drivers/video/backlight/kb3886_bl.c: use usleep_range() instead of msleep() for small...
[linux-2.6/btrfs-unstable.git] / net / ipv6 / addrconf.c
blob480e68422efb3c0f3ff7267ac89b8fe2b0d42fee
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 void addrconf_add_lroute(struct net_device *dev)
1774 struct in6_addr addr;
1776 ipv6_addr_set(&addr, htonl(0xFE800000), 0, 0, 0);
1777 addrconf_prefix_route(&addr, 64, dev, 0, 0);
1780 static struct inet6_dev *addrconf_add_dev(struct net_device *dev)
1782 struct inet6_dev *idev;
1784 ASSERT_RTNL();
1786 idev = ipv6_find_idev(dev);
1787 if (!idev)
1788 return ERR_PTR(-ENOBUFS);
1790 if (idev->cnf.disable_ipv6)
1791 return ERR_PTR(-EACCES);
1793 /* Add default multicast route */
1794 if (!(dev->flags & IFF_LOOPBACK))
1795 addrconf_add_mroute(dev);
1797 /* Add link local route */
1798 addrconf_add_lroute(dev);
1799 return idev;
1802 void addrconf_prefix_rcv(struct net_device *dev, u8 *opt, int len, bool sllao)
1804 struct prefix_info *pinfo;
1805 __u32 valid_lft;
1806 __u32 prefered_lft;
1807 int addr_type;
1808 struct inet6_dev *in6_dev;
1809 struct net *net = dev_net(dev);
1811 pinfo = (struct prefix_info *) opt;
1813 if (len < sizeof(struct prefix_info)) {
1814 ADBG(("addrconf: prefix option too short\n"));
1815 return;
1819 * Validation checks ([ADDRCONF], page 19)
1822 addr_type = ipv6_addr_type(&pinfo->prefix);
1824 if (addr_type & (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL))
1825 return;
1827 valid_lft = ntohl(pinfo->valid);
1828 prefered_lft = ntohl(pinfo->prefered);
1830 if (prefered_lft > valid_lft) {
1831 net_warn_ratelimited("addrconf: prefix option has invalid lifetime\n");
1832 return;
1835 in6_dev = in6_dev_get(dev);
1837 if (in6_dev == NULL) {
1838 net_dbg_ratelimited("addrconf: device %s not configured\n",
1839 dev->name);
1840 return;
1844 * Two things going on here:
1845 * 1) Add routes for on-link prefixes
1846 * 2) Configure prefixes with the auto flag set
1849 if (pinfo->onlink) {
1850 struct rt6_info *rt;
1851 unsigned long rt_expires;
1853 /* Avoid arithmetic overflow. Really, we could
1854 * save rt_expires in seconds, likely valid_lft,
1855 * but it would require division in fib gc, that it
1856 * not good.
1858 if (HZ > USER_HZ)
1859 rt_expires = addrconf_timeout_fixup(valid_lft, HZ);
1860 else
1861 rt_expires = addrconf_timeout_fixup(valid_lft, USER_HZ);
1863 if (addrconf_finite_timeout(rt_expires))
1864 rt_expires *= HZ;
1866 rt = addrconf_get_prefix_route(&pinfo->prefix,
1867 pinfo->prefix_len,
1868 dev,
1869 RTF_ADDRCONF | RTF_PREFIX_RT,
1870 RTF_GATEWAY | RTF_DEFAULT);
1872 if (rt) {
1873 /* Autoconf prefix route */
1874 if (valid_lft == 0) {
1875 ip6_del_rt(rt);
1876 rt = NULL;
1877 } else if (addrconf_finite_timeout(rt_expires)) {
1878 /* not infinity */
1879 rt6_set_expires(rt, jiffies + rt_expires);
1880 } else {
1881 rt6_clean_expires(rt);
1883 } else if (valid_lft) {
1884 clock_t expires = 0;
1885 int flags = RTF_ADDRCONF | RTF_PREFIX_RT;
1886 if (addrconf_finite_timeout(rt_expires)) {
1887 /* not infinity */
1888 flags |= RTF_EXPIRES;
1889 expires = jiffies_to_clock_t(rt_expires);
1891 addrconf_prefix_route(&pinfo->prefix, pinfo->prefix_len,
1892 dev, expires, flags);
1894 if (rt)
1895 dst_release(&rt->dst);
1898 /* Try to figure out our local address for this prefix */
1900 if (pinfo->autoconf && in6_dev->cnf.autoconf) {
1901 struct inet6_ifaddr *ifp;
1902 struct in6_addr addr;
1903 int create = 0, update_lft = 0;
1905 if (pinfo->prefix_len == 64) {
1906 memcpy(&addr, &pinfo->prefix, 8);
1907 if (ipv6_generate_eui64(addr.s6_addr + 8, dev) &&
1908 ipv6_inherit_eui64(addr.s6_addr + 8, in6_dev)) {
1909 in6_dev_put(in6_dev);
1910 return;
1912 goto ok;
1914 net_dbg_ratelimited("IPv6 addrconf: prefix with wrong length %d\n",
1915 pinfo->prefix_len);
1916 in6_dev_put(in6_dev);
1917 return;
1921 ifp = ipv6_get_ifaddr(net, &addr, dev, 1);
1923 if (ifp == NULL && valid_lft) {
1924 int max_addresses = in6_dev->cnf.max_addresses;
1925 u32 addr_flags = 0;
1927 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1928 if (in6_dev->cnf.optimistic_dad &&
1929 !net->ipv6.devconf_all->forwarding && sllao)
1930 addr_flags = IFA_F_OPTIMISTIC;
1931 #endif
1933 /* Do not allow to create too much of autoconfigured
1934 * addresses; this would be too easy way to crash kernel.
1936 if (!max_addresses ||
1937 ipv6_count_addresses(in6_dev) < max_addresses)
1938 ifp = ipv6_add_addr(in6_dev, &addr, pinfo->prefix_len,
1939 addr_type&IPV6_ADDR_SCOPE_MASK,
1940 addr_flags);
1942 if (!ifp || IS_ERR(ifp)) {
1943 in6_dev_put(in6_dev);
1944 return;
1947 update_lft = create = 1;
1948 ifp->cstamp = jiffies;
1949 addrconf_dad_start(ifp);
1952 if (ifp) {
1953 int flags;
1954 unsigned long now;
1955 #ifdef CONFIG_IPV6_PRIVACY
1956 struct inet6_ifaddr *ift;
1957 #endif
1958 u32 stored_lft;
1960 /* update lifetime (RFC2462 5.5.3 e) */
1961 spin_lock(&ifp->lock);
1962 now = jiffies;
1963 if (ifp->valid_lft > (now - ifp->tstamp) / HZ)
1964 stored_lft = ifp->valid_lft - (now - ifp->tstamp) / HZ;
1965 else
1966 stored_lft = 0;
1967 if (!update_lft && stored_lft) {
1968 if (valid_lft > MIN_VALID_LIFETIME ||
1969 valid_lft > stored_lft)
1970 update_lft = 1;
1971 else if (stored_lft <= MIN_VALID_LIFETIME) {
1972 /* valid_lft <= stored_lft is always true */
1974 * RFC 4862 Section 5.5.3e:
1975 * "Note that the preferred lifetime of
1976 * the corresponding address is always
1977 * reset to the Preferred Lifetime in
1978 * the received Prefix Information
1979 * option, regardless of whether the
1980 * valid lifetime is also reset or
1981 * ignored."
1983 * So if the preferred lifetime in
1984 * this advertisement is different
1985 * than what we have stored, but the
1986 * valid lifetime is invalid, just
1987 * reset prefered_lft.
1989 * We must set the valid lifetime
1990 * to the stored lifetime since we'll
1991 * be updating the timestamp below,
1992 * else we'll set it back to the
1993 * minimum.
1995 if (prefered_lft != ifp->prefered_lft) {
1996 valid_lft = stored_lft;
1997 update_lft = 1;
1999 } else {
2000 valid_lft = MIN_VALID_LIFETIME;
2001 if (valid_lft < prefered_lft)
2002 prefered_lft = valid_lft;
2003 update_lft = 1;
2007 if (update_lft) {
2008 ifp->valid_lft = valid_lft;
2009 ifp->prefered_lft = prefered_lft;
2010 ifp->tstamp = now;
2011 flags = ifp->flags;
2012 ifp->flags &= ~IFA_F_DEPRECATED;
2013 spin_unlock(&ifp->lock);
2015 if (!(flags&IFA_F_TENTATIVE))
2016 ipv6_ifa_notify(0, ifp);
2017 } else
2018 spin_unlock(&ifp->lock);
2020 #ifdef CONFIG_IPV6_PRIVACY
2021 read_lock_bh(&in6_dev->lock);
2022 /* update all temporary addresses in the list */
2023 list_for_each_entry(ift, &in6_dev->tempaddr_list,
2024 tmp_list) {
2025 int age, max_valid, max_prefered;
2027 if (ifp != ift->ifpub)
2028 continue;
2031 * RFC 4941 section 3.3:
2032 * If a received option will extend the lifetime
2033 * of a public address, the lifetimes of
2034 * temporary addresses should be extended,
2035 * subject to the overall constraint that no
2036 * temporary addresses should ever remain
2037 * "valid" or "preferred" for a time longer than
2038 * (TEMP_VALID_LIFETIME) or
2039 * (TEMP_PREFERRED_LIFETIME - DESYNC_FACTOR),
2040 * respectively.
2042 age = (now - ift->cstamp) / HZ;
2043 max_valid = in6_dev->cnf.temp_valid_lft - age;
2044 if (max_valid < 0)
2045 max_valid = 0;
2047 max_prefered = in6_dev->cnf.temp_prefered_lft -
2048 in6_dev->cnf.max_desync_factor -
2049 age;
2050 if (max_prefered < 0)
2051 max_prefered = 0;
2053 if (valid_lft > max_valid)
2054 valid_lft = max_valid;
2056 if (prefered_lft > max_prefered)
2057 prefered_lft = max_prefered;
2059 spin_lock(&ift->lock);
2060 flags = ift->flags;
2061 ift->valid_lft = valid_lft;
2062 ift->prefered_lft = prefered_lft;
2063 ift->tstamp = now;
2064 if (prefered_lft > 0)
2065 ift->flags &= ~IFA_F_DEPRECATED;
2067 spin_unlock(&ift->lock);
2068 if (!(flags&IFA_F_TENTATIVE))
2069 ipv6_ifa_notify(0, ift);
2072 if ((create || list_empty(&in6_dev->tempaddr_list)) && in6_dev->cnf.use_tempaddr > 0) {
2074 * When a new public address is created as
2075 * described in [ADDRCONF], also create a new
2076 * temporary address. Also create a temporary
2077 * address if it's enabled but no temporary
2078 * address currently exists.
2080 read_unlock_bh(&in6_dev->lock);
2081 ipv6_create_tempaddr(ifp, NULL);
2082 } else {
2083 read_unlock_bh(&in6_dev->lock);
2085 #endif
2086 in6_ifa_put(ifp);
2087 addrconf_verify(0);
2090 inet6_prefix_notify(RTM_NEWPREFIX, in6_dev, pinfo);
2091 in6_dev_put(in6_dev);
2095 * Set destination address.
2096 * Special case for SIT interfaces where we create a new "virtual"
2097 * device.
2099 int addrconf_set_dstaddr(struct net *net, void __user *arg)
2101 struct in6_ifreq ireq;
2102 struct net_device *dev;
2103 int err = -EINVAL;
2105 rtnl_lock();
2107 err = -EFAULT;
2108 if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2109 goto err_exit;
2111 dev = __dev_get_by_index(net, ireq.ifr6_ifindex);
2113 err = -ENODEV;
2114 if (dev == NULL)
2115 goto err_exit;
2117 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2118 if (dev->type == ARPHRD_SIT) {
2119 const struct net_device_ops *ops = dev->netdev_ops;
2120 struct ifreq ifr;
2121 struct ip_tunnel_parm p;
2123 err = -EADDRNOTAVAIL;
2124 if (!(ipv6_addr_type(&ireq.ifr6_addr) & IPV6_ADDR_COMPATv4))
2125 goto err_exit;
2127 memset(&p, 0, sizeof(p));
2128 p.iph.daddr = ireq.ifr6_addr.s6_addr32[3];
2129 p.iph.saddr = 0;
2130 p.iph.version = 4;
2131 p.iph.ihl = 5;
2132 p.iph.protocol = IPPROTO_IPV6;
2133 p.iph.ttl = 64;
2134 ifr.ifr_ifru.ifru_data = (__force void __user *)&p;
2136 if (ops->ndo_do_ioctl) {
2137 mm_segment_t oldfs = get_fs();
2139 set_fs(KERNEL_DS);
2140 err = ops->ndo_do_ioctl(dev, &ifr, SIOCADDTUNNEL);
2141 set_fs(oldfs);
2142 } else
2143 err = -EOPNOTSUPP;
2145 if (err == 0) {
2146 err = -ENOBUFS;
2147 dev = __dev_get_by_name(net, p.name);
2148 if (!dev)
2149 goto err_exit;
2150 err = dev_open(dev);
2153 #endif
2155 err_exit:
2156 rtnl_unlock();
2157 return err;
2161 * Manual configuration of address on an interface
2163 static int inet6_addr_add(struct net *net, int ifindex, const struct in6_addr *pfx,
2164 unsigned int plen, __u8 ifa_flags, __u32 prefered_lft,
2165 __u32 valid_lft)
2167 struct inet6_ifaddr *ifp;
2168 struct inet6_dev *idev;
2169 struct net_device *dev;
2170 int scope;
2171 u32 flags;
2172 clock_t expires;
2173 unsigned long timeout;
2175 ASSERT_RTNL();
2177 if (plen > 128)
2178 return -EINVAL;
2180 /* check the lifetime */
2181 if (!valid_lft || prefered_lft > valid_lft)
2182 return -EINVAL;
2184 dev = __dev_get_by_index(net, ifindex);
2185 if (!dev)
2186 return -ENODEV;
2188 idev = addrconf_add_dev(dev);
2189 if (IS_ERR(idev))
2190 return PTR_ERR(idev);
2192 scope = ipv6_addr_scope(pfx);
2194 timeout = addrconf_timeout_fixup(valid_lft, HZ);
2195 if (addrconf_finite_timeout(timeout)) {
2196 expires = jiffies_to_clock_t(timeout * HZ);
2197 valid_lft = timeout;
2198 flags = RTF_EXPIRES;
2199 } else {
2200 expires = 0;
2201 flags = 0;
2202 ifa_flags |= IFA_F_PERMANENT;
2205 timeout = addrconf_timeout_fixup(prefered_lft, HZ);
2206 if (addrconf_finite_timeout(timeout)) {
2207 if (timeout == 0)
2208 ifa_flags |= IFA_F_DEPRECATED;
2209 prefered_lft = timeout;
2212 ifp = ipv6_add_addr(idev, pfx, plen, scope, ifa_flags);
2214 if (!IS_ERR(ifp)) {
2215 spin_lock_bh(&ifp->lock);
2216 ifp->valid_lft = valid_lft;
2217 ifp->prefered_lft = prefered_lft;
2218 ifp->tstamp = jiffies;
2219 spin_unlock_bh(&ifp->lock);
2221 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, dev,
2222 expires, flags);
2224 * Note that section 3.1 of RFC 4429 indicates
2225 * that the Optimistic flag should not be set for
2226 * manually configured addresses
2228 addrconf_dad_start(ifp);
2229 in6_ifa_put(ifp);
2230 addrconf_verify(0);
2231 return 0;
2234 return PTR_ERR(ifp);
2237 static int inet6_addr_del(struct net *net, int ifindex, const struct in6_addr *pfx,
2238 unsigned int plen)
2240 struct inet6_ifaddr *ifp;
2241 struct inet6_dev *idev;
2242 struct net_device *dev;
2244 if (plen > 128)
2245 return -EINVAL;
2247 dev = __dev_get_by_index(net, ifindex);
2248 if (!dev)
2249 return -ENODEV;
2251 if ((idev = __in6_dev_get(dev)) == NULL)
2252 return -ENXIO;
2254 read_lock_bh(&idev->lock);
2255 list_for_each_entry(ifp, &idev->addr_list, if_list) {
2256 if (ifp->prefix_len == plen &&
2257 ipv6_addr_equal(pfx, &ifp->addr)) {
2258 in6_ifa_hold(ifp);
2259 read_unlock_bh(&idev->lock);
2261 ipv6_del_addr(ifp);
2263 /* If the last address is deleted administratively,
2264 disable IPv6 on this interface.
2266 if (list_empty(&idev->addr_list))
2267 addrconf_ifdown(idev->dev, 1);
2268 return 0;
2271 read_unlock_bh(&idev->lock);
2272 return -EADDRNOTAVAIL;
2276 int addrconf_add_ifaddr(struct net *net, void __user *arg)
2278 struct in6_ifreq ireq;
2279 int err;
2281 if (!capable(CAP_NET_ADMIN))
2282 return -EPERM;
2284 if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2285 return -EFAULT;
2287 rtnl_lock();
2288 err = inet6_addr_add(net, ireq.ifr6_ifindex, &ireq.ifr6_addr,
2289 ireq.ifr6_prefixlen, IFA_F_PERMANENT,
2290 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
2291 rtnl_unlock();
2292 return err;
2295 int addrconf_del_ifaddr(struct net *net, void __user *arg)
2297 struct in6_ifreq ireq;
2298 int err;
2300 if (!capable(CAP_NET_ADMIN))
2301 return -EPERM;
2303 if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2304 return -EFAULT;
2306 rtnl_lock();
2307 err = inet6_addr_del(net, ireq.ifr6_ifindex, &ireq.ifr6_addr,
2308 ireq.ifr6_prefixlen);
2309 rtnl_unlock();
2310 return err;
2313 static void add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
2314 int plen, int scope)
2316 struct inet6_ifaddr *ifp;
2318 ifp = ipv6_add_addr(idev, addr, plen, scope, IFA_F_PERMANENT);
2319 if (!IS_ERR(ifp)) {
2320 spin_lock_bh(&ifp->lock);
2321 ifp->flags &= ~IFA_F_TENTATIVE;
2322 spin_unlock_bh(&ifp->lock);
2323 ipv6_ifa_notify(RTM_NEWADDR, ifp);
2324 in6_ifa_put(ifp);
2328 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2329 static void sit_add_v4_addrs(struct inet6_dev *idev)
2331 struct in6_addr addr;
2332 struct net_device *dev;
2333 struct net *net = dev_net(idev->dev);
2334 int scope;
2336 ASSERT_RTNL();
2338 memset(&addr, 0, sizeof(struct in6_addr));
2339 memcpy(&addr.s6_addr32[3], idev->dev->dev_addr, 4);
2341 if (idev->dev->flags&IFF_POINTOPOINT) {
2342 addr.s6_addr32[0] = htonl(0xfe800000);
2343 scope = IFA_LINK;
2344 } else {
2345 scope = IPV6_ADDR_COMPATv4;
2348 if (addr.s6_addr32[3]) {
2349 add_addr(idev, &addr, 128, scope);
2350 return;
2353 for_each_netdev(net, dev) {
2354 struct in_device *in_dev = __in_dev_get_rtnl(dev);
2355 if (in_dev && (dev->flags & IFF_UP)) {
2356 struct in_ifaddr *ifa;
2358 int flag = scope;
2360 for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) {
2361 int plen;
2363 addr.s6_addr32[3] = ifa->ifa_local;
2365 if (ifa->ifa_scope == RT_SCOPE_LINK)
2366 continue;
2367 if (ifa->ifa_scope >= RT_SCOPE_HOST) {
2368 if (idev->dev->flags&IFF_POINTOPOINT)
2369 continue;
2370 flag |= IFA_HOST;
2372 if (idev->dev->flags&IFF_POINTOPOINT)
2373 plen = 64;
2374 else
2375 plen = 96;
2377 add_addr(idev, &addr, plen, flag);
2382 #endif
2384 static void init_loopback(struct net_device *dev)
2386 struct inet6_dev *idev;
2388 /* ::1 */
2390 ASSERT_RTNL();
2392 if ((idev = ipv6_find_idev(dev)) == NULL) {
2393 pr_debug("%s: add_dev failed\n", __func__);
2394 return;
2397 add_addr(idev, &in6addr_loopback, 128, IFA_HOST);
2400 static void addrconf_add_linklocal(struct inet6_dev *idev, const struct in6_addr *addr)
2402 struct inet6_ifaddr *ifp;
2403 u32 addr_flags = IFA_F_PERMANENT;
2405 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2406 if (idev->cnf.optimistic_dad &&
2407 !dev_net(idev->dev)->ipv6.devconf_all->forwarding)
2408 addr_flags |= IFA_F_OPTIMISTIC;
2409 #endif
2412 ifp = ipv6_add_addr(idev, addr, 64, IFA_LINK, addr_flags);
2413 if (!IS_ERR(ifp)) {
2414 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, idev->dev, 0, 0);
2415 addrconf_dad_start(ifp);
2416 in6_ifa_put(ifp);
2420 static void addrconf_dev_config(struct net_device *dev)
2422 struct in6_addr addr;
2423 struct inet6_dev *idev;
2425 ASSERT_RTNL();
2427 if ((dev->type != ARPHRD_ETHER) &&
2428 (dev->type != ARPHRD_FDDI) &&
2429 (dev->type != ARPHRD_ARCNET) &&
2430 (dev->type != ARPHRD_INFINIBAND) &&
2431 (dev->type != ARPHRD_IEEE802154)) {
2432 /* Alas, we support only Ethernet autoconfiguration. */
2433 return;
2436 idev = addrconf_add_dev(dev);
2437 if (IS_ERR(idev))
2438 return;
2440 memset(&addr, 0, sizeof(struct in6_addr));
2441 addr.s6_addr32[0] = htonl(0xFE800000);
2443 if (ipv6_generate_eui64(addr.s6_addr + 8, dev) == 0)
2444 addrconf_add_linklocal(idev, &addr);
2447 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2448 static void addrconf_sit_config(struct net_device *dev)
2450 struct inet6_dev *idev;
2452 ASSERT_RTNL();
2455 * Configure the tunnel with one of our IPv4
2456 * addresses... we should configure all of
2457 * our v4 addrs in the tunnel
2460 if ((idev = ipv6_find_idev(dev)) == NULL) {
2461 pr_debug("%s: add_dev failed\n", __func__);
2462 return;
2465 if (dev->priv_flags & IFF_ISATAP) {
2466 struct in6_addr addr;
2468 ipv6_addr_set(&addr, htonl(0xFE800000), 0, 0, 0);
2469 addrconf_prefix_route(&addr, 64, dev, 0, 0);
2470 if (!ipv6_generate_eui64(addr.s6_addr + 8, dev))
2471 addrconf_add_linklocal(idev, &addr);
2472 return;
2475 sit_add_v4_addrs(idev);
2477 if (dev->flags&IFF_POINTOPOINT) {
2478 addrconf_add_mroute(dev);
2479 addrconf_add_lroute(dev);
2480 } else
2481 sit_route_add(dev);
2483 #endif
2485 #if defined(CONFIG_NET_IPGRE) || defined(CONFIG_NET_IPGRE_MODULE)
2486 static void addrconf_gre_config(struct net_device *dev)
2488 struct inet6_dev *idev;
2489 struct in6_addr addr;
2491 pr_info("%s(%s)\n", __func__, dev->name);
2493 ASSERT_RTNL();
2495 if ((idev = ipv6_find_idev(dev)) == NULL) {
2496 pr_debug("%s: add_dev failed\n", __func__);
2497 return;
2500 ipv6_addr_set(&addr, htonl(0xFE800000), 0, 0, 0);
2501 addrconf_prefix_route(&addr, 64, dev, 0, 0);
2503 if (!ipv6_generate_eui64(addr.s6_addr + 8, dev))
2504 addrconf_add_linklocal(idev, &addr);
2506 #endif
2508 static inline int
2509 ipv6_inherit_linklocal(struct inet6_dev *idev, struct net_device *link_dev)
2511 struct in6_addr lladdr;
2513 if (!ipv6_get_lladdr(link_dev, &lladdr, IFA_F_TENTATIVE)) {
2514 addrconf_add_linklocal(idev, &lladdr);
2515 return 0;
2517 return -1;
2520 static void ip6_tnl_add_linklocal(struct inet6_dev *idev)
2522 struct net_device *link_dev;
2523 struct net *net = dev_net(idev->dev);
2525 /* first try to inherit the link-local address from the link device */
2526 if (idev->dev->iflink &&
2527 (link_dev = __dev_get_by_index(net, idev->dev->iflink))) {
2528 if (!ipv6_inherit_linklocal(idev, link_dev))
2529 return;
2531 /* then try to inherit it from any device */
2532 for_each_netdev(net, link_dev) {
2533 if (!ipv6_inherit_linklocal(idev, link_dev))
2534 return;
2536 pr_debug("init ip6-ip6: add_linklocal failed\n");
2540 * Autoconfigure tunnel with a link-local address so routing protocols,
2541 * DHCPv6, MLD etc. can be run over the virtual link
2544 static void addrconf_ip6_tnl_config(struct net_device *dev)
2546 struct inet6_dev *idev;
2548 ASSERT_RTNL();
2550 idev = addrconf_add_dev(dev);
2551 if (IS_ERR(idev)) {
2552 pr_debug("init ip6-ip6: add_dev failed\n");
2553 return;
2555 ip6_tnl_add_linklocal(idev);
2558 static int addrconf_notify(struct notifier_block *this, unsigned long event,
2559 void *data)
2561 struct net_device *dev = (struct net_device *) data;
2562 struct inet6_dev *idev = __in6_dev_get(dev);
2563 int run_pending = 0;
2564 int err;
2566 switch (event) {
2567 case NETDEV_REGISTER:
2568 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2569 idev = ipv6_add_dev(dev);
2570 if (!idev)
2571 return notifier_from_errno(-ENOMEM);
2573 break;
2575 case NETDEV_UP:
2576 case NETDEV_CHANGE:
2577 if (dev->flags & IFF_SLAVE)
2578 break;
2580 if (event == NETDEV_UP) {
2581 if (!addrconf_qdisc_ok(dev)) {
2582 /* device is not ready yet. */
2583 pr_info("ADDRCONF(NETDEV_UP): %s: link is not ready\n",
2584 dev->name);
2585 break;
2588 if (!idev && dev->mtu >= IPV6_MIN_MTU)
2589 idev = ipv6_add_dev(dev);
2591 if (idev) {
2592 idev->if_flags |= IF_READY;
2593 run_pending = 1;
2595 } else {
2596 if (!addrconf_qdisc_ok(dev)) {
2597 /* device is still not ready. */
2598 break;
2601 if (idev) {
2602 if (idev->if_flags & IF_READY)
2603 /* device is already configured. */
2604 break;
2605 idev->if_flags |= IF_READY;
2608 pr_info("ADDRCONF(NETDEV_CHANGE): %s: link becomes ready\n",
2609 dev->name);
2611 run_pending = 1;
2614 switch (dev->type) {
2615 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2616 case ARPHRD_SIT:
2617 addrconf_sit_config(dev);
2618 break;
2619 #endif
2620 #if defined(CONFIG_NET_IPGRE) || defined(CONFIG_NET_IPGRE_MODULE)
2621 case ARPHRD_IPGRE:
2622 addrconf_gre_config(dev);
2623 break;
2624 #endif
2625 case ARPHRD_TUNNEL6:
2626 addrconf_ip6_tnl_config(dev);
2627 break;
2628 case ARPHRD_LOOPBACK:
2629 init_loopback(dev);
2630 break;
2632 default:
2633 addrconf_dev_config(dev);
2634 break;
2637 if (idev) {
2638 if (run_pending)
2639 addrconf_dad_run(idev);
2642 * If the MTU changed during the interface down,
2643 * when the interface up, the changed MTU must be
2644 * reflected in the idev as well as routers.
2646 if (idev->cnf.mtu6 != dev->mtu &&
2647 dev->mtu >= IPV6_MIN_MTU) {
2648 rt6_mtu_change(dev, dev->mtu);
2649 idev->cnf.mtu6 = dev->mtu;
2651 idev->tstamp = jiffies;
2652 inet6_ifinfo_notify(RTM_NEWLINK, idev);
2655 * If the changed mtu during down is lower than
2656 * IPV6_MIN_MTU stop IPv6 on this interface.
2658 if (dev->mtu < IPV6_MIN_MTU)
2659 addrconf_ifdown(dev, 1);
2661 break;
2663 case NETDEV_CHANGEMTU:
2664 if (idev && dev->mtu >= IPV6_MIN_MTU) {
2665 rt6_mtu_change(dev, dev->mtu);
2666 idev->cnf.mtu6 = dev->mtu;
2667 break;
2670 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2671 idev = ipv6_add_dev(dev);
2672 if (idev)
2673 break;
2677 * MTU falled under IPV6_MIN_MTU.
2678 * Stop IPv6 on this interface.
2681 case NETDEV_DOWN:
2682 case NETDEV_UNREGISTER:
2684 * Remove all addresses from this interface.
2686 addrconf_ifdown(dev, event != NETDEV_DOWN);
2687 break;
2689 case NETDEV_CHANGENAME:
2690 if (idev) {
2691 snmp6_unregister_dev(idev);
2692 addrconf_sysctl_unregister(idev);
2693 addrconf_sysctl_register(idev);
2694 err = snmp6_register_dev(idev);
2695 if (err)
2696 return notifier_from_errno(err);
2698 break;
2700 case NETDEV_PRE_TYPE_CHANGE:
2701 case NETDEV_POST_TYPE_CHANGE:
2702 addrconf_type_change(dev, event);
2703 break;
2706 return NOTIFY_OK;
2710 * addrconf module should be notified of a device going up
2712 static struct notifier_block ipv6_dev_notf = {
2713 .notifier_call = addrconf_notify,
2716 static void addrconf_type_change(struct net_device *dev, unsigned long event)
2718 struct inet6_dev *idev;
2719 ASSERT_RTNL();
2721 idev = __in6_dev_get(dev);
2723 if (event == NETDEV_POST_TYPE_CHANGE)
2724 ipv6_mc_remap(idev);
2725 else if (event == NETDEV_PRE_TYPE_CHANGE)
2726 ipv6_mc_unmap(idev);
2729 static int addrconf_ifdown(struct net_device *dev, int how)
2731 struct net *net = dev_net(dev);
2732 struct inet6_dev *idev;
2733 struct inet6_ifaddr *ifa;
2734 int state, i;
2736 ASSERT_RTNL();
2738 rt6_ifdown(net, dev);
2739 neigh_ifdown(&nd_tbl, dev);
2741 idev = __in6_dev_get(dev);
2742 if (idev == NULL)
2743 return -ENODEV;
2746 * Step 1: remove reference to ipv6 device from parent device.
2747 * Do not dev_put!
2749 if (how) {
2750 idev->dead = 1;
2752 /* protected by rtnl_lock */
2753 RCU_INIT_POINTER(dev->ip6_ptr, NULL);
2755 /* Step 1.5: remove snmp6 entry */
2756 snmp6_unregister_dev(idev);
2760 /* Step 2: clear hash table */
2761 for (i = 0; i < IN6_ADDR_HSIZE; i++) {
2762 struct hlist_head *h = &inet6_addr_lst[i];
2763 struct hlist_node *n;
2765 spin_lock_bh(&addrconf_hash_lock);
2766 restart:
2767 hlist_for_each_entry_rcu(ifa, n, h, addr_lst) {
2768 if (ifa->idev == idev) {
2769 hlist_del_init_rcu(&ifa->addr_lst);
2770 addrconf_del_timer(ifa);
2771 goto restart;
2774 spin_unlock_bh(&addrconf_hash_lock);
2777 write_lock_bh(&idev->lock);
2779 /* Step 2: clear flags for stateless addrconf */
2780 if (!how)
2781 idev->if_flags &= ~(IF_RS_SENT|IF_RA_RCVD|IF_READY);
2783 #ifdef CONFIG_IPV6_PRIVACY
2784 if (how && del_timer(&idev->regen_timer))
2785 in6_dev_put(idev);
2787 /* Step 3: clear tempaddr list */
2788 while (!list_empty(&idev->tempaddr_list)) {
2789 ifa = list_first_entry(&idev->tempaddr_list,
2790 struct inet6_ifaddr, tmp_list);
2791 list_del(&ifa->tmp_list);
2792 write_unlock_bh(&idev->lock);
2793 spin_lock_bh(&ifa->lock);
2795 if (ifa->ifpub) {
2796 in6_ifa_put(ifa->ifpub);
2797 ifa->ifpub = NULL;
2799 spin_unlock_bh(&ifa->lock);
2800 in6_ifa_put(ifa);
2801 write_lock_bh(&idev->lock);
2803 #endif
2805 while (!list_empty(&idev->addr_list)) {
2806 ifa = list_first_entry(&idev->addr_list,
2807 struct inet6_ifaddr, if_list);
2808 addrconf_del_timer(ifa);
2810 list_del(&ifa->if_list);
2812 write_unlock_bh(&idev->lock);
2814 spin_lock_bh(&ifa->state_lock);
2815 state = ifa->state;
2816 ifa->state = INET6_IFADDR_STATE_DEAD;
2817 spin_unlock_bh(&ifa->state_lock);
2819 if (state != INET6_IFADDR_STATE_DEAD) {
2820 __ipv6_ifa_notify(RTM_DELADDR, ifa);
2821 atomic_notifier_call_chain(&inet6addr_chain, NETDEV_DOWN, ifa);
2823 in6_ifa_put(ifa);
2825 write_lock_bh(&idev->lock);
2828 write_unlock_bh(&idev->lock);
2830 /* Step 5: Discard multicast list */
2831 if (how)
2832 ipv6_mc_destroy_dev(idev);
2833 else
2834 ipv6_mc_down(idev);
2836 idev->tstamp = jiffies;
2838 /* Last: Shot the device (if unregistered) */
2839 if (how) {
2840 addrconf_sysctl_unregister(idev);
2841 neigh_parms_release(&nd_tbl, idev->nd_parms);
2842 neigh_ifdown(&nd_tbl, dev);
2843 in6_dev_put(idev);
2845 return 0;
2848 static void addrconf_rs_timer(unsigned long data)
2850 struct inet6_ifaddr *ifp = (struct inet6_ifaddr *) data;
2851 struct inet6_dev *idev = ifp->idev;
2853 read_lock(&idev->lock);
2854 if (idev->dead || !(idev->if_flags & IF_READY))
2855 goto out;
2857 if (idev->cnf.forwarding)
2858 goto out;
2860 /* Announcement received after solicitation was sent */
2861 if (idev->if_flags & IF_RA_RCVD)
2862 goto out;
2864 spin_lock(&ifp->lock);
2865 if (ifp->probes++ < idev->cnf.rtr_solicits) {
2866 /* The wait after the last probe can be shorter */
2867 addrconf_mod_timer(ifp, AC_RS,
2868 (ifp->probes == idev->cnf.rtr_solicits) ?
2869 idev->cnf.rtr_solicit_delay :
2870 idev->cnf.rtr_solicit_interval);
2871 spin_unlock(&ifp->lock);
2873 ndisc_send_rs(idev->dev, &ifp->addr, &in6addr_linklocal_allrouters);
2874 } else {
2875 spin_unlock(&ifp->lock);
2877 * Note: we do not support deprecated "all on-link"
2878 * assumption any longer.
2880 pr_debug("%s: no IPv6 routers present\n", idev->dev->name);
2883 out:
2884 read_unlock(&idev->lock);
2885 in6_ifa_put(ifp);
2889 * Duplicate Address Detection
2891 static void addrconf_dad_kick(struct inet6_ifaddr *ifp)
2893 unsigned long rand_num;
2894 struct inet6_dev *idev = ifp->idev;
2896 if (ifp->flags & IFA_F_OPTIMISTIC)
2897 rand_num = 0;
2898 else
2899 rand_num = net_random() % (idev->cnf.rtr_solicit_delay ? : 1);
2901 ifp->probes = idev->cnf.dad_transmits;
2902 addrconf_mod_timer(ifp, AC_DAD, rand_num);
2905 static void addrconf_dad_start(struct inet6_ifaddr *ifp)
2907 struct inet6_dev *idev = ifp->idev;
2908 struct net_device *dev = idev->dev;
2910 addrconf_join_solict(dev, &ifp->addr);
2912 net_srandom(ifp->addr.s6_addr32[3]);
2914 read_lock_bh(&idev->lock);
2915 spin_lock(&ifp->lock);
2916 if (ifp->state == INET6_IFADDR_STATE_DEAD)
2917 goto out;
2919 if (dev->flags&(IFF_NOARP|IFF_LOOPBACK) ||
2920 idev->cnf.accept_dad < 1 ||
2921 !(ifp->flags&IFA_F_TENTATIVE) ||
2922 ifp->flags & IFA_F_NODAD) {
2923 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
2924 spin_unlock(&ifp->lock);
2925 read_unlock_bh(&idev->lock);
2927 addrconf_dad_completed(ifp);
2928 return;
2931 if (!(idev->if_flags & IF_READY)) {
2932 spin_unlock(&ifp->lock);
2933 read_unlock_bh(&idev->lock);
2935 * If the device is not ready:
2936 * - keep it tentative if it is a permanent address.
2937 * - otherwise, kill it.
2939 in6_ifa_hold(ifp);
2940 addrconf_dad_stop(ifp, 0);
2941 return;
2945 * Optimistic nodes can start receiving
2946 * Frames right away
2948 if (ifp->flags & IFA_F_OPTIMISTIC)
2949 ip6_ins_rt(ifp->rt);
2951 addrconf_dad_kick(ifp);
2952 out:
2953 spin_unlock(&ifp->lock);
2954 read_unlock_bh(&idev->lock);
2957 static void addrconf_dad_timer(unsigned long data)
2959 struct inet6_ifaddr *ifp = (struct inet6_ifaddr *) data;
2960 struct inet6_dev *idev = ifp->idev;
2961 struct in6_addr mcaddr;
2963 if (!ifp->probes && addrconf_dad_end(ifp))
2964 goto out;
2966 read_lock(&idev->lock);
2967 if (idev->dead || !(idev->if_flags & IF_READY)) {
2968 read_unlock(&idev->lock);
2969 goto out;
2972 spin_lock(&ifp->lock);
2973 if (ifp->state == INET6_IFADDR_STATE_DEAD) {
2974 spin_unlock(&ifp->lock);
2975 read_unlock(&idev->lock);
2976 goto out;
2979 if (ifp->probes == 0) {
2981 * DAD was successful
2984 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
2985 spin_unlock(&ifp->lock);
2986 read_unlock(&idev->lock);
2988 addrconf_dad_completed(ifp);
2990 goto out;
2993 ifp->probes--;
2994 addrconf_mod_timer(ifp, AC_DAD, ifp->idev->nd_parms->retrans_time);
2995 spin_unlock(&ifp->lock);
2996 read_unlock(&idev->lock);
2998 /* send a neighbour solicitation for our addr */
2999 addrconf_addr_solict_mult(&ifp->addr, &mcaddr);
3000 ndisc_send_ns(ifp->idev->dev, NULL, &ifp->addr, &mcaddr, &in6addr_any);
3001 out:
3002 in6_ifa_put(ifp);
3005 static void addrconf_dad_completed(struct inet6_ifaddr *ifp)
3007 struct net_device *dev = ifp->idev->dev;
3010 * Configure the address for reception. Now it is valid.
3013 ipv6_ifa_notify(RTM_NEWADDR, ifp);
3015 /* If added prefix is link local and we are prepared to process
3016 router advertisements, start sending router solicitations.
3019 if (((ifp->idev->cnf.accept_ra == 1 && !ifp->idev->cnf.forwarding) ||
3020 ifp->idev->cnf.accept_ra == 2) &&
3021 ifp->idev->cnf.rtr_solicits > 0 &&
3022 (dev->flags&IFF_LOOPBACK) == 0 &&
3023 (ipv6_addr_type(&ifp->addr) & IPV6_ADDR_LINKLOCAL)) {
3025 * If a host as already performed a random delay
3026 * [...] as part of DAD [...] there is no need
3027 * to delay again before sending the first RS
3029 ndisc_send_rs(ifp->idev->dev, &ifp->addr, &in6addr_linklocal_allrouters);
3031 spin_lock_bh(&ifp->lock);
3032 ifp->probes = 1;
3033 ifp->idev->if_flags |= IF_RS_SENT;
3034 addrconf_mod_timer(ifp, AC_RS, ifp->idev->cnf.rtr_solicit_interval);
3035 spin_unlock_bh(&ifp->lock);
3039 static void addrconf_dad_run(struct inet6_dev *idev)
3041 struct inet6_ifaddr *ifp;
3043 read_lock_bh(&idev->lock);
3044 list_for_each_entry(ifp, &idev->addr_list, if_list) {
3045 spin_lock(&ifp->lock);
3046 if (ifp->flags & IFA_F_TENTATIVE &&
3047 ifp->state == INET6_IFADDR_STATE_DAD)
3048 addrconf_dad_kick(ifp);
3049 spin_unlock(&ifp->lock);
3051 read_unlock_bh(&idev->lock);
3054 #ifdef CONFIG_PROC_FS
3055 struct if6_iter_state {
3056 struct seq_net_private p;
3057 int bucket;
3058 int offset;
3061 static struct inet6_ifaddr *if6_get_first(struct seq_file *seq, loff_t pos)
3063 struct inet6_ifaddr *ifa = NULL;
3064 struct if6_iter_state *state = seq->private;
3065 struct net *net = seq_file_net(seq);
3066 int p = 0;
3068 /* initial bucket if pos is 0 */
3069 if (pos == 0) {
3070 state->bucket = 0;
3071 state->offset = 0;
3074 for (; state->bucket < IN6_ADDR_HSIZE; ++state->bucket) {
3075 struct hlist_node *n;
3076 hlist_for_each_entry_rcu_bh(ifa, n, &inet6_addr_lst[state->bucket],
3077 addr_lst) {
3078 /* sync with offset */
3079 if (p < state->offset) {
3080 p++;
3081 continue;
3083 state->offset++;
3084 if (net_eq(dev_net(ifa->idev->dev), net))
3085 return ifa;
3088 /* prepare for next bucket */
3089 state->offset = 0;
3090 p = 0;
3092 return NULL;
3095 static struct inet6_ifaddr *if6_get_next(struct seq_file *seq,
3096 struct inet6_ifaddr *ifa)
3098 struct if6_iter_state *state = seq->private;
3099 struct net *net = seq_file_net(seq);
3100 struct hlist_node *n = &ifa->addr_lst;
3102 hlist_for_each_entry_continue_rcu_bh(ifa, n, addr_lst) {
3103 state->offset++;
3104 if (net_eq(dev_net(ifa->idev->dev), net))
3105 return ifa;
3108 while (++state->bucket < IN6_ADDR_HSIZE) {
3109 state->offset = 0;
3110 hlist_for_each_entry_rcu_bh(ifa, n,
3111 &inet6_addr_lst[state->bucket], addr_lst) {
3112 state->offset++;
3113 if (net_eq(dev_net(ifa->idev->dev), net))
3114 return ifa;
3118 return NULL;
3121 static void *if6_seq_start(struct seq_file *seq, loff_t *pos)
3122 __acquires(rcu_bh)
3124 rcu_read_lock_bh();
3125 return if6_get_first(seq, *pos);
3128 static void *if6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
3130 struct inet6_ifaddr *ifa;
3132 ifa = if6_get_next(seq, v);
3133 ++*pos;
3134 return ifa;
3137 static void if6_seq_stop(struct seq_file *seq, void *v)
3138 __releases(rcu_bh)
3140 rcu_read_unlock_bh();
3143 static int if6_seq_show(struct seq_file *seq, void *v)
3145 struct inet6_ifaddr *ifp = (struct inet6_ifaddr *)v;
3146 seq_printf(seq, "%pi6 %02x %02x %02x %02x %8s\n",
3147 &ifp->addr,
3148 ifp->idev->dev->ifindex,
3149 ifp->prefix_len,
3150 ifp->scope,
3151 ifp->flags,
3152 ifp->idev->dev->name);
3153 return 0;
3156 static const struct seq_operations if6_seq_ops = {
3157 .start = if6_seq_start,
3158 .next = if6_seq_next,
3159 .show = if6_seq_show,
3160 .stop = if6_seq_stop,
3163 static int if6_seq_open(struct inode *inode, struct file *file)
3165 return seq_open_net(inode, file, &if6_seq_ops,
3166 sizeof(struct if6_iter_state));
3169 static const struct file_operations if6_fops = {
3170 .owner = THIS_MODULE,
3171 .open = if6_seq_open,
3172 .read = seq_read,
3173 .llseek = seq_lseek,
3174 .release = seq_release_net,
3177 static int __net_init if6_proc_net_init(struct net *net)
3179 if (!proc_net_fops_create(net, "if_inet6", S_IRUGO, &if6_fops))
3180 return -ENOMEM;
3181 return 0;
3184 static void __net_exit if6_proc_net_exit(struct net *net)
3186 proc_net_remove(net, "if_inet6");
3189 static struct pernet_operations if6_proc_net_ops = {
3190 .init = if6_proc_net_init,
3191 .exit = if6_proc_net_exit,
3194 int __init if6_proc_init(void)
3196 return register_pernet_subsys(&if6_proc_net_ops);
3199 void if6_proc_exit(void)
3201 unregister_pernet_subsys(&if6_proc_net_ops);
3203 #endif /* CONFIG_PROC_FS */
3205 #if defined(CONFIG_IPV6_MIP6) || defined(CONFIG_IPV6_MIP6_MODULE)
3206 /* Check if address is a home address configured on any interface. */
3207 int ipv6_chk_home_addr(struct net *net, const struct in6_addr *addr)
3209 int ret = 0;
3210 struct inet6_ifaddr *ifp = NULL;
3211 struct hlist_node *n;
3212 unsigned int hash = inet6_addr_hash(addr);
3214 rcu_read_lock_bh();
3215 hlist_for_each_entry_rcu_bh(ifp, n, &inet6_addr_lst[hash], addr_lst) {
3216 if (!net_eq(dev_net(ifp->idev->dev), net))
3217 continue;
3218 if (ipv6_addr_equal(&ifp->addr, addr) &&
3219 (ifp->flags & IFA_F_HOMEADDRESS)) {
3220 ret = 1;
3221 break;
3224 rcu_read_unlock_bh();
3225 return ret;
3227 #endif
3230 * Periodic address status verification
3233 static void addrconf_verify(unsigned long foo)
3235 unsigned long now, next, next_sec, next_sched;
3236 struct inet6_ifaddr *ifp;
3237 struct hlist_node *node;
3238 int i;
3240 rcu_read_lock_bh();
3241 spin_lock(&addrconf_verify_lock);
3242 now = jiffies;
3243 next = round_jiffies_up(now + ADDR_CHECK_FREQUENCY);
3245 del_timer(&addr_chk_timer);
3247 for (i = 0; i < IN6_ADDR_HSIZE; i++) {
3248 restart:
3249 hlist_for_each_entry_rcu_bh(ifp, node,
3250 &inet6_addr_lst[i], addr_lst) {
3251 unsigned long age;
3253 if (ifp->flags & IFA_F_PERMANENT)
3254 continue;
3256 spin_lock(&ifp->lock);
3257 /* We try to batch several events at once. */
3258 age = (now - ifp->tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
3260 if (ifp->valid_lft != INFINITY_LIFE_TIME &&
3261 age >= ifp->valid_lft) {
3262 spin_unlock(&ifp->lock);
3263 in6_ifa_hold(ifp);
3264 ipv6_del_addr(ifp);
3265 goto restart;
3266 } else if (ifp->prefered_lft == INFINITY_LIFE_TIME) {
3267 spin_unlock(&ifp->lock);
3268 continue;
3269 } else if (age >= ifp->prefered_lft) {
3270 /* jiffies - ifp->tstamp > age >= ifp->prefered_lft */
3271 int deprecate = 0;
3273 if (!(ifp->flags&IFA_F_DEPRECATED)) {
3274 deprecate = 1;
3275 ifp->flags |= IFA_F_DEPRECATED;
3278 if (time_before(ifp->tstamp + ifp->valid_lft * HZ, next))
3279 next = ifp->tstamp + ifp->valid_lft * HZ;
3281 spin_unlock(&ifp->lock);
3283 if (deprecate) {
3284 in6_ifa_hold(ifp);
3286 ipv6_ifa_notify(0, ifp);
3287 in6_ifa_put(ifp);
3288 goto restart;
3290 #ifdef CONFIG_IPV6_PRIVACY
3291 } else if ((ifp->flags&IFA_F_TEMPORARY) &&
3292 !(ifp->flags&IFA_F_TENTATIVE)) {
3293 unsigned long regen_advance = ifp->idev->cnf.regen_max_retry *
3294 ifp->idev->cnf.dad_transmits *
3295 ifp->idev->nd_parms->retrans_time / HZ;
3297 if (age >= ifp->prefered_lft - regen_advance) {
3298 struct inet6_ifaddr *ifpub = ifp->ifpub;
3299 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3300 next = ifp->tstamp + ifp->prefered_lft * HZ;
3301 if (!ifp->regen_count && ifpub) {
3302 ifp->regen_count++;
3303 in6_ifa_hold(ifp);
3304 in6_ifa_hold(ifpub);
3305 spin_unlock(&ifp->lock);
3307 spin_lock(&ifpub->lock);
3308 ifpub->regen_count = 0;
3309 spin_unlock(&ifpub->lock);
3310 ipv6_create_tempaddr(ifpub, ifp);
3311 in6_ifa_put(ifpub);
3312 in6_ifa_put(ifp);
3313 goto restart;
3315 } else if (time_before(ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ, next))
3316 next = ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ;
3317 spin_unlock(&ifp->lock);
3318 #endif
3319 } else {
3320 /* ifp->prefered_lft <= ifp->valid_lft */
3321 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3322 next = ifp->tstamp + ifp->prefered_lft * HZ;
3323 spin_unlock(&ifp->lock);
3328 next_sec = round_jiffies_up(next);
3329 next_sched = next;
3331 /* If rounded timeout is accurate enough, accept it. */
3332 if (time_before(next_sec, next + ADDRCONF_TIMER_FUZZ))
3333 next_sched = next_sec;
3335 /* And minimum interval is ADDRCONF_TIMER_FUZZ_MAX. */
3336 if (time_before(next_sched, jiffies + ADDRCONF_TIMER_FUZZ_MAX))
3337 next_sched = jiffies + ADDRCONF_TIMER_FUZZ_MAX;
3339 ADBG((KERN_DEBUG "now = %lu, schedule = %lu, rounded schedule = %lu => %lu\n",
3340 now, next, next_sec, next_sched));
3342 addr_chk_timer.expires = next_sched;
3343 add_timer(&addr_chk_timer);
3344 spin_unlock(&addrconf_verify_lock);
3345 rcu_read_unlock_bh();
3348 static struct in6_addr *extract_addr(struct nlattr *addr, struct nlattr *local)
3350 struct in6_addr *pfx = NULL;
3352 if (addr)
3353 pfx = nla_data(addr);
3355 if (local) {
3356 if (pfx && nla_memcmp(local, pfx, sizeof(*pfx)))
3357 pfx = NULL;
3358 else
3359 pfx = nla_data(local);
3362 return pfx;
3365 static const struct nla_policy ifa_ipv6_policy[IFA_MAX+1] = {
3366 [IFA_ADDRESS] = { .len = sizeof(struct in6_addr) },
3367 [IFA_LOCAL] = { .len = sizeof(struct in6_addr) },
3368 [IFA_CACHEINFO] = { .len = sizeof(struct ifa_cacheinfo) },
3371 static int
3372 inet6_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
3374 struct net *net = sock_net(skb->sk);
3375 struct ifaddrmsg *ifm;
3376 struct nlattr *tb[IFA_MAX+1];
3377 struct in6_addr *pfx;
3378 int err;
3380 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3381 if (err < 0)
3382 return err;
3384 ifm = nlmsg_data(nlh);
3385 pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3386 if (pfx == NULL)
3387 return -EINVAL;
3389 return inet6_addr_del(net, ifm->ifa_index, pfx, ifm->ifa_prefixlen);
3392 static int inet6_addr_modify(struct inet6_ifaddr *ifp, u8 ifa_flags,
3393 u32 prefered_lft, u32 valid_lft)
3395 u32 flags;
3396 clock_t expires;
3397 unsigned long timeout;
3399 if (!valid_lft || (prefered_lft > valid_lft))
3400 return -EINVAL;
3402 timeout = addrconf_timeout_fixup(valid_lft, HZ);
3403 if (addrconf_finite_timeout(timeout)) {
3404 expires = jiffies_to_clock_t(timeout * HZ);
3405 valid_lft = timeout;
3406 flags = RTF_EXPIRES;
3407 } else {
3408 expires = 0;
3409 flags = 0;
3410 ifa_flags |= IFA_F_PERMANENT;
3413 timeout = addrconf_timeout_fixup(prefered_lft, HZ);
3414 if (addrconf_finite_timeout(timeout)) {
3415 if (timeout == 0)
3416 ifa_flags |= IFA_F_DEPRECATED;
3417 prefered_lft = timeout;
3420 spin_lock_bh(&ifp->lock);
3421 ifp->flags = (ifp->flags & ~(IFA_F_DEPRECATED | IFA_F_PERMANENT | IFA_F_NODAD | IFA_F_HOMEADDRESS)) | ifa_flags;
3422 ifp->tstamp = jiffies;
3423 ifp->valid_lft = valid_lft;
3424 ifp->prefered_lft = prefered_lft;
3426 spin_unlock_bh(&ifp->lock);
3427 if (!(ifp->flags&IFA_F_TENTATIVE))
3428 ipv6_ifa_notify(0, ifp);
3430 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, ifp->idev->dev,
3431 expires, flags);
3432 addrconf_verify(0);
3434 return 0;
3437 static int
3438 inet6_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
3440 struct net *net = sock_net(skb->sk);
3441 struct ifaddrmsg *ifm;
3442 struct nlattr *tb[IFA_MAX+1];
3443 struct in6_addr *pfx;
3444 struct inet6_ifaddr *ifa;
3445 struct net_device *dev;
3446 u32 valid_lft = INFINITY_LIFE_TIME, preferred_lft = INFINITY_LIFE_TIME;
3447 u8 ifa_flags;
3448 int err;
3450 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3451 if (err < 0)
3452 return err;
3454 ifm = nlmsg_data(nlh);
3455 pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3456 if (pfx == NULL)
3457 return -EINVAL;
3459 if (tb[IFA_CACHEINFO]) {
3460 struct ifa_cacheinfo *ci;
3462 ci = nla_data(tb[IFA_CACHEINFO]);
3463 valid_lft = ci->ifa_valid;
3464 preferred_lft = ci->ifa_prefered;
3465 } else {
3466 preferred_lft = INFINITY_LIFE_TIME;
3467 valid_lft = INFINITY_LIFE_TIME;
3470 dev = __dev_get_by_index(net, ifm->ifa_index);
3471 if (dev == NULL)
3472 return -ENODEV;
3474 /* We ignore other flags so far. */
3475 ifa_flags = ifm->ifa_flags & (IFA_F_NODAD | IFA_F_HOMEADDRESS);
3477 ifa = ipv6_get_ifaddr(net, pfx, dev, 1);
3478 if (ifa == NULL) {
3480 * It would be best to check for !NLM_F_CREATE here but
3481 * userspace alreay relies on not having to provide this.
3483 return inet6_addr_add(net, ifm->ifa_index, pfx,
3484 ifm->ifa_prefixlen, ifa_flags,
3485 preferred_lft, valid_lft);
3488 if (nlh->nlmsg_flags & NLM_F_EXCL ||
3489 !(nlh->nlmsg_flags & NLM_F_REPLACE))
3490 err = -EEXIST;
3491 else
3492 err = inet6_addr_modify(ifa, ifa_flags, preferred_lft, valid_lft);
3494 in6_ifa_put(ifa);
3496 return err;
3499 static void put_ifaddrmsg(struct nlmsghdr *nlh, u8 prefixlen, u8 flags,
3500 u8 scope, int ifindex)
3502 struct ifaddrmsg *ifm;
3504 ifm = nlmsg_data(nlh);
3505 ifm->ifa_family = AF_INET6;
3506 ifm->ifa_prefixlen = prefixlen;
3507 ifm->ifa_flags = flags;
3508 ifm->ifa_scope = scope;
3509 ifm->ifa_index = ifindex;
3512 static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
3513 unsigned long tstamp, u32 preferred, u32 valid)
3515 struct ifa_cacheinfo ci;
3517 ci.cstamp = cstamp_delta(cstamp);
3518 ci.tstamp = cstamp_delta(tstamp);
3519 ci.ifa_prefered = preferred;
3520 ci.ifa_valid = valid;
3522 return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
3525 static inline int rt_scope(int ifa_scope)
3527 if (ifa_scope & IFA_HOST)
3528 return RT_SCOPE_HOST;
3529 else if (ifa_scope & IFA_LINK)
3530 return RT_SCOPE_LINK;
3531 else if (ifa_scope & IFA_SITE)
3532 return RT_SCOPE_SITE;
3533 else
3534 return RT_SCOPE_UNIVERSE;
3537 static inline int inet6_ifaddr_msgsize(void)
3539 return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
3540 + nla_total_size(16) /* IFA_ADDRESS */
3541 + nla_total_size(sizeof(struct ifa_cacheinfo));
3544 static int inet6_fill_ifaddr(struct sk_buff *skb, struct inet6_ifaddr *ifa,
3545 u32 portid, u32 seq, int event, unsigned int flags)
3547 struct nlmsghdr *nlh;
3548 u32 preferred, valid;
3550 nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
3551 if (nlh == NULL)
3552 return -EMSGSIZE;
3554 put_ifaddrmsg(nlh, ifa->prefix_len, ifa->flags, rt_scope(ifa->scope),
3555 ifa->idev->dev->ifindex);
3557 if (!(ifa->flags&IFA_F_PERMANENT)) {
3558 preferred = ifa->prefered_lft;
3559 valid = ifa->valid_lft;
3560 if (preferred != INFINITY_LIFE_TIME) {
3561 long tval = (jiffies - ifa->tstamp)/HZ;
3562 if (preferred > tval)
3563 preferred -= tval;
3564 else
3565 preferred = 0;
3566 if (valid != INFINITY_LIFE_TIME) {
3567 if (valid > tval)
3568 valid -= tval;
3569 else
3570 valid = 0;
3573 } else {
3574 preferred = INFINITY_LIFE_TIME;
3575 valid = INFINITY_LIFE_TIME;
3578 if (nla_put(skb, IFA_ADDRESS, 16, &ifa->addr) < 0 ||
3579 put_cacheinfo(skb, ifa->cstamp, ifa->tstamp, preferred, valid) < 0) {
3580 nlmsg_cancel(skb, nlh);
3581 return -EMSGSIZE;
3584 return nlmsg_end(skb, nlh);
3587 static int inet6_fill_ifmcaddr(struct sk_buff *skb, struct ifmcaddr6 *ifmca,
3588 u32 portid, u32 seq, int event, u16 flags)
3590 struct nlmsghdr *nlh;
3591 u8 scope = RT_SCOPE_UNIVERSE;
3592 int ifindex = ifmca->idev->dev->ifindex;
3594 if (ipv6_addr_scope(&ifmca->mca_addr) & IFA_SITE)
3595 scope = RT_SCOPE_SITE;
3597 nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
3598 if (nlh == NULL)
3599 return -EMSGSIZE;
3601 put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3602 if (nla_put(skb, IFA_MULTICAST, 16, &ifmca->mca_addr) < 0 ||
3603 put_cacheinfo(skb, ifmca->mca_cstamp, ifmca->mca_tstamp,
3604 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3605 nlmsg_cancel(skb, nlh);
3606 return -EMSGSIZE;
3609 return nlmsg_end(skb, nlh);
3612 static int inet6_fill_ifacaddr(struct sk_buff *skb, struct ifacaddr6 *ifaca,
3613 u32 portid, u32 seq, int event, unsigned int flags)
3615 struct nlmsghdr *nlh;
3616 u8 scope = RT_SCOPE_UNIVERSE;
3617 int ifindex = ifaca->aca_idev->dev->ifindex;
3619 if (ipv6_addr_scope(&ifaca->aca_addr) & IFA_SITE)
3620 scope = RT_SCOPE_SITE;
3622 nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
3623 if (nlh == NULL)
3624 return -EMSGSIZE;
3626 put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3627 if (nla_put(skb, IFA_ANYCAST, 16, &ifaca->aca_addr) < 0 ||
3628 put_cacheinfo(skb, ifaca->aca_cstamp, ifaca->aca_tstamp,
3629 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3630 nlmsg_cancel(skb, nlh);
3631 return -EMSGSIZE;
3634 return nlmsg_end(skb, nlh);
3637 enum addr_type_t {
3638 UNICAST_ADDR,
3639 MULTICAST_ADDR,
3640 ANYCAST_ADDR,
3643 /* called with rcu_read_lock() */
3644 static int in6_dump_addrs(struct inet6_dev *idev, struct sk_buff *skb,
3645 struct netlink_callback *cb, enum addr_type_t type,
3646 int s_ip_idx, int *p_ip_idx)
3648 struct ifmcaddr6 *ifmca;
3649 struct ifacaddr6 *ifaca;
3650 int err = 1;
3651 int ip_idx = *p_ip_idx;
3653 read_lock_bh(&idev->lock);
3654 switch (type) {
3655 case UNICAST_ADDR: {
3656 struct inet6_ifaddr *ifa;
3658 /* unicast address incl. temp addr */
3659 list_for_each_entry(ifa, &idev->addr_list, if_list) {
3660 if (++ip_idx < s_ip_idx)
3661 continue;
3662 err = inet6_fill_ifaddr(skb, ifa,
3663 NETLINK_CB(cb->skb).portid,
3664 cb->nlh->nlmsg_seq,
3665 RTM_NEWADDR,
3666 NLM_F_MULTI);
3667 if (err <= 0)
3668 break;
3670 break;
3672 case MULTICAST_ADDR:
3673 /* multicast address */
3674 for (ifmca = idev->mc_list; ifmca;
3675 ifmca = ifmca->next, ip_idx++) {
3676 if (ip_idx < s_ip_idx)
3677 continue;
3678 err = inet6_fill_ifmcaddr(skb, ifmca,
3679 NETLINK_CB(cb->skb).portid,
3680 cb->nlh->nlmsg_seq,
3681 RTM_GETMULTICAST,
3682 NLM_F_MULTI);
3683 if (err <= 0)
3684 break;
3686 break;
3687 case ANYCAST_ADDR:
3688 /* anycast address */
3689 for (ifaca = idev->ac_list; ifaca;
3690 ifaca = ifaca->aca_next, ip_idx++) {
3691 if (ip_idx < s_ip_idx)
3692 continue;
3693 err = inet6_fill_ifacaddr(skb, ifaca,
3694 NETLINK_CB(cb->skb).portid,
3695 cb->nlh->nlmsg_seq,
3696 RTM_GETANYCAST,
3697 NLM_F_MULTI);
3698 if (err <= 0)
3699 break;
3701 break;
3702 default:
3703 break;
3705 read_unlock_bh(&idev->lock);
3706 *p_ip_idx = ip_idx;
3707 return err;
3710 static int inet6_dump_addr(struct sk_buff *skb, struct netlink_callback *cb,
3711 enum addr_type_t type)
3713 struct net *net = sock_net(skb->sk);
3714 int h, s_h;
3715 int idx, ip_idx;
3716 int s_idx, s_ip_idx;
3717 struct net_device *dev;
3718 struct inet6_dev *idev;
3719 struct hlist_head *head;
3720 struct hlist_node *node;
3722 s_h = cb->args[0];
3723 s_idx = idx = cb->args[1];
3724 s_ip_idx = ip_idx = cb->args[2];
3726 rcu_read_lock();
3727 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
3728 idx = 0;
3729 head = &net->dev_index_head[h];
3730 hlist_for_each_entry_rcu(dev, node, head, index_hlist) {
3731 if (idx < s_idx)
3732 goto cont;
3733 if (h > s_h || idx > s_idx)
3734 s_ip_idx = 0;
3735 ip_idx = 0;
3736 idev = __in6_dev_get(dev);
3737 if (!idev)
3738 goto cont;
3740 if (in6_dump_addrs(idev, skb, cb, type,
3741 s_ip_idx, &ip_idx) <= 0)
3742 goto done;
3743 cont:
3744 idx++;
3747 done:
3748 rcu_read_unlock();
3749 cb->args[0] = h;
3750 cb->args[1] = idx;
3751 cb->args[2] = ip_idx;
3753 return skb->len;
3756 static int inet6_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
3758 enum addr_type_t type = UNICAST_ADDR;
3760 return inet6_dump_addr(skb, cb, type);
3763 static int inet6_dump_ifmcaddr(struct sk_buff *skb, struct netlink_callback *cb)
3765 enum addr_type_t type = MULTICAST_ADDR;
3767 return inet6_dump_addr(skb, cb, type);
3771 static int inet6_dump_ifacaddr(struct sk_buff *skb, struct netlink_callback *cb)
3773 enum addr_type_t type = ANYCAST_ADDR;
3775 return inet6_dump_addr(skb, cb, type);
3778 static int inet6_rtm_getaddr(struct sk_buff *in_skb, struct nlmsghdr *nlh,
3779 void *arg)
3781 struct net *net = sock_net(in_skb->sk);
3782 struct ifaddrmsg *ifm;
3783 struct nlattr *tb[IFA_MAX+1];
3784 struct in6_addr *addr = NULL;
3785 struct net_device *dev = NULL;
3786 struct inet6_ifaddr *ifa;
3787 struct sk_buff *skb;
3788 int err;
3790 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3791 if (err < 0)
3792 goto errout;
3794 addr = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3795 if (addr == NULL) {
3796 err = -EINVAL;
3797 goto errout;
3800 ifm = nlmsg_data(nlh);
3801 if (ifm->ifa_index)
3802 dev = __dev_get_by_index(net, ifm->ifa_index);
3804 ifa = ipv6_get_ifaddr(net, addr, dev, 1);
3805 if (!ifa) {
3806 err = -EADDRNOTAVAIL;
3807 goto errout;
3810 skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_KERNEL);
3811 if (!skb) {
3812 err = -ENOBUFS;
3813 goto errout_ifa;
3816 err = inet6_fill_ifaddr(skb, ifa, NETLINK_CB(in_skb).portid,
3817 nlh->nlmsg_seq, RTM_NEWADDR, 0);
3818 if (err < 0) {
3819 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
3820 WARN_ON(err == -EMSGSIZE);
3821 kfree_skb(skb);
3822 goto errout_ifa;
3824 err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
3825 errout_ifa:
3826 in6_ifa_put(ifa);
3827 errout:
3828 return err;
3831 static void inet6_ifa_notify(int event, struct inet6_ifaddr *ifa)
3833 struct sk_buff *skb;
3834 struct net *net = dev_net(ifa->idev->dev);
3835 int err = -ENOBUFS;
3837 skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_ATOMIC);
3838 if (skb == NULL)
3839 goto errout;
3841 err = inet6_fill_ifaddr(skb, ifa, 0, 0, event, 0);
3842 if (err < 0) {
3843 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
3844 WARN_ON(err == -EMSGSIZE);
3845 kfree_skb(skb);
3846 goto errout;
3848 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
3849 return;
3850 errout:
3851 if (err < 0)
3852 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
3855 static inline void ipv6_store_devconf(struct ipv6_devconf *cnf,
3856 __s32 *array, int bytes)
3858 BUG_ON(bytes < (DEVCONF_MAX * 4));
3860 memset(array, 0, bytes);
3861 array[DEVCONF_FORWARDING] = cnf->forwarding;
3862 array[DEVCONF_HOPLIMIT] = cnf->hop_limit;
3863 array[DEVCONF_MTU6] = cnf->mtu6;
3864 array[DEVCONF_ACCEPT_RA] = cnf->accept_ra;
3865 array[DEVCONF_ACCEPT_REDIRECTS] = cnf->accept_redirects;
3866 array[DEVCONF_AUTOCONF] = cnf->autoconf;
3867 array[DEVCONF_DAD_TRANSMITS] = cnf->dad_transmits;
3868 array[DEVCONF_RTR_SOLICITS] = cnf->rtr_solicits;
3869 array[DEVCONF_RTR_SOLICIT_INTERVAL] =
3870 jiffies_to_msecs(cnf->rtr_solicit_interval);
3871 array[DEVCONF_RTR_SOLICIT_DELAY] =
3872 jiffies_to_msecs(cnf->rtr_solicit_delay);
3873 array[DEVCONF_FORCE_MLD_VERSION] = cnf->force_mld_version;
3874 #ifdef CONFIG_IPV6_PRIVACY
3875 array[DEVCONF_USE_TEMPADDR] = cnf->use_tempaddr;
3876 array[DEVCONF_TEMP_VALID_LFT] = cnf->temp_valid_lft;
3877 array[DEVCONF_TEMP_PREFERED_LFT] = cnf->temp_prefered_lft;
3878 array[DEVCONF_REGEN_MAX_RETRY] = cnf->regen_max_retry;
3879 array[DEVCONF_MAX_DESYNC_FACTOR] = cnf->max_desync_factor;
3880 #endif
3881 array[DEVCONF_MAX_ADDRESSES] = cnf->max_addresses;
3882 array[DEVCONF_ACCEPT_RA_DEFRTR] = cnf->accept_ra_defrtr;
3883 array[DEVCONF_ACCEPT_RA_PINFO] = cnf->accept_ra_pinfo;
3884 #ifdef CONFIG_IPV6_ROUTER_PREF
3885 array[DEVCONF_ACCEPT_RA_RTR_PREF] = cnf->accept_ra_rtr_pref;
3886 array[DEVCONF_RTR_PROBE_INTERVAL] =
3887 jiffies_to_msecs(cnf->rtr_probe_interval);
3888 #ifdef CONFIG_IPV6_ROUTE_INFO
3889 array[DEVCONF_ACCEPT_RA_RT_INFO_MAX_PLEN] = cnf->accept_ra_rt_info_max_plen;
3890 #endif
3891 #endif
3892 array[DEVCONF_PROXY_NDP] = cnf->proxy_ndp;
3893 array[DEVCONF_ACCEPT_SOURCE_ROUTE] = cnf->accept_source_route;
3894 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
3895 array[DEVCONF_OPTIMISTIC_DAD] = cnf->optimistic_dad;
3896 #endif
3897 #ifdef CONFIG_IPV6_MROUTE
3898 array[DEVCONF_MC_FORWARDING] = cnf->mc_forwarding;
3899 #endif
3900 array[DEVCONF_DISABLE_IPV6] = cnf->disable_ipv6;
3901 array[DEVCONF_ACCEPT_DAD] = cnf->accept_dad;
3902 array[DEVCONF_FORCE_TLLAO] = cnf->force_tllao;
3905 static inline size_t inet6_ifla6_size(void)
3907 return nla_total_size(4) /* IFLA_INET6_FLAGS */
3908 + nla_total_size(sizeof(struct ifla_cacheinfo))
3909 + nla_total_size(DEVCONF_MAX * 4) /* IFLA_INET6_CONF */
3910 + nla_total_size(IPSTATS_MIB_MAX * 8) /* IFLA_INET6_STATS */
3911 + nla_total_size(ICMP6_MIB_MAX * 8); /* IFLA_INET6_ICMP6STATS */
3914 static inline size_t inet6_if_nlmsg_size(void)
3916 return NLMSG_ALIGN(sizeof(struct ifinfomsg))
3917 + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
3918 + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
3919 + nla_total_size(4) /* IFLA_MTU */
3920 + nla_total_size(4) /* IFLA_LINK */
3921 + nla_total_size(inet6_ifla6_size()); /* IFLA_PROTINFO */
3924 static inline void __snmp6_fill_statsdev(u64 *stats, atomic_long_t *mib,
3925 int items, int bytes)
3927 int i;
3928 int pad = bytes - sizeof(u64) * items;
3929 BUG_ON(pad < 0);
3931 /* Use put_unaligned() because stats may not be aligned for u64. */
3932 put_unaligned(items, &stats[0]);
3933 for (i = 1; i < items; i++)
3934 put_unaligned(atomic_long_read(&mib[i]), &stats[i]);
3936 memset(&stats[items], 0, pad);
3939 static inline void __snmp6_fill_stats64(u64 *stats, void __percpu **mib,
3940 int items, int bytes, size_t syncpoff)
3942 int i;
3943 int pad = bytes - sizeof(u64) * items;
3944 BUG_ON(pad < 0);
3946 /* Use put_unaligned() because stats may not be aligned for u64. */
3947 put_unaligned(items, &stats[0]);
3948 for (i = 1; i < items; i++)
3949 put_unaligned(snmp_fold_field64(mib, i, syncpoff), &stats[i]);
3951 memset(&stats[items], 0, pad);
3954 static void snmp6_fill_stats(u64 *stats, struct inet6_dev *idev, int attrtype,
3955 int bytes)
3957 switch (attrtype) {
3958 case IFLA_INET6_STATS:
3959 __snmp6_fill_stats64(stats, (void __percpu **)idev->stats.ipv6,
3960 IPSTATS_MIB_MAX, bytes, offsetof(struct ipstats_mib, syncp));
3961 break;
3962 case IFLA_INET6_ICMP6STATS:
3963 __snmp6_fill_statsdev(stats, idev->stats.icmpv6dev->mibs, ICMP6_MIB_MAX, bytes);
3964 break;
3968 static int inet6_fill_ifla6_attrs(struct sk_buff *skb, struct inet6_dev *idev)
3970 struct nlattr *nla;
3971 struct ifla_cacheinfo ci;
3973 if (nla_put_u32(skb, IFLA_INET6_FLAGS, idev->if_flags))
3974 goto nla_put_failure;
3975 ci.max_reasm_len = IPV6_MAXPLEN;
3976 ci.tstamp = cstamp_delta(idev->tstamp);
3977 ci.reachable_time = jiffies_to_msecs(idev->nd_parms->reachable_time);
3978 ci.retrans_time = jiffies_to_msecs(idev->nd_parms->retrans_time);
3979 if (nla_put(skb, IFLA_INET6_CACHEINFO, sizeof(ci), &ci))
3980 goto nla_put_failure;
3981 nla = nla_reserve(skb, IFLA_INET6_CONF, DEVCONF_MAX * sizeof(s32));
3982 if (nla == NULL)
3983 goto nla_put_failure;
3984 ipv6_store_devconf(&idev->cnf, nla_data(nla), nla_len(nla));
3986 /* XXX - MC not implemented */
3988 nla = nla_reserve(skb, IFLA_INET6_STATS, IPSTATS_MIB_MAX * sizeof(u64));
3989 if (nla == NULL)
3990 goto nla_put_failure;
3991 snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_STATS, nla_len(nla));
3993 nla = nla_reserve(skb, IFLA_INET6_ICMP6STATS, ICMP6_MIB_MAX * sizeof(u64));
3994 if (nla == NULL)
3995 goto nla_put_failure;
3996 snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_ICMP6STATS, nla_len(nla));
3998 return 0;
4000 nla_put_failure:
4001 return -EMSGSIZE;
4004 static size_t inet6_get_link_af_size(const struct net_device *dev)
4006 if (!__in6_dev_get(dev))
4007 return 0;
4009 return inet6_ifla6_size();
4012 static int inet6_fill_link_af(struct sk_buff *skb, const struct net_device *dev)
4014 struct inet6_dev *idev = __in6_dev_get(dev);
4016 if (!idev)
4017 return -ENODATA;
4019 if (inet6_fill_ifla6_attrs(skb, idev) < 0)
4020 return -EMSGSIZE;
4022 return 0;
4025 static int inet6_fill_ifinfo(struct sk_buff *skb, struct inet6_dev *idev,
4026 u32 portid, u32 seq, int event, unsigned int flags)
4028 struct net_device *dev = idev->dev;
4029 struct ifinfomsg *hdr;
4030 struct nlmsghdr *nlh;
4031 void *protoinfo;
4033 nlh = nlmsg_put(skb, portid, seq, event, sizeof(*hdr), flags);
4034 if (nlh == NULL)
4035 return -EMSGSIZE;
4037 hdr = nlmsg_data(nlh);
4038 hdr->ifi_family = AF_INET6;
4039 hdr->__ifi_pad = 0;
4040 hdr->ifi_type = dev->type;
4041 hdr->ifi_index = dev->ifindex;
4042 hdr->ifi_flags = dev_get_flags(dev);
4043 hdr->ifi_change = 0;
4045 if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
4046 (dev->addr_len &&
4047 nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
4048 nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
4049 (dev->ifindex != dev->iflink &&
4050 nla_put_u32(skb, IFLA_LINK, dev->iflink)))
4051 goto nla_put_failure;
4052 protoinfo = nla_nest_start(skb, IFLA_PROTINFO);
4053 if (protoinfo == NULL)
4054 goto nla_put_failure;
4056 if (inet6_fill_ifla6_attrs(skb, idev) < 0)
4057 goto nla_put_failure;
4059 nla_nest_end(skb, protoinfo);
4060 return nlmsg_end(skb, nlh);
4062 nla_put_failure:
4063 nlmsg_cancel(skb, nlh);
4064 return -EMSGSIZE;
4067 static int inet6_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
4069 struct net *net = sock_net(skb->sk);
4070 int h, s_h;
4071 int idx = 0, s_idx;
4072 struct net_device *dev;
4073 struct inet6_dev *idev;
4074 struct hlist_head *head;
4075 struct hlist_node *node;
4077 s_h = cb->args[0];
4078 s_idx = cb->args[1];
4080 rcu_read_lock();
4081 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
4082 idx = 0;
4083 head = &net->dev_index_head[h];
4084 hlist_for_each_entry_rcu(dev, node, head, index_hlist) {
4085 if (idx < s_idx)
4086 goto cont;
4087 idev = __in6_dev_get(dev);
4088 if (!idev)
4089 goto cont;
4090 if (inet6_fill_ifinfo(skb, idev,
4091 NETLINK_CB(cb->skb).portid,
4092 cb->nlh->nlmsg_seq,
4093 RTM_NEWLINK, NLM_F_MULTI) <= 0)
4094 goto out;
4095 cont:
4096 idx++;
4099 out:
4100 rcu_read_unlock();
4101 cb->args[1] = idx;
4102 cb->args[0] = h;
4104 return skb->len;
4107 void inet6_ifinfo_notify(int event, struct inet6_dev *idev)
4109 struct sk_buff *skb;
4110 struct net *net = dev_net(idev->dev);
4111 int err = -ENOBUFS;
4113 skb = nlmsg_new(inet6_if_nlmsg_size(), GFP_ATOMIC);
4114 if (skb == NULL)
4115 goto errout;
4117 err = inet6_fill_ifinfo(skb, idev, 0, 0, event, 0);
4118 if (err < 0) {
4119 /* -EMSGSIZE implies BUG in inet6_if_nlmsg_size() */
4120 WARN_ON(err == -EMSGSIZE);
4121 kfree_skb(skb);
4122 goto errout;
4124 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFINFO, NULL, GFP_ATOMIC);
4125 return;
4126 errout:
4127 if (err < 0)
4128 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFINFO, err);
4131 static inline size_t inet6_prefix_nlmsg_size(void)
4133 return NLMSG_ALIGN(sizeof(struct prefixmsg))
4134 + nla_total_size(sizeof(struct in6_addr))
4135 + nla_total_size(sizeof(struct prefix_cacheinfo));
4138 static int inet6_fill_prefix(struct sk_buff *skb, struct inet6_dev *idev,
4139 struct prefix_info *pinfo, u32 portid, u32 seq,
4140 int event, unsigned int flags)
4142 struct prefixmsg *pmsg;
4143 struct nlmsghdr *nlh;
4144 struct prefix_cacheinfo ci;
4146 nlh = nlmsg_put(skb, portid, seq, event, sizeof(*pmsg), flags);
4147 if (nlh == NULL)
4148 return -EMSGSIZE;
4150 pmsg = nlmsg_data(nlh);
4151 pmsg->prefix_family = AF_INET6;
4152 pmsg->prefix_pad1 = 0;
4153 pmsg->prefix_pad2 = 0;
4154 pmsg->prefix_ifindex = idev->dev->ifindex;
4155 pmsg->prefix_len = pinfo->prefix_len;
4156 pmsg->prefix_type = pinfo->type;
4157 pmsg->prefix_pad3 = 0;
4158 pmsg->prefix_flags = 0;
4159 if (pinfo->onlink)
4160 pmsg->prefix_flags |= IF_PREFIX_ONLINK;
4161 if (pinfo->autoconf)
4162 pmsg->prefix_flags |= IF_PREFIX_AUTOCONF;
4164 if (nla_put(skb, PREFIX_ADDRESS, sizeof(pinfo->prefix), &pinfo->prefix))
4165 goto nla_put_failure;
4166 ci.preferred_time = ntohl(pinfo->prefered);
4167 ci.valid_time = ntohl(pinfo->valid);
4168 if (nla_put(skb, PREFIX_CACHEINFO, sizeof(ci), &ci))
4169 goto nla_put_failure;
4170 return nlmsg_end(skb, nlh);
4172 nla_put_failure:
4173 nlmsg_cancel(skb, nlh);
4174 return -EMSGSIZE;
4177 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
4178 struct prefix_info *pinfo)
4180 struct sk_buff *skb;
4181 struct net *net = dev_net(idev->dev);
4182 int err = -ENOBUFS;
4184 skb = nlmsg_new(inet6_prefix_nlmsg_size(), GFP_ATOMIC);
4185 if (skb == NULL)
4186 goto errout;
4188 err = inet6_fill_prefix(skb, idev, pinfo, 0, 0, event, 0);
4189 if (err < 0) {
4190 /* -EMSGSIZE implies BUG in inet6_prefix_nlmsg_size() */
4191 WARN_ON(err == -EMSGSIZE);
4192 kfree_skb(skb);
4193 goto errout;
4195 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_PREFIX, NULL, GFP_ATOMIC);
4196 return;
4197 errout:
4198 if (err < 0)
4199 rtnl_set_sk_err(net, RTNLGRP_IPV6_PREFIX, err);
4202 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
4204 inet6_ifa_notify(event ? : RTM_NEWADDR, ifp);
4206 switch (event) {
4207 case RTM_NEWADDR:
4209 * If the address was optimistic
4210 * we inserted the route at the start of
4211 * our DAD process, so we don't need
4212 * to do it again
4214 if (!(ifp->rt->rt6i_node))
4215 ip6_ins_rt(ifp->rt);
4216 if (ifp->idev->cnf.forwarding)
4217 addrconf_join_anycast(ifp);
4218 break;
4219 case RTM_DELADDR:
4220 if (ifp->idev->cnf.forwarding)
4221 addrconf_leave_anycast(ifp);
4222 addrconf_leave_solict(ifp->idev, &ifp->addr);
4223 dst_hold(&ifp->rt->dst);
4225 if (ip6_del_rt(ifp->rt))
4226 dst_free(&ifp->rt->dst);
4227 break;
4231 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
4233 rcu_read_lock_bh();
4234 if (likely(ifp->idev->dead == 0))
4235 __ipv6_ifa_notify(event, ifp);
4236 rcu_read_unlock_bh();
4239 #ifdef CONFIG_SYSCTL
4241 static
4242 int addrconf_sysctl_forward(ctl_table *ctl, int write,
4243 void __user *buffer, size_t *lenp, loff_t *ppos)
4245 int *valp = ctl->data;
4246 int val = *valp;
4247 loff_t pos = *ppos;
4248 ctl_table lctl;
4249 int ret;
4252 * ctl->data points to idev->cnf.forwarding, we should
4253 * not modify it until we get the rtnl lock.
4255 lctl = *ctl;
4256 lctl.data = &val;
4258 ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
4260 if (write)
4261 ret = addrconf_fixup_forwarding(ctl, valp, val);
4262 if (ret)
4263 *ppos = pos;
4264 return ret;
4267 static void dev_disable_change(struct inet6_dev *idev)
4269 if (!idev || !idev->dev)
4270 return;
4272 if (idev->cnf.disable_ipv6)
4273 addrconf_notify(NULL, NETDEV_DOWN, idev->dev);
4274 else
4275 addrconf_notify(NULL, NETDEV_UP, idev->dev);
4278 static void addrconf_disable_change(struct net *net, __s32 newf)
4280 struct net_device *dev;
4281 struct inet6_dev *idev;
4283 rcu_read_lock();
4284 for_each_netdev_rcu(net, dev) {
4285 idev = __in6_dev_get(dev);
4286 if (idev) {
4287 int changed = (!idev->cnf.disable_ipv6) ^ (!newf);
4288 idev->cnf.disable_ipv6 = newf;
4289 if (changed)
4290 dev_disable_change(idev);
4293 rcu_read_unlock();
4296 static int addrconf_disable_ipv6(struct ctl_table *table, int *p, int newf)
4298 struct net *net;
4299 int old;
4301 if (!rtnl_trylock())
4302 return restart_syscall();
4304 net = (struct net *)table->extra2;
4305 old = *p;
4306 *p = newf;
4308 if (p == &net->ipv6.devconf_dflt->disable_ipv6) {
4309 rtnl_unlock();
4310 return 0;
4313 if (p == &net->ipv6.devconf_all->disable_ipv6) {
4314 net->ipv6.devconf_dflt->disable_ipv6 = newf;
4315 addrconf_disable_change(net, newf);
4316 } else if ((!newf) ^ (!old))
4317 dev_disable_change((struct inet6_dev *)table->extra1);
4319 rtnl_unlock();
4320 return 0;
4323 static
4324 int addrconf_sysctl_disable(ctl_table *ctl, int write,
4325 void __user *buffer, size_t *lenp, loff_t *ppos)
4327 int *valp = ctl->data;
4328 int val = *valp;
4329 loff_t pos = *ppos;
4330 ctl_table lctl;
4331 int ret;
4334 * ctl->data points to idev->cnf.disable_ipv6, we should
4335 * not modify it until we get the rtnl lock.
4337 lctl = *ctl;
4338 lctl.data = &val;
4340 ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
4342 if (write)
4343 ret = addrconf_disable_ipv6(ctl, valp, val);
4344 if (ret)
4345 *ppos = pos;
4346 return ret;
4349 static struct addrconf_sysctl_table
4351 struct ctl_table_header *sysctl_header;
4352 ctl_table addrconf_vars[DEVCONF_MAX+1];
4353 } addrconf_sysctl __read_mostly = {
4354 .sysctl_header = NULL,
4355 .addrconf_vars = {
4357 .procname = "forwarding",
4358 .data = &ipv6_devconf.forwarding,
4359 .maxlen = sizeof(int),
4360 .mode = 0644,
4361 .proc_handler = addrconf_sysctl_forward,
4364 .procname = "hop_limit",
4365 .data = &ipv6_devconf.hop_limit,
4366 .maxlen = sizeof(int),
4367 .mode = 0644,
4368 .proc_handler = proc_dointvec,
4371 .procname = "mtu",
4372 .data = &ipv6_devconf.mtu6,
4373 .maxlen = sizeof(int),
4374 .mode = 0644,
4375 .proc_handler = proc_dointvec,
4378 .procname = "accept_ra",
4379 .data = &ipv6_devconf.accept_ra,
4380 .maxlen = sizeof(int),
4381 .mode = 0644,
4382 .proc_handler = proc_dointvec,
4385 .procname = "accept_redirects",
4386 .data = &ipv6_devconf.accept_redirects,
4387 .maxlen = sizeof(int),
4388 .mode = 0644,
4389 .proc_handler = proc_dointvec,
4392 .procname = "autoconf",
4393 .data = &ipv6_devconf.autoconf,
4394 .maxlen = sizeof(int),
4395 .mode = 0644,
4396 .proc_handler = proc_dointvec,
4399 .procname = "dad_transmits",
4400 .data = &ipv6_devconf.dad_transmits,
4401 .maxlen = sizeof(int),
4402 .mode = 0644,
4403 .proc_handler = proc_dointvec,
4406 .procname = "router_solicitations",
4407 .data = &ipv6_devconf.rtr_solicits,
4408 .maxlen = sizeof(int),
4409 .mode = 0644,
4410 .proc_handler = proc_dointvec,
4413 .procname = "router_solicitation_interval",
4414 .data = &ipv6_devconf.rtr_solicit_interval,
4415 .maxlen = sizeof(int),
4416 .mode = 0644,
4417 .proc_handler = proc_dointvec_jiffies,
4420 .procname = "router_solicitation_delay",
4421 .data = &ipv6_devconf.rtr_solicit_delay,
4422 .maxlen = sizeof(int),
4423 .mode = 0644,
4424 .proc_handler = proc_dointvec_jiffies,
4427 .procname = "force_mld_version",
4428 .data = &ipv6_devconf.force_mld_version,
4429 .maxlen = sizeof(int),
4430 .mode = 0644,
4431 .proc_handler = proc_dointvec,
4433 #ifdef CONFIG_IPV6_PRIVACY
4435 .procname = "use_tempaddr",
4436 .data = &ipv6_devconf.use_tempaddr,
4437 .maxlen = sizeof(int),
4438 .mode = 0644,
4439 .proc_handler = proc_dointvec,
4442 .procname = "temp_valid_lft",
4443 .data = &ipv6_devconf.temp_valid_lft,
4444 .maxlen = sizeof(int),
4445 .mode = 0644,
4446 .proc_handler = proc_dointvec,
4449 .procname = "temp_prefered_lft",
4450 .data = &ipv6_devconf.temp_prefered_lft,
4451 .maxlen = sizeof(int),
4452 .mode = 0644,
4453 .proc_handler = proc_dointvec,
4456 .procname = "regen_max_retry",
4457 .data = &ipv6_devconf.regen_max_retry,
4458 .maxlen = sizeof(int),
4459 .mode = 0644,
4460 .proc_handler = proc_dointvec,
4463 .procname = "max_desync_factor",
4464 .data = &ipv6_devconf.max_desync_factor,
4465 .maxlen = sizeof(int),
4466 .mode = 0644,
4467 .proc_handler = proc_dointvec,
4469 #endif
4471 .procname = "max_addresses",
4472 .data = &ipv6_devconf.max_addresses,
4473 .maxlen = sizeof(int),
4474 .mode = 0644,
4475 .proc_handler = proc_dointvec,
4478 .procname = "accept_ra_defrtr",
4479 .data = &ipv6_devconf.accept_ra_defrtr,
4480 .maxlen = sizeof(int),
4481 .mode = 0644,
4482 .proc_handler = proc_dointvec,
4485 .procname = "accept_ra_pinfo",
4486 .data = &ipv6_devconf.accept_ra_pinfo,
4487 .maxlen = sizeof(int),
4488 .mode = 0644,
4489 .proc_handler = proc_dointvec,
4491 #ifdef CONFIG_IPV6_ROUTER_PREF
4493 .procname = "accept_ra_rtr_pref",
4494 .data = &ipv6_devconf.accept_ra_rtr_pref,
4495 .maxlen = sizeof(int),
4496 .mode = 0644,
4497 .proc_handler = proc_dointvec,
4500 .procname = "router_probe_interval",
4501 .data = &ipv6_devconf.rtr_probe_interval,
4502 .maxlen = sizeof(int),
4503 .mode = 0644,
4504 .proc_handler = proc_dointvec_jiffies,
4506 #ifdef CONFIG_IPV6_ROUTE_INFO
4508 .procname = "accept_ra_rt_info_max_plen",
4509 .data = &ipv6_devconf.accept_ra_rt_info_max_plen,
4510 .maxlen = sizeof(int),
4511 .mode = 0644,
4512 .proc_handler = proc_dointvec,
4514 #endif
4515 #endif
4517 .procname = "proxy_ndp",
4518 .data = &ipv6_devconf.proxy_ndp,
4519 .maxlen = sizeof(int),
4520 .mode = 0644,
4521 .proc_handler = proc_dointvec,
4524 .procname = "accept_source_route",
4525 .data = &ipv6_devconf.accept_source_route,
4526 .maxlen = sizeof(int),
4527 .mode = 0644,
4528 .proc_handler = proc_dointvec,
4530 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
4532 .procname = "optimistic_dad",
4533 .data = &ipv6_devconf.optimistic_dad,
4534 .maxlen = sizeof(int),
4535 .mode = 0644,
4536 .proc_handler = proc_dointvec,
4539 #endif
4540 #ifdef CONFIG_IPV6_MROUTE
4542 .procname = "mc_forwarding",
4543 .data = &ipv6_devconf.mc_forwarding,
4544 .maxlen = sizeof(int),
4545 .mode = 0444,
4546 .proc_handler = proc_dointvec,
4548 #endif
4550 .procname = "disable_ipv6",
4551 .data = &ipv6_devconf.disable_ipv6,
4552 .maxlen = sizeof(int),
4553 .mode = 0644,
4554 .proc_handler = addrconf_sysctl_disable,
4557 .procname = "accept_dad",
4558 .data = &ipv6_devconf.accept_dad,
4559 .maxlen = sizeof(int),
4560 .mode = 0644,
4561 .proc_handler = proc_dointvec,
4564 .procname = "force_tllao",
4565 .data = &ipv6_devconf.force_tllao,
4566 .maxlen = sizeof(int),
4567 .mode = 0644,
4568 .proc_handler = proc_dointvec
4571 /* sentinel */
4576 static int __addrconf_sysctl_register(struct net *net, char *dev_name,
4577 struct inet6_dev *idev, struct ipv6_devconf *p)
4579 int i;
4580 struct addrconf_sysctl_table *t;
4581 char path[sizeof("net/ipv6/conf/") + IFNAMSIZ];
4583 t = kmemdup(&addrconf_sysctl, sizeof(*t), GFP_KERNEL);
4584 if (t == NULL)
4585 goto out;
4587 for (i = 0; t->addrconf_vars[i].data; i++) {
4588 t->addrconf_vars[i].data += (char *)p - (char *)&ipv6_devconf;
4589 t->addrconf_vars[i].extra1 = idev; /* embedded; no ref */
4590 t->addrconf_vars[i].extra2 = net;
4593 snprintf(path, sizeof(path), "net/ipv6/conf/%s", dev_name);
4595 t->sysctl_header = register_net_sysctl(net, path, t->addrconf_vars);
4596 if (t->sysctl_header == NULL)
4597 goto free;
4599 p->sysctl = t;
4600 return 0;
4602 free:
4603 kfree(t);
4604 out:
4605 return -ENOBUFS;
4608 static void __addrconf_sysctl_unregister(struct ipv6_devconf *p)
4610 struct addrconf_sysctl_table *t;
4612 if (p->sysctl == NULL)
4613 return;
4615 t = p->sysctl;
4616 p->sysctl = NULL;
4617 unregister_net_sysctl_table(t->sysctl_header);
4618 kfree(t);
4621 static void addrconf_sysctl_register(struct inet6_dev *idev)
4623 neigh_sysctl_register(idev->dev, idev->nd_parms, "ipv6",
4624 &ndisc_ifinfo_sysctl_change);
4625 __addrconf_sysctl_register(dev_net(idev->dev), idev->dev->name,
4626 idev, &idev->cnf);
4629 static void addrconf_sysctl_unregister(struct inet6_dev *idev)
4631 __addrconf_sysctl_unregister(&idev->cnf);
4632 neigh_sysctl_unregister(idev->nd_parms);
4636 #endif
4638 static int __net_init addrconf_init_net(struct net *net)
4640 int err;
4641 struct ipv6_devconf *all, *dflt;
4643 err = -ENOMEM;
4644 all = &ipv6_devconf;
4645 dflt = &ipv6_devconf_dflt;
4647 if (!net_eq(net, &init_net)) {
4648 all = kmemdup(all, sizeof(ipv6_devconf), GFP_KERNEL);
4649 if (all == NULL)
4650 goto err_alloc_all;
4652 dflt = kmemdup(dflt, sizeof(ipv6_devconf_dflt), GFP_KERNEL);
4653 if (dflt == NULL)
4654 goto err_alloc_dflt;
4655 } else {
4656 /* these will be inherited by all namespaces */
4657 dflt->autoconf = ipv6_defaults.autoconf;
4658 dflt->disable_ipv6 = ipv6_defaults.disable_ipv6;
4661 net->ipv6.devconf_all = all;
4662 net->ipv6.devconf_dflt = dflt;
4664 #ifdef CONFIG_SYSCTL
4665 err = __addrconf_sysctl_register(net, "all", NULL, all);
4666 if (err < 0)
4667 goto err_reg_all;
4669 err = __addrconf_sysctl_register(net, "default", NULL, dflt);
4670 if (err < 0)
4671 goto err_reg_dflt;
4672 #endif
4673 return 0;
4675 #ifdef CONFIG_SYSCTL
4676 err_reg_dflt:
4677 __addrconf_sysctl_unregister(all);
4678 err_reg_all:
4679 kfree(dflt);
4680 #endif
4681 err_alloc_dflt:
4682 kfree(all);
4683 err_alloc_all:
4684 return err;
4687 static void __net_exit addrconf_exit_net(struct net *net)
4689 #ifdef CONFIG_SYSCTL
4690 __addrconf_sysctl_unregister(net->ipv6.devconf_dflt);
4691 __addrconf_sysctl_unregister(net->ipv6.devconf_all);
4692 #endif
4693 if (!net_eq(net, &init_net)) {
4694 kfree(net->ipv6.devconf_dflt);
4695 kfree(net->ipv6.devconf_all);
4699 static struct pernet_operations addrconf_ops = {
4700 .init = addrconf_init_net,
4701 .exit = addrconf_exit_net,
4705 * Device notifier
4708 int register_inet6addr_notifier(struct notifier_block *nb)
4710 return atomic_notifier_chain_register(&inet6addr_chain, nb);
4712 EXPORT_SYMBOL(register_inet6addr_notifier);
4714 int unregister_inet6addr_notifier(struct notifier_block *nb)
4716 return atomic_notifier_chain_unregister(&inet6addr_chain, nb);
4718 EXPORT_SYMBOL(unregister_inet6addr_notifier);
4720 static struct rtnl_af_ops inet6_ops = {
4721 .family = AF_INET6,
4722 .fill_link_af = inet6_fill_link_af,
4723 .get_link_af_size = inet6_get_link_af_size,
4727 * Init / cleanup code
4730 int __init addrconf_init(void)
4732 int i, err;
4734 err = ipv6_addr_label_init();
4735 if (err < 0) {
4736 pr_crit("%s: cannot initialize default policy table: %d\n",
4737 __func__, err);
4738 goto out;
4741 err = register_pernet_subsys(&addrconf_ops);
4742 if (err < 0)
4743 goto out_addrlabel;
4745 /* The addrconf netdev notifier requires that loopback_dev
4746 * has it's ipv6 private information allocated and setup
4747 * before it can bring up and give link-local addresses
4748 * to other devices which are up.
4750 * Unfortunately, loopback_dev is not necessarily the first
4751 * entry in the global dev_base list of net devices. In fact,
4752 * it is likely to be the very last entry on that list.
4753 * So this causes the notifier registry below to try and
4754 * give link-local addresses to all devices besides loopback_dev
4755 * first, then loopback_dev, which cases all the non-loopback_dev
4756 * devices to fail to get a link-local address.
4758 * So, as a temporary fix, allocate the ipv6 structure for
4759 * loopback_dev first by hand.
4760 * Longer term, all of the dependencies ipv6 has upon the loopback
4761 * device and it being up should be removed.
4763 rtnl_lock();
4764 if (!ipv6_add_dev(init_net.loopback_dev))
4765 err = -ENOMEM;
4766 rtnl_unlock();
4767 if (err)
4768 goto errlo;
4770 for (i = 0; i < IN6_ADDR_HSIZE; i++)
4771 INIT_HLIST_HEAD(&inet6_addr_lst[i]);
4773 register_netdevice_notifier(&ipv6_dev_notf);
4775 addrconf_verify(0);
4777 err = rtnl_af_register(&inet6_ops);
4778 if (err < 0)
4779 goto errout_af;
4781 err = __rtnl_register(PF_INET6, RTM_GETLINK, NULL, inet6_dump_ifinfo,
4782 NULL);
4783 if (err < 0)
4784 goto errout;
4786 /* Only the first call to __rtnl_register can fail */
4787 __rtnl_register(PF_INET6, RTM_NEWADDR, inet6_rtm_newaddr, NULL, NULL);
4788 __rtnl_register(PF_INET6, RTM_DELADDR, inet6_rtm_deladdr, NULL, NULL);
4789 __rtnl_register(PF_INET6, RTM_GETADDR, inet6_rtm_getaddr,
4790 inet6_dump_ifaddr, NULL);
4791 __rtnl_register(PF_INET6, RTM_GETMULTICAST, NULL,
4792 inet6_dump_ifmcaddr, NULL);
4793 __rtnl_register(PF_INET6, RTM_GETANYCAST, NULL,
4794 inet6_dump_ifacaddr, NULL);
4796 ipv6_addr_label_rtnl_register();
4798 return 0;
4799 errout:
4800 rtnl_af_unregister(&inet6_ops);
4801 errout_af:
4802 unregister_netdevice_notifier(&ipv6_dev_notf);
4803 errlo:
4804 unregister_pernet_subsys(&addrconf_ops);
4805 out_addrlabel:
4806 ipv6_addr_label_cleanup();
4807 out:
4808 return err;
4811 void addrconf_cleanup(void)
4813 struct net_device *dev;
4814 int i;
4816 unregister_netdevice_notifier(&ipv6_dev_notf);
4817 unregister_pernet_subsys(&addrconf_ops);
4818 ipv6_addr_label_cleanup();
4820 rtnl_lock();
4822 __rtnl_af_unregister(&inet6_ops);
4824 /* clean dev list */
4825 for_each_netdev(&init_net, dev) {
4826 if (__in6_dev_get(dev) == NULL)
4827 continue;
4828 addrconf_ifdown(dev, 1);
4830 addrconf_ifdown(init_net.loopback_dev, 2);
4833 * Check hash table.
4835 spin_lock_bh(&addrconf_hash_lock);
4836 for (i = 0; i < IN6_ADDR_HSIZE; i++)
4837 WARN_ON(!hlist_empty(&inet6_addr_lst[i]));
4838 spin_unlock_bh(&addrconf_hash_lock);
4840 del_timer(&addr_chk_timer);
4841 rtnl_unlock();