Loading...
1/*
2 * IPv6 Address [auto]configuration
3 * Linux INET6 implementation
4 *
5 * Authors:
6 * Pedro Roque <roque@di.fc.ul.pt>
7 * Alexey Kuznetsov <kuznet@ms2.inr.ac.ru>
8 *
9 * This program is free software; you can redistribute it and/or
10 * modify it under the terms of the GNU General Public License
11 * as published by the Free Software Foundation; either version
12 * 2 of the License, or (at your option) any later version.
13 */
14
15/*
16 * Changes:
17 *
18 * Janos Farkas : delete timer on ifdown
19 * <chexum@bankinf.banki.hu>
20 * Andi Kleen : kill double kfree on module
21 * unload.
22 * Maciej W. Rozycki : FDDI support
23 * sekiya@USAGI : Don't send too many RS
24 * packets.
25 * yoshfuji@USAGI : Fixed interval between DAD
26 * packets.
27 * YOSHIFUJI Hideaki @USAGI : improved accuracy of
28 * address validation timer.
29 * YOSHIFUJI Hideaki @USAGI : Privacy Extensions (RFC3041)
30 * support.
31 * Yuji SEKIYA @USAGI : Don't assign a same IPv6
32 * address on a same interface.
33 * YOSHIFUJI Hideaki @USAGI : ARCnet support
34 * YOSHIFUJI Hideaki @USAGI : convert /proc/net/if_inet6 to
35 * seq_file.
36 * YOSHIFUJI Hideaki @USAGI : improved source address
37 * selection; consider scope,
38 * status etc.
39 */
40
41#include <linux/errno.h>
42#include <linux/types.h>
43#include <linux/kernel.h>
44#include <linux/socket.h>
45#include <linux/sockios.h>
46#include <linux/net.h>
47#include <linux/in6.h>
48#include <linux/netdevice.h>
49#include <linux/if_addr.h>
50#include <linux/if_arp.h>
51#include <linux/if_arcnet.h>
52#include <linux/if_infiniband.h>
53#include <linux/route.h>
54#include <linux/inetdevice.h>
55#include <linux/init.h>
56#include <linux/slab.h>
57#ifdef CONFIG_SYSCTL
58#include <linux/sysctl.h>
59#endif
60#include <linux/capability.h>
61#include <linux/delay.h>
62#include <linux/notifier.h>
63#include <linux/string.h>
64
65#include <net/net_namespace.h>
66#include <net/sock.h>
67#include <net/snmp.h>
68
69#include <net/ipv6.h>
70#include <net/protocol.h>
71#include <net/ndisc.h>
72#include <net/ip6_route.h>
73#include <net/addrconf.h>
74#include <net/tcp.h>
75#include <net/ip.h>
76#include <net/netlink.h>
77#include <net/pkt_sched.h>
78#include <linux/if_tunnel.h>
79#include <linux/rtnetlink.h>
80
81#ifdef CONFIG_IPV6_PRIVACY
82#include <linux/random.h>
83#endif
84
85#include <linux/uaccess.h>
86#include <asm/unaligned.h>
87
88#include <linux/proc_fs.h>
89#include <linux/seq_file.h>
90
91/* Set to 3 to get tracing... */
92#define ACONF_DEBUG 2
93
94#if ACONF_DEBUG >= 3
95#define ADBG(x) printk x
96#else
97#define ADBG(x)
98#endif
99
100#define INFINITY_LIFE_TIME 0xFFFFFFFF
101
102static inline u32 cstamp_delta(unsigned long cstamp)
103{
104 return (cstamp - INITIAL_JIFFIES) * 100UL / HZ;
105}
106
107#define ADDRCONF_TIMER_FUZZ_MINUS (HZ > 50 ? HZ/50 : 1)
108#define ADDRCONF_TIMER_FUZZ (HZ / 4)
109#define ADDRCONF_TIMER_FUZZ_MAX (HZ)
110
111#ifdef CONFIG_SYSCTL
112static void addrconf_sysctl_register(struct inet6_dev *idev);
113static void addrconf_sysctl_unregister(struct inet6_dev *idev);
114#else
115static inline void addrconf_sysctl_register(struct inet6_dev *idev)
116{
117}
118
119static inline void addrconf_sysctl_unregister(struct inet6_dev *idev)
120{
121}
122#endif
123
124#ifdef CONFIG_IPV6_PRIVACY
125static int __ipv6_regen_rndid(struct inet6_dev *idev);
126static int __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr);
127static void ipv6_regen_rndid(unsigned long data);
128#endif
129
130static int ipv6_generate_eui64(u8 *eui, struct net_device *dev);
131static int ipv6_count_addresses(struct inet6_dev *idev);
132
133/*
134 * Configured unicast address hash table
135 */
136static struct hlist_head inet6_addr_lst[IN6_ADDR_HSIZE];
137static DEFINE_SPINLOCK(addrconf_hash_lock);
138
139static void addrconf_verify(unsigned long);
140
141static DEFINE_TIMER(addr_chk_timer, addrconf_verify, 0, 0);
142static DEFINE_SPINLOCK(addrconf_verify_lock);
143
144static void addrconf_join_anycast(struct inet6_ifaddr *ifp);
145static void addrconf_leave_anycast(struct inet6_ifaddr *ifp);
146
147static void addrconf_type_change(struct net_device *dev,
148 unsigned long event);
149static int addrconf_ifdown(struct net_device *dev, int how);
150
151static void addrconf_dad_start(struct inet6_ifaddr *ifp, u32 flags);
152static void addrconf_dad_timer(unsigned long data);
153static void addrconf_dad_completed(struct inet6_ifaddr *ifp);
154static void addrconf_dad_run(struct inet6_dev *idev);
155static void addrconf_rs_timer(unsigned long data);
156static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
157static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
158
159static void inet6_prefix_notify(int event, struct inet6_dev *idev,
160 struct prefix_info *pinfo);
161static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
162 struct net_device *dev);
163
164static ATOMIC_NOTIFIER_HEAD(inet6addr_chain);
165
166static struct ipv6_devconf ipv6_devconf __read_mostly = {
167 .forwarding = 0,
168 .hop_limit = IPV6_DEFAULT_HOPLIMIT,
169 .mtu6 = IPV6_MIN_MTU,
170 .accept_ra = 1,
171 .accept_redirects = 1,
172 .autoconf = 1,
173 .force_mld_version = 0,
174 .dad_transmits = 1,
175 .rtr_solicits = MAX_RTR_SOLICITATIONS,
176 .rtr_solicit_interval = RTR_SOLICITATION_INTERVAL,
177 .rtr_solicit_delay = MAX_RTR_SOLICITATION_DELAY,
178#ifdef CONFIG_IPV6_PRIVACY
179 .use_tempaddr = 0,
180 .temp_valid_lft = TEMP_VALID_LIFETIME,
181 .temp_prefered_lft = TEMP_PREFERRED_LIFETIME,
182 .regen_max_retry = REGEN_MAX_RETRY,
183 .max_desync_factor = MAX_DESYNC_FACTOR,
184#endif
185 .max_addresses = IPV6_MAX_ADDRESSES,
186 .accept_ra_defrtr = 1,
187 .accept_ra_pinfo = 1,
188#ifdef CONFIG_IPV6_ROUTER_PREF
189 .accept_ra_rtr_pref = 1,
190 .rtr_probe_interval = 60 * HZ,
191#ifdef CONFIG_IPV6_ROUTE_INFO
192 .accept_ra_rt_info_max_plen = 0,
193#endif
194#endif
195 .proxy_ndp = 0,
196 .accept_source_route = 0, /* we do not accept RH0 by default. */
197 .disable_ipv6 = 0,
198 .accept_dad = 1,
199};
200
201static struct ipv6_devconf ipv6_devconf_dflt __read_mostly = {
202 .forwarding = 0,
203 .hop_limit = IPV6_DEFAULT_HOPLIMIT,
204 .mtu6 = IPV6_MIN_MTU,
205 .accept_ra = 1,
206 .accept_redirects = 1,
207 .autoconf = 1,
208 .dad_transmits = 1,
209 .rtr_solicits = MAX_RTR_SOLICITATIONS,
210 .rtr_solicit_interval = RTR_SOLICITATION_INTERVAL,
211 .rtr_solicit_delay = MAX_RTR_SOLICITATION_DELAY,
212#ifdef CONFIG_IPV6_PRIVACY
213 .use_tempaddr = 0,
214 .temp_valid_lft = TEMP_VALID_LIFETIME,
215 .temp_prefered_lft = TEMP_PREFERRED_LIFETIME,
216 .regen_max_retry = REGEN_MAX_RETRY,
217 .max_desync_factor = MAX_DESYNC_FACTOR,
218#endif
219 .max_addresses = IPV6_MAX_ADDRESSES,
220 .accept_ra_defrtr = 1,
221 .accept_ra_pinfo = 1,
222#ifdef CONFIG_IPV6_ROUTER_PREF
223 .accept_ra_rtr_pref = 1,
224 .rtr_probe_interval = 60 * HZ,
225#ifdef CONFIG_IPV6_ROUTE_INFO
226 .accept_ra_rt_info_max_plen = 0,
227#endif
228#endif
229 .proxy_ndp = 0,
230 .accept_source_route = 0, /* we do not accept RH0 by default. */
231 .disable_ipv6 = 0,
232 .accept_dad = 1,
233};
234
235/* IPv6 Wildcard Address and Loopback Address defined by RFC2553 */
236const struct in6_addr in6addr_any = IN6ADDR_ANY_INIT;
237const struct in6_addr in6addr_loopback = IN6ADDR_LOOPBACK_INIT;
238const struct in6_addr in6addr_linklocal_allnodes = IN6ADDR_LINKLOCAL_ALLNODES_INIT;
239const struct in6_addr in6addr_linklocal_allrouters = IN6ADDR_LINKLOCAL_ALLROUTERS_INIT;
240
241/* Check if a valid qdisc is available */
242static inline bool addrconf_qdisc_ok(const struct net_device *dev)
243{
244 return !qdisc_tx_is_noop(dev);
245}
246
247/* Check if a route is valid prefix route */
248static inline int addrconf_is_prefix_route(const struct rt6_info *rt)
249{
250 return (rt->rt6i_flags & (RTF_GATEWAY | RTF_DEFAULT)) == 0;
251}
252
253static void addrconf_del_timer(struct inet6_ifaddr *ifp)
254{
255 if (del_timer(&ifp->timer))
256 __in6_ifa_put(ifp);
257}
258
259enum addrconf_timer_t {
260 AC_NONE,
261 AC_DAD,
262 AC_RS,
263};
264
265static void addrconf_mod_timer(struct inet6_ifaddr *ifp,
266 enum addrconf_timer_t what,
267 unsigned long when)
268{
269 if (!del_timer(&ifp->timer))
270 in6_ifa_hold(ifp);
271
272 switch (what) {
273 case AC_DAD:
274 ifp->timer.function = addrconf_dad_timer;
275 break;
276 case AC_RS:
277 ifp->timer.function = addrconf_rs_timer;
278 break;
279 default:
280 break;
281 }
282 ifp->timer.expires = jiffies + when;
283 add_timer(&ifp->timer);
284}
285
286static int snmp6_alloc_dev(struct inet6_dev *idev)
287{
288 if (snmp_mib_init((void __percpu **)idev->stats.ipv6,
289 sizeof(struct ipstats_mib),
290 __alignof__(struct ipstats_mib)) < 0)
291 goto err_ip;
292 idev->stats.icmpv6dev = kzalloc(sizeof(struct icmpv6_mib_device),
293 GFP_KERNEL);
294 if (!idev->stats.icmpv6dev)
295 goto err_icmp;
296 idev->stats.icmpv6msgdev = kzalloc(sizeof(struct icmpv6msg_mib_device),
297 GFP_KERNEL);
298 if (!idev->stats.icmpv6msgdev)
299 goto err_icmpmsg;
300
301 return 0;
302
303err_icmpmsg:
304 kfree(idev->stats.icmpv6dev);
305err_icmp:
306 snmp_mib_free((void __percpu **)idev->stats.ipv6);
307err_ip:
308 return -ENOMEM;
309}
310
311static void snmp6_free_dev(struct inet6_dev *idev)
312{
313 kfree(idev->stats.icmpv6msgdev);
314 kfree(idev->stats.icmpv6dev);
315 snmp_mib_free((void __percpu **)idev->stats.ipv6);
316}
317
318/* Nobody refers to this device, we may destroy it. */
319
320void in6_dev_finish_destroy(struct inet6_dev *idev)
321{
322 struct net_device *dev = idev->dev;
323
324 WARN_ON(!list_empty(&idev->addr_list));
325 WARN_ON(idev->mc_list != NULL);
326
327#ifdef NET_REFCNT_DEBUG
328 printk(KERN_DEBUG "in6_dev_finish_destroy: %s\n", dev ? dev->name : "NIL");
329#endif
330 dev_put(dev);
331 if (!idev->dead) {
332 pr_warning("Freeing alive inet6 device %p\n", idev);
333 return;
334 }
335 snmp6_free_dev(idev);
336 kfree_rcu(idev, rcu);
337}
338
339EXPORT_SYMBOL(in6_dev_finish_destroy);
340
341static struct inet6_dev * ipv6_add_dev(struct net_device *dev)
342{
343 struct inet6_dev *ndev;
344
345 ASSERT_RTNL();
346
347 if (dev->mtu < IPV6_MIN_MTU)
348 return NULL;
349
350 ndev = kzalloc(sizeof(struct inet6_dev), GFP_KERNEL);
351
352 if (ndev == NULL)
353 return NULL;
354
355 rwlock_init(&ndev->lock);
356 ndev->dev = dev;
357 INIT_LIST_HEAD(&ndev->addr_list);
358
359 memcpy(&ndev->cnf, dev_net(dev)->ipv6.devconf_dflt, sizeof(ndev->cnf));
360 ndev->cnf.mtu6 = dev->mtu;
361 ndev->cnf.sysctl = NULL;
362 ndev->nd_parms = neigh_parms_alloc(dev, &nd_tbl);
363 if (ndev->nd_parms == NULL) {
364 kfree(ndev);
365 return NULL;
366 }
367 if (ndev->cnf.forwarding)
368 dev_disable_lro(dev);
369 /* We refer to the device */
370 dev_hold(dev);
371
372 if (snmp6_alloc_dev(ndev) < 0) {
373 ADBG((KERN_WARNING
374 "%s(): cannot allocate memory for statistics; dev=%s.\n",
375 __func__, dev->name));
376 neigh_parms_release(&nd_tbl, ndev->nd_parms);
377 dev_put(dev);
378 kfree(ndev);
379 return NULL;
380 }
381
382 if (snmp6_register_dev(ndev) < 0) {
383 ADBG((KERN_WARNING
384 "%s(): cannot create /proc/net/dev_snmp6/%s\n",
385 __func__, dev->name));
386 neigh_parms_release(&nd_tbl, ndev->nd_parms);
387 ndev->dead = 1;
388 in6_dev_finish_destroy(ndev);
389 return NULL;
390 }
391
392 /* One reference from device. We must do this before
393 * we invoke __ipv6_regen_rndid().
394 */
395 in6_dev_hold(ndev);
396
397 if (dev->flags & (IFF_NOARP | IFF_LOOPBACK))
398 ndev->cnf.accept_dad = -1;
399
400#if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
401 if (dev->type == ARPHRD_SIT && (dev->priv_flags & IFF_ISATAP)) {
402 printk(KERN_INFO
403 "%s: Disabled Multicast RS\n",
404 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
424 if (netif_running(dev) && addrconf_qdisc_ok(dev))
425 ndev->if_flags |= IF_READY;
426
427 ipv6_mc_init_dev(ndev);
428 ndev->tstamp = jiffies;
429 addrconf_sysctl_register(ndev);
430 /* protected by rtnl_lock */
431 rcu_assign_pointer(dev->ip6_ptr, ndev);
432
433 /* Join all-node multicast group */
434 ipv6_dev_mc_inc(dev, &in6addr_linklocal_allnodes);
435
436 return ndev;
437}
438
439static struct inet6_dev * ipv6_find_idev(struct net_device *dev)
440{
441 struct inet6_dev *idev;
442
443 ASSERT_RTNL();
444
445 idev = __in6_dev_get(dev);
446 if (!idev) {
447 idev = ipv6_add_dev(dev);
448 if (!idev)
449 return NULL;
450 }
451
452 if (dev->flags&IFF_UP)
453 ipv6_mc_up(idev);
454 return idev;
455}
456
457#ifdef CONFIG_SYSCTL
458static void dev_forward_change(struct inet6_dev *idev)
459{
460 struct net_device *dev;
461 struct inet6_ifaddr *ifa;
462
463 if (!idev)
464 return;
465 dev = idev->dev;
466 if (idev->cnf.forwarding)
467 dev_disable_lro(dev);
468 if (dev && (dev->flags & IFF_MULTICAST)) {
469 if (idev->cnf.forwarding)
470 ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
471 else
472 ipv6_dev_mc_dec(dev, &in6addr_linklocal_allrouters);
473 }
474
475 list_for_each_entry(ifa, &idev->addr_list, if_list) {
476 if (ifa->flags&IFA_F_TENTATIVE)
477 continue;
478 if (idev->cnf.forwarding)
479 addrconf_join_anycast(ifa);
480 else
481 addrconf_leave_anycast(ifa);
482 }
483}
484
485
486static void addrconf_forward_change(struct net *net, __s32 newf)
487{
488 struct net_device *dev;
489 struct inet6_dev *idev;
490
491 rcu_read_lock();
492 for_each_netdev_rcu(net, dev) {
493 idev = __in6_dev_get(dev);
494 if (idev) {
495 int changed = (!idev->cnf.forwarding) ^ (!newf);
496 idev->cnf.forwarding = newf;
497 if (changed)
498 dev_forward_change(idev);
499 }
500 }
501 rcu_read_unlock();
502}
503
504static int addrconf_fixup_forwarding(struct ctl_table *table, int *p, int old)
505{
506 struct net *net;
507
508 net = (struct net *)table->extra2;
509 if (p == &net->ipv6.devconf_dflt->forwarding)
510 return 0;
511
512 if (!rtnl_trylock()) {
513 /* Restore the original values before restarting */
514 *p = old;
515 return restart_syscall();
516 }
517
518 if (p == &net->ipv6.devconf_all->forwarding) {
519 __s32 newf = net->ipv6.devconf_all->forwarding;
520 net->ipv6.devconf_dflt->forwarding = newf;
521 addrconf_forward_change(net, newf);
522 } else if ((!*p) ^ (!old))
523 dev_forward_change((struct inet6_dev *)table->extra1);
524 rtnl_unlock();
525
526 if (*p)
527 rt6_purge_dflt_routers(net);
528 return 1;
529}
530#endif
531
532/* Nobody refers to this ifaddr, destroy it */
533void inet6_ifa_finish_destroy(struct inet6_ifaddr *ifp)
534{
535 WARN_ON(!hlist_unhashed(&ifp->addr_lst));
536
537#ifdef NET_REFCNT_DEBUG
538 printk(KERN_DEBUG "inet6_ifa_finish_destroy\n");
539#endif
540
541 in6_dev_put(ifp->idev);
542
543 if (del_timer(&ifp->timer))
544 pr_notice("Timer is still running, when freeing ifa=%p\n", ifp);
545
546 if (ifp->state != INET6_IFADDR_STATE_DEAD) {
547 pr_warning("Freeing alive inet6 address %p\n", ifp);
548 return;
549 }
550 dst_release(&ifp->rt->dst);
551
552 kfree_rcu(ifp, rcu);
553}
554
555static void
556ipv6_link_dev_addr(struct inet6_dev *idev, struct inet6_ifaddr *ifp)
557{
558 struct list_head *p;
559 int ifp_scope = ipv6_addr_src_scope(&ifp->addr);
560
561 /*
562 * Each device address list is sorted in order of scope -
563 * global before linklocal.
564 */
565 list_for_each(p, &idev->addr_list) {
566 struct inet6_ifaddr *ifa
567 = list_entry(p, struct inet6_ifaddr, if_list);
568 if (ifp_scope >= ipv6_addr_src_scope(&ifa->addr))
569 break;
570 }
571
572 list_add_tail(&ifp->if_list, p);
573}
574
575static u32 ipv6_addr_hash(const struct in6_addr *addr)
576{
577 /*
578 * We perform the hash function over the last 64 bits of the address
579 * This will include the IEEE address token on links that support it.
580 */
581 return jhash_2words((__force u32)addr->s6_addr32[2],
582 (__force u32)addr->s6_addr32[3], 0)
583 & (IN6_ADDR_HSIZE - 1);
584}
585
586/* On success it returns ifp with increased reference count */
587
588static struct inet6_ifaddr *
589ipv6_add_addr(struct inet6_dev *idev, const struct in6_addr *addr, int pfxlen,
590 int scope, u32 flags)
591{
592 struct inet6_ifaddr *ifa = NULL;
593 struct rt6_info *rt;
594 unsigned int hash;
595 int err = 0;
596 int addr_type = ipv6_addr_type(addr);
597
598 if (addr_type == IPV6_ADDR_ANY ||
599 addr_type & IPV6_ADDR_MULTICAST ||
600 (!(idev->dev->flags & IFF_LOOPBACK) &&
601 addr_type & IPV6_ADDR_LOOPBACK))
602 return ERR_PTR(-EADDRNOTAVAIL);
603
604 rcu_read_lock_bh();
605 if (idev->dead) {
606 err = -ENODEV; /*XXX*/
607 goto out2;
608 }
609
610 if (idev->cnf.disable_ipv6) {
611 err = -EACCES;
612 goto out2;
613 }
614
615 spin_lock(&addrconf_hash_lock);
616
617 /* Ignore adding duplicate addresses on an interface */
618 if (ipv6_chk_same_addr(dev_net(idev->dev), addr, idev->dev)) {
619 ADBG(("ipv6_add_addr: already assigned\n"));
620 err = -EEXIST;
621 goto out;
622 }
623
624 ifa = kzalloc(sizeof(struct inet6_ifaddr), GFP_ATOMIC);
625
626 if (ifa == NULL) {
627 ADBG(("ipv6_add_addr: malloc failed\n"));
628 err = -ENOBUFS;
629 goto out;
630 }
631
632 rt = addrconf_dst_alloc(idev, addr, 0);
633 if (IS_ERR(rt)) {
634 err = PTR_ERR(rt);
635 goto out;
636 }
637
638 ipv6_addr_copy(&ifa->addr, addr);
639
640 spin_lock_init(&ifa->lock);
641 spin_lock_init(&ifa->state_lock);
642 init_timer(&ifa->timer);
643 INIT_HLIST_NODE(&ifa->addr_lst);
644 ifa->timer.data = (unsigned long) ifa;
645 ifa->scope = scope;
646 ifa->prefix_len = pfxlen;
647 ifa->flags = flags | IFA_F_TENTATIVE;
648 ifa->cstamp = ifa->tstamp = jiffies;
649
650 ifa->rt = rt;
651
652 /*
653 * part one of RFC 4429, section 3.3
654 * We should not configure an address as
655 * optimistic if we do not yet know the link
656 * layer address of our nexhop router
657 */
658
659 if (dst_get_neighbour_raw(&rt->dst) == NULL)
660 ifa->flags &= ~IFA_F_OPTIMISTIC;
661
662 ifa->idev = idev;
663 in6_dev_hold(idev);
664 /* For caller */
665 in6_ifa_hold(ifa);
666
667 /* Add to big hash table */
668 hash = ipv6_addr_hash(addr);
669
670 hlist_add_head_rcu(&ifa->addr_lst, &inet6_addr_lst[hash]);
671 spin_unlock(&addrconf_hash_lock);
672
673 write_lock(&idev->lock);
674 /* Add to inet6_dev unicast addr list. */
675 ipv6_link_dev_addr(idev, ifa);
676
677#ifdef CONFIG_IPV6_PRIVACY
678 if (ifa->flags&IFA_F_TEMPORARY) {
679 list_add(&ifa->tmp_list, &idev->tempaddr_list);
680 in6_ifa_hold(ifa);
681 }
682#endif
683
684 in6_ifa_hold(ifa);
685 write_unlock(&idev->lock);
686out2:
687 rcu_read_unlock_bh();
688
689 if (likely(err == 0))
690 atomic_notifier_call_chain(&inet6addr_chain, NETDEV_UP, ifa);
691 else {
692 kfree(ifa);
693 ifa = ERR_PTR(err);
694 }
695
696 return ifa;
697out:
698 spin_unlock(&addrconf_hash_lock);
699 goto out2;
700}
701
702/* This function wants to get referenced ifp and releases it before return */
703
704static void ipv6_del_addr(struct inet6_ifaddr *ifp)
705{
706 struct inet6_ifaddr *ifa, *ifn;
707 struct inet6_dev *idev = ifp->idev;
708 int state;
709 int deleted = 0, onlink = 0;
710 unsigned long expires = jiffies;
711
712 spin_lock_bh(&ifp->state_lock);
713 state = ifp->state;
714 ifp->state = INET6_IFADDR_STATE_DEAD;
715 spin_unlock_bh(&ifp->state_lock);
716
717 if (state == INET6_IFADDR_STATE_DEAD)
718 goto out;
719
720 spin_lock_bh(&addrconf_hash_lock);
721 hlist_del_init_rcu(&ifp->addr_lst);
722 spin_unlock_bh(&addrconf_hash_lock);
723
724 write_lock_bh(&idev->lock);
725#ifdef CONFIG_IPV6_PRIVACY
726 if (ifp->flags&IFA_F_TEMPORARY) {
727 list_del(&ifp->tmp_list);
728 if (ifp->ifpub) {
729 in6_ifa_put(ifp->ifpub);
730 ifp->ifpub = NULL;
731 }
732 __in6_ifa_put(ifp);
733 }
734#endif
735
736 list_for_each_entry_safe(ifa, ifn, &idev->addr_list, if_list) {
737 if (ifa == ifp) {
738 list_del_init(&ifp->if_list);
739 __in6_ifa_put(ifp);
740
741 if (!(ifp->flags & IFA_F_PERMANENT) || onlink > 0)
742 break;
743 deleted = 1;
744 continue;
745 } else if (ifp->flags & IFA_F_PERMANENT) {
746 if (ipv6_prefix_equal(&ifa->addr, &ifp->addr,
747 ifp->prefix_len)) {
748 if (ifa->flags & IFA_F_PERMANENT) {
749 onlink = 1;
750 if (deleted)
751 break;
752 } else {
753 unsigned long lifetime;
754
755 if (!onlink)
756 onlink = -1;
757
758 spin_lock(&ifa->lock);
759
760 lifetime = addrconf_timeout_fixup(ifa->valid_lft, HZ);
761 /*
762 * Note: Because this address is
763 * not permanent, lifetime <
764 * LONG_MAX / HZ here.
765 */
766 if (time_before(expires,
767 ifa->tstamp + lifetime * HZ))
768 expires = ifa->tstamp + lifetime * HZ;
769 spin_unlock(&ifa->lock);
770 }
771 }
772 }
773 }
774 write_unlock_bh(&idev->lock);
775
776 addrconf_del_timer(ifp);
777
778 ipv6_ifa_notify(RTM_DELADDR, ifp);
779
780 atomic_notifier_call_chain(&inet6addr_chain, NETDEV_DOWN, ifp);
781
782 /*
783 * Purge or update corresponding prefix
784 *
785 * 1) we don't purge prefix here if address was not permanent.
786 * prefix is managed by its own lifetime.
787 * 2) if there're no addresses, delete prefix.
788 * 3) if there're still other permanent address(es),
789 * corresponding prefix is still permanent.
790 * 4) otherwise, update prefix lifetime to the
791 * longest valid lifetime among the corresponding
792 * addresses on the device.
793 * Note: subsequent RA will update lifetime.
794 *
795 * --yoshfuji
796 */
797 if ((ifp->flags & IFA_F_PERMANENT) && onlink < 1) {
798 struct in6_addr prefix;
799 struct rt6_info *rt;
800 struct net *net = dev_net(ifp->idev->dev);
801 ipv6_addr_prefix(&prefix, &ifp->addr, ifp->prefix_len);
802 rt = rt6_lookup(net, &prefix, NULL, ifp->idev->dev->ifindex, 1);
803
804 if (rt && addrconf_is_prefix_route(rt)) {
805 if (onlink == 0) {
806 ip6_del_rt(rt);
807 rt = NULL;
808 } else if (!(rt->rt6i_flags & RTF_EXPIRES)) {
809 rt->rt6i_expires = expires;
810 rt->rt6i_flags |= RTF_EXPIRES;
811 }
812 }
813 dst_release(&rt->dst);
814 }
815
816 /* clean up prefsrc entries */
817 rt6_remove_prefsrc(ifp);
818out:
819 in6_ifa_put(ifp);
820}
821
822#ifdef CONFIG_IPV6_PRIVACY
823static int ipv6_create_tempaddr(struct inet6_ifaddr *ifp, struct inet6_ifaddr *ift)
824{
825 struct inet6_dev *idev = ifp->idev;
826 struct in6_addr addr, *tmpaddr;
827 unsigned long tmp_prefered_lft, tmp_valid_lft, tmp_cstamp, tmp_tstamp, age;
828 unsigned long regen_advance;
829 int tmp_plen;
830 int ret = 0;
831 int max_addresses;
832 u32 addr_flags;
833
834 write_lock(&idev->lock);
835 if (ift) {
836 spin_lock_bh(&ift->lock);
837 memcpy(&addr.s6_addr[8], &ift->addr.s6_addr[8], 8);
838 spin_unlock_bh(&ift->lock);
839 tmpaddr = &addr;
840 } else {
841 tmpaddr = NULL;
842 }
843retry:
844 in6_dev_hold(idev);
845 if (idev->cnf.use_tempaddr <= 0) {
846 write_unlock(&idev->lock);
847 printk(KERN_INFO
848 "ipv6_create_tempaddr(): use_tempaddr is disabled.\n");
849 in6_dev_put(idev);
850 ret = -1;
851 goto out;
852 }
853 spin_lock_bh(&ifp->lock);
854 if (ifp->regen_count++ >= idev->cnf.regen_max_retry) {
855 idev->cnf.use_tempaddr = -1; /*XXX*/
856 spin_unlock_bh(&ifp->lock);
857 write_unlock(&idev->lock);
858 printk(KERN_WARNING
859 "ipv6_create_tempaddr(): regeneration time exceeded. disabled temporary address support.\n");
860 in6_dev_put(idev);
861 ret = -1;
862 goto out;
863 }
864 in6_ifa_hold(ifp);
865 memcpy(addr.s6_addr, ifp->addr.s6_addr, 8);
866 if (__ipv6_try_regen_rndid(idev, tmpaddr) < 0) {
867 spin_unlock_bh(&ifp->lock);
868 write_unlock(&idev->lock);
869 printk(KERN_WARNING
870 "ipv6_create_tempaddr(): regeneration of randomized interface id failed.\n");
871 in6_ifa_put(ifp);
872 in6_dev_put(idev);
873 ret = -1;
874 goto out;
875 }
876 memcpy(&addr.s6_addr[8], idev->rndid, 8);
877 age = (jiffies - ifp->tstamp) / HZ;
878 tmp_valid_lft = min_t(__u32,
879 ifp->valid_lft,
880 idev->cnf.temp_valid_lft + age);
881 tmp_prefered_lft = min_t(__u32,
882 ifp->prefered_lft,
883 idev->cnf.temp_prefered_lft + age -
884 idev->cnf.max_desync_factor);
885 tmp_plen = ifp->prefix_len;
886 max_addresses = idev->cnf.max_addresses;
887 tmp_cstamp = ifp->cstamp;
888 tmp_tstamp = ifp->tstamp;
889 spin_unlock_bh(&ifp->lock);
890
891 regen_advance = idev->cnf.regen_max_retry *
892 idev->cnf.dad_transmits *
893 idev->nd_parms->retrans_time / HZ;
894 write_unlock(&idev->lock);
895
896 /* A temporary address is created only if this calculated Preferred
897 * Lifetime is greater than REGEN_ADVANCE time units. In particular,
898 * an implementation must not create a temporary address with a zero
899 * Preferred Lifetime.
900 */
901 if (tmp_prefered_lft <= regen_advance) {
902 in6_ifa_put(ifp);
903 in6_dev_put(idev);
904 ret = -1;
905 goto out;
906 }
907
908 addr_flags = IFA_F_TEMPORARY;
909 /* set in addrconf_prefix_rcv() */
910 if (ifp->flags & IFA_F_OPTIMISTIC)
911 addr_flags |= IFA_F_OPTIMISTIC;
912
913 ift = !max_addresses ||
914 ipv6_count_addresses(idev) < max_addresses ?
915 ipv6_add_addr(idev, &addr, tmp_plen,
916 ipv6_addr_type(&addr)&IPV6_ADDR_SCOPE_MASK,
917 addr_flags) : NULL;
918 if (!ift || IS_ERR(ift)) {
919 in6_ifa_put(ifp);
920 in6_dev_put(idev);
921 printk(KERN_INFO
922 "ipv6_create_tempaddr(): retry temporary address regeneration.\n");
923 tmpaddr = &addr;
924 write_lock(&idev->lock);
925 goto retry;
926 }
927
928 spin_lock_bh(&ift->lock);
929 ift->ifpub = ifp;
930 ift->valid_lft = tmp_valid_lft;
931 ift->prefered_lft = tmp_prefered_lft;
932 ift->cstamp = tmp_cstamp;
933 ift->tstamp = tmp_tstamp;
934 spin_unlock_bh(&ift->lock);
935
936 addrconf_dad_start(ift, 0);
937 in6_ifa_put(ift);
938 in6_dev_put(idev);
939out:
940 return ret;
941}
942#endif
943
944/*
945 * Choose an appropriate source address (RFC3484)
946 */
947enum {
948 IPV6_SADDR_RULE_INIT = 0,
949 IPV6_SADDR_RULE_LOCAL,
950 IPV6_SADDR_RULE_SCOPE,
951 IPV6_SADDR_RULE_PREFERRED,
952#ifdef CONFIG_IPV6_MIP6
953 IPV6_SADDR_RULE_HOA,
954#endif
955 IPV6_SADDR_RULE_OIF,
956 IPV6_SADDR_RULE_LABEL,
957#ifdef CONFIG_IPV6_PRIVACY
958 IPV6_SADDR_RULE_PRIVACY,
959#endif
960 IPV6_SADDR_RULE_ORCHID,
961 IPV6_SADDR_RULE_PREFIX,
962 IPV6_SADDR_RULE_MAX
963};
964
965struct ipv6_saddr_score {
966 int rule;
967 int addr_type;
968 struct inet6_ifaddr *ifa;
969 DECLARE_BITMAP(scorebits, IPV6_SADDR_RULE_MAX);
970 int scopedist;
971 int matchlen;
972};
973
974struct ipv6_saddr_dst {
975 const struct in6_addr *addr;
976 int ifindex;
977 int scope;
978 int label;
979 unsigned int prefs;
980};
981
982static inline int ipv6_saddr_preferred(int type)
983{
984 if (type & (IPV6_ADDR_MAPPED|IPV6_ADDR_COMPATv4|IPV6_ADDR_LOOPBACK))
985 return 1;
986 return 0;
987}
988
989static int ipv6_get_saddr_eval(struct net *net,
990 struct ipv6_saddr_score *score,
991 struct ipv6_saddr_dst *dst,
992 int i)
993{
994 int ret;
995
996 if (i <= score->rule) {
997 switch (i) {
998 case IPV6_SADDR_RULE_SCOPE:
999 ret = score->scopedist;
1000 break;
1001 case IPV6_SADDR_RULE_PREFIX:
1002 ret = score->matchlen;
1003 break;
1004 default:
1005 ret = !!test_bit(i, score->scorebits);
1006 }
1007 goto out;
1008 }
1009
1010 switch (i) {
1011 case IPV6_SADDR_RULE_INIT:
1012 /* Rule 0: remember if hiscore is not ready yet */
1013 ret = !!score->ifa;
1014 break;
1015 case IPV6_SADDR_RULE_LOCAL:
1016 /* Rule 1: Prefer same address */
1017 ret = ipv6_addr_equal(&score->ifa->addr, dst->addr);
1018 break;
1019 case IPV6_SADDR_RULE_SCOPE:
1020 /* Rule 2: Prefer appropriate scope
1021 *
1022 * ret
1023 * ^
1024 * -1 | d 15
1025 * ---+--+-+---> scope
1026 * |
1027 * | d is scope of the destination.
1028 * B-d | \
1029 * | \ <- smaller scope is better if
1030 * B-15 | \ if scope is enough for destinaion.
1031 * | ret = B - scope (-1 <= scope >= d <= 15).
1032 * d-C-1 | /
1033 * |/ <- greater is better
1034 * -C / if scope is not enough for destination.
1035 * /| ret = scope - C (-1 <= d < scope <= 15).
1036 *
1037 * d - C - 1 < B -15 (for all -1 <= d <= 15).
1038 * C > d + 14 - B >= 15 + 14 - B = 29 - B.
1039 * Assume B = 0 and we get C > 29.
1040 */
1041 ret = __ipv6_addr_src_scope(score->addr_type);
1042 if (ret >= dst->scope)
1043 ret = -ret;
1044 else
1045 ret -= 128; /* 30 is enough */
1046 score->scopedist = ret;
1047 break;
1048 case IPV6_SADDR_RULE_PREFERRED:
1049 /* Rule 3: Avoid deprecated and optimistic addresses */
1050 ret = ipv6_saddr_preferred(score->addr_type) ||
1051 !(score->ifa->flags & (IFA_F_DEPRECATED|IFA_F_OPTIMISTIC));
1052 break;
1053#ifdef CONFIG_IPV6_MIP6
1054 case IPV6_SADDR_RULE_HOA:
1055 {
1056 /* Rule 4: Prefer home address */
1057 int prefhome = !(dst->prefs & IPV6_PREFER_SRC_COA);
1058 ret = !(score->ifa->flags & IFA_F_HOMEADDRESS) ^ prefhome;
1059 break;
1060 }
1061#endif
1062 case IPV6_SADDR_RULE_OIF:
1063 /* Rule 5: Prefer outgoing interface */
1064 ret = (!dst->ifindex ||
1065 dst->ifindex == score->ifa->idev->dev->ifindex);
1066 break;
1067 case IPV6_SADDR_RULE_LABEL:
1068 /* Rule 6: Prefer matching label */
1069 ret = ipv6_addr_label(net,
1070 &score->ifa->addr, score->addr_type,
1071 score->ifa->idev->dev->ifindex) == dst->label;
1072 break;
1073#ifdef CONFIG_IPV6_PRIVACY
1074 case IPV6_SADDR_RULE_PRIVACY:
1075 {
1076 /* Rule 7: Prefer public address
1077 * Note: prefer temporary address if use_tempaddr >= 2
1078 */
1079 int preftmp = dst->prefs & (IPV6_PREFER_SRC_PUBLIC|IPV6_PREFER_SRC_TMP) ?
1080 !!(dst->prefs & IPV6_PREFER_SRC_TMP) :
1081 score->ifa->idev->cnf.use_tempaddr >= 2;
1082 ret = (!(score->ifa->flags & IFA_F_TEMPORARY)) ^ preftmp;
1083 break;
1084 }
1085#endif
1086 case IPV6_SADDR_RULE_ORCHID:
1087 /* Rule 8-: Prefer ORCHID vs ORCHID or
1088 * non-ORCHID vs non-ORCHID
1089 */
1090 ret = !(ipv6_addr_orchid(&score->ifa->addr) ^
1091 ipv6_addr_orchid(dst->addr));
1092 break;
1093 case IPV6_SADDR_RULE_PREFIX:
1094 /* Rule 8: Use longest matching prefix */
1095 score->matchlen = ret = ipv6_addr_diff(&score->ifa->addr,
1096 dst->addr);
1097 break;
1098 default:
1099 ret = 0;
1100 }
1101
1102 if (ret)
1103 __set_bit(i, score->scorebits);
1104 score->rule = i;
1105out:
1106 return ret;
1107}
1108
1109int ipv6_dev_get_saddr(struct net *net, struct net_device *dst_dev,
1110 const struct in6_addr *daddr, unsigned int prefs,
1111 struct in6_addr *saddr)
1112{
1113 struct ipv6_saddr_score scores[2],
1114 *score = &scores[0], *hiscore = &scores[1];
1115 struct ipv6_saddr_dst dst;
1116 struct net_device *dev;
1117 int dst_type;
1118
1119 dst_type = __ipv6_addr_type(daddr);
1120 dst.addr = daddr;
1121 dst.ifindex = dst_dev ? dst_dev->ifindex : 0;
1122 dst.scope = __ipv6_addr_src_scope(dst_type);
1123 dst.label = ipv6_addr_label(net, daddr, dst_type, dst.ifindex);
1124 dst.prefs = prefs;
1125
1126 hiscore->rule = -1;
1127 hiscore->ifa = NULL;
1128
1129 rcu_read_lock();
1130
1131 for_each_netdev_rcu(net, dev) {
1132 struct inet6_dev *idev;
1133
1134 /* Candidate Source Address (section 4)
1135 * - multicast and link-local destination address,
1136 * the set of candidate source address MUST only
1137 * include addresses assigned to interfaces
1138 * belonging to the same link as the outgoing
1139 * interface.
1140 * (- For site-local destination addresses, the
1141 * set of candidate source addresses MUST only
1142 * include addresses assigned to interfaces
1143 * belonging to the same site as the outgoing
1144 * interface.)
1145 */
1146 if (((dst_type & IPV6_ADDR_MULTICAST) ||
1147 dst.scope <= IPV6_ADDR_SCOPE_LINKLOCAL) &&
1148 dst.ifindex && dev->ifindex != dst.ifindex)
1149 continue;
1150
1151 idev = __in6_dev_get(dev);
1152 if (!idev)
1153 continue;
1154
1155 read_lock_bh(&idev->lock);
1156 list_for_each_entry(score->ifa, &idev->addr_list, if_list) {
1157 int i;
1158
1159 /*
1160 * - Tentative Address (RFC2462 section 5.4)
1161 * - A tentative address is not considered
1162 * "assigned to an interface" in the traditional
1163 * sense, unless it is also flagged as optimistic.
1164 * - Candidate Source Address (section 4)
1165 * - In any case, anycast addresses, multicast
1166 * addresses, and the unspecified address MUST
1167 * NOT be included in a candidate set.
1168 */
1169 if ((score->ifa->flags & IFA_F_TENTATIVE) &&
1170 (!(score->ifa->flags & IFA_F_OPTIMISTIC)))
1171 continue;
1172
1173 score->addr_type = __ipv6_addr_type(&score->ifa->addr);
1174
1175 if (unlikely(score->addr_type == IPV6_ADDR_ANY ||
1176 score->addr_type & IPV6_ADDR_MULTICAST)) {
1177 LIMIT_NETDEBUG(KERN_DEBUG
1178 "ADDRCONF: unspecified / multicast address "
1179 "assigned as unicast address on %s",
1180 dev->name);
1181 continue;
1182 }
1183
1184 score->rule = -1;
1185 bitmap_zero(score->scorebits, IPV6_SADDR_RULE_MAX);
1186
1187 for (i = 0; i < IPV6_SADDR_RULE_MAX; i++) {
1188 int minihiscore, miniscore;
1189
1190 minihiscore = ipv6_get_saddr_eval(net, hiscore, &dst, i);
1191 miniscore = ipv6_get_saddr_eval(net, score, &dst, i);
1192
1193 if (minihiscore > miniscore) {
1194 if (i == IPV6_SADDR_RULE_SCOPE &&
1195 score->scopedist > 0) {
1196 /*
1197 * special case:
1198 * each remaining entry
1199 * has too small (not enough)
1200 * scope, because ifa entries
1201 * are sorted by their scope
1202 * values.
1203 */
1204 goto try_nextdev;
1205 }
1206 break;
1207 } else if (minihiscore < miniscore) {
1208 if (hiscore->ifa)
1209 in6_ifa_put(hiscore->ifa);
1210
1211 in6_ifa_hold(score->ifa);
1212
1213 swap(hiscore, score);
1214
1215 /* restore our iterator */
1216 score->ifa = hiscore->ifa;
1217
1218 break;
1219 }
1220 }
1221 }
1222try_nextdev:
1223 read_unlock_bh(&idev->lock);
1224 }
1225 rcu_read_unlock();
1226
1227 if (!hiscore->ifa)
1228 return -EADDRNOTAVAIL;
1229
1230 ipv6_addr_copy(saddr, &hiscore->ifa->addr);
1231 in6_ifa_put(hiscore->ifa);
1232 return 0;
1233}
1234EXPORT_SYMBOL(ipv6_dev_get_saddr);
1235
1236int ipv6_get_lladdr(struct net_device *dev, struct in6_addr *addr,
1237 unsigned char banned_flags)
1238{
1239 struct inet6_dev *idev;
1240 int err = -EADDRNOTAVAIL;
1241
1242 rcu_read_lock();
1243 idev = __in6_dev_get(dev);
1244 if (idev) {
1245 struct inet6_ifaddr *ifp;
1246
1247 read_lock_bh(&idev->lock);
1248 list_for_each_entry(ifp, &idev->addr_list, if_list) {
1249 if (ifp->scope == IFA_LINK &&
1250 !(ifp->flags & banned_flags)) {
1251 ipv6_addr_copy(addr, &ifp->addr);
1252 err = 0;
1253 break;
1254 }
1255 }
1256 read_unlock_bh(&idev->lock);
1257 }
1258 rcu_read_unlock();
1259 return err;
1260}
1261
1262static int ipv6_count_addresses(struct inet6_dev *idev)
1263{
1264 int cnt = 0;
1265 struct inet6_ifaddr *ifp;
1266
1267 read_lock_bh(&idev->lock);
1268 list_for_each_entry(ifp, &idev->addr_list, if_list)
1269 cnt++;
1270 read_unlock_bh(&idev->lock);
1271 return cnt;
1272}
1273
1274int ipv6_chk_addr(struct net *net, const struct in6_addr *addr,
1275 struct net_device *dev, int strict)
1276{
1277 struct inet6_ifaddr *ifp;
1278 struct hlist_node *node;
1279 unsigned int hash = ipv6_addr_hash(addr);
1280
1281 rcu_read_lock_bh();
1282 hlist_for_each_entry_rcu(ifp, node, &inet6_addr_lst[hash], addr_lst) {
1283 if (!net_eq(dev_net(ifp->idev->dev), net))
1284 continue;
1285 if (ipv6_addr_equal(&ifp->addr, addr) &&
1286 !(ifp->flags&IFA_F_TENTATIVE) &&
1287 (dev == NULL || ifp->idev->dev == dev ||
1288 !(ifp->scope&(IFA_LINK|IFA_HOST) || strict))) {
1289 rcu_read_unlock_bh();
1290 return 1;
1291 }
1292 }
1293
1294 rcu_read_unlock_bh();
1295 return 0;
1296}
1297EXPORT_SYMBOL(ipv6_chk_addr);
1298
1299static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
1300 struct net_device *dev)
1301{
1302 unsigned int hash = ipv6_addr_hash(addr);
1303 struct inet6_ifaddr *ifp;
1304 struct hlist_node *node;
1305
1306 hlist_for_each_entry(ifp, node, &inet6_addr_lst[hash], addr_lst) {
1307 if (!net_eq(dev_net(ifp->idev->dev), net))
1308 continue;
1309 if (ipv6_addr_equal(&ifp->addr, addr)) {
1310 if (dev == NULL || ifp->idev->dev == dev)
1311 return true;
1312 }
1313 }
1314 return false;
1315}
1316
1317int ipv6_chk_prefix(const struct in6_addr *addr, struct net_device *dev)
1318{
1319 struct inet6_dev *idev;
1320 struct inet6_ifaddr *ifa;
1321 int onlink;
1322
1323 onlink = 0;
1324 rcu_read_lock();
1325 idev = __in6_dev_get(dev);
1326 if (idev) {
1327 read_lock_bh(&idev->lock);
1328 list_for_each_entry(ifa, &idev->addr_list, if_list) {
1329 onlink = ipv6_prefix_equal(addr, &ifa->addr,
1330 ifa->prefix_len);
1331 if (onlink)
1332 break;
1333 }
1334 read_unlock_bh(&idev->lock);
1335 }
1336 rcu_read_unlock();
1337 return onlink;
1338}
1339
1340EXPORT_SYMBOL(ipv6_chk_prefix);
1341
1342struct inet6_ifaddr *ipv6_get_ifaddr(struct net *net, const struct in6_addr *addr,
1343 struct net_device *dev, int strict)
1344{
1345 struct inet6_ifaddr *ifp, *result = NULL;
1346 unsigned int hash = ipv6_addr_hash(addr);
1347 struct hlist_node *node;
1348
1349 rcu_read_lock_bh();
1350 hlist_for_each_entry_rcu_bh(ifp, node, &inet6_addr_lst[hash], addr_lst) {
1351 if (!net_eq(dev_net(ifp->idev->dev), net))
1352 continue;
1353 if (ipv6_addr_equal(&ifp->addr, addr)) {
1354 if (dev == NULL || ifp->idev->dev == dev ||
1355 !(ifp->scope&(IFA_LINK|IFA_HOST) || strict)) {
1356 result = ifp;
1357 in6_ifa_hold(ifp);
1358 break;
1359 }
1360 }
1361 }
1362 rcu_read_unlock_bh();
1363
1364 return result;
1365}
1366
1367/* Gets referenced address, destroys ifaddr */
1368
1369static void addrconf_dad_stop(struct inet6_ifaddr *ifp, int dad_failed)
1370{
1371 if (ifp->flags&IFA_F_PERMANENT) {
1372 spin_lock_bh(&ifp->lock);
1373 addrconf_del_timer(ifp);
1374 ifp->flags |= IFA_F_TENTATIVE;
1375 if (dad_failed)
1376 ifp->flags |= IFA_F_DADFAILED;
1377 spin_unlock_bh(&ifp->lock);
1378 if (dad_failed)
1379 ipv6_ifa_notify(0, ifp);
1380 in6_ifa_put(ifp);
1381#ifdef CONFIG_IPV6_PRIVACY
1382 } else if (ifp->flags&IFA_F_TEMPORARY) {
1383 struct inet6_ifaddr *ifpub;
1384 spin_lock_bh(&ifp->lock);
1385 ifpub = ifp->ifpub;
1386 if (ifpub) {
1387 in6_ifa_hold(ifpub);
1388 spin_unlock_bh(&ifp->lock);
1389 ipv6_create_tempaddr(ifpub, ifp);
1390 in6_ifa_put(ifpub);
1391 } else {
1392 spin_unlock_bh(&ifp->lock);
1393 }
1394 ipv6_del_addr(ifp);
1395#endif
1396 } else
1397 ipv6_del_addr(ifp);
1398}
1399
1400static int addrconf_dad_end(struct inet6_ifaddr *ifp)
1401{
1402 int err = -ENOENT;
1403
1404 spin_lock(&ifp->state_lock);
1405 if (ifp->state == INET6_IFADDR_STATE_DAD) {
1406 ifp->state = INET6_IFADDR_STATE_POSTDAD;
1407 err = 0;
1408 }
1409 spin_unlock(&ifp->state_lock);
1410
1411 return err;
1412}
1413
1414void addrconf_dad_failure(struct inet6_ifaddr *ifp)
1415{
1416 struct inet6_dev *idev = ifp->idev;
1417
1418 if (addrconf_dad_end(ifp)) {
1419 in6_ifa_put(ifp);
1420 return;
1421 }
1422
1423 if (net_ratelimit())
1424 printk(KERN_INFO "%s: IPv6 duplicate address %pI6c detected!\n",
1425 ifp->idev->dev->name, &ifp->addr);
1426
1427 if (idev->cnf.accept_dad > 1 && !idev->cnf.disable_ipv6) {
1428 struct in6_addr addr;
1429
1430 addr.s6_addr32[0] = htonl(0xfe800000);
1431 addr.s6_addr32[1] = 0;
1432
1433 if (!ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) &&
1434 ipv6_addr_equal(&ifp->addr, &addr)) {
1435 /* DAD failed for link-local based on MAC address */
1436 idev->cnf.disable_ipv6 = 1;
1437
1438 printk(KERN_INFO "%s: IPv6 being disabled!\n",
1439 ifp->idev->dev->name);
1440 }
1441 }
1442
1443 addrconf_dad_stop(ifp, 1);
1444}
1445
1446/* Join to solicited addr multicast group. */
1447
1448void addrconf_join_solict(struct net_device *dev, const struct in6_addr *addr)
1449{
1450 struct in6_addr maddr;
1451
1452 if (dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1453 return;
1454
1455 addrconf_addr_solict_mult(addr, &maddr);
1456 ipv6_dev_mc_inc(dev, &maddr);
1457}
1458
1459void addrconf_leave_solict(struct inet6_dev *idev, const struct in6_addr *addr)
1460{
1461 struct in6_addr maddr;
1462
1463 if (idev->dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1464 return;
1465
1466 addrconf_addr_solict_mult(addr, &maddr);
1467 __ipv6_dev_mc_dec(idev, &maddr);
1468}
1469
1470static void addrconf_join_anycast(struct inet6_ifaddr *ifp)
1471{
1472 struct in6_addr addr;
1473 if (ifp->prefix_len == 127) /* RFC 6164 */
1474 return;
1475 ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1476 if (ipv6_addr_any(&addr))
1477 return;
1478 ipv6_dev_ac_inc(ifp->idev->dev, &addr);
1479}
1480
1481static void addrconf_leave_anycast(struct inet6_ifaddr *ifp)
1482{
1483 struct in6_addr addr;
1484 if (ifp->prefix_len == 127) /* RFC 6164 */
1485 return;
1486 ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1487 if (ipv6_addr_any(&addr))
1488 return;
1489 __ipv6_dev_ac_dec(ifp->idev, &addr);
1490}
1491
1492static int addrconf_ifid_eui48(u8 *eui, struct net_device *dev)
1493{
1494 if (dev->addr_len != ETH_ALEN)
1495 return -1;
1496 memcpy(eui, dev->dev_addr, 3);
1497 memcpy(eui + 5, dev->dev_addr + 3, 3);
1498
1499 /*
1500 * The zSeries OSA network cards can be shared among various
1501 * OS instances, but the OSA cards have only one MAC address.
1502 * This leads to duplicate address conflicts in conjunction
1503 * with IPv6 if more than one instance uses the same card.
1504 *
1505 * The driver for these cards can deliver a unique 16-bit
1506 * identifier for each instance sharing the same card. It is
1507 * placed instead of 0xFFFE in the interface identifier. The
1508 * "u" bit of the interface identifier is not inverted in this
1509 * case. Hence the resulting interface identifier has local
1510 * scope according to RFC2373.
1511 */
1512 if (dev->dev_id) {
1513 eui[3] = (dev->dev_id >> 8) & 0xFF;
1514 eui[4] = dev->dev_id & 0xFF;
1515 } else {
1516 eui[3] = 0xFF;
1517 eui[4] = 0xFE;
1518 eui[0] ^= 2;
1519 }
1520 return 0;
1521}
1522
1523static int addrconf_ifid_arcnet(u8 *eui, struct net_device *dev)
1524{
1525 /* XXX: inherit EUI-64 from other interface -- yoshfuji */
1526 if (dev->addr_len != ARCNET_ALEN)
1527 return -1;
1528 memset(eui, 0, 7);
1529 eui[7] = *(u8*)dev->dev_addr;
1530 return 0;
1531}
1532
1533static int addrconf_ifid_infiniband(u8 *eui, struct net_device *dev)
1534{
1535 if (dev->addr_len != INFINIBAND_ALEN)
1536 return -1;
1537 memcpy(eui, dev->dev_addr + 12, 8);
1538 eui[0] |= 2;
1539 return 0;
1540}
1541
1542static int __ipv6_isatap_ifid(u8 *eui, __be32 addr)
1543{
1544 if (addr == 0)
1545 return -1;
1546 eui[0] = (ipv4_is_zeronet(addr) || ipv4_is_private_10(addr) ||
1547 ipv4_is_loopback(addr) || ipv4_is_linklocal_169(addr) ||
1548 ipv4_is_private_172(addr) || ipv4_is_test_192(addr) ||
1549 ipv4_is_anycast_6to4(addr) || ipv4_is_private_192(addr) ||
1550 ipv4_is_test_198(addr) || ipv4_is_multicast(addr) ||
1551 ipv4_is_lbcast(addr)) ? 0x00 : 0x02;
1552 eui[1] = 0;
1553 eui[2] = 0x5E;
1554 eui[3] = 0xFE;
1555 memcpy(eui + 4, &addr, 4);
1556 return 0;
1557}
1558
1559static int addrconf_ifid_sit(u8 *eui, struct net_device *dev)
1560{
1561 if (dev->priv_flags & IFF_ISATAP)
1562 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
1563 return -1;
1564}
1565
1566static int addrconf_ifid_gre(u8 *eui, struct net_device *dev)
1567{
1568 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
1569}
1570
1571static int ipv6_generate_eui64(u8 *eui, struct net_device *dev)
1572{
1573 switch (dev->type) {
1574 case ARPHRD_ETHER:
1575 case ARPHRD_FDDI:
1576 case ARPHRD_IEEE802_TR:
1577 return addrconf_ifid_eui48(eui, dev);
1578 case ARPHRD_ARCNET:
1579 return addrconf_ifid_arcnet(eui, dev);
1580 case ARPHRD_INFINIBAND:
1581 return addrconf_ifid_infiniband(eui, dev);
1582 case ARPHRD_SIT:
1583 return addrconf_ifid_sit(eui, dev);
1584 case ARPHRD_IPGRE:
1585 return addrconf_ifid_gre(eui, dev);
1586 }
1587 return -1;
1588}
1589
1590static int ipv6_inherit_eui64(u8 *eui, struct inet6_dev *idev)
1591{
1592 int err = -1;
1593 struct inet6_ifaddr *ifp;
1594
1595 read_lock_bh(&idev->lock);
1596 list_for_each_entry(ifp, &idev->addr_list, if_list) {
1597 if (ifp->scope == IFA_LINK && !(ifp->flags&IFA_F_TENTATIVE)) {
1598 memcpy(eui, ifp->addr.s6_addr+8, 8);
1599 err = 0;
1600 break;
1601 }
1602 }
1603 read_unlock_bh(&idev->lock);
1604 return err;
1605}
1606
1607#ifdef CONFIG_IPV6_PRIVACY
1608/* (re)generation of randomized interface identifier (RFC 3041 3.2, 3.5) */
1609static int __ipv6_regen_rndid(struct inet6_dev *idev)
1610{
1611regen:
1612 get_random_bytes(idev->rndid, sizeof(idev->rndid));
1613 idev->rndid[0] &= ~0x02;
1614
1615 /*
1616 * <draft-ietf-ipngwg-temp-addresses-v2-00.txt>:
1617 * check if generated address is not inappropriate
1618 *
1619 * - Reserved subnet anycast (RFC 2526)
1620 * 11111101 11....11 1xxxxxxx
1621 * - ISATAP (RFC4214) 6.1
1622 * 00-00-5E-FE-xx-xx-xx-xx
1623 * - value 0
1624 * - XXX: already assigned to an address on the device
1625 */
1626 if (idev->rndid[0] == 0xfd &&
1627 (idev->rndid[1]&idev->rndid[2]&idev->rndid[3]&idev->rndid[4]&idev->rndid[5]&idev->rndid[6]) == 0xff &&
1628 (idev->rndid[7]&0x80))
1629 goto regen;
1630 if ((idev->rndid[0]|idev->rndid[1]) == 0) {
1631 if (idev->rndid[2] == 0x5e && idev->rndid[3] == 0xfe)
1632 goto regen;
1633 if ((idev->rndid[2]|idev->rndid[3]|idev->rndid[4]|idev->rndid[5]|idev->rndid[6]|idev->rndid[7]) == 0x00)
1634 goto regen;
1635 }
1636
1637 return 0;
1638}
1639
1640static void ipv6_regen_rndid(unsigned long data)
1641{
1642 struct inet6_dev *idev = (struct inet6_dev *) data;
1643 unsigned long expires;
1644
1645 rcu_read_lock_bh();
1646 write_lock_bh(&idev->lock);
1647
1648 if (idev->dead)
1649 goto out;
1650
1651 if (__ipv6_regen_rndid(idev) < 0)
1652 goto out;
1653
1654 expires = jiffies +
1655 idev->cnf.temp_prefered_lft * HZ -
1656 idev->cnf.regen_max_retry * idev->cnf.dad_transmits * idev->nd_parms->retrans_time -
1657 idev->cnf.max_desync_factor * HZ;
1658 if (time_before(expires, jiffies)) {
1659 printk(KERN_WARNING
1660 "ipv6_regen_rndid(): too short regeneration interval; timer disabled for %s.\n",
1661 idev->dev->name);
1662 goto out;
1663 }
1664
1665 if (!mod_timer(&idev->regen_timer, expires))
1666 in6_dev_hold(idev);
1667
1668out:
1669 write_unlock_bh(&idev->lock);
1670 rcu_read_unlock_bh();
1671 in6_dev_put(idev);
1672}
1673
1674static int __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr) {
1675 int ret = 0;
1676
1677 if (tmpaddr && memcmp(idev->rndid, &tmpaddr->s6_addr[8], 8) == 0)
1678 ret = __ipv6_regen_rndid(idev);
1679 return ret;
1680}
1681#endif
1682
1683/*
1684 * Add prefix route.
1685 */
1686
1687static void
1688addrconf_prefix_route(struct in6_addr *pfx, int plen, struct net_device *dev,
1689 unsigned long expires, u32 flags)
1690{
1691 struct fib6_config cfg = {
1692 .fc_table = RT6_TABLE_PREFIX,
1693 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1694 .fc_ifindex = dev->ifindex,
1695 .fc_expires = expires,
1696 .fc_dst_len = plen,
1697 .fc_flags = RTF_UP | flags,
1698 .fc_nlinfo.nl_net = dev_net(dev),
1699 .fc_protocol = RTPROT_KERNEL,
1700 };
1701
1702 ipv6_addr_copy(&cfg.fc_dst, pfx);
1703
1704 /* Prevent useless cloning on PtP SIT.
1705 This thing is done here expecting that the whole
1706 class of non-broadcast devices need not cloning.
1707 */
1708#if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
1709 if (dev->type == ARPHRD_SIT && (dev->flags & IFF_POINTOPOINT))
1710 cfg.fc_flags |= RTF_NONEXTHOP;
1711#endif
1712
1713 ip6_route_add(&cfg);
1714}
1715
1716/* Create "default" multicast route to the interface */
1717
1718static void addrconf_add_mroute(struct net_device *dev)
1719{
1720 struct fib6_config cfg = {
1721 .fc_table = RT6_TABLE_LOCAL,
1722 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1723 .fc_ifindex = dev->ifindex,
1724 .fc_dst_len = 8,
1725 .fc_flags = RTF_UP,
1726 .fc_nlinfo.nl_net = dev_net(dev),
1727 };
1728
1729 ipv6_addr_set(&cfg.fc_dst, htonl(0xFF000000), 0, 0, 0);
1730
1731 ip6_route_add(&cfg);
1732}
1733
1734#if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
1735static void sit_route_add(struct net_device *dev)
1736{
1737 struct fib6_config cfg = {
1738 .fc_table = RT6_TABLE_MAIN,
1739 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1740 .fc_ifindex = dev->ifindex,
1741 .fc_dst_len = 96,
1742 .fc_flags = RTF_UP | RTF_NONEXTHOP,
1743 .fc_nlinfo.nl_net = dev_net(dev),
1744 };
1745
1746 /* prefix length - 96 bits "::d.d.d.d" */
1747 ip6_route_add(&cfg);
1748}
1749#endif
1750
1751static void addrconf_add_lroute(struct net_device *dev)
1752{
1753 struct in6_addr addr;
1754
1755 ipv6_addr_set(&addr, htonl(0xFE800000), 0, 0, 0);
1756 addrconf_prefix_route(&addr, 64, dev, 0, 0);
1757}
1758
1759static struct inet6_dev *addrconf_add_dev(struct net_device *dev)
1760{
1761 struct inet6_dev *idev;
1762
1763 ASSERT_RTNL();
1764
1765 idev = ipv6_find_idev(dev);
1766 if (!idev)
1767 return ERR_PTR(-ENOBUFS);
1768
1769 if (idev->cnf.disable_ipv6)
1770 return ERR_PTR(-EACCES);
1771
1772 /* Add default multicast route */
1773 addrconf_add_mroute(dev);
1774
1775 /* Add link local route */
1776 addrconf_add_lroute(dev);
1777 return idev;
1778}
1779
1780void addrconf_prefix_rcv(struct net_device *dev, u8 *opt, int len)
1781{
1782 struct prefix_info *pinfo;
1783 __u32 valid_lft;
1784 __u32 prefered_lft;
1785 int addr_type;
1786 struct inet6_dev *in6_dev;
1787 struct net *net = dev_net(dev);
1788
1789 pinfo = (struct prefix_info *) opt;
1790
1791 if (len < sizeof(struct prefix_info)) {
1792 ADBG(("addrconf: prefix option too short\n"));
1793 return;
1794 }
1795
1796 /*
1797 * Validation checks ([ADDRCONF], page 19)
1798 */
1799
1800 addr_type = ipv6_addr_type(&pinfo->prefix);
1801
1802 if (addr_type & (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL))
1803 return;
1804
1805 valid_lft = ntohl(pinfo->valid);
1806 prefered_lft = ntohl(pinfo->prefered);
1807
1808 if (prefered_lft > valid_lft) {
1809 if (net_ratelimit())
1810 printk(KERN_WARNING "addrconf: prefix option has invalid lifetime\n");
1811 return;
1812 }
1813
1814 in6_dev = in6_dev_get(dev);
1815
1816 if (in6_dev == NULL) {
1817 if (net_ratelimit())
1818 printk(KERN_DEBUG "addrconf: device %s not configured\n", dev->name);
1819 return;
1820 }
1821
1822 /*
1823 * Two things going on here:
1824 * 1) Add routes for on-link prefixes
1825 * 2) Configure prefixes with the auto flag set
1826 */
1827
1828 if (pinfo->onlink) {
1829 struct rt6_info *rt;
1830 unsigned long rt_expires;
1831
1832 /* Avoid arithmetic overflow. Really, we could
1833 * save rt_expires in seconds, likely valid_lft,
1834 * but it would require division in fib gc, that it
1835 * not good.
1836 */
1837 if (HZ > USER_HZ)
1838 rt_expires = addrconf_timeout_fixup(valid_lft, HZ);
1839 else
1840 rt_expires = addrconf_timeout_fixup(valid_lft, USER_HZ);
1841
1842 if (addrconf_finite_timeout(rt_expires))
1843 rt_expires *= HZ;
1844
1845 rt = rt6_lookup(net, &pinfo->prefix, NULL,
1846 dev->ifindex, 1);
1847
1848 if (rt && addrconf_is_prefix_route(rt)) {
1849 /* Autoconf prefix route */
1850 if (valid_lft == 0) {
1851 ip6_del_rt(rt);
1852 rt = NULL;
1853 } else if (addrconf_finite_timeout(rt_expires)) {
1854 /* not infinity */
1855 rt->rt6i_expires = jiffies + rt_expires;
1856 rt->rt6i_flags |= RTF_EXPIRES;
1857 } else {
1858 rt->rt6i_flags &= ~RTF_EXPIRES;
1859 rt->rt6i_expires = 0;
1860 }
1861 } else if (valid_lft) {
1862 clock_t expires = 0;
1863 int flags = RTF_ADDRCONF | RTF_PREFIX_RT;
1864 if (addrconf_finite_timeout(rt_expires)) {
1865 /* not infinity */
1866 flags |= RTF_EXPIRES;
1867 expires = jiffies_to_clock_t(rt_expires);
1868 }
1869 addrconf_prefix_route(&pinfo->prefix, pinfo->prefix_len,
1870 dev, expires, flags);
1871 }
1872 if (rt)
1873 dst_release(&rt->dst);
1874 }
1875
1876 /* Try to figure out our local address for this prefix */
1877
1878 if (pinfo->autoconf && in6_dev->cnf.autoconf) {
1879 struct inet6_ifaddr * ifp;
1880 struct in6_addr addr;
1881 int create = 0, update_lft = 0;
1882
1883 if (pinfo->prefix_len == 64) {
1884 memcpy(&addr, &pinfo->prefix, 8);
1885 if (ipv6_generate_eui64(addr.s6_addr + 8, dev) &&
1886 ipv6_inherit_eui64(addr.s6_addr + 8, in6_dev)) {
1887 in6_dev_put(in6_dev);
1888 return;
1889 }
1890 goto ok;
1891 }
1892 if (net_ratelimit())
1893 printk(KERN_DEBUG "IPv6 addrconf: prefix with wrong length %d\n",
1894 pinfo->prefix_len);
1895 in6_dev_put(in6_dev);
1896 return;
1897
1898ok:
1899
1900 ifp = ipv6_get_ifaddr(net, &addr, dev, 1);
1901
1902 if (ifp == NULL && valid_lft) {
1903 int max_addresses = in6_dev->cnf.max_addresses;
1904 u32 addr_flags = 0;
1905
1906#ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1907 if (in6_dev->cnf.optimistic_dad &&
1908 !net->ipv6.devconf_all->forwarding)
1909 addr_flags = IFA_F_OPTIMISTIC;
1910#endif
1911
1912 /* Do not allow to create too much of autoconfigured
1913 * addresses; this would be too easy way to crash kernel.
1914 */
1915 if (!max_addresses ||
1916 ipv6_count_addresses(in6_dev) < max_addresses)
1917 ifp = ipv6_add_addr(in6_dev, &addr, pinfo->prefix_len,
1918 addr_type&IPV6_ADDR_SCOPE_MASK,
1919 addr_flags);
1920
1921 if (!ifp || IS_ERR(ifp)) {
1922 in6_dev_put(in6_dev);
1923 return;
1924 }
1925
1926 update_lft = create = 1;
1927 ifp->cstamp = jiffies;
1928 addrconf_dad_start(ifp, RTF_ADDRCONF|RTF_PREFIX_RT);
1929 }
1930
1931 if (ifp) {
1932 int flags;
1933 unsigned long now;
1934#ifdef CONFIG_IPV6_PRIVACY
1935 struct inet6_ifaddr *ift;
1936#endif
1937 u32 stored_lft;
1938
1939 /* update lifetime (RFC2462 5.5.3 e) */
1940 spin_lock(&ifp->lock);
1941 now = jiffies;
1942 if (ifp->valid_lft > (now - ifp->tstamp) / HZ)
1943 stored_lft = ifp->valid_lft - (now - ifp->tstamp) / HZ;
1944 else
1945 stored_lft = 0;
1946 if (!update_lft && stored_lft) {
1947 if (valid_lft > MIN_VALID_LIFETIME ||
1948 valid_lft > stored_lft)
1949 update_lft = 1;
1950 else if (stored_lft <= MIN_VALID_LIFETIME) {
1951 /* valid_lft <= stored_lft is always true */
1952 /*
1953 * RFC 4862 Section 5.5.3e:
1954 * "Note that the preferred lifetime of
1955 * the corresponding address is always
1956 * reset to the Preferred Lifetime in
1957 * the received Prefix Information
1958 * option, regardless of whether the
1959 * valid lifetime is also reset or
1960 * ignored."
1961 *
1962 * So if the preferred lifetime in
1963 * this advertisement is different
1964 * than what we have stored, but the
1965 * valid lifetime is invalid, just
1966 * reset prefered_lft.
1967 *
1968 * We must set the valid lifetime
1969 * to the stored lifetime since we'll
1970 * be updating the timestamp below,
1971 * else we'll set it back to the
1972 * minimum.
1973 */
1974 if (prefered_lft != ifp->prefered_lft) {
1975 valid_lft = stored_lft;
1976 update_lft = 1;
1977 }
1978 } else {
1979 valid_lft = MIN_VALID_LIFETIME;
1980 if (valid_lft < prefered_lft)
1981 prefered_lft = valid_lft;
1982 update_lft = 1;
1983 }
1984 }
1985
1986 if (update_lft) {
1987 ifp->valid_lft = valid_lft;
1988 ifp->prefered_lft = prefered_lft;
1989 ifp->tstamp = now;
1990 flags = ifp->flags;
1991 ifp->flags &= ~IFA_F_DEPRECATED;
1992 spin_unlock(&ifp->lock);
1993
1994 if (!(flags&IFA_F_TENTATIVE))
1995 ipv6_ifa_notify(0, ifp);
1996 } else
1997 spin_unlock(&ifp->lock);
1998
1999#ifdef CONFIG_IPV6_PRIVACY
2000 read_lock_bh(&in6_dev->lock);
2001 /* update all temporary addresses in the list */
2002 list_for_each_entry(ift, &in6_dev->tempaddr_list, tmp_list) {
2003 /*
2004 * When adjusting the lifetimes of an existing
2005 * temporary address, only lower the lifetimes.
2006 * Implementations must not increase the
2007 * lifetimes of an existing temporary address
2008 * when processing a Prefix Information Option.
2009 */
2010 if (ifp != ift->ifpub)
2011 continue;
2012
2013 spin_lock(&ift->lock);
2014 flags = ift->flags;
2015 if (ift->valid_lft > valid_lft &&
2016 ift->valid_lft - valid_lft > (jiffies - ift->tstamp) / HZ)
2017 ift->valid_lft = valid_lft + (jiffies - ift->tstamp) / HZ;
2018 if (ift->prefered_lft > prefered_lft &&
2019 ift->prefered_lft - prefered_lft > (jiffies - ift->tstamp) / HZ)
2020 ift->prefered_lft = prefered_lft + (jiffies - ift->tstamp) / HZ;
2021 spin_unlock(&ift->lock);
2022 if (!(flags&IFA_F_TENTATIVE))
2023 ipv6_ifa_notify(0, ift);
2024 }
2025
2026 if ((create || list_empty(&in6_dev->tempaddr_list)) && in6_dev->cnf.use_tempaddr > 0) {
2027 /*
2028 * When a new public address is created as described in [ADDRCONF],
2029 * also create a new temporary address. Also create a temporary
2030 * address if it's enabled but no temporary address currently exists.
2031 */
2032 read_unlock_bh(&in6_dev->lock);
2033 ipv6_create_tempaddr(ifp, NULL);
2034 } else {
2035 read_unlock_bh(&in6_dev->lock);
2036 }
2037#endif
2038 in6_ifa_put(ifp);
2039 addrconf_verify(0);
2040 }
2041 }
2042 inet6_prefix_notify(RTM_NEWPREFIX, in6_dev, pinfo);
2043 in6_dev_put(in6_dev);
2044}
2045
2046/*
2047 * Set destination address.
2048 * Special case for SIT interfaces where we create a new "virtual"
2049 * device.
2050 */
2051int addrconf_set_dstaddr(struct net *net, void __user *arg)
2052{
2053 struct in6_ifreq ireq;
2054 struct net_device *dev;
2055 int err = -EINVAL;
2056
2057 rtnl_lock();
2058
2059 err = -EFAULT;
2060 if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2061 goto err_exit;
2062
2063 dev = __dev_get_by_index(net, ireq.ifr6_ifindex);
2064
2065 err = -ENODEV;
2066 if (dev == NULL)
2067 goto err_exit;
2068
2069#if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2070 if (dev->type == ARPHRD_SIT) {
2071 const struct net_device_ops *ops = dev->netdev_ops;
2072 struct ifreq ifr;
2073 struct ip_tunnel_parm p;
2074
2075 err = -EADDRNOTAVAIL;
2076 if (!(ipv6_addr_type(&ireq.ifr6_addr) & IPV6_ADDR_COMPATv4))
2077 goto err_exit;
2078
2079 memset(&p, 0, sizeof(p));
2080 p.iph.daddr = ireq.ifr6_addr.s6_addr32[3];
2081 p.iph.saddr = 0;
2082 p.iph.version = 4;
2083 p.iph.ihl = 5;
2084 p.iph.protocol = IPPROTO_IPV6;
2085 p.iph.ttl = 64;
2086 ifr.ifr_ifru.ifru_data = (__force void __user *)&p;
2087
2088 if (ops->ndo_do_ioctl) {
2089 mm_segment_t oldfs = get_fs();
2090
2091 set_fs(KERNEL_DS);
2092 err = ops->ndo_do_ioctl(dev, &ifr, SIOCADDTUNNEL);
2093 set_fs(oldfs);
2094 } else
2095 err = -EOPNOTSUPP;
2096
2097 if (err == 0) {
2098 err = -ENOBUFS;
2099 dev = __dev_get_by_name(net, p.name);
2100 if (!dev)
2101 goto err_exit;
2102 err = dev_open(dev);
2103 }
2104 }
2105#endif
2106
2107err_exit:
2108 rtnl_unlock();
2109 return err;
2110}
2111
2112/*
2113 * Manual configuration of address on an interface
2114 */
2115static int inet6_addr_add(struct net *net, int ifindex, const struct in6_addr *pfx,
2116 unsigned int plen, __u8 ifa_flags, __u32 prefered_lft,
2117 __u32 valid_lft)
2118{
2119 struct inet6_ifaddr *ifp;
2120 struct inet6_dev *idev;
2121 struct net_device *dev;
2122 int scope;
2123 u32 flags;
2124 clock_t expires;
2125 unsigned long timeout;
2126
2127 ASSERT_RTNL();
2128
2129 if (plen > 128)
2130 return -EINVAL;
2131
2132 /* check the lifetime */
2133 if (!valid_lft || prefered_lft > valid_lft)
2134 return -EINVAL;
2135
2136 dev = __dev_get_by_index(net, ifindex);
2137 if (!dev)
2138 return -ENODEV;
2139
2140 idev = addrconf_add_dev(dev);
2141 if (IS_ERR(idev))
2142 return PTR_ERR(idev);
2143
2144 scope = ipv6_addr_scope(pfx);
2145
2146 timeout = addrconf_timeout_fixup(valid_lft, HZ);
2147 if (addrconf_finite_timeout(timeout)) {
2148 expires = jiffies_to_clock_t(timeout * HZ);
2149 valid_lft = timeout;
2150 flags = RTF_EXPIRES;
2151 } else {
2152 expires = 0;
2153 flags = 0;
2154 ifa_flags |= IFA_F_PERMANENT;
2155 }
2156
2157 timeout = addrconf_timeout_fixup(prefered_lft, HZ);
2158 if (addrconf_finite_timeout(timeout)) {
2159 if (timeout == 0)
2160 ifa_flags |= IFA_F_DEPRECATED;
2161 prefered_lft = timeout;
2162 }
2163
2164 ifp = ipv6_add_addr(idev, pfx, plen, scope, ifa_flags);
2165
2166 if (!IS_ERR(ifp)) {
2167 spin_lock_bh(&ifp->lock);
2168 ifp->valid_lft = valid_lft;
2169 ifp->prefered_lft = prefered_lft;
2170 ifp->tstamp = jiffies;
2171 spin_unlock_bh(&ifp->lock);
2172
2173 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, dev,
2174 expires, flags);
2175 /*
2176 * Note that section 3.1 of RFC 4429 indicates
2177 * that the Optimistic flag should not be set for
2178 * manually configured addresses
2179 */
2180 addrconf_dad_start(ifp, 0);
2181 in6_ifa_put(ifp);
2182 addrconf_verify(0);
2183 return 0;
2184 }
2185
2186 return PTR_ERR(ifp);
2187}
2188
2189static int inet6_addr_del(struct net *net, int ifindex, const struct in6_addr *pfx,
2190 unsigned int plen)
2191{
2192 struct inet6_ifaddr *ifp;
2193 struct inet6_dev *idev;
2194 struct net_device *dev;
2195
2196 if (plen > 128)
2197 return -EINVAL;
2198
2199 dev = __dev_get_by_index(net, ifindex);
2200 if (!dev)
2201 return -ENODEV;
2202
2203 if ((idev = __in6_dev_get(dev)) == NULL)
2204 return -ENXIO;
2205
2206 read_lock_bh(&idev->lock);
2207 list_for_each_entry(ifp, &idev->addr_list, if_list) {
2208 if (ifp->prefix_len == plen &&
2209 ipv6_addr_equal(pfx, &ifp->addr)) {
2210 in6_ifa_hold(ifp);
2211 read_unlock_bh(&idev->lock);
2212
2213 ipv6_del_addr(ifp);
2214
2215 /* If the last address is deleted administratively,
2216 disable IPv6 on this interface.
2217 */
2218 if (list_empty(&idev->addr_list))
2219 addrconf_ifdown(idev->dev, 1);
2220 return 0;
2221 }
2222 }
2223 read_unlock_bh(&idev->lock);
2224 return -EADDRNOTAVAIL;
2225}
2226
2227
2228int addrconf_add_ifaddr(struct net *net, void __user *arg)
2229{
2230 struct in6_ifreq ireq;
2231 int err;
2232
2233 if (!capable(CAP_NET_ADMIN))
2234 return -EPERM;
2235
2236 if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2237 return -EFAULT;
2238
2239 rtnl_lock();
2240 err = inet6_addr_add(net, ireq.ifr6_ifindex, &ireq.ifr6_addr,
2241 ireq.ifr6_prefixlen, IFA_F_PERMANENT,
2242 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
2243 rtnl_unlock();
2244 return err;
2245}
2246
2247int addrconf_del_ifaddr(struct net *net, void __user *arg)
2248{
2249 struct in6_ifreq ireq;
2250 int err;
2251
2252 if (!capable(CAP_NET_ADMIN))
2253 return -EPERM;
2254
2255 if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2256 return -EFAULT;
2257
2258 rtnl_lock();
2259 err = inet6_addr_del(net, ireq.ifr6_ifindex, &ireq.ifr6_addr,
2260 ireq.ifr6_prefixlen);
2261 rtnl_unlock();
2262 return err;
2263}
2264
2265static void add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
2266 int plen, int scope)
2267{
2268 struct inet6_ifaddr *ifp;
2269
2270 ifp = ipv6_add_addr(idev, addr, plen, scope, IFA_F_PERMANENT);
2271 if (!IS_ERR(ifp)) {
2272 spin_lock_bh(&ifp->lock);
2273 ifp->flags &= ~IFA_F_TENTATIVE;
2274 spin_unlock_bh(&ifp->lock);
2275 ipv6_ifa_notify(RTM_NEWADDR, ifp);
2276 in6_ifa_put(ifp);
2277 }
2278}
2279
2280#if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2281static void sit_add_v4_addrs(struct inet6_dev *idev)
2282{
2283 struct in6_addr addr;
2284 struct net_device *dev;
2285 struct net *net = dev_net(idev->dev);
2286 int scope;
2287
2288 ASSERT_RTNL();
2289
2290 memset(&addr, 0, sizeof(struct in6_addr));
2291 memcpy(&addr.s6_addr32[3], idev->dev->dev_addr, 4);
2292
2293 if (idev->dev->flags&IFF_POINTOPOINT) {
2294 addr.s6_addr32[0] = htonl(0xfe800000);
2295 scope = IFA_LINK;
2296 } else {
2297 scope = IPV6_ADDR_COMPATv4;
2298 }
2299
2300 if (addr.s6_addr32[3]) {
2301 add_addr(idev, &addr, 128, scope);
2302 return;
2303 }
2304
2305 for_each_netdev(net, dev) {
2306 struct in_device * in_dev = __in_dev_get_rtnl(dev);
2307 if (in_dev && (dev->flags & IFF_UP)) {
2308 struct in_ifaddr * ifa;
2309
2310 int flag = scope;
2311
2312 for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) {
2313 int plen;
2314
2315 addr.s6_addr32[3] = ifa->ifa_local;
2316
2317 if (ifa->ifa_scope == RT_SCOPE_LINK)
2318 continue;
2319 if (ifa->ifa_scope >= RT_SCOPE_HOST) {
2320 if (idev->dev->flags&IFF_POINTOPOINT)
2321 continue;
2322 flag |= IFA_HOST;
2323 }
2324 if (idev->dev->flags&IFF_POINTOPOINT)
2325 plen = 64;
2326 else
2327 plen = 96;
2328
2329 add_addr(idev, &addr, plen, flag);
2330 }
2331 }
2332 }
2333}
2334#endif
2335
2336static void init_loopback(struct net_device *dev)
2337{
2338 struct inet6_dev *idev;
2339
2340 /* ::1 */
2341
2342 ASSERT_RTNL();
2343
2344 if ((idev = ipv6_find_idev(dev)) == NULL) {
2345 printk(KERN_DEBUG "init loopback: add_dev failed\n");
2346 return;
2347 }
2348
2349 add_addr(idev, &in6addr_loopback, 128, IFA_HOST);
2350}
2351
2352static void addrconf_add_linklocal(struct inet6_dev *idev, const struct in6_addr *addr)
2353{
2354 struct inet6_ifaddr * ifp;
2355 u32 addr_flags = IFA_F_PERMANENT;
2356
2357#ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2358 if (idev->cnf.optimistic_dad &&
2359 !dev_net(idev->dev)->ipv6.devconf_all->forwarding)
2360 addr_flags |= IFA_F_OPTIMISTIC;
2361#endif
2362
2363
2364 ifp = ipv6_add_addr(idev, addr, 64, IFA_LINK, addr_flags);
2365 if (!IS_ERR(ifp)) {
2366 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, idev->dev, 0, 0);
2367 addrconf_dad_start(ifp, 0);
2368 in6_ifa_put(ifp);
2369 }
2370}
2371
2372static void addrconf_dev_config(struct net_device *dev)
2373{
2374 struct in6_addr addr;
2375 struct inet6_dev * idev;
2376
2377 ASSERT_RTNL();
2378
2379 if ((dev->type != ARPHRD_ETHER) &&
2380 (dev->type != ARPHRD_FDDI) &&
2381 (dev->type != ARPHRD_IEEE802_TR) &&
2382 (dev->type != ARPHRD_ARCNET) &&
2383 (dev->type != ARPHRD_INFINIBAND)) {
2384 /* Alas, we support only Ethernet autoconfiguration. */
2385 return;
2386 }
2387
2388 idev = addrconf_add_dev(dev);
2389 if (IS_ERR(idev))
2390 return;
2391
2392 memset(&addr, 0, sizeof(struct in6_addr));
2393 addr.s6_addr32[0] = htonl(0xFE800000);
2394
2395 if (ipv6_generate_eui64(addr.s6_addr + 8, dev) == 0)
2396 addrconf_add_linklocal(idev, &addr);
2397}
2398
2399#if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2400static void addrconf_sit_config(struct net_device *dev)
2401{
2402 struct inet6_dev *idev;
2403
2404 ASSERT_RTNL();
2405
2406 /*
2407 * Configure the tunnel with one of our IPv4
2408 * addresses... we should configure all of
2409 * our v4 addrs in the tunnel
2410 */
2411
2412 if ((idev = ipv6_find_idev(dev)) == NULL) {
2413 printk(KERN_DEBUG "init sit: add_dev failed\n");
2414 return;
2415 }
2416
2417 if (dev->priv_flags & IFF_ISATAP) {
2418 struct in6_addr addr;
2419
2420 ipv6_addr_set(&addr, htonl(0xFE800000), 0, 0, 0);
2421 addrconf_prefix_route(&addr, 64, dev, 0, 0);
2422 if (!ipv6_generate_eui64(addr.s6_addr + 8, dev))
2423 addrconf_add_linklocal(idev, &addr);
2424 return;
2425 }
2426
2427 sit_add_v4_addrs(idev);
2428
2429 if (dev->flags&IFF_POINTOPOINT) {
2430 addrconf_add_mroute(dev);
2431 addrconf_add_lroute(dev);
2432 } else
2433 sit_route_add(dev);
2434}
2435#endif
2436
2437#if defined(CONFIG_NET_IPGRE) || defined(CONFIG_NET_IPGRE_MODULE)
2438static void addrconf_gre_config(struct net_device *dev)
2439{
2440 struct inet6_dev *idev;
2441 struct in6_addr addr;
2442
2443 pr_info("ipv6: addrconf_gre_config(%s)\n", dev->name);
2444
2445 ASSERT_RTNL();
2446
2447 if ((idev = ipv6_find_idev(dev)) == NULL) {
2448 printk(KERN_DEBUG "init gre: add_dev failed\n");
2449 return;
2450 }
2451
2452 ipv6_addr_set(&addr, htonl(0xFE800000), 0, 0, 0);
2453 addrconf_prefix_route(&addr, 64, dev, 0, 0);
2454
2455 if (!ipv6_generate_eui64(addr.s6_addr + 8, dev))
2456 addrconf_add_linklocal(idev, &addr);
2457}
2458#endif
2459
2460static inline int
2461ipv6_inherit_linklocal(struct inet6_dev *idev, struct net_device *link_dev)
2462{
2463 struct in6_addr lladdr;
2464
2465 if (!ipv6_get_lladdr(link_dev, &lladdr, IFA_F_TENTATIVE)) {
2466 addrconf_add_linklocal(idev, &lladdr);
2467 return 0;
2468 }
2469 return -1;
2470}
2471
2472static void ip6_tnl_add_linklocal(struct inet6_dev *idev)
2473{
2474 struct net_device *link_dev;
2475 struct net *net = dev_net(idev->dev);
2476
2477 /* first try to inherit the link-local address from the link device */
2478 if (idev->dev->iflink &&
2479 (link_dev = __dev_get_by_index(net, idev->dev->iflink))) {
2480 if (!ipv6_inherit_linklocal(idev, link_dev))
2481 return;
2482 }
2483 /* then try to inherit it from any device */
2484 for_each_netdev(net, link_dev) {
2485 if (!ipv6_inherit_linklocal(idev, link_dev))
2486 return;
2487 }
2488 printk(KERN_DEBUG "init ip6-ip6: add_linklocal failed\n");
2489}
2490
2491/*
2492 * Autoconfigure tunnel with a link-local address so routing protocols,
2493 * DHCPv6, MLD etc. can be run over the virtual link
2494 */
2495
2496static void addrconf_ip6_tnl_config(struct net_device *dev)
2497{
2498 struct inet6_dev *idev;
2499
2500 ASSERT_RTNL();
2501
2502 idev = addrconf_add_dev(dev);
2503 if (IS_ERR(idev)) {
2504 printk(KERN_DEBUG "init ip6-ip6: add_dev failed\n");
2505 return;
2506 }
2507 ip6_tnl_add_linklocal(idev);
2508}
2509
2510static int addrconf_notify(struct notifier_block *this, unsigned long event,
2511 void * data)
2512{
2513 struct net_device *dev = (struct net_device *) data;
2514 struct inet6_dev *idev = __in6_dev_get(dev);
2515 int run_pending = 0;
2516 int err;
2517
2518 switch (event) {
2519 case NETDEV_REGISTER:
2520 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2521 idev = ipv6_add_dev(dev);
2522 if (!idev)
2523 return notifier_from_errno(-ENOMEM);
2524 }
2525 break;
2526
2527 case NETDEV_UP:
2528 case NETDEV_CHANGE:
2529 if (dev->flags & IFF_SLAVE)
2530 break;
2531
2532 if (event == NETDEV_UP) {
2533 if (!addrconf_qdisc_ok(dev)) {
2534 /* device is not ready yet. */
2535 printk(KERN_INFO
2536 "ADDRCONF(NETDEV_UP): %s: "
2537 "link is not ready\n",
2538 dev->name);
2539 break;
2540 }
2541
2542 if (!idev && dev->mtu >= IPV6_MIN_MTU)
2543 idev = ipv6_add_dev(dev);
2544
2545 if (idev) {
2546 idev->if_flags |= IF_READY;
2547 run_pending = 1;
2548 }
2549 } else {
2550 if (!addrconf_qdisc_ok(dev)) {
2551 /* device is still not ready. */
2552 break;
2553 }
2554
2555 if (idev) {
2556 if (idev->if_flags & IF_READY)
2557 /* device is already configured. */
2558 break;
2559 idev->if_flags |= IF_READY;
2560 }
2561
2562 printk(KERN_INFO
2563 "ADDRCONF(NETDEV_CHANGE): %s: "
2564 "link becomes ready\n",
2565 dev->name);
2566
2567 run_pending = 1;
2568 }
2569
2570 switch (dev->type) {
2571#if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2572 case ARPHRD_SIT:
2573 addrconf_sit_config(dev);
2574 break;
2575#endif
2576#if defined(CONFIG_NET_IPGRE) || defined(CONFIG_NET_IPGRE_MODULE)
2577 case ARPHRD_IPGRE:
2578 addrconf_gre_config(dev);
2579 break;
2580#endif
2581 case ARPHRD_TUNNEL6:
2582 addrconf_ip6_tnl_config(dev);
2583 break;
2584 case ARPHRD_LOOPBACK:
2585 init_loopback(dev);
2586 break;
2587
2588 default:
2589 addrconf_dev_config(dev);
2590 break;
2591 }
2592
2593 if (idev) {
2594 if (run_pending)
2595 addrconf_dad_run(idev);
2596
2597 /*
2598 * If the MTU changed during the interface down,
2599 * when the interface up, the changed MTU must be
2600 * reflected in the idev as well as routers.
2601 */
2602 if (idev->cnf.mtu6 != dev->mtu &&
2603 dev->mtu >= IPV6_MIN_MTU) {
2604 rt6_mtu_change(dev, dev->mtu);
2605 idev->cnf.mtu6 = dev->mtu;
2606 }
2607 idev->tstamp = jiffies;
2608 inet6_ifinfo_notify(RTM_NEWLINK, idev);
2609
2610 /*
2611 * If the changed mtu during down is lower than
2612 * IPV6_MIN_MTU stop IPv6 on this interface.
2613 */
2614 if (dev->mtu < IPV6_MIN_MTU)
2615 addrconf_ifdown(dev, 1);
2616 }
2617 break;
2618
2619 case NETDEV_CHANGEMTU:
2620 if (idev && dev->mtu >= IPV6_MIN_MTU) {
2621 rt6_mtu_change(dev, dev->mtu);
2622 idev->cnf.mtu6 = dev->mtu;
2623 break;
2624 }
2625
2626 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2627 idev = ipv6_add_dev(dev);
2628 if (idev)
2629 break;
2630 }
2631
2632 /*
2633 * MTU falled under IPV6_MIN_MTU.
2634 * Stop IPv6 on this interface.
2635 */
2636
2637 case NETDEV_DOWN:
2638 case NETDEV_UNREGISTER:
2639 /*
2640 * Remove all addresses from this interface.
2641 */
2642 addrconf_ifdown(dev, event != NETDEV_DOWN);
2643 break;
2644
2645 case NETDEV_CHANGENAME:
2646 if (idev) {
2647 snmp6_unregister_dev(idev);
2648 addrconf_sysctl_unregister(idev);
2649 addrconf_sysctl_register(idev);
2650 err = snmp6_register_dev(idev);
2651 if (err)
2652 return notifier_from_errno(err);
2653 }
2654 break;
2655
2656 case NETDEV_PRE_TYPE_CHANGE:
2657 case NETDEV_POST_TYPE_CHANGE:
2658 addrconf_type_change(dev, event);
2659 break;
2660 }
2661
2662 return NOTIFY_OK;
2663}
2664
2665/*
2666 * addrconf module should be notified of a device going up
2667 */
2668static struct notifier_block ipv6_dev_notf = {
2669 .notifier_call = addrconf_notify,
2670};
2671
2672static void addrconf_type_change(struct net_device *dev, unsigned long event)
2673{
2674 struct inet6_dev *idev;
2675 ASSERT_RTNL();
2676
2677 idev = __in6_dev_get(dev);
2678
2679 if (event == NETDEV_POST_TYPE_CHANGE)
2680 ipv6_mc_remap(idev);
2681 else if (event == NETDEV_PRE_TYPE_CHANGE)
2682 ipv6_mc_unmap(idev);
2683}
2684
2685static int addrconf_ifdown(struct net_device *dev, int how)
2686{
2687 struct net *net = dev_net(dev);
2688 struct inet6_dev *idev;
2689 struct inet6_ifaddr *ifa;
2690 int state, i;
2691
2692 ASSERT_RTNL();
2693
2694 rt6_ifdown(net, dev);
2695 neigh_ifdown(&nd_tbl, dev);
2696
2697 idev = __in6_dev_get(dev);
2698 if (idev == NULL)
2699 return -ENODEV;
2700
2701 /*
2702 * Step 1: remove reference to ipv6 device from parent device.
2703 * Do not dev_put!
2704 */
2705 if (how) {
2706 idev->dead = 1;
2707
2708 /* protected by rtnl_lock */
2709 rcu_assign_pointer(dev->ip6_ptr, NULL);
2710
2711 /* Step 1.5: remove snmp6 entry */
2712 snmp6_unregister_dev(idev);
2713
2714 }
2715
2716 /* Step 2: clear hash table */
2717 for (i = 0; i < IN6_ADDR_HSIZE; i++) {
2718 struct hlist_head *h = &inet6_addr_lst[i];
2719 struct hlist_node *n;
2720
2721 spin_lock_bh(&addrconf_hash_lock);
2722 restart:
2723 hlist_for_each_entry_rcu(ifa, n, h, addr_lst) {
2724 if (ifa->idev == idev) {
2725 hlist_del_init_rcu(&ifa->addr_lst);
2726 addrconf_del_timer(ifa);
2727 goto restart;
2728 }
2729 }
2730 spin_unlock_bh(&addrconf_hash_lock);
2731 }
2732
2733 write_lock_bh(&idev->lock);
2734
2735 /* Step 2: clear flags for stateless addrconf */
2736 if (!how)
2737 idev->if_flags &= ~(IF_RS_SENT|IF_RA_RCVD|IF_READY);
2738
2739#ifdef CONFIG_IPV6_PRIVACY
2740 if (how && del_timer(&idev->regen_timer))
2741 in6_dev_put(idev);
2742
2743 /* Step 3: clear tempaddr list */
2744 while (!list_empty(&idev->tempaddr_list)) {
2745 ifa = list_first_entry(&idev->tempaddr_list,
2746 struct inet6_ifaddr, tmp_list);
2747 list_del(&ifa->tmp_list);
2748 write_unlock_bh(&idev->lock);
2749 spin_lock_bh(&ifa->lock);
2750
2751 if (ifa->ifpub) {
2752 in6_ifa_put(ifa->ifpub);
2753 ifa->ifpub = NULL;
2754 }
2755 spin_unlock_bh(&ifa->lock);
2756 in6_ifa_put(ifa);
2757 write_lock_bh(&idev->lock);
2758 }
2759#endif
2760
2761 while (!list_empty(&idev->addr_list)) {
2762 ifa = list_first_entry(&idev->addr_list,
2763 struct inet6_ifaddr, if_list);
2764 addrconf_del_timer(ifa);
2765
2766 list_del(&ifa->if_list);
2767
2768 write_unlock_bh(&idev->lock);
2769
2770 spin_lock_bh(&ifa->state_lock);
2771 state = ifa->state;
2772 ifa->state = INET6_IFADDR_STATE_DEAD;
2773 spin_unlock_bh(&ifa->state_lock);
2774
2775 if (state != INET6_IFADDR_STATE_DEAD) {
2776 __ipv6_ifa_notify(RTM_DELADDR, ifa);
2777 atomic_notifier_call_chain(&inet6addr_chain, NETDEV_DOWN, ifa);
2778 }
2779 in6_ifa_put(ifa);
2780
2781 write_lock_bh(&idev->lock);
2782 }
2783
2784 write_unlock_bh(&idev->lock);
2785
2786 /* Step 5: Discard multicast list */
2787 if (how)
2788 ipv6_mc_destroy_dev(idev);
2789 else
2790 ipv6_mc_down(idev);
2791
2792 idev->tstamp = jiffies;
2793
2794 /* Last: Shot the device (if unregistered) */
2795 if (how) {
2796 addrconf_sysctl_unregister(idev);
2797 neigh_parms_release(&nd_tbl, idev->nd_parms);
2798 neigh_ifdown(&nd_tbl, dev);
2799 in6_dev_put(idev);
2800 }
2801 return 0;
2802}
2803
2804static void addrconf_rs_timer(unsigned long data)
2805{
2806 struct inet6_ifaddr *ifp = (struct inet6_ifaddr *) data;
2807 struct inet6_dev *idev = ifp->idev;
2808
2809 read_lock(&idev->lock);
2810 if (idev->dead || !(idev->if_flags & IF_READY))
2811 goto out;
2812
2813 if (idev->cnf.forwarding)
2814 goto out;
2815
2816 /* Announcement received after solicitation was sent */
2817 if (idev->if_flags & IF_RA_RCVD)
2818 goto out;
2819
2820 spin_lock(&ifp->lock);
2821 if (ifp->probes++ < idev->cnf.rtr_solicits) {
2822 /* The wait after the last probe can be shorter */
2823 addrconf_mod_timer(ifp, AC_RS,
2824 (ifp->probes == idev->cnf.rtr_solicits) ?
2825 idev->cnf.rtr_solicit_delay :
2826 idev->cnf.rtr_solicit_interval);
2827 spin_unlock(&ifp->lock);
2828
2829 ndisc_send_rs(idev->dev, &ifp->addr, &in6addr_linklocal_allrouters);
2830 } else {
2831 spin_unlock(&ifp->lock);
2832 /*
2833 * Note: we do not support deprecated "all on-link"
2834 * assumption any longer.
2835 */
2836 printk(KERN_DEBUG "%s: no IPv6 routers present\n",
2837 idev->dev->name);
2838 }
2839
2840out:
2841 read_unlock(&idev->lock);
2842 in6_ifa_put(ifp);
2843}
2844
2845/*
2846 * Duplicate Address Detection
2847 */
2848static void addrconf_dad_kick(struct inet6_ifaddr *ifp)
2849{
2850 unsigned long rand_num;
2851 struct inet6_dev *idev = ifp->idev;
2852
2853 if (ifp->flags & IFA_F_OPTIMISTIC)
2854 rand_num = 0;
2855 else
2856 rand_num = net_random() % (idev->cnf.rtr_solicit_delay ? : 1);
2857
2858 ifp->probes = idev->cnf.dad_transmits;
2859 addrconf_mod_timer(ifp, AC_DAD, rand_num);
2860}
2861
2862static void addrconf_dad_start(struct inet6_ifaddr *ifp, u32 flags)
2863{
2864 struct inet6_dev *idev = ifp->idev;
2865 struct net_device *dev = idev->dev;
2866
2867 addrconf_join_solict(dev, &ifp->addr);
2868
2869 net_srandom(ifp->addr.s6_addr32[3]);
2870
2871 read_lock_bh(&idev->lock);
2872 spin_lock(&ifp->lock);
2873 if (ifp->state == INET6_IFADDR_STATE_DEAD)
2874 goto out;
2875
2876 if (dev->flags&(IFF_NOARP|IFF_LOOPBACK) ||
2877 idev->cnf.accept_dad < 1 ||
2878 !(ifp->flags&IFA_F_TENTATIVE) ||
2879 ifp->flags & IFA_F_NODAD) {
2880 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
2881 spin_unlock(&ifp->lock);
2882 read_unlock_bh(&idev->lock);
2883
2884 addrconf_dad_completed(ifp);
2885 return;
2886 }
2887
2888 if (!(idev->if_flags & IF_READY)) {
2889 spin_unlock(&ifp->lock);
2890 read_unlock_bh(&idev->lock);
2891 /*
2892 * If the device is not ready:
2893 * - keep it tentative if it is a permanent address.
2894 * - otherwise, kill it.
2895 */
2896 in6_ifa_hold(ifp);
2897 addrconf_dad_stop(ifp, 0);
2898 return;
2899 }
2900
2901 /*
2902 * Optimistic nodes can start receiving
2903 * Frames right away
2904 */
2905 if (ifp->flags & IFA_F_OPTIMISTIC)
2906 ip6_ins_rt(ifp->rt);
2907
2908 addrconf_dad_kick(ifp);
2909out:
2910 spin_unlock(&ifp->lock);
2911 read_unlock_bh(&idev->lock);
2912}
2913
2914static void addrconf_dad_timer(unsigned long data)
2915{
2916 struct inet6_ifaddr *ifp = (struct inet6_ifaddr *) data;
2917 struct inet6_dev *idev = ifp->idev;
2918 struct in6_addr mcaddr;
2919
2920 if (!ifp->probes && addrconf_dad_end(ifp))
2921 goto out;
2922
2923 read_lock(&idev->lock);
2924 if (idev->dead || !(idev->if_flags & IF_READY)) {
2925 read_unlock(&idev->lock);
2926 goto out;
2927 }
2928
2929 spin_lock(&ifp->lock);
2930 if (ifp->state == INET6_IFADDR_STATE_DEAD) {
2931 spin_unlock(&ifp->lock);
2932 read_unlock(&idev->lock);
2933 goto out;
2934 }
2935
2936 if (ifp->probes == 0) {
2937 /*
2938 * DAD was successful
2939 */
2940
2941 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
2942 spin_unlock(&ifp->lock);
2943 read_unlock(&idev->lock);
2944
2945 addrconf_dad_completed(ifp);
2946
2947 goto out;
2948 }
2949
2950 ifp->probes--;
2951 addrconf_mod_timer(ifp, AC_DAD, ifp->idev->nd_parms->retrans_time);
2952 spin_unlock(&ifp->lock);
2953 read_unlock(&idev->lock);
2954
2955 /* send a neighbour solicitation for our addr */
2956 addrconf_addr_solict_mult(&ifp->addr, &mcaddr);
2957 ndisc_send_ns(ifp->idev->dev, NULL, &ifp->addr, &mcaddr, &in6addr_any);
2958out:
2959 in6_ifa_put(ifp);
2960}
2961
2962static void addrconf_dad_completed(struct inet6_ifaddr *ifp)
2963{
2964 struct net_device *dev = ifp->idev->dev;
2965
2966 /*
2967 * Configure the address for reception. Now it is valid.
2968 */
2969
2970 ipv6_ifa_notify(RTM_NEWADDR, ifp);
2971
2972 /* If added prefix is link local and forwarding is off,
2973 start sending router solicitations.
2974 */
2975
2976 if ((ifp->idev->cnf.forwarding == 0 ||
2977 ifp->idev->cnf.forwarding == 2) &&
2978 ifp->idev->cnf.rtr_solicits > 0 &&
2979 (dev->flags&IFF_LOOPBACK) == 0 &&
2980 (ipv6_addr_type(&ifp->addr) & IPV6_ADDR_LINKLOCAL)) {
2981 /*
2982 * If a host as already performed a random delay
2983 * [...] as part of DAD [...] there is no need
2984 * to delay again before sending the first RS
2985 */
2986 ndisc_send_rs(ifp->idev->dev, &ifp->addr, &in6addr_linklocal_allrouters);
2987
2988 spin_lock_bh(&ifp->lock);
2989 ifp->probes = 1;
2990 ifp->idev->if_flags |= IF_RS_SENT;
2991 addrconf_mod_timer(ifp, AC_RS, ifp->idev->cnf.rtr_solicit_interval);
2992 spin_unlock_bh(&ifp->lock);
2993 }
2994}
2995
2996static void addrconf_dad_run(struct inet6_dev *idev)
2997{
2998 struct inet6_ifaddr *ifp;
2999
3000 read_lock_bh(&idev->lock);
3001 list_for_each_entry(ifp, &idev->addr_list, if_list) {
3002 spin_lock(&ifp->lock);
3003 if (ifp->flags & IFA_F_TENTATIVE &&
3004 ifp->state == INET6_IFADDR_STATE_DAD)
3005 addrconf_dad_kick(ifp);
3006 spin_unlock(&ifp->lock);
3007 }
3008 read_unlock_bh(&idev->lock);
3009}
3010
3011#ifdef CONFIG_PROC_FS
3012struct if6_iter_state {
3013 struct seq_net_private p;
3014 int bucket;
3015};
3016
3017static struct inet6_ifaddr *if6_get_first(struct seq_file *seq)
3018{
3019 struct inet6_ifaddr *ifa = NULL;
3020 struct if6_iter_state *state = seq->private;
3021 struct net *net = seq_file_net(seq);
3022
3023 for (state->bucket = 0; state->bucket < IN6_ADDR_HSIZE; ++state->bucket) {
3024 struct hlist_node *n;
3025 hlist_for_each_entry_rcu_bh(ifa, n, &inet6_addr_lst[state->bucket],
3026 addr_lst)
3027 if (net_eq(dev_net(ifa->idev->dev), net))
3028 return ifa;
3029 }
3030 return NULL;
3031}
3032
3033static struct inet6_ifaddr *if6_get_next(struct seq_file *seq,
3034 struct inet6_ifaddr *ifa)
3035{
3036 struct if6_iter_state *state = seq->private;
3037 struct net *net = seq_file_net(seq);
3038 struct hlist_node *n = &ifa->addr_lst;
3039
3040 hlist_for_each_entry_continue_rcu_bh(ifa, n, addr_lst)
3041 if (net_eq(dev_net(ifa->idev->dev), net))
3042 return ifa;
3043
3044 while (++state->bucket < IN6_ADDR_HSIZE) {
3045 hlist_for_each_entry_rcu_bh(ifa, n,
3046 &inet6_addr_lst[state->bucket], addr_lst) {
3047 if (net_eq(dev_net(ifa->idev->dev), net))
3048 return ifa;
3049 }
3050 }
3051
3052 return NULL;
3053}
3054
3055static struct inet6_ifaddr *if6_get_idx(struct seq_file *seq, loff_t pos)
3056{
3057 struct inet6_ifaddr *ifa = if6_get_first(seq);
3058
3059 if (ifa)
3060 while (pos && (ifa = if6_get_next(seq, ifa)) != NULL)
3061 --pos;
3062 return pos ? NULL : ifa;
3063}
3064
3065static void *if6_seq_start(struct seq_file *seq, loff_t *pos)
3066 __acquires(rcu_bh)
3067{
3068 rcu_read_lock_bh();
3069 return if6_get_idx(seq, *pos);
3070}
3071
3072static void *if6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
3073{
3074 struct inet6_ifaddr *ifa;
3075
3076 ifa = if6_get_next(seq, v);
3077 ++*pos;
3078 return ifa;
3079}
3080
3081static void if6_seq_stop(struct seq_file *seq, void *v)
3082 __releases(rcu_bh)
3083{
3084 rcu_read_unlock_bh();
3085}
3086
3087static int if6_seq_show(struct seq_file *seq, void *v)
3088{
3089 struct inet6_ifaddr *ifp = (struct inet6_ifaddr *)v;
3090 seq_printf(seq, "%pi6 %02x %02x %02x %02x %8s\n",
3091 &ifp->addr,
3092 ifp->idev->dev->ifindex,
3093 ifp->prefix_len,
3094 ifp->scope,
3095 ifp->flags,
3096 ifp->idev->dev->name);
3097 return 0;
3098}
3099
3100static const struct seq_operations if6_seq_ops = {
3101 .start = if6_seq_start,
3102 .next = if6_seq_next,
3103 .show = if6_seq_show,
3104 .stop = if6_seq_stop,
3105};
3106
3107static int if6_seq_open(struct inode *inode, struct file *file)
3108{
3109 return seq_open_net(inode, file, &if6_seq_ops,
3110 sizeof(struct if6_iter_state));
3111}
3112
3113static const struct file_operations if6_fops = {
3114 .owner = THIS_MODULE,
3115 .open = if6_seq_open,
3116 .read = seq_read,
3117 .llseek = seq_lseek,
3118 .release = seq_release_net,
3119};
3120
3121static int __net_init if6_proc_net_init(struct net *net)
3122{
3123 if (!proc_net_fops_create(net, "if_inet6", S_IRUGO, &if6_fops))
3124 return -ENOMEM;
3125 return 0;
3126}
3127
3128static void __net_exit if6_proc_net_exit(struct net *net)
3129{
3130 proc_net_remove(net, "if_inet6");
3131}
3132
3133static struct pernet_operations if6_proc_net_ops = {
3134 .init = if6_proc_net_init,
3135 .exit = if6_proc_net_exit,
3136};
3137
3138int __init if6_proc_init(void)
3139{
3140 return register_pernet_subsys(&if6_proc_net_ops);
3141}
3142
3143void if6_proc_exit(void)
3144{
3145 unregister_pernet_subsys(&if6_proc_net_ops);
3146}
3147#endif /* CONFIG_PROC_FS */
3148
3149#if defined(CONFIG_IPV6_MIP6) || defined(CONFIG_IPV6_MIP6_MODULE)
3150/* Check if address is a home address configured on any interface. */
3151int ipv6_chk_home_addr(struct net *net, const struct in6_addr *addr)
3152{
3153 int ret = 0;
3154 struct inet6_ifaddr *ifp = NULL;
3155 struct hlist_node *n;
3156 unsigned int hash = ipv6_addr_hash(addr);
3157
3158 rcu_read_lock_bh();
3159 hlist_for_each_entry_rcu_bh(ifp, n, &inet6_addr_lst[hash], addr_lst) {
3160 if (!net_eq(dev_net(ifp->idev->dev), net))
3161 continue;
3162 if (ipv6_addr_equal(&ifp->addr, addr) &&
3163 (ifp->flags & IFA_F_HOMEADDRESS)) {
3164 ret = 1;
3165 break;
3166 }
3167 }
3168 rcu_read_unlock_bh();
3169 return ret;
3170}
3171#endif
3172
3173/*
3174 * Periodic address status verification
3175 */
3176
3177static void addrconf_verify(unsigned long foo)
3178{
3179 unsigned long now, next, next_sec, next_sched;
3180 struct inet6_ifaddr *ifp;
3181 struct hlist_node *node;
3182 int i;
3183
3184 rcu_read_lock_bh();
3185 spin_lock(&addrconf_verify_lock);
3186 now = jiffies;
3187 next = round_jiffies_up(now + ADDR_CHECK_FREQUENCY);
3188
3189 del_timer(&addr_chk_timer);
3190
3191 for (i = 0; i < IN6_ADDR_HSIZE; i++) {
3192restart:
3193 hlist_for_each_entry_rcu_bh(ifp, node,
3194 &inet6_addr_lst[i], addr_lst) {
3195 unsigned long age;
3196
3197 if (ifp->flags & IFA_F_PERMANENT)
3198 continue;
3199
3200 spin_lock(&ifp->lock);
3201 /* We try to batch several events at once. */
3202 age = (now - ifp->tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
3203
3204 if (ifp->valid_lft != INFINITY_LIFE_TIME &&
3205 age >= ifp->valid_lft) {
3206 spin_unlock(&ifp->lock);
3207 in6_ifa_hold(ifp);
3208 ipv6_del_addr(ifp);
3209 goto restart;
3210 } else if (ifp->prefered_lft == INFINITY_LIFE_TIME) {
3211 spin_unlock(&ifp->lock);
3212 continue;
3213 } else if (age >= ifp->prefered_lft) {
3214 /* jiffies - ifp->tstamp > age >= ifp->prefered_lft */
3215 int deprecate = 0;
3216
3217 if (!(ifp->flags&IFA_F_DEPRECATED)) {
3218 deprecate = 1;
3219 ifp->flags |= IFA_F_DEPRECATED;
3220 }
3221
3222 if (time_before(ifp->tstamp + ifp->valid_lft * HZ, next))
3223 next = ifp->tstamp + ifp->valid_lft * HZ;
3224
3225 spin_unlock(&ifp->lock);
3226
3227 if (deprecate) {
3228 in6_ifa_hold(ifp);
3229
3230 ipv6_ifa_notify(0, ifp);
3231 in6_ifa_put(ifp);
3232 goto restart;
3233 }
3234#ifdef CONFIG_IPV6_PRIVACY
3235 } else if ((ifp->flags&IFA_F_TEMPORARY) &&
3236 !(ifp->flags&IFA_F_TENTATIVE)) {
3237 unsigned long regen_advance = ifp->idev->cnf.regen_max_retry *
3238 ifp->idev->cnf.dad_transmits *
3239 ifp->idev->nd_parms->retrans_time / HZ;
3240
3241 if (age >= ifp->prefered_lft - regen_advance) {
3242 struct inet6_ifaddr *ifpub = ifp->ifpub;
3243 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3244 next = ifp->tstamp + ifp->prefered_lft * HZ;
3245 if (!ifp->regen_count && ifpub) {
3246 ifp->regen_count++;
3247 in6_ifa_hold(ifp);
3248 in6_ifa_hold(ifpub);
3249 spin_unlock(&ifp->lock);
3250
3251 spin_lock(&ifpub->lock);
3252 ifpub->regen_count = 0;
3253 spin_unlock(&ifpub->lock);
3254 ipv6_create_tempaddr(ifpub, ifp);
3255 in6_ifa_put(ifpub);
3256 in6_ifa_put(ifp);
3257 goto restart;
3258 }
3259 } else if (time_before(ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ, next))
3260 next = ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ;
3261 spin_unlock(&ifp->lock);
3262#endif
3263 } else {
3264 /* ifp->prefered_lft <= ifp->valid_lft */
3265 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3266 next = ifp->tstamp + ifp->prefered_lft * HZ;
3267 spin_unlock(&ifp->lock);
3268 }
3269 }
3270 }
3271
3272 next_sec = round_jiffies_up(next);
3273 next_sched = next;
3274
3275 /* If rounded timeout is accurate enough, accept it. */
3276 if (time_before(next_sec, next + ADDRCONF_TIMER_FUZZ))
3277 next_sched = next_sec;
3278
3279 /* And minimum interval is ADDRCONF_TIMER_FUZZ_MAX. */
3280 if (time_before(next_sched, jiffies + ADDRCONF_TIMER_FUZZ_MAX))
3281 next_sched = jiffies + ADDRCONF_TIMER_FUZZ_MAX;
3282
3283 ADBG((KERN_DEBUG "now = %lu, schedule = %lu, rounded schedule = %lu => %lu\n",
3284 now, next, next_sec, next_sched));
3285
3286 addr_chk_timer.expires = next_sched;
3287 add_timer(&addr_chk_timer);
3288 spin_unlock(&addrconf_verify_lock);
3289 rcu_read_unlock_bh();
3290}
3291
3292static struct in6_addr *extract_addr(struct nlattr *addr, struct nlattr *local)
3293{
3294 struct in6_addr *pfx = NULL;
3295
3296 if (addr)
3297 pfx = nla_data(addr);
3298
3299 if (local) {
3300 if (pfx && nla_memcmp(local, pfx, sizeof(*pfx)))
3301 pfx = NULL;
3302 else
3303 pfx = nla_data(local);
3304 }
3305
3306 return pfx;
3307}
3308
3309static const struct nla_policy ifa_ipv6_policy[IFA_MAX+1] = {
3310 [IFA_ADDRESS] = { .len = sizeof(struct in6_addr) },
3311 [IFA_LOCAL] = { .len = sizeof(struct in6_addr) },
3312 [IFA_CACHEINFO] = { .len = sizeof(struct ifa_cacheinfo) },
3313};
3314
3315static int
3316inet6_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
3317{
3318 struct net *net = sock_net(skb->sk);
3319 struct ifaddrmsg *ifm;
3320 struct nlattr *tb[IFA_MAX+1];
3321 struct in6_addr *pfx;
3322 int err;
3323
3324 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3325 if (err < 0)
3326 return err;
3327
3328 ifm = nlmsg_data(nlh);
3329 pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3330 if (pfx == NULL)
3331 return -EINVAL;
3332
3333 return inet6_addr_del(net, ifm->ifa_index, pfx, ifm->ifa_prefixlen);
3334}
3335
3336static int inet6_addr_modify(struct inet6_ifaddr *ifp, u8 ifa_flags,
3337 u32 prefered_lft, u32 valid_lft)
3338{
3339 u32 flags;
3340 clock_t expires;
3341 unsigned long timeout;
3342
3343 if (!valid_lft || (prefered_lft > valid_lft))
3344 return -EINVAL;
3345
3346 timeout = addrconf_timeout_fixup(valid_lft, HZ);
3347 if (addrconf_finite_timeout(timeout)) {
3348 expires = jiffies_to_clock_t(timeout * HZ);
3349 valid_lft = timeout;
3350 flags = RTF_EXPIRES;
3351 } else {
3352 expires = 0;
3353 flags = 0;
3354 ifa_flags |= IFA_F_PERMANENT;
3355 }
3356
3357 timeout = addrconf_timeout_fixup(prefered_lft, HZ);
3358 if (addrconf_finite_timeout(timeout)) {
3359 if (timeout == 0)
3360 ifa_flags |= IFA_F_DEPRECATED;
3361 prefered_lft = timeout;
3362 }
3363
3364 spin_lock_bh(&ifp->lock);
3365 ifp->flags = (ifp->flags & ~(IFA_F_DEPRECATED | IFA_F_PERMANENT | IFA_F_NODAD | IFA_F_HOMEADDRESS)) | ifa_flags;
3366 ifp->tstamp = jiffies;
3367 ifp->valid_lft = valid_lft;
3368 ifp->prefered_lft = prefered_lft;
3369
3370 spin_unlock_bh(&ifp->lock);
3371 if (!(ifp->flags&IFA_F_TENTATIVE))
3372 ipv6_ifa_notify(0, ifp);
3373
3374 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, ifp->idev->dev,
3375 expires, flags);
3376 addrconf_verify(0);
3377
3378 return 0;
3379}
3380
3381static int
3382inet6_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
3383{
3384 struct net *net = sock_net(skb->sk);
3385 struct ifaddrmsg *ifm;
3386 struct nlattr *tb[IFA_MAX+1];
3387 struct in6_addr *pfx;
3388 struct inet6_ifaddr *ifa;
3389 struct net_device *dev;
3390 u32 valid_lft = INFINITY_LIFE_TIME, preferred_lft = INFINITY_LIFE_TIME;
3391 u8 ifa_flags;
3392 int err;
3393
3394 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3395 if (err < 0)
3396 return err;
3397
3398 ifm = nlmsg_data(nlh);
3399 pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3400 if (pfx == NULL)
3401 return -EINVAL;
3402
3403 if (tb[IFA_CACHEINFO]) {
3404 struct ifa_cacheinfo *ci;
3405
3406 ci = nla_data(tb[IFA_CACHEINFO]);
3407 valid_lft = ci->ifa_valid;
3408 preferred_lft = ci->ifa_prefered;
3409 } else {
3410 preferred_lft = INFINITY_LIFE_TIME;
3411 valid_lft = INFINITY_LIFE_TIME;
3412 }
3413
3414 dev = __dev_get_by_index(net, ifm->ifa_index);
3415 if (dev == NULL)
3416 return -ENODEV;
3417
3418 /* We ignore other flags so far. */
3419 ifa_flags = ifm->ifa_flags & (IFA_F_NODAD | IFA_F_HOMEADDRESS);
3420
3421 ifa = ipv6_get_ifaddr(net, pfx, dev, 1);
3422 if (ifa == NULL) {
3423 /*
3424 * It would be best to check for !NLM_F_CREATE here but
3425 * userspace alreay relies on not having to provide this.
3426 */
3427 return inet6_addr_add(net, ifm->ifa_index, pfx,
3428 ifm->ifa_prefixlen, ifa_flags,
3429 preferred_lft, valid_lft);
3430 }
3431
3432 if (nlh->nlmsg_flags & NLM_F_EXCL ||
3433 !(nlh->nlmsg_flags & NLM_F_REPLACE))
3434 err = -EEXIST;
3435 else
3436 err = inet6_addr_modify(ifa, ifa_flags, preferred_lft, valid_lft);
3437
3438 in6_ifa_put(ifa);
3439
3440 return err;
3441}
3442
3443static void put_ifaddrmsg(struct nlmsghdr *nlh, u8 prefixlen, u8 flags,
3444 u8 scope, int ifindex)
3445{
3446 struct ifaddrmsg *ifm;
3447
3448 ifm = nlmsg_data(nlh);
3449 ifm->ifa_family = AF_INET6;
3450 ifm->ifa_prefixlen = prefixlen;
3451 ifm->ifa_flags = flags;
3452 ifm->ifa_scope = scope;
3453 ifm->ifa_index = ifindex;
3454}
3455
3456static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
3457 unsigned long tstamp, u32 preferred, u32 valid)
3458{
3459 struct ifa_cacheinfo ci;
3460
3461 ci.cstamp = cstamp_delta(cstamp);
3462 ci.tstamp = cstamp_delta(tstamp);
3463 ci.ifa_prefered = preferred;
3464 ci.ifa_valid = valid;
3465
3466 return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
3467}
3468
3469static inline int rt_scope(int ifa_scope)
3470{
3471 if (ifa_scope & IFA_HOST)
3472 return RT_SCOPE_HOST;
3473 else if (ifa_scope & IFA_LINK)
3474 return RT_SCOPE_LINK;
3475 else if (ifa_scope & IFA_SITE)
3476 return RT_SCOPE_SITE;
3477 else
3478 return RT_SCOPE_UNIVERSE;
3479}
3480
3481static inline int inet6_ifaddr_msgsize(void)
3482{
3483 return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
3484 + nla_total_size(16) /* IFA_ADDRESS */
3485 + nla_total_size(sizeof(struct ifa_cacheinfo));
3486}
3487
3488static int inet6_fill_ifaddr(struct sk_buff *skb, struct inet6_ifaddr *ifa,
3489 u32 pid, u32 seq, int event, unsigned int flags)
3490{
3491 struct nlmsghdr *nlh;
3492 u32 preferred, valid;
3493
3494 nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3495 if (nlh == NULL)
3496 return -EMSGSIZE;
3497
3498 put_ifaddrmsg(nlh, ifa->prefix_len, ifa->flags, rt_scope(ifa->scope),
3499 ifa->idev->dev->ifindex);
3500
3501 if (!(ifa->flags&IFA_F_PERMANENT)) {
3502 preferred = ifa->prefered_lft;
3503 valid = ifa->valid_lft;
3504 if (preferred != INFINITY_LIFE_TIME) {
3505 long tval = (jiffies - ifa->tstamp)/HZ;
3506 if (preferred > tval)
3507 preferred -= tval;
3508 else
3509 preferred = 0;
3510 if (valid != INFINITY_LIFE_TIME) {
3511 if (valid > tval)
3512 valid -= tval;
3513 else
3514 valid = 0;
3515 }
3516 }
3517 } else {
3518 preferred = INFINITY_LIFE_TIME;
3519 valid = INFINITY_LIFE_TIME;
3520 }
3521
3522 if (nla_put(skb, IFA_ADDRESS, 16, &ifa->addr) < 0 ||
3523 put_cacheinfo(skb, ifa->cstamp, ifa->tstamp, preferred, valid) < 0) {
3524 nlmsg_cancel(skb, nlh);
3525 return -EMSGSIZE;
3526 }
3527
3528 return nlmsg_end(skb, nlh);
3529}
3530
3531static int inet6_fill_ifmcaddr(struct sk_buff *skb, struct ifmcaddr6 *ifmca,
3532 u32 pid, u32 seq, int event, u16 flags)
3533{
3534 struct nlmsghdr *nlh;
3535 u8 scope = RT_SCOPE_UNIVERSE;
3536 int ifindex = ifmca->idev->dev->ifindex;
3537
3538 if (ipv6_addr_scope(&ifmca->mca_addr) & IFA_SITE)
3539 scope = RT_SCOPE_SITE;
3540
3541 nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3542 if (nlh == NULL)
3543 return -EMSGSIZE;
3544
3545 put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3546 if (nla_put(skb, IFA_MULTICAST, 16, &ifmca->mca_addr) < 0 ||
3547 put_cacheinfo(skb, ifmca->mca_cstamp, ifmca->mca_tstamp,
3548 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3549 nlmsg_cancel(skb, nlh);
3550 return -EMSGSIZE;
3551 }
3552
3553 return nlmsg_end(skb, nlh);
3554}
3555
3556static int inet6_fill_ifacaddr(struct sk_buff *skb, struct ifacaddr6 *ifaca,
3557 u32 pid, u32 seq, int event, unsigned int flags)
3558{
3559 struct nlmsghdr *nlh;
3560 u8 scope = RT_SCOPE_UNIVERSE;
3561 int ifindex = ifaca->aca_idev->dev->ifindex;
3562
3563 if (ipv6_addr_scope(&ifaca->aca_addr) & IFA_SITE)
3564 scope = RT_SCOPE_SITE;
3565
3566 nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3567 if (nlh == NULL)
3568 return -EMSGSIZE;
3569
3570 put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3571 if (nla_put(skb, IFA_ANYCAST, 16, &ifaca->aca_addr) < 0 ||
3572 put_cacheinfo(skb, ifaca->aca_cstamp, ifaca->aca_tstamp,
3573 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3574 nlmsg_cancel(skb, nlh);
3575 return -EMSGSIZE;
3576 }
3577
3578 return nlmsg_end(skb, nlh);
3579}
3580
3581enum addr_type_t {
3582 UNICAST_ADDR,
3583 MULTICAST_ADDR,
3584 ANYCAST_ADDR,
3585};
3586
3587/* called with rcu_read_lock() */
3588static int in6_dump_addrs(struct inet6_dev *idev, struct sk_buff *skb,
3589 struct netlink_callback *cb, enum addr_type_t type,
3590 int s_ip_idx, int *p_ip_idx)
3591{
3592 struct ifmcaddr6 *ifmca;
3593 struct ifacaddr6 *ifaca;
3594 int err = 1;
3595 int ip_idx = *p_ip_idx;
3596
3597 read_lock_bh(&idev->lock);
3598 switch (type) {
3599 case UNICAST_ADDR: {
3600 struct inet6_ifaddr *ifa;
3601
3602 /* unicast address incl. temp addr */
3603 list_for_each_entry(ifa, &idev->addr_list, if_list) {
3604 if (++ip_idx < s_ip_idx)
3605 continue;
3606 err = inet6_fill_ifaddr(skb, ifa,
3607 NETLINK_CB(cb->skb).pid,
3608 cb->nlh->nlmsg_seq,
3609 RTM_NEWADDR,
3610 NLM_F_MULTI);
3611 if (err <= 0)
3612 break;
3613 }
3614 break;
3615 }
3616 case MULTICAST_ADDR:
3617 /* multicast address */
3618 for (ifmca = idev->mc_list; ifmca;
3619 ifmca = ifmca->next, ip_idx++) {
3620 if (ip_idx < s_ip_idx)
3621 continue;
3622 err = inet6_fill_ifmcaddr(skb, ifmca,
3623 NETLINK_CB(cb->skb).pid,
3624 cb->nlh->nlmsg_seq,
3625 RTM_GETMULTICAST,
3626 NLM_F_MULTI);
3627 if (err <= 0)
3628 break;
3629 }
3630 break;
3631 case ANYCAST_ADDR:
3632 /* anycast address */
3633 for (ifaca = idev->ac_list; ifaca;
3634 ifaca = ifaca->aca_next, ip_idx++) {
3635 if (ip_idx < s_ip_idx)
3636 continue;
3637 err = inet6_fill_ifacaddr(skb, ifaca,
3638 NETLINK_CB(cb->skb).pid,
3639 cb->nlh->nlmsg_seq,
3640 RTM_GETANYCAST,
3641 NLM_F_MULTI);
3642 if (err <= 0)
3643 break;
3644 }
3645 break;
3646 default:
3647 break;
3648 }
3649 read_unlock_bh(&idev->lock);
3650 *p_ip_idx = ip_idx;
3651 return err;
3652}
3653
3654static int inet6_dump_addr(struct sk_buff *skb, struct netlink_callback *cb,
3655 enum addr_type_t type)
3656{
3657 struct net *net = sock_net(skb->sk);
3658 int h, s_h;
3659 int idx, ip_idx;
3660 int s_idx, s_ip_idx;
3661 struct net_device *dev;
3662 struct inet6_dev *idev;
3663 struct hlist_head *head;
3664 struct hlist_node *node;
3665
3666 s_h = cb->args[0];
3667 s_idx = idx = cb->args[1];
3668 s_ip_idx = ip_idx = cb->args[2];
3669
3670 rcu_read_lock();
3671 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
3672 idx = 0;
3673 head = &net->dev_index_head[h];
3674 hlist_for_each_entry_rcu(dev, node, head, index_hlist) {
3675 if (idx < s_idx)
3676 goto cont;
3677 if (h > s_h || idx > s_idx)
3678 s_ip_idx = 0;
3679 ip_idx = 0;
3680 idev = __in6_dev_get(dev);
3681 if (!idev)
3682 goto cont;
3683
3684 if (in6_dump_addrs(idev, skb, cb, type,
3685 s_ip_idx, &ip_idx) <= 0)
3686 goto done;
3687cont:
3688 idx++;
3689 }
3690 }
3691done:
3692 rcu_read_unlock();
3693 cb->args[0] = h;
3694 cb->args[1] = idx;
3695 cb->args[2] = ip_idx;
3696
3697 return skb->len;
3698}
3699
3700static int inet6_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
3701{
3702 enum addr_type_t type = UNICAST_ADDR;
3703
3704 return inet6_dump_addr(skb, cb, type);
3705}
3706
3707static int inet6_dump_ifmcaddr(struct sk_buff *skb, struct netlink_callback *cb)
3708{
3709 enum addr_type_t type = MULTICAST_ADDR;
3710
3711 return inet6_dump_addr(skb, cb, type);
3712}
3713
3714
3715static int inet6_dump_ifacaddr(struct sk_buff *skb, struct netlink_callback *cb)
3716{
3717 enum addr_type_t type = ANYCAST_ADDR;
3718
3719 return inet6_dump_addr(skb, cb, type);
3720}
3721
3722static int inet6_rtm_getaddr(struct sk_buff *in_skb, struct nlmsghdr* nlh,
3723 void *arg)
3724{
3725 struct net *net = sock_net(in_skb->sk);
3726 struct ifaddrmsg *ifm;
3727 struct nlattr *tb[IFA_MAX+1];
3728 struct in6_addr *addr = NULL;
3729 struct net_device *dev = NULL;
3730 struct inet6_ifaddr *ifa;
3731 struct sk_buff *skb;
3732 int err;
3733
3734 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3735 if (err < 0)
3736 goto errout;
3737
3738 addr = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3739 if (addr == NULL) {
3740 err = -EINVAL;
3741 goto errout;
3742 }
3743
3744 ifm = nlmsg_data(nlh);
3745 if (ifm->ifa_index)
3746 dev = __dev_get_by_index(net, ifm->ifa_index);
3747
3748 ifa = ipv6_get_ifaddr(net, addr, dev, 1);
3749 if (!ifa) {
3750 err = -EADDRNOTAVAIL;
3751 goto errout;
3752 }
3753
3754 skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_KERNEL);
3755 if (!skb) {
3756 err = -ENOBUFS;
3757 goto errout_ifa;
3758 }
3759
3760 err = inet6_fill_ifaddr(skb, ifa, NETLINK_CB(in_skb).pid,
3761 nlh->nlmsg_seq, RTM_NEWADDR, 0);
3762 if (err < 0) {
3763 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
3764 WARN_ON(err == -EMSGSIZE);
3765 kfree_skb(skb);
3766 goto errout_ifa;
3767 }
3768 err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).pid);
3769errout_ifa:
3770 in6_ifa_put(ifa);
3771errout:
3772 return err;
3773}
3774
3775static void inet6_ifa_notify(int event, struct inet6_ifaddr *ifa)
3776{
3777 struct sk_buff *skb;
3778 struct net *net = dev_net(ifa->idev->dev);
3779 int err = -ENOBUFS;
3780
3781 skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_ATOMIC);
3782 if (skb == NULL)
3783 goto errout;
3784
3785 err = inet6_fill_ifaddr(skb, ifa, 0, 0, event, 0);
3786 if (err < 0) {
3787 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
3788 WARN_ON(err == -EMSGSIZE);
3789 kfree_skb(skb);
3790 goto errout;
3791 }
3792 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
3793 return;
3794errout:
3795 if (err < 0)
3796 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
3797}
3798
3799static inline void ipv6_store_devconf(struct ipv6_devconf *cnf,
3800 __s32 *array, int bytes)
3801{
3802 BUG_ON(bytes < (DEVCONF_MAX * 4));
3803
3804 memset(array, 0, bytes);
3805 array[DEVCONF_FORWARDING] = cnf->forwarding;
3806 array[DEVCONF_HOPLIMIT] = cnf->hop_limit;
3807 array[DEVCONF_MTU6] = cnf->mtu6;
3808 array[DEVCONF_ACCEPT_RA] = cnf->accept_ra;
3809 array[DEVCONF_ACCEPT_REDIRECTS] = cnf->accept_redirects;
3810 array[DEVCONF_AUTOCONF] = cnf->autoconf;
3811 array[DEVCONF_DAD_TRANSMITS] = cnf->dad_transmits;
3812 array[DEVCONF_RTR_SOLICITS] = cnf->rtr_solicits;
3813 array[DEVCONF_RTR_SOLICIT_INTERVAL] =
3814 jiffies_to_msecs(cnf->rtr_solicit_interval);
3815 array[DEVCONF_RTR_SOLICIT_DELAY] =
3816 jiffies_to_msecs(cnf->rtr_solicit_delay);
3817 array[DEVCONF_FORCE_MLD_VERSION] = cnf->force_mld_version;
3818#ifdef CONFIG_IPV6_PRIVACY
3819 array[DEVCONF_USE_TEMPADDR] = cnf->use_tempaddr;
3820 array[DEVCONF_TEMP_VALID_LFT] = cnf->temp_valid_lft;
3821 array[DEVCONF_TEMP_PREFERED_LFT] = cnf->temp_prefered_lft;
3822 array[DEVCONF_REGEN_MAX_RETRY] = cnf->regen_max_retry;
3823 array[DEVCONF_MAX_DESYNC_FACTOR] = cnf->max_desync_factor;
3824#endif
3825 array[DEVCONF_MAX_ADDRESSES] = cnf->max_addresses;
3826 array[DEVCONF_ACCEPT_RA_DEFRTR] = cnf->accept_ra_defrtr;
3827 array[DEVCONF_ACCEPT_RA_PINFO] = cnf->accept_ra_pinfo;
3828#ifdef CONFIG_IPV6_ROUTER_PREF
3829 array[DEVCONF_ACCEPT_RA_RTR_PREF] = cnf->accept_ra_rtr_pref;
3830 array[DEVCONF_RTR_PROBE_INTERVAL] =
3831 jiffies_to_msecs(cnf->rtr_probe_interval);
3832#ifdef CONFIG_IPV6_ROUTE_INFO
3833 array[DEVCONF_ACCEPT_RA_RT_INFO_MAX_PLEN] = cnf->accept_ra_rt_info_max_plen;
3834#endif
3835#endif
3836 array[DEVCONF_PROXY_NDP] = cnf->proxy_ndp;
3837 array[DEVCONF_ACCEPT_SOURCE_ROUTE] = cnf->accept_source_route;
3838#ifdef CONFIG_IPV6_OPTIMISTIC_DAD
3839 array[DEVCONF_OPTIMISTIC_DAD] = cnf->optimistic_dad;
3840#endif
3841#ifdef CONFIG_IPV6_MROUTE
3842 array[DEVCONF_MC_FORWARDING] = cnf->mc_forwarding;
3843#endif
3844 array[DEVCONF_DISABLE_IPV6] = cnf->disable_ipv6;
3845 array[DEVCONF_ACCEPT_DAD] = cnf->accept_dad;
3846 array[DEVCONF_FORCE_TLLAO] = cnf->force_tllao;
3847}
3848
3849static inline size_t inet6_ifla6_size(void)
3850{
3851 return nla_total_size(4) /* IFLA_INET6_FLAGS */
3852 + nla_total_size(sizeof(struct ifla_cacheinfo))
3853 + nla_total_size(DEVCONF_MAX * 4) /* IFLA_INET6_CONF */
3854 + nla_total_size(IPSTATS_MIB_MAX * 8) /* IFLA_INET6_STATS */
3855 + nla_total_size(ICMP6_MIB_MAX * 8); /* IFLA_INET6_ICMP6STATS */
3856}
3857
3858static inline size_t inet6_if_nlmsg_size(void)
3859{
3860 return NLMSG_ALIGN(sizeof(struct ifinfomsg))
3861 + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
3862 + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
3863 + nla_total_size(4) /* IFLA_MTU */
3864 + nla_total_size(4) /* IFLA_LINK */
3865 + nla_total_size(inet6_ifla6_size()); /* IFLA_PROTINFO */
3866}
3867
3868static inline void __snmp6_fill_statsdev(u64 *stats, atomic_long_t *mib,
3869 int items, int bytes)
3870{
3871 int i;
3872 int pad = bytes - sizeof(u64) * items;
3873 BUG_ON(pad < 0);
3874
3875 /* Use put_unaligned() because stats may not be aligned for u64. */
3876 put_unaligned(items, &stats[0]);
3877 for (i = 1; i < items; i++)
3878 put_unaligned(atomic_long_read(&mib[i]), &stats[i]);
3879
3880 memset(&stats[items], 0, pad);
3881}
3882
3883static inline void __snmp6_fill_stats64(u64 *stats, void __percpu **mib,
3884 int items, int bytes, size_t syncpoff)
3885{
3886 int i;
3887 int pad = bytes - sizeof(u64) * items;
3888 BUG_ON(pad < 0);
3889
3890 /* Use put_unaligned() because stats may not be aligned for u64. */
3891 put_unaligned(items, &stats[0]);
3892 for (i = 1; i < items; i++)
3893 put_unaligned(snmp_fold_field64(mib, i, syncpoff), &stats[i]);
3894
3895 memset(&stats[items], 0, pad);
3896}
3897
3898static void snmp6_fill_stats(u64 *stats, struct inet6_dev *idev, int attrtype,
3899 int bytes)
3900{
3901 switch (attrtype) {
3902 case IFLA_INET6_STATS:
3903 __snmp6_fill_stats64(stats, (void __percpu **)idev->stats.ipv6,
3904 IPSTATS_MIB_MAX, bytes, offsetof(struct ipstats_mib, syncp));
3905 break;
3906 case IFLA_INET6_ICMP6STATS:
3907 __snmp6_fill_statsdev(stats, idev->stats.icmpv6dev->mibs, ICMP6_MIB_MAX, bytes);
3908 break;
3909 }
3910}
3911
3912static int inet6_fill_ifla6_attrs(struct sk_buff *skb, struct inet6_dev *idev)
3913{
3914 struct nlattr *nla;
3915 struct ifla_cacheinfo ci;
3916
3917 NLA_PUT_U32(skb, IFLA_INET6_FLAGS, idev->if_flags);
3918
3919 ci.max_reasm_len = IPV6_MAXPLEN;
3920 ci.tstamp = cstamp_delta(idev->tstamp);
3921 ci.reachable_time = jiffies_to_msecs(idev->nd_parms->reachable_time);
3922 ci.retrans_time = jiffies_to_msecs(idev->nd_parms->retrans_time);
3923 NLA_PUT(skb, IFLA_INET6_CACHEINFO, sizeof(ci), &ci);
3924
3925 nla = nla_reserve(skb, IFLA_INET6_CONF, DEVCONF_MAX * sizeof(s32));
3926 if (nla == NULL)
3927 goto nla_put_failure;
3928 ipv6_store_devconf(&idev->cnf, nla_data(nla), nla_len(nla));
3929
3930 /* XXX - MC not implemented */
3931
3932 nla = nla_reserve(skb, IFLA_INET6_STATS, IPSTATS_MIB_MAX * sizeof(u64));
3933 if (nla == NULL)
3934 goto nla_put_failure;
3935 snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_STATS, nla_len(nla));
3936
3937 nla = nla_reserve(skb, IFLA_INET6_ICMP6STATS, ICMP6_MIB_MAX * sizeof(u64));
3938 if (nla == NULL)
3939 goto nla_put_failure;
3940 snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_ICMP6STATS, nla_len(nla));
3941
3942 return 0;
3943
3944nla_put_failure:
3945 return -EMSGSIZE;
3946}
3947
3948static size_t inet6_get_link_af_size(const struct net_device *dev)
3949{
3950 if (!__in6_dev_get(dev))
3951 return 0;
3952
3953 return inet6_ifla6_size();
3954}
3955
3956static int inet6_fill_link_af(struct sk_buff *skb, const struct net_device *dev)
3957{
3958 struct inet6_dev *idev = __in6_dev_get(dev);
3959
3960 if (!idev)
3961 return -ENODATA;
3962
3963 if (inet6_fill_ifla6_attrs(skb, idev) < 0)
3964 return -EMSGSIZE;
3965
3966 return 0;
3967}
3968
3969static int inet6_fill_ifinfo(struct sk_buff *skb, struct inet6_dev *idev,
3970 u32 pid, u32 seq, int event, unsigned int flags)
3971{
3972 struct net_device *dev = idev->dev;
3973 struct ifinfomsg *hdr;
3974 struct nlmsghdr *nlh;
3975 void *protoinfo;
3976
3977 nlh = nlmsg_put(skb, pid, seq, event, sizeof(*hdr), flags);
3978 if (nlh == NULL)
3979 return -EMSGSIZE;
3980
3981 hdr = nlmsg_data(nlh);
3982 hdr->ifi_family = AF_INET6;
3983 hdr->__ifi_pad = 0;
3984 hdr->ifi_type = dev->type;
3985 hdr->ifi_index = dev->ifindex;
3986 hdr->ifi_flags = dev_get_flags(dev);
3987 hdr->ifi_change = 0;
3988
3989 NLA_PUT_STRING(skb, IFLA_IFNAME, dev->name);
3990
3991 if (dev->addr_len)
3992 NLA_PUT(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr);
3993
3994 NLA_PUT_U32(skb, IFLA_MTU, dev->mtu);
3995 if (dev->ifindex != dev->iflink)
3996 NLA_PUT_U32(skb, IFLA_LINK, dev->iflink);
3997
3998 protoinfo = nla_nest_start(skb, IFLA_PROTINFO);
3999 if (protoinfo == NULL)
4000 goto nla_put_failure;
4001
4002 if (inet6_fill_ifla6_attrs(skb, idev) < 0)
4003 goto nla_put_failure;
4004
4005 nla_nest_end(skb, protoinfo);
4006 return nlmsg_end(skb, nlh);
4007
4008nla_put_failure:
4009 nlmsg_cancel(skb, nlh);
4010 return -EMSGSIZE;
4011}
4012
4013static int inet6_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
4014{
4015 struct net *net = sock_net(skb->sk);
4016 int h, s_h;
4017 int idx = 0, s_idx;
4018 struct net_device *dev;
4019 struct inet6_dev *idev;
4020 struct hlist_head *head;
4021 struct hlist_node *node;
4022
4023 s_h = cb->args[0];
4024 s_idx = cb->args[1];
4025
4026 rcu_read_lock();
4027 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
4028 idx = 0;
4029 head = &net->dev_index_head[h];
4030 hlist_for_each_entry_rcu(dev, node, head, index_hlist) {
4031 if (idx < s_idx)
4032 goto cont;
4033 idev = __in6_dev_get(dev);
4034 if (!idev)
4035 goto cont;
4036 if (inet6_fill_ifinfo(skb, idev,
4037 NETLINK_CB(cb->skb).pid,
4038 cb->nlh->nlmsg_seq,
4039 RTM_NEWLINK, NLM_F_MULTI) <= 0)
4040 goto out;
4041cont:
4042 idx++;
4043 }
4044 }
4045out:
4046 rcu_read_unlock();
4047 cb->args[1] = idx;
4048 cb->args[0] = h;
4049
4050 return skb->len;
4051}
4052
4053void inet6_ifinfo_notify(int event, struct inet6_dev *idev)
4054{
4055 struct sk_buff *skb;
4056 struct net *net = dev_net(idev->dev);
4057 int err = -ENOBUFS;
4058
4059 skb = nlmsg_new(inet6_if_nlmsg_size(), GFP_ATOMIC);
4060 if (skb == NULL)
4061 goto errout;
4062
4063 err = inet6_fill_ifinfo(skb, idev, 0, 0, event, 0);
4064 if (err < 0) {
4065 /* -EMSGSIZE implies BUG in inet6_if_nlmsg_size() */
4066 WARN_ON(err == -EMSGSIZE);
4067 kfree_skb(skb);
4068 goto errout;
4069 }
4070 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFINFO, NULL, GFP_ATOMIC);
4071 return;
4072errout:
4073 if (err < 0)
4074 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFINFO, err);
4075}
4076
4077static inline size_t inet6_prefix_nlmsg_size(void)
4078{
4079 return NLMSG_ALIGN(sizeof(struct prefixmsg))
4080 + nla_total_size(sizeof(struct in6_addr))
4081 + nla_total_size(sizeof(struct prefix_cacheinfo));
4082}
4083
4084static int inet6_fill_prefix(struct sk_buff *skb, struct inet6_dev *idev,
4085 struct prefix_info *pinfo, u32 pid, u32 seq,
4086 int event, unsigned int flags)
4087{
4088 struct prefixmsg *pmsg;
4089 struct nlmsghdr *nlh;
4090 struct prefix_cacheinfo ci;
4091
4092 nlh = nlmsg_put(skb, pid, seq, event, sizeof(*pmsg), flags);
4093 if (nlh == NULL)
4094 return -EMSGSIZE;
4095
4096 pmsg = nlmsg_data(nlh);
4097 pmsg->prefix_family = AF_INET6;
4098 pmsg->prefix_pad1 = 0;
4099 pmsg->prefix_pad2 = 0;
4100 pmsg->prefix_ifindex = idev->dev->ifindex;
4101 pmsg->prefix_len = pinfo->prefix_len;
4102 pmsg->prefix_type = pinfo->type;
4103 pmsg->prefix_pad3 = 0;
4104 pmsg->prefix_flags = 0;
4105 if (pinfo->onlink)
4106 pmsg->prefix_flags |= IF_PREFIX_ONLINK;
4107 if (pinfo->autoconf)
4108 pmsg->prefix_flags |= IF_PREFIX_AUTOCONF;
4109
4110 NLA_PUT(skb, PREFIX_ADDRESS, sizeof(pinfo->prefix), &pinfo->prefix);
4111
4112 ci.preferred_time = ntohl(pinfo->prefered);
4113 ci.valid_time = ntohl(pinfo->valid);
4114 NLA_PUT(skb, PREFIX_CACHEINFO, sizeof(ci), &ci);
4115
4116 return nlmsg_end(skb, nlh);
4117
4118nla_put_failure:
4119 nlmsg_cancel(skb, nlh);
4120 return -EMSGSIZE;
4121}
4122
4123static void inet6_prefix_notify(int event, struct inet6_dev *idev,
4124 struct prefix_info *pinfo)
4125{
4126 struct sk_buff *skb;
4127 struct net *net = dev_net(idev->dev);
4128 int err = -ENOBUFS;
4129
4130 skb = nlmsg_new(inet6_prefix_nlmsg_size(), GFP_ATOMIC);
4131 if (skb == NULL)
4132 goto errout;
4133
4134 err = inet6_fill_prefix(skb, idev, pinfo, 0, 0, event, 0);
4135 if (err < 0) {
4136 /* -EMSGSIZE implies BUG in inet6_prefix_nlmsg_size() */
4137 WARN_ON(err == -EMSGSIZE);
4138 kfree_skb(skb);
4139 goto errout;
4140 }
4141 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_PREFIX, NULL, GFP_ATOMIC);
4142 return;
4143errout:
4144 if (err < 0)
4145 rtnl_set_sk_err(net, RTNLGRP_IPV6_PREFIX, err);
4146}
4147
4148static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
4149{
4150 inet6_ifa_notify(event ? : RTM_NEWADDR, ifp);
4151
4152 switch (event) {
4153 case RTM_NEWADDR:
4154 /*
4155 * If the address was optimistic
4156 * we inserted the route at the start of
4157 * our DAD process, so we don't need
4158 * to do it again
4159 */
4160 if (!(ifp->rt->rt6i_node))
4161 ip6_ins_rt(ifp->rt);
4162 if (ifp->idev->cnf.forwarding)
4163 addrconf_join_anycast(ifp);
4164 break;
4165 case RTM_DELADDR:
4166 if (ifp->idev->cnf.forwarding)
4167 addrconf_leave_anycast(ifp);
4168 addrconf_leave_solict(ifp->idev, &ifp->addr);
4169 dst_hold(&ifp->rt->dst);
4170
4171 if (ip6_del_rt(ifp->rt))
4172 dst_free(&ifp->rt->dst);
4173 break;
4174 }
4175}
4176
4177static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
4178{
4179 rcu_read_lock_bh();
4180 if (likely(ifp->idev->dead == 0))
4181 __ipv6_ifa_notify(event, ifp);
4182 rcu_read_unlock_bh();
4183}
4184
4185#ifdef CONFIG_SYSCTL
4186
4187static
4188int addrconf_sysctl_forward(ctl_table *ctl, int write,
4189 void __user *buffer, size_t *lenp, loff_t *ppos)
4190{
4191 int *valp = ctl->data;
4192 int val = *valp;
4193 loff_t pos = *ppos;
4194 int ret;
4195
4196 ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
4197
4198 if (write)
4199 ret = addrconf_fixup_forwarding(ctl, valp, val);
4200 if (ret)
4201 *ppos = pos;
4202 return ret;
4203}
4204
4205static void dev_disable_change(struct inet6_dev *idev)
4206{
4207 if (!idev || !idev->dev)
4208 return;
4209
4210 if (idev->cnf.disable_ipv6)
4211 addrconf_notify(NULL, NETDEV_DOWN, idev->dev);
4212 else
4213 addrconf_notify(NULL, NETDEV_UP, idev->dev);
4214}
4215
4216static void addrconf_disable_change(struct net *net, __s32 newf)
4217{
4218 struct net_device *dev;
4219 struct inet6_dev *idev;
4220
4221 rcu_read_lock();
4222 for_each_netdev_rcu(net, dev) {
4223 idev = __in6_dev_get(dev);
4224 if (idev) {
4225 int changed = (!idev->cnf.disable_ipv6) ^ (!newf);
4226 idev->cnf.disable_ipv6 = newf;
4227 if (changed)
4228 dev_disable_change(idev);
4229 }
4230 }
4231 rcu_read_unlock();
4232}
4233
4234static int addrconf_disable_ipv6(struct ctl_table *table, int *p, int old)
4235{
4236 struct net *net;
4237
4238 net = (struct net *)table->extra2;
4239
4240 if (p == &net->ipv6.devconf_dflt->disable_ipv6)
4241 return 0;
4242
4243 if (!rtnl_trylock()) {
4244 /* Restore the original values before restarting */
4245 *p = old;
4246 return restart_syscall();
4247 }
4248
4249 if (p == &net->ipv6.devconf_all->disable_ipv6) {
4250 __s32 newf = net->ipv6.devconf_all->disable_ipv6;
4251 net->ipv6.devconf_dflt->disable_ipv6 = newf;
4252 addrconf_disable_change(net, newf);
4253 } else if ((!*p) ^ (!old))
4254 dev_disable_change((struct inet6_dev *)table->extra1);
4255
4256 rtnl_unlock();
4257 return 0;
4258}
4259
4260static
4261int addrconf_sysctl_disable(ctl_table *ctl, int write,
4262 void __user *buffer, size_t *lenp, loff_t *ppos)
4263{
4264 int *valp = ctl->data;
4265 int val = *valp;
4266 loff_t pos = *ppos;
4267 int ret;
4268
4269 ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
4270
4271 if (write)
4272 ret = addrconf_disable_ipv6(ctl, valp, val);
4273 if (ret)
4274 *ppos = pos;
4275 return ret;
4276}
4277
4278static struct addrconf_sysctl_table
4279{
4280 struct ctl_table_header *sysctl_header;
4281 ctl_table addrconf_vars[DEVCONF_MAX+1];
4282 char *dev_name;
4283} addrconf_sysctl __read_mostly = {
4284 .sysctl_header = NULL,
4285 .addrconf_vars = {
4286 {
4287 .procname = "forwarding",
4288 .data = &ipv6_devconf.forwarding,
4289 .maxlen = sizeof(int),
4290 .mode = 0644,
4291 .proc_handler = addrconf_sysctl_forward,
4292 },
4293 {
4294 .procname = "hop_limit",
4295 .data = &ipv6_devconf.hop_limit,
4296 .maxlen = sizeof(int),
4297 .mode = 0644,
4298 .proc_handler = proc_dointvec,
4299 },
4300 {
4301 .procname = "mtu",
4302 .data = &ipv6_devconf.mtu6,
4303 .maxlen = sizeof(int),
4304 .mode = 0644,
4305 .proc_handler = proc_dointvec,
4306 },
4307 {
4308 .procname = "accept_ra",
4309 .data = &ipv6_devconf.accept_ra,
4310 .maxlen = sizeof(int),
4311 .mode = 0644,
4312 .proc_handler = proc_dointvec,
4313 },
4314 {
4315 .procname = "accept_redirects",
4316 .data = &ipv6_devconf.accept_redirects,
4317 .maxlen = sizeof(int),
4318 .mode = 0644,
4319 .proc_handler = proc_dointvec,
4320 },
4321 {
4322 .procname = "autoconf",
4323 .data = &ipv6_devconf.autoconf,
4324 .maxlen = sizeof(int),
4325 .mode = 0644,
4326 .proc_handler = proc_dointvec,
4327 },
4328 {
4329 .procname = "dad_transmits",
4330 .data = &ipv6_devconf.dad_transmits,
4331 .maxlen = sizeof(int),
4332 .mode = 0644,
4333 .proc_handler = proc_dointvec,
4334 },
4335 {
4336 .procname = "router_solicitations",
4337 .data = &ipv6_devconf.rtr_solicits,
4338 .maxlen = sizeof(int),
4339 .mode = 0644,
4340 .proc_handler = proc_dointvec,
4341 },
4342 {
4343 .procname = "router_solicitation_interval",
4344 .data = &ipv6_devconf.rtr_solicit_interval,
4345 .maxlen = sizeof(int),
4346 .mode = 0644,
4347 .proc_handler = proc_dointvec_jiffies,
4348 },
4349 {
4350 .procname = "router_solicitation_delay",
4351 .data = &ipv6_devconf.rtr_solicit_delay,
4352 .maxlen = sizeof(int),
4353 .mode = 0644,
4354 .proc_handler = proc_dointvec_jiffies,
4355 },
4356 {
4357 .procname = "force_mld_version",
4358 .data = &ipv6_devconf.force_mld_version,
4359 .maxlen = sizeof(int),
4360 .mode = 0644,
4361 .proc_handler = proc_dointvec,
4362 },
4363#ifdef CONFIG_IPV6_PRIVACY
4364 {
4365 .procname = "use_tempaddr",
4366 .data = &ipv6_devconf.use_tempaddr,
4367 .maxlen = sizeof(int),
4368 .mode = 0644,
4369 .proc_handler = proc_dointvec,
4370 },
4371 {
4372 .procname = "temp_valid_lft",
4373 .data = &ipv6_devconf.temp_valid_lft,
4374 .maxlen = sizeof(int),
4375 .mode = 0644,
4376 .proc_handler = proc_dointvec,
4377 },
4378 {
4379 .procname = "temp_prefered_lft",
4380 .data = &ipv6_devconf.temp_prefered_lft,
4381 .maxlen = sizeof(int),
4382 .mode = 0644,
4383 .proc_handler = proc_dointvec,
4384 },
4385 {
4386 .procname = "regen_max_retry",
4387 .data = &ipv6_devconf.regen_max_retry,
4388 .maxlen = sizeof(int),
4389 .mode = 0644,
4390 .proc_handler = proc_dointvec,
4391 },
4392 {
4393 .procname = "max_desync_factor",
4394 .data = &ipv6_devconf.max_desync_factor,
4395 .maxlen = sizeof(int),
4396 .mode = 0644,
4397 .proc_handler = proc_dointvec,
4398 },
4399#endif
4400 {
4401 .procname = "max_addresses",
4402 .data = &ipv6_devconf.max_addresses,
4403 .maxlen = sizeof(int),
4404 .mode = 0644,
4405 .proc_handler = proc_dointvec,
4406 },
4407 {
4408 .procname = "accept_ra_defrtr",
4409 .data = &ipv6_devconf.accept_ra_defrtr,
4410 .maxlen = sizeof(int),
4411 .mode = 0644,
4412 .proc_handler = proc_dointvec,
4413 },
4414 {
4415 .procname = "accept_ra_pinfo",
4416 .data = &ipv6_devconf.accept_ra_pinfo,
4417 .maxlen = sizeof(int),
4418 .mode = 0644,
4419 .proc_handler = proc_dointvec,
4420 },
4421#ifdef CONFIG_IPV6_ROUTER_PREF
4422 {
4423 .procname = "accept_ra_rtr_pref",
4424 .data = &ipv6_devconf.accept_ra_rtr_pref,
4425 .maxlen = sizeof(int),
4426 .mode = 0644,
4427 .proc_handler = proc_dointvec,
4428 },
4429 {
4430 .procname = "router_probe_interval",
4431 .data = &ipv6_devconf.rtr_probe_interval,
4432 .maxlen = sizeof(int),
4433 .mode = 0644,
4434 .proc_handler = proc_dointvec_jiffies,
4435 },
4436#ifdef CONFIG_IPV6_ROUTE_INFO
4437 {
4438 .procname = "accept_ra_rt_info_max_plen",
4439 .data = &ipv6_devconf.accept_ra_rt_info_max_plen,
4440 .maxlen = sizeof(int),
4441 .mode = 0644,
4442 .proc_handler = proc_dointvec,
4443 },
4444#endif
4445#endif
4446 {
4447 .procname = "proxy_ndp",
4448 .data = &ipv6_devconf.proxy_ndp,
4449 .maxlen = sizeof(int),
4450 .mode = 0644,
4451 .proc_handler = proc_dointvec,
4452 },
4453 {
4454 .procname = "accept_source_route",
4455 .data = &ipv6_devconf.accept_source_route,
4456 .maxlen = sizeof(int),
4457 .mode = 0644,
4458 .proc_handler = proc_dointvec,
4459 },
4460#ifdef CONFIG_IPV6_OPTIMISTIC_DAD
4461 {
4462 .procname = "optimistic_dad",
4463 .data = &ipv6_devconf.optimistic_dad,
4464 .maxlen = sizeof(int),
4465 .mode = 0644,
4466 .proc_handler = proc_dointvec,
4467
4468 },
4469#endif
4470#ifdef CONFIG_IPV6_MROUTE
4471 {
4472 .procname = "mc_forwarding",
4473 .data = &ipv6_devconf.mc_forwarding,
4474 .maxlen = sizeof(int),
4475 .mode = 0444,
4476 .proc_handler = proc_dointvec,
4477 },
4478#endif
4479 {
4480 .procname = "disable_ipv6",
4481 .data = &ipv6_devconf.disable_ipv6,
4482 .maxlen = sizeof(int),
4483 .mode = 0644,
4484 .proc_handler = addrconf_sysctl_disable,
4485 },
4486 {
4487 .procname = "accept_dad",
4488 .data = &ipv6_devconf.accept_dad,
4489 .maxlen = sizeof(int),
4490 .mode = 0644,
4491 .proc_handler = proc_dointvec,
4492 },
4493 {
4494 .procname = "force_tllao",
4495 .data = &ipv6_devconf.force_tllao,
4496 .maxlen = sizeof(int),
4497 .mode = 0644,
4498 .proc_handler = proc_dointvec
4499 },
4500 {
4501 /* sentinel */
4502 }
4503 },
4504};
4505
4506static int __addrconf_sysctl_register(struct net *net, char *dev_name,
4507 struct inet6_dev *idev, struct ipv6_devconf *p)
4508{
4509 int i;
4510 struct addrconf_sysctl_table *t;
4511
4512#define ADDRCONF_CTL_PATH_DEV 3
4513
4514 struct ctl_path addrconf_ctl_path[] = {
4515 { .procname = "net", },
4516 { .procname = "ipv6", },
4517 { .procname = "conf", },
4518 { /* to be set */ },
4519 { },
4520 };
4521
4522
4523 t = kmemdup(&addrconf_sysctl, sizeof(*t), GFP_KERNEL);
4524 if (t == NULL)
4525 goto out;
4526
4527 for (i = 0; t->addrconf_vars[i].data; i++) {
4528 t->addrconf_vars[i].data += (char *)p - (char *)&ipv6_devconf;
4529 t->addrconf_vars[i].extra1 = idev; /* embedded; no ref */
4530 t->addrconf_vars[i].extra2 = net;
4531 }
4532
4533 /*
4534 * Make a copy of dev_name, because '.procname' is regarded as const
4535 * by sysctl and we wouldn't want anyone to change it under our feet
4536 * (see SIOCSIFNAME).
4537 */
4538 t->dev_name = kstrdup(dev_name, GFP_KERNEL);
4539 if (!t->dev_name)
4540 goto free;
4541
4542 addrconf_ctl_path[ADDRCONF_CTL_PATH_DEV].procname = t->dev_name;
4543
4544 t->sysctl_header = register_net_sysctl_table(net, addrconf_ctl_path,
4545 t->addrconf_vars);
4546 if (t->sysctl_header == NULL)
4547 goto free_procname;
4548
4549 p->sysctl = t;
4550 return 0;
4551
4552free_procname:
4553 kfree(t->dev_name);
4554free:
4555 kfree(t);
4556out:
4557 return -ENOBUFS;
4558}
4559
4560static void __addrconf_sysctl_unregister(struct ipv6_devconf *p)
4561{
4562 struct addrconf_sysctl_table *t;
4563
4564 if (p->sysctl == NULL)
4565 return;
4566
4567 t = p->sysctl;
4568 p->sysctl = NULL;
4569 unregister_net_sysctl_table(t->sysctl_header);
4570 kfree(t->dev_name);
4571 kfree(t);
4572}
4573
4574static void addrconf_sysctl_register(struct inet6_dev *idev)
4575{
4576 neigh_sysctl_register(idev->dev, idev->nd_parms, "ipv6",
4577 &ndisc_ifinfo_sysctl_change);
4578 __addrconf_sysctl_register(dev_net(idev->dev), idev->dev->name,
4579 idev, &idev->cnf);
4580}
4581
4582static void addrconf_sysctl_unregister(struct inet6_dev *idev)
4583{
4584 __addrconf_sysctl_unregister(&idev->cnf);
4585 neigh_sysctl_unregister(idev->nd_parms);
4586}
4587
4588
4589#endif
4590
4591static int __net_init addrconf_init_net(struct net *net)
4592{
4593 int err;
4594 struct ipv6_devconf *all, *dflt;
4595
4596 err = -ENOMEM;
4597 all = &ipv6_devconf;
4598 dflt = &ipv6_devconf_dflt;
4599
4600 if (!net_eq(net, &init_net)) {
4601 all = kmemdup(all, sizeof(ipv6_devconf), GFP_KERNEL);
4602 if (all == NULL)
4603 goto err_alloc_all;
4604
4605 dflt = kmemdup(dflt, sizeof(ipv6_devconf_dflt), GFP_KERNEL);
4606 if (dflt == NULL)
4607 goto err_alloc_dflt;
4608 } else {
4609 /* these will be inherited by all namespaces */
4610 dflt->autoconf = ipv6_defaults.autoconf;
4611 dflt->disable_ipv6 = ipv6_defaults.disable_ipv6;
4612 }
4613
4614 net->ipv6.devconf_all = all;
4615 net->ipv6.devconf_dflt = dflt;
4616
4617#ifdef CONFIG_SYSCTL
4618 err = __addrconf_sysctl_register(net, "all", NULL, all);
4619 if (err < 0)
4620 goto err_reg_all;
4621
4622 err = __addrconf_sysctl_register(net, "default", NULL, dflt);
4623 if (err < 0)
4624 goto err_reg_dflt;
4625#endif
4626 return 0;
4627
4628#ifdef CONFIG_SYSCTL
4629err_reg_dflt:
4630 __addrconf_sysctl_unregister(all);
4631err_reg_all:
4632 kfree(dflt);
4633#endif
4634err_alloc_dflt:
4635 kfree(all);
4636err_alloc_all:
4637 return err;
4638}
4639
4640static void __net_exit addrconf_exit_net(struct net *net)
4641{
4642#ifdef CONFIG_SYSCTL
4643 __addrconf_sysctl_unregister(net->ipv6.devconf_dflt);
4644 __addrconf_sysctl_unregister(net->ipv6.devconf_all);
4645#endif
4646 if (!net_eq(net, &init_net)) {
4647 kfree(net->ipv6.devconf_dflt);
4648 kfree(net->ipv6.devconf_all);
4649 }
4650}
4651
4652static struct pernet_operations addrconf_ops = {
4653 .init = addrconf_init_net,
4654 .exit = addrconf_exit_net,
4655};
4656
4657/*
4658 * Device notifier
4659 */
4660
4661int register_inet6addr_notifier(struct notifier_block *nb)
4662{
4663 return atomic_notifier_chain_register(&inet6addr_chain, nb);
4664}
4665EXPORT_SYMBOL(register_inet6addr_notifier);
4666
4667int unregister_inet6addr_notifier(struct notifier_block *nb)
4668{
4669 return atomic_notifier_chain_unregister(&inet6addr_chain, nb);
4670}
4671EXPORT_SYMBOL(unregister_inet6addr_notifier);
4672
4673static struct rtnl_af_ops inet6_ops = {
4674 .family = AF_INET6,
4675 .fill_link_af = inet6_fill_link_af,
4676 .get_link_af_size = inet6_get_link_af_size,
4677};
4678
4679/*
4680 * Init / cleanup code
4681 */
4682
4683int __init addrconf_init(void)
4684{
4685 int i, err;
4686
4687 err = ipv6_addr_label_init();
4688 if (err < 0) {
4689 printk(KERN_CRIT "IPv6 Addrconf:"
4690 " cannot initialize default policy table: %d.\n", err);
4691 goto out;
4692 }
4693
4694 err = register_pernet_subsys(&addrconf_ops);
4695 if (err < 0)
4696 goto out_addrlabel;
4697
4698 /* The addrconf netdev notifier requires that loopback_dev
4699 * has it's ipv6 private information allocated and setup
4700 * before it can bring up and give link-local addresses
4701 * to other devices which are up.
4702 *
4703 * Unfortunately, loopback_dev is not necessarily the first
4704 * entry in the global dev_base list of net devices. In fact,
4705 * it is likely to be the very last entry on that list.
4706 * So this causes the notifier registry below to try and
4707 * give link-local addresses to all devices besides loopback_dev
4708 * first, then loopback_dev, which cases all the non-loopback_dev
4709 * devices to fail to get a link-local address.
4710 *
4711 * So, as a temporary fix, allocate the ipv6 structure for
4712 * loopback_dev first by hand.
4713 * Longer term, all of the dependencies ipv6 has upon the loopback
4714 * device and it being up should be removed.
4715 */
4716 rtnl_lock();
4717 if (!ipv6_add_dev(init_net.loopback_dev))
4718 err = -ENOMEM;
4719 rtnl_unlock();
4720 if (err)
4721 goto errlo;
4722
4723 for (i = 0; i < IN6_ADDR_HSIZE; i++)
4724 INIT_HLIST_HEAD(&inet6_addr_lst[i]);
4725
4726 register_netdevice_notifier(&ipv6_dev_notf);
4727
4728 addrconf_verify(0);
4729
4730 err = rtnl_af_register(&inet6_ops);
4731 if (err < 0)
4732 goto errout_af;
4733
4734 err = __rtnl_register(PF_INET6, RTM_GETLINK, NULL, inet6_dump_ifinfo,
4735 NULL);
4736 if (err < 0)
4737 goto errout;
4738
4739 /* Only the first call to __rtnl_register can fail */
4740 __rtnl_register(PF_INET6, RTM_NEWADDR, inet6_rtm_newaddr, NULL, NULL);
4741 __rtnl_register(PF_INET6, RTM_DELADDR, inet6_rtm_deladdr, NULL, NULL);
4742 __rtnl_register(PF_INET6, RTM_GETADDR, inet6_rtm_getaddr,
4743 inet6_dump_ifaddr, NULL);
4744 __rtnl_register(PF_INET6, RTM_GETMULTICAST, NULL,
4745 inet6_dump_ifmcaddr, NULL);
4746 __rtnl_register(PF_INET6, RTM_GETANYCAST, NULL,
4747 inet6_dump_ifacaddr, NULL);
4748
4749 ipv6_addr_label_rtnl_register();
4750
4751 return 0;
4752errout:
4753 rtnl_af_unregister(&inet6_ops);
4754errout_af:
4755 unregister_netdevice_notifier(&ipv6_dev_notf);
4756errlo:
4757 unregister_pernet_subsys(&addrconf_ops);
4758out_addrlabel:
4759 ipv6_addr_label_cleanup();
4760out:
4761 return err;
4762}
4763
4764void addrconf_cleanup(void)
4765{
4766 struct net_device *dev;
4767 int i;
4768
4769 unregister_netdevice_notifier(&ipv6_dev_notf);
4770 unregister_pernet_subsys(&addrconf_ops);
4771 ipv6_addr_label_cleanup();
4772
4773 rtnl_lock();
4774
4775 __rtnl_af_unregister(&inet6_ops);
4776
4777 /* clean dev list */
4778 for_each_netdev(&init_net, dev) {
4779 if (__in6_dev_get(dev) == NULL)
4780 continue;
4781 addrconf_ifdown(dev, 1);
4782 }
4783 addrconf_ifdown(init_net.loopback_dev, 2);
4784
4785 /*
4786 * Check hash table.
4787 */
4788 spin_lock_bh(&addrconf_hash_lock);
4789 for (i = 0; i < IN6_ADDR_HSIZE; i++)
4790 WARN_ON(!hlist_empty(&inet6_addr_lst[i]));
4791 spin_unlock_bh(&addrconf_hash_lock);
4792
4793 del_timer(&addr_chk_timer);
4794 rtnl_unlock();
4795}
1// SPDX-License-Identifier: GPL-2.0-or-later
2/*
3 * IPv6 Address [auto]configuration
4 * Linux INET6 implementation
5 *
6 * Authors:
7 * Pedro Roque <roque@di.fc.ul.pt>
8 * Alexey Kuznetsov <kuznet@ms2.inr.ac.ru>
9 */
10
11/*
12 * Changes:
13 *
14 * Janos Farkas : delete timer on ifdown
15 * <chexum@bankinf.banki.hu>
16 * Andi Kleen : kill double kfree on module
17 * unload.
18 * Maciej W. Rozycki : FDDI support
19 * sekiya@USAGI : Don't send too many RS
20 * packets.
21 * yoshfuji@USAGI : Fixed interval between DAD
22 * packets.
23 * YOSHIFUJI Hideaki @USAGI : improved accuracy of
24 * address validation timer.
25 * YOSHIFUJI Hideaki @USAGI : Privacy Extensions (RFC3041)
26 * support.
27 * Yuji SEKIYA @USAGI : Don't assign a same IPv6
28 * address on a same interface.
29 * YOSHIFUJI Hideaki @USAGI : ARCnet support
30 * YOSHIFUJI Hideaki @USAGI : convert /proc/net/if_inet6 to
31 * seq_file.
32 * YOSHIFUJI Hideaki @USAGI : improved source address
33 * selection; consider scope,
34 * status etc.
35 */
36
37#define pr_fmt(fmt) "IPv6: " fmt
38
39#include <linux/errno.h>
40#include <linux/types.h>
41#include <linux/kernel.h>
42#include <linux/sched/signal.h>
43#include <linux/socket.h>
44#include <linux/sockios.h>
45#include <linux/net.h>
46#include <linux/inet.h>
47#include <linux/in6.h>
48#include <linux/netdevice.h>
49#include <linux/if_addr.h>
50#include <linux/if_arp.h>
51#include <linux/if_arcnet.h>
52#include <linux/if_infiniband.h>
53#include <linux/route.h>
54#include <linux/inetdevice.h>
55#include <linux/init.h>
56#include <linux/slab.h>
57#ifdef CONFIG_SYSCTL
58#include <linux/sysctl.h>
59#endif
60#include <linux/capability.h>
61#include <linux/delay.h>
62#include <linux/notifier.h>
63#include <linux/string.h>
64#include <linux/hash.h>
65
66#include <net/net_namespace.h>
67#include <net/sock.h>
68#include <net/snmp.h>
69
70#include <net/6lowpan.h>
71#include <net/firewire.h>
72#include <net/ipv6.h>
73#include <net/protocol.h>
74#include <net/ndisc.h>
75#include <net/ip6_route.h>
76#include <net/addrconf.h>
77#include <net/tcp.h>
78#include <net/ip.h>
79#include <net/netlink.h>
80#include <net/pkt_sched.h>
81#include <net/l3mdev.h>
82#include <linux/if_tunnel.h>
83#include <linux/rtnetlink.h>
84#include <linux/netconf.h>
85#include <linux/random.h>
86#include <linux/uaccess.h>
87#include <asm/unaligned.h>
88
89#include <linux/proc_fs.h>
90#include <linux/seq_file.h>
91#include <linux/export.h>
92#include <linux/ioam6.h>
93
94#define INFINITY_LIFE_TIME 0xFFFFFFFF
95
96#define IPV6_MAX_STRLEN \
97 sizeof("ffff:ffff:ffff:ffff:ffff:ffff:255.255.255.255")
98
99static inline u32 cstamp_delta(unsigned long cstamp)
100{
101 return (cstamp - INITIAL_JIFFIES) * 100UL / HZ;
102}
103
104static inline s32 rfc3315_s14_backoff_init(s32 irt)
105{
106 /* multiply 'initial retransmission time' by 0.9 .. 1.1 */
107 u64 tmp = get_random_u32_inclusive(900000, 1100000) * (u64)irt;
108 do_div(tmp, 1000000);
109 return (s32)tmp;
110}
111
112static inline s32 rfc3315_s14_backoff_update(s32 rt, s32 mrt)
113{
114 /* multiply 'retransmission timeout' by 1.9 .. 2.1 */
115 u64 tmp = get_random_u32_inclusive(1900000, 2100000) * (u64)rt;
116 do_div(tmp, 1000000);
117 if ((s32)tmp > mrt) {
118 /* multiply 'maximum retransmission time' by 0.9 .. 1.1 */
119 tmp = get_random_u32_inclusive(900000, 1100000) * (u64)mrt;
120 do_div(tmp, 1000000);
121 }
122 return (s32)tmp;
123}
124
125#ifdef CONFIG_SYSCTL
126static int addrconf_sysctl_register(struct inet6_dev *idev);
127static void addrconf_sysctl_unregister(struct inet6_dev *idev);
128#else
129static inline int addrconf_sysctl_register(struct inet6_dev *idev)
130{
131 return 0;
132}
133
134static inline void addrconf_sysctl_unregister(struct inet6_dev *idev)
135{
136}
137#endif
138
139static void ipv6_gen_rnd_iid(struct in6_addr *addr);
140
141static int ipv6_generate_eui64(u8 *eui, struct net_device *dev);
142static int ipv6_count_addresses(const struct inet6_dev *idev);
143static int ipv6_generate_stable_address(struct in6_addr *addr,
144 u8 dad_count,
145 const struct inet6_dev *idev);
146
147#define IN6_ADDR_HSIZE_SHIFT 8
148#define IN6_ADDR_HSIZE (1 << IN6_ADDR_HSIZE_SHIFT)
149
150static void addrconf_verify(struct net *net);
151static void addrconf_verify_rtnl(struct net *net);
152
153static struct workqueue_struct *addrconf_wq;
154
155static void addrconf_join_anycast(struct inet6_ifaddr *ifp);
156static void addrconf_leave_anycast(struct inet6_ifaddr *ifp);
157
158static void addrconf_type_change(struct net_device *dev,
159 unsigned long event);
160static int addrconf_ifdown(struct net_device *dev, bool unregister);
161
162static struct fib6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
163 int plen,
164 const struct net_device *dev,
165 u32 flags, u32 noflags,
166 bool no_gw);
167
168static void addrconf_dad_start(struct inet6_ifaddr *ifp);
169static void addrconf_dad_work(struct work_struct *w);
170static void addrconf_dad_completed(struct inet6_ifaddr *ifp, bool bump_id,
171 bool send_na);
172static void addrconf_dad_run(struct inet6_dev *idev, bool restart);
173static void addrconf_rs_timer(struct timer_list *t);
174static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
175static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
176
177static void inet6_prefix_notify(int event, struct inet6_dev *idev,
178 struct prefix_info *pinfo);
179
180static struct ipv6_devconf ipv6_devconf __read_mostly = {
181 .forwarding = 0,
182 .hop_limit = IPV6_DEFAULT_HOPLIMIT,
183 .mtu6 = IPV6_MIN_MTU,
184 .accept_ra = 1,
185 .accept_redirects = 1,
186 .autoconf = 1,
187 .force_mld_version = 0,
188 .mldv1_unsolicited_report_interval = 10 * HZ,
189 .mldv2_unsolicited_report_interval = HZ,
190 .dad_transmits = 1,
191 .rtr_solicits = MAX_RTR_SOLICITATIONS,
192 .rtr_solicit_interval = RTR_SOLICITATION_INTERVAL,
193 .rtr_solicit_max_interval = RTR_SOLICITATION_MAX_INTERVAL,
194 .rtr_solicit_delay = MAX_RTR_SOLICITATION_DELAY,
195 .use_tempaddr = 0,
196 .temp_valid_lft = TEMP_VALID_LIFETIME,
197 .temp_prefered_lft = TEMP_PREFERRED_LIFETIME,
198 .regen_max_retry = REGEN_MAX_RETRY,
199 .max_desync_factor = MAX_DESYNC_FACTOR,
200 .max_addresses = IPV6_MAX_ADDRESSES,
201 .accept_ra_defrtr = 1,
202 .ra_defrtr_metric = IP6_RT_PRIO_USER,
203 .accept_ra_from_local = 0,
204 .accept_ra_min_hop_limit= 1,
205 .accept_ra_pinfo = 1,
206#ifdef CONFIG_IPV6_ROUTER_PREF
207 .accept_ra_rtr_pref = 1,
208 .rtr_probe_interval = 60 * HZ,
209#ifdef CONFIG_IPV6_ROUTE_INFO
210 .accept_ra_rt_info_min_plen = 0,
211 .accept_ra_rt_info_max_plen = 0,
212#endif
213#endif
214 .proxy_ndp = 0,
215 .accept_source_route = 0, /* we do not accept RH0 by default. */
216 .disable_ipv6 = 0,
217 .accept_dad = 0,
218 .suppress_frag_ndisc = 1,
219 .accept_ra_mtu = 1,
220 .stable_secret = {
221 .initialized = false,
222 },
223 .use_oif_addrs_only = 0,
224 .ignore_routes_with_linkdown = 0,
225 .keep_addr_on_down = 0,
226 .seg6_enabled = 0,
227#ifdef CONFIG_IPV6_SEG6_HMAC
228 .seg6_require_hmac = 0,
229#endif
230 .enhanced_dad = 1,
231 .addr_gen_mode = IN6_ADDR_GEN_MODE_EUI64,
232 .disable_policy = 0,
233 .rpl_seg_enabled = 0,
234 .ioam6_enabled = 0,
235 .ioam6_id = IOAM6_DEFAULT_IF_ID,
236 .ioam6_id_wide = IOAM6_DEFAULT_IF_ID_WIDE,
237 .ndisc_evict_nocarrier = 1,
238};
239
240static struct ipv6_devconf ipv6_devconf_dflt __read_mostly = {
241 .forwarding = 0,
242 .hop_limit = IPV6_DEFAULT_HOPLIMIT,
243 .mtu6 = IPV6_MIN_MTU,
244 .accept_ra = 1,
245 .accept_redirects = 1,
246 .autoconf = 1,
247 .force_mld_version = 0,
248 .mldv1_unsolicited_report_interval = 10 * HZ,
249 .mldv2_unsolicited_report_interval = HZ,
250 .dad_transmits = 1,
251 .rtr_solicits = MAX_RTR_SOLICITATIONS,
252 .rtr_solicit_interval = RTR_SOLICITATION_INTERVAL,
253 .rtr_solicit_max_interval = RTR_SOLICITATION_MAX_INTERVAL,
254 .rtr_solicit_delay = MAX_RTR_SOLICITATION_DELAY,
255 .use_tempaddr = 0,
256 .temp_valid_lft = TEMP_VALID_LIFETIME,
257 .temp_prefered_lft = TEMP_PREFERRED_LIFETIME,
258 .regen_max_retry = REGEN_MAX_RETRY,
259 .max_desync_factor = MAX_DESYNC_FACTOR,
260 .max_addresses = IPV6_MAX_ADDRESSES,
261 .accept_ra_defrtr = 1,
262 .ra_defrtr_metric = IP6_RT_PRIO_USER,
263 .accept_ra_from_local = 0,
264 .accept_ra_min_hop_limit= 1,
265 .accept_ra_pinfo = 1,
266#ifdef CONFIG_IPV6_ROUTER_PREF
267 .accept_ra_rtr_pref = 1,
268 .rtr_probe_interval = 60 * HZ,
269#ifdef CONFIG_IPV6_ROUTE_INFO
270 .accept_ra_rt_info_min_plen = 0,
271 .accept_ra_rt_info_max_plen = 0,
272#endif
273#endif
274 .proxy_ndp = 0,
275 .accept_source_route = 0, /* we do not accept RH0 by default. */
276 .disable_ipv6 = 0,
277 .accept_dad = 1,
278 .suppress_frag_ndisc = 1,
279 .accept_ra_mtu = 1,
280 .stable_secret = {
281 .initialized = false,
282 },
283 .use_oif_addrs_only = 0,
284 .ignore_routes_with_linkdown = 0,
285 .keep_addr_on_down = 0,
286 .seg6_enabled = 0,
287#ifdef CONFIG_IPV6_SEG6_HMAC
288 .seg6_require_hmac = 0,
289#endif
290 .enhanced_dad = 1,
291 .addr_gen_mode = IN6_ADDR_GEN_MODE_EUI64,
292 .disable_policy = 0,
293 .rpl_seg_enabled = 0,
294 .ioam6_enabled = 0,
295 .ioam6_id = IOAM6_DEFAULT_IF_ID,
296 .ioam6_id_wide = IOAM6_DEFAULT_IF_ID_WIDE,
297 .ndisc_evict_nocarrier = 1,
298};
299
300/* Check if link is ready: is it up and is a valid qdisc available */
301static inline bool addrconf_link_ready(const struct net_device *dev)
302{
303 return netif_oper_up(dev) && !qdisc_tx_is_noop(dev);
304}
305
306static void addrconf_del_rs_timer(struct inet6_dev *idev)
307{
308 if (del_timer(&idev->rs_timer))
309 __in6_dev_put(idev);
310}
311
312static void addrconf_del_dad_work(struct inet6_ifaddr *ifp)
313{
314 if (cancel_delayed_work(&ifp->dad_work))
315 __in6_ifa_put(ifp);
316}
317
318static void addrconf_mod_rs_timer(struct inet6_dev *idev,
319 unsigned long when)
320{
321 if (!timer_pending(&idev->rs_timer))
322 in6_dev_hold(idev);
323 mod_timer(&idev->rs_timer, jiffies + when);
324}
325
326static void addrconf_mod_dad_work(struct inet6_ifaddr *ifp,
327 unsigned long delay)
328{
329 in6_ifa_hold(ifp);
330 if (mod_delayed_work(addrconf_wq, &ifp->dad_work, delay))
331 in6_ifa_put(ifp);
332}
333
334static int snmp6_alloc_dev(struct inet6_dev *idev)
335{
336 int i;
337
338 idev->stats.ipv6 = alloc_percpu_gfp(struct ipstats_mib, GFP_KERNEL_ACCOUNT);
339 if (!idev->stats.ipv6)
340 goto err_ip;
341
342 for_each_possible_cpu(i) {
343 struct ipstats_mib *addrconf_stats;
344 addrconf_stats = per_cpu_ptr(idev->stats.ipv6, i);
345 u64_stats_init(&addrconf_stats->syncp);
346 }
347
348
349 idev->stats.icmpv6dev = kzalloc(sizeof(struct icmpv6_mib_device),
350 GFP_KERNEL);
351 if (!idev->stats.icmpv6dev)
352 goto err_icmp;
353 idev->stats.icmpv6msgdev = kzalloc(sizeof(struct icmpv6msg_mib_device),
354 GFP_KERNEL_ACCOUNT);
355 if (!idev->stats.icmpv6msgdev)
356 goto err_icmpmsg;
357
358 return 0;
359
360err_icmpmsg:
361 kfree(idev->stats.icmpv6dev);
362err_icmp:
363 free_percpu(idev->stats.ipv6);
364err_ip:
365 return -ENOMEM;
366}
367
368static struct inet6_dev *ipv6_add_dev(struct net_device *dev)
369{
370 struct inet6_dev *ndev;
371 int err = -ENOMEM;
372
373 ASSERT_RTNL();
374
375 if (dev->mtu < IPV6_MIN_MTU && dev != blackhole_netdev)
376 return ERR_PTR(-EINVAL);
377
378 ndev = kzalloc(sizeof(*ndev), GFP_KERNEL_ACCOUNT);
379 if (!ndev)
380 return ERR_PTR(err);
381
382 rwlock_init(&ndev->lock);
383 ndev->dev = dev;
384 INIT_LIST_HEAD(&ndev->addr_list);
385 timer_setup(&ndev->rs_timer, addrconf_rs_timer, 0);
386 memcpy(&ndev->cnf, dev_net(dev)->ipv6.devconf_dflt, sizeof(ndev->cnf));
387
388 if (ndev->cnf.stable_secret.initialized)
389 ndev->cnf.addr_gen_mode = IN6_ADDR_GEN_MODE_STABLE_PRIVACY;
390
391 ndev->cnf.mtu6 = dev->mtu;
392 ndev->ra_mtu = 0;
393 ndev->nd_parms = neigh_parms_alloc(dev, &nd_tbl);
394 if (!ndev->nd_parms) {
395 kfree(ndev);
396 return ERR_PTR(err);
397 }
398 if (ndev->cnf.forwarding)
399 dev_disable_lro(dev);
400 /* We refer to the device */
401 netdev_hold(dev, &ndev->dev_tracker, GFP_KERNEL);
402
403 if (snmp6_alloc_dev(ndev) < 0) {
404 netdev_dbg(dev, "%s: cannot allocate memory for statistics\n",
405 __func__);
406 neigh_parms_release(&nd_tbl, ndev->nd_parms);
407 netdev_put(dev, &ndev->dev_tracker);
408 kfree(ndev);
409 return ERR_PTR(err);
410 }
411
412 if (dev != blackhole_netdev) {
413 if (snmp6_register_dev(ndev) < 0) {
414 netdev_dbg(dev, "%s: cannot create /proc/net/dev_snmp6/%s\n",
415 __func__, dev->name);
416 goto err_release;
417 }
418 }
419 /* One reference from device. */
420 refcount_set(&ndev->refcnt, 1);
421
422 if (dev->flags & (IFF_NOARP | IFF_LOOPBACK))
423 ndev->cnf.accept_dad = -1;
424
425#if IS_ENABLED(CONFIG_IPV6_SIT)
426 if (dev->type == ARPHRD_SIT && (dev->priv_flags & IFF_ISATAP)) {
427 pr_info("%s: Disabled Multicast RS\n", dev->name);
428 ndev->cnf.rtr_solicits = 0;
429 }
430#endif
431
432 INIT_LIST_HEAD(&ndev->tempaddr_list);
433 ndev->desync_factor = U32_MAX;
434 if ((dev->flags&IFF_LOOPBACK) ||
435 dev->type == ARPHRD_TUNNEL ||
436 dev->type == ARPHRD_TUNNEL6 ||
437 dev->type == ARPHRD_SIT ||
438 dev->type == ARPHRD_NONE) {
439 ndev->cnf.use_tempaddr = -1;
440 }
441
442 ndev->token = in6addr_any;
443
444 if (netif_running(dev) && addrconf_link_ready(dev))
445 ndev->if_flags |= IF_READY;
446
447 ipv6_mc_init_dev(ndev);
448 ndev->tstamp = jiffies;
449 if (dev != blackhole_netdev) {
450 err = addrconf_sysctl_register(ndev);
451 if (err) {
452 ipv6_mc_destroy_dev(ndev);
453 snmp6_unregister_dev(ndev);
454 goto err_release;
455 }
456 }
457 /* protected by rtnl_lock */
458 rcu_assign_pointer(dev->ip6_ptr, ndev);
459
460 if (dev != blackhole_netdev) {
461 /* Join interface-local all-node multicast group */
462 ipv6_dev_mc_inc(dev, &in6addr_interfacelocal_allnodes);
463
464 /* Join all-node multicast group */
465 ipv6_dev_mc_inc(dev, &in6addr_linklocal_allnodes);
466
467 /* Join all-router multicast group if forwarding is set */
468 if (ndev->cnf.forwarding && (dev->flags & IFF_MULTICAST))
469 ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
470 }
471 return ndev;
472
473err_release:
474 neigh_parms_release(&nd_tbl, ndev->nd_parms);
475 ndev->dead = 1;
476 in6_dev_finish_destroy(ndev);
477 return ERR_PTR(err);
478}
479
480static struct inet6_dev *ipv6_find_idev(struct net_device *dev)
481{
482 struct inet6_dev *idev;
483
484 ASSERT_RTNL();
485
486 idev = __in6_dev_get(dev);
487 if (!idev) {
488 idev = ipv6_add_dev(dev);
489 if (IS_ERR(idev))
490 return idev;
491 }
492
493 if (dev->flags&IFF_UP)
494 ipv6_mc_up(idev);
495 return idev;
496}
497
498static int inet6_netconf_msgsize_devconf(int type)
499{
500 int size = NLMSG_ALIGN(sizeof(struct netconfmsg))
501 + nla_total_size(4); /* NETCONFA_IFINDEX */
502 bool all = false;
503
504 if (type == NETCONFA_ALL)
505 all = true;
506
507 if (all || type == NETCONFA_FORWARDING)
508 size += nla_total_size(4);
509#ifdef CONFIG_IPV6_MROUTE
510 if (all || type == NETCONFA_MC_FORWARDING)
511 size += nla_total_size(4);
512#endif
513 if (all || type == NETCONFA_PROXY_NEIGH)
514 size += nla_total_size(4);
515
516 if (all || type == NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN)
517 size += nla_total_size(4);
518
519 return size;
520}
521
522static int inet6_netconf_fill_devconf(struct sk_buff *skb, int ifindex,
523 struct ipv6_devconf *devconf, u32 portid,
524 u32 seq, int event, unsigned int flags,
525 int type)
526{
527 struct nlmsghdr *nlh;
528 struct netconfmsg *ncm;
529 bool all = false;
530
531 nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct netconfmsg),
532 flags);
533 if (!nlh)
534 return -EMSGSIZE;
535
536 if (type == NETCONFA_ALL)
537 all = true;
538
539 ncm = nlmsg_data(nlh);
540 ncm->ncm_family = AF_INET6;
541
542 if (nla_put_s32(skb, NETCONFA_IFINDEX, ifindex) < 0)
543 goto nla_put_failure;
544
545 if (!devconf)
546 goto out;
547
548 if ((all || type == NETCONFA_FORWARDING) &&
549 nla_put_s32(skb, NETCONFA_FORWARDING, devconf->forwarding) < 0)
550 goto nla_put_failure;
551#ifdef CONFIG_IPV6_MROUTE
552 if ((all || type == NETCONFA_MC_FORWARDING) &&
553 nla_put_s32(skb, NETCONFA_MC_FORWARDING,
554 atomic_read(&devconf->mc_forwarding)) < 0)
555 goto nla_put_failure;
556#endif
557 if ((all || type == NETCONFA_PROXY_NEIGH) &&
558 nla_put_s32(skb, NETCONFA_PROXY_NEIGH, devconf->proxy_ndp) < 0)
559 goto nla_put_failure;
560
561 if ((all || type == NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN) &&
562 nla_put_s32(skb, NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
563 devconf->ignore_routes_with_linkdown) < 0)
564 goto nla_put_failure;
565
566out:
567 nlmsg_end(skb, nlh);
568 return 0;
569
570nla_put_failure:
571 nlmsg_cancel(skb, nlh);
572 return -EMSGSIZE;
573}
574
575void inet6_netconf_notify_devconf(struct net *net, int event, int type,
576 int ifindex, struct ipv6_devconf *devconf)
577{
578 struct sk_buff *skb;
579 int err = -ENOBUFS;
580
581 skb = nlmsg_new(inet6_netconf_msgsize_devconf(type), GFP_KERNEL);
582 if (!skb)
583 goto errout;
584
585 err = inet6_netconf_fill_devconf(skb, ifindex, devconf, 0, 0,
586 event, 0, type);
587 if (err < 0) {
588 /* -EMSGSIZE implies BUG in inet6_netconf_msgsize_devconf() */
589 WARN_ON(err == -EMSGSIZE);
590 kfree_skb(skb);
591 goto errout;
592 }
593 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_NETCONF, NULL, GFP_KERNEL);
594 return;
595errout:
596 rtnl_set_sk_err(net, RTNLGRP_IPV6_NETCONF, err);
597}
598
599static const struct nla_policy devconf_ipv6_policy[NETCONFA_MAX+1] = {
600 [NETCONFA_IFINDEX] = { .len = sizeof(int) },
601 [NETCONFA_FORWARDING] = { .len = sizeof(int) },
602 [NETCONFA_PROXY_NEIGH] = { .len = sizeof(int) },
603 [NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN] = { .len = sizeof(int) },
604};
605
606static int inet6_netconf_valid_get_req(struct sk_buff *skb,
607 const struct nlmsghdr *nlh,
608 struct nlattr **tb,
609 struct netlink_ext_ack *extack)
610{
611 int i, err;
612
613 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(struct netconfmsg))) {
614 NL_SET_ERR_MSG_MOD(extack, "Invalid header for netconf get request");
615 return -EINVAL;
616 }
617
618 if (!netlink_strict_get_check(skb))
619 return nlmsg_parse_deprecated(nlh, sizeof(struct netconfmsg),
620 tb, NETCONFA_MAX,
621 devconf_ipv6_policy, extack);
622
623 err = nlmsg_parse_deprecated_strict(nlh, sizeof(struct netconfmsg),
624 tb, NETCONFA_MAX,
625 devconf_ipv6_policy, extack);
626 if (err)
627 return err;
628
629 for (i = 0; i <= NETCONFA_MAX; i++) {
630 if (!tb[i])
631 continue;
632
633 switch (i) {
634 case NETCONFA_IFINDEX:
635 break;
636 default:
637 NL_SET_ERR_MSG_MOD(extack, "Unsupported attribute in netconf get request");
638 return -EINVAL;
639 }
640 }
641
642 return 0;
643}
644
645static int inet6_netconf_get_devconf(struct sk_buff *in_skb,
646 struct nlmsghdr *nlh,
647 struct netlink_ext_ack *extack)
648{
649 struct net *net = sock_net(in_skb->sk);
650 struct nlattr *tb[NETCONFA_MAX+1];
651 struct inet6_dev *in6_dev = NULL;
652 struct net_device *dev = NULL;
653 struct sk_buff *skb;
654 struct ipv6_devconf *devconf;
655 int ifindex;
656 int err;
657
658 err = inet6_netconf_valid_get_req(in_skb, nlh, tb, extack);
659 if (err < 0)
660 return err;
661
662 if (!tb[NETCONFA_IFINDEX])
663 return -EINVAL;
664
665 err = -EINVAL;
666 ifindex = nla_get_s32(tb[NETCONFA_IFINDEX]);
667 switch (ifindex) {
668 case NETCONFA_IFINDEX_ALL:
669 devconf = net->ipv6.devconf_all;
670 break;
671 case NETCONFA_IFINDEX_DEFAULT:
672 devconf = net->ipv6.devconf_dflt;
673 break;
674 default:
675 dev = dev_get_by_index(net, ifindex);
676 if (!dev)
677 return -EINVAL;
678 in6_dev = in6_dev_get(dev);
679 if (!in6_dev)
680 goto errout;
681 devconf = &in6_dev->cnf;
682 break;
683 }
684
685 err = -ENOBUFS;
686 skb = nlmsg_new(inet6_netconf_msgsize_devconf(NETCONFA_ALL), GFP_KERNEL);
687 if (!skb)
688 goto errout;
689
690 err = inet6_netconf_fill_devconf(skb, ifindex, devconf,
691 NETLINK_CB(in_skb).portid,
692 nlh->nlmsg_seq, RTM_NEWNETCONF, 0,
693 NETCONFA_ALL);
694 if (err < 0) {
695 /* -EMSGSIZE implies BUG in inet6_netconf_msgsize_devconf() */
696 WARN_ON(err == -EMSGSIZE);
697 kfree_skb(skb);
698 goto errout;
699 }
700 err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
701errout:
702 if (in6_dev)
703 in6_dev_put(in6_dev);
704 dev_put(dev);
705 return err;
706}
707
708static int inet6_netconf_dump_devconf(struct sk_buff *skb,
709 struct netlink_callback *cb)
710{
711 const struct nlmsghdr *nlh = cb->nlh;
712 struct net *net = sock_net(skb->sk);
713 int h, s_h;
714 int idx, s_idx;
715 struct net_device *dev;
716 struct inet6_dev *idev;
717 struct hlist_head *head;
718
719 if (cb->strict_check) {
720 struct netlink_ext_ack *extack = cb->extack;
721 struct netconfmsg *ncm;
722
723 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ncm))) {
724 NL_SET_ERR_MSG_MOD(extack, "Invalid header for netconf dump request");
725 return -EINVAL;
726 }
727
728 if (nlmsg_attrlen(nlh, sizeof(*ncm))) {
729 NL_SET_ERR_MSG_MOD(extack, "Invalid data after header in netconf dump request");
730 return -EINVAL;
731 }
732 }
733
734 s_h = cb->args[0];
735 s_idx = idx = cb->args[1];
736
737 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
738 idx = 0;
739 head = &net->dev_index_head[h];
740 rcu_read_lock();
741 cb->seq = atomic_read(&net->ipv6.dev_addr_genid) ^
742 net->dev_base_seq;
743 hlist_for_each_entry_rcu(dev, head, index_hlist) {
744 if (idx < s_idx)
745 goto cont;
746 idev = __in6_dev_get(dev);
747 if (!idev)
748 goto cont;
749
750 if (inet6_netconf_fill_devconf(skb, dev->ifindex,
751 &idev->cnf,
752 NETLINK_CB(cb->skb).portid,
753 nlh->nlmsg_seq,
754 RTM_NEWNETCONF,
755 NLM_F_MULTI,
756 NETCONFA_ALL) < 0) {
757 rcu_read_unlock();
758 goto done;
759 }
760 nl_dump_check_consistent(cb, nlmsg_hdr(skb));
761cont:
762 idx++;
763 }
764 rcu_read_unlock();
765 }
766 if (h == NETDEV_HASHENTRIES) {
767 if (inet6_netconf_fill_devconf(skb, NETCONFA_IFINDEX_ALL,
768 net->ipv6.devconf_all,
769 NETLINK_CB(cb->skb).portid,
770 nlh->nlmsg_seq,
771 RTM_NEWNETCONF, NLM_F_MULTI,
772 NETCONFA_ALL) < 0)
773 goto done;
774 else
775 h++;
776 }
777 if (h == NETDEV_HASHENTRIES + 1) {
778 if (inet6_netconf_fill_devconf(skb, NETCONFA_IFINDEX_DEFAULT,
779 net->ipv6.devconf_dflt,
780 NETLINK_CB(cb->skb).portid,
781 nlh->nlmsg_seq,
782 RTM_NEWNETCONF, NLM_F_MULTI,
783 NETCONFA_ALL) < 0)
784 goto done;
785 else
786 h++;
787 }
788done:
789 cb->args[0] = h;
790 cb->args[1] = idx;
791
792 return skb->len;
793}
794
795#ifdef CONFIG_SYSCTL
796static void dev_forward_change(struct inet6_dev *idev)
797{
798 struct net_device *dev;
799 struct inet6_ifaddr *ifa;
800 LIST_HEAD(tmp_addr_list);
801
802 if (!idev)
803 return;
804 dev = idev->dev;
805 if (idev->cnf.forwarding)
806 dev_disable_lro(dev);
807 if (dev->flags & IFF_MULTICAST) {
808 if (idev->cnf.forwarding) {
809 ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
810 ipv6_dev_mc_inc(dev, &in6addr_interfacelocal_allrouters);
811 ipv6_dev_mc_inc(dev, &in6addr_sitelocal_allrouters);
812 } else {
813 ipv6_dev_mc_dec(dev, &in6addr_linklocal_allrouters);
814 ipv6_dev_mc_dec(dev, &in6addr_interfacelocal_allrouters);
815 ipv6_dev_mc_dec(dev, &in6addr_sitelocal_allrouters);
816 }
817 }
818
819 read_lock_bh(&idev->lock);
820 list_for_each_entry(ifa, &idev->addr_list, if_list) {
821 if (ifa->flags&IFA_F_TENTATIVE)
822 continue;
823 list_add_tail(&ifa->if_list_aux, &tmp_addr_list);
824 }
825 read_unlock_bh(&idev->lock);
826
827 while (!list_empty(&tmp_addr_list)) {
828 ifa = list_first_entry(&tmp_addr_list,
829 struct inet6_ifaddr, if_list_aux);
830 list_del(&ifa->if_list_aux);
831 if (idev->cnf.forwarding)
832 addrconf_join_anycast(ifa);
833 else
834 addrconf_leave_anycast(ifa);
835 }
836
837 inet6_netconf_notify_devconf(dev_net(dev), RTM_NEWNETCONF,
838 NETCONFA_FORWARDING,
839 dev->ifindex, &idev->cnf);
840}
841
842
843static void addrconf_forward_change(struct net *net, __s32 newf)
844{
845 struct net_device *dev;
846 struct inet6_dev *idev;
847
848 for_each_netdev(net, dev) {
849 idev = __in6_dev_get(dev);
850 if (idev) {
851 int changed = (!idev->cnf.forwarding) ^ (!newf);
852 idev->cnf.forwarding = newf;
853 if (changed)
854 dev_forward_change(idev);
855 }
856 }
857}
858
859static int addrconf_fixup_forwarding(struct ctl_table *table, int *p, int newf)
860{
861 struct net *net;
862 int old;
863
864 if (!rtnl_trylock())
865 return restart_syscall();
866
867 net = (struct net *)table->extra2;
868 old = *p;
869 *p = newf;
870
871 if (p == &net->ipv6.devconf_dflt->forwarding) {
872 if ((!newf) ^ (!old))
873 inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
874 NETCONFA_FORWARDING,
875 NETCONFA_IFINDEX_DEFAULT,
876 net->ipv6.devconf_dflt);
877 rtnl_unlock();
878 return 0;
879 }
880
881 if (p == &net->ipv6.devconf_all->forwarding) {
882 int old_dflt = net->ipv6.devconf_dflt->forwarding;
883
884 net->ipv6.devconf_dflt->forwarding = newf;
885 if ((!newf) ^ (!old_dflt))
886 inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
887 NETCONFA_FORWARDING,
888 NETCONFA_IFINDEX_DEFAULT,
889 net->ipv6.devconf_dflt);
890
891 addrconf_forward_change(net, newf);
892 if ((!newf) ^ (!old))
893 inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
894 NETCONFA_FORWARDING,
895 NETCONFA_IFINDEX_ALL,
896 net->ipv6.devconf_all);
897 } else if ((!newf) ^ (!old))
898 dev_forward_change((struct inet6_dev *)table->extra1);
899 rtnl_unlock();
900
901 if (newf)
902 rt6_purge_dflt_routers(net);
903 return 1;
904}
905
906static void addrconf_linkdown_change(struct net *net, __s32 newf)
907{
908 struct net_device *dev;
909 struct inet6_dev *idev;
910
911 for_each_netdev(net, dev) {
912 idev = __in6_dev_get(dev);
913 if (idev) {
914 int changed = (!idev->cnf.ignore_routes_with_linkdown) ^ (!newf);
915
916 idev->cnf.ignore_routes_with_linkdown = newf;
917 if (changed)
918 inet6_netconf_notify_devconf(dev_net(dev),
919 RTM_NEWNETCONF,
920 NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
921 dev->ifindex,
922 &idev->cnf);
923 }
924 }
925}
926
927static int addrconf_fixup_linkdown(struct ctl_table *table, int *p, int newf)
928{
929 struct net *net;
930 int old;
931
932 if (!rtnl_trylock())
933 return restart_syscall();
934
935 net = (struct net *)table->extra2;
936 old = *p;
937 *p = newf;
938
939 if (p == &net->ipv6.devconf_dflt->ignore_routes_with_linkdown) {
940 if ((!newf) ^ (!old))
941 inet6_netconf_notify_devconf(net,
942 RTM_NEWNETCONF,
943 NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
944 NETCONFA_IFINDEX_DEFAULT,
945 net->ipv6.devconf_dflt);
946 rtnl_unlock();
947 return 0;
948 }
949
950 if (p == &net->ipv6.devconf_all->ignore_routes_with_linkdown) {
951 net->ipv6.devconf_dflt->ignore_routes_with_linkdown = newf;
952 addrconf_linkdown_change(net, newf);
953 if ((!newf) ^ (!old))
954 inet6_netconf_notify_devconf(net,
955 RTM_NEWNETCONF,
956 NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
957 NETCONFA_IFINDEX_ALL,
958 net->ipv6.devconf_all);
959 }
960 rtnl_unlock();
961
962 return 1;
963}
964
965#endif
966
967/* Nobody refers to this ifaddr, destroy it */
968void inet6_ifa_finish_destroy(struct inet6_ifaddr *ifp)
969{
970 WARN_ON(!hlist_unhashed(&ifp->addr_lst));
971
972#ifdef NET_REFCNT_DEBUG
973 pr_debug("%s\n", __func__);
974#endif
975
976 in6_dev_put(ifp->idev);
977
978 if (cancel_delayed_work(&ifp->dad_work))
979 pr_notice("delayed DAD work was pending while freeing ifa=%p\n",
980 ifp);
981
982 if (ifp->state != INET6_IFADDR_STATE_DEAD) {
983 pr_warn("Freeing alive inet6 address %p\n", ifp);
984 return;
985 }
986
987 kfree_rcu(ifp, rcu);
988}
989
990static void
991ipv6_link_dev_addr(struct inet6_dev *idev, struct inet6_ifaddr *ifp)
992{
993 struct list_head *p;
994 int ifp_scope = ipv6_addr_src_scope(&ifp->addr);
995
996 /*
997 * Each device address list is sorted in order of scope -
998 * global before linklocal.
999 */
1000 list_for_each(p, &idev->addr_list) {
1001 struct inet6_ifaddr *ifa
1002 = list_entry(p, struct inet6_ifaddr, if_list);
1003 if (ifp_scope >= ipv6_addr_src_scope(&ifa->addr))
1004 break;
1005 }
1006
1007 list_add_tail_rcu(&ifp->if_list, p);
1008}
1009
1010static u32 inet6_addr_hash(const struct net *net, const struct in6_addr *addr)
1011{
1012 u32 val = ipv6_addr_hash(addr) ^ net_hash_mix(net);
1013
1014 return hash_32(val, IN6_ADDR_HSIZE_SHIFT);
1015}
1016
1017static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
1018 struct net_device *dev, unsigned int hash)
1019{
1020 struct inet6_ifaddr *ifp;
1021
1022 hlist_for_each_entry(ifp, &net->ipv6.inet6_addr_lst[hash], addr_lst) {
1023 if (ipv6_addr_equal(&ifp->addr, addr)) {
1024 if (!dev || ifp->idev->dev == dev)
1025 return true;
1026 }
1027 }
1028 return false;
1029}
1030
1031static int ipv6_add_addr_hash(struct net_device *dev, struct inet6_ifaddr *ifa)
1032{
1033 struct net *net = dev_net(dev);
1034 unsigned int hash = inet6_addr_hash(net, &ifa->addr);
1035 int err = 0;
1036
1037 spin_lock(&net->ipv6.addrconf_hash_lock);
1038
1039 /* Ignore adding duplicate addresses on an interface */
1040 if (ipv6_chk_same_addr(net, &ifa->addr, dev, hash)) {
1041 netdev_dbg(dev, "ipv6_add_addr: already assigned\n");
1042 err = -EEXIST;
1043 } else {
1044 hlist_add_head_rcu(&ifa->addr_lst, &net->ipv6.inet6_addr_lst[hash]);
1045 }
1046
1047 spin_unlock(&net->ipv6.addrconf_hash_lock);
1048
1049 return err;
1050}
1051
1052/* On success it returns ifp with increased reference count */
1053
1054static struct inet6_ifaddr *
1055ipv6_add_addr(struct inet6_dev *idev, struct ifa6_config *cfg,
1056 bool can_block, struct netlink_ext_ack *extack)
1057{
1058 gfp_t gfp_flags = can_block ? GFP_KERNEL : GFP_ATOMIC;
1059 int addr_type = ipv6_addr_type(cfg->pfx);
1060 struct net *net = dev_net(idev->dev);
1061 struct inet6_ifaddr *ifa = NULL;
1062 struct fib6_info *f6i = NULL;
1063 int err = 0;
1064
1065 if (addr_type == IPV6_ADDR_ANY ||
1066 (addr_type & IPV6_ADDR_MULTICAST &&
1067 !(cfg->ifa_flags & IFA_F_MCAUTOJOIN)) ||
1068 (!(idev->dev->flags & IFF_LOOPBACK) &&
1069 !netif_is_l3_master(idev->dev) &&
1070 addr_type & IPV6_ADDR_LOOPBACK))
1071 return ERR_PTR(-EADDRNOTAVAIL);
1072
1073 if (idev->dead) {
1074 err = -ENODEV; /*XXX*/
1075 goto out;
1076 }
1077
1078 if (idev->cnf.disable_ipv6) {
1079 err = -EACCES;
1080 goto out;
1081 }
1082
1083 /* validator notifier needs to be blocking;
1084 * do not call in atomic context
1085 */
1086 if (can_block) {
1087 struct in6_validator_info i6vi = {
1088 .i6vi_addr = *cfg->pfx,
1089 .i6vi_dev = idev,
1090 .extack = extack,
1091 };
1092
1093 err = inet6addr_validator_notifier_call_chain(NETDEV_UP, &i6vi);
1094 err = notifier_to_errno(err);
1095 if (err < 0)
1096 goto out;
1097 }
1098
1099 ifa = kzalloc(sizeof(*ifa), gfp_flags | __GFP_ACCOUNT);
1100 if (!ifa) {
1101 err = -ENOBUFS;
1102 goto out;
1103 }
1104
1105 f6i = addrconf_f6i_alloc(net, idev, cfg->pfx, false, gfp_flags);
1106 if (IS_ERR(f6i)) {
1107 err = PTR_ERR(f6i);
1108 f6i = NULL;
1109 goto out;
1110 }
1111
1112 neigh_parms_data_state_setall(idev->nd_parms);
1113
1114 ifa->addr = *cfg->pfx;
1115 if (cfg->peer_pfx)
1116 ifa->peer_addr = *cfg->peer_pfx;
1117
1118 spin_lock_init(&ifa->lock);
1119 INIT_DELAYED_WORK(&ifa->dad_work, addrconf_dad_work);
1120 INIT_HLIST_NODE(&ifa->addr_lst);
1121 ifa->scope = cfg->scope;
1122 ifa->prefix_len = cfg->plen;
1123 ifa->rt_priority = cfg->rt_priority;
1124 ifa->flags = cfg->ifa_flags;
1125 ifa->ifa_proto = cfg->ifa_proto;
1126 /* No need to add the TENTATIVE flag for addresses with NODAD */
1127 if (!(cfg->ifa_flags & IFA_F_NODAD))
1128 ifa->flags |= IFA_F_TENTATIVE;
1129 ifa->valid_lft = cfg->valid_lft;
1130 ifa->prefered_lft = cfg->preferred_lft;
1131 ifa->cstamp = ifa->tstamp = jiffies;
1132 ifa->tokenized = false;
1133
1134 ifa->rt = f6i;
1135
1136 ifa->idev = idev;
1137 in6_dev_hold(idev);
1138
1139 /* For caller */
1140 refcount_set(&ifa->refcnt, 1);
1141
1142 rcu_read_lock_bh();
1143
1144 err = ipv6_add_addr_hash(idev->dev, ifa);
1145 if (err < 0) {
1146 rcu_read_unlock_bh();
1147 goto out;
1148 }
1149
1150 write_lock(&idev->lock);
1151
1152 /* Add to inet6_dev unicast addr list. */
1153 ipv6_link_dev_addr(idev, ifa);
1154
1155 if (ifa->flags&IFA_F_TEMPORARY) {
1156 list_add(&ifa->tmp_list, &idev->tempaddr_list);
1157 in6_ifa_hold(ifa);
1158 }
1159
1160 in6_ifa_hold(ifa);
1161 write_unlock(&idev->lock);
1162
1163 rcu_read_unlock_bh();
1164
1165 inet6addr_notifier_call_chain(NETDEV_UP, ifa);
1166out:
1167 if (unlikely(err < 0)) {
1168 fib6_info_release(f6i);
1169
1170 if (ifa) {
1171 if (ifa->idev)
1172 in6_dev_put(ifa->idev);
1173 kfree(ifa);
1174 }
1175 ifa = ERR_PTR(err);
1176 }
1177
1178 return ifa;
1179}
1180
1181enum cleanup_prefix_rt_t {
1182 CLEANUP_PREFIX_RT_NOP, /* no cleanup action for prefix route */
1183 CLEANUP_PREFIX_RT_DEL, /* delete the prefix route */
1184 CLEANUP_PREFIX_RT_EXPIRE, /* update the lifetime of the prefix route */
1185};
1186
1187/*
1188 * Check, whether the prefix for ifp would still need a prefix route
1189 * after deleting ifp. The function returns one of the CLEANUP_PREFIX_RT_*
1190 * constants.
1191 *
1192 * 1) we don't purge prefix if address was not permanent.
1193 * prefix is managed by its own lifetime.
1194 * 2) we also don't purge, if the address was IFA_F_NOPREFIXROUTE.
1195 * 3) if there are no addresses, delete prefix.
1196 * 4) if there are still other permanent address(es),
1197 * corresponding prefix is still permanent.
1198 * 5) if there are still other addresses with IFA_F_NOPREFIXROUTE,
1199 * don't purge the prefix, assume user space is managing it.
1200 * 6) otherwise, update prefix lifetime to the
1201 * longest valid lifetime among the corresponding
1202 * addresses on the device.
1203 * Note: subsequent RA will update lifetime.
1204 **/
1205static enum cleanup_prefix_rt_t
1206check_cleanup_prefix_route(struct inet6_ifaddr *ifp, unsigned long *expires)
1207{
1208 struct inet6_ifaddr *ifa;
1209 struct inet6_dev *idev = ifp->idev;
1210 unsigned long lifetime;
1211 enum cleanup_prefix_rt_t action = CLEANUP_PREFIX_RT_DEL;
1212
1213 *expires = jiffies;
1214
1215 list_for_each_entry(ifa, &idev->addr_list, if_list) {
1216 if (ifa == ifp)
1217 continue;
1218 if (ifa->prefix_len != ifp->prefix_len ||
1219 !ipv6_prefix_equal(&ifa->addr, &ifp->addr,
1220 ifp->prefix_len))
1221 continue;
1222 if (ifa->flags & (IFA_F_PERMANENT | IFA_F_NOPREFIXROUTE))
1223 return CLEANUP_PREFIX_RT_NOP;
1224
1225 action = CLEANUP_PREFIX_RT_EXPIRE;
1226
1227 spin_lock(&ifa->lock);
1228
1229 lifetime = addrconf_timeout_fixup(ifa->valid_lft, HZ);
1230 /*
1231 * Note: Because this address is
1232 * not permanent, lifetime <
1233 * LONG_MAX / HZ here.
1234 */
1235 if (time_before(*expires, ifa->tstamp + lifetime * HZ))
1236 *expires = ifa->tstamp + lifetime * HZ;
1237 spin_unlock(&ifa->lock);
1238 }
1239
1240 return action;
1241}
1242
1243static void
1244cleanup_prefix_route(struct inet6_ifaddr *ifp, unsigned long expires,
1245 bool del_rt, bool del_peer)
1246{
1247 struct fib6_info *f6i;
1248
1249 f6i = addrconf_get_prefix_route(del_peer ? &ifp->peer_addr : &ifp->addr,
1250 ifp->prefix_len,
1251 ifp->idev->dev, 0, RTF_DEFAULT, true);
1252 if (f6i) {
1253 if (del_rt)
1254 ip6_del_rt(dev_net(ifp->idev->dev), f6i, false);
1255 else {
1256 if (!(f6i->fib6_flags & RTF_EXPIRES))
1257 fib6_set_expires(f6i, expires);
1258 fib6_info_release(f6i);
1259 }
1260 }
1261}
1262
1263
1264/* This function wants to get referenced ifp and releases it before return */
1265
1266static void ipv6_del_addr(struct inet6_ifaddr *ifp)
1267{
1268 enum cleanup_prefix_rt_t action = CLEANUP_PREFIX_RT_NOP;
1269 struct net *net = dev_net(ifp->idev->dev);
1270 unsigned long expires;
1271 int state;
1272
1273 ASSERT_RTNL();
1274
1275 spin_lock_bh(&ifp->lock);
1276 state = ifp->state;
1277 ifp->state = INET6_IFADDR_STATE_DEAD;
1278 spin_unlock_bh(&ifp->lock);
1279
1280 if (state == INET6_IFADDR_STATE_DEAD)
1281 goto out;
1282
1283 spin_lock_bh(&net->ipv6.addrconf_hash_lock);
1284 hlist_del_init_rcu(&ifp->addr_lst);
1285 spin_unlock_bh(&net->ipv6.addrconf_hash_lock);
1286
1287 write_lock_bh(&ifp->idev->lock);
1288
1289 if (ifp->flags&IFA_F_TEMPORARY) {
1290 list_del(&ifp->tmp_list);
1291 if (ifp->ifpub) {
1292 in6_ifa_put(ifp->ifpub);
1293 ifp->ifpub = NULL;
1294 }
1295 __in6_ifa_put(ifp);
1296 }
1297
1298 if (ifp->flags & IFA_F_PERMANENT && !(ifp->flags & IFA_F_NOPREFIXROUTE))
1299 action = check_cleanup_prefix_route(ifp, &expires);
1300
1301 list_del_rcu(&ifp->if_list);
1302 __in6_ifa_put(ifp);
1303
1304 write_unlock_bh(&ifp->idev->lock);
1305
1306 addrconf_del_dad_work(ifp);
1307
1308 ipv6_ifa_notify(RTM_DELADDR, ifp);
1309
1310 inet6addr_notifier_call_chain(NETDEV_DOWN, ifp);
1311
1312 if (action != CLEANUP_PREFIX_RT_NOP) {
1313 cleanup_prefix_route(ifp, expires,
1314 action == CLEANUP_PREFIX_RT_DEL, false);
1315 }
1316
1317 /* clean up prefsrc entries */
1318 rt6_remove_prefsrc(ifp);
1319out:
1320 in6_ifa_put(ifp);
1321}
1322
1323static int ipv6_create_tempaddr(struct inet6_ifaddr *ifp, bool block)
1324{
1325 struct inet6_dev *idev = ifp->idev;
1326 unsigned long tmp_tstamp, age;
1327 unsigned long regen_advance;
1328 unsigned long now = jiffies;
1329 s32 cnf_temp_preferred_lft;
1330 struct inet6_ifaddr *ift;
1331 struct ifa6_config cfg;
1332 long max_desync_factor;
1333 struct in6_addr addr;
1334 int ret = 0;
1335
1336 write_lock_bh(&idev->lock);
1337
1338retry:
1339 in6_dev_hold(idev);
1340 if (idev->cnf.use_tempaddr <= 0) {
1341 write_unlock_bh(&idev->lock);
1342 pr_info("%s: use_tempaddr is disabled\n", __func__);
1343 in6_dev_put(idev);
1344 ret = -1;
1345 goto out;
1346 }
1347 spin_lock_bh(&ifp->lock);
1348 if (ifp->regen_count++ >= idev->cnf.regen_max_retry) {
1349 idev->cnf.use_tempaddr = -1; /*XXX*/
1350 spin_unlock_bh(&ifp->lock);
1351 write_unlock_bh(&idev->lock);
1352 pr_warn("%s: regeneration time exceeded - disabled temporary address support\n",
1353 __func__);
1354 in6_dev_put(idev);
1355 ret = -1;
1356 goto out;
1357 }
1358 in6_ifa_hold(ifp);
1359 memcpy(addr.s6_addr, ifp->addr.s6_addr, 8);
1360 ipv6_gen_rnd_iid(&addr);
1361
1362 age = (now - ifp->tstamp) / HZ;
1363
1364 regen_advance = idev->cnf.regen_max_retry *
1365 idev->cnf.dad_transmits *
1366 max(NEIGH_VAR(idev->nd_parms, RETRANS_TIME), HZ/100) / HZ;
1367
1368 /* recalculate max_desync_factor each time and update
1369 * idev->desync_factor if it's larger
1370 */
1371 cnf_temp_preferred_lft = READ_ONCE(idev->cnf.temp_prefered_lft);
1372 max_desync_factor = min_t(__u32,
1373 idev->cnf.max_desync_factor,
1374 cnf_temp_preferred_lft - regen_advance);
1375
1376 if (unlikely(idev->desync_factor > max_desync_factor)) {
1377 if (max_desync_factor > 0) {
1378 get_random_bytes(&idev->desync_factor,
1379 sizeof(idev->desync_factor));
1380 idev->desync_factor %= max_desync_factor;
1381 } else {
1382 idev->desync_factor = 0;
1383 }
1384 }
1385
1386 memset(&cfg, 0, sizeof(cfg));
1387 cfg.valid_lft = min_t(__u32, ifp->valid_lft,
1388 idev->cnf.temp_valid_lft + age);
1389 cfg.preferred_lft = cnf_temp_preferred_lft + age - idev->desync_factor;
1390 cfg.preferred_lft = min_t(__u32, ifp->prefered_lft, cfg.preferred_lft);
1391
1392 cfg.plen = ifp->prefix_len;
1393 tmp_tstamp = ifp->tstamp;
1394 spin_unlock_bh(&ifp->lock);
1395
1396 write_unlock_bh(&idev->lock);
1397
1398 /* A temporary address is created only if this calculated Preferred
1399 * Lifetime is greater than REGEN_ADVANCE time units. In particular,
1400 * an implementation must not create a temporary address with a zero
1401 * Preferred Lifetime.
1402 * Use age calculation as in addrconf_verify to avoid unnecessary
1403 * temporary addresses being generated.
1404 */
1405 age = (now - tmp_tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
1406 if (cfg.preferred_lft <= regen_advance + age) {
1407 in6_ifa_put(ifp);
1408 in6_dev_put(idev);
1409 ret = -1;
1410 goto out;
1411 }
1412
1413 cfg.ifa_flags = IFA_F_TEMPORARY;
1414 /* set in addrconf_prefix_rcv() */
1415 if (ifp->flags & IFA_F_OPTIMISTIC)
1416 cfg.ifa_flags |= IFA_F_OPTIMISTIC;
1417
1418 cfg.pfx = &addr;
1419 cfg.scope = ipv6_addr_scope(cfg.pfx);
1420
1421 ift = ipv6_add_addr(idev, &cfg, block, NULL);
1422 if (IS_ERR(ift)) {
1423 in6_ifa_put(ifp);
1424 in6_dev_put(idev);
1425 pr_info("%s: retry temporary address regeneration\n", __func__);
1426 write_lock_bh(&idev->lock);
1427 goto retry;
1428 }
1429
1430 spin_lock_bh(&ift->lock);
1431 ift->ifpub = ifp;
1432 ift->cstamp = now;
1433 ift->tstamp = tmp_tstamp;
1434 spin_unlock_bh(&ift->lock);
1435
1436 addrconf_dad_start(ift);
1437 in6_ifa_put(ift);
1438 in6_dev_put(idev);
1439out:
1440 return ret;
1441}
1442
1443/*
1444 * Choose an appropriate source address (RFC3484)
1445 */
1446enum {
1447 IPV6_SADDR_RULE_INIT = 0,
1448 IPV6_SADDR_RULE_LOCAL,
1449 IPV6_SADDR_RULE_SCOPE,
1450 IPV6_SADDR_RULE_PREFERRED,
1451#ifdef CONFIG_IPV6_MIP6
1452 IPV6_SADDR_RULE_HOA,
1453#endif
1454 IPV6_SADDR_RULE_OIF,
1455 IPV6_SADDR_RULE_LABEL,
1456 IPV6_SADDR_RULE_PRIVACY,
1457 IPV6_SADDR_RULE_ORCHID,
1458 IPV6_SADDR_RULE_PREFIX,
1459#ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1460 IPV6_SADDR_RULE_NOT_OPTIMISTIC,
1461#endif
1462 IPV6_SADDR_RULE_MAX
1463};
1464
1465struct ipv6_saddr_score {
1466 int rule;
1467 int addr_type;
1468 struct inet6_ifaddr *ifa;
1469 DECLARE_BITMAP(scorebits, IPV6_SADDR_RULE_MAX);
1470 int scopedist;
1471 int matchlen;
1472};
1473
1474struct ipv6_saddr_dst {
1475 const struct in6_addr *addr;
1476 int ifindex;
1477 int scope;
1478 int label;
1479 unsigned int prefs;
1480};
1481
1482static inline int ipv6_saddr_preferred(int type)
1483{
1484 if (type & (IPV6_ADDR_MAPPED|IPV6_ADDR_COMPATv4|IPV6_ADDR_LOOPBACK))
1485 return 1;
1486 return 0;
1487}
1488
1489static bool ipv6_use_optimistic_addr(struct net *net,
1490 struct inet6_dev *idev)
1491{
1492#ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1493 if (!idev)
1494 return false;
1495 if (!net->ipv6.devconf_all->optimistic_dad && !idev->cnf.optimistic_dad)
1496 return false;
1497 if (!net->ipv6.devconf_all->use_optimistic && !idev->cnf.use_optimistic)
1498 return false;
1499
1500 return true;
1501#else
1502 return false;
1503#endif
1504}
1505
1506static bool ipv6_allow_optimistic_dad(struct net *net,
1507 struct inet6_dev *idev)
1508{
1509#ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1510 if (!idev)
1511 return false;
1512 if (!net->ipv6.devconf_all->optimistic_dad && !idev->cnf.optimistic_dad)
1513 return false;
1514
1515 return true;
1516#else
1517 return false;
1518#endif
1519}
1520
1521static int ipv6_get_saddr_eval(struct net *net,
1522 struct ipv6_saddr_score *score,
1523 struct ipv6_saddr_dst *dst,
1524 int i)
1525{
1526 int ret;
1527
1528 if (i <= score->rule) {
1529 switch (i) {
1530 case IPV6_SADDR_RULE_SCOPE:
1531 ret = score->scopedist;
1532 break;
1533 case IPV6_SADDR_RULE_PREFIX:
1534 ret = score->matchlen;
1535 break;
1536 default:
1537 ret = !!test_bit(i, score->scorebits);
1538 }
1539 goto out;
1540 }
1541
1542 switch (i) {
1543 case IPV6_SADDR_RULE_INIT:
1544 /* Rule 0: remember if hiscore is not ready yet */
1545 ret = !!score->ifa;
1546 break;
1547 case IPV6_SADDR_RULE_LOCAL:
1548 /* Rule 1: Prefer same address */
1549 ret = ipv6_addr_equal(&score->ifa->addr, dst->addr);
1550 break;
1551 case IPV6_SADDR_RULE_SCOPE:
1552 /* Rule 2: Prefer appropriate scope
1553 *
1554 * ret
1555 * ^
1556 * -1 | d 15
1557 * ---+--+-+---> scope
1558 * |
1559 * | d is scope of the destination.
1560 * B-d | \
1561 * | \ <- smaller scope is better if
1562 * B-15 | \ if scope is enough for destination.
1563 * | ret = B - scope (-1 <= scope >= d <= 15).
1564 * d-C-1 | /
1565 * |/ <- greater is better
1566 * -C / if scope is not enough for destination.
1567 * /| ret = scope - C (-1 <= d < scope <= 15).
1568 *
1569 * d - C - 1 < B -15 (for all -1 <= d <= 15).
1570 * C > d + 14 - B >= 15 + 14 - B = 29 - B.
1571 * Assume B = 0 and we get C > 29.
1572 */
1573 ret = __ipv6_addr_src_scope(score->addr_type);
1574 if (ret >= dst->scope)
1575 ret = -ret;
1576 else
1577 ret -= 128; /* 30 is enough */
1578 score->scopedist = ret;
1579 break;
1580 case IPV6_SADDR_RULE_PREFERRED:
1581 {
1582 /* Rule 3: Avoid deprecated and optimistic addresses */
1583 u8 avoid = IFA_F_DEPRECATED;
1584
1585 if (!ipv6_use_optimistic_addr(net, score->ifa->idev))
1586 avoid |= IFA_F_OPTIMISTIC;
1587 ret = ipv6_saddr_preferred(score->addr_type) ||
1588 !(score->ifa->flags & avoid);
1589 break;
1590 }
1591#ifdef CONFIG_IPV6_MIP6
1592 case IPV6_SADDR_RULE_HOA:
1593 {
1594 /* Rule 4: Prefer home address */
1595 int prefhome = !(dst->prefs & IPV6_PREFER_SRC_COA);
1596 ret = !(score->ifa->flags & IFA_F_HOMEADDRESS) ^ prefhome;
1597 break;
1598 }
1599#endif
1600 case IPV6_SADDR_RULE_OIF:
1601 /* Rule 5: Prefer outgoing interface */
1602 ret = (!dst->ifindex ||
1603 dst->ifindex == score->ifa->idev->dev->ifindex);
1604 break;
1605 case IPV6_SADDR_RULE_LABEL:
1606 /* Rule 6: Prefer matching label */
1607 ret = ipv6_addr_label(net,
1608 &score->ifa->addr, score->addr_type,
1609 score->ifa->idev->dev->ifindex) == dst->label;
1610 break;
1611 case IPV6_SADDR_RULE_PRIVACY:
1612 {
1613 /* Rule 7: Prefer public address
1614 * Note: prefer temporary address if use_tempaddr >= 2
1615 */
1616 int preftmp = dst->prefs & (IPV6_PREFER_SRC_PUBLIC|IPV6_PREFER_SRC_TMP) ?
1617 !!(dst->prefs & IPV6_PREFER_SRC_TMP) :
1618 score->ifa->idev->cnf.use_tempaddr >= 2;
1619 ret = (!(score->ifa->flags & IFA_F_TEMPORARY)) ^ preftmp;
1620 break;
1621 }
1622 case IPV6_SADDR_RULE_ORCHID:
1623 /* Rule 8-: Prefer ORCHID vs ORCHID or
1624 * non-ORCHID vs non-ORCHID
1625 */
1626 ret = !(ipv6_addr_orchid(&score->ifa->addr) ^
1627 ipv6_addr_orchid(dst->addr));
1628 break;
1629 case IPV6_SADDR_RULE_PREFIX:
1630 /* Rule 8: Use longest matching prefix */
1631 ret = ipv6_addr_diff(&score->ifa->addr, dst->addr);
1632 if (ret > score->ifa->prefix_len)
1633 ret = score->ifa->prefix_len;
1634 score->matchlen = ret;
1635 break;
1636#ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1637 case IPV6_SADDR_RULE_NOT_OPTIMISTIC:
1638 /* Optimistic addresses still have lower precedence than other
1639 * preferred addresses.
1640 */
1641 ret = !(score->ifa->flags & IFA_F_OPTIMISTIC);
1642 break;
1643#endif
1644 default:
1645 ret = 0;
1646 }
1647
1648 if (ret)
1649 __set_bit(i, score->scorebits);
1650 score->rule = i;
1651out:
1652 return ret;
1653}
1654
1655static int __ipv6_dev_get_saddr(struct net *net,
1656 struct ipv6_saddr_dst *dst,
1657 struct inet6_dev *idev,
1658 struct ipv6_saddr_score *scores,
1659 int hiscore_idx)
1660{
1661 struct ipv6_saddr_score *score = &scores[1 - hiscore_idx], *hiscore = &scores[hiscore_idx];
1662
1663 list_for_each_entry_rcu(score->ifa, &idev->addr_list, if_list) {
1664 int i;
1665
1666 /*
1667 * - Tentative Address (RFC2462 section 5.4)
1668 * - A tentative address is not considered
1669 * "assigned to an interface" in the traditional
1670 * sense, unless it is also flagged as optimistic.
1671 * - Candidate Source Address (section 4)
1672 * - In any case, anycast addresses, multicast
1673 * addresses, and the unspecified address MUST
1674 * NOT be included in a candidate set.
1675 */
1676 if ((score->ifa->flags & IFA_F_TENTATIVE) &&
1677 (!(score->ifa->flags & IFA_F_OPTIMISTIC)))
1678 continue;
1679
1680 score->addr_type = __ipv6_addr_type(&score->ifa->addr);
1681
1682 if (unlikely(score->addr_type == IPV6_ADDR_ANY ||
1683 score->addr_type & IPV6_ADDR_MULTICAST)) {
1684 net_dbg_ratelimited("ADDRCONF: unspecified / multicast address assigned as unicast address on %s",
1685 idev->dev->name);
1686 continue;
1687 }
1688
1689 score->rule = -1;
1690 bitmap_zero(score->scorebits, IPV6_SADDR_RULE_MAX);
1691
1692 for (i = 0; i < IPV6_SADDR_RULE_MAX; i++) {
1693 int minihiscore, miniscore;
1694
1695 minihiscore = ipv6_get_saddr_eval(net, hiscore, dst, i);
1696 miniscore = ipv6_get_saddr_eval(net, score, dst, i);
1697
1698 if (minihiscore > miniscore) {
1699 if (i == IPV6_SADDR_RULE_SCOPE &&
1700 score->scopedist > 0) {
1701 /*
1702 * special case:
1703 * each remaining entry
1704 * has too small (not enough)
1705 * scope, because ifa entries
1706 * are sorted by their scope
1707 * values.
1708 */
1709 goto out;
1710 }
1711 break;
1712 } else if (minihiscore < miniscore) {
1713 swap(hiscore, score);
1714 hiscore_idx = 1 - hiscore_idx;
1715
1716 /* restore our iterator */
1717 score->ifa = hiscore->ifa;
1718
1719 break;
1720 }
1721 }
1722 }
1723out:
1724 return hiscore_idx;
1725}
1726
1727static int ipv6_get_saddr_master(struct net *net,
1728 const struct net_device *dst_dev,
1729 const struct net_device *master,
1730 struct ipv6_saddr_dst *dst,
1731 struct ipv6_saddr_score *scores,
1732 int hiscore_idx)
1733{
1734 struct inet6_dev *idev;
1735
1736 idev = __in6_dev_get(dst_dev);
1737 if (idev)
1738 hiscore_idx = __ipv6_dev_get_saddr(net, dst, idev,
1739 scores, hiscore_idx);
1740
1741 idev = __in6_dev_get(master);
1742 if (idev)
1743 hiscore_idx = __ipv6_dev_get_saddr(net, dst, idev,
1744 scores, hiscore_idx);
1745
1746 return hiscore_idx;
1747}
1748
1749int ipv6_dev_get_saddr(struct net *net, const struct net_device *dst_dev,
1750 const struct in6_addr *daddr, unsigned int prefs,
1751 struct in6_addr *saddr)
1752{
1753 struct ipv6_saddr_score scores[2], *hiscore;
1754 struct ipv6_saddr_dst dst;
1755 struct inet6_dev *idev;
1756 struct net_device *dev;
1757 int dst_type;
1758 bool use_oif_addr = false;
1759 int hiscore_idx = 0;
1760 int ret = 0;
1761
1762 dst_type = __ipv6_addr_type(daddr);
1763 dst.addr = daddr;
1764 dst.ifindex = dst_dev ? dst_dev->ifindex : 0;
1765 dst.scope = __ipv6_addr_src_scope(dst_type);
1766 dst.label = ipv6_addr_label(net, daddr, dst_type, dst.ifindex);
1767 dst.prefs = prefs;
1768
1769 scores[hiscore_idx].rule = -1;
1770 scores[hiscore_idx].ifa = NULL;
1771
1772 rcu_read_lock();
1773
1774 /* Candidate Source Address (section 4)
1775 * - multicast and link-local destination address,
1776 * the set of candidate source address MUST only
1777 * include addresses assigned to interfaces
1778 * belonging to the same link as the outgoing
1779 * interface.
1780 * (- For site-local destination addresses, the
1781 * set of candidate source addresses MUST only
1782 * include addresses assigned to interfaces
1783 * belonging to the same site as the outgoing
1784 * interface.)
1785 * - "It is RECOMMENDED that the candidate source addresses
1786 * be the set of unicast addresses assigned to the
1787 * interface that will be used to send to the destination
1788 * (the 'outgoing' interface)." (RFC 6724)
1789 */
1790 if (dst_dev) {
1791 idev = __in6_dev_get(dst_dev);
1792 if ((dst_type & IPV6_ADDR_MULTICAST) ||
1793 dst.scope <= IPV6_ADDR_SCOPE_LINKLOCAL ||
1794 (idev && idev->cnf.use_oif_addrs_only)) {
1795 use_oif_addr = true;
1796 }
1797 }
1798
1799 if (use_oif_addr) {
1800 if (idev)
1801 hiscore_idx = __ipv6_dev_get_saddr(net, &dst, idev, scores, hiscore_idx);
1802 } else {
1803 const struct net_device *master;
1804 int master_idx = 0;
1805
1806 /* if dst_dev exists and is enslaved to an L3 device, then
1807 * prefer addresses from dst_dev and then the master over
1808 * any other enslaved devices in the L3 domain.
1809 */
1810 master = l3mdev_master_dev_rcu(dst_dev);
1811 if (master) {
1812 master_idx = master->ifindex;
1813
1814 hiscore_idx = ipv6_get_saddr_master(net, dst_dev,
1815 master, &dst,
1816 scores, hiscore_idx);
1817
1818 if (scores[hiscore_idx].ifa)
1819 goto out;
1820 }
1821
1822 for_each_netdev_rcu(net, dev) {
1823 /* only consider addresses on devices in the
1824 * same L3 domain
1825 */
1826 if (l3mdev_master_ifindex_rcu(dev) != master_idx)
1827 continue;
1828 idev = __in6_dev_get(dev);
1829 if (!idev)
1830 continue;
1831 hiscore_idx = __ipv6_dev_get_saddr(net, &dst, idev, scores, hiscore_idx);
1832 }
1833 }
1834
1835out:
1836 hiscore = &scores[hiscore_idx];
1837 if (!hiscore->ifa)
1838 ret = -EADDRNOTAVAIL;
1839 else
1840 *saddr = hiscore->ifa->addr;
1841
1842 rcu_read_unlock();
1843 return ret;
1844}
1845EXPORT_SYMBOL(ipv6_dev_get_saddr);
1846
1847static int __ipv6_get_lladdr(struct inet6_dev *idev, struct in6_addr *addr,
1848 u32 banned_flags)
1849{
1850 struct inet6_ifaddr *ifp;
1851 int err = -EADDRNOTAVAIL;
1852
1853 list_for_each_entry_reverse(ifp, &idev->addr_list, if_list) {
1854 if (ifp->scope > IFA_LINK)
1855 break;
1856 if (ifp->scope == IFA_LINK &&
1857 !(ifp->flags & banned_flags)) {
1858 *addr = ifp->addr;
1859 err = 0;
1860 break;
1861 }
1862 }
1863 return err;
1864}
1865
1866int ipv6_get_lladdr(struct net_device *dev, struct in6_addr *addr,
1867 u32 banned_flags)
1868{
1869 struct inet6_dev *idev;
1870 int err = -EADDRNOTAVAIL;
1871
1872 rcu_read_lock();
1873 idev = __in6_dev_get(dev);
1874 if (idev) {
1875 read_lock_bh(&idev->lock);
1876 err = __ipv6_get_lladdr(idev, addr, banned_flags);
1877 read_unlock_bh(&idev->lock);
1878 }
1879 rcu_read_unlock();
1880 return err;
1881}
1882
1883static int ipv6_count_addresses(const struct inet6_dev *idev)
1884{
1885 const struct inet6_ifaddr *ifp;
1886 int cnt = 0;
1887
1888 rcu_read_lock();
1889 list_for_each_entry_rcu(ifp, &idev->addr_list, if_list)
1890 cnt++;
1891 rcu_read_unlock();
1892 return cnt;
1893}
1894
1895int ipv6_chk_addr(struct net *net, const struct in6_addr *addr,
1896 const struct net_device *dev, int strict)
1897{
1898 return ipv6_chk_addr_and_flags(net, addr, dev, !dev,
1899 strict, IFA_F_TENTATIVE);
1900}
1901EXPORT_SYMBOL(ipv6_chk_addr);
1902
1903/* device argument is used to find the L3 domain of interest. If
1904 * skip_dev_check is set, then the ifp device is not checked against
1905 * the passed in dev argument. So the 2 cases for addresses checks are:
1906 * 1. does the address exist in the L3 domain that dev is part of
1907 * (skip_dev_check = true), or
1908 *
1909 * 2. does the address exist on the specific device
1910 * (skip_dev_check = false)
1911 */
1912static struct net_device *
1913__ipv6_chk_addr_and_flags(struct net *net, const struct in6_addr *addr,
1914 const struct net_device *dev, bool skip_dev_check,
1915 int strict, u32 banned_flags)
1916{
1917 unsigned int hash = inet6_addr_hash(net, addr);
1918 struct net_device *l3mdev, *ndev;
1919 struct inet6_ifaddr *ifp;
1920 u32 ifp_flags;
1921
1922 rcu_read_lock();
1923
1924 l3mdev = l3mdev_master_dev_rcu(dev);
1925 if (skip_dev_check)
1926 dev = NULL;
1927
1928 hlist_for_each_entry_rcu(ifp, &net->ipv6.inet6_addr_lst[hash], addr_lst) {
1929 ndev = ifp->idev->dev;
1930
1931 if (l3mdev_master_dev_rcu(ndev) != l3mdev)
1932 continue;
1933
1934 /* Decouple optimistic from tentative for evaluation here.
1935 * Ban optimistic addresses explicitly, when required.
1936 */
1937 ifp_flags = (ifp->flags&IFA_F_OPTIMISTIC)
1938 ? (ifp->flags&~IFA_F_TENTATIVE)
1939 : ifp->flags;
1940 if (ipv6_addr_equal(&ifp->addr, addr) &&
1941 !(ifp_flags&banned_flags) &&
1942 (!dev || ndev == dev ||
1943 !(ifp->scope&(IFA_LINK|IFA_HOST) || strict))) {
1944 rcu_read_unlock();
1945 return ndev;
1946 }
1947 }
1948
1949 rcu_read_unlock();
1950 return NULL;
1951}
1952
1953int ipv6_chk_addr_and_flags(struct net *net, const struct in6_addr *addr,
1954 const struct net_device *dev, bool skip_dev_check,
1955 int strict, u32 banned_flags)
1956{
1957 return __ipv6_chk_addr_and_flags(net, addr, dev, skip_dev_check,
1958 strict, banned_flags) ? 1 : 0;
1959}
1960EXPORT_SYMBOL(ipv6_chk_addr_and_flags);
1961
1962
1963/* Compares an address/prefix_len with addresses on device @dev.
1964 * If one is found it returns true.
1965 */
1966bool ipv6_chk_custom_prefix(const struct in6_addr *addr,
1967 const unsigned int prefix_len, struct net_device *dev)
1968{
1969 const struct inet6_ifaddr *ifa;
1970 const struct inet6_dev *idev;
1971 bool ret = false;
1972
1973 rcu_read_lock();
1974 idev = __in6_dev_get(dev);
1975 if (idev) {
1976 list_for_each_entry_rcu(ifa, &idev->addr_list, if_list) {
1977 ret = ipv6_prefix_equal(addr, &ifa->addr, prefix_len);
1978 if (ret)
1979 break;
1980 }
1981 }
1982 rcu_read_unlock();
1983
1984 return ret;
1985}
1986EXPORT_SYMBOL(ipv6_chk_custom_prefix);
1987
1988int ipv6_chk_prefix(const struct in6_addr *addr, struct net_device *dev)
1989{
1990 const struct inet6_ifaddr *ifa;
1991 const struct inet6_dev *idev;
1992 int onlink;
1993
1994 onlink = 0;
1995 rcu_read_lock();
1996 idev = __in6_dev_get(dev);
1997 if (idev) {
1998 list_for_each_entry_rcu(ifa, &idev->addr_list, if_list) {
1999 onlink = ipv6_prefix_equal(addr, &ifa->addr,
2000 ifa->prefix_len);
2001 if (onlink)
2002 break;
2003 }
2004 }
2005 rcu_read_unlock();
2006 return onlink;
2007}
2008EXPORT_SYMBOL(ipv6_chk_prefix);
2009
2010/**
2011 * ipv6_dev_find - find the first device with a given source address.
2012 * @net: the net namespace
2013 * @addr: the source address
2014 * @dev: used to find the L3 domain of interest
2015 *
2016 * The caller should be protected by RCU, or RTNL.
2017 */
2018struct net_device *ipv6_dev_find(struct net *net, const struct in6_addr *addr,
2019 struct net_device *dev)
2020{
2021 return __ipv6_chk_addr_and_flags(net, addr, dev, !dev, 1,
2022 IFA_F_TENTATIVE);
2023}
2024EXPORT_SYMBOL(ipv6_dev_find);
2025
2026struct inet6_ifaddr *ipv6_get_ifaddr(struct net *net, const struct in6_addr *addr,
2027 struct net_device *dev, int strict)
2028{
2029 unsigned int hash = inet6_addr_hash(net, addr);
2030 struct inet6_ifaddr *ifp, *result = NULL;
2031
2032 rcu_read_lock();
2033 hlist_for_each_entry_rcu(ifp, &net->ipv6.inet6_addr_lst[hash], addr_lst) {
2034 if (ipv6_addr_equal(&ifp->addr, addr)) {
2035 if (!dev || ifp->idev->dev == dev ||
2036 !(ifp->scope&(IFA_LINK|IFA_HOST) || strict)) {
2037 result = ifp;
2038 in6_ifa_hold(ifp);
2039 break;
2040 }
2041 }
2042 }
2043 rcu_read_unlock();
2044
2045 return result;
2046}
2047
2048/* Gets referenced address, destroys ifaddr */
2049
2050static void addrconf_dad_stop(struct inet6_ifaddr *ifp, int dad_failed)
2051{
2052 if (dad_failed)
2053 ifp->flags |= IFA_F_DADFAILED;
2054
2055 if (ifp->flags&IFA_F_TEMPORARY) {
2056 struct inet6_ifaddr *ifpub;
2057 spin_lock_bh(&ifp->lock);
2058 ifpub = ifp->ifpub;
2059 if (ifpub) {
2060 in6_ifa_hold(ifpub);
2061 spin_unlock_bh(&ifp->lock);
2062 ipv6_create_tempaddr(ifpub, true);
2063 in6_ifa_put(ifpub);
2064 } else {
2065 spin_unlock_bh(&ifp->lock);
2066 }
2067 ipv6_del_addr(ifp);
2068 } else if (ifp->flags&IFA_F_PERMANENT || !dad_failed) {
2069 spin_lock_bh(&ifp->lock);
2070 addrconf_del_dad_work(ifp);
2071 ifp->flags |= IFA_F_TENTATIVE;
2072 if (dad_failed)
2073 ifp->flags &= ~IFA_F_OPTIMISTIC;
2074 spin_unlock_bh(&ifp->lock);
2075 if (dad_failed)
2076 ipv6_ifa_notify(0, ifp);
2077 in6_ifa_put(ifp);
2078 } else {
2079 ipv6_del_addr(ifp);
2080 }
2081}
2082
2083static int addrconf_dad_end(struct inet6_ifaddr *ifp)
2084{
2085 int err = -ENOENT;
2086
2087 spin_lock_bh(&ifp->lock);
2088 if (ifp->state == INET6_IFADDR_STATE_DAD) {
2089 ifp->state = INET6_IFADDR_STATE_POSTDAD;
2090 err = 0;
2091 }
2092 spin_unlock_bh(&ifp->lock);
2093
2094 return err;
2095}
2096
2097void addrconf_dad_failure(struct sk_buff *skb, struct inet6_ifaddr *ifp)
2098{
2099 struct inet6_dev *idev = ifp->idev;
2100 struct net *net = dev_net(idev->dev);
2101
2102 if (addrconf_dad_end(ifp)) {
2103 in6_ifa_put(ifp);
2104 return;
2105 }
2106
2107 net_info_ratelimited("%s: IPv6 duplicate address %pI6c used by %pM detected!\n",
2108 ifp->idev->dev->name, &ifp->addr, eth_hdr(skb)->h_source);
2109
2110 spin_lock_bh(&ifp->lock);
2111
2112 if (ifp->flags & IFA_F_STABLE_PRIVACY) {
2113 struct in6_addr new_addr;
2114 struct inet6_ifaddr *ifp2;
2115 int retries = ifp->stable_privacy_retry + 1;
2116 struct ifa6_config cfg = {
2117 .pfx = &new_addr,
2118 .plen = ifp->prefix_len,
2119 .ifa_flags = ifp->flags,
2120 .valid_lft = ifp->valid_lft,
2121 .preferred_lft = ifp->prefered_lft,
2122 .scope = ifp->scope,
2123 };
2124
2125 if (retries > net->ipv6.sysctl.idgen_retries) {
2126 net_info_ratelimited("%s: privacy stable address generation failed because of DAD conflicts!\n",
2127 ifp->idev->dev->name);
2128 goto errdad;
2129 }
2130
2131 new_addr = ifp->addr;
2132 if (ipv6_generate_stable_address(&new_addr, retries,
2133 idev))
2134 goto errdad;
2135
2136 spin_unlock_bh(&ifp->lock);
2137
2138 if (idev->cnf.max_addresses &&
2139 ipv6_count_addresses(idev) >=
2140 idev->cnf.max_addresses)
2141 goto lock_errdad;
2142
2143 net_info_ratelimited("%s: generating new stable privacy address because of DAD conflict\n",
2144 ifp->idev->dev->name);
2145
2146 ifp2 = ipv6_add_addr(idev, &cfg, false, NULL);
2147 if (IS_ERR(ifp2))
2148 goto lock_errdad;
2149
2150 spin_lock_bh(&ifp2->lock);
2151 ifp2->stable_privacy_retry = retries;
2152 ifp2->state = INET6_IFADDR_STATE_PREDAD;
2153 spin_unlock_bh(&ifp2->lock);
2154
2155 addrconf_mod_dad_work(ifp2, net->ipv6.sysctl.idgen_delay);
2156 in6_ifa_put(ifp2);
2157lock_errdad:
2158 spin_lock_bh(&ifp->lock);
2159 }
2160
2161errdad:
2162 /* transition from _POSTDAD to _ERRDAD */
2163 ifp->state = INET6_IFADDR_STATE_ERRDAD;
2164 spin_unlock_bh(&ifp->lock);
2165
2166 addrconf_mod_dad_work(ifp, 0);
2167 in6_ifa_put(ifp);
2168}
2169
2170/* Join to solicited addr multicast group.
2171 * caller must hold RTNL */
2172void addrconf_join_solict(struct net_device *dev, const struct in6_addr *addr)
2173{
2174 struct in6_addr maddr;
2175
2176 if (dev->flags&(IFF_LOOPBACK|IFF_NOARP))
2177 return;
2178
2179 addrconf_addr_solict_mult(addr, &maddr);
2180 ipv6_dev_mc_inc(dev, &maddr);
2181}
2182
2183/* caller must hold RTNL */
2184void addrconf_leave_solict(struct inet6_dev *idev, const struct in6_addr *addr)
2185{
2186 struct in6_addr maddr;
2187
2188 if (idev->dev->flags&(IFF_LOOPBACK|IFF_NOARP))
2189 return;
2190
2191 addrconf_addr_solict_mult(addr, &maddr);
2192 __ipv6_dev_mc_dec(idev, &maddr);
2193}
2194
2195/* caller must hold RTNL */
2196static void addrconf_join_anycast(struct inet6_ifaddr *ifp)
2197{
2198 struct in6_addr addr;
2199
2200 if (ifp->prefix_len >= 127) /* RFC 6164 */
2201 return;
2202 ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
2203 if (ipv6_addr_any(&addr))
2204 return;
2205 __ipv6_dev_ac_inc(ifp->idev, &addr);
2206}
2207
2208/* caller must hold RTNL */
2209static void addrconf_leave_anycast(struct inet6_ifaddr *ifp)
2210{
2211 struct in6_addr addr;
2212
2213 if (ifp->prefix_len >= 127) /* RFC 6164 */
2214 return;
2215 ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
2216 if (ipv6_addr_any(&addr))
2217 return;
2218 __ipv6_dev_ac_dec(ifp->idev, &addr);
2219}
2220
2221static int addrconf_ifid_6lowpan(u8 *eui, struct net_device *dev)
2222{
2223 switch (dev->addr_len) {
2224 case ETH_ALEN:
2225 memcpy(eui, dev->dev_addr, 3);
2226 eui[3] = 0xFF;
2227 eui[4] = 0xFE;
2228 memcpy(eui + 5, dev->dev_addr + 3, 3);
2229 break;
2230 case EUI64_ADDR_LEN:
2231 memcpy(eui, dev->dev_addr, EUI64_ADDR_LEN);
2232 eui[0] ^= 2;
2233 break;
2234 default:
2235 return -1;
2236 }
2237
2238 return 0;
2239}
2240
2241static int addrconf_ifid_ieee1394(u8 *eui, struct net_device *dev)
2242{
2243 const union fwnet_hwaddr *ha;
2244
2245 if (dev->addr_len != FWNET_ALEN)
2246 return -1;
2247
2248 ha = (const union fwnet_hwaddr *)dev->dev_addr;
2249
2250 memcpy(eui, &ha->uc.uniq_id, sizeof(ha->uc.uniq_id));
2251 eui[0] ^= 2;
2252 return 0;
2253}
2254
2255static int addrconf_ifid_arcnet(u8 *eui, struct net_device *dev)
2256{
2257 /* XXX: inherit EUI-64 from other interface -- yoshfuji */
2258 if (dev->addr_len != ARCNET_ALEN)
2259 return -1;
2260 memset(eui, 0, 7);
2261 eui[7] = *(u8 *)dev->dev_addr;
2262 return 0;
2263}
2264
2265static int addrconf_ifid_infiniband(u8 *eui, struct net_device *dev)
2266{
2267 if (dev->addr_len != INFINIBAND_ALEN)
2268 return -1;
2269 memcpy(eui, dev->dev_addr + 12, 8);
2270 eui[0] |= 2;
2271 return 0;
2272}
2273
2274static int __ipv6_isatap_ifid(u8 *eui, __be32 addr)
2275{
2276 if (addr == 0)
2277 return -1;
2278 eui[0] = (ipv4_is_zeronet(addr) || ipv4_is_private_10(addr) ||
2279 ipv4_is_loopback(addr) || ipv4_is_linklocal_169(addr) ||
2280 ipv4_is_private_172(addr) || ipv4_is_test_192(addr) ||
2281 ipv4_is_anycast_6to4(addr) || ipv4_is_private_192(addr) ||
2282 ipv4_is_test_198(addr) || ipv4_is_multicast(addr) ||
2283 ipv4_is_lbcast(addr)) ? 0x00 : 0x02;
2284 eui[1] = 0;
2285 eui[2] = 0x5E;
2286 eui[3] = 0xFE;
2287 memcpy(eui + 4, &addr, 4);
2288 return 0;
2289}
2290
2291static int addrconf_ifid_sit(u8 *eui, struct net_device *dev)
2292{
2293 if (dev->priv_flags & IFF_ISATAP)
2294 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
2295 return -1;
2296}
2297
2298static int addrconf_ifid_gre(u8 *eui, struct net_device *dev)
2299{
2300 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
2301}
2302
2303static int addrconf_ifid_ip6tnl(u8 *eui, struct net_device *dev)
2304{
2305 memcpy(eui, dev->perm_addr, 3);
2306 memcpy(eui + 5, dev->perm_addr + 3, 3);
2307 eui[3] = 0xFF;
2308 eui[4] = 0xFE;
2309 eui[0] ^= 2;
2310 return 0;
2311}
2312
2313static int ipv6_generate_eui64(u8 *eui, struct net_device *dev)
2314{
2315 switch (dev->type) {
2316 case ARPHRD_ETHER:
2317 case ARPHRD_FDDI:
2318 return addrconf_ifid_eui48(eui, dev);
2319 case ARPHRD_ARCNET:
2320 return addrconf_ifid_arcnet(eui, dev);
2321 case ARPHRD_INFINIBAND:
2322 return addrconf_ifid_infiniband(eui, dev);
2323 case ARPHRD_SIT:
2324 return addrconf_ifid_sit(eui, dev);
2325 case ARPHRD_IPGRE:
2326 case ARPHRD_TUNNEL:
2327 return addrconf_ifid_gre(eui, dev);
2328 case ARPHRD_6LOWPAN:
2329 return addrconf_ifid_6lowpan(eui, dev);
2330 case ARPHRD_IEEE1394:
2331 return addrconf_ifid_ieee1394(eui, dev);
2332 case ARPHRD_TUNNEL6:
2333 case ARPHRD_IP6GRE:
2334 case ARPHRD_RAWIP:
2335 return addrconf_ifid_ip6tnl(eui, dev);
2336 }
2337 return -1;
2338}
2339
2340static int ipv6_inherit_eui64(u8 *eui, struct inet6_dev *idev)
2341{
2342 int err = -1;
2343 struct inet6_ifaddr *ifp;
2344
2345 read_lock_bh(&idev->lock);
2346 list_for_each_entry_reverse(ifp, &idev->addr_list, if_list) {
2347 if (ifp->scope > IFA_LINK)
2348 break;
2349 if (ifp->scope == IFA_LINK && !(ifp->flags&IFA_F_TENTATIVE)) {
2350 memcpy(eui, ifp->addr.s6_addr+8, 8);
2351 err = 0;
2352 break;
2353 }
2354 }
2355 read_unlock_bh(&idev->lock);
2356 return err;
2357}
2358
2359/* Generation of a randomized Interface Identifier
2360 * draft-ietf-6man-rfc4941bis, Section 3.3.1
2361 */
2362
2363static void ipv6_gen_rnd_iid(struct in6_addr *addr)
2364{
2365regen:
2366 get_random_bytes(&addr->s6_addr[8], 8);
2367
2368 /* <draft-ietf-6man-rfc4941bis-08.txt>, Section 3.3.1:
2369 * check if generated address is not inappropriate:
2370 *
2371 * - Reserved IPv6 Interface Identifiers
2372 * - XXX: already assigned to an address on the device
2373 */
2374
2375 /* Subnet-router anycast: 0000:0000:0000:0000 */
2376 if (!(addr->s6_addr32[2] | addr->s6_addr32[3]))
2377 goto regen;
2378
2379 /* IANA Ethernet block: 0200:5EFF:FE00:0000-0200:5EFF:FE00:5212
2380 * Proxy Mobile IPv6: 0200:5EFF:FE00:5213
2381 * IANA Ethernet block: 0200:5EFF:FE00:5214-0200:5EFF:FEFF:FFFF
2382 */
2383 if (ntohl(addr->s6_addr32[2]) == 0x02005eff &&
2384 (ntohl(addr->s6_addr32[3]) & 0Xff000000) == 0xfe000000)
2385 goto regen;
2386
2387 /* Reserved subnet anycast addresses */
2388 if (ntohl(addr->s6_addr32[2]) == 0xfdffffff &&
2389 ntohl(addr->s6_addr32[3]) >= 0Xffffff80)
2390 goto regen;
2391}
2392
2393/*
2394 * Add prefix route.
2395 */
2396
2397static void
2398addrconf_prefix_route(struct in6_addr *pfx, int plen, u32 metric,
2399 struct net_device *dev, unsigned long expires,
2400 u32 flags, gfp_t gfp_flags)
2401{
2402 struct fib6_config cfg = {
2403 .fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_PREFIX,
2404 .fc_metric = metric ? : IP6_RT_PRIO_ADDRCONF,
2405 .fc_ifindex = dev->ifindex,
2406 .fc_expires = expires,
2407 .fc_dst_len = plen,
2408 .fc_flags = RTF_UP | flags,
2409 .fc_nlinfo.nl_net = dev_net(dev),
2410 .fc_protocol = RTPROT_KERNEL,
2411 .fc_type = RTN_UNICAST,
2412 };
2413
2414 cfg.fc_dst = *pfx;
2415
2416 /* Prevent useless cloning on PtP SIT.
2417 This thing is done here expecting that the whole
2418 class of non-broadcast devices need not cloning.
2419 */
2420#if IS_ENABLED(CONFIG_IPV6_SIT)
2421 if (dev->type == ARPHRD_SIT && (dev->flags & IFF_POINTOPOINT))
2422 cfg.fc_flags |= RTF_NONEXTHOP;
2423#endif
2424
2425 ip6_route_add(&cfg, gfp_flags, NULL);
2426}
2427
2428
2429static struct fib6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
2430 int plen,
2431 const struct net_device *dev,
2432 u32 flags, u32 noflags,
2433 bool no_gw)
2434{
2435 struct fib6_node *fn;
2436 struct fib6_info *rt = NULL;
2437 struct fib6_table *table;
2438 u32 tb_id = l3mdev_fib_table(dev) ? : RT6_TABLE_PREFIX;
2439
2440 table = fib6_get_table(dev_net(dev), tb_id);
2441 if (!table)
2442 return NULL;
2443
2444 rcu_read_lock();
2445 fn = fib6_locate(&table->tb6_root, pfx, plen, NULL, 0, true);
2446 if (!fn)
2447 goto out;
2448
2449 for_each_fib6_node_rt_rcu(fn) {
2450 /* prefix routes only use builtin fib6_nh */
2451 if (rt->nh)
2452 continue;
2453
2454 if (rt->fib6_nh->fib_nh_dev->ifindex != dev->ifindex)
2455 continue;
2456 if (no_gw && rt->fib6_nh->fib_nh_gw_family)
2457 continue;
2458 if ((rt->fib6_flags & flags) != flags)
2459 continue;
2460 if ((rt->fib6_flags & noflags) != 0)
2461 continue;
2462 if (!fib6_info_hold_safe(rt))
2463 continue;
2464 break;
2465 }
2466out:
2467 rcu_read_unlock();
2468 return rt;
2469}
2470
2471
2472/* Create "default" multicast route to the interface */
2473
2474static void addrconf_add_mroute(struct net_device *dev)
2475{
2476 struct fib6_config cfg = {
2477 .fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_LOCAL,
2478 .fc_metric = IP6_RT_PRIO_ADDRCONF,
2479 .fc_ifindex = dev->ifindex,
2480 .fc_dst_len = 8,
2481 .fc_flags = RTF_UP,
2482 .fc_type = RTN_MULTICAST,
2483 .fc_nlinfo.nl_net = dev_net(dev),
2484 .fc_protocol = RTPROT_KERNEL,
2485 };
2486
2487 ipv6_addr_set(&cfg.fc_dst, htonl(0xFF000000), 0, 0, 0);
2488
2489 ip6_route_add(&cfg, GFP_KERNEL, NULL);
2490}
2491
2492static struct inet6_dev *addrconf_add_dev(struct net_device *dev)
2493{
2494 struct inet6_dev *idev;
2495
2496 ASSERT_RTNL();
2497
2498 idev = ipv6_find_idev(dev);
2499 if (IS_ERR(idev))
2500 return idev;
2501
2502 if (idev->cnf.disable_ipv6)
2503 return ERR_PTR(-EACCES);
2504
2505 /* Add default multicast route */
2506 if (!(dev->flags & IFF_LOOPBACK) && !netif_is_l3_master(dev))
2507 addrconf_add_mroute(dev);
2508
2509 return idev;
2510}
2511
2512static void manage_tempaddrs(struct inet6_dev *idev,
2513 struct inet6_ifaddr *ifp,
2514 __u32 valid_lft, __u32 prefered_lft,
2515 bool create, unsigned long now)
2516{
2517 u32 flags;
2518 struct inet6_ifaddr *ift;
2519
2520 read_lock_bh(&idev->lock);
2521 /* update all temporary addresses in the list */
2522 list_for_each_entry(ift, &idev->tempaddr_list, tmp_list) {
2523 int age, max_valid, max_prefered;
2524
2525 if (ifp != ift->ifpub)
2526 continue;
2527
2528 /* RFC 4941 section 3.3:
2529 * If a received option will extend the lifetime of a public
2530 * address, the lifetimes of temporary addresses should
2531 * be extended, subject to the overall constraint that no
2532 * temporary addresses should ever remain "valid" or "preferred"
2533 * for a time longer than (TEMP_VALID_LIFETIME) or
2534 * (TEMP_PREFERRED_LIFETIME - DESYNC_FACTOR), respectively.
2535 */
2536 age = (now - ift->cstamp) / HZ;
2537 max_valid = idev->cnf.temp_valid_lft - age;
2538 if (max_valid < 0)
2539 max_valid = 0;
2540
2541 max_prefered = idev->cnf.temp_prefered_lft -
2542 idev->desync_factor - age;
2543 if (max_prefered < 0)
2544 max_prefered = 0;
2545
2546 if (valid_lft > max_valid)
2547 valid_lft = max_valid;
2548
2549 if (prefered_lft > max_prefered)
2550 prefered_lft = max_prefered;
2551
2552 spin_lock(&ift->lock);
2553 flags = ift->flags;
2554 ift->valid_lft = valid_lft;
2555 ift->prefered_lft = prefered_lft;
2556 ift->tstamp = now;
2557 if (prefered_lft > 0)
2558 ift->flags &= ~IFA_F_DEPRECATED;
2559
2560 spin_unlock(&ift->lock);
2561 if (!(flags&IFA_F_TENTATIVE))
2562 ipv6_ifa_notify(0, ift);
2563 }
2564
2565 if ((create || list_empty(&idev->tempaddr_list)) &&
2566 idev->cnf.use_tempaddr > 0) {
2567 /* When a new public address is created as described
2568 * in [ADDRCONF], also create a new temporary address.
2569 * Also create a temporary address if it's enabled but
2570 * no temporary address currently exists.
2571 */
2572 read_unlock_bh(&idev->lock);
2573 ipv6_create_tempaddr(ifp, false);
2574 } else {
2575 read_unlock_bh(&idev->lock);
2576 }
2577}
2578
2579static bool is_addr_mode_generate_stable(struct inet6_dev *idev)
2580{
2581 return idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_STABLE_PRIVACY ||
2582 idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_RANDOM;
2583}
2584
2585int addrconf_prefix_rcv_add_addr(struct net *net, struct net_device *dev,
2586 const struct prefix_info *pinfo,
2587 struct inet6_dev *in6_dev,
2588 const struct in6_addr *addr, int addr_type,
2589 u32 addr_flags, bool sllao, bool tokenized,
2590 __u32 valid_lft, u32 prefered_lft)
2591{
2592 struct inet6_ifaddr *ifp = ipv6_get_ifaddr(net, addr, dev, 1);
2593 int create = 0, update_lft = 0;
2594
2595 if (!ifp && valid_lft) {
2596 int max_addresses = in6_dev->cnf.max_addresses;
2597 struct ifa6_config cfg = {
2598 .pfx = addr,
2599 .plen = pinfo->prefix_len,
2600 .ifa_flags = addr_flags,
2601 .valid_lft = valid_lft,
2602 .preferred_lft = prefered_lft,
2603 .scope = addr_type & IPV6_ADDR_SCOPE_MASK,
2604 .ifa_proto = IFAPROT_KERNEL_RA
2605 };
2606
2607#ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2608 if ((net->ipv6.devconf_all->optimistic_dad ||
2609 in6_dev->cnf.optimistic_dad) &&
2610 !net->ipv6.devconf_all->forwarding && sllao)
2611 cfg.ifa_flags |= IFA_F_OPTIMISTIC;
2612#endif
2613
2614 /* Do not allow to create too much of autoconfigured
2615 * addresses; this would be too easy way to crash kernel.
2616 */
2617 if (!max_addresses ||
2618 ipv6_count_addresses(in6_dev) < max_addresses)
2619 ifp = ipv6_add_addr(in6_dev, &cfg, false, NULL);
2620
2621 if (IS_ERR_OR_NULL(ifp))
2622 return -1;
2623
2624 create = 1;
2625 spin_lock_bh(&ifp->lock);
2626 ifp->flags |= IFA_F_MANAGETEMPADDR;
2627 ifp->cstamp = jiffies;
2628 ifp->tokenized = tokenized;
2629 spin_unlock_bh(&ifp->lock);
2630 addrconf_dad_start(ifp);
2631 }
2632
2633 if (ifp) {
2634 u32 flags;
2635 unsigned long now;
2636 u32 stored_lft;
2637
2638 /* update lifetime (RFC2462 5.5.3 e) */
2639 spin_lock_bh(&ifp->lock);
2640 now = jiffies;
2641 if (ifp->valid_lft > (now - ifp->tstamp) / HZ)
2642 stored_lft = ifp->valid_lft - (now - ifp->tstamp) / HZ;
2643 else
2644 stored_lft = 0;
2645 if (!create && stored_lft) {
2646 const u32 minimum_lft = min_t(u32,
2647 stored_lft, MIN_VALID_LIFETIME);
2648 valid_lft = max(valid_lft, minimum_lft);
2649
2650 /* RFC4862 Section 5.5.3e:
2651 * "Note that the preferred lifetime of the
2652 * corresponding address is always reset to
2653 * the Preferred Lifetime in the received
2654 * Prefix Information option, regardless of
2655 * whether the valid lifetime is also reset or
2656 * ignored."
2657 *
2658 * So we should always update prefered_lft here.
2659 */
2660 update_lft = 1;
2661 }
2662
2663 if (update_lft) {
2664 ifp->valid_lft = valid_lft;
2665 ifp->prefered_lft = prefered_lft;
2666 ifp->tstamp = now;
2667 flags = ifp->flags;
2668 ifp->flags &= ~IFA_F_DEPRECATED;
2669 spin_unlock_bh(&ifp->lock);
2670
2671 if (!(flags&IFA_F_TENTATIVE))
2672 ipv6_ifa_notify(0, ifp);
2673 } else
2674 spin_unlock_bh(&ifp->lock);
2675
2676 manage_tempaddrs(in6_dev, ifp, valid_lft, prefered_lft,
2677 create, now);
2678
2679 in6_ifa_put(ifp);
2680 addrconf_verify(net);
2681 }
2682
2683 return 0;
2684}
2685EXPORT_SYMBOL_GPL(addrconf_prefix_rcv_add_addr);
2686
2687void addrconf_prefix_rcv(struct net_device *dev, u8 *opt, int len, bool sllao)
2688{
2689 struct prefix_info *pinfo;
2690 __u32 valid_lft;
2691 __u32 prefered_lft;
2692 int addr_type, err;
2693 u32 addr_flags = 0;
2694 struct inet6_dev *in6_dev;
2695 struct net *net = dev_net(dev);
2696
2697 pinfo = (struct prefix_info *) opt;
2698
2699 if (len < sizeof(struct prefix_info)) {
2700 netdev_dbg(dev, "addrconf: prefix option too short\n");
2701 return;
2702 }
2703
2704 /*
2705 * Validation checks ([ADDRCONF], page 19)
2706 */
2707
2708 addr_type = ipv6_addr_type(&pinfo->prefix);
2709
2710 if (addr_type & (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL))
2711 return;
2712
2713 valid_lft = ntohl(pinfo->valid);
2714 prefered_lft = ntohl(pinfo->prefered);
2715
2716 if (prefered_lft > valid_lft) {
2717 net_warn_ratelimited("addrconf: prefix option has invalid lifetime\n");
2718 return;
2719 }
2720
2721 in6_dev = in6_dev_get(dev);
2722
2723 if (!in6_dev) {
2724 net_dbg_ratelimited("addrconf: device %s not configured\n",
2725 dev->name);
2726 return;
2727 }
2728
2729 /*
2730 * Two things going on here:
2731 * 1) Add routes for on-link prefixes
2732 * 2) Configure prefixes with the auto flag set
2733 */
2734
2735 if (pinfo->onlink) {
2736 struct fib6_info *rt;
2737 unsigned long rt_expires;
2738
2739 /* Avoid arithmetic overflow. Really, we could
2740 * save rt_expires in seconds, likely valid_lft,
2741 * but it would require division in fib gc, that it
2742 * not good.
2743 */
2744 if (HZ > USER_HZ)
2745 rt_expires = addrconf_timeout_fixup(valid_lft, HZ);
2746 else
2747 rt_expires = addrconf_timeout_fixup(valid_lft, USER_HZ);
2748
2749 if (addrconf_finite_timeout(rt_expires))
2750 rt_expires *= HZ;
2751
2752 rt = addrconf_get_prefix_route(&pinfo->prefix,
2753 pinfo->prefix_len,
2754 dev,
2755 RTF_ADDRCONF | RTF_PREFIX_RT,
2756 RTF_DEFAULT, true);
2757
2758 if (rt) {
2759 /* Autoconf prefix route */
2760 if (valid_lft == 0) {
2761 ip6_del_rt(net, rt, false);
2762 rt = NULL;
2763 } else if (addrconf_finite_timeout(rt_expires)) {
2764 /* not infinity */
2765 fib6_set_expires(rt, jiffies + rt_expires);
2766 } else {
2767 fib6_clean_expires(rt);
2768 }
2769 } else if (valid_lft) {
2770 clock_t expires = 0;
2771 int flags = RTF_ADDRCONF | RTF_PREFIX_RT;
2772 if (addrconf_finite_timeout(rt_expires)) {
2773 /* not infinity */
2774 flags |= RTF_EXPIRES;
2775 expires = jiffies_to_clock_t(rt_expires);
2776 }
2777 addrconf_prefix_route(&pinfo->prefix, pinfo->prefix_len,
2778 0, dev, expires, flags,
2779 GFP_ATOMIC);
2780 }
2781 fib6_info_release(rt);
2782 }
2783
2784 /* Try to figure out our local address for this prefix */
2785
2786 if (pinfo->autoconf && in6_dev->cnf.autoconf) {
2787 struct in6_addr addr;
2788 bool tokenized = false, dev_addr_generated = false;
2789
2790 if (pinfo->prefix_len == 64) {
2791 memcpy(&addr, &pinfo->prefix, 8);
2792
2793 if (!ipv6_addr_any(&in6_dev->token)) {
2794 read_lock_bh(&in6_dev->lock);
2795 memcpy(addr.s6_addr + 8,
2796 in6_dev->token.s6_addr + 8, 8);
2797 read_unlock_bh(&in6_dev->lock);
2798 tokenized = true;
2799 } else if (is_addr_mode_generate_stable(in6_dev) &&
2800 !ipv6_generate_stable_address(&addr, 0,
2801 in6_dev)) {
2802 addr_flags |= IFA_F_STABLE_PRIVACY;
2803 goto ok;
2804 } else if (ipv6_generate_eui64(addr.s6_addr + 8, dev) &&
2805 ipv6_inherit_eui64(addr.s6_addr + 8, in6_dev)) {
2806 goto put;
2807 } else {
2808 dev_addr_generated = true;
2809 }
2810 goto ok;
2811 }
2812 net_dbg_ratelimited("IPv6 addrconf: prefix with wrong length %d\n",
2813 pinfo->prefix_len);
2814 goto put;
2815
2816ok:
2817 err = addrconf_prefix_rcv_add_addr(net, dev, pinfo, in6_dev,
2818 &addr, addr_type,
2819 addr_flags, sllao,
2820 tokenized, valid_lft,
2821 prefered_lft);
2822 if (err)
2823 goto put;
2824
2825 /* Ignore error case here because previous prefix add addr was
2826 * successful which will be notified.
2827 */
2828 ndisc_ops_prefix_rcv_add_addr(net, dev, pinfo, in6_dev, &addr,
2829 addr_type, addr_flags, sllao,
2830 tokenized, valid_lft,
2831 prefered_lft,
2832 dev_addr_generated);
2833 }
2834 inet6_prefix_notify(RTM_NEWPREFIX, in6_dev, pinfo);
2835put:
2836 in6_dev_put(in6_dev);
2837}
2838
2839static int addrconf_set_sit_dstaddr(struct net *net, struct net_device *dev,
2840 struct in6_ifreq *ireq)
2841{
2842 struct ip_tunnel_parm p = { };
2843 int err;
2844
2845 if (!(ipv6_addr_type(&ireq->ifr6_addr) & IPV6_ADDR_COMPATv4))
2846 return -EADDRNOTAVAIL;
2847
2848 p.iph.daddr = ireq->ifr6_addr.s6_addr32[3];
2849 p.iph.version = 4;
2850 p.iph.ihl = 5;
2851 p.iph.protocol = IPPROTO_IPV6;
2852 p.iph.ttl = 64;
2853
2854 if (!dev->netdev_ops->ndo_tunnel_ctl)
2855 return -EOPNOTSUPP;
2856 err = dev->netdev_ops->ndo_tunnel_ctl(dev, &p, SIOCADDTUNNEL);
2857 if (err)
2858 return err;
2859
2860 dev = __dev_get_by_name(net, p.name);
2861 if (!dev)
2862 return -ENOBUFS;
2863 return dev_open(dev, NULL);
2864}
2865
2866/*
2867 * Set destination address.
2868 * Special case for SIT interfaces where we create a new "virtual"
2869 * device.
2870 */
2871int addrconf_set_dstaddr(struct net *net, void __user *arg)
2872{
2873 struct net_device *dev;
2874 struct in6_ifreq ireq;
2875 int err = -ENODEV;
2876
2877 if (!IS_ENABLED(CONFIG_IPV6_SIT))
2878 return -ENODEV;
2879 if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2880 return -EFAULT;
2881
2882 rtnl_lock();
2883 dev = __dev_get_by_index(net, ireq.ifr6_ifindex);
2884 if (dev && dev->type == ARPHRD_SIT)
2885 err = addrconf_set_sit_dstaddr(net, dev, &ireq);
2886 rtnl_unlock();
2887 return err;
2888}
2889
2890static int ipv6_mc_config(struct sock *sk, bool join,
2891 const struct in6_addr *addr, int ifindex)
2892{
2893 int ret;
2894
2895 ASSERT_RTNL();
2896
2897 lock_sock(sk);
2898 if (join)
2899 ret = ipv6_sock_mc_join(sk, ifindex, addr);
2900 else
2901 ret = ipv6_sock_mc_drop(sk, ifindex, addr);
2902 release_sock(sk);
2903
2904 return ret;
2905}
2906
2907/*
2908 * Manual configuration of address on an interface
2909 */
2910static int inet6_addr_add(struct net *net, int ifindex,
2911 struct ifa6_config *cfg,
2912 struct netlink_ext_ack *extack)
2913{
2914 struct inet6_ifaddr *ifp;
2915 struct inet6_dev *idev;
2916 struct net_device *dev;
2917 unsigned long timeout;
2918 clock_t expires;
2919 u32 flags;
2920
2921 ASSERT_RTNL();
2922
2923 if (cfg->plen > 128)
2924 return -EINVAL;
2925
2926 /* check the lifetime */
2927 if (!cfg->valid_lft || cfg->preferred_lft > cfg->valid_lft)
2928 return -EINVAL;
2929
2930 if (cfg->ifa_flags & IFA_F_MANAGETEMPADDR && cfg->plen != 64)
2931 return -EINVAL;
2932
2933 dev = __dev_get_by_index(net, ifindex);
2934 if (!dev)
2935 return -ENODEV;
2936
2937 idev = addrconf_add_dev(dev);
2938 if (IS_ERR(idev))
2939 return PTR_ERR(idev);
2940
2941 if (cfg->ifa_flags & IFA_F_MCAUTOJOIN) {
2942 int ret = ipv6_mc_config(net->ipv6.mc_autojoin_sk,
2943 true, cfg->pfx, ifindex);
2944
2945 if (ret < 0)
2946 return ret;
2947 }
2948
2949 cfg->scope = ipv6_addr_scope(cfg->pfx);
2950
2951 timeout = addrconf_timeout_fixup(cfg->valid_lft, HZ);
2952 if (addrconf_finite_timeout(timeout)) {
2953 expires = jiffies_to_clock_t(timeout * HZ);
2954 cfg->valid_lft = timeout;
2955 flags = RTF_EXPIRES;
2956 } else {
2957 expires = 0;
2958 flags = 0;
2959 cfg->ifa_flags |= IFA_F_PERMANENT;
2960 }
2961
2962 timeout = addrconf_timeout_fixup(cfg->preferred_lft, HZ);
2963 if (addrconf_finite_timeout(timeout)) {
2964 if (timeout == 0)
2965 cfg->ifa_flags |= IFA_F_DEPRECATED;
2966 cfg->preferred_lft = timeout;
2967 }
2968
2969 ifp = ipv6_add_addr(idev, cfg, true, extack);
2970 if (!IS_ERR(ifp)) {
2971 if (!(cfg->ifa_flags & IFA_F_NOPREFIXROUTE)) {
2972 addrconf_prefix_route(&ifp->addr, ifp->prefix_len,
2973 ifp->rt_priority, dev, expires,
2974 flags, GFP_KERNEL);
2975 }
2976
2977 /* Send a netlink notification if DAD is enabled and
2978 * optimistic flag is not set
2979 */
2980 if (!(ifp->flags & (IFA_F_OPTIMISTIC | IFA_F_NODAD)))
2981 ipv6_ifa_notify(0, ifp);
2982 /*
2983 * Note that section 3.1 of RFC 4429 indicates
2984 * that the Optimistic flag should not be set for
2985 * manually configured addresses
2986 */
2987 addrconf_dad_start(ifp);
2988 if (cfg->ifa_flags & IFA_F_MANAGETEMPADDR)
2989 manage_tempaddrs(idev, ifp, cfg->valid_lft,
2990 cfg->preferred_lft, true, jiffies);
2991 in6_ifa_put(ifp);
2992 addrconf_verify_rtnl(net);
2993 return 0;
2994 } else if (cfg->ifa_flags & IFA_F_MCAUTOJOIN) {
2995 ipv6_mc_config(net->ipv6.mc_autojoin_sk, false,
2996 cfg->pfx, ifindex);
2997 }
2998
2999 return PTR_ERR(ifp);
3000}
3001
3002static int inet6_addr_del(struct net *net, int ifindex, u32 ifa_flags,
3003 const struct in6_addr *pfx, unsigned int plen)
3004{
3005 struct inet6_ifaddr *ifp;
3006 struct inet6_dev *idev;
3007 struct net_device *dev;
3008
3009 if (plen > 128)
3010 return -EINVAL;
3011
3012 dev = __dev_get_by_index(net, ifindex);
3013 if (!dev)
3014 return -ENODEV;
3015
3016 idev = __in6_dev_get(dev);
3017 if (!idev)
3018 return -ENXIO;
3019
3020 read_lock_bh(&idev->lock);
3021 list_for_each_entry(ifp, &idev->addr_list, if_list) {
3022 if (ifp->prefix_len == plen &&
3023 ipv6_addr_equal(pfx, &ifp->addr)) {
3024 in6_ifa_hold(ifp);
3025 read_unlock_bh(&idev->lock);
3026
3027 if (!(ifp->flags & IFA_F_TEMPORARY) &&
3028 (ifa_flags & IFA_F_MANAGETEMPADDR))
3029 manage_tempaddrs(idev, ifp, 0, 0, false,
3030 jiffies);
3031 ipv6_del_addr(ifp);
3032 addrconf_verify_rtnl(net);
3033 if (ipv6_addr_is_multicast(pfx)) {
3034 ipv6_mc_config(net->ipv6.mc_autojoin_sk,
3035 false, pfx, dev->ifindex);
3036 }
3037 return 0;
3038 }
3039 }
3040 read_unlock_bh(&idev->lock);
3041 return -EADDRNOTAVAIL;
3042}
3043
3044
3045int addrconf_add_ifaddr(struct net *net, void __user *arg)
3046{
3047 struct ifa6_config cfg = {
3048 .ifa_flags = IFA_F_PERMANENT,
3049 .preferred_lft = INFINITY_LIFE_TIME,
3050 .valid_lft = INFINITY_LIFE_TIME,
3051 };
3052 struct in6_ifreq ireq;
3053 int err;
3054
3055 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
3056 return -EPERM;
3057
3058 if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
3059 return -EFAULT;
3060
3061 cfg.pfx = &ireq.ifr6_addr;
3062 cfg.plen = ireq.ifr6_prefixlen;
3063
3064 rtnl_lock();
3065 err = inet6_addr_add(net, ireq.ifr6_ifindex, &cfg, NULL);
3066 rtnl_unlock();
3067 return err;
3068}
3069
3070int addrconf_del_ifaddr(struct net *net, void __user *arg)
3071{
3072 struct in6_ifreq ireq;
3073 int err;
3074
3075 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
3076 return -EPERM;
3077
3078 if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
3079 return -EFAULT;
3080
3081 rtnl_lock();
3082 err = inet6_addr_del(net, ireq.ifr6_ifindex, 0, &ireq.ifr6_addr,
3083 ireq.ifr6_prefixlen);
3084 rtnl_unlock();
3085 return err;
3086}
3087
3088static void add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
3089 int plen, int scope, u8 proto)
3090{
3091 struct inet6_ifaddr *ifp;
3092 struct ifa6_config cfg = {
3093 .pfx = addr,
3094 .plen = plen,
3095 .ifa_flags = IFA_F_PERMANENT,
3096 .valid_lft = INFINITY_LIFE_TIME,
3097 .preferred_lft = INFINITY_LIFE_TIME,
3098 .scope = scope,
3099 .ifa_proto = proto
3100 };
3101
3102 ifp = ipv6_add_addr(idev, &cfg, true, NULL);
3103 if (!IS_ERR(ifp)) {
3104 spin_lock_bh(&ifp->lock);
3105 ifp->flags &= ~IFA_F_TENTATIVE;
3106 spin_unlock_bh(&ifp->lock);
3107 rt_genid_bump_ipv6(dev_net(idev->dev));
3108 ipv6_ifa_notify(RTM_NEWADDR, ifp);
3109 in6_ifa_put(ifp);
3110 }
3111}
3112
3113#if IS_ENABLED(CONFIG_IPV6_SIT) || IS_ENABLED(CONFIG_NET_IPGRE) || IS_ENABLED(CONFIG_IPV6_GRE)
3114static void add_v4_addrs(struct inet6_dev *idev)
3115{
3116 struct in6_addr addr;
3117 struct net_device *dev;
3118 struct net *net = dev_net(idev->dev);
3119 int scope, plen, offset = 0;
3120 u32 pflags = 0;
3121
3122 ASSERT_RTNL();
3123
3124 memset(&addr, 0, sizeof(struct in6_addr));
3125 /* in case of IP6GRE the dev_addr is an IPv6 and therefore we use only the last 4 bytes */
3126 if (idev->dev->addr_len == sizeof(struct in6_addr))
3127 offset = sizeof(struct in6_addr) - 4;
3128 memcpy(&addr.s6_addr32[3], idev->dev->dev_addr + offset, 4);
3129
3130 if (!(idev->dev->flags & IFF_POINTOPOINT) && idev->dev->type == ARPHRD_SIT) {
3131 scope = IPV6_ADDR_COMPATv4;
3132 plen = 96;
3133 pflags |= RTF_NONEXTHOP;
3134 } else {
3135 if (idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_NONE)
3136 return;
3137
3138 addr.s6_addr32[0] = htonl(0xfe800000);
3139 scope = IFA_LINK;
3140 plen = 64;
3141 }
3142
3143 if (addr.s6_addr32[3]) {
3144 add_addr(idev, &addr, plen, scope, IFAPROT_UNSPEC);
3145 addrconf_prefix_route(&addr, plen, 0, idev->dev, 0, pflags,
3146 GFP_KERNEL);
3147 return;
3148 }
3149
3150 for_each_netdev(net, dev) {
3151 struct in_device *in_dev = __in_dev_get_rtnl(dev);
3152 if (in_dev && (dev->flags & IFF_UP)) {
3153 struct in_ifaddr *ifa;
3154 int flag = scope;
3155
3156 in_dev_for_each_ifa_rtnl(ifa, in_dev) {
3157 addr.s6_addr32[3] = ifa->ifa_local;
3158
3159 if (ifa->ifa_scope == RT_SCOPE_LINK)
3160 continue;
3161 if (ifa->ifa_scope >= RT_SCOPE_HOST) {
3162 if (idev->dev->flags&IFF_POINTOPOINT)
3163 continue;
3164 flag |= IFA_HOST;
3165 }
3166
3167 add_addr(idev, &addr, plen, flag,
3168 IFAPROT_UNSPEC);
3169 addrconf_prefix_route(&addr, plen, 0, idev->dev,
3170 0, pflags, GFP_KERNEL);
3171 }
3172 }
3173 }
3174}
3175#endif
3176
3177static void init_loopback(struct net_device *dev)
3178{
3179 struct inet6_dev *idev;
3180
3181 /* ::1 */
3182
3183 ASSERT_RTNL();
3184
3185 idev = ipv6_find_idev(dev);
3186 if (IS_ERR(idev)) {
3187 pr_debug("%s: add_dev failed\n", __func__);
3188 return;
3189 }
3190
3191 add_addr(idev, &in6addr_loopback, 128, IFA_HOST, IFAPROT_KERNEL_LO);
3192}
3193
3194void addrconf_add_linklocal(struct inet6_dev *idev,
3195 const struct in6_addr *addr, u32 flags)
3196{
3197 struct ifa6_config cfg = {
3198 .pfx = addr,
3199 .plen = 64,
3200 .ifa_flags = flags | IFA_F_PERMANENT,
3201 .valid_lft = INFINITY_LIFE_TIME,
3202 .preferred_lft = INFINITY_LIFE_TIME,
3203 .scope = IFA_LINK,
3204 .ifa_proto = IFAPROT_KERNEL_LL
3205 };
3206 struct inet6_ifaddr *ifp;
3207
3208#ifdef CONFIG_IPV6_OPTIMISTIC_DAD
3209 if ((dev_net(idev->dev)->ipv6.devconf_all->optimistic_dad ||
3210 idev->cnf.optimistic_dad) &&
3211 !dev_net(idev->dev)->ipv6.devconf_all->forwarding)
3212 cfg.ifa_flags |= IFA_F_OPTIMISTIC;
3213#endif
3214
3215 ifp = ipv6_add_addr(idev, &cfg, true, NULL);
3216 if (!IS_ERR(ifp)) {
3217 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, 0, idev->dev,
3218 0, 0, GFP_ATOMIC);
3219 addrconf_dad_start(ifp);
3220 in6_ifa_put(ifp);
3221 }
3222}
3223EXPORT_SYMBOL_GPL(addrconf_add_linklocal);
3224
3225static bool ipv6_reserved_interfaceid(struct in6_addr address)
3226{
3227 if ((address.s6_addr32[2] | address.s6_addr32[3]) == 0)
3228 return true;
3229
3230 if (address.s6_addr32[2] == htonl(0x02005eff) &&
3231 ((address.s6_addr32[3] & htonl(0xfe000000)) == htonl(0xfe000000)))
3232 return true;
3233
3234 if (address.s6_addr32[2] == htonl(0xfdffffff) &&
3235 ((address.s6_addr32[3] & htonl(0xffffff80)) == htonl(0xffffff80)))
3236 return true;
3237
3238 return false;
3239}
3240
3241static int ipv6_generate_stable_address(struct in6_addr *address,
3242 u8 dad_count,
3243 const struct inet6_dev *idev)
3244{
3245 static DEFINE_SPINLOCK(lock);
3246 static __u32 digest[SHA1_DIGEST_WORDS];
3247 static __u32 workspace[SHA1_WORKSPACE_WORDS];
3248
3249 static union {
3250 char __data[SHA1_BLOCK_SIZE];
3251 struct {
3252 struct in6_addr secret;
3253 __be32 prefix[2];
3254 unsigned char hwaddr[MAX_ADDR_LEN];
3255 u8 dad_count;
3256 } __packed;
3257 } data;
3258
3259 struct in6_addr secret;
3260 struct in6_addr temp;
3261 struct net *net = dev_net(idev->dev);
3262
3263 BUILD_BUG_ON(sizeof(data.__data) != sizeof(data));
3264
3265 if (idev->cnf.stable_secret.initialized)
3266 secret = idev->cnf.stable_secret.secret;
3267 else if (net->ipv6.devconf_dflt->stable_secret.initialized)
3268 secret = net->ipv6.devconf_dflt->stable_secret.secret;
3269 else
3270 return -1;
3271
3272retry:
3273 spin_lock_bh(&lock);
3274
3275 sha1_init(digest);
3276 memset(&data, 0, sizeof(data));
3277 memset(workspace, 0, sizeof(workspace));
3278 memcpy(data.hwaddr, idev->dev->perm_addr, idev->dev->addr_len);
3279 data.prefix[0] = address->s6_addr32[0];
3280 data.prefix[1] = address->s6_addr32[1];
3281 data.secret = secret;
3282 data.dad_count = dad_count;
3283
3284 sha1_transform(digest, data.__data, workspace);
3285
3286 temp = *address;
3287 temp.s6_addr32[2] = (__force __be32)digest[0];
3288 temp.s6_addr32[3] = (__force __be32)digest[1];
3289
3290 spin_unlock_bh(&lock);
3291
3292 if (ipv6_reserved_interfaceid(temp)) {
3293 dad_count++;
3294 if (dad_count > dev_net(idev->dev)->ipv6.sysctl.idgen_retries)
3295 return -1;
3296 goto retry;
3297 }
3298
3299 *address = temp;
3300 return 0;
3301}
3302
3303static void ipv6_gen_mode_random_init(struct inet6_dev *idev)
3304{
3305 struct ipv6_stable_secret *s = &idev->cnf.stable_secret;
3306
3307 if (s->initialized)
3308 return;
3309 s = &idev->cnf.stable_secret;
3310 get_random_bytes(&s->secret, sizeof(s->secret));
3311 s->initialized = true;
3312}
3313
3314static void addrconf_addr_gen(struct inet6_dev *idev, bool prefix_route)
3315{
3316 struct in6_addr addr;
3317
3318 /* no link local addresses on L3 master devices */
3319 if (netif_is_l3_master(idev->dev))
3320 return;
3321
3322 /* no link local addresses on devices flagged as slaves */
3323 if (idev->dev->priv_flags & IFF_NO_ADDRCONF)
3324 return;
3325
3326 ipv6_addr_set(&addr, htonl(0xFE800000), 0, 0, 0);
3327
3328 switch (idev->cnf.addr_gen_mode) {
3329 case IN6_ADDR_GEN_MODE_RANDOM:
3330 ipv6_gen_mode_random_init(idev);
3331 fallthrough;
3332 case IN6_ADDR_GEN_MODE_STABLE_PRIVACY:
3333 if (!ipv6_generate_stable_address(&addr, 0, idev))
3334 addrconf_add_linklocal(idev, &addr,
3335 IFA_F_STABLE_PRIVACY);
3336 else if (prefix_route)
3337 addrconf_prefix_route(&addr, 64, 0, idev->dev,
3338 0, 0, GFP_KERNEL);
3339 break;
3340 case IN6_ADDR_GEN_MODE_EUI64:
3341 /* addrconf_add_linklocal also adds a prefix_route and we
3342 * only need to care about prefix routes if ipv6_generate_eui64
3343 * couldn't generate one.
3344 */
3345 if (ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) == 0)
3346 addrconf_add_linklocal(idev, &addr, 0);
3347 else if (prefix_route)
3348 addrconf_prefix_route(&addr, 64, 0, idev->dev,
3349 0, 0, GFP_KERNEL);
3350 break;
3351 case IN6_ADDR_GEN_MODE_NONE:
3352 default:
3353 /* will not add any link local address */
3354 break;
3355 }
3356}
3357
3358static void addrconf_dev_config(struct net_device *dev)
3359{
3360 struct inet6_dev *idev;
3361
3362 ASSERT_RTNL();
3363
3364 if ((dev->type != ARPHRD_ETHER) &&
3365 (dev->type != ARPHRD_FDDI) &&
3366 (dev->type != ARPHRD_ARCNET) &&
3367 (dev->type != ARPHRD_INFINIBAND) &&
3368 (dev->type != ARPHRD_IEEE1394) &&
3369 (dev->type != ARPHRD_TUNNEL6) &&
3370 (dev->type != ARPHRD_6LOWPAN) &&
3371 (dev->type != ARPHRD_TUNNEL) &&
3372 (dev->type != ARPHRD_NONE) &&
3373 (dev->type != ARPHRD_RAWIP)) {
3374 /* Alas, we support only Ethernet autoconfiguration. */
3375 idev = __in6_dev_get(dev);
3376 if (!IS_ERR_OR_NULL(idev) && dev->flags & IFF_UP &&
3377 dev->flags & IFF_MULTICAST)
3378 ipv6_mc_up(idev);
3379 return;
3380 }
3381
3382 idev = addrconf_add_dev(dev);
3383 if (IS_ERR(idev))
3384 return;
3385
3386 /* this device type has no EUI support */
3387 if (dev->type == ARPHRD_NONE &&
3388 idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_EUI64)
3389 idev->cnf.addr_gen_mode = IN6_ADDR_GEN_MODE_RANDOM;
3390
3391 addrconf_addr_gen(idev, false);
3392}
3393
3394#if IS_ENABLED(CONFIG_IPV6_SIT)
3395static void addrconf_sit_config(struct net_device *dev)
3396{
3397 struct inet6_dev *idev;
3398
3399 ASSERT_RTNL();
3400
3401 /*
3402 * Configure the tunnel with one of our IPv4
3403 * addresses... we should configure all of
3404 * our v4 addrs in the tunnel
3405 */
3406
3407 idev = ipv6_find_idev(dev);
3408 if (IS_ERR(idev)) {
3409 pr_debug("%s: add_dev failed\n", __func__);
3410 return;
3411 }
3412
3413 if (dev->priv_flags & IFF_ISATAP) {
3414 addrconf_addr_gen(idev, false);
3415 return;
3416 }
3417
3418 add_v4_addrs(idev);
3419
3420 if (dev->flags&IFF_POINTOPOINT)
3421 addrconf_add_mroute(dev);
3422}
3423#endif
3424
3425#if IS_ENABLED(CONFIG_NET_IPGRE) || IS_ENABLED(CONFIG_IPV6_GRE)
3426static void addrconf_gre_config(struct net_device *dev)
3427{
3428 struct inet6_dev *idev;
3429
3430 ASSERT_RTNL();
3431
3432 idev = ipv6_find_idev(dev);
3433 if (IS_ERR(idev)) {
3434 pr_debug("%s: add_dev failed\n", __func__);
3435 return;
3436 }
3437
3438 if (dev->type == ARPHRD_ETHER) {
3439 addrconf_addr_gen(idev, true);
3440 return;
3441 }
3442
3443 add_v4_addrs(idev);
3444
3445 if (dev->flags & IFF_POINTOPOINT)
3446 addrconf_add_mroute(dev);
3447}
3448#endif
3449
3450static void addrconf_init_auto_addrs(struct net_device *dev)
3451{
3452 switch (dev->type) {
3453#if IS_ENABLED(CONFIG_IPV6_SIT)
3454 case ARPHRD_SIT:
3455 addrconf_sit_config(dev);
3456 break;
3457#endif
3458#if IS_ENABLED(CONFIG_NET_IPGRE) || IS_ENABLED(CONFIG_IPV6_GRE)
3459 case ARPHRD_IP6GRE:
3460 case ARPHRD_IPGRE:
3461 addrconf_gre_config(dev);
3462 break;
3463#endif
3464 case ARPHRD_LOOPBACK:
3465 init_loopback(dev);
3466 break;
3467
3468 default:
3469 addrconf_dev_config(dev);
3470 break;
3471 }
3472}
3473
3474static int fixup_permanent_addr(struct net *net,
3475 struct inet6_dev *idev,
3476 struct inet6_ifaddr *ifp)
3477{
3478 /* !fib6_node means the host route was removed from the
3479 * FIB, for example, if 'lo' device is taken down. In that
3480 * case regenerate the host route.
3481 */
3482 if (!ifp->rt || !ifp->rt->fib6_node) {
3483 struct fib6_info *f6i, *prev;
3484
3485 f6i = addrconf_f6i_alloc(net, idev, &ifp->addr, false,
3486 GFP_ATOMIC);
3487 if (IS_ERR(f6i))
3488 return PTR_ERR(f6i);
3489
3490 /* ifp->rt can be accessed outside of rtnl */
3491 spin_lock(&ifp->lock);
3492 prev = ifp->rt;
3493 ifp->rt = f6i;
3494 spin_unlock(&ifp->lock);
3495
3496 fib6_info_release(prev);
3497 }
3498
3499 if (!(ifp->flags & IFA_F_NOPREFIXROUTE)) {
3500 addrconf_prefix_route(&ifp->addr, ifp->prefix_len,
3501 ifp->rt_priority, idev->dev, 0, 0,
3502 GFP_ATOMIC);
3503 }
3504
3505 if (ifp->state == INET6_IFADDR_STATE_PREDAD)
3506 addrconf_dad_start(ifp);
3507
3508 return 0;
3509}
3510
3511static void addrconf_permanent_addr(struct net *net, struct net_device *dev)
3512{
3513 struct inet6_ifaddr *ifp, *tmp;
3514 struct inet6_dev *idev;
3515
3516 idev = __in6_dev_get(dev);
3517 if (!idev)
3518 return;
3519
3520 write_lock_bh(&idev->lock);
3521
3522 list_for_each_entry_safe(ifp, tmp, &idev->addr_list, if_list) {
3523 if ((ifp->flags & IFA_F_PERMANENT) &&
3524 fixup_permanent_addr(net, idev, ifp) < 0) {
3525 write_unlock_bh(&idev->lock);
3526 in6_ifa_hold(ifp);
3527 ipv6_del_addr(ifp);
3528 write_lock_bh(&idev->lock);
3529
3530 net_info_ratelimited("%s: Failed to add prefix route for address %pI6c; dropping\n",
3531 idev->dev->name, &ifp->addr);
3532 }
3533 }
3534
3535 write_unlock_bh(&idev->lock);
3536}
3537
3538static int addrconf_notify(struct notifier_block *this, unsigned long event,
3539 void *ptr)
3540{
3541 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
3542 struct netdev_notifier_change_info *change_info;
3543 struct netdev_notifier_changeupper_info *info;
3544 struct inet6_dev *idev = __in6_dev_get(dev);
3545 struct net *net = dev_net(dev);
3546 int run_pending = 0;
3547 int err;
3548
3549 switch (event) {
3550 case NETDEV_REGISTER:
3551 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
3552 idev = ipv6_add_dev(dev);
3553 if (IS_ERR(idev))
3554 return notifier_from_errno(PTR_ERR(idev));
3555 }
3556 break;
3557
3558 case NETDEV_CHANGEMTU:
3559 /* if MTU under IPV6_MIN_MTU stop IPv6 on this interface. */
3560 if (dev->mtu < IPV6_MIN_MTU) {
3561 addrconf_ifdown(dev, dev != net->loopback_dev);
3562 break;
3563 }
3564
3565 if (idev) {
3566 rt6_mtu_change(dev, dev->mtu);
3567 idev->cnf.mtu6 = dev->mtu;
3568 break;
3569 }
3570
3571 /* allocate new idev */
3572 idev = ipv6_add_dev(dev);
3573 if (IS_ERR(idev))
3574 break;
3575
3576 /* device is still not ready */
3577 if (!(idev->if_flags & IF_READY))
3578 break;
3579
3580 run_pending = 1;
3581 fallthrough;
3582 case NETDEV_UP:
3583 case NETDEV_CHANGE:
3584 if (idev && idev->cnf.disable_ipv6)
3585 break;
3586
3587 if (dev->priv_flags & IFF_NO_ADDRCONF) {
3588 if (event == NETDEV_UP && !IS_ERR_OR_NULL(idev) &&
3589 dev->flags & IFF_UP && dev->flags & IFF_MULTICAST)
3590 ipv6_mc_up(idev);
3591 break;
3592 }
3593
3594 if (event == NETDEV_UP) {
3595 /* restore routes for permanent addresses */
3596 addrconf_permanent_addr(net, dev);
3597
3598 if (!addrconf_link_ready(dev)) {
3599 /* device is not ready yet. */
3600 pr_debug("ADDRCONF(NETDEV_UP): %s: link is not ready\n",
3601 dev->name);
3602 break;
3603 }
3604
3605 if (!idev && dev->mtu >= IPV6_MIN_MTU)
3606 idev = ipv6_add_dev(dev);
3607
3608 if (!IS_ERR_OR_NULL(idev)) {
3609 idev->if_flags |= IF_READY;
3610 run_pending = 1;
3611 }
3612 } else if (event == NETDEV_CHANGE) {
3613 if (!addrconf_link_ready(dev)) {
3614 /* device is still not ready. */
3615 rt6_sync_down_dev(dev, event);
3616 break;
3617 }
3618
3619 if (!IS_ERR_OR_NULL(idev)) {
3620 if (idev->if_flags & IF_READY) {
3621 /* device is already configured -
3622 * but resend MLD reports, we might
3623 * have roamed and need to update
3624 * multicast snooping switches
3625 */
3626 ipv6_mc_up(idev);
3627 change_info = ptr;
3628 if (change_info->flags_changed & IFF_NOARP)
3629 addrconf_dad_run(idev, true);
3630 rt6_sync_up(dev, RTNH_F_LINKDOWN);
3631 break;
3632 }
3633 idev->if_flags |= IF_READY;
3634 }
3635
3636 pr_info("ADDRCONF(NETDEV_CHANGE): %s: link becomes ready\n",
3637 dev->name);
3638
3639 run_pending = 1;
3640 }
3641
3642 addrconf_init_auto_addrs(dev);
3643
3644 if (!IS_ERR_OR_NULL(idev)) {
3645 if (run_pending)
3646 addrconf_dad_run(idev, false);
3647
3648 /* Device has an address by now */
3649 rt6_sync_up(dev, RTNH_F_DEAD);
3650
3651 /*
3652 * If the MTU changed during the interface down,
3653 * when the interface up, the changed MTU must be
3654 * reflected in the idev as well as routers.
3655 */
3656 if (idev->cnf.mtu6 != dev->mtu &&
3657 dev->mtu >= IPV6_MIN_MTU) {
3658 rt6_mtu_change(dev, dev->mtu);
3659 idev->cnf.mtu6 = dev->mtu;
3660 }
3661 idev->tstamp = jiffies;
3662 inet6_ifinfo_notify(RTM_NEWLINK, idev);
3663
3664 /*
3665 * If the changed mtu during down is lower than
3666 * IPV6_MIN_MTU stop IPv6 on this interface.
3667 */
3668 if (dev->mtu < IPV6_MIN_MTU)
3669 addrconf_ifdown(dev, dev != net->loopback_dev);
3670 }
3671 break;
3672
3673 case NETDEV_DOWN:
3674 case NETDEV_UNREGISTER:
3675 /*
3676 * Remove all addresses from this interface.
3677 */
3678 addrconf_ifdown(dev, event != NETDEV_DOWN);
3679 break;
3680
3681 case NETDEV_CHANGENAME:
3682 if (idev) {
3683 snmp6_unregister_dev(idev);
3684 addrconf_sysctl_unregister(idev);
3685 err = addrconf_sysctl_register(idev);
3686 if (err)
3687 return notifier_from_errno(err);
3688 err = snmp6_register_dev(idev);
3689 if (err) {
3690 addrconf_sysctl_unregister(idev);
3691 return notifier_from_errno(err);
3692 }
3693 }
3694 break;
3695
3696 case NETDEV_PRE_TYPE_CHANGE:
3697 case NETDEV_POST_TYPE_CHANGE:
3698 if (idev)
3699 addrconf_type_change(dev, event);
3700 break;
3701
3702 case NETDEV_CHANGEUPPER:
3703 info = ptr;
3704
3705 /* flush all routes if dev is linked to or unlinked from
3706 * an L3 master device (e.g., VRF)
3707 */
3708 if (info->upper_dev && netif_is_l3_master(info->upper_dev))
3709 addrconf_ifdown(dev, false);
3710 }
3711
3712 return NOTIFY_OK;
3713}
3714
3715/*
3716 * addrconf module should be notified of a device going up
3717 */
3718static struct notifier_block ipv6_dev_notf = {
3719 .notifier_call = addrconf_notify,
3720 .priority = ADDRCONF_NOTIFY_PRIORITY,
3721};
3722
3723static void addrconf_type_change(struct net_device *dev, unsigned long event)
3724{
3725 struct inet6_dev *idev;
3726 ASSERT_RTNL();
3727
3728 idev = __in6_dev_get(dev);
3729
3730 if (event == NETDEV_POST_TYPE_CHANGE)
3731 ipv6_mc_remap(idev);
3732 else if (event == NETDEV_PRE_TYPE_CHANGE)
3733 ipv6_mc_unmap(idev);
3734}
3735
3736static bool addr_is_local(const struct in6_addr *addr)
3737{
3738 return ipv6_addr_type(addr) &
3739 (IPV6_ADDR_LINKLOCAL | IPV6_ADDR_LOOPBACK);
3740}
3741
3742static int addrconf_ifdown(struct net_device *dev, bool unregister)
3743{
3744 unsigned long event = unregister ? NETDEV_UNREGISTER : NETDEV_DOWN;
3745 struct net *net = dev_net(dev);
3746 struct inet6_dev *idev;
3747 struct inet6_ifaddr *ifa;
3748 LIST_HEAD(tmp_addr_list);
3749 bool keep_addr = false;
3750 bool was_ready;
3751 int state, i;
3752
3753 ASSERT_RTNL();
3754
3755 rt6_disable_ip(dev, event);
3756
3757 idev = __in6_dev_get(dev);
3758 if (!idev)
3759 return -ENODEV;
3760
3761 /*
3762 * Step 1: remove reference to ipv6 device from parent device.
3763 * Do not dev_put!
3764 */
3765 if (unregister) {
3766 idev->dead = 1;
3767
3768 /* protected by rtnl_lock */
3769 RCU_INIT_POINTER(dev->ip6_ptr, NULL);
3770
3771 /* Step 1.5: remove snmp6 entry */
3772 snmp6_unregister_dev(idev);
3773
3774 }
3775
3776 /* combine the user config with event to determine if permanent
3777 * addresses are to be removed from address hash table
3778 */
3779 if (!unregister && !idev->cnf.disable_ipv6) {
3780 /* aggregate the system setting and interface setting */
3781 int _keep_addr = net->ipv6.devconf_all->keep_addr_on_down;
3782
3783 if (!_keep_addr)
3784 _keep_addr = idev->cnf.keep_addr_on_down;
3785
3786 keep_addr = (_keep_addr > 0);
3787 }
3788
3789 /* Step 2: clear hash table */
3790 for (i = 0; i < IN6_ADDR_HSIZE; i++) {
3791 struct hlist_head *h = &net->ipv6.inet6_addr_lst[i];
3792
3793 spin_lock_bh(&net->ipv6.addrconf_hash_lock);
3794restart:
3795 hlist_for_each_entry_rcu(ifa, h, addr_lst) {
3796 if (ifa->idev == idev) {
3797 addrconf_del_dad_work(ifa);
3798 /* combined flag + permanent flag decide if
3799 * address is retained on a down event
3800 */
3801 if (!keep_addr ||
3802 !(ifa->flags & IFA_F_PERMANENT) ||
3803 addr_is_local(&ifa->addr)) {
3804 hlist_del_init_rcu(&ifa->addr_lst);
3805 goto restart;
3806 }
3807 }
3808 }
3809 spin_unlock_bh(&net->ipv6.addrconf_hash_lock);
3810 }
3811
3812 write_lock_bh(&idev->lock);
3813
3814 addrconf_del_rs_timer(idev);
3815
3816 /* Step 2: clear flags for stateless addrconf, repeated down
3817 * detection
3818 */
3819 was_ready = idev->if_flags & IF_READY;
3820 if (!unregister)
3821 idev->if_flags &= ~(IF_RS_SENT|IF_RA_RCVD|IF_READY);
3822
3823 /* Step 3: clear tempaddr list */
3824 while (!list_empty(&idev->tempaddr_list)) {
3825 ifa = list_first_entry(&idev->tempaddr_list,
3826 struct inet6_ifaddr, tmp_list);
3827 list_del(&ifa->tmp_list);
3828 write_unlock_bh(&idev->lock);
3829 spin_lock_bh(&ifa->lock);
3830
3831 if (ifa->ifpub) {
3832 in6_ifa_put(ifa->ifpub);
3833 ifa->ifpub = NULL;
3834 }
3835 spin_unlock_bh(&ifa->lock);
3836 in6_ifa_put(ifa);
3837 write_lock_bh(&idev->lock);
3838 }
3839
3840 list_for_each_entry(ifa, &idev->addr_list, if_list)
3841 list_add_tail(&ifa->if_list_aux, &tmp_addr_list);
3842 write_unlock_bh(&idev->lock);
3843
3844 while (!list_empty(&tmp_addr_list)) {
3845 struct fib6_info *rt = NULL;
3846 bool keep;
3847
3848 ifa = list_first_entry(&tmp_addr_list,
3849 struct inet6_ifaddr, if_list_aux);
3850 list_del(&ifa->if_list_aux);
3851
3852 addrconf_del_dad_work(ifa);
3853
3854 keep = keep_addr && (ifa->flags & IFA_F_PERMANENT) &&
3855 !addr_is_local(&ifa->addr);
3856
3857 spin_lock_bh(&ifa->lock);
3858
3859 if (keep) {
3860 /* set state to skip the notifier below */
3861 state = INET6_IFADDR_STATE_DEAD;
3862 ifa->state = INET6_IFADDR_STATE_PREDAD;
3863 if (!(ifa->flags & IFA_F_NODAD))
3864 ifa->flags |= IFA_F_TENTATIVE;
3865
3866 rt = ifa->rt;
3867 ifa->rt = NULL;
3868 } else {
3869 state = ifa->state;
3870 ifa->state = INET6_IFADDR_STATE_DEAD;
3871 }
3872
3873 spin_unlock_bh(&ifa->lock);
3874
3875 if (rt)
3876 ip6_del_rt(net, rt, false);
3877
3878 if (state != INET6_IFADDR_STATE_DEAD) {
3879 __ipv6_ifa_notify(RTM_DELADDR, ifa);
3880 inet6addr_notifier_call_chain(NETDEV_DOWN, ifa);
3881 } else {
3882 if (idev->cnf.forwarding)
3883 addrconf_leave_anycast(ifa);
3884 addrconf_leave_solict(ifa->idev, &ifa->addr);
3885 }
3886
3887 if (!keep) {
3888 write_lock_bh(&idev->lock);
3889 list_del_rcu(&ifa->if_list);
3890 write_unlock_bh(&idev->lock);
3891 in6_ifa_put(ifa);
3892 }
3893 }
3894
3895 /* Step 5: Discard anycast and multicast list */
3896 if (unregister) {
3897 ipv6_ac_destroy_dev(idev);
3898 ipv6_mc_destroy_dev(idev);
3899 } else if (was_ready) {
3900 ipv6_mc_down(idev);
3901 }
3902
3903 idev->tstamp = jiffies;
3904 idev->ra_mtu = 0;
3905
3906 /* Last: Shot the device (if unregistered) */
3907 if (unregister) {
3908 addrconf_sysctl_unregister(idev);
3909 neigh_parms_release(&nd_tbl, idev->nd_parms);
3910 neigh_ifdown(&nd_tbl, dev);
3911 in6_dev_put(idev);
3912 }
3913 return 0;
3914}
3915
3916static void addrconf_rs_timer(struct timer_list *t)
3917{
3918 struct inet6_dev *idev = from_timer(idev, t, rs_timer);
3919 struct net_device *dev = idev->dev;
3920 struct in6_addr lladdr;
3921
3922 write_lock(&idev->lock);
3923 if (idev->dead || !(idev->if_flags & IF_READY))
3924 goto out;
3925
3926 if (!ipv6_accept_ra(idev))
3927 goto out;
3928
3929 /* Announcement received after solicitation was sent */
3930 if (idev->if_flags & IF_RA_RCVD)
3931 goto out;
3932
3933 if (idev->rs_probes++ < idev->cnf.rtr_solicits || idev->cnf.rtr_solicits < 0) {
3934 write_unlock(&idev->lock);
3935 if (!ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE))
3936 ndisc_send_rs(dev, &lladdr,
3937 &in6addr_linklocal_allrouters);
3938 else
3939 goto put;
3940
3941 write_lock(&idev->lock);
3942 idev->rs_interval = rfc3315_s14_backoff_update(
3943 idev->rs_interval, idev->cnf.rtr_solicit_max_interval);
3944 /* The wait after the last probe can be shorter */
3945 addrconf_mod_rs_timer(idev, (idev->rs_probes ==
3946 idev->cnf.rtr_solicits) ?
3947 idev->cnf.rtr_solicit_delay :
3948 idev->rs_interval);
3949 } else {
3950 /*
3951 * Note: we do not support deprecated "all on-link"
3952 * assumption any longer.
3953 */
3954 pr_debug("%s: no IPv6 routers present\n", idev->dev->name);
3955 }
3956
3957out:
3958 write_unlock(&idev->lock);
3959put:
3960 in6_dev_put(idev);
3961}
3962
3963/*
3964 * Duplicate Address Detection
3965 */
3966static void addrconf_dad_kick(struct inet6_ifaddr *ifp)
3967{
3968 unsigned long rand_num;
3969 struct inet6_dev *idev = ifp->idev;
3970 u64 nonce;
3971
3972 if (ifp->flags & IFA_F_OPTIMISTIC)
3973 rand_num = 0;
3974 else
3975 rand_num = get_random_u32_below(idev->cnf.rtr_solicit_delay ? : 1);
3976
3977 nonce = 0;
3978 if (idev->cnf.enhanced_dad ||
3979 dev_net(idev->dev)->ipv6.devconf_all->enhanced_dad) {
3980 do
3981 get_random_bytes(&nonce, 6);
3982 while (nonce == 0);
3983 }
3984 ifp->dad_nonce = nonce;
3985 ifp->dad_probes = idev->cnf.dad_transmits;
3986 addrconf_mod_dad_work(ifp, rand_num);
3987}
3988
3989static void addrconf_dad_begin(struct inet6_ifaddr *ifp)
3990{
3991 struct inet6_dev *idev = ifp->idev;
3992 struct net_device *dev = idev->dev;
3993 bool bump_id, notify = false;
3994 struct net *net;
3995
3996 addrconf_join_solict(dev, &ifp->addr);
3997
3998 read_lock_bh(&idev->lock);
3999 spin_lock(&ifp->lock);
4000 if (ifp->state == INET6_IFADDR_STATE_DEAD)
4001 goto out;
4002
4003 net = dev_net(dev);
4004 if (dev->flags&(IFF_NOARP|IFF_LOOPBACK) ||
4005 (net->ipv6.devconf_all->accept_dad < 1 &&
4006 idev->cnf.accept_dad < 1) ||
4007 !(ifp->flags&IFA_F_TENTATIVE) ||
4008 ifp->flags & IFA_F_NODAD) {
4009 bool send_na = false;
4010
4011 if (ifp->flags & IFA_F_TENTATIVE &&
4012 !(ifp->flags & IFA_F_OPTIMISTIC))
4013 send_na = true;
4014 bump_id = ifp->flags & IFA_F_TENTATIVE;
4015 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
4016 spin_unlock(&ifp->lock);
4017 read_unlock_bh(&idev->lock);
4018
4019 addrconf_dad_completed(ifp, bump_id, send_na);
4020 return;
4021 }
4022
4023 if (!(idev->if_flags & IF_READY)) {
4024 spin_unlock(&ifp->lock);
4025 read_unlock_bh(&idev->lock);
4026 /*
4027 * If the device is not ready:
4028 * - keep it tentative if it is a permanent address.
4029 * - otherwise, kill it.
4030 */
4031 in6_ifa_hold(ifp);
4032 addrconf_dad_stop(ifp, 0);
4033 return;
4034 }
4035
4036 /*
4037 * Optimistic nodes can start receiving
4038 * Frames right away
4039 */
4040 if (ifp->flags & IFA_F_OPTIMISTIC) {
4041 ip6_ins_rt(net, ifp->rt);
4042 if (ipv6_use_optimistic_addr(net, idev)) {
4043 /* Because optimistic nodes can use this address,
4044 * notify listeners. If DAD fails, RTM_DELADDR is sent.
4045 */
4046 notify = true;
4047 }
4048 }
4049
4050 addrconf_dad_kick(ifp);
4051out:
4052 spin_unlock(&ifp->lock);
4053 read_unlock_bh(&idev->lock);
4054 if (notify)
4055 ipv6_ifa_notify(RTM_NEWADDR, ifp);
4056}
4057
4058static void addrconf_dad_start(struct inet6_ifaddr *ifp)
4059{
4060 bool begin_dad = false;
4061
4062 spin_lock_bh(&ifp->lock);
4063 if (ifp->state != INET6_IFADDR_STATE_DEAD) {
4064 ifp->state = INET6_IFADDR_STATE_PREDAD;
4065 begin_dad = true;
4066 }
4067 spin_unlock_bh(&ifp->lock);
4068
4069 if (begin_dad)
4070 addrconf_mod_dad_work(ifp, 0);
4071}
4072
4073static void addrconf_dad_work(struct work_struct *w)
4074{
4075 struct inet6_ifaddr *ifp = container_of(to_delayed_work(w),
4076 struct inet6_ifaddr,
4077 dad_work);
4078 struct inet6_dev *idev = ifp->idev;
4079 bool bump_id, disable_ipv6 = false;
4080 struct in6_addr mcaddr;
4081
4082 enum {
4083 DAD_PROCESS,
4084 DAD_BEGIN,
4085 DAD_ABORT,
4086 } action = DAD_PROCESS;
4087
4088 rtnl_lock();
4089
4090 spin_lock_bh(&ifp->lock);
4091 if (ifp->state == INET6_IFADDR_STATE_PREDAD) {
4092 action = DAD_BEGIN;
4093 ifp->state = INET6_IFADDR_STATE_DAD;
4094 } else if (ifp->state == INET6_IFADDR_STATE_ERRDAD) {
4095 action = DAD_ABORT;
4096 ifp->state = INET6_IFADDR_STATE_POSTDAD;
4097
4098 if ((dev_net(idev->dev)->ipv6.devconf_all->accept_dad > 1 ||
4099 idev->cnf.accept_dad > 1) &&
4100 !idev->cnf.disable_ipv6 &&
4101 !(ifp->flags & IFA_F_STABLE_PRIVACY)) {
4102 struct in6_addr addr;
4103
4104 addr.s6_addr32[0] = htonl(0xfe800000);
4105 addr.s6_addr32[1] = 0;
4106
4107 if (!ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) &&
4108 ipv6_addr_equal(&ifp->addr, &addr)) {
4109 /* DAD failed for link-local based on MAC */
4110 idev->cnf.disable_ipv6 = 1;
4111
4112 pr_info("%s: IPv6 being disabled!\n",
4113 ifp->idev->dev->name);
4114 disable_ipv6 = true;
4115 }
4116 }
4117 }
4118 spin_unlock_bh(&ifp->lock);
4119
4120 if (action == DAD_BEGIN) {
4121 addrconf_dad_begin(ifp);
4122 goto out;
4123 } else if (action == DAD_ABORT) {
4124 in6_ifa_hold(ifp);
4125 addrconf_dad_stop(ifp, 1);
4126 if (disable_ipv6)
4127 addrconf_ifdown(idev->dev, false);
4128 goto out;
4129 }
4130
4131 if (!ifp->dad_probes && addrconf_dad_end(ifp))
4132 goto out;
4133
4134 write_lock_bh(&idev->lock);
4135 if (idev->dead || !(idev->if_flags & IF_READY)) {
4136 write_unlock_bh(&idev->lock);
4137 goto out;
4138 }
4139
4140 spin_lock(&ifp->lock);
4141 if (ifp->state == INET6_IFADDR_STATE_DEAD) {
4142 spin_unlock(&ifp->lock);
4143 write_unlock_bh(&idev->lock);
4144 goto out;
4145 }
4146
4147 if (ifp->dad_probes == 0) {
4148 bool send_na = false;
4149
4150 /*
4151 * DAD was successful
4152 */
4153
4154 if (ifp->flags & IFA_F_TENTATIVE &&
4155 !(ifp->flags & IFA_F_OPTIMISTIC))
4156 send_na = true;
4157 bump_id = ifp->flags & IFA_F_TENTATIVE;
4158 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
4159 spin_unlock(&ifp->lock);
4160 write_unlock_bh(&idev->lock);
4161
4162 addrconf_dad_completed(ifp, bump_id, send_na);
4163
4164 goto out;
4165 }
4166
4167 ifp->dad_probes--;
4168 addrconf_mod_dad_work(ifp,
4169 max(NEIGH_VAR(ifp->idev->nd_parms, RETRANS_TIME),
4170 HZ/100));
4171 spin_unlock(&ifp->lock);
4172 write_unlock_bh(&idev->lock);
4173
4174 /* send a neighbour solicitation for our addr */
4175 addrconf_addr_solict_mult(&ifp->addr, &mcaddr);
4176 ndisc_send_ns(ifp->idev->dev, &ifp->addr, &mcaddr, &in6addr_any,
4177 ifp->dad_nonce);
4178out:
4179 in6_ifa_put(ifp);
4180 rtnl_unlock();
4181}
4182
4183/* ifp->idev must be at least read locked */
4184static bool ipv6_lonely_lladdr(struct inet6_ifaddr *ifp)
4185{
4186 struct inet6_ifaddr *ifpiter;
4187 struct inet6_dev *idev = ifp->idev;
4188
4189 list_for_each_entry_reverse(ifpiter, &idev->addr_list, if_list) {
4190 if (ifpiter->scope > IFA_LINK)
4191 break;
4192 if (ifp != ifpiter && ifpiter->scope == IFA_LINK &&
4193 (ifpiter->flags & (IFA_F_PERMANENT|IFA_F_TENTATIVE|
4194 IFA_F_OPTIMISTIC|IFA_F_DADFAILED)) ==
4195 IFA_F_PERMANENT)
4196 return false;
4197 }
4198 return true;
4199}
4200
4201static void addrconf_dad_completed(struct inet6_ifaddr *ifp, bool bump_id,
4202 bool send_na)
4203{
4204 struct net_device *dev = ifp->idev->dev;
4205 struct in6_addr lladdr;
4206 bool send_rs, send_mld;
4207
4208 addrconf_del_dad_work(ifp);
4209
4210 /*
4211 * Configure the address for reception. Now it is valid.
4212 */
4213
4214 ipv6_ifa_notify(RTM_NEWADDR, ifp);
4215
4216 /* If added prefix is link local and we are prepared to process
4217 router advertisements, start sending router solicitations.
4218 */
4219
4220 read_lock_bh(&ifp->idev->lock);
4221 send_mld = ifp->scope == IFA_LINK && ipv6_lonely_lladdr(ifp);
4222 send_rs = send_mld &&
4223 ipv6_accept_ra(ifp->idev) &&
4224 ifp->idev->cnf.rtr_solicits != 0 &&
4225 (dev->flags & IFF_LOOPBACK) == 0 &&
4226 (dev->type != ARPHRD_TUNNEL);
4227 read_unlock_bh(&ifp->idev->lock);
4228
4229 /* While dad is in progress mld report's source address is in6_addrany.
4230 * Resend with proper ll now.
4231 */
4232 if (send_mld)
4233 ipv6_mc_dad_complete(ifp->idev);
4234
4235 /* send unsolicited NA if enabled */
4236 if (send_na &&
4237 (ifp->idev->cnf.ndisc_notify ||
4238 dev_net(dev)->ipv6.devconf_all->ndisc_notify)) {
4239 ndisc_send_na(dev, &in6addr_linklocal_allnodes, &ifp->addr,
4240 /*router=*/ !!ifp->idev->cnf.forwarding,
4241 /*solicited=*/ false, /*override=*/ true,
4242 /*inc_opt=*/ true);
4243 }
4244
4245 if (send_rs) {
4246 /*
4247 * If a host as already performed a random delay
4248 * [...] as part of DAD [...] there is no need
4249 * to delay again before sending the first RS
4250 */
4251 if (ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE))
4252 return;
4253 ndisc_send_rs(dev, &lladdr, &in6addr_linklocal_allrouters);
4254
4255 write_lock_bh(&ifp->idev->lock);
4256 spin_lock(&ifp->lock);
4257 ifp->idev->rs_interval = rfc3315_s14_backoff_init(
4258 ifp->idev->cnf.rtr_solicit_interval);
4259 ifp->idev->rs_probes = 1;
4260 ifp->idev->if_flags |= IF_RS_SENT;
4261 addrconf_mod_rs_timer(ifp->idev, ifp->idev->rs_interval);
4262 spin_unlock(&ifp->lock);
4263 write_unlock_bh(&ifp->idev->lock);
4264 }
4265
4266 if (bump_id)
4267 rt_genid_bump_ipv6(dev_net(dev));
4268
4269 /* Make sure that a new temporary address will be created
4270 * before this temporary address becomes deprecated.
4271 */
4272 if (ifp->flags & IFA_F_TEMPORARY)
4273 addrconf_verify_rtnl(dev_net(dev));
4274}
4275
4276static void addrconf_dad_run(struct inet6_dev *idev, bool restart)
4277{
4278 struct inet6_ifaddr *ifp;
4279
4280 read_lock_bh(&idev->lock);
4281 list_for_each_entry(ifp, &idev->addr_list, if_list) {
4282 spin_lock(&ifp->lock);
4283 if ((ifp->flags & IFA_F_TENTATIVE &&
4284 ifp->state == INET6_IFADDR_STATE_DAD) || restart) {
4285 if (restart)
4286 ifp->state = INET6_IFADDR_STATE_PREDAD;
4287 addrconf_dad_kick(ifp);
4288 }
4289 spin_unlock(&ifp->lock);
4290 }
4291 read_unlock_bh(&idev->lock);
4292}
4293
4294#ifdef CONFIG_PROC_FS
4295struct if6_iter_state {
4296 struct seq_net_private p;
4297 int bucket;
4298 int offset;
4299};
4300
4301static struct inet6_ifaddr *if6_get_first(struct seq_file *seq, loff_t pos)
4302{
4303 struct if6_iter_state *state = seq->private;
4304 struct net *net = seq_file_net(seq);
4305 struct inet6_ifaddr *ifa = NULL;
4306 int p = 0;
4307
4308 /* initial bucket if pos is 0 */
4309 if (pos == 0) {
4310 state->bucket = 0;
4311 state->offset = 0;
4312 }
4313
4314 for (; state->bucket < IN6_ADDR_HSIZE; ++state->bucket) {
4315 hlist_for_each_entry_rcu(ifa, &net->ipv6.inet6_addr_lst[state->bucket],
4316 addr_lst) {
4317 /* sync with offset */
4318 if (p < state->offset) {
4319 p++;
4320 continue;
4321 }
4322 return ifa;
4323 }
4324
4325 /* prepare for next bucket */
4326 state->offset = 0;
4327 p = 0;
4328 }
4329 return NULL;
4330}
4331
4332static struct inet6_ifaddr *if6_get_next(struct seq_file *seq,
4333 struct inet6_ifaddr *ifa)
4334{
4335 struct if6_iter_state *state = seq->private;
4336 struct net *net = seq_file_net(seq);
4337
4338 hlist_for_each_entry_continue_rcu(ifa, addr_lst) {
4339 state->offset++;
4340 return ifa;
4341 }
4342
4343 state->offset = 0;
4344 while (++state->bucket < IN6_ADDR_HSIZE) {
4345 hlist_for_each_entry_rcu(ifa,
4346 &net->ipv6.inet6_addr_lst[state->bucket], addr_lst) {
4347 return ifa;
4348 }
4349 }
4350
4351 return NULL;
4352}
4353
4354static void *if6_seq_start(struct seq_file *seq, loff_t *pos)
4355 __acquires(rcu)
4356{
4357 rcu_read_lock();
4358 return if6_get_first(seq, *pos);
4359}
4360
4361static void *if6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
4362{
4363 struct inet6_ifaddr *ifa;
4364
4365 ifa = if6_get_next(seq, v);
4366 ++*pos;
4367 return ifa;
4368}
4369
4370static void if6_seq_stop(struct seq_file *seq, void *v)
4371 __releases(rcu)
4372{
4373 rcu_read_unlock();
4374}
4375
4376static int if6_seq_show(struct seq_file *seq, void *v)
4377{
4378 struct inet6_ifaddr *ifp = (struct inet6_ifaddr *)v;
4379 seq_printf(seq, "%pi6 %02x %02x %02x %02x %8s\n",
4380 &ifp->addr,
4381 ifp->idev->dev->ifindex,
4382 ifp->prefix_len,
4383 ifp->scope,
4384 (u8) ifp->flags,
4385 ifp->idev->dev->name);
4386 return 0;
4387}
4388
4389static const struct seq_operations if6_seq_ops = {
4390 .start = if6_seq_start,
4391 .next = if6_seq_next,
4392 .show = if6_seq_show,
4393 .stop = if6_seq_stop,
4394};
4395
4396static int __net_init if6_proc_net_init(struct net *net)
4397{
4398 if (!proc_create_net("if_inet6", 0444, net->proc_net, &if6_seq_ops,
4399 sizeof(struct if6_iter_state)))
4400 return -ENOMEM;
4401 return 0;
4402}
4403
4404static void __net_exit if6_proc_net_exit(struct net *net)
4405{
4406 remove_proc_entry("if_inet6", net->proc_net);
4407}
4408
4409static struct pernet_operations if6_proc_net_ops = {
4410 .init = if6_proc_net_init,
4411 .exit = if6_proc_net_exit,
4412};
4413
4414int __init if6_proc_init(void)
4415{
4416 return register_pernet_subsys(&if6_proc_net_ops);
4417}
4418
4419void if6_proc_exit(void)
4420{
4421 unregister_pernet_subsys(&if6_proc_net_ops);
4422}
4423#endif /* CONFIG_PROC_FS */
4424
4425#if IS_ENABLED(CONFIG_IPV6_MIP6)
4426/* Check if address is a home address configured on any interface. */
4427int ipv6_chk_home_addr(struct net *net, const struct in6_addr *addr)
4428{
4429 unsigned int hash = inet6_addr_hash(net, addr);
4430 struct inet6_ifaddr *ifp = NULL;
4431 int ret = 0;
4432
4433 rcu_read_lock();
4434 hlist_for_each_entry_rcu(ifp, &net->ipv6.inet6_addr_lst[hash], addr_lst) {
4435 if (ipv6_addr_equal(&ifp->addr, addr) &&
4436 (ifp->flags & IFA_F_HOMEADDRESS)) {
4437 ret = 1;
4438 break;
4439 }
4440 }
4441 rcu_read_unlock();
4442 return ret;
4443}
4444#endif
4445
4446/* RFC6554 has some algorithm to avoid loops in segment routing by
4447 * checking if the segments contains any of a local interface address.
4448 *
4449 * Quote:
4450 *
4451 * To detect loops in the SRH, a router MUST determine if the SRH
4452 * includes multiple addresses assigned to any interface on that router.
4453 * If such addresses appear more than once and are separated by at least
4454 * one address not assigned to that router.
4455 */
4456int ipv6_chk_rpl_srh_loop(struct net *net, const struct in6_addr *segs,
4457 unsigned char nsegs)
4458{
4459 const struct in6_addr *addr;
4460 int i, ret = 0, found = 0;
4461 struct inet6_ifaddr *ifp;
4462 bool separated = false;
4463 unsigned int hash;
4464 bool hash_found;
4465
4466 rcu_read_lock();
4467 for (i = 0; i < nsegs; i++) {
4468 addr = &segs[i];
4469 hash = inet6_addr_hash(net, addr);
4470
4471 hash_found = false;
4472 hlist_for_each_entry_rcu(ifp, &net->ipv6.inet6_addr_lst[hash], addr_lst) {
4473
4474 if (ipv6_addr_equal(&ifp->addr, addr)) {
4475 hash_found = true;
4476 break;
4477 }
4478 }
4479
4480 if (hash_found) {
4481 if (found > 1 && separated) {
4482 ret = 1;
4483 break;
4484 }
4485
4486 separated = false;
4487 found++;
4488 } else {
4489 separated = true;
4490 }
4491 }
4492 rcu_read_unlock();
4493
4494 return ret;
4495}
4496
4497/*
4498 * Periodic address status verification
4499 */
4500
4501static void addrconf_verify_rtnl(struct net *net)
4502{
4503 unsigned long now, next, next_sec, next_sched;
4504 struct inet6_ifaddr *ifp;
4505 int i;
4506
4507 ASSERT_RTNL();
4508
4509 rcu_read_lock_bh();
4510 now = jiffies;
4511 next = round_jiffies_up(now + ADDR_CHECK_FREQUENCY);
4512
4513 cancel_delayed_work(&net->ipv6.addr_chk_work);
4514
4515 for (i = 0; i < IN6_ADDR_HSIZE; i++) {
4516restart:
4517 hlist_for_each_entry_rcu_bh(ifp, &net->ipv6.inet6_addr_lst[i], addr_lst) {
4518 unsigned long age;
4519
4520 /* When setting preferred_lft to a value not zero or
4521 * infinity, while valid_lft is infinity
4522 * IFA_F_PERMANENT has a non-infinity life time.
4523 */
4524 if ((ifp->flags & IFA_F_PERMANENT) &&
4525 (ifp->prefered_lft == INFINITY_LIFE_TIME))
4526 continue;
4527
4528 spin_lock(&ifp->lock);
4529 /* We try to batch several events at once. */
4530 age = (now - ifp->tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
4531
4532 if ((ifp->flags&IFA_F_TEMPORARY) &&
4533 !(ifp->flags&IFA_F_TENTATIVE) &&
4534 ifp->prefered_lft != INFINITY_LIFE_TIME &&
4535 !ifp->regen_count && ifp->ifpub) {
4536 /* This is a non-regenerated temporary addr. */
4537
4538 unsigned long regen_advance = ifp->idev->cnf.regen_max_retry *
4539 ifp->idev->cnf.dad_transmits *
4540 max(NEIGH_VAR(ifp->idev->nd_parms, RETRANS_TIME), HZ/100) / HZ;
4541
4542 if (age + regen_advance >= ifp->prefered_lft) {
4543 struct inet6_ifaddr *ifpub = ifp->ifpub;
4544 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
4545 next = ifp->tstamp + ifp->prefered_lft * HZ;
4546
4547 ifp->regen_count++;
4548 in6_ifa_hold(ifp);
4549 in6_ifa_hold(ifpub);
4550 spin_unlock(&ifp->lock);
4551
4552 spin_lock(&ifpub->lock);
4553 ifpub->regen_count = 0;
4554 spin_unlock(&ifpub->lock);
4555 rcu_read_unlock_bh();
4556 ipv6_create_tempaddr(ifpub, true);
4557 in6_ifa_put(ifpub);
4558 in6_ifa_put(ifp);
4559 rcu_read_lock_bh();
4560 goto restart;
4561 } else if (time_before(ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ, next))
4562 next = ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ;
4563 }
4564
4565 if (ifp->valid_lft != INFINITY_LIFE_TIME &&
4566 age >= ifp->valid_lft) {
4567 spin_unlock(&ifp->lock);
4568 in6_ifa_hold(ifp);
4569 rcu_read_unlock_bh();
4570 ipv6_del_addr(ifp);
4571 rcu_read_lock_bh();
4572 goto restart;
4573 } else if (ifp->prefered_lft == INFINITY_LIFE_TIME) {
4574 spin_unlock(&ifp->lock);
4575 continue;
4576 } else if (age >= ifp->prefered_lft) {
4577 /* jiffies - ifp->tstamp > age >= ifp->prefered_lft */
4578 int deprecate = 0;
4579
4580 if (!(ifp->flags&IFA_F_DEPRECATED)) {
4581 deprecate = 1;
4582 ifp->flags |= IFA_F_DEPRECATED;
4583 }
4584
4585 if ((ifp->valid_lft != INFINITY_LIFE_TIME) &&
4586 (time_before(ifp->tstamp + ifp->valid_lft * HZ, next)))
4587 next = ifp->tstamp + ifp->valid_lft * HZ;
4588
4589 spin_unlock(&ifp->lock);
4590
4591 if (deprecate) {
4592 in6_ifa_hold(ifp);
4593
4594 ipv6_ifa_notify(0, ifp);
4595 in6_ifa_put(ifp);
4596 goto restart;
4597 }
4598 } else {
4599 /* ifp->prefered_lft <= ifp->valid_lft */
4600 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
4601 next = ifp->tstamp + ifp->prefered_lft * HZ;
4602 spin_unlock(&ifp->lock);
4603 }
4604 }
4605 }
4606
4607 next_sec = round_jiffies_up(next);
4608 next_sched = next;
4609
4610 /* If rounded timeout is accurate enough, accept it. */
4611 if (time_before(next_sec, next + ADDRCONF_TIMER_FUZZ))
4612 next_sched = next_sec;
4613
4614 /* And minimum interval is ADDRCONF_TIMER_FUZZ_MAX. */
4615 if (time_before(next_sched, jiffies + ADDRCONF_TIMER_FUZZ_MAX))
4616 next_sched = jiffies + ADDRCONF_TIMER_FUZZ_MAX;
4617
4618 pr_debug("now = %lu, schedule = %lu, rounded schedule = %lu => %lu\n",
4619 now, next, next_sec, next_sched);
4620 mod_delayed_work(addrconf_wq, &net->ipv6.addr_chk_work, next_sched - now);
4621 rcu_read_unlock_bh();
4622}
4623
4624static void addrconf_verify_work(struct work_struct *w)
4625{
4626 struct net *net = container_of(to_delayed_work(w), struct net,
4627 ipv6.addr_chk_work);
4628
4629 rtnl_lock();
4630 addrconf_verify_rtnl(net);
4631 rtnl_unlock();
4632}
4633
4634static void addrconf_verify(struct net *net)
4635{
4636 mod_delayed_work(addrconf_wq, &net->ipv6.addr_chk_work, 0);
4637}
4638
4639static struct in6_addr *extract_addr(struct nlattr *addr, struct nlattr *local,
4640 struct in6_addr **peer_pfx)
4641{
4642 struct in6_addr *pfx = NULL;
4643
4644 *peer_pfx = NULL;
4645
4646 if (addr)
4647 pfx = nla_data(addr);
4648
4649 if (local) {
4650 if (pfx && nla_memcmp(local, pfx, sizeof(*pfx)))
4651 *peer_pfx = pfx;
4652 pfx = nla_data(local);
4653 }
4654
4655 return pfx;
4656}
4657
4658static const struct nla_policy ifa_ipv6_policy[IFA_MAX+1] = {
4659 [IFA_ADDRESS] = { .len = sizeof(struct in6_addr) },
4660 [IFA_LOCAL] = { .len = sizeof(struct in6_addr) },
4661 [IFA_CACHEINFO] = { .len = sizeof(struct ifa_cacheinfo) },
4662 [IFA_FLAGS] = { .len = sizeof(u32) },
4663 [IFA_RT_PRIORITY] = { .len = sizeof(u32) },
4664 [IFA_TARGET_NETNSID] = { .type = NLA_S32 },
4665 [IFA_PROTO] = { .type = NLA_U8 },
4666};
4667
4668static int
4669inet6_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh,
4670 struct netlink_ext_ack *extack)
4671{
4672 struct net *net = sock_net(skb->sk);
4673 struct ifaddrmsg *ifm;
4674 struct nlattr *tb[IFA_MAX+1];
4675 struct in6_addr *pfx, *peer_pfx;
4676 u32 ifa_flags;
4677 int err;
4678
4679 err = nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFA_MAX,
4680 ifa_ipv6_policy, extack);
4681 if (err < 0)
4682 return err;
4683
4684 ifm = nlmsg_data(nlh);
4685 pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
4686 if (!pfx)
4687 return -EINVAL;
4688
4689 ifa_flags = tb[IFA_FLAGS] ? nla_get_u32(tb[IFA_FLAGS]) : ifm->ifa_flags;
4690
4691 /* We ignore other flags so far. */
4692 ifa_flags &= IFA_F_MANAGETEMPADDR;
4693
4694 return inet6_addr_del(net, ifm->ifa_index, ifa_flags, pfx,
4695 ifm->ifa_prefixlen);
4696}
4697
4698static int modify_prefix_route(struct inet6_ifaddr *ifp,
4699 unsigned long expires, u32 flags,
4700 bool modify_peer)
4701{
4702 struct fib6_info *f6i;
4703 u32 prio;
4704
4705 f6i = addrconf_get_prefix_route(modify_peer ? &ifp->peer_addr : &ifp->addr,
4706 ifp->prefix_len,
4707 ifp->idev->dev, 0, RTF_DEFAULT, true);
4708 if (!f6i)
4709 return -ENOENT;
4710
4711 prio = ifp->rt_priority ? : IP6_RT_PRIO_ADDRCONF;
4712 if (f6i->fib6_metric != prio) {
4713 /* delete old one */
4714 ip6_del_rt(dev_net(ifp->idev->dev), f6i, false);
4715
4716 /* add new one */
4717 addrconf_prefix_route(modify_peer ? &ifp->peer_addr : &ifp->addr,
4718 ifp->prefix_len,
4719 ifp->rt_priority, ifp->idev->dev,
4720 expires, flags, GFP_KERNEL);
4721 } else {
4722 if (!expires)
4723 fib6_clean_expires(f6i);
4724 else
4725 fib6_set_expires(f6i, expires);
4726
4727 fib6_info_release(f6i);
4728 }
4729
4730 return 0;
4731}
4732
4733static int inet6_addr_modify(struct net *net, struct inet6_ifaddr *ifp,
4734 struct ifa6_config *cfg)
4735{
4736 u32 flags;
4737 clock_t expires;
4738 unsigned long timeout;
4739 bool was_managetempaddr;
4740 bool had_prefixroute;
4741 bool new_peer = false;
4742
4743 ASSERT_RTNL();
4744
4745 if (!cfg->valid_lft || cfg->preferred_lft > cfg->valid_lft)
4746 return -EINVAL;
4747
4748 if (cfg->ifa_flags & IFA_F_MANAGETEMPADDR &&
4749 (ifp->flags & IFA_F_TEMPORARY || ifp->prefix_len != 64))
4750 return -EINVAL;
4751
4752 if (!(ifp->flags & IFA_F_TENTATIVE) || ifp->flags & IFA_F_DADFAILED)
4753 cfg->ifa_flags &= ~IFA_F_OPTIMISTIC;
4754
4755 timeout = addrconf_timeout_fixup(cfg->valid_lft, HZ);
4756 if (addrconf_finite_timeout(timeout)) {
4757 expires = jiffies_to_clock_t(timeout * HZ);
4758 cfg->valid_lft = timeout;
4759 flags = RTF_EXPIRES;
4760 } else {
4761 expires = 0;
4762 flags = 0;
4763 cfg->ifa_flags |= IFA_F_PERMANENT;
4764 }
4765
4766 timeout = addrconf_timeout_fixup(cfg->preferred_lft, HZ);
4767 if (addrconf_finite_timeout(timeout)) {
4768 if (timeout == 0)
4769 cfg->ifa_flags |= IFA_F_DEPRECATED;
4770 cfg->preferred_lft = timeout;
4771 }
4772
4773 if (cfg->peer_pfx &&
4774 memcmp(&ifp->peer_addr, cfg->peer_pfx, sizeof(struct in6_addr))) {
4775 if (!ipv6_addr_any(&ifp->peer_addr))
4776 cleanup_prefix_route(ifp, expires, true, true);
4777 new_peer = true;
4778 }
4779
4780 spin_lock_bh(&ifp->lock);
4781 was_managetempaddr = ifp->flags & IFA_F_MANAGETEMPADDR;
4782 had_prefixroute = ifp->flags & IFA_F_PERMANENT &&
4783 !(ifp->flags & IFA_F_NOPREFIXROUTE);
4784 ifp->flags &= ~(IFA_F_DEPRECATED | IFA_F_PERMANENT | IFA_F_NODAD |
4785 IFA_F_HOMEADDRESS | IFA_F_MANAGETEMPADDR |
4786 IFA_F_NOPREFIXROUTE);
4787 ifp->flags |= cfg->ifa_flags;
4788 ifp->tstamp = jiffies;
4789 ifp->valid_lft = cfg->valid_lft;
4790 ifp->prefered_lft = cfg->preferred_lft;
4791 ifp->ifa_proto = cfg->ifa_proto;
4792
4793 if (cfg->rt_priority && cfg->rt_priority != ifp->rt_priority)
4794 ifp->rt_priority = cfg->rt_priority;
4795
4796 if (new_peer)
4797 ifp->peer_addr = *cfg->peer_pfx;
4798
4799 spin_unlock_bh(&ifp->lock);
4800 if (!(ifp->flags&IFA_F_TENTATIVE))
4801 ipv6_ifa_notify(0, ifp);
4802
4803 if (!(cfg->ifa_flags & IFA_F_NOPREFIXROUTE)) {
4804 int rc = -ENOENT;
4805
4806 if (had_prefixroute)
4807 rc = modify_prefix_route(ifp, expires, flags, false);
4808
4809 /* prefix route could have been deleted; if so restore it */
4810 if (rc == -ENOENT) {
4811 addrconf_prefix_route(&ifp->addr, ifp->prefix_len,
4812 ifp->rt_priority, ifp->idev->dev,
4813 expires, flags, GFP_KERNEL);
4814 }
4815
4816 if (had_prefixroute && !ipv6_addr_any(&ifp->peer_addr))
4817 rc = modify_prefix_route(ifp, expires, flags, true);
4818
4819 if (rc == -ENOENT && !ipv6_addr_any(&ifp->peer_addr)) {
4820 addrconf_prefix_route(&ifp->peer_addr, ifp->prefix_len,
4821 ifp->rt_priority, ifp->idev->dev,
4822 expires, flags, GFP_KERNEL);
4823 }
4824 } else if (had_prefixroute) {
4825 enum cleanup_prefix_rt_t action;
4826 unsigned long rt_expires;
4827
4828 write_lock_bh(&ifp->idev->lock);
4829 action = check_cleanup_prefix_route(ifp, &rt_expires);
4830 write_unlock_bh(&ifp->idev->lock);
4831
4832 if (action != CLEANUP_PREFIX_RT_NOP) {
4833 cleanup_prefix_route(ifp, rt_expires,
4834 action == CLEANUP_PREFIX_RT_DEL, false);
4835 }
4836 }
4837
4838 if (was_managetempaddr || ifp->flags & IFA_F_MANAGETEMPADDR) {
4839 if (was_managetempaddr &&
4840 !(ifp->flags & IFA_F_MANAGETEMPADDR)) {
4841 cfg->valid_lft = 0;
4842 cfg->preferred_lft = 0;
4843 }
4844 manage_tempaddrs(ifp->idev, ifp, cfg->valid_lft,
4845 cfg->preferred_lft, !was_managetempaddr,
4846 jiffies);
4847 }
4848
4849 addrconf_verify_rtnl(net);
4850
4851 return 0;
4852}
4853
4854static int
4855inet6_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh,
4856 struct netlink_ext_ack *extack)
4857{
4858 struct net *net = sock_net(skb->sk);
4859 struct ifaddrmsg *ifm;
4860 struct nlattr *tb[IFA_MAX+1];
4861 struct in6_addr *peer_pfx;
4862 struct inet6_ifaddr *ifa;
4863 struct net_device *dev;
4864 struct inet6_dev *idev;
4865 struct ifa6_config cfg;
4866 int err;
4867
4868 err = nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFA_MAX,
4869 ifa_ipv6_policy, extack);
4870 if (err < 0)
4871 return err;
4872
4873 memset(&cfg, 0, sizeof(cfg));
4874
4875 ifm = nlmsg_data(nlh);
4876 cfg.pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
4877 if (!cfg.pfx)
4878 return -EINVAL;
4879
4880 cfg.peer_pfx = peer_pfx;
4881 cfg.plen = ifm->ifa_prefixlen;
4882 if (tb[IFA_RT_PRIORITY])
4883 cfg.rt_priority = nla_get_u32(tb[IFA_RT_PRIORITY]);
4884
4885 if (tb[IFA_PROTO])
4886 cfg.ifa_proto = nla_get_u8(tb[IFA_PROTO]);
4887
4888 cfg.valid_lft = INFINITY_LIFE_TIME;
4889 cfg.preferred_lft = INFINITY_LIFE_TIME;
4890
4891 if (tb[IFA_CACHEINFO]) {
4892 struct ifa_cacheinfo *ci;
4893
4894 ci = nla_data(tb[IFA_CACHEINFO]);
4895 cfg.valid_lft = ci->ifa_valid;
4896 cfg.preferred_lft = ci->ifa_prefered;
4897 }
4898
4899 dev = __dev_get_by_index(net, ifm->ifa_index);
4900 if (!dev)
4901 return -ENODEV;
4902
4903 if (tb[IFA_FLAGS])
4904 cfg.ifa_flags = nla_get_u32(tb[IFA_FLAGS]);
4905 else
4906 cfg.ifa_flags = ifm->ifa_flags;
4907
4908 /* We ignore other flags so far. */
4909 cfg.ifa_flags &= IFA_F_NODAD | IFA_F_HOMEADDRESS |
4910 IFA_F_MANAGETEMPADDR | IFA_F_NOPREFIXROUTE |
4911 IFA_F_MCAUTOJOIN | IFA_F_OPTIMISTIC;
4912
4913 idev = ipv6_find_idev(dev);
4914 if (IS_ERR(idev))
4915 return PTR_ERR(idev);
4916
4917 if (!ipv6_allow_optimistic_dad(net, idev))
4918 cfg.ifa_flags &= ~IFA_F_OPTIMISTIC;
4919
4920 if (cfg.ifa_flags & IFA_F_NODAD &&
4921 cfg.ifa_flags & IFA_F_OPTIMISTIC) {
4922 NL_SET_ERR_MSG(extack, "IFA_F_NODAD and IFA_F_OPTIMISTIC are mutually exclusive");
4923 return -EINVAL;
4924 }
4925
4926 ifa = ipv6_get_ifaddr(net, cfg.pfx, dev, 1);
4927 if (!ifa) {
4928 /*
4929 * It would be best to check for !NLM_F_CREATE here but
4930 * userspace already relies on not having to provide this.
4931 */
4932 return inet6_addr_add(net, ifm->ifa_index, &cfg, extack);
4933 }
4934
4935 if (nlh->nlmsg_flags & NLM_F_EXCL ||
4936 !(nlh->nlmsg_flags & NLM_F_REPLACE))
4937 err = -EEXIST;
4938 else
4939 err = inet6_addr_modify(net, ifa, &cfg);
4940
4941 in6_ifa_put(ifa);
4942
4943 return err;
4944}
4945
4946static void put_ifaddrmsg(struct nlmsghdr *nlh, u8 prefixlen, u32 flags,
4947 u8 scope, int ifindex)
4948{
4949 struct ifaddrmsg *ifm;
4950
4951 ifm = nlmsg_data(nlh);
4952 ifm->ifa_family = AF_INET6;
4953 ifm->ifa_prefixlen = prefixlen;
4954 ifm->ifa_flags = flags;
4955 ifm->ifa_scope = scope;
4956 ifm->ifa_index = ifindex;
4957}
4958
4959static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
4960 unsigned long tstamp, u32 preferred, u32 valid)
4961{
4962 struct ifa_cacheinfo ci;
4963
4964 ci.cstamp = cstamp_delta(cstamp);
4965 ci.tstamp = cstamp_delta(tstamp);
4966 ci.ifa_prefered = preferred;
4967 ci.ifa_valid = valid;
4968
4969 return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
4970}
4971
4972static inline int rt_scope(int ifa_scope)
4973{
4974 if (ifa_scope & IFA_HOST)
4975 return RT_SCOPE_HOST;
4976 else if (ifa_scope & IFA_LINK)
4977 return RT_SCOPE_LINK;
4978 else if (ifa_scope & IFA_SITE)
4979 return RT_SCOPE_SITE;
4980 else
4981 return RT_SCOPE_UNIVERSE;
4982}
4983
4984static inline int inet6_ifaddr_msgsize(void)
4985{
4986 return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
4987 + nla_total_size(16) /* IFA_LOCAL */
4988 + nla_total_size(16) /* IFA_ADDRESS */
4989 + nla_total_size(sizeof(struct ifa_cacheinfo))
4990 + nla_total_size(4) /* IFA_FLAGS */
4991 + nla_total_size(1) /* IFA_PROTO */
4992 + nla_total_size(4) /* IFA_RT_PRIORITY */;
4993}
4994
4995enum addr_type_t {
4996 UNICAST_ADDR,
4997 MULTICAST_ADDR,
4998 ANYCAST_ADDR,
4999};
5000
5001struct inet6_fill_args {
5002 u32 portid;
5003 u32 seq;
5004 int event;
5005 unsigned int flags;
5006 int netnsid;
5007 int ifindex;
5008 enum addr_type_t type;
5009};
5010
5011static int inet6_fill_ifaddr(struct sk_buff *skb, struct inet6_ifaddr *ifa,
5012 struct inet6_fill_args *args)
5013{
5014 struct nlmsghdr *nlh;
5015 u32 preferred, valid;
5016
5017 nlh = nlmsg_put(skb, args->portid, args->seq, args->event,
5018 sizeof(struct ifaddrmsg), args->flags);
5019 if (!nlh)
5020 return -EMSGSIZE;
5021
5022 put_ifaddrmsg(nlh, ifa->prefix_len, ifa->flags, rt_scope(ifa->scope),
5023 ifa->idev->dev->ifindex);
5024
5025 if (args->netnsid >= 0 &&
5026 nla_put_s32(skb, IFA_TARGET_NETNSID, args->netnsid))
5027 goto error;
5028
5029 spin_lock_bh(&ifa->lock);
5030 if (!((ifa->flags&IFA_F_PERMANENT) &&
5031 (ifa->prefered_lft == INFINITY_LIFE_TIME))) {
5032 preferred = ifa->prefered_lft;
5033 valid = ifa->valid_lft;
5034 if (preferred != INFINITY_LIFE_TIME) {
5035 long tval = (jiffies - ifa->tstamp)/HZ;
5036 if (preferred > tval)
5037 preferred -= tval;
5038 else
5039 preferred = 0;
5040 if (valid != INFINITY_LIFE_TIME) {
5041 if (valid > tval)
5042 valid -= tval;
5043 else
5044 valid = 0;
5045 }
5046 }
5047 } else {
5048 preferred = INFINITY_LIFE_TIME;
5049 valid = INFINITY_LIFE_TIME;
5050 }
5051 spin_unlock_bh(&ifa->lock);
5052
5053 if (!ipv6_addr_any(&ifa->peer_addr)) {
5054 if (nla_put_in6_addr(skb, IFA_LOCAL, &ifa->addr) < 0 ||
5055 nla_put_in6_addr(skb, IFA_ADDRESS, &ifa->peer_addr) < 0)
5056 goto error;
5057 } else
5058 if (nla_put_in6_addr(skb, IFA_ADDRESS, &ifa->addr) < 0)
5059 goto error;
5060
5061 if (ifa->rt_priority &&
5062 nla_put_u32(skb, IFA_RT_PRIORITY, ifa->rt_priority))
5063 goto error;
5064
5065 if (put_cacheinfo(skb, ifa->cstamp, ifa->tstamp, preferred, valid) < 0)
5066 goto error;
5067
5068 if (nla_put_u32(skb, IFA_FLAGS, ifa->flags) < 0)
5069 goto error;
5070
5071 if (ifa->ifa_proto &&
5072 nla_put_u8(skb, IFA_PROTO, ifa->ifa_proto))
5073 goto error;
5074
5075 nlmsg_end(skb, nlh);
5076 return 0;
5077
5078error:
5079 nlmsg_cancel(skb, nlh);
5080 return -EMSGSIZE;
5081}
5082
5083static int inet6_fill_ifmcaddr(struct sk_buff *skb, struct ifmcaddr6 *ifmca,
5084 struct inet6_fill_args *args)
5085{
5086 struct nlmsghdr *nlh;
5087 u8 scope = RT_SCOPE_UNIVERSE;
5088 int ifindex = ifmca->idev->dev->ifindex;
5089
5090 if (ipv6_addr_scope(&ifmca->mca_addr) & IFA_SITE)
5091 scope = RT_SCOPE_SITE;
5092
5093 nlh = nlmsg_put(skb, args->portid, args->seq, args->event,
5094 sizeof(struct ifaddrmsg), args->flags);
5095 if (!nlh)
5096 return -EMSGSIZE;
5097
5098 if (args->netnsid >= 0 &&
5099 nla_put_s32(skb, IFA_TARGET_NETNSID, args->netnsid)) {
5100 nlmsg_cancel(skb, nlh);
5101 return -EMSGSIZE;
5102 }
5103
5104 put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
5105 if (nla_put_in6_addr(skb, IFA_MULTICAST, &ifmca->mca_addr) < 0 ||
5106 put_cacheinfo(skb, ifmca->mca_cstamp, ifmca->mca_tstamp,
5107 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
5108 nlmsg_cancel(skb, nlh);
5109 return -EMSGSIZE;
5110 }
5111
5112 nlmsg_end(skb, nlh);
5113 return 0;
5114}
5115
5116static int inet6_fill_ifacaddr(struct sk_buff *skb, struct ifacaddr6 *ifaca,
5117 struct inet6_fill_args *args)
5118{
5119 struct net_device *dev = fib6_info_nh_dev(ifaca->aca_rt);
5120 int ifindex = dev ? dev->ifindex : 1;
5121 struct nlmsghdr *nlh;
5122 u8 scope = RT_SCOPE_UNIVERSE;
5123
5124 if (ipv6_addr_scope(&ifaca->aca_addr) & IFA_SITE)
5125 scope = RT_SCOPE_SITE;
5126
5127 nlh = nlmsg_put(skb, args->portid, args->seq, args->event,
5128 sizeof(struct ifaddrmsg), args->flags);
5129 if (!nlh)
5130 return -EMSGSIZE;
5131
5132 if (args->netnsid >= 0 &&
5133 nla_put_s32(skb, IFA_TARGET_NETNSID, args->netnsid)) {
5134 nlmsg_cancel(skb, nlh);
5135 return -EMSGSIZE;
5136 }
5137
5138 put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
5139 if (nla_put_in6_addr(skb, IFA_ANYCAST, &ifaca->aca_addr) < 0 ||
5140 put_cacheinfo(skb, ifaca->aca_cstamp, ifaca->aca_tstamp,
5141 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
5142 nlmsg_cancel(skb, nlh);
5143 return -EMSGSIZE;
5144 }
5145
5146 nlmsg_end(skb, nlh);
5147 return 0;
5148}
5149
5150/* called with rcu_read_lock() */
5151static int in6_dump_addrs(struct inet6_dev *idev, struct sk_buff *skb,
5152 struct netlink_callback *cb, int s_ip_idx,
5153 struct inet6_fill_args *fillargs)
5154{
5155 struct ifmcaddr6 *ifmca;
5156 struct ifacaddr6 *ifaca;
5157 int ip_idx = 0;
5158 int err = 1;
5159
5160 read_lock_bh(&idev->lock);
5161 switch (fillargs->type) {
5162 case UNICAST_ADDR: {
5163 struct inet6_ifaddr *ifa;
5164 fillargs->event = RTM_NEWADDR;
5165
5166 /* unicast address incl. temp addr */
5167 list_for_each_entry(ifa, &idev->addr_list, if_list) {
5168 if (ip_idx < s_ip_idx)
5169 goto next;
5170 err = inet6_fill_ifaddr(skb, ifa, fillargs);
5171 if (err < 0)
5172 break;
5173 nl_dump_check_consistent(cb, nlmsg_hdr(skb));
5174next:
5175 ip_idx++;
5176 }
5177 break;
5178 }
5179 case MULTICAST_ADDR:
5180 read_unlock_bh(&idev->lock);
5181 fillargs->event = RTM_GETMULTICAST;
5182
5183 /* multicast address */
5184 for (ifmca = rtnl_dereference(idev->mc_list);
5185 ifmca;
5186 ifmca = rtnl_dereference(ifmca->next), ip_idx++) {
5187 if (ip_idx < s_ip_idx)
5188 continue;
5189 err = inet6_fill_ifmcaddr(skb, ifmca, fillargs);
5190 if (err < 0)
5191 break;
5192 }
5193 read_lock_bh(&idev->lock);
5194 break;
5195 case ANYCAST_ADDR:
5196 fillargs->event = RTM_GETANYCAST;
5197 /* anycast address */
5198 for (ifaca = idev->ac_list; ifaca;
5199 ifaca = ifaca->aca_next, ip_idx++) {
5200 if (ip_idx < s_ip_idx)
5201 continue;
5202 err = inet6_fill_ifacaddr(skb, ifaca, fillargs);
5203 if (err < 0)
5204 break;
5205 }
5206 break;
5207 default:
5208 break;
5209 }
5210 read_unlock_bh(&idev->lock);
5211 cb->args[2] = ip_idx;
5212 return err;
5213}
5214
5215static int inet6_valid_dump_ifaddr_req(const struct nlmsghdr *nlh,
5216 struct inet6_fill_args *fillargs,
5217 struct net **tgt_net, struct sock *sk,
5218 struct netlink_callback *cb)
5219{
5220 struct netlink_ext_ack *extack = cb->extack;
5221 struct nlattr *tb[IFA_MAX+1];
5222 struct ifaddrmsg *ifm;
5223 int err, i;
5224
5225 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) {
5226 NL_SET_ERR_MSG_MOD(extack, "Invalid header for address dump request");
5227 return -EINVAL;
5228 }
5229
5230 ifm = nlmsg_data(nlh);
5231 if (ifm->ifa_prefixlen || ifm->ifa_flags || ifm->ifa_scope) {
5232 NL_SET_ERR_MSG_MOD(extack, "Invalid values in header for address dump request");
5233 return -EINVAL;
5234 }
5235
5236 fillargs->ifindex = ifm->ifa_index;
5237 if (fillargs->ifindex) {
5238 cb->answer_flags |= NLM_F_DUMP_FILTERED;
5239 fillargs->flags |= NLM_F_DUMP_FILTERED;
5240 }
5241
5242 err = nlmsg_parse_deprecated_strict(nlh, sizeof(*ifm), tb, IFA_MAX,
5243 ifa_ipv6_policy, extack);
5244 if (err < 0)
5245 return err;
5246
5247 for (i = 0; i <= IFA_MAX; ++i) {
5248 if (!tb[i])
5249 continue;
5250
5251 if (i == IFA_TARGET_NETNSID) {
5252 struct net *net;
5253
5254 fillargs->netnsid = nla_get_s32(tb[i]);
5255 net = rtnl_get_net_ns_capable(sk, fillargs->netnsid);
5256 if (IS_ERR(net)) {
5257 fillargs->netnsid = -1;
5258 NL_SET_ERR_MSG_MOD(extack, "Invalid target network namespace id");
5259 return PTR_ERR(net);
5260 }
5261 *tgt_net = net;
5262 } else {
5263 NL_SET_ERR_MSG_MOD(extack, "Unsupported attribute in dump request");
5264 return -EINVAL;
5265 }
5266 }
5267
5268 return 0;
5269}
5270
5271static int inet6_dump_addr(struct sk_buff *skb, struct netlink_callback *cb,
5272 enum addr_type_t type)
5273{
5274 const struct nlmsghdr *nlh = cb->nlh;
5275 struct inet6_fill_args fillargs = {
5276 .portid = NETLINK_CB(cb->skb).portid,
5277 .seq = cb->nlh->nlmsg_seq,
5278 .flags = NLM_F_MULTI,
5279 .netnsid = -1,
5280 .type = type,
5281 };
5282 struct net *tgt_net = sock_net(skb->sk);
5283 int idx, s_idx, s_ip_idx;
5284 int h, s_h;
5285 struct net_device *dev;
5286 struct inet6_dev *idev;
5287 struct hlist_head *head;
5288 int err = 0;
5289
5290 s_h = cb->args[0];
5291 s_idx = idx = cb->args[1];
5292 s_ip_idx = cb->args[2];
5293
5294 if (cb->strict_check) {
5295 err = inet6_valid_dump_ifaddr_req(nlh, &fillargs, &tgt_net,
5296 skb->sk, cb);
5297 if (err < 0)
5298 goto put_tgt_net;
5299
5300 err = 0;
5301 if (fillargs.ifindex) {
5302 dev = __dev_get_by_index(tgt_net, fillargs.ifindex);
5303 if (!dev) {
5304 err = -ENODEV;
5305 goto put_tgt_net;
5306 }
5307 idev = __in6_dev_get(dev);
5308 if (idev) {
5309 err = in6_dump_addrs(idev, skb, cb, s_ip_idx,
5310 &fillargs);
5311 if (err > 0)
5312 err = 0;
5313 }
5314 goto put_tgt_net;
5315 }
5316 }
5317
5318 rcu_read_lock();
5319 cb->seq = atomic_read(&tgt_net->ipv6.dev_addr_genid) ^ tgt_net->dev_base_seq;
5320 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
5321 idx = 0;
5322 head = &tgt_net->dev_index_head[h];
5323 hlist_for_each_entry_rcu(dev, head, index_hlist) {
5324 if (idx < s_idx)
5325 goto cont;
5326 if (h > s_h || idx > s_idx)
5327 s_ip_idx = 0;
5328 idev = __in6_dev_get(dev);
5329 if (!idev)
5330 goto cont;
5331
5332 if (in6_dump_addrs(idev, skb, cb, s_ip_idx,
5333 &fillargs) < 0)
5334 goto done;
5335cont:
5336 idx++;
5337 }
5338 }
5339done:
5340 rcu_read_unlock();
5341 cb->args[0] = h;
5342 cb->args[1] = idx;
5343put_tgt_net:
5344 if (fillargs.netnsid >= 0)
5345 put_net(tgt_net);
5346
5347 return skb->len ? : err;
5348}
5349
5350static int inet6_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
5351{
5352 enum addr_type_t type = UNICAST_ADDR;
5353
5354 return inet6_dump_addr(skb, cb, type);
5355}
5356
5357static int inet6_dump_ifmcaddr(struct sk_buff *skb, struct netlink_callback *cb)
5358{
5359 enum addr_type_t type = MULTICAST_ADDR;
5360
5361 return inet6_dump_addr(skb, cb, type);
5362}
5363
5364
5365static int inet6_dump_ifacaddr(struct sk_buff *skb, struct netlink_callback *cb)
5366{
5367 enum addr_type_t type = ANYCAST_ADDR;
5368
5369 return inet6_dump_addr(skb, cb, type);
5370}
5371
5372static int inet6_rtm_valid_getaddr_req(struct sk_buff *skb,
5373 const struct nlmsghdr *nlh,
5374 struct nlattr **tb,
5375 struct netlink_ext_ack *extack)
5376{
5377 struct ifaddrmsg *ifm;
5378 int i, err;
5379
5380 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) {
5381 NL_SET_ERR_MSG_MOD(extack, "Invalid header for get address request");
5382 return -EINVAL;
5383 }
5384
5385 if (!netlink_strict_get_check(skb))
5386 return nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFA_MAX,
5387 ifa_ipv6_policy, extack);
5388
5389 ifm = nlmsg_data(nlh);
5390 if (ifm->ifa_prefixlen || ifm->ifa_flags || ifm->ifa_scope) {
5391 NL_SET_ERR_MSG_MOD(extack, "Invalid values in header for get address request");
5392 return -EINVAL;
5393 }
5394
5395 err = nlmsg_parse_deprecated_strict(nlh, sizeof(*ifm), tb, IFA_MAX,
5396 ifa_ipv6_policy, extack);
5397 if (err)
5398 return err;
5399
5400 for (i = 0; i <= IFA_MAX; i++) {
5401 if (!tb[i])
5402 continue;
5403
5404 switch (i) {
5405 case IFA_TARGET_NETNSID:
5406 case IFA_ADDRESS:
5407 case IFA_LOCAL:
5408 break;
5409 default:
5410 NL_SET_ERR_MSG_MOD(extack, "Unsupported attribute in get address request");
5411 return -EINVAL;
5412 }
5413 }
5414
5415 return 0;
5416}
5417
5418static int inet6_rtm_getaddr(struct sk_buff *in_skb, struct nlmsghdr *nlh,
5419 struct netlink_ext_ack *extack)
5420{
5421 struct net *tgt_net = sock_net(in_skb->sk);
5422 struct inet6_fill_args fillargs = {
5423 .portid = NETLINK_CB(in_skb).portid,
5424 .seq = nlh->nlmsg_seq,
5425 .event = RTM_NEWADDR,
5426 .flags = 0,
5427 .netnsid = -1,
5428 };
5429 struct ifaddrmsg *ifm;
5430 struct nlattr *tb[IFA_MAX+1];
5431 struct in6_addr *addr = NULL, *peer;
5432 struct net_device *dev = NULL;
5433 struct inet6_ifaddr *ifa;
5434 struct sk_buff *skb;
5435 int err;
5436
5437 err = inet6_rtm_valid_getaddr_req(in_skb, nlh, tb, extack);
5438 if (err < 0)
5439 return err;
5440
5441 if (tb[IFA_TARGET_NETNSID]) {
5442 fillargs.netnsid = nla_get_s32(tb[IFA_TARGET_NETNSID]);
5443
5444 tgt_net = rtnl_get_net_ns_capable(NETLINK_CB(in_skb).sk,
5445 fillargs.netnsid);
5446 if (IS_ERR(tgt_net))
5447 return PTR_ERR(tgt_net);
5448 }
5449
5450 addr = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer);
5451 if (!addr)
5452 return -EINVAL;
5453
5454 ifm = nlmsg_data(nlh);
5455 if (ifm->ifa_index)
5456 dev = dev_get_by_index(tgt_net, ifm->ifa_index);
5457
5458 ifa = ipv6_get_ifaddr(tgt_net, addr, dev, 1);
5459 if (!ifa) {
5460 err = -EADDRNOTAVAIL;
5461 goto errout;
5462 }
5463
5464 skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_KERNEL);
5465 if (!skb) {
5466 err = -ENOBUFS;
5467 goto errout_ifa;
5468 }
5469
5470 err = inet6_fill_ifaddr(skb, ifa, &fillargs);
5471 if (err < 0) {
5472 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
5473 WARN_ON(err == -EMSGSIZE);
5474 kfree_skb(skb);
5475 goto errout_ifa;
5476 }
5477 err = rtnl_unicast(skb, tgt_net, NETLINK_CB(in_skb).portid);
5478errout_ifa:
5479 in6_ifa_put(ifa);
5480errout:
5481 dev_put(dev);
5482 if (fillargs.netnsid >= 0)
5483 put_net(tgt_net);
5484
5485 return err;
5486}
5487
5488static void inet6_ifa_notify(int event, struct inet6_ifaddr *ifa)
5489{
5490 struct sk_buff *skb;
5491 struct net *net = dev_net(ifa->idev->dev);
5492 struct inet6_fill_args fillargs = {
5493 .portid = 0,
5494 .seq = 0,
5495 .event = event,
5496 .flags = 0,
5497 .netnsid = -1,
5498 };
5499 int err = -ENOBUFS;
5500
5501 skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_ATOMIC);
5502 if (!skb)
5503 goto errout;
5504
5505 err = inet6_fill_ifaddr(skb, ifa, &fillargs);
5506 if (err < 0) {
5507 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
5508 WARN_ON(err == -EMSGSIZE);
5509 kfree_skb(skb);
5510 goto errout;
5511 }
5512 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
5513 return;
5514errout:
5515 if (err < 0)
5516 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
5517}
5518
5519static inline void ipv6_store_devconf(struct ipv6_devconf *cnf,
5520 __s32 *array, int bytes)
5521{
5522 BUG_ON(bytes < (DEVCONF_MAX * 4));
5523
5524 memset(array, 0, bytes);
5525 array[DEVCONF_FORWARDING] = cnf->forwarding;
5526 array[DEVCONF_HOPLIMIT] = cnf->hop_limit;
5527 array[DEVCONF_MTU6] = cnf->mtu6;
5528 array[DEVCONF_ACCEPT_RA] = cnf->accept_ra;
5529 array[DEVCONF_ACCEPT_REDIRECTS] = cnf->accept_redirects;
5530 array[DEVCONF_AUTOCONF] = cnf->autoconf;
5531 array[DEVCONF_DAD_TRANSMITS] = cnf->dad_transmits;
5532 array[DEVCONF_RTR_SOLICITS] = cnf->rtr_solicits;
5533 array[DEVCONF_RTR_SOLICIT_INTERVAL] =
5534 jiffies_to_msecs(cnf->rtr_solicit_interval);
5535 array[DEVCONF_RTR_SOLICIT_MAX_INTERVAL] =
5536 jiffies_to_msecs(cnf->rtr_solicit_max_interval);
5537 array[DEVCONF_RTR_SOLICIT_DELAY] =
5538 jiffies_to_msecs(cnf->rtr_solicit_delay);
5539 array[DEVCONF_FORCE_MLD_VERSION] = cnf->force_mld_version;
5540 array[DEVCONF_MLDV1_UNSOLICITED_REPORT_INTERVAL] =
5541 jiffies_to_msecs(cnf->mldv1_unsolicited_report_interval);
5542 array[DEVCONF_MLDV2_UNSOLICITED_REPORT_INTERVAL] =
5543 jiffies_to_msecs(cnf->mldv2_unsolicited_report_interval);
5544 array[DEVCONF_USE_TEMPADDR] = cnf->use_tempaddr;
5545 array[DEVCONF_TEMP_VALID_LFT] = cnf->temp_valid_lft;
5546 array[DEVCONF_TEMP_PREFERED_LFT] = cnf->temp_prefered_lft;
5547 array[DEVCONF_REGEN_MAX_RETRY] = cnf->regen_max_retry;
5548 array[DEVCONF_MAX_DESYNC_FACTOR] = cnf->max_desync_factor;
5549 array[DEVCONF_MAX_ADDRESSES] = cnf->max_addresses;
5550 array[DEVCONF_ACCEPT_RA_DEFRTR] = cnf->accept_ra_defrtr;
5551 array[DEVCONF_RA_DEFRTR_METRIC] = cnf->ra_defrtr_metric;
5552 array[DEVCONF_ACCEPT_RA_MIN_HOP_LIMIT] = cnf->accept_ra_min_hop_limit;
5553 array[DEVCONF_ACCEPT_RA_PINFO] = cnf->accept_ra_pinfo;
5554#ifdef CONFIG_IPV6_ROUTER_PREF
5555 array[DEVCONF_ACCEPT_RA_RTR_PREF] = cnf->accept_ra_rtr_pref;
5556 array[DEVCONF_RTR_PROBE_INTERVAL] =
5557 jiffies_to_msecs(cnf->rtr_probe_interval);
5558#ifdef CONFIG_IPV6_ROUTE_INFO
5559 array[DEVCONF_ACCEPT_RA_RT_INFO_MIN_PLEN] = cnf->accept_ra_rt_info_min_plen;
5560 array[DEVCONF_ACCEPT_RA_RT_INFO_MAX_PLEN] = cnf->accept_ra_rt_info_max_plen;
5561#endif
5562#endif
5563 array[DEVCONF_PROXY_NDP] = cnf->proxy_ndp;
5564 array[DEVCONF_ACCEPT_SOURCE_ROUTE] = cnf->accept_source_route;
5565#ifdef CONFIG_IPV6_OPTIMISTIC_DAD
5566 array[DEVCONF_OPTIMISTIC_DAD] = cnf->optimistic_dad;
5567 array[DEVCONF_USE_OPTIMISTIC] = cnf->use_optimistic;
5568#endif
5569#ifdef CONFIG_IPV6_MROUTE
5570 array[DEVCONF_MC_FORWARDING] = atomic_read(&cnf->mc_forwarding);
5571#endif
5572 array[DEVCONF_DISABLE_IPV6] = cnf->disable_ipv6;
5573 array[DEVCONF_ACCEPT_DAD] = cnf->accept_dad;
5574 array[DEVCONF_FORCE_TLLAO] = cnf->force_tllao;
5575 array[DEVCONF_NDISC_NOTIFY] = cnf->ndisc_notify;
5576 array[DEVCONF_SUPPRESS_FRAG_NDISC] = cnf->suppress_frag_ndisc;
5577 array[DEVCONF_ACCEPT_RA_FROM_LOCAL] = cnf->accept_ra_from_local;
5578 array[DEVCONF_ACCEPT_RA_MTU] = cnf->accept_ra_mtu;
5579 array[DEVCONF_IGNORE_ROUTES_WITH_LINKDOWN] = cnf->ignore_routes_with_linkdown;
5580 /* we omit DEVCONF_STABLE_SECRET for now */
5581 array[DEVCONF_USE_OIF_ADDRS_ONLY] = cnf->use_oif_addrs_only;
5582 array[DEVCONF_DROP_UNICAST_IN_L2_MULTICAST] = cnf->drop_unicast_in_l2_multicast;
5583 array[DEVCONF_DROP_UNSOLICITED_NA] = cnf->drop_unsolicited_na;
5584 array[DEVCONF_KEEP_ADDR_ON_DOWN] = cnf->keep_addr_on_down;
5585 array[DEVCONF_SEG6_ENABLED] = cnf->seg6_enabled;
5586#ifdef CONFIG_IPV6_SEG6_HMAC
5587 array[DEVCONF_SEG6_REQUIRE_HMAC] = cnf->seg6_require_hmac;
5588#endif
5589 array[DEVCONF_ENHANCED_DAD] = cnf->enhanced_dad;
5590 array[DEVCONF_ADDR_GEN_MODE] = cnf->addr_gen_mode;
5591 array[DEVCONF_DISABLE_POLICY] = cnf->disable_policy;
5592 array[DEVCONF_NDISC_TCLASS] = cnf->ndisc_tclass;
5593 array[DEVCONF_RPL_SEG_ENABLED] = cnf->rpl_seg_enabled;
5594 array[DEVCONF_IOAM6_ENABLED] = cnf->ioam6_enabled;
5595 array[DEVCONF_IOAM6_ID] = cnf->ioam6_id;
5596 array[DEVCONF_IOAM6_ID_WIDE] = cnf->ioam6_id_wide;
5597 array[DEVCONF_NDISC_EVICT_NOCARRIER] = cnf->ndisc_evict_nocarrier;
5598 array[DEVCONF_ACCEPT_UNTRACKED_NA] = cnf->accept_untracked_na;
5599}
5600
5601static inline size_t inet6_ifla6_size(void)
5602{
5603 return nla_total_size(4) /* IFLA_INET6_FLAGS */
5604 + nla_total_size(sizeof(struct ifla_cacheinfo))
5605 + nla_total_size(DEVCONF_MAX * 4) /* IFLA_INET6_CONF */
5606 + nla_total_size(IPSTATS_MIB_MAX * 8) /* IFLA_INET6_STATS */
5607 + nla_total_size(ICMP6_MIB_MAX * 8) /* IFLA_INET6_ICMP6STATS */
5608 + nla_total_size(sizeof(struct in6_addr)) /* IFLA_INET6_TOKEN */
5609 + nla_total_size(1) /* IFLA_INET6_ADDR_GEN_MODE */
5610 + nla_total_size(4) /* IFLA_INET6_RA_MTU */
5611 + 0;
5612}
5613
5614static inline size_t inet6_if_nlmsg_size(void)
5615{
5616 return NLMSG_ALIGN(sizeof(struct ifinfomsg))
5617 + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
5618 + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
5619 + nla_total_size(4) /* IFLA_MTU */
5620 + nla_total_size(4) /* IFLA_LINK */
5621 + nla_total_size(1) /* IFLA_OPERSTATE */
5622 + nla_total_size(inet6_ifla6_size()); /* IFLA_PROTINFO */
5623}
5624
5625static inline void __snmp6_fill_statsdev(u64 *stats, atomic_long_t *mib,
5626 int bytes)
5627{
5628 int i;
5629 int pad = bytes - sizeof(u64) * ICMP6_MIB_MAX;
5630 BUG_ON(pad < 0);
5631
5632 /* Use put_unaligned() because stats may not be aligned for u64. */
5633 put_unaligned(ICMP6_MIB_MAX, &stats[0]);
5634 for (i = 1; i < ICMP6_MIB_MAX; i++)
5635 put_unaligned(atomic_long_read(&mib[i]), &stats[i]);
5636
5637 memset(&stats[ICMP6_MIB_MAX], 0, pad);
5638}
5639
5640static inline void __snmp6_fill_stats64(u64 *stats, void __percpu *mib,
5641 int bytes, size_t syncpoff)
5642{
5643 int i, c;
5644 u64 buff[IPSTATS_MIB_MAX];
5645 int pad = bytes - sizeof(u64) * IPSTATS_MIB_MAX;
5646
5647 BUG_ON(pad < 0);
5648
5649 memset(buff, 0, sizeof(buff));
5650 buff[0] = IPSTATS_MIB_MAX;
5651
5652 for_each_possible_cpu(c) {
5653 for (i = 1; i < IPSTATS_MIB_MAX; i++)
5654 buff[i] += snmp_get_cpu_field64(mib, c, i, syncpoff);
5655 }
5656
5657 memcpy(stats, buff, IPSTATS_MIB_MAX * sizeof(u64));
5658 memset(&stats[IPSTATS_MIB_MAX], 0, pad);
5659}
5660
5661static void snmp6_fill_stats(u64 *stats, struct inet6_dev *idev, int attrtype,
5662 int bytes)
5663{
5664 switch (attrtype) {
5665 case IFLA_INET6_STATS:
5666 __snmp6_fill_stats64(stats, idev->stats.ipv6, bytes,
5667 offsetof(struct ipstats_mib, syncp));
5668 break;
5669 case IFLA_INET6_ICMP6STATS:
5670 __snmp6_fill_statsdev(stats, idev->stats.icmpv6dev->mibs, bytes);
5671 break;
5672 }
5673}
5674
5675static int inet6_fill_ifla6_attrs(struct sk_buff *skb, struct inet6_dev *idev,
5676 u32 ext_filter_mask)
5677{
5678 struct nlattr *nla;
5679 struct ifla_cacheinfo ci;
5680
5681 if (nla_put_u32(skb, IFLA_INET6_FLAGS, idev->if_flags))
5682 goto nla_put_failure;
5683 ci.max_reasm_len = IPV6_MAXPLEN;
5684 ci.tstamp = cstamp_delta(idev->tstamp);
5685 ci.reachable_time = jiffies_to_msecs(idev->nd_parms->reachable_time);
5686 ci.retrans_time = jiffies_to_msecs(NEIGH_VAR(idev->nd_parms, RETRANS_TIME));
5687 if (nla_put(skb, IFLA_INET6_CACHEINFO, sizeof(ci), &ci))
5688 goto nla_put_failure;
5689 nla = nla_reserve(skb, IFLA_INET6_CONF, DEVCONF_MAX * sizeof(s32));
5690 if (!nla)
5691 goto nla_put_failure;
5692 ipv6_store_devconf(&idev->cnf, nla_data(nla), nla_len(nla));
5693
5694 /* XXX - MC not implemented */
5695
5696 if (ext_filter_mask & RTEXT_FILTER_SKIP_STATS)
5697 return 0;
5698
5699 nla = nla_reserve(skb, IFLA_INET6_STATS, IPSTATS_MIB_MAX * sizeof(u64));
5700 if (!nla)
5701 goto nla_put_failure;
5702 snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_STATS, nla_len(nla));
5703
5704 nla = nla_reserve(skb, IFLA_INET6_ICMP6STATS, ICMP6_MIB_MAX * sizeof(u64));
5705 if (!nla)
5706 goto nla_put_failure;
5707 snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_ICMP6STATS, nla_len(nla));
5708
5709 nla = nla_reserve(skb, IFLA_INET6_TOKEN, sizeof(struct in6_addr));
5710 if (!nla)
5711 goto nla_put_failure;
5712 read_lock_bh(&idev->lock);
5713 memcpy(nla_data(nla), idev->token.s6_addr, nla_len(nla));
5714 read_unlock_bh(&idev->lock);
5715
5716 if (nla_put_u8(skb, IFLA_INET6_ADDR_GEN_MODE, idev->cnf.addr_gen_mode))
5717 goto nla_put_failure;
5718
5719 if (idev->ra_mtu &&
5720 nla_put_u32(skb, IFLA_INET6_RA_MTU, idev->ra_mtu))
5721 goto nla_put_failure;
5722
5723 return 0;
5724
5725nla_put_failure:
5726 return -EMSGSIZE;
5727}
5728
5729static size_t inet6_get_link_af_size(const struct net_device *dev,
5730 u32 ext_filter_mask)
5731{
5732 if (!__in6_dev_get(dev))
5733 return 0;
5734
5735 return inet6_ifla6_size();
5736}
5737
5738static int inet6_fill_link_af(struct sk_buff *skb, const struct net_device *dev,
5739 u32 ext_filter_mask)
5740{
5741 struct inet6_dev *idev = __in6_dev_get(dev);
5742
5743 if (!idev)
5744 return -ENODATA;
5745
5746 if (inet6_fill_ifla6_attrs(skb, idev, ext_filter_mask) < 0)
5747 return -EMSGSIZE;
5748
5749 return 0;
5750}
5751
5752static int inet6_set_iftoken(struct inet6_dev *idev, struct in6_addr *token,
5753 struct netlink_ext_ack *extack)
5754{
5755 struct inet6_ifaddr *ifp;
5756 struct net_device *dev = idev->dev;
5757 bool clear_token, update_rs = false;
5758 struct in6_addr ll_addr;
5759
5760 ASSERT_RTNL();
5761
5762 if (!token)
5763 return -EINVAL;
5764
5765 if (dev->flags & IFF_LOOPBACK) {
5766 NL_SET_ERR_MSG_MOD(extack, "Device is loopback");
5767 return -EINVAL;
5768 }
5769
5770 if (dev->flags & IFF_NOARP) {
5771 NL_SET_ERR_MSG_MOD(extack,
5772 "Device does not do neighbour discovery");
5773 return -EINVAL;
5774 }
5775
5776 if (!ipv6_accept_ra(idev)) {
5777 NL_SET_ERR_MSG_MOD(extack,
5778 "Router advertisement is disabled on device");
5779 return -EINVAL;
5780 }
5781
5782 if (idev->cnf.rtr_solicits == 0) {
5783 NL_SET_ERR_MSG(extack,
5784 "Router solicitation is disabled on device");
5785 return -EINVAL;
5786 }
5787
5788 write_lock_bh(&idev->lock);
5789
5790 BUILD_BUG_ON(sizeof(token->s6_addr) != 16);
5791 memcpy(idev->token.s6_addr + 8, token->s6_addr + 8, 8);
5792
5793 write_unlock_bh(&idev->lock);
5794
5795 clear_token = ipv6_addr_any(token);
5796 if (clear_token)
5797 goto update_lft;
5798
5799 if (!idev->dead && (idev->if_flags & IF_READY) &&
5800 !ipv6_get_lladdr(dev, &ll_addr, IFA_F_TENTATIVE |
5801 IFA_F_OPTIMISTIC)) {
5802 /* If we're not ready, then normal ifup will take care
5803 * of this. Otherwise, we need to request our rs here.
5804 */
5805 ndisc_send_rs(dev, &ll_addr, &in6addr_linklocal_allrouters);
5806 update_rs = true;
5807 }
5808
5809update_lft:
5810 write_lock_bh(&idev->lock);
5811
5812 if (update_rs) {
5813 idev->if_flags |= IF_RS_SENT;
5814 idev->rs_interval = rfc3315_s14_backoff_init(
5815 idev->cnf.rtr_solicit_interval);
5816 idev->rs_probes = 1;
5817 addrconf_mod_rs_timer(idev, idev->rs_interval);
5818 }
5819
5820 /* Well, that's kinda nasty ... */
5821 list_for_each_entry(ifp, &idev->addr_list, if_list) {
5822 spin_lock(&ifp->lock);
5823 if (ifp->tokenized) {
5824 ifp->valid_lft = 0;
5825 ifp->prefered_lft = 0;
5826 }
5827 spin_unlock(&ifp->lock);
5828 }
5829
5830 write_unlock_bh(&idev->lock);
5831 inet6_ifinfo_notify(RTM_NEWLINK, idev);
5832 addrconf_verify_rtnl(dev_net(dev));
5833 return 0;
5834}
5835
5836static const struct nla_policy inet6_af_policy[IFLA_INET6_MAX + 1] = {
5837 [IFLA_INET6_ADDR_GEN_MODE] = { .type = NLA_U8 },
5838 [IFLA_INET6_TOKEN] = { .len = sizeof(struct in6_addr) },
5839 [IFLA_INET6_RA_MTU] = { .type = NLA_REJECT,
5840 .reject_message =
5841 "IFLA_INET6_RA_MTU can not be set" },
5842};
5843
5844static int check_addr_gen_mode(int mode)
5845{
5846 if (mode != IN6_ADDR_GEN_MODE_EUI64 &&
5847 mode != IN6_ADDR_GEN_MODE_NONE &&
5848 mode != IN6_ADDR_GEN_MODE_STABLE_PRIVACY &&
5849 mode != IN6_ADDR_GEN_MODE_RANDOM)
5850 return -EINVAL;
5851 return 1;
5852}
5853
5854static int check_stable_privacy(struct inet6_dev *idev, struct net *net,
5855 int mode)
5856{
5857 if (mode == IN6_ADDR_GEN_MODE_STABLE_PRIVACY &&
5858 !idev->cnf.stable_secret.initialized &&
5859 !net->ipv6.devconf_dflt->stable_secret.initialized)
5860 return -EINVAL;
5861 return 1;
5862}
5863
5864static int inet6_validate_link_af(const struct net_device *dev,
5865 const struct nlattr *nla,
5866 struct netlink_ext_ack *extack)
5867{
5868 struct nlattr *tb[IFLA_INET6_MAX + 1];
5869 struct inet6_dev *idev = NULL;
5870 int err;
5871
5872 if (dev) {
5873 idev = __in6_dev_get(dev);
5874 if (!idev)
5875 return -EAFNOSUPPORT;
5876 }
5877
5878 err = nla_parse_nested_deprecated(tb, IFLA_INET6_MAX, nla,
5879 inet6_af_policy, extack);
5880 if (err)
5881 return err;
5882
5883 if (!tb[IFLA_INET6_TOKEN] && !tb[IFLA_INET6_ADDR_GEN_MODE])
5884 return -EINVAL;
5885
5886 if (tb[IFLA_INET6_ADDR_GEN_MODE]) {
5887 u8 mode = nla_get_u8(tb[IFLA_INET6_ADDR_GEN_MODE]);
5888
5889 if (check_addr_gen_mode(mode) < 0)
5890 return -EINVAL;
5891 if (dev && check_stable_privacy(idev, dev_net(dev), mode) < 0)
5892 return -EINVAL;
5893 }
5894
5895 return 0;
5896}
5897
5898static int inet6_set_link_af(struct net_device *dev, const struct nlattr *nla,
5899 struct netlink_ext_ack *extack)
5900{
5901 struct inet6_dev *idev = __in6_dev_get(dev);
5902 struct nlattr *tb[IFLA_INET6_MAX + 1];
5903 int err;
5904
5905 if (!idev)
5906 return -EAFNOSUPPORT;
5907
5908 if (nla_parse_nested_deprecated(tb, IFLA_INET6_MAX, nla, NULL, NULL) < 0)
5909 return -EINVAL;
5910
5911 if (tb[IFLA_INET6_TOKEN]) {
5912 err = inet6_set_iftoken(idev, nla_data(tb[IFLA_INET6_TOKEN]),
5913 extack);
5914 if (err)
5915 return err;
5916 }
5917
5918 if (tb[IFLA_INET6_ADDR_GEN_MODE]) {
5919 u8 mode = nla_get_u8(tb[IFLA_INET6_ADDR_GEN_MODE]);
5920
5921 idev->cnf.addr_gen_mode = mode;
5922 }
5923
5924 return 0;
5925}
5926
5927static int inet6_fill_ifinfo(struct sk_buff *skb, struct inet6_dev *idev,
5928 u32 portid, u32 seq, int event, unsigned int flags)
5929{
5930 struct net_device *dev = idev->dev;
5931 struct ifinfomsg *hdr;
5932 struct nlmsghdr *nlh;
5933 void *protoinfo;
5934
5935 nlh = nlmsg_put(skb, portid, seq, event, sizeof(*hdr), flags);
5936 if (!nlh)
5937 return -EMSGSIZE;
5938
5939 hdr = nlmsg_data(nlh);
5940 hdr->ifi_family = AF_INET6;
5941 hdr->__ifi_pad = 0;
5942 hdr->ifi_type = dev->type;
5943 hdr->ifi_index = dev->ifindex;
5944 hdr->ifi_flags = dev_get_flags(dev);
5945 hdr->ifi_change = 0;
5946
5947 if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
5948 (dev->addr_len &&
5949 nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
5950 nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
5951 (dev->ifindex != dev_get_iflink(dev) &&
5952 nla_put_u32(skb, IFLA_LINK, dev_get_iflink(dev))) ||
5953 nla_put_u8(skb, IFLA_OPERSTATE,
5954 netif_running(dev) ? dev->operstate : IF_OPER_DOWN))
5955 goto nla_put_failure;
5956 protoinfo = nla_nest_start_noflag(skb, IFLA_PROTINFO);
5957 if (!protoinfo)
5958 goto nla_put_failure;
5959
5960 if (inet6_fill_ifla6_attrs(skb, idev, 0) < 0)
5961 goto nla_put_failure;
5962
5963 nla_nest_end(skb, protoinfo);
5964 nlmsg_end(skb, nlh);
5965 return 0;
5966
5967nla_put_failure:
5968 nlmsg_cancel(skb, nlh);
5969 return -EMSGSIZE;
5970}
5971
5972static int inet6_valid_dump_ifinfo(const struct nlmsghdr *nlh,
5973 struct netlink_ext_ack *extack)
5974{
5975 struct ifinfomsg *ifm;
5976
5977 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) {
5978 NL_SET_ERR_MSG_MOD(extack, "Invalid header for link dump request");
5979 return -EINVAL;
5980 }
5981
5982 if (nlmsg_attrlen(nlh, sizeof(*ifm))) {
5983 NL_SET_ERR_MSG_MOD(extack, "Invalid data after header");
5984 return -EINVAL;
5985 }
5986
5987 ifm = nlmsg_data(nlh);
5988 if (ifm->__ifi_pad || ifm->ifi_type || ifm->ifi_flags ||
5989 ifm->ifi_change || ifm->ifi_index) {
5990 NL_SET_ERR_MSG_MOD(extack, "Invalid values in header for dump request");
5991 return -EINVAL;
5992 }
5993
5994 return 0;
5995}
5996
5997static int inet6_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
5998{
5999 struct net *net = sock_net(skb->sk);
6000 int h, s_h;
6001 int idx = 0, s_idx;
6002 struct net_device *dev;
6003 struct inet6_dev *idev;
6004 struct hlist_head *head;
6005
6006 /* only requests using strict checking can pass data to
6007 * influence the dump
6008 */
6009 if (cb->strict_check) {
6010 int err = inet6_valid_dump_ifinfo(cb->nlh, cb->extack);
6011
6012 if (err < 0)
6013 return err;
6014 }
6015
6016 s_h = cb->args[0];
6017 s_idx = cb->args[1];
6018
6019 rcu_read_lock();
6020 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
6021 idx = 0;
6022 head = &net->dev_index_head[h];
6023 hlist_for_each_entry_rcu(dev, head, index_hlist) {
6024 if (idx < s_idx)
6025 goto cont;
6026 idev = __in6_dev_get(dev);
6027 if (!idev)
6028 goto cont;
6029 if (inet6_fill_ifinfo(skb, idev,
6030 NETLINK_CB(cb->skb).portid,
6031 cb->nlh->nlmsg_seq,
6032 RTM_NEWLINK, NLM_F_MULTI) < 0)
6033 goto out;
6034cont:
6035 idx++;
6036 }
6037 }
6038out:
6039 rcu_read_unlock();
6040 cb->args[1] = idx;
6041 cb->args[0] = h;
6042
6043 return skb->len;
6044}
6045
6046void inet6_ifinfo_notify(int event, struct inet6_dev *idev)
6047{
6048 struct sk_buff *skb;
6049 struct net *net = dev_net(idev->dev);
6050 int err = -ENOBUFS;
6051
6052 skb = nlmsg_new(inet6_if_nlmsg_size(), GFP_ATOMIC);
6053 if (!skb)
6054 goto errout;
6055
6056 err = inet6_fill_ifinfo(skb, idev, 0, 0, event, 0);
6057 if (err < 0) {
6058 /* -EMSGSIZE implies BUG in inet6_if_nlmsg_size() */
6059 WARN_ON(err == -EMSGSIZE);
6060 kfree_skb(skb);
6061 goto errout;
6062 }
6063 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFINFO, NULL, GFP_ATOMIC);
6064 return;
6065errout:
6066 if (err < 0)
6067 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFINFO, err);
6068}
6069
6070static inline size_t inet6_prefix_nlmsg_size(void)
6071{
6072 return NLMSG_ALIGN(sizeof(struct prefixmsg))
6073 + nla_total_size(sizeof(struct in6_addr))
6074 + nla_total_size(sizeof(struct prefix_cacheinfo));
6075}
6076
6077static int inet6_fill_prefix(struct sk_buff *skb, struct inet6_dev *idev,
6078 struct prefix_info *pinfo, u32 portid, u32 seq,
6079 int event, unsigned int flags)
6080{
6081 struct prefixmsg *pmsg;
6082 struct nlmsghdr *nlh;
6083 struct prefix_cacheinfo ci;
6084
6085 nlh = nlmsg_put(skb, portid, seq, event, sizeof(*pmsg), flags);
6086 if (!nlh)
6087 return -EMSGSIZE;
6088
6089 pmsg = nlmsg_data(nlh);
6090 pmsg->prefix_family = AF_INET6;
6091 pmsg->prefix_pad1 = 0;
6092 pmsg->prefix_pad2 = 0;
6093 pmsg->prefix_ifindex = idev->dev->ifindex;
6094 pmsg->prefix_len = pinfo->prefix_len;
6095 pmsg->prefix_type = pinfo->type;
6096 pmsg->prefix_pad3 = 0;
6097 pmsg->prefix_flags = 0;
6098 if (pinfo->onlink)
6099 pmsg->prefix_flags |= IF_PREFIX_ONLINK;
6100 if (pinfo->autoconf)
6101 pmsg->prefix_flags |= IF_PREFIX_AUTOCONF;
6102
6103 if (nla_put(skb, PREFIX_ADDRESS, sizeof(pinfo->prefix), &pinfo->prefix))
6104 goto nla_put_failure;
6105 ci.preferred_time = ntohl(pinfo->prefered);
6106 ci.valid_time = ntohl(pinfo->valid);
6107 if (nla_put(skb, PREFIX_CACHEINFO, sizeof(ci), &ci))
6108 goto nla_put_failure;
6109 nlmsg_end(skb, nlh);
6110 return 0;
6111
6112nla_put_failure:
6113 nlmsg_cancel(skb, nlh);
6114 return -EMSGSIZE;
6115}
6116
6117static void inet6_prefix_notify(int event, struct inet6_dev *idev,
6118 struct prefix_info *pinfo)
6119{
6120 struct sk_buff *skb;
6121 struct net *net = dev_net(idev->dev);
6122 int err = -ENOBUFS;
6123
6124 skb = nlmsg_new(inet6_prefix_nlmsg_size(), GFP_ATOMIC);
6125 if (!skb)
6126 goto errout;
6127
6128 err = inet6_fill_prefix(skb, idev, pinfo, 0, 0, event, 0);
6129 if (err < 0) {
6130 /* -EMSGSIZE implies BUG in inet6_prefix_nlmsg_size() */
6131 WARN_ON(err == -EMSGSIZE);
6132 kfree_skb(skb);
6133 goto errout;
6134 }
6135 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_PREFIX, NULL, GFP_ATOMIC);
6136 return;
6137errout:
6138 if (err < 0)
6139 rtnl_set_sk_err(net, RTNLGRP_IPV6_PREFIX, err);
6140}
6141
6142static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
6143{
6144 struct net *net = dev_net(ifp->idev->dev);
6145
6146 if (event)
6147 ASSERT_RTNL();
6148
6149 inet6_ifa_notify(event ? : RTM_NEWADDR, ifp);
6150
6151 switch (event) {
6152 case RTM_NEWADDR:
6153 /*
6154 * If the address was optimistic we inserted the route at the
6155 * start of our DAD process, so we don't need to do it again.
6156 * If the device was taken down in the middle of the DAD
6157 * cycle there is a race where we could get here without a
6158 * host route, so nothing to insert. That will be fixed when
6159 * the device is brought up.
6160 */
6161 if (ifp->rt && !rcu_access_pointer(ifp->rt->fib6_node)) {
6162 ip6_ins_rt(net, ifp->rt);
6163 } else if (!ifp->rt && (ifp->idev->dev->flags & IFF_UP)) {
6164 pr_warn("BUG: Address %pI6c on device %s is missing its host route.\n",
6165 &ifp->addr, ifp->idev->dev->name);
6166 }
6167
6168 if (ifp->idev->cnf.forwarding)
6169 addrconf_join_anycast(ifp);
6170 if (!ipv6_addr_any(&ifp->peer_addr))
6171 addrconf_prefix_route(&ifp->peer_addr, 128,
6172 ifp->rt_priority, ifp->idev->dev,
6173 0, 0, GFP_ATOMIC);
6174 break;
6175 case RTM_DELADDR:
6176 if (ifp->idev->cnf.forwarding)
6177 addrconf_leave_anycast(ifp);
6178 addrconf_leave_solict(ifp->idev, &ifp->addr);
6179 if (!ipv6_addr_any(&ifp->peer_addr)) {
6180 struct fib6_info *rt;
6181
6182 rt = addrconf_get_prefix_route(&ifp->peer_addr, 128,
6183 ifp->idev->dev, 0, 0,
6184 false);
6185 if (rt)
6186 ip6_del_rt(net, rt, false);
6187 }
6188 if (ifp->rt) {
6189 ip6_del_rt(net, ifp->rt, false);
6190 ifp->rt = NULL;
6191 }
6192 rt_genid_bump_ipv6(net);
6193 break;
6194 }
6195 atomic_inc(&net->ipv6.dev_addr_genid);
6196}
6197
6198static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
6199{
6200 if (likely(ifp->idev->dead == 0))
6201 __ipv6_ifa_notify(event, ifp);
6202}
6203
6204#ifdef CONFIG_SYSCTL
6205
6206static int addrconf_sysctl_forward(struct ctl_table *ctl, int write,
6207 void *buffer, size_t *lenp, loff_t *ppos)
6208{
6209 int *valp = ctl->data;
6210 int val = *valp;
6211 loff_t pos = *ppos;
6212 struct ctl_table lctl;
6213 int ret;
6214
6215 /*
6216 * ctl->data points to idev->cnf.forwarding, we should
6217 * not modify it until we get the rtnl lock.
6218 */
6219 lctl = *ctl;
6220 lctl.data = &val;
6221
6222 ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
6223
6224 if (write)
6225 ret = addrconf_fixup_forwarding(ctl, valp, val);
6226 if (ret)
6227 *ppos = pos;
6228 return ret;
6229}
6230
6231static int addrconf_sysctl_mtu(struct ctl_table *ctl, int write,
6232 void *buffer, size_t *lenp, loff_t *ppos)
6233{
6234 struct inet6_dev *idev = ctl->extra1;
6235 int min_mtu = IPV6_MIN_MTU;
6236 struct ctl_table lctl;
6237
6238 lctl = *ctl;
6239 lctl.extra1 = &min_mtu;
6240 lctl.extra2 = idev ? &idev->dev->mtu : NULL;
6241
6242 return proc_dointvec_minmax(&lctl, write, buffer, lenp, ppos);
6243}
6244
6245static void dev_disable_change(struct inet6_dev *idev)
6246{
6247 struct netdev_notifier_info info;
6248
6249 if (!idev || !idev->dev)
6250 return;
6251
6252 netdev_notifier_info_init(&info, idev->dev);
6253 if (idev->cnf.disable_ipv6)
6254 addrconf_notify(NULL, NETDEV_DOWN, &info);
6255 else
6256 addrconf_notify(NULL, NETDEV_UP, &info);
6257}
6258
6259static void addrconf_disable_change(struct net *net, __s32 newf)
6260{
6261 struct net_device *dev;
6262 struct inet6_dev *idev;
6263
6264 for_each_netdev(net, dev) {
6265 idev = __in6_dev_get(dev);
6266 if (idev) {
6267 int changed = (!idev->cnf.disable_ipv6) ^ (!newf);
6268 idev->cnf.disable_ipv6 = newf;
6269 if (changed)
6270 dev_disable_change(idev);
6271 }
6272 }
6273}
6274
6275static int addrconf_disable_ipv6(struct ctl_table *table, int *p, int newf)
6276{
6277 struct net *net;
6278 int old;
6279
6280 if (!rtnl_trylock())
6281 return restart_syscall();
6282
6283 net = (struct net *)table->extra2;
6284 old = *p;
6285 *p = newf;
6286
6287 if (p == &net->ipv6.devconf_dflt->disable_ipv6) {
6288 rtnl_unlock();
6289 return 0;
6290 }
6291
6292 if (p == &net->ipv6.devconf_all->disable_ipv6) {
6293 net->ipv6.devconf_dflt->disable_ipv6 = newf;
6294 addrconf_disable_change(net, newf);
6295 } else if ((!newf) ^ (!old))
6296 dev_disable_change((struct inet6_dev *)table->extra1);
6297
6298 rtnl_unlock();
6299 return 0;
6300}
6301
6302static int addrconf_sysctl_disable(struct ctl_table *ctl, int write,
6303 void *buffer, size_t *lenp, loff_t *ppos)
6304{
6305 int *valp = ctl->data;
6306 int val = *valp;
6307 loff_t pos = *ppos;
6308 struct ctl_table lctl;
6309 int ret;
6310
6311 /*
6312 * ctl->data points to idev->cnf.disable_ipv6, we should
6313 * not modify it until we get the rtnl lock.
6314 */
6315 lctl = *ctl;
6316 lctl.data = &val;
6317
6318 ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
6319
6320 if (write)
6321 ret = addrconf_disable_ipv6(ctl, valp, val);
6322 if (ret)
6323 *ppos = pos;
6324 return ret;
6325}
6326
6327static int addrconf_sysctl_proxy_ndp(struct ctl_table *ctl, int write,
6328 void *buffer, size_t *lenp, loff_t *ppos)
6329{
6330 int *valp = ctl->data;
6331 int ret;
6332 int old, new;
6333
6334 old = *valp;
6335 ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
6336 new = *valp;
6337
6338 if (write && old != new) {
6339 struct net *net = ctl->extra2;
6340
6341 if (!rtnl_trylock())
6342 return restart_syscall();
6343
6344 if (valp == &net->ipv6.devconf_dflt->proxy_ndp)
6345 inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
6346 NETCONFA_PROXY_NEIGH,
6347 NETCONFA_IFINDEX_DEFAULT,
6348 net->ipv6.devconf_dflt);
6349 else if (valp == &net->ipv6.devconf_all->proxy_ndp)
6350 inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
6351 NETCONFA_PROXY_NEIGH,
6352 NETCONFA_IFINDEX_ALL,
6353 net->ipv6.devconf_all);
6354 else {
6355 struct inet6_dev *idev = ctl->extra1;
6356
6357 inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
6358 NETCONFA_PROXY_NEIGH,
6359 idev->dev->ifindex,
6360 &idev->cnf);
6361 }
6362 rtnl_unlock();
6363 }
6364
6365 return ret;
6366}
6367
6368static int addrconf_sysctl_addr_gen_mode(struct ctl_table *ctl, int write,
6369 void *buffer, size_t *lenp,
6370 loff_t *ppos)
6371{
6372 int ret = 0;
6373 u32 new_val;
6374 struct inet6_dev *idev = (struct inet6_dev *)ctl->extra1;
6375 struct net *net = (struct net *)ctl->extra2;
6376 struct ctl_table tmp = {
6377 .data = &new_val,
6378 .maxlen = sizeof(new_val),
6379 .mode = ctl->mode,
6380 };
6381
6382 if (!rtnl_trylock())
6383 return restart_syscall();
6384
6385 new_val = *((u32 *)ctl->data);
6386
6387 ret = proc_douintvec(&tmp, write, buffer, lenp, ppos);
6388 if (ret != 0)
6389 goto out;
6390
6391 if (write) {
6392 if (check_addr_gen_mode(new_val) < 0) {
6393 ret = -EINVAL;
6394 goto out;
6395 }
6396
6397 if (idev) {
6398 if (check_stable_privacy(idev, net, new_val) < 0) {
6399 ret = -EINVAL;
6400 goto out;
6401 }
6402
6403 if (idev->cnf.addr_gen_mode != new_val) {
6404 idev->cnf.addr_gen_mode = new_val;
6405 addrconf_init_auto_addrs(idev->dev);
6406 }
6407 } else if (&net->ipv6.devconf_all->addr_gen_mode == ctl->data) {
6408 struct net_device *dev;
6409
6410 net->ipv6.devconf_dflt->addr_gen_mode = new_val;
6411 for_each_netdev(net, dev) {
6412 idev = __in6_dev_get(dev);
6413 if (idev &&
6414 idev->cnf.addr_gen_mode != new_val) {
6415 idev->cnf.addr_gen_mode = new_val;
6416 addrconf_init_auto_addrs(idev->dev);
6417 }
6418 }
6419 }
6420
6421 *((u32 *)ctl->data) = new_val;
6422 }
6423
6424out:
6425 rtnl_unlock();
6426
6427 return ret;
6428}
6429
6430static int addrconf_sysctl_stable_secret(struct ctl_table *ctl, int write,
6431 void *buffer, size_t *lenp,
6432 loff_t *ppos)
6433{
6434 int err;
6435 struct in6_addr addr;
6436 char str[IPV6_MAX_STRLEN];
6437 struct ctl_table lctl = *ctl;
6438 struct net *net = ctl->extra2;
6439 struct ipv6_stable_secret *secret = ctl->data;
6440
6441 if (&net->ipv6.devconf_all->stable_secret == ctl->data)
6442 return -EIO;
6443
6444 lctl.maxlen = IPV6_MAX_STRLEN;
6445 lctl.data = str;
6446
6447 if (!rtnl_trylock())
6448 return restart_syscall();
6449
6450 if (!write && !secret->initialized) {
6451 err = -EIO;
6452 goto out;
6453 }
6454
6455 err = snprintf(str, sizeof(str), "%pI6", &secret->secret);
6456 if (err >= sizeof(str)) {
6457 err = -EIO;
6458 goto out;
6459 }
6460
6461 err = proc_dostring(&lctl, write, buffer, lenp, ppos);
6462 if (err || !write)
6463 goto out;
6464
6465 if (in6_pton(str, -1, addr.in6_u.u6_addr8, -1, NULL) != 1) {
6466 err = -EIO;
6467 goto out;
6468 }
6469
6470 secret->initialized = true;
6471 secret->secret = addr;
6472
6473 if (&net->ipv6.devconf_dflt->stable_secret == ctl->data) {
6474 struct net_device *dev;
6475
6476 for_each_netdev(net, dev) {
6477 struct inet6_dev *idev = __in6_dev_get(dev);
6478
6479 if (idev) {
6480 idev->cnf.addr_gen_mode =
6481 IN6_ADDR_GEN_MODE_STABLE_PRIVACY;
6482 }
6483 }
6484 } else {
6485 struct inet6_dev *idev = ctl->extra1;
6486
6487 idev->cnf.addr_gen_mode = IN6_ADDR_GEN_MODE_STABLE_PRIVACY;
6488 }
6489
6490out:
6491 rtnl_unlock();
6492
6493 return err;
6494}
6495
6496static
6497int addrconf_sysctl_ignore_routes_with_linkdown(struct ctl_table *ctl,
6498 int write, void *buffer,
6499 size_t *lenp,
6500 loff_t *ppos)
6501{
6502 int *valp = ctl->data;
6503 int val = *valp;
6504 loff_t pos = *ppos;
6505 struct ctl_table lctl;
6506 int ret;
6507
6508 /* ctl->data points to idev->cnf.ignore_routes_when_linkdown
6509 * we should not modify it until we get the rtnl lock.
6510 */
6511 lctl = *ctl;
6512 lctl.data = &val;
6513
6514 ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
6515
6516 if (write)
6517 ret = addrconf_fixup_linkdown(ctl, valp, val);
6518 if (ret)
6519 *ppos = pos;
6520 return ret;
6521}
6522
6523static
6524void addrconf_set_nopolicy(struct rt6_info *rt, int action)
6525{
6526 if (rt) {
6527 if (action)
6528 rt->dst.flags |= DST_NOPOLICY;
6529 else
6530 rt->dst.flags &= ~DST_NOPOLICY;
6531 }
6532}
6533
6534static
6535void addrconf_disable_policy_idev(struct inet6_dev *idev, int val)
6536{
6537 struct inet6_ifaddr *ifa;
6538
6539 read_lock_bh(&idev->lock);
6540 list_for_each_entry(ifa, &idev->addr_list, if_list) {
6541 spin_lock(&ifa->lock);
6542 if (ifa->rt) {
6543 /* host routes only use builtin fib6_nh */
6544 struct fib6_nh *nh = ifa->rt->fib6_nh;
6545 int cpu;
6546
6547 rcu_read_lock();
6548 ifa->rt->dst_nopolicy = val ? true : false;
6549 if (nh->rt6i_pcpu) {
6550 for_each_possible_cpu(cpu) {
6551 struct rt6_info **rtp;
6552
6553 rtp = per_cpu_ptr(nh->rt6i_pcpu, cpu);
6554 addrconf_set_nopolicy(*rtp, val);
6555 }
6556 }
6557 rcu_read_unlock();
6558 }
6559 spin_unlock(&ifa->lock);
6560 }
6561 read_unlock_bh(&idev->lock);
6562}
6563
6564static
6565int addrconf_disable_policy(struct ctl_table *ctl, int *valp, int val)
6566{
6567 struct inet6_dev *idev;
6568 struct net *net;
6569
6570 if (!rtnl_trylock())
6571 return restart_syscall();
6572
6573 *valp = val;
6574
6575 net = (struct net *)ctl->extra2;
6576 if (valp == &net->ipv6.devconf_dflt->disable_policy) {
6577 rtnl_unlock();
6578 return 0;
6579 }
6580
6581 if (valp == &net->ipv6.devconf_all->disable_policy) {
6582 struct net_device *dev;
6583
6584 for_each_netdev(net, dev) {
6585 idev = __in6_dev_get(dev);
6586 if (idev)
6587 addrconf_disable_policy_idev(idev, val);
6588 }
6589 } else {
6590 idev = (struct inet6_dev *)ctl->extra1;
6591 addrconf_disable_policy_idev(idev, val);
6592 }
6593
6594 rtnl_unlock();
6595 return 0;
6596}
6597
6598static int addrconf_sysctl_disable_policy(struct ctl_table *ctl, int write,
6599 void *buffer, size_t *lenp, loff_t *ppos)
6600{
6601 int *valp = ctl->data;
6602 int val = *valp;
6603 loff_t pos = *ppos;
6604 struct ctl_table lctl;
6605 int ret;
6606
6607 lctl = *ctl;
6608 lctl.data = &val;
6609 ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
6610
6611 if (write && (*valp != val))
6612 ret = addrconf_disable_policy(ctl, valp, val);
6613
6614 if (ret)
6615 *ppos = pos;
6616
6617 return ret;
6618}
6619
6620static int minus_one = -1;
6621static const int two_five_five = 255;
6622static u32 ioam6_if_id_max = U16_MAX;
6623
6624static const struct ctl_table addrconf_sysctl[] = {
6625 {
6626 .procname = "forwarding",
6627 .data = &ipv6_devconf.forwarding,
6628 .maxlen = sizeof(int),
6629 .mode = 0644,
6630 .proc_handler = addrconf_sysctl_forward,
6631 },
6632 {
6633 .procname = "hop_limit",
6634 .data = &ipv6_devconf.hop_limit,
6635 .maxlen = sizeof(int),
6636 .mode = 0644,
6637 .proc_handler = proc_dointvec_minmax,
6638 .extra1 = (void *)SYSCTL_ONE,
6639 .extra2 = (void *)&two_five_five,
6640 },
6641 {
6642 .procname = "mtu",
6643 .data = &ipv6_devconf.mtu6,
6644 .maxlen = sizeof(int),
6645 .mode = 0644,
6646 .proc_handler = addrconf_sysctl_mtu,
6647 },
6648 {
6649 .procname = "accept_ra",
6650 .data = &ipv6_devconf.accept_ra,
6651 .maxlen = sizeof(int),
6652 .mode = 0644,
6653 .proc_handler = proc_dointvec,
6654 },
6655 {
6656 .procname = "accept_redirects",
6657 .data = &ipv6_devconf.accept_redirects,
6658 .maxlen = sizeof(int),
6659 .mode = 0644,
6660 .proc_handler = proc_dointvec,
6661 },
6662 {
6663 .procname = "autoconf",
6664 .data = &ipv6_devconf.autoconf,
6665 .maxlen = sizeof(int),
6666 .mode = 0644,
6667 .proc_handler = proc_dointvec,
6668 },
6669 {
6670 .procname = "dad_transmits",
6671 .data = &ipv6_devconf.dad_transmits,
6672 .maxlen = sizeof(int),
6673 .mode = 0644,
6674 .proc_handler = proc_dointvec,
6675 },
6676 {
6677 .procname = "router_solicitations",
6678 .data = &ipv6_devconf.rtr_solicits,
6679 .maxlen = sizeof(int),
6680 .mode = 0644,
6681 .proc_handler = proc_dointvec_minmax,
6682 .extra1 = &minus_one,
6683 },
6684 {
6685 .procname = "router_solicitation_interval",
6686 .data = &ipv6_devconf.rtr_solicit_interval,
6687 .maxlen = sizeof(int),
6688 .mode = 0644,
6689 .proc_handler = proc_dointvec_jiffies,
6690 },
6691 {
6692 .procname = "router_solicitation_max_interval",
6693 .data = &ipv6_devconf.rtr_solicit_max_interval,
6694 .maxlen = sizeof(int),
6695 .mode = 0644,
6696 .proc_handler = proc_dointvec_jiffies,
6697 },
6698 {
6699 .procname = "router_solicitation_delay",
6700 .data = &ipv6_devconf.rtr_solicit_delay,
6701 .maxlen = sizeof(int),
6702 .mode = 0644,
6703 .proc_handler = proc_dointvec_jiffies,
6704 },
6705 {
6706 .procname = "force_mld_version",
6707 .data = &ipv6_devconf.force_mld_version,
6708 .maxlen = sizeof(int),
6709 .mode = 0644,
6710 .proc_handler = proc_dointvec,
6711 },
6712 {
6713 .procname = "mldv1_unsolicited_report_interval",
6714 .data =
6715 &ipv6_devconf.mldv1_unsolicited_report_interval,
6716 .maxlen = sizeof(int),
6717 .mode = 0644,
6718 .proc_handler = proc_dointvec_ms_jiffies,
6719 },
6720 {
6721 .procname = "mldv2_unsolicited_report_interval",
6722 .data =
6723 &ipv6_devconf.mldv2_unsolicited_report_interval,
6724 .maxlen = sizeof(int),
6725 .mode = 0644,
6726 .proc_handler = proc_dointvec_ms_jiffies,
6727 },
6728 {
6729 .procname = "use_tempaddr",
6730 .data = &ipv6_devconf.use_tempaddr,
6731 .maxlen = sizeof(int),
6732 .mode = 0644,
6733 .proc_handler = proc_dointvec,
6734 },
6735 {
6736 .procname = "temp_valid_lft",
6737 .data = &ipv6_devconf.temp_valid_lft,
6738 .maxlen = sizeof(int),
6739 .mode = 0644,
6740 .proc_handler = proc_dointvec,
6741 },
6742 {
6743 .procname = "temp_prefered_lft",
6744 .data = &ipv6_devconf.temp_prefered_lft,
6745 .maxlen = sizeof(int),
6746 .mode = 0644,
6747 .proc_handler = proc_dointvec,
6748 },
6749 {
6750 .procname = "regen_max_retry",
6751 .data = &ipv6_devconf.regen_max_retry,
6752 .maxlen = sizeof(int),
6753 .mode = 0644,
6754 .proc_handler = proc_dointvec,
6755 },
6756 {
6757 .procname = "max_desync_factor",
6758 .data = &ipv6_devconf.max_desync_factor,
6759 .maxlen = sizeof(int),
6760 .mode = 0644,
6761 .proc_handler = proc_dointvec,
6762 },
6763 {
6764 .procname = "max_addresses",
6765 .data = &ipv6_devconf.max_addresses,
6766 .maxlen = sizeof(int),
6767 .mode = 0644,
6768 .proc_handler = proc_dointvec,
6769 },
6770 {
6771 .procname = "accept_ra_defrtr",
6772 .data = &ipv6_devconf.accept_ra_defrtr,
6773 .maxlen = sizeof(int),
6774 .mode = 0644,
6775 .proc_handler = proc_dointvec,
6776 },
6777 {
6778 .procname = "ra_defrtr_metric",
6779 .data = &ipv6_devconf.ra_defrtr_metric,
6780 .maxlen = sizeof(u32),
6781 .mode = 0644,
6782 .proc_handler = proc_douintvec_minmax,
6783 .extra1 = (void *)SYSCTL_ONE,
6784 },
6785 {
6786 .procname = "accept_ra_min_hop_limit",
6787 .data = &ipv6_devconf.accept_ra_min_hop_limit,
6788 .maxlen = sizeof(int),
6789 .mode = 0644,
6790 .proc_handler = proc_dointvec,
6791 },
6792 {
6793 .procname = "accept_ra_pinfo",
6794 .data = &ipv6_devconf.accept_ra_pinfo,
6795 .maxlen = sizeof(int),
6796 .mode = 0644,
6797 .proc_handler = proc_dointvec,
6798 },
6799#ifdef CONFIG_IPV6_ROUTER_PREF
6800 {
6801 .procname = "accept_ra_rtr_pref",
6802 .data = &ipv6_devconf.accept_ra_rtr_pref,
6803 .maxlen = sizeof(int),
6804 .mode = 0644,
6805 .proc_handler = proc_dointvec,
6806 },
6807 {
6808 .procname = "router_probe_interval",
6809 .data = &ipv6_devconf.rtr_probe_interval,
6810 .maxlen = sizeof(int),
6811 .mode = 0644,
6812 .proc_handler = proc_dointvec_jiffies,
6813 },
6814#ifdef CONFIG_IPV6_ROUTE_INFO
6815 {
6816 .procname = "accept_ra_rt_info_min_plen",
6817 .data = &ipv6_devconf.accept_ra_rt_info_min_plen,
6818 .maxlen = sizeof(int),
6819 .mode = 0644,
6820 .proc_handler = proc_dointvec,
6821 },
6822 {
6823 .procname = "accept_ra_rt_info_max_plen",
6824 .data = &ipv6_devconf.accept_ra_rt_info_max_plen,
6825 .maxlen = sizeof(int),
6826 .mode = 0644,
6827 .proc_handler = proc_dointvec,
6828 },
6829#endif
6830#endif
6831 {
6832 .procname = "proxy_ndp",
6833 .data = &ipv6_devconf.proxy_ndp,
6834 .maxlen = sizeof(int),
6835 .mode = 0644,
6836 .proc_handler = addrconf_sysctl_proxy_ndp,
6837 },
6838 {
6839 .procname = "accept_source_route",
6840 .data = &ipv6_devconf.accept_source_route,
6841 .maxlen = sizeof(int),
6842 .mode = 0644,
6843 .proc_handler = proc_dointvec,
6844 },
6845#ifdef CONFIG_IPV6_OPTIMISTIC_DAD
6846 {
6847 .procname = "optimistic_dad",
6848 .data = &ipv6_devconf.optimistic_dad,
6849 .maxlen = sizeof(int),
6850 .mode = 0644,
6851 .proc_handler = proc_dointvec,
6852 },
6853 {
6854 .procname = "use_optimistic",
6855 .data = &ipv6_devconf.use_optimistic,
6856 .maxlen = sizeof(int),
6857 .mode = 0644,
6858 .proc_handler = proc_dointvec,
6859 },
6860#endif
6861#ifdef CONFIG_IPV6_MROUTE
6862 {
6863 .procname = "mc_forwarding",
6864 .data = &ipv6_devconf.mc_forwarding,
6865 .maxlen = sizeof(int),
6866 .mode = 0444,
6867 .proc_handler = proc_dointvec,
6868 },
6869#endif
6870 {
6871 .procname = "disable_ipv6",
6872 .data = &ipv6_devconf.disable_ipv6,
6873 .maxlen = sizeof(int),
6874 .mode = 0644,
6875 .proc_handler = addrconf_sysctl_disable,
6876 },
6877 {
6878 .procname = "accept_dad",
6879 .data = &ipv6_devconf.accept_dad,
6880 .maxlen = sizeof(int),
6881 .mode = 0644,
6882 .proc_handler = proc_dointvec,
6883 },
6884 {
6885 .procname = "force_tllao",
6886 .data = &ipv6_devconf.force_tllao,
6887 .maxlen = sizeof(int),
6888 .mode = 0644,
6889 .proc_handler = proc_dointvec
6890 },
6891 {
6892 .procname = "ndisc_notify",
6893 .data = &ipv6_devconf.ndisc_notify,
6894 .maxlen = sizeof(int),
6895 .mode = 0644,
6896 .proc_handler = proc_dointvec
6897 },
6898 {
6899 .procname = "suppress_frag_ndisc",
6900 .data = &ipv6_devconf.suppress_frag_ndisc,
6901 .maxlen = sizeof(int),
6902 .mode = 0644,
6903 .proc_handler = proc_dointvec
6904 },
6905 {
6906 .procname = "accept_ra_from_local",
6907 .data = &ipv6_devconf.accept_ra_from_local,
6908 .maxlen = sizeof(int),
6909 .mode = 0644,
6910 .proc_handler = proc_dointvec,
6911 },
6912 {
6913 .procname = "accept_ra_mtu",
6914 .data = &ipv6_devconf.accept_ra_mtu,
6915 .maxlen = sizeof(int),
6916 .mode = 0644,
6917 .proc_handler = proc_dointvec,
6918 },
6919 {
6920 .procname = "stable_secret",
6921 .data = &ipv6_devconf.stable_secret,
6922 .maxlen = IPV6_MAX_STRLEN,
6923 .mode = 0600,
6924 .proc_handler = addrconf_sysctl_stable_secret,
6925 },
6926 {
6927 .procname = "use_oif_addrs_only",
6928 .data = &ipv6_devconf.use_oif_addrs_only,
6929 .maxlen = sizeof(int),
6930 .mode = 0644,
6931 .proc_handler = proc_dointvec,
6932 },
6933 {
6934 .procname = "ignore_routes_with_linkdown",
6935 .data = &ipv6_devconf.ignore_routes_with_linkdown,
6936 .maxlen = sizeof(int),
6937 .mode = 0644,
6938 .proc_handler = addrconf_sysctl_ignore_routes_with_linkdown,
6939 },
6940 {
6941 .procname = "drop_unicast_in_l2_multicast",
6942 .data = &ipv6_devconf.drop_unicast_in_l2_multicast,
6943 .maxlen = sizeof(int),
6944 .mode = 0644,
6945 .proc_handler = proc_dointvec,
6946 },
6947 {
6948 .procname = "drop_unsolicited_na",
6949 .data = &ipv6_devconf.drop_unsolicited_na,
6950 .maxlen = sizeof(int),
6951 .mode = 0644,
6952 .proc_handler = proc_dointvec,
6953 },
6954 {
6955 .procname = "keep_addr_on_down",
6956 .data = &ipv6_devconf.keep_addr_on_down,
6957 .maxlen = sizeof(int),
6958 .mode = 0644,
6959 .proc_handler = proc_dointvec,
6960
6961 },
6962 {
6963 .procname = "seg6_enabled",
6964 .data = &ipv6_devconf.seg6_enabled,
6965 .maxlen = sizeof(int),
6966 .mode = 0644,
6967 .proc_handler = proc_dointvec,
6968 },
6969#ifdef CONFIG_IPV6_SEG6_HMAC
6970 {
6971 .procname = "seg6_require_hmac",
6972 .data = &ipv6_devconf.seg6_require_hmac,
6973 .maxlen = sizeof(int),
6974 .mode = 0644,
6975 .proc_handler = proc_dointvec,
6976 },
6977#endif
6978 {
6979 .procname = "enhanced_dad",
6980 .data = &ipv6_devconf.enhanced_dad,
6981 .maxlen = sizeof(int),
6982 .mode = 0644,
6983 .proc_handler = proc_dointvec,
6984 },
6985 {
6986 .procname = "addr_gen_mode",
6987 .data = &ipv6_devconf.addr_gen_mode,
6988 .maxlen = sizeof(int),
6989 .mode = 0644,
6990 .proc_handler = addrconf_sysctl_addr_gen_mode,
6991 },
6992 {
6993 .procname = "disable_policy",
6994 .data = &ipv6_devconf.disable_policy,
6995 .maxlen = sizeof(int),
6996 .mode = 0644,
6997 .proc_handler = addrconf_sysctl_disable_policy,
6998 },
6999 {
7000 .procname = "ndisc_tclass",
7001 .data = &ipv6_devconf.ndisc_tclass,
7002 .maxlen = sizeof(int),
7003 .mode = 0644,
7004 .proc_handler = proc_dointvec_minmax,
7005 .extra1 = (void *)SYSCTL_ZERO,
7006 .extra2 = (void *)&two_five_five,
7007 },
7008 {
7009 .procname = "rpl_seg_enabled",
7010 .data = &ipv6_devconf.rpl_seg_enabled,
7011 .maxlen = sizeof(int),
7012 .mode = 0644,
7013 .proc_handler = proc_dointvec,
7014 },
7015 {
7016 .procname = "ioam6_enabled",
7017 .data = &ipv6_devconf.ioam6_enabled,
7018 .maxlen = sizeof(u8),
7019 .mode = 0644,
7020 .proc_handler = proc_dou8vec_minmax,
7021 .extra1 = (void *)SYSCTL_ZERO,
7022 .extra2 = (void *)SYSCTL_ONE,
7023 },
7024 {
7025 .procname = "ioam6_id",
7026 .data = &ipv6_devconf.ioam6_id,
7027 .maxlen = sizeof(u32),
7028 .mode = 0644,
7029 .proc_handler = proc_douintvec_minmax,
7030 .extra1 = (void *)SYSCTL_ZERO,
7031 .extra2 = (void *)&ioam6_if_id_max,
7032 },
7033 {
7034 .procname = "ioam6_id_wide",
7035 .data = &ipv6_devconf.ioam6_id_wide,
7036 .maxlen = sizeof(u32),
7037 .mode = 0644,
7038 .proc_handler = proc_douintvec,
7039 },
7040 {
7041 .procname = "ndisc_evict_nocarrier",
7042 .data = &ipv6_devconf.ndisc_evict_nocarrier,
7043 .maxlen = sizeof(u8),
7044 .mode = 0644,
7045 .proc_handler = proc_dou8vec_minmax,
7046 .extra1 = (void *)SYSCTL_ZERO,
7047 .extra2 = (void *)SYSCTL_ONE,
7048 },
7049 {
7050 .procname = "accept_untracked_na",
7051 .data = &ipv6_devconf.accept_untracked_na,
7052 .maxlen = sizeof(int),
7053 .mode = 0644,
7054 .proc_handler = proc_dointvec_minmax,
7055 .extra1 = SYSCTL_ZERO,
7056 .extra2 = SYSCTL_TWO,
7057 },
7058 {
7059 /* sentinel */
7060 }
7061};
7062
7063static int __addrconf_sysctl_register(struct net *net, char *dev_name,
7064 struct inet6_dev *idev, struct ipv6_devconf *p)
7065{
7066 int i, ifindex;
7067 struct ctl_table *table;
7068 char path[sizeof("net/ipv6/conf/") + IFNAMSIZ];
7069
7070 table = kmemdup(addrconf_sysctl, sizeof(addrconf_sysctl), GFP_KERNEL_ACCOUNT);
7071 if (!table)
7072 goto out;
7073
7074 for (i = 0; table[i].data; i++) {
7075 table[i].data += (char *)p - (char *)&ipv6_devconf;
7076 /* If one of these is already set, then it is not safe to
7077 * overwrite either of them: this makes proc_dointvec_minmax
7078 * usable.
7079 */
7080 if (!table[i].extra1 && !table[i].extra2) {
7081 table[i].extra1 = idev; /* embedded; no ref */
7082 table[i].extra2 = net;
7083 }
7084 }
7085
7086 snprintf(path, sizeof(path), "net/ipv6/conf/%s", dev_name);
7087
7088 p->sysctl_header = register_net_sysctl(net, path, table);
7089 if (!p->sysctl_header)
7090 goto free;
7091
7092 if (!strcmp(dev_name, "all"))
7093 ifindex = NETCONFA_IFINDEX_ALL;
7094 else if (!strcmp(dev_name, "default"))
7095 ifindex = NETCONFA_IFINDEX_DEFAULT;
7096 else
7097 ifindex = idev->dev->ifindex;
7098 inet6_netconf_notify_devconf(net, RTM_NEWNETCONF, NETCONFA_ALL,
7099 ifindex, p);
7100 return 0;
7101
7102free:
7103 kfree(table);
7104out:
7105 return -ENOBUFS;
7106}
7107
7108static void __addrconf_sysctl_unregister(struct net *net,
7109 struct ipv6_devconf *p, int ifindex)
7110{
7111 struct ctl_table *table;
7112
7113 if (!p->sysctl_header)
7114 return;
7115
7116 table = p->sysctl_header->ctl_table_arg;
7117 unregister_net_sysctl_table(p->sysctl_header);
7118 p->sysctl_header = NULL;
7119 kfree(table);
7120
7121 inet6_netconf_notify_devconf(net, RTM_DELNETCONF, 0, ifindex, NULL);
7122}
7123
7124static int addrconf_sysctl_register(struct inet6_dev *idev)
7125{
7126 int err;
7127
7128 if (!sysctl_dev_name_is_allowed(idev->dev->name))
7129 return -EINVAL;
7130
7131 err = neigh_sysctl_register(idev->dev, idev->nd_parms,
7132 &ndisc_ifinfo_sysctl_change);
7133 if (err)
7134 return err;
7135 err = __addrconf_sysctl_register(dev_net(idev->dev), idev->dev->name,
7136 idev, &idev->cnf);
7137 if (err)
7138 neigh_sysctl_unregister(idev->nd_parms);
7139
7140 return err;
7141}
7142
7143static void addrconf_sysctl_unregister(struct inet6_dev *idev)
7144{
7145 __addrconf_sysctl_unregister(dev_net(idev->dev), &idev->cnf,
7146 idev->dev->ifindex);
7147 neigh_sysctl_unregister(idev->nd_parms);
7148}
7149
7150
7151#endif
7152
7153static int __net_init addrconf_init_net(struct net *net)
7154{
7155 int err = -ENOMEM;
7156 struct ipv6_devconf *all, *dflt;
7157
7158 spin_lock_init(&net->ipv6.addrconf_hash_lock);
7159 INIT_DEFERRABLE_WORK(&net->ipv6.addr_chk_work, addrconf_verify_work);
7160 net->ipv6.inet6_addr_lst = kcalloc(IN6_ADDR_HSIZE,
7161 sizeof(struct hlist_head),
7162 GFP_KERNEL);
7163 if (!net->ipv6.inet6_addr_lst)
7164 goto err_alloc_addr;
7165
7166 all = kmemdup(&ipv6_devconf, sizeof(ipv6_devconf), GFP_KERNEL);
7167 if (!all)
7168 goto err_alloc_all;
7169
7170 dflt = kmemdup(&ipv6_devconf_dflt, sizeof(ipv6_devconf_dflt), GFP_KERNEL);
7171 if (!dflt)
7172 goto err_alloc_dflt;
7173
7174 if (!net_eq(net, &init_net)) {
7175 switch (net_inherit_devconf()) {
7176 case 1: /* copy from init_net */
7177 memcpy(all, init_net.ipv6.devconf_all,
7178 sizeof(ipv6_devconf));
7179 memcpy(dflt, init_net.ipv6.devconf_dflt,
7180 sizeof(ipv6_devconf_dflt));
7181 break;
7182 case 3: /* copy from the current netns */
7183 memcpy(all, current->nsproxy->net_ns->ipv6.devconf_all,
7184 sizeof(ipv6_devconf));
7185 memcpy(dflt,
7186 current->nsproxy->net_ns->ipv6.devconf_dflt,
7187 sizeof(ipv6_devconf_dflt));
7188 break;
7189 case 0:
7190 case 2:
7191 /* use compiled values */
7192 break;
7193 }
7194 }
7195
7196 /* these will be inherited by all namespaces */
7197 dflt->autoconf = ipv6_defaults.autoconf;
7198 dflt->disable_ipv6 = ipv6_defaults.disable_ipv6;
7199
7200 dflt->stable_secret.initialized = false;
7201 all->stable_secret.initialized = false;
7202
7203 net->ipv6.devconf_all = all;
7204 net->ipv6.devconf_dflt = dflt;
7205
7206#ifdef CONFIG_SYSCTL
7207 err = __addrconf_sysctl_register(net, "all", NULL, all);
7208 if (err < 0)
7209 goto err_reg_all;
7210
7211 err = __addrconf_sysctl_register(net, "default", NULL, dflt);
7212 if (err < 0)
7213 goto err_reg_dflt;
7214#endif
7215 return 0;
7216
7217#ifdef CONFIG_SYSCTL
7218err_reg_dflt:
7219 __addrconf_sysctl_unregister(net, all, NETCONFA_IFINDEX_ALL);
7220err_reg_all:
7221 kfree(dflt);
7222 net->ipv6.devconf_dflt = NULL;
7223#endif
7224err_alloc_dflt:
7225 kfree(all);
7226 net->ipv6.devconf_all = NULL;
7227err_alloc_all:
7228 kfree(net->ipv6.inet6_addr_lst);
7229err_alloc_addr:
7230 return err;
7231}
7232
7233static void __net_exit addrconf_exit_net(struct net *net)
7234{
7235 int i;
7236
7237#ifdef CONFIG_SYSCTL
7238 __addrconf_sysctl_unregister(net, net->ipv6.devconf_dflt,
7239 NETCONFA_IFINDEX_DEFAULT);
7240 __addrconf_sysctl_unregister(net, net->ipv6.devconf_all,
7241 NETCONFA_IFINDEX_ALL);
7242#endif
7243 kfree(net->ipv6.devconf_dflt);
7244 net->ipv6.devconf_dflt = NULL;
7245 kfree(net->ipv6.devconf_all);
7246 net->ipv6.devconf_all = NULL;
7247
7248 cancel_delayed_work_sync(&net->ipv6.addr_chk_work);
7249 /*
7250 * Check hash table, then free it.
7251 */
7252 for (i = 0; i < IN6_ADDR_HSIZE; i++)
7253 WARN_ON_ONCE(!hlist_empty(&net->ipv6.inet6_addr_lst[i]));
7254
7255 kfree(net->ipv6.inet6_addr_lst);
7256 net->ipv6.inet6_addr_lst = NULL;
7257}
7258
7259static struct pernet_operations addrconf_ops = {
7260 .init = addrconf_init_net,
7261 .exit = addrconf_exit_net,
7262};
7263
7264static struct rtnl_af_ops inet6_ops __read_mostly = {
7265 .family = AF_INET6,
7266 .fill_link_af = inet6_fill_link_af,
7267 .get_link_af_size = inet6_get_link_af_size,
7268 .validate_link_af = inet6_validate_link_af,
7269 .set_link_af = inet6_set_link_af,
7270};
7271
7272/*
7273 * Init / cleanup code
7274 */
7275
7276int __init addrconf_init(void)
7277{
7278 struct inet6_dev *idev;
7279 int err;
7280
7281 err = ipv6_addr_label_init();
7282 if (err < 0) {
7283 pr_crit("%s: cannot initialize default policy table: %d\n",
7284 __func__, err);
7285 goto out;
7286 }
7287
7288 err = register_pernet_subsys(&addrconf_ops);
7289 if (err < 0)
7290 goto out_addrlabel;
7291
7292 addrconf_wq = create_workqueue("ipv6_addrconf");
7293 if (!addrconf_wq) {
7294 err = -ENOMEM;
7295 goto out_nowq;
7296 }
7297
7298 rtnl_lock();
7299 idev = ipv6_add_dev(blackhole_netdev);
7300 rtnl_unlock();
7301 if (IS_ERR(idev)) {
7302 err = PTR_ERR(idev);
7303 goto errlo;
7304 }
7305
7306 ip6_route_init_special_entries();
7307
7308 register_netdevice_notifier(&ipv6_dev_notf);
7309
7310 addrconf_verify(&init_net);
7311
7312 rtnl_af_register(&inet6_ops);
7313
7314 err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETLINK,
7315 NULL, inet6_dump_ifinfo, 0);
7316 if (err < 0)
7317 goto errout;
7318
7319 err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_NEWADDR,
7320 inet6_rtm_newaddr, NULL, 0);
7321 if (err < 0)
7322 goto errout;
7323 err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_DELADDR,
7324 inet6_rtm_deladdr, NULL, 0);
7325 if (err < 0)
7326 goto errout;
7327 err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETADDR,
7328 inet6_rtm_getaddr, inet6_dump_ifaddr,
7329 RTNL_FLAG_DOIT_UNLOCKED);
7330 if (err < 0)
7331 goto errout;
7332 err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETMULTICAST,
7333 NULL, inet6_dump_ifmcaddr, 0);
7334 if (err < 0)
7335 goto errout;
7336 err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETANYCAST,
7337 NULL, inet6_dump_ifacaddr, 0);
7338 if (err < 0)
7339 goto errout;
7340 err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETNETCONF,
7341 inet6_netconf_get_devconf,
7342 inet6_netconf_dump_devconf,
7343 RTNL_FLAG_DOIT_UNLOCKED);
7344 if (err < 0)
7345 goto errout;
7346 err = ipv6_addr_label_rtnl_register();
7347 if (err < 0)
7348 goto errout;
7349
7350 return 0;
7351errout:
7352 rtnl_unregister_all(PF_INET6);
7353 rtnl_af_unregister(&inet6_ops);
7354 unregister_netdevice_notifier(&ipv6_dev_notf);
7355errlo:
7356 destroy_workqueue(addrconf_wq);
7357out_nowq:
7358 unregister_pernet_subsys(&addrconf_ops);
7359out_addrlabel:
7360 ipv6_addr_label_cleanup();
7361out:
7362 return err;
7363}
7364
7365void addrconf_cleanup(void)
7366{
7367 struct net_device *dev;
7368
7369 unregister_netdevice_notifier(&ipv6_dev_notf);
7370 unregister_pernet_subsys(&addrconf_ops);
7371 ipv6_addr_label_cleanup();
7372
7373 rtnl_af_unregister(&inet6_ops);
7374
7375 rtnl_lock();
7376
7377 /* clean dev list */
7378 for_each_netdev(&init_net, dev) {
7379 if (__in6_dev_get(dev) == NULL)
7380 continue;
7381 addrconf_ifdown(dev, true);
7382 }
7383 addrconf_ifdown(init_net.loopback_dev, true);
7384
7385 rtnl_unlock();
7386
7387 destroy_workqueue(addrconf_wq);
7388}