Linux 3.7.4
[linux/fpc-iii.git] / net / ipv6 / addrconf.c
blob0424e4e27414d3bd6c79f5094d0d3907e62b02f5
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 #define pr_fmt(fmt) "IPv6: " fmt
43 #include <linux/errno.h>
44 #include <linux/types.h>
45 #include <linux/kernel.h>
46 #include <linux/socket.h>
47 #include <linux/sockios.h>
48 #include <linux/net.h>
49 #include <linux/in6.h>
50 #include <linux/netdevice.h>
51 #include <linux/if_addr.h>
52 #include <linux/if_arp.h>
53 #include <linux/if_arcnet.h>
54 #include <linux/if_infiniband.h>
55 #include <linux/route.h>
56 #include <linux/inetdevice.h>
57 #include <linux/init.h>
58 #include <linux/slab.h>
59 #ifdef CONFIG_SYSCTL
60 #include <linux/sysctl.h>
61 #endif
62 #include <linux/capability.h>
63 #include <linux/delay.h>
64 #include <linux/notifier.h>
65 #include <linux/string.h>
66 #include <linux/hash.h>
68 #include <net/net_namespace.h>
69 #include <net/sock.h>
70 #include <net/snmp.h>
72 #include <net/af_ieee802154.h>
73 #include <net/ipv6.h>
74 #include <net/protocol.h>
75 #include <net/ndisc.h>
76 #include <net/ip6_route.h>
77 #include <net/addrconf.h>
78 #include <net/tcp.h>
79 #include <net/ip.h>
80 #include <net/netlink.h>
81 #include <net/pkt_sched.h>
82 #include <linux/if_tunnel.h>
83 #include <linux/rtnetlink.h>
85 #ifdef CONFIG_IPV6_PRIVACY
86 #include <linux/random.h>
87 #endif
89 #include <linux/uaccess.h>
90 #include <asm/unaligned.h>
92 #include <linux/proc_fs.h>
93 #include <linux/seq_file.h>
94 #include <linux/export.h>
96 /* Set to 3 to get tracing... */
97 #define ACONF_DEBUG 2
99 #if ACONF_DEBUG >= 3
100 #define ADBG(x) printk x
101 #else
102 #define ADBG(x)
103 #endif
105 #define INFINITY_LIFE_TIME 0xFFFFFFFF
107 static inline u32 cstamp_delta(unsigned long cstamp)
109 return (cstamp - INITIAL_JIFFIES) * 100UL / HZ;
112 #define ADDRCONF_TIMER_FUZZ_MINUS (HZ > 50 ? HZ/50 : 1)
113 #define ADDRCONF_TIMER_FUZZ (HZ / 4)
114 #define ADDRCONF_TIMER_FUZZ_MAX (HZ)
116 #ifdef CONFIG_SYSCTL
117 static void addrconf_sysctl_register(struct inet6_dev *idev);
118 static void addrconf_sysctl_unregister(struct inet6_dev *idev);
119 #else
120 static inline void addrconf_sysctl_register(struct inet6_dev *idev)
124 static inline void addrconf_sysctl_unregister(struct inet6_dev *idev)
127 #endif
129 #ifdef CONFIG_IPV6_PRIVACY
130 static void __ipv6_regen_rndid(struct inet6_dev *idev);
131 static void __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr);
132 static void ipv6_regen_rndid(unsigned long data);
133 #endif
135 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev);
136 static int ipv6_count_addresses(struct inet6_dev *idev);
139 * Configured unicast address hash table
141 static struct hlist_head inet6_addr_lst[IN6_ADDR_HSIZE];
142 static DEFINE_SPINLOCK(addrconf_hash_lock);
144 static void addrconf_verify(unsigned long);
146 static DEFINE_TIMER(addr_chk_timer, addrconf_verify, 0, 0);
147 static DEFINE_SPINLOCK(addrconf_verify_lock);
149 static void addrconf_join_anycast(struct inet6_ifaddr *ifp);
150 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp);
152 static void addrconf_type_change(struct net_device *dev,
153 unsigned long event);
154 static int addrconf_ifdown(struct net_device *dev, int how);
156 static void addrconf_dad_start(struct inet6_ifaddr *ifp);
157 static void addrconf_dad_timer(unsigned long data);
158 static void addrconf_dad_completed(struct inet6_ifaddr *ifp);
159 static void addrconf_dad_run(struct inet6_dev *idev);
160 static void addrconf_rs_timer(unsigned long data);
161 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
162 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
164 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
165 struct prefix_info *pinfo);
166 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
167 struct net_device *dev);
169 static ATOMIC_NOTIFIER_HEAD(inet6addr_chain);
171 static struct ipv6_devconf ipv6_devconf __read_mostly = {
172 .forwarding = 0,
173 .hop_limit = IPV6_DEFAULT_HOPLIMIT,
174 .mtu6 = IPV6_MIN_MTU,
175 .accept_ra = 1,
176 .accept_redirects = 1,
177 .autoconf = 1,
178 .force_mld_version = 0,
179 .dad_transmits = 1,
180 .rtr_solicits = MAX_RTR_SOLICITATIONS,
181 .rtr_solicit_interval = RTR_SOLICITATION_INTERVAL,
182 .rtr_solicit_delay = MAX_RTR_SOLICITATION_DELAY,
183 #ifdef CONFIG_IPV6_PRIVACY
184 .use_tempaddr = 0,
185 .temp_valid_lft = TEMP_VALID_LIFETIME,
186 .temp_prefered_lft = TEMP_PREFERRED_LIFETIME,
187 .regen_max_retry = REGEN_MAX_RETRY,
188 .max_desync_factor = MAX_DESYNC_FACTOR,
189 #endif
190 .max_addresses = IPV6_MAX_ADDRESSES,
191 .accept_ra_defrtr = 1,
192 .accept_ra_pinfo = 1,
193 #ifdef CONFIG_IPV6_ROUTER_PREF
194 .accept_ra_rtr_pref = 1,
195 .rtr_probe_interval = 60 * HZ,
196 #ifdef CONFIG_IPV6_ROUTE_INFO
197 .accept_ra_rt_info_max_plen = 0,
198 #endif
199 #endif
200 .proxy_ndp = 0,
201 .accept_source_route = 0, /* we do not accept RH0 by default. */
202 .disable_ipv6 = 0,
203 .accept_dad = 1,
206 static struct ipv6_devconf ipv6_devconf_dflt __read_mostly = {
207 .forwarding = 0,
208 .hop_limit = IPV6_DEFAULT_HOPLIMIT,
209 .mtu6 = IPV6_MIN_MTU,
210 .accept_ra = 1,
211 .accept_redirects = 1,
212 .autoconf = 1,
213 .dad_transmits = 1,
214 .rtr_solicits = MAX_RTR_SOLICITATIONS,
215 .rtr_solicit_interval = RTR_SOLICITATION_INTERVAL,
216 .rtr_solicit_delay = MAX_RTR_SOLICITATION_DELAY,
217 #ifdef CONFIG_IPV6_PRIVACY
218 .use_tempaddr = 0,
219 .temp_valid_lft = TEMP_VALID_LIFETIME,
220 .temp_prefered_lft = TEMP_PREFERRED_LIFETIME,
221 .regen_max_retry = REGEN_MAX_RETRY,
222 .max_desync_factor = MAX_DESYNC_FACTOR,
223 #endif
224 .max_addresses = IPV6_MAX_ADDRESSES,
225 .accept_ra_defrtr = 1,
226 .accept_ra_pinfo = 1,
227 #ifdef CONFIG_IPV6_ROUTER_PREF
228 .accept_ra_rtr_pref = 1,
229 .rtr_probe_interval = 60 * HZ,
230 #ifdef CONFIG_IPV6_ROUTE_INFO
231 .accept_ra_rt_info_max_plen = 0,
232 #endif
233 #endif
234 .proxy_ndp = 0,
235 .accept_source_route = 0, /* we do not accept RH0 by default. */
236 .disable_ipv6 = 0,
237 .accept_dad = 1,
240 /* IPv6 Wildcard Address and Loopback Address defined by RFC2553 */
241 const struct in6_addr in6addr_any = IN6ADDR_ANY_INIT;
242 const struct in6_addr in6addr_loopback = IN6ADDR_LOOPBACK_INIT;
243 const struct in6_addr in6addr_linklocal_allnodes = IN6ADDR_LINKLOCAL_ALLNODES_INIT;
244 const struct in6_addr in6addr_linklocal_allrouters = IN6ADDR_LINKLOCAL_ALLROUTERS_INIT;
246 /* Check if a valid qdisc is available */
247 static inline bool addrconf_qdisc_ok(const struct net_device *dev)
249 return !qdisc_tx_is_noop(dev);
252 /* Check if a route is valid prefix route */
253 static inline int addrconf_is_prefix_route(const struct rt6_info *rt)
255 return (rt->rt6i_flags & (RTF_GATEWAY | RTF_DEFAULT)) == 0;
258 static void addrconf_del_timer(struct inet6_ifaddr *ifp)
260 if (del_timer(&ifp->timer))
261 __in6_ifa_put(ifp);
264 enum addrconf_timer_t {
265 AC_NONE,
266 AC_DAD,
267 AC_RS,
270 static void addrconf_mod_timer(struct inet6_ifaddr *ifp,
271 enum addrconf_timer_t what,
272 unsigned long when)
274 if (!del_timer(&ifp->timer))
275 in6_ifa_hold(ifp);
277 switch (what) {
278 case AC_DAD:
279 ifp->timer.function = addrconf_dad_timer;
280 break;
281 case AC_RS:
282 ifp->timer.function = addrconf_rs_timer;
283 break;
284 default:
285 break;
287 ifp->timer.expires = jiffies + when;
288 add_timer(&ifp->timer);
291 static int snmp6_alloc_dev(struct inet6_dev *idev)
293 if (snmp_mib_init((void __percpu **)idev->stats.ipv6,
294 sizeof(struct ipstats_mib),
295 __alignof__(struct ipstats_mib)) < 0)
296 goto err_ip;
297 idev->stats.icmpv6dev = kzalloc(sizeof(struct icmpv6_mib_device),
298 GFP_KERNEL);
299 if (!idev->stats.icmpv6dev)
300 goto err_icmp;
301 idev->stats.icmpv6msgdev = kzalloc(sizeof(struct icmpv6msg_mib_device),
302 GFP_KERNEL);
303 if (!idev->stats.icmpv6msgdev)
304 goto err_icmpmsg;
306 return 0;
308 err_icmpmsg:
309 kfree(idev->stats.icmpv6dev);
310 err_icmp:
311 snmp_mib_free((void __percpu **)idev->stats.ipv6);
312 err_ip:
313 return -ENOMEM;
316 static void snmp6_free_dev(struct inet6_dev *idev)
318 kfree(idev->stats.icmpv6msgdev);
319 kfree(idev->stats.icmpv6dev);
320 snmp_mib_free((void __percpu **)idev->stats.ipv6);
323 /* Nobody refers to this device, we may destroy it. */
325 void in6_dev_finish_destroy(struct inet6_dev *idev)
327 struct net_device *dev = idev->dev;
329 WARN_ON(!list_empty(&idev->addr_list));
330 WARN_ON(idev->mc_list != NULL);
332 #ifdef NET_REFCNT_DEBUG
333 pr_debug("%s: %s\n", __func__, dev ? dev->name : "NIL");
334 #endif
335 dev_put(dev);
336 if (!idev->dead) {
337 pr_warn("Freeing alive inet6 device %p\n", idev);
338 return;
340 snmp6_free_dev(idev);
341 kfree_rcu(idev, rcu);
343 EXPORT_SYMBOL(in6_dev_finish_destroy);
345 static struct inet6_dev *ipv6_add_dev(struct net_device *dev)
347 struct inet6_dev *ndev;
349 ASSERT_RTNL();
351 if (dev->mtu < IPV6_MIN_MTU)
352 return NULL;
354 ndev = kzalloc(sizeof(struct inet6_dev), GFP_KERNEL);
356 if (ndev == NULL)
357 return NULL;
359 rwlock_init(&ndev->lock);
360 ndev->dev = dev;
361 INIT_LIST_HEAD(&ndev->addr_list);
363 memcpy(&ndev->cnf, dev_net(dev)->ipv6.devconf_dflt, sizeof(ndev->cnf));
364 ndev->cnf.mtu6 = dev->mtu;
365 ndev->cnf.sysctl = NULL;
366 ndev->nd_parms = neigh_parms_alloc(dev, &nd_tbl);
367 if (ndev->nd_parms == NULL) {
368 kfree(ndev);
369 return NULL;
371 if (ndev->cnf.forwarding)
372 dev_disable_lro(dev);
373 /* We refer to the device */
374 dev_hold(dev);
376 if (snmp6_alloc_dev(ndev) < 0) {
377 ADBG((KERN_WARNING
378 "%s: cannot allocate memory for statistics; dev=%s.\n",
379 __func__, dev->name));
380 neigh_parms_release(&nd_tbl, ndev->nd_parms);
381 dev_put(dev);
382 kfree(ndev);
383 return NULL;
386 if (snmp6_register_dev(ndev) < 0) {
387 ADBG((KERN_WARNING
388 "%s: cannot create /proc/net/dev_snmp6/%s\n",
389 __func__, dev->name));
390 neigh_parms_release(&nd_tbl, ndev->nd_parms);
391 ndev->dead = 1;
392 in6_dev_finish_destroy(ndev);
393 return NULL;
396 /* One reference from device. We must do this before
397 * we invoke __ipv6_regen_rndid().
399 in6_dev_hold(ndev);
401 if (dev->flags & (IFF_NOARP | IFF_LOOPBACK))
402 ndev->cnf.accept_dad = -1;
404 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
405 if (dev->type == ARPHRD_SIT && (dev->priv_flags & IFF_ISATAP)) {
406 pr_info("%s: Disabled Multicast RS\n", 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 ndev->cnf.use_tempaddr = -1;
420 } else {
421 in6_dev_hold(ndev);
422 ipv6_regen_rndid((unsigned long) ndev);
424 #endif
426 if (netif_running(dev) && addrconf_qdisc_ok(dev))
427 ndev->if_flags |= IF_READY;
429 ipv6_mc_init_dev(ndev);
430 ndev->tstamp = jiffies;
431 addrconf_sysctl_register(ndev);
432 /* protected by rtnl_lock */
433 rcu_assign_pointer(dev->ip6_ptr, ndev);
435 /* Join all-node multicast group */
436 ipv6_dev_mc_inc(dev, &in6addr_linklocal_allnodes);
438 /* Join all-router multicast group if forwarding is set */
439 if (ndev->cnf.forwarding && (dev->flags & IFF_MULTICAST))
440 ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
442 return ndev;
445 static struct inet6_dev *ipv6_find_idev(struct net_device *dev)
447 struct inet6_dev *idev;
449 ASSERT_RTNL();
451 idev = __in6_dev_get(dev);
452 if (!idev) {
453 idev = ipv6_add_dev(dev);
454 if (!idev)
455 return NULL;
458 if (dev->flags&IFF_UP)
459 ipv6_mc_up(idev);
460 return idev;
463 #ifdef CONFIG_SYSCTL
464 static void dev_forward_change(struct inet6_dev *idev)
466 struct net_device *dev;
467 struct inet6_ifaddr *ifa;
469 if (!idev)
470 return;
471 dev = idev->dev;
472 if (idev->cnf.forwarding)
473 dev_disable_lro(dev);
474 if (dev && (dev->flags & IFF_MULTICAST)) {
475 if (idev->cnf.forwarding)
476 ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
477 else
478 ipv6_dev_mc_dec(dev, &in6addr_linklocal_allrouters);
481 list_for_each_entry(ifa, &idev->addr_list, if_list) {
482 if (ifa->flags&IFA_F_TENTATIVE)
483 continue;
484 if (idev->cnf.forwarding)
485 addrconf_join_anycast(ifa);
486 else
487 addrconf_leave_anycast(ifa);
492 static void addrconf_forward_change(struct net *net, __s32 newf)
494 struct net_device *dev;
495 struct inet6_dev *idev;
497 for_each_netdev(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);
508 static int addrconf_fixup_forwarding(struct ctl_table *table, int *p, int newf)
510 struct net *net;
511 int old;
513 if (!rtnl_trylock())
514 return restart_syscall();
516 net = (struct net *)table->extra2;
517 old = *p;
518 *p = newf;
520 if (p == &net->ipv6.devconf_dflt->forwarding) {
521 rtnl_unlock();
522 return 0;
525 if (p == &net->ipv6.devconf_all->forwarding) {
526 net->ipv6.devconf_dflt->forwarding = newf;
527 addrconf_forward_change(net, newf);
528 } else if ((!newf) ^ (!old))
529 dev_forward_change((struct inet6_dev *)table->extra1);
530 rtnl_unlock();
532 if (newf)
533 rt6_purge_dflt_routers(net);
534 return 1;
536 #endif
538 /* Nobody refers to this ifaddr, destroy it */
539 void inet6_ifa_finish_destroy(struct inet6_ifaddr *ifp)
541 WARN_ON(!hlist_unhashed(&ifp->addr_lst));
543 #ifdef NET_REFCNT_DEBUG
544 pr_debug("%s\n", __func__);
545 #endif
547 in6_dev_put(ifp->idev);
549 if (del_timer(&ifp->timer))
550 pr_notice("Timer is still running, when freeing ifa=%p\n", ifp);
552 if (ifp->state != INET6_IFADDR_STATE_DEAD) {
553 pr_warn("Freeing alive inet6 address %p\n", ifp);
554 return;
556 dst_release(&ifp->rt->dst);
558 kfree_rcu(ifp, rcu);
561 static void
562 ipv6_link_dev_addr(struct inet6_dev *idev, struct inet6_ifaddr *ifp)
564 struct list_head *p;
565 int ifp_scope = ipv6_addr_src_scope(&ifp->addr);
568 * Each device address list is sorted in order of scope -
569 * global before linklocal.
571 list_for_each(p, &idev->addr_list) {
572 struct inet6_ifaddr *ifa
573 = list_entry(p, struct inet6_ifaddr, if_list);
574 if (ifp_scope >= ipv6_addr_src_scope(&ifa->addr))
575 break;
578 list_add_tail(&ifp->if_list, p);
581 static u32 inet6_addr_hash(const struct in6_addr *addr)
583 return hash_32(ipv6_addr_hash(addr), IN6_ADDR_HSIZE_SHIFT);
586 /* On success it returns ifp with increased reference count */
588 static struct inet6_ifaddr *
589 ipv6_add_addr(struct inet6_dev *idev, const struct in6_addr *addr, int pfxlen,
590 int scope, u32 flags)
592 struct inet6_ifaddr *ifa = NULL;
593 struct rt6_info *rt;
594 unsigned int hash;
595 int err = 0;
596 int addr_type = ipv6_addr_type(addr);
598 if (addr_type == IPV6_ADDR_ANY ||
599 addr_type & IPV6_ADDR_MULTICAST ||
600 (!(idev->dev->flags & IFF_LOOPBACK) &&
601 addr_type & IPV6_ADDR_LOOPBACK))
602 return ERR_PTR(-EADDRNOTAVAIL);
604 rcu_read_lock_bh();
605 if (idev->dead) {
606 err = -ENODEV; /*XXX*/
607 goto out2;
610 if (idev->cnf.disable_ipv6) {
611 err = -EACCES;
612 goto out2;
615 spin_lock(&addrconf_hash_lock);
617 /* Ignore adding duplicate addresses on an interface */
618 if (ipv6_chk_same_addr(dev_net(idev->dev), addr, idev->dev)) {
619 ADBG(("ipv6_add_addr: already assigned\n"));
620 err = -EEXIST;
621 goto out;
624 ifa = kzalloc(sizeof(struct inet6_ifaddr), GFP_ATOMIC);
626 if (ifa == NULL) {
627 ADBG(("ipv6_add_addr: malloc failed\n"));
628 err = -ENOBUFS;
629 goto out;
632 rt = addrconf_dst_alloc(idev, addr, false);
633 if (IS_ERR(rt)) {
634 err = PTR_ERR(rt);
635 goto out;
638 ifa->addr = *addr;
640 spin_lock_init(&ifa->lock);
641 spin_lock_init(&ifa->state_lock);
642 init_timer(&ifa->timer);
643 INIT_HLIST_NODE(&ifa->addr_lst);
644 ifa->timer.data = (unsigned long) ifa;
645 ifa->scope = scope;
646 ifa->prefix_len = pfxlen;
647 ifa->flags = flags | IFA_F_TENTATIVE;
648 ifa->cstamp = ifa->tstamp = jiffies;
650 ifa->rt = rt;
652 ifa->idev = idev;
653 in6_dev_hold(idev);
654 /* For caller */
655 in6_ifa_hold(ifa);
657 /* Add to big hash table */
658 hash = inet6_addr_hash(addr);
660 hlist_add_head_rcu(&ifa->addr_lst, &inet6_addr_lst[hash]);
661 spin_unlock(&addrconf_hash_lock);
663 write_lock(&idev->lock);
664 /* Add to inet6_dev unicast addr list. */
665 ipv6_link_dev_addr(idev, ifa);
667 #ifdef CONFIG_IPV6_PRIVACY
668 if (ifa->flags&IFA_F_TEMPORARY) {
669 list_add(&ifa->tmp_list, &idev->tempaddr_list);
670 in6_ifa_hold(ifa);
672 #endif
674 in6_ifa_hold(ifa);
675 write_unlock(&idev->lock);
676 out2:
677 rcu_read_unlock_bh();
679 if (likely(err == 0))
680 atomic_notifier_call_chain(&inet6addr_chain, NETDEV_UP, ifa);
681 else {
682 kfree(ifa);
683 ifa = ERR_PTR(err);
686 return ifa;
687 out:
688 spin_unlock(&addrconf_hash_lock);
689 goto out2;
692 /* This function wants to get referenced ifp and releases it before return */
694 static void ipv6_del_addr(struct inet6_ifaddr *ifp)
696 struct inet6_ifaddr *ifa, *ifn;
697 struct inet6_dev *idev = ifp->idev;
698 int state;
699 int deleted = 0, onlink = 0;
700 unsigned long expires = jiffies;
702 spin_lock_bh(&ifp->state_lock);
703 state = ifp->state;
704 ifp->state = INET6_IFADDR_STATE_DEAD;
705 spin_unlock_bh(&ifp->state_lock);
707 if (state == INET6_IFADDR_STATE_DEAD)
708 goto out;
710 spin_lock_bh(&addrconf_hash_lock);
711 hlist_del_init_rcu(&ifp->addr_lst);
712 spin_unlock_bh(&addrconf_hash_lock);
714 write_lock_bh(&idev->lock);
715 #ifdef CONFIG_IPV6_PRIVACY
716 if (ifp->flags&IFA_F_TEMPORARY) {
717 list_del(&ifp->tmp_list);
718 if (ifp->ifpub) {
719 in6_ifa_put(ifp->ifpub);
720 ifp->ifpub = NULL;
722 __in6_ifa_put(ifp);
724 #endif
726 list_for_each_entry_safe(ifa, ifn, &idev->addr_list, if_list) {
727 if (ifa == ifp) {
728 list_del_init(&ifp->if_list);
729 __in6_ifa_put(ifp);
731 if (!(ifp->flags & IFA_F_PERMANENT) || onlink > 0)
732 break;
733 deleted = 1;
734 continue;
735 } else if (ifp->flags & IFA_F_PERMANENT) {
736 if (ipv6_prefix_equal(&ifa->addr, &ifp->addr,
737 ifp->prefix_len)) {
738 if (ifa->flags & IFA_F_PERMANENT) {
739 onlink = 1;
740 if (deleted)
741 break;
742 } else {
743 unsigned long lifetime;
745 if (!onlink)
746 onlink = -1;
748 spin_lock(&ifa->lock);
750 lifetime = addrconf_timeout_fixup(ifa->valid_lft, HZ);
752 * Note: Because this address is
753 * not permanent, lifetime <
754 * LONG_MAX / HZ here.
756 if (time_before(expires,
757 ifa->tstamp + lifetime * HZ))
758 expires = ifa->tstamp + lifetime * HZ;
759 spin_unlock(&ifa->lock);
764 write_unlock_bh(&idev->lock);
766 addrconf_del_timer(ifp);
768 ipv6_ifa_notify(RTM_DELADDR, ifp);
770 atomic_notifier_call_chain(&inet6addr_chain, NETDEV_DOWN, ifp);
773 * Purge or update corresponding prefix
775 * 1) we don't purge prefix here if address was not permanent.
776 * prefix is managed by its own lifetime.
777 * 2) if there're no addresses, delete prefix.
778 * 3) if there're still other permanent address(es),
779 * corresponding prefix is still permanent.
780 * 4) otherwise, update prefix lifetime to the
781 * longest valid lifetime among the corresponding
782 * addresses on the device.
783 * Note: subsequent RA will update lifetime.
785 * --yoshfuji
787 if ((ifp->flags & IFA_F_PERMANENT) && onlink < 1) {
788 struct in6_addr prefix;
789 struct rt6_info *rt;
790 struct net *net = dev_net(ifp->idev->dev);
791 struct flowi6 fl6 = {};
793 ipv6_addr_prefix(&prefix, &ifp->addr, ifp->prefix_len);
794 fl6.flowi6_oif = ifp->idev->dev->ifindex;
795 fl6.daddr = prefix;
796 rt = (struct rt6_info *)ip6_route_lookup(net, &fl6,
797 RT6_LOOKUP_F_IFACE);
799 if (rt != net->ipv6.ip6_null_entry &&
800 addrconf_is_prefix_route(rt)) {
801 if (onlink == 0) {
802 ip6_del_rt(rt);
803 rt = NULL;
804 } else if (!(rt->rt6i_flags & RTF_EXPIRES)) {
805 rt6_set_expires(rt, expires);
808 dst_release(&rt->dst);
811 /* clean up prefsrc entries */
812 rt6_remove_prefsrc(ifp);
813 out:
814 in6_ifa_put(ifp);
817 #ifdef CONFIG_IPV6_PRIVACY
818 static int ipv6_create_tempaddr(struct inet6_ifaddr *ifp, struct inet6_ifaddr *ift)
820 struct inet6_dev *idev = ifp->idev;
821 struct in6_addr addr, *tmpaddr;
822 unsigned long tmp_prefered_lft, tmp_valid_lft, tmp_tstamp, age;
823 unsigned long regen_advance;
824 int tmp_plen;
825 int ret = 0;
826 int max_addresses;
827 u32 addr_flags;
828 unsigned long now = jiffies;
830 write_lock(&idev->lock);
831 if (ift) {
832 spin_lock_bh(&ift->lock);
833 memcpy(&addr.s6_addr[8], &ift->addr.s6_addr[8], 8);
834 spin_unlock_bh(&ift->lock);
835 tmpaddr = &addr;
836 } else {
837 tmpaddr = NULL;
839 retry:
840 in6_dev_hold(idev);
841 if (idev->cnf.use_tempaddr <= 0) {
842 write_unlock(&idev->lock);
843 pr_info("%s: use_tempaddr is disabled\n", __func__);
844 in6_dev_put(idev);
845 ret = -1;
846 goto out;
848 spin_lock_bh(&ifp->lock);
849 if (ifp->regen_count++ >= idev->cnf.regen_max_retry) {
850 idev->cnf.use_tempaddr = -1; /*XXX*/
851 spin_unlock_bh(&ifp->lock);
852 write_unlock(&idev->lock);
853 pr_warn("%s: regeneration time exceeded - disabled temporary address support\n",
854 __func__);
855 in6_dev_put(idev);
856 ret = -1;
857 goto out;
859 in6_ifa_hold(ifp);
860 memcpy(addr.s6_addr, ifp->addr.s6_addr, 8);
861 __ipv6_try_regen_rndid(idev, tmpaddr);
862 memcpy(&addr.s6_addr[8], idev->rndid, 8);
863 age = (now - ifp->tstamp) / HZ;
864 tmp_valid_lft = min_t(__u32,
865 ifp->valid_lft,
866 idev->cnf.temp_valid_lft + age);
867 tmp_prefered_lft = min_t(__u32,
868 ifp->prefered_lft,
869 idev->cnf.temp_prefered_lft + age -
870 idev->cnf.max_desync_factor);
871 tmp_plen = ifp->prefix_len;
872 max_addresses = idev->cnf.max_addresses;
873 tmp_tstamp = ifp->tstamp;
874 spin_unlock_bh(&ifp->lock);
876 regen_advance = idev->cnf.regen_max_retry *
877 idev->cnf.dad_transmits *
878 idev->nd_parms->retrans_time / HZ;
879 write_unlock(&idev->lock);
881 /* A temporary address is created only if this calculated Preferred
882 * Lifetime is greater than REGEN_ADVANCE time units. In particular,
883 * an implementation must not create a temporary address with a zero
884 * Preferred Lifetime.
886 if (tmp_prefered_lft <= regen_advance) {
887 in6_ifa_put(ifp);
888 in6_dev_put(idev);
889 ret = -1;
890 goto out;
893 addr_flags = IFA_F_TEMPORARY;
894 /* set in addrconf_prefix_rcv() */
895 if (ifp->flags & IFA_F_OPTIMISTIC)
896 addr_flags |= IFA_F_OPTIMISTIC;
898 ift = !max_addresses ||
899 ipv6_count_addresses(idev) < max_addresses ?
900 ipv6_add_addr(idev, &addr, tmp_plen,
901 ipv6_addr_type(&addr)&IPV6_ADDR_SCOPE_MASK,
902 addr_flags) : NULL;
903 if (!ift || IS_ERR(ift)) {
904 in6_ifa_put(ifp);
905 in6_dev_put(idev);
906 pr_info("%s: retry temporary address regeneration\n", __func__);
907 tmpaddr = &addr;
908 write_lock(&idev->lock);
909 goto retry;
912 spin_lock_bh(&ift->lock);
913 ift->ifpub = ifp;
914 ift->valid_lft = tmp_valid_lft;
915 ift->prefered_lft = tmp_prefered_lft;
916 ift->cstamp = now;
917 ift->tstamp = tmp_tstamp;
918 spin_unlock_bh(&ift->lock);
920 addrconf_dad_start(ift);
921 in6_ifa_put(ift);
922 in6_dev_put(idev);
923 out:
924 return ret;
926 #endif
929 * Choose an appropriate source address (RFC3484)
931 enum {
932 IPV6_SADDR_RULE_INIT = 0,
933 IPV6_SADDR_RULE_LOCAL,
934 IPV6_SADDR_RULE_SCOPE,
935 IPV6_SADDR_RULE_PREFERRED,
936 #ifdef CONFIG_IPV6_MIP6
937 IPV6_SADDR_RULE_HOA,
938 #endif
939 IPV6_SADDR_RULE_OIF,
940 IPV6_SADDR_RULE_LABEL,
941 #ifdef CONFIG_IPV6_PRIVACY
942 IPV6_SADDR_RULE_PRIVACY,
943 #endif
944 IPV6_SADDR_RULE_ORCHID,
945 IPV6_SADDR_RULE_PREFIX,
946 IPV6_SADDR_RULE_MAX
949 struct ipv6_saddr_score {
950 int rule;
951 int addr_type;
952 struct inet6_ifaddr *ifa;
953 DECLARE_BITMAP(scorebits, IPV6_SADDR_RULE_MAX);
954 int scopedist;
955 int matchlen;
958 struct ipv6_saddr_dst {
959 const struct in6_addr *addr;
960 int ifindex;
961 int scope;
962 int label;
963 unsigned int prefs;
966 static inline int ipv6_saddr_preferred(int type)
968 if (type & (IPV6_ADDR_MAPPED|IPV6_ADDR_COMPATv4|IPV6_ADDR_LOOPBACK))
969 return 1;
970 return 0;
973 static int ipv6_get_saddr_eval(struct net *net,
974 struct ipv6_saddr_score *score,
975 struct ipv6_saddr_dst *dst,
976 int i)
978 int ret;
980 if (i <= score->rule) {
981 switch (i) {
982 case IPV6_SADDR_RULE_SCOPE:
983 ret = score->scopedist;
984 break;
985 case IPV6_SADDR_RULE_PREFIX:
986 ret = score->matchlen;
987 break;
988 default:
989 ret = !!test_bit(i, score->scorebits);
991 goto out;
994 switch (i) {
995 case IPV6_SADDR_RULE_INIT:
996 /* Rule 0: remember if hiscore is not ready yet */
997 ret = !!score->ifa;
998 break;
999 case IPV6_SADDR_RULE_LOCAL:
1000 /* Rule 1: Prefer same address */
1001 ret = ipv6_addr_equal(&score->ifa->addr, dst->addr);
1002 break;
1003 case IPV6_SADDR_RULE_SCOPE:
1004 /* Rule 2: Prefer appropriate scope
1006 * ret
1008 * -1 | d 15
1009 * ---+--+-+---> scope
1011 * | d is scope of the destination.
1012 * B-d | \
1013 * | \ <- smaller scope is better if
1014 * B-15 | \ if scope is enough for destinaion.
1015 * | ret = B - scope (-1 <= scope >= d <= 15).
1016 * d-C-1 | /
1017 * |/ <- greater is better
1018 * -C / if scope is not enough for destination.
1019 * /| ret = scope - C (-1 <= d < scope <= 15).
1021 * d - C - 1 < B -15 (for all -1 <= d <= 15).
1022 * C > d + 14 - B >= 15 + 14 - B = 29 - B.
1023 * Assume B = 0 and we get C > 29.
1025 ret = __ipv6_addr_src_scope(score->addr_type);
1026 if (ret >= dst->scope)
1027 ret = -ret;
1028 else
1029 ret -= 128; /* 30 is enough */
1030 score->scopedist = ret;
1031 break;
1032 case IPV6_SADDR_RULE_PREFERRED:
1033 /* Rule 3: Avoid deprecated and optimistic addresses */
1034 ret = ipv6_saddr_preferred(score->addr_type) ||
1035 !(score->ifa->flags & (IFA_F_DEPRECATED|IFA_F_OPTIMISTIC));
1036 break;
1037 #ifdef CONFIG_IPV6_MIP6
1038 case IPV6_SADDR_RULE_HOA:
1040 /* Rule 4: Prefer home address */
1041 int prefhome = !(dst->prefs & IPV6_PREFER_SRC_COA);
1042 ret = !(score->ifa->flags & IFA_F_HOMEADDRESS) ^ prefhome;
1043 break;
1045 #endif
1046 case IPV6_SADDR_RULE_OIF:
1047 /* Rule 5: Prefer outgoing interface */
1048 ret = (!dst->ifindex ||
1049 dst->ifindex == score->ifa->idev->dev->ifindex);
1050 break;
1051 case IPV6_SADDR_RULE_LABEL:
1052 /* Rule 6: Prefer matching label */
1053 ret = ipv6_addr_label(net,
1054 &score->ifa->addr, score->addr_type,
1055 score->ifa->idev->dev->ifindex) == dst->label;
1056 break;
1057 #ifdef CONFIG_IPV6_PRIVACY
1058 case IPV6_SADDR_RULE_PRIVACY:
1060 /* Rule 7: Prefer public address
1061 * Note: prefer temporary address if use_tempaddr >= 2
1063 int preftmp = dst->prefs & (IPV6_PREFER_SRC_PUBLIC|IPV6_PREFER_SRC_TMP) ?
1064 !!(dst->prefs & IPV6_PREFER_SRC_TMP) :
1065 score->ifa->idev->cnf.use_tempaddr >= 2;
1066 ret = (!(score->ifa->flags & IFA_F_TEMPORARY)) ^ preftmp;
1067 break;
1069 #endif
1070 case IPV6_SADDR_RULE_ORCHID:
1071 /* Rule 8-: Prefer ORCHID vs ORCHID or
1072 * non-ORCHID vs non-ORCHID
1074 ret = !(ipv6_addr_orchid(&score->ifa->addr) ^
1075 ipv6_addr_orchid(dst->addr));
1076 break;
1077 case IPV6_SADDR_RULE_PREFIX:
1078 /* Rule 8: Use longest matching prefix */
1079 ret = ipv6_addr_diff(&score->ifa->addr, dst->addr);
1080 if (ret > score->ifa->prefix_len)
1081 ret = score->ifa->prefix_len;
1082 score->matchlen = ret;
1083 break;
1084 default:
1085 ret = 0;
1088 if (ret)
1089 __set_bit(i, score->scorebits);
1090 score->rule = i;
1091 out:
1092 return ret;
1095 int ipv6_dev_get_saddr(struct net *net, const struct net_device *dst_dev,
1096 const struct in6_addr *daddr, unsigned int prefs,
1097 struct in6_addr *saddr)
1099 struct ipv6_saddr_score scores[2],
1100 *score = &scores[0], *hiscore = &scores[1];
1101 struct ipv6_saddr_dst dst;
1102 struct net_device *dev;
1103 int dst_type;
1105 dst_type = __ipv6_addr_type(daddr);
1106 dst.addr = daddr;
1107 dst.ifindex = dst_dev ? dst_dev->ifindex : 0;
1108 dst.scope = __ipv6_addr_src_scope(dst_type);
1109 dst.label = ipv6_addr_label(net, daddr, dst_type, dst.ifindex);
1110 dst.prefs = prefs;
1112 hiscore->rule = -1;
1113 hiscore->ifa = NULL;
1115 rcu_read_lock();
1117 for_each_netdev_rcu(net, dev) {
1118 struct inet6_dev *idev;
1120 /* Candidate Source Address (section 4)
1121 * - multicast and link-local destination address,
1122 * the set of candidate source address MUST only
1123 * include addresses assigned to interfaces
1124 * belonging to the same link as the outgoing
1125 * interface.
1126 * (- For site-local destination addresses, the
1127 * set of candidate source addresses MUST only
1128 * include addresses assigned to interfaces
1129 * belonging to the same site as the outgoing
1130 * interface.)
1132 if (((dst_type & IPV6_ADDR_MULTICAST) ||
1133 dst.scope <= IPV6_ADDR_SCOPE_LINKLOCAL) &&
1134 dst.ifindex && dev->ifindex != dst.ifindex)
1135 continue;
1137 idev = __in6_dev_get(dev);
1138 if (!idev)
1139 continue;
1141 read_lock_bh(&idev->lock);
1142 list_for_each_entry(score->ifa, &idev->addr_list, if_list) {
1143 int i;
1146 * - Tentative Address (RFC2462 section 5.4)
1147 * - A tentative address is not considered
1148 * "assigned to an interface" in the traditional
1149 * sense, unless it is also flagged as optimistic.
1150 * - Candidate Source Address (section 4)
1151 * - In any case, anycast addresses, multicast
1152 * addresses, and the unspecified address MUST
1153 * NOT be included in a candidate set.
1155 if ((score->ifa->flags & IFA_F_TENTATIVE) &&
1156 (!(score->ifa->flags & IFA_F_OPTIMISTIC)))
1157 continue;
1159 score->addr_type = __ipv6_addr_type(&score->ifa->addr);
1161 if (unlikely(score->addr_type == IPV6_ADDR_ANY ||
1162 score->addr_type & IPV6_ADDR_MULTICAST)) {
1163 LIMIT_NETDEBUG(KERN_DEBUG
1164 "ADDRCONF: unspecified / multicast address "
1165 "assigned as unicast address on %s",
1166 dev->name);
1167 continue;
1170 score->rule = -1;
1171 bitmap_zero(score->scorebits, IPV6_SADDR_RULE_MAX);
1173 for (i = 0; i < IPV6_SADDR_RULE_MAX; i++) {
1174 int minihiscore, miniscore;
1176 minihiscore = ipv6_get_saddr_eval(net, hiscore, &dst, i);
1177 miniscore = ipv6_get_saddr_eval(net, score, &dst, i);
1179 if (minihiscore > miniscore) {
1180 if (i == IPV6_SADDR_RULE_SCOPE &&
1181 score->scopedist > 0) {
1183 * special case:
1184 * each remaining entry
1185 * has too small (not enough)
1186 * scope, because ifa entries
1187 * are sorted by their scope
1188 * values.
1190 goto try_nextdev;
1192 break;
1193 } else if (minihiscore < miniscore) {
1194 if (hiscore->ifa)
1195 in6_ifa_put(hiscore->ifa);
1197 in6_ifa_hold(score->ifa);
1199 swap(hiscore, score);
1201 /* restore our iterator */
1202 score->ifa = hiscore->ifa;
1204 break;
1208 try_nextdev:
1209 read_unlock_bh(&idev->lock);
1211 rcu_read_unlock();
1213 if (!hiscore->ifa)
1214 return -EADDRNOTAVAIL;
1216 *saddr = hiscore->ifa->addr;
1217 in6_ifa_put(hiscore->ifa);
1218 return 0;
1220 EXPORT_SYMBOL(ipv6_dev_get_saddr);
1222 int ipv6_get_lladdr(struct net_device *dev, struct in6_addr *addr,
1223 unsigned char banned_flags)
1225 struct inet6_dev *idev;
1226 int err = -EADDRNOTAVAIL;
1228 rcu_read_lock();
1229 idev = __in6_dev_get(dev);
1230 if (idev) {
1231 struct inet6_ifaddr *ifp;
1233 read_lock_bh(&idev->lock);
1234 list_for_each_entry(ifp, &idev->addr_list, if_list) {
1235 if (ifp->scope == IFA_LINK &&
1236 !(ifp->flags & banned_flags)) {
1237 *addr = ifp->addr;
1238 err = 0;
1239 break;
1242 read_unlock_bh(&idev->lock);
1244 rcu_read_unlock();
1245 return err;
1248 static int ipv6_count_addresses(struct inet6_dev *idev)
1250 int cnt = 0;
1251 struct inet6_ifaddr *ifp;
1253 read_lock_bh(&idev->lock);
1254 list_for_each_entry(ifp, &idev->addr_list, if_list)
1255 cnt++;
1256 read_unlock_bh(&idev->lock);
1257 return cnt;
1260 int ipv6_chk_addr(struct net *net, const struct in6_addr *addr,
1261 struct net_device *dev, int strict)
1263 struct inet6_ifaddr *ifp;
1264 struct hlist_node *node;
1265 unsigned int hash = inet6_addr_hash(addr);
1267 rcu_read_lock_bh();
1268 hlist_for_each_entry_rcu(ifp, node, &inet6_addr_lst[hash], addr_lst) {
1269 if (!net_eq(dev_net(ifp->idev->dev), net))
1270 continue;
1271 if (ipv6_addr_equal(&ifp->addr, addr) &&
1272 !(ifp->flags&IFA_F_TENTATIVE) &&
1273 (dev == NULL || ifp->idev->dev == dev ||
1274 !(ifp->scope&(IFA_LINK|IFA_HOST) || strict))) {
1275 rcu_read_unlock_bh();
1276 return 1;
1280 rcu_read_unlock_bh();
1281 return 0;
1283 EXPORT_SYMBOL(ipv6_chk_addr);
1285 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
1286 struct net_device *dev)
1288 unsigned int hash = inet6_addr_hash(addr);
1289 struct inet6_ifaddr *ifp;
1290 struct hlist_node *node;
1292 hlist_for_each_entry(ifp, node, &inet6_addr_lst[hash], addr_lst) {
1293 if (!net_eq(dev_net(ifp->idev->dev), net))
1294 continue;
1295 if (ipv6_addr_equal(&ifp->addr, addr)) {
1296 if (dev == NULL || ifp->idev->dev == dev)
1297 return true;
1300 return false;
1303 int ipv6_chk_prefix(const struct in6_addr *addr, struct net_device *dev)
1305 struct inet6_dev *idev;
1306 struct inet6_ifaddr *ifa;
1307 int onlink;
1309 onlink = 0;
1310 rcu_read_lock();
1311 idev = __in6_dev_get(dev);
1312 if (idev) {
1313 read_lock_bh(&idev->lock);
1314 list_for_each_entry(ifa, &idev->addr_list, if_list) {
1315 onlink = ipv6_prefix_equal(addr, &ifa->addr,
1316 ifa->prefix_len);
1317 if (onlink)
1318 break;
1320 read_unlock_bh(&idev->lock);
1322 rcu_read_unlock();
1323 return onlink;
1325 EXPORT_SYMBOL(ipv6_chk_prefix);
1327 struct inet6_ifaddr *ipv6_get_ifaddr(struct net *net, const struct in6_addr *addr,
1328 struct net_device *dev, int strict)
1330 struct inet6_ifaddr *ifp, *result = NULL;
1331 unsigned int hash = inet6_addr_hash(addr);
1332 struct hlist_node *node;
1334 rcu_read_lock_bh();
1335 hlist_for_each_entry_rcu_bh(ifp, node, &inet6_addr_lst[hash], addr_lst) {
1336 if (!net_eq(dev_net(ifp->idev->dev), net))
1337 continue;
1338 if (ipv6_addr_equal(&ifp->addr, addr)) {
1339 if (dev == NULL || ifp->idev->dev == dev ||
1340 !(ifp->scope&(IFA_LINK|IFA_HOST) || strict)) {
1341 result = ifp;
1342 in6_ifa_hold(ifp);
1343 break;
1347 rcu_read_unlock_bh();
1349 return result;
1352 /* Gets referenced address, destroys ifaddr */
1354 static void addrconf_dad_stop(struct inet6_ifaddr *ifp, int dad_failed)
1356 if (ifp->flags&IFA_F_PERMANENT) {
1357 spin_lock_bh(&ifp->lock);
1358 addrconf_del_timer(ifp);
1359 ifp->flags |= IFA_F_TENTATIVE;
1360 if (dad_failed)
1361 ifp->flags |= IFA_F_DADFAILED;
1362 spin_unlock_bh(&ifp->lock);
1363 if (dad_failed)
1364 ipv6_ifa_notify(0, ifp);
1365 in6_ifa_put(ifp);
1366 #ifdef CONFIG_IPV6_PRIVACY
1367 } else if (ifp->flags&IFA_F_TEMPORARY) {
1368 struct inet6_ifaddr *ifpub;
1369 spin_lock_bh(&ifp->lock);
1370 ifpub = ifp->ifpub;
1371 if (ifpub) {
1372 in6_ifa_hold(ifpub);
1373 spin_unlock_bh(&ifp->lock);
1374 ipv6_create_tempaddr(ifpub, ifp);
1375 in6_ifa_put(ifpub);
1376 } else {
1377 spin_unlock_bh(&ifp->lock);
1379 ipv6_del_addr(ifp);
1380 #endif
1381 } else
1382 ipv6_del_addr(ifp);
1385 static int addrconf_dad_end(struct inet6_ifaddr *ifp)
1387 int err = -ENOENT;
1389 spin_lock(&ifp->state_lock);
1390 if (ifp->state == INET6_IFADDR_STATE_DAD) {
1391 ifp->state = INET6_IFADDR_STATE_POSTDAD;
1392 err = 0;
1394 spin_unlock(&ifp->state_lock);
1396 return err;
1399 void addrconf_dad_failure(struct inet6_ifaddr *ifp)
1401 struct inet6_dev *idev = ifp->idev;
1403 if (addrconf_dad_end(ifp)) {
1404 in6_ifa_put(ifp);
1405 return;
1408 net_info_ratelimited("%s: IPv6 duplicate address %pI6c detected!\n",
1409 ifp->idev->dev->name, &ifp->addr);
1411 if (idev->cnf.accept_dad > 1 && !idev->cnf.disable_ipv6) {
1412 struct in6_addr addr;
1414 addr.s6_addr32[0] = htonl(0xfe800000);
1415 addr.s6_addr32[1] = 0;
1417 if (!ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) &&
1418 ipv6_addr_equal(&ifp->addr, &addr)) {
1419 /* DAD failed for link-local based on MAC address */
1420 idev->cnf.disable_ipv6 = 1;
1422 pr_info("%s: IPv6 being disabled!\n",
1423 ifp->idev->dev->name);
1427 addrconf_dad_stop(ifp, 1);
1430 /* Join to solicited addr multicast group. */
1432 void addrconf_join_solict(struct net_device *dev, const struct in6_addr *addr)
1434 struct in6_addr maddr;
1436 if (dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1437 return;
1439 addrconf_addr_solict_mult(addr, &maddr);
1440 ipv6_dev_mc_inc(dev, &maddr);
1443 void addrconf_leave_solict(struct inet6_dev *idev, const struct in6_addr *addr)
1445 struct in6_addr maddr;
1447 if (idev->dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1448 return;
1450 addrconf_addr_solict_mult(addr, &maddr);
1451 __ipv6_dev_mc_dec(idev, &maddr);
1454 static void addrconf_join_anycast(struct inet6_ifaddr *ifp)
1456 struct in6_addr addr;
1457 if (ifp->prefix_len == 127) /* RFC 6164 */
1458 return;
1459 ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1460 if (ipv6_addr_any(&addr))
1461 return;
1462 ipv6_dev_ac_inc(ifp->idev->dev, &addr);
1465 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp)
1467 struct in6_addr addr;
1468 if (ifp->prefix_len == 127) /* RFC 6164 */
1469 return;
1470 ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1471 if (ipv6_addr_any(&addr))
1472 return;
1473 __ipv6_dev_ac_dec(ifp->idev, &addr);
1476 static int addrconf_ifid_eui48(u8 *eui, struct net_device *dev)
1478 if (dev->addr_len != ETH_ALEN)
1479 return -1;
1480 memcpy(eui, dev->dev_addr, 3);
1481 memcpy(eui + 5, dev->dev_addr + 3, 3);
1484 * The zSeries OSA network cards can be shared among various
1485 * OS instances, but the OSA cards have only one MAC address.
1486 * This leads to duplicate address conflicts in conjunction
1487 * with IPv6 if more than one instance uses the same card.
1489 * The driver for these cards can deliver a unique 16-bit
1490 * identifier for each instance sharing the same card. It is
1491 * placed instead of 0xFFFE in the interface identifier. The
1492 * "u" bit of the interface identifier is not inverted in this
1493 * case. Hence the resulting interface identifier has local
1494 * scope according to RFC2373.
1496 if (dev->dev_id) {
1497 eui[3] = (dev->dev_id >> 8) & 0xFF;
1498 eui[4] = dev->dev_id & 0xFF;
1499 } else {
1500 eui[3] = 0xFF;
1501 eui[4] = 0xFE;
1502 eui[0] ^= 2;
1504 return 0;
1507 static int addrconf_ifid_eui64(u8 *eui, struct net_device *dev)
1509 if (dev->addr_len != IEEE802154_ADDR_LEN)
1510 return -1;
1511 memcpy(eui, dev->dev_addr, 8);
1512 return 0;
1515 static int addrconf_ifid_arcnet(u8 *eui, struct net_device *dev)
1517 /* XXX: inherit EUI-64 from other interface -- yoshfuji */
1518 if (dev->addr_len != ARCNET_ALEN)
1519 return -1;
1520 memset(eui, 0, 7);
1521 eui[7] = *(u8 *)dev->dev_addr;
1522 return 0;
1525 static int addrconf_ifid_infiniband(u8 *eui, struct net_device *dev)
1527 if (dev->addr_len != INFINIBAND_ALEN)
1528 return -1;
1529 memcpy(eui, dev->dev_addr + 12, 8);
1530 eui[0] |= 2;
1531 return 0;
1534 static int __ipv6_isatap_ifid(u8 *eui, __be32 addr)
1536 if (addr == 0)
1537 return -1;
1538 eui[0] = (ipv4_is_zeronet(addr) || ipv4_is_private_10(addr) ||
1539 ipv4_is_loopback(addr) || ipv4_is_linklocal_169(addr) ||
1540 ipv4_is_private_172(addr) || ipv4_is_test_192(addr) ||
1541 ipv4_is_anycast_6to4(addr) || ipv4_is_private_192(addr) ||
1542 ipv4_is_test_198(addr) || ipv4_is_multicast(addr) ||
1543 ipv4_is_lbcast(addr)) ? 0x00 : 0x02;
1544 eui[1] = 0;
1545 eui[2] = 0x5E;
1546 eui[3] = 0xFE;
1547 memcpy(eui + 4, &addr, 4);
1548 return 0;
1551 static int addrconf_ifid_sit(u8 *eui, struct net_device *dev)
1553 if (dev->priv_flags & IFF_ISATAP)
1554 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
1555 return -1;
1558 static int addrconf_ifid_gre(u8 *eui, struct net_device *dev)
1560 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
1563 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev)
1565 switch (dev->type) {
1566 case ARPHRD_ETHER:
1567 case ARPHRD_FDDI:
1568 return addrconf_ifid_eui48(eui, dev);
1569 case ARPHRD_ARCNET:
1570 return addrconf_ifid_arcnet(eui, dev);
1571 case ARPHRD_INFINIBAND:
1572 return addrconf_ifid_infiniband(eui, dev);
1573 case ARPHRD_SIT:
1574 return addrconf_ifid_sit(eui, dev);
1575 case ARPHRD_IPGRE:
1576 return addrconf_ifid_gre(eui, dev);
1577 case ARPHRD_IEEE802154:
1578 return addrconf_ifid_eui64(eui, dev);
1580 return -1;
1583 static int ipv6_inherit_eui64(u8 *eui, struct inet6_dev *idev)
1585 int err = -1;
1586 struct inet6_ifaddr *ifp;
1588 read_lock_bh(&idev->lock);
1589 list_for_each_entry(ifp, &idev->addr_list, if_list) {
1590 if (ifp->scope == IFA_LINK && !(ifp->flags&IFA_F_TENTATIVE)) {
1591 memcpy(eui, ifp->addr.s6_addr+8, 8);
1592 err = 0;
1593 break;
1596 read_unlock_bh(&idev->lock);
1597 return err;
1600 #ifdef CONFIG_IPV6_PRIVACY
1601 /* (re)generation of randomized interface identifier (RFC 3041 3.2, 3.5) */
1602 static void __ipv6_regen_rndid(struct inet6_dev *idev)
1604 regen:
1605 get_random_bytes(idev->rndid, sizeof(idev->rndid));
1606 idev->rndid[0] &= ~0x02;
1609 * <draft-ietf-ipngwg-temp-addresses-v2-00.txt>:
1610 * check if generated address is not inappropriate
1612 * - Reserved subnet anycast (RFC 2526)
1613 * 11111101 11....11 1xxxxxxx
1614 * - ISATAP (RFC4214) 6.1
1615 * 00-00-5E-FE-xx-xx-xx-xx
1616 * - value 0
1617 * - XXX: already assigned to an address on the device
1619 if (idev->rndid[0] == 0xfd &&
1620 (idev->rndid[1]&idev->rndid[2]&idev->rndid[3]&idev->rndid[4]&idev->rndid[5]&idev->rndid[6]) == 0xff &&
1621 (idev->rndid[7]&0x80))
1622 goto regen;
1623 if ((idev->rndid[0]|idev->rndid[1]) == 0) {
1624 if (idev->rndid[2] == 0x5e && idev->rndid[3] == 0xfe)
1625 goto regen;
1626 if ((idev->rndid[2]|idev->rndid[3]|idev->rndid[4]|idev->rndid[5]|idev->rndid[6]|idev->rndid[7]) == 0x00)
1627 goto regen;
1631 static void ipv6_regen_rndid(unsigned long data)
1633 struct inet6_dev *idev = (struct inet6_dev *) data;
1634 unsigned long expires;
1636 rcu_read_lock_bh();
1637 write_lock_bh(&idev->lock);
1639 if (idev->dead)
1640 goto out;
1642 __ipv6_regen_rndid(idev);
1644 expires = jiffies +
1645 idev->cnf.temp_prefered_lft * HZ -
1646 idev->cnf.regen_max_retry * idev->cnf.dad_transmits * idev->nd_parms->retrans_time -
1647 idev->cnf.max_desync_factor * HZ;
1648 if (time_before(expires, jiffies)) {
1649 pr_warn("%s: too short regeneration interval; timer disabled for %s\n",
1650 __func__, idev->dev->name);
1651 goto out;
1654 if (!mod_timer(&idev->regen_timer, expires))
1655 in6_dev_hold(idev);
1657 out:
1658 write_unlock_bh(&idev->lock);
1659 rcu_read_unlock_bh();
1660 in6_dev_put(idev);
1663 static void __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr)
1665 if (tmpaddr && memcmp(idev->rndid, &tmpaddr->s6_addr[8], 8) == 0)
1666 __ipv6_regen_rndid(idev);
1668 #endif
1671 * Add prefix route.
1674 static void
1675 addrconf_prefix_route(struct in6_addr *pfx, int plen, struct net_device *dev,
1676 unsigned long expires, u32 flags)
1678 struct fib6_config cfg = {
1679 .fc_table = RT6_TABLE_PREFIX,
1680 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1681 .fc_ifindex = dev->ifindex,
1682 .fc_expires = expires,
1683 .fc_dst_len = plen,
1684 .fc_flags = RTF_UP | flags,
1685 .fc_nlinfo.nl_net = dev_net(dev),
1686 .fc_protocol = RTPROT_KERNEL,
1689 cfg.fc_dst = *pfx;
1691 /* Prevent useless cloning on PtP SIT.
1692 This thing is done here expecting that the whole
1693 class of non-broadcast devices need not cloning.
1695 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
1696 if (dev->type == ARPHRD_SIT && (dev->flags & IFF_POINTOPOINT))
1697 cfg.fc_flags |= RTF_NONEXTHOP;
1698 #endif
1700 ip6_route_add(&cfg);
1704 static struct rt6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
1705 int plen,
1706 const struct net_device *dev,
1707 u32 flags, u32 noflags)
1709 struct fib6_node *fn;
1710 struct rt6_info *rt = NULL;
1711 struct fib6_table *table;
1713 table = fib6_get_table(dev_net(dev), RT6_TABLE_PREFIX);
1714 if (table == NULL)
1715 return NULL;
1717 read_lock_bh(&table->tb6_lock);
1718 fn = fib6_locate(&table->tb6_root, pfx, plen, NULL, 0);
1719 if (!fn)
1720 goto out;
1721 for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
1722 if (rt->dst.dev->ifindex != dev->ifindex)
1723 continue;
1724 if ((rt->rt6i_flags & flags) != flags)
1725 continue;
1726 if ((noflags != 0) && ((rt->rt6i_flags & flags) != 0))
1727 continue;
1728 dst_hold(&rt->dst);
1729 break;
1731 out:
1732 read_unlock_bh(&table->tb6_lock);
1733 return rt;
1737 /* Create "default" multicast route to the interface */
1739 static void addrconf_add_mroute(struct net_device *dev)
1741 struct fib6_config cfg = {
1742 .fc_table = RT6_TABLE_LOCAL,
1743 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1744 .fc_ifindex = dev->ifindex,
1745 .fc_dst_len = 8,
1746 .fc_flags = RTF_UP,
1747 .fc_nlinfo.nl_net = dev_net(dev),
1750 ipv6_addr_set(&cfg.fc_dst, htonl(0xFF000000), 0, 0, 0);
1752 ip6_route_add(&cfg);
1755 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
1756 static void sit_route_add(struct net_device *dev)
1758 struct fib6_config cfg = {
1759 .fc_table = RT6_TABLE_MAIN,
1760 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1761 .fc_ifindex = dev->ifindex,
1762 .fc_dst_len = 96,
1763 .fc_flags = RTF_UP | RTF_NONEXTHOP,
1764 .fc_nlinfo.nl_net = dev_net(dev),
1767 /* prefix length - 96 bits "::d.d.d.d" */
1768 ip6_route_add(&cfg);
1770 #endif
1772 static struct inet6_dev *addrconf_add_dev(struct net_device *dev)
1774 struct inet6_dev *idev;
1776 ASSERT_RTNL();
1778 idev = ipv6_find_idev(dev);
1779 if (!idev)
1780 return ERR_PTR(-ENOBUFS);
1782 if (idev->cnf.disable_ipv6)
1783 return ERR_PTR(-EACCES);
1785 /* Add default multicast route */
1786 if (!(dev->flags & IFF_LOOPBACK))
1787 addrconf_add_mroute(dev);
1789 return idev;
1792 void addrconf_prefix_rcv(struct net_device *dev, u8 *opt, int len, bool sllao)
1794 struct prefix_info *pinfo;
1795 __u32 valid_lft;
1796 __u32 prefered_lft;
1797 int addr_type;
1798 struct inet6_dev *in6_dev;
1799 struct net *net = dev_net(dev);
1801 pinfo = (struct prefix_info *) opt;
1803 if (len < sizeof(struct prefix_info)) {
1804 ADBG(("addrconf: prefix option too short\n"));
1805 return;
1809 * Validation checks ([ADDRCONF], page 19)
1812 addr_type = ipv6_addr_type(&pinfo->prefix);
1814 if (addr_type & (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL))
1815 return;
1817 valid_lft = ntohl(pinfo->valid);
1818 prefered_lft = ntohl(pinfo->prefered);
1820 if (prefered_lft > valid_lft) {
1821 net_warn_ratelimited("addrconf: prefix option has invalid lifetime\n");
1822 return;
1825 in6_dev = in6_dev_get(dev);
1827 if (in6_dev == NULL) {
1828 net_dbg_ratelimited("addrconf: device %s not configured\n",
1829 dev->name);
1830 return;
1834 * Two things going on here:
1835 * 1) Add routes for on-link prefixes
1836 * 2) Configure prefixes with the auto flag set
1839 if (pinfo->onlink) {
1840 struct rt6_info *rt;
1841 unsigned long rt_expires;
1843 /* Avoid arithmetic overflow. Really, we could
1844 * save rt_expires in seconds, likely valid_lft,
1845 * but it would require division in fib gc, that it
1846 * not good.
1848 if (HZ > USER_HZ)
1849 rt_expires = addrconf_timeout_fixup(valid_lft, HZ);
1850 else
1851 rt_expires = addrconf_timeout_fixup(valid_lft, USER_HZ);
1853 if (addrconf_finite_timeout(rt_expires))
1854 rt_expires *= HZ;
1856 rt = addrconf_get_prefix_route(&pinfo->prefix,
1857 pinfo->prefix_len,
1858 dev,
1859 RTF_ADDRCONF | RTF_PREFIX_RT,
1860 RTF_GATEWAY | RTF_DEFAULT);
1862 if (rt) {
1863 /* Autoconf prefix route */
1864 if (valid_lft == 0) {
1865 ip6_del_rt(rt);
1866 rt = NULL;
1867 } else if (addrconf_finite_timeout(rt_expires)) {
1868 /* not infinity */
1869 rt6_set_expires(rt, jiffies + rt_expires);
1870 } else {
1871 rt6_clean_expires(rt);
1873 } else if (valid_lft) {
1874 clock_t expires = 0;
1875 int flags = RTF_ADDRCONF | RTF_PREFIX_RT;
1876 if (addrconf_finite_timeout(rt_expires)) {
1877 /* not infinity */
1878 flags |= RTF_EXPIRES;
1879 expires = jiffies_to_clock_t(rt_expires);
1881 addrconf_prefix_route(&pinfo->prefix, pinfo->prefix_len,
1882 dev, expires, flags);
1884 if (rt)
1885 dst_release(&rt->dst);
1888 /* Try to figure out our local address for this prefix */
1890 if (pinfo->autoconf && in6_dev->cnf.autoconf) {
1891 struct inet6_ifaddr *ifp;
1892 struct in6_addr addr;
1893 int create = 0, update_lft = 0;
1895 if (pinfo->prefix_len == 64) {
1896 memcpy(&addr, &pinfo->prefix, 8);
1897 if (ipv6_generate_eui64(addr.s6_addr + 8, dev) &&
1898 ipv6_inherit_eui64(addr.s6_addr + 8, in6_dev)) {
1899 in6_dev_put(in6_dev);
1900 return;
1902 goto ok;
1904 net_dbg_ratelimited("IPv6 addrconf: prefix with wrong length %d\n",
1905 pinfo->prefix_len);
1906 in6_dev_put(in6_dev);
1907 return;
1911 ifp = ipv6_get_ifaddr(net, &addr, dev, 1);
1913 if (ifp == NULL && valid_lft) {
1914 int max_addresses = in6_dev->cnf.max_addresses;
1915 u32 addr_flags = 0;
1917 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1918 if (in6_dev->cnf.optimistic_dad &&
1919 !net->ipv6.devconf_all->forwarding && sllao)
1920 addr_flags = IFA_F_OPTIMISTIC;
1921 #endif
1923 /* Do not allow to create too much of autoconfigured
1924 * addresses; this would be too easy way to crash kernel.
1926 if (!max_addresses ||
1927 ipv6_count_addresses(in6_dev) < max_addresses)
1928 ifp = ipv6_add_addr(in6_dev, &addr, pinfo->prefix_len,
1929 addr_type&IPV6_ADDR_SCOPE_MASK,
1930 addr_flags);
1932 if (!ifp || IS_ERR(ifp)) {
1933 in6_dev_put(in6_dev);
1934 return;
1937 update_lft = create = 1;
1938 ifp->cstamp = jiffies;
1939 addrconf_dad_start(ifp);
1942 if (ifp) {
1943 int flags;
1944 unsigned long now;
1945 #ifdef CONFIG_IPV6_PRIVACY
1946 struct inet6_ifaddr *ift;
1947 #endif
1948 u32 stored_lft;
1950 /* update lifetime (RFC2462 5.5.3 e) */
1951 spin_lock(&ifp->lock);
1952 now = jiffies;
1953 if (ifp->valid_lft > (now - ifp->tstamp) / HZ)
1954 stored_lft = ifp->valid_lft - (now - ifp->tstamp) / HZ;
1955 else
1956 stored_lft = 0;
1957 if (!update_lft && stored_lft) {
1958 if (valid_lft > MIN_VALID_LIFETIME ||
1959 valid_lft > stored_lft)
1960 update_lft = 1;
1961 else if (stored_lft <= MIN_VALID_LIFETIME) {
1962 /* valid_lft <= stored_lft is always true */
1964 * RFC 4862 Section 5.5.3e:
1965 * "Note that the preferred lifetime of
1966 * the corresponding address is always
1967 * reset to the Preferred Lifetime in
1968 * the received Prefix Information
1969 * option, regardless of whether the
1970 * valid lifetime is also reset or
1971 * ignored."
1973 * So if the preferred lifetime in
1974 * this advertisement is different
1975 * than what we have stored, but the
1976 * valid lifetime is invalid, just
1977 * reset prefered_lft.
1979 * We must set the valid lifetime
1980 * to the stored lifetime since we'll
1981 * be updating the timestamp below,
1982 * else we'll set it back to the
1983 * minimum.
1985 if (prefered_lft != ifp->prefered_lft) {
1986 valid_lft = stored_lft;
1987 update_lft = 1;
1989 } else {
1990 valid_lft = MIN_VALID_LIFETIME;
1991 if (valid_lft < prefered_lft)
1992 prefered_lft = valid_lft;
1993 update_lft = 1;
1997 if (update_lft) {
1998 ifp->valid_lft = valid_lft;
1999 ifp->prefered_lft = prefered_lft;
2000 ifp->tstamp = now;
2001 flags = ifp->flags;
2002 ifp->flags &= ~IFA_F_DEPRECATED;
2003 spin_unlock(&ifp->lock);
2005 if (!(flags&IFA_F_TENTATIVE))
2006 ipv6_ifa_notify(0, ifp);
2007 } else
2008 spin_unlock(&ifp->lock);
2010 #ifdef CONFIG_IPV6_PRIVACY
2011 read_lock_bh(&in6_dev->lock);
2012 /* update all temporary addresses in the list */
2013 list_for_each_entry(ift, &in6_dev->tempaddr_list,
2014 tmp_list) {
2015 int age, max_valid, max_prefered;
2017 if (ifp != ift->ifpub)
2018 continue;
2021 * RFC 4941 section 3.3:
2022 * If a received option will extend the lifetime
2023 * of a public address, the lifetimes of
2024 * temporary addresses should be extended,
2025 * subject to the overall constraint that no
2026 * temporary addresses should ever remain
2027 * "valid" or "preferred" for a time longer than
2028 * (TEMP_VALID_LIFETIME) or
2029 * (TEMP_PREFERRED_LIFETIME - DESYNC_FACTOR),
2030 * respectively.
2032 age = (now - ift->cstamp) / HZ;
2033 max_valid = in6_dev->cnf.temp_valid_lft - age;
2034 if (max_valid < 0)
2035 max_valid = 0;
2037 max_prefered = in6_dev->cnf.temp_prefered_lft -
2038 in6_dev->cnf.max_desync_factor -
2039 age;
2040 if (max_prefered < 0)
2041 max_prefered = 0;
2043 if (valid_lft > max_valid)
2044 valid_lft = max_valid;
2046 if (prefered_lft > max_prefered)
2047 prefered_lft = max_prefered;
2049 spin_lock(&ift->lock);
2050 flags = ift->flags;
2051 ift->valid_lft = valid_lft;
2052 ift->prefered_lft = prefered_lft;
2053 ift->tstamp = now;
2054 if (prefered_lft > 0)
2055 ift->flags &= ~IFA_F_DEPRECATED;
2057 spin_unlock(&ift->lock);
2058 if (!(flags&IFA_F_TENTATIVE))
2059 ipv6_ifa_notify(0, ift);
2062 if ((create || list_empty(&in6_dev->tempaddr_list)) && in6_dev->cnf.use_tempaddr > 0) {
2064 * When a new public address is created as
2065 * described in [ADDRCONF], also create a new
2066 * temporary address. Also create a temporary
2067 * address if it's enabled but no temporary
2068 * address currently exists.
2070 read_unlock_bh(&in6_dev->lock);
2071 ipv6_create_tempaddr(ifp, NULL);
2072 } else {
2073 read_unlock_bh(&in6_dev->lock);
2075 #endif
2076 in6_ifa_put(ifp);
2077 addrconf_verify(0);
2080 inet6_prefix_notify(RTM_NEWPREFIX, in6_dev, pinfo);
2081 in6_dev_put(in6_dev);
2085 * Set destination address.
2086 * Special case for SIT interfaces where we create a new "virtual"
2087 * device.
2089 int addrconf_set_dstaddr(struct net *net, void __user *arg)
2091 struct in6_ifreq ireq;
2092 struct net_device *dev;
2093 int err = -EINVAL;
2095 rtnl_lock();
2097 err = -EFAULT;
2098 if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2099 goto err_exit;
2101 dev = __dev_get_by_index(net, ireq.ifr6_ifindex);
2103 err = -ENODEV;
2104 if (dev == NULL)
2105 goto err_exit;
2107 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2108 if (dev->type == ARPHRD_SIT) {
2109 const struct net_device_ops *ops = dev->netdev_ops;
2110 struct ifreq ifr;
2111 struct ip_tunnel_parm p;
2113 err = -EADDRNOTAVAIL;
2114 if (!(ipv6_addr_type(&ireq.ifr6_addr) & IPV6_ADDR_COMPATv4))
2115 goto err_exit;
2117 memset(&p, 0, sizeof(p));
2118 p.iph.daddr = ireq.ifr6_addr.s6_addr32[3];
2119 p.iph.saddr = 0;
2120 p.iph.version = 4;
2121 p.iph.ihl = 5;
2122 p.iph.protocol = IPPROTO_IPV6;
2123 p.iph.ttl = 64;
2124 ifr.ifr_ifru.ifru_data = (__force void __user *)&p;
2126 if (ops->ndo_do_ioctl) {
2127 mm_segment_t oldfs = get_fs();
2129 set_fs(KERNEL_DS);
2130 err = ops->ndo_do_ioctl(dev, &ifr, SIOCADDTUNNEL);
2131 set_fs(oldfs);
2132 } else
2133 err = -EOPNOTSUPP;
2135 if (err == 0) {
2136 err = -ENOBUFS;
2137 dev = __dev_get_by_name(net, p.name);
2138 if (!dev)
2139 goto err_exit;
2140 err = dev_open(dev);
2143 #endif
2145 err_exit:
2146 rtnl_unlock();
2147 return err;
2151 * Manual configuration of address on an interface
2153 static int inet6_addr_add(struct net *net, int ifindex, const struct in6_addr *pfx,
2154 unsigned int plen, __u8 ifa_flags, __u32 prefered_lft,
2155 __u32 valid_lft)
2157 struct inet6_ifaddr *ifp;
2158 struct inet6_dev *idev;
2159 struct net_device *dev;
2160 int scope;
2161 u32 flags;
2162 clock_t expires;
2163 unsigned long timeout;
2165 ASSERT_RTNL();
2167 if (plen > 128)
2168 return -EINVAL;
2170 /* check the lifetime */
2171 if (!valid_lft || prefered_lft > valid_lft)
2172 return -EINVAL;
2174 dev = __dev_get_by_index(net, ifindex);
2175 if (!dev)
2176 return -ENODEV;
2178 idev = addrconf_add_dev(dev);
2179 if (IS_ERR(idev))
2180 return PTR_ERR(idev);
2182 scope = ipv6_addr_scope(pfx);
2184 timeout = addrconf_timeout_fixup(valid_lft, HZ);
2185 if (addrconf_finite_timeout(timeout)) {
2186 expires = jiffies_to_clock_t(timeout * HZ);
2187 valid_lft = timeout;
2188 flags = RTF_EXPIRES;
2189 } else {
2190 expires = 0;
2191 flags = 0;
2192 ifa_flags |= IFA_F_PERMANENT;
2195 timeout = addrconf_timeout_fixup(prefered_lft, HZ);
2196 if (addrconf_finite_timeout(timeout)) {
2197 if (timeout == 0)
2198 ifa_flags |= IFA_F_DEPRECATED;
2199 prefered_lft = timeout;
2202 ifp = ipv6_add_addr(idev, pfx, plen, scope, ifa_flags);
2204 if (!IS_ERR(ifp)) {
2205 spin_lock_bh(&ifp->lock);
2206 ifp->valid_lft = valid_lft;
2207 ifp->prefered_lft = prefered_lft;
2208 ifp->tstamp = jiffies;
2209 spin_unlock_bh(&ifp->lock);
2211 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, dev,
2212 expires, flags);
2214 * Note that section 3.1 of RFC 4429 indicates
2215 * that the Optimistic flag should not be set for
2216 * manually configured addresses
2218 addrconf_dad_start(ifp);
2219 in6_ifa_put(ifp);
2220 addrconf_verify(0);
2221 return 0;
2224 return PTR_ERR(ifp);
2227 static int inet6_addr_del(struct net *net, int ifindex, const struct in6_addr *pfx,
2228 unsigned int plen)
2230 struct inet6_ifaddr *ifp;
2231 struct inet6_dev *idev;
2232 struct net_device *dev;
2234 if (plen > 128)
2235 return -EINVAL;
2237 dev = __dev_get_by_index(net, ifindex);
2238 if (!dev)
2239 return -ENODEV;
2241 if ((idev = __in6_dev_get(dev)) == NULL)
2242 return -ENXIO;
2244 read_lock_bh(&idev->lock);
2245 list_for_each_entry(ifp, &idev->addr_list, if_list) {
2246 if (ifp->prefix_len == plen &&
2247 ipv6_addr_equal(pfx, &ifp->addr)) {
2248 in6_ifa_hold(ifp);
2249 read_unlock_bh(&idev->lock);
2251 ipv6_del_addr(ifp);
2253 /* If the last address is deleted administratively,
2254 disable IPv6 on this interface.
2256 if (list_empty(&idev->addr_list))
2257 addrconf_ifdown(idev->dev, 1);
2258 return 0;
2261 read_unlock_bh(&idev->lock);
2262 return -EADDRNOTAVAIL;
2266 int addrconf_add_ifaddr(struct net *net, void __user *arg)
2268 struct in6_ifreq ireq;
2269 int err;
2271 if (!capable(CAP_NET_ADMIN))
2272 return -EPERM;
2274 if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2275 return -EFAULT;
2277 rtnl_lock();
2278 err = inet6_addr_add(net, ireq.ifr6_ifindex, &ireq.ifr6_addr,
2279 ireq.ifr6_prefixlen, IFA_F_PERMANENT,
2280 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
2281 rtnl_unlock();
2282 return err;
2285 int addrconf_del_ifaddr(struct net *net, void __user *arg)
2287 struct in6_ifreq ireq;
2288 int err;
2290 if (!capable(CAP_NET_ADMIN))
2291 return -EPERM;
2293 if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2294 return -EFAULT;
2296 rtnl_lock();
2297 err = inet6_addr_del(net, ireq.ifr6_ifindex, &ireq.ifr6_addr,
2298 ireq.ifr6_prefixlen);
2299 rtnl_unlock();
2300 return err;
2303 static void add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
2304 int plen, int scope)
2306 struct inet6_ifaddr *ifp;
2308 ifp = ipv6_add_addr(idev, addr, plen, scope, IFA_F_PERMANENT);
2309 if (!IS_ERR(ifp)) {
2310 spin_lock_bh(&ifp->lock);
2311 ifp->flags &= ~IFA_F_TENTATIVE;
2312 spin_unlock_bh(&ifp->lock);
2313 ipv6_ifa_notify(RTM_NEWADDR, ifp);
2314 in6_ifa_put(ifp);
2318 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2319 static void sit_add_v4_addrs(struct inet6_dev *idev)
2321 struct in6_addr addr;
2322 struct net_device *dev;
2323 struct net *net = dev_net(idev->dev);
2324 int scope;
2326 ASSERT_RTNL();
2328 memset(&addr, 0, sizeof(struct in6_addr));
2329 memcpy(&addr.s6_addr32[3], idev->dev->dev_addr, 4);
2331 if (idev->dev->flags&IFF_POINTOPOINT) {
2332 addr.s6_addr32[0] = htonl(0xfe800000);
2333 scope = IFA_LINK;
2334 } else {
2335 scope = IPV6_ADDR_COMPATv4;
2338 if (addr.s6_addr32[3]) {
2339 add_addr(idev, &addr, 128, scope);
2340 return;
2343 for_each_netdev(net, dev) {
2344 struct in_device *in_dev = __in_dev_get_rtnl(dev);
2345 if (in_dev && (dev->flags & IFF_UP)) {
2346 struct in_ifaddr *ifa;
2348 int flag = scope;
2350 for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) {
2351 int plen;
2353 addr.s6_addr32[3] = ifa->ifa_local;
2355 if (ifa->ifa_scope == RT_SCOPE_LINK)
2356 continue;
2357 if (ifa->ifa_scope >= RT_SCOPE_HOST) {
2358 if (idev->dev->flags&IFF_POINTOPOINT)
2359 continue;
2360 flag |= IFA_HOST;
2362 if (idev->dev->flags&IFF_POINTOPOINT)
2363 plen = 64;
2364 else
2365 plen = 96;
2367 add_addr(idev, &addr, plen, flag);
2372 #endif
2374 static void init_loopback(struct net_device *dev)
2376 struct inet6_dev *idev;
2378 /* ::1 */
2380 ASSERT_RTNL();
2382 if ((idev = ipv6_find_idev(dev)) == NULL) {
2383 pr_debug("%s: add_dev failed\n", __func__);
2384 return;
2387 add_addr(idev, &in6addr_loopback, 128, IFA_HOST);
2390 static void addrconf_add_linklocal(struct inet6_dev *idev, const struct in6_addr *addr)
2392 struct inet6_ifaddr *ifp;
2393 u32 addr_flags = IFA_F_PERMANENT;
2395 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2396 if (idev->cnf.optimistic_dad &&
2397 !dev_net(idev->dev)->ipv6.devconf_all->forwarding)
2398 addr_flags |= IFA_F_OPTIMISTIC;
2399 #endif
2402 ifp = ipv6_add_addr(idev, addr, 64, IFA_LINK, addr_flags);
2403 if (!IS_ERR(ifp)) {
2404 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, idev->dev, 0, 0);
2405 addrconf_dad_start(ifp);
2406 in6_ifa_put(ifp);
2410 static void addrconf_dev_config(struct net_device *dev)
2412 struct in6_addr addr;
2413 struct inet6_dev *idev;
2415 ASSERT_RTNL();
2417 if ((dev->type != ARPHRD_ETHER) &&
2418 (dev->type != ARPHRD_FDDI) &&
2419 (dev->type != ARPHRD_ARCNET) &&
2420 (dev->type != ARPHRD_INFINIBAND) &&
2421 (dev->type != ARPHRD_IEEE802154)) {
2422 /* Alas, we support only Ethernet autoconfiguration. */
2423 return;
2426 idev = addrconf_add_dev(dev);
2427 if (IS_ERR(idev))
2428 return;
2430 memset(&addr, 0, sizeof(struct in6_addr));
2431 addr.s6_addr32[0] = htonl(0xFE800000);
2433 if (ipv6_generate_eui64(addr.s6_addr + 8, dev) == 0)
2434 addrconf_add_linklocal(idev, &addr);
2437 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2438 static void addrconf_sit_config(struct net_device *dev)
2440 struct inet6_dev *idev;
2442 ASSERT_RTNL();
2445 * Configure the tunnel with one of our IPv4
2446 * addresses... we should configure all of
2447 * our v4 addrs in the tunnel
2450 if ((idev = ipv6_find_idev(dev)) == NULL) {
2451 pr_debug("%s: add_dev failed\n", __func__);
2452 return;
2455 if (dev->priv_flags & IFF_ISATAP) {
2456 struct in6_addr addr;
2458 ipv6_addr_set(&addr, htonl(0xFE800000), 0, 0, 0);
2459 addrconf_prefix_route(&addr, 64, dev, 0, 0);
2460 if (!ipv6_generate_eui64(addr.s6_addr + 8, dev))
2461 addrconf_add_linklocal(idev, &addr);
2462 return;
2465 sit_add_v4_addrs(idev);
2467 if (dev->flags&IFF_POINTOPOINT)
2468 addrconf_add_mroute(dev);
2469 else
2470 sit_route_add(dev);
2472 #endif
2474 #if defined(CONFIG_NET_IPGRE) || defined(CONFIG_NET_IPGRE_MODULE)
2475 static void addrconf_gre_config(struct net_device *dev)
2477 struct inet6_dev *idev;
2478 struct in6_addr addr;
2480 pr_info("%s(%s)\n", __func__, dev->name);
2482 ASSERT_RTNL();
2484 if ((idev = ipv6_find_idev(dev)) == NULL) {
2485 pr_debug("%s: add_dev failed\n", __func__);
2486 return;
2489 ipv6_addr_set(&addr, htonl(0xFE800000), 0, 0, 0);
2490 addrconf_prefix_route(&addr, 64, dev, 0, 0);
2492 if (!ipv6_generate_eui64(addr.s6_addr + 8, dev))
2493 addrconf_add_linklocal(idev, &addr);
2495 #endif
2497 static inline int
2498 ipv6_inherit_linklocal(struct inet6_dev *idev, struct net_device *link_dev)
2500 struct in6_addr lladdr;
2502 if (!ipv6_get_lladdr(link_dev, &lladdr, IFA_F_TENTATIVE)) {
2503 addrconf_add_linklocal(idev, &lladdr);
2504 return 0;
2506 return -1;
2509 static void ip6_tnl_add_linklocal(struct inet6_dev *idev)
2511 struct net_device *link_dev;
2512 struct net *net = dev_net(idev->dev);
2514 /* first try to inherit the link-local address from the link device */
2515 if (idev->dev->iflink &&
2516 (link_dev = __dev_get_by_index(net, idev->dev->iflink))) {
2517 if (!ipv6_inherit_linklocal(idev, link_dev))
2518 return;
2520 /* then try to inherit it from any device */
2521 for_each_netdev(net, link_dev) {
2522 if (!ipv6_inherit_linklocal(idev, link_dev))
2523 return;
2525 pr_debug("init ip6-ip6: add_linklocal failed\n");
2529 * Autoconfigure tunnel with a link-local address so routing protocols,
2530 * DHCPv6, MLD etc. can be run over the virtual link
2533 static void addrconf_ip6_tnl_config(struct net_device *dev)
2535 struct inet6_dev *idev;
2537 ASSERT_RTNL();
2539 idev = addrconf_add_dev(dev);
2540 if (IS_ERR(idev)) {
2541 pr_debug("init ip6-ip6: add_dev failed\n");
2542 return;
2544 ip6_tnl_add_linklocal(idev);
2547 static int addrconf_notify(struct notifier_block *this, unsigned long event,
2548 void *data)
2550 struct net_device *dev = (struct net_device *) data;
2551 struct inet6_dev *idev = __in6_dev_get(dev);
2552 int run_pending = 0;
2553 int err;
2555 switch (event) {
2556 case NETDEV_REGISTER:
2557 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2558 idev = ipv6_add_dev(dev);
2559 if (!idev)
2560 return notifier_from_errno(-ENOMEM);
2562 break;
2564 case NETDEV_UP:
2565 case NETDEV_CHANGE:
2566 if (dev->flags & IFF_SLAVE)
2567 break;
2569 if (event == NETDEV_UP) {
2570 if (!addrconf_qdisc_ok(dev)) {
2571 /* device is not ready yet. */
2572 pr_info("ADDRCONF(NETDEV_UP): %s: link is not ready\n",
2573 dev->name);
2574 break;
2577 if (!idev && dev->mtu >= IPV6_MIN_MTU)
2578 idev = ipv6_add_dev(dev);
2580 if (idev) {
2581 idev->if_flags |= IF_READY;
2582 run_pending = 1;
2584 } else {
2585 if (!addrconf_qdisc_ok(dev)) {
2586 /* device is still not ready. */
2587 break;
2590 if (idev) {
2591 if (idev->if_flags & IF_READY)
2592 /* device is already configured. */
2593 break;
2594 idev->if_flags |= IF_READY;
2597 pr_info("ADDRCONF(NETDEV_CHANGE): %s: link becomes ready\n",
2598 dev->name);
2600 run_pending = 1;
2603 switch (dev->type) {
2604 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2605 case ARPHRD_SIT:
2606 addrconf_sit_config(dev);
2607 break;
2608 #endif
2609 #if defined(CONFIG_NET_IPGRE) || defined(CONFIG_NET_IPGRE_MODULE)
2610 case ARPHRD_IPGRE:
2611 addrconf_gre_config(dev);
2612 break;
2613 #endif
2614 case ARPHRD_TUNNEL6:
2615 addrconf_ip6_tnl_config(dev);
2616 break;
2617 case ARPHRD_LOOPBACK:
2618 init_loopback(dev);
2619 break;
2621 default:
2622 addrconf_dev_config(dev);
2623 break;
2626 if (idev) {
2627 if (run_pending)
2628 addrconf_dad_run(idev);
2631 * If the MTU changed during the interface down,
2632 * when the interface up, the changed MTU must be
2633 * reflected in the idev as well as routers.
2635 if (idev->cnf.mtu6 != dev->mtu &&
2636 dev->mtu >= IPV6_MIN_MTU) {
2637 rt6_mtu_change(dev, dev->mtu);
2638 idev->cnf.mtu6 = dev->mtu;
2640 idev->tstamp = jiffies;
2641 inet6_ifinfo_notify(RTM_NEWLINK, idev);
2644 * If the changed mtu during down is lower than
2645 * IPV6_MIN_MTU stop IPv6 on this interface.
2647 if (dev->mtu < IPV6_MIN_MTU)
2648 addrconf_ifdown(dev, 1);
2650 break;
2652 case NETDEV_CHANGEMTU:
2653 if (idev && dev->mtu >= IPV6_MIN_MTU) {
2654 rt6_mtu_change(dev, dev->mtu);
2655 idev->cnf.mtu6 = dev->mtu;
2656 break;
2659 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2660 idev = ipv6_add_dev(dev);
2661 if (idev)
2662 break;
2666 * MTU falled under IPV6_MIN_MTU.
2667 * Stop IPv6 on this interface.
2670 case NETDEV_DOWN:
2671 case NETDEV_UNREGISTER:
2673 * Remove all addresses from this interface.
2675 addrconf_ifdown(dev, event != NETDEV_DOWN);
2676 break;
2678 case NETDEV_CHANGENAME:
2679 if (idev) {
2680 snmp6_unregister_dev(idev);
2681 addrconf_sysctl_unregister(idev);
2682 addrconf_sysctl_register(idev);
2683 err = snmp6_register_dev(idev);
2684 if (err)
2685 return notifier_from_errno(err);
2687 break;
2689 case NETDEV_PRE_TYPE_CHANGE:
2690 case NETDEV_POST_TYPE_CHANGE:
2691 addrconf_type_change(dev, event);
2692 break;
2695 return NOTIFY_OK;
2699 * addrconf module should be notified of a device going up
2701 static struct notifier_block ipv6_dev_notf = {
2702 .notifier_call = addrconf_notify,
2705 static void addrconf_type_change(struct net_device *dev, unsigned long event)
2707 struct inet6_dev *idev;
2708 ASSERT_RTNL();
2710 idev = __in6_dev_get(dev);
2712 if (event == NETDEV_POST_TYPE_CHANGE)
2713 ipv6_mc_remap(idev);
2714 else if (event == NETDEV_PRE_TYPE_CHANGE)
2715 ipv6_mc_unmap(idev);
2718 static int addrconf_ifdown(struct net_device *dev, int how)
2720 struct net *net = dev_net(dev);
2721 struct inet6_dev *idev;
2722 struct inet6_ifaddr *ifa;
2723 int state, i;
2725 ASSERT_RTNL();
2727 rt6_ifdown(net, dev);
2728 neigh_ifdown(&nd_tbl, dev);
2730 idev = __in6_dev_get(dev);
2731 if (idev == NULL)
2732 return -ENODEV;
2735 * Step 1: remove reference to ipv6 device from parent device.
2736 * Do not dev_put!
2738 if (how) {
2739 idev->dead = 1;
2741 /* protected by rtnl_lock */
2742 RCU_INIT_POINTER(dev->ip6_ptr, NULL);
2744 /* Step 1.5: remove snmp6 entry */
2745 snmp6_unregister_dev(idev);
2749 /* Step 2: clear hash table */
2750 for (i = 0; i < IN6_ADDR_HSIZE; i++) {
2751 struct hlist_head *h = &inet6_addr_lst[i];
2752 struct hlist_node *n;
2754 spin_lock_bh(&addrconf_hash_lock);
2755 restart:
2756 hlist_for_each_entry_rcu(ifa, n, h, addr_lst) {
2757 if (ifa->idev == idev) {
2758 hlist_del_init_rcu(&ifa->addr_lst);
2759 addrconf_del_timer(ifa);
2760 goto restart;
2763 spin_unlock_bh(&addrconf_hash_lock);
2766 write_lock_bh(&idev->lock);
2768 /* Step 2: clear flags for stateless addrconf */
2769 if (!how)
2770 idev->if_flags &= ~(IF_RS_SENT|IF_RA_RCVD|IF_READY);
2772 #ifdef CONFIG_IPV6_PRIVACY
2773 if (how && del_timer(&idev->regen_timer))
2774 in6_dev_put(idev);
2776 /* Step 3: clear tempaddr list */
2777 while (!list_empty(&idev->tempaddr_list)) {
2778 ifa = list_first_entry(&idev->tempaddr_list,
2779 struct inet6_ifaddr, tmp_list);
2780 list_del(&ifa->tmp_list);
2781 write_unlock_bh(&idev->lock);
2782 spin_lock_bh(&ifa->lock);
2784 if (ifa->ifpub) {
2785 in6_ifa_put(ifa->ifpub);
2786 ifa->ifpub = NULL;
2788 spin_unlock_bh(&ifa->lock);
2789 in6_ifa_put(ifa);
2790 write_lock_bh(&idev->lock);
2792 #endif
2794 while (!list_empty(&idev->addr_list)) {
2795 ifa = list_first_entry(&idev->addr_list,
2796 struct inet6_ifaddr, if_list);
2797 addrconf_del_timer(ifa);
2799 list_del(&ifa->if_list);
2801 write_unlock_bh(&idev->lock);
2803 spin_lock_bh(&ifa->state_lock);
2804 state = ifa->state;
2805 ifa->state = INET6_IFADDR_STATE_DEAD;
2806 spin_unlock_bh(&ifa->state_lock);
2808 if (state != INET6_IFADDR_STATE_DEAD) {
2809 __ipv6_ifa_notify(RTM_DELADDR, ifa);
2810 atomic_notifier_call_chain(&inet6addr_chain, NETDEV_DOWN, ifa);
2812 in6_ifa_put(ifa);
2814 write_lock_bh(&idev->lock);
2817 write_unlock_bh(&idev->lock);
2819 /* Step 5: Discard multicast list */
2820 if (how)
2821 ipv6_mc_destroy_dev(idev);
2822 else
2823 ipv6_mc_down(idev);
2825 idev->tstamp = jiffies;
2827 /* Last: Shot the device (if unregistered) */
2828 if (how) {
2829 addrconf_sysctl_unregister(idev);
2830 neigh_parms_release(&nd_tbl, idev->nd_parms);
2831 neigh_ifdown(&nd_tbl, dev);
2832 in6_dev_put(idev);
2834 return 0;
2837 static void addrconf_rs_timer(unsigned long data)
2839 struct inet6_ifaddr *ifp = (struct inet6_ifaddr *) data;
2840 struct inet6_dev *idev = ifp->idev;
2842 read_lock(&idev->lock);
2843 if (idev->dead || !(idev->if_flags & IF_READY))
2844 goto out;
2846 if (idev->cnf.forwarding)
2847 goto out;
2849 /* Announcement received after solicitation was sent */
2850 if (idev->if_flags & IF_RA_RCVD)
2851 goto out;
2853 spin_lock(&ifp->lock);
2854 if (ifp->probes++ < idev->cnf.rtr_solicits) {
2855 /* The wait after the last probe can be shorter */
2856 addrconf_mod_timer(ifp, AC_RS,
2857 (ifp->probes == idev->cnf.rtr_solicits) ?
2858 idev->cnf.rtr_solicit_delay :
2859 idev->cnf.rtr_solicit_interval);
2860 spin_unlock(&ifp->lock);
2862 ndisc_send_rs(idev->dev, &ifp->addr, &in6addr_linklocal_allrouters);
2863 } else {
2864 spin_unlock(&ifp->lock);
2866 * Note: we do not support deprecated "all on-link"
2867 * assumption any longer.
2869 pr_debug("%s: no IPv6 routers present\n", idev->dev->name);
2872 out:
2873 read_unlock(&idev->lock);
2874 in6_ifa_put(ifp);
2878 * Duplicate Address Detection
2880 static void addrconf_dad_kick(struct inet6_ifaddr *ifp)
2882 unsigned long rand_num;
2883 struct inet6_dev *idev = ifp->idev;
2885 if (ifp->flags & IFA_F_OPTIMISTIC)
2886 rand_num = 0;
2887 else
2888 rand_num = net_random() % (idev->cnf.rtr_solicit_delay ? : 1);
2890 ifp->probes = idev->cnf.dad_transmits;
2891 addrconf_mod_timer(ifp, AC_DAD, rand_num);
2894 static void addrconf_dad_start(struct inet6_ifaddr *ifp)
2896 struct inet6_dev *idev = ifp->idev;
2897 struct net_device *dev = idev->dev;
2899 addrconf_join_solict(dev, &ifp->addr);
2901 net_srandom(ifp->addr.s6_addr32[3]);
2903 read_lock_bh(&idev->lock);
2904 spin_lock(&ifp->lock);
2905 if (ifp->state == INET6_IFADDR_STATE_DEAD)
2906 goto out;
2908 if (dev->flags&(IFF_NOARP|IFF_LOOPBACK) ||
2909 idev->cnf.accept_dad < 1 ||
2910 !(ifp->flags&IFA_F_TENTATIVE) ||
2911 ifp->flags & IFA_F_NODAD) {
2912 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
2913 spin_unlock(&ifp->lock);
2914 read_unlock_bh(&idev->lock);
2916 addrconf_dad_completed(ifp);
2917 return;
2920 if (!(idev->if_flags & IF_READY)) {
2921 spin_unlock(&ifp->lock);
2922 read_unlock_bh(&idev->lock);
2924 * If the device is not ready:
2925 * - keep it tentative if it is a permanent address.
2926 * - otherwise, kill it.
2928 in6_ifa_hold(ifp);
2929 addrconf_dad_stop(ifp, 0);
2930 return;
2934 * Optimistic nodes can start receiving
2935 * Frames right away
2937 if (ifp->flags & IFA_F_OPTIMISTIC)
2938 ip6_ins_rt(ifp->rt);
2940 addrconf_dad_kick(ifp);
2941 out:
2942 spin_unlock(&ifp->lock);
2943 read_unlock_bh(&idev->lock);
2946 static void addrconf_dad_timer(unsigned long data)
2948 struct inet6_ifaddr *ifp = (struct inet6_ifaddr *) data;
2949 struct inet6_dev *idev = ifp->idev;
2950 struct in6_addr mcaddr;
2952 if (!ifp->probes && addrconf_dad_end(ifp))
2953 goto out;
2955 read_lock(&idev->lock);
2956 if (idev->dead || !(idev->if_flags & IF_READY)) {
2957 read_unlock(&idev->lock);
2958 goto out;
2961 spin_lock(&ifp->lock);
2962 if (ifp->state == INET6_IFADDR_STATE_DEAD) {
2963 spin_unlock(&ifp->lock);
2964 read_unlock(&idev->lock);
2965 goto out;
2968 if (ifp->probes == 0) {
2970 * DAD was successful
2973 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
2974 spin_unlock(&ifp->lock);
2975 read_unlock(&idev->lock);
2977 addrconf_dad_completed(ifp);
2979 goto out;
2982 ifp->probes--;
2983 addrconf_mod_timer(ifp, AC_DAD, ifp->idev->nd_parms->retrans_time);
2984 spin_unlock(&ifp->lock);
2985 read_unlock(&idev->lock);
2987 /* send a neighbour solicitation for our addr */
2988 addrconf_addr_solict_mult(&ifp->addr, &mcaddr);
2989 ndisc_send_ns(ifp->idev->dev, NULL, &ifp->addr, &mcaddr, &in6addr_any);
2990 out:
2991 in6_ifa_put(ifp);
2994 static void addrconf_dad_completed(struct inet6_ifaddr *ifp)
2996 struct net_device *dev = ifp->idev->dev;
2999 * Configure the address for reception. Now it is valid.
3002 ipv6_ifa_notify(RTM_NEWADDR, ifp);
3004 /* If added prefix is link local and we are prepared to process
3005 router advertisements, start sending router solicitations.
3008 if (((ifp->idev->cnf.accept_ra == 1 && !ifp->idev->cnf.forwarding) ||
3009 ifp->idev->cnf.accept_ra == 2) &&
3010 ifp->idev->cnf.rtr_solicits > 0 &&
3011 (dev->flags&IFF_LOOPBACK) == 0 &&
3012 (ipv6_addr_type(&ifp->addr) & IPV6_ADDR_LINKLOCAL)) {
3014 * If a host as already performed a random delay
3015 * [...] as part of DAD [...] there is no need
3016 * to delay again before sending the first RS
3018 ndisc_send_rs(ifp->idev->dev, &ifp->addr, &in6addr_linklocal_allrouters);
3020 spin_lock_bh(&ifp->lock);
3021 ifp->probes = 1;
3022 ifp->idev->if_flags |= IF_RS_SENT;
3023 addrconf_mod_timer(ifp, AC_RS, ifp->idev->cnf.rtr_solicit_interval);
3024 spin_unlock_bh(&ifp->lock);
3028 static void addrconf_dad_run(struct inet6_dev *idev)
3030 struct inet6_ifaddr *ifp;
3032 read_lock_bh(&idev->lock);
3033 list_for_each_entry(ifp, &idev->addr_list, if_list) {
3034 spin_lock(&ifp->lock);
3035 if (ifp->flags & IFA_F_TENTATIVE &&
3036 ifp->state == INET6_IFADDR_STATE_DAD)
3037 addrconf_dad_kick(ifp);
3038 spin_unlock(&ifp->lock);
3040 read_unlock_bh(&idev->lock);
3043 #ifdef CONFIG_PROC_FS
3044 struct if6_iter_state {
3045 struct seq_net_private p;
3046 int bucket;
3047 int offset;
3050 static struct inet6_ifaddr *if6_get_first(struct seq_file *seq, loff_t pos)
3052 struct inet6_ifaddr *ifa = NULL;
3053 struct if6_iter_state *state = seq->private;
3054 struct net *net = seq_file_net(seq);
3055 int p = 0;
3057 /* initial bucket if pos is 0 */
3058 if (pos == 0) {
3059 state->bucket = 0;
3060 state->offset = 0;
3063 for (; state->bucket < IN6_ADDR_HSIZE; ++state->bucket) {
3064 struct hlist_node *n;
3065 hlist_for_each_entry_rcu_bh(ifa, n, &inet6_addr_lst[state->bucket],
3066 addr_lst) {
3067 if (!net_eq(dev_net(ifa->idev->dev), net))
3068 continue;
3069 /* sync with offset */
3070 if (p < state->offset) {
3071 p++;
3072 continue;
3074 state->offset++;
3075 return ifa;
3078 /* prepare for next bucket */
3079 state->offset = 0;
3080 p = 0;
3082 return NULL;
3085 static struct inet6_ifaddr *if6_get_next(struct seq_file *seq,
3086 struct inet6_ifaddr *ifa)
3088 struct if6_iter_state *state = seq->private;
3089 struct net *net = seq_file_net(seq);
3090 struct hlist_node *n = &ifa->addr_lst;
3092 hlist_for_each_entry_continue_rcu_bh(ifa, n, addr_lst) {
3093 if (!net_eq(dev_net(ifa->idev->dev), net))
3094 continue;
3095 state->offset++;
3096 return ifa;
3099 while (++state->bucket < IN6_ADDR_HSIZE) {
3100 state->offset = 0;
3101 hlist_for_each_entry_rcu_bh(ifa, n,
3102 &inet6_addr_lst[state->bucket], addr_lst) {
3103 if (!net_eq(dev_net(ifa->idev->dev), net))
3104 continue;
3105 state->offset++;
3106 return ifa;
3110 return NULL;
3113 static void *if6_seq_start(struct seq_file *seq, loff_t *pos)
3114 __acquires(rcu_bh)
3116 rcu_read_lock_bh();
3117 return if6_get_first(seq, *pos);
3120 static void *if6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
3122 struct inet6_ifaddr *ifa;
3124 ifa = if6_get_next(seq, v);
3125 ++*pos;
3126 return ifa;
3129 static void if6_seq_stop(struct seq_file *seq, void *v)
3130 __releases(rcu_bh)
3132 rcu_read_unlock_bh();
3135 static int if6_seq_show(struct seq_file *seq, void *v)
3137 struct inet6_ifaddr *ifp = (struct inet6_ifaddr *)v;
3138 seq_printf(seq, "%pi6 %02x %02x %02x %02x %8s\n",
3139 &ifp->addr,
3140 ifp->idev->dev->ifindex,
3141 ifp->prefix_len,
3142 ifp->scope,
3143 ifp->flags,
3144 ifp->idev->dev->name);
3145 return 0;
3148 static const struct seq_operations if6_seq_ops = {
3149 .start = if6_seq_start,
3150 .next = if6_seq_next,
3151 .show = if6_seq_show,
3152 .stop = if6_seq_stop,
3155 static int if6_seq_open(struct inode *inode, struct file *file)
3157 return seq_open_net(inode, file, &if6_seq_ops,
3158 sizeof(struct if6_iter_state));
3161 static const struct file_operations if6_fops = {
3162 .owner = THIS_MODULE,
3163 .open = if6_seq_open,
3164 .read = seq_read,
3165 .llseek = seq_lseek,
3166 .release = seq_release_net,
3169 static int __net_init if6_proc_net_init(struct net *net)
3171 if (!proc_net_fops_create(net, "if_inet6", S_IRUGO, &if6_fops))
3172 return -ENOMEM;
3173 return 0;
3176 static void __net_exit if6_proc_net_exit(struct net *net)
3178 proc_net_remove(net, "if_inet6");
3181 static struct pernet_operations if6_proc_net_ops = {
3182 .init = if6_proc_net_init,
3183 .exit = if6_proc_net_exit,
3186 int __init if6_proc_init(void)
3188 return register_pernet_subsys(&if6_proc_net_ops);
3191 void if6_proc_exit(void)
3193 unregister_pernet_subsys(&if6_proc_net_ops);
3195 #endif /* CONFIG_PROC_FS */
3197 #if defined(CONFIG_IPV6_MIP6) || defined(CONFIG_IPV6_MIP6_MODULE)
3198 /* Check if address is a home address configured on any interface. */
3199 int ipv6_chk_home_addr(struct net *net, const struct in6_addr *addr)
3201 int ret = 0;
3202 struct inet6_ifaddr *ifp = NULL;
3203 struct hlist_node *n;
3204 unsigned int hash = inet6_addr_hash(addr);
3206 rcu_read_lock_bh();
3207 hlist_for_each_entry_rcu_bh(ifp, n, &inet6_addr_lst[hash], addr_lst) {
3208 if (!net_eq(dev_net(ifp->idev->dev), net))
3209 continue;
3210 if (ipv6_addr_equal(&ifp->addr, addr) &&
3211 (ifp->flags & IFA_F_HOMEADDRESS)) {
3212 ret = 1;
3213 break;
3216 rcu_read_unlock_bh();
3217 return ret;
3219 #endif
3222 * Periodic address status verification
3225 static void addrconf_verify(unsigned long foo)
3227 unsigned long now, next, next_sec, next_sched;
3228 struct inet6_ifaddr *ifp;
3229 struct hlist_node *node;
3230 int i;
3232 rcu_read_lock_bh();
3233 spin_lock(&addrconf_verify_lock);
3234 now = jiffies;
3235 next = round_jiffies_up(now + ADDR_CHECK_FREQUENCY);
3237 del_timer(&addr_chk_timer);
3239 for (i = 0; i < IN6_ADDR_HSIZE; i++) {
3240 restart:
3241 hlist_for_each_entry_rcu_bh(ifp, node,
3242 &inet6_addr_lst[i], addr_lst) {
3243 unsigned long age;
3245 if (ifp->flags & IFA_F_PERMANENT)
3246 continue;
3248 spin_lock(&ifp->lock);
3249 /* We try to batch several events at once. */
3250 age = (now - ifp->tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
3252 if (ifp->valid_lft != INFINITY_LIFE_TIME &&
3253 age >= ifp->valid_lft) {
3254 spin_unlock(&ifp->lock);
3255 in6_ifa_hold(ifp);
3256 ipv6_del_addr(ifp);
3257 goto restart;
3258 } else if (ifp->prefered_lft == INFINITY_LIFE_TIME) {
3259 spin_unlock(&ifp->lock);
3260 continue;
3261 } else if (age >= ifp->prefered_lft) {
3262 /* jiffies - ifp->tstamp > age >= ifp->prefered_lft */
3263 int deprecate = 0;
3265 if (!(ifp->flags&IFA_F_DEPRECATED)) {
3266 deprecate = 1;
3267 ifp->flags |= IFA_F_DEPRECATED;
3270 if (time_before(ifp->tstamp + ifp->valid_lft * HZ, next))
3271 next = ifp->tstamp + ifp->valid_lft * HZ;
3273 spin_unlock(&ifp->lock);
3275 if (deprecate) {
3276 in6_ifa_hold(ifp);
3278 ipv6_ifa_notify(0, ifp);
3279 in6_ifa_put(ifp);
3280 goto restart;
3282 #ifdef CONFIG_IPV6_PRIVACY
3283 } else if ((ifp->flags&IFA_F_TEMPORARY) &&
3284 !(ifp->flags&IFA_F_TENTATIVE)) {
3285 unsigned long regen_advance = ifp->idev->cnf.regen_max_retry *
3286 ifp->idev->cnf.dad_transmits *
3287 ifp->idev->nd_parms->retrans_time / HZ;
3289 if (age >= ifp->prefered_lft - regen_advance) {
3290 struct inet6_ifaddr *ifpub = ifp->ifpub;
3291 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3292 next = ifp->tstamp + ifp->prefered_lft * HZ;
3293 if (!ifp->regen_count && ifpub) {
3294 ifp->regen_count++;
3295 in6_ifa_hold(ifp);
3296 in6_ifa_hold(ifpub);
3297 spin_unlock(&ifp->lock);
3299 spin_lock(&ifpub->lock);
3300 ifpub->regen_count = 0;
3301 spin_unlock(&ifpub->lock);
3302 ipv6_create_tempaddr(ifpub, ifp);
3303 in6_ifa_put(ifpub);
3304 in6_ifa_put(ifp);
3305 goto restart;
3307 } else if (time_before(ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ, next))
3308 next = ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ;
3309 spin_unlock(&ifp->lock);
3310 #endif
3311 } else {
3312 /* ifp->prefered_lft <= ifp->valid_lft */
3313 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3314 next = ifp->tstamp + ifp->prefered_lft * HZ;
3315 spin_unlock(&ifp->lock);
3320 next_sec = round_jiffies_up(next);
3321 next_sched = next;
3323 /* If rounded timeout is accurate enough, accept it. */
3324 if (time_before(next_sec, next + ADDRCONF_TIMER_FUZZ))
3325 next_sched = next_sec;
3327 /* And minimum interval is ADDRCONF_TIMER_FUZZ_MAX. */
3328 if (time_before(next_sched, jiffies + ADDRCONF_TIMER_FUZZ_MAX))
3329 next_sched = jiffies + ADDRCONF_TIMER_FUZZ_MAX;
3331 ADBG((KERN_DEBUG "now = %lu, schedule = %lu, rounded schedule = %lu => %lu\n",
3332 now, next, next_sec, next_sched));
3334 addr_chk_timer.expires = next_sched;
3335 add_timer(&addr_chk_timer);
3336 spin_unlock(&addrconf_verify_lock);
3337 rcu_read_unlock_bh();
3340 static struct in6_addr *extract_addr(struct nlattr *addr, struct nlattr *local)
3342 struct in6_addr *pfx = NULL;
3344 if (addr)
3345 pfx = nla_data(addr);
3347 if (local) {
3348 if (pfx && nla_memcmp(local, pfx, sizeof(*pfx)))
3349 pfx = NULL;
3350 else
3351 pfx = nla_data(local);
3354 return pfx;
3357 static const struct nla_policy ifa_ipv6_policy[IFA_MAX+1] = {
3358 [IFA_ADDRESS] = { .len = sizeof(struct in6_addr) },
3359 [IFA_LOCAL] = { .len = sizeof(struct in6_addr) },
3360 [IFA_CACHEINFO] = { .len = sizeof(struct ifa_cacheinfo) },
3363 static int
3364 inet6_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
3366 struct net *net = sock_net(skb->sk);
3367 struct ifaddrmsg *ifm;
3368 struct nlattr *tb[IFA_MAX+1];
3369 struct in6_addr *pfx;
3370 int err;
3372 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3373 if (err < 0)
3374 return err;
3376 ifm = nlmsg_data(nlh);
3377 pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3378 if (pfx == NULL)
3379 return -EINVAL;
3381 return inet6_addr_del(net, ifm->ifa_index, pfx, ifm->ifa_prefixlen);
3384 static int inet6_addr_modify(struct inet6_ifaddr *ifp, u8 ifa_flags,
3385 u32 prefered_lft, u32 valid_lft)
3387 u32 flags;
3388 clock_t expires;
3389 unsigned long timeout;
3391 if (!valid_lft || (prefered_lft > valid_lft))
3392 return -EINVAL;
3394 timeout = addrconf_timeout_fixup(valid_lft, HZ);
3395 if (addrconf_finite_timeout(timeout)) {
3396 expires = jiffies_to_clock_t(timeout * HZ);
3397 valid_lft = timeout;
3398 flags = RTF_EXPIRES;
3399 } else {
3400 expires = 0;
3401 flags = 0;
3402 ifa_flags |= IFA_F_PERMANENT;
3405 timeout = addrconf_timeout_fixup(prefered_lft, HZ);
3406 if (addrconf_finite_timeout(timeout)) {
3407 if (timeout == 0)
3408 ifa_flags |= IFA_F_DEPRECATED;
3409 prefered_lft = timeout;
3412 spin_lock_bh(&ifp->lock);
3413 ifp->flags = (ifp->flags & ~(IFA_F_DEPRECATED | IFA_F_PERMANENT | IFA_F_NODAD | IFA_F_HOMEADDRESS)) | ifa_flags;
3414 ifp->tstamp = jiffies;
3415 ifp->valid_lft = valid_lft;
3416 ifp->prefered_lft = prefered_lft;
3418 spin_unlock_bh(&ifp->lock);
3419 if (!(ifp->flags&IFA_F_TENTATIVE))
3420 ipv6_ifa_notify(0, ifp);
3422 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, ifp->idev->dev,
3423 expires, flags);
3424 addrconf_verify(0);
3426 return 0;
3429 static int
3430 inet6_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
3432 struct net *net = sock_net(skb->sk);
3433 struct ifaddrmsg *ifm;
3434 struct nlattr *tb[IFA_MAX+1];
3435 struct in6_addr *pfx;
3436 struct inet6_ifaddr *ifa;
3437 struct net_device *dev;
3438 u32 valid_lft = INFINITY_LIFE_TIME, preferred_lft = INFINITY_LIFE_TIME;
3439 u8 ifa_flags;
3440 int err;
3442 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3443 if (err < 0)
3444 return err;
3446 ifm = nlmsg_data(nlh);
3447 pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3448 if (pfx == NULL)
3449 return -EINVAL;
3451 if (tb[IFA_CACHEINFO]) {
3452 struct ifa_cacheinfo *ci;
3454 ci = nla_data(tb[IFA_CACHEINFO]);
3455 valid_lft = ci->ifa_valid;
3456 preferred_lft = ci->ifa_prefered;
3457 } else {
3458 preferred_lft = INFINITY_LIFE_TIME;
3459 valid_lft = INFINITY_LIFE_TIME;
3462 dev = __dev_get_by_index(net, ifm->ifa_index);
3463 if (dev == NULL)
3464 return -ENODEV;
3466 /* We ignore other flags so far. */
3467 ifa_flags = ifm->ifa_flags & (IFA_F_NODAD | IFA_F_HOMEADDRESS);
3469 ifa = ipv6_get_ifaddr(net, pfx, dev, 1);
3470 if (ifa == NULL) {
3472 * It would be best to check for !NLM_F_CREATE here but
3473 * userspace alreay relies on not having to provide this.
3475 return inet6_addr_add(net, ifm->ifa_index, pfx,
3476 ifm->ifa_prefixlen, ifa_flags,
3477 preferred_lft, valid_lft);
3480 if (nlh->nlmsg_flags & NLM_F_EXCL ||
3481 !(nlh->nlmsg_flags & NLM_F_REPLACE))
3482 err = -EEXIST;
3483 else
3484 err = inet6_addr_modify(ifa, ifa_flags, preferred_lft, valid_lft);
3486 in6_ifa_put(ifa);
3488 return err;
3491 static void put_ifaddrmsg(struct nlmsghdr *nlh, u8 prefixlen, u8 flags,
3492 u8 scope, int ifindex)
3494 struct ifaddrmsg *ifm;
3496 ifm = nlmsg_data(nlh);
3497 ifm->ifa_family = AF_INET6;
3498 ifm->ifa_prefixlen = prefixlen;
3499 ifm->ifa_flags = flags;
3500 ifm->ifa_scope = scope;
3501 ifm->ifa_index = ifindex;
3504 static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
3505 unsigned long tstamp, u32 preferred, u32 valid)
3507 struct ifa_cacheinfo ci;
3509 ci.cstamp = cstamp_delta(cstamp);
3510 ci.tstamp = cstamp_delta(tstamp);
3511 ci.ifa_prefered = preferred;
3512 ci.ifa_valid = valid;
3514 return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
3517 static inline int rt_scope(int ifa_scope)
3519 if (ifa_scope & IFA_HOST)
3520 return RT_SCOPE_HOST;
3521 else if (ifa_scope & IFA_LINK)
3522 return RT_SCOPE_LINK;
3523 else if (ifa_scope & IFA_SITE)
3524 return RT_SCOPE_SITE;
3525 else
3526 return RT_SCOPE_UNIVERSE;
3529 static inline int inet6_ifaddr_msgsize(void)
3531 return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
3532 + nla_total_size(16) /* IFA_ADDRESS */
3533 + nla_total_size(sizeof(struct ifa_cacheinfo));
3536 static int inet6_fill_ifaddr(struct sk_buff *skb, struct inet6_ifaddr *ifa,
3537 u32 portid, u32 seq, int event, unsigned int flags)
3539 struct nlmsghdr *nlh;
3540 u32 preferred, valid;
3542 nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
3543 if (nlh == NULL)
3544 return -EMSGSIZE;
3546 put_ifaddrmsg(nlh, ifa->prefix_len, ifa->flags, rt_scope(ifa->scope),
3547 ifa->idev->dev->ifindex);
3549 if (!(ifa->flags&IFA_F_PERMANENT)) {
3550 preferred = ifa->prefered_lft;
3551 valid = ifa->valid_lft;
3552 if (preferred != INFINITY_LIFE_TIME) {
3553 long tval = (jiffies - ifa->tstamp)/HZ;
3554 if (preferred > tval)
3555 preferred -= tval;
3556 else
3557 preferred = 0;
3558 if (valid != INFINITY_LIFE_TIME) {
3559 if (valid > tval)
3560 valid -= tval;
3561 else
3562 valid = 0;
3565 } else {
3566 preferred = INFINITY_LIFE_TIME;
3567 valid = INFINITY_LIFE_TIME;
3570 if (nla_put(skb, IFA_ADDRESS, 16, &ifa->addr) < 0 ||
3571 put_cacheinfo(skb, ifa->cstamp, ifa->tstamp, preferred, valid) < 0) {
3572 nlmsg_cancel(skb, nlh);
3573 return -EMSGSIZE;
3576 return nlmsg_end(skb, nlh);
3579 static int inet6_fill_ifmcaddr(struct sk_buff *skb, struct ifmcaddr6 *ifmca,
3580 u32 portid, u32 seq, int event, u16 flags)
3582 struct nlmsghdr *nlh;
3583 u8 scope = RT_SCOPE_UNIVERSE;
3584 int ifindex = ifmca->idev->dev->ifindex;
3586 if (ipv6_addr_scope(&ifmca->mca_addr) & IFA_SITE)
3587 scope = RT_SCOPE_SITE;
3589 nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
3590 if (nlh == NULL)
3591 return -EMSGSIZE;
3593 put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3594 if (nla_put(skb, IFA_MULTICAST, 16, &ifmca->mca_addr) < 0 ||
3595 put_cacheinfo(skb, ifmca->mca_cstamp, ifmca->mca_tstamp,
3596 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3597 nlmsg_cancel(skb, nlh);
3598 return -EMSGSIZE;
3601 return nlmsg_end(skb, nlh);
3604 static int inet6_fill_ifacaddr(struct sk_buff *skb, struct ifacaddr6 *ifaca,
3605 u32 portid, u32 seq, int event, unsigned int flags)
3607 struct nlmsghdr *nlh;
3608 u8 scope = RT_SCOPE_UNIVERSE;
3609 int ifindex = ifaca->aca_idev->dev->ifindex;
3611 if (ipv6_addr_scope(&ifaca->aca_addr) & IFA_SITE)
3612 scope = RT_SCOPE_SITE;
3614 nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
3615 if (nlh == NULL)
3616 return -EMSGSIZE;
3618 put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3619 if (nla_put(skb, IFA_ANYCAST, 16, &ifaca->aca_addr) < 0 ||
3620 put_cacheinfo(skb, ifaca->aca_cstamp, ifaca->aca_tstamp,
3621 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3622 nlmsg_cancel(skb, nlh);
3623 return -EMSGSIZE;
3626 return nlmsg_end(skb, nlh);
3629 enum addr_type_t {
3630 UNICAST_ADDR,
3631 MULTICAST_ADDR,
3632 ANYCAST_ADDR,
3635 /* called with rcu_read_lock() */
3636 static int in6_dump_addrs(struct inet6_dev *idev, struct sk_buff *skb,
3637 struct netlink_callback *cb, enum addr_type_t type,
3638 int s_ip_idx, int *p_ip_idx)
3640 struct ifmcaddr6 *ifmca;
3641 struct ifacaddr6 *ifaca;
3642 int err = 1;
3643 int ip_idx = *p_ip_idx;
3645 read_lock_bh(&idev->lock);
3646 switch (type) {
3647 case UNICAST_ADDR: {
3648 struct inet6_ifaddr *ifa;
3650 /* unicast address incl. temp addr */
3651 list_for_each_entry(ifa, &idev->addr_list, if_list) {
3652 if (++ip_idx < s_ip_idx)
3653 continue;
3654 err = inet6_fill_ifaddr(skb, ifa,
3655 NETLINK_CB(cb->skb).portid,
3656 cb->nlh->nlmsg_seq,
3657 RTM_NEWADDR,
3658 NLM_F_MULTI);
3659 if (err <= 0)
3660 break;
3662 break;
3664 case MULTICAST_ADDR:
3665 /* multicast address */
3666 for (ifmca = idev->mc_list; ifmca;
3667 ifmca = ifmca->next, ip_idx++) {
3668 if (ip_idx < s_ip_idx)
3669 continue;
3670 err = inet6_fill_ifmcaddr(skb, ifmca,
3671 NETLINK_CB(cb->skb).portid,
3672 cb->nlh->nlmsg_seq,
3673 RTM_GETMULTICAST,
3674 NLM_F_MULTI);
3675 if (err <= 0)
3676 break;
3678 break;
3679 case ANYCAST_ADDR:
3680 /* anycast address */
3681 for (ifaca = idev->ac_list; ifaca;
3682 ifaca = ifaca->aca_next, ip_idx++) {
3683 if (ip_idx < s_ip_idx)
3684 continue;
3685 err = inet6_fill_ifacaddr(skb, ifaca,
3686 NETLINK_CB(cb->skb).portid,
3687 cb->nlh->nlmsg_seq,
3688 RTM_GETANYCAST,
3689 NLM_F_MULTI);
3690 if (err <= 0)
3691 break;
3693 break;
3694 default:
3695 break;
3697 read_unlock_bh(&idev->lock);
3698 *p_ip_idx = ip_idx;
3699 return err;
3702 static int inet6_dump_addr(struct sk_buff *skb, struct netlink_callback *cb,
3703 enum addr_type_t type)
3705 struct net *net = sock_net(skb->sk);
3706 int h, s_h;
3707 int idx, ip_idx;
3708 int s_idx, s_ip_idx;
3709 struct net_device *dev;
3710 struct inet6_dev *idev;
3711 struct hlist_head *head;
3712 struct hlist_node *node;
3714 s_h = cb->args[0];
3715 s_idx = idx = cb->args[1];
3716 s_ip_idx = ip_idx = cb->args[2];
3718 rcu_read_lock();
3719 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
3720 idx = 0;
3721 head = &net->dev_index_head[h];
3722 hlist_for_each_entry_rcu(dev, node, head, index_hlist) {
3723 if (idx < s_idx)
3724 goto cont;
3725 if (h > s_h || idx > s_idx)
3726 s_ip_idx = 0;
3727 ip_idx = 0;
3728 idev = __in6_dev_get(dev);
3729 if (!idev)
3730 goto cont;
3732 if (in6_dump_addrs(idev, skb, cb, type,
3733 s_ip_idx, &ip_idx) <= 0)
3734 goto done;
3735 cont:
3736 idx++;
3739 done:
3740 rcu_read_unlock();
3741 cb->args[0] = h;
3742 cb->args[1] = idx;
3743 cb->args[2] = ip_idx;
3745 return skb->len;
3748 static int inet6_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
3750 enum addr_type_t type = UNICAST_ADDR;
3752 return inet6_dump_addr(skb, cb, type);
3755 static int inet6_dump_ifmcaddr(struct sk_buff *skb, struct netlink_callback *cb)
3757 enum addr_type_t type = MULTICAST_ADDR;
3759 return inet6_dump_addr(skb, cb, type);
3763 static int inet6_dump_ifacaddr(struct sk_buff *skb, struct netlink_callback *cb)
3765 enum addr_type_t type = ANYCAST_ADDR;
3767 return inet6_dump_addr(skb, cb, type);
3770 static int inet6_rtm_getaddr(struct sk_buff *in_skb, struct nlmsghdr *nlh,
3771 void *arg)
3773 struct net *net = sock_net(in_skb->sk);
3774 struct ifaddrmsg *ifm;
3775 struct nlattr *tb[IFA_MAX+1];
3776 struct in6_addr *addr = NULL;
3777 struct net_device *dev = NULL;
3778 struct inet6_ifaddr *ifa;
3779 struct sk_buff *skb;
3780 int err;
3782 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3783 if (err < 0)
3784 goto errout;
3786 addr = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3787 if (addr == NULL) {
3788 err = -EINVAL;
3789 goto errout;
3792 ifm = nlmsg_data(nlh);
3793 if (ifm->ifa_index)
3794 dev = __dev_get_by_index(net, ifm->ifa_index);
3796 ifa = ipv6_get_ifaddr(net, addr, dev, 1);
3797 if (!ifa) {
3798 err = -EADDRNOTAVAIL;
3799 goto errout;
3802 skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_KERNEL);
3803 if (!skb) {
3804 err = -ENOBUFS;
3805 goto errout_ifa;
3808 err = inet6_fill_ifaddr(skb, ifa, NETLINK_CB(in_skb).portid,
3809 nlh->nlmsg_seq, RTM_NEWADDR, 0);
3810 if (err < 0) {
3811 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
3812 WARN_ON(err == -EMSGSIZE);
3813 kfree_skb(skb);
3814 goto errout_ifa;
3816 err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
3817 errout_ifa:
3818 in6_ifa_put(ifa);
3819 errout:
3820 return err;
3823 static void inet6_ifa_notify(int event, struct inet6_ifaddr *ifa)
3825 struct sk_buff *skb;
3826 struct net *net = dev_net(ifa->idev->dev);
3827 int err = -ENOBUFS;
3829 skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_ATOMIC);
3830 if (skb == NULL)
3831 goto errout;
3833 err = inet6_fill_ifaddr(skb, ifa, 0, 0, event, 0);
3834 if (err < 0) {
3835 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
3836 WARN_ON(err == -EMSGSIZE);
3837 kfree_skb(skb);
3838 goto errout;
3840 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
3841 return;
3842 errout:
3843 if (err < 0)
3844 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
3847 static inline void ipv6_store_devconf(struct ipv6_devconf *cnf,
3848 __s32 *array, int bytes)
3850 BUG_ON(bytes < (DEVCONF_MAX * 4));
3852 memset(array, 0, bytes);
3853 array[DEVCONF_FORWARDING] = cnf->forwarding;
3854 array[DEVCONF_HOPLIMIT] = cnf->hop_limit;
3855 array[DEVCONF_MTU6] = cnf->mtu6;
3856 array[DEVCONF_ACCEPT_RA] = cnf->accept_ra;
3857 array[DEVCONF_ACCEPT_REDIRECTS] = cnf->accept_redirects;
3858 array[DEVCONF_AUTOCONF] = cnf->autoconf;
3859 array[DEVCONF_DAD_TRANSMITS] = cnf->dad_transmits;
3860 array[DEVCONF_RTR_SOLICITS] = cnf->rtr_solicits;
3861 array[DEVCONF_RTR_SOLICIT_INTERVAL] =
3862 jiffies_to_msecs(cnf->rtr_solicit_interval);
3863 array[DEVCONF_RTR_SOLICIT_DELAY] =
3864 jiffies_to_msecs(cnf->rtr_solicit_delay);
3865 array[DEVCONF_FORCE_MLD_VERSION] = cnf->force_mld_version;
3866 #ifdef CONFIG_IPV6_PRIVACY
3867 array[DEVCONF_USE_TEMPADDR] = cnf->use_tempaddr;
3868 array[DEVCONF_TEMP_VALID_LFT] = cnf->temp_valid_lft;
3869 array[DEVCONF_TEMP_PREFERED_LFT] = cnf->temp_prefered_lft;
3870 array[DEVCONF_REGEN_MAX_RETRY] = cnf->regen_max_retry;
3871 array[DEVCONF_MAX_DESYNC_FACTOR] = cnf->max_desync_factor;
3872 #endif
3873 array[DEVCONF_MAX_ADDRESSES] = cnf->max_addresses;
3874 array[DEVCONF_ACCEPT_RA_DEFRTR] = cnf->accept_ra_defrtr;
3875 array[DEVCONF_ACCEPT_RA_PINFO] = cnf->accept_ra_pinfo;
3876 #ifdef CONFIG_IPV6_ROUTER_PREF
3877 array[DEVCONF_ACCEPT_RA_RTR_PREF] = cnf->accept_ra_rtr_pref;
3878 array[DEVCONF_RTR_PROBE_INTERVAL] =
3879 jiffies_to_msecs(cnf->rtr_probe_interval);
3880 #ifdef CONFIG_IPV6_ROUTE_INFO
3881 array[DEVCONF_ACCEPT_RA_RT_INFO_MAX_PLEN] = cnf->accept_ra_rt_info_max_plen;
3882 #endif
3883 #endif
3884 array[DEVCONF_PROXY_NDP] = cnf->proxy_ndp;
3885 array[DEVCONF_ACCEPT_SOURCE_ROUTE] = cnf->accept_source_route;
3886 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
3887 array[DEVCONF_OPTIMISTIC_DAD] = cnf->optimistic_dad;
3888 #endif
3889 #ifdef CONFIG_IPV6_MROUTE
3890 array[DEVCONF_MC_FORWARDING] = cnf->mc_forwarding;
3891 #endif
3892 array[DEVCONF_DISABLE_IPV6] = cnf->disable_ipv6;
3893 array[DEVCONF_ACCEPT_DAD] = cnf->accept_dad;
3894 array[DEVCONF_FORCE_TLLAO] = cnf->force_tllao;
3897 static inline size_t inet6_ifla6_size(void)
3899 return nla_total_size(4) /* IFLA_INET6_FLAGS */
3900 + nla_total_size(sizeof(struct ifla_cacheinfo))
3901 + nla_total_size(DEVCONF_MAX * 4) /* IFLA_INET6_CONF */
3902 + nla_total_size(IPSTATS_MIB_MAX * 8) /* IFLA_INET6_STATS */
3903 + nla_total_size(ICMP6_MIB_MAX * 8); /* IFLA_INET6_ICMP6STATS */
3906 static inline size_t inet6_if_nlmsg_size(void)
3908 return NLMSG_ALIGN(sizeof(struct ifinfomsg))
3909 + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
3910 + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
3911 + nla_total_size(4) /* IFLA_MTU */
3912 + nla_total_size(4) /* IFLA_LINK */
3913 + nla_total_size(inet6_ifla6_size()); /* IFLA_PROTINFO */
3916 static inline void __snmp6_fill_statsdev(u64 *stats, atomic_long_t *mib,
3917 int items, int bytes)
3919 int i;
3920 int pad = bytes - sizeof(u64) * items;
3921 BUG_ON(pad < 0);
3923 /* Use put_unaligned() because stats may not be aligned for u64. */
3924 put_unaligned(items, &stats[0]);
3925 for (i = 1; i < items; i++)
3926 put_unaligned(atomic_long_read(&mib[i]), &stats[i]);
3928 memset(&stats[items], 0, pad);
3931 static inline void __snmp6_fill_stats64(u64 *stats, void __percpu **mib,
3932 int items, int bytes, size_t syncpoff)
3934 int i;
3935 int pad = bytes - sizeof(u64) * items;
3936 BUG_ON(pad < 0);
3938 /* Use put_unaligned() because stats may not be aligned for u64. */
3939 put_unaligned(items, &stats[0]);
3940 for (i = 1; i < items; i++)
3941 put_unaligned(snmp_fold_field64(mib, i, syncpoff), &stats[i]);
3943 memset(&stats[items], 0, pad);
3946 static void snmp6_fill_stats(u64 *stats, struct inet6_dev *idev, int attrtype,
3947 int bytes)
3949 switch (attrtype) {
3950 case IFLA_INET6_STATS:
3951 __snmp6_fill_stats64(stats, (void __percpu **)idev->stats.ipv6,
3952 IPSTATS_MIB_MAX, bytes, offsetof(struct ipstats_mib, syncp));
3953 break;
3954 case IFLA_INET6_ICMP6STATS:
3955 __snmp6_fill_statsdev(stats, idev->stats.icmpv6dev->mibs, ICMP6_MIB_MAX, bytes);
3956 break;
3960 static int inet6_fill_ifla6_attrs(struct sk_buff *skb, struct inet6_dev *idev)
3962 struct nlattr *nla;
3963 struct ifla_cacheinfo ci;
3965 if (nla_put_u32(skb, IFLA_INET6_FLAGS, idev->if_flags))
3966 goto nla_put_failure;
3967 ci.max_reasm_len = IPV6_MAXPLEN;
3968 ci.tstamp = cstamp_delta(idev->tstamp);
3969 ci.reachable_time = jiffies_to_msecs(idev->nd_parms->reachable_time);
3970 ci.retrans_time = jiffies_to_msecs(idev->nd_parms->retrans_time);
3971 if (nla_put(skb, IFLA_INET6_CACHEINFO, sizeof(ci), &ci))
3972 goto nla_put_failure;
3973 nla = nla_reserve(skb, IFLA_INET6_CONF, DEVCONF_MAX * sizeof(s32));
3974 if (nla == NULL)
3975 goto nla_put_failure;
3976 ipv6_store_devconf(&idev->cnf, nla_data(nla), nla_len(nla));
3978 /* XXX - MC not implemented */
3980 nla = nla_reserve(skb, IFLA_INET6_STATS, IPSTATS_MIB_MAX * sizeof(u64));
3981 if (nla == NULL)
3982 goto nla_put_failure;
3983 snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_STATS, nla_len(nla));
3985 nla = nla_reserve(skb, IFLA_INET6_ICMP6STATS, ICMP6_MIB_MAX * sizeof(u64));
3986 if (nla == NULL)
3987 goto nla_put_failure;
3988 snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_ICMP6STATS, nla_len(nla));
3990 return 0;
3992 nla_put_failure:
3993 return -EMSGSIZE;
3996 static size_t inet6_get_link_af_size(const struct net_device *dev)
3998 if (!__in6_dev_get(dev))
3999 return 0;
4001 return inet6_ifla6_size();
4004 static int inet6_fill_link_af(struct sk_buff *skb, const struct net_device *dev)
4006 struct inet6_dev *idev = __in6_dev_get(dev);
4008 if (!idev)
4009 return -ENODATA;
4011 if (inet6_fill_ifla6_attrs(skb, idev) < 0)
4012 return -EMSGSIZE;
4014 return 0;
4017 static int inet6_fill_ifinfo(struct sk_buff *skb, struct inet6_dev *idev,
4018 u32 portid, u32 seq, int event, unsigned int flags)
4020 struct net_device *dev = idev->dev;
4021 struct ifinfomsg *hdr;
4022 struct nlmsghdr *nlh;
4023 void *protoinfo;
4025 nlh = nlmsg_put(skb, portid, seq, event, sizeof(*hdr), flags);
4026 if (nlh == NULL)
4027 return -EMSGSIZE;
4029 hdr = nlmsg_data(nlh);
4030 hdr->ifi_family = AF_INET6;
4031 hdr->__ifi_pad = 0;
4032 hdr->ifi_type = dev->type;
4033 hdr->ifi_index = dev->ifindex;
4034 hdr->ifi_flags = dev_get_flags(dev);
4035 hdr->ifi_change = 0;
4037 if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
4038 (dev->addr_len &&
4039 nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
4040 nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
4041 (dev->ifindex != dev->iflink &&
4042 nla_put_u32(skb, IFLA_LINK, dev->iflink)))
4043 goto nla_put_failure;
4044 protoinfo = nla_nest_start(skb, IFLA_PROTINFO);
4045 if (protoinfo == NULL)
4046 goto nla_put_failure;
4048 if (inet6_fill_ifla6_attrs(skb, idev) < 0)
4049 goto nla_put_failure;
4051 nla_nest_end(skb, protoinfo);
4052 return nlmsg_end(skb, nlh);
4054 nla_put_failure:
4055 nlmsg_cancel(skb, nlh);
4056 return -EMSGSIZE;
4059 static int inet6_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
4061 struct net *net = sock_net(skb->sk);
4062 int h, s_h;
4063 int idx = 0, s_idx;
4064 struct net_device *dev;
4065 struct inet6_dev *idev;
4066 struct hlist_head *head;
4067 struct hlist_node *node;
4069 s_h = cb->args[0];
4070 s_idx = cb->args[1];
4072 rcu_read_lock();
4073 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
4074 idx = 0;
4075 head = &net->dev_index_head[h];
4076 hlist_for_each_entry_rcu(dev, node, head, index_hlist) {
4077 if (idx < s_idx)
4078 goto cont;
4079 idev = __in6_dev_get(dev);
4080 if (!idev)
4081 goto cont;
4082 if (inet6_fill_ifinfo(skb, idev,
4083 NETLINK_CB(cb->skb).portid,
4084 cb->nlh->nlmsg_seq,
4085 RTM_NEWLINK, NLM_F_MULTI) <= 0)
4086 goto out;
4087 cont:
4088 idx++;
4091 out:
4092 rcu_read_unlock();
4093 cb->args[1] = idx;
4094 cb->args[0] = h;
4096 return skb->len;
4099 void inet6_ifinfo_notify(int event, struct inet6_dev *idev)
4101 struct sk_buff *skb;
4102 struct net *net = dev_net(idev->dev);
4103 int err = -ENOBUFS;
4105 skb = nlmsg_new(inet6_if_nlmsg_size(), GFP_ATOMIC);
4106 if (skb == NULL)
4107 goto errout;
4109 err = inet6_fill_ifinfo(skb, idev, 0, 0, event, 0);
4110 if (err < 0) {
4111 /* -EMSGSIZE implies BUG in inet6_if_nlmsg_size() */
4112 WARN_ON(err == -EMSGSIZE);
4113 kfree_skb(skb);
4114 goto errout;
4116 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFINFO, NULL, GFP_ATOMIC);
4117 return;
4118 errout:
4119 if (err < 0)
4120 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFINFO, err);
4123 static inline size_t inet6_prefix_nlmsg_size(void)
4125 return NLMSG_ALIGN(sizeof(struct prefixmsg))
4126 + nla_total_size(sizeof(struct in6_addr))
4127 + nla_total_size(sizeof(struct prefix_cacheinfo));
4130 static int inet6_fill_prefix(struct sk_buff *skb, struct inet6_dev *idev,
4131 struct prefix_info *pinfo, u32 portid, u32 seq,
4132 int event, unsigned int flags)
4134 struct prefixmsg *pmsg;
4135 struct nlmsghdr *nlh;
4136 struct prefix_cacheinfo ci;
4138 nlh = nlmsg_put(skb, portid, seq, event, sizeof(*pmsg), flags);
4139 if (nlh == NULL)
4140 return -EMSGSIZE;
4142 pmsg = nlmsg_data(nlh);
4143 pmsg->prefix_family = AF_INET6;
4144 pmsg->prefix_pad1 = 0;
4145 pmsg->prefix_pad2 = 0;
4146 pmsg->prefix_ifindex = idev->dev->ifindex;
4147 pmsg->prefix_len = pinfo->prefix_len;
4148 pmsg->prefix_type = pinfo->type;
4149 pmsg->prefix_pad3 = 0;
4150 pmsg->prefix_flags = 0;
4151 if (pinfo->onlink)
4152 pmsg->prefix_flags |= IF_PREFIX_ONLINK;
4153 if (pinfo->autoconf)
4154 pmsg->prefix_flags |= IF_PREFIX_AUTOCONF;
4156 if (nla_put(skb, PREFIX_ADDRESS, sizeof(pinfo->prefix), &pinfo->prefix))
4157 goto nla_put_failure;
4158 ci.preferred_time = ntohl(pinfo->prefered);
4159 ci.valid_time = ntohl(pinfo->valid);
4160 if (nla_put(skb, PREFIX_CACHEINFO, sizeof(ci), &ci))
4161 goto nla_put_failure;
4162 return nlmsg_end(skb, nlh);
4164 nla_put_failure:
4165 nlmsg_cancel(skb, nlh);
4166 return -EMSGSIZE;
4169 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
4170 struct prefix_info *pinfo)
4172 struct sk_buff *skb;
4173 struct net *net = dev_net(idev->dev);
4174 int err = -ENOBUFS;
4176 skb = nlmsg_new(inet6_prefix_nlmsg_size(), GFP_ATOMIC);
4177 if (skb == NULL)
4178 goto errout;
4180 err = inet6_fill_prefix(skb, idev, pinfo, 0, 0, event, 0);
4181 if (err < 0) {
4182 /* -EMSGSIZE implies BUG in inet6_prefix_nlmsg_size() */
4183 WARN_ON(err == -EMSGSIZE);
4184 kfree_skb(skb);
4185 goto errout;
4187 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_PREFIX, NULL, GFP_ATOMIC);
4188 return;
4189 errout:
4190 if (err < 0)
4191 rtnl_set_sk_err(net, RTNLGRP_IPV6_PREFIX, err);
4194 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
4196 inet6_ifa_notify(event ? : RTM_NEWADDR, ifp);
4198 switch (event) {
4199 case RTM_NEWADDR:
4201 * If the address was optimistic
4202 * we inserted the route at the start of
4203 * our DAD process, so we don't need
4204 * to do it again
4206 if (!(ifp->rt->rt6i_node))
4207 ip6_ins_rt(ifp->rt);
4208 if (ifp->idev->cnf.forwarding)
4209 addrconf_join_anycast(ifp);
4210 break;
4211 case RTM_DELADDR:
4212 if (ifp->idev->cnf.forwarding)
4213 addrconf_leave_anycast(ifp);
4214 addrconf_leave_solict(ifp->idev, &ifp->addr);
4215 dst_hold(&ifp->rt->dst);
4217 if (ip6_del_rt(ifp->rt))
4218 dst_free(&ifp->rt->dst);
4219 break;
4223 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
4225 rcu_read_lock_bh();
4226 if (likely(ifp->idev->dead == 0))
4227 __ipv6_ifa_notify(event, ifp);
4228 rcu_read_unlock_bh();
4231 #ifdef CONFIG_SYSCTL
4233 static
4234 int addrconf_sysctl_forward(ctl_table *ctl, int write,
4235 void __user *buffer, size_t *lenp, loff_t *ppos)
4237 int *valp = ctl->data;
4238 int val = *valp;
4239 loff_t pos = *ppos;
4240 ctl_table lctl;
4241 int ret;
4244 * ctl->data points to idev->cnf.forwarding, we should
4245 * not modify it until we get the rtnl lock.
4247 lctl = *ctl;
4248 lctl.data = &val;
4250 ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
4252 if (write)
4253 ret = addrconf_fixup_forwarding(ctl, valp, val);
4254 if (ret)
4255 *ppos = pos;
4256 return ret;
4259 static void dev_disable_change(struct inet6_dev *idev)
4261 if (!idev || !idev->dev)
4262 return;
4264 if (idev->cnf.disable_ipv6)
4265 addrconf_notify(NULL, NETDEV_DOWN, idev->dev);
4266 else
4267 addrconf_notify(NULL, NETDEV_UP, idev->dev);
4270 static void addrconf_disable_change(struct net *net, __s32 newf)
4272 struct net_device *dev;
4273 struct inet6_dev *idev;
4275 rcu_read_lock();
4276 for_each_netdev_rcu(net, dev) {
4277 idev = __in6_dev_get(dev);
4278 if (idev) {
4279 int changed = (!idev->cnf.disable_ipv6) ^ (!newf);
4280 idev->cnf.disable_ipv6 = newf;
4281 if (changed)
4282 dev_disable_change(idev);
4285 rcu_read_unlock();
4288 static int addrconf_disable_ipv6(struct ctl_table *table, int *p, int newf)
4290 struct net *net;
4291 int old;
4293 if (!rtnl_trylock())
4294 return restart_syscall();
4296 net = (struct net *)table->extra2;
4297 old = *p;
4298 *p = newf;
4300 if (p == &net->ipv6.devconf_dflt->disable_ipv6) {
4301 rtnl_unlock();
4302 return 0;
4305 if (p == &net->ipv6.devconf_all->disable_ipv6) {
4306 net->ipv6.devconf_dflt->disable_ipv6 = newf;
4307 addrconf_disable_change(net, newf);
4308 } else if ((!newf) ^ (!old))
4309 dev_disable_change((struct inet6_dev *)table->extra1);
4311 rtnl_unlock();
4312 return 0;
4315 static
4316 int addrconf_sysctl_disable(ctl_table *ctl, int write,
4317 void __user *buffer, size_t *lenp, loff_t *ppos)
4319 int *valp = ctl->data;
4320 int val = *valp;
4321 loff_t pos = *ppos;
4322 ctl_table lctl;
4323 int ret;
4326 * ctl->data points to idev->cnf.disable_ipv6, we should
4327 * not modify it until we get the rtnl lock.
4329 lctl = *ctl;
4330 lctl.data = &val;
4332 ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
4334 if (write)
4335 ret = addrconf_disable_ipv6(ctl, valp, val);
4336 if (ret)
4337 *ppos = pos;
4338 return ret;
4341 static struct addrconf_sysctl_table
4343 struct ctl_table_header *sysctl_header;
4344 ctl_table addrconf_vars[DEVCONF_MAX+1];
4345 } addrconf_sysctl __read_mostly = {
4346 .sysctl_header = NULL,
4347 .addrconf_vars = {
4349 .procname = "forwarding",
4350 .data = &ipv6_devconf.forwarding,
4351 .maxlen = sizeof(int),
4352 .mode = 0644,
4353 .proc_handler = addrconf_sysctl_forward,
4356 .procname = "hop_limit",
4357 .data = &ipv6_devconf.hop_limit,
4358 .maxlen = sizeof(int),
4359 .mode = 0644,
4360 .proc_handler = proc_dointvec,
4363 .procname = "mtu",
4364 .data = &ipv6_devconf.mtu6,
4365 .maxlen = sizeof(int),
4366 .mode = 0644,
4367 .proc_handler = proc_dointvec,
4370 .procname = "accept_ra",
4371 .data = &ipv6_devconf.accept_ra,
4372 .maxlen = sizeof(int),
4373 .mode = 0644,
4374 .proc_handler = proc_dointvec,
4377 .procname = "accept_redirects",
4378 .data = &ipv6_devconf.accept_redirects,
4379 .maxlen = sizeof(int),
4380 .mode = 0644,
4381 .proc_handler = proc_dointvec,
4384 .procname = "autoconf",
4385 .data = &ipv6_devconf.autoconf,
4386 .maxlen = sizeof(int),
4387 .mode = 0644,
4388 .proc_handler = proc_dointvec,
4391 .procname = "dad_transmits",
4392 .data = &ipv6_devconf.dad_transmits,
4393 .maxlen = sizeof(int),
4394 .mode = 0644,
4395 .proc_handler = proc_dointvec,
4398 .procname = "router_solicitations",
4399 .data = &ipv6_devconf.rtr_solicits,
4400 .maxlen = sizeof(int),
4401 .mode = 0644,
4402 .proc_handler = proc_dointvec,
4405 .procname = "router_solicitation_interval",
4406 .data = &ipv6_devconf.rtr_solicit_interval,
4407 .maxlen = sizeof(int),
4408 .mode = 0644,
4409 .proc_handler = proc_dointvec_jiffies,
4412 .procname = "router_solicitation_delay",
4413 .data = &ipv6_devconf.rtr_solicit_delay,
4414 .maxlen = sizeof(int),
4415 .mode = 0644,
4416 .proc_handler = proc_dointvec_jiffies,
4419 .procname = "force_mld_version",
4420 .data = &ipv6_devconf.force_mld_version,
4421 .maxlen = sizeof(int),
4422 .mode = 0644,
4423 .proc_handler = proc_dointvec,
4425 #ifdef CONFIG_IPV6_PRIVACY
4427 .procname = "use_tempaddr",
4428 .data = &ipv6_devconf.use_tempaddr,
4429 .maxlen = sizeof(int),
4430 .mode = 0644,
4431 .proc_handler = proc_dointvec,
4434 .procname = "temp_valid_lft",
4435 .data = &ipv6_devconf.temp_valid_lft,
4436 .maxlen = sizeof(int),
4437 .mode = 0644,
4438 .proc_handler = proc_dointvec,
4441 .procname = "temp_prefered_lft",
4442 .data = &ipv6_devconf.temp_prefered_lft,
4443 .maxlen = sizeof(int),
4444 .mode = 0644,
4445 .proc_handler = proc_dointvec,
4448 .procname = "regen_max_retry",
4449 .data = &ipv6_devconf.regen_max_retry,
4450 .maxlen = sizeof(int),
4451 .mode = 0644,
4452 .proc_handler = proc_dointvec,
4455 .procname = "max_desync_factor",
4456 .data = &ipv6_devconf.max_desync_factor,
4457 .maxlen = sizeof(int),
4458 .mode = 0644,
4459 .proc_handler = proc_dointvec,
4461 #endif
4463 .procname = "max_addresses",
4464 .data = &ipv6_devconf.max_addresses,
4465 .maxlen = sizeof(int),
4466 .mode = 0644,
4467 .proc_handler = proc_dointvec,
4470 .procname = "accept_ra_defrtr",
4471 .data = &ipv6_devconf.accept_ra_defrtr,
4472 .maxlen = sizeof(int),
4473 .mode = 0644,
4474 .proc_handler = proc_dointvec,
4477 .procname = "accept_ra_pinfo",
4478 .data = &ipv6_devconf.accept_ra_pinfo,
4479 .maxlen = sizeof(int),
4480 .mode = 0644,
4481 .proc_handler = proc_dointvec,
4483 #ifdef CONFIG_IPV6_ROUTER_PREF
4485 .procname = "accept_ra_rtr_pref",
4486 .data = &ipv6_devconf.accept_ra_rtr_pref,
4487 .maxlen = sizeof(int),
4488 .mode = 0644,
4489 .proc_handler = proc_dointvec,
4492 .procname = "router_probe_interval",
4493 .data = &ipv6_devconf.rtr_probe_interval,
4494 .maxlen = sizeof(int),
4495 .mode = 0644,
4496 .proc_handler = proc_dointvec_jiffies,
4498 #ifdef CONFIG_IPV6_ROUTE_INFO
4500 .procname = "accept_ra_rt_info_max_plen",
4501 .data = &ipv6_devconf.accept_ra_rt_info_max_plen,
4502 .maxlen = sizeof(int),
4503 .mode = 0644,
4504 .proc_handler = proc_dointvec,
4506 #endif
4507 #endif
4509 .procname = "proxy_ndp",
4510 .data = &ipv6_devconf.proxy_ndp,
4511 .maxlen = sizeof(int),
4512 .mode = 0644,
4513 .proc_handler = proc_dointvec,
4516 .procname = "accept_source_route",
4517 .data = &ipv6_devconf.accept_source_route,
4518 .maxlen = sizeof(int),
4519 .mode = 0644,
4520 .proc_handler = proc_dointvec,
4522 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
4524 .procname = "optimistic_dad",
4525 .data = &ipv6_devconf.optimistic_dad,
4526 .maxlen = sizeof(int),
4527 .mode = 0644,
4528 .proc_handler = proc_dointvec,
4531 #endif
4532 #ifdef CONFIG_IPV6_MROUTE
4534 .procname = "mc_forwarding",
4535 .data = &ipv6_devconf.mc_forwarding,
4536 .maxlen = sizeof(int),
4537 .mode = 0444,
4538 .proc_handler = proc_dointvec,
4540 #endif
4542 .procname = "disable_ipv6",
4543 .data = &ipv6_devconf.disable_ipv6,
4544 .maxlen = sizeof(int),
4545 .mode = 0644,
4546 .proc_handler = addrconf_sysctl_disable,
4549 .procname = "accept_dad",
4550 .data = &ipv6_devconf.accept_dad,
4551 .maxlen = sizeof(int),
4552 .mode = 0644,
4553 .proc_handler = proc_dointvec,
4556 .procname = "force_tllao",
4557 .data = &ipv6_devconf.force_tllao,
4558 .maxlen = sizeof(int),
4559 .mode = 0644,
4560 .proc_handler = proc_dointvec
4563 /* sentinel */
4568 static int __addrconf_sysctl_register(struct net *net, char *dev_name,
4569 struct inet6_dev *idev, struct ipv6_devconf *p)
4571 int i;
4572 struct addrconf_sysctl_table *t;
4573 char path[sizeof("net/ipv6/conf/") + IFNAMSIZ];
4575 t = kmemdup(&addrconf_sysctl, sizeof(*t), GFP_KERNEL);
4576 if (t == NULL)
4577 goto out;
4579 for (i = 0; t->addrconf_vars[i].data; i++) {
4580 t->addrconf_vars[i].data += (char *)p - (char *)&ipv6_devconf;
4581 t->addrconf_vars[i].extra1 = idev; /* embedded; no ref */
4582 t->addrconf_vars[i].extra2 = net;
4585 snprintf(path, sizeof(path), "net/ipv6/conf/%s", dev_name);
4587 t->sysctl_header = register_net_sysctl(net, path, t->addrconf_vars);
4588 if (t->sysctl_header == NULL)
4589 goto free;
4591 p->sysctl = t;
4592 return 0;
4594 free:
4595 kfree(t);
4596 out:
4597 return -ENOBUFS;
4600 static void __addrconf_sysctl_unregister(struct ipv6_devconf *p)
4602 struct addrconf_sysctl_table *t;
4604 if (p->sysctl == NULL)
4605 return;
4607 t = p->sysctl;
4608 p->sysctl = NULL;
4609 unregister_net_sysctl_table(t->sysctl_header);
4610 kfree(t);
4613 static void addrconf_sysctl_register(struct inet6_dev *idev)
4615 neigh_sysctl_register(idev->dev, idev->nd_parms, "ipv6",
4616 &ndisc_ifinfo_sysctl_change);
4617 __addrconf_sysctl_register(dev_net(idev->dev), idev->dev->name,
4618 idev, &idev->cnf);
4621 static void addrconf_sysctl_unregister(struct inet6_dev *idev)
4623 __addrconf_sysctl_unregister(&idev->cnf);
4624 neigh_sysctl_unregister(idev->nd_parms);
4628 #endif
4630 static int __net_init addrconf_init_net(struct net *net)
4632 int err;
4633 struct ipv6_devconf *all, *dflt;
4635 err = -ENOMEM;
4636 all = &ipv6_devconf;
4637 dflt = &ipv6_devconf_dflt;
4639 if (!net_eq(net, &init_net)) {
4640 all = kmemdup(all, sizeof(ipv6_devconf), GFP_KERNEL);
4641 if (all == NULL)
4642 goto err_alloc_all;
4644 dflt = kmemdup(dflt, sizeof(ipv6_devconf_dflt), GFP_KERNEL);
4645 if (dflt == NULL)
4646 goto err_alloc_dflt;
4647 } else {
4648 /* these will be inherited by all namespaces */
4649 dflt->autoconf = ipv6_defaults.autoconf;
4650 dflt->disable_ipv6 = ipv6_defaults.disable_ipv6;
4653 net->ipv6.devconf_all = all;
4654 net->ipv6.devconf_dflt = dflt;
4656 #ifdef CONFIG_SYSCTL
4657 err = __addrconf_sysctl_register(net, "all", NULL, all);
4658 if (err < 0)
4659 goto err_reg_all;
4661 err = __addrconf_sysctl_register(net, "default", NULL, dflt);
4662 if (err < 0)
4663 goto err_reg_dflt;
4664 #endif
4665 return 0;
4667 #ifdef CONFIG_SYSCTL
4668 err_reg_dflt:
4669 __addrconf_sysctl_unregister(all);
4670 err_reg_all:
4671 kfree(dflt);
4672 #endif
4673 err_alloc_dflt:
4674 kfree(all);
4675 err_alloc_all:
4676 return err;
4679 static void __net_exit addrconf_exit_net(struct net *net)
4681 #ifdef CONFIG_SYSCTL
4682 __addrconf_sysctl_unregister(net->ipv6.devconf_dflt);
4683 __addrconf_sysctl_unregister(net->ipv6.devconf_all);
4684 #endif
4685 if (!net_eq(net, &init_net)) {
4686 kfree(net->ipv6.devconf_dflt);
4687 kfree(net->ipv6.devconf_all);
4691 static struct pernet_operations addrconf_ops = {
4692 .init = addrconf_init_net,
4693 .exit = addrconf_exit_net,
4697 * Device notifier
4700 int register_inet6addr_notifier(struct notifier_block *nb)
4702 return atomic_notifier_chain_register(&inet6addr_chain, nb);
4704 EXPORT_SYMBOL(register_inet6addr_notifier);
4706 int unregister_inet6addr_notifier(struct notifier_block *nb)
4708 return atomic_notifier_chain_unregister(&inet6addr_chain, nb);
4710 EXPORT_SYMBOL(unregister_inet6addr_notifier);
4712 static struct rtnl_af_ops inet6_ops = {
4713 .family = AF_INET6,
4714 .fill_link_af = inet6_fill_link_af,
4715 .get_link_af_size = inet6_get_link_af_size,
4719 * Init / cleanup code
4722 int __init addrconf_init(void)
4724 int i, err;
4726 err = ipv6_addr_label_init();
4727 if (err < 0) {
4728 pr_crit("%s: cannot initialize default policy table: %d\n",
4729 __func__, err);
4730 goto out;
4733 err = register_pernet_subsys(&addrconf_ops);
4734 if (err < 0)
4735 goto out_addrlabel;
4737 /* The addrconf netdev notifier requires that loopback_dev
4738 * has it's ipv6 private information allocated and setup
4739 * before it can bring up and give link-local addresses
4740 * to other devices which are up.
4742 * Unfortunately, loopback_dev is not necessarily the first
4743 * entry in the global dev_base list of net devices. In fact,
4744 * it is likely to be the very last entry on that list.
4745 * So this causes the notifier registry below to try and
4746 * give link-local addresses to all devices besides loopback_dev
4747 * first, then loopback_dev, which cases all the non-loopback_dev
4748 * devices to fail to get a link-local address.
4750 * So, as a temporary fix, allocate the ipv6 structure for
4751 * loopback_dev first by hand.
4752 * Longer term, all of the dependencies ipv6 has upon the loopback
4753 * device and it being up should be removed.
4755 rtnl_lock();
4756 if (!ipv6_add_dev(init_net.loopback_dev))
4757 err = -ENOMEM;
4758 rtnl_unlock();
4759 if (err)
4760 goto errlo;
4762 for (i = 0; i < IN6_ADDR_HSIZE; i++)
4763 INIT_HLIST_HEAD(&inet6_addr_lst[i]);
4765 register_netdevice_notifier(&ipv6_dev_notf);
4767 addrconf_verify(0);
4769 err = rtnl_af_register(&inet6_ops);
4770 if (err < 0)
4771 goto errout_af;
4773 err = __rtnl_register(PF_INET6, RTM_GETLINK, NULL, inet6_dump_ifinfo,
4774 NULL);
4775 if (err < 0)
4776 goto errout;
4778 /* Only the first call to __rtnl_register can fail */
4779 __rtnl_register(PF_INET6, RTM_NEWADDR, inet6_rtm_newaddr, NULL, NULL);
4780 __rtnl_register(PF_INET6, RTM_DELADDR, inet6_rtm_deladdr, NULL, NULL);
4781 __rtnl_register(PF_INET6, RTM_GETADDR, inet6_rtm_getaddr,
4782 inet6_dump_ifaddr, NULL);
4783 __rtnl_register(PF_INET6, RTM_GETMULTICAST, NULL,
4784 inet6_dump_ifmcaddr, NULL);
4785 __rtnl_register(PF_INET6, RTM_GETANYCAST, NULL,
4786 inet6_dump_ifacaddr, NULL);
4788 ipv6_addr_label_rtnl_register();
4790 return 0;
4791 errout:
4792 rtnl_af_unregister(&inet6_ops);
4793 errout_af:
4794 unregister_netdevice_notifier(&ipv6_dev_notf);
4795 errlo:
4796 unregister_pernet_subsys(&addrconf_ops);
4797 out_addrlabel:
4798 ipv6_addr_label_cleanup();
4799 out:
4800 return err;
4803 void addrconf_cleanup(void)
4805 struct net_device *dev;
4806 int i;
4808 unregister_netdevice_notifier(&ipv6_dev_notf);
4809 unregister_pernet_subsys(&addrconf_ops);
4810 ipv6_addr_label_cleanup();
4812 rtnl_lock();
4814 __rtnl_af_unregister(&inet6_ops);
4816 /* clean dev list */
4817 for_each_netdev(&init_net, dev) {
4818 if (__in6_dev_get(dev) == NULL)
4819 continue;
4820 addrconf_ifdown(dev, 1);
4822 addrconf_ifdown(init_net.loopback_dev, 2);
4825 * Check hash table.
4827 spin_lock_bh(&addrconf_hash_lock);
4828 for (i = 0; i < IN6_ADDR_HSIZE; i++)
4829 WARN_ON(!hlist_empty(&inet6_addr_lst[i]));
4830 spin_unlock_bh(&addrconf_hash_lock);
4832 del_timer(&addr_chk_timer);
4833 rtnl_unlock();