Merge remote-tracking branch 'pfdo/drm-fixes' into drm-next
[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 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 ip6_ins_rt(sp_rt);
2663 }
2664 read_unlock_bh(&idev->lock);
2665 }
2666 }
2667
2668 static void addrconf_add_linklocal(struct inet6_dev *idev, const struct in6_addr *addr)
2669 {
2670 struct inet6_ifaddr *ifp;
2671 u32 addr_flags = IFA_F_PERMANENT;
2672
2673 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2674 if (idev->cnf.optimistic_dad &&
2675 !dev_net(idev->dev)->ipv6.devconf_all->forwarding)
2676 addr_flags |= IFA_F_OPTIMISTIC;
2677 #endif
2678
2679
2680 ifp = ipv6_add_addr(idev, addr, 64, IFA_LINK, addr_flags);
2681 if (!IS_ERR(ifp)) {
2682 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, idev->dev, 0, 0);
2683 addrconf_dad_start(ifp);
2684 in6_ifa_put(ifp);
2685 }
2686 }
2687
2688 static void addrconf_dev_config(struct net_device *dev)
2689 {
2690 struct in6_addr addr;
2691 struct inet6_dev *idev;
2692
2693 ASSERT_RTNL();
2694
2695 if ((dev->type != ARPHRD_ETHER) &&
2696 (dev->type != ARPHRD_FDDI) &&
2697 (dev->type != ARPHRD_ARCNET) &&
2698 (dev->type != ARPHRD_INFINIBAND) &&
2699 (dev->type != ARPHRD_IEEE802154) &&
2700 (dev->type != ARPHRD_IEEE1394)) {
2701 /* Alas, we support only Ethernet autoconfiguration. */
2702 return;
2703 }
2704
2705 idev = addrconf_add_dev(dev);
2706 if (IS_ERR(idev))
2707 return;
2708
2709 memset(&addr, 0, sizeof(struct in6_addr));
2710 addr.s6_addr32[0] = htonl(0xFE800000);
2711
2712 if (ipv6_generate_eui64(addr.s6_addr + 8, dev) == 0)
2713 addrconf_add_linklocal(idev, &addr);
2714 }
2715
2716 #if IS_ENABLED(CONFIG_IPV6_SIT)
2717 static void addrconf_sit_config(struct net_device *dev)
2718 {
2719 struct inet6_dev *idev;
2720
2721 ASSERT_RTNL();
2722
2723 /*
2724 * Configure the tunnel with one of our IPv4
2725 * addresses... we should configure all of
2726 * our v4 addrs in the tunnel
2727 */
2728
2729 if ((idev = ipv6_find_idev(dev)) == NULL) {
2730 pr_debug("%s: add_dev failed\n", __func__);
2731 return;
2732 }
2733
2734 if (dev->priv_flags & IFF_ISATAP) {
2735 struct in6_addr addr;
2736
2737 ipv6_addr_set(&addr, htonl(0xFE800000), 0, 0, 0);
2738 addrconf_prefix_route(&addr, 64, dev, 0, 0);
2739 if (!ipv6_generate_eui64(addr.s6_addr + 8, dev))
2740 addrconf_add_linklocal(idev, &addr);
2741 return;
2742 }
2743
2744 sit_add_v4_addrs(idev);
2745
2746 if (dev->flags&IFF_POINTOPOINT)
2747 addrconf_add_mroute(dev);
2748 else
2749 sit_route_add(dev);
2750 }
2751 #endif
2752
2753 #if IS_ENABLED(CONFIG_NET_IPGRE)
2754 static void addrconf_gre_config(struct net_device *dev)
2755 {
2756 struct inet6_dev *idev;
2757 struct in6_addr addr;
2758
2759 pr_info("%s(%s)\n", __func__, dev->name);
2760
2761 ASSERT_RTNL();
2762
2763 if ((idev = ipv6_find_idev(dev)) == NULL) {
2764 pr_debug("%s: add_dev failed\n", __func__);
2765 return;
2766 }
2767
2768 ipv6_addr_set(&addr, htonl(0xFE800000), 0, 0, 0);
2769 addrconf_prefix_route(&addr, 64, dev, 0, 0);
2770
2771 if (!ipv6_generate_eui64(addr.s6_addr + 8, dev))
2772 addrconf_add_linklocal(idev, &addr);
2773 }
2774 #endif
2775
2776 static inline int
2777 ipv6_inherit_linklocal(struct inet6_dev *idev, struct net_device *link_dev)
2778 {
2779 struct in6_addr lladdr;
2780
2781 if (!ipv6_get_lladdr(link_dev, &lladdr, IFA_F_TENTATIVE)) {
2782 addrconf_add_linklocal(idev, &lladdr);
2783 return 0;
2784 }
2785 return -1;
2786 }
2787
2788 static void ip6_tnl_add_linklocal(struct inet6_dev *idev)
2789 {
2790 struct net_device *link_dev;
2791 struct net *net = dev_net(idev->dev);
2792
2793 /* first try to inherit the link-local address from the link device */
2794 if (idev->dev->iflink &&
2795 (link_dev = __dev_get_by_index(net, idev->dev->iflink))) {
2796 if (!ipv6_inherit_linklocal(idev, link_dev))
2797 return;
2798 }
2799 /* then try to inherit it from any device */
2800 for_each_netdev(net, link_dev) {
2801 if (!ipv6_inherit_linklocal(idev, link_dev))
2802 return;
2803 }
2804 pr_debug("init ip6-ip6: add_linklocal failed\n");
2805 }
2806
2807 /*
2808 * Autoconfigure tunnel with a link-local address so routing protocols,
2809 * DHCPv6, MLD etc. can be run over the virtual link
2810 */
2811
2812 static void addrconf_ip6_tnl_config(struct net_device *dev)
2813 {
2814 struct inet6_dev *idev;
2815
2816 ASSERT_RTNL();
2817
2818 idev = addrconf_add_dev(dev);
2819 if (IS_ERR(idev)) {
2820 pr_debug("init ip6-ip6: add_dev failed\n");
2821 return;
2822 }
2823 ip6_tnl_add_linklocal(idev);
2824 }
2825
2826 static int addrconf_notify(struct notifier_block *this, unsigned long event,
2827 void *data)
2828 {
2829 struct net_device *dev = (struct net_device *) data;
2830 struct inet6_dev *idev = __in6_dev_get(dev);
2831 int run_pending = 0;
2832 int err;
2833
2834 switch (event) {
2835 case NETDEV_REGISTER:
2836 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2837 idev = ipv6_add_dev(dev);
2838 if (!idev)
2839 return notifier_from_errno(-ENOMEM);
2840 }
2841 break;
2842
2843 case NETDEV_UP:
2844 case NETDEV_CHANGE:
2845 if (dev->flags & IFF_SLAVE)
2846 break;
2847
2848 if (event == NETDEV_UP) {
2849 if (!addrconf_qdisc_ok(dev)) {
2850 /* device is not ready yet. */
2851 pr_info("ADDRCONF(NETDEV_UP): %s: link is not ready\n",
2852 dev->name);
2853 break;
2854 }
2855
2856 if (!idev && dev->mtu >= IPV6_MIN_MTU)
2857 idev = ipv6_add_dev(dev);
2858
2859 if (idev) {
2860 idev->if_flags |= IF_READY;
2861 run_pending = 1;
2862 }
2863 } else {
2864 if (!addrconf_qdisc_ok(dev)) {
2865 /* device is still not ready. */
2866 break;
2867 }
2868
2869 if (idev) {
2870 if (idev->if_flags & IF_READY)
2871 /* device is already configured. */
2872 break;
2873 idev->if_flags |= IF_READY;
2874 }
2875
2876 pr_info("ADDRCONF(NETDEV_CHANGE): %s: link becomes ready\n",
2877 dev->name);
2878
2879 run_pending = 1;
2880 }
2881
2882 switch (dev->type) {
2883 #if IS_ENABLED(CONFIG_IPV6_SIT)
2884 case ARPHRD_SIT:
2885 addrconf_sit_config(dev);
2886 break;
2887 #endif
2888 #if IS_ENABLED(CONFIG_NET_IPGRE)
2889 case ARPHRD_IPGRE:
2890 addrconf_gre_config(dev);
2891 break;
2892 #endif
2893 case ARPHRD_TUNNEL6:
2894 addrconf_ip6_tnl_config(dev);
2895 break;
2896 case ARPHRD_LOOPBACK:
2897 init_loopback(dev);
2898 break;
2899
2900 default:
2901 addrconf_dev_config(dev);
2902 break;
2903 }
2904
2905 if (idev) {
2906 if (run_pending)
2907 addrconf_dad_run(idev);
2908
2909 /*
2910 * If the MTU changed during the interface down,
2911 * when the interface up, the changed MTU must be
2912 * reflected in the idev as well as routers.
2913 */
2914 if (idev->cnf.mtu6 != dev->mtu &&
2915 dev->mtu >= IPV6_MIN_MTU) {
2916 rt6_mtu_change(dev, dev->mtu);
2917 idev->cnf.mtu6 = dev->mtu;
2918 }
2919 idev->tstamp = jiffies;
2920 inet6_ifinfo_notify(RTM_NEWLINK, idev);
2921
2922 /*
2923 * If the changed mtu during down is lower than
2924 * IPV6_MIN_MTU stop IPv6 on this interface.
2925 */
2926 if (dev->mtu < IPV6_MIN_MTU)
2927 addrconf_ifdown(dev, 1);
2928 }
2929 break;
2930
2931 case NETDEV_CHANGEMTU:
2932 if (idev && dev->mtu >= IPV6_MIN_MTU) {
2933 rt6_mtu_change(dev, dev->mtu);
2934 idev->cnf.mtu6 = dev->mtu;
2935 break;
2936 }
2937
2938 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2939 idev = ipv6_add_dev(dev);
2940 if (idev)
2941 break;
2942 }
2943
2944 /*
2945 * MTU falled under IPV6_MIN_MTU.
2946 * Stop IPv6 on this interface.
2947 */
2948
2949 case NETDEV_DOWN:
2950 case NETDEV_UNREGISTER:
2951 /*
2952 * Remove all addresses from this interface.
2953 */
2954 addrconf_ifdown(dev, event != NETDEV_DOWN);
2955 break;
2956
2957 case NETDEV_CHANGENAME:
2958 if (idev) {
2959 snmp6_unregister_dev(idev);
2960 addrconf_sysctl_unregister(idev);
2961 addrconf_sysctl_register(idev);
2962 err = snmp6_register_dev(idev);
2963 if (err)
2964 return notifier_from_errno(err);
2965 }
2966 break;
2967
2968 case NETDEV_PRE_TYPE_CHANGE:
2969 case NETDEV_POST_TYPE_CHANGE:
2970 addrconf_type_change(dev, event);
2971 break;
2972 }
2973
2974 return NOTIFY_OK;
2975 }
2976
2977 /*
2978 * addrconf module should be notified of a device going up
2979 */
2980 static struct notifier_block ipv6_dev_notf = {
2981 .notifier_call = addrconf_notify,
2982 };
2983
2984 static void addrconf_type_change(struct net_device *dev, unsigned long event)
2985 {
2986 struct inet6_dev *idev;
2987 ASSERT_RTNL();
2988
2989 idev = __in6_dev_get(dev);
2990
2991 if (event == NETDEV_POST_TYPE_CHANGE)
2992 ipv6_mc_remap(idev);
2993 else if (event == NETDEV_PRE_TYPE_CHANGE)
2994 ipv6_mc_unmap(idev);
2995 }
2996
2997 static int addrconf_ifdown(struct net_device *dev, int how)
2998 {
2999 struct net *net = dev_net(dev);
3000 struct inet6_dev *idev;
3001 struct inet6_ifaddr *ifa;
3002 int state, i;
3003
3004 ASSERT_RTNL();
3005
3006 rt6_ifdown(net, dev);
3007 neigh_ifdown(&nd_tbl, dev);
3008
3009 idev = __in6_dev_get(dev);
3010 if (idev == NULL)
3011 return -ENODEV;
3012
3013 /*
3014 * Step 1: remove reference to ipv6 device from parent device.
3015 * Do not dev_put!
3016 */
3017 if (how) {
3018 idev->dead = 1;
3019
3020 /* protected by rtnl_lock */
3021 RCU_INIT_POINTER(dev->ip6_ptr, NULL);
3022
3023 /* Step 1.5: remove snmp6 entry */
3024 snmp6_unregister_dev(idev);
3025
3026 }
3027
3028 /* Step 2: clear hash table */
3029 for (i = 0; i < IN6_ADDR_HSIZE; i++) {
3030 struct hlist_head *h = &inet6_addr_lst[i];
3031
3032 spin_lock_bh(&addrconf_hash_lock);
3033 restart:
3034 hlist_for_each_entry_rcu(ifa, h, addr_lst) {
3035 if (ifa->idev == idev) {
3036 hlist_del_init_rcu(&ifa->addr_lst);
3037 addrconf_del_timer(ifa);
3038 goto restart;
3039 }
3040 }
3041 spin_unlock_bh(&addrconf_hash_lock);
3042 }
3043
3044 write_lock_bh(&idev->lock);
3045
3046 /* Step 2: clear flags for stateless addrconf */
3047 if (!how)
3048 idev->if_flags &= ~(IF_RS_SENT|IF_RA_RCVD|IF_READY);
3049
3050 #ifdef CONFIG_IPV6_PRIVACY
3051 if (how && del_timer(&idev->regen_timer))
3052 in6_dev_put(idev);
3053
3054 /* Step 3: clear tempaddr list */
3055 while (!list_empty(&idev->tempaddr_list)) {
3056 ifa = list_first_entry(&idev->tempaddr_list,
3057 struct inet6_ifaddr, tmp_list);
3058 list_del(&ifa->tmp_list);
3059 write_unlock_bh(&idev->lock);
3060 spin_lock_bh(&ifa->lock);
3061
3062 if (ifa->ifpub) {
3063 in6_ifa_put(ifa->ifpub);
3064 ifa->ifpub = NULL;
3065 }
3066 spin_unlock_bh(&ifa->lock);
3067 in6_ifa_put(ifa);
3068 write_lock_bh(&idev->lock);
3069 }
3070 #endif
3071
3072 while (!list_empty(&idev->addr_list)) {
3073 ifa = list_first_entry(&idev->addr_list,
3074 struct inet6_ifaddr, if_list);
3075 addrconf_del_timer(ifa);
3076
3077 list_del(&ifa->if_list);
3078
3079 write_unlock_bh(&idev->lock);
3080
3081 spin_lock_bh(&ifa->state_lock);
3082 state = ifa->state;
3083 ifa->state = INET6_IFADDR_STATE_DEAD;
3084 spin_unlock_bh(&ifa->state_lock);
3085
3086 if (state != INET6_IFADDR_STATE_DEAD) {
3087 __ipv6_ifa_notify(RTM_DELADDR, ifa);
3088 inet6addr_notifier_call_chain(NETDEV_DOWN, ifa);
3089 }
3090 in6_ifa_put(ifa);
3091
3092 write_lock_bh(&idev->lock);
3093 }
3094
3095 write_unlock_bh(&idev->lock);
3096
3097 /* Step 5: Discard multicast list */
3098 if (how)
3099 ipv6_mc_destroy_dev(idev);
3100 else
3101 ipv6_mc_down(idev);
3102
3103 idev->tstamp = jiffies;
3104
3105 /* Last: Shot the device (if unregistered) */
3106 if (how) {
3107 addrconf_sysctl_unregister(idev);
3108 neigh_parms_release(&nd_tbl, idev->nd_parms);
3109 neigh_ifdown(&nd_tbl, dev);
3110 in6_dev_put(idev);
3111 }
3112 return 0;
3113 }
3114
3115 static void addrconf_rs_timer(unsigned long data)
3116 {
3117 struct inet6_ifaddr *ifp = (struct inet6_ifaddr *) data;
3118 struct inet6_dev *idev = ifp->idev;
3119
3120 read_lock(&idev->lock);
3121 if (idev->dead || !(idev->if_flags & IF_READY))
3122 goto out;
3123
3124 if (!ipv6_accept_ra(idev))
3125 goto out;
3126
3127 /* Announcement received after solicitation was sent */
3128 if (idev->if_flags & IF_RA_RCVD)
3129 goto out;
3130
3131 spin_lock(&ifp->lock);
3132 if (ifp->probes++ < idev->cnf.rtr_solicits) {
3133 /* The wait after the last probe can be shorter */
3134 addrconf_mod_timer(ifp, AC_RS,
3135 (ifp->probes == idev->cnf.rtr_solicits) ?
3136 idev->cnf.rtr_solicit_delay :
3137 idev->cnf.rtr_solicit_interval);
3138 spin_unlock(&ifp->lock);
3139
3140 ndisc_send_rs(idev->dev, &ifp->addr, &in6addr_linklocal_allrouters);
3141 } else {
3142 spin_unlock(&ifp->lock);
3143 /*
3144 * Note: we do not support deprecated "all on-link"
3145 * assumption any longer.
3146 */
3147 pr_debug("%s: no IPv6 routers present\n", idev->dev->name);
3148 }
3149
3150 out:
3151 read_unlock(&idev->lock);
3152 in6_ifa_put(ifp);
3153 }
3154
3155 /*
3156 * Duplicate Address Detection
3157 */
3158 static void addrconf_dad_kick(struct inet6_ifaddr *ifp)
3159 {
3160 unsigned long rand_num;
3161 struct inet6_dev *idev = ifp->idev;
3162
3163 if (ifp->flags & IFA_F_OPTIMISTIC)
3164 rand_num = 0;
3165 else
3166 rand_num = net_random() % (idev->cnf.rtr_solicit_delay ? : 1);
3167
3168 ifp->probes = idev->cnf.dad_transmits;
3169 addrconf_mod_timer(ifp, AC_DAD, rand_num);
3170 }
3171
3172 static void addrconf_dad_start(struct inet6_ifaddr *ifp)
3173 {
3174 struct inet6_dev *idev = ifp->idev;
3175 struct net_device *dev = idev->dev;
3176
3177 addrconf_join_solict(dev, &ifp->addr);
3178
3179 net_srandom(ifp->addr.s6_addr32[3]);
3180
3181 read_lock_bh(&idev->lock);
3182 spin_lock(&ifp->lock);
3183 if (ifp->state == INET6_IFADDR_STATE_DEAD)
3184 goto out;
3185
3186 if (dev->flags&(IFF_NOARP|IFF_LOOPBACK) ||
3187 idev->cnf.accept_dad < 1 ||
3188 !(ifp->flags&IFA_F_TENTATIVE) ||
3189 ifp->flags & IFA_F_NODAD) {
3190 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
3191 spin_unlock(&ifp->lock);
3192 read_unlock_bh(&idev->lock);
3193
3194 addrconf_dad_completed(ifp);
3195 return;
3196 }
3197
3198 if (!(idev->if_flags & IF_READY)) {
3199 spin_unlock(&ifp->lock);
3200 read_unlock_bh(&idev->lock);
3201 /*
3202 * If the device is not ready:
3203 * - keep it tentative if it is a permanent address.
3204 * - otherwise, kill it.
3205 */
3206 in6_ifa_hold(ifp);
3207 addrconf_dad_stop(ifp, 0);
3208 return;
3209 }
3210
3211 /*
3212 * Optimistic nodes can start receiving
3213 * Frames right away
3214 */
3215 if (ifp->flags & IFA_F_OPTIMISTIC)
3216 ip6_ins_rt(ifp->rt);
3217
3218 addrconf_dad_kick(ifp);
3219 out:
3220 spin_unlock(&ifp->lock);
3221 read_unlock_bh(&idev->lock);
3222 }
3223
3224 static void addrconf_dad_timer(unsigned long data)
3225 {
3226 struct inet6_ifaddr *ifp = (struct inet6_ifaddr *) data;
3227 struct inet6_dev *idev = ifp->idev;
3228 struct in6_addr mcaddr;
3229
3230 if (!ifp->probes && addrconf_dad_end(ifp))
3231 goto out;
3232
3233 read_lock(&idev->lock);
3234 if (idev->dead || !(idev->if_flags & IF_READY)) {
3235 read_unlock(&idev->lock);
3236 goto out;
3237 }
3238
3239 spin_lock(&ifp->lock);
3240 if (ifp->state == INET6_IFADDR_STATE_DEAD) {
3241 spin_unlock(&ifp->lock);
3242 read_unlock(&idev->lock);
3243 goto out;
3244 }
3245
3246 if (ifp->probes == 0) {
3247 /*
3248 * DAD was successful
3249 */
3250
3251 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
3252 spin_unlock(&ifp->lock);
3253 read_unlock(&idev->lock);
3254
3255 addrconf_dad_completed(ifp);
3256
3257 goto out;
3258 }
3259
3260 ifp->probes--;
3261 addrconf_mod_timer(ifp, AC_DAD, ifp->idev->nd_parms->retrans_time);
3262 spin_unlock(&ifp->lock);
3263 read_unlock(&idev->lock);
3264
3265 /* send a neighbour solicitation for our addr */
3266 addrconf_addr_solict_mult(&ifp->addr, &mcaddr);
3267 ndisc_send_ns(ifp->idev->dev, NULL, &ifp->addr, &mcaddr, &in6addr_any);
3268 out:
3269 in6_ifa_put(ifp);
3270 }
3271
3272 static void addrconf_dad_completed(struct inet6_ifaddr *ifp)
3273 {
3274 struct net_device *dev = ifp->idev->dev;
3275
3276 /*
3277 * Configure the address for reception. Now it is valid.
3278 */
3279
3280 ipv6_ifa_notify(RTM_NEWADDR, ifp);
3281
3282 /* If added prefix is link local and we are prepared to process
3283 router advertisements, start sending router solicitations.
3284 */
3285
3286 if (ipv6_accept_ra(ifp->idev) &&
3287 ifp->idev->cnf.rtr_solicits > 0 &&
3288 (dev->flags&IFF_LOOPBACK) == 0 &&
3289 (ipv6_addr_type(&ifp->addr) & IPV6_ADDR_LINKLOCAL)) {
3290 /*
3291 * If a host as already performed a random delay
3292 * [...] as part of DAD [...] there is no need
3293 * to delay again before sending the first RS
3294 */
3295 ndisc_send_rs(ifp->idev->dev, &ifp->addr, &in6addr_linklocal_allrouters);
3296
3297 spin_lock_bh(&ifp->lock);
3298 ifp->probes = 1;
3299 ifp->idev->if_flags |= IF_RS_SENT;
3300 addrconf_mod_timer(ifp, AC_RS, ifp->idev->cnf.rtr_solicit_interval);
3301 spin_unlock_bh(&ifp->lock);
3302 }
3303 }
3304
3305 static void addrconf_dad_run(struct inet6_dev *idev)
3306 {
3307 struct inet6_ifaddr *ifp;
3308
3309 read_lock_bh(&idev->lock);
3310 list_for_each_entry(ifp, &idev->addr_list, if_list) {
3311 spin_lock(&ifp->lock);
3312 if (ifp->flags & IFA_F_TENTATIVE &&
3313 ifp->state == INET6_IFADDR_STATE_DAD)
3314 addrconf_dad_kick(ifp);
3315 spin_unlock(&ifp->lock);
3316 }
3317 read_unlock_bh(&idev->lock);
3318 }
3319
3320 #ifdef CONFIG_PROC_FS
3321 struct if6_iter_state {
3322 struct seq_net_private p;
3323 int bucket;
3324 int offset;
3325 };
3326
3327 static struct inet6_ifaddr *if6_get_first(struct seq_file *seq, loff_t pos)
3328 {
3329 struct inet6_ifaddr *ifa = NULL;
3330 struct if6_iter_state *state = seq->private;
3331 struct net *net = seq_file_net(seq);
3332 int p = 0;
3333
3334 /* initial bucket if pos is 0 */
3335 if (pos == 0) {
3336 state->bucket = 0;
3337 state->offset = 0;
3338 }
3339
3340 for (; state->bucket < IN6_ADDR_HSIZE; ++state->bucket) {
3341 hlist_for_each_entry_rcu_bh(ifa, &inet6_addr_lst[state->bucket],
3342 addr_lst) {
3343 if (!net_eq(dev_net(ifa->idev->dev), net))
3344 continue;
3345 /* sync with offset */
3346 if (p < state->offset) {
3347 p++;
3348 continue;
3349 }
3350 state->offset++;
3351 return ifa;
3352 }
3353
3354 /* prepare for next bucket */
3355 state->offset = 0;
3356 p = 0;
3357 }
3358 return NULL;
3359 }
3360
3361 static struct inet6_ifaddr *if6_get_next(struct seq_file *seq,
3362 struct inet6_ifaddr *ifa)
3363 {
3364 struct if6_iter_state *state = seq->private;
3365 struct net *net = seq_file_net(seq);
3366
3367 hlist_for_each_entry_continue_rcu_bh(ifa, addr_lst) {
3368 if (!net_eq(dev_net(ifa->idev->dev), net))
3369 continue;
3370 state->offset++;
3371 return ifa;
3372 }
3373
3374 while (++state->bucket < IN6_ADDR_HSIZE) {
3375 state->offset = 0;
3376 hlist_for_each_entry_rcu_bh(ifa,
3377 &inet6_addr_lst[state->bucket], addr_lst) {
3378 if (!net_eq(dev_net(ifa->idev->dev), net))
3379 continue;
3380 state->offset++;
3381 return ifa;
3382 }
3383 }
3384
3385 return NULL;
3386 }
3387
3388 static void *if6_seq_start(struct seq_file *seq, loff_t *pos)
3389 __acquires(rcu_bh)
3390 {
3391 rcu_read_lock_bh();
3392 return if6_get_first(seq, *pos);
3393 }
3394
3395 static void *if6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
3396 {
3397 struct inet6_ifaddr *ifa;
3398
3399 ifa = if6_get_next(seq, v);
3400 ++*pos;
3401 return ifa;
3402 }
3403
3404 static void if6_seq_stop(struct seq_file *seq, void *v)
3405 __releases(rcu_bh)
3406 {
3407 rcu_read_unlock_bh();
3408 }
3409
3410 static int if6_seq_show(struct seq_file *seq, void *v)
3411 {
3412 struct inet6_ifaddr *ifp = (struct inet6_ifaddr *)v;
3413 seq_printf(seq, "%pi6 %02x %02x %02x %02x %8s\n",
3414 &ifp->addr,
3415 ifp->idev->dev->ifindex,
3416 ifp->prefix_len,
3417 ifp->scope,
3418 ifp->flags,
3419 ifp->idev->dev->name);
3420 return 0;
3421 }
3422
3423 static const struct seq_operations if6_seq_ops = {
3424 .start = if6_seq_start,
3425 .next = if6_seq_next,
3426 .show = if6_seq_show,
3427 .stop = if6_seq_stop,
3428 };
3429
3430 static int if6_seq_open(struct inode *inode, struct file *file)
3431 {
3432 return seq_open_net(inode, file, &if6_seq_ops,
3433 sizeof(struct if6_iter_state));
3434 }
3435
3436 static const struct file_operations if6_fops = {
3437 .owner = THIS_MODULE,
3438 .open = if6_seq_open,
3439 .read = seq_read,
3440 .llseek = seq_lseek,
3441 .release = seq_release_net,
3442 };
3443
3444 static int __net_init if6_proc_net_init(struct net *net)
3445 {
3446 if (!proc_create("if_inet6", S_IRUGO, net->proc_net, &if6_fops))
3447 return -ENOMEM;
3448 return 0;
3449 }
3450
3451 static void __net_exit if6_proc_net_exit(struct net *net)
3452 {
3453 remove_proc_entry("if_inet6", net->proc_net);
3454 }
3455
3456 static struct pernet_operations if6_proc_net_ops = {
3457 .init = if6_proc_net_init,
3458 .exit = if6_proc_net_exit,
3459 };
3460
3461 int __init if6_proc_init(void)
3462 {
3463 return register_pernet_subsys(&if6_proc_net_ops);
3464 }
3465
3466 void if6_proc_exit(void)
3467 {
3468 unregister_pernet_subsys(&if6_proc_net_ops);
3469 }
3470 #endif /* CONFIG_PROC_FS */
3471
3472 #if IS_ENABLED(CONFIG_IPV6_MIP6)
3473 /* Check if address is a home address configured on any interface. */
3474 int ipv6_chk_home_addr(struct net *net, const struct in6_addr *addr)
3475 {
3476 int ret = 0;
3477 struct inet6_ifaddr *ifp = NULL;
3478 unsigned int hash = inet6_addr_hash(addr);
3479
3480 rcu_read_lock_bh();
3481 hlist_for_each_entry_rcu_bh(ifp, &inet6_addr_lst[hash], addr_lst) {
3482 if (!net_eq(dev_net(ifp->idev->dev), net))
3483 continue;
3484 if (ipv6_addr_equal(&ifp->addr, addr) &&
3485 (ifp->flags & IFA_F_HOMEADDRESS)) {
3486 ret = 1;
3487 break;
3488 }
3489 }
3490 rcu_read_unlock_bh();
3491 return ret;
3492 }
3493 #endif
3494
3495 /*
3496 * Periodic address status verification
3497 */
3498
3499 static void addrconf_verify(unsigned long foo)
3500 {
3501 unsigned long now, next, next_sec, next_sched;
3502 struct inet6_ifaddr *ifp;
3503 int i;
3504
3505 rcu_read_lock_bh();
3506 spin_lock(&addrconf_verify_lock);
3507 now = jiffies;
3508 next = round_jiffies_up(now + ADDR_CHECK_FREQUENCY);
3509
3510 del_timer(&addr_chk_timer);
3511
3512 for (i = 0; i < IN6_ADDR_HSIZE; i++) {
3513 restart:
3514 hlist_for_each_entry_rcu_bh(ifp,
3515 &inet6_addr_lst[i], addr_lst) {
3516 unsigned long age;
3517
3518 if (ifp->flags & IFA_F_PERMANENT)
3519 continue;
3520
3521 spin_lock(&ifp->lock);
3522 /* We try to batch several events at once. */
3523 age = (now - ifp->tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
3524
3525 if (ifp->valid_lft != INFINITY_LIFE_TIME &&
3526 age >= ifp->valid_lft) {
3527 spin_unlock(&ifp->lock);
3528 in6_ifa_hold(ifp);
3529 ipv6_del_addr(ifp);
3530 goto restart;
3531 } else if (ifp->prefered_lft == INFINITY_LIFE_TIME) {
3532 spin_unlock(&ifp->lock);
3533 continue;
3534 } else if (age >= ifp->prefered_lft) {
3535 /* jiffies - ifp->tstamp > age >= ifp->prefered_lft */
3536 int deprecate = 0;
3537
3538 if (!(ifp->flags&IFA_F_DEPRECATED)) {
3539 deprecate = 1;
3540 ifp->flags |= IFA_F_DEPRECATED;
3541 }
3542
3543 if (time_before(ifp->tstamp + ifp->valid_lft * HZ, next))
3544 next = ifp->tstamp + ifp->valid_lft * HZ;
3545
3546 spin_unlock(&ifp->lock);
3547
3548 if (deprecate) {
3549 in6_ifa_hold(ifp);
3550
3551 ipv6_ifa_notify(0, ifp);
3552 in6_ifa_put(ifp);
3553 goto restart;
3554 }
3555 #ifdef CONFIG_IPV6_PRIVACY
3556 } else if ((ifp->flags&IFA_F_TEMPORARY) &&
3557 !(ifp->flags&IFA_F_TENTATIVE)) {
3558 unsigned long regen_advance = ifp->idev->cnf.regen_max_retry *
3559 ifp->idev->cnf.dad_transmits *
3560 ifp->idev->nd_parms->retrans_time / HZ;
3561
3562 if (age >= ifp->prefered_lft - regen_advance) {
3563 struct inet6_ifaddr *ifpub = ifp->ifpub;
3564 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3565 next = ifp->tstamp + ifp->prefered_lft * HZ;
3566 if (!ifp->regen_count && ifpub) {
3567 ifp->regen_count++;
3568 in6_ifa_hold(ifp);
3569 in6_ifa_hold(ifpub);
3570 spin_unlock(&ifp->lock);
3571
3572 spin_lock(&ifpub->lock);
3573 ifpub->regen_count = 0;
3574 spin_unlock(&ifpub->lock);
3575 ipv6_create_tempaddr(ifpub, ifp);
3576 in6_ifa_put(ifpub);
3577 in6_ifa_put(ifp);
3578 goto restart;
3579 }
3580 } else if (time_before(ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ, next))
3581 next = ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ;
3582 spin_unlock(&ifp->lock);
3583 #endif
3584 } else {
3585 /* ifp->prefered_lft <= ifp->valid_lft */
3586 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3587 next = ifp->tstamp + ifp->prefered_lft * HZ;
3588 spin_unlock(&ifp->lock);
3589 }
3590 }
3591 }
3592
3593 next_sec = round_jiffies_up(next);
3594 next_sched = next;
3595
3596 /* If rounded timeout is accurate enough, accept it. */
3597 if (time_before(next_sec, next + ADDRCONF_TIMER_FUZZ))
3598 next_sched = next_sec;
3599
3600 /* And minimum interval is ADDRCONF_TIMER_FUZZ_MAX. */
3601 if (time_before(next_sched, jiffies + ADDRCONF_TIMER_FUZZ_MAX))
3602 next_sched = jiffies + ADDRCONF_TIMER_FUZZ_MAX;
3603
3604 ADBG((KERN_DEBUG "now = %lu, schedule = %lu, rounded schedule = %lu => %lu\n",
3605 now, next, next_sec, next_sched));
3606
3607 addr_chk_timer.expires = next_sched;
3608 add_timer(&addr_chk_timer);
3609 spin_unlock(&addrconf_verify_lock);
3610 rcu_read_unlock_bh();
3611 }
3612
3613 static struct in6_addr *extract_addr(struct nlattr *addr, struct nlattr *local)
3614 {
3615 struct in6_addr *pfx = NULL;
3616
3617 if (addr)
3618 pfx = nla_data(addr);
3619
3620 if (local) {
3621 if (pfx && nla_memcmp(local, pfx, sizeof(*pfx)))
3622 pfx = NULL;
3623 else
3624 pfx = nla_data(local);
3625 }
3626
3627 return pfx;
3628 }
3629
3630 static const struct nla_policy ifa_ipv6_policy[IFA_MAX+1] = {
3631 [IFA_ADDRESS] = { .len = sizeof(struct in6_addr) },
3632 [IFA_LOCAL] = { .len = sizeof(struct in6_addr) },
3633 [IFA_CACHEINFO] = { .len = sizeof(struct ifa_cacheinfo) },
3634 };
3635
3636 static int
3637 inet6_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh)
3638 {
3639 struct net *net = sock_net(skb->sk);
3640 struct ifaddrmsg *ifm;
3641 struct nlattr *tb[IFA_MAX+1];
3642 struct in6_addr *pfx;
3643 int err;
3644
3645 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3646 if (err < 0)
3647 return err;
3648
3649 ifm = nlmsg_data(nlh);
3650 pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3651 if (pfx == NULL)
3652 return -EINVAL;
3653
3654 return inet6_addr_del(net, ifm->ifa_index, pfx, ifm->ifa_prefixlen);
3655 }
3656
3657 static int inet6_addr_modify(struct inet6_ifaddr *ifp, u8 ifa_flags,
3658 u32 prefered_lft, u32 valid_lft)
3659 {
3660 u32 flags;
3661 clock_t expires;
3662 unsigned long timeout;
3663
3664 if (!valid_lft || (prefered_lft > valid_lft))
3665 return -EINVAL;
3666
3667 timeout = addrconf_timeout_fixup(valid_lft, HZ);
3668 if (addrconf_finite_timeout(timeout)) {
3669 expires = jiffies_to_clock_t(timeout * HZ);
3670 valid_lft = timeout;
3671 flags = RTF_EXPIRES;
3672 } else {
3673 expires = 0;
3674 flags = 0;
3675 ifa_flags |= IFA_F_PERMANENT;
3676 }
3677
3678 timeout = addrconf_timeout_fixup(prefered_lft, HZ);
3679 if (addrconf_finite_timeout(timeout)) {
3680 if (timeout == 0)
3681 ifa_flags |= IFA_F_DEPRECATED;
3682 prefered_lft = timeout;
3683 }
3684
3685 spin_lock_bh(&ifp->lock);
3686 ifp->flags = (ifp->flags & ~(IFA_F_DEPRECATED | IFA_F_PERMANENT | IFA_F_NODAD | IFA_F_HOMEADDRESS)) | ifa_flags;
3687 ifp->tstamp = jiffies;
3688 ifp->valid_lft = valid_lft;
3689 ifp->prefered_lft = prefered_lft;
3690
3691 spin_unlock_bh(&ifp->lock);
3692 if (!(ifp->flags&IFA_F_TENTATIVE))
3693 ipv6_ifa_notify(0, ifp);
3694
3695 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, ifp->idev->dev,
3696 expires, flags);
3697 addrconf_verify(0);
3698
3699 return 0;
3700 }
3701
3702 static int
3703 inet6_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh)
3704 {
3705 struct net *net = sock_net(skb->sk);
3706 struct ifaddrmsg *ifm;
3707 struct nlattr *tb[IFA_MAX+1];
3708 struct in6_addr *pfx;
3709 struct inet6_ifaddr *ifa;
3710 struct net_device *dev;
3711 u32 valid_lft = INFINITY_LIFE_TIME, preferred_lft = INFINITY_LIFE_TIME;
3712 u8 ifa_flags;
3713 int err;
3714
3715 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3716 if (err < 0)
3717 return err;
3718
3719 ifm = nlmsg_data(nlh);
3720 pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3721 if (pfx == NULL)
3722 return -EINVAL;
3723
3724 if (tb[IFA_CACHEINFO]) {
3725 struct ifa_cacheinfo *ci;
3726
3727 ci = nla_data(tb[IFA_CACHEINFO]);
3728 valid_lft = ci->ifa_valid;
3729 preferred_lft = ci->ifa_prefered;
3730 } else {
3731 preferred_lft = INFINITY_LIFE_TIME;
3732 valid_lft = INFINITY_LIFE_TIME;
3733 }
3734
3735 dev = __dev_get_by_index(net, ifm->ifa_index);
3736 if (dev == NULL)
3737 return -ENODEV;
3738
3739 /* We ignore other flags so far. */
3740 ifa_flags = ifm->ifa_flags & (IFA_F_NODAD | IFA_F_HOMEADDRESS);
3741
3742 ifa = ipv6_get_ifaddr(net, pfx, dev, 1);
3743 if (ifa == NULL) {
3744 /*
3745 * It would be best to check for !NLM_F_CREATE here but
3746 * userspace alreay relies on not having to provide this.
3747 */
3748 return inet6_addr_add(net, ifm->ifa_index, pfx,
3749 ifm->ifa_prefixlen, ifa_flags,
3750 preferred_lft, valid_lft);
3751 }
3752
3753 if (nlh->nlmsg_flags & NLM_F_EXCL ||
3754 !(nlh->nlmsg_flags & NLM_F_REPLACE))
3755 err = -EEXIST;
3756 else
3757 err = inet6_addr_modify(ifa, ifa_flags, preferred_lft, valid_lft);
3758
3759 in6_ifa_put(ifa);
3760
3761 return err;
3762 }
3763
3764 static void put_ifaddrmsg(struct nlmsghdr *nlh, u8 prefixlen, u8 flags,
3765 u8 scope, int ifindex)
3766 {
3767 struct ifaddrmsg *ifm;
3768
3769 ifm = nlmsg_data(nlh);
3770 ifm->ifa_family = AF_INET6;
3771 ifm->ifa_prefixlen = prefixlen;
3772 ifm->ifa_flags = flags;
3773 ifm->ifa_scope = scope;
3774 ifm->ifa_index = ifindex;
3775 }
3776
3777 static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
3778 unsigned long tstamp, u32 preferred, u32 valid)
3779 {
3780 struct ifa_cacheinfo ci;
3781
3782 ci.cstamp = cstamp_delta(cstamp);
3783 ci.tstamp = cstamp_delta(tstamp);
3784 ci.ifa_prefered = preferred;
3785 ci.ifa_valid = valid;
3786
3787 return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
3788 }
3789
3790 static inline int rt_scope(int ifa_scope)
3791 {
3792 if (ifa_scope & IFA_HOST)
3793 return RT_SCOPE_HOST;
3794 else if (ifa_scope & IFA_LINK)
3795 return RT_SCOPE_LINK;
3796 else if (ifa_scope & IFA_SITE)
3797 return RT_SCOPE_SITE;
3798 else
3799 return RT_SCOPE_UNIVERSE;
3800 }
3801
3802 static inline int inet6_ifaddr_msgsize(void)
3803 {
3804 return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
3805 + nla_total_size(16) /* IFA_ADDRESS */
3806 + nla_total_size(sizeof(struct ifa_cacheinfo));
3807 }
3808
3809 static int inet6_fill_ifaddr(struct sk_buff *skb, struct inet6_ifaddr *ifa,
3810 u32 portid, u32 seq, int event, unsigned int flags)
3811 {
3812 struct nlmsghdr *nlh;
3813 u32 preferred, valid;
3814
3815 nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
3816 if (nlh == NULL)
3817 return -EMSGSIZE;
3818
3819 put_ifaddrmsg(nlh, ifa->prefix_len, ifa->flags, rt_scope(ifa->scope),
3820 ifa->idev->dev->ifindex);
3821
3822 if (!(ifa->flags&IFA_F_PERMANENT)) {
3823 preferred = ifa->prefered_lft;
3824 valid = ifa->valid_lft;
3825 if (preferred != INFINITY_LIFE_TIME) {
3826 long tval = (jiffies - ifa->tstamp)/HZ;
3827 if (preferred > tval)
3828 preferred -= tval;
3829 else
3830 preferred = 0;
3831 if (valid != INFINITY_LIFE_TIME) {
3832 if (valid > tval)
3833 valid -= tval;
3834 else
3835 valid = 0;
3836 }
3837 }
3838 } else {
3839 preferred = INFINITY_LIFE_TIME;
3840 valid = INFINITY_LIFE_TIME;
3841 }
3842
3843 if (nla_put(skb, IFA_ADDRESS, 16, &ifa->addr) < 0 ||
3844 put_cacheinfo(skb, ifa->cstamp, ifa->tstamp, preferred, valid) < 0) {
3845 nlmsg_cancel(skb, nlh);
3846 return -EMSGSIZE;
3847 }
3848
3849 return nlmsg_end(skb, nlh);
3850 }
3851
3852 static int inet6_fill_ifmcaddr(struct sk_buff *skb, struct ifmcaddr6 *ifmca,
3853 u32 portid, u32 seq, int event, u16 flags)
3854 {
3855 struct nlmsghdr *nlh;
3856 u8 scope = RT_SCOPE_UNIVERSE;
3857 int ifindex = ifmca->idev->dev->ifindex;
3858
3859 if (ipv6_addr_scope(&ifmca->mca_addr) & IFA_SITE)
3860 scope = RT_SCOPE_SITE;
3861
3862 nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
3863 if (nlh == NULL)
3864 return -EMSGSIZE;
3865
3866 put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3867 if (nla_put(skb, IFA_MULTICAST, 16, &ifmca->mca_addr) < 0 ||
3868 put_cacheinfo(skb, ifmca->mca_cstamp, ifmca->mca_tstamp,
3869 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3870 nlmsg_cancel(skb, nlh);
3871 return -EMSGSIZE;
3872 }
3873
3874 return nlmsg_end(skb, nlh);
3875 }
3876
3877 static int inet6_fill_ifacaddr(struct sk_buff *skb, struct ifacaddr6 *ifaca,
3878 u32 portid, u32 seq, int event, unsigned int flags)
3879 {
3880 struct nlmsghdr *nlh;
3881 u8 scope = RT_SCOPE_UNIVERSE;
3882 int ifindex = ifaca->aca_idev->dev->ifindex;
3883
3884 if (ipv6_addr_scope(&ifaca->aca_addr) & IFA_SITE)
3885 scope = RT_SCOPE_SITE;
3886
3887 nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
3888 if (nlh == NULL)
3889 return -EMSGSIZE;
3890
3891 put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3892 if (nla_put(skb, IFA_ANYCAST, 16, &ifaca->aca_addr) < 0 ||
3893 put_cacheinfo(skb, ifaca->aca_cstamp, ifaca->aca_tstamp,
3894 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3895 nlmsg_cancel(skb, nlh);
3896 return -EMSGSIZE;
3897 }
3898
3899 return nlmsg_end(skb, nlh);
3900 }
3901
3902 enum addr_type_t {
3903 UNICAST_ADDR,
3904 MULTICAST_ADDR,
3905 ANYCAST_ADDR,
3906 };
3907
3908 /* called with rcu_read_lock() */
3909 static int in6_dump_addrs(struct inet6_dev *idev, struct sk_buff *skb,
3910 struct netlink_callback *cb, enum addr_type_t type,
3911 int s_ip_idx, int *p_ip_idx)
3912 {
3913 struct ifmcaddr6 *ifmca;
3914 struct ifacaddr6 *ifaca;
3915 int err = 1;
3916 int ip_idx = *p_ip_idx;
3917
3918 read_lock_bh(&idev->lock);
3919 switch (type) {
3920 case UNICAST_ADDR: {
3921 struct inet6_ifaddr *ifa;
3922
3923 /* unicast address incl. temp addr */
3924 list_for_each_entry(ifa, &idev->addr_list, if_list) {
3925 if (++ip_idx < s_ip_idx)
3926 continue;
3927 err = inet6_fill_ifaddr(skb, ifa,
3928 NETLINK_CB(cb->skb).portid,
3929 cb->nlh->nlmsg_seq,
3930 RTM_NEWADDR,
3931 NLM_F_MULTI);
3932 if (err <= 0)
3933 break;
3934 nl_dump_check_consistent(cb, nlmsg_hdr(skb));
3935 }
3936 break;
3937 }
3938 case MULTICAST_ADDR:
3939 /* multicast address */
3940 for (ifmca = idev->mc_list; ifmca;
3941 ifmca = ifmca->next, ip_idx++) {
3942 if (ip_idx < s_ip_idx)
3943 continue;
3944 err = inet6_fill_ifmcaddr(skb, ifmca,
3945 NETLINK_CB(cb->skb).portid,
3946 cb->nlh->nlmsg_seq,
3947 RTM_GETMULTICAST,
3948 NLM_F_MULTI);
3949 if (err <= 0)
3950 break;
3951 }
3952 break;
3953 case ANYCAST_ADDR:
3954 /* anycast address */
3955 for (ifaca = idev->ac_list; ifaca;
3956 ifaca = ifaca->aca_next, ip_idx++) {
3957 if (ip_idx < s_ip_idx)
3958 continue;
3959 err = inet6_fill_ifacaddr(skb, ifaca,
3960 NETLINK_CB(cb->skb).portid,
3961 cb->nlh->nlmsg_seq,
3962 RTM_GETANYCAST,
3963 NLM_F_MULTI);
3964 if (err <= 0)
3965 break;
3966 }
3967 break;
3968 default:
3969 break;
3970 }
3971 read_unlock_bh(&idev->lock);
3972 *p_ip_idx = ip_idx;
3973 return err;
3974 }
3975
3976 static int inet6_dump_addr(struct sk_buff *skb, struct netlink_callback *cb,
3977 enum addr_type_t type)
3978 {
3979 struct net *net = sock_net(skb->sk);
3980 int h, s_h;
3981 int idx, ip_idx;
3982 int s_idx, s_ip_idx;
3983 struct net_device *dev;
3984 struct inet6_dev *idev;
3985 struct hlist_head *head;
3986
3987 s_h = cb->args[0];
3988 s_idx = idx = cb->args[1];
3989 s_ip_idx = ip_idx = cb->args[2];
3990
3991 rcu_read_lock();
3992 cb->seq = atomic_read(&net->ipv6.dev_addr_genid) ^ net->dev_base_seq;
3993 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
3994 idx = 0;
3995 head = &net->dev_index_head[h];
3996 hlist_for_each_entry_rcu(dev, head, index_hlist) {
3997 if (idx < s_idx)
3998 goto cont;
3999 if (h > s_h || idx > s_idx)
4000 s_ip_idx = 0;
4001 ip_idx = 0;
4002 idev = __in6_dev_get(dev);
4003 if (!idev)
4004 goto cont;
4005
4006 if (in6_dump_addrs(idev, skb, cb, type,
4007 s_ip_idx, &ip_idx) <= 0)
4008 goto done;
4009 cont:
4010 idx++;
4011 }
4012 }
4013 done:
4014 rcu_read_unlock();
4015 cb->args[0] = h;
4016 cb->args[1] = idx;
4017 cb->args[2] = ip_idx;
4018
4019 return skb->len;
4020 }
4021
4022 static int inet6_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
4023 {
4024 enum addr_type_t type = UNICAST_ADDR;
4025
4026 return inet6_dump_addr(skb, cb, type);
4027 }
4028
4029 static int inet6_dump_ifmcaddr(struct sk_buff *skb, struct netlink_callback *cb)
4030 {
4031 enum addr_type_t type = MULTICAST_ADDR;
4032
4033 return inet6_dump_addr(skb, cb, type);
4034 }
4035
4036
4037 static int inet6_dump_ifacaddr(struct sk_buff *skb, struct netlink_callback *cb)
4038 {
4039 enum addr_type_t type = ANYCAST_ADDR;
4040
4041 return inet6_dump_addr(skb, cb, type);
4042 }
4043
4044 static int inet6_rtm_getaddr(struct sk_buff *in_skb, struct nlmsghdr *nlh)
4045 {
4046 struct net *net = sock_net(in_skb->sk);
4047 struct ifaddrmsg *ifm;
4048 struct nlattr *tb[IFA_MAX+1];
4049 struct in6_addr *addr = NULL;
4050 struct net_device *dev = NULL;
4051 struct inet6_ifaddr *ifa;
4052 struct sk_buff *skb;
4053 int err;
4054
4055 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
4056 if (err < 0)
4057 goto errout;
4058
4059 addr = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
4060 if (addr == NULL) {
4061 err = -EINVAL;
4062 goto errout;
4063 }
4064
4065 ifm = nlmsg_data(nlh);
4066 if (ifm->ifa_index)
4067 dev = __dev_get_by_index(net, ifm->ifa_index);
4068
4069 ifa = ipv6_get_ifaddr(net, addr, dev, 1);
4070 if (!ifa) {
4071 err = -EADDRNOTAVAIL;
4072 goto errout;
4073 }
4074
4075 skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_KERNEL);
4076 if (!skb) {
4077 err = -ENOBUFS;
4078 goto errout_ifa;
4079 }
4080
4081 err = inet6_fill_ifaddr(skb, ifa, NETLINK_CB(in_skb).portid,
4082 nlh->nlmsg_seq, RTM_NEWADDR, 0);
4083 if (err < 0) {
4084 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
4085 WARN_ON(err == -EMSGSIZE);
4086 kfree_skb(skb);
4087 goto errout_ifa;
4088 }
4089 err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
4090 errout_ifa:
4091 in6_ifa_put(ifa);
4092 errout:
4093 return err;
4094 }
4095
4096 static void inet6_ifa_notify(int event, struct inet6_ifaddr *ifa)
4097 {
4098 struct sk_buff *skb;
4099 struct net *net = dev_net(ifa->idev->dev);
4100 int err = -ENOBUFS;
4101
4102 skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_ATOMIC);
4103 if (skb == NULL)
4104 goto errout;
4105
4106 err = inet6_fill_ifaddr(skb, ifa, 0, 0, event, 0);
4107 if (err < 0) {
4108 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
4109 WARN_ON(err == -EMSGSIZE);
4110 kfree_skb(skb);
4111 goto errout;
4112 }
4113 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
4114 return;
4115 errout:
4116 if (err < 0)
4117 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
4118 }
4119
4120 static inline void ipv6_store_devconf(struct ipv6_devconf *cnf,
4121 __s32 *array, int bytes)
4122 {
4123 BUG_ON(bytes < (DEVCONF_MAX * 4));
4124
4125 memset(array, 0, bytes);
4126 array[DEVCONF_FORWARDING] = cnf->forwarding;
4127 array[DEVCONF_HOPLIMIT] = cnf->hop_limit;
4128 array[DEVCONF_MTU6] = cnf->mtu6;
4129 array[DEVCONF_ACCEPT_RA] = cnf->accept_ra;
4130 array[DEVCONF_ACCEPT_REDIRECTS] = cnf->accept_redirects;
4131 array[DEVCONF_AUTOCONF] = cnf->autoconf;
4132 array[DEVCONF_DAD_TRANSMITS] = cnf->dad_transmits;
4133 array[DEVCONF_RTR_SOLICITS] = cnf->rtr_solicits;
4134 array[DEVCONF_RTR_SOLICIT_INTERVAL] =
4135 jiffies_to_msecs(cnf->rtr_solicit_interval);
4136 array[DEVCONF_RTR_SOLICIT_DELAY] =
4137 jiffies_to_msecs(cnf->rtr_solicit_delay);
4138 array[DEVCONF_FORCE_MLD_VERSION] = cnf->force_mld_version;
4139 #ifdef CONFIG_IPV6_PRIVACY
4140 array[DEVCONF_USE_TEMPADDR] = cnf->use_tempaddr;
4141 array[DEVCONF_TEMP_VALID_LFT] = cnf->temp_valid_lft;
4142 array[DEVCONF_TEMP_PREFERED_LFT] = cnf->temp_prefered_lft;
4143 array[DEVCONF_REGEN_MAX_RETRY] = cnf->regen_max_retry;
4144 array[DEVCONF_MAX_DESYNC_FACTOR] = cnf->max_desync_factor;
4145 #endif
4146 array[DEVCONF_MAX_ADDRESSES] = cnf->max_addresses;
4147 array[DEVCONF_ACCEPT_RA_DEFRTR] = cnf->accept_ra_defrtr;
4148 array[DEVCONF_ACCEPT_RA_PINFO] = cnf->accept_ra_pinfo;
4149 #ifdef CONFIG_IPV6_ROUTER_PREF
4150 array[DEVCONF_ACCEPT_RA_RTR_PREF] = cnf->accept_ra_rtr_pref;
4151 array[DEVCONF_RTR_PROBE_INTERVAL] =
4152 jiffies_to_msecs(cnf->rtr_probe_interval);
4153 #ifdef CONFIG_IPV6_ROUTE_INFO
4154 array[DEVCONF_ACCEPT_RA_RT_INFO_MAX_PLEN] = cnf->accept_ra_rt_info_max_plen;
4155 #endif
4156 #endif
4157 array[DEVCONF_PROXY_NDP] = cnf->proxy_ndp;
4158 array[DEVCONF_ACCEPT_SOURCE_ROUTE] = cnf->accept_source_route;
4159 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
4160 array[DEVCONF_OPTIMISTIC_DAD] = cnf->optimistic_dad;
4161 #endif
4162 #ifdef CONFIG_IPV6_MROUTE
4163 array[DEVCONF_MC_FORWARDING] = cnf->mc_forwarding;
4164 #endif
4165 array[DEVCONF_DISABLE_IPV6] = cnf->disable_ipv6;
4166 array[DEVCONF_ACCEPT_DAD] = cnf->accept_dad;
4167 array[DEVCONF_FORCE_TLLAO] = cnf->force_tllao;
4168 array[DEVCONF_NDISC_NOTIFY] = cnf->ndisc_notify;
4169 }
4170
4171 static inline size_t inet6_ifla6_size(void)
4172 {
4173 return nla_total_size(4) /* IFLA_INET6_FLAGS */
4174 + nla_total_size(sizeof(struct ifla_cacheinfo))
4175 + nla_total_size(DEVCONF_MAX * 4) /* IFLA_INET6_CONF */
4176 + nla_total_size(IPSTATS_MIB_MAX * 8) /* IFLA_INET6_STATS */
4177 + nla_total_size(ICMP6_MIB_MAX * 8) /* IFLA_INET6_ICMP6STATS */
4178 + nla_total_size(sizeof(struct in6_addr)); /* IFLA_INET6_TOKEN */
4179 }
4180
4181 static inline size_t inet6_if_nlmsg_size(void)
4182 {
4183 return NLMSG_ALIGN(sizeof(struct ifinfomsg))
4184 + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
4185 + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
4186 + nla_total_size(4) /* IFLA_MTU */
4187 + nla_total_size(4) /* IFLA_LINK */
4188 + nla_total_size(inet6_ifla6_size()); /* IFLA_PROTINFO */
4189 }
4190
4191 static inline void __snmp6_fill_statsdev(u64 *stats, atomic_long_t *mib,
4192 int items, int bytes)
4193 {
4194 int i;
4195 int pad = bytes - sizeof(u64) * items;
4196 BUG_ON(pad < 0);
4197
4198 /* Use put_unaligned() because stats may not be aligned for u64. */
4199 put_unaligned(items, &stats[0]);
4200 for (i = 1; i < items; i++)
4201 put_unaligned(atomic_long_read(&mib[i]), &stats[i]);
4202
4203 memset(&stats[items], 0, pad);
4204 }
4205
4206 static inline void __snmp6_fill_stats64(u64 *stats, void __percpu **mib,
4207 int items, int bytes, size_t syncpoff)
4208 {
4209 int i;
4210 int pad = bytes - sizeof(u64) * items;
4211 BUG_ON(pad < 0);
4212
4213 /* Use put_unaligned() because stats may not be aligned for u64. */
4214 put_unaligned(items, &stats[0]);
4215 for (i = 1; i < items; i++)
4216 put_unaligned(snmp_fold_field64(mib, i, syncpoff), &stats[i]);
4217
4218 memset(&stats[items], 0, pad);
4219 }
4220
4221 static void snmp6_fill_stats(u64 *stats, struct inet6_dev *idev, int attrtype,
4222 int bytes)
4223 {
4224 switch (attrtype) {
4225 case IFLA_INET6_STATS:
4226 __snmp6_fill_stats64(stats, (void __percpu **)idev->stats.ipv6,
4227 IPSTATS_MIB_MAX, bytes, offsetof(struct ipstats_mib, syncp));
4228 break;
4229 case IFLA_INET6_ICMP6STATS:
4230 __snmp6_fill_statsdev(stats, idev->stats.icmpv6dev->mibs, ICMP6_MIB_MAX, bytes);
4231 break;
4232 }
4233 }
4234
4235 static int inet6_fill_ifla6_attrs(struct sk_buff *skb, struct inet6_dev *idev)
4236 {
4237 struct nlattr *nla;
4238 struct ifla_cacheinfo ci;
4239
4240 if (nla_put_u32(skb, IFLA_INET6_FLAGS, idev->if_flags))
4241 goto nla_put_failure;
4242 ci.max_reasm_len = IPV6_MAXPLEN;
4243 ci.tstamp = cstamp_delta(idev->tstamp);
4244 ci.reachable_time = jiffies_to_msecs(idev->nd_parms->reachable_time);
4245 ci.retrans_time = jiffies_to_msecs(idev->nd_parms->retrans_time);
4246 if (nla_put(skb, IFLA_INET6_CACHEINFO, sizeof(ci), &ci))
4247 goto nla_put_failure;
4248 nla = nla_reserve(skb, IFLA_INET6_CONF, DEVCONF_MAX * sizeof(s32));
4249 if (nla == NULL)
4250 goto nla_put_failure;
4251 ipv6_store_devconf(&idev->cnf, nla_data(nla), nla_len(nla));
4252
4253 /* XXX - MC not implemented */
4254
4255 nla = nla_reserve(skb, IFLA_INET6_STATS, IPSTATS_MIB_MAX * sizeof(u64));
4256 if (nla == NULL)
4257 goto nla_put_failure;
4258 snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_STATS, nla_len(nla));
4259
4260 nla = nla_reserve(skb, IFLA_INET6_ICMP6STATS, ICMP6_MIB_MAX * sizeof(u64));
4261 if (nla == NULL)
4262 goto nla_put_failure;
4263 snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_ICMP6STATS, nla_len(nla));
4264
4265 nla = nla_reserve(skb, IFLA_INET6_TOKEN, sizeof(struct in6_addr));
4266 if (nla == NULL)
4267 goto nla_put_failure;
4268 read_lock_bh(&idev->lock);
4269 memcpy(nla_data(nla), idev->token.s6_addr, nla_len(nla));
4270 read_unlock_bh(&idev->lock);
4271
4272 return 0;
4273
4274 nla_put_failure:
4275 return -EMSGSIZE;
4276 }
4277
4278 static size_t inet6_get_link_af_size(const struct net_device *dev)
4279 {
4280 if (!__in6_dev_get(dev))
4281 return 0;
4282
4283 return inet6_ifla6_size();
4284 }
4285
4286 static int inet6_fill_link_af(struct sk_buff *skb, const struct net_device *dev)
4287 {
4288 struct inet6_dev *idev = __in6_dev_get(dev);
4289
4290 if (!idev)
4291 return -ENODATA;
4292
4293 if (inet6_fill_ifla6_attrs(skb, idev) < 0)
4294 return -EMSGSIZE;
4295
4296 return 0;
4297 }
4298
4299 static int inet6_set_iftoken(struct inet6_dev *idev, struct in6_addr *token)
4300 {
4301 struct inet6_ifaddr *ifp;
4302 struct net_device *dev = idev->dev;
4303 bool update_rs = false;
4304
4305 if (token == NULL)
4306 return -EINVAL;
4307 if (ipv6_addr_any(token))
4308 return -EINVAL;
4309 if (dev->flags & (IFF_LOOPBACK | IFF_NOARP))
4310 return -EINVAL;
4311 if (!ipv6_accept_ra(idev))
4312 return -EINVAL;
4313 if (idev->cnf.rtr_solicits <= 0)
4314 return -EINVAL;
4315
4316 write_lock_bh(&idev->lock);
4317
4318 BUILD_BUG_ON(sizeof(token->s6_addr) != 16);
4319 memcpy(idev->token.s6_addr + 8, token->s6_addr + 8, 8);
4320
4321 write_unlock_bh(&idev->lock);
4322
4323 if (!idev->dead && (idev->if_flags & IF_READY)) {
4324 struct in6_addr ll_addr;
4325
4326 ipv6_get_lladdr(dev, &ll_addr, IFA_F_TENTATIVE |
4327 IFA_F_OPTIMISTIC);
4328
4329 /* If we're not ready, then normal ifup will take care
4330 * of this. Otherwise, we need to request our rs here.
4331 */
4332 ndisc_send_rs(dev, &ll_addr, &in6addr_linklocal_allrouters);
4333 update_rs = true;
4334 }
4335
4336 write_lock_bh(&idev->lock);
4337
4338 if (update_rs)
4339 idev->if_flags |= IF_RS_SENT;
4340
4341 /* Well, that's kinda nasty ... */
4342 list_for_each_entry(ifp, &idev->addr_list, if_list) {
4343 spin_lock(&ifp->lock);
4344 if (ifp->tokenized) {
4345 ifp->valid_lft = 0;
4346 ifp->prefered_lft = 0;
4347 }
4348 spin_unlock(&ifp->lock);
4349 }
4350
4351 write_unlock_bh(&idev->lock);
4352 return 0;
4353 }
4354
4355 static int inet6_set_link_af(struct net_device *dev, const struct nlattr *nla)
4356 {
4357 int err = -EINVAL;
4358 struct inet6_dev *idev = __in6_dev_get(dev);
4359 struct nlattr *tb[IFLA_INET6_MAX + 1];
4360
4361 if (!idev)
4362 return -EAFNOSUPPORT;
4363
4364 if (nla_parse_nested(tb, IFLA_INET6_MAX, nla, NULL) < 0)
4365 BUG();
4366
4367 if (tb[IFLA_INET6_TOKEN])
4368 err = inet6_set_iftoken(idev, nla_data(tb[IFLA_INET6_TOKEN]));
4369
4370 return err;
4371 }
4372
4373 static int inet6_fill_ifinfo(struct sk_buff *skb, struct inet6_dev *idev,
4374 u32 portid, u32 seq, int event, unsigned int flags)
4375 {
4376 struct net_device *dev = idev->dev;
4377 struct ifinfomsg *hdr;
4378 struct nlmsghdr *nlh;
4379 void *protoinfo;
4380
4381 nlh = nlmsg_put(skb, portid, seq, event, sizeof(*hdr), flags);
4382 if (nlh == NULL)
4383 return -EMSGSIZE;
4384
4385 hdr = nlmsg_data(nlh);
4386 hdr->ifi_family = AF_INET6;
4387 hdr->__ifi_pad = 0;
4388 hdr->ifi_type = dev->type;
4389 hdr->ifi_index = dev->ifindex;
4390 hdr->ifi_flags = dev_get_flags(dev);
4391 hdr->ifi_change = 0;
4392
4393 if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
4394 (dev->addr_len &&
4395 nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
4396 nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
4397 (dev->ifindex != dev->iflink &&
4398 nla_put_u32(skb, IFLA_LINK, dev->iflink)))
4399 goto nla_put_failure;
4400 protoinfo = nla_nest_start(skb, IFLA_PROTINFO);
4401 if (protoinfo == NULL)
4402 goto nla_put_failure;
4403
4404 if (inet6_fill_ifla6_attrs(skb, idev) < 0)
4405 goto nla_put_failure;
4406
4407 nla_nest_end(skb, protoinfo);
4408 return nlmsg_end(skb, nlh);
4409
4410 nla_put_failure:
4411 nlmsg_cancel(skb, nlh);
4412 return -EMSGSIZE;
4413 }
4414
4415 static int inet6_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
4416 {
4417 struct net *net = sock_net(skb->sk);
4418 int h, s_h;
4419 int idx = 0, s_idx;
4420 struct net_device *dev;
4421 struct inet6_dev *idev;
4422 struct hlist_head *head;
4423
4424 s_h = cb->args[0];
4425 s_idx = cb->args[1];
4426
4427 rcu_read_lock();
4428 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
4429 idx = 0;
4430 head = &net->dev_index_head[h];
4431 hlist_for_each_entry_rcu(dev, head, index_hlist) {
4432 if (idx < s_idx)
4433 goto cont;
4434 idev = __in6_dev_get(dev);
4435 if (!idev)
4436 goto cont;
4437 if (inet6_fill_ifinfo(skb, idev,
4438 NETLINK_CB(cb->skb).portid,
4439 cb->nlh->nlmsg_seq,
4440 RTM_NEWLINK, NLM_F_MULTI) <= 0)
4441 goto out;
4442 cont:
4443 idx++;
4444 }
4445 }
4446 out:
4447 rcu_read_unlock();
4448 cb->args[1] = idx;
4449 cb->args[0] = h;
4450
4451 return skb->len;
4452 }
4453
4454 void inet6_ifinfo_notify(int event, struct inet6_dev *idev)
4455 {
4456 struct sk_buff *skb;
4457 struct net *net = dev_net(idev->dev);
4458 int err = -ENOBUFS;
4459
4460 skb = nlmsg_new(inet6_if_nlmsg_size(), GFP_ATOMIC);
4461 if (skb == NULL)
4462 goto errout;
4463
4464 err = inet6_fill_ifinfo(skb, idev, 0, 0, event, 0);
4465 if (err < 0) {
4466 /* -EMSGSIZE implies BUG in inet6_if_nlmsg_size() */
4467 WARN_ON(err == -EMSGSIZE);
4468 kfree_skb(skb);
4469 goto errout;
4470 }
4471 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFINFO, NULL, GFP_ATOMIC);
4472 return;
4473 errout:
4474 if (err < 0)
4475 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFINFO, err);
4476 }
4477
4478 static inline size_t inet6_prefix_nlmsg_size(void)
4479 {
4480 return NLMSG_ALIGN(sizeof(struct prefixmsg))
4481 + nla_total_size(sizeof(struct in6_addr))
4482 + nla_total_size(sizeof(struct prefix_cacheinfo));
4483 }
4484
4485 static int inet6_fill_prefix(struct sk_buff *skb, struct inet6_dev *idev,
4486 struct prefix_info *pinfo, u32 portid, u32 seq,
4487 int event, unsigned int flags)
4488 {
4489 struct prefixmsg *pmsg;
4490 struct nlmsghdr *nlh;
4491 struct prefix_cacheinfo ci;
4492
4493 nlh = nlmsg_put(skb, portid, seq, event, sizeof(*pmsg), flags);
4494 if (nlh == NULL)
4495 return -EMSGSIZE;
4496
4497 pmsg = nlmsg_data(nlh);
4498 pmsg->prefix_family = AF_INET6;
4499 pmsg->prefix_pad1 = 0;
4500 pmsg->prefix_pad2 = 0;
4501 pmsg->prefix_ifindex = idev->dev->ifindex;
4502 pmsg->prefix_len = pinfo->prefix_len;
4503 pmsg->prefix_type = pinfo->type;
4504 pmsg->prefix_pad3 = 0;
4505 pmsg->prefix_flags = 0;
4506 if (pinfo->onlink)
4507 pmsg->prefix_flags |= IF_PREFIX_ONLINK;
4508 if (pinfo->autoconf)
4509 pmsg->prefix_flags |= IF_PREFIX_AUTOCONF;
4510
4511 if (nla_put(skb, PREFIX_ADDRESS, sizeof(pinfo->prefix), &pinfo->prefix))
4512 goto nla_put_failure;
4513 ci.preferred_time = ntohl(pinfo->prefered);
4514 ci.valid_time = ntohl(pinfo->valid);
4515 if (nla_put(skb, PREFIX_CACHEINFO, sizeof(ci), &ci))
4516 goto nla_put_failure;
4517 return nlmsg_end(skb, nlh);
4518
4519 nla_put_failure:
4520 nlmsg_cancel(skb, nlh);
4521 return -EMSGSIZE;
4522 }
4523
4524 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
4525 struct prefix_info *pinfo)
4526 {
4527 struct sk_buff *skb;
4528 struct net *net = dev_net(idev->dev);
4529 int err = -ENOBUFS;
4530
4531 skb = nlmsg_new(inet6_prefix_nlmsg_size(), GFP_ATOMIC);
4532 if (skb == NULL)
4533 goto errout;
4534
4535 err = inet6_fill_prefix(skb, idev, pinfo, 0, 0, event, 0);
4536 if (err < 0) {
4537 /* -EMSGSIZE implies BUG in inet6_prefix_nlmsg_size() */
4538 WARN_ON(err == -EMSGSIZE);
4539 kfree_skb(skb);
4540 goto errout;
4541 }
4542 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_PREFIX, NULL, GFP_ATOMIC);
4543 return;
4544 errout:
4545 if (err < 0)
4546 rtnl_set_sk_err(net, RTNLGRP_IPV6_PREFIX, err);
4547 }
4548
4549 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
4550 {
4551 struct net *net = dev_net(ifp->idev->dev);
4552
4553 inet6_ifa_notify(event ? : RTM_NEWADDR, ifp);
4554
4555 switch (event) {
4556 case RTM_NEWADDR:
4557 /*
4558 * If the address was optimistic
4559 * we inserted the route at the start of
4560 * our DAD process, so we don't need
4561 * to do it again
4562 */
4563 if (!(ifp->rt->rt6i_node))
4564 ip6_ins_rt(ifp->rt);
4565 if (ifp->idev->cnf.forwarding)
4566 addrconf_join_anycast(ifp);
4567 break;
4568 case RTM_DELADDR:
4569 if (ifp->idev->cnf.forwarding)
4570 addrconf_leave_anycast(ifp);
4571 addrconf_leave_solict(ifp->idev, &ifp->addr);
4572 dst_hold(&ifp->rt->dst);
4573
4574 if (ip6_del_rt(ifp->rt))
4575 dst_free(&ifp->rt->dst);
4576 break;
4577 }
4578 atomic_inc(&net->ipv6.dev_addr_genid);
4579 }
4580
4581 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
4582 {
4583 rcu_read_lock_bh();
4584 if (likely(ifp->idev->dead == 0))
4585 __ipv6_ifa_notify(event, ifp);
4586 rcu_read_unlock_bh();
4587 }
4588
4589 #ifdef CONFIG_SYSCTL
4590
4591 static
4592 int addrconf_sysctl_forward(ctl_table *ctl, int write,
4593 void __user *buffer, size_t *lenp, loff_t *ppos)
4594 {
4595 int *valp = ctl->data;
4596 int val = *valp;
4597 loff_t pos = *ppos;
4598 ctl_table lctl;
4599 int ret;
4600
4601 /*
4602 * ctl->data points to idev->cnf.forwarding, we should
4603 * not modify it until we get the rtnl lock.
4604 */
4605 lctl = *ctl;
4606 lctl.data = &val;
4607
4608 ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
4609
4610 if (write)
4611 ret = addrconf_fixup_forwarding(ctl, valp, val);
4612 if (ret)
4613 *ppos = pos;
4614 return ret;
4615 }
4616
4617 static void dev_disable_change(struct inet6_dev *idev)
4618 {
4619 if (!idev || !idev->dev)
4620 return;
4621
4622 if (idev->cnf.disable_ipv6)
4623 addrconf_notify(NULL, NETDEV_DOWN, idev->dev);
4624 else
4625 addrconf_notify(NULL, NETDEV_UP, idev->dev);
4626 }
4627
4628 static void addrconf_disable_change(struct net *net, __s32 newf)
4629 {
4630 struct net_device *dev;
4631 struct inet6_dev *idev;
4632
4633 rcu_read_lock();
4634 for_each_netdev_rcu(net, dev) {
4635 idev = __in6_dev_get(dev);
4636 if (idev) {
4637 int changed = (!idev->cnf.disable_ipv6) ^ (!newf);
4638 idev->cnf.disable_ipv6 = newf;
4639 if (changed)
4640 dev_disable_change(idev);
4641 }
4642 }
4643 rcu_read_unlock();
4644 }
4645
4646 static int addrconf_disable_ipv6(struct ctl_table *table, int *p, int newf)
4647 {
4648 struct net *net;
4649 int old;
4650
4651 if (!rtnl_trylock())
4652 return restart_syscall();
4653
4654 net = (struct net *)table->extra2;
4655 old = *p;
4656 *p = newf;
4657
4658 if (p == &net->ipv6.devconf_dflt->disable_ipv6) {
4659 rtnl_unlock();
4660 return 0;
4661 }
4662
4663 if (p == &net->ipv6.devconf_all->disable_ipv6) {
4664 net->ipv6.devconf_dflt->disable_ipv6 = newf;
4665 addrconf_disable_change(net, newf);
4666 } else if ((!newf) ^ (!old))
4667 dev_disable_change((struct inet6_dev *)table->extra1);
4668
4669 rtnl_unlock();
4670 return 0;
4671 }
4672
4673 static
4674 int addrconf_sysctl_disable(ctl_table *ctl, int write,
4675 void __user *buffer, size_t *lenp, loff_t *ppos)
4676 {
4677 int *valp = ctl->data;
4678 int val = *valp;
4679 loff_t pos = *ppos;
4680 ctl_table lctl;
4681 int ret;
4682
4683 /*
4684 * ctl->data points to idev->cnf.disable_ipv6, we should
4685 * not modify it until we get the rtnl lock.
4686 */
4687 lctl = *ctl;
4688 lctl.data = &val;
4689
4690 ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
4691
4692 if (write)
4693 ret = addrconf_disable_ipv6(ctl, valp, val);
4694 if (ret)
4695 *ppos = pos;
4696 return ret;
4697 }
4698
4699 static struct addrconf_sysctl_table
4700 {
4701 struct ctl_table_header *sysctl_header;
4702 ctl_table addrconf_vars[DEVCONF_MAX+1];
4703 } addrconf_sysctl __read_mostly = {
4704 .sysctl_header = NULL,
4705 .addrconf_vars = {
4706 {
4707 .procname = "forwarding",
4708 .data = &ipv6_devconf.forwarding,
4709 .maxlen = sizeof(int),
4710 .mode = 0644,
4711 .proc_handler = addrconf_sysctl_forward,
4712 },
4713 {
4714 .procname = "hop_limit",
4715 .data = &ipv6_devconf.hop_limit,
4716 .maxlen = sizeof(int),
4717 .mode = 0644,
4718 .proc_handler = proc_dointvec,
4719 },
4720 {
4721 .procname = "mtu",
4722 .data = &ipv6_devconf.mtu6,
4723 .maxlen = sizeof(int),
4724 .mode = 0644,
4725 .proc_handler = proc_dointvec,
4726 },
4727 {
4728 .procname = "accept_ra",
4729 .data = &ipv6_devconf.accept_ra,
4730 .maxlen = sizeof(int),
4731 .mode = 0644,
4732 .proc_handler = proc_dointvec,
4733 },
4734 {
4735 .procname = "accept_redirects",
4736 .data = &ipv6_devconf.accept_redirects,
4737 .maxlen = sizeof(int),
4738 .mode = 0644,
4739 .proc_handler = proc_dointvec,
4740 },
4741 {
4742 .procname = "autoconf",
4743 .data = &ipv6_devconf.autoconf,
4744 .maxlen = sizeof(int),
4745 .mode = 0644,
4746 .proc_handler = proc_dointvec,
4747 },
4748 {
4749 .procname = "dad_transmits",
4750 .data = &ipv6_devconf.dad_transmits,
4751 .maxlen = sizeof(int),
4752 .mode = 0644,
4753 .proc_handler = proc_dointvec,
4754 },
4755 {
4756 .procname = "router_solicitations",
4757 .data = &ipv6_devconf.rtr_solicits,
4758 .maxlen = sizeof(int),
4759 .mode = 0644,
4760 .proc_handler = proc_dointvec,
4761 },
4762 {
4763 .procname = "router_solicitation_interval",
4764 .data = &ipv6_devconf.rtr_solicit_interval,
4765 .maxlen = sizeof(int),
4766 .mode = 0644,
4767 .proc_handler = proc_dointvec_jiffies,
4768 },
4769 {
4770 .procname = "router_solicitation_delay",
4771 .data = &ipv6_devconf.rtr_solicit_delay,
4772 .maxlen = sizeof(int),
4773 .mode = 0644,
4774 .proc_handler = proc_dointvec_jiffies,
4775 },
4776 {
4777 .procname = "force_mld_version",
4778 .data = &ipv6_devconf.force_mld_version,
4779 .maxlen = sizeof(int),
4780 .mode = 0644,
4781 .proc_handler = proc_dointvec,
4782 },
4783 #ifdef CONFIG_IPV6_PRIVACY
4784 {
4785 .procname = "use_tempaddr",
4786 .data = &ipv6_devconf.use_tempaddr,
4787 .maxlen = sizeof(int),
4788 .mode = 0644,
4789 .proc_handler = proc_dointvec,
4790 },
4791 {
4792 .procname = "temp_valid_lft",
4793 .data = &ipv6_devconf.temp_valid_lft,
4794 .maxlen = sizeof(int),
4795 .mode = 0644,
4796 .proc_handler = proc_dointvec,
4797 },
4798 {
4799 .procname = "temp_prefered_lft",
4800 .data = &ipv6_devconf.temp_prefered_lft,
4801 .maxlen = sizeof(int),
4802 .mode = 0644,
4803 .proc_handler = proc_dointvec,
4804 },
4805 {
4806 .procname = "regen_max_retry",
4807 .data = &ipv6_devconf.regen_max_retry,
4808 .maxlen = sizeof(int),
4809 .mode = 0644,
4810 .proc_handler = proc_dointvec,
4811 },
4812 {
4813 .procname = "max_desync_factor",
4814 .data = &ipv6_devconf.max_desync_factor,
4815 .maxlen = sizeof(int),
4816 .mode = 0644,
4817 .proc_handler = proc_dointvec,
4818 },
4819 #endif
4820 {
4821 .procname = "max_addresses",
4822 .data = &ipv6_devconf.max_addresses,
4823 .maxlen = sizeof(int),
4824 .mode = 0644,
4825 .proc_handler = proc_dointvec,
4826 },
4827 {
4828 .procname = "accept_ra_defrtr",
4829 .data = &ipv6_devconf.accept_ra_defrtr,
4830 .maxlen = sizeof(int),
4831 .mode = 0644,
4832 .proc_handler = proc_dointvec,
4833 },
4834 {
4835 .procname = "accept_ra_pinfo",
4836 .data = &ipv6_devconf.accept_ra_pinfo,
4837 .maxlen = sizeof(int),
4838 .mode = 0644,
4839 .proc_handler = proc_dointvec,
4840 },
4841 #ifdef CONFIG_IPV6_ROUTER_PREF
4842 {
4843 .procname = "accept_ra_rtr_pref",
4844 .data = &ipv6_devconf.accept_ra_rtr_pref,
4845 .maxlen = sizeof(int),
4846 .mode = 0644,
4847 .proc_handler = proc_dointvec,
4848 },
4849 {
4850 .procname = "router_probe_interval",
4851 .data = &ipv6_devconf.rtr_probe_interval,
4852 .maxlen = sizeof(int),
4853 .mode = 0644,
4854 .proc_handler = proc_dointvec_jiffies,
4855 },
4856 #ifdef CONFIG_IPV6_ROUTE_INFO
4857 {
4858 .procname = "accept_ra_rt_info_max_plen",
4859 .data = &ipv6_devconf.accept_ra_rt_info_max_plen,
4860 .maxlen = sizeof(int),
4861 .mode = 0644,
4862 .proc_handler = proc_dointvec,
4863 },
4864 #endif
4865 #endif
4866 {
4867 .procname = "proxy_ndp",
4868 .data = &ipv6_devconf.proxy_ndp,
4869 .maxlen = sizeof(int),
4870 .mode = 0644,
4871 .proc_handler = proc_dointvec,
4872 },
4873 {
4874 .procname = "accept_source_route",
4875 .data = &ipv6_devconf.accept_source_route,
4876 .maxlen = sizeof(int),
4877 .mode = 0644,
4878 .proc_handler = proc_dointvec,
4879 },
4880 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
4881 {
4882 .procname = "optimistic_dad",
4883 .data = &ipv6_devconf.optimistic_dad,
4884 .maxlen = sizeof(int),
4885 .mode = 0644,
4886 .proc_handler = proc_dointvec,
4887
4888 },
4889 #endif
4890 #ifdef CONFIG_IPV6_MROUTE
4891 {
4892 .procname = "mc_forwarding",
4893 .data = &ipv6_devconf.mc_forwarding,
4894 .maxlen = sizeof(int),
4895 .mode = 0444,
4896 .proc_handler = proc_dointvec,
4897 },
4898 #endif
4899 {
4900 .procname = "disable_ipv6",
4901 .data = &ipv6_devconf.disable_ipv6,
4902 .maxlen = sizeof(int),
4903 .mode = 0644,
4904 .proc_handler = addrconf_sysctl_disable,
4905 },
4906 {
4907 .procname = "accept_dad",
4908 .data = &ipv6_devconf.accept_dad,
4909 .maxlen = sizeof(int),
4910 .mode = 0644,
4911 .proc_handler = proc_dointvec,
4912 },
4913 {
4914 .procname = "force_tllao",
4915 .data = &ipv6_devconf.force_tllao,
4916 .maxlen = sizeof(int),
4917 .mode = 0644,
4918 .proc_handler = proc_dointvec
4919 },
4920 {
4921 .procname = "ndisc_notify",
4922 .data = &ipv6_devconf.ndisc_notify,
4923 .maxlen = sizeof(int),
4924 .mode = 0644,
4925 .proc_handler = proc_dointvec
4926 },
4927 {
4928 /* sentinel */
4929 }
4930 },
4931 };
4932
4933 static int __addrconf_sysctl_register(struct net *net, char *dev_name,
4934 struct inet6_dev *idev, struct ipv6_devconf *p)
4935 {
4936 int i;
4937 struct addrconf_sysctl_table *t;
4938 char path[sizeof("net/ipv6/conf/") + IFNAMSIZ];
4939
4940 t = kmemdup(&addrconf_sysctl, sizeof(*t), GFP_KERNEL);
4941 if (t == NULL)
4942 goto out;
4943
4944 for (i = 0; t->addrconf_vars[i].data; i++) {
4945 t->addrconf_vars[i].data += (char *)p - (char *)&ipv6_devconf;
4946 t->addrconf_vars[i].extra1 = idev; /* embedded; no ref */
4947 t->addrconf_vars[i].extra2 = net;
4948 }
4949
4950 snprintf(path, sizeof(path), "net/ipv6/conf/%s", dev_name);
4951
4952 t->sysctl_header = register_net_sysctl(net, path, t->addrconf_vars);
4953 if (t->sysctl_header == NULL)
4954 goto free;
4955
4956 p->sysctl = t;
4957 return 0;
4958
4959 free:
4960 kfree(t);
4961 out:
4962 return -ENOBUFS;
4963 }
4964
4965 static void __addrconf_sysctl_unregister(struct ipv6_devconf *p)
4966 {
4967 struct addrconf_sysctl_table *t;
4968
4969 if (p->sysctl == NULL)
4970 return;
4971
4972 t = p->sysctl;
4973 p->sysctl = NULL;
4974 unregister_net_sysctl_table(t->sysctl_header);
4975 kfree(t);
4976 }
4977
4978 static void addrconf_sysctl_register(struct inet6_dev *idev)
4979 {
4980 neigh_sysctl_register(idev->dev, idev->nd_parms, "ipv6",
4981 &ndisc_ifinfo_sysctl_change);
4982 __addrconf_sysctl_register(dev_net(idev->dev), idev->dev->name,
4983 idev, &idev->cnf);
4984 }
4985
4986 static void addrconf_sysctl_unregister(struct inet6_dev *idev)
4987 {
4988 __addrconf_sysctl_unregister(&idev->cnf);
4989 neigh_sysctl_unregister(idev->nd_parms);
4990 }
4991
4992
4993 #endif
4994
4995 static int __net_init addrconf_init_net(struct net *net)
4996 {
4997 int err = -ENOMEM;
4998 struct ipv6_devconf *all, *dflt;
4999
5000 all = kmemdup(&ipv6_devconf, sizeof(ipv6_devconf), GFP_KERNEL);
5001 if (all == NULL)
5002 goto err_alloc_all;
5003
5004 dflt = kmemdup(&ipv6_devconf_dflt, sizeof(ipv6_devconf_dflt), GFP_KERNEL);
5005 if (dflt == NULL)
5006 goto err_alloc_dflt;
5007
5008 /* these will be inherited by all namespaces */
5009 dflt->autoconf = ipv6_defaults.autoconf;
5010 dflt->disable_ipv6 = ipv6_defaults.disable_ipv6;
5011
5012 net->ipv6.devconf_all = all;
5013 net->ipv6.devconf_dflt = dflt;
5014
5015 #ifdef CONFIG_SYSCTL
5016 err = __addrconf_sysctl_register(net, "all", NULL, all);
5017 if (err < 0)
5018 goto err_reg_all;
5019
5020 err = __addrconf_sysctl_register(net, "default", NULL, dflt);
5021 if (err < 0)
5022 goto err_reg_dflt;
5023 #endif
5024 return 0;
5025
5026 #ifdef CONFIG_SYSCTL
5027 err_reg_dflt:
5028 __addrconf_sysctl_unregister(all);
5029 err_reg_all:
5030 kfree(dflt);
5031 #endif
5032 err_alloc_dflt:
5033 kfree(all);
5034 err_alloc_all:
5035 return err;
5036 }
5037
5038 static void __net_exit addrconf_exit_net(struct net *net)
5039 {
5040 #ifdef CONFIG_SYSCTL
5041 __addrconf_sysctl_unregister(net->ipv6.devconf_dflt);
5042 __addrconf_sysctl_unregister(net->ipv6.devconf_all);
5043 #endif
5044 if (!net_eq(net, &init_net)) {
5045 kfree(net->ipv6.devconf_dflt);
5046 kfree(net->ipv6.devconf_all);
5047 }
5048 }
5049
5050 static struct pernet_operations addrconf_ops = {
5051 .init = addrconf_init_net,
5052 .exit = addrconf_exit_net,
5053 };
5054
5055 static struct rtnl_af_ops inet6_ops = {
5056 .family = AF_INET6,
5057 .fill_link_af = inet6_fill_link_af,
5058 .get_link_af_size = inet6_get_link_af_size,
5059 .set_link_af = inet6_set_link_af,
5060 };
5061
5062 /*
5063 * Init / cleanup code
5064 */
5065
5066 int __init addrconf_init(void)
5067 {
5068 int i, err;
5069
5070 err = ipv6_addr_label_init();
5071 if (err < 0) {
5072 pr_crit("%s: cannot initialize default policy table: %d\n",
5073 __func__, err);
5074 goto out;
5075 }
5076
5077 err = register_pernet_subsys(&addrconf_ops);
5078 if (err < 0)
5079 goto out_addrlabel;
5080
5081 /* The addrconf netdev notifier requires that loopback_dev
5082 * has it's ipv6 private information allocated and setup
5083 * before it can bring up and give link-local addresses
5084 * to other devices which are up.
5085 *
5086 * Unfortunately, loopback_dev is not necessarily the first
5087 * entry in the global dev_base list of net devices. In fact,
5088 * it is likely to be the very last entry on that list.
5089 * So this causes the notifier registry below to try and
5090 * give link-local addresses to all devices besides loopback_dev
5091 * first, then loopback_dev, which cases all the non-loopback_dev
5092 * devices to fail to get a link-local address.
5093 *
5094 * So, as a temporary fix, allocate the ipv6 structure for
5095 * loopback_dev first by hand.
5096 * Longer term, all of the dependencies ipv6 has upon the loopback
5097 * device and it being up should be removed.
5098 */
5099 rtnl_lock();
5100 if (!ipv6_add_dev(init_net.loopback_dev))
5101 err = -ENOMEM;
5102 rtnl_unlock();
5103 if (err)
5104 goto errlo;
5105
5106 for (i = 0; i < IN6_ADDR_HSIZE; i++)
5107 INIT_HLIST_HEAD(&inet6_addr_lst[i]);
5108
5109 register_netdevice_notifier(&ipv6_dev_notf);
5110
5111 addrconf_verify(0);
5112
5113 err = rtnl_af_register(&inet6_ops);
5114 if (err < 0)
5115 goto errout_af;
5116
5117 err = __rtnl_register(PF_INET6, RTM_GETLINK, NULL, inet6_dump_ifinfo,
5118 NULL);
5119 if (err < 0)
5120 goto errout;
5121
5122 /* Only the first call to __rtnl_register can fail */
5123 __rtnl_register(PF_INET6, RTM_NEWADDR, inet6_rtm_newaddr, NULL, NULL);
5124 __rtnl_register(PF_INET6, RTM_DELADDR, inet6_rtm_deladdr, NULL, NULL);
5125 __rtnl_register(PF_INET6, RTM_GETADDR, inet6_rtm_getaddr,
5126 inet6_dump_ifaddr, NULL);
5127 __rtnl_register(PF_INET6, RTM_GETMULTICAST, NULL,
5128 inet6_dump_ifmcaddr, NULL);
5129 __rtnl_register(PF_INET6, RTM_GETANYCAST, NULL,
5130 inet6_dump_ifacaddr, NULL);
5131 __rtnl_register(PF_INET6, RTM_GETNETCONF, inet6_netconf_get_devconf,
5132 inet6_netconf_dump_devconf, NULL);
5133
5134 ipv6_addr_label_rtnl_register();
5135
5136 return 0;
5137 errout:
5138 rtnl_af_unregister(&inet6_ops);
5139 errout_af:
5140 unregister_netdevice_notifier(&ipv6_dev_notf);
5141 errlo:
5142 unregister_pernet_subsys(&addrconf_ops);
5143 out_addrlabel:
5144 ipv6_addr_label_cleanup();
5145 out:
5146 return err;
5147 }
5148
5149 void addrconf_cleanup(void)
5150 {
5151 struct net_device *dev;
5152 int i;
5153
5154 unregister_netdevice_notifier(&ipv6_dev_notf);
5155 unregister_pernet_subsys(&addrconf_ops);
5156 ipv6_addr_label_cleanup();
5157
5158 rtnl_lock();
5159
5160 __rtnl_af_unregister(&inet6_ops);
5161
5162 /* clean dev list */
5163 for_each_netdev(&init_net, dev) {
5164 if (__in6_dev_get(dev) == NULL)
5165 continue;
5166 addrconf_ifdown(dev, 1);
5167 }
5168 addrconf_ifdown(init_net.loopback_dev, 2);
5169
5170 /*
5171 * Check hash table.
5172 */
5173 spin_lock_bh(&addrconf_hash_lock);
5174 for (i = 0; i < IN6_ADDR_HSIZE; i++)
5175 WARN_ON(!hlist_empty(&inet6_addr_lst[i]));
5176 spin_unlock_bh(&addrconf_hash_lock);
5177
5178 del_timer(&addr_chk_timer);
5179 rtnl_unlock();
5180 }