Staging: hv: remove OnChildDeviceRemove vmbus_driver callback
[zen-stable.git] / net / ipv6 / addrconf.c
blobb41ce0f0d514154597753d549b5a29a3bc7e7ae0
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; /*XXX*/
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, age;
840 unsigned long regen_advance;
841 int tmp_plen;
842 int ret = 0;
843 int max_addresses;
844 u32 addr_flags;
846 write_lock(&idev->lock);
847 if (ift) {
848 spin_lock_bh(&ift->lock);
849 memcpy(&addr.s6_addr[8], &ift->addr.s6_addr[8], 8);
850 spin_unlock_bh(&ift->lock);
851 tmpaddr = &addr;
852 } else {
853 tmpaddr = NULL;
855 retry:
856 in6_dev_hold(idev);
857 if (idev->cnf.use_tempaddr <= 0) {
858 write_unlock(&idev->lock);
859 printk(KERN_INFO
860 "ipv6_create_tempaddr(): use_tempaddr is disabled.\n");
861 in6_dev_put(idev);
862 ret = -1;
863 goto out;
865 spin_lock_bh(&ifp->lock);
866 if (ifp->regen_count++ >= idev->cnf.regen_max_retry) {
867 idev->cnf.use_tempaddr = -1; /*XXX*/
868 spin_unlock_bh(&ifp->lock);
869 write_unlock(&idev->lock);
870 printk(KERN_WARNING
871 "ipv6_create_tempaddr(): regeneration time exceeded. disabled temporary address support.\n");
872 in6_dev_put(idev);
873 ret = -1;
874 goto out;
876 in6_ifa_hold(ifp);
877 memcpy(addr.s6_addr, ifp->addr.s6_addr, 8);
878 if (__ipv6_try_regen_rndid(idev, tmpaddr) < 0) {
879 spin_unlock_bh(&ifp->lock);
880 write_unlock(&idev->lock);
881 printk(KERN_WARNING
882 "ipv6_create_tempaddr(): regeneration of randomized interface id failed.\n");
883 in6_ifa_put(ifp);
884 in6_dev_put(idev);
885 ret = -1;
886 goto out;
888 memcpy(&addr.s6_addr[8], idev->rndid, 8);
889 age = (jiffies - ifp->tstamp) / HZ;
890 tmp_valid_lft = min_t(__u32,
891 ifp->valid_lft,
892 idev->cnf.temp_valid_lft + age);
893 tmp_prefered_lft = min_t(__u32,
894 ifp->prefered_lft,
895 idev->cnf.temp_prefered_lft + age -
896 idev->cnf.max_desync_factor);
897 tmp_plen = ifp->prefix_len;
898 max_addresses = idev->cnf.max_addresses;
899 tmp_cstamp = ifp->cstamp;
900 tmp_tstamp = ifp->tstamp;
901 spin_unlock_bh(&ifp->lock);
903 regen_advance = idev->cnf.regen_max_retry *
904 idev->cnf.dad_transmits *
905 idev->nd_parms->retrans_time / HZ;
906 write_unlock(&idev->lock);
908 /* A temporary address is created only if this calculated Preferred
909 * Lifetime is greater than REGEN_ADVANCE time units. In particular,
910 * an implementation must not create a temporary address with a zero
911 * Preferred Lifetime.
913 if (tmp_prefered_lft <= regen_advance) {
914 in6_ifa_put(ifp);
915 in6_dev_put(idev);
916 ret = -1;
917 goto out;
920 addr_flags = IFA_F_TEMPORARY;
921 /* set in addrconf_prefix_rcv() */
922 if (ifp->flags & IFA_F_OPTIMISTIC)
923 addr_flags |= IFA_F_OPTIMISTIC;
925 ift = !max_addresses ||
926 ipv6_count_addresses(idev) < max_addresses ?
927 ipv6_add_addr(idev, &addr, tmp_plen,
928 ipv6_addr_type(&addr)&IPV6_ADDR_SCOPE_MASK,
929 addr_flags) : NULL;
930 if (!ift || IS_ERR(ift)) {
931 in6_ifa_put(ifp);
932 in6_dev_put(idev);
933 printk(KERN_INFO
934 "ipv6_create_tempaddr(): retry temporary address regeneration.\n");
935 tmpaddr = &addr;
936 write_lock(&idev->lock);
937 goto retry;
940 spin_lock_bh(&ift->lock);
941 ift->ifpub = ifp;
942 ift->valid_lft = tmp_valid_lft;
943 ift->prefered_lft = tmp_prefered_lft;
944 ift->cstamp = tmp_cstamp;
945 ift->tstamp = tmp_tstamp;
946 spin_unlock_bh(&ift->lock);
948 addrconf_dad_start(ift, 0);
949 in6_ifa_put(ift);
950 in6_dev_put(idev);
951 out:
952 return ret;
954 #endif
957 * Choose an appropriate source address (RFC3484)
959 enum {
960 IPV6_SADDR_RULE_INIT = 0,
961 IPV6_SADDR_RULE_LOCAL,
962 IPV6_SADDR_RULE_SCOPE,
963 IPV6_SADDR_RULE_PREFERRED,
964 #ifdef CONFIG_IPV6_MIP6
965 IPV6_SADDR_RULE_HOA,
966 #endif
967 IPV6_SADDR_RULE_OIF,
968 IPV6_SADDR_RULE_LABEL,
969 #ifdef CONFIG_IPV6_PRIVACY
970 IPV6_SADDR_RULE_PRIVACY,
971 #endif
972 IPV6_SADDR_RULE_ORCHID,
973 IPV6_SADDR_RULE_PREFIX,
974 IPV6_SADDR_RULE_MAX
977 struct ipv6_saddr_score {
978 int rule;
979 int addr_type;
980 struct inet6_ifaddr *ifa;
981 DECLARE_BITMAP(scorebits, IPV6_SADDR_RULE_MAX);
982 int scopedist;
983 int matchlen;
986 struct ipv6_saddr_dst {
987 const struct in6_addr *addr;
988 int ifindex;
989 int scope;
990 int label;
991 unsigned int prefs;
994 static inline int ipv6_saddr_preferred(int type)
996 if (type & (IPV6_ADDR_MAPPED|IPV6_ADDR_COMPATv4|IPV6_ADDR_LOOPBACK))
997 return 1;
998 return 0;
1001 static int ipv6_get_saddr_eval(struct net *net,
1002 struct ipv6_saddr_score *score,
1003 struct ipv6_saddr_dst *dst,
1004 int i)
1006 int ret;
1008 if (i <= score->rule) {
1009 switch (i) {
1010 case IPV6_SADDR_RULE_SCOPE:
1011 ret = score->scopedist;
1012 break;
1013 case IPV6_SADDR_RULE_PREFIX:
1014 ret = score->matchlen;
1015 break;
1016 default:
1017 ret = !!test_bit(i, score->scorebits);
1019 goto out;
1022 switch (i) {
1023 case IPV6_SADDR_RULE_INIT:
1024 /* Rule 0: remember if hiscore is not ready yet */
1025 ret = !!score->ifa;
1026 break;
1027 case IPV6_SADDR_RULE_LOCAL:
1028 /* Rule 1: Prefer same address */
1029 ret = ipv6_addr_equal(&score->ifa->addr, dst->addr);
1030 break;
1031 case IPV6_SADDR_RULE_SCOPE:
1032 /* Rule 2: Prefer appropriate scope
1034 * ret
1036 * -1 | d 15
1037 * ---+--+-+---> scope
1039 * | d is scope of the destination.
1040 * B-d | \
1041 * | \ <- smaller scope is better if
1042 * B-15 | \ if scope is enough for destinaion.
1043 * | ret = B - scope (-1 <= scope >= d <= 15).
1044 * d-C-1 | /
1045 * |/ <- greater is better
1046 * -C / if scope is not enough for destination.
1047 * /| ret = scope - C (-1 <= d < scope <= 15).
1049 * d - C - 1 < B -15 (for all -1 <= d <= 15).
1050 * C > d + 14 - B >= 15 + 14 - B = 29 - B.
1051 * Assume B = 0 and we get C > 29.
1053 ret = __ipv6_addr_src_scope(score->addr_type);
1054 if (ret >= dst->scope)
1055 ret = -ret;
1056 else
1057 ret -= 128; /* 30 is enough */
1058 score->scopedist = ret;
1059 break;
1060 case IPV6_SADDR_RULE_PREFERRED:
1061 /* Rule 3: Avoid deprecated and optimistic addresses */
1062 ret = ipv6_saddr_preferred(score->addr_type) ||
1063 !(score->ifa->flags & (IFA_F_DEPRECATED|IFA_F_OPTIMISTIC));
1064 break;
1065 #ifdef CONFIG_IPV6_MIP6
1066 case IPV6_SADDR_RULE_HOA:
1068 /* Rule 4: Prefer home address */
1069 int prefhome = !(dst->prefs & IPV6_PREFER_SRC_COA);
1070 ret = !(score->ifa->flags & IFA_F_HOMEADDRESS) ^ prefhome;
1071 break;
1073 #endif
1074 case IPV6_SADDR_RULE_OIF:
1075 /* Rule 5: Prefer outgoing interface */
1076 ret = (!dst->ifindex ||
1077 dst->ifindex == score->ifa->idev->dev->ifindex);
1078 break;
1079 case IPV6_SADDR_RULE_LABEL:
1080 /* Rule 6: Prefer matching label */
1081 ret = ipv6_addr_label(net,
1082 &score->ifa->addr, score->addr_type,
1083 score->ifa->idev->dev->ifindex) == dst->label;
1084 break;
1085 #ifdef CONFIG_IPV6_PRIVACY
1086 case IPV6_SADDR_RULE_PRIVACY:
1088 /* Rule 7: Prefer public address
1089 * Note: prefer temprary address if use_tempaddr >= 2
1091 int preftmp = dst->prefs & (IPV6_PREFER_SRC_PUBLIC|IPV6_PREFER_SRC_TMP) ?
1092 !!(dst->prefs & IPV6_PREFER_SRC_TMP) :
1093 score->ifa->idev->cnf.use_tempaddr >= 2;
1094 ret = (!(score->ifa->flags & IFA_F_TEMPORARY)) ^ preftmp;
1095 break;
1097 #endif
1098 case IPV6_SADDR_RULE_ORCHID:
1099 /* Rule 8-: Prefer ORCHID vs ORCHID or
1100 * non-ORCHID vs non-ORCHID
1102 ret = !(ipv6_addr_orchid(&score->ifa->addr) ^
1103 ipv6_addr_orchid(dst->addr));
1104 break;
1105 case IPV6_SADDR_RULE_PREFIX:
1106 /* Rule 8: Use longest matching prefix */
1107 score->matchlen = ret = ipv6_addr_diff(&score->ifa->addr,
1108 dst->addr);
1109 break;
1110 default:
1111 ret = 0;
1114 if (ret)
1115 __set_bit(i, score->scorebits);
1116 score->rule = i;
1117 out:
1118 return ret;
1121 int ipv6_dev_get_saddr(struct net *net, struct net_device *dst_dev,
1122 const struct in6_addr *daddr, unsigned int prefs,
1123 struct in6_addr *saddr)
1125 struct ipv6_saddr_score scores[2],
1126 *score = &scores[0], *hiscore = &scores[1];
1127 struct ipv6_saddr_dst dst;
1128 struct net_device *dev;
1129 int dst_type;
1131 dst_type = __ipv6_addr_type(daddr);
1132 dst.addr = daddr;
1133 dst.ifindex = dst_dev ? dst_dev->ifindex : 0;
1134 dst.scope = __ipv6_addr_src_scope(dst_type);
1135 dst.label = ipv6_addr_label(net, daddr, dst_type, dst.ifindex);
1136 dst.prefs = prefs;
1138 hiscore->rule = -1;
1139 hiscore->ifa = NULL;
1141 rcu_read_lock();
1143 for_each_netdev_rcu(net, dev) {
1144 struct inet6_dev *idev;
1146 /* Candidate Source Address (section 4)
1147 * - multicast and link-local destination address,
1148 * the set of candidate source address MUST only
1149 * include addresses assigned to interfaces
1150 * belonging to the same link as the outgoing
1151 * interface.
1152 * (- For site-local destination addresses, the
1153 * set of candidate source addresses MUST only
1154 * include addresses assigned to interfaces
1155 * belonging to the same site as the outgoing
1156 * interface.)
1158 if (((dst_type & IPV6_ADDR_MULTICAST) ||
1159 dst.scope <= IPV6_ADDR_SCOPE_LINKLOCAL) &&
1160 dst.ifindex && dev->ifindex != dst.ifindex)
1161 continue;
1163 idev = __in6_dev_get(dev);
1164 if (!idev)
1165 continue;
1167 read_lock_bh(&idev->lock);
1168 list_for_each_entry(score->ifa, &idev->addr_list, if_list) {
1169 int i;
1172 * - Tentative Address (RFC2462 section 5.4)
1173 * - A tentative address is not considered
1174 * "assigned to an interface" in the traditional
1175 * sense, unless it is also flagged as optimistic.
1176 * - Candidate Source Address (section 4)
1177 * - In any case, anycast addresses, multicast
1178 * addresses, and the unspecified address MUST
1179 * NOT be included in a candidate set.
1181 if ((score->ifa->flags & IFA_F_TENTATIVE) &&
1182 (!(score->ifa->flags & IFA_F_OPTIMISTIC)))
1183 continue;
1185 score->addr_type = __ipv6_addr_type(&score->ifa->addr);
1187 if (unlikely(score->addr_type == IPV6_ADDR_ANY ||
1188 score->addr_type & IPV6_ADDR_MULTICAST)) {
1189 LIMIT_NETDEBUG(KERN_DEBUG
1190 "ADDRCONF: unspecified / multicast address "
1191 "assigned as unicast address on %s",
1192 dev->name);
1193 continue;
1196 score->rule = -1;
1197 bitmap_zero(score->scorebits, IPV6_SADDR_RULE_MAX);
1199 for (i = 0; i < IPV6_SADDR_RULE_MAX; i++) {
1200 int minihiscore, miniscore;
1202 minihiscore = ipv6_get_saddr_eval(net, hiscore, &dst, i);
1203 miniscore = ipv6_get_saddr_eval(net, score, &dst, i);
1205 if (minihiscore > miniscore) {
1206 if (i == IPV6_SADDR_RULE_SCOPE &&
1207 score->scopedist > 0) {
1209 * special case:
1210 * each remaining entry
1211 * has too small (not enough)
1212 * scope, because ifa entries
1213 * are sorted by their scope
1214 * values.
1216 goto try_nextdev;
1218 break;
1219 } else if (minihiscore < miniscore) {
1220 if (hiscore->ifa)
1221 in6_ifa_put(hiscore->ifa);
1223 in6_ifa_hold(score->ifa);
1225 swap(hiscore, score);
1227 /* restore our iterator */
1228 score->ifa = hiscore->ifa;
1230 break;
1234 try_nextdev:
1235 read_unlock_bh(&idev->lock);
1237 rcu_read_unlock();
1239 if (!hiscore->ifa)
1240 return -EADDRNOTAVAIL;
1242 ipv6_addr_copy(saddr, &hiscore->ifa->addr);
1243 in6_ifa_put(hiscore->ifa);
1244 return 0;
1246 EXPORT_SYMBOL(ipv6_dev_get_saddr);
1248 int ipv6_get_lladdr(struct net_device *dev, struct in6_addr *addr,
1249 unsigned char banned_flags)
1251 struct inet6_dev *idev;
1252 int err = -EADDRNOTAVAIL;
1254 rcu_read_lock();
1255 idev = __in6_dev_get(dev);
1256 if (idev) {
1257 struct inet6_ifaddr *ifp;
1259 read_lock_bh(&idev->lock);
1260 list_for_each_entry(ifp, &idev->addr_list, if_list) {
1261 if (ifp->scope == IFA_LINK &&
1262 !(ifp->flags & banned_flags)) {
1263 ipv6_addr_copy(addr, &ifp->addr);
1264 err = 0;
1265 break;
1268 read_unlock_bh(&idev->lock);
1270 rcu_read_unlock();
1271 return err;
1274 static int ipv6_count_addresses(struct inet6_dev *idev)
1276 int cnt = 0;
1277 struct inet6_ifaddr *ifp;
1279 read_lock_bh(&idev->lock);
1280 list_for_each_entry(ifp, &idev->addr_list, if_list)
1281 cnt++;
1282 read_unlock_bh(&idev->lock);
1283 return cnt;
1286 int ipv6_chk_addr(struct net *net, struct in6_addr *addr,
1287 struct net_device *dev, int strict)
1289 struct inet6_ifaddr *ifp;
1290 struct hlist_node *node;
1291 unsigned int hash = ipv6_addr_hash(addr);
1293 rcu_read_lock_bh();
1294 hlist_for_each_entry_rcu(ifp, node, &inet6_addr_lst[hash], addr_lst) {
1295 if (!net_eq(dev_net(ifp->idev->dev), net))
1296 continue;
1297 if (ipv6_addr_equal(&ifp->addr, addr) &&
1298 !(ifp->flags&IFA_F_TENTATIVE) &&
1299 (dev == NULL || ifp->idev->dev == dev ||
1300 !(ifp->scope&(IFA_LINK|IFA_HOST) || strict))) {
1301 rcu_read_unlock_bh();
1302 return 1;
1306 rcu_read_unlock_bh();
1307 return 0;
1309 EXPORT_SYMBOL(ipv6_chk_addr);
1311 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
1312 struct net_device *dev)
1314 unsigned int hash = ipv6_addr_hash(addr);
1315 struct inet6_ifaddr *ifp;
1316 struct hlist_node *node;
1318 hlist_for_each_entry(ifp, node, &inet6_addr_lst[hash], addr_lst) {
1319 if (!net_eq(dev_net(ifp->idev->dev), net))
1320 continue;
1321 if (ipv6_addr_equal(&ifp->addr, addr)) {
1322 if (dev == NULL || ifp->idev->dev == dev)
1323 return true;
1326 return false;
1329 int ipv6_chk_prefix(struct in6_addr *addr, struct net_device *dev)
1331 struct inet6_dev *idev;
1332 struct inet6_ifaddr *ifa;
1333 int onlink;
1335 onlink = 0;
1336 rcu_read_lock();
1337 idev = __in6_dev_get(dev);
1338 if (idev) {
1339 read_lock_bh(&idev->lock);
1340 list_for_each_entry(ifa, &idev->addr_list, if_list) {
1341 onlink = ipv6_prefix_equal(addr, &ifa->addr,
1342 ifa->prefix_len);
1343 if (onlink)
1344 break;
1346 read_unlock_bh(&idev->lock);
1348 rcu_read_unlock();
1349 return onlink;
1352 EXPORT_SYMBOL(ipv6_chk_prefix);
1354 struct inet6_ifaddr *ipv6_get_ifaddr(struct net *net, const struct in6_addr *addr,
1355 struct net_device *dev, int strict)
1357 struct inet6_ifaddr *ifp, *result = NULL;
1358 unsigned int hash = ipv6_addr_hash(addr);
1359 struct hlist_node *node;
1361 rcu_read_lock_bh();
1362 hlist_for_each_entry_rcu_bh(ifp, node, &inet6_addr_lst[hash], addr_lst) {
1363 if (!net_eq(dev_net(ifp->idev->dev), net))
1364 continue;
1365 if (ipv6_addr_equal(&ifp->addr, addr)) {
1366 if (dev == NULL || ifp->idev->dev == dev ||
1367 !(ifp->scope&(IFA_LINK|IFA_HOST) || strict)) {
1368 result = ifp;
1369 in6_ifa_hold(ifp);
1370 break;
1374 rcu_read_unlock_bh();
1376 return result;
1379 /* Gets referenced address, destroys ifaddr */
1381 static void addrconf_dad_stop(struct inet6_ifaddr *ifp, int dad_failed)
1383 if (ifp->flags&IFA_F_PERMANENT) {
1384 spin_lock_bh(&ifp->lock);
1385 addrconf_del_timer(ifp);
1386 ifp->flags |= IFA_F_TENTATIVE;
1387 if (dad_failed)
1388 ifp->flags |= IFA_F_DADFAILED;
1389 spin_unlock_bh(&ifp->lock);
1390 if (dad_failed)
1391 ipv6_ifa_notify(0, ifp);
1392 in6_ifa_put(ifp);
1393 #ifdef CONFIG_IPV6_PRIVACY
1394 } else if (ifp->flags&IFA_F_TEMPORARY) {
1395 struct inet6_ifaddr *ifpub;
1396 spin_lock_bh(&ifp->lock);
1397 ifpub = ifp->ifpub;
1398 if (ifpub) {
1399 in6_ifa_hold(ifpub);
1400 spin_unlock_bh(&ifp->lock);
1401 ipv6_create_tempaddr(ifpub, ifp);
1402 in6_ifa_put(ifpub);
1403 } else {
1404 spin_unlock_bh(&ifp->lock);
1406 ipv6_del_addr(ifp);
1407 #endif
1408 } else
1409 ipv6_del_addr(ifp);
1412 static int addrconf_dad_end(struct inet6_ifaddr *ifp)
1414 int err = -ENOENT;
1416 spin_lock(&ifp->state_lock);
1417 if (ifp->state == INET6_IFADDR_STATE_DAD) {
1418 ifp->state = INET6_IFADDR_STATE_POSTDAD;
1419 err = 0;
1421 spin_unlock(&ifp->state_lock);
1423 return err;
1426 void addrconf_dad_failure(struct inet6_ifaddr *ifp)
1428 struct inet6_dev *idev = ifp->idev;
1430 if (addrconf_dad_end(ifp)) {
1431 in6_ifa_put(ifp);
1432 return;
1435 if (net_ratelimit())
1436 printk(KERN_INFO "%s: IPv6 duplicate address %pI6c detected!\n",
1437 ifp->idev->dev->name, &ifp->addr);
1439 if (idev->cnf.accept_dad > 1 && !idev->cnf.disable_ipv6) {
1440 struct in6_addr addr;
1442 addr.s6_addr32[0] = htonl(0xfe800000);
1443 addr.s6_addr32[1] = 0;
1445 if (!ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) &&
1446 ipv6_addr_equal(&ifp->addr, &addr)) {
1447 /* DAD failed for link-local based on MAC address */
1448 idev->cnf.disable_ipv6 = 1;
1450 printk(KERN_INFO "%s: IPv6 being disabled!\n",
1451 ifp->idev->dev->name);
1455 addrconf_dad_stop(ifp, 1);
1458 /* Join to solicited addr multicast group. */
1460 void addrconf_join_solict(struct net_device *dev, struct in6_addr *addr)
1462 struct in6_addr maddr;
1464 if (dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1465 return;
1467 addrconf_addr_solict_mult(addr, &maddr);
1468 ipv6_dev_mc_inc(dev, &maddr);
1471 void addrconf_leave_solict(struct inet6_dev *idev, struct in6_addr *addr)
1473 struct in6_addr maddr;
1475 if (idev->dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1476 return;
1478 addrconf_addr_solict_mult(addr, &maddr);
1479 __ipv6_dev_mc_dec(idev, &maddr);
1482 static void addrconf_join_anycast(struct inet6_ifaddr *ifp)
1484 struct in6_addr addr;
1485 ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1486 if (ipv6_addr_any(&addr))
1487 return;
1488 ipv6_dev_ac_inc(ifp->idev->dev, &addr);
1491 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp)
1493 struct in6_addr addr;
1494 ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1495 if (ipv6_addr_any(&addr))
1496 return;
1497 __ipv6_dev_ac_dec(ifp->idev, &addr);
1500 static int addrconf_ifid_eui48(u8 *eui, struct net_device *dev)
1502 if (dev->addr_len != ETH_ALEN)
1503 return -1;
1504 memcpy(eui, dev->dev_addr, 3);
1505 memcpy(eui + 5, dev->dev_addr + 3, 3);
1508 * The zSeries OSA network cards can be shared among various
1509 * OS instances, but the OSA cards have only one MAC address.
1510 * This leads to duplicate address conflicts in conjunction
1511 * with IPv6 if more than one instance uses the same card.
1513 * The driver for these cards can deliver a unique 16-bit
1514 * identifier for each instance sharing the same card. It is
1515 * placed instead of 0xFFFE in the interface identifier. The
1516 * "u" bit of the interface identifier is not inverted in this
1517 * case. Hence the resulting interface identifier has local
1518 * scope according to RFC2373.
1520 if (dev->dev_id) {
1521 eui[3] = (dev->dev_id >> 8) & 0xFF;
1522 eui[4] = dev->dev_id & 0xFF;
1523 } else {
1524 eui[3] = 0xFF;
1525 eui[4] = 0xFE;
1526 eui[0] ^= 2;
1528 return 0;
1531 static int addrconf_ifid_arcnet(u8 *eui, struct net_device *dev)
1533 /* XXX: inherit EUI-64 from other interface -- yoshfuji */
1534 if (dev->addr_len != ARCNET_ALEN)
1535 return -1;
1536 memset(eui, 0, 7);
1537 eui[7] = *(u8*)dev->dev_addr;
1538 return 0;
1541 static int addrconf_ifid_infiniband(u8 *eui, struct net_device *dev)
1543 if (dev->addr_len != INFINIBAND_ALEN)
1544 return -1;
1545 memcpy(eui, dev->dev_addr + 12, 8);
1546 eui[0] |= 2;
1547 return 0;
1550 static int __ipv6_isatap_ifid(u8 *eui, __be32 addr)
1552 if (addr == 0)
1553 return -1;
1554 eui[0] = (ipv4_is_zeronet(addr) || ipv4_is_private_10(addr) ||
1555 ipv4_is_loopback(addr) || ipv4_is_linklocal_169(addr) ||
1556 ipv4_is_private_172(addr) || ipv4_is_test_192(addr) ||
1557 ipv4_is_anycast_6to4(addr) || ipv4_is_private_192(addr) ||
1558 ipv4_is_test_198(addr) || ipv4_is_multicast(addr) ||
1559 ipv4_is_lbcast(addr)) ? 0x00 : 0x02;
1560 eui[1] = 0;
1561 eui[2] = 0x5E;
1562 eui[3] = 0xFE;
1563 memcpy(eui + 4, &addr, 4);
1564 return 0;
1567 static int addrconf_ifid_sit(u8 *eui, struct net_device *dev)
1569 if (dev->priv_flags & IFF_ISATAP)
1570 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
1571 return -1;
1574 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev)
1576 switch (dev->type) {
1577 case ARPHRD_ETHER:
1578 case ARPHRD_FDDI:
1579 case ARPHRD_IEEE802_TR:
1580 return addrconf_ifid_eui48(eui, dev);
1581 case ARPHRD_ARCNET:
1582 return addrconf_ifid_arcnet(eui, dev);
1583 case ARPHRD_INFINIBAND:
1584 return addrconf_ifid_infiniband(eui, dev);
1585 case ARPHRD_SIT:
1586 return addrconf_ifid_sit(eui, dev);
1588 return -1;
1591 static int ipv6_inherit_eui64(u8 *eui, struct inet6_dev *idev)
1593 int err = -1;
1594 struct inet6_ifaddr *ifp;
1596 read_lock_bh(&idev->lock);
1597 list_for_each_entry(ifp, &idev->addr_list, if_list) {
1598 if (ifp->scope == IFA_LINK && !(ifp->flags&IFA_F_TENTATIVE)) {
1599 memcpy(eui, ifp->addr.s6_addr+8, 8);
1600 err = 0;
1601 break;
1604 read_unlock_bh(&idev->lock);
1605 return err;
1608 #ifdef CONFIG_IPV6_PRIVACY
1609 /* (re)generation of randomized interface identifier (RFC 3041 3.2, 3.5) */
1610 static int __ipv6_regen_rndid(struct inet6_dev *idev)
1612 regen:
1613 get_random_bytes(idev->rndid, sizeof(idev->rndid));
1614 idev->rndid[0] &= ~0x02;
1617 * <draft-ietf-ipngwg-temp-addresses-v2-00.txt>:
1618 * check if generated address is not inappropriate
1620 * - Reserved subnet anycast (RFC 2526)
1621 * 11111101 11....11 1xxxxxxx
1622 * - ISATAP (RFC4214) 6.1
1623 * 00-00-5E-FE-xx-xx-xx-xx
1624 * - value 0
1625 * - XXX: already assigned to an address on the device
1627 if (idev->rndid[0] == 0xfd &&
1628 (idev->rndid[1]&idev->rndid[2]&idev->rndid[3]&idev->rndid[4]&idev->rndid[5]&idev->rndid[6]) == 0xff &&
1629 (idev->rndid[7]&0x80))
1630 goto regen;
1631 if ((idev->rndid[0]|idev->rndid[1]) == 0) {
1632 if (idev->rndid[2] == 0x5e && idev->rndid[3] == 0xfe)
1633 goto regen;
1634 if ((idev->rndid[2]|idev->rndid[3]|idev->rndid[4]|idev->rndid[5]|idev->rndid[6]|idev->rndid[7]) == 0x00)
1635 goto regen;
1638 return 0;
1641 static void ipv6_regen_rndid(unsigned long data)
1643 struct inet6_dev *idev = (struct inet6_dev *) data;
1644 unsigned long expires;
1646 rcu_read_lock_bh();
1647 write_lock_bh(&idev->lock);
1649 if (idev->dead)
1650 goto out;
1652 if (__ipv6_regen_rndid(idev) < 0)
1653 goto out;
1655 expires = jiffies +
1656 idev->cnf.temp_prefered_lft * HZ -
1657 idev->cnf.regen_max_retry * idev->cnf.dad_transmits * idev->nd_parms->retrans_time -
1658 idev->cnf.max_desync_factor * HZ;
1659 if (time_before(expires, jiffies)) {
1660 printk(KERN_WARNING
1661 "ipv6_regen_rndid(): too short regeneration interval; timer disabled for %s.\n",
1662 idev->dev->name);
1663 goto out;
1666 if (!mod_timer(&idev->regen_timer, expires))
1667 in6_dev_hold(idev);
1669 out:
1670 write_unlock_bh(&idev->lock);
1671 rcu_read_unlock_bh();
1672 in6_dev_put(idev);
1675 static int __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr) {
1676 int ret = 0;
1678 if (tmpaddr && memcmp(idev->rndid, &tmpaddr->s6_addr[8], 8) == 0)
1679 ret = __ipv6_regen_rndid(idev);
1680 return ret;
1682 #endif
1685 * Add prefix route.
1688 static void
1689 addrconf_prefix_route(struct in6_addr *pfx, int plen, struct net_device *dev,
1690 unsigned long expires, u32 flags)
1692 struct fib6_config cfg = {
1693 .fc_table = RT6_TABLE_PREFIX,
1694 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1695 .fc_ifindex = dev->ifindex,
1696 .fc_expires = expires,
1697 .fc_dst_len = plen,
1698 .fc_flags = RTF_UP | flags,
1699 .fc_nlinfo.nl_net = dev_net(dev),
1700 .fc_protocol = RTPROT_KERNEL,
1703 ipv6_addr_copy(&cfg.fc_dst, pfx);
1705 /* Prevent useless cloning on PtP SIT.
1706 This thing is done here expecting that the whole
1707 class of non-broadcast devices need not cloning.
1709 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
1710 if (dev->type == ARPHRD_SIT && (dev->flags & IFF_POINTOPOINT))
1711 cfg.fc_flags |= RTF_NONEXTHOP;
1712 #endif
1714 ip6_route_add(&cfg);
1717 /* Create "default" multicast route to the interface */
1719 static void addrconf_add_mroute(struct net_device *dev)
1721 struct fib6_config cfg = {
1722 .fc_table = RT6_TABLE_LOCAL,
1723 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1724 .fc_ifindex = dev->ifindex,
1725 .fc_dst_len = 8,
1726 .fc_flags = RTF_UP,
1727 .fc_nlinfo.nl_net = dev_net(dev),
1730 ipv6_addr_set(&cfg.fc_dst, htonl(0xFF000000), 0, 0, 0);
1732 ip6_route_add(&cfg);
1735 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
1736 static void sit_route_add(struct net_device *dev)
1738 struct fib6_config cfg = {
1739 .fc_table = RT6_TABLE_MAIN,
1740 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1741 .fc_ifindex = dev->ifindex,
1742 .fc_dst_len = 96,
1743 .fc_flags = RTF_UP | RTF_NONEXTHOP,
1744 .fc_nlinfo.nl_net = dev_net(dev),
1747 /* prefix length - 96 bits "::d.d.d.d" */
1748 ip6_route_add(&cfg);
1750 #endif
1752 static void addrconf_add_lroute(struct net_device *dev)
1754 struct in6_addr addr;
1756 ipv6_addr_set(&addr, htonl(0xFE800000), 0, 0, 0);
1757 addrconf_prefix_route(&addr, 64, dev, 0, 0);
1760 static struct inet6_dev *addrconf_add_dev(struct net_device *dev)
1762 struct inet6_dev *idev;
1764 ASSERT_RTNL();
1766 idev = ipv6_find_idev(dev);
1767 if (!idev)
1768 return ERR_PTR(-ENOBUFS);
1770 if (idev->cnf.disable_ipv6)
1771 return ERR_PTR(-EACCES);
1773 /* Add default multicast route */
1774 addrconf_add_mroute(dev);
1776 /* Add link local route */
1777 addrconf_add_lroute(dev);
1778 return idev;
1781 void addrconf_prefix_rcv(struct net_device *dev, u8 *opt, int len)
1783 struct prefix_info *pinfo;
1784 __u32 valid_lft;
1785 __u32 prefered_lft;
1786 int addr_type;
1787 struct inet6_dev *in6_dev;
1788 struct net *net = dev_net(dev);
1790 pinfo = (struct prefix_info *) opt;
1792 if (len < sizeof(struct prefix_info)) {
1793 ADBG(("addrconf: prefix option too short\n"));
1794 return;
1798 * Validation checks ([ADDRCONF], page 19)
1801 addr_type = ipv6_addr_type(&pinfo->prefix);
1803 if (addr_type & (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL))
1804 return;
1806 valid_lft = ntohl(pinfo->valid);
1807 prefered_lft = ntohl(pinfo->prefered);
1809 if (prefered_lft > valid_lft) {
1810 if (net_ratelimit())
1811 printk(KERN_WARNING "addrconf: prefix option has invalid lifetime\n");
1812 return;
1815 in6_dev = in6_dev_get(dev);
1817 if (in6_dev == NULL) {
1818 if (net_ratelimit())
1819 printk(KERN_DEBUG "addrconf: device %s not configured\n", dev->name);
1820 return;
1824 * Two things going on here:
1825 * 1) Add routes for on-link prefixes
1826 * 2) Configure prefixes with the auto flag set
1829 if (pinfo->onlink) {
1830 struct rt6_info *rt;
1831 unsigned long rt_expires;
1833 /* Avoid arithmetic overflow. Really, we could
1834 * save rt_expires in seconds, likely valid_lft,
1835 * but it would require division in fib gc, that it
1836 * not good.
1838 if (HZ > USER_HZ)
1839 rt_expires = addrconf_timeout_fixup(valid_lft, HZ);
1840 else
1841 rt_expires = addrconf_timeout_fixup(valid_lft, USER_HZ);
1843 if (addrconf_finite_timeout(rt_expires))
1844 rt_expires *= HZ;
1846 rt = rt6_lookup(net, &pinfo->prefix, NULL,
1847 dev->ifindex, 1);
1849 if (rt && addrconf_is_prefix_route(rt)) {
1850 /* Autoconf prefix route */
1851 if (valid_lft == 0) {
1852 ip6_del_rt(rt);
1853 rt = NULL;
1854 } else if (addrconf_finite_timeout(rt_expires)) {
1855 /* not infinity */
1856 rt->rt6i_expires = jiffies + rt_expires;
1857 rt->rt6i_flags |= RTF_EXPIRES;
1858 } else {
1859 rt->rt6i_flags &= ~RTF_EXPIRES;
1860 rt->rt6i_expires = 0;
1862 } else if (valid_lft) {
1863 clock_t expires = 0;
1864 int flags = RTF_ADDRCONF | RTF_PREFIX_RT;
1865 if (addrconf_finite_timeout(rt_expires)) {
1866 /* not infinity */
1867 flags |= RTF_EXPIRES;
1868 expires = jiffies_to_clock_t(rt_expires);
1870 addrconf_prefix_route(&pinfo->prefix, pinfo->prefix_len,
1871 dev, expires, flags);
1873 if (rt)
1874 dst_release(&rt->dst);
1877 /* Try to figure out our local address for this prefix */
1879 if (pinfo->autoconf && in6_dev->cnf.autoconf) {
1880 struct inet6_ifaddr * ifp;
1881 struct in6_addr addr;
1882 int create = 0, update_lft = 0;
1884 if (pinfo->prefix_len == 64) {
1885 memcpy(&addr, &pinfo->prefix, 8);
1886 if (ipv6_generate_eui64(addr.s6_addr + 8, dev) &&
1887 ipv6_inherit_eui64(addr.s6_addr + 8, in6_dev)) {
1888 in6_dev_put(in6_dev);
1889 return;
1891 goto ok;
1893 if (net_ratelimit())
1894 printk(KERN_DEBUG "IPv6 addrconf: prefix with wrong length %d\n",
1895 pinfo->prefix_len);
1896 in6_dev_put(in6_dev);
1897 return;
1901 ifp = ipv6_get_ifaddr(net, &addr, dev, 1);
1903 if (ifp == NULL && valid_lft) {
1904 int max_addresses = in6_dev->cnf.max_addresses;
1905 u32 addr_flags = 0;
1907 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1908 if (in6_dev->cnf.optimistic_dad &&
1909 !net->ipv6.devconf_all->forwarding)
1910 addr_flags = IFA_F_OPTIMISTIC;
1911 #endif
1913 /* Do not allow to create too much of autoconfigured
1914 * addresses; this would be too easy way to crash kernel.
1916 if (!max_addresses ||
1917 ipv6_count_addresses(in6_dev) < max_addresses)
1918 ifp = ipv6_add_addr(in6_dev, &addr, pinfo->prefix_len,
1919 addr_type&IPV6_ADDR_SCOPE_MASK,
1920 addr_flags);
1922 if (!ifp || IS_ERR(ifp)) {
1923 in6_dev_put(in6_dev);
1924 return;
1927 update_lft = create = 1;
1928 ifp->cstamp = jiffies;
1929 addrconf_dad_start(ifp, RTF_ADDRCONF|RTF_PREFIX_RT);
1932 if (ifp) {
1933 int flags;
1934 unsigned long now;
1935 #ifdef CONFIG_IPV6_PRIVACY
1936 struct inet6_ifaddr *ift;
1937 #endif
1938 u32 stored_lft;
1940 /* update lifetime (RFC2462 5.5.3 e) */
1941 spin_lock(&ifp->lock);
1942 now = jiffies;
1943 if (ifp->valid_lft > (now - ifp->tstamp) / HZ)
1944 stored_lft = ifp->valid_lft - (now - ifp->tstamp) / HZ;
1945 else
1946 stored_lft = 0;
1947 if (!update_lft && stored_lft) {
1948 if (valid_lft > MIN_VALID_LIFETIME ||
1949 valid_lft > stored_lft)
1950 update_lft = 1;
1951 else if (stored_lft <= MIN_VALID_LIFETIME) {
1952 /* valid_lft <= stored_lft is always true */
1954 * RFC 4862 Section 5.5.3e:
1955 * "Note that the preferred lifetime of
1956 * the corresponding address is always
1957 * reset to the Preferred Lifetime in
1958 * the received Prefix Information
1959 * option, regardless of whether the
1960 * valid lifetime is also reset or
1961 * ignored."
1963 * So if the preferred lifetime in
1964 * this advertisement is different
1965 * than what we have stored, but the
1966 * valid lifetime is invalid, just
1967 * reset prefered_lft.
1969 * We must set the valid lifetime
1970 * to the stored lifetime since we'll
1971 * be updating the timestamp below,
1972 * else we'll set it back to the
1973 * minumum.
1975 if (prefered_lft != ifp->prefered_lft) {
1976 valid_lft = stored_lft;
1977 update_lft = 1;
1979 } else {
1980 valid_lft = MIN_VALID_LIFETIME;
1981 if (valid_lft < prefered_lft)
1982 prefered_lft = valid_lft;
1983 update_lft = 1;
1987 if (update_lft) {
1988 ifp->valid_lft = valid_lft;
1989 ifp->prefered_lft = prefered_lft;
1990 ifp->tstamp = now;
1991 flags = ifp->flags;
1992 ifp->flags &= ~IFA_F_DEPRECATED;
1993 spin_unlock(&ifp->lock);
1995 if (!(flags&IFA_F_TENTATIVE))
1996 ipv6_ifa_notify(0, ifp);
1997 } else
1998 spin_unlock(&ifp->lock);
2000 #ifdef CONFIG_IPV6_PRIVACY
2001 read_lock_bh(&in6_dev->lock);
2002 /* update all temporary addresses in the list */
2003 list_for_each_entry(ift, &in6_dev->tempaddr_list, tmp_list) {
2005 * When adjusting the lifetimes of an existing
2006 * temporary address, only lower the lifetimes.
2007 * Implementations must not increase the
2008 * lifetimes of an existing temporary address
2009 * when processing a Prefix Information Option.
2011 if (ifp != ift->ifpub)
2012 continue;
2014 spin_lock(&ift->lock);
2015 flags = ift->flags;
2016 if (ift->valid_lft > valid_lft &&
2017 ift->valid_lft - valid_lft > (jiffies - ift->tstamp) / HZ)
2018 ift->valid_lft = valid_lft + (jiffies - ift->tstamp) / HZ;
2019 if (ift->prefered_lft > prefered_lft &&
2020 ift->prefered_lft - prefered_lft > (jiffies - ift->tstamp) / HZ)
2021 ift->prefered_lft = prefered_lft + (jiffies - ift->tstamp) / HZ;
2022 spin_unlock(&ift->lock);
2023 if (!(flags&IFA_F_TENTATIVE))
2024 ipv6_ifa_notify(0, ift);
2027 if ((create || list_empty(&in6_dev->tempaddr_list)) && in6_dev->cnf.use_tempaddr > 0) {
2029 * When a new public address is created as described in [ADDRCONF],
2030 * also create a new temporary address. Also create a temporary
2031 * address if it's enabled but no temporary address currently exists.
2033 read_unlock_bh(&in6_dev->lock);
2034 ipv6_create_tempaddr(ifp, NULL);
2035 } else {
2036 read_unlock_bh(&in6_dev->lock);
2038 #endif
2039 in6_ifa_put(ifp);
2040 addrconf_verify(0);
2043 inet6_prefix_notify(RTM_NEWPREFIX, in6_dev, pinfo);
2044 in6_dev_put(in6_dev);
2048 * Set destination address.
2049 * Special case for SIT interfaces where we create a new "virtual"
2050 * device.
2052 int addrconf_set_dstaddr(struct net *net, void __user *arg)
2054 struct in6_ifreq ireq;
2055 struct net_device *dev;
2056 int err = -EINVAL;
2058 rtnl_lock();
2060 err = -EFAULT;
2061 if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2062 goto err_exit;
2064 dev = __dev_get_by_index(net, ireq.ifr6_ifindex);
2066 err = -ENODEV;
2067 if (dev == NULL)
2068 goto err_exit;
2070 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2071 if (dev->type == ARPHRD_SIT) {
2072 const struct net_device_ops *ops = dev->netdev_ops;
2073 struct ifreq ifr;
2074 struct ip_tunnel_parm p;
2076 err = -EADDRNOTAVAIL;
2077 if (!(ipv6_addr_type(&ireq.ifr6_addr) & IPV6_ADDR_COMPATv4))
2078 goto err_exit;
2080 memset(&p, 0, sizeof(p));
2081 p.iph.daddr = ireq.ifr6_addr.s6_addr32[3];
2082 p.iph.saddr = 0;
2083 p.iph.version = 4;
2084 p.iph.ihl = 5;
2085 p.iph.protocol = IPPROTO_IPV6;
2086 p.iph.ttl = 64;
2087 ifr.ifr_ifru.ifru_data = (__force void __user *)&p;
2089 if (ops->ndo_do_ioctl) {
2090 mm_segment_t oldfs = get_fs();
2092 set_fs(KERNEL_DS);
2093 err = ops->ndo_do_ioctl(dev, &ifr, SIOCADDTUNNEL);
2094 set_fs(oldfs);
2095 } else
2096 err = -EOPNOTSUPP;
2098 if (err == 0) {
2099 err = -ENOBUFS;
2100 dev = __dev_get_by_name(net, p.name);
2101 if (!dev)
2102 goto err_exit;
2103 err = dev_open(dev);
2106 #endif
2108 err_exit:
2109 rtnl_unlock();
2110 return err;
2114 * Manual configuration of address on an interface
2116 static int inet6_addr_add(struct net *net, int ifindex, struct in6_addr *pfx,
2117 unsigned int plen, __u8 ifa_flags, __u32 prefered_lft,
2118 __u32 valid_lft)
2120 struct inet6_ifaddr *ifp;
2121 struct inet6_dev *idev;
2122 struct net_device *dev;
2123 int scope;
2124 u32 flags;
2125 clock_t expires;
2126 unsigned long timeout;
2128 ASSERT_RTNL();
2130 if (plen > 128)
2131 return -EINVAL;
2133 /* check the lifetime */
2134 if (!valid_lft || prefered_lft > valid_lft)
2135 return -EINVAL;
2137 dev = __dev_get_by_index(net, ifindex);
2138 if (!dev)
2139 return -ENODEV;
2141 idev = addrconf_add_dev(dev);
2142 if (IS_ERR(idev))
2143 return PTR_ERR(idev);
2145 scope = ipv6_addr_scope(pfx);
2147 timeout = addrconf_timeout_fixup(valid_lft, HZ);
2148 if (addrconf_finite_timeout(timeout)) {
2149 expires = jiffies_to_clock_t(timeout * HZ);
2150 valid_lft = timeout;
2151 flags = RTF_EXPIRES;
2152 } else {
2153 expires = 0;
2154 flags = 0;
2155 ifa_flags |= IFA_F_PERMANENT;
2158 timeout = addrconf_timeout_fixup(prefered_lft, HZ);
2159 if (addrconf_finite_timeout(timeout)) {
2160 if (timeout == 0)
2161 ifa_flags |= IFA_F_DEPRECATED;
2162 prefered_lft = timeout;
2165 ifp = ipv6_add_addr(idev, pfx, plen, scope, ifa_flags);
2167 if (!IS_ERR(ifp)) {
2168 spin_lock_bh(&ifp->lock);
2169 ifp->valid_lft = valid_lft;
2170 ifp->prefered_lft = prefered_lft;
2171 ifp->tstamp = jiffies;
2172 spin_unlock_bh(&ifp->lock);
2174 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, dev,
2175 expires, flags);
2177 * Note that section 3.1 of RFC 4429 indicates
2178 * that the Optimistic flag should not be set for
2179 * manually configured addresses
2181 addrconf_dad_start(ifp, 0);
2182 in6_ifa_put(ifp);
2183 addrconf_verify(0);
2184 return 0;
2187 return PTR_ERR(ifp);
2190 static int inet6_addr_del(struct net *net, int ifindex, struct in6_addr *pfx,
2191 unsigned int plen)
2193 struct inet6_ifaddr *ifp;
2194 struct inet6_dev *idev;
2195 struct net_device *dev;
2197 if (plen > 128)
2198 return -EINVAL;
2200 dev = __dev_get_by_index(net, ifindex);
2201 if (!dev)
2202 return -ENODEV;
2204 if ((idev = __in6_dev_get(dev)) == NULL)
2205 return -ENXIO;
2207 read_lock_bh(&idev->lock);
2208 list_for_each_entry(ifp, &idev->addr_list, if_list) {
2209 if (ifp->prefix_len == plen &&
2210 ipv6_addr_equal(pfx, &ifp->addr)) {
2211 in6_ifa_hold(ifp);
2212 read_unlock_bh(&idev->lock);
2214 ipv6_del_addr(ifp);
2216 /* If the last address is deleted administratively,
2217 disable IPv6 on this interface.
2219 if (list_empty(&idev->addr_list))
2220 addrconf_ifdown(idev->dev, 1);
2221 return 0;
2224 read_unlock_bh(&idev->lock);
2225 return -EADDRNOTAVAIL;
2229 int addrconf_add_ifaddr(struct net *net, void __user *arg)
2231 struct in6_ifreq ireq;
2232 int err;
2234 if (!capable(CAP_NET_ADMIN))
2235 return -EPERM;
2237 if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2238 return -EFAULT;
2240 rtnl_lock();
2241 err = inet6_addr_add(net, ireq.ifr6_ifindex, &ireq.ifr6_addr,
2242 ireq.ifr6_prefixlen, IFA_F_PERMANENT,
2243 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
2244 rtnl_unlock();
2245 return err;
2248 int addrconf_del_ifaddr(struct net *net, void __user *arg)
2250 struct in6_ifreq ireq;
2251 int err;
2253 if (!capable(CAP_NET_ADMIN))
2254 return -EPERM;
2256 if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2257 return -EFAULT;
2259 rtnl_lock();
2260 err = inet6_addr_del(net, ireq.ifr6_ifindex, &ireq.ifr6_addr,
2261 ireq.ifr6_prefixlen);
2262 rtnl_unlock();
2263 return err;
2266 static void add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
2267 int plen, int scope)
2269 struct inet6_ifaddr *ifp;
2271 ifp = ipv6_add_addr(idev, addr, plen, scope, IFA_F_PERMANENT);
2272 if (!IS_ERR(ifp)) {
2273 spin_lock_bh(&ifp->lock);
2274 ifp->flags &= ~IFA_F_TENTATIVE;
2275 spin_unlock_bh(&ifp->lock);
2276 ipv6_ifa_notify(RTM_NEWADDR, ifp);
2277 in6_ifa_put(ifp);
2281 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2282 static void sit_add_v4_addrs(struct inet6_dev *idev)
2284 struct in6_addr addr;
2285 struct net_device *dev;
2286 struct net *net = dev_net(idev->dev);
2287 int scope;
2289 ASSERT_RTNL();
2291 memset(&addr, 0, sizeof(struct in6_addr));
2292 memcpy(&addr.s6_addr32[3], idev->dev->dev_addr, 4);
2294 if (idev->dev->flags&IFF_POINTOPOINT) {
2295 addr.s6_addr32[0] = htonl(0xfe800000);
2296 scope = IFA_LINK;
2297 } else {
2298 scope = IPV6_ADDR_COMPATv4;
2301 if (addr.s6_addr32[3]) {
2302 add_addr(idev, &addr, 128, scope);
2303 return;
2306 for_each_netdev(net, dev) {
2307 struct in_device * in_dev = __in_dev_get_rtnl(dev);
2308 if (in_dev && (dev->flags & IFF_UP)) {
2309 struct in_ifaddr * ifa;
2311 int flag = scope;
2313 for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) {
2314 int plen;
2316 addr.s6_addr32[3] = ifa->ifa_local;
2318 if (ifa->ifa_scope == RT_SCOPE_LINK)
2319 continue;
2320 if (ifa->ifa_scope >= RT_SCOPE_HOST) {
2321 if (idev->dev->flags&IFF_POINTOPOINT)
2322 continue;
2323 flag |= IFA_HOST;
2325 if (idev->dev->flags&IFF_POINTOPOINT)
2326 plen = 64;
2327 else
2328 plen = 96;
2330 add_addr(idev, &addr, plen, flag);
2335 #endif
2337 static void init_loopback(struct net_device *dev)
2339 struct inet6_dev *idev;
2341 /* ::1 */
2343 ASSERT_RTNL();
2345 if ((idev = ipv6_find_idev(dev)) == NULL) {
2346 printk(KERN_DEBUG "init loopback: add_dev failed\n");
2347 return;
2350 add_addr(idev, &in6addr_loopback, 128, IFA_HOST);
2353 static void addrconf_add_linklocal(struct inet6_dev *idev, struct in6_addr *addr)
2355 struct inet6_ifaddr * ifp;
2356 u32 addr_flags = IFA_F_PERMANENT;
2358 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2359 if (idev->cnf.optimistic_dad &&
2360 !dev_net(idev->dev)->ipv6.devconf_all->forwarding)
2361 addr_flags |= IFA_F_OPTIMISTIC;
2362 #endif
2365 ifp = ipv6_add_addr(idev, addr, 64, IFA_LINK, addr_flags);
2366 if (!IS_ERR(ifp)) {
2367 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, idev->dev, 0, 0);
2368 addrconf_dad_start(ifp, 0);
2369 in6_ifa_put(ifp);
2373 static void addrconf_dev_config(struct net_device *dev)
2375 struct in6_addr addr;
2376 struct inet6_dev * idev;
2378 ASSERT_RTNL();
2380 if ((dev->type != ARPHRD_ETHER) &&
2381 (dev->type != ARPHRD_FDDI) &&
2382 (dev->type != ARPHRD_IEEE802_TR) &&
2383 (dev->type != ARPHRD_ARCNET) &&
2384 (dev->type != ARPHRD_INFINIBAND)) {
2385 /* Alas, we support only Ethernet autoconfiguration. */
2386 return;
2389 idev = addrconf_add_dev(dev);
2390 if (IS_ERR(idev))
2391 return;
2393 memset(&addr, 0, sizeof(struct in6_addr));
2394 addr.s6_addr32[0] = htonl(0xFE800000);
2396 if (ipv6_generate_eui64(addr.s6_addr + 8, dev) == 0)
2397 addrconf_add_linklocal(idev, &addr);
2400 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2401 static void addrconf_sit_config(struct net_device *dev)
2403 struct inet6_dev *idev;
2405 ASSERT_RTNL();
2408 * Configure the tunnel with one of our IPv4
2409 * addresses... we should configure all of
2410 * our v4 addrs in the tunnel
2413 if ((idev = ipv6_find_idev(dev)) == NULL) {
2414 printk(KERN_DEBUG "init sit: add_dev failed\n");
2415 return;
2418 if (dev->priv_flags & IFF_ISATAP) {
2419 struct in6_addr addr;
2421 ipv6_addr_set(&addr, htonl(0xFE800000), 0, 0, 0);
2422 addrconf_prefix_route(&addr, 64, dev, 0, 0);
2423 if (!ipv6_generate_eui64(addr.s6_addr + 8, dev))
2424 addrconf_add_linklocal(idev, &addr);
2425 return;
2428 sit_add_v4_addrs(idev);
2430 if (dev->flags&IFF_POINTOPOINT) {
2431 addrconf_add_mroute(dev);
2432 addrconf_add_lroute(dev);
2433 } else
2434 sit_route_add(dev);
2436 #endif
2438 static inline int
2439 ipv6_inherit_linklocal(struct inet6_dev *idev, struct net_device *link_dev)
2441 struct in6_addr lladdr;
2443 if (!ipv6_get_lladdr(link_dev, &lladdr, IFA_F_TENTATIVE)) {
2444 addrconf_add_linklocal(idev, &lladdr);
2445 return 0;
2447 return -1;
2450 static void ip6_tnl_add_linklocal(struct inet6_dev *idev)
2452 struct net_device *link_dev;
2453 struct net *net = dev_net(idev->dev);
2455 /* first try to inherit the link-local address from the link device */
2456 if (idev->dev->iflink &&
2457 (link_dev = __dev_get_by_index(net, idev->dev->iflink))) {
2458 if (!ipv6_inherit_linklocal(idev, link_dev))
2459 return;
2461 /* then try to inherit it from any device */
2462 for_each_netdev(net, link_dev) {
2463 if (!ipv6_inherit_linklocal(idev, link_dev))
2464 return;
2466 printk(KERN_DEBUG "init ip6-ip6: add_linklocal failed\n");
2470 * Autoconfigure tunnel with a link-local address so routing protocols,
2471 * DHCPv6, MLD etc. can be run over the virtual link
2474 static void addrconf_ip6_tnl_config(struct net_device *dev)
2476 struct inet6_dev *idev;
2478 ASSERT_RTNL();
2480 idev = addrconf_add_dev(dev);
2481 if (IS_ERR(idev)) {
2482 printk(KERN_DEBUG "init ip6-ip6: add_dev failed\n");
2483 return;
2485 ip6_tnl_add_linklocal(idev);
2488 static int addrconf_notify(struct notifier_block *this, unsigned long event,
2489 void * data)
2491 struct net_device *dev = (struct net_device *) data;
2492 struct inet6_dev *idev = __in6_dev_get(dev);
2493 int run_pending = 0;
2494 int err;
2496 switch (event) {
2497 case NETDEV_REGISTER:
2498 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2499 idev = ipv6_add_dev(dev);
2500 if (!idev)
2501 return notifier_from_errno(-ENOMEM);
2503 break;
2505 case NETDEV_UP:
2506 case NETDEV_CHANGE:
2507 if (dev->flags & IFF_SLAVE)
2508 break;
2510 if (event == NETDEV_UP) {
2511 if (!addrconf_qdisc_ok(dev)) {
2512 /* device is not ready yet. */
2513 printk(KERN_INFO
2514 "ADDRCONF(NETDEV_UP): %s: "
2515 "link is not ready\n",
2516 dev->name);
2517 break;
2520 if (!idev && dev->mtu >= IPV6_MIN_MTU)
2521 idev = ipv6_add_dev(dev);
2523 if (idev) {
2524 idev->if_flags |= IF_READY;
2525 run_pending = 1;
2527 } else {
2528 if (!addrconf_qdisc_ok(dev)) {
2529 /* device is still not ready. */
2530 break;
2533 if (idev) {
2534 if (idev->if_flags & IF_READY)
2535 /* device is already configured. */
2536 break;
2537 idev->if_flags |= IF_READY;
2540 printk(KERN_INFO
2541 "ADDRCONF(NETDEV_CHANGE): %s: "
2542 "link becomes ready\n",
2543 dev->name);
2545 run_pending = 1;
2548 switch (dev->type) {
2549 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2550 case ARPHRD_SIT:
2551 addrconf_sit_config(dev);
2552 break;
2553 #endif
2554 case ARPHRD_TUNNEL6:
2555 addrconf_ip6_tnl_config(dev);
2556 break;
2557 case ARPHRD_LOOPBACK:
2558 init_loopback(dev);
2559 break;
2561 default:
2562 addrconf_dev_config(dev);
2563 break;
2566 if (idev) {
2567 if (run_pending)
2568 addrconf_dad_run(idev);
2571 * If the MTU changed during the interface down,
2572 * when the interface up, the changed MTU must be
2573 * reflected in the idev as well as routers.
2575 if (idev->cnf.mtu6 != dev->mtu &&
2576 dev->mtu >= IPV6_MIN_MTU) {
2577 rt6_mtu_change(dev, dev->mtu);
2578 idev->cnf.mtu6 = dev->mtu;
2580 idev->tstamp = jiffies;
2581 inet6_ifinfo_notify(RTM_NEWLINK, idev);
2584 * If the changed mtu during down is lower than
2585 * IPV6_MIN_MTU stop IPv6 on this interface.
2587 if (dev->mtu < IPV6_MIN_MTU)
2588 addrconf_ifdown(dev, 1);
2590 break;
2592 case NETDEV_CHANGEMTU:
2593 if (idev && dev->mtu >= IPV6_MIN_MTU) {
2594 rt6_mtu_change(dev, dev->mtu);
2595 idev->cnf.mtu6 = dev->mtu;
2596 break;
2599 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2600 idev = ipv6_add_dev(dev);
2601 if (idev)
2602 break;
2606 * MTU falled under IPV6_MIN_MTU.
2607 * Stop IPv6 on this interface.
2610 case NETDEV_DOWN:
2611 case NETDEV_UNREGISTER:
2613 * Remove all addresses from this interface.
2615 addrconf_ifdown(dev, event != NETDEV_DOWN);
2616 break;
2618 case NETDEV_CHANGENAME:
2619 if (idev) {
2620 snmp6_unregister_dev(idev);
2621 addrconf_sysctl_unregister(idev);
2622 addrconf_sysctl_register(idev);
2623 err = snmp6_register_dev(idev);
2624 if (err)
2625 return notifier_from_errno(err);
2627 break;
2629 case NETDEV_PRE_TYPE_CHANGE:
2630 case NETDEV_POST_TYPE_CHANGE:
2631 addrconf_type_change(dev, event);
2632 break;
2635 return NOTIFY_OK;
2639 * addrconf module should be notified of a device going up
2641 static struct notifier_block ipv6_dev_notf = {
2642 .notifier_call = addrconf_notify,
2645 static void addrconf_type_change(struct net_device *dev, unsigned long event)
2647 struct inet6_dev *idev;
2648 ASSERT_RTNL();
2650 idev = __in6_dev_get(dev);
2652 if (event == NETDEV_POST_TYPE_CHANGE)
2653 ipv6_mc_remap(idev);
2654 else if (event == NETDEV_PRE_TYPE_CHANGE)
2655 ipv6_mc_unmap(idev);
2658 static int addrconf_ifdown(struct net_device *dev, int how)
2660 struct net *net = dev_net(dev);
2661 struct inet6_dev *idev;
2662 struct inet6_ifaddr *ifa;
2663 LIST_HEAD(keep_list);
2664 int state;
2666 ASSERT_RTNL();
2668 rt6_ifdown(net, dev);
2669 neigh_ifdown(&nd_tbl, dev);
2671 idev = __in6_dev_get(dev);
2672 if (idev == NULL)
2673 return -ENODEV;
2676 * Step 1: remove reference to ipv6 device from parent device.
2677 * Do not dev_put!
2679 if (how) {
2680 idev->dead = 1;
2682 /* protected by rtnl_lock */
2683 rcu_assign_pointer(dev->ip6_ptr, NULL);
2685 /* Step 1.5: remove snmp6 entry */
2686 snmp6_unregister_dev(idev);
2690 write_lock_bh(&idev->lock);
2692 /* Step 2: clear flags for stateless addrconf */
2693 if (!how)
2694 idev->if_flags &= ~(IF_RS_SENT|IF_RA_RCVD|IF_READY);
2696 #ifdef CONFIG_IPV6_PRIVACY
2697 if (how && del_timer(&idev->regen_timer))
2698 in6_dev_put(idev);
2700 /* Step 3: clear tempaddr list */
2701 while (!list_empty(&idev->tempaddr_list)) {
2702 ifa = list_first_entry(&idev->tempaddr_list,
2703 struct inet6_ifaddr, tmp_list);
2704 list_del(&ifa->tmp_list);
2705 write_unlock_bh(&idev->lock);
2706 spin_lock_bh(&ifa->lock);
2708 if (ifa->ifpub) {
2709 in6_ifa_put(ifa->ifpub);
2710 ifa->ifpub = NULL;
2712 spin_unlock_bh(&ifa->lock);
2713 in6_ifa_put(ifa);
2714 write_lock_bh(&idev->lock);
2716 #endif
2718 while (!list_empty(&idev->addr_list)) {
2719 ifa = list_first_entry(&idev->addr_list,
2720 struct inet6_ifaddr, if_list);
2721 addrconf_del_timer(ifa);
2723 /* If just doing link down, and address is permanent
2724 and not link-local, then retain it. */
2725 if (!how &&
2726 (ifa->flags&IFA_F_PERMANENT) &&
2727 !(ipv6_addr_type(&ifa->addr) & IPV6_ADDR_LINKLOCAL)) {
2728 list_move_tail(&ifa->if_list, &keep_list);
2730 /* If not doing DAD on this address, just keep it. */
2731 if ((dev->flags&(IFF_NOARP|IFF_LOOPBACK)) ||
2732 idev->cnf.accept_dad <= 0 ||
2733 (ifa->flags & IFA_F_NODAD))
2734 continue;
2736 /* If it was tentative already, no need to notify */
2737 if (ifa->flags & IFA_F_TENTATIVE)
2738 continue;
2740 /* Flag it for later restoration when link comes up */
2741 ifa->flags |= IFA_F_TENTATIVE;
2742 ifa->state = INET6_IFADDR_STATE_DAD;
2743 } else {
2744 list_del(&ifa->if_list);
2746 /* clear hash table */
2747 spin_lock_bh(&addrconf_hash_lock);
2748 hlist_del_init_rcu(&ifa->addr_lst);
2749 spin_unlock_bh(&addrconf_hash_lock);
2751 write_unlock_bh(&idev->lock);
2752 spin_lock_bh(&ifa->state_lock);
2753 state = ifa->state;
2754 ifa->state = INET6_IFADDR_STATE_DEAD;
2755 spin_unlock_bh(&ifa->state_lock);
2757 if (state == INET6_IFADDR_STATE_DEAD) {
2758 in6_ifa_put(ifa);
2759 } else {
2760 __ipv6_ifa_notify(RTM_DELADDR, ifa);
2761 atomic_notifier_call_chain(&inet6addr_chain,
2762 NETDEV_DOWN, ifa);
2764 write_lock_bh(&idev->lock);
2768 list_splice(&keep_list, &idev->addr_list);
2770 write_unlock_bh(&idev->lock);
2772 /* Step 5: Discard multicast list */
2773 if (how)
2774 ipv6_mc_destroy_dev(idev);
2775 else
2776 ipv6_mc_down(idev);
2778 idev->tstamp = jiffies;
2780 /* Last: Shot the device (if unregistered) */
2781 if (how) {
2782 addrconf_sysctl_unregister(idev);
2783 neigh_parms_release(&nd_tbl, idev->nd_parms);
2784 neigh_ifdown(&nd_tbl, dev);
2785 in6_dev_put(idev);
2787 return 0;
2790 static void addrconf_rs_timer(unsigned long data)
2792 struct inet6_ifaddr *ifp = (struct inet6_ifaddr *) data;
2793 struct inet6_dev *idev = ifp->idev;
2795 read_lock(&idev->lock);
2796 if (idev->dead || !(idev->if_flags & IF_READY))
2797 goto out;
2799 if (idev->cnf.forwarding)
2800 goto out;
2802 /* Announcement received after solicitation was sent */
2803 if (idev->if_flags & IF_RA_RCVD)
2804 goto out;
2806 spin_lock(&ifp->lock);
2807 if (ifp->probes++ < idev->cnf.rtr_solicits) {
2808 /* The wait after the last probe can be shorter */
2809 addrconf_mod_timer(ifp, AC_RS,
2810 (ifp->probes == idev->cnf.rtr_solicits) ?
2811 idev->cnf.rtr_solicit_delay :
2812 idev->cnf.rtr_solicit_interval);
2813 spin_unlock(&ifp->lock);
2815 ndisc_send_rs(idev->dev, &ifp->addr, &in6addr_linklocal_allrouters);
2816 } else {
2817 spin_unlock(&ifp->lock);
2819 * Note: we do not support deprecated "all on-link"
2820 * assumption any longer.
2822 printk(KERN_DEBUG "%s: no IPv6 routers present\n",
2823 idev->dev->name);
2826 out:
2827 read_unlock(&idev->lock);
2828 in6_ifa_put(ifp);
2832 * Duplicate Address Detection
2834 static void addrconf_dad_kick(struct inet6_ifaddr *ifp)
2836 unsigned long rand_num;
2837 struct inet6_dev *idev = ifp->idev;
2839 if (ifp->flags & IFA_F_OPTIMISTIC)
2840 rand_num = 0;
2841 else
2842 rand_num = net_random() % (idev->cnf.rtr_solicit_delay ? : 1);
2844 ifp->probes = idev->cnf.dad_transmits;
2845 addrconf_mod_timer(ifp, AC_DAD, rand_num);
2848 static void addrconf_dad_start(struct inet6_ifaddr *ifp, u32 flags)
2850 struct inet6_dev *idev = ifp->idev;
2851 struct net_device *dev = idev->dev;
2853 addrconf_join_solict(dev, &ifp->addr);
2855 net_srandom(ifp->addr.s6_addr32[3]);
2857 read_lock_bh(&idev->lock);
2858 spin_lock(&ifp->lock);
2859 if (ifp->state == INET6_IFADDR_STATE_DEAD)
2860 goto out;
2862 if (dev->flags&(IFF_NOARP|IFF_LOOPBACK) ||
2863 idev->cnf.accept_dad < 1 ||
2864 !(ifp->flags&IFA_F_TENTATIVE) ||
2865 ifp->flags & IFA_F_NODAD) {
2866 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
2867 spin_unlock(&ifp->lock);
2868 read_unlock_bh(&idev->lock);
2870 addrconf_dad_completed(ifp);
2871 return;
2874 if (!(idev->if_flags & IF_READY)) {
2875 spin_unlock(&ifp->lock);
2876 read_unlock_bh(&idev->lock);
2878 * If the device is not ready:
2879 * - keep it tentative if it is a permanent address.
2880 * - otherwise, kill it.
2882 in6_ifa_hold(ifp);
2883 addrconf_dad_stop(ifp, 0);
2884 return;
2888 * Optimistic nodes can start receiving
2889 * Frames right away
2891 if (ifp->flags & IFA_F_OPTIMISTIC)
2892 ip6_ins_rt(ifp->rt);
2894 addrconf_dad_kick(ifp);
2895 out:
2896 spin_unlock(&ifp->lock);
2897 read_unlock_bh(&idev->lock);
2900 static void addrconf_dad_timer(unsigned long data)
2902 struct inet6_ifaddr *ifp = (struct inet6_ifaddr *) data;
2903 struct inet6_dev *idev = ifp->idev;
2904 struct in6_addr mcaddr;
2906 if (!ifp->probes && addrconf_dad_end(ifp))
2907 goto out;
2909 read_lock(&idev->lock);
2910 if (idev->dead || !(idev->if_flags & IF_READY)) {
2911 read_unlock(&idev->lock);
2912 goto out;
2915 spin_lock(&ifp->lock);
2916 if (ifp->state == INET6_IFADDR_STATE_DEAD) {
2917 spin_unlock(&ifp->lock);
2918 read_unlock(&idev->lock);
2919 goto out;
2922 if (ifp->probes == 0) {
2924 * DAD was successful
2927 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
2928 spin_unlock(&ifp->lock);
2929 read_unlock(&idev->lock);
2931 addrconf_dad_completed(ifp);
2933 goto out;
2936 ifp->probes--;
2937 addrconf_mod_timer(ifp, AC_DAD, ifp->idev->nd_parms->retrans_time);
2938 spin_unlock(&ifp->lock);
2939 read_unlock(&idev->lock);
2941 /* send a neighbour solicitation for our addr */
2942 addrconf_addr_solict_mult(&ifp->addr, &mcaddr);
2943 ndisc_send_ns(ifp->idev->dev, NULL, &ifp->addr, &mcaddr, &in6addr_any);
2944 out:
2945 in6_ifa_put(ifp);
2948 static void addrconf_dad_completed(struct inet6_ifaddr *ifp)
2950 struct net_device *dev = ifp->idev->dev;
2953 * Configure the address for reception. Now it is valid.
2956 ipv6_ifa_notify(RTM_NEWADDR, ifp);
2958 /* If added prefix is link local and forwarding is off,
2959 start sending router solicitations.
2962 if ((ifp->idev->cnf.forwarding == 0 ||
2963 ifp->idev->cnf.forwarding == 2) &&
2964 ifp->idev->cnf.rtr_solicits > 0 &&
2965 (dev->flags&IFF_LOOPBACK) == 0 &&
2966 (ipv6_addr_type(&ifp->addr) & IPV6_ADDR_LINKLOCAL)) {
2968 * If a host as already performed a random delay
2969 * [...] as part of DAD [...] there is no need
2970 * to delay again before sending the first RS
2972 ndisc_send_rs(ifp->idev->dev, &ifp->addr, &in6addr_linklocal_allrouters);
2974 spin_lock_bh(&ifp->lock);
2975 ifp->probes = 1;
2976 ifp->idev->if_flags |= IF_RS_SENT;
2977 addrconf_mod_timer(ifp, AC_RS, ifp->idev->cnf.rtr_solicit_interval);
2978 spin_unlock_bh(&ifp->lock);
2982 static void addrconf_dad_run(struct inet6_dev *idev)
2984 struct inet6_ifaddr *ifp;
2986 read_lock_bh(&idev->lock);
2987 list_for_each_entry(ifp, &idev->addr_list, if_list) {
2988 spin_lock(&ifp->lock);
2989 if (ifp->flags & IFA_F_TENTATIVE &&
2990 ifp->state == INET6_IFADDR_STATE_DAD)
2991 addrconf_dad_kick(ifp);
2992 spin_unlock(&ifp->lock);
2994 read_unlock_bh(&idev->lock);
2997 #ifdef CONFIG_PROC_FS
2998 struct if6_iter_state {
2999 struct seq_net_private p;
3000 int bucket;
3003 static struct inet6_ifaddr *if6_get_first(struct seq_file *seq)
3005 struct inet6_ifaddr *ifa = NULL;
3006 struct if6_iter_state *state = seq->private;
3007 struct net *net = seq_file_net(seq);
3009 for (state->bucket = 0; state->bucket < IN6_ADDR_HSIZE; ++state->bucket) {
3010 struct hlist_node *n;
3011 hlist_for_each_entry_rcu_bh(ifa, n, &inet6_addr_lst[state->bucket],
3012 addr_lst)
3013 if (net_eq(dev_net(ifa->idev->dev), net))
3014 return ifa;
3016 return NULL;
3019 static struct inet6_ifaddr *if6_get_next(struct seq_file *seq,
3020 struct inet6_ifaddr *ifa)
3022 struct if6_iter_state *state = seq->private;
3023 struct net *net = seq_file_net(seq);
3024 struct hlist_node *n = &ifa->addr_lst;
3026 hlist_for_each_entry_continue_rcu_bh(ifa, n, addr_lst)
3027 if (net_eq(dev_net(ifa->idev->dev), net))
3028 return ifa;
3030 while (++state->bucket < IN6_ADDR_HSIZE) {
3031 hlist_for_each_entry_rcu_bh(ifa, n,
3032 &inet6_addr_lst[state->bucket], addr_lst) {
3033 if (net_eq(dev_net(ifa->idev->dev), net))
3034 return ifa;
3038 return NULL;
3041 static struct inet6_ifaddr *if6_get_idx(struct seq_file *seq, loff_t pos)
3043 struct inet6_ifaddr *ifa = if6_get_first(seq);
3045 if (ifa)
3046 while (pos && (ifa = if6_get_next(seq, ifa)) != NULL)
3047 --pos;
3048 return pos ? NULL : ifa;
3051 static void *if6_seq_start(struct seq_file *seq, loff_t *pos)
3052 __acquires(rcu_bh)
3054 rcu_read_lock_bh();
3055 return if6_get_idx(seq, *pos);
3058 static void *if6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
3060 struct inet6_ifaddr *ifa;
3062 ifa = if6_get_next(seq, v);
3063 ++*pos;
3064 return ifa;
3067 static void if6_seq_stop(struct seq_file *seq, void *v)
3068 __releases(rcu_bh)
3070 rcu_read_unlock_bh();
3073 static int if6_seq_show(struct seq_file *seq, void *v)
3075 struct inet6_ifaddr *ifp = (struct inet6_ifaddr *)v;
3076 seq_printf(seq, "%pi6 %02x %02x %02x %02x %8s\n",
3077 &ifp->addr,
3078 ifp->idev->dev->ifindex,
3079 ifp->prefix_len,
3080 ifp->scope,
3081 ifp->flags,
3082 ifp->idev->dev->name);
3083 return 0;
3086 static const struct seq_operations if6_seq_ops = {
3087 .start = if6_seq_start,
3088 .next = if6_seq_next,
3089 .show = if6_seq_show,
3090 .stop = if6_seq_stop,
3093 static int if6_seq_open(struct inode *inode, struct file *file)
3095 return seq_open_net(inode, file, &if6_seq_ops,
3096 sizeof(struct if6_iter_state));
3099 static const struct file_operations if6_fops = {
3100 .owner = THIS_MODULE,
3101 .open = if6_seq_open,
3102 .read = seq_read,
3103 .llseek = seq_lseek,
3104 .release = seq_release_net,
3107 static int __net_init if6_proc_net_init(struct net *net)
3109 if (!proc_net_fops_create(net, "if_inet6", S_IRUGO, &if6_fops))
3110 return -ENOMEM;
3111 return 0;
3114 static void __net_exit if6_proc_net_exit(struct net *net)
3116 proc_net_remove(net, "if_inet6");
3119 static struct pernet_operations if6_proc_net_ops = {
3120 .init = if6_proc_net_init,
3121 .exit = if6_proc_net_exit,
3124 int __init if6_proc_init(void)
3126 return register_pernet_subsys(&if6_proc_net_ops);
3129 void if6_proc_exit(void)
3131 unregister_pernet_subsys(&if6_proc_net_ops);
3133 #endif /* CONFIG_PROC_FS */
3135 #if defined(CONFIG_IPV6_MIP6) || defined(CONFIG_IPV6_MIP6_MODULE)
3136 /* Check if address is a home address configured on any interface. */
3137 int ipv6_chk_home_addr(struct net *net, struct in6_addr *addr)
3139 int ret = 0;
3140 struct inet6_ifaddr *ifp = NULL;
3141 struct hlist_node *n;
3142 unsigned int hash = ipv6_addr_hash(addr);
3144 rcu_read_lock_bh();
3145 hlist_for_each_entry_rcu_bh(ifp, n, &inet6_addr_lst[hash], addr_lst) {
3146 if (!net_eq(dev_net(ifp->idev->dev), net))
3147 continue;
3148 if (ipv6_addr_equal(&ifp->addr, addr) &&
3149 (ifp->flags & IFA_F_HOMEADDRESS)) {
3150 ret = 1;
3151 break;
3154 rcu_read_unlock_bh();
3155 return ret;
3157 #endif
3160 * Periodic address status verification
3163 static void addrconf_verify(unsigned long foo)
3165 unsigned long now, next, next_sec, next_sched;
3166 struct inet6_ifaddr *ifp;
3167 struct hlist_node *node;
3168 int i;
3170 rcu_read_lock_bh();
3171 spin_lock(&addrconf_verify_lock);
3172 now = jiffies;
3173 next = round_jiffies_up(now + ADDR_CHECK_FREQUENCY);
3175 del_timer(&addr_chk_timer);
3177 for (i = 0; i < IN6_ADDR_HSIZE; i++) {
3178 restart:
3179 hlist_for_each_entry_rcu_bh(ifp, node,
3180 &inet6_addr_lst[i], addr_lst) {
3181 unsigned long age;
3183 if (ifp->flags & IFA_F_PERMANENT)
3184 continue;
3186 spin_lock(&ifp->lock);
3187 /* We try to batch several events at once. */
3188 age = (now - ifp->tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
3190 if (ifp->valid_lft != INFINITY_LIFE_TIME &&
3191 age >= ifp->valid_lft) {
3192 spin_unlock(&ifp->lock);
3193 in6_ifa_hold(ifp);
3194 ipv6_del_addr(ifp);
3195 goto restart;
3196 } else if (ifp->prefered_lft == INFINITY_LIFE_TIME) {
3197 spin_unlock(&ifp->lock);
3198 continue;
3199 } else if (age >= ifp->prefered_lft) {
3200 /* jiffies - ifp->tstamp > age >= ifp->prefered_lft */
3201 int deprecate = 0;
3203 if (!(ifp->flags&IFA_F_DEPRECATED)) {
3204 deprecate = 1;
3205 ifp->flags |= IFA_F_DEPRECATED;
3208 if (time_before(ifp->tstamp + ifp->valid_lft * HZ, next))
3209 next = ifp->tstamp + ifp->valid_lft * HZ;
3211 spin_unlock(&ifp->lock);
3213 if (deprecate) {
3214 in6_ifa_hold(ifp);
3216 ipv6_ifa_notify(0, ifp);
3217 in6_ifa_put(ifp);
3218 goto restart;
3220 #ifdef CONFIG_IPV6_PRIVACY
3221 } else if ((ifp->flags&IFA_F_TEMPORARY) &&
3222 !(ifp->flags&IFA_F_TENTATIVE)) {
3223 unsigned long regen_advance = ifp->idev->cnf.regen_max_retry *
3224 ifp->idev->cnf.dad_transmits *
3225 ifp->idev->nd_parms->retrans_time / HZ;
3227 if (age >= ifp->prefered_lft - regen_advance) {
3228 struct inet6_ifaddr *ifpub = ifp->ifpub;
3229 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3230 next = ifp->tstamp + ifp->prefered_lft * HZ;
3231 if (!ifp->regen_count && ifpub) {
3232 ifp->regen_count++;
3233 in6_ifa_hold(ifp);
3234 in6_ifa_hold(ifpub);
3235 spin_unlock(&ifp->lock);
3237 spin_lock(&ifpub->lock);
3238 ifpub->regen_count = 0;
3239 spin_unlock(&ifpub->lock);
3240 ipv6_create_tempaddr(ifpub, ifp);
3241 in6_ifa_put(ifpub);
3242 in6_ifa_put(ifp);
3243 goto restart;
3245 } else if (time_before(ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ, next))
3246 next = ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ;
3247 spin_unlock(&ifp->lock);
3248 #endif
3249 } else {
3250 /* ifp->prefered_lft <= ifp->valid_lft */
3251 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3252 next = ifp->tstamp + ifp->prefered_lft * HZ;
3253 spin_unlock(&ifp->lock);
3258 next_sec = round_jiffies_up(next);
3259 next_sched = next;
3261 /* If rounded timeout is accurate enough, accept it. */
3262 if (time_before(next_sec, next + ADDRCONF_TIMER_FUZZ))
3263 next_sched = next_sec;
3265 /* And minimum interval is ADDRCONF_TIMER_FUZZ_MAX. */
3266 if (time_before(next_sched, jiffies + ADDRCONF_TIMER_FUZZ_MAX))
3267 next_sched = jiffies + ADDRCONF_TIMER_FUZZ_MAX;
3269 ADBG((KERN_DEBUG "now = %lu, schedule = %lu, rounded schedule = %lu => %lu\n",
3270 now, next, next_sec, next_sched));
3272 addr_chk_timer.expires = next_sched;
3273 add_timer(&addr_chk_timer);
3274 spin_unlock(&addrconf_verify_lock);
3275 rcu_read_unlock_bh();
3278 static struct in6_addr *extract_addr(struct nlattr *addr, struct nlattr *local)
3280 struct in6_addr *pfx = NULL;
3282 if (addr)
3283 pfx = nla_data(addr);
3285 if (local) {
3286 if (pfx && nla_memcmp(local, pfx, sizeof(*pfx)))
3287 pfx = NULL;
3288 else
3289 pfx = nla_data(local);
3292 return pfx;
3295 static const struct nla_policy ifa_ipv6_policy[IFA_MAX+1] = {
3296 [IFA_ADDRESS] = { .len = sizeof(struct in6_addr) },
3297 [IFA_LOCAL] = { .len = sizeof(struct in6_addr) },
3298 [IFA_CACHEINFO] = { .len = sizeof(struct ifa_cacheinfo) },
3301 static int
3302 inet6_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
3304 struct net *net = sock_net(skb->sk);
3305 struct ifaddrmsg *ifm;
3306 struct nlattr *tb[IFA_MAX+1];
3307 struct in6_addr *pfx;
3308 int err;
3310 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3311 if (err < 0)
3312 return err;
3314 ifm = nlmsg_data(nlh);
3315 pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3316 if (pfx == NULL)
3317 return -EINVAL;
3319 return inet6_addr_del(net, ifm->ifa_index, pfx, ifm->ifa_prefixlen);
3322 static int inet6_addr_modify(struct inet6_ifaddr *ifp, u8 ifa_flags,
3323 u32 prefered_lft, u32 valid_lft)
3325 u32 flags;
3326 clock_t expires;
3327 unsigned long timeout;
3329 if (!valid_lft || (prefered_lft > valid_lft))
3330 return -EINVAL;
3332 timeout = addrconf_timeout_fixup(valid_lft, HZ);
3333 if (addrconf_finite_timeout(timeout)) {
3334 expires = jiffies_to_clock_t(timeout * HZ);
3335 valid_lft = timeout;
3336 flags = RTF_EXPIRES;
3337 } else {
3338 expires = 0;
3339 flags = 0;
3340 ifa_flags |= IFA_F_PERMANENT;
3343 timeout = addrconf_timeout_fixup(prefered_lft, HZ);
3344 if (addrconf_finite_timeout(timeout)) {
3345 if (timeout == 0)
3346 ifa_flags |= IFA_F_DEPRECATED;
3347 prefered_lft = timeout;
3350 spin_lock_bh(&ifp->lock);
3351 ifp->flags = (ifp->flags & ~(IFA_F_DEPRECATED | IFA_F_PERMANENT | IFA_F_NODAD | IFA_F_HOMEADDRESS)) | ifa_flags;
3352 ifp->tstamp = jiffies;
3353 ifp->valid_lft = valid_lft;
3354 ifp->prefered_lft = prefered_lft;
3356 spin_unlock_bh(&ifp->lock);
3357 if (!(ifp->flags&IFA_F_TENTATIVE))
3358 ipv6_ifa_notify(0, ifp);
3360 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, ifp->idev->dev,
3361 expires, flags);
3362 addrconf_verify(0);
3364 return 0;
3367 static int
3368 inet6_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
3370 struct net *net = sock_net(skb->sk);
3371 struct ifaddrmsg *ifm;
3372 struct nlattr *tb[IFA_MAX+1];
3373 struct in6_addr *pfx;
3374 struct inet6_ifaddr *ifa;
3375 struct net_device *dev;
3376 u32 valid_lft = INFINITY_LIFE_TIME, preferred_lft = INFINITY_LIFE_TIME;
3377 u8 ifa_flags;
3378 int err;
3380 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3381 if (err < 0)
3382 return err;
3384 ifm = nlmsg_data(nlh);
3385 pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3386 if (pfx == NULL)
3387 return -EINVAL;
3389 if (tb[IFA_CACHEINFO]) {
3390 struct ifa_cacheinfo *ci;
3392 ci = nla_data(tb[IFA_CACHEINFO]);
3393 valid_lft = ci->ifa_valid;
3394 preferred_lft = ci->ifa_prefered;
3395 } else {
3396 preferred_lft = INFINITY_LIFE_TIME;
3397 valid_lft = INFINITY_LIFE_TIME;
3400 dev = __dev_get_by_index(net, ifm->ifa_index);
3401 if (dev == NULL)
3402 return -ENODEV;
3404 /* We ignore other flags so far. */
3405 ifa_flags = ifm->ifa_flags & (IFA_F_NODAD | IFA_F_HOMEADDRESS);
3407 ifa = ipv6_get_ifaddr(net, pfx, dev, 1);
3408 if (ifa == NULL) {
3410 * It would be best to check for !NLM_F_CREATE here but
3411 * userspace alreay relies on not having to provide this.
3413 return inet6_addr_add(net, ifm->ifa_index, pfx,
3414 ifm->ifa_prefixlen, ifa_flags,
3415 preferred_lft, valid_lft);
3418 if (nlh->nlmsg_flags & NLM_F_EXCL ||
3419 !(nlh->nlmsg_flags & NLM_F_REPLACE))
3420 err = -EEXIST;
3421 else
3422 err = inet6_addr_modify(ifa, ifa_flags, preferred_lft, valid_lft);
3424 in6_ifa_put(ifa);
3426 return err;
3429 static void put_ifaddrmsg(struct nlmsghdr *nlh, u8 prefixlen, u8 flags,
3430 u8 scope, int ifindex)
3432 struct ifaddrmsg *ifm;
3434 ifm = nlmsg_data(nlh);
3435 ifm->ifa_family = AF_INET6;
3436 ifm->ifa_prefixlen = prefixlen;
3437 ifm->ifa_flags = flags;
3438 ifm->ifa_scope = scope;
3439 ifm->ifa_index = ifindex;
3442 static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
3443 unsigned long tstamp, u32 preferred, u32 valid)
3445 struct ifa_cacheinfo ci;
3447 ci.cstamp = (u32)(TIME_DELTA(cstamp, INITIAL_JIFFIES) / HZ * 100
3448 + TIME_DELTA(cstamp, INITIAL_JIFFIES) % HZ * 100 / HZ);
3449 ci.tstamp = (u32)(TIME_DELTA(tstamp, INITIAL_JIFFIES) / HZ * 100
3450 + TIME_DELTA(tstamp, INITIAL_JIFFIES) % HZ * 100 / HZ);
3451 ci.ifa_prefered = preferred;
3452 ci.ifa_valid = valid;
3454 return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
3457 static inline int rt_scope(int ifa_scope)
3459 if (ifa_scope & IFA_HOST)
3460 return RT_SCOPE_HOST;
3461 else if (ifa_scope & IFA_LINK)
3462 return RT_SCOPE_LINK;
3463 else if (ifa_scope & IFA_SITE)
3464 return RT_SCOPE_SITE;
3465 else
3466 return RT_SCOPE_UNIVERSE;
3469 static inline int inet6_ifaddr_msgsize(void)
3471 return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
3472 + nla_total_size(16) /* IFA_ADDRESS */
3473 + nla_total_size(sizeof(struct ifa_cacheinfo));
3476 static int inet6_fill_ifaddr(struct sk_buff *skb, struct inet6_ifaddr *ifa,
3477 u32 pid, u32 seq, int event, unsigned int flags)
3479 struct nlmsghdr *nlh;
3480 u32 preferred, valid;
3482 nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3483 if (nlh == NULL)
3484 return -EMSGSIZE;
3486 put_ifaddrmsg(nlh, ifa->prefix_len, ifa->flags, rt_scope(ifa->scope),
3487 ifa->idev->dev->ifindex);
3489 if (!(ifa->flags&IFA_F_PERMANENT)) {
3490 preferred = ifa->prefered_lft;
3491 valid = ifa->valid_lft;
3492 if (preferred != INFINITY_LIFE_TIME) {
3493 long tval = (jiffies - ifa->tstamp)/HZ;
3494 if (preferred > tval)
3495 preferred -= tval;
3496 else
3497 preferred = 0;
3498 if (valid != INFINITY_LIFE_TIME) {
3499 if (valid > tval)
3500 valid -= tval;
3501 else
3502 valid = 0;
3505 } else {
3506 preferred = INFINITY_LIFE_TIME;
3507 valid = INFINITY_LIFE_TIME;
3510 if (nla_put(skb, IFA_ADDRESS, 16, &ifa->addr) < 0 ||
3511 put_cacheinfo(skb, ifa->cstamp, ifa->tstamp, preferred, valid) < 0) {
3512 nlmsg_cancel(skb, nlh);
3513 return -EMSGSIZE;
3516 return nlmsg_end(skb, nlh);
3519 static int inet6_fill_ifmcaddr(struct sk_buff *skb, struct ifmcaddr6 *ifmca,
3520 u32 pid, u32 seq, int event, u16 flags)
3522 struct nlmsghdr *nlh;
3523 u8 scope = RT_SCOPE_UNIVERSE;
3524 int ifindex = ifmca->idev->dev->ifindex;
3526 if (ipv6_addr_scope(&ifmca->mca_addr) & IFA_SITE)
3527 scope = RT_SCOPE_SITE;
3529 nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3530 if (nlh == NULL)
3531 return -EMSGSIZE;
3533 put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3534 if (nla_put(skb, IFA_MULTICAST, 16, &ifmca->mca_addr) < 0 ||
3535 put_cacheinfo(skb, ifmca->mca_cstamp, ifmca->mca_tstamp,
3536 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3537 nlmsg_cancel(skb, nlh);
3538 return -EMSGSIZE;
3541 return nlmsg_end(skb, nlh);
3544 static int inet6_fill_ifacaddr(struct sk_buff *skb, struct ifacaddr6 *ifaca,
3545 u32 pid, u32 seq, int event, unsigned int flags)
3547 struct nlmsghdr *nlh;
3548 u8 scope = RT_SCOPE_UNIVERSE;
3549 int ifindex = ifaca->aca_idev->dev->ifindex;
3551 if (ipv6_addr_scope(&ifaca->aca_addr) & IFA_SITE)
3552 scope = RT_SCOPE_SITE;
3554 nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3555 if (nlh == NULL)
3556 return -EMSGSIZE;
3558 put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3559 if (nla_put(skb, IFA_ANYCAST, 16, &ifaca->aca_addr) < 0 ||
3560 put_cacheinfo(skb, ifaca->aca_cstamp, ifaca->aca_tstamp,
3561 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3562 nlmsg_cancel(skb, nlh);
3563 return -EMSGSIZE;
3566 return nlmsg_end(skb, nlh);
3569 enum addr_type_t {
3570 UNICAST_ADDR,
3571 MULTICAST_ADDR,
3572 ANYCAST_ADDR,
3575 /* called with rcu_read_lock() */
3576 static int in6_dump_addrs(struct inet6_dev *idev, struct sk_buff *skb,
3577 struct netlink_callback *cb, enum addr_type_t type,
3578 int s_ip_idx, int *p_ip_idx)
3580 struct ifmcaddr6 *ifmca;
3581 struct ifacaddr6 *ifaca;
3582 int err = 1;
3583 int ip_idx = *p_ip_idx;
3585 read_lock_bh(&idev->lock);
3586 switch (type) {
3587 case UNICAST_ADDR: {
3588 struct inet6_ifaddr *ifa;
3590 /* unicast address incl. temp addr */
3591 list_for_each_entry(ifa, &idev->addr_list, if_list) {
3592 if (++ip_idx < s_ip_idx)
3593 continue;
3594 err = inet6_fill_ifaddr(skb, ifa,
3595 NETLINK_CB(cb->skb).pid,
3596 cb->nlh->nlmsg_seq,
3597 RTM_NEWADDR,
3598 NLM_F_MULTI);
3599 if (err <= 0)
3600 break;
3602 break;
3604 case MULTICAST_ADDR:
3605 /* multicast address */
3606 for (ifmca = idev->mc_list; ifmca;
3607 ifmca = ifmca->next, ip_idx++) {
3608 if (ip_idx < s_ip_idx)
3609 continue;
3610 err = inet6_fill_ifmcaddr(skb, ifmca,
3611 NETLINK_CB(cb->skb).pid,
3612 cb->nlh->nlmsg_seq,
3613 RTM_GETMULTICAST,
3614 NLM_F_MULTI);
3615 if (err <= 0)
3616 break;
3618 break;
3619 case ANYCAST_ADDR:
3620 /* anycast address */
3621 for (ifaca = idev->ac_list; ifaca;
3622 ifaca = ifaca->aca_next, ip_idx++) {
3623 if (ip_idx < s_ip_idx)
3624 continue;
3625 err = inet6_fill_ifacaddr(skb, ifaca,
3626 NETLINK_CB(cb->skb).pid,
3627 cb->nlh->nlmsg_seq,
3628 RTM_GETANYCAST,
3629 NLM_F_MULTI);
3630 if (err <= 0)
3631 break;
3633 break;
3634 default:
3635 break;
3637 read_unlock_bh(&idev->lock);
3638 *p_ip_idx = ip_idx;
3639 return err;
3642 static int inet6_dump_addr(struct sk_buff *skb, struct netlink_callback *cb,
3643 enum addr_type_t type)
3645 struct net *net = sock_net(skb->sk);
3646 int h, s_h;
3647 int idx, ip_idx;
3648 int s_idx, s_ip_idx;
3649 struct net_device *dev;
3650 struct inet6_dev *idev;
3651 struct hlist_head *head;
3652 struct hlist_node *node;
3654 s_h = cb->args[0];
3655 s_idx = idx = cb->args[1];
3656 s_ip_idx = ip_idx = cb->args[2];
3658 rcu_read_lock();
3659 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
3660 idx = 0;
3661 head = &net->dev_index_head[h];
3662 hlist_for_each_entry_rcu(dev, node, head, index_hlist) {
3663 if (idx < s_idx)
3664 goto cont;
3665 if (h > s_h || idx > s_idx)
3666 s_ip_idx = 0;
3667 ip_idx = 0;
3668 idev = __in6_dev_get(dev);
3669 if (!idev)
3670 goto cont;
3672 if (in6_dump_addrs(idev, skb, cb, type,
3673 s_ip_idx, &ip_idx) <= 0)
3674 goto done;
3675 cont:
3676 idx++;
3679 done:
3680 rcu_read_unlock();
3681 cb->args[0] = h;
3682 cb->args[1] = idx;
3683 cb->args[2] = ip_idx;
3685 return skb->len;
3688 static int inet6_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
3690 enum addr_type_t type = UNICAST_ADDR;
3692 return inet6_dump_addr(skb, cb, type);
3695 static int inet6_dump_ifmcaddr(struct sk_buff *skb, struct netlink_callback *cb)
3697 enum addr_type_t type = MULTICAST_ADDR;
3699 return inet6_dump_addr(skb, cb, type);
3703 static int inet6_dump_ifacaddr(struct sk_buff *skb, struct netlink_callback *cb)
3705 enum addr_type_t type = ANYCAST_ADDR;
3707 return inet6_dump_addr(skb, cb, type);
3710 static int inet6_rtm_getaddr(struct sk_buff *in_skb, struct nlmsghdr* nlh,
3711 void *arg)
3713 struct net *net = sock_net(in_skb->sk);
3714 struct ifaddrmsg *ifm;
3715 struct nlattr *tb[IFA_MAX+1];
3716 struct in6_addr *addr = NULL;
3717 struct net_device *dev = NULL;
3718 struct inet6_ifaddr *ifa;
3719 struct sk_buff *skb;
3720 int err;
3722 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3723 if (err < 0)
3724 goto errout;
3726 addr = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3727 if (addr == NULL) {
3728 err = -EINVAL;
3729 goto errout;
3732 ifm = nlmsg_data(nlh);
3733 if (ifm->ifa_index)
3734 dev = __dev_get_by_index(net, ifm->ifa_index);
3736 ifa = ipv6_get_ifaddr(net, addr, dev, 1);
3737 if (!ifa) {
3738 err = -EADDRNOTAVAIL;
3739 goto errout;
3742 skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_KERNEL);
3743 if (!skb) {
3744 err = -ENOBUFS;
3745 goto errout_ifa;
3748 err = inet6_fill_ifaddr(skb, ifa, NETLINK_CB(in_skb).pid,
3749 nlh->nlmsg_seq, RTM_NEWADDR, 0);
3750 if (err < 0) {
3751 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
3752 WARN_ON(err == -EMSGSIZE);
3753 kfree_skb(skb);
3754 goto errout_ifa;
3756 err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).pid);
3757 errout_ifa:
3758 in6_ifa_put(ifa);
3759 errout:
3760 return err;
3763 static void inet6_ifa_notify(int event, struct inet6_ifaddr *ifa)
3765 struct sk_buff *skb;
3766 struct net *net = dev_net(ifa->idev->dev);
3767 int err = -ENOBUFS;
3769 skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_ATOMIC);
3770 if (skb == NULL)
3771 goto errout;
3773 err = inet6_fill_ifaddr(skb, ifa, 0, 0, event, 0);
3774 if (err < 0) {
3775 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
3776 WARN_ON(err == -EMSGSIZE);
3777 kfree_skb(skb);
3778 goto errout;
3780 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
3781 return;
3782 errout:
3783 if (err < 0)
3784 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
3787 static inline void ipv6_store_devconf(struct ipv6_devconf *cnf,
3788 __s32 *array, int bytes)
3790 BUG_ON(bytes < (DEVCONF_MAX * 4));
3792 memset(array, 0, bytes);
3793 array[DEVCONF_FORWARDING] = cnf->forwarding;
3794 array[DEVCONF_HOPLIMIT] = cnf->hop_limit;
3795 array[DEVCONF_MTU6] = cnf->mtu6;
3796 array[DEVCONF_ACCEPT_RA] = cnf->accept_ra;
3797 array[DEVCONF_ACCEPT_REDIRECTS] = cnf->accept_redirects;
3798 array[DEVCONF_AUTOCONF] = cnf->autoconf;
3799 array[DEVCONF_DAD_TRANSMITS] = cnf->dad_transmits;
3800 array[DEVCONF_RTR_SOLICITS] = cnf->rtr_solicits;
3801 array[DEVCONF_RTR_SOLICIT_INTERVAL] = cnf->rtr_solicit_interval;
3802 array[DEVCONF_RTR_SOLICIT_DELAY] = cnf->rtr_solicit_delay;
3803 array[DEVCONF_FORCE_MLD_VERSION] = cnf->force_mld_version;
3804 #ifdef CONFIG_IPV6_PRIVACY
3805 array[DEVCONF_USE_TEMPADDR] = cnf->use_tempaddr;
3806 array[DEVCONF_TEMP_VALID_LFT] = cnf->temp_valid_lft;
3807 array[DEVCONF_TEMP_PREFERED_LFT] = cnf->temp_prefered_lft;
3808 array[DEVCONF_REGEN_MAX_RETRY] = cnf->regen_max_retry;
3809 array[DEVCONF_MAX_DESYNC_FACTOR] = cnf->max_desync_factor;
3810 #endif
3811 array[DEVCONF_MAX_ADDRESSES] = cnf->max_addresses;
3812 array[DEVCONF_ACCEPT_RA_DEFRTR] = cnf->accept_ra_defrtr;
3813 array[DEVCONF_ACCEPT_RA_PINFO] = cnf->accept_ra_pinfo;
3814 #ifdef CONFIG_IPV6_ROUTER_PREF
3815 array[DEVCONF_ACCEPT_RA_RTR_PREF] = cnf->accept_ra_rtr_pref;
3816 array[DEVCONF_RTR_PROBE_INTERVAL] = cnf->rtr_probe_interval;
3817 #ifdef CONFIG_IPV6_ROUTE_INFO
3818 array[DEVCONF_ACCEPT_RA_RT_INFO_MAX_PLEN] = cnf->accept_ra_rt_info_max_plen;
3819 #endif
3820 #endif
3821 array[DEVCONF_PROXY_NDP] = cnf->proxy_ndp;
3822 array[DEVCONF_ACCEPT_SOURCE_ROUTE] = cnf->accept_source_route;
3823 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
3824 array[DEVCONF_OPTIMISTIC_DAD] = cnf->optimistic_dad;
3825 #endif
3826 #ifdef CONFIG_IPV6_MROUTE
3827 array[DEVCONF_MC_FORWARDING] = cnf->mc_forwarding;
3828 #endif
3829 array[DEVCONF_DISABLE_IPV6] = cnf->disable_ipv6;
3830 array[DEVCONF_ACCEPT_DAD] = cnf->accept_dad;
3831 array[DEVCONF_FORCE_TLLAO] = cnf->force_tllao;
3834 static inline size_t inet6_if_nlmsg_size(void)
3836 return NLMSG_ALIGN(sizeof(struct ifinfomsg))
3837 + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
3838 + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
3839 + nla_total_size(4) /* IFLA_MTU */
3840 + nla_total_size(4) /* IFLA_LINK */
3841 + nla_total_size( /* IFLA_PROTINFO */
3842 nla_total_size(4) /* IFLA_INET6_FLAGS */
3843 + nla_total_size(sizeof(struct ifla_cacheinfo))
3844 + nla_total_size(DEVCONF_MAX * 4) /* IFLA_INET6_CONF */
3845 + nla_total_size(IPSTATS_MIB_MAX * 8) /* IFLA_INET6_STATS */
3846 + nla_total_size(ICMP6_MIB_MAX * 8) /* IFLA_INET6_ICMP6STATS */
3850 static inline void __snmp6_fill_stats(u64 *stats, void __percpu **mib,
3851 int items, int bytes)
3853 int i;
3854 int pad = bytes - sizeof(u64) * items;
3855 BUG_ON(pad < 0);
3857 /* Use put_unaligned() because stats may not be aligned for u64. */
3858 put_unaligned(items, &stats[0]);
3859 for (i = 1; i < items; i++)
3860 put_unaligned(snmp_fold_field(mib, i), &stats[i]);
3862 memset(&stats[items], 0, pad);
3865 static inline void __snmp6_fill_stats64(u64 *stats, void __percpu **mib,
3866 int items, int bytes, size_t syncpoff)
3868 int i;
3869 int pad = bytes - sizeof(u64) * items;
3870 BUG_ON(pad < 0);
3872 /* Use put_unaligned() because stats may not be aligned for u64. */
3873 put_unaligned(items, &stats[0]);
3874 for (i = 1; i < items; i++)
3875 put_unaligned(snmp_fold_field64(mib, i, syncpoff), &stats[i]);
3877 memset(&stats[items], 0, pad);
3880 static void snmp6_fill_stats(u64 *stats, struct inet6_dev *idev, int attrtype,
3881 int bytes)
3883 switch (attrtype) {
3884 case IFLA_INET6_STATS:
3885 __snmp6_fill_stats64(stats, (void __percpu **)idev->stats.ipv6,
3886 IPSTATS_MIB_MAX, bytes, offsetof(struct ipstats_mib, syncp));
3887 break;
3888 case IFLA_INET6_ICMP6STATS:
3889 __snmp6_fill_stats(stats, (void __percpu **)idev->stats.icmpv6, ICMP6_MIB_MAX, bytes);
3890 break;
3894 static int inet6_fill_ifinfo(struct sk_buff *skb, struct inet6_dev *idev,
3895 u32 pid, u32 seq, int event, unsigned int flags)
3897 struct net_device *dev = idev->dev;
3898 struct nlattr *nla;
3899 struct ifinfomsg *hdr;
3900 struct nlmsghdr *nlh;
3901 void *protoinfo;
3902 struct ifla_cacheinfo ci;
3904 nlh = nlmsg_put(skb, pid, seq, event, sizeof(*hdr), flags);
3905 if (nlh == NULL)
3906 return -EMSGSIZE;
3908 hdr = nlmsg_data(nlh);
3909 hdr->ifi_family = AF_INET6;
3910 hdr->__ifi_pad = 0;
3911 hdr->ifi_type = dev->type;
3912 hdr->ifi_index = dev->ifindex;
3913 hdr->ifi_flags = dev_get_flags(dev);
3914 hdr->ifi_change = 0;
3916 NLA_PUT_STRING(skb, IFLA_IFNAME, dev->name);
3918 if (dev->addr_len)
3919 NLA_PUT(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr);
3921 NLA_PUT_U32(skb, IFLA_MTU, dev->mtu);
3922 if (dev->ifindex != dev->iflink)
3923 NLA_PUT_U32(skb, IFLA_LINK, dev->iflink);
3925 protoinfo = nla_nest_start(skb, IFLA_PROTINFO);
3926 if (protoinfo == NULL)
3927 goto nla_put_failure;
3929 NLA_PUT_U32(skb, IFLA_INET6_FLAGS, idev->if_flags);
3931 ci.max_reasm_len = IPV6_MAXPLEN;
3932 ci.tstamp = (__u32)(TIME_DELTA(idev->tstamp, INITIAL_JIFFIES) / HZ * 100
3933 + TIME_DELTA(idev->tstamp, INITIAL_JIFFIES) % HZ * 100 / HZ);
3934 ci.reachable_time = idev->nd_parms->reachable_time;
3935 ci.retrans_time = idev->nd_parms->retrans_time;
3936 NLA_PUT(skb, IFLA_INET6_CACHEINFO, sizeof(ci), &ci);
3938 nla = nla_reserve(skb, IFLA_INET6_CONF, DEVCONF_MAX * sizeof(s32));
3939 if (nla == NULL)
3940 goto nla_put_failure;
3941 ipv6_store_devconf(&idev->cnf, nla_data(nla), nla_len(nla));
3943 /* XXX - MC not implemented */
3945 nla = nla_reserve(skb, IFLA_INET6_STATS, IPSTATS_MIB_MAX * sizeof(u64));
3946 if (nla == NULL)
3947 goto nla_put_failure;
3948 snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_STATS, nla_len(nla));
3950 nla = nla_reserve(skb, IFLA_INET6_ICMP6STATS, ICMP6_MIB_MAX * sizeof(u64));
3951 if (nla == NULL)
3952 goto nla_put_failure;
3953 snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_ICMP6STATS, nla_len(nla));
3955 nla_nest_end(skb, protoinfo);
3956 return nlmsg_end(skb, nlh);
3958 nla_put_failure:
3959 nlmsg_cancel(skb, nlh);
3960 return -EMSGSIZE;
3963 static int inet6_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
3965 struct net *net = sock_net(skb->sk);
3966 int h, s_h;
3967 int idx = 0, s_idx;
3968 struct net_device *dev;
3969 struct inet6_dev *idev;
3970 struct hlist_head *head;
3971 struct hlist_node *node;
3973 s_h = cb->args[0];
3974 s_idx = cb->args[1];
3976 rcu_read_lock();
3977 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
3978 idx = 0;
3979 head = &net->dev_index_head[h];
3980 hlist_for_each_entry_rcu(dev, node, head, index_hlist) {
3981 if (idx < s_idx)
3982 goto cont;
3983 idev = __in6_dev_get(dev);
3984 if (!idev)
3985 goto cont;
3986 if (inet6_fill_ifinfo(skb, idev,
3987 NETLINK_CB(cb->skb).pid,
3988 cb->nlh->nlmsg_seq,
3989 RTM_NEWLINK, NLM_F_MULTI) <= 0)
3990 goto out;
3991 cont:
3992 idx++;
3995 out:
3996 rcu_read_unlock();
3997 cb->args[1] = idx;
3998 cb->args[0] = h;
4000 return skb->len;
4003 void inet6_ifinfo_notify(int event, struct inet6_dev *idev)
4005 struct sk_buff *skb;
4006 struct net *net = dev_net(idev->dev);
4007 int err = -ENOBUFS;
4009 skb = nlmsg_new(inet6_if_nlmsg_size(), GFP_ATOMIC);
4010 if (skb == NULL)
4011 goto errout;
4013 err = inet6_fill_ifinfo(skb, idev, 0, 0, event, 0);
4014 if (err < 0) {
4015 /* -EMSGSIZE implies BUG in inet6_if_nlmsg_size() */
4016 WARN_ON(err == -EMSGSIZE);
4017 kfree_skb(skb);
4018 goto errout;
4020 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
4021 return;
4022 errout:
4023 if (err < 0)
4024 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
4027 static inline size_t inet6_prefix_nlmsg_size(void)
4029 return NLMSG_ALIGN(sizeof(struct prefixmsg))
4030 + nla_total_size(sizeof(struct in6_addr))
4031 + nla_total_size(sizeof(struct prefix_cacheinfo));
4034 static int inet6_fill_prefix(struct sk_buff *skb, struct inet6_dev *idev,
4035 struct prefix_info *pinfo, u32 pid, u32 seq,
4036 int event, unsigned int flags)
4038 struct prefixmsg *pmsg;
4039 struct nlmsghdr *nlh;
4040 struct prefix_cacheinfo ci;
4042 nlh = nlmsg_put(skb, pid, seq, event, sizeof(*pmsg), flags);
4043 if (nlh == NULL)
4044 return -EMSGSIZE;
4046 pmsg = nlmsg_data(nlh);
4047 pmsg->prefix_family = AF_INET6;
4048 pmsg->prefix_pad1 = 0;
4049 pmsg->prefix_pad2 = 0;
4050 pmsg->prefix_ifindex = idev->dev->ifindex;
4051 pmsg->prefix_len = pinfo->prefix_len;
4052 pmsg->prefix_type = pinfo->type;
4053 pmsg->prefix_pad3 = 0;
4054 pmsg->prefix_flags = 0;
4055 if (pinfo->onlink)
4056 pmsg->prefix_flags |= IF_PREFIX_ONLINK;
4057 if (pinfo->autoconf)
4058 pmsg->prefix_flags |= IF_PREFIX_AUTOCONF;
4060 NLA_PUT(skb, PREFIX_ADDRESS, sizeof(pinfo->prefix), &pinfo->prefix);
4062 ci.preferred_time = ntohl(pinfo->prefered);
4063 ci.valid_time = ntohl(pinfo->valid);
4064 NLA_PUT(skb, PREFIX_CACHEINFO, sizeof(ci), &ci);
4066 return nlmsg_end(skb, nlh);
4068 nla_put_failure:
4069 nlmsg_cancel(skb, nlh);
4070 return -EMSGSIZE;
4073 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
4074 struct prefix_info *pinfo)
4076 struct sk_buff *skb;
4077 struct net *net = dev_net(idev->dev);
4078 int err = -ENOBUFS;
4080 skb = nlmsg_new(inet6_prefix_nlmsg_size(), GFP_ATOMIC);
4081 if (skb == NULL)
4082 goto errout;
4084 err = inet6_fill_prefix(skb, idev, pinfo, 0, 0, event, 0);
4085 if (err < 0) {
4086 /* -EMSGSIZE implies BUG in inet6_prefix_nlmsg_size() */
4087 WARN_ON(err == -EMSGSIZE);
4088 kfree_skb(skb);
4089 goto errout;
4091 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_PREFIX, NULL, GFP_ATOMIC);
4092 return;
4093 errout:
4094 if (err < 0)
4095 rtnl_set_sk_err(net, RTNLGRP_IPV6_PREFIX, err);
4098 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
4100 inet6_ifa_notify(event ? : RTM_NEWADDR, ifp);
4102 switch (event) {
4103 case RTM_NEWADDR:
4105 * If the address was optimistic
4106 * we inserted the route at the start of
4107 * our DAD process, so we don't need
4108 * to do it again
4110 if (!(ifp->rt->rt6i_node))
4111 ip6_ins_rt(ifp->rt);
4112 if (ifp->idev->cnf.forwarding)
4113 addrconf_join_anycast(ifp);
4114 break;
4115 case RTM_DELADDR:
4116 if (ifp->idev->cnf.forwarding)
4117 addrconf_leave_anycast(ifp);
4118 addrconf_leave_solict(ifp->idev, &ifp->addr);
4119 dst_hold(&ifp->rt->dst);
4121 if (ifp->state == INET6_IFADDR_STATE_DEAD &&
4122 ip6_del_rt(ifp->rt))
4123 dst_free(&ifp->rt->dst);
4124 break;
4128 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
4130 rcu_read_lock_bh();
4131 if (likely(ifp->idev->dead == 0))
4132 __ipv6_ifa_notify(event, ifp);
4133 rcu_read_unlock_bh();
4136 #ifdef CONFIG_SYSCTL
4138 static
4139 int addrconf_sysctl_forward(ctl_table *ctl, int write,
4140 void __user *buffer, size_t *lenp, loff_t *ppos)
4142 int *valp = ctl->data;
4143 int val = *valp;
4144 loff_t pos = *ppos;
4145 int ret;
4147 ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
4149 if (write)
4150 ret = addrconf_fixup_forwarding(ctl, valp, val);
4151 if (ret)
4152 *ppos = pos;
4153 return ret;
4156 static void dev_disable_change(struct inet6_dev *idev)
4158 if (!idev || !idev->dev)
4159 return;
4161 if (idev->cnf.disable_ipv6)
4162 addrconf_notify(NULL, NETDEV_DOWN, idev->dev);
4163 else
4164 addrconf_notify(NULL, NETDEV_UP, idev->dev);
4167 static void addrconf_disable_change(struct net *net, __s32 newf)
4169 struct net_device *dev;
4170 struct inet6_dev *idev;
4172 rcu_read_lock();
4173 for_each_netdev_rcu(net, dev) {
4174 idev = __in6_dev_get(dev);
4175 if (idev) {
4176 int changed = (!idev->cnf.disable_ipv6) ^ (!newf);
4177 idev->cnf.disable_ipv6 = newf;
4178 if (changed)
4179 dev_disable_change(idev);
4182 rcu_read_unlock();
4185 static int addrconf_disable_ipv6(struct ctl_table *table, int *p, int old)
4187 struct net *net;
4189 net = (struct net *)table->extra2;
4191 if (p == &net->ipv6.devconf_dflt->disable_ipv6)
4192 return 0;
4194 if (!rtnl_trylock()) {
4195 /* Restore the original values before restarting */
4196 *p = old;
4197 return restart_syscall();
4200 if (p == &net->ipv6.devconf_all->disable_ipv6) {
4201 __s32 newf = net->ipv6.devconf_all->disable_ipv6;
4202 net->ipv6.devconf_dflt->disable_ipv6 = newf;
4203 addrconf_disable_change(net, newf);
4204 } else if ((!*p) ^ (!old))
4205 dev_disable_change((struct inet6_dev *)table->extra1);
4207 rtnl_unlock();
4208 return 0;
4211 static
4212 int addrconf_sysctl_disable(ctl_table *ctl, int write,
4213 void __user *buffer, size_t *lenp, loff_t *ppos)
4215 int *valp = ctl->data;
4216 int val = *valp;
4217 loff_t pos = *ppos;
4218 int ret;
4220 ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
4222 if (write)
4223 ret = addrconf_disable_ipv6(ctl, valp, val);
4224 if (ret)
4225 *ppos = pos;
4226 return ret;
4229 static struct addrconf_sysctl_table
4231 struct ctl_table_header *sysctl_header;
4232 ctl_table addrconf_vars[DEVCONF_MAX+1];
4233 char *dev_name;
4234 } addrconf_sysctl __read_mostly = {
4235 .sysctl_header = NULL,
4236 .addrconf_vars = {
4238 .procname = "forwarding",
4239 .data = &ipv6_devconf.forwarding,
4240 .maxlen = sizeof(int),
4241 .mode = 0644,
4242 .proc_handler = addrconf_sysctl_forward,
4245 .procname = "hop_limit",
4246 .data = &ipv6_devconf.hop_limit,
4247 .maxlen = sizeof(int),
4248 .mode = 0644,
4249 .proc_handler = proc_dointvec,
4252 .procname = "mtu",
4253 .data = &ipv6_devconf.mtu6,
4254 .maxlen = sizeof(int),
4255 .mode = 0644,
4256 .proc_handler = proc_dointvec,
4259 .procname = "accept_ra",
4260 .data = &ipv6_devconf.accept_ra,
4261 .maxlen = sizeof(int),
4262 .mode = 0644,
4263 .proc_handler = proc_dointvec,
4266 .procname = "accept_redirects",
4267 .data = &ipv6_devconf.accept_redirects,
4268 .maxlen = sizeof(int),
4269 .mode = 0644,
4270 .proc_handler = proc_dointvec,
4273 .procname = "autoconf",
4274 .data = &ipv6_devconf.autoconf,
4275 .maxlen = sizeof(int),
4276 .mode = 0644,
4277 .proc_handler = proc_dointvec,
4280 .procname = "dad_transmits",
4281 .data = &ipv6_devconf.dad_transmits,
4282 .maxlen = sizeof(int),
4283 .mode = 0644,
4284 .proc_handler = proc_dointvec,
4287 .procname = "router_solicitations",
4288 .data = &ipv6_devconf.rtr_solicits,
4289 .maxlen = sizeof(int),
4290 .mode = 0644,
4291 .proc_handler = proc_dointvec,
4294 .procname = "router_solicitation_interval",
4295 .data = &ipv6_devconf.rtr_solicit_interval,
4296 .maxlen = sizeof(int),
4297 .mode = 0644,
4298 .proc_handler = proc_dointvec_jiffies,
4301 .procname = "router_solicitation_delay",
4302 .data = &ipv6_devconf.rtr_solicit_delay,
4303 .maxlen = sizeof(int),
4304 .mode = 0644,
4305 .proc_handler = proc_dointvec_jiffies,
4308 .procname = "force_mld_version",
4309 .data = &ipv6_devconf.force_mld_version,
4310 .maxlen = sizeof(int),
4311 .mode = 0644,
4312 .proc_handler = proc_dointvec,
4314 #ifdef CONFIG_IPV6_PRIVACY
4316 .procname = "use_tempaddr",
4317 .data = &ipv6_devconf.use_tempaddr,
4318 .maxlen = sizeof(int),
4319 .mode = 0644,
4320 .proc_handler = proc_dointvec,
4323 .procname = "temp_valid_lft",
4324 .data = &ipv6_devconf.temp_valid_lft,
4325 .maxlen = sizeof(int),
4326 .mode = 0644,
4327 .proc_handler = proc_dointvec,
4330 .procname = "temp_prefered_lft",
4331 .data = &ipv6_devconf.temp_prefered_lft,
4332 .maxlen = sizeof(int),
4333 .mode = 0644,
4334 .proc_handler = proc_dointvec,
4337 .procname = "regen_max_retry",
4338 .data = &ipv6_devconf.regen_max_retry,
4339 .maxlen = sizeof(int),
4340 .mode = 0644,
4341 .proc_handler = proc_dointvec,
4344 .procname = "max_desync_factor",
4345 .data = &ipv6_devconf.max_desync_factor,
4346 .maxlen = sizeof(int),
4347 .mode = 0644,
4348 .proc_handler = proc_dointvec,
4350 #endif
4352 .procname = "max_addresses",
4353 .data = &ipv6_devconf.max_addresses,
4354 .maxlen = sizeof(int),
4355 .mode = 0644,
4356 .proc_handler = proc_dointvec,
4359 .procname = "accept_ra_defrtr",
4360 .data = &ipv6_devconf.accept_ra_defrtr,
4361 .maxlen = sizeof(int),
4362 .mode = 0644,
4363 .proc_handler = proc_dointvec,
4366 .procname = "accept_ra_pinfo",
4367 .data = &ipv6_devconf.accept_ra_pinfo,
4368 .maxlen = sizeof(int),
4369 .mode = 0644,
4370 .proc_handler = proc_dointvec,
4372 #ifdef CONFIG_IPV6_ROUTER_PREF
4374 .procname = "accept_ra_rtr_pref",
4375 .data = &ipv6_devconf.accept_ra_rtr_pref,
4376 .maxlen = sizeof(int),
4377 .mode = 0644,
4378 .proc_handler = proc_dointvec,
4381 .procname = "router_probe_interval",
4382 .data = &ipv6_devconf.rtr_probe_interval,
4383 .maxlen = sizeof(int),
4384 .mode = 0644,
4385 .proc_handler = proc_dointvec_jiffies,
4387 #ifdef CONFIG_IPV6_ROUTE_INFO
4389 .procname = "accept_ra_rt_info_max_plen",
4390 .data = &ipv6_devconf.accept_ra_rt_info_max_plen,
4391 .maxlen = sizeof(int),
4392 .mode = 0644,
4393 .proc_handler = proc_dointvec,
4395 #endif
4396 #endif
4398 .procname = "proxy_ndp",
4399 .data = &ipv6_devconf.proxy_ndp,
4400 .maxlen = sizeof(int),
4401 .mode = 0644,
4402 .proc_handler = proc_dointvec,
4405 .procname = "accept_source_route",
4406 .data = &ipv6_devconf.accept_source_route,
4407 .maxlen = sizeof(int),
4408 .mode = 0644,
4409 .proc_handler = proc_dointvec,
4411 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
4413 .procname = "optimistic_dad",
4414 .data = &ipv6_devconf.optimistic_dad,
4415 .maxlen = sizeof(int),
4416 .mode = 0644,
4417 .proc_handler = proc_dointvec,
4420 #endif
4421 #ifdef CONFIG_IPV6_MROUTE
4423 .procname = "mc_forwarding",
4424 .data = &ipv6_devconf.mc_forwarding,
4425 .maxlen = sizeof(int),
4426 .mode = 0444,
4427 .proc_handler = proc_dointvec,
4429 #endif
4431 .procname = "disable_ipv6",
4432 .data = &ipv6_devconf.disable_ipv6,
4433 .maxlen = sizeof(int),
4434 .mode = 0644,
4435 .proc_handler = addrconf_sysctl_disable,
4438 .procname = "accept_dad",
4439 .data = &ipv6_devconf.accept_dad,
4440 .maxlen = sizeof(int),
4441 .mode = 0644,
4442 .proc_handler = proc_dointvec,
4445 .procname = "force_tllao",
4446 .data = &ipv6_devconf.force_tllao,
4447 .maxlen = sizeof(int),
4448 .mode = 0644,
4449 .proc_handler = proc_dointvec
4452 /* sentinel */
4457 static int __addrconf_sysctl_register(struct net *net, char *dev_name,
4458 struct inet6_dev *idev, struct ipv6_devconf *p)
4460 int i;
4461 struct addrconf_sysctl_table *t;
4463 #define ADDRCONF_CTL_PATH_DEV 3
4465 struct ctl_path addrconf_ctl_path[] = {
4466 { .procname = "net", },
4467 { .procname = "ipv6", },
4468 { .procname = "conf", },
4469 { /* to be set */ },
4470 { },
4474 t = kmemdup(&addrconf_sysctl, sizeof(*t), GFP_KERNEL);
4475 if (t == NULL)
4476 goto out;
4478 for (i = 0; t->addrconf_vars[i].data; i++) {
4479 t->addrconf_vars[i].data += (char *)p - (char *)&ipv6_devconf;
4480 t->addrconf_vars[i].extra1 = idev; /* embedded; no ref */
4481 t->addrconf_vars[i].extra2 = net;
4485 * Make a copy of dev_name, because '.procname' is regarded as const
4486 * by sysctl and we wouldn't want anyone to change it under our feet
4487 * (see SIOCSIFNAME).
4489 t->dev_name = kstrdup(dev_name, GFP_KERNEL);
4490 if (!t->dev_name)
4491 goto free;
4493 addrconf_ctl_path[ADDRCONF_CTL_PATH_DEV].procname = t->dev_name;
4495 t->sysctl_header = register_net_sysctl_table(net, addrconf_ctl_path,
4496 t->addrconf_vars);
4497 if (t->sysctl_header == NULL)
4498 goto free_procname;
4500 p->sysctl = t;
4501 return 0;
4503 free_procname:
4504 kfree(t->dev_name);
4505 free:
4506 kfree(t);
4507 out:
4508 return -ENOBUFS;
4511 static void __addrconf_sysctl_unregister(struct ipv6_devconf *p)
4513 struct addrconf_sysctl_table *t;
4515 if (p->sysctl == NULL)
4516 return;
4518 t = p->sysctl;
4519 p->sysctl = NULL;
4520 unregister_sysctl_table(t->sysctl_header);
4521 kfree(t->dev_name);
4522 kfree(t);
4525 static void addrconf_sysctl_register(struct inet6_dev *idev)
4527 neigh_sysctl_register(idev->dev, idev->nd_parms, "ipv6",
4528 &ndisc_ifinfo_sysctl_change);
4529 __addrconf_sysctl_register(dev_net(idev->dev), idev->dev->name,
4530 idev, &idev->cnf);
4533 static void addrconf_sysctl_unregister(struct inet6_dev *idev)
4535 __addrconf_sysctl_unregister(&idev->cnf);
4536 neigh_sysctl_unregister(idev->nd_parms);
4540 #endif
4542 static int __net_init addrconf_init_net(struct net *net)
4544 int err;
4545 struct ipv6_devconf *all, *dflt;
4547 err = -ENOMEM;
4548 all = &ipv6_devconf;
4549 dflt = &ipv6_devconf_dflt;
4551 if (!net_eq(net, &init_net)) {
4552 all = kmemdup(all, sizeof(ipv6_devconf), GFP_KERNEL);
4553 if (all == NULL)
4554 goto err_alloc_all;
4556 dflt = kmemdup(dflt, sizeof(ipv6_devconf_dflt), GFP_KERNEL);
4557 if (dflt == NULL)
4558 goto err_alloc_dflt;
4559 } else {
4560 /* these will be inherited by all namespaces */
4561 dflt->autoconf = ipv6_defaults.autoconf;
4562 dflt->disable_ipv6 = ipv6_defaults.disable_ipv6;
4565 net->ipv6.devconf_all = all;
4566 net->ipv6.devconf_dflt = dflt;
4568 #ifdef CONFIG_SYSCTL
4569 err = __addrconf_sysctl_register(net, "all", NULL, all);
4570 if (err < 0)
4571 goto err_reg_all;
4573 err = __addrconf_sysctl_register(net, "default", NULL, dflt);
4574 if (err < 0)
4575 goto err_reg_dflt;
4576 #endif
4577 return 0;
4579 #ifdef CONFIG_SYSCTL
4580 err_reg_dflt:
4581 __addrconf_sysctl_unregister(all);
4582 err_reg_all:
4583 kfree(dflt);
4584 #endif
4585 err_alloc_dflt:
4586 kfree(all);
4587 err_alloc_all:
4588 return err;
4591 static void __net_exit addrconf_exit_net(struct net *net)
4593 #ifdef CONFIG_SYSCTL
4594 __addrconf_sysctl_unregister(net->ipv6.devconf_dflt);
4595 __addrconf_sysctl_unregister(net->ipv6.devconf_all);
4596 #endif
4597 if (!net_eq(net, &init_net)) {
4598 kfree(net->ipv6.devconf_dflt);
4599 kfree(net->ipv6.devconf_all);
4603 static struct pernet_operations addrconf_ops = {
4604 .init = addrconf_init_net,
4605 .exit = addrconf_exit_net,
4609 * Device notifier
4612 int register_inet6addr_notifier(struct notifier_block *nb)
4614 return atomic_notifier_chain_register(&inet6addr_chain, nb);
4616 EXPORT_SYMBOL(register_inet6addr_notifier);
4618 int unregister_inet6addr_notifier(struct notifier_block *nb)
4620 return atomic_notifier_chain_unregister(&inet6addr_chain, nb);
4622 EXPORT_SYMBOL(unregister_inet6addr_notifier);
4625 * Init / cleanup code
4628 int __init addrconf_init(void)
4630 int i, err;
4632 err = ipv6_addr_label_init();
4633 if (err < 0) {
4634 printk(KERN_CRIT "IPv6 Addrconf:"
4635 " cannot initialize default policy table: %d.\n", err);
4636 goto out;
4639 err = register_pernet_subsys(&addrconf_ops);
4640 if (err < 0)
4641 goto out_addrlabel;
4643 /* The addrconf netdev notifier requires that loopback_dev
4644 * has it's ipv6 private information allocated and setup
4645 * before it can bring up and give link-local addresses
4646 * to other devices which are up.
4648 * Unfortunately, loopback_dev is not necessarily the first
4649 * entry in the global dev_base list of net devices. In fact,
4650 * it is likely to be the very last entry on that list.
4651 * So this causes the notifier registry below to try and
4652 * give link-local addresses to all devices besides loopback_dev
4653 * first, then loopback_dev, which cases all the non-loopback_dev
4654 * devices to fail to get a link-local address.
4656 * So, as a temporary fix, allocate the ipv6 structure for
4657 * loopback_dev first by hand.
4658 * Longer term, all of the dependencies ipv6 has upon the loopback
4659 * device and it being up should be removed.
4661 rtnl_lock();
4662 if (!ipv6_add_dev(init_net.loopback_dev))
4663 err = -ENOMEM;
4664 rtnl_unlock();
4665 if (err)
4666 goto errlo;
4668 for (i = 0; i < IN6_ADDR_HSIZE; i++)
4669 INIT_HLIST_HEAD(&inet6_addr_lst[i]);
4671 register_netdevice_notifier(&ipv6_dev_notf);
4673 addrconf_verify(0);
4675 err = __rtnl_register(PF_INET6, RTM_GETLINK, NULL, inet6_dump_ifinfo);
4676 if (err < 0)
4677 goto errout;
4679 /* Only the first call to __rtnl_register can fail */
4680 __rtnl_register(PF_INET6, RTM_NEWADDR, inet6_rtm_newaddr, NULL);
4681 __rtnl_register(PF_INET6, RTM_DELADDR, inet6_rtm_deladdr, NULL);
4682 __rtnl_register(PF_INET6, RTM_GETADDR, inet6_rtm_getaddr, inet6_dump_ifaddr);
4683 __rtnl_register(PF_INET6, RTM_GETMULTICAST, NULL, inet6_dump_ifmcaddr);
4684 __rtnl_register(PF_INET6, RTM_GETANYCAST, NULL, inet6_dump_ifacaddr);
4686 ipv6_addr_label_rtnl_register();
4688 return 0;
4689 errout:
4690 unregister_netdevice_notifier(&ipv6_dev_notf);
4691 errlo:
4692 unregister_pernet_subsys(&addrconf_ops);
4693 out_addrlabel:
4694 ipv6_addr_label_cleanup();
4695 out:
4696 return err;
4699 void addrconf_cleanup(void)
4701 struct net_device *dev;
4702 int i;
4704 unregister_netdevice_notifier(&ipv6_dev_notf);
4705 unregister_pernet_subsys(&addrconf_ops);
4706 ipv6_addr_label_cleanup();
4708 rtnl_lock();
4710 /* clean dev list */
4711 for_each_netdev(&init_net, dev) {
4712 if (__in6_dev_get(dev) == NULL)
4713 continue;
4714 addrconf_ifdown(dev, 1);
4716 addrconf_ifdown(init_net.loopback_dev, 2);
4719 * Check hash table.
4721 spin_lock_bh(&addrconf_hash_lock);
4722 for (i = 0; i < IN6_ADDR_HSIZE; i++)
4723 WARN_ON(!hlist_empty(&inet6_addr_lst[i]));
4724 spin_unlock_bh(&addrconf_hash_lock);
4726 del_timer(&addr_chk_timer);
4727 rtnl_unlock();