USB: cp210x: call generic open last in open
[zen-stable.git] / net / ipv6 / addrconf.c
bloba225d5ee3c2fc877f25e7af417c58de124fe8cb9
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 #include <linux/slab.h>
57 #ifdef CONFIG_SYSCTL
58 #include <linux/sysctl.h>
59 #endif
60 #include <linux/capability.h>
61 #include <linux/delay.h>
62 #include <linux/notifier.h>
63 #include <linux/string.h>
65 #include <net/net_namespace.h>
66 #include <net/sock.h>
67 #include <net/snmp.h>
69 #include <net/ipv6.h>
70 #include <net/protocol.h>
71 #include <net/ndisc.h>
72 #include <net/ip6_route.h>
73 #include <net/addrconf.h>
74 #include <net/tcp.h>
75 #include <net/ip.h>
76 #include <net/netlink.h>
77 #include <net/pkt_sched.h>
78 #include <linux/if_tunnel.h>
79 #include <linux/rtnetlink.h>
81 #ifdef CONFIG_IPV6_PRIVACY
82 #include <linux/random.h>
83 #endif
85 #include <linux/uaccess.h>
86 #include <asm/unaligned.h>
88 #include <linux/proc_fs.h>
89 #include <linux/seq_file.h>
90 #include <linux/export.h>
92 /* Set to 3 to get tracing... */
93 #define ACONF_DEBUG 2
95 #if ACONF_DEBUG >= 3
96 #define ADBG(x) printk x
97 #else
98 #define ADBG(x)
99 #endif
101 #define INFINITY_LIFE_TIME 0xFFFFFFFF
103 static inline u32 cstamp_delta(unsigned long cstamp)
105 return (cstamp - INITIAL_JIFFIES) * 100UL / HZ;
108 #define ADDRCONF_TIMER_FUZZ_MINUS (HZ > 50 ? HZ/50 : 1)
109 #define ADDRCONF_TIMER_FUZZ (HZ / 4)
110 #define ADDRCONF_TIMER_FUZZ_MAX (HZ)
112 #ifdef CONFIG_SYSCTL
113 static void addrconf_sysctl_register(struct inet6_dev *idev);
114 static void addrconf_sysctl_unregister(struct inet6_dev *idev);
115 #else
116 static inline void addrconf_sysctl_register(struct inet6_dev *idev)
120 static inline void addrconf_sysctl_unregister(struct inet6_dev *idev)
123 #endif
125 #ifdef CONFIG_IPV6_PRIVACY
126 static int __ipv6_regen_rndid(struct inet6_dev *idev);
127 static int __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr);
128 static void ipv6_regen_rndid(unsigned long data);
129 #endif
131 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev);
132 static int ipv6_count_addresses(struct inet6_dev *idev);
135 * Configured unicast address hash table
137 static struct hlist_head inet6_addr_lst[IN6_ADDR_HSIZE];
138 static DEFINE_SPINLOCK(addrconf_hash_lock);
140 static void addrconf_verify(unsigned long);
142 static DEFINE_TIMER(addr_chk_timer, addrconf_verify, 0, 0);
143 static DEFINE_SPINLOCK(addrconf_verify_lock);
145 static void addrconf_join_anycast(struct inet6_ifaddr *ifp);
146 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp);
148 static void addrconf_type_change(struct net_device *dev,
149 unsigned long event);
150 static int addrconf_ifdown(struct net_device *dev, int how);
152 static void addrconf_dad_start(struct inet6_ifaddr *ifp, u32 flags);
153 static void addrconf_dad_timer(unsigned long data);
154 static void addrconf_dad_completed(struct inet6_ifaddr *ifp);
155 static void addrconf_dad_run(struct inet6_dev *idev);
156 static void addrconf_rs_timer(unsigned long data);
157 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
158 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
160 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
161 struct prefix_info *pinfo);
162 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
163 struct net_device *dev);
165 static ATOMIC_NOTIFIER_HEAD(inet6addr_chain);
167 static struct ipv6_devconf ipv6_devconf __read_mostly = {
168 .forwarding = 0,
169 .hop_limit = IPV6_DEFAULT_HOPLIMIT,
170 .mtu6 = IPV6_MIN_MTU,
171 .accept_ra = 1,
172 .accept_redirects = 1,
173 .autoconf = 1,
174 .force_mld_version = 0,
175 .dad_transmits = 1,
176 .rtr_solicits = MAX_RTR_SOLICITATIONS,
177 .rtr_solicit_interval = RTR_SOLICITATION_INTERVAL,
178 .rtr_solicit_delay = MAX_RTR_SOLICITATION_DELAY,
179 #ifdef CONFIG_IPV6_PRIVACY
180 .use_tempaddr = 0,
181 .temp_valid_lft = TEMP_VALID_LIFETIME,
182 .temp_prefered_lft = TEMP_PREFERRED_LIFETIME,
183 .regen_max_retry = REGEN_MAX_RETRY,
184 .max_desync_factor = MAX_DESYNC_FACTOR,
185 #endif
186 .max_addresses = IPV6_MAX_ADDRESSES,
187 .accept_ra_defrtr = 1,
188 .accept_ra_pinfo = 1,
189 #ifdef CONFIG_IPV6_ROUTER_PREF
190 .accept_ra_rtr_pref = 1,
191 .rtr_probe_interval = 60 * HZ,
192 #ifdef CONFIG_IPV6_ROUTE_INFO
193 .accept_ra_rt_info_max_plen = 0,
194 #endif
195 #endif
196 .proxy_ndp = 0,
197 .accept_source_route = 0, /* we do not accept RH0 by default. */
198 .disable_ipv6 = 0,
199 .accept_dad = 1,
202 static struct ipv6_devconf ipv6_devconf_dflt __read_mostly = {
203 .forwarding = 0,
204 .hop_limit = IPV6_DEFAULT_HOPLIMIT,
205 .mtu6 = IPV6_MIN_MTU,
206 .accept_ra = 1,
207 .accept_redirects = 1,
208 .autoconf = 1,
209 .dad_transmits = 1,
210 .rtr_solicits = MAX_RTR_SOLICITATIONS,
211 .rtr_solicit_interval = RTR_SOLICITATION_INTERVAL,
212 .rtr_solicit_delay = MAX_RTR_SOLICITATION_DELAY,
213 #ifdef CONFIG_IPV6_PRIVACY
214 .use_tempaddr = 0,
215 .temp_valid_lft = TEMP_VALID_LIFETIME,
216 .temp_prefered_lft = TEMP_PREFERRED_LIFETIME,
217 .regen_max_retry = REGEN_MAX_RETRY,
218 .max_desync_factor = MAX_DESYNC_FACTOR,
219 #endif
220 .max_addresses = IPV6_MAX_ADDRESSES,
221 .accept_ra_defrtr = 1,
222 .accept_ra_pinfo = 1,
223 #ifdef CONFIG_IPV6_ROUTER_PREF
224 .accept_ra_rtr_pref = 1,
225 .rtr_probe_interval = 60 * HZ,
226 #ifdef CONFIG_IPV6_ROUTE_INFO
227 .accept_ra_rt_info_max_plen = 0,
228 #endif
229 #endif
230 .proxy_ndp = 0,
231 .accept_source_route = 0, /* we do not accept RH0 by default. */
232 .disable_ipv6 = 0,
233 .accept_dad = 1,
236 /* IPv6 Wildcard Address and Loopback Address defined by RFC2553 */
237 const struct in6_addr in6addr_any = IN6ADDR_ANY_INIT;
238 const struct in6_addr in6addr_loopback = IN6ADDR_LOOPBACK_INIT;
239 const struct in6_addr in6addr_linklocal_allnodes = IN6ADDR_LINKLOCAL_ALLNODES_INIT;
240 const struct in6_addr in6addr_linklocal_allrouters = IN6ADDR_LINKLOCAL_ALLROUTERS_INIT;
242 /* Check if a valid qdisc is available */
243 static inline bool addrconf_qdisc_ok(const struct net_device *dev)
245 return !qdisc_tx_is_noop(dev);
248 /* Check if a route is valid prefix route */
249 static inline int addrconf_is_prefix_route(const struct rt6_info *rt)
251 return (rt->rt6i_flags & (RTF_GATEWAY | RTF_DEFAULT)) == 0;
254 static void addrconf_del_timer(struct inet6_ifaddr *ifp)
256 if (del_timer(&ifp->timer))
257 __in6_ifa_put(ifp);
260 enum addrconf_timer_t {
261 AC_NONE,
262 AC_DAD,
263 AC_RS,
266 static void addrconf_mod_timer(struct inet6_ifaddr *ifp,
267 enum addrconf_timer_t what,
268 unsigned long when)
270 if (!del_timer(&ifp->timer))
271 in6_ifa_hold(ifp);
273 switch (what) {
274 case AC_DAD:
275 ifp->timer.function = addrconf_dad_timer;
276 break;
277 case AC_RS:
278 ifp->timer.function = addrconf_rs_timer;
279 break;
280 default:
281 break;
283 ifp->timer.expires = jiffies + when;
284 add_timer(&ifp->timer);
287 static int snmp6_alloc_dev(struct inet6_dev *idev)
289 if (snmp_mib_init((void __percpu **)idev->stats.ipv6,
290 sizeof(struct ipstats_mib),
291 __alignof__(struct ipstats_mib)) < 0)
292 goto err_ip;
293 idev->stats.icmpv6dev = kzalloc(sizeof(struct icmpv6_mib_device),
294 GFP_KERNEL);
295 if (!idev->stats.icmpv6dev)
296 goto err_icmp;
297 idev->stats.icmpv6msgdev = kzalloc(sizeof(struct icmpv6msg_mib_device),
298 GFP_KERNEL);
299 if (!idev->stats.icmpv6msgdev)
300 goto err_icmpmsg;
302 return 0;
304 err_icmpmsg:
305 kfree(idev->stats.icmpv6dev);
306 err_icmp:
307 snmp_mib_free((void __percpu **)idev->stats.ipv6);
308 err_ip:
309 return -ENOMEM;
312 static void snmp6_free_dev(struct inet6_dev *idev)
314 kfree(idev->stats.icmpv6msgdev);
315 kfree(idev->stats.icmpv6dev);
316 snmp_mib_free((void __percpu **)idev->stats.ipv6);
319 /* Nobody refers to this device, we may destroy it. */
321 void in6_dev_finish_destroy(struct inet6_dev *idev)
323 struct net_device *dev = idev->dev;
325 WARN_ON(!list_empty(&idev->addr_list));
326 WARN_ON(idev->mc_list != NULL);
328 #ifdef NET_REFCNT_DEBUG
329 printk(KERN_DEBUG "in6_dev_finish_destroy: %s\n", dev ? dev->name : "NIL");
330 #endif
331 dev_put(dev);
332 if (!idev->dead) {
333 pr_warning("Freeing alive inet6 device %p\n", idev);
334 return;
336 snmp6_free_dev(idev);
337 kfree_rcu(idev, rcu);
340 EXPORT_SYMBOL(in6_dev_finish_destroy);
342 static struct inet6_dev * ipv6_add_dev(struct net_device *dev)
344 struct inet6_dev *ndev;
346 ASSERT_RTNL();
348 if (dev->mtu < IPV6_MIN_MTU)
349 return NULL;
351 ndev = kzalloc(sizeof(struct inet6_dev), GFP_KERNEL);
353 if (ndev == NULL)
354 return NULL;
356 rwlock_init(&ndev->lock);
357 ndev->dev = dev;
358 INIT_LIST_HEAD(&ndev->addr_list);
360 memcpy(&ndev->cnf, dev_net(dev)->ipv6.devconf_dflt, sizeof(ndev->cnf));
361 ndev->cnf.mtu6 = dev->mtu;
362 ndev->cnf.sysctl = NULL;
363 ndev->nd_parms = neigh_parms_alloc(dev, &nd_tbl);
364 if (ndev->nd_parms == NULL) {
365 kfree(ndev);
366 return NULL;
368 if (ndev->cnf.forwarding)
369 dev_disable_lro(dev);
370 /* We refer to the device */
371 dev_hold(dev);
373 if (snmp6_alloc_dev(ndev) < 0) {
374 ADBG((KERN_WARNING
375 "%s(): cannot allocate memory for statistics; dev=%s.\n",
376 __func__, dev->name));
377 neigh_parms_release(&nd_tbl, ndev->nd_parms);
378 dev_put(dev);
379 kfree(ndev);
380 return NULL;
383 if (snmp6_register_dev(ndev) < 0) {
384 ADBG((KERN_WARNING
385 "%s(): cannot create /proc/net/dev_snmp6/%s\n",
386 __func__, dev->name));
387 neigh_parms_release(&nd_tbl, ndev->nd_parms);
388 ndev->dead = 1;
389 in6_dev_finish_destroy(ndev);
390 return NULL;
393 /* One reference from device. We must do this before
394 * we invoke __ipv6_regen_rndid().
396 in6_dev_hold(ndev);
398 if (dev->flags & (IFF_NOARP | IFF_LOOPBACK))
399 ndev->cnf.accept_dad = -1;
401 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
402 if (dev->type == ARPHRD_SIT && (dev->priv_flags & IFF_ISATAP)) {
403 printk(KERN_INFO
404 "%s: Disabled Multicast RS\n",
405 dev->name);
406 ndev->cnf.rtr_solicits = 0;
408 #endif
410 #ifdef CONFIG_IPV6_PRIVACY
411 INIT_LIST_HEAD(&ndev->tempaddr_list);
412 setup_timer(&ndev->regen_timer, ipv6_regen_rndid, (unsigned long)ndev);
413 if ((dev->flags&IFF_LOOPBACK) ||
414 dev->type == ARPHRD_TUNNEL ||
415 dev->type == ARPHRD_TUNNEL6 ||
416 dev->type == ARPHRD_SIT ||
417 dev->type == ARPHRD_NONE) {
418 ndev->cnf.use_tempaddr = -1;
419 } else {
420 in6_dev_hold(ndev);
421 ipv6_regen_rndid((unsigned long) ndev);
423 #endif
425 if (netif_running(dev) && addrconf_qdisc_ok(dev))
426 ndev->if_flags |= IF_READY;
428 ipv6_mc_init_dev(ndev);
429 ndev->tstamp = jiffies;
430 addrconf_sysctl_register(ndev);
431 /* protected by rtnl_lock */
432 rcu_assign_pointer(dev->ip6_ptr, ndev);
434 /* Join all-node multicast group */
435 ipv6_dev_mc_inc(dev, &in6addr_linklocal_allnodes);
437 return ndev;
440 static struct inet6_dev * ipv6_find_idev(struct net_device *dev)
442 struct inet6_dev *idev;
444 ASSERT_RTNL();
446 idev = __in6_dev_get(dev);
447 if (!idev) {
448 idev = ipv6_add_dev(dev);
449 if (!idev)
450 return NULL;
453 if (dev->flags&IFF_UP)
454 ipv6_mc_up(idev);
455 return idev;
458 #ifdef CONFIG_SYSCTL
459 static void dev_forward_change(struct inet6_dev *idev)
461 struct net_device *dev;
462 struct inet6_ifaddr *ifa;
464 if (!idev)
465 return;
466 dev = idev->dev;
467 if (idev->cnf.forwarding)
468 dev_disable_lro(dev);
469 if (dev && (dev->flags & IFF_MULTICAST)) {
470 if (idev->cnf.forwarding)
471 ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
472 else
473 ipv6_dev_mc_dec(dev, &in6addr_linklocal_allrouters);
476 list_for_each_entry(ifa, &idev->addr_list, if_list) {
477 if (ifa->flags&IFA_F_TENTATIVE)
478 continue;
479 if (idev->cnf.forwarding)
480 addrconf_join_anycast(ifa);
481 else
482 addrconf_leave_anycast(ifa);
487 static void addrconf_forward_change(struct net *net, __s32 newf)
489 struct net_device *dev;
490 struct inet6_dev *idev;
492 rcu_read_lock();
493 for_each_netdev_rcu(net, dev) {
494 idev = __in6_dev_get(dev);
495 if (idev) {
496 int changed = (!idev->cnf.forwarding) ^ (!newf);
497 idev->cnf.forwarding = newf;
498 if (changed)
499 dev_forward_change(idev);
502 rcu_read_unlock();
505 static int addrconf_fixup_forwarding(struct ctl_table *table, int *p, int old)
507 struct net *net;
509 net = (struct net *)table->extra2;
510 if (p == &net->ipv6.devconf_dflt->forwarding)
511 return 0;
513 if (!rtnl_trylock()) {
514 /* Restore the original values before restarting */
515 *p = old;
516 return restart_syscall();
519 if (p == &net->ipv6.devconf_all->forwarding) {
520 __s32 newf = net->ipv6.devconf_all->forwarding;
521 net->ipv6.devconf_dflt->forwarding = newf;
522 addrconf_forward_change(net, newf);
523 } else if ((!*p) ^ (!old))
524 dev_forward_change((struct inet6_dev *)table->extra1);
525 rtnl_unlock();
527 if (*p)
528 rt6_purge_dflt_routers(net);
529 return 1;
531 #endif
533 /* Nobody refers to this ifaddr, destroy it */
534 void inet6_ifa_finish_destroy(struct inet6_ifaddr *ifp)
536 WARN_ON(!hlist_unhashed(&ifp->addr_lst));
538 #ifdef NET_REFCNT_DEBUG
539 printk(KERN_DEBUG "inet6_ifa_finish_destroy\n");
540 #endif
542 in6_dev_put(ifp->idev);
544 if (del_timer(&ifp->timer))
545 pr_notice("Timer is still running, when freeing ifa=%p\n", ifp);
547 if (ifp->state != INET6_IFADDR_STATE_DEAD) {
548 pr_warning("Freeing alive inet6 address %p\n", ifp);
549 return;
551 dst_release(&ifp->rt->dst);
553 kfree_rcu(ifp, rcu);
556 static void
557 ipv6_link_dev_addr(struct inet6_dev *idev, struct inet6_ifaddr *ifp)
559 struct list_head *p;
560 int ifp_scope = ipv6_addr_src_scope(&ifp->addr);
563 * Each device address list is sorted in order of scope -
564 * global before linklocal.
566 list_for_each(p, &idev->addr_list) {
567 struct inet6_ifaddr *ifa
568 = list_entry(p, struct inet6_ifaddr, if_list);
569 if (ifp_scope >= ipv6_addr_src_scope(&ifa->addr))
570 break;
573 list_add_tail(&ifp->if_list, p);
576 static u32 ipv6_addr_hash(const struct in6_addr *addr)
579 * We perform the hash function over the last 64 bits of the address
580 * This will include the IEEE address token on links that support it.
582 return jhash_2words((__force u32)addr->s6_addr32[2],
583 (__force u32)addr->s6_addr32[3], 0)
584 & (IN6_ADDR_HSIZE - 1);
587 /* On success it returns ifp with increased reference count */
589 static struct inet6_ifaddr *
590 ipv6_add_addr(struct inet6_dev *idev, const struct in6_addr *addr, int pfxlen,
591 int scope, u32 flags)
593 struct inet6_ifaddr *ifa = NULL;
594 struct rt6_info *rt;
595 unsigned int hash;
596 int err = 0;
597 int addr_type = ipv6_addr_type(addr);
599 if (addr_type == IPV6_ADDR_ANY ||
600 addr_type & IPV6_ADDR_MULTICAST ||
601 (!(idev->dev->flags & IFF_LOOPBACK) &&
602 addr_type & IPV6_ADDR_LOOPBACK))
603 return ERR_PTR(-EADDRNOTAVAIL);
605 rcu_read_lock_bh();
606 if (idev->dead) {
607 err = -ENODEV; /*XXX*/
608 goto out2;
611 if (idev->cnf.disable_ipv6) {
612 err = -EACCES;
613 goto out2;
616 spin_lock(&addrconf_hash_lock);
618 /* Ignore adding duplicate addresses on an interface */
619 if (ipv6_chk_same_addr(dev_net(idev->dev), addr, idev->dev)) {
620 ADBG(("ipv6_add_addr: already assigned\n"));
621 err = -EEXIST;
622 goto out;
625 ifa = kzalloc(sizeof(struct inet6_ifaddr), GFP_ATOMIC);
627 if (ifa == NULL) {
628 ADBG(("ipv6_add_addr: malloc failed\n"));
629 err = -ENOBUFS;
630 goto out;
633 rt = addrconf_dst_alloc(idev, addr, false);
634 if (IS_ERR(rt)) {
635 err = PTR_ERR(rt);
636 goto out;
639 ifa->addr = *addr;
641 spin_lock_init(&ifa->lock);
642 spin_lock_init(&ifa->state_lock);
643 init_timer(&ifa->timer);
644 INIT_HLIST_NODE(&ifa->addr_lst);
645 ifa->timer.data = (unsigned long) ifa;
646 ifa->scope = scope;
647 ifa->prefix_len = pfxlen;
648 ifa->flags = flags | IFA_F_TENTATIVE;
649 ifa->cstamp = ifa->tstamp = jiffies;
651 ifa->rt = rt;
653 ifa->idev = idev;
654 in6_dev_hold(idev);
655 /* For caller */
656 in6_ifa_hold(ifa);
658 /* Add to big hash table */
659 hash = ipv6_addr_hash(addr);
661 hlist_add_head_rcu(&ifa->addr_lst, &inet6_addr_lst[hash]);
662 spin_unlock(&addrconf_hash_lock);
664 write_lock(&idev->lock);
665 /* Add to inet6_dev unicast addr list. */
666 ipv6_link_dev_addr(idev, ifa);
668 #ifdef CONFIG_IPV6_PRIVACY
669 if (ifa->flags&IFA_F_TEMPORARY) {
670 list_add(&ifa->tmp_list, &idev->tempaddr_list);
671 in6_ifa_hold(ifa);
673 #endif
675 in6_ifa_hold(ifa);
676 write_unlock(&idev->lock);
677 out2:
678 rcu_read_unlock_bh();
680 if (likely(err == 0))
681 atomic_notifier_call_chain(&inet6addr_chain, NETDEV_UP, ifa);
682 else {
683 kfree(ifa);
684 ifa = ERR_PTR(err);
687 return ifa;
688 out:
689 spin_unlock(&addrconf_hash_lock);
690 goto out2;
693 /* This function wants to get referenced ifp and releases it before return */
695 static void ipv6_del_addr(struct inet6_ifaddr *ifp)
697 struct inet6_ifaddr *ifa, *ifn;
698 struct inet6_dev *idev = ifp->idev;
699 int state;
700 int deleted = 0, onlink = 0;
701 unsigned long expires = jiffies;
703 spin_lock_bh(&ifp->state_lock);
704 state = ifp->state;
705 ifp->state = INET6_IFADDR_STATE_DEAD;
706 spin_unlock_bh(&ifp->state_lock);
708 if (state == INET6_IFADDR_STATE_DEAD)
709 goto out;
711 spin_lock_bh(&addrconf_hash_lock);
712 hlist_del_init_rcu(&ifp->addr_lst);
713 spin_unlock_bh(&addrconf_hash_lock);
715 write_lock_bh(&idev->lock);
716 #ifdef CONFIG_IPV6_PRIVACY
717 if (ifp->flags&IFA_F_TEMPORARY) {
718 list_del(&ifp->tmp_list);
719 if (ifp->ifpub) {
720 in6_ifa_put(ifp->ifpub);
721 ifp->ifpub = NULL;
723 __in6_ifa_put(ifp);
725 #endif
727 list_for_each_entry_safe(ifa, ifn, &idev->addr_list, if_list) {
728 if (ifa == ifp) {
729 list_del_init(&ifp->if_list);
730 __in6_ifa_put(ifp);
732 if (!(ifp->flags & IFA_F_PERMANENT) || onlink > 0)
733 break;
734 deleted = 1;
735 continue;
736 } else if (ifp->flags & IFA_F_PERMANENT) {
737 if (ipv6_prefix_equal(&ifa->addr, &ifp->addr,
738 ifp->prefix_len)) {
739 if (ifa->flags & IFA_F_PERMANENT) {
740 onlink = 1;
741 if (deleted)
742 break;
743 } else {
744 unsigned long lifetime;
746 if (!onlink)
747 onlink = -1;
749 spin_lock(&ifa->lock);
751 lifetime = addrconf_timeout_fixup(ifa->valid_lft, HZ);
753 * Note: Because this address is
754 * not permanent, lifetime <
755 * LONG_MAX / HZ here.
757 if (time_before(expires,
758 ifa->tstamp + lifetime * HZ))
759 expires = ifa->tstamp + lifetime * HZ;
760 spin_unlock(&ifa->lock);
765 write_unlock_bh(&idev->lock);
767 addrconf_del_timer(ifp);
769 ipv6_ifa_notify(RTM_DELADDR, ifp);
771 atomic_notifier_call_chain(&inet6addr_chain, NETDEV_DOWN, ifp);
774 * Purge or update corresponding prefix
776 * 1) we don't purge prefix here if address was not permanent.
777 * prefix is managed by its own lifetime.
778 * 2) if there're no addresses, delete prefix.
779 * 3) if there're still other permanent address(es),
780 * corresponding prefix is still permanent.
781 * 4) otherwise, update prefix lifetime to the
782 * longest valid lifetime among the corresponding
783 * addresses on the device.
784 * Note: subsequent RA will update lifetime.
786 * --yoshfuji
788 if ((ifp->flags & IFA_F_PERMANENT) && onlink < 1) {
789 struct in6_addr prefix;
790 struct rt6_info *rt;
791 struct net *net = dev_net(ifp->idev->dev);
792 ipv6_addr_prefix(&prefix, &ifp->addr, ifp->prefix_len);
793 rt = rt6_lookup(net, &prefix, NULL, ifp->idev->dev->ifindex, 1);
795 if (rt && addrconf_is_prefix_route(rt)) {
796 if (onlink == 0) {
797 ip6_del_rt(rt);
798 rt = NULL;
799 } else if (!(rt->rt6i_flags & RTF_EXPIRES)) {
800 rt->dst.expires = expires;
801 rt->rt6i_flags |= RTF_EXPIRES;
804 dst_release(&rt->dst);
807 /* clean up prefsrc entries */
808 rt6_remove_prefsrc(ifp);
809 out:
810 in6_ifa_put(ifp);
813 #ifdef CONFIG_IPV6_PRIVACY
814 static int ipv6_create_tempaddr(struct inet6_ifaddr *ifp, struct inet6_ifaddr *ift)
816 struct inet6_dev *idev = ifp->idev;
817 struct in6_addr addr, *tmpaddr;
818 unsigned long tmp_prefered_lft, tmp_valid_lft, tmp_tstamp, age;
819 unsigned long regen_advance;
820 int tmp_plen;
821 int ret = 0;
822 int max_addresses;
823 u32 addr_flags;
824 unsigned long now = jiffies;
826 write_lock(&idev->lock);
827 if (ift) {
828 spin_lock_bh(&ift->lock);
829 memcpy(&addr.s6_addr[8], &ift->addr.s6_addr[8], 8);
830 spin_unlock_bh(&ift->lock);
831 tmpaddr = &addr;
832 } else {
833 tmpaddr = NULL;
835 retry:
836 in6_dev_hold(idev);
837 if (idev->cnf.use_tempaddr <= 0) {
838 write_unlock(&idev->lock);
839 printk(KERN_INFO
840 "ipv6_create_tempaddr(): use_tempaddr is disabled.\n");
841 in6_dev_put(idev);
842 ret = -1;
843 goto out;
845 spin_lock_bh(&ifp->lock);
846 if (ifp->regen_count++ >= idev->cnf.regen_max_retry) {
847 idev->cnf.use_tempaddr = -1; /*XXX*/
848 spin_unlock_bh(&ifp->lock);
849 write_unlock(&idev->lock);
850 printk(KERN_WARNING
851 "ipv6_create_tempaddr(): regeneration time exceeded. disabled temporary address support.\n");
852 in6_dev_put(idev);
853 ret = -1;
854 goto out;
856 in6_ifa_hold(ifp);
857 memcpy(addr.s6_addr, ifp->addr.s6_addr, 8);
858 if (__ipv6_try_regen_rndid(idev, tmpaddr) < 0) {
859 spin_unlock_bh(&ifp->lock);
860 write_unlock(&idev->lock);
861 printk(KERN_WARNING
862 "ipv6_create_tempaddr(): regeneration of randomized interface id failed.\n");
863 in6_ifa_put(ifp);
864 in6_dev_put(idev);
865 ret = -1;
866 goto out;
868 memcpy(&addr.s6_addr[8], idev->rndid, 8);
869 age = (now - ifp->tstamp) / HZ;
870 tmp_valid_lft = min_t(__u32,
871 ifp->valid_lft,
872 idev->cnf.temp_valid_lft + age);
873 tmp_prefered_lft = min_t(__u32,
874 ifp->prefered_lft,
875 idev->cnf.temp_prefered_lft + age -
876 idev->cnf.max_desync_factor);
877 tmp_plen = ifp->prefix_len;
878 max_addresses = idev->cnf.max_addresses;
879 tmp_tstamp = ifp->tstamp;
880 spin_unlock_bh(&ifp->lock);
882 regen_advance = idev->cnf.regen_max_retry *
883 idev->cnf.dad_transmits *
884 idev->nd_parms->retrans_time / HZ;
885 write_unlock(&idev->lock);
887 /* A temporary address is created only if this calculated Preferred
888 * Lifetime is greater than REGEN_ADVANCE time units. In particular,
889 * an implementation must not create a temporary address with a zero
890 * Preferred Lifetime.
892 if (tmp_prefered_lft <= regen_advance) {
893 in6_ifa_put(ifp);
894 in6_dev_put(idev);
895 ret = -1;
896 goto out;
899 addr_flags = IFA_F_TEMPORARY;
900 /* set in addrconf_prefix_rcv() */
901 if (ifp->flags & IFA_F_OPTIMISTIC)
902 addr_flags |= IFA_F_OPTIMISTIC;
904 ift = !max_addresses ||
905 ipv6_count_addresses(idev) < max_addresses ?
906 ipv6_add_addr(idev, &addr, tmp_plen,
907 ipv6_addr_type(&addr)&IPV6_ADDR_SCOPE_MASK,
908 addr_flags) : NULL;
909 if (!ift || IS_ERR(ift)) {
910 in6_ifa_put(ifp);
911 in6_dev_put(idev);
912 printk(KERN_INFO
913 "ipv6_create_tempaddr(): retry temporary address regeneration.\n");
914 tmpaddr = &addr;
915 write_lock(&idev->lock);
916 goto retry;
919 spin_lock_bh(&ift->lock);
920 ift->ifpub = ifp;
921 ift->valid_lft = tmp_valid_lft;
922 ift->prefered_lft = tmp_prefered_lft;
923 ift->cstamp = now;
924 ift->tstamp = tmp_tstamp;
925 spin_unlock_bh(&ift->lock);
927 addrconf_dad_start(ift, 0);
928 in6_ifa_put(ift);
929 in6_dev_put(idev);
930 out:
931 return ret;
933 #endif
936 * Choose an appropriate source address (RFC3484)
938 enum {
939 IPV6_SADDR_RULE_INIT = 0,
940 IPV6_SADDR_RULE_LOCAL,
941 IPV6_SADDR_RULE_SCOPE,
942 IPV6_SADDR_RULE_PREFERRED,
943 #ifdef CONFIG_IPV6_MIP6
944 IPV6_SADDR_RULE_HOA,
945 #endif
946 IPV6_SADDR_RULE_OIF,
947 IPV6_SADDR_RULE_LABEL,
948 #ifdef CONFIG_IPV6_PRIVACY
949 IPV6_SADDR_RULE_PRIVACY,
950 #endif
951 IPV6_SADDR_RULE_ORCHID,
952 IPV6_SADDR_RULE_PREFIX,
953 IPV6_SADDR_RULE_MAX
956 struct ipv6_saddr_score {
957 int rule;
958 int addr_type;
959 struct inet6_ifaddr *ifa;
960 DECLARE_BITMAP(scorebits, IPV6_SADDR_RULE_MAX);
961 int scopedist;
962 int matchlen;
965 struct ipv6_saddr_dst {
966 const struct in6_addr *addr;
967 int ifindex;
968 int scope;
969 int label;
970 unsigned int prefs;
973 static inline int ipv6_saddr_preferred(int type)
975 if (type & (IPV6_ADDR_MAPPED|IPV6_ADDR_COMPATv4|IPV6_ADDR_LOOPBACK))
976 return 1;
977 return 0;
980 static int ipv6_get_saddr_eval(struct net *net,
981 struct ipv6_saddr_score *score,
982 struct ipv6_saddr_dst *dst,
983 int i)
985 int ret;
987 if (i <= score->rule) {
988 switch (i) {
989 case IPV6_SADDR_RULE_SCOPE:
990 ret = score->scopedist;
991 break;
992 case IPV6_SADDR_RULE_PREFIX:
993 ret = score->matchlen;
994 break;
995 default:
996 ret = !!test_bit(i, score->scorebits);
998 goto out;
1001 switch (i) {
1002 case IPV6_SADDR_RULE_INIT:
1003 /* Rule 0: remember if hiscore is not ready yet */
1004 ret = !!score->ifa;
1005 break;
1006 case IPV6_SADDR_RULE_LOCAL:
1007 /* Rule 1: Prefer same address */
1008 ret = ipv6_addr_equal(&score->ifa->addr, dst->addr);
1009 break;
1010 case IPV6_SADDR_RULE_SCOPE:
1011 /* Rule 2: Prefer appropriate scope
1013 * ret
1015 * -1 | d 15
1016 * ---+--+-+---> scope
1018 * | d is scope of the destination.
1019 * B-d | \
1020 * | \ <- smaller scope is better if
1021 * B-15 | \ if scope is enough for destinaion.
1022 * | ret = B - scope (-1 <= scope >= d <= 15).
1023 * d-C-1 | /
1024 * |/ <- greater is better
1025 * -C / if scope is not enough for destination.
1026 * /| ret = scope - C (-1 <= d < scope <= 15).
1028 * d - C - 1 < B -15 (for all -1 <= d <= 15).
1029 * C > d + 14 - B >= 15 + 14 - B = 29 - B.
1030 * Assume B = 0 and we get C > 29.
1032 ret = __ipv6_addr_src_scope(score->addr_type);
1033 if (ret >= dst->scope)
1034 ret = -ret;
1035 else
1036 ret -= 128; /* 30 is enough */
1037 score->scopedist = ret;
1038 break;
1039 case IPV6_SADDR_RULE_PREFERRED:
1040 /* Rule 3: Avoid deprecated and optimistic addresses */
1041 ret = ipv6_saddr_preferred(score->addr_type) ||
1042 !(score->ifa->flags & (IFA_F_DEPRECATED|IFA_F_OPTIMISTIC));
1043 break;
1044 #ifdef CONFIG_IPV6_MIP6
1045 case IPV6_SADDR_RULE_HOA:
1047 /* Rule 4: Prefer home address */
1048 int prefhome = !(dst->prefs & IPV6_PREFER_SRC_COA);
1049 ret = !(score->ifa->flags & IFA_F_HOMEADDRESS) ^ prefhome;
1050 break;
1052 #endif
1053 case IPV6_SADDR_RULE_OIF:
1054 /* Rule 5: Prefer outgoing interface */
1055 ret = (!dst->ifindex ||
1056 dst->ifindex == score->ifa->idev->dev->ifindex);
1057 break;
1058 case IPV6_SADDR_RULE_LABEL:
1059 /* Rule 6: Prefer matching label */
1060 ret = ipv6_addr_label(net,
1061 &score->ifa->addr, score->addr_type,
1062 score->ifa->idev->dev->ifindex) == dst->label;
1063 break;
1064 #ifdef CONFIG_IPV6_PRIVACY
1065 case IPV6_SADDR_RULE_PRIVACY:
1067 /* Rule 7: Prefer public address
1068 * Note: prefer temporary address if use_tempaddr >= 2
1070 int preftmp = dst->prefs & (IPV6_PREFER_SRC_PUBLIC|IPV6_PREFER_SRC_TMP) ?
1071 !!(dst->prefs & IPV6_PREFER_SRC_TMP) :
1072 score->ifa->idev->cnf.use_tempaddr >= 2;
1073 ret = (!(score->ifa->flags & IFA_F_TEMPORARY)) ^ preftmp;
1074 break;
1076 #endif
1077 case IPV6_SADDR_RULE_ORCHID:
1078 /* Rule 8-: Prefer ORCHID vs ORCHID or
1079 * non-ORCHID vs non-ORCHID
1081 ret = !(ipv6_addr_orchid(&score->ifa->addr) ^
1082 ipv6_addr_orchid(dst->addr));
1083 break;
1084 case IPV6_SADDR_RULE_PREFIX:
1085 /* Rule 8: Use longest matching prefix */
1086 score->matchlen = ret = ipv6_addr_diff(&score->ifa->addr,
1087 dst->addr);
1088 break;
1089 default:
1090 ret = 0;
1093 if (ret)
1094 __set_bit(i, score->scorebits);
1095 score->rule = i;
1096 out:
1097 return ret;
1100 int ipv6_dev_get_saddr(struct net *net, struct net_device *dst_dev,
1101 const struct in6_addr *daddr, unsigned int prefs,
1102 struct in6_addr *saddr)
1104 struct ipv6_saddr_score scores[2],
1105 *score = &scores[0], *hiscore = &scores[1];
1106 struct ipv6_saddr_dst dst;
1107 struct net_device *dev;
1108 int dst_type;
1110 dst_type = __ipv6_addr_type(daddr);
1111 dst.addr = daddr;
1112 dst.ifindex = dst_dev ? dst_dev->ifindex : 0;
1113 dst.scope = __ipv6_addr_src_scope(dst_type);
1114 dst.label = ipv6_addr_label(net, daddr, dst_type, dst.ifindex);
1115 dst.prefs = prefs;
1117 hiscore->rule = -1;
1118 hiscore->ifa = NULL;
1120 rcu_read_lock();
1122 for_each_netdev_rcu(net, dev) {
1123 struct inet6_dev *idev;
1125 /* Candidate Source Address (section 4)
1126 * - multicast and link-local destination address,
1127 * the set of candidate source address MUST only
1128 * include addresses assigned to interfaces
1129 * belonging to the same link as the outgoing
1130 * interface.
1131 * (- For site-local destination addresses, the
1132 * set of candidate source addresses MUST only
1133 * include addresses assigned to interfaces
1134 * belonging to the same site as the outgoing
1135 * interface.)
1137 if (((dst_type & IPV6_ADDR_MULTICAST) ||
1138 dst.scope <= IPV6_ADDR_SCOPE_LINKLOCAL) &&
1139 dst.ifindex && dev->ifindex != dst.ifindex)
1140 continue;
1142 idev = __in6_dev_get(dev);
1143 if (!idev)
1144 continue;
1146 read_lock_bh(&idev->lock);
1147 list_for_each_entry(score->ifa, &idev->addr_list, if_list) {
1148 int i;
1151 * - Tentative Address (RFC2462 section 5.4)
1152 * - A tentative address is not considered
1153 * "assigned to an interface" in the traditional
1154 * sense, unless it is also flagged as optimistic.
1155 * - Candidate Source Address (section 4)
1156 * - In any case, anycast addresses, multicast
1157 * addresses, and the unspecified address MUST
1158 * NOT be included in a candidate set.
1160 if ((score->ifa->flags & IFA_F_TENTATIVE) &&
1161 (!(score->ifa->flags & IFA_F_OPTIMISTIC)))
1162 continue;
1164 score->addr_type = __ipv6_addr_type(&score->ifa->addr);
1166 if (unlikely(score->addr_type == IPV6_ADDR_ANY ||
1167 score->addr_type & IPV6_ADDR_MULTICAST)) {
1168 LIMIT_NETDEBUG(KERN_DEBUG
1169 "ADDRCONF: unspecified / multicast address "
1170 "assigned as unicast address on %s",
1171 dev->name);
1172 continue;
1175 score->rule = -1;
1176 bitmap_zero(score->scorebits, IPV6_SADDR_RULE_MAX);
1178 for (i = 0; i < IPV6_SADDR_RULE_MAX; i++) {
1179 int minihiscore, miniscore;
1181 minihiscore = ipv6_get_saddr_eval(net, hiscore, &dst, i);
1182 miniscore = ipv6_get_saddr_eval(net, score, &dst, i);
1184 if (minihiscore > miniscore) {
1185 if (i == IPV6_SADDR_RULE_SCOPE &&
1186 score->scopedist > 0) {
1188 * special case:
1189 * each remaining entry
1190 * has too small (not enough)
1191 * scope, because ifa entries
1192 * are sorted by their scope
1193 * values.
1195 goto try_nextdev;
1197 break;
1198 } else if (minihiscore < miniscore) {
1199 if (hiscore->ifa)
1200 in6_ifa_put(hiscore->ifa);
1202 in6_ifa_hold(score->ifa);
1204 swap(hiscore, score);
1206 /* restore our iterator */
1207 score->ifa = hiscore->ifa;
1209 break;
1213 try_nextdev:
1214 read_unlock_bh(&idev->lock);
1216 rcu_read_unlock();
1218 if (!hiscore->ifa)
1219 return -EADDRNOTAVAIL;
1221 *saddr = hiscore->ifa->addr;
1222 in6_ifa_put(hiscore->ifa);
1223 return 0;
1225 EXPORT_SYMBOL(ipv6_dev_get_saddr);
1227 int ipv6_get_lladdr(struct net_device *dev, struct in6_addr *addr,
1228 unsigned char banned_flags)
1230 struct inet6_dev *idev;
1231 int err = -EADDRNOTAVAIL;
1233 rcu_read_lock();
1234 idev = __in6_dev_get(dev);
1235 if (idev) {
1236 struct inet6_ifaddr *ifp;
1238 read_lock_bh(&idev->lock);
1239 list_for_each_entry(ifp, &idev->addr_list, if_list) {
1240 if (ifp->scope == IFA_LINK &&
1241 !(ifp->flags & banned_flags)) {
1242 *addr = ifp->addr;
1243 err = 0;
1244 break;
1247 read_unlock_bh(&idev->lock);
1249 rcu_read_unlock();
1250 return err;
1253 static int ipv6_count_addresses(struct inet6_dev *idev)
1255 int cnt = 0;
1256 struct inet6_ifaddr *ifp;
1258 read_lock_bh(&idev->lock);
1259 list_for_each_entry(ifp, &idev->addr_list, if_list)
1260 cnt++;
1261 read_unlock_bh(&idev->lock);
1262 return cnt;
1265 int ipv6_chk_addr(struct net *net, const struct in6_addr *addr,
1266 struct net_device *dev, int strict)
1268 struct inet6_ifaddr *ifp;
1269 struct hlist_node *node;
1270 unsigned int hash = ipv6_addr_hash(addr);
1272 rcu_read_lock_bh();
1273 hlist_for_each_entry_rcu(ifp, node, &inet6_addr_lst[hash], addr_lst) {
1274 if (!net_eq(dev_net(ifp->idev->dev), net))
1275 continue;
1276 if (ipv6_addr_equal(&ifp->addr, addr) &&
1277 !(ifp->flags&IFA_F_TENTATIVE) &&
1278 (dev == NULL || ifp->idev->dev == dev ||
1279 !(ifp->scope&(IFA_LINK|IFA_HOST) || strict))) {
1280 rcu_read_unlock_bh();
1281 return 1;
1285 rcu_read_unlock_bh();
1286 return 0;
1288 EXPORT_SYMBOL(ipv6_chk_addr);
1290 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
1291 struct net_device *dev)
1293 unsigned int hash = ipv6_addr_hash(addr);
1294 struct inet6_ifaddr *ifp;
1295 struct hlist_node *node;
1297 hlist_for_each_entry(ifp, node, &inet6_addr_lst[hash], addr_lst) {
1298 if (!net_eq(dev_net(ifp->idev->dev), net))
1299 continue;
1300 if (ipv6_addr_equal(&ifp->addr, addr)) {
1301 if (dev == NULL || ifp->idev->dev == dev)
1302 return true;
1305 return false;
1308 int ipv6_chk_prefix(const struct in6_addr *addr, struct net_device *dev)
1310 struct inet6_dev *idev;
1311 struct inet6_ifaddr *ifa;
1312 int onlink;
1314 onlink = 0;
1315 rcu_read_lock();
1316 idev = __in6_dev_get(dev);
1317 if (idev) {
1318 read_lock_bh(&idev->lock);
1319 list_for_each_entry(ifa, &idev->addr_list, if_list) {
1320 onlink = ipv6_prefix_equal(addr, &ifa->addr,
1321 ifa->prefix_len);
1322 if (onlink)
1323 break;
1325 read_unlock_bh(&idev->lock);
1327 rcu_read_unlock();
1328 return onlink;
1331 EXPORT_SYMBOL(ipv6_chk_prefix);
1333 struct inet6_ifaddr *ipv6_get_ifaddr(struct net *net, const struct in6_addr *addr,
1334 struct net_device *dev, int strict)
1336 struct inet6_ifaddr *ifp, *result = NULL;
1337 unsigned int hash = ipv6_addr_hash(addr);
1338 struct hlist_node *node;
1340 rcu_read_lock_bh();
1341 hlist_for_each_entry_rcu_bh(ifp, node, &inet6_addr_lst[hash], addr_lst) {
1342 if (!net_eq(dev_net(ifp->idev->dev), net))
1343 continue;
1344 if (ipv6_addr_equal(&ifp->addr, addr)) {
1345 if (dev == NULL || ifp->idev->dev == dev ||
1346 !(ifp->scope&(IFA_LINK|IFA_HOST) || strict)) {
1347 result = ifp;
1348 in6_ifa_hold(ifp);
1349 break;
1353 rcu_read_unlock_bh();
1355 return result;
1358 /* Gets referenced address, destroys ifaddr */
1360 static void addrconf_dad_stop(struct inet6_ifaddr *ifp, int dad_failed)
1362 if (ifp->flags&IFA_F_PERMANENT) {
1363 spin_lock_bh(&ifp->lock);
1364 addrconf_del_timer(ifp);
1365 ifp->flags |= IFA_F_TENTATIVE;
1366 if (dad_failed)
1367 ifp->flags |= IFA_F_DADFAILED;
1368 spin_unlock_bh(&ifp->lock);
1369 if (dad_failed)
1370 ipv6_ifa_notify(0, ifp);
1371 in6_ifa_put(ifp);
1372 #ifdef CONFIG_IPV6_PRIVACY
1373 } else if (ifp->flags&IFA_F_TEMPORARY) {
1374 struct inet6_ifaddr *ifpub;
1375 spin_lock_bh(&ifp->lock);
1376 ifpub = ifp->ifpub;
1377 if (ifpub) {
1378 in6_ifa_hold(ifpub);
1379 spin_unlock_bh(&ifp->lock);
1380 ipv6_create_tempaddr(ifpub, ifp);
1381 in6_ifa_put(ifpub);
1382 } else {
1383 spin_unlock_bh(&ifp->lock);
1385 ipv6_del_addr(ifp);
1386 #endif
1387 } else
1388 ipv6_del_addr(ifp);
1391 static int addrconf_dad_end(struct inet6_ifaddr *ifp)
1393 int err = -ENOENT;
1395 spin_lock(&ifp->state_lock);
1396 if (ifp->state == INET6_IFADDR_STATE_DAD) {
1397 ifp->state = INET6_IFADDR_STATE_POSTDAD;
1398 err = 0;
1400 spin_unlock(&ifp->state_lock);
1402 return err;
1405 void addrconf_dad_failure(struct inet6_ifaddr *ifp)
1407 struct inet6_dev *idev = ifp->idev;
1409 if (addrconf_dad_end(ifp)) {
1410 in6_ifa_put(ifp);
1411 return;
1414 if (net_ratelimit())
1415 printk(KERN_INFO "%s: IPv6 duplicate address %pI6c detected!\n",
1416 ifp->idev->dev->name, &ifp->addr);
1418 if (idev->cnf.accept_dad > 1 && !idev->cnf.disable_ipv6) {
1419 struct in6_addr addr;
1421 addr.s6_addr32[0] = htonl(0xfe800000);
1422 addr.s6_addr32[1] = 0;
1424 if (!ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) &&
1425 ipv6_addr_equal(&ifp->addr, &addr)) {
1426 /* DAD failed for link-local based on MAC address */
1427 idev->cnf.disable_ipv6 = 1;
1429 printk(KERN_INFO "%s: IPv6 being disabled!\n",
1430 ifp->idev->dev->name);
1434 addrconf_dad_stop(ifp, 1);
1437 /* Join to solicited addr multicast group. */
1439 void addrconf_join_solict(struct net_device *dev, const struct in6_addr *addr)
1441 struct in6_addr maddr;
1443 if (dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1444 return;
1446 addrconf_addr_solict_mult(addr, &maddr);
1447 ipv6_dev_mc_inc(dev, &maddr);
1450 void addrconf_leave_solict(struct inet6_dev *idev, const struct in6_addr *addr)
1452 struct in6_addr maddr;
1454 if (idev->dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1455 return;
1457 addrconf_addr_solict_mult(addr, &maddr);
1458 __ipv6_dev_mc_dec(idev, &maddr);
1461 static void addrconf_join_anycast(struct inet6_ifaddr *ifp)
1463 struct in6_addr addr;
1464 if (ifp->prefix_len == 127) /* RFC 6164 */
1465 return;
1466 ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1467 if (ipv6_addr_any(&addr))
1468 return;
1469 ipv6_dev_ac_inc(ifp->idev->dev, &addr);
1472 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp)
1474 struct in6_addr addr;
1475 if (ifp->prefix_len == 127) /* RFC 6164 */
1476 return;
1477 ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1478 if (ipv6_addr_any(&addr))
1479 return;
1480 __ipv6_dev_ac_dec(ifp->idev, &addr);
1483 static int addrconf_ifid_eui48(u8 *eui, struct net_device *dev)
1485 if (dev->addr_len != ETH_ALEN)
1486 return -1;
1487 memcpy(eui, dev->dev_addr, 3);
1488 memcpy(eui + 5, dev->dev_addr + 3, 3);
1491 * The zSeries OSA network cards can be shared among various
1492 * OS instances, but the OSA cards have only one MAC address.
1493 * This leads to duplicate address conflicts in conjunction
1494 * with IPv6 if more than one instance uses the same card.
1496 * The driver for these cards can deliver a unique 16-bit
1497 * identifier for each instance sharing the same card. It is
1498 * placed instead of 0xFFFE in the interface identifier. The
1499 * "u" bit of the interface identifier is not inverted in this
1500 * case. Hence the resulting interface identifier has local
1501 * scope according to RFC2373.
1503 if (dev->dev_id) {
1504 eui[3] = (dev->dev_id >> 8) & 0xFF;
1505 eui[4] = dev->dev_id & 0xFF;
1506 } else {
1507 eui[3] = 0xFF;
1508 eui[4] = 0xFE;
1509 eui[0] ^= 2;
1511 return 0;
1514 static int addrconf_ifid_arcnet(u8 *eui, struct net_device *dev)
1516 /* XXX: inherit EUI-64 from other interface -- yoshfuji */
1517 if (dev->addr_len != ARCNET_ALEN)
1518 return -1;
1519 memset(eui, 0, 7);
1520 eui[7] = *(u8*)dev->dev_addr;
1521 return 0;
1524 static int addrconf_ifid_infiniband(u8 *eui, struct net_device *dev)
1526 if (dev->addr_len != INFINIBAND_ALEN)
1527 return -1;
1528 memcpy(eui, dev->dev_addr + 12, 8);
1529 eui[0] |= 2;
1530 return 0;
1533 static int __ipv6_isatap_ifid(u8 *eui, __be32 addr)
1535 if (addr == 0)
1536 return -1;
1537 eui[0] = (ipv4_is_zeronet(addr) || ipv4_is_private_10(addr) ||
1538 ipv4_is_loopback(addr) || ipv4_is_linklocal_169(addr) ||
1539 ipv4_is_private_172(addr) || ipv4_is_test_192(addr) ||
1540 ipv4_is_anycast_6to4(addr) || ipv4_is_private_192(addr) ||
1541 ipv4_is_test_198(addr) || ipv4_is_multicast(addr) ||
1542 ipv4_is_lbcast(addr)) ? 0x00 : 0x02;
1543 eui[1] = 0;
1544 eui[2] = 0x5E;
1545 eui[3] = 0xFE;
1546 memcpy(eui + 4, &addr, 4);
1547 return 0;
1550 static int addrconf_ifid_sit(u8 *eui, struct net_device *dev)
1552 if (dev->priv_flags & IFF_ISATAP)
1553 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
1554 return -1;
1557 static int addrconf_ifid_gre(u8 *eui, struct net_device *dev)
1559 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
1562 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev)
1564 switch (dev->type) {
1565 case ARPHRD_ETHER:
1566 case ARPHRD_FDDI:
1567 case ARPHRD_IEEE802_TR:
1568 return addrconf_ifid_eui48(eui, dev);
1569 case ARPHRD_ARCNET:
1570 return addrconf_ifid_arcnet(eui, dev);
1571 case ARPHRD_INFINIBAND:
1572 return addrconf_ifid_infiniband(eui, dev);
1573 case ARPHRD_SIT:
1574 return addrconf_ifid_sit(eui, dev);
1575 case ARPHRD_IPGRE:
1576 return addrconf_ifid_gre(eui, dev);
1578 return -1;
1581 static int ipv6_inherit_eui64(u8 *eui, struct inet6_dev *idev)
1583 int err = -1;
1584 struct inet6_ifaddr *ifp;
1586 read_lock_bh(&idev->lock);
1587 list_for_each_entry(ifp, &idev->addr_list, if_list) {
1588 if (ifp->scope == IFA_LINK && !(ifp->flags&IFA_F_TENTATIVE)) {
1589 memcpy(eui, ifp->addr.s6_addr+8, 8);
1590 err = 0;
1591 break;
1594 read_unlock_bh(&idev->lock);
1595 return err;
1598 #ifdef CONFIG_IPV6_PRIVACY
1599 /* (re)generation of randomized interface identifier (RFC 3041 3.2, 3.5) */
1600 static int __ipv6_regen_rndid(struct inet6_dev *idev)
1602 regen:
1603 get_random_bytes(idev->rndid, sizeof(idev->rndid));
1604 idev->rndid[0] &= ~0x02;
1607 * <draft-ietf-ipngwg-temp-addresses-v2-00.txt>:
1608 * check if generated address is not inappropriate
1610 * - Reserved subnet anycast (RFC 2526)
1611 * 11111101 11....11 1xxxxxxx
1612 * - ISATAP (RFC4214) 6.1
1613 * 00-00-5E-FE-xx-xx-xx-xx
1614 * - value 0
1615 * - XXX: already assigned to an address on the device
1617 if (idev->rndid[0] == 0xfd &&
1618 (idev->rndid[1]&idev->rndid[2]&idev->rndid[3]&idev->rndid[4]&idev->rndid[5]&idev->rndid[6]) == 0xff &&
1619 (idev->rndid[7]&0x80))
1620 goto regen;
1621 if ((idev->rndid[0]|idev->rndid[1]) == 0) {
1622 if (idev->rndid[2] == 0x5e && idev->rndid[3] == 0xfe)
1623 goto regen;
1624 if ((idev->rndid[2]|idev->rndid[3]|idev->rndid[4]|idev->rndid[5]|idev->rndid[6]|idev->rndid[7]) == 0x00)
1625 goto regen;
1628 return 0;
1631 static void ipv6_regen_rndid(unsigned long data)
1633 struct inet6_dev *idev = (struct inet6_dev *) data;
1634 unsigned long expires;
1636 rcu_read_lock_bh();
1637 write_lock_bh(&idev->lock);
1639 if (idev->dead)
1640 goto out;
1642 if (__ipv6_regen_rndid(idev) < 0)
1643 goto out;
1645 expires = jiffies +
1646 idev->cnf.temp_prefered_lft * HZ -
1647 idev->cnf.regen_max_retry * idev->cnf.dad_transmits * idev->nd_parms->retrans_time -
1648 idev->cnf.max_desync_factor * HZ;
1649 if (time_before(expires, jiffies)) {
1650 printk(KERN_WARNING
1651 "ipv6_regen_rndid(): too short regeneration interval; timer disabled for %s.\n",
1652 idev->dev->name);
1653 goto out;
1656 if (!mod_timer(&idev->regen_timer, expires))
1657 in6_dev_hold(idev);
1659 out:
1660 write_unlock_bh(&idev->lock);
1661 rcu_read_unlock_bh();
1662 in6_dev_put(idev);
1665 static int __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr) {
1666 int ret = 0;
1668 if (tmpaddr && memcmp(idev->rndid, &tmpaddr->s6_addr[8], 8) == 0)
1669 ret = __ipv6_regen_rndid(idev);
1670 return ret;
1672 #endif
1675 * Add prefix route.
1678 static void
1679 addrconf_prefix_route(struct in6_addr *pfx, int plen, struct net_device *dev,
1680 unsigned long expires, u32 flags)
1682 struct fib6_config cfg = {
1683 .fc_table = RT6_TABLE_PREFIX,
1684 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1685 .fc_ifindex = dev->ifindex,
1686 .fc_expires = expires,
1687 .fc_dst_len = plen,
1688 .fc_flags = RTF_UP | flags,
1689 .fc_nlinfo.nl_net = dev_net(dev),
1690 .fc_protocol = RTPROT_KERNEL,
1693 cfg.fc_dst = *pfx;
1695 /* Prevent useless cloning on PtP SIT.
1696 This thing is done here expecting that the whole
1697 class of non-broadcast devices need not cloning.
1699 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
1700 if (dev->type == ARPHRD_SIT && (dev->flags & IFF_POINTOPOINT))
1701 cfg.fc_flags |= RTF_NONEXTHOP;
1702 #endif
1704 ip6_route_add(&cfg);
1708 static struct rt6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
1709 int plen,
1710 const struct net_device *dev,
1711 u32 flags, u32 noflags)
1713 struct fib6_node *fn;
1714 struct rt6_info *rt = NULL;
1715 struct fib6_table *table;
1717 table = fib6_get_table(dev_net(dev), RT6_TABLE_PREFIX);
1718 if (table == NULL)
1719 return NULL;
1721 write_lock_bh(&table->tb6_lock);
1722 fn = fib6_locate(&table->tb6_root, pfx, plen, NULL, 0);
1723 if (!fn)
1724 goto out;
1725 for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
1726 if (rt->dst.dev->ifindex != dev->ifindex)
1727 continue;
1728 if ((rt->rt6i_flags & flags) != flags)
1729 continue;
1730 if ((noflags != 0) && ((rt->rt6i_flags & flags) != 0))
1731 continue;
1732 dst_hold(&rt->dst);
1733 break;
1735 out:
1736 write_unlock_bh(&table->tb6_lock);
1737 return rt;
1741 /* Create "default" multicast route to the interface */
1743 static void addrconf_add_mroute(struct net_device *dev)
1745 struct fib6_config cfg = {
1746 .fc_table = RT6_TABLE_LOCAL,
1747 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1748 .fc_ifindex = dev->ifindex,
1749 .fc_dst_len = 8,
1750 .fc_flags = RTF_UP,
1751 .fc_nlinfo.nl_net = dev_net(dev),
1754 ipv6_addr_set(&cfg.fc_dst, htonl(0xFF000000), 0, 0, 0);
1756 ip6_route_add(&cfg);
1759 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
1760 static void sit_route_add(struct net_device *dev)
1762 struct fib6_config cfg = {
1763 .fc_table = RT6_TABLE_MAIN,
1764 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1765 .fc_ifindex = dev->ifindex,
1766 .fc_dst_len = 96,
1767 .fc_flags = RTF_UP | RTF_NONEXTHOP,
1768 .fc_nlinfo.nl_net = dev_net(dev),
1771 /* prefix length - 96 bits "::d.d.d.d" */
1772 ip6_route_add(&cfg);
1774 #endif
1776 static void addrconf_add_lroute(struct net_device *dev)
1778 struct in6_addr addr;
1780 ipv6_addr_set(&addr, htonl(0xFE800000), 0, 0, 0);
1781 addrconf_prefix_route(&addr, 64, dev, 0, 0);
1784 static struct inet6_dev *addrconf_add_dev(struct net_device *dev)
1786 struct inet6_dev *idev;
1788 ASSERT_RTNL();
1790 idev = ipv6_find_idev(dev);
1791 if (!idev)
1792 return ERR_PTR(-ENOBUFS);
1794 if (idev->cnf.disable_ipv6)
1795 return ERR_PTR(-EACCES);
1797 /* Add default multicast route */
1798 if (!(dev->flags & IFF_LOOPBACK))
1799 addrconf_add_mroute(dev);
1801 /* Add link local route */
1802 addrconf_add_lroute(dev);
1803 return idev;
1806 void addrconf_prefix_rcv(struct net_device *dev, u8 *opt, int len, bool sllao)
1808 struct prefix_info *pinfo;
1809 __u32 valid_lft;
1810 __u32 prefered_lft;
1811 int addr_type;
1812 struct inet6_dev *in6_dev;
1813 struct net *net = dev_net(dev);
1815 pinfo = (struct prefix_info *) opt;
1817 if (len < sizeof(struct prefix_info)) {
1818 ADBG(("addrconf: prefix option too short\n"));
1819 return;
1823 * Validation checks ([ADDRCONF], page 19)
1826 addr_type = ipv6_addr_type(&pinfo->prefix);
1828 if (addr_type & (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL))
1829 return;
1831 valid_lft = ntohl(pinfo->valid);
1832 prefered_lft = ntohl(pinfo->prefered);
1834 if (prefered_lft > valid_lft) {
1835 if (net_ratelimit())
1836 printk(KERN_WARNING "addrconf: prefix option has invalid lifetime\n");
1837 return;
1840 in6_dev = in6_dev_get(dev);
1842 if (in6_dev == NULL) {
1843 if (net_ratelimit())
1844 printk(KERN_DEBUG "addrconf: device %s not configured\n", dev->name);
1845 return;
1849 * Two things going on here:
1850 * 1) Add routes for on-link prefixes
1851 * 2) Configure prefixes with the auto flag set
1854 if (pinfo->onlink) {
1855 struct rt6_info *rt;
1856 unsigned long rt_expires;
1858 /* Avoid arithmetic overflow. Really, we could
1859 * save rt_expires in seconds, likely valid_lft,
1860 * but it would require division in fib gc, that it
1861 * not good.
1863 if (HZ > USER_HZ)
1864 rt_expires = addrconf_timeout_fixup(valid_lft, HZ);
1865 else
1866 rt_expires = addrconf_timeout_fixup(valid_lft, USER_HZ);
1868 if (addrconf_finite_timeout(rt_expires))
1869 rt_expires *= HZ;
1871 rt = addrconf_get_prefix_route(&pinfo->prefix,
1872 pinfo->prefix_len,
1873 dev,
1874 RTF_ADDRCONF | RTF_PREFIX_RT,
1875 RTF_GATEWAY | RTF_DEFAULT);
1877 if (rt) {
1878 /* Autoconf prefix route */
1879 if (valid_lft == 0) {
1880 ip6_del_rt(rt);
1881 rt = NULL;
1882 } else if (addrconf_finite_timeout(rt_expires)) {
1883 /* not infinity */
1884 rt->dst.expires = jiffies + rt_expires;
1885 rt->rt6i_flags |= RTF_EXPIRES;
1886 } else {
1887 rt->rt6i_flags &= ~RTF_EXPIRES;
1888 rt->dst.expires = 0;
1890 } else if (valid_lft) {
1891 clock_t expires = 0;
1892 int flags = RTF_ADDRCONF | RTF_PREFIX_RT;
1893 if (addrconf_finite_timeout(rt_expires)) {
1894 /* not infinity */
1895 flags |= RTF_EXPIRES;
1896 expires = jiffies_to_clock_t(rt_expires);
1898 addrconf_prefix_route(&pinfo->prefix, pinfo->prefix_len,
1899 dev, expires, flags);
1901 if (rt)
1902 dst_release(&rt->dst);
1905 /* Try to figure out our local address for this prefix */
1907 if (pinfo->autoconf && in6_dev->cnf.autoconf) {
1908 struct inet6_ifaddr * ifp;
1909 struct in6_addr addr;
1910 int create = 0, update_lft = 0;
1912 if (pinfo->prefix_len == 64) {
1913 memcpy(&addr, &pinfo->prefix, 8);
1914 if (ipv6_generate_eui64(addr.s6_addr + 8, dev) &&
1915 ipv6_inherit_eui64(addr.s6_addr + 8, in6_dev)) {
1916 in6_dev_put(in6_dev);
1917 return;
1919 goto ok;
1921 if (net_ratelimit())
1922 printk(KERN_DEBUG "IPv6 addrconf: prefix with wrong length %d\n",
1923 pinfo->prefix_len);
1924 in6_dev_put(in6_dev);
1925 return;
1929 ifp = ipv6_get_ifaddr(net, &addr, dev, 1);
1931 if (ifp == NULL && valid_lft) {
1932 int max_addresses = in6_dev->cnf.max_addresses;
1933 u32 addr_flags = 0;
1935 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1936 if (in6_dev->cnf.optimistic_dad &&
1937 !net->ipv6.devconf_all->forwarding && sllao)
1938 addr_flags = IFA_F_OPTIMISTIC;
1939 #endif
1941 /* Do not allow to create too much of autoconfigured
1942 * addresses; this would be too easy way to crash kernel.
1944 if (!max_addresses ||
1945 ipv6_count_addresses(in6_dev) < max_addresses)
1946 ifp = ipv6_add_addr(in6_dev, &addr, pinfo->prefix_len,
1947 addr_type&IPV6_ADDR_SCOPE_MASK,
1948 addr_flags);
1950 if (!ifp || IS_ERR(ifp)) {
1951 in6_dev_put(in6_dev);
1952 return;
1955 update_lft = create = 1;
1956 ifp->cstamp = jiffies;
1957 addrconf_dad_start(ifp, RTF_ADDRCONF|RTF_PREFIX_RT);
1960 if (ifp) {
1961 int flags;
1962 unsigned long now;
1963 #ifdef CONFIG_IPV6_PRIVACY
1964 struct inet6_ifaddr *ift;
1965 #endif
1966 u32 stored_lft;
1968 /* update lifetime (RFC2462 5.5.3 e) */
1969 spin_lock(&ifp->lock);
1970 now = jiffies;
1971 if (ifp->valid_lft > (now - ifp->tstamp) / HZ)
1972 stored_lft = ifp->valid_lft - (now - ifp->tstamp) / HZ;
1973 else
1974 stored_lft = 0;
1975 if (!update_lft && stored_lft) {
1976 if (valid_lft > MIN_VALID_LIFETIME ||
1977 valid_lft > stored_lft)
1978 update_lft = 1;
1979 else if (stored_lft <= MIN_VALID_LIFETIME) {
1980 /* valid_lft <= stored_lft is always true */
1982 * RFC 4862 Section 5.5.3e:
1983 * "Note that the preferred lifetime of
1984 * the corresponding address is always
1985 * reset to the Preferred Lifetime in
1986 * the received Prefix Information
1987 * option, regardless of whether the
1988 * valid lifetime is also reset or
1989 * ignored."
1991 * So if the preferred lifetime in
1992 * this advertisement is different
1993 * than what we have stored, but the
1994 * valid lifetime is invalid, just
1995 * reset prefered_lft.
1997 * We must set the valid lifetime
1998 * to the stored lifetime since we'll
1999 * be updating the timestamp below,
2000 * else we'll set it back to the
2001 * minimum.
2003 if (prefered_lft != ifp->prefered_lft) {
2004 valid_lft = stored_lft;
2005 update_lft = 1;
2007 } else {
2008 valid_lft = MIN_VALID_LIFETIME;
2009 if (valid_lft < prefered_lft)
2010 prefered_lft = valid_lft;
2011 update_lft = 1;
2015 if (update_lft) {
2016 ifp->valid_lft = valid_lft;
2017 ifp->prefered_lft = prefered_lft;
2018 ifp->tstamp = now;
2019 flags = ifp->flags;
2020 ifp->flags &= ~IFA_F_DEPRECATED;
2021 spin_unlock(&ifp->lock);
2023 if (!(flags&IFA_F_TENTATIVE))
2024 ipv6_ifa_notify(0, ifp);
2025 } else
2026 spin_unlock(&ifp->lock);
2028 #ifdef CONFIG_IPV6_PRIVACY
2029 read_lock_bh(&in6_dev->lock);
2030 /* update all temporary addresses in the list */
2031 list_for_each_entry(ift, &in6_dev->tempaddr_list,
2032 tmp_list) {
2033 int age, max_valid, max_prefered;
2035 if (ifp != ift->ifpub)
2036 continue;
2039 * RFC 4941 section 3.3:
2040 * If a received option will extend the lifetime
2041 * of a public address, the lifetimes of
2042 * temporary addresses should be extended,
2043 * subject to the overall constraint that no
2044 * temporary addresses should ever remain
2045 * "valid" or "preferred" for a time longer than
2046 * (TEMP_VALID_LIFETIME) or
2047 * (TEMP_PREFERRED_LIFETIME - DESYNC_FACTOR),
2048 * respectively.
2050 age = (now - ift->cstamp) / HZ;
2051 max_valid = in6_dev->cnf.temp_valid_lft - age;
2052 if (max_valid < 0)
2053 max_valid = 0;
2055 max_prefered = in6_dev->cnf.temp_prefered_lft -
2056 in6_dev->cnf.max_desync_factor -
2057 age;
2058 if (max_prefered < 0)
2059 max_prefered = 0;
2061 if (valid_lft > max_valid)
2062 valid_lft = max_valid;
2064 if (prefered_lft > max_prefered)
2065 prefered_lft = max_prefered;
2067 spin_lock(&ift->lock);
2068 flags = ift->flags;
2069 ift->valid_lft = valid_lft;
2070 ift->prefered_lft = prefered_lft;
2071 ift->tstamp = now;
2072 if (prefered_lft > 0)
2073 ift->flags &= ~IFA_F_DEPRECATED;
2075 spin_unlock(&ift->lock);
2076 if (!(flags&IFA_F_TENTATIVE))
2077 ipv6_ifa_notify(0, ift);
2080 if ((create || list_empty(&in6_dev->tempaddr_list)) && in6_dev->cnf.use_tempaddr > 0) {
2082 * When a new public address is created as
2083 * described in [ADDRCONF], also create a new
2084 * temporary address. Also create a temporary
2085 * address if it's enabled but no temporary
2086 * address currently exists.
2088 read_unlock_bh(&in6_dev->lock);
2089 ipv6_create_tempaddr(ifp, NULL);
2090 } else {
2091 read_unlock_bh(&in6_dev->lock);
2093 #endif
2094 in6_ifa_put(ifp);
2095 addrconf_verify(0);
2098 inet6_prefix_notify(RTM_NEWPREFIX, in6_dev, pinfo);
2099 in6_dev_put(in6_dev);
2103 * Set destination address.
2104 * Special case for SIT interfaces where we create a new "virtual"
2105 * device.
2107 int addrconf_set_dstaddr(struct net *net, void __user *arg)
2109 struct in6_ifreq ireq;
2110 struct net_device *dev;
2111 int err = -EINVAL;
2113 rtnl_lock();
2115 err = -EFAULT;
2116 if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2117 goto err_exit;
2119 dev = __dev_get_by_index(net, ireq.ifr6_ifindex);
2121 err = -ENODEV;
2122 if (dev == NULL)
2123 goto err_exit;
2125 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2126 if (dev->type == ARPHRD_SIT) {
2127 const struct net_device_ops *ops = dev->netdev_ops;
2128 struct ifreq ifr;
2129 struct ip_tunnel_parm p;
2131 err = -EADDRNOTAVAIL;
2132 if (!(ipv6_addr_type(&ireq.ifr6_addr) & IPV6_ADDR_COMPATv4))
2133 goto err_exit;
2135 memset(&p, 0, sizeof(p));
2136 p.iph.daddr = ireq.ifr6_addr.s6_addr32[3];
2137 p.iph.saddr = 0;
2138 p.iph.version = 4;
2139 p.iph.ihl = 5;
2140 p.iph.protocol = IPPROTO_IPV6;
2141 p.iph.ttl = 64;
2142 ifr.ifr_ifru.ifru_data = (__force void __user *)&p;
2144 if (ops->ndo_do_ioctl) {
2145 mm_segment_t oldfs = get_fs();
2147 set_fs(KERNEL_DS);
2148 err = ops->ndo_do_ioctl(dev, &ifr, SIOCADDTUNNEL);
2149 set_fs(oldfs);
2150 } else
2151 err = -EOPNOTSUPP;
2153 if (err == 0) {
2154 err = -ENOBUFS;
2155 dev = __dev_get_by_name(net, p.name);
2156 if (!dev)
2157 goto err_exit;
2158 err = dev_open(dev);
2161 #endif
2163 err_exit:
2164 rtnl_unlock();
2165 return err;
2169 * Manual configuration of address on an interface
2171 static int inet6_addr_add(struct net *net, int ifindex, const struct in6_addr *pfx,
2172 unsigned int plen, __u8 ifa_flags, __u32 prefered_lft,
2173 __u32 valid_lft)
2175 struct inet6_ifaddr *ifp;
2176 struct inet6_dev *idev;
2177 struct net_device *dev;
2178 int scope;
2179 u32 flags;
2180 clock_t expires;
2181 unsigned long timeout;
2183 ASSERT_RTNL();
2185 if (plen > 128)
2186 return -EINVAL;
2188 /* check the lifetime */
2189 if (!valid_lft || prefered_lft > valid_lft)
2190 return -EINVAL;
2192 dev = __dev_get_by_index(net, ifindex);
2193 if (!dev)
2194 return -ENODEV;
2196 idev = addrconf_add_dev(dev);
2197 if (IS_ERR(idev))
2198 return PTR_ERR(idev);
2200 scope = ipv6_addr_scope(pfx);
2202 timeout = addrconf_timeout_fixup(valid_lft, HZ);
2203 if (addrconf_finite_timeout(timeout)) {
2204 expires = jiffies_to_clock_t(timeout * HZ);
2205 valid_lft = timeout;
2206 flags = RTF_EXPIRES;
2207 } else {
2208 expires = 0;
2209 flags = 0;
2210 ifa_flags |= IFA_F_PERMANENT;
2213 timeout = addrconf_timeout_fixup(prefered_lft, HZ);
2214 if (addrconf_finite_timeout(timeout)) {
2215 if (timeout == 0)
2216 ifa_flags |= IFA_F_DEPRECATED;
2217 prefered_lft = timeout;
2220 ifp = ipv6_add_addr(idev, pfx, plen, scope, ifa_flags);
2222 if (!IS_ERR(ifp)) {
2223 spin_lock_bh(&ifp->lock);
2224 ifp->valid_lft = valid_lft;
2225 ifp->prefered_lft = prefered_lft;
2226 ifp->tstamp = jiffies;
2227 spin_unlock_bh(&ifp->lock);
2229 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, dev,
2230 expires, flags);
2232 * Note that section 3.1 of RFC 4429 indicates
2233 * that the Optimistic flag should not be set for
2234 * manually configured addresses
2236 addrconf_dad_start(ifp, 0);
2237 in6_ifa_put(ifp);
2238 addrconf_verify(0);
2239 return 0;
2242 return PTR_ERR(ifp);
2245 static int inet6_addr_del(struct net *net, int ifindex, const struct in6_addr *pfx,
2246 unsigned int plen)
2248 struct inet6_ifaddr *ifp;
2249 struct inet6_dev *idev;
2250 struct net_device *dev;
2252 if (plen > 128)
2253 return -EINVAL;
2255 dev = __dev_get_by_index(net, ifindex);
2256 if (!dev)
2257 return -ENODEV;
2259 if ((idev = __in6_dev_get(dev)) == NULL)
2260 return -ENXIO;
2262 read_lock_bh(&idev->lock);
2263 list_for_each_entry(ifp, &idev->addr_list, if_list) {
2264 if (ifp->prefix_len == plen &&
2265 ipv6_addr_equal(pfx, &ifp->addr)) {
2266 in6_ifa_hold(ifp);
2267 read_unlock_bh(&idev->lock);
2269 ipv6_del_addr(ifp);
2271 /* If the last address is deleted administratively,
2272 disable IPv6 on this interface.
2274 if (list_empty(&idev->addr_list))
2275 addrconf_ifdown(idev->dev, 1);
2276 return 0;
2279 read_unlock_bh(&idev->lock);
2280 return -EADDRNOTAVAIL;
2284 int addrconf_add_ifaddr(struct net *net, void __user *arg)
2286 struct in6_ifreq ireq;
2287 int err;
2289 if (!capable(CAP_NET_ADMIN))
2290 return -EPERM;
2292 if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2293 return -EFAULT;
2295 rtnl_lock();
2296 err = inet6_addr_add(net, ireq.ifr6_ifindex, &ireq.ifr6_addr,
2297 ireq.ifr6_prefixlen, IFA_F_PERMANENT,
2298 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
2299 rtnl_unlock();
2300 return err;
2303 int addrconf_del_ifaddr(struct net *net, void __user *arg)
2305 struct in6_ifreq ireq;
2306 int err;
2308 if (!capable(CAP_NET_ADMIN))
2309 return -EPERM;
2311 if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2312 return -EFAULT;
2314 rtnl_lock();
2315 err = inet6_addr_del(net, ireq.ifr6_ifindex, &ireq.ifr6_addr,
2316 ireq.ifr6_prefixlen);
2317 rtnl_unlock();
2318 return err;
2321 static void add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
2322 int plen, int scope)
2324 struct inet6_ifaddr *ifp;
2326 ifp = ipv6_add_addr(idev, addr, plen, scope, IFA_F_PERMANENT);
2327 if (!IS_ERR(ifp)) {
2328 spin_lock_bh(&ifp->lock);
2329 ifp->flags &= ~IFA_F_TENTATIVE;
2330 spin_unlock_bh(&ifp->lock);
2331 ipv6_ifa_notify(RTM_NEWADDR, ifp);
2332 in6_ifa_put(ifp);
2336 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2337 static void sit_add_v4_addrs(struct inet6_dev *idev)
2339 struct in6_addr addr;
2340 struct net_device *dev;
2341 struct net *net = dev_net(idev->dev);
2342 int scope;
2344 ASSERT_RTNL();
2346 memset(&addr, 0, sizeof(struct in6_addr));
2347 memcpy(&addr.s6_addr32[3], idev->dev->dev_addr, 4);
2349 if (idev->dev->flags&IFF_POINTOPOINT) {
2350 addr.s6_addr32[0] = htonl(0xfe800000);
2351 scope = IFA_LINK;
2352 } else {
2353 scope = IPV6_ADDR_COMPATv4;
2356 if (addr.s6_addr32[3]) {
2357 add_addr(idev, &addr, 128, scope);
2358 return;
2361 for_each_netdev(net, dev) {
2362 struct in_device * in_dev = __in_dev_get_rtnl(dev);
2363 if (in_dev && (dev->flags & IFF_UP)) {
2364 struct in_ifaddr * ifa;
2366 int flag = scope;
2368 for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) {
2369 int plen;
2371 addr.s6_addr32[3] = ifa->ifa_local;
2373 if (ifa->ifa_scope == RT_SCOPE_LINK)
2374 continue;
2375 if (ifa->ifa_scope >= RT_SCOPE_HOST) {
2376 if (idev->dev->flags&IFF_POINTOPOINT)
2377 continue;
2378 flag |= IFA_HOST;
2380 if (idev->dev->flags&IFF_POINTOPOINT)
2381 plen = 64;
2382 else
2383 plen = 96;
2385 add_addr(idev, &addr, plen, flag);
2390 #endif
2392 static void init_loopback(struct net_device *dev)
2394 struct inet6_dev *idev;
2396 /* ::1 */
2398 ASSERT_RTNL();
2400 if ((idev = ipv6_find_idev(dev)) == NULL) {
2401 printk(KERN_DEBUG "init loopback: add_dev failed\n");
2402 return;
2405 add_addr(idev, &in6addr_loopback, 128, IFA_HOST);
2408 static void addrconf_add_linklocal(struct inet6_dev *idev, const struct in6_addr *addr)
2410 struct inet6_ifaddr * ifp;
2411 u32 addr_flags = IFA_F_PERMANENT;
2413 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2414 if (idev->cnf.optimistic_dad &&
2415 !dev_net(idev->dev)->ipv6.devconf_all->forwarding)
2416 addr_flags |= IFA_F_OPTIMISTIC;
2417 #endif
2420 ifp = ipv6_add_addr(idev, addr, 64, IFA_LINK, addr_flags);
2421 if (!IS_ERR(ifp)) {
2422 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, idev->dev, 0, 0);
2423 addrconf_dad_start(ifp, 0);
2424 in6_ifa_put(ifp);
2428 static void addrconf_dev_config(struct net_device *dev)
2430 struct in6_addr addr;
2431 struct inet6_dev * idev;
2433 ASSERT_RTNL();
2435 if ((dev->type != ARPHRD_ETHER) &&
2436 (dev->type != ARPHRD_FDDI) &&
2437 (dev->type != ARPHRD_IEEE802_TR) &&
2438 (dev->type != ARPHRD_ARCNET) &&
2439 (dev->type != ARPHRD_INFINIBAND)) {
2440 /* Alas, we support only Ethernet autoconfiguration. */
2441 return;
2444 idev = addrconf_add_dev(dev);
2445 if (IS_ERR(idev))
2446 return;
2448 memset(&addr, 0, sizeof(struct in6_addr));
2449 addr.s6_addr32[0] = htonl(0xFE800000);
2451 if (ipv6_generate_eui64(addr.s6_addr + 8, dev) == 0)
2452 addrconf_add_linklocal(idev, &addr);
2455 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2456 static void addrconf_sit_config(struct net_device *dev)
2458 struct inet6_dev *idev;
2460 ASSERT_RTNL();
2463 * Configure the tunnel with one of our IPv4
2464 * addresses... we should configure all of
2465 * our v4 addrs in the tunnel
2468 if ((idev = ipv6_find_idev(dev)) == NULL) {
2469 printk(KERN_DEBUG "init sit: add_dev failed\n");
2470 return;
2473 if (dev->priv_flags & IFF_ISATAP) {
2474 struct in6_addr addr;
2476 ipv6_addr_set(&addr, htonl(0xFE800000), 0, 0, 0);
2477 addrconf_prefix_route(&addr, 64, dev, 0, 0);
2478 if (!ipv6_generate_eui64(addr.s6_addr + 8, dev))
2479 addrconf_add_linklocal(idev, &addr);
2480 return;
2483 sit_add_v4_addrs(idev);
2485 if (dev->flags&IFF_POINTOPOINT) {
2486 addrconf_add_mroute(dev);
2487 addrconf_add_lroute(dev);
2488 } else
2489 sit_route_add(dev);
2491 #endif
2493 #if defined(CONFIG_NET_IPGRE) || defined(CONFIG_NET_IPGRE_MODULE)
2494 static void addrconf_gre_config(struct net_device *dev)
2496 struct inet6_dev *idev;
2497 struct in6_addr addr;
2499 pr_info("ipv6: addrconf_gre_config(%s)\n", dev->name);
2501 ASSERT_RTNL();
2503 if ((idev = ipv6_find_idev(dev)) == NULL) {
2504 printk(KERN_DEBUG "init gre: add_dev failed\n");
2505 return;
2508 ipv6_addr_set(&addr, htonl(0xFE800000), 0, 0, 0);
2509 addrconf_prefix_route(&addr, 64, dev, 0, 0);
2511 if (!ipv6_generate_eui64(addr.s6_addr + 8, dev))
2512 addrconf_add_linklocal(idev, &addr);
2514 #endif
2516 static inline int
2517 ipv6_inherit_linklocal(struct inet6_dev *idev, struct net_device *link_dev)
2519 struct in6_addr lladdr;
2521 if (!ipv6_get_lladdr(link_dev, &lladdr, IFA_F_TENTATIVE)) {
2522 addrconf_add_linklocal(idev, &lladdr);
2523 return 0;
2525 return -1;
2528 static void ip6_tnl_add_linklocal(struct inet6_dev *idev)
2530 struct net_device *link_dev;
2531 struct net *net = dev_net(idev->dev);
2533 /* first try to inherit the link-local address from the link device */
2534 if (idev->dev->iflink &&
2535 (link_dev = __dev_get_by_index(net, idev->dev->iflink))) {
2536 if (!ipv6_inherit_linklocal(idev, link_dev))
2537 return;
2539 /* then try to inherit it from any device */
2540 for_each_netdev(net, link_dev) {
2541 if (!ipv6_inherit_linklocal(idev, link_dev))
2542 return;
2544 printk(KERN_DEBUG "init ip6-ip6: add_linklocal failed\n");
2548 * Autoconfigure tunnel with a link-local address so routing protocols,
2549 * DHCPv6, MLD etc. can be run over the virtual link
2552 static void addrconf_ip6_tnl_config(struct net_device *dev)
2554 struct inet6_dev *idev;
2556 ASSERT_RTNL();
2558 idev = addrconf_add_dev(dev);
2559 if (IS_ERR(idev)) {
2560 printk(KERN_DEBUG "init ip6-ip6: add_dev failed\n");
2561 return;
2563 ip6_tnl_add_linklocal(idev);
2566 static int addrconf_notify(struct notifier_block *this, unsigned long event,
2567 void * data)
2569 struct net_device *dev = (struct net_device *) data;
2570 struct inet6_dev *idev = __in6_dev_get(dev);
2571 int run_pending = 0;
2572 int err;
2574 switch (event) {
2575 case NETDEV_REGISTER:
2576 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2577 idev = ipv6_add_dev(dev);
2578 if (!idev)
2579 return notifier_from_errno(-ENOMEM);
2581 break;
2583 case NETDEV_UP:
2584 case NETDEV_CHANGE:
2585 if (dev->flags & IFF_SLAVE)
2586 break;
2588 if (event == NETDEV_UP) {
2589 if (!addrconf_qdisc_ok(dev)) {
2590 /* device is not ready yet. */
2591 printk(KERN_INFO
2592 "ADDRCONF(NETDEV_UP): %s: "
2593 "link is not ready\n",
2594 dev->name);
2595 break;
2598 if (!idev && dev->mtu >= IPV6_MIN_MTU)
2599 idev = ipv6_add_dev(dev);
2601 if (idev) {
2602 idev->if_flags |= IF_READY;
2603 run_pending = 1;
2605 } else {
2606 if (!addrconf_qdisc_ok(dev)) {
2607 /* device is still not ready. */
2608 break;
2611 if (idev) {
2612 if (idev->if_flags & IF_READY)
2613 /* device is already configured. */
2614 break;
2615 idev->if_flags |= IF_READY;
2618 printk(KERN_INFO
2619 "ADDRCONF(NETDEV_CHANGE): %s: "
2620 "link becomes ready\n",
2621 dev->name);
2623 run_pending = 1;
2626 switch (dev->type) {
2627 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2628 case ARPHRD_SIT:
2629 addrconf_sit_config(dev);
2630 break;
2631 #endif
2632 #if defined(CONFIG_NET_IPGRE) || defined(CONFIG_NET_IPGRE_MODULE)
2633 case ARPHRD_IPGRE:
2634 addrconf_gre_config(dev);
2635 break;
2636 #endif
2637 case ARPHRD_TUNNEL6:
2638 addrconf_ip6_tnl_config(dev);
2639 break;
2640 case ARPHRD_LOOPBACK:
2641 init_loopback(dev);
2642 break;
2644 default:
2645 addrconf_dev_config(dev);
2646 break;
2649 if (idev) {
2650 if (run_pending)
2651 addrconf_dad_run(idev);
2654 * If the MTU changed during the interface down,
2655 * when the interface up, the changed MTU must be
2656 * reflected in the idev as well as routers.
2658 if (idev->cnf.mtu6 != dev->mtu &&
2659 dev->mtu >= IPV6_MIN_MTU) {
2660 rt6_mtu_change(dev, dev->mtu);
2661 idev->cnf.mtu6 = dev->mtu;
2663 idev->tstamp = jiffies;
2664 inet6_ifinfo_notify(RTM_NEWLINK, idev);
2667 * If the changed mtu during down is lower than
2668 * IPV6_MIN_MTU stop IPv6 on this interface.
2670 if (dev->mtu < IPV6_MIN_MTU)
2671 addrconf_ifdown(dev, 1);
2673 break;
2675 case NETDEV_CHANGEMTU:
2676 if (idev && dev->mtu >= IPV6_MIN_MTU) {
2677 rt6_mtu_change(dev, dev->mtu);
2678 idev->cnf.mtu6 = dev->mtu;
2679 break;
2682 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2683 idev = ipv6_add_dev(dev);
2684 if (idev)
2685 break;
2689 * MTU falled under IPV6_MIN_MTU.
2690 * Stop IPv6 on this interface.
2693 case NETDEV_DOWN:
2694 case NETDEV_UNREGISTER:
2696 * Remove all addresses from this interface.
2698 addrconf_ifdown(dev, event != NETDEV_DOWN);
2699 break;
2701 case NETDEV_CHANGENAME:
2702 if (idev) {
2703 snmp6_unregister_dev(idev);
2704 addrconf_sysctl_unregister(idev);
2705 addrconf_sysctl_register(idev);
2706 err = snmp6_register_dev(idev);
2707 if (err)
2708 return notifier_from_errno(err);
2710 break;
2712 case NETDEV_PRE_TYPE_CHANGE:
2713 case NETDEV_POST_TYPE_CHANGE:
2714 addrconf_type_change(dev, event);
2715 break;
2718 return NOTIFY_OK;
2722 * addrconf module should be notified of a device going up
2724 static struct notifier_block ipv6_dev_notf = {
2725 .notifier_call = addrconf_notify,
2728 static void addrconf_type_change(struct net_device *dev, unsigned long event)
2730 struct inet6_dev *idev;
2731 ASSERT_RTNL();
2733 idev = __in6_dev_get(dev);
2735 if (event == NETDEV_POST_TYPE_CHANGE)
2736 ipv6_mc_remap(idev);
2737 else if (event == NETDEV_PRE_TYPE_CHANGE)
2738 ipv6_mc_unmap(idev);
2741 static int addrconf_ifdown(struct net_device *dev, int how)
2743 struct net *net = dev_net(dev);
2744 struct inet6_dev *idev;
2745 struct inet6_ifaddr *ifa;
2746 int state, i;
2748 ASSERT_RTNL();
2750 rt6_ifdown(net, dev);
2751 neigh_ifdown(&nd_tbl, dev);
2753 idev = __in6_dev_get(dev);
2754 if (idev == NULL)
2755 return -ENODEV;
2758 * Step 1: remove reference to ipv6 device from parent device.
2759 * Do not dev_put!
2761 if (how) {
2762 idev->dead = 1;
2764 /* protected by rtnl_lock */
2765 RCU_INIT_POINTER(dev->ip6_ptr, NULL);
2767 /* Step 1.5: remove snmp6 entry */
2768 snmp6_unregister_dev(idev);
2772 /* Step 2: clear hash table */
2773 for (i = 0; i < IN6_ADDR_HSIZE; i++) {
2774 struct hlist_head *h = &inet6_addr_lst[i];
2775 struct hlist_node *n;
2777 spin_lock_bh(&addrconf_hash_lock);
2778 restart:
2779 hlist_for_each_entry_rcu(ifa, n, h, addr_lst) {
2780 if (ifa->idev == idev) {
2781 hlist_del_init_rcu(&ifa->addr_lst);
2782 addrconf_del_timer(ifa);
2783 goto restart;
2786 spin_unlock_bh(&addrconf_hash_lock);
2789 write_lock_bh(&idev->lock);
2791 /* Step 2: clear flags for stateless addrconf */
2792 if (!how)
2793 idev->if_flags &= ~(IF_RS_SENT|IF_RA_RCVD|IF_READY);
2795 #ifdef CONFIG_IPV6_PRIVACY
2796 if (how && del_timer(&idev->regen_timer))
2797 in6_dev_put(idev);
2799 /* Step 3: clear tempaddr list */
2800 while (!list_empty(&idev->tempaddr_list)) {
2801 ifa = list_first_entry(&idev->tempaddr_list,
2802 struct inet6_ifaddr, tmp_list);
2803 list_del(&ifa->tmp_list);
2804 write_unlock_bh(&idev->lock);
2805 spin_lock_bh(&ifa->lock);
2807 if (ifa->ifpub) {
2808 in6_ifa_put(ifa->ifpub);
2809 ifa->ifpub = NULL;
2811 spin_unlock_bh(&ifa->lock);
2812 in6_ifa_put(ifa);
2813 write_lock_bh(&idev->lock);
2815 #endif
2817 while (!list_empty(&idev->addr_list)) {
2818 ifa = list_first_entry(&idev->addr_list,
2819 struct inet6_ifaddr, if_list);
2820 addrconf_del_timer(ifa);
2822 list_del(&ifa->if_list);
2824 write_unlock_bh(&idev->lock);
2826 spin_lock_bh(&ifa->state_lock);
2827 state = ifa->state;
2828 ifa->state = INET6_IFADDR_STATE_DEAD;
2829 spin_unlock_bh(&ifa->state_lock);
2831 if (state != INET6_IFADDR_STATE_DEAD) {
2832 __ipv6_ifa_notify(RTM_DELADDR, ifa);
2833 atomic_notifier_call_chain(&inet6addr_chain, NETDEV_DOWN, ifa);
2835 in6_ifa_put(ifa);
2837 write_lock_bh(&idev->lock);
2840 write_unlock_bh(&idev->lock);
2842 /* Step 5: Discard multicast list */
2843 if (how)
2844 ipv6_mc_destroy_dev(idev);
2845 else
2846 ipv6_mc_down(idev);
2848 idev->tstamp = jiffies;
2850 /* Last: Shot the device (if unregistered) */
2851 if (how) {
2852 addrconf_sysctl_unregister(idev);
2853 neigh_parms_release(&nd_tbl, idev->nd_parms);
2854 neigh_ifdown(&nd_tbl, dev);
2855 in6_dev_put(idev);
2857 return 0;
2860 static void addrconf_rs_timer(unsigned long data)
2862 struct inet6_ifaddr *ifp = (struct inet6_ifaddr *) data;
2863 struct inet6_dev *idev = ifp->idev;
2865 read_lock(&idev->lock);
2866 if (idev->dead || !(idev->if_flags & IF_READY))
2867 goto out;
2869 if (idev->cnf.forwarding)
2870 goto out;
2872 /* Announcement received after solicitation was sent */
2873 if (idev->if_flags & IF_RA_RCVD)
2874 goto out;
2876 spin_lock(&ifp->lock);
2877 if (ifp->probes++ < idev->cnf.rtr_solicits) {
2878 /* The wait after the last probe can be shorter */
2879 addrconf_mod_timer(ifp, AC_RS,
2880 (ifp->probes == idev->cnf.rtr_solicits) ?
2881 idev->cnf.rtr_solicit_delay :
2882 idev->cnf.rtr_solicit_interval);
2883 spin_unlock(&ifp->lock);
2885 ndisc_send_rs(idev->dev, &ifp->addr, &in6addr_linklocal_allrouters);
2886 } else {
2887 spin_unlock(&ifp->lock);
2889 * Note: we do not support deprecated "all on-link"
2890 * assumption any longer.
2892 printk(KERN_DEBUG "%s: no IPv6 routers present\n",
2893 idev->dev->name);
2896 out:
2897 read_unlock(&idev->lock);
2898 in6_ifa_put(ifp);
2902 * Duplicate Address Detection
2904 static void addrconf_dad_kick(struct inet6_ifaddr *ifp)
2906 unsigned long rand_num;
2907 struct inet6_dev *idev = ifp->idev;
2909 if (ifp->flags & IFA_F_OPTIMISTIC)
2910 rand_num = 0;
2911 else
2912 rand_num = net_random() % (idev->cnf.rtr_solicit_delay ? : 1);
2914 ifp->probes = idev->cnf.dad_transmits;
2915 addrconf_mod_timer(ifp, AC_DAD, rand_num);
2918 static void addrconf_dad_start(struct inet6_ifaddr *ifp, u32 flags)
2920 struct inet6_dev *idev = ifp->idev;
2921 struct net_device *dev = idev->dev;
2923 addrconf_join_solict(dev, &ifp->addr);
2925 net_srandom(ifp->addr.s6_addr32[3]);
2927 read_lock_bh(&idev->lock);
2928 spin_lock(&ifp->lock);
2929 if (ifp->state == INET6_IFADDR_STATE_DEAD)
2930 goto out;
2932 if (dev->flags&(IFF_NOARP|IFF_LOOPBACK) ||
2933 idev->cnf.accept_dad < 1 ||
2934 !(ifp->flags&IFA_F_TENTATIVE) ||
2935 ifp->flags & IFA_F_NODAD) {
2936 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
2937 spin_unlock(&ifp->lock);
2938 read_unlock_bh(&idev->lock);
2940 addrconf_dad_completed(ifp);
2941 return;
2944 if (!(idev->if_flags & IF_READY)) {
2945 spin_unlock(&ifp->lock);
2946 read_unlock_bh(&idev->lock);
2948 * If the device is not ready:
2949 * - keep it tentative if it is a permanent address.
2950 * - otherwise, kill it.
2952 in6_ifa_hold(ifp);
2953 addrconf_dad_stop(ifp, 0);
2954 return;
2958 * Optimistic nodes can start receiving
2959 * Frames right away
2961 if (ifp->flags & IFA_F_OPTIMISTIC)
2962 ip6_ins_rt(ifp->rt);
2964 addrconf_dad_kick(ifp);
2965 out:
2966 spin_unlock(&ifp->lock);
2967 read_unlock_bh(&idev->lock);
2970 static void addrconf_dad_timer(unsigned long data)
2972 struct inet6_ifaddr *ifp = (struct inet6_ifaddr *) data;
2973 struct inet6_dev *idev = ifp->idev;
2974 struct in6_addr mcaddr;
2976 if (!ifp->probes && addrconf_dad_end(ifp))
2977 goto out;
2979 read_lock(&idev->lock);
2980 if (idev->dead || !(idev->if_flags & IF_READY)) {
2981 read_unlock(&idev->lock);
2982 goto out;
2985 spin_lock(&ifp->lock);
2986 if (ifp->state == INET6_IFADDR_STATE_DEAD) {
2987 spin_unlock(&ifp->lock);
2988 read_unlock(&idev->lock);
2989 goto out;
2992 if (ifp->probes == 0) {
2994 * DAD was successful
2997 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
2998 spin_unlock(&ifp->lock);
2999 read_unlock(&idev->lock);
3001 addrconf_dad_completed(ifp);
3003 goto out;
3006 ifp->probes--;
3007 addrconf_mod_timer(ifp, AC_DAD, ifp->idev->nd_parms->retrans_time);
3008 spin_unlock(&ifp->lock);
3009 read_unlock(&idev->lock);
3011 /* send a neighbour solicitation for our addr */
3012 addrconf_addr_solict_mult(&ifp->addr, &mcaddr);
3013 ndisc_send_ns(ifp->idev->dev, NULL, &ifp->addr, &mcaddr, &in6addr_any);
3014 out:
3015 in6_ifa_put(ifp);
3018 static void addrconf_dad_completed(struct inet6_ifaddr *ifp)
3020 struct net_device *dev = ifp->idev->dev;
3023 * Configure the address for reception. Now it is valid.
3026 ipv6_ifa_notify(RTM_NEWADDR, ifp);
3028 /* If added prefix is link local and we are prepared to process
3029 router advertisements, start sending router solicitations.
3032 if (((ifp->idev->cnf.accept_ra == 1 && !ifp->idev->cnf.forwarding) ||
3033 ifp->idev->cnf.accept_ra == 2) &&
3034 ifp->idev->cnf.rtr_solicits > 0 &&
3035 (dev->flags&IFF_LOOPBACK) == 0 &&
3036 (ipv6_addr_type(&ifp->addr) & IPV6_ADDR_LINKLOCAL)) {
3038 * If a host as already performed a random delay
3039 * [...] as part of DAD [...] there is no need
3040 * to delay again before sending the first RS
3042 ndisc_send_rs(ifp->idev->dev, &ifp->addr, &in6addr_linklocal_allrouters);
3044 spin_lock_bh(&ifp->lock);
3045 ifp->probes = 1;
3046 ifp->idev->if_flags |= IF_RS_SENT;
3047 addrconf_mod_timer(ifp, AC_RS, ifp->idev->cnf.rtr_solicit_interval);
3048 spin_unlock_bh(&ifp->lock);
3052 static void addrconf_dad_run(struct inet6_dev *idev)
3054 struct inet6_ifaddr *ifp;
3056 read_lock_bh(&idev->lock);
3057 list_for_each_entry(ifp, &idev->addr_list, if_list) {
3058 spin_lock(&ifp->lock);
3059 if (ifp->flags & IFA_F_TENTATIVE &&
3060 ifp->state == INET6_IFADDR_STATE_DAD)
3061 addrconf_dad_kick(ifp);
3062 spin_unlock(&ifp->lock);
3064 read_unlock_bh(&idev->lock);
3067 #ifdef CONFIG_PROC_FS
3068 struct if6_iter_state {
3069 struct seq_net_private p;
3070 int bucket;
3071 int offset;
3074 static struct inet6_ifaddr *if6_get_first(struct seq_file *seq, loff_t pos)
3076 struct inet6_ifaddr *ifa = NULL;
3077 struct if6_iter_state *state = seq->private;
3078 struct net *net = seq_file_net(seq);
3079 int p = 0;
3081 /* initial bucket if pos is 0 */
3082 if (pos == 0) {
3083 state->bucket = 0;
3084 state->offset = 0;
3087 for (; state->bucket < IN6_ADDR_HSIZE; ++state->bucket) {
3088 struct hlist_node *n;
3089 hlist_for_each_entry_rcu_bh(ifa, n, &inet6_addr_lst[state->bucket],
3090 addr_lst) {
3091 /* sync with offset */
3092 if (p < state->offset) {
3093 p++;
3094 continue;
3096 state->offset++;
3097 if (net_eq(dev_net(ifa->idev->dev), net))
3098 return ifa;
3101 /* prepare for next bucket */
3102 state->offset = 0;
3103 p = 0;
3105 return NULL;
3108 static struct inet6_ifaddr *if6_get_next(struct seq_file *seq,
3109 struct inet6_ifaddr *ifa)
3111 struct if6_iter_state *state = seq->private;
3112 struct net *net = seq_file_net(seq);
3113 struct hlist_node *n = &ifa->addr_lst;
3115 hlist_for_each_entry_continue_rcu_bh(ifa, n, addr_lst) {
3116 state->offset++;
3117 if (net_eq(dev_net(ifa->idev->dev), net))
3118 return ifa;
3121 while (++state->bucket < IN6_ADDR_HSIZE) {
3122 state->offset = 0;
3123 hlist_for_each_entry_rcu_bh(ifa, n,
3124 &inet6_addr_lst[state->bucket], addr_lst) {
3125 state->offset++;
3126 if (net_eq(dev_net(ifa->idev->dev), net))
3127 return ifa;
3131 return NULL;
3134 static void *if6_seq_start(struct seq_file *seq, loff_t *pos)
3135 __acquires(rcu_bh)
3137 rcu_read_lock_bh();
3138 return if6_get_first(seq, *pos);
3141 static void *if6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
3143 struct inet6_ifaddr *ifa;
3145 ifa = if6_get_next(seq, v);
3146 ++*pos;
3147 return ifa;
3150 static void if6_seq_stop(struct seq_file *seq, void *v)
3151 __releases(rcu_bh)
3153 rcu_read_unlock_bh();
3156 static int if6_seq_show(struct seq_file *seq, void *v)
3158 struct inet6_ifaddr *ifp = (struct inet6_ifaddr *)v;
3159 seq_printf(seq, "%pi6 %02x %02x %02x %02x %8s\n",
3160 &ifp->addr,
3161 ifp->idev->dev->ifindex,
3162 ifp->prefix_len,
3163 ifp->scope,
3164 ifp->flags,
3165 ifp->idev->dev->name);
3166 return 0;
3169 static const struct seq_operations if6_seq_ops = {
3170 .start = if6_seq_start,
3171 .next = if6_seq_next,
3172 .show = if6_seq_show,
3173 .stop = if6_seq_stop,
3176 static int if6_seq_open(struct inode *inode, struct file *file)
3178 return seq_open_net(inode, file, &if6_seq_ops,
3179 sizeof(struct if6_iter_state));
3182 static const struct file_operations if6_fops = {
3183 .owner = THIS_MODULE,
3184 .open = if6_seq_open,
3185 .read = seq_read,
3186 .llseek = seq_lseek,
3187 .release = seq_release_net,
3190 static int __net_init if6_proc_net_init(struct net *net)
3192 if (!proc_net_fops_create(net, "if_inet6", S_IRUGO, &if6_fops))
3193 return -ENOMEM;
3194 return 0;
3197 static void __net_exit if6_proc_net_exit(struct net *net)
3199 proc_net_remove(net, "if_inet6");
3202 static struct pernet_operations if6_proc_net_ops = {
3203 .init = if6_proc_net_init,
3204 .exit = if6_proc_net_exit,
3207 int __init if6_proc_init(void)
3209 return register_pernet_subsys(&if6_proc_net_ops);
3212 void if6_proc_exit(void)
3214 unregister_pernet_subsys(&if6_proc_net_ops);
3216 #endif /* CONFIG_PROC_FS */
3218 #if defined(CONFIG_IPV6_MIP6) || defined(CONFIG_IPV6_MIP6_MODULE)
3219 /* Check if address is a home address configured on any interface. */
3220 int ipv6_chk_home_addr(struct net *net, const struct in6_addr *addr)
3222 int ret = 0;
3223 struct inet6_ifaddr *ifp = NULL;
3224 struct hlist_node *n;
3225 unsigned int hash = ipv6_addr_hash(addr);
3227 rcu_read_lock_bh();
3228 hlist_for_each_entry_rcu_bh(ifp, n, &inet6_addr_lst[hash], addr_lst) {
3229 if (!net_eq(dev_net(ifp->idev->dev), net))
3230 continue;
3231 if (ipv6_addr_equal(&ifp->addr, addr) &&
3232 (ifp->flags & IFA_F_HOMEADDRESS)) {
3233 ret = 1;
3234 break;
3237 rcu_read_unlock_bh();
3238 return ret;
3240 #endif
3243 * Periodic address status verification
3246 static void addrconf_verify(unsigned long foo)
3248 unsigned long now, next, next_sec, next_sched;
3249 struct inet6_ifaddr *ifp;
3250 struct hlist_node *node;
3251 int i;
3253 rcu_read_lock_bh();
3254 spin_lock(&addrconf_verify_lock);
3255 now = jiffies;
3256 next = round_jiffies_up(now + ADDR_CHECK_FREQUENCY);
3258 del_timer(&addr_chk_timer);
3260 for (i = 0; i < IN6_ADDR_HSIZE; i++) {
3261 restart:
3262 hlist_for_each_entry_rcu_bh(ifp, node,
3263 &inet6_addr_lst[i], addr_lst) {
3264 unsigned long age;
3266 if (ifp->flags & IFA_F_PERMANENT)
3267 continue;
3269 spin_lock(&ifp->lock);
3270 /* We try to batch several events at once. */
3271 age = (now - ifp->tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
3273 if (ifp->valid_lft != INFINITY_LIFE_TIME &&
3274 age >= ifp->valid_lft) {
3275 spin_unlock(&ifp->lock);
3276 in6_ifa_hold(ifp);
3277 ipv6_del_addr(ifp);
3278 goto restart;
3279 } else if (ifp->prefered_lft == INFINITY_LIFE_TIME) {
3280 spin_unlock(&ifp->lock);
3281 continue;
3282 } else if (age >= ifp->prefered_lft) {
3283 /* jiffies - ifp->tstamp > age >= ifp->prefered_lft */
3284 int deprecate = 0;
3286 if (!(ifp->flags&IFA_F_DEPRECATED)) {
3287 deprecate = 1;
3288 ifp->flags |= IFA_F_DEPRECATED;
3291 if (time_before(ifp->tstamp + ifp->valid_lft * HZ, next))
3292 next = ifp->tstamp + ifp->valid_lft * HZ;
3294 spin_unlock(&ifp->lock);
3296 if (deprecate) {
3297 in6_ifa_hold(ifp);
3299 ipv6_ifa_notify(0, ifp);
3300 in6_ifa_put(ifp);
3301 goto restart;
3303 #ifdef CONFIG_IPV6_PRIVACY
3304 } else if ((ifp->flags&IFA_F_TEMPORARY) &&
3305 !(ifp->flags&IFA_F_TENTATIVE)) {
3306 unsigned long regen_advance = ifp->idev->cnf.regen_max_retry *
3307 ifp->idev->cnf.dad_transmits *
3308 ifp->idev->nd_parms->retrans_time / HZ;
3310 if (age >= ifp->prefered_lft - regen_advance) {
3311 struct inet6_ifaddr *ifpub = ifp->ifpub;
3312 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3313 next = ifp->tstamp + ifp->prefered_lft * HZ;
3314 if (!ifp->regen_count && ifpub) {
3315 ifp->regen_count++;
3316 in6_ifa_hold(ifp);
3317 in6_ifa_hold(ifpub);
3318 spin_unlock(&ifp->lock);
3320 spin_lock(&ifpub->lock);
3321 ifpub->regen_count = 0;
3322 spin_unlock(&ifpub->lock);
3323 ipv6_create_tempaddr(ifpub, ifp);
3324 in6_ifa_put(ifpub);
3325 in6_ifa_put(ifp);
3326 goto restart;
3328 } else if (time_before(ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ, next))
3329 next = ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ;
3330 spin_unlock(&ifp->lock);
3331 #endif
3332 } else {
3333 /* ifp->prefered_lft <= ifp->valid_lft */
3334 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3335 next = ifp->tstamp + ifp->prefered_lft * HZ;
3336 spin_unlock(&ifp->lock);
3341 next_sec = round_jiffies_up(next);
3342 next_sched = next;
3344 /* If rounded timeout is accurate enough, accept it. */
3345 if (time_before(next_sec, next + ADDRCONF_TIMER_FUZZ))
3346 next_sched = next_sec;
3348 /* And minimum interval is ADDRCONF_TIMER_FUZZ_MAX. */
3349 if (time_before(next_sched, jiffies + ADDRCONF_TIMER_FUZZ_MAX))
3350 next_sched = jiffies + ADDRCONF_TIMER_FUZZ_MAX;
3352 ADBG((KERN_DEBUG "now = %lu, schedule = %lu, rounded schedule = %lu => %lu\n",
3353 now, next, next_sec, next_sched));
3355 addr_chk_timer.expires = next_sched;
3356 add_timer(&addr_chk_timer);
3357 spin_unlock(&addrconf_verify_lock);
3358 rcu_read_unlock_bh();
3361 static struct in6_addr *extract_addr(struct nlattr *addr, struct nlattr *local)
3363 struct in6_addr *pfx = NULL;
3365 if (addr)
3366 pfx = nla_data(addr);
3368 if (local) {
3369 if (pfx && nla_memcmp(local, pfx, sizeof(*pfx)))
3370 pfx = NULL;
3371 else
3372 pfx = nla_data(local);
3375 return pfx;
3378 static const struct nla_policy ifa_ipv6_policy[IFA_MAX+1] = {
3379 [IFA_ADDRESS] = { .len = sizeof(struct in6_addr) },
3380 [IFA_LOCAL] = { .len = sizeof(struct in6_addr) },
3381 [IFA_CACHEINFO] = { .len = sizeof(struct ifa_cacheinfo) },
3384 static int
3385 inet6_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
3387 struct net *net = sock_net(skb->sk);
3388 struct ifaddrmsg *ifm;
3389 struct nlattr *tb[IFA_MAX+1];
3390 struct in6_addr *pfx;
3391 int err;
3393 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3394 if (err < 0)
3395 return err;
3397 ifm = nlmsg_data(nlh);
3398 pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3399 if (pfx == NULL)
3400 return -EINVAL;
3402 return inet6_addr_del(net, ifm->ifa_index, pfx, ifm->ifa_prefixlen);
3405 static int inet6_addr_modify(struct inet6_ifaddr *ifp, u8 ifa_flags,
3406 u32 prefered_lft, u32 valid_lft)
3408 u32 flags;
3409 clock_t expires;
3410 unsigned long timeout;
3412 if (!valid_lft || (prefered_lft > valid_lft))
3413 return -EINVAL;
3415 timeout = addrconf_timeout_fixup(valid_lft, HZ);
3416 if (addrconf_finite_timeout(timeout)) {
3417 expires = jiffies_to_clock_t(timeout * HZ);
3418 valid_lft = timeout;
3419 flags = RTF_EXPIRES;
3420 } else {
3421 expires = 0;
3422 flags = 0;
3423 ifa_flags |= IFA_F_PERMANENT;
3426 timeout = addrconf_timeout_fixup(prefered_lft, HZ);
3427 if (addrconf_finite_timeout(timeout)) {
3428 if (timeout == 0)
3429 ifa_flags |= IFA_F_DEPRECATED;
3430 prefered_lft = timeout;
3433 spin_lock_bh(&ifp->lock);
3434 ifp->flags = (ifp->flags & ~(IFA_F_DEPRECATED | IFA_F_PERMANENT | IFA_F_NODAD | IFA_F_HOMEADDRESS)) | ifa_flags;
3435 ifp->tstamp = jiffies;
3436 ifp->valid_lft = valid_lft;
3437 ifp->prefered_lft = prefered_lft;
3439 spin_unlock_bh(&ifp->lock);
3440 if (!(ifp->flags&IFA_F_TENTATIVE))
3441 ipv6_ifa_notify(0, ifp);
3443 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, ifp->idev->dev,
3444 expires, flags);
3445 addrconf_verify(0);
3447 return 0;
3450 static int
3451 inet6_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
3453 struct net *net = sock_net(skb->sk);
3454 struct ifaddrmsg *ifm;
3455 struct nlattr *tb[IFA_MAX+1];
3456 struct in6_addr *pfx;
3457 struct inet6_ifaddr *ifa;
3458 struct net_device *dev;
3459 u32 valid_lft = INFINITY_LIFE_TIME, preferred_lft = INFINITY_LIFE_TIME;
3460 u8 ifa_flags;
3461 int err;
3463 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3464 if (err < 0)
3465 return err;
3467 ifm = nlmsg_data(nlh);
3468 pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3469 if (pfx == NULL)
3470 return -EINVAL;
3472 if (tb[IFA_CACHEINFO]) {
3473 struct ifa_cacheinfo *ci;
3475 ci = nla_data(tb[IFA_CACHEINFO]);
3476 valid_lft = ci->ifa_valid;
3477 preferred_lft = ci->ifa_prefered;
3478 } else {
3479 preferred_lft = INFINITY_LIFE_TIME;
3480 valid_lft = INFINITY_LIFE_TIME;
3483 dev = __dev_get_by_index(net, ifm->ifa_index);
3484 if (dev == NULL)
3485 return -ENODEV;
3487 /* We ignore other flags so far. */
3488 ifa_flags = ifm->ifa_flags & (IFA_F_NODAD | IFA_F_HOMEADDRESS);
3490 ifa = ipv6_get_ifaddr(net, pfx, dev, 1);
3491 if (ifa == NULL) {
3493 * It would be best to check for !NLM_F_CREATE here but
3494 * userspace alreay relies on not having to provide this.
3496 return inet6_addr_add(net, ifm->ifa_index, pfx,
3497 ifm->ifa_prefixlen, ifa_flags,
3498 preferred_lft, valid_lft);
3501 if (nlh->nlmsg_flags & NLM_F_EXCL ||
3502 !(nlh->nlmsg_flags & NLM_F_REPLACE))
3503 err = -EEXIST;
3504 else
3505 err = inet6_addr_modify(ifa, ifa_flags, preferred_lft, valid_lft);
3507 in6_ifa_put(ifa);
3509 return err;
3512 static void put_ifaddrmsg(struct nlmsghdr *nlh, u8 prefixlen, u8 flags,
3513 u8 scope, int ifindex)
3515 struct ifaddrmsg *ifm;
3517 ifm = nlmsg_data(nlh);
3518 ifm->ifa_family = AF_INET6;
3519 ifm->ifa_prefixlen = prefixlen;
3520 ifm->ifa_flags = flags;
3521 ifm->ifa_scope = scope;
3522 ifm->ifa_index = ifindex;
3525 static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
3526 unsigned long tstamp, u32 preferred, u32 valid)
3528 struct ifa_cacheinfo ci;
3530 ci.cstamp = cstamp_delta(cstamp);
3531 ci.tstamp = cstamp_delta(tstamp);
3532 ci.ifa_prefered = preferred;
3533 ci.ifa_valid = valid;
3535 return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
3538 static inline int rt_scope(int ifa_scope)
3540 if (ifa_scope & IFA_HOST)
3541 return RT_SCOPE_HOST;
3542 else if (ifa_scope & IFA_LINK)
3543 return RT_SCOPE_LINK;
3544 else if (ifa_scope & IFA_SITE)
3545 return RT_SCOPE_SITE;
3546 else
3547 return RT_SCOPE_UNIVERSE;
3550 static inline int inet6_ifaddr_msgsize(void)
3552 return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
3553 + nla_total_size(16) /* IFA_ADDRESS */
3554 + nla_total_size(sizeof(struct ifa_cacheinfo));
3557 static int inet6_fill_ifaddr(struct sk_buff *skb, struct inet6_ifaddr *ifa,
3558 u32 pid, u32 seq, int event, unsigned int flags)
3560 struct nlmsghdr *nlh;
3561 u32 preferred, valid;
3563 nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3564 if (nlh == NULL)
3565 return -EMSGSIZE;
3567 put_ifaddrmsg(nlh, ifa->prefix_len, ifa->flags, rt_scope(ifa->scope),
3568 ifa->idev->dev->ifindex);
3570 if (!(ifa->flags&IFA_F_PERMANENT)) {
3571 preferred = ifa->prefered_lft;
3572 valid = ifa->valid_lft;
3573 if (preferred != INFINITY_LIFE_TIME) {
3574 long tval = (jiffies - ifa->tstamp)/HZ;
3575 if (preferred > tval)
3576 preferred -= tval;
3577 else
3578 preferred = 0;
3579 if (valid != INFINITY_LIFE_TIME) {
3580 if (valid > tval)
3581 valid -= tval;
3582 else
3583 valid = 0;
3586 } else {
3587 preferred = INFINITY_LIFE_TIME;
3588 valid = INFINITY_LIFE_TIME;
3591 if (nla_put(skb, IFA_ADDRESS, 16, &ifa->addr) < 0 ||
3592 put_cacheinfo(skb, ifa->cstamp, ifa->tstamp, preferred, valid) < 0) {
3593 nlmsg_cancel(skb, nlh);
3594 return -EMSGSIZE;
3597 return nlmsg_end(skb, nlh);
3600 static int inet6_fill_ifmcaddr(struct sk_buff *skb, struct ifmcaddr6 *ifmca,
3601 u32 pid, u32 seq, int event, u16 flags)
3603 struct nlmsghdr *nlh;
3604 u8 scope = RT_SCOPE_UNIVERSE;
3605 int ifindex = ifmca->idev->dev->ifindex;
3607 if (ipv6_addr_scope(&ifmca->mca_addr) & IFA_SITE)
3608 scope = RT_SCOPE_SITE;
3610 nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3611 if (nlh == NULL)
3612 return -EMSGSIZE;
3614 put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3615 if (nla_put(skb, IFA_MULTICAST, 16, &ifmca->mca_addr) < 0 ||
3616 put_cacheinfo(skb, ifmca->mca_cstamp, ifmca->mca_tstamp,
3617 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3618 nlmsg_cancel(skb, nlh);
3619 return -EMSGSIZE;
3622 return nlmsg_end(skb, nlh);
3625 static int inet6_fill_ifacaddr(struct sk_buff *skb, struct ifacaddr6 *ifaca,
3626 u32 pid, u32 seq, int event, unsigned int flags)
3628 struct nlmsghdr *nlh;
3629 u8 scope = RT_SCOPE_UNIVERSE;
3630 int ifindex = ifaca->aca_idev->dev->ifindex;
3632 if (ipv6_addr_scope(&ifaca->aca_addr) & IFA_SITE)
3633 scope = RT_SCOPE_SITE;
3635 nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3636 if (nlh == NULL)
3637 return -EMSGSIZE;
3639 put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3640 if (nla_put(skb, IFA_ANYCAST, 16, &ifaca->aca_addr) < 0 ||
3641 put_cacheinfo(skb, ifaca->aca_cstamp, ifaca->aca_tstamp,
3642 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3643 nlmsg_cancel(skb, nlh);
3644 return -EMSGSIZE;
3647 return nlmsg_end(skb, nlh);
3650 enum addr_type_t {
3651 UNICAST_ADDR,
3652 MULTICAST_ADDR,
3653 ANYCAST_ADDR,
3656 /* called with rcu_read_lock() */
3657 static int in6_dump_addrs(struct inet6_dev *idev, struct sk_buff *skb,
3658 struct netlink_callback *cb, enum addr_type_t type,
3659 int s_ip_idx, int *p_ip_idx)
3661 struct ifmcaddr6 *ifmca;
3662 struct ifacaddr6 *ifaca;
3663 int err = 1;
3664 int ip_idx = *p_ip_idx;
3666 read_lock_bh(&idev->lock);
3667 switch (type) {
3668 case UNICAST_ADDR: {
3669 struct inet6_ifaddr *ifa;
3671 /* unicast address incl. temp addr */
3672 list_for_each_entry(ifa, &idev->addr_list, if_list) {
3673 if (++ip_idx < s_ip_idx)
3674 continue;
3675 err = inet6_fill_ifaddr(skb, ifa,
3676 NETLINK_CB(cb->skb).pid,
3677 cb->nlh->nlmsg_seq,
3678 RTM_NEWADDR,
3679 NLM_F_MULTI);
3680 if (err <= 0)
3681 break;
3683 break;
3685 case MULTICAST_ADDR:
3686 /* multicast address */
3687 for (ifmca = idev->mc_list; ifmca;
3688 ifmca = ifmca->next, ip_idx++) {
3689 if (ip_idx < s_ip_idx)
3690 continue;
3691 err = inet6_fill_ifmcaddr(skb, ifmca,
3692 NETLINK_CB(cb->skb).pid,
3693 cb->nlh->nlmsg_seq,
3694 RTM_GETMULTICAST,
3695 NLM_F_MULTI);
3696 if (err <= 0)
3697 break;
3699 break;
3700 case ANYCAST_ADDR:
3701 /* anycast address */
3702 for (ifaca = idev->ac_list; ifaca;
3703 ifaca = ifaca->aca_next, ip_idx++) {
3704 if (ip_idx < s_ip_idx)
3705 continue;
3706 err = inet6_fill_ifacaddr(skb, ifaca,
3707 NETLINK_CB(cb->skb).pid,
3708 cb->nlh->nlmsg_seq,
3709 RTM_GETANYCAST,
3710 NLM_F_MULTI);
3711 if (err <= 0)
3712 break;
3714 break;
3715 default:
3716 break;
3718 read_unlock_bh(&idev->lock);
3719 *p_ip_idx = ip_idx;
3720 return err;
3723 static int inet6_dump_addr(struct sk_buff *skb, struct netlink_callback *cb,
3724 enum addr_type_t type)
3726 struct net *net = sock_net(skb->sk);
3727 int h, s_h;
3728 int idx, ip_idx;
3729 int s_idx, s_ip_idx;
3730 struct net_device *dev;
3731 struct inet6_dev *idev;
3732 struct hlist_head *head;
3733 struct hlist_node *node;
3735 s_h = cb->args[0];
3736 s_idx = idx = cb->args[1];
3737 s_ip_idx = ip_idx = cb->args[2];
3739 rcu_read_lock();
3740 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
3741 idx = 0;
3742 head = &net->dev_index_head[h];
3743 hlist_for_each_entry_rcu(dev, node, head, index_hlist) {
3744 if (idx < s_idx)
3745 goto cont;
3746 if (h > s_h || idx > s_idx)
3747 s_ip_idx = 0;
3748 ip_idx = 0;
3749 idev = __in6_dev_get(dev);
3750 if (!idev)
3751 goto cont;
3753 if (in6_dump_addrs(idev, skb, cb, type,
3754 s_ip_idx, &ip_idx) <= 0)
3755 goto done;
3756 cont:
3757 idx++;
3760 done:
3761 rcu_read_unlock();
3762 cb->args[0] = h;
3763 cb->args[1] = idx;
3764 cb->args[2] = ip_idx;
3766 return skb->len;
3769 static int inet6_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
3771 enum addr_type_t type = UNICAST_ADDR;
3773 return inet6_dump_addr(skb, cb, type);
3776 static int inet6_dump_ifmcaddr(struct sk_buff *skb, struct netlink_callback *cb)
3778 enum addr_type_t type = MULTICAST_ADDR;
3780 return inet6_dump_addr(skb, cb, type);
3784 static int inet6_dump_ifacaddr(struct sk_buff *skb, struct netlink_callback *cb)
3786 enum addr_type_t type = ANYCAST_ADDR;
3788 return inet6_dump_addr(skb, cb, type);
3791 static int inet6_rtm_getaddr(struct sk_buff *in_skb, struct nlmsghdr* nlh,
3792 void *arg)
3794 struct net *net = sock_net(in_skb->sk);
3795 struct ifaddrmsg *ifm;
3796 struct nlattr *tb[IFA_MAX+1];
3797 struct in6_addr *addr = NULL;
3798 struct net_device *dev = NULL;
3799 struct inet6_ifaddr *ifa;
3800 struct sk_buff *skb;
3801 int err;
3803 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3804 if (err < 0)
3805 goto errout;
3807 addr = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3808 if (addr == NULL) {
3809 err = -EINVAL;
3810 goto errout;
3813 ifm = nlmsg_data(nlh);
3814 if (ifm->ifa_index)
3815 dev = __dev_get_by_index(net, ifm->ifa_index);
3817 ifa = ipv6_get_ifaddr(net, addr, dev, 1);
3818 if (!ifa) {
3819 err = -EADDRNOTAVAIL;
3820 goto errout;
3823 skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_KERNEL);
3824 if (!skb) {
3825 err = -ENOBUFS;
3826 goto errout_ifa;
3829 err = inet6_fill_ifaddr(skb, ifa, NETLINK_CB(in_skb).pid,
3830 nlh->nlmsg_seq, RTM_NEWADDR, 0);
3831 if (err < 0) {
3832 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
3833 WARN_ON(err == -EMSGSIZE);
3834 kfree_skb(skb);
3835 goto errout_ifa;
3837 err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).pid);
3838 errout_ifa:
3839 in6_ifa_put(ifa);
3840 errout:
3841 return err;
3844 static void inet6_ifa_notify(int event, struct inet6_ifaddr *ifa)
3846 struct sk_buff *skb;
3847 struct net *net = dev_net(ifa->idev->dev);
3848 int err = -ENOBUFS;
3850 skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_ATOMIC);
3851 if (skb == NULL)
3852 goto errout;
3854 err = inet6_fill_ifaddr(skb, ifa, 0, 0, event, 0);
3855 if (err < 0) {
3856 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
3857 WARN_ON(err == -EMSGSIZE);
3858 kfree_skb(skb);
3859 goto errout;
3861 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
3862 return;
3863 errout:
3864 if (err < 0)
3865 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
3868 static inline void ipv6_store_devconf(struct ipv6_devconf *cnf,
3869 __s32 *array, int bytes)
3871 BUG_ON(bytes < (DEVCONF_MAX * 4));
3873 memset(array, 0, bytes);
3874 array[DEVCONF_FORWARDING] = cnf->forwarding;
3875 array[DEVCONF_HOPLIMIT] = cnf->hop_limit;
3876 array[DEVCONF_MTU6] = cnf->mtu6;
3877 array[DEVCONF_ACCEPT_RA] = cnf->accept_ra;
3878 array[DEVCONF_ACCEPT_REDIRECTS] = cnf->accept_redirects;
3879 array[DEVCONF_AUTOCONF] = cnf->autoconf;
3880 array[DEVCONF_DAD_TRANSMITS] = cnf->dad_transmits;
3881 array[DEVCONF_RTR_SOLICITS] = cnf->rtr_solicits;
3882 array[DEVCONF_RTR_SOLICIT_INTERVAL] =
3883 jiffies_to_msecs(cnf->rtr_solicit_interval);
3884 array[DEVCONF_RTR_SOLICIT_DELAY] =
3885 jiffies_to_msecs(cnf->rtr_solicit_delay);
3886 array[DEVCONF_FORCE_MLD_VERSION] = cnf->force_mld_version;
3887 #ifdef CONFIG_IPV6_PRIVACY
3888 array[DEVCONF_USE_TEMPADDR] = cnf->use_tempaddr;
3889 array[DEVCONF_TEMP_VALID_LFT] = cnf->temp_valid_lft;
3890 array[DEVCONF_TEMP_PREFERED_LFT] = cnf->temp_prefered_lft;
3891 array[DEVCONF_REGEN_MAX_RETRY] = cnf->regen_max_retry;
3892 array[DEVCONF_MAX_DESYNC_FACTOR] = cnf->max_desync_factor;
3893 #endif
3894 array[DEVCONF_MAX_ADDRESSES] = cnf->max_addresses;
3895 array[DEVCONF_ACCEPT_RA_DEFRTR] = cnf->accept_ra_defrtr;
3896 array[DEVCONF_ACCEPT_RA_PINFO] = cnf->accept_ra_pinfo;
3897 #ifdef CONFIG_IPV6_ROUTER_PREF
3898 array[DEVCONF_ACCEPT_RA_RTR_PREF] = cnf->accept_ra_rtr_pref;
3899 array[DEVCONF_RTR_PROBE_INTERVAL] =
3900 jiffies_to_msecs(cnf->rtr_probe_interval);
3901 #ifdef CONFIG_IPV6_ROUTE_INFO
3902 array[DEVCONF_ACCEPT_RA_RT_INFO_MAX_PLEN] = cnf->accept_ra_rt_info_max_plen;
3903 #endif
3904 #endif
3905 array[DEVCONF_PROXY_NDP] = cnf->proxy_ndp;
3906 array[DEVCONF_ACCEPT_SOURCE_ROUTE] = cnf->accept_source_route;
3907 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
3908 array[DEVCONF_OPTIMISTIC_DAD] = cnf->optimistic_dad;
3909 #endif
3910 #ifdef CONFIG_IPV6_MROUTE
3911 array[DEVCONF_MC_FORWARDING] = cnf->mc_forwarding;
3912 #endif
3913 array[DEVCONF_DISABLE_IPV6] = cnf->disable_ipv6;
3914 array[DEVCONF_ACCEPT_DAD] = cnf->accept_dad;
3915 array[DEVCONF_FORCE_TLLAO] = cnf->force_tllao;
3918 static inline size_t inet6_ifla6_size(void)
3920 return nla_total_size(4) /* IFLA_INET6_FLAGS */
3921 + nla_total_size(sizeof(struct ifla_cacheinfo))
3922 + nla_total_size(DEVCONF_MAX * 4) /* IFLA_INET6_CONF */
3923 + nla_total_size(IPSTATS_MIB_MAX * 8) /* IFLA_INET6_STATS */
3924 + nla_total_size(ICMP6_MIB_MAX * 8); /* IFLA_INET6_ICMP6STATS */
3927 static inline size_t inet6_if_nlmsg_size(void)
3929 return NLMSG_ALIGN(sizeof(struct ifinfomsg))
3930 + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
3931 + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
3932 + nla_total_size(4) /* IFLA_MTU */
3933 + nla_total_size(4) /* IFLA_LINK */
3934 + nla_total_size(inet6_ifla6_size()); /* IFLA_PROTINFO */
3937 static inline void __snmp6_fill_statsdev(u64 *stats, atomic_long_t *mib,
3938 int items, int bytes)
3940 int i;
3941 int pad = bytes - sizeof(u64) * items;
3942 BUG_ON(pad < 0);
3944 /* Use put_unaligned() because stats may not be aligned for u64. */
3945 put_unaligned(items, &stats[0]);
3946 for (i = 1; i < items; i++)
3947 put_unaligned(atomic_long_read(&mib[i]), &stats[i]);
3949 memset(&stats[items], 0, pad);
3952 static inline void __snmp6_fill_stats64(u64 *stats, void __percpu **mib,
3953 int items, int bytes, size_t syncpoff)
3955 int i;
3956 int pad = bytes - sizeof(u64) * items;
3957 BUG_ON(pad < 0);
3959 /* Use put_unaligned() because stats may not be aligned for u64. */
3960 put_unaligned(items, &stats[0]);
3961 for (i = 1; i < items; i++)
3962 put_unaligned(snmp_fold_field64(mib, i, syncpoff), &stats[i]);
3964 memset(&stats[items], 0, pad);
3967 static void snmp6_fill_stats(u64 *stats, struct inet6_dev *idev, int attrtype,
3968 int bytes)
3970 switch (attrtype) {
3971 case IFLA_INET6_STATS:
3972 __snmp6_fill_stats64(stats, (void __percpu **)idev->stats.ipv6,
3973 IPSTATS_MIB_MAX, bytes, offsetof(struct ipstats_mib, syncp));
3974 break;
3975 case IFLA_INET6_ICMP6STATS:
3976 __snmp6_fill_statsdev(stats, idev->stats.icmpv6dev->mibs, ICMP6_MIB_MAX, bytes);
3977 break;
3981 static int inet6_fill_ifla6_attrs(struct sk_buff *skb, struct inet6_dev *idev)
3983 struct nlattr *nla;
3984 struct ifla_cacheinfo ci;
3986 NLA_PUT_U32(skb, IFLA_INET6_FLAGS, idev->if_flags);
3988 ci.max_reasm_len = IPV6_MAXPLEN;
3989 ci.tstamp = cstamp_delta(idev->tstamp);
3990 ci.reachable_time = jiffies_to_msecs(idev->nd_parms->reachable_time);
3991 ci.retrans_time = jiffies_to_msecs(idev->nd_parms->retrans_time);
3992 NLA_PUT(skb, IFLA_INET6_CACHEINFO, sizeof(ci), &ci);
3994 nla = nla_reserve(skb, IFLA_INET6_CONF, DEVCONF_MAX * sizeof(s32));
3995 if (nla == NULL)
3996 goto nla_put_failure;
3997 ipv6_store_devconf(&idev->cnf, nla_data(nla), nla_len(nla));
3999 /* XXX - MC not implemented */
4001 nla = nla_reserve(skb, IFLA_INET6_STATS, IPSTATS_MIB_MAX * sizeof(u64));
4002 if (nla == NULL)
4003 goto nla_put_failure;
4004 snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_STATS, nla_len(nla));
4006 nla = nla_reserve(skb, IFLA_INET6_ICMP6STATS, ICMP6_MIB_MAX * sizeof(u64));
4007 if (nla == NULL)
4008 goto nla_put_failure;
4009 snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_ICMP6STATS, nla_len(nla));
4011 return 0;
4013 nla_put_failure:
4014 return -EMSGSIZE;
4017 static size_t inet6_get_link_af_size(const struct net_device *dev)
4019 if (!__in6_dev_get(dev))
4020 return 0;
4022 return inet6_ifla6_size();
4025 static int inet6_fill_link_af(struct sk_buff *skb, const struct net_device *dev)
4027 struct inet6_dev *idev = __in6_dev_get(dev);
4029 if (!idev)
4030 return -ENODATA;
4032 if (inet6_fill_ifla6_attrs(skb, idev) < 0)
4033 return -EMSGSIZE;
4035 return 0;
4038 static int inet6_fill_ifinfo(struct sk_buff *skb, struct inet6_dev *idev,
4039 u32 pid, u32 seq, int event, unsigned int flags)
4041 struct net_device *dev = idev->dev;
4042 struct ifinfomsg *hdr;
4043 struct nlmsghdr *nlh;
4044 void *protoinfo;
4046 nlh = nlmsg_put(skb, pid, seq, event, sizeof(*hdr), flags);
4047 if (nlh == NULL)
4048 return -EMSGSIZE;
4050 hdr = nlmsg_data(nlh);
4051 hdr->ifi_family = AF_INET6;
4052 hdr->__ifi_pad = 0;
4053 hdr->ifi_type = dev->type;
4054 hdr->ifi_index = dev->ifindex;
4055 hdr->ifi_flags = dev_get_flags(dev);
4056 hdr->ifi_change = 0;
4058 NLA_PUT_STRING(skb, IFLA_IFNAME, dev->name);
4060 if (dev->addr_len)
4061 NLA_PUT(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr);
4063 NLA_PUT_U32(skb, IFLA_MTU, dev->mtu);
4064 if (dev->ifindex != dev->iflink)
4065 NLA_PUT_U32(skb, IFLA_LINK, dev->iflink);
4067 protoinfo = nla_nest_start(skb, IFLA_PROTINFO);
4068 if (protoinfo == NULL)
4069 goto nla_put_failure;
4071 if (inet6_fill_ifla6_attrs(skb, idev) < 0)
4072 goto nla_put_failure;
4074 nla_nest_end(skb, protoinfo);
4075 return nlmsg_end(skb, nlh);
4077 nla_put_failure:
4078 nlmsg_cancel(skb, nlh);
4079 return -EMSGSIZE;
4082 static int inet6_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
4084 struct net *net = sock_net(skb->sk);
4085 int h, s_h;
4086 int idx = 0, s_idx;
4087 struct net_device *dev;
4088 struct inet6_dev *idev;
4089 struct hlist_head *head;
4090 struct hlist_node *node;
4092 s_h = cb->args[0];
4093 s_idx = cb->args[1];
4095 rcu_read_lock();
4096 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
4097 idx = 0;
4098 head = &net->dev_index_head[h];
4099 hlist_for_each_entry_rcu(dev, node, head, index_hlist) {
4100 if (idx < s_idx)
4101 goto cont;
4102 idev = __in6_dev_get(dev);
4103 if (!idev)
4104 goto cont;
4105 if (inet6_fill_ifinfo(skb, idev,
4106 NETLINK_CB(cb->skb).pid,
4107 cb->nlh->nlmsg_seq,
4108 RTM_NEWLINK, NLM_F_MULTI) <= 0)
4109 goto out;
4110 cont:
4111 idx++;
4114 out:
4115 rcu_read_unlock();
4116 cb->args[1] = idx;
4117 cb->args[0] = h;
4119 return skb->len;
4122 void inet6_ifinfo_notify(int event, struct inet6_dev *idev)
4124 struct sk_buff *skb;
4125 struct net *net = dev_net(idev->dev);
4126 int err = -ENOBUFS;
4128 skb = nlmsg_new(inet6_if_nlmsg_size(), GFP_ATOMIC);
4129 if (skb == NULL)
4130 goto errout;
4132 err = inet6_fill_ifinfo(skb, idev, 0, 0, event, 0);
4133 if (err < 0) {
4134 /* -EMSGSIZE implies BUG in inet6_if_nlmsg_size() */
4135 WARN_ON(err == -EMSGSIZE);
4136 kfree_skb(skb);
4137 goto errout;
4139 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFINFO, NULL, GFP_ATOMIC);
4140 return;
4141 errout:
4142 if (err < 0)
4143 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFINFO, err);
4146 static inline size_t inet6_prefix_nlmsg_size(void)
4148 return NLMSG_ALIGN(sizeof(struct prefixmsg))
4149 + nla_total_size(sizeof(struct in6_addr))
4150 + nla_total_size(sizeof(struct prefix_cacheinfo));
4153 static int inet6_fill_prefix(struct sk_buff *skb, struct inet6_dev *idev,
4154 struct prefix_info *pinfo, u32 pid, u32 seq,
4155 int event, unsigned int flags)
4157 struct prefixmsg *pmsg;
4158 struct nlmsghdr *nlh;
4159 struct prefix_cacheinfo ci;
4161 nlh = nlmsg_put(skb, pid, seq, event, sizeof(*pmsg), flags);
4162 if (nlh == NULL)
4163 return -EMSGSIZE;
4165 pmsg = nlmsg_data(nlh);
4166 pmsg->prefix_family = AF_INET6;
4167 pmsg->prefix_pad1 = 0;
4168 pmsg->prefix_pad2 = 0;
4169 pmsg->prefix_ifindex = idev->dev->ifindex;
4170 pmsg->prefix_len = pinfo->prefix_len;
4171 pmsg->prefix_type = pinfo->type;
4172 pmsg->prefix_pad3 = 0;
4173 pmsg->prefix_flags = 0;
4174 if (pinfo->onlink)
4175 pmsg->prefix_flags |= IF_PREFIX_ONLINK;
4176 if (pinfo->autoconf)
4177 pmsg->prefix_flags |= IF_PREFIX_AUTOCONF;
4179 NLA_PUT(skb, PREFIX_ADDRESS, sizeof(pinfo->prefix), &pinfo->prefix);
4181 ci.preferred_time = ntohl(pinfo->prefered);
4182 ci.valid_time = ntohl(pinfo->valid);
4183 NLA_PUT(skb, PREFIX_CACHEINFO, sizeof(ci), &ci);
4185 return nlmsg_end(skb, nlh);
4187 nla_put_failure:
4188 nlmsg_cancel(skb, nlh);
4189 return -EMSGSIZE;
4192 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
4193 struct prefix_info *pinfo)
4195 struct sk_buff *skb;
4196 struct net *net = dev_net(idev->dev);
4197 int err = -ENOBUFS;
4199 skb = nlmsg_new(inet6_prefix_nlmsg_size(), GFP_ATOMIC);
4200 if (skb == NULL)
4201 goto errout;
4203 err = inet6_fill_prefix(skb, idev, pinfo, 0, 0, event, 0);
4204 if (err < 0) {
4205 /* -EMSGSIZE implies BUG in inet6_prefix_nlmsg_size() */
4206 WARN_ON(err == -EMSGSIZE);
4207 kfree_skb(skb);
4208 goto errout;
4210 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_PREFIX, NULL, GFP_ATOMIC);
4211 return;
4212 errout:
4213 if (err < 0)
4214 rtnl_set_sk_err(net, RTNLGRP_IPV6_PREFIX, err);
4217 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
4219 inet6_ifa_notify(event ? : RTM_NEWADDR, ifp);
4221 switch (event) {
4222 case RTM_NEWADDR:
4224 * If the address was optimistic
4225 * we inserted the route at the start of
4226 * our DAD process, so we don't need
4227 * to do it again
4229 if (!(ifp->rt->rt6i_node))
4230 ip6_ins_rt(ifp->rt);
4231 if (ifp->idev->cnf.forwarding)
4232 addrconf_join_anycast(ifp);
4233 break;
4234 case RTM_DELADDR:
4235 if (ifp->idev->cnf.forwarding)
4236 addrconf_leave_anycast(ifp);
4237 addrconf_leave_solict(ifp->idev, &ifp->addr);
4238 dst_hold(&ifp->rt->dst);
4240 if (ip6_del_rt(ifp->rt))
4241 dst_free(&ifp->rt->dst);
4242 break;
4246 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
4248 rcu_read_lock_bh();
4249 if (likely(ifp->idev->dead == 0))
4250 __ipv6_ifa_notify(event, ifp);
4251 rcu_read_unlock_bh();
4254 #ifdef CONFIG_SYSCTL
4256 static
4257 int addrconf_sysctl_forward(ctl_table *ctl, int write,
4258 void __user *buffer, size_t *lenp, loff_t *ppos)
4260 int *valp = ctl->data;
4261 int val = *valp;
4262 loff_t pos = *ppos;
4263 int ret;
4265 ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
4267 if (write)
4268 ret = addrconf_fixup_forwarding(ctl, valp, val);
4269 if (ret)
4270 *ppos = pos;
4271 return ret;
4274 static void dev_disable_change(struct inet6_dev *idev)
4276 if (!idev || !idev->dev)
4277 return;
4279 if (idev->cnf.disable_ipv6)
4280 addrconf_notify(NULL, NETDEV_DOWN, idev->dev);
4281 else
4282 addrconf_notify(NULL, NETDEV_UP, idev->dev);
4285 static void addrconf_disable_change(struct net *net, __s32 newf)
4287 struct net_device *dev;
4288 struct inet6_dev *idev;
4290 rcu_read_lock();
4291 for_each_netdev_rcu(net, dev) {
4292 idev = __in6_dev_get(dev);
4293 if (idev) {
4294 int changed = (!idev->cnf.disable_ipv6) ^ (!newf);
4295 idev->cnf.disable_ipv6 = newf;
4296 if (changed)
4297 dev_disable_change(idev);
4300 rcu_read_unlock();
4303 static int addrconf_disable_ipv6(struct ctl_table *table, int *p, int old)
4305 struct net *net;
4307 net = (struct net *)table->extra2;
4309 if (p == &net->ipv6.devconf_dflt->disable_ipv6)
4310 return 0;
4312 if (!rtnl_trylock()) {
4313 /* Restore the original values before restarting */
4314 *p = old;
4315 return restart_syscall();
4318 if (p == &net->ipv6.devconf_all->disable_ipv6) {
4319 __s32 newf = net->ipv6.devconf_all->disable_ipv6;
4320 net->ipv6.devconf_dflt->disable_ipv6 = newf;
4321 addrconf_disable_change(net, newf);
4322 } else if ((!*p) ^ (!old))
4323 dev_disable_change((struct inet6_dev *)table->extra1);
4325 rtnl_unlock();
4326 return 0;
4329 static
4330 int addrconf_sysctl_disable(ctl_table *ctl, int write,
4331 void __user *buffer, size_t *lenp, loff_t *ppos)
4333 int *valp = ctl->data;
4334 int val = *valp;
4335 loff_t pos = *ppos;
4336 int ret;
4338 ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
4340 if (write)
4341 ret = addrconf_disable_ipv6(ctl, valp, val);
4342 if (ret)
4343 *ppos = pos;
4344 return ret;
4347 static struct addrconf_sysctl_table
4349 struct ctl_table_header *sysctl_header;
4350 ctl_table addrconf_vars[DEVCONF_MAX+1];
4351 char *dev_name;
4352 } addrconf_sysctl __read_mostly = {
4353 .sysctl_header = NULL,
4354 .addrconf_vars = {
4356 .procname = "forwarding",
4357 .data = &ipv6_devconf.forwarding,
4358 .maxlen = sizeof(int),
4359 .mode = 0644,
4360 .proc_handler = addrconf_sysctl_forward,
4363 .procname = "hop_limit",
4364 .data = &ipv6_devconf.hop_limit,
4365 .maxlen = sizeof(int),
4366 .mode = 0644,
4367 .proc_handler = proc_dointvec,
4370 .procname = "mtu",
4371 .data = &ipv6_devconf.mtu6,
4372 .maxlen = sizeof(int),
4373 .mode = 0644,
4374 .proc_handler = proc_dointvec,
4377 .procname = "accept_ra",
4378 .data = &ipv6_devconf.accept_ra,
4379 .maxlen = sizeof(int),
4380 .mode = 0644,
4381 .proc_handler = proc_dointvec,
4384 .procname = "accept_redirects",
4385 .data = &ipv6_devconf.accept_redirects,
4386 .maxlen = sizeof(int),
4387 .mode = 0644,
4388 .proc_handler = proc_dointvec,
4391 .procname = "autoconf",
4392 .data = &ipv6_devconf.autoconf,
4393 .maxlen = sizeof(int),
4394 .mode = 0644,
4395 .proc_handler = proc_dointvec,
4398 .procname = "dad_transmits",
4399 .data = &ipv6_devconf.dad_transmits,
4400 .maxlen = sizeof(int),
4401 .mode = 0644,
4402 .proc_handler = proc_dointvec,
4405 .procname = "router_solicitations",
4406 .data = &ipv6_devconf.rtr_solicits,
4407 .maxlen = sizeof(int),
4408 .mode = 0644,
4409 .proc_handler = proc_dointvec,
4412 .procname = "router_solicitation_interval",
4413 .data = &ipv6_devconf.rtr_solicit_interval,
4414 .maxlen = sizeof(int),
4415 .mode = 0644,
4416 .proc_handler = proc_dointvec_jiffies,
4419 .procname = "router_solicitation_delay",
4420 .data = &ipv6_devconf.rtr_solicit_delay,
4421 .maxlen = sizeof(int),
4422 .mode = 0644,
4423 .proc_handler = proc_dointvec_jiffies,
4426 .procname = "force_mld_version",
4427 .data = &ipv6_devconf.force_mld_version,
4428 .maxlen = sizeof(int),
4429 .mode = 0644,
4430 .proc_handler = proc_dointvec,
4432 #ifdef CONFIG_IPV6_PRIVACY
4434 .procname = "use_tempaddr",
4435 .data = &ipv6_devconf.use_tempaddr,
4436 .maxlen = sizeof(int),
4437 .mode = 0644,
4438 .proc_handler = proc_dointvec,
4441 .procname = "temp_valid_lft",
4442 .data = &ipv6_devconf.temp_valid_lft,
4443 .maxlen = sizeof(int),
4444 .mode = 0644,
4445 .proc_handler = proc_dointvec,
4448 .procname = "temp_prefered_lft",
4449 .data = &ipv6_devconf.temp_prefered_lft,
4450 .maxlen = sizeof(int),
4451 .mode = 0644,
4452 .proc_handler = proc_dointvec,
4455 .procname = "regen_max_retry",
4456 .data = &ipv6_devconf.regen_max_retry,
4457 .maxlen = sizeof(int),
4458 .mode = 0644,
4459 .proc_handler = proc_dointvec,
4462 .procname = "max_desync_factor",
4463 .data = &ipv6_devconf.max_desync_factor,
4464 .maxlen = sizeof(int),
4465 .mode = 0644,
4466 .proc_handler = proc_dointvec,
4468 #endif
4470 .procname = "max_addresses",
4471 .data = &ipv6_devconf.max_addresses,
4472 .maxlen = sizeof(int),
4473 .mode = 0644,
4474 .proc_handler = proc_dointvec,
4477 .procname = "accept_ra_defrtr",
4478 .data = &ipv6_devconf.accept_ra_defrtr,
4479 .maxlen = sizeof(int),
4480 .mode = 0644,
4481 .proc_handler = proc_dointvec,
4484 .procname = "accept_ra_pinfo",
4485 .data = &ipv6_devconf.accept_ra_pinfo,
4486 .maxlen = sizeof(int),
4487 .mode = 0644,
4488 .proc_handler = proc_dointvec,
4490 #ifdef CONFIG_IPV6_ROUTER_PREF
4492 .procname = "accept_ra_rtr_pref",
4493 .data = &ipv6_devconf.accept_ra_rtr_pref,
4494 .maxlen = sizeof(int),
4495 .mode = 0644,
4496 .proc_handler = proc_dointvec,
4499 .procname = "router_probe_interval",
4500 .data = &ipv6_devconf.rtr_probe_interval,
4501 .maxlen = sizeof(int),
4502 .mode = 0644,
4503 .proc_handler = proc_dointvec_jiffies,
4505 #ifdef CONFIG_IPV6_ROUTE_INFO
4507 .procname = "accept_ra_rt_info_max_plen",
4508 .data = &ipv6_devconf.accept_ra_rt_info_max_plen,
4509 .maxlen = sizeof(int),
4510 .mode = 0644,
4511 .proc_handler = proc_dointvec,
4513 #endif
4514 #endif
4516 .procname = "proxy_ndp",
4517 .data = &ipv6_devconf.proxy_ndp,
4518 .maxlen = sizeof(int),
4519 .mode = 0644,
4520 .proc_handler = proc_dointvec,
4523 .procname = "accept_source_route",
4524 .data = &ipv6_devconf.accept_source_route,
4525 .maxlen = sizeof(int),
4526 .mode = 0644,
4527 .proc_handler = proc_dointvec,
4529 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
4531 .procname = "optimistic_dad",
4532 .data = &ipv6_devconf.optimistic_dad,
4533 .maxlen = sizeof(int),
4534 .mode = 0644,
4535 .proc_handler = proc_dointvec,
4538 #endif
4539 #ifdef CONFIG_IPV6_MROUTE
4541 .procname = "mc_forwarding",
4542 .data = &ipv6_devconf.mc_forwarding,
4543 .maxlen = sizeof(int),
4544 .mode = 0444,
4545 .proc_handler = proc_dointvec,
4547 #endif
4549 .procname = "disable_ipv6",
4550 .data = &ipv6_devconf.disable_ipv6,
4551 .maxlen = sizeof(int),
4552 .mode = 0644,
4553 .proc_handler = addrconf_sysctl_disable,
4556 .procname = "accept_dad",
4557 .data = &ipv6_devconf.accept_dad,
4558 .maxlen = sizeof(int),
4559 .mode = 0644,
4560 .proc_handler = proc_dointvec,
4563 .procname = "force_tllao",
4564 .data = &ipv6_devconf.force_tllao,
4565 .maxlen = sizeof(int),
4566 .mode = 0644,
4567 .proc_handler = proc_dointvec
4570 /* sentinel */
4575 static int __addrconf_sysctl_register(struct net *net, char *dev_name,
4576 struct inet6_dev *idev, struct ipv6_devconf *p)
4578 int i;
4579 struct addrconf_sysctl_table *t;
4581 #define ADDRCONF_CTL_PATH_DEV 3
4583 struct ctl_path addrconf_ctl_path[] = {
4584 { .procname = "net", },
4585 { .procname = "ipv6", },
4586 { .procname = "conf", },
4587 { /* to be set */ },
4588 { },
4592 t = kmemdup(&addrconf_sysctl, sizeof(*t), GFP_KERNEL);
4593 if (t == NULL)
4594 goto out;
4596 for (i = 0; t->addrconf_vars[i].data; i++) {
4597 t->addrconf_vars[i].data += (char *)p - (char *)&ipv6_devconf;
4598 t->addrconf_vars[i].extra1 = idev; /* embedded; no ref */
4599 t->addrconf_vars[i].extra2 = net;
4603 * Make a copy of dev_name, because '.procname' is regarded as const
4604 * by sysctl and we wouldn't want anyone to change it under our feet
4605 * (see SIOCSIFNAME).
4607 t->dev_name = kstrdup(dev_name, GFP_KERNEL);
4608 if (!t->dev_name)
4609 goto free;
4611 addrconf_ctl_path[ADDRCONF_CTL_PATH_DEV].procname = t->dev_name;
4613 t->sysctl_header = register_net_sysctl_table(net, addrconf_ctl_path,
4614 t->addrconf_vars);
4615 if (t->sysctl_header == NULL)
4616 goto free_procname;
4618 p->sysctl = t;
4619 return 0;
4621 free_procname:
4622 kfree(t->dev_name);
4623 free:
4624 kfree(t);
4625 out:
4626 return -ENOBUFS;
4629 static void __addrconf_sysctl_unregister(struct ipv6_devconf *p)
4631 struct addrconf_sysctl_table *t;
4633 if (p->sysctl == NULL)
4634 return;
4636 t = p->sysctl;
4637 p->sysctl = NULL;
4638 unregister_net_sysctl_table(t->sysctl_header);
4639 kfree(t->dev_name);
4640 kfree(t);
4643 static void addrconf_sysctl_register(struct inet6_dev *idev)
4645 neigh_sysctl_register(idev->dev, idev->nd_parms, "ipv6",
4646 &ndisc_ifinfo_sysctl_change);
4647 __addrconf_sysctl_register(dev_net(idev->dev), idev->dev->name,
4648 idev, &idev->cnf);
4651 static void addrconf_sysctl_unregister(struct inet6_dev *idev)
4653 __addrconf_sysctl_unregister(&idev->cnf);
4654 neigh_sysctl_unregister(idev->nd_parms);
4658 #endif
4660 static int __net_init addrconf_init_net(struct net *net)
4662 int err;
4663 struct ipv6_devconf *all, *dflt;
4665 err = -ENOMEM;
4666 all = &ipv6_devconf;
4667 dflt = &ipv6_devconf_dflt;
4669 if (!net_eq(net, &init_net)) {
4670 all = kmemdup(all, sizeof(ipv6_devconf), GFP_KERNEL);
4671 if (all == NULL)
4672 goto err_alloc_all;
4674 dflt = kmemdup(dflt, sizeof(ipv6_devconf_dflt), GFP_KERNEL);
4675 if (dflt == NULL)
4676 goto err_alloc_dflt;
4677 } else {
4678 /* these will be inherited by all namespaces */
4679 dflt->autoconf = ipv6_defaults.autoconf;
4680 dflt->disable_ipv6 = ipv6_defaults.disable_ipv6;
4683 net->ipv6.devconf_all = all;
4684 net->ipv6.devconf_dflt = dflt;
4686 #ifdef CONFIG_SYSCTL
4687 err = __addrconf_sysctl_register(net, "all", NULL, all);
4688 if (err < 0)
4689 goto err_reg_all;
4691 err = __addrconf_sysctl_register(net, "default", NULL, dflt);
4692 if (err < 0)
4693 goto err_reg_dflt;
4694 #endif
4695 return 0;
4697 #ifdef CONFIG_SYSCTL
4698 err_reg_dflt:
4699 __addrconf_sysctl_unregister(all);
4700 err_reg_all:
4701 kfree(dflt);
4702 #endif
4703 err_alloc_dflt:
4704 kfree(all);
4705 err_alloc_all:
4706 return err;
4709 static void __net_exit addrconf_exit_net(struct net *net)
4711 #ifdef CONFIG_SYSCTL
4712 __addrconf_sysctl_unregister(net->ipv6.devconf_dflt);
4713 __addrconf_sysctl_unregister(net->ipv6.devconf_all);
4714 #endif
4715 if (!net_eq(net, &init_net)) {
4716 kfree(net->ipv6.devconf_dflt);
4717 kfree(net->ipv6.devconf_all);
4721 static struct pernet_operations addrconf_ops = {
4722 .init = addrconf_init_net,
4723 .exit = addrconf_exit_net,
4727 * Device notifier
4730 int register_inet6addr_notifier(struct notifier_block *nb)
4732 return atomic_notifier_chain_register(&inet6addr_chain, nb);
4734 EXPORT_SYMBOL(register_inet6addr_notifier);
4736 int unregister_inet6addr_notifier(struct notifier_block *nb)
4738 return atomic_notifier_chain_unregister(&inet6addr_chain, nb);
4740 EXPORT_SYMBOL(unregister_inet6addr_notifier);
4742 static struct rtnl_af_ops inet6_ops = {
4743 .family = AF_INET6,
4744 .fill_link_af = inet6_fill_link_af,
4745 .get_link_af_size = inet6_get_link_af_size,
4749 * Init / cleanup code
4752 int __init addrconf_init(void)
4754 int i, err;
4756 err = ipv6_addr_label_init();
4757 if (err < 0) {
4758 printk(KERN_CRIT "IPv6 Addrconf:"
4759 " cannot initialize default policy table: %d.\n", err);
4760 goto out;
4763 err = register_pernet_subsys(&addrconf_ops);
4764 if (err < 0)
4765 goto out_addrlabel;
4767 /* The addrconf netdev notifier requires that loopback_dev
4768 * has it's ipv6 private information allocated and setup
4769 * before it can bring up and give link-local addresses
4770 * to other devices which are up.
4772 * Unfortunately, loopback_dev is not necessarily the first
4773 * entry in the global dev_base list of net devices. In fact,
4774 * it is likely to be the very last entry on that list.
4775 * So this causes the notifier registry below to try and
4776 * give link-local addresses to all devices besides loopback_dev
4777 * first, then loopback_dev, which cases all the non-loopback_dev
4778 * devices to fail to get a link-local address.
4780 * So, as a temporary fix, allocate the ipv6 structure for
4781 * loopback_dev first by hand.
4782 * Longer term, all of the dependencies ipv6 has upon the loopback
4783 * device and it being up should be removed.
4785 rtnl_lock();
4786 if (!ipv6_add_dev(init_net.loopback_dev))
4787 err = -ENOMEM;
4788 rtnl_unlock();
4789 if (err)
4790 goto errlo;
4792 for (i = 0; i < IN6_ADDR_HSIZE; i++)
4793 INIT_HLIST_HEAD(&inet6_addr_lst[i]);
4795 register_netdevice_notifier(&ipv6_dev_notf);
4797 addrconf_verify(0);
4799 err = rtnl_af_register(&inet6_ops);
4800 if (err < 0)
4801 goto errout_af;
4803 err = __rtnl_register(PF_INET6, RTM_GETLINK, NULL, inet6_dump_ifinfo,
4804 NULL);
4805 if (err < 0)
4806 goto errout;
4808 /* Only the first call to __rtnl_register can fail */
4809 __rtnl_register(PF_INET6, RTM_NEWADDR, inet6_rtm_newaddr, NULL, NULL);
4810 __rtnl_register(PF_INET6, RTM_DELADDR, inet6_rtm_deladdr, NULL, NULL);
4811 __rtnl_register(PF_INET6, RTM_GETADDR, inet6_rtm_getaddr,
4812 inet6_dump_ifaddr, NULL);
4813 __rtnl_register(PF_INET6, RTM_GETMULTICAST, NULL,
4814 inet6_dump_ifmcaddr, NULL);
4815 __rtnl_register(PF_INET6, RTM_GETANYCAST, NULL,
4816 inet6_dump_ifacaddr, NULL);
4818 ipv6_addr_label_rtnl_register();
4820 return 0;
4821 errout:
4822 rtnl_af_unregister(&inet6_ops);
4823 errout_af:
4824 unregister_netdevice_notifier(&ipv6_dev_notf);
4825 errlo:
4826 unregister_pernet_subsys(&addrconf_ops);
4827 out_addrlabel:
4828 ipv6_addr_label_cleanup();
4829 out:
4830 return err;
4833 void addrconf_cleanup(void)
4835 struct net_device *dev;
4836 int i;
4838 unregister_netdevice_notifier(&ipv6_dev_notf);
4839 unregister_pernet_subsys(&addrconf_ops);
4840 ipv6_addr_label_cleanup();
4842 rtnl_lock();
4844 __rtnl_af_unregister(&inet6_ops);
4846 /* clean dev list */
4847 for_each_netdev(&init_net, dev) {
4848 if (__in6_dev_get(dev) == NULL)
4849 continue;
4850 addrconf_ifdown(dev, 1);
4852 addrconf_ifdown(init_net.loopback_dev, 2);
4855 * Check hash table.
4857 spin_lock_bh(&addrconf_hash_lock);
4858 for (i = 0; i < IN6_ADDR_HSIZE; i++)
4859 WARN_ON(!hlist_empty(&inet6_addr_lst[i]));
4860 spin_unlock_bh(&addrconf_hash_lock);
4862 del_timer(&addr_chk_timer);
4863 rtnl_unlock();