RT-AC56 3.0.0.4.374.37 core
[tomato.git] / release / src-rt-6.x.4708 / linux / linux-2.6.36 / net / ipv6 / addrconf.c
blob2089911896df273309f4070f0b3254aef88af6b6
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>
91 /* Set to 3 to get tracing... */
92 #define ACONF_DEBUG 2
94 #if ACONF_DEBUG >= 3
95 #define ADBG(x) printk x
96 #else
97 #define ADBG(x)
98 #endif
100 #define INFINITY_LIFE_TIME 0xFFFFFFFF
101 #define TIME_DELTA(a, b) ((unsigned long)((long)(a) - (long)(b)))
103 #define ADDRCONF_TIMER_FUZZ_MINUS (HZ > 50 ? HZ/50 : 1)
104 #define ADDRCONF_TIMER_FUZZ (HZ / 4)
105 #define ADDRCONF_TIMER_FUZZ_MAX (HZ)
107 #ifdef CONFIG_SYSCTL
108 static void addrconf_sysctl_register(struct inet6_dev *idev);
109 static void addrconf_sysctl_unregister(struct inet6_dev *idev);
110 #else
111 static inline void addrconf_sysctl_register(struct inet6_dev *idev)
115 static inline void addrconf_sysctl_unregister(struct inet6_dev *idev)
118 #endif
120 #ifdef CONFIG_IPV6_PRIVACY
121 static int __ipv6_regen_rndid(struct inet6_dev *idev);
122 static int __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr);
123 static void ipv6_regen_rndid(unsigned long data);
124 #endif
126 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev);
127 static int ipv6_count_addresses(struct inet6_dev *idev);
130 * Configured unicast address hash table
132 static struct hlist_head inet6_addr_lst[IN6_ADDR_HSIZE];
133 static DEFINE_SPINLOCK(addrconf_hash_lock);
135 static void addrconf_verify(unsigned long);
137 static DEFINE_TIMER(addr_chk_timer, addrconf_verify, 0, 0);
138 static DEFINE_SPINLOCK(addrconf_verify_lock);
140 static void addrconf_join_anycast(struct inet6_ifaddr *ifp);
141 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp);
143 static void addrconf_type_change(struct net_device *dev,
144 unsigned long event);
145 static int addrconf_ifdown(struct net_device *dev, int how);
147 static void addrconf_dad_start(struct inet6_ifaddr *ifp, u32 flags);
148 static void addrconf_dad_timer(unsigned long data);
149 static void addrconf_dad_completed(struct inet6_ifaddr *ifp);
150 static void addrconf_dad_run(struct inet6_dev *idev);
151 static void addrconf_rs_timer(unsigned long data);
152 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
153 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
155 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
156 struct prefix_info *pinfo);
157 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
158 struct net_device *dev);
160 static ATOMIC_NOTIFIER_HEAD(inet6addr_chain);
162 static struct ipv6_devconf ipv6_devconf __read_mostly = {
163 .forwarding = 0,
164 .hop_limit = IPV6_DEFAULT_HOPLIMIT,
165 .mtu6 = IPV6_MIN_MTU,
166 .accept_ra = 1,
167 .accept_redirects = 1,
168 .autoconf = 1,
169 .force_mld_version = 0,
170 .dad_transmits = 1,
171 .rtr_solicits = MAX_RTR_SOLICITATIONS,
172 .rtr_solicit_interval = RTR_SOLICITATION_INTERVAL,
173 .rtr_solicit_delay = MAX_RTR_SOLICITATION_DELAY,
174 #ifdef CONFIG_IPV6_PRIVACY
175 .use_tempaddr = 0,
176 .temp_valid_lft = TEMP_VALID_LIFETIME,
177 .temp_prefered_lft = TEMP_PREFERRED_LIFETIME,
178 .regen_max_retry = REGEN_MAX_RETRY,
179 .max_desync_factor = MAX_DESYNC_FACTOR,
180 #endif
181 .max_addresses = IPV6_MAX_ADDRESSES,
182 .accept_ra_defrtr = 1,
183 .accept_ra_pinfo = 1,
184 #ifdef CONFIG_IPV6_ROUTER_PREF
185 .accept_ra_rtr_pref = 1,
186 .rtr_probe_interval = 60 * HZ,
187 #ifdef CONFIG_IPV6_ROUTE_INFO
188 .accept_ra_rt_info_max_plen = 0,
189 #endif
190 #endif
191 .proxy_ndp = 0,
192 .accept_source_route = 0, /* we do not accept RH0 by default. */
193 .disable_ipv6 = 0,
194 .accept_dad = 1,
197 static struct ipv6_devconf ipv6_devconf_dflt __read_mostly = {
198 .forwarding = 0,
199 .hop_limit = IPV6_DEFAULT_HOPLIMIT,
200 .mtu6 = IPV6_MIN_MTU,
201 .accept_ra = 1,
202 .accept_redirects = 1,
203 .autoconf = 1,
204 .dad_transmits = 1,
205 .rtr_solicits = MAX_RTR_SOLICITATIONS,
206 .rtr_solicit_interval = RTR_SOLICITATION_INTERVAL,
207 .rtr_solicit_delay = MAX_RTR_SOLICITATION_DELAY,
208 #ifdef CONFIG_IPV6_PRIVACY
209 .use_tempaddr = 0,
210 .temp_valid_lft = TEMP_VALID_LIFETIME,
211 .temp_prefered_lft = TEMP_PREFERRED_LIFETIME,
212 .regen_max_retry = REGEN_MAX_RETRY,
213 .max_desync_factor = MAX_DESYNC_FACTOR,
214 #endif
215 .max_addresses = IPV6_MAX_ADDRESSES,
216 .accept_ra_defrtr = 1,
217 .accept_ra_pinfo = 1,
218 #ifdef CONFIG_IPV6_ROUTER_PREF
219 .accept_ra_rtr_pref = 1,
220 .rtr_probe_interval = 60 * HZ,
221 #ifdef CONFIG_IPV6_ROUTE_INFO
222 .accept_ra_rt_info_max_plen = 0,
223 #endif
224 #endif
225 .proxy_ndp = 0,
226 .accept_source_route = 0, /* we do not accept RH0 by default. */
227 .disable_ipv6 = 0,
228 .accept_dad = 1,
231 /* IPv6 Wildcard Address and Loopback Address defined by RFC2553 */
232 const struct in6_addr in6addr_any = IN6ADDR_ANY_INIT;
233 const struct in6_addr in6addr_loopback = IN6ADDR_LOOPBACK_INIT;
234 const struct in6_addr in6addr_linklocal_allnodes = IN6ADDR_LINKLOCAL_ALLNODES_INIT;
235 const struct in6_addr in6addr_linklocal_allrouters = IN6ADDR_LINKLOCAL_ALLROUTERS_INIT;
237 /* Check if a valid qdisc is available */
238 static inline bool addrconf_qdisc_ok(const struct net_device *dev)
240 return !qdisc_tx_is_noop(dev);
243 /* Check if a route is valid prefix route */
244 static inline int addrconf_is_prefix_route(const struct rt6_info *rt)
246 return ((rt->rt6i_flags & (RTF_GATEWAY | RTF_DEFAULT)) == 0);
249 static void addrconf_del_timer(struct inet6_ifaddr *ifp)
251 if (del_timer(&ifp->timer))
252 __in6_ifa_put(ifp);
255 enum addrconf_timer_t {
256 AC_NONE,
257 AC_DAD,
258 AC_RS,
261 static void addrconf_mod_timer(struct inet6_ifaddr *ifp,
262 enum addrconf_timer_t what,
263 unsigned long when)
265 if (!del_timer(&ifp->timer))
266 in6_ifa_hold(ifp);
268 switch (what) {
269 case AC_DAD:
270 ifp->timer.function = addrconf_dad_timer;
271 break;
272 case AC_RS:
273 ifp->timer.function = addrconf_rs_timer;
274 break;
275 default:
276 break;
278 ifp->timer.expires = jiffies + when;
279 add_timer(&ifp->timer);
282 static int snmp6_alloc_dev(struct inet6_dev *idev)
284 if (snmp_mib_init((void __percpu **)idev->stats.ipv6,
285 sizeof(struct ipstats_mib),
286 __alignof__(struct ipstats_mib)) < 0)
287 goto err_ip;
288 if (snmp_mib_init((void __percpu **)idev->stats.icmpv6,
289 sizeof(struct icmpv6_mib),
290 __alignof__(struct icmpv6_mib)) < 0)
291 goto err_icmp;
292 if (snmp_mib_init((void __percpu **)idev->stats.icmpv6msg,
293 sizeof(struct icmpv6msg_mib),
294 __alignof__(struct icmpv6msg_mib)) < 0)
295 goto err_icmpmsg;
297 return 0;
299 err_icmpmsg:
300 snmp_mib_free((void __percpu **)idev->stats.icmpv6);
301 err_icmp:
302 snmp_mib_free((void __percpu **)idev->stats.ipv6);
303 err_ip:
304 return -ENOMEM;
307 static void snmp6_free_dev(struct inet6_dev *idev)
309 snmp_mib_free((void __percpu **)idev->stats.icmpv6msg);
310 snmp_mib_free((void __percpu **)idev->stats.icmpv6);
311 snmp_mib_free((void __percpu **)idev->stats.ipv6);
314 /* Nobody refers to this device, we may destroy it. */
316 static void in6_dev_finish_destroy_rcu(struct rcu_head *head)
318 struct inet6_dev *idev = container_of(head, struct inet6_dev, rcu);
319 kfree(idev);
322 void in6_dev_finish_destroy(struct inet6_dev *idev)
324 struct net_device *dev = idev->dev;
326 WARN_ON(!list_empty(&idev->addr_list));
327 WARN_ON(idev->mc_list != NULL);
329 #ifdef NET_REFCNT_DEBUG
330 printk(KERN_DEBUG "in6_dev_finish_destroy: %s\n", dev ? dev->name : "NIL");
331 #endif
332 dev_put(dev);
333 if (!idev->dead) {
334 pr_warning("Freeing alive inet6 device %p\n", idev);
335 return;
337 snmp6_free_dev(idev);
338 call_rcu(&idev->rcu, in6_dev_finish_destroy_rcu);
341 EXPORT_SYMBOL(in6_dev_finish_destroy);
343 static struct inet6_dev * ipv6_add_dev(struct net_device *dev)
345 struct inet6_dev *ndev;
347 ASSERT_RTNL();
349 if (dev->mtu < IPV6_MIN_MTU)
350 return NULL;
352 ndev = kzalloc(sizeof(struct inet6_dev), GFP_KERNEL);
354 if (ndev == NULL)
355 return NULL;
357 rwlock_init(&ndev->lock);
358 ndev->dev = dev;
359 INIT_LIST_HEAD(&ndev->addr_list);
361 memcpy(&ndev->cnf, dev_net(dev)->ipv6.devconf_dflt, sizeof(ndev->cnf));
362 ndev->cnf.mtu6 = dev->mtu;
363 ndev->cnf.sysctl = NULL;
364 ndev->nd_parms = neigh_parms_alloc(dev, &nd_tbl);
365 if (ndev->nd_parms == NULL) {
366 kfree(ndev);
367 return NULL;
369 if (ndev->cnf.forwarding)
370 dev_disable_lro(dev);
371 /* We refer to the device */
372 dev_hold(dev);
374 if (snmp6_alloc_dev(ndev) < 0) {
375 ADBG((KERN_WARNING
376 "%s(): cannot allocate memory for statistics; dev=%s.\n",
377 __func__, dev->name));
378 neigh_parms_release(&nd_tbl, ndev->nd_parms);
379 ndev->dead = 1;
380 in6_dev_finish_destroy(ndev);
381 return NULL;
384 if (snmp6_register_dev(ndev) < 0) {
385 ADBG((KERN_WARNING
386 "%s(): cannot create /proc/net/dev_snmp6/%s\n",
387 __func__, dev->name));
388 neigh_parms_release(&nd_tbl, ndev->nd_parms);
389 ndev->dead = 1;
390 in6_dev_finish_destroy(ndev);
391 return NULL;
394 /* One reference from device. We must do this before
395 * we invoke __ipv6_regen_rndid().
397 in6_dev_hold(ndev);
399 if (dev->flags & (IFF_NOARP | IFF_LOOPBACK))
400 ndev->cnf.accept_dad = -1;
402 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
403 if (dev->type == ARPHRD_SIT && (dev->priv_flags & IFF_ISATAP)) {
404 printk(KERN_INFO
405 "%s: Disabled Multicast RS\n",
406 dev->name);
407 ndev->cnf.rtr_solicits = 0;
409 #endif
411 #ifdef CONFIG_IPV6_PRIVACY
412 INIT_LIST_HEAD(&ndev->tempaddr_list);
413 setup_timer(&ndev->regen_timer, ipv6_regen_rndid, (unsigned long)ndev);
414 if ((dev->flags&IFF_LOOPBACK) ||
415 dev->type == ARPHRD_TUNNEL ||
416 dev->type == ARPHRD_TUNNEL6 ||
417 dev->type == ARPHRD_SIT ||
418 dev->type == ARPHRD_NONE) {
419 printk(KERN_INFO
420 "%s: Disabled Privacy Extensions\n",
421 dev->name);
422 ndev->cnf.use_tempaddr = -1;
423 } else {
424 in6_dev_hold(ndev);
425 ipv6_regen_rndid((unsigned long) ndev);
427 #endif
429 if (netif_running(dev) && addrconf_qdisc_ok(dev))
430 ndev->if_flags |= IF_READY;
432 ipv6_mc_init_dev(ndev);
433 ndev->tstamp = jiffies;
434 addrconf_sysctl_register(ndev);
435 /* protected by rtnl_lock */
436 rcu_assign_pointer(dev->ip6_ptr, ndev);
438 /* Join all-node multicast group */
439 ipv6_dev_mc_inc(dev, &in6addr_linklocal_allnodes);
441 return ndev;
444 static struct inet6_dev * ipv6_find_idev(struct net_device *dev)
446 struct inet6_dev *idev;
448 ASSERT_RTNL();
450 idev = __in6_dev_get(dev);
451 if (!idev) {
452 idev = ipv6_add_dev(dev);
453 if (!idev)
454 return NULL;
457 if (dev->flags&IFF_UP)
458 ipv6_mc_up(idev);
459 return idev;
462 #ifdef CONFIG_SYSCTL
463 static void dev_forward_change(struct inet6_dev *idev)
465 struct net_device *dev;
466 struct inet6_ifaddr *ifa;
468 if (!idev)
469 return;
470 dev = idev->dev;
471 if (idev->cnf.forwarding)
472 dev_disable_lro(dev);
473 if (dev && (dev->flags & IFF_MULTICAST)) {
474 if (idev->cnf.forwarding)
475 ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
476 else
477 ipv6_dev_mc_dec(dev, &in6addr_linklocal_allrouters);
480 list_for_each_entry(ifa, &idev->addr_list, if_list) {
481 if (ifa->flags&IFA_F_TENTATIVE)
482 continue;
483 if (idev->cnf.forwarding)
484 addrconf_join_anycast(ifa);
485 else
486 addrconf_leave_anycast(ifa);
491 static void addrconf_forward_change(struct net *net, __s32 newf)
493 struct net_device *dev;
494 struct inet6_dev *idev;
496 rcu_read_lock();
497 for_each_netdev_rcu(net, dev) {
498 idev = __in6_dev_get(dev);
499 if (idev) {
500 int changed = (!idev->cnf.forwarding) ^ (!newf);
501 idev->cnf.forwarding = newf;
502 if (changed)
503 dev_forward_change(idev);
506 rcu_read_unlock();
509 static int addrconf_fixup_forwarding(struct ctl_table *table, int *p, int old)
511 struct net *net;
513 net = (struct net *)table->extra2;
514 if (p == &net->ipv6.devconf_dflt->forwarding)
515 return 0;
517 if (!rtnl_trylock()) {
518 /* Restore the original values before restarting */
519 *p = old;
520 return restart_syscall();
523 if (p == &net->ipv6.devconf_all->forwarding) {
524 __s32 newf = net->ipv6.devconf_all->forwarding;
525 net->ipv6.devconf_dflt->forwarding = newf;
526 addrconf_forward_change(net, newf);
527 } else if ((!*p) ^ (!old))
528 dev_forward_change((struct inet6_dev *)table->extra1);
529 rtnl_unlock();
531 if (*p)
532 rt6_purge_dflt_routers(net);
533 return 1;
535 #endif
537 static void inet6_ifa_finish_destroy_rcu(struct rcu_head *head)
539 struct inet6_ifaddr *ifp = container_of(head, struct inet6_ifaddr, rcu);
540 kfree(ifp);
543 /* Nobody refers to this ifaddr, destroy it */
544 void inet6_ifa_finish_destroy(struct inet6_ifaddr *ifp)
546 WARN_ON(!hlist_unhashed(&ifp->addr_lst));
548 #ifdef NET_REFCNT_DEBUG
549 printk(KERN_DEBUG "inet6_ifa_finish_destroy\n");
550 #endif
552 in6_dev_put(ifp->idev);
554 if (del_timer(&ifp->timer))
555 pr_notice("Timer is still running, when freeing ifa=%p\n", ifp);
557 if (ifp->state != INET6_IFADDR_STATE_DEAD) {
558 pr_warning("Freeing alive inet6 address %p\n", ifp);
559 return;
561 dst_release(&ifp->rt->dst);
563 call_rcu(&ifp->rcu, inet6_ifa_finish_destroy_rcu);
566 static void
567 ipv6_link_dev_addr(struct inet6_dev *idev, struct inet6_ifaddr *ifp)
569 struct list_head *p;
570 int ifp_scope = ipv6_addr_src_scope(&ifp->addr);
573 * Each device address list is sorted in order of scope -
574 * global before linklocal.
576 list_for_each(p, &idev->addr_list) {
577 struct inet6_ifaddr *ifa
578 = list_entry(p, struct inet6_ifaddr, if_list);
579 if (ifp_scope >= ipv6_addr_src_scope(&ifa->addr))
580 break;
583 list_add_tail(&ifp->if_list, p);
586 static u32 ipv6_addr_hash(const struct in6_addr *addr)
589 * We perform the hash function over the last 64 bits of the address
590 * This will include the IEEE address token on links that support it.
592 return jhash_2words((__force u32)addr->s6_addr32[2],
593 (__force u32)addr->s6_addr32[3], 0)
594 & (IN6_ADDR_HSIZE - 1);
597 /* On success it returns ifp with increased reference count */
599 static struct inet6_ifaddr *
600 ipv6_add_addr(struct inet6_dev *idev, const struct in6_addr *addr, int pfxlen,
601 int scope, u32 flags)
603 struct inet6_ifaddr *ifa = NULL;
604 struct rt6_info *rt;
605 unsigned int hash;
606 int err = 0;
607 int addr_type = ipv6_addr_type(addr);
609 if (addr_type == IPV6_ADDR_ANY ||
610 addr_type & IPV6_ADDR_MULTICAST ||
611 (!(idev->dev->flags & IFF_LOOPBACK) &&
612 addr_type & IPV6_ADDR_LOOPBACK))
613 return ERR_PTR(-EADDRNOTAVAIL);
615 rcu_read_lock_bh();
616 if (idev->dead) {
617 err = -ENODEV;
618 goto out2;
621 if (idev->cnf.disable_ipv6) {
622 err = -EACCES;
623 goto out2;
626 spin_lock(&addrconf_hash_lock);
628 /* Ignore adding duplicate addresses on an interface */
629 if (ipv6_chk_same_addr(dev_net(idev->dev), addr, idev->dev)) {
630 ADBG(("ipv6_add_addr: already assigned\n"));
631 err = -EEXIST;
632 goto out;
635 ifa = kzalloc(sizeof(struct inet6_ifaddr), GFP_ATOMIC);
637 if (ifa == NULL) {
638 ADBG(("ipv6_add_addr: malloc failed\n"));
639 err = -ENOBUFS;
640 goto out;
643 rt = addrconf_dst_alloc(idev, addr, 0);
644 if (IS_ERR(rt)) {
645 err = PTR_ERR(rt);
646 goto out;
649 ipv6_addr_copy(&ifa->addr, addr);
651 spin_lock_init(&ifa->lock);
652 spin_lock_init(&ifa->state_lock);
653 init_timer(&ifa->timer);
654 INIT_HLIST_NODE(&ifa->addr_lst);
655 ifa->timer.data = (unsigned long) ifa;
656 ifa->scope = scope;
657 ifa->prefix_len = pfxlen;
658 ifa->flags = flags | IFA_F_TENTATIVE;
659 ifa->cstamp = ifa->tstamp = jiffies;
661 ifa->rt = rt;
664 * part one of RFC 4429, section 3.3
665 * We should not configure an address as
666 * optimistic if we do not yet know the link
667 * layer address of our nexhop router
670 if (rt->rt6i_nexthop == NULL)
671 ifa->flags &= ~IFA_F_OPTIMISTIC;
673 ifa->idev = idev;
674 in6_dev_hold(idev);
675 /* For caller */
676 in6_ifa_hold(ifa);
678 /* Add to big hash table */
679 hash = ipv6_addr_hash(addr);
681 hlist_add_head_rcu(&ifa->addr_lst, &inet6_addr_lst[hash]);
682 spin_unlock(&addrconf_hash_lock);
684 write_lock(&idev->lock);
685 /* Add to inet6_dev unicast addr list. */
686 ipv6_link_dev_addr(idev, ifa);
688 #ifdef CONFIG_IPV6_PRIVACY
689 if (ifa->flags&IFA_F_TEMPORARY) {
690 list_add(&ifa->tmp_list, &idev->tempaddr_list);
691 in6_ifa_hold(ifa);
693 #endif
695 in6_ifa_hold(ifa);
696 write_unlock(&idev->lock);
697 out2:
698 rcu_read_unlock_bh();
700 if (likely(err == 0))
701 atomic_notifier_call_chain(&inet6addr_chain, NETDEV_UP, ifa);
702 else {
703 kfree(ifa);
704 ifa = ERR_PTR(err);
707 return ifa;
708 out:
709 spin_unlock(&addrconf_hash_lock);
710 goto out2;
713 /* This function wants to get referenced ifp and releases it before return */
715 static void ipv6_del_addr(struct inet6_ifaddr *ifp)
717 struct inet6_ifaddr *ifa, *ifn;
718 struct inet6_dev *idev = ifp->idev;
719 int state;
720 int hash;
721 int deleted = 0, onlink = 0;
722 unsigned long expires = jiffies;
724 hash = ipv6_addr_hash(&ifp->addr);
726 spin_lock_bh(&ifp->state_lock);
727 state = ifp->state;
728 ifp->state = INET6_IFADDR_STATE_DEAD;
729 spin_unlock_bh(&ifp->state_lock);
731 if (state == INET6_IFADDR_STATE_DEAD)
732 goto out;
734 spin_lock_bh(&addrconf_hash_lock);
735 hlist_del_init_rcu(&ifp->addr_lst);
736 spin_unlock_bh(&addrconf_hash_lock);
738 write_lock_bh(&idev->lock);
739 #ifdef CONFIG_IPV6_PRIVACY
740 if (ifp->flags&IFA_F_TEMPORARY) {
741 list_del(&ifp->tmp_list);
742 if (ifp->ifpub) {
743 in6_ifa_put(ifp->ifpub);
744 ifp->ifpub = NULL;
746 __in6_ifa_put(ifp);
748 #endif
750 list_for_each_entry_safe(ifa, ifn, &idev->addr_list, if_list) {
751 if (ifa == ifp) {
752 list_del_init(&ifp->if_list);
753 __in6_ifa_put(ifp);
755 if (!(ifp->flags & IFA_F_PERMANENT) || onlink > 0)
756 break;
757 deleted = 1;
758 continue;
759 } else if (ifp->flags & IFA_F_PERMANENT) {
760 if (ipv6_prefix_equal(&ifa->addr, &ifp->addr,
761 ifp->prefix_len)) {
762 if (ifa->flags & IFA_F_PERMANENT) {
763 onlink = 1;
764 if (deleted)
765 break;
766 } else {
767 unsigned long lifetime;
769 if (!onlink)
770 onlink = -1;
772 spin_lock(&ifa->lock);
774 lifetime = addrconf_timeout_fixup(ifa->valid_lft, HZ);
776 * Note: Because this address is
777 * not permanent, lifetime <
778 * LONG_MAX / HZ here.
780 if (time_before(expires,
781 ifa->tstamp + lifetime * HZ))
782 expires = ifa->tstamp + lifetime * HZ;
783 spin_unlock(&ifa->lock);
788 write_unlock_bh(&idev->lock);
790 addrconf_del_timer(ifp);
792 ipv6_ifa_notify(RTM_DELADDR, ifp);
794 atomic_notifier_call_chain(&inet6addr_chain, NETDEV_DOWN, ifp);
797 * Purge or update corresponding prefix
799 * 1) we don't purge prefix here if address was not permanent.
800 * prefix is managed by its own lifetime.
801 * 2) if there're no addresses, delete prefix.
802 * 3) if there're still other permanent address(es),
803 * corresponding prefix is still permanent.
804 * 4) otherwise, update prefix lifetime to the
805 * longest valid lifetime among the corresponding
806 * addresses on the device.
807 * Note: subsequent RA will update lifetime.
809 * --yoshfuji
811 if ((ifp->flags & IFA_F_PERMANENT) && onlink < 1) {
812 struct in6_addr prefix;
813 struct rt6_info *rt;
814 struct net *net = dev_net(ifp->idev->dev);
815 ipv6_addr_prefix(&prefix, &ifp->addr, ifp->prefix_len);
816 rt = rt6_lookup(net, &prefix, NULL, ifp->idev->dev->ifindex, 1);
818 if (rt && addrconf_is_prefix_route(rt)) {
819 if (onlink == 0) {
820 ip6_del_rt(rt);
821 rt = NULL;
822 } else if (!(rt->rt6i_flags & RTF_EXPIRES)) {
823 rt->rt6i_expires = expires;
824 rt->rt6i_flags |= RTF_EXPIRES;
827 dst_release(&rt->dst);
830 out:
831 in6_ifa_put(ifp);
834 #ifdef CONFIG_IPV6_PRIVACY
835 static int ipv6_create_tempaddr(struct inet6_ifaddr *ifp, struct inet6_ifaddr *ift)
837 struct inet6_dev *idev = ifp->idev;
838 struct in6_addr addr, *tmpaddr;
839 unsigned long tmp_prefered_lft, tmp_valid_lft, tmp_cstamp, tmp_tstamp;
840 unsigned long regen_advance;
841 int tmp_plen;
842 int ret = 0;
843 int max_addresses;
844 u32 addr_flags;
846 write_lock(&idev->lock);
847 if (ift) {
848 spin_lock_bh(&ift->lock);
849 memcpy(&addr.s6_addr[8], &ift->addr.s6_addr[8], 8);
850 spin_unlock_bh(&ift->lock);
851 tmpaddr = &addr;
852 } else {
853 tmpaddr = NULL;
855 retry:
856 in6_dev_hold(idev);
857 if (idev->cnf.use_tempaddr <= 0) {
858 write_unlock(&idev->lock);
859 printk(KERN_INFO
860 "ipv6_create_tempaddr(): use_tempaddr is disabled.\n");
861 in6_dev_put(idev);
862 ret = -1;
863 goto out;
865 spin_lock_bh(&ifp->lock);
866 if (ifp->regen_count++ >= idev->cnf.regen_max_retry) {
867 idev->cnf.use_tempaddr = -1;
868 spin_unlock_bh(&ifp->lock);
869 write_unlock(&idev->lock);
870 printk(KERN_WARNING
871 "ipv6_create_tempaddr(): regeneration time exceeded. disabled temporary address support.\n");
872 in6_dev_put(idev);
873 ret = -1;
874 goto out;
876 in6_ifa_hold(ifp);
877 memcpy(addr.s6_addr, ifp->addr.s6_addr, 8);
878 if (__ipv6_try_regen_rndid(idev, tmpaddr) < 0) {
879 spin_unlock_bh(&ifp->lock);
880 write_unlock(&idev->lock);
881 printk(KERN_WARNING
882 "ipv6_create_tempaddr(): regeneration of randomized interface id failed.\n");
883 in6_ifa_put(ifp);
884 in6_dev_put(idev);
885 ret = -1;
886 goto out;
888 memcpy(&addr.s6_addr[8], idev->rndid, 8);
889 tmp_valid_lft = min_t(__u32,
890 ifp->valid_lft,
891 idev->cnf.temp_valid_lft);
892 tmp_prefered_lft = min_t(__u32,
893 ifp->prefered_lft,
894 idev->cnf.temp_prefered_lft -
895 idev->cnf.max_desync_factor);
896 tmp_plen = ifp->prefix_len;
897 max_addresses = idev->cnf.max_addresses;
898 tmp_cstamp = ifp->cstamp;
899 tmp_tstamp = ifp->tstamp;
900 spin_unlock_bh(&ifp->lock);
902 regen_advance = idev->cnf.regen_max_retry *
903 idev->cnf.dad_transmits *
904 idev->nd_parms->retrans_time / HZ;
905 write_unlock(&idev->lock);
907 /* A temporary address is created only if this calculated Preferred
908 * Lifetime is greater than REGEN_ADVANCE time units. In particular,
909 * an implementation must not create a temporary address with a zero
910 * Preferred Lifetime.
912 if (tmp_prefered_lft <= regen_advance) {
913 in6_ifa_put(ifp);
914 in6_dev_put(idev);
915 ret = -1;
916 goto out;
919 addr_flags = IFA_F_TEMPORARY;
920 /* set in addrconf_prefix_rcv() */
921 if (ifp->flags & IFA_F_OPTIMISTIC)
922 addr_flags |= IFA_F_OPTIMISTIC;
924 ift = !max_addresses ||
925 ipv6_count_addresses(idev) < max_addresses ?
926 ipv6_add_addr(idev, &addr, tmp_plen,
927 ipv6_addr_type(&addr)&IPV6_ADDR_SCOPE_MASK,
928 addr_flags) : NULL;
929 if (!ift || IS_ERR(ift)) {
930 in6_ifa_put(ifp);
931 in6_dev_put(idev);
932 printk(KERN_INFO
933 "ipv6_create_tempaddr(): retry temporary address regeneration.\n");
934 tmpaddr = &addr;
935 write_lock(&idev->lock);
936 goto retry;
939 spin_lock_bh(&ift->lock);
940 ift->ifpub = ifp;
941 ift->valid_lft = tmp_valid_lft;
942 ift->prefered_lft = tmp_prefered_lft;
943 ift->cstamp = tmp_cstamp;
944 ift->tstamp = tmp_tstamp;
945 spin_unlock_bh(&ift->lock);
947 addrconf_dad_start(ift, 0);
948 in6_ifa_put(ift);
949 in6_dev_put(idev);
950 out:
951 return ret;
953 #endif
956 * Choose an appropriate source address (RFC3484)
958 enum {
959 IPV6_SADDR_RULE_INIT = 0,
960 IPV6_SADDR_RULE_LOCAL,
961 IPV6_SADDR_RULE_SCOPE,
962 IPV6_SADDR_RULE_PREFERRED,
963 #ifdef CONFIG_IPV6_MIP6
964 IPV6_SADDR_RULE_HOA,
965 #endif
966 IPV6_SADDR_RULE_OIF,
967 IPV6_SADDR_RULE_LABEL,
968 #ifdef CONFIG_IPV6_PRIVACY
969 IPV6_SADDR_RULE_PRIVACY,
970 #endif
971 IPV6_SADDR_RULE_ORCHID,
972 IPV6_SADDR_RULE_PREFIX,
973 IPV6_SADDR_RULE_MAX
976 struct ipv6_saddr_score {
977 int rule;
978 int addr_type;
979 struct inet6_ifaddr *ifa;
980 DECLARE_BITMAP(scorebits, IPV6_SADDR_RULE_MAX);
981 int scopedist;
982 int matchlen;
985 struct ipv6_saddr_dst {
986 const struct in6_addr *addr;
987 int ifindex;
988 int scope;
989 int label;
990 unsigned int prefs;
993 static inline int ipv6_saddr_preferred(int type)
995 if (type & (IPV6_ADDR_MAPPED|IPV6_ADDR_COMPATv4|IPV6_ADDR_LOOPBACK))
996 return 1;
997 return 0;
1000 static int ipv6_get_saddr_eval(struct net *net,
1001 struct ipv6_saddr_score *score,
1002 struct ipv6_saddr_dst *dst,
1003 int i)
1005 int ret;
1007 if (i <= score->rule) {
1008 switch (i) {
1009 case IPV6_SADDR_RULE_SCOPE:
1010 ret = score->scopedist;
1011 break;
1012 case IPV6_SADDR_RULE_PREFIX:
1013 ret = score->matchlen;
1014 break;
1015 default:
1016 ret = !!test_bit(i, score->scorebits);
1018 goto out;
1021 switch (i) {
1022 case IPV6_SADDR_RULE_INIT:
1023 /* Rule 0: remember if hiscore is not ready yet */
1024 ret = !!score->ifa;
1025 break;
1026 case IPV6_SADDR_RULE_LOCAL:
1027 /* Rule 1: Prefer same address */
1028 ret = ipv6_addr_equal(&score->ifa->addr, dst->addr);
1029 break;
1030 case IPV6_SADDR_RULE_SCOPE:
1031 /* Rule 2: Prefer appropriate scope
1033 * ret
1035 * -1 | d 15
1036 * ---+--+-+---> scope
1038 * | d is scope of the destination.
1039 * B-d | \
1040 * | \ <- smaller scope is better if
1041 * B-15 | \ if scope is enough for destinaion.
1042 * | ret = B - scope (-1 <= scope >= d <= 15).
1043 * d-C-1 | /
1044 * |/ <- greater is better
1045 * -C / if scope is not enough for destination.
1046 * /| ret = scope - C (-1 <= d < scope <= 15).
1048 * d - C - 1 < B -15 (for all -1 <= d <= 15).
1049 * C > d + 14 - B >= 15 + 14 - B = 29 - B.
1050 * Assume B = 0 and we get C > 29.
1052 ret = __ipv6_addr_src_scope(score->addr_type);
1053 if (ret >= dst->scope)
1054 ret = -ret;
1055 else
1056 ret -= 128; /* 30 is enough */
1057 score->scopedist = ret;
1058 break;
1059 case IPV6_SADDR_RULE_PREFERRED:
1060 /* Rule 3: Avoid deprecated and optimistic addresses */
1061 ret = ipv6_saddr_preferred(score->addr_type) ||
1062 !(score->ifa->flags & (IFA_F_DEPRECATED|IFA_F_OPTIMISTIC));
1063 break;
1064 #ifdef CONFIG_IPV6_MIP6
1065 case IPV6_SADDR_RULE_HOA:
1067 /* Rule 4: Prefer home address */
1068 int prefhome = !(dst->prefs & IPV6_PREFER_SRC_COA);
1069 ret = !(score->ifa->flags & IFA_F_HOMEADDRESS) ^ prefhome;
1070 break;
1072 #endif
1073 case IPV6_SADDR_RULE_OIF:
1074 /* Rule 5: Prefer outgoing interface */
1075 ret = (!dst->ifindex ||
1076 dst->ifindex == score->ifa->idev->dev->ifindex);
1077 break;
1078 case IPV6_SADDR_RULE_LABEL:
1079 /* Rule 6: Prefer matching label */
1080 ret = ipv6_addr_label(net,
1081 &score->ifa->addr, score->addr_type,
1082 score->ifa->idev->dev->ifindex) == dst->label;
1083 break;
1084 #ifdef CONFIG_IPV6_PRIVACY
1085 case IPV6_SADDR_RULE_PRIVACY:
1087 /* Rule 7: Prefer public address
1088 * Note: prefer temprary address if use_tempaddr >= 2
1090 int preftmp = dst->prefs & (IPV6_PREFER_SRC_PUBLIC|IPV6_PREFER_SRC_TMP) ?
1091 !!(dst->prefs & IPV6_PREFER_SRC_TMP) :
1092 score->ifa->idev->cnf.use_tempaddr >= 2;
1093 ret = (!(score->ifa->flags & IFA_F_TEMPORARY)) ^ preftmp;
1094 break;
1096 #endif
1097 case IPV6_SADDR_RULE_ORCHID:
1098 /* Rule 8-: Prefer ORCHID vs ORCHID or
1099 * non-ORCHID vs non-ORCHID
1101 ret = !(ipv6_addr_orchid(&score->ifa->addr) ^
1102 ipv6_addr_orchid(dst->addr));
1103 break;
1104 case IPV6_SADDR_RULE_PREFIX:
1105 /* Rule 8: Use longest matching prefix */
1106 score->matchlen = ret = ipv6_addr_diff(&score->ifa->addr,
1107 dst->addr);
1108 break;
1109 default:
1110 ret = 0;
1113 if (ret)
1114 __set_bit(i, score->scorebits);
1115 score->rule = i;
1116 out:
1117 return ret;
1120 int ipv6_dev_get_saddr(struct net *net, struct net_device *dst_dev,
1121 const struct in6_addr *daddr, unsigned int prefs,
1122 struct in6_addr *saddr)
1124 struct ipv6_saddr_score scores[2],
1125 *score = &scores[0], *hiscore = &scores[1];
1126 struct ipv6_saddr_dst dst;
1127 struct net_device *dev;
1128 int dst_type;
1130 dst_type = __ipv6_addr_type(daddr);
1131 dst.addr = daddr;
1132 dst.ifindex = dst_dev ? dst_dev->ifindex : 0;
1133 dst.scope = __ipv6_addr_src_scope(dst_type);
1134 dst.label = ipv6_addr_label(net, daddr, dst_type, dst.ifindex);
1135 dst.prefs = prefs;
1137 hiscore->rule = -1;
1138 hiscore->ifa = NULL;
1140 rcu_read_lock();
1142 for_each_netdev_rcu(net, dev) {
1143 struct inet6_dev *idev;
1145 /* Candidate Source Address (section 4)
1146 * - multicast and link-local destination address,
1147 * the set of candidate source address MUST only
1148 * include addresses assigned to interfaces
1149 * belonging to the same link as the outgoing
1150 * interface.
1151 * (- For site-local destination addresses, the
1152 * set of candidate source addresses MUST only
1153 * include addresses assigned to interfaces
1154 * belonging to the same site as the outgoing
1155 * interface.)
1157 if (((dst_type & IPV6_ADDR_MULTICAST) ||
1158 dst.scope <= IPV6_ADDR_SCOPE_LINKLOCAL) &&
1159 dst.ifindex && dev->ifindex != dst.ifindex)
1160 continue;
1162 idev = __in6_dev_get(dev);
1163 if (!idev)
1164 continue;
1166 read_lock_bh(&idev->lock);
1167 list_for_each_entry(score->ifa, &idev->addr_list, if_list) {
1168 int i;
1171 * - Tentative Address (RFC2462 section 5.4)
1172 * - A tentative address is not considered
1173 * "assigned to an interface" in the traditional
1174 * sense, unless it is also flagged as optimistic.
1175 * - Candidate Source Address (section 4)
1176 * - In any case, anycast addresses, multicast
1177 * addresses, and the unspecified address MUST
1178 * NOT be included in a candidate set.
1180 if ((score->ifa->flags & IFA_F_TENTATIVE) &&
1181 (!(score->ifa->flags & IFA_F_OPTIMISTIC)))
1182 continue;
1184 score->addr_type = __ipv6_addr_type(&score->ifa->addr);
1186 if (unlikely(score->addr_type == IPV6_ADDR_ANY ||
1187 score->addr_type & IPV6_ADDR_MULTICAST)) {
1188 LIMIT_NETDEBUG(KERN_DEBUG
1189 "ADDRCONF: unspecified / multicast address "
1190 "assigned as unicast address on %s",
1191 dev->name);
1192 continue;
1195 score->rule = -1;
1196 bitmap_zero(score->scorebits, IPV6_SADDR_RULE_MAX);
1198 for (i = 0; i < IPV6_SADDR_RULE_MAX; i++) {
1199 int minihiscore, miniscore;
1201 minihiscore = ipv6_get_saddr_eval(net, hiscore, &dst, i);
1202 miniscore = ipv6_get_saddr_eval(net, score, &dst, i);
1204 if (minihiscore > miniscore) {
1205 if (i == IPV6_SADDR_RULE_SCOPE &&
1206 score->scopedist > 0) {
1208 * special case:
1209 * each remaining entry
1210 * has too small (not enough)
1211 * scope, because ifa entries
1212 * are sorted by their scope
1213 * values.
1215 goto try_nextdev;
1217 break;
1218 } else if (minihiscore < miniscore) {
1219 if (hiscore->ifa)
1220 in6_ifa_put(hiscore->ifa);
1222 in6_ifa_hold(score->ifa);
1224 swap(hiscore, score);
1226 /* restore our iterator */
1227 score->ifa = hiscore->ifa;
1229 break;
1233 try_nextdev:
1234 read_unlock_bh(&idev->lock);
1236 rcu_read_unlock();
1238 if (!hiscore->ifa)
1239 return -EADDRNOTAVAIL;
1241 ipv6_addr_copy(saddr, &hiscore->ifa->addr);
1242 in6_ifa_put(hiscore->ifa);
1243 return 0;
1245 EXPORT_SYMBOL(ipv6_dev_get_saddr);
1247 int ipv6_get_lladdr(struct net_device *dev, struct in6_addr *addr,
1248 unsigned char banned_flags)
1250 struct inet6_dev *idev;
1251 int err = -EADDRNOTAVAIL;
1253 rcu_read_lock();
1254 idev = __in6_dev_get(dev);
1255 if (idev) {
1256 struct inet6_ifaddr *ifp;
1258 read_lock_bh(&idev->lock);
1259 list_for_each_entry(ifp, &idev->addr_list, if_list) {
1260 if (ifp->scope == IFA_LINK &&
1261 !(ifp->flags & banned_flags)) {
1262 ipv6_addr_copy(addr, &ifp->addr);
1263 err = 0;
1264 break;
1267 read_unlock_bh(&idev->lock);
1269 rcu_read_unlock();
1270 return err;
1273 static int ipv6_count_addresses(struct inet6_dev *idev)
1275 int cnt = 0;
1276 struct inet6_ifaddr *ifp;
1278 read_lock_bh(&idev->lock);
1279 list_for_each_entry(ifp, &idev->addr_list, if_list)
1280 cnt++;
1281 read_unlock_bh(&idev->lock);
1282 return cnt;
1285 int ipv6_chk_addr(struct net *net, struct in6_addr *addr,
1286 struct net_device *dev, int strict)
1288 struct inet6_ifaddr *ifp;
1289 struct hlist_node *node;
1290 unsigned int hash = ipv6_addr_hash(addr);
1292 rcu_read_lock_bh();
1293 hlist_for_each_entry_rcu(ifp, node, &inet6_addr_lst[hash], addr_lst) {
1294 if (!net_eq(dev_net(ifp->idev->dev), net))
1295 continue;
1296 if (ipv6_addr_equal(&ifp->addr, addr) &&
1297 !(ifp->flags&IFA_F_TENTATIVE) &&
1298 (dev == NULL || ifp->idev->dev == dev ||
1299 !(ifp->scope&(IFA_LINK|IFA_HOST) || strict))) {
1300 rcu_read_unlock_bh();
1301 return 1;
1305 rcu_read_unlock_bh();
1306 return 0;
1308 EXPORT_SYMBOL(ipv6_chk_addr);
1310 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
1311 struct net_device *dev)
1313 unsigned int hash = ipv6_addr_hash(addr);
1314 struct inet6_ifaddr *ifp;
1315 struct hlist_node *node;
1317 hlist_for_each_entry(ifp, node, &inet6_addr_lst[hash], addr_lst) {
1318 if (!net_eq(dev_net(ifp->idev->dev), net))
1319 continue;
1320 if (ipv6_addr_equal(&ifp->addr, addr)) {
1321 if (dev == NULL || ifp->idev->dev == dev)
1322 return true;
1325 return false;
1328 int ipv6_chk_prefix(struct in6_addr *addr, struct net_device *dev)
1330 struct inet6_dev *idev;
1331 struct inet6_ifaddr *ifa;
1332 int onlink;
1334 onlink = 0;
1335 rcu_read_lock();
1336 idev = __in6_dev_get(dev);
1337 if (idev) {
1338 read_lock_bh(&idev->lock);
1339 list_for_each_entry(ifa, &idev->addr_list, if_list) {
1340 onlink = ipv6_prefix_equal(addr, &ifa->addr,
1341 ifa->prefix_len);
1342 if (onlink)
1343 break;
1345 read_unlock_bh(&idev->lock);
1347 rcu_read_unlock();
1348 return onlink;
1351 EXPORT_SYMBOL(ipv6_chk_prefix);
1353 struct inet6_ifaddr *ipv6_get_ifaddr(struct net *net, const struct in6_addr *addr,
1354 struct net_device *dev, int strict)
1356 struct inet6_ifaddr *ifp, *result = NULL;
1357 unsigned int hash = ipv6_addr_hash(addr);
1358 struct hlist_node *node;
1360 rcu_read_lock_bh();
1361 hlist_for_each_entry_rcu_bh(ifp, node, &inet6_addr_lst[hash], addr_lst) {
1362 if (!net_eq(dev_net(ifp->idev->dev), net))
1363 continue;
1364 if (ipv6_addr_equal(&ifp->addr, addr)) {
1365 if (dev == NULL || ifp->idev->dev == dev ||
1366 !(ifp->scope&(IFA_LINK|IFA_HOST) || strict)) {
1367 result = ifp;
1368 in6_ifa_hold(ifp);
1369 break;
1373 rcu_read_unlock_bh();
1375 return result;
1378 /* Gets referenced address, destroys ifaddr */
1380 static void addrconf_dad_stop(struct inet6_ifaddr *ifp, int dad_failed)
1382 if (ifp->flags&IFA_F_PERMANENT) {
1383 spin_lock_bh(&ifp->lock);
1384 addrconf_del_timer(ifp);
1385 ifp->flags |= IFA_F_TENTATIVE;
1386 if (dad_failed)
1387 ifp->flags |= IFA_F_DADFAILED;
1388 spin_unlock_bh(&ifp->lock);
1389 if (dad_failed)
1390 ipv6_ifa_notify(0, ifp);
1391 in6_ifa_put(ifp);
1392 #ifdef CONFIG_IPV6_PRIVACY
1393 } else if (ifp->flags&IFA_F_TEMPORARY) {
1394 struct inet6_ifaddr *ifpub;
1395 spin_lock_bh(&ifp->lock);
1396 ifpub = ifp->ifpub;
1397 if (ifpub) {
1398 in6_ifa_hold(ifpub);
1399 spin_unlock_bh(&ifp->lock);
1400 ipv6_create_tempaddr(ifpub, ifp);
1401 in6_ifa_put(ifpub);
1402 } else {
1403 spin_unlock_bh(&ifp->lock);
1405 ipv6_del_addr(ifp);
1406 #endif
1407 } else
1408 ipv6_del_addr(ifp);
1411 static int addrconf_dad_end(struct inet6_ifaddr *ifp)
1413 int err = -ENOENT;
1415 spin_lock(&ifp->state_lock);
1416 if (ifp->state == INET6_IFADDR_STATE_DAD) {
1417 ifp->state = INET6_IFADDR_STATE_POSTDAD;
1418 err = 0;
1420 spin_unlock(&ifp->state_lock);
1422 return err;
1425 void addrconf_dad_failure(struct inet6_ifaddr *ifp)
1427 struct inet6_dev *idev = ifp->idev;
1429 if (addrconf_dad_end(ifp))
1430 return;
1432 if (net_ratelimit())
1433 printk(KERN_INFO "%s: IPv6 duplicate address %pI6c detected!\n",
1434 ifp->idev->dev->name, &ifp->addr);
1436 if (idev->cnf.accept_dad > 1 && !idev->cnf.disable_ipv6) {
1437 struct in6_addr addr;
1439 addr.s6_addr32[0] = htonl(0xfe800000);
1440 addr.s6_addr32[1] = 0;
1442 if (!ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) &&
1443 ipv6_addr_equal(&ifp->addr, &addr)) {
1444 /* DAD failed for link-local based on MAC address */
1445 idev->cnf.disable_ipv6 = 1;
1447 printk(KERN_INFO "%s: IPv6 being disabled!\n",
1448 ifp->idev->dev->name);
1452 addrconf_dad_stop(ifp, 1);
1455 /* Join to solicited addr multicast group. */
1457 void addrconf_join_solict(struct net_device *dev, struct in6_addr *addr)
1459 struct in6_addr maddr;
1461 if (dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1462 return;
1464 addrconf_addr_solict_mult(addr, &maddr);
1465 ipv6_dev_mc_inc(dev, &maddr);
1468 void addrconf_leave_solict(struct inet6_dev *idev, struct in6_addr *addr)
1470 struct in6_addr maddr;
1472 if (idev->dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1473 return;
1475 addrconf_addr_solict_mult(addr, &maddr);
1476 __ipv6_dev_mc_dec(idev, &maddr);
1479 static void addrconf_join_anycast(struct inet6_ifaddr *ifp)
1481 struct in6_addr addr;
1482 ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1483 if (ipv6_addr_any(&addr))
1484 return;
1485 ipv6_dev_ac_inc(ifp->idev->dev, &addr);
1488 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp)
1490 struct in6_addr addr;
1491 ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1492 if (ipv6_addr_any(&addr))
1493 return;
1494 __ipv6_dev_ac_dec(ifp->idev, &addr);
1497 static int addrconf_ifid_eui48(u8 *eui, struct net_device *dev)
1499 if (dev->addr_len != ETH_ALEN)
1500 return -1;
1501 memcpy(eui, dev->dev_addr, 3);
1502 memcpy(eui + 5, dev->dev_addr + 3, 3);
1505 * The zSeries OSA network cards can be shared among various
1506 * OS instances, but the OSA cards have only one MAC address.
1507 * This leads to duplicate address conflicts in conjunction
1508 * with IPv6 if more than one instance uses the same card.
1510 * The driver for these cards can deliver a unique 16-bit
1511 * identifier for each instance sharing the same card. It is
1512 * placed instead of 0xFFFE in the interface identifier. The
1513 * "u" bit of the interface identifier is not inverted in this
1514 * case. Hence the resulting interface identifier has local
1515 * scope according to RFC2373.
1517 if (dev->dev_id) {
1518 eui[3] = (dev->dev_id >> 8) & 0xFF;
1519 eui[4] = dev->dev_id & 0xFF;
1520 } else {
1521 eui[3] = 0xFF;
1522 eui[4] = 0xFE;
1523 eui[0] ^= 2;
1525 return 0;
1528 static int addrconf_ifid_arcnet(u8 *eui, struct net_device *dev)
1530 if (dev->addr_len != ARCNET_ALEN)
1531 return -1;
1532 memset(eui, 0, 7);
1533 eui[7] = *(u8*)dev->dev_addr;
1534 return 0;
1537 static int addrconf_ifid_infiniband(u8 *eui, struct net_device *dev)
1539 if (dev->addr_len != INFINIBAND_ALEN)
1540 return -1;
1541 memcpy(eui, dev->dev_addr + 12, 8);
1542 eui[0] |= 2;
1543 return 0;
1546 int __ipv6_isatap_ifid(u8 *eui, __be32 addr)
1548 if (addr == 0)
1549 return -1;
1550 eui[0] = (ipv4_is_zeronet(addr) || ipv4_is_private_10(addr) ||
1551 ipv4_is_loopback(addr) || ipv4_is_linklocal_169(addr) ||
1552 ipv4_is_private_172(addr) || ipv4_is_test_192(addr) ||
1553 ipv4_is_anycast_6to4(addr) || ipv4_is_private_192(addr) ||
1554 ipv4_is_test_198(addr) || ipv4_is_multicast(addr) ||
1555 ipv4_is_lbcast(addr)) ? 0x00 : 0x02;
1556 eui[1] = 0;
1557 eui[2] = 0x5E;
1558 eui[3] = 0xFE;
1559 memcpy(eui + 4, &addr, 4);
1560 return 0;
1562 EXPORT_SYMBOL(__ipv6_isatap_ifid);
1564 static int addrconf_ifid_sit(u8 *eui, struct net_device *dev)
1566 if (dev->priv_flags & IFF_ISATAP)
1567 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
1568 return -1;
1571 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev)
1573 switch (dev->type) {
1574 case ARPHRD_ETHER:
1575 case ARPHRD_FDDI:
1576 case ARPHRD_IEEE802_TR:
1577 return addrconf_ifid_eui48(eui, dev);
1578 case ARPHRD_ARCNET:
1579 return addrconf_ifid_arcnet(eui, dev);
1580 case ARPHRD_INFINIBAND:
1581 return addrconf_ifid_infiniband(eui, dev);
1582 case ARPHRD_SIT:
1583 return addrconf_ifid_sit(eui, dev);
1585 return -1;
1588 static int ipv6_inherit_eui64(u8 *eui, struct inet6_dev *idev)
1590 int err = -1;
1591 struct inet6_ifaddr *ifp;
1593 read_lock_bh(&idev->lock);
1594 list_for_each_entry(ifp, &idev->addr_list, if_list) {
1595 if (ifp->scope == IFA_LINK && !(ifp->flags&IFA_F_TENTATIVE)) {
1596 memcpy(eui, ifp->addr.s6_addr+8, 8);
1597 err = 0;
1598 break;
1601 read_unlock_bh(&idev->lock);
1602 return err;
1605 #ifdef CONFIG_IPV6_PRIVACY
1606 /* (re)generation of randomized interface identifier (RFC 3041 3.2, 3.5) */
1607 static int __ipv6_regen_rndid(struct inet6_dev *idev)
1609 regen:
1610 get_random_bytes(idev->rndid, sizeof(idev->rndid));
1611 idev->rndid[0] &= ~0x02;
1613 if (idev->rndid[0] == 0xfd &&
1614 (idev->rndid[1]&idev->rndid[2]&idev->rndid[3]&idev->rndid[4]&idev->rndid[5]&idev->rndid[6]) == 0xff &&
1615 (idev->rndid[7]&0x80))
1616 goto regen;
1617 if ((idev->rndid[0]|idev->rndid[1]) == 0) {
1618 if (idev->rndid[2] == 0x5e && idev->rndid[3] == 0xfe)
1619 goto regen;
1620 if ((idev->rndid[2]|idev->rndid[3]|idev->rndid[4]|idev->rndid[5]|idev->rndid[6]|idev->rndid[7]) == 0x00)
1621 goto regen;
1624 return 0;
1627 static void ipv6_regen_rndid(unsigned long data)
1629 struct inet6_dev *idev = (struct inet6_dev *) data;
1630 unsigned long expires;
1632 rcu_read_lock_bh();
1633 write_lock_bh(&idev->lock);
1635 if (idev->dead)
1636 goto out;
1638 if (__ipv6_regen_rndid(idev) < 0)
1639 goto out;
1641 expires = jiffies +
1642 idev->cnf.temp_prefered_lft * HZ -
1643 idev->cnf.regen_max_retry * idev->cnf.dad_transmits * idev->nd_parms->retrans_time -
1644 idev->cnf.max_desync_factor * HZ;
1645 if (time_before(expires, jiffies)) {
1646 printk(KERN_WARNING
1647 "ipv6_regen_rndid(): too short regeneration interval; timer disabled for %s.\n",
1648 idev->dev->name);
1649 goto out;
1652 if (!mod_timer(&idev->regen_timer, expires))
1653 in6_dev_hold(idev);
1655 out:
1656 write_unlock_bh(&idev->lock);
1657 rcu_read_unlock_bh();
1658 in6_dev_put(idev);
1661 static int __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr) {
1662 int ret = 0;
1664 if (tmpaddr && memcmp(idev->rndid, &tmpaddr->s6_addr[8], 8) == 0)
1665 ret = __ipv6_regen_rndid(idev);
1666 return ret;
1668 #endif
1671 * Add prefix route.
1674 static void
1675 addrconf_prefix_route(struct in6_addr *pfx, int plen, struct net_device *dev,
1676 unsigned long expires, u32 flags)
1678 struct fib6_config cfg = {
1679 .fc_table = RT6_TABLE_PREFIX,
1680 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1681 .fc_ifindex = dev->ifindex,
1682 .fc_expires = expires,
1683 .fc_dst_len = plen,
1684 .fc_flags = RTF_UP | flags,
1685 .fc_nlinfo.nl_net = dev_net(dev),
1686 .fc_protocol = RTPROT_KERNEL,
1689 ipv6_addr_copy(&cfg.fc_dst, pfx);
1691 /* Prevent useless cloning on PtP SIT.
1692 This thing is done here expecting that the whole
1693 class of non-broadcast devices need not cloning.
1695 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
1696 if (dev->type == ARPHRD_SIT && (dev->flags & IFF_POINTOPOINT))
1697 cfg.fc_flags |= RTF_NONEXTHOP;
1698 #endif
1700 ip6_route_add(&cfg);
1703 /* Create "default" multicast route to the interface */
1705 static void addrconf_add_mroute(struct net_device *dev)
1707 struct fib6_config cfg = {
1708 .fc_table = RT6_TABLE_LOCAL,
1709 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1710 .fc_ifindex = dev->ifindex,
1711 .fc_dst_len = 8,
1712 .fc_flags = RTF_UP,
1713 .fc_nlinfo.nl_net = dev_net(dev),
1716 ipv6_addr_set(&cfg.fc_dst, htonl(0xFF000000), 0, 0, 0);
1718 ip6_route_add(&cfg);
1721 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
1722 static void sit_route_add(struct net_device *dev)
1724 struct fib6_config cfg = {
1725 .fc_table = RT6_TABLE_MAIN,
1726 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1727 .fc_ifindex = dev->ifindex,
1728 .fc_dst_len = 96,
1729 .fc_flags = RTF_UP | RTF_NONEXTHOP,
1730 .fc_nlinfo.nl_net = dev_net(dev),
1733 /* prefix length - 96 bits "::d.d.d.d" */
1734 ip6_route_add(&cfg);
1736 #endif
1738 static void addrconf_add_lroute(struct net_device *dev)
1740 struct in6_addr addr;
1742 ipv6_addr_set(&addr, htonl(0xFE800000), 0, 0, 0);
1743 addrconf_prefix_route(&addr, 64, dev, 0, 0);
1746 static struct inet6_dev *addrconf_add_dev(struct net_device *dev)
1748 struct inet6_dev *idev;
1750 ASSERT_RTNL();
1752 idev = ipv6_find_idev(dev);
1753 if (!idev)
1754 return ERR_PTR(-ENOBUFS);
1756 if (idev->cnf.disable_ipv6)
1757 return ERR_PTR(-EACCES);
1759 /* Add default multicast route */
1760 addrconf_add_mroute(dev);
1762 /* Add link local route */
1763 addrconf_add_lroute(dev);
1764 return idev;
1767 void addrconf_prefix_rcv(struct net_device *dev, u8 *opt, int len)
1769 struct prefix_info *pinfo;
1770 __u32 valid_lft;
1771 __u32 prefered_lft;
1772 int addr_type;
1773 struct inet6_dev *in6_dev;
1774 struct net *net = dev_net(dev);
1776 pinfo = (struct prefix_info *) opt;
1778 if (len < sizeof(struct prefix_info)) {
1779 ADBG(("addrconf: prefix option too short\n"));
1780 return;
1784 * Validation checks ([ADDRCONF], page 19)
1787 addr_type = ipv6_addr_type(&pinfo->prefix);
1789 if (addr_type & (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL))
1790 return;
1792 valid_lft = ntohl(pinfo->valid);
1793 prefered_lft = ntohl(pinfo->prefered);
1795 if (prefered_lft > valid_lft) {
1796 if (net_ratelimit())
1797 printk(KERN_WARNING "addrconf: prefix option has invalid lifetime\n");
1798 return;
1801 in6_dev = in6_dev_get(dev);
1803 if (in6_dev == NULL) {
1804 if (net_ratelimit())
1805 printk(KERN_DEBUG "addrconf: device %s not configured\n", dev->name);
1806 return;
1810 * Two things going on here:
1811 * 1) Add routes for on-link prefixes
1812 * 2) Configure prefixes with the auto flag set
1815 if (pinfo->onlink) {
1816 struct rt6_info *rt;
1817 unsigned long rt_expires;
1819 /* Avoid arithmetic overflow. Really, we could
1820 * save rt_expires in seconds, likely valid_lft,
1821 * but it would require division in fib gc, that it
1822 * not good.
1824 if (HZ > USER_HZ)
1825 rt_expires = addrconf_timeout_fixup(valid_lft, HZ);
1826 else
1827 rt_expires = addrconf_timeout_fixup(valid_lft, USER_HZ);
1829 if (addrconf_finite_timeout(rt_expires))
1830 rt_expires *= HZ;
1832 rt = rt6_lookup(net, &pinfo->prefix, NULL,
1833 dev->ifindex, 1);
1834 #if 0
1835 if (rt && addrconf_is_prefix_route(rt)) {
1836 /* Autoconf prefix route */
1837 if (valid_lft == 0) {
1838 ip6_del_rt(rt);
1839 rt = NULL;
1840 } else if (addrconf_finite_timeout(rt_expires)) {
1841 /* not infinity */
1842 rt->rt6i_expires = jiffies + rt_expires;
1843 rt->rt6i_flags |= RTF_EXPIRES;
1844 } else {
1845 rt->rt6i_flags &= ~RTF_EXPIRES;
1846 rt->rt6i_expires = 0;
1848 } else if (valid_lft) {
1849 #endif
1850 if (valid_lft) {
1851 clock_t expires = 0;
1852 int flags = RTF_ADDRCONF | RTF_PREFIX_RT;
1853 if (addrconf_finite_timeout(rt_expires)) {
1854 /* not infinity */
1855 flags |= RTF_EXPIRES;
1856 expires = jiffies_to_clock_t(rt_expires);
1858 addrconf_prefix_route(&pinfo->prefix, pinfo->prefix_len,
1859 dev, expires, flags);
1861 if (rt)
1862 dst_release(&rt->dst);
1865 /* Try to figure out our local address for this prefix */
1867 if (pinfo->autoconf && in6_dev->cnf.autoconf) {
1868 struct inet6_ifaddr * ifp;
1869 struct in6_addr addr;
1870 int create = 0, update_lft = 0;
1872 if (pinfo->prefix_len == 64) {
1873 memcpy(&addr, &pinfo->prefix, 8);
1874 if (ipv6_generate_eui64(addr.s6_addr + 8, dev) &&
1875 ipv6_inherit_eui64(addr.s6_addr + 8, in6_dev)) {
1876 in6_dev_put(in6_dev);
1877 return;
1879 goto ok;
1881 if (net_ratelimit())
1882 printk(KERN_DEBUG "IPv6 addrconf: prefix with wrong length %d\n",
1883 pinfo->prefix_len);
1884 in6_dev_put(in6_dev);
1885 return;
1889 ifp = ipv6_get_ifaddr(net, &addr, dev, 1);
1891 if (ifp == NULL && valid_lft) {
1892 int max_addresses = in6_dev->cnf.max_addresses;
1893 u32 addr_flags = 0;
1895 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1896 if (in6_dev->cnf.optimistic_dad &&
1897 !net->ipv6.devconf_all->forwarding)
1898 addr_flags = IFA_F_OPTIMISTIC;
1899 #endif
1901 /* Do not allow to create too much of autoconfigured
1902 * addresses; this would be too easy way to crash kernel.
1904 if (!max_addresses ||
1905 ipv6_count_addresses(in6_dev) < max_addresses)
1906 ifp = ipv6_add_addr(in6_dev, &addr, pinfo->prefix_len,
1907 addr_type&IPV6_ADDR_SCOPE_MASK,
1908 addr_flags);
1910 if (!ifp || IS_ERR(ifp)) {
1911 in6_dev_put(in6_dev);
1912 return;
1915 update_lft = create = 1;
1916 ifp->cstamp = jiffies;
1917 addrconf_dad_start(ifp, RTF_ADDRCONF|RTF_PREFIX_RT);
1920 if (ifp) {
1921 int flags;
1922 unsigned long now;
1923 #ifdef CONFIG_IPV6_PRIVACY
1924 struct inet6_ifaddr *ift;
1925 #endif
1926 u32 stored_lft;
1928 /* update lifetime (RFC2462 5.5.3 e) */
1929 spin_lock(&ifp->lock);
1930 now = jiffies;
1931 if (ifp->valid_lft > (now - ifp->tstamp) / HZ)
1932 stored_lft = ifp->valid_lft - (now - ifp->tstamp) / HZ;
1933 else
1934 stored_lft = 0;
1935 if (!update_lft && stored_lft) {
1936 if (valid_lft > MIN_VALID_LIFETIME ||
1937 valid_lft > stored_lft)
1938 update_lft = 1;
1939 else if (stored_lft <= MIN_VALID_LIFETIME) {
1940 /* valid_lft <= stored_lft is always true */
1942 * RFC 4862 Section 5.5.3e:
1943 * "Note that the preferred lifetime of
1944 * the corresponding address is always
1945 * reset to the Preferred Lifetime in
1946 * the received Prefix Information
1947 * option, regardless of whether the
1948 * valid lifetime is also reset or
1949 * ignored."
1951 * So if the preferred lifetime in
1952 * this advertisement is different
1953 * than what we have stored, but the
1954 * valid lifetime is invalid, just
1955 * reset prefered_lft.
1957 * We must set the valid lifetime
1958 * to the stored lifetime since we'll
1959 * be updating the timestamp below,
1960 * else we'll set it back to the
1961 * minumum.
1963 if (prefered_lft != ifp->prefered_lft) {
1964 valid_lft = stored_lft;
1965 update_lft = 1;
1967 } else {
1968 valid_lft = MIN_VALID_LIFETIME;
1969 if (valid_lft < prefered_lft)
1970 prefered_lft = valid_lft;
1971 update_lft = 1;
1975 if (update_lft) {
1976 ifp->valid_lft = valid_lft;
1977 ifp->prefered_lft = prefered_lft;
1978 ifp->tstamp = now;
1979 flags = ifp->flags;
1980 ifp->flags &= ~IFA_F_DEPRECATED;
1981 spin_unlock(&ifp->lock);
1983 if (!(flags&IFA_F_TENTATIVE))
1984 ipv6_ifa_notify(0, ifp);
1985 } else
1986 spin_unlock(&ifp->lock);
1988 #ifdef CONFIG_IPV6_PRIVACY
1989 read_lock_bh(&in6_dev->lock);
1990 /* update all temporary addresses in the list */
1991 list_for_each_entry(ift, &in6_dev->tempaddr_list, tmp_list) {
1993 * When adjusting the lifetimes of an existing
1994 * temporary address, only lower the lifetimes.
1995 * Implementations must not increase the
1996 * lifetimes of an existing temporary address
1997 * when processing a Prefix Information Option.
1999 if (ifp != ift->ifpub)
2000 continue;
2002 spin_lock(&ift->lock);
2003 flags = ift->flags;
2004 if (ift->valid_lft > valid_lft &&
2005 ift->valid_lft - valid_lft > (jiffies - ift->tstamp) / HZ)
2006 ift->valid_lft = valid_lft + (jiffies - ift->tstamp) / HZ;
2007 if (ift->prefered_lft > prefered_lft &&
2008 ift->prefered_lft - prefered_lft > (jiffies - ift->tstamp) / HZ)
2009 ift->prefered_lft = prefered_lft + (jiffies - ift->tstamp) / HZ;
2010 spin_unlock(&ift->lock);
2011 if (!(flags&IFA_F_TENTATIVE))
2012 ipv6_ifa_notify(0, ift);
2015 if (create && in6_dev->cnf.use_tempaddr > 0) {
2017 * When a new public address is created as described in [ADDRCONF],
2018 * also create a new temporary address.
2020 read_unlock_bh(&in6_dev->lock);
2021 ipv6_create_tempaddr(ifp, NULL);
2022 } else {
2023 read_unlock_bh(&in6_dev->lock);
2025 #endif
2026 in6_ifa_put(ifp);
2027 addrconf_verify(0);
2030 inet6_prefix_notify(RTM_NEWPREFIX, in6_dev, pinfo);
2031 in6_dev_put(in6_dev);
2035 * Set destination address.
2036 * Special case for SIT interfaces where we create a new "virtual"
2037 * device.
2039 int addrconf_set_dstaddr(struct net *net, void __user *arg)
2041 struct in6_ifreq ireq;
2042 struct net_device *dev;
2043 int err = -EINVAL;
2045 rtnl_lock();
2047 err = -EFAULT;
2048 if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2049 goto err_exit;
2051 dev = __dev_get_by_index(net, ireq.ifr6_ifindex);
2053 err = -ENODEV;
2054 if (dev == NULL)
2055 goto err_exit;
2057 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2058 if (dev->type == ARPHRD_SIT) {
2059 const struct net_device_ops *ops = dev->netdev_ops;
2060 struct ifreq ifr;
2061 struct ip_tunnel_parm p;
2063 err = -EADDRNOTAVAIL;
2064 if (!(ipv6_addr_type(&ireq.ifr6_addr) & IPV6_ADDR_COMPATv4))
2065 goto err_exit;
2067 memset(&p, 0, sizeof(p));
2068 p.iph.daddr = ireq.ifr6_addr.s6_addr32[3];
2069 p.iph.saddr = 0;
2070 p.iph.version = 4;
2071 p.iph.ihl = 5;
2072 p.iph.protocol = IPPROTO_IPV6;
2073 p.iph.ttl = 64;
2074 ifr.ifr_ifru.ifru_data = (__force void __user *)&p;
2076 if (ops->ndo_do_ioctl) {
2077 mm_segment_t oldfs = get_fs();
2079 set_fs(KERNEL_DS);
2080 err = ops->ndo_do_ioctl(dev, &ifr, SIOCADDTUNNEL);
2081 set_fs(oldfs);
2082 } else
2083 err = -EOPNOTSUPP;
2085 if (err == 0) {
2086 err = -ENOBUFS;
2087 dev = __dev_get_by_name(net, p.name);
2088 if (!dev)
2089 goto err_exit;
2090 err = dev_open(dev);
2093 #endif
2095 err_exit:
2096 rtnl_unlock();
2097 return err;
2101 * Manual configuration of address on an interface
2103 static int inet6_addr_add(struct net *net, int ifindex, struct in6_addr *pfx,
2104 unsigned int plen, __u8 ifa_flags, __u32 prefered_lft,
2105 __u32 valid_lft)
2107 struct inet6_ifaddr *ifp;
2108 struct inet6_dev *idev;
2109 struct net_device *dev;
2110 int scope;
2111 u32 flags;
2112 clock_t expires;
2113 unsigned long timeout;
2115 ASSERT_RTNL();
2117 if (plen > 128)
2118 return -EINVAL;
2120 /* check the lifetime */
2121 if (!valid_lft || prefered_lft > valid_lft)
2122 return -EINVAL;
2124 dev = __dev_get_by_index(net, ifindex);
2125 if (!dev)
2126 return -ENODEV;
2128 idev = addrconf_add_dev(dev);
2129 if (IS_ERR(idev))
2130 return PTR_ERR(idev);
2132 scope = ipv6_addr_scope(pfx);
2134 timeout = addrconf_timeout_fixup(valid_lft, HZ);
2135 if (addrconf_finite_timeout(timeout)) {
2136 expires = jiffies_to_clock_t(timeout * HZ);
2137 valid_lft = timeout;
2138 flags = RTF_EXPIRES;
2139 } else {
2140 expires = 0;
2141 flags = 0;
2142 ifa_flags |= IFA_F_PERMANENT;
2145 timeout = addrconf_timeout_fixup(prefered_lft, HZ);
2146 if (addrconf_finite_timeout(timeout)) {
2147 if (timeout == 0)
2148 ifa_flags |= IFA_F_DEPRECATED;
2149 prefered_lft = timeout;
2152 ifp = ipv6_add_addr(idev, pfx, plen, scope, ifa_flags);
2154 if (!IS_ERR(ifp)) {
2155 spin_lock_bh(&ifp->lock);
2156 ifp->valid_lft = valid_lft;
2157 ifp->prefered_lft = prefered_lft;
2158 ifp->tstamp = jiffies;
2159 spin_unlock_bh(&ifp->lock);
2161 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, dev,
2162 expires, flags);
2164 * Note that section 3.1 of RFC 4429 indicates
2165 * that the Optimistic flag should not be set for
2166 * manually configured addresses
2168 addrconf_dad_start(ifp, 0);
2169 in6_ifa_put(ifp);
2170 addrconf_verify(0);
2171 return 0;
2174 return PTR_ERR(ifp);
2177 static int inet6_addr_del(struct net *net, int ifindex, struct in6_addr *pfx,
2178 unsigned int plen)
2180 struct inet6_ifaddr *ifp;
2181 struct inet6_dev *idev;
2182 struct net_device *dev;
2184 if (plen > 128)
2185 return -EINVAL;
2187 dev = __dev_get_by_index(net, ifindex);
2188 if (!dev)
2189 return -ENODEV;
2191 if ((idev = __in6_dev_get(dev)) == NULL)
2192 return -ENXIO;
2194 read_lock_bh(&idev->lock);
2195 list_for_each_entry(ifp, &idev->addr_list, if_list) {
2196 if (ifp->prefix_len == plen &&
2197 ipv6_addr_equal(pfx, &ifp->addr)) {
2198 in6_ifa_hold(ifp);
2199 read_unlock_bh(&idev->lock);
2201 ipv6_del_addr(ifp);
2203 /* If the last address is deleted administratively,
2204 disable IPv6 on this interface.
2206 if (list_empty(&idev->addr_list))
2207 addrconf_ifdown(idev->dev, 1);
2208 return 0;
2211 read_unlock_bh(&idev->lock);
2212 return -EADDRNOTAVAIL;
2216 int addrconf_add_ifaddr(struct net *net, void __user *arg)
2218 struct in6_ifreq ireq;
2219 int err;
2221 if (!capable(CAP_NET_ADMIN))
2222 return -EPERM;
2224 if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2225 return -EFAULT;
2227 rtnl_lock();
2228 err = inet6_addr_add(net, ireq.ifr6_ifindex, &ireq.ifr6_addr,
2229 ireq.ifr6_prefixlen, IFA_F_PERMANENT,
2230 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
2231 rtnl_unlock();
2232 return err;
2235 int addrconf_del_ifaddr(struct net *net, void __user *arg)
2237 struct in6_ifreq ireq;
2238 int err;
2240 if (!capable(CAP_NET_ADMIN))
2241 return -EPERM;
2243 if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2244 return -EFAULT;
2246 rtnl_lock();
2247 err = inet6_addr_del(net, ireq.ifr6_ifindex, &ireq.ifr6_addr,
2248 ireq.ifr6_prefixlen);
2249 rtnl_unlock();
2250 return err;
2253 static void add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
2254 int plen, int scope)
2256 struct inet6_ifaddr *ifp;
2258 ifp = ipv6_add_addr(idev, addr, plen, scope, IFA_F_PERMANENT);
2259 if (!IS_ERR(ifp)) {
2260 spin_lock_bh(&ifp->lock);
2261 ifp->flags &= ~IFA_F_TENTATIVE;
2262 spin_unlock_bh(&ifp->lock);
2263 ipv6_ifa_notify(RTM_NEWADDR, ifp);
2264 in6_ifa_put(ifp);
2268 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2269 static void sit_add_v4_addrs(struct inet6_dev *idev)
2271 struct in6_addr addr;
2272 struct net_device *dev;
2273 struct net *net = dev_net(idev->dev);
2274 int scope;
2276 ASSERT_RTNL();
2278 memset(&addr, 0, sizeof(struct in6_addr));
2279 memcpy(&addr.s6_addr32[3], idev->dev->dev_addr, 4);
2281 if (idev->dev->flags&IFF_POINTOPOINT) {
2282 addr.s6_addr32[0] = htonl(0xfe800000);
2283 scope = IFA_LINK;
2284 } else {
2285 scope = IPV6_ADDR_COMPATv4;
2288 if (addr.s6_addr32[3]) {
2289 add_addr(idev, &addr, 128, scope);
2290 return;
2293 for_each_netdev(net, dev) {
2294 struct in_device * in_dev = __in_dev_get_rtnl(dev);
2295 if (in_dev && (dev->flags & IFF_UP)) {
2296 struct in_ifaddr * ifa;
2298 int flag = scope;
2300 for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) {
2301 int plen;
2303 addr.s6_addr32[3] = ifa->ifa_local;
2305 if (ifa->ifa_scope == RT_SCOPE_LINK)
2306 continue;
2307 if (ifa->ifa_scope >= RT_SCOPE_HOST) {
2308 if (idev->dev->flags&IFF_POINTOPOINT)
2309 continue;
2310 flag |= IFA_HOST;
2312 if (idev->dev->flags&IFF_POINTOPOINT)
2313 plen = 64;
2314 else
2315 plen = 96;
2317 add_addr(idev, &addr, plen, flag);
2322 #endif
2324 static void init_loopback(struct net_device *dev)
2326 struct inet6_dev *idev;
2328 /* ::1 */
2330 ASSERT_RTNL();
2332 if ((idev = ipv6_find_idev(dev)) == NULL) {
2333 printk(KERN_DEBUG "init loopback: add_dev failed\n");
2334 return;
2337 add_addr(idev, &in6addr_loopback, 128, IFA_HOST);
2340 static void addrconf_add_linklocal(struct inet6_dev *idev, struct in6_addr *addr)
2342 struct inet6_ifaddr * ifp;
2343 u32 addr_flags = IFA_F_PERMANENT;
2345 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2346 if (idev->cnf.optimistic_dad &&
2347 !dev_net(idev->dev)->ipv6.devconf_all->forwarding)
2348 addr_flags |= IFA_F_OPTIMISTIC;
2349 #endif
2352 ifp = ipv6_add_addr(idev, addr, 64, IFA_LINK, addr_flags);
2353 if (!IS_ERR(ifp)) {
2354 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, idev->dev, 0, 0);
2355 addrconf_dad_start(ifp, 0);
2356 in6_ifa_put(ifp);
2360 static void addrconf_dev_config(struct net_device *dev)
2362 struct in6_addr addr;
2363 struct inet6_dev * idev;
2365 ASSERT_RTNL();
2367 if ((dev->type != ARPHRD_ETHER) &&
2368 (dev->type != ARPHRD_FDDI) &&
2369 (dev->type != ARPHRD_IEEE802_TR) &&
2370 (dev->type != ARPHRD_ARCNET) &&
2371 (dev->type != ARPHRD_INFINIBAND)) {
2372 /* Alas, we support only Ethernet autoconfiguration. */
2373 return;
2376 idev = addrconf_add_dev(dev);
2377 if (IS_ERR(idev))
2378 return;
2380 memset(&addr, 0, sizeof(struct in6_addr));
2381 addr.s6_addr32[0] = htonl(0xFE800000);
2383 if (ipv6_generate_eui64(addr.s6_addr + 8, dev) == 0)
2384 addrconf_add_linklocal(idev, &addr);
2387 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2388 static void addrconf_sit_config(struct net_device *dev)
2390 struct inet6_dev *idev;
2392 ASSERT_RTNL();
2395 * Configure the tunnel with one of our IPv4
2396 * addresses... we should configure all of
2397 * our v4 addrs in the tunnel
2400 if ((idev = ipv6_find_idev(dev)) == NULL) {
2401 printk(KERN_DEBUG "init sit: add_dev failed\n");
2402 return;
2405 if (dev->priv_flags & IFF_ISATAP) {
2406 struct in6_addr addr;
2408 ipv6_addr_set(&addr, htonl(0xFE800000), 0, 0, 0);
2409 addrconf_prefix_route(&addr, 64, dev, 0, 0);
2410 if (!ipv6_generate_eui64(addr.s6_addr + 8, dev))
2411 addrconf_add_linklocal(idev, &addr);
2412 return;
2415 sit_add_v4_addrs(idev);
2417 if (dev->flags&IFF_POINTOPOINT) {
2418 addrconf_add_mroute(dev);
2419 addrconf_add_lroute(dev);
2420 } else
2421 sit_route_add(dev);
2423 #endif
2425 static inline int
2426 ipv6_inherit_linklocal(struct inet6_dev *idev, struct net_device *link_dev)
2428 struct in6_addr lladdr;
2430 if (!ipv6_get_lladdr(link_dev, &lladdr, IFA_F_TENTATIVE)) {
2431 addrconf_add_linklocal(idev, &lladdr);
2432 return 0;
2434 return -1;
2437 static void ip6_tnl_add_linklocal(struct inet6_dev *idev)
2439 struct net_device *link_dev;
2440 struct net *net = dev_net(idev->dev);
2442 /* first try to inherit the link-local address from the link device */
2443 if (idev->dev->iflink &&
2444 (link_dev = __dev_get_by_index(net, idev->dev->iflink))) {
2445 if (!ipv6_inherit_linklocal(idev, link_dev))
2446 return;
2448 /* then try to inherit it from any device */
2449 for_each_netdev(net, link_dev) {
2450 if (!ipv6_inherit_linklocal(idev, link_dev))
2451 return;
2453 printk(KERN_DEBUG "init ip6-ip6: add_linklocal failed\n");
2457 * Autoconfigure tunnel with a link-local address so routing protocols,
2458 * DHCPv6, MLD etc. can be run over the virtual link
2461 static void addrconf_ip6_tnl_config(struct net_device *dev)
2463 struct inet6_dev *idev;
2465 ASSERT_RTNL();
2467 idev = addrconf_add_dev(dev);
2468 if (IS_ERR(idev)) {
2469 printk(KERN_DEBUG "init ip6-ip6: add_dev failed\n");
2470 return;
2472 ip6_tnl_add_linklocal(idev);
2475 static int addrconf_notify(struct notifier_block *this, unsigned long event,
2476 void * data)
2478 struct net_device *dev = (struct net_device *) data;
2479 struct inet6_dev *idev = __in6_dev_get(dev);
2480 int run_pending = 0;
2481 int err;
2483 switch (event) {
2484 case NETDEV_REGISTER:
2485 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2486 idev = ipv6_add_dev(dev);
2487 if (!idev)
2488 return notifier_from_errno(-ENOMEM);
2490 break;
2492 case NETDEV_UP:
2493 case NETDEV_CHANGE:
2494 if (dev->flags & IFF_SLAVE)
2495 break;
2497 if (event == NETDEV_UP) {
2498 if (!addrconf_qdisc_ok(dev)) {
2499 /* device is not ready yet. */
2500 printk(KERN_INFO
2501 "ADDRCONF(NETDEV_UP): %s: "
2502 "link is not ready\n",
2503 dev->name);
2504 break;
2507 if (!idev && dev->mtu >= IPV6_MIN_MTU)
2508 idev = ipv6_add_dev(dev);
2510 if (idev) {
2511 idev->if_flags |= IF_READY;
2512 run_pending = 1;
2514 } else {
2515 if (!addrconf_qdisc_ok(dev)) {
2516 /* device is still not ready. */
2517 break;
2520 if (idev) {
2521 if (idev->if_flags & IF_READY)
2522 /* device is already configured. */
2523 break;
2524 idev->if_flags |= IF_READY;
2527 printk(KERN_INFO
2528 "ADDRCONF(NETDEV_CHANGE): %s: "
2529 "link becomes ready\n",
2530 dev->name);
2532 run_pending = 1;
2535 switch (dev->type) {
2536 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2537 case ARPHRD_SIT:
2538 addrconf_sit_config(dev);
2539 break;
2540 #endif
2541 case ARPHRD_TUNNEL6:
2542 addrconf_ip6_tnl_config(dev);
2543 break;
2544 case ARPHRD_LOOPBACK:
2545 init_loopback(dev);
2546 break;
2548 default:
2549 addrconf_dev_config(dev);
2550 break;
2553 if (idev) {
2554 if (run_pending)
2555 addrconf_dad_run(idev);
2558 * If the MTU changed during the interface down,
2559 * when the interface up, the changed MTU must be
2560 * reflected in the idev as well as routers.
2562 if (idev->cnf.mtu6 != dev->mtu &&
2563 dev->mtu >= IPV6_MIN_MTU) {
2564 rt6_mtu_change(dev, dev->mtu);
2565 idev->cnf.mtu6 = dev->mtu;
2567 idev->tstamp = jiffies;
2568 inet6_ifinfo_notify(RTM_NEWLINK, idev);
2571 * If the changed mtu during down is lower than
2572 * IPV6_MIN_MTU stop IPv6 on this interface.
2574 if (dev->mtu < IPV6_MIN_MTU)
2575 addrconf_ifdown(dev, 1);
2577 break;
2579 case NETDEV_CHANGEMTU:
2580 if (idev && dev->mtu >= IPV6_MIN_MTU) {
2581 rt6_mtu_change(dev, dev->mtu);
2582 idev->cnf.mtu6 = dev->mtu;
2583 break;
2586 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2587 idev = ipv6_add_dev(dev);
2588 if (idev)
2589 break;
2593 * MTU falled under IPV6_MIN_MTU.
2594 * Stop IPv6 on this interface.
2597 case NETDEV_DOWN:
2598 case NETDEV_UNREGISTER:
2600 * Remove all addresses from this interface.
2602 addrconf_ifdown(dev, event != NETDEV_DOWN);
2603 break;
2605 case NETDEV_CHANGENAME:
2606 if (idev) {
2607 snmp6_unregister_dev(idev);
2608 addrconf_sysctl_unregister(idev);
2609 addrconf_sysctl_register(idev);
2610 err = snmp6_register_dev(idev);
2611 if (err)
2612 return notifier_from_errno(err);
2614 break;
2616 case NETDEV_PRE_TYPE_CHANGE:
2617 case NETDEV_POST_TYPE_CHANGE:
2618 addrconf_type_change(dev, event);
2619 break;
2622 return NOTIFY_OK;
2626 * addrconf module should be notified of a device going up
2628 static struct notifier_block ipv6_dev_notf = {
2629 .notifier_call = addrconf_notify,
2632 static void addrconf_type_change(struct net_device *dev, unsigned long event)
2634 struct inet6_dev *idev;
2635 ASSERT_RTNL();
2637 idev = __in6_dev_get(dev);
2639 if (event == NETDEV_POST_TYPE_CHANGE)
2640 ipv6_mc_remap(idev);
2641 else if (event == NETDEV_PRE_TYPE_CHANGE)
2642 ipv6_mc_unmap(idev);
2645 static int addrconf_ifdown(struct net_device *dev, int how)
2647 struct net *net = dev_net(dev);
2648 struct inet6_dev *idev;
2649 struct inet6_ifaddr *ifa;
2650 LIST_HEAD(keep_list);
2651 int state;
2653 ASSERT_RTNL();
2655 rt6_ifdown(net, dev);
2656 neigh_ifdown(&nd_tbl, dev);
2658 idev = __in6_dev_get(dev);
2659 if (idev == NULL)
2660 return -ENODEV;
2663 * Step 1: remove reference to ipv6 device from parent device.
2664 * Do not dev_put!
2666 if (how) {
2667 idev->dead = 1;
2669 /* protected by rtnl_lock */
2670 rcu_assign_pointer(dev->ip6_ptr, NULL);
2672 /* Step 1.5: remove snmp6 entry */
2673 snmp6_unregister_dev(idev);
2677 write_lock_bh(&idev->lock);
2679 /* Step 2: clear flags for stateless addrconf */
2680 if (!how)
2681 idev->if_flags &= ~(IF_RS_SENT|IF_RA_RCVD|IF_READY);
2683 #ifdef CONFIG_IPV6_PRIVACY
2684 if (how && del_timer(&idev->regen_timer))
2685 in6_dev_put(idev);
2687 /* Step 3: clear tempaddr list */
2688 while (!list_empty(&idev->tempaddr_list)) {
2689 ifa = list_first_entry(&idev->tempaddr_list,
2690 struct inet6_ifaddr, tmp_list);
2691 list_del(&ifa->tmp_list);
2692 write_unlock_bh(&idev->lock);
2693 spin_lock_bh(&ifa->lock);
2695 if (ifa->ifpub) {
2696 in6_ifa_put(ifa->ifpub);
2697 ifa->ifpub = NULL;
2699 spin_unlock_bh(&ifa->lock);
2700 in6_ifa_put(ifa);
2701 write_lock_bh(&idev->lock);
2703 #endif
2705 while (!list_empty(&idev->addr_list)) {
2706 ifa = list_first_entry(&idev->addr_list,
2707 struct inet6_ifaddr, if_list);
2708 addrconf_del_timer(ifa);
2710 /* If just doing link down, and address is permanent
2711 and not link-local, then retain it. */
2712 if (!how &&
2713 (ifa->flags&IFA_F_PERMANENT) &&
2714 !(ipv6_addr_type(&ifa->addr) & IPV6_ADDR_LINKLOCAL)) {
2715 list_move_tail(&ifa->if_list, &keep_list);
2717 /* If not doing DAD on this address, just keep it. */
2718 if ((dev->flags&(IFF_NOARP|IFF_LOOPBACK)) ||
2719 idev->cnf.accept_dad <= 0 ||
2720 (ifa->flags & IFA_F_NODAD))
2721 continue;
2723 /* If it was tentative already, no need to notify */
2724 if (ifa->flags & IFA_F_TENTATIVE)
2725 continue;
2727 /* Flag it for later restoration when link comes up */
2728 ifa->flags |= IFA_F_TENTATIVE;
2729 ifa->state = INET6_IFADDR_STATE_DAD;
2731 write_unlock_bh(&idev->lock);
2733 in6_ifa_hold(ifa);
2734 } else {
2735 list_del(&ifa->if_list);
2737 /* clear hash table */
2738 spin_lock_bh(&addrconf_hash_lock);
2739 hlist_del_init_rcu(&ifa->addr_lst);
2740 spin_unlock_bh(&addrconf_hash_lock);
2742 write_unlock_bh(&idev->lock);
2743 spin_lock_bh(&ifa->state_lock);
2744 state = ifa->state;
2745 ifa->state = INET6_IFADDR_STATE_DEAD;
2746 spin_unlock_bh(&ifa->state_lock);
2748 if (state == INET6_IFADDR_STATE_DEAD)
2749 goto put_ifa;
2752 __ipv6_ifa_notify(RTM_DELADDR, ifa);
2753 if (ifa->state == INET6_IFADDR_STATE_DEAD)
2754 atomic_notifier_call_chain(&inet6addr_chain,
2755 NETDEV_DOWN, ifa);
2757 put_ifa:
2758 in6_ifa_put(ifa);
2760 write_lock_bh(&idev->lock);
2763 list_splice(&keep_list, &idev->addr_list);
2765 write_unlock_bh(&idev->lock);
2767 /* Step 5: Discard multicast list */
2768 if (how)
2769 ipv6_mc_destroy_dev(idev);
2770 else
2771 ipv6_mc_down(idev);
2773 idev->tstamp = jiffies;
2775 /* Last: Shot the device (if unregistered) */
2776 if (how) {
2777 addrconf_sysctl_unregister(idev);
2778 neigh_parms_release(&nd_tbl, idev->nd_parms);
2779 neigh_ifdown(&nd_tbl, dev);
2780 in6_dev_put(idev);
2782 return 0;
2785 static void addrconf_rs_timer(unsigned long data)
2787 struct inet6_ifaddr *ifp = (struct inet6_ifaddr *) data;
2788 struct inet6_dev *idev = ifp->idev;
2790 read_lock(&idev->lock);
2791 if (idev->dead || !(idev->if_flags & IF_READY))
2792 goto out;
2794 if (idev->cnf.forwarding)
2795 goto out;
2797 /* Announcement received after solicitation was sent */
2798 if (idev->if_flags & IF_RA_RCVD)
2799 goto out;
2801 spin_lock(&ifp->lock);
2802 if (ifp->probes++ < idev->cnf.rtr_solicits) {
2803 /* The wait after the last probe can be shorter */
2804 addrconf_mod_timer(ifp, AC_RS,
2805 (ifp->probes == idev->cnf.rtr_solicits) ?
2806 idev->cnf.rtr_solicit_delay :
2807 idev->cnf.rtr_solicit_interval);
2808 spin_unlock(&ifp->lock);
2810 ndisc_send_rs(idev->dev, &ifp->addr, &in6addr_linklocal_allrouters);
2811 } else {
2812 spin_unlock(&ifp->lock);
2814 * Note: we do not support deprecated "all on-link"
2815 * assumption any longer.
2817 printk(KERN_DEBUG "%s: no IPv6 routers present\n",
2818 idev->dev->name);
2821 out:
2822 read_unlock(&idev->lock);
2823 in6_ifa_put(ifp);
2827 * Duplicate Address Detection
2829 static void addrconf_dad_kick(struct inet6_ifaddr *ifp)
2831 unsigned long rand_num;
2832 struct inet6_dev *idev = ifp->idev;
2834 if (ifp->flags & IFA_F_OPTIMISTIC)
2835 rand_num = 0;
2836 else
2837 rand_num = net_random() % (idev->cnf.rtr_solicit_delay ? : 1);
2839 ifp->probes = idev->cnf.dad_transmits;
2840 addrconf_mod_timer(ifp, AC_DAD, rand_num);
2843 static void addrconf_dad_start(struct inet6_ifaddr *ifp, u32 flags)
2845 struct inet6_dev *idev = ifp->idev;
2846 struct net_device *dev = idev->dev;
2848 addrconf_join_solict(dev, &ifp->addr);
2850 net_srandom(ifp->addr.s6_addr32[3]);
2852 read_lock_bh(&idev->lock);
2853 spin_lock(&ifp->lock);
2854 if (ifp->state == INET6_IFADDR_STATE_DEAD)
2855 goto out;
2857 if (dev->flags&(IFF_NOARP|IFF_LOOPBACK) ||
2858 idev->cnf.accept_dad < 1 ||
2859 !(ifp->flags&IFA_F_TENTATIVE) ||
2860 ifp->flags & IFA_F_NODAD) {
2861 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
2862 spin_unlock(&ifp->lock);
2863 read_unlock_bh(&idev->lock);
2865 addrconf_dad_completed(ifp);
2866 return;
2869 if (!(idev->if_flags & IF_READY)) {
2870 spin_unlock(&ifp->lock);
2871 read_unlock_bh(&idev->lock);
2873 * If the device is not ready:
2874 * - keep it tentative if it is a permanent address.
2875 * - otherwise, kill it.
2877 in6_ifa_hold(ifp);
2878 addrconf_dad_stop(ifp, 0);
2879 return;
2883 * Optimistic nodes can start receiving
2884 * Frames right away
2886 if (ifp->flags & IFA_F_OPTIMISTIC)
2887 ip6_ins_rt(ifp->rt);
2889 addrconf_dad_kick(ifp);
2890 out:
2891 spin_unlock(&ifp->lock);
2892 read_unlock_bh(&idev->lock);
2895 static void addrconf_dad_timer(unsigned long data)
2897 struct inet6_ifaddr *ifp = (struct inet6_ifaddr *) data;
2898 struct inet6_dev *idev = ifp->idev;
2899 struct in6_addr mcaddr;
2901 if (!ifp->probes && addrconf_dad_end(ifp))
2902 goto out;
2904 read_lock(&idev->lock);
2905 if (idev->dead || !(idev->if_flags & IF_READY)) {
2906 read_unlock(&idev->lock);
2907 goto out;
2910 spin_lock(&ifp->lock);
2911 if (ifp->state == INET6_IFADDR_STATE_DEAD) {
2912 spin_unlock(&ifp->lock);
2913 read_unlock(&idev->lock);
2914 goto out;
2917 if (ifp->probes == 0) {
2919 * DAD was successful
2922 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
2923 spin_unlock(&ifp->lock);
2924 read_unlock(&idev->lock);
2926 addrconf_dad_completed(ifp);
2928 goto out;
2931 ifp->probes--;
2932 addrconf_mod_timer(ifp, AC_DAD, ifp->idev->nd_parms->retrans_time);
2933 spin_unlock(&ifp->lock);
2934 read_unlock(&idev->lock);
2936 /* send a neighbour solicitation for our addr */
2937 addrconf_addr_solict_mult(&ifp->addr, &mcaddr);
2938 ndisc_send_ns(ifp->idev->dev, NULL, &ifp->addr, &mcaddr, &in6addr_any);
2939 out:
2940 in6_ifa_put(ifp);
2943 static void addrconf_dad_completed(struct inet6_ifaddr *ifp)
2945 struct net_device *dev = ifp->idev->dev;
2948 * Configure the address for reception. Now it is valid.
2951 ipv6_ifa_notify(RTM_NEWADDR, ifp);
2953 /* If added prefix is link local and forwarding is off,
2954 start sending router solicitations.
2957 if (ifp->idev->cnf.forwarding == 0 &&
2958 ifp->idev->cnf.rtr_solicits > 0 &&
2959 (dev->flags&IFF_LOOPBACK) == 0 &&
2960 (ipv6_addr_type(&ifp->addr) & IPV6_ADDR_LINKLOCAL)) {
2962 * If a host as already performed a random delay
2963 * [...] as part of DAD [...] there is no need
2964 * to delay again before sending the first RS
2966 ndisc_send_rs(ifp->idev->dev, &ifp->addr, &in6addr_linklocal_allrouters);
2968 spin_lock_bh(&ifp->lock);
2969 ifp->probes = 1;
2970 ifp->idev->if_flags |= IF_RS_SENT;
2971 addrconf_mod_timer(ifp, AC_RS, ifp->idev->cnf.rtr_solicit_interval);
2972 spin_unlock_bh(&ifp->lock);
2976 static void addrconf_dad_run(struct inet6_dev *idev)
2978 struct inet6_ifaddr *ifp;
2980 read_lock_bh(&idev->lock);
2981 list_for_each_entry(ifp, &idev->addr_list, if_list) {
2982 spin_lock(&ifp->lock);
2983 if (ifp->flags & IFA_F_TENTATIVE &&
2984 ifp->state == INET6_IFADDR_STATE_DAD)
2985 addrconf_dad_kick(ifp);
2986 spin_unlock(&ifp->lock);
2988 read_unlock_bh(&idev->lock);
2991 #ifdef CONFIG_PROC_FS
2992 struct if6_iter_state {
2993 struct seq_net_private p;
2994 int bucket;
2997 static struct inet6_ifaddr *if6_get_first(struct seq_file *seq)
2999 struct inet6_ifaddr *ifa = NULL;
3000 struct if6_iter_state *state = seq->private;
3001 struct net *net = seq_file_net(seq);
3003 for (state->bucket = 0; state->bucket < IN6_ADDR_HSIZE; ++state->bucket) {
3004 struct hlist_node *n;
3005 hlist_for_each_entry_rcu_bh(ifa, n, &inet6_addr_lst[state->bucket],
3006 addr_lst)
3007 if (net_eq(dev_net(ifa->idev->dev), net))
3008 return ifa;
3010 return NULL;
3013 static struct inet6_ifaddr *if6_get_next(struct seq_file *seq,
3014 struct inet6_ifaddr *ifa)
3016 struct if6_iter_state *state = seq->private;
3017 struct net *net = seq_file_net(seq);
3018 struct hlist_node *n = &ifa->addr_lst;
3020 hlist_for_each_entry_continue_rcu_bh(ifa, n, addr_lst)
3021 if (net_eq(dev_net(ifa->idev->dev), net))
3022 return ifa;
3024 while (++state->bucket < IN6_ADDR_HSIZE) {
3025 hlist_for_each_entry_rcu_bh(ifa, n,
3026 &inet6_addr_lst[state->bucket], addr_lst) {
3027 if (net_eq(dev_net(ifa->idev->dev), net))
3028 return ifa;
3032 return NULL;
3035 static struct inet6_ifaddr *if6_get_idx(struct seq_file *seq, loff_t pos)
3037 struct inet6_ifaddr *ifa = if6_get_first(seq);
3039 if (ifa)
3040 while (pos && (ifa = if6_get_next(seq, ifa)) != NULL)
3041 --pos;
3042 return pos ? NULL : ifa;
3045 static void *if6_seq_start(struct seq_file *seq, loff_t *pos)
3046 __acquires(rcu_bh)
3048 rcu_read_lock_bh();
3049 return if6_get_idx(seq, *pos);
3052 static void *if6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
3054 struct inet6_ifaddr *ifa;
3056 ifa = if6_get_next(seq, v);
3057 ++*pos;
3058 return ifa;
3061 static void if6_seq_stop(struct seq_file *seq, void *v)
3062 __releases(rcu_bh)
3064 rcu_read_unlock_bh();
3067 static int if6_seq_show(struct seq_file *seq, void *v)
3069 struct inet6_ifaddr *ifp = (struct inet6_ifaddr *)v;
3070 seq_printf(seq, "%pi6 %02x %02x %02x %02x %8s\n",
3071 &ifp->addr,
3072 ifp->idev->dev->ifindex,
3073 ifp->prefix_len,
3074 ifp->scope,
3075 ifp->flags,
3076 ifp->idev->dev->name);
3077 return 0;
3080 static const struct seq_operations if6_seq_ops = {
3081 .start = if6_seq_start,
3082 .next = if6_seq_next,
3083 .show = if6_seq_show,
3084 .stop = if6_seq_stop,
3087 static int if6_seq_open(struct inode *inode, struct file *file)
3089 return seq_open_net(inode, file, &if6_seq_ops,
3090 sizeof(struct if6_iter_state));
3093 static const struct file_operations if6_fops = {
3094 .owner = THIS_MODULE,
3095 .open = if6_seq_open,
3096 .read = seq_read,
3097 .llseek = seq_lseek,
3098 .release = seq_release_net,
3101 static int __net_init if6_proc_net_init(struct net *net)
3103 if (!proc_net_fops_create(net, "if_inet6", S_IRUGO, &if6_fops))
3104 return -ENOMEM;
3105 return 0;
3108 static void __net_exit if6_proc_net_exit(struct net *net)
3110 proc_net_remove(net, "if_inet6");
3113 static struct pernet_operations if6_proc_net_ops = {
3114 .init = if6_proc_net_init,
3115 .exit = if6_proc_net_exit,
3118 int __init if6_proc_init(void)
3120 return register_pernet_subsys(&if6_proc_net_ops);
3123 void if6_proc_exit(void)
3125 unregister_pernet_subsys(&if6_proc_net_ops);
3127 #endif /* CONFIG_PROC_FS */
3129 #if defined(CONFIG_IPV6_MIP6) || defined(CONFIG_IPV6_MIP6_MODULE)
3130 /* Check if address is a home address configured on any interface. */
3131 int ipv6_chk_home_addr(struct net *net, struct in6_addr *addr)
3133 int ret = 0;
3134 struct inet6_ifaddr *ifp = NULL;
3135 struct hlist_node *n;
3136 unsigned int hash = ipv6_addr_hash(addr);
3138 rcu_read_lock_bh();
3139 hlist_for_each_entry_rcu_bh(ifp, n, &inet6_addr_lst[hash], addr_lst) {
3140 if (!net_eq(dev_net(ifp->idev->dev), net))
3141 continue;
3142 if (ipv6_addr_equal(&ifp->addr, addr) &&
3143 (ifp->flags & IFA_F_HOMEADDRESS)) {
3144 ret = 1;
3145 break;
3148 rcu_read_unlock_bh();
3149 return ret;
3151 #endif
3154 * Periodic address status verification
3157 static void addrconf_verify(unsigned long foo)
3159 unsigned long now, next, next_sec, next_sched;
3160 struct inet6_ifaddr *ifp;
3161 struct hlist_node *node;
3162 int i;
3164 rcu_read_lock_bh();
3165 spin_lock(&addrconf_verify_lock);
3166 now = jiffies;
3167 next = round_jiffies_up(now + ADDR_CHECK_FREQUENCY);
3169 del_timer(&addr_chk_timer);
3171 for (i = 0; i < IN6_ADDR_HSIZE; i++) {
3172 restart:
3173 hlist_for_each_entry_rcu_bh(ifp, node,
3174 &inet6_addr_lst[i], addr_lst) {
3175 unsigned long age;
3177 if (ifp->flags & IFA_F_PERMANENT)
3178 continue;
3180 spin_lock(&ifp->lock);
3181 /* We try to batch several events at once. */
3182 age = (now - ifp->tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
3184 if (ifp->valid_lft != INFINITY_LIFE_TIME &&
3185 age >= ifp->valid_lft) {
3186 spin_unlock(&ifp->lock);
3187 in6_ifa_hold(ifp);
3188 ipv6_del_addr(ifp);
3189 goto restart;
3190 } else if (ifp->prefered_lft == INFINITY_LIFE_TIME) {
3191 spin_unlock(&ifp->lock);
3192 continue;
3193 } else if (age >= ifp->prefered_lft) {
3194 /* jiffies - ifp->tstamp > age >= ifp->prefered_lft */
3195 int deprecate = 0;
3197 if (!(ifp->flags&IFA_F_DEPRECATED)) {
3198 deprecate = 1;
3199 ifp->flags |= IFA_F_DEPRECATED;
3202 if (time_before(ifp->tstamp + ifp->valid_lft * HZ, next))
3203 next = ifp->tstamp + ifp->valid_lft * HZ;
3205 spin_unlock(&ifp->lock);
3207 if (deprecate) {
3208 in6_ifa_hold(ifp);
3210 ipv6_ifa_notify(0, ifp);
3211 in6_ifa_put(ifp);
3212 goto restart;
3214 #ifdef CONFIG_IPV6_PRIVACY
3215 } else if ((ifp->flags&IFA_F_TEMPORARY) &&
3216 !(ifp->flags&IFA_F_TENTATIVE)) {
3217 unsigned long regen_advance = ifp->idev->cnf.regen_max_retry *
3218 ifp->idev->cnf.dad_transmits *
3219 ifp->idev->nd_parms->retrans_time / HZ;
3221 if (age >= ifp->prefered_lft - regen_advance) {
3222 struct inet6_ifaddr *ifpub = ifp->ifpub;
3223 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3224 next = ifp->tstamp + ifp->prefered_lft * HZ;
3225 if (!ifp->regen_count && ifpub) {
3226 ifp->regen_count++;
3227 in6_ifa_hold(ifp);
3228 in6_ifa_hold(ifpub);
3229 spin_unlock(&ifp->lock);
3231 spin_lock(&ifpub->lock);
3232 ifpub->regen_count = 0;
3233 spin_unlock(&ifpub->lock);
3234 ipv6_create_tempaddr(ifpub, ifp);
3235 in6_ifa_put(ifpub);
3236 in6_ifa_put(ifp);
3237 goto restart;
3239 } else if (time_before(ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ, next))
3240 next = ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ;
3241 spin_unlock(&ifp->lock);
3242 #endif
3243 } else {
3244 /* ifp->prefered_lft <= ifp->valid_lft */
3245 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3246 next = ifp->tstamp + ifp->prefered_lft * HZ;
3247 spin_unlock(&ifp->lock);
3252 next_sec = round_jiffies_up(next);
3253 next_sched = next;
3255 /* If rounded timeout is accurate enough, accept it. */
3256 if (time_before(next_sec, next + ADDRCONF_TIMER_FUZZ))
3257 next_sched = next_sec;
3259 /* And minimum interval is ADDRCONF_TIMER_FUZZ_MAX. */
3260 if (time_before(next_sched, jiffies + ADDRCONF_TIMER_FUZZ_MAX))
3261 next_sched = jiffies + ADDRCONF_TIMER_FUZZ_MAX;
3263 ADBG((KERN_DEBUG "now = %lu, schedule = %lu, rounded schedule = %lu => %lu\n",
3264 now, next, next_sec, next_sched));
3266 addr_chk_timer.expires = next_sched;
3267 add_timer(&addr_chk_timer);
3268 spin_unlock(&addrconf_verify_lock);
3269 rcu_read_unlock_bh();
3272 static struct in6_addr *extract_addr(struct nlattr *addr, struct nlattr *local)
3274 struct in6_addr *pfx = NULL;
3276 if (addr)
3277 pfx = nla_data(addr);
3279 if (local) {
3280 if (pfx && nla_memcmp(local, pfx, sizeof(*pfx)))
3281 pfx = NULL;
3282 else
3283 pfx = nla_data(local);
3286 return pfx;
3289 static const struct nla_policy ifa_ipv6_policy[IFA_MAX+1] = {
3290 [IFA_ADDRESS] = { .len = sizeof(struct in6_addr) },
3291 [IFA_LOCAL] = { .len = sizeof(struct in6_addr) },
3292 [IFA_CACHEINFO] = { .len = sizeof(struct ifa_cacheinfo) },
3295 static int
3296 inet6_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
3298 struct net *net = sock_net(skb->sk);
3299 struct ifaddrmsg *ifm;
3300 struct nlattr *tb[IFA_MAX+1];
3301 struct in6_addr *pfx;
3302 int err;
3304 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3305 if (err < 0)
3306 return err;
3308 ifm = nlmsg_data(nlh);
3309 pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3310 if (pfx == NULL)
3311 return -EINVAL;
3313 return inet6_addr_del(net, ifm->ifa_index, pfx, ifm->ifa_prefixlen);
3316 static int inet6_addr_modify(struct inet6_ifaddr *ifp, u8 ifa_flags,
3317 u32 prefered_lft, u32 valid_lft)
3319 u32 flags;
3320 clock_t expires;
3321 unsigned long timeout;
3323 if (!valid_lft || (prefered_lft > valid_lft))
3324 return -EINVAL;
3326 timeout = addrconf_timeout_fixup(valid_lft, HZ);
3327 if (addrconf_finite_timeout(timeout)) {
3328 expires = jiffies_to_clock_t(timeout * HZ);
3329 valid_lft = timeout;
3330 flags = RTF_EXPIRES;
3331 } else {
3332 expires = 0;
3333 flags = 0;
3334 ifa_flags |= IFA_F_PERMANENT;
3337 timeout = addrconf_timeout_fixup(prefered_lft, HZ);
3338 if (addrconf_finite_timeout(timeout)) {
3339 if (timeout == 0)
3340 ifa_flags |= IFA_F_DEPRECATED;
3341 prefered_lft = timeout;
3344 spin_lock_bh(&ifp->lock);
3345 ifp->flags = (ifp->flags & ~(IFA_F_DEPRECATED | IFA_F_PERMANENT | IFA_F_NODAD | IFA_F_HOMEADDRESS)) | ifa_flags;
3346 ifp->tstamp = jiffies;
3347 ifp->valid_lft = valid_lft;
3348 ifp->prefered_lft = prefered_lft;
3350 spin_unlock_bh(&ifp->lock);
3351 if (!(ifp->flags&IFA_F_TENTATIVE))
3352 ipv6_ifa_notify(0, ifp);
3354 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, ifp->idev->dev,
3355 expires, flags);
3356 addrconf_verify(0);
3358 return 0;
3361 static int
3362 inet6_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
3364 struct net *net = sock_net(skb->sk);
3365 struct ifaddrmsg *ifm;
3366 struct nlattr *tb[IFA_MAX+1];
3367 struct in6_addr *pfx;
3368 struct inet6_ifaddr *ifa;
3369 struct net_device *dev;
3370 u32 valid_lft = INFINITY_LIFE_TIME, preferred_lft = INFINITY_LIFE_TIME;
3371 u8 ifa_flags;
3372 int err;
3374 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3375 if (err < 0)
3376 return err;
3378 ifm = nlmsg_data(nlh);
3379 pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3380 if (pfx == NULL)
3381 return -EINVAL;
3383 if (tb[IFA_CACHEINFO]) {
3384 struct ifa_cacheinfo *ci;
3386 ci = nla_data(tb[IFA_CACHEINFO]);
3387 valid_lft = ci->ifa_valid;
3388 preferred_lft = ci->ifa_prefered;
3389 } else {
3390 preferred_lft = INFINITY_LIFE_TIME;
3391 valid_lft = INFINITY_LIFE_TIME;
3394 dev = __dev_get_by_index(net, ifm->ifa_index);
3395 if (dev == NULL)
3396 return -ENODEV;
3398 /* We ignore other flags so far. */
3399 ifa_flags = ifm->ifa_flags & (IFA_F_NODAD | IFA_F_HOMEADDRESS);
3401 ifa = ipv6_get_ifaddr(net, pfx, dev, 1);
3402 if (ifa == NULL) {
3404 * It would be best to check for !NLM_F_CREATE here but
3405 * userspace alreay relies on not having to provide this.
3407 return inet6_addr_add(net, ifm->ifa_index, pfx,
3408 ifm->ifa_prefixlen, ifa_flags,
3409 preferred_lft, valid_lft);
3412 if (nlh->nlmsg_flags & NLM_F_EXCL ||
3413 !(nlh->nlmsg_flags & NLM_F_REPLACE))
3414 err = -EEXIST;
3415 else
3416 err = inet6_addr_modify(ifa, ifa_flags, preferred_lft, valid_lft);
3418 in6_ifa_put(ifa);
3420 return err;
3423 static void put_ifaddrmsg(struct nlmsghdr *nlh, u8 prefixlen, u8 flags,
3424 u8 scope, int ifindex)
3426 struct ifaddrmsg *ifm;
3428 ifm = nlmsg_data(nlh);
3429 ifm->ifa_family = AF_INET6;
3430 ifm->ifa_prefixlen = prefixlen;
3431 ifm->ifa_flags = flags;
3432 ifm->ifa_scope = scope;
3433 ifm->ifa_index = ifindex;
3436 static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
3437 unsigned long tstamp, u32 preferred, u32 valid)
3439 struct ifa_cacheinfo ci;
3441 ci.cstamp = (u32)(TIME_DELTA(cstamp, INITIAL_JIFFIES) / HZ * 100
3442 + TIME_DELTA(cstamp, INITIAL_JIFFIES) % HZ * 100 / HZ);
3443 ci.tstamp = (u32)(TIME_DELTA(tstamp, INITIAL_JIFFIES) / HZ * 100
3444 + TIME_DELTA(tstamp, INITIAL_JIFFIES) % HZ * 100 / HZ);
3445 ci.ifa_prefered = preferred;
3446 ci.ifa_valid = valid;
3448 return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
3451 static inline int rt_scope(int ifa_scope)
3453 if (ifa_scope & IFA_HOST)
3454 return RT_SCOPE_HOST;
3455 else if (ifa_scope & IFA_LINK)
3456 return RT_SCOPE_LINK;
3457 else if (ifa_scope & IFA_SITE)
3458 return RT_SCOPE_SITE;
3459 else
3460 return RT_SCOPE_UNIVERSE;
3463 static inline int inet6_ifaddr_msgsize(void)
3465 return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
3466 + nla_total_size(16) /* IFA_ADDRESS */
3467 + nla_total_size(sizeof(struct ifa_cacheinfo));
3470 static int inet6_fill_ifaddr(struct sk_buff *skb, struct inet6_ifaddr *ifa,
3471 u32 pid, u32 seq, int event, unsigned int flags)
3473 struct nlmsghdr *nlh;
3474 u32 preferred, valid;
3476 nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3477 if (nlh == NULL)
3478 return -EMSGSIZE;
3480 put_ifaddrmsg(nlh, ifa->prefix_len, ifa->flags, rt_scope(ifa->scope),
3481 ifa->idev->dev->ifindex);
3483 if (!(ifa->flags&IFA_F_PERMANENT)) {
3484 preferred = ifa->prefered_lft;
3485 valid = ifa->valid_lft;
3486 if (preferred != INFINITY_LIFE_TIME) {
3487 long tval = (jiffies - ifa->tstamp)/HZ;
3488 if (preferred > tval)
3489 preferred -= tval;
3490 else
3491 preferred = 0;
3492 if (valid != INFINITY_LIFE_TIME) {
3493 if (valid > tval)
3494 valid -= tval;
3495 else
3496 valid = 0;
3499 } else {
3500 preferred = INFINITY_LIFE_TIME;
3501 valid = INFINITY_LIFE_TIME;
3504 if (nla_put(skb, IFA_ADDRESS, 16, &ifa->addr) < 0 ||
3505 put_cacheinfo(skb, ifa->cstamp, ifa->tstamp, preferred, valid) < 0) {
3506 nlmsg_cancel(skb, nlh);
3507 return -EMSGSIZE;
3510 return nlmsg_end(skb, nlh);
3513 static int inet6_fill_ifmcaddr(struct sk_buff *skb, struct ifmcaddr6 *ifmca,
3514 u32 pid, u32 seq, int event, u16 flags)
3516 struct nlmsghdr *nlh;
3517 u8 scope = RT_SCOPE_UNIVERSE;
3518 int ifindex = ifmca->idev->dev->ifindex;
3520 if (ipv6_addr_scope(&ifmca->mca_addr) & IFA_SITE)
3521 scope = RT_SCOPE_SITE;
3523 nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3524 if (nlh == NULL)
3525 return -EMSGSIZE;
3527 put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3528 if (nla_put(skb, IFA_MULTICAST, 16, &ifmca->mca_addr) < 0 ||
3529 put_cacheinfo(skb, ifmca->mca_cstamp, ifmca->mca_tstamp,
3530 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3531 nlmsg_cancel(skb, nlh);
3532 return -EMSGSIZE;
3535 return nlmsg_end(skb, nlh);
3538 static int inet6_fill_ifacaddr(struct sk_buff *skb, struct ifacaddr6 *ifaca,
3539 u32 pid, u32 seq, int event, unsigned int flags)
3541 struct nlmsghdr *nlh;
3542 u8 scope = RT_SCOPE_UNIVERSE;
3543 int ifindex = ifaca->aca_idev->dev->ifindex;
3545 if (ipv6_addr_scope(&ifaca->aca_addr) & IFA_SITE)
3546 scope = RT_SCOPE_SITE;
3548 nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3549 if (nlh == NULL)
3550 return -EMSGSIZE;
3552 put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3553 if (nla_put(skb, IFA_ANYCAST, 16, &ifaca->aca_addr) < 0 ||
3554 put_cacheinfo(skb, ifaca->aca_cstamp, ifaca->aca_tstamp,
3555 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3556 nlmsg_cancel(skb, nlh);
3557 return -EMSGSIZE;
3560 return nlmsg_end(skb, nlh);
3563 enum addr_type_t {
3564 UNICAST_ADDR,
3565 MULTICAST_ADDR,
3566 ANYCAST_ADDR,
3569 /* called with rcu_read_lock() */
3570 static int in6_dump_addrs(struct inet6_dev *idev, struct sk_buff *skb,
3571 struct netlink_callback *cb, enum addr_type_t type,
3572 int s_ip_idx, int *p_ip_idx)
3574 struct ifmcaddr6 *ifmca;
3575 struct ifacaddr6 *ifaca;
3576 int err = 1;
3577 int ip_idx = *p_ip_idx;
3579 read_lock_bh(&idev->lock);
3580 switch (type) {
3581 case UNICAST_ADDR: {
3582 struct inet6_ifaddr *ifa;
3584 /* unicast address incl. temp addr */
3585 list_for_each_entry(ifa, &idev->addr_list, if_list) {
3586 if (++ip_idx < s_ip_idx)
3587 continue;
3588 err = inet6_fill_ifaddr(skb, ifa,
3589 NETLINK_CB(cb->skb).pid,
3590 cb->nlh->nlmsg_seq,
3591 RTM_NEWADDR,
3592 NLM_F_MULTI);
3593 if (err <= 0)
3594 break;
3596 break;
3598 case MULTICAST_ADDR:
3599 /* multicast address */
3600 for (ifmca = idev->mc_list; ifmca;
3601 ifmca = ifmca->next, ip_idx++) {
3602 if (ip_idx < s_ip_idx)
3603 continue;
3604 err = inet6_fill_ifmcaddr(skb, ifmca,
3605 NETLINK_CB(cb->skb).pid,
3606 cb->nlh->nlmsg_seq,
3607 RTM_GETMULTICAST,
3608 NLM_F_MULTI);
3609 if (err <= 0)
3610 break;
3612 break;
3613 case ANYCAST_ADDR:
3614 /* anycast address */
3615 for (ifaca = idev->ac_list; ifaca;
3616 ifaca = ifaca->aca_next, ip_idx++) {
3617 if (ip_idx < s_ip_idx)
3618 continue;
3619 err = inet6_fill_ifacaddr(skb, ifaca,
3620 NETLINK_CB(cb->skb).pid,
3621 cb->nlh->nlmsg_seq,
3622 RTM_GETANYCAST,
3623 NLM_F_MULTI);
3624 if (err <= 0)
3625 break;
3627 break;
3628 default:
3629 break;
3631 read_unlock_bh(&idev->lock);
3632 *p_ip_idx = ip_idx;
3633 return err;
3636 static int inet6_dump_addr(struct sk_buff *skb, struct netlink_callback *cb,
3637 enum addr_type_t type)
3639 struct net *net = sock_net(skb->sk);
3640 int h, s_h;
3641 int idx, ip_idx;
3642 int s_idx, s_ip_idx;
3643 struct net_device *dev;
3644 struct inet6_dev *idev;
3645 struct hlist_head *head;
3646 struct hlist_node *node;
3648 s_h = cb->args[0];
3649 s_idx = idx = cb->args[1];
3650 s_ip_idx = ip_idx = cb->args[2];
3652 rcu_read_lock();
3653 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
3654 idx = 0;
3655 head = &net->dev_index_head[h];
3656 hlist_for_each_entry_rcu(dev, node, head, index_hlist) {
3657 if (idx < s_idx)
3658 goto cont;
3659 if (h > s_h || idx > s_idx)
3660 s_ip_idx = 0;
3661 ip_idx = 0;
3662 idev = __in6_dev_get(dev);
3663 if (!idev)
3664 goto cont;
3666 if (in6_dump_addrs(idev, skb, cb, type,
3667 s_ip_idx, &ip_idx) <= 0)
3668 goto done;
3669 cont:
3670 idx++;
3673 done:
3674 rcu_read_unlock();
3675 cb->args[0] = h;
3676 cb->args[1] = idx;
3677 cb->args[2] = ip_idx;
3679 return skb->len;
3682 static int inet6_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
3684 enum addr_type_t type = UNICAST_ADDR;
3686 return inet6_dump_addr(skb, cb, type);
3689 static int inet6_dump_ifmcaddr(struct sk_buff *skb, struct netlink_callback *cb)
3691 enum addr_type_t type = MULTICAST_ADDR;
3693 return inet6_dump_addr(skb, cb, type);
3697 static int inet6_dump_ifacaddr(struct sk_buff *skb, struct netlink_callback *cb)
3699 enum addr_type_t type = ANYCAST_ADDR;
3701 return inet6_dump_addr(skb, cb, type);
3704 static int inet6_rtm_getaddr(struct sk_buff *in_skb, struct nlmsghdr* nlh,
3705 void *arg)
3707 struct net *net = sock_net(in_skb->sk);
3708 struct ifaddrmsg *ifm;
3709 struct nlattr *tb[IFA_MAX+1];
3710 struct in6_addr *addr = NULL;
3711 struct net_device *dev = NULL;
3712 struct inet6_ifaddr *ifa;
3713 struct sk_buff *skb;
3714 int err;
3716 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3717 if (err < 0)
3718 goto errout;
3720 addr = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3721 if (addr == NULL) {
3722 err = -EINVAL;
3723 goto errout;
3726 ifm = nlmsg_data(nlh);
3727 if (ifm->ifa_index)
3728 dev = __dev_get_by_index(net, ifm->ifa_index);
3730 ifa = ipv6_get_ifaddr(net, addr, dev, 1);
3731 if (!ifa) {
3732 err = -EADDRNOTAVAIL;
3733 goto errout;
3736 skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_KERNEL);
3737 if (!skb) {
3738 err = -ENOBUFS;
3739 goto errout_ifa;
3742 err = inet6_fill_ifaddr(skb, ifa, NETLINK_CB(in_skb).pid,
3743 nlh->nlmsg_seq, RTM_NEWADDR, 0);
3744 if (err < 0) {
3745 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
3746 WARN_ON(err == -EMSGSIZE);
3747 kfree_skb(skb);
3748 goto errout_ifa;
3750 err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).pid);
3751 errout_ifa:
3752 in6_ifa_put(ifa);
3753 errout:
3754 return err;
3757 static void inet6_ifa_notify(int event, struct inet6_ifaddr *ifa)
3759 struct sk_buff *skb;
3760 struct net *net = dev_net(ifa->idev->dev);
3761 int err = -ENOBUFS;
3763 skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_ATOMIC);
3764 if (skb == NULL)
3765 goto errout;
3767 err = inet6_fill_ifaddr(skb, ifa, 0, 0, event, 0);
3768 if (err < 0) {
3769 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
3770 WARN_ON(err == -EMSGSIZE);
3771 kfree_skb(skb);
3772 goto errout;
3774 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
3775 return;
3776 errout:
3777 if (err < 0)
3778 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
3781 static inline void ipv6_store_devconf(struct ipv6_devconf *cnf,
3782 __s32 *array, int bytes)
3784 BUG_ON(bytes < (DEVCONF_MAX * 4));
3786 memset(array, 0, bytes);
3787 array[DEVCONF_FORWARDING] = cnf->forwarding;
3788 array[DEVCONF_HOPLIMIT] = cnf->hop_limit;
3789 array[DEVCONF_MTU6] = cnf->mtu6;
3790 array[DEVCONF_ACCEPT_RA] = cnf->accept_ra;
3791 array[DEVCONF_ACCEPT_REDIRECTS] = cnf->accept_redirects;
3792 array[DEVCONF_AUTOCONF] = cnf->autoconf;
3793 array[DEVCONF_DAD_TRANSMITS] = cnf->dad_transmits;
3794 array[DEVCONF_RTR_SOLICITS] = cnf->rtr_solicits;
3795 array[DEVCONF_RTR_SOLICIT_INTERVAL] = cnf->rtr_solicit_interval;
3796 array[DEVCONF_RTR_SOLICIT_DELAY] = cnf->rtr_solicit_delay;
3797 array[DEVCONF_FORCE_MLD_VERSION] = cnf->force_mld_version;
3798 #ifdef CONFIG_IPV6_PRIVACY
3799 array[DEVCONF_USE_TEMPADDR] = cnf->use_tempaddr;
3800 array[DEVCONF_TEMP_VALID_LFT] = cnf->temp_valid_lft;
3801 array[DEVCONF_TEMP_PREFERED_LFT] = cnf->temp_prefered_lft;
3802 array[DEVCONF_REGEN_MAX_RETRY] = cnf->regen_max_retry;
3803 array[DEVCONF_MAX_DESYNC_FACTOR] = cnf->max_desync_factor;
3804 #endif
3805 array[DEVCONF_MAX_ADDRESSES] = cnf->max_addresses;
3806 array[DEVCONF_ACCEPT_RA_DEFRTR] = cnf->accept_ra_defrtr;
3807 array[DEVCONF_ACCEPT_RA_PINFO] = cnf->accept_ra_pinfo;
3808 #ifdef CONFIG_IPV6_ROUTER_PREF
3809 array[DEVCONF_ACCEPT_RA_RTR_PREF] = cnf->accept_ra_rtr_pref;
3810 array[DEVCONF_RTR_PROBE_INTERVAL] = cnf->rtr_probe_interval;
3811 #ifdef CONFIG_IPV6_ROUTE_INFO
3812 array[DEVCONF_ACCEPT_RA_RT_INFO_MAX_PLEN] = cnf->accept_ra_rt_info_max_plen;
3813 #endif
3814 #endif
3815 array[DEVCONF_PROXY_NDP] = cnf->proxy_ndp;
3816 array[DEVCONF_ACCEPT_SOURCE_ROUTE] = cnf->accept_source_route;
3817 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
3818 array[DEVCONF_OPTIMISTIC_DAD] = cnf->optimistic_dad;
3819 #endif
3820 #ifdef CONFIG_IPV6_MROUTE
3821 array[DEVCONF_MC_FORWARDING] = cnf->mc_forwarding;
3822 #endif
3823 array[DEVCONF_DISABLE_IPV6] = cnf->disable_ipv6;
3824 array[DEVCONF_ACCEPT_DAD] = cnf->accept_dad;
3825 array[DEVCONF_FORCE_TLLAO] = cnf->force_tllao;
3828 static inline size_t inet6_if_nlmsg_size(void)
3830 return NLMSG_ALIGN(sizeof(struct ifinfomsg))
3831 + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
3832 + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
3833 + nla_total_size(4) /* IFLA_MTU */
3834 + nla_total_size(4) /* IFLA_LINK */
3835 + nla_total_size( /* IFLA_PROTINFO */
3836 nla_total_size(4) /* IFLA_INET6_FLAGS */
3837 + nla_total_size(sizeof(struct ifla_cacheinfo))
3838 + nla_total_size(DEVCONF_MAX * 4) /* IFLA_INET6_CONF */
3839 + nla_total_size(IPSTATS_MIB_MAX * 8) /* IFLA_INET6_STATS */
3840 + nla_total_size(ICMP6_MIB_MAX * 8) /* IFLA_INET6_ICMP6STATS */
3844 static inline void __snmp6_fill_stats(u64 *stats, void __percpu **mib,
3845 int items, int bytes)
3847 int i;
3848 int pad = bytes - sizeof(u64) * items;
3849 BUG_ON(pad < 0);
3851 /* Use put_unaligned() because stats may not be aligned for u64. */
3852 put_unaligned(items, &stats[0]);
3853 for (i = 1; i < items; i++)
3854 put_unaligned(snmp_fold_field(mib, i), &stats[i]);
3856 memset(&stats[items], 0, pad);
3859 static inline void __snmp6_fill_stats64(u64 *stats, void __percpu **mib,
3860 int items, int bytes, size_t syncpoff)
3862 int i;
3863 int pad = bytes - sizeof(u64) * items;
3864 BUG_ON(pad < 0);
3866 /* Use put_unaligned() because stats may not be aligned for u64. */
3867 put_unaligned(items, &stats[0]);
3868 for (i = 1; i < items; i++)
3869 put_unaligned(snmp_fold_field64(mib, i, syncpoff), &stats[i]);
3871 memset(&stats[items], 0, pad);
3874 static void snmp6_fill_stats(u64 *stats, struct inet6_dev *idev, int attrtype,
3875 int bytes)
3877 switch (attrtype) {
3878 case IFLA_INET6_STATS:
3879 __snmp6_fill_stats64(stats, (void __percpu **)idev->stats.ipv6,
3880 IPSTATS_MIB_MAX, bytes, offsetof(struct ipstats_mib, syncp));
3881 break;
3882 case IFLA_INET6_ICMP6STATS:
3883 __snmp6_fill_stats(stats, (void __percpu **)idev->stats.icmpv6, ICMP6_MIB_MAX, bytes);
3884 break;
3888 static int inet6_fill_ifinfo(struct sk_buff *skb, struct inet6_dev *idev,
3889 u32 pid, u32 seq, int event, unsigned int flags)
3891 struct net_device *dev = idev->dev;
3892 struct nlattr *nla;
3893 struct ifinfomsg *hdr;
3894 struct nlmsghdr *nlh;
3895 void *protoinfo;
3896 struct ifla_cacheinfo ci;
3898 nlh = nlmsg_put(skb, pid, seq, event, sizeof(*hdr), flags);
3899 if (nlh == NULL)
3900 return -EMSGSIZE;
3902 hdr = nlmsg_data(nlh);
3903 hdr->ifi_family = AF_INET6;
3904 hdr->__ifi_pad = 0;
3905 hdr->ifi_type = dev->type;
3906 hdr->ifi_index = dev->ifindex;
3907 hdr->ifi_flags = dev_get_flags(dev);
3908 hdr->ifi_change = 0;
3910 NLA_PUT_STRING(skb, IFLA_IFNAME, dev->name);
3912 if (dev->addr_len)
3913 NLA_PUT(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr);
3915 NLA_PUT_U32(skb, IFLA_MTU, dev->mtu);
3916 if (dev->ifindex != dev->iflink)
3917 NLA_PUT_U32(skb, IFLA_LINK, dev->iflink);
3919 protoinfo = nla_nest_start(skb, IFLA_PROTINFO);
3920 if (protoinfo == NULL)
3921 goto nla_put_failure;
3923 NLA_PUT_U32(skb, IFLA_INET6_FLAGS, idev->if_flags);
3925 ci.max_reasm_len = IPV6_MAXPLEN;
3926 ci.tstamp = (__u32)(TIME_DELTA(idev->tstamp, INITIAL_JIFFIES) / HZ * 100
3927 + TIME_DELTA(idev->tstamp, INITIAL_JIFFIES) % HZ * 100 / HZ);
3928 ci.reachable_time = idev->nd_parms->reachable_time;
3929 ci.retrans_time = idev->nd_parms->retrans_time;
3930 NLA_PUT(skb, IFLA_INET6_CACHEINFO, sizeof(ci), &ci);
3932 nla = nla_reserve(skb, IFLA_INET6_CONF, DEVCONF_MAX * sizeof(s32));
3933 if (nla == NULL)
3934 goto nla_put_failure;
3935 ipv6_store_devconf(&idev->cnf, nla_data(nla), nla_len(nla));
3938 nla = nla_reserve(skb, IFLA_INET6_STATS, IPSTATS_MIB_MAX * sizeof(u64));
3939 if (nla == NULL)
3940 goto nla_put_failure;
3941 snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_STATS, nla_len(nla));
3943 nla = nla_reserve(skb, IFLA_INET6_ICMP6STATS, ICMP6_MIB_MAX * sizeof(u64));
3944 if (nla == NULL)
3945 goto nla_put_failure;
3946 snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_ICMP6STATS, nla_len(nla));
3948 nla_nest_end(skb, protoinfo);
3949 return nlmsg_end(skb, nlh);
3951 nla_put_failure:
3952 nlmsg_cancel(skb, nlh);
3953 return -EMSGSIZE;
3956 static int inet6_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
3958 struct net *net = sock_net(skb->sk);
3959 int h, s_h;
3960 int idx = 0, s_idx;
3961 struct net_device *dev;
3962 struct inet6_dev *idev;
3963 struct hlist_head *head;
3964 struct hlist_node *node;
3966 s_h = cb->args[0];
3967 s_idx = cb->args[1];
3969 rcu_read_lock();
3970 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
3971 idx = 0;
3972 head = &net->dev_index_head[h];
3973 hlist_for_each_entry_rcu(dev, node, head, index_hlist) {
3974 if (idx < s_idx)
3975 goto cont;
3976 idev = __in6_dev_get(dev);
3977 if (!idev)
3978 goto cont;
3979 if (inet6_fill_ifinfo(skb, idev,
3980 NETLINK_CB(cb->skb).pid,
3981 cb->nlh->nlmsg_seq,
3982 RTM_NEWLINK, NLM_F_MULTI) <= 0)
3983 goto out;
3984 cont:
3985 idx++;
3988 out:
3989 rcu_read_unlock();
3990 cb->args[1] = idx;
3991 cb->args[0] = h;
3993 return skb->len;
3996 void inet6_ifinfo_notify(int event, struct inet6_dev *idev)
3998 struct sk_buff *skb;
3999 struct net *net = dev_net(idev->dev);
4000 int err = -ENOBUFS;
4002 skb = nlmsg_new(inet6_if_nlmsg_size(), GFP_ATOMIC);
4003 if (skb == NULL)
4004 goto errout;
4006 err = inet6_fill_ifinfo(skb, idev, 0, 0, event, 0);
4007 if (err < 0) {
4008 /* -EMSGSIZE implies BUG in inet6_if_nlmsg_size() */
4009 WARN_ON(err == -EMSGSIZE);
4010 kfree_skb(skb);
4011 goto errout;
4013 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
4014 return;
4015 errout:
4016 if (err < 0)
4017 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
4020 static inline size_t inet6_prefix_nlmsg_size(void)
4022 return NLMSG_ALIGN(sizeof(struct prefixmsg))
4023 + nla_total_size(sizeof(struct in6_addr))
4024 + nla_total_size(sizeof(struct prefix_cacheinfo));
4027 static int inet6_fill_prefix(struct sk_buff *skb, struct inet6_dev *idev,
4028 struct prefix_info *pinfo, u32 pid, u32 seq,
4029 int event, unsigned int flags)
4031 struct prefixmsg *pmsg;
4032 struct nlmsghdr *nlh;
4033 struct prefix_cacheinfo ci;
4035 nlh = nlmsg_put(skb, pid, seq, event, sizeof(*pmsg), flags);
4036 if (nlh == NULL)
4037 return -EMSGSIZE;
4039 pmsg = nlmsg_data(nlh);
4040 pmsg->prefix_family = AF_INET6;
4041 pmsg->prefix_pad1 = 0;
4042 pmsg->prefix_pad2 = 0;
4043 pmsg->prefix_ifindex = idev->dev->ifindex;
4044 pmsg->prefix_len = pinfo->prefix_len;
4045 pmsg->prefix_type = pinfo->type;
4046 pmsg->prefix_pad3 = 0;
4047 pmsg->prefix_flags = 0;
4048 if (pinfo->onlink)
4049 pmsg->prefix_flags |= IF_PREFIX_ONLINK;
4050 if (pinfo->autoconf)
4051 pmsg->prefix_flags |= IF_PREFIX_AUTOCONF;
4053 NLA_PUT(skb, PREFIX_ADDRESS, sizeof(pinfo->prefix), &pinfo->prefix);
4055 ci.preferred_time = ntohl(pinfo->prefered);
4056 ci.valid_time = ntohl(pinfo->valid);
4057 NLA_PUT(skb, PREFIX_CACHEINFO, sizeof(ci), &ci);
4059 return nlmsg_end(skb, nlh);
4061 nla_put_failure:
4062 nlmsg_cancel(skb, nlh);
4063 return -EMSGSIZE;
4066 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
4067 struct prefix_info *pinfo)
4069 struct sk_buff *skb;
4070 struct net *net = dev_net(idev->dev);
4071 int err = -ENOBUFS;
4073 skb = nlmsg_new(inet6_prefix_nlmsg_size(), GFP_ATOMIC);
4074 if (skb == NULL)
4075 goto errout;
4077 err = inet6_fill_prefix(skb, idev, pinfo, 0, 0, event, 0);
4078 if (err < 0) {
4079 /* -EMSGSIZE implies BUG in inet6_prefix_nlmsg_size() */
4080 WARN_ON(err == -EMSGSIZE);
4081 kfree_skb(skb);
4082 goto errout;
4084 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_PREFIX, NULL, GFP_ATOMIC);
4085 return;
4086 errout:
4087 if (err < 0)
4088 rtnl_set_sk_err(net, RTNLGRP_IPV6_PREFIX, err);
4091 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
4093 inet6_ifa_notify(event ? : RTM_NEWADDR, ifp);
4095 switch (event) {
4096 case RTM_NEWADDR:
4098 * If the address was optimistic
4099 * we inserted the route at the start of
4100 * our DAD process, so we don't need
4101 * to do it again
4103 if (!(ifp->rt->rt6i_node))
4104 ip6_ins_rt(ifp->rt);
4105 if (ifp->idev->cnf.forwarding)
4106 addrconf_join_anycast(ifp);
4107 break;
4108 case RTM_DELADDR:
4109 if (ifp->idev->cnf.forwarding)
4110 addrconf_leave_anycast(ifp);
4111 addrconf_leave_solict(ifp->idev, &ifp->addr);
4112 dst_hold(&ifp->rt->dst);
4114 if (ifp->state == INET6_IFADDR_STATE_DEAD &&
4115 ip6_del_rt(ifp->rt))
4116 dst_free(&ifp->rt->dst);
4117 break;
4121 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
4123 rcu_read_lock_bh();
4124 if (likely(ifp->idev->dead == 0))
4125 __ipv6_ifa_notify(event, ifp);
4126 rcu_read_unlock_bh();
4129 #ifdef CONFIG_SYSCTL
4131 static
4132 int addrconf_sysctl_forward(ctl_table *ctl, int write,
4133 void __user *buffer, size_t *lenp, loff_t *ppos)
4135 int *valp = ctl->data;
4136 int val = *valp;
4137 loff_t pos = *ppos;
4138 int ret;
4140 ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
4142 if (write)
4143 ret = addrconf_fixup_forwarding(ctl, valp, val);
4144 if (ret)
4145 *ppos = pos;
4146 return ret;
4149 static void dev_disable_change(struct inet6_dev *idev)
4151 if (!idev || !idev->dev)
4152 return;
4154 if (idev->cnf.disable_ipv6)
4155 addrconf_notify(NULL, NETDEV_DOWN, idev->dev);
4156 else
4157 addrconf_notify(NULL, NETDEV_UP, idev->dev);
4160 static void addrconf_disable_change(struct net *net, __s32 newf)
4162 struct net_device *dev;
4163 struct inet6_dev *idev;
4165 rcu_read_lock();
4166 for_each_netdev_rcu(net, dev) {
4167 idev = __in6_dev_get(dev);
4168 if (idev) {
4169 int changed = (!idev->cnf.disable_ipv6) ^ (!newf);
4170 idev->cnf.disable_ipv6 = newf;
4171 if (changed)
4172 dev_disable_change(idev);
4175 rcu_read_unlock();
4178 static int addrconf_disable_ipv6(struct ctl_table *table, int *p, int old)
4180 struct net *net;
4182 net = (struct net *)table->extra2;
4184 if (p == &net->ipv6.devconf_dflt->disable_ipv6)
4185 return 0;
4187 if (!rtnl_trylock()) {
4188 /* Restore the original values before restarting */
4189 *p = old;
4190 return restart_syscall();
4193 if (p == &net->ipv6.devconf_all->disable_ipv6) {
4194 __s32 newf = net->ipv6.devconf_all->disable_ipv6;
4195 net->ipv6.devconf_dflt->disable_ipv6 = newf;
4196 addrconf_disable_change(net, newf);
4197 } else if ((!*p) ^ (!old))
4198 dev_disable_change((struct inet6_dev *)table->extra1);
4200 rtnl_unlock();
4201 return 0;
4204 static
4205 int addrconf_sysctl_disable(ctl_table *ctl, int write,
4206 void __user *buffer, size_t *lenp, loff_t *ppos)
4208 int *valp = ctl->data;
4209 int val = *valp;
4210 loff_t pos = *ppos;
4211 int ret;
4213 ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
4215 if (write)
4216 ret = addrconf_disable_ipv6(ctl, valp, val);
4217 if (ret)
4218 *ppos = pos;
4219 return ret;
4222 static struct addrconf_sysctl_table
4224 struct ctl_table_header *sysctl_header;
4225 ctl_table addrconf_vars[DEVCONF_MAX+1];
4226 char *dev_name;
4227 } addrconf_sysctl __read_mostly = {
4228 .sysctl_header = NULL,
4229 .addrconf_vars = {
4231 .procname = "forwarding",
4232 .data = &ipv6_devconf.forwarding,
4233 .maxlen = sizeof(int),
4234 .mode = 0644,
4235 .proc_handler = addrconf_sysctl_forward,
4238 .procname = "hop_limit",
4239 .data = &ipv6_devconf.hop_limit,
4240 .maxlen = sizeof(int),
4241 .mode = 0644,
4242 .proc_handler = proc_dointvec,
4245 .procname = "mtu",
4246 .data = &ipv6_devconf.mtu6,
4247 .maxlen = sizeof(int),
4248 .mode = 0644,
4249 .proc_handler = proc_dointvec,
4252 .procname = "accept_ra",
4253 .data = &ipv6_devconf.accept_ra,
4254 .maxlen = sizeof(int),
4255 .mode = 0644,
4256 .proc_handler = proc_dointvec,
4259 .procname = "accept_redirects",
4260 .data = &ipv6_devconf.accept_redirects,
4261 .maxlen = sizeof(int),
4262 .mode = 0644,
4263 .proc_handler = proc_dointvec,
4266 .procname = "autoconf",
4267 .data = &ipv6_devconf.autoconf,
4268 .maxlen = sizeof(int),
4269 .mode = 0644,
4270 .proc_handler = proc_dointvec,
4273 .procname = "dad_transmits",
4274 .data = &ipv6_devconf.dad_transmits,
4275 .maxlen = sizeof(int),
4276 .mode = 0644,
4277 .proc_handler = proc_dointvec,
4280 .procname = "router_solicitations",
4281 .data = &ipv6_devconf.rtr_solicits,
4282 .maxlen = sizeof(int),
4283 .mode = 0644,
4284 .proc_handler = proc_dointvec,
4287 .procname = "router_solicitation_interval",
4288 .data = &ipv6_devconf.rtr_solicit_interval,
4289 .maxlen = sizeof(int),
4290 .mode = 0644,
4291 .proc_handler = proc_dointvec_jiffies,
4294 .procname = "router_solicitation_delay",
4295 .data = &ipv6_devconf.rtr_solicit_delay,
4296 .maxlen = sizeof(int),
4297 .mode = 0644,
4298 .proc_handler = proc_dointvec_jiffies,
4301 .procname = "force_mld_version",
4302 .data = &ipv6_devconf.force_mld_version,
4303 .maxlen = sizeof(int),
4304 .mode = 0644,
4305 .proc_handler = proc_dointvec,
4307 #ifdef CONFIG_IPV6_PRIVACY
4309 .procname = "use_tempaddr",
4310 .data = &ipv6_devconf.use_tempaddr,
4311 .maxlen = sizeof(int),
4312 .mode = 0644,
4313 .proc_handler = proc_dointvec,
4316 .procname = "temp_valid_lft",
4317 .data = &ipv6_devconf.temp_valid_lft,
4318 .maxlen = sizeof(int),
4319 .mode = 0644,
4320 .proc_handler = proc_dointvec,
4323 .procname = "temp_prefered_lft",
4324 .data = &ipv6_devconf.temp_prefered_lft,
4325 .maxlen = sizeof(int),
4326 .mode = 0644,
4327 .proc_handler = proc_dointvec,
4330 .procname = "regen_max_retry",
4331 .data = &ipv6_devconf.regen_max_retry,
4332 .maxlen = sizeof(int),
4333 .mode = 0644,
4334 .proc_handler = proc_dointvec,
4337 .procname = "max_desync_factor",
4338 .data = &ipv6_devconf.max_desync_factor,
4339 .maxlen = sizeof(int),
4340 .mode = 0644,
4341 .proc_handler = proc_dointvec,
4343 #endif
4345 .procname = "max_addresses",
4346 .data = &ipv6_devconf.max_addresses,
4347 .maxlen = sizeof(int),
4348 .mode = 0644,
4349 .proc_handler = proc_dointvec,
4352 .procname = "accept_ra_defrtr",
4353 .data = &ipv6_devconf.accept_ra_defrtr,
4354 .maxlen = sizeof(int),
4355 .mode = 0644,
4356 .proc_handler = proc_dointvec,
4359 .procname = "accept_ra_pinfo",
4360 .data = &ipv6_devconf.accept_ra_pinfo,
4361 .maxlen = sizeof(int),
4362 .mode = 0644,
4363 .proc_handler = proc_dointvec,
4365 #ifdef CONFIG_IPV6_ROUTER_PREF
4367 .procname = "accept_ra_rtr_pref",
4368 .data = &ipv6_devconf.accept_ra_rtr_pref,
4369 .maxlen = sizeof(int),
4370 .mode = 0644,
4371 .proc_handler = proc_dointvec,
4374 .procname = "router_probe_interval",
4375 .data = &ipv6_devconf.rtr_probe_interval,
4376 .maxlen = sizeof(int),
4377 .mode = 0644,
4378 .proc_handler = proc_dointvec_jiffies,
4380 #ifdef CONFIG_IPV6_ROUTE_INFO
4382 .procname = "accept_ra_rt_info_max_plen",
4383 .data = &ipv6_devconf.accept_ra_rt_info_max_plen,
4384 .maxlen = sizeof(int),
4385 .mode = 0644,
4386 .proc_handler = proc_dointvec,
4388 #endif
4389 #endif
4391 .procname = "proxy_ndp",
4392 .data = &ipv6_devconf.proxy_ndp,
4393 .maxlen = sizeof(int),
4394 .mode = 0644,
4395 .proc_handler = proc_dointvec,
4398 .procname = "accept_source_route",
4399 .data = &ipv6_devconf.accept_source_route,
4400 .maxlen = sizeof(int),
4401 .mode = 0644,
4402 .proc_handler = proc_dointvec,
4404 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
4406 .procname = "optimistic_dad",
4407 .data = &ipv6_devconf.optimistic_dad,
4408 .maxlen = sizeof(int),
4409 .mode = 0644,
4410 .proc_handler = proc_dointvec,
4413 #endif
4414 #ifdef CONFIG_IPV6_MROUTE
4416 .procname = "mc_forwarding",
4417 .data = &ipv6_devconf.mc_forwarding,
4418 .maxlen = sizeof(int),
4419 .mode = 0444,
4420 .proc_handler = proc_dointvec,
4422 #endif
4424 .procname = "disable_ipv6",
4425 .data = &ipv6_devconf.disable_ipv6,
4426 .maxlen = sizeof(int),
4427 .mode = 0644,
4428 .proc_handler = addrconf_sysctl_disable,
4431 .procname = "accept_dad",
4432 .data = &ipv6_devconf.accept_dad,
4433 .maxlen = sizeof(int),
4434 .mode = 0644,
4435 .proc_handler = proc_dointvec,
4438 .procname = "force_tllao",
4439 .data = &ipv6_devconf.force_tllao,
4440 .maxlen = sizeof(int),
4441 .mode = 0644,
4442 .proc_handler = proc_dointvec
4445 /* sentinel */
4450 static int __addrconf_sysctl_register(struct net *net, char *dev_name,
4451 struct inet6_dev *idev, struct ipv6_devconf *p)
4453 int i;
4454 struct addrconf_sysctl_table *t;
4456 #define ADDRCONF_CTL_PATH_DEV 3
4458 struct ctl_path addrconf_ctl_path[] = {
4459 { .procname = "net", },
4460 { .procname = "ipv6", },
4461 { .procname = "conf", },
4462 { /* to be set */ },
4463 { },
4467 t = kmemdup(&addrconf_sysctl, sizeof(*t), GFP_KERNEL);
4468 if (t == NULL)
4469 goto out;
4471 for (i = 0; t->addrconf_vars[i].data; i++) {
4472 t->addrconf_vars[i].data += (char *)p - (char *)&ipv6_devconf;
4473 t->addrconf_vars[i].extra1 = idev; /* embedded; no ref */
4474 t->addrconf_vars[i].extra2 = net;
4478 * Make a copy of dev_name, because '.procname' is regarded as const
4479 * by sysctl and we wouldn't want anyone to change it under our feet
4480 * (see SIOCSIFNAME).
4482 t->dev_name = kstrdup(dev_name, GFP_KERNEL);
4483 if (!t->dev_name)
4484 goto free;
4486 addrconf_ctl_path[ADDRCONF_CTL_PATH_DEV].procname = t->dev_name;
4488 t->sysctl_header = register_net_sysctl_table(net, addrconf_ctl_path,
4489 t->addrconf_vars);
4490 if (t->sysctl_header == NULL)
4491 goto free_procname;
4493 p->sysctl = t;
4494 return 0;
4496 free_procname:
4497 kfree(t->dev_name);
4498 free:
4499 kfree(t);
4500 out:
4501 return -ENOBUFS;
4504 static void __addrconf_sysctl_unregister(struct ipv6_devconf *p)
4506 struct addrconf_sysctl_table *t;
4508 if (p->sysctl == NULL)
4509 return;
4511 t = p->sysctl;
4512 p->sysctl = NULL;
4513 unregister_sysctl_table(t->sysctl_header);
4514 kfree(t->dev_name);
4515 kfree(t);
4518 static void addrconf_sysctl_register(struct inet6_dev *idev)
4520 neigh_sysctl_register(idev->dev, idev->nd_parms, "ipv6",
4521 &ndisc_ifinfo_sysctl_change);
4522 __addrconf_sysctl_register(dev_net(idev->dev), idev->dev->name,
4523 idev, &idev->cnf);
4526 static void addrconf_sysctl_unregister(struct inet6_dev *idev)
4528 __addrconf_sysctl_unregister(&idev->cnf);
4529 neigh_sysctl_unregister(idev->nd_parms);
4533 #endif
4535 static int __net_init addrconf_init_net(struct net *net)
4537 int err;
4538 struct ipv6_devconf *all, *dflt;
4540 err = -ENOMEM;
4541 all = &ipv6_devconf;
4542 dflt = &ipv6_devconf_dflt;
4544 if (!net_eq(net, &init_net)) {
4545 all = kmemdup(all, sizeof(ipv6_devconf), GFP_KERNEL);
4546 if (all == NULL)
4547 goto err_alloc_all;
4549 dflt = kmemdup(dflt, sizeof(ipv6_devconf_dflt), GFP_KERNEL);
4550 if (dflt == NULL)
4551 goto err_alloc_dflt;
4552 } else {
4553 /* these will be inherited by all namespaces */
4554 dflt->autoconf = ipv6_defaults.autoconf;
4555 dflt->disable_ipv6 = ipv6_defaults.disable_ipv6;
4558 net->ipv6.devconf_all = all;
4559 net->ipv6.devconf_dflt = dflt;
4561 #ifdef CONFIG_SYSCTL
4562 err = __addrconf_sysctl_register(net, "all", NULL, all);
4563 if (err < 0)
4564 goto err_reg_all;
4566 err = __addrconf_sysctl_register(net, "default", NULL, dflt);
4567 if (err < 0)
4568 goto err_reg_dflt;
4569 #endif
4570 return 0;
4572 #ifdef CONFIG_SYSCTL
4573 err_reg_dflt:
4574 __addrconf_sysctl_unregister(all);
4575 err_reg_all:
4576 kfree(dflt);
4577 #endif
4578 err_alloc_dflt:
4579 kfree(all);
4580 err_alloc_all:
4581 return err;
4584 static void __net_exit addrconf_exit_net(struct net *net)
4586 #ifdef CONFIG_SYSCTL
4587 __addrconf_sysctl_unregister(net->ipv6.devconf_dflt);
4588 __addrconf_sysctl_unregister(net->ipv6.devconf_all);
4589 #endif
4590 if (!net_eq(net, &init_net)) {
4591 kfree(net->ipv6.devconf_dflt);
4592 kfree(net->ipv6.devconf_all);
4596 static struct pernet_operations addrconf_ops = {
4597 .init = addrconf_init_net,
4598 .exit = addrconf_exit_net,
4602 * Device notifier
4605 int register_inet6addr_notifier(struct notifier_block *nb)
4607 return atomic_notifier_chain_register(&inet6addr_chain, nb);
4609 EXPORT_SYMBOL(register_inet6addr_notifier);
4611 int unregister_inet6addr_notifier(struct notifier_block *nb)
4613 return atomic_notifier_chain_unregister(&inet6addr_chain, nb);
4615 EXPORT_SYMBOL(unregister_inet6addr_notifier);
4618 * Init / cleanup code
4621 int __init addrconf_init(void)
4623 int i, err;
4625 err = ipv6_addr_label_init();
4626 if (err < 0) {
4627 printk(KERN_CRIT "IPv6 Addrconf:"
4628 " cannot initialize default policy table: %d.\n", err);
4629 goto out;
4632 err = register_pernet_subsys(&addrconf_ops);
4633 if (err < 0)
4634 goto out_addrlabel;
4636 /* The addrconf netdev notifier requires that loopback_dev
4637 * has it's ipv6 private information allocated and setup
4638 * before it can bring up and give link-local addresses
4639 * to other devices which are up.
4641 * Unfortunately, loopback_dev is not necessarily the first
4642 * entry in the global dev_base list of net devices. In fact,
4643 * it is likely to be the very last entry on that list.
4644 * So this causes the notifier registry below to try and
4645 * give link-local addresses to all devices besides loopback_dev
4646 * first, then loopback_dev, which cases all the non-loopback_dev
4647 * devices to fail to get a link-local address.
4649 * So, as a temporary fix, allocate the ipv6 structure for
4650 * loopback_dev first by hand.
4651 * Longer term, all of the dependencies ipv6 has upon the loopback
4652 * device and it being up should be removed.
4654 rtnl_lock();
4655 if (!ipv6_add_dev(init_net.loopback_dev))
4656 err = -ENOMEM;
4657 rtnl_unlock();
4658 if (err)
4659 goto errlo;
4661 for (i = 0; i < IN6_ADDR_HSIZE; i++)
4662 INIT_HLIST_HEAD(&inet6_addr_lst[i]);
4664 register_netdevice_notifier(&ipv6_dev_notf);
4666 addrconf_verify(0);
4668 err = __rtnl_register(PF_INET6, RTM_GETLINK, NULL, inet6_dump_ifinfo);
4669 if (err < 0)
4670 goto errout;
4672 /* Only the first call to __rtnl_register can fail */
4673 __rtnl_register(PF_INET6, RTM_NEWADDR, inet6_rtm_newaddr, NULL);
4674 __rtnl_register(PF_INET6, RTM_DELADDR, inet6_rtm_deladdr, NULL);
4675 __rtnl_register(PF_INET6, RTM_GETADDR, inet6_rtm_getaddr, inet6_dump_ifaddr);
4676 __rtnl_register(PF_INET6, RTM_GETMULTICAST, NULL, inet6_dump_ifmcaddr);
4677 __rtnl_register(PF_INET6, RTM_GETANYCAST, NULL, inet6_dump_ifacaddr);
4679 ipv6_addr_label_rtnl_register();
4681 return 0;
4682 errout:
4683 unregister_netdevice_notifier(&ipv6_dev_notf);
4684 errlo:
4685 unregister_pernet_subsys(&addrconf_ops);
4686 out_addrlabel:
4687 ipv6_addr_label_cleanup();
4688 out:
4689 return err;
4692 void addrconf_cleanup(void)
4694 struct net_device *dev;
4695 int i;
4697 unregister_netdevice_notifier(&ipv6_dev_notf);
4698 unregister_pernet_subsys(&addrconf_ops);
4699 ipv6_addr_label_cleanup();
4701 rtnl_lock();
4703 /* clean dev list */
4704 for_each_netdev(&init_net, dev) {
4705 if (__in6_dev_get(dev) == NULL)
4706 continue;
4707 addrconf_ifdown(dev, 1);
4709 addrconf_ifdown(init_net.loopback_dev, 2);
4712 * Check hash table.
4714 spin_lock_bh(&addrconf_hash_lock);
4715 for (i = 0; i < IN6_ADDR_HSIZE; i++)
4716 WARN_ON(!hlist_empty(&inet6_addr_lst[i]));
4717 spin_unlock_bh(&addrconf_hash_lock);
4719 del_timer(&addr_chk_timer);
4720 rtnl_unlock();