?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
Merge tag 'remoteproc-3.11-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git...
[linux-2.6.git] / net / ipv6 / addrconf.c
blobcfdcf7b2daf68ff681608351e0bca7202923824b
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 #define pr_fmt(fmt) "IPv6: " fmt
42
43 #include <linux/errno.h>
44 #include <linux/types.h>
45 #include <linux/kernel.h>
46 #include <linux/socket.h>
47 #include <linux/sockios.h>
48 #include <linux/net.h>
49 #include <linux/in6.h>
50 #include <linux/netdevice.h>
51 #include <linux/if_addr.h>
52 #include <linux/if_arp.h>
53 #include <linux/if_arcnet.h>
54 #include <linux/if_infiniband.h>
55 #include <linux/route.h>
56 #include <linux/inetdevice.h>
57 #include <linux/init.h>
58 #include <linux/slab.h>
59 #ifdef CONFIG_SYSCTL
60 #include <linux/sysctl.h>
61 #endif
62 #include <linux/capability.h>
63 #include <linux/delay.h>
64 #include <linux/notifier.h>
65 #include <linux/string.h>
66 #include <linux/hash.h>
67
68 #include <net/net_namespace.h>
69 #include <net/sock.h>
70 #include <net/snmp.h>
71
72 #include <net/af_ieee802154.h>
73 #include <net/firewire.h>
74 #include <net/ipv6.h>
75 #include <net/protocol.h>
76 #include <net/ndisc.h>
77 #include <net/ip6_route.h>
78 #include <net/addrconf.h>
79 #include <net/tcp.h>
80 #include <net/ip.h>
81 #include <net/netlink.h>
82 #include <net/pkt_sched.h>
83 #include <linux/if_tunnel.h>
84 #include <linux/rtnetlink.h>
85 #include <linux/netconf.h>
86
87 #ifdef CONFIG_IPV6_PRIVACY
88 #include <linux/random.h>
89 #endif
90
91 #include <linux/uaccess.h>
92 #include <asm/unaligned.h>
93
94 #include <linux/proc_fs.h>
95 #include <linux/seq_file.h>
96 #include <linux/export.h>
97
98 /* Set to 3 to get tracing... */
99 #define ACONF_DEBUG 2
100
101 #if ACONF_DEBUG >= 3
102 #define ADBG(x) printk x
103 #else
104 #define ADBG(x)
105 #endif
106
107 #define INFINITY_LIFE_TIME 0xFFFFFFFF
108
109 static inline u32 cstamp_delta(unsigned long cstamp)
110 {
111 return (cstamp - INITIAL_JIFFIES) * 100UL / HZ;
112 }
113
114 #ifdef CONFIG_SYSCTL
115 static void addrconf_sysctl_register(struct inet6_dev *idev);
116 static void addrconf_sysctl_unregister(struct inet6_dev *idev);
117 #else
118 static inline void addrconf_sysctl_register(struct inet6_dev *idev)
119 {
120 }
121
122 static inline void addrconf_sysctl_unregister(struct inet6_dev *idev)
123 {
124 }
125 #endif
126
127 #ifdef CONFIG_IPV6_PRIVACY
128 static void __ipv6_regen_rndid(struct inet6_dev *idev);
129 static void __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr);
130 static void ipv6_regen_rndid(unsigned long data);
131 #endif
132
133 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev);
134 static int ipv6_count_addresses(struct inet6_dev *idev);
135
136 /*
137 * Configured unicast address hash table
138 */
139 static struct hlist_head inet6_addr_lst[IN6_ADDR_HSIZE];
140 static DEFINE_SPINLOCK(addrconf_hash_lock);
141
142 static void addrconf_verify(unsigned long);
143
144 static DEFINE_TIMER(addr_chk_timer, addrconf_verify, 0, 0);
145 static DEFINE_SPINLOCK(addrconf_verify_lock);
146
147 static void addrconf_join_anycast(struct inet6_ifaddr *ifp);
148 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp);
149
150 static void addrconf_type_change(struct net_device *dev,
151 unsigned long event);
152 static int addrconf_ifdown(struct net_device *dev, int how);
153
154 static struct rt6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
155 int plen,
156 const struct net_device *dev,
157 u32 flags, u32 noflags);
158
159 static void addrconf_dad_start(struct inet6_ifaddr *ifp);
160 static void addrconf_dad_timer(unsigned long data);
161 static void addrconf_dad_completed(struct inet6_ifaddr *ifp);
162 static void addrconf_dad_run(struct inet6_dev *idev);
163 static void addrconf_rs_timer(unsigned long data);
164 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
165 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
166
167 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
168 struct prefix_info *pinfo);
169 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
170 struct net_device *dev);
171
172 static struct ipv6_devconf ipv6_devconf __read_mostly = {
173 .forwarding = 0,
174 .hop_limit = IPV6_DEFAULT_HOPLIMIT,
175 .mtu6 = IPV6_MIN_MTU,
176 .accept_ra = 1,
177 .accept_redirects = 1,
178 .autoconf = 1,
179 .force_mld_version = 0,
180 .dad_transmits = 1,
181 .rtr_solicits = MAX_RTR_SOLICITATIONS,
182 .rtr_solicit_interval = RTR_SOLICITATION_INTERVAL,
183 .rtr_solicit_delay = MAX_RTR_SOLICITATION_DELAY,
184 #ifdef CONFIG_IPV6_PRIVACY
185 .use_tempaddr = 0,
186 .temp_valid_lft = TEMP_VALID_LIFETIME,
187 .temp_prefered_lft = TEMP_PREFERRED_LIFETIME,
188 .regen_max_retry = REGEN_MAX_RETRY,
189 .max_desync_factor = MAX_DESYNC_FACTOR,
190 #endif
191 .max_addresses = IPV6_MAX_ADDRESSES,
192 .accept_ra_defrtr = 1,
193 .accept_ra_pinfo = 1,
194 #ifdef CONFIG_IPV6_ROUTER_PREF
195 .accept_ra_rtr_pref = 1,
196 .rtr_probe_interval = 60 * HZ,
197 #ifdef CONFIG_IPV6_ROUTE_INFO
198 .accept_ra_rt_info_max_plen = 0,
199 #endif
200 #endif
201 .proxy_ndp = 0,
202 .accept_source_route = 0, /* we do not accept RH0 by default. */
203 .disable_ipv6 = 0,
204 .accept_dad = 1,
205 };
206
207 static struct ipv6_devconf ipv6_devconf_dflt __read_mostly = {
208 .forwarding = 0,
209 .hop_limit = IPV6_DEFAULT_HOPLIMIT,
210 .mtu6 = IPV6_MIN_MTU,
211 .accept_ra = 1,
212 .accept_redirects = 1,
213 .autoconf = 1,
214 .dad_transmits = 1,
215 .rtr_solicits = MAX_RTR_SOLICITATIONS,
216 .rtr_solicit_interval = RTR_SOLICITATION_INTERVAL,
217 .rtr_solicit_delay = MAX_RTR_SOLICITATION_DELAY,
218 #ifdef CONFIG_IPV6_PRIVACY
219 .use_tempaddr = 0,
220 .temp_valid_lft = TEMP_VALID_LIFETIME,
221 .temp_prefered_lft = TEMP_PREFERRED_LIFETIME,
222 .regen_max_retry = REGEN_MAX_RETRY,
223 .max_desync_factor = MAX_DESYNC_FACTOR,
224 #endif
225 .max_addresses = IPV6_MAX_ADDRESSES,
226 .accept_ra_defrtr = 1,
227 .accept_ra_pinfo = 1,
228 #ifdef CONFIG_IPV6_ROUTER_PREF
229 .accept_ra_rtr_pref = 1,
230 .rtr_probe_interval = 60 * HZ,
231 #ifdef CONFIG_IPV6_ROUTE_INFO
232 .accept_ra_rt_info_max_plen = 0,
233 #endif
234 #endif
235 .proxy_ndp = 0,
236 .accept_source_route = 0, /* we do not accept RH0 by default. */
237 .disable_ipv6 = 0,
238 .accept_dad = 1,
239 };
240
241 /* IPv6 Wildcard Address and Loopback Address defined by RFC2553 */
242 const struct in6_addr in6addr_any = IN6ADDR_ANY_INIT;
243 const struct in6_addr in6addr_loopback = IN6ADDR_LOOPBACK_INIT;
244 const struct in6_addr in6addr_linklocal_allnodes = IN6ADDR_LINKLOCAL_ALLNODES_INIT;
245 const struct in6_addr in6addr_linklocal_allrouters = IN6ADDR_LINKLOCAL_ALLROUTERS_INIT;
246 const struct in6_addr in6addr_interfacelocal_allnodes = IN6ADDR_INTERFACELOCAL_ALLNODES_INIT;
247 const struct in6_addr in6addr_interfacelocal_allrouters = IN6ADDR_INTERFACELOCAL_ALLROUTERS_INIT;
248 const struct in6_addr in6addr_sitelocal_allrouters = IN6ADDR_SITELOCAL_ALLROUTERS_INIT;
249
250 /* Check if a valid qdisc is available */
251 static inline bool addrconf_qdisc_ok(const struct net_device *dev)
252 {
253 return !qdisc_tx_is_noop(dev);
254 }
255
256 static void addrconf_del_rs_timer(struct inet6_dev *idev)
257 {
258 if (del_timer(&idev->rs_timer))
259 __in6_dev_put(idev);
260 }
261
262 static void addrconf_del_dad_timer(struct inet6_ifaddr *ifp)
263 {
264 if (del_timer(&ifp->dad_timer))
265 __in6_ifa_put(ifp);
266 }
267
268 static void addrconf_mod_rs_timer(struct inet6_dev *idev,
269 unsigned long when)
270 {
271 if (!timer_pending(&idev->rs_timer))
272 in6_dev_hold(idev);
273 mod_timer(&idev->rs_timer, jiffies + when);
274 }
275
276 static void addrconf_mod_dad_timer(struct inet6_ifaddr *ifp,
277 unsigned long when)
278 {
279 if (!timer_pending(&ifp->dad_timer))
280 in6_ifa_hold(ifp);
281 mod_timer(&ifp->dad_timer, jiffies + when);
282 }
283
284 static int snmp6_alloc_dev(struct inet6_dev *idev)
285 {
286 if (snmp_mib_init((void __percpu **)idev->stats.ipv6,
287 sizeof(struct ipstats_mib),
288 __alignof__(struct ipstats_mib)) < 0)
289 goto err_ip;
290 idev->stats.icmpv6dev = kzalloc(sizeof(struct icmpv6_mib_device),
291 GFP_KERNEL);
292 if (!idev->stats.icmpv6dev)
293 goto err_icmp;
294 idev->stats.icmpv6msgdev = kzalloc(sizeof(struct icmpv6msg_mib_device),
295 GFP_KERNEL);
296 if (!idev->stats.icmpv6msgdev)
297 goto err_icmpmsg;
298
299 return 0;
300
301 err_icmpmsg:
302 kfree(idev->stats.icmpv6dev);
303 err_icmp:
304 snmp_mib_free((void __percpu **)idev->stats.ipv6);
305 err_ip:
306 return -ENOMEM;
307 }
308
309 static void snmp6_free_dev(struct inet6_dev *idev)
310 {
311 kfree(idev->stats.icmpv6msgdev);
312 kfree(idev->stats.icmpv6dev);
313 snmp_mib_free((void __percpu **)idev->stats.ipv6);
314 }
315
316 /* Nobody refers to this device, we may destroy it. */
317
318 void in6_dev_finish_destroy(struct inet6_dev *idev)
319 {
320 struct net_device *dev = idev->dev;
321
322 WARN_ON(!list_empty(&idev->addr_list));
323 WARN_ON(idev->mc_list != NULL);
324 WARN_ON(timer_pending(&idev->rs_timer));
325
326 #ifdef NET_REFCNT_DEBUG
327 pr_debug("%s: %s\n", __func__, dev ? dev->name : "NIL");
328 #endif
329 dev_put(dev);
330 if (!idev->dead) {
331 pr_warn("Freeing alive inet6 device %p\n", idev);
332 return;
333 }
334 snmp6_free_dev(idev);
335 kfree_rcu(idev, rcu);
336 }
337 EXPORT_SYMBOL(in6_dev_finish_destroy);
338
339 static struct inet6_dev *ipv6_add_dev(struct net_device *dev)
340 {
341 struct inet6_dev *ndev;
342
343 ASSERT_RTNL();
344
345 if (dev->mtu < IPV6_MIN_MTU)
346 return NULL;
347
348 ndev = kzalloc(sizeof(struct inet6_dev), GFP_KERNEL);
349
350 if (ndev == NULL)
351 return NULL;
352
353 rwlock_init(&ndev->lock);
354 ndev->dev = dev;
355 INIT_LIST_HEAD(&ndev->addr_list);
356 setup_timer(&ndev->rs_timer, addrconf_rs_timer,
357 (unsigned long)ndev);
358 memcpy(&ndev->cnf, dev_net(dev)->ipv6.devconf_dflt, sizeof(ndev->cnf));
359 ndev->cnf.mtu6 = dev->mtu;
360 ndev->cnf.sysctl = NULL;
361 ndev->nd_parms = neigh_parms_alloc(dev, &nd_tbl);
362 if (ndev->nd_parms == NULL) {
363 kfree(ndev);
364 return NULL;
365 }
366 if (ndev->cnf.forwarding)
367 dev_disable_lro(dev);
368 /* We refer to the device */
369 dev_hold(dev);
370
371 if (snmp6_alloc_dev(ndev) < 0) {
372 ADBG((KERN_WARNING
373 "%s: cannot allocate memory for statistics; dev=%s.\n",
374 __func__, dev->name));
375 neigh_parms_release(&nd_tbl, ndev->nd_parms);
376 dev_put(dev);
377 kfree(ndev);
378 return NULL;
379 }
380
381 if (snmp6_register_dev(ndev) < 0) {
382 ADBG((KERN_WARNING
383 "%s: cannot create /proc/net/dev_snmp6/%s\n",
384 __func__, dev->name));
385 neigh_parms_release(&nd_tbl, ndev->nd_parms);
386 ndev->dead = 1;
387 in6_dev_finish_destroy(ndev);
388 return NULL;
389 }
390
391 /* One reference from device. We must do this before
392 * we invoke __ipv6_regen_rndid().
393 */
394 in6_dev_hold(ndev);
395
396 if (dev->flags & (IFF_NOARP | IFF_LOOPBACK))
397 ndev->cnf.accept_dad = -1;
398
399 #if IS_ENABLED(CONFIG_IPV6_SIT)
400 if (dev->type == ARPHRD_SIT && (dev->priv_flags & IFF_ISATAP)) {
401 pr_info("%s: Disabled Multicast RS\n", dev->name);
402 ndev->cnf.rtr_solicits = 0;
403 }
404 #endif
405
406 #ifdef CONFIG_IPV6_PRIVACY
407 INIT_LIST_HEAD(&ndev->tempaddr_list);
408 setup_timer(&ndev->regen_timer, ipv6_regen_rndid, (unsigned long)ndev);
409 if ((dev->flags&IFF_LOOPBACK) ||
410 dev->type == ARPHRD_TUNNEL ||
411 dev->type == ARPHRD_TUNNEL6 ||
412 dev->type == ARPHRD_SIT ||
413 dev->type == ARPHRD_NONE) {
414 ndev->cnf.use_tempaddr = -1;
415 } else {
416 in6_dev_hold(ndev);
417 ipv6_regen_rndid((unsigned long) ndev);
418 }
419 #endif
420 ndev->token = in6addr_any;
421
422 if (netif_running(dev) && addrconf_qdisc_ok(dev))
423 ndev->if_flags |= IF_READY;
424
425 ipv6_mc_init_dev(ndev);
426 ndev->tstamp = jiffies;
427 addrconf_sysctl_register(ndev);
428 /* protected by rtnl_lock */
429 rcu_assign_pointer(dev->ip6_ptr, ndev);
430
431 /* Join interface-local all-node multicast group */
432 ipv6_dev_mc_inc(dev, &in6addr_interfacelocal_allnodes);
433
434 /* Join all-node multicast group */
435 ipv6_dev_mc_inc(dev, &in6addr_linklocal_allnodes);
436
437 /* Join all-router multicast group if forwarding is set */
438 if (ndev->cnf.forwarding && (dev->flags & IFF_MULTICAST))
439 ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
440
441 return ndev;
442 }
443
444 static struct inet6_dev *ipv6_find_idev(struct net_device *dev)
445 {
446 struct inet6_dev *idev;
447
448 ASSERT_RTNL();
449
450 idev = __in6_dev_get(dev);
451 if (!idev) {
452 idev = ipv6_add_dev(dev);
453 if (!idev)
454 return NULL;
455 }
456
457 if (dev->flags&IFF_UP)
458 ipv6_mc_up(idev);
459 return idev;
460 }
461
462 static int inet6_netconf_msgsize_devconf(int type)
463 {
464 int size = NLMSG_ALIGN(sizeof(struct netconfmsg))
465 + nla_total_size(4); /* NETCONFA_IFINDEX */
466
467 /* type -1 is used for ALL */
468 if (type == -1 || type == NETCONFA_FORWARDING)
469 size += nla_total_size(4);
470 #ifdef CONFIG_IPV6_MROUTE
471 if (type == -1 || type == NETCONFA_MC_FORWARDING)
472 size += nla_total_size(4);
473 #endif
474
475 return size;
476 }
477
478 static int inet6_netconf_fill_devconf(struct sk_buff *skb, int ifindex,
479 struct ipv6_devconf *devconf, u32 portid,
480 u32 seq, int event, unsigned int flags,
481 int type)
482 {
483 struct nlmsghdr *nlh;
484 struct netconfmsg *ncm;
485
486 nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct netconfmsg),
487 flags);
488 if (nlh == NULL)
489 return -EMSGSIZE;
490
491 ncm = nlmsg_data(nlh);
492 ncm->ncm_family = AF_INET6;
493
494 if (nla_put_s32(skb, NETCONFA_IFINDEX, ifindex) < 0)
495 goto nla_put_failure;
496
497 /* type -1 is used for ALL */
498 if ((type == -1 || type == NETCONFA_FORWARDING) &&
499 nla_put_s32(skb, NETCONFA_FORWARDING, devconf->forwarding) < 0)
500 goto nla_put_failure;
501 #ifdef CONFIG_IPV6_MROUTE
502 if ((type == -1 || type == NETCONFA_MC_FORWARDING) &&
503 nla_put_s32(skb, NETCONFA_MC_FORWARDING,
504 devconf->mc_forwarding) < 0)
505 goto nla_put_failure;
506 #endif
507 return nlmsg_end(skb, nlh);
508
509 nla_put_failure:
510 nlmsg_cancel(skb, nlh);
511 return -EMSGSIZE;
512 }
513
514 void inet6_netconf_notify_devconf(struct net *net, int type, int ifindex,
515 struct ipv6_devconf *devconf)
516 {
517 struct sk_buff *skb;
518 int err = -ENOBUFS;
519
520 skb = nlmsg_new(inet6_netconf_msgsize_devconf(type), GFP_ATOMIC);
521 if (skb == NULL)
522 goto errout;
523
524 err = inet6_netconf_fill_devconf(skb, ifindex, devconf, 0, 0,
525 RTM_NEWNETCONF, 0, type);
526 if (err < 0) {
527 /* -EMSGSIZE implies BUG in inet6_netconf_msgsize_devconf() */
528 WARN_ON(err == -EMSGSIZE);
529 kfree_skb(skb);
530 goto errout;
531 }
532 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_NETCONF, NULL, GFP_ATOMIC);
533 return;
534 errout:
535 rtnl_set_sk_err(net, RTNLGRP_IPV6_NETCONF, err);
536 }
537
538 static const struct nla_policy devconf_ipv6_policy[NETCONFA_MAX+1] = {
539 [NETCONFA_IFINDEX] = { .len = sizeof(int) },
540 [NETCONFA_FORWARDING] = { .len = sizeof(int) },
541 };
542
543 static int inet6_netconf_get_devconf(struct sk_buff *in_skb,
544 struct nlmsghdr *nlh)
545 {
546 struct net *net = sock_net(in_skb->sk);
547 struct nlattr *tb[NETCONFA_MAX+1];
548 struct netconfmsg *ncm;
549 struct sk_buff *skb;
550 struct ipv6_devconf *devconf;
551 struct inet6_dev *in6_dev;
552 struct net_device *dev;
553 int ifindex;
554 int err;
555
556 err = nlmsg_parse(nlh, sizeof(*ncm), tb, NETCONFA_MAX,
557 devconf_ipv6_policy);
558 if (err < 0)
559 goto errout;
560
561 err = EINVAL;
562 if (!tb[NETCONFA_IFINDEX])
563 goto errout;
564
565 ifindex = nla_get_s32(tb[NETCONFA_IFINDEX]);
566 switch (ifindex) {
567 case NETCONFA_IFINDEX_ALL:
568 devconf = net->ipv6.devconf_all;
569 break;
570 case NETCONFA_IFINDEX_DEFAULT:
571 devconf = net->ipv6.devconf_dflt;
572 break;
573 default:
574 dev = __dev_get_by_index(net, ifindex);
575 if (dev == NULL)
576 goto errout;
577 in6_dev = __in6_dev_get(dev);
578 if (in6_dev == NULL)
579 goto errout;
580 devconf = &in6_dev->cnf;
581 break;
582 }
583
584 err = -ENOBUFS;
585 skb = nlmsg_new(inet6_netconf_msgsize_devconf(-1), GFP_ATOMIC);
586 if (skb == NULL)
587 goto errout;
588
589 err = inet6_netconf_fill_devconf(skb, ifindex, devconf,
590 NETLINK_CB(in_skb).portid,
591 nlh->nlmsg_seq, RTM_NEWNETCONF, 0,
592 -1);
593 if (err < 0) {
594 /* -EMSGSIZE implies BUG in inet6_netconf_msgsize_devconf() */
595 WARN_ON(err == -EMSGSIZE);
596 kfree_skb(skb);
597 goto errout;
598 }
599 err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
600 errout:
601 return err;
602 }
603
604 static int inet6_netconf_dump_devconf(struct sk_buff *skb,
605 struct netlink_callback *cb)
606 {
607 struct net *net = sock_net(skb->sk);
608 int h, s_h;
609 int idx, s_idx;
610 struct net_device *dev;
611 struct inet6_dev *idev;
612 struct hlist_head *head;
613
614 s_h = cb->args[0];
615 s_idx = idx = cb->args[1];
616
617 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
618 idx = 0;
619 head = &net->dev_index_head[h];
620 rcu_read_lock();
621 cb->seq = atomic_read(&net->ipv6.dev_addr_genid) ^
622 net->dev_base_seq;
623 hlist_for_each_entry_rcu(dev, head, index_hlist) {
624 if (idx < s_idx)
625 goto cont;
626 idev = __in6_dev_get(dev);
627 if (!idev)
628 goto cont;
629
630 if (inet6_netconf_fill_devconf(skb, dev->ifindex,
631 &idev->cnf,
632 NETLINK_CB(cb->skb).portid,
633 cb->nlh->nlmsg_seq,
634 RTM_NEWNETCONF,
635 NLM_F_MULTI,
636 -1) <= 0) {
637 rcu_read_unlock();
638 goto done;
639 }
640 nl_dump_check_consistent(cb, nlmsg_hdr(skb));
641 cont:
642 idx++;
643 }
644 rcu_read_unlock();
645 }
646 if (h == NETDEV_HASHENTRIES) {
647 if (inet6_netconf_fill_devconf(skb, NETCONFA_IFINDEX_ALL,
648 net->ipv6.devconf_all,
649 NETLINK_CB(cb->skb).portid,
650 cb->nlh->nlmsg_seq,
651 RTM_NEWNETCONF, NLM_F_MULTI,
652 -1) <= 0)
653 goto done;
654 else
655 h++;
656 }
657 if (h == NETDEV_HASHENTRIES + 1) {
658 if (inet6_netconf_fill_devconf(skb, NETCONFA_IFINDEX_DEFAULT,
659 net->ipv6.devconf_dflt,
660 NETLINK_CB(cb->skb).portid,
661 cb->nlh->nlmsg_seq,
662 RTM_NEWNETCONF, NLM_F_MULTI,
663 -1) <= 0)
664 goto done;
665 else
666 h++;
667 }
668 done:
669 cb->args[0] = h;
670 cb->args[1] = idx;
671
672 return skb->len;
673 }
674
675 #ifdef CONFIG_SYSCTL
676 static void dev_forward_change(struct inet6_dev *idev)
677 {
678 struct net_device *dev;
679 struct inet6_ifaddr *ifa;
680
681 if (!idev)
682 return;
683 dev = idev->dev;
684 if (idev->cnf.forwarding)
685 dev_disable_lro(dev);
686 if (dev->flags & IFF_MULTICAST) {
687 if (idev->cnf.forwarding) {
688 ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
689 ipv6_dev_mc_inc(dev, &in6addr_interfacelocal_allrouters);
690 ipv6_dev_mc_inc(dev, &in6addr_sitelocal_allrouters);
691 } else {
692 ipv6_dev_mc_dec(dev, &in6addr_linklocal_allrouters);
693 ipv6_dev_mc_dec(dev, &in6addr_interfacelocal_allrouters);
694 ipv6_dev_mc_dec(dev, &in6addr_sitelocal_allrouters);
695 }
696 }
697
698 list_for_each_entry(ifa, &idev->addr_list, if_list) {
699 if (ifa->flags&IFA_F_TENTATIVE)
700 continue;
701 if (idev->cnf.forwarding)
702 addrconf_join_anycast(ifa);
703 else
704 addrconf_leave_anycast(ifa);
705 }
706 inet6_netconf_notify_devconf(dev_net(dev), NETCONFA_FORWARDING,
707 dev->ifindex, &idev->cnf);
708 }
709
710
711 static void addrconf_forward_change(struct net *net, __s32 newf)
712 {
713 struct net_device *dev;
714 struct inet6_dev *idev;
715
716 for_each_netdev(net, dev) {
717 idev = __in6_dev_get(dev);
718 if (idev) {
719 int changed = (!idev->cnf.forwarding) ^ (!newf);
720 idev->cnf.forwarding = newf;
721 if (changed)
722 dev_forward_change(idev);
723 }
724 }
725 }
726
727 static int addrconf_fixup_forwarding(struct ctl_table *table, int *p, int newf)
728 {
729 struct net *net;
730 int old;
731
732 if (!rtnl_trylock())
733 return restart_syscall();
734
735 net = (struct net *)table->extra2;
736 old = *p;
737 *p = newf;
738
739 if (p == &net->ipv6.devconf_dflt->forwarding) {
740 if ((!newf) ^ (!old))
741 inet6_netconf_notify_devconf(net, NETCONFA_FORWARDING,
742 NETCONFA_IFINDEX_DEFAULT,
743 net->ipv6.devconf_dflt);
744 rtnl_unlock();
745 return 0;
746 }
747
748 if (p == &net->ipv6.devconf_all->forwarding) {
749 net->ipv6.devconf_dflt->forwarding = newf;
750 addrconf_forward_change(net, newf);
751 if ((!newf) ^ (!old))
752 inet6_netconf_notify_devconf(net, NETCONFA_FORWARDING,
753 NETCONFA_IFINDEX_ALL,
754 net->ipv6.devconf_all);
755 } else if ((!newf) ^ (!old))
756 dev_forward_change((struct inet6_dev *)table->extra1);
757 rtnl_unlock();
758
759 if (newf)
760 rt6_purge_dflt_routers(net);
761 return 1;
762 }
763 #endif
764
765 /* Nobody refers to this ifaddr, destroy it */
766 void inet6_ifa_finish_destroy(struct inet6_ifaddr *ifp)
767 {
768 WARN_ON(!hlist_unhashed(&ifp->addr_lst));
769
770 #ifdef NET_REFCNT_DEBUG
771 pr_debug("%s\n", __func__);
772 #endif
773
774 in6_dev_put(ifp->idev);
775
776 if (del_timer(&ifp->dad_timer))
777 pr_notice("Timer is still running, when freeing ifa=%p\n", ifp);
778
779 if (ifp->state != INET6_IFADDR_STATE_DEAD) {
780 pr_warn("Freeing alive inet6 address %p\n", ifp);
781 return;
782 }
783 ip6_rt_put(ifp->rt);
784
785 kfree_rcu(ifp, rcu);
786 }
787
788 static void
789 ipv6_link_dev_addr(struct inet6_dev *idev, struct inet6_ifaddr *ifp)
790 {
791 struct list_head *p;
792 int ifp_scope = ipv6_addr_src_scope(&ifp->addr);
793
794 /*
795 * Each device address list is sorted in order of scope -
796 * global before linklocal.
797 */
798 list_for_each(p, &idev->addr_list) {
799 struct inet6_ifaddr *ifa
800 = list_entry(p, struct inet6_ifaddr, if_list);
801 if (ifp_scope >= ipv6_addr_src_scope(&ifa->addr))
802 break;
803 }
804
805 list_add_tail(&ifp->if_list, p);
806 }
807
808 static u32 inet6_addr_hash(const struct in6_addr *addr)
809 {
810 return hash_32(ipv6_addr_hash(addr), IN6_ADDR_HSIZE_SHIFT);
811 }
812
813 /* On success it returns ifp with increased reference count */
814
815 static struct inet6_ifaddr *
816 ipv6_add_addr(struct inet6_dev *idev, const struct in6_addr *addr, int pfxlen,
817 int scope, u32 flags)
818 {
819 struct inet6_ifaddr *ifa = NULL;
820 struct rt6_info *rt;
821 unsigned int hash;
822 int err = 0;
823 int addr_type = ipv6_addr_type(addr);
824
825 if (addr_type == IPV6_ADDR_ANY ||
826 addr_type & IPV6_ADDR_MULTICAST ||
827 (!(idev->dev->flags & IFF_LOOPBACK) &&
828 addr_type & IPV6_ADDR_LOOPBACK))
829 return ERR_PTR(-EADDRNOTAVAIL);
830
831 rcu_read_lock_bh();
832 if (idev->dead) {
833 err = -ENODEV; /*XXX*/
834 goto out2;
835 }
836
837 if (idev->cnf.disable_ipv6) {
838 err = -EACCES;
839 goto out2;
840 }
841
842 spin_lock(&addrconf_hash_lock);
843
844 /* Ignore adding duplicate addresses on an interface */
845 if (ipv6_chk_same_addr(dev_net(idev->dev), addr, idev->dev)) {
846 ADBG(("ipv6_add_addr: already assigned\n"));
847 err = -EEXIST;
848 goto out;
849 }
850
851 ifa = kzalloc(sizeof(struct inet6_ifaddr), GFP_ATOMIC);
852
853 if (ifa == NULL) {
854 ADBG(("ipv6_add_addr: malloc failed\n"));
855 err = -ENOBUFS;
856 goto out;
857 }
858
859 rt = addrconf_dst_alloc(idev, addr, false);
860 if (IS_ERR(rt)) {
861 err = PTR_ERR(rt);
862 goto out;
863 }
864
865 ifa->addr = *addr;
866
867 spin_lock_init(&ifa->lock);
868 spin_lock_init(&ifa->state_lock);
869 setup_timer(&ifa->dad_timer, addrconf_dad_timer,
870 (unsigned long)ifa);
871 INIT_HLIST_NODE(&ifa->addr_lst);
872 ifa->scope = scope;
873 ifa->prefix_len = pfxlen;
874 ifa->flags = flags | IFA_F_TENTATIVE;
875 ifa->cstamp = ifa->tstamp = jiffies;
876 ifa->tokenized = false;
877
878 ifa->rt = rt;
879
880 ifa->idev = idev;
881 in6_dev_hold(idev);
882 /* For caller */
883 in6_ifa_hold(ifa);
884
885 /* Add to big hash table */
886 hash = inet6_addr_hash(addr);
887
888 hlist_add_head_rcu(&ifa->addr_lst, &inet6_addr_lst[hash]);
889 spin_unlock(&addrconf_hash_lock);
890
891 write_lock(&idev->lock);
892 /* Add to inet6_dev unicast addr list. */
893 ipv6_link_dev_addr(idev, ifa);
894
895 #ifdef CONFIG_IPV6_PRIVACY
896 if (ifa->flags&IFA_F_TEMPORARY) {
897 list_add(&ifa->tmp_list, &idev->tempaddr_list);
898 in6_ifa_hold(ifa);
899 }
900 #endif
901
902 in6_ifa_hold(ifa);
903 write_unlock(&idev->lock);
904 out2:
905 rcu_read_unlock_bh();
906
907 if (likely(err == 0))
908 inet6addr_notifier_call_chain(NETDEV_UP, ifa);
909 else {
910 kfree(ifa);
911 ifa = ERR_PTR(err);
912 }
913
914 return ifa;
915 out:
916 spin_unlock(&addrconf_hash_lock);
917 goto out2;
918 }
919
920 /* This function wants to get referenced ifp and releases it before return */
921
922 static void ipv6_del_addr(struct inet6_ifaddr *ifp)
923 {
924 struct inet6_ifaddr *ifa, *ifn;
925 struct inet6_dev *idev = ifp->idev;
926 int state;
927 int deleted = 0, onlink = 0;
928 unsigned long expires = jiffies;
929
930 spin_lock_bh(&ifp->state_lock);
931 state = ifp->state;
932 ifp->state = INET6_IFADDR_STATE_DEAD;
933 spin_unlock_bh(&ifp->state_lock);
934
935 if (state == INET6_IFADDR_STATE_DEAD)
936 goto out;
937
938 spin_lock_bh(&addrconf_hash_lock);
939 hlist_del_init_rcu(&ifp->addr_lst);
940 spin_unlock_bh(&addrconf_hash_lock);
941
942 write_lock_bh(&idev->lock);
943 #ifdef CONFIG_IPV6_PRIVACY
944 if (ifp->flags&IFA_F_TEMPORARY) {
945 list_del(&ifp->tmp_list);
946 if (ifp->ifpub) {
947 in6_ifa_put(ifp->ifpub);
948 ifp->ifpub = NULL;
949 }
950 __in6_ifa_put(ifp);
951 }
952 #endif
953
954 list_for_each_entry_safe(ifa, ifn, &idev->addr_list, if_list) {
955 if (ifa == ifp) {
956 list_del_init(&ifp->if_list);
957 __in6_ifa_put(ifp);
958
959 if (!(ifp->flags & IFA_F_PERMANENT) || onlink > 0)
960 break;
961 deleted = 1;
962 continue;
963 } else if (ifp->flags & IFA_F_PERMANENT) {
964 if (ipv6_prefix_equal(&ifa->addr, &ifp->addr,
965 ifp->prefix_len)) {
966 if (ifa->flags & IFA_F_PERMANENT) {
967 onlink = 1;
968 if (deleted)
969 break;
970 } else {
971 unsigned long lifetime;
972
973 if (!onlink)
974 onlink = -1;
975
976 spin_lock(&ifa->lock);
977
978 lifetime = addrconf_timeout_fixup(ifa->valid_lft, HZ);
979 /*
980 * Note: Because this address is
981 * not permanent, lifetime <
982 * LONG_MAX / HZ here.
983 */
984 if (time_before(expires,
985 ifa->tstamp + lifetime * HZ))
986 expires = ifa->tstamp + lifetime * HZ;
987 spin_unlock(&ifa->lock);
988 }
989 }
990 }
991 }
992 write_unlock_bh(&idev->lock);
993
994 addrconf_del_dad_timer(ifp);
995
996 ipv6_ifa_notify(RTM_DELADDR, ifp);
997
998 inet6addr_notifier_call_chain(NETDEV_DOWN, ifp);
999
1000 /*
1001 * Purge or update corresponding prefix
1002 *
1003 * 1) we don't purge prefix here if address was not permanent.
1004 * prefix is managed by its own lifetime.
1005 * 2) if there're no addresses, delete prefix.
1006 * 3) if there're still other permanent address(es),
1007 * corresponding prefix is still permanent.
1008 * 4) otherwise, update prefix lifetime to the
1009 * longest valid lifetime among the corresponding
1010 * addresses on the device.
1011 * Note: subsequent RA will update lifetime.
1012 *
1013 * --yoshfuji
1014 */
1015 if ((ifp->flags & IFA_F_PERMANENT) && onlink < 1) {
1016 struct in6_addr prefix;
1017 struct rt6_info *rt;
1018
1019 ipv6_addr_prefix(&prefix, &ifp->addr, ifp->prefix_len);
1020
1021 rt = addrconf_get_prefix_route(&prefix,
1022 ifp->prefix_len,
1023 ifp->idev->dev,
1024 0, RTF_GATEWAY | RTF_DEFAULT);
1025
1026 if (rt) {
1027 if (onlink == 0) {
1028 ip6_del_rt(rt);
1029 rt = NULL;
1030 } else if (!(rt->rt6i_flags & RTF_EXPIRES)) {
1031 rt6_set_expires(rt, expires);
1032 }
1033 }
1034 ip6_rt_put(rt);
1035 }
1036
1037 /* clean up prefsrc entries */
1038 rt6_remove_prefsrc(ifp);
1039 out:
1040 in6_ifa_put(ifp);
1041 }
1042
1043 #ifdef CONFIG_IPV6_PRIVACY
1044 static int ipv6_create_tempaddr(struct inet6_ifaddr *ifp, struct inet6_ifaddr *ift)
1045 {
1046 struct inet6_dev *idev = ifp->idev;
1047 struct in6_addr addr, *tmpaddr;
1048 unsigned long tmp_prefered_lft, tmp_valid_lft, tmp_tstamp, age;
1049 unsigned long regen_advance;
1050 int tmp_plen;
1051 int ret = 0;
1052 int max_addresses;
1053 u32 addr_flags;
1054 unsigned long now = jiffies;
1055
1056 write_lock(&idev->lock);
1057 if (ift) {
1058 spin_lock_bh(&ift->lock);
1059 memcpy(&addr.s6_addr[8], &ift->addr.s6_addr[8], 8);
1060 spin_unlock_bh(&ift->lock);
1061 tmpaddr = &addr;
1062 } else {
1063 tmpaddr = NULL;
1064 }
1065 retry:
1066 in6_dev_hold(idev);
1067 if (idev->cnf.use_tempaddr <= 0) {
1068 write_unlock(&idev->lock);
1069 pr_info("%s: use_tempaddr is disabled\n", __func__);
1070 in6_dev_put(idev);
1071 ret = -1;
1072 goto out;
1073 }
1074 spin_lock_bh(&ifp->lock);
1075 if (ifp->regen_count++ >= idev->cnf.regen_max_retry) {
1076 idev->cnf.use_tempaddr = -1; /*XXX*/
1077 spin_unlock_bh(&ifp->lock);
1078 write_unlock(&idev->lock);
1079 pr_warn("%s: regeneration time exceeded - disabled temporary address support\n",
1080 __func__);
1081 in6_dev_put(idev);
1082 ret = -1;
1083 goto out;
1084 }
1085 in6_ifa_hold(ifp);
1086 memcpy(addr.s6_addr, ifp->addr.s6_addr, 8);
1087 __ipv6_try_regen_rndid(idev, tmpaddr);
1088 memcpy(&addr.s6_addr[8], idev->rndid, 8);
1089 age = (now - ifp->tstamp) / HZ;
1090 tmp_valid_lft = min_t(__u32,
1091 ifp->valid_lft,
1092 idev->cnf.temp_valid_lft + age);
1093 tmp_prefered_lft = min_t(__u32,
1094 ifp->prefered_lft,
1095 idev->cnf.temp_prefered_lft + age -
1096 idev->cnf.max_desync_factor);
1097 tmp_plen = ifp->prefix_len;
1098 max_addresses = idev->cnf.max_addresses;
1099 tmp_tstamp = ifp->tstamp;
1100 spin_unlock_bh(&ifp->lock);
1101
1102 regen_advance = idev->cnf.regen_max_retry *
1103 idev->cnf.dad_transmits *
1104 idev->nd_parms->retrans_time / HZ;
1105 write_unlock(&idev->lock);
1106
1107 /* A temporary address is created only if this calculated Preferred
1108 * Lifetime is greater than REGEN_ADVANCE time units. In particular,
1109 * an implementation must not create a temporary address with a zero
1110 * Preferred Lifetime.
1111 */
1112 if (tmp_prefered_lft <= regen_advance) {
1113 in6_ifa_put(ifp);
1114 in6_dev_put(idev);
1115 ret = -1;
1116 goto out;
1117 }
1118
1119 addr_flags = IFA_F_TEMPORARY;
1120 /* set in addrconf_prefix_rcv() */
1121 if (ifp->flags & IFA_F_OPTIMISTIC)
1122 addr_flags |= IFA_F_OPTIMISTIC;
1123
1124 ift = !max_addresses ||
1125 ipv6_count_addresses(idev) < max_addresses ?
1126 ipv6_add_addr(idev, &addr, tmp_plen, ipv6_addr_scope(&addr),
1127 addr_flags) : NULL;
1128 if (IS_ERR_OR_NULL(ift)) {
1129 in6_ifa_put(ifp);
1130 in6_dev_put(idev);
1131 pr_info("%s: retry temporary address regeneration\n", __func__);
1132 tmpaddr = &addr;
1133 write_lock(&idev->lock);
1134 goto retry;
1135 }
1136
1137 spin_lock_bh(&ift->lock);
1138 ift->ifpub = ifp;
1139 ift->valid_lft = tmp_valid_lft;
1140 ift->prefered_lft = tmp_prefered_lft;
1141 ift->cstamp = now;
1142 ift->tstamp = tmp_tstamp;
1143 spin_unlock_bh(&ift->lock);
1144
1145 addrconf_dad_start(ift);
1146 in6_ifa_put(ift);
1147 in6_dev_put(idev);
1148 out:
1149 return ret;
1150 }
1151 #endif
1152
1153 /*
1154 * Choose an appropriate source address (RFC3484)
1155 */
1156 enum {
1157 IPV6_SADDR_RULE_INIT = 0,
1158 IPV6_SADDR_RULE_LOCAL,
1159 IPV6_SADDR_RULE_SCOPE,
1160 IPV6_SADDR_RULE_PREFERRED,
1161 #ifdef CONFIG_IPV6_MIP6
1162 IPV6_SADDR_RULE_HOA,
1163 #endif
1164 IPV6_SADDR_RULE_OIF,
1165 IPV6_SADDR_RULE_LABEL,
1166 #ifdef CONFIG_IPV6_PRIVACY
1167 IPV6_SADDR_RULE_PRIVACY,
1168 #endif
1169 IPV6_SADDR_RULE_ORCHID,
1170 IPV6_SADDR_RULE_PREFIX,
1171 IPV6_SADDR_RULE_MAX
1172 };
1173
1174 struct ipv6_saddr_score {
1175 int rule;
1176 int addr_type;
1177 struct inet6_ifaddr *ifa;
1178 DECLARE_BITMAP(scorebits, IPV6_SADDR_RULE_MAX);
1179 int scopedist;
1180 int matchlen;
1181 };
1182
1183 struct ipv6_saddr_dst {
1184 const struct in6_addr *addr;
1185 int ifindex;
1186 int scope;
1187 int label;
1188 unsigned int prefs;
1189 };
1190
1191 static inline int ipv6_saddr_preferred(int type)
1192 {
1193 if (type & (IPV6_ADDR_MAPPED|IPV6_ADDR_COMPATv4|IPV6_ADDR_LOOPBACK))
1194 return 1;
1195 return 0;
1196 }
1197
1198 static int ipv6_get_saddr_eval(struct net *net,
1199 struct ipv6_saddr_score *score,
1200 struct ipv6_saddr_dst *dst,
1201 int i)
1202 {
1203 int ret;
1204
1205 if (i <= score->rule) {
1206 switch (i) {
1207 case IPV6_SADDR_RULE_SCOPE:
1208 ret = score->scopedist;
1209 break;
1210 case IPV6_SADDR_RULE_PREFIX:
1211 ret = score->matchlen;
1212 break;
1213 default:
1214 ret = !!test_bit(i, score->scorebits);
1215 }
1216 goto out;
1217 }
1218
1219 switch (i) {
1220 case IPV6_SADDR_RULE_INIT:
1221 /* Rule 0: remember if hiscore is not ready yet */
1222 ret = !!score->ifa;
1223 break;
1224 case IPV6_SADDR_RULE_LOCAL:
1225 /* Rule 1: Prefer same address */
1226 ret = ipv6_addr_equal(&score->ifa->addr, dst->addr);
1227 break;
1228 case IPV6_SADDR_RULE_SCOPE:
1229 /* Rule 2: Prefer appropriate scope
1230 *
1231 * ret
1232 * ^
1233 * -1 | d 15
1234 * ---+--+-+---> scope
1235 * |
1236 * | d is scope of the destination.
1237 * B-d | \
1238 * | \ <- smaller scope is better if
1239 * B-15 | \ if scope is enough for destinaion.
1240 * | ret = B - scope (-1 <= scope >= d <= 15).
1241 * d-C-1 | /
1242 * |/ <- greater is better
1243 * -C / if scope is not enough for destination.
1244 * /| ret = scope - C (-1 <= d < scope <= 15).
1245 *
1246 * d - C - 1 < B -15 (for all -1 <= d <= 15).
1247 * C > d + 14 - B >= 15 + 14 - B = 29 - B.
1248 * Assume B = 0 and we get C > 29.
1249 */
1250 ret = __ipv6_addr_src_scope(score->addr_type);
1251 if (ret >= dst->scope)
1252 ret = -ret;
1253 else
1254 ret -= 128; /* 30 is enough */
1255 score->scopedist = ret;
1256 break;
1257 case IPV6_SADDR_RULE_PREFERRED:
1258 /* Rule 3: Avoid deprecated and optimistic addresses */
1259 ret = ipv6_saddr_preferred(score->addr_type) ||
1260 !(score->ifa->flags & (IFA_F_DEPRECATED|IFA_F_OPTIMISTIC));
1261 break;
1262 #ifdef CONFIG_IPV6_MIP6
1263 case IPV6_SADDR_RULE_HOA:
1264 {
1265 /* Rule 4: Prefer home address */
1266 int prefhome = !(dst->prefs & IPV6_PREFER_SRC_COA);
1267 ret = !(score->ifa->flags & IFA_F_HOMEADDRESS) ^ prefhome;
1268 break;
1269 }
1270 #endif
1271 case IPV6_SADDR_RULE_OIF:
1272 /* Rule 5: Prefer outgoing interface */
1273 ret = (!dst->ifindex ||
1274 dst->ifindex == score->ifa->idev->dev->ifindex);
1275 break;
1276 case IPV6_SADDR_RULE_LABEL:
1277 /* Rule 6: Prefer matching label */
1278 ret = ipv6_addr_label(net,
1279 &score->ifa->addr, score->addr_type,
1280 score->ifa->idev->dev->ifindex) == dst->label;
1281 break;
1282 #ifdef CONFIG_IPV6_PRIVACY
1283 case IPV6_SADDR_RULE_PRIVACY:
1284 {
1285 /* Rule 7: Prefer public address
1286 * Note: prefer temporary address if use_tempaddr >= 2
1287 */
1288 int preftmp = dst->prefs & (IPV6_PREFER_SRC_PUBLIC|IPV6_PREFER_SRC_TMP) ?
1289 !!(dst->prefs & IPV6_PREFER_SRC_TMP) :
1290 score->ifa->idev->cnf.use_tempaddr >= 2;
1291 ret = (!(score->ifa->flags & IFA_F_TEMPORARY)) ^ preftmp;
1292 break;
1293 }
1294 #endif
1295 case IPV6_SADDR_RULE_ORCHID:
1296 /* Rule 8-: Prefer ORCHID vs ORCHID or
1297 * non-ORCHID vs non-ORCHID
1298 */
1299 ret = !(ipv6_addr_orchid(&score->ifa->addr) ^
1300 ipv6_addr_orchid(dst->addr));
1301 break;
1302 case IPV6_SADDR_RULE_PREFIX:
1303 /* Rule 8: Use longest matching prefix */
1304 ret = ipv6_addr_diff(&score->ifa->addr, dst->addr);
1305 if (ret > score->ifa->prefix_len)
1306 ret = score->ifa->prefix_len;
1307 score->matchlen = ret;
1308 break;
1309 default:
1310 ret = 0;
1311 }
1312
1313 if (ret)
1314 __set_bit(i, score->scorebits);
1315 score->rule = i;
1316 out:
1317 return ret;
1318 }
1319
1320 int ipv6_dev_get_saddr(struct net *net, const struct net_device *dst_dev,
1321 const struct in6_addr *daddr, unsigned int prefs,
1322 struct in6_addr *saddr)
1323 {
1324 struct ipv6_saddr_score scores[2],
1325 *score = &scores[0], *hiscore = &scores[1];
1326 struct ipv6_saddr_dst dst;
1327 struct net_device *dev;
1328 int dst_type;
1329
1330 dst_type = __ipv6_addr_type(daddr);
1331 dst.addr = daddr;
1332 dst.ifindex = dst_dev ? dst_dev->ifindex : 0;
1333 dst.scope = __ipv6_addr_src_scope(dst_type);
1334 dst.label = ipv6_addr_label(net, daddr, dst_type, dst.ifindex);
1335 dst.prefs = prefs;
1336
1337 hiscore->rule = -1;
1338 hiscore->ifa = NULL;
1339
1340 rcu_read_lock();
1341
1342 for_each_netdev_rcu(net, dev) {
1343 struct inet6_dev *idev;
1344
1345 /* Candidate Source Address (section 4)
1346 * - multicast and link-local destination address,
1347 * the set of candidate source address MUST only
1348 * include addresses assigned to interfaces
1349 * belonging to the same link as the outgoing
1350 * interface.
1351 * (- For site-local destination addresses, the
1352 * set of candidate source addresses MUST only
1353 * include addresses assigned to interfaces
1354 * belonging to the same site as the outgoing
1355 * interface.)
1356 */
1357 if (((dst_type & IPV6_ADDR_MULTICAST) ||
1358 dst.scope <= IPV6_ADDR_SCOPE_LINKLOCAL) &&
1359 dst.ifindex && dev->ifindex != dst.ifindex)
1360 continue;
1361
1362 idev = __in6_dev_get(dev);
1363 if (!idev)
1364 continue;
1365
1366 read_lock_bh(&idev->lock);
1367 list_for_each_entry(score->ifa, &idev->addr_list, if_list) {
1368 int i;
1369
1370 /*
1371 * - Tentative Address (RFC2462 section 5.4)
1372 * - A tentative address is not considered
1373 * "assigned to an interface" in the traditional
1374 * sense, unless it is also flagged as optimistic.
1375 * - Candidate Source Address (section 4)
1376 * - In any case, anycast addresses, multicast
1377 * addresses, and the unspecified address MUST
1378 * NOT be included in a candidate set.
1379 */
1380 if ((score->ifa->flags & IFA_F_TENTATIVE) &&
1381 (!(score->ifa->flags & IFA_F_OPTIMISTIC)))
1382 continue;
1383
1384 score->addr_type = __ipv6_addr_type(&score->ifa->addr);
1385
1386 if (unlikely(score->addr_type == IPV6_ADDR_ANY ||
1387 score->addr_type & IPV6_ADDR_MULTICAST)) {
1388 LIMIT_NETDEBUG(KERN_DEBUG
1389 "ADDRCONF: unspecified / multicast address "
1390 "assigned as unicast address on %s",
1391 dev->name);
1392 continue;
1393 }
1394
1395 score->rule = -1;
1396 bitmap_zero(score->scorebits, IPV6_SADDR_RULE_MAX);
1397
1398 for (i = 0; i < IPV6_SADDR_RULE_MAX; i++) {
1399 int minihiscore, miniscore;
1400
1401 minihiscore = ipv6_get_saddr_eval(net, hiscore, &dst, i);
1402 miniscore = ipv6_get_saddr_eval(net, score, &dst, i);
1403
1404 if (minihiscore > miniscore) {
1405 if (i == IPV6_SADDR_RULE_SCOPE &&
1406 score->scopedist > 0) {
1407 /*
1408 * special case:
1409 * each remaining entry
1410 * has too small (not enough)
1411 * scope, because ifa entries
1412 * are sorted by their scope
1413 * values.
1414 */
1415 goto try_nextdev;
1416 }
1417 break;
1418 } else if (minihiscore < miniscore) {
1419 if (hiscore->ifa)
1420 in6_ifa_put(hiscore->ifa);
1421
1422 in6_ifa_hold(score->ifa);
1423
1424 swap(hiscore, score);
1425
1426 /* restore our iterator */
1427 score->ifa = hiscore->ifa;
1428
1429 break;
1430 }
1431 }
1432 }
1433 try_nextdev:
1434 read_unlock_bh(&idev->lock);
1435 }
1436 rcu_read_unlock();
1437
1438 if (!hiscore->ifa)
1439 return -EADDRNOTAVAIL;
1440
1441 *saddr = hiscore->ifa->addr;
1442 in6_ifa_put(hiscore->ifa);
1443 return 0;
1444 }
1445 EXPORT_SYMBOL(ipv6_dev_get_saddr);
1446
1447 int __ipv6_get_lladdr(struct inet6_dev *idev, struct in6_addr *addr,
1448 unsigned char banned_flags)
1449 {
1450 struct inet6_ifaddr *ifp;
1451 int err = -EADDRNOTAVAIL;
1452
1453 list_for_each_entry(ifp, &idev->addr_list, if_list) {
1454 if (ifp->scope == IFA_LINK &&
1455 !(ifp->flags & banned_flags)) {
1456 *addr = ifp->addr;
1457 err = 0;
1458 break;
1459 }
1460 }
1461 return err;
1462 }
1463
1464 int ipv6_get_lladdr(struct net_device *dev, struct in6_addr *addr,
1465 unsigned char banned_flags)
1466 {
1467 struct inet6_dev *idev;
1468 int err = -EADDRNOTAVAIL;
1469
1470 rcu_read_lock();
1471 idev = __in6_dev_get(dev);
1472 if (idev) {
1473 read_lock_bh(&idev->lock);
1474 err = __ipv6_get_lladdr(idev, addr, banned_flags);
1475 read_unlock_bh(&idev->lock);
1476 }
1477 rcu_read_unlock();
1478 return err;
1479 }
1480
1481 static int ipv6_count_addresses(struct inet6_dev *idev)
1482 {
1483 int cnt = 0;
1484 struct inet6_ifaddr *ifp;
1485
1486 read_lock_bh(&idev->lock);
1487 list_for_each_entry(ifp, &idev->addr_list, if_list)
1488 cnt++;
1489 read_unlock_bh(&idev->lock);
1490 return cnt;
1491 }
1492
1493 int ipv6_chk_addr(struct net *net, const struct in6_addr *addr,
1494 const struct net_device *dev, int strict)
1495 {
1496 struct inet6_ifaddr *ifp;
1497 unsigned int hash = inet6_addr_hash(addr);
1498
1499 rcu_read_lock_bh();
1500 hlist_for_each_entry_rcu(ifp, &inet6_addr_lst[hash], addr_lst) {
1501 if (!net_eq(dev_net(ifp->idev->dev), net))
1502 continue;
1503 if (ipv6_addr_equal(&ifp->addr, addr) &&
1504 !(ifp->flags&IFA_F_TENTATIVE) &&
1505 (dev == NULL || ifp->idev->dev == dev ||
1506 !(ifp->scope&(IFA_LINK|IFA_HOST) || strict))) {
1507 rcu_read_unlock_bh();
1508 return 1;
1509 }
1510 }
1511
1512 rcu_read_unlock_bh();
1513 return 0;
1514 }
1515 EXPORT_SYMBOL(ipv6_chk_addr);
1516
1517 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
1518 struct net_device *dev)
1519 {
1520 unsigned int hash = inet6_addr_hash(addr);
1521 struct inet6_ifaddr *ifp;
1522
1523 hlist_for_each_entry(ifp, &inet6_addr_lst[hash], addr_lst) {
1524 if (!net_eq(dev_net(ifp->idev->dev), net))
1525 continue;
1526 if (ipv6_addr_equal(&ifp->addr, addr)) {
1527 if (dev == NULL || ifp->idev->dev == dev)
1528 return true;
1529 }
1530 }
1531 return false;
1532 }
1533
1534 int ipv6_chk_prefix(const struct in6_addr *addr, struct net_device *dev)
1535 {
1536 struct inet6_dev *idev;
1537 struct inet6_ifaddr *ifa;
1538 int onlink;
1539
1540 onlink = 0;
1541 rcu_read_lock();
1542 idev = __in6_dev_get(dev);
1543 if (idev) {
1544 read_lock_bh(&idev->lock);
1545 list_for_each_entry(ifa, &idev->addr_list, if_list) {
1546 onlink = ipv6_prefix_equal(addr, &ifa->addr,
1547 ifa->prefix_len);
1548 if (onlink)
1549 break;
1550 }
1551 read_unlock_bh(&idev->lock);
1552 }
1553 rcu_read_unlock();
1554 return onlink;
1555 }
1556 EXPORT_SYMBOL(ipv6_chk_prefix);
1557
1558 struct inet6_ifaddr *ipv6_get_ifaddr(struct net *net, const struct in6_addr *addr,
1559 struct net_device *dev, int strict)
1560 {
1561 struct inet6_ifaddr *ifp, *result = NULL;
1562 unsigned int hash = inet6_addr_hash(addr);
1563
1564 rcu_read_lock_bh();
1565 hlist_for_each_entry_rcu_bh(ifp, &inet6_addr_lst[hash], addr_lst) {
1566 if (!net_eq(dev_net(ifp->idev->dev), net))
1567 continue;
1568 if (ipv6_addr_equal(&ifp->addr, addr)) {
1569 if (dev == NULL || ifp->idev->dev == dev ||
1570 !(ifp->scope&(IFA_LINK|IFA_HOST) || strict)) {
1571 result = ifp;
1572 in6_ifa_hold(ifp);
1573 break;
1574 }
1575 }
1576 }
1577 rcu_read_unlock_bh();
1578
1579 return result;
1580 }
1581
1582 /* Gets referenced address, destroys ifaddr */
1583
1584 static void addrconf_dad_stop(struct inet6_ifaddr *ifp, int dad_failed)
1585 {
1586 if (ifp->flags&IFA_F_PERMANENT) {
1587 spin_lock_bh(&ifp->lock);
1588 addrconf_del_dad_timer(ifp);
1589 ifp->flags |= IFA_F_TENTATIVE;
1590 if (dad_failed)
1591 ifp->flags |= IFA_F_DADFAILED;
1592 spin_unlock_bh(&ifp->lock);
1593 if (dad_failed)
1594 ipv6_ifa_notify(0, ifp);
1595 in6_ifa_put(ifp);
1596 #ifdef CONFIG_IPV6_PRIVACY
1597 } else if (ifp->flags&IFA_F_TEMPORARY) {
1598 struct inet6_ifaddr *ifpub;
1599 spin_lock_bh(&ifp->lock);
1600 ifpub = ifp->ifpub;
1601 if (ifpub) {
1602 in6_ifa_hold(ifpub);
1603 spin_unlock_bh(&ifp->lock);
1604 ipv6_create_tempaddr(ifpub, ifp);
1605 in6_ifa_put(ifpub);
1606 } else {
1607 spin_unlock_bh(&ifp->lock);
1608 }
1609 ipv6_del_addr(ifp);
1610 #endif
1611 } else
1612 ipv6_del_addr(ifp);
1613 }
1614
1615 static int addrconf_dad_end(struct inet6_ifaddr *ifp)
1616 {
1617 int err = -ENOENT;
1618
1619 spin_lock(&ifp->state_lock);
1620 if (ifp->state == INET6_IFADDR_STATE_DAD) {
1621 ifp->state = INET6_IFADDR_STATE_POSTDAD;
1622 err = 0;
1623 }
1624 spin_unlock(&ifp->state_lock);
1625
1626 return err;
1627 }
1628
1629 void addrconf_dad_failure(struct inet6_ifaddr *ifp)
1630 {
1631 struct inet6_dev *idev = ifp->idev;
1632
1633 if (addrconf_dad_end(ifp)) {
1634 in6_ifa_put(ifp);
1635 return;
1636 }
1637
1638 net_info_ratelimited("%s: IPv6 duplicate address %pI6c detected!\n",
1639 ifp->idev->dev->name, &ifp->addr);
1640
1641 if (idev->cnf.accept_dad > 1 && !idev->cnf.disable_ipv6) {
1642 struct in6_addr addr;
1643
1644 addr.s6_addr32[0] = htonl(0xfe800000);
1645 addr.s6_addr32[1] = 0;
1646
1647 if (!ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) &&
1648 ipv6_addr_equal(&ifp->addr, &addr)) {
1649 /* DAD failed for link-local based on MAC address */
1650 idev->cnf.disable_ipv6 = 1;
1651
1652 pr_info("%s: IPv6 being disabled!\n",
1653 ifp->idev->dev->name);
1654 }
1655 }
1656
1657 addrconf_dad_stop(ifp, 1);
1658 }
1659
1660 /* Join to solicited addr multicast group. */
1661
1662 void addrconf_join_solict(struct net_device *dev, const struct in6_addr *addr)
1663 {
1664 struct in6_addr maddr;
1665
1666 if (dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1667 return;
1668
1669 addrconf_addr_solict_mult(addr, &maddr);
1670 ipv6_dev_mc_inc(dev, &maddr);
1671 }
1672
1673 void addrconf_leave_solict(struct inet6_dev *idev, const struct in6_addr *addr)
1674 {
1675 struct in6_addr maddr;
1676
1677 if (idev->dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1678 return;
1679
1680 addrconf_addr_solict_mult(addr, &maddr);
1681 __ipv6_dev_mc_dec(idev, &maddr);
1682 }
1683
1684 static void addrconf_join_anycast(struct inet6_ifaddr *ifp)
1685 {
1686 struct in6_addr addr;
1687 if (ifp->prefix_len == 127) /* RFC 6164 */
1688 return;
1689 ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1690 if (ipv6_addr_any(&addr))
1691 return;
1692 ipv6_dev_ac_inc(ifp->idev->dev, &addr);
1693 }
1694
1695 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp)
1696 {
1697 struct in6_addr addr;
1698 if (ifp->prefix_len == 127) /* RFC 6164 */
1699 return;
1700 ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1701 if (ipv6_addr_any(&addr))
1702 return;
1703 __ipv6_dev_ac_dec(ifp->idev, &addr);
1704 }
1705
1706 static int addrconf_ifid_eui48(u8 *eui, struct net_device *dev)
1707 {
1708 if (dev->addr_len != ETH_ALEN)
1709 return -1;
1710 memcpy(eui, dev->dev_addr, 3);
1711 memcpy(eui + 5, dev->dev_addr + 3, 3);
1712
1713 /*
1714 * The zSeries OSA network cards can be shared among various
1715 * OS instances, but the OSA cards have only one MAC address.
1716 * This leads to duplicate address conflicts in conjunction
1717 * with IPv6 if more than one instance uses the same card.
1718 *
1719 * The driver for these cards can deliver a unique 16-bit
1720 * identifier for each instance sharing the same card. It is
1721 * placed instead of 0xFFFE in the interface identifier. The
1722 * "u" bit of the interface identifier is not inverted in this
1723 * case. Hence the resulting interface identifier has local
1724 * scope according to RFC2373.
1725 */
1726 if (dev->dev_id) {
1727 eui[3] = (dev->dev_id >> 8) & 0xFF;
1728 eui[4] = dev->dev_id & 0xFF;
1729 } else {
1730 eui[3] = 0xFF;
1731 eui[4] = 0xFE;
1732 eui[0] ^= 2;
1733 }
1734 return 0;
1735 }
1736
1737 static int addrconf_ifid_eui64(u8 *eui, struct net_device *dev)
1738 {
1739 if (dev->addr_len != IEEE802154_ADDR_LEN)
1740 return -1;
1741 memcpy(eui, dev->dev_addr, 8);
1742 eui[0] ^= 2;
1743 return 0;
1744 }
1745
1746 static int addrconf_ifid_ieee1394(u8 *eui, struct net_device *dev)
1747 {
1748 union fwnet_hwaddr *ha;
1749
1750 if (dev->addr_len != FWNET_ALEN)
1751 return -1;
1752
1753 ha = (union fwnet_hwaddr *)dev->dev_addr;
1754
1755 memcpy(eui, &ha->uc.uniq_id, sizeof(ha->uc.uniq_id));
1756 eui[0] ^= 2;
1757 return 0;
1758 }
1759
1760 static int addrconf_ifid_arcnet(u8 *eui, struct net_device *dev)
1761 {
1762 /* XXX: inherit EUI-64 from other interface -- yoshfuji */
1763 if (dev->addr_len != ARCNET_ALEN)
1764 return -1;
1765 memset(eui, 0, 7);
1766 eui[7] = *(u8 *)dev->dev_addr;
1767 return 0;
1768 }
1769
1770 static int addrconf_ifid_infiniband(u8 *eui, struct net_device *dev)
1771 {
1772 if (dev->addr_len != INFINIBAND_ALEN)
1773 return -1;
1774 memcpy(eui, dev->dev_addr + 12, 8);
1775 eui[0] |= 2;
1776 return 0;
1777 }
1778
1779 static int __ipv6_isatap_ifid(u8 *eui, __be32 addr)
1780 {
1781 if (addr == 0)
1782 return -1;
1783 eui[0] = (ipv4_is_zeronet(addr) || ipv4_is_private_10(addr) ||
1784 ipv4_is_loopback(addr) || ipv4_is_linklocal_169(addr) ||
1785 ipv4_is_private_172(addr) || ipv4_is_test_192(addr) ||
1786 ipv4_is_anycast_6to4(addr) || ipv4_is_private_192(addr) ||
1787 ipv4_is_test_198(addr) || ipv4_is_multicast(addr) ||
1788 ipv4_is_lbcast(addr)) ? 0x00 : 0x02;
1789 eui[1] = 0;
1790 eui[2] = 0x5E;
1791 eui[3] = 0xFE;
1792 memcpy(eui + 4, &addr, 4);
1793 return 0;
1794 }
1795
1796 static int addrconf_ifid_sit(u8 *eui, struct net_device *dev)
1797 {
1798 if (dev->priv_flags & IFF_ISATAP)
1799 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
1800 return -1;
1801 }
1802
1803 static int addrconf_ifid_gre(u8 *eui, struct net_device *dev)
1804 {
1805 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
1806 }
1807
1808 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev)
1809 {
1810 switch (dev->type) {
1811 case ARPHRD_ETHER:
1812 case ARPHRD_FDDI:
1813 return addrconf_ifid_eui48(eui, dev);
1814 case ARPHRD_ARCNET:
1815 return addrconf_ifid_arcnet(eui, dev);
1816 case ARPHRD_INFINIBAND:
1817 return addrconf_ifid_infiniband(eui, dev);
1818 case ARPHRD_SIT:
1819 return addrconf_ifid_sit(eui, dev);
1820 case ARPHRD_IPGRE:
1821 return addrconf_ifid_gre(eui, dev);
1822 case ARPHRD_IEEE802154:
1823 return addrconf_ifid_eui64(eui, dev);
1824 case ARPHRD_IEEE1394:
1825 return addrconf_ifid_ieee1394(eui, dev);
1826 }
1827 return -1;
1828 }
1829
1830 static int ipv6_inherit_eui64(u8 *eui, struct inet6_dev *idev)
1831 {
1832 int err = -1;
1833 struct inet6_ifaddr *ifp;
1834
1835 read_lock_bh(&idev->lock);
1836 list_for_each_entry(ifp, &idev->addr_list, if_list) {
1837 if (ifp->scope == IFA_LINK && !(ifp->flags&IFA_F_TENTATIVE)) {
1838 memcpy(eui, ifp->addr.s6_addr+8, 8);
1839 err = 0;
1840 break;
1841 }
1842 }
1843 read_unlock_bh(&idev->lock);
1844 return err;
1845 }
1846
1847 #ifdef CONFIG_IPV6_PRIVACY
1848 /* (re)generation of randomized interface identifier (RFC 3041 3.2, 3.5) */
1849 static void __ipv6_regen_rndid(struct inet6_dev *idev)
1850 {
1851 regen:
1852 get_random_bytes(idev->rndid, sizeof(idev->rndid));
1853 idev->rndid[0] &= ~0x02;
1854
1855 /*
1856 * <draft-ietf-ipngwg-temp-addresses-v2-00.txt>:
1857 * check if generated address is not inappropriate
1858 *
1859 * - Reserved subnet anycast (RFC 2526)
1860 * 11111101 11....11 1xxxxxxx
1861 * - ISATAP (RFC4214) 6.1
1862 * 00-00-5E-FE-xx-xx-xx-xx
1863 * - value 0
1864 * - XXX: already assigned to an address on the device
1865 */
1866 if (idev->rndid[0] == 0xfd &&
1867 (idev->rndid[1]&idev->rndid[2]&idev->rndid[3]&idev->rndid[4]&idev->rndid[5]&idev->rndid[6]) == 0xff &&
1868 (idev->rndid[7]&0x80))
1869 goto regen;
1870 if ((idev->rndid[0]|idev->rndid[1]) == 0) {
1871 if (idev->rndid[2] == 0x5e && idev->rndid[3] == 0xfe)
1872 goto regen;
1873 if ((idev->rndid[2]|idev->rndid[3]|idev->rndid[4]|idev->rndid[5]|idev->rndid[6]|idev->rndid[7]) == 0x00)
1874 goto regen;
1875 }
1876 }
1877
1878 static void ipv6_regen_rndid(unsigned long data)
1879 {
1880 struct inet6_dev *idev = (struct inet6_dev *) data;
1881 unsigned long expires;
1882
1883 rcu_read_lock_bh();
1884 write_lock_bh(&idev->lock);
1885
1886 if (idev->dead)
1887 goto out;
1888
1889 __ipv6_regen_rndid(idev);
1890
1891 expires = jiffies +
1892 idev->cnf.temp_prefered_lft * HZ -
1893 idev->cnf.regen_max_retry * idev->cnf.dad_transmits * idev->nd_parms->retrans_time -
1894 idev->cnf.max_desync_factor * HZ;
1895 if (time_before(expires, jiffies)) {
1896 pr_warn("%s: too short regeneration interval; timer disabled for %s\n",
1897 __func__, idev->dev->name);
1898 goto out;
1899 }
1900
1901 if (!mod_timer(&idev->regen_timer, expires))
1902 in6_dev_hold(idev);
1903
1904 out:
1905 write_unlock_bh(&idev->lock);
1906 rcu_read_unlock_bh();
1907 in6_dev_put(idev);
1908 }
1909
1910 static void __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr)
1911 {
1912 if (tmpaddr && memcmp(idev->rndid, &tmpaddr->s6_addr[8], 8) == 0)
1913 __ipv6_regen_rndid(idev);
1914 }
1915 #endif
1916
1917 /*
1918 * Add prefix route.
1919 */
1920
1921 static void
1922 addrconf_prefix_route(struct in6_addr *pfx, int plen, struct net_device *dev,
1923 unsigned long expires, u32 flags)
1924 {
1925 struct fib6_config cfg = {
1926 .fc_table = RT6_TABLE_PREFIX,
1927 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1928 .fc_ifindex = dev->ifindex,
1929 .fc_expires = expires,
1930 .fc_dst_len = plen,
1931 .fc_flags = RTF_UP | flags,
1932 .fc_nlinfo.nl_net = dev_net(dev),
1933 .fc_protocol = RTPROT_KERNEL,
1934 };
1935
1936 cfg.fc_dst = *pfx;
1937
1938 /* Prevent useless cloning on PtP SIT.
1939 This thing is done here expecting that the whole
1940 class of non-broadcast devices need not cloning.
1941 */
1942 #if IS_ENABLED(CONFIG_IPV6_SIT)
1943 if (dev->type == ARPHRD_SIT && (dev->flags & IFF_POINTOPOINT))
1944 cfg.fc_flags |= RTF_NONEXTHOP;
1945 #endif
1946
1947 ip6_route_add(&cfg);
1948 }
1949
1950
1951 static struct rt6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
1952 int plen,
1953 const struct net_device *dev,
1954 u32 flags, u32 noflags)
1955 {
1956 struct fib6_node *fn;
1957 struct rt6_info *rt = NULL;
1958 struct fib6_table *table;
1959
1960 table = fib6_get_table(dev_net(dev), RT6_TABLE_PREFIX);
1961 if (table == NULL)
1962 return NULL;
1963
1964 read_lock_bh(&table->tb6_lock);
1965 fn = fib6_locate(&table->tb6_root, pfx, plen, NULL, 0);
1966 if (!fn)
1967 goto out;
1968 for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
1969 if (rt->dst.dev->ifindex != dev->ifindex)
1970 continue;
1971 if ((rt->rt6i_flags & flags) != flags)
1972 continue;
1973 if ((rt->rt6i_flags & noflags) != 0)
1974 continue;
1975 dst_hold(&rt->dst);
1976 break;
1977 }
1978 out:
1979 read_unlock_bh(&table->tb6_lock);
1980 return rt;
1981 }
1982
1983
1984 /* Create "default" multicast route to the interface */
1985
1986 static void addrconf_add_mroute(struct net_device *dev)
1987 {
1988 struct fib6_config cfg = {
1989 .fc_table = RT6_TABLE_LOCAL,
1990 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1991 .fc_ifindex = dev->ifindex,
1992 .fc_dst_len = 8,
1993 .fc_flags = RTF_UP,
1994 .fc_nlinfo.nl_net = dev_net(dev),
1995 };
1996
1997 ipv6_addr_set(&cfg.fc_dst, htonl(0xFF000000), 0, 0, 0);
1998
1999 ip6_route_add(&cfg);
2000 }
2001
2002 #if IS_ENABLED(CONFIG_IPV6_SIT)
2003 static void sit_route_add(struct net_device *dev)
2004 {
2005 struct fib6_config cfg = {
2006 .fc_table = RT6_TABLE_MAIN,
2007 .fc_metric = IP6_RT_PRIO_ADDRCONF,
2008 .fc_ifindex = dev->ifindex,
2009 .fc_dst_len = 96,
2010 .fc_flags = RTF_UP | RTF_NONEXTHOP,
2011 .fc_nlinfo.nl_net = dev_net(dev),
2012 };
2013
2014 /* prefix length - 96 bits "::d.d.d.d" */
2015 ip6_route_add(&cfg);
2016 }
2017 #endif
2018
2019 static struct inet6_dev *addrconf_add_dev(struct net_device *dev)
2020 {
2021 struct inet6_dev *idev;
2022
2023 ASSERT_RTNL();
2024
2025 idev = ipv6_find_idev(dev);
2026 if (!idev)
2027 return ERR_PTR(-ENOBUFS);
2028
2029 if (idev->cnf.disable_ipv6)
2030 return ERR_PTR(-EACCES);
2031
2032 /* Add default multicast route */
2033 if (!(dev->flags & IFF_LOOPBACK))
2034 addrconf_add_mroute(dev);
2035
2036 return idev;
2037 }
2038
2039 void addrconf_prefix_rcv(struct net_device *dev, u8 *opt, int len, bool sllao)
2040 {
2041 struct prefix_info *pinfo;
2042 __u32 valid_lft;
2043 __u32 prefered_lft;
2044 int addr_type;
2045 struct inet6_dev *in6_dev;
2046 struct net *net = dev_net(dev);
2047
2048 pinfo = (struct prefix_info *) opt;
2049
2050 if (len < sizeof(struct prefix_info)) {
2051 ADBG(("addrconf: prefix option too short\n"));
2052 return;
2053 }
2054
2055 /*
2056 * Validation checks ([ADDRCONF], page 19)
2057 */
2058
2059 addr_type = ipv6_addr_type(&pinfo->prefix);
2060
2061 if (addr_type & (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL))
2062 return;
2063
2064 valid_lft = ntohl(pinfo->valid);
2065 prefered_lft = ntohl(pinfo->prefered);
2066
2067 if (prefered_lft > valid_lft) {
2068 net_warn_ratelimited("addrconf: prefix option has invalid lifetime\n");
2069 return;
2070 }
2071
2072 in6_dev = in6_dev_get(dev);
2073
2074 if (in6_dev == NULL) {
2075 net_dbg_ratelimited("addrconf: device %s not configured\n",
2076 dev->name);
2077 return;
2078 }
2079
2080 /*
2081 * Two things going on here:
2082 * 1) Add routes for on-link prefixes
2083 * 2) Configure prefixes with the auto flag set
2084 */
2085
2086 if (pinfo->onlink) {
2087 struct rt6_info *rt;
2088 unsigned long rt_expires;
2089
2090 /* Avoid arithmetic overflow. Really, we could
2091 * save rt_expires in seconds, likely valid_lft,
2092 * but it would require division in fib gc, that it
2093 * not good.
2094 */
2095 if (HZ > USER_HZ)
2096 rt_expires = addrconf_timeout_fixup(valid_lft, HZ);
2097 else
2098 rt_expires = addrconf_timeout_fixup(valid_lft, USER_HZ);
2099
2100 if (addrconf_finite_timeout(rt_expires))
2101 rt_expires *= HZ;
2102
2103 rt = addrconf_get_prefix_route(&pinfo->prefix,
2104 pinfo->prefix_len,
2105 dev,
2106 RTF_ADDRCONF | RTF_PREFIX_RT,
2107 RTF_GATEWAY | RTF_DEFAULT);
2108
2109 if (rt) {
2110 /* Autoconf prefix route */
2111 if (valid_lft == 0) {
2112 ip6_del_rt(rt);
2113 rt = NULL;
2114 } else if (addrconf_finite_timeout(rt_expires)) {
2115 /* not infinity */
2116 rt6_set_expires(rt, jiffies + rt_expires);
2117 } else {
2118 rt6_clean_expires(rt);
2119 }
2120 } else if (valid_lft) {
2121 clock_t expires = 0;
2122 int flags = RTF_ADDRCONF | RTF_PREFIX_RT;
2123 if (addrconf_finite_timeout(rt_expires)) {
2124 /* not infinity */
2125 flags |= RTF_EXPIRES;
2126 expires = jiffies_to_clock_t(rt_expires);
2127 }
2128 addrconf_prefix_route(&pinfo->prefix, pinfo->prefix_len,
2129 dev, expires, flags);
2130 }
2131 ip6_rt_put(rt);
2132 }
2133
2134 /* Try to figure out our local address for this prefix */
2135
2136 if (pinfo->autoconf && in6_dev->cnf.autoconf) {
2137 struct inet6_ifaddr *ifp;
2138 struct in6_addr addr;
2139 int create = 0, update_lft = 0;
2140 bool tokenized = false;
2141
2142 if (pinfo->prefix_len == 64) {
2143 memcpy(&addr, &pinfo->prefix, 8);
2144
2145 if (!ipv6_addr_any(&in6_dev->token)) {
2146 read_lock_bh(&in6_dev->lock);
2147 memcpy(addr.s6_addr + 8,
2148 in6_dev->token.s6_addr + 8, 8);
2149 read_unlock_bh(&in6_dev->lock);
2150 tokenized = true;
2151 } else if (ipv6_generate_eui64(addr.s6_addr + 8, dev) &&
2152 ipv6_inherit_eui64(addr.s6_addr + 8, in6_dev)) {
2153 in6_dev_put(in6_dev);
2154 return;
2155 }
2156 goto ok;
2157 }
2158 net_dbg_ratelimited("IPv6 addrconf: prefix with wrong length %d\n",
2159 pinfo->prefix_len);
2160 in6_dev_put(in6_dev);
2161 return;
2162
2163 ok:
2164
2165 ifp = ipv6_get_ifaddr(net, &addr, dev, 1);
2166
2167 if (ifp == NULL && valid_lft) {
2168 int max_addresses = in6_dev->cnf.max_addresses;
2169 u32 addr_flags = 0;
2170
2171 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2172 if (in6_dev->cnf.optimistic_dad &&
2173 !net->ipv6.devconf_all->forwarding && sllao)
2174 addr_flags = IFA_F_OPTIMISTIC;
2175 #endif
2176
2177 /* Do not allow to create too much of autoconfigured
2178 * addresses; this would be too easy way to crash kernel.
2179 */
2180 if (!max_addresses ||
2181 ipv6_count_addresses(in6_dev) < max_addresses)
2182 ifp = ipv6_add_addr(in6_dev, &addr, pinfo->prefix_len,
2183 addr_type&IPV6_ADDR_SCOPE_MASK,
2184 addr_flags);
2185
2186 if (IS_ERR_OR_NULL(ifp)) {
2187 in6_dev_put(in6_dev);
2188 return;
2189 }
2190
2191 update_lft = create = 1;
2192 ifp->cstamp = jiffies;
2193 ifp->tokenized = tokenized;
2194 addrconf_dad_start(ifp);
2195 }
2196
2197 if (ifp) {
2198 int flags;
2199 unsigned long now;
2200 #ifdef CONFIG_IPV6_PRIVACY
2201 struct inet6_ifaddr *ift;
2202 #endif
2203 u32 stored_lft;
2204
2205 /* update lifetime (RFC2462 5.5.3 e) */
2206 spin_lock(&ifp->lock);
2207 now = jiffies;
2208 if (ifp->valid_lft > (now - ifp->tstamp) / HZ)
2209 stored_lft = ifp->valid_lft - (now - ifp->tstamp) / HZ;
2210 else
2211 stored_lft = 0;
2212 if (!update_lft && stored_lft) {
2213 if (valid_lft > MIN_VALID_LIFETIME ||
2214 valid_lft > stored_lft)
2215 update_lft = 1;
2216 else if (stored_lft <= MIN_VALID_LIFETIME) {
2217 /* valid_lft <= stored_lft is always true */
2218 /*
2219 * RFC 4862 Section 5.5.3e:
2220 * "Note that the preferred lifetime of
2221 * the corresponding address is always
2222 * reset to the Preferred Lifetime in
2223 * the received Prefix Information
2224 * option, regardless of whether the
2225 * valid lifetime is also reset or
2226 * ignored."
2227 *
2228 * So if the preferred lifetime in
2229 * this advertisement is different
2230 * than what we have stored, but the
2231 * valid lifetime is invalid, just
2232 * reset prefered_lft.
2233 *
2234 * We must set the valid lifetime
2235 * to the stored lifetime since we'll
2236 * be updating the timestamp below,
2237 * else we'll set it back to the
2238 * minimum.
2239 */
2240 if (prefered_lft != ifp->prefered_lft) {
2241 valid_lft = stored_lft;
2242 update_lft = 1;
2243 }
2244 } else {
2245 valid_lft = MIN_VALID_LIFETIME;
2246 if (valid_lft < prefered_lft)
2247 prefered_lft = valid_lft;
2248 update_lft = 1;
2249 }
2250 }
2251
2252 if (update_lft) {
2253 ifp->valid_lft = valid_lft;
2254 ifp->prefered_lft = prefered_lft;
2255 ifp->tstamp = now;
2256 flags = ifp->flags;
2257 ifp->flags &= ~IFA_F_DEPRECATED;
2258 spin_unlock(&ifp->lock);
2259
2260 if (!(flags&IFA_F_TENTATIVE))
2261 ipv6_ifa_notify(0, ifp);
2262 } else
2263 spin_unlock(&ifp->lock);
2264
2265 #ifdef CONFIG_IPV6_PRIVACY
2266 read_lock_bh(&in6_dev->lock);
2267 /* update all temporary addresses in the list */
2268 list_for_each_entry(ift, &in6_dev->tempaddr_list,
2269 tmp_list) {
2270 int age, max_valid, max_prefered;
2271
2272 if (ifp != ift->ifpub)
2273 continue;
2274
2275 /*
2276 * RFC 4941 section 3.3:
2277 * If a received option will extend the lifetime
2278 * of a public address, the lifetimes of
2279 * temporary addresses should be extended,
2280 * subject to the overall constraint that no
2281 * temporary addresses should ever remain
2282 * "valid" or "preferred" for a time longer than
2283 * (TEMP_VALID_LIFETIME) or
2284 * (TEMP_PREFERRED_LIFETIME - DESYNC_FACTOR),
2285 * respectively.
2286 */
2287 age = (now - ift->cstamp) / HZ;
2288 max_valid = in6_dev->cnf.temp_valid_lft - age;
2289 if (max_valid < 0)
2290 max_valid = 0;
2291
2292 max_prefered = in6_dev->cnf.temp_prefered_lft -
2293 in6_dev->cnf.max_desync_factor -
2294 age;
2295 if (max_prefered < 0)
2296 max_prefered = 0;
2297
2298 if (valid_lft > max_valid)
2299 valid_lft = max_valid;
2300
2301 if (prefered_lft > max_prefered)
2302 prefered_lft = max_prefered;
2303
2304 spin_lock(&ift->lock);
2305 flags = ift->flags;
2306 ift->valid_lft = valid_lft;
2307 ift->prefered_lft = prefered_lft;
2308 ift->tstamp = now;
2309 if (prefered_lft > 0)
2310 ift->flags &= ~IFA_F_DEPRECATED;
2311
2312 spin_unlock(&ift->lock);
2313 if (!(flags&IFA_F_TENTATIVE))
2314 ipv6_ifa_notify(0, ift);
2315 }
2316
2317 if ((create || list_empty(&in6_dev->tempaddr_list)) && in6_dev->cnf.use_tempaddr > 0) {
2318 /*
2319 * When a new public address is created as
2320 * described in [ADDRCONF], also create a new
2321 * temporary address. Also create a temporary
2322 * address if it's enabled but no temporary
2323 * address currently exists.
2324 */
2325 read_unlock_bh(&in6_dev->lock);
2326 ipv6_create_tempaddr(ifp, NULL);
2327 } else {
2328 read_unlock_bh(&in6_dev->lock);
2329 }
2330 #endif
2331 in6_ifa_put(ifp);
2332 addrconf_verify(0);
2333 }
2334 }
2335 inet6_prefix_notify(RTM_NEWPREFIX, in6_dev, pinfo);
2336 in6_dev_put(in6_dev);
2337 }
2338
2339 /*
2340 * Set destination address.
2341 * Special case for SIT interfaces where we create a new "virtual"
2342 * device.
2343 */
2344 int addrconf_set_dstaddr(struct net *net, void __user *arg)
2345 {
2346 struct in6_ifreq ireq;
2347 struct net_device *dev;
2348 int err = -EINVAL;
2349
2350 rtnl_lock();
2351
2352 err = -EFAULT;
2353 if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2354 goto err_exit;
2355
2356 dev = __dev_get_by_index(net, ireq.ifr6_ifindex);
2357
2358 err = -ENODEV;
2359 if (dev == NULL)
2360 goto err_exit;
2361
2362 #if IS_ENABLED(CONFIG_IPV6_SIT)
2363 if (dev->type == ARPHRD_SIT) {
2364 const struct net_device_ops *ops = dev->netdev_ops;
2365 struct ifreq ifr;
2366 struct ip_tunnel_parm p;
2367
2368 err = -EADDRNOTAVAIL;
2369 if (!(ipv6_addr_type(&ireq.ifr6_addr) & IPV6_ADDR_COMPATv4))
2370 goto err_exit;
2371
2372 memset(&p, 0, sizeof(p));
2373 p.iph.daddr = ireq.ifr6_addr.s6_addr32[3];
2374 p.iph.saddr = 0;
2375 p.iph.version = 4;
2376 p.iph.ihl = 5;
2377 p.iph.protocol = IPPROTO_IPV6;
2378 p.iph.ttl = 64;
2379 ifr.ifr_ifru.ifru_data = (__force void __user *)&p;
2380
2381 if (ops->ndo_do_ioctl) {
2382 mm_segment_t oldfs = get_fs();
2383
2384 set_fs(KERNEL_DS);
2385 err = ops->ndo_do_ioctl(dev, &ifr, SIOCADDTUNNEL);
2386 set_fs(oldfs);
2387 } else
2388 err = -EOPNOTSUPP;
2389
2390 if (err == 0) {
2391 err = -ENOBUFS;
2392 dev = __dev_get_by_name(net, p.name);
2393 if (!dev)
2394 goto err_exit;
2395 err = dev_open(dev);
2396 }
2397 }
2398 #endif
2399
2400 err_exit:
2401 rtnl_unlock();
2402 return err;
2403 }
2404
2405 /*
2406 * Manual configuration of address on an interface
2407 */
2408 static int inet6_addr_add(struct net *net, int ifindex, const struct in6_addr *pfx,
2409 const struct in6_addr *peer_pfx,
2410 unsigned int plen, __u8 ifa_flags, __u32 prefered_lft,
2411 __u32 valid_lft)
2412 {
2413 struct inet6_ifaddr *ifp;
2414 struct inet6_dev *idev;
2415 struct net_device *dev;
2416 int scope;
2417 u32 flags;
2418 clock_t expires;
2419 unsigned long timeout;
2420
2421 ASSERT_RTNL();
2422
2423 if (plen > 128)
2424 return -EINVAL;
2425
2426 /* check the lifetime */
2427 if (!valid_lft || prefered_lft > valid_lft)
2428 return -EINVAL;
2429
2430 dev = __dev_get_by_index(net, ifindex);
2431 if (!dev)
2432 return -ENODEV;
2433
2434 idev = addrconf_add_dev(dev);
2435 if (IS_ERR(idev))
2436 return PTR_ERR(idev);
2437
2438 scope = ipv6_addr_scope(pfx);
2439
2440 timeout = addrconf_timeout_fixup(valid_lft, HZ);
2441 if (addrconf_finite_timeout(timeout)) {
2442 expires = jiffies_to_clock_t(timeout * HZ);
2443 valid_lft = timeout;
2444 flags = RTF_EXPIRES;
2445 } else {
2446 expires = 0;
2447 flags = 0;
2448 ifa_flags |= IFA_F_PERMANENT;
2449 }
2450
2451 timeout = addrconf_timeout_fixup(prefered_lft, HZ);
2452 if (addrconf_finite_timeout(timeout)) {
2453 if (timeout == 0)
2454 ifa_flags |= IFA_F_DEPRECATED;
2455 prefered_lft = timeout;
2456 }
2457
2458 ifp = ipv6_add_addr(idev, pfx, plen, scope, ifa_flags);
2459
2460 if (!IS_ERR(ifp)) {
2461 spin_lock_bh(&ifp->lock);
2462 ifp->valid_lft = valid_lft;
2463 ifp->prefered_lft = prefered_lft;
2464 ifp->tstamp = jiffies;
2465 if (peer_pfx)
2466 ifp->peer_addr = *peer_pfx;
2467 spin_unlock_bh(&ifp->lock);
2468
2469 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, dev,
2470 expires, flags);
2471 /*
2472 * Note that section 3.1 of RFC 4429 indicates
2473 * that the Optimistic flag should not be set for
2474 * manually configured addresses
2475 */
2476 addrconf_dad_start(ifp);
2477 in6_ifa_put(ifp);
2478 addrconf_verify(0);
2479 return 0;
2480 }
2481
2482 return PTR_ERR(ifp);
2483 }
2484
2485 static int inet6_addr_del(struct net *net, int ifindex, const struct in6_addr *pfx,
2486 unsigned int plen)
2487 {
2488 struct inet6_ifaddr *ifp;
2489 struct inet6_dev *idev;
2490 struct net_device *dev;
2491
2492 if (plen > 128)
2493 return -EINVAL;
2494
2495 dev = __dev_get_by_index(net, ifindex);
2496 if (!dev)
2497 return -ENODEV;
2498
2499 if ((idev = __in6_dev_get(dev)) == NULL)
2500 return -ENXIO;
2501
2502 read_lock_bh(&idev->lock);
2503 list_for_each_entry(ifp, &idev->addr_list, if_list) {
2504 if (ifp->prefix_len == plen &&
2505 ipv6_addr_equal(pfx, &ifp->addr)) {
2506 in6_ifa_hold(ifp);
2507 read_unlock_bh(&idev->lock);
2508
2509 ipv6_del_addr(ifp);
2510 return 0;
2511 }
2512 }
2513 read_unlock_bh(&idev->lock);
2514 return -EADDRNOTAVAIL;
2515 }
2516
2517
2518 int addrconf_add_ifaddr(struct net *net, void __user *arg)
2519 {
2520 struct in6_ifreq ireq;
2521 int err;
2522
2523 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2524 return -EPERM;
2525
2526 if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2527 return -EFAULT;
2528
2529 rtnl_lock();
2530 err = inet6_addr_add(net, ireq.ifr6_ifindex, &ireq.ifr6_addr, NULL,
2531 ireq.ifr6_prefixlen, IFA_F_PERMANENT,
2532 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
2533 rtnl_unlock();
2534 return err;
2535 }
2536
2537 int addrconf_del_ifaddr(struct net *net, void __user *arg)
2538 {
2539 struct in6_ifreq ireq;
2540 int err;
2541
2542 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2543 return -EPERM;
2544
2545 if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2546 return -EFAULT;
2547
2548 rtnl_lock();
2549 err = inet6_addr_del(net, ireq.ifr6_ifindex, &ireq.ifr6_addr,
2550 ireq.ifr6_prefixlen);
2551 rtnl_unlock();
2552 return err;
2553 }
2554
2555 static void add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
2556 int plen, int scope)
2557 {
2558 struct inet6_ifaddr *ifp;
2559
2560 ifp = ipv6_add_addr(idev, addr, plen, scope, IFA_F_PERMANENT);
2561 if (!IS_ERR(ifp)) {
2562 spin_lock_bh(&ifp->lock);
2563 ifp->flags &= ~IFA_F_TENTATIVE;
2564 spin_unlock_bh(&ifp->lock);
2565 ipv6_ifa_notify(RTM_NEWADDR, ifp);
2566 in6_ifa_put(ifp);
2567 }
2568 }
2569
2570 #if IS_ENABLED(CONFIG_IPV6_SIT)
2571 static void sit_add_v4_addrs(struct inet6_dev *idev)
2572 {
2573 struct in6_addr addr;
2574 struct net_device *dev;
2575 struct net *net = dev_net(idev->dev);
2576 int scope;
2577
2578 ASSERT_RTNL();
2579
2580 memset(&addr, 0, sizeof(struct in6_addr));
2581 memcpy(&addr.s6_addr32[3], idev->dev->dev_addr, 4);
2582
2583 if (idev->dev->flags&IFF_POINTOPOINT) {
2584 addr.s6_addr32[0] = htonl(0xfe800000);
2585 scope = IFA_LINK;
2586 } else {
2587 scope = IPV6_ADDR_COMPATv4;
2588 }
2589
2590 if (addr.s6_addr32[3]) {
2591 add_addr(idev, &addr, 128, scope);
2592 return;
2593 }
2594
2595 for_each_netdev(net, dev) {
2596 struct in_device *in_dev = __in_dev_get_rtnl(dev);
2597 if (in_dev && (dev->flags & IFF_UP)) {
2598 struct in_ifaddr *ifa;
2599
2600 int flag = scope;
2601
2602 for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) {
2603 int plen;
2604
2605 addr.s6_addr32[3] = ifa->ifa_local;
2606
2607 if (ifa->ifa_scope == RT_SCOPE_LINK)
2608 continue;
2609 if (ifa->ifa_scope >= RT_SCOPE_HOST) {
2610 if (idev->dev->flags&IFF_POINTOPOINT)
2611 continue;
2612 flag |= IFA_HOST;
2613 }
2614 if (idev->dev->flags&IFF_POINTOPOINT)
2615 plen = 64;
2616 else
2617 plen = 96;
2618
2619 add_addr(idev, &addr, plen, flag);
2620 }
2621 }
2622 }
2623 }
2624 #endif
2625
2626 static void init_loopback(struct net_device *dev)
2627 {
2628 struct inet6_dev *idev;
2629 struct net_device *sp_dev;
2630 struct inet6_ifaddr *sp_ifa;
2631 struct rt6_info *sp_rt;
2632
2633 /* ::1 */
2634
2635 ASSERT_RTNL();
2636
2637 if ((idev = ipv6_find_idev(dev)) == NULL) {
2638 pr_debug("%s: add_dev failed\n", __func__);
2639 return;
2640 }
2641
2642 add_addr(idev, &in6addr_loopback, 128, IFA_HOST);
2643
2644 /* Add routes to other interface's IPv6 addresses */
2645 for_each_netdev(dev_net(dev), sp_dev) {
2646 if (!strcmp(sp_dev->name, dev->name))
2647 continue;
2648
2649 idev = __in6_dev_get(sp_dev);
2650 if (!idev)
2651 continue;
2652
2653 read_lock_bh(&idev->lock);
2654 list_for_each_entry(sp_ifa, &idev->addr_list, if_list) {
2655
2656 if (sp_ifa->flags & (IFA_F_DADFAILED | IFA_F_TENTATIVE))
2657 continue;
2658
2659 if (sp_ifa->rt)
2660 continue;
2661
2662 sp_rt = addrconf_dst_alloc(idev, &sp_ifa->addr, 0);
2663
2664 /* Failure cases are ignored */
2665 if (!IS_ERR(sp_rt)) {
2666 sp_ifa->rt = sp_rt;
2667 ip6_ins_rt(sp_rt);
2668 }
2669 }
2670 read_unlock_bh(&idev->lock);
2671 }
2672 }
2673
2674 static void addrconf_add_linklocal(struct inet6_dev *idev, const struct in6_addr *addr)
2675 {
2676 struct inet6_ifaddr *ifp;
2677 u32 addr_flags = IFA_F_PERMANENT;
2678
2679 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2680 if (idev->cnf.optimistic_dad &&
2681 !dev_net(idev->dev)->ipv6.devconf_all->forwarding)
2682 addr_flags |= IFA_F_OPTIMISTIC;
2683 #endif
2684
2685
2686 ifp = ipv6_add_addr(idev, addr, 64, IFA_LINK, addr_flags);
2687 if (!IS_ERR(ifp)) {
2688 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, idev->dev, 0, 0);
2689 addrconf_dad_start(ifp);
2690 in6_ifa_put(ifp);
2691 }
2692 }
2693
2694 static void addrconf_dev_config(struct net_device *dev)
2695 {
2696 struct in6_addr addr;
2697 struct inet6_dev *idev;
2698
2699 ASSERT_RTNL();
2700
2701 if ((dev->type != ARPHRD_ETHER) &&
2702 (dev->type != ARPHRD_FDDI) &&
2703 (dev->type != ARPHRD_ARCNET) &&
2704 (dev->type != ARPHRD_INFINIBAND) &&
2705 (dev->type != ARPHRD_IEEE802154) &&
2706 (dev->type != ARPHRD_IEEE1394)) {
2707 /* Alas, we support only Ethernet autoconfiguration. */
2708 return;
2709 }
2710
2711 idev = addrconf_add_dev(dev);
2712 if (IS_ERR(idev))
2713 return;
2714
2715 memset(&addr, 0, sizeof(struct in6_addr));
2716 addr.s6_addr32[0] = htonl(0xFE800000);
2717
2718 if (ipv6_generate_eui64(addr.s6_addr + 8, dev) == 0)
2719 addrconf_add_linklocal(idev, &addr);
2720 }
2721
2722 #if IS_ENABLED(CONFIG_IPV6_SIT)
2723 static void addrconf_sit_config(struct net_device *dev)
2724 {
2725 struct inet6_dev *idev;
2726
2727 ASSERT_RTNL();
2728
2729 /*
2730 * Configure the tunnel with one of our IPv4
2731 * addresses... we should configure all of
2732 * our v4 addrs in the tunnel
2733 */
2734
2735 if ((idev = ipv6_find_idev(dev)) == NULL) {
2736 pr_debug("%s: add_dev failed\n", __func__);
2737 return;
2738 }
2739
2740 if (dev->priv_flags & IFF_ISATAP) {
2741 struct in6_addr addr;
2742
2743 ipv6_addr_set(&addr, htonl(0xFE800000), 0, 0, 0);
2744 addrconf_prefix_route(&addr, 64, dev, 0, 0);
2745 if (!ipv6_generate_eui64(addr.s6_addr + 8, dev))
2746 addrconf_add_linklocal(idev, &addr);
2747 return;
2748 }
2749
2750 sit_add_v4_addrs(idev);
2751
2752 if (dev->flags&IFF_POINTOPOINT)
2753 addrconf_add_mroute(dev);
2754 else
2755 sit_route_add(dev);
2756 }
2757 #endif
2758
2759 #if IS_ENABLED(CONFIG_NET_IPGRE)
2760 static void addrconf_gre_config(struct net_device *dev)
2761 {
2762 struct inet6_dev *idev;
2763 struct in6_addr addr;
2764
2765 ASSERT_RTNL();
2766
2767 if ((idev = ipv6_find_idev(dev)) == NULL) {
2768 pr_debug("%s: add_dev failed\n", __func__);
2769 return;
2770 }
2771
2772 ipv6_addr_set(&addr, htonl(0xFE800000), 0, 0, 0);
2773 addrconf_prefix_route(&addr, 64, dev, 0, 0);
2774
2775 if (!ipv6_generate_eui64(addr.s6_addr + 8, dev))
2776 addrconf_add_linklocal(idev, &addr);
2777 }
2778 #endif
2779
2780 static inline int
2781 ipv6_inherit_linklocal(struct inet6_dev *idev, struct net_device *link_dev)
2782 {
2783 struct in6_addr lladdr;
2784
2785 if (!ipv6_get_lladdr(link_dev, &lladdr, IFA_F_TENTATIVE)) {
2786 addrconf_add_linklocal(idev, &lladdr);
2787 return 0;
2788 }
2789 return -1;
2790 }
2791
2792 static void ip6_tnl_add_linklocal(struct inet6_dev *idev)
2793 {
2794 struct net_device *link_dev;
2795 struct net *net = dev_net(idev->dev);
2796
2797 /* first try to inherit the link-local address from the link device */
2798 if (idev->dev->iflink &&
2799 (link_dev = __dev_get_by_index(net, idev->dev->iflink))) {
2800 if (!ipv6_inherit_linklocal(idev, link_dev))
2801 return;
2802 }
2803 /* then try to inherit it from any device */
2804 for_each_netdev(net, link_dev) {
2805 if (!ipv6_inherit_linklocal(idev, link_dev))
2806 return;
2807 }
2808 pr_debug("init ip6-ip6: add_linklocal failed\n");
2809 }
2810
2811 /*
2812 * Autoconfigure tunnel with a link-local address so routing protocols,
2813 * DHCPv6, MLD etc. can be run over the virtual link
2814 */
2815
2816 static void addrconf_ip6_tnl_config(struct net_device *dev)
2817 {
2818 struct inet6_dev *idev;
2819
2820 ASSERT_RTNL();
2821
2822 idev = addrconf_add_dev(dev);
2823 if (IS_ERR(idev)) {
2824 pr_debug("init ip6-ip6: add_dev failed\n");
2825 return;
2826 }
2827 ip6_tnl_add_linklocal(idev);
2828 }
2829
2830 static int addrconf_notify(struct notifier_block *this, unsigned long event,
2831 void *ptr)
2832 {
2833 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
2834 struct inet6_dev *idev = __in6_dev_get(dev);
2835 int run_pending = 0;
2836 int err;
2837
2838 switch (event) {
2839 case NETDEV_REGISTER:
2840 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2841 idev = ipv6_add_dev(dev);
2842 if (!idev)
2843 return notifier_from_errno(-ENOMEM);
2844 }
2845 break;
2846
2847 case NETDEV_UP:
2848 case NETDEV_CHANGE:
2849 if (dev->flags & IFF_SLAVE)
2850 break;
2851
2852 if (event == NETDEV_UP) {
2853 if (!addrconf_qdisc_ok(dev)) {
2854 /* device is not ready yet. */
2855 pr_info("ADDRCONF(NETDEV_UP): %s: link is not ready\n",
2856 dev->name);
2857 break;
2858 }
2859
2860 if (!idev && dev->mtu >= IPV6_MIN_MTU)
2861 idev = ipv6_add_dev(dev);
2862
2863 if (idev) {
2864 idev->if_flags |= IF_READY;
2865 run_pending = 1;
2866 }
2867 } else {
2868 if (!addrconf_qdisc_ok(dev)) {
2869 /* device is still not ready. */
2870 break;
2871 }
2872
2873 if (idev) {
2874 if (idev->if_flags & IF_READY)
2875 /* device is already configured. */
2876 break;
2877 idev->if_flags |= IF_READY;
2878 }
2879
2880 pr_info("ADDRCONF(NETDEV_CHANGE): %s: link becomes ready\n",
2881 dev->name);
2882
2883 run_pending = 1;
2884 }
2885
2886 switch (dev->type) {
2887 #if IS_ENABLED(CONFIG_IPV6_SIT)
2888 case ARPHRD_SIT:
2889 addrconf_sit_config(dev);
2890 break;
2891 #endif
2892 #if IS_ENABLED(CONFIG_NET_IPGRE)
2893 case ARPHRD_IPGRE:
2894 addrconf_gre_config(dev);
2895 break;
2896 #endif
2897 case ARPHRD_TUNNEL6:
2898 addrconf_ip6_tnl_config(dev);
2899 break;
2900 case ARPHRD_LOOPBACK:
2901 init_loopback(dev);
2902 break;
2903
2904 default:
2905 addrconf_dev_config(dev);
2906 break;
2907 }
2908
2909 if (idev) {
2910 if (run_pending)
2911 addrconf_dad_run(idev);
2912
2913 /*
2914 * If the MTU changed during the interface down,
2915 * when the interface up, the changed MTU must be
2916 * reflected in the idev as well as routers.
2917 */
2918 if (idev->cnf.mtu6 != dev->mtu &&
2919 dev->mtu >= IPV6_MIN_MTU) {
2920 rt6_mtu_change(dev, dev->mtu);
2921 idev->cnf.mtu6 = dev->mtu;
2922 }
2923 idev->tstamp = jiffies;
2924 inet6_ifinfo_notify(RTM_NEWLINK, idev);
2925
2926 /*
2927 * If the changed mtu during down is lower than
2928 * IPV6_MIN_MTU stop IPv6 on this interface.
2929 */
2930 if (dev->mtu < IPV6_MIN_MTU)
2931 addrconf_ifdown(dev, 1);
2932 }
2933 break;
2934
2935 case NETDEV_CHANGEMTU:
2936 if (idev && dev->mtu >= IPV6_MIN_MTU) {
2937 rt6_mtu_change(dev, dev->mtu);
2938 idev->cnf.mtu6 = dev->mtu;
2939 break;
2940 }
2941
2942 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2943 idev = ipv6_add_dev(dev);
2944 if (idev)
2945 break;
2946 }
2947
2948 /*
2949 * MTU falled under IPV6_MIN_MTU.
2950 * Stop IPv6 on this interface.
2951 */
2952
2953 case NETDEV_DOWN:
2954 case NETDEV_UNREGISTER:
2955 /*
2956 * Remove all addresses from this interface.
2957 */
2958 addrconf_ifdown(dev, event != NETDEV_DOWN);
2959 break;
2960
2961 case NETDEV_CHANGENAME:
2962 if (idev) {
2963 snmp6_unregister_dev(idev);
2964 addrconf_sysctl_unregister(idev);
2965 addrconf_sysctl_register(idev);
2966 err = snmp6_register_dev(idev);
2967 if (err)
2968 return notifier_from_errno(err);
2969 }
2970 break;
2971
2972 case NETDEV_PRE_TYPE_CHANGE:
2973 case NETDEV_POST_TYPE_CHANGE:
2974 addrconf_type_change(dev, event);
2975 break;
2976 }
2977
2978 return NOTIFY_OK;
2979 }
2980
2981 /*
2982 * addrconf module should be notified of a device going up
2983 */
2984 static struct notifier_block ipv6_dev_notf = {
2985 .notifier_call = addrconf_notify,
2986 };
2987
2988 static void addrconf_type_change(struct net_device *dev, unsigned long event)
2989 {
2990 struct inet6_dev *idev;
2991 ASSERT_RTNL();
2992
2993 idev = __in6_dev_get(dev);
2994
2995 if (event == NETDEV_POST_TYPE_CHANGE)
2996 ipv6_mc_remap(idev);
2997 else if (event == NETDEV_PRE_TYPE_CHANGE)
2998 ipv6_mc_unmap(idev);
2999 }
3000
3001 static int addrconf_ifdown(struct net_device *dev, int how)
3002 {
3003 struct net *net = dev_net(dev);
3004 struct inet6_dev *idev;
3005 struct inet6_ifaddr *ifa;
3006 int state, i;
3007
3008 ASSERT_RTNL();
3009
3010 rt6_ifdown(net, dev);
3011 neigh_ifdown(&nd_tbl, dev);
3012
3013 idev = __in6_dev_get(dev);
3014 if (idev == NULL)
3015 return -ENODEV;
3016
3017 /*
3018 * Step 1: remove reference to ipv6 device from parent device.
3019 * Do not dev_put!
3020 */
3021 if (how) {
3022 idev->dead = 1;
3023
3024 /* protected by rtnl_lock */
3025 RCU_INIT_POINTER(dev->ip6_ptr, NULL);
3026
3027 /* Step 1.5: remove snmp6 entry */
3028 snmp6_unregister_dev(idev);
3029
3030 }
3031
3032 /* Step 2: clear hash table */
3033 for (i = 0; i < IN6_ADDR_HSIZE; i++) {
3034 struct hlist_head *h = &inet6_addr_lst[i];
3035
3036 spin_lock_bh(&addrconf_hash_lock);
3037 restart:
3038 hlist_for_each_entry_rcu(ifa, h, addr_lst) {
3039 if (ifa->idev == idev) {
3040 hlist_del_init_rcu(&ifa->addr_lst);
3041 addrconf_del_dad_timer(ifa);
3042 goto restart;
3043 }
3044 }
3045 spin_unlock_bh(&addrconf_hash_lock);
3046 }
3047
3048 write_lock_bh(&idev->lock);
3049
3050 addrconf_del_rs_timer(idev);
3051
3052 /* Step 2: clear flags for stateless addrconf */
3053 if (!how)
3054 idev->if_flags &= ~(IF_RS_SENT|IF_RA_RCVD|IF_READY);
3055
3056 #ifdef CONFIG_IPV6_PRIVACY
3057 if (how && del_timer(&idev->regen_timer))
3058 in6_dev_put(idev);
3059
3060 /* Step 3: clear tempaddr list */
3061 while (!list_empty(&idev->tempaddr_list)) {
3062 ifa = list_first_entry(&idev->tempaddr_list,
3063 struct inet6_ifaddr, tmp_list);
3064 list_del(&ifa->tmp_list);
3065 write_unlock_bh(&idev->lock);
3066 spin_lock_bh(&ifa->lock);
3067
3068 if (ifa->ifpub) {
3069 in6_ifa_put(ifa->ifpub);
3070 ifa->ifpub = NULL;
3071 }
3072 spin_unlock_bh(&ifa->lock);
3073 in6_ifa_put(ifa);
3074 write_lock_bh(&idev->lock);
3075 }
3076 #endif
3077
3078 while (!list_empty(&idev->addr_list)) {
3079 ifa = list_first_entry(&idev->addr_list,
3080 struct inet6_ifaddr, if_list);
3081 addrconf_del_dad_timer(ifa);
3082
3083 list_del(&ifa->if_list);
3084
3085 write_unlock_bh(&idev->lock);
3086
3087 spin_lock_bh(&ifa->state_lock);
3088 state = ifa->state;
3089 ifa->state = INET6_IFADDR_STATE_DEAD;
3090 spin_unlock_bh(&ifa->state_lock);
3091
3092 if (state != INET6_IFADDR_STATE_DEAD) {
3093 __ipv6_ifa_notify(RTM_DELADDR, ifa);
3094 inet6addr_notifier_call_chain(NETDEV_DOWN, ifa);
3095 }
3096 in6_ifa_put(ifa);
3097
3098 write_lock_bh(&idev->lock);
3099 }
3100
3101 write_unlock_bh(&idev->lock);
3102
3103 /* Step 5: Discard multicast list */
3104 if (how)
3105 ipv6_mc_destroy_dev(idev);
3106 else
3107 ipv6_mc_down(idev);
3108
3109 idev->tstamp = jiffies;
3110
3111 /* Last: Shot the device (if unregistered) */
3112 if (how) {
3113 addrconf_sysctl_unregister(idev);
3114 neigh_parms_release(&nd_tbl, idev->nd_parms);
3115 neigh_ifdown(&nd_tbl, dev);
3116 in6_dev_put(idev);
3117 }
3118 return 0;
3119 }
3120
3121 static void addrconf_rs_timer(unsigned long data)
3122 {
3123 struct inet6_dev *idev = (struct inet6_dev *)data;
3124 struct in6_addr lladdr;
3125
3126 write_lock(&idev->lock);
3127 if (idev->dead || !(idev->if_flags & IF_READY))
3128 goto out;
3129
3130 if (!ipv6_accept_ra(idev))
3131 goto out;
3132
3133 /* Announcement received after solicitation was sent */
3134 if (idev->if_flags & IF_RA_RCVD)
3135 goto out;
3136
3137 if (idev->rs_probes++ < idev->cnf.rtr_solicits) {
3138 if (!__ipv6_get_lladdr(idev, &lladdr, IFA_F_TENTATIVE))
3139 ndisc_send_rs(idev->dev, &lladdr,
3140 &in6addr_linklocal_allrouters);
3141 else
3142 goto out;
3143
3144 /* The wait after the last probe can be shorter */
3145 addrconf_mod_rs_timer(idev, (idev->rs_probes ==
3146 idev->cnf.rtr_solicits) ?
3147 idev->cnf.rtr_solicit_delay :
3148 idev->cnf.rtr_solicit_interval);
3149 } else {
3150 /*
3151 * Note: we do not support deprecated "all on-link"
3152 * assumption any longer.
3153 */
3154 pr_debug("%s: no IPv6 routers present\n", idev->dev->name);
3155 }
3156
3157 out:
3158 write_unlock(&idev->lock);
3159 in6_dev_put(idev);
3160 }
3161
3162 /*
3163 * Duplicate Address Detection
3164 */
3165 static void addrconf_dad_kick(struct inet6_ifaddr *ifp)
3166 {
3167 unsigned long rand_num;
3168 struct inet6_dev *idev = ifp->idev;
3169
3170 if (ifp->flags & IFA_F_OPTIMISTIC)
3171 rand_num = 0;
3172 else
3173 rand_num = net_random() % (idev->cnf.rtr_solicit_delay ? : 1);
3174
3175 ifp->dad_probes = idev->cnf.dad_transmits;
3176 addrconf_mod_dad_timer(ifp, rand_num);
3177 }
3178
3179 static void addrconf_dad_start(struct inet6_ifaddr *ifp)
3180 {
3181 struct inet6_dev *idev = ifp->idev;
3182 struct net_device *dev = idev->dev;
3183
3184 addrconf_join_solict(dev, &ifp->addr);
3185
3186 net_srandom(ifp->addr.s6_addr32[3]);
3187
3188 read_lock_bh(&idev->lock);
3189 spin_lock(&ifp->lock);
3190 if (ifp->state == INET6_IFADDR_STATE_DEAD)
3191 goto out;
3192
3193 if (dev->flags&(IFF_NOARP|IFF_LOOPBACK) ||
3194 idev->cnf.accept_dad < 1 ||
3195 !(ifp->flags&IFA_F_TENTATIVE) ||
3196 ifp->flags & IFA_F_NODAD) {
3197 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
3198 spin_unlock(&ifp->lock);
3199 read_unlock_bh(&idev->lock);
3200
3201 addrconf_dad_completed(ifp);
3202 return;
3203 }
3204
3205 if (!(idev->if_flags & IF_READY)) {
3206 spin_unlock(&ifp->lock);
3207 read_unlock_bh(&idev->lock);
3208 /*
3209 * If the device is not ready:
3210 * - keep it tentative if it is a permanent address.
3211 * - otherwise, kill it.
3212 */
3213 in6_ifa_hold(ifp);
3214 addrconf_dad_stop(ifp, 0);
3215 return;
3216 }
3217
3218 /*
3219 * Optimistic nodes can start receiving
3220 * Frames right away
3221 */
3222 if (ifp->flags & IFA_F_OPTIMISTIC)
3223 ip6_ins_rt(ifp->rt);
3224
3225 addrconf_dad_kick(ifp);
3226 out:
3227 spin_unlock(&ifp->lock);
3228 read_unlock_bh(&idev->lock);
3229 }
3230
3231 static void addrconf_dad_timer(unsigned long data)
3232 {
3233 struct inet6_ifaddr *ifp = (struct inet6_ifaddr *) data;
3234 struct inet6_dev *idev = ifp->idev;
3235 struct in6_addr mcaddr;
3236
3237 if (!ifp->dad_probes && addrconf_dad_end(ifp))
3238 goto out;
3239
3240 write_lock(&idev->lock);
3241 if (idev->dead || !(idev->if_flags & IF_READY)) {
3242 write_unlock(&idev->lock);
3243 goto out;
3244 }
3245
3246 spin_lock(&ifp->lock);
3247 if (ifp->state == INET6_IFADDR_STATE_DEAD) {
3248 spin_unlock(&ifp->lock);
3249 write_unlock(&idev->lock);
3250 goto out;
3251 }
3252
3253 if (ifp->dad_probes == 0) {
3254 /*
3255 * DAD was successful
3256 */
3257
3258 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
3259 spin_unlock(&ifp->lock);
3260 write_unlock(&idev->lock);
3261
3262 addrconf_dad_completed(ifp);
3263
3264 goto out;
3265 }
3266
3267 ifp->dad_probes--;
3268 addrconf_mod_dad_timer(ifp, ifp->idev->nd_parms->retrans_time);
3269 spin_unlock(&ifp->lock);
3270 write_unlock(&idev->lock);
3271
3272 /* send a neighbour solicitation for our addr */
3273 addrconf_addr_solict_mult(&ifp->addr, &mcaddr);
3274 ndisc_send_ns(ifp->idev->dev, NULL, &ifp->addr, &mcaddr, &in6addr_any);
3275 out:
3276 in6_ifa_put(ifp);
3277 }
3278
3279 static void addrconf_dad_completed(struct inet6_ifaddr *ifp)
3280 {
3281 struct net_device *dev = ifp->idev->dev;
3282 struct in6_addr lladdr;
3283 bool send_rs, send_mld;
3284
3285 addrconf_del_dad_timer(ifp);
3286
3287 /*
3288 * Configure the address for reception. Now it is valid.
3289 */
3290
3291 ipv6_ifa_notify(RTM_NEWADDR, ifp);
3292
3293 /* If added prefix is link local and we are prepared to process
3294 router advertisements, start sending router solicitations.
3295 */
3296
3297 read_lock_bh(&ifp->idev->lock);
3298 spin_lock(&ifp->lock);
3299 send_mld = ipv6_addr_type(&ifp->addr) & IPV6_ADDR_LINKLOCAL &&
3300 ifp->idev->valid_ll_addr_cnt == 1;
3301 send_rs = send_mld &&
3302 ipv6_accept_ra(ifp->idev) &&
3303 ifp->idev->cnf.rtr_solicits > 0 &&
3304 (dev->flags&IFF_LOOPBACK) == 0;
3305 spin_unlock(&ifp->lock);
3306 read_unlock_bh(&ifp->idev->lock);
3307
3308 /* While dad is in progress mld report's source address is in6_addrany.
3309 * Resend with proper ll now.
3310 */
3311 if (send_mld)
3312 ipv6_mc_dad_complete(ifp->idev);
3313
3314 if (send_rs) {
3315 /*
3316 * If a host as already performed a random delay
3317 * [...] as part of DAD [...] there is no need
3318 * to delay again before sending the first RS
3319 */
3320 if (ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE))
3321 return;
3322 ndisc_send_rs(dev, &lladdr, &in6addr_linklocal_allrouters);
3323
3324 write_lock_bh(&ifp->idev->lock);
3325 spin_lock(&ifp->lock);
3326 ifp->idev->rs_probes = 1;
3327 ifp->idev->if_flags |= IF_RS_SENT;
3328 addrconf_mod_rs_timer(ifp->idev,
3329 ifp->idev->cnf.rtr_solicit_interval);
3330 spin_unlock(&ifp->lock);
3331 write_unlock_bh(&ifp->idev->lock);
3332 }
3333 }
3334
3335 static void addrconf_dad_run(struct inet6_dev *idev)
3336 {
3337 struct inet6_ifaddr *ifp;
3338
3339 read_lock_bh(&idev->lock);
3340 list_for_each_entry(ifp, &idev->addr_list, if_list) {
3341 spin_lock(&ifp->lock);
3342 if (ifp->flags & IFA_F_TENTATIVE &&
3343 ifp->state == INET6_IFADDR_STATE_DAD)
3344 addrconf_dad_kick(ifp);
3345 spin_unlock(&ifp->lock);
3346 }
3347 read_unlock_bh(&idev->lock);
3348 }
3349
3350 #ifdef CONFIG_PROC_FS
3351 struct if6_iter_state {
3352 struct seq_net_private p;
3353 int bucket;
3354 int offset;
3355 };
3356
3357 static struct inet6_ifaddr *if6_get_first(struct seq_file *seq, loff_t pos)
3358 {
3359 struct inet6_ifaddr *ifa = NULL;
3360 struct if6_iter_state *state = seq->private;
3361 struct net *net = seq_file_net(seq);
3362 int p = 0;
3363
3364 /* initial bucket if pos is 0 */
3365 if (pos == 0) {
3366 state->bucket = 0;
3367 state->offset = 0;
3368 }
3369
3370 for (; state->bucket < IN6_ADDR_HSIZE; ++state->bucket) {
3371 hlist_for_each_entry_rcu_bh(ifa, &inet6_addr_lst[state->bucket],
3372 addr_lst) {
3373 if (!net_eq(dev_net(ifa->idev->dev), net))
3374 continue;
3375 /* sync with offset */
3376 if (p < state->offset) {
3377 p++;
3378 continue;
3379 }
3380 state->offset++;
3381 return ifa;
3382 }
3383
3384 /* prepare for next bucket */
3385 state->offset = 0;
3386 p = 0;
3387 }
3388 return NULL;
3389 }
3390
3391 static struct inet6_ifaddr *if6_get_next(struct seq_file *seq,
3392 struct inet6_ifaddr *ifa)
3393 {
3394 struct if6_iter_state *state = seq->private;
3395 struct net *net = seq_file_net(seq);
3396
3397 hlist_for_each_entry_continue_rcu_bh(ifa, addr_lst) {
3398 if (!net_eq(dev_net(ifa->idev->dev), net))
3399 continue;
3400 state->offset++;
3401 return ifa;
3402 }
3403
3404 while (++state->bucket < IN6_ADDR_HSIZE) {
3405 state->offset = 0;
3406 hlist_for_each_entry_rcu_bh(ifa,
3407 &inet6_addr_lst[state->bucket], addr_lst) {
3408 if (!net_eq(dev_net(ifa->idev->dev), net))
3409 continue;
3410 state->offset++;
3411 return ifa;
3412 }
3413 }
3414
3415 return NULL;
3416 }
3417
3418 static void *if6_seq_start(struct seq_file *seq, loff_t *pos)
3419 __acquires(rcu_bh)
3420 {
3421 rcu_read_lock_bh();
3422 return if6_get_first(seq, *pos);
3423 }
3424
3425 static void *if6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
3426 {
3427 struct inet6_ifaddr *ifa;
3428
3429 ifa = if6_get_next(seq, v);
3430 ++*pos;
3431 return ifa;
3432 }
3433
3434 static void if6_seq_stop(struct seq_file *seq, void *v)
3435 __releases(rcu_bh)
3436 {
3437 rcu_read_unlock_bh();
3438 }
3439
3440 static int if6_seq_show(struct seq_file *seq, void *v)
3441 {
3442 struct inet6_ifaddr *ifp = (struct inet6_ifaddr *)v;
3443 seq_printf(seq, "%pi6 %02x %02x %02x %02x %8s\n",
3444 &ifp->addr,
3445 ifp->idev->dev->ifindex,
3446 ifp->prefix_len,
3447 ifp->scope,
3448 ifp->flags,
3449 ifp->idev->dev->name);
3450 return 0;
3451 }
3452
3453 static const struct seq_operations if6_seq_ops = {
3454 .start = if6_seq_start,
3455 .next = if6_seq_next,
3456 .show = if6_seq_show,
3457 .stop = if6_seq_stop,
3458 };
3459
3460 static int if6_seq_open(struct inode *inode, struct file *file)
3461 {
3462 return seq_open_net(inode, file, &if6_seq_ops,
3463 sizeof(struct if6_iter_state));
3464 }
3465
3466 static const struct file_operations if6_fops = {
3467 .owner = THIS_MODULE,
3468 .open = if6_seq_open,
3469 .read = seq_read,
3470 .llseek = seq_lseek,
3471 .release = seq_release_net,
3472 };
3473
3474 static int __net_init if6_proc_net_init(struct net *net)
3475 {
3476 if (!proc_create("if_inet6", S_IRUGO, net->proc_net, &if6_fops))
3477 return -ENOMEM;
3478 return 0;
3479 }
3480
3481 static void __net_exit if6_proc_net_exit(struct net *net)
3482 {
3483 remove_proc_entry("if_inet6", net->proc_net);
3484 }
3485
3486 static struct pernet_operations if6_proc_net_ops = {
3487 .init = if6_proc_net_init,
3488 .exit = if6_proc_net_exit,
3489 };
3490
3491 int __init if6_proc_init(void)
3492 {
3493 return register_pernet_subsys(&if6_proc_net_ops);
3494 }
3495
3496 void if6_proc_exit(void)
3497 {
3498 unregister_pernet_subsys(&if6_proc_net_ops);
3499 }
3500 #endif /* CONFIG_PROC_FS */
3501
3502 #if IS_ENABLED(CONFIG_IPV6_MIP6)
3503 /* Check if address is a home address configured on any interface. */
3504 int ipv6_chk_home_addr(struct net *net, const struct in6_addr *addr)
3505 {
3506 int ret = 0;
3507 struct inet6_ifaddr *ifp = NULL;
3508 unsigned int hash = inet6_addr_hash(addr);
3509
3510 rcu_read_lock_bh();
3511 hlist_for_each_entry_rcu_bh(ifp, &inet6_addr_lst[hash], addr_lst) {
3512 if (!net_eq(dev_net(ifp->idev->dev), net))
3513 continue;
3514 if (ipv6_addr_equal(&ifp->addr, addr) &&
3515 (ifp->flags & IFA_F_HOMEADDRESS)) {
3516 ret = 1;
3517 break;
3518 }
3519 }
3520 rcu_read_unlock_bh();
3521 return ret;
3522 }
3523 #endif
3524
3525 /*
3526 * Periodic address status verification
3527 */
3528
3529 static void addrconf_verify(unsigned long foo)
3530 {
3531 unsigned long now, next, next_sec, next_sched;
3532 struct inet6_ifaddr *ifp;
3533 int i;
3534
3535 rcu_read_lock_bh();
3536 spin_lock(&addrconf_verify_lock);
3537 now = jiffies;
3538 next = round_jiffies_up(now + ADDR_CHECK_FREQUENCY);
3539
3540 del_timer(&addr_chk_timer);
3541
3542 for (i = 0; i < IN6_ADDR_HSIZE; i++) {
3543 restart:
3544 hlist_for_each_entry_rcu_bh(ifp,
3545 &inet6_addr_lst[i], addr_lst) {
3546 unsigned long age;
3547
3548 if (ifp->flags & IFA_F_PERMANENT)
3549 continue;
3550
3551 spin_lock(&ifp->lock);
3552 /* We try to batch several events at once. */
3553 age = (now - ifp->tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
3554
3555 if (ifp->valid_lft != INFINITY_LIFE_TIME &&
3556 age >= ifp->valid_lft) {
3557 spin_unlock(&ifp->lock);
3558 in6_ifa_hold(ifp);
3559 ipv6_del_addr(ifp);
3560 goto restart;
3561 } else if (ifp->prefered_lft == INFINITY_LIFE_TIME) {
3562 spin_unlock(&ifp->lock);
3563 continue;
3564 } else if (age >= ifp->prefered_lft) {
3565 /* jiffies - ifp->tstamp > age >= ifp->prefered_lft */
3566 int deprecate = 0;
3567
3568 if (!(ifp->flags&IFA_F_DEPRECATED)) {
3569 deprecate = 1;
3570 ifp->flags |= IFA_F_DEPRECATED;
3571 }
3572
3573 if (time_before(ifp->tstamp + ifp->valid_lft * HZ, next))
3574 next = ifp->tstamp + ifp->valid_lft * HZ;
3575
3576 spin_unlock(&ifp->lock);
3577
3578 if (deprecate) {
3579 in6_ifa_hold(ifp);
3580
3581 ipv6_ifa_notify(0, ifp);
3582 in6_ifa_put(ifp);
3583 goto restart;
3584 }
3585 #ifdef CONFIG_IPV6_PRIVACY
3586 } else if ((ifp->flags&IFA_F_TEMPORARY) &&
3587 !(ifp->flags&IFA_F_TENTATIVE)) {
3588 unsigned long regen_advance = ifp->idev->cnf.regen_max_retry *
3589 ifp->idev->cnf.dad_transmits *
3590 ifp->idev->nd_parms->retrans_time / HZ;
3591
3592 if (age >= ifp->prefered_lft - regen_advance) {
3593 struct inet6_ifaddr *ifpub = ifp->ifpub;
3594 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3595 next = ifp->tstamp + ifp->prefered_lft * HZ;
3596 if (!ifp->regen_count && ifpub) {
3597 ifp->regen_count++;
3598 in6_ifa_hold(ifp);
3599 in6_ifa_hold(ifpub);
3600 spin_unlock(&ifp->lock);
3601
3602 spin_lock(&ifpub->lock);
3603 ifpub->regen_count = 0;
3604 spin_unlock(&ifpub->lock);
3605 ipv6_create_tempaddr(ifpub, ifp);
3606 in6_ifa_put(ifpub);
3607 in6_ifa_put(ifp);
3608 goto restart;
3609 }
3610 } else if (time_before(ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ, next))
3611 next = ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ;
3612 spin_unlock(&ifp->lock);
3613 #endif
3614 } else {
3615 /* ifp->prefered_lft <= ifp->valid_lft */
3616 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3617 next = ifp->tstamp + ifp->prefered_lft * HZ;
3618 spin_unlock(&ifp->lock);
3619 }
3620 }
3621 }
3622
3623 next_sec = round_jiffies_up(next);
3624 next_sched = next;
3625
3626 /* If rounded timeout is accurate enough, accept it. */
3627 if (time_before(next_sec, next + ADDRCONF_TIMER_FUZZ))
3628 next_sched = next_sec;
3629
3630 /* And minimum interval is ADDRCONF_TIMER_FUZZ_MAX. */
3631 if (time_before(next_sched, jiffies + ADDRCONF_TIMER_FUZZ_MAX))
3632 next_sched = jiffies + ADDRCONF_TIMER_FUZZ_MAX;
3633
3634 ADBG((KERN_DEBUG "now = %lu, schedule = %lu, rounded schedule = %lu => %lu\n",
3635 now, next, next_sec, next_sched));
3636
3637 addr_chk_timer.expires = next_sched;
3638 add_timer(&addr_chk_timer);
3639 spin_unlock(&addrconf_verify_lock);
3640 rcu_read_unlock_bh();
3641 }
3642
3643 static struct in6_addr *extract_addr(struct nlattr *addr, struct nlattr *local,
3644 struct in6_addr **peer_pfx)
3645 {
3646 struct in6_addr *pfx = NULL;
3647
3648 *peer_pfx = NULL;
3649
3650 if (addr)
3651 pfx = nla_data(addr);
3652
3653 if (local) {
3654 if (pfx && nla_memcmp(local, pfx, sizeof(*pfx)))
3655 *peer_pfx = pfx;
3656 pfx = nla_data(local);
3657 }
3658
3659 return pfx;
3660 }
3661
3662 static const struct nla_policy ifa_ipv6_policy[IFA_MAX+1] = {
3663 [IFA_ADDRESS] = { .len = sizeof(struct in6_addr) },
3664 [IFA_LOCAL] = { .len = sizeof(struct in6_addr) },
3665 [IFA_CACHEINFO] = { .len = sizeof(struct ifa_cacheinfo) },
3666 };
3667
3668 static int
3669 inet6_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh)
3670 {
3671 struct net *net = sock_net(skb->sk);
3672 struct ifaddrmsg *ifm;
3673 struct nlattr *tb[IFA_MAX+1];
3674 struct in6_addr *pfx, *peer_pfx;
3675 int err;
3676
3677 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3678 if (err < 0)
3679 return err;
3680
3681 ifm = nlmsg_data(nlh);
3682 pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
3683 if (pfx == NULL)
3684 return -EINVAL;
3685
3686 return inet6_addr_del(net, ifm->ifa_index, pfx, ifm->ifa_prefixlen);
3687 }
3688
3689 static int inet6_addr_modify(struct inet6_ifaddr *ifp, u8 ifa_flags,
3690 u32 prefered_lft, u32 valid_lft)
3691 {
3692 u32 flags;
3693 clock_t expires;
3694 unsigned long timeout;
3695
3696 if (!valid_lft || (prefered_lft > valid_lft))
3697 return -EINVAL;
3698
3699 timeout = addrconf_timeout_fixup(valid_lft, HZ);
3700 if (addrconf_finite_timeout(timeout)) {
3701 expires = jiffies_to_clock_t(timeout * HZ);
3702 valid_lft = timeout;
3703 flags = RTF_EXPIRES;
3704 } else {
3705 expires = 0;
3706 flags = 0;
3707 ifa_flags |= IFA_F_PERMANENT;
3708 }
3709
3710 timeout = addrconf_timeout_fixup(prefered_lft, HZ);
3711 if (addrconf_finite_timeout(timeout)) {
3712 if (timeout == 0)
3713 ifa_flags |= IFA_F_DEPRECATED;
3714 prefered_lft = timeout;
3715 }
3716
3717 spin_lock_bh(&ifp->lock);
3718 ifp->flags = (ifp->flags & ~(IFA_F_DEPRECATED | IFA_F_PERMANENT | IFA_F_NODAD | IFA_F_HOMEADDRESS)) | ifa_flags;
3719 ifp->tstamp = jiffies;
3720 ifp->valid_lft = valid_lft;
3721 ifp->prefered_lft = prefered_lft;
3722
3723 spin_unlock_bh(&ifp->lock);
3724 if (!(ifp->flags&IFA_F_TENTATIVE))
3725 ipv6_ifa_notify(0, ifp);
3726
3727 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, ifp->idev->dev,
3728 expires, flags);
3729 addrconf_verify(0);
3730
3731 return 0;
3732 }
3733
3734 static int
3735 inet6_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh)
3736 {
3737 struct net *net = sock_net(skb->sk);
3738 struct ifaddrmsg *ifm;
3739 struct nlattr *tb[IFA_MAX+1];
3740 struct in6_addr *pfx, *peer_pfx;
3741 struct inet6_ifaddr *ifa;
3742 struct net_device *dev;
3743 u32 valid_lft = INFINITY_LIFE_TIME, preferred_lft = INFINITY_LIFE_TIME;
3744 u8 ifa_flags;
3745 int err;
3746
3747 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3748 if (err < 0)
3749 return err;
3750
3751 ifm = nlmsg_data(nlh);
3752 pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
3753 if (pfx == NULL)
3754 return -EINVAL;
3755
3756 if (tb[IFA_CACHEINFO]) {
3757 struct ifa_cacheinfo *ci;
3758
3759 ci = nla_data(tb[IFA_CACHEINFO]);
3760 valid_lft = ci->ifa_valid;
3761 preferred_lft = ci->ifa_prefered;
3762 } else {
3763 preferred_lft = INFINITY_LIFE_TIME;
3764 valid_lft = INFINITY_LIFE_TIME;
3765 }
3766
3767 dev = __dev_get_by_index(net, ifm->ifa_index);
3768 if (dev == NULL)
3769 return -ENODEV;
3770
3771 /* We ignore other flags so far. */
3772 ifa_flags = ifm->ifa_flags & (IFA_F_NODAD | IFA_F_HOMEADDRESS);
3773
3774 ifa = ipv6_get_ifaddr(net, pfx, dev, 1);
3775 if (ifa == NULL) {
3776 /*
3777 * It would be best to check for !NLM_F_CREATE here but
3778 * userspace alreay relies on not having to provide this.
3779 */
3780 return inet6_addr_add(net, ifm->ifa_index, pfx, peer_pfx,
3781 ifm->ifa_prefixlen, ifa_flags,
3782 preferred_lft, valid_lft);
3783 }
3784
3785 if (nlh->nlmsg_flags & NLM_F_EXCL ||
3786 !(nlh->nlmsg_flags & NLM_F_REPLACE))
3787 err = -EEXIST;
3788 else
3789 err = inet6_addr_modify(ifa, ifa_flags, preferred_lft, valid_lft);
3790
3791 in6_ifa_put(ifa);
3792
3793 return err;
3794 }
3795
3796 static void put_ifaddrmsg(struct nlmsghdr *nlh, u8 prefixlen, u8 flags,
3797 u8 scope, int ifindex)
3798 {
3799 struct ifaddrmsg *ifm;
3800
3801 ifm = nlmsg_data(nlh);
3802 ifm->ifa_family = AF_INET6;
3803 ifm->ifa_prefixlen = prefixlen;
3804 ifm->ifa_flags = flags;
3805 ifm->ifa_scope = scope;
3806 ifm->ifa_index = ifindex;
3807 }
3808
3809 static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
3810 unsigned long tstamp, u32 preferred, u32 valid)
3811 {
3812 struct ifa_cacheinfo ci;
3813
3814 ci.cstamp = cstamp_delta(cstamp);
3815 ci.tstamp = cstamp_delta(tstamp);
3816 ci.ifa_prefered = preferred;
3817 ci.ifa_valid = valid;
3818
3819 return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
3820 }
3821
3822 static inline int rt_scope(int ifa_scope)
3823 {
3824 if (ifa_scope & IFA_HOST)
3825 return RT_SCOPE_HOST;
3826 else if (ifa_scope & IFA_LINK)
3827 return RT_SCOPE_LINK;
3828 else if (ifa_scope & IFA_SITE)
3829 return RT_SCOPE_SITE;
3830 else
3831 return RT_SCOPE_UNIVERSE;
3832 }
3833
3834 static inline int inet6_ifaddr_msgsize(void)
3835 {
3836 return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
3837 + nla_total_size(16) /* IFA_LOCAL */
3838 + nla_total_size(16) /* IFA_ADDRESS */
3839 + nla_total_size(sizeof(struct ifa_cacheinfo));
3840 }
3841
3842 static int inet6_fill_ifaddr(struct sk_buff *skb, struct inet6_ifaddr *ifa,
3843 u32 portid, u32 seq, int event, unsigned int flags)
3844 {
3845 struct nlmsghdr *nlh;
3846 u32 preferred, valid;
3847
3848 nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
3849 if (nlh == NULL)
3850 return -EMSGSIZE;
3851
3852 put_ifaddrmsg(nlh, ifa->prefix_len, ifa->flags, rt_scope(ifa->scope),
3853 ifa->idev->dev->ifindex);
3854
3855 if (!(ifa->flags&IFA_F_PERMANENT)) {
3856 preferred = ifa->prefered_lft;
3857 valid = ifa->valid_lft;
3858 if (preferred != INFINITY_LIFE_TIME) {
3859 long tval = (jiffies - ifa->tstamp)/HZ;
3860 if (preferred > tval)
3861 preferred -= tval;
3862 else
3863 preferred = 0;
3864 if (valid != INFINITY_LIFE_TIME) {
3865 if (valid > tval)
3866 valid -= tval;
3867 else
3868 valid = 0;
3869 }
3870 }
3871 } else {
3872 preferred = INFINITY_LIFE_TIME;
3873 valid = INFINITY_LIFE_TIME;
3874 }
3875
3876 if (!ipv6_addr_any(&ifa->peer_addr)) {
3877 if (nla_put(skb, IFA_LOCAL, 16, &ifa->addr) < 0 ||
3878 nla_put(skb, IFA_ADDRESS, 16, &ifa->peer_addr) < 0)
3879 goto error;
3880 } else
3881 if (nla_put(skb, IFA_ADDRESS, 16, &ifa->addr) < 0)
3882 goto error;
3883
3884 if (put_cacheinfo(skb, ifa->cstamp, ifa->tstamp, preferred, valid) < 0)
3885 goto error;
3886
3887 return nlmsg_end(skb, nlh);
3888
3889 error:
3890 nlmsg_cancel(skb, nlh);
3891 return -EMSGSIZE;
3892 }
3893
3894 static int inet6_fill_ifmcaddr(struct sk_buff *skb, struct ifmcaddr6 *ifmca,
3895 u32 portid, u32 seq, int event, u16 flags)
3896 {
3897 struct nlmsghdr *nlh;
3898 u8 scope = RT_SCOPE_UNIVERSE;
3899 int ifindex = ifmca->idev->dev->ifindex;
3900
3901 if (ipv6_addr_scope(&ifmca->mca_addr) & IFA_SITE)
3902 scope = RT_SCOPE_SITE;
3903
3904 nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
3905 if (nlh == NULL)
3906 return -EMSGSIZE;
3907
3908 put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3909 if (nla_put(skb, IFA_MULTICAST, 16, &ifmca->mca_addr) < 0 ||
3910 put_cacheinfo(skb, ifmca->mca_cstamp, ifmca->mca_tstamp,
3911 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3912 nlmsg_cancel(skb, nlh);
3913 return -EMSGSIZE;
3914 }
3915
3916 return nlmsg_end(skb, nlh);
3917 }
3918
3919 static int inet6_fill_ifacaddr(struct sk_buff *skb, struct ifacaddr6 *ifaca,
3920 u32 portid, u32 seq, int event, unsigned int flags)
3921 {
3922 struct nlmsghdr *nlh;
3923 u8 scope = RT_SCOPE_UNIVERSE;
3924 int ifindex = ifaca->aca_idev->dev->ifindex;
3925
3926 if (ipv6_addr_scope(&ifaca->aca_addr) & IFA_SITE)
3927 scope = RT_SCOPE_SITE;
3928
3929 nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
3930 if (nlh == NULL)
3931 return -EMSGSIZE;
3932
3933 put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3934 if (nla_put(skb, IFA_ANYCAST, 16, &ifaca->aca_addr) < 0 ||
3935 put_cacheinfo(skb, ifaca->aca_cstamp, ifaca->aca_tstamp,
3936 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3937 nlmsg_cancel(skb, nlh);
3938 return -EMSGSIZE;
3939 }
3940
3941 return nlmsg_end(skb, nlh);
3942 }
3943
3944 enum addr_type_t {
3945 UNICAST_ADDR,
3946 MULTICAST_ADDR,
3947 ANYCAST_ADDR,
3948 };
3949
3950 /* called with rcu_read_lock() */
3951 static int in6_dump_addrs(struct inet6_dev *idev, struct sk_buff *skb,
3952 struct netlink_callback *cb, enum addr_type_t type,
3953 int s_ip_idx, int *p_ip_idx)
3954 {
3955 struct ifmcaddr6 *ifmca;
3956 struct ifacaddr6 *ifaca;
3957 int err = 1;
3958 int ip_idx = *p_ip_idx;
3959
3960 read_lock_bh(&idev->lock);
3961 switch (type) {
3962 case UNICAST_ADDR: {
3963 struct inet6_ifaddr *ifa;
3964
3965 /* unicast address incl. temp addr */
3966 list_for_each_entry(ifa, &idev->addr_list, if_list) {
3967 if (++ip_idx < s_ip_idx)
3968 continue;
3969 err = inet6_fill_ifaddr(skb, ifa,
3970 NETLINK_CB(cb->skb).portid,
3971 cb->nlh->nlmsg_seq,
3972 RTM_NEWADDR,
3973 NLM_F_MULTI);
3974 if (err <= 0)
3975 break;
3976 nl_dump_check_consistent(cb, nlmsg_hdr(skb));
3977 }
3978 break;
3979 }
3980 case MULTICAST_ADDR:
3981 /* multicast address */
3982 for (ifmca = idev->mc_list; ifmca;
3983 ifmca = ifmca->next, ip_idx++) {
3984 if (ip_idx < s_ip_idx)
3985 continue;
3986 err = inet6_fill_ifmcaddr(skb, ifmca,
3987 NETLINK_CB(cb->skb).portid,
3988 cb->nlh->nlmsg_seq,
3989 RTM_GETMULTICAST,
3990 NLM_F_MULTI);
3991 if (err <= 0)
3992 break;
3993 }
3994 break;
3995 case ANYCAST_ADDR:
3996 /* anycast address */
3997 for (ifaca = idev->ac_list; ifaca;
3998 ifaca = ifaca->aca_next, ip_idx++) {
3999 if (ip_idx < s_ip_idx)
4000 continue;
4001 err = inet6_fill_ifacaddr(skb, ifaca,
4002 NETLINK_CB(cb->skb).portid,
4003 cb->nlh->nlmsg_seq,
4004 RTM_GETANYCAST,
4005 NLM_F_MULTI);
4006 if (err <= 0)
4007 break;
4008 }
4009 break;
4010 default:
4011 break;
4012 }
4013 read_unlock_bh(&idev->lock);
4014 *p_ip_idx = ip_idx;
4015 return err;
4016 }
4017
4018 static int inet6_dump_addr(struct sk_buff *skb, struct netlink_callback *cb,
4019 enum addr_type_t type)
4020 {
4021 struct net *net = sock_net(skb->sk);
4022 int h, s_h;
4023 int idx, ip_idx;
4024 int s_idx, s_ip_idx;
4025 struct net_device *dev;
4026 struct inet6_dev *idev;
4027 struct hlist_head *head;
4028
4029 s_h = cb->args[0];
4030 s_idx = idx = cb->args[1];
4031 s_ip_idx = ip_idx = cb->args[2];
4032
4033 rcu_read_lock();
4034 cb->seq = atomic_read(&net->ipv6.dev_addr_genid) ^ net->dev_base_seq;
4035 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
4036 idx = 0;
4037 head = &net->dev_index_head[h];
4038 hlist_for_each_entry_rcu(dev, head, index_hlist) {
4039 if (idx < s_idx)
4040 goto cont;
4041 if (h > s_h || idx > s_idx)
4042 s_ip_idx = 0;
4043 ip_idx = 0;
4044 idev = __in6_dev_get(dev);
4045 if (!idev)
4046 goto cont;
4047
4048 if (in6_dump_addrs(idev, skb, cb, type,
4049 s_ip_idx, &ip_idx) <= 0)
4050 goto done;
4051 cont:
4052 idx++;
4053 }
4054 }
4055 done:
4056 rcu_read_unlock();
4057 cb->args[0] = h;
4058 cb->args[1] = idx;
4059 cb->args[2] = ip_idx;
4060
4061 return skb->len;
4062 }
4063
4064 static int inet6_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
4065 {
4066 enum addr_type_t type = UNICAST_ADDR;
4067
4068 return inet6_dump_addr(skb, cb, type);
4069 }
4070
4071 static int inet6_dump_ifmcaddr(struct sk_buff *skb, struct netlink_callback *cb)
4072 {
4073 enum addr_type_t type = MULTICAST_ADDR;
4074
4075 return inet6_dump_addr(skb, cb, type);
4076 }
4077
4078
4079 static int inet6_dump_ifacaddr(struct sk_buff *skb, struct netlink_callback *cb)
4080 {
4081 enum addr_type_t type = ANYCAST_ADDR;
4082
4083 return inet6_dump_addr(skb, cb, type);
4084 }
4085
4086 static int inet6_rtm_getaddr(struct sk_buff *in_skb, struct nlmsghdr *nlh)
4087 {
4088 struct net *net = sock_net(in_skb->sk);
4089 struct ifaddrmsg *ifm;
4090 struct nlattr *tb[IFA_MAX+1];
4091 struct in6_addr *addr = NULL, *peer;
4092 struct net_device *dev = NULL;
4093 struct inet6_ifaddr *ifa;
4094 struct sk_buff *skb;
4095 int err;
4096
4097 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
4098 if (err < 0)
4099 goto errout;
4100
4101 addr = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer);
4102 if (addr == NULL) {
4103 err = -EINVAL;
4104 goto errout;
4105 }
4106
4107 ifm = nlmsg_data(nlh);
4108 if (ifm->ifa_index)
4109 dev = __dev_get_by_index(net, ifm->ifa_index);
4110
4111 ifa = ipv6_get_ifaddr(net, addr, dev, 1);
4112 if (!ifa) {
4113 err = -EADDRNOTAVAIL;
4114 goto errout;
4115 }
4116
4117 skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_KERNEL);
4118 if (!skb) {
4119 err = -ENOBUFS;
4120 goto errout_ifa;
4121 }
4122
4123 err = inet6_fill_ifaddr(skb, ifa, NETLINK_CB(in_skb).portid,
4124 nlh->nlmsg_seq, RTM_NEWADDR, 0);
4125 if (err < 0) {
4126 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
4127 WARN_ON(err == -EMSGSIZE);
4128 kfree_skb(skb);
4129 goto errout_ifa;
4130 }
4131 err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
4132 errout_ifa:
4133 in6_ifa_put(ifa);
4134 errout:
4135 return err;
4136 }
4137
4138 static void inet6_ifa_notify(int event, struct inet6_ifaddr *ifa)
4139 {
4140 struct sk_buff *skb;
4141 struct net *net = dev_net(ifa->idev->dev);
4142 int err = -ENOBUFS;
4143
4144 skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_ATOMIC);
4145 if (skb == NULL)
4146 goto errout;
4147
4148 err = inet6_fill_ifaddr(skb, ifa, 0, 0, event, 0);
4149 if (err < 0) {
4150 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
4151 WARN_ON(err == -EMSGSIZE);
4152 kfree_skb(skb);
4153 goto errout;
4154 }
4155 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
4156 return;
4157 errout:
4158 if (err < 0)
4159 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
4160 }
4161
4162 static inline void ipv6_store_devconf(struct ipv6_devconf *cnf,
4163 __s32 *array, int bytes)
4164 {
4165 BUG_ON(bytes < (DEVCONF_MAX * 4));
4166
4167 memset(array, 0, bytes);
4168 array[DEVCONF_FORWARDING] = cnf->forwarding;
4169 array[DEVCONF_HOPLIMIT] = cnf->hop_limit;
4170 array[DEVCONF_MTU6] = cnf->mtu6;
4171 array[DEVCONF_ACCEPT_RA] = cnf->accept_ra;
4172 array[DEVCONF_ACCEPT_REDIRECTS] = cnf->accept_redirects;
4173 array[DEVCONF_AUTOCONF] = cnf->autoconf;
4174 array[DEVCONF_DAD_TRANSMITS] = cnf->dad_transmits;
4175 array[DEVCONF_RTR_SOLICITS] = cnf->rtr_solicits;
4176 array[DEVCONF_RTR_SOLICIT_INTERVAL] =
4177 jiffies_to_msecs(cnf->rtr_solicit_interval);
4178 array[DEVCONF_RTR_SOLICIT_DELAY] =
4179 jiffies_to_msecs(cnf->rtr_solicit_delay);
4180 array[DEVCONF_FORCE_MLD_VERSION] = cnf->force_mld_version;
4181 #ifdef CONFIG_IPV6_PRIVACY
4182 array[DEVCONF_USE_TEMPADDR] = cnf->use_tempaddr;
4183 array[DEVCONF_TEMP_VALID_LFT] = cnf->temp_valid_lft;
4184 array[DEVCONF_TEMP_PREFERED_LFT] = cnf->temp_prefered_lft;
4185 array[DEVCONF_REGEN_MAX_RETRY] = cnf->regen_max_retry;
4186 array[DEVCONF_MAX_DESYNC_FACTOR] = cnf->max_desync_factor;
4187 #endif
4188 array[DEVCONF_MAX_ADDRESSES] = cnf->max_addresses;
4189 array[DEVCONF_ACCEPT_RA_DEFRTR] = cnf->accept_ra_defrtr;
4190 array[DEVCONF_ACCEPT_RA_PINFO] = cnf->accept_ra_pinfo;
4191 #ifdef CONFIG_IPV6_ROUTER_PREF
4192 array[DEVCONF_ACCEPT_RA_RTR_PREF] = cnf->accept_ra_rtr_pref;
4193 array[DEVCONF_RTR_PROBE_INTERVAL] =
4194 jiffies_to_msecs(cnf->rtr_probe_interval);
4195 #ifdef CONFIG_IPV6_ROUTE_INFO
4196 array[DEVCONF_ACCEPT_RA_RT_INFO_MAX_PLEN] = cnf->accept_ra_rt_info_max_plen;
4197 #endif
4198 #endif
4199 array[DEVCONF_PROXY_NDP] = cnf->proxy_ndp;
4200 array[DEVCONF_ACCEPT_SOURCE_ROUTE] = cnf->accept_source_route;
4201 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
4202 array[DEVCONF_OPTIMISTIC_DAD] = cnf->optimistic_dad;
4203 #endif
4204 #ifdef CONFIG_IPV6_MROUTE
4205 array[DEVCONF_MC_FORWARDING] = cnf->mc_forwarding;
4206 #endif
4207 array[DEVCONF_DISABLE_IPV6] = cnf->disable_ipv6;
4208 array[DEVCONF_ACCEPT_DAD] = cnf->accept_dad;
4209 array[DEVCONF_FORCE_TLLAO] = cnf->force_tllao;
4210 array[DEVCONF_NDISC_NOTIFY] = cnf->ndisc_notify;
4211 }
4212
4213 static inline size_t inet6_ifla6_size(void)
4214 {
4215 return nla_total_size(4) /* IFLA_INET6_FLAGS */
4216 + nla_total_size(sizeof(struct ifla_cacheinfo))
4217 + nla_total_size(DEVCONF_MAX * 4) /* IFLA_INET6_CONF */
4218 + nla_total_size(IPSTATS_MIB_MAX * 8) /* IFLA_INET6_STATS */
4219 + nla_total_size(ICMP6_MIB_MAX * 8) /* IFLA_INET6_ICMP6STATS */
4220 + nla_total_size(sizeof(struct in6_addr)); /* IFLA_INET6_TOKEN */
4221 }
4222
4223 static inline size_t inet6_if_nlmsg_size(void)
4224 {
4225 return NLMSG_ALIGN(sizeof(struct ifinfomsg))
4226 + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
4227 + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
4228 + nla_total_size(4) /* IFLA_MTU */
4229 + nla_total_size(4) /* IFLA_LINK */
4230 + nla_total_size(inet6_ifla6_size()); /* IFLA_PROTINFO */
4231 }
4232
4233 static inline void __snmp6_fill_statsdev(u64 *stats, atomic_long_t *mib,
4234 int items, int bytes)
4235 {
4236 int i;
4237 int pad = bytes - sizeof(u64) * items;
4238 BUG_ON(pad < 0);
4239
4240 /* Use put_unaligned() because stats may not be aligned for u64. */
4241 put_unaligned(items, &stats[0]);
4242 for (i = 1; i < items; i++)
4243 put_unaligned(atomic_long_read(&mib[i]), &stats[i]);
4244
4245 memset(&stats[items], 0, pad);
4246 }
4247
4248 static inline void __snmp6_fill_stats64(u64 *stats, void __percpu **mib,
4249 int items, int bytes, size_t syncpoff)
4250 {
4251 int i;
4252 int pad = bytes - sizeof(u64) * items;
4253 BUG_ON(pad < 0);
4254
4255 /* Use put_unaligned() because stats may not be aligned for u64. */
4256 put_unaligned(items, &stats[0]);
4257 for (i = 1; i < items; i++)
4258 put_unaligned(snmp_fold_field64(mib, i, syncpoff), &stats[i]);
4259
4260 memset(&stats[items], 0, pad);
4261 }
4262
4263 static void snmp6_fill_stats(u64 *stats, struct inet6_dev *idev, int attrtype,
4264 int bytes)
4265 {
4266 switch (attrtype) {
4267 case IFLA_INET6_STATS:
4268 __snmp6_fill_stats64(stats, (void __percpu **)idev->stats.ipv6,
4269 IPSTATS_MIB_MAX, bytes, offsetof(struct ipstats_mib, syncp));
4270 break;
4271 case IFLA_INET6_ICMP6STATS:
4272 __snmp6_fill_statsdev(stats, idev->stats.icmpv6dev->mibs, ICMP6_MIB_MAX, bytes);
4273 break;
4274 }
4275 }
4276
4277 static int inet6_fill_ifla6_attrs(struct sk_buff *skb, struct inet6_dev *idev)
4278 {
4279 struct nlattr *nla;
4280 struct ifla_cacheinfo ci;
4281
4282 if (nla_put_u32(skb, IFLA_INET6_FLAGS, idev->if_flags))
4283 goto nla_put_failure;
4284 ci.max_reasm_len = IPV6_MAXPLEN;
4285 ci.tstamp = cstamp_delta(idev->tstamp);
4286 ci.reachable_time = jiffies_to_msecs(idev->nd_parms->reachable_time);
4287 ci.retrans_time = jiffies_to_msecs(idev->nd_parms->retrans_time);
4288 if (nla_put(skb, IFLA_INET6_CACHEINFO, sizeof(ci), &ci))
4289 goto nla_put_failure;
4290 nla = nla_reserve(skb, IFLA_INET6_CONF, DEVCONF_MAX * sizeof(s32));
4291 if (nla == NULL)
4292 goto nla_put_failure;
4293 ipv6_store_devconf(&idev->cnf, nla_data(nla), nla_len(nla));
4294
4295 /* XXX - MC not implemented */
4296
4297 nla = nla_reserve(skb, IFLA_INET6_STATS, IPSTATS_MIB_MAX * sizeof(u64));
4298 if (nla == NULL)
4299 goto nla_put_failure;
4300 snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_STATS, nla_len(nla));
4301
4302 nla = nla_reserve(skb, IFLA_INET6_ICMP6STATS, ICMP6_MIB_MAX * sizeof(u64));
4303 if (nla == NULL)
4304 goto nla_put_failure;
4305 snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_ICMP6STATS, nla_len(nla));
4306
4307 nla = nla_reserve(skb, IFLA_INET6_TOKEN, sizeof(struct in6_addr));
4308 if (nla == NULL)
4309 goto nla_put_failure;
4310 read_lock_bh(&idev->lock);
4311 memcpy(nla_data(nla), idev->token.s6_addr, nla_len(nla));
4312 read_unlock_bh(&idev->lock);
4313
4314 return 0;
4315
4316 nla_put_failure:
4317 return -EMSGSIZE;
4318 }
4319
4320 static size_t inet6_get_link_af_size(const struct net_device *dev)
4321 {
4322 if (!__in6_dev_get(dev))
4323 return 0;
4324
4325 return inet6_ifla6_size();
4326 }
4327
4328 static int inet6_fill_link_af(struct sk_buff *skb, const struct net_device *dev)
4329 {
4330 struct inet6_dev *idev = __in6_dev_get(dev);
4331
4332 if (!idev)
4333 return -ENODATA;
4334
4335 if (inet6_fill_ifla6_attrs(skb, idev) < 0)
4336 return -EMSGSIZE;
4337
4338 return 0;
4339 }
4340
4341 static int inet6_set_iftoken(struct inet6_dev *idev, struct in6_addr *token)
4342 {
4343 struct inet6_ifaddr *ifp;
4344 struct net_device *dev = idev->dev;
4345 bool update_rs = false;
4346 struct in6_addr ll_addr;
4347
4348 if (token == NULL)
4349 return -EINVAL;
4350 if (ipv6_addr_any(token))
4351 return -EINVAL;
4352 if (dev->flags & (IFF_LOOPBACK | IFF_NOARP))
4353 return -EINVAL;
4354 if (!ipv6_accept_ra(idev))
4355 return -EINVAL;
4356 if (idev->cnf.rtr_solicits <= 0)
4357 return -EINVAL;
4358
4359 write_lock_bh(&idev->lock);
4360
4361 BUILD_BUG_ON(sizeof(token->s6_addr) != 16);
4362 memcpy(idev->token.s6_addr + 8, token->s6_addr + 8, 8);
4363
4364 write_unlock_bh(&idev->lock);
4365
4366 if (!idev->dead && (idev->if_flags & IF_READY) &&
4367 !ipv6_get_lladdr(dev, &ll_addr, IFA_F_TENTATIVE |
4368 IFA_F_OPTIMISTIC)) {
4369
4370 /* If we're not ready, then normal ifup will take care
4371 * of this. Otherwise, we need to request our rs here.
4372 */
4373 ndisc_send_rs(dev, &ll_addr, &in6addr_linklocal_allrouters);
4374 update_rs = true;
4375 }
4376
4377 write_lock_bh(&idev->lock);
4378
4379 if (update_rs) {
4380 idev->if_flags |= IF_RS_SENT;
4381 idev->rs_probes = 1;
4382 addrconf_mod_rs_timer(idev, idev->cnf.rtr_solicit_interval);
4383 }
4384
4385 /* Well, that's kinda nasty ... */
4386 list_for_each_entry(ifp, &idev->addr_list, if_list) {
4387 spin_lock(&ifp->lock);
4388 if (ifp->tokenized) {
4389 ifp->valid_lft = 0;
4390 ifp->prefered_lft = 0;
4391 }
4392 spin_unlock(&ifp->lock);
4393 }
4394
4395 write_unlock_bh(&idev->lock);
4396 addrconf_verify(0);
4397 return 0;
4398 }
4399
4400 static int inet6_set_link_af(struct net_device *dev, const struct nlattr *nla)
4401 {
4402 int err = -EINVAL;
4403 struct inet6_dev *idev = __in6_dev_get(dev);
4404 struct nlattr *tb[IFLA_INET6_MAX + 1];
4405
4406 if (!idev)
4407 return -EAFNOSUPPORT;
4408
4409 if (nla_parse_nested(tb, IFLA_INET6_MAX, nla, NULL) < 0)
4410 BUG();
4411
4412 if (tb[IFLA_INET6_TOKEN])
4413 err = inet6_set_iftoken(idev, nla_data(tb[IFLA_INET6_TOKEN]));
4414
4415 return err;
4416 }
4417
4418 static int inet6_fill_ifinfo(struct sk_buff *skb, struct inet6_dev *idev,
4419 u32 portid, u32 seq, int event, unsigned int flags)
4420 {
4421 struct net_device *dev = idev->dev;
4422 struct ifinfomsg *hdr;
4423 struct nlmsghdr *nlh;
4424 void *protoinfo;
4425
4426 nlh = nlmsg_put(skb, portid, seq, event, sizeof(*hdr), flags);
4427 if (nlh == NULL)
4428 return -EMSGSIZE;
4429
4430 hdr = nlmsg_data(nlh);
4431 hdr->ifi_family = AF_INET6;
4432 hdr->__ifi_pad = 0;
4433 hdr->ifi_type = dev->type;
4434 hdr->ifi_index = dev->ifindex;
4435 hdr->ifi_flags = dev_get_flags(dev);
4436 hdr->ifi_change = 0;
4437
4438 if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
4439 (dev->addr_len &&
4440 nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
4441 nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
4442 (dev->ifindex != dev->iflink &&
4443 nla_put_u32(skb, IFLA_LINK, dev->iflink)))
4444 goto nla_put_failure;
4445 protoinfo = nla_nest_start(skb, IFLA_PROTINFO);
4446 if (protoinfo == NULL)
4447 goto nla_put_failure;
4448
4449 if (inet6_fill_ifla6_attrs(skb, idev) < 0)
4450 goto nla_put_failure;
4451
4452 nla_nest_end(skb, protoinfo);
4453 return nlmsg_end(skb, nlh);
4454
4455 nla_put_failure:
4456 nlmsg_cancel(skb, nlh);
4457 return -EMSGSIZE;
4458 }
4459
4460 static int inet6_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
4461 {
4462 struct net *net = sock_net(skb->sk);
4463 int h, s_h;
4464 int idx = 0, s_idx;
4465 struct net_device *dev;
4466 struct inet6_dev *idev;
4467 struct hlist_head *head;
4468
4469 s_h = cb->args[0];
4470 s_idx = cb->args[1];
4471
4472 rcu_read_lock();
4473 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
4474 idx = 0;
4475 head = &net->dev_index_head[h];
4476 hlist_for_each_entry_rcu(dev, head, index_hlist) {
4477 if (idx < s_idx)
4478 goto cont;
4479 idev = __in6_dev_get(dev);
4480 if (!idev)
4481 goto cont;
4482 if (inet6_fill_ifinfo(skb, idev,
4483 NETLINK_CB(cb->skb).portid,
4484 cb->nlh->nlmsg_seq,
4485 RTM_NEWLINK, NLM_F_MULTI) <= 0)
4486 goto out;
4487 cont:
4488 idx++;
4489 }
4490 }
4491 out:
4492 rcu_read_unlock();
4493 cb->args[1] = idx;
4494 cb->args[0] = h;
4495
4496 return skb->len;
4497 }
4498
4499 void inet6_ifinfo_notify(int event, struct inet6_dev *idev)
4500 {
4501 struct sk_buff *skb;
4502 struct net *net = dev_net(idev->dev);
4503 int err = -ENOBUFS;
4504
4505 skb = nlmsg_new(inet6_if_nlmsg_size(), GFP_ATOMIC);
4506 if (skb == NULL)
4507 goto errout;
4508
4509 err = inet6_fill_ifinfo(skb, idev, 0, 0, event, 0);
4510 if (err < 0) {
4511 /* -EMSGSIZE implies BUG in inet6_if_nlmsg_size() */
4512 WARN_ON(err == -EMSGSIZE);
4513 kfree_skb(skb);
4514 goto errout;
4515 }
4516 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFINFO, NULL, GFP_ATOMIC);
4517 return;
4518 errout:
4519 if (err < 0)
4520 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFINFO, err);
4521 }
4522
4523 static inline size_t inet6_prefix_nlmsg_size(void)
4524 {
4525 return NLMSG_ALIGN(sizeof(struct prefixmsg))
4526 + nla_total_size(sizeof(struct in6_addr))
4527 + nla_total_size(sizeof(struct prefix_cacheinfo));
4528 }
4529
4530 static int inet6_fill_prefix(struct sk_buff *skb, struct inet6_dev *idev,
4531 struct prefix_info *pinfo, u32 portid, u32 seq,
4532 int event, unsigned int flags)
4533 {
4534 struct prefixmsg *pmsg;
4535 struct nlmsghdr *nlh;
4536 struct prefix_cacheinfo ci;
4537
4538 nlh = nlmsg_put(skb, portid, seq, event, sizeof(*pmsg), flags);
4539 if (nlh == NULL)
4540 return -EMSGSIZE;
4541
4542 pmsg = nlmsg_data(nlh);
4543 pmsg->prefix_family = AF_INET6;
4544 pmsg->prefix_pad1 = 0;
4545 pmsg->prefix_pad2 = 0;
4546 pmsg->prefix_ifindex = idev->dev->ifindex;
4547 pmsg->prefix_len = pinfo->prefix_len;
4548 pmsg->prefix_type = pinfo->type;
4549 pmsg->prefix_pad3 = 0;
4550 pmsg->prefix_flags = 0;
4551 if (pinfo->onlink)
4552 pmsg->prefix_flags |= IF_PREFIX_ONLINK;
4553 if (pinfo->autoconf)
4554 pmsg->prefix_flags |= IF_PREFIX_AUTOCONF;
4555
4556 if (nla_put(skb, PREFIX_ADDRESS, sizeof(pinfo->prefix), &pinfo->prefix))
4557 goto nla_put_failure;
4558 ci.preferred_time = ntohl(pinfo->prefered);
4559 ci.valid_time = ntohl(pinfo->valid);
4560 if (nla_put(skb, PREFIX_CACHEINFO, sizeof(ci), &ci))
4561 goto nla_put_failure;
4562 return nlmsg_end(skb, nlh);
4563
4564 nla_put_failure:
4565 nlmsg_cancel(skb, nlh);
4566 return -EMSGSIZE;
4567 }
4568
4569 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
4570 struct prefix_info *pinfo)
4571 {
4572 struct sk_buff *skb;
4573 struct net *net = dev_net(idev->dev);
4574 int err = -ENOBUFS;
4575
4576 skb = nlmsg_new(inet6_prefix_nlmsg_size(), GFP_ATOMIC);
4577 if (skb == NULL)
4578 goto errout;
4579
4580 err = inet6_fill_prefix(skb, idev, pinfo, 0, 0, event, 0);
4581 if (err < 0) {
4582 /* -EMSGSIZE implies BUG in inet6_prefix_nlmsg_size() */
4583 WARN_ON(err == -EMSGSIZE);
4584 kfree_skb(skb);
4585 goto errout;
4586 }
4587 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_PREFIX, NULL, GFP_ATOMIC);
4588 return;
4589 errout:
4590 if (err < 0)
4591 rtnl_set_sk_err(net, RTNLGRP_IPV6_PREFIX, err);
4592 }
4593
4594 static void update_valid_ll_addr_cnt(struct inet6_ifaddr *ifp, int count)
4595 {
4596 write_lock_bh(&ifp->idev->lock);
4597 spin_lock(&ifp->lock);
4598 if (((ifp->flags & (IFA_F_PERMANENT|IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|
4599 IFA_F_DADFAILED)) == IFA_F_PERMANENT) &&
4600 (ipv6_addr_type(&ifp->addr) & IPV6_ADDR_LINKLOCAL))
4601 ifp->idev->valid_ll_addr_cnt += count;
4602 WARN_ON(ifp->idev->valid_ll_addr_cnt < 0);
4603 spin_unlock(&ifp->lock);
4604 write_unlock_bh(&ifp->idev->lock);
4605 }
4606
4607 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
4608 {
4609 struct net *net = dev_net(ifp->idev->dev);
4610
4611 inet6_ifa_notify(event ? : RTM_NEWADDR, ifp);
4612
4613 switch (event) {
4614 case RTM_NEWADDR:
4615 update_valid_ll_addr_cnt(ifp, 1);
4616
4617 /*
4618 * If the address was optimistic
4619 * we inserted the route at the start of
4620 * our DAD process, so we don't need
4621 * to do it again
4622 */
4623 if (!(ifp->rt->rt6i_node))
4624 ip6_ins_rt(ifp->rt);
4625 if (ifp->idev->cnf.forwarding)
4626 addrconf_join_anycast(ifp);
4627 if (!ipv6_addr_any(&ifp->peer_addr))
4628 addrconf_prefix_route(&ifp->peer_addr, 128,
4629 ifp->idev->dev, 0, 0);
4630 break;
4631 case RTM_DELADDR:
4632 update_valid_ll_addr_cnt(ifp, -1);
4633
4634 if (ifp->idev->cnf.forwarding)
4635 addrconf_leave_anycast(ifp);
4636 addrconf_leave_solict(ifp->idev, &ifp->addr);
4637 if (!ipv6_addr_any(&ifp->peer_addr)) {
4638 struct rt6_info *rt;
4639 struct net_device *dev = ifp->idev->dev;
4640
4641 rt = rt6_lookup(dev_net(dev), &ifp->peer_addr, NULL,
4642 dev->ifindex, 1);
4643 if (rt) {
4644 dst_hold(&rt->dst);
4645 if (ip6_del_rt(rt))
4646 dst_free(&rt->dst);
4647 }
4648 }
4649 dst_hold(&ifp->rt->dst);
4650
4651 if (ip6_del_rt(ifp->rt))
4652 dst_free(&ifp->rt->dst);
4653 break;
4654 }
4655 atomic_inc(&net->ipv6.dev_addr_genid);
4656 }
4657
4658 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
4659 {
4660 rcu_read_lock_bh();
4661 if (likely(ifp->idev->dead == 0))
4662 __ipv6_ifa_notify(event, ifp);
4663 rcu_read_unlock_bh();
4664 }
4665
4666 #ifdef CONFIG_SYSCTL
4667
4668 static
4669 int addrconf_sysctl_forward(struct ctl_table *ctl, int write,
4670 void __user *buffer, size_t *lenp, loff_t *ppos)
4671 {
4672 int *valp = ctl->data;
4673 int val = *valp;
4674 loff_t pos = *ppos;
4675 struct ctl_table lctl;
4676 int ret;
4677
4678 /*
4679 * ctl->data points to idev->cnf.forwarding, we should
4680 * not modify it until we get the rtnl lock.
4681 */
4682 lctl = *ctl;
4683 lctl.data = &val;
4684
4685 ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
4686
4687 if (write)
4688 ret = addrconf_fixup_forwarding(ctl, valp, val);
4689 if (ret)
4690 *ppos = pos;
4691 return ret;
4692 }
4693
4694 static void dev_disable_change(struct inet6_dev *idev)
4695 {
4696 struct netdev_notifier_info info;
4697
4698 if (!idev || !idev->dev)
4699 return;
4700
4701 netdev_notifier_info_init(&info, idev->dev);
4702 if (idev->cnf.disable_ipv6)
4703 addrconf_notify(NULL, NETDEV_DOWN, &info);
4704 else
4705 addrconf_notify(NULL, NETDEV_UP, &info);
4706 }
4707
4708 static void addrconf_disable_change(struct net *net, __s32 newf)
4709 {
4710 struct net_device *dev;
4711 struct inet6_dev *idev;
4712
4713 rcu_read_lock();
4714 for_each_netdev_rcu(net, dev) {
4715 idev = __in6_dev_get(dev);
4716 if (idev) {
4717 int changed = (!idev->cnf.disable_ipv6) ^ (!newf);
4718 idev->cnf.disable_ipv6 = newf;
4719 if (changed)
4720 dev_disable_change(idev);
4721 }
4722 }
4723 rcu_read_unlock();
4724 }
4725
4726 static int addrconf_disable_ipv6(struct ctl_table *table, int *p, int newf)
4727 {
4728 struct net *net;
4729 int old;
4730
4731 if (!rtnl_trylock())
4732 return restart_syscall();
4733
4734 net = (struct net *)table->extra2;
4735 old = *p;
4736 *p = newf;
4737
4738 if (p == &net->ipv6.devconf_dflt->disable_ipv6) {
4739 rtnl_unlock();
4740 return 0;
4741 }
4742
4743 if (p == &net->ipv6.devconf_all->disable_ipv6) {
4744 net->ipv6.devconf_dflt->disable_ipv6 = newf;
4745 addrconf_disable_change(net, newf);
4746 } else if ((!newf) ^ (!old))
4747 dev_disable_change((struct inet6_dev *)table->extra1);
4748
4749 rtnl_unlock();
4750 return 0;
4751 }
4752
4753 static
4754 int addrconf_sysctl_disable(struct ctl_table *ctl, int write,
4755 void __user *buffer, size_t *lenp, loff_t *ppos)
4756 {
4757 int *valp = ctl->data;
4758 int val = *valp;
4759 loff_t pos = *ppos;
4760 struct ctl_table lctl;
4761 int ret;
4762
4763 /*
4764 * ctl->data points to idev->cnf.disable_ipv6, we should
4765 * not modify it until we get the rtnl lock.
4766 */
4767 lctl = *ctl;
4768 lctl.data = &val;
4769
4770 ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
4771
4772 if (write)
4773 ret = addrconf_disable_ipv6(ctl, valp, val);
4774 if (ret)
4775 *ppos = pos;
4776 return ret;
4777 }
4778
4779 static struct addrconf_sysctl_table
4780 {
4781 struct ctl_table_header *sysctl_header;
4782 struct ctl_table addrconf_vars[DEVCONF_MAX+1];
4783 } addrconf_sysctl __read_mostly = {
4784 .sysctl_header = NULL,
4785 .addrconf_vars = {
4786 {
4787 .procname = "forwarding",
4788 .data = &ipv6_devconf.forwarding,
4789 .maxlen = sizeof(int),
4790 .mode = 0644,
4791 .proc_handler = addrconf_sysctl_forward,
4792 },
4793 {
4794 .procname = "hop_limit",
4795 .data = &ipv6_devconf.hop_limit,
4796 .maxlen = sizeof(int),
4797 .mode = 0644,
4798 .proc_handler = proc_dointvec,
4799 },
4800 {
4801 .procname = "mtu",
4802 .data = &ipv6_devconf.mtu6,
4803 .maxlen = sizeof(int),
4804 .mode = 0644,
4805 .proc_handler = proc_dointvec,
4806 },
4807 {
4808 .procname = "accept_ra",
4809 .data = &ipv6_devconf.accept_ra,
4810 .maxlen = sizeof(int),
4811 .mode = 0644,
4812 .proc_handler = proc_dointvec,
4813 },
4814 {
4815 .procname = "accept_redirects",
4816 .data = &ipv6_devconf.accept_redirects,
4817 .maxlen = sizeof(int),
4818 .mode = 0644,
4819 .proc_handler = proc_dointvec,
4820 },
4821 {
4822 .procname = "autoconf",
4823 .data = &ipv6_devconf.autoconf,
4824 .maxlen = sizeof(int),
4825 .mode = 0644,
4826 .proc_handler = proc_dointvec,
4827 },
4828 {
4829 .procname = "dad_transmits",
4830 .data = &ipv6_devconf.dad_transmits,
4831 .maxlen = sizeof(int),
4832 .mode = 0644,
4833 .proc_handler = proc_dointvec,
4834 },
4835 {
4836 .procname = "router_solicitations",
4837 .data = &ipv6_devconf.rtr_solicits,
4838 .maxlen = sizeof(int),
4839 .mode = 0644,
4840 .proc_handler = proc_dointvec,
4841 },
4842 {
4843 .procname = "router_solicitation_interval",
4844 .data = &ipv6_devconf.rtr_solicit_interval,
4845 .maxlen = sizeof(int),
4846 .mode = 0644,
4847 .proc_handler = proc_dointvec_jiffies,
4848 },
4849 {
4850 .procname = "router_solicitation_delay",
4851 .data = &ipv6_devconf.rtr_solicit_delay,
4852 .maxlen = sizeof(int),
4853 .mode = 0644,
4854 .proc_handler = proc_dointvec_jiffies,
4855 },
4856 {
4857 .procname = "force_mld_version",
4858 .data = &ipv6_devconf.force_mld_version,
4859 .maxlen = sizeof(int),
4860 .mode = 0644,
4861 .proc_handler = proc_dointvec,
4862 },
4863 #ifdef CONFIG_IPV6_PRIVACY
4864 {
4865 .procname = "use_tempaddr",
4866 .data = &ipv6_devconf.use_tempaddr,
4867 .maxlen = sizeof(int),
4868 .mode = 0644,
4869 .proc_handler = proc_dointvec,
4870 },
4871 {
4872 .procname = "temp_valid_lft",
4873 .data = &ipv6_devconf.temp_valid_lft,
4874 .maxlen = sizeof(int),
4875 .mode = 0644,
4876 .proc_handler = proc_dointvec,
4877 },
4878 {
4879 .procname = "temp_prefered_lft",
4880 .data = &ipv6_devconf.temp_prefered_lft,
4881 .maxlen = sizeof(int),
4882 .mode = 0644,
4883 .proc_handler = proc_dointvec,
4884 },
4885 {
4886 .procname = "regen_max_retry",
4887 .data = &ipv6_devconf.regen_max_retry,
4888 .maxlen = sizeof(int),
4889 .mode = 0644,
4890 .proc_handler = proc_dointvec,
4891 },
4892 {
4893 .procname = "max_desync_factor",
4894 .data = &ipv6_devconf.max_desync_factor,
4895 .maxlen = sizeof(int),
4896 .mode = 0644,
4897 .proc_handler = proc_dointvec,
4898 },
4899 #endif
4900 {
4901 .procname = "max_addresses",
4902 .data = &ipv6_devconf.max_addresses,
4903 .maxlen = sizeof(int),
4904 .mode = 0644,
4905 .proc_handler = proc_dointvec,
4906 },
4907 {
4908 .procname = "accept_ra_defrtr",
4909 .data = &ipv6_devconf.accept_ra_defrtr,
4910 .maxlen = sizeof(int),
4911 .mode = 0644,
4912 .proc_handler = proc_dointvec,
4913 },
4914 {
4915 .procname = "accept_ra_pinfo",
4916 .data = &ipv6_devconf.accept_ra_pinfo,
4917 .maxlen = sizeof(int),
4918 .mode = 0644,
4919 .proc_handler = proc_dointvec,
4920 },
4921 #ifdef CONFIG_IPV6_ROUTER_PREF
4922 {
4923 .procname = "accept_ra_rtr_pref",
4924 .data = &ipv6_devconf.accept_ra_rtr_pref,
4925 .maxlen = sizeof(int),
4926 .mode = 0644,
4927 .proc_handler = proc_dointvec,
4928 },
4929 {
4930 .procname = "router_probe_interval",
4931 .data = &ipv6_devconf.rtr_probe_interval,
4932 .maxlen = sizeof(int),
4933 .mode = 0644,
4934 .proc_handler = proc_dointvec_jiffies,
4935 },
4936 #ifdef CONFIG_IPV6_ROUTE_INFO
4937 {
4938 .procname = "accept_ra_rt_info_max_plen",
4939 .data = &ipv6_devconf.accept_ra_rt_info_max_plen,
4940 .maxlen = sizeof(int),
4941 .mode = 0644,
4942 .proc_handler = proc_dointvec,
4943 },
4944 #endif
4945 #endif
4946 {
4947 .procname = "proxy_ndp",
4948 .data = &ipv6_devconf.proxy_ndp,
4949 .maxlen = sizeof(int),
4950 .mode = 0644,
4951 .proc_handler = proc_dointvec,
4952 },
4953 {
4954 .procname = "accept_source_route",
4955 .data = &ipv6_devconf.accept_source_route,
4956 .maxlen = sizeof(int),
4957 .mode = 0644,
4958 .proc_handler = proc_dointvec,
4959 },
4960 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
4961 {
4962 .procname = "optimistic_dad",
4963 .data = &ipv6_devconf.optimistic_dad,
4964 .maxlen = sizeof(int),
4965 .mode = 0644,
4966 .proc_handler = proc_dointvec,
4967
4968 },
4969 #endif
4970 #ifdef CONFIG_IPV6_MROUTE
4971 {
4972 .procname = "mc_forwarding",
4973 .data = &ipv6_devconf.mc_forwarding,
4974 .maxlen = sizeof(int),
4975 .mode = 0444,
4976 .proc_handler = proc_dointvec,
4977 },
4978 #endif
4979 {
4980 .procname = "disable_ipv6",
4981 .data = &ipv6_devconf.disable_ipv6,
4982 .maxlen = sizeof(int),
4983 .mode = 0644,
4984 .proc_handler = addrconf_sysctl_disable,
4985 },
4986 {
4987 .procname = "accept_dad",
4988 .data = &ipv6_devconf.accept_dad,
4989 .maxlen = sizeof(int),
4990 .mode = 0644,
4991 .proc_handler = proc_dointvec,
4992 },
4993 {
4994 .procname = "force_tllao",
4995 .data = &ipv6_devconf.force_tllao,
4996 .maxlen = sizeof(int),
4997 .mode = 0644,
4998 .proc_handler = proc_dointvec
4999 },
5000 {
5001 .procname = "ndisc_notify",
5002 .data = &ipv6_devconf.ndisc_notify,
5003 .maxlen = sizeof(int),
5004 .mode = 0644,
5005 .proc_handler = proc_dointvec
5006 },
5007 {
5008 /* sentinel */
5009 }
5010 },
5011 };
5012
5013 static int __addrconf_sysctl_register(struct net *net, char *dev_name,
5014 struct inet6_dev *idev, struct ipv6_devconf *p)
5015 {
5016 int i;
5017 struct addrconf_sysctl_table *t;
5018 char path[sizeof("net/ipv6/conf/") + IFNAMSIZ];
5019
5020 t = kmemdup(&addrconf_sysctl, sizeof(*t), GFP_KERNEL);
5021 if (t == NULL)
5022 goto out;
5023
5024 for (i = 0; t->addrconf_vars[i].data; i++) {
5025 t->addrconf_vars[i].data += (char *)p - (char *)&ipv6_devconf;
5026 t->addrconf_vars[i].extra1 = idev; /* embedded; no ref */
5027 t->addrconf_vars[i].extra2 = net;
5028 }
5029
5030 snprintf(path, sizeof(path), "net/ipv6/conf/%s", dev_name);
5031
5032 t->sysctl_header = register_net_sysctl(net, path, t->addrconf_vars);
5033 if (t->sysctl_header == NULL)
5034 goto free;
5035
5036 p->sysctl = t;
5037 return 0;
5038
5039 free:
5040 kfree(t);
5041 out:
5042 return -ENOBUFS;
5043 }
5044
5045 static void __addrconf_sysctl_unregister(struct ipv6_devconf *p)
5046 {
5047 struct addrconf_sysctl_table *t;
5048
5049 if (p->sysctl == NULL)
5050 return;
5051
5052 t = p->sysctl;
5053 p->sysctl = NULL;
5054 unregister_net_sysctl_table(t->sysctl_header);
5055 kfree(t);
5056 }
5057
5058 static void addrconf_sysctl_register(struct inet6_dev *idev)
5059 {
5060 neigh_sysctl_register(idev->dev, idev->nd_parms, "ipv6",
5061 &ndisc_ifinfo_sysctl_change);
5062 __addrconf_sysctl_register(dev_net(idev->dev), idev->dev->name,
5063 idev, &idev->cnf);
5064 }
5065
5066 static void addrconf_sysctl_unregister(struct inet6_dev *idev)
5067 {
5068 __addrconf_sysctl_unregister(&idev->cnf);
5069 neigh_sysctl_unregister(idev->nd_parms);
5070 }
5071
5072
5073 #endif
5074
5075 static int __net_init addrconf_init_net(struct net *net)
5076 {
5077 int err = -ENOMEM;
5078 struct ipv6_devconf *all, *dflt;
5079
5080 all = kmemdup(&ipv6_devconf, sizeof(ipv6_devconf), GFP_KERNEL);
5081 if (all == NULL)
5082 goto err_alloc_all;
5083
5084 dflt = kmemdup(&ipv6_devconf_dflt, sizeof(ipv6_devconf_dflt), GFP_KERNEL);
5085 if (dflt == NULL)
5086 goto err_alloc_dflt;
5087
5088 /* these will be inherited by all namespaces */
5089 dflt->autoconf = ipv6_defaults.autoconf;
5090 dflt->disable_ipv6 = ipv6_defaults.disable_ipv6;
5091
5092 net->ipv6.devconf_all = all;
5093 net->ipv6.devconf_dflt = dflt;
5094
5095 #ifdef CONFIG_SYSCTL
5096 err = __addrconf_sysctl_register(net, "all", NULL, all);
5097 if (err < 0)
5098 goto err_reg_all;
5099
5100 err = __addrconf_sysctl_register(net, "default", NULL, dflt);
5101 if (err < 0)
5102 goto err_reg_dflt;
5103 #endif
5104 return 0;
5105
5106 #ifdef CONFIG_SYSCTL
5107 err_reg_dflt:
5108 __addrconf_sysctl_unregister(all);
5109 err_reg_all:
5110 kfree(dflt);
5111 #endif
5112 err_alloc_dflt:
5113 kfree(all);
5114 err_alloc_all:
5115 return err;
5116 }
5117
5118 static void __net_exit addrconf_exit_net(struct net *net)
5119 {
5120 #ifdef CONFIG_SYSCTL
5121 __addrconf_sysctl_unregister(net->ipv6.devconf_dflt);
5122 __addrconf_sysctl_unregister(net->ipv6.devconf_all);
5123 #endif
5124 if (!net_eq(net, &init_net)) {
5125 kfree(net->ipv6.devconf_dflt);
5126 kfree(net->ipv6.devconf_all);
5127 }
5128 }
5129
5130 static struct pernet_operations addrconf_ops = {
5131 .init = addrconf_init_net,
5132 .exit = addrconf_exit_net,
5133 };
5134
5135 static struct rtnl_af_ops inet6_ops = {
5136 .family = AF_INET6,
5137 .fill_link_af = inet6_fill_link_af,
5138 .get_link_af_size = inet6_get_link_af_size,
5139 .set_link_af = inet6_set_link_af,
5140 };
5141
5142 /*
5143 * Init / cleanup code
5144 */
5145
5146 int __init addrconf_init(void)
5147 {
5148 int i, err;
5149
5150 err = ipv6_addr_label_init();
5151 if (err < 0) {
5152 pr_crit("%s: cannot initialize default policy table: %d\n",
5153 __func__, err);
5154 goto out;
5155 }
5156
5157 err = register_pernet_subsys(&addrconf_ops);
5158 if (err < 0)
5159 goto out_addrlabel;
5160
5161 /* The addrconf netdev notifier requires that loopback_dev
5162 * has it's ipv6 private information allocated and setup
5163 * before it can bring up and give link-local addresses
5164 * to other devices which are up.
5165 *
5166 * Unfortunately, loopback_dev is not necessarily the first
5167 * entry in the global dev_base list of net devices. In fact,
5168 * it is likely to be the very last entry on that list.
5169 * So this causes the notifier registry below to try and
5170 * give link-local addresses to all devices besides loopback_dev
5171 * first, then loopback_dev, which cases all the non-loopback_dev
5172 * devices to fail to get a link-local address.
5173 *
5174 * So, as a temporary fix, allocate the ipv6 structure for
5175 * loopback_dev first by hand.
5176 * Longer term, all of the dependencies ipv6 has upon the loopback
5177 * device and it being up should be removed.
5178 */
5179 rtnl_lock();
5180 if (!ipv6_add_dev(init_net.loopback_dev))
5181 err = -ENOMEM;
5182 rtnl_unlock();
5183 if (err)
5184 goto errlo;
5185
5186 for (i = 0; i < IN6_ADDR_HSIZE; i++)
5187 INIT_HLIST_HEAD(&inet6_addr_lst[i]);
5188
5189 register_netdevice_notifier(&ipv6_dev_notf);
5190
5191 addrconf_verify(0);
5192
5193 err = rtnl_af_register(&inet6_ops);
5194 if (err < 0)
5195 goto errout_af;
5196
5197 err = __rtnl_register(PF_INET6, RTM_GETLINK, NULL, inet6_dump_ifinfo,
5198 NULL);
5199 if (err < 0)
5200 goto errout;
5201
5202 /* Only the first call to __rtnl_register can fail */
5203 __rtnl_register(PF_INET6, RTM_NEWADDR, inet6_rtm_newaddr, NULL, NULL);
5204 __rtnl_register(PF_INET6, RTM_DELADDR, inet6_rtm_deladdr, NULL, NULL);
5205 __rtnl_register(PF_INET6, RTM_GETADDR, inet6_rtm_getaddr,
5206 inet6_dump_ifaddr, NULL);
5207 __rtnl_register(PF_INET6, RTM_GETMULTICAST, NULL,
5208 inet6_dump_ifmcaddr, NULL);
5209 __rtnl_register(PF_INET6, RTM_GETANYCAST, NULL,
5210 inet6_dump_ifacaddr, NULL);
5211 __rtnl_register(PF_INET6, RTM_GETNETCONF, inet6_netconf_get_devconf,
5212 inet6_netconf_dump_devconf, NULL);
5213
5214 ipv6_addr_label_rtnl_register();
5215
5216 return 0;
5217 errout:
5218 rtnl_af_unregister(&inet6_ops);
5219 errout_af:
5220 unregister_netdevice_notifier(&ipv6_dev_notf);
5221 errlo:
5222 unregister_pernet_subsys(&addrconf_ops);
5223 out_addrlabel:
5224 ipv6_addr_label_cleanup();
5225 out:
5226 return err;
5227 }
5228
5229 void addrconf_cleanup(void)
5230 {
5231 struct net_device *dev;
5232 int i;
5233
5234 unregister_netdevice_notifier(&ipv6_dev_notf);
5235 unregister_pernet_subsys(&addrconf_ops);
5236 ipv6_addr_label_cleanup();
5237
5238 rtnl_lock();
5239
5240 __rtnl_af_unregister(&inet6_ops);
5241
5242 /* clean dev list */
5243 for_each_netdev(&init_net, dev) {
5244 if (__in6_dev_get(dev) == NULL)
5245 continue;
5246 addrconf_ifdown(dev, 1);
5247 }
5248 addrconf_ifdown(init_net.loopback_dev, 2);
5249
5250 /*
5251 * Check hash table.
5252 */
5253 spin_lock_bh(&addrconf_hash_lock);
5254 for (i = 0; i < IN6_ADDR_HSIZE; i++)
5255 WARN_ON(!hlist_empty(&inet6_addr_lst[i]));
5256 spin_unlock_bh(&addrconf_hash_lock);
5257
5258 del_timer(&addr_chk_timer);
5259 rtnl_unlock();
5260 }
Linus' kernel tree