?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
ipv6: add special mode forwarding=2 to send RS while
[linux-2.6/linux-acpi-2.6/ibm-acpi-2.6.git] / net / ipv6 / addrconf.c
blob5d59520ae162af8b665116f70b027780327af61b
1 /*
2 * IPv6 Address [auto]configuration
3 * Linux INET6 implementation
4 *
5 * Authors:
6 * Pedro Roque <roque@di.fc.ul.pt>
7 * Alexey Kuznetsov <kuznet@ms2.inr.ac.ru>
8 *
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.
13 */
14
15 /*
16 * Changes:
17 *
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.
39 */
40
41 #include <linux/errno.h>
42 #include <linux/types.h>
43 #include <linux/kernel.h>
44 #include <linux/socket.h>
45 #include <linux/sockios.h>
46 #include <linux/net.h>
47 #include <linux/in6.h>
48 #include <linux/netdevice.h>
49 #include <linux/if_addr.h>
50 #include <linux/if_arp.h>
51 #include <linux/if_arcnet.h>
52 #include <linux/if_infiniband.h>
53 #include <linux/route.h>
54 #include <linux/inetdevice.h>
55 #include <linux/init.h>
56 #include <linux/slab.h>
57 #ifdef CONFIG_SYSCTL
58 #include <linux/sysctl.h>
59 #endif
60 #include <linux/capability.h>
61 #include <linux/delay.h>
62 #include <linux/notifier.h>
63 #include <linux/string.h>
64
65 #include <net/net_namespace.h>
66 #include <net/sock.h>
67 #include <net/snmp.h>
68
69 #include <net/ipv6.h>
70 #include <net/protocol.h>
71 #include <net/ndisc.h>
72 #include <net/ip6_route.h>
73 #include <net/addrconf.h>
74 #include <net/tcp.h>
75 #include <net/ip.h>
76 #include <net/netlink.h>
77 #include <net/pkt_sched.h>
78 #include <linux/if_tunnel.h>
79 #include <linux/rtnetlink.h>
80
81 #ifdef CONFIG_IPV6_PRIVACY
82 #include <linux/random.h>
83 #endif
84
85 #include <linux/uaccess.h>
86 #include <asm/unaligned.h>
87
88 #include <linux/proc_fs.h>
89 #include <linux/seq_file.h>
90
91 /* Set to 3 to get tracing... */
92 #define ACONF_DEBUG 2
93
94 #if ACONF_DEBUG >= 3
95 #define ADBG(x) printk x
96 #else
97 #define ADBG(x)
98 #endif
99
100 #define INFINITY_LIFE_TIME 0xFFFFFFFF
101 #define TIME_DELTA(a, b) ((unsigned long)((long)(a) - (long)(b)))
102
103 #define ADDRCONF_TIMER_FUZZ_MINUS (HZ > 50 ? HZ/50 : 1)
104 #define ADDRCONF_TIMER_FUZZ (HZ / 4)
105 #define ADDRCONF_TIMER_FUZZ_MAX (HZ)
106
107 #ifdef CONFIG_SYSCTL
108 static void addrconf_sysctl_register(struct inet6_dev *idev);
109 static void addrconf_sysctl_unregister(struct inet6_dev *idev);
110 #else
111 static inline void addrconf_sysctl_register(struct inet6_dev *idev)
112 {
113 }
114
115 static inline void addrconf_sysctl_unregister(struct inet6_dev *idev)
116 {
117 }
118 #endif
119
120 #ifdef CONFIG_IPV6_PRIVACY
121 static int __ipv6_regen_rndid(struct inet6_dev *idev);
122 static int __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr);
123 static void ipv6_regen_rndid(unsigned long data);
124
125 static int desync_factor = MAX_DESYNC_FACTOR * HZ;
126 #endif
127
128 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev);
129 static int ipv6_count_addresses(struct inet6_dev *idev);
130
131 /*
132 * Configured unicast address hash table
133 */
134 static struct hlist_head inet6_addr_lst[IN6_ADDR_HSIZE];
135 static DEFINE_SPINLOCK(addrconf_hash_lock);
136
137 static void addrconf_verify(unsigned long);
138
139 static DEFINE_TIMER(addr_chk_timer, addrconf_verify, 0, 0);
140 static DEFINE_SPINLOCK(addrconf_verify_lock);
141
142 static void addrconf_join_anycast(struct inet6_ifaddr *ifp);
143 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp);
144
145 static void addrconf_type_change(struct net_device *dev,
146 unsigned long event);
147 static int addrconf_ifdown(struct net_device *dev, int how);
148
149 static void addrconf_dad_start(struct inet6_ifaddr *ifp, u32 flags);
150 static void addrconf_dad_timer(unsigned long data);
151 static void addrconf_dad_completed(struct inet6_ifaddr *ifp);
152 static void addrconf_dad_run(struct inet6_dev *idev);
153 static void addrconf_rs_timer(unsigned long data);
154 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
155 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
156
157 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
158 struct prefix_info *pinfo);
159 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
160 struct net_device *dev);
161
162 static ATOMIC_NOTIFIER_HEAD(inet6addr_chain);
163
164 static struct ipv6_devconf ipv6_devconf __read_mostly = {
165 .forwarding = 0,
166 .hop_limit = IPV6_DEFAULT_HOPLIMIT,
167 .mtu6 = IPV6_MIN_MTU,
168 .accept_ra = 1,
169 .accept_redirects = 1,
170 .autoconf = 1,
171 .force_mld_version = 0,
172 .dad_transmits = 1,
173 .rtr_solicits = MAX_RTR_SOLICITATIONS,
174 .rtr_solicit_interval = RTR_SOLICITATION_INTERVAL,
175 .rtr_solicit_delay = MAX_RTR_SOLICITATION_DELAY,
176 #ifdef CONFIG_IPV6_PRIVACY
177 .use_tempaddr = 0,
178 .temp_valid_lft = TEMP_VALID_LIFETIME,
179 .temp_prefered_lft = TEMP_PREFERRED_LIFETIME,
180 .regen_max_retry = REGEN_MAX_RETRY,
181 .max_desync_factor = MAX_DESYNC_FACTOR,
182 #endif
183 .max_addresses = IPV6_MAX_ADDRESSES,
184 .accept_ra_defrtr = 1,
185 .accept_ra_pinfo = 1,
186 #ifdef CONFIG_IPV6_ROUTER_PREF
187 .accept_ra_rtr_pref = 1,
188 .rtr_probe_interval = 60 * HZ,
189 #ifdef CONFIG_IPV6_ROUTE_INFO
190 .accept_ra_rt_info_max_plen = 0,
191 #endif
192 #endif
193 .proxy_ndp = 0,
194 .accept_source_route = 0, /* we do not accept RH0 by default. */
195 .disable_ipv6 = 0,
196 .accept_dad = 1,
197 };
198
199 static struct ipv6_devconf ipv6_devconf_dflt __read_mostly = {
200 .forwarding = 0,
201 .hop_limit = IPV6_DEFAULT_HOPLIMIT,
202 .mtu6 = IPV6_MIN_MTU,
203 .accept_ra = 1,
204 .accept_redirects = 1,
205 .autoconf = 1,
206 .dad_transmits = 1,
207 .rtr_solicits = MAX_RTR_SOLICITATIONS,
208 .rtr_solicit_interval = RTR_SOLICITATION_INTERVAL,
209 .rtr_solicit_delay = MAX_RTR_SOLICITATION_DELAY,
210 #ifdef CONFIG_IPV6_PRIVACY
211 .use_tempaddr = 0,
212 .temp_valid_lft = TEMP_VALID_LIFETIME,
213 .temp_prefered_lft = TEMP_PREFERRED_LIFETIME,
214 .regen_max_retry = REGEN_MAX_RETRY,
215 .max_desync_factor = MAX_DESYNC_FACTOR,
216 #endif
217 .max_addresses = IPV6_MAX_ADDRESSES,
218 .accept_ra_defrtr = 1,
219 .accept_ra_pinfo = 1,
220 #ifdef CONFIG_IPV6_ROUTER_PREF
221 .accept_ra_rtr_pref = 1,
222 .rtr_probe_interval = 60 * HZ,
223 #ifdef CONFIG_IPV6_ROUTE_INFO
224 .accept_ra_rt_info_max_plen = 0,
225 #endif
226 #endif
227 .proxy_ndp = 0,
228 .accept_source_route = 0, /* we do not accept RH0 by default. */
229 .disable_ipv6 = 0,
230 .accept_dad = 1,
231 };
232
233 /* IPv6 Wildcard Address and Loopback Address defined by RFC2553 */
234 const struct in6_addr in6addr_any = IN6ADDR_ANY_INIT;
235 const struct in6_addr in6addr_loopback = IN6ADDR_LOOPBACK_INIT;
236 const struct in6_addr in6addr_linklocal_allnodes = IN6ADDR_LINKLOCAL_ALLNODES_INIT;
237 const struct in6_addr in6addr_linklocal_allrouters = IN6ADDR_LINKLOCAL_ALLROUTERS_INIT;
238
239 /* Check if a valid qdisc is available */
240 static inline bool addrconf_qdisc_ok(const struct net_device *dev)
241 {
242 return !qdisc_tx_is_noop(dev);
243 }
244
245 /* Check if a route is valid prefix route */
246 static inline int addrconf_is_prefix_route(const struct rt6_info *rt)
247 {
248 return ((rt->rt6i_flags & (RTF_GATEWAY | RTF_DEFAULT)) == 0);
249 }
250
251 static void addrconf_del_timer(struct inet6_ifaddr *ifp)
252 {
253 if (del_timer(&ifp->timer))
254 __in6_ifa_put(ifp);
255 }
256
257 enum addrconf_timer_t {
258 AC_NONE,
259 AC_DAD,
260 AC_RS,
261 };
262
263 static void addrconf_mod_timer(struct inet6_ifaddr *ifp,
264 enum addrconf_timer_t what,
265 unsigned long when)
266 {
267 if (!del_timer(&ifp->timer))
268 in6_ifa_hold(ifp);
269
270 switch (what) {
271 case AC_DAD:
272 ifp->timer.function = addrconf_dad_timer;
273 break;
274 case AC_RS:
275 ifp->timer.function = addrconf_rs_timer;
276 break;
277 default:
278 break;
279 }
280 ifp->timer.expires = jiffies + when;
281 add_timer(&ifp->timer);
282 }
283
284 static int snmp6_alloc_dev(struct inet6_dev *idev)
285 {
286 if (snmp_mib_init((void __percpu **)idev->stats.ipv6,
287 sizeof(struct ipstats_mib)) < 0)
288 goto err_ip;
289 if (snmp_mib_init((void __percpu **)idev->stats.icmpv6,
290 sizeof(struct icmpv6_mib)) < 0)
291 goto err_icmp;
292 if (snmp_mib_init((void __percpu **)idev->stats.icmpv6msg,
293 sizeof(struct icmpv6msg_mib)) < 0)
294 goto err_icmpmsg;
295
296 return 0;
297
298 err_icmpmsg:
299 snmp_mib_free((void __percpu **)idev->stats.icmpv6);
300 err_icmp:
301 snmp_mib_free((void __percpu **)idev->stats.ipv6);
302 err_ip:
303 return -ENOMEM;
304 }
305
306 static void snmp6_free_dev(struct inet6_dev *idev)
307 {
308 snmp_mib_free((void __percpu **)idev->stats.icmpv6msg);
309 snmp_mib_free((void __percpu **)idev->stats.icmpv6);
310 snmp_mib_free((void __percpu **)idev->stats.ipv6);
311 }
312
313 /* Nobody refers to this device, we may destroy it. */
314
315 static void in6_dev_finish_destroy_rcu(struct rcu_head *head)
316 {
317 struct inet6_dev *idev = container_of(head, struct inet6_dev, rcu);
318 kfree(idev);
319 }
320
321 void in6_dev_finish_destroy(struct inet6_dev *idev)
322 {
323 struct net_device *dev = idev->dev;
324
325 WARN_ON(!list_empty(&idev->addr_list));
326 WARN_ON(idev->mc_list != NULL);
327
328 #ifdef NET_REFCNT_DEBUG
329 printk(KERN_DEBUG "in6_dev_finish_destroy: %s\n", dev ? dev->name : "NIL");
330 #endif
331 dev_put(dev);
332 if (!idev->dead) {
333 pr_warning("Freeing alive inet6 device %p\n", idev);
334 return;
335 }
336 snmp6_free_dev(idev);
337 call_rcu(&idev->rcu, in6_dev_finish_destroy_rcu);
338 }
339
340 EXPORT_SYMBOL(in6_dev_finish_destroy);
341
342 static struct inet6_dev * ipv6_add_dev(struct net_device *dev)
343 {
344 struct inet6_dev *ndev;
345
346 ASSERT_RTNL();
347
348 if (dev->mtu < IPV6_MIN_MTU)
349 return NULL;
350
351 ndev = kzalloc(sizeof(struct inet6_dev), GFP_KERNEL);
352
353 if (ndev == NULL)
354 return NULL;
355
356 rwlock_init(&ndev->lock);
357 ndev->dev = dev;
358 INIT_LIST_HEAD(&ndev->addr_list);
359
360 memcpy(&ndev->cnf, dev_net(dev)->ipv6.devconf_dflt, sizeof(ndev->cnf));
361 ndev->cnf.mtu6 = dev->mtu;
362 ndev->cnf.sysctl = NULL;
363 ndev->nd_parms = neigh_parms_alloc(dev, &nd_tbl);
364 if (ndev->nd_parms == NULL) {
365 kfree(ndev);
366 return NULL;
367 }
368 if (ndev->cnf.forwarding)
369 dev_disable_lro(dev);
370 /* We refer to the device */
371 dev_hold(dev);
372
373 if (snmp6_alloc_dev(ndev) < 0) {
374 ADBG((KERN_WARNING
375 "%s(): cannot allocate memory for statistics; dev=%s.\n",
376 __func__, dev->name));
377 neigh_parms_release(&nd_tbl, ndev->nd_parms);
378 ndev->dead = 1;
379 in6_dev_finish_destroy(ndev);
380 return NULL;
381 }
382
383 if (snmp6_register_dev(ndev) < 0) {
384 ADBG((KERN_WARNING
385 "%s(): cannot create /proc/net/dev_snmp6/%s\n",
386 __func__, dev->name));
387 neigh_parms_release(&nd_tbl, ndev->nd_parms);
388 ndev->dead = 1;
389 in6_dev_finish_destroy(ndev);
390 return NULL;
391 }
392
393 /* One reference from device. We must do this before
394 * we invoke __ipv6_regen_rndid().
395 */
396 in6_dev_hold(ndev);
397
398 if (dev->flags & (IFF_NOARP | IFF_LOOPBACK))
399 ndev->cnf.accept_dad = -1;
400
401 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
402 if (dev->type == ARPHRD_SIT && (dev->priv_flags & IFF_ISATAP)) {
403 printk(KERN_INFO
404 "%s: Disabled Multicast RS\n",
405 dev->name);
406 ndev->cnf.rtr_solicits = 0;
407 }
408 #endif
409
410 #ifdef CONFIG_IPV6_PRIVACY
411 INIT_LIST_HEAD(&ndev->tempaddr_list);
412 setup_timer(&ndev->regen_timer, ipv6_regen_rndid, (unsigned long)ndev);
413 if ((dev->flags&IFF_LOOPBACK) ||
414 dev->type == ARPHRD_TUNNEL ||
415 dev->type == ARPHRD_TUNNEL6 ||
416 dev->type == ARPHRD_SIT ||
417 dev->type == ARPHRD_NONE) {
418 ndev->cnf.use_tempaddr = -1;
419 } else {
420 in6_dev_hold(ndev);
421 ipv6_regen_rndid((unsigned long) ndev);
422 }
423 #endif
424
425 if (netif_running(dev) && addrconf_qdisc_ok(dev))
426 ndev->if_flags |= IF_READY;
427
428 ipv6_mc_init_dev(ndev);
429 ndev->tstamp = jiffies;
430 addrconf_sysctl_register(ndev);
431 /* protected by rtnl_lock */
432 rcu_assign_pointer(dev->ip6_ptr, ndev);
433
434 /* Join all-node multicast group */
435 ipv6_dev_mc_inc(dev, &in6addr_linklocal_allnodes);
436
437 return ndev;
438 }
439
440 static struct inet6_dev * ipv6_find_idev(struct net_device *dev)
441 {
442 struct inet6_dev *idev;
443
444 ASSERT_RTNL();
445
446 idev = __in6_dev_get(dev);
447 if (!idev) {
448 idev = ipv6_add_dev(dev);
449 if (!idev)
450 return NULL;
451 }
452
453 if (dev->flags&IFF_UP)
454 ipv6_mc_up(idev);
455 return idev;
456 }
457
458 #ifdef CONFIG_SYSCTL
459 static void dev_forward_change(struct inet6_dev *idev)
460 {
461 struct net_device *dev;
462 struct inet6_ifaddr *ifa;
463
464 if (!idev)
465 return;
466 dev = idev->dev;
467 if (idev->cnf.forwarding)
468 dev_disable_lro(dev);
469 if (dev && (dev->flags & IFF_MULTICAST)) {
470 if (idev->cnf.forwarding)
471 ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
472 else
473 ipv6_dev_mc_dec(dev, &in6addr_linklocal_allrouters);
474 }
475
476 list_for_each_entry(ifa, &idev->addr_list, if_list) {
477 if (ifa->flags&IFA_F_TENTATIVE)
478 continue;
479 if (idev->cnf.forwarding)
480 addrconf_join_anycast(ifa);
481 else
482 addrconf_leave_anycast(ifa);
483 }
484 }
485
486
487 static void addrconf_forward_change(struct net *net, __s32 newf)
488 {
489 struct net_device *dev;
490 struct inet6_dev *idev;
491
492 rcu_read_lock();
493 for_each_netdev_rcu(net, dev) {
494 idev = __in6_dev_get(dev);
495 if (idev) {
496 int changed = (!idev->cnf.forwarding) ^ (!newf);
497 idev->cnf.forwarding = newf;
498 if (changed)
499 dev_forward_change(idev);
500 }
501 }
502 rcu_read_unlock();
503 }
504
505 static int addrconf_fixup_forwarding(struct ctl_table *table, int *p, int old)
506 {
507 struct net *net;
508
509 net = (struct net *)table->extra2;
510 if (p == &net->ipv6.devconf_dflt->forwarding)
511 return 0;
512
513 if (!rtnl_trylock()) {
514 /* Restore the original values before restarting */
515 *p = old;
516 return restart_syscall();
517 }
518
519 if (p == &net->ipv6.devconf_all->forwarding) {
520 __s32 newf = net->ipv6.devconf_all->forwarding;
521 net->ipv6.devconf_dflt->forwarding = newf;
522 addrconf_forward_change(net, newf);
523 } else if ((!*p) ^ (!old))
524 dev_forward_change((struct inet6_dev *)table->extra1);
525 rtnl_unlock();
526
527 if (*p)
528 rt6_purge_dflt_routers(net);
529 return 1;
530 }
531 #endif
532
533 static void inet6_ifa_finish_destroy_rcu(struct rcu_head *head)
534 {
535 struct inet6_ifaddr *ifp = container_of(head, struct inet6_ifaddr, rcu);
536 kfree(ifp);
537 }
538
539 /* Nobody refers to this ifaddr, destroy it */
540 void inet6_ifa_finish_destroy(struct inet6_ifaddr *ifp)
541 {
542 WARN_ON(!hlist_unhashed(&ifp->addr_lst));
543
544 #ifdef NET_REFCNT_DEBUG
545 printk(KERN_DEBUG "inet6_ifa_finish_destroy\n");
546 #endif
547
548 in6_dev_put(ifp->idev);
549
550 if (del_timer(&ifp->timer))
551 pr_notice("Timer is still running, when freeing ifa=%p\n", ifp);
552
553 if (ifp->state != INET6_IFADDR_STATE_DEAD) {
554 pr_warning("Freeing alive inet6 address %p\n", ifp);
555 return;
556 }
557 dst_release(&ifp->rt->u.dst);
558
559 call_rcu(&ifp->rcu, inet6_ifa_finish_destroy_rcu);
560 }
561
562 static void
563 ipv6_link_dev_addr(struct inet6_dev *idev, struct inet6_ifaddr *ifp)
564 {
565 struct list_head *p;
566 int ifp_scope = ipv6_addr_src_scope(&ifp->addr);
567
568 /*
569 * Each device address list is sorted in order of scope -
570 * global before linklocal.
571 */
572 list_for_each(p, &idev->addr_list) {
573 struct inet6_ifaddr *ifa
574 = list_entry(p, struct inet6_ifaddr, if_list);
575 if (ifp_scope >= ipv6_addr_src_scope(&ifa->addr))
576 break;
577 }
578
579 list_add_tail(&ifp->if_list, p);
580 }
581
582 static u32 ipv6_addr_hash(const struct in6_addr *addr)
583 {
584 /*
585 * We perform the hash function over the last 64 bits of the address
586 * This will include the IEEE address token on links that support it.
587 */
588 return jhash_2words((__force u32)addr->s6_addr32[2],
589 (__force u32)addr->s6_addr32[3], 0)
590 & (IN6_ADDR_HSIZE - 1);
591 }
592
593 /* On success it returns ifp with increased reference count */
594
595 static struct inet6_ifaddr *
596 ipv6_add_addr(struct inet6_dev *idev, const struct in6_addr *addr, int pfxlen,
597 int scope, u32 flags)
598 {
599 struct inet6_ifaddr *ifa = NULL;
600 struct rt6_info *rt;
601 unsigned int hash;
602 int err = 0;
603 int addr_type = ipv6_addr_type(addr);
604
605 if (addr_type == IPV6_ADDR_ANY ||
606 addr_type & IPV6_ADDR_MULTICAST ||
607 (!(idev->dev->flags & IFF_LOOPBACK) &&
608 addr_type & IPV6_ADDR_LOOPBACK))
609 return ERR_PTR(-EADDRNOTAVAIL);
610
611 rcu_read_lock_bh();
612 if (idev->dead) {
613 err = -ENODEV; /*XXX*/
614 goto out2;
615 }
616
617 if (idev->cnf.disable_ipv6) {
618 err = -EACCES;
619 goto out2;
620 }
621
622 spin_lock(&addrconf_hash_lock);
623
624 /* Ignore adding duplicate addresses on an interface */
625 if (ipv6_chk_same_addr(dev_net(idev->dev), addr, idev->dev)) {
626 ADBG(("ipv6_add_addr: already assigned\n"));
627 err = -EEXIST;
628 goto out;
629 }
630
631 ifa = kzalloc(sizeof(struct inet6_ifaddr), GFP_ATOMIC);
632
633 if (ifa == NULL) {
634 ADBG(("ipv6_add_addr: malloc failed\n"));
635 err = -ENOBUFS;
636 goto out;
637 }
638
639 rt = addrconf_dst_alloc(idev, addr, 0);
640 if (IS_ERR(rt)) {
641 err = PTR_ERR(rt);
642 goto out;
643 }
644
645 ipv6_addr_copy(&ifa->addr, addr);
646
647 spin_lock_init(&ifa->lock);
648 spin_lock_init(&ifa->state_lock);
649 init_timer(&ifa->timer);
650 INIT_HLIST_NODE(&ifa->addr_lst);
651 ifa->timer.data = (unsigned long) ifa;
652 ifa->scope = scope;
653 ifa->prefix_len = pfxlen;
654 ifa->flags = flags | IFA_F_TENTATIVE;
655 ifa->cstamp = ifa->tstamp = jiffies;
656
657 ifa->rt = rt;
658
659 /*
660 * part one of RFC 4429, section 3.3
661 * We should not configure an address as
662 * optimistic if we do not yet know the link
663 * layer address of our nexhop router
664 */
665
666 if (rt->rt6i_nexthop == NULL)
667 ifa->flags &= ~IFA_F_OPTIMISTIC;
668
669 ifa->idev = idev;
670 in6_dev_hold(idev);
671 /* For caller */
672 in6_ifa_hold(ifa);
673
674 /* Add to big hash table */
675 hash = ipv6_addr_hash(addr);
676
677 hlist_add_head_rcu(&ifa->addr_lst, &inet6_addr_lst[hash]);
678 spin_unlock(&addrconf_hash_lock);
679
680 write_lock(&idev->lock);
681 /* Add to inet6_dev unicast addr list. */
682 ipv6_link_dev_addr(idev, ifa);
683
684 #ifdef CONFIG_IPV6_PRIVACY
685 if (ifa->flags&IFA_F_TEMPORARY) {
686 list_add(&ifa->tmp_list, &idev->tempaddr_list);
687 in6_ifa_hold(ifa);
688 }
689 #endif
690
691 in6_ifa_hold(ifa);
692 write_unlock(&idev->lock);
693 out2:
694 rcu_read_unlock_bh();
695
696 if (likely(err == 0))
697 atomic_notifier_call_chain(&inet6addr_chain, NETDEV_UP, ifa);
698 else {
699 kfree(ifa);
700 ifa = ERR_PTR(err);
701 }
702
703 return ifa;
704 out:
705 spin_unlock(&addrconf_hash_lock);
706 goto out2;
707 }
708
709 /* This function wants to get referenced ifp and releases it before return */
710
711 static void ipv6_del_addr(struct inet6_ifaddr *ifp)
712 {
713 struct inet6_ifaddr *ifa, *ifn;
714 struct inet6_dev *idev = ifp->idev;
715 int state;
716 int hash;
717 int deleted = 0, onlink = 0;
718 unsigned long expires = jiffies;
719
720 hash = ipv6_addr_hash(&ifp->addr);
721
722 spin_lock_bh(&ifp->state_lock);
723 state = ifp->state;
724 ifp->state = INET6_IFADDR_STATE_DEAD;
725 spin_unlock_bh(&ifp->state_lock);
726
727 if (state == INET6_IFADDR_STATE_DEAD)
728 goto out;
729
730 spin_lock_bh(&addrconf_hash_lock);
731 hlist_del_init_rcu(&ifp->addr_lst);
732 spin_unlock_bh(&addrconf_hash_lock);
733
734 write_lock_bh(&idev->lock);
735 #ifdef CONFIG_IPV6_PRIVACY
736 if (ifp->flags&IFA_F_TEMPORARY) {
737 list_del(&ifp->tmp_list);
738 if (ifp->ifpub) {
739 in6_ifa_put(ifp->ifpub);
740 ifp->ifpub = NULL;
741 }
742 __in6_ifa_put(ifp);
743 }
744 #endif
745
746 list_for_each_entry_safe(ifa, ifn, &idev->addr_list, if_list) {
747 if (ifa == ifp) {
748 list_del_init(&ifp->if_list);
749 __in6_ifa_put(ifp);
750
751 if (!(ifp->flags & IFA_F_PERMANENT) || onlink > 0)
752 break;
753 deleted = 1;
754 continue;
755 } else if (ifp->flags & IFA_F_PERMANENT) {
756 if (ipv6_prefix_equal(&ifa->addr, &ifp->addr,
757 ifp->prefix_len)) {
758 if (ifa->flags & IFA_F_PERMANENT) {
759 onlink = 1;
760 if (deleted)
761 break;
762 } else {
763 unsigned long lifetime;
764
765 if (!onlink)
766 onlink = -1;
767
768 spin_lock(&ifa->lock);
769
770 lifetime = addrconf_timeout_fixup(ifa->valid_lft, HZ);
771 /*
772 * Note: Because this address is
773 * not permanent, lifetime <
774 * LONG_MAX / HZ here.
775 */
776 if (time_before(expires,
777 ifa->tstamp + lifetime * HZ))
778 expires = ifa->tstamp + lifetime * HZ;
779 spin_unlock(&ifa->lock);
780 }
781 }
782 }
783 }
784 write_unlock_bh(&idev->lock);
785
786 addrconf_del_timer(ifp);
787
788 ipv6_ifa_notify(RTM_DELADDR, ifp);
789
790 atomic_notifier_call_chain(&inet6addr_chain, NETDEV_DOWN, ifp);
791
792 /*
793 * Purge or update corresponding prefix
794 *
795 * 1) we don't purge prefix here if address was not permanent.
796 * prefix is managed by its own lifetime.
797 * 2) if there're no addresses, delete prefix.
798 * 3) if there're still other permanent address(es),
799 * corresponding prefix is still permanent.
800 * 4) otherwise, update prefix lifetime to the
801 * longest valid lifetime among the corresponding
802 * addresses on the device.
803 * Note: subsequent RA will update lifetime.
804 *
805 * --yoshfuji
806 */
807 if ((ifp->flags & IFA_F_PERMANENT) && onlink < 1) {
808 struct in6_addr prefix;
809 struct rt6_info *rt;
810 struct net *net = dev_net(ifp->idev->dev);
811 ipv6_addr_prefix(&prefix, &ifp->addr, ifp->prefix_len);
812 rt = rt6_lookup(net, &prefix, NULL, ifp->idev->dev->ifindex, 1);
813
814 if (rt && addrconf_is_prefix_route(rt)) {
815 if (onlink == 0) {
816 ip6_del_rt(rt);
817 rt = NULL;
818 } else if (!(rt->rt6i_flags & RTF_EXPIRES)) {
819 rt->rt6i_expires = expires;
820 rt->rt6i_flags |= RTF_EXPIRES;
821 }
822 }
823 dst_release(&rt->u.dst);
824 }
825
826 out:
827 in6_ifa_put(ifp);
828 }
829
830 #ifdef CONFIG_IPV6_PRIVACY
831 static int ipv6_create_tempaddr(struct inet6_ifaddr *ifp, struct inet6_ifaddr *ift)
832 {
833 struct inet6_dev *idev = ifp->idev;
834 struct in6_addr addr, *tmpaddr;
835 unsigned long tmp_prefered_lft, tmp_valid_lft, tmp_cstamp, tmp_tstamp;
836 unsigned long regen_advance;
837 int tmp_plen;
838 int ret = 0;
839 int max_addresses;
840 u32 addr_flags;
841
842 write_lock(&idev->lock);
843 if (ift) {
844 spin_lock_bh(&ift->lock);
845 memcpy(&addr.s6_addr[8], &ift->addr.s6_addr[8], 8);
846 spin_unlock_bh(&ift->lock);
847 tmpaddr = &addr;
848 } else {
849 tmpaddr = NULL;
850 }
851 retry:
852 in6_dev_hold(idev);
853 if (idev->cnf.use_tempaddr <= 0) {
854 write_unlock(&idev->lock);
855 printk(KERN_INFO
856 "ipv6_create_tempaddr(): use_tempaddr is disabled.\n");
857 in6_dev_put(idev);
858 ret = -1;
859 goto out;
860 }
861 spin_lock_bh(&ifp->lock);
862 if (ifp->regen_count++ >= idev->cnf.regen_max_retry) {
863 idev->cnf.use_tempaddr = -1; /*XXX*/
864 spin_unlock_bh(&ifp->lock);
865 write_unlock(&idev->lock);
866 printk(KERN_WARNING
867 "ipv6_create_tempaddr(): regeneration time exceeded. disabled temporary address support.\n");
868 in6_dev_put(idev);
869 ret = -1;
870 goto out;
871 }
872 in6_ifa_hold(ifp);
873 memcpy(addr.s6_addr, ifp->addr.s6_addr, 8);
874 if (__ipv6_try_regen_rndid(idev, tmpaddr) < 0) {
875 spin_unlock_bh(&ifp->lock);
876 write_unlock(&idev->lock);
877 printk(KERN_WARNING
878 "ipv6_create_tempaddr(): regeneration of randomized interface id failed.\n");
879 in6_ifa_put(ifp);
880 in6_dev_put(idev);
881 ret = -1;
882 goto out;
883 }
884 memcpy(&addr.s6_addr[8], idev->rndid, 8);
885 tmp_valid_lft = min_t(__u32,
886 ifp->valid_lft,
887 idev->cnf.temp_valid_lft);
888 tmp_prefered_lft = min_t(__u32,
889 ifp->prefered_lft,
890 idev->cnf.temp_prefered_lft - desync_factor / HZ);
891 tmp_plen = ifp->prefix_len;
892 max_addresses = idev->cnf.max_addresses;
893 tmp_cstamp = ifp->cstamp;
894 tmp_tstamp = ifp->tstamp;
895 spin_unlock_bh(&ifp->lock);
896
897 regen_advance = idev->cnf.regen_max_retry *
898 idev->cnf.dad_transmits *
899 idev->nd_parms->retrans_time / HZ;
900 write_unlock(&idev->lock);
901
902 /* A temporary address is created only if this calculated Preferred
903 * Lifetime is greater than REGEN_ADVANCE time units. In particular,
904 * an implementation must not create a temporary address with a zero
905 * Preferred Lifetime.
906 */
907 if (tmp_prefered_lft <= regen_advance) {
908 in6_ifa_put(ifp);
909 in6_dev_put(idev);
910 ret = -1;
911 goto out;
912 }
913
914 addr_flags = IFA_F_TEMPORARY;
915 /* set in addrconf_prefix_rcv() */
916 if (ifp->flags & IFA_F_OPTIMISTIC)
917 addr_flags |= IFA_F_OPTIMISTIC;
918
919 ift = !max_addresses ||
920 ipv6_count_addresses(idev) < max_addresses ?
921 ipv6_add_addr(idev, &addr, tmp_plen,
922 ipv6_addr_type(&addr)&IPV6_ADDR_SCOPE_MASK,
923 addr_flags) : NULL;
924 if (!ift || IS_ERR(ift)) {
925 in6_ifa_put(ifp);
926 in6_dev_put(idev);
927 printk(KERN_INFO
928 "ipv6_create_tempaddr(): retry temporary address regeneration.\n");
929 tmpaddr = &addr;
930 write_lock(&idev->lock);
931 goto retry;
932 }
933
934 spin_lock_bh(&ift->lock);
935 ift->ifpub = ifp;
936 ift->valid_lft = tmp_valid_lft;
937 ift->prefered_lft = tmp_prefered_lft;
938 ift->cstamp = tmp_cstamp;
939 ift->tstamp = tmp_tstamp;
940 spin_unlock_bh(&ift->lock);
941
942 addrconf_dad_start(ift, 0);
943 in6_ifa_put(ift);
944 in6_dev_put(idev);
945 out:
946 return ret;
947 }
948 #endif
949
950 /*
951 * Choose an appropriate source address (RFC3484)
952 */
953 enum {
954 IPV6_SADDR_RULE_INIT = 0,
955 IPV6_SADDR_RULE_LOCAL,
956 IPV6_SADDR_RULE_SCOPE,
957 IPV6_SADDR_RULE_PREFERRED,
958 #ifdef CONFIG_IPV6_MIP6
959 IPV6_SADDR_RULE_HOA,
960 #endif
961 IPV6_SADDR_RULE_OIF,
962 IPV6_SADDR_RULE_LABEL,
963 #ifdef CONFIG_IPV6_PRIVACY
964 IPV6_SADDR_RULE_PRIVACY,
965 #endif
966 IPV6_SADDR_RULE_ORCHID,
967 IPV6_SADDR_RULE_PREFIX,
968 IPV6_SADDR_RULE_MAX
969 };
970
971 struct ipv6_saddr_score {
972 int rule;
973 int addr_type;
974 struct inet6_ifaddr *ifa;
975 DECLARE_BITMAP(scorebits, IPV6_SADDR_RULE_MAX);
976 int scopedist;
977 int matchlen;
978 };
979
980 struct ipv6_saddr_dst {
981 const struct in6_addr *addr;
982 int ifindex;
983 int scope;
984 int label;
985 unsigned int prefs;
986 };
987
988 static inline int ipv6_saddr_preferred(int type)
989 {
990 if (type & (IPV6_ADDR_MAPPED|IPV6_ADDR_COMPATv4|IPV6_ADDR_LOOPBACK))
991 return 1;
992 return 0;
993 }
994
995 static int ipv6_get_saddr_eval(struct net *net,
996 struct ipv6_saddr_score *score,
997 struct ipv6_saddr_dst *dst,
998 int i)
999 {
1000 int ret;
1001
1002 if (i <= score->rule) {
1003 switch (i) {
1004 case IPV6_SADDR_RULE_SCOPE:
1005 ret = score->scopedist;
1006 break;
1007 case IPV6_SADDR_RULE_PREFIX:
1008 ret = score->matchlen;
1009 break;
1010 default:
1011 ret = !!test_bit(i, score->scorebits);
1012 }
1013 goto out;
1014 }
1015
1016 switch (i) {
1017 case IPV6_SADDR_RULE_INIT:
1018 /* Rule 0: remember if hiscore is not ready yet */
1019 ret = !!score->ifa;
1020 break;
1021 case IPV6_SADDR_RULE_LOCAL:
1022 /* Rule 1: Prefer same address */
1023 ret = ipv6_addr_equal(&score->ifa->addr, dst->addr);
1024 break;
1025 case IPV6_SADDR_RULE_SCOPE:
1026 /* Rule 2: Prefer appropriate scope
1027 *
1028 * ret
1029 * ^
1030 * -1 | d 15
1031 * ---+--+-+---> scope
1032 * |
1033 * | d is scope of the destination.
1034 * B-d | \
1035 * | \ <- smaller scope is better if
1036 * B-15 | \ if scope is enough for destinaion.
1037 * | ret = B - scope (-1 <= scope >= d <= 15).
1038 * d-C-1 | /
1039 * |/ <- greater is better
1040 * -C / if scope is not enough for destination.
1041 * /| ret = scope - C (-1 <= d < scope <= 15).
1042 *
1043 * d - C - 1 < B -15 (for all -1 <= d <= 15).
1044 * C > d + 14 - B >= 15 + 14 - B = 29 - B.
1045 * Assume B = 0 and we get C > 29.
1046 */
1047 ret = __ipv6_addr_src_scope(score->addr_type);
1048 if (ret >= dst->scope)
1049 ret = -ret;
1050 else
1051 ret -= 128; /* 30 is enough */
1052 score->scopedist = ret;
1053 break;
1054 case IPV6_SADDR_RULE_PREFERRED:
1055 /* Rule 3: Avoid deprecated and optimistic addresses */
1056 ret = ipv6_saddr_preferred(score->addr_type) ||
1057 !(score->ifa->flags & (IFA_F_DEPRECATED|IFA_F_OPTIMISTIC));
1058 break;
1059 #ifdef CONFIG_IPV6_MIP6
1060 case IPV6_SADDR_RULE_HOA:
1061 {
1062 /* Rule 4: Prefer home address */
1063 int prefhome = !(dst->prefs & IPV6_PREFER_SRC_COA);
1064 ret = !(score->ifa->flags & IFA_F_HOMEADDRESS) ^ prefhome;
1065 break;
1066 }
1067 #endif
1068 case IPV6_SADDR_RULE_OIF:
1069 /* Rule 5: Prefer outgoing interface */
1070 ret = (!dst->ifindex ||
1071 dst->ifindex == score->ifa->idev->dev->ifindex);
1072 break;
1073 case IPV6_SADDR_RULE_LABEL:
1074 /* Rule 6: Prefer matching label */
1075 ret = ipv6_addr_label(net,
1076 &score->ifa->addr, score->addr_type,
1077 score->ifa->idev->dev->ifindex) == dst->label;
1078 break;
1079 #ifdef CONFIG_IPV6_PRIVACY
1080 case IPV6_SADDR_RULE_PRIVACY:
1081 {
1082 /* Rule 7: Prefer public address
1083 * Note: prefer temprary address if use_tempaddr >= 2
1084 */
1085 int preftmp = dst->prefs & (IPV6_PREFER_SRC_PUBLIC|IPV6_PREFER_SRC_TMP) ?
1086 !!(dst->prefs & IPV6_PREFER_SRC_TMP) :
1087 score->ifa->idev->cnf.use_tempaddr >= 2;
1088 ret = (!(score->ifa->flags & IFA_F_TEMPORARY)) ^ preftmp;
1089 break;
1090 }
1091 #endif
1092 case IPV6_SADDR_RULE_ORCHID:
1093 /* Rule 8-: Prefer ORCHID vs ORCHID or
1094 * non-ORCHID vs non-ORCHID
1095 */
1096 ret = !(ipv6_addr_orchid(&score->ifa->addr) ^
1097 ipv6_addr_orchid(dst->addr));
1098 break;
1099 case IPV6_SADDR_RULE_PREFIX:
1100 /* Rule 8: Use longest matching prefix */
1101 score->matchlen = ret = ipv6_addr_diff(&score->ifa->addr,
1102 dst->addr);
1103 break;
1104 default:
1105 ret = 0;
1106 }
1107
1108 if (ret)
1109 __set_bit(i, score->scorebits);
1110 score->rule = i;
1111 out:
1112 return ret;
1113 }
1114
1115 int ipv6_dev_get_saddr(struct net *net, struct net_device *dst_dev,
1116 const struct in6_addr *daddr, unsigned int prefs,
1117 struct in6_addr *saddr)
1118 {
1119 struct ipv6_saddr_score scores[2],
1120 *score = &scores[0], *hiscore = &scores[1];
1121 struct ipv6_saddr_dst dst;
1122 struct net_device *dev;
1123 int dst_type;
1124
1125 dst_type = __ipv6_addr_type(daddr);
1126 dst.addr = daddr;
1127 dst.ifindex = dst_dev ? dst_dev->ifindex : 0;
1128 dst.scope = __ipv6_addr_src_scope(dst_type);
1129 dst.label = ipv6_addr_label(net, daddr, dst_type, dst.ifindex);
1130 dst.prefs = prefs;
1131
1132 hiscore->rule = -1;
1133 hiscore->ifa = NULL;
1134
1135 rcu_read_lock();
1136
1137 for_each_netdev_rcu(net, dev) {
1138 struct inet6_dev *idev;
1139
1140 /* Candidate Source Address (section 4)
1141 * - multicast and link-local destination address,
1142 * the set of candidate source address MUST only
1143 * include addresses assigned to interfaces
1144 * belonging to the same link as the outgoing
1145 * interface.
1146 * (- For site-local destination addresses, the
1147 * set of candidate source addresses MUST only
1148 * include addresses assigned to interfaces
1149 * belonging to the same site as the outgoing
1150 * interface.)
1151 */
1152 if (((dst_type & IPV6_ADDR_MULTICAST) ||
1153 dst.scope <= IPV6_ADDR_SCOPE_LINKLOCAL) &&
1154 dst.ifindex && dev->ifindex != dst.ifindex)
1155 continue;
1156
1157 idev = __in6_dev_get(dev);
1158 if (!idev)
1159 continue;
1160
1161 read_lock_bh(&idev->lock);
1162 list_for_each_entry(score->ifa, &idev->addr_list, if_list) {
1163 int i;
1164
1165 /*
1166 * - Tentative Address (RFC2462 section 5.4)
1167 * - A tentative address is not considered
1168 * "assigned to an interface" in the traditional
1169 * sense, unless it is also flagged as optimistic.
1170 * - Candidate Source Address (section 4)
1171 * - In any case, anycast addresses, multicast
1172 * addresses, and the unspecified address MUST
1173 * NOT be included in a candidate set.
1174 */
1175 if ((score->ifa->flags & IFA_F_TENTATIVE) &&
1176 (!(score->ifa->flags & IFA_F_OPTIMISTIC)))
1177 continue;
1178
1179 score->addr_type = __ipv6_addr_type(&score->ifa->addr);
1180
1181 if (unlikely(score->addr_type == IPV6_ADDR_ANY ||
1182 score->addr_type & IPV6_ADDR_MULTICAST)) {
1183 LIMIT_NETDEBUG(KERN_DEBUG
1184 "ADDRCONF: unspecified / multicast address "
1185 "assigned as unicast address on %s",
1186 dev->name);
1187 continue;
1188 }
1189
1190 score->rule = -1;
1191 bitmap_zero(score->scorebits, IPV6_SADDR_RULE_MAX);
1192
1193 for (i = 0; i < IPV6_SADDR_RULE_MAX; i++) {
1194 int minihiscore, miniscore;
1195
1196 minihiscore = ipv6_get_saddr_eval(net, hiscore, &dst, i);
1197 miniscore = ipv6_get_saddr_eval(net, score, &dst, i);
1198
1199 if (minihiscore > miniscore) {
1200 if (i == IPV6_SADDR_RULE_SCOPE &&
1201 score->scopedist > 0) {
1202 /*
1203 * special case:
1204 * each remaining entry
1205 * has too small (not enough)
1206 * scope, because ifa entries
1207 * are sorted by their scope
1208 * values.
1209 */
1210 goto try_nextdev;
1211 }
1212 break;
1213 } else if (minihiscore < miniscore) {
1214 if (hiscore->ifa)
1215 in6_ifa_put(hiscore->ifa);
1216
1217 in6_ifa_hold(score->ifa);
1218
1219 swap(hiscore, score);
1220
1221 /* restore our iterator */
1222 score->ifa = hiscore->ifa;
1223
1224 break;
1225 }
1226 }
1227 }
1228 try_nextdev:
1229 read_unlock_bh(&idev->lock);
1230 }
1231 rcu_read_unlock();
1232
1233 if (!hiscore->ifa)
1234 return -EADDRNOTAVAIL;
1235
1236 ipv6_addr_copy(saddr, &hiscore->ifa->addr);
1237 in6_ifa_put(hiscore->ifa);
1238 return 0;
1239 }
1240 EXPORT_SYMBOL(ipv6_dev_get_saddr);
1241
1242 int ipv6_get_lladdr(struct net_device *dev, struct in6_addr *addr,
1243 unsigned char banned_flags)
1244 {
1245 struct inet6_dev *idev;
1246 int err = -EADDRNOTAVAIL;
1247
1248 rcu_read_lock();
1249 idev = __in6_dev_get(dev);
1250 if (idev) {
1251 struct inet6_ifaddr *ifp;
1252
1253 read_lock_bh(&idev->lock);
1254 list_for_each_entry(ifp, &idev->addr_list, if_list) {
1255 if (ifp->scope == IFA_LINK &&
1256 !(ifp->flags & banned_flags)) {
1257 ipv6_addr_copy(addr, &ifp->addr);
1258 err = 0;
1259 break;
1260 }
1261 }
1262 read_unlock_bh(&idev->lock);
1263 }
1264 rcu_read_unlock();
1265 return err;
1266 }
1267
1268 static int ipv6_count_addresses(struct inet6_dev *idev)
1269 {
1270 int cnt = 0;
1271 struct inet6_ifaddr *ifp;
1272
1273 read_lock_bh(&idev->lock);
1274 list_for_each_entry(ifp, &idev->addr_list, if_list)
1275 cnt++;
1276 read_unlock_bh(&idev->lock);
1277 return cnt;
1278 }
1279
1280 int ipv6_chk_addr(struct net *net, struct in6_addr *addr,
1281 struct net_device *dev, int strict)
1282 {
1283 struct inet6_ifaddr *ifp;
1284 struct hlist_node *node;
1285 unsigned int hash = ipv6_addr_hash(addr);
1286
1287 rcu_read_lock_bh();
1288 hlist_for_each_entry_rcu(ifp, node, &inet6_addr_lst[hash], addr_lst) {
1289 if (!net_eq(dev_net(ifp->idev->dev), net))
1290 continue;
1291 if (ipv6_addr_equal(&ifp->addr, addr) &&
1292 !(ifp->flags&IFA_F_TENTATIVE) &&
1293 (dev == NULL || ifp->idev->dev == dev ||
1294 !(ifp->scope&(IFA_LINK|IFA_HOST) || strict))) {
1295 rcu_read_unlock_bh();
1296 return 1;
1297 }
1298 }
1299
1300 rcu_read_unlock_bh();
1301 return 0;
1302 }
1303 EXPORT_SYMBOL(ipv6_chk_addr);
1304
1305 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
1306 struct net_device *dev)
1307 {
1308 unsigned int hash = ipv6_addr_hash(addr);
1309 struct inet6_ifaddr *ifp;
1310 struct hlist_node *node;
1311
1312 hlist_for_each_entry(ifp, node, &inet6_addr_lst[hash], addr_lst) {
1313 if (!net_eq(dev_net(ifp->idev->dev), net))
1314 continue;
1315 if (ipv6_addr_equal(&ifp->addr, addr)) {
1316 if (dev == NULL || ifp->idev->dev == dev)
1317 return true;
1318 }
1319 }
1320 return false;
1321 }
1322
1323 int ipv6_chk_prefix(struct in6_addr *addr, struct net_device *dev)
1324 {
1325 struct inet6_dev *idev;
1326 struct inet6_ifaddr *ifa;
1327 int onlink;
1328
1329 onlink = 0;
1330 rcu_read_lock();
1331 idev = __in6_dev_get(dev);
1332 if (idev) {
1333 read_lock_bh(&idev->lock);
1334 list_for_each_entry(ifa, &idev->addr_list, if_list) {
1335 onlink = ipv6_prefix_equal(addr, &ifa->addr,
1336 ifa->prefix_len);
1337 if (onlink)
1338 break;
1339 }
1340 read_unlock_bh(&idev->lock);
1341 }
1342 rcu_read_unlock();
1343 return onlink;
1344 }
1345
1346 EXPORT_SYMBOL(ipv6_chk_prefix);
1347
1348 struct inet6_ifaddr *ipv6_get_ifaddr(struct net *net, const struct in6_addr *addr,
1349 struct net_device *dev, int strict)
1350 {
1351 struct inet6_ifaddr *ifp, *result = NULL;
1352 unsigned int hash = ipv6_addr_hash(addr);
1353 struct hlist_node *node;
1354
1355 rcu_read_lock_bh();
1356 hlist_for_each_entry_rcu_bh(ifp, node, &inet6_addr_lst[hash], addr_lst) {
1357 if (!net_eq(dev_net(ifp->idev->dev), net))
1358 continue;
1359 if (ipv6_addr_equal(&ifp->addr, addr)) {
1360 if (dev == NULL || ifp->idev->dev == dev ||
1361 !(ifp->scope&(IFA_LINK|IFA_HOST) || strict)) {
1362 result = ifp;
1363 in6_ifa_hold(ifp);
1364 break;
1365 }
1366 }
1367 }
1368 rcu_read_unlock_bh();
1369
1370 return result;
1371 }
1372
1373 /* Gets referenced address, destroys ifaddr */
1374
1375 static void addrconf_dad_stop(struct inet6_ifaddr *ifp, int dad_failed)
1376 {
1377 if (ifp->flags&IFA_F_PERMANENT) {
1378 spin_lock_bh(&ifp->lock);
1379 addrconf_del_timer(ifp);
1380 ifp->flags |= IFA_F_TENTATIVE;
1381 if (dad_failed)
1382 ifp->flags |= IFA_F_DADFAILED;
1383 spin_unlock_bh(&ifp->lock);
1384 if (dad_failed)
1385 ipv6_ifa_notify(0, ifp);
1386 in6_ifa_put(ifp);
1387 #ifdef CONFIG_IPV6_PRIVACY
1388 } else if (ifp->flags&IFA_F_TEMPORARY) {
1389 struct inet6_ifaddr *ifpub;
1390 spin_lock_bh(&ifp->lock);
1391 ifpub = ifp->ifpub;
1392 if (ifpub) {
1393 in6_ifa_hold(ifpub);
1394 spin_unlock_bh(&ifp->lock);
1395 ipv6_create_tempaddr(ifpub, ifp);
1396 in6_ifa_put(ifpub);
1397 } else {
1398 spin_unlock_bh(&ifp->lock);
1399 }
1400 ipv6_del_addr(ifp);
1401 #endif
1402 } else
1403 ipv6_del_addr(ifp);
1404 }
1405
1406 static int addrconf_dad_end(struct inet6_ifaddr *ifp)
1407 {
1408 int err = -ENOENT;
1409
1410 spin_lock(&ifp->state_lock);
1411 if (ifp->state == INET6_IFADDR_STATE_DAD) {
1412 ifp->state = INET6_IFADDR_STATE_POSTDAD;
1413 err = 0;
1414 }
1415 spin_unlock(&ifp->state_lock);
1416
1417 return err;
1418 }
1419
1420 void addrconf_dad_failure(struct inet6_ifaddr *ifp)
1421 {
1422 struct inet6_dev *idev = ifp->idev;
1423
1424 if (addrconf_dad_end(ifp)) {
1425 in6_ifa_put(ifp);
1426 return;
1427 }
1428
1429 if (net_ratelimit())
1430 printk(KERN_INFO "%s: IPv6 duplicate address %pI6c detected!\n",
1431 ifp->idev->dev->name, &ifp->addr);
1432
1433 if (idev->cnf.accept_dad > 1 && !idev->cnf.disable_ipv6) {
1434 struct in6_addr addr;
1435
1436 addr.s6_addr32[0] = htonl(0xfe800000);
1437 addr.s6_addr32[1] = 0;
1438
1439 if (!ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) &&
1440 ipv6_addr_equal(&ifp->addr, &addr)) {
1441 /* DAD failed for link-local based on MAC address */
1442 idev->cnf.disable_ipv6 = 1;
1443
1444 printk(KERN_INFO "%s: IPv6 being disabled!\n",
1445 ifp->idev->dev->name);
1446 }
1447 }
1448
1449 addrconf_dad_stop(ifp, 1);
1450 }
1451
1452 /* Join to solicited addr multicast group. */
1453
1454 void addrconf_join_solict(struct net_device *dev, struct in6_addr *addr)
1455 {
1456 struct in6_addr maddr;
1457
1458 if (dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1459 return;
1460
1461 addrconf_addr_solict_mult(addr, &maddr);
1462 ipv6_dev_mc_inc(dev, &maddr);
1463 }
1464
1465 void addrconf_leave_solict(struct inet6_dev *idev, struct in6_addr *addr)
1466 {
1467 struct in6_addr maddr;
1468
1469 if (idev->dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1470 return;
1471
1472 addrconf_addr_solict_mult(addr, &maddr);
1473 __ipv6_dev_mc_dec(idev, &maddr);
1474 }
1475
1476 static void addrconf_join_anycast(struct inet6_ifaddr *ifp)
1477 {
1478 struct in6_addr addr;
1479 ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1480 if (ipv6_addr_any(&addr))
1481 return;
1482 ipv6_dev_ac_inc(ifp->idev->dev, &addr);
1483 }
1484
1485 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp)
1486 {
1487 struct in6_addr addr;
1488 ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1489 if (ipv6_addr_any(&addr))
1490 return;
1491 __ipv6_dev_ac_dec(ifp->idev, &addr);
1492 }
1493
1494 static int addrconf_ifid_eui48(u8 *eui, struct net_device *dev)
1495 {
1496 if (dev->addr_len != ETH_ALEN)
1497 return -1;
1498 memcpy(eui, dev->dev_addr, 3);
1499 memcpy(eui + 5, dev->dev_addr + 3, 3);
1500
1501 /*
1502 * The zSeries OSA network cards can be shared among various
1503 * OS instances, but the OSA cards have only one MAC address.
1504 * This leads to duplicate address conflicts in conjunction
1505 * with IPv6 if more than one instance uses the same card.
1506 *
1507 * The driver for these cards can deliver a unique 16-bit
1508 * identifier for each instance sharing the same card. It is
1509 * placed instead of 0xFFFE in the interface identifier. The
1510 * "u" bit of the interface identifier is not inverted in this
1511 * case. Hence the resulting interface identifier has local
1512 * scope according to RFC2373.
1513 */
1514 if (dev->dev_id) {
1515 eui[3] = (dev->dev_id >> 8) & 0xFF;
1516 eui[4] = dev->dev_id & 0xFF;
1517 } else {
1518 eui[3] = 0xFF;
1519 eui[4] = 0xFE;
1520 eui[0] ^= 2;
1521 }
1522 return 0;
1523 }
1524
1525 static int addrconf_ifid_arcnet(u8 *eui, struct net_device *dev)
1526 {
1527 /* XXX: inherit EUI-64 from other interface -- yoshfuji */
1528 if (dev->addr_len != ARCNET_ALEN)
1529 return -1;
1530 memset(eui, 0, 7);
1531 eui[7] = *(u8*)dev->dev_addr;
1532 return 0;
1533 }
1534
1535 static int addrconf_ifid_infiniband(u8 *eui, struct net_device *dev)
1536 {
1537 if (dev->addr_len != INFINIBAND_ALEN)
1538 return -1;
1539 memcpy(eui, dev->dev_addr + 12, 8);
1540 eui[0] |= 2;
1541 return 0;
1542 }
1543
1544 int __ipv6_isatap_ifid(u8 *eui, __be32 addr)
1545 {
1546 if (addr == 0)
1547 return -1;
1548 eui[0] = (ipv4_is_zeronet(addr) || ipv4_is_private_10(addr) ||
1549 ipv4_is_loopback(addr) || ipv4_is_linklocal_169(addr) ||
1550 ipv4_is_private_172(addr) || ipv4_is_test_192(addr) ||
1551 ipv4_is_anycast_6to4(addr) || ipv4_is_private_192(addr) ||
1552 ipv4_is_test_198(addr) || ipv4_is_multicast(addr) ||
1553 ipv4_is_lbcast(addr)) ? 0x00 : 0x02;
1554 eui[1] = 0;
1555 eui[2] = 0x5E;
1556 eui[3] = 0xFE;
1557 memcpy(eui + 4, &addr, 4);
1558 return 0;
1559 }
1560 EXPORT_SYMBOL(__ipv6_isatap_ifid);
1561
1562 static int addrconf_ifid_sit(u8 *eui, struct net_device *dev)
1563 {
1564 if (dev->priv_flags & IFF_ISATAP)
1565 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
1566 return -1;
1567 }
1568
1569 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev)
1570 {
1571 switch (dev->type) {
1572 case ARPHRD_ETHER:
1573 case ARPHRD_FDDI:
1574 case ARPHRD_IEEE802_TR:
1575 return addrconf_ifid_eui48(eui, dev);
1576 case ARPHRD_ARCNET:
1577 return addrconf_ifid_arcnet(eui, dev);
1578 case ARPHRD_INFINIBAND:
1579 return addrconf_ifid_infiniband(eui, dev);
1580 case ARPHRD_SIT:
1581 return addrconf_ifid_sit(eui, dev);
1582 }
1583 return -1;
1584 }
1585
1586 static int ipv6_inherit_eui64(u8 *eui, struct inet6_dev *idev)
1587 {
1588 int err = -1;
1589 struct inet6_ifaddr *ifp;
1590
1591 read_lock_bh(&idev->lock);
1592 list_for_each_entry(ifp, &idev->addr_list, if_list) {
1593 if (ifp->scope == IFA_LINK && !(ifp->flags&IFA_F_TENTATIVE)) {
1594 memcpy(eui, ifp->addr.s6_addr+8, 8);
1595 err = 0;
1596 break;
1597 }
1598 }
1599 read_unlock_bh(&idev->lock);
1600 return err;
1601 }
1602
1603 #ifdef CONFIG_IPV6_PRIVACY
1604 /* (re)generation of randomized interface identifier (RFC 3041 3.2, 3.5) */
1605 static int __ipv6_regen_rndid(struct inet6_dev *idev)
1606 {
1607 regen:
1608 get_random_bytes(idev->rndid, sizeof(idev->rndid));
1609 idev->rndid[0] &= ~0x02;
1610
1611 /*
1612 * <draft-ietf-ipngwg-temp-addresses-v2-00.txt>:
1613 * check if generated address is not inappropriate
1614 *
1615 * - Reserved subnet anycast (RFC 2526)
1616 * 11111101 11....11 1xxxxxxx
1617 * - ISATAP (RFC4214) 6.1
1618 * 00-00-5E-FE-xx-xx-xx-xx
1619 * - value 0
1620 * - XXX: already assigned to an address on the device
1621 */
1622 if (idev->rndid[0] == 0xfd &&
1623 (idev->rndid[1]&idev->rndid[2]&idev->rndid[3]&idev->rndid[4]&idev->rndid[5]&idev->rndid[6]) == 0xff &&
1624 (idev->rndid[7]&0x80))
1625 goto regen;
1626 if ((idev->rndid[0]|idev->rndid[1]) == 0) {
1627 if (idev->rndid[2] == 0x5e && idev->rndid[3] == 0xfe)
1628 goto regen;
1629 if ((idev->rndid[2]|idev->rndid[3]|idev->rndid[4]|idev->rndid[5]|idev->rndid[6]|idev->rndid[7]) == 0x00)
1630 goto regen;
1631 }
1632
1633 return 0;
1634 }
1635
1636 static void ipv6_regen_rndid(unsigned long data)
1637 {
1638 struct inet6_dev *idev = (struct inet6_dev *) data;
1639 unsigned long expires;
1640
1641 rcu_read_lock_bh();
1642 write_lock_bh(&idev->lock);
1643
1644 if (idev->dead)
1645 goto out;
1646
1647 if (__ipv6_regen_rndid(idev) < 0)
1648 goto out;
1649
1650 expires = jiffies +
1651 idev->cnf.temp_prefered_lft * HZ -
1652 idev->cnf.regen_max_retry * idev->cnf.dad_transmits * idev->nd_parms->retrans_time - desync_factor;
1653 if (time_before(expires, jiffies)) {
1654 printk(KERN_WARNING
1655 "ipv6_regen_rndid(): too short regeneration interval; timer disabled for %s.\n",
1656 idev->dev->name);
1657 goto out;
1658 }
1659
1660 if (!mod_timer(&idev->regen_timer, expires))
1661 in6_dev_hold(idev);
1662
1663 out:
1664 write_unlock_bh(&idev->lock);
1665 rcu_read_unlock_bh();
1666 in6_dev_put(idev);
1667 }
1668
1669 static int __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr) {
1670 int ret = 0;
1671
1672 if (tmpaddr && memcmp(idev->rndid, &tmpaddr->s6_addr[8], 8) == 0)
1673 ret = __ipv6_regen_rndid(idev);
1674 return ret;
1675 }
1676 #endif
1677
1678 /*
1679 * Add prefix route.
1680 */
1681
1682 static void
1683 addrconf_prefix_route(struct in6_addr *pfx, int plen, struct net_device *dev,
1684 unsigned long expires, u32 flags)
1685 {
1686 struct fib6_config cfg = {
1687 .fc_table = RT6_TABLE_PREFIX,
1688 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1689 .fc_ifindex = dev->ifindex,
1690 .fc_expires = expires,
1691 .fc_dst_len = plen,
1692 .fc_flags = RTF_UP | flags,
1693 .fc_nlinfo.nl_net = dev_net(dev),
1694 .fc_protocol = RTPROT_KERNEL,
1695 };
1696
1697 ipv6_addr_copy(&cfg.fc_dst, pfx);
1698
1699 /* Prevent useless cloning on PtP SIT.
1700 This thing is done here expecting that the whole
1701 class of non-broadcast devices need not cloning.
1702 */
1703 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
1704 if (dev->type == ARPHRD_SIT && (dev->flags & IFF_POINTOPOINT))
1705 cfg.fc_flags |= RTF_NONEXTHOP;
1706 #endif
1707
1708 ip6_route_add(&cfg);
1709 }
1710
1711 /* Create "default" multicast route to the interface */
1712
1713 static void addrconf_add_mroute(struct net_device *dev)
1714 {
1715 struct fib6_config cfg = {
1716 .fc_table = RT6_TABLE_LOCAL,
1717 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1718 .fc_ifindex = dev->ifindex,
1719 .fc_dst_len = 8,
1720 .fc_flags = RTF_UP,
1721 .fc_nlinfo.nl_net = dev_net(dev),
1722 };
1723
1724 ipv6_addr_set(&cfg.fc_dst, htonl(0xFF000000), 0, 0, 0);
1725
1726 ip6_route_add(&cfg);
1727 }
1728
1729 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
1730 static void sit_route_add(struct net_device *dev)
1731 {
1732 struct fib6_config cfg = {
1733 .fc_table = RT6_TABLE_MAIN,
1734 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1735 .fc_ifindex = dev->ifindex,
1736 .fc_dst_len = 96,
1737 .fc_flags = RTF_UP | RTF_NONEXTHOP,
1738 .fc_nlinfo.nl_net = dev_net(dev),
1739 };
1740
1741 /* prefix length - 96 bits "::d.d.d.d" */
1742 ip6_route_add(&cfg);
1743 }
1744 #endif
1745
1746 static void addrconf_add_lroute(struct net_device *dev)
1747 {
1748 struct in6_addr addr;
1749
1750 ipv6_addr_set(&addr, htonl(0xFE800000), 0, 0, 0);
1751 addrconf_prefix_route(&addr, 64, dev, 0, 0);
1752 }
1753
1754 static struct inet6_dev *addrconf_add_dev(struct net_device *dev)
1755 {
1756 struct inet6_dev *idev;
1757
1758 ASSERT_RTNL();
1759
1760 idev = ipv6_find_idev(dev);
1761 if (!idev)
1762 return ERR_PTR(-ENOBUFS);
1763
1764 if (idev->cnf.disable_ipv6)
1765 return ERR_PTR(-EACCES);
1766
1767 /* Add default multicast route */
1768 addrconf_add_mroute(dev);
1769
1770 /* Add link local route */
1771 addrconf_add_lroute(dev);
1772 return idev;
1773 }
1774
1775 void addrconf_prefix_rcv(struct net_device *dev, u8 *opt, int len)
1776 {
1777 struct prefix_info *pinfo;
1778 __u32 valid_lft;
1779 __u32 prefered_lft;
1780 int addr_type;
1781 struct inet6_dev *in6_dev;
1782 struct net *net = dev_net(dev);
1783
1784 pinfo = (struct prefix_info *) opt;
1785
1786 if (len < sizeof(struct prefix_info)) {
1787 ADBG(("addrconf: prefix option too short\n"));
1788 return;
1789 }
1790
1791 /*
1792 * Validation checks ([ADDRCONF], page 19)
1793 */
1794
1795 addr_type = ipv6_addr_type(&pinfo->prefix);
1796
1797 if (addr_type & (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL))
1798 return;
1799
1800 valid_lft = ntohl(pinfo->valid);
1801 prefered_lft = ntohl(pinfo->prefered);
1802
1803 if (prefered_lft > valid_lft) {
1804 if (net_ratelimit())
1805 printk(KERN_WARNING "addrconf: prefix option has invalid lifetime\n");
1806 return;
1807 }
1808
1809 in6_dev = in6_dev_get(dev);
1810
1811 if (in6_dev == NULL) {
1812 if (net_ratelimit())
1813 printk(KERN_DEBUG "addrconf: device %s not configured\n", dev->name);
1814 return;
1815 }
1816
1817 /*
1818 * Two things going on here:
1819 * 1) Add routes for on-link prefixes
1820 * 2) Configure prefixes with the auto flag set
1821 */
1822
1823 if (pinfo->onlink) {
1824 struct rt6_info *rt;
1825 unsigned long rt_expires;
1826
1827 /* Avoid arithmetic overflow. Really, we could
1828 * save rt_expires in seconds, likely valid_lft,
1829 * but it would require division in fib gc, that it
1830 * not good.
1831 */
1832 if (HZ > USER_HZ)
1833 rt_expires = addrconf_timeout_fixup(valid_lft, HZ);
1834 else
1835 rt_expires = addrconf_timeout_fixup(valid_lft, USER_HZ);
1836
1837 if (addrconf_finite_timeout(rt_expires))
1838 rt_expires *= HZ;
1839
1840 rt = rt6_lookup(net, &pinfo->prefix, NULL,
1841 dev->ifindex, 1);
1842
1843 if (rt && addrconf_is_prefix_route(rt)) {
1844 /* Autoconf prefix route */
1845 if (valid_lft == 0) {
1846 ip6_del_rt(rt);
1847 rt = NULL;
1848 } else if (addrconf_finite_timeout(rt_expires)) {
1849 /* not infinity */
1850 rt->rt6i_expires = jiffies + rt_expires;
1851 rt->rt6i_flags |= RTF_EXPIRES;
1852 } else {
1853 rt->rt6i_flags &= ~RTF_EXPIRES;
1854 rt->rt6i_expires = 0;
1855 }
1856 } else if (valid_lft) {
1857 clock_t expires = 0;
1858 int flags = RTF_ADDRCONF | RTF_PREFIX_RT;
1859 if (addrconf_finite_timeout(rt_expires)) {
1860 /* not infinity */
1861 flags |= RTF_EXPIRES;
1862 expires = jiffies_to_clock_t(rt_expires);
1863 }
1864 addrconf_prefix_route(&pinfo->prefix, pinfo->prefix_len,
1865 dev, expires, flags);
1866 }
1867 if (rt)
1868 dst_release(&rt->u.dst);
1869 }
1870
1871 /* Try to figure out our local address for this prefix */
1872
1873 if (pinfo->autoconf && in6_dev->cnf.autoconf) {
1874 struct inet6_ifaddr * ifp;
1875 struct in6_addr addr;
1876 int create = 0, update_lft = 0;
1877
1878 if (pinfo->prefix_len == 64) {
1879 memcpy(&addr, &pinfo->prefix, 8);
1880 if (ipv6_generate_eui64(addr.s6_addr + 8, dev) &&
1881 ipv6_inherit_eui64(addr.s6_addr + 8, in6_dev)) {
1882 in6_dev_put(in6_dev);
1883 return;
1884 }
1885 goto ok;
1886 }
1887 if (net_ratelimit())
1888 printk(KERN_DEBUG "IPv6 addrconf: prefix with wrong length %d\n",
1889 pinfo->prefix_len);
1890 in6_dev_put(in6_dev);
1891 return;
1892
1893 ok:
1894
1895 ifp = ipv6_get_ifaddr(net, &addr, dev, 1);
1896
1897 if (ifp == NULL && valid_lft) {
1898 int max_addresses = in6_dev->cnf.max_addresses;
1899 u32 addr_flags = 0;
1900
1901 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1902 if (in6_dev->cnf.optimistic_dad &&
1903 !net->ipv6.devconf_all->forwarding)
1904 addr_flags = IFA_F_OPTIMISTIC;
1905 #endif
1906
1907 /* Do not allow to create too much of autoconfigured
1908 * addresses; this would be too easy way to crash kernel.
1909 */
1910 if (!max_addresses ||
1911 ipv6_count_addresses(in6_dev) < max_addresses)
1912 ifp = ipv6_add_addr(in6_dev, &addr, pinfo->prefix_len,
1913 addr_type&IPV6_ADDR_SCOPE_MASK,
1914 addr_flags);
1915
1916 if (!ifp || IS_ERR(ifp)) {
1917 in6_dev_put(in6_dev);
1918 return;
1919 }
1920
1921 update_lft = create = 1;
1922 ifp->cstamp = jiffies;
1923 addrconf_dad_start(ifp, RTF_ADDRCONF|RTF_PREFIX_RT);
1924 }
1925
1926 if (ifp) {
1927 int flags;
1928 unsigned long now;
1929 #ifdef CONFIG_IPV6_PRIVACY
1930 struct inet6_ifaddr *ift;
1931 #endif
1932 u32 stored_lft;
1933
1934 /* update lifetime (RFC2462 5.5.3 e) */
1935 spin_lock(&ifp->lock);
1936 now = jiffies;
1937 if (ifp->valid_lft > (now - ifp->tstamp) / HZ)
1938 stored_lft = ifp->valid_lft - (now - ifp->tstamp) / HZ;
1939 else
1940 stored_lft = 0;
1941 if (!update_lft && stored_lft) {
1942 if (valid_lft > MIN_VALID_LIFETIME ||
1943 valid_lft > stored_lft)
1944 update_lft = 1;
1945 else if (stored_lft <= MIN_VALID_LIFETIME) {
1946 /* valid_lft <= stored_lft is always true */
1947 /*
1948 * RFC 4862 Section 5.5.3e:
1949 * "Note that the preferred lifetime of
1950 * the corresponding address is always
1951 * reset to the Preferred Lifetime in
1952 * the received Prefix Information
1953 * option, regardless of whether the
1954 * valid lifetime is also reset or
1955 * ignored."
1956 *
1957 * So if the preferred lifetime in
1958 * this advertisement is different
1959 * than what we have stored, but the
1960 * valid lifetime is invalid, just
1961 * reset prefered_lft.
1962 *
1963 * We must set the valid lifetime
1964 * to the stored lifetime since we'll
1965 * be updating the timestamp below,
1966 * else we'll set it back to the
1967 * minumum.
1968 */
1969 if (prefered_lft != ifp->prefered_lft) {
1970 valid_lft = stored_lft;
1971 update_lft = 1;
1972 }
1973 } else {
1974 valid_lft = MIN_VALID_LIFETIME;
1975 if (valid_lft < prefered_lft)
1976 prefered_lft = valid_lft;
1977 update_lft = 1;
1978 }
1979 }
1980
1981 if (update_lft) {
1982 ifp->valid_lft = valid_lft;
1983 ifp->prefered_lft = prefered_lft;
1984 ifp->tstamp = now;
1985 flags = ifp->flags;
1986 ifp->flags &= ~IFA_F_DEPRECATED;
1987 spin_unlock(&ifp->lock);
1988
1989 if (!(flags&IFA_F_TENTATIVE))
1990 ipv6_ifa_notify(0, ifp);
1991 } else
1992 spin_unlock(&ifp->lock);
1993
1994 #ifdef CONFIG_IPV6_PRIVACY
1995 read_lock_bh(&in6_dev->lock);
1996 /* update all temporary addresses in the list */
1997 list_for_each_entry(ift, &in6_dev->tempaddr_list, tmp_list) {
1998 /*
1999 * When adjusting the lifetimes of an existing
2000 * temporary address, only lower the lifetimes.
2001 * Implementations must not increase the
2002 * lifetimes of an existing temporary address
2003 * when processing a Prefix Information Option.
2004 */
2005 if (ifp != ift->ifpub)
2006 continue;
2007
2008 spin_lock(&ift->lock);
2009 flags = ift->flags;
2010 if (ift->valid_lft > valid_lft &&
2011 ift->valid_lft - valid_lft > (jiffies - ift->tstamp) / HZ)
2012 ift->valid_lft = valid_lft + (jiffies - ift->tstamp) / HZ;
2013 if (ift->prefered_lft > prefered_lft &&
2014 ift->prefered_lft - prefered_lft > (jiffies - ift->tstamp) / HZ)
2015 ift->prefered_lft = prefered_lft + (jiffies - ift->tstamp) / HZ;
2016 spin_unlock(&ift->lock);
2017 if (!(flags&IFA_F_TENTATIVE))
2018 ipv6_ifa_notify(0, ift);
2019 }
2020
2021 if (create && in6_dev->cnf.use_tempaddr > 0) {
2022 /*
2023 * When a new public address is created as described in [ADDRCONF],
2024 * also create a new temporary address.
2025 */
2026 read_unlock_bh(&in6_dev->lock);
2027 ipv6_create_tempaddr(ifp, NULL);
2028 } else {
2029 read_unlock_bh(&in6_dev->lock);
2030 }
2031 #endif
2032 in6_ifa_put(ifp);
2033 addrconf_verify(0);
2034 }
2035 }
2036 inet6_prefix_notify(RTM_NEWPREFIX, in6_dev, pinfo);
2037 in6_dev_put(in6_dev);
2038 }
2039
2040 /*
2041 * Set destination address.
2042 * Special case for SIT interfaces where we create a new "virtual"
2043 * device.
2044 */
2045 int addrconf_set_dstaddr(struct net *net, void __user *arg)
2046 {
2047 struct in6_ifreq ireq;
2048 struct net_device *dev;
2049 int err = -EINVAL;
2050
2051 rtnl_lock();
2052
2053 err = -EFAULT;
2054 if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2055 goto err_exit;
2056
2057 dev = __dev_get_by_index(net, ireq.ifr6_ifindex);
2058
2059 err = -ENODEV;
2060 if (dev == NULL)
2061 goto err_exit;
2062
2063 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2064 if (dev->type == ARPHRD_SIT) {
2065 const struct net_device_ops *ops = dev->netdev_ops;
2066 struct ifreq ifr;
2067 struct ip_tunnel_parm p;
2068
2069 err = -EADDRNOTAVAIL;
2070 if (!(ipv6_addr_type(&ireq.ifr6_addr) & IPV6_ADDR_COMPATv4))
2071 goto err_exit;
2072
2073 memset(&p, 0, sizeof(p));
2074 p.iph.daddr = ireq.ifr6_addr.s6_addr32[3];
2075 p.iph.saddr = 0;
2076 p.iph.version = 4;
2077 p.iph.ihl = 5;
2078 p.iph.protocol = IPPROTO_IPV6;
2079 p.iph.ttl = 64;
2080 ifr.ifr_ifru.ifru_data = (__force void __user *)&p;
2081
2082 if (ops->ndo_do_ioctl) {
2083 mm_segment_t oldfs = get_fs();
2084
2085 set_fs(KERNEL_DS);
2086 err = ops->ndo_do_ioctl(dev, &ifr, SIOCADDTUNNEL);
2087 set_fs(oldfs);
2088 } else
2089 err = -EOPNOTSUPP;
2090
2091 if (err == 0) {
2092 err = -ENOBUFS;
2093 dev = __dev_get_by_name(net, p.name);
2094 if (!dev)
2095 goto err_exit;
2096 err = dev_open(dev);
2097 }
2098 }
2099 #endif
2100
2101 err_exit:
2102 rtnl_unlock();
2103 return err;
2104 }
2105
2106 /*
2107 * Manual configuration of address on an interface
2108 */
2109 static int inet6_addr_add(struct net *net, int ifindex, struct in6_addr *pfx,
2110 unsigned int plen, __u8 ifa_flags, __u32 prefered_lft,
2111 __u32 valid_lft)
2112 {
2113 struct inet6_ifaddr *ifp;
2114 struct inet6_dev *idev;
2115 struct net_device *dev;
2116 int scope;
2117 u32 flags;
2118 clock_t expires;
2119 unsigned long timeout;
2120
2121 ASSERT_RTNL();
2122
2123 if (plen > 128)
2124 return -EINVAL;
2125
2126 /* check the lifetime */
2127 if (!valid_lft || prefered_lft > valid_lft)
2128 return -EINVAL;
2129
2130 dev = __dev_get_by_index(net, ifindex);
2131 if (!dev)
2132 return -ENODEV;
2133
2134 idev = addrconf_add_dev(dev);
2135 if (IS_ERR(idev))
2136 return PTR_ERR(idev);
2137
2138 scope = ipv6_addr_scope(pfx);
2139
2140 timeout = addrconf_timeout_fixup(valid_lft, HZ);
2141 if (addrconf_finite_timeout(timeout)) {
2142 expires = jiffies_to_clock_t(timeout * HZ);
2143 valid_lft = timeout;
2144 flags = RTF_EXPIRES;
2145 } else {
2146 expires = 0;
2147 flags = 0;
2148 ifa_flags |= IFA_F_PERMANENT;
2149 }
2150
2151 timeout = addrconf_timeout_fixup(prefered_lft, HZ);
2152 if (addrconf_finite_timeout(timeout)) {
2153 if (timeout == 0)
2154 ifa_flags |= IFA_F_DEPRECATED;
2155 prefered_lft = timeout;
2156 }
2157
2158 ifp = ipv6_add_addr(idev, pfx, plen, scope, ifa_flags);
2159
2160 if (!IS_ERR(ifp)) {
2161 spin_lock_bh(&ifp->lock);
2162 ifp->valid_lft = valid_lft;
2163 ifp->prefered_lft = prefered_lft;
2164 ifp->tstamp = jiffies;
2165 spin_unlock_bh(&ifp->lock);
2166
2167 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, dev,
2168 expires, flags);
2169 /*
2170 * Note that section 3.1 of RFC 4429 indicates
2171 * that the Optimistic flag should not be set for
2172 * manually configured addresses
2173 */
2174 addrconf_dad_start(ifp, 0);
2175 in6_ifa_put(ifp);
2176 addrconf_verify(0);
2177 return 0;
2178 }
2179
2180 return PTR_ERR(ifp);
2181 }
2182
2183 static int inet6_addr_del(struct net *net, int ifindex, struct in6_addr *pfx,
2184 unsigned int plen)
2185 {
2186 struct inet6_ifaddr *ifp;
2187 struct inet6_dev *idev;
2188 struct net_device *dev;
2189
2190 if (plen > 128)
2191 return -EINVAL;
2192
2193 dev = __dev_get_by_index(net, ifindex);
2194 if (!dev)
2195 return -ENODEV;
2196
2197 if ((idev = __in6_dev_get(dev)) == NULL)
2198 return -ENXIO;
2199
2200 read_lock_bh(&idev->lock);
2201 list_for_each_entry(ifp, &idev->addr_list, if_list) {
2202 if (ifp->prefix_len == plen &&
2203 ipv6_addr_equal(pfx, &ifp->addr)) {
2204 in6_ifa_hold(ifp);
2205 read_unlock_bh(&idev->lock);
2206
2207 ipv6_del_addr(ifp);
2208
2209 /* If the last address is deleted administratively,
2210 disable IPv6 on this interface.
2211 */
2212 if (list_empty(&idev->addr_list))
2213 addrconf_ifdown(idev->dev, 1);
2214 return 0;
2215 }
2216 }
2217 read_unlock_bh(&idev->lock);
2218 return -EADDRNOTAVAIL;
2219 }
2220
2221
2222 int addrconf_add_ifaddr(struct net *net, void __user *arg)
2223 {
2224 struct in6_ifreq ireq;
2225 int err;
2226
2227 if (!capable(CAP_NET_ADMIN))
2228 return -EPERM;
2229
2230 if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2231 return -EFAULT;
2232
2233 rtnl_lock();
2234 err = inet6_addr_add(net, ireq.ifr6_ifindex, &ireq.ifr6_addr,
2235 ireq.ifr6_prefixlen, IFA_F_PERMANENT,
2236 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
2237 rtnl_unlock();
2238 return err;
2239 }
2240
2241 int addrconf_del_ifaddr(struct net *net, void __user *arg)
2242 {
2243 struct in6_ifreq ireq;
2244 int err;
2245
2246 if (!capable(CAP_NET_ADMIN))
2247 return -EPERM;
2248
2249 if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2250 return -EFAULT;
2251
2252 rtnl_lock();
2253 err = inet6_addr_del(net, ireq.ifr6_ifindex, &ireq.ifr6_addr,
2254 ireq.ifr6_prefixlen);
2255 rtnl_unlock();
2256 return err;
2257 }
2258
2259 static void add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
2260 int plen, int scope)
2261 {
2262 struct inet6_ifaddr *ifp;
2263
2264 ifp = ipv6_add_addr(idev, addr, plen, scope, IFA_F_PERMANENT);
2265 if (!IS_ERR(ifp)) {
2266 spin_lock_bh(&ifp->lock);
2267 ifp->flags &= ~IFA_F_TENTATIVE;
2268 spin_unlock_bh(&ifp->lock);
2269 ipv6_ifa_notify(RTM_NEWADDR, ifp);
2270 in6_ifa_put(ifp);
2271 }
2272 }
2273
2274 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2275 static void sit_add_v4_addrs(struct inet6_dev *idev)
2276 {
2277 struct in6_addr addr;
2278 struct net_device *dev;
2279 struct net *net = dev_net(idev->dev);
2280 int scope;
2281
2282 ASSERT_RTNL();
2283
2284 memset(&addr, 0, sizeof(struct in6_addr));
2285 memcpy(&addr.s6_addr32[3], idev->dev->dev_addr, 4);
2286
2287 if (idev->dev->flags&IFF_POINTOPOINT) {
2288 addr.s6_addr32[0] = htonl(0xfe800000);
2289 scope = IFA_LINK;
2290 } else {
2291 scope = IPV6_ADDR_COMPATv4;
2292 }
2293
2294 if (addr.s6_addr32[3]) {
2295 add_addr(idev, &addr, 128, scope);
2296 return;
2297 }
2298
2299 for_each_netdev(net, dev) {
2300 struct in_device * in_dev = __in_dev_get_rtnl(dev);
2301 if (in_dev && (dev->flags & IFF_UP)) {
2302 struct in_ifaddr * ifa;
2303
2304 int flag = scope;
2305
2306 for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) {
2307 int plen;
2308
2309 addr.s6_addr32[3] = ifa->ifa_local;
2310
2311 if (ifa->ifa_scope == RT_SCOPE_LINK)
2312 continue;
2313 if (ifa->ifa_scope >= RT_SCOPE_HOST) {
2314 if (idev->dev->flags&IFF_POINTOPOINT)
2315 continue;
2316 flag |= IFA_HOST;
2317 }
2318 if (idev->dev->flags&IFF_POINTOPOINT)
2319 plen = 64;
2320 else
2321 plen = 96;
2322
2323 add_addr(idev, &addr, plen, flag);
2324 }
2325 }
2326 }
2327 }
2328 #endif
2329
2330 static void init_loopback(struct net_device *dev)
2331 {
2332 struct inet6_dev *idev;
2333
2334 /* ::1 */
2335
2336 ASSERT_RTNL();
2337
2338 if ((idev = ipv6_find_idev(dev)) == NULL) {
2339 printk(KERN_DEBUG "init loopback: add_dev failed\n");
2340 return;
2341 }
2342
2343 add_addr(idev, &in6addr_loopback, 128, IFA_HOST);
2344 }
2345
2346 static void addrconf_add_linklocal(struct inet6_dev *idev, struct in6_addr *addr)
2347 {
2348 struct inet6_ifaddr * ifp;
2349 u32 addr_flags = IFA_F_PERMANENT;
2350
2351 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2352 if (idev->cnf.optimistic_dad &&
2353 !dev_net(idev->dev)->ipv6.devconf_all->forwarding)
2354 addr_flags |= IFA_F_OPTIMISTIC;
2355 #endif
2356
2357
2358 ifp = ipv6_add_addr(idev, addr, 64, IFA_LINK, addr_flags);
2359 if (!IS_ERR(ifp)) {
2360 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, idev->dev, 0, 0);
2361 addrconf_dad_start(ifp, 0);
2362 in6_ifa_put(ifp);
2363 }
2364 }
2365
2366 static void addrconf_dev_config(struct net_device *dev)
2367 {
2368 struct in6_addr addr;
2369 struct inet6_dev * idev;
2370
2371 ASSERT_RTNL();
2372
2373 if ((dev->type != ARPHRD_ETHER) &&
2374 (dev->type != ARPHRD_FDDI) &&
2375 (dev->type != ARPHRD_IEEE802_TR) &&
2376 (dev->type != ARPHRD_ARCNET) &&
2377 (dev->type != ARPHRD_INFINIBAND)) {
2378 /* Alas, we support only Ethernet autoconfiguration. */
2379 return;
2380 }
2381
2382 idev = addrconf_add_dev(dev);
2383 if (IS_ERR(idev))
2384 return;
2385
2386 memset(&addr, 0, sizeof(struct in6_addr));
2387 addr.s6_addr32[0] = htonl(0xFE800000);
2388
2389 if (ipv6_generate_eui64(addr.s6_addr + 8, dev) == 0)
2390 addrconf_add_linklocal(idev, &addr);
2391 }
2392
2393 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2394 static void addrconf_sit_config(struct net_device *dev)
2395 {
2396 struct inet6_dev *idev;
2397
2398 ASSERT_RTNL();
2399
2400 /*
2401 * Configure the tunnel with one of our IPv4
2402 * addresses... we should configure all of
2403 * our v4 addrs in the tunnel
2404 */
2405
2406 if ((idev = ipv6_find_idev(dev)) == NULL) {
2407 printk(KERN_DEBUG "init sit: add_dev failed\n");
2408 return;
2409 }
2410
2411 if (dev->priv_flags & IFF_ISATAP) {
2412 struct in6_addr addr;
2413
2414 ipv6_addr_set(&addr, htonl(0xFE800000), 0, 0, 0);
2415 addrconf_prefix_route(&addr, 64, dev, 0, 0);
2416 if (!ipv6_generate_eui64(addr.s6_addr + 8, dev))
2417 addrconf_add_linklocal(idev, &addr);
2418 return;
2419 }
2420
2421 sit_add_v4_addrs(idev);
2422
2423 if (dev->flags&IFF_POINTOPOINT) {
2424 addrconf_add_mroute(dev);
2425 addrconf_add_lroute(dev);
2426 } else
2427 sit_route_add(dev);
2428 }
2429 #endif
2430
2431 static inline int
2432 ipv6_inherit_linklocal(struct inet6_dev *idev, struct net_device *link_dev)
2433 {
2434 struct in6_addr lladdr;
2435
2436 if (!ipv6_get_lladdr(link_dev, &lladdr, IFA_F_TENTATIVE)) {
2437 addrconf_add_linklocal(idev, &lladdr);
2438 return 0;
2439 }
2440 return -1;
2441 }
2442
2443 static void ip6_tnl_add_linklocal(struct inet6_dev *idev)
2444 {
2445 struct net_device *link_dev;
2446 struct net *net = dev_net(idev->dev);
2447
2448 /* first try to inherit the link-local address from the link device */
2449 if (idev->dev->iflink &&
2450 (link_dev = __dev_get_by_index(net, idev->dev->iflink))) {
2451 if (!ipv6_inherit_linklocal(idev, link_dev))
2452 return;
2453 }
2454 /* then try to inherit it from any device */
2455 for_each_netdev(net, link_dev) {
2456 if (!ipv6_inherit_linklocal(idev, link_dev))
2457 return;
2458 }
2459 printk(KERN_DEBUG "init ip6-ip6: add_linklocal failed\n");
2460 }
2461
2462 /*
2463 * Autoconfigure tunnel with a link-local address so routing protocols,
2464 * DHCPv6, MLD etc. can be run over the virtual link
2465 */
2466
2467 static void addrconf_ip6_tnl_config(struct net_device *dev)
2468 {
2469 struct inet6_dev *idev;
2470
2471 ASSERT_RTNL();
2472
2473 idev = addrconf_add_dev(dev);
2474 if (IS_ERR(idev)) {
2475 printk(KERN_DEBUG "init ip6-ip6: add_dev failed\n");
2476 return;
2477 }
2478 ip6_tnl_add_linklocal(idev);
2479 }
2480
2481 static int addrconf_notify(struct notifier_block *this, unsigned long event,
2482 void * data)
2483 {
2484 struct net_device *dev = (struct net_device *) data;
2485 struct inet6_dev *idev = __in6_dev_get(dev);
2486 int run_pending = 0;
2487 int err;
2488
2489 switch (event) {
2490 case NETDEV_REGISTER:
2491 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2492 idev = ipv6_add_dev(dev);
2493 if (!idev)
2494 return notifier_from_errno(-ENOMEM);
2495 }
2496 break;
2497
2498 case NETDEV_UP:
2499 case NETDEV_CHANGE:
2500 if (dev->flags & IFF_SLAVE)
2501 break;
2502
2503 if (event == NETDEV_UP) {
2504 if (!addrconf_qdisc_ok(dev)) {
2505 /* device is not ready yet. */
2506 printk(KERN_INFO
2507 "ADDRCONF(NETDEV_UP): %s: "
2508 "link is not ready\n",
2509 dev->name);
2510 break;
2511 }
2512
2513 if (!idev && dev->mtu >= IPV6_MIN_MTU)
2514 idev = ipv6_add_dev(dev);
2515
2516 if (idev) {
2517 idev->if_flags |= IF_READY;
2518 run_pending = 1;
2519 }
2520 } else {
2521 if (!addrconf_qdisc_ok(dev)) {
2522 /* device is still not ready. */
2523 break;
2524 }
2525
2526 if (idev) {
2527 if (idev->if_flags & IF_READY)
2528 /* device is already configured. */
2529 break;
2530 idev->if_flags |= IF_READY;
2531 }
2532
2533 printk(KERN_INFO
2534 "ADDRCONF(NETDEV_CHANGE): %s: "
2535 "link becomes ready\n",
2536 dev->name);
2537
2538 run_pending = 1;
2539 }
2540
2541 switch (dev->type) {
2542 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2543 case ARPHRD_SIT:
2544 addrconf_sit_config(dev);
2545 break;
2546 #endif
2547 case ARPHRD_TUNNEL6:
2548 addrconf_ip6_tnl_config(dev);
2549 break;
2550 case ARPHRD_LOOPBACK:
2551 init_loopback(dev);
2552 break;
2553
2554 default:
2555 addrconf_dev_config(dev);
2556 break;
2557 }
2558
2559 if (idev) {
2560 if (run_pending)
2561 addrconf_dad_run(idev);
2562
2563 /*
2564 * If the MTU changed during the interface down,
2565 * when the interface up, the changed MTU must be
2566 * reflected in the idev as well as routers.
2567 */
2568 if (idev->cnf.mtu6 != dev->mtu &&
2569 dev->mtu >= IPV6_MIN_MTU) {
2570 rt6_mtu_change(dev, dev->mtu);
2571 idev->cnf.mtu6 = dev->mtu;
2572 }
2573 idev->tstamp = jiffies;
2574 inet6_ifinfo_notify(RTM_NEWLINK, idev);
2575
2576 /*
2577 * If the changed mtu during down is lower than
2578 * IPV6_MIN_MTU stop IPv6 on this interface.
2579 */
2580 if (dev->mtu < IPV6_MIN_MTU)
2581 addrconf_ifdown(dev, 1);
2582 }
2583 break;
2584
2585 case NETDEV_CHANGEMTU:
2586 if (idev && dev->mtu >= IPV6_MIN_MTU) {
2587 rt6_mtu_change(dev, dev->mtu);
2588 idev->cnf.mtu6 = dev->mtu;
2589 break;
2590 }
2591
2592 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2593 idev = ipv6_add_dev(dev);
2594 if (idev)
2595 break;
2596 }
2597
2598 /*
2599 * MTU falled under IPV6_MIN_MTU.
2600 * Stop IPv6 on this interface.
2601 */
2602
2603 case NETDEV_DOWN:
2604 case NETDEV_UNREGISTER:
2605 /*
2606 * Remove all addresses from this interface.
2607 */
2608 addrconf_ifdown(dev, event != NETDEV_DOWN);
2609 break;
2610
2611 case NETDEV_CHANGENAME:
2612 if (idev) {
2613 snmp6_unregister_dev(idev);
2614 addrconf_sysctl_unregister(idev);
2615 addrconf_sysctl_register(idev);
2616 err = snmp6_register_dev(idev);
2617 if (err)
2618 return notifier_from_errno(err);
2619 }
2620 break;
2621
2622 case NETDEV_PRE_TYPE_CHANGE:
2623 case NETDEV_POST_TYPE_CHANGE:
2624 addrconf_type_change(dev, event);
2625 break;
2626 }
2627
2628 return NOTIFY_OK;
2629 }
2630
2631 /*
2632 * addrconf module should be notified of a device going up
2633 */
2634 static struct notifier_block ipv6_dev_notf = {
2635 .notifier_call = addrconf_notify,
2636 };
2637
2638 static void addrconf_type_change(struct net_device *dev, unsigned long event)
2639 {
2640 struct inet6_dev *idev;
2641 ASSERT_RTNL();
2642
2643 idev = __in6_dev_get(dev);
2644
2645 if (event == NETDEV_POST_TYPE_CHANGE)
2646 ipv6_mc_remap(idev);
2647 else if (event == NETDEV_PRE_TYPE_CHANGE)
2648 ipv6_mc_unmap(idev);
2649 }
2650
2651 static int addrconf_ifdown(struct net_device *dev, int how)
2652 {
2653 struct net *net = dev_net(dev);
2654 struct inet6_dev *idev;
2655 struct inet6_ifaddr *ifa;
2656 LIST_HEAD(keep_list);
2657 int state;
2658
2659 ASSERT_RTNL();
2660
2661 rt6_ifdown(net, dev);
2662 neigh_ifdown(&nd_tbl, dev);
2663
2664 idev = __in6_dev_get(dev);
2665 if (idev == NULL)
2666 return -ENODEV;
2667
2668 /*
2669 * Step 1: remove reference to ipv6 device from parent device.
2670 * Do not dev_put!
2671 */
2672 if (how) {
2673 idev->dead = 1;
2674
2675 /* protected by rtnl_lock */
2676 rcu_assign_pointer(dev->ip6_ptr, NULL);
2677
2678 /* Step 1.5: remove snmp6 entry */
2679 snmp6_unregister_dev(idev);
2680
2681 }
2682
2683 write_lock_bh(&idev->lock);
2684
2685 /* Step 2: clear flags for stateless addrconf */
2686 if (!how)
2687 idev->if_flags &= ~(IF_RS_SENT|IF_RA_RCVD|IF_READY);
2688
2689 #ifdef CONFIG_IPV6_PRIVACY
2690 if (how && del_timer(&idev->regen_timer))
2691 in6_dev_put(idev);
2692
2693 /* Step 3: clear tempaddr list */
2694 while (!list_empty(&idev->tempaddr_list)) {
2695 ifa = list_first_entry(&idev->tempaddr_list,
2696 struct inet6_ifaddr, tmp_list);
2697 list_del(&ifa->tmp_list);
2698 write_unlock_bh(&idev->lock);
2699 spin_lock_bh(&ifa->lock);
2700
2701 if (ifa->ifpub) {
2702 in6_ifa_put(ifa->ifpub);
2703 ifa->ifpub = NULL;
2704 }
2705 spin_unlock_bh(&ifa->lock);
2706 in6_ifa_put(ifa);
2707 write_lock_bh(&idev->lock);
2708 }
2709 #endif
2710
2711 while (!list_empty(&idev->addr_list)) {
2712 ifa = list_first_entry(&idev->addr_list,
2713 struct inet6_ifaddr, if_list);
2714 addrconf_del_timer(ifa);
2715
2716 /* If just doing link down, and address is permanent
2717 and not link-local, then retain it. */
2718 if (!how &&
2719 (ifa->flags&IFA_F_PERMANENT) &&
2720 !(ipv6_addr_type(&ifa->addr) & IPV6_ADDR_LINKLOCAL)) {
2721 list_move_tail(&ifa->if_list, &keep_list);
2722
2723 /* If not doing DAD on this address, just keep it. */
2724 if ((dev->flags&(IFF_NOARP|IFF_LOOPBACK)) ||
2725 idev->cnf.accept_dad <= 0 ||
2726 (ifa->flags & IFA_F_NODAD))
2727 continue;
2728
2729 /* If it was tentative already, no need to notify */
2730 if (ifa->flags & IFA_F_TENTATIVE)
2731 continue;
2732
2733 /* Flag it for later restoration when link comes up */
2734 ifa->flags |= IFA_F_TENTATIVE;
2735 ifa->state = INET6_IFADDR_STATE_DAD;
2736
2737 write_unlock_bh(&idev->lock);
2738
2739 in6_ifa_hold(ifa);
2740 } else {
2741 list_del(&ifa->if_list);
2742
2743 /* clear hash table */
2744 spin_lock_bh(&addrconf_hash_lock);
2745 hlist_del_init_rcu(&ifa->addr_lst);
2746 spin_unlock_bh(&addrconf_hash_lock);
2747
2748 write_unlock_bh(&idev->lock);
2749 spin_lock_bh(&ifa->state_lock);
2750 state = ifa->state;
2751 ifa->state = INET6_IFADDR_STATE_DEAD;
2752 spin_unlock_bh(&ifa->state_lock);
2753
2754 if (state == INET6_IFADDR_STATE_DEAD)
2755 goto put_ifa;
2756 }
2757
2758 __ipv6_ifa_notify(RTM_DELADDR, ifa);
2759 if (ifa->state == INET6_IFADDR_STATE_DEAD)
2760 atomic_notifier_call_chain(&inet6addr_chain,
2761 NETDEV_DOWN, ifa);
2762
2763 put_ifa:
2764 in6_ifa_put(ifa);
2765
2766 write_lock_bh(&idev->lock);
2767 }
2768
2769 list_splice(&keep_list, &idev->addr_list);
2770
2771 write_unlock_bh(&idev->lock);
2772
2773 /* Step 5: Discard multicast list */
2774 if (how)
2775 ipv6_mc_destroy_dev(idev);
2776 else
2777 ipv6_mc_down(idev);
2778
2779 idev->tstamp = jiffies;
2780
2781 /* Last: Shot the device (if unregistered) */
2782 if (how) {
2783 addrconf_sysctl_unregister(idev);
2784 neigh_parms_release(&nd_tbl, idev->nd_parms);
2785 neigh_ifdown(&nd_tbl, dev);
2786 in6_dev_put(idev);
2787 }
2788 return 0;
2789 }
2790
2791 static void addrconf_rs_timer(unsigned long data)
2792 {
2793 struct inet6_ifaddr *ifp = (struct inet6_ifaddr *) data;
2794 struct inet6_dev *idev = ifp->idev;
2795
2796 read_lock(&idev->lock);
2797 if (idev->dead || !(idev->if_flags & IF_READY))
2798 goto out;
2799
2800 if (idev->cnf.forwarding)
2801 goto out;
2802
2803 /* Announcement received after solicitation was sent */
2804 if (idev->if_flags & IF_RA_RCVD)
2805 goto out;
2806
2807 spin_lock(&ifp->lock);
2808 if (ifp->probes++ < idev->cnf.rtr_solicits) {
2809 /* The wait after the last probe can be shorter */
2810 addrconf_mod_timer(ifp, AC_RS,
2811 (ifp->probes == idev->cnf.rtr_solicits) ?
2812 idev->cnf.rtr_solicit_delay :
2813 idev->cnf.rtr_solicit_interval);
2814 spin_unlock(&ifp->lock);
2815
2816 ndisc_send_rs(idev->dev, &ifp->addr, &in6addr_linklocal_allrouters);
2817 } else {
2818 spin_unlock(&ifp->lock);
2819 /*
2820 * Note: we do not support deprecated "all on-link"
2821 * assumption any longer.
2822 */
2823 printk(KERN_DEBUG "%s: no IPv6 routers present\n",
2824 idev->dev->name);
2825 }
2826
2827 out:
2828 read_unlock(&idev->lock);
2829 in6_ifa_put(ifp);
2830 }
2831
2832 /*
2833 * Duplicate Address Detection
2834 */
2835 static void addrconf_dad_kick(struct inet6_ifaddr *ifp)
2836 {
2837 unsigned long rand_num;
2838 struct inet6_dev *idev = ifp->idev;
2839
2840 if (ifp->flags & IFA_F_OPTIMISTIC)
2841 rand_num = 0;
2842 else
2843 rand_num = net_random() % (idev->cnf.rtr_solicit_delay ? : 1);
2844
2845 ifp->probes = idev->cnf.dad_transmits;
2846 addrconf_mod_timer(ifp, AC_DAD, rand_num);
2847 }
2848
2849 static void addrconf_dad_start(struct inet6_ifaddr *ifp, u32 flags)
2850 {
2851 struct inet6_dev *idev = ifp->idev;
2852 struct net_device *dev = idev->dev;
2853
2854 addrconf_join_solict(dev, &ifp->addr);
2855
2856 net_srandom(ifp->addr.s6_addr32[3]);
2857
2858 read_lock_bh(&idev->lock);
2859 spin_lock(&ifp->lock);
2860 if (ifp->state == INET6_IFADDR_STATE_DEAD)
2861 goto out;
2862
2863 if (dev->flags&(IFF_NOARP|IFF_LOOPBACK) ||
2864 idev->cnf.accept_dad < 1 ||
2865 !(ifp->flags&IFA_F_TENTATIVE) ||
2866 ifp->flags & IFA_F_NODAD) {
2867 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
2868 spin_unlock(&ifp->lock);
2869 read_unlock_bh(&idev->lock);
2870
2871 addrconf_dad_completed(ifp);
2872 return;
2873 }
2874
2875 if (!(idev->if_flags & IF_READY)) {
2876 spin_unlock(&ifp->lock);
2877 read_unlock_bh(&idev->lock);
2878 /*
2879 * If the device is not ready:
2880 * - keep it tentative if it is a permanent address.
2881 * - otherwise, kill it.
2882 */
2883 in6_ifa_hold(ifp);
2884 addrconf_dad_stop(ifp, 0);
2885 return;
2886 }
2887
2888 /*
2889 * Optimistic nodes can start receiving
2890 * Frames right away
2891 */
2892 if (ifp->flags & IFA_F_OPTIMISTIC)
2893 ip6_ins_rt(ifp->rt);
2894
2895 addrconf_dad_kick(ifp);
2896 out:
2897 spin_unlock(&ifp->lock);
2898 read_unlock_bh(&idev->lock);
2899 }
2900
2901 static void addrconf_dad_timer(unsigned long data)
2902 {
2903 struct inet6_ifaddr *ifp = (struct inet6_ifaddr *) data;
2904 struct inet6_dev *idev = ifp->idev;
2905 struct in6_addr mcaddr;
2906
2907 if (!ifp->probes && addrconf_dad_end(ifp))
2908 goto out;
2909
2910 read_lock(&idev->lock);
2911 if (idev->dead || !(idev->if_flags & IF_READY)) {
2912 read_unlock(&idev->lock);
2913 goto out;
2914 }
2915
2916 spin_lock(&ifp->lock);
2917 if (ifp->state == INET6_IFADDR_STATE_DEAD) {
2918 spin_unlock(&ifp->lock);
2919 read_unlock(&idev->lock);
2920 goto out;
2921 }
2922
2923 if (ifp->probes == 0) {
2924 /*
2925 * DAD was successful
2926 */
2927
2928 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
2929 spin_unlock(&ifp->lock);
2930 read_unlock(&idev->lock);
2931
2932 addrconf_dad_completed(ifp);
2933
2934 goto out;
2935 }
2936
2937 ifp->probes--;
2938 addrconf_mod_timer(ifp, AC_DAD, ifp->idev->nd_parms->retrans_time);
2939 spin_unlock(&ifp->lock);
2940 read_unlock(&idev->lock);
2941
2942 /* send a neighbour solicitation for our addr */
2943 addrconf_addr_solict_mult(&ifp->addr, &mcaddr);
2944 ndisc_send_ns(ifp->idev->dev, NULL, &ifp->addr, &mcaddr, &in6addr_any);
2945 out:
2946 in6_ifa_put(ifp);
2947 }
2948
2949 static void addrconf_dad_completed(struct inet6_ifaddr *ifp)
2950 {
2951 struct net_device *dev = ifp->idev->dev;
2952
2953 /*
2954 * Configure the address for reception. Now it is valid.
2955 */
2956
2957 ipv6_ifa_notify(RTM_NEWADDR, ifp);
2958
2959 /* If added prefix is link local and forwarding is off,
2960 start sending router solicitations.
2961 */
2962
2963 if ((ifp->idev->cnf.forwarding == 0 ||
2964 ifp->idev->cnf.forwarding == 2) &&
2965 ifp->idev->cnf.rtr_solicits > 0 &&
2966 (dev->flags&IFF_LOOPBACK) == 0 &&
2967 (ipv6_addr_type(&ifp->addr) & IPV6_ADDR_LINKLOCAL)) {
2968 /*
2969 * If a host as already performed a random delay
2970 * [...] as part of DAD [...] there is no need
2971 * to delay again before sending the first RS
2972 */
2973 ndisc_send_rs(ifp->idev->dev, &ifp->addr, &in6addr_linklocal_allrouters);
2974
2975 spin_lock_bh(&ifp->lock);
2976 ifp->probes = 1;
2977 ifp->idev->if_flags |= IF_RS_SENT;
2978 addrconf_mod_timer(ifp, AC_RS, ifp->idev->cnf.rtr_solicit_interval);
2979 spin_unlock_bh(&ifp->lock);
2980 }
2981 }
2982
2983 static void addrconf_dad_run(struct inet6_dev *idev)
2984 {
2985 struct inet6_ifaddr *ifp;
2986
2987 read_lock_bh(&idev->lock);
2988 list_for_each_entry(ifp, &idev->addr_list, if_list) {
2989 spin_lock(&ifp->lock);
2990 if (ifp->flags & IFA_F_TENTATIVE &&
2991 ifp->state == INET6_IFADDR_STATE_DAD)
2992 addrconf_dad_kick(ifp);
2993 spin_unlock(&ifp->lock);
2994 }
2995 read_unlock_bh(&idev->lock);
2996 }
2997
2998 #ifdef CONFIG_PROC_FS
2999 struct if6_iter_state {
3000 struct seq_net_private p;
3001 int bucket;
3002 };
3003
3004 static struct inet6_ifaddr *if6_get_first(struct seq_file *seq)
3005 {
3006 struct inet6_ifaddr *ifa = NULL;
3007 struct if6_iter_state *state = seq->private;
3008 struct net *net = seq_file_net(seq);
3009
3010 for (state->bucket = 0; state->bucket < IN6_ADDR_HSIZE; ++state->bucket) {
3011 struct hlist_node *n;
3012 hlist_for_each_entry_rcu_bh(ifa, n, &inet6_addr_lst[state->bucket],
3013 addr_lst)
3014 if (net_eq(dev_net(ifa->idev->dev), net))
3015 return ifa;
3016 }
3017 return NULL;
3018 }
3019
3020 static struct inet6_ifaddr *if6_get_next(struct seq_file *seq,
3021 struct inet6_ifaddr *ifa)
3022 {
3023 struct if6_iter_state *state = seq->private;
3024 struct net *net = seq_file_net(seq);
3025 struct hlist_node *n = &ifa->addr_lst;
3026
3027 hlist_for_each_entry_continue_rcu_bh(ifa, n, addr_lst)
3028 if (net_eq(dev_net(ifa->idev->dev), net))
3029 return ifa;
3030
3031 while (++state->bucket < IN6_ADDR_HSIZE) {
3032 hlist_for_each_entry_rcu_bh(ifa, n,
3033 &inet6_addr_lst[state->bucket], addr_lst) {
3034 if (net_eq(dev_net(ifa->idev->dev), net))
3035 return ifa;
3036 }
3037 }
3038
3039 return NULL;
3040 }
3041
3042 static struct inet6_ifaddr *if6_get_idx(struct seq_file *seq, loff_t pos)
3043 {
3044 struct inet6_ifaddr *ifa = if6_get_first(seq);
3045
3046 if (ifa)
3047 while (pos && (ifa = if6_get_next(seq, ifa)) != NULL)
3048 --pos;
3049 return pos ? NULL : ifa;
3050 }
3051
3052 static void *if6_seq_start(struct seq_file *seq, loff_t *pos)
3053 __acquires(rcu_bh)
3054 {
3055 rcu_read_lock_bh();
3056 return if6_get_idx(seq, *pos);
3057 }
3058
3059 static void *if6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
3060 {
3061 struct inet6_ifaddr *ifa;
3062
3063 ifa = if6_get_next(seq, v);
3064 ++*pos;
3065 return ifa;
3066 }
3067
3068 static void if6_seq_stop(struct seq_file *seq, void *v)
3069 __releases(rcu_bh)
3070 {
3071 rcu_read_unlock_bh();
3072 }
3073
3074 static int if6_seq_show(struct seq_file *seq, void *v)
3075 {
3076 struct inet6_ifaddr *ifp = (struct inet6_ifaddr *)v;
3077 seq_printf(seq, "%pi6 %02x %02x %02x %02x %8s\n",
3078 &ifp->addr,
3079 ifp->idev->dev->ifindex,
3080 ifp->prefix_len,
3081 ifp->scope,
3082 ifp->flags,
3083 ifp->idev->dev->name);
3084 return 0;
3085 }
3086
3087 static const struct seq_operations if6_seq_ops = {
3088 .start = if6_seq_start,
3089 .next = if6_seq_next,
3090 .show = if6_seq_show,
3091 .stop = if6_seq_stop,
3092 };
3093
3094 static int if6_seq_open(struct inode *inode, struct file *file)
3095 {
3096 return seq_open_net(inode, file, &if6_seq_ops,
3097 sizeof(struct if6_iter_state));
3098 }
3099
3100 static const struct file_operations if6_fops = {
3101 .owner = THIS_MODULE,
3102 .open = if6_seq_open,
3103 .read = seq_read,
3104 .llseek = seq_lseek,
3105 .release = seq_release_net,
3106 };
3107
3108 static int __net_init if6_proc_net_init(struct net *net)
3109 {
3110 if (!proc_net_fops_create(net, "if_inet6", S_IRUGO, &if6_fops))
3111 return -ENOMEM;
3112 return 0;
3113 }
3114
3115 static void __net_exit if6_proc_net_exit(struct net *net)
3116 {
3117 proc_net_remove(net, "if_inet6");
3118 }
3119
3120 static struct pernet_operations if6_proc_net_ops = {
3121 .init = if6_proc_net_init,
3122 .exit = if6_proc_net_exit,
3123 };
3124
3125 int __init if6_proc_init(void)
3126 {
3127 return register_pernet_subsys(&if6_proc_net_ops);
3128 }
3129
3130 void if6_proc_exit(void)
3131 {
3132 unregister_pernet_subsys(&if6_proc_net_ops);
3133 }
3134 #endif /* CONFIG_PROC_FS */
3135
3136 #if defined(CONFIG_IPV6_MIP6) || defined(CONFIG_IPV6_MIP6_MODULE)
3137 /* Check if address is a home address configured on any interface. */
3138 int ipv6_chk_home_addr(struct net *net, struct in6_addr *addr)
3139 {
3140 int ret = 0;
3141 struct inet6_ifaddr *ifp = NULL;
3142 struct hlist_node *n;
3143 unsigned int hash = ipv6_addr_hash(addr);
3144
3145 rcu_read_lock_bh();
3146 hlist_for_each_entry_rcu_bh(ifp, n, &inet6_addr_lst[hash], addr_lst) {
3147 if (!net_eq(dev_net(ifp->idev->dev), net))
3148 continue;
3149 if (ipv6_addr_equal(&ifp->addr, addr) &&
3150 (ifp->flags & IFA_F_HOMEADDRESS)) {
3151 ret = 1;
3152 break;
3153 }
3154 }
3155 rcu_read_unlock_bh();
3156 return ret;
3157 }
3158 #endif
3159
3160 /*
3161 * Periodic address status verification
3162 */
3163
3164 static void addrconf_verify(unsigned long foo)
3165 {
3166 unsigned long now, next, next_sec, next_sched;
3167 struct inet6_ifaddr *ifp;
3168 struct hlist_node *node;
3169 int i;
3170
3171 rcu_read_lock_bh();
3172 spin_lock(&addrconf_verify_lock);
3173 now = jiffies;
3174 next = round_jiffies_up(now + ADDR_CHECK_FREQUENCY);
3175
3176 del_timer(&addr_chk_timer);
3177
3178 for (i = 0; i < IN6_ADDR_HSIZE; i++) {
3179 restart:
3180 hlist_for_each_entry_rcu_bh(ifp, node,
3181 &inet6_addr_lst[i], addr_lst) {
3182 unsigned long age;
3183
3184 if (ifp->flags & IFA_F_PERMANENT)
3185 continue;
3186
3187 spin_lock(&ifp->lock);
3188 /* We try to batch several events at once. */
3189 age = (now - ifp->tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
3190
3191 if (ifp->valid_lft != INFINITY_LIFE_TIME &&
3192 age >= ifp->valid_lft) {
3193 spin_unlock(&ifp->lock);
3194 in6_ifa_hold(ifp);
3195 ipv6_del_addr(ifp);
3196 goto restart;
3197 } else if (ifp->prefered_lft == INFINITY_LIFE_TIME) {
3198 spin_unlock(&ifp->lock);
3199 continue;
3200 } else if (age >= ifp->prefered_lft) {
3201 /* jiffies - ifp->tstamp > age >= ifp->prefered_lft */
3202 int deprecate = 0;
3203
3204 if (!(ifp->flags&IFA_F_DEPRECATED)) {
3205 deprecate = 1;
3206 ifp->flags |= IFA_F_DEPRECATED;
3207 }
3208
3209 if (time_before(ifp->tstamp + ifp->valid_lft * HZ, next))
3210 next = ifp->tstamp + ifp->valid_lft * HZ;
3211
3212 spin_unlock(&ifp->lock);
3213
3214 if (deprecate) {
3215 in6_ifa_hold(ifp);
3216
3217 ipv6_ifa_notify(0, ifp);
3218 in6_ifa_put(ifp);
3219 goto restart;
3220 }
3221 #ifdef CONFIG_IPV6_PRIVACY
3222 } else if ((ifp->flags&IFA_F_TEMPORARY) &&
3223 !(ifp->flags&IFA_F_TENTATIVE)) {
3224 unsigned long regen_advance = ifp->idev->cnf.regen_max_retry *
3225 ifp->idev->cnf.dad_transmits *
3226 ifp->idev->nd_parms->retrans_time / HZ;
3227
3228 if (age >= ifp->prefered_lft - regen_advance) {
3229 struct inet6_ifaddr *ifpub = ifp->ifpub;
3230 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3231 next = ifp->tstamp + ifp->prefered_lft * HZ;
3232 if (!ifp->regen_count && ifpub) {
3233 ifp->regen_count++;
3234 in6_ifa_hold(ifp);
3235 in6_ifa_hold(ifpub);
3236 spin_unlock(&ifp->lock);
3237
3238 spin_lock(&ifpub->lock);
3239 ifpub->regen_count = 0;
3240 spin_unlock(&ifpub->lock);
3241 ipv6_create_tempaddr(ifpub, ifp);
3242 in6_ifa_put(ifpub);
3243 in6_ifa_put(ifp);
3244 goto restart;
3245 }
3246 } else if (time_before(ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ, next))
3247 next = ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ;
3248 spin_unlock(&ifp->lock);
3249 #endif
3250 } else {
3251 /* ifp->prefered_lft <= ifp->valid_lft */
3252 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3253 next = ifp->tstamp + ifp->prefered_lft * HZ;
3254 spin_unlock(&ifp->lock);
3255 }
3256 }
3257 }
3258
3259 next_sec = round_jiffies_up(next);
3260 next_sched = next;
3261
3262 /* If rounded timeout is accurate enough, accept it. */
3263 if (time_before(next_sec, next + ADDRCONF_TIMER_FUZZ))
3264 next_sched = next_sec;
3265
3266 /* And minimum interval is ADDRCONF_TIMER_FUZZ_MAX. */
3267 if (time_before(next_sched, jiffies + ADDRCONF_TIMER_FUZZ_MAX))
3268 next_sched = jiffies + ADDRCONF_TIMER_FUZZ_MAX;
3269
3270 ADBG((KERN_DEBUG "now = %lu, schedule = %lu, rounded schedule = %lu => %lu\n",
3271 now, next, next_sec, next_sched));
3272
3273 addr_chk_timer.expires = next_sched;
3274 add_timer(&addr_chk_timer);
3275 spin_unlock(&addrconf_verify_lock);
3276 rcu_read_unlock_bh();
3277 }
3278
3279 static struct in6_addr *extract_addr(struct nlattr *addr, struct nlattr *local)
3280 {
3281 struct in6_addr *pfx = NULL;
3282
3283 if (addr)
3284 pfx = nla_data(addr);
3285
3286 if (local) {
3287 if (pfx && nla_memcmp(local, pfx, sizeof(*pfx)))
3288 pfx = NULL;
3289 else
3290 pfx = nla_data(local);
3291 }
3292
3293 return pfx;
3294 }
3295
3296 static const struct nla_policy ifa_ipv6_policy[IFA_MAX+1] = {
3297 [IFA_ADDRESS] = { .len = sizeof(struct in6_addr) },
3298 [IFA_LOCAL] = { .len = sizeof(struct in6_addr) },
3299 [IFA_CACHEINFO] = { .len = sizeof(struct ifa_cacheinfo) },
3300 };
3301
3302 static int
3303 inet6_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
3304 {
3305 struct net *net = sock_net(skb->sk);
3306 struct ifaddrmsg *ifm;
3307 struct nlattr *tb[IFA_MAX+1];
3308 struct in6_addr *pfx;
3309 int err;
3310
3311 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3312 if (err < 0)
3313 return err;
3314
3315 ifm = nlmsg_data(nlh);
3316 pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3317 if (pfx == NULL)
3318 return -EINVAL;
3319
3320 return inet6_addr_del(net, ifm->ifa_index, pfx, ifm->ifa_prefixlen);
3321 }
3322
3323 static int inet6_addr_modify(struct inet6_ifaddr *ifp, u8 ifa_flags,
3324 u32 prefered_lft, u32 valid_lft)
3325 {
3326 u32 flags;
3327 clock_t expires;
3328 unsigned long timeout;
3329
3330 if (!valid_lft || (prefered_lft > valid_lft))
3331 return -EINVAL;
3332
3333 timeout = addrconf_timeout_fixup(valid_lft, HZ);
3334 if (addrconf_finite_timeout(timeout)) {
3335 expires = jiffies_to_clock_t(timeout * HZ);
3336 valid_lft = timeout;
3337 flags = RTF_EXPIRES;
3338 } else {
3339 expires = 0;
3340 flags = 0;
3341 ifa_flags |= IFA_F_PERMANENT;
3342 }
3343
3344 timeout = addrconf_timeout_fixup(prefered_lft, HZ);
3345 if (addrconf_finite_timeout(timeout)) {
3346 if (timeout == 0)
3347 ifa_flags |= IFA_F_DEPRECATED;
3348 prefered_lft = timeout;
3349 }
3350
3351 spin_lock_bh(&ifp->lock);
3352 ifp->flags = (ifp->flags & ~(IFA_F_DEPRECATED | IFA_F_PERMANENT | IFA_F_NODAD | IFA_F_HOMEADDRESS)) | ifa_flags;
3353 ifp->tstamp = jiffies;
3354 ifp->valid_lft = valid_lft;
3355 ifp->prefered_lft = prefered_lft;
3356
3357 spin_unlock_bh(&ifp->lock);
3358 if (!(ifp->flags&IFA_F_TENTATIVE))
3359 ipv6_ifa_notify(0, ifp);
3360
3361 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, ifp->idev->dev,
3362 expires, flags);
3363 addrconf_verify(0);
3364
3365 return 0;
3366 }
3367
3368 static int
3369 inet6_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
3370 {
3371 struct net *net = sock_net(skb->sk);
3372 struct ifaddrmsg *ifm;
3373 struct nlattr *tb[IFA_MAX+1];
3374 struct in6_addr *pfx;
3375 struct inet6_ifaddr *ifa;
3376 struct net_device *dev;
3377 u32 valid_lft = INFINITY_LIFE_TIME, preferred_lft = INFINITY_LIFE_TIME;
3378 u8 ifa_flags;
3379 int err;
3380
3381 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3382 if (err < 0)
3383 return err;
3384
3385 ifm = nlmsg_data(nlh);
3386 pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3387 if (pfx == NULL)
3388 return -EINVAL;
3389
3390 if (tb[IFA_CACHEINFO]) {
3391 struct ifa_cacheinfo *ci;
3392
3393 ci = nla_data(tb[IFA_CACHEINFO]);
3394 valid_lft = ci->ifa_valid;
3395 preferred_lft = ci->ifa_prefered;
3396 } else {
3397 preferred_lft = INFINITY_LIFE_TIME;
3398 valid_lft = INFINITY_LIFE_TIME;
3399 }
3400
3401 dev = __dev_get_by_index(net, ifm->ifa_index);
3402 if (dev == NULL)
3403 return -ENODEV;
3404
3405 /* We ignore other flags so far. */
3406 ifa_flags = ifm->ifa_flags & (IFA_F_NODAD | IFA_F_HOMEADDRESS);
3407
3408 ifa = ipv6_get_ifaddr(net, pfx, dev, 1);
3409 if (ifa == NULL) {
3410 /*
3411 * It would be best to check for !NLM_F_CREATE here but
3412 * userspace alreay relies on not having to provide this.
3413 */
3414 return inet6_addr_add(net, ifm->ifa_index, pfx,
3415 ifm->ifa_prefixlen, ifa_flags,
3416 preferred_lft, valid_lft);
3417 }
3418
3419 if (nlh->nlmsg_flags & NLM_F_EXCL ||
3420 !(nlh->nlmsg_flags & NLM_F_REPLACE))
3421 err = -EEXIST;
3422 else
3423 err = inet6_addr_modify(ifa, ifa_flags, preferred_lft, valid_lft);
3424
3425 in6_ifa_put(ifa);
3426
3427 return err;
3428 }
3429
3430 static void put_ifaddrmsg(struct nlmsghdr *nlh, u8 prefixlen, u8 flags,
3431 u8 scope, int ifindex)
3432 {
3433 struct ifaddrmsg *ifm;
3434
3435 ifm = nlmsg_data(nlh);
3436 ifm->ifa_family = AF_INET6;
3437 ifm->ifa_prefixlen = prefixlen;
3438 ifm->ifa_flags = flags;
3439 ifm->ifa_scope = scope;
3440 ifm->ifa_index = ifindex;
3441 }
3442
3443 static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
3444 unsigned long tstamp, u32 preferred, u32 valid)
3445 {
3446 struct ifa_cacheinfo ci;
3447
3448 ci.cstamp = (u32)(TIME_DELTA(cstamp, INITIAL_JIFFIES) / HZ * 100
3449 + TIME_DELTA(cstamp, INITIAL_JIFFIES) % HZ * 100 / HZ);
3450 ci.tstamp = (u32)(TIME_DELTA(tstamp, INITIAL_JIFFIES) / HZ * 100
3451 + TIME_DELTA(tstamp, INITIAL_JIFFIES) % HZ * 100 / HZ);
3452 ci.ifa_prefered = preferred;
3453 ci.ifa_valid = valid;
3454
3455 return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
3456 }
3457
3458 static inline int rt_scope(int ifa_scope)
3459 {
3460 if (ifa_scope & IFA_HOST)
3461 return RT_SCOPE_HOST;
3462 else if (ifa_scope & IFA_LINK)
3463 return RT_SCOPE_LINK;
3464 else if (ifa_scope & IFA_SITE)
3465 return RT_SCOPE_SITE;
3466 else
3467 return RT_SCOPE_UNIVERSE;
3468 }
3469
3470 static inline int inet6_ifaddr_msgsize(void)
3471 {
3472 return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
3473 + nla_total_size(16) /* IFA_ADDRESS */
3474 + nla_total_size(sizeof(struct ifa_cacheinfo));
3475 }
3476
3477 static int inet6_fill_ifaddr(struct sk_buff *skb, struct inet6_ifaddr *ifa,
3478 u32 pid, u32 seq, int event, unsigned int flags)
3479 {
3480 struct nlmsghdr *nlh;
3481 u32 preferred, valid;
3482
3483 nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3484 if (nlh == NULL)
3485 return -EMSGSIZE;
3486
3487 put_ifaddrmsg(nlh, ifa->prefix_len, ifa->flags, rt_scope(ifa->scope),
3488 ifa->idev->dev->ifindex);
3489
3490 if (!(ifa->flags&IFA_F_PERMANENT)) {
3491 preferred = ifa->prefered_lft;
3492 valid = ifa->valid_lft;
3493 if (preferred != INFINITY_LIFE_TIME) {
3494 long tval = (jiffies - ifa->tstamp)/HZ;
3495 if (preferred > tval)
3496 preferred -= tval;
3497 else
3498 preferred = 0;
3499 if (valid != INFINITY_LIFE_TIME)
3500 valid -= tval;
3501 }
3502 } else {
3503 preferred = INFINITY_LIFE_TIME;
3504 valid = INFINITY_LIFE_TIME;
3505 }
3506
3507 if (nla_put(skb, IFA_ADDRESS, 16, &ifa->addr) < 0 ||
3508 put_cacheinfo(skb, ifa->cstamp, ifa->tstamp, preferred, valid) < 0) {
3509 nlmsg_cancel(skb, nlh);
3510 return -EMSGSIZE;
3511 }
3512
3513 return nlmsg_end(skb, nlh);
3514 }
3515
3516 static int inet6_fill_ifmcaddr(struct sk_buff *skb, struct ifmcaddr6 *ifmca,
3517 u32 pid, u32 seq, int event, u16 flags)
3518 {
3519 struct nlmsghdr *nlh;
3520 u8 scope = RT_SCOPE_UNIVERSE;
3521 int ifindex = ifmca->idev->dev->ifindex;
3522
3523 if (ipv6_addr_scope(&ifmca->mca_addr) & IFA_SITE)
3524 scope = RT_SCOPE_SITE;
3525
3526 nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3527 if (nlh == NULL)
3528 return -EMSGSIZE;
3529
3530 put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3531 if (nla_put(skb, IFA_MULTICAST, 16, &ifmca->mca_addr) < 0 ||
3532 put_cacheinfo(skb, ifmca->mca_cstamp, ifmca->mca_tstamp,
3533 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3534 nlmsg_cancel(skb, nlh);
3535 return -EMSGSIZE;
3536 }
3537
3538 return nlmsg_end(skb, nlh);
3539 }
3540
3541 static int inet6_fill_ifacaddr(struct sk_buff *skb, struct ifacaddr6 *ifaca,
3542 u32 pid, u32 seq, int event, unsigned int flags)
3543 {
3544 struct nlmsghdr *nlh;
3545 u8 scope = RT_SCOPE_UNIVERSE;
3546 int ifindex = ifaca->aca_idev->dev->ifindex;
3547
3548 if (ipv6_addr_scope(&ifaca->aca_addr) & IFA_SITE)
3549 scope = RT_SCOPE_SITE;
3550
3551 nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3552 if (nlh == NULL)
3553 return -EMSGSIZE;
3554
3555 put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3556 if (nla_put(skb, IFA_ANYCAST, 16, &ifaca->aca_addr) < 0 ||
3557 put_cacheinfo(skb, ifaca->aca_cstamp, ifaca->aca_tstamp,
3558 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3559 nlmsg_cancel(skb, nlh);
3560 return -EMSGSIZE;
3561 }
3562
3563 return nlmsg_end(skb, nlh);
3564 }
3565
3566 enum addr_type_t {
3567 UNICAST_ADDR,
3568 MULTICAST_ADDR,
3569 ANYCAST_ADDR,
3570 };
3571
3572 /* called with rcu_read_lock() */
3573 static int in6_dump_addrs(struct inet6_dev *idev, struct sk_buff *skb,
3574 struct netlink_callback *cb, enum addr_type_t type,
3575 int s_ip_idx, int *p_ip_idx)
3576 {
3577 struct ifmcaddr6 *ifmca;
3578 struct ifacaddr6 *ifaca;
3579 int err = 1;
3580 int ip_idx = *p_ip_idx;
3581
3582 read_lock_bh(&idev->lock);
3583 switch (type) {
3584 case UNICAST_ADDR: {
3585 struct inet6_ifaddr *ifa;
3586
3587 /* unicast address incl. temp addr */
3588 list_for_each_entry(ifa, &idev->addr_list, if_list) {
3589 if (++ip_idx < s_ip_idx)
3590 continue;
3591 err = inet6_fill_ifaddr(skb, ifa,
3592 NETLINK_CB(cb->skb).pid,
3593 cb->nlh->nlmsg_seq,
3594 RTM_NEWADDR,
3595 NLM_F_MULTI);
3596 if (err <= 0)
3597 break;
3598 }
3599 break;
3600 }
3601 case MULTICAST_ADDR:
3602 /* multicast address */
3603 for (ifmca = idev->mc_list; ifmca;
3604 ifmca = ifmca->next, ip_idx++) {
3605 if (ip_idx < s_ip_idx)
3606 continue;
3607 err = inet6_fill_ifmcaddr(skb, ifmca,
3608 NETLINK_CB(cb->skb).pid,
3609 cb->nlh->nlmsg_seq,
3610 RTM_GETMULTICAST,
3611 NLM_F_MULTI);
3612 if (err <= 0)
3613 break;
3614 }
3615 break;
3616 case ANYCAST_ADDR:
3617 /* anycast address */
3618 for (ifaca = idev->ac_list; ifaca;
3619 ifaca = ifaca->aca_next, ip_idx++) {
3620 if (ip_idx < s_ip_idx)
3621 continue;
3622 err = inet6_fill_ifacaddr(skb, ifaca,
3623 NETLINK_CB(cb->skb).pid,
3624 cb->nlh->nlmsg_seq,
3625 RTM_GETANYCAST,
3626 NLM_F_MULTI);
3627 if (err <= 0)
3628 break;
3629 }
3630 break;
3631 default:
3632 break;
3633 }
3634 read_unlock_bh(&idev->lock);
3635 *p_ip_idx = ip_idx;
3636 return err;
3637 }
3638
3639 static int inet6_dump_addr(struct sk_buff *skb, struct netlink_callback *cb,
3640 enum addr_type_t type)
3641 {
3642 struct net *net = sock_net(skb->sk);
3643 int h, s_h;
3644 int idx, ip_idx;
3645 int s_idx, s_ip_idx;
3646 struct net_device *dev;
3647 struct inet6_dev *idev;
3648 struct hlist_head *head;
3649 struct hlist_node *node;
3650
3651 s_h = cb->args[0];
3652 s_idx = idx = cb->args[1];
3653 s_ip_idx = ip_idx = cb->args[2];
3654
3655 rcu_read_lock();
3656 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
3657 idx = 0;
3658 head = &net->dev_index_head[h];
3659 hlist_for_each_entry_rcu(dev, node, head, index_hlist) {
3660 if (idx < s_idx)
3661 goto cont;
3662 if (h > s_h || idx > s_idx)
3663 s_ip_idx = 0;
3664 ip_idx = 0;
3665 idev = __in6_dev_get(dev);
3666 if (!idev)
3667 goto cont;
3668
3669 if (in6_dump_addrs(idev, skb, cb, type,
3670 s_ip_idx, &ip_idx) <= 0)
3671 goto done;
3672 cont:
3673 idx++;
3674 }
3675 }
3676 done:
3677 rcu_read_unlock();
3678 cb->args[0] = h;
3679 cb->args[1] = idx;
3680 cb->args[2] = ip_idx;
3681
3682 return skb->len;
3683 }
3684
3685 static int inet6_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
3686 {
3687 enum addr_type_t type = UNICAST_ADDR;
3688
3689 return inet6_dump_addr(skb, cb, type);
3690 }
3691
3692 static int inet6_dump_ifmcaddr(struct sk_buff *skb, struct netlink_callback *cb)
3693 {
3694 enum addr_type_t type = MULTICAST_ADDR;
3695
3696 return inet6_dump_addr(skb, cb, type);
3697 }
3698
3699
3700 static int inet6_dump_ifacaddr(struct sk_buff *skb, struct netlink_callback *cb)
3701 {
3702 enum addr_type_t type = ANYCAST_ADDR;
3703
3704 return inet6_dump_addr(skb, cb, type);
3705 }
3706
3707 static int inet6_rtm_getaddr(struct sk_buff *in_skb, struct nlmsghdr* nlh,
3708 void *arg)
3709 {
3710 struct net *net = sock_net(in_skb->sk);
3711 struct ifaddrmsg *ifm;
3712 struct nlattr *tb[IFA_MAX+1];
3713 struct in6_addr *addr = NULL;
3714 struct net_device *dev = NULL;
3715 struct inet6_ifaddr *ifa;
3716 struct sk_buff *skb;
3717 int err;
3718
3719 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3720 if (err < 0)
3721 goto errout;
3722
3723 addr = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3724 if (addr == NULL) {
3725 err = -EINVAL;
3726 goto errout;
3727 }
3728
3729 ifm = nlmsg_data(nlh);
3730 if (ifm->ifa_index)
3731 dev = __dev_get_by_index(net, ifm->ifa_index);
3732
3733 ifa = ipv6_get_ifaddr(net, addr, dev, 1);
3734 if (!ifa) {
3735 err = -EADDRNOTAVAIL;
3736 goto errout;
3737 }
3738
3739 skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_KERNEL);
3740 if (!skb) {
3741 err = -ENOBUFS;
3742 goto errout_ifa;
3743 }
3744
3745 err = inet6_fill_ifaddr(skb, ifa, NETLINK_CB(in_skb).pid,
3746 nlh->nlmsg_seq, RTM_NEWADDR, 0);
3747 if (err < 0) {
3748 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
3749 WARN_ON(err == -EMSGSIZE);
3750 kfree_skb(skb);
3751 goto errout_ifa;
3752 }
3753 err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).pid);
3754 errout_ifa:
3755 in6_ifa_put(ifa);
3756 errout:
3757 return err;
3758 }
3759
3760 static void inet6_ifa_notify(int event, struct inet6_ifaddr *ifa)
3761 {
3762 struct sk_buff *skb;
3763 struct net *net = dev_net(ifa->idev->dev);
3764 int err = -ENOBUFS;
3765
3766 skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_ATOMIC);
3767 if (skb == NULL)
3768 goto errout;
3769
3770 err = inet6_fill_ifaddr(skb, ifa, 0, 0, event, 0);
3771 if (err < 0) {
3772 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
3773 WARN_ON(err == -EMSGSIZE);
3774 kfree_skb(skb);
3775 goto errout;
3776 }
3777 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
3778 return;
3779 errout:
3780 if (err < 0)
3781 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
3782 }
3783
3784 static inline void ipv6_store_devconf(struct ipv6_devconf *cnf,
3785 __s32 *array, int bytes)
3786 {
3787 BUG_ON(bytes < (DEVCONF_MAX * 4));
3788
3789 memset(array, 0, bytes);
3790 array[DEVCONF_FORWARDING] = cnf->forwarding;
3791 array[DEVCONF_HOPLIMIT] = cnf->hop_limit;
3792 array[DEVCONF_MTU6] = cnf->mtu6;
3793 array[DEVCONF_ACCEPT_RA] = cnf->accept_ra;
3794 array[DEVCONF_ACCEPT_REDIRECTS] = cnf->accept_redirects;
3795 array[DEVCONF_AUTOCONF] = cnf->autoconf;
3796 array[DEVCONF_DAD_TRANSMITS] = cnf->dad_transmits;
3797 array[DEVCONF_RTR_SOLICITS] = cnf->rtr_solicits;
3798 array[DEVCONF_RTR_SOLICIT_INTERVAL] = cnf->rtr_solicit_interval;
3799 array[DEVCONF_RTR_SOLICIT_DELAY] = cnf->rtr_solicit_delay;
3800 array[DEVCONF_FORCE_MLD_VERSION] = cnf->force_mld_version;
3801 #ifdef CONFIG_IPV6_PRIVACY
3802 array[DEVCONF_USE_TEMPADDR] = cnf->use_tempaddr;
3803 array[DEVCONF_TEMP_VALID_LFT] = cnf->temp_valid_lft;
3804 array[DEVCONF_TEMP_PREFERED_LFT] = cnf->temp_prefered_lft;
3805 array[DEVCONF_REGEN_MAX_RETRY] = cnf->regen_max_retry;
3806 array[DEVCONF_MAX_DESYNC_FACTOR] = cnf->max_desync_factor;
3807 #endif
3808 array[DEVCONF_MAX_ADDRESSES] = cnf->max_addresses;
3809 array[DEVCONF_ACCEPT_RA_DEFRTR] = cnf->accept_ra_defrtr;
3810 array[DEVCONF_ACCEPT_RA_PINFO] = cnf->accept_ra_pinfo;
3811 #ifdef CONFIG_IPV6_ROUTER_PREF
3812 array[DEVCONF_ACCEPT_RA_RTR_PREF] = cnf->accept_ra_rtr_pref;
3813 array[DEVCONF_RTR_PROBE_INTERVAL] = cnf->rtr_probe_interval;
3814 #ifdef CONFIG_IPV6_ROUTE_INFO
3815 array[DEVCONF_ACCEPT_RA_RT_INFO_MAX_PLEN] = cnf->accept_ra_rt_info_max_plen;
3816 #endif
3817 #endif
3818 array[DEVCONF_PROXY_NDP] = cnf->proxy_ndp;
3819 array[DEVCONF_ACCEPT_SOURCE_ROUTE] = cnf->accept_source_route;
3820 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
3821 array[DEVCONF_OPTIMISTIC_DAD] = cnf->optimistic_dad;
3822 #endif
3823 #ifdef CONFIG_IPV6_MROUTE
3824 array[DEVCONF_MC_FORWARDING] = cnf->mc_forwarding;
3825 #endif
3826 array[DEVCONF_DISABLE_IPV6] = cnf->disable_ipv6;
3827 array[DEVCONF_ACCEPT_DAD] = cnf->accept_dad;
3828 array[DEVCONF_FORCE_TLLAO] = cnf->force_tllao;
3829 }
3830
3831 static inline size_t inet6_if_nlmsg_size(void)
3832 {
3833 return NLMSG_ALIGN(sizeof(struct ifinfomsg))
3834 + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
3835 + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
3836 + nla_total_size(4) /* IFLA_MTU */
3837 + nla_total_size(4) /* IFLA_LINK */
3838 + nla_total_size( /* IFLA_PROTINFO */
3839 nla_total_size(4) /* IFLA_INET6_FLAGS */
3840 + nla_total_size(sizeof(struct ifla_cacheinfo))
3841 + nla_total_size(DEVCONF_MAX * 4) /* IFLA_INET6_CONF */
3842 + nla_total_size(IPSTATS_MIB_MAX * 8) /* IFLA_INET6_STATS */
3843 + nla_total_size(ICMP6_MIB_MAX * 8) /* IFLA_INET6_ICMP6STATS */
3844 );
3845 }
3846
3847 static inline void __snmp6_fill_stats(u64 *stats, void __percpu **mib,
3848 int items, int bytes)
3849 {
3850 int i;
3851 int pad = bytes - sizeof(u64) * items;
3852 BUG_ON(pad < 0);
3853
3854 /* Use put_unaligned() because stats may not be aligned for u64. */
3855 put_unaligned(items, &stats[0]);
3856 for (i = 1; i < items; i++)
3857 put_unaligned(snmp_fold_field(mib, i), &stats[i]);
3858
3859 memset(&stats[items], 0, pad);
3860 }
3861
3862 static void snmp6_fill_stats(u64 *stats, struct inet6_dev *idev, int attrtype,
3863 int bytes)
3864 {
3865 switch (attrtype) {
3866 case IFLA_INET6_STATS:
3867 __snmp6_fill_stats(stats, (void __percpu **)idev->stats.ipv6, IPSTATS_MIB_MAX, bytes);
3868 break;
3869 case IFLA_INET6_ICMP6STATS:
3870 __snmp6_fill_stats(stats, (void __percpu **)idev->stats.icmpv6, ICMP6_MIB_MAX, bytes);
3871 break;
3872 }
3873 }
3874
3875 static int inet6_fill_ifinfo(struct sk_buff *skb, struct inet6_dev *idev,
3876 u32 pid, u32 seq, int event, unsigned int flags)
3877 {
3878 struct net_device *dev = idev->dev;
3879 struct nlattr *nla;
3880 struct ifinfomsg *hdr;
3881 struct nlmsghdr *nlh;
3882 void *protoinfo;
3883 struct ifla_cacheinfo ci;
3884
3885 nlh = nlmsg_put(skb, pid, seq, event, sizeof(*hdr), flags);
3886 if (nlh == NULL)
3887 return -EMSGSIZE;
3888
3889 hdr = nlmsg_data(nlh);
3890 hdr->ifi_family = AF_INET6;
3891 hdr->__ifi_pad = 0;
3892 hdr->ifi_type = dev->type;
3893 hdr->ifi_index = dev->ifindex;
3894 hdr->ifi_flags = dev_get_flags(dev);
3895 hdr->ifi_change = 0;
3896
3897 NLA_PUT_STRING(skb, IFLA_IFNAME, dev->name);
3898
3899 if (dev->addr_len)
3900 NLA_PUT(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr);
3901
3902 NLA_PUT_U32(skb, IFLA_MTU, dev->mtu);
3903 if (dev->ifindex != dev->iflink)
3904 NLA_PUT_U32(skb, IFLA_LINK, dev->iflink);
3905
3906 protoinfo = nla_nest_start(skb, IFLA_PROTINFO);
3907 if (protoinfo == NULL)
3908 goto nla_put_failure;
3909
3910 NLA_PUT_U32(skb, IFLA_INET6_FLAGS, idev->if_flags);
3911
3912 ci.max_reasm_len = IPV6_MAXPLEN;
3913 ci.tstamp = (__u32)(TIME_DELTA(idev->tstamp, INITIAL_JIFFIES) / HZ * 100
3914 + TIME_DELTA(idev->tstamp, INITIAL_JIFFIES) % HZ * 100 / HZ);
3915 ci.reachable_time = idev->nd_parms->reachable_time;
3916 ci.retrans_time = idev->nd_parms->retrans_time;
3917 NLA_PUT(skb, IFLA_INET6_CACHEINFO, sizeof(ci), &ci);
3918
3919 nla = nla_reserve(skb, IFLA_INET6_CONF, DEVCONF_MAX * sizeof(s32));
3920 if (nla == NULL)
3921 goto nla_put_failure;
3922 ipv6_store_devconf(&idev->cnf, nla_data(nla), nla_len(nla));
3923
3924 /* XXX - MC not implemented */
3925
3926 nla = nla_reserve(skb, IFLA_INET6_STATS, IPSTATS_MIB_MAX * sizeof(u64));
3927 if (nla == NULL)
3928 goto nla_put_failure;
3929 snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_STATS, nla_len(nla));
3930
3931 nla = nla_reserve(skb, IFLA_INET6_ICMP6STATS, ICMP6_MIB_MAX * sizeof(u64));
3932 if (nla == NULL)
3933 goto nla_put_failure;
3934 snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_ICMP6STATS, nla_len(nla));
3935
3936 nla_nest_end(skb, protoinfo);
3937 return nlmsg_end(skb, nlh);
3938
3939 nla_put_failure:
3940 nlmsg_cancel(skb, nlh);
3941 return -EMSGSIZE;
3942 }
3943
3944 static int inet6_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
3945 {
3946 struct net *net = sock_net(skb->sk);
3947 int h, s_h;
3948 int idx = 0, s_idx;
3949 struct net_device *dev;
3950 struct inet6_dev *idev;
3951 struct hlist_head *head;
3952 struct hlist_node *node;
3953
3954 s_h = cb->args[0];
3955 s_idx = cb->args[1];
3956
3957 rcu_read_lock();
3958 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
3959 idx = 0;
3960 head = &net->dev_index_head[h];
3961 hlist_for_each_entry_rcu(dev, node, head, index_hlist) {
3962 if (idx < s_idx)
3963 goto cont;
3964 idev = __in6_dev_get(dev);
3965 if (!idev)
3966 goto cont;
3967 if (inet6_fill_ifinfo(skb, idev,
3968 NETLINK_CB(cb->skb).pid,
3969 cb->nlh->nlmsg_seq,
3970 RTM_NEWLINK, NLM_F_MULTI) <= 0)
3971 goto out;
3972 cont:
3973 idx++;
3974 }
3975 }
3976 out:
3977 rcu_read_unlock();
3978 cb->args[1] = idx;
3979 cb->args[0] = h;
3980
3981 return skb->len;
3982 }
3983
3984 void inet6_ifinfo_notify(int event, struct inet6_dev *idev)
3985 {
3986 struct sk_buff *skb;
3987 struct net *net = dev_net(idev->dev);
3988 int err = -ENOBUFS;
3989
3990 skb = nlmsg_new(inet6_if_nlmsg_size(), GFP_ATOMIC);
3991 if (skb == NULL)
3992 goto errout;
3993
3994 err = inet6_fill_ifinfo(skb, idev, 0, 0, event, 0);
3995 if (err < 0) {
3996 /* -EMSGSIZE implies BUG in inet6_if_nlmsg_size() */
3997 WARN_ON(err == -EMSGSIZE);
3998 kfree_skb(skb);
3999 goto errout;
4000 }
4001 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
4002 return;
4003 errout:
4004 if (err < 0)
4005 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
4006 }
4007
4008 static inline size_t inet6_prefix_nlmsg_size(void)
4009 {
4010 return NLMSG_ALIGN(sizeof(struct prefixmsg))
4011 + nla_total_size(sizeof(struct in6_addr))
4012 + nla_total_size(sizeof(struct prefix_cacheinfo));
4013 }
4014
4015 static int inet6_fill_prefix(struct sk_buff *skb, struct inet6_dev *idev,
4016 struct prefix_info *pinfo, u32 pid, u32 seq,
4017 int event, unsigned int flags)
4018 {
4019 struct prefixmsg *pmsg;
4020 struct nlmsghdr *nlh;
4021 struct prefix_cacheinfo ci;
4022
4023 nlh = nlmsg_put(skb, pid, seq, event, sizeof(*pmsg), flags);
4024 if (nlh == NULL)
4025 return -EMSGSIZE;
4026
4027 pmsg = nlmsg_data(nlh);
4028 pmsg->prefix_family = AF_INET6;
4029 pmsg->prefix_pad1 = 0;
4030 pmsg->prefix_pad2 = 0;
4031 pmsg->prefix_ifindex = idev->dev->ifindex;
4032 pmsg->prefix_len = pinfo->prefix_len;
4033 pmsg->prefix_type = pinfo->type;
4034 pmsg->prefix_pad3 = 0;
4035 pmsg->prefix_flags = 0;
4036 if (pinfo->onlink)
4037 pmsg->prefix_flags |= IF_PREFIX_ONLINK;
4038 if (pinfo->autoconf)
4039 pmsg->prefix_flags |= IF_PREFIX_AUTOCONF;
4040
4041 NLA_PUT(skb, PREFIX_ADDRESS, sizeof(pinfo->prefix), &pinfo->prefix);
4042
4043 ci.preferred_time = ntohl(pinfo->prefered);
4044 ci.valid_time = ntohl(pinfo->valid);
4045 NLA_PUT(skb, PREFIX_CACHEINFO, sizeof(ci), &ci);
4046
4047 return nlmsg_end(skb, nlh);
4048
4049 nla_put_failure:
4050 nlmsg_cancel(skb, nlh);
4051 return -EMSGSIZE;
4052 }
4053
4054 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
4055 struct prefix_info *pinfo)
4056 {
4057 struct sk_buff *skb;
4058 struct net *net = dev_net(idev->dev);
4059 int err = -ENOBUFS;
4060
4061 skb = nlmsg_new(inet6_prefix_nlmsg_size(), GFP_ATOMIC);
4062 if (skb == NULL)
4063 goto errout;
4064
4065 err = inet6_fill_prefix(skb, idev, pinfo, 0, 0, event, 0);
4066 if (err < 0) {
4067 /* -EMSGSIZE implies BUG in inet6_prefix_nlmsg_size() */
4068 WARN_ON(err == -EMSGSIZE);
4069 kfree_skb(skb);
4070 goto errout;
4071 }
4072 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_PREFIX, NULL, GFP_ATOMIC);
4073 return;
4074 errout:
4075 if (err < 0)
4076 rtnl_set_sk_err(net, RTNLGRP_IPV6_PREFIX, err);
4077 }
4078
4079 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
4080 {
4081 inet6_ifa_notify(event ? : RTM_NEWADDR, ifp);
4082
4083 switch (event) {
4084 case RTM_NEWADDR:
4085 /*
4086 * If the address was optimistic
4087 * we inserted the route at the start of
4088 * our DAD process, so we don't need
4089 * to do it again
4090 */
4091 if (!(ifp->rt->rt6i_node))
4092 ip6_ins_rt(ifp->rt);
4093 if (ifp->idev->cnf.forwarding)
4094 addrconf_join_anycast(ifp);
4095 break;
4096 case RTM_DELADDR:
4097 if (ifp->idev->cnf.forwarding)
4098 addrconf_leave_anycast(ifp);
4099 addrconf_leave_solict(ifp->idev, &ifp->addr);
4100 dst_hold(&ifp->rt->u.dst);
4101
4102 if (ifp->state == INET6_IFADDR_STATE_DEAD &&
4103 ip6_del_rt(ifp->rt))
4104 dst_free(&ifp->rt->u.dst);
4105 break;
4106 }
4107 }
4108
4109 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
4110 {
4111 rcu_read_lock_bh();
4112 if (likely(ifp->idev->dead == 0))
4113 __ipv6_ifa_notify(event, ifp);
4114 rcu_read_unlock_bh();
4115 }
4116
4117 #ifdef CONFIG_SYSCTL
4118
4119 static
4120 int addrconf_sysctl_forward(ctl_table *ctl, int write,
4121 void __user *buffer, size_t *lenp, loff_t *ppos)
4122 {
4123 int *valp = ctl->data;
4124 int val = *valp;
4125 loff_t pos = *ppos;
4126 int ret;
4127
4128 ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
4129
4130 if (write)
4131 ret = addrconf_fixup_forwarding(ctl, valp, val);
4132 if (ret)
4133 *ppos = pos;
4134 return ret;
4135 }
4136
4137 static void dev_disable_change(struct inet6_dev *idev)
4138 {
4139 if (!idev || !idev->dev)
4140 return;
4141
4142 if (idev->cnf.disable_ipv6)
4143 addrconf_notify(NULL, NETDEV_DOWN, idev->dev);
4144 else
4145 addrconf_notify(NULL, NETDEV_UP, idev->dev);
4146 }
4147
4148 static void addrconf_disable_change(struct net *net, __s32 newf)
4149 {
4150 struct net_device *dev;
4151 struct inet6_dev *idev;
4152
4153 rcu_read_lock();
4154 for_each_netdev_rcu(net, dev) {
4155 idev = __in6_dev_get(dev);
4156 if (idev) {
4157 int changed = (!idev->cnf.disable_ipv6) ^ (!newf);
4158 idev->cnf.disable_ipv6 = newf;
4159 if (changed)
4160 dev_disable_change(idev);
4161 }
4162 }
4163 rcu_read_unlock();
4164 }
4165
4166 static int addrconf_disable_ipv6(struct ctl_table *table, int *p, int old)
4167 {
4168 struct net *net;
4169
4170 net = (struct net *)table->extra2;
4171
4172 if (p == &net->ipv6.devconf_dflt->disable_ipv6)
4173 return 0;
4174
4175 if (!rtnl_trylock()) {
4176 /* Restore the original values before restarting */
4177 *p = old;
4178 return restart_syscall();
4179 }
4180
4181 if (p == &net->ipv6.devconf_all->disable_ipv6) {
4182 __s32 newf = net->ipv6.devconf_all->disable_ipv6;
4183 net->ipv6.devconf_dflt->disable_ipv6 = newf;
4184 addrconf_disable_change(net, newf);
4185 } else if ((!*p) ^ (!old))
4186 dev_disable_change((struct inet6_dev *)table->extra1);
4187
4188 rtnl_unlock();
4189 return 0;
4190 }
4191
4192 static
4193 int addrconf_sysctl_disable(ctl_table *ctl, int write,
4194 void __user *buffer, size_t *lenp, loff_t *ppos)
4195 {
4196 int *valp = ctl->data;
4197 int val = *valp;
4198 loff_t pos = *ppos;
4199 int ret;
4200
4201 ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
4202
4203 if (write)
4204 ret = addrconf_disable_ipv6(ctl, valp, val);
4205 if (ret)
4206 *ppos = pos;
4207 return ret;
4208 }
4209
4210 static struct addrconf_sysctl_table
4211 {
4212 struct ctl_table_header *sysctl_header;
4213 ctl_table addrconf_vars[DEVCONF_MAX+1];
4214 char *dev_name;
4215 } addrconf_sysctl __read_mostly = {
4216 .sysctl_header = NULL,
4217 .addrconf_vars = {
4218 {
4219 .procname = "forwarding",
4220 .data = &ipv6_devconf.forwarding,
4221 .maxlen = sizeof(int),
4222 .mode = 0644,
4223 .proc_handler = addrconf_sysctl_forward,
4224 },
4225 {
4226 .procname = "hop_limit",
4227 .data = &ipv6_devconf.hop_limit,
4228 .maxlen = sizeof(int),
4229 .mode = 0644,
4230 .proc_handler = proc_dointvec,
4231 },
4232 {
4233 .procname = "mtu",
4234 .data = &ipv6_devconf.mtu6,
4235 .maxlen = sizeof(int),
4236 .mode = 0644,
4237 .proc_handler = proc_dointvec,
4238 },
4239 {
4240 .procname = "accept_ra",
4241 .data = &ipv6_devconf.accept_ra,
4242 .maxlen = sizeof(int),
4243 .mode = 0644,
4244 .proc_handler = proc_dointvec,
4245 },
4246 {
4247 .procname = "accept_redirects",
4248 .data = &ipv6_devconf.accept_redirects,
4249 .maxlen = sizeof(int),
4250 .mode = 0644,
4251 .proc_handler = proc_dointvec,
4252 },
4253 {
4254 .procname = "autoconf",
4255 .data = &ipv6_devconf.autoconf,
4256 .maxlen = sizeof(int),
4257 .mode = 0644,
4258 .proc_handler = proc_dointvec,
4259 },
4260 {
4261 .procname = "dad_transmits",
4262 .data = &ipv6_devconf.dad_transmits,
4263 .maxlen = sizeof(int),
4264 .mode = 0644,
4265 .proc_handler = proc_dointvec,
4266 },
4267 {
4268 .procname = "router_solicitations",
4269 .data = &ipv6_devconf.rtr_solicits,
4270 .maxlen = sizeof(int),
4271 .mode = 0644,
4272 .proc_handler = proc_dointvec,
4273 },
4274 {
4275 .procname = "router_solicitation_interval",
4276 .data = &ipv6_devconf.rtr_solicit_interval,
4277 .maxlen = sizeof(int),
4278 .mode = 0644,
4279 .proc_handler = proc_dointvec_jiffies,
4280 },
4281 {
4282 .procname = "router_solicitation_delay",
4283 .data = &ipv6_devconf.rtr_solicit_delay,
4284 .maxlen = sizeof(int),
4285 .mode = 0644,
4286 .proc_handler = proc_dointvec_jiffies,
4287 },
4288 {
4289 .procname = "force_mld_version",
4290 .data = &ipv6_devconf.force_mld_version,
4291 .maxlen = sizeof(int),
4292 .mode = 0644,
4293 .proc_handler = proc_dointvec,
4294 },
4295 #ifdef CONFIG_IPV6_PRIVACY
4296 {
4297 .procname = "use_tempaddr",
4298 .data = &ipv6_devconf.use_tempaddr,
4299 .maxlen = sizeof(int),
4300 .mode = 0644,
4301 .proc_handler = proc_dointvec,
4302 },
4303 {
4304 .procname = "temp_valid_lft",
4305 .data = &ipv6_devconf.temp_valid_lft,
4306 .maxlen = sizeof(int),
4307 .mode = 0644,
4308 .proc_handler = proc_dointvec,
4309 },
4310 {
4311 .procname = "temp_prefered_lft",
4312 .data = &ipv6_devconf.temp_prefered_lft,
4313 .maxlen = sizeof(int),
4314 .mode = 0644,
4315 .proc_handler = proc_dointvec,
4316 },
4317 {
4318 .procname = "regen_max_retry",
4319 .data = &ipv6_devconf.regen_max_retry,
4320 .maxlen = sizeof(int),
4321 .mode = 0644,
4322 .proc_handler = proc_dointvec,
4323 },
4324 {
4325 .procname = "max_desync_factor",
4326 .data = &ipv6_devconf.max_desync_factor,
4327 .maxlen = sizeof(int),
4328 .mode = 0644,
4329 .proc_handler = proc_dointvec,
4330 },
4331 #endif
4332 {
4333 .procname = "max_addresses",
4334 .data = &ipv6_devconf.max_addresses,
4335 .maxlen = sizeof(int),
4336 .mode = 0644,
4337 .proc_handler = proc_dointvec,
4338 },
4339 {
4340 .procname = "accept_ra_defrtr",
4341 .data = &ipv6_devconf.accept_ra_defrtr,
4342 .maxlen = sizeof(int),
4343 .mode = 0644,
4344 .proc_handler = proc_dointvec,
4345 },
4346 {
4347 .procname = "accept_ra_pinfo",
4348 .data = &ipv6_devconf.accept_ra_pinfo,
4349 .maxlen = sizeof(int),
4350 .mode = 0644,
4351 .proc_handler = proc_dointvec,
4352 },
4353 #ifdef CONFIG_IPV6_ROUTER_PREF
4354 {
4355 .procname = "accept_ra_rtr_pref",
4356 .data = &ipv6_devconf.accept_ra_rtr_pref,
4357 .maxlen = sizeof(int),
4358 .mode = 0644,
4359 .proc_handler = proc_dointvec,
4360 },
4361 {
4362 .procname = "router_probe_interval",
4363 .data = &ipv6_devconf.rtr_probe_interval,
4364 .maxlen = sizeof(int),
4365 .mode = 0644,
4366 .proc_handler = proc_dointvec_jiffies,
4367 },
4368 #ifdef CONFIG_IPV6_ROUTE_INFO
4369 {
4370 .procname = "accept_ra_rt_info_max_plen",
4371 .data = &ipv6_devconf.accept_ra_rt_info_max_plen,
4372 .maxlen = sizeof(int),
4373 .mode = 0644,
4374 .proc_handler = proc_dointvec,
4375 },
4376 #endif
4377 #endif
4378 {
4379 .procname = "proxy_ndp",
4380 .data = &ipv6_devconf.proxy_ndp,
4381 .maxlen = sizeof(int),
4382 .mode = 0644,
4383 .proc_handler = proc_dointvec,
4384 },
4385 {
4386 .procname = "accept_source_route",
4387 .data = &ipv6_devconf.accept_source_route,
4388 .maxlen = sizeof(int),
4389 .mode = 0644,
4390 .proc_handler = proc_dointvec,
4391 },
4392 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
4393 {
4394 .procname = "optimistic_dad",
4395 .data = &ipv6_devconf.optimistic_dad,
4396 .maxlen = sizeof(int),
4397 .mode = 0644,
4398 .proc_handler = proc_dointvec,
4399
4400 },
4401 #endif
4402 #ifdef CONFIG_IPV6_MROUTE
4403 {
4404 .procname = "mc_forwarding",
4405 .data = &ipv6_devconf.mc_forwarding,
4406 .maxlen = sizeof(int),
4407 .mode = 0444,
4408 .proc_handler = proc_dointvec,
4409 },
4410 #endif
4411 {
4412 .procname = "disable_ipv6",
4413 .data = &ipv6_devconf.disable_ipv6,
4414 .maxlen = sizeof(int),
4415 .mode = 0644,
4416 .proc_handler = addrconf_sysctl_disable,
4417 },
4418 {
4419 .procname = "accept_dad",
4420 .data = &ipv6_devconf.accept_dad,
4421 .maxlen = sizeof(int),
4422 .mode = 0644,
4423 .proc_handler = proc_dointvec,
4424 },
4425 {
4426 .procname = "force_tllao",
4427 .data = &ipv6_devconf.force_tllao,
4428 .maxlen = sizeof(int),
4429 .mode = 0644,
4430 .proc_handler = proc_dointvec
4431 },
4432 {
4433 /* sentinel */
4434 }
4435 },
4436 };
4437
4438 static int __addrconf_sysctl_register(struct net *net, char *dev_name,
4439 struct inet6_dev *idev, struct ipv6_devconf *p)
4440 {
4441 int i;
4442 struct addrconf_sysctl_table *t;
4443
4444 #define ADDRCONF_CTL_PATH_DEV 3
4445
4446 struct ctl_path addrconf_ctl_path[] = {
4447 { .procname = "net", },
4448 { .procname = "ipv6", },
4449 { .procname = "conf", },
4450 { /* to be set */ },
4451 { },
4452 };
4453
4454
4455 t = kmemdup(&addrconf_sysctl, sizeof(*t), GFP_KERNEL);
4456 if (t == NULL)
4457 goto out;
4458
4459 for (i = 0; t->addrconf_vars[i].data; i++) {
4460 t->addrconf_vars[i].data += (char *)p - (char *)&ipv6_devconf;
4461 t->addrconf_vars[i].extra1 = idev; /* embedded; no ref */
4462 t->addrconf_vars[i].extra2 = net;
4463 }
4464
4465 /*
4466 * Make a copy of dev_name, because '.procname' is regarded as const
4467 * by sysctl and we wouldn't want anyone to change it under our feet
4468 * (see SIOCSIFNAME).
4469 */
4470 t->dev_name = kstrdup(dev_name, GFP_KERNEL);
4471 if (!t->dev_name)
4472 goto free;
4473
4474 addrconf_ctl_path[ADDRCONF_CTL_PATH_DEV].procname = t->dev_name;
4475
4476 t->sysctl_header = register_net_sysctl_table(net, addrconf_ctl_path,
4477 t->addrconf_vars);
4478 if (t->sysctl_header == NULL)
4479 goto free_procname;
4480
4481 p->sysctl = t;
4482 return 0;
4483
4484 free_procname:
4485 kfree(t->dev_name);
4486 free:
4487 kfree(t);
4488 out:
4489 return -ENOBUFS;
4490 }
4491
4492 static void __addrconf_sysctl_unregister(struct ipv6_devconf *p)
4493 {
4494 struct addrconf_sysctl_table *t;
4495
4496 if (p->sysctl == NULL)
4497 return;
4498
4499 t = p->sysctl;
4500 p->sysctl = NULL;
4501 unregister_sysctl_table(t->sysctl_header);
4502 kfree(t->dev_name);
4503 kfree(t);
4504 }
4505
4506 static void addrconf_sysctl_register(struct inet6_dev *idev)
4507 {
4508 neigh_sysctl_register(idev->dev, idev->nd_parms, "ipv6",
4509 &ndisc_ifinfo_sysctl_change);
4510 __addrconf_sysctl_register(dev_net(idev->dev), idev->dev->name,
4511 idev, &idev->cnf);
4512 }
4513
4514 static void addrconf_sysctl_unregister(struct inet6_dev *idev)
4515 {
4516 __addrconf_sysctl_unregister(&idev->cnf);
4517 neigh_sysctl_unregister(idev->nd_parms);
4518 }
4519
4520
4521 #endif
4522
4523 static int __net_init addrconf_init_net(struct net *net)
4524 {
4525 int err;
4526 struct ipv6_devconf *all, *dflt;
4527
4528 err = -ENOMEM;
4529 all = &ipv6_devconf;
4530 dflt = &ipv6_devconf_dflt;
4531
4532 if (!net_eq(net, &init_net)) {
4533 all = kmemdup(all, sizeof(ipv6_devconf), GFP_KERNEL);
4534 if (all == NULL)
4535 goto err_alloc_all;
4536
4537 dflt = kmemdup(dflt, sizeof(ipv6_devconf_dflt), GFP_KERNEL);
4538 if (dflt == NULL)
4539 goto err_alloc_dflt;
4540 } else {
4541 /* these will be inherited by all namespaces */
4542 dflt->autoconf = ipv6_defaults.autoconf;
4543 dflt->disable_ipv6 = ipv6_defaults.disable_ipv6;
4544 }
4545
4546 net->ipv6.devconf_all = all;
4547 net->ipv6.devconf_dflt = dflt;
4548
4549 #ifdef CONFIG_SYSCTL
4550 err = __addrconf_sysctl_register(net, "all", NULL, all);
4551 if (err < 0)
4552 goto err_reg_all;
4553
4554 err = __addrconf_sysctl_register(net, "default", NULL, dflt);
4555 if (err < 0)
4556 goto err_reg_dflt;
4557 #endif
4558 return 0;
4559
4560 #ifdef CONFIG_SYSCTL
4561 err_reg_dflt:
4562 __addrconf_sysctl_unregister(all);
4563 err_reg_all:
4564 kfree(dflt);
4565 #endif
4566 err_alloc_dflt:
4567 kfree(all);
4568 err_alloc_all:
4569 return err;
4570 }
4571
4572 static void __net_exit addrconf_exit_net(struct net *net)
4573 {
4574 #ifdef CONFIG_SYSCTL
4575 __addrconf_sysctl_unregister(net->ipv6.devconf_dflt);
4576 __addrconf_sysctl_unregister(net->ipv6.devconf_all);
4577 #endif
4578 if (!net_eq(net, &init_net)) {
4579 kfree(net->ipv6.devconf_dflt);
4580 kfree(net->ipv6.devconf_all);
4581 }
4582 }
4583
4584 static struct pernet_operations addrconf_ops = {
4585 .init = addrconf_init_net,
4586 .exit = addrconf_exit_net,
4587 };
4588
4589 /*
4590 * Device notifier
4591 */
4592
4593 int register_inet6addr_notifier(struct notifier_block *nb)
4594 {
4595 return atomic_notifier_chain_register(&inet6addr_chain, nb);
4596 }
4597 EXPORT_SYMBOL(register_inet6addr_notifier);
4598
4599 int unregister_inet6addr_notifier(struct notifier_block *nb)
4600 {
4601 return atomic_notifier_chain_unregister(&inet6addr_chain, nb);
4602 }
4603 EXPORT_SYMBOL(unregister_inet6addr_notifier);
4604
4605 /*
4606 * Init / cleanup code
4607 */
4608
4609 int __init addrconf_init(void)
4610 {
4611 int i, err;
4612
4613 err = ipv6_addr_label_init();
4614 if (err < 0) {
4615 printk(KERN_CRIT "IPv6 Addrconf:"
4616 " cannot initialize default policy table: %d.\n", err);
4617 return err;
4618 }
4619
4620 register_pernet_subsys(&addrconf_ops);
4621
4622 /* The addrconf netdev notifier requires that loopback_dev
4623 * has it's ipv6 private information allocated and setup
4624 * before it can bring up and give link-local addresses
4625 * to other devices which are up.
4626 *
4627 * Unfortunately, loopback_dev is not necessarily the first
4628 * entry in the global dev_base list of net devices. In fact,
4629 * it is likely to be the very last entry on that list.
4630 * So this causes the notifier registry below to try and
4631 * give link-local addresses to all devices besides loopback_dev
4632 * first, then loopback_dev, which cases all the non-loopback_dev
4633 * devices to fail to get a link-local address.
4634 *
4635 * So, as a temporary fix, allocate the ipv6 structure for
4636 * loopback_dev first by hand.
4637 * Longer term, all of the dependencies ipv6 has upon the loopback
4638 * device and it being up should be removed.
4639 */
4640 rtnl_lock();
4641 if (!ipv6_add_dev(init_net.loopback_dev))
4642 err = -ENOMEM;
4643 rtnl_unlock();
4644 if (err)
4645 goto errlo;
4646
4647 for (i = 0; i < IN6_ADDR_HSIZE; i++)
4648 INIT_HLIST_HEAD(&inet6_addr_lst[i]);
4649
4650 register_netdevice_notifier(&ipv6_dev_notf);
4651
4652 addrconf_verify(0);
4653
4654 err = __rtnl_register(PF_INET6, RTM_GETLINK, NULL, inet6_dump_ifinfo);
4655 if (err < 0)
4656 goto errout;
4657
4658 /* Only the first call to __rtnl_register can fail */
4659 __rtnl_register(PF_INET6, RTM_NEWADDR, inet6_rtm_newaddr, NULL);
4660 __rtnl_register(PF_INET6, RTM_DELADDR, inet6_rtm_deladdr, NULL);
4661 __rtnl_register(PF_INET6, RTM_GETADDR, inet6_rtm_getaddr, inet6_dump_ifaddr);
4662 __rtnl_register(PF_INET6, RTM_GETMULTICAST, NULL, inet6_dump_ifmcaddr);
4663 __rtnl_register(PF_INET6, RTM_GETANYCAST, NULL, inet6_dump_ifacaddr);
4664
4665 ipv6_addr_label_rtnl_register();
4666
4667 return 0;
4668 errout:
4669 unregister_netdevice_notifier(&ipv6_dev_notf);
4670 errlo:
4671 unregister_pernet_subsys(&addrconf_ops);
4672
4673 return err;
4674 }
4675
4676 void addrconf_cleanup(void)
4677 {
4678 struct net_device *dev;
4679 int i;
4680
4681 unregister_netdevice_notifier(&ipv6_dev_notf);
4682 unregister_pernet_subsys(&addrconf_ops);
4683
4684 rtnl_lock();
4685
4686 /* clean dev list */
4687 for_each_netdev(&init_net, dev) {
4688 if (__in6_dev_get(dev) == NULL)
4689 continue;
4690 addrconf_ifdown(dev, 1);
4691 }
4692 addrconf_ifdown(init_net.loopback_dev, 2);
4693
4694 /*
4695 * Check hash table.
4696 */
4697 spin_lock_bh(&addrconf_hash_lock);
4698 for (i = 0; i < IN6_ADDR_HSIZE; i++)
4699 WARN_ON(!hlist_empty(&inet6_addr_lst[i]));
4700 spin_unlock_bh(&addrconf_hash_lock);
4701
4702 del_timer(&addr_chk_timer);
4703 rtnl_unlock();
4704 }
Linux 2.6 ibm-acpi/thinkpad-acpi driver git tree