Linux kernel mirror (for testing)
git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git
kernel
os
linux
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/ip_tunnels.h>
67#include <net/net_namespace.h>
68#include <net/sock.h>
69#include <net/snmp.h>
70
71#include <net/6lowpan.h>
72#include <net/firewire.h>
73#include <net/ipv6.h>
74#include <net/protocol.h>
75#include <net/ndisc.h>
76#include <net/ip6_route.h>
77#include <net/addrconf.h>
78#include <net/tcp.h>
79#include <net/ip.h>
80#include <net/netlink.h>
81#include <net/pkt_sched.h>
82#include <net/l3mdev.h>
83#include <net/netdev_lock.h>
84#include <linux/if_tunnel.h>
85#include <linux/rtnetlink.h>
86#include <linux/netconf.h>
87#include <linux/random.h>
88#include <linux/uaccess.h>
89#include <linux/unaligned.h>
90
91#include <linux/proc_fs.h>
92#include <linux/seq_file.h>
93#include <linux/export.h>
94#include <linux/ioam6.h>
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_min_advance = REGEN_MIN_ADVANCE,
199 .regen_max_retry = REGEN_MAX_RETRY,
200 .max_desync_factor = MAX_DESYNC_FACTOR,
201 .max_addresses = IPV6_MAX_ADDRESSES,
202 .accept_ra_defrtr = 1,
203 .ra_defrtr_metric = IP6_RT_PRIO_USER,
204 .accept_ra_from_local = 0,
205 .accept_ra_min_hop_limit= 1,
206 .accept_ra_min_lft = 0,
207 .accept_ra_pinfo = 1,
208#ifdef CONFIG_IPV6_ROUTER_PREF
209 .accept_ra_rtr_pref = 1,
210 .rtr_probe_interval = 60 * HZ,
211#ifdef CONFIG_IPV6_ROUTE_INFO
212 .accept_ra_rt_info_min_plen = 0,
213 .accept_ra_rt_info_max_plen = 0,
214#endif
215#endif
216 .proxy_ndp = 0,
217 .accept_source_route = 0, /* we do not accept RH0 by default. */
218 .disable_ipv6 = 0,
219 .accept_dad = 0,
220 .suppress_frag_ndisc = 1,
221 .accept_ra_mtu = 1,
222 .stable_secret = {
223 .initialized = false,
224 },
225 .use_oif_addrs_only = 0,
226 .ignore_routes_with_linkdown = 0,
227 .keep_addr_on_down = 0,
228 .seg6_enabled = 0,
229#ifdef CONFIG_IPV6_SEG6_HMAC
230 .seg6_require_hmac = 0,
231#endif
232 .enhanced_dad = 1,
233 .addr_gen_mode = IN6_ADDR_GEN_MODE_EUI64,
234 .disable_policy = 0,
235 .rpl_seg_enabled = 0,
236 .ioam6_enabled = 0,
237 .ioam6_id = IOAM6_DEFAULT_IF_ID,
238 .ioam6_id_wide = IOAM6_DEFAULT_IF_ID_WIDE,
239 .ndisc_evict_nocarrier = 1,
240 .ra_honor_pio_life = 0,
241 .ra_honor_pio_pflag = 0,
242 .force_forwarding = 0,
243};
244
245static struct ipv6_devconf ipv6_devconf_dflt __read_mostly = {
246 .forwarding = 0,
247 .hop_limit = IPV6_DEFAULT_HOPLIMIT,
248 .mtu6 = IPV6_MIN_MTU,
249 .accept_ra = 1,
250 .accept_redirects = 1,
251 .autoconf = 1,
252 .force_mld_version = 0,
253 .mldv1_unsolicited_report_interval = 10 * HZ,
254 .mldv2_unsolicited_report_interval = HZ,
255 .dad_transmits = 1,
256 .rtr_solicits = MAX_RTR_SOLICITATIONS,
257 .rtr_solicit_interval = RTR_SOLICITATION_INTERVAL,
258 .rtr_solicit_max_interval = RTR_SOLICITATION_MAX_INTERVAL,
259 .rtr_solicit_delay = MAX_RTR_SOLICITATION_DELAY,
260 .use_tempaddr = 0,
261 .temp_valid_lft = TEMP_VALID_LIFETIME,
262 .temp_prefered_lft = TEMP_PREFERRED_LIFETIME,
263 .regen_min_advance = REGEN_MIN_ADVANCE,
264 .regen_max_retry = REGEN_MAX_RETRY,
265 .max_desync_factor = MAX_DESYNC_FACTOR,
266 .max_addresses = IPV6_MAX_ADDRESSES,
267 .accept_ra_defrtr = 1,
268 .ra_defrtr_metric = IP6_RT_PRIO_USER,
269 .accept_ra_from_local = 0,
270 .accept_ra_min_hop_limit= 1,
271 .accept_ra_min_lft = 0,
272 .accept_ra_pinfo = 1,
273#ifdef CONFIG_IPV6_ROUTER_PREF
274 .accept_ra_rtr_pref = 1,
275 .rtr_probe_interval = 60 * HZ,
276#ifdef CONFIG_IPV6_ROUTE_INFO
277 .accept_ra_rt_info_min_plen = 0,
278 .accept_ra_rt_info_max_plen = 0,
279#endif
280#endif
281 .proxy_ndp = 0,
282 .accept_source_route = 0, /* we do not accept RH0 by default. */
283 .disable_ipv6 = 0,
284 .accept_dad = 1,
285 .suppress_frag_ndisc = 1,
286 .accept_ra_mtu = 1,
287 .stable_secret = {
288 .initialized = false,
289 },
290 .use_oif_addrs_only = 0,
291 .ignore_routes_with_linkdown = 0,
292 .keep_addr_on_down = 0,
293 .seg6_enabled = 0,
294#ifdef CONFIG_IPV6_SEG6_HMAC
295 .seg6_require_hmac = 0,
296#endif
297 .enhanced_dad = 1,
298 .addr_gen_mode = IN6_ADDR_GEN_MODE_EUI64,
299 .disable_policy = 0,
300 .rpl_seg_enabled = 0,
301 .ioam6_enabled = 0,
302 .ioam6_id = IOAM6_DEFAULT_IF_ID,
303 .ioam6_id_wide = IOAM6_DEFAULT_IF_ID_WIDE,
304 .ndisc_evict_nocarrier = 1,
305 .ra_honor_pio_life = 0,
306 .ra_honor_pio_pflag = 0,
307 .force_forwarding = 0,
308};
309
310/* Check if link is ready: is it up and is a valid qdisc available */
311static inline bool addrconf_link_ready(const struct net_device *dev)
312{
313 return netif_oper_up(dev) && !qdisc_tx_is_noop(dev);
314}
315
316static void addrconf_del_rs_timer(struct inet6_dev *idev)
317{
318 if (timer_delete(&idev->rs_timer))
319 __in6_dev_put(idev);
320}
321
322static void addrconf_del_dad_work(struct inet6_ifaddr *ifp)
323{
324 if (cancel_delayed_work(&ifp->dad_work))
325 __in6_ifa_put(ifp);
326}
327
328static void addrconf_mod_rs_timer(struct inet6_dev *idev,
329 unsigned long when)
330{
331 if (!mod_timer(&idev->rs_timer, jiffies + when))
332 in6_dev_hold(idev);
333}
334
335static void addrconf_mod_dad_work(struct inet6_ifaddr *ifp,
336 unsigned long delay)
337{
338 in6_ifa_hold(ifp);
339 if (mod_delayed_work(addrconf_wq, &ifp->dad_work, delay))
340 in6_ifa_put(ifp);
341}
342
343static int snmp6_alloc_dev(struct inet6_dev *idev)
344{
345 int i;
346
347 idev->stats.ipv6 = alloc_percpu_gfp(struct ipstats_mib, GFP_KERNEL_ACCOUNT);
348 if (!idev->stats.ipv6)
349 goto err_ip;
350
351 for_each_possible_cpu(i) {
352 struct ipstats_mib *addrconf_stats;
353 addrconf_stats = per_cpu_ptr(idev->stats.ipv6, i);
354 u64_stats_init(&addrconf_stats->syncp);
355 }
356
357
358 idev->stats.icmpv6dev = kzalloc(sizeof(struct icmpv6_mib_device),
359 GFP_KERNEL);
360 if (!idev->stats.icmpv6dev)
361 goto err_icmp;
362 idev->stats.icmpv6msgdev = kzalloc(sizeof(struct icmpv6msg_mib_device),
363 GFP_KERNEL_ACCOUNT);
364 if (!idev->stats.icmpv6msgdev)
365 goto err_icmpmsg;
366
367 return 0;
368
369err_icmpmsg:
370 kfree(idev->stats.icmpv6dev);
371err_icmp:
372 free_percpu(idev->stats.ipv6);
373err_ip:
374 return -ENOMEM;
375}
376
377static struct inet6_dev *ipv6_add_dev(struct net_device *dev)
378{
379 struct inet6_dev *ndev;
380 int err = -ENOMEM;
381
382 ASSERT_RTNL();
383 netdev_ops_assert_locked(dev);
384
385 if (dev->mtu < IPV6_MIN_MTU && dev != blackhole_netdev)
386 return ERR_PTR(-EINVAL);
387
388 ndev = kzalloc(sizeof(*ndev), GFP_KERNEL_ACCOUNT);
389 if (!ndev)
390 return ERR_PTR(err);
391
392 rwlock_init(&ndev->lock);
393 ndev->dev = dev;
394 INIT_LIST_HEAD(&ndev->addr_list);
395 timer_setup(&ndev->rs_timer, addrconf_rs_timer, 0);
396 memcpy(&ndev->cnf, dev_net(dev)->ipv6.devconf_dflt, sizeof(ndev->cnf));
397
398 if (ndev->cnf.stable_secret.initialized)
399 ndev->cnf.addr_gen_mode = IN6_ADDR_GEN_MODE_STABLE_PRIVACY;
400
401 ndev->cnf.mtu6 = dev->mtu;
402 ndev->ra_mtu = 0;
403 ndev->nd_parms = neigh_parms_alloc(dev, &nd_tbl);
404 if (!ndev->nd_parms) {
405 kfree(ndev);
406 return ERR_PTR(err);
407 }
408 if (ndev->cnf.forwarding)
409 netif_disable_lro(dev);
410 /* We refer to the device */
411 netdev_hold(dev, &ndev->dev_tracker, GFP_KERNEL);
412
413 if (snmp6_alloc_dev(ndev) < 0) {
414 netdev_dbg(dev, "%s: cannot allocate memory for statistics\n",
415 __func__);
416 neigh_parms_release(&nd_tbl, ndev->nd_parms);
417 netdev_put(dev, &ndev->dev_tracker);
418 kfree(ndev);
419 return ERR_PTR(err);
420 }
421
422 if (dev != blackhole_netdev) {
423 if (snmp6_register_dev(ndev) < 0) {
424 netdev_dbg(dev, "%s: cannot create /proc/net/dev_snmp6/%s\n",
425 __func__, dev->name);
426 goto err_release;
427 }
428 }
429 /* One reference from device. */
430 refcount_set(&ndev->refcnt, 1);
431
432 if (dev->flags & (IFF_NOARP | IFF_LOOPBACK))
433 ndev->cnf.accept_dad = -1;
434
435#if IS_ENABLED(CONFIG_IPV6_SIT)
436 if (dev->type == ARPHRD_SIT && (dev->priv_flags & IFF_ISATAP)) {
437 pr_info("%s: Disabled Multicast RS\n", dev->name);
438 ndev->cnf.rtr_solicits = 0;
439 }
440#endif
441
442 INIT_LIST_HEAD(&ndev->tempaddr_list);
443 ndev->desync_factor = U32_MAX;
444 if ((dev->flags&IFF_LOOPBACK) ||
445 dev->type == ARPHRD_TUNNEL ||
446 dev->type == ARPHRD_TUNNEL6 ||
447 dev->type == ARPHRD_SIT ||
448 dev->type == ARPHRD_NONE) {
449 ndev->cnf.use_tempaddr = -1;
450 }
451
452 ndev->token = in6addr_any;
453
454 if (netif_running(dev) && addrconf_link_ready(dev))
455 ndev->if_flags |= IF_READY;
456
457 ipv6_mc_init_dev(ndev);
458 ndev->tstamp = jiffies;
459 if (dev != blackhole_netdev) {
460 err = addrconf_sysctl_register(ndev);
461 if (err) {
462 ipv6_mc_destroy_dev(ndev);
463 snmp6_unregister_dev(ndev);
464 goto err_release;
465 }
466 }
467 /* protected by rtnl_lock */
468 rcu_assign_pointer(dev->ip6_ptr, ndev);
469
470 if (dev != blackhole_netdev) {
471 /* Join interface-local all-node multicast group */
472 ipv6_dev_mc_inc(dev, &in6addr_interfacelocal_allnodes);
473
474 /* Join all-node multicast group */
475 ipv6_dev_mc_inc(dev, &in6addr_linklocal_allnodes);
476
477 /* Join all-router multicast group if forwarding is set */
478 if (ndev->cnf.forwarding && (dev->flags & IFF_MULTICAST))
479 ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
480 }
481 return ndev;
482
483err_release:
484 neigh_parms_release(&nd_tbl, ndev->nd_parms);
485 ndev->dead = 1;
486 in6_dev_finish_destroy(ndev);
487 return ERR_PTR(err);
488}
489
490static struct inet6_dev *ipv6_find_idev(struct net_device *dev)
491{
492 struct inet6_dev *idev;
493
494 ASSERT_RTNL();
495
496 idev = __in6_dev_get(dev);
497 if (!idev) {
498 idev = ipv6_add_dev(dev);
499 if (IS_ERR(idev))
500 return idev;
501 }
502
503 if (dev->flags&IFF_UP)
504 ipv6_mc_up(idev);
505 return idev;
506}
507
508static int inet6_netconf_msgsize_devconf(int type)
509{
510 int size = NLMSG_ALIGN(sizeof(struct netconfmsg))
511 + nla_total_size(4); /* NETCONFA_IFINDEX */
512 bool all = false;
513
514 if (type == NETCONFA_ALL)
515 all = true;
516
517 if (all || type == NETCONFA_FORWARDING)
518 size += nla_total_size(4);
519#ifdef CONFIG_IPV6_MROUTE
520 if (all || type == NETCONFA_MC_FORWARDING)
521 size += nla_total_size(4);
522#endif
523 if (all || type == NETCONFA_PROXY_NEIGH)
524 size += nla_total_size(4);
525
526 if (all || type == NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN)
527 size += nla_total_size(4);
528
529 return size;
530}
531
532static int inet6_netconf_fill_devconf(struct sk_buff *skb, int ifindex,
533 struct ipv6_devconf *devconf, u32 portid,
534 u32 seq, int event, unsigned int flags,
535 int type)
536{
537 struct nlmsghdr *nlh;
538 struct netconfmsg *ncm;
539 bool all = false;
540
541 nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct netconfmsg),
542 flags);
543 if (!nlh)
544 return -EMSGSIZE;
545
546 if (type == NETCONFA_ALL)
547 all = true;
548
549 ncm = nlmsg_data(nlh);
550 ncm->ncm_family = AF_INET6;
551
552 if (nla_put_s32(skb, NETCONFA_IFINDEX, ifindex) < 0)
553 goto nla_put_failure;
554
555 if (!devconf)
556 goto out;
557
558 if ((all || type == NETCONFA_FORWARDING) &&
559 nla_put_s32(skb, NETCONFA_FORWARDING,
560 READ_ONCE(devconf->forwarding)) < 0)
561 goto nla_put_failure;
562#ifdef CONFIG_IPV6_MROUTE
563 if ((all || type == NETCONFA_MC_FORWARDING) &&
564 nla_put_s32(skb, NETCONFA_MC_FORWARDING,
565 atomic_read(&devconf->mc_forwarding)) < 0)
566 goto nla_put_failure;
567#endif
568 if ((all || type == NETCONFA_PROXY_NEIGH) &&
569 nla_put_s32(skb, NETCONFA_PROXY_NEIGH,
570 READ_ONCE(devconf->proxy_ndp)) < 0)
571 goto nla_put_failure;
572
573 if ((all || type == NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN) &&
574 nla_put_s32(skb, NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
575 READ_ONCE(devconf->ignore_routes_with_linkdown)) < 0)
576 goto nla_put_failure;
577
578out:
579 nlmsg_end(skb, nlh);
580 return 0;
581
582nla_put_failure:
583 nlmsg_cancel(skb, nlh);
584 return -EMSGSIZE;
585}
586
587void inet6_netconf_notify_devconf(struct net *net, int event, int type,
588 int ifindex, struct ipv6_devconf *devconf)
589{
590 struct sk_buff *skb;
591 int err = -ENOBUFS;
592
593 skb = nlmsg_new(inet6_netconf_msgsize_devconf(type), GFP_KERNEL);
594 if (!skb)
595 goto errout;
596
597 err = inet6_netconf_fill_devconf(skb, ifindex, devconf, 0, 0,
598 event, 0, type);
599 if (err < 0) {
600 /* -EMSGSIZE implies BUG in inet6_netconf_msgsize_devconf() */
601 WARN_ON(err == -EMSGSIZE);
602 kfree_skb(skb);
603 goto errout;
604 }
605 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_NETCONF, NULL, GFP_KERNEL);
606 return;
607errout:
608 rtnl_set_sk_err(net, RTNLGRP_IPV6_NETCONF, err);
609}
610
611static const struct nla_policy devconf_ipv6_policy[NETCONFA_MAX+1] = {
612 [NETCONFA_IFINDEX] = { .len = sizeof(int) },
613 [NETCONFA_FORWARDING] = { .len = sizeof(int) },
614 [NETCONFA_PROXY_NEIGH] = { .len = sizeof(int) },
615 [NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN] = { .len = sizeof(int) },
616};
617
618static int inet6_netconf_valid_get_req(struct sk_buff *skb,
619 const struct nlmsghdr *nlh,
620 struct nlattr **tb,
621 struct netlink_ext_ack *extack)
622{
623 int i, err;
624
625 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(struct netconfmsg))) {
626 NL_SET_ERR_MSG_MOD(extack, "Invalid header for netconf get request");
627 return -EINVAL;
628 }
629
630 if (!netlink_strict_get_check(skb))
631 return nlmsg_parse_deprecated(nlh, sizeof(struct netconfmsg),
632 tb, NETCONFA_MAX,
633 devconf_ipv6_policy, extack);
634
635 err = nlmsg_parse_deprecated_strict(nlh, sizeof(struct netconfmsg),
636 tb, NETCONFA_MAX,
637 devconf_ipv6_policy, extack);
638 if (err)
639 return err;
640
641 for (i = 0; i <= NETCONFA_MAX; i++) {
642 if (!tb[i])
643 continue;
644
645 switch (i) {
646 case NETCONFA_IFINDEX:
647 break;
648 default:
649 NL_SET_ERR_MSG_MOD(extack, "Unsupported attribute in netconf get request");
650 return -EINVAL;
651 }
652 }
653
654 return 0;
655}
656
657static int inet6_netconf_get_devconf(struct sk_buff *in_skb,
658 struct nlmsghdr *nlh,
659 struct netlink_ext_ack *extack)
660{
661 struct net *net = sock_net(in_skb->sk);
662 struct nlattr *tb[NETCONFA_MAX+1];
663 struct inet6_dev *in6_dev = NULL;
664 struct net_device *dev = NULL;
665 struct sk_buff *skb;
666 struct ipv6_devconf *devconf;
667 int ifindex;
668 int err;
669
670 err = inet6_netconf_valid_get_req(in_skb, nlh, tb, extack);
671 if (err < 0)
672 return err;
673
674 if (!tb[NETCONFA_IFINDEX])
675 return -EINVAL;
676
677 err = -EINVAL;
678 ifindex = nla_get_s32(tb[NETCONFA_IFINDEX]);
679 switch (ifindex) {
680 case NETCONFA_IFINDEX_ALL:
681 devconf = net->ipv6.devconf_all;
682 break;
683 case NETCONFA_IFINDEX_DEFAULT:
684 devconf = net->ipv6.devconf_dflt;
685 break;
686 default:
687 dev = dev_get_by_index(net, ifindex);
688 if (!dev)
689 return -EINVAL;
690 in6_dev = in6_dev_get(dev);
691 if (!in6_dev)
692 goto errout;
693 devconf = &in6_dev->cnf;
694 break;
695 }
696
697 err = -ENOBUFS;
698 skb = nlmsg_new(inet6_netconf_msgsize_devconf(NETCONFA_ALL), GFP_KERNEL);
699 if (!skb)
700 goto errout;
701
702 err = inet6_netconf_fill_devconf(skb, ifindex, devconf,
703 NETLINK_CB(in_skb).portid,
704 nlh->nlmsg_seq, RTM_NEWNETCONF, 0,
705 NETCONFA_ALL);
706 if (err < 0) {
707 /* -EMSGSIZE implies BUG in inet6_netconf_msgsize_devconf() */
708 WARN_ON(err == -EMSGSIZE);
709 kfree_skb(skb);
710 goto errout;
711 }
712 err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
713errout:
714 if (in6_dev)
715 in6_dev_put(in6_dev);
716 dev_put(dev);
717 return err;
718}
719
720/* Combine dev_addr_genid and dev_base_seq to detect changes.
721 */
722static u32 inet6_base_seq(const struct net *net)
723{
724 u32 res = atomic_read(&net->ipv6.dev_addr_genid) +
725 READ_ONCE(net->dev_base_seq);
726
727 /* Must not return 0 (see nl_dump_check_consistent()).
728 * Chose a value far away from 0.
729 */
730 if (!res)
731 res = 0x80000000;
732 return res;
733}
734
735static int inet6_netconf_dump_devconf(struct sk_buff *skb,
736 struct netlink_callback *cb)
737{
738 const struct nlmsghdr *nlh = cb->nlh;
739 struct net *net = sock_net(skb->sk);
740 struct {
741 unsigned long ifindex;
742 unsigned int all_default;
743 } *ctx = (void *)cb->ctx;
744 struct net_device *dev;
745 struct inet6_dev *idev;
746 int err = 0;
747
748 if (cb->strict_check) {
749 struct netlink_ext_ack *extack = cb->extack;
750 struct netconfmsg *ncm;
751
752 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ncm))) {
753 NL_SET_ERR_MSG_MOD(extack, "Invalid header for netconf dump request");
754 return -EINVAL;
755 }
756
757 if (nlmsg_attrlen(nlh, sizeof(*ncm))) {
758 NL_SET_ERR_MSG_MOD(extack, "Invalid data after header in netconf dump request");
759 return -EINVAL;
760 }
761 }
762
763 rcu_read_lock();
764 for_each_netdev_dump(net, dev, ctx->ifindex) {
765 idev = __in6_dev_get(dev);
766 if (!idev)
767 continue;
768 err = inet6_netconf_fill_devconf(skb, dev->ifindex,
769 &idev->cnf,
770 NETLINK_CB(cb->skb).portid,
771 nlh->nlmsg_seq,
772 RTM_NEWNETCONF,
773 NLM_F_MULTI,
774 NETCONFA_ALL);
775 if (err < 0)
776 goto done;
777 }
778 if (ctx->all_default == 0) {
779 err = inet6_netconf_fill_devconf(skb, NETCONFA_IFINDEX_ALL,
780 net->ipv6.devconf_all,
781 NETLINK_CB(cb->skb).portid,
782 nlh->nlmsg_seq,
783 RTM_NEWNETCONF, NLM_F_MULTI,
784 NETCONFA_ALL);
785 if (err < 0)
786 goto done;
787 ctx->all_default++;
788 }
789 if (ctx->all_default == 1) {
790 err = inet6_netconf_fill_devconf(skb, NETCONFA_IFINDEX_DEFAULT,
791 net->ipv6.devconf_dflt,
792 NETLINK_CB(cb->skb).portid,
793 nlh->nlmsg_seq,
794 RTM_NEWNETCONF, NLM_F_MULTI,
795 NETCONFA_ALL);
796 if (err < 0)
797 goto done;
798 ctx->all_default++;
799 }
800done:
801 rcu_read_unlock();
802 return err;
803}
804
805#ifdef CONFIG_SYSCTL
806static void dev_forward_change(struct inet6_dev *idev)
807{
808 struct net_device *dev;
809 struct inet6_ifaddr *ifa;
810 LIST_HEAD(tmp_addr_list);
811
812 if (!idev)
813 return;
814 dev = idev->dev;
815 if (idev->cnf.forwarding)
816 dev_disable_lro(dev);
817 if (dev->flags & IFF_MULTICAST) {
818 if (idev->cnf.forwarding) {
819 ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
820 ipv6_dev_mc_inc(dev, &in6addr_interfacelocal_allrouters);
821 ipv6_dev_mc_inc(dev, &in6addr_sitelocal_allrouters);
822 } else {
823 ipv6_dev_mc_dec(dev, &in6addr_linklocal_allrouters);
824 ipv6_dev_mc_dec(dev, &in6addr_interfacelocal_allrouters);
825 ipv6_dev_mc_dec(dev, &in6addr_sitelocal_allrouters);
826 }
827 }
828
829 read_lock_bh(&idev->lock);
830 list_for_each_entry(ifa, &idev->addr_list, if_list) {
831 if (ifa->flags&IFA_F_TENTATIVE)
832 continue;
833 list_add_tail(&ifa->if_list_aux, &tmp_addr_list);
834 }
835 read_unlock_bh(&idev->lock);
836
837 while (!list_empty(&tmp_addr_list)) {
838 ifa = list_first_entry(&tmp_addr_list,
839 struct inet6_ifaddr, if_list_aux);
840 list_del(&ifa->if_list_aux);
841 if (idev->cnf.forwarding)
842 addrconf_join_anycast(ifa);
843 else
844 addrconf_leave_anycast(ifa);
845 }
846
847 inet6_netconf_notify_devconf(dev_net(dev), RTM_NEWNETCONF,
848 NETCONFA_FORWARDING,
849 dev->ifindex, &idev->cnf);
850}
851
852
853static void addrconf_forward_change(struct net *net, __s32 newf)
854{
855 struct net_device *dev;
856 struct inet6_dev *idev;
857
858 for_each_netdev(net, dev) {
859 idev = __in6_dev_get_rtnl_net(dev);
860 if (idev) {
861 int changed = (!idev->cnf.forwarding) ^ (!newf);
862 /* Disabling all.forwarding sets 0 to force_forwarding for all interfaces */
863 if (newf == 0)
864 WRITE_ONCE(idev->cnf.force_forwarding, 0);
865
866 WRITE_ONCE(idev->cnf.forwarding, newf);
867 if (changed)
868 dev_forward_change(idev);
869 }
870 }
871}
872
873static int addrconf_fixup_forwarding(const struct ctl_table *table, int *p, int newf)
874{
875 struct net *net = (struct net *)table->extra2;
876 int old;
877
878 if (!rtnl_net_trylock(net))
879 return restart_syscall();
880
881 old = *p;
882 WRITE_ONCE(*p, newf);
883
884 if (p == &net->ipv6.devconf_dflt->forwarding) {
885 if ((!newf) ^ (!old))
886 inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
887 NETCONFA_FORWARDING,
888 NETCONFA_IFINDEX_DEFAULT,
889 net->ipv6.devconf_dflt);
890 rtnl_net_unlock(net);
891 return 0;
892 }
893
894 if (p == &net->ipv6.devconf_all->forwarding) {
895 int old_dflt = net->ipv6.devconf_dflt->forwarding;
896
897 WRITE_ONCE(net->ipv6.devconf_dflt->forwarding, newf);
898 if ((!newf) ^ (!old_dflt))
899 inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
900 NETCONFA_FORWARDING,
901 NETCONFA_IFINDEX_DEFAULT,
902 net->ipv6.devconf_dflt);
903
904 addrconf_forward_change(net, newf);
905 if ((!newf) ^ (!old))
906 inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
907 NETCONFA_FORWARDING,
908 NETCONFA_IFINDEX_ALL,
909 net->ipv6.devconf_all);
910 } else if ((!newf) ^ (!old))
911 dev_forward_change((struct inet6_dev *)table->extra1);
912 rtnl_net_unlock(net);
913
914 if (newf)
915 rt6_purge_dflt_routers(net);
916 return 1;
917}
918
919static void addrconf_linkdown_change(struct net *net, __s32 newf)
920{
921 struct net_device *dev;
922 struct inet6_dev *idev;
923
924 for_each_netdev(net, dev) {
925 idev = __in6_dev_get_rtnl_net(dev);
926 if (idev) {
927 int changed = (!idev->cnf.ignore_routes_with_linkdown) ^ (!newf);
928
929 WRITE_ONCE(idev->cnf.ignore_routes_with_linkdown, newf);
930 if (changed)
931 inet6_netconf_notify_devconf(dev_net(dev),
932 RTM_NEWNETCONF,
933 NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
934 dev->ifindex,
935 &idev->cnf);
936 }
937 }
938}
939
940static int addrconf_fixup_linkdown(const struct ctl_table *table, int *p, int newf)
941{
942 struct net *net = (struct net *)table->extra2;
943 int old;
944
945 if (!rtnl_net_trylock(net))
946 return restart_syscall();
947
948 old = *p;
949 WRITE_ONCE(*p, newf);
950
951 if (p == &net->ipv6.devconf_dflt->ignore_routes_with_linkdown) {
952 if ((!newf) ^ (!old))
953 inet6_netconf_notify_devconf(net,
954 RTM_NEWNETCONF,
955 NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
956 NETCONFA_IFINDEX_DEFAULT,
957 net->ipv6.devconf_dflt);
958 rtnl_net_unlock(net);
959 return 0;
960 }
961
962 if (p == &net->ipv6.devconf_all->ignore_routes_with_linkdown) {
963 WRITE_ONCE(net->ipv6.devconf_dflt->ignore_routes_with_linkdown, newf);
964 addrconf_linkdown_change(net, newf);
965 if ((!newf) ^ (!old))
966 inet6_netconf_notify_devconf(net,
967 RTM_NEWNETCONF,
968 NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
969 NETCONFA_IFINDEX_ALL,
970 net->ipv6.devconf_all);
971 }
972
973 rtnl_net_unlock(net);
974
975 return 1;
976}
977
978#endif
979
980/* Nobody refers to this ifaddr, destroy it */
981void inet6_ifa_finish_destroy(struct inet6_ifaddr *ifp)
982{
983 WARN_ON(!hlist_unhashed(&ifp->addr_lst));
984
985#ifdef NET_REFCNT_DEBUG
986 pr_debug("%s\n", __func__);
987#endif
988
989 in6_dev_put(ifp->idev);
990
991 if (cancel_delayed_work(&ifp->dad_work))
992 pr_notice("delayed DAD work was pending while freeing ifa=%p\n",
993 ifp);
994
995 if (ifp->state != INET6_IFADDR_STATE_DEAD) {
996 pr_warn("Freeing alive inet6 address %p\n", ifp);
997 return;
998 }
999
1000 kfree_rcu(ifp, rcu);
1001}
1002
1003static void
1004ipv6_link_dev_addr(struct inet6_dev *idev, struct inet6_ifaddr *ifp)
1005{
1006 struct list_head *p;
1007 int ifp_scope = ipv6_addr_src_scope(&ifp->addr);
1008
1009 /*
1010 * Each device address list is sorted in order of scope -
1011 * global before linklocal.
1012 */
1013 list_for_each(p, &idev->addr_list) {
1014 struct inet6_ifaddr *ifa
1015 = list_entry(p, struct inet6_ifaddr, if_list);
1016 if (ifp_scope > ipv6_addr_src_scope(&ifa->addr))
1017 break;
1018 }
1019
1020 list_add_tail_rcu(&ifp->if_list, p);
1021}
1022
1023static u32 inet6_addr_hash(const struct net *net, const struct in6_addr *addr)
1024{
1025 u32 val = __ipv6_addr_jhash(addr, net_hash_mix(net));
1026
1027 return hash_32(val, IN6_ADDR_HSIZE_SHIFT);
1028}
1029
1030static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
1031 struct net_device *dev, unsigned int hash)
1032{
1033 struct inet6_ifaddr *ifp;
1034
1035 hlist_for_each_entry(ifp, &net->ipv6.inet6_addr_lst[hash], addr_lst) {
1036 if (ipv6_addr_equal(&ifp->addr, addr)) {
1037 if (!dev || ifp->idev->dev == dev)
1038 return true;
1039 }
1040 }
1041 return false;
1042}
1043
1044static int ipv6_add_addr_hash(struct net_device *dev, struct inet6_ifaddr *ifa)
1045{
1046 struct net *net = dev_net(dev);
1047 unsigned int hash = inet6_addr_hash(net, &ifa->addr);
1048 int err = 0;
1049
1050 spin_lock_bh(&net->ipv6.addrconf_hash_lock);
1051
1052 /* Ignore adding duplicate addresses on an interface */
1053 if (ipv6_chk_same_addr(net, &ifa->addr, dev, hash)) {
1054 netdev_dbg(dev, "ipv6_add_addr: already assigned\n");
1055 err = -EEXIST;
1056 } else {
1057 hlist_add_head_rcu(&ifa->addr_lst, &net->ipv6.inet6_addr_lst[hash]);
1058 }
1059
1060 spin_unlock_bh(&net->ipv6.addrconf_hash_lock);
1061
1062 return err;
1063}
1064
1065/* On success it returns ifp with increased reference count */
1066
1067static struct inet6_ifaddr *
1068ipv6_add_addr(struct inet6_dev *idev, struct ifa6_config *cfg,
1069 bool can_block, struct netlink_ext_ack *extack)
1070{
1071 gfp_t gfp_flags = can_block ? GFP_KERNEL : GFP_ATOMIC;
1072 int addr_type = ipv6_addr_type(cfg->pfx);
1073 struct net *net = dev_net(idev->dev);
1074 struct inet6_ifaddr *ifa = NULL;
1075 struct fib6_info *f6i = NULL;
1076 int err = 0;
1077
1078 if (addr_type == IPV6_ADDR_ANY) {
1079 NL_SET_ERR_MSG_MOD(extack, "Invalid address");
1080 return ERR_PTR(-EADDRNOTAVAIL);
1081 } else if (addr_type & IPV6_ADDR_MULTICAST &&
1082 !(cfg->ifa_flags & IFA_F_MCAUTOJOIN)) {
1083 NL_SET_ERR_MSG_MOD(extack, "Cannot assign multicast address without \"IFA_F_MCAUTOJOIN\" flag");
1084 return ERR_PTR(-EADDRNOTAVAIL);
1085 } else if (!(idev->dev->flags & IFF_LOOPBACK) &&
1086 !netif_is_l3_master(idev->dev) &&
1087 addr_type & IPV6_ADDR_LOOPBACK) {
1088 NL_SET_ERR_MSG_MOD(extack, "Cannot assign loopback address on this device");
1089 return ERR_PTR(-EADDRNOTAVAIL);
1090 }
1091
1092 if (idev->dead) {
1093 NL_SET_ERR_MSG_MOD(extack, "device is going away");
1094 err = -ENODEV;
1095 goto out;
1096 }
1097
1098 if (idev->cnf.disable_ipv6) {
1099 NL_SET_ERR_MSG_MOD(extack, "IPv6 is disabled on this device");
1100 err = -EACCES;
1101 goto out;
1102 }
1103
1104 /* validator notifier needs to be blocking;
1105 * do not call in atomic context
1106 */
1107 if (can_block) {
1108 struct in6_validator_info i6vi = {
1109 .i6vi_addr = *cfg->pfx,
1110 .i6vi_dev = idev,
1111 .extack = extack,
1112 };
1113
1114 err = inet6addr_validator_notifier_call_chain(NETDEV_UP, &i6vi);
1115 err = notifier_to_errno(err);
1116 if (err < 0)
1117 goto out;
1118 }
1119
1120 ifa = kzalloc(sizeof(*ifa), gfp_flags | __GFP_ACCOUNT);
1121 if (!ifa) {
1122 err = -ENOBUFS;
1123 goto out;
1124 }
1125
1126 f6i = addrconf_f6i_alloc(net, idev, cfg->pfx, false, gfp_flags, extack);
1127 if (IS_ERR(f6i)) {
1128 err = PTR_ERR(f6i);
1129 f6i = NULL;
1130 goto out;
1131 }
1132
1133 neigh_parms_data_state_setall(idev->nd_parms);
1134
1135 ifa->addr = *cfg->pfx;
1136 if (cfg->peer_pfx)
1137 ifa->peer_addr = *cfg->peer_pfx;
1138
1139 spin_lock_init(&ifa->lock);
1140 INIT_DELAYED_WORK(&ifa->dad_work, addrconf_dad_work);
1141 INIT_HLIST_NODE(&ifa->addr_lst);
1142 ifa->scope = cfg->scope;
1143 ifa->prefix_len = cfg->plen;
1144 ifa->rt_priority = cfg->rt_priority;
1145 ifa->flags = cfg->ifa_flags;
1146 ifa->ifa_proto = cfg->ifa_proto;
1147 /* No need to add the TENTATIVE flag for addresses with NODAD */
1148 if (!(cfg->ifa_flags & IFA_F_NODAD))
1149 ifa->flags |= IFA_F_TENTATIVE;
1150 ifa->valid_lft = cfg->valid_lft;
1151 ifa->prefered_lft = cfg->preferred_lft;
1152 ifa->cstamp = ifa->tstamp = jiffies;
1153 ifa->tokenized = false;
1154
1155 ifa->rt = f6i;
1156
1157 ifa->idev = idev;
1158 in6_dev_hold(idev);
1159
1160 /* For caller */
1161 refcount_set(&ifa->refcnt, 1);
1162
1163 rcu_read_lock();
1164
1165 err = ipv6_add_addr_hash(idev->dev, ifa);
1166 if (err < 0) {
1167 rcu_read_unlock();
1168 goto out;
1169 }
1170
1171 write_lock_bh(&idev->lock);
1172
1173 /* Add to inet6_dev unicast addr list. */
1174 ipv6_link_dev_addr(idev, ifa);
1175
1176 if (ifa->flags&IFA_F_TEMPORARY) {
1177 list_add(&ifa->tmp_list, &idev->tempaddr_list);
1178 in6_ifa_hold(ifa);
1179 }
1180
1181 in6_ifa_hold(ifa);
1182 write_unlock_bh(&idev->lock);
1183
1184 rcu_read_unlock();
1185
1186 inet6addr_notifier_call_chain(NETDEV_UP, ifa);
1187out:
1188 if (unlikely(err < 0)) {
1189 fib6_info_release(f6i);
1190
1191 if (ifa) {
1192 if (ifa->idev)
1193 in6_dev_put(ifa->idev);
1194 kfree(ifa);
1195 }
1196 ifa = ERR_PTR(err);
1197 }
1198
1199 return ifa;
1200}
1201
1202enum cleanup_prefix_rt_t {
1203 CLEANUP_PREFIX_RT_NOP, /* no cleanup action for prefix route */
1204 CLEANUP_PREFIX_RT_DEL, /* delete the prefix route */
1205 CLEANUP_PREFIX_RT_EXPIRE, /* update the lifetime of the prefix route */
1206};
1207
1208/*
1209 * Check, whether the prefix for ifp would still need a prefix route
1210 * after deleting ifp. The function returns one of the CLEANUP_PREFIX_RT_*
1211 * constants.
1212 *
1213 * 1) we don't purge prefix if address was not permanent.
1214 * prefix is managed by its own lifetime.
1215 * 2) we also don't purge, if the address was IFA_F_NOPREFIXROUTE.
1216 * 3) if there are no addresses, delete prefix.
1217 * 4) if there are still other permanent address(es),
1218 * corresponding prefix is still permanent.
1219 * 5) if there are still other addresses with IFA_F_NOPREFIXROUTE,
1220 * don't purge the prefix, assume user space is managing it.
1221 * 6) otherwise, update prefix lifetime to the
1222 * longest valid lifetime among the corresponding
1223 * addresses on the device.
1224 * Note: subsequent RA will update lifetime.
1225 **/
1226static enum cleanup_prefix_rt_t
1227check_cleanup_prefix_route(struct inet6_ifaddr *ifp, unsigned long *expires)
1228{
1229 struct inet6_ifaddr *ifa;
1230 struct inet6_dev *idev = ifp->idev;
1231 unsigned long lifetime;
1232 enum cleanup_prefix_rt_t action = CLEANUP_PREFIX_RT_DEL;
1233
1234 *expires = jiffies;
1235
1236 list_for_each_entry(ifa, &idev->addr_list, if_list) {
1237 if (ifa == ifp)
1238 continue;
1239 if (ifa->prefix_len != ifp->prefix_len ||
1240 !ipv6_prefix_equal(&ifa->addr, &ifp->addr,
1241 ifp->prefix_len))
1242 continue;
1243 if (ifa->flags & (IFA_F_PERMANENT | IFA_F_NOPREFIXROUTE))
1244 return CLEANUP_PREFIX_RT_NOP;
1245
1246 action = CLEANUP_PREFIX_RT_EXPIRE;
1247
1248 spin_lock(&ifa->lock);
1249
1250 lifetime = addrconf_timeout_fixup(ifa->valid_lft, HZ);
1251 /*
1252 * Note: Because this address is
1253 * not permanent, lifetime <
1254 * LONG_MAX / HZ here.
1255 */
1256 if (time_before(*expires, ifa->tstamp + lifetime * HZ))
1257 *expires = ifa->tstamp + lifetime * HZ;
1258 spin_unlock(&ifa->lock);
1259 }
1260
1261 return action;
1262}
1263
1264static void
1265cleanup_prefix_route(struct inet6_ifaddr *ifp, unsigned long expires,
1266 bool del_rt, bool del_peer)
1267{
1268 struct fib6_table *table;
1269 struct fib6_info *f6i;
1270
1271 f6i = addrconf_get_prefix_route(del_peer ? &ifp->peer_addr : &ifp->addr,
1272 ifp->prefix_len,
1273 ifp->idev->dev, 0, RTF_DEFAULT, true);
1274 if (f6i) {
1275 if (del_rt)
1276 ip6_del_rt(dev_net(ifp->idev->dev), f6i, false);
1277 else {
1278 if (!(f6i->fib6_flags & RTF_EXPIRES)) {
1279 table = f6i->fib6_table;
1280 spin_lock_bh(&table->tb6_lock);
1281
1282 fib6_set_expires(f6i, expires);
1283 fib6_add_gc_list(f6i);
1284
1285 spin_unlock_bh(&table->tb6_lock);
1286 }
1287 fib6_info_release(f6i);
1288 }
1289 }
1290}
1291
1292
1293/* This function wants to get referenced ifp and releases it before return */
1294
1295static void ipv6_del_addr(struct inet6_ifaddr *ifp)
1296{
1297 enum cleanup_prefix_rt_t action = CLEANUP_PREFIX_RT_NOP;
1298 struct net *net = dev_net(ifp->idev->dev);
1299 unsigned long expires;
1300 int state;
1301
1302 ASSERT_RTNL();
1303
1304 spin_lock_bh(&ifp->lock);
1305 state = ifp->state;
1306 ifp->state = INET6_IFADDR_STATE_DEAD;
1307 spin_unlock_bh(&ifp->lock);
1308
1309 if (state == INET6_IFADDR_STATE_DEAD)
1310 goto out;
1311
1312 spin_lock_bh(&net->ipv6.addrconf_hash_lock);
1313 hlist_del_init_rcu(&ifp->addr_lst);
1314 spin_unlock_bh(&net->ipv6.addrconf_hash_lock);
1315
1316 write_lock_bh(&ifp->idev->lock);
1317
1318 if (ifp->flags&IFA_F_TEMPORARY) {
1319 list_del(&ifp->tmp_list);
1320 if (ifp->ifpub) {
1321 in6_ifa_put(ifp->ifpub);
1322 ifp->ifpub = NULL;
1323 }
1324 __in6_ifa_put(ifp);
1325 }
1326
1327 if (!(ifp->flags & IFA_F_NOPREFIXROUTE))
1328 action = check_cleanup_prefix_route(ifp, &expires);
1329
1330 list_del_rcu(&ifp->if_list);
1331 __in6_ifa_put(ifp);
1332
1333 write_unlock_bh(&ifp->idev->lock);
1334
1335 addrconf_del_dad_work(ifp);
1336
1337 ipv6_ifa_notify(RTM_DELADDR, ifp);
1338
1339 inet6addr_notifier_call_chain(NETDEV_DOWN, ifp);
1340
1341 if (action != CLEANUP_PREFIX_RT_NOP) {
1342 cleanup_prefix_route(ifp, expires,
1343 action == CLEANUP_PREFIX_RT_DEL, false);
1344 }
1345
1346 /* clean up prefsrc entries */
1347 rt6_remove_prefsrc(ifp);
1348out:
1349 in6_ifa_put(ifp);
1350}
1351
1352static unsigned long ipv6_get_regen_advance(const struct inet6_dev *idev)
1353{
1354 return READ_ONCE(idev->cnf.regen_min_advance) +
1355 READ_ONCE(idev->cnf.regen_max_retry) *
1356 READ_ONCE(idev->cnf.dad_transmits) *
1357 max(NEIGH_VAR(idev->nd_parms, RETRANS_TIME), HZ/100) / HZ;
1358}
1359
1360static int ipv6_create_tempaddr(struct inet6_ifaddr *ifp, bool block)
1361{
1362 struct inet6_dev *idev = ifp->idev;
1363 unsigned long tmp_tstamp, age;
1364 unsigned long regen_advance;
1365 unsigned long now = jiffies;
1366 u32 if_public_preferred_lft;
1367 s32 cnf_temp_preferred_lft;
1368 struct inet6_ifaddr *ift;
1369 struct ifa6_config cfg;
1370 long max_desync_factor;
1371 struct in6_addr addr;
1372 int ret = 0;
1373
1374 write_lock_bh(&idev->lock);
1375
1376retry:
1377 in6_dev_hold(idev);
1378 if (READ_ONCE(idev->cnf.use_tempaddr) <= 0) {
1379 write_unlock_bh(&idev->lock);
1380 pr_info("%s: use_tempaddr is disabled\n", __func__);
1381 in6_dev_put(idev);
1382 ret = -1;
1383 goto out;
1384 }
1385 spin_lock_bh(&ifp->lock);
1386 if (ifp->regen_count++ >= READ_ONCE(idev->cnf.regen_max_retry)) {
1387 WRITE_ONCE(idev->cnf.use_tempaddr, -1); /*XXX*/
1388 spin_unlock_bh(&ifp->lock);
1389 write_unlock_bh(&idev->lock);
1390 pr_warn("%s: regeneration time exceeded - disabled temporary address support\n",
1391 __func__);
1392 in6_dev_put(idev);
1393 ret = -1;
1394 goto out;
1395 }
1396 in6_ifa_hold(ifp);
1397 memcpy(addr.s6_addr, ifp->addr.s6_addr, 8);
1398 ipv6_gen_rnd_iid(&addr);
1399
1400 age = (now - ifp->tstamp) / HZ;
1401
1402 regen_advance = ipv6_get_regen_advance(idev);
1403
1404 /* recalculate max_desync_factor each time and update
1405 * idev->desync_factor if it's larger
1406 */
1407 cnf_temp_preferred_lft = READ_ONCE(idev->cnf.temp_prefered_lft);
1408 max_desync_factor = min_t(long,
1409 READ_ONCE(idev->cnf.max_desync_factor),
1410 cnf_temp_preferred_lft - regen_advance);
1411
1412 if (unlikely(idev->desync_factor > max_desync_factor)) {
1413 if (max_desync_factor > 0) {
1414 get_random_bytes(&idev->desync_factor,
1415 sizeof(idev->desync_factor));
1416 idev->desync_factor %= max_desync_factor;
1417 } else {
1418 idev->desync_factor = 0;
1419 }
1420 }
1421
1422 if_public_preferred_lft = ifp->prefered_lft;
1423
1424 memset(&cfg, 0, sizeof(cfg));
1425 cfg.valid_lft = min_t(__u32, ifp->valid_lft,
1426 READ_ONCE(idev->cnf.temp_valid_lft) + age);
1427 cfg.preferred_lft = cnf_temp_preferred_lft + age - idev->desync_factor;
1428 cfg.preferred_lft = min_t(__u32, if_public_preferred_lft, cfg.preferred_lft);
1429 cfg.preferred_lft = min_t(__u32, cfg.valid_lft, cfg.preferred_lft);
1430
1431 cfg.plen = ifp->prefix_len;
1432 tmp_tstamp = ifp->tstamp;
1433 spin_unlock_bh(&ifp->lock);
1434
1435 write_unlock_bh(&idev->lock);
1436
1437 /* From RFC 4941:
1438 *
1439 * A temporary address is created only if this calculated Preferred
1440 * Lifetime is greater than REGEN_ADVANCE time units. In
1441 * particular, an implementation must not create a temporary address
1442 * with a zero Preferred Lifetime.
1443 *
1444 * ...
1445 *
1446 * When creating a temporary address, the lifetime values MUST be
1447 * derived from the corresponding prefix as follows:
1448 *
1449 * ...
1450 *
1451 * * Its Preferred Lifetime is the lower of the Preferred Lifetime
1452 * of the public address or TEMP_PREFERRED_LIFETIME -
1453 * DESYNC_FACTOR.
1454 *
1455 * To comply with the RFC's requirements, clamp the preferred lifetime
1456 * to a minimum of regen_advance, unless that would exceed valid_lft or
1457 * ifp->prefered_lft.
1458 *
1459 * Use age calculation as in addrconf_verify to avoid unnecessary
1460 * temporary addresses being generated.
1461 */
1462 age = (now - tmp_tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
1463 if (cfg.preferred_lft <= regen_advance + age) {
1464 cfg.preferred_lft = regen_advance + age + 1;
1465 if (cfg.preferred_lft > cfg.valid_lft ||
1466 cfg.preferred_lft > if_public_preferred_lft) {
1467 in6_ifa_put(ifp);
1468 in6_dev_put(idev);
1469 ret = -1;
1470 goto out;
1471 }
1472 }
1473
1474 cfg.ifa_flags = IFA_F_TEMPORARY;
1475 /* set in addrconf_prefix_rcv() */
1476 if (ifp->flags & IFA_F_OPTIMISTIC)
1477 cfg.ifa_flags |= IFA_F_OPTIMISTIC;
1478
1479 cfg.pfx = &addr;
1480 cfg.scope = ipv6_addr_scope(cfg.pfx);
1481
1482 ift = ipv6_add_addr(idev, &cfg, block, NULL);
1483 if (IS_ERR(ift)) {
1484 in6_ifa_put(ifp);
1485 in6_dev_put(idev);
1486 pr_info("%s: retry temporary address regeneration\n", __func__);
1487 write_lock_bh(&idev->lock);
1488 goto retry;
1489 }
1490
1491 spin_lock_bh(&ift->lock);
1492 ift->ifpub = ifp;
1493 ift->cstamp = now;
1494 ift->tstamp = tmp_tstamp;
1495 spin_unlock_bh(&ift->lock);
1496
1497 addrconf_dad_start(ift);
1498 in6_ifa_put(ift);
1499 in6_dev_put(idev);
1500out:
1501 return ret;
1502}
1503
1504/*
1505 * Choose an appropriate source address (RFC3484)
1506 */
1507enum {
1508 IPV6_SADDR_RULE_INIT = 0,
1509 IPV6_SADDR_RULE_LOCAL,
1510 IPV6_SADDR_RULE_SCOPE,
1511 IPV6_SADDR_RULE_PREFERRED,
1512#ifdef CONFIG_IPV6_MIP6
1513 IPV6_SADDR_RULE_HOA,
1514#endif
1515 IPV6_SADDR_RULE_OIF,
1516 IPV6_SADDR_RULE_LABEL,
1517 IPV6_SADDR_RULE_PRIVACY,
1518 IPV6_SADDR_RULE_ORCHID,
1519 IPV6_SADDR_RULE_PREFIX,
1520#ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1521 IPV6_SADDR_RULE_NOT_OPTIMISTIC,
1522#endif
1523 IPV6_SADDR_RULE_MAX
1524};
1525
1526struct ipv6_saddr_score {
1527 int rule;
1528 int addr_type;
1529 struct inet6_ifaddr *ifa;
1530 DECLARE_BITMAP(scorebits, IPV6_SADDR_RULE_MAX);
1531 int scopedist;
1532 int matchlen;
1533};
1534
1535struct ipv6_saddr_dst {
1536 const struct in6_addr *addr;
1537 int ifindex;
1538 int scope;
1539 int label;
1540 unsigned int prefs;
1541};
1542
1543static inline int ipv6_saddr_preferred(int type)
1544{
1545 if (type & (IPV6_ADDR_MAPPED|IPV6_ADDR_COMPATv4|IPV6_ADDR_LOOPBACK))
1546 return 1;
1547 return 0;
1548}
1549
1550static bool ipv6_use_optimistic_addr(const struct net *net,
1551 const struct inet6_dev *idev)
1552{
1553#ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1554 if (!idev)
1555 return false;
1556 if (!READ_ONCE(net->ipv6.devconf_all->optimistic_dad) &&
1557 !READ_ONCE(idev->cnf.optimistic_dad))
1558 return false;
1559 if (!READ_ONCE(net->ipv6.devconf_all->use_optimistic) &&
1560 !READ_ONCE(idev->cnf.use_optimistic))
1561 return false;
1562
1563 return true;
1564#else
1565 return false;
1566#endif
1567}
1568
1569static bool ipv6_allow_optimistic_dad(const struct net *net,
1570 const struct inet6_dev *idev)
1571{
1572#ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1573 if (!idev)
1574 return false;
1575 if (!READ_ONCE(net->ipv6.devconf_all->optimistic_dad) &&
1576 !READ_ONCE(idev->cnf.optimistic_dad))
1577 return false;
1578
1579 return true;
1580#else
1581 return false;
1582#endif
1583}
1584
1585static int ipv6_get_saddr_eval(struct net *net,
1586 struct ipv6_saddr_score *score,
1587 struct ipv6_saddr_dst *dst,
1588 int i)
1589{
1590 int ret;
1591
1592 if (i <= score->rule) {
1593 switch (i) {
1594 case IPV6_SADDR_RULE_SCOPE:
1595 ret = score->scopedist;
1596 break;
1597 case IPV6_SADDR_RULE_PREFIX:
1598 ret = score->matchlen;
1599 break;
1600 default:
1601 ret = !!test_bit(i, score->scorebits);
1602 }
1603 goto out;
1604 }
1605
1606 switch (i) {
1607 case IPV6_SADDR_RULE_INIT:
1608 /* Rule 0: remember if hiscore is not ready yet */
1609 ret = !!score->ifa;
1610 break;
1611 case IPV6_SADDR_RULE_LOCAL:
1612 /* Rule 1: Prefer same address */
1613 ret = ipv6_addr_equal(&score->ifa->addr, dst->addr);
1614 break;
1615 case IPV6_SADDR_RULE_SCOPE:
1616 /* Rule 2: Prefer appropriate scope
1617 *
1618 * ret
1619 * ^
1620 * -1 | d 15
1621 * ---+--+-+---> scope
1622 * |
1623 * | d is scope of the destination.
1624 * B-d | \
1625 * | \ <- smaller scope is better if
1626 * B-15 | \ if scope is enough for destination.
1627 * | ret = B - scope (-1 <= scope >= d <= 15).
1628 * d-C-1 | /
1629 * |/ <- greater is better
1630 * -C / if scope is not enough for destination.
1631 * /| ret = scope - C (-1 <= d < scope <= 15).
1632 *
1633 * d - C - 1 < B -15 (for all -1 <= d <= 15).
1634 * C > d + 14 - B >= 15 + 14 - B = 29 - B.
1635 * Assume B = 0 and we get C > 29.
1636 */
1637 ret = __ipv6_addr_src_scope(score->addr_type);
1638 if (ret >= dst->scope)
1639 ret = -ret;
1640 else
1641 ret -= 128; /* 30 is enough */
1642 score->scopedist = ret;
1643 break;
1644 case IPV6_SADDR_RULE_PREFERRED:
1645 {
1646 /* Rule 3: Avoid deprecated and optimistic addresses */
1647 u8 avoid = IFA_F_DEPRECATED;
1648
1649 if (!ipv6_use_optimistic_addr(net, score->ifa->idev))
1650 avoid |= IFA_F_OPTIMISTIC;
1651 ret = ipv6_saddr_preferred(score->addr_type) ||
1652 !(score->ifa->flags & avoid);
1653 break;
1654 }
1655#ifdef CONFIG_IPV6_MIP6
1656 case IPV6_SADDR_RULE_HOA:
1657 {
1658 /* Rule 4: Prefer home address */
1659 int prefhome = !(dst->prefs & IPV6_PREFER_SRC_COA);
1660 ret = !(score->ifa->flags & IFA_F_HOMEADDRESS) ^ prefhome;
1661 break;
1662 }
1663#endif
1664 case IPV6_SADDR_RULE_OIF:
1665 /* Rule 5: Prefer outgoing interface */
1666 ret = (!dst->ifindex ||
1667 dst->ifindex == score->ifa->idev->dev->ifindex);
1668 break;
1669 case IPV6_SADDR_RULE_LABEL:
1670 /* Rule 6: Prefer matching label */
1671 ret = ipv6_addr_label(net,
1672 &score->ifa->addr, score->addr_type,
1673 score->ifa->idev->dev->ifindex) == dst->label;
1674 break;
1675 case IPV6_SADDR_RULE_PRIVACY:
1676 {
1677 /* Rule 7: Prefer public address
1678 * Note: prefer temporary address if use_tempaddr >= 2
1679 */
1680 int preftmp = dst->prefs & (IPV6_PREFER_SRC_PUBLIC|IPV6_PREFER_SRC_TMP) ?
1681 !!(dst->prefs & IPV6_PREFER_SRC_TMP) :
1682 READ_ONCE(score->ifa->idev->cnf.use_tempaddr) >= 2;
1683 ret = (!(score->ifa->flags & IFA_F_TEMPORARY)) ^ preftmp;
1684 break;
1685 }
1686 case IPV6_SADDR_RULE_ORCHID:
1687 /* Rule 8-: Prefer ORCHID vs ORCHID or
1688 * non-ORCHID vs non-ORCHID
1689 */
1690 ret = !(ipv6_addr_orchid(&score->ifa->addr) ^
1691 ipv6_addr_orchid(dst->addr));
1692 break;
1693 case IPV6_SADDR_RULE_PREFIX:
1694 /* Rule 8: Use longest matching prefix */
1695 ret = ipv6_addr_diff(&score->ifa->addr, dst->addr);
1696 if (ret > score->ifa->prefix_len)
1697 ret = score->ifa->prefix_len;
1698 score->matchlen = ret;
1699 break;
1700#ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1701 case IPV6_SADDR_RULE_NOT_OPTIMISTIC:
1702 /* Optimistic addresses still have lower precedence than other
1703 * preferred addresses.
1704 */
1705 ret = !(score->ifa->flags & IFA_F_OPTIMISTIC);
1706 break;
1707#endif
1708 default:
1709 ret = 0;
1710 }
1711
1712 if (ret)
1713 __set_bit(i, score->scorebits);
1714 score->rule = i;
1715out:
1716 return ret;
1717}
1718
1719static int __ipv6_dev_get_saddr(struct net *net,
1720 struct ipv6_saddr_dst *dst,
1721 struct inet6_dev *idev,
1722 struct ipv6_saddr_score *scores,
1723 int hiscore_idx)
1724{
1725 struct ipv6_saddr_score *score = &scores[1 - hiscore_idx], *hiscore = &scores[hiscore_idx];
1726
1727 list_for_each_entry_rcu(score->ifa, &idev->addr_list, if_list) {
1728 int i;
1729
1730 /*
1731 * - Tentative Address (RFC2462 section 5.4)
1732 * - A tentative address is not considered
1733 * "assigned to an interface" in the traditional
1734 * sense, unless it is also flagged as optimistic.
1735 * - Candidate Source Address (section 4)
1736 * - In any case, anycast addresses, multicast
1737 * addresses, and the unspecified address MUST
1738 * NOT be included in a candidate set.
1739 */
1740 if ((score->ifa->flags & IFA_F_TENTATIVE) &&
1741 (!(score->ifa->flags & IFA_F_OPTIMISTIC)))
1742 continue;
1743
1744 score->addr_type = __ipv6_addr_type(&score->ifa->addr);
1745
1746 if (unlikely(score->addr_type == IPV6_ADDR_ANY ||
1747 score->addr_type & IPV6_ADDR_MULTICAST)) {
1748 net_dbg_ratelimited("ADDRCONF: unspecified / multicast address assigned as unicast address on %s",
1749 idev->dev->name);
1750 continue;
1751 }
1752
1753 score->rule = -1;
1754 bitmap_zero(score->scorebits, IPV6_SADDR_RULE_MAX);
1755
1756 for (i = 0; i < IPV6_SADDR_RULE_MAX; i++) {
1757 int minihiscore, miniscore;
1758
1759 minihiscore = ipv6_get_saddr_eval(net, hiscore, dst, i);
1760 miniscore = ipv6_get_saddr_eval(net, score, dst, i);
1761
1762 if (minihiscore > miniscore) {
1763 if (i == IPV6_SADDR_RULE_SCOPE &&
1764 score->scopedist > 0) {
1765 /*
1766 * special case:
1767 * each remaining entry
1768 * has too small (not enough)
1769 * scope, because ifa entries
1770 * are sorted by their scope
1771 * values.
1772 */
1773 goto out;
1774 }
1775 break;
1776 } else if (minihiscore < miniscore) {
1777 swap(hiscore, score);
1778 hiscore_idx = 1 - hiscore_idx;
1779
1780 /* restore our iterator */
1781 score->ifa = hiscore->ifa;
1782
1783 break;
1784 }
1785 }
1786 }
1787out:
1788 return hiscore_idx;
1789}
1790
1791static int ipv6_get_saddr_master(struct net *net,
1792 const struct net_device *dst_dev,
1793 const struct net_device *master,
1794 struct ipv6_saddr_dst *dst,
1795 struct ipv6_saddr_score *scores,
1796 int hiscore_idx)
1797{
1798 struct inet6_dev *idev;
1799
1800 idev = __in6_dev_get(dst_dev);
1801 if (idev)
1802 hiscore_idx = __ipv6_dev_get_saddr(net, dst, idev,
1803 scores, hiscore_idx);
1804
1805 idev = __in6_dev_get(master);
1806 if (idev)
1807 hiscore_idx = __ipv6_dev_get_saddr(net, dst, idev,
1808 scores, hiscore_idx);
1809
1810 return hiscore_idx;
1811}
1812
1813int ipv6_dev_get_saddr(struct net *net, const struct net_device *dst_dev,
1814 const struct in6_addr *daddr, unsigned int prefs,
1815 struct in6_addr *saddr)
1816{
1817 struct ipv6_saddr_score scores[2], *hiscore;
1818 struct ipv6_saddr_dst dst;
1819 struct inet6_dev *idev;
1820 struct net_device *dev;
1821 int dst_type;
1822 bool use_oif_addr = false;
1823 int hiscore_idx = 0;
1824 int ret = 0;
1825
1826 dst_type = __ipv6_addr_type(daddr);
1827 dst.addr = daddr;
1828 dst.ifindex = dst_dev ? dst_dev->ifindex : 0;
1829 dst.scope = __ipv6_addr_src_scope(dst_type);
1830 dst.label = ipv6_addr_label(net, daddr, dst_type, dst.ifindex);
1831 dst.prefs = prefs;
1832
1833 scores[hiscore_idx].rule = -1;
1834 scores[hiscore_idx].ifa = NULL;
1835
1836 rcu_read_lock();
1837
1838 /* Candidate Source Address (section 4)
1839 * - multicast and link-local destination address,
1840 * the set of candidate source address MUST only
1841 * include addresses assigned to interfaces
1842 * belonging to the same link as the outgoing
1843 * interface.
1844 * (- For site-local destination addresses, the
1845 * set of candidate source addresses MUST only
1846 * include addresses assigned to interfaces
1847 * belonging to the same site as the outgoing
1848 * interface.)
1849 * - "It is RECOMMENDED that the candidate source addresses
1850 * be the set of unicast addresses assigned to the
1851 * interface that will be used to send to the destination
1852 * (the 'outgoing' interface)." (RFC 6724)
1853 */
1854 if (dst_dev) {
1855 idev = __in6_dev_get(dst_dev);
1856 if ((dst_type & IPV6_ADDR_MULTICAST) ||
1857 dst.scope <= IPV6_ADDR_SCOPE_LINKLOCAL ||
1858 (idev && READ_ONCE(idev->cnf.use_oif_addrs_only))) {
1859 use_oif_addr = true;
1860 }
1861 }
1862
1863 if (use_oif_addr) {
1864 if (idev)
1865 hiscore_idx = __ipv6_dev_get_saddr(net, &dst, idev, scores, hiscore_idx);
1866 } else {
1867 const struct net_device *master;
1868 int master_idx = 0;
1869
1870 /* if dst_dev exists and is enslaved to an L3 device, then
1871 * prefer addresses from dst_dev and then the master over
1872 * any other enslaved devices in the L3 domain.
1873 */
1874 master = l3mdev_master_dev_rcu(dst_dev);
1875 if (master) {
1876 master_idx = master->ifindex;
1877
1878 hiscore_idx = ipv6_get_saddr_master(net, dst_dev,
1879 master, &dst,
1880 scores, hiscore_idx);
1881
1882 if (scores[hiscore_idx].ifa &&
1883 scores[hiscore_idx].scopedist >= 0)
1884 goto out;
1885 }
1886
1887 for_each_netdev_rcu(net, dev) {
1888 /* only consider addresses on devices in the
1889 * same L3 domain
1890 */
1891 if (l3mdev_master_ifindex_rcu(dev) != master_idx)
1892 continue;
1893 idev = __in6_dev_get(dev);
1894 if (!idev)
1895 continue;
1896 hiscore_idx = __ipv6_dev_get_saddr(net, &dst, idev, scores, hiscore_idx);
1897 }
1898 }
1899
1900out:
1901 hiscore = &scores[hiscore_idx];
1902 if (!hiscore->ifa)
1903 ret = -EADDRNOTAVAIL;
1904 else
1905 *saddr = hiscore->ifa->addr;
1906
1907 rcu_read_unlock();
1908 return ret;
1909}
1910EXPORT_SYMBOL(ipv6_dev_get_saddr);
1911
1912static int __ipv6_get_lladdr(struct inet6_dev *idev, struct in6_addr *addr,
1913 u32 banned_flags)
1914{
1915 struct inet6_ifaddr *ifp;
1916 int err = -EADDRNOTAVAIL;
1917
1918 list_for_each_entry_reverse(ifp, &idev->addr_list, if_list) {
1919 if (ifp->scope > IFA_LINK)
1920 break;
1921 if (ifp->scope == IFA_LINK &&
1922 !(ifp->flags & banned_flags)) {
1923 *addr = ifp->addr;
1924 err = 0;
1925 break;
1926 }
1927 }
1928 return err;
1929}
1930
1931int ipv6_get_lladdr(struct net_device *dev, struct in6_addr *addr,
1932 u32 banned_flags)
1933{
1934 struct inet6_dev *idev;
1935 int err = -EADDRNOTAVAIL;
1936
1937 rcu_read_lock();
1938 idev = __in6_dev_get(dev);
1939 if (idev) {
1940 read_lock_bh(&idev->lock);
1941 err = __ipv6_get_lladdr(idev, addr, banned_flags);
1942 read_unlock_bh(&idev->lock);
1943 }
1944 rcu_read_unlock();
1945 return err;
1946}
1947
1948static int ipv6_count_addresses(const struct inet6_dev *idev)
1949{
1950 const struct inet6_ifaddr *ifp;
1951 int cnt = 0;
1952
1953 rcu_read_lock();
1954 list_for_each_entry_rcu(ifp, &idev->addr_list, if_list)
1955 cnt++;
1956 rcu_read_unlock();
1957 return cnt;
1958}
1959
1960int ipv6_chk_addr(struct net *net, const struct in6_addr *addr,
1961 const struct net_device *dev, int strict)
1962{
1963 return ipv6_chk_addr_and_flags(net, addr, dev, !dev,
1964 strict, IFA_F_TENTATIVE);
1965}
1966EXPORT_SYMBOL(ipv6_chk_addr);
1967
1968/* device argument is used to find the L3 domain of interest. If
1969 * skip_dev_check is set, then the ifp device is not checked against
1970 * the passed in dev argument. So the 2 cases for addresses checks are:
1971 * 1. does the address exist in the L3 domain that dev is part of
1972 * (skip_dev_check = true), or
1973 *
1974 * 2. does the address exist on the specific device
1975 * (skip_dev_check = false)
1976 */
1977static struct net_device *
1978__ipv6_chk_addr_and_flags(struct net *net, const struct in6_addr *addr,
1979 const struct net_device *dev, bool skip_dev_check,
1980 int strict, u32 banned_flags)
1981{
1982 unsigned int hash = inet6_addr_hash(net, addr);
1983 struct net_device *l3mdev, *ndev;
1984 struct inet6_ifaddr *ifp;
1985 u32 ifp_flags;
1986
1987 rcu_read_lock();
1988
1989 l3mdev = l3mdev_master_dev_rcu(dev);
1990 if (skip_dev_check)
1991 dev = NULL;
1992
1993 hlist_for_each_entry_rcu(ifp, &net->ipv6.inet6_addr_lst[hash], addr_lst) {
1994 ndev = ifp->idev->dev;
1995
1996 if (l3mdev_master_dev_rcu(ndev) != l3mdev)
1997 continue;
1998
1999 /* Decouple optimistic from tentative for evaluation here.
2000 * Ban optimistic addresses explicitly, when required.
2001 */
2002 ifp_flags = (ifp->flags&IFA_F_OPTIMISTIC)
2003 ? (ifp->flags&~IFA_F_TENTATIVE)
2004 : ifp->flags;
2005 if (ipv6_addr_equal(&ifp->addr, addr) &&
2006 !(ifp_flags&banned_flags) &&
2007 (!dev || ndev == dev ||
2008 !(ifp->scope&(IFA_LINK|IFA_HOST) || strict))) {
2009 rcu_read_unlock();
2010 return ndev;
2011 }
2012 }
2013
2014 rcu_read_unlock();
2015 return NULL;
2016}
2017
2018int ipv6_chk_addr_and_flags(struct net *net, const struct in6_addr *addr,
2019 const struct net_device *dev, bool skip_dev_check,
2020 int strict, u32 banned_flags)
2021{
2022 return __ipv6_chk_addr_and_flags(net, addr, dev, skip_dev_check,
2023 strict, banned_flags) ? 1 : 0;
2024}
2025EXPORT_SYMBOL(ipv6_chk_addr_and_flags);
2026
2027
2028/* Compares an address/prefix_len with addresses on device @dev.
2029 * If one is found it returns true.
2030 */
2031bool ipv6_chk_custom_prefix(const struct in6_addr *addr,
2032 const unsigned int prefix_len, struct net_device *dev)
2033{
2034 const struct inet6_ifaddr *ifa;
2035 const struct inet6_dev *idev;
2036 bool ret = false;
2037
2038 rcu_read_lock();
2039 idev = __in6_dev_get(dev);
2040 if (idev) {
2041 list_for_each_entry_rcu(ifa, &idev->addr_list, if_list) {
2042 ret = ipv6_prefix_equal(addr, &ifa->addr, prefix_len);
2043 if (ret)
2044 break;
2045 }
2046 }
2047 rcu_read_unlock();
2048
2049 return ret;
2050}
2051EXPORT_SYMBOL(ipv6_chk_custom_prefix);
2052
2053int ipv6_chk_prefix(const struct in6_addr *addr, struct net_device *dev)
2054{
2055 const struct inet6_ifaddr *ifa;
2056 const struct inet6_dev *idev;
2057 int onlink;
2058
2059 onlink = 0;
2060 rcu_read_lock();
2061 idev = __in6_dev_get(dev);
2062 if (idev) {
2063 list_for_each_entry_rcu(ifa, &idev->addr_list, if_list) {
2064 onlink = ipv6_prefix_equal(addr, &ifa->addr,
2065 ifa->prefix_len);
2066 if (onlink)
2067 break;
2068 }
2069 }
2070 rcu_read_unlock();
2071 return onlink;
2072}
2073EXPORT_SYMBOL(ipv6_chk_prefix);
2074
2075/**
2076 * ipv6_dev_find - find the first device with a given source address.
2077 * @net: the net namespace
2078 * @addr: the source address
2079 * @dev: used to find the L3 domain of interest
2080 *
2081 * The caller should be protected by RCU, or RTNL.
2082 */
2083struct net_device *ipv6_dev_find(struct net *net, const struct in6_addr *addr,
2084 struct net_device *dev)
2085{
2086 return __ipv6_chk_addr_and_flags(net, addr, dev, !dev, 1,
2087 IFA_F_TENTATIVE);
2088}
2089EXPORT_SYMBOL(ipv6_dev_find);
2090
2091struct inet6_ifaddr *ipv6_get_ifaddr(struct net *net, const struct in6_addr *addr,
2092 struct net_device *dev, int strict)
2093{
2094 unsigned int hash = inet6_addr_hash(net, addr);
2095 struct inet6_ifaddr *ifp, *result = NULL;
2096
2097 rcu_read_lock();
2098 hlist_for_each_entry_rcu(ifp, &net->ipv6.inet6_addr_lst[hash], addr_lst) {
2099 if (ipv6_addr_equal(&ifp->addr, addr)) {
2100 if (!dev || ifp->idev->dev == dev ||
2101 !(ifp->scope&(IFA_LINK|IFA_HOST) || strict)) {
2102 if (in6_ifa_hold_safe(ifp)) {
2103 result = ifp;
2104 break;
2105 }
2106 }
2107 }
2108 }
2109 rcu_read_unlock();
2110
2111 return result;
2112}
2113
2114/* Gets referenced address, destroys ifaddr */
2115
2116static void addrconf_dad_stop(struct inet6_ifaddr *ifp, int dad_failed)
2117{
2118 if (dad_failed)
2119 ifp->flags |= IFA_F_DADFAILED;
2120
2121 if (ifp->flags&IFA_F_TEMPORARY) {
2122 struct inet6_ifaddr *ifpub;
2123 spin_lock_bh(&ifp->lock);
2124 ifpub = ifp->ifpub;
2125 if (ifpub) {
2126 in6_ifa_hold(ifpub);
2127 spin_unlock_bh(&ifp->lock);
2128 ipv6_create_tempaddr(ifpub, true);
2129 in6_ifa_put(ifpub);
2130 } else {
2131 spin_unlock_bh(&ifp->lock);
2132 }
2133 ipv6_del_addr(ifp);
2134 } else if (ifp->flags&IFA_F_PERMANENT || !dad_failed) {
2135 spin_lock_bh(&ifp->lock);
2136 addrconf_del_dad_work(ifp);
2137 ifp->flags |= IFA_F_TENTATIVE;
2138 if (dad_failed)
2139 ifp->flags &= ~IFA_F_OPTIMISTIC;
2140 spin_unlock_bh(&ifp->lock);
2141 if (dad_failed)
2142 ipv6_ifa_notify(0, ifp);
2143 in6_ifa_put(ifp);
2144 } else {
2145 ipv6_del_addr(ifp);
2146 }
2147}
2148
2149static int addrconf_dad_end(struct inet6_ifaddr *ifp)
2150{
2151 int err = -ENOENT;
2152
2153 spin_lock_bh(&ifp->lock);
2154 if (ifp->state == INET6_IFADDR_STATE_DAD) {
2155 ifp->state = INET6_IFADDR_STATE_POSTDAD;
2156 err = 0;
2157 }
2158 spin_unlock_bh(&ifp->lock);
2159
2160 return err;
2161}
2162
2163void addrconf_dad_failure(struct sk_buff *skb, struct inet6_ifaddr *ifp)
2164{
2165 struct inet6_dev *idev = ifp->idev;
2166 struct net *net = dev_net(idev->dev);
2167 int max_addresses;
2168
2169 if (addrconf_dad_end(ifp)) {
2170 in6_ifa_put(ifp);
2171 return;
2172 }
2173
2174 net_info_ratelimited("%s: IPv6 duplicate address %pI6c used by %pM detected!\n",
2175 ifp->idev->dev->name, &ifp->addr, eth_hdr(skb)->h_source);
2176
2177 spin_lock_bh(&ifp->lock);
2178
2179 if (ifp->flags & IFA_F_STABLE_PRIVACY) {
2180 struct in6_addr new_addr;
2181 struct inet6_ifaddr *ifp2;
2182 int retries = ifp->stable_privacy_retry + 1;
2183 struct ifa6_config cfg = {
2184 .pfx = &new_addr,
2185 .plen = ifp->prefix_len,
2186 .ifa_flags = ifp->flags,
2187 .valid_lft = ifp->valid_lft,
2188 .preferred_lft = ifp->prefered_lft,
2189 .scope = ifp->scope,
2190 };
2191
2192 if (retries > net->ipv6.sysctl.idgen_retries) {
2193 net_info_ratelimited("%s: privacy stable address generation failed because of DAD conflicts!\n",
2194 ifp->idev->dev->name);
2195 goto errdad;
2196 }
2197
2198 new_addr = ifp->addr;
2199 if (ipv6_generate_stable_address(&new_addr, retries,
2200 idev))
2201 goto errdad;
2202
2203 spin_unlock_bh(&ifp->lock);
2204
2205 max_addresses = READ_ONCE(idev->cnf.max_addresses);
2206 if (max_addresses &&
2207 ipv6_count_addresses(idev) >= max_addresses)
2208 goto lock_errdad;
2209
2210 net_info_ratelimited("%s: generating new stable privacy address because of DAD conflict\n",
2211 ifp->idev->dev->name);
2212
2213 ifp2 = ipv6_add_addr(idev, &cfg, false, NULL);
2214 if (IS_ERR(ifp2))
2215 goto lock_errdad;
2216
2217 spin_lock_bh(&ifp2->lock);
2218 ifp2->stable_privacy_retry = retries;
2219 ifp2->state = INET6_IFADDR_STATE_PREDAD;
2220 spin_unlock_bh(&ifp2->lock);
2221
2222 addrconf_mod_dad_work(ifp2, net->ipv6.sysctl.idgen_delay);
2223 in6_ifa_put(ifp2);
2224lock_errdad:
2225 spin_lock_bh(&ifp->lock);
2226 }
2227
2228errdad:
2229 /* transition from _POSTDAD to _ERRDAD */
2230 ifp->state = INET6_IFADDR_STATE_ERRDAD;
2231 spin_unlock_bh(&ifp->lock);
2232
2233 addrconf_mod_dad_work(ifp, 0);
2234 in6_ifa_put(ifp);
2235}
2236
2237/* Join to solicited addr multicast group. */
2238void addrconf_join_solict(struct net_device *dev, const struct in6_addr *addr)
2239{
2240 struct in6_addr maddr;
2241
2242 if (READ_ONCE(dev->flags) & (IFF_LOOPBACK | IFF_NOARP))
2243 return;
2244
2245 addrconf_addr_solict_mult(addr, &maddr);
2246 ipv6_dev_mc_inc(dev, &maddr);
2247}
2248
2249void addrconf_leave_solict(struct inet6_dev *idev, const struct in6_addr *addr)
2250{
2251 struct in6_addr maddr;
2252
2253 if (READ_ONCE(idev->dev->flags) & (IFF_LOOPBACK | IFF_NOARP))
2254 return;
2255
2256 addrconf_addr_solict_mult(addr, &maddr);
2257 __ipv6_dev_mc_dec(idev, &maddr);
2258}
2259
2260static void addrconf_join_anycast(struct inet6_ifaddr *ifp)
2261{
2262 struct in6_addr addr;
2263
2264 if (ifp->prefix_len >= 127) /* RFC 6164 */
2265 return;
2266 ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
2267 if (ipv6_addr_any(&addr))
2268 return;
2269 __ipv6_dev_ac_inc(ifp->idev, &addr);
2270}
2271
2272static void addrconf_leave_anycast(struct inet6_ifaddr *ifp)
2273{
2274 struct in6_addr addr;
2275
2276 if (ifp->prefix_len >= 127) /* RFC 6164 */
2277 return;
2278 ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
2279 if (ipv6_addr_any(&addr))
2280 return;
2281 __ipv6_dev_ac_dec(ifp->idev, &addr);
2282}
2283
2284static int addrconf_ifid_6lowpan(u8 *eui, struct net_device *dev)
2285{
2286 switch (dev->addr_len) {
2287 case ETH_ALEN:
2288 memcpy(eui, dev->dev_addr, 3);
2289 eui[3] = 0xFF;
2290 eui[4] = 0xFE;
2291 memcpy(eui + 5, dev->dev_addr + 3, 3);
2292 break;
2293 case EUI64_ADDR_LEN:
2294 memcpy(eui, dev->dev_addr, EUI64_ADDR_LEN);
2295 eui[0] ^= 2;
2296 break;
2297 default:
2298 return -1;
2299 }
2300
2301 return 0;
2302}
2303
2304static int addrconf_ifid_ieee1394(u8 *eui, struct net_device *dev)
2305{
2306 const union fwnet_hwaddr *ha;
2307
2308 if (dev->addr_len != FWNET_ALEN)
2309 return -1;
2310
2311 ha = (const union fwnet_hwaddr *)dev->dev_addr;
2312
2313 memcpy(eui, &ha->uc.uniq_id, sizeof(ha->uc.uniq_id));
2314 eui[0] ^= 2;
2315 return 0;
2316}
2317
2318static int addrconf_ifid_arcnet(u8 *eui, struct net_device *dev)
2319{
2320 /* XXX: inherit EUI-64 from other interface -- yoshfuji */
2321 if (dev->addr_len != ARCNET_ALEN)
2322 return -1;
2323 memset(eui, 0, 7);
2324 eui[7] = *(u8 *)dev->dev_addr;
2325 return 0;
2326}
2327
2328static int addrconf_ifid_infiniband(u8 *eui, struct net_device *dev)
2329{
2330 if (dev->addr_len != INFINIBAND_ALEN)
2331 return -1;
2332 memcpy(eui, dev->dev_addr + 12, 8);
2333 eui[0] |= 2;
2334 return 0;
2335}
2336
2337static int __ipv6_isatap_ifid(u8 *eui, __be32 addr)
2338{
2339 if (addr == 0)
2340 return -1;
2341 eui[0] = (ipv4_is_zeronet(addr) || ipv4_is_private_10(addr) ||
2342 ipv4_is_loopback(addr) || ipv4_is_linklocal_169(addr) ||
2343 ipv4_is_private_172(addr) || ipv4_is_test_192(addr) ||
2344 ipv4_is_anycast_6to4(addr) || ipv4_is_private_192(addr) ||
2345 ipv4_is_test_198(addr) || ipv4_is_multicast(addr) ||
2346 ipv4_is_lbcast(addr)) ? 0x00 : 0x02;
2347 eui[1] = 0;
2348 eui[2] = 0x5E;
2349 eui[3] = 0xFE;
2350 memcpy(eui + 4, &addr, 4);
2351 return 0;
2352}
2353
2354static int addrconf_ifid_sit(u8 *eui, struct net_device *dev)
2355{
2356 if (dev->priv_flags & IFF_ISATAP)
2357 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
2358 return -1;
2359}
2360
2361static int addrconf_ifid_gre(u8 *eui, struct net_device *dev)
2362{
2363 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
2364}
2365
2366static int addrconf_ifid_ip6tnl(u8 *eui, struct net_device *dev)
2367{
2368 memcpy(eui, dev->perm_addr, 3);
2369 memcpy(eui + 5, dev->perm_addr + 3, 3);
2370 eui[3] = 0xFF;
2371 eui[4] = 0xFE;
2372 eui[0] ^= 2;
2373 return 0;
2374}
2375
2376static int ipv6_generate_eui64(u8 *eui, struct net_device *dev)
2377{
2378 switch (dev->type) {
2379 case ARPHRD_ETHER:
2380 case ARPHRD_FDDI:
2381 return addrconf_ifid_eui48(eui, dev);
2382 case ARPHRD_ARCNET:
2383 return addrconf_ifid_arcnet(eui, dev);
2384 case ARPHRD_INFINIBAND:
2385 return addrconf_ifid_infiniband(eui, dev);
2386 case ARPHRD_SIT:
2387 return addrconf_ifid_sit(eui, dev);
2388 case ARPHRD_IPGRE:
2389 case ARPHRD_TUNNEL:
2390 return addrconf_ifid_gre(eui, dev);
2391 case ARPHRD_6LOWPAN:
2392 return addrconf_ifid_6lowpan(eui, dev);
2393 case ARPHRD_IEEE1394:
2394 return addrconf_ifid_ieee1394(eui, dev);
2395 case ARPHRD_TUNNEL6:
2396 case ARPHRD_IP6GRE:
2397 case ARPHRD_RAWIP:
2398 return addrconf_ifid_ip6tnl(eui, dev);
2399 }
2400 return -1;
2401}
2402
2403static int ipv6_inherit_eui64(u8 *eui, struct inet6_dev *idev)
2404{
2405 int err = -1;
2406 struct inet6_ifaddr *ifp;
2407
2408 read_lock_bh(&idev->lock);
2409 list_for_each_entry_reverse(ifp, &idev->addr_list, if_list) {
2410 if (ifp->scope > IFA_LINK)
2411 break;
2412 if (ifp->scope == IFA_LINK && !(ifp->flags&IFA_F_TENTATIVE)) {
2413 memcpy(eui, ifp->addr.s6_addr+8, 8);
2414 err = 0;
2415 break;
2416 }
2417 }
2418 read_unlock_bh(&idev->lock);
2419 return err;
2420}
2421
2422/* Generation of a randomized Interface Identifier
2423 * draft-ietf-6man-rfc4941bis, Section 3.3.1
2424 */
2425
2426static void ipv6_gen_rnd_iid(struct in6_addr *addr)
2427{
2428regen:
2429 get_random_bytes(&addr->s6_addr[8], 8);
2430
2431 /* <draft-ietf-6man-rfc4941bis-08.txt>, Section 3.3.1:
2432 * check if generated address is not inappropriate:
2433 *
2434 * - Reserved IPv6 Interface Identifiers
2435 * - XXX: already assigned to an address on the device
2436 */
2437
2438 /* Subnet-router anycast: 0000:0000:0000:0000 */
2439 if (!(addr->s6_addr32[2] | addr->s6_addr32[3]))
2440 goto regen;
2441
2442 /* IANA Ethernet block: 0200:5EFF:FE00:0000-0200:5EFF:FE00:5212
2443 * Proxy Mobile IPv6: 0200:5EFF:FE00:5213
2444 * IANA Ethernet block: 0200:5EFF:FE00:5214-0200:5EFF:FEFF:FFFF
2445 */
2446 if (ntohl(addr->s6_addr32[2]) == 0x02005eff &&
2447 (ntohl(addr->s6_addr32[3]) & 0Xff000000) == 0xfe000000)
2448 goto regen;
2449
2450 /* Reserved subnet anycast addresses */
2451 if (ntohl(addr->s6_addr32[2]) == 0xfdffffff &&
2452 ntohl(addr->s6_addr32[3]) >= 0Xffffff80)
2453 goto regen;
2454}
2455
2456/*
2457 * Add prefix route.
2458 */
2459
2460static void
2461addrconf_prefix_route(struct in6_addr *pfx, int plen, u32 metric,
2462 struct net_device *dev, unsigned long expires,
2463 u32 flags, gfp_t gfp_flags)
2464{
2465 struct fib6_config cfg = {
2466 .fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_PREFIX,
2467 .fc_metric = metric ? : IP6_RT_PRIO_ADDRCONF,
2468 .fc_ifindex = dev->ifindex,
2469 .fc_expires = expires,
2470 .fc_dst_len = plen,
2471 .fc_flags = RTF_UP | flags,
2472 .fc_nlinfo.nl_net = dev_net(dev),
2473 .fc_protocol = RTPROT_KERNEL,
2474 .fc_type = RTN_UNICAST,
2475 };
2476
2477 cfg.fc_dst = *pfx;
2478
2479 /* Prevent useless cloning on PtP SIT.
2480 This thing is done here expecting that the whole
2481 class of non-broadcast devices need not cloning.
2482 */
2483#if IS_ENABLED(CONFIG_IPV6_SIT)
2484 if (dev->type == ARPHRD_SIT && (dev->flags & IFF_POINTOPOINT))
2485 cfg.fc_flags |= RTF_NONEXTHOP;
2486#endif
2487
2488 ip6_route_add(&cfg, gfp_flags, NULL);
2489}
2490
2491
2492static struct fib6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
2493 int plen,
2494 const struct net_device *dev,
2495 u32 flags, u32 noflags,
2496 bool no_gw)
2497{
2498 struct fib6_node *fn;
2499 struct fib6_info *rt = NULL;
2500 struct fib6_table *table;
2501 u32 tb_id = l3mdev_fib_table(dev) ? : RT6_TABLE_PREFIX;
2502
2503 table = fib6_get_table(dev_net(dev), tb_id);
2504 if (!table)
2505 return NULL;
2506
2507 rcu_read_lock();
2508 fn = fib6_locate(&table->tb6_root, pfx, plen, NULL, 0, true);
2509 if (!fn)
2510 goto out;
2511
2512 for_each_fib6_node_rt_rcu(fn) {
2513 /* prefix routes only use builtin fib6_nh */
2514 if (rt->nh)
2515 continue;
2516
2517 if (rt->fib6_nh->fib_nh_dev->ifindex != dev->ifindex)
2518 continue;
2519 if (no_gw && rt->fib6_nh->fib_nh_gw_family)
2520 continue;
2521 if ((rt->fib6_flags & flags) != flags)
2522 continue;
2523 if ((rt->fib6_flags & noflags) != 0)
2524 continue;
2525 if (!fib6_info_hold_safe(rt))
2526 continue;
2527 break;
2528 }
2529out:
2530 rcu_read_unlock();
2531 return rt;
2532}
2533
2534
2535/* Create "default" multicast route to the interface */
2536
2537static void addrconf_add_mroute(struct net_device *dev)
2538{
2539 struct fib6_config cfg = {
2540 .fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_LOCAL,
2541 .fc_metric = IP6_RT_PRIO_ADDRCONF,
2542 .fc_ifindex = dev->ifindex,
2543 .fc_dst_len = 8,
2544 .fc_flags = RTF_UP,
2545 .fc_type = RTN_MULTICAST,
2546 .fc_nlinfo.nl_net = dev_net(dev),
2547 .fc_protocol = RTPROT_KERNEL,
2548 };
2549
2550 ipv6_addr_set(&cfg.fc_dst, htonl(0xFF000000), 0, 0, 0);
2551
2552 ip6_route_add(&cfg, GFP_KERNEL, NULL);
2553}
2554
2555static struct inet6_dev *addrconf_add_dev(struct net_device *dev)
2556{
2557 struct inet6_dev *idev;
2558
2559 ASSERT_RTNL();
2560
2561 idev = ipv6_find_idev(dev);
2562 if (IS_ERR(idev))
2563 return idev;
2564
2565 if (idev->cnf.disable_ipv6)
2566 return ERR_PTR(-EACCES);
2567
2568 /* Add default multicast route */
2569 if (!(dev->flags & IFF_LOOPBACK) && !netif_is_l3_master(dev))
2570 addrconf_add_mroute(dev);
2571
2572 return idev;
2573}
2574
2575static void delete_tempaddrs(struct inet6_dev *idev,
2576 struct inet6_ifaddr *ifp)
2577{
2578 struct inet6_ifaddr *ift, *tmp;
2579
2580 write_lock_bh(&idev->lock);
2581 list_for_each_entry_safe(ift, tmp, &idev->tempaddr_list, tmp_list) {
2582 if (ift->ifpub != ifp)
2583 continue;
2584
2585 in6_ifa_hold(ift);
2586 write_unlock_bh(&idev->lock);
2587 ipv6_del_addr(ift);
2588 write_lock_bh(&idev->lock);
2589 }
2590 write_unlock_bh(&idev->lock);
2591}
2592
2593static void manage_tempaddrs(struct inet6_dev *idev,
2594 struct inet6_ifaddr *ifp,
2595 __u32 valid_lft, __u32 prefered_lft,
2596 bool create, unsigned long now)
2597{
2598 u32 flags;
2599 struct inet6_ifaddr *ift;
2600
2601 read_lock_bh(&idev->lock);
2602 /* update all temporary addresses in the list */
2603 list_for_each_entry(ift, &idev->tempaddr_list, tmp_list) {
2604 int age, max_valid, max_prefered;
2605
2606 if (ifp != ift->ifpub)
2607 continue;
2608
2609 /* RFC 4941 section 3.3:
2610 * If a received option will extend the lifetime of a public
2611 * address, the lifetimes of temporary addresses should
2612 * be extended, subject to the overall constraint that no
2613 * temporary addresses should ever remain "valid" or "preferred"
2614 * for a time longer than (TEMP_VALID_LIFETIME) or
2615 * (TEMP_PREFERRED_LIFETIME - DESYNC_FACTOR), respectively.
2616 */
2617 age = (now - ift->cstamp) / HZ;
2618 max_valid = READ_ONCE(idev->cnf.temp_valid_lft) - age;
2619 if (max_valid < 0)
2620 max_valid = 0;
2621
2622 max_prefered = READ_ONCE(idev->cnf.temp_prefered_lft) -
2623 idev->desync_factor - age;
2624 if (max_prefered < 0)
2625 max_prefered = 0;
2626
2627 if (valid_lft > max_valid)
2628 valid_lft = max_valid;
2629
2630 if (prefered_lft > max_prefered)
2631 prefered_lft = max_prefered;
2632
2633 spin_lock(&ift->lock);
2634 flags = ift->flags;
2635 ift->valid_lft = valid_lft;
2636 ift->prefered_lft = prefered_lft;
2637 ift->tstamp = now;
2638 if (prefered_lft > 0)
2639 ift->flags &= ~IFA_F_DEPRECATED;
2640
2641 spin_unlock(&ift->lock);
2642 if (!(flags&IFA_F_TENTATIVE))
2643 ipv6_ifa_notify(0, ift);
2644 }
2645
2646 /* Also create a temporary address if it's enabled but no temporary
2647 * address currently exists.
2648 * However, we get called with valid_lft == 0, prefered_lft == 0, create == false
2649 * as part of cleanup (ie. deleting the mngtmpaddr).
2650 * We don't want that to result in creating a new temporary ip address.
2651 */
2652 if (list_empty(&idev->tempaddr_list) && (valid_lft || prefered_lft))
2653 create = true;
2654
2655 if (create && READ_ONCE(idev->cnf.use_tempaddr) > 0) {
2656 /* When a new public address is created as described
2657 * in [ADDRCONF], also create a new temporary address.
2658 */
2659 read_unlock_bh(&idev->lock);
2660 ipv6_create_tempaddr(ifp, false);
2661 } else {
2662 read_unlock_bh(&idev->lock);
2663 }
2664}
2665
2666static bool is_addr_mode_generate_stable(struct inet6_dev *idev)
2667{
2668 return idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_STABLE_PRIVACY ||
2669 idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_RANDOM;
2670}
2671
2672int addrconf_prefix_rcv_add_addr(struct net *net, struct net_device *dev,
2673 const struct prefix_info *pinfo,
2674 struct inet6_dev *in6_dev,
2675 const struct in6_addr *addr, int addr_type,
2676 u32 addr_flags, bool sllao, bool tokenized,
2677 __u32 valid_lft, u32 prefered_lft)
2678{
2679 struct inet6_ifaddr *ifp = ipv6_get_ifaddr(net, addr, dev, 1);
2680 int create = 0, update_lft = 0;
2681
2682 if (!ifp && valid_lft) {
2683 int max_addresses = READ_ONCE(in6_dev->cnf.max_addresses);
2684 struct ifa6_config cfg = {
2685 .pfx = addr,
2686 .plen = pinfo->prefix_len,
2687 .ifa_flags = addr_flags,
2688 .valid_lft = valid_lft,
2689 .preferred_lft = prefered_lft,
2690 .scope = addr_type & IPV6_ADDR_SCOPE_MASK,
2691 .ifa_proto = IFAPROT_KERNEL_RA
2692 };
2693
2694#ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2695 if ((READ_ONCE(net->ipv6.devconf_all->optimistic_dad) ||
2696 READ_ONCE(in6_dev->cnf.optimistic_dad)) &&
2697 !net->ipv6.devconf_all->forwarding && sllao)
2698 cfg.ifa_flags |= IFA_F_OPTIMISTIC;
2699#endif
2700
2701 /* Do not allow to create too much of autoconfigured
2702 * addresses; this would be too easy way to crash kernel.
2703 */
2704 if (!max_addresses ||
2705 ipv6_count_addresses(in6_dev) < max_addresses)
2706 ifp = ipv6_add_addr(in6_dev, &cfg, false, NULL);
2707
2708 if (IS_ERR_OR_NULL(ifp))
2709 return -1;
2710
2711 create = 1;
2712 spin_lock_bh(&ifp->lock);
2713 ifp->flags |= IFA_F_MANAGETEMPADDR;
2714 ifp->cstamp = jiffies;
2715 ifp->tokenized = tokenized;
2716 spin_unlock_bh(&ifp->lock);
2717 addrconf_dad_start(ifp);
2718 }
2719
2720 if (ifp) {
2721 u32 flags;
2722 unsigned long now;
2723 u32 stored_lft;
2724
2725 /* update lifetime (RFC2462 5.5.3 e) */
2726 spin_lock_bh(&ifp->lock);
2727 now = jiffies;
2728 if (ifp->valid_lft > (now - ifp->tstamp) / HZ)
2729 stored_lft = ifp->valid_lft - (now - ifp->tstamp) / HZ;
2730 else
2731 stored_lft = 0;
2732
2733 /* RFC4862 Section 5.5.3e:
2734 * "Note that the preferred lifetime of the
2735 * corresponding address is always reset to
2736 * the Preferred Lifetime in the received
2737 * Prefix Information option, regardless of
2738 * whether the valid lifetime is also reset or
2739 * ignored."
2740 *
2741 * So we should always update prefered_lft here.
2742 */
2743 update_lft = !create && stored_lft;
2744
2745 if (update_lft && !READ_ONCE(in6_dev->cnf.ra_honor_pio_life)) {
2746 const u32 minimum_lft = min_t(u32,
2747 stored_lft, MIN_VALID_LIFETIME);
2748 valid_lft = max(valid_lft, minimum_lft);
2749 }
2750
2751 if (update_lft) {
2752 ifp->valid_lft = valid_lft;
2753 ifp->prefered_lft = prefered_lft;
2754 WRITE_ONCE(ifp->tstamp, now);
2755 flags = ifp->flags;
2756 ifp->flags &= ~IFA_F_DEPRECATED;
2757 spin_unlock_bh(&ifp->lock);
2758
2759 if (!(flags&IFA_F_TENTATIVE))
2760 ipv6_ifa_notify(0, ifp);
2761 } else
2762 spin_unlock_bh(&ifp->lock);
2763
2764 manage_tempaddrs(in6_dev, ifp, valid_lft, prefered_lft,
2765 create, now);
2766
2767 in6_ifa_put(ifp);
2768 addrconf_verify(net);
2769 }
2770
2771 return 0;
2772}
2773EXPORT_SYMBOL_GPL(addrconf_prefix_rcv_add_addr);
2774
2775void addrconf_prefix_rcv(struct net_device *dev, u8 *opt, int len, bool sllao)
2776{
2777 struct prefix_info *pinfo;
2778 struct fib6_table *table;
2779 __u32 valid_lft;
2780 __u32 prefered_lft;
2781 int addr_type, err;
2782 u32 addr_flags = 0;
2783 struct inet6_dev *in6_dev;
2784 struct net *net = dev_net(dev);
2785 bool ignore_autoconf = false;
2786
2787 pinfo = (struct prefix_info *) opt;
2788
2789 if (len < sizeof(struct prefix_info)) {
2790 netdev_dbg(dev, "addrconf: prefix option too short\n");
2791 return;
2792 }
2793
2794 /*
2795 * Validation checks ([ADDRCONF], page 19)
2796 */
2797
2798 addr_type = ipv6_addr_type(&pinfo->prefix);
2799
2800 if (addr_type & (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL))
2801 return;
2802
2803 valid_lft = ntohl(pinfo->valid);
2804 prefered_lft = ntohl(pinfo->prefered);
2805
2806 if (prefered_lft > valid_lft) {
2807 net_warn_ratelimited("addrconf: prefix option has invalid lifetime\n");
2808 return;
2809 }
2810
2811 in6_dev = in6_dev_get(dev);
2812
2813 if (!in6_dev) {
2814 net_dbg_ratelimited("addrconf: device %s not configured\n",
2815 dev->name);
2816 return;
2817 }
2818
2819 if (valid_lft != 0 && valid_lft < in6_dev->cnf.accept_ra_min_lft)
2820 goto put;
2821
2822 /*
2823 * Two things going on here:
2824 * 1) Add routes for on-link prefixes
2825 * 2) Configure prefixes with the auto flag set
2826 */
2827
2828 if (pinfo->onlink) {
2829 struct fib6_info *rt;
2830 unsigned long rt_expires;
2831
2832 /* Avoid arithmetic overflow. Really, we could
2833 * save rt_expires in seconds, likely valid_lft,
2834 * but it would require division in fib gc, that it
2835 * not good.
2836 */
2837 if (HZ > USER_HZ)
2838 rt_expires = addrconf_timeout_fixup(valid_lft, HZ);
2839 else
2840 rt_expires = addrconf_timeout_fixup(valid_lft, USER_HZ);
2841
2842 if (addrconf_finite_timeout(rt_expires))
2843 rt_expires *= HZ;
2844
2845 rt = addrconf_get_prefix_route(&pinfo->prefix,
2846 pinfo->prefix_len,
2847 dev,
2848 RTF_ADDRCONF | RTF_PREFIX_RT,
2849 RTF_DEFAULT, true);
2850
2851 if (rt) {
2852 /* Autoconf prefix route */
2853 if (valid_lft == 0) {
2854 ip6_del_rt(net, rt, false);
2855 rt = NULL;
2856 } else {
2857 table = rt->fib6_table;
2858 spin_lock_bh(&table->tb6_lock);
2859
2860 if (addrconf_finite_timeout(rt_expires)) {
2861 /* not infinity */
2862 fib6_set_expires(rt, jiffies + rt_expires);
2863 fib6_add_gc_list(rt);
2864 } else {
2865 fib6_clean_expires(rt);
2866 fib6_remove_gc_list(rt);
2867 }
2868
2869 spin_unlock_bh(&table->tb6_lock);
2870 }
2871 } else if (valid_lft) {
2872 clock_t expires = 0;
2873 int flags = RTF_ADDRCONF | RTF_PREFIX_RT;
2874 if (addrconf_finite_timeout(rt_expires)) {
2875 /* not infinity */
2876 flags |= RTF_EXPIRES;
2877 expires = jiffies_to_clock_t(rt_expires);
2878 }
2879 addrconf_prefix_route(&pinfo->prefix, pinfo->prefix_len,
2880 0, dev, expires, flags,
2881 GFP_ATOMIC);
2882 }
2883 fib6_info_release(rt);
2884 }
2885
2886 /* Try to figure out our local address for this prefix */
2887
2888 ignore_autoconf = READ_ONCE(in6_dev->cnf.ra_honor_pio_pflag) && pinfo->preferpd;
2889 if (pinfo->autoconf && in6_dev->cnf.autoconf && !ignore_autoconf) {
2890 struct in6_addr addr;
2891 bool tokenized = false, dev_addr_generated = false;
2892
2893 if (pinfo->prefix_len == 64) {
2894 memcpy(&addr, &pinfo->prefix, 8);
2895
2896 if (!ipv6_addr_any(&in6_dev->token)) {
2897 read_lock_bh(&in6_dev->lock);
2898 memcpy(addr.s6_addr + 8,
2899 in6_dev->token.s6_addr + 8, 8);
2900 read_unlock_bh(&in6_dev->lock);
2901 tokenized = true;
2902 } else if (is_addr_mode_generate_stable(in6_dev) &&
2903 !ipv6_generate_stable_address(&addr, 0,
2904 in6_dev)) {
2905 addr_flags |= IFA_F_STABLE_PRIVACY;
2906 goto ok;
2907 } else if (ipv6_generate_eui64(addr.s6_addr + 8, dev) &&
2908 ipv6_inherit_eui64(addr.s6_addr + 8, in6_dev)) {
2909 goto put;
2910 } else {
2911 dev_addr_generated = true;
2912 }
2913 goto ok;
2914 }
2915 net_dbg_ratelimited("IPv6 addrconf: prefix with wrong length %d\n",
2916 pinfo->prefix_len);
2917 goto put;
2918
2919ok:
2920 err = addrconf_prefix_rcv_add_addr(net, dev, pinfo, in6_dev,
2921 &addr, addr_type,
2922 addr_flags, sllao,
2923 tokenized, valid_lft,
2924 prefered_lft);
2925 if (err)
2926 goto put;
2927
2928 /* Ignore error case here because previous prefix add addr was
2929 * successful which will be notified.
2930 */
2931 ndisc_ops_prefix_rcv_add_addr(net, dev, pinfo, in6_dev, &addr,
2932 addr_type, addr_flags, sllao,
2933 tokenized, valid_lft,
2934 prefered_lft,
2935 dev_addr_generated);
2936 }
2937 inet6_prefix_notify(RTM_NEWPREFIX, in6_dev, pinfo);
2938put:
2939 in6_dev_put(in6_dev);
2940}
2941
2942static int addrconf_set_sit_dstaddr(struct net *net, struct net_device *dev,
2943 struct in6_ifreq *ireq)
2944{
2945 struct ip_tunnel_parm_kern p = { };
2946 int err;
2947
2948 if (!(ipv6_addr_type(&ireq->ifr6_addr) & IPV6_ADDR_COMPATv4))
2949 return -EADDRNOTAVAIL;
2950
2951 p.iph.daddr = ireq->ifr6_addr.s6_addr32[3];
2952 p.iph.version = 4;
2953 p.iph.ihl = 5;
2954 p.iph.protocol = IPPROTO_IPV6;
2955 p.iph.ttl = 64;
2956
2957 if (!dev->netdev_ops->ndo_tunnel_ctl)
2958 return -EOPNOTSUPP;
2959 err = dev->netdev_ops->ndo_tunnel_ctl(dev, &p, SIOCADDTUNNEL);
2960 if (err)
2961 return err;
2962
2963 dev = __dev_get_by_name(net, p.name);
2964 if (!dev)
2965 return -ENOBUFS;
2966 return dev_open(dev, NULL);
2967}
2968
2969/*
2970 * Set destination address.
2971 * Special case for SIT interfaces where we create a new "virtual"
2972 * device.
2973 */
2974int addrconf_set_dstaddr(struct net *net, void __user *arg)
2975{
2976 struct net_device *dev;
2977 struct in6_ifreq ireq;
2978 int err = -ENODEV;
2979
2980 if (!IS_ENABLED(CONFIG_IPV6_SIT))
2981 return -ENODEV;
2982 if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2983 return -EFAULT;
2984
2985 rtnl_net_lock(net);
2986 dev = __dev_get_by_index(net, ireq.ifr6_ifindex);
2987 if (dev && dev->type == ARPHRD_SIT)
2988 err = addrconf_set_sit_dstaddr(net, dev, &ireq);
2989 rtnl_net_unlock(net);
2990 return err;
2991}
2992
2993static int ipv6_mc_config(struct sock *sk, bool join,
2994 const struct in6_addr *addr, int ifindex)
2995{
2996 int ret;
2997
2998 ASSERT_RTNL();
2999
3000 lock_sock(sk);
3001 if (join)
3002 ret = ipv6_sock_mc_join(sk, ifindex, addr);
3003 else
3004 ret = ipv6_sock_mc_drop(sk, ifindex, addr);
3005 release_sock(sk);
3006
3007 return ret;
3008}
3009
3010/*
3011 * Manual configuration of address on an interface
3012 */
3013static int inet6_addr_add(struct net *net, struct net_device *dev,
3014 struct ifa6_config *cfg, clock_t expires, u32 flags,
3015 struct netlink_ext_ack *extack)
3016{
3017 struct inet6_ifaddr *ifp;
3018 struct inet6_dev *idev;
3019
3020 ASSERT_RTNL_NET(net);
3021
3022 if (cfg->plen > 128) {
3023 NL_SET_ERR_MSG_MOD(extack, "Invalid prefix length");
3024 return -EINVAL;
3025 }
3026
3027 if (cfg->ifa_flags & IFA_F_MANAGETEMPADDR && cfg->plen != 64) {
3028 NL_SET_ERR_MSG_MOD(extack, "address with \"mngtmpaddr\" flag must have a prefix length of 64");
3029 return -EINVAL;
3030 }
3031
3032 idev = addrconf_add_dev(dev);
3033 if (IS_ERR(idev)) {
3034 NL_SET_ERR_MSG_MOD(extack, "IPv6 is disabled on this device");
3035 return PTR_ERR(idev);
3036 }
3037
3038 if (cfg->ifa_flags & IFA_F_MCAUTOJOIN) {
3039 int ret = ipv6_mc_config(net->ipv6.mc_autojoin_sk,
3040 true, cfg->pfx, dev->ifindex);
3041
3042 if (ret < 0) {
3043 NL_SET_ERR_MSG_MOD(extack, "Multicast auto join failed");
3044 return ret;
3045 }
3046 }
3047
3048 cfg->scope = ipv6_addr_scope(cfg->pfx);
3049
3050 ifp = ipv6_add_addr(idev, cfg, true, extack);
3051 if (!IS_ERR(ifp)) {
3052 if (!(cfg->ifa_flags & IFA_F_NOPREFIXROUTE)) {
3053 addrconf_prefix_route(&ifp->addr, ifp->prefix_len,
3054 ifp->rt_priority, dev, expires,
3055 flags, GFP_KERNEL);
3056 }
3057
3058 /* Send a netlink notification if DAD is enabled and
3059 * optimistic flag is not set
3060 */
3061 if (!(ifp->flags & (IFA_F_OPTIMISTIC | IFA_F_NODAD)))
3062 ipv6_ifa_notify(0, ifp);
3063 /*
3064 * Note that section 3.1 of RFC 4429 indicates
3065 * that the Optimistic flag should not be set for
3066 * manually configured addresses
3067 */
3068 addrconf_dad_start(ifp);
3069 if (cfg->ifa_flags & IFA_F_MANAGETEMPADDR)
3070 manage_tempaddrs(idev, ifp, cfg->valid_lft,
3071 cfg->preferred_lft, true, jiffies);
3072 in6_ifa_put(ifp);
3073 addrconf_verify_rtnl(net);
3074 return 0;
3075 } else if (cfg->ifa_flags & IFA_F_MCAUTOJOIN) {
3076 ipv6_mc_config(net->ipv6.mc_autojoin_sk, false,
3077 cfg->pfx, dev->ifindex);
3078 }
3079
3080 return PTR_ERR(ifp);
3081}
3082
3083static int inet6_addr_del(struct net *net, int ifindex, u32 ifa_flags,
3084 const struct in6_addr *pfx, unsigned int plen,
3085 struct netlink_ext_ack *extack)
3086{
3087 struct inet6_ifaddr *ifp;
3088 struct inet6_dev *idev;
3089 struct net_device *dev;
3090
3091 if (plen > 128) {
3092 NL_SET_ERR_MSG_MOD(extack, "Invalid prefix length");
3093 return -EINVAL;
3094 }
3095
3096 dev = __dev_get_by_index(net, ifindex);
3097 if (!dev) {
3098 NL_SET_ERR_MSG_MOD(extack, "Unable to find the interface");
3099 return -ENODEV;
3100 }
3101
3102 idev = __in6_dev_get_rtnl_net(dev);
3103 if (!idev) {
3104 NL_SET_ERR_MSG_MOD(extack, "IPv6 is disabled on this device");
3105 return -ENXIO;
3106 }
3107
3108 read_lock_bh(&idev->lock);
3109 list_for_each_entry(ifp, &idev->addr_list, if_list) {
3110 if (ifp->prefix_len == plen &&
3111 ipv6_addr_equal(pfx, &ifp->addr)) {
3112 in6_ifa_hold(ifp);
3113 read_unlock_bh(&idev->lock);
3114
3115 if (!(ifp->flags & IFA_F_TEMPORARY) &&
3116 (ifp->flags & IFA_F_MANAGETEMPADDR))
3117 delete_tempaddrs(idev, ifp);
3118
3119 ipv6_del_addr(ifp);
3120
3121 addrconf_verify_rtnl(net);
3122 if (ipv6_addr_is_multicast(pfx)) {
3123 ipv6_mc_config(net->ipv6.mc_autojoin_sk,
3124 false, pfx, dev->ifindex);
3125 }
3126 return 0;
3127 }
3128 }
3129 read_unlock_bh(&idev->lock);
3130
3131 NL_SET_ERR_MSG_MOD(extack, "address not found");
3132 return -EADDRNOTAVAIL;
3133}
3134
3135
3136int addrconf_add_ifaddr(struct net *net, void __user *arg)
3137{
3138 struct ifa6_config cfg = {
3139 .ifa_flags = IFA_F_PERMANENT,
3140 .preferred_lft = INFINITY_LIFE_TIME,
3141 .valid_lft = INFINITY_LIFE_TIME,
3142 };
3143 struct net_device *dev;
3144 struct in6_ifreq ireq;
3145 int err;
3146
3147 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
3148 return -EPERM;
3149
3150 if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
3151 return -EFAULT;
3152
3153 cfg.pfx = &ireq.ifr6_addr;
3154 cfg.plen = ireq.ifr6_prefixlen;
3155
3156 rtnl_net_lock(net);
3157 dev = __dev_get_by_index(net, ireq.ifr6_ifindex);
3158 if (dev) {
3159 netdev_lock_ops(dev);
3160 err = inet6_addr_add(net, dev, &cfg, 0, 0, NULL);
3161 netdev_unlock_ops(dev);
3162 } else {
3163 err = -ENODEV;
3164 }
3165 rtnl_net_unlock(net);
3166 return err;
3167}
3168
3169int addrconf_del_ifaddr(struct net *net, void __user *arg)
3170{
3171 struct in6_ifreq ireq;
3172 int err;
3173
3174 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
3175 return -EPERM;
3176
3177 if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
3178 return -EFAULT;
3179
3180 rtnl_net_lock(net);
3181 err = inet6_addr_del(net, ireq.ifr6_ifindex, 0, &ireq.ifr6_addr,
3182 ireq.ifr6_prefixlen, NULL);
3183 rtnl_net_unlock(net);
3184 return err;
3185}
3186
3187static void add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
3188 int plen, int scope, u8 proto)
3189{
3190 struct inet6_ifaddr *ifp;
3191 struct ifa6_config cfg = {
3192 .pfx = addr,
3193 .plen = plen,
3194 .ifa_flags = IFA_F_PERMANENT,
3195 .valid_lft = INFINITY_LIFE_TIME,
3196 .preferred_lft = INFINITY_LIFE_TIME,
3197 .scope = scope,
3198 .ifa_proto = proto
3199 };
3200
3201 ifp = ipv6_add_addr(idev, &cfg, true, NULL);
3202 if (!IS_ERR(ifp)) {
3203 spin_lock_bh(&ifp->lock);
3204 ifp->flags &= ~IFA_F_TENTATIVE;
3205 spin_unlock_bh(&ifp->lock);
3206 rt_genid_bump_ipv6(dev_net(idev->dev));
3207 ipv6_ifa_notify(RTM_NEWADDR, ifp);
3208 in6_ifa_put(ifp);
3209 }
3210}
3211
3212#if IS_ENABLED(CONFIG_IPV6_SIT) || IS_ENABLED(CONFIG_NET_IPGRE)
3213static void add_v4_addrs(struct inet6_dev *idev)
3214{
3215 struct in6_addr addr;
3216 struct net_device *dev;
3217 struct net *net = dev_net(idev->dev);
3218 int scope, plen;
3219 u32 pflags = 0;
3220
3221 ASSERT_RTNL();
3222
3223 memset(&addr, 0, sizeof(struct in6_addr));
3224 memcpy(&addr.s6_addr32[3], idev->dev->dev_addr, 4);
3225
3226 if (!(idev->dev->flags & IFF_POINTOPOINT) && idev->dev->type == ARPHRD_SIT) {
3227 scope = IPV6_ADDR_COMPATv4;
3228 plen = 96;
3229 pflags |= RTF_NONEXTHOP;
3230 } else {
3231 if (idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_NONE)
3232 return;
3233
3234 addr.s6_addr32[0] = htonl(0xfe800000);
3235 scope = IFA_LINK;
3236 plen = 64;
3237 }
3238
3239 if (addr.s6_addr32[3]) {
3240 add_addr(idev, &addr, plen, scope, IFAPROT_UNSPEC);
3241 addrconf_prefix_route(&addr, plen, 0, idev->dev, 0, pflags,
3242 GFP_KERNEL);
3243 return;
3244 }
3245
3246 for_each_netdev(net, dev) {
3247 struct in_device *in_dev = __in_dev_get_rtnl(dev);
3248 if (in_dev && (dev->flags & IFF_UP)) {
3249 struct in_ifaddr *ifa;
3250 int flag = scope;
3251
3252 in_dev_for_each_ifa_rtnl(ifa, in_dev) {
3253 addr.s6_addr32[3] = ifa->ifa_local;
3254
3255 if (ifa->ifa_scope == RT_SCOPE_LINK)
3256 continue;
3257 if (ifa->ifa_scope >= RT_SCOPE_HOST) {
3258 if (idev->dev->flags&IFF_POINTOPOINT)
3259 continue;
3260 flag |= IFA_HOST;
3261 }
3262
3263 add_addr(idev, &addr, plen, flag,
3264 IFAPROT_UNSPEC);
3265 addrconf_prefix_route(&addr, plen, 0, idev->dev,
3266 0, pflags, GFP_KERNEL);
3267 }
3268 }
3269 }
3270}
3271#endif
3272
3273static void init_loopback(struct net_device *dev)
3274{
3275 struct inet6_dev *idev;
3276
3277 /* ::1 */
3278
3279 ASSERT_RTNL();
3280
3281 idev = ipv6_find_idev(dev);
3282 if (IS_ERR(idev)) {
3283 pr_debug("%s: add_dev failed\n", __func__);
3284 return;
3285 }
3286
3287 add_addr(idev, &in6addr_loopback, 128, IFA_HOST, IFAPROT_KERNEL_LO);
3288}
3289
3290void addrconf_add_linklocal(struct inet6_dev *idev,
3291 const struct in6_addr *addr, u32 flags)
3292{
3293 struct ifa6_config cfg = {
3294 .pfx = addr,
3295 .plen = 64,
3296 .ifa_flags = flags | IFA_F_PERMANENT,
3297 .valid_lft = INFINITY_LIFE_TIME,
3298 .preferred_lft = INFINITY_LIFE_TIME,
3299 .scope = IFA_LINK,
3300 .ifa_proto = IFAPROT_KERNEL_LL
3301 };
3302 struct inet6_ifaddr *ifp;
3303
3304#ifdef CONFIG_IPV6_OPTIMISTIC_DAD
3305 if ((READ_ONCE(dev_net(idev->dev)->ipv6.devconf_all->optimistic_dad) ||
3306 READ_ONCE(idev->cnf.optimistic_dad)) &&
3307 !dev_net(idev->dev)->ipv6.devconf_all->forwarding)
3308 cfg.ifa_flags |= IFA_F_OPTIMISTIC;
3309#endif
3310
3311 ifp = ipv6_add_addr(idev, &cfg, true, NULL);
3312 if (!IS_ERR(ifp)) {
3313 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, 0, idev->dev,
3314 0, 0, GFP_ATOMIC);
3315 addrconf_dad_start(ifp);
3316 in6_ifa_put(ifp);
3317 }
3318}
3319EXPORT_SYMBOL_GPL(addrconf_add_linklocal);
3320
3321static bool ipv6_reserved_interfaceid(struct in6_addr address)
3322{
3323 if ((address.s6_addr32[2] | address.s6_addr32[3]) == 0)
3324 return true;
3325
3326 if (address.s6_addr32[2] == htonl(0x02005eff) &&
3327 ((address.s6_addr32[3] & htonl(0xfe000000)) == htonl(0xfe000000)))
3328 return true;
3329
3330 if (address.s6_addr32[2] == htonl(0xfdffffff) &&
3331 ((address.s6_addr32[3] & htonl(0xffffff80)) == htonl(0xffffff80)))
3332 return true;
3333
3334 return false;
3335}
3336
3337static int ipv6_generate_stable_address(struct in6_addr *address,
3338 u8 dad_count,
3339 const struct inet6_dev *idev)
3340{
3341 static DEFINE_SPINLOCK(lock);
3342 static struct sha1_ctx sha_ctx;
3343
3344 static union {
3345 u8 __data[SHA1_BLOCK_SIZE];
3346 struct {
3347 struct in6_addr secret;
3348 __be32 prefix[2];
3349 unsigned char hwaddr[MAX_ADDR_LEN];
3350 u8 dad_count;
3351 } __packed;
3352 } data;
3353
3354 struct in6_addr secret;
3355 struct in6_addr temp;
3356 struct net *net = dev_net(idev->dev);
3357
3358 BUILD_BUG_ON(sizeof(data.__data) != sizeof(data));
3359
3360 if (idev->cnf.stable_secret.initialized)
3361 secret = idev->cnf.stable_secret.secret;
3362 else if (net->ipv6.devconf_dflt->stable_secret.initialized)
3363 secret = net->ipv6.devconf_dflt->stable_secret.secret;
3364 else
3365 return -1;
3366
3367retry:
3368 spin_lock_bh(&lock);
3369
3370 sha1_init(&sha_ctx);
3371
3372 memset(&data, 0, sizeof(data));
3373 memcpy(data.hwaddr, idev->dev->perm_addr, idev->dev->addr_len);
3374 data.prefix[0] = address->s6_addr32[0];
3375 data.prefix[1] = address->s6_addr32[1];
3376 data.secret = secret;
3377 data.dad_count = dad_count;
3378
3379 sha1_update(&sha_ctx, data.__data, sizeof(data));
3380
3381 /*
3382 * Note that the SHA-1 finalization is omitted here, and the digest is
3383 * pulled directly from the internal SHA-1 state (making it incompatible
3384 * with standard SHA-1). Unusual, but technically okay since the data
3385 * length is fixed and is a multiple of the SHA-1 block size.
3386 */
3387 temp = *address;
3388 temp.s6_addr32[2] = (__force __be32)sha_ctx.state.h[0];
3389 temp.s6_addr32[3] = (__force __be32)sha_ctx.state.h[1];
3390
3391 spin_unlock_bh(&lock);
3392
3393 if (ipv6_reserved_interfaceid(temp)) {
3394 dad_count++;
3395 if (dad_count > dev_net(idev->dev)->ipv6.sysctl.idgen_retries)
3396 return -1;
3397 goto retry;
3398 }
3399
3400 *address = temp;
3401 return 0;
3402}
3403
3404static void ipv6_gen_mode_random_init(struct inet6_dev *idev)
3405{
3406 struct ipv6_stable_secret *s = &idev->cnf.stable_secret;
3407
3408 if (s->initialized)
3409 return;
3410 s = &idev->cnf.stable_secret;
3411 get_random_bytes(&s->secret, sizeof(s->secret));
3412 s->initialized = true;
3413}
3414
3415static void addrconf_addr_gen(struct inet6_dev *idev, bool prefix_route)
3416{
3417 struct in6_addr addr;
3418
3419 /* no link local addresses on L3 master devices */
3420 if (netif_is_l3_master(idev->dev))
3421 return;
3422
3423 /* no link local addresses on devices flagged as slaves */
3424 if (idev->dev->priv_flags & IFF_NO_ADDRCONF)
3425 return;
3426
3427 ipv6_addr_set(&addr, htonl(0xFE800000), 0, 0, 0);
3428
3429 switch (idev->cnf.addr_gen_mode) {
3430 case IN6_ADDR_GEN_MODE_RANDOM:
3431 ipv6_gen_mode_random_init(idev);
3432 fallthrough;
3433 case IN6_ADDR_GEN_MODE_STABLE_PRIVACY:
3434 if (!ipv6_generate_stable_address(&addr, 0, idev))
3435 addrconf_add_linklocal(idev, &addr,
3436 IFA_F_STABLE_PRIVACY);
3437 else if (prefix_route)
3438 addrconf_prefix_route(&addr, 64, 0, idev->dev,
3439 0, 0, GFP_KERNEL);
3440 break;
3441 case IN6_ADDR_GEN_MODE_EUI64:
3442 /* addrconf_add_linklocal also adds a prefix_route and we
3443 * only need to care about prefix routes if ipv6_generate_eui64
3444 * couldn't generate one.
3445 */
3446 if (ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) == 0)
3447 addrconf_add_linklocal(idev, &addr, 0);
3448 else if (prefix_route)
3449 addrconf_prefix_route(&addr, 64, 0, idev->dev,
3450 0, 0, GFP_KERNEL);
3451 break;
3452 case IN6_ADDR_GEN_MODE_NONE:
3453 default:
3454 /* will not add any link local address */
3455 break;
3456 }
3457}
3458
3459static void addrconf_dev_config(struct net_device *dev)
3460{
3461 struct inet6_dev *idev;
3462
3463 ASSERT_RTNL();
3464
3465 if ((dev->type != ARPHRD_ETHER) &&
3466 (dev->type != ARPHRD_FDDI) &&
3467 (dev->type != ARPHRD_ARCNET) &&
3468 (dev->type != ARPHRD_INFINIBAND) &&
3469 (dev->type != ARPHRD_IEEE1394) &&
3470 (dev->type != ARPHRD_TUNNEL6) &&
3471 (dev->type != ARPHRD_6LOWPAN) &&
3472 (dev->type != ARPHRD_IP6GRE) &&
3473 (dev->type != ARPHRD_TUNNEL) &&
3474 (dev->type != ARPHRD_NONE) &&
3475 (dev->type != ARPHRD_RAWIP)) {
3476 /* Alas, we support only Ethernet autoconfiguration. */
3477 idev = __in6_dev_get(dev);
3478 if (!IS_ERR_OR_NULL(idev) && dev->flags & IFF_UP &&
3479 dev->flags & IFF_MULTICAST)
3480 ipv6_mc_up(idev);
3481 return;
3482 }
3483
3484 idev = addrconf_add_dev(dev);
3485 if (IS_ERR(idev))
3486 return;
3487
3488 /* this device type has no EUI support */
3489 if (dev->type == ARPHRD_NONE &&
3490 idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_EUI64)
3491 WRITE_ONCE(idev->cnf.addr_gen_mode,
3492 IN6_ADDR_GEN_MODE_RANDOM);
3493
3494 addrconf_addr_gen(idev, false);
3495}
3496
3497#if IS_ENABLED(CONFIG_IPV6_SIT)
3498static void addrconf_sit_config(struct net_device *dev)
3499{
3500 struct inet6_dev *idev;
3501
3502 ASSERT_RTNL();
3503
3504 /*
3505 * Configure the tunnel with one of our IPv4
3506 * addresses... we should configure all of
3507 * our v4 addrs in the tunnel
3508 */
3509
3510 idev = ipv6_find_idev(dev);
3511 if (IS_ERR(idev)) {
3512 pr_debug("%s: add_dev failed\n", __func__);
3513 return;
3514 }
3515
3516 if (dev->priv_flags & IFF_ISATAP) {
3517 addrconf_addr_gen(idev, false);
3518 return;
3519 }
3520
3521 add_v4_addrs(idev);
3522
3523 if (dev->flags&IFF_POINTOPOINT)
3524 addrconf_add_mroute(dev);
3525}
3526#endif
3527
3528#if IS_ENABLED(CONFIG_NET_IPGRE)
3529static void addrconf_gre_config(struct net_device *dev)
3530{
3531 struct inet6_dev *idev;
3532
3533 ASSERT_RTNL();
3534
3535 idev = addrconf_add_dev(dev);
3536 if (IS_ERR(idev))
3537 return;
3538
3539 /* Generate the IPv6 link-local address using addrconf_addr_gen(),
3540 * unless we have an IPv4 GRE device not bound to an IP address and
3541 * which is in EUI64 mode (as __ipv6_isatap_ifid() would fail in this
3542 * case). Such devices fall back to add_v4_addrs() instead.
3543 */
3544 if (!(*(__be32 *)dev->dev_addr == 0 &&
3545 idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_EUI64)) {
3546 addrconf_addr_gen(idev, true);
3547 return;
3548 }
3549
3550 add_v4_addrs(idev);
3551}
3552#endif
3553
3554static void addrconf_init_auto_addrs(struct net_device *dev)
3555{
3556 switch (dev->type) {
3557#if IS_ENABLED(CONFIG_IPV6_SIT)
3558 case ARPHRD_SIT:
3559 addrconf_sit_config(dev);
3560 break;
3561#endif
3562#if IS_ENABLED(CONFIG_NET_IPGRE)
3563 case ARPHRD_IPGRE:
3564 addrconf_gre_config(dev);
3565 break;
3566#endif
3567 case ARPHRD_LOOPBACK:
3568 init_loopback(dev);
3569 break;
3570
3571 default:
3572 addrconf_dev_config(dev);
3573 break;
3574 }
3575}
3576
3577static int fixup_permanent_addr(struct net *net,
3578 struct inet6_dev *idev,
3579 struct inet6_ifaddr *ifp)
3580{
3581 /* !fib6_node means the host route was removed from the
3582 * FIB, for example, if 'lo' device is taken down. In that
3583 * case regenerate the host route.
3584 */
3585 if (!ifp->rt || !ifp->rt->fib6_node) {
3586 struct fib6_info *f6i, *prev;
3587
3588 f6i = addrconf_f6i_alloc(net, idev, &ifp->addr, false,
3589 GFP_ATOMIC, NULL);
3590 if (IS_ERR(f6i))
3591 return PTR_ERR(f6i);
3592
3593 /* ifp->rt can be accessed outside of rtnl */
3594 spin_lock(&ifp->lock);
3595 prev = ifp->rt;
3596 ifp->rt = f6i;
3597 spin_unlock(&ifp->lock);
3598
3599 fib6_info_release(prev);
3600 }
3601
3602 if (!(ifp->flags & IFA_F_NOPREFIXROUTE)) {
3603 addrconf_prefix_route(&ifp->addr, ifp->prefix_len,
3604 ifp->rt_priority, idev->dev, 0, 0,
3605 GFP_ATOMIC);
3606 }
3607
3608 if (ifp->state == INET6_IFADDR_STATE_PREDAD)
3609 addrconf_dad_start(ifp);
3610
3611 return 0;
3612}
3613
3614static void addrconf_permanent_addr(struct net *net, struct net_device *dev)
3615{
3616 struct inet6_ifaddr *ifp, *tmp;
3617 struct inet6_dev *idev;
3618
3619 idev = __in6_dev_get(dev);
3620 if (!idev)
3621 return;
3622
3623 write_lock_bh(&idev->lock);
3624
3625 list_for_each_entry_safe(ifp, tmp, &idev->addr_list, if_list) {
3626 if ((ifp->flags & IFA_F_PERMANENT) &&
3627 fixup_permanent_addr(net, idev, ifp) < 0) {
3628 write_unlock_bh(&idev->lock);
3629 in6_ifa_hold(ifp);
3630 ipv6_del_addr(ifp);
3631 write_lock_bh(&idev->lock);
3632
3633 net_info_ratelimited("%s: Failed to add prefix route for address %pI6c; dropping\n",
3634 idev->dev->name, &ifp->addr);
3635 }
3636 }
3637
3638 write_unlock_bh(&idev->lock);
3639}
3640
3641static int addrconf_notify(struct notifier_block *this, unsigned long event,
3642 void *ptr)
3643{
3644 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
3645 struct netdev_notifier_change_info *change_info;
3646 struct netdev_notifier_changeupper_info *info;
3647 struct inet6_dev *idev = __in6_dev_get(dev);
3648 struct net *net = dev_net(dev);
3649 int run_pending = 0;
3650 int err;
3651
3652 switch (event) {
3653 case NETDEV_REGISTER:
3654 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
3655 idev = ipv6_add_dev(dev);
3656 if (IS_ERR(idev))
3657 return notifier_from_errno(PTR_ERR(idev));
3658 }
3659 break;
3660
3661 case NETDEV_CHANGEMTU:
3662 /* if MTU under IPV6_MIN_MTU stop IPv6 on this interface. */
3663 if (dev->mtu < IPV6_MIN_MTU) {
3664 addrconf_ifdown(dev, dev != net->loopback_dev);
3665 break;
3666 }
3667
3668 if (idev) {
3669 rt6_mtu_change(dev, dev->mtu);
3670 WRITE_ONCE(idev->cnf.mtu6, dev->mtu);
3671 break;
3672 }
3673
3674 /* allocate new idev */
3675 idev = ipv6_add_dev(dev);
3676 if (IS_ERR(idev))
3677 break;
3678
3679 /* device is still not ready */
3680 if (!(idev->if_flags & IF_READY))
3681 break;
3682
3683 run_pending = 1;
3684 fallthrough;
3685 case NETDEV_UP:
3686 case NETDEV_CHANGE:
3687 if (idev && idev->cnf.disable_ipv6)
3688 break;
3689
3690 if (dev->priv_flags & IFF_NO_ADDRCONF) {
3691 if (event == NETDEV_UP && !IS_ERR_OR_NULL(idev) &&
3692 dev->flags & IFF_UP && dev->flags & IFF_MULTICAST)
3693 ipv6_mc_up(idev);
3694 break;
3695 }
3696
3697 if (event == NETDEV_UP) {
3698 /* restore routes for permanent addresses */
3699 addrconf_permanent_addr(net, dev);
3700
3701 if (!addrconf_link_ready(dev)) {
3702 /* device is not ready yet. */
3703 pr_debug("ADDRCONF(NETDEV_UP): %s: link is not ready\n",
3704 dev->name);
3705 break;
3706 }
3707
3708 if (!idev && dev->mtu >= IPV6_MIN_MTU)
3709 idev = ipv6_add_dev(dev);
3710
3711 if (!IS_ERR_OR_NULL(idev)) {
3712 idev->if_flags |= IF_READY;
3713 run_pending = 1;
3714 }
3715 } else if (event == NETDEV_CHANGE) {
3716 if (!addrconf_link_ready(dev)) {
3717 /* device is still not ready. */
3718 rt6_sync_down_dev(dev, event);
3719 break;
3720 }
3721
3722 if (!IS_ERR_OR_NULL(idev)) {
3723 if (idev->if_flags & IF_READY) {
3724 /* device is already configured -
3725 * but resend MLD reports, we might
3726 * have roamed and need to update
3727 * multicast snooping switches
3728 */
3729 ipv6_mc_up(idev);
3730 change_info = ptr;
3731 if (change_info->flags_changed & IFF_NOARP)
3732 addrconf_dad_run(idev, true);
3733 rt6_sync_up(dev, RTNH_F_LINKDOWN);
3734 break;
3735 }
3736 idev->if_flags |= IF_READY;
3737 }
3738
3739 pr_debug("ADDRCONF(NETDEV_CHANGE): %s: link becomes ready\n",
3740 dev->name);
3741
3742 run_pending = 1;
3743 }
3744
3745 addrconf_init_auto_addrs(dev);
3746
3747 if (!IS_ERR_OR_NULL(idev)) {
3748 if (run_pending)
3749 addrconf_dad_run(idev, false);
3750
3751 /* Device has an address by now */
3752 rt6_sync_up(dev, RTNH_F_DEAD);
3753
3754 /*
3755 * If the MTU changed during the interface down,
3756 * when the interface up, the changed MTU must be
3757 * reflected in the idev as well as routers.
3758 */
3759 if (idev->cnf.mtu6 != dev->mtu &&
3760 dev->mtu >= IPV6_MIN_MTU) {
3761 rt6_mtu_change(dev, dev->mtu);
3762 WRITE_ONCE(idev->cnf.mtu6, dev->mtu);
3763 }
3764 WRITE_ONCE(idev->tstamp, jiffies);
3765 inet6_ifinfo_notify(RTM_NEWLINK, idev);
3766
3767 /*
3768 * If the changed mtu during down is lower than
3769 * IPV6_MIN_MTU stop IPv6 on this interface.
3770 */
3771 if (dev->mtu < IPV6_MIN_MTU)
3772 addrconf_ifdown(dev, dev != net->loopback_dev);
3773 }
3774 break;
3775
3776 case NETDEV_DOWN:
3777 case NETDEV_UNREGISTER:
3778 /*
3779 * Remove all addresses from this interface.
3780 */
3781 addrconf_ifdown(dev, event != NETDEV_DOWN);
3782 break;
3783
3784 case NETDEV_CHANGENAME:
3785 if (idev) {
3786 snmp6_unregister_dev(idev);
3787 addrconf_sysctl_unregister(idev);
3788 err = addrconf_sysctl_register(idev);
3789 if (err)
3790 return notifier_from_errno(err);
3791 err = snmp6_register_dev(idev);
3792 if (err) {
3793 addrconf_sysctl_unregister(idev);
3794 return notifier_from_errno(err);
3795 }
3796 }
3797 break;
3798
3799 case NETDEV_PRE_TYPE_CHANGE:
3800 case NETDEV_POST_TYPE_CHANGE:
3801 if (idev)
3802 addrconf_type_change(dev, event);
3803 break;
3804
3805 case NETDEV_CHANGEUPPER:
3806 info = ptr;
3807
3808 /* flush all routes if dev is linked to or unlinked from
3809 * an L3 master device (e.g., VRF)
3810 */
3811 if (info->upper_dev && netif_is_l3_master(info->upper_dev))
3812 addrconf_ifdown(dev, false);
3813 }
3814
3815 return NOTIFY_OK;
3816}
3817
3818/*
3819 * addrconf module should be notified of a device going up
3820 */
3821static struct notifier_block ipv6_dev_notf = {
3822 .notifier_call = addrconf_notify,
3823 .priority = ADDRCONF_NOTIFY_PRIORITY,
3824};
3825
3826static void addrconf_type_change(struct net_device *dev, unsigned long event)
3827{
3828 struct inet6_dev *idev;
3829 ASSERT_RTNL();
3830
3831 idev = __in6_dev_get(dev);
3832
3833 if (event == NETDEV_POST_TYPE_CHANGE)
3834 ipv6_mc_remap(idev);
3835 else if (event == NETDEV_PRE_TYPE_CHANGE)
3836 ipv6_mc_unmap(idev);
3837}
3838
3839static bool addr_is_local(const struct in6_addr *addr)
3840{
3841 return ipv6_addr_type(addr) &
3842 (IPV6_ADDR_LINKLOCAL | IPV6_ADDR_LOOPBACK);
3843}
3844
3845static int addrconf_ifdown(struct net_device *dev, bool unregister)
3846{
3847 unsigned long event = unregister ? NETDEV_UNREGISTER : NETDEV_DOWN;
3848 struct net *net = dev_net(dev);
3849 struct inet6_dev *idev;
3850 struct inet6_ifaddr *ifa;
3851 LIST_HEAD(tmp_addr_list);
3852 bool keep_addr = false;
3853 bool was_ready;
3854 int state, i;
3855
3856 ASSERT_RTNL();
3857
3858 rt6_disable_ip(dev, event);
3859
3860 idev = __in6_dev_get(dev);
3861 if (!idev)
3862 return -ENODEV;
3863
3864 /*
3865 * Step 1: remove reference to ipv6 device from parent device.
3866 * Do not dev_put!
3867 */
3868 if (unregister) {
3869 WRITE_ONCE(idev->dead, 1);
3870
3871 /* protected by rtnl_lock */
3872 RCU_INIT_POINTER(dev->ip6_ptr, NULL);
3873
3874 /* Step 1.5: remove snmp6 entry */
3875 snmp6_unregister_dev(idev);
3876
3877 }
3878
3879 /* combine the user config with event to determine if permanent
3880 * addresses are to be removed from address hash table
3881 */
3882 if (!unregister && !idev->cnf.disable_ipv6) {
3883 /* aggregate the system setting and interface setting */
3884 int _keep_addr = READ_ONCE(net->ipv6.devconf_all->keep_addr_on_down);
3885
3886 if (!_keep_addr)
3887 _keep_addr = READ_ONCE(idev->cnf.keep_addr_on_down);
3888
3889 keep_addr = (_keep_addr > 0);
3890 }
3891
3892 /* Step 2: clear hash table */
3893 for (i = 0; i < IN6_ADDR_HSIZE; i++) {
3894 struct hlist_head *h = &net->ipv6.inet6_addr_lst[i];
3895
3896 spin_lock_bh(&net->ipv6.addrconf_hash_lock);
3897restart:
3898 hlist_for_each_entry_rcu(ifa, h, addr_lst) {
3899 if (ifa->idev == idev) {
3900 addrconf_del_dad_work(ifa);
3901 /* combined flag + permanent flag decide if
3902 * address is retained on a down event
3903 */
3904 if (!keep_addr ||
3905 !(ifa->flags & IFA_F_PERMANENT) ||
3906 addr_is_local(&ifa->addr)) {
3907 hlist_del_init_rcu(&ifa->addr_lst);
3908 goto restart;
3909 }
3910 }
3911 }
3912 spin_unlock_bh(&net->ipv6.addrconf_hash_lock);
3913 }
3914
3915 write_lock_bh(&idev->lock);
3916
3917 addrconf_del_rs_timer(idev);
3918
3919 /* Step 2: clear flags for stateless addrconf, repeated down
3920 * detection
3921 */
3922 was_ready = idev->if_flags & IF_READY;
3923 if (!unregister)
3924 idev->if_flags &= ~(IF_RS_SENT|IF_RA_RCVD|IF_READY);
3925
3926 /* Step 3: clear tempaddr list */
3927 while (!list_empty(&idev->tempaddr_list)) {
3928 ifa = list_first_entry(&idev->tempaddr_list,
3929 struct inet6_ifaddr, tmp_list);
3930 list_del(&ifa->tmp_list);
3931 write_unlock_bh(&idev->lock);
3932 spin_lock_bh(&ifa->lock);
3933
3934 if (ifa->ifpub) {
3935 in6_ifa_put(ifa->ifpub);
3936 ifa->ifpub = NULL;
3937 }
3938 spin_unlock_bh(&ifa->lock);
3939 in6_ifa_put(ifa);
3940 write_lock_bh(&idev->lock);
3941 }
3942
3943 list_for_each_entry(ifa, &idev->addr_list, if_list)
3944 list_add_tail(&ifa->if_list_aux, &tmp_addr_list);
3945 write_unlock_bh(&idev->lock);
3946
3947 while (!list_empty(&tmp_addr_list)) {
3948 struct fib6_info *rt = NULL;
3949 bool keep;
3950
3951 ifa = list_first_entry(&tmp_addr_list,
3952 struct inet6_ifaddr, if_list_aux);
3953 list_del(&ifa->if_list_aux);
3954
3955 addrconf_del_dad_work(ifa);
3956
3957 keep = keep_addr && (ifa->flags & IFA_F_PERMANENT) &&
3958 !addr_is_local(&ifa->addr);
3959
3960 spin_lock_bh(&ifa->lock);
3961
3962 if (keep) {
3963 /* set state to skip the notifier below */
3964 state = INET6_IFADDR_STATE_DEAD;
3965 ifa->state = INET6_IFADDR_STATE_PREDAD;
3966 if (!(ifa->flags & IFA_F_NODAD))
3967 ifa->flags |= IFA_F_TENTATIVE;
3968
3969 rt = ifa->rt;
3970 ifa->rt = NULL;
3971 } else {
3972 state = ifa->state;
3973 ifa->state = INET6_IFADDR_STATE_DEAD;
3974 }
3975
3976 spin_unlock_bh(&ifa->lock);
3977
3978 if (rt)
3979 ip6_del_rt(net, rt, false);
3980
3981 if (state != INET6_IFADDR_STATE_DEAD) {
3982 __ipv6_ifa_notify(RTM_DELADDR, ifa);
3983 inet6addr_notifier_call_chain(NETDEV_DOWN, ifa);
3984 } else {
3985 if (idev->cnf.forwarding)
3986 addrconf_leave_anycast(ifa);
3987 addrconf_leave_solict(ifa->idev, &ifa->addr);
3988 }
3989
3990 if (!keep) {
3991 write_lock_bh(&idev->lock);
3992 list_del_rcu(&ifa->if_list);
3993 write_unlock_bh(&idev->lock);
3994 in6_ifa_put(ifa);
3995 }
3996 }
3997
3998 /* Step 5: Discard anycast and multicast list */
3999 if (unregister) {
4000 ipv6_ac_destroy_dev(idev);
4001 ipv6_mc_destroy_dev(idev);
4002 } else if (was_ready) {
4003 ipv6_mc_down(idev);
4004 }
4005
4006 WRITE_ONCE(idev->tstamp, jiffies);
4007 idev->ra_mtu = 0;
4008
4009 /* Last: Shot the device (if unregistered) */
4010 if (unregister) {
4011 addrconf_sysctl_unregister(idev);
4012 neigh_parms_release(&nd_tbl, idev->nd_parms);
4013 neigh_ifdown(&nd_tbl, dev);
4014 in6_dev_put(idev);
4015 }
4016 return 0;
4017}
4018
4019static void addrconf_rs_timer(struct timer_list *t)
4020{
4021 struct inet6_dev *idev = timer_container_of(idev, t, rs_timer);
4022 struct net_device *dev = idev->dev;
4023 struct in6_addr lladdr;
4024 int rtr_solicits;
4025
4026 write_lock(&idev->lock);
4027 if (idev->dead || !(idev->if_flags & IF_READY))
4028 goto out;
4029
4030 if (!ipv6_accept_ra(idev))
4031 goto out;
4032
4033 /* Announcement received after solicitation was sent */
4034 if (idev->if_flags & IF_RA_RCVD)
4035 goto out;
4036
4037 rtr_solicits = READ_ONCE(idev->cnf.rtr_solicits);
4038
4039 if (idev->rs_probes++ < rtr_solicits || rtr_solicits < 0) {
4040 write_unlock(&idev->lock);
4041 if (!ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE))
4042 ndisc_send_rs(dev, &lladdr,
4043 &in6addr_linklocal_allrouters);
4044 else
4045 goto put;
4046
4047 write_lock(&idev->lock);
4048 idev->rs_interval = rfc3315_s14_backoff_update(
4049 idev->rs_interval,
4050 READ_ONCE(idev->cnf.rtr_solicit_max_interval));
4051 /* The wait after the last probe can be shorter */
4052 addrconf_mod_rs_timer(idev, (idev->rs_probes ==
4053 READ_ONCE(idev->cnf.rtr_solicits)) ?
4054 READ_ONCE(idev->cnf.rtr_solicit_delay) :
4055 idev->rs_interval);
4056 } else {
4057 /*
4058 * Note: we do not support deprecated "all on-link"
4059 * assumption any longer.
4060 */
4061 pr_debug("%s: no IPv6 routers present\n", idev->dev->name);
4062 }
4063
4064out:
4065 write_unlock(&idev->lock);
4066put:
4067 in6_dev_put(idev);
4068}
4069
4070/*
4071 * Duplicate Address Detection
4072 */
4073static void addrconf_dad_kick(struct inet6_ifaddr *ifp)
4074{
4075 struct inet6_dev *idev = ifp->idev;
4076 unsigned long rand_num;
4077 u64 nonce;
4078
4079 if (ifp->flags & IFA_F_OPTIMISTIC)
4080 rand_num = 0;
4081 else
4082 rand_num = get_random_u32_below(
4083 READ_ONCE(idev->cnf.rtr_solicit_delay) ? : 1);
4084
4085 nonce = 0;
4086 if (READ_ONCE(idev->cnf.enhanced_dad) ||
4087 READ_ONCE(dev_net(idev->dev)->ipv6.devconf_all->enhanced_dad)) {
4088 do
4089 get_random_bytes(&nonce, 6);
4090 while (nonce == 0);
4091 }
4092 ifp->dad_nonce = nonce;
4093 ifp->dad_probes = READ_ONCE(idev->cnf.dad_transmits);
4094 addrconf_mod_dad_work(ifp, rand_num);
4095}
4096
4097static void addrconf_dad_begin(struct inet6_ifaddr *ifp)
4098{
4099 struct inet6_dev *idev = ifp->idev;
4100 struct net_device *dev = idev->dev;
4101 bool bump_id, notify = false;
4102 struct net *net;
4103
4104 addrconf_join_solict(dev, &ifp->addr);
4105
4106 read_lock_bh(&idev->lock);
4107 spin_lock(&ifp->lock);
4108 if (ifp->state == INET6_IFADDR_STATE_DEAD)
4109 goto out;
4110
4111 net = dev_net(dev);
4112 if (dev->flags&(IFF_NOARP|IFF_LOOPBACK) ||
4113 (READ_ONCE(net->ipv6.devconf_all->accept_dad) < 1 &&
4114 READ_ONCE(idev->cnf.accept_dad) < 1) ||
4115 !(ifp->flags&IFA_F_TENTATIVE) ||
4116 ifp->flags & IFA_F_NODAD) {
4117 bool send_na = false;
4118
4119 if (ifp->flags & IFA_F_TENTATIVE &&
4120 !(ifp->flags & IFA_F_OPTIMISTIC))
4121 send_na = true;
4122 bump_id = ifp->flags & IFA_F_TENTATIVE;
4123 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
4124 spin_unlock(&ifp->lock);
4125 read_unlock_bh(&idev->lock);
4126
4127 addrconf_dad_completed(ifp, bump_id, send_na);
4128 return;
4129 }
4130
4131 if (!(idev->if_flags & IF_READY)) {
4132 spin_unlock(&ifp->lock);
4133 read_unlock_bh(&idev->lock);
4134 /*
4135 * If the device is not ready:
4136 * - keep it tentative if it is a permanent address.
4137 * - otherwise, kill it.
4138 */
4139 in6_ifa_hold(ifp);
4140 addrconf_dad_stop(ifp, 0);
4141 return;
4142 }
4143
4144 /*
4145 * Optimistic nodes can start receiving
4146 * Frames right away
4147 */
4148 if (ifp->flags & IFA_F_OPTIMISTIC) {
4149 ip6_ins_rt(net, ifp->rt);
4150 if (ipv6_use_optimistic_addr(net, idev)) {
4151 /* Because optimistic nodes can use this address,
4152 * notify listeners. If DAD fails, RTM_DELADDR is sent.
4153 */
4154 notify = true;
4155 }
4156 }
4157
4158 addrconf_dad_kick(ifp);
4159out:
4160 spin_unlock(&ifp->lock);
4161 read_unlock_bh(&idev->lock);
4162 if (notify)
4163 ipv6_ifa_notify(RTM_NEWADDR, ifp);
4164}
4165
4166static void addrconf_dad_start(struct inet6_ifaddr *ifp)
4167{
4168 bool begin_dad = false;
4169
4170 spin_lock_bh(&ifp->lock);
4171 if (ifp->state != INET6_IFADDR_STATE_DEAD) {
4172 ifp->state = INET6_IFADDR_STATE_PREDAD;
4173 begin_dad = true;
4174 }
4175 spin_unlock_bh(&ifp->lock);
4176
4177 if (begin_dad)
4178 addrconf_mod_dad_work(ifp, 0);
4179}
4180
4181static void addrconf_dad_work(struct work_struct *w)
4182{
4183 struct inet6_ifaddr *ifp = container_of(to_delayed_work(w),
4184 struct inet6_ifaddr,
4185 dad_work);
4186 struct inet6_dev *idev = ifp->idev;
4187 bool bump_id, disable_ipv6 = false;
4188 struct in6_addr mcaddr;
4189 struct net *net;
4190
4191 enum {
4192 DAD_PROCESS,
4193 DAD_BEGIN,
4194 DAD_ABORT,
4195 } action = DAD_PROCESS;
4196
4197 net = dev_net(idev->dev);
4198
4199 rtnl_net_lock(net);
4200
4201 spin_lock_bh(&ifp->lock);
4202 if (ifp->state == INET6_IFADDR_STATE_PREDAD) {
4203 action = DAD_BEGIN;
4204 ifp->state = INET6_IFADDR_STATE_DAD;
4205 } else if (ifp->state == INET6_IFADDR_STATE_ERRDAD) {
4206 action = DAD_ABORT;
4207 ifp->state = INET6_IFADDR_STATE_POSTDAD;
4208
4209 if ((READ_ONCE(net->ipv6.devconf_all->accept_dad) > 1 ||
4210 READ_ONCE(idev->cnf.accept_dad) > 1) &&
4211 !idev->cnf.disable_ipv6 &&
4212 !(ifp->flags & IFA_F_STABLE_PRIVACY)) {
4213 struct in6_addr addr;
4214
4215 addr.s6_addr32[0] = htonl(0xfe800000);
4216 addr.s6_addr32[1] = 0;
4217
4218 if (!ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) &&
4219 ipv6_addr_equal(&ifp->addr, &addr)) {
4220 /* DAD failed for link-local based on MAC */
4221 WRITE_ONCE(idev->cnf.disable_ipv6, 1);
4222
4223 pr_info("%s: IPv6 being disabled!\n",
4224 ifp->idev->dev->name);
4225 disable_ipv6 = true;
4226 }
4227 }
4228 }
4229 spin_unlock_bh(&ifp->lock);
4230
4231 if (action == DAD_BEGIN) {
4232 addrconf_dad_begin(ifp);
4233 goto out;
4234 } else if (action == DAD_ABORT) {
4235 in6_ifa_hold(ifp);
4236 addrconf_dad_stop(ifp, 1);
4237 if (disable_ipv6)
4238 addrconf_ifdown(idev->dev, false);
4239 goto out;
4240 }
4241
4242 if (!ifp->dad_probes && addrconf_dad_end(ifp))
4243 goto out;
4244
4245 write_lock_bh(&idev->lock);
4246 if (idev->dead || !(idev->if_flags & IF_READY)) {
4247 write_unlock_bh(&idev->lock);
4248 goto out;
4249 }
4250
4251 spin_lock(&ifp->lock);
4252 if (ifp->state == INET6_IFADDR_STATE_DEAD) {
4253 spin_unlock(&ifp->lock);
4254 write_unlock_bh(&idev->lock);
4255 goto out;
4256 }
4257
4258 if (ifp->dad_probes == 0) {
4259 bool send_na = false;
4260
4261 /*
4262 * DAD was successful
4263 */
4264
4265 if (ifp->flags & IFA_F_TENTATIVE &&
4266 !(ifp->flags & IFA_F_OPTIMISTIC))
4267 send_na = true;
4268 bump_id = ifp->flags & IFA_F_TENTATIVE;
4269 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
4270 spin_unlock(&ifp->lock);
4271 write_unlock_bh(&idev->lock);
4272
4273 addrconf_dad_completed(ifp, bump_id, send_na);
4274
4275 goto out;
4276 }
4277
4278 ifp->dad_probes--;
4279 addrconf_mod_dad_work(ifp,
4280 max(NEIGH_VAR(ifp->idev->nd_parms, RETRANS_TIME),
4281 HZ/100));
4282 spin_unlock(&ifp->lock);
4283 write_unlock_bh(&idev->lock);
4284
4285 /* send a neighbour solicitation for our addr */
4286 addrconf_addr_solict_mult(&ifp->addr, &mcaddr);
4287 ndisc_send_ns(ifp->idev->dev, &ifp->addr, &mcaddr, &in6addr_any,
4288 ifp->dad_nonce);
4289out:
4290 in6_ifa_put(ifp);
4291 rtnl_net_unlock(net);
4292}
4293
4294/* ifp->idev must be at least read locked */
4295static bool ipv6_lonely_lladdr(struct inet6_ifaddr *ifp)
4296{
4297 struct inet6_ifaddr *ifpiter;
4298 struct inet6_dev *idev = ifp->idev;
4299
4300 list_for_each_entry_reverse(ifpiter, &idev->addr_list, if_list) {
4301 if (ifpiter->scope > IFA_LINK)
4302 break;
4303 if (ifp != ifpiter && ifpiter->scope == IFA_LINK &&
4304 (ifpiter->flags & (IFA_F_PERMANENT|IFA_F_TENTATIVE|
4305 IFA_F_OPTIMISTIC|IFA_F_DADFAILED)) ==
4306 IFA_F_PERMANENT)
4307 return false;
4308 }
4309 return true;
4310}
4311
4312static void addrconf_dad_completed(struct inet6_ifaddr *ifp, bool bump_id,
4313 bool send_na)
4314{
4315 struct net_device *dev = ifp->idev->dev;
4316 struct in6_addr lladdr;
4317 bool send_rs, send_mld;
4318
4319 addrconf_del_dad_work(ifp);
4320
4321 /*
4322 * Configure the address for reception. Now it is valid.
4323 */
4324
4325 ipv6_ifa_notify(RTM_NEWADDR, ifp);
4326
4327 /* If added prefix is link local and we are prepared to process
4328 router advertisements, start sending router solicitations.
4329 */
4330
4331 read_lock_bh(&ifp->idev->lock);
4332 send_mld = ifp->scope == IFA_LINK && ipv6_lonely_lladdr(ifp);
4333 send_rs = send_mld &&
4334 ipv6_accept_ra(ifp->idev) &&
4335 READ_ONCE(ifp->idev->cnf.rtr_solicits) != 0 &&
4336 (dev->flags & IFF_LOOPBACK) == 0 &&
4337 (dev->type != ARPHRD_TUNNEL) &&
4338 !netif_is_team_port(dev);
4339 read_unlock_bh(&ifp->idev->lock);
4340
4341 /* While dad is in progress mld report's source address is in6_addrany.
4342 * Resend with proper ll now.
4343 */
4344 if (send_mld)
4345 ipv6_mc_dad_complete(ifp->idev);
4346
4347 /* send unsolicited NA if enabled */
4348 if (send_na &&
4349 (READ_ONCE(ifp->idev->cnf.ndisc_notify) ||
4350 READ_ONCE(dev_net(dev)->ipv6.devconf_all->ndisc_notify))) {
4351 ndisc_send_na(dev, &in6addr_linklocal_allnodes, &ifp->addr,
4352 /*router=*/ !!ifp->idev->cnf.forwarding,
4353 /*solicited=*/ false, /*override=*/ true,
4354 /*inc_opt=*/ true);
4355 }
4356
4357 if (send_rs) {
4358 /*
4359 * If a host as already performed a random delay
4360 * [...] as part of DAD [...] there is no need
4361 * to delay again before sending the first RS
4362 */
4363 if (ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE))
4364 return;
4365 ndisc_send_rs(dev, &lladdr, &in6addr_linklocal_allrouters);
4366
4367 write_lock_bh(&ifp->idev->lock);
4368 spin_lock(&ifp->lock);
4369 ifp->idev->rs_interval = rfc3315_s14_backoff_init(
4370 READ_ONCE(ifp->idev->cnf.rtr_solicit_interval));
4371 ifp->idev->rs_probes = 1;
4372 ifp->idev->if_flags |= IF_RS_SENT;
4373 addrconf_mod_rs_timer(ifp->idev, ifp->idev->rs_interval);
4374 spin_unlock(&ifp->lock);
4375 write_unlock_bh(&ifp->idev->lock);
4376 }
4377
4378 if (bump_id)
4379 rt_genid_bump_ipv6(dev_net(dev));
4380
4381 /* Make sure that a new temporary address will be created
4382 * before this temporary address becomes deprecated.
4383 */
4384 if (ifp->flags & IFA_F_TEMPORARY)
4385 addrconf_verify_rtnl(dev_net(dev));
4386}
4387
4388static void addrconf_dad_run(struct inet6_dev *idev, bool restart)
4389{
4390 struct inet6_ifaddr *ifp;
4391
4392 read_lock_bh(&idev->lock);
4393 list_for_each_entry(ifp, &idev->addr_list, if_list) {
4394 spin_lock(&ifp->lock);
4395 if ((ifp->flags & IFA_F_TENTATIVE &&
4396 ifp->state == INET6_IFADDR_STATE_DAD) || restart) {
4397 if (restart)
4398 ifp->state = INET6_IFADDR_STATE_PREDAD;
4399 addrconf_dad_kick(ifp);
4400 }
4401 spin_unlock(&ifp->lock);
4402 }
4403 read_unlock_bh(&idev->lock);
4404}
4405
4406#ifdef CONFIG_PROC_FS
4407struct if6_iter_state {
4408 struct seq_net_private p;
4409 int bucket;
4410 int offset;
4411};
4412
4413static struct inet6_ifaddr *if6_get_first(struct seq_file *seq, loff_t pos)
4414{
4415 struct if6_iter_state *state = seq->private;
4416 struct net *net = seq_file_net(seq);
4417 struct inet6_ifaddr *ifa = NULL;
4418 int p = 0;
4419
4420 /* initial bucket if pos is 0 */
4421 if (pos == 0) {
4422 state->bucket = 0;
4423 state->offset = 0;
4424 }
4425
4426 for (; state->bucket < IN6_ADDR_HSIZE; ++state->bucket) {
4427 hlist_for_each_entry_rcu(ifa, &net->ipv6.inet6_addr_lst[state->bucket],
4428 addr_lst) {
4429 /* sync with offset */
4430 if (p < state->offset) {
4431 p++;
4432 continue;
4433 }
4434 return ifa;
4435 }
4436
4437 /* prepare for next bucket */
4438 state->offset = 0;
4439 p = 0;
4440 }
4441 return NULL;
4442}
4443
4444static struct inet6_ifaddr *if6_get_next(struct seq_file *seq,
4445 struct inet6_ifaddr *ifa)
4446{
4447 struct if6_iter_state *state = seq->private;
4448 struct net *net = seq_file_net(seq);
4449
4450 hlist_for_each_entry_continue_rcu(ifa, addr_lst) {
4451 state->offset++;
4452 return ifa;
4453 }
4454
4455 state->offset = 0;
4456 while (++state->bucket < IN6_ADDR_HSIZE) {
4457 hlist_for_each_entry_rcu(ifa,
4458 &net->ipv6.inet6_addr_lst[state->bucket], addr_lst) {
4459 return ifa;
4460 }
4461 }
4462
4463 return NULL;
4464}
4465
4466static void *if6_seq_start(struct seq_file *seq, loff_t *pos)
4467 __acquires(rcu)
4468{
4469 rcu_read_lock();
4470 return if6_get_first(seq, *pos);
4471}
4472
4473static void *if6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
4474{
4475 struct inet6_ifaddr *ifa;
4476
4477 ifa = if6_get_next(seq, v);
4478 ++*pos;
4479 return ifa;
4480}
4481
4482static void if6_seq_stop(struct seq_file *seq, void *v)
4483 __releases(rcu)
4484{
4485 rcu_read_unlock();
4486}
4487
4488static int if6_seq_show(struct seq_file *seq, void *v)
4489{
4490 struct inet6_ifaddr *ifp = (struct inet6_ifaddr *)v;
4491 seq_printf(seq, "%pi6 %02x %02x %02x %02x %8s\n",
4492 &ifp->addr,
4493 ifp->idev->dev->ifindex,
4494 ifp->prefix_len,
4495 ifp->scope,
4496 (u8) ifp->flags,
4497 ifp->idev->dev->name);
4498 return 0;
4499}
4500
4501static const struct seq_operations if6_seq_ops = {
4502 .start = if6_seq_start,
4503 .next = if6_seq_next,
4504 .show = if6_seq_show,
4505 .stop = if6_seq_stop,
4506};
4507
4508static int __net_init if6_proc_net_init(struct net *net)
4509{
4510 if (!proc_create_net("if_inet6", 0444, net->proc_net, &if6_seq_ops,
4511 sizeof(struct if6_iter_state)))
4512 return -ENOMEM;
4513 return 0;
4514}
4515
4516static void __net_exit if6_proc_net_exit(struct net *net)
4517{
4518 remove_proc_entry("if_inet6", net->proc_net);
4519}
4520
4521static struct pernet_operations if6_proc_net_ops = {
4522 .init = if6_proc_net_init,
4523 .exit = if6_proc_net_exit,
4524};
4525
4526int __init if6_proc_init(void)
4527{
4528 return register_pernet_subsys(&if6_proc_net_ops);
4529}
4530
4531void if6_proc_exit(void)
4532{
4533 unregister_pernet_subsys(&if6_proc_net_ops);
4534}
4535#endif /* CONFIG_PROC_FS */
4536
4537#if IS_ENABLED(CONFIG_IPV6_MIP6)
4538/* Check if address is a home address configured on any interface. */
4539int ipv6_chk_home_addr(struct net *net, const struct in6_addr *addr)
4540{
4541 unsigned int hash = inet6_addr_hash(net, addr);
4542 struct inet6_ifaddr *ifp = NULL;
4543 int ret = 0;
4544
4545 rcu_read_lock();
4546 hlist_for_each_entry_rcu(ifp, &net->ipv6.inet6_addr_lst[hash], addr_lst) {
4547 if (ipv6_addr_equal(&ifp->addr, addr) &&
4548 (ifp->flags & IFA_F_HOMEADDRESS)) {
4549 ret = 1;
4550 break;
4551 }
4552 }
4553 rcu_read_unlock();
4554 return ret;
4555}
4556#endif
4557
4558/* RFC6554 has some algorithm to avoid loops in segment routing by
4559 * checking if the segments contains any of a local interface address.
4560 *
4561 * Quote:
4562 *
4563 * To detect loops in the SRH, a router MUST determine if the SRH
4564 * includes multiple addresses assigned to any interface on that router.
4565 * If such addresses appear more than once and are separated by at least
4566 * one address not assigned to that router.
4567 */
4568int ipv6_chk_rpl_srh_loop(struct net *net, const struct in6_addr *segs,
4569 unsigned char nsegs)
4570{
4571 const struct in6_addr *addr;
4572 int i, ret = 0, found = 0;
4573 struct inet6_ifaddr *ifp;
4574 bool separated = false;
4575 unsigned int hash;
4576 bool hash_found;
4577
4578 rcu_read_lock();
4579 for (i = 0; i < nsegs; i++) {
4580 addr = &segs[i];
4581 hash = inet6_addr_hash(net, addr);
4582
4583 hash_found = false;
4584 hlist_for_each_entry_rcu(ifp, &net->ipv6.inet6_addr_lst[hash], addr_lst) {
4585
4586 if (ipv6_addr_equal(&ifp->addr, addr)) {
4587 hash_found = true;
4588 break;
4589 }
4590 }
4591
4592 if (hash_found) {
4593 if (found > 1 && separated) {
4594 ret = 1;
4595 break;
4596 }
4597
4598 separated = false;
4599 found++;
4600 } else {
4601 separated = true;
4602 }
4603 }
4604 rcu_read_unlock();
4605
4606 return ret;
4607}
4608
4609/*
4610 * Periodic address status verification
4611 */
4612
4613static void addrconf_verify_rtnl(struct net *net)
4614{
4615 unsigned long now, next, next_sec, next_sched;
4616 struct inet6_ifaddr *ifp;
4617 int i;
4618
4619 ASSERT_RTNL();
4620
4621 rcu_read_lock_bh();
4622 now = jiffies;
4623 next = round_jiffies_up(now + ADDR_CHECK_FREQUENCY);
4624
4625 cancel_delayed_work(&net->ipv6.addr_chk_work);
4626
4627 for (i = 0; i < IN6_ADDR_HSIZE; i++) {
4628restart:
4629 hlist_for_each_entry_rcu_bh(ifp, &net->ipv6.inet6_addr_lst[i], addr_lst) {
4630 unsigned long age;
4631
4632 /* When setting preferred_lft to a value not zero or
4633 * infinity, while valid_lft is infinity
4634 * IFA_F_PERMANENT has a non-infinity life time.
4635 */
4636 if ((ifp->flags & IFA_F_PERMANENT) &&
4637 (ifp->prefered_lft == INFINITY_LIFE_TIME))
4638 continue;
4639
4640 spin_lock(&ifp->lock);
4641 /* We try to batch several events at once. */
4642 age = (now - ifp->tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
4643
4644 if ((ifp->flags&IFA_F_TEMPORARY) &&
4645 !(ifp->flags&IFA_F_TENTATIVE) &&
4646 ifp->prefered_lft != INFINITY_LIFE_TIME &&
4647 !ifp->regen_count && ifp->ifpub) {
4648 /* This is a non-regenerated temporary addr. */
4649
4650 unsigned long regen_advance = ipv6_get_regen_advance(ifp->idev);
4651
4652 if (age + regen_advance >= ifp->prefered_lft) {
4653 struct inet6_ifaddr *ifpub = ifp->ifpub;
4654 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
4655 next = ifp->tstamp + ifp->prefered_lft * HZ;
4656
4657 ifp->regen_count++;
4658 in6_ifa_hold(ifp);
4659 in6_ifa_hold(ifpub);
4660 spin_unlock(&ifp->lock);
4661
4662 spin_lock(&ifpub->lock);
4663 ifpub->regen_count = 0;
4664 spin_unlock(&ifpub->lock);
4665 rcu_read_unlock_bh();
4666 ipv6_create_tempaddr(ifpub, true);
4667 in6_ifa_put(ifpub);
4668 in6_ifa_put(ifp);
4669 rcu_read_lock_bh();
4670 goto restart;
4671 } else if (time_before(ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ, next))
4672 next = ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ;
4673 }
4674
4675 if (ifp->valid_lft != INFINITY_LIFE_TIME &&
4676 age >= ifp->valid_lft) {
4677 spin_unlock(&ifp->lock);
4678 in6_ifa_hold(ifp);
4679 rcu_read_unlock_bh();
4680 ipv6_del_addr(ifp);
4681 rcu_read_lock_bh();
4682 goto restart;
4683 } else if (ifp->prefered_lft == INFINITY_LIFE_TIME) {
4684 spin_unlock(&ifp->lock);
4685 continue;
4686 } else if (age >= ifp->prefered_lft) {
4687 /* jiffies - ifp->tstamp > age >= ifp->prefered_lft */
4688 int deprecate = 0;
4689
4690 if (!(ifp->flags&IFA_F_DEPRECATED)) {
4691 deprecate = 1;
4692 ifp->flags |= IFA_F_DEPRECATED;
4693 }
4694
4695 if ((ifp->valid_lft != INFINITY_LIFE_TIME) &&
4696 (time_before(ifp->tstamp + ifp->valid_lft * HZ, next)))
4697 next = ifp->tstamp + ifp->valid_lft * HZ;
4698
4699 spin_unlock(&ifp->lock);
4700
4701 if (deprecate) {
4702 in6_ifa_hold(ifp);
4703
4704 ipv6_ifa_notify(0, ifp);
4705 in6_ifa_put(ifp);
4706 goto restart;
4707 }
4708 } else {
4709 /* ifp->prefered_lft <= ifp->valid_lft */
4710 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
4711 next = ifp->tstamp + ifp->prefered_lft * HZ;
4712 spin_unlock(&ifp->lock);
4713 }
4714 }
4715 }
4716
4717 next_sec = round_jiffies_up(next);
4718 next_sched = next;
4719
4720 /* If rounded timeout is accurate enough, accept it. */
4721 if (time_before(next_sec, next + ADDRCONF_TIMER_FUZZ))
4722 next_sched = next_sec;
4723
4724 /* And minimum interval is ADDRCONF_TIMER_FUZZ_MAX. */
4725 if (time_before(next_sched, jiffies + ADDRCONF_TIMER_FUZZ_MAX))
4726 next_sched = jiffies + ADDRCONF_TIMER_FUZZ_MAX;
4727
4728 pr_debug("now = %lu, schedule = %lu, rounded schedule = %lu => %lu\n",
4729 now, next, next_sec, next_sched);
4730 mod_delayed_work(addrconf_wq, &net->ipv6.addr_chk_work, next_sched - now);
4731 rcu_read_unlock_bh();
4732}
4733
4734static void addrconf_verify_work(struct work_struct *w)
4735{
4736 struct net *net = container_of(to_delayed_work(w), struct net,
4737 ipv6.addr_chk_work);
4738
4739 rtnl_net_lock(net);
4740 addrconf_verify_rtnl(net);
4741 rtnl_net_unlock(net);
4742}
4743
4744static void addrconf_verify(struct net *net)
4745{
4746 mod_delayed_work(addrconf_wq, &net->ipv6.addr_chk_work, 0);
4747}
4748
4749static struct in6_addr *extract_addr(struct nlattr *addr, struct nlattr *local,
4750 struct in6_addr **peer_pfx)
4751{
4752 struct in6_addr *pfx = NULL;
4753
4754 *peer_pfx = NULL;
4755
4756 if (addr)
4757 pfx = nla_data(addr);
4758
4759 if (local) {
4760 if (pfx && nla_memcmp(local, pfx, sizeof(*pfx)))
4761 *peer_pfx = pfx;
4762 pfx = nla_data(local);
4763 }
4764
4765 return pfx;
4766}
4767
4768static const struct nla_policy ifa_ipv6_policy[IFA_MAX+1] = {
4769 [IFA_ADDRESS] = { .len = sizeof(struct in6_addr) },
4770 [IFA_LOCAL] = { .len = sizeof(struct in6_addr) },
4771 [IFA_CACHEINFO] = { .len = sizeof(struct ifa_cacheinfo) },
4772 [IFA_FLAGS] = { .len = sizeof(u32) },
4773 [IFA_RT_PRIORITY] = { .len = sizeof(u32) },
4774 [IFA_TARGET_NETNSID] = { .type = NLA_S32 },
4775 [IFA_PROTO] = { .type = NLA_U8 },
4776};
4777
4778static int
4779inet6_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh,
4780 struct netlink_ext_ack *extack)
4781{
4782 struct net *net = sock_net(skb->sk);
4783 struct ifaddrmsg *ifm;
4784 struct nlattr *tb[IFA_MAX+1];
4785 struct in6_addr *pfx, *peer_pfx;
4786 u32 ifa_flags;
4787 int err;
4788
4789 err = nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFA_MAX,
4790 ifa_ipv6_policy, extack);
4791 if (err < 0)
4792 return err;
4793
4794 ifm = nlmsg_data(nlh);
4795 pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
4796 if (!pfx)
4797 return -EINVAL;
4798
4799 ifa_flags = nla_get_u32_default(tb[IFA_FLAGS], ifm->ifa_flags);
4800
4801 /* We ignore other flags so far. */
4802 ifa_flags &= IFA_F_MANAGETEMPADDR;
4803
4804 rtnl_net_lock(net);
4805 err = inet6_addr_del(net, ifm->ifa_index, ifa_flags, pfx,
4806 ifm->ifa_prefixlen, extack);
4807 rtnl_net_unlock(net);
4808
4809 return err;
4810}
4811
4812static int modify_prefix_route(struct net *net, struct inet6_ifaddr *ifp,
4813 unsigned long expires, u32 flags,
4814 bool modify_peer)
4815{
4816 struct fib6_table *table;
4817 struct fib6_info *f6i;
4818 u32 prio;
4819
4820 f6i = addrconf_get_prefix_route(modify_peer ? &ifp->peer_addr : &ifp->addr,
4821 ifp->prefix_len,
4822 ifp->idev->dev, 0, RTF_DEFAULT, true);
4823 if (!f6i)
4824 return -ENOENT;
4825
4826 prio = ifp->rt_priority ? : IP6_RT_PRIO_ADDRCONF;
4827 if (f6i->fib6_metric != prio) {
4828 /* delete old one */
4829 ip6_del_rt(dev_net(ifp->idev->dev), f6i, false);
4830
4831 /* add new one */
4832 addrconf_prefix_route(modify_peer ? &ifp->peer_addr : &ifp->addr,
4833 ifp->prefix_len,
4834 ifp->rt_priority, ifp->idev->dev,
4835 expires, flags, GFP_KERNEL);
4836 return 0;
4837 }
4838 if (f6i != net->ipv6.fib6_null_entry) {
4839 table = f6i->fib6_table;
4840 spin_lock_bh(&table->tb6_lock);
4841
4842 if (!(flags & RTF_EXPIRES)) {
4843 fib6_clean_expires(f6i);
4844 fib6_remove_gc_list(f6i);
4845 } else {
4846 fib6_set_expires(f6i, expires);
4847 fib6_add_gc_list(f6i);
4848 }
4849
4850 spin_unlock_bh(&table->tb6_lock);
4851 }
4852 fib6_info_release(f6i);
4853
4854 return 0;
4855}
4856
4857static int inet6_addr_modify(struct net *net, struct inet6_ifaddr *ifp,
4858 struct ifa6_config *cfg, clock_t expires,
4859 u32 flags)
4860{
4861 bool was_managetempaddr;
4862 bool new_peer = false;
4863 bool had_prefixroute;
4864
4865 ASSERT_RTNL_NET(net);
4866
4867 if (cfg->ifa_flags & IFA_F_MANAGETEMPADDR &&
4868 (ifp->flags & IFA_F_TEMPORARY || ifp->prefix_len != 64))
4869 return -EINVAL;
4870
4871 if (!(ifp->flags & IFA_F_TENTATIVE) || ifp->flags & IFA_F_DADFAILED)
4872 cfg->ifa_flags &= ~IFA_F_OPTIMISTIC;
4873
4874 if (cfg->peer_pfx &&
4875 memcmp(&ifp->peer_addr, cfg->peer_pfx, sizeof(struct in6_addr))) {
4876 if (!ipv6_addr_any(&ifp->peer_addr))
4877 cleanup_prefix_route(ifp, expires, true, true);
4878 new_peer = true;
4879 }
4880
4881 spin_lock_bh(&ifp->lock);
4882 was_managetempaddr = ifp->flags & IFA_F_MANAGETEMPADDR;
4883 had_prefixroute = ifp->flags & IFA_F_PERMANENT &&
4884 !(ifp->flags & IFA_F_NOPREFIXROUTE);
4885 ifp->flags &= ~(IFA_F_DEPRECATED | IFA_F_PERMANENT | IFA_F_NODAD |
4886 IFA_F_HOMEADDRESS | IFA_F_MANAGETEMPADDR |
4887 IFA_F_NOPREFIXROUTE);
4888 ifp->flags |= cfg->ifa_flags;
4889 WRITE_ONCE(ifp->tstamp, jiffies);
4890 WRITE_ONCE(ifp->valid_lft, cfg->valid_lft);
4891 WRITE_ONCE(ifp->prefered_lft, cfg->preferred_lft);
4892 WRITE_ONCE(ifp->ifa_proto, cfg->ifa_proto);
4893
4894 if (cfg->rt_priority && cfg->rt_priority != ifp->rt_priority)
4895 WRITE_ONCE(ifp->rt_priority, cfg->rt_priority);
4896
4897 if (new_peer)
4898 ifp->peer_addr = *cfg->peer_pfx;
4899
4900 spin_unlock_bh(&ifp->lock);
4901 if (!(ifp->flags&IFA_F_TENTATIVE))
4902 ipv6_ifa_notify(0, ifp);
4903
4904 if (!(cfg->ifa_flags & IFA_F_NOPREFIXROUTE)) {
4905 int rc = -ENOENT;
4906
4907 if (had_prefixroute)
4908 rc = modify_prefix_route(net, ifp, expires, flags, false);
4909
4910 /* prefix route could have been deleted; if so restore it */
4911 if (rc == -ENOENT) {
4912 addrconf_prefix_route(&ifp->addr, ifp->prefix_len,
4913 ifp->rt_priority, ifp->idev->dev,
4914 expires, flags, GFP_KERNEL);
4915 }
4916
4917 if (had_prefixroute && !ipv6_addr_any(&ifp->peer_addr))
4918 rc = modify_prefix_route(net, ifp, expires, flags, true);
4919
4920 if (rc == -ENOENT && !ipv6_addr_any(&ifp->peer_addr)) {
4921 addrconf_prefix_route(&ifp->peer_addr, ifp->prefix_len,
4922 ifp->rt_priority, ifp->idev->dev,
4923 expires, flags, GFP_KERNEL);
4924 }
4925 } else if (had_prefixroute) {
4926 enum cleanup_prefix_rt_t action;
4927 unsigned long rt_expires;
4928
4929 write_lock_bh(&ifp->idev->lock);
4930 action = check_cleanup_prefix_route(ifp, &rt_expires);
4931 write_unlock_bh(&ifp->idev->lock);
4932
4933 if (action != CLEANUP_PREFIX_RT_NOP) {
4934 cleanup_prefix_route(ifp, rt_expires,
4935 action == CLEANUP_PREFIX_RT_DEL, false);
4936 }
4937 }
4938
4939 if (was_managetempaddr || ifp->flags & IFA_F_MANAGETEMPADDR) {
4940 if (was_managetempaddr && !(ifp->flags & IFA_F_MANAGETEMPADDR))
4941 delete_tempaddrs(ifp->idev, ifp);
4942 else
4943 manage_tempaddrs(ifp->idev, ifp, cfg->valid_lft,
4944 cfg->preferred_lft, !was_managetempaddr,
4945 jiffies);
4946 }
4947
4948 addrconf_verify_rtnl(net);
4949
4950 return 0;
4951}
4952
4953static int
4954inet6_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh,
4955 struct netlink_ext_ack *extack)
4956{
4957 struct net *net = sock_net(skb->sk);
4958 struct nlattr *tb[IFA_MAX+1];
4959 struct in6_addr *peer_pfx;
4960 struct inet6_ifaddr *ifa;
4961 struct net_device *dev;
4962 struct inet6_dev *idev;
4963 struct ifa6_config cfg;
4964 struct ifaddrmsg *ifm;
4965 unsigned long timeout;
4966 clock_t expires;
4967 u32 flags;
4968 int err;
4969
4970 err = nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFA_MAX,
4971 ifa_ipv6_policy, extack);
4972 if (err < 0)
4973 return err;
4974
4975 memset(&cfg, 0, sizeof(cfg));
4976
4977 ifm = nlmsg_data(nlh);
4978 cfg.pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
4979 if (!cfg.pfx)
4980 return -EINVAL;
4981
4982 cfg.peer_pfx = peer_pfx;
4983 cfg.plen = ifm->ifa_prefixlen;
4984 if (tb[IFA_RT_PRIORITY])
4985 cfg.rt_priority = nla_get_u32(tb[IFA_RT_PRIORITY]);
4986
4987 if (tb[IFA_PROTO])
4988 cfg.ifa_proto = nla_get_u8(tb[IFA_PROTO]);
4989
4990 cfg.ifa_flags = nla_get_u32_default(tb[IFA_FLAGS], ifm->ifa_flags);
4991
4992 /* We ignore other flags so far. */
4993 cfg.ifa_flags &= IFA_F_NODAD | IFA_F_HOMEADDRESS |
4994 IFA_F_MANAGETEMPADDR | IFA_F_NOPREFIXROUTE |
4995 IFA_F_MCAUTOJOIN | IFA_F_OPTIMISTIC;
4996
4997 cfg.ifa_flags |= IFA_F_PERMANENT;
4998 cfg.valid_lft = INFINITY_LIFE_TIME;
4999 cfg.preferred_lft = INFINITY_LIFE_TIME;
5000 expires = 0;
5001 flags = 0;
5002
5003 if (tb[IFA_CACHEINFO]) {
5004 struct ifa_cacheinfo *ci;
5005
5006 ci = nla_data(tb[IFA_CACHEINFO]);
5007 cfg.valid_lft = ci->ifa_valid;
5008 cfg.preferred_lft = ci->ifa_prefered;
5009
5010 if (!cfg.valid_lft || cfg.preferred_lft > cfg.valid_lft) {
5011 NL_SET_ERR_MSG_MOD(extack, "address lifetime invalid");
5012 return -EINVAL;
5013 }
5014
5015 timeout = addrconf_timeout_fixup(cfg.valid_lft, HZ);
5016 if (addrconf_finite_timeout(timeout)) {
5017 cfg.ifa_flags &= ~IFA_F_PERMANENT;
5018 cfg.valid_lft = timeout;
5019 expires = jiffies_to_clock_t(timeout * HZ);
5020 flags = RTF_EXPIRES;
5021 }
5022
5023 timeout = addrconf_timeout_fixup(cfg.preferred_lft, HZ);
5024 if (addrconf_finite_timeout(timeout)) {
5025 if (timeout == 0)
5026 cfg.ifa_flags |= IFA_F_DEPRECATED;
5027
5028 cfg.preferred_lft = timeout;
5029 }
5030 }
5031
5032 rtnl_net_lock(net);
5033
5034 dev = __dev_get_by_index(net, ifm->ifa_index);
5035 if (!dev) {
5036 NL_SET_ERR_MSG_MOD(extack, "Unable to find the interface");
5037 err = -ENODEV;
5038 goto unlock_rtnl;
5039 }
5040
5041 netdev_lock_ops(dev);
5042 idev = ipv6_find_idev(dev);
5043 if (IS_ERR(idev)) {
5044 err = PTR_ERR(idev);
5045 goto unlock;
5046 }
5047
5048 if (!ipv6_allow_optimistic_dad(net, idev))
5049 cfg.ifa_flags &= ~IFA_F_OPTIMISTIC;
5050
5051 if (cfg.ifa_flags & IFA_F_NODAD &&
5052 cfg.ifa_flags & IFA_F_OPTIMISTIC) {
5053 NL_SET_ERR_MSG(extack, "IFA_F_NODAD and IFA_F_OPTIMISTIC are mutually exclusive");
5054 err = -EINVAL;
5055 goto unlock;
5056 }
5057
5058 ifa = ipv6_get_ifaddr(net, cfg.pfx, dev, 1);
5059 if (!ifa) {
5060 /*
5061 * It would be best to check for !NLM_F_CREATE here but
5062 * userspace already relies on not having to provide this.
5063 */
5064 err = inet6_addr_add(net, dev, &cfg, expires, flags, extack);
5065 goto unlock;
5066 }
5067
5068 if (nlh->nlmsg_flags & NLM_F_EXCL ||
5069 !(nlh->nlmsg_flags & NLM_F_REPLACE)) {
5070 NL_SET_ERR_MSG_MOD(extack, "address already assigned");
5071 err = -EEXIST;
5072 } else {
5073 err = inet6_addr_modify(net, ifa, &cfg, expires, flags);
5074 }
5075
5076 in6_ifa_put(ifa);
5077unlock:
5078 netdev_unlock_ops(dev);
5079unlock_rtnl:
5080 rtnl_net_unlock(net);
5081
5082 return err;
5083}
5084
5085static void put_ifaddrmsg(struct nlmsghdr *nlh, u8 prefixlen, u32 flags,
5086 u8 scope, int ifindex)
5087{
5088 struct ifaddrmsg *ifm;
5089
5090 ifm = nlmsg_data(nlh);
5091 ifm->ifa_family = AF_INET6;
5092 ifm->ifa_prefixlen = prefixlen;
5093 ifm->ifa_flags = flags;
5094 ifm->ifa_scope = scope;
5095 ifm->ifa_index = ifindex;
5096}
5097
5098static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
5099 unsigned long tstamp, u32 preferred, u32 valid)
5100{
5101 struct ifa_cacheinfo ci;
5102
5103 ci.cstamp = cstamp_delta(cstamp);
5104 ci.tstamp = cstamp_delta(tstamp);
5105 ci.ifa_prefered = preferred;
5106 ci.ifa_valid = valid;
5107
5108 return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
5109}
5110
5111static inline int rt_scope(int ifa_scope)
5112{
5113 if (ifa_scope & IFA_HOST)
5114 return RT_SCOPE_HOST;
5115 else if (ifa_scope & IFA_LINK)
5116 return RT_SCOPE_LINK;
5117 else if (ifa_scope & IFA_SITE)
5118 return RT_SCOPE_SITE;
5119 else
5120 return RT_SCOPE_UNIVERSE;
5121}
5122
5123static inline int inet6_ifaddr_msgsize(void)
5124{
5125 return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
5126 + nla_total_size(16) /* IFA_LOCAL */
5127 + nla_total_size(16) /* IFA_ADDRESS */
5128 + nla_total_size(sizeof(struct ifa_cacheinfo))
5129 + nla_total_size(4) /* IFA_FLAGS */
5130 + nla_total_size(1) /* IFA_PROTO */
5131 + nla_total_size(4) /* IFA_RT_PRIORITY */;
5132}
5133
5134static int inet6_fill_ifaddr(struct sk_buff *skb,
5135 const struct inet6_ifaddr *ifa,
5136 struct inet6_fill_args *args)
5137{
5138 struct nlmsghdr *nlh;
5139 u32 preferred, valid;
5140 u32 flags, priority;
5141 u8 proto;
5142
5143 nlh = nlmsg_put(skb, args->portid, args->seq, args->event,
5144 sizeof(struct ifaddrmsg), args->flags);
5145 if (!nlh)
5146 return -EMSGSIZE;
5147
5148 flags = READ_ONCE(ifa->flags);
5149 put_ifaddrmsg(nlh, ifa->prefix_len, ifa->flags, rt_scope(ifa->scope),
5150 ifa->idev->dev->ifindex);
5151
5152 if (args->netnsid >= 0 &&
5153 nla_put_s32(skb, IFA_TARGET_NETNSID, args->netnsid))
5154 goto error;
5155
5156 preferred = READ_ONCE(ifa->prefered_lft);
5157 valid = READ_ONCE(ifa->valid_lft);
5158
5159 if (!((flags & IFA_F_PERMANENT) &&
5160 (preferred == INFINITY_LIFE_TIME))) {
5161 if (preferred != INFINITY_LIFE_TIME) {
5162 long tval = (jiffies - READ_ONCE(ifa->tstamp)) / HZ;
5163
5164 if (preferred > tval)
5165 preferred -= tval;
5166 else
5167 preferred = 0;
5168 if (valid != INFINITY_LIFE_TIME) {
5169 if (valid > tval)
5170 valid -= tval;
5171 else
5172 valid = 0;
5173 }
5174 }
5175 } else {
5176 preferred = INFINITY_LIFE_TIME;
5177 valid = INFINITY_LIFE_TIME;
5178 }
5179
5180 if (!ipv6_addr_any(&ifa->peer_addr)) {
5181 if (nla_put_in6_addr(skb, IFA_LOCAL, &ifa->addr) < 0 ||
5182 nla_put_in6_addr(skb, IFA_ADDRESS, &ifa->peer_addr) < 0)
5183 goto error;
5184 } else {
5185 if (nla_put_in6_addr(skb, IFA_ADDRESS, &ifa->addr) < 0)
5186 goto error;
5187 }
5188
5189 priority = READ_ONCE(ifa->rt_priority);
5190 if (priority && nla_put_u32(skb, IFA_RT_PRIORITY, priority))
5191 goto error;
5192
5193 if (put_cacheinfo(skb, ifa->cstamp, READ_ONCE(ifa->tstamp),
5194 preferred, valid) < 0)
5195 goto error;
5196
5197 if (nla_put_u32(skb, IFA_FLAGS, flags) < 0)
5198 goto error;
5199
5200 proto = READ_ONCE(ifa->ifa_proto);
5201 if (proto && nla_put_u8(skb, IFA_PROTO, proto))
5202 goto error;
5203
5204 nlmsg_end(skb, nlh);
5205 return 0;
5206
5207error:
5208 nlmsg_cancel(skb, nlh);
5209 return -EMSGSIZE;
5210}
5211
5212int inet6_fill_ifmcaddr(struct sk_buff *skb,
5213 const struct ifmcaddr6 *ifmca,
5214 struct inet6_fill_args *args)
5215{
5216 int ifindex = ifmca->idev->dev->ifindex;
5217 u8 scope = RT_SCOPE_UNIVERSE;
5218 struct nlmsghdr *nlh;
5219
5220 if (!args->force_rt_scope_universe &&
5221 ipv6_addr_scope(&ifmca->mca_addr) & IFA_SITE)
5222 scope = RT_SCOPE_SITE;
5223
5224 nlh = nlmsg_put(skb, args->portid, args->seq, args->event,
5225 sizeof(struct ifaddrmsg), args->flags);
5226 if (!nlh)
5227 return -EMSGSIZE;
5228
5229 if (args->netnsid >= 0 &&
5230 nla_put_s32(skb, IFA_TARGET_NETNSID, args->netnsid)) {
5231 nlmsg_cancel(skb, nlh);
5232 return -EMSGSIZE;
5233 }
5234
5235 put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
5236 if (nla_put_in6_addr(skb, IFA_MULTICAST, &ifmca->mca_addr) < 0 ||
5237 put_cacheinfo(skb, ifmca->mca_cstamp, READ_ONCE(ifmca->mca_tstamp),
5238 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
5239 nlmsg_cancel(skb, nlh);
5240 return -EMSGSIZE;
5241 }
5242
5243 nlmsg_end(skb, nlh);
5244 return 0;
5245}
5246
5247int inet6_fill_ifacaddr(struct sk_buff *skb,
5248 const struct ifacaddr6 *ifaca,
5249 struct inet6_fill_args *args)
5250{
5251 struct net_device *dev = fib6_info_nh_dev(ifaca->aca_rt);
5252 int ifindex = dev ? dev->ifindex : 1;
5253 u8 scope = RT_SCOPE_UNIVERSE;
5254 struct nlmsghdr *nlh;
5255
5256 if (ipv6_addr_scope(&ifaca->aca_addr) & IFA_SITE)
5257 scope = RT_SCOPE_SITE;
5258
5259 nlh = nlmsg_put(skb, args->portid, args->seq, args->event,
5260 sizeof(struct ifaddrmsg), args->flags);
5261 if (!nlh)
5262 return -EMSGSIZE;
5263
5264 if (args->netnsid >= 0 &&
5265 nla_put_s32(skb, IFA_TARGET_NETNSID, args->netnsid)) {
5266 nlmsg_cancel(skb, nlh);
5267 return -EMSGSIZE;
5268 }
5269
5270 put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
5271 if (nla_put_in6_addr(skb, IFA_ANYCAST, &ifaca->aca_addr) < 0 ||
5272 put_cacheinfo(skb, ifaca->aca_cstamp, READ_ONCE(ifaca->aca_tstamp),
5273 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
5274 nlmsg_cancel(skb, nlh);
5275 return -EMSGSIZE;
5276 }
5277
5278 nlmsg_end(skb, nlh);
5279 return 0;
5280}
5281
5282/* called with rcu_read_lock() */
5283static int in6_dump_addrs(const struct inet6_dev *idev, struct sk_buff *skb,
5284 struct netlink_callback *cb, int *s_ip_idx,
5285 struct inet6_fill_args *fillargs)
5286{
5287 const struct ifmcaddr6 *ifmca;
5288 const struct ifacaddr6 *ifaca;
5289 int ip_idx = 0;
5290 int err = 0;
5291
5292 switch (fillargs->type) {
5293 case UNICAST_ADDR: {
5294 const struct inet6_ifaddr *ifa;
5295 fillargs->event = RTM_NEWADDR;
5296
5297 /* unicast address incl. temp addr */
5298 list_for_each_entry_rcu(ifa, &idev->addr_list, if_list) {
5299 if (ip_idx < *s_ip_idx)
5300 goto next;
5301 err = inet6_fill_ifaddr(skb, ifa, fillargs);
5302 if (err < 0)
5303 break;
5304 nl_dump_check_consistent(cb, nlmsg_hdr(skb));
5305next:
5306 ip_idx++;
5307 }
5308 break;
5309 }
5310 case MULTICAST_ADDR:
5311 fillargs->event = RTM_GETMULTICAST;
5312
5313 /* multicast address */
5314 for (ifmca = rcu_dereference(idev->mc_list);
5315 ifmca;
5316 ifmca = rcu_dereference(ifmca->next), ip_idx++) {
5317 if (ip_idx < *s_ip_idx)
5318 continue;
5319 err = inet6_fill_ifmcaddr(skb, ifmca, fillargs);
5320 if (err < 0)
5321 break;
5322 }
5323 break;
5324 case ANYCAST_ADDR:
5325 fillargs->event = RTM_GETANYCAST;
5326 /* anycast address */
5327 for (ifaca = rcu_dereference(idev->ac_list); ifaca;
5328 ifaca = rcu_dereference(ifaca->aca_next), ip_idx++) {
5329 if (ip_idx < *s_ip_idx)
5330 continue;
5331 err = inet6_fill_ifacaddr(skb, ifaca, fillargs);
5332 if (err < 0)
5333 break;
5334 }
5335 break;
5336 default:
5337 break;
5338 }
5339 *s_ip_idx = err ? ip_idx : 0;
5340 return err;
5341}
5342
5343static int inet6_valid_dump_ifaddr_req(const struct nlmsghdr *nlh,
5344 struct inet6_fill_args *fillargs,
5345 struct net **tgt_net, struct sock *sk,
5346 struct netlink_callback *cb)
5347{
5348 struct netlink_ext_ack *extack = cb->extack;
5349 struct nlattr *tb[IFA_MAX+1];
5350 struct ifaddrmsg *ifm;
5351 int err, i;
5352
5353 ifm = nlmsg_payload(nlh, sizeof(*ifm));
5354 if (!ifm) {
5355 NL_SET_ERR_MSG_MOD(extack, "Invalid header for address dump request");
5356 return -EINVAL;
5357 }
5358
5359 if (ifm->ifa_prefixlen || ifm->ifa_flags || ifm->ifa_scope) {
5360 NL_SET_ERR_MSG_MOD(extack, "Invalid values in header for address dump request");
5361 return -EINVAL;
5362 }
5363
5364 fillargs->ifindex = ifm->ifa_index;
5365 if (fillargs->ifindex) {
5366 cb->answer_flags |= NLM_F_DUMP_FILTERED;
5367 fillargs->flags |= NLM_F_DUMP_FILTERED;
5368 }
5369
5370 err = nlmsg_parse_deprecated_strict(nlh, sizeof(*ifm), tb, IFA_MAX,
5371 ifa_ipv6_policy, extack);
5372 if (err < 0)
5373 return err;
5374
5375 for (i = 0; i <= IFA_MAX; ++i) {
5376 if (!tb[i])
5377 continue;
5378
5379 if (i == IFA_TARGET_NETNSID) {
5380 struct net *net;
5381
5382 fillargs->netnsid = nla_get_s32(tb[i]);
5383 net = rtnl_get_net_ns_capable(sk, fillargs->netnsid);
5384 if (IS_ERR(net)) {
5385 fillargs->netnsid = -1;
5386 NL_SET_ERR_MSG_MOD(extack, "Invalid target network namespace id");
5387 return PTR_ERR(net);
5388 }
5389 *tgt_net = net;
5390 } else {
5391 NL_SET_ERR_MSG_MOD(extack, "Unsupported attribute in dump request");
5392 return -EINVAL;
5393 }
5394 }
5395
5396 return 0;
5397}
5398
5399static int inet6_dump_addr(struct sk_buff *skb, struct netlink_callback *cb,
5400 enum addr_type_t type)
5401{
5402 struct net *tgt_net = sock_net(skb->sk);
5403 const struct nlmsghdr *nlh = cb->nlh;
5404 struct inet6_fill_args fillargs = {
5405 .portid = NETLINK_CB(cb->skb).portid,
5406 .seq = cb->nlh->nlmsg_seq,
5407 .flags = NLM_F_MULTI,
5408 .netnsid = -1,
5409 .type = type,
5410 .force_rt_scope_universe = false,
5411 };
5412 struct {
5413 unsigned long ifindex;
5414 int ip_idx;
5415 } *ctx = (void *)cb->ctx;
5416 struct net_device *dev;
5417 struct inet6_dev *idev;
5418 int err = 0;
5419
5420 rcu_read_lock();
5421 if (cb->strict_check) {
5422 err = inet6_valid_dump_ifaddr_req(nlh, &fillargs, &tgt_net,
5423 skb->sk, cb);
5424 if (err < 0)
5425 goto done;
5426
5427 err = 0;
5428 if (fillargs.ifindex) {
5429 dev = dev_get_by_index_rcu(tgt_net, fillargs.ifindex);
5430 if (!dev) {
5431 err = -ENODEV;
5432 goto done;
5433 }
5434 idev = __in6_dev_get(dev);
5435 if (idev)
5436 err = in6_dump_addrs(idev, skb, cb,
5437 &ctx->ip_idx,
5438 &fillargs);
5439 goto done;
5440 }
5441 }
5442
5443 cb->seq = inet6_base_seq(tgt_net);
5444 for_each_netdev_dump(tgt_net, dev, ctx->ifindex) {
5445 idev = __in6_dev_get(dev);
5446 if (!idev)
5447 continue;
5448 err = in6_dump_addrs(idev, skb, cb, &ctx->ip_idx,
5449 &fillargs);
5450 if (err < 0)
5451 goto done;
5452 }
5453done:
5454 rcu_read_unlock();
5455 if (fillargs.netnsid >= 0)
5456 put_net(tgt_net);
5457
5458 return err;
5459}
5460
5461static int inet6_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
5462{
5463 enum addr_type_t type = UNICAST_ADDR;
5464
5465 return inet6_dump_addr(skb, cb, type);
5466}
5467
5468static int inet6_dump_ifmcaddr(struct sk_buff *skb, struct netlink_callback *cb)
5469{
5470 enum addr_type_t type = MULTICAST_ADDR;
5471
5472 return inet6_dump_addr(skb, cb, type);
5473}
5474
5475
5476static int inet6_dump_ifacaddr(struct sk_buff *skb, struct netlink_callback *cb)
5477{
5478 enum addr_type_t type = ANYCAST_ADDR;
5479
5480 return inet6_dump_addr(skb, cb, type);
5481}
5482
5483static int inet6_rtm_valid_getaddr_req(struct sk_buff *skb,
5484 const struct nlmsghdr *nlh,
5485 struct nlattr **tb,
5486 struct netlink_ext_ack *extack)
5487{
5488 struct ifaddrmsg *ifm;
5489 int i, err;
5490
5491 ifm = nlmsg_payload(nlh, sizeof(*ifm));
5492 if (!ifm) {
5493 NL_SET_ERR_MSG_MOD(extack, "Invalid header for get address request");
5494 return -EINVAL;
5495 }
5496
5497 if (!netlink_strict_get_check(skb))
5498 return nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFA_MAX,
5499 ifa_ipv6_policy, extack);
5500
5501 if (ifm->ifa_prefixlen || ifm->ifa_flags || ifm->ifa_scope) {
5502 NL_SET_ERR_MSG_MOD(extack, "Invalid values in header for get address request");
5503 return -EINVAL;
5504 }
5505
5506 err = nlmsg_parse_deprecated_strict(nlh, sizeof(*ifm), tb, IFA_MAX,
5507 ifa_ipv6_policy, extack);
5508 if (err)
5509 return err;
5510
5511 for (i = 0; i <= IFA_MAX; i++) {
5512 if (!tb[i])
5513 continue;
5514
5515 switch (i) {
5516 case IFA_TARGET_NETNSID:
5517 case IFA_ADDRESS:
5518 case IFA_LOCAL:
5519 break;
5520 default:
5521 NL_SET_ERR_MSG_MOD(extack, "Unsupported attribute in get address request");
5522 return -EINVAL;
5523 }
5524 }
5525
5526 return 0;
5527}
5528
5529static int inet6_rtm_getaddr(struct sk_buff *in_skb, struct nlmsghdr *nlh,
5530 struct netlink_ext_ack *extack)
5531{
5532 struct net *tgt_net = sock_net(in_skb->sk);
5533 struct inet6_fill_args fillargs = {
5534 .portid = NETLINK_CB(in_skb).portid,
5535 .seq = nlh->nlmsg_seq,
5536 .event = RTM_NEWADDR,
5537 .flags = 0,
5538 .netnsid = -1,
5539 .force_rt_scope_universe = false,
5540 };
5541 struct ifaddrmsg *ifm;
5542 struct nlattr *tb[IFA_MAX+1];
5543 struct in6_addr *addr = NULL, *peer;
5544 struct net_device *dev = NULL;
5545 struct inet6_ifaddr *ifa;
5546 struct sk_buff *skb;
5547 int err;
5548
5549 err = inet6_rtm_valid_getaddr_req(in_skb, nlh, tb, extack);
5550 if (err < 0)
5551 return err;
5552
5553 if (tb[IFA_TARGET_NETNSID]) {
5554 fillargs.netnsid = nla_get_s32(tb[IFA_TARGET_NETNSID]);
5555
5556 tgt_net = rtnl_get_net_ns_capable(NETLINK_CB(in_skb).sk,
5557 fillargs.netnsid);
5558 if (IS_ERR(tgt_net))
5559 return PTR_ERR(tgt_net);
5560 }
5561
5562 addr = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer);
5563 if (!addr) {
5564 err = -EINVAL;
5565 goto errout;
5566 }
5567 ifm = nlmsg_data(nlh);
5568 if (ifm->ifa_index)
5569 dev = dev_get_by_index(tgt_net, ifm->ifa_index);
5570
5571 ifa = ipv6_get_ifaddr(tgt_net, addr, dev, 1);
5572 if (!ifa) {
5573 err = -EADDRNOTAVAIL;
5574 goto errout;
5575 }
5576
5577 skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_KERNEL);
5578 if (!skb) {
5579 err = -ENOBUFS;
5580 goto errout_ifa;
5581 }
5582
5583 err = inet6_fill_ifaddr(skb, ifa, &fillargs);
5584 if (err < 0) {
5585 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
5586 WARN_ON(err == -EMSGSIZE);
5587 kfree_skb(skb);
5588 goto errout_ifa;
5589 }
5590 err = rtnl_unicast(skb, tgt_net, NETLINK_CB(in_skb).portid);
5591errout_ifa:
5592 in6_ifa_put(ifa);
5593errout:
5594 dev_put(dev);
5595 if (fillargs.netnsid >= 0)
5596 put_net(tgt_net);
5597
5598 return err;
5599}
5600
5601static void inet6_ifa_notify(int event, struct inet6_ifaddr *ifa)
5602{
5603 struct sk_buff *skb;
5604 struct net *net = dev_net(ifa->idev->dev);
5605 struct inet6_fill_args fillargs = {
5606 .portid = 0,
5607 .seq = 0,
5608 .event = event,
5609 .flags = 0,
5610 .netnsid = -1,
5611 .force_rt_scope_universe = false,
5612 };
5613 int err = -ENOBUFS;
5614
5615 skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_ATOMIC);
5616 if (!skb)
5617 goto errout;
5618
5619 err = inet6_fill_ifaddr(skb, ifa, &fillargs);
5620 if (err < 0) {
5621 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
5622 WARN_ON(err == -EMSGSIZE);
5623 kfree_skb(skb);
5624 goto errout;
5625 }
5626 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
5627 return;
5628errout:
5629 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
5630}
5631
5632static void ipv6_store_devconf(const struct ipv6_devconf *cnf,
5633 __s32 *array, int bytes)
5634{
5635 BUG_ON(bytes < (DEVCONF_MAX * 4));
5636
5637 memset(array, 0, bytes);
5638 array[DEVCONF_FORWARDING] = READ_ONCE(cnf->forwarding);
5639 array[DEVCONF_HOPLIMIT] = READ_ONCE(cnf->hop_limit);
5640 array[DEVCONF_MTU6] = READ_ONCE(cnf->mtu6);
5641 array[DEVCONF_ACCEPT_RA] = READ_ONCE(cnf->accept_ra);
5642 array[DEVCONF_ACCEPT_REDIRECTS] = READ_ONCE(cnf->accept_redirects);
5643 array[DEVCONF_AUTOCONF] = READ_ONCE(cnf->autoconf);
5644 array[DEVCONF_DAD_TRANSMITS] = READ_ONCE(cnf->dad_transmits);
5645 array[DEVCONF_RTR_SOLICITS] = READ_ONCE(cnf->rtr_solicits);
5646 array[DEVCONF_RTR_SOLICIT_INTERVAL] =
5647 jiffies_to_msecs(READ_ONCE(cnf->rtr_solicit_interval));
5648 array[DEVCONF_RTR_SOLICIT_MAX_INTERVAL] =
5649 jiffies_to_msecs(READ_ONCE(cnf->rtr_solicit_max_interval));
5650 array[DEVCONF_RTR_SOLICIT_DELAY] =
5651 jiffies_to_msecs(READ_ONCE(cnf->rtr_solicit_delay));
5652 array[DEVCONF_FORCE_MLD_VERSION] = READ_ONCE(cnf->force_mld_version);
5653 array[DEVCONF_MLDV1_UNSOLICITED_REPORT_INTERVAL] =
5654 jiffies_to_msecs(READ_ONCE(cnf->mldv1_unsolicited_report_interval));
5655 array[DEVCONF_MLDV2_UNSOLICITED_REPORT_INTERVAL] =
5656 jiffies_to_msecs(READ_ONCE(cnf->mldv2_unsolicited_report_interval));
5657 array[DEVCONF_USE_TEMPADDR] = READ_ONCE(cnf->use_tempaddr);
5658 array[DEVCONF_TEMP_VALID_LFT] = READ_ONCE(cnf->temp_valid_lft);
5659 array[DEVCONF_TEMP_PREFERED_LFT] = READ_ONCE(cnf->temp_prefered_lft);
5660 array[DEVCONF_REGEN_MAX_RETRY] = READ_ONCE(cnf->regen_max_retry);
5661 array[DEVCONF_MAX_DESYNC_FACTOR] = READ_ONCE(cnf->max_desync_factor);
5662 array[DEVCONF_MAX_ADDRESSES] = READ_ONCE(cnf->max_addresses);
5663 array[DEVCONF_ACCEPT_RA_DEFRTR] = READ_ONCE(cnf->accept_ra_defrtr);
5664 array[DEVCONF_RA_DEFRTR_METRIC] = READ_ONCE(cnf->ra_defrtr_metric);
5665 array[DEVCONF_ACCEPT_RA_MIN_HOP_LIMIT] =
5666 READ_ONCE(cnf->accept_ra_min_hop_limit);
5667 array[DEVCONF_ACCEPT_RA_PINFO] = READ_ONCE(cnf->accept_ra_pinfo);
5668#ifdef CONFIG_IPV6_ROUTER_PREF
5669 array[DEVCONF_ACCEPT_RA_RTR_PREF] = READ_ONCE(cnf->accept_ra_rtr_pref);
5670 array[DEVCONF_RTR_PROBE_INTERVAL] =
5671 jiffies_to_msecs(READ_ONCE(cnf->rtr_probe_interval));
5672#ifdef CONFIG_IPV6_ROUTE_INFO
5673 array[DEVCONF_ACCEPT_RA_RT_INFO_MIN_PLEN] =
5674 READ_ONCE(cnf->accept_ra_rt_info_min_plen);
5675 array[DEVCONF_ACCEPT_RA_RT_INFO_MAX_PLEN] =
5676 READ_ONCE(cnf->accept_ra_rt_info_max_plen);
5677#endif
5678#endif
5679 array[DEVCONF_PROXY_NDP] = READ_ONCE(cnf->proxy_ndp);
5680 array[DEVCONF_ACCEPT_SOURCE_ROUTE] =
5681 READ_ONCE(cnf->accept_source_route);
5682#ifdef CONFIG_IPV6_OPTIMISTIC_DAD
5683 array[DEVCONF_OPTIMISTIC_DAD] = READ_ONCE(cnf->optimistic_dad);
5684 array[DEVCONF_USE_OPTIMISTIC] = READ_ONCE(cnf->use_optimistic);
5685#endif
5686#ifdef CONFIG_IPV6_MROUTE
5687 array[DEVCONF_MC_FORWARDING] = atomic_read(&cnf->mc_forwarding);
5688#endif
5689 array[DEVCONF_DISABLE_IPV6] = READ_ONCE(cnf->disable_ipv6);
5690 array[DEVCONF_ACCEPT_DAD] = READ_ONCE(cnf->accept_dad);
5691 array[DEVCONF_FORCE_TLLAO] = READ_ONCE(cnf->force_tllao);
5692 array[DEVCONF_NDISC_NOTIFY] = READ_ONCE(cnf->ndisc_notify);
5693 array[DEVCONF_SUPPRESS_FRAG_NDISC] =
5694 READ_ONCE(cnf->suppress_frag_ndisc);
5695 array[DEVCONF_ACCEPT_RA_FROM_LOCAL] =
5696 READ_ONCE(cnf->accept_ra_from_local);
5697 array[DEVCONF_ACCEPT_RA_MTU] = READ_ONCE(cnf->accept_ra_mtu);
5698 array[DEVCONF_IGNORE_ROUTES_WITH_LINKDOWN] =
5699 READ_ONCE(cnf->ignore_routes_with_linkdown);
5700 /* we omit DEVCONF_STABLE_SECRET for now */
5701 array[DEVCONF_USE_OIF_ADDRS_ONLY] = READ_ONCE(cnf->use_oif_addrs_only);
5702 array[DEVCONF_DROP_UNICAST_IN_L2_MULTICAST] =
5703 READ_ONCE(cnf->drop_unicast_in_l2_multicast);
5704 array[DEVCONF_DROP_UNSOLICITED_NA] = READ_ONCE(cnf->drop_unsolicited_na);
5705 array[DEVCONF_KEEP_ADDR_ON_DOWN] = READ_ONCE(cnf->keep_addr_on_down);
5706 array[DEVCONF_SEG6_ENABLED] = READ_ONCE(cnf->seg6_enabled);
5707#ifdef CONFIG_IPV6_SEG6_HMAC
5708 array[DEVCONF_SEG6_REQUIRE_HMAC] = READ_ONCE(cnf->seg6_require_hmac);
5709#endif
5710 array[DEVCONF_ENHANCED_DAD] = READ_ONCE(cnf->enhanced_dad);
5711 array[DEVCONF_ADDR_GEN_MODE] = READ_ONCE(cnf->addr_gen_mode);
5712 array[DEVCONF_DISABLE_POLICY] = READ_ONCE(cnf->disable_policy);
5713 array[DEVCONF_NDISC_TCLASS] = READ_ONCE(cnf->ndisc_tclass);
5714 array[DEVCONF_RPL_SEG_ENABLED] = READ_ONCE(cnf->rpl_seg_enabled);
5715 array[DEVCONF_IOAM6_ENABLED] = READ_ONCE(cnf->ioam6_enabled);
5716 array[DEVCONF_IOAM6_ID] = READ_ONCE(cnf->ioam6_id);
5717 array[DEVCONF_IOAM6_ID_WIDE] = READ_ONCE(cnf->ioam6_id_wide);
5718 array[DEVCONF_NDISC_EVICT_NOCARRIER] =
5719 READ_ONCE(cnf->ndisc_evict_nocarrier);
5720 array[DEVCONF_ACCEPT_UNTRACKED_NA] =
5721 READ_ONCE(cnf->accept_untracked_na);
5722 array[DEVCONF_ACCEPT_RA_MIN_LFT] = READ_ONCE(cnf->accept_ra_min_lft);
5723 array[DEVCONF_FORCE_FORWARDING] = READ_ONCE(cnf->force_forwarding);
5724}
5725
5726static inline size_t inet6_ifla6_size(void)
5727{
5728 return nla_total_size(4) /* IFLA_INET6_FLAGS */
5729 + nla_total_size(sizeof(struct ifla_cacheinfo))
5730 + nla_total_size(DEVCONF_MAX * 4) /* IFLA_INET6_CONF */
5731 + nla_total_size(IPSTATS_MIB_MAX * 8) /* IFLA_INET6_STATS */
5732 + nla_total_size(ICMP6_MIB_MAX * 8) /* IFLA_INET6_ICMP6STATS */
5733 + nla_total_size(sizeof(struct in6_addr)) /* IFLA_INET6_TOKEN */
5734 + nla_total_size(1) /* IFLA_INET6_ADDR_GEN_MODE */
5735 + nla_total_size(4) /* IFLA_INET6_RA_MTU */
5736 + 0;
5737}
5738
5739static inline size_t inet6_if_nlmsg_size(void)
5740{
5741 return NLMSG_ALIGN(sizeof(struct ifinfomsg))
5742 + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
5743 + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
5744 + nla_total_size(4) /* IFLA_MTU */
5745 + nla_total_size(4) /* IFLA_LINK */
5746 + nla_total_size(1) /* IFLA_OPERSTATE */
5747 + nla_total_size(inet6_ifla6_size()); /* IFLA_PROTINFO */
5748}
5749
5750static inline void __snmp6_fill_statsdev(u64 *stats, atomic_long_t *mib,
5751 int bytes)
5752{
5753 int i;
5754 int pad = bytes - sizeof(u64) * ICMP6_MIB_MAX;
5755 BUG_ON(pad < 0);
5756
5757 /* Use put_unaligned() because stats may not be aligned for u64. */
5758 put_unaligned(ICMP6_MIB_MAX, &stats[0]);
5759 for (i = 1; i < ICMP6_MIB_MAX; i++)
5760 put_unaligned(atomic_long_read(&mib[i]), &stats[i]);
5761
5762 memset(&stats[ICMP6_MIB_MAX], 0, pad);
5763}
5764
5765static inline void __snmp6_fill_stats64(u64 *stats, void __percpu *mib,
5766 int bytes, size_t syncpoff)
5767{
5768 int i, c;
5769 u64 buff[IPSTATS_MIB_MAX];
5770 int pad = bytes - sizeof(u64) * IPSTATS_MIB_MAX;
5771
5772 BUG_ON(pad < 0);
5773
5774 memset(buff, 0, sizeof(buff));
5775 buff[0] = IPSTATS_MIB_MAX;
5776
5777 for_each_possible_cpu(c) {
5778 for (i = 1; i < IPSTATS_MIB_MAX; i++)
5779 buff[i] += snmp_get_cpu_field64(mib, c, i, syncpoff);
5780 }
5781
5782 memcpy(stats, buff, IPSTATS_MIB_MAX * sizeof(u64));
5783 memset(&stats[IPSTATS_MIB_MAX], 0, pad);
5784}
5785
5786static void snmp6_fill_stats(u64 *stats, struct inet6_dev *idev, int attrtype,
5787 int bytes)
5788{
5789 switch (attrtype) {
5790 case IFLA_INET6_STATS:
5791 __snmp6_fill_stats64(stats, idev->stats.ipv6, bytes,
5792 offsetof(struct ipstats_mib, syncp));
5793 break;
5794 case IFLA_INET6_ICMP6STATS:
5795 __snmp6_fill_statsdev(stats, idev->stats.icmpv6dev->mibs, bytes);
5796 break;
5797 }
5798}
5799
5800static int inet6_fill_ifla6_stats_attrs(struct sk_buff *skb,
5801 struct inet6_dev *idev)
5802{
5803 struct nlattr *nla;
5804
5805 nla = nla_reserve(skb, IFLA_INET6_STATS, IPSTATS_MIB_MAX * sizeof(u64));
5806 if (!nla)
5807 goto nla_put_failure;
5808 snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_STATS, nla_len(nla));
5809
5810 nla = nla_reserve(skb, IFLA_INET6_ICMP6STATS, ICMP6_MIB_MAX * sizeof(u64));
5811 if (!nla)
5812 goto nla_put_failure;
5813 snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_ICMP6STATS, nla_len(nla));
5814
5815 return 0;
5816
5817nla_put_failure:
5818 return -EMSGSIZE;
5819}
5820
5821static int inet6_fill_ifla6_attrs(struct sk_buff *skb, struct inet6_dev *idev,
5822 u32 ext_filter_mask)
5823{
5824 struct ifla_cacheinfo ci;
5825 struct nlattr *nla;
5826 u32 ra_mtu;
5827
5828 if (nla_put_u32(skb, IFLA_INET6_FLAGS, READ_ONCE(idev->if_flags)))
5829 goto nla_put_failure;
5830 ci.max_reasm_len = IPV6_MAXPLEN;
5831 ci.tstamp = cstamp_delta(READ_ONCE(idev->tstamp));
5832 ci.reachable_time = jiffies_to_msecs(idev->nd_parms->reachable_time);
5833 ci.retrans_time = jiffies_to_msecs(NEIGH_VAR(idev->nd_parms, RETRANS_TIME));
5834 if (nla_put(skb, IFLA_INET6_CACHEINFO, sizeof(ci), &ci))
5835 goto nla_put_failure;
5836 nla = nla_reserve(skb, IFLA_INET6_CONF, DEVCONF_MAX * sizeof(s32));
5837 if (!nla)
5838 goto nla_put_failure;
5839 ipv6_store_devconf(&idev->cnf, nla_data(nla), nla_len(nla));
5840
5841 /* XXX - MC not implemented */
5842
5843 if (!(ext_filter_mask & RTEXT_FILTER_SKIP_STATS)) {
5844 if (inet6_fill_ifla6_stats_attrs(skb, idev) < 0)
5845 goto nla_put_failure;
5846 }
5847
5848 nla = nla_reserve(skb, IFLA_INET6_TOKEN, sizeof(struct in6_addr));
5849 if (!nla)
5850 goto nla_put_failure;
5851 read_lock_bh(&idev->lock);
5852 memcpy(nla_data(nla), idev->token.s6_addr, nla_len(nla));
5853 read_unlock_bh(&idev->lock);
5854
5855 if (nla_put_u8(skb, IFLA_INET6_ADDR_GEN_MODE,
5856 READ_ONCE(idev->cnf.addr_gen_mode)))
5857 goto nla_put_failure;
5858
5859 ra_mtu = READ_ONCE(idev->ra_mtu);
5860 if (ra_mtu && nla_put_u32(skb, IFLA_INET6_RA_MTU, ra_mtu))
5861 goto nla_put_failure;
5862
5863 return 0;
5864
5865nla_put_failure:
5866 return -EMSGSIZE;
5867}
5868
5869static size_t inet6_get_link_af_size(const struct net_device *dev,
5870 u32 ext_filter_mask)
5871{
5872 if (!__in6_dev_get(dev))
5873 return 0;
5874
5875 return inet6_ifla6_size();
5876}
5877
5878static int inet6_fill_link_af(struct sk_buff *skb, const struct net_device *dev,
5879 u32 ext_filter_mask)
5880{
5881 struct inet6_dev *idev = __in6_dev_get(dev);
5882
5883 if (!idev)
5884 return -ENODATA;
5885
5886 if (inet6_fill_ifla6_attrs(skb, idev, ext_filter_mask) < 0)
5887 return -EMSGSIZE;
5888
5889 return 0;
5890}
5891
5892static int inet6_set_iftoken(struct inet6_dev *idev, struct in6_addr *token,
5893 struct netlink_ext_ack *extack)
5894{
5895 struct inet6_ifaddr *ifp;
5896 struct net_device *dev = idev->dev;
5897 bool clear_token, update_rs = false;
5898 struct in6_addr ll_addr;
5899
5900 ASSERT_RTNL();
5901
5902 if (!token)
5903 return -EINVAL;
5904
5905 if (dev->flags & IFF_LOOPBACK) {
5906 NL_SET_ERR_MSG_MOD(extack, "Device is loopback");
5907 return -EINVAL;
5908 }
5909
5910 if (dev->flags & IFF_NOARP) {
5911 NL_SET_ERR_MSG_MOD(extack,
5912 "Device does not do neighbour discovery");
5913 return -EINVAL;
5914 }
5915
5916 if (!ipv6_accept_ra(idev)) {
5917 NL_SET_ERR_MSG_MOD(extack,
5918 "Router advertisement is disabled on device");
5919 return -EINVAL;
5920 }
5921
5922 if (READ_ONCE(idev->cnf.rtr_solicits) == 0) {
5923 NL_SET_ERR_MSG(extack,
5924 "Router solicitation is disabled on device");
5925 return -EINVAL;
5926 }
5927
5928 write_lock_bh(&idev->lock);
5929
5930 BUILD_BUG_ON(sizeof(token->s6_addr) != 16);
5931 memcpy(idev->token.s6_addr + 8, token->s6_addr + 8, 8);
5932
5933 write_unlock_bh(&idev->lock);
5934
5935 clear_token = ipv6_addr_any(token);
5936 if (clear_token)
5937 goto update_lft;
5938
5939 if (!idev->dead && (idev->if_flags & IF_READY) &&
5940 !ipv6_get_lladdr(dev, &ll_addr, IFA_F_TENTATIVE |
5941 IFA_F_OPTIMISTIC)) {
5942 /* If we're not ready, then normal ifup will take care
5943 * of this. Otherwise, we need to request our rs here.
5944 */
5945 ndisc_send_rs(dev, &ll_addr, &in6addr_linklocal_allrouters);
5946 update_rs = true;
5947 }
5948
5949update_lft:
5950 write_lock_bh(&idev->lock);
5951
5952 if (update_rs) {
5953 idev->if_flags |= IF_RS_SENT;
5954 idev->rs_interval = rfc3315_s14_backoff_init(
5955 READ_ONCE(idev->cnf.rtr_solicit_interval));
5956 idev->rs_probes = 1;
5957 addrconf_mod_rs_timer(idev, idev->rs_interval);
5958 }
5959
5960 /* Well, that's kinda nasty ... */
5961 list_for_each_entry(ifp, &idev->addr_list, if_list) {
5962 spin_lock(&ifp->lock);
5963 if (ifp->tokenized) {
5964 ifp->valid_lft = 0;
5965 ifp->prefered_lft = 0;
5966 }
5967 spin_unlock(&ifp->lock);
5968 }
5969
5970 write_unlock_bh(&idev->lock);
5971 inet6_ifinfo_notify(RTM_NEWLINK, idev);
5972 addrconf_verify_rtnl(dev_net(dev));
5973 return 0;
5974}
5975
5976static const struct nla_policy inet6_af_policy[IFLA_INET6_MAX + 1] = {
5977 [IFLA_INET6_ADDR_GEN_MODE] = { .type = NLA_U8 },
5978 [IFLA_INET6_TOKEN] = { .len = sizeof(struct in6_addr) },
5979 [IFLA_INET6_RA_MTU] = { .type = NLA_REJECT,
5980 .reject_message =
5981 "IFLA_INET6_RA_MTU can not be set" },
5982};
5983
5984static int check_addr_gen_mode(int mode)
5985{
5986 if (mode != IN6_ADDR_GEN_MODE_EUI64 &&
5987 mode != IN6_ADDR_GEN_MODE_NONE &&
5988 mode != IN6_ADDR_GEN_MODE_STABLE_PRIVACY &&
5989 mode != IN6_ADDR_GEN_MODE_RANDOM)
5990 return -EINVAL;
5991 return 1;
5992}
5993
5994static int check_stable_privacy(struct inet6_dev *idev, struct net *net,
5995 int mode)
5996{
5997 if (mode == IN6_ADDR_GEN_MODE_STABLE_PRIVACY &&
5998 !idev->cnf.stable_secret.initialized &&
5999 !net->ipv6.devconf_dflt->stable_secret.initialized)
6000 return -EINVAL;
6001 return 1;
6002}
6003
6004static int inet6_validate_link_af(const struct net_device *dev,
6005 const struct nlattr *nla,
6006 struct netlink_ext_ack *extack)
6007{
6008 struct nlattr *tb[IFLA_INET6_MAX + 1];
6009 struct inet6_dev *idev = NULL;
6010 int err;
6011
6012 if (dev) {
6013 idev = __in6_dev_get(dev);
6014 if (!idev)
6015 return -EAFNOSUPPORT;
6016 }
6017
6018 err = nla_parse_nested_deprecated(tb, IFLA_INET6_MAX, nla,
6019 inet6_af_policy, extack);
6020 if (err)
6021 return err;
6022
6023 if (!tb[IFLA_INET6_TOKEN] && !tb[IFLA_INET6_ADDR_GEN_MODE])
6024 return -EINVAL;
6025
6026 if (tb[IFLA_INET6_ADDR_GEN_MODE]) {
6027 u8 mode = nla_get_u8(tb[IFLA_INET6_ADDR_GEN_MODE]);
6028
6029 if (check_addr_gen_mode(mode) < 0)
6030 return -EINVAL;
6031 if (dev && check_stable_privacy(idev, dev_net(dev), mode) < 0)
6032 return -EINVAL;
6033 }
6034
6035 return 0;
6036}
6037
6038static int inet6_set_link_af(struct net_device *dev, const struct nlattr *nla,
6039 struct netlink_ext_ack *extack)
6040{
6041 struct inet6_dev *idev = __in6_dev_get(dev);
6042 struct nlattr *tb[IFLA_INET6_MAX + 1];
6043 int err;
6044
6045 if (!idev)
6046 return -EAFNOSUPPORT;
6047
6048 if (nla_parse_nested_deprecated(tb, IFLA_INET6_MAX, nla, NULL, NULL) < 0)
6049 return -EINVAL;
6050
6051 if (tb[IFLA_INET6_TOKEN]) {
6052 err = inet6_set_iftoken(idev, nla_data(tb[IFLA_INET6_TOKEN]),
6053 extack);
6054 if (err)
6055 return err;
6056 }
6057
6058 if (tb[IFLA_INET6_ADDR_GEN_MODE]) {
6059 u8 mode = nla_get_u8(tb[IFLA_INET6_ADDR_GEN_MODE]);
6060
6061 WRITE_ONCE(idev->cnf.addr_gen_mode, mode);
6062 }
6063
6064 return 0;
6065}
6066
6067static int inet6_fill_ifinfo(struct sk_buff *skb, struct inet6_dev *idev,
6068 u32 portid, u32 seq, int event, unsigned int flags)
6069{
6070 struct net_device *dev = idev->dev;
6071 struct ifinfomsg *hdr;
6072 struct nlmsghdr *nlh;
6073 int ifindex, iflink;
6074 void *protoinfo;
6075
6076 nlh = nlmsg_put(skb, portid, seq, event, sizeof(*hdr), flags);
6077 if (!nlh)
6078 return -EMSGSIZE;
6079
6080 hdr = nlmsg_data(nlh);
6081 hdr->ifi_family = AF_INET6;
6082 hdr->__ifi_pad = 0;
6083 hdr->ifi_type = dev->type;
6084 ifindex = READ_ONCE(dev->ifindex);
6085 hdr->ifi_index = ifindex;
6086 hdr->ifi_flags = netif_get_flags(dev);
6087 hdr->ifi_change = 0;
6088
6089 iflink = dev_get_iflink(dev);
6090 if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
6091 (dev->addr_len &&
6092 nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
6093 nla_put_u32(skb, IFLA_MTU, READ_ONCE(dev->mtu)) ||
6094 (ifindex != iflink &&
6095 nla_put_u32(skb, IFLA_LINK, iflink)) ||
6096 nla_put_u8(skb, IFLA_OPERSTATE,
6097 netif_running(dev) ? READ_ONCE(dev->operstate) : IF_OPER_DOWN))
6098 goto nla_put_failure;
6099 protoinfo = nla_nest_start_noflag(skb, IFLA_PROTINFO);
6100 if (!protoinfo)
6101 goto nla_put_failure;
6102
6103 if (inet6_fill_ifla6_attrs(skb, idev, 0) < 0)
6104 goto nla_put_failure;
6105
6106 nla_nest_end(skb, protoinfo);
6107 nlmsg_end(skb, nlh);
6108 return 0;
6109
6110nla_put_failure:
6111 nlmsg_cancel(skb, nlh);
6112 return -EMSGSIZE;
6113}
6114
6115static int inet6_valid_dump_ifinfo(const struct nlmsghdr *nlh,
6116 struct netlink_ext_ack *extack)
6117{
6118 struct ifinfomsg *ifm;
6119
6120 ifm = nlmsg_payload(nlh, sizeof(*ifm));
6121 if (!ifm) {
6122 NL_SET_ERR_MSG_MOD(extack, "Invalid header for link dump request");
6123 return -EINVAL;
6124 }
6125
6126 if (nlmsg_attrlen(nlh, sizeof(*ifm))) {
6127 NL_SET_ERR_MSG_MOD(extack, "Invalid data after header");
6128 return -EINVAL;
6129 }
6130
6131 if (ifm->__ifi_pad || ifm->ifi_type || ifm->ifi_flags ||
6132 ifm->ifi_change || ifm->ifi_index) {
6133 NL_SET_ERR_MSG_MOD(extack, "Invalid values in header for dump request");
6134 return -EINVAL;
6135 }
6136
6137 return 0;
6138}
6139
6140static int inet6_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
6141{
6142 struct net *net = sock_net(skb->sk);
6143 struct {
6144 unsigned long ifindex;
6145 } *ctx = (void *)cb->ctx;
6146 struct net_device *dev;
6147 struct inet6_dev *idev;
6148 int err;
6149
6150 /* only requests using strict checking can pass data to
6151 * influence the dump
6152 */
6153 if (cb->strict_check) {
6154 err = inet6_valid_dump_ifinfo(cb->nlh, cb->extack);
6155
6156 if (err < 0)
6157 return err;
6158 }
6159
6160 err = 0;
6161 rcu_read_lock();
6162 for_each_netdev_dump(net, dev, ctx->ifindex) {
6163 idev = __in6_dev_get(dev);
6164 if (!idev)
6165 continue;
6166 err = inet6_fill_ifinfo(skb, idev,
6167 NETLINK_CB(cb->skb).portid,
6168 cb->nlh->nlmsg_seq,
6169 RTM_NEWLINK, NLM_F_MULTI);
6170 if (err < 0)
6171 break;
6172 }
6173 rcu_read_unlock();
6174
6175 return err;
6176}
6177
6178void inet6_ifinfo_notify(int event, struct inet6_dev *idev)
6179{
6180 struct sk_buff *skb;
6181 struct net *net = dev_net(idev->dev);
6182 int err = -ENOBUFS;
6183
6184 skb = nlmsg_new(inet6_if_nlmsg_size(), GFP_ATOMIC);
6185 if (!skb)
6186 goto errout;
6187
6188 err = inet6_fill_ifinfo(skb, idev, 0, 0, event, 0);
6189 if (err < 0) {
6190 /* -EMSGSIZE implies BUG in inet6_if_nlmsg_size() */
6191 WARN_ON(err == -EMSGSIZE);
6192 kfree_skb(skb);
6193 goto errout;
6194 }
6195 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFINFO, NULL, GFP_ATOMIC);
6196 return;
6197errout:
6198 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFINFO, err);
6199}
6200
6201static inline size_t inet6_prefix_nlmsg_size(void)
6202{
6203 return NLMSG_ALIGN(sizeof(struct prefixmsg))
6204 + nla_total_size(sizeof(struct in6_addr))
6205 + nla_total_size(sizeof(struct prefix_cacheinfo));
6206}
6207
6208static int inet6_fill_prefix(struct sk_buff *skb, struct inet6_dev *idev,
6209 struct prefix_info *pinfo, u32 portid, u32 seq,
6210 int event, unsigned int flags)
6211{
6212 struct prefixmsg *pmsg;
6213 struct nlmsghdr *nlh;
6214 struct prefix_cacheinfo ci;
6215
6216 nlh = nlmsg_put(skb, portid, seq, event, sizeof(*pmsg), flags);
6217 if (!nlh)
6218 return -EMSGSIZE;
6219
6220 pmsg = nlmsg_data(nlh);
6221 pmsg->prefix_family = AF_INET6;
6222 pmsg->prefix_pad1 = 0;
6223 pmsg->prefix_pad2 = 0;
6224 pmsg->prefix_ifindex = idev->dev->ifindex;
6225 pmsg->prefix_len = pinfo->prefix_len;
6226 pmsg->prefix_type = pinfo->type;
6227 pmsg->prefix_pad3 = 0;
6228 pmsg->prefix_flags = pinfo->flags;
6229
6230 if (nla_put(skb, PREFIX_ADDRESS, sizeof(pinfo->prefix), &pinfo->prefix))
6231 goto nla_put_failure;
6232 ci.preferred_time = ntohl(pinfo->prefered);
6233 ci.valid_time = ntohl(pinfo->valid);
6234 if (nla_put(skb, PREFIX_CACHEINFO, sizeof(ci), &ci))
6235 goto nla_put_failure;
6236 nlmsg_end(skb, nlh);
6237 return 0;
6238
6239nla_put_failure:
6240 nlmsg_cancel(skb, nlh);
6241 return -EMSGSIZE;
6242}
6243
6244static void inet6_prefix_notify(int event, struct inet6_dev *idev,
6245 struct prefix_info *pinfo)
6246{
6247 struct sk_buff *skb;
6248 struct net *net = dev_net(idev->dev);
6249 int err = -ENOBUFS;
6250
6251 skb = nlmsg_new(inet6_prefix_nlmsg_size(), GFP_ATOMIC);
6252 if (!skb)
6253 goto errout;
6254
6255 err = inet6_fill_prefix(skb, idev, pinfo, 0, 0, event, 0);
6256 if (err < 0) {
6257 /* -EMSGSIZE implies BUG in inet6_prefix_nlmsg_size() */
6258 WARN_ON(err == -EMSGSIZE);
6259 kfree_skb(skb);
6260 goto errout;
6261 }
6262 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_PREFIX, NULL, GFP_ATOMIC);
6263 return;
6264errout:
6265 rtnl_set_sk_err(net, RTNLGRP_IPV6_PREFIX, err);
6266}
6267
6268static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
6269{
6270 struct net *net = dev_net(ifp->idev->dev);
6271
6272 if (event)
6273 ASSERT_RTNL();
6274
6275 inet6_ifa_notify(event ? : RTM_NEWADDR, ifp);
6276
6277 switch (event) {
6278 case RTM_NEWADDR:
6279 /*
6280 * If the address was optimistic we inserted the route at the
6281 * start of our DAD process, so we don't need to do it again.
6282 * If the device was taken down in the middle of the DAD
6283 * cycle there is a race where we could get here without a
6284 * host route, so nothing to insert. That will be fixed when
6285 * the device is brought up.
6286 */
6287 if (ifp->rt && !rcu_access_pointer(ifp->rt->fib6_node)) {
6288 ip6_ins_rt(net, ifp->rt);
6289 } else if (!ifp->rt && (ifp->idev->dev->flags & IFF_UP)) {
6290 pr_warn("BUG: Address %pI6c on device %s is missing its host route.\n",
6291 &ifp->addr, ifp->idev->dev->name);
6292 }
6293
6294 if (ifp->idev->cnf.forwarding)
6295 addrconf_join_anycast(ifp);
6296 if (!ipv6_addr_any(&ifp->peer_addr))
6297 addrconf_prefix_route(&ifp->peer_addr, 128,
6298 ifp->rt_priority, ifp->idev->dev,
6299 0, 0, GFP_ATOMIC);
6300 break;
6301 case RTM_DELADDR:
6302 if (ifp->idev->cnf.forwarding)
6303 addrconf_leave_anycast(ifp);
6304 addrconf_leave_solict(ifp->idev, &ifp->addr);
6305 if (!ipv6_addr_any(&ifp->peer_addr)) {
6306 struct fib6_info *rt;
6307
6308 rt = addrconf_get_prefix_route(&ifp->peer_addr, 128,
6309 ifp->idev->dev, 0, 0,
6310 false);
6311 if (rt)
6312 ip6_del_rt(net, rt, false);
6313 }
6314 if (ifp->rt) {
6315 ip6_del_rt(net, ifp->rt, false);
6316 ifp->rt = NULL;
6317 }
6318 rt_genid_bump_ipv6(net);
6319 break;
6320 }
6321 atomic_inc(&net->ipv6.dev_addr_genid);
6322}
6323
6324static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
6325{
6326 if (likely(ifp->idev->dead == 0))
6327 __ipv6_ifa_notify(event, ifp);
6328}
6329
6330#ifdef CONFIG_SYSCTL
6331
6332static int addrconf_sysctl_forward(const struct ctl_table *ctl, int write,
6333 void *buffer, size_t *lenp, loff_t *ppos)
6334{
6335 int *valp = ctl->data;
6336 int val = *valp;
6337 loff_t pos = *ppos;
6338 struct ctl_table lctl;
6339 int ret;
6340
6341 /*
6342 * ctl->data points to idev->cnf.forwarding, we should
6343 * not modify it until we get the rtnl lock.
6344 */
6345 lctl = *ctl;
6346 lctl.data = &val;
6347
6348 ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
6349
6350 if (write)
6351 ret = addrconf_fixup_forwarding(ctl, valp, val);
6352 if (ret)
6353 *ppos = pos;
6354 return ret;
6355}
6356
6357static int addrconf_sysctl_mtu(const struct ctl_table *ctl, int write,
6358 void *buffer, size_t *lenp, loff_t *ppos)
6359{
6360 struct inet6_dev *idev = ctl->extra1;
6361 int min_mtu = IPV6_MIN_MTU;
6362 struct ctl_table lctl;
6363
6364 lctl = *ctl;
6365 lctl.extra1 = &min_mtu;
6366 lctl.extra2 = idev ? &idev->dev->mtu : NULL;
6367
6368 return proc_dointvec_minmax(&lctl, write, buffer, lenp, ppos);
6369}
6370
6371static void dev_disable_change(struct inet6_dev *idev)
6372{
6373 struct netdev_notifier_info info;
6374
6375 if (!idev || !idev->dev)
6376 return;
6377
6378 netdev_notifier_info_init(&info, idev->dev);
6379 if (idev->cnf.disable_ipv6)
6380 addrconf_notify(NULL, NETDEV_DOWN, &info);
6381 else
6382 addrconf_notify(NULL, NETDEV_UP, &info);
6383}
6384
6385static void addrconf_disable_change(struct net *net, __s32 newf)
6386{
6387 struct net_device *dev;
6388 struct inet6_dev *idev;
6389
6390 for_each_netdev(net, dev) {
6391 idev = __in6_dev_get_rtnl_net(dev);
6392 if (idev) {
6393 int changed = (!idev->cnf.disable_ipv6) ^ (!newf);
6394
6395 WRITE_ONCE(idev->cnf.disable_ipv6, newf);
6396 if (changed)
6397 dev_disable_change(idev);
6398 }
6399 }
6400}
6401
6402static int addrconf_disable_ipv6(const struct ctl_table *table, int *p, int newf)
6403{
6404 struct net *net = (struct net *)table->extra2;
6405 int old;
6406
6407 if (p == &net->ipv6.devconf_dflt->disable_ipv6) {
6408 WRITE_ONCE(*p, newf);
6409 return 0;
6410 }
6411
6412 if (!rtnl_net_trylock(net))
6413 return restart_syscall();
6414
6415 old = *p;
6416 WRITE_ONCE(*p, newf);
6417
6418 if (p == &net->ipv6.devconf_all->disable_ipv6) {
6419 WRITE_ONCE(net->ipv6.devconf_dflt->disable_ipv6, newf);
6420 addrconf_disable_change(net, newf);
6421 } else if ((!newf) ^ (!old)) {
6422 dev_disable_change((struct inet6_dev *)table->extra1);
6423 }
6424
6425 rtnl_net_unlock(net);
6426 return 0;
6427}
6428
6429static int addrconf_sysctl_disable(const struct ctl_table *ctl, int write,
6430 void *buffer, size_t *lenp, loff_t *ppos)
6431{
6432 int *valp = ctl->data;
6433 int val = *valp;
6434 loff_t pos = *ppos;
6435 struct ctl_table lctl;
6436 int ret;
6437
6438 /*
6439 * ctl->data points to idev->cnf.disable_ipv6, we should
6440 * not modify it until we get the rtnl lock.
6441 */
6442 lctl = *ctl;
6443 lctl.data = &val;
6444
6445 ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
6446
6447 if (write)
6448 ret = addrconf_disable_ipv6(ctl, valp, val);
6449 if (ret)
6450 *ppos = pos;
6451 return ret;
6452}
6453
6454static int addrconf_sysctl_proxy_ndp(const struct ctl_table *ctl, int write,
6455 void *buffer, size_t *lenp, loff_t *ppos)
6456{
6457 int *valp = ctl->data;
6458 int ret;
6459 int old, new;
6460
6461 old = *valp;
6462 ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
6463 new = *valp;
6464
6465 if (write && old != new) {
6466 struct net *net = ctl->extra2;
6467
6468 if (!rtnl_net_trylock(net))
6469 return restart_syscall();
6470
6471 if (valp == &net->ipv6.devconf_dflt->proxy_ndp) {
6472 inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
6473 NETCONFA_PROXY_NEIGH,
6474 NETCONFA_IFINDEX_DEFAULT,
6475 net->ipv6.devconf_dflt);
6476 } else if (valp == &net->ipv6.devconf_all->proxy_ndp) {
6477 inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
6478 NETCONFA_PROXY_NEIGH,
6479 NETCONFA_IFINDEX_ALL,
6480 net->ipv6.devconf_all);
6481 } else {
6482 struct inet6_dev *idev = ctl->extra1;
6483
6484 inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
6485 NETCONFA_PROXY_NEIGH,
6486 idev->dev->ifindex,
6487 &idev->cnf);
6488 }
6489 rtnl_net_unlock(net);
6490 }
6491
6492 return ret;
6493}
6494
6495static int addrconf_sysctl_addr_gen_mode(const struct ctl_table *ctl, int write,
6496 void *buffer, size_t *lenp,
6497 loff_t *ppos)
6498{
6499 int ret = 0;
6500 u32 new_val;
6501 struct inet6_dev *idev = (struct inet6_dev *)ctl->extra1;
6502 struct net *net = (struct net *)ctl->extra2;
6503 struct ctl_table tmp = {
6504 .data = &new_val,
6505 .maxlen = sizeof(new_val),
6506 .mode = ctl->mode,
6507 };
6508
6509 if (!rtnl_net_trylock(net))
6510 return restart_syscall();
6511
6512 new_val = *((u32 *)ctl->data);
6513
6514 ret = proc_douintvec(&tmp, write, buffer, lenp, ppos);
6515 if (ret != 0)
6516 goto out;
6517
6518 if (write) {
6519 if (check_addr_gen_mode(new_val) < 0) {
6520 ret = -EINVAL;
6521 goto out;
6522 }
6523
6524 if (idev) {
6525 if (check_stable_privacy(idev, net, new_val) < 0) {
6526 ret = -EINVAL;
6527 goto out;
6528 }
6529
6530 if (idev->cnf.addr_gen_mode != new_val) {
6531 WRITE_ONCE(idev->cnf.addr_gen_mode, new_val);
6532 netdev_lock_ops(idev->dev);
6533 addrconf_init_auto_addrs(idev->dev);
6534 netdev_unlock_ops(idev->dev);
6535 }
6536 } else if (&net->ipv6.devconf_all->addr_gen_mode == ctl->data) {
6537 struct net_device *dev;
6538
6539 WRITE_ONCE(net->ipv6.devconf_dflt->addr_gen_mode, new_val);
6540 for_each_netdev(net, dev) {
6541 idev = __in6_dev_get_rtnl_net(dev);
6542 if (idev &&
6543 idev->cnf.addr_gen_mode != new_val) {
6544 WRITE_ONCE(idev->cnf.addr_gen_mode,
6545 new_val);
6546 netdev_lock_ops(idev->dev);
6547 addrconf_init_auto_addrs(idev->dev);
6548 netdev_unlock_ops(idev->dev);
6549 }
6550 }
6551 }
6552
6553 WRITE_ONCE(*((u32 *)ctl->data), new_val);
6554 }
6555
6556out:
6557 rtnl_net_unlock(net);
6558
6559 return ret;
6560}
6561
6562static int addrconf_sysctl_stable_secret(const struct ctl_table *ctl, int write,
6563 void *buffer, size_t *lenp,
6564 loff_t *ppos)
6565{
6566 int err;
6567 struct in6_addr addr;
6568 char str[IPV6_MAX_STRLEN];
6569 struct ctl_table lctl = *ctl;
6570 struct net *net = ctl->extra2;
6571 struct ipv6_stable_secret *secret = ctl->data;
6572
6573 if (&net->ipv6.devconf_all->stable_secret == ctl->data)
6574 return -EIO;
6575
6576 lctl.maxlen = IPV6_MAX_STRLEN;
6577 lctl.data = str;
6578
6579 if (!rtnl_net_trylock(net))
6580 return restart_syscall();
6581
6582 if (!write && !secret->initialized) {
6583 err = -EIO;
6584 goto out;
6585 }
6586
6587 err = snprintf(str, sizeof(str), "%pI6", &secret->secret);
6588 if (err >= sizeof(str)) {
6589 err = -EIO;
6590 goto out;
6591 }
6592
6593 err = proc_dostring(&lctl, write, buffer, lenp, ppos);
6594 if (err || !write)
6595 goto out;
6596
6597 if (in6_pton(str, -1, addr.in6_u.u6_addr8, -1, NULL) != 1) {
6598 err = -EIO;
6599 goto out;
6600 }
6601
6602 secret->initialized = true;
6603 secret->secret = addr;
6604
6605 if (&net->ipv6.devconf_dflt->stable_secret == ctl->data) {
6606 struct net_device *dev;
6607
6608 for_each_netdev(net, dev) {
6609 struct inet6_dev *idev = __in6_dev_get_rtnl_net(dev);
6610
6611 if (idev) {
6612 WRITE_ONCE(idev->cnf.addr_gen_mode,
6613 IN6_ADDR_GEN_MODE_STABLE_PRIVACY);
6614 }
6615 }
6616 } else {
6617 struct inet6_dev *idev = ctl->extra1;
6618
6619 WRITE_ONCE(idev->cnf.addr_gen_mode,
6620 IN6_ADDR_GEN_MODE_STABLE_PRIVACY);
6621 }
6622
6623out:
6624 rtnl_net_unlock(net);
6625
6626 return err;
6627}
6628
6629static
6630int addrconf_sysctl_ignore_routes_with_linkdown(const struct ctl_table *ctl,
6631 int write, void *buffer,
6632 size_t *lenp,
6633 loff_t *ppos)
6634{
6635 int *valp = ctl->data;
6636 int val = *valp;
6637 loff_t pos = *ppos;
6638 struct ctl_table lctl;
6639 int ret;
6640
6641 /* ctl->data points to idev->cnf.ignore_routes_when_linkdown
6642 * we should not modify it until we get the rtnl lock.
6643 */
6644 lctl = *ctl;
6645 lctl.data = &val;
6646
6647 ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
6648
6649 if (write)
6650 ret = addrconf_fixup_linkdown(ctl, valp, val);
6651 if (ret)
6652 *ppos = pos;
6653 return ret;
6654}
6655
6656static
6657void addrconf_set_nopolicy(struct rt6_info *rt, int action)
6658{
6659 if (rt) {
6660 if (action)
6661 rt->dst.flags |= DST_NOPOLICY;
6662 else
6663 rt->dst.flags &= ~DST_NOPOLICY;
6664 }
6665}
6666
6667static
6668void addrconf_disable_policy_idev(struct inet6_dev *idev, int val)
6669{
6670 struct inet6_ifaddr *ifa;
6671
6672 read_lock_bh(&idev->lock);
6673 list_for_each_entry(ifa, &idev->addr_list, if_list) {
6674 spin_lock(&ifa->lock);
6675 if (ifa->rt) {
6676 /* host routes only use builtin fib6_nh */
6677 struct fib6_nh *nh = ifa->rt->fib6_nh;
6678 int cpu;
6679
6680 rcu_read_lock();
6681 ifa->rt->dst_nopolicy = val ? true : false;
6682 if (nh->rt6i_pcpu) {
6683 for_each_possible_cpu(cpu) {
6684 struct rt6_info **rtp;
6685
6686 rtp = per_cpu_ptr(nh->rt6i_pcpu, cpu);
6687 addrconf_set_nopolicy(*rtp, val);
6688 }
6689 }
6690 rcu_read_unlock();
6691 }
6692 spin_unlock(&ifa->lock);
6693 }
6694 read_unlock_bh(&idev->lock);
6695}
6696
6697static
6698int addrconf_disable_policy(const struct ctl_table *ctl, int *valp, int val)
6699{
6700 struct net *net = (struct net *)ctl->extra2;
6701 struct inet6_dev *idev;
6702
6703 if (valp == &net->ipv6.devconf_dflt->disable_policy) {
6704 WRITE_ONCE(*valp, val);
6705 return 0;
6706 }
6707
6708 if (!rtnl_net_trylock(net))
6709 return restart_syscall();
6710
6711 WRITE_ONCE(*valp, val);
6712
6713 if (valp == &net->ipv6.devconf_all->disable_policy) {
6714 struct net_device *dev;
6715
6716 for_each_netdev(net, dev) {
6717 idev = __in6_dev_get_rtnl_net(dev);
6718 if (idev)
6719 addrconf_disable_policy_idev(idev, val);
6720 }
6721 } else {
6722 idev = (struct inet6_dev *)ctl->extra1;
6723 addrconf_disable_policy_idev(idev, val);
6724 }
6725
6726 rtnl_net_unlock(net);
6727 return 0;
6728}
6729
6730static int addrconf_sysctl_disable_policy(const struct ctl_table *ctl, int write,
6731 void *buffer, size_t *lenp, loff_t *ppos)
6732{
6733 int *valp = ctl->data;
6734 int val = *valp;
6735 loff_t pos = *ppos;
6736 struct ctl_table lctl;
6737 int ret;
6738
6739 lctl = *ctl;
6740 lctl.data = &val;
6741 ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
6742
6743 if (write && (*valp != val))
6744 ret = addrconf_disable_policy(ctl, valp, val);
6745
6746 if (ret)
6747 *ppos = pos;
6748
6749 return ret;
6750}
6751
6752static void addrconf_force_forward_change(struct net *net, __s32 newf)
6753{
6754 struct net_device *dev;
6755 struct inet6_dev *idev;
6756
6757 for_each_netdev(net, dev) {
6758 idev = __in6_dev_get_rtnl_net(dev);
6759 if (idev) {
6760 int changed = (!idev->cnf.force_forwarding) ^ (!newf);
6761
6762 WRITE_ONCE(idev->cnf.force_forwarding, newf);
6763 if (changed)
6764 inet6_netconf_notify_devconf(dev_net(dev), RTM_NEWNETCONF,
6765 NETCONFA_FORCE_FORWARDING,
6766 dev->ifindex, &idev->cnf);
6767 }
6768 }
6769}
6770
6771static int addrconf_sysctl_force_forwarding(const struct ctl_table *ctl, int write,
6772 void *buffer, size_t *lenp, loff_t *ppos)
6773{
6774 struct inet6_dev *idev = ctl->extra1;
6775 struct ctl_table tmp_ctl = *ctl;
6776 struct net *net = ctl->extra2;
6777 int *valp = ctl->data;
6778 int new_val = *valp;
6779 int old_val = *valp;
6780 loff_t pos = *ppos;
6781 int ret;
6782
6783 tmp_ctl.extra1 = SYSCTL_ZERO;
6784 tmp_ctl.extra2 = SYSCTL_ONE;
6785 tmp_ctl.data = &new_val;
6786
6787 ret = proc_douintvec_minmax(&tmp_ctl, write, buffer, lenp, ppos);
6788
6789 if (write && old_val != new_val) {
6790 if (!rtnl_net_trylock(net))
6791 return restart_syscall();
6792
6793 WRITE_ONCE(*valp, new_val);
6794
6795 if (valp == &net->ipv6.devconf_dflt->force_forwarding) {
6796 inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
6797 NETCONFA_FORCE_FORWARDING,
6798 NETCONFA_IFINDEX_DEFAULT,
6799 net->ipv6.devconf_dflt);
6800 } else if (valp == &net->ipv6.devconf_all->force_forwarding) {
6801 inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
6802 NETCONFA_FORCE_FORWARDING,
6803 NETCONFA_IFINDEX_ALL,
6804 net->ipv6.devconf_all);
6805
6806 addrconf_force_forward_change(net, new_val);
6807 } else {
6808 inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
6809 NETCONFA_FORCE_FORWARDING,
6810 idev->dev->ifindex,
6811 &idev->cnf);
6812 }
6813 rtnl_net_unlock(net);
6814 }
6815
6816 if (ret)
6817 *ppos = pos;
6818 return ret;
6819}
6820
6821static int minus_one = -1;
6822static const int two_five_five = 255;
6823static u32 ioam6_if_id_max = U16_MAX;
6824
6825static const struct ctl_table addrconf_sysctl[] = {
6826 {
6827 .procname = "forwarding",
6828 .data = &ipv6_devconf.forwarding,
6829 .maxlen = sizeof(int),
6830 .mode = 0644,
6831 .proc_handler = addrconf_sysctl_forward,
6832 },
6833 {
6834 .procname = "hop_limit",
6835 .data = &ipv6_devconf.hop_limit,
6836 .maxlen = sizeof(int),
6837 .mode = 0644,
6838 .proc_handler = proc_dointvec_minmax,
6839 .extra1 = (void *)SYSCTL_ONE,
6840 .extra2 = (void *)&two_five_five,
6841 },
6842 {
6843 .procname = "mtu",
6844 .data = &ipv6_devconf.mtu6,
6845 .maxlen = sizeof(int),
6846 .mode = 0644,
6847 .proc_handler = addrconf_sysctl_mtu,
6848 },
6849 {
6850 .procname = "accept_ra",
6851 .data = &ipv6_devconf.accept_ra,
6852 .maxlen = sizeof(int),
6853 .mode = 0644,
6854 .proc_handler = proc_dointvec,
6855 },
6856 {
6857 .procname = "accept_redirects",
6858 .data = &ipv6_devconf.accept_redirects,
6859 .maxlen = sizeof(int),
6860 .mode = 0644,
6861 .proc_handler = proc_dointvec,
6862 },
6863 {
6864 .procname = "autoconf",
6865 .data = &ipv6_devconf.autoconf,
6866 .maxlen = sizeof(int),
6867 .mode = 0644,
6868 .proc_handler = proc_dointvec,
6869 },
6870 {
6871 .procname = "dad_transmits",
6872 .data = &ipv6_devconf.dad_transmits,
6873 .maxlen = sizeof(int),
6874 .mode = 0644,
6875 .proc_handler = proc_dointvec,
6876 },
6877 {
6878 .procname = "router_solicitations",
6879 .data = &ipv6_devconf.rtr_solicits,
6880 .maxlen = sizeof(int),
6881 .mode = 0644,
6882 .proc_handler = proc_dointvec_minmax,
6883 .extra1 = &minus_one,
6884 },
6885 {
6886 .procname = "router_solicitation_interval",
6887 .data = &ipv6_devconf.rtr_solicit_interval,
6888 .maxlen = sizeof(int),
6889 .mode = 0644,
6890 .proc_handler = proc_dointvec_jiffies,
6891 },
6892 {
6893 .procname = "router_solicitation_max_interval",
6894 .data = &ipv6_devconf.rtr_solicit_max_interval,
6895 .maxlen = sizeof(int),
6896 .mode = 0644,
6897 .proc_handler = proc_dointvec_jiffies,
6898 },
6899 {
6900 .procname = "router_solicitation_delay",
6901 .data = &ipv6_devconf.rtr_solicit_delay,
6902 .maxlen = sizeof(int),
6903 .mode = 0644,
6904 .proc_handler = proc_dointvec_jiffies,
6905 },
6906 {
6907 .procname = "force_mld_version",
6908 .data = &ipv6_devconf.force_mld_version,
6909 .maxlen = sizeof(int),
6910 .mode = 0644,
6911 .proc_handler = proc_dointvec,
6912 },
6913 {
6914 .procname = "mldv1_unsolicited_report_interval",
6915 .data =
6916 &ipv6_devconf.mldv1_unsolicited_report_interval,
6917 .maxlen = sizeof(int),
6918 .mode = 0644,
6919 .proc_handler = proc_dointvec_ms_jiffies,
6920 },
6921 {
6922 .procname = "mldv2_unsolicited_report_interval",
6923 .data =
6924 &ipv6_devconf.mldv2_unsolicited_report_interval,
6925 .maxlen = sizeof(int),
6926 .mode = 0644,
6927 .proc_handler = proc_dointvec_ms_jiffies,
6928 },
6929 {
6930 .procname = "use_tempaddr",
6931 .data = &ipv6_devconf.use_tempaddr,
6932 .maxlen = sizeof(int),
6933 .mode = 0644,
6934 .proc_handler = proc_dointvec,
6935 },
6936 {
6937 .procname = "temp_valid_lft",
6938 .data = &ipv6_devconf.temp_valid_lft,
6939 .maxlen = sizeof(int),
6940 .mode = 0644,
6941 .proc_handler = proc_dointvec,
6942 },
6943 {
6944 .procname = "temp_prefered_lft",
6945 .data = &ipv6_devconf.temp_prefered_lft,
6946 .maxlen = sizeof(int),
6947 .mode = 0644,
6948 .proc_handler = proc_dointvec,
6949 },
6950 {
6951 .procname = "regen_min_advance",
6952 .data = &ipv6_devconf.regen_min_advance,
6953 .maxlen = sizeof(int),
6954 .mode = 0644,
6955 .proc_handler = proc_dointvec,
6956 },
6957 {
6958 .procname = "regen_max_retry",
6959 .data = &ipv6_devconf.regen_max_retry,
6960 .maxlen = sizeof(int),
6961 .mode = 0644,
6962 .proc_handler = proc_dointvec,
6963 },
6964 {
6965 .procname = "max_desync_factor",
6966 .data = &ipv6_devconf.max_desync_factor,
6967 .maxlen = sizeof(int),
6968 .mode = 0644,
6969 .proc_handler = proc_dointvec,
6970 },
6971 {
6972 .procname = "max_addresses",
6973 .data = &ipv6_devconf.max_addresses,
6974 .maxlen = sizeof(int),
6975 .mode = 0644,
6976 .proc_handler = proc_dointvec,
6977 },
6978 {
6979 .procname = "accept_ra_defrtr",
6980 .data = &ipv6_devconf.accept_ra_defrtr,
6981 .maxlen = sizeof(int),
6982 .mode = 0644,
6983 .proc_handler = proc_dointvec,
6984 },
6985 {
6986 .procname = "ra_defrtr_metric",
6987 .data = &ipv6_devconf.ra_defrtr_metric,
6988 .maxlen = sizeof(u32),
6989 .mode = 0644,
6990 .proc_handler = proc_douintvec_minmax,
6991 .extra1 = (void *)SYSCTL_ONE,
6992 },
6993 {
6994 .procname = "accept_ra_min_hop_limit",
6995 .data = &ipv6_devconf.accept_ra_min_hop_limit,
6996 .maxlen = sizeof(int),
6997 .mode = 0644,
6998 .proc_handler = proc_dointvec,
6999 },
7000 {
7001 .procname = "accept_ra_min_lft",
7002 .data = &ipv6_devconf.accept_ra_min_lft,
7003 .maxlen = sizeof(int),
7004 .mode = 0644,
7005 .proc_handler = proc_dointvec,
7006 },
7007 {
7008 .procname = "accept_ra_pinfo",
7009 .data = &ipv6_devconf.accept_ra_pinfo,
7010 .maxlen = sizeof(int),
7011 .mode = 0644,
7012 .proc_handler = proc_dointvec,
7013 },
7014 {
7015 .procname = "ra_honor_pio_life",
7016 .data = &ipv6_devconf.ra_honor_pio_life,
7017 .maxlen = sizeof(u8),
7018 .mode = 0644,
7019 .proc_handler = proc_dou8vec_minmax,
7020 .extra1 = SYSCTL_ZERO,
7021 .extra2 = SYSCTL_ONE,
7022 },
7023 {
7024 .procname = "ra_honor_pio_pflag",
7025 .data = &ipv6_devconf.ra_honor_pio_pflag,
7026 .maxlen = sizeof(u8),
7027 .mode = 0644,
7028 .proc_handler = proc_dou8vec_minmax,
7029 .extra1 = SYSCTL_ZERO,
7030 .extra2 = SYSCTL_ONE,
7031 },
7032#ifdef CONFIG_IPV6_ROUTER_PREF
7033 {
7034 .procname = "accept_ra_rtr_pref",
7035 .data = &ipv6_devconf.accept_ra_rtr_pref,
7036 .maxlen = sizeof(int),
7037 .mode = 0644,
7038 .proc_handler = proc_dointvec,
7039 },
7040 {
7041 .procname = "router_probe_interval",
7042 .data = &ipv6_devconf.rtr_probe_interval,
7043 .maxlen = sizeof(int),
7044 .mode = 0644,
7045 .proc_handler = proc_dointvec_jiffies,
7046 },
7047#ifdef CONFIG_IPV6_ROUTE_INFO
7048 {
7049 .procname = "accept_ra_rt_info_min_plen",
7050 .data = &ipv6_devconf.accept_ra_rt_info_min_plen,
7051 .maxlen = sizeof(int),
7052 .mode = 0644,
7053 .proc_handler = proc_dointvec,
7054 },
7055 {
7056 .procname = "accept_ra_rt_info_max_plen",
7057 .data = &ipv6_devconf.accept_ra_rt_info_max_plen,
7058 .maxlen = sizeof(int),
7059 .mode = 0644,
7060 .proc_handler = proc_dointvec,
7061 },
7062#endif
7063#endif
7064 {
7065 .procname = "proxy_ndp",
7066 .data = &ipv6_devconf.proxy_ndp,
7067 .maxlen = sizeof(int),
7068 .mode = 0644,
7069 .proc_handler = addrconf_sysctl_proxy_ndp,
7070 },
7071 {
7072 .procname = "accept_source_route",
7073 .data = &ipv6_devconf.accept_source_route,
7074 .maxlen = sizeof(int),
7075 .mode = 0644,
7076 .proc_handler = proc_dointvec,
7077 },
7078#ifdef CONFIG_IPV6_OPTIMISTIC_DAD
7079 {
7080 .procname = "optimistic_dad",
7081 .data = &ipv6_devconf.optimistic_dad,
7082 .maxlen = sizeof(int),
7083 .mode = 0644,
7084 .proc_handler = proc_dointvec,
7085 },
7086 {
7087 .procname = "use_optimistic",
7088 .data = &ipv6_devconf.use_optimistic,
7089 .maxlen = sizeof(int),
7090 .mode = 0644,
7091 .proc_handler = proc_dointvec,
7092 },
7093#endif
7094#ifdef CONFIG_IPV6_MROUTE
7095 {
7096 .procname = "mc_forwarding",
7097 .data = &ipv6_devconf.mc_forwarding,
7098 .maxlen = sizeof(int),
7099 .mode = 0444,
7100 .proc_handler = proc_dointvec,
7101 },
7102#endif
7103 {
7104 .procname = "disable_ipv6",
7105 .data = &ipv6_devconf.disable_ipv6,
7106 .maxlen = sizeof(int),
7107 .mode = 0644,
7108 .proc_handler = addrconf_sysctl_disable,
7109 },
7110 {
7111 .procname = "accept_dad",
7112 .data = &ipv6_devconf.accept_dad,
7113 .maxlen = sizeof(int),
7114 .mode = 0644,
7115 .proc_handler = proc_dointvec,
7116 },
7117 {
7118 .procname = "force_tllao",
7119 .data = &ipv6_devconf.force_tllao,
7120 .maxlen = sizeof(int),
7121 .mode = 0644,
7122 .proc_handler = proc_dointvec
7123 },
7124 {
7125 .procname = "ndisc_notify",
7126 .data = &ipv6_devconf.ndisc_notify,
7127 .maxlen = sizeof(int),
7128 .mode = 0644,
7129 .proc_handler = proc_dointvec
7130 },
7131 {
7132 .procname = "suppress_frag_ndisc",
7133 .data = &ipv6_devconf.suppress_frag_ndisc,
7134 .maxlen = sizeof(int),
7135 .mode = 0644,
7136 .proc_handler = proc_dointvec
7137 },
7138 {
7139 .procname = "accept_ra_from_local",
7140 .data = &ipv6_devconf.accept_ra_from_local,
7141 .maxlen = sizeof(int),
7142 .mode = 0644,
7143 .proc_handler = proc_dointvec,
7144 },
7145 {
7146 .procname = "accept_ra_mtu",
7147 .data = &ipv6_devconf.accept_ra_mtu,
7148 .maxlen = sizeof(int),
7149 .mode = 0644,
7150 .proc_handler = proc_dointvec,
7151 },
7152 {
7153 .procname = "stable_secret",
7154 .data = &ipv6_devconf.stable_secret,
7155 .maxlen = IPV6_MAX_STRLEN,
7156 .mode = 0600,
7157 .proc_handler = addrconf_sysctl_stable_secret,
7158 },
7159 {
7160 .procname = "use_oif_addrs_only",
7161 .data = &ipv6_devconf.use_oif_addrs_only,
7162 .maxlen = sizeof(int),
7163 .mode = 0644,
7164 .proc_handler = proc_dointvec,
7165 },
7166 {
7167 .procname = "ignore_routes_with_linkdown",
7168 .data = &ipv6_devconf.ignore_routes_with_linkdown,
7169 .maxlen = sizeof(int),
7170 .mode = 0644,
7171 .proc_handler = addrconf_sysctl_ignore_routes_with_linkdown,
7172 },
7173 {
7174 .procname = "drop_unicast_in_l2_multicast",
7175 .data = &ipv6_devconf.drop_unicast_in_l2_multicast,
7176 .maxlen = sizeof(int),
7177 .mode = 0644,
7178 .proc_handler = proc_dointvec,
7179 },
7180 {
7181 .procname = "drop_unsolicited_na",
7182 .data = &ipv6_devconf.drop_unsolicited_na,
7183 .maxlen = sizeof(int),
7184 .mode = 0644,
7185 .proc_handler = proc_dointvec,
7186 },
7187 {
7188 .procname = "keep_addr_on_down",
7189 .data = &ipv6_devconf.keep_addr_on_down,
7190 .maxlen = sizeof(int),
7191 .mode = 0644,
7192 .proc_handler = proc_dointvec,
7193
7194 },
7195 {
7196 .procname = "seg6_enabled",
7197 .data = &ipv6_devconf.seg6_enabled,
7198 .maxlen = sizeof(int),
7199 .mode = 0644,
7200 .proc_handler = proc_dointvec,
7201 },
7202#ifdef CONFIG_IPV6_SEG6_HMAC
7203 {
7204 .procname = "seg6_require_hmac",
7205 .data = &ipv6_devconf.seg6_require_hmac,
7206 .maxlen = sizeof(int),
7207 .mode = 0644,
7208 .proc_handler = proc_dointvec,
7209 },
7210#endif
7211 {
7212 .procname = "enhanced_dad",
7213 .data = &ipv6_devconf.enhanced_dad,
7214 .maxlen = sizeof(int),
7215 .mode = 0644,
7216 .proc_handler = proc_dointvec,
7217 },
7218 {
7219 .procname = "addr_gen_mode",
7220 .data = &ipv6_devconf.addr_gen_mode,
7221 .maxlen = sizeof(int),
7222 .mode = 0644,
7223 .proc_handler = addrconf_sysctl_addr_gen_mode,
7224 },
7225 {
7226 .procname = "disable_policy",
7227 .data = &ipv6_devconf.disable_policy,
7228 .maxlen = sizeof(int),
7229 .mode = 0644,
7230 .proc_handler = addrconf_sysctl_disable_policy,
7231 },
7232 {
7233 .procname = "ndisc_tclass",
7234 .data = &ipv6_devconf.ndisc_tclass,
7235 .maxlen = sizeof(int),
7236 .mode = 0644,
7237 .proc_handler = proc_dointvec_minmax,
7238 .extra1 = (void *)SYSCTL_ZERO,
7239 .extra2 = (void *)&two_five_five,
7240 },
7241 {
7242 .procname = "rpl_seg_enabled",
7243 .data = &ipv6_devconf.rpl_seg_enabled,
7244 .maxlen = sizeof(int),
7245 .mode = 0644,
7246 .proc_handler = proc_dointvec_minmax,
7247 .extra1 = SYSCTL_ZERO,
7248 .extra2 = SYSCTL_ONE,
7249 },
7250 {
7251 .procname = "ioam6_enabled",
7252 .data = &ipv6_devconf.ioam6_enabled,
7253 .maxlen = sizeof(u8),
7254 .mode = 0644,
7255 .proc_handler = proc_dou8vec_minmax,
7256 .extra1 = (void *)SYSCTL_ZERO,
7257 .extra2 = (void *)SYSCTL_ONE,
7258 },
7259 {
7260 .procname = "ioam6_id",
7261 .data = &ipv6_devconf.ioam6_id,
7262 .maxlen = sizeof(u32),
7263 .mode = 0644,
7264 .proc_handler = proc_douintvec_minmax,
7265 .extra1 = (void *)SYSCTL_ZERO,
7266 .extra2 = (void *)&ioam6_if_id_max,
7267 },
7268 {
7269 .procname = "ioam6_id_wide",
7270 .data = &ipv6_devconf.ioam6_id_wide,
7271 .maxlen = sizeof(u32),
7272 .mode = 0644,
7273 .proc_handler = proc_douintvec,
7274 },
7275 {
7276 .procname = "ndisc_evict_nocarrier",
7277 .data = &ipv6_devconf.ndisc_evict_nocarrier,
7278 .maxlen = sizeof(u8),
7279 .mode = 0644,
7280 .proc_handler = proc_dou8vec_minmax,
7281 .extra1 = (void *)SYSCTL_ZERO,
7282 .extra2 = (void *)SYSCTL_ONE,
7283 },
7284 {
7285 .procname = "accept_untracked_na",
7286 .data = &ipv6_devconf.accept_untracked_na,
7287 .maxlen = sizeof(int),
7288 .mode = 0644,
7289 .proc_handler = proc_dointvec_minmax,
7290 .extra1 = SYSCTL_ZERO,
7291 .extra2 = SYSCTL_TWO,
7292 },
7293 {
7294 .procname = "force_forwarding",
7295 .data = &ipv6_devconf.force_forwarding,
7296 .maxlen = sizeof(int),
7297 .mode = 0644,
7298 .proc_handler = addrconf_sysctl_force_forwarding,
7299 },
7300};
7301
7302static int __addrconf_sysctl_register(struct net *net, char *dev_name,
7303 struct inet6_dev *idev, struct ipv6_devconf *p)
7304{
7305 size_t table_size = ARRAY_SIZE(addrconf_sysctl);
7306 int i, ifindex;
7307 struct ctl_table *table;
7308 char path[sizeof("net/ipv6/conf/") + IFNAMSIZ];
7309
7310 table = kmemdup(addrconf_sysctl, sizeof(addrconf_sysctl), GFP_KERNEL_ACCOUNT);
7311 if (!table)
7312 goto out;
7313
7314 for (i = 0; i < table_size; i++) {
7315 table[i].data += (char *)p - (char *)&ipv6_devconf;
7316 /* If one of these is already set, then it is not safe to
7317 * overwrite either of them: this makes proc_dointvec_minmax
7318 * usable.
7319 */
7320 if (!table[i].extra1 && !table[i].extra2) {
7321 table[i].extra1 = idev; /* embedded; no ref */
7322 table[i].extra2 = net;
7323 }
7324 }
7325
7326 snprintf(path, sizeof(path), "net/ipv6/conf/%s", dev_name);
7327
7328 p->sysctl_header = register_net_sysctl_sz(net, path, table,
7329 table_size);
7330 if (!p->sysctl_header)
7331 goto free;
7332
7333 if (!strcmp(dev_name, "all"))
7334 ifindex = NETCONFA_IFINDEX_ALL;
7335 else if (!strcmp(dev_name, "default"))
7336 ifindex = NETCONFA_IFINDEX_DEFAULT;
7337 else
7338 ifindex = idev->dev->ifindex;
7339 inet6_netconf_notify_devconf(net, RTM_NEWNETCONF, NETCONFA_ALL,
7340 ifindex, p);
7341 return 0;
7342
7343free:
7344 kfree(table);
7345out:
7346 return -ENOBUFS;
7347}
7348
7349static void __addrconf_sysctl_unregister(struct net *net,
7350 struct ipv6_devconf *p, int ifindex)
7351{
7352 const struct ctl_table *table;
7353
7354 if (!p->sysctl_header)
7355 return;
7356
7357 table = p->sysctl_header->ctl_table_arg;
7358 unregister_net_sysctl_table(p->sysctl_header);
7359 p->sysctl_header = NULL;
7360 kfree(table);
7361
7362 inet6_netconf_notify_devconf(net, RTM_DELNETCONF, 0, ifindex, NULL);
7363}
7364
7365static int addrconf_sysctl_register(struct inet6_dev *idev)
7366{
7367 int err;
7368
7369 if (!sysctl_dev_name_is_allowed(idev->dev->name))
7370 return -EINVAL;
7371
7372 err = neigh_sysctl_register(idev->dev, idev->nd_parms,
7373 &ndisc_ifinfo_sysctl_change);
7374 if (err)
7375 return err;
7376 err = __addrconf_sysctl_register(dev_net(idev->dev), idev->dev->name,
7377 idev, &idev->cnf);
7378 if (err)
7379 neigh_sysctl_unregister(idev->nd_parms);
7380
7381 return err;
7382}
7383
7384static void addrconf_sysctl_unregister(struct inet6_dev *idev)
7385{
7386 __addrconf_sysctl_unregister(dev_net(idev->dev), &idev->cnf,
7387 idev->dev->ifindex);
7388 neigh_sysctl_unregister(idev->nd_parms);
7389}
7390
7391
7392#endif
7393
7394static int __net_init addrconf_init_net(struct net *net)
7395{
7396 int err = -ENOMEM;
7397 struct ipv6_devconf *all, *dflt;
7398
7399 spin_lock_init(&net->ipv6.addrconf_hash_lock);
7400 INIT_DEFERRABLE_WORK(&net->ipv6.addr_chk_work, addrconf_verify_work);
7401 net->ipv6.inet6_addr_lst = kcalloc(IN6_ADDR_HSIZE,
7402 sizeof(struct hlist_head),
7403 GFP_KERNEL);
7404 if (!net->ipv6.inet6_addr_lst)
7405 goto err_alloc_addr;
7406
7407 all = kmemdup(&ipv6_devconf, sizeof(ipv6_devconf), GFP_KERNEL);
7408 if (!all)
7409 goto err_alloc_all;
7410
7411 dflt = kmemdup(&ipv6_devconf_dflt, sizeof(ipv6_devconf_dflt), GFP_KERNEL);
7412 if (!dflt)
7413 goto err_alloc_dflt;
7414
7415 if (!net_eq(net, &init_net)) {
7416 switch (net_inherit_devconf()) {
7417 case 1: /* copy from init_net */
7418 memcpy(all, init_net.ipv6.devconf_all,
7419 sizeof(ipv6_devconf));
7420 memcpy(dflt, init_net.ipv6.devconf_dflt,
7421 sizeof(ipv6_devconf_dflt));
7422 break;
7423 case 3: /* copy from the current netns */
7424 memcpy(all, current->nsproxy->net_ns->ipv6.devconf_all,
7425 sizeof(ipv6_devconf));
7426 memcpy(dflt,
7427 current->nsproxy->net_ns->ipv6.devconf_dflt,
7428 sizeof(ipv6_devconf_dflt));
7429 break;
7430 case 0:
7431 case 2:
7432 /* use compiled values */
7433 break;
7434 }
7435 }
7436
7437 /* these will be inherited by all namespaces */
7438 dflt->autoconf = ipv6_defaults.autoconf;
7439 dflt->disable_ipv6 = ipv6_defaults.disable_ipv6;
7440
7441 dflt->stable_secret.initialized = false;
7442 all->stable_secret.initialized = false;
7443
7444 net->ipv6.devconf_all = all;
7445 net->ipv6.devconf_dflt = dflt;
7446
7447#ifdef CONFIG_SYSCTL
7448 err = __addrconf_sysctl_register(net, "all", NULL, all);
7449 if (err < 0)
7450 goto err_reg_all;
7451
7452 err = __addrconf_sysctl_register(net, "default", NULL, dflt);
7453 if (err < 0)
7454 goto err_reg_dflt;
7455#endif
7456 return 0;
7457
7458#ifdef CONFIG_SYSCTL
7459err_reg_dflt:
7460 __addrconf_sysctl_unregister(net, all, NETCONFA_IFINDEX_ALL);
7461err_reg_all:
7462 kfree(dflt);
7463 net->ipv6.devconf_dflt = NULL;
7464#endif
7465err_alloc_dflt:
7466 kfree(all);
7467 net->ipv6.devconf_all = NULL;
7468err_alloc_all:
7469 kfree(net->ipv6.inet6_addr_lst);
7470err_alloc_addr:
7471 return err;
7472}
7473
7474static void __net_exit addrconf_exit_net(struct net *net)
7475{
7476 int i;
7477
7478#ifdef CONFIG_SYSCTL
7479 __addrconf_sysctl_unregister(net, net->ipv6.devconf_dflt,
7480 NETCONFA_IFINDEX_DEFAULT);
7481 __addrconf_sysctl_unregister(net, net->ipv6.devconf_all,
7482 NETCONFA_IFINDEX_ALL);
7483#endif
7484 kfree(net->ipv6.devconf_dflt);
7485 net->ipv6.devconf_dflt = NULL;
7486 kfree(net->ipv6.devconf_all);
7487 net->ipv6.devconf_all = NULL;
7488
7489 cancel_delayed_work_sync(&net->ipv6.addr_chk_work);
7490 /*
7491 * Check hash table, then free it.
7492 */
7493 for (i = 0; i < IN6_ADDR_HSIZE; i++)
7494 WARN_ON_ONCE(!hlist_empty(&net->ipv6.inet6_addr_lst[i]));
7495
7496 kfree(net->ipv6.inet6_addr_lst);
7497 net->ipv6.inet6_addr_lst = NULL;
7498}
7499
7500static struct pernet_operations addrconf_ops = {
7501 .init = addrconf_init_net,
7502 .exit = addrconf_exit_net,
7503};
7504
7505static struct rtnl_af_ops inet6_ops __read_mostly = {
7506 .family = AF_INET6,
7507 .fill_link_af = inet6_fill_link_af,
7508 .get_link_af_size = inet6_get_link_af_size,
7509 .validate_link_af = inet6_validate_link_af,
7510 .set_link_af = inet6_set_link_af,
7511};
7512
7513static const struct rtnl_msg_handler addrconf_rtnl_msg_handlers[] __initconst_or_module = {
7514 {.owner = THIS_MODULE, .protocol = PF_INET6, .msgtype = RTM_GETLINK,
7515 .dumpit = inet6_dump_ifinfo, .flags = RTNL_FLAG_DUMP_UNLOCKED},
7516 {.owner = THIS_MODULE, .protocol = PF_INET6, .msgtype = RTM_NEWADDR,
7517 .doit = inet6_rtm_newaddr, .flags = RTNL_FLAG_DOIT_PERNET},
7518 {.owner = THIS_MODULE, .protocol = PF_INET6, .msgtype = RTM_DELADDR,
7519 .doit = inet6_rtm_deladdr, .flags = RTNL_FLAG_DOIT_PERNET},
7520 {.owner = THIS_MODULE, .protocol = PF_INET6, .msgtype = RTM_GETADDR,
7521 .doit = inet6_rtm_getaddr, .dumpit = inet6_dump_ifaddr,
7522 .flags = RTNL_FLAG_DOIT_UNLOCKED | RTNL_FLAG_DUMP_UNLOCKED},
7523 {.owner = THIS_MODULE, .protocol = PF_INET6, .msgtype = RTM_GETMULTICAST,
7524 .dumpit = inet6_dump_ifmcaddr,
7525 .flags = RTNL_FLAG_DUMP_UNLOCKED},
7526 {.owner = THIS_MODULE, .protocol = PF_INET6, .msgtype = RTM_GETANYCAST,
7527 .dumpit = inet6_dump_ifacaddr,
7528 .flags = RTNL_FLAG_DUMP_UNLOCKED},
7529 {.owner = THIS_MODULE, .protocol = PF_INET6, .msgtype = RTM_GETNETCONF,
7530 .doit = inet6_netconf_get_devconf, .dumpit = inet6_netconf_dump_devconf,
7531 .flags = RTNL_FLAG_DOIT_UNLOCKED | RTNL_FLAG_DUMP_UNLOCKED},
7532};
7533
7534/*
7535 * Init / cleanup code
7536 */
7537
7538int __init addrconf_init(void)
7539{
7540 struct inet6_dev *idev;
7541 int err;
7542
7543 err = ipv6_addr_label_init();
7544 if (err < 0) {
7545 pr_crit("%s: cannot initialize default policy table: %d\n",
7546 __func__, err);
7547 goto out;
7548 }
7549
7550 err = register_pernet_subsys(&addrconf_ops);
7551 if (err < 0)
7552 goto out_addrlabel;
7553
7554 /* All works using addrconf_wq need to lock rtnl. */
7555 addrconf_wq = create_singlethread_workqueue("ipv6_addrconf");
7556 if (!addrconf_wq) {
7557 err = -ENOMEM;
7558 goto out_nowq;
7559 }
7560
7561 rtnl_net_lock(&init_net);
7562 idev = ipv6_add_dev(blackhole_netdev);
7563 rtnl_net_unlock(&init_net);
7564 if (IS_ERR(idev)) {
7565 err = PTR_ERR(idev);
7566 goto errlo;
7567 }
7568
7569 ip6_route_init_special_entries();
7570
7571 register_netdevice_notifier(&ipv6_dev_notf);
7572
7573 addrconf_verify(&init_net);
7574
7575 err = rtnl_af_register(&inet6_ops);
7576 if (err)
7577 goto erraf;
7578
7579 err = rtnl_register_many(addrconf_rtnl_msg_handlers);
7580 if (err)
7581 goto errout;
7582
7583 err = ipv6_addr_label_rtnl_register();
7584 if (err < 0)
7585 goto errout;
7586
7587 return 0;
7588errout:
7589 rtnl_unregister_all(PF_INET6);
7590 rtnl_af_unregister(&inet6_ops);
7591erraf:
7592 unregister_netdevice_notifier(&ipv6_dev_notf);
7593errlo:
7594 destroy_workqueue(addrconf_wq);
7595out_nowq:
7596 unregister_pernet_subsys(&addrconf_ops);
7597out_addrlabel:
7598 ipv6_addr_label_cleanup();
7599out:
7600 return err;
7601}
7602
7603void addrconf_cleanup(void)
7604{
7605 struct net_device *dev;
7606
7607 unregister_netdevice_notifier(&ipv6_dev_notf);
7608 unregister_pernet_subsys(&addrconf_ops);
7609 ipv6_addr_label_cleanup();
7610
7611 rtnl_af_unregister(&inet6_ops);
7612
7613 rtnl_net_lock(&init_net);
7614
7615 /* clean dev list */
7616 for_each_netdev(&init_net, dev) {
7617 if (!__in6_dev_get_rtnl_net(dev))
7618 continue;
7619 addrconf_ifdown(dev, true);
7620 }
7621 addrconf_ifdown(init_net.loopback_dev, true);
7622
7623 rtnl_net_unlock(&init_net);
7624
7625 destroy_workqueue(addrconf_wq);
7626}