Merge tag 'pm+acpi-3.10-rc5' of git://git.kernel.org/pub/scm/linux/kernel/git/rafael...
[GitHub/mt8127/android_kernel_alcatel_ttab.git] / net / ipv6 / addrconf.c
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_timer(struct inet6_ifaddr *ifp)
257 {
258 if (del_timer(&ifp->timer))
259 __in6_ifa_put(ifp);
260 }
261
262 enum addrconf_timer_t {
263 AC_NONE,
264 AC_DAD,
265 AC_RS,
266 };
267
268 static void addrconf_mod_timer(struct inet6_ifaddr *ifp,
269 enum addrconf_timer_t what,
270 unsigned long when)
271 {
272 if (!del_timer(&ifp->timer))
273 in6_ifa_hold(ifp);
274
275 switch (what) {
276 case AC_DAD:
277 ifp->timer.function = addrconf_dad_timer;
278 break;
279 case AC_RS:
280 ifp->timer.function = addrconf_rs_timer;
281 break;
282 default:
283 break;
284 }
285 ifp->timer.expires = jiffies + when;
286 add_timer(&ifp->timer);
287 }
288
289 static int snmp6_alloc_dev(struct inet6_dev *idev)
290 {
291 if (snmp_mib_init((void __percpu **)idev->stats.ipv6,
292 sizeof(struct ipstats_mib),
293 __alignof__(struct ipstats_mib)) < 0)
294 goto err_ip;
295 idev->stats.icmpv6dev = kzalloc(sizeof(struct icmpv6_mib_device),
296 GFP_KERNEL);
297 if (!idev->stats.icmpv6dev)
298 goto err_icmp;
299 idev->stats.icmpv6msgdev = kzalloc(sizeof(struct icmpv6msg_mib_device),
300 GFP_KERNEL);
301 if (!idev->stats.icmpv6msgdev)
302 goto err_icmpmsg;
303
304 return 0;
305
306 err_icmpmsg:
307 kfree(idev->stats.icmpv6dev);
308 err_icmp:
309 snmp_mib_free((void __percpu **)idev->stats.ipv6);
310 err_ip:
311 return -ENOMEM;
312 }
313
314 static void snmp6_free_dev(struct inet6_dev *idev)
315 {
316 kfree(idev->stats.icmpv6msgdev);
317 kfree(idev->stats.icmpv6dev);
318 snmp_mib_free((void __percpu **)idev->stats.ipv6);
319 }
320
321 /* Nobody refers to this device, we may destroy it. */
322
323 void in6_dev_finish_destroy(struct inet6_dev *idev)
324 {
325 struct net_device *dev = idev->dev;
326
327 WARN_ON(!list_empty(&idev->addr_list));
328 WARN_ON(idev->mc_list != NULL);
329
330 #ifdef NET_REFCNT_DEBUG
331 pr_debug("%s: %s\n", __func__, dev ? dev->name : "NIL");
332 #endif
333 dev_put(dev);
334 if (!idev->dead) {
335 pr_warn("Freeing alive inet6 device %p\n", idev);
336 return;
337 }
338 snmp6_free_dev(idev);
339 kfree_rcu(idev, rcu);
340 }
341 EXPORT_SYMBOL(in6_dev_finish_destroy);
342
343 static struct inet6_dev *ipv6_add_dev(struct net_device *dev)
344 {
345 struct inet6_dev *ndev;
346
347 ASSERT_RTNL();
348
349 if (dev->mtu < IPV6_MIN_MTU)
350 return NULL;
351
352 ndev = kzalloc(sizeof(struct inet6_dev), GFP_KERNEL);
353
354 if (ndev == NULL)
355 return NULL;
356
357 rwlock_init(&ndev->lock);
358 ndev->dev = dev;
359 INIT_LIST_HEAD(&ndev->addr_list);
360
361 memcpy(&ndev->cnf, dev_net(dev)->ipv6.devconf_dflt, sizeof(ndev->cnf));
362 ndev->cnf.mtu6 = dev->mtu;
363 ndev->cnf.sysctl = NULL;
364 ndev->nd_parms = neigh_parms_alloc(dev, &nd_tbl);
365 if (ndev->nd_parms == NULL) {
366 kfree(ndev);
367 return NULL;
368 }
369 if (ndev->cnf.forwarding)
370 dev_disable_lro(dev);
371 /* We refer to the device */
372 dev_hold(dev);
373
374 if (snmp6_alloc_dev(ndev) < 0) {
375 ADBG((KERN_WARNING
376 "%s: cannot allocate memory for statistics; dev=%s.\n",
377 __func__, dev->name));
378 neigh_parms_release(&nd_tbl, ndev->nd_parms);
379 dev_put(dev);
380 kfree(ndev);
381 return NULL;
382 }
383
384 if (snmp6_register_dev(ndev) < 0) {
385 ADBG((KERN_WARNING
386 "%s: cannot create /proc/net/dev_snmp6/%s\n",
387 __func__, dev->name));
388 neigh_parms_release(&nd_tbl, ndev->nd_parms);
389 ndev->dead = 1;
390 in6_dev_finish_destroy(ndev);
391 return NULL;
392 }
393
394 /* One reference from device. We must do this before
395 * we invoke __ipv6_regen_rndid().
396 */
397 in6_dev_hold(ndev);
398
399 if (dev->flags & (IFF_NOARP | IFF_LOOPBACK))
400 ndev->cnf.accept_dad = -1;
401
402 #if IS_ENABLED(CONFIG_IPV6_SIT)
403 if (dev->type == ARPHRD_SIT && (dev->priv_flags & IFF_ISATAP)) {
404 pr_info("%s: Disabled Multicast RS\n", dev->name);
405 ndev->cnf.rtr_solicits = 0;
406 }
407 #endif
408
409 #ifdef CONFIG_IPV6_PRIVACY
410 INIT_LIST_HEAD(&ndev->tempaddr_list);
411 setup_timer(&ndev->regen_timer, ipv6_regen_rndid, (unsigned long)ndev);
412 if ((dev->flags&IFF_LOOPBACK) ||
413 dev->type == ARPHRD_TUNNEL ||
414 dev->type == ARPHRD_TUNNEL6 ||
415 dev->type == ARPHRD_SIT ||
416 dev->type == ARPHRD_NONE) {
417 ndev->cnf.use_tempaddr = -1;
418 } else {
419 in6_dev_hold(ndev);
420 ipv6_regen_rndid((unsigned long) ndev);
421 }
422 #endif
423 ndev->token = in6addr_any;
424
425 if (netif_running(dev) && addrconf_qdisc_ok(dev))
426 ndev->if_flags |= IF_READY;
427
428 ipv6_mc_init_dev(ndev);
429 ndev->tstamp = jiffies;
430 addrconf_sysctl_register(ndev);
431 /* protected by rtnl_lock */
432 rcu_assign_pointer(dev->ip6_ptr, ndev);
433
434 /* Join interface-local all-node multicast group */
435 ipv6_dev_mc_inc(dev, &in6addr_interfacelocal_allnodes);
436
437 /* Join all-node multicast group */
438 ipv6_dev_mc_inc(dev, &in6addr_linklocal_allnodes);
439
440 /* Join all-router multicast group if forwarding is set */
441 if (ndev->cnf.forwarding && (dev->flags & IFF_MULTICAST))
442 ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
443
444 return ndev;
445 }
446
447 static struct inet6_dev *ipv6_find_idev(struct net_device *dev)
448 {
449 struct inet6_dev *idev;
450
451 ASSERT_RTNL();
452
453 idev = __in6_dev_get(dev);
454 if (!idev) {
455 idev = ipv6_add_dev(dev);
456 if (!idev)
457 return NULL;
458 }
459
460 if (dev->flags&IFF_UP)
461 ipv6_mc_up(idev);
462 return idev;
463 }
464
465 static int inet6_netconf_msgsize_devconf(int type)
466 {
467 int size = NLMSG_ALIGN(sizeof(struct netconfmsg))
468 + nla_total_size(4); /* NETCONFA_IFINDEX */
469
470 /* type -1 is used for ALL */
471 if (type == -1 || type == NETCONFA_FORWARDING)
472 size += nla_total_size(4);
473 #ifdef CONFIG_IPV6_MROUTE
474 if (type == -1 || type == NETCONFA_MC_FORWARDING)
475 size += nla_total_size(4);
476 #endif
477
478 return size;
479 }
480
481 static int inet6_netconf_fill_devconf(struct sk_buff *skb, int ifindex,
482 struct ipv6_devconf *devconf, u32 portid,
483 u32 seq, int event, unsigned int flags,
484 int type)
485 {
486 struct nlmsghdr *nlh;
487 struct netconfmsg *ncm;
488
489 nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct netconfmsg),
490 flags);
491 if (nlh == NULL)
492 return -EMSGSIZE;
493
494 ncm = nlmsg_data(nlh);
495 ncm->ncm_family = AF_INET6;
496
497 if (nla_put_s32(skb, NETCONFA_IFINDEX, ifindex) < 0)
498 goto nla_put_failure;
499
500 /* type -1 is used for ALL */
501 if ((type == -1 || type == NETCONFA_FORWARDING) &&
502 nla_put_s32(skb, NETCONFA_FORWARDING, devconf->forwarding) < 0)
503 goto nla_put_failure;
504 #ifdef CONFIG_IPV6_MROUTE
505 if ((type == -1 || type == NETCONFA_MC_FORWARDING) &&
506 nla_put_s32(skb, NETCONFA_MC_FORWARDING,
507 devconf->mc_forwarding) < 0)
508 goto nla_put_failure;
509 #endif
510 return nlmsg_end(skb, nlh);
511
512 nla_put_failure:
513 nlmsg_cancel(skb, nlh);
514 return -EMSGSIZE;
515 }
516
517 void inet6_netconf_notify_devconf(struct net *net, int type, int ifindex,
518 struct ipv6_devconf *devconf)
519 {
520 struct sk_buff *skb;
521 int err = -ENOBUFS;
522
523 skb = nlmsg_new(inet6_netconf_msgsize_devconf(type), GFP_ATOMIC);
524 if (skb == NULL)
525 goto errout;
526
527 err = inet6_netconf_fill_devconf(skb, ifindex, devconf, 0, 0,
528 RTM_NEWNETCONF, 0, type);
529 if (err < 0) {
530 /* -EMSGSIZE implies BUG in inet6_netconf_msgsize_devconf() */
531 WARN_ON(err == -EMSGSIZE);
532 kfree_skb(skb);
533 goto errout;
534 }
535 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_NETCONF, NULL, GFP_ATOMIC);
536 return;
537 errout:
538 rtnl_set_sk_err(net, RTNLGRP_IPV6_NETCONF, err);
539 }
540
541 static const struct nla_policy devconf_ipv6_policy[NETCONFA_MAX+1] = {
542 [NETCONFA_IFINDEX] = { .len = sizeof(int) },
543 [NETCONFA_FORWARDING] = { .len = sizeof(int) },
544 };
545
546 static int inet6_netconf_get_devconf(struct sk_buff *in_skb,
547 struct nlmsghdr *nlh)
548 {
549 struct net *net = sock_net(in_skb->sk);
550 struct nlattr *tb[NETCONFA_MAX+1];
551 struct netconfmsg *ncm;
552 struct sk_buff *skb;
553 struct ipv6_devconf *devconf;
554 struct inet6_dev *in6_dev;
555 struct net_device *dev;
556 int ifindex;
557 int err;
558
559 err = nlmsg_parse(nlh, sizeof(*ncm), tb, NETCONFA_MAX,
560 devconf_ipv6_policy);
561 if (err < 0)
562 goto errout;
563
564 err = EINVAL;
565 if (!tb[NETCONFA_IFINDEX])
566 goto errout;
567
568 ifindex = nla_get_s32(tb[NETCONFA_IFINDEX]);
569 switch (ifindex) {
570 case NETCONFA_IFINDEX_ALL:
571 devconf = net->ipv6.devconf_all;
572 break;
573 case NETCONFA_IFINDEX_DEFAULT:
574 devconf = net->ipv6.devconf_dflt;
575 break;
576 default:
577 dev = __dev_get_by_index(net, ifindex);
578 if (dev == NULL)
579 goto errout;
580 in6_dev = __in6_dev_get(dev);
581 if (in6_dev == NULL)
582 goto errout;
583 devconf = &in6_dev->cnf;
584 break;
585 }
586
587 err = -ENOBUFS;
588 skb = nlmsg_new(inet6_netconf_msgsize_devconf(-1), GFP_ATOMIC);
589 if (skb == NULL)
590 goto errout;
591
592 err = inet6_netconf_fill_devconf(skb, ifindex, devconf,
593 NETLINK_CB(in_skb).portid,
594 nlh->nlmsg_seq, RTM_NEWNETCONF, 0,
595 -1);
596 if (err < 0) {
597 /* -EMSGSIZE implies BUG in inet6_netconf_msgsize_devconf() */
598 WARN_ON(err == -EMSGSIZE);
599 kfree_skb(skb);
600 goto errout;
601 }
602 err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
603 errout:
604 return err;
605 }
606
607 static int inet6_netconf_dump_devconf(struct sk_buff *skb,
608 struct netlink_callback *cb)
609 {
610 struct net *net = sock_net(skb->sk);
611 int h, s_h;
612 int idx, s_idx;
613 struct net_device *dev;
614 struct inet6_dev *idev;
615 struct hlist_head *head;
616
617 s_h = cb->args[0];
618 s_idx = idx = cb->args[1];
619
620 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
621 idx = 0;
622 head = &net->dev_index_head[h];
623 rcu_read_lock();
624 cb->seq = atomic_read(&net->ipv6.dev_addr_genid) ^
625 net->dev_base_seq;
626 hlist_for_each_entry_rcu(dev, head, index_hlist) {
627 if (idx < s_idx)
628 goto cont;
629 idev = __in6_dev_get(dev);
630 if (!idev)
631 goto cont;
632
633 if (inet6_netconf_fill_devconf(skb, dev->ifindex,
634 &idev->cnf,
635 NETLINK_CB(cb->skb).portid,
636 cb->nlh->nlmsg_seq,
637 RTM_NEWNETCONF,
638 NLM_F_MULTI,
639 -1) <= 0) {
640 rcu_read_unlock();
641 goto done;
642 }
643 nl_dump_check_consistent(cb, nlmsg_hdr(skb));
644 cont:
645 idx++;
646 }
647 rcu_read_unlock();
648 }
649 if (h == NETDEV_HASHENTRIES) {
650 if (inet6_netconf_fill_devconf(skb, NETCONFA_IFINDEX_ALL,
651 net->ipv6.devconf_all,
652 NETLINK_CB(cb->skb).portid,
653 cb->nlh->nlmsg_seq,
654 RTM_NEWNETCONF, NLM_F_MULTI,
655 -1) <= 0)
656 goto done;
657 else
658 h++;
659 }
660 if (h == NETDEV_HASHENTRIES + 1) {
661 if (inet6_netconf_fill_devconf(skb, NETCONFA_IFINDEX_DEFAULT,
662 net->ipv6.devconf_dflt,
663 NETLINK_CB(cb->skb).portid,
664 cb->nlh->nlmsg_seq,
665 RTM_NEWNETCONF, NLM_F_MULTI,
666 -1) <= 0)
667 goto done;
668 else
669 h++;
670 }
671 done:
672 cb->args[0] = h;
673 cb->args[1] = idx;
674
675 return skb->len;
676 }
677
678 #ifdef CONFIG_SYSCTL
679 static void dev_forward_change(struct inet6_dev *idev)
680 {
681 struct net_device *dev;
682 struct inet6_ifaddr *ifa;
683
684 if (!idev)
685 return;
686 dev = idev->dev;
687 if (idev->cnf.forwarding)
688 dev_disable_lro(dev);
689 if (dev->flags & IFF_MULTICAST) {
690 if (idev->cnf.forwarding) {
691 ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
692 ipv6_dev_mc_inc(dev, &in6addr_interfacelocal_allrouters);
693 ipv6_dev_mc_inc(dev, &in6addr_sitelocal_allrouters);
694 } else {
695 ipv6_dev_mc_dec(dev, &in6addr_linklocal_allrouters);
696 ipv6_dev_mc_dec(dev, &in6addr_interfacelocal_allrouters);
697 ipv6_dev_mc_dec(dev, &in6addr_sitelocal_allrouters);
698 }
699 }
700
701 list_for_each_entry(ifa, &idev->addr_list, if_list) {
702 if (ifa->flags&IFA_F_TENTATIVE)
703 continue;
704 if (idev->cnf.forwarding)
705 addrconf_join_anycast(ifa);
706 else
707 addrconf_leave_anycast(ifa);
708 }
709 inet6_netconf_notify_devconf(dev_net(dev), NETCONFA_FORWARDING,
710 dev->ifindex, &idev->cnf);
711 }
712
713
714 static void addrconf_forward_change(struct net *net, __s32 newf)
715 {
716 struct net_device *dev;
717 struct inet6_dev *idev;
718
719 for_each_netdev(net, dev) {
720 idev = __in6_dev_get(dev);
721 if (idev) {
722 int changed = (!idev->cnf.forwarding) ^ (!newf);
723 idev->cnf.forwarding = newf;
724 if (changed)
725 dev_forward_change(idev);
726 }
727 }
728 }
729
730 static int addrconf_fixup_forwarding(struct ctl_table *table, int *p, int newf)
731 {
732 struct net *net;
733 int old;
734
735 if (!rtnl_trylock())
736 return restart_syscall();
737
738 net = (struct net *)table->extra2;
739 old = *p;
740 *p = newf;
741
742 if (p == &net->ipv6.devconf_dflt->forwarding) {
743 if ((!newf) ^ (!old))
744 inet6_netconf_notify_devconf(net, NETCONFA_FORWARDING,
745 NETCONFA_IFINDEX_DEFAULT,
746 net->ipv6.devconf_dflt);
747 rtnl_unlock();
748 return 0;
749 }
750
751 if (p == &net->ipv6.devconf_all->forwarding) {
752 net->ipv6.devconf_dflt->forwarding = newf;
753 addrconf_forward_change(net, newf);
754 if ((!newf) ^ (!old))
755 inet6_netconf_notify_devconf(net, NETCONFA_FORWARDING,
756 NETCONFA_IFINDEX_ALL,
757 net->ipv6.devconf_all);
758 } else if ((!newf) ^ (!old))
759 dev_forward_change((struct inet6_dev *)table->extra1);
760 rtnl_unlock();
761
762 if (newf)
763 rt6_purge_dflt_routers(net);
764 return 1;
765 }
766 #endif
767
768 /* Nobody refers to this ifaddr, destroy it */
769 void inet6_ifa_finish_destroy(struct inet6_ifaddr *ifp)
770 {
771 WARN_ON(!hlist_unhashed(&ifp->addr_lst));
772
773 #ifdef NET_REFCNT_DEBUG
774 pr_debug("%s\n", __func__);
775 #endif
776
777 in6_dev_put(ifp->idev);
778
779 if (del_timer(&ifp->timer))
780 pr_notice("Timer is still running, when freeing ifa=%p\n", ifp);
781
782 if (ifp->state != INET6_IFADDR_STATE_DEAD) {
783 pr_warn("Freeing alive inet6 address %p\n", ifp);
784 return;
785 }
786 ip6_rt_put(ifp->rt);
787
788 kfree_rcu(ifp, rcu);
789 }
790
791 static void
792 ipv6_link_dev_addr(struct inet6_dev *idev, struct inet6_ifaddr *ifp)
793 {
794 struct list_head *p;
795 int ifp_scope = ipv6_addr_src_scope(&ifp->addr);
796
797 /*
798 * Each device address list is sorted in order of scope -
799 * global before linklocal.
800 */
801 list_for_each(p, &idev->addr_list) {
802 struct inet6_ifaddr *ifa
803 = list_entry(p, struct inet6_ifaddr, if_list);
804 if (ifp_scope >= ipv6_addr_src_scope(&ifa->addr))
805 break;
806 }
807
808 list_add_tail(&ifp->if_list, p);
809 }
810
811 static u32 inet6_addr_hash(const struct in6_addr *addr)
812 {
813 return hash_32(ipv6_addr_hash(addr), IN6_ADDR_HSIZE_SHIFT);
814 }
815
816 /* On success it returns ifp with increased reference count */
817
818 static struct inet6_ifaddr *
819 ipv6_add_addr(struct inet6_dev *idev, const struct in6_addr *addr, int pfxlen,
820 int scope, u32 flags)
821 {
822 struct inet6_ifaddr *ifa = NULL;
823 struct rt6_info *rt;
824 unsigned int hash;
825 int err = 0;
826 int addr_type = ipv6_addr_type(addr);
827
828 if (addr_type == IPV6_ADDR_ANY ||
829 addr_type & IPV6_ADDR_MULTICAST ||
830 (!(idev->dev->flags & IFF_LOOPBACK) &&
831 addr_type & IPV6_ADDR_LOOPBACK))
832 return ERR_PTR(-EADDRNOTAVAIL);
833
834 rcu_read_lock_bh();
835 if (idev->dead) {
836 err = -ENODEV; /*XXX*/
837 goto out2;
838 }
839
840 if (idev->cnf.disable_ipv6) {
841 err = -EACCES;
842 goto out2;
843 }
844
845 spin_lock(&addrconf_hash_lock);
846
847 /* Ignore adding duplicate addresses on an interface */
848 if (ipv6_chk_same_addr(dev_net(idev->dev), addr, idev->dev)) {
849 ADBG(("ipv6_add_addr: already assigned\n"));
850 err = -EEXIST;
851 goto out;
852 }
853
854 ifa = kzalloc(sizeof(struct inet6_ifaddr), GFP_ATOMIC);
855
856 if (ifa == NULL) {
857 ADBG(("ipv6_add_addr: malloc failed\n"));
858 err = -ENOBUFS;
859 goto out;
860 }
861
862 rt = addrconf_dst_alloc(idev, addr, false);
863 if (IS_ERR(rt)) {
864 err = PTR_ERR(rt);
865 goto out;
866 }
867
868 ifa->addr = *addr;
869
870 spin_lock_init(&ifa->lock);
871 spin_lock_init(&ifa->state_lock);
872 init_timer(&ifa->timer);
873 INIT_HLIST_NODE(&ifa->addr_lst);
874 ifa->timer.data = (unsigned long) ifa;
875 ifa->scope = scope;
876 ifa->prefix_len = pfxlen;
877 ifa->flags = flags | IFA_F_TENTATIVE;
878 ifa->cstamp = ifa->tstamp = jiffies;
879 ifa->tokenized = false;
880
881 ifa->rt = rt;
882
883 ifa->idev = idev;
884 in6_dev_hold(idev);
885 /* For caller */
886 in6_ifa_hold(ifa);
887
888 /* Add to big hash table */
889 hash = inet6_addr_hash(addr);
890
891 hlist_add_head_rcu(&ifa->addr_lst, &inet6_addr_lst[hash]);
892 spin_unlock(&addrconf_hash_lock);
893
894 write_lock(&idev->lock);
895 /* Add to inet6_dev unicast addr list. */
896 ipv6_link_dev_addr(idev, ifa);
897
898 #ifdef CONFIG_IPV6_PRIVACY
899 if (ifa->flags&IFA_F_TEMPORARY) {
900 list_add(&ifa->tmp_list, &idev->tempaddr_list);
901 in6_ifa_hold(ifa);
902 }
903 #endif
904
905 in6_ifa_hold(ifa);
906 write_unlock(&idev->lock);
907 out2:
908 rcu_read_unlock_bh();
909
910 if (likely(err == 0))
911 inet6addr_notifier_call_chain(NETDEV_UP, ifa);
912 else {
913 kfree(ifa);
914 ifa = ERR_PTR(err);
915 }
916
917 return ifa;
918 out:
919 spin_unlock(&addrconf_hash_lock);
920 goto out2;
921 }
922
923 /* This function wants to get referenced ifp and releases it before return */
924
925 static void ipv6_del_addr(struct inet6_ifaddr *ifp)
926 {
927 struct inet6_ifaddr *ifa, *ifn;
928 struct inet6_dev *idev = ifp->idev;
929 int state;
930 int deleted = 0, onlink = 0;
931 unsigned long expires = jiffies;
932
933 spin_lock_bh(&ifp->state_lock);
934 state = ifp->state;
935 ifp->state = INET6_IFADDR_STATE_DEAD;
936 spin_unlock_bh(&ifp->state_lock);
937
938 if (state == INET6_IFADDR_STATE_DEAD)
939 goto out;
940
941 spin_lock_bh(&addrconf_hash_lock);
942 hlist_del_init_rcu(&ifp->addr_lst);
943 spin_unlock_bh(&addrconf_hash_lock);
944
945 write_lock_bh(&idev->lock);
946 #ifdef CONFIG_IPV6_PRIVACY
947 if (ifp->flags&IFA_F_TEMPORARY) {
948 list_del(&ifp->tmp_list);
949 if (ifp->ifpub) {
950 in6_ifa_put(ifp->ifpub);
951 ifp->ifpub = NULL;
952 }
953 __in6_ifa_put(ifp);
954 }
955 #endif
956
957 list_for_each_entry_safe(ifa, ifn, &idev->addr_list, if_list) {
958 if (ifa == ifp) {
959 list_del_init(&ifp->if_list);
960 __in6_ifa_put(ifp);
961
962 if (!(ifp->flags & IFA_F_PERMANENT) || onlink > 0)
963 break;
964 deleted = 1;
965 continue;
966 } else if (ifp->flags & IFA_F_PERMANENT) {
967 if (ipv6_prefix_equal(&ifa->addr, &ifp->addr,
968 ifp->prefix_len)) {
969 if (ifa->flags & IFA_F_PERMANENT) {
970 onlink = 1;
971 if (deleted)
972 break;
973 } else {
974 unsigned long lifetime;
975
976 if (!onlink)
977 onlink = -1;
978
979 spin_lock(&ifa->lock);
980
981 lifetime = addrconf_timeout_fixup(ifa->valid_lft, HZ);
982 /*
983 * Note: Because this address is
984 * not permanent, lifetime <
985 * LONG_MAX / HZ here.
986 */
987 if (time_before(expires,
988 ifa->tstamp + lifetime * HZ))
989 expires = ifa->tstamp + lifetime * HZ;
990 spin_unlock(&ifa->lock);
991 }
992 }
993 }
994 }
995 write_unlock_bh(&idev->lock);
996
997 addrconf_del_timer(ifp);
998
999 ipv6_ifa_notify(RTM_DELADDR, ifp);
1000
1001 inet6addr_notifier_call_chain(NETDEV_DOWN, ifp);
1002
1003 /*
1004 * Purge or update corresponding prefix
1005 *
1006 * 1) we don't purge prefix here if address was not permanent.
1007 * prefix is managed by its own lifetime.
1008 * 2) if there're no addresses, delete prefix.
1009 * 3) if there're still other permanent address(es),
1010 * corresponding prefix is still permanent.
1011 * 4) otherwise, update prefix lifetime to the
1012 * longest valid lifetime among the corresponding
1013 * addresses on the device.
1014 * Note: subsequent RA will update lifetime.
1015 *
1016 * --yoshfuji
1017 */
1018 if ((ifp->flags & IFA_F_PERMANENT) && onlink < 1) {
1019 struct in6_addr prefix;
1020 struct rt6_info *rt;
1021
1022 ipv6_addr_prefix(&prefix, &ifp->addr, ifp->prefix_len);
1023
1024 rt = addrconf_get_prefix_route(&prefix,
1025 ifp->prefix_len,
1026 ifp->idev->dev,
1027 0, RTF_GATEWAY | RTF_DEFAULT);
1028
1029 if (rt) {
1030 if (onlink == 0) {
1031 ip6_del_rt(rt);
1032 rt = NULL;
1033 } else if (!(rt->rt6i_flags & RTF_EXPIRES)) {
1034 rt6_set_expires(rt, expires);
1035 }
1036 }
1037 ip6_rt_put(rt);
1038 }
1039
1040 /* clean up prefsrc entries */
1041 rt6_remove_prefsrc(ifp);
1042 out:
1043 in6_ifa_put(ifp);
1044 }
1045
1046 #ifdef CONFIG_IPV6_PRIVACY
1047 static int ipv6_create_tempaddr(struct inet6_ifaddr *ifp, struct inet6_ifaddr *ift)
1048 {
1049 struct inet6_dev *idev = ifp->idev;
1050 struct in6_addr addr, *tmpaddr;
1051 unsigned long tmp_prefered_lft, tmp_valid_lft, tmp_tstamp, age;
1052 unsigned long regen_advance;
1053 int tmp_plen;
1054 int ret = 0;
1055 int max_addresses;
1056 u32 addr_flags;
1057 unsigned long now = jiffies;
1058
1059 write_lock(&idev->lock);
1060 if (ift) {
1061 spin_lock_bh(&ift->lock);
1062 memcpy(&addr.s6_addr[8], &ift->addr.s6_addr[8], 8);
1063 spin_unlock_bh(&ift->lock);
1064 tmpaddr = &addr;
1065 } else {
1066 tmpaddr = NULL;
1067 }
1068 retry:
1069 in6_dev_hold(idev);
1070 if (idev->cnf.use_tempaddr <= 0) {
1071 write_unlock(&idev->lock);
1072 pr_info("%s: use_tempaddr is disabled\n", __func__);
1073 in6_dev_put(idev);
1074 ret = -1;
1075 goto out;
1076 }
1077 spin_lock_bh(&ifp->lock);
1078 if (ifp->regen_count++ >= idev->cnf.regen_max_retry) {
1079 idev->cnf.use_tempaddr = -1; /*XXX*/
1080 spin_unlock_bh(&ifp->lock);
1081 write_unlock(&idev->lock);
1082 pr_warn("%s: regeneration time exceeded - disabled temporary address support\n",
1083 __func__);
1084 in6_dev_put(idev);
1085 ret = -1;
1086 goto out;
1087 }
1088 in6_ifa_hold(ifp);
1089 memcpy(addr.s6_addr, ifp->addr.s6_addr, 8);
1090 __ipv6_try_regen_rndid(idev, tmpaddr);
1091 memcpy(&addr.s6_addr[8], idev->rndid, 8);
1092 age = (now - ifp->tstamp) / HZ;
1093 tmp_valid_lft = min_t(__u32,
1094 ifp->valid_lft,
1095 idev->cnf.temp_valid_lft + age);
1096 tmp_prefered_lft = min_t(__u32,
1097 ifp->prefered_lft,
1098 idev->cnf.temp_prefered_lft + age -
1099 idev->cnf.max_desync_factor);
1100 tmp_plen = ifp->prefix_len;
1101 max_addresses = idev->cnf.max_addresses;
1102 tmp_tstamp = ifp->tstamp;
1103 spin_unlock_bh(&ifp->lock);
1104
1105 regen_advance = idev->cnf.regen_max_retry *
1106 idev->cnf.dad_transmits *
1107 idev->nd_parms->retrans_time / HZ;
1108 write_unlock(&idev->lock);
1109
1110 /* A temporary address is created only if this calculated Preferred
1111 * Lifetime is greater than REGEN_ADVANCE time units. In particular,
1112 * an implementation must not create a temporary address with a zero
1113 * Preferred Lifetime.
1114 */
1115 if (tmp_prefered_lft <= regen_advance) {
1116 in6_ifa_put(ifp);
1117 in6_dev_put(idev);
1118 ret = -1;
1119 goto out;
1120 }
1121
1122 addr_flags = IFA_F_TEMPORARY;
1123 /* set in addrconf_prefix_rcv() */
1124 if (ifp->flags & IFA_F_OPTIMISTIC)
1125 addr_flags |= IFA_F_OPTIMISTIC;
1126
1127 ift = !max_addresses ||
1128 ipv6_count_addresses(idev) < max_addresses ?
1129 ipv6_add_addr(idev, &addr, tmp_plen,
1130 ipv6_addr_type(&addr)&IPV6_ADDR_SCOPE_MASK,
1131 addr_flags) : NULL;
1132 if (IS_ERR_OR_NULL(ift)) {
1133 in6_ifa_put(ifp);
1134 in6_dev_put(idev);
1135 pr_info("%s: retry temporary address regeneration\n", __func__);
1136 tmpaddr = &addr;
1137 write_lock(&idev->lock);
1138 goto retry;
1139 }
1140
1141 spin_lock_bh(&ift->lock);
1142 ift->ifpub = ifp;
1143 ift->valid_lft = tmp_valid_lft;
1144 ift->prefered_lft = tmp_prefered_lft;
1145 ift->cstamp = now;
1146 ift->tstamp = tmp_tstamp;
1147 spin_unlock_bh(&ift->lock);
1148
1149 addrconf_dad_start(ift);
1150 in6_ifa_put(ift);
1151 in6_dev_put(idev);
1152 out:
1153 return ret;
1154 }
1155 #endif
1156
1157 /*
1158 * Choose an appropriate source address (RFC3484)
1159 */
1160 enum {
1161 IPV6_SADDR_RULE_INIT = 0,
1162 IPV6_SADDR_RULE_LOCAL,
1163 IPV6_SADDR_RULE_SCOPE,
1164 IPV6_SADDR_RULE_PREFERRED,
1165 #ifdef CONFIG_IPV6_MIP6
1166 IPV6_SADDR_RULE_HOA,
1167 #endif
1168 IPV6_SADDR_RULE_OIF,
1169 IPV6_SADDR_RULE_LABEL,
1170 #ifdef CONFIG_IPV6_PRIVACY
1171 IPV6_SADDR_RULE_PRIVACY,
1172 #endif
1173 IPV6_SADDR_RULE_ORCHID,
1174 IPV6_SADDR_RULE_PREFIX,
1175 IPV6_SADDR_RULE_MAX
1176 };
1177
1178 struct ipv6_saddr_score {
1179 int rule;
1180 int addr_type;
1181 struct inet6_ifaddr *ifa;
1182 DECLARE_BITMAP(scorebits, IPV6_SADDR_RULE_MAX);
1183 int scopedist;
1184 int matchlen;
1185 };
1186
1187 struct ipv6_saddr_dst {
1188 const struct in6_addr *addr;
1189 int ifindex;
1190 int scope;
1191 int label;
1192 unsigned int prefs;
1193 };
1194
1195 static inline int ipv6_saddr_preferred(int type)
1196 {
1197 if (type & (IPV6_ADDR_MAPPED|IPV6_ADDR_COMPATv4|IPV6_ADDR_LOOPBACK))
1198 return 1;
1199 return 0;
1200 }
1201
1202 static int ipv6_get_saddr_eval(struct net *net,
1203 struct ipv6_saddr_score *score,
1204 struct ipv6_saddr_dst *dst,
1205 int i)
1206 {
1207 int ret;
1208
1209 if (i <= score->rule) {
1210 switch (i) {
1211 case IPV6_SADDR_RULE_SCOPE:
1212 ret = score->scopedist;
1213 break;
1214 case IPV6_SADDR_RULE_PREFIX:
1215 ret = score->matchlen;
1216 break;
1217 default:
1218 ret = !!test_bit(i, score->scorebits);
1219 }
1220 goto out;
1221 }
1222
1223 switch (i) {
1224 case IPV6_SADDR_RULE_INIT:
1225 /* Rule 0: remember if hiscore is not ready yet */
1226 ret = !!score->ifa;
1227 break;
1228 case IPV6_SADDR_RULE_LOCAL:
1229 /* Rule 1: Prefer same address */
1230 ret = ipv6_addr_equal(&score->ifa->addr, dst->addr);
1231 break;
1232 case IPV6_SADDR_RULE_SCOPE:
1233 /* Rule 2: Prefer appropriate scope
1234 *
1235 * ret
1236 * ^
1237 * -1 | d 15
1238 * ---+--+-+---> scope
1239 * |
1240 * | d is scope of the destination.
1241 * B-d | \
1242 * | \ <- smaller scope is better if
1243 * B-15 | \ if scope is enough for destinaion.
1244 * | ret = B - scope (-1 <= scope >= d <= 15).
1245 * d-C-1 | /
1246 * |/ <- greater is better
1247 * -C / if scope is not enough for destination.
1248 * /| ret = scope - C (-1 <= d < scope <= 15).
1249 *
1250 * d - C - 1 < B -15 (for all -1 <= d <= 15).
1251 * C > d + 14 - B >= 15 + 14 - B = 29 - B.
1252 * Assume B = 0 and we get C > 29.
1253 */
1254 ret = __ipv6_addr_src_scope(score->addr_type);
1255 if (ret >= dst->scope)
1256 ret = -ret;
1257 else
1258 ret -= 128; /* 30 is enough */
1259 score->scopedist = ret;
1260 break;
1261 case IPV6_SADDR_RULE_PREFERRED:
1262 /* Rule 3: Avoid deprecated and optimistic addresses */
1263 ret = ipv6_saddr_preferred(score->addr_type) ||
1264 !(score->ifa->flags & (IFA_F_DEPRECATED|IFA_F_OPTIMISTIC));
1265 break;
1266 #ifdef CONFIG_IPV6_MIP6
1267 case IPV6_SADDR_RULE_HOA:
1268 {
1269 /* Rule 4: Prefer home address */
1270 int prefhome = !(dst->prefs & IPV6_PREFER_SRC_COA);
1271 ret = !(score->ifa->flags & IFA_F_HOMEADDRESS) ^ prefhome;
1272 break;
1273 }
1274 #endif
1275 case IPV6_SADDR_RULE_OIF:
1276 /* Rule 5: Prefer outgoing interface */
1277 ret = (!dst->ifindex ||
1278 dst->ifindex == score->ifa->idev->dev->ifindex);
1279 break;
1280 case IPV6_SADDR_RULE_LABEL:
1281 /* Rule 6: Prefer matching label */
1282 ret = ipv6_addr_label(net,
1283 &score->ifa->addr, score->addr_type,
1284 score->ifa->idev->dev->ifindex) == dst->label;
1285 break;
1286 #ifdef CONFIG_IPV6_PRIVACY
1287 case IPV6_SADDR_RULE_PRIVACY:
1288 {
1289 /* Rule 7: Prefer public address
1290 * Note: prefer temporary address if use_tempaddr >= 2
1291 */
1292 int preftmp = dst->prefs & (IPV6_PREFER_SRC_PUBLIC|IPV6_PREFER_SRC_TMP) ?
1293 !!(dst->prefs & IPV6_PREFER_SRC_TMP) :
1294 score->ifa->idev->cnf.use_tempaddr >= 2;
1295 ret = (!(score->ifa->flags & IFA_F_TEMPORARY)) ^ preftmp;
1296 break;
1297 }
1298 #endif
1299 case IPV6_SADDR_RULE_ORCHID:
1300 /* Rule 8-: Prefer ORCHID vs ORCHID or
1301 * non-ORCHID vs non-ORCHID
1302 */
1303 ret = !(ipv6_addr_orchid(&score->ifa->addr) ^
1304 ipv6_addr_orchid(dst->addr));
1305 break;
1306 case IPV6_SADDR_RULE_PREFIX:
1307 /* Rule 8: Use longest matching prefix */
1308 ret = ipv6_addr_diff(&score->ifa->addr, dst->addr);
1309 if (ret > score->ifa->prefix_len)
1310 ret = score->ifa->prefix_len;
1311 score->matchlen = ret;
1312 break;
1313 default:
1314 ret = 0;
1315 }
1316
1317 if (ret)
1318 __set_bit(i, score->scorebits);
1319 score->rule = i;
1320 out:
1321 return ret;
1322 }
1323
1324 int ipv6_dev_get_saddr(struct net *net, const struct net_device *dst_dev,
1325 const struct in6_addr *daddr, unsigned int prefs,
1326 struct in6_addr *saddr)
1327 {
1328 struct ipv6_saddr_score scores[2],
1329 *score = &scores[0], *hiscore = &scores[1];
1330 struct ipv6_saddr_dst dst;
1331 struct net_device *dev;
1332 int dst_type;
1333
1334 dst_type = __ipv6_addr_type(daddr);
1335 dst.addr = daddr;
1336 dst.ifindex = dst_dev ? dst_dev->ifindex : 0;
1337 dst.scope = __ipv6_addr_src_scope(dst_type);
1338 dst.label = ipv6_addr_label(net, daddr, dst_type, dst.ifindex);
1339 dst.prefs = prefs;
1340
1341 hiscore->rule = -1;
1342 hiscore->ifa = NULL;
1343
1344 rcu_read_lock();
1345
1346 for_each_netdev_rcu(net, dev) {
1347 struct inet6_dev *idev;
1348
1349 /* Candidate Source Address (section 4)
1350 * - multicast and link-local destination address,
1351 * the set of candidate source address MUST only
1352 * include addresses assigned to interfaces
1353 * belonging to the same link as the outgoing
1354 * interface.
1355 * (- For site-local destination addresses, the
1356 * set of candidate source addresses MUST only
1357 * include addresses assigned to interfaces
1358 * belonging to the same site as the outgoing
1359 * interface.)
1360 */
1361 if (((dst_type & IPV6_ADDR_MULTICAST) ||
1362 dst.scope <= IPV6_ADDR_SCOPE_LINKLOCAL) &&
1363 dst.ifindex && dev->ifindex != dst.ifindex)
1364 continue;
1365
1366 idev = __in6_dev_get(dev);
1367 if (!idev)
1368 continue;
1369
1370 read_lock_bh(&idev->lock);
1371 list_for_each_entry(score->ifa, &idev->addr_list, if_list) {
1372 int i;
1373
1374 /*
1375 * - Tentative Address (RFC2462 section 5.4)
1376 * - A tentative address is not considered
1377 * "assigned to an interface" in the traditional
1378 * sense, unless it is also flagged as optimistic.
1379 * - Candidate Source Address (section 4)
1380 * - In any case, anycast addresses, multicast
1381 * addresses, and the unspecified address MUST
1382 * NOT be included in a candidate set.
1383 */
1384 if ((score->ifa->flags & IFA_F_TENTATIVE) &&
1385 (!(score->ifa->flags & IFA_F_OPTIMISTIC)))
1386 continue;
1387
1388 score->addr_type = __ipv6_addr_type(&score->ifa->addr);
1389
1390 if (unlikely(score->addr_type == IPV6_ADDR_ANY ||
1391 score->addr_type & IPV6_ADDR_MULTICAST)) {
1392 LIMIT_NETDEBUG(KERN_DEBUG
1393 "ADDRCONF: unspecified / multicast address "
1394 "assigned as unicast address on %s",
1395 dev->name);
1396 continue;
1397 }
1398
1399 score->rule = -1;
1400 bitmap_zero(score->scorebits, IPV6_SADDR_RULE_MAX);
1401
1402 for (i = 0; i < IPV6_SADDR_RULE_MAX; i++) {
1403 int minihiscore, miniscore;
1404
1405 minihiscore = ipv6_get_saddr_eval(net, hiscore, &dst, i);
1406 miniscore = ipv6_get_saddr_eval(net, score, &dst, i);
1407
1408 if (minihiscore > miniscore) {
1409 if (i == IPV6_SADDR_RULE_SCOPE &&
1410 score->scopedist > 0) {
1411 /*
1412 * special case:
1413 * each remaining entry
1414 * has too small (not enough)
1415 * scope, because ifa entries
1416 * are sorted by their scope
1417 * values.
1418 */
1419 goto try_nextdev;
1420 }
1421 break;
1422 } else if (minihiscore < miniscore) {
1423 if (hiscore->ifa)
1424 in6_ifa_put(hiscore->ifa);
1425
1426 in6_ifa_hold(score->ifa);
1427
1428 swap(hiscore, score);
1429
1430 /* restore our iterator */
1431 score->ifa = hiscore->ifa;
1432
1433 break;
1434 }
1435 }
1436 }
1437 try_nextdev:
1438 read_unlock_bh(&idev->lock);
1439 }
1440 rcu_read_unlock();
1441
1442 if (!hiscore->ifa)
1443 return -EADDRNOTAVAIL;
1444
1445 *saddr = hiscore->ifa->addr;
1446 in6_ifa_put(hiscore->ifa);
1447 return 0;
1448 }
1449 EXPORT_SYMBOL(ipv6_dev_get_saddr);
1450
1451 int ipv6_get_lladdr(struct net_device *dev, struct in6_addr *addr,
1452 unsigned char banned_flags)
1453 {
1454 struct inet6_dev *idev;
1455 int err = -EADDRNOTAVAIL;
1456
1457 rcu_read_lock();
1458 idev = __in6_dev_get(dev);
1459 if (idev) {
1460 struct inet6_ifaddr *ifp;
1461
1462 read_lock_bh(&idev->lock);
1463 list_for_each_entry(ifp, &idev->addr_list, if_list) {
1464 if (ifp->scope == IFA_LINK &&
1465 !(ifp->flags & banned_flags)) {
1466 *addr = ifp->addr;
1467 err = 0;
1468 break;
1469 }
1470 }
1471 read_unlock_bh(&idev->lock);
1472 }
1473 rcu_read_unlock();
1474 return err;
1475 }
1476
1477 static int ipv6_count_addresses(struct inet6_dev *idev)
1478 {
1479 int cnt = 0;
1480 struct inet6_ifaddr *ifp;
1481
1482 read_lock_bh(&idev->lock);
1483 list_for_each_entry(ifp, &idev->addr_list, if_list)
1484 cnt++;
1485 read_unlock_bh(&idev->lock);
1486 return cnt;
1487 }
1488
1489 int ipv6_chk_addr(struct net *net, const struct in6_addr *addr,
1490 const struct net_device *dev, int strict)
1491 {
1492 struct inet6_ifaddr *ifp;
1493 unsigned int hash = inet6_addr_hash(addr);
1494
1495 rcu_read_lock_bh();
1496 hlist_for_each_entry_rcu(ifp, &inet6_addr_lst[hash], addr_lst) {
1497 if (!net_eq(dev_net(ifp->idev->dev), net))
1498 continue;
1499 if (ipv6_addr_equal(&ifp->addr, addr) &&
1500 !(ifp->flags&IFA_F_TENTATIVE) &&
1501 (dev == NULL || ifp->idev->dev == dev ||
1502 !(ifp->scope&(IFA_LINK|IFA_HOST) || strict))) {
1503 rcu_read_unlock_bh();
1504 return 1;
1505 }
1506 }
1507
1508 rcu_read_unlock_bh();
1509 return 0;
1510 }
1511 EXPORT_SYMBOL(ipv6_chk_addr);
1512
1513 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
1514 struct net_device *dev)
1515 {
1516 unsigned int hash = inet6_addr_hash(addr);
1517 struct inet6_ifaddr *ifp;
1518
1519 hlist_for_each_entry(ifp, &inet6_addr_lst[hash], addr_lst) {
1520 if (!net_eq(dev_net(ifp->idev->dev), net))
1521 continue;
1522 if (ipv6_addr_equal(&ifp->addr, addr)) {
1523 if (dev == NULL || ifp->idev->dev == dev)
1524 return true;
1525 }
1526 }
1527 return false;
1528 }
1529
1530 int ipv6_chk_prefix(const struct in6_addr *addr, struct net_device *dev)
1531 {
1532 struct inet6_dev *idev;
1533 struct inet6_ifaddr *ifa;
1534 int onlink;
1535
1536 onlink = 0;
1537 rcu_read_lock();
1538 idev = __in6_dev_get(dev);
1539 if (idev) {
1540 read_lock_bh(&idev->lock);
1541 list_for_each_entry(ifa, &idev->addr_list, if_list) {
1542 onlink = ipv6_prefix_equal(addr, &ifa->addr,
1543 ifa->prefix_len);
1544 if (onlink)
1545 break;
1546 }
1547 read_unlock_bh(&idev->lock);
1548 }
1549 rcu_read_unlock();
1550 return onlink;
1551 }
1552 EXPORT_SYMBOL(ipv6_chk_prefix);
1553
1554 struct inet6_ifaddr *ipv6_get_ifaddr(struct net *net, const struct in6_addr *addr,
1555 struct net_device *dev, int strict)
1556 {
1557 struct inet6_ifaddr *ifp, *result = NULL;
1558 unsigned int hash = inet6_addr_hash(addr);
1559
1560 rcu_read_lock_bh();
1561 hlist_for_each_entry_rcu_bh(ifp, &inet6_addr_lst[hash], addr_lst) {
1562 if (!net_eq(dev_net(ifp->idev->dev), net))
1563 continue;
1564 if (ipv6_addr_equal(&ifp->addr, addr)) {
1565 if (dev == NULL || ifp->idev->dev == dev ||
1566 !(ifp->scope&(IFA_LINK|IFA_HOST) || strict)) {
1567 result = ifp;
1568 in6_ifa_hold(ifp);
1569 break;
1570 }
1571 }
1572 }
1573 rcu_read_unlock_bh();
1574
1575 return result;
1576 }
1577
1578 /* Gets referenced address, destroys ifaddr */
1579
1580 static void addrconf_dad_stop(struct inet6_ifaddr *ifp, int dad_failed)
1581 {
1582 if (ifp->flags&IFA_F_PERMANENT) {
1583 spin_lock_bh(&ifp->lock);
1584 addrconf_del_timer(ifp);
1585 ifp->flags |= IFA_F_TENTATIVE;
1586 if (dad_failed)
1587 ifp->flags |= IFA_F_DADFAILED;
1588 spin_unlock_bh(&ifp->lock);
1589 if (dad_failed)
1590 ipv6_ifa_notify(0, ifp);
1591 in6_ifa_put(ifp);
1592 #ifdef CONFIG_IPV6_PRIVACY
1593 } else if (ifp->flags&IFA_F_TEMPORARY) {
1594 struct inet6_ifaddr *ifpub;
1595 spin_lock_bh(&ifp->lock);
1596 ifpub = ifp->ifpub;
1597 if (ifpub) {
1598 in6_ifa_hold(ifpub);
1599 spin_unlock_bh(&ifp->lock);
1600 ipv6_create_tempaddr(ifpub, ifp);
1601 in6_ifa_put(ifpub);
1602 } else {
1603 spin_unlock_bh(&ifp->lock);
1604 }
1605 ipv6_del_addr(ifp);
1606 #endif
1607 } else
1608 ipv6_del_addr(ifp);
1609 }
1610
1611 static int addrconf_dad_end(struct inet6_ifaddr *ifp)
1612 {
1613 int err = -ENOENT;
1614
1615 spin_lock(&ifp->state_lock);
1616 if (ifp->state == INET6_IFADDR_STATE_DAD) {
1617 ifp->state = INET6_IFADDR_STATE_POSTDAD;
1618 err = 0;
1619 }
1620 spin_unlock(&ifp->state_lock);
1621
1622 return err;
1623 }
1624
1625 void addrconf_dad_failure(struct inet6_ifaddr *ifp)
1626 {
1627 struct inet6_dev *idev = ifp->idev;
1628
1629 if (addrconf_dad_end(ifp)) {
1630 in6_ifa_put(ifp);
1631 return;
1632 }
1633
1634 net_info_ratelimited("%s: IPv6 duplicate address %pI6c detected!\n",
1635 ifp->idev->dev->name, &ifp->addr);
1636
1637 if (idev->cnf.accept_dad > 1 && !idev->cnf.disable_ipv6) {
1638 struct in6_addr addr;
1639
1640 addr.s6_addr32[0] = htonl(0xfe800000);
1641 addr.s6_addr32[1] = 0;
1642
1643 if (!ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) &&
1644 ipv6_addr_equal(&ifp->addr, &addr)) {
1645 /* DAD failed for link-local based on MAC address */
1646 idev->cnf.disable_ipv6 = 1;
1647
1648 pr_info("%s: IPv6 being disabled!\n",
1649 ifp->idev->dev->name);
1650 }
1651 }
1652
1653 addrconf_dad_stop(ifp, 1);
1654 }
1655
1656 /* Join to solicited addr multicast group. */
1657
1658 void addrconf_join_solict(struct net_device *dev, const struct in6_addr *addr)
1659 {
1660 struct in6_addr maddr;
1661
1662 if (dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1663 return;
1664
1665 addrconf_addr_solict_mult(addr, &maddr);
1666 ipv6_dev_mc_inc(dev, &maddr);
1667 }
1668
1669 void addrconf_leave_solict(struct inet6_dev *idev, const struct in6_addr *addr)
1670 {
1671 struct in6_addr maddr;
1672
1673 if (idev->dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1674 return;
1675
1676 addrconf_addr_solict_mult(addr, &maddr);
1677 __ipv6_dev_mc_dec(idev, &maddr);
1678 }
1679
1680 static void addrconf_join_anycast(struct inet6_ifaddr *ifp)
1681 {
1682 struct in6_addr addr;
1683 if (ifp->prefix_len == 127) /* RFC 6164 */
1684 return;
1685 ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1686 if (ipv6_addr_any(&addr))
1687 return;
1688 ipv6_dev_ac_inc(ifp->idev->dev, &addr);
1689 }
1690
1691 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp)
1692 {
1693 struct in6_addr addr;
1694 if (ifp->prefix_len == 127) /* RFC 6164 */
1695 return;
1696 ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1697 if (ipv6_addr_any(&addr))
1698 return;
1699 __ipv6_dev_ac_dec(ifp->idev, &addr);
1700 }
1701
1702 static int addrconf_ifid_eui48(u8 *eui, struct net_device *dev)
1703 {
1704 if (dev->addr_len != ETH_ALEN)
1705 return -1;
1706 memcpy(eui, dev->dev_addr, 3);
1707 memcpy(eui + 5, dev->dev_addr + 3, 3);
1708
1709 /*
1710 * The zSeries OSA network cards can be shared among various
1711 * OS instances, but the OSA cards have only one MAC address.
1712 * This leads to duplicate address conflicts in conjunction
1713 * with IPv6 if more than one instance uses the same card.
1714 *
1715 * The driver for these cards can deliver a unique 16-bit
1716 * identifier for each instance sharing the same card. It is
1717 * placed instead of 0xFFFE in the interface identifier. The
1718 * "u" bit of the interface identifier is not inverted in this
1719 * case. Hence the resulting interface identifier has local
1720 * scope according to RFC2373.
1721 */
1722 if (dev->dev_id) {
1723 eui[3] = (dev->dev_id >> 8) & 0xFF;
1724 eui[4] = dev->dev_id & 0xFF;
1725 } else {
1726 eui[3] = 0xFF;
1727 eui[4] = 0xFE;
1728 eui[0] ^= 2;
1729 }
1730 return 0;
1731 }
1732
1733 static int addrconf_ifid_eui64(u8 *eui, struct net_device *dev)
1734 {
1735 if (dev->addr_len != IEEE802154_ADDR_LEN)
1736 return -1;
1737 memcpy(eui, dev->dev_addr, 8);
1738 eui[0] ^= 2;
1739 return 0;
1740 }
1741
1742 static int addrconf_ifid_ieee1394(u8 *eui, struct net_device *dev)
1743 {
1744 union fwnet_hwaddr *ha;
1745
1746 if (dev->addr_len != FWNET_ALEN)
1747 return -1;
1748
1749 ha = (union fwnet_hwaddr *)dev->dev_addr;
1750
1751 memcpy(eui, &ha->uc.uniq_id, sizeof(ha->uc.uniq_id));
1752 eui[0] ^= 2;
1753 return 0;
1754 }
1755
1756 static int addrconf_ifid_arcnet(u8 *eui, struct net_device *dev)
1757 {
1758 /* XXX: inherit EUI-64 from other interface -- yoshfuji */
1759 if (dev->addr_len != ARCNET_ALEN)
1760 return -1;
1761 memset(eui, 0, 7);
1762 eui[7] = *(u8 *)dev->dev_addr;
1763 return 0;
1764 }
1765
1766 static int addrconf_ifid_infiniband(u8 *eui, struct net_device *dev)
1767 {
1768 if (dev->addr_len != INFINIBAND_ALEN)
1769 return -1;
1770 memcpy(eui, dev->dev_addr + 12, 8);
1771 eui[0] |= 2;
1772 return 0;
1773 }
1774
1775 static int __ipv6_isatap_ifid(u8 *eui, __be32 addr)
1776 {
1777 if (addr == 0)
1778 return -1;
1779 eui[0] = (ipv4_is_zeronet(addr) || ipv4_is_private_10(addr) ||
1780 ipv4_is_loopback(addr) || ipv4_is_linklocal_169(addr) ||
1781 ipv4_is_private_172(addr) || ipv4_is_test_192(addr) ||
1782 ipv4_is_anycast_6to4(addr) || ipv4_is_private_192(addr) ||
1783 ipv4_is_test_198(addr) || ipv4_is_multicast(addr) ||
1784 ipv4_is_lbcast(addr)) ? 0x00 : 0x02;
1785 eui[1] = 0;
1786 eui[2] = 0x5E;
1787 eui[3] = 0xFE;
1788 memcpy(eui + 4, &addr, 4);
1789 return 0;
1790 }
1791
1792 static int addrconf_ifid_sit(u8 *eui, struct net_device *dev)
1793 {
1794 if (dev->priv_flags & IFF_ISATAP)
1795 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
1796 return -1;
1797 }
1798
1799 static int addrconf_ifid_gre(u8 *eui, struct net_device *dev)
1800 {
1801 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
1802 }
1803
1804 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev)
1805 {
1806 switch (dev->type) {
1807 case ARPHRD_ETHER:
1808 case ARPHRD_FDDI:
1809 return addrconf_ifid_eui48(eui, dev);
1810 case ARPHRD_ARCNET:
1811 return addrconf_ifid_arcnet(eui, dev);
1812 case ARPHRD_INFINIBAND:
1813 return addrconf_ifid_infiniband(eui, dev);
1814 case ARPHRD_SIT:
1815 return addrconf_ifid_sit(eui, dev);
1816 case ARPHRD_IPGRE:
1817 return addrconf_ifid_gre(eui, dev);
1818 case ARPHRD_IEEE802154:
1819 return addrconf_ifid_eui64(eui, dev);
1820 case ARPHRD_IEEE1394:
1821 return addrconf_ifid_ieee1394(eui, dev);
1822 }
1823 return -1;
1824 }
1825
1826 static int ipv6_inherit_eui64(u8 *eui, struct inet6_dev *idev)
1827 {
1828 int err = -1;
1829 struct inet6_ifaddr *ifp;
1830
1831 read_lock_bh(&idev->lock);
1832 list_for_each_entry(ifp, &idev->addr_list, if_list) {
1833 if (ifp->scope == IFA_LINK && !(ifp->flags&IFA_F_TENTATIVE)) {
1834 memcpy(eui, ifp->addr.s6_addr+8, 8);
1835 err = 0;
1836 break;
1837 }
1838 }
1839 read_unlock_bh(&idev->lock);
1840 return err;
1841 }
1842
1843 #ifdef CONFIG_IPV6_PRIVACY
1844 /* (re)generation of randomized interface identifier (RFC 3041 3.2, 3.5) */
1845 static void __ipv6_regen_rndid(struct inet6_dev *idev)
1846 {
1847 regen:
1848 get_random_bytes(idev->rndid, sizeof(idev->rndid));
1849 idev->rndid[0] &= ~0x02;
1850
1851 /*
1852 * <draft-ietf-ipngwg-temp-addresses-v2-00.txt>:
1853 * check if generated address is not inappropriate
1854 *
1855 * - Reserved subnet anycast (RFC 2526)
1856 * 11111101 11....11 1xxxxxxx
1857 * - ISATAP (RFC4214) 6.1
1858 * 00-00-5E-FE-xx-xx-xx-xx
1859 * - value 0
1860 * - XXX: already assigned to an address on the device
1861 */
1862 if (idev->rndid[0] == 0xfd &&
1863 (idev->rndid[1]&idev->rndid[2]&idev->rndid[3]&idev->rndid[4]&idev->rndid[5]&idev->rndid[6]) == 0xff &&
1864 (idev->rndid[7]&0x80))
1865 goto regen;
1866 if ((idev->rndid[0]|idev->rndid[1]) == 0) {
1867 if (idev->rndid[2] == 0x5e && idev->rndid[3] == 0xfe)
1868 goto regen;
1869 if ((idev->rndid[2]|idev->rndid[3]|idev->rndid[4]|idev->rndid[5]|idev->rndid[6]|idev->rndid[7]) == 0x00)
1870 goto regen;
1871 }
1872 }
1873
1874 static void ipv6_regen_rndid(unsigned long data)
1875 {
1876 struct inet6_dev *idev = (struct inet6_dev *) data;
1877 unsigned long expires;
1878
1879 rcu_read_lock_bh();
1880 write_lock_bh(&idev->lock);
1881
1882 if (idev->dead)
1883 goto out;
1884
1885 __ipv6_regen_rndid(idev);
1886
1887 expires = jiffies +
1888 idev->cnf.temp_prefered_lft * HZ -
1889 idev->cnf.regen_max_retry * idev->cnf.dad_transmits * idev->nd_parms->retrans_time -
1890 idev->cnf.max_desync_factor * HZ;
1891 if (time_before(expires, jiffies)) {
1892 pr_warn("%s: too short regeneration interval; timer disabled for %s\n",
1893 __func__, idev->dev->name);
1894 goto out;
1895 }
1896
1897 if (!mod_timer(&idev->regen_timer, expires))
1898 in6_dev_hold(idev);
1899
1900 out:
1901 write_unlock_bh(&idev->lock);
1902 rcu_read_unlock_bh();
1903 in6_dev_put(idev);
1904 }
1905
1906 static void __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr)
1907 {
1908 if (tmpaddr && memcmp(idev->rndid, &tmpaddr->s6_addr[8], 8) == 0)
1909 __ipv6_regen_rndid(idev);
1910 }
1911 #endif
1912
1913 /*
1914 * Add prefix route.
1915 */
1916
1917 static void
1918 addrconf_prefix_route(struct in6_addr *pfx, int plen, struct net_device *dev,
1919 unsigned long expires, u32 flags)
1920 {
1921 struct fib6_config cfg = {
1922 .fc_table = RT6_TABLE_PREFIX,
1923 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1924 .fc_ifindex = dev->ifindex,
1925 .fc_expires = expires,
1926 .fc_dst_len = plen,
1927 .fc_flags = RTF_UP | flags,
1928 .fc_nlinfo.nl_net = dev_net(dev),
1929 .fc_protocol = RTPROT_KERNEL,
1930 };
1931
1932 cfg.fc_dst = *pfx;
1933
1934 /* Prevent useless cloning on PtP SIT.
1935 This thing is done here expecting that the whole
1936 class of non-broadcast devices need not cloning.
1937 */
1938 #if IS_ENABLED(CONFIG_IPV6_SIT)
1939 if (dev->type == ARPHRD_SIT && (dev->flags & IFF_POINTOPOINT))
1940 cfg.fc_flags |= RTF_NONEXTHOP;
1941 #endif
1942
1943 ip6_route_add(&cfg);
1944 }
1945
1946
1947 static struct rt6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
1948 int plen,
1949 const struct net_device *dev,
1950 u32 flags, u32 noflags)
1951 {
1952 struct fib6_node *fn;
1953 struct rt6_info *rt = NULL;
1954 struct fib6_table *table;
1955
1956 table = fib6_get_table(dev_net(dev), RT6_TABLE_PREFIX);
1957 if (table == NULL)
1958 return NULL;
1959
1960 read_lock_bh(&table->tb6_lock);
1961 fn = fib6_locate(&table->tb6_root, pfx, plen, NULL, 0);
1962 if (!fn)
1963 goto out;
1964 for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
1965 if (rt->dst.dev->ifindex != dev->ifindex)
1966 continue;
1967 if ((rt->rt6i_flags & flags) != flags)
1968 continue;
1969 if ((rt->rt6i_flags & noflags) != 0)
1970 continue;
1971 dst_hold(&rt->dst);
1972 break;
1973 }
1974 out:
1975 read_unlock_bh(&table->tb6_lock);
1976 return rt;
1977 }
1978
1979
1980 /* Create "default" multicast route to the interface */
1981
1982 static void addrconf_add_mroute(struct net_device *dev)
1983 {
1984 struct fib6_config cfg = {
1985 .fc_table = RT6_TABLE_LOCAL,
1986 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1987 .fc_ifindex = dev->ifindex,
1988 .fc_dst_len = 8,
1989 .fc_flags = RTF_UP,
1990 .fc_nlinfo.nl_net = dev_net(dev),
1991 };
1992
1993 ipv6_addr_set(&cfg.fc_dst, htonl(0xFF000000), 0, 0, 0);
1994
1995 ip6_route_add(&cfg);
1996 }
1997
1998 #if IS_ENABLED(CONFIG_IPV6_SIT)
1999 static void sit_route_add(struct net_device *dev)
2000 {
2001 struct fib6_config cfg = {
2002 .fc_table = RT6_TABLE_MAIN,
2003 .fc_metric = IP6_RT_PRIO_ADDRCONF,
2004 .fc_ifindex = dev->ifindex,
2005 .fc_dst_len = 96,
2006 .fc_flags = RTF_UP | RTF_NONEXTHOP,
2007 .fc_nlinfo.nl_net = dev_net(dev),
2008 };
2009
2010 /* prefix length - 96 bits "::d.d.d.d" */
2011 ip6_route_add(&cfg);
2012 }
2013 #endif
2014
2015 static struct inet6_dev *addrconf_add_dev(struct net_device *dev)
2016 {
2017 struct inet6_dev *idev;
2018
2019 ASSERT_RTNL();
2020
2021 idev = ipv6_find_idev(dev);
2022 if (!idev)
2023 return ERR_PTR(-ENOBUFS);
2024
2025 if (idev->cnf.disable_ipv6)
2026 return ERR_PTR(-EACCES);
2027
2028 /* Add default multicast route */
2029 if (!(dev->flags & IFF_LOOPBACK))
2030 addrconf_add_mroute(dev);
2031
2032 return idev;
2033 }
2034
2035 void addrconf_prefix_rcv(struct net_device *dev, u8 *opt, int len, bool sllao)
2036 {
2037 struct prefix_info *pinfo;
2038 __u32 valid_lft;
2039 __u32 prefered_lft;
2040 int addr_type;
2041 struct inet6_dev *in6_dev;
2042 struct net *net = dev_net(dev);
2043
2044 pinfo = (struct prefix_info *) opt;
2045
2046 if (len < sizeof(struct prefix_info)) {
2047 ADBG(("addrconf: prefix option too short\n"));
2048 return;
2049 }
2050
2051 /*
2052 * Validation checks ([ADDRCONF], page 19)
2053 */
2054
2055 addr_type = ipv6_addr_type(&pinfo->prefix);
2056
2057 if (addr_type & (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL))
2058 return;
2059
2060 valid_lft = ntohl(pinfo->valid);
2061 prefered_lft = ntohl(pinfo->prefered);
2062
2063 if (prefered_lft > valid_lft) {
2064 net_warn_ratelimited("addrconf: prefix option has invalid lifetime\n");
2065 return;
2066 }
2067
2068 in6_dev = in6_dev_get(dev);
2069
2070 if (in6_dev == NULL) {
2071 net_dbg_ratelimited("addrconf: device %s not configured\n",
2072 dev->name);
2073 return;
2074 }
2075
2076 /*
2077 * Two things going on here:
2078 * 1) Add routes for on-link prefixes
2079 * 2) Configure prefixes with the auto flag set
2080 */
2081
2082 if (pinfo->onlink) {
2083 struct rt6_info *rt;
2084 unsigned long rt_expires;
2085
2086 /* Avoid arithmetic overflow. Really, we could
2087 * save rt_expires in seconds, likely valid_lft,
2088 * but it would require division in fib gc, that it
2089 * not good.
2090 */
2091 if (HZ > USER_HZ)
2092 rt_expires = addrconf_timeout_fixup(valid_lft, HZ);
2093 else
2094 rt_expires = addrconf_timeout_fixup(valid_lft, USER_HZ);
2095
2096 if (addrconf_finite_timeout(rt_expires))
2097 rt_expires *= HZ;
2098
2099 rt = addrconf_get_prefix_route(&pinfo->prefix,
2100 pinfo->prefix_len,
2101 dev,
2102 RTF_ADDRCONF | RTF_PREFIX_RT,
2103 RTF_GATEWAY | RTF_DEFAULT);
2104
2105 if (rt) {
2106 /* Autoconf prefix route */
2107 if (valid_lft == 0) {
2108 ip6_del_rt(rt);
2109 rt = NULL;
2110 } else if (addrconf_finite_timeout(rt_expires)) {
2111 /* not infinity */
2112 rt6_set_expires(rt, jiffies + rt_expires);
2113 } else {
2114 rt6_clean_expires(rt);
2115 }
2116 } else if (valid_lft) {
2117 clock_t expires = 0;
2118 int flags = RTF_ADDRCONF | RTF_PREFIX_RT;
2119 if (addrconf_finite_timeout(rt_expires)) {
2120 /* not infinity */
2121 flags |= RTF_EXPIRES;
2122 expires = jiffies_to_clock_t(rt_expires);
2123 }
2124 addrconf_prefix_route(&pinfo->prefix, pinfo->prefix_len,
2125 dev, expires, flags);
2126 }
2127 ip6_rt_put(rt);
2128 }
2129
2130 /* Try to figure out our local address for this prefix */
2131
2132 if (pinfo->autoconf && in6_dev->cnf.autoconf) {
2133 struct inet6_ifaddr *ifp;
2134 struct in6_addr addr;
2135 int create = 0, update_lft = 0;
2136 bool tokenized = false;
2137
2138 if (pinfo->prefix_len == 64) {
2139 memcpy(&addr, &pinfo->prefix, 8);
2140
2141 if (!ipv6_addr_any(&in6_dev->token)) {
2142 read_lock_bh(&in6_dev->lock);
2143 memcpy(addr.s6_addr + 8,
2144 in6_dev->token.s6_addr + 8, 8);
2145 read_unlock_bh(&in6_dev->lock);
2146 tokenized = true;
2147 } else if (ipv6_generate_eui64(addr.s6_addr + 8, dev) &&
2148 ipv6_inherit_eui64(addr.s6_addr + 8, in6_dev)) {
2149 in6_dev_put(in6_dev);
2150 return;
2151 }
2152 goto ok;
2153 }
2154 net_dbg_ratelimited("IPv6 addrconf: prefix with wrong length %d\n",
2155 pinfo->prefix_len);
2156 in6_dev_put(in6_dev);
2157 return;
2158
2159 ok:
2160
2161 ifp = ipv6_get_ifaddr(net, &addr, dev, 1);
2162
2163 if (ifp == NULL && valid_lft) {
2164 int max_addresses = in6_dev->cnf.max_addresses;
2165 u32 addr_flags = 0;
2166
2167 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2168 if (in6_dev->cnf.optimistic_dad &&
2169 !net->ipv6.devconf_all->forwarding && sllao)
2170 addr_flags = IFA_F_OPTIMISTIC;
2171 #endif
2172
2173 /* Do not allow to create too much of autoconfigured
2174 * addresses; this would be too easy way to crash kernel.
2175 */
2176 if (!max_addresses ||
2177 ipv6_count_addresses(in6_dev) < max_addresses)
2178 ifp = ipv6_add_addr(in6_dev, &addr, pinfo->prefix_len,
2179 addr_type&IPV6_ADDR_SCOPE_MASK,
2180 addr_flags);
2181
2182 if (IS_ERR_OR_NULL(ifp)) {
2183 in6_dev_put(in6_dev);
2184 return;
2185 }
2186
2187 update_lft = create = 1;
2188 ifp->cstamp = jiffies;
2189 ifp->tokenized = tokenized;
2190 addrconf_dad_start(ifp);
2191 }
2192
2193 if (ifp) {
2194 int flags;
2195 unsigned long now;
2196 #ifdef CONFIG_IPV6_PRIVACY
2197 struct inet6_ifaddr *ift;
2198 #endif
2199 u32 stored_lft;
2200
2201 /* update lifetime (RFC2462 5.5.3 e) */
2202 spin_lock(&ifp->lock);
2203 now = jiffies;
2204 if (ifp->valid_lft > (now - ifp->tstamp) / HZ)
2205 stored_lft = ifp->valid_lft - (now - ifp->tstamp) / HZ;
2206 else
2207 stored_lft = 0;
2208 if (!update_lft && stored_lft) {
2209 if (valid_lft > MIN_VALID_LIFETIME ||
2210 valid_lft > stored_lft)
2211 update_lft = 1;
2212 else if (stored_lft <= MIN_VALID_LIFETIME) {
2213 /* valid_lft <= stored_lft is always true */
2214 /*
2215 * RFC 4862 Section 5.5.3e:
2216 * "Note that the preferred lifetime of
2217 * the corresponding address is always
2218 * reset to the Preferred Lifetime in
2219 * the received Prefix Information
2220 * option, regardless of whether the
2221 * valid lifetime is also reset or
2222 * ignored."
2223 *
2224 * So if the preferred lifetime in
2225 * this advertisement is different
2226 * than what we have stored, but the
2227 * valid lifetime is invalid, just
2228 * reset prefered_lft.
2229 *
2230 * We must set the valid lifetime
2231 * to the stored lifetime since we'll
2232 * be updating the timestamp below,
2233 * else we'll set it back to the
2234 * minimum.
2235 */
2236 if (prefered_lft != ifp->prefered_lft) {
2237 valid_lft = stored_lft;
2238 update_lft = 1;
2239 }
2240 } else {
2241 valid_lft = MIN_VALID_LIFETIME;
2242 if (valid_lft < prefered_lft)
2243 prefered_lft = valid_lft;
2244 update_lft = 1;
2245 }
2246 }
2247
2248 if (update_lft) {
2249 ifp->valid_lft = valid_lft;
2250 ifp->prefered_lft = prefered_lft;
2251 ifp->tstamp = now;
2252 flags = ifp->flags;
2253 ifp->flags &= ~IFA_F_DEPRECATED;
2254 spin_unlock(&ifp->lock);
2255
2256 if (!(flags&IFA_F_TENTATIVE))
2257 ipv6_ifa_notify(0, ifp);
2258 } else
2259 spin_unlock(&ifp->lock);
2260
2261 #ifdef CONFIG_IPV6_PRIVACY
2262 read_lock_bh(&in6_dev->lock);
2263 /* update all temporary addresses in the list */
2264 list_for_each_entry(ift, &in6_dev->tempaddr_list,
2265 tmp_list) {
2266 int age, max_valid, max_prefered;
2267
2268 if (ifp != ift->ifpub)
2269 continue;
2270
2271 /*
2272 * RFC 4941 section 3.3:
2273 * If a received option will extend the lifetime
2274 * of a public address, the lifetimes of
2275 * temporary addresses should be extended,
2276 * subject to the overall constraint that no
2277 * temporary addresses should ever remain
2278 * "valid" or "preferred" for a time longer than
2279 * (TEMP_VALID_LIFETIME) or
2280 * (TEMP_PREFERRED_LIFETIME - DESYNC_FACTOR),
2281 * respectively.
2282 */
2283 age = (now - ift->cstamp) / HZ;
2284 max_valid = in6_dev->cnf.temp_valid_lft - age;
2285 if (max_valid < 0)
2286 max_valid = 0;
2287
2288 max_prefered = in6_dev->cnf.temp_prefered_lft -
2289 in6_dev->cnf.max_desync_factor -
2290 age;
2291 if (max_prefered < 0)
2292 max_prefered = 0;
2293
2294 if (valid_lft > max_valid)
2295 valid_lft = max_valid;
2296
2297 if (prefered_lft > max_prefered)
2298 prefered_lft = max_prefered;
2299
2300 spin_lock(&ift->lock);
2301 flags = ift->flags;
2302 ift->valid_lft = valid_lft;
2303 ift->prefered_lft = prefered_lft;
2304 ift->tstamp = now;
2305 if (prefered_lft > 0)
2306 ift->flags &= ~IFA_F_DEPRECATED;
2307
2308 spin_unlock(&ift->lock);
2309 if (!(flags&IFA_F_TENTATIVE))
2310 ipv6_ifa_notify(0, ift);
2311 }
2312
2313 if ((create || list_empty(&in6_dev->tempaddr_list)) && in6_dev->cnf.use_tempaddr > 0) {
2314 /*
2315 * When a new public address is created as
2316 * described in [ADDRCONF], also create a new
2317 * temporary address. Also create a temporary
2318 * address if it's enabled but no temporary
2319 * address currently exists.
2320 */
2321 read_unlock_bh(&in6_dev->lock);
2322 ipv6_create_tempaddr(ifp, NULL);
2323 } else {
2324 read_unlock_bh(&in6_dev->lock);
2325 }
2326 #endif
2327 in6_ifa_put(ifp);
2328 addrconf_verify(0);
2329 }
2330 }
2331 inet6_prefix_notify(RTM_NEWPREFIX, in6_dev, pinfo);
2332 in6_dev_put(in6_dev);
2333 }
2334
2335 /*
2336 * Set destination address.
2337 * Special case for SIT interfaces where we create a new "virtual"
2338 * device.
2339 */
2340 int addrconf_set_dstaddr(struct net *net, void __user *arg)
2341 {
2342 struct in6_ifreq ireq;
2343 struct net_device *dev;
2344 int err = -EINVAL;
2345
2346 rtnl_lock();
2347
2348 err = -EFAULT;
2349 if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2350 goto err_exit;
2351
2352 dev = __dev_get_by_index(net, ireq.ifr6_ifindex);
2353
2354 err = -ENODEV;
2355 if (dev == NULL)
2356 goto err_exit;
2357
2358 #if IS_ENABLED(CONFIG_IPV6_SIT)
2359 if (dev->type == ARPHRD_SIT) {
2360 const struct net_device_ops *ops = dev->netdev_ops;
2361 struct ifreq ifr;
2362 struct ip_tunnel_parm p;
2363
2364 err = -EADDRNOTAVAIL;
2365 if (!(ipv6_addr_type(&ireq.ifr6_addr) & IPV6_ADDR_COMPATv4))
2366 goto err_exit;
2367
2368 memset(&p, 0, sizeof(p));
2369 p.iph.daddr = ireq.ifr6_addr.s6_addr32[3];
2370 p.iph.saddr = 0;
2371 p.iph.version = 4;
2372 p.iph.ihl = 5;
2373 p.iph.protocol = IPPROTO_IPV6;
2374 p.iph.ttl = 64;
2375 ifr.ifr_ifru.ifru_data = (__force void __user *)&p;
2376
2377 if (ops->ndo_do_ioctl) {
2378 mm_segment_t oldfs = get_fs();
2379
2380 set_fs(KERNEL_DS);
2381 err = ops->ndo_do_ioctl(dev, &ifr, SIOCADDTUNNEL);
2382 set_fs(oldfs);
2383 } else
2384 err = -EOPNOTSUPP;
2385
2386 if (err == 0) {
2387 err = -ENOBUFS;
2388 dev = __dev_get_by_name(net, p.name);
2389 if (!dev)
2390 goto err_exit;
2391 err = dev_open(dev);
2392 }
2393 }
2394 #endif
2395
2396 err_exit:
2397 rtnl_unlock();
2398 return err;
2399 }
2400
2401 /*
2402 * Manual configuration of address on an interface
2403 */
2404 static int inet6_addr_add(struct net *net, int ifindex, const struct in6_addr *pfx,
2405 unsigned int plen, __u8 ifa_flags, __u32 prefered_lft,
2406 __u32 valid_lft)
2407 {
2408 struct inet6_ifaddr *ifp;
2409 struct inet6_dev *idev;
2410 struct net_device *dev;
2411 int scope;
2412 u32 flags;
2413 clock_t expires;
2414 unsigned long timeout;
2415
2416 ASSERT_RTNL();
2417
2418 if (plen > 128)
2419 return -EINVAL;
2420
2421 /* check the lifetime */
2422 if (!valid_lft || prefered_lft > valid_lft)
2423 return -EINVAL;
2424
2425 dev = __dev_get_by_index(net, ifindex);
2426 if (!dev)
2427 return -ENODEV;
2428
2429 idev = addrconf_add_dev(dev);
2430 if (IS_ERR(idev))
2431 return PTR_ERR(idev);
2432
2433 scope = ipv6_addr_scope(pfx);
2434
2435 timeout = addrconf_timeout_fixup(valid_lft, HZ);
2436 if (addrconf_finite_timeout(timeout)) {
2437 expires = jiffies_to_clock_t(timeout * HZ);
2438 valid_lft = timeout;
2439 flags = RTF_EXPIRES;
2440 } else {
2441 expires = 0;
2442 flags = 0;
2443 ifa_flags |= IFA_F_PERMANENT;
2444 }
2445
2446 timeout = addrconf_timeout_fixup(prefered_lft, HZ);
2447 if (addrconf_finite_timeout(timeout)) {
2448 if (timeout == 0)
2449 ifa_flags |= IFA_F_DEPRECATED;
2450 prefered_lft = timeout;
2451 }
2452
2453 ifp = ipv6_add_addr(idev, pfx, plen, scope, ifa_flags);
2454
2455 if (!IS_ERR(ifp)) {
2456 spin_lock_bh(&ifp->lock);
2457 ifp->valid_lft = valid_lft;
2458 ifp->prefered_lft = prefered_lft;
2459 ifp->tstamp = jiffies;
2460 spin_unlock_bh(&ifp->lock);
2461
2462 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, dev,
2463 expires, flags);
2464 /*
2465 * Note that section 3.1 of RFC 4429 indicates
2466 * that the Optimistic flag should not be set for
2467 * manually configured addresses
2468 */
2469 addrconf_dad_start(ifp);
2470 in6_ifa_put(ifp);
2471 addrconf_verify(0);
2472 return 0;
2473 }
2474
2475 return PTR_ERR(ifp);
2476 }
2477
2478 static int inet6_addr_del(struct net *net, int ifindex, const struct in6_addr *pfx,
2479 unsigned int plen)
2480 {
2481 struct inet6_ifaddr *ifp;
2482 struct inet6_dev *idev;
2483 struct net_device *dev;
2484
2485 if (plen > 128)
2486 return -EINVAL;
2487
2488 dev = __dev_get_by_index(net, ifindex);
2489 if (!dev)
2490 return -ENODEV;
2491
2492 if ((idev = __in6_dev_get(dev)) == NULL)
2493 return -ENXIO;
2494
2495 read_lock_bh(&idev->lock);
2496 list_for_each_entry(ifp, &idev->addr_list, if_list) {
2497 if (ifp->prefix_len == plen &&
2498 ipv6_addr_equal(pfx, &ifp->addr)) {
2499 in6_ifa_hold(ifp);
2500 read_unlock_bh(&idev->lock);
2501
2502 ipv6_del_addr(ifp);
2503
2504 /* If the last address is deleted administratively,
2505 disable IPv6 on this interface.
2506 */
2507 if (list_empty(&idev->addr_list))
2508 addrconf_ifdown(idev->dev, 1);
2509 return 0;
2510 }
2511 }
2512 read_unlock_bh(&idev->lock);
2513 return -EADDRNOTAVAIL;
2514 }
2515
2516
2517 int addrconf_add_ifaddr(struct net *net, void __user *arg)
2518 {
2519 struct in6_ifreq ireq;
2520 int err;
2521
2522 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2523 return -EPERM;
2524
2525 if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2526 return -EFAULT;
2527
2528 rtnl_lock();
2529 err = inet6_addr_add(net, ireq.ifr6_ifindex, &ireq.ifr6_addr,
2530 ireq.ifr6_prefixlen, IFA_F_PERMANENT,
2531 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
2532 rtnl_unlock();
2533 return err;
2534 }
2535
2536 int addrconf_del_ifaddr(struct net *net, void __user *arg)
2537 {
2538 struct in6_ifreq ireq;
2539 int err;
2540
2541 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2542 return -EPERM;
2543
2544 if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2545 return -EFAULT;
2546
2547 rtnl_lock();
2548 err = inet6_addr_del(net, ireq.ifr6_ifindex, &ireq.ifr6_addr,
2549 ireq.ifr6_prefixlen);
2550 rtnl_unlock();
2551 return err;
2552 }
2553
2554 static void add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
2555 int plen, int scope)
2556 {
2557 struct inet6_ifaddr *ifp;
2558
2559 ifp = ipv6_add_addr(idev, addr, plen, scope, IFA_F_PERMANENT);
2560 if (!IS_ERR(ifp)) {
2561 spin_lock_bh(&ifp->lock);
2562 ifp->flags &= ~IFA_F_TENTATIVE;
2563 spin_unlock_bh(&ifp->lock);
2564 ipv6_ifa_notify(RTM_NEWADDR, ifp);
2565 in6_ifa_put(ifp);
2566 }
2567 }
2568
2569 #if IS_ENABLED(CONFIG_IPV6_SIT)
2570 static void sit_add_v4_addrs(struct inet6_dev *idev)
2571 {
2572 struct in6_addr addr;
2573 struct net_device *dev;
2574 struct net *net = dev_net(idev->dev);
2575 int scope;
2576
2577 ASSERT_RTNL();
2578
2579 memset(&addr, 0, sizeof(struct in6_addr));
2580 memcpy(&addr.s6_addr32[3], idev->dev->dev_addr, 4);
2581
2582 if (idev->dev->flags&IFF_POINTOPOINT) {
2583 addr.s6_addr32[0] = htonl(0xfe800000);
2584 scope = IFA_LINK;
2585 } else {
2586 scope = IPV6_ADDR_COMPATv4;
2587 }
2588
2589 if (addr.s6_addr32[3]) {
2590 add_addr(idev, &addr, 128, scope);
2591 return;
2592 }
2593
2594 for_each_netdev(net, dev) {
2595 struct in_device *in_dev = __in_dev_get_rtnl(dev);
2596 if (in_dev && (dev->flags & IFF_UP)) {
2597 struct in_ifaddr *ifa;
2598
2599 int flag = scope;
2600
2601 for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) {
2602 int plen;
2603
2604 addr.s6_addr32[3] = ifa->ifa_local;
2605
2606 if (ifa->ifa_scope == RT_SCOPE_LINK)
2607 continue;
2608 if (ifa->ifa_scope >= RT_SCOPE_HOST) {
2609 if (idev->dev->flags&IFF_POINTOPOINT)
2610 continue;
2611 flag |= IFA_HOST;
2612 }
2613 if (idev->dev->flags&IFF_POINTOPOINT)
2614 plen = 64;
2615 else
2616 plen = 96;
2617
2618 add_addr(idev, &addr, plen, flag);
2619 }
2620 }
2621 }
2622 }
2623 #endif
2624
2625 static void init_loopback(struct net_device *dev)
2626 {
2627 struct inet6_dev *idev;
2628 struct net_device *sp_dev;
2629 struct inet6_ifaddr *sp_ifa;
2630 struct rt6_info *sp_rt;
2631
2632 /* ::1 */
2633
2634 ASSERT_RTNL();
2635
2636 if ((idev = ipv6_find_idev(dev)) == NULL) {
2637 pr_debug("%s: add_dev failed\n", __func__);
2638 return;
2639 }
2640
2641 add_addr(idev, &in6addr_loopback, 128, IFA_HOST);
2642
2643 /* Add routes to other interface's IPv6 addresses */
2644 for_each_netdev(dev_net(dev), sp_dev) {
2645 if (!strcmp(sp_dev->name, dev->name))
2646 continue;
2647
2648 idev = __in6_dev_get(sp_dev);
2649 if (!idev)
2650 continue;
2651
2652 read_lock_bh(&idev->lock);
2653 list_for_each_entry(sp_ifa, &idev->addr_list, if_list) {
2654
2655 if (sp_ifa->flags & (IFA_F_DADFAILED | IFA_F_TENTATIVE))
2656 continue;
2657
2658 sp_rt = addrconf_dst_alloc(idev, &sp_ifa->addr, 0);
2659
2660 /* Failure cases are ignored */
2661 if (!IS_ERR(sp_rt)) {
2662 sp_ifa->rt = sp_rt;
2663 ip6_ins_rt(sp_rt);
2664 }
2665 }
2666 read_unlock_bh(&idev->lock);
2667 }
2668 }
2669
2670 static void addrconf_add_linklocal(struct inet6_dev *idev, const struct in6_addr *addr)
2671 {
2672 struct inet6_ifaddr *ifp;
2673 u32 addr_flags = IFA_F_PERMANENT;
2674
2675 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2676 if (idev->cnf.optimistic_dad &&
2677 !dev_net(idev->dev)->ipv6.devconf_all->forwarding)
2678 addr_flags |= IFA_F_OPTIMISTIC;
2679 #endif
2680
2681
2682 ifp = ipv6_add_addr(idev, addr, 64, IFA_LINK, addr_flags);
2683 if (!IS_ERR(ifp)) {
2684 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, idev->dev, 0, 0);
2685 addrconf_dad_start(ifp);
2686 in6_ifa_put(ifp);
2687 }
2688 }
2689
2690 static void addrconf_dev_config(struct net_device *dev)
2691 {
2692 struct in6_addr addr;
2693 struct inet6_dev *idev;
2694
2695 ASSERT_RTNL();
2696
2697 if ((dev->type != ARPHRD_ETHER) &&
2698 (dev->type != ARPHRD_FDDI) &&
2699 (dev->type != ARPHRD_ARCNET) &&
2700 (dev->type != ARPHRD_INFINIBAND) &&
2701 (dev->type != ARPHRD_IEEE802154) &&
2702 (dev->type != ARPHRD_IEEE1394)) {
2703 /* Alas, we support only Ethernet autoconfiguration. */
2704 return;
2705 }
2706
2707 idev = addrconf_add_dev(dev);
2708 if (IS_ERR(idev))
2709 return;
2710
2711 memset(&addr, 0, sizeof(struct in6_addr));
2712 addr.s6_addr32[0] = htonl(0xFE800000);
2713
2714 if (ipv6_generate_eui64(addr.s6_addr + 8, dev) == 0)
2715 addrconf_add_linklocal(idev, &addr);
2716 }
2717
2718 #if IS_ENABLED(CONFIG_IPV6_SIT)
2719 static void addrconf_sit_config(struct net_device *dev)
2720 {
2721 struct inet6_dev *idev;
2722
2723 ASSERT_RTNL();
2724
2725 /*
2726 * Configure the tunnel with one of our IPv4
2727 * addresses... we should configure all of
2728 * our v4 addrs in the tunnel
2729 */
2730
2731 if ((idev = ipv6_find_idev(dev)) == NULL) {
2732 pr_debug("%s: add_dev failed\n", __func__);
2733 return;
2734 }
2735
2736 if (dev->priv_flags & IFF_ISATAP) {
2737 struct in6_addr addr;
2738
2739 ipv6_addr_set(&addr, htonl(0xFE800000), 0, 0, 0);
2740 addrconf_prefix_route(&addr, 64, dev, 0, 0);
2741 if (!ipv6_generate_eui64(addr.s6_addr + 8, dev))
2742 addrconf_add_linklocal(idev, &addr);
2743 return;
2744 }
2745
2746 sit_add_v4_addrs(idev);
2747
2748 if (dev->flags&IFF_POINTOPOINT)
2749 addrconf_add_mroute(dev);
2750 else
2751 sit_route_add(dev);
2752 }
2753 #endif
2754
2755 #if IS_ENABLED(CONFIG_NET_IPGRE)
2756 static void addrconf_gre_config(struct net_device *dev)
2757 {
2758 struct inet6_dev *idev;
2759 struct in6_addr addr;
2760
2761 pr_info("%s(%s)\n", __func__, dev->name);
2762
2763 ASSERT_RTNL();
2764
2765 if ((idev = ipv6_find_idev(dev)) == NULL) {
2766 pr_debug("%s: add_dev failed\n", __func__);
2767 return;
2768 }
2769
2770 ipv6_addr_set(&addr, htonl(0xFE800000), 0, 0, 0);
2771 addrconf_prefix_route(&addr, 64, dev, 0, 0);
2772
2773 if (!ipv6_generate_eui64(addr.s6_addr + 8, dev))
2774 addrconf_add_linklocal(idev, &addr);
2775 }
2776 #endif
2777
2778 static inline int
2779 ipv6_inherit_linklocal(struct inet6_dev *idev, struct net_device *link_dev)
2780 {
2781 struct in6_addr lladdr;
2782
2783 if (!ipv6_get_lladdr(link_dev, &lladdr, IFA_F_TENTATIVE)) {
2784 addrconf_add_linklocal(idev, &lladdr);
2785 return 0;
2786 }
2787 return -1;
2788 }
2789
2790 static void ip6_tnl_add_linklocal(struct inet6_dev *idev)
2791 {
2792 struct net_device *link_dev;
2793 struct net *net = dev_net(idev->dev);
2794
2795 /* first try to inherit the link-local address from the link device */
2796 if (idev->dev->iflink &&
2797 (link_dev = __dev_get_by_index(net, idev->dev->iflink))) {
2798 if (!ipv6_inherit_linklocal(idev, link_dev))
2799 return;
2800 }
2801 /* then try to inherit it from any device */
2802 for_each_netdev(net, link_dev) {
2803 if (!ipv6_inherit_linklocal(idev, link_dev))
2804 return;
2805 }
2806 pr_debug("init ip6-ip6: add_linklocal failed\n");
2807 }
2808
2809 /*
2810 * Autoconfigure tunnel with a link-local address so routing protocols,
2811 * DHCPv6, MLD etc. can be run over the virtual link
2812 */
2813
2814 static void addrconf_ip6_tnl_config(struct net_device *dev)
2815 {
2816 struct inet6_dev *idev;
2817
2818 ASSERT_RTNL();
2819
2820 idev = addrconf_add_dev(dev);
2821 if (IS_ERR(idev)) {
2822 pr_debug("init ip6-ip6: add_dev failed\n");
2823 return;
2824 }
2825 ip6_tnl_add_linklocal(idev);
2826 }
2827
2828 static int addrconf_notify(struct notifier_block *this, unsigned long event,
2829 void *data)
2830 {
2831 struct net_device *dev = (struct net_device *) data;
2832 struct inet6_dev *idev = __in6_dev_get(dev);
2833 int run_pending = 0;
2834 int err;
2835
2836 switch (event) {
2837 case NETDEV_REGISTER:
2838 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2839 idev = ipv6_add_dev(dev);
2840 if (!idev)
2841 return notifier_from_errno(-ENOMEM);
2842 }
2843 break;
2844
2845 case NETDEV_UP:
2846 case NETDEV_CHANGE:
2847 if (dev->flags & IFF_SLAVE)
2848 break;
2849
2850 if (event == NETDEV_UP) {
2851 if (!addrconf_qdisc_ok(dev)) {
2852 /* device is not ready yet. */
2853 pr_info("ADDRCONF(NETDEV_UP): %s: link is not ready\n",
2854 dev->name);
2855 break;
2856 }
2857
2858 if (!idev && dev->mtu >= IPV6_MIN_MTU)
2859 idev = ipv6_add_dev(dev);
2860
2861 if (idev) {
2862 idev->if_flags |= IF_READY;
2863 run_pending = 1;
2864 }
2865 } else {
2866 if (!addrconf_qdisc_ok(dev)) {
2867 /* device is still not ready. */
2868 break;
2869 }
2870
2871 if (idev) {
2872 if (idev->if_flags & IF_READY)
2873 /* device is already configured. */
2874 break;
2875 idev->if_flags |= IF_READY;
2876 }
2877
2878 pr_info("ADDRCONF(NETDEV_CHANGE): %s: link becomes ready\n",
2879 dev->name);
2880
2881 run_pending = 1;
2882 }
2883
2884 switch (dev->type) {
2885 #if IS_ENABLED(CONFIG_IPV6_SIT)
2886 case ARPHRD_SIT:
2887 addrconf_sit_config(dev);
2888 break;
2889 #endif
2890 #if IS_ENABLED(CONFIG_NET_IPGRE)
2891 case ARPHRD_IPGRE:
2892 addrconf_gre_config(dev);
2893 break;
2894 #endif
2895 case ARPHRD_TUNNEL6:
2896 addrconf_ip6_tnl_config(dev);
2897 break;
2898 case ARPHRD_LOOPBACK:
2899 init_loopback(dev);
2900 break;
2901
2902 default:
2903 addrconf_dev_config(dev);
2904 break;
2905 }
2906
2907 if (idev) {
2908 if (run_pending)
2909 addrconf_dad_run(idev);
2910
2911 /*
2912 * If the MTU changed during the interface down,
2913 * when the interface up, the changed MTU must be
2914 * reflected in the idev as well as routers.
2915 */
2916 if (idev->cnf.mtu6 != dev->mtu &&
2917 dev->mtu >= IPV6_MIN_MTU) {
2918 rt6_mtu_change(dev, dev->mtu);
2919 idev->cnf.mtu6 = dev->mtu;
2920 }
2921 idev->tstamp = jiffies;
2922 inet6_ifinfo_notify(RTM_NEWLINK, idev);
2923
2924 /*
2925 * If the changed mtu during down is lower than
2926 * IPV6_MIN_MTU stop IPv6 on this interface.
2927 */
2928 if (dev->mtu < IPV6_MIN_MTU)
2929 addrconf_ifdown(dev, 1);
2930 }
2931 break;
2932
2933 case NETDEV_CHANGEMTU:
2934 if (idev && dev->mtu >= IPV6_MIN_MTU) {
2935 rt6_mtu_change(dev, dev->mtu);
2936 idev->cnf.mtu6 = dev->mtu;
2937 break;
2938 }
2939
2940 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2941 idev = ipv6_add_dev(dev);
2942 if (idev)
2943 break;
2944 }
2945
2946 /*
2947 * MTU falled under IPV6_MIN_MTU.
2948 * Stop IPv6 on this interface.
2949 */
2950
2951 case NETDEV_DOWN:
2952 case NETDEV_UNREGISTER:
2953 /*
2954 * Remove all addresses from this interface.
2955 */
2956 addrconf_ifdown(dev, event != NETDEV_DOWN);
2957 break;
2958
2959 case NETDEV_CHANGENAME:
2960 if (idev) {
2961 snmp6_unregister_dev(idev);
2962 addrconf_sysctl_unregister(idev);
2963 addrconf_sysctl_register(idev);
2964 err = snmp6_register_dev(idev);
2965 if (err)
2966 return notifier_from_errno(err);
2967 }
2968 break;
2969
2970 case NETDEV_PRE_TYPE_CHANGE:
2971 case NETDEV_POST_TYPE_CHANGE:
2972 addrconf_type_change(dev, event);
2973 break;
2974 }
2975
2976 return NOTIFY_OK;
2977 }
2978
2979 /*
2980 * addrconf module should be notified of a device going up
2981 */
2982 static struct notifier_block ipv6_dev_notf = {
2983 .notifier_call = addrconf_notify,
2984 };
2985
2986 static void addrconf_type_change(struct net_device *dev, unsigned long event)
2987 {
2988 struct inet6_dev *idev;
2989 ASSERT_RTNL();
2990
2991 idev = __in6_dev_get(dev);
2992
2993 if (event == NETDEV_POST_TYPE_CHANGE)
2994 ipv6_mc_remap(idev);
2995 else if (event == NETDEV_PRE_TYPE_CHANGE)
2996 ipv6_mc_unmap(idev);
2997 }
2998
2999 static int addrconf_ifdown(struct net_device *dev, int how)
3000 {
3001 struct net *net = dev_net(dev);
3002 struct inet6_dev *idev;
3003 struct inet6_ifaddr *ifa;
3004 int state, i;
3005
3006 ASSERT_RTNL();
3007
3008 rt6_ifdown(net, dev);
3009 neigh_ifdown(&nd_tbl, dev);
3010
3011 idev = __in6_dev_get(dev);
3012 if (idev == NULL)
3013 return -ENODEV;
3014
3015 /*
3016 * Step 1: remove reference to ipv6 device from parent device.
3017 * Do not dev_put!
3018 */
3019 if (how) {
3020 idev->dead = 1;
3021
3022 /* protected by rtnl_lock */
3023 RCU_INIT_POINTER(dev->ip6_ptr, NULL);
3024
3025 /* Step 1.5: remove snmp6 entry */
3026 snmp6_unregister_dev(idev);
3027
3028 }
3029
3030 /* Step 2: clear hash table */
3031 for (i = 0; i < IN6_ADDR_HSIZE; i++) {
3032 struct hlist_head *h = &inet6_addr_lst[i];
3033
3034 spin_lock_bh(&addrconf_hash_lock);
3035 restart:
3036 hlist_for_each_entry_rcu(ifa, h, addr_lst) {
3037 if (ifa->idev == idev) {
3038 hlist_del_init_rcu(&ifa->addr_lst);
3039 addrconf_del_timer(ifa);
3040 goto restart;
3041 }
3042 }
3043 spin_unlock_bh(&addrconf_hash_lock);
3044 }
3045
3046 write_lock_bh(&idev->lock);
3047
3048 /* Step 2: clear flags for stateless addrconf */
3049 if (!how)
3050 idev->if_flags &= ~(IF_RS_SENT|IF_RA_RCVD|IF_READY);
3051
3052 #ifdef CONFIG_IPV6_PRIVACY
3053 if (how && del_timer(&idev->regen_timer))
3054 in6_dev_put(idev);
3055
3056 /* Step 3: clear tempaddr list */
3057 while (!list_empty(&idev->tempaddr_list)) {
3058 ifa = list_first_entry(&idev->tempaddr_list,
3059 struct inet6_ifaddr, tmp_list);
3060 list_del(&ifa->tmp_list);
3061 write_unlock_bh(&idev->lock);
3062 spin_lock_bh(&ifa->lock);
3063
3064 if (ifa->ifpub) {
3065 in6_ifa_put(ifa->ifpub);
3066 ifa->ifpub = NULL;
3067 }
3068 spin_unlock_bh(&ifa->lock);
3069 in6_ifa_put(ifa);
3070 write_lock_bh(&idev->lock);
3071 }
3072 #endif
3073
3074 while (!list_empty(&idev->addr_list)) {
3075 ifa = list_first_entry(&idev->addr_list,
3076 struct inet6_ifaddr, if_list);
3077 addrconf_del_timer(ifa);
3078
3079 list_del(&ifa->if_list);
3080
3081 write_unlock_bh(&idev->lock);
3082
3083 spin_lock_bh(&ifa->state_lock);
3084 state = ifa->state;
3085 ifa->state = INET6_IFADDR_STATE_DEAD;
3086 spin_unlock_bh(&ifa->state_lock);
3087
3088 if (state != INET6_IFADDR_STATE_DEAD) {
3089 __ipv6_ifa_notify(RTM_DELADDR, ifa);
3090 inet6addr_notifier_call_chain(NETDEV_DOWN, ifa);
3091 }
3092 in6_ifa_put(ifa);
3093
3094 write_lock_bh(&idev->lock);
3095 }
3096
3097 write_unlock_bh(&idev->lock);
3098
3099 /* Step 5: Discard multicast list */
3100 if (how)
3101 ipv6_mc_destroy_dev(idev);
3102 else
3103 ipv6_mc_down(idev);
3104
3105 idev->tstamp = jiffies;
3106
3107 /* Last: Shot the device (if unregistered) */
3108 if (how) {
3109 addrconf_sysctl_unregister(idev);
3110 neigh_parms_release(&nd_tbl, idev->nd_parms);
3111 neigh_ifdown(&nd_tbl, dev);
3112 in6_dev_put(idev);
3113 }
3114 return 0;
3115 }
3116
3117 static void addrconf_rs_timer(unsigned long data)
3118 {
3119 struct inet6_ifaddr *ifp = (struct inet6_ifaddr *) data;
3120 struct inet6_dev *idev = ifp->idev;
3121
3122 read_lock(&idev->lock);
3123 if (idev->dead || !(idev->if_flags & IF_READY))
3124 goto out;
3125
3126 if (!ipv6_accept_ra(idev))
3127 goto out;
3128
3129 /* Announcement received after solicitation was sent */
3130 if (idev->if_flags & IF_RA_RCVD)
3131 goto out;
3132
3133 spin_lock(&ifp->lock);
3134 if (ifp->probes++ < idev->cnf.rtr_solicits) {
3135 /* The wait after the last probe can be shorter */
3136 addrconf_mod_timer(ifp, AC_RS,
3137 (ifp->probes == idev->cnf.rtr_solicits) ?
3138 idev->cnf.rtr_solicit_delay :
3139 idev->cnf.rtr_solicit_interval);
3140 spin_unlock(&ifp->lock);
3141
3142 ndisc_send_rs(idev->dev, &ifp->addr, &in6addr_linklocal_allrouters);
3143 } else {
3144 spin_unlock(&ifp->lock);
3145 /*
3146 * Note: we do not support deprecated "all on-link"
3147 * assumption any longer.
3148 */
3149 pr_debug("%s: no IPv6 routers present\n", idev->dev->name);
3150 }
3151
3152 out:
3153 read_unlock(&idev->lock);
3154 in6_ifa_put(ifp);
3155 }
3156
3157 /*
3158 * Duplicate Address Detection
3159 */
3160 static void addrconf_dad_kick(struct inet6_ifaddr *ifp)
3161 {
3162 unsigned long rand_num;
3163 struct inet6_dev *idev = ifp->idev;
3164
3165 if (ifp->flags & IFA_F_OPTIMISTIC)
3166 rand_num = 0;
3167 else
3168 rand_num = net_random() % (idev->cnf.rtr_solicit_delay ? : 1);
3169
3170 ifp->probes = idev->cnf.dad_transmits;
3171 addrconf_mod_timer(ifp, AC_DAD, rand_num);
3172 }
3173
3174 static void addrconf_dad_start(struct inet6_ifaddr *ifp)
3175 {
3176 struct inet6_dev *idev = ifp->idev;
3177 struct net_device *dev = idev->dev;
3178
3179 addrconf_join_solict(dev, &ifp->addr);
3180
3181 net_srandom(ifp->addr.s6_addr32[3]);
3182
3183 read_lock_bh(&idev->lock);
3184 spin_lock(&ifp->lock);
3185 if (ifp->state == INET6_IFADDR_STATE_DEAD)
3186 goto out;
3187
3188 if (dev->flags&(IFF_NOARP|IFF_LOOPBACK) ||
3189 idev->cnf.accept_dad < 1 ||
3190 !(ifp->flags&IFA_F_TENTATIVE) ||
3191 ifp->flags & IFA_F_NODAD) {
3192 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
3193 spin_unlock(&ifp->lock);
3194 read_unlock_bh(&idev->lock);
3195
3196 addrconf_dad_completed(ifp);
3197 return;
3198 }
3199
3200 if (!(idev->if_flags & IF_READY)) {
3201 spin_unlock(&ifp->lock);
3202 read_unlock_bh(&idev->lock);
3203 /*
3204 * If the device is not ready:
3205 * - keep it tentative if it is a permanent address.
3206 * - otherwise, kill it.
3207 */
3208 in6_ifa_hold(ifp);
3209 addrconf_dad_stop(ifp, 0);
3210 return;
3211 }
3212
3213 /*
3214 * Optimistic nodes can start receiving
3215 * Frames right away
3216 */
3217 if (ifp->flags & IFA_F_OPTIMISTIC)
3218 ip6_ins_rt(ifp->rt);
3219
3220 addrconf_dad_kick(ifp);
3221 out:
3222 spin_unlock(&ifp->lock);
3223 read_unlock_bh(&idev->lock);
3224 }
3225
3226 static void addrconf_dad_timer(unsigned long data)
3227 {
3228 struct inet6_ifaddr *ifp = (struct inet6_ifaddr *) data;
3229 struct inet6_dev *idev = ifp->idev;
3230 struct in6_addr mcaddr;
3231
3232 if (!ifp->probes && addrconf_dad_end(ifp))
3233 goto out;
3234
3235 read_lock(&idev->lock);
3236 if (idev->dead || !(idev->if_flags & IF_READY)) {
3237 read_unlock(&idev->lock);
3238 goto out;
3239 }
3240
3241 spin_lock(&ifp->lock);
3242 if (ifp->state == INET6_IFADDR_STATE_DEAD) {
3243 spin_unlock(&ifp->lock);
3244 read_unlock(&idev->lock);
3245 goto out;
3246 }
3247
3248 if (ifp->probes == 0) {
3249 /*
3250 * DAD was successful
3251 */
3252
3253 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
3254 spin_unlock(&ifp->lock);
3255 read_unlock(&idev->lock);
3256
3257 addrconf_dad_completed(ifp);
3258
3259 goto out;
3260 }
3261
3262 ifp->probes--;
3263 addrconf_mod_timer(ifp, AC_DAD, ifp->idev->nd_parms->retrans_time);
3264 spin_unlock(&ifp->lock);
3265 read_unlock(&idev->lock);
3266
3267 /* send a neighbour solicitation for our addr */
3268 addrconf_addr_solict_mult(&ifp->addr, &mcaddr);
3269 ndisc_send_ns(ifp->idev->dev, NULL, &ifp->addr, &mcaddr, &in6addr_any);
3270 out:
3271 in6_ifa_put(ifp);
3272 }
3273
3274 static void addrconf_dad_completed(struct inet6_ifaddr *ifp)
3275 {
3276 struct net_device *dev = ifp->idev->dev;
3277
3278 /*
3279 * Configure the address for reception. Now it is valid.
3280 */
3281
3282 ipv6_ifa_notify(RTM_NEWADDR, ifp);
3283
3284 /* If added prefix is link local and we are prepared to process
3285 router advertisements, start sending router solicitations.
3286 */
3287
3288 if (ipv6_accept_ra(ifp->idev) &&
3289 ifp->idev->cnf.rtr_solicits > 0 &&
3290 (dev->flags&IFF_LOOPBACK) == 0 &&
3291 (ipv6_addr_type(&ifp->addr) & IPV6_ADDR_LINKLOCAL)) {
3292 /*
3293 * If a host as already performed a random delay
3294 * [...] as part of DAD [...] there is no need
3295 * to delay again before sending the first RS
3296 */
3297 ndisc_send_rs(ifp->idev->dev, &ifp->addr, &in6addr_linklocal_allrouters);
3298
3299 spin_lock_bh(&ifp->lock);
3300 ifp->probes = 1;
3301 ifp->idev->if_flags |= IF_RS_SENT;
3302 addrconf_mod_timer(ifp, AC_RS, ifp->idev->cnf.rtr_solicit_interval);
3303 spin_unlock_bh(&ifp->lock);
3304 }
3305 }
3306
3307 static void addrconf_dad_run(struct inet6_dev *idev)
3308 {
3309 struct inet6_ifaddr *ifp;
3310
3311 read_lock_bh(&idev->lock);
3312 list_for_each_entry(ifp, &idev->addr_list, if_list) {
3313 spin_lock(&ifp->lock);
3314 if (ifp->flags & IFA_F_TENTATIVE &&
3315 ifp->state == INET6_IFADDR_STATE_DAD)
3316 addrconf_dad_kick(ifp);
3317 spin_unlock(&ifp->lock);
3318 }
3319 read_unlock_bh(&idev->lock);
3320 }
3321
3322 #ifdef CONFIG_PROC_FS
3323 struct if6_iter_state {
3324 struct seq_net_private p;
3325 int bucket;
3326 int offset;
3327 };
3328
3329 static struct inet6_ifaddr *if6_get_first(struct seq_file *seq, loff_t pos)
3330 {
3331 struct inet6_ifaddr *ifa = NULL;
3332 struct if6_iter_state *state = seq->private;
3333 struct net *net = seq_file_net(seq);
3334 int p = 0;
3335
3336 /* initial bucket if pos is 0 */
3337 if (pos == 0) {
3338 state->bucket = 0;
3339 state->offset = 0;
3340 }
3341
3342 for (; state->bucket < IN6_ADDR_HSIZE; ++state->bucket) {
3343 hlist_for_each_entry_rcu_bh(ifa, &inet6_addr_lst[state->bucket],
3344 addr_lst) {
3345 if (!net_eq(dev_net(ifa->idev->dev), net))
3346 continue;
3347 /* sync with offset */
3348 if (p < state->offset) {
3349 p++;
3350 continue;
3351 }
3352 state->offset++;
3353 return ifa;
3354 }
3355
3356 /* prepare for next bucket */
3357 state->offset = 0;
3358 p = 0;
3359 }
3360 return NULL;
3361 }
3362
3363 static struct inet6_ifaddr *if6_get_next(struct seq_file *seq,
3364 struct inet6_ifaddr *ifa)
3365 {
3366 struct if6_iter_state *state = seq->private;
3367 struct net *net = seq_file_net(seq);
3368
3369 hlist_for_each_entry_continue_rcu_bh(ifa, addr_lst) {
3370 if (!net_eq(dev_net(ifa->idev->dev), net))
3371 continue;
3372 state->offset++;
3373 return ifa;
3374 }
3375
3376 while (++state->bucket < IN6_ADDR_HSIZE) {
3377 state->offset = 0;
3378 hlist_for_each_entry_rcu_bh(ifa,
3379 &inet6_addr_lst[state->bucket], addr_lst) {
3380 if (!net_eq(dev_net(ifa->idev->dev), net))
3381 continue;
3382 state->offset++;
3383 return ifa;
3384 }
3385 }
3386
3387 return NULL;
3388 }
3389
3390 static void *if6_seq_start(struct seq_file *seq, loff_t *pos)
3391 __acquires(rcu_bh)
3392 {
3393 rcu_read_lock_bh();
3394 return if6_get_first(seq, *pos);
3395 }
3396
3397 static void *if6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
3398 {
3399 struct inet6_ifaddr *ifa;
3400
3401 ifa = if6_get_next(seq, v);
3402 ++*pos;
3403 return ifa;
3404 }
3405
3406 static void if6_seq_stop(struct seq_file *seq, void *v)
3407 __releases(rcu_bh)
3408 {
3409 rcu_read_unlock_bh();
3410 }
3411
3412 static int if6_seq_show(struct seq_file *seq, void *v)
3413 {
3414 struct inet6_ifaddr *ifp = (struct inet6_ifaddr *)v;
3415 seq_printf(seq, "%pi6 %02x %02x %02x %02x %8s\n",
3416 &ifp->addr,
3417 ifp->idev->dev->ifindex,
3418 ifp->prefix_len,
3419 ifp->scope,
3420 ifp->flags,
3421 ifp->idev->dev->name);
3422 return 0;
3423 }
3424
3425 static const struct seq_operations if6_seq_ops = {
3426 .start = if6_seq_start,
3427 .next = if6_seq_next,
3428 .show = if6_seq_show,
3429 .stop = if6_seq_stop,
3430 };
3431
3432 static int if6_seq_open(struct inode *inode, struct file *file)
3433 {
3434 return seq_open_net(inode, file, &if6_seq_ops,
3435 sizeof(struct if6_iter_state));
3436 }
3437
3438 static const struct file_operations if6_fops = {
3439 .owner = THIS_MODULE,
3440 .open = if6_seq_open,
3441 .read = seq_read,
3442 .llseek = seq_lseek,
3443 .release = seq_release_net,
3444 };
3445
3446 static int __net_init if6_proc_net_init(struct net *net)
3447 {
3448 if (!proc_create("if_inet6", S_IRUGO, net->proc_net, &if6_fops))
3449 return -ENOMEM;
3450 return 0;
3451 }
3452
3453 static void __net_exit if6_proc_net_exit(struct net *net)
3454 {
3455 remove_proc_entry("if_inet6", net->proc_net);
3456 }
3457
3458 static struct pernet_operations if6_proc_net_ops = {
3459 .init = if6_proc_net_init,
3460 .exit = if6_proc_net_exit,
3461 };
3462
3463 int __init if6_proc_init(void)
3464 {
3465 return register_pernet_subsys(&if6_proc_net_ops);
3466 }
3467
3468 void if6_proc_exit(void)
3469 {
3470 unregister_pernet_subsys(&if6_proc_net_ops);
3471 }
3472 #endif /* CONFIG_PROC_FS */
3473
3474 #if IS_ENABLED(CONFIG_IPV6_MIP6)
3475 /* Check if address is a home address configured on any interface. */
3476 int ipv6_chk_home_addr(struct net *net, const struct in6_addr *addr)
3477 {
3478 int ret = 0;
3479 struct inet6_ifaddr *ifp = NULL;
3480 unsigned int hash = inet6_addr_hash(addr);
3481
3482 rcu_read_lock_bh();
3483 hlist_for_each_entry_rcu_bh(ifp, &inet6_addr_lst[hash], addr_lst) {
3484 if (!net_eq(dev_net(ifp->idev->dev), net))
3485 continue;
3486 if (ipv6_addr_equal(&ifp->addr, addr) &&
3487 (ifp->flags & IFA_F_HOMEADDRESS)) {
3488 ret = 1;
3489 break;
3490 }
3491 }
3492 rcu_read_unlock_bh();
3493 return ret;
3494 }
3495 #endif
3496
3497 /*
3498 * Periodic address status verification
3499 */
3500
3501 static void addrconf_verify(unsigned long foo)
3502 {
3503 unsigned long now, next, next_sec, next_sched;
3504 struct inet6_ifaddr *ifp;
3505 int i;
3506
3507 rcu_read_lock_bh();
3508 spin_lock(&addrconf_verify_lock);
3509 now = jiffies;
3510 next = round_jiffies_up(now + ADDR_CHECK_FREQUENCY);
3511
3512 del_timer(&addr_chk_timer);
3513
3514 for (i = 0; i < IN6_ADDR_HSIZE; i++) {
3515 restart:
3516 hlist_for_each_entry_rcu_bh(ifp,
3517 &inet6_addr_lst[i], addr_lst) {
3518 unsigned long age;
3519
3520 if (ifp->flags & IFA_F_PERMANENT)
3521 continue;
3522
3523 spin_lock(&ifp->lock);
3524 /* We try to batch several events at once. */
3525 age = (now - ifp->tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
3526
3527 if (ifp->valid_lft != INFINITY_LIFE_TIME &&
3528 age >= ifp->valid_lft) {
3529 spin_unlock(&ifp->lock);
3530 in6_ifa_hold(ifp);
3531 ipv6_del_addr(ifp);
3532 goto restart;
3533 } else if (ifp->prefered_lft == INFINITY_LIFE_TIME) {
3534 spin_unlock(&ifp->lock);
3535 continue;
3536 } else if (age >= ifp->prefered_lft) {
3537 /* jiffies - ifp->tstamp > age >= ifp->prefered_lft */
3538 int deprecate = 0;
3539
3540 if (!(ifp->flags&IFA_F_DEPRECATED)) {
3541 deprecate = 1;
3542 ifp->flags |= IFA_F_DEPRECATED;
3543 }
3544
3545 if (time_before(ifp->tstamp + ifp->valid_lft * HZ, next))
3546 next = ifp->tstamp + ifp->valid_lft * HZ;
3547
3548 spin_unlock(&ifp->lock);
3549
3550 if (deprecate) {
3551 in6_ifa_hold(ifp);
3552
3553 ipv6_ifa_notify(0, ifp);
3554 in6_ifa_put(ifp);
3555 goto restart;
3556 }
3557 #ifdef CONFIG_IPV6_PRIVACY
3558 } else if ((ifp->flags&IFA_F_TEMPORARY) &&
3559 !(ifp->flags&IFA_F_TENTATIVE)) {
3560 unsigned long regen_advance = ifp->idev->cnf.regen_max_retry *
3561 ifp->idev->cnf.dad_transmits *
3562 ifp->idev->nd_parms->retrans_time / HZ;
3563
3564 if (age >= ifp->prefered_lft - regen_advance) {
3565 struct inet6_ifaddr *ifpub = ifp->ifpub;
3566 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3567 next = ifp->tstamp + ifp->prefered_lft * HZ;
3568 if (!ifp->regen_count && ifpub) {
3569 ifp->regen_count++;
3570 in6_ifa_hold(ifp);
3571 in6_ifa_hold(ifpub);
3572 spin_unlock(&ifp->lock);
3573
3574 spin_lock(&ifpub->lock);
3575 ifpub->regen_count = 0;
3576 spin_unlock(&ifpub->lock);
3577 ipv6_create_tempaddr(ifpub, ifp);
3578 in6_ifa_put(ifpub);
3579 in6_ifa_put(ifp);
3580 goto restart;
3581 }
3582 } else if (time_before(ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ, next))
3583 next = ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ;
3584 spin_unlock(&ifp->lock);
3585 #endif
3586 } else {
3587 /* ifp->prefered_lft <= ifp->valid_lft */
3588 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3589 next = ifp->tstamp + ifp->prefered_lft * HZ;
3590 spin_unlock(&ifp->lock);
3591 }
3592 }
3593 }
3594
3595 next_sec = round_jiffies_up(next);
3596 next_sched = next;
3597
3598 /* If rounded timeout is accurate enough, accept it. */
3599 if (time_before(next_sec, next + ADDRCONF_TIMER_FUZZ))
3600 next_sched = next_sec;
3601
3602 /* And minimum interval is ADDRCONF_TIMER_FUZZ_MAX. */
3603 if (time_before(next_sched, jiffies + ADDRCONF_TIMER_FUZZ_MAX))
3604 next_sched = jiffies + ADDRCONF_TIMER_FUZZ_MAX;
3605
3606 ADBG((KERN_DEBUG "now = %lu, schedule = %lu, rounded schedule = %lu => %lu\n",
3607 now, next, next_sec, next_sched));
3608
3609 addr_chk_timer.expires = next_sched;
3610 add_timer(&addr_chk_timer);
3611 spin_unlock(&addrconf_verify_lock);
3612 rcu_read_unlock_bh();
3613 }
3614
3615 static struct in6_addr *extract_addr(struct nlattr *addr, struct nlattr *local)
3616 {
3617 struct in6_addr *pfx = NULL;
3618
3619 if (addr)
3620 pfx = nla_data(addr);
3621
3622 if (local) {
3623 if (pfx && nla_memcmp(local, pfx, sizeof(*pfx)))
3624 pfx = NULL;
3625 else
3626 pfx = nla_data(local);
3627 }
3628
3629 return pfx;
3630 }
3631
3632 static const struct nla_policy ifa_ipv6_policy[IFA_MAX+1] = {
3633 [IFA_ADDRESS] = { .len = sizeof(struct in6_addr) },
3634 [IFA_LOCAL] = { .len = sizeof(struct in6_addr) },
3635 [IFA_CACHEINFO] = { .len = sizeof(struct ifa_cacheinfo) },
3636 };
3637
3638 static int
3639 inet6_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh)
3640 {
3641 struct net *net = sock_net(skb->sk);
3642 struct ifaddrmsg *ifm;
3643 struct nlattr *tb[IFA_MAX+1];
3644 struct in6_addr *pfx;
3645 int err;
3646
3647 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3648 if (err < 0)
3649 return err;
3650
3651 ifm = nlmsg_data(nlh);
3652 pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3653 if (pfx == NULL)
3654 return -EINVAL;
3655
3656 return inet6_addr_del(net, ifm->ifa_index, pfx, ifm->ifa_prefixlen);
3657 }
3658
3659 static int inet6_addr_modify(struct inet6_ifaddr *ifp, u8 ifa_flags,
3660 u32 prefered_lft, u32 valid_lft)
3661 {
3662 u32 flags;
3663 clock_t expires;
3664 unsigned long timeout;
3665
3666 if (!valid_lft || (prefered_lft > valid_lft))
3667 return -EINVAL;
3668
3669 timeout = addrconf_timeout_fixup(valid_lft, HZ);
3670 if (addrconf_finite_timeout(timeout)) {
3671 expires = jiffies_to_clock_t(timeout * HZ);
3672 valid_lft = timeout;
3673 flags = RTF_EXPIRES;
3674 } else {
3675 expires = 0;
3676 flags = 0;
3677 ifa_flags |= IFA_F_PERMANENT;
3678 }
3679
3680 timeout = addrconf_timeout_fixup(prefered_lft, HZ);
3681 if (addrconf_finite_timeout(timeout)) {
3682 if (timeout == 0)
3683 ifa_flags |= IFA_F_DEPRECATED;
3684 prefered_lft = timeout;
3685 }
3686
3687 spin_lock_bh(&ifp->lock);
3688 ifp->flags = (ifp->flags & ~(IFA_F_DEPRECATED | IFA_F_PERMANENT | IFA_F_NODAD | IFA_F_HOMEADDRESS)) | ifa_flags;
3689 ifp->tstamp = jiffies;
3690 ifp->valid_lft = valid_lft;
3691 ifp->prefered_lft = prefered_lft;
3692
3693 spin_unlock_bh(&ifp->lock);
3694 if (!(ifp->flags&IFA_F_TENTATIVE))
3695 ipv6_ifa_notify(0, ifp);
3696
3697 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, ifp->idev->dev,
3698 expires, flags);
3699 addrconf_verify(0);
3700
3701 return 0;
3702 }
3703
3704 static int
3705 inet6_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh)
3706 {
3707 struct net *net = sock_net(skb->sk);
3708 struct ifaddrmsg *ifm;
3709 struct nlattr *tb[IFA_MAX+1];
3710 struct in6_addr *pfx;
3711 struct inet6_ifaddr *ifa;
3712 struct net_device *dev;
3713 u32 valid_lft = INFINITY_LIFE_TIME, preferred_lft = INFINITY_LIFE_TIME;
3714 u8 ifa_flags;
3715 int err;
3716
3717 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3718 if (err < 0)
3719 return err;
3720
3721 ifm = nlmsg_data(nlh);
3722 pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3723 if (pfx == NULL)
3724 return -EINVAL;
3725
3726 if (tb[IFA_CACHEINFO]) {
3727 struct ifa_cacheinfo *ci;
3728
3729 ci = nla_data(tb[IFA_CACHEINFO]);
3730 valid_lft = ci->ifa_valid;
3731 preferred_lft = ci->ifa_prefered;
3732 } else {
3733 preferred_lft = INFINITY_LIFE_TIME;
3734 valid_lft = INFINITY_LIFE_TIME;
3735 }
3736
3737 dev = __dev_get_by_index(net, ifm->ifa_index);
3738 if (dev == NULL)
3739 return -ENODEV;
3740
3741 /* We ignore other flags so far. */
3742 ifa_flags = ifm->ifa_flags & (IFA_F_NODAD | IFA_F_HOMEADDRESS);
3743
3744 ifa = ipv6_get_ifaddr(net, pfx, dev, 1);
3745 if (ifa == NULL) {
3746 /*
3747 * It would be best to check for !NLM_F_CREATE here but
3748 * userspace alreay relies on not having to provide this.
3749 */
3750 return inet6_addr_add(net, ifm->ifa_index, pfx,
3751 ifm->ifa_prefixlen, ifa_flags,
3752 preferred_lft, valid_lft);
3753 }
3754
3755 if (nlh->nlmsg_flags & NLM_F_EXCL ||
3756 !(nlh->nlmsg_flags & NLM_F_REPLACE))
3757 err = -EEXIST;
3758 else
3759 err = inet6_addr_modify(ifa, ifa_flags, preferred_lft, valid_lft);
3760
3761 in6_ifa_put(ifa);
3762
3763 return err;
3764 }
3765
3766 static void put_ifaddrmsg(struct nlmsghdr *nlh, u8 prefixlen, u8 flags,
3767 u8 scope, int ifindex)
3768 {
3769 struct ifaddrmsg *ifm;
3770
3771 ifm = nlmsg_data(nlh);
3772 ifm->ifa_family = AF_INET6;
3773 ifm->ifa_prefixlen = prefixlen;
3774 ifm->ifa_flags = flags;
3775 ifm->ifa_scope = scope;
3776 ifm->ifa_index = ifindex;
3777 }
3778
3779 static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
3780 unsigned long tstamp, u32 preferred, u32 valid)
3781 {
3782 struct ifa_cacheinfo ci;
3783
3784 ci.cstamp = cstamp_delta(cstamp);
3785 ci.tstamp = cstamp_delta(tstamp);
3786 ci.ifa_prefered = preferred;
3787 ci.ifa_valid = valid;
3788
3789 return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
3790 }
3791
3792 static inline int rt_scope(int ifa_scope)
3793 {
3794 if (ifa_scope & IFA_HOST)
3795 return RT_SCOPE_HOST;
3796 else if (ifa_scope & IFA_LINK)
3797 return RT_SCOPE_LINK;
3798 else if (ifa_scope & IFA_SITE)
3799 return RT_SCOPE_SITE;
3800 else
3801 return RT_SCOPE_UNIVERSE;
3802 }
3803
3804 static inline int inet6_ifaddr_msgsize(void)
3805 {
3806 return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
3807 + nla_total_size(16) /* IFA_ADDRESS */
3808 + nla_total_size(sizeof(struct ifa_cacheinfo));
3809 }
3810
3811 static int inet6_fill_ifaddr(struct sk_buff *skb, struct inet6_ifaddr *ifa,
3812 u32 portid, u32 seq, int event, unsigned int flags)
3813 {
3814 struct nlmsghdr *nlh;
3815 u32 preferred, valid;
3816
3817 nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
3818 if (nlh == NULL)
3819 return -EMSGSIZE;
3820
3821 put_ifaddrmsg(nlh, ifa->prefix_len, ifa->flags, rt_scope(ifa->scope),
3822 ifa->idev->dev->ifindex);
3823
3824 if (!(ifa->flags&IFA_F_PERMANENT)) {
3825 preferred = ifa->prefered_lft;
3826 valid = ifa->valid_lft;
3827 if (preferred != INFINITY_LIFE_TIME) {
3828 long tval = (jiffies - ifa->tstamp)/HZ;
3829 if (preferred > tval)
3830 preferred -= tval;
3831 else
3832 preferred = 0;
3833 if (valid != INFINITY_LIFE_TIME) {
3834 if (valid > tval)
3835 valid -= tval;
3836 else
3837 valid = 0;
3838 }
3839 }
3840 } else {
3841 preferred = INFINITY_LIFE_TIME;
3842 valid = INFINITY_LIFE_TIME;
3843 }
3844
3845 if (nla_put(skb, IFA_ADDRESS, 16, &ifa->addr) < 0 ||
3846 put_cacheinfo(skb, ifa->cstamp, ifa->tstamp, preferred, valid) < 0) {
3847 nlmsg_cancel(skb, nlh);
3848 return -EMSGSIZE;
3849 }
3850
3851 return nlmsg_end(skb, nlh);
3852 }
3853
3854 static int inet6_fill_ifmcaddr(struct sk_buff *skb, struct ifmcaddr6 *ifmca,
3855 u32 portid, u32 seq, int event, u16 flags)
3856 {
3857 struct nlmsghdr *nlh;
3858 u8 scope = RT_SCOPE_UNIVERSE;
3859 int ifindex = ifmca->idev->dev->ifindex;
3860
3861 if (ipv6_addr_scope(&ifmca->mca_addr) & IFA_SITE)
3862 scope = RT_SCOPE_SITE;
3863
3864 nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
3865 if (nlh == NULL)
3866 return -EMSGSIZE;
3867
3868 put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3869 if (nla_put(skb, IFA_MULTICAST, 16, &ifmca->mca_addr) < 0 ||
3870 put_cacheinfo(skb, ifmca->mca_cstamp, ifmca->mca_tstamp,
3871 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3872 nlmsg_cancel(skb, nlh);
3873 return -EMSGSIZE;
3874 }
3875
3876 return nlmsg_end(skb, nlh);
3877 }
3878
3879 static int inet6_fill_ifacaddr(struct sk_buff *skb, struct ifacaddr6 *ifaca,
3880 u32 portid, u32 seq, int event, unsigned int flags)
3881 {
3882 struct nlmsghdr *nlh;
3883 u8 scope = RT_SCOPE_UNIVERSE;
3884 int ifindex = ifaca->aca_idev->dev->ifindex;
3885
3886 if (ipv6_addr_scope(&ifaca->aca_addr) & IFA_SITE)
3887 scope = RT_SCOPE_SITE;
3888
3889 nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
3890 if (nlh == NULL)
3891 return -EMSGSIZE;
3892
3893 put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3894 if (nla_put(skb, IFA_ANYCAST, 16, &ifaca->aca_addr) < 0 ||
3895 put_cacheinfo(skb, ifaca->aca_cstamp, ifaca->aca_tstamp,
3896 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3897 nlmsg_cancel(skb, nlh);
3898 return -EMSGSIZE;
3899 }
3900
3901 return nlmsg_end(skb, nlh);
3902 }
3903
3904 enum addr_type_t {
3905 UNICAST_ADDR,
3906 MULTICAST_ADDR,
3907 ANYCAST_ADDR,
3908 };
3909
3910 /* called with rcu_read_lock() */
3911 static int in6_dump_addrs(struct inet6_dev *idev, struct sk_buff *skb,
3912 struct netlink_callback *cb, enum addr_type_t type,
3913 int s_ip_idx, int *p_ip_idx)
3914 {
3915 struct ifmcaddr6 *ifmca;
3916 struct ifacaddr6 *ifaca;
3917 int err = 1;
3918 int ip_idx = *p_ip_idx;
3919
3920 read_lock_bh(&idev->lock);
3921 switch (type) {
3922 case UNICAST_ADDR: {
3923 struct inet6_ifaddr *ifa;
3924
3925 /* unicast address incl. temp addr */
3926 list_for_each_entry(ifa, &idev->addr_list, if_list) {
3927 if (++ip_idx < s_ip_idx)
3928 continue;
3929 err = inet6_fill_ifaddr(skb, ifa,
3930 NETLINK_CB(cb->skb).portid,
3931 cb->nlh->nlmsg_seq,
3932 RTM_NEWADDR,
3933 NLM_F_MULTI);
3934 if (err <= 0)
3935 break;
3936 nl_dump_check_consistent(cb, nlmsg_hdr(skb));
3937 }
3938 break;
3939 }
3940 case MULTICAST_ADDR:
3941 /* multicast address */
3942 for (ifmca = idev->mc_list; ifmca;
3943 ifmca = ifmca->next, ip_idx++) {
3944 if (ip_idx < s_ip_idx)
3945 continue;
3946 err = inet6_fill_ifmcaddr(skb, ifmca,
3947 NETLINK_CB(cb->skb).portid,
3948 cb->nlh->nlmsg_seq,
3949 RTM_GETMULTICAST,
3950 NLM_F_MULTI);
3951 if (err <= 0)
3952 break;
3953 }
3954 break;
3955 case ANYCAST_ADDR:
3956 /* anycast address */
3957 for (ifaca = idev->ac_list; ifaca;
3958 ifaca = ifaca->aca_next, ip_idx++) {
3959 if (ip_idx < s_ip_idx)
3960 continue;
3961 err = inet6_fill_ifacaddr(skb, ifaca,
3962 NETLINK_CB(cb->skb).portid,
3963 cb->nlh->nlmsg_seq,
3964 RTM_GETANYCAST,
3965 NLM_F_MULTI);
3966 if (err <= 0)
3967 break;
3968 }
3969 break;
3970 default:
3971 break;
3972 }
3973 read_unlock_bh(&idev->lock);
3974 *p_ip_idx = ip_idx;
3975 return err;
3976 }
3977
3978 static int inet6_dump_addr(struct sk_buff *skb, struct netlink_callback *cb,
3979 enum addr_type_t type)
3980 {
3981 struct net *net = sock_net(skb->sk);
3982 int h, s_h;
3983 int idx, ip_idx;
3984 int s_idx, s_ip_idx;
3985 struct net_device *dev;
3986 struct inet6_dev *idev;
3987 struct hlist_head *head;
3988
3989 s_h = cb->args[0];
3990 s_idx = idx = cb->args[1];
3991 s_ip_idx = ip_idx = cb->args[2];
3992
3993 rcu_read_lock();
3994 cb->seq = atomic_read(&net->ipv6.dev_addr_genid) ^ net->dev_base_seq;
3995 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
3996 idx = 0;
3997 head = &net->dev_index_head[h];
3998 hlist_for_each_entry_rcu(dev, head, index_hlist) {
3999 if (idx < s_idx)
4000 goto cont;
4001 if (h > s_h || idx > s_idx)
4002 s_ip_idx = 0;
4003 ip_idx = 0;
4004 idev = __in6_dev_get(dev);
4005 if (!idev)
4006 goto cont;
4007
4008 if (in6_dump_addrs(idev, skb, cb, type,
4009 s_ip_idx, &ip_idx) <= 0)
4010 goto done;
4011 cont:
4012 idx++;
4013 }
4014 }
4015 done:
4016 rcu_read_unlock();
4017 cb->args[0] = h;
4018 cb->args[1] = idx;
4019 cb->args[2] = ip_idx;
4020
4021 return skb->len;
4022 }
4023
4024 static int inet6_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
4025 {
4026 enum addr_type_t type = UNICAST_ADDR;
4027
4028 return inet6_dump_addr(skb, cb, type);
4029 }
4030
4031 static int inet6_dump_ifmcaddr(struct sk_buff *skb, struct netlink_callback *cb)
4032 {
4033 enum addr_type_t type = MULTICAST_ADDR;
4034
4035 return inet6_dump_addr(skb, cb, type);
4036 }
4037
4038
4039 static int inet6_dump_ifacaddr(struct sk_buff *skb, struct netlink_callback *cb)
4040 {
4041 enum addr_type_t type = ANYCAST_ADDR;
4042
4043 return inet6_dump_addr(skb, cb, type);
4044 }
4045
4046 static int inet6_rtm_getaddr(struct sk_buff *in_skb, struct nlmsghdr *nlh)
4047 {
4048 struct net *net = sock_net(in_skb->sk);
4049 struct ifaddrmsg *ifm;
4050 struct nlattr *tb[IFA_MAX+1];
4051 struct in6_addr *addr = NULL;
4052 struct net_device *dev = NULL;
4053 struct inet6_ifaddr *ifa;
4054 struct sk_buff *skb;
4055 int err;
4056
4057 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
4058 if (err < 0)
4059 goto errout;
4060
4061 addr = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
4062 if (addr == NULL) {
4063 err = -EINVAL;
4064 goto errout;
4065 }
4066
4067 ifm = nlmsg_data(nlh);
4068 if (ifm->ifa_index)
4069 dev = __dev_get_by_index(net, ifm->ifa_index);
4070
4071 ifa = ipv6_get_ifaddr(net, addr, dev, 1);
4072 if (!ifa) {
4073 err = -EADDRNOTAVAIL;
4074 goto errout;
4075 }
4076
4077 skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_KERNEL);
4078 if (!skb) {
4079 err = -ENOBUFS;
4080 goto errout_ifa;
4081 }
4082
4083 err = inet6_fill_ifaddr(skb, ifa, NETLINK_CB(in_skb).portid,
4084 nlh->nlmsg_seq, RTM_NEWADDR, 0);
4085 if (err < 0) {
4086 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
4087 WARN_ON(err == -EMSGSIZE);
4088 kfree_skb(skb);
4089 goto errout_ifa;
4090 }
4091 err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
4092 errout_ifa:
4093 in6_ifa_put(ifa);
4094 errout:
4095 return err;
4096 }
4097
4098 static void inet6_ifa_notify(int event, struct inet6_ifaddr *ifa)
4099 {
4100 struct sk_buff *skb;
4101 struct net *net = dev_net(ifa->idev->dev);
4102 int err = -ENOBUFS;
4103
4104 skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_ATOMIC);
4105 if (skb == NULL)
4106 goto errout;
4107
4108 err = inet6_fill_ifaddr(skb, ifa, 0, 0, event, 0);
4109 if (err < 0) {
4110 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
4111 WARN_ON(err == -EMSGSIZE);
4112 kfree_skb(skb);
4113 goto errout;
4114 }
4115 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
4116 return;
4117 errout:
4118 if (err < 0)
4119 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
4120 }
4121
4122 static inline void ipv6_store_devconf(struct ipv6_devconf *cnf,
4123 __s32 *array, int bytes)
4124 {
4125 BUG_ON(bytes < (DEVCONF_MAX * 4));
4126
4127 memset(array, 0, bytes);
4128 array[DEVCONF_FORWARDING] = cnf->forwarding;
4129 array[DEVCONF_HOPLIMIT] = cnf->hop_limit;
4130 array[DEVCONF_MTU6] = cnf->mtu6;
4131 array[DEVCONF_ACCEPT_RA] = cnf->accept_ra;
4132 array[DEVCONF_ACCEPT_REDIRECTS] = cnf->accept_redirects;
4133 array[DEVCONF_AUTOCONF] = cnf->autoconf;
4134 array[DEVCONF_DAD_TRANSMITS] = cnf->dad_transmits;
4135 array[DEVCONF_RTR_SOLICITS] = cnf->rtr_solicits;
4136 array[DEVCONF_RTR_SOLICIT_INTERVAL] =
4137 jiffies_to_msecs(cnf->rtr_solicit_interval);
4138 array[DEVCONF_RTR_SOLICIT_DELAY] =
4139 jiffies_to_msecs(cnf->rtr_solicit_delay);
4140 array[DEVCONF_FORCE_MLD_VERSION] = cnf->force_mld_version;
4141 #ifdef CONFIG_IPV6_PRIVACY
4142 array[DEVCONF_USE_TEMPADDR] = cnf->use_tempaddr;
4143 array[DEVCONF_TEMP_VALID_LFT] = cnf->temp_valid_lft;
4144 array[DEVCONF_TEMP_PREFERED_LFT] = cnf->temp_prefered_lft;
4145 array[DEVCONF_REGEN_MAX_RETRY] = cnf->regen_max_retry;
4146 array[DEVCONF_MAX_DESYNC_FACTOR] = cnf->max_desync_factor;
4147 #endif
4148 array[DEVCONF_MAX_ADDRESSES] = cnf->max_addresses;
4149 array[DEVCONF_ACCEPT_RA_DEFRTR] = cnf->accept_ra_defrtr;
4150 array[DEVCONF_ACCEPT_RA_PINFO] = cnf->accept_ra_pinfo;
4151 #ifdef CONFIG_IPV6_ROUTER_PREF
4152 array[DEVCONF_ACCEPT_RA_RTR_PREF] = cnf->accept_ra_rtr_pref;
4153 array[DEVCONF_RTR_PROBE_INTERVAL] =
4154 jiffies_to_msecs(cnf->rtr_probe_interval);
4155 #ifdef CONFIG_IPV6_ROUTE_INFO
4156 array[DEVCONF_ACCEPT_RA_RT_INFO_MAX_PLEN] = cnf->accept_ra_rt_info_max_plen;
4157 #endif
4158 #endif
4159 array[DEVCONF_PROXY_NDP] = cnf->proxy_ndp;
4160 array[DEVCONF_ACCEPT_SOURCE_ROUTE] = cnf->accept_source_route;
4161 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
4162 array[DEVCONF_OPTIMISTIC_DAD] = cnf->optimistic_dad;
4163 #endif
4164 #ifdef CONFIG_IPV6_MROUTE
4165 array[DEVCONF_MC_FORWARDING] = cnf->mc_forwarding;
4166 #endif
4167 array[DEVCONF_DISABLE_IPV6] = cnf->disable_ipv6;
4168 array[DEVCONF_ACCEPT_DAD] = cnf->accept_dad;
4169 array[DEVCONF_FORCE_TLLAO] = cnf->force_tllao;
4170 array[DEVCONF_NDISC_NOTIFY] = cnf->ndisc_notify;
4171 }
4172
4173 static inline size_t inet6_ifla6_size(void)
4174 {
4175 return nla_total_size(4) /* IFLA_INET6_FLAGS */
4176 + nla_total_size(sizeof(struct ifla_cacheinfo))
4177 + nla_total_size(DEVCONF_MAX * 4) /* IFLA_INET6_CONF */
4178 + nla_total_size(IPSTATS_MIB_MAX * 8) /* IFLA_INET6_STATS */
4179 + nla_total_size(ICMP6_MIB_MAX * 8) /* IFLA_INET6_ICMP6STATS */
4180 + nla_total_size(sizeof(struct in6_addr)); /* IFLA_INET6_TOKEN */
4181 }
4182
4183 static inline size_t inet6_if_nlmsg_size(void)
4184 {
4185 return NLMSG_ALIGN(sizeof(struct ifinfomsg))
4186 + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
4187 + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
4188 + nla_total_size(4) /* IFLA_MTU */
4189 + nla_total_size(4) /* IFLA_LINK */
4190 + nla_total_size(inet6_ifla6_size()); /* IFLA_PROTINFO */
4191 }
4192
4193 static inline void __snmp6_fill_statsdev(u64 *stats, atomic_long_t *mib,
4194 int items, int bytes)
4195 {
4196 int i;
4197 int pad = bytes - sizeof(u64) * items;
4198 BUG_ON(pad < 0);
4199
4200 /* Use put_unaligned() because stats may not be aligned for u64. */
4201 put_unaligned(items, &stats[0]);
4202 for (i = 1; i < items; i++)
4203 put_unaligned(atomic_long_read(&mib[i]), &stats[i]);
4204
4205 memset(&stats[items], 0, pad);
4206 }
4207
4208 static inline void __snmp6_fill_stats64(u64 *stats, void __percpu **mib,
4209 int items, int bytes, size_t syncpoff)
4210 {
4211 int i;
4212 int pad = bytes - sizeof(u64) * items;
4213 BUG_ON(pad < 0);
4214
4215 /* Use put_unaligned() because stats may not be aligned for u64. */
4216 put_unaligned(items, &stats[0]);
4217 for (i = 1; i < items; i++)
4218 put_unaligned(snmp_fold_field64(mib, i, syncpoff), &stats[i]);
4219
4220 memset(&stats[items], 0, pad);
4221 }
4222
4223 static void snmp6_fill_stats(u64 *stats, struct inet6_dev *idev, int attrtype,
4224 int bytes)
4225 {
4226 switch (attrtype) {
4227 case IFLA_INET6_STATS:
4228 __snmp6_fill_stats64(stats, (void __percpu **)idev->stats.ipv6,
4229 IPSTATS_MIB_MAX, bytes, offsetof(struct ipstats_mib, syncp));
4230 break;
4231 case IFLA_INET6_ICMP6STATS:
4232 __snmp6_fill_statsdev(stats, idev->stats.icmpv6dev->mibs, ICMP6_MIB_MAX, bytes);
4233 break;
4234 }
4235 }
4236
4237 static int inet6_fill_ifla6_attrs(struct sk_buff *skb, struct inet6_dev *idev)
4238 {
4239 struct nlattr *nla;
4240 struct ifla_cacheinfo ci;
4241
4242 if (nla_put_u32(skb, IFLA_INET6_FLAGS, idev->if_flags))
4243 goto nla_put_failure;
4244 ci.max_reasm_len = IPV6_MAXPLEN;
4245 ci.tstamp = cstamp_delta(idev->tstamp);
4246 ci.reachable_time = jiffies_to_msecs(idev->nd_parms->reachable_time);
4247 ci.retrans_time = jiffies_to_msecs(idev->nd_parms->retrans_time);
4248 if (nla_put(skb, IFLA_INET6_CACHEINFO, sizeof(ci), &ci))
4249 goto nla_put_failure;
4250 nla = nla_reserve(skb, IFLA_INET6_CONF, DEVCONF_MAX * sizeof(s32));
4251 if (nla == NULL)
4252 goto nla_put_failure;
4253 ipv6_store_devconf(&idev->cnf, nla_data(nla), nla_len(nla));
4254
4255 /* XXX - MC not implemented */
4256
4257 nla = nla_reserve(skb, IFLA_INET6_STATS, IPSTATS_MIB_MAX * sizeof(u64));
4258 if (nla == NULL)
4259 goto nla_put_failure;
4260 snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_STATS, nla_len(nla));
4261
4262 nla = nla_reserve(skb, IFLA_INET6_ICMP6STATS, ICMP6_MIB_MAX * sizeof(u64));
4263 if (nla == NULL)
4264 goto nla_put_failure;
4265 snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_ICMP6STATS, nla_len(nla));
4266
4267 nla = nla_reserve(skb, IFLA_INET6_TOKEN, sizeof(struct in6_addr));
4268 if (nla == NULL)
4269 goto nla_put_failure;
4270 read_lock_bh(&idev->lock);
4271 memcpy(nla_data(nla), idev->token.s6_addr, nla_len(nla));
4272 read_unlock_bh(&idev->lock);
4273
4274 return 0;
4275
4276 nla_put_failure:
4277 return -EMSGSIZE;
4278 }
4279
4280 static size_t inet6_get_link_af_size(const struct net_device *dev)
4281 {
4282 if (!__in6_dev_get(dev))
4283 return 0;
4284
4285 return inet6_ifla6_size();
4286 }
4287
4288 static int inet6_fill_link_af(struct sk_buff *skb, const struct net_device *dev)
4289 {
4290 struct inet6_dev *idev = __in6_dev_get(dev);
4291
4292 if (!idev)
4293 return -ENODATA;
4294
4295 if (inet6_fill_ifla6_attrs(skb, idev) < 0)
4296 return -EMSGSIZE;
4297
4298 return 0;
4299 }
4300
4301 static int inet6_set_iftoken(struct inet6_dev *idev, struct in6_addr *token)
4302 {
4303 struct inet6_ifaddr *ifp;
4304 struct net_device *dev = idev->dev;
4305 bool update_rs = false;
4306
4307 if (token == NULL)
4308 return -EINVAL;
4309 if (ipv6_addr_any(token))
4310 return -EINVAL;
4311 if (dev->flags & (IFF_LOOPBACK | IFF_NOARP))
4312 return -EINVAL;
4313 if (!ipv6_accept_ra(idev))
4314 return -EINVAL;
4315 if (idev->cnf.rtr_solicits <= 0)
4316 return -EINVAL;
4317
4318 write_lock_bh(&idev->lock);
4319
4320 BUILD_BUG_ON(sizeof(token->s6_addr) != 16);
4321 memcpy(idev->token.s6_addr + 8, token->s6_addr + 8, 8);
4322
4323 write_unlock_bh(&idev->lock);
4324
4325 if (!idev->dead && (idev->if_flags & IF_READY)) {
4326 struct in6_addr ll_addr;
4327
4328 ipv6_get_lladdr(dev, &ll_addr, IFA_F_TENTATIVE |
4329 IFA_F_OPTIMISTIC);
4330
4331 /* If we're not ready, then normal ifup will take care
4332 * of this. Otherwise, we need to request our rs here.
4333 */
4334 ndisc_send_rs(dev, &ll_addr, &in6addr_linklocal_allrouters);
4335 update_rs = true;
4336 }
4337
4338 write_lock_bh(&idev->lock);
4339
4340 if (update_rs)
4341 idev->if_flags |= IF_RS_SENT;
4342
4343 /* Well, that's kinda nasty ... */
4344 list_for_each_entry(ifp, &idev->addr_list, if_list) {
4345 spin_lock(&ifp->lock);
4346 if (ifp->tokenized) {
4347 ifp->valid_lft = 0;
4348 ifp->prefered_lft = 0;
4349 }
4350 spin_unlock(&ifp->lock);
4351 }
4352
4353 write_unlock_bh(&idev->lock);
4354 return 0;
4355 }
4356
4357 static int inet6_set_link_af(struct net_device *dev, const struct nlattr *nla)
4358 {
4359 int err = -EINVAL;
4360 struct inet6_dev *idev = __in6_dev_get(dev);
4361 struct nlattr *tb[IFLA_INET6_MAX + 1];
4362
4363 if (!idev)
4364 return -EAFNOSUPPORT;
4365
4366 if (nla_parse_nested(tb, IFLA_INET6_MAX, nla, NULL) < 0)
4367 BUG();
4368
4369 if (tb[IFLA_INET6_TOKEN])
4370 err = inet6_set_iftoken(idev, nla_data(tb[IFLA_INET6_TOKEN]));
4371
4372 return err;
4373 }
4374
4375 static int inet6_fill_ifinfo(struct sk_buff *skb, struct inet6_dev *idev,
4376 u32 portid, u32 seq, int event, unsigned int flags)
4377 {
4378 struct net_device *dev = idev->dev;
4379 struct ifinfomsg *hdr;
4380 struct nlmsghdr *nlh;
4381 void *protoinfo;
4382
4383 nlh = nlmsg_put(skb, portid, seq, event, sizeof(*hdr), flags);
4384 if (nlh == NULL)
4385 return -EMSGSIZE;
4386
4387 hdr = nlmsg_data(nlh);
4388 hdr->ifi_family = AF_INET6;
4389 hdr->__ifi_pad = 0;
4390 hdr->ifi_type = dev->type;
4391 hdr->ifi_index = dev->ifindex;
4392 hdr->ifi_flags = dev_get_flags(dev);
4393 hdr->ifi_change = 0;
4394
4395 if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
4396 (dev->addr_len &&
4397 nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
4398 nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
4399 (dev->ifindex != dev->iflink &&
4400 nla_put_u32(skb, IFLA_LINK, dev->iflink)))
4401 goto nla_put_failure;
4402 protoinfo = nla_nest_start(skb, IFLA_PROTINFO);
4403 if (protoinfo == NULL)
4404 goto nla_put_failure;
4405
4406 if (inet6_fill_ifla6_attrs(skb, idev) < 0)
4407 goto nla_put_failure;
4408
4409 nla_nest_end(skb, protoinfo);
4410 return nlmsg_end(skb, nlh);
4411
4412 nla_put_failure:
4413 nlmsg_cancel(skb, nlh);
4414 return -EMSGSIZE;
4415 }
4416
4417 static int inet6_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
4418 {
4419 struct net *net = sock_net(skb->sk);
4420 int h, s_h;
4421 int idx = 0, s_idx;
4422 struct net_device *dev;
4423 struct inet6_dev *idev;
4424 struct hlist_head *head;
4425
4426 s_h = cb->args[0];
4427 s_idx = cb->args[1];
4428
4429 rcu_read_lock();
4430 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
4431 idx = 0;
4432 head = &net->dev_index_head[h];
4433 hlist_for_each_entry_rcu(dev, head, index_hlist) {
4434 if (idx < s_idx)
4435 goto cont;
4436 idev = __in6_dev_get(dev);
4437 if (!idev)
4438 goto cont;
4439 if (inet6_fill_ifinfo(skb, idev,
4440 NETLINK_CB(cb->skb).portid,
4441 cb->nlh->nlmsg_seq,
4442 RTM_NEWLINK, NLM_F_MULTI) <= 0)
4443 goto out;
4444 cont:
4445 idx++;
4446 }
4447 }
4448 out:
4449 rcu_read_unlock();
4450 cb->args[1] = idx;
4451 cb->args[0] = h;
4452
4453 return skb->len;
4454 }
4455
4456 void inet6_ifinfo_notify(int event, struct inet6_dev *idev)
4457 {
4458 struct sk_buff *skb;
4459 struct net *net = dev_net(idev->dev);
4460 int err = -ENOBUFS;
4461
4462 skb = nlmsg_new(inet6_if_nlmsg_size(), GFP_ATOMIC);
4463 if (skb == NULL)
4464 goto errout;
4465
4466 err = inet6_fill_ifinfo(skb, idev, 0, 0, event, 0);
4467 if (err < 0) {
4468 /* -EMSGSIZE implies BUG in inet6_if_nlmsg_size() */
4469 WARN_ON(err == -EMSGSIZE);
4470 kfree_skb(skb);
4471 goto errout;
4472 }
4473 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFINFO, NULL, GFP_ATOMIC);
4474 return;
4475 errout:
4476 if (err < 0)
4477 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFINFO, err);
4478 }
4479
4480 static inline size_t inet6_prefix_nlmsg_size(void)
4481 {
4482 return NLMSG_ALIGN(sizeof(struct prefixmsg))
4483 + nla_total_size(sizeof(struct in6_addr))
4484 + nla_total_size(sizeof(struct prefix_cacheinfo));
4485 }
4486
4487 static int inet6_fill_prefix(struct sk_buff *skb, struct inet6_dev *idev,
4488 struct prefix_info *pinfo, u32 portid, u32 seq,
4489 int event, unsigned int flags)
4490 {
4491 struct prefixmsg *pmsg;
4492 struct nlmsghdr *nlh;
4493 struct prefix_cacheinfo ci;
4494
4495 nlh = nlmsg_put(skb, portid, seq, event, sizeof(*pmsg), flags);
4496 if (nlh == NULL)
4497 return -EMSGSIZE;
4498
4499 pmsg = nlmsg_data(nlh);
4500 pmsg->prefix_family = AF_INET6;
4501 pmsg->prefix_pad1 = 0;
4502 pmsg->prefix_pad2 = 0;
4503 pmsg->prefix_ifindex = idev->dev->ifindex;
4504 pmsg->prefix_len = pinfo->prefix_len;
4505 pmsg->prefix_type = pinfo->type;
4506 pmsg->prefix_pad3 = 0;
4507 pmsg->prefix_flags = 0;
4508 if (pinfo->onlink)
4509 pmsg->prefix_flags |= IF_PREFIX_ONLINK;
4510 if (pinfo->autoconf)
4511 pmsg->prefix_flags |= IF_PREFIX_AUTOCONF;
4512
4513 if (nla_put(skb, PREFIX_ADDRESS, sizeof(pinfo->prefix), &pinfo->prefix))
4514 goto nla_put_failure;
4515 ci.preferred_time = ntohl(pinfo->prefered);
4516 ci.valid_time = ntohl(pinfo->valid);
4517 if (nla_put(skb, PREFIX_CACHEINFO, sizeof(ci), &ci))
4518 goto nla_put_failure;
4519 return nlmsg_end(skb, nlh);
4520
4521 nla_put_failure:
4522 nlmsg_cancel(skb, nlh);
4523 return -EMSGSIZE;
4524 }
4525
4526 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
4527 struct prefix_info *pinfo)
4528 {
4529 struct sk_buff *skb;
4530 struct net *net = dev_net(idev->dev);
4531 int err = -ENOBUFS;
4532
4533 skb = nlmsg_new(inet6_prefix_nlmsg_size(), GFP_ATOMIC);
4534 if (skb == NULL)
4535 goto errout;
4536
4537 err = inet6_fill_prefix(skb, idev, pinfo, 0, 0, event, 0);
4538 if (err < 0) {
4539 /* -EMSGSIZE implies BUG in inet6_prefix_nlmsg_size() */
4540 WARN_ON(err == -EMSGSIZE);
4541 kfree_skb(skb);
4542 goto errout;
4543 }
4544 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_PREFIX, NULL, GFP_ATOMIC);
4545 return;
4546 errout:
4547 if (err < 0)
4548 rtnl_set_sk_err(net, RTNLGRP_IPV6_PREFIX, err);
4549 }
4550
4551 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
4552 {
4553 struct net *net = dev_net(ifp->idev->dev);
4554
4555 inet6_ifa_notify(event ? : RTM_NEWADDR, ifp);
4556
4557 switch (event) {
4558 case RTM_NEWADDR:
4559 /*
4560 * If the address was optimistic
4561 * we inserted the route at the start of
4562 * our DAD process, so we don't need
4563 * to do it again
4564 */
4565 if (!(ifp->rt->rt6i_node))
4566 ip6_ins_rt(ifp->rt);
4567 if (ifp->idev->cnf.forwarding)
4568 addrconf_join_anycast(ifp);
4569 break;
4570 case RTM_DELADDR:
4571 if (ifp->idev->cnf.forwarding)
4572 addrconf_leave_anycast(ifp);
4573 addrconf_leave_solict(ifp->idev, &ifp->addr);
4574 dst_hold(&ifp->rt->dst);
4575
4576 if (ip6_del_rt(ifp->rt))
4577 dst_free(&ifp->rt->dst);
4578 break;
4579 }
4580 atomic_inc(&net->ipv6.dev_addr_genid);
4581 }
4582
4583 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
4584 {
4585 rcu_read_lock_bh();
4586 if (likely(ifp->idev->dead == 0))
4587 __ipv6_ifa_notify(event, ifp);
4588 rcu_read_unlock_bh();
4589 }
4590
4591 #ifdef CONFIG_SYSCTL
4592
4593 static
4594 int addrconf_sysctl_forward(ctl_table *ctl, int write,
4595 void __user *buffer, size_t *lenp, loff_t *ppos)
4596 {
4597 int *valp = ctl->data;
4598 int val = *valp;
4599 loff_t pos = *ppos;
4600 ctl_table lctl;
4601 int ret;
4602
4603 /*
4604 * ctl->data points to idev->cnf.forwarding, we should
4605 * not modify it until we get the rtnl lock.
4606 */
4607 lctl = *ctl;
4608 lctl.data = &val;
4609
4610 ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
4611
4612 if (write)
4613 ret = addrconf_fixup_forwarding(ctl, valp, val);
4614 if (ret)
4615 *ppos = pos;
4616 return ret;
4617 }
4618
4619 static void dev_disable_change(struct inet6_dev *idev)
4620 {
4621 if (!idev || !idev->dev)
4622 return;
4623
4624 if (idev->cnf.disable_ipv6)
4625 addrconf_notify(NULL, NETDEV_DOWN, idev->dev);
4626 else
4627 addrconf_notify(NULL, NETDEV_UP, idev->dev);
4628 }
4629
4630 static void addrconf_disable_change(struct net *net, __s32 newf)
4631 {
4632 struct net_device *dev;
4633 struct inet6_dev *idev;
4634
4635 rcu_read_lock();
4636 for_each_netdev_rcu(net, dev) {
4637 idev = __in6_dev_get(dev);
4638 if (idev) {
4639 int changed = (!idev->cnf.disable_ipv6) ^ (!newf);
4640 idev->cnf.disable_ipv6 = newf;
4641 if (changed)
4642 dev_disable_change(idev);
4643 }
4644 }
4645 rcu_read_unlock();
4646 }
4647
4648 static int addrconf_disable_ipv6(struct ctl_table *table, int *p, int newf)
4649 {
4650 struct net *net;
4651 int old;
4652
4653 if (!rtnl_trylock())
4654 return restart_syscall();
4655
4656 net = (struct net *)table->extra2;
4657 old = *p;
4658 *p = newf;
4659
4660 if (p == &net->ipv6.devconf_dflt->disable_ipv6) {
4661 rtnl_unlock();
4662 return 0;
4663 }
4664
4665 if (p == &net->ipv6.devconf_all->disable_ipv6) {
4666 net->ipv6.devconf_dflt->disable_ipv6 = newf;
4667 addrconf_disable_change(net, newf);
4668 } else if ((!newf) ^ (!old))
4669 dev_disable_change((struct inet6_dev *)table->extra1);
4670
4671 rtnl_unlock();
4672 return 0;
4673 }
4674
4675 static
4676 int addrconf_sysctl_disable(ctl_table *ctl, int write,
4677 void __user *buffer, size_t *lenp, loff_t *ppos)
4678 {
4679 int *valp = ctl->data;
4680 int val = *valp;
4681 loff_t pos = *ppos;
4682 ctl_table lctl;
4683 int ret;
4684
4685 /*
4686 * ctl->data points to idev->cnf.disable_ipv6, we should
4687 * not modify it until we get the rtnl lock.
4688 */
4689 lctl = *ctl;
4690 lctl.data = &val;
4691
4692 ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
4693
4694 if (write)
4695 ret = addrconf_disable_ipv6(ctl, valp, val);
4696 if (ret)
4697 *ppos = pos;
4698 return ret;
4699 }
4700
4701 static struct addrconf_sysctl_table
4702 {
4703 struct ctl_table_header *sysctl_header;
4704 ctl_table addrconf_vars[DEVCONF_MAX+1];
4705 } addrconf_sysctl __read_mostly = {
4706 .sysctl_header = NULL,
4707 .addrconf_vars = {
4708 {
4709 .procname = "forwarding",
4710 .data = &ipv6_devconf.forwarding,
4711 .maxlen = sizeof(int),
4712 .mode = 0644,
4713 .proc_handler = addrconf_sysctl_forward,
4714 },
4715 {
4716 .procname = "hop_limit",
4717 .data = &ipv6_devconf.hop_limit,
4718 .maxlen = sizeof(int),
4719 .mode = 0644,
4720 .proc_handler = proc_dointvec,
4721 },
4722 {
4723 .procname = "mtu",
4724 .data = &ipv6_devconf.mtu6,
4725 .maxlen = sizeof(int),
4726 .mode = 0644,
4727 .proc_handler = proc_dointvec,
4728 },
4729 {
4730 .procname = "accept_ra",
4731 .data = &ipv6_devconf.accept_ra,
4732 .maxlen = sizeof(int),
4733 .mode = 0644,
4734 .proc_handler = proc_dointvec,
4735 },
4736 {
4737 .procname = "accept_redirects",
4738 .data = &ipv6_devconf.accept_redirects,
4739 .maxlen = sizeof(int),
4740 .mode = 0644,
4741 .proc_handler = proc_dointvec,
4742 },
4743 {
4744 .procname = "autoconf",
4745 .data = &ipv6_devconf.autoconf,
4746 .maxlen = sizeof(int),
4747 .mode = 0644,
4748 .proc_handler = proc_dointvec,
4749 },
4750 {
4751 .procname = "dad_transmits",
4752 .data = &ipv6_devconf.dad_transmits,
4753 .maxlen = sizeof(int),
4754 .mode = 0644,
4755 .proc_handler = proc_dointvec,
4756 },
4757 {
4758 .procname = "router_solicitations",
4759 .data = &ipv6_devconf.rtr_solicits,
4760 .maxlen = sizeof(int),
4761 .mode = 0644,
4762 .proc_handler = proc_dointvec,
4763 },
4764 {
4765 .procname = "router_solicitation_interval",
4766 .data = &ipv6_devconf.rtr_solicit_interval,
4767 .maxlen = sizeof(int),
4768 .mode = 0644,
4769 .proc_handler = proc_dointvec_jiffies,
4770 },
4771 {
4772 .procname = "router_solicitation_delay",
4773 .data = &ipv6_devconf.rtr_solicit_delay,
4774 .maxlen = sizeof(int),
4775 .mode = 0644,
4776 .proc_handler = proc_dointvec_jiffies,
4777 },
4778 {
4779 .procname = "force_mld_version",
4780 .data = &ipv6_devconf.force_mld_version,
4781 .maxlen = sizeof(int),
4782 .mode = 0644,
4783 .proc_handler = proc_dointvec,
4784 },
4785 #ifdef CONFIG_IPV6_PRIVACY
4786 {
4787 .procname = "use_tempaddr",
4788 .data = &ipv6_devconf.use_tempaddr,
4789 .maxlen = sizeof(int),
4790 .mode = 0644,
4791 .proc_handler = proc_dointvec,
4792 },
4793 {
4794 .procname = "temp_valid_lft",
4795 .data = &ipv6_devconf.temp_valid_lft,
4796 .maxlen = sizeof(int),
4797 .mode = 0644,
4798 .proc_handler = proc_dointvec,
4799 },
4800 {
4801 .procname = "temp_prefered_lft",
4802 .data = &ipv6_devconf.temp_prefered_lft,
4803 .maxlen = sizeof(int),
4804 .mode = 0644,
4805 .proc_handler = proc_dointvec,
4806 },
4807 {
4808 .procname = "regen_max_retry",
4809 .data = &ipv6_devconf.regen_max_retry,
4810 .maxlen = sizeof(int),
4811 .mode = 0644,
4812 .proc_handler = proc_dointvec,
4813 },
4814 {
4815 .procname = "max_desync_factor",
4816 .data = &ipv6_devconf.max_desync_factor,
4817 .maxlen = sizeof(int),
4818 .mode = 0644,
4819 .proc_handler = proc_dointvec,
4820 },
4821 #endif
4822 {
4823 .procname = "max_addresses",
4824 .data = &ipv6_devconf.max_addresses,
4825 .maxlen = sizeof(int),
4826 .mode = 0644,
4827 .proc_handler = proc_dointvec,
4828 },
4829 {
4830 .procname = "accept_ra_defrtr",
4831 .data = &ipv6_devconf.accept_ra_defrtr,
4832 .maxlen = sizeof(int),
4833 .mode = 0644,
4834 .proc_handler = proc_dointvec,
4835 },
4836 {
4837 .procname = "accept_ra_pinfo",
4838 .data = &ipv6_devconf.accept_ra_pinfo,
4839 .maxlen = sizeof(int),
4840 .mode = 0644,
4841 .proc_handler = proc_dointvec,
4842 },
4843 #ifdef CONFIG_IPV6_ROUTER_PREF
4844 {
4845 .procname = "accept_ra_rtr_pref",
4846 .data = &ipv6_devconf.accept_ra_rtr_pref,
4847 .maxlen = sizeof(int),
4848 .mode = 0644,
4849 .proc_handler = proc_dointvec,
4850 },
4851 {
4852 .procname = "router_probe_interval",
4853 .data = &ipv6_devconf.rtr_probe_interval,
4854 .maxlen = sizeof(int),
4855 .mode = 0644,
4856 .proc_handler = proc_dointvec_jiffies,
4857 },
4858 #ifdef CONFIG_IPV6_ROUTE_INFO
4859 {
4860 .procname = "accept_ra_rt_info_max_plen",
4861 .data = &ipv6_devconf.accept_ra_rt_info_max_plen,
4862 .maxlen = sizeof(int),
4863 .mode = 0644,
4864 .proc_handler = proc_dointvec,
4865 },
4866 #endif
4867 #endif
4868 {
4869 .procname = "proxy_ndp",
4870 .data = &ipv6_devconf.proxy_ndp,
4871 .maxlen = sizeof(int),
4872 .mode = 0644,
4873 .proc_handler = proc_dointvec,
4874 },
4875 {
4876 .procname = "accept_source_route",
4877 .data = &ipv6_devconf.accept_source_route,
4878 .maxlen = sizeof(int),
4879 .mode = 0644,
4880 .proc_handler = proc_dointvec,
4881 },
4882 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
4883 {
4884 .procname = "optimistic_dad",
4885 .data = &ipv6_devconf.optimistic_dad,
4886 .maxlen = sizeof(int),
4887 .mode = 0644,
4888 .proc_handler = proc_dointvec,
4889
4890 },
4891 #endif
4892 #ifdef CONFIG_IPV6_MROUTE
4893 {
4894 .procname = "mc_forwarding",
4895 .data = &ipv6_devconf.mc_forwarding,
4896 .maxlen = sizeof(int),
4897 .mode = 0444,
4898 .proc_handler = proc_dointvec,
4899 },
4900 #endif
4901 {
4902 .procname = "disable_ipv6",
4903 .data = &ipv6_devconf.disable_ipv6,
4904 .maxlen = sizeof(int),
4905 .mode = 0644,
4906 .proc_handler = addrconf_sysctl_disable,
4907 },
4908 {
4909 .procname = "accept_dad",
4910 .data = &ipv6_devconf.accept_dad,
4911 .maxlen = sizeof(int),
4912 .mode = 0644,
4913 .proc_handler = proc_dointvec,
4914 },
4915 {
4916 .procname = "force_tllao",
4917 .data = &ipv6_devconf.force_tllao,
4918 .maxlen = sizeof(int),
4919 .mode = 0644,
4920 .proc_handler = proc_dointvec
4921 },
4922 {
4923 .procname = "ndisc_notify",
4924 .data = &ipv6_devconf.ndisc_notify,
4925 .maxlen = sizeof(int),
4926 .mode = 0644,
4927 .proc_handler = proc_dointvec
4928 },
4929 {
4930 /* sentinel */
4931 }
4932 },
4933 };
4934
4935 static int __addrconf_sysctl_register(struct net *net, char *dev_name,
4936 struct inet6_dev *idev, struct ipv6_devconf *p)
4937 {
4938 int i;
4939 struct addrconf_sysctl_table *t;
4940 char path[sizeof("net/ipv6/conf/") + IFNAMSIZ];
4941
4942 t = kmemdup(&addrconf_sysctl, sizeof(*t), GFP_KERNEL);
4943 if (t == NULL)
4944 goto out;
4945
4946 for (i = 0; t->addrconf_vars[i].data; i++) {
4947 t->addrconf_vars[i].data += (char *)p - (char *)&ipv6_devconf;
4948 t->addrconf_vars[i].extra1 = idev; /* embedded; no ref */
4949 t->addrconf_vars[i].extra2 = net;
4950 }
4951
4952 snprintf(path, sizeof(path), "net/ipv6/conf/%s", dev_name);
4953
4954 t->sysctl_header = register_net_sysctl(net, path, t->addrconf_vars);
4955 if (t->sysctl_header == NULL)
4956 goto free;
4957
4958 p->sysctl = t;
4959 return 0;
4960
4961 free:
4962 kfree(t);
4963 out:
4964 return -ENOBUFS;
4965 }
4966
4967 static void __addrconf_sysctl_unregister(struct ipv6_devconf *p)
4968 {
4969 struct addrconf_sysctl_table *t;
4970
4971 if (p->sysctl == NULL)
4972 return;
4973
4974 t = p->sysctl;
4975 p->sysctl = NULL;
4976 unregister_net_sysctl_table(t->sysctl_header);
4977 kfree(t);
4978 }
4979
4980 static void addrconf_sysctl_register(struct inet6_dev *idev)
4981 {
4982 neigh_sysctl_register(idev->dev, idev->nd_parms, "ipv6",
4983 &ndisc_ifinfo_sysctl_change);
4984 __addrconf_sysctl_register(dev_net(idev->dev), idev->dev->name,
4985 idev, &idev->cnf);
4986 }
4987
4988 static void addrconf_sysctl_unregister(struct inet6_dev *idev)
4989 {
4990 __addrconf_sysctl_unregister(&idev->cnf);
4991 neigh_sysctl_unregister(idev->nd_parms);
4992 }
4993
4994
4995 #endif
4996
4997 static int __net_init addrconf_init_net(struct net *net)
4998 {
4999 int err = -ENOMEM;
5000 struct ipv6_devconf *all, *dflt;
5001
5002 all = kmemdup(&ipv6_devconf, sizeof(ipv6_devconf), GFP_KERNEL);
5003 if (all == NULL)
5004 goto err_alloc_all;
5005
5006 dflt = kmemdup(&ipv6_devconf_dflt, sizeof(ipv6_devconf_dflt), GFP_KERNEL);
5007 if (dflt == NULL)
5008 goto err_alloc_dflt;
5009
5010 /* these will be inherited by all namespaces */
5011 dflt->autoconf = ipv6_defaults.autoconf;
5012 dflt->disable_ipv6 = ipv6_defaults.disable_ipv6;
5013
5014 net->ipv6.devconf_all = all;
5015 net->ipv6.devconf_dflt = dflt;
5016
5017 #ifdef CONFIG_SYSCTL
5018 err = __addrconf_sysctl_register(net, "all", NULL, all);
5019 if (err < 0)
5020 goto err_reg_all;
5021
5022 err = __addrconf_sysctl_register(net, "default", NULL, dflt);
5023 if (err < 0)
5024 goto err_reg_dflt;
5025 #endif
5026 return 0;
5027
5028 #ifdef CONFIG_SYSCTL
5029 err_reg_dflt:
5030 __addrconf_sysctl_unregister(all);
5031 err_reg_all:
5032 kfree(dflt);
5033 #endif
5034 err_alloc_dflt:
5035 kfree(all);
5036 err_alloc_all:
5037 return err;
5038 }
5039
5040 static void __net_exit addrconf_exit_net(struct net *net)
5041 {
5042 #ifdef CONFIG_SYSCTL
5043 __addrconf_sysctl_unregister(net->ipv6.devconf_dflt);
5044 __addrconf_sysctl_unregister(net->ipv6.devconf_all);
5045 #endif
5046 if (!net_eq(net, &init_net)) {
5047 kfree(net->ipv6.devconf_dflt);
5048 kfree(net->ipv6.devconf_all);
5049 }
5050 }
5051
5052 static struct pernet_operations addrconf_ops = {
5053 .init = addrconf_init_net,
5054 .exit = addrconf_exit_net,
5055 };
5056
5057 static struct rtnl_af_ops inet6_ops = {
5058 .family = AF_INET6,
5059 .fill_link_af = inet6_fill_link_af,
5060 .get_link_af_size = inet6_get_link_af_size,
5061 .set_link_af = inet6_set_link_af,
5062 };
5063
5064 /*
5065 * Init / cleanup code
5066 */
5067
5068 int __init addrconf_init(void)
5069 {
5070 int i, err;
5071
5072 err = ipv6_addr_label_init();
5073 if (err < 0) {
5074 pr_crit("%s: cannot initialize default policy table: %d\n",
5075 __func__, err);
5076 goto out;
5077 }
5078
5079 err = register_pernet_subsys(&addrconf_ops);
5080 if (err < 0)
5081 goto out_addrlabel;
5082
5083 /* The addrconf netdev notifier requires that loopback_dev
5084 * has it's ipv6 private information allocated and setup
5085 * before it can bring up and give link-local addresses
5086 * to other devices which are up.
5087 *
5088 * Unfortunately, loopback_dev is not necessarily the first
5089 * entry in the global dev_base list of net devices. In fact,
5090 * it is likely to be the very last entry on that list.
5091 * So this causes the notifier registry below to try and
5092 * give link-local addresses to all devices besides loopback_dev
5093 * first, then loopback_dev, which cases all the non-loopback_dev
5094 * devices to fail to get a link-local address.
5095 *
5096 * So, as a temporary fix, allocate the ipv6 structure for
5097 * loopback_dev first by hand.
5098 * Longer term, all of the dependencies ipv6 has upon the loopback
5099 * device and it being up should be removed.
5100 */
5101 rtnl_lock();
5102 if (!ipv6_add_dev(init_net.loopback_dev))
5103 err = -ENOMEM;
5104 rtnl_unlock();
5105 if (err)
5106 goto errlo;
5107
5108 for (i = 0; i < IN6_ADDR_HSIZE; i++)
5109 INIT_HLIST_HEAD(&inet6_addr_lst[i]);
5110
5111 register_netdevice_notifier(&ipv6_dev_notf);
5112
5113 addrconf_verify(0);
5114
5115 err = rtnl_af_register(&inet6_ops);
5116 if (err < 0)
5117 goto errout_af;
5118
5119 err = __rtnl_register(PF_INET6, RTM_GETLINK, NULL, inet6_dump_ifinfo,
5120 NULL);
5121 if (err < 0)
5122 goto errout;
5123
5124 /* Only the first call to __rtnl_register can fail */
5125 __rtnl_register(PF_INET6, RTM_NEWADDR, inet6_rtm_newaddr, NULL, NULL);
5126 __rtnl_register(PF_INET6, RTM_DELADDR, inet6_rtm_deladdr, NULL, NULL);
5127 __rtnl_register(PF_INET6, RTM_GETADDR, inet6_rtm_getaddr,
5128 inet6_dump_ifaddr, NULL);
5129 __rtnl_register(PF_INET6, RTM_GETMULTICAST, NULL,
5130 inet6_dump_ifmcaddr, NULL);
5131 __rtnl_register(PF_INET6, RTM_GETANYCAST, NULL,
5132 inet6_dump_ifacaddr, NULL);
5133 __rtnl_register(PF_INET6, RTM_GETNETCONF, inet6_netconf_get_devconf,
5134 inet6_netconf_dump_devconf, NULL);
5135
5136 ipv6_addr_label_rtnl_register();
5137
5138 return 0;
5139 errout:
5140 rtnl_af_unregister(&inet6_ops);
5141 errout_af:
5142 unregister_netdevice_notifier(&ipv6_dev_notf);
5143 errlo:
5144 unregister_pernet_subsys(&addrconf_ops);
5145 out_addrlabel:
5146 ipv6_addr_label_cleanup();
5147 out:
5148 return err;
5149 }
5150
5151 void addrconf_cleanup(void)
5152 {
5153 struct net_device *dev;
5154 int i;
5155
5156 unregister_netdevice_notifier(&ipv6_dev_notf);
5157 unregister_pernet_subsys(&addrconf_ops);
5158 ipv6_addr_label_cleanup();
5159
5160 rtnl_lock();
5161
5162 __rtnl_af_unregister(&inet6_ops);
5163
5164 /* clean dev list */
5165 for_each_netdev(&init_net, dev) {
5166 if (__in6_dev_get(dev) == NULL)
5167 continue;
5168 addrconf_ifdown(dev, 1);
5169 }
5170 addrconf_ifdown(init_net.loopback_dev, 2);
5171
5172 /*
5173 * Check hash table.
5174 */
5175 spin_lock_bh(&addrconf_hash_lock);
5176 for (i = 0; i < IN6_ADDR_HSIZE; i++)
5177 WARN_ON(!hlist_empty(&inet6_addr_lst[i]));
5178 spin_unlock_bh(&addrconf_hash_lock);
5179
5180 del_timer(&addr_chk_timer);
5181 rtnl_unlock();
5182 }