?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
Add linux-next specific files for 20110421
[linux-2.6/next.git] / net / ipv6 / addrconf.c
blob129d7e1f311c72f5554bce198ecf1b75e5ceb2b2
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
102 static inline u32 cstamp_delta(unsigned long cstamp)
103 {
104 return (cstamp - INITIAL_JIFFIES) * 100UL / HZ;
105 }
106
107 #define ADDRCONF_TIMER_FUZZ_MINUS (HZ > 50 ? HZ/50 : 1)
108 #define ADDRCONF_TIMER_FUZZ (HZ / 4)
109 #define ADDRCONF_TIMER_FUZZ_MAX (HZ)
110
111 #ifdef CONFIG_SYSCTL
112 static void addrconf_sysctl_register(struct inet6_dev *idev);
113 static void addrconf_sysctl_unregister(struct inet6_dev *idev);
114 #else
115 static inline void addrconf_sysctl_register(struct inet6_dev *idev)
116 {
117 }
118
119 static inline void addrconf_sysctl_unregister(struct inet6_dev *idev)
120 {
121 }
122 #endif
123
124 #ifdef CONFIG_IPV6_PRIVACY
125 static int __ipv6_regen_rndid(struct inet6_dev *idev);
126 static int __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr);
127 static void ipv6_regen_rndid(unsigned long data);
128 #endif
129
130 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev);
131 static int ipv6_count_addresses(struct inet6_dev *idev);
132
133 /*
134 * Configured unicast address hash table
135 */
136 static struct hlist_head inet6_addr_lst[IN6_ADDR_HSIZE];
137 static DEFINE_SPINLOCK(addrconf_hash_lock);
138
139 static void addrconf_verify(unsigned long);
140
141 static DEFINE_TIMER(addr_chk_timer, addrconf_verify, 0, 0);
142 static DEFINE_SPINLOCK(addrconf_verify_lock);
143
144 static void addrconf_join_anycast(struct inet6_ifaddr *ifp);
145 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp);
146
147 static void addrconf_type_change(struct net_device *dev,
148 unsigned long event);
149 static int addrconf_ifdown(struct net_device *dev, int how);
150
151 static void addrconf_dad_start(struct inet6_ifaddr *ifp, u32 flags);
152 static void addrconf_dad_timer(unsigned long data);
153 static void addrconf_dad_completed(struct inet6_ifaddr *ifp);
154 static void addrconf_dad_run(struct inet6_dev *idev);
155 static void addrconf_rs_timer(unsigned long data);
156 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
157 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
158
159 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
160 struct prefix_info *pinfo);
161 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
162 struct net_device *dev);
163
164 static ATOMIC_NOTIFIER_HEAD(inet6addr_chain);
165
166 static struct ipv6_devconf ipv6_devconf __read_mostly = {
167 .forwarding = 0,
168 .hop_limit = IPV6_DEFAULT_HOPLIMIT,
169 .mtu6 = IPV6_MIN_MTU,
170 .accept_ra = 1,
171 .accept_redirects = 1,
172 .autoconf = 1,
173 .force_mld_version = 0,
174 .dad_transmits = 1,
175 .rtr_solicits = MAX_RTR_SOLICITATIONS,
176 .rtr_solicit_interval = RTR_SOLICITATION_INTERVAL,
177 .rtr_solicit_delay = MAX_RTR_SOLICITATION_DELAY,
178 #ifdef CONFIG_IPV6_PRIVACY
179 .use_tempaddr = 0,
180 .temp_valid_lft = TEMP_VALID_LIFETIME,
181 .temp_prefered_lft = TEMP_PREFERRED_LIFETIME,
182 .regen_max_retry = REGEN_MAX_RETRY,
183 .max_desync_factor = MAX_DESYNC_FACTOR,
184 #endif
185 .max_addresses = IPV6_MAX_ADDRESSES,
186 .accept_ra_defrtr = 1,
187 .accept_ra_pinfo = 1,
188 #ifdef CONFIG_IPV6_ROUTER_PREF
189 .accept_ra_rtr_pref = 1,
190 .rtr_probe_interval = 60 * HZ,
191 #ifdef CONFIG_IPV6_ROUTE_INFO
192 .accept_ra_rt_info_max_plen = 0,
193 #endif
194 #endif
195 .proxy_ndp = 0,
196 .accept_source_route = 0, /* we do not accept RH0 by default. */
197 .disable_ipv6 = 0,
198 .accept_dad = 1,
199 };
200
201 static struct ipv6_devconf ipv6_devconf_dflt __read_mostly = {
202 .forwarding = 0,
203 .hop_limit = IPV6_DEFAULT_HOPLIMIT,
204 .mtu6 = IPV6_MIN_MTU,
205 .accept_ra = 1,
206 .accept_redirects = 1,
207 .autoconf = 1,
208 .dad_transmits = 1,
209 .rtr_solicits = MAX_RTR_SOLICITATIONS,
210 .rtr_solicit_interval = RTR_SOLICITATION_INTERVAL,
211 .rtr_solicit_delay = MAX_RTR_SOLICITATION_DELAY,
212 #ifdef CONFIG_IPV6_PRIVACY
213 .use_tempaddr = 0,
214 .temp_valid_lft = TEMP_VALID_LIFETIME,
215 .temp_prefered_lft = TEMP_PREFERRED_LIFETIME,
216 .regen_max_retry = REGEN_MAX_RETRY,
217 .max_desync_factor = MAX_DESYNC_FACTOR,
218 #endif
219 .max_addresses = IPV6_MAX_ADDRESSES,
220 .accept_ra_defrtr = 1,
221 .accept_ra_pinfo = 1,
222 #ifdef CONFIG_IPV6_ROUTER_PREF
223 .accept_ra_rtr_pref = 1,
224 .rtr_probe_interval = 60 * HZ,
225 #ifdef CONFIG_IPV6_ROUTE_INFO
226 .accept_ra_rt_info_max_plen = 0,
227 #endif
228 #endif
229 .proxy_ndp = 0,
230 .accept_source_route = 0, /* we do not accept RH0 by default. */
231 .disable_ipv6 = 0,
232 .accept_dad = 1,
233 };
234
235 /* IPv6 Wildcard Address and Loopback Address defined by RFC2553 */
236 const struct in6_addr in6addr_any = IN6ADDR_ANY_INIT;
237 const struct in6_addr in6addr_loopback = IN6ADDR_LOOPBACK_INIT;
238 const struct in6_addr in6addr_linklocal_allnodes = IN6ADDR_LINKLOCAL_ALLNODES_INIT;
239 const struct in6_addr in6addr_linklocal_allrouters = IN6ADDR_LINKLOCAL_ALLROUTERS_INIT;
240
241 /* Check if a valid qdisc is available */
242 static inline bool addrconf_qdisc_ok(const struct net_device *dev)
243 {
244 return !qdisc_tx_is_noop(dev);
245 }
246
247 /* Check if a route is valid prefix route */
248 static inline int addrconf_is_prefix_route(const struct rt6_info *rt)
249 {
250 return (rt->rt6i_flags & (RTF_GATEWAY | RTF_DEFAULT)) == 0;
251 }
252
253 static void addrconf_del_timer(struct inet6_ifaddr *ifp)
254 {
255 if (del_timer(&ifp->timer))
256 __in6_ifa_put(ifp);
257 }
258
259 enum addrconf_timer_t {
260 AC_NONE,
261 AC_DAD,
262 AC_RS,
263 };
264
265 static void addrconf_mod_timer(struct inet6_ifaddr *ifp,
266 enum addrconf_timer_t what,
267 unsigned long when)
268 {
269 if (!del_timer(&ifp->timer))
270 in6_ifa_hold(ifp);
271
272 switch (what) {
273 case AC_DAD:
274 ifp->timer.function = addrconf_dad_timer;
275 break;
276 case AC_RS:
277 ifp->timer.function = addrconf_rs_timer;
278 break;
279 default:
280 break;
281 }
282 ifp->timer.expires = jiffies + when;
283 add_timer(&ifp->timer);
284 }
285
286 static int snmp6_alloc_dev(struct inet6_dev *idev)
287 {
288 if (snmp_mib_init((void __percpu **)idev->stats.ipv6,
289 sizeof(struct ipstats_mib),
290 __alignof__(struct ipstats_mib)) < 0)
291 goto err_ip;
292 if (snmp_mib_init((void __percpu **)idev->stats.icmpv6,
293 sizeof(struct icmpv6_mib),
294 __alignof__(struct icmpv6_mib)) < 0)
295 goto err_icmp;
296 if (snmp_mib_init((void __percpu **)idev->stats.icmpv6msg,
297 sizeof(struct icmpv6msg_mib),
298 __alignof__(struct icmpv6msg_mib)) < 0)
299 goto err_icmpmsg;
300
301 return 0;
302
303 err_icmpmsg:
304 snmp_mib_free((void __percpu **)idev->stats.icmpv6);
305 err_icmp:
306 snmp_mib_free((void __percpu **)idev->stats.ipv6);
307 err_ip:
308 return -ENOMEM;
309 }
310
311 static void snmp6_free_dev(struct inet6_dev *idev)
312 {
313 snmp_mib_free((void __percpu **)idev->stats.icmpv6msg);
314 snmp_mib_free((void __percpu **)idev->stats.icmpv6);
315 snmp_mib_free((void __percpu **)idev->stats.ipv6);
316 }
317
318 /* Nobody refers to this device, we may destroy it. */
319
320 static void in6_dev_finish_destroy_rcu(struct rcu_head *head)
321 {
322 struct inet6_dev *idev = container_of(head, struct inet6_dev, rcu);
323 kfree(idev);
324 }
325
326 void in6_dev_finish_destroy(struct inet6_dev *idev)
327 {
328 struct net_device *dev = idev->dev;
329
330 WARN_ON(!list_empty(&idev->addr_list));
331 WARN_ON(idev->mc_list != NULL);
332
333 #ifdef NET_REFCNT_DEBUG
334 printk(KERN_DEBUG "in6_dev_finish_destroy: %s\n", dev ? dev->name : "NIL");
335 #endif
336 dev_put(dev);
337 if (!idev->dead) {
338 pr_warning("Freeing alive inet6 device %p\n", idev);
339 return;
340 }
341 snmp6_free_dev(idev);
342 call_rcu(&idev->rcu, in6_dev_finish_destroy_rcu);
343 }
344
345 EXPORT_SYMBOL(in6_dev_finish_destroy);
346
347 static struct inet6_dev * ipv6_add_dev(struct net_device *dev)
348 {
349 struct inet6_dev *ndev;
350
351 ASSERT_RTNL();
352
353 if (dev->mtu < IPV6_MIN_MTU)
354 return NULL;
355
356 ndev = kzalloc(sizeof(struct inet6_dev), GFP_KERNEL);
357
358 if (ndev == NULL)
359 return NULL;
360
361 rwlock_init(&ndev->lock);
362 ndev->dev = dev;
363 INIT_LIST_HEAD(&ndev->addr_list);
364
365 memcpy(&ndev->cnf, dev_net(dev)->ipv6.devconf_dflt, sizeof(ndev->cnf));
366 ndev->cnf.mtu6 = dev->mtu;
367 ndev->cnf.sysctl = NULL;
368 ndev->nd_parms = neigh_parms_alloc(dev, &nd_tbl);
369 if (ndev->nd_parms == NULL) {
370 kfree(ndev);
371 return NULL;
372 }
373 if (ndev->cnf.forwarding)
374 dev_disable_lro(dev);
375 /* We refer to the device */
376 dev_hold(dev);
377
378 if (snmp6_alloc_dev(ndev) < 0) {
379 ADBG((KERN_WARNING
380 "%s(): cannot allocate memory for statistics; dev=%s.\n",
381 __func__, dev->name));
382 neigh_parms_release(&nd_tbl, ndev->nd_parms);
383 ndev->dead = 1;
384 in6_dev_finish_destroy(ndev);
385 return NULL;
386 }
387
388 if (snmp6_register_dev(ndev) < 0) {
389 ADBG((KERN_WARNING
390 "%s(): cannot create /proc/net/dev_snmp6/%s\n",
391 __func__, dev->name));
392 neigh_parms_release(&nd_tbl, ndev->nd_parms);
393 ndev->dead = 1;
394 in6_dev_finish_destroy(ndev);
395 return NULL;
396 }
397
398 /* One reference from device. We must do this before
399 * we invoke __ipv6_regen_rndid().
400 */
401 in6_dev_hold(ndev);
402
403 if (dev->flags & (IFF_NOARP | IFF_LOOPBACK))
404 ndev->cnf.accept_dad = -1;
405
406 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
407 if (dev->type == ARPHRD_SIT && (dev->priv_flags & IFF_ISATAP)) {
408 printk(KERN_INFO
409 "%s: Disabled Multicast RS\n",
410 dev->name);
411 ndev->cnf.rtr_solicits = 0;
412 }
413 #endif
414
415 #ifdef CONFIG_IPV6_PRIVACY
416 INIT_LIST_HEAD(&ndev->tempaddr_list);
417 setup_timer(&ndev->regen_timer, ipv6_regen_rndid, (unsigned long)ndev);
418 if ((dev->flags&IFF_LOOPBACK) ||
419 dev->type == ARPHRD_TUNNEL ||
420 dev->type == ARPHRD_TUNNEL6 ||
421 dev->type == ARPHRD_SIT ||
422 dev->type == ARPHRD_NONE) {
423 ndev->cnf.use_tempaddr = -1;
424 } else {
425 in6_dev_hold(ndev);
426 ipv6_regen_rndid((unsigned long) ndev);
427 }
428 #endif
429
430 if (netif_running(dev) && addrconf_qdisc_ok(dev))
431 ndev->if_flags |= IF_READY;
432
433 ipv6_mc_init_dev(ndev);
434 ndev->tstamp = jiffies;
435 addrconf_sysctl_register(ndev);
436 /* protected by rtnl_lock */
437 rcu_assign_pointer(dev->ip6_ptr, ndev);
438
439 /* Join all-node multicast group */
440 ipv6_dev_mc_inc(dev, &in6addr_linklocal_allnodes);
441
442 return ndev;
443 }
444
445 static struct inet6_dev * ipv6_find_idev(struct net_device *dev)
446 {
447 struct inet6_dev *idev;
448
449 ASSERT_RTNL();
450
451 idev = __in6_dev_get(dev);
452 if (!idev) {
453 idev = ipv6_add_dev(dev);
454 if (!idev)
455 return NULL;
456 }
457
458 if (dev->flags&IFF_UP)
459 ipv6_mc_up(idev);
460 return idev;
461 }
462
463 #ifdef CONFIG_SYSCTL
464 static void dev_forward_change(struct inet6_dev *idev)
465 {
466 struct net_device *dev;
467 struct inet6_ifaddr *ifa;
468
469 if (!idev)
470 return;
471 dev = idev->dev;
472 if (idev->cnf.forwarding)
473 dev_disable_lro(dev);
474 if (dev && (dev->flags & IFF_MULTICAST)) {
475 if (idev->cnf.forwarding)
476 ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
477 else
478 ipv6_dev_mc_dec(dev, &in6addr_linklocal_allrouters);
479 }
480
481 list_for_each_entry(ifa, &idev->addr_list, if_list) {
482 if (ifa->flags&IFA_F_TENTATIVE)
483 continue;
484 if (idev->cnf.forwarding)
485 addrconf_join_anycast(ifa);
486 else
487 addrconf_leave_anycast(ifa);
488 }
489 }
490
491
492 static void addrconf_forward_change(struct net *net, __s32 newf)
493 {
494 struct net_device *dev;
495 struct inet6_dev *idev;
496
497 rcu_read_lock();
498 for_each_netdev_rcu(net, dev) {
499 idev = __in6_dev_get(dev);
500 if (idev) {
501 int changed = (!idev->cnf.forwarding) ^ (!newf);
502 idev->cnf.forwarding = newf;
503 if (changed)
504 dev_forward_change(idev);
505 }
506 }
507 rcu_read_unlock();
508 }
509
510 static int addrconf_fixup_forwarding(struct ctl_table *table, int *p, int old)
511 {
512 struct net *net;
513
514 net = (struct net *)table->extra2;
515 if (p == &net->ipv6.devconf_dflt->forwarding)
516 return 0;
517
518 if (!rtnl_trylock()) {
519 /* Restore the original values before restarting */
520 *p = old;
521 return restart_syscall();
522 }
523
524 if (p == &net->ipv6.devconf_all->forwarding) {
525 __s32 newf = net->ipv6.devconf_all->forwarding;
526 net->ipv6.devconf_dflt->forwarding = newf;
527 addrconf_forward_change(net, newf);
528 } else if ((!*p) ^ (!old))
529 dev_forward_change((struct inet6_dev *)table->extra1);
530 rtnl_unlock();
531
532 if (*p)
533 rt6_purge_dflt_routers(net);
534 return 1;
535 }
536 #endif
537
538 static void inet6_ifa_finish_destroy_rcu(struct rcu_head *head)
539 {
540 struct inet6_ifaddr *ifp = container_of(head, struct inet6_ifaddr, rcu);
541 kfree(ifp);
542 }
543
544 /* Nobody refers to this ifaddr, destroy it */
545 void inet6_ifa_finish_destroy(struct inet6_ifaddr *ifp)
546 {
547 WARN_ON(!hlist_unhashed(&ifp->addr_lst));
548
549 #ifdef NET_REFCNT_DEBUG
550 printk(KERN_DEBUG "inet6_ifa_finish_destroy\n");
551 #endif
552
553 in6_dev_put(ifp->idev);
554
555 if (del_timer(&ifp->timer))
556 pr_notice("Timer is still running, when freeing ifa=%p\n", ifp);
557
558 if (ifp->state != INET6_IFADDR_STATE_DEAD) {
559 pr_warning("Freeing alive inet6 address %p\n", ifp);
560 return;
561 }
562 dst_release(&ifp->rt->dst);
563
564 call_rcu(&ifp->rcu, inet6_ifa_finish_destroy_rcu);
565 }
566
567 static void
568 ipv6_link_dev_addr(struct inet6_dev *idev, struct inet6_ifaddr *ifp)
569 {
570 struct list_head *p;
571 int ifp_scope = ipv6_addr_src_scope(&ifp->addr);
572
573 /*
574 * Each device address list is sorted in order of scope -
575 * global before linklocal.
576 */
577 list_for_each(p, &idev->addr_list) {
578 struct inet6_ifaddr *ifa
579 = list_entry(p, struct inet6_ifaddr, if_list);
580 if (ifp_scope >= ipv6_addr_src_scope(&ifa->addr))
581 break;
582 }
583
584 list_add_tail(&ifp->if_list, p);
585 }
586
587 static u32 ipv6_addr_hash(const struct in6_addr *addr)
588 {
589 /*
590 * We perform the hash function over the last 64 bits of the address
591 * This will include the IEEE address token on links that support it.
592 */
593 return jhash_2words((__force u32)addr->s6_addr32[2],
594 (__force u32)addr->s6_addr32[3], 0)
595 & (IN6_ADDR_HSIZE - 1);
596 }
597
598 /* On success it returns ifp with increased reference count */
599
600 static struct inet6_ifaddr *
601 ipv6_add_addr(struct inet6_dev *idev, const struct in6_addr *addr, int pfxlen,
602 int scope, u32 flags)
603 {
604 struct inet6_ifaddr *ifa = NULL;
605 struct rt6_info *rt;
606 unsigned int hash;
607 int err = 0;
608 int addr_type = ipv6_addr_type(addr);
609
610 if (addr_type == IPV6_ADDR_ANY ||
611 addr_type & IPV6_ADDR_MULTICAST ||
612 (!(idev->dev->flags & IFF_LOOPBACK) &&
613 addr_type & IPV6_ADDR_LOOPBACK))
614 return ERR_PTR(-EADDRNOTAVAIL);
615
616 rcu_read_lock_bh();
617 if (idev->dead) {
618 err = -ENODEV; /*XXX*/
619 goto out2;
620 }
621
622 if (idev->cnf.disable_ipv6) {
623 err = -EACCES;
624 goto out2;
625 }
626
627 spin_lock(&addrconf_hash_lock);
628
629 /* Ignore adding duplicate addresses on an interface */
630 if (ipv6_chk_same_addr(dev_net(idev->dev), addr, idev->dev)) {
631 ADBG(("ipv6_add_addr: already assigned\n"));
632 err = -EEXIST;
633 goto out;
634 }
635
636 ifa = kzalloc(sizeof(struct inet6_ifaddr), GFP_ATOMIC);
637
638 if (ifa == NULL) {
639 ADBG(("ipv6_add_addr: malloc failed\n"));
640 err = -ENOBUFS;
641 goto out;
642 }
643
644 rt = addrconf_dst_alloc(idev, addr, 0);
645 if (IS_ERR(rt)) {
646 err = PTR_ERR(rt);
647 goto out;
648 }
649
650 ipv6_addr_copy(&ifa->addr, addr);
651
652 spin_lock_init(&ifa->lock);
653 spin_lock_init(&ifa->state_lock);
654 init_timer(&ifa->timer);
655 INIT_HLIST_NODE(&ifa->addr_lst);
656 ifa->timer.data = (unsigned long) ifa;
657 ifa->scope = scope;
658 ifa->prefix_len = pfxlen;
659 ifa->flags = flags | IFA_F_TENTATIVE;
660 ifa->cstamp = ifa->tstamp = jiffies;
661
662 ifa->rt = rt;
663
664 /*
665 * part one of RFC 4429, section 3.3
666 * We should not configure an address as
667 * optimistic if we do not yet know the link
668 * layer address of our nexhop router
669 */
670
671 if (rt->rt6i_nexthop == NULL)
672 ifa->flags &= ~IFA_F_OPTIMISTIC;
673
674 ifa->idev = idev;
675 in6_dev_hold(idev);
676 /* For caller */
677 in6_ifa_hold(ifa);
678
679 /* Add to big hash table */
680 hash = ipv6_addr_hash(addr);
681
682 hlist_add_head_rcu(&ifa->addr_lst, &inet6_addr_lst[hash]);
683 spin_unlock(&addrconf_hash_lock);
684
685 write_lock(&idev->lock);
686 /* Add to inet6_dev unicast addr list. */
687 ipv6_link_dev_addr(idev, ifa);
688
689 #ifdef CONFIG_IPV6_PRIVACY
690 if (ifa->flags&IFA_F_TEMPORARY) {
691 list_add(&ifa->tmp_list, &idev->tempaddr_list);
692 in6_ifa_hold(ifa);
693 }
694 #endif
695
696 in6_ifa_hold(ifa);
697 write_unlock(&idev->lock);
698 out2:
699 rcu_read_unlock_bh();
700
701 if (likely(err == 0))
702 atomic_notifier_call_chain(&inet6addr_chain, NETDEV_UP, ifa);
703 else {
704 kfree(ifa);
705 ifa = ERR_PTR(err);
706 }
707
708 return ifa;
709 out:
710 spin_unlock(&addrconf_hash_lock);
711 goto out2;
712 }
713
714 /* This function wants to get referenced ifp and releases it before return */
715
716 static void ipv6_del_addr(struct inet6_ifaddr *ifp)
717 {
718 struct inet6_ifaddr *ifa, *ifn;
719 struct inet6_dev *idev = ifp->idev;
720 int state;
721 int deleted = 0, onlink = 0;
722 unsigned long expires = jiffies;
723
724 spin_lock_bh(&ifp->state_lock);
725 state = ifp->state;
726 ifp->state = INET6_IFADDR_STATE_DEAD;
727 spin_unlock_bh(&ifp->state_lock);
728
729 if (state == INET6_IFADDR_STATE_DEAD)
730 goto out;
731
732 spin_lock_bh(&addrconf_hash_lock);
733 hlist_del_init_rcu(&ifp->addr_lst);
734 spin_unlock_bh(&addrconf_hash_lock);
735
736 write_lock_bh(&idev->lock);
737 #ifdef CONFIG_IPV6_PRIVACY
738 if (ifp->flags&IFA_F_TEMPORARY) {
739 list_del(&ifp->tmp_list);
740 if (ifp->ifpub) {
741 in6_ifa_put(ifp->ifpub);
742 ifp->ifpub = NULL;
743 }
744 __in6_ifa_put(ifp);
745 }
746 #endif
747
748 list_for_each_entry_safe(ifa, ifn, &idev->addr_list, if_list) {
749 if (ifa == ifp) {
750 list_del_init(&ifp->if_list);
751 __in6_ifa_put(ifp);
752
753 if (!(ifp->flags & IFA_F_PERMANENT) || onlink > 0)
754 break;
755 deleted = 1;
756 continue;
757 } else if (ifp->flags & IFA_F_PERMANENT) {
758 if (ipv6_prefix_equal(&ifa->addr, &ifp->addr,
759 ifp->prefix_len)) {
760 if (ifa->flags & IFA_F_PERMANENT) {
761 onlink = 1;
762 if (deleted)
763 break;
764 } else {
765 unsigned long lifetime;
766
767 if (!onlink)
768 onlink = -1;
769
770 spin_lock(&ifa->lock);
771
772 lifetime = addrconf_timeout_fixup(ifa->valid_lft, HZ);
773 /*
774 * Note: Because this address is
775 * not permanent, lifetime <
776 * LONG_MAX / HZ here.
777 */
778 if (time_before(expires,
779 ifa->tstamp + lifetime * HZ))
780 expires = ifa->tstamp + lifetime * HZ;
781 spin_unlock(&ifa->lock);
782 }
783 }
784 }
785 }
786 write_unlock_bh(&idev->lock);
787
788 addrconf_del_timer(ifp);
789
790 ipv6_ifa_notify(RTM_DELADDR, ifp);
791
792 atomic_notifier_call_chain(&inet6addr_chain, NETDEV_DOWN, ifp);
793
794 /*
795 * Purge or update corresponding prefix
796 *
797 * 1) we don't purge prefix here if address was not permanent.
798 * prefix is managed by its own lifetime.
799 * 2) if there're no addresses, delete prefix.
800 * 3) if there're still other permanent address(es),
801 * corresponding prefix is still permanent.
802 * 4) otherwise, update prefix lifetime to the
803 * longest valid lifetime among the corresponding
804 * addresses on the device.
805 * Note: subsequent RA will update lifetime.
806 *
807 * --yoshfuji
808 */
809 if ((ifp->flags & IFA_F_PERMANENT) && onlink < 1) {
810 struct in6_addr prefix;
811 struct rt6_info *rt;
812 struct net *net = dev_net(ifp->idev->dev);
813 ipv6_addr_prefix(&prefix, &ifp->addr, ifp->prefix_len);
814 rt = rt6_lookup(net, &prefix, NULL, ifp->idev->dev->ifindex, 1);
815
816 if (rt && addrconf_is_prefix_route(rt)) {
817 if (onlink == 0) {
818 ip6_del_rt(rt);
819 rt = NULL;
820 } else if (!(rt->rt6i_flags & RTF_EXPIRES)) {
821 rt->rt6i_expires = expires;
822 rt->rt6i_flags |= RTF_EXPIRES;
823 }
824 }
825 dst_release(&rt->dst);
826 }
827
828 /* clean up prefsrc entries */
829 rt6_remove_prefsrc(ifp);
830 out:
831 in6_ifa_put(ifp);
832 }
833
834 #ifdef CONFIG_IPV6_PRIVACY
835 static int ipv6_create_tempaddr(struct inet6_ifaddr *ifp, struct inet6_ifaddr *ift)
836 {
837 struct inet6_dev *idev = ifp->idev;
838 struct in6_addr addr, *tmpaddr;
839 unsigned long tmp_prefered_lft, tmp_valid_lft, tmp_cstamp, tmp_tstamp, age;
840 unsigned long regen_advance;
841 int tmp_plen;
842 int ret = 0;
843 int max_addresses;
844 u32 addr_flags;
845
846 write_lock(&idev->lock);
847 if (ift) {
848 spin_lock_bh(&ift->lock);
849 memcpy(&addr.s6_addr[8], &ift->addr.s6_addr[8], 8);
850 spin_unlock_bh(&ift->lock);
851 tmpaddr = &addr;
852 } else {
853 tmpaddr = NULL;
854 }
855 retry:
856 in6_dev_hold(idev);
857 if (idev->cnf.use_tempaddr <= 0) {
858 write_unlock(&idev->lock);
859 printk(KERN_INFO
860 "ipv6_create_tempaddr(): use_tempaddr is disabled.\n");
861 in6_dev_put(idev);
862 ret = -1;
863 goto out;
864 }
865 spin_lock_bh(&ifp->lock);
866 if (ifp->regen_count++ >= idev->cnf.regen_max_retry) {
867 idev->cnf.use_tempaddr = -1; /*XXX*/
868 spin_unlock_bh(&ifp->lock);
869 write_unlock(&idev->lock);
870 printk(KERN_WARNING
871 "ipv6_create_tempaddr(): regeneration time exceeded. disabled temporary address support.\n");
872 in6_dev_put(idev);
873 ret = -1;
874 goto out;
875 }
876 in6_ifa_hold(ifp);
877 memcpy(addr.s6_addr, ifp->addr.s6_addr, 8);
878 if (__ipv6_try_regen_rndid(idev, tmpaddr) < 0) {
879 spin_unlock_bh(&ifp->lock);
880 write_unlock(&idev->lock);
881 printk(KERN_WARNING
882 "ipv6_create_tempaddr(): regeneration of randomized interface id failed.\n");
883 in6_ifa_put(ifp);
884 in6_dev_put(idev);
885 ret = -1;
886 goto out;
887 }
888 memcpy(&addr.s6_addr[8], idev->rndid, 8);
889 age = (jiffies - ifp->tstamp) / HZ;
890 tmp_valid_lft = min_t(__u32,
891 ifp->valid_lft,
892 idev->cnf.temp_valid_lft + age);
893 tmp_prefered_lft = min_t(__u32,
894 ifp->prefered_lft,
895 idev->cnf.temp_prefered_lft + age -
896 idev->cnf.max_desync_factor);
897 tmp_plen = ifp->prefix_len;
898 max_addresses = idev->cnf.max_addresses;
899 tmp_cstamp = ifp->cstamp;
900 tmp_tstamp = ifp->tstamp;
901 spin_unlock_bh(&ifp->lock);
902
903 regen_advance = idev->cnf.regen_max_retry *
904 idev->cnf.dad_transmits *
905 idev->nd_parms->retrans_time / HZ;
906 write_unlock(&idev->lock);
907
908 /* A temporary address is created only if this calculated Preferred
909 * Lifetime is greater than REGEN_ADVANCE time units. In particular,
910 * an implementation must not create a temporary address with a zero
911 * Preferred Lifetime.
912 */
913 if (tmp_prefered_lft <= regen_advance) {
914 in6_ifa_put(ifp);
915 in6_dev_put(idev);
916 ret = -1;
917 goto out;
918 }
919
920 addr_flags = IFA_F_TEMPORARY;
921 /* set in addrconf_prefix_rcv() */
922 if (ifp->flags & IFA_F_OPTIMISTIC)
923 addr_flags |= IFA_F_OPTIMISTIC;
924
925 ift = !max_addresses ||
926 ipv6_count_addresses(idev) < max_addresses ?
927 ipv6_add_addr(idev, &addr, tmp_plen,
928 ipv6_addr_type(&addr)&IPV6_ADDR_SCOPE_MASK,
929 addr_flags) : NULL;
930 if (!ift || IS_ERR(ift)) {
931 in6_ifa_put(ifp);
932 in6_dev_put(idev);
933 printk(KERN_INFO
934 "ipv6_create_tempaddr(): retry temporary address regeneration.\n");
935 tmpaddr = &addr;
936 write_lock(&idev->lock);
937 goto retry;
938 }
939
940 spin_lock_bh(&ift->lock);
941 ift->ifpub = ifp;
942 ift->valid_lft = tmp_valid_lft;
943 ift->prefered_lft = tmp_prefered_lft;
944 ift->cstamp = tmp_cstamp;
945 ift->tstamp = tmp_tstamp;
946 spin_unlock_bh(&ift->lock);
947
948 addrconf_dad_start(ift, 0);
949 in6_ifa_put(ift);
950 in6_dev_put(idev);
951 out:
952 return ret;
953 }
954 #endif
955
956 /*
957 * Choose an appropriate source address (RFC3484)
958 */
959 enum {
960 IPV6_SADDR_RULE_INIT = 0,
961 IPV6_SADDR_RULE_LOCAL,
962 IPV6_SADDR_RULE_SCOPE,
963 IPV6_SADDR_RULE_PREFERRED,
964 #ifdef CONFIG_IPV6_MIP6
965 IPV6_SADDR_RULE_HOA,
966 #endif
967 IPV6_SADDR_RULE_OIF,
968 IPV6_SADDR_RULE_LABEL,
969 #ifdef CONFIG_IPV6_PRIVACY
970 IPV6_SADDR_RULE_PRIVACY,
971 #endif
972 IPV6_SADDR_RULE_ORCHID,
973 IPV6_SADDR_RULE_PREFIX,
974 IPV6_SADDR_RULE_MAX
975 };
976
977 struct ipv6_saddr_score {
978 int rule;
979 int addr_type;
980 struct inet6_ifaddr *ifa;
981 DECLARE_BITMAP(scorebits, IPV6_SADDR_RULE_MAX);
982 int scopedist;
983 int matchlen;
984 };
985
986 struct ipv6_saddr_dst {
987 const struct in6_addr *addr;
988 int ifindex;
989 int scope;
990 int label;
991 unsigned int prefs;
992 };
993
994 static inline int ipv6_saddr_preferred(int type)
995 {
996 if (type & (IPV6_ADDR_MAPPED|IPV6_ADDR_COMPATv4|IPV6_ADDR_LOOPBACK))
997 return 1;
998 return 0;
999 }
1000
1001 static int ipv6_get_saddr_eval(struct net *net,
1002 struct ipv6_saddr_score *score,
1003 struct ipv6_saddr_dst *dst,
1004 int i)
1005 {
1006 int ret;
1007
1008 if (i <= score->rule) {
1009 switch (i) {
1010 case IPV6_SADDR_RULE_SCOPE:
1011 ret = score->scopedist;
1012 break;
1013 case IPV6_SADDR_RULE_PREFIX:
1014 ret = score->matchlen;
1015 break;
1016 default:
1017 ret = !!test_bit(i, score->scorebits);
1018 }
1019 goto out;
1020 }
1021
1022 switch (i) {
1023 case IPV6_SADDR_RULE_INIT:
1024 /* Rule 0: remember if hiscore is not ready yet */
1025 ret = !!score->ifa;
1026 break;
1027 case IPV6_SADDR_RULE_LOCAL:
1028 /* Rule 1: Prefer same address */
1029 ret = ipv6_addr_equal(&score->ifa->addr, dst->addr);
1030 break;
1031 case IPV6_SADDR_RULE_SCOPE:
1032 /* Rule 2: Prefer appropriate scope
1033 *
1034 * ret
1035 * ^
1036 * -1 | d 15
1037 * ---+--+-+---> scope
1038 * |
1039 * | d is scope of the destination.
1040 * B-d | \
1041 * | \ <- smaller scope is better if
1042 * B-15 | \ if scope is enough for destinaion.
1043 * | ret = B - scope (-1 <= scope >= d <= 15).
1044 * d-C-1 | /
1045 * |/ <- greater is better
1046 * -C / if scope is not enough for destination.
1047 * /| ret = scope - C (-1 <= d < scope <= 15).
1048 *
1049 * d - C - 1 < B -15 (for all -1 <= d <= 15).
1050 * C > d + 14 - B >= 15 + 14 - B = 29 - B.
1051 * Assume B = 0 and we get C > 29.
1052 */
1053 ret = __ipv6_addr_src_scope(score->addr_type);
1054 if (ret >= dst->scope)
1055 ret = -ret;
1056 else
1057 ret -= 128; /* 30 is enough */
1058 score->scopedist = ret;
1059 break;
1060 case IPV6_SADDR_RULE_PREFERRED:
1061 /* Rule 3: Avoid deprecated and optimistic addresses */
1062 ret = ipv6_saddr_preferred(score->addr_type) ||
1063 !(score->ifa->flags & (IFA_F_DEPRECATED|IFA_F_OPTIMISTIC));
1064 break;
1065 #ifdef CONFIG_IPV6_MIP6
1066 case IPV6_SADDR_RULE_HOA:
1067 {
1068 /* Rule 4: Prefer home address */
1069 int prefhome = !(dst->prefs & IPV6_PREFER_SRC_COA);
1070 ret = !(score->ifa->flags & IFA_F_HOMEADDRESS) ^ prefhome;
1071 break;
1072 }
1073 #endif
1074 case IPV6_SADDR_RULE_OIF:
1075 /* Rule 5: Prefer outgoing interface */
1076 ret = (!dst->ifindex ||
1077 dst->ifindex == score->ifa->idev->dev->ifindex);
1078 break;
1079 case IPV6_SADDR_RULE_LABEL:
1080 /* Rule 6: Prefer matching label */
1081 ret = ipv6_addr_label(net,
1082 &score->ifa->addr, score->addr_type,
1083 score->ifa->idev->dev->ifindex) == dst->label;
1084 break;
1085 #ifdef CONFIG_IPV6_PRIVACY
1086 case IPV6_SADDR_RULE_PRIVACY:
1087 {
1088 /* Rule 7: Prefer public address
1089 * Note: prefer temporary address if use_tempaddr >= 2
1090 */
1091 int preftmp = dst->prefs & (IPV6_PREFER_SRC_PUBLIC|IPV6_PREFER_SRC_TMP) ?
1092 !!(dst->prefs & IPV6_PREFER_SRC_TMP) :
1093 score->ifa->idev->cnf.use_tempaddr >= 2;
1094 ret = (!(score->ifa->flags & IFA_F_TEMPORARY)) ^ preftmp;
1095 break;
1096 }
1097 #endif
1098 case IPV6_SADDR_RULE_ORCHID:
1099 /* Rule 8-: Prefer ORCHID vs ORCHID or
1100 * non-ORCHID vs non-ORCHID
1101 */
1102 ret = !(ipv6_addr_orchid(&score->ifa->addr) ^
1103 ipv6_addr_orchid(dst->addr));
1104 break;
1105 case IPV6_SADDR_RULE_PREFIX:
1106 /* Rule 8: Use longest matching prefix */
1107 score->matchlen = ret = ipv6_addr_diff(&score->ifa->addr,
1108 dst->addr);
1109 break;
1110 default:
1111 ret = 0;
1112 }
1113
1114 if (ret)
1115 __set_bit(i, score->scorebits);
1116 score->rule = i;
1117 out:
1118 return ret;
1119 }
1120
1121 int ipv6_dev_get_saddr(struct net *net, struct net_device *dst_dev,
1122 const struct in6_addr *daddr, unsigned int prefs,
1123 struct in6_addr *saddr)
1124 {
1125 struct ipv6_saddr_score scores[2],
1126 *score = &scores[0], *hiscore = &scores[1];
1127 struct ipv6_saddr_dst dst;
1128 struct net_device *dev;
1129 int dst_type;
1130
1131 dst_type = __ipv6_addr_type(daddr);
1132 dst.addr = daddr;
1133 dst.ifindex = dst_dev ? dst_dev->ifindex : 0;
1134 dst.scope = __ipv6_addr_src_scope(dst_type);
1135 dst.label = ipv6_addr_label(net, daddr, dst_type, dst.ifindex);
1136 dst.prefs = prefs;
1137
1138 hiscore->rule = -1;
1139 hiscore->ifa = NULL;
1140
1141 rcu_read_lock();
1142
1143 for_each_netdev_rcu(net, dev) {
1144 struct inet6_dev *idev;
1145
1146 /* Candidate Source Address (section 4)
1147 * - multicast and link-local destination address,
1148 * the set of candidate source address MUST only
1149 * include addresses assigned to interfaces
1150 * belonging to the same link as the outgoing
1151 * interface.
1152 * (- For site-local destination addresses, the
1153 * set of candidate source addresses MUST only
1154 * include addresses assigned to interfaces
1155 * belonging to the same site as the outgoing
1156 * interface.)
1157 */
1158 if (((dst_type & IPV6_ADDR_MULTICAST) ||
1159 dst.scope <= IPV6_ADDR_SCOPE_LINKLOCAL) &&
1160 dst.ifindex && dev->ifindex != dst.ifindex)
1161 continue;
1162
1163 idev = __in6_dev_get(dev);
1164 if (!idev)
1165 continue;
1166
1167 read_lock_bh(&idev->lock);
1168 list_for_each_entry(score->ifa, &idev->addr_list, if_list) {
1169 int i;
1170
1171 /*
1172 * - Tentative Address (RFC2462 section 5.4)
1173 * - A tentative address is not considered
1174 * "assigned to an interface" in the traditional
1175 * sense, unless it is also flagged as optimistic.
1176 * - Candidate Source Address (section 4)
1177 * - In any case, anycast addresses, multicast
1178 * addresses, and the unspecified address MUST
1179 * NOT be included in a candidate set.
1180 */
1181 if ((score->ifa->flags & IFA_F_TENTATIVE) &&
1182 (!(score->ifa->flags & IFA_F_OPTIMISTIC)))
1183 continue;
1184
1185 score->addr_type = __ipv6_addr_type(&score->ifa->addr);
1186
1187 if (unlikely(score->addr_type == IPV6_ADDR_ANY ||
1188 score->addr_type & IPV6_ADDR_MULTICAST)) {
1189 LIMIT_NETDEBUG(KERN_DEBUG
1190 "ADDRCONF: unspecified / multicast address "
1191 "assigned as unicast address on %s",
1192 dev->name);
1193 continue;
1194 }
1195
1196 score->rule = -1;
1197 bitmap_zero(score->scorebits, IPV6_SADDR_RULE_MAX);
1198
1199 for (i = 0; i < IPV6_SADDR_RULE_MAX; i++) {
1200 int minihiscore, miniscore;
1201
1202 minihiscore = ipv6_get_saddr_eval(net, hiscore, &dst, i);
1203 miniscore = ipv6_get_saddr_eval(net, score, &dst, i);
1204
1205 if (minihiscore > miniscore) {
1206 if (i == IPV6_SADDR_RULE_SCOPE &&
1207 score->scopedist > 0) {
1208 /*
1209 * special case:
1210 * each remaining entry
1211 * has too small (not enough)
1212 * scope, because ifa entries
1213 * are sorted by their scope
1214 * values.
1215 */
1216 goto try_nextdev;
1217 }
1218 break;
1219 } else if (minihiscore < miniscore) {
1220 if (hiscore->ifa)
1221 in6_ifa_put(hiscore->ifa);
1222
1223 in6_ifa_hold(score->ifa);
1224
1225 swap(hiscore, score);
1226
1227 /* restore our iterator */
1228 score->ifa = hiscore->ifa;
1229
1230 break;
1231 }
1232 }
1233 }
1234 try_nextdev:
1235 read_unlock_bh(&idev->lock);
1236 }
1237 rcu_read_unlock();
1238
1239 if (!hiscore->ifa)
1240 return -EADDRNOTAVAIL;
1241
1242 ipv6_addr_copy(saddr, &hiscore->ifa->addr);
1243 in6_ifa_put(hiscore->ifa);
1244 return 0;
1245 }
1246 EXPORT_SYMBOL(ipv6_dev_get_saddr);
1247
1248 int ipv6_get_lladdr(struct net_device *dev, struct in6_addr *addr,
1249 unsigned char banned_flags)
1250 {
1251 struct inet6_dev *idev;
1252 int err = -EADDRNOTAVAIL;
1253
1254 rcu_read_lock();
1255 idev = __in6_dev_get(dev);
1256 if (idev) {
1257 struct inet6_ifaddr *ifp;
1258
1259 read_lock_bh(&idev->lock);
1260 list_for_each_entry(ifp, &idev->addr_list, if_list) {
1261 if (ifp->scope == IFA_LINK &&
1262 !(ifp->flags & banned_flags)) {
1263 ipv6_addr_copy(addr, &ifp->addr);
1264 err = 0;
1265 break;
1266 }
1267 }
1268 read_unlock_bh(&idev->lock);
1269 }
1270 rcu_read_unlock();
1271 return err;
1272 }
1273
1274 static int ipv6_count_addresses(struct inet6_dev *idev)
1275 {
1276 int cnt = 0;
1277 struct inet6_ifaddr *ifp;
1278
1279 read_lock_bh(&idev->lock);
1280 list_for_each_entry(ifp, &idev->addr_list, if_list)
1281 cnt++;
1282 read_unlock_bh(&idev->lock);
1283 return cnt;
1284 }
1285
1286 int ipv6_chk_addr(struct net *net, struct in6_addr *addr,
1287 struct net_device *dev, int strict)
1288 {
1289 struct inet6_ifaddr *ifp;
1290 struct hlist_node *node;
1291 unsigned int hash = ipv6_addr_hash(addr);
1292
1293 rcu_read_lock_bh();
1294 hlist_for_each_entry_rcu(ifp, node, &inet6_addr_lst[hash], addr_lst) {
1295 if (!net_eq(dev_net(ifp->idev->dev), net))
1296 continue;
1297 if (ipv6_addr_equal(&ifp->addr, addr) &&
1298 !(ifp->flags&IFA_F_TENTATIVE) &&
1299 (dev == NULL || ifp->idev->dev == dev ||
1300 !(ifp->scope&(IFA_LINK|IFA_HOST) || strict))) {
1301 rcu_read_unlock_bh();
1302 return 1;
1303 }
1304 }
1305
1306 rcu_read_unlock_bh();
1307 return 0;
1308 }
1309 EXPORT_SYMBOL(ipv6_chk_addr);
1310
1311 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
1312 struct net_device *dev)
1313 {
1314 unsigned int hash = ipv6_addr_hash(addr);
1315 struct inet6_ifaddr *ifp;
1316 struct hlist_node *node;
1317
1318 hlist_for_each_entry(ifp, node, &inet6_addr_lst[hash], addr_lst) {
1319 if (!net_eq(dev_net(ifp->idev->dev), net))
1320 continue;
1321 if (ipv6_addr_equal(&ifp->addr, addr)) {
1322 if (dev == NULL || ifp->idev->dev == dev)
1323 return true;
1324 }
1325 }
1326 return false;
1327 }
1328
1329 int ipv6_chk_prefix(struct in6_addr *addr, struct net_device *dev)
1330 {
1331 struct inet6_dev *idev;
1332 struct inet6_ifaddr *ifa;
1333 int onlink;
1334
1335 onlink = 0;
1336 rcu_read_lock();
1337 idev = __in6_dev_get(dev);
1338 if (idev) {
1339 read_lock_bh(&idev->lock);
1340 list_for_each_entry(ifa, &idev->addr_list, if_list) {
1341 onlink = ipv6_prefix_equal(addr, &ifa->addr,
1342 ifa->prefix_len);
1343 if (onlink)
1344 break;
1345 }
1346 read_unlock_bh(&idev->lock);
1347 }
1348 rcu_read_unlock();
1349 return onlink;
1350 }
1351
1352 EXPORT_SYMBOL(ipv6_chk_prefix);
1353
1354 struct inet6_ifaddr *ipv6_get_ifaddr(struct net *net, const struct in6_addr *addr,
1355 struct net_device *dev, int strict)
1356 {
1357 struct inet6_ifaddr *ifp, *result = NULL;
1358 unsigned int hash = ipv6_addr_hash(addr);
1359 struct hlist_node *node;
1360
1361 rcu_read_lock_bh();
1362 hlist_for_each_entry_rcu_bh(ifp, node, &inet6_addr_lst[hash], addr_lst) {
1363 if (!net_eq(dev_net(ifp->idev->dev), net))
1364 continue;
1365 if (ipv6_addr_equal(&ifp->addr, addr)) {
1366 if (dev == NULL || ifp->idev->dev == dev ||
1367 !(ifp->scope&(IFA_LINK|IFA_HOST) || strict)) {
1368 result = ifp;
1369 in6_ifa_hold(ifp);
1370 break;
1371 }
1372 }
1373 }
1374 rcu_read_unlock_bh();
1375
1376 return result;
1377 }
1378
1379 /* Gets referenced address, destroys ifaddr */
1380
1381 static void addrconf_dad_stop(struct inet6_ifaddr *ifp, int dad_failed)
1382 {
1383 if (ifp->flags&IFA_F_PERMANENT) {
1384 spin_lock_bh(&ifp->lock);
1385 addrconf_del_timer(ifp);
1386 ifp->flags |= IFA_F_TENTATIVE;
1387 if (dad_failed)
1388 ifp->flags |= IFA_F_DADFAILED;
1389 spin_unlock_bh(&ifp->lock);
1390 if (dad_failed)
1391 ipv6_ifa_notify(0, ifp);
1392 in6_ifa_put(ifp);
1393 #ifdef CONFIG_IPV6_PRIVACY
1394 } else if (ifp->flags&IFA_F_TEMPORARY) {
1395 struct inet6_ifaddr *ifpub;
1396 spin_lock_bh(&ifp->lock);
1397 ifpub = ifp->ifpub;
1398 if (ifpub) {
1399 in6_ifa_hold(ifpub);
1400 spin_unlock_bh(&ifp->lock);
1401 ipv6_create_tempaddr(ifpub, ifp);
1402 in6_ifa_put(ifpub);
1403 } else {
1404 spin_unlock_bh(&ifp->lock);
1405 }
1406 ipv6_del_addr(ifp);
1407 #endif
1408 } else
1409 ipv6_del_addr(ifp);
1410 }
1411
1412 static int addrconf_dad_end(struct inet6_ifaddr *ifp)
1413 {
1414 int err = -ENOENT;
1415
1416 spin_lock(&ifp->state_lock);
1417 if (ifp->state == INET6_IFADDR_STATE_DAD) {
1418 ifp->state = INET6_IFADDR_STATE_POSTDAD;
1419 err = 0;
1420 }
1421 spin_unlock(&ifp->state_lock);
1422
1423 return err;
1424 }
1425
1426 void addrconf_dad_failure(struct inet6_ifaddr *ifp)
1427 {
1428 struct inet6_dev *idev = ifp->idev;
1429
1430 if (addrconf_dad_end(ifp)) {
1431 in6_ifa_put(ifp);
1432 return;
1433 }
1434
1435 if (net_ratelimit())
1436 printk(KERN_INFO "%s: IPv6 duplicate address %pI6c detected!\n",
1437 ifp->idev->dev->name, &ifp->addr);
1438
1439 if (idev->cnf.accept_dad > 1 && !idev->cnf.disable_ipv6) {
1440 struct in6_addr addr;
1441
1442 addr.s6_addr32[0] = htonl(0xfe800000);
1443 addr.s6_addr32[1] = 0;
1444
1445 if (!ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) &&
1446 ipv6_addr_equal(&ifp->addr, &addr)) {
1447 /* DAD failed for link-local based on MAC address */
1448 idev->cnf.disable_ipv6 = 1;
1449
1450 printk(KERN_INFO "%s: IPv6 being disabled!\n",
1451 ifp->idev->dev->name);
1452 }
1453 }
1454
1455 addrconf_dad_stop(ifp, 1);
1456 }
1457
1458 /* Join to solicited addr multicast group. */
1459
1460 void addrconf_join_solict(struct net_device *dev, struct in6_addr *addr)
1461 {
1462 struct in6_addr maddr;
1463
1464 if (dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1465 return;
1466
1467 addrconf_addr_solict_mult(addr, &maddr);
1468 ipv6_dev_mc_inc(dev, &maddr);
1469 }
1470
1471 void addrconf_leave_solict(struct inet6_dev *idev, struct in6_addr *addr)
1472 {
1473 struct in6_addr maddr;
1474
1475 if (idev->dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1476 return;
1477
1478 addrconf_addr_solict_mult(addr, &maddr);
1479 __ipv6_dev_mc_dec(idev, &maddr);
1480 }
1481
1482 static void addrconf_join_anycast(struct inet6_ifaddr *ifp)
1483 {
1484 struct in6_addr addr;
1485 ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1486 if (ipv6_addr_any(&addr))
1487 return;
1488 ipv6_dev_ac_inc(ifp->idev->dev, &addr);
1489 }
1490
1491 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp)
1492 {
1493 struct in6_addr addr;
1494 ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1495 if (ipv6_addr_any(&addr))
1496 return;
1497 __ipv6_dev_ac_dec(ifp->idev, &addr);
1498 }
1499
1500 static int addrconf_ifid_eui48(u8 *eui, struct net_device *dev)
1501 {
1502 if (dev->addr_len != ETH_ALEN)
1503 return -1;
1504 memcpy(eui, dev->dev_addr, 3);
1505 memcpy(eui + 5, dev->dev_addr + 3, 3);
1506
1507 /*
1508 * The zSeries OSA network cards can be shared among various
1509 * OS instances, but the OSA cards have only one MAC address.
1510 * This leads to duplicate address conflicts in conjunction
1511 * with IPv6 if more than one instance uses the same card.
1512 *
1513 * The driver for these cards can deliver a unique 16-bit
1514 * identifier for each instance sharing the same card. It is
1515 * placed instead of 0xFFFE in the interface identifier. The
1516 * "u" bit of the interface identifier is not inverted in this
1517 * case. Hence the resulting interface identifier has local
1518 * scope according to RFC2373.
1519 */
1520 if (dev->dev_id) {
1521 eui[3] = (dev->dev_id >> 8) & 0xFF;
1522 eui[4] = dev->dev_id & 0xFF;
1523 } else {
1524 eui[3] = 0xFF;
1525 eui[4] = 0xFE;
1526 eui[0] ^= 2;
1527 }
1528 return 0;
1529 }
1530
1531 static int addrconf_ifid_arcnet(u8 *eui, struct net_device *dev)
1532 {
1533 /* XXX: inherit EUI-64 from other interface -- yoshfuji */
1534 if (dev->addr_len != ARCNET_ALEN)
1535 return -1;
1536 memset(eui, 0, 7);
1537 eui[7] = *(u8*)dev->dev_addr;
1538 return 0;
1539 }
1540
1541 static int addrconf_ifid_infiniband(u8 *eui, struct net_device *dev)
1542 {
1543 if (dev->addr_len != INFINIBAND_ALEN)
1544 return -1;
1545 memcpy(eui, dev->dev_addr + 12, 8);
1546 eui[0] |= 2;
1547 return 0;
1548 }
1549
1550 static int __ipv6_isatap_ifid(u8 *eui, __be32 addr)
1551 {
1552 if (addr == 0)
1553 return -1;
1554 eui[0] = (ipv4_is_zeronet(addr) || ipv4_is_private_10(addr) ||
1555 ipv4_is_loopback(addr) || ipv4_is_linklocal_169(addr) ||
1556 ipv4_is_private_172(addr) || ipv4_is_test_192(addr) ||
1557 ipv4_is_anycast_6to4(addr) || ipv4_is_private_192(addr) ||
1558 ipv4_is_test_198(addr) || ipv4_is_multicast(addr) ||
1559 ipv4_is_lbcast(addr)) ? 0x00 : 0x02;
1560 eui[1] = 0;
1561 eui[2] = 0x5E;
1562 eui[3] = 0xFE;
1563 memcpy(eui + 4, &addr, 4);
1564 return 0;
1565 }
1566
1567 static int addrconf_ifid_sit(u8 *eui, struct net_device *dev)
1568 {
1569 if (dev->priv_flags & IFF_ISATAP)
1570 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
1571 return -1;
1572 }
1573
1574 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev)
1575 {
1576 switch (dev->type) {
1577 case ARPHRD_ETHER:
1578 case ARPHRD_FDDI:
1579 case ARPHRD_IEEE802_TR:
1580 return addrconf_ifid_eui48(eui, dev);
1581 case ARPHRD_ARCNET:
1582 return addrconf_ifid_arcnet(eui, dev);
1583 case ARPHRD_INFINIBAND:
1584 return addrconf_ifid_infiniband(eui, dev);
1585 case ARPHRD_SIT:
1586 return addrconf_ifid_sit(eui, dev);
1587 }
1588 return -1;
1589 }
1590
1591 static int ipv6_inherit_eui64(u8 *eui, struct inet6_dev *idev)
1592 {
1593 int err = -1;
1594 struct inet6_ifaddr *ifp;
1595
1596 read_lock_bh(&idev->lock);
1597 list_for_each_entry(ifp, &idev->addr_list, if_list) {
1598 if (ifp->scope == IFA_LINK && !(ifp->flags&IFA_F_TENTATIVE)) {
1599 memcpy(eui, ifp->addr.s6_addr+8, 8);
1600 err = 0;
1601 break;
1602 }
1603 }
1604 read_unlock_bh(&idev->lock);
1605 return err;
1606 }
1607
1608 #ifdef CONFIG_IPV6_PRIVACY
1609 /* (re)generation of randomized interface identifier (RFC 3041 3.2, 3.5) */
1610 static int __ipv6_regen_rndid(struct inet6_dev *idev)
1611 {
1612 regen:
1613 get_random_bytes(idev->rndid, sizeof(idev->rndid));
1614 idev->rndid[0] &= ~0x02;
1615
1616 /*
1617 * <draft-ietf-ipngwg-temp-addresses-v2-00.txt>:
1618 * check if generated address is not inappropriate
1619 *
1620 * - Reserved subnet anycast (RFC 2526)
1621 * 11111101 11....11 1xxxxxxx
1622 * - ISATAP (RFC4214) 6.1
1623 * 00-00-5E-FE-xx-xx-xx-xx
1624 * - value 0
1625 * - XXX: already assigned to an address on the device
1626 */
1627 if (idev->rndid[0] == 0xfd &&
1628 (idev->rndid[1]&idev->rndid[2]&idev->rndid[3]&idev->rndid[4]&idev->rndid[5]&idev->rndid[6]) == 0xff &&
1629 (idev->rndid[7]&0x80))
1630 goto regen;
1631 if ((idev->rndid[0]|idev->rndid[1]) == 0) {
1632 if (idev->rndid[2] == 0x5e && idev->rndid[3] == 0xfe)
1633 goto regen;
1634 if ((idev->rndid[2]|idev->rndid[3]|idev->rndid[4]|idev->rndid[5]|idev->rndid[6]|idev->rndid[7]) == 0x00)
1635 goto regen;
1636 }
1637
1638 return 0;
1639 }
1640
1641 static void ipv6_regen_rndid(unsigned long data)
1642 {
1643 struct inet6_dev *idev = (struct inet6_dev *) data;
1644 unsigned long expires;
1645
1646 rcu_read_lock_bh();
1647 write_lock_bh(&idev->lock);
1648
1649 if (idev->dead)
1650 goto out;
1651
1652 if (__ipv6_regen_rndid(idev) < 0)
1653 goto out;
1654
1655 expires = jiffies +
1656 idev->cnf.temp_prefered_lft * HZ -
1657 idev->cnf.regen_max_retry * idev->cnf.dad_transmits * idev->nd_parms->retrans_time -
1658 idev->cnf.max_desync_factor * HZ;
1659 if (time_before(expires, jiffies)) {
1660 printk(KERN_WARNING
1661 "ipv6_regen_rndid(): too short regeneration interval; timer disabled for %s.\n",
1662 idev->dev->name);
1663 goto out;
1664 }
1665
1666 if (!mod_timer(&idev->regen_timer, expires))
1667 in6_dev_hold(idev);
1668
1669 out:
1670 write_unlock_bh(&idev->lock);
1671 rcu_read_unlock_bh();
1672 in6_dev_put(idev);
1673 }
1674
1675 static int __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr) {
1676 int ret = 0;
1677
1678 if (tmpaddr && memcmp(idev->rndid, &tmpaddr->s6_addr[8], 8) == 0)
1679 ret = __ipv6_regen_rndid(idev);
1680 return ret;
1681 }
1682 #endif
1683
1684 /*
1685 * Add prefix route.
1686 */
1687
1688 static void
1689 addrconf_prefix_route(struct in6_addr *pfx, int plen, struct net_device *dev,
1690 unsigned long expires, u32 flags)
1691 {
1692 struct fib6_config cfg = {
1693 .fc_table = RT6_TABLE_PREFIX,
1694 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1695 .fc_ifindex = dev->ifindex,
1696 .fc_expires = expires,
1697 .fc_dst_len = plen,
1698 .fc_flags = RTF_UP | flags,
1699 .fc_nlinfo.nl_net = dev_net(dev),
1700 .fc_protocol = RTPROT_KERNEL,
1701 };
1702
1703 ipv6_addr_copy(&cfg.fc_dst, pfx);
1704
1705 /* Prevent useless cloning on PtP SIT.
1706 This thing is done here expecting that the whole
1707 class of non-broadcast devices need not cloning.
1708 */
1709 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
1710 if (dev->type == ARPHRD_SIT && (dev->flags & IFF_POINTOPOINT))
1711 cfg.fc_flags |= RTF_NONEXTHOP;
1712 #endif
1713
1714 ip6_route_add(&cfg);
1715 }
1716
1717 /* Create "default" multicast route to the interface */
1718
1719 static void addrconf_add_mroute(struct net_device *dev)
1720 {
1721 struct fib6_config cfg = {
1722 .fc_table = RT6_TABLE_LOCAL,
1723 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1724 .fc_ifindex = dev->ifindex,
1725 .fc_dst_len = 8,
1726 .fc_flags = RTF_UP,
1727 .fc_nlinfo.nl_net = dev_net(dev),
1728 };
1729
1730 ipv6_addr_set(&cfg.fc_dst, htonl(0xFF000000), 0, 0, 0);
1731
1732 ip6_route_add(&cfg);
1733 }
1734
1735 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
1736 static void sit_route_add(struct net_device *dev)
1737 {
1738 struct fib6_config cfg = {
1739 .fc_table = RT6_TABLE_MAIN,
1740 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1741 .fc_ifindex = dev->ifindex,
1742 .fc_dst_len = 96,
1743 .fc_flags = RTF_UP | RTF_NONEXTHOP,
1744 .fc_nlinfo.nl_net = dev_net(dev),
1745 };
1746
1747 /* prefix length - 96 bits "::d.d.d.d" */
1748 ip6_route_add(&cfg);
1749 }
1750 #endif
1751
1752 static void addrconf_add_lroute(struct net_device *dev)
1753 {
1754 struct in6_addr addr;
1755
1756 ipv6_addr_set(&addr, htonl(0xFE800000), 0, 0, 0);
1757 addrconf_prefix_route(&addr, 64, dev, 0, 0);
1758 }
1759
1760 static struct inet6_dev *addrconf_add_dev(struct net_device *dev)
1761 {
1762 struct inet6_dev *idev;
1763
1764 ASSERT_RTNL();
1765
1766 idev = ipv6_find_idev(dev);
1767 if (!idev)
1768 return ERR_PTR(-ENOBUFS);
1769
1770 if (idev->cnf.disable_ipv6)
1771 return ERR_PTR(-EACCES);
1772
1773 /* Add default multicast route */
1774 addrconf_add_mroute(dev);
1775
1776 /* Add link local route */
1777 addrconf_add_lroute(dev);
1778 return idev;
1779 }
1780
1781 void addrconf_prefix_rcv(struct net_device *dev, u8 *opt, int len)
1782 {
1783 struct prefix_info *pinfo;
1784 __u32 valid_lft;
1785 __u32 prefered_lft;
1786 int addr_type;
1787 struct inet6_dev *in6_dev;
1788 struct net *net = dev_net(dev);
1789
1790 pinfo = (struct prefix_info *) opt;
1791
1792 if (len < sizeof(struct prefix_info)) {
1793 ADBG(("addrconf: prefix option too short\n"));
1794 return;
1795 }
1796
1797 /*
1798 * Validation checks ([ADDRCONF], page 19)
1799 */
1800
1801 addr_type = ipv6_addr_type(&pinfo->prefix);
1802
1803 if (addr_type & (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL))
1804 return;
1805
1806 valid_lft = ntohl(pinfo->valid);
1807 prefered_lft = ntohl(pinfo->prefered);
1808
1809 if (prefered_lft > valid_lft) {
1810 if (net_ratelimit())
1811 printk(KERN_WARNING "addrconf: prefix option has invalid lifetime\n");
1812 return;
1813 }
1814
1815 in6_dev = in6_dev_get(dev);
1816
1817 if (in6_dev == NULL) {
1818 if (net_ratelimit())
1819 printk(KERN_DEBUG "addrconf: device %s not configured\n", dev->name);
1820 return;
1821 }
1822
1823 /*
1824 * Two things going on here:
1825 * 1) Add routes for on-link prefixes
1826 * 2) Configure prefixes with the auto flag set
1827 */
1828
1829 if (pinfo->onlink) {
1830 struct rt6_info *rt;
1831 unsigned long rt_expires;
1832
1833 /* Avoid arithmetic overflow. Really, we could
1834 * save rt_expires in seconds, likely valid_lft,
1835 * but it would require division in fib gc, that it
1836 * not good.
1837 */
1838 if (HZ > USER_HZ)
1839 rt_expires = addrconf_timeout_fixup(valid_lft, HZ);
1840 else
1841 rt_expires = addrconf_timeout_fixup(valid_lft, USER_HZ);
1842
1843 if (addrconf_finite_timeout(rt_expires))
1844 rt_expires *= HZ;
1845
1846 rt = rt6_lookup(net, &pinfo->prefix, NULL,
1847 dev->ifindex, 1);
1848
1849 if (rt && addrconf_is_prefix_route(rt)) {
1850 /* Autoconf prefix route */
1851 if (valid_lft == 0) {
1852 ip6_del_rt(rt);
1853 rt = NULL;
1854 } else if (addrconf_finite_timeout(rt_expires)) {
1855 /* not infinity */
1856 rt->rt6i_expires = jiffies + rt_expires;
1857 rt->rt6i_flags |= RTF_EXPIRES;
1858 } else {
1859 rt->rt6i_flags &= ~RTF_EXPIRES;
1860 rt->rt6i_expires = 0;
1861 }
1862 } else if (valid_lft) {
1863 clock_t expires = 0;
1864 int flags = RTF_ADDRCONF | RTF_PREFIX_RT;
1865 if (addrconf_finite_timeout(rt_expires)) {
1866 /* not infinity */
1867 flags |= RTF_EXPIRES;
1868 expires = jiffies_to_clock_t(rt_expires);
1869 }
1870 addrconf_prefix_route(&pinfo->prefix, pinfo->prefix_len,
1871 dev, expires, flags);
1872 }
1873 if (rt)
1874 dst_release(&rt->dst);
1875 }
1876
1877 /* Try to figure out our local address for this prefix */
1878
1879 if (pinfo->autoconf && in6_dev->cnf.autoconf) {
1880 struct inet6_ifaddr * ifp;
1881 struct in6_addr addr;
1882 int create = 0, update_lft = 0;
1883
1884 if (pinfo->prefix_len == 64) {
1885 memcpy(&addr, &pinfo->prefix, 8);
1886 if (ipv6_generate_eui64(addr.s6_addr + 8, dev) &&
1887 ipv6_inherit_eui64(addr.s6_addr + 8, in6_dev)) {
1888 in6_dev_put(in6_dev);
1889 return;
1890 }
1891 goto ok;
1892 }
1893 if (net_ratelimit())
1894 printk(KERN_DEBUG "IPv6 addrconf: prefix with wrong length %d\n",
1895 pinfo->prefix_len);
1896 in6_dev_put(in6_dev);
1897 return;
1898
1899 ok:
1900
1901 ifp = ipv6_get_ifaddr(net, &addr, dev, 1);
1902
1903 if (ifp == NULL && valid_lft) {
1904 int max_addresses = in6_dev->cnf.max_addresses;
1905 u32 addr_flags = 0;
1906
1907 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1908 if (in6_dev->cnf.optimistic_dad &&
1909 !net->ipv6.devconf_all->forwarding)
1910 addr_flags = IFA_F_OPTIMISTIC;
1911 #endif
1912
1913 /* Do not allow to create too much of autoconfigured
1914 * addresses; this would be too easy way to crash kernel.
1915 */
1916 if (!max_addresses ||
1917 ipv6_count_addresses(in6_dev) < max_addresses)
1918 ifp = ipv6_add_addr(in6_dev, &addr, pinfo->prefix_len,
1919 addr_type&IPV6_ADDR_SCOPE_MASK,
1920 addr_flags);
1921
1922 if (!ifp || IS_ERR(ifp)) {
1923 in6_dev_put(in6_dev);
1924 return;
1925 }
1926
1927 update_lft = create = 1;
1928 ifp->cstamp = jiffies;
1929 addrconf_dad_start(ifp, RTF_ADDRCONF|RTF_PREFIX_RT);
1930 }
1931
1932 if (ifp) {
1933 int flags;
1934 unsigned long now;
1935 #ifdef CONFIG_IPV6_PRIVACY
1936 struct inet6_ifaddr *ift;
1937 #endif
1938 u32 stored_lft;
1939
1940 /* update lifetime (RFC2462 5.5.3 e) */
1941 spin_lock(&ifp->lock);
1942 now = jiffies;
1943 if (ifp->valid_lft > (now - ifp->tstamp) / HZ)
1944 stored_lft = ifp->valid_lft - (now - ifp->tstamp) / HZ;
1945 else
1946 stored_lft = 0;
1947 if (!update_lft && stored_lft) {
1948 if (valid_lft > MIN_VALID_LIFETIME ||
1949 valid_lft > stored_lft)
1950 update_lft = 1;
1951 else if (stored_lft <= MIN_VALID_LIFETIME) {
1952 /* valid_lft <= stored_lft is always true */
1953 /*
1954 * RFC 4862 Section 5.5.3e:
1955 * "Note that the preferred lifetime of
1956 * the corresponding address is always
1957 * reset to the Preferred Lifetime in
1958 * the received Prefix Information
1959 * option, regardless of whether the
1960 * valid lifetime is also reset or
1961 * ignored."
1962 *
1963 * So if the preferred lifetime in
1964 * this advertisement is different
1965 * than what we have stored, but the
1966 * valid lifetime is invalid, just
1967 * reset prefered_lft.
1968 *
1969 * We must set the valid lifetime
1970 * to the stored lifetime since we'll
1971 * be updating the timestamp below,
1972 * else we'll set it back to the
1973 * minimum.
1974 */
1975 if (prefered_lft != ifp->prefered_lft) {
1976 valid_lft = stored_lft;
1977 update_lft = 1;
1978 }
1979 } else {
1980 valid_lft = MIN_VALID_LIFETIME;
1981 if (valid_lft < prefered_lft)
1982 prefered_lft = valid_lft;
1983 update_lft = 1;
1984 }
1985 }
1986
1987 if (update_lft) {
1988 ifp->valid_lft = valid_lft;
1989 ifp->prefered_lft = prefered_lft;
1990 ifp->tstamp = now;
1991 flags = ifp->flags;
1992 ifp->flags &= ~IFA_F_DEPRECATED;
1993 spin_unlock(&ifp->lock);
1994
1995 if (!(flags&IFA_F_TENTATIVE))
1996 ipv6_ifa_notify(0, ifp);
1997 } else
1998 spin_unlock(&ifp->lock);
1999
2000 #ifdef CONFIG_IPV6_PRIVACY
2001 read_lock_bh(&in6_dev->lock);
2002 /* update all temporary addresses in the list */
2003 list_for_each_entry(ift, &in6_dev->tempaddr_list, tmp_list) {
2004 /*
2005 * When adjusting the lifetimes of an existing
2006 * temporary address, only lower the lifetimes.
2007 * Implementations must not increase the
2008 * lifetimes of an existing temporary address
2009 * when processing a Prefix Information Option.
2010 */
2011 if (ifp != ift->ifpub)
2012 continue;
2013
2014 spin_lock(&ift->lock);
2015 flags = ift->flags;
2016 if (ift->valid_lft > valid_lft &&
2017 ift->valid_lft - valid_lft > (jiffies - ift->tstamp) / HZ)
2018 ift->valid_lft = valid_lft + (jiffies - ift->tstamp) / HZ;
2019 if (ift->prefered_lft > prefered_lft &&
2020 ift->prefered_lft - prefered_lft > (jiffies - ift->tstamp) / HZ)
2021 ift->prefered_lft = prefered_lft + (jiffies - ift->tstamp) / HZ;
2022 spin_unlock(&ift->lock);
2023 if (!(flags&IFA_F_TENTATIVE))
2024 ipv6_ifa_notify(0, ift);
2025 }
2026
2027 if ((create || list_empty(&in6_dev->tempaddr_list)) && in6_dev->cnf.use_tempaddr > 0) {
2028 /*
2029 * When a new public address is created as described in [ADDRCONF],
2030 * also create a new temporary address. Also create a temporary
2031 * address if it's enabled but no temporary address currently exists.
2032 */
2033 read_unlock_bh(&in6_dev->lock);
2034 ipv6_create_tempaddr(ifp, NULL);
2035 } else {
2036 read_unlock_bh(&in6_dev->lock);
2037 }
2038 #endif
2039 in6_ifa_put(ifp);
2040 addrconf_verify(0);
2041 }
2042 }
2043 inet6_prefix_notify(RTM_NEWPREFIX, in6_dev, pinfo);
2044 in6_dev_put(in6_dev);
2045 }
2046
2047 /*
2048 * Set destination address.
2049 * Special case for SIT interfaces where we create a new "virtual"
2050 * device.
2051 */
2052 int addrconf_set_dstaddr(struct net *net, void __user *arg)
2053 {
2054 struct in6_ifreq ireq;
2055 struct net_device *dev;
2056 int err = -EINVAL;
2057
2058 rtnl_lock();
2059
2060 err = -EFAULT;
2061 if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2062 goto err_exit;
2063
2064 dev = __dev_get_by_index(net, ireq.ifr6_ifindex);
2065
2066 err = -ENODEV;
2067 if (dev == NULL)
2068 goto err_exit;
2069
2070 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2071 if (dev->type == ARPHRD_SIT) {
2072 const struct net_device_ops *ops = dev->netdev_ops;
2073 struct ifreq ifr;
2074 struct ip_tunnel_parm p;
2075
2076 err = -EADDRNOTAVAIL;
2077 if (!(ipv6_addr_type(&ireq.ifr6_addr) & IPV6_ADDR_COMPATv4))
2078 goto err_exit;
2079
2080 memset(&p, 0, sizeof(p));
2081 p.iph.daddr = ireq.ifr6_addr.s6_addr32[3];
2082 p.iph.saddr = 0;
2083 p.iph.version = 4;
2084 p.iph.ihl = 5;
2085 p.iph.protocol = IPPROTO_IPV6;
2086 p.iph.ttl = 64;
2087 ifr.ifr_ifru.ifru_data = (__force void __user *)&p;
2088
2089 if (ops->ndo_do_ioctl) {
2090 mm_segment_t oldfs = get_fs();
2091
2092 set_fs(KERNEL_DS);
2093 err = ops->ndo_do_ioctl(dev, &ifr, SIOCADDTUNNEL);
2094 set_fs(oldfs);
2095 } else
2096 err = -EOPNOTSUPP;
2097
2098 if (err == 0) {
2099 err = -ENOBUFS;
2100 dev = __dev_get_by_name(net, p.name);
2101 if (!dev)
2102 goto err_exit;
2103 err = dev_open(dev);
2104 }
2105 }
2106 #endif
2107
2108 err_exit:
2109 rtnl_unlock();
2110 return err;
2111 }
2112
2113 /*
2114 * Manual configuration of address on an interface
2115 */
2116 static int inet6_addr_add(struct net *net, int ifindex, struct in6_addr *pfx,
2117 unsigned int plen, __u8 ifa_flags, __u32 prefered_lft,
2118 __u32 valid_lft)
2119 {
2120 struct inet6_ifaddr *ifp;
2121 struct inet6_dev *idev;
2122 struct net_device *dev;
2123 int scope;
2124 u32 flags;
2125 clock_t expires;
2126 unsigned long timeout;
2127
2128 ASSERT_RTNL();
2129
2130 if (plen > 128)
2131 return -EINVAL;
2132
2133 /* check the lifetime */
2134 if (!valid_lft || prefered_lft > valid_lft)
2135 return -EINVAL;
2136
2137 dev = __dev_get_by_index(net, ifindex);
2138 if (!dev)
2139 return -ENODEV;
2140
2141 idev = addrconf_add_dev(dev);
2142 if (IS_ERR(idev))
2143 return PTR_ERR(idev);
2144
2145 scope = ipv6_addr_scope(pfx);
2146
2147 timeout = addrconf_timeout_fixup(valid_lft, HZ);
2148 if (addrconf_finite_timeout(timeout)) {
2149 expires = jiffies_to_clock_t(timeout * HZ);
2150 valid_lft = timeout;
2151 flags = RTF_EXPIRES;
2152 } else {
2153 expires = 0;
2154 flags = 0;
2155 ifa_flags |= IFA_F_PERMANENT;
2156 }
2157
2158 timeout = addrconf_timeout_fixup(prefered_lft, HZ);
2159 if (addrconf_finite_timeout(timeout)) {
2160 if (timeout == 0)
2161 ifa_flags |= IFA_F_DEPRECATED;
2162 prefered_lft = timeout;
2163 }
2164
2165 ifp = ipv6_add_addr(idev, pfx, plen, scope, ifa_flags);
2166
2167 if (!IS_ERR(ifp)) {
2168 spin_lock_bh(&ifp->lock);
2169 ifp->valid_lft = valid_lft;
2170 ifp->prefered_lft = prefered_lft;
2171 ifp->tstamp = jiffies;
2172 spin_unlock_bh(&ifp->lock);
2173
2174 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, dev,
2175 expires, flags);
2176 /*
2177 * Note that section 3.1 of RFC 4429 indicates
2178 * that the Optimistic flag should not be set for
2179 * manually configured addresses
2180 */
2181 addrconf_dad_start(ifp, 0);
2182 in6_ifa_put(ifp);
2183 addrconf_verify(0);
2184 return 0;
2185 }
2186
2187 return PTR_ERR(ifp);
2188 }
2189
2190 static int inet6_addr_del(struct net *net, int ifindex, struct in6_addr *pfx,
2191 unsigned int plen)
2192 {
2193 struct inet6_ifaddr *ifp;
2194 struct inet6_dev *idev;
2195 struct net_device *dev;
2196
2197 if (plen > 128)
2198 return -EINVAL;
2199
2200 dev = __dev_get_by_index(net, ifindex);
2201 if (!dev)
2202 return -ENODEV;
2203
2204 if ((idev = __in6_dev_get(dev)) == NULL)
2205 return -ENXIO;
2206
2207 read_lock_bh(&idev->lock);
2208 list_for_each_entry(ifp, &idev->addr_list, if_list) {
2209 if (ifp->prefix_len == plen &&
2210 ipv6_addr_equal(pfx, &ifp->addr)) {
2211 in6_ifa_hold(ifp);
2212 read_unlock_bh(&idev->lock);
2213
2214 ipv6_del_addr(ifp);
2215
2216 /* If the last address is deleted administratively,
2217 disable IPv6 on this interface.
2218 */
2219 if (list_empty(&idev->addr_list))
2220 addrconf_ifdown(idev->dev, 1);
2221 return 0;
2222 }
2223 }
2224 read_unlock_bh(&idev->lock);
2225 return -EADDRNOTAVAIL;
2226 }
2227
2228
2229 int addrconf_add_ifaddr(struct net *net, void __user *arg)
2230 {
2231 struct in6_ifreq ireq;
2232 int err;
2233
2234 if (!capable(CAP_NET_ADMIN))
2235 return -EPERM;
2236
2237 if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2238 return -EFAULT;
2239
2240 rtnl_lock();
2241 err = inet6_addr_add(net, ireq.ifr6_ifindex, &ireq.ifr6_addr,
2242 ireq.ifr6_prefixlen, IFA_F_PERMANENT,
2243 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
2244 rtnl_unlock();
2245 return err;
2246 }
2247
2248 int addrconf_del_ifaddr(struct net *net, void __user *arg)
2249 {
2250 struct in6_ifreq ireq;
2251 int err;
2252
2253 if (!capable(CAP_NET_ADMIN))
2254 return -EPERM;
2255
2256 if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2257 return -EFAULT;
2258
2259 rtnl_lock();
2260 err = inet6_addr_del(net, ireq.ifr6_ifindex, &ireq.ifr6_addr,
2261 ireq.ifr6_prefixlen);
2262 rtnl_unlock();
2263 return err;
2264 }
2265
2266 static void add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
2267 int plen, int scope)
2268 {
2269 struct inet6_ifaddr *ifp;
2270
2271 ifp = ipv6_add_addr(idev, addr, plen, scope, IFA_F_PERMANENT);
2272 if (!IS_ERR(ifp)) {
2273 spin_lock_bh(&ifp->lock);
2274 ifp->flags &= ~IFA_F_TENTATIVE;
2275 spin_unlock_bh(&ifp->lock);
2276 ipv6_ifa_notify(RTM_NEWADDR, ifp);
2277 in6_ifa_put(ifp);
2278 }
2279 }
2280
2281 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2282 static void sit_add_v4_addrs(struct inet6_dev *idev)
2283 {
2284 struct in6_addr addr;
2285 struct net_device *dev;
2286 struct net *net = dev_net(idev->dev);
2287 int scope;
2288
2289 ASSERT_RTNL();
2290
2291 memset(&addr, 0, sizeof(struct in6_addr));
2292 memcpy(&addr.s6_addr32[3], idev->dev->dev_addr, 4);
2293
2294 if (idev->dev->flags&IFF_POINTOPOINT) {
2295 addr.s6_addr32[0] = htonl(0xfe800000);
2296 scope = IFA_LINK;
2297 } else {
2298 scope = IPV6_ADDR_COMPATv4;
2299 }
2300
2301 if (addr.s6_addr32[3]) {
2302 add_addr(idev, &addr, 128, scope);
2303 return;
2304 }
2305
2306 for_each_netdev(net, dev) {
2307 struct in_device * in_dev = __in_dev_get_rtnl(dev);
2308 if (in_dev && (dev->flags & IFF_UP)) {
2309 struct in_ifaddr * ifa;
2310
2311 int flag = scope;
2312
2313 for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) {
2314 int plen;
2315
2316 addr.s6_addr32[3] = ifa->ifa_local;
2317
2318 if (ifa->ifa_scope == RT_SCOPE_LINK)
2319 continue;
2320 if (ifa->ifa_scope >= RT_SCOPE_HOST) {
2321 if (idev->dev->flags&IFF_POINTOPOINT)
2322 continue;
2323 flag |= IFA_HOST;
2324 }
2325 if (idev->dev->flags&IFF_POINTOPOINT)
2326 plen = 64;
2327 else
2328 plen = 96;
2329
2330 add_addr(idev, &addr, plen, flag);
2331 }
2332 }
2333 }
2334 }
2335 #endif
2336
2337 static void init_loopback(struct net_device *dev)
2338 {
2339 struct inet6_dev *idev;
2340
2341 /* ::1 */
2342
2343 ASSERT_RTNL();
2344
2345 if ((idev = ipv6_find_idev(dev)) == NULL) {
2346 printk(KERN_DEBUG "init loopback: add_dev failed\n");
2347 return;
2348 }
2349
2350 add_addr(idev, &in6addr_loopback, 128, IFA_HOST);
2351 }
2352
2353 static void addrconf_add_linklocal(struct inet6_dev *idev, struct in6_addr *addr)
2354 {
2355 struct inet6_ifaddr * ifp;
2356 u32 addr_flags = IFA_F_PERMANENT;
2357
2358 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2359 if (idev->cnf.optimistic_dad &&
2360 !dev_net(idev->dev)->ipv6.devconf_all->forwarding)
2361 addr_flags |= IFA_F_OPTIMISTIC;
2362 #endif
2363
2364
2365 ifp = ipv6_add_addr(idev, addr, 64, IFA_LINK, addr_flags);
2366 if (!IS_ERR(ifp)) {
2367 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, idev->dev, 0, 0);
2368 addrconf_dad_start(ifp, 0);
2369 in6_ifa_put(ifp);
2370 }
2371 }
2372
2373 static void addrconf_dev_config(struct net_device *dev)
2374 {
2375 struct in6_addr addr;
2376 struct inet6_dev * idev;
2377
2378 ASSERT_RTNL();
2379
2380 if ((dev->type != ARPHRD_ETHER) &&
2381 (dev->type != ARPHRD_FDDI) &&
2382 (dev->type != ARPHRD_IEEE802_TR) &&
2383 (dev->type != ARPHRD_ARCNET) &&
2384 (dev->type != ARPHRD_INFINIBAND)) {
2385 /* Alas, we support only Ethernet autoconfiguration. */
2386 return;
2387 }
2388
2389 idev = addrconf_add_dev(dev);
2390 if (IS_ERR(idev))
2391 return;
2392
2393 memset(&addr, 0, sizeof(struct in6_addr));
2394 addr.s6_addr32[0] = htonl(0xFE800000);
2395
2396 if (ipv6_generate_eui64(addr.s6_addr + 8, dev) == 0)
2397 addrconf_add_linklocal(idev, &addr);
2398 }
2399
2400 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2401 static void addrconf_sit_config(struct net_device *dev)
2402 {
2403 struct inet6_dev *idev;
2404
2405 ASSERT_RTNL();
2406
2407 /*
2408 * Configure the tunnel with one of our IPv4
2409 * addresses... we should configure all of
2410 * our v4 addrs in the tunnel
2411 */
2412
2413 if ((idev = ipv6_find_idev(dev)) == NULL) {
2414 printk(KERN_DEBUG "init sit: add_dev failed\n");
2415 return;
2416 }
2417
2418 if (dev->priv_flags & IFF_ISATAP) {
2419 struct in6_addr addr;
2420
2421 ipv6_addr_set(&addr, htonl(0xFE800000), 0, 0, 0);
2422 addrconf_prefix_route(&addr, 64, dev, 0, 0);
2423 if (!ipv6_generate_eui64(addr.s6_addr + 8, dev))
2424 addrconf_add_linklocal(idev, &addr);
2425 return;
2426 }
2427
2428 sit_add_v4_addrs(idev);
2429
2430 if (dev->flags&IFF_POINTOPOINT) {
2431 addrconf_add_mroute(dev);
2432 addrconf_add_lroute(dev);
2433 } else
2434 sit_route_add(dev);
2435 }
2436 #endif
2437
2438 static inline int
2439 ipv6_inherit_linklocal(struct inet6_dev *idev, struct net_device *link_dev)
2440 {
2441 struct in6_addr lladdr;
2442
2443 if (!ipv6_get_lladdr(link_dev, &lladdr, IFA_F_TENTATIVE)) {
2444 addrconf_add_linklocal(idev, &lladdr);
2445 return 0;
2446 }
2447 return -1;
2448 }
2449
2450 static void ip6_tnl_add_linklocal(struct inet6_dev *idev)
2451 {
2452 struct net_device *link_dev;
2453 struct net *net = dev_net(idev->dev);
2454
2455 /* first try to inherit the link-local address from the link device */
2456 if (idev->dev->iflink &&
2457 (link_dev = __dev_get_by_index(net, idev->dev->iflink))) {
2458 if (!ipv6_inherit_linklocal(idev, link_dev))
2459 return;
2460 }
2461 /* then try to inherit it from any device */
2462 for_each_netdev(net, link_dev) {
2463 if (!ipv6_inherit_linklocal(idev, link_dev))
2464 return;
2465 }
2466 printk(KERN_DEBUG "init ip6-ip6: add_linklocal failed\n");
2467 }
2468
2469 /*
2470 * Autoconfigure tunnel with a link-local address so routing protocols,
2471 * DHCPv6, MLD etc. can be run over the virtual link
2472 */
2473
2474 static void addrconf_ip6_tnl_config(struct net_device *dev)
2475 {
2476 struct inet6_dev *idev;
2477
2478 ASSERT_RTNL();
2479
2480 idev = addrconf_add_dev(dev);
2481 if (IS_ERR(idev)) {
2482 printk(KERN_DEBUG "init ip6-ip6: add_dev failed\n");
2483 return;
2484 }
2485 ip6_tnl_add_linklocal(idev);
2486 }
2487
2488 static int addrconf_notify(struct notifier_block *this, unsigned long event,
2489 void * data)
2490 {
2491 struct net_device *dev = (struct net_device *) data;
2492 struct inet6_dev *idev = __in6_dev_get(dev);
2493 int run_pending = 0;
2494 int err;
2495
2496 switch (event) {
2497 case NETDEV_REGISTER:
2498 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2499 idev = ipv6_add_dev(dev);
2500 if (!idev)
2501 return notifier_from_errno(-ENOMEM);
2502 }
2503 break;
2504
2505 case NETDEV_UP:
2506 case NETDEV_CHANGE:
2507 if (dev->flags & IFF_SLAVE)
2508 break;
2509
2510 if (event == NETDEV_UP) {
2511 if (!addrconf_qdisc_ok(dev)) {
2512 /* device is not ready yet. */
2513 printk(KERN_INFO
2514 "ADDRCONF(NETDEV_UP): %s: "
2515 "link is not ready\n",
2516 dev->name);
2517 break;
2518 }
2519
2520 if (!idev && dev->mtu >= IPV6_MIN_MTU)
2521 idev = ipv6_add_dev(dev);
2522
2523 if (idev) {
2524 idev->if_flags |= IF_READY;
2525 run_pending = 1;
2526 }
2527 } else {
2528 if (!addrconf_qdisc_ok(dev)) {
2529 /* device is still not ready. */
2530 break;
2531 }
2532
2533 if (idev) {
2534 if (idev->if_flags & IF_READY)
2535 /* device is already configured. */
2536 break;
2537 idev->if_flags |= IF_READY;
2538 }
2539
2540 printk(KERN_INFO
2541 "ADDRCONF(NETDEV_CHANGE): %s: "
2542 "link becomes ready\n",
2543 dev->name);
2544
2545 run_pending = 1;
2546 }
2547
2548 switch (dev->type) {
2549 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2550 case ARPHRD_SIT:
2551 addrconf_sit_config(dev);
2552 break;
2553 #endif
2554 case ARPHRD_TUNNEL6:
2555 addrconf_ip6_tnl_config(dev);
2556 break;
2557 case ARPHRD_LOOPBACK:
2558 init_loopback(dev);
2559 break;
2560
2561 default:
2562 addrconf_dev_config(dev);
2563 break;
2564 }
2565
2566 if (idev) {
2567 if (run_pending)
2568 addrconf_dad_run(idev);
2569
2570 /*
2571 * If the MTU changed during the interface down,
2572 * when the interface up, the changed MTU must be
2573 * reflected in the idev as well as routers.
2574 */
2575 if (idev->cnf.mtu6 != dev->mtu &&
2576 dev->mtu >= IPV6_MIN_MTU) {
2577 rt6_mtu_change(dev, dev->mtu);
2578 idev->cnf.mtu6 = dev->mtu;
2579 }
2580 idev->tstamp = jiffies;
2581 inet6_ifinfo_notify(RTM_NEWLINK, idev);
2582
2583 /*
2584 * If the changed mtu during down is lower than
2585 * IPV6_MIN_MTU stop IPv6 on this interface.
2586 */
2587 if (dev->mtu < IPV6_MIN_MTU)
2588 addrconf_ifdown(dev, 1);
2589 }
2590 break;
2591
2592 case NETDEV_CHANGEMTU:
2593 if (idev && dev->mtu >= IPV6_MIN_MTU) {
2594 rt6_mtu_change(dev, dev->mtu);
2595 idev->cnf.mtu6 = dev->mtu;
2596 break;
2597 }
2598
2599 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2600 idev = ipv6_add_dev(dev);
2601 if (idev)
2602 break;
2603 }
2604
2605 /*
2606 * MTU falled under IPV6_MIN_MTU.
2607 * Stop IPv6 on this interface.
2608 */
2609
2610 case NETDEV_DOWN:
2611 case NETDEV_UNREGISTER:
2612 /*
2613 * Remove all addresses from this interface.
2614 */
2615 addrconf_ifdown(dev, event != NETDEV_DOWN);
2616 break;
2617
2618 case NETDEV_CHANGENAME:
2619 if (idev) {
2620 snmp6_unregister_dev(idev);
2621 addrconf_sysctl_unregister(idev);
2622 addrconf_sysctl_register(idev);
2623 err = snmp6_register_dev(idev);
2624 if (err)
2625 return notifier_from_errno(err);
2626 }
2627 break;
2628
2629 case NETDEV_PRE_TYPE_CHANGE:
2630 case NETDEV_POST_TYPE_CHANGE:
2631 addrconf_type_change(dev, event);
2632 break;
2633 }
2634
2635 return NOTIFY_OK;
2636 }
2637
2638 /*
2639 * addrconf module should be notified of a device going up
2640 */
2641 static struct notifier_block ipv6_dev_notf = {
2642 .notifier_call = addrconf_notify,
2643 };
2644
2645 static void addrconf_type_change(struct net_device *dev, unsigned long event)
2646 {
2647 struct inet6_dev *idev;
2648 ASSERT_RTNL();
2649
2650 idev = __in6_dev_get(dev);
2651
2652 if (event == NETDEV_POST_TYPE_CHANGE)
2653 ipv6_mc_remap(idev);
2654 else if (event == NETDEV_PRE_TYPE_CHANGE)
2655 ipv6_mc_unmap(idev);
2656 }
2657
2658 static int addrconf_ifdown(struct net_device *dev, int how)
2659 {
2660 struct net *net = dev_net(dev);
2661 struct inet6_dev *idev;
2662 struct inet6_ifaddr *ifa;
2663 int state, i;
2664
2665 ASSERT_RTNL();
2666
2667 rt6_ifdown(net, dev);
2668 neigh_ifdown(&nd_tbl, dev);
2669
2670 idev = __in6_dev_get(dev);
2671 if (idev == NULL)
2672 return -ENODEV;
2673
2674 /*
2675 * Step 1: remove reference to ipv6 device from parent device.
2676 * Do not dev_put!
2677 */
2678 if (how) {
2679 idev->dead = 1;
2680
2681 /* protected by rtnl_lock */
2682 rcu_assign_pointer(dev->ip6_ptr, NULL);
2683
2684 /* Step 1.5: remove snmp6 entry */
2685 snmp6_unregister_dev(idev);
2686
2687 }
2688
2689 /* Step 2: clear hash table */
2690 for (i = 0; i < IN6_ADDR_HSIZE; i++) {
2691 struct hlist_head *h = &inet6_addr_lst[i];
2692 struct hlist_node *n;
2693
2694 spin_lock_bh(&addrconf_hash_lock);
2695 restart:
2696 hlist_for_each_entry_rcu(ifa, n, h, addr_lst) {
2697 if (ifa->idev == idev) {
2698 hlist_del_init_rcu(&ifa->addr_lst);
2699 addrconf_del_timer(ifa);
2700 goto restart;
2701 }
2702 }
2703 spin_unlock_bh(&addrconf_hash_lock);
2704 }
2705
2706 write_lock_bh(&idev->lock);
2707
2708 /* Step 2: clear flags for stateless addrconf */
2709 if (!how)
2710 idev->if_flags &= ~(IF_RS_SENT|IF_RA_RCVD|IF_READY);
2711
2712 #ifdef CONFIG_IPV6_PRIVACY
2713 if (how && del_timer(&idev->regen_timer))
2714 in6_dev_put(idev);
2715
2716 /* Step 3: clear tempaddr list */
2717 while (!list_empty(&idev->tempaddr_list)) {
2718 ifa = list_first_entry(&idev->tempaddr_list,
2719 struct inet6_ifaddr, tmp_list);
2720 list_del(&ifa->tmp_list);
2721 write_unlock_bh(&idev->lock);
2722 spin_lock_bh(&ifa->lock);
2723
2724 if (ifa->ifpub) {
2725 in6_ifa_put(ifa->ifpub);
2726 ifa->ifpub = NULL;
2727 }
2728 spin_unlock_bh(&ifa->lock);
2729 in6_ifa_put(ifa);
2730 write_lock_bh(&idev->lock);
2731 }
2732 #endif
2733
2734 while (!list_empty(&idev->addr_list)) {
2735 ifa = list_first_entry(&idev->addr_list,
2736 struct inet6_ifaddr, if_list);
2737 addrconf_del_timer(ifa);
2738
2739 list_del(&ifa->if_list);
2740
2741 write_unlock_bh(&idev->lock);
2742
2743 spin_lock_bh(&ifa->state_lock);
2744 state = ifa->state;
2745 ifa->state = INET6_IFADDR_STATE_DEAD;
2746 spin_unlock_bh(&ifa->state_lock);
2747
2748 if (state != INET6_IFADDR_STATE_DEAD) {
2749 __ipv6_ifa_notify(RTM_DELADDR, ifa);
2750 atomic_notifier_call_chain(&inet6addr_chain, NETDEV_DOWN, ifa);
2751 }
2752 in6_ifa_put(ifa);
2753
2754 write_lock_bh(&idev->lock);
2755 }
2756
2757 write_unlock_bh(&idev->lock);
2758
2759 /* Step 5: Discard multicast list */
2760 if (how)
2761 ipv6_mc_destroy_dev(idev);
2762 else
2763 ipv6_mc_down(idev);
2764
2765 idev->tstamp = jiffies;
2766
2767 /* Last: Shot the device (if unregistered) */
2768 if (how) {
2769 addrconf_sysctl_unregister(idev);
2770 neigh_parms_release(&nd_tbl, idev->nd_parms);
2771 neigh_ifdown(&nd_tbl, dev);
2772 in6_dev_put(idev);
2773 }
2774 return 0;
2775 }
2776
2777 static void addrconf_rs_timer(unsigned long data)
2778 {
2779 struct inet6_ifaddr *ifp = (struct inet6_ifaddr *) data;
2780 struct inet6_dev *idev = ifp->idev;
2781
2782 read_lock(&idev->lock);
2783 if (idev->dead || !(idev->if_flags & IF_READY))
2784 goto out;
2785
2786 if (idev->cnf.forwarding)
2787 goto out;
2788
2789 /* Announcement received after solicitation was sent */
2790 if (idev->if_flags & IF_RA_RCVD)
2791 goto out;
2792
2793 spin_lock(&ifp->lock);
2794 if (ifp->probes++ < idev->cnf.rtr_solicits) {
2795 /* The wait after the last probe can be shorter */
2796 addrconf_mod_timer(ifp, AC_RS,
2797 (ifp->probes == idev->cnf.rtr_solicits) ?
2798 idev->cnf.rtr_solicit_delay :
2799 idev->cnf.rtr_solicit_interval);
2800 spin_unlock(&ifp->lock);
2801
2802 ndisc_send_rs(idev->dev, &ifp->addr, &in6addr_linklocal_allrouters);
2803 } else {
2804 spin_unlock(&ifp->lock);
2805 /*
2806 * Note: we do not support deprecated "all on-link"
2807 * assumption any longer.
2808 */
2809 printk(KERN_DEBUG "%s: no IPv6 routers present\n",
2810 idev->dev->name);
2811 }
2812
2813 out:
2814 read_unlock(&idev->lock);
2815 in6_ifa_put(ifp);
2816 }
2817
2818 /*
2819 * Duplicate Address Detection
2820 */
2821 static void addrconf_dad_kick(struct inet6_ifaddr *ifp)
2822 {
2823 unsigned long rand_num;
2824 struct inet6_dev *idev = ifp->idev;
2825
2826 if (ifp->flags & IFA_F_OPTIMISTIC)
2827 rand_num = 0;
2828 else
2829 rand_num = net_random() % (idev->cnf.rtr_solicit_delay ? : 1);
2830
2831 ifp->probes = idev->cnf.dad_transmits;
2832 addrconf_mod_timer(ifp, AC_DAD, rand_num);
2833 }
2834
2835 static void addrconf_dad_start(struct inet6_ifaddr *ifp, u32 flags)
2836 {
2837 struct inet6_dev *idev = ifp->idev;
2838 struct net_device *dev = idev->dev;
2839
2840 addrconf_join_solict(dev, &ifp->addr);
2841
2842 net_srandom(ifp->addr.s6_addr32[3]);
2843
2844 read_lock_bh(&idev->lock);
2845 spin_lock(&ifp->lock);
2846 if (ifp->state == INET6_IFADDR_STATE_DEAD)
2847 goto out;
2848
2849 if (dev->flags&(IFF_NOARP|IFF_LOOPBACK) ||
2850 idev->cnf.accept_dad < 1 ||
2851 !(ifp->flags&IFA_F_TENTATIVE) ||
2852 ifp->flags & IFA_F_NODAD) {
2853 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
2854 spin_unlock(&ifp->lock);
2855 read_unlock_bh(&idev->lock);
2856
2857 addrconf_dad_completed(ifp);
2858 return;
2859 }
2860
2861 if (!(idev->if_flags & IF_READY)) {
2862 spin_unlock(&ifp->lock);
2863 read_unlock_bh(&idev->lock);
2864 /*
2865 * If the device is not ready:
2866 * - keep it tentative if it is a permanent address.
2867 * - otherwise, kill it.
2868 */
2869 in6_ifa_hold(ifp);
2870 addrconf_dad_stop(ifp, 0);
2871 return;
2872 }
2873
2874 /*
2875 * Optimistic nodes can start receiving
2876 * Frames right away
2877 */
2878 if (ifp->flags & IFA_F_OPTIMISTIC)
2879 ip6_ins_rt(ifp->rt);
2880
2881 addrconf_dad_kick(ifp);
2882 out:
2883 spin_unlock(&ifp->lock);
2884 read_unlock_bh(&idev->lock);
2885 }
2886
2887 static void addrconf_dad_timer(unsigned long data)
2888 {
2889 struct inet6_ifaddr *ifp = (struct inet6_ifaddr *) data;
2890 struct inet6_dev *idev = ifp->idev;
2891 struct in6_addr mcaddr;
2892
2893 if (!ifp->probes && addrconf_dad_end(ifp))
2894 goto out;
2895
2896 read_lock(&idev->lock);
2897 if (idev->dead || !(idev->if_flags & IF_READY)) {
2898 read_unlock(&idev->lock);
2899 goto out;
2900 }
2901
2902 spin_lock(&ifp->lock);
2903 if (ifp->state == INET6_IFADDR_STATE_DEAD) {
2904 spin_unlock(&ifp->lock);
2905 read_unlock(&idev->lock);
2906 goto out;
2907 }
2908
2909 if (ifp->probes == 0) {
2910 /*
2911 * DAD was successful
2912 */
2913
2914 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
2915 spin_unlock(&ifp->lock);
2916 read_unlock(&idev->lock);
2917
2918 addrconf_dad_completed(ifp);
2919
2920 goto out;
2921 }
2922
2923 ifp->probes--;
2924 addrconf_mod_timer(ifp, AC_DAD, ifp->idev->nd_parms->retrans_time);
2925 spin_unlock(&ifp->lock);
2926 read_unlock(&idev->lock);
2927
2928 /* send a neighbour solicitation for our addr */
2929 addrconf_addr_solict_mult(&ifp->addr, &mcaddr);
2930 ndisc_send_ns(ifp->idev->dev, NULL, &ifp->addr, &mcaddr, &in6addr_any);
2931 out:
2932 in6_ifa_put(ifp);
2933 }
2934
2935 static void addrconf_dad_completed(struct inet6_ifaddr *ifp)
2936 {
2937 struct net_device *dev = ifp->idev->dev;
2938
2939 /*
2940 * Configure the address for reception. Now it is valid.
2941 */
2942
2943 ipv6_ifa_notify(RTM_NEWADDR, ifp);
2944
2945 /* If added prefix is link local and forwarding is off,
2946 start sending router solicitations.
2947 */
2948
2949 if ((ifp->idev->cnf.forwarding == 0 ||
2950 ifp->idev->cnf.forwarding == 2) &&
2951 ifp->idev->cnf.rtr_solicits > 0 &&
2952 (dev->flags&IFF_LOOPBACK) == 0 &&
2953 (ipv6_addr_type(&ifp->addr) & IPV6_ADDR_LINKLOCAL)) {
2954 /*
2955 * If a host as already performed a random delay
2956 * [...] as part of DAD [...] there is no need
2957 * to delay again before sending the first RS
2958 */
2959 ndisc_send_rs(ifp->idev->dev, &ifp->addr, &in6addr_linklocal_allrouters);
2960
2961 spin_lock_bh(&ifp->lock);
2962 ifp->probes = 1;
2963 ifp->idev->if_flags |= IF_RS_SENT;
2964 addrconf_mod_timer(ifp, AC_RS, ifp->idev->cnf.rtr_solicit_interval);
2965 spin_unlock_bh(&ifp->lock);
2966 }
2967 }
2968
2969 static void addrconf_dad_run(struct inet6_dev *idev)
2970 {
2971 struct inet6_ifaddr *ifp;
2972
2973 read_lock_bh(&idev->lock);
2974 list_for_each_entry(ifp, &idev->addr_list, if_list) {
2975 spin_lock(&ifp->lock);
2976 if (ifp->flags & IFA_F_TENTATIVE &&
2977 ifp->state == INET6_IFADDR_STATE_DAD)
2978 addrconf_dad_kick(ifp);
2979 spin_unlock(&ifp->lock);
2980 }
2981 read_unlock_bh(&idev->lock);
2982 }
2983
2984 #ifdef CONFIG_PROC_FS
2985 struct if6_iter_state {
2986 struct seq_net_private p;
2987 int bucket;
2988 };
2989
2990 static struct inet6_ifaddr *if6_get_first(struct seq_file *seq)
2991 {
2992 struct inet6_ifaddr *ifa = NULL;
2993 struct if6_iter_state *state = seq->private;
2994 struct net *net = seq_file_net(seq);
2995
2996 for (state->bucket = 0; state->bucket < IN6_ADDR_HSIZE; ++state->bucket) {
2997 struct hlist_node *n;
2998 hlist_for_each_entry_rcu_bh(ifa, n, &inet6_addr_lst[state->bucket],
2999 addr_lst)
3000 if (net_eq(dev_net(ifa->idev->dev), net))
3001 return ifa;
3002 }
3003 return NULL;
3004 }
3005
3006 static struct inet6_ifaddr *if6_get_next(struct seq_file *seq,
3007 struct inet6_ifaddr *ifa)
3008 {
3009 struct if6_iter_state *state = seq->private;
3010 struct net *net = seq_file_net(seq);
3011 struct hlist_node *n = &ifa->addr_lst;
3012
3013 hlist_for_each_entry_continue_rcu_bh(ifa, n, addr_lst)
3014 if (net_eq(dev_net(ifa->idev->dev), net))
3015 return ifa;
3016
3017 while (++state->bucket < IN6_ADDR_HSIZE) {
3018 hlist_for_each_entry_rcu_bh(ifa, n,
3019 &inet6_addr_lst[state->bucket], addr_lst) {
3020 if (net_eq(dev_net(ifa->idev->dev), net))
3021 return ifa;
3022 }
3023 }
3024
3025 return NULL;
3026 }
3027
3028 static struct inet6_ifaddr *if6_get_idx(struct seq_file *seq, loff_t pos)
3029 {
3030 struct inet6_ifaddr *ifa = if6_get_first(seq);
3031
3032 if (ifa)
3033 while (pos && (ifa = if6_get_next(seq, ifa)) != NULL)
3034 --pos;
3035 return pos ? NULL : ifa;
3036 }
3037
3038 static void *if6_seq_start(struct seq_file *seq, loff_t *pos)
3039 __acquires(rcu_bh)
3040 {
3041 rcu_read_lock_bh();
3042 return if6_get_idx(seq, *pos);
3043 }
3044
3045 static void *if6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
3046 {
3047 struct inet6_ifaddr *ifa;
3048
3049 ifa = if6_get_next(seq, v);
3050 ++*pos;
3051 return ifa;
3052 }
3053
3054 static void if6_seq_stop(struct seq_file *seq, void *v)
3055 __releases(rcu_bh)
3056 {
3057 rcu_read_unlock_bh();
3058 }
3059
3060 static int if6_seq_show(struct seq_file *seq, void *v)
3061 {
3062 struct inet6_ifaddr *ifp = (struct inet6_ifaddr *)v;
3063 seq_printf(seq, "%pi6 %02x %02x %02x %02x %8s\n",
3064 &ifp->addr,
3065 ifp->idev->dev->ifindex,
3066 ifp->prefix_len,
3067 ifp->scope,
3068 ifp->flags,
3069 ifp->idev->dev->name);
3070 return 0;
3071 }
3072
3073 static const struct seq_operations if6_seq_ops = {
3074 .start = if6_seq_start,
3075 .next = if6_seq_next,
3076 .show = if6_seq_show,
3077 .stop = if6_seq_stop,
3078 };
3079
3080 static int if6_seq_open(struct inode *inode, struct file *file)
3081 {
3082 return seq_open_net(inode, file, &if6_seq_ops,
3083 sizeof(struct if6_iter_state));
3084 }
3085
3086 static const struct file_operations if6_fops = {
3087 .owner = THIS_MODULE,
3088 .open = if6_seq_open,
3089 .read = seq_read,
3090 .llseek = seq_lseek,
3091 .release = seq_release_net,
3092 };
3093
3094 static int __net_init if6_proc_net_init(struct net *net)
3095 {
3096 if (!proc_net_fops_create(net, "if_inet6", S_IRUGO, &if6_fops))
3097 return -ENOMEM;
3098 return 0;
3099 }
3100
3101 static void __net_exit if6_proc_net_exit(struct net *net)
3102 {
3103 proc_net_remove(net, "if_inet6");
3104 }
3105
3106 static struct pernet_operations if6_proc_net_ops = {
3107 .init = if6_proc_net_init,
3108 .exit = if6_proc_net_exit,
3109 };
3110
3111 int __init if6_proc_init(void)
3112 {
3113 return register_pernet_subsys(&if6_proc_net_ops);
3114 }
3115
3116 void if6_proc_exit(void)
3117 {
3118 unregister_pernet_subsys(&if6_proc_net_ops);
3119 }
3120 #endif /* CONFIG_PROC_FS */
3121
3122 #if defined(CONFIG_IPV6_MIP6) || defined(CONFIG_IPV6_MIP6_MODULE)
3123 /* Check if address is a home address configured on any interface. */
3124 int ipv6_chk_home_addr(struct net *net, struct in6_addr *addr)
3125 {
3126 int ret = 0;
3127 struct inet6_ifaddr *ifp = NULL;
3128 struct hlist_node *n;
3129 unsigned int hash = ipv6_addr_hash(addr);
3130
3131 rcu_read_lock_bh();
3132 hlist_for_each_entry_rcu_bh(ifp, n, &inet6_addr_lst[hash], addr_lst) {
3133 if (!net_eq(dev_net(ifp->idev->dev), net))
3134 continue;
3135 if (ipv6_addr_equal(&ifp->addr, addr) &&
3136 (ifp->flags & IFA_F_HOMEADDRESS)) {
3137 ret = 1;
3138 break;
3139 }
3140 }
3141 rcu_read_unlock_bh();
3142 return ret;
3143 }
3144 #endif
3145
3146 /*
3147 * Periodic address status verification
3148 */
3149
3150 static void addrconf_verify(unsigned long foo)
3151 {
3152 unsigned long now, next, next_sec, next_sched;
3153 struct inet6_ifaddr *ifp;
3154 struct hlist_node *node;
3155 int i;
3156
3157 rcu_read_lock_bh();
3158 spin_lock(&addrconf_verify_lock);
3159 now = jiffies;
3160 next = round_jiffies_up(now + ADDR_CHECK_FREQUENCY);
3161
3162 del_timer(&addr_chk_timer);
3163
3164 for (i = 0; i < IN6_ADDR_HSIZE; i++) {
3165 restart:
3166 hlist_for_each_entry_rcu_bh(ifp, node,
3167 &inet6_addr_lst[i], addr_lst) {
3168 unsigned long age;
3169
3170 if (ifp->flags & IFA_F_PERMANENT)
3171 continue;
3172
3173 spin_lock(&ifp->lock);
3174 /* We try to batch several events at once. */
3175 age = (now - ifp->tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
3176
3177 if (ifp->valid_lft != INFINITY_LIFE_TIME &&
3178 age >= ifp->valid_lft) {
3179 spin_unlock(&ifp->lock);
3180 in6_ifa_hold(ifp);
3181 ipv6_del_addr(ifp);
3182 goto restart;
3183 } else if (ifp->prefered_lft == INFINITY_LIFE_TIME) {
3184 spin_unlock(&ifp->lock);
3185 continue;
3186 } else if (age >= ifp->prefered_lft) {
3187 /* jiffies - ifp->tstamp > age >= ifp->prefered_lft */
3188 int deprecate = 0;
3189
3190 if (!(ifp->flags&IFA_F_DEPRECATED)) {
3191 deprecate = 1;
3192 ifp->flags |= IFA_F_DEPRECATED;
3193 }
3194
3195 if (time_before(ifp->tstamp + ifp->valid_lft * HZ, next))
3196 next = ifp->tstamp + ifp->valid_lft * HZ;
3197
3198 spin_unlock(&ifp->lock);
3199
3200 if (deprecate) {
3201 in6_ifa_hold(ifp);
3202
3203 ipv6_ifa_notify(0, ifp);
3204 in6_ifa_put(ifp);
3205 goto restart;
3206 }
3207 #ifdef CONFIG_IPV6_PRIVACY
3208 } else if ((ifp->flags&IFA_F_TEMPORARY) &&
3209 !(ifp->flags&IFA_F_TENTATIVE)) {
3210 unsigned long regen_advance = ifp->idev->cnf.regen_max_retry *
3211 ifp->idev->cnf.dad_transmits *
3212 ifp->idev->nd_parms->retrans_time / HZ;
3213
3214 if (age >= ifp->prefered_lft - regen_advance) {
3215 struct inet6_ifaddr *ifpub = ifp->ifpub;
3216 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3217 next = ifp->tstamp + ifp->prefered_lft * HZ;
3218 if (!ifp->regen_count && ifpub) {
3219 ifp->regen_count++;
3220 in6_ifa_hold(ifp);
3221 in6_ifa_hold(ifpub);
3222 spin_unlock(&ifp->lock);
3223
3224 spin_lock(&ifpub->lock);
3225 ifpub->regen_count = 0;
3226 spin_unlock(&ifpub->lock);
3227 ipv6_create_tempaddr(ifpub, ifp);
3228 in6_ifa_put(ifpub);
3229 in6_ifa_put(ifp);
3230 goto restart;
3231 }
3232 } else if (time_before(ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ, next))
3233 next = ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ;
3234 spin_unlock(&ifp->lock);
3235 #endif
3236 } else {
3237 /* ifp->prefered_lft <= ifp->valid_lft */
3238 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3239 next = ifp->tstamp + ifp->prefered_lft * HZ;
3240 spin_unlock(&ifp->lock);
3241 }
3242 }
3243 }
3244
3245 next_sec = round_jiffies_up(next);
3246 next_sched = next;
3247
3248 /* If rounded timeout is accurate enough, accept it. */
3249 if (time_before(next_sec, next + ADDRCONF_TIMER_FUZZ))
3250 next_sched = next_sec;
3251
3252 /* And minimum interval is ADDRCONF_TIMER_FUZZ_MAX. */
3253 if (time_before(next_sched, jiffies + ADDRCONF_TIMER_FUZZ_MAX))
3254 next_sched = jiffies + ADDRCONF_TIMER_FUZZ_MAX;
3255
3256 ADBG((KERN_DEBUG "now = %lu, schedule = %lu, rounded schedule = %lu => %lu\n",
3257 now, next, next_sec, next_sched));
3258
3259 addr_chk_timer.expires = next_sched;
3260 add_timer(&addr_chk_timer);
3261 spin_unlock(&addrconf_verify_lock);
3262 rcu_read_unlock_bh();
3263 }
3264
3265 static struct in6_addr *extract_addr(struct nlattr *addr, struct nlattr *local)
3266 {
3267 struct in6_addr *pfx = NULL;
3268
3269 if (addr)
3270 pfx = nla_data(addr);
3271
3272 if (local) {
3273 if (pfx && nla_memcmp(local, pfx, sizeof(*pfx)))
3274 pfx = NULL;
3275 else
3276 pfx = nla_data(local);
3277 }
3278
3279 return pfx;
3280 }
3281
3282 static const struct nla_policy ifa_ipv6_policy[IFA_MAX+1] = {
3283 [IFA_ADDRESS] = { .len = sizeof(struct in6_addr) },
3284 [IFA_LOCAL] = { .len = sizeof(struct in6_addr) },
3285 [IFA_CACHEINFO] = { .len = sizeof(struct ifa_cacheinfo) },
3286 };
3287
3288 static int
3289 inet6_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
3290 {
3291 struct net *net = sock_net(skb->sk);
3292 struct ifaddrmsg *ifm;
3293 struct nlattr *tb[IFA_MAX+1];
3294 struct in6_addr *pfx;
3295 int err;
3296
3297 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3298 if (err < 0)
3299 return err;
3300
3301 ifm = nlmsg_data(nlh);
3302 pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3303 if (pfx == NULL)
3304 return -EINVAL;
3305
3306 return inet6_addr_del(net, ifm->ifa_index, pfx, ifm->ifa_prefixlen);
3307 }
3308
3309 static int inet6_addr_modify(struct inet6_ifaddr *ifp, u8 ifa_flags,
3310 u32 prefered_lft, u32 valid_lft)
3311 {
3312 u32 flags;
3313 clock_t expires;
3314 unsigned long timeout;
3315
3316 if (!valid_lft || (prefered_lft > valid_lft))
3317 return -EINVAL;
3318
3319 timeout = addrconf_timeout_fixup(valid_lft, HZ);
3320 if (addrconf_finite_timeout(timeout)) {
3321 expires = jiffies_to_clock_t(timeout * HZ);
3322 valid_lft = timeout;
3323 flags = RTF_EXPIRES;
3324 } else {
3325 expires = 0;
3326 flags = 0;
3327 ifa_flags |= IFA_F_PERMANENT;
3328 }
3329
3330 timeout = addrconf_timeout_fixup(prefered_lft, HZ);
3331 if (addrconf_finite_timeout(timeout)) {
3332 if (timeout == 0)
3333 ifa_flags |= IFA_F_DEPRECATED;
3334 prefered_lft = timeout;
3335 }
3336
3337 spin_lock_bh(&ifp->lock);
3338 ifp->flags = (ifp->flags & ~(IFA_F_DEPRECATED | IFA_F_PERMANENT | IFA_F_NODAD | IFA_F_HOMEADDRESS)) | ifa_flags;
3339 ifp->tstamp = jiffies;
3340 ifp->valid_lft = valid_lft;
3341 ifp->prefered_lft = prefered_lft;
3342
3343 spin_unlock_bh(&ifp->lock);
3344 if (!(ifp->flags&IFA_F_TENTATIVE))
3345 ipv6_ifa_notify(0, ifp);
3346
3347 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, ifp->idev->dev,
3348 expires, flags);
3349 addrconf_verify(0);
3350
3351 return 0;
3352 }
3353
3354 static int
3355 inet6_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
3356 {
3357 struct net *net = sock_net(skb->sk);
3358 struct ifaddrmsg *ifm;
3359 struct nlattr *tb[IFA_MAX+1];
3360 struct in6_addr *pfx;
3361 struct inet6_ifaddr *ifa;
3362 struct net_device *dev;
3363 u32 valid_lft = INFINITY_LIFE_TIME, preferred_lft = INFINITY_LIFE_TIME;
3364 u8 ifa_flags;
3365 int err;
3366
3367 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3368 if (err < 0)
3369 return err;
3370
3371 ifm = nlmsg_data(nlh);
3372 pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3373 if (pfx == NULL)
3374 return -EINVAL;
3375
3376 if (tb[IFA_CACHEINFO]) {
3377 struct ifa_cacheinfo *ci;
3378
3379 ci = nla_data(tb[IFA_CACHEINFO]);
3380 valid_lft = ci->ifa_valid;
3381 preferred_lft = ci->ifa_prefered;
3382 } else {
3383 preferred_lft = INFINITY_LIFE_TIME;
3384 valid_lft = INFINITY_LIFE_TIME;
3385 }
3386
3387 dev = __dev_get_by_index(net, ifm->ifa_index);
3388 if (dev == NULL)
3389 return -ENODEV;
3390
3391 /* We ignore other flags so far. */
3392 ifa_flags = ifm->ifa_flags & (IFA_F_NODAD | IFA_F_HOMEADDRESS);
3393
3394 ifa = ipv6_get_ifaddr(net, pfx, dev, 1);
3395 if (ifa == NULL) {
3396 /*
3397 * It would be best to check for !NLM_F_CREATE here but
3398 * userspace alreay relies on not having to provide this.
3399 */
3400 return inet6_addr_add(net, ifm->ifa_index, pfx,
3401 ifm->ifa_prefixlen, ifa_flags,
3402 preferred_lft, valid_lft);
3403 }
3404
3405 if (nlh->nlmsg_flags & NLM_F_EXCL ||
3406 !(nlh->nlmsg_flags & NLM_F_REPLACE))
3407 err = -EEXIST;
3408 else
3409 err = inet6_addr_modify(ifa, ifa_flags, preferred_lft, valid_lft);
3410
3411 in6_ifa_put(ifa);
3412
3413 return err;
3414 }
3415
3416 static void put_ifaddrmsg(struct nlmsghdr *nlh, u8 prefixlen, u8 flags,
3417 u8 scope, int ifindex)
3418 {
3419 struct ifaddrmsg *ifm;
3420
3421 ifm = nlmsg_data(nlh);
3422 ifm->ifa_family = AF_INET6;
3423 ifm->ifa_prefixlen = prefixlen;
3424 ifm->ifa_flags = flags;
3425 ifm->ifa_scope = scope;
3426 ifm->ifa_index = ifindex;
3427 }
3428
3429 static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
3430 unsigned long tstamp, u32 preferred, u32 valid)
3431 {
3432 struct ifa_cacheinfo ci;
3433
3434 ci.cstamp = cstamp_delta(cstamp);
3435 ci.tstamp = cstamp_delta(tstamp);
3436 ci.ifa_prefered = preferred;
3437 ci.ifa_valid = valid;
3438
3439 return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
3440 }
3441
3442 static inline int rt_scope(int ifa_scope)
3443 {
3444 if (ifa_scope & IFA_HOST)
3445 return RT_SCOPE_HOST;
3446 else if (ifa_scope & IFA_LINK)
3447 return RT_SCOPE_LINK;
3448 else if (ifa_scope & IFA_SITE)
3449 return RT_SCOPE_SITE;
3450 else
3451 return RT_SCOPE_UNIVERSE;
3452 }
3453
3454 static inline int inet6_ifaddr_msgsize(void)
3455 {
3456 return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
3457 + nla_total_size(16) /* IFA_ADDRESS */
3458 + nla_total_size(sizeof(struct ifa_cacheinfo));
3459 }
3460
3461 static int inet6_fill_ifaddr(struct sk_buff *skb, struct inet6_ifaddr *ifa,
3462 u32 pid, u32 seq, int event, unsigned int flags)
3463 {
3464 struct nlmsghdr *nlh;
3465 u32 preferred, valid;
3466
3467 nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3468 if (nlh == NULL)
3469 return -EMSGSIZE;
3470
3471 put_ifaddrmsg(nlh, ifa->prefix_len, ifa->flags, rt_scope(ifa->scope),
3472 ifa->idev->dev->ifindex);
3473
3474 if (!(ifa->flags&IFA_F_PERMANENT)) {
3475 preferred = ifa->prefered_lft;
3476 valid = ifa->valid_lft;
3477 if (preferred != INFINITY_LIFE_TIME) {
3478 long tval = (jiffies - ifa->tstamp)/HZ;
3479 if (preferred > tval)
3480 preferred -= tval;
3481 else
3482 preferred = 0;
3483 if (valid != INFINITY_LIFE_TIME) {
3484 if (valid > tval)
3485 valid -= tval;
3486 else
3487 valid = 0;
3488 }
3489 }
3490 } else {
3491 preferred = INFINITY_LIFE_TIME;
3492 valid = INFINITY_LIFE_TIME;
3493 }
3494
3495 if (nla_put(skb, IFA_ADDRESS, 16, &ifa->addr) < 0 ||
3496 put_cacheinfo(skb, ifa->cstamp, ifa->tstamp, preferred, valid) < 0) {
3497 nlmsg_cancel(skb, nlh);
3498 return -EMSGSIZE;
3499 }
3500
3501 return nlmsg_end(skb, nlh);
3502 }
3503
3504 static int inet6_fill_ifmcaddr(struct sk_buff *skb, struct ifmcaddr6 *ifmca,
3505 u32 pid, u32 seq, int event, u16 flags)
3506 {
3507 struct nlmsghdr *nlh;
3508 u8 scope = RT_SCOPE_UNIVERSE;
3509 int ifindex = ifmca->idev->dev->ifindex;
3510
3511 if (ipv6_addr_scope(&ifmca->mca_addr) & IFA_SITE)
3512 scope = RT_SCOPE_SITE;
3513
3514 nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3515 if (nlh == NULL)
3516 return -EMSGSIZE;
3517
3518 put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3519 if (nla_put(skb, IFA_MULTICAST, 16, &ifmca->mca_addr) < 0 ||
3520 put_cacheinfo(skb, ifmca->mca_cstamp, ifmca->mca_tstamp,
3521 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3522 nlmsg_cancel(skb, nlh);
3523 return -EMSGSIZE;
3524 }
3525
3526 return nlmsg_end(skb, nlh);
3527 }
3528
3529 static int inet6_fill_ifacaddr(struct sk_buff *skb, struct ifacaddr6 *ifaca,
3530 u32 pid, u32 seq, int event, unsigned int flags)
3531 {
3532 struct nlmsghdr *nlh;
3533 u8 scope = RT_SCOPE_UNIVERSE;
3534 int ifindex = ifaca->aca_idev->dev->ifindex;
3535
3536 if (ipv6_addr_scope(&ifaca->aca_addr) & IFA_SITE)
3537 scope = RT_SCOPE_SITE;
3538
3539 nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3540 if (nlh == NULL)
3541 return -EMSGSIZE;
3542
3543 put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3544 if (nla_put(skb, IFA_ANYCAST, 16, &ifaca->aca_addr) < 0 ||
3545 put_cacheinfo(skb, ifaca->aca_cstamp, ifaca->aca_tstamp,
3546 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3547 nlmsg_cancel(skb, nlh);
3548 return -EMSGSIZE;
3549 }
3550
3551 return nlmsg_end(skb, nlh);
3552 }
3553
3554 enum addr_type_t {
3555 UNICAST_ADDR,
3556 MULTICAST_ADDR,
3557 ANYCAST_ADDR,
3558 };
3559
3560 /* called with rcu_read_lock() */
3561 static int in6_dump_addrs(struct inet6_dev *idev, struct sk_buff *skb,
3562 struct netlink_callback *cb, enum addr_type_t type,
3563 int s_ip_idx, int *p_ip_idx)
3564 {
3565 struct ifmcaddr6 *ifmca;
3566 struct ifacaddr6 *ifaca;
3567 int err = 1;
3568 int ip_idx = *p_ip_idx;
3569
3570 read_lock_bh(&idev->lock);
3571 switch (type) {
3572 case UNICAST_ADDR: {
3573 struct inet6_ifaddr *ifa;
3574
3575 /* unicast address incl. temp addr */
3576 list_for_each_entry(ifa, &idev->addr_list, if_list) {
3577 if (++ip_idx < s_ip_idx)
3578 continue;
3579 err = inet6_fill_ifaddr(skb, ifa,
3580 NETLINK_CB(cb->skb).pid,
3581 cb->nlh->nlmsg_seq,
3582 RTM_NEWADDR,
3583 NLM_F_MULTI);
3584 if (err <= 0)
3585 break;
3586 }
3587 break;
3588 }
3589 case MULTICAST_ADDR:
3590 /* multicast address */
3591 for (ifmca = idev->mc_list; ifmca;
3592 ifmca = ifmca->next, ip_idx++) {
3593 if (ip_idx < s_ip_idx)
3594 continue;
3595 err = inet6_fill_ifmcaddr(skb, ifmca,
3596 NETLINK_CB(cb->skb).pid,
3597 cb->nlh->nlmsg_seq,
3598 RTM_GETMULTICAST,
3599 NLM_F_MULTI);
3600 if (err <= 0)
3601 break;
3602 }
3603 break;
3604 case ANYCAST_ADDR:
3605 /* anycast address */
3606 for (ifaca = idev->ac_list; ifaca;
3607 ifaca = ifaca->aca_next, ip_idx++) {
3608 if (ip_idx < s_ip_idx)
3609 continue;
3610 err = inet6_fill_ifacaddr(skb, ifaca,
3611 NETLINK_CB(cb->skb).pid,
3612 cb->nlh->nlmsg_seq,
3613 RTM_GETANYCAST,
3614 NLM_F_MULTI);
3615 if (err <= 0)
3616 break;
3617 }
3618 break;
3619 default:
3620 break;
3621 }
3622 read_unlock_bh(&idev->lock);
3623 *p_ip_idx = ip_idx;
3624 return err;
3625 }
3626
3627 static int inet6_dump_addr(struct sk_buff *skb, struct netlink_callback *cb,
3628 enum addr_type_t type)
3629 {
3630 struct net *net = sock_net(skb->sk);
3631 int h, s_h;
3632 int idx, ip_idx;
3633 int s_idx, s_ip_idx;
3634 struct net_device *dev;
3635 struct inet6_dev *idev;
3636 struct hlist_head *head;
3637 struct hlist_node *node;
3638
3639 s_h = cb->args[0];
3640 s_idx = idx = cb->args[1];
3641 s_ip_idx = ip_idx = cb->args[2];
3642
3643 rcu_read_lock();
3644 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
3645 idx = 0;
3646 head = &net->dev_index_head[h];
3647 hlist_for_each_entry_rcu(dev, node, head, index_hlist) {
3648 if (idx < s_idx)
3649 goto cont;
3650 if (h > s_h || idx > s_idx)
3651 s_ip_idx = 0;
3652 ip_idx = 0;
3653 idev = __in6_dev_get(dev);
3654 if (!idev)
3655 goto cont;
3656
3657 if (in6_dump_addrs(idev, skb, cb, type,
3658 s_ip_idx, &ip_idx) <= 0)
3659 goto done;
3660 cont:
3661 idx++;
3662 }
3663 }
3664 done:
3665 rcu_read_unlock();
3666 cb->args[0] = h;
3667 cb->args[1] = idx;
3668 cb->args[2] = ip_idx;
3669
3670 return skb->len;
3671 }
3672
3673 static int inet6_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
3674 {
3675 enum addr_type_t type = UNICAST_ADDR;
3676
3677 return inet6_dump_addr(skb, cb, type);
3678 }
3679
3680 static int inet6_dump_ifmcaddr(struct sk_buff *skb, struct netlink_callback *cb)
3681 {
3682 enum addr_type_t type = MULTICAST_ADDR;
3683
3684 return inet6_dump_addr(skb, cb, type);
3685 }
3686
3687
3688 static int inet6_dump_ifacaddr(struct sk_buff *skb, struct netlink_callback *cb)
3689 {
3690 enum addr_type_t type = ANYCAST_ADDR;
3691
3692 return inet6_dump_addr(skb, cb, type);
3693 }
3694
3695 static int inet6_rtm_getaddr(struct sk_buff *in_skb, struct nlmsghdr* nlh,
3696 void *arg)
3697 {
3698 struct net *net = sock_net(in_skb->sk);
3699 struct ifaddrmsg *ifm;
3700 struct nlattr *tb[IFA_MAX+1];
3701 struct in6_addr *addr = NULL;
3702 struct net_device *dev = NULL;
3703 struct inet6_ifaddr *ifa;
3704 struct sk_buff *skb;
3705 int err;
3706
3707 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3708 if (err < 0)
3709 goto errout;
3710
3711 addr = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3712 if (addr == NULL) {
3713 err = -EINVAL;
3714 goto errout;
3715 }
3716
3717 ifm = nlmsg_data(nlh);
3718 if (ifm->ifa_index)
3719 dev = __dev_get_by_index(net, ifm->ifa_index);
3720
3721 ifa = ipv6_get_ifaddr(net, addr, dev, 1);
3722 if (!ifa) {
3723 err = -EADDRNOTAVAIL;
3724 goto errout;
3725 }
3726
3727 skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_KERNEL);
3728 if (!skb) {
3729 err = -ENOBUFS;
3730 goto errout_ifa;
3731 }
3732
3733 err = inet6_fill_ifaddr(skb, ifa, NETLINK_CB(in_skb).pid,
3734 nlh->nlmsg_seq, RTM_NEWADDR, 0);
3735 if (err < 0) {
3736 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
3737 WARN_ON(err == -EMSGSIZE);
3738 kfree_skb(skb);
3739 goto errout_ifa;
3740 }
3741 err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).pid);
3742 errout_ifa:
3743 in6_ifa_put(ifa);
3744 errout:
3745 return err;
3746 }
3747
3748 static void inet6_ifa_notify(int event, struct inet6_ifaddr *ifa)
3749 {
3750 struct sk_buff *skb;
3751 struct net *net = dev_net(ifa->idev->dev);
3752 int err = -ENOBUFS;
3753
3754 skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_ATOMIC);
3755 if (skb == NULL)
3756 goto errout;
3757
3758 err = inet6_fill_ifaddr(skb, ifa, 0, 0, event, 0);
3759 if (err < 0) {
3760 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
3761 WARN_ON(err == -EMSGSIZE);
3762 kfree_skb(skb);
3763 goto errout;
3764 }
3765 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
3766 return;
3767 errout:
3768 if (err < 0)
3769 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
3770 }
3771
3772 static inline void ipv6_store_devconf(struct ipv6_devconf *cnf,
3773 __s32 *array, int bytes)
3774 {
3775 BUG_ON(bytes < (DEVCONF_MAX * 4));
3776
3777 memset(array, 0, bytes);
3778 array[DEVCONF_FORWARDING] = cnf->forwarding;
3779 array[DEVCONF_HOPLIMIT] = cnf->hop_limit;
3780 array[DEVCONF_MTU6] = cnf->mtu6;
3781 array[DEVCONF_ACCEPT_RA] = cnf->accept_ra;
3782 array[DEVCONF_ACCEPT_REDIRECTS] = cnf->accept_redirects;
3783 array[DEVCONF_AUTOCONF] = cnf->autoconf;
3784 array[DEVCONF_DAD_TRANSMITS] = cnf->dad_transmits;
3785 array[DEVCONF_RTR_SOLICITS] = cnf->rtr_solicits;
3786 array[DEVCONF_RTR_SOLICIT_INTERVAL] =
3787 jiffies_to_msecs(cnf->rtr_solicit_interval);
3788 array[DEVCONF_RTR_SOLICIT_DELAY] =
3789 jiffies_to_msecs(cnf->rtr_solicit_delay);
3790 array[DEVCONF_FORCE_MLD_VERSION] = cnf->force_mld_version;
3791 #ifdef CONFIG_IPV6_PRIVACY
3792 array[DEVCONF_USE_TEMPADDR] = cnf->use_tempaddr;
3793 array[DEVCONF_TEMP_VALID_LFT] = cnf->temp_valid_lft;
3794 array[DEVCONF_TEMP_PREFERED_LFT] = cnf->temp_prefered_lft;
3795 array[DEVCONF_REGEN_MAX_RETRY] = cnf->regen_max_retry;
3796 array[DEVCONF_MAX_DESYNC_FACTOR] = cnf->max_desync_factor;
3797 #endif
3798 array[DEVCONF_MAX_ADDRESSES] = cnf->max_addresses;
3799 array[DEVCONF_ACCEPT_RA_DEFRTR] = cnf->accept_ra_defrtr;
3800 array[DEVCONF_ACCEPT_RA_PINFO] = cnf->accept_ra_pinfo;
3801 #ifdef CONFIG_IPV6_ROUTER_PREF
3802 array[DEVCONF_ACCEPT_RA_RTR_PREF] = cnf->accept_ra_rtr_pref;
3803 array[DEVCONF_RTR_PROBE_INTERVAL] =
3804 jiffies_to_msecs(cnf->rtr_probe_interval);
3805 #ifdef CONFIG_IPV6_ROUTE_INFO
3806 array[DEVCONF_ACCEPT_RA_RT_INFO_MAX_PLEN] = cnf->accept_ra_rt_info_max_plen;
3807 #endif
3808 #endif
3809 array[DEVCONF_PROXY_NDP] = cnf->proxy_ndp;
3810 array[DEVCONF_ACCEPT_SOURCE_ROUTE] = cnf->accept_source_route;
3811 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
3812 array[DEVCONF_OPTIMISTIC_DAD] = cnf->optimistic_dad;
3813 #endif
3814 #ifdef CONFIG_IPV6_MROUTE
3815 array[DEVCONF_MC_FORWARDING] = cnf->mc_forwarding;
3816 #endif
3817 array[DEVCONF_DISABLE_IPV6] = cnf->disable_ipv6;
3818 array[DEVCONF_ACCEPT_DAD] = cnf->accept_dad;
3819 array[DEVCONF_FORCE_TLLAO] = cnf->force_tllao;
3820 }
3821
3822 static inline size_t inet6_ifla6_size(void)
3823 {
3824 return nla_total_size(4) /* IFLA_INET6_FLAGS */
3825 + nla_total_size(sizeof(struct ifla_cacheinfo))
3826 + nla_total_size(DEVCONF_MAX * 4) /* IFLA_INET6_CONF */
3827 + nla_total_size(IPSTATS_MIB_MAX * 8) /* IFLA_INET6_STATS */
3828 + nla_total_size(ICMP6_MIB_MAX * 8); /* IFLA_INET6_ICMP6STATS */
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(inet6_ifla6_size()); /* IFLA_PROTINFO */
3839 }
3840
3841 static inline void __snmp6_fill_stats(u64 *stats, void __percpu **mib,
3842 int items, int bytes)
3843 {
3844 int i;
3845 int pad = bytes - sizeof(u64) * items;
3846 BUG_ON(pad < 0);
3847
3848 /* Use put_unaligned() because stats may not be aligned for u64. */
3849 put_unaligned(items, &stats[0]);
3850 for (i = 1; i < items; i++)
3851 put_unaligned(snmp_fold_field(mib, i), &stats[i]);
3852
3853 memset(&stats[items], 0, pad);
3854 }
3855
3856 static inline void __snmp6_fill_stats64(u64 *stats, void __percpu **mib,
3857 int items, int bytes, size_t syncpoff)
3858 {
3859 int i;
3860 int pad = bytes - sizeof(u64) * items;
3861 BUG_ON(pad < 0);
3862
3863 /* Use put_unaligned() because stats may not be aligned for u64. */
3864 put_unaligned(items, &stats[0]);
3865 for (i = 1; i < items; i++)
3866 put_unaligned(snmp_fold_field64(mib, i, syncpoff), &stats[i]);
3867
3868 memset(&stats[items], 0, pad);
3869 }
3870
3871 static void snmp6_fill_stats(u64 *stats, struct inet6_dev *idev, int attrtype,
3872 int bytes)
3873 {
3874 switch (attrtype) {
3875 case IFLA_INET6_STATS:
3876 __snmp6_fill_stats64(stats, (void __percpu **)idev->stats.ipv6,
3877 IPSTATS_MIB_MAX, bytes, offsetof(struct ipstats_mib, syncp));
3878 break;
3879 case IFLA_INET6_ICMP6STATS:
3880 __snmp6_fill_stats(stats, (void __percpu **)idev->stats.icmpv6, ICMP6_MIB_MAX, bytes);
3881 break;
3882 }
3883 }
3884
3885 static int inet6_fill_ifla6_attrs(struct sk_buff *skb, struct inet6_dev *idev)
3886 {
3887 struct nlattr *nla;
3888 struct ifla_cacheinfo ci;
3889
3890 NLA_PUT_U32(skb, IFLA_INET6_FLAGS, idev->if_flags);
3891
3892 ci.max_reasm_len = IPV6_MAXPLEN;
3893 ci.tstamp = cstamp_delta(idev->tstamp);
3894 ci.reachable_time = jiffies_to_msecs(idev->nd_parms->reachable_time);
3895 ci.retrans_time = jiffies_to_msecs(idev->nd_parms->retrans_time);
3896 NLA_PUT(skb, IFLA_INET6_CACHEINFO, sizeof(ci), &ci);
3897
3898 nla = nla_reserve(skb, IFLA_INET6_CONF, DEVCONF_MAX * sizeof(s32));
3899 if (nla == NULL)
3900 goto nla_put_failure;
3901 ipv6_store_devconf(&idev->cnf, nla_data(nla), nla_len(nla));
3902
3903 /* XXX - MC not implemented */
3904
3905 nla = nla_reserve(skb, IFLA_INET6_STATS, IPSTATS_MIB_MAX * sizeof(u64));
3906 if (nla == NULL)
3907 goto nla_put_failure;
3908 snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_STATS, nla_len(nla));
3909
3910 nla = nla_reserve(skb, IFLA_INET6_ICMP6STATS, ICMP6_MIB_MAX * sizeof(u64));
3911 if (nla == NULL)
3912 goto nla_put_failure;
3913 snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_ICMP6STATS, nla_len(nla));
3914
3915 return 0;
3916
3917 nla_put_failure:
3918 return -EMSGSIZE;
3919 }
3920
3921 static size_t inet6_get_link_af_size(const struct net_device *dev)
3922 {
3923 if (!__in6_dev_get(dev))
3924 return 0;
3925
3926 return inet6_ifla6_size();
3927 }
3928
3929 static int inet6_fill_link_af(struct sk_buff *skb, const struct net_device *dev)
3930 {
3931 struct inet6_dev *idev = __in6_dev_get(dev);
3932
3933 if (!idev)
3934 return -ENODATA;
3935
3936 if (inet6_fill_ifla6_attrs(skb, idev) < 0)
3937 return -EMSGSIZE;
3938
3939 return 0;
3940 }
3941
3942 static int inet6_fill_ifinfo(struct sk_buff *skb, struct inet6_dev *idev,
3943 u32 pid, u32 seq, int event, unsigned int flags)
3944 {
3945 struct net_device *dev = idev->dev;
3946 struct ifinfomsg *hdr;
3947 struct nlmsghdr *nlh;
3948 void *protoinfo;
3949
3950 nlh = nlmsg_put(skb, pid, seq, event, sizeof(*hdr), flags);
3951 if (nlh == NULL)
3952 return -EMSGSIZE;
3953
3954 hdr = nlmsg_data(nlh);
3955 hdr->ifi_family = AF_INET6;
3956 hdr->__ifi_pad = 0;
3957 hdr->ifi_type = dev->type;
3958 hdr->ifi_index = dev->ifindex;
3959 hdr->ifi_flags = dev_get_flags(dev);
3960 hdr->ifi_change = 0;
3961
3962 NLA_PUT_STRING(skb, IFLA_IFNAME, dev->name);
3963
3964 if (dev->addr_len)
3965 NLA_PUT(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr);
3966
3967 NLA_PUT_U32(skb, IFLA_MTU, dev->mtu);
3968 if (dev->ifindex != dev->iflink)
3969 NLA_PUT_U32(skb, IFLA_LINK, dev->iflink);
3970
3971 protoinfo = nla_nest_start(skb, IFLA_PROTINFO);
3972 if (protoinfo == NULL)
3973 goto nla_put_failure;
3974
3975 if (inet6_fill_ifla6_attrs(skb, idev) < 0)
3976 goto nla_put_failure;
3977
3978 nla_nest_end(skb, protoinfo);
3979 return nlmsg_end(skb, nlh);
3980
3981 nla_put_failure:
3982 nlmsg_cancel(skb, nlh);
3983 return -EMSGSIZE;
3984 }
3985
3986 static int inet6_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
3987 {
3988 struct net *net = sock_net(skb->sk);
3989 int h, s_h;
3990 int idx = 0, s_idx;
3991 struct net_device *dev;
3992 struct inet6_dev *idev;
3993 struct hlist_head *head;
3994 struct hlist_node *node;
3995
3996 s_h = cb->args[0];
3997 s_idx = cb->args[1];
3998
3999 rcu_read_lock();
4000 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
4001 idx = 0;
4002 head = &net->dev_index_head[h];
4003 hlist_for_each_entry_rcu(dev, node, head, index_hlist) {
4004 if (idx < s_idx)
4005 goto cont;
4006 idev = __in6_dev_get(dev);
4007 if (!idev)
4008 goto cont;
4009 if (inet6_fill_ifinfo(skb, idev,
4010 NETLINK_CB(cb->skb).pid,
4011 cb->nlh->nlmsg_seq,
4012 RTM_NEWLINK, NLM_F_MULTI) <= 0)
4013 goto out;
4014 cont:
4015 idx++;
4016 }
4017 }
4018 out:
4019 rcu_read_unlock();
4020 cb->args[1] = idx;
4021 cb->args[0] = h;
4022
4023 return skb->len;
4024 }
4025
4026 void inet6_ifinfo_notify(int event, struct inet6_dev *idev)
4027 {
4028 struct sk_buff *skb;
4029 struct net *net = dev_net(idev->dev);
4030 int err = -ENOBUFS;
4031
4032 skb = nlmsg_new(inet6_if_nlmsg_size(), GFP_ATOMIC);
4033 if (skb == NULL)
4034 goto errout;
4035
4036 err = inet6_fill_ifinfo(skb, idev, 0, 0, event, 0);
4037 if (err < 0) {
4038 /* -EMSGSIZE implies BUG in inet6_if_nlmsg_size() */
4039 WARN_ON(err == -EMSGSIZE);
4040 kfree_skb(skb);
4041 goto errout;
4042 }
4043 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFINFO, NULL, GFP_ATOMIC);
4044 return;
4045 errout:
4046 if (err < 0)
4047 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFINFO, err);
4048 }
4049
4050 static inline size_t inet6_prefix_nlmsg_size(void)
4051 {
4052 return NLMSG_ALIGN(sizeof(struct prefixmsg))
4053 + nla_total_size(sizeof(struct in6_addr))
4054 + nla_total_size(sizeof(struct prefix_cacheinfo));
4055 }
4056
4057 static int inet6_fill_prefix(struct sk_buff *skb, struct inet6_dev *idev,
4058 struct prefix_info *pinfo, u32 pid, u32 seq,
4059 int event, unsigned int flags)
4060 {
4061 struct prefixmsg *pmsg;
4062 struct nlmsghdr *nlh;
4063 struct prefix_cacheinfo ci;
4064
4065 nlh = nlmsg_put(skb, pid, seq, event, sizeof(*pmsg), flags);
4066 if (nlh == NULL)
4067 return -EMSGSIZE;
4068
4069 pmsg = nlmsg_data(nlh);
4070 pmsg->prefix_family = AF_INET6;
4071 pmsg->prefix_pad1 = 0;
4072 pmsg->prefix_pad2 = 0;
4073 pmsg->prefix_ifindex = idev->dev->ifindex;
4074 pmsg->prefix_len = pinfo->prefix_len;
4075 pmsg->prefix_type = pinfo->type;
4076 pmsg->prefix_pad3 = 0;
4077 pmsg->prefix_flags = 0;
4078 if (pinfo->onlink)
4079 pmsg->prefix_flags |= IF_PREFIX_ONLINK;
4080 if (pinfo->autoconf)
4081 pmsg->prefix_flags |= IF_PREFIX_AUTOCONF;
4082
4083 NLA_PUT(skb, PREFIX_ADDRESS, sizeof(pinfo->prefix), &pinfo->prefix);
4084
4085 ci.preferred_time = ntohl(pinfo->prefered);
4086 ci.valid_time = ntohl(pinfo->valid);
4087 NLA_PUT(skb, PREFIX_CACHEINFO, sizeof(ci), &ci);
4088
4089 return nlmsg_end(skb, nlh);
4090
4091 nla_put_failure:
4092 nlmsg_cancel(skb, nlh);
4093 return -EMSGSIZE;
4094 }
4095
4096 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
4097 struct prefix_info *pinfo)
4098 {
4099 struct sk_buff *skb;
4100 struct net *net = dev_net(idev->dev);
4101 int err = -ENOBUFS;
4102
4103 skb = nlmsg_new(inet6_prefix_nlmsg_size(), GFP_ATOMIC);
4104 if (skb == NULL)
4105 goto errout;
4106
4107 err = inet6_fill_prefix(skb, idev, pinfo, 0, 0, event, 0);
4108 if (err < 0) {
4109 /* -EMSGSIZE implies BUG in inet6_prefix_nlmsg_size() */
4110 WARN_ON(err == -EMSGSIZE);
4111 kfree_skb(skb);
4112 goto errout;
4113 }
4114 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_PREFIX, NULL, GFP_ATOMIC);
4115 return;
4116 errout:
4117 if (err < 0)
4118 rtnl_set_sk_err(net, RTNLGRP_IPV6_PREFIX, err);
4119 }
4120
4121 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
4122 {
4123 inet6_ifa_notify(event ? : RTM_NEWADDR, ifp);
4124
4125 switch (event) {
4126 case RTM_NEWADDR:
4127 /*
4128 * If the address was optimistic
4129 * we inserted the route at the start of
4130 * our DAD process, so we don't need
4131 * to do it again
4132 */
4133 if (!(ifp->rt->rt6i_node))
4134 ip6_ins_rt(ifp->rt);
4135 if (ifp->idev->cnf.forwarding)
4136 addrconf_join_anycast(ifp);
4137 break;
4138 case RTM_DELADDR:
4139 if (ifp->idev->cnf.forwarding)
4140 addrconf_leave_anycast(ifp);
4141 addrconf_leave_solict(ifp->idev, &ifp->addr);
4142 dst_hold(&ifp->rt->dst);
4143
4144 if (ip6_del_rt(ifp->rt))
4145 dst_free(&ifp->rt->dst);
4146 break;
4147 }
4148 }
4149
4150 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
4151 {
4152 rcu_read_lock_bh();
4153 if (likely(ifp->idev->dead == 0))
4154 __ipv6_ifa_notify(event, ifp);
4155 rcu_read_unlock_bh();
4156 }
4157
4158 #ifdef CONFIG_SYSCTL
4159
4160 static
4161 int addrconf_sysctl_forward(ctl_table *ctl, int write,
4162 void __user *buffer, size_t *lenp, loff_t *ppos)
4163 {
4164 int *valp = ctl->data;
4165 int val = *valp;
4166 loff_t pos = *ppos;
4167 int ret;
4168
4169 ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
4170
4171 if (write)
4172 ret = addrconf_fixup_forwarding(ctl, valp, val);
4173 if (ret)
4174 *ppos = pos;
4175 return ret;
4176 }
4177
4178 static void dev_disable_change(struct inet6_dev *idev)
4179 {
4180 if (!idev || !idev->dev)
4181 return;
4182
4183 if (idev->cnf.disable_ipv6)
4184 addrconf_notify(NULL, NETDEV_DOWN, idev->dev);
4185 else
4186 addrconf_notify(NULL, NETDEV_UP, idev->dev);
4187 }
4188
4189 static void addrconf_disable_change(struct net *net, __s32 newf)
4190 {
4191 struct net_device *dev;
4192 struct inet6_dev *idev;
4193
4194 rcu_read_lock();
4195 for_each_netdev_rcu(net, dev) {
4196 idev = __in6_dev_get(dev);
4197 if (idev) {
4198 int changed = (!idev->cnf.disable_ipv6) ^ (!newf);
4199 idev->cnf.disable_ipv6 = newf;
4200 if (changed)
4201 dev_disable_change(idev);
4202 }
4203 }
4204 rcu_read_unlock();
4205 }
4206
4207 static int addrconf_disable_ipv6(struct ctl_table *table, int *p, int old)
4208 {
4209 struct net *net;
4210
4211 net = (struct net *)table->extra2;
4212
4213 if (p == &net->ipv6.devconf_dflt->disable_ipv6)
4214 return 0;
4215
4216 if (!rtnl_trylock()) {
4217 /* Restore the original values before restarting */
4218 *p = old;
4219 return restart_syscall();
4220 }
4221
4222 if (p == &net->ipv6.devconf_all->disable_ipv6) {
4223 __s32 newf = net->ipv6.devconf_all->disable_ipv6;
4224 net->ipv6.devconf_dflt->disable_ipv6 = newf;
4225 addrconf_disable_change(net, newf);
4226 } else if ((!*p) ^ (!old))
4227 dev_disable_change((struct inet6_dev *)table->extra1);
4228
4229 rtnl_unlock();
4230 return 0;
4231 }
4232
4233 static
4234 int addrconf_sysctl_disable(ctl_table *ctl, int write,
4235 void __user *buffer, size_t *lenp, loff_t *ppos)
4236 {
4237 int *valp = ctl->data;
4238 int val = *valp;
4239 loff_t pos = *ppos;
4240 int ret;
4241
4242 ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
4243
4244 if (write)
4245 ret = addrconf_disable_ipv6(ctl, valp, val);
4246 if (ret)
4247 *ppos = pos;
4248 return ret;
4249 }
4250
4251 static struct addrconf_sysctl_table
4252 {
4253 struct ctl_table_header *sysctl_header;
4254 ctl_table addrconf_vars[DEVCONF_MAX+1];
4255 char *dev_name;
4256 } addrconf_sysctl __read_mostly = {
4257 .sysctl_header = NULL,
4258 .addrconf_vars = {
4259 {
4260 .procname = "forwarding",
4261 .data = &ipv6_devconf.forwarding,
4262 .maxlen = sizeof(int),
4263 .mode = 0644,
4264 .proc_handler = addrconf_sysctl_forward,
4265 },
4266 {
4267 .procname = "hop_limit",
4268 .data = &ipv6_devconf.hop_limit,
4269 .maxlen = sizeof(int),
4270 .mode = 0644,
4271 .proc_handler = proc_dointvec,
4272 },
4273 {
4274 .procname = "mtu",
4275 .data = &ipv6_devconf.mtu6,
4276 .maxlen = sizeof(int),
4277 .mode = 0644,
4278 .proc_handler = proc_dointvec,
4279 },
4280 {
4281 .procname = "accept_ra",
4282 .data = &ipv6_devconf.accept_ra,
4283 .maxlen = sizeof(int),
4284 .mode = 0644,
4285 .proc_handler = proc_dointvec,
4286 },
4287 {
4288 .procname = "accept_redirects",
4289 .data = &ipv6_devconf.accept_redirects,
4290 .maxlen = sizeof(int),
4291 .mode = 0644,
4292 .proc_handler = proc_dointvec,
4293 },
4294 {
4295 .procname = "autoconf",
4296 .data = &ipv6_devconf.autoconf,
4297 .maxlen = sizeof(int),
4298 .mode = 0644,
4299 .proc_handler = proc_dointvec,
4300 },
4301 {
4302 .procname = "dad_transmits",
4303 .data = &ipv6_devconf.dad_transmits,
4304 .maxlen = sizeof(int),
4305 .mode = 0644,
4306 .proc_handler = proc_dointvec,
4307 },
4308 {
4309 .procname = "router_solicitations",
4310 .data = &ipv6_devconf.rtr_solicits,
4311 .maxlen = sizeof(int),
4312 .mode = 0644,
4313 .proc_handler = proc_dointvec,
4314 },
4315 {
4316 .procname = "router_solicitation_interval",
4317 .data = &ipv6_devconf.rtr_solicit_interval,
4318 .maxlen = sizeof(int),
4319 .mode = 0644,
4320 .proc_handler = proc_dointvec_jiffies,
4321 },
4322 {
4323 .procname = "router_solicitation_delay",
4324 .data = &ipv6_devconf.rtr_solicit_delay,
4325 .maxlen = sizeof(int),
4326 .mode = 0644,
4327 .proc_handler = proc_dointvec_jiffies,
4328 },
4329 {
4330 .procname = "force_mld_version",
4331 .data = &ipv6_devconf.force_mld_version,
4332 .maxlen = sizeof(int),
4333 .mode = 0644,
4334 .proc_handler = proc_dointvec,
4335 },
4336 #ifdef CONFIG_IPV6_PRIVACY
4337 {
4338 .procname = "use_tempaddr",
4339 .data = &ipv6_devconf.use_tempaddr,
4340 .maxlen = sizeof(int),
4341 .mode = 0644,
4342 .proc_handler = proc_dointvec,
4343 },
4344 {
4345 .procname = "temp_valid_lft",
4346 .data = &ipv6_devconf.temp_valid_lft,
4347 .maxlen = sizeof(int),
4348 .mode = 0644,
4349 .proc_handler = proc_dointvec,
4350 },
4351 {
4352 .procname = "temp_prefered_lft",
4353 .data = &ipv6_devconf.temp_prefered_lft,
4354 .maxlen = sizeof(int),
4355 .mode = 0644,
4356 .proc_handler = proc_dointvec,
4357 },
4358 {
4359 .procname = "regen_max_retry",
4360 .data = &ipv6_devconf.regen_max_retry,
4361 .maxlen = sizeof(int),
4362 .mode = 0644,
4363 .proc_handler = proc_dointvec,
4364 },
4365 {
4366 .procname = "max_desync_factor",
4367 .data = &ipv6_devconf.max_desync_factor,
4368 .maxlen = sizeof(int),
4369 .mode = 0644,
4370 .proc_handler = proc_dointvec,
4371 },
4372 #endif
4373 {
4374 .procname = "max_addresses",
4375 .data = &ipv6_devconf.max_addresses,
4376 .maxlen = sizeof(int),
4377 .mode = 0644,
4378 .proc_handler = proc_dointvec,
4379 },
4380 {
4381 .procname = "accept_ra_defrtr",
4382 .data = &ipv6_devconf.accept_ra_defrtr,
4383 .maxlen = sizeof(int),
4384 .mode = 0644,
4385 .proc_handler = proc_dointvec,
4386 },
4387 {
4388 .procname = "accept_ra_pinfo",
4389 .data = &ipv6_devconf.accept_ra_pinfo,
4390 .maxlen = sizeof(int),
4391 .mode = 0644,
4392 .proc_handler = proc_dointvec,
4393 },
4394 #ifdef CONFIG_IPV6_ROUTER_PREF
4395 {
4396 .procname = "accept_ra_rtr_pref",
4397 .data = &ipv6_devconf.accept_ra_rtr_pref,
4398 .maxlen = sizeof(int),
4399 .mode = 0644,
4400 .proc_handler = proc_dointvec,
4401 },
4402 {
4403 .procname = "router_probe_interval",
4404 .data = &ipv6_devconf.rtr_probe_interval,
4405 .maxlen = sizeof(int),
4406 .mode = 0644,
4407 .proc_handler = proc_dointvec_jiffies,
4408 },
4409 #ifdef CONFIG_IPV6_ROUTE_INFO
4410 {
4411 .procname = "accept_ra_rt_info_max_plen",
4412 .data = &ipv6_devconf.accept_ra_rt_info_max_plen,
4413 .maxlen = sizeof(int),
4414 .mode = 0644,
4415 .proc_handler = proc_dointvec,
4416 },
4417 #endif
4418 #endif
4419 {
4420 .procname = "proxy_ndp",
4421 .data = &ipv6_devconf.proxy_ndp,
4422 .maxlen = sizeof(int),
4423 .mode = 0644,
4424 .proc_handler = proc_dointvec,
4425 },
4426 {
4427 .procname = "accept_source_route",
4428 .data = &ipv6_devconf.accept_source_route,
4429 .maxlen = sizeof(int),
4430 .mode = 0644,
4431 .proc_handler = proc_dointvec,
4432 },
4433 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
4434 {
4435 .procname = "optimistic_dad",
4436 .data = &ipv6_devconf.optimistic_dad,
4437 .maxlen = sizeof(int),
4438 .mode = 0644,
4439 .proc_handler = proc_dointvec,
4440
4441 },
4442 #endif
4443 #ifdef CONFIG_IPV6_MROUTE
4444 {
4445 .procname = "mc_forwarding",
4446 .data = &ipv6_devconf.mc_forwarding,
4447 .maxlen = sizeof(int),
4448 .mode = 0444,
4449 .proc_handler = proc_dointvec,
4450 },
4451 #endif
4452 {
4453 .procname = "disable_ipv6",
4454 .data = &ipv6_devconf.disable_ipv6,
4455 .maxlen = sizeof(int),
4456 .mode = 0644,
4457 .proc_handler = addrconf_sysctl_disable,
4458 },
4459 {
4460 .procname = "accept_dad",
4461 .data = &ipv6_devconf.accept_dad,
4462 .maxlen = sizeof(int),
4463 .mode = 0644,
4464 .proc_handler = proc_dointvec,
4465 },
4466 {
4467 .procname = "force_tllao",
4468 .data = &ipv6_devconf.force_tllao,
4469 .maxlen = sizeof(int),
4470 .mode = 0644,
4471 .proc_handler = proc_dointvec
4472 },
4473 {
4474 /* sentinel */
4475 }
4476 },
4477 };
4478
4479 static int __addrconf_sysctl_register(struct net *net, char *dev_name,
4480 struct inet6_dev *idev, struct ipv6_devconf *p)
4481 {
4482 int i;
4483 struct addrconf_sysctl_table *t;
4484
4485 #define ADDRCONF_CTL_PATH_DEV 3
4486
4487 struct ctl_path addrconf_ctl_path[] = {
4488 { .procname = "net", },
4489 { .procname = "ipv6", },
4490 { .procname = "conf", },
4491 { /* to be set */ },
4492 { },
4493 };
4494
4495
4496 t = kmemdup(&addrconf_sysctl, sizeof(*t), GFP_KERNEL);
4497 if (t == NULL)
4498 goto out;
4499
4500 for (i = 0; t->addrconf_vars[i].data; i++) {
4501 t->addrconf_vars[i].data += (char *)p - (char *)&ipv6_devconf;
4502 t->addrconf_vars[i].extra1 = idev; /* embedded; no ref */
4503 t->addrconf_vars[i].extra2 = net;
4504 }
4505
4506 /*
4507 * Make a copy of dev_name, because '.procname' is regarded as const
4508 * by sysctl and we wouldn't want anyone to change it under our feet
4509 * (see SIOCSIFNAME).
4510 */
4511 t->dev_name = kstrdup(dev_name, GFP_KERNEL);
4512 if (!t->dev_name)
4513 goto free;
4514
4515 addrconf_ctl_path[ADDRCONF_CTL_PATH_DEV].procname = t->dev_name;
4516
4517 t->sysctl_header = register_net_sysctl_table(net, addrconf_ctl_path,
4518 t->addrconf_vars);
4519 if (t->sysctl_header == NULL)
4520 goto free_procname;
4521
4522 p->sysctl = t;
4523 return 0;
4524
4525 free_procname:
4526 kfree(t->dev_name);
4527 free:
4528 kfree(t);
4529 out:
4530 return -ENOBUFS;
4531 }
4532
4533 static void __addrconf_sysctl_unregister(struct ipv6_devconf *p)
4534 {
4535 struct addrconf_sysctl_table *t;
4536
4537 if (p->sysctl == NULL)
4538 return;
4539
4540 t = p->sysctl;
4541 p->sysctl = NULL;
4542 unregister_sysctl_table(t->sysctl_header);
4543 kfree(t->dev_name);
4544 kfree(t);
4545 }
4546
4547 static void addrconf_sysctl_register(struct inet6_dev *idev)
4548 {
4549 neigh_sysctl_register(idev->dev, idev->nd_parms, "ipv6",
4550 &ndisc_ifinfo_sysctl_change);
4551 __addrconf_sysctl_register(dev_net(idev->dev), idev->dev->name,
4552 idev, &idev->cnf);
4553 }
4554
4555 static void addrconf_sysctl_unregister(struct inet6_dev *idev)
4556 {
4557 __addrconf_sysctl_unregister(&idev->cnf);
4558 neigh_sysctl_unregister(idev->nd_parms);
4559 }
4560
4561
4562 #endif
4563
4564 static int __net_init addrconf_init_net(struct net *net)
4565 {
4566 int err;
4567 struct ipv6_devconf *all, *dflt;
4568
4569 err = -ENOMEM;
4570 all = &ipv6_devconf;
4571 dflt = &ipv6_devconf_dflt;
4572
4573 if (!net_eq(net, &init_net)) {
4574 all = kmemdup(all, sizeof(ipv6_devconf), GFP_KERNEL);
4575 if (all == NULL)
4576 goto err_alloc_all;
4577
4578 dflt = kmemdup(dflt, sizeof(ipv6_devconf_dflt), GFP_KERNEL);
4579 if (dflt == NULL)
4580 goto err_alloc_dflt;
4581 } else {
4582 /* these will be inherited by all namespaces */
4583 dflt->autoconf = ipv6_defaults.autoconf;
4584 dflt->disable_ipv6 = ipv6_defaults.disable_ipv6;
4585 }
4586
4587 net->ipv6.devconf_all = all;
4588 net->ipv6.devconf_dflt = dflt;
4589
4590 #ifdef CONFIG_SYSCTL
4591 err = __addrconf_sysctl_register(net, "all", NULL, all);
4592 if (err < 0)
4593 goto err_reg_all;
4594
4595 err = __addrconf_sysctl_register(net, "default", NULL, dflt);
4596 if (err < 0)
4597 goto err_reg_dflt;
4598 #endif
4599 return 0;
4600
4601 #ifdef CONFIG_SYSCTL
4602 err_reg_dflt:
4603 __addrconf_sysctl_unregister(all);
4604 err_reg_all:
4605 kfree(dflt);
4606 #endif
4607 err_alloc_dflt:
4608 kfree(all);
4609 err_alloc_all:
4610 return err;
4611 }
4612
4613 static void __net_exit addrconf_exit_net(struct net *net)
4614 {
4615 #ifdef CONFIG_SYSCTL
4616 __addrconf_sysctl_unregister(net->ipv6.devconf_dflt);
4617 __addrconf_sysctl_unregister(net->ipv6.devconf_all);
4618 #endif
4619 if (!net_eq(net, &init_net)) {
4620 kfree(net->ipv6.devconf_dflt);
4621 kfree(net->ipv6.devconf_all);
4622 }
4623 }
4624
4625 static struct pernet_operations addrconf_ops = {
4626 .init = addrconf_init_net,
4627 .exit = addrconf_exit_net,
4628 };
4629
4630 /*
4631 * Device notifier
4632 */
4633
4634 int register_inet6addr_notifier(struct notifier_block *nb)
4635 {
4636 return atomic_notifier_chain_register(&inet6addr_chain, nb);
4637 }
4638 EXPORT_SYMBOL(register_inet6addr_notifier);
4639
4640 int unregister_inet6addr_notifier(struct notifier_block *nb)
4641 {
4642 return atomic_notifier_chain_unregister(&inet6addr_chain, nb);
4643 }
4644 EXPORT_SYMBOL(unregister_inet6addr_notifier);
4645
4646 static struct rtnl_af_ops inet6_ops = {
4647 .family = AF_INET6,
4648 .fill_link_af = inet6_fill_link_af,
4649 .get_link_af_size = inet6_get_link_af_size,
4650 };
4651
4652 /*
4653 * Init / cleanup code
4654 */
4655
4656 int __init addrconf_init(void)
4657 {
4658 int i, err;
4659
4660 err = ipv6_addr_label_init();
4661 if (err < 0) {
4662 printk(KERN_CRIT "IPv6 Addrconf:"
4663 " cannot initialize default policy table: %d.\n", err);
4664 goto out;
4665 }
4666
4667 err = register_pernet_subsys(&addrconf_ops);
4668 if (err < 0)
4669 goto out_addrlabel;
4670
4671 /* The addrconf netdev notifier requires that loopback_dev
4672 * has it's ipv6 private information allocated and setup
4673 * before it can bring up and give link-local addresses
4674 * to other devices which are up.
4675 *
4676 * Unfortunately, loopback_dev is not necessarily the first
4677 * entry in the global dev_base list of net devices. In fact,
4678 * it is likely to be the very last entry on that list.
4679 * So this causes the notifier registry below to try and
4680 * give link-local addresses to all devices besides loopback_dev
4681 * first, then loopback_dev, which cases all the non-loopback_dev
4682 * devices to fail to get a link-local address.
4683 *
4684 * So, as a temporary fix, allocate the ipv6 structure for
4685 * loopback_dev first by hand.
4686 * Longer term, all of the dependencies ipv6 has upon the loopback
4687 * device and it being up should be removed.
4688 */
4689 rtnl_lock();
4690 if (!ipv6_add_dev(init_net.loopback_dev))
4691 err = -ENOMEM;
4692 rtnl_unlock();
4693 if (err)
4694 goto errlo;
4695
4696 for (i = 0; i < IN6_ADDR_HSIZE; i++)
4697 INIT_HLIST_HEAD(&inet6_addr_lst[i]);
4698
4699 register_netdevice_notifier(&ipv6_dev_notf);
4700
4701 addrconf_verify(0);
4702
4703 err = rtnl_af_register(&inet6_ops);
4704 if (err < 0)
4705 goto errout_af;
4706
4707 err = __rtnl_register(PF_INET6, RTM_GETLINK, NULL, inet6_dump_ifinfo);
4708 if (err < 0)
4709 goto errout;
4710
4711 /* Only the first call to __rtnl_register can fail */
4712 __rtnl_register(PF_INET6, RTM_NEWADDR, inet6_rtm_newaddr, NULL);
4713 __rtnl_register(PF_INET6, RTM_DELADDR, inet6_rtm_deladdr, NULL);
4714 __rtnl_register(PF_INET6, RTM_GETADDR, inet6_rtm_getaddr, inet6_dump_ifaddr);
4715 __rtnl_register(PF_INET6, RTM_GETMULTICAST, NULL, inet6_dump_ifmcaddr);
4716 __rtnl_register(PF_INET6, RTM_GETANYCAST, NULL, inet6_dump_ifacaddr);
4717
4718 ipv6_addr_label_rtnl_register();
4719
4720 return 0;
4721 errout:
4722 rtnl_af_unregister(&inet6_ops);
4723 errout_af:
4724 unregister_netdevice_notifier(&ipv6_dev_notf);
4725 errlo:
4726 unregister_pernet_subsys(&addrconf_ops);
4727 out_addrlabel:
4728 ipv6_addr_label_cleanup();
4729 out:
4730 return err;
4731 }
4732
4733 void addrconf_cleanup(void)
4734 {
4735 struct net_device *dev;
4736 int i;
4737
4738 unregister_netdevice_notifier(&ipv6_dev_notf);
4739 unregister_pernet_subsys(&addrconf_ops);
4740 ipv6_addr_label_cleanup();
4741
4742 rtnl_lock();
4743
4744 __rtnl_af_unregister(&inet6_ops);
4745
4746 /* clean dev list */
4747 for_each_netdev(&init_net, dev) {
4748 if (__in6_dev_get(dev) == NULL)
4749 continue;
4750 addrconf_ifdown(dev, 1);
4751 }
4752 addrconf_ifdown(init_net.loopback_dev, 2);
4753
4754 /*
4755 * Check hash table.
4756 */
4757 spin_lock_bh(&addrconf_hash_lock);
4758 for (i = 0; i < IN6_ADDR_HSIZE; i++)
4759 WARN_ON(!hlist_empty(&inet6_addr_lst[i]));
4760 spin_unlock_bh(&addrconf_hash_lock);
4761
4762 del_timer(&addr_chk_timer);
4763 rtnl_unlock();
4764 }
linux-next