {
811 struct in6_addr prefix;
812 struct rt6_info *rt;
813 struct net *net = dev_net(ifp->idev->dev);
814 ipv6_addr_prefix(&prefix, &ifp->addr, ifp->prefix_len);
815 rt = rt6_lookup(net, &prefix, NULL, ifp->idev->dev->ifindex, 1);
817 if (rt && addrconf_is_prefix_route(rt)) {
818 if (onlink == 0) {
819 ip6_del_rt(rt);
820 rt = NULL;
821 } else if (!(rt->rt6i_flags & RTF_EXPIRES)) {
822 rt->rt6i_expires = expires;
823 rt->rt6i_flags |= RTF_EXPIRES;
826 dst_release(&rt->u.dst);
829 in6_ifa_put(ifp);
832 #ifdef CONFIG_IPV6_PRIVACY
833 static int ipv6_create_tempaddr(struct inet6_ifaddr *ifp, struct inet6_ifaddr *ift)
835 struct inet6_dev *idev = ifp->idev;
836 struct in6_addr addr, *tmpaddr;
837 unsigned long tmp_prefered_lft, tmp_valid_lft, tmp_cstamp, tmp_tstamp;
838 unsigned long regen_advance;
839 int tmp_plen;
840 int ret = 0;
841 int max_addresses;
842 u32 addr_flags;
844 write_lock(&idev->lock);
845 if (ift) {
846 spin_lock_bh(&ift->lock);
847 memcpy(&addr.s6_addr[8], &ift->addr.s6_addr[8], 8);
848 spin_unlock_bh(&ift->lock);
849 tmpaddr = &addr;
850 } else {
851 tmpaddr = NULL;
853 retry:
854 in6_dev_hold(idev);
855 if (idev->cnf.use_tempaddr <= 0) {
856 write_unlock(&idev->lock);
857 printk(KERN_INFO
858 "ipv6_create_tempaddr(): use_tempaddr is disabled.\n");
859 in6_dev_put(idev);
860 ret = -1;
861 goto out;
863 spin_lock_bh(&ifp->lock);
864 if (ifp->regen_count++ >= idev->cnf.regen_max_retry) {
865 idev->cnf.use_tempaddr = -1; /*XXX*/
866 spin_unlock_bh(&ifp->lock);
867 write_unlock(&idev->lock);
868 printk(KERN_WARNING
869 "ipv6_create_tempaddr(): regeneration time exceeded. disabled temporary address support.\n");
870 in6_dev_put(idev);
871 ret = -1;
872 goto out;
874 in6_ifa_hold(ifp);
875 memcpy(addr.s6_addr, ifp->addr.s6_addr, 8);
876 if (__ipv6_try_regen_rndid(idev, tmpaddr) < 0) {
877 spin_unlock_bh(&ifp->lock);
878 write_unlock(&idev->lock);
879 printk(KERN_WARNING
880 "ipv6_create_tempaddr(): regeneration of randomized interface id failed.\n");
881 in6_ifa_put(ifp);
882 in6_dev_put(idev);
883 ret = -1;
884 goto out;
886 memcpy(&addr.s6_addr[8], idev->rndid, 8);
887 tmp_valid_lft = min_t(__u32,
888 ifp->valid_lft,
889 idev->cnf.temp_valid_lft);
890 tmp_prefered_lft = min_t(__u32,
891 ifp->prefered_lft,
892 idev->cnf.temp_prefered_lft - desync_factor / HZ);
893 tmp_plen = ifp->prefix_len;
894 max_addresses = idev->cnf.max_addresses;
895 tmp_cstamp = ifp->cstamp;
896 tmp_tstamp = ifp->tstamp;
897 spin_unlock_bh(&ifp->lock);
899 regen_advance = idev->cnf.regen_max_retry *
900 idev->cnf.dad_transmits *
901 idev->nd_parms->retrans_time / HZ;
902 write_unlock(&idev->lock);
904 /* A temporary address is created only if this calculated Preferred
905 * Lifetime is greater than REGEN_ADVANCE time units. In particular,
906 * an implementation must not create a temporary address with a zero
907 * Preferred Lifetime.
909 if (tmp_prefered_lft <= regen_advance) {
910 in6_ifa_put(ifp);
911 in6_dev_put(idev);
912 ret = -1;
913 goto out;
916 addr_flags = IFA_F_TEMPORARY;
917 /* set in addrconf_prefix_rcv() */
918 if (ifp->flags & IFA_F_OPTIMISTIC)
919 addr_flags |= IFA_F_OPTIMISTIC;
921 ift = !max_addresses ||
922 ipv6_count_addresses(idev) < max_addresses ?
923 ipv6_add_addr(idev, &addr, tmp_plen,
924 ipv6_addr_type(&addr)&IPV6_ADDR_SCOPE_MASK,
925 addr_flags) : NULL;
926 if (!ift || IS_ERR(ift)) {
927 in6_ifa_put(ifp);
928 in6_dev_put(idev);
929 printk(KERN_INFO
930 "ipv6_create_tempaddr(): retry temporary address regeneration.\n");
931 tmpaddr = &addr;
932 write_lock(&idev->lock);
933 goto retry;
936 spin_lock_bh(&ift->lock);
937 ift->ifpub = ifp;
938 ift->valid_lft = tmp_valid_lft;
939 ift->prefered_lft = tmp_prefered_lft;
940 ift->cstamp = tmp_cstamp;
941 ift->tstamp = tmp_tstamp;
942 spin_unlock_bh(&ift->lock);
944 addrconf_dad_start(ift, 0);
945 in6_ifa_put(ift);
946 in6_dev_put(idev);
947 out:
948 return ret;
950 #endif
953 * Choose an appropriate source address (RFC3484)
955 enum {
956 IPV6_SADDR_RULE_INIT = 0,
957 IPV6_SADDR_RULE_LOCAL,
958 IPV6_SADDR_RULE_SCOPE,
959 IPV6_SADDR_RULE_PREFERRED,
960 #ifdef CONFIG_IPV6_MIP6
961 IPV6_SADDR_RULE_HOA,
962 #endif
963 IPV6_SADDR_RULE_OIF,
964 IPV6_SADDR_RULE_LABEL,
965 #ifdef CONFIG_IPV6_PRIVACY
966 IPV6_SADDR_RULE_PRIVACY,
967 #endif
968 IPV6_SADDR_RULE_ORCHID,
969 IPV6_SADDR_RULE_PREFIX,
970 IPV6_SADDR_RULE_MAX
973 struct ipv6_saddr_score {
974 int rule;
975 int addr_type;
976 struct inet6_ifaddr *ifa;
977 DECLARE_BITMAP(scorebits, IPV6_SADDR_RULE_MAX);
978 int scopedist;
979 int matchlen;
982 struct ipv6_saddr_dst {
983 const struct in6_addr *addr;
984 int ifindex;
985 int scope;
986 int label;
987 unsigned int prefs;
990 static inline int ipv6_saddr_preferred(int type)
992 if (type & (IPV6_ADDR_MAPPED|IPV6_ADDR_COMPATv4|
993 IPV6_ADDR_LOOPBACK|IPV6_ADDR_RESERVED))
994 return 1;
995 return 0;
998 static int ipv6_get_saddr_eval(struct net *net,
999 struct ipv6_saddr_score *score,
1000 struct ipv6_saddr_dst *dst,
1001 int i)
1003 int ret;
1005 if (i <= score->rule) {
1006 switch (i) {
1007 case IPV6_SADDR_RULE_SCOPE:
1008 ret = score->scopedist;
1009 break;
1010 case IPV6_SADDR_RULE_PREFIX:
1011 ret = score->matchlen;
1012 break;
1013 default:
1014 ret = !!test_bit(i, score->scorebits);
1016 goto out;
1019 switch (i) {
1020 case IPV6_SADDR_RULE_INIT:
1021 /* Rule 0: remember if hiscore is not ready yet */
1022 ret = !!score->ifa;
1023 break;
1024 case IPV6_SADDR_RULE_LOCAL:
1025 /* Rule 1: Prefer same address */
1026 ret = ipv6_addr_equal(&score->ifa->addr, dst->addr);
1027 break;
1028 case IPV6_SADDR_RULE_SCOPE:
1029 /* Rule 2: Prefer appropriate scope
1031 * ret
1033 * -1 | d 15
1034 * ---+--+-+---> scope
1036 * | d is scope of the destination.
1037 * B-d | \
1038 * | \ <- smaller scope is better if
1039 * B-15 | \ if scope is enough for destinaion.
1040 * | ret = B - scope (-1 <= scope >= d <= 15).
1041 * d-C-1 | /
1042 * |/ <- greater is better
1043 * -C / if scope is not enough for destination.
1044 * /| ret = scope - C (-1 <= d < scope <= 15).
1046 * d - C - 1 < B -15 (for all -1 <= d <= 15).
1047 * C > d + 14 - B >= 15 + 14 - B = 29 - B.
1048 * Assume B = 0 and we get C > 29.
1050 ret = __ipv6_addr_src_scope(score->addr_type);
1051 if (ret >= dst->scope)
1052 ret = -ret;
1053 else
1054 ret -= 128; /* 30 is enough */
1055 score->scopedist = ret;
1056 break;
1057 case IPV6_SADDR_RULE_PREFERRED:
1058 /* Rule 3: Avoid deprecated and optimistic addresses */
1059 ret = ipv6_saddr_preferred(score->addr_type) ||
1060 !(score->ifa->flags & (IFA_F_DEPRECATED|IFA_F_OPTIMISTIC));
1061 break;
1062 #ifdef CONFIG_IPV6_MIP6
1063 case IPV6_SADDR_RULE_HOA:
1065 /* Rule 4: Prefer home address */
1066 int prefhome = !(dst->prefs & IPV6_PREFER_SRC_COA);
1067 ret = !(score->ifa->flags & IFA_F_HOMEADDRESS) ^ prefhome;
1068 break;
1070 #endif
1071 case IPV6_SADDR_RULE_OIF:
1072 /* Rule 5: Prefer outgoing interface */
1073 ret = (!dst->ifindex ||
1074 dst->ifindex == score->ifa->idev->dev->ifindex);
1075 break;
1076 case IPV6_SADDR_RULE_LABEL:
1077 /* Rule 6: Prefer matching label */
1078 ret = ipv6_addr_label(net,
1079 &score->ifa->addr, score->addr_type,
1080 score->ifa->idev->dev->ifindex) == dst->label;
1081 break;
1082 #ifdef CONFIG_IPV6_PRIVACY
1083 case IPV6_SADDR_RULE_PRIVACY:
1085 /* Rule 7: Prefer public address
1086 * Note: prefer temprary address if use_tempaddr >= 2
1088 int preftmp = dst->prefs & (IPV6_PREFER_SRC_PUBLIC|IPV6_PREFER_SRC_TMP) ?
1089 !!(dst->prefs & IPV6_PREFER_SRC_TMP) :
1090 score->ifa->idev->cnf.use_tempaddr >= 2;
1091 ret = (!(score->ifa->flags & IFA_F_TEMPORARY)) ^ preftmp;
1092 break;
1094 #endif
1095 case IPV6_SADDR_RULE_ORCHID:
1096 /* Rule 8-: Prefer ORCHID vs ORCHID or
1097 * non-ORCHID vs non-ORCHID
1099 ret = !(ipv6_addr_orchid(&score->ifa->addr) ^
1100 ipv6_addr_orchid(dst->addr));
1101 break;
1102 case IPV6_SADDR_RULE_PREFIX:
1103 /* Rule 8: Use longest matching prefix */
1104 score->matchlen = ret = ipv6_addr_diff(&score->ifa->addr,
1105 dst->addr);
1106 break;
1107 default:
1108 ret = 0;
1111 if (ret)
1112 __set_bit(i, score->scorebits);
1113 score->rule = i;
1114 out:
1115 return ret;
1118 int ipv6_dev_get_saddr(struct net *net, struct net_device *dst_dev,
1119 const struct in6_addr *daddr, unsigned int prefs,
1120 struct in6_addr *saddr)
1122 struct ipv6_saddr_score scores[2],
1123 *score = &scores[0], *hiscore = &scores[1];
1124 struct ipv6_saddr_dst dst;
1125 struct net_device *dev;
1126 int dst_type;
1128 dst_type = __ipv6_addr_type(daddr);
1129 dst.addr = daddr;
1130 dst.ifindex = dst_dev ? dst_dev->ifindex : 0;
1131 dst.scope = __ipv6_addr_src_scope(dst_type);
1132 dst.label = ipv6_addr_label(net, daddr, dst_type, dst.ifindex);
1133 dst.prefs = prefs;
1135 hiscore->rule = -1;
1136 hiscore->ifa = NULL;
1138 read_lock(&dev_base_lock);
1139 rcu_read_lock();
1141 for_each_netdev(net, dev) {
1142 struct inet6_dev *idev;
1144 /* Candidate Source Address (section 4)
1145 * - multicast and link-local destination address,
1146 * the set of candidate source address MUST only
1147 * include addresses assigned to interfaces
1148 * belonging to the same link as the outgoing
1149 * interface.
1150 * (- For site-local destination addresses, the
1151 * set of candidate source addresses MUST only
1152 * include addresses assigned to interfaces
1153 * belonging to the same site as the outgoing
1154 * interface.)
1156 if (((dst_type & IPV6_ADDR_MULTICAST) ||
1157 dst.scope <= IPV6_ADDR_SCOPE_LINKLOCAL) &&
1158 dst.ifindex && dev->ifindex != dst.ifindex)
1159 continue;
1161 idev = __in6_dev_get(dev);
1162 if (!idev)
1163 continue;
1165 read_lock_bh(&idev->lock);
1166 for (score->ifa = idev->addr_list; score->ifa; score->ifa = score->ifa->if_next) {
1167 int i;
1170 * - Tentative Address (RFC2462 section 5.4)
1171 * - A tentative address is not considered
1172 * "assigned to an interface" in the traditional
1173 * sense, unless it is also flagged as optimistic.
1174 * - Candidate Source Address (section 4)
1175 * - In any case, anycast addresses, multicast
1176 * addresses, and the unspecified address MUST
1177 * NOT be included in a candidate set.
1179 if ((score->ifa->flags & IFA_F_TENTATIVE) &&
1180 (!(score->ifa->flags & IFA_F_OPTIMISTIC)))
1181 continue;
1183 score->addr_type = __ipv6_addr_type(&score->ifa->addr);
1185 if (unlikely(score->addr_type == IPV6_ADDR_ANY ||
1186 score->addr_type & IPV6_ADDR_MULTICAST)) {
1187 LIMIT_NETDEBUG(KERN_DEBUG
1188 "ADDRCONF: unspecified / multicast address "
1189 "assigned as unicast address on %s",
1190 dev->name);
1191 continue;
1194 score->rule = -1;
1195 bitmap_zero(score->scorebits, IPV6_SADDR_RULE_MAX);
1197 for (i = 0; i < IPV6_SADDR_RULE_MAX; i++) {
1198 int minihiscore, miniscore;
1200 minihiscore = ipv6_get_saddr_eval(net, hiscore, &dst, i);
1201 miniscore = ipv6_get_saddr_eval(net, score, &dst, i);
1203 if (minihiscore > miniscore) {
1204 if (i == IPV6_SADDR_RULE_SCOPE &&
1205 score->scopedist > 0) {
1207 * special case:
1208 * each remaining entry
1209 * has too small (not enough)
1210 * scope, because ifa entries
1211 * are sorted by their scope
1212 * values.
1214 goto try_nextdev;
1216 break;
1217 } else if (minihiscore < miniscore) {
1218 if (hiscore->ifa)
1219 in6_ifa_put(hiscore->ifa);
1221 in6_ifa_hold(score->ifa);
1223 swap(hiscore, score);
1225 /* restore our iterator */
1226 score->ifa = hiscore->ifa;
1228 break;
1232 try_nextdev:
1233 read_unlock_bh(&idev->lock);
1235 rcu_read_unlock();
1236 read_unlock(&dev_base_lock);
1238 if (!hiscore->ifa)
1239 return -EADDRNOTAVAIL;
1241 ipv6_addr_copy(saddr, &hiscore->ifa->addr);
1242 in6_ifa_put(hiscore->ifa);
1243 return 0;
1246 EXPORT_SYMBOL(ipv6_dev_get_saddr);
1248 int ipv6_get_lladdr(struct net_device *dev, struct in6_addr *addr,
1249 unsigned char banned_flags)
1251 struct inet6_dev *idev;
1252 int err = -EADDRNOTAVAIL;
1254 rcu_read_lock();
1255 if ((idev = __in6_dev_get(dev)) != NULL) {
1256 struct inet6_ifaddr *ifp;
1258 read_lock_bh(&idev->lock);
1259 for (ifp=idev->addr_list; ifp; ifp=ifp->if_next) {
1260 if (ifp->scope == IFA_LINK && !(ifp->flags & banned_flags)) {
1261 ipv6_addr_copy(addr, &ifp->addr);
1262 err = 0;
1263 break;
1266 read_unlock_bh(&idev->lock);
1268 rcu_read_unlock();
1269 return err;
1272 static int ipv6_count_addresses(struct inet6_dev *idev)
1274 int cnt = 0;
1275 struct inet6_ifaddr *ifp;
1277 read_lock_bh(&idev->lock);
1278 for (ifp=idev->addr_list; ifp; ifp=ifp->if_next)
1279 cnt++;
1280 read_unlock_bh(&idev->lock);
1281 return cnt;
1284 int ipv6_chk_addr(struct net *net, struct in6_addr *addr,
1285 struct net_device *dev, int strict)
1287 struct inet6_ifaddr * ifp;
1288 u8 hash = ipv6_addr_hash(addr);
1290 read_lock_bh(&addrconf_hash_lock);
1291 for(ifp = inet6_addr_lst[hash]; ifp; ifp=ifp->lst_next) {
1292 if (!net_eq(dev_net(ifp->idev->dev), net))
1293 continue;
1294 if (ipv6_addr_equal(&ifp->addr, addr) &&
1295 !(ifp->flags&IFA_F_TENTATIVE)) {
1296 if (dev == NULL || ifp->idev->dev == dev ||
1297 !(ifp->scope&(IFA_LINK|IFA_HOST) || strict))
1298 break;
1301 read_unlock_bh(&addrconf_hash_lock);
1302 return ifp != NULL;
1304 EXPORT_SYMBOL(ipv6_chk_addr);
1306 static
1307 int ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
1308 struct net_device *dev)
1310 struct inet6_ifaddr * ifp;
1311 u8 hash = ipv6_addr_hash(addr);
1313 for(ifp = inet6_addr_lst[hash]; ifp; ifp=ifp->lst_next) {
1314 if (!net_eq(dev_net(ifp->idev->dev), net))
1315 continue;
1316 if (ipv6_addr_equal(&ifp->addr, addr)) {
1317 if (dev == NULL || ifp->idev->dev == dev)
1318 break;
1321 return ifp != NULL;
1324 int ipv6_chk_prefix(struct in6_addr *addr, struct net_device *dev)
1326 struct inet6_dev *idev;
1327 struct inet6_ifaddr *ifa;
1328 int onlink;
1330 onlink = 0;
1331 rcu_read_lock();
1332 idev = __in6_dev_get(dev);
1333 if (idev) {
1334 read_lock_bh(&idev->lock);
1335 for (ifa = idev->addr_list; ifa; ifa = ifa->if_next) {
1336 onlink = ipv6_prefix_equal(addr, &ifa->addr,
1337 ifa->prefix_len);
1338 if (onlink)
1339 break;
1341 read_unlock_bh(&idev->lock);
1343 rcu_read_unlock();
1344 return onlink;
1347 EXPORT_SYMBOL(ipv6_chk_prefix);
1349 struct inet6_ifaddr *ipv6_get_ifaddr(struct net *net, const struct in6_addr *addr,
1350 struct net_device *dev, int strict)
1352 struct inet6_ifaddr * ifp;
1353 u8 hash = ipv6_addr_hash(addr);
1355 read_lock_bh(&addrconf_hash_lock);
1356 for(ifp = inet6_addr_lst[hash]; ifp; ifp=ifp->lst_next) {
1357 if (!net_eq(dev_net(ifp->idev->dev), net))
1358 continue;
1359 if (ipv6_addr_equal(&ifp->addr, addr)) {
1360 if (dev == NULL || ifp->idev->dev == dev ||
1361 !(ifp->scope&(IFA_LINK|IFA_HOST) || strict)) {
1362 in6_ifa_hold(ifp);
1363 break;
1367 read_unlock_bh(&addrconf_hash_lock);
1369 return ifp;
1372 /* Gets referenced address, destroys ifaddr */
1374 static void addrconf_dad_stop(struct inet6_ifaddr *ifp)
1376 if (ifp->flags&IFA_F_PERMANENT) {
1377 spin_lock_bh(&ifp->lock);
1378 addrconf_del_timer(ifp);
1379 ifp->flags |= IFA_F_TENTATIVE;
1380 spin_unlock_bh(&ifp->lock);
1381 in6_ifa_put(ifp);
1382 #ifdef CONFIG_IPV6_PRIVACY
1383 } else if (ifp->flags&IFA_F_TEMPORARY) {
1384 struct inet6_ifaddr *ifpub;
1385 spin_lock_bh(&ifp->lock);
1386 ifpub = ifp->ifpub;
1387 if (ifpub) {
1388 in6_ifa_hold(ifpub);
1389 spin_unlock_bh(&ifp->lock);
1390 ipv6_create_tempaddr(ifpub, ifp);
1391 in6_ifa_put(ifpub);
1392 } else {
1393 spin_unlock_bh(&ifp->lock);
1395 ipv6_del_addr(ifp);
1396 #endif
1397 } else
1398 ipv6_del_addr(ifp);
1401 void addrconf_dad_failure(struct inet6_ifaddr *ifp)
1403 struct inet6_dev *idev = ifp->idev;
1405 if (net_ratelimit())
1406 printk(KERN_INFO "%s: IPv6 duplicate address detected!\n",
1407 ifp->idev->dev->name);
1409 if (idev->cnf.accept_dad > 1 && !idev->cnf.disable_ipv6) {
1410 struct in6_addr addr;
1412 addr.s6_addr32[0] = htonl(0xfe800000);
1413 addr.s6_addr32[1] = 0;
1415 if (!ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) &&
1416 ipv6_addr_equal(&ifp->addr, &addr)) {
1417 /* DAD failed for link-local based on MAC address */
1418 idev->cnf.disable_ipv6 = 1;
1420 printk(KERN_INFO "%s: IPv6 being disabled!\n",
1421 ifp->idev->dev->name);
1425 addrconf_dad_stop(ifp);
1428 /* Join to solicited addr multicast group. */
1430 void addrconf_join_solict(struct net_device *dev, struct in6_addr *addr)
1432 struct in6_addr maddr;
1434 if (dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1435 return;
1437 addrconf_addr_solict_mult(addr, &maddr);
1438 ipv6_dev_mc_inc(dev, &maddr);
1441 void addrconf_leave_solict(struct inet6_dev *idev, struct in6_addr *addr)
1443 struct in6_addr maddr;
1445 if (idev->dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1446 return;
1448 addrconf_addr_solict_mult(addr, &maddr);
1449 __ipv6_dev_mc_dec(idev, &maddr);
1452 static void addrconf_join_anycast(struct inet6_ifaddr *ifp)
1454 struct in6_addr addr;
1455 ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1456 if (ipv6_addr_any(&addr))
1457 return;
1458 ipv6_dev_ac_inc(ifp->idev->dev, &addr);
1461 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp)
1463 struct in6_addr addr;
1464 ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1465 if (ipv6_addr_any(&addr))
1466 return;
1467 __ipv6_dev_ac_dec(ifp->idev, &addr);
1470 static int addrconf_ifid_eui48(u8 *eui, struct net_device *dev)
1472 if (dev->addr_len != ETH_ALEN)
1473 return -1;
1474 memcpy(eui, dev->dev_addr, 3);
1475 memcpy(eui + 5, dev->dev_addr + 3, 3);
1478 * The zSeries OSA network cards can be shared among various
1479 * OS instances, but the OSA cards have only one MAC address.
1480 * This leads to duplicate address conflicts in conjunction
1481 * with IPv6 if more than one instance uses the same card.
1483 * The driver for these cards can deliver a unique 16-bit
1484 * identifier for each instance sharing the same card. It is
1485 * placed instead of 0xFFFE in the interface identifier. The
1486 * "u" bit of the interface identifier is not inverted in this
1487 * case. Hence the resulting interface identifier has local
1488 * scope according to RFC2373.
1490 if (dev->dev_id) {
1491 eui[3] = (dev->dev_id >> 8) & 0xFF;
1492 eui[4] = dev->dev_id & 0xFF;
1493 } else {
1494 eui[3] = 0xFF;
1495 eui[4] = 0xFE;
1496 eui[0] ^= 2;
1498 return 0;
1501 static int addrconf_ifid_arcnet(u8 *eui, struct net_device *dev)
1503 /* XXX: inherit EUI-64 from other interface -- yoshfuji */
1504 if (dev->addr_len != ARCNET_ALEN)
1505 return -1;
1506 memset(eui, 0, 7);
1507 eui[7] = *(u8*)dev->dev_addr;
1508 return 0;
1511 static int addrconf_ifid_infiniband(u8 *eui, struct net_device *dev)
1513 if (dev->addr_len != INFINIBAND_ALEN)
1514 return -1;
1515 memcpy(eui, dev->dev_addr + 12, 8);
1516 eui[0] |= 2;
1517 return 0;
1520 int __ipv6_isatap_ifid(u8 *eui, __be32 addr)
1522 if (addr == 0)
1523 return -1;
1524 eui[0] = (ipv4_is_zeronet(addr) || ipv4_is_private_10(addr) ||
1525 ipv4_is_loopback(addr) || ipv4_is_linklocal_169(addr) ||
1526 ipv4_is_private_172(addr) || ipv4_is_test_192(addr) ||
1527 ipv4_is_anycast_6to4(addr) || ipv4_is_private_192(addr) ||
1528 ipv4_is_test_198(addr) || ipv4_is_multicast(addr) ||
1529 ipv4_is_lbcast(addr)) ? 0x00 : 0x02;
1530 eui[1] = 0;
1531 eui[2] = 0x5E;
1532 eui[3] = 0xFE;
1533 memcpy(eui + 4, &addr, 4);
1534 return 0;
1536 EXPORT_SYMBOL(__ipv6_isatap_ifid);
1538 static int addrconf_ifid_sit(u8 *eui, struct net_device *dev)
1540 if (dev->priv_flags & IFF_ISATAP)
1541 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
1542 return -1;
1545 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev)
1547 switch (dev->type) {
1548 case ARPHRD_ETHER:
1549 case ARPHRD_FDDI:
1550 case ARPHRD_IEEE802_TR:
1551 return addrconf_ifid_eui48(eui, dev);
1552 case ARPHRD_ARCNET:
1553 return addrconf_ifid_arcnet(eui, dev);
1554 case ARPHRD_INFINIBAND:
1555 return addrconf_ifid_infiniband(eui, dev);
1556 case ARPHRD_SIT:
1557 return addrconf_ifid_sit(eui, dev);
1559 return -1;
1562 static int ipv6_inherit_eui64(u8 *eui, struct inet6_dev *idev)
1564 int err = -1;
1565 struct inet6_ifaddr *ifp;
1567 read_lock_bh(&idev->lock);
1568 for (ifp=idev->addr_list; ifp; ifp=ifp->if_next) {
1569 if (ifp->scope == IFA_LINK && !(ifp->flags&IFA_F_TENTATIVE)) {
1570 memcpy(eui, ifp->addr.s6_addr+8, 8);
1571 err = 0;
1572 break;
1575 read_unlock_bh(&idev->lock);
1576 return err;
1579 #ifdef CONFIG_IPV6_PRIVACY
1580 /* (re)generation of randomized interface identifier (RFC 3041 3.2, 3.5) */
1581 static int __ipv6_regen_rndid(struct inet6_dev *idev)
1583 regen:
1584 get_random_bytes(idev->rndid, sizeof(idev->rndid));
1585 idev->rndid[0] &= ~0x02;
1588 * <draft-ietf-ipngwg-temp-addresses-v2-00.txt>:
1589 * check if generated address is not inappropriate
1591 * - Reserved subnet anycast (RFC 2526)
1592 * 11111101 11....11 1xxxxxxx
1593 * - ISATAP (RFC4214) 6.1
1594 * 00-00-5E-FE-xx-xx-xx-xx
1595 * - value 0
1596 * - XXX: already assigned to an address on the device
1598 if (idev->rndid[0] == 0xfd &&
1599 (idev->rndid[1]&idev->rndid[2]&idev->rndid[3]&idev->rndid[4]&idev->rndid[5]&idev->rndid[6]) == 0xff &&
1600 (idev->rndid[7]&0x80))
1601 goto regen;
1602 if ((idev->rndid[0]|idev->rndid[1]) == 0) {
1603 if (idev->rndid[2] == 0x5e && idev->rndid[3] == 0xfe)
1604 goto regen;
1605 if ((idev->rndid[2]|idev->rndid[3]|idev->rndid[4]|idev->rndid[5]|idev->rndid[6]|idev->rndid[7]) == 0x00)
1606 goto regen;
1609 return 0;
1612 static void ipv6_regen_rndid(unsigned long data)
1614 struct inet6_dev *idev = (struct inet6_dev *) data;
1615 unsigned long expires;
1617 rcu_read_lock_bh();
1618 write_lock_bh(&idev->lock);
1620 if (idev->dead)
1621 goto out;
1623 if (__ipv6_regen_rndid(idev) < 0)
1624 goto out;
1626 expires = jiffies +
1627 idev->cnf.temp_prefered_lft * HZ -
1628 idev->cnf.regen_max_retry * idev->cnf.dad_transmits * idev->nd_parms->retrans_time - desync_factor;
1629 if (time_before(expires, jiffies)) {
1630 printk(KERN_WARNING
1631 "ipv6_regen_rndid(): too short regeneration interval; timer disabled for %s.\n",
1632 idev->dev->name);
1633 goto out;
1636 if (!mod_timer(&idev->regen_timer, expires))
1637 in6_dev_hold(idev);
1639 out:
1640 write_unlock_bh(&idev->lock);
1641 rcu_read_unlock_bh();
1642 in6_dev_put(idev);
1645 static int __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr) {
1646 int ret = 0;
1648 if (tmpaddr && memcmp(idev->rndid, &tmpaddr->s6_addr[8], 8) == 0)
1649 ret = __ipv6_regen_rndid(idev);
1650 return ret;
1652 #endif
1655 * Add prefix route.
1658 static void
1659 addrconf_prefix_route(struct in6_addr *pfx, int plen, struct net_device *dev,
1660 unsigned long expires, u32 flags)
1662 struct fib6_config cfg = {
1663 .fc_table = RT6_TABLE_PREFIX,
1664 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1665 .fc_ifindex = dev->ifindex,
1666 .fc_expires = expires,
1667 .fc_dst_len = plen,
1668 .fc_flags = RTF_UP | flags,
1669 .fc_nlinfo.nl_net = dev_net(dev),
1670 .fc_protocol = RTPROT_KERNEL,
1673 ipv6_addr_copy(&cfg.fc_dst, pfx);
1675 /* Prevent useless cloning on PtP SIT.
1676 This thing is done here expecting that the whole
1677 class of non-broadcast devices need not cloning.
1679 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
1680 if (dev->type == ARPHRD_SIT && (dev->flags & IFF_POINTOPOINT))
1681 cfg.fc_flags |= RTF_NONEXTHOP;
1682 #endif
1684 ip6_route_add(&cfg);
1687 /* Create "default" multicast route to the interface */
1689 static void addrconf_add_mroute(struct net_device *dev)
1691 struct fib6_config cfg = {
1692 .fc_table = RT6_TABLE_LOCAL,
1693 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1694 .fc_ifindex = dev->ifindex,
1695 .fc_dst_len = 8,
1696 .fc_flags = RTF_UP,
1697 .fc_nlinfo.nl_net = dev_net(dev),
1700 ipv6_addr_set(&cfg.fc_dst, htonl(0xFF000000), 0, 0, 0);
1702 ip6_route_add(&cfg);
1705 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
1706 static void sit_route_add(struct net_device *dev)
1708 struct fib6_config cfg = {
1709 .fc_table = RT6_TABLE_MAIN,
1710 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1711 .fc_ifindex = dev->ifindex,
1712 .fc_dst_len = 96,
1713 .fc_flags = RTF_UP | RTF_NONEXTHOP,
1714 .fc_nlinfo.nl_net = dev_net(dev),
1717 /* prefix length - 96 bits "::d.d.d.d" */
1718 ip6_route_add(&cfg);
1720 #endif
1722 static void addrconf_add_lroute(struct net_device *dev)
1724 struct in6_addr addr;
1726 ipv6_addr_set(&addr, htonl(0xFE800000), 0, 0, 0);
1727 addrconf_prefix_route(&addr, 64, dev, 0, 0);
1730 static struct inet6_dev *addrconf_add_dev(struct net_device *dev)
1732 struct inet6_dev *idev;
1734 ASSERT_RTNL();
1736 if ((idev = ipv6_find_idev(dev)) == NULL)
1737 return NULL;
1739 /* Add default multicast route */
1740 addrconf_add_mroute(dev);
1742 /* Add link local route */
1743 addrconf_add_lroute(dev);
1744 return idev;
1747 void addrconf_prefix_rcv(struct net_device *dev, u8 *opt, int len)
1749 struct prefix_info *pinfo;
1750 __u32 valid_lft;
1751 __u32 prefered_lft;
1752 int addr_type;
1753 struct inet6_dev *in6_dev;
1754 struct net *net = dev_net(dev);
1756 pinfo = (struct prefix_info *) opt;
1758 if (len < sizeof(struct prefix_info)) {
1759 ADBG(("addrconf: prefix option too short\n"));
1760 return;
1764 * Validation checks ([ADDRCONF], page 19)
1767 addr_type = ipv6_addr_type(&pinfo->prefix);
1769 if (addr_type & (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL))
1770 return;
1772 valid_lft = ntohl(pinfo->valid);
1773 prefered_lft = ntohl(pinfo->prefered);
1775 if (prefered_lft > valid_lft) {
1776 if (net_ratelimit())
1777 printk(KERN_WARNING "addrconf: prefix option has invalid lifetime\n");
1778 return;
1781 in6_dev = in6_dev_get(dev);
1783 if (in6_dev == NULL) {
1784 if (net_ratelimit())
1785 printk(KERN_DEBUG "addrconf: device %s not configured\n", dev->name);
1786 return;
1790 * Two things going on here:
1791 * 1) Add routes for on-link prefixes
1792 * 2) Configure prefixes with the auto flag set
1795 if (pinfo->onlink) {
1796 struct rt6_info *rt;
1797 unsigned long rt_expires;
1799 /* Avoid arithmetic overflow. Really, we could
1800 * save rt_expires in seconds, likely valid_lft,
1801 * but it would require division in fib gc, that it
1802 * not good.
1804 if (HZ > USER_HZ)
1805 rt_expires = addrconf_timeout_fixup(valid_lft, HZ);
1806 else
1807 rt_expires = addrconf_timeout_fixup(valid_lft, USER_HZ);
1809 if (addrconf_finite_timeout(rt_expires))
1810 rt_expires *= HZ;
1812 rt = rt6_lookup(net, &pinfo->prefix, NULL,
1813 dev->ifindex, 1);
1815 if (rt && addrconf_is_prefix_route(rt)) {
1816 /* Autoconf prefix route */
1817 if (valid_lft == 0) {
1818 ip6_del_rt(rt);
1819 rt = NULL;
1820 } else if (addrconf_finite_timeout(rt_expires)) {
1821 /* not infinity */
1822 rt->rt6i_expires = jiffies + rt_expires;
1823 rt->rt6i_flags |= RTF_EXPIRES;
1824 } else {
1825 rt->rt6i_flags &= ~RTF_EXPIRES;
1826 rt->rt6i_expires = 0;
1828 } else if (valid_lft) {
1829 clock_t expires = 0;
1830 int flags = RTF_ADDRCONF | RTF_PREFIX_RT;
1831 if (addrconf_finite_timeout(rt_expires)) {
1832 /* not infinity */
1833 flags |= RTF_EXPIRES;
1834 expires = jiffies_to_clock_t(rt_expires);
1836 addrconf_prefix_route(&pinfo->prefix, pinfo->prefix_len,
1837 dev, expires, flags);
1839 if (rt)
1840 dst_release(&rt->u.dst);
1843 /* Try to figure out our local address for this prefix */
1845 if (pinfo->autoconf && in6_dev->cnf.autoconf) {
1846 struct inet6_ifaddr * ifp;
1847 struct in6_addr addr;
1848 int create = 0, update_lft = 0;
1850 if (pinfo->prefix_len == 64) {
1851 memcpy(&addr, &pinfo->prefix, 8);
1852 if (ipv6_generate_eui64(addr.s6_addr + 8, dev) &&
1853 ipv6_inherit_eui64(addr.s6_addr + 8, in6_dev)) {
1854 in6_dev_put(in6_dev);
1855 return;
1857 goto ok;
1859 if (net_ratelimit())
1860 printk(KERN_DEBUG "IPv6 addrconf: prefix with wrong length %d\n",
1861 pinfo->prefix_len);
1862 in6_dev_put(in6_dev);
1863 return;
1867 ifp = ipv6_get_ifaddr(net, &addr, dev, 1);
1869 if (ifp == NULL && valid_lft) {
1870 int max_addresses = in6_dev->cnf.max_addresses;
1871 u32 addr_flags = 0;
1873 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1874 if (in6_dev->cnf.optimistic_dad &&
1875 !net->ipv6.devconf_all->forwarding)
1876 addr_flags = IFA_F_OPTIMISTIC;
1877 #endif
1879 /* Do not allow to create too much of autoconfigured
1880 * addresses; this would be too easy way to crash kernel.
1882 if (!max_addresses ||
1883 ipv6_count_addresses(in6_dev) < max_addresses)
1884 ifp = ipv6_add_addr(in6_dev, &addr, pinfo->prefix_len,
1885 addr_type&IPV6_ADDR_SCOPE_MASK,
1886 addr_flags);
1888 if (!ifp || IS_ERR(ifp)) {
1889 in6_dev_put(in6_dev);
1890 return;
1893 update_lft = create = 1;
1894 ifp->cstamp = jiffies;
1895 addrconf_dad_start(ifp, RTF_ADDRCONF|RTF_PREFIX_RT);
1898 if (ifp) {
1899 int flags;
1900 unsigned long now;
1901 #ifdef CONFIG_IPV6_PRIVACY
1902 struct inet6_ifaddr *ift;
1903 #endif
1904 u32 stored_lft;
1906 /* update lifetime (RFC2462 5.5.3 e) */
1907 spin_lock(&ifp->lock);
1908 now = jiffies;
1909 if (ifp->valid_lft > (now - ifp->tstamp) / HZ)
1910 stored_lft = ifp->valid_lft - (now - ifp->tstamp) / HZ;
1911 else
1912 stored_lft = 0;
1913 if (!update_lft && stored_lft) {
1914 if (valid_lft > MIN_VALID_LIFETIME ||
1915 valid_lft > stored_lft)
1916 update_lft = 1;
1917 else if (stored_lft <= MIN_VALID_LIFETIME) {
1918 /* valid_lft <= stored_lft is always true */
1919 /* XXX: IPsec */
1920 update_lft = 0;
1921 } else {
1922 valid_lft = MIN_VALID_LIFETIME;
1923 if (valid_lft < prefered_lft)
1924 prefered_lft = valid_lft;
1925 update_lft = 1;
1929 if (update_lft) {
1930 ifp->valid_lft = valid_lft;
1931 ifp->prefered_lft = prefered_lft;
1932 ifp->tstamp = now;
1933 flags = ifp->flags;
1934 ifp->flags &= ~IFA_F_DEPRECATED;
1935 spin_unlock(&ifp->lock);
1937 if (!(flags&IFA_F_TENTATIVE))
1938 ipv6_ifa_notify(0, ifp);
1939 } else
1940 spin_unlock(&ifp->lock);
1942 #ifdef CONFIG_IPV6_PRIVACY
1943 read_lock_bh(&in6_dev->lock);
1944 /* update all temporary addresses in the list */
1945 for (ift=in6_dev->tempaddr_list; ift; ift=ift->tmp_next) {
1947 * When adjusting the lifetimes of an existing
1948 * temporary address, only lower the lifetimes.
1949 * Implementations must not increase the
1950 * lifetimes of an existing temporary address
1951 * when processing a Prefix Information Option.
1953 if (ifp != ift->ifpub)
1954 continue;
1956 spin_lock(&ift->lock);
1957 flags = ift->flags;
1958 if (ift->valid_lft > valid_lft &&
1959 ift->valid_lft - valid_lft > (jiffies - ift->tstamp) / HZ)
1960 ift->valid_lft = valid_lft + (jiffies - ift->tstamp) / HZ;
1961 if (ift->prefered_lft > prefered_lft &&
1962 ift->prefered_lft - prefered_lft > (jiffies - ift->tstamp) / HZ)
1963 ift->prefered_lft = prefered_lft + (jiffies - ift->tstamp) / HZ;
1964 spin_unlock(&ift->lock);
1965 if (!(flags&IFA_F_TENTATIVE))
1966 ipv6_ifa_notify(0, ift);
1969 if (create && in6_dev->cnf.use_tempaddr > 0) {
1971 * When a new public address is created as described in [ADDRCONF],
1972 * also create a new temporary address.
1974 read_unlock_bh(&in6_dev->lock);
1975 ipv6_create_tempaddr(ifp, NULL);
1976 } else {
1977 read_unlock_bh(&in6_dev->lock);
1979 #endif
1980 in6_ifa_put(ifp);
1981 addrconf_verify(0);
1984 inet6_prefix_notify(RTM_NEWPREFIX, in6_dev, pinfo);
1985 in6_dev_put(in6_dev);
1989 * Set destination address.
1990 * Special case for SIT interfaces where we create a new "virtual"
1991 * device.
1993 int addrconf_set_dstaddr(struct net *net, void __user *arg)
1995 struct in6_ifreq ireq;
1996 struct net_device *dev;
1997 int err = -EINVAL;
1999 rtnl_lock();
2001 err = -EFAULT;
2002 if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2003 goto err_exit;
2005 dev = __dev_get_by_index(net, ireq.ifr6_ifindex);
2007 err = -ENODEV;
2008 if (dev == NULL)
2009 goto err_exit;
2011 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2012 if (dev->type == ARPHRD_SIT) {
2013 const struct net_device_ops *ops = dev->netdev_ops;
2014 struct ifreq ifr;
2015 struct ip_tunnel_parm p;
2017 err = -EADDRNOTAVAIL;
2018 if (!(ipv6_addr_type(&ireq.ifr6_addr) & IPV6_ADDR_COMPATv4))
2019 goto err_exit;
2021 memset(&p, 0, sizeof(p));
2022 p.iph.daddr = ireq.ifr6_addr.s6_addr32[3];
2023 p.iph.saddr = 0;
2024 p.iph.version = 4;
2025 p.iph.ihl = 5;
2026 p.iph.protocol = IPPROTO_IPV6;
2027 p.iph.ttl = 64;
2028 ifr.ifr_ifru.ifru_data = (__force void __user *)&p;
2030 if (ops->ndo_do_ioctl) {
2031 mm_segment_t oldfs = get_fs();
2033 set_fs(KERNEL_DS);
2034 err = ops->ndo_do_ioctl(dev, &ifr, SIOCADDTUNNEL);
2035 set_fs(oldfs);
2036 } else
2037 err = -EOPNOTSUPP;
2039 if (err == 0) {
2040 err = -ENOBUFS;
2041 dev = __dev_get_by_name(net, p.name);
2042 if (!dev)
2043 goto err_exit;
2044 err = dev_open(dev);
2047 #endif
2049 err_exit:
2050 rtnl_unlock();
2051 return err;
2055 * Manual configuration of address on an interface
2057 static int inet6_addr_add(struct net *net, int ifindex, struct in6_addr *pfx,
2058 unsigned int plen, __u8 ifa_flags, __u32 prefered_lft,
2059 __u32 valid_lft)
2061 struct inet6_ifaddr *ifp;
2062 struct inet6_dev *idev;
2063 struct net_device *dev;
2064 int scope;
2065 u32 flags;
2066 clock_t expires;
2067 unsigned long timeout;
2069 ASSERT_RTNL();
2071 if (plen > 128)
2072 return -EINVAL;
2074 /* check the lifetime */
2075 if (!valid_lft || prefered_lft > valid_lft)
2076 return -EINVAL;
2078 dev = __dev_get_by_index(net, ifindex);
2079 if (!dev)
2080 return -ENODEV;
2082 if ((idev = addrconf_add_dev(dev)) == NULL)
2083 return -ENOBUFS;
2085 scope = ipv6_addr_scope(pfx);
2087 timeout = addrconf_timeout_fixup(valid_lft, HZ);
2088 if (addrconf_finite_timeout(timeout)) {
2089 expires = jiffies_to_clock_t(timeout * HZ);
2090 valid_lft = timeout;
2091 flags = RTF_EXPIRES;
2092 } else {
2093 expires = 0;
2094 flags = 0;
2095 ifa_flags |= IFA_F_PERMANENT;
2098 timeout = addrconf_timeout_fixup(prefered_lft, HZ);
2099 if (addrconf_finite_timeout(timeout)) {
2100 if (timeout == 0)
2101 ifa_flags |= IFA_F_DEPRECATED;
2102 prefered_lft = timeout;
2105 ifp = ipv6_add_addr(idev, pfx, plen, scope, ifa_flags);
2107 if (!IS_ERR(ifp)) {
2108 spin_lock_bh(&ifp->lock);
2109 ifp->valid_lft = valid_lft;
2110 ifp->prefered_lft = prefered_lft;
2111 ifp->tstamp = jiffies;
2112 spin_unlock_bh(&ifp->lock);
2114 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, dev,
2115 expires, flags);
2117 * Note that section 3.1 of RFC 4429 indicates
2118 * that the Optimistic flag should not be set for
2119 * manually configured addresses
2121 addrconf_dad_start(ifp, 0);
2122 in6_ifa_put(ifp);
2123 addrconf_verify(0);
2124 return 0;
2127 return PTR_ERR(ifp);
2130 static int inet6_addr_del(struct net *net, int ifindex, struct in6_addr *pfx,
2131 unsigned int plen)
2133 struct inet6_ifaddr *ifp;
2134 struct inet6_dev *idev;
2135 struct net_device *dev;
2137 if (plen > 128)
2138 return -EINVAL;
2140 dev = __dev_get_by_index(net, ifindex);
2141 if (!dev)
2142 return -ENODEV;
2144 if ((idev = __in6_dev_get(dev)) == NULL)
2145 return -ENXIO;
2147 read_lock_bh(&idev->lock);
2148 for (ifp = idev->addr_list; ifp; ifp=ifp->if_next) {
2149 if (ifp->prefix_len == plen &&
2150 ipv6_addr_equal(pfx, &ifp->addr)) {
2151 in6_ifa_hold(ifp);
2152 read_unlock_bh(&idev->lock);
2154 ipv6_del_addr(ifp);
2156 /* If the last address is deleted administratively,
2157 disable IPv6 on this interface.
2159 if (idev->addr_list == NULL)
2160 addrconf_ifdown(idev->dev, 1);
2161 return 0;
2164 read_unlock_bh(&idev->lock);
2165 return -EADDRNOTAVAIL;
2169 int addrconf_add_ifaddr(struct net *net, void __user *arg)
2171 struct in6_ifreq ireq;
2172 int err;
2174 if (!capable(CAP_NET_ADMIN))
2175 return -EPERM;
2177 if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2178 return -EFAULT;
2180 rtnl_lock();
2181 err = inet6_addr_add(net, ireq.ifr6_ifindex, &ireq.ifr6_addr,
2182 ireq.ifr6_prefixlen, IFA_F_PERMANENT,
2183 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
2184 rtnl_unlock();
2185 return err;
2188 int addrconf_del_ifaddr(struct net *net, void __user *arg)
2190 struct in6_ifreq ireq;
2191 int err;
2193 if (!capable(CAP_NET_ADMIN))
2194 return -EPERM;
2196 if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2197 return -EFAULT;
2199 rtnl_lock();
2200 err = inet6_addr_del(net, ireq.ifr6_ifindex, &ireq.ifr6_addr,
2201 ireq.ifr6_prefixlen);
2202 rtnl_unlock();
2203 return err;
2206 static void add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
2207 int plen, int scope)
2209 struct inet6_ifaddr *ifp;
2211 ifp = ipv6_add_addr(idev, addr, plen, scope, IFA_F_PERMANENT);
2212 if (!IS_ERR(ifp)) {
2213 spin_lock_bh(&ifp->lock);
2214 ifp->flags &= ~IFA_F_TENTATIVE;
2215 spin_unlock_bh(&ifp->lock);
2216 ipv6_ifa_notify(RTM_NEWADDR, ifp);
2217 in6_ifa_put(ifp);
2221 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2222 static void sit_add_v4_addrs(struct inet6_dev *idev)
2224 struct in6_addr addr;
2225 struct net_device *dev;
2226 struct net *net = dev_net(idev->dev);
2227 int scope;
2229 ASSERT_RTNL();
2231 memset(&addr, 0, sizeof(struct in6_addr));
2232 memcpy(&addr.s6_addr32[3], idev->dev->dev_addr, 4);
2234 if (idev->dev->flags&IFF_POINTOPOINT) {
2235 addr.s6_addr32[0] = htonl(0xfe800000);
2236 scope = IFA_LINK;
2237 } else {
2238 scope = IPV6_ADDR_COMPATv4;
2241 if (addr.s6_addr32[3]) {
2242 add_addr(idev, &addr, 128, scope);
2243 return;
2246 for_each_netdev(net, dev) {
2247 struct in_device * in_dev = __in_dev_get_rtnl(dev);
2248 if (in_dev && (dev->flags & IFF_UP)) {
2249 struct in_ifaddr * ifa;
2251 int flag = scope;
2253 for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) {
2254 int plen;
2256 addr.s6_addr32[3] = ifa->ifa_local;
2258 if (ifa->ifa_scope == RT_SCOPE_LINK)
2259 continue;
2260 if (ifa->ifa_scope >= RT_SCOPE_HOST) {
2261 if (idev->dev->flags&IFF_POINTOPOINT)
2262 continue;
2263 flag |= IFA_HOST;
2265 if (idev->dev->flags&IFF_POINTOPOINT)
2266 plen = 64;
2267 else
2268 plen = 96;
2270 add_addr(idev, &addr, plen, flag);
2275 #endif
2277 static void init_loopback(struct net_device *dev)
2279 struct inet6_dev *idev;
2281 /* ::1 */
2283 ASSERT_RTNL();
2285 if ((idev = ipv6_find_idev(dev)) == NULL) {
2286 printk(KERN_DEBUG "init loopback: add_dev failed\n");
2287 return;
2290 add_addr(idev, &in6addr_loopback, 128, IFA_HOST);
2293 static void addrconf_add_linklocal(struct inet6_dev *idev, struct in6_addr *addr)
2295 struct inet6_ifaddr * ifp;
2296 u32 addr_flags = IFA_F_PERMANENT;
2298 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2299 if (idev->cnf.optimistic_dad &&
2300 !dev_net(idev->dev)->ipv6.devconf_all->forwarding)
2301 addr_flags |= IFA_F_OPTIMISTIC;
2302 #endif
2305 ifp = ipv6_add_addr(idev, addr, 64, IFA_LINK, addr_flags);
2306 if (!IS_ERR(ifp)) {
2307 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, idev->dev, 0, 0);
2308 addrconf_dad_start(ifp, 0);
2309 in6_ifa_put(ifp);
2313 static void addrconf_dev_config(struct net_device *dev)
2315 struct in6_addr addr;
2316 struct inet6_dev * idev;
2318 ASSERT_RTNL();
2320 if ((dev->type != ARPHRD_ETHER) &&
2321 (dev->type != ARPHRD_FDDI) &&
2322 (dev->type != ARPHRD_IEEE802_TR) &&
2323 (dev->type != ARPHRD_ARCNET) &&
2324 (dev->type != ARPHRD_INFINIBAND)) {
2325 /* Alas, we support only Ethernet autoconfiguration. */
2326 return;
2329 idev = addrconf_add_dev(dev);
2330 if (idev == NULL)
2331 return;
2333 memset(&addr, 0, sizeof(struct in6_addr));
2334 addr.s6_addr32[0] = htonl(0xFE800000);
2336 if (ipv6_generate_eui64(addr.s6_addr + 8, dev) == 0)
2337 addrconf_add_linklocal(idev, &addr);
2340 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2341 static void addrconf_sit_config(struct net_device *dev)
2343 struct inet6_dev *idev;
2345 ASSERT_RTNL();
2348 * Configure the tunnel with one of our IPv4
2349 * addresses... we should configure all of
2350 * our v4 addrs in the tunnel
2353 if ((idev = ipv6_find_idev(dev)) == NULL) {
2354 printk(KERN_DEBUG "init sit: add_dev failed\n");
2355 return;
2358 if (dev->priv_flags & IFF_ISATAP) {
2359 struct in6_addr addr;
2361 ipv6_addr_set(&addr, htonl(0xFE800000), 0, 0, 0);
2362 addrconf_prefix_route(&addr, 64, dev, 0, 0);
2363 if (!ipv6_generate_eui64(addr.s6_addr + 8, dev))
2364 addrconf_add_linklocal(idev, &addr);
2365 return;
2368 sit_add_v4_addrs(idev);
2370 if (dev->flags&IFF_POINTOPOINT) {
2371 addrconf_add_mroute(dev);
2372 addrconf_add_lroute(dev);
2373 } else
2374 sit_route_add(dev);
2376 #endif
2378 static inline int
2379 ipv6_inherit_linklocal(struct inet6_dev *idev, struct net_device *link_dev)
2381 struct in6_addr lladdr;
2383 if (!ipv6_get_lladdr(link_dev, &lladdr, IFA_F_TENTATIVE)) {
2384 addrconf_add_linklocal(idev, &lladdr);
2385 return 0;
2387 return -1;
2390 static void ip6_tnl_add_linklocal(struct inet6_dev *idev)
2392 struct net_device *link_dev;
2393 struct net *net = dev_net(idev->dev);
2395 /* first try to inherit the link-local address from the link device */
2396 if (idev->dev->iflink &&
2397 (link_dev = __dev_get_by_index(net, idev->dev->iflink))) {
2398 if (!ipv6_inherit_linklocal(idev, link_dev))
2399 return;
2401 /* then try to inherit it from any device */
2402 for_each_netdev(net, link_dev) {
2403 if (!ipv6_inherit_linklocal(idev, link_dev))
2404 return;
2406 printk(KERN_DEBUG "init ip6-ip6: add_linklocal failed\n");
2410 * Autoconfigure tunnel with a link-local address so routing protocols,
2411 * DHCPv6, MLD etc. can be run over the virtual link
2414 static void addrconf_ip6_tnl_config(struct net_device *dev)
2416 struct inet6_dev *idev;
2418 ASSERT_RTNL();
2420 if ((idev = addrconf_add_dev(dev)) == NULL) {
2421 printk(KERN_DEBUG "init ip6-ip6: add_dev failed\n");
2422 return;
2424 ip6_tnl_add_linklocal(idev);
2427 static int addrconf_notify(struct notifier_block *this, unsigned long event,
2428 void * data)
2430 struct net_device *dev = (struct net_device *) data;
2431 struct inet6_dev *idev = __in6_dev_get(dev);
2432 int run_pending = 0;
2433 int err;
2435 switch(event) {
2436 case NETDEV_REGISTER:
2437 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2438 idev = ipv6_add_dev(dev);
2439 if (!idev)
2440 return notifier_from_errno(-ENOMEM);
2442 break;
2443 case NETDEV_UP:
2444 case NETDEV_CHANGE:
2445 if (dev->flags & IFF_SLAVE)
2446 break;
2448 if (event == NETDEV_UP) {
2449 if (!addrconf_qdisc_ok(dev)) {
2450 /* device is not ready yet. */
2451 printk(KERN_INFO
2452 "ADDRCONF(NETDEV_UP): %s: "
2453 "link is not ready\n",
2454 dev->name);
2455 break;
2458 if (!idev && dev->mtu >= IPV6_MIN_MTU)
2459 idev = ipv6_add_dev(dev);
2461 if (idev) {
2462 idev->if_flags |= IF_READY;
2463 run_pending = 1;
2465 } else {
2466 if (!addrconf_qdisc_ok(dev)) {
2467 /* device is still not ready. */
2468 break;
2471 if (idev) {
2472 if (idev->if_flags & IF_READY) {
2473 /* device is already configured. */
2474 break;
2476 idev->if_flags |= IF_READY;
2479 printk(KERN_INFO
2480 "ADDRCONF(NETDEV_CHANGE): %s: "
2481 "link becomes ready\n",
2482 dev->name);
2484 run_pending = 1;
2487 switch(dev->type) {
2488 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2489 case ARPHRD_SIT:
2490 addrconf_sit_config(dev);
2491 break;
2492 #endif
2493 case ARPHRD_TUNNEL6:
2494 addrconf_ip6_tnl_config(dev);
2495 break;
2496 case ARPHRD_LOOPBACK:
2497 init_loopback(dev);
2498 break;
2500 default:
2501 addrconf_dev_config(dev);
2502 break;
2504 if (idev) {
2505 if (run_pending)
2506 addrconf_dad_run(idev);
2508 /* If the MTU changed during the interface down, when the
2509 interface up, the changed MTU must be reflected in the
2510 idev as well as routers.
2512 if (idev->cnf.mtu6 != dev->mtu && dev->mtu >= IPV6_MIN_MTU) {
2513 rt6_mtu_change(dev, dev->mtu);
2514 idev->cnf.mtu6 = dev->mtu;
2516 idev->tstamp = jiffies;
2517 inet6_ifinfo_notify(RTM_NEWLINK, idev);
2518 /* If the changed mtu during down is lower than IPV6_MIN_MTU
2519 stop IPv6 on this interface.
2521 if (dev->mtu < IPV6_MIN_MTU)
2522 addrconf_ifdown(dev, event != NETDEV_DOWN);
2524 break;
2526 case NETDEV_CHANGEMTU:
2527 if (idev && dev->mtu >= IPV6_MIN_MTU) {
2528 rt6_mtu_change(dev, dev->mtu);
2529 idev->cnf.mtu6 = dev->mtu;
2530 break;
2533 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2534 idev = ipv6_add_dev(dev);
2535 if (idev)
2536 break;
2539 /* MTU falled under IPV6_MIN_MTU. Stop IPv6 on this interface. */
2541 case NETDEV_DOWN:
2542 case NETDEV_UNREGISTER:
2544 * Remove all addresses from this interface.
2546 addrconf_ifdown(dev, event != NETDEV_DOWN);
2547 break;
2549 case NETDEV_CHANGENAME:
2550 if (idev) {
2551 snmp6_unregister_dev(idev);
2552 addrconf_sysctl_unregister(idev);
2553 addrconf_sysctl_register(idev);
2554 err = snmp6_register_dev(idev);
2555 if (err)
2556 return notifier_from_errno(err);
2558 break;
2561 return NOTIFY_OK;
2565 * addrconf module should be notified of a device going up
2567 static struct notifier_block ipv6_dev_notf = {
2568 .notifier_call = addrconf_notify,
2569 .priority = 0
2572 static int addrconf_ifdown(struct net_device *dev, int how)
2574 struct inet6_dev *idev;
2575 struct inet6_ifaddr *ifa, **bifa;
2576 struct net *net = dev_net(dev);
2577 int i;
2579 ASSERT_RTNL();
2581 rt6_ifdown(net, dev);
2582 neigh_ifdown(&nd_tbl, dev);
2584 idev = __in6_dev_get(dev);
2585 if (idev == NULL)
2586 return -ENODEV;
2588 /* Step 1: remove reference to ipv6 device from parent device.
2589 Do not dev_put!
2591 if (how) {
2592 idev->dead = 1;
2594 /* protected by rtnl_lock */
2595 rcu_assign_pointer(dev->ip6_ptr, NULL);
2597 /* Step 1.5: remove snmp6 entry */
2598 snmp6_unregister_dev(idev);
2602 /* Step 2: clear hash table */
2603 for (i=0; i<IN6_ADDR_HSIZE; i++) {
2604 bifa = &inet6_addr_lst[i];
2606 write_lock_bh(&addrconf_hash_lock);
2607 while ((ifa = *bifa) != NULL) {
2608 if (ifa->idev == idev) {
2609 *bifa = ifa->lst_next;
2610 ifa->lst_next = NULL;
2611 addrconf_del_timer(ifa);
2612 in6_ifa_put(ifa);
2613 continue;
2615 bifa = &ifa->lst_next;
2617 write_unlock_bh(&addrconf_hash_lock);
2620 write_lock_bh(&idev->lock);
2622 /* Step 3: clear flags for stateless addrconf */
2623 if (!how)
2624 idev->if_flags &= ~(IF_RS_SENT|IF_RA_RCVD|IF_READY);
2626 /* Step 4: clear address list */
2627 #ifdef CONFIG_IPV6_PRIVACY
2628 if (how && del_timer(&idev->regen_timer))
2629 in6_dev_put(idev);
2631 /* clear tempaddr list */
2632 while ((ifa = idev->tempaddr_list) != NULL) {
2633 idev->tempaddr_list = ifa->tmp_next;
2634 ifa->tmp_next = NULL;
2635 ifa->dead = 1;
2636 write_unlock_bh(&idev->lock);
2637 spin_lock_bh(&ifa->lock);
2639 if (ifa->ifpub) {
2640 in6_ifa_put(ifa->ifpub);
2641 ifa->ifpub = NULL;
2643 spin_unlock_bh(&ifa->lock);
2644 in6_ifa_put(ifa);
2645 write_lock_bh(&idev->lock);
2647 #endif
2648 while ((ifa = idev->addr_list) != NULL) {
2649 idev->addr_list = ifa->if_next;
2650 ifa->if_next = NULL;
2651 ifa->dead = 1;
2652 addrconf_del_timer(ifa);
2653 write_unlock_bh(&idev->lock);
2655 __ipv6_ifa_notify(RTM_DELADDR, ifa);
2656 atomic_notifier_call_chain(&inet6addr_chain, NETDEV_DOWN, ifa);
2657 in6_ifa_put(ifa);
2659 write_lock_bh(&idev->lock);
2661 write_unlock_bh(&idev->lock);
2663 /* Step 5: Discard multicast list */
2665 if (how)
2666 ipv6_mc_destroy_dev(idev);
2667 else
2668 ipv6_mc_down(idev);
2670 idev->tstamp = jiffies;
2672 /* Shot the device (if unregistered) */
2674 if (how) {
2675 addrconf_sysctl_unregister(idev);
2676 neigh_parms_release(&nd_tbl, idev->nd_parms);
2677 neigh_ifdown(&nd_tbl, dev);
2678 in6_dev_put(idev);
2680 return 0;
2683 static void addrconf_rs_timer(unsigned long data)
2685 struct inet6_ifaddr *ifp = (struct inet6_ifaddr *) data;
2687 if (ifp->idev->cnf.forwarding)
2688 goto out;
2690 if (ifp->idev->if_flags & IF_RA_RCVD) {
2692 * Announcement received after solicitation
2693 * was sent
2695 goto out;
2698 spin_lock(&ifp->lock);
2699 if (ifp->probes++ < ifp->idev->cnf.rtr_solicits) {
2700 /* The wait after the last probe can be shorter */
2701 addrconf_mod_timer(ifp, AC_RS,
2702 (ifp->probes == ifp->idev->cnf.rtr_solicits) ?
2703 ifp->idev->cnf.rtr_solicit_delay :
2704 ifp->idev->cnf.rtr_solicit_interval);
2705 spin_unlock(&ifp->lock);
2707 ndisc_send_rs(ifp->idev->dev, &ifp->addr, &in6addr_linklocal_allrouters);
2708 } else {
2709 spin_unlock(&ifp->lock);
2711 * Note: we do not support deprecated "all on-link"
2712 * assumption any longer.
2714 printk(KERN_DEBUG "%s: no IPv6 routers present\n",
2715 ifp->idev->dev->name);
2718 out:
2719 in6_ifa_put(ifp);
2723 * Duplicate Address Detection
2725 static void addrconf_dad_kick(struct inet6_ifaddr *ifp)
2727 unsigned long rand_num;
2728 struct inet6_dev *idev = ifp->idev;
2730 if (ifp->flags & IFA_F_OPTIMISTIC)
2731 rand_num = 0;
2732 else
2733 rand_num = net_random() % (idev->cnf.rtr_solicit_delay ? : 1);
2735 ifp->probes = idev->cnf.dad_transmits;
2736 addrconf_mod_timer(ifp, AC_DAD, rand_num);
2739 static void addrconf_dad_start(struct inet6_ifaddr *ifp, u32 flags)
2741 struct inet6_dev *idev = ifp->idev;
2742 struct net_device *dev = idev->dev;
2744 addrconf_join_solict(dev, &ifp->addr);
2746 net_srandom(ifp->addr.s6_addr32[3]);
2748 read_lock_bh(&idev->lock);
2749 if (ifp->dead)
2750 goto out;
2751 spin_lock_bh(&ifp->lock);
2753 if (dev->flags&(IFF_NOARP|IFF_LOOPBACK) ||
2754 idev->cnf.accept_dad < 1 ||
2755 !(ifp->flags&IFA_F_TENTATIVE) ||
2756 ifp->flags & IFA_F_NODAD) {
2757 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC);
2758 spin_unlock_bh(&ifp->lock);
2759 read_unlock_bh(&idev->lock);
2761 addrconf_dad_completed(ifp);
2762 return;
2765 if (!(idev->if_flags & IF_READY)) {
2766 spin_unlock_bh(&ifp->lock);
2767 read_unlock_bh(&idev->lock);
2769 * If the device is not ready:
2770 * - keep it tentative if it is a permanent address.
2771 * - otherwise, kill it.
2773 in6_ifa_hold(ifp);
2774 addrconf_dad_stop(ifp);
2775 return;
2779 * Optimistic nodes can start receiving
2780 * Frames right away
2782 if(ifp->flags & IFA_F_OPTIMISTIC)
2783 ip6_ins_rt(ifp->rt);
2785 addrconf_dad_kick(ifp);
2786 spin_unlock_bh(&ifp->lock);
2787 out:
2788 read_unlock_bh(&idev->lock);
2791 static void addrconf_dad_timer(unsigned long data)
2793 struct inet6_ifaddr *ifp = (struct inet6_ifaddr *) data;
2794 struct inet6_dev *idev = ifp->idev;
2795 struct in6_addr mcaddr;
2797 read_lock_bh(&idev->lock);
2798 if (idev->dead) {
2799 read_unlock_bh(&idev->lock);
2800 goto out;
2802 spin_lock_bh(&ifp->lock);
2803 if (ifp->probes == 0) {
2805 * DAD was successful
2808 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC);
2809 spin_unlock_bh(&ifp->lock);
2810 read_unlock_bh(&idev->lock);
2812 addrconf_dad_completed(ifp);
2814 goto out;
2817 ifp->probes--;
2818 addrconf_mod_timer(ifp, AC_DAD, ifp->idev->nd_parms->retrans_time);
2819 spin_unlock_bh(&ifp->lock);
2820 read_unlock_bh(&idev->lock);
2822 /* send a neighbour solicitation for our addr */
2823 addrconf_addr_solict_mult(&ifp->addr, &mcaddr);
2824 ndisc_send_ns(ifp->idev->dev, NULL, &ifp->addr, &mcaddr, &in6addr_any);
2825 out:
2826 in6_ifa_put(ifp);
2829 static void addrconf_dad_completed(struct inet6_ifaddr *ifp)
2831 struct net_device * dev = ifp->idev->dev;
2834 * Configure the address for reception. Now it is valid.
2837 ipv6_ifa_notify(RTM_NEWADDR, ifp);
2839 /* If added prefix is link local and forwarding is off,
2840 start sending router solicitations.
2843 if (ifp->idev->cnf.forwarding == 0 &&
2844 ifp->idev->cnf.rtr_solicits > 0 &&
2845 (dev->flags&IFF_LOOPBACK) == 0 &&
2846 (ipv6_addr_type(&ifp->addr) & IPV6_ADDR_LINKLOCAL)) {
2848 * If a host as already performed a random delay
2849 * [...] as part of DAD [...] there is no need
2850 * to delay again before sending the first RS
2852 ndisc_send_rs(ifp->idev->dev, &ifp->addr, &in6addr_linklocal_allrouters);
2854 spin_lock_bh(&ifp->lock);
2855 ifp->probes = 1;
2856 ifp->idev->if_flags |= IF_RS_SENT;
2857 addrconf_mod_timer(ifp, AC_RS, ifp->idev->cnf.rtr_solicit_interval);
2858 spin_unlock_bh(&ifp->lock);
2862 static void addrconf_dad_run(struct inet6_dev *idev) {
2863 struct inet6_ifaddr *ifp;
2865 read_lock_bh(&idev->lock);
2866 for (ifp = idev->addr_list; ifp; ifp = ifp->if_next) {
2867 spin_lock_bh(&ifp->lock);
2868 if (!(ifp->flags & IFA_F_TENTATIVE)) {
2869 spin_unlock_bh(&ifp->lock);
2870 continue;
2872 spin_unlock_bh(&ifp->lock);
2873 addrconf_dad_kick(ifp);
2875 read_unlock_bh(&idev->lock);
2878 #ifdef CONFIG_PROC_FS
2879 struct if6_iter_state {
2880 struct seq_net_private p;
2881 int bucket;
2884 static struct inet6_ifaddr *if6_get_first(struct seq_file *seq)
2886 struct inet6_ifaddr *ifa = NULL;
2887 struct if6_iter_state *state = seq->private;
2888 struct net *net = seq_file_net(seq);
2890 for (state->bucket = 0; state->bucket < IN6_ADDR_HSIZE; ++state->bucket) {
2891 ifa = inet6_addr_lst[state->bucket];
2893 while (ifa && !net_eq(dev_net(ifa->idev->dev), net))
2894 ifa = ifa->lst_next;
2895 if (ifa)
2896 break;
2898 return ifa;
2901 static struct inet6_ifaddr *if6_get_next(struct seq_file *seq, struct inet6_ifaddr *ifa)
2903 struct if6_iter_state *state = seq->private;
2904 struct net *net = seq_file_net(seq);
2906 ifa = ifa->lst_next;
2907 try_again:
2908 if (ifa) {
2909 if (!net_eq(dev_net(ifa->idev->dev), net)) {
2910 ifa = ifa->lst_next;
2911 goto try_again;
2915 if (!ifa && ++state->bucket < IN6_ADDR_HSIZE) {
2916 ifa = inet6_addr_lst[state->bucket];
2917 goto try_again;
2920 return ifa;
2923 static struct inet6_ifaddr *if6_get_idx(struct seq_file *seq, loff_t pos)
2925 struct inet6_ifaddr *ifa = if6_get_first(seq);
2927 if (ifa)
2928 while(pos && (ifa = if6_get_next(seq, ifa)) != NULL)
2929 --pos;
2930 return pos ? NULL : ifa;
2933 static void *if6_seq_start(struct seq_file *seq, loff_t *pos)
2934 __acquires(addrconf_hash_lock)
2936 read_lock_bh(&addrconf_hash_lock);
2937 return if6_get_idx(seq, *pos);
2940 static void *if6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
2942 struct inet6_ifaddr *ifa;
2944 ifa = if6_get_next(seq, v);
2945 ++*pos;
2946 return ifa;
2949 static void if6_seq_stop(struct seq_file *seq, void *v)
2950 __releases(addrconf_hash_lock)
2952 read_unlock_bh(&addrconf_hash_lock);
2955 static int if6_seq_show(struct seq_file *seq, void *v)
2957 struct inet6_ifaddr *ifp = (struct inet6_ifaddr *)v;
2958 seq_printf(seq, "%pi6 %02x %02x %02x %02x %8s\n",
2959 &ifp->addr,
2960 ifp->idev->dev->ifindex,
2961 ifp->prefix_len,
2962 ifp->scope,
2963 ifp->flags,
2964 ifp->idev->dev->name);
2965 return 0;
2968 static const struct seq_operations if6_seq_ops = {
2969 .start = if6_seq_start,
2970 .next = if6_seq_next,
2971 .show = if6_seq_show,
2972 .stop = if6_seq_stop,
2975 static int if6_seq_open(struct inode *inode, struct file *file)
2977 return seq_open_net(inode, file, &if6_seq_ops,
2978 sizeof(struct if6_iter_state));
2981 static const struct file_operations if6_fops = {
2982 .owner = THIS_MODULE,
2983 .open = if6_seq_open,
2984 .read = seq_read,
2985 .llseek = seq_lseek,
2986 .release = seq_release_net,
2989 static int if6_proc_net_init(struct net *net)
2991 if (!proc_net_fops_create(net, "if_inet6", S_IRUGO, &if6_fops))
2992 return -ENOMEM;
2993 return 0;
2996 static void if6_proc_net_exit(struct net *net)
2998 proc_net_remove(net, "if_inet6");
3001 static struct pernet_operations if6_proc_net_ops = {
3002 .init = if6_proc_net_init,
3003 .exit = if6_proc_net_exit,
3006 int __init if6_proc_init(void)
3008 return register_pernet_subsys(&if6_proc_net_ops);
3011 void if6_proc_exit(void)
3013 unregister_pernet_subsys(&if6_proc_net_ops);
3015 #endif /* CONFIG_PROC_FS */
3017 #if defined(CONFIG_IPV6_MIP6) || defined(CONFIG_IPV6_MIP6_MODULE)
3018 /* Check if address is a home address configured on any interface. */
3019 int ipv6_chk_home_addr(struct net *net, struct in6_addr *addr)
3021 int ret = 0;
3022 struct inet6_ifaddr * ifp;
3023 u8 hash = ipv6_addr_hash(addr);
3024 read_lock_bh(&addrconf_hash_lock);
3025 for (ifp = inet6_addr_lst[hash]; ifp; ifp = ifp->lst_next) {
3026 if (!net_eq(dev_net(ifp->idev->dev), net))
3027 continue;
3028 if (ipv6_addr_equal(&ifp->addr, addr) &&
3029 (ifp->flags & IFA_F_HOMEADDRESS)) {
3030 ret = 1;
3031 break;
3034 read_unlock_bh(&addrconf_hash_lock);
3035 return ret;
3037 #endif
3040 * Periodic address status verification
3043 static void addrconf_verify(unsigned long foo)
3045 struct inet6_ifaddr *ifp;
3046 unsigned long now, next;
3047 int i;
3049 spin_lock_bh(&addrconf_verify_lock);
3050 now = jiffies;
3051 next = now + ADDR_CHECK_FREQUENCY;
3053 del_timer(&addr_chk_timer);
3055 for (i=0; i < IN6_ADDR_HSIZE; i++) {
3057 restart:
3058 read_lock(&addrconf_hash_lock);
3059 for (ifp=inet6_addr_lst[i]; ifp; ifp=ifp->lst_next) {
3060 unsigned long age;
3061 #ifdef CONFIG_IPV6_PRIVACY
3062 unsigned long regen_advance;
3063 #endif
3065 if (ifp->flags & IFA_F_PERMANENT)
3066 continue;
3068 spin_lock(&ifp->lock);
3069 age = (now - ifp->tstamp) / HZ;
3071 #ifdef CONFIG_IPV6_PRIVACY
3072 regen_advance = ifp->idev->cnf.regen_max_retry *
3073 ifp->idev->cnf.dad_transmits *
3074 ifp->idev->nd_parms->retrans_time / HZ;
3075 #endif
3077 if (ifp->valid_lft != INFINITY_LIFE_TIME &&
3078 age >= ifp->valid_lft) {
3079 spin_unlock(&ifp->lock);
3080 in6_ifa_hold(ifp);
3081 read_unlock(&addrconf_hash_lock);
3082 ipv6_del_addr(ifp);
3083 goto restart;
3084 } else if (ifp->prefered_lft == INFINITY_LIFE_TIME) {
3085 spin_unlock(&ifp->lock);
3086 continue;
3087 } else if (age >= ifp->prefered_lft) {
3088 /* jiffies - ifp->tsamp > age >= ifp->prefered_lft */
3089 int deprecate = 0;
3091 if (!(ifp->flags&IFA_F_DEPRECATED)) {
3092 deprecate = 1;
3093 ifp->flags |= IFA_F_DEPRECATED;
3096 if (time_before(ifp->tstamp + ifp->valid_lft * HZ, next))
3097 next = ifp->tstamp + ifp->valid_lft * HZ;
3099 spin_unlock(&ifp->lock);
3101 if (deprecate) {
3102 in6_ifa_hold(ifp);
3103 read_unlock(&addrconf_hash_lock);
3105 ipv6_ifa_notify(0, ifp);
3106 in6_ifa_put(ifp);
3107 goto restart;
3109 #ifdef CONFIG_IPV6_PRIVACY
3110 } else if ((ifp->flags&IFA_F_TEMPORARY) &&
3111 !(ifp->flags&IFA_F_TENTATIVE)) {
3112 if (age >= ifp->prefered_lft - regen_advance) {
3113 struct inet6_ifaddr *ifpub = ifp->ifpub;
3114 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3115 next = ifp->tstamp + ifp->prefered_lft * HZ;
3116 if (!ifp->regen_count && ifpub) {
3117 ifp->regen_count++;
3118 in6_ifa_hold(ifp);
3119 in6_ifa_hold(ifpub);
3120 spin_unlock(&ifp->lock);
3121 read_unlock(&addrconf_hash_lock);
3122 spin_lock(&ifpub->lock);
3123 ifpub->regen_count = 0;
3124 spin_unlock(&ifpub->lock);
3125 ipv6_create_tempaddr(ifpub, ifp);
3126 in6_ifa_put(ifpub);
3127 in6_ifa_put(ifp);
3128 goto restart;
3130 } else if (time_before(ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ, next))
3131 next = ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ;
3132 spin_unlock(&ifp->lock);
3133 #endif
3134 } else {
3135 /* ifp->prefered_lft <= ifp->valid_lft */
3136 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3137 next = ifp->tstamp + ifp->prefered_lft * HZ;
3138 spin_unlock(&ifp->lock);
3141 read_unlock(&addrconf_hash_lock);
3144 addr_chk_timer.expires = time_before(next, jiffies + HZ) ? jiffies + HZ : next;
3145 add_timer(&addr_chk_timer);
3146 spin_unlock_bh(&addrconf_verify_lock);
3149 static struct in6_addr *extract_addr(struct nlattr *addr, struct nlattr *local)
3151 struct in6_addr *pfx = NULL;
3153 if (addr)
3154 pfx = nla_data(addr);
3156 if (local) {
3157 if (pfx && nla_memcmp(local, pfx, sizeof(*pfx)))
3158 pfx = NULL;
3159 else
3160 pfx = nla_data(local);
3163 return pfx;
3166 static const struct nla_policy ifa_ipv6_policy[IFA_MAX+1] = {
3167 [IFA_ADDRESS] = { .len = sizeof(struct in6_addr) },
3168 [IFA_LOCAL] = { .len = sizeof(struct in6_addr) },
3169 [IFA_CACHEINFO] = { .len = sizeof(struct ifa_cacheinfo) },
3172 static int
3173 inet6_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
3175 struct net *net = sock_net(skb->sk);
3176 struct ifaddrmsg *ifm;
3177 struct nlattr *tb[IFA_MAX+1];
3178 struct in6_addr *pfx;
3179 int err;
3181 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3182 if (err < 0)
3183 return err;
3185 ifm = nlmsg_data(nlh);
3186 pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3187 if (pfx == NULL)
3188 return -EINVAL;
3190 return inet6_addr_del(net, ifm->ifa_index, pfx, ifm->ifa_prefixlen);
3193 static int inet6_addr_modify(struct inet6_ifaddr *ifp, u8 ifa_flags,
3194 u32 prefered_lft, u32 valid_lft)
3196 u32 flags;
3197 clock_t expires;
3198 unsigned long timeout;
3200 if (!valid_lft || (prefered_lft > valid_lft))
3201 return -EINVAL;
3203 timeout = addrconf_timeout_fixup(valid_lft, HZ);
3204 if (addrconf_finite_timeout(timeout)) {
3205 expires = jiffies_to_clock_t(timeout * HZ);
3206 valid_lft = timeout;
3207 flags = RTF_EXPIRES;
3208 } else {
3209 expires = 0;
3210 flags = 0;
3211 ifa_flags |= IFA_F_PERMANENT;
3214 timeout = addrconf_timeout_fixup(prefered_lft, HZ);
3215 if (addrconf_finite_timeout(timeout)) {
3216 if (timeout == 0)
3217 ifa_flags |= IFA_F_DEPRECATED;
3218 prefered_lft = timeout;
3221 spin_lock_bh(&ifp->lock);
3222 ifp->flags = (ifp->flags & ~(IFA_F_DEPRECATED | IFA_F_PERMANENT | IFA_F_NODAD | IFA_F_HOMEADDRESS)) | ifa_flags;
3223 ifp->tstamp = jiffies;
3224 ifp->valid_lft = valid_lft;
3225 ifp->prefered_lft = prefered_lft;
3227 spin_unlock_bh(&ifp->lock);
3228 if (!(ifp->flags&IFA_F_TENTATIVE))
3229 ipv6_ifa_notify(0, ifp);
3231 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, ifp->idev->dev,
3232 expires, flags);
3233 addrconf_verify(0);
3235 return 0;
3238 static int
3239 inet6_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
3241 struct net *net = sock_net(skb->sk);
3242 struct ifaddrmsg *ifm;
3243 struct nlattr *tb[IFA_MAX+1];
3244 struct in6_addr *pfx;
3245 struct inet6_ifaddr *ifa;
3246 struct net_device *dev;
3247 u32 valid_lft = INFINITY_LIFE_TIME, preferred_lft = INFINITY_LIFE_TIME;
3248 u8 ifa_flags;
3249 int err;
3251 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3252 if (err < 0)
3253 return err;
3255 ifm = nlmsg_data(nlh);
3256 pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3257 if (pfx == NULL)
3258 return -EINVAL;
3260 if (tb[IFA_CACHEINFO]) {
3261 struct ifa_cacheinfo *ci;
3263 ci = nla_data(tb[IFA_CACHEINFO]);
3264 valid_lft = ci->ifa_valid;
3265 preferred_lft = ci->ifa_prefered;
3266 } else {
3267 preferred_lft = INFINITY_LIFE_TIME;
3268 valid_lft = INFINITY_LIFE_TIME;
3271 dev = __dev_get_by_index(net, ifm->ifa_index);
3272 if (dev == NULL)
3273 return -ENODEV;
3275 /* We ignore other flags so far. */
3276 ifa_flags = ifm->ifa_flags & (IFA_F_NODAD | IFA_F_HOMEADDRESS);
3278 ifa = ipv6_get_ifaddr(net, pfx, dev, 1);
3279 if (ifa == NULL) {
3281 * It would be best to check for !NLM_F_CREATE here but
3282 * userspace alreay relies on not having to provide this.
3284 return inet6_addr_add(net, ifm->ifa_index, pfx,
3285 ifm->ifa_prefixlen, ifa_flags,
3286 preferred_lft, valid_lft);
3289 if (nlh->nlmsg_flags & NLM_F_EXCL ||
3290 !(nlh->nlmsg_flags & NLM_F_REPLACE))
3291 err = -EEXIST;
3292 else
3293 err = inet6_addr_modify(ifa, ifa_flags, preferred_lft, valid_lft);
3295 in6_ifa_put(ifa);
3297 return err;
3300 static void put_ifaddrmsg(struct nlmsghdr *nlh, u8 prefixlen, u8 flags,
3301 u8 scope, int ifindex)
3303 struct ifaddrmsg *ifm;
3305 ifm = nlmsg_data(nlh);
3306 ifm->ifa_family = AF_INET6;
3307 ifm->ifa_prefixlen = prefixlen;
3308 ifm->ifa_flags = flags;
3309 ifm->ifa_scope = scope;
3310 ifm->ifa_index = ifindex;
3313 static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
3314 unsigned long tstamp, u32 preferred, u32 valid)
3316 struct ifa_cacheinfo ci;
3318 ci.cstamp = (u32)(TIME_DELTA(cstamp, INITIAL_JIFFIES) / HZ * 100
3319 + TIME_DELTA(cstamp, INITIAL_JIFFIES) % HZ * 100 / HZ);
3320 ci.tstamp = (u32)(TIME_DELTA(tstamp, INITIAL_JIFFIES) / HZ * 100
3321 + TIME_DELTA(tstamp, INITIAL_JIFFIES) % HZ * 100 / HZ);
3322 ci.ifa_prefered = preferred;
3323 ci.ifa_valid = valid;
3325 return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
3328 static inline int rt_scope(int ifa_scope)
3330 if (ifa_scope & IFA_HOST)
3331 return RT_SCOPE_HOST;
3332 else if (ifa_scope & IFA_LINK)
3333 return RT_SCOPE_LINK;
3334 else if (ifa_scope & IFA_SITE)
3335 return RT_SCOPE_SITE;
3336 else
3337 return RT_SCOPE_UNIVERSE;
3340 static inline int inet6_ifaddr_msgsize(void)
3342 return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
3343 + nla_total_size(16) /* IFA_ADDRESS */
3344 + nla_total_size(sizeof(struct ifa_cacheinfo));
3347 static int inet6_fill_ifaddr(struct sk_buff *skb, struct inet6_ifaddr *ifa,
3348 u32 pid, u32 seq, int event, unsigned int flags)
3350 struct nlmsghdr *nlh;
3351 u32 preferred, valid;
3353 nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3354 if (nlh == NULL)
3355 return -EMSGSIZE;
3357 put_ifaddrmsg(nlh, ifa->prefix_len, ifa->flags, rt_scope(ifa->scope),
3358 ifa->idev->dev->ifindex);
3360 if (!(ifa->flags&IFA_F_PERMANENT)) {
3361 preferred = ifa->prefered_lft;
3362 valid = ifa->valid_lft;
3363 if (preferred != INFINITY_LIFE_TIME) {
3364 long tval = (jiffies - ifa->tstamp)/HZ;
3365 preferred -= tval;
3366 if (valid != INFINITY_LIFE_TIME)
3367 valid -= tval;
3369 } else {
3370 preferred = INFINITY_LIFE_TIME;
3371 valid = INFINITY_LIFE_TIME;
3374 if (nla_put(skb, IFA_ADDRESS, 16, &ifa->addr) < 0 ||
3375 put_cacheinfo(skb, ifa->cstamp, ifa->tstamp, preferred, valid) < 0) {
3376 nlmsg_cancel(skb, nlh);
3377 return -EMSGSIZE;
3380 return nlmsg_end(skb, nlh);
3383 static int inet6_fill_ifmcaddr(struct sk_buff *skb, struct ifmcaddr6 *ifmca,
3384 u32 pid, u32 seq, int event, u16 flags)
3386 struct nlmsghdr *nlh;
3387 u8 scope = RT_SCOPE_UNIVERSE;
3388 int ifindex = ifmca->idev->dev->ifindex;
3390 if (ipv6_addr_scope(&ifmca->mca_addr) & IFA_SITE)
3391 scope = RT_SCOPE_SITE;
3393 nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3394 if (nlh == NULL)
3395 return -EMSGSIZE;
3397 put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3398 if (nla_put(skb, IFA_MULTICAST, 16, &ifmca->mca_addr) < 0 ||
3399 put_cacheinfo(skb, ifmca->mca_cstamp, ifmca->mca_tstamp,
3400 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3401 nlmsg_cancel(skb, nlh);
3402 return -EMSGSIZE;
3405 return nlmsg_end(skb, nlh);
3408 static int inet6_fill_ifacaddr(struct sk_buff *skb, struct ifacaddr6 *ifaca,
3409 u32 pid, u32 seq, int event, unsigned int flags)
3411 struct nlmsghdr *nlh;
3412 u8 scope = RT_SCOPE_UNIVERSE;
3413 int ifindex = ifaca->aca_idev->dev->ifindex;
3415 if (ipv6_addr_scope(&ifaca->aca_addr) & IFA_SITE)
3416 scope = RT_SCOPE_SITE;
3418 nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3419 if (nlh == NULL)
3420 return -EMSGSIZE;
3422 put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3423 if (nla_put(skb, IFA_ANYCAST, 16, &ifaca->aca_addr) < 0 ||
3424 put_cacheinfo(skb, ifaca->aca_cstamp, ifaca->aca_tstamp,
3425 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3426 nlmsg_cancel(skb, nlh);
3427 return -EMSGSIZE;
3430 return nlmsg_end(skb, nlh);
3433 enum addr_type_t
3435 UNICAST_ADDR,
3436 MULTICAST_ADDR,
3437 ANYCAST_ADDR,
3440 static int inet6_dump_addr(struct sk_buff *skb, struct netlink_callback *cb,
3441 enum addr_type_t type)
3443 int idx, ip_idx;
3444 int s_idx, s_ip_idx;
3445 int err = 1;
3446 struct net_device *dev;
3447 struct inet6_dev *idev = NULL;
3448 struct inet6_ifaddr *ifa;
3449 struct ifmcaddr6 *ifmca;
3450 struct ifacaddr6 *ifaca;
3451 struct net *net = sock_net(skb->sk);
3453 s_idx = cb->args[0];
3454 s_ip_idx = ip_idx = cb->args[1];
3456 idx = 0;
3457 for_each_netdev(net, dev) {
3458 if (idx < s_idx)
3459 goto cont;
3460 if (idx > s_idx)
3461 s_ip_idx = 0;
3462 ip_idx = 0;
3463 if ((idev = in6_dev_get(dev)) == NULL)
3464 goto cont;
3465 read_lock_bh(&idev->lock);
3466 switch (type) {
3467 case UNICAST_ADDR:
3468 /* unicast address incl. temp addr */
3469 for (ifa = idev->addr_list; ifa;
3470 ifa = ifa->if_next, ip_idx++) {
3471 if (ip_idx < s_ip_idx)
3472 continue;
3473 err = inet6_fill_ifaddr(skb, ifa,
3474 NETLINK_CB(cb->skb).pid,
3475 cb->nlh->nlmsg_seq,
3476 RTM_NEWADDR,
3477 NLM_F_MULTI);
3479 break;
3480 case MULTICAST_ADDR:
3481 /* multicast address */
3482 for (ifmca = idev->mc_list; ifmca;
3483 ifmca = ifmca->next, ip_idx++) {
3484 if (ip_idx < s_ip_idx)
3485 continue;
3486 err = inet6_fill_ifmcaddr(skb, ifmca,
3487 NETLINK_CB(cb->skb).pid,
3488 cb->nlh->nlmsg_seq,
3489 RTM_GETMULTICAST,
3490 NLM_F_MULTI);
3492 break;
3493 case ANYCAST_ADDR:
3494 /* anycast address */
3495 for (ifaca = idev->ac_list; ifaca;
3496 ifaca = ifaca->aca_next, ip_idx++) {
3497 if (ip_idx < s_ip_idx)
3498 continue;
3499 err = inet6_fill_ifacaddr(skb, ifaca,
3500 NETLINK_CB(cb->skb).pid,
3501 cb->nlh->nlmsg_seq,
3502 RTM_GETANYCAST,
3503 NLM_F_MULTI);
3505 break;
3506 default:
3507 break;
3509 read_unlock_bh(&idev->lock);
3510 in6_dev_put(idev);
3512 if (err <= 0)
3513 break;
3514 cont:
3515 idx++;
3517 cb->args[0] = idx;
3518 cb->args[1] = ip_idx;
3519 return skb->len;
3522 static int inet6_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
3524 enum addr_type_t type = UNICAST_ADDR;
3526 return inet6_dump_addr(skb, cb, type);
3529 static int inet6_dump_ifmcaddr(struct sk_buff *skb, struct netlink_callback *cb)
3531 enum addr_type_t type = MULTICAST_ADDR;
3533 return inet6_dump_addr(skb, cb, type);
3537 static int inet6_dump_ifacaddr(struct sk_buff *skb, struct netlink_callback *cb)
3539 enum addr_type_t type = ANYCAST_ADDR;
3541 return inet6_dump_addr(skb, cb, type);
3544 static int inet6_rtm_getaddr(struct sk_buff *in_skb, struct nlmsghdr* nlh,
3545 void *arg)
3547 struct net *net = sock_net(in_skb->sk);
3548 struct ifaddrmsg *ifm;
3549 struct nlattr *tb[IFA_MAX+1];
3550 struct in6_addr *addr = NULL;
3551 struct net_device *dev = NULL;
3552 struct inet6_ifaddr *ifa;
3553 struct sk_buff *skb;
3554 int err;
3556 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3557 if (err < 0)
3558 goto errout;
3560 addr = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3561 if (addr == NULL) {
3562 err = -EINVAL;
3563 goto errout;
3566 ifm = nlmsg_data(nlh);
3567 if (ifm->ifa_index)
3568 dev = __dev_get_by_index(net, ifm->ifa_index);
3570 if ((ifa = ipv6_get_ifaddr(net, addr, dev, 1)) == NULL) {
3571 err = -EADDRNOTAVAIL;
3572 goto errout;
3575 if ((skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_KERNEL)) == NULL) {
3576 err = -ENOBUFS;
3577 goto errout_ifa;
3580 err = inet6_fill_ifaddr(skb, ifa, NETLINK_CB(in_skb).pid,
3581 nlh->nlmsg_seq, RTM_NEWADDR, 0);
3582 if (err < 0) {
3583 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
3584 WARN_ON(err == -EMSGSIZE);
3585 kfree_skb(skb);
3586 goto errout_ifa;
3588 err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).pid);
3589 errout_ifa:
3590 in6_ifa_put(ifa);
3591 errout:
3592 return err;
3595 static void inet6_ifa_notify(int event, struct inet6_ifaddr *ifa)
3597 struct sk_buff *skb;
3598 struct net *net = dev_net(ifa->idev->dev);
3599 int err = -ENOBUFS;
3601 skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_ATOMIC);
3602 if (skb == NULL)
3603 goto errout;
3605 err = inet6_fill_ifaddr(skb, ifa, 0, 0, event, 0);
3606 if (err < 0) {
3607 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
3608 WARN_ON(err == -EMSGSIZE);
3609 kfree_skb(skb);
3610 goto errout;
3612 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
3613 return;
3614 errout:
3615 if (err < 0)
3616 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
3619 static inline void ipv6_store_devconf(struct ipv6_devconf *cnf,
3620 __s32 *array, int bytes)
3622 BUG_ON(bytes < (DEVCONF_MAX * 4));
3624 memset(array, 0, bytes);
3625 array[DEVCONF_FORWARDING] = cnf->forwarding;
3626 array[DEVCONF_HOPLIMIT] = cnf->hop_limit;
3627 array[DEVCONF_MTU6] = cnf->mtu6;
3628 array[DEVCONF_ACCEPT_RA] = cnf->accept_ra;
3629 array[DEVCONF_ACCEPT_REDIRECTS] = cnf->accept_redirects;
3630 array[DEVCONF_AUTOCONF] = cnf->autoconf;
3631 array[DEVCONF_DAD_TRANSMITS] = cnf->dad_transmits;
3632 array[DEVCONF_RTR_SOLICITS] = cnf->rtr_solicits;
3633 array[DEVCONF_RTR_SOLICIT_INTERVAL] = cnf->rtr_solicit_interval;
3634 array[DEVCONF_RTR_SOLICIT_DELAY] = cnf->rtr_solicit_delay;
3635 array[DEVCONF_FORCE_MLD_VERSION] = cnf->force_mld_version;
3636 #ifdef CONFIG_IPV6_PRIVACY
3637 array[DEVCONF_USE_TEMPADDR] = cnf->use_tempaddr;
3638 array[DEVCONF_TEMP_VALID_LFT] = cnf->temp_valid_lft;
3639 array[DEVCONF_TEMP_PREFERED_LFT] = cnf->temp_prefered_lft;
3640 array[DEVCONF_REGEN_MAX_RETRY] = cnf->regen_max_retry;
3641 array[DEVCONF_MAX_DESYNC_FACTOR] = cnf->max_desync_factor;
3642 #endif
3643 array[DEVCONF_MAX_ADDRESSES] = cnf->max_addresses;
3644 array[DEVCONF_ACCEPT_RA_DEFRTR] = cnf->accept_ra_defrtr;
3645 array[DEVCONF_ACCEPT_RA_PINFO] = cnf->accept_ra_pinfo;
3646 #ifdef CONFIG_IPV6_ROUTER_PREF
3647 array[DEVCONF_ACCEPT_RA_RTR_PREF] = cnf->accept_ra_rtr_pref;
3648 array[DEVCONF_RTR_PROBE_INTERVAL] = cnf->rtr_probe_interval;
3649 #ifdef CONFIG_IPV6_ROUTE_INFO
3650 array[DEVCONF_ACCEPT_RA_RT_INFO_MAX_PLEN] = cnf->accept_ra_rt_info_max_plen;
3651 #endif
3652 #endif
3653 array[DEVCONF_PROXY_NDP] = cnf->proxy_ndp;
3654 array[DEVCONF_ACCEPT_SOURCE_ROUTE] = cnf->accept_source_route;
3655 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
3656 array[DEVCONF_OPTIMISTIC_DAD] = cnf->optimistic_dad;
3657 #endif
3658 #ifdef CONFIG_IPV6_MROUTE
3659 array[DEVCONF_MC_FORWARDING] = cnf->mc_forwarding;
3660 #endif
3661 array[DEVCONF_DISABLE_IPV6] = cnf->disable_ipv6;
3662 array[DEVCONF_ACCEPT_DAD] = cnf->accept_dad;
3665 static inline size_t inet6_if_nlmsg_size(void)
3667 return NLMSG_ALIGN(sizeof(struct ifinfomsg))
3668 + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
3669 + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
3670 + nla_total_size(4) /* IFLA_MTU */
3671 + nla_total_size(4) /* IFLA_LINK */
3672 + nla_total_size( /* IFLA_PROTINFO */
3673 nla_total_size(4) /* IFLA_INET6_FLAGS */
3674 + nla_total_size(sizeof(struct ifla_cacheinfo))
3675 + nla_total_size(DEVCONF_MAX * 4) /* IFLA_INET6_CONF */
3676 + nla_total_size(IPSTATS_MIB_MAX * 8) /* IFLA_INET6_STATS */
3677 + nla_total_size(ICMP6_MIB_MAX * 8) /* IFLA_INET6_ICMP6STATS */
3681 static inline void __snmp6_fill_stats(u64 *stats, void **mib, int items,
3682 int bytes)
3684 int i;
3685 int pad = bytes - sizeof(u64) * items;
3686 BUG_ON(pad < 0);
3688 /* Use put_unaligned() because stats may not be aligned for u64. */
3689 put_unaligned(items, &stats[0]);
3690 for (i = 1; i < items; i++)
3691 put_unaligned(snmp_fold_field(mib, i), &stats[i]);
3693 memset(&stats[items], 0, pad);
3696 static void snmp6_fill_stats(u64 *stats, struct inet6_dev *idev, int attrtype,
3697 int bytes)
3699 switch(attrtype) {
3700 case IFLA_INET6_STATS:
3701 __snmp6_fill_stats(stats, (void **)idev->stats.ipv6, IPSTATS_MIB_MAX, bytes);
3702 break;
3703 case IFLA_INET6_ICMP6STATS:
3704 __snmp6_fill_stats(stats, (void **)idev->stats.icmpv6, ICMP6_MIB_MAX, bytes);
3705 break;
3709 static int inet6_fill_ifinfo(struct sk_buff *skb, struct inet6_dev *idev,
3710 u32 pid, u32 seq, int event, unsigned int flags)
3712 struct net_device *dev = idev->dev;
3713 struct nlattr *nla;
3714 struct ifinfomsg *hdr;
3715 struct nlmsghdr *nlh;
3716 void *protoinfo;
3717 struct ifla_cacheinfo ci;
3719 nlh = nlmsg_put(skb, pid, seq, event, sizeof(*hdr), flags);
3720 if (nlh == NULL)
3721 return -EMSGSIZE;
3723 hdr = nlmsg_data(nlh);
3724 hdr->ifi_family = AF_INET6;
3725 hdr->__ifi_pad = 0;
3726 hdr->ifi_type = dev->type;
3727 hdr->ifi_index = dev->ifindex;
3728 hdr->ifi_flags = dev_get_flags(dev);
3729 hdr->ifi_change = 0;
3731 NLA_PUT_STRING(skb, IFLA_IFNAME, dev->name);
3733 if (dev->addr_len)
3734 NLA_PUT(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr);
3736 NLA_PUT_U32(skb, IFLA_MTU, dev->mtu);
3737 if (dev->ifindex != dev->iflink)
3738 NLA_PUT_U32(skb, IFLA_LINK, dev->iflink);
3740 protoinfo = nla_nest_start(skb, IFLA_PROTINFO);
3741 if (protoinfo == NULL)
3742 goto nla_put_failure;
3744 NLA_PUT_U32(skb, IFLA_INET6_FLAGS, idev->if_flags);
3746 ci.max_reasm_len = IPV6_MAXPLEN;
3747 ci.tstamp = (__u32)(TIME_DELTA(idev->tstamp, INITIAL_JIFFIES) / HZ * 100
3748 + TIME_DELTA(idev->tstamp, INITIAL_JIFFIES) % HZ * 100 / HZ);
3749 ci.reachable_time = idev->nd_parms->reachable_time;
3750 ci.retrans_time = idev->nd_parms->retrans_time;
3751 NLA_PUT(skb, IFLA_INET6_CACHEINFO, sizeof(ci), &ci);
3753 nla = nla_reserve(skb, IFLA_INET6_CONF, DEVCONF_MAX * sizeof(s32));
3754 if (nla == NULL)
3755 goto nla_put_failure;
3756 ipv6_store_devconf(&idev->cnf, nla_data(nla), nla_len(nla));
3758 /* XXX - MC not implemented */
3760 nla = nla_reserve(skb, IFLA_INET6_STATS, IPSTATS_MIB_MAX * sizeof(u64));
3761 if (nla == NULL)
3762 goto nla_put_failure;
3763 snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_STATS, nla_len(nla));
3765 nla = nla_reserve(skb, IFLA_INET6_ICMP6STATS, ICMP6_MIB_MAX * sizeof(u64));
3766 if (nla == NULL)
3767 goto nla_put_failure;
3768 snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_ICMP6STATS, nla_len(nla));
3770 nla_nest_end(skb, protoinfo);
3771 return nlmsg_end(skb, nlh);
3773 nla_put_failure:
3774 nlmsg_cancel(skb, nlh);
3775 return -EMSGSIZE;
3778 static int inet6_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
3780 struct net *net = sock_net(skb->sk);
3781 int idx, err;
3782 int s_idx = cb->args[0];
3783 struct net_device *dev;
3784 struct inet6_dev *idev;
3786 read_lock(&dev_base_lock);
3787 idx = 0;
3788 for_each_netdev(net, dev) {
3789 if (idx < s_idx)
3790 goto cont;
3791 if ((idev = in6_dev_get(dev)) == NULL)
3792 goto cont;
3793 err = inet6_fill_ifinfo(skb, idev, NETLINK_CB(cb->skb).pid,
3794 cb->nlh->nlmsg_seq, RTM_NEWLINK, NLM_F_MULTI);
3795 in6_dev_put(idev);
3796 if (err <= 0)
3797 break;
3798 cont:
3799 idx++;
3801 read_unlock(&dev_base_lock);
3802 cb->args[0] = idx;
3804 return skb->len;
3807 void inet6_ifinfo_notify(int event, struct inet6_dev *idev)
3809 struct sk_buff *skb;
3810 struct net *net = dev_net(idev->dev);
3811 int err = -ENOBUFS;
3813 skb = nlmsg_new(inet6_if_nlmsg_size(), GFP_ATOMIC);
3814 if (skb == NULL)
3815 goto errout;
3817 err = inet6_fill_ifinfo(skb, idev, 0, 0, event, 0);
3818 if (err < 0) {
3819 /* -EMSGSIZE implies BUG in inet6_if_nlmsg_size() */
3820 WARN_ON(err == -EMSGSIZE);
3821 kfree_skb(skb);
3822 goto errout;
3824 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
3825 return;
3826 errout:
3827 if (err < 0)
3828 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
3831 static inline size_t inet6_prefix_nlmsg_size(void)
3833 return NLMSG_ALIGN(sizeof(struct prefixmsg))
3834 + nla_total_size(sizeof(struct in6_addr))
3835 + nla_total_size(sizeof(struct prefix_cacheinfo));
3838 static int inet6_fill_prefix(struct sk_buff *skb, struct inet6_dev *idev,
3839 struct prefix_info *pinfo, u32 pid, u32 seq,
3840 int event, unsigned int flags)
3842 struct prefixmsg *pmsg;
3843 struct nlmsghdr *nlh;
3844 struct prefix_cacheinfo ci;
3846 nlh = nlmsg_put(skb, pid, seq, event, sizeof(*pmsg), flags);
3847 if (nlh == NULL)
3848 return -EMSGSIZE;
3850 pmsg = nlmsg_data(nlh);
3851 pmsg->prefix_family = AF_INET6;
3852 pmsg->prefix_pad1 = 0;
3853 pmsg->prefix_pad2 = 0;
3854 pmsg->prefix_ifindex = idev->dev->ifindex;
3855 pmsg->prefix_len = pinfo->prefix_len;
3856 pmsg->prefix_type = pinfo->type;
3857 pmsg->prefix_pad3 = 0;
3858 pmsg->prefix_flags = 0;
3859 if (pinfo->onlink)
3860 pmsg->prefix_flags |= IF_PREFIX_ONLINK;
3861 if (pinfo->autoconf)
3862 pmsg->prefix_flags |= IF_PREFIX_AUTOCONF;
3864 NLA_PUT(skb, PREFIX_ADDRESS, sizeof(pinfo->prefix), &pinfo->prefix);
3866 ci.preferred_time = ntohl(pinfo->prefered);
3867 ci.valid_time = ntohl(pinfo->valid);
3868 NLA_PUT(skb, PREFIX_CACHEINFO, sizeof(ci), &ci);
3870 return nlmsg_end(skb, nlh);
3872 nla_put_failure:
3873 nlmsg_cancel(skb, nlh);
3874 return -EMSGSIZE;
3877 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
3878 struct prefix_info *pinfo)
3880 struct sk_buff *skb;
3881 struct net *net = dev_net(idev->dev);
3882 int err = -ENOBUFS;
3884 skb = nlmsg_new(inet6_prefix_nlmsg_size(), GFP_ATOMIC);
3885 if (skb == NULL)
3886 goto errout;
3888 err = inet6_fill_prefix(skb, idev, pinfo, 0, 0, event, 0);
3889 if (err < 0) {
3890 /* -EMSGSIZE implies BUG in inet6_prefix_nlmsg_size() */
3891 WARN_ON(err == -EMSGSIZE);
3892 kfree_skb(skb);
3893 goto errout;
3895 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_PREFIX, NULL, GFP_ATOMIC);
3896 return;
3897 errout:
3898 if (err < 0)
3899 rtnl_set_sk_err(net, RTNLGRP_IPV6_PREFIX, err);
3902 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
3904 inet6_ifa_notify(event ? : RTM_NEWADDR, ifp);
3906 switch (event) {
3907 case RTM_NEWADDR:
3909 * If the address was optimistic
3910 * we inserted the route at the start of
3911 * our DAD process, so we don't need
3912 * to do it again
3914 if (!(ifp->rt->rt6i_node))
3915 ip6_ins_rt(ifp->rt);
3916 if (ifp->idev->cnf.forwarding)
3917 addrconf_join_anycast(ifp);
3918 break;
3919 case RTM_DELADDR:
3920 if (ifp->idev->cnf.forwarding)
3921 addrconf_leave_anycast(ifp);
3922 addrconf_leave_solict(ifp->idev, &ifp->addr);
3923 dst_hold(&ifp->rt->u.dst);
3924 if (ip6_del_rt(ifp->rt))
3925 dst_free(&ifp->rt->u.dst);
3926 break;
3930 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
3932 rcu_read_lock_bh();
3933 if (likely(ifp->idev->dead == 0))
3934 __ipv6_ifa_notify(event, ifp);
3935 rcu_read_unlock_bh();
3938 #ifdef CONFIG_SYSCTL
3940 static
3941 int addrconf_sysctl_forward(ctl_table *ctl, int write, struct file * filp,
3942 void __user *buffer, size_t *lenp, loff_t *ppos)
3944 int *valp = ctl->data;
3945 int val = *valp;
3946 int ret;
3948 ret = proc_dointvec(ctl, write, filp, buffer, lenp, ppos);
3950 if (write)
3951 ret = addrconf_fixup_forwarding(ctl, valp, val);
3952 return ret;
3955 static int addrconf_sysctl_forward_strategy(ctl_table *table,
3956 void __user *oldval,
3957 size_t __user *oldlenp,
3958 void __user *newval, size_t newlen)
3960 int *valp = table->data;
3961 int val = *valp;
3962 int new;
3964 if (!newval || !newlen)
3965 return 0;
3966 if (newlen != sizeof(int))
3967 return -EINVAL;
3968 if (get_user(new, (int __user *)newval))
3969 return -EFAULT;
3970 if (new == *valp)
3971 return 0;
3972 if (oldval && oldlenp) {
3973 size_t len;
3974 if (get_user(len, oldlenp))
3975 return -EFAULT;
3976 if (len) {
3977 if (len > table->maxlen)
3978 len = table->maxlen;
3979 if (copy_to_user(oldval, valp, len))
3980 return -EFAULT;
3981 if (put_user(len, oldlenp))
3982 return -EFAULT;
3986 *valp = new;
3987 return addrconf_fixup_forwarding(table, valp, val);
3990 static void dev_disable_change(struct inet6_dev *idev)
3992 if (!idev || !idev->dev)
3993 return;
3995 if (idev->cnf.disable_ipv6)
3996 addrconf_notify(NULL, NETDEV_DOWN, idev->dev);
3997 else
3998 addrconf_notify(NULL, NETDEV_UP, idev->dev);
4001 static void addrconf_disable_change(struct net *net, __s32 newf)
4003 struct net_device *dev;
4004 struct inet6_dev *idev;
4006 read_lock(&dev_base_lock);
4007 for_each_netdev(net, dev) {
4008 rcu_read_lock();
4009 idev = __in6_dev_get(dev);
4010 if (idev) {
4011 int changed = (!idev->cnf.disable_ipv6) ^ (!newf);
4012 idev->cnf.disable_ipv6 = newf;
4013 if (changed)
4014 dev_disable_change(idev);
4016 rcu_read_unlock();
4018 read_unlock(&dev_base_lock);
4021 static int addrconf_disable_ipv6(struct ctl_table *table, int *p, int old)
4023 struct net *net;
4025 net = (struct net *)table->extra2;
4027 if (p == &net->ipv6.devconf_dflt->disable_ipv6)
4028 return 0;
4030 if (!rtnl_trylock())
4031 return restart_syscall();
4033 if (p == &net->ipv6.devconf_all->disable_ipv6) {
4034 __s32 newf = net->ipv6.devconf_all->disable_ipv6;
4035 net->ipv6.devconf_dflt->disable_ipv6 = newf;
4036 addrconf_disable_change(net, newf);
4037 } else if ((!*p) ^ (!old))
4038 dev_disable_change((struct inet6_dev *)table->extra1);
4040 rtnl_unlock();
4041 return 0;
4044 static
4045 int addrconf_sysctl_disable(ctl_table *ctl, int write, struct file * filp,
4046 void __user *buffer, size_t *lenp, loff_t *ppos)
4048 int *valp = ctl->data;
4049 int val = *valp;
4050 int ret;
4052 ret = proc_dointvec(ctl, write, filp, buffer, lenp, ppos);
4054 if (write)
4055 ret = addrconf_disable_ipv6(ctl, valp, val);
4056 return ret;
4059 static struct addrconf_sysctl_table
4061 struct ctl_table_header *sysctl_header;
4062 ctl_table addrconf_vars[DEVCONF_MAX+1];
4063 char *dev_name;
4064 } addrconf_sysctl __read_mostly = {
4065 .sysctl_header = NULL,
4066 .addrconf_vars = {
4068 .ctl_name = NET_IPV6_FORWARDING,
4069 .procname = "forwarding",
4070 .data = &ipv6_devconf.forwarding,
4071 .maxlen = sizeof(int),
4072 .mode = 0644,
4073 .proc_handler = addrconf_sysctl_forward,
4074 .strategy = addrconf_sysctl_forward_strategy,
4077 .ctl_name = NET_IPV6_HOP_LIMIT,
4078 .procname = "hop_limit",
4079 .data = &ipv6_devconf.hop_limit,
4080 .maxlen = sizeof(int),
4081 .mode = 0644,
4082 .proc_handler = proc_dointvec,
4085 .ctl_name = NET_IPV6_MTU,
4086 .procname = "mtu",
4087 .data = &ipv6_devconf.mtu6,
4088 .maxlen = sizeof(int),
4089 .mode = 0644,
4090 .proc_handler = proc_dointvec,
4093 .ctl_name = NET_IPV6_ACCEPT_RA,
4094 .procname = "accept_ra",
4095 .data = &ipv6_devconf.accept_ra,
4096 .maxlen = sizeof(int),
4097 .mode = 0644,
4098 .proc_handler = proc_dointvec,
4101 .ctl_name = NET_IPV6_ACCEPT_REDIRECTS,
4102 .procname = "accept_redirects",
4103 .data = &ipv6_devconf.accept_redirects,
4104 .maxlen = sizeof(int),
4105 .mode = 0644,
4106 .proc_handler = proc_dointvec,
4109 .ctl_name = NET_IPV6_AUTOCONF,
4110 .procname = "autoconf",
4111 .data = &ipv6_devconf.autoconf,
4112 .maxlen = sizeof(int),
4113 .mode = 0644,
4114 .proc_handler = proc_dointvec,
4117 .ctl_name = NET_IPV6_DAD_TRANSMITS,
4118 .procname = "dad_transmits",
4119 .data = &ipv6_devconf.dad_transmits,
4120 .maxlen = sizeof(int),
4121 .mode = 0644,
4122 .proc_handler = proc_dointvec,
4125 .ctl_name = NET_IPV6_RTR_SOLICITS,
4126 .procname = "router_solicitations",
4127 .data = &ipv6_devconf.rtr_solicits,
4128 .maxlen = sizeof(int),
4129 .mode = 0644,
4130 .proc_handler = proc_dointvec,
4133 .ctl_name = NET_IPV6_RTR_SOLICIT_INTERVAL,
4134 .procname = "router_solicitation_interval",
4135 .data = &ipv6_devconf.rtr_solicit_interval,
4136 .maxlen = sizeof(int),
4137 .mode = 0644,
4138 .proc_handler = proc_dointvec_jiffies,
4139 .strategy = sysctl_jiffies,
4142 .ctl_name = NET_IPV6_RTR_SOLICIT_DELAY,
4143 .procname = "router_solicitation_delay",
4144 .data = &ipv6_devconf.rtr_solicit_delay,
4145 .maxlen = sizeof(int),
4146 .mode = 0644,
4147 .proc_handler = proc_dointvec_jiffies,
4148 .strategy = sysctl_jiffies,
4151 .ctl_name = NET_IPV6_FORCE_MLD_VERSION,
4152 .procname = "force_mld_version",
4153 .data = &ipv6_devconf.force_mld_version,
4154 .maxlen = sizeof(int),
4155 .mode = 0644,
4156 .proc_handler = proc_dointvec,
4158 #ifdef CONFIG_IPV6_PRIVACY
4160 .ctl_name = NET_IPV6_USE_TEMPADDR,
4161 .procname = "use_tempaddr",
4162 .data = &ipv6_devconf.use_tempaddr,
4163 .maxlen = sizeof(int),
4164 .mode = 0644,
4165 .proc_handler = proc_dointvec,
4168 .ctl_name = NET_IPV6_TEMP_VALID_LFT,
4169 .procname = "temp_valid_lft",
4170 .data = &ipv6_devconf.temp_valid_lft,
4171 .maxlen = sizeof(int),
4172 .mode = 0644,
4173 .proc_handler = proc_dointvec,
4176 .ctl_name = NET_IPV6_TEMP_PREFERED_LFT,
4177 .procname = "temp_prefered_lft",
4178 .data = &ipv6_devconf.temp_prefered_lft,
4179 .maxlen = sizeof(int),
4180 .mode = 0644,
4181 .proc_handler = proc_dointvec,
4184 .ctl_name = NET_IPV6_REGEN_MAX_RETRY,
4185 .procname = "regen_max_retry",
4186 .data = &ipv6_devconf.regen_max_retry,
4187 .maxlen = sizeof(int),
4188 .mode = 0644,
4189 .proc_handler = proc_dointvec,
4192 .ctl_name = NET_IPV6_MAX_DESYNC_FACTOR,
4193 .procname = "max_desync_factor",
4194 .data = &ipv6_devconf.max_desync_factor,
4195 .maxlen = sizeof(int),
4196 .mode = 0644,
4197 .proc_handler = proc_dointvec,
4199 #endif
4201 .ctl_name = NET_IPV6_MAX_ADDRESSES,
4202 .procname = "max_addresses",
4203 .data = &ipv6_devconf.max_addresses,
4204 .maxlen = sizeof(int),
4205 .mode = 0644,
4206 .proc_handler = proc_dointvec,
4209 .ctl_name = NET_IPV6_ACCEPT_RA_DEFRTR,
4210 .procname = "accept_ra_defrtr",
4211 .data = &ipv6_devconf.accept_ra_defrtr,
4212 .maxlen = sizeof(int),
4213 .mode = 0644,
4214 .proc_handler = proc_dointvec,
4217 .ctl_name = NET_IPV6_ACCEPT_RA_PINFO,
4218 .procname = "accept_ra_pinfo",
4219 .data = &ipv6_devconf.accept_ra_pinfo,
4220 .maxlen = sizeof(int),
4221 .mode = 0644,
4222 .proc_handler = proc_dointvec,
4224 #ifdef CONFIG_IPV6_ROUTER_PREF
4226 .ctl_name = NET_IPV6_ACCEPT_RA_RTR_PREF,
4227 .procname = "accept_ra_rtr_pref",
4228 .data = &ipv6_devconf.accept_ra_rtr_pref,
4229 .maxlen = sizeof(int),
4230 .mode = 0644,
4231 .proc_handler = proc_dointvec,
4234 .ctl_name = NET_IPV6_RTR_PROBE_INTERVAL,
4235 .procname = "router_probe_interval",
4236 .data = &ipv6_devconf.rtr_probe_interval,
4237 .maxlen = sizeof(int),
4238 .mode = 0644,
4239 .proc_handler = proc_dointvec_jiffies,
4240 .strategy = sysctl_jiffies,
4242 #ifdef CONFIG_IPV6_ROUTE_INFO
4244 .ctl_name = NET_IPV6_ACCEPT_RA_RT_INFO_MAX_PLEN,
4245 .procname = "accept_ra_rt_info_max_plen",
4246 .data = &ipv6_devconf.accept_ra_rt_info_max_plen,
4247 .maxlen = sizeof(int),
4248 .mode = 0644,
4249 .proc_handler = proc_dointvec,
4251 #endif
4252 #endif
4254 .ctl_name = NET_IPV6_PROXY_NDP,
4255 .procname = "proxy_ndp",
4256 .data = &ipv6_devconf.proxy_ndp,
4257 .maxlen = sizeof(int),
4258 .mode = 0644,
4259 .proc_handler = proc_dointvec,
4262 .ctl_name = NET_IPV6_ACCEPT_SOURCE_ROUTE,
4263 .procname = "accept_source_route",
4264 .data = &ipv6_devconf.accept_source_route,
4265 .maxlen = sizeof(int),
4266 .mode = 0644,
4267 .proc_handler = proc_dointvec,
4269 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
4271 .ctl_name = CTL_UNNUMBERED,
4272 .procname = "optimistic_dad",
4273 .data = &ipv6_devconf.optimistic_dad,
4274 .maxlen = sizeof(int),
4275 .mode = 0644,
4276 .proc_handler = proc_dointvec,
4279 #endif
4280 #ifdef CONFIG_IPV6_MROUTE
4282 .ctl_name = CTL_UNNUMBERED,
4283 .procname = "mc_forwarding",
4284 .data = &ipv6_devconf.mc_forwarding,
4285 .maxlen = sizeof(int),
4286 .mode = 0444,
4287 .proc_handler = proc_dointvec,
4289 #endif
4291 .ctl_name = CTL_UNNUMBERED,
4292 .procname = "disable_ipv6",
4293 .data = &ipv6_devconf.disable_ipv6,
4294 .maxlen = sizeof(int),
4295 .mode = 0644,
4296 .proc_handler = addrconf_sysctl_disable,
4297 .strategy = sysctl_intvec,
4300 .ctl_name = CTL_UNNUMBERED,
4301 .procname = "accept_dad",
4302 .data = &ipv6_devconf.accept_dad,
4303 .maxlen = sizeof(int),
4304 .mode = 0644,
4305 .proc_handler = proc_dointvec,
4308 .ctl_name = 0, /* sentinel */
4313 static int __addrconf_sysctl_register(struct net *net, char *dev_name,
4314 int ctl_name, struct inet6_dev *idev, struct ipv6_devconf *p)
4316 int i;
4317 struct addrconf_sysctl_table *t;
4319 #define ADDRCONF_CTL_PATH_DEV 3
4321 struct ctl_path addrconf_ctl_path[] = {
4322 { .procname = "net", .ctl_name = CTL_NET, },
4323 { .procname = "ipv6", .ctl_name = NET_IPV6, },
4324 { .procname = "conf", .ctl_name = NET_IPV6_CONF, },
4325 { /* to be set */ },
4326 { },
4330 t = kmemdup(&addrconf_sysctl, sizeof(*t), GFP_KERNEL);
4331 if (t == NULL)
4332 goto out;
4334 for (i=0; t->addrconf_vars[i].data; i++) {
4335 t->addrconf_vars[i].data += (char*)p - (char*)&ipv6_devconf;
4336 t->addrconf_vars[i].extra1 = idev; /* embedded; no ref */
4337 t->addrconf_vars[i].extra2 = net;
4341 * Make a copy of dev_name, because '.procname' is regarded as const
4342 * by sysctl and we wouldn't want anyone to change it under our feet
4343 * (see SIOCSIFNAME).
4345 t->dev_name = kstrdup(dev_name, GFP_KERNEL);
4346 if (!t->dev_name)
4347 goto free;
4349 addrconf_ctl_path[ADDRCONF_CTL_PATH_DEV].procname = t->dev_name;
4350 addrconf_ctl_path[ADDRCONF_CTL_PATH_DEV].ctl_name = ctl_name;
4352 t->sysctl_header = register_net_sysctl_table(net, addrconf_ctl_path,
4353 t->addrconf_vars);
4354 if (t->sysctl_header == NULL)
4355 goto free_procname;
4357 p->sysctl = t;
4358 return 0;
4360 free_procname:
4361 kfree(t->dev_name);
4362 free:
4363 kfree(t);
4364 out:
4365 return -ENOBUFS;
4368 static void __addrconf_sysctl_unregister(struct ipv6_devconf *p)
4370 struct addrconf_sysctl_table *t;
4372 if (p->sysctl == NULL)
4373 return;
4375 t = p->sysctl;
4376 p->sysctl = NULL;
4377 unregister_sysctl_table(t->sysctl_header);
4378 kfree(t->dev_name);
4379 kfree(t);
4382 static void addrconf_sysctl_register(struct inet6_dev *idev)
4384 neigh_sysctl_register(idev->dev, idev->nd_parms, NET_IPV6,
4385 NET_IPV6_NEIGH, "ipv6",
4386 &ndisc_ifinfo_sysctl_change,
4387 ndisc_ifinfo_sysctl_strategy);
4388 __addrconf_sysctl_register(dev_net(idev->dev), idev->dev->name,
4389 idev->dev->ifindex, idev, &idev->cnf);
4392 static void addrconf_sysctl_unregister(struct inet6_dev *idev)
4394 __addrconf_sysctl_unregister(&idev->cnf);
4395 neigh_sysctl_unregister(idev->nd_parms);
4399 #endif
4401 static int addrconf_init_net(struct net *net)
4403 int err;
4404 struct ipv6_devconf *all, *dflt;
4406 err = -ENOMEM;
4407 all = &ipv6_devconf;
4408 dflt = &ipv6_devconf_dflt;
4410 if (net != &init_net) {
4411 all = kmemdup(all, sizeof(ipv6_devconf), GFP_KERNEL);
4412 if (all == NULL)
4413 goto err_alloc_all;
4415 dflt = kmemdup(dflt, sizeof(ipv6_devconf_dflt), GFP_KERNEL);
4416 if (dflt == NULL)
4417 goto err_alloc_dflt;
4418 } else {
4419 /* these will be inherited by all namespaces */
4420 dflt->autoconf = ipv6_defaults.autoconf;
4421 dflt->disable_ipv6 = ipv6_defaults.disable_ipv6;
4424 net->ipv6.devconf_all = all;
4425 net->ipv6.devconf_dflt = dflt;
4427 #ifdef CONFIG_SYSCTL
4428 err = __addrconf_sysctl_register(net, "all", NET_PROTO_CONF_ALL,
4429 NULL, all);
4430 if (err < 0)
4431 goto err_reg_all;
4433 err = __addrconf_sysctl_register(net, "default", NET_PROTO_CONF_DEFAULT,
4434 NULL, dflt);
4435 if (err < 0)
4436 goto err_reg_dflt;
4437 #endif
4438 return 0;
4440 #ifdef CONFIG_SYSCTL
4441 err_reg_dflt:
4442 __addrconf_sysctl_unregister(all);
4443 err_reg_all:
4444 kfree(dflt);
4445 #endif
4446 err_alloc_dflt:
4447 kfree(all);
4448 err_alloc_all:
4449 return err;
4452 static void addrconf_exit_net(struct net *net)
4454 #ifdef CONFIG_SYSCTL
4455 __addrconf_sysctl_unregister(net->ipv6.devconf_dflt);
4456 __addrconf_sysctl_unregister(net->ipv6.devconf_all);
4457 #endif
4458 if (net != &init_net) {
4459 kfree(net->ipv6.devconf_dflt);
4460 kfree(net->ipv6.devconf_all);
4464 static struct pernet_operations addrconf_ops = {
4465 .init = addrconf_init_net,
4466 .exit = addrconf_exit_net,
4470 * Device notifier
4473 int register_inet6addr_notifier(struct notifier_block *nb)
4475 return atomic_notifier_chain_register(&inet6addr_chain, nb);
4478 EXPORT_SYMBOL(register_inet6addr_notifier);
4480 int unregister_inet6addr_notifier(struct notifier_block *nb)
4482 return atomic_notifier_chain_unregister(&inet6addr_chain,nb);
4485 EXPORT_SYMBOL(unregister_inet6addr_notifier);
4488 * Init / cleanup code
4491 int __init addrconf_init(void)
4493 int err;
4495 if ((err = ipv6_addr_label_init()) < 0) {
4496 printk(KERN_CRIT "IPv6 Addrconf: cannot initialize default policy table: %d.\n",
4497 err);
4498 return err;
4501 register_pernet_subsys(&addrconf_ops);
4503 /* The addrconf netdev notifier requires that loopback_dev
4504 * has it's ipv6 private information allocated and setup
4505 * before it can bring up and give link-local addresses
4506 * to other devices which are up.
4508 * Unfortunately, loopback_dev is not necessarily the first
4509 * entry in the global dev_base list of net devices. In fact,
4510 * it is likely to be the very last entry on that list.
4511 * So this causes the notifier registry below to try and
4512 * give link-local addresses to all devices besides loopback_dev
4513 * first, then loopback_dev, which cases all the non-loopback_dev
4514 * devices to fail to get a link-local address.
4516 * So, as a temporary fix, allocate the ipv6 structure for
4517 * loopback_dev first by hand.
4518 * Longer term, all of the dependencies ipv6 has upon the loopback
4519 * device and it being up should be removed.
4521 rtnl_lock();
4522 if (!ipv6_add_dev(init_net.loopback_dev))
4523 err = -ENOMEM;
4524 rtnl_unlock();
4525 if (err)
4526 goto errlo;
4528 register_netdevice_notifier(&ipv6_dev_notf);
4530 addrconf_verify(0);
4532 err = __rtnl_register(PF_INET6, RTM_GETLINK, NULL, inet6_dump_ifinfo);
4533 if (err < 0)
4534 goto errout;
4536 /* Only the first call to __rtnl_register can fail */
4537 __rtnl_register(PF_INET6, RTM_NEWADDR, inet6_rtm_newaddr, NULL);
4538 __rtnl_register(PF_INET6, RTM_DELADDR, inet6_rtm_deladdr, NULL);
4539 __rtnl_register(PF_INET6, RTM_GETADDR, inet6_rtm_getaddr, inet6_dump_ifaddr);
4540 __rtnl_register(PF_INET6, RTM_GETMULTICAST, NULL, inet6_dump_ifmcaddr);
4541 __rtnl_register(PF_INET6, RTM_GETANYCAST, NULL, inet6_dump_ifacaddr);
4543 ipv6_addr_label_rtnl_register();
4545 return 0;
4546 errout:
4547 unregister_netdevice_notifier(&ipv6_dev_notf);
4548 errlo:
4549 unregister_pernet_subsys(&addrconf_ops);
4551 return err;
4554 void addrconf_cleanup(void)
4556 struct inet6_ifaddr *ifa;
4557 struct net_device *dev;
4558 int i;
4560 unregister_netdevice_notifier(&ipv6_dev_notf);
4561 unregister_pernet_subsys(&addrconf_ops);
4563 rtnl_lock();
4565 /* clean dev list */
4566 for_each_netdev(&init_net, dev) {
4567 if (__in6_dev_get(dev) == NULL)
4568 continue;
4569 addrconf_ifdown(dev, 1);
4571 addrconf_ifdown(init_net.loopback_dev, 2);
4574 * Check hash table.
4576 write_lock_bh(&addrconf_hash_lock);
4577 for (i=0; i < IN6_ADDR_HSIZE; i++) {
4578 for (ifa=inet6_addr_lst[i]; ifa; ) {
4579 struct inet6_ifaddr *bifa;
4581 bifa = ifa;
4582 ifa = ifa->lst_next;
4583 printk(KERN_DEBUG "bug: IPv6 address leakage detected: ifa=%p\n", bifa);
4584 /* Do not free it; something is wrong.
4585 Now we can investigate it with debugger.
4589 write_unlock_bh(&addrconf_hash_lock);
4591 del_timer(&addr_chk_timer);
4592 rtnl_unlock();