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 <crypto/sha1.h>
40 #include <linux/errno.h>
41 #include <linux/types.h>
42 #include <linux/kernel.h>
43 #include <linux/sched/signal.h>
44 #include <linux/socket.h>
45 #include <linux/sockios.h>
46 #include <linux/net.h>
47 #include <linux/inet.h>
48 #include <linux/in6.h>
49 #include <linux/netdevice.h>
50 #include <linux/if_addr.h>
51 #include <linux/if_arp.h>
52 #include <linux/if_arcnet.h>
53 #include <linux/if_infiniband.h>
54 #include <linux/route.h>
55 #include <linux/inetdevice.h>
56 #include <linux/init.h>
57 #include <linux/slab.h>
58 #ifdef CONFIG_SYSCTL
59 #include <linux/sysctl.h>
60 #endif
61 #include <linux/capability.h>
62 #include <linux/delay.h>
63 #include <linux/notifier.h>
64 #include <linux/string.h>
65 #include <linux/hash.h>
66
67 #include <net/ip_tunnels.h>
68 #include <net/net_namespace.h>
69 #include <net/sock.h>
70 #include <net/snmp.h>
71
72 #include <net/6lowpan.h>
73 #include <net/firewire.h>
74 #include <net/ipv6.h>
75 #include <net/protocol.h>
76 #include <net/ndisc.h>
77 #include <net/ip6_route.h>
78 #include <net/addrconf.h>
79 #include <net/tcp.h>
80 #include <net/ip.h>
81 #include <net/netlink.h>
82 #include <net/pkt_sched.h>
83 #include <net/l3mdev.h>
84 #include <net/netdev_lock.h>
85 #include <linux/if_tunnel.h>
86 #include <linux/rtnetlink.h>
87 #include <linux/netconf.h>
88 #include <linux/random.h>
89 #include <linux/uaccess.h>
90 #include <linux/unaligned.h>
91
92 #include <linux/proc_fs.h>
93 #include <linux/seq_file.h>
94 #include <linux/export.h>
95 #include <linux/ioam6.h>
96
97 #define IPV6_MAX_STRLEN \
98 sizeof("ffff:ffff:ffff:ffff:ffff:ffff:255.255.255.255")
99
cstamp_delta(unsigned long cstamp)100 static inline u32 cstamp_delta(unsigned long cstamp)
101 {
102 return (cstamp - INITIAL_JIFFIES) * 100UL / HZ;
103 }
104
rfc3315_s14_backoff_init(s32 irt)105 static inline s32 rfc3315_s14_backoff_init(s32 irt)
106 {
107 /* multiply 'initial retransmission time' by 0.9 .. 1.1 */
108 u64 tmp = get_random_u32_inclusive(900000, 1100000) * (u64)irt;
109 do_div(tmp, 1000000);
110 return (s32)tmp;
111 }
112
rfc3315_s14_backoff_update(s32 rt,s32 mrt)113 static inline s32 rfc3315_s14_backoff_update(s32 rt, s32 mrt)
114 {
115 /* multiply 'retransmission timeout' by 1.9 .. 2.1 */
116 u64 tmp = get_random_u32_inclusive(1900000, 2100000) * (u64)rt;
117 do_div(tmp, 1000000);
118 if ((s32)tmp > mrt) {
119 /* multiply 'maximum retransmission time' by 0.9 .. 1.1 */
120 tmp = get_random_u32_inclusive(900000, 1100000) * (u64)mrt;
121 do_div(tmp, 1000000);
122 }
123 return (s32)tmp;
124 }
125
126 #ifdef CONFIG_SYSCTL
127 static int addrconf_sysctl_register(struct inet6_dev *idev);
128 static void addrconf_sysctl_unregister(struct inet6_dev *idev);
129 #else
addrconf_sysctl_register(struct inet6_dev * idev)130 static inline int addrconf_sysctl_register(struct inet6_dev *idev)
131 {
132 return 0;
133 }
134
addrconf_sysctl_unregister(struct inet6_dev * idev)135 static inline void addrconf_sysctl_unregister(struct inet6_dev *idev)
136 {
137 }
138 #endif
139
140 static void ipv6_gen_rnd_iid(struct in6_addr *addr);
141
142 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev);
143 static int ipv6_count_addresses(const struct inet6_dev *idev);
144 static int ipv6_generate_stable_address(struct in6_addr *addr,
145 u8 dad_count,
146 const struct inet6_dev *idev);
147
148 #define IN6_ADDR_HSIZE_SHIFT 8
149 #define IN6_ADDR_HSIZE (1 << IN6_ADDR_HSIZE_SHIFT)
150
151 static void addrconf_verify(struct net *net);
152 static void addrconf_verify_rtnl(struct net *net);
153
154 static struct workqueue_struct *addrconf_wq;
155
156 static void addrconf_join_anycast(struct inet6_ifaddr *ifp);
157 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp);
158
159 static void addrconf_type_change(struct net_device *dev,
160 unsigned long event);
161 static int addrconf_ifdown(struct net_device *dev, bool unregister);
162
163 static struct fib6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
164 int plen,
165 const struct net_device *dev,
166 u32 flags, u32 noflags,
167 bool no_gw);
168
169 static void addrconf_dad_start(struct inet6_ifaddr *ifp);
170 static void addrconf_dad_work(struct work_struct *w);
171 static void addrconf_dad_completed(struct inet6_ifaddr *ifp, bool bump_id,
172 bool send_na);
173 static void addrconf_dad_run(struct inet6_dev *idev, bool restart);
174 static void addrconf_rs_timer(struct timer_list *t);
175 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
176 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
177
178 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
179 struct prefix_info *pinfo);
180
181 static struct ipv6_devconf ipv6_devconf __read_mostly = {
182 .forwarding = 0,
183 .hop_limit = IPV6_DEFAULT_HOPLIMIT,
184 .mtu6 = IPV6_MIN_MTU,
185 .accept_ra = 1,
186 .accept_redirects = 1,
187 .autoconf = 1,
188 .force_mld_version = 0,
189 .mldv1_unsolicited_report_interval = 10 * HZ,
190 .mldv2_unsolicited_report_interval = HZ,
191 .dad_transmits = 1,
192 .rtr_solicits = MAX_RTR_SOLICITATIONS,
193 .rtr_solicit_interval = RTR_SOLICITATION_INTERVAL,
194 .rtr_solicit_max_interval = RTR_SOLICITATION_MAX_INTERVAL,
195 .rtr_solicit_delay = MAX_RTR_SOLICITATION_DELAY,
196 .use_tempaddr = 0,
197 .temp_valid_lft = TEMP_VALID_LIFETIME,
198 .temp_prefered_lft = TEMP_PREFERRED_LIFETIME,
199 .regen_min_advance = REGEN_MIN_ADVANCE,
200 .regen_max_retry = REGEN_MAX_RETRY,
201 .max_desync_factor = MAX_DESYNC_FACTOR,
202 .max_addresses = IPV6_MAX_ADDRESSES,
203 .accept_ra_defrtr = 1,
204 .ra_defrtr_metric = IP6_RT_PRIO_USER,
205 .accept_ra_from_local = 0,
206 .accept_ra_min_hop_limit= 1,
207 .accept_ra_min_lft = 0,
208 .accept_ra_pinfo = 1,
209 #ifdef CONFIG_IPV6_ROUTER_PREF
210 .accept_ra_rtr_pref = 1,
211 .rtr_probe_interval = 60 * HZ,
212 #ifdef CONFIG_IPV6_ROUTE_INFO
213 .accept_ra_rt_info_min_plen = 0,
214 .accept_ra_rt_info_max_plen = 0,
215 #endif
216 #endif
217 .proxy_ndp = 0,
218 .accept_source_route = 0, /* we do not accept RH0 by default. */
219 .disable_ipv6 = 0,
220 .accept_dad = 0,
221 .suppress_frag_ndisc = 1,
222 .accept_ra_mtu = 1,
223 .stable_secret = {
224 .initialized = false,
225 },
226 .use_oif_addrs_only = 0,
227 .ignore_routes_with_linkdown = 0,
228 .keep_addr_on_down = 0,
229 .seg6_enabled = 0,
230 #ifdef CONFIG_IPV6_SEG6_HMAC
231 .seg6_require_hmac = 0,
232 #endif
233 .enhanced_dad = 1,
234 .addr_gen_mode = IN6_ADDR_GEN_MODE_EUI64,
235 .disable_policy = 0,
236 .rpl_seg_enabled = 0,
237 .ioam6_enabled = 0,
238 .ioam6_id = IOAM6_DEFAULT_IF_ID,
239 .ioam6_id_wide = IOAM6_DEFAULT_IF_ID_WIDE,
240 .ndisc_evict_nocarrier = 1,
241 .ra_honor_pio_life = 0,
242 .ra_honor_pio_pflag = 0,
243 .force_forwarding = 0,
244 };
245
246 static struct ipv6_devconf ipv6_devconf_dflt __read_mostly = {
247 .forwarding = 0,
248 .hop_limit = IPV6_DEFAULT_HOPLIMIT,
249 .mtu6 = IPV6_MIN_MTU,
250 .accept_ra = 1,
251 .accept_redirects = 1,
252 .autoconf = 1,
253 .force_mld_version = 0,
254 .mldv1_unsolicited_report_interval = 10 * HZ,
255 .mldv2_unsolicited_report_interval = HZ,
256 .dad_transmits = 1,
257 .rtr_solicits = MAX_RTR_SOLICITATIONS,
258 .rtr_solicit_interval = RTR_SOLICITATION_INTERVAL,
259 .rtr_solicit_max_interval = RTR_SOLICITATION_MAX_INTERVAL,
260 .rtr_solicit_delay = MAX_RTR_SOLICITATION_DELAY,
261 .use_tempaddr = 0,
262 .temp_valid_lft = TEMP_VALID_LIFETIME,
263 .temp_prefered_lft = TEMP_PREFERRED_LIFETIME,
264 .regen_min_advance = REGEN_MIN_ADVANCE,
265 .regen_max_retry = REGEN_MAX_RETRY,
266 .max_desync_factor = MAX_DESYNC_FACTOR,
267 .max_addresses = IPV6_MAX_ADDRESSES,
268 .accept_ra_defrtr = 1,
269 .ra_defrtr_metric = IP6_RT_PRIO_USER,
270 .accept_ra_from_local = 0,
271 .accept_ra_min_hop_limit= 1,
272 .accept_ra_min_lft = 0,
273 .accept_ra_pinfo = 1,
274 #ifdef CONFIG_IPV6_ROUTER_PREF
275 .accept_ra_rtr_pref = 1,
276 .rtr_probe_interval = 60 * HZ,
277 #ifdef CONFIG_IPV6_ROUTE_INFO
278 .accept_ra_rt_info_min_plen = 0,
279 .accept_ra_rt_info_max_plen = 0,
280 #endif
281 #endif
282 .proxy_ndp = 0,
283 .accept_source_route = 0, /* we do not accept RH0 by default. */
284 .disable_ipv6 = 0,
285 .accept_dad = 1,
286 .suppress_frag_ndisc = 1,
287 .accept_ra_mtu = 1,
288 .stable_secret = {
289 .initialized = false,
290 },
291 .use_oif_addrs_only = 0,
292 .ignore_routes_with_linkdown = 0,
293 .keep_addr_on_down = 0,
294 .seg6_enabled = 0,
295 #ifdef CONFIG_IPV6_SEG6_HMAC
296 .seg6_require_hmac = 0,
297 #endif
298 .enhanced_dad = 1,
299 .addr_gen_mode = IN6_ADDR_GEN_MODE_EUI64,
300 .disable_policy = 0,
301 .rpl_seg_enabled = 0,
302 .ioam6_enabled = 0,
303 .ioam6_id = IOAM6_DEFAULT_IF_ID,
304 .ioam6_id_wide = IOAM6_DEFAULT_IF_ID_WIDE,
305 .ndisc_evict_nocarrier = 1,
306 .ra_honor_pio_life = 0,
307 .ra_honor_pio_pflag = 0,
308 .force_forwarding = 0,
309 };
310
311 /* Check if link is ready: is it up and is a valid qdisc available */
addrconf_link_ready(const struct net_device * dev)312 static inline bool addrconf_link_ready(const struct net_device *dev)
313 {
314 return netif_oper_up(dev) && !qdisc_tx_is_noop(dev);
315 }
316
addrconf_del_rs_timer(struct inet6_dev * idev)317 static void addrconf_del_rs_timer(struct inet6_dev *idev)
318 {
319 if (timer_delete(&idev->rs_timer))
320 __in6_dev_put(idev);
321 }
322
addrconf_del_dad_work(struct inet6_ifaddr * ifp)323 static void addrconf_del_dad_work(struct inet6_ifaddr *ifp)
324 {
325 if (cancel_delayed_work(&ifp->dad_work))
326 __in6_ifa_put(ifp);
327 }
328
addrconf_mod_rs_timer(struct inet6_dev * idev,unsigned long when)329 static void addrconf_mod_rs_timer(struct inet6_dev *idev,
330 unsigned long when)
331 {
332 if (!mod_timer(&idev->rs_timer, jiffies + when))
333 in6_dev_hold(idev);
334 }
335
addrconf_mod_dad_work(struct inet6_ifaddr * ifp,unsigned long delay)336 static void addrconf_mod_dad_work(struct inet6_ifaddr *ifp,
337 unsigned long delay)
338 {
339 in6_ifa_hold(ifp);
340 if (mod_delayed_work(addrconf_wq, &ifp->dad_work, delay))
341 in6_ifa_put(ifp);
342 }
343
snmp6_alloc_dev(struct inet6_dev * idev)344 static int snmp6_alloc_dev(struct inet6_dev *idev)
345 {
346 int i;
347
348 idev->stats.ipv6 = alloc_percpu_gfp(struct ipstats_mib, GFP_KERNEL_ACCOUNT);
349 if (!idev->stats.ipv6)
350 goto err_ip;
351
352 for_each_possible_cpu(i) {
353 struct ipstats_mib *addrconf_stats;
354 addrconf_stats = per_cpu_ptr(idev->stats.ipv6, i);
355 u64_stats_init(&addrconf_stats->syncp);
356 }
357
358
359 idev->stats.icmpv6dev = kzalloc_obj(struct icmpv6_mib_device);
360 if (!idev->stats.icmpv6dev)
361 goto err_icmp;
362 idev->stats.icmpv6msgdev = kzalloc_obj(struct icmpv6msg_mib_device,
363 GFP_KERNEL_ACCOUNT);
364 if (!idev->stats.icmpv6msgdev)
365 goto err_icmpmsg;
366
367 return 0;
368
369 err_icmpmsg:
370 kfree(idev->stats.icmpv6dev);
371 err_icmp:
372 free_percpu(idev->stats.ipv6);
373 err_ip:
374 return -ENOMEM;
375 }
376
ipv6_add_dev(struct net_device * dev)377 static 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_obj(*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
483 err_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
ipv6_find_idev(struct net_device * dev)490 static 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
inet6_netconf_msgsize_devconf(int type)508 static 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
inet6_netconf_fill_devconf(struct sk_buff * skb,int ifindex,struct ipv6_devconf * devconf,u32 portid,u32 seq,int event,unsigned int flags,int type)532 static 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
578 out:
579 nlmsg_end(skb, nlh);
580 return 0;
581
582 nla_put_failure:
583 nlmsg_cancel(skb, nlh);
584 return -EMSGSIZE;
585 }
586
inet6_netconf_notify_devconf(struct net * net,int event,int type,int ifindex,struct ipv6_devconf * devconf)587 void 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;
607 errout:
608 rtnl_set_sk_err(net, RTNLGRP_IPV6_NETCONF, err);
609 }
610
611 static 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
inet6_netconf_valid_get_req(struct sk_buff * skb,const struct nlmsghdr * nlh,struct nlattr ** tb,struct netlink_ext_ack * extack)618 static 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
inet6_netconf_get_devconf(struct sk_buff * in_skb,struct nlmsghdr * nlh,struct netlink_ext_ack * extack)657 static 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);
713 errout:
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 */
inet6_base_seq(const struct net * net)722 static 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
inet6_netconf_dump_devconf(struct sk_buff * skb,struct netlink_callback * cb)735 static 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 }
800 done:
801 rcu_read_unlock();
802 return err;
803 }
804
805 #ifdef CONFIG_SYSCTL
dev_forward_change(struct inet6_dev * idev)806 static 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
addrconf_forward_change(struct net * net,__s32 newf)853 static 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
addrconf_fixup_forwarding(const struct ctl_table * table,int * p,int newf)873 static 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
addrconf_linkdown_change(struct net * net,__s32 newf)919 static 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
addrconf_fixup_linkdown(const struct ctl_table * table,int * p,int newf)940 static 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 */
inet6_ifa_finish_destroy(struct inet6_ifaddr * ifp)981 void 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
1003 static void
ipv6_link_dev_addr(struct inet6_dev * idev,struct inet6_ifaddr * ifp)1004 ipv6_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
inet6_addr_hash(const struct net * net,const struct in6_addr * addr)1023 static 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
ipv6_chk_same_addr(struct net * net,const struct in6_addr * addr,struct net_device * dev,unsigned int hash)1030 static 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
ipv6_add_addr_hash(struct net_device * dev,struct inet6_ifaddr * ifa)1044 static 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
1067 static struct inet6_ifaddr *
ipv6_add_addr(struct inet6_dev * idev,struct ifa6_config * cfg,bool can_block,struct netlink_ext_ack * extack)1068 ipv6_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_obj(*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);
1187 out:
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
1202 enum 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 **/
1226 static enum cleanup_prefix_rt_t
check_cleanup_prefix_route(struct inet6_ifaddr * ifp,unsigned long * expires)1227 check_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
1264 static void
cleanup_prefix_route(struct inet6_ifaddr * ifp,unsigned long expires,bool del_rt,bool del_peer)1265 cleanup_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
ipv6_del_addr(struct inet6_ifaddr * ifp)1295 static 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);
1348 out:
1349 in6_ifa_put(ifp);
1350 }
1351
ipv6_get_regen_advance(const struct inet6_dev * idev)1352 static 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
ipv6_create_tempaddr(struct inet6_ifaddr * ifp,bool block)1360 static 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
1376 retry:
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);
1500 out:
1501 return ret;
1502 }
1503
1504 /*
1505 * Choose an appropriate source address (RFC3484)
1506 */
1507 enum {
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
1526 struct 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
1535 struct ipv6_saddr_dst {
1536 const struct in6_addr *addr;
1537 int ifindex;
1538 int scope;
1539 int label;
1540 unsigned int prefs;
1541 };
1542
ipv6_saddr_preferred(int type)1543 static 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
ipv6_use_optimistic_addr(const struct net * net,const struct inet6_dev * idev)1550 static 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
ipv6_allow_optimistic_dad(const struct net * net,const struct inet6_dev * idev)1569 static 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
ipv6_get_saddr_eval(struct net * net,struct ipv6_saddr_score * score,struct ipv6_saddr_dst * dst,int i)1585 static 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;
1715 out:
1716 return ret;
1717 }
1718
__ipv6_dev_get_saddr(struct net * net,struct ipv6_saddr_dst * dst,struct inet6_dev * idev,struct ipv6_saddr_score * scores,int hiscore_idx)1719 static 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 }
1787 out:
1788 return hiscore_idx;
1789 }
1790
ipv6_get_saddr_master(struct net * net,const struct net_device * dst_dev,const struct net_device * master,struct ipv6_saddr_dst * dst,struct ipv6_saddr_score * scores,int hiscore_idx)1791 static 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
ipv6_dev_get_saddr(struct net * net,const struct net_device * dst_dev,const struct in6_addr * daddr,unsigned int prefs,struct in6_addr * saddr)1813 int 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
1900 out:
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 }
1910 EXPORT_SYMBOL(ipv6_dev_get_saddr);
1911
__ipv6_get_lladdr(struct inet6_dev * idev,struct in6_addr * addr,u32 banned_flags)1912 static 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
ipv6_get_lladdr(struct net_device * dev,struct in6_addr * addr,u32 banned_flags)1931 int 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
ipv6_count_addresses(const struct inet6_dev * idev)1948 static 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
ipv6_chk_addr(struct net * net,const struct in6_addr * addr,const struct net_device * dev,int strict)1960 int 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 }
1966 EXPORT_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 */
1977 static struct net_device *
__ipv6_chk_addr_and_flags(struct net * net,const struct in6_addr * addr,const struct net_device * dev,bool skip_dev_check,int strict,u32 banned_flags)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
ipv6_chk_addr_and_flags(struct net * net,const struct in6_addr * addr,const struct net_device * dev,bool skip_dev_check,int strict,u32 banned_flags)2018 int 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 }
2025 EXPORT_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 */
ipv6_chk_custom_prefix(const struct in6_addr * addr,const unsigned int prefix_len,struct net_device * dev)2031 bool 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 }
2051 EXPORT_SYMBOL(ipv6_chk_custom_prefix);
2052
ipv6_chk_prefix(const struct in6_addr * addr,struct net_device * dev)2053 int 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 }
2073 EXPORT_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 */
ipv6_dev_find(struct net * net,const struct in6_addr * addr,struct net_device * dev)2083 struct 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 }
2089 EXPORT_SYMBOL(ipv6_dev_find);
2090
ipv6_get_ifaddr(struct net * net,const struct in6_addr * addr,struct net_device * dev,int strict)2091 struct 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
addrconf_dad_stop(struct inet6_ifaddr * ifp,int dad_failed)2116 static 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
addrconf_dad_end(struct inet6_ifaddr * ifp)2149 static 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
addrconf_dad_failure(struct sk_buff * skb,struct inet6_ifaddr * ifp)2163 void 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);
2224 lock_errdad:
2225 spin_lock_bh(&ifp->lock);
2226 }
2227
2228 errdad:
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. */
addrconf_join_solict(struct net_device * dev,const struct in6_addr * addr)2238 void 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
addrconf_leave_solict(struct inet6_dev * idev,const struct in6_addr * addr)2249 void 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
addrconf_join_anycast(struct inet6_ifaddr * ifp)2260 static 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
addrconf_leave_anycast(struct inet6_ifaddr * ifp)2272 static 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
addrconf_ifid_6lowpan(u8 * eui,struct net_device * dev)2284 static 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
addrconf_ifid_ieee1394(u8 * eui,struct net_device * dev)2304 static 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
addrconf_ifid_arcnet(u8 * eui,struct net_device * dev)2318 static 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
addrconf_ifid_infiniband(u8 * eui,struct net_device * dev)2328 static 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
__ipv6_isatap_ifid(u8 * eui,__be32 addr)2337 static 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
addrconf_ifid_sit(u8 * eui,struct net_device * dev)2354 static 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
addrconf_ifid_gre(u8 * eui,struct net_device * dev)2361 static int addrconf_ifid_gre(u8 *eui, struct net_device *dev)
2362 {
2363 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
2364 }
2365
addrconf_ifid_ip6tnl(u8 * eui,struct net_device * dev)2366 static 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
ipv6_generate_eui64(u8 * eui,struct net_device * dev)2376 static 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
ipv6_inherit_eui64(u8 * eui,struct inet6_dev * idev)2403 static 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
ipv6_gen_rnd_iid(struct in6_addr * addr)2426 static void ipv6_gen_rnd_iid(struct in6_addr *addr)
2427 {
2428 regen:
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
2460 static void
addrconf_prefix_route(struct in6_addr * pfx,int plen,u32 metric,struct net_device * dev,unsigned long expires,u32 flags,gfp_t gfp_flags)2461 addrconf_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
addrconf_get_prefix_route(const struct in6_addr * pfx,int plen,const struct net_device * dev,u32 flags,u32 noflags,bool no_gw)2492 static 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 }
2529 out:
2530 rcu_read_unlock();
2531 return rt;
2532 }
2533
2534
2535 /* Create "default" multicast route to the interface */
2536
addrconf_add_mroute(struct net_device * dev)2537 static 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
addrconf_add_dev(struct net_device * dev)2555 static 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
delete_tempaddrs(struct inet6_dev * idev,struct inet6_ifaddr * ifp)2575 static 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
manage_tempaddrs(struct inet6_dev * idev,struct inet6_ifaddr * ifp,__u32 valid_lft,__u32 prefered_lft,bool create,unsigned long now)2593 static 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
is_addr_mode_generate_stable(struct inet6_dev * idev)2666 static 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
addrconf_prefix_rcv_add_addr(struct net * net,struct net_device * dev,const struct prefix_info * pinfo,struct inet6_dev * in6_dev,const struct in6_addr * addr,int addr_type,u32 addr_flags,bool sllao,bool tokenized,__u32 valid_lft,u32 prefered_lft)2672 int 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 }
2773 EXPORT_SYMBOL_GPL(addrconf_prefix_rcv_add_addr);
2774
addrconf_prefix_rcv(struct net_device * dev,u8 * opt,int len,bool sllao)2775 void 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_may_remove_gc_list(net, 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
2919 ok:
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);
2938 put:
2939 in6_dev_put(in6_dev);
2940 }
2941
addrconf_set_sit_dstaddr(struct net * net,struct net_device * dev,struct in6_ifreq * ireq)2942 static 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 */
addrconf_set_dstaddr(struct net * net,void __user * arg)2974 int 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
ipv6_mc_config(struct sock * sk,bool join,const struct in6_addr * addr,int ifindex)2993 static 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 */
inet6_addr_add(struct net * net,struct net_device * dev,struct ifa6_config * cfg,clock_t expires,u32 flags,struct netlink_ext_ack * extack)3013 static 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
inet6_addr_del(struct net * net,int ifindex,u32 ifa_flags,const struct in6_addr * pfx,unsigned int plen,struct netlink_ext_ack * extack)3083 static 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
addrconf_add_ifaddr(struct net * net,void __user * arg)3136 int 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
addrconf_del_ifaddr(struct net * net,void __user * arg)3169 int 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
add_addr(struct inet6_dev * idev,const struct in6_addr * addr,int plen,int scope,u8 proto)3187 static 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)
add_v4_addrs(struct inet6_dev * idev)3213 static 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
init_loopback(struct net_device * dev)3273 static 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
addrconf_add_linklocal(struct inet6_dev * idev,const struct in6_addr * addr,u32 flags)3290 void 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 }
3319 EXPORT_SYMBOL_GPL(addrconf_add_linklocal);
3320
ipv6_reserved_interfaceid(struct in6_addr address)3321 static 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
ipv6_generate_stable_address(struct in6_addr * address,u8 dad_count,const struct inet6_dev * idev)3337 static 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
3367 retry:
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
ipv6_gen_mode_random_init(struct inet6_dev * idev)3404 static 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
addrconf_addr_gen(struct inet6_dev * idev,bool prefix_route)3415 static 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
addrconf_dev_config(struct net_device * dev)3459 static 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)
addrconf_sit_config(struct net_device * dev)3498 static 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)
addrconf_gre_config(struct net_device * dev)3529 static 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
addrconf_init_auto_addrs(struct net_device * dev)3554 static 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
fixup_permanent_addr(struct net * net,struct inet6_dev * idev,struct inet6_ifaddr * ifp)3577 static 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_KERNEL, NULL);
3590 if (IS_ERR(f6i))
3591 return PTR_ERR(f6i);
3592
3593 /* ifp->rt can be accessed outside of rtnl */
3594 spin_lock_bh(&ifp->lock);
3595 prev = ifp->rt;
3596 ifp->rt = f6i;
3597 spin_unlock_bh(&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_KERNEL);
3606 }
3607
3608 if (ifp->state == INET6_IFADDR_STATE_PREDAD)
3609 addrconf_dad_start(ifp);
3610
3611 return 0;
3612 }
3613
addrconf_permanent_addr(struct net * net,struct net_device * dev)3614 static void addrconf_permanent_addr(struct net *net, struct net_device *dev)
3615 {
3616 struct inet6_ifaddr *ifp;
3617 LIST_HEAD(tmp_addr_list);
3618 struct inet6_dev *idev;
3619
3620 /* Mutual exclusion with other if_list_aux users. */
3621 ASSERT_RTNL();
3622
3623 idev = __in6_dev_get(dev);
3624 if (!idev)
3625 return;
3626
3627 write_lock_bh(&idev->lock);
3628 list_for_each_entry(ifp, &idev->addr_list, if_list) {
3629 if (ifp->flags & IFA_F_PERMANENT)
3630 list_add_tail(&ifp->if_list_aux, &tmp_addr_list);
3631 }
3632 write_unlock_bh(&idev->lock);
3633
3634 while (!list_empty(&tmp_addr_list)) {
3635 ifp = list_first_entry(&tmp_addr_list,
3636 struct inet6_ifaddr, if_list_aux);
3637 list_del(&ifp->if_list_aux);
3638
3639 if (fixup_permanent_addr(net, idev, ifp) < 0) {
3640 net_info_ratelimited("%s: Failed to add prefix route for address %pI6c; dropping\n",
3641 idev->dev->name, &ifp->addr);
3642 in6_ifa_hold(ifp);
3643 ipv6_del_addr(ifp);
3644 }
3645 }
3646 }
3647
addrconf_notify(struct notifier_block * this,unsigned long event,void * ptr)3648 static int addrconf_notify(struct notifier_block *this, unsigned long event,
3649 void *ptr)
3650 {
3651 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
3652 struct netdev_notifier_change_info *change_info;
3653 struct netdev_notifier_changeupper_info *info;
3654 struct inet6_dev *idev = __in6_dev_get(dev);
3655 struct net *net = dev_net(dev);
3656 int run_pending = 0;
3657 int err;
3658
3659 switch (event) {
3660 case NETDEV_REGISTER:
3661 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
3662 idev = ipv6_add_dev(dev);
3663 if (IS_ERR(idev))
3664 return notifier_from_errno(PTR_ERR(idev));
3665 }
3666 break;
3667
3668 case NETDEV_CHANGEMTU:
3669 /* if MTU under IPV6_MIN_MTU stop IPv6 on this interface. */
3670 if (dev->mtu < IPV6_MIN_MTU) {
3671 addrconf_ifdown(dev, dev != net->loopback_dev);
3672 break;
3673 }
3674
3675 if (idev) {
3676 rt6_mtu_change(dev, dev->mtu);
3677 WRITE_ONCE(idev->cnf.mtu6, dev->mtu);
3678 break;
3679 }
3680
3681 /* allocate new idev */
3682 idev = ipv6_add_dev(dev);
3683 if (IS_ERR(idev))
3684 break;
3685
3686 /* device is still not ready */
3687 if (!(idev->if_flags & IF_READY))
3688 break;
3689
3690 run_pending = 1;
3691 fallthrough;
3692 case NETDEV_UP:
3693 case NETDEV_CHANGE:
3694 if (idev && idev->cnf.disable_ipv6)
3695 break;
3696
3697 if (dev->priv_flags & IFF_NO_ADDRCONF) {
3698 if (event == NETDEV_UP && !IS_ERR_OR_NULL(idev) &&
3699 dev->flags & IFF_UP && dev->flags & IFF_MULTICAST)
3700 ipv6_mc_up(idev);
3701 break;
3702 }
3703
3704 if (event == NETDEV_UP) {
3705 /* restore routes for permanent addresses */
3706 addrconf_permanent_addr(net, dev);
3707
3708 if (!addrconf_link_ready(dev)) {
3709 /* device is not ready yet. */
3710 pr_debug("ADDRCONF(NETDEV_UP): %s: link is not ready\n",
3711 dev->name);
3712 break;
3713 }
3714
3715 if (!idev && dev->mtu >= IPV6_MIN_MTU)
3716 idev = ipv6_add_dev(dev);
3717
3718 if (!IS_ERR_OR_NULL(idev)) {
3719 idev->if_flags |= IF_READY;
3720 run_pending = 1;
3721 }
3722 } else if (event == NETDEV_CHANGE) {
3723 if (!addrconf_link_ready(dev)) {
3724 /* device is still not ready. */
3725 rt6_sync_down_dev(dev, event);
3726 break;
3727 }
3728
3729 if (!IS_ERR_OR_NULL(idev)) {
3730 if (idev->if_flags & IF_READY) {
3731 /* device is already configured -
3732 * but resend MLD reports, we might
3733 * have roamed and need to update
3734 * multicast snooping switches
3735 */
3736 ipv6_mc_up(idev);
3737 change_info = ptr;
3738 if (change_info->flags_changed & IFF_NOARP)
3739 addrconf_dad_run(idev, true);
3740 rt6_sync_up(dev, RTNH_F_LINKDOWN);
3741 break;
3742 }
3743 idev->if_flags |= IF_READY;
3744 }
3745
3746 pr_debug("ADDRCONF(NETDEV_CHANGE): %s: link becomes ready\n",
3747 dev->name);
3748
3749 run_pending = 1;
3750 }
3751
3752 addrconf_init_auto_addrs(dev);
3753
3754 if (!IS_ERR_OR_NULL(idev)) {
3755 if (run_pending)
3756 addrconf_dad_run(idev, false);
3757
3758 /* Device has an address by now */
3759 rt6_sync_up(dev, RTNH_F_DEAD);
3760
3761 /*
3762 * If the MTU changed during the interface down,
3763 * when the interface up, the changed MTU must be
3764 * reflected in the idev as well as routers.
3765 */
3766 if (idev->cnf.mtu6 != dev->mtu &&
3767 dev->mtu >= IPV6_MIN_MTU) {
3768 rt6_mtu_change(dev, dev->mtu);
3769 WRITE_ONCE(idev->cnf.mtu6, dev->mtu);
3770 }
3771 WRITE_ONCE(idev->tstamp, jiffies);
3772 inet6_ifinfo_notify(RTM_NEWLINK, idev);
3773
3774 /*
3775 * If the changed mtu during down is lower than
3776 * IPV6_MIN_MTU stop IPv6 on this interface.
3777 */
3778 if (dev->mtu < IPV6_MIN_MTU)
3779 addrconf_ifdown(dev, dev != net->loopback_dev);
3780 }
3781 break;
3782
3783 case NETDEV_DOWN:
3784 case NETDEV_UNREGISTER:
3785 /*
3786 * Remove all addresses from this interface.
3787 */
3788 addrconf_ifdown(dev, event != NETDEV_DOWN);
3789 break;
3790
3791 case NETDEV_CHANGENAME:
3792 if (idev) {
3793 snmp6_unregister_dev(idev);
3794 addrconf_sysctl_unregister(idev);
3795 err = addrconf_sysctl_register(idev);
3796 if (err)
3797 return notifier_from_errno(err);
3798 err = snmp6_register_dev(idev);
3799 if (err) {
3800 addrconf_sysctl_unregister(idev);
3801 return notifier_from_errno(err);
3802 }
3803 }
3804 break;
3805
3806 case NETDEV_PRE_TYPE_CHANGE:
3807 case NETDEV_POST_TYPE_CHANGE:
3808 if (idev)
3809 addrconf_type_change(dev, event);
3810 break;
3811
3812 case NETDEV_CHANGEUPPER:
3813 info = ptr;
3814
3815 /* flush all routes if dev is linked to or unlinked from
3816 * an L3 master device (e.g., VRF)
3817 */
3818 if (info->upper_dev && netif_is_l3_master(info->upper_dev))
3819 addrconf_ifdown(dev, false);
3820 }
3821
3822 return NOTIFY_OK;
3823 }
3824
3825 /*
3826 * addrconf module should be notified of a device going up
3827 */
3828 static struct notifier_block ipv6_dev_notf = {
3829 .notifier_call = addrconf_notify,
3830 .priority = ADDRCONF_NOTIFY_PRIORITY,
3831 };
3832
addrconf_type_change(struct net_device * dev,unsigned long event)3833 static void addrconf_type_change(struct net_device *dev, unsigned long event)
3834 {
3835 struct inet6_dev *idev;
3836 ASSERT_RTNL();
3837
3838 idev = __in6_dev_get(dev);
3839
3840 if (event == NETDEV_POST_TYPE_CHANGE)
3841 ipv6_mc_remap(idev);
3842 else if (event == NETDEV_PRE_TYPE_CHANGE)
3843 ipv6_mc_unmap(idev);
3844 }
3845
addr_is_local(const struct in6_addr * addr)3846 static bool addr_is_local(const struct in6_addr *addr)
3847 {
3848 return ipv6_addr_type(addr) &
3849 (IPV6_ADDR_LINKLOCAL | IPV6_ADDR_LOOPBACK);
3850 }
3851
addrconf_ifdown(struct net_device * dev,bool unregister)3852 static int addrconf_ifdown(struct net_device *dev, bool unregister)
3853 {
3854 unsigned long event = unregister ? NETDEV_UNREGISTER : NETDEV_DOWN;
3855 struct net *net = dev_net(dev);
3856 struct inet6_dev *idev;
3857 struct inet6_ifaddr *ifa;
3858 LIST_HEAD(tmp_addr_list);
3859 bool keep_addr = false;
3860 bool was_ready;
3861 int state, i;
3862
3863 ASSERT_RTNL();
3864
3865 rt6_disable_ip(dev, event);
3866
3867 idev = __in6_dev_get(dev);
3868 if (!idev)
3869 return -ENODEV;
3870
3871 /*
3872 * Step 1: remove reference to ipv6 device from parent device.
3873 * Do not dev_put!
3874 */
3875 if (unregister) {
3876 WRITE_ONCE(idev->dead, 1);
3877
3878 /* protected by rtnl_lock */
3879 RCU_INIT_POINTER(dev->ip6_ptr, NULL);
3880
3881 /* Step 1.5: remove snmp6 entry */
3882 snmp6_unregister_dev(idev);
3883
3884 }
3885
3886 /* combine the user config with event to determine if permanent
3887 * addresses are to be removed from address hash table
3888 */
3889 if (!unregister && !idev->cnf.disable_ipv6) {
3890 /* aggregate the system setting and interface setting */
3891 int _keep_addr = READ_ONCE(net->ipv6.devconf_all->keep_addr_on_down);
3892
3893 if (!_keep_addr)
3894 _keep_addr = READ_ONCE(idev->cnf.keep_addr_on_down);
3895
3896 keep_addr = (_keep_addr > 0);
3897 }
3898
3899 /* Step 2: clear hash table */
3900 for (i = 0; i < IN6_ADDR_HSIZE; i++) {
3901 struct hlist_head *h = &net->ipv6.inet6_addr_lst[i];
3902
3903 spin_lock_bh(&net->ipv6.addrconf_hash_lock);
3904 restart:
3905 hlist_for_each_entry_rcu(ifa, h, addr_lst) {
3906 if (ifa->idev == idev) {
3907 addrconf_del_dad_work(ifa);
3908 /* combined flag + permanent flag decide if
3909 * address is retained on a down event
3910 */
3911 if (!keep_addr ||
3912 !(ifa->flags & IFA_F_PERMANENT) ||
3913 addr_is_local(&ifa->addr)) {
3914 hlist_del_init_rcu(&ifa->addr_lst);
3915 goto restart;
3916 }
3917 }
3918 }
3919 spin_unlock_bh(&net->ipv6.addrconf_hash_lock);
3920 }
3921
3922 write_lock_bh(&idev->lock);
3923
3924 addrconf_del_rs_timer(idev);
3925
3926 /* Step 2: clear flags for stateless addrconf, repeated down
3927 * detection
3928 */
3929 was_ready = idev->if_flags & IF_READY;
3930 if (!unregister)
3931 idev->if_flags &= ~(IF_RS_SENT|IF_RA_RCVD|IF_READY);
3932
3933 /* Step 3: clear tempaddr list */
3934 while (!list_empty(&idev->tempaddr_list)) {
3935 ifa = list_first_entry(&idev->tempaddr_list,
3936 struct inet6_ifaddr, tmp_list);
3937 list_del(&ifa->tmp_list);
3938 write_unlock_bh(&idev->lock);
3939 spin_lock_bh(&ifa->lock);
3940
3941 if (ifa->ifpub) {
3942 in6_ifa_put(ifa->ifpub);
3943 ifa->ifpub = NULL;
3944 }
3945 spin_unlock_bh(&ifa->lock);
3946 in6_ifa_put(ifa);
3947 write_lock_bh(&idev->lock);
3948 }
3949
3950 list_for_each_entry(ifa, &idev->addr_list, if_list)
3951 list_add_tail(&ifa->if_list_aux, &tmp_addr_list);
3952 write_unlock_bh(&idev->lock);
3953
3954 while (!list_empty(&tmp_addr_list)) {
3955 struct fib6_info *rt = NULL;
3956 bool keep;
3957
3958 ifa = list_first_entry(&tmp_addr_list,
3959 struct inet6_ifaddr, if_list_aux);
3960 list_del(&ifa->if_list_aux);
3961
3962 addrconf_del_dad_work(ifa);
3963
3964 keep = keep_addr && (ifa->flags & IFA_F_PERMANENT) &&
3965 !addr_is_local(&ifa->addr);
3966
3967 spin_lock_bh(&ifa->lock);
3968
3969 if (keep) {
3970 /* set state to skip the notifier below */
3971 state = INET6_IFADDR_STATE_DEAD;
3972 ifa->state = INET6_IFADDR_STATE_PREDAD;
3973 if (!(ifa->flags & IFA_F_NODAD))
3974 ifa->flags |= IFA_F_TENTATIVE;
3975
3976 rt = ifa->rt;
3977 ifa->rt = NULL;
3978 } else {
3979 state = ifa->state;
3980 ifa->state = INET6_IFADDR_STATE_DEAD;
3981 }
3982
3983 spin_unlock_bh(&ifa->lock);
3984
3985 if (rt)
3986 ip6_del_rt(net, rt, false);
3987
3988 if (state != INET6_IFADDR_STATE_DEAD) {
3989 __ipv6_ifa_notify(RTM_DELADDR, ifa);
3990 inet6addr_notifier_call_chain(NETDEV_DOWN, ifa);
3991 } else {
3992 if (idev->cnf.forwarding)
3993 addrconf_leave_anycast(ifa);
3994 addrconf_leave_solict(ifa->idev, &ifa->addr);
3995 }
3996
3997 if (!keep) {
3998 write_lock_bh(&idev->lock);
3999 list_del_rcu(&ifa->if_list);
4000 write_unlock_bh(&idev->lock);
4001 in6_ifa_put(ifa);
4002 }
4003 }
4004
4005 /* Step 5: Discard anycast and multicast list */
4006 if (unregister) {
4007 ipv6_ac_destroy_dev(idev);
4008 ipv6_mc_destroy_dev(idev);
4009 } else if (was_ready) {
4010 ipv6_mc_down(idev);
4011 }
4012
4013 WRITE_ONCE(idev->tstamp, jiffies);
4014 idev->ra_mtu = 0;
4015
4016 /* Last: Shot the device (if unregistered) */
4017 if (unregister) {
4018 addrconf_sysctl_unregister(idev);
4019 neigh_parms_release(&nd_tbl, idev->nd_parms);
4020 neigh_ifdown(&nd_tbl, dev);
4021 in6_dev_put(idev);
4022 }
4023 return 0;
4024 }
4025
addrconf_rs_timer(struct timer_list * t)4026 static void addrconf_rs_timer(struct timer_list *t)
4027 {
4028 struct inet6_dev *idev = timer_container_of(idev, t, rs_timer);
4029 struct net_device *dev = idev->dev;
4030 struct in6_addr lladdr;
4031 int rtr_solicits;
4032
4033 write_lock(&idev->lock);
4034 if (idev->dead || !(idev->if_flags & IF_READY))
4035 goto out;
4036
4037 if (!ipv6_accept_ra(idev))
4038 goto out;
4039
4040 /* Announcement received after solicitation was sent */
4041 if (idev->if_flags & IF_RA_RCVD)
4042 goto out;
4043
4044 rtr_solicits = READ_ONCE(idev->cnf.rtr_solicits);
4045
4046 if (idev->rs_probes++ < rtr_solicits || rtr_solicits < 0) {
4047 write_unlock(&idev->lock);
4048 if (!ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE))
4049 ndisc_send_rs(dev, &lladdr,
4050 &in6addr_linklocal_allrouters);
4051 else
4052 goto put;
4053
4054 write_lock(&idev->lock);
4055 idev->rs_interval = rfc3315_s14_backoff_update(
4056 idev->rs_interval,
4057 READ_ONCE(idev->cnf.rtr_solicit_max_interval));
4058 /* The wait after the last probe can be shorter */
4059 addrconf_mod_rs_timer(idev, (idev->rs_probes ==
4060 READ_ONCE(idev->cnf.rtr_solicits)) ?
4061 READ_ONCE(idev->cnf.rtr_solicit_delay) :
4062 idev->rs_interval);
4063 } else {
4064 /*
4065 * Note: we do not support deprecated "all on-link"
4066 * assumption any longer.
4067 */
4068 pr_debug("%s: no IPv6 routers present\n", idev->dev->name);
4069 }
4070
4071 out:
4072 write_unlock(&idev->lock);
4073 put:
4074 in6_dev_put(idev);
4075 }
4076
4077 /*
4078 * Duplicate Address Detection
4079 */
addrconf_dad_kick(struct inet6_ifaddr * ifp)4080 static void addrconf_dad_kick(struct inet6_ifaddr *ifp)
4081 {
4082 struct inet6_dev *idev = ifp->idev;
4083 unsigned long rand_num;
4084 u64 nonce;
4085
4086 if (ifp->flags & IFA_F_OPTIMISTIC)
4087 rand_num = 0;
4088 else
4089 rand_num = get_random_u32_below(
4090 READ_ONCE(idev->cnf.rtr_solicit_delay) ? : 1);
4091
4092 nonce = 0;
4093 if (READ_ONCE(idev->cnf.enhanced_dad) ||
4094 READ_ONCE(dev_net(idev->dev)->ipv6.devconf_all->enhanced_dad)) {
4095 do
4096 get_random_bytes(&nonce, 6);
4097 while (nonce == 0);
4098 }
4099 ifp->dad_nonce = nonce;
4100 ifp->dad_probes = READ_ONCE(idev->cnf.dad_transmits);
4101 addrconf_mod_dad_work(ifp, rand_num);
4102 }
4103
addrconf_dad_begin(struct inet6_ifaddr * ifp)4104 static void addrconf_dad_begin(struct inet6_ifaddr *ifp)
4105 {
4106 struct inet6_dev *idev = ifp->idev;
4107 struct net_device *dev = idev->dev;
4108 bool bump_id, notify = false;
4109 struct net *net;
4110
4111 addrconf_join_solict(dev, &ifp->addr);
4112
4113 read_lock_bh(&idev->lock);
4114 spin_lock(&ifp->lock);
4115 if (ifp->state == INET6_IFADDR_STATE_DEAD)
4116 goto out;
4117
4118 net = dev_net(dev);
4119 if (dev->flags&(IFF_NOARP|IFF_LOOPBACK) ||
4120 (READ_ONCE(net->ipv6.devconf_all->accept_dad) < 1 &&
4121 READ_ONCE(idev->cnf.accept_dad) < 1) ||
4122 !(ifp->flags&IFA_F_TENTATIVE) ||
4123 ifp->flags & IFA_F_NODAD) {
4124 bool send_na = false;
4125
4126 if (ifp->flags & IFA_F_TENTATIVE &&
4127 !(ifp->flags & IFA_F_OPTIMISTIC))
4128 send_na = true;
4129 bump_id = ifp->flags & IFA_F_TENTATIVE;
4130 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
4131 spin_unlock(&ifp->lock);
4132 read_unlock_bh(&idev->lock);
4133
4134 addrconf_dad_completed(ifp, bump_id, send_na);
4135 return;
4136 }
4137
4138 if (!(idev->if_flags & IF_READY)) {
4139 spin_unlock(&ifp->lock);
4140 read_unlock_bh(&idev->lock);
4141 /*
4142 * If the device is not ready:
4143 * - keep it tentative if it is a permanent address.
4144 * - otherwise, kill it.
4145 */
4146 in6_ifa_hold(ifp);
4147 addrconf_dad_stop(ifp, 0);
4148 return;
4149 }
4150
4151 /*
4152 * Optimistic nodes can start receiving
4153 * Frames right away
4154 */
4155 if (ifp->flags & IFA_F_OPTIMISTIC) {
4156 ip6_ins_rt(net, ifp->rt);
4157 if (ipv6_use_optimistic_addr(net, idev)) {
4158 /* Because optimistic nodes can use this address,
4159 * notify listeners. If DAD fails, RTM_DELADDR is sent.
4160 */
4161 notify = true;
4162 }
4163 }
4164
4165 addrconf_dad_kick(ifp);
4166 out:
4167 spin_unlock(&ifp->lock);
4168 read_unlock_bh(&idev->lock);
4169 if (notify)
4170 ipv6_ifa_notify(RTM_NEWADDR, ifp);
4171 }
4172
addrconf_dad_start(struct inet6_ifaddr * ifp)4173 static void addrconf_dad_start(struct inet6_ifaddr *ifp)
4174 {
4175 bool begin_dad = false;
4176
4177 spin_lock_bh(&ifp->lock);
4178 if (ifp->state != INET6_IFADDR_STATE_DEAD) {
4179 ifp->state = INET6_IFADDR_STATE_PREDAD;
4180 begin_dad = true;
4181 }
4182 spin_unlock_bh(&ifp->lock);
4183
4184 if (begin_dad)
4185 addrconf_mod_dad_work(ifp, 0);
4186 }
4187
addrconf_dad_work(struct work_struct * w)4188 static void addrconf_dad_work(struct work_struct *w)
4189 {
4190 struct inet6_ifaddr *ifp = container_of(to_delayed_work(w),
4191 struct inet6_ifaddr,
4192 dad_work);
4193 struct inet6_dev *idev = ifp->idev;
4194 bool bump_id, disable_ipv6 = false;
4195 struct in6_addr mcaddr;
4196 struct net *net;
4197
4198 enum {
4199 DAD_PROCESS,
4200 DAD_BEGIN,
4201 DAD_ABORT,
4202 } action = DAD_PROCESS;
4203
4204 net = dev_net(idev->dev);
4205
4206 rtnl_net_lock(net);
4207
4208 spin_lock_bh(&ifp->lock);
4209 if (ifp->state == INET6_IFADDR_STATE_PREDAD) {
4210 action = DAD_BEGIN;
4211 ifp->state = INET6_IFADDR_STATE_DAD;
4212 } else if (ifp->state == INET6_IFADDR_STATE_ERRDAD) {
4213 action = DAD_ABORT;
4214 ifp->state = INET6_IFADDR_STATE_POSTDAD;
4215
4216 if ((READ_ONCE(net->ipv6.devconf_all->accept_dad) > 1 ||
4217 READ_ONCE(idev->cnf.accept_dad) > 1) &&
4218 !idev->cnf.disable_ipv6 &&
4219 !(ifp->flags & IFA_F_STABLE_PRIVACY)) {
4220 struct in6_addr addr;
4221
4222 addr.s6_addr32[0] = htonl(0xfe800000);
4223 addr.s6_addr32[1] = 0;
4224
4225 if (!ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) &&
4226 ipv6_addr_equal(&ifp->addr, &addr)) {
4227 /* DAD failed for link-local based on MAC */
4228 WRITE_ONCE(idev->cnf.disable_ipv6, 1);
4229
4230 pr_info("%s: IPv6 being disabled!\n",
4231 ifp->idev->dev->name);
4232 disable_ipv6 = true;
4233 }
4234 }
4235 }
4236 spin_unlock_bh(&ifp->lock);
4237
4238 if (action == DAD_BEGIN) {
4239 addrconf_dad_begin(ifp);
4240 goto out;
4241 } else if (action == DAD_ABORT) {
4242 in6_ifa_hold(ifp);
4243 addrconf_dad_stop(ifp, 1);
4244 if (disable_ipv6)
4245 addrconf_ifdown(idev->dev, false);
4246 goto out;
4247 }
4248
4249 if (!ifp->dad_probes && addrconf_dad_end(ifp))
4250 goto out;
4251
4252 write_lock_bh(&idev->lock);
4253 if (idev->dead || !(idev->if_flags & IF_READY)) {
4254 write_unlock_bh(&idev->lock);
4255 goto out;
4256 }
4257
4258 spin_lock(&ifp->lock);
4259 if (ifp->state == INET6_IFADDR_STATE_DEAD) {
4260 spin_unlock(&ifp->lock);
4261 write_unlock_bh(&idev->lock);
4262 goto out;
4263 }
4264
4265 if (ifp->dad_probes == 0) {
4266 bool send_na = false;
4267
4268 /*
4269 * DAD was successful
4270 */
4271
4272 if (ifp->flags & IFA_F_TENTATIVE &&
4273 !(ifp->flags & IFA_F_OPTIMISTIC))
4274 send_na = true;
4275 bump_id = ifp->flags & IFA_F_TENTATIVE;
4276 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
4277 spin_unlock(&ifp->lock);
4278 write_unlock_bh(&idev->lock);
4279
4280 addrconf_dad_completed(ifp, bump_id, send_na);
4281
4282 goto out;
4283 }
4284
4285 ifp->dad_probes--;
4286 addrconf_mod_dad_work(ifp,
4287 max(NEIGH_VAR(ifp->idev->nd_parms, RETRANS_TIME),
4288 HZ/100));
4289 spin_unlock(&ifp->lock);
4290 write_unlock_bh(&idev->lock);
4291
4292 /* send a neighbour solicitation for our addr */
4293 addrconf_addr_solict_mult(&ifp->addr, &mcaddr);
4294 ndisc_send_ns(ifp->idev->dev, &ifp->addr, &mcaddr, &in6addr_any,
4295 ifp->dad_nonce);
4296 out:
4297 in6_ifa_put(ifp);
4298 rtnl_net_unlock(net);
4299 }
4300
4301 /* ifp->idev must be at least read locked */
ipv6_lonely_lladdr(struct inet6_ifaddr * ifp)4302 static bool ipv6_lonely_lladdr(struct inet6_ifaddr *ifp)
4303 {
4304 struct inet6_ifaddr *ifpiter;
4305 struct inet6_dev *idev = ifp->idev;
4306
4307 list_for_each_entry_reverse(ifpiter, &idev->addr_list, if_list) {
4308 if (ifpiter->scope > IFA_LINK)
4309 break;
4310 if (ifp != ifpiter && ifpiter->scope == IFA_LINK &&
4311 (ifpiter->flags & (IFA_F_PERMANENT|IFA_F_TENTATIVE|
4312 IFA_F_OPTIMISTIC|IFA_F_DADFAILED)) ==
4313 IFA_F_PERMANENT)
4314 return false;
4315 }
4316 return true;
4317 }
4318
addrconf_dad_completed(struct inet6_ifaddr * ifp,bool bump_id,bool send_na)4319 static void addrconf_dad_completed(struct inet6_ifaddr *ifp, bool bump_id,
4320 bool send_na)
4321 {
4322 struct net_device *dev = ifp->idev->dev;
4323 struct in6_addr lladdr;
4324 bool send_rs, send_mld;
4325
4326 addrconf_del_dad_work(ifp);
4327
4328 /*
4329 * Configure the address for reception. Now it is valid.
4330 */
4331
4332 ipv6_ifa_notify(RTM_NEWADDR, ifp);
4333
4334 /* If added prefix is link local and we are prepared to process
4335 router advertisements, start sending router solicitations.
4336 */
4337
4338 read_lock_bh(&ifp->idev->lock);
4339 send_mld = ifp->scope == IFA_LINK && ipv6_lonely_lladdr(ifp);
4340 send_rs = send_mld &&
4341 ipv6_accept_ra(ifp->idev) &&
4342 READ_ONCE(ifp->idev->cnf.rtr_solicits) != 0 &&
4343 (dev->flags & IFF_LOOPBACK) == 0 &&
4344 (dev->type != ARPHRD_TUNNEL) &&
4345 !netif_is_team_port(dev);
4346 read_unlock_bh(&ifp->idev->lock);
4347
4348 /* While dad is in progress mld report's source address is in6_addrany.
4349 * Resend with proper ll now.
4350 */
4351 if (send_mld)
4352 ipv6_mc_dad_complete(ifp->idev);
4353
4354 /* send unsolicited NA if enabled */
4355 if (send_na &&
4356 (READ_ONCE(ifp->idev->cnf.ndisc_notify) ||
4357 READ_ONCE(dev_net(dev)->ipv6.devconf_all->ndisc_notify))) {
4358 ndisc_send_na(dev, &in6addr_linklocal_allnodes, &ifp->addr,
4359 /*router=*/ !!ifp->idev->cnf.forwarding,
4360 /*solicited=*/ false, /*override=*/ true,
4361 /*inc_opt=*/ true);
4362 }
4363
4364 if (send_rs) {
4365 /*
4366 * If a host as already performed a random delay
4367 * [...] as part of DAD [...] there is no need
4368 * to delay again before sending the first RS
4369 */
4370 if (ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE))
4371 return;
4372 ndisc_send_rs(dev, &lladdr, &in6addr_linklocal_allrouters);
4373
4374 write_lock_bh(&ifp->idev->lock);
4375 spin_lock(&ifp->lock);
4376 ifp->idev->rs_interval = rfc3315_s14_backoff_init(
4377 READ_ONCE(ifp->idev->cnf.rtr_solicit_interval));
4378 ifp->idev->rs_probes = 1;
4379 ifp->idev->if_flags |= IF_RS_SENT;
4380 addrconf_mod_rs_timer(ifp->idev, ifp->idev->rs_interval);
4381 spin_unlock(&ifp->lock);
4382 write_unlock_bh(&ifp->idev->lock);
4383 }
4384
4385 if (bump_id)
4386 rt_genid_bump_ipv6(dev_net(dev));
4387
4388 /* Make sure that a new temporary address will be created
4389 * before this temporary address becomes deprecated.
4390 */
4391 if (ifp->flags & IFA_F_TEMPORARY)
4392 addrconf_verify_rtnl(dev_net(dev));
4393 }
4394
addrconf_dad_run(struct inet6_dev * idev,bool restart)4395 static void addrconf_dad_run(struct inet6_dev *idev, bool restart)
4396 {
4397 struct inet6_ifaddr *ifp;
4398
4399 read_lock_bh(&idev->lock);
4400 list_for_each_entry(ifp, &idev->addr_list, if_list) {
4401 spin_lock(&ifp->lock);
4402 if ((ifp->flags & IFA_F_TENTATIVE &&
4403 ifp->state == INET6_IFADDR_STATE_DAD) || restart) {
4404 if (restart)
4405 ifp->state = INET6_IFADDR_STATE_PREDAD;
4406 addrconf_dad_kick(ifp);
4407 }
4408 spin_unlock(&ifp->lock);
4409 }
4410 read_unlock_bh(&idev->lock);
4411 }
4412
4413 #ifdef CONFIG_PROC_FS
4414 struct if6_iter_state {
4415 struct seq_net_private p;
4416 int bucket;
4417 int offset;
4418 };
4419
if6_get_first(struct seq_file * seq,loff_t pos)4420 static struct inet6_ifaddr *if6_get_first(struct seq_file *seq, loff_t pos)
4421 {
4422 struct if6_iter_state *state = seq->private;
4423 struct net *net = seq_file_net(seq);
4424 struct inet6_ifaddr *ifa = NULL;
4425 int p = 0;
4426
4427 /* initial bucket if pos is 0 */
4428 if (pos == 0) {
4429 state->bucket = 0;
4430 state->offset = 0;
4431 }
4432
4433 for (; state->bucket < IN6_ADDR_HSIZE; ++state->bucket) {
4434 hlist_for_each_entry_rcu(ifa, &net->ipv6.inet6_addr_lst[state->bucket],
4435 addr_lst) {
4436 /* sync with offset */
4437 if (p < state->offset) {
4438 p++;
4439 continue;
4440 }
4441 return ifa;
4442 }
4443
4444 /* prepare for next bucket */
4445 state->offset = 0;
4446 p = 0;
4447 }
4448 return NULL;
4449 }
4450
if6_get_next(struct seq_file * seq,struct inet6_ifaddr * ifa)4451 static struct inet6_ifaddr *if6_get_next(struct seq_file *seq,
4452 struct inet6_ifaddr *ifa)
4453 {
4454 struct if6_iter_state *state = seq->private;
4455 struct net *net = seq_file_net(seq);
4456
4457 hlist_for_each_entry_continue_rcu(ifa, addr_lst) {
4458 state->offset++;
4459 return ifa;
4460 }
4461
4462 state->offset = 0;
4463 while (++state->bucket < IN6_ADDR_HSIZE) {
4464 hlist_for_each_entry_rcu(ifa,
4465 &net->ipv6.inet6_addr_lst[state->bucket], addr_lst) {
4466 return ifa;
4467 }
4468 }
4469
4470 return NULL;
4471 }
4472
if6_seq_start(struct seq_file * seq,loff_t * pos)4473 static void *if6_seq_start(struct seq_file *seq, loff_t *pos)
4474 __acquires(rcu)
4475 {
4476 rcu_read_lock();
4477 return if6_get_first(seq, *pos);
4478 }
4479
if6_seq_next(struct seq_file * seq,void * v,loff_t * pos)4480 static void *if6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
4481 {
4482 struct inet6_ifaddr *ifa;
4483
4484 ifa = if6_get_next(seq, v);
4485 ++*pos;
4486 return ifa;
4487 }
4488
if6_seq_stop(struct seq_file * seq,void * v)4489 static void if6_seq_stop(struct seq_file *seq, void *v)
4490 __releases(rcu)
4491 {
4492 rcu_read_unlock();
4493 }
4494
if6_seq_show(struct seq_file * seq,void * v)4495 static int if6_seq_show(struct seq_file *seq, void *v)
4496 {
4497 struct inet6_ifaddr *ifp = (struct inet6_ifaddr *)v;
4498 seq_printf(seq, "%pi6 %02x %02x %02x %02x %8s\n",
4499 &ifp->addr,
4500 ifp->idev->dev->ifindex,
4501 ifp->prefix_len,
4502 ifp->scope,
4503 (u8) ifp->flags,
4504 ifp->idev->dev->name);
4505 return 0;
4506 }
4507
4508 static const struct seq_operations if6_seq_ops = {
4509 .start = if6_seq_start,
4510 .next = if6_seq_next,
4511 .show = if6_seq_show,
4512 .stop = if6_seq_stop,
4513 };
4514
if6_proc_net_init(struct net * net)4515 static int __net_init if6_proc_net_init(struct net *net)
4516 {
4517 if (!proc_create_net("if_inet6", 0444, net->proc_net, &if6_seq_ops,
4518 sizeof(struct if6_iter_state)))
4519 return -ENOMEM;
4520 return 0;
4521 }
4522
if6_proc_net_exit(struct net * net)4523 static void __net_exit if6_proc_net_exit(struct net *net)
4524 {
4525 remove_proc_entry("if_inet6", net->proc_net);
4526 }
4527
4528 static struct pernet_operations if6_proc_net_ops = {
4529 .init = if6_proc_net_init,
4530 .exit = if6_proc_net_exit,
4531 };
4532
if6_proc_init(void)4533 int __init if6_proc_init(void)
4534 {
4535 return register_pernet_subsys(&if6_proc_net_ops);
4536 }
4537
if6_proc_exit(void)4538 void if6_proc_exit(void)
4539 {
4540 unregister_pernet_subsys(&if6_proc_net_ops);
4541 }
4542 #endif /* CONFIG_PROC_FS */
4543
4544 #if IS_ENABLED(CONFIG_IPV6_MIP6)
4545 /* Check if address is a home address configured on any interface. */
ipv6_chk_home_addr(struct net * net,const struct in6_addr * addr)4546 int ipv6_chk_home_addr(struct net *net, const struct in6_addr *addr)
4547 {
4548 unsigned int hash = inet6_addr_hash(net, addr);
4549 struct inet6_ifaddr *ifp = NULL;
4550 int ret = 0;
4551
4552 rcu_read_lock();
4553 hlist_for_each_entry_rcu(ifp, &net->ipv6.inet6_addr_lst[hash], addr_lst) {
4554 if (ipv6_addr_equal(&ifp->addr, addr) &&
4555 (ifp->flags & IFA_F_HOMEADDRESS)) {
4556 ret = 1;
4557 break;
4558 }
4559 }
4560 rcu_read_unlock();
4561 return ret;
4562 }
4563 #endif
4564
4565 /* RFC6554 has some algorithm to avoid loops in segment routing by
4566 * checking if the segments contains any of a local interface address.
4567 *
4568 * Quote:
4569 *
4570 * To detect loops in the SRH, a router MUST determine if the SRH
4571 * includes multiple addresses assigned to any interface on that router.
4572 * If such addresses appear more than once and are separated by at least
4573 * one address not assigned to that router.
4574 */
ipv6_chk_rpl_srh_loop(struct net * net,const struct in6_addr * segs,unsigned char nsegs)4575 int ipv6_chk_rpl_srh_loop(struct net *net, const struct in6_addr *segs,
4576 unsigned char nsegs)
4577 {
4578 const struct in6_addr *addr;
4579 int i, ret = 0, found = 0;
4580 struct inet6_ifaddr *ifp;
4581 bool separated = false;
4582 unsigned int hash;
4583 bool hash_found;
4584
4585 rcu_read_lock();
4586 for (i = 0; i < nsegs; i++) {
4587 addr = &segs[i];
4588 hash = inet6_addr_hash(net, addr);
4589
4590 hash_found = false;
4591 hlist_for_each_entry_rcu(ifp, &net->ipv6.inet6_addr_lst[hash], addr_lst) {
4592
4593 if (ipv6_addr_equal(&ifp->addr, addr)) {
4594 hash_found = true;
4595 break;
4596 }
4597 }
4598
4599 if (hash_found) {
4600 if (found > 1 && separated) {
4601 ret = 1;
4602 break;
4603 }
4604
4605 separated = false;
4606 found++;
4607 } else {
4608 separated = true;
4609 }
4610 }
4611 rcu_read_unlock();
4612
4613 return ret;
4614 }
4615
4616 /*
4617 * Periodic address status verification
4618 */
4619
addrconf_verify_rtnl(struct net * net)4620 static void addrconf_verify_rtnl(struct net *net)
4621 {
4622 unsigned long now, next, next_sec, next_sched;
4623 struct inet6_ifaddr *ifp;
4624 int i;
4625
4626 ASSERT_RTNL();
4627
4628 rcu_read_lock_bh();
4629 now = jiffies;
4630 next = round_jiffies_up(now + ADDR_CHECK_FREQUENCY);
4631
4632 cancel_delayed_work(&net->ipv6.addr_chk_work);
4633
4634 for (i = 0; i < IN6_ADDR_HSIZE; i++) {
4635 restart:
4636 hlist_for_each_entry_rcu_bh(ifp, &net->ipv6.inet6_addr_lst[i], addr_lst) {
4637 unsigned long age;
4638
4639 /* When setting preferred_lft to a value not zero or
4640 * infinity, while valid_lft is infinity
4641 * IFA_F_PERMANENT has a non-infinity life time.
4642 */
4643 if ((ifp->flags & IFA_F_PERMANENT) &&
4644 (ifp->prefered_lft == INFINITY_LIFE_TIME))
4645 continue;
4646
4647 spin_lock(&ifp->lock);
4648 /* We try to batch several events at once. */
4649 age = (now - ifp->tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
4650
4651 if ((ifp->flags&IFA_F_TEMPORARY) &&
4652 !(ifp->flags&IFA_F_TENTATIVE) &&
4653 ifp->prefered_lft != INFINITY_LIFE_TIME &&
4654 !ifp->regen_count && ifp->ifpub) {
4655 /* This is a non-regenerated temporary addr. */
4656
4657 unsigned long regen_advance = ipv6_get_regen_advance(ifp->idev);
4658
4659 if (age + regen_advance >= ifp->prefered_lft) {
4660 struct inet6_ifaddr *ifpub = ifp->ifpub;
4661 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
4662 next = ifp->tstamp + ifp->prefered_lft * HZ;
4663
4664 ifp->regen_count++;
4665 in6_ifa_hold(ifp);
4666 in6_ifa_hold(ifpub);
4667 spin_unlock(&ifp->lock);
4668
4669 spin_lock(&ifpub->lock);
4670 ifpub->regen_count = 0;
4671 spin_unlock(&ifpub->lock);
4672 rcu_read_unlock_bh();
4673 ipv6_create_tempaddr(ifpub, true);
4674 in6_ifa_put(ifpub);
4675 in6_ifa_put(ifp);
4676 rcu_read_lock_bh();
4677 goto restart;
4678 } else if (time_before(ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ, next))
4679 next = ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ;
4680 }
4681
4682 if (ifp->valid_lft != INFINITY_LIFE_TIME &&
4683 age >= ifp->valid_lft) {
4684 spin_unlock(&ifp->lock);
4685 in6_ifa_hold(ifp);
4686 rcu_read_unlock_bh();
4687 ipv6_del_addr(ifp);
4688 rcu_read_lock_bh();
4689 goto restart;
4690 } else if (ifp->prefered_lft == INFINITY_LIFE_TIME) {
4691 spin_unlock(&ifp->lock);
4692 continue;
4693 } else if (age >= ifp->prefered_lft) {
4694 /* jiffies - ifp->tstamp > age >= ifp->prefered_lft */
4695 int deprecate = 0;
4696
4697 if (!(ifp->flags&IFA_F_DEPRECATED)) {
4698 deprecate = 1;
4699 ifp->flags |= IFA_F_DEPRECATED;
4700 }
4701
4702 if ((ifp->valid_lft != INFINITY_LIFE_TIME) &&
4703 (time_before(ifp->tstamp + ifp->valid_lft * HZ, next)))
4704 next = ifp->tstamp + ifp->valid_lft * HZ;
4705
4706 spin_unlock(&ifp->lock);
4707
4708 if (deprecate) {
4709 in6_ifa_hold(ifp);
4710
4711 ipv6_ifa_notify(0, ifp);
4712 in6_ifa_put(ifp);
4713 goto restart;
4714 }
4715 } else {
4716 /* ifp->prefered_lft <= ifp->valid_lft */
4717 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
4718 next = ifp->tstamp + ifp->prefered_lft * HZ;
4719 spin_unlock(&ifp->lock);
4720 }
4721 }
4722 }
4723
4724 next_sec = round_jiffies_up(next);
4725 next_sched = next;
4726
4727 /* If rounded timeout is accurate enough, accept it. */
4728 if (time_before(next_sec, next + ADDRCONF_TIMER_FUZZ))
4729 next_sched = next_sec;
4730
4731 /* And minimum interval is ADDRCONF_TIMER_FUZZ_MAX. */
4732 if (time_before(next_sched, jiffies + ADDRCONF_TIMER_FUZZ_MAX))
4733 next_sched = jiffies + ADDRCONF_TIMER_FUZZ_MAX;
4734
4735 pr_debug("now = %lu, schedule = %lu, rounded schedule = %lu => %lu\n",
4736 now, next, next_sec, next_sched);
4737 mod_delayed_work(addrconf_wq, &net->ipv6.addr_chk_work, next_sched - now);
4738 rcu_read_unlock_bh();
4739 }
4740
addrconf_verify_work(struct work_struct * w)4741 static void addrconf_verify_work(struct work_struct *w)
4742 {
4743 struct net *net = container_of(to_delayed_work(w), struct net,
4744 ipv6.addr_chk_work);
4745
4746 rtnl_net_lock(net);
4747 addrconf_verify_rtnl(net);
4748 rtnl_net_unlock(net);
4749 }
4750
addrconf_verify(struct net * net)4751 static void addrconf_verify(struct net *net)
4752 {
4753 mod_delayed_work(addrconf_wq, &net->ipv6.addr_chk_work, 0);
4754 }
4755
extract_addr(struct nlattr * addr,struct nlattr * local,struct in6_addr ** peer_pfx)4756 static struct in6_addr *extract_addr(struct nlattr *addr, struct nlattr *local,
4757 struct in6_addr **peer_pfx)
4758 {
4759 struct in6_addr *pfx = NULL;
4760
4761 *peer_pfx = NULL;
4762
4763 if (addr)
4764 pfx = nla_data(addr);
4765
4766 if (local) {
4767 if (pfx && nla_memcmp(local, pfx, sizeof(*pfx)))
4768 *peer_pfx = pfx;
4769 pfx = nla_data(local);
4770 }
4771
4772 return pfx;
4773 }
4774
4775 static const struct nla_policy ifa_ipv6_policy[IFA_MAX+1] = {
4776 [IFA_ADDRESS] = { .len = sizeof(struct in6_addr) },
4777 [IFA_LOCAL] = { .len = sizeof(struct in6_addr) },
4778 [IFA_CACHEINFO] = { .len = sizeof(struct ifa_cacheinfo) },
4779 [IFA_FLAGS] = { .len = sizeof(u32) },
4780 [IFA_RT_PRIORITY] = { .len = sizeof(u32) },
4781 [IFA_TARGET_NETNSID] = { .type = NLA_S32 },
4782 [IFA_PROTO] = { .type = NLA_U8 },
4783 };
4784
4785 static int
inet6_rtm_deladdr(struct sk_buff * skb,struct nlmsghdr * nlh,struct netlink_ext_ack * extack)4786 inet6_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh,
4787 struct netlink_ext_ack *extack)
4788 {
4789 struct net *net = sock_net(skb->sk);
4790 struct ifaddrmsg *ifm;
4791 struct nlattr *tb[IFA_MAX+1];
4792 struct in6_addr *pfx, *peer_pfx;
4793 u32 ifa_flags;
4794 int err;
4795
4796 err = nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFA_MAX,
4797 ifa_ipv6_policy, extack);
4798 if (err < 0)
4799 return err;
4800
4801 ifm = nlmsg_data(nlh);
4802 pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
4803 if (!pfx)
4804 return -EINVAL;
4805
4806 ifa_flags = nla_get_u32_default(tb[IFA_FLAGS], ifm->ifa_flags);
4807
4808 /* We ignore other flags so far. */
4809 ifa_flags &= IFA_F_MANAGETEMPADDR;
4810
4811 rtnl_net_lock(net);
4812 err = inet6_addr_del(net, ifm->ifa_index, ifa_flags, pfx,
4813 ifm->ifa_prefixlen, extack);
4814 rtnl_net_unlock(net);
4815
4816 return err;
4817 }
4818
modify_prefix_route(struct net * net,struct inet6_ifaddr * ifp,unsigned long expires,u32 flags,bool modify_peer)4819 static int modify_prefix_route(struct net *net, struct inet6_ifaddr *ifp,
4820 unsigned long expires, u32 flags,
4821 bool modify_peer)
4822 {
4823 struct fib6_table *table;
4824 struct fib6_info *f6i;
4825 u32 prio;
4826
4827 f6i = addrconf_get_prefix_route(modify_peer ? &ifp->peer_addr : &ifp->addr,
4828 ifp->prefix_len,
4829 ifp->idev->dev, 0, RTF_DEFAULT, true);
4830 if (!f6i)
4831 return -ENOENT;
4832
4833 prio = ifp->rt_priority ? : IP6_RT_PRIO_ADDRCONF;
4834 if (f6i->fib6_metric != prio) {
4835 /* delete old one */
4836 ip6_del_rt(dev_net(ifp->idev->dev), f6i, false);
4837
4838 /* add new one */
4839 addrconf_prefix_route(modify_peer ? &ifp->peer_addr : &ifp->addr,
4840 ifp->prefix_len,
4841 ifp->rt_priority, ifp->idev->dev,
4842 expires, flags, GFP_KERNEL);
4843 return 0;
4844 }
4845 if (f6i != net->ipv6.fib6_null_entry) {
4846 table = f6i->fib6_table;
4847 spin_lock_bh(&table->tb6_lock);
4848
4849 if (!(flags & RTF_EXPIRES)) {
4850 fib6_clean_expires(f6i);
4851 fib6_may_remove_gc_list(net, f6i);
4852 } else {
4853 fib6_set_expires(f6i, expires);
4854 fib6_add_gc_list(f6i);
4855 }
4856
4857 spin_unlock_bh(&table->tb6_lock);
4858 }
4859 fib6_info_release(f6i);
4860
4861 return 0;
4862 }
4863
inet6_addr_modify(struct net * net,struct inet6_ifaddr * ifp,struct ifa6_config * cfg,clock_t expires,u32 flags)4864 static int inet6_addr_modify(struct net *net, struct inet6_ifaddr *ifp,
4865 struct ifa6_config *cfg, clock_t expires,
4866 u32 flags)
4867 {
4868 bool was_managetempaddr;
4869 bool new_peer = false;
4870 bool had_prefixroute;
4871
4872 ASSERT_RTNL_NET(net);
4873
4874 if (cfg->ifa_flags & IFA_F_MANAGETEMPADDR &&
4875 (ifp->flags & IFA_F_TEMPORARY || ifp->prefix_len != 64))
4876 return -EINVAL;
4877
4878 if (!(ifp->flags & IFA_F_TENTATIVE) || ifp->flags & IFA_F_DADFAILED)
4879 cfg->ifa_flags &= ~IFA_F_OPTIMISTIC;
4880
4881 if (cfg->peer_pfx &&
4882 memcmp(&ifp->peer_addr, cfg->peer_pfx, sizeof(struct in6_addr))) {
4883 if (!ipv6_addr_any(&ifp->peer_addr))
4884 cleanup_prefix_route(ifp, expires, true, true);
4885 new_peer = true;
4886 }
4887
4888 spin_lock_bh(&ifp->lock);
4889 was_managetempaddr = ifp->flags & IFA_F_MANAGETEMPADDR;
4890 had_prefixroute = ifp->flags & IFA_F_PERMANENT &&
4891 !(ifp->flags & IFA_F_NOPREFIXROUTE);
4892 ifp->flags &= ~(IFA_F_DEPRECATED | IFA_F_PERMANENT | IFA_F_NODAD |
4893 IFA_F_HOMEADDRESS | IFA_F_MANAGETEMPADDR |
4894 IFA_F_NOPREFIXROUTE);
4895 ifp->flags |= cfg->ifa_flags;
4896 WRITE_ONCE(ifp->tstamp, jiffies);
4897 WRITE_ONCE(ifp->valid_lft, cfg->valid_lft);
4898 WRITE_ONCE(ifp->prefered_lft, cfg->preferred_lft);
4899 WRITE_ONCE(ifp->ifa_proto, cfg->ifa_proto);
4900
4901 if (cfg->rt_priority && cfg->rt_priority != ifp->rt_priority)
4902 WRITE_ONCE(ifp->rt_priority, cfg->rt_priority);
4903
4904 if (new_peer)
4905 ifp->peer_addr = *cfg->peer_pfx;
4906
4907 spin_unlock_bh(&ifp->lock);
4908 if (!(ifp->flags&IFA_F_TENTATIVE))
4909 ipv6_ifa_notify(0, ifp);
4910
4911 if (!(cfg->ifa_flags & IFA_F_NOPREFIXROUTE)) {
4912 int rc = -ENOENT;
4913
4914 if (had_prefixroute)
4915 rc = modify_prefix_route(net, ifp, expires, flags, false);
4916
4917 /* prefix route could have been deleted; if so restore it */
4918 if (rc == -ENOENT) {
4919 addrconf_prefix_route(&ifp->addr, ifp->prefix_len,
4920 ifp->rt_priority, ifp->idev->dev,
4921 expires, flags, GFP_KERNEL);
4922 }
4923
4924 if (had_prefixroute && !ipv6_addr_any(&ifp->peer_addr))
4925 rc = modify_prefix_route(net, ifp, expires, flags, true);
4926
4927 if (rc == -ENOENT && !ipv6_addr_any(&ifp->peer_addr)) {
4928 addrconf_prefix_route(&ifp->peer_addr, ifp->prefix_len,
4929 ifp->rt_priority, ifp->idev->dev,
4930 expires, flags, GFP_KERNEL);
4931 }
4932 } else if (had_prefixroute) {
4933 enum cleanup_prefix_rt_t action;
4934 unsigned long rt_expires;
4935
4936 write_lock_bh(&ifp->idev->lock);
4937 action = check_cleanup_prefix_route(ifp, &rt_expires);
4938 write_unlock_bh(&ifp->idev->lock);
4939
4940 if (action != CLEANUP_PREFIX_RT_NOP) {
4941 cleanup_prefix_route(ifp, rt_expires,
4942 action == CLEANUP_PREFIX_RT_DEL, false);
4943 }
4944 }
4945
4946 if (was_managetempaddr || ifp->flags & IFA_F_MANAGETEMPADDR) {
4947 if (was_managetempaddr && !(ifp->flags & IFA_F_MANAGETEMPADDR))
4948 delete_tempaddrs(ifp->idev, ifp);
4949 else
4950 manage_tempaddrs(ifp->idev, ifp, cfg->valid_lft,
4951 cfg->preferred_lft, !was_managetempaddr,
4952 jiffies);
4953 }
4954
4955 addrconf_verify_rtnl(net);
4956
4957 return 0;
4958 }
4959
4960 static int
inet6_rtm_newaddr(struct sk_buff * skb,struct nlmsghdr * nlh,struct netlink_ext_ack * extack)4961 inet6_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh,
4962 struct netlink_ext_ack *extack)
4963 {
4964 struct net *net = sock_net(skb->sk);
4965 struct nlattr *tb[IFA_MAX+1];
4966 struct in6_addr *peer_pfx;
4967 struct inet6_ifaddr *ifa;
4968 struct net_device *dev;
4969 struct inet6_dev *idev;
4970 struct ifa6_config cfg;
4971 struct ifaddrmsg *ifm;
4972 unsigned long timeout;
4973 clock_t expires;
4974 u32 flags;
4975 int err;
4976
4977 err = nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFA_MAX,
4978 ifa_ipv6_policy, extack);
4979 if (err < 0)
4980 return err;
4981
4982 memset(&cfg, 0, sizeof(cfg));
4983
4984 ifm = nlmsg_data(nlh);
4985 cfg.pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
4986 if (!cfg.pfx)
4987 return -EINVAL;
4988
4989 cfg.peer_pfx = peer_pfx;
4990 cfg.plen = ifm->ifa_prefixlen;
4991 if (tb[IFA_RT_PRIORITY])
4992 cfg.rt_priority = nla_get_u32(tb[IFA_RT_PRIORITY]);
4993
4994 if (tb[IFA_PROTO])
4995 cfg.ifa_proto = nla_get_u8(tb[IFA_PROTO]);
4996
4997 cfg.ifa_flags = nla_get_u32_default(tb[IFA_FLAGS], ifm->ifa_flags);
4998
4999 /* We ignore other flags so far. */
5000 cfg.ifa_flags &= IFA_F_NODAD | IFA_F_HOMEADDRESS |
5001 IFA_F_MANAGETEMPADDR | IFA_F_NOPREFIXROUTE |
5002 IFA_F_MCAUTOJOIN | IFA_F_OPTIMISTIC;
5003
5004 cfg.ifa_flags |= IFA_F_PERMANENT;
5005 cfg.valid_lft = INFINITY_LIFE_TIME;
5006 cfg.preferred_lft = INFINITY_LIFE_TIME;
5007 expires = 0;
5008 flags = 0;
5009
5010 if (tb[IFA_CACHEINFO]) {
5011 struct ifa_cacheinfo *ci;
5012
5013 ci = nla_data(tb[IFA_CACHEINFO]);
5014 cfg.valid_lft = ci->ifa_valid;
5015 cfg.preferred_lft = ci->ifa_prefered;
5016
5017 if (!cfg.valid_lft || cfg.preferred_lft > cfg.valid_lft) {
5018 NL_SET_ERR_MSG_MOD(extack, "address lifetime invalid");
5019 return -EINVAL;
5020 }
5021
5022 timeout = addrconf_timeout_fixup(cfg.valid_lft, HZ);
5023 if (addrconf_finite_timeout(timeout)) {
5024 cfg.ifa_flags &= ~IFA_F_PERMANENT;
5025 cfg.valid_lft = timeout;
5026 expires = jiffies_to_clock_t(timeout * HZ);
5027 flags = RTF_EXPIRES;
5028 }
5029
5030 timeout = addrconf_timeout_fixup(cfg.preferred_lft, HZ);
5031 if (addrconf_finite_timeout(timeout)) {
5032 if (timeout == 0)
5033 cfg.ifa_flags |= IFA_F_DEPRECATED;
5034
5035 cfg.preferred_lft = timeout;
5036 }
5037 }
5038
5039 rtnl_net_lock(net);
5040
5041 dev = __dev_get_by_index(net, ifm->ifa_index);
5042 if (!dev) {
5043 NL_SET_ERR_MSG_MOD(extack, "Unable to find the interface");
5044 err = -ENODEV;
5045 goto unlock_rtnl;
5046 }
5047
5048 netdev_lock_ops(dev);
5049 idev = ipv6_find_idev(dev);
5050 if (IS_ERR(idev)) {
5051 err = PTR_ERR(idev);
5052 goto unlock;
5053 }
5054
5055 if (!ipv6_allow_optimistic_dad(net, idev))
5056 cfg.ifa_flags &= ~IFA_F_OPTIMISTIC;
5057
5058 if (cfg.ifa_flags & IFA_F_NODAD &&
5059 cfg.ifa_flags & IFA_F_OPTIMISTIC) {
5060 NL_SET_ERR_MSG(extack, "IFA_F_NODAD and IFA_F_OPTIMISTIC are mutually exclusive");
5061 err = -EINVAL;
5062 goto unlock;
5063 }
5064
5065 ifa = ipv6_get_ifaddr(net, cfg.pfx, dev, 1);
5066 if (!ifa) {
5067 /*
5068 * It would be best to check for !NLM_F_CREATE here but
5069 * userspace already relies on not having to provide this.
5070 */
5071 err = inet6_addr_add(net, dev, &cfg, expires, flags, extack);
5072 goto unlock;
5073 }
5074
5075 if (nlh->nlmsg_flags & NLM_F_EXCL ||
5076 !(nlh->nlmsg_flags & NLM_F_REPLACE)) {
5077 NL_SET_ERR_MSG_MOD(extack, "address already assigned");
5078 err = -EEXIST;
5079 } else {
5080 err = inet6_addr_modify(net, ifa, &cfg, expires, flags);
5081 }
5082
5083 in6_ifa_put(ifa);
5084 unlock:
5085 netdev_unlock_ops(dev);
5086 unlock_rtnl:
5087 rtnl_net_unlock(net);
5088
5089 return err;
5090 }
5091
put_ifaddrmsg(struct nlmsghdr * nlh,u8 prefixlen,u32 flags,u8 scope,int ifindex)5092 static void put_ifaddrmsg(struct nlmsghdr *nlh, u8 prefixlen, u32 flags,
5093 u8 scope, int ifindex)
5094 {
5095 struct ifaddrmsg *ifm;
5096
5097 ifm = nlmsg_data(nlh);
5098 ifm->ifa_family = AF_INET6;
5099 ifm->ifa_prefixlen = prefixlen;
5100 ifm->ifa_flags = flags;
5101 ifm->ifa_scope = scope;
5102 ifm->ifa_index = ifindex;
5103 }
5104
put_cacheinfo(struct sk_buff * skb,unsigned long cstamp,unsigned long tstamp,u32 preferred,u32 valid)5105 static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
5106 unsigned long tstamp, u32 preferred, u32 valid)
5107 {
5108 struct ifa_cacheinfo ci;
5109
5110 ci.cstamp = cstamp_delta(cstamp);
5111 ci.tstamp = cstamp_delta(tstamp);
5112 ci.ifa_prefered = preferred;
5113 ci.ifa_valid = valid;
5114
5115 return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
5116 }
5117
rt_scope(int ifa_scope)5118 static inline int rt_scope(int ifa_scope)
5119 {
5120 if (ifa_scope & IFA_HOST)
5121 return RT_SCOPE_HOST;
5122 else if (ifa_scope & IFA_LINK)
5123 return RT_SCOPE_LINK;
5124 else if (ifa_scope & IFA_SITE)
5125 return RT_SCOPE_SITE;
5126 else
5127 return RT_SCOPE_UNIVERSE;
5128 }
5129
inet6_ifaddr_msgsize(void)5130 static inline int inet6_ifaddr_msgsize(void)
5131 {
5132 return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
5133 + nla_total_size(16) /* IFA_LOCAL */
5134 + nla_total_size(16) /* IFA_ADDRESS */
5135 + nla_total_size(sizeof(struct ifa_cacheinfo))
5136 + nla_total_size(4) /* IFA_FLAGS */
5137 + nla_total_size(1) /* IFA_PROTO */
5138 + nla_total_size(4) /* IFA_RT_PRIORITY */;
5139 }
5140
inet6_fill_ifaddr(struct sk_buff * skb,const struct inet6_ifaddr * ifa,struct inet6_fill_args * args)5141 static int inet6_fill_ifaddr(struct sk_buff *skb,
5142 const struct inet6_ifaddr *ifa,
5143 struct inet6_fill_args *args)
5144 {
5145 struct nlmsghdr *nlh;
5146 u32 preferred, valid;
5147 u32 flags, priority;
5148 u8 proto;
5149
5150 nlh = nlmsg_put(skb, args->portid, args->seq, args->event,
5151 sizeof(struct ifaddrmsg), args->flags);
5152 if (!nlh)
5153 return -EMSGSIZE;
5154
5155 flags = READ_ONCE(ifa->flags);
5156 put_ifaddrmsg(nlh, ifa->prefix_len, ifa->flags, rt_scope(ifa->scope),
5157 ifa->idev->dev->ifindex);
5158
5159 if (args->netnsid >= 0 &&
5160 nla_put_s32(skb, IFA_TARGET_NETNSID, args->netnsid))
5161 goto error;
5162
5163 preferred = READ_ONCE(ifa->prefered_lft);
5164 valid = READ_ONCE(ifa->valid_lft);
5165
5166 if (!((flags & IFA_F_PERMANENT) &&
5167 (preferred == INFINITY_LIFE_TIME))) {
5168 if (preferred != INFINITY_LIFE_TIME) {
5169 long tval = (jiffies - READ_ONCE(ifa->tstamp)) / HZ;
5170
5171 if (preferred > tval)
5172 preferred -= tval;
5173 else
5174 preferred = 0;
5175 if (valid != INFINITY_LIFE_TIME) {
5176 if (valid > tval)
5177 valid -= tval;
5178 else
5179 valid = 0;
5180 }
5181 }
5182 } else {
5183 preferred = INFINITY_LIFE_TIME;
5184 valid = INFINITY_LIFE_TIME;
5185 }
5186
5187 if (!ipv6_addr_any(&ifa->peer_addr)) {
5188 if (nla_put_in6_addr(skb, IFA_LOCAL, &ifa->addr) < 0 ||
5189 nla_put_in6_addr(skb, IFA_ADDRESS, &ifa->peer_addr) < 0)
5190 goto error;
5191 } else {
5192 if (nla_put_in6_addr(skb, IFA_ADDRESS, &ifa->addr) < 0)
5193 goto error;
5194 }
5195
5196 priority = READ_ONCE(ifa->rt_priority);
5197 if (priority && nla_put_u32(skb, IFA_RT_PRIORITY, priority))
5198 goto error;
5199
5200 if (put_cacheinfo(skb, ifa->cstamp, READ_ONCE(ifa->tstamp),
5201 preferred, valid) < 0)
5202 goto error;
5203
5204 if (nla_put_u32(skb, IFA_FLAGS, flags) < 0)
5205 goto error;
5206
5207 proto = READ_ONCE(ifa->ifa_proto);
5208 if (proto && nla_put_u8(skb, IFA_PROTO, proto))
5209 goto error;
5210
5211 nlmsg_end(skb, nlh);
5212 return 0;
5213
5214 error:
5215 nlmsg_cancel(skb, nlh);
5216 return -EMSGSIZE;
5217 }
5218
inet6_fill_ifmcaddr(struct sk_buff * skb,const struct ifmcaddr6 * ifmca,struct inet6_fill_args * args)5219 int inet6_fill_ifmcaddr(struct sk_buff *skb,
5220 const struct ifmcaddr6 *ifmca,
5221 struct inet6_fill_args *args)
5222 {
5223 int ifindex = ifmca->idev->dev->ifindex;
5224 u8 scope = RT_SCOPE_UNIVERSE;
5225 struct nlmsghdr *nlh;
5226
5227 if (!args->force_rt_scope_universe &&
5228 ipv6_addr_scope(&ifmca->mca_addr) & IFA_SITE)
5229 scope = RT_SCOPE_SITE;
5230
5231 nlh = nlmsg_put(skb, args->portid, args->seq, args->event,
5232 sizeof(struct ifaddrmsg), args->flags);
5233 if (!nlh)
5234 return -EMSGSIZE;
5235
5236 if (args->netnsid >= 0 &&
5237 nla_put_s32(skb, IFA_TARGET_NETNSID, args->netnsid)) {
5238 nlmsg_cancel(skb, nlh);
5239 return -EMSGSIZE;
5240 }
5241
5242 put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
5243 if (nla_put_in6_addr(skb, IFA_MULTICAST, &ifmca->mca_addr) < 0 ||
5244 put_cacheinfo(skb, ifmca->mca_cstamp, READ_ONCE(ifmca->mca_tstamp),
5245 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
5246 nlmsg_cancel(skb, nlh);
5247 return -EMSGSIZE;
5248 }
5249
5250 nlmsg_end(skb, nlh);
5251 return 0;
5252 }
5253
inet6_fill_ifacaddr(struct sk_buff * skb,const struct ifacaddr6 * ifaca,struct inet6_fill_args * args)5254 int inet6_fill_ifacaddr(struct sk_buff *skb,
5255 const struct ifacaddr6 *ifaca,
5256 struct inet6_fill_args *args)
5257 {
5258 struct net_device *dev = fib6_info_nh_dev(ifaca->aca_rt);
5259 int ifindex = dev ? dev->ifindex : 1;
5260 u8 scope = RT_SCOPE_UNIVERSE;
5261 struct nlmsghdr *nlh;
5262
5263 if (ipv6_addr_scope(&ifaca->aca_addr) & IFA_SITE)
5264 scope = RT_SCOPE_SITE;
5265
5266 nlh = nlmsg_put(skb, args->portid, args->seq, args->event,
5267 sizeof(struct ifaddrmsg), args->flags);
5268 if (!nlh)
5269 return -EMSGSIZE;
5270
5271 if (args->netnsid >= 0 &&
5272 nla_put_s32(skb, IFA_TARGET_NETNSID, args->netnsid)) {
5273 nlmsg_cancel(skb, nlh);
5274 return -EMSGSIZE;
5275 }
5276
5277 put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
5278 if (nla_put_in6_addr(skb, IFA_ANYCAST, &ifaca->aca_addr) < 0 ||
5279 put_cacheinfo(skb, ifaca->aca_cstamp, READ_ONCE(ifaca->aca_tstamp),
5280 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
5281 nlmsg_cancel(skb, nlh);
5282 return -EMSGSIZE;
5283 }
5284
5285 nlmsg_end(skb, nlh);
5286 return 0;
5287 }
5288
5289 /* called with rcu_read_lock() */
in6_dump_addrs(const struct inet6_dev * idev,struct sk_buff * skb,struct netlink_callback * cb,int * s_ip_idx,struct inet6_fill_args * fillargs)5290 static int in6_dump_addrs(const struct inet6_dev *idev, struct sk_buff *skb,
5291 struct netlink_callback *cb, int *s_ip_idx,
5292 struct inet6_fill_args *fillargs)
5293 {
5294 const struct ifmcaddr6 *ifmca;
5295 const struct ifacaddr6 *ifaca;
5296 int ip_idx = 0;
5297 int err = 0;
5298
5299 switch (fillargs->type) {
5300 case UNICAST_ADDR: {
5301 const struct inet6_ifaddr *ifa;
5302 fillargs->event = RTM_NEWADDR;
5303
5304 /* unicast address incl. temp addr */
5305 list_for_each_entry_rcu(ifa, &idev->addr_list, if_list) {
5306 if (ip_idx < *s_ip_idx)
5307 goto next;
5308 err = inet6_fill_ifaddr(skb, ifa, fillargs);
5309 if (err < 0)
5310 break;
5311 nl_dump_check_consistent(cb, nlmsg_hdr(skb));
5312 next:
5313 ip_idx++;
5314 }
5315 break;
5316 }
5317 case MULTICAST_ADDR:
5318 fillargs->event = RTM_GETMULTICAST;
5319
5320 /* multicast address */
5321 for (ifmca = rcu_dereference(idev->mc_list);
5322 ifmca;
5323 ifmca = rcu_dereference(ifmca->next), ip_idx++) {
5324 if (ip_idx < *s_ip_idx)
5325 continue;
5326 err = inet6_fill_ifmcaddr(skb, ifmca, fillargs);
5327 if (err < 0)
5328 break;
5329 }
5330 break;
5331 case ANYCAST_ADDR:
5332 fillargs->event = RTM_GETANYCAST;
5333 /* anycast address */
5334 for (ifaca = rcu_dereference(idev->ac_list); ifaca;
5335 ifaca = rcu_dereference(ifaca->aca_next), ip_idx++) {
5336 if (ip_idx < *s_ip_idx)
5337 continue;
5338 err = inet6_fill_ifacaddr(skb, ifaca, fillargs);
5339 if (err < 0)
5340 break;
5341 }
5342 break;
5343 default:
5344 break;
5345 }
5346 *s_ip_idx = err ? ip_idx : 0;
5347 return err;
5348 }
5349
inet6_valid_dump_ifaddr_req(const struct nlmsghdr * nlh,struct inet6_fill_args * fillargs,struct net ** tgt_net,struct sock * sk,struct netlink_callback * cb)5350 static int inet6_valid_dump_ifaddr_req(const struct nlmsghdr *nlh,
5351 struct inet6_fill_args *fillargs,
5352 struct net **tgt_net, struct sock *sk,
5353 struct netlink_callback *cb)
5354 {
5355 struct netlink_ext_ack *extack = cb->extack;
5356 struct nlattr *tb[IFA_MAX+1];
5357 struct ifaddrmsg *ifm;
5358 int err, i;
5359
5360 ifm = nlmsg_payload(nlh, sizeof(*ifm));
5361 if (!ifm) {
5362 NL_SET_ERR_MSG_MOD(extack, "Invalid header for address dump request");
5363 return -EINVAL;
5364 }
5365
5366 if (ifm->ifa_prefixlen || ifm->ifa_flags || ifm->ifa_scope) {
5367 NL_SET_ERR_MSG_MOD(extack, "Invalid values in header for address dump request");
5368 return -EINVAL;
5369 }
5370
5371 fillargs->ifindex = ifm->ifa_index;
5372 if (fillargs->ifindex) {
5373 cb->answer_flags |= NLM_F_DUMP_FILTERED;
5374 fillargs->flags |= NLM_F_DUMP_FILTERED;
5375 }
5376
5377 err = nlmsg_parse_deprecated_strict(nlh, sizeof(*ifm), tb, IFA_MAX,
5378 ifa_ipv6_policy, extack);
5379 if (err < 0)
5380 return err;
5381
5382 for (i = 0; i <= IFA_MAX; ++i) {
5383 if (!tb[i])
5384 continue;
5385
5386 if (i == IFA_TARGET_NETNSID) {
5387 struct net *net;
5388
5389 fillargs->netnsid = nla_get_s32(tb[i]);
5390 net = rtnl_get_net_ns_capable(sk, fillargs->netnsid);
5391 if (IS_ERR(net)) {
5392 fillargs->netnsid = -1;
5393 NL_SET_ERR_MSG_MOD(extack, "Invalid target network namespace id");
5394 return PTR_ERR(net);
5395 }
5396 *tgt_net = net;
5397 } else {
5398 NL_SET_ERR_MSG_MOD(extack, "Unsupported attribute in dump request");
5399 return -EINVAL;
5400 }
5401 }
5402
5403 return 0;
5404 }
5405
inet6_dump_addr(struct sk_buff * skb,struct netlink_callback * cb,enum addr_type_t type)5406 static int inet6_dump_addr(struct sk_buff *skb, struct netlink_callback *cb,
5407 enum addr_type_t type)
5408 {
5409 struct net *tgt_net = sock_net(skb->sk);
5410 const struct nlmsghdr *nlh = cb->nlh;
5411 struct inet6_fill_args fillargs = {
5412 .portid = NETLINK_CB(cb->skb).portid,
5413 .seq = cb->nlh->nlmsg_seq,
5414 .flags = NLM_F_MULTI,
5415 .netnsid = -1,
5416 .type = type,
5417 .force_rt_scope_universe = false,
5418 };
5419 struct {
5420 unsigned long ifindex;
5421 int ip_idx;
5422 } *ctx = (void *)cb->ctx;
5423 struct net_device *dev;
5424 struct inet6_dev *idev;
5425 int err = 0;
5426
5427 rcu_read_lock();
5428 if (cb->strict_check) {
5429 err = inet6_valid_dump_ifaddr_req(nlh, &fillargs, &tgt_net,
5430 skb->sk, cb);
5431 if (err < 0)
5432 goto done;
5433
5434 err = 0;
5435 if (fillargs.ifindex) {
5436 dev = dev_get_by_index_rcu(tgt_net, fillargs.ifindex);
5437 if (!dev) {
5438 err = -ENODEV;
5439 goto done;
5440 }
5441 idev = __in6_dev_get(dev);
5442 if (idev)
5443 err = in6_dump_addrs(idev, skb, cb,
5444 &ctx->ip_idx,
5445 &fillargs);
5446 goto done;
5447 }
5448 }
5449
5450 cb->seq = inet6_base_seq(tgt_net);
5451 for_each_netdev_dump(tgt_net, dev, ctx->ifindex) {
5452 idev = __in6_dev_get(dev);
5453 if (!idev)
5454 continue;
5455 err = in6_dump_addrs(idev, skb, cb, &ctx->ip_idx,
5456 &fillargs);
5457 if (err < 0)
5458 goto done;
5459 }
5460 done:
5461 rcu_read_unlock();
5462 if (fillargs.netnsid >= 0)
5463 put_net(tgt_net);
5464
5465 return err;
5466 }
5467
inet6_dump_ifaddr(struct sk_buff * skb,struct netlink_callback * cb)5468 static int inet6_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
5469 {
5470 enum addr_type_t type = UNICAST_ADDR;
5471
5472 return inet6_dump_addr(skb, cb, type);
5473 }
5474
inet6_dump_ifmcaddr(struct sk_buff * skb,struct netlink_callback * cb)5475 static int inet6_dump_ifmcaddr(struct sk_buff *skb, struct netlink_callback *cb)
5476 {
5477 enum addr_type_t type = MULTICAST_ADDR;
5478
5479 return inet6_dump_addr(skb, cb, type);
5480 }
5481
5482
inet6_dump_ifacaddr(struct sk_buff * skb,struct netlink_callback * cb)5483 static int inet6_dump_ifacaddr(struct sk_buff *skb, struct netlink_callback *cb)
5484 {
5485 enum addr_type_t type = ANYCAST_ADDR;
5486
5487 return inet6_dump_addr(skb, cb, type);
5488 }
5489
inet6_rtm_valid_getaddr_req(struct sk_buff * skb,const struct nlmsghdr * nlh,struct nlattr ** tb,struct netlink_ext_ack * extack)5490 static int inet6_rtm_valid_getaddr_req(struct sk_buff *skb,
5491 const struct nlmsghdr *nlh,
5492 struct nlattr **tb,
5493 struct netlink_ext_ack *extack)
5494 {
5495 struct ifaddrmsg *ifm;
5496 int i, err;
5497
5498 ifm = nlmsg_payload(nlh, sizeof(*ifm));
5499 if (!ifm) {
5500 NL_SET_ERR_MSG_MOD(extack, "Invalid header for get address request");
5501 return -EINVAL;
5502 }
5503
5504 if (!netlink_strict_get_check(skb))
5505 return nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFA_MAX,
5506 ifa_ipv6_policy, extack);
5507
5508 if (ifm->ifa_prefixlen || ifm->ifa_flags || ifm->ifa_scope) {
5509 NL_SET_ERR_MSG_MOD(extack, "Invalid values in header for get address request");
5510 return -EINVAL;
5511 }
5512
5513 err = nlmsg_parse_deprecated_strict(nlh, sizeof(*ifm), tb, IFA_MAX,
5514 ifa_ipv6_policy, extack);
5515 if (err)
5516 return err;
5517
5518 for (i = 0; i <= IFA_MAX; i++) {
5519 if (!tb[i])
5520 continue;
5521
5522 switch (i) {
5523 case IFA_TARGET_NETNSID:
5524 case IFA_ADDRESS:
5525 case IFA_LOCAL:
5526 break;
5527 default:
5528 NL_SET_ERR_MSG_MOD(extack, "Unsupported attribute in get address request");
5529 return -EINVAL;
5530 }
5531 }
5532
5533 return 0;
5534 }
5535
inet6_rtm_getaddr(struct sk_buff * in_skb,struct nlmsghdr * nlh,struct netlink_ext_ack * extack)5536 static int inet6_rtm_getaddr(struct sk_buff *in_skb, struct nlmsghdr *nlh,
5537 struct netlink_ext_ack *extack)
5538 {
5539 struct net *tgt_net = sock_net(in_skb->sk);
5540 struct inet6_fill_args fillargs = {
5541 .portid = NETLINK_CB(in_skb).portid,
5542 .seq = nlh->nlmsg_seq,
5543 .event = RTM_NEWADDR,
5544 .flags = 0,
5545 .netnsid = -1,
5546 .force_rt_scope_universe = false,
5547 };
5548 struct ifaddrmsg *ifm;
5549 struct nlattr *tb[IFA_MAX+1];
5550 struct in6_addr *addr = NULL, *peer;
5551 struct net_device *dev = NULL;
5552 struct inet6_ifaddr *ifa;
5553 struct sk_buff *skb;
5554 int err;
5555
5556 err = inet6_rtm_valid_getaddr_req(in_skb, nlh, tb, extack);
5557 if (err < 0)
5558 return err;
5559
5560 if (tb[IFA_TARGET_NETNSID]) {
5561 fillargs.netnsid = nla_get_s32(tb[IFA_TARGET_NETNSID]);
5562
5563 tgt_net = rtnl_get_net_ns_capable(NETLINK_CB(in_skb).sk,
5564 fillargs.netnsid);
5565 if (IS_ERR(tgt_net))
5566 return PTR_ERR(tgt_net);
5567 }
5568
5569 addr = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer);
5570 if (!addr) {
5571 err = -EINVAL;
5572 goto errout;
5573 }
5574 ifm = nlmsg_data(nlh);
5575 if (ifm->ifa_index)
5576 dev = dev_get_by_index(tgt_net, ifm->ifa_index);
5577
5578 ifa = ipv6_get_ifaddr(tgt_net, addr, dev, 1);
5579 if (!ifa) {
5580 err = -EADDRNOTAVAIL;
5581 goto errout;
5582 }
5583
5584 skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_KERNEL);
5585 if (!skb) {
5586 err = -ENOBUFS;
5587 goto errout_ifa;
5588 }
5589
5590 err = inet6_fill_ifaddr(skb, ifa, &fillargs);
5591 if (err < 0) {
5592 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
5593 WARN_ON(err == -EMSGSIZE);
5594 kfree_skb(skb);
5595 goto errout_ifa;
5596 }
5597 err = rtnl_unicast(skb, tgt_net, NETLINK_CB(in_skb).portid);
5598 errout_ifa:
5599 in6_ifa_put(ifa);
5600 errout:
5601 dev_put(dev);
5602 if (fillargs.netnsid >= 0)
5603 put_net(tgt_net);
5604
5605 return err;
5606 }
5607
inet6_ifa_notify(int event,struct inet6_ifaddr * ifa)5608 static void inet6_ifa_notify(int event, struct inet6_ifaddr *ifa)
5609 {
5610 struct sk_buff *skb;
5611 struct net *net = dev_net(ifa->idev->dev);
5612 struct inet6_fill_args fillargs = {
5613 .portid = 0,
5614 .seq = 0,
5615 .event = event,
5616 .flags = 0,
5617 .netnsid = -1,
5618 .force_rt_scope_universe = false,
5619 };
5620 int err = -ENOBUFS;
5621
5622 skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_ATOMIC);
5623 if (!skb)
5624 goto errout;
5625
5626 err = inet6_fill_ifaddr(skb, ifa, &fillargs);
5627 if (err < 0) {
5628 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
5629 WARN_ON(err == -EMSGSIZE);
5630 kfree_skb(skb);
5631 goto errout;
5632 }
5633 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
5634 return;
5635 errout:
5636 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
5637 }
5638
ipv6_store_devconf(const struct ipv6_devconf * cnf,__s32 * array,int bytes)5639 static void ipv6_store_devconf(const struct ipv6_devconf *cnf,
5640 __s32 *array, int bytes)
5641 {
5642 BUG_ON(bytes < (DEVCONF_MAX * 4));
5643
5644 memset(array, 0, bytes);
5645 array[DEVCONF_FORWARDING] = READ_ONCE(cnf->forwarding);
5646 array[DEVCONF_HOPLIMIT] = READ_ONCE(cnf->hop_limit);
5647 array[DEVCONF_MTU6] = READ_ONCE(cnf->mtu6);
5648 array[DEVCONF_ACCEPT_RA] = READ_ONCE(cnf->accept_ra);
5649 array[DEVCONF_ACCEPT_REDIRECTS] = READ_ONCE(cnf->accept_redirects);
5650 array[DEVCONF_AUTOCONF] = READ_ONCE(cnf->autoconf);
5651 array[DEVCONF_DAD_TRANSMITS] = READ_ONCE(cnf->dad_transmits);
5652 array[DEVCONF_RTR_SOLICITS] = READ_ONCE(cnf->rtr_solicits);
5653 array[DEVCONF_RTR_SOLICIT_INTERVAL] =
5654 jiffies_to_msecs(READ_ONCE(cnf->rtr_solicit_interval));
5655 array[DEVCONF_RTR_SOLICIT_MAX_INTERVAL] =
5656 jiffies_to_msecs(READ_ONCE(cnf->rtr_solicit_max_interval));
5657 array[DEVCONF_RTR_SOLICIT_DELAY] =
5658 jiffies_to_msecs(READ_ONCE(cnf->rtr_solicit_delay));
5659 array[DEVCONF_FORCE_MLD_VERSION] = READ_ONCE(cnf->force_mld_version);
5660 array[DEVCONF_MLDV1_UNSOLICITED_REPORT_INTERVAL] =
5661 jiffies_to_msecs(READ_ONCE(cnf->mldv1_unsolicited_report_interval));
5662 array[DEVCONF_MLDV2_UNSOLICITED_REPORT_INTERVAL] =
5663 jiffies_to_msecs(READ_ONCE(cnf->mldv2_unsolicited_report_interval));
5664 array[DEVCONF_USE_TEMPADDR] = READ_ONCE(cnf->use_tempaddr);
5665 array[DEVCONF_TEMP_VALID_LFT] = READ_ONCE(cnf->temp_valid_lft);
5666 array[DEVCONF_TEMP_PREFERED_LFT] = READ_ONCE(cnf->temp_prefered_lft);
5667 array[DEVCONF_REGEN_MAX_RETRY] = READ_ONCE(cnf->regen_max_retry);
5668 array[DEVCONF_MAX_DESYNC_FACTOR] = READ_ONCE(cnf->max_desync_factor);
5669 array[DEVCONF_MAX_ADDRESSES] = READ_ONCE(cnf->max_addresses);
5670 array[DEVCONF_ACCEPT_RA_DEFRTR] = READ_ONCE(cnf->accept_ra_defrtr);
5671 array[DEVCONF_RA_DEFRTR_METRIC] = READ_ONCE(cnf->ra_defrtr_metric);
5672 array[DEVCONF_ACCEPT_RA_MIN_HOP_LIMIT] =
5673 READ_ONCE(cnf->accept_ra_min_hop_limit);
5674 array[DEVCONF_ACCEPT_RA_PINFO] = READ_ONCE(cnf->accept_ra_pinfo);
5675 #ifdef CONFIG_IPV6_ROUTER_PREF
5676 array[DEVCONF_ACCEPT_RA_RTR_PREF] = READ_ONCE(cnf->accept_ra_rtr_pref);
5677 array[DEVCONF_RTR_PROBE_INTERVAL] =
5678 jiffies_to_msecs(READ_ONCE(cnf->rtr_probe_interval));
5679 #ifdef CONFIG_IPV6_ROUTE_INFO
5680 array[DEVCONF_ACCEPT_RA_RT_INFO_MIN_PLEN] =
5681 READ_ONCE(cnf->accept_ra_rt_info_min_plen);
5682 array[DEVCONF_ACCEPT_RA_RT_INFO_MAX_PLEN] =
5683 READ_ONCE(cnf->accept_ra_rt_info_max_plen);
5684 #endif
5685 #endif
5686 array[DEVCONF_PROXY_NDP] = READ_ONCE(cnf->proxy_ndp);
5687 array[DEVCONF_ACCEPT_SOURCE_ROUTE] =
5688 READ_ONCE(cnf->accept_source_route);
5689 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
5690 array[DEVCONF_OPTIMISTIC_DAD] = READ_ONCE(cnf->optimistic_dad);
5691 array[DEVCONF_USE_OPTIMISTIC] = READ_ONCE(cnf->use_optimistic);
5692 #endif
5693 #ifdef CONFIG_IPV6_MROUTE
5694 array[DEVCONF_MC_FORWARDING] = atomic_read(&cnf->mc_forwarding);
5695 #endif
5696 array[DEVCONF_DISABLE_IPV6] = READ_ONCE(cnf->disable_ipv6);
5697 array[DEVCONF_ACCEPT_DAD] = READ_ONCE(cnf->accept_dad);
5698 array[DEVCONF_FORCE_TLLAO] = READ_ONCE(cnf->force_tllao);
5699 array[DEVCONF_NDISC_NOTIFY] = READ_ONCE(cnf->ndisc_notify);
5700 array[DEVCONF_SUPPRESS_FRAG_NDISC] =
5701 READ_ONCE(cnf->suppress_frag_ndisc);
5702 array[DEVCONF_ACCEPT_RA_FROM_LOCAL] =
5703 READ_ONCE(cnf->accept_ra_from_local);
5704 array[DEVCONF_ACCEPT_RA_MTU] = READ_ONCE(cnf->accept_ra_mtu);
5705 array[DEVCONF_IGNORE_ROUTES_WITH_LINKDOWN] =
5706 READ_ONCE(cnf->ignore_routes_with_linkdown);
5707 /* we omit DEVCONF_STABLE_SECRET for now */
5708 array[DEVCONF_USE_OIF_ADDRS_ONLY] = READ_ONCE(cnf->use_oif_addrs_only);
5709 array[DEVCONF_DROP_UNICAST_IN_L2_MULTICAST] =
5710 READ_ONCE(cnf->drop_unicast_in_l2_multicast);
5711 array[DEVCONF_DROP_UNSOLICITED_NA] = READ_ONCE(cnf->drop_unsolicited_na);
5712 array[DEVCONF_KEEP_ADDR_ON_DOWN] = READ_ONCE(cnf->keep_addr_on_down);
5713 array[DEVCONF_SEG6_ENABLED] = READ_ONCE(cnf->seg6_enabled);
5714 #ifdef CONFIG_IPV6_SEG6_HMAC
5715 array[DEVCONF_SEG6_REQUIRE_HMAC] = READ_ONCE(cnf->seg6_require_hmac);
5716 #endif
5717 array[DEVCONF_ENHANCED_DAD] = READ_ONCE(cnf->enhanced_dad);
5718 array[DEVCONF_ADDR_GEN_MODE] = READ_ONCE(cnf->addr_gen_mode);
5719 array[DEVCONF_DISABLE_POLICY] = READ_ONCE(cnf->disable_policy);
5720 array[DEVCONF_NDISC_TCLASS] = READ_ONCE(cnf->ndisc_tclass);
5721 array[DEVCONF_RPL_SEG_ENABLED] = READ_ONCE(cnf->rpl_seg_enabled);
5722 array[DEVCONF_IOAM6_ENABLED] = READ_ONCE(cnf->ioam6_enabled);
5723 array[DEVCONF_IOAM6_ID] = READ_ONCE(cnf->ioam6_id);
5724 array[DEVCONF_IOAM6_ID_WIDE] = READ_ONCE(cnf->ioam6_id_wide);
5725 array[DEVCONF_NDISC_EVICT_NOCARRIER] =
5726 READ_ONCE(cnf->ndisc_evict_nocarrier);
5727 array[DEVCONF_ACCEPT_UNTRACKED_NA] =
5728 READ_ONCE(cnf->accept_untracked_na);
5729 array[DEVCONF_ACCEPT_RA_MIN_LFT] = READ_ONCE(cnf->accept_ra_min_lft);
5730 array[DEVCONF_FORCE_FORWARDING] = READ_ONCE(cnf->force_forwarding);
5731 }
5732
inet6_ifla6_size(void)5733 static inline size_t inet6_ifla6_size(void)
5734 {
5735 return nla_total_size(4) /* IFLA_INET6_FLAGS */
5736 + nla_total_size(sizeof(struct ifla_cacheinfo))
5737 + nla_total_size(DEVCONF_MAX * 4) /* IFLA_INET6_CONF */
5738 + nla_total_size(IPSTATS_MIB_MAX * 8) /* IFLA_INET6_STATS */
5739 + nla_total_size(ICMP6_MIB_MAX * 8) /* IFLA_INET6_ICMP6STATS */
5740 + nla_total_size(sizeof(struct in6_addr)) /* IFLA_INET6_TOKEN */
5741 + nla_total_size(1) /* IFLA_INET6_ADDR_GEN_MODE */
5742 + nla_total_size(4) /* IFLA_INET6_RA_MTU */
5743 + 0;
5744 }
5745
inet6_if_nlmsg_size(void)5746 static inline size_t inet6_if_nlmsg_size(void)
5747 {
5748 return NLMSG_ALIGN(sizeof(struct ifinfomsg))
5749 + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
5750 + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
5751 + nla_total_size(4) /* IFLA_MTU */
5752 + nla_total_size(4) /* IFLA_LINK */
5753 + nla_total_size(1) /* IFLA_OPERSTATE */
5754 + nla_total_size(inet6_ifla6_size()); /* IFLA_PROTINFO */
5755 }
5756
__snmp6_fill_statsdev(u64 * stats,atomic_long_t * mib,int bytes)5757 static inline void __snmp6_fill_statsdev(u64 *stats, atomic_long_t *mib,
5758 int bytes)
5759 {
5760 int i;
5761 int pad = bytes - sizeof(u64) * ICMP6_MIB_MAX;
5762 BUG_ON(pad < 0);
5763
5764 /* Use put_unaligned() because stats may not be aligned for u64. */
5765 put_unaligned(ICMP6_MIB_MAX, &stats[0]);
5766 for (i = 1; i < ICMP6_MIB_MAX; i++)
5767 put_unaligned(atomic_long_read(&mib[i]), &stats[i]);
5768
5769 memset(&stats[ICMP6_MIB_MAX], 0, pad);
5770 }
5771
__snmp6_fill_stats64(u64 * stats,void __percpu * mib,int bytes,size_t syncpoff)5772 static inline void __snmp6_fill_stats64(u64 *stats, void __percpu *mib,
5773 int bytes, size_t syncpoff)
5774 {
5775 int i, c;
5776 u64 buff[IPSTATS_MIB_MAX];
5777 int pad = bytes - sizeof(u64) * IPSTATS_MIB_MAX;
5778
5779 BUG_ON(pad < 0);
5780
5781 memset(buff, 0, sizeof(buff));
5782 buff[0] = IPSTATS_MIB_MAX;
5783
5784 for_each_possible_cpu(c) {
5785 for (i = 1; i < IPSTATS_MIB_MAX; i++)
5786 buff[i] += snmp_get_cpu_field64(mib, c, i, syncpoff);
5787 }
5788
5789 memcpy(stats, buff, IPSTATS_MIB_MAX * sizeof(u64));
5790 memset(&stats[IPSTATS_MIB_MAX], 0, pad);
5791 }
5792
snmp6_fill_stats(u64 * stats,struct inet6_dev * idev,int attrtype,int bytes)5793 static void snmp6_fill_stats(u64 *stats, struct inet6_dev *idev, int attrtype,
5794 int bytes)
5795 {
5796 switch (attrtype) {
5797 case IFLA_INET6_STATS:
5798 __snmp6_fill_stats64(stats, idev->stats.ipv6, bytes,
5799 offsetof(struct ipstats_mib, syncp));
5800 break;
5801 case IFLA_INET6_ICMP6STATS:
5802 __snmp6_fill_statsdev(stats, idev->stats.icmpv6dev->mibs, bytes);
5803 break;
5804 }
5805 }
5806
inet6_fill_ifla6_stats_attrs(struct sk_buff * skb,struct inet6_dev * idev)5807 static int inet6_fill_ifla6_stats_attrs(struct sk_buff *skb,
5808 struct inet6_dev *idev)
5809 {
5810 struct nlattr *nla;
5811
5812 nla = nla_reserve(skb, IFLA_INET6_STATS, IPSTATS_MIB_MAX * sizeof(u64));
5813 if (!nla)
5814 goto nla_put_failure;
5815 snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_STATS, nla_len(nla));
5816
5817 nla = nla_reserve(skb, IFLA_INET6_ICMP6STATS, ICMP6_MIB_MAX * sizeof(u64));
5818 if (!nla)
5819 goto nla_put_failure;
5820 snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_ICMP6STATS, nla_len(nla));
5821
5822 return 0;
5823
5824 nla_put_failure:
5825 return -EMSGSIZE;
5826 }
5827
inet6_fill_ifla6_attrs(struct sk_buff * skb,struct inet6_dev * idev,u32 ext_filter_mask)5828 static int inet6_fill_ifla6_attrs(struct sk_buff *skb, struct inet6_dev *idev,
5829 u32 ext_filter_mask)
5830 {
5831 struct ifla_cacheinfo ci;
5832 struct nlattr *nla;
5833 u32 ra_mtu;
5834
5835 if (nla_put_u32(skb, IFLA_INET6_FLAGS, READ_ONCE(idev->if_flags)))
5836 goto nla_put_failure;
5837 ci.max_reasm_len = IPV6_MAXPLEN;
5838 ci.tstamp = cstamp_delta(READ_ONCE(idev->tstamp));
5839 ci.reachable_time = jiffies_to_msecs(idev->nd_parms->reachable_time);
5840 ci.retrans_time = jiffies_to_msecs(NEIGH_VAR(idev->nd_parms, RETRANS_TIME));
5841 if (nla_put(skb, IFLA_INET6_CACHEINFO, sizeof(ci), &ci))
5842 goto nla_put_failure;
5843 nla = nla_reserve(skb, IFLA_INET6_CONF, DEVCONF_MAX * sizeof(s32));
5844 if (!nla)
5845 goto nla_put_failure;
5846 ipv6_store_devconf(&idev->cnf, nla_data(nla), nla_len(nla));
5847
5848 /* XXX - MC not implemented */
5849
5850 if (!(ext_filter_mask & RTEXT_FILTER_SKIP_STATS)) {
5851 if (inet6_fill_ifla6_stats_attrs(skb, idev) < 0)
5852 goto nla_put_failure;
5853 }
5854
5855 nla = nla_reserve(skb, IFLA_INET6_TOKEN, sizeof(struct in6_addr));
5856 if (!nla)
5857 goto nla_put_failure;
5858 read_lock_bh(&idev->lock);
5859 memcpy(nla_data(nla), idev->token.s6_addr, nla_len(nla));
5860 read_unlock_bh(&idev->lock);
5861
5862 if (nla_put_u8(skb, IFLA_INET6_ADDR_GEN_MODE,
5863 READ_ONCE(idev->cnf.addr_gen_mode)))
5864 goto nla_put_failure;
5865
5866 ra_mtu = READ_ONCE(idev->ra_mtu);
5867 if (ra_mtu && nla_put_u32(skb, IFLA_INET6_RA_MTU, ra_mtu))
5868 goto nla_put_failure;
5869
5870 return 0;
5871
5872 nla_put_failure:
5873 return -EMSGSIZE;
5874 }
5875
inet6_get_link_af_size(const struct net_device * dev,u32 ext_filter_mask)5876 static size_t inet6_get_link_af_size(const struct net_device *dev,
5877 u32 ext_filter_mask)
5878 {
5879 if (!__in6_dev_get(dev))
5880 return 0;
5881
5882 return inet6_ifla6_size();
5883 }
5884
inet6_fill_link_af(struct sk_buff * skb,const struct net_device * dev,u32 ext_filter_mask)5885 static int inet6_fill_link_af(struct sk_buff *skb, const struct net_device *dev,
5886 u32 ext_filter_mask)
5887 {
5888 struct inet6_dev *idev = __in6_dev_get(dev);
5889
5890 if (!idev)
5891 return -ENODATA;
5892
5893 if (inet6_fill_ifla6_attrs(skb, idev, ext_filter_mask) < 0)
5894 return -EMSGSIZE;
5895
5896 return 0;
5897 }
5898
inet6_set_iftoken(struct inet6_dev * idev,struct in6_addr * token,struct netlink_ext_ack * extack)5899 static int inet6_set_iftoken(struct inet6_dev *idev, struct in6_addr *token,
5900 struct netlink_ext_ack *extack)
5901 {
5902 struct inet6_ifaddr *ifp;
5903 struct net_device *dev = idev->dev;
5904 bool clear_token, update_rs = false;
5905 struct in6_addr ll_addr;
5906
5907 ASSERT_RTNL();
5908
5909 if (!token)
5910 return -EINVAL;
5911
5912 if (dev->flags & IFF_LOOPBACK) {
5913 NL_SET_ERR_MSG_MOD(extack, "Device is loopback");
5914 return -EINVAL;
5915 }
5916
5917 if (dev->flags & IFF_NOARP) {
5918 NL_SET_ERR_MSG_MOD(extack,
5919 "Device does not do neighbour discovery");
5920 return -EINVAL;
5921 }
5922
5923 if (!ipv6_accept_ra(idev)) {
5924 NL_SET_ERR_MSG_MOD(extack,
5925 "Router advertisement is disabled on device");
5926 return -EINVAL;
5927 }
5928
5929 if (READ_ONCE(idev->cnf.rtr_solicits) == 0) {
5930 NL_SET_ERR_MSG(extack,
5931 "Router solicitation is disabled on device");
5932 return -EINVAL;
5933 }
5934
5935 write_lock_bh(&idev->lock);
5936
5937 BUILD_BUG_ON(sizeof(token->s6_addr) != 16);
5938 memcpy(idev->token.s6_addr + 8, token->s6_addr + 8, 8);
5939
5940 write_unlock_bh(&idev->lock);
5941
5942 clear_token = ipv6_addr_any(token);
5943 if (clear_token)
5944 goto update_lft;
5945
5946 if (!idev->dead && (idev->if_flags & IF_READY) &&
5947 !ipv6_get_lladdr(dev, &ll_addr, IFA_F_TENTATIVE |
5948 IFA_F_OPTIMISTIC)) {
5949 /* If we're not ready, then normal ifup will take care
5950 * of this. Otherwise, we need to request our rs here.
5951 */
5952 ndisc_send_rs(dev, &ll_addr, &in6addr_linklocal_allrouters);
5953 update_rs = true;
5954 }
5955
5956 update_lft:
5957 write_lock_bh(&idev->lock);
5958
5959 if (update_rs) {
5960 idev->if_flags |= IF_RS_SENT;
5961 idev->rs_interval = rfc3315_s14_backoff_init(
5962 READ_ONCE(idev->cnf.rtr_solicit_interval));
5963 idev->rs_probes = 1;
5964 addrconf_mod_rs_timer(idev, idev->rs_interval);
5965 }
5966
5967 /* Well, that's kinda nasty ... */
5968 list_for_each_entry(ifp, &idev->addr_list, if_list) {
5969 spin_lock(&ifp->lock);
5970 if (ifp->tokenized) {
5971 ifp->valid_lft = 0;
5972 ifp->prefered_lft = 0;
5973 }
5974 spin_unlock(&ifp->lock);
5975 }
5976
5977 write_unlock_bh(&idev->lock);
5978 inet6_ifinfo_notify(RTM_NEWLINK, idev);
5979 addrconf_verify_rtnl(dev_net(dev));
5980 return 0;
5981 }
5982
5983 static const struct nla_policy inet6_af_policy[IFLA_INET6_MAX + 1] = {
5984 [IFLA_INET6_ADDR_GEN_MODE] = { .type = NLA_U8 },
5985 [IFLA_INET6_TOKEN] = { .len = sizeof(struct in6_addr) },
5986 [IFLA_INET6_RA_MTU] = { .type = NLA_REJECT,
5987 .reject_message =
5988 "IFLA_INET6_RA_MTU can not be set" },
5989 };
5990
check_addr_gen_mode(int mode)5991 static int check_addr_gen_mode(int mode)
5992 {
5993 if (mode != IN6_ADDR_GEN_MODE_EUI64 &&
5994 mode != IN6_ADDR_GEN_MODE_NONE &&
5995 mode != IN6_ADDR_GEN_MODE_STABLE_PRIVACY &&
5996 mode != IN6_ADDR_GEN_MODE_RANDOM)
5997 return -EINVAL;
5998 return 1;
5999 }
6000
check_stable_privacy(struct inet6_dev * idev,struct net * net,int mode)6001 static int check_stable_privacy(struct inet6_dev *idev, struct net *net,
6002 int mode)
6003 {
6004 if (mode == IN6_ADDR_GEN_MODE_STABLE_PRIVACY &&
6005 !idev->cnf.stable_secret.initialized &&
6006 !net->ipv6.devconf_dflt->stable_secret.initialized)
6007 return -EINVAL;
6008 return 1;
6009 }
6010
inet6_validate_link_af(const struct net_device * dev,const struct nlattr * nla,struct netlink_ext_ack * extack)6011 static int inet6_validate_link_af(const struct net_device *dev,
6012 const struct nlattr *nla,
6013 struct netlink_ext_ack *extack)
6014 {
6015 struct nlattr *tb[IFLA_INET6_MAX + 1];
6016 struct inet6_dev *idev = NULL;
6017 int err;
6018
6019 if (dev) {
6020 idev = __in6_dev_get(dev);
6021 if (!idev)
6022 return -EAFNOSUPPORT;
6023 }
6024
6025 err = nla_parse_nested_deprecated(tb, IFLA_INET6_MAX, nla,
6026 inet6_af_policy, extack);
6027 if (err)
6028 return err;
6029
6030 if (!tb[IFLA_INET6_TOKEN] && !tb[IFLA_INET6_ADDR_GEN_MODE])
6031 return -EINVAL;
6032
6033 if (tb[IFLA_INET6_ADDR_GEN_MODE]) {
6034 u8 mode = nla_get_u8(tb[IFLA_INET6_ADDR_GEN_MODE]);
6035
6036 if (check_addr_gen_mode(mode) < 0)
6037 return -EINVAL;
6038 if (dev && check_stable_privacy(idev, dev_net(dev), mode) < 0)
6039 return -EINVAL;
6040 }
6041
6042 return 0;
6043 }
6044
inet6_set_link_af(struct net_device * dev,const struct nlattr * nla,struct netlink_ext_ack * extack)6045 static int inet6_set_link_af(struct net_device *dev, const struct nlattr *nla,
6046 struct netlink_ext_ack *extack)
6047 {
6048 struct inet6_dev *idev = __in6_dev_get(dev);
6049 struct nlattr *tb[IFLA_INET6_MAX + 1];
6050 int err;
6051
6052 if (!idev)
6053 return -EAFNOSUPPORT;
6054
6055 if (nla_parse_nested_deprecated(tb, IFLA_INET6_MAX, nla, NULL, NULL) < 0)
6056 return -EINVAL;
6057
6058 if (tb[IFLA_INET6_TOKEN]) {
6059 err = inet6_set_iftoken(idev, nla_data(tb[IFLA_INET6_TOKEN]),
6060 extack);
6061 if (err)
6062 return err;
6063 }
6064
6065 if (tb[IFLA_INET6_ADDR_GEN_MODE]) {
6066 u8 mode = nla_get_u8(tb[IFLA_INET6_ADDR_GEN_MODE]);
6067
6068 WRITE_ONCE(idev->cnf.addr_gen_mode, mode);
6069 }
6070
6071 return 0;
6072 }
6073
inet6_fill_ifinfo(struct sk_buff * skb,struct inet6_dev * idev,u32 portid,u32 seq,int event,unsigned int flags)6074 static int inet6_fill_ifinfo(struct sk_buff *skb, struct inet6_dev *idev,
6075 u32 portid, u32 seq, int event, unsigned int flags)
6076 {
6077 struct net_device *dev = idev->dev;
6078 struct ifinfomsg *hdr;
6079 struct nlmsghdr *nlh;
6080 int ifindex, iflink;
6081 void *protoinfo;
6082
6083 nlh = nlmsg_put(skb, portid, seq, event, sizeof(*hdr), flags);
6084 if (!nlh)
6085 return -EMSGSIZE;
6086
6087 hdr = nlmsg_data(nlh);
6088 hdr->ifi_family = AF_INET6;
6089 hdr->__ifi_pad = 0;
6090 hdr->ifi_type = dev->type;
6091 ifindex = READ_ONCE(dev->ifindex);
6092 hdr->ifi_index = ifindex;
6093 hdr->ifi_flags = netif_get_flags(dev);
6094 hdr->ifi_change = 0;
6095
6096 iflink = dev_get_iflink(dev);
6097 if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
6098 (dev->addr_len &&
6099 nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
6100 nla_put_u32(skb, IFLA_MTU, READ_ONCE(dev->mtu)) ||
6101 (ifindex != iflink &&
6102 nla_put_u32(skb, IFLA_LINK, iflink)) ||
6103 nla_put_u8(skb, IFLA_OPERSTATE,
6104 netif_running(dev) ? READ_ONCE(dev->operstate) : IF_OPER_DOWN))
6105 goto nla_put_failure;
6106 protoinfo = nla_nest_start_noflag(skb, IFLA_PROTINFO);
6107 if (!protoinfo)
6108 goto nla_put_failure;
6109
6110 if (inet6_fill_ifla6_attrs(skb, idev, 0) < 0)
6111 goto nla_put_failure;
6112
6113 nla_nest_end(skb, protoinfo);
6114 nlmsg_end(skb, nlh);
6115 return 0;
6116
6117 nla_put_failure:
6118 nlmsg_cancel(skb, nlh);
6119 return -EMSGSIZE;
6120 }
6121
inet6_valid_dump_ifinfo(const struct nlmsghdr * nlh,struct netlink_ext_ack * extack)6122 static int inet6_valid_dump_ifinfo(const struct nlmsghdr *nlh,
6123 struct netlink_ext_ack *extack)
6124 {
6125 struct ifinfomsg *ifm;
6126
6127 ifm = nlmsg_payload(nlh, sizeof(*ifm));
6128 if (!ifm) {
6129 NL_SET_ERR_MSG_MOD(extack, "Invalid header for link dump request");
6130 return -EINVAL;
6131 }
6132
6133 if (nlmsg_attrlen(nlh, sizeof(*ifm))) {
6134 NL_SET_ERR_MSG_MOD(extack, "Invalid data after header");
6135 return -EINVAL;
6136 }
6137
6138 if (ifm->__ifi_pad || ifm->ifi_type || ifm->ifi_flags ||
6139 ifm->ifi_change || ifm->ifi_index) {
6140 NL_SET_ERR_MSG_MOD(extack, "Invalid values in header for dump request");
6141 return -EINVAL;
6142 }
6143
6144 return 0;
6145 }
6146
inet6_dump_ifinfo(struct sk_buff * skb,struct netlink_callback * cb)6147 static int inet6_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
6148 {
6149 struct net *net = sock_net(skb->sk);
6150 struct {
6151 unsigned long ifindex;
6152 } *ctx = (void *)cb->ctx;
6153 struct net_device *dev;
6154 struct inet6_dev *idev;
6155 int err;
6156
6157 /* only requests using strict checking can pass data to
6158 * influence the dump
6159 */
6160 if (cb->strict_check) {
6161 err = inet6_valid_dump_ifinfo(cb->nlh, cb->extack);
6162
6163 if (err < 0)
6164 return err;
6165 }
6166
6167 err = 0;
6168 rcu_read_lock();
6169 for_each_netdev_dump(net, dev, ctx->ifindex) {
6170 idev = __in6_dev_get(dev);
6171 if (!idev)
6172 continue;
6173 err = inet6_fill_ifinfo(skb, idev,
6174 NETLINK_CB(cb->skb).portid,
6175 cb->nlh->nlmsg_seq,
6176 RTM_NEWLINK, NLM_F_MULTI);
6177 if (err < 0)
6178 break;
6179 }
6180 rcu_read_unlock();
6181
6182 return err;
6183 }
6184
inet6_ifinfo_notify(int event,struct inet6_dev * idev)6185 void inet6_ifinfo_notify(int event, struct inet6_dev *idev)
6186 {
6187 struct sk_buff *skb;
6188 struct net *net = dev_net(idev->dev);
6189 int err = -ENOBUFS;
6190
6191 skb = nlmsg_new(inet6_if_nlmsg_size(), GFP_ATOMIC);
6192 if (!skb)
6193 goto errout;
6194
6195 err = inet6_fill_ifinfo(skb, idev, 0, 0, event, 0);
6196 if (err < 0) {
6197 /* -EMSGSIZE implies BUG in inet6_if_nlmsg_size() */
6198 WARN_ON(err == -EMSGSIZE);
6199 kfree_skb(skb);
6200 goto errout;
6201 }
6202 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFINFO, NULL, GFP_ATOMIC);
6203 return;
6204 errout:
6205 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFINFO, err);
6206 }
6207
inet6_prefix_nlmsg_size(void)6208 static inline size_t inet6_prefix_nlmsg_size(void)
6209 {
6210 return NLMSG_ALIGN(sizeof(struct prefixmsg))
6211 + nla_total_size(sizeof(struct in6_addr))
6212 + nla_total_size(sizeof(struct prefix_cacheinfo));
6213 }
6214
inet6_fill_prefix(struct sk_buff * skb,struct inet6_dev * idev,struct prefix_info * pinfo,u32 portid,u32 seq,int event,unsigned int flags)6215 static int inet6_fill_prefix(struct sk_buff *skb, struct inet6_dev *idev,
6216 struct prefix_info *pinfo, u32 portid, u32 seq,
6217 int event, unsigned int flags)
6218 {
6219 struct prefixmsg *pmsg;
6220 struct nlmsghdr *nlh;
6221 struct prefix_cacheinfo ci;
6222
6223 nlh = nlmsg_put(skb, portid, seq, event, sizeof(*pmsg), flags);
6224 if (!nlh)
6225 return -EMSGSIZE;
6226
6227 pmsg = nlmsg_data(nlh);
6228 pmsg->prefix_family = AF_INET6;
6229 pmsg->prefix_pad1 = 0;
6230 pmsg->prefix_pad2 = 0;
6231 pmsg->prefix_ifindex = idev->dev->ifindex;
6232 pmsg->prefix_len = pinfo->prefix_len;
6233 pmsg->prefix_type = pinfo->type;
6234 pmsg->prefix_pad3 = 0;
6235 pmsg->prefix_flags = pinfo->flags;
6236
6237 if (nla_put(skb, PREFIX_ADDRESS, sizeof(pinfo->prefix), &pinfo->prefix))
6238 goto nla_put_failure;
6239 ci.preferred_time = ntohl(pinfo->prefered);
6240 ci.valid_time = ntohl(pinfo->valid);
6241 if (nla_put(skb, PREFIX_CACHEINFO, sizeof(ci), &ci))
6242 goto nla_put_failure;
6243 nlmsg_end(skb, nlh);
6244 return 0;
6245
6246 nla_put_failure:
6247 nlmsg_cancel(skb, nlh);
6248 return -EMSGSIZE;
6249 }
6250
inet6_prefix_notify(int event,struct inet6_dev * idev,struct prefix_info * pinfo)6251 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
6252 struct prefix_info *pinfo)
6253 {
6254 struct sk_buff *skb;
6255 struct net *net = dev_net(idev->dev);
6256 int err = -ENOBUFS;
6257
6258 skb = nlmsg_new(inet6_prefix_nlmsg_size(), GFP_ATOMIC);
6259 if (!skb)
6260 goto errout;
6261
6262 err = inet6_fill_prefix(skb, idev, pinfo, 0, 0, event, 0);
6263 if (err < 0) {
6264 /* -EMSGSIZE implies BUG in inet6_prefix_nlmsg_size() */
6265 WARN_ON(err == -EMSGSIZE);
6266 kfree_skb(skb);
6267 goto errout;
6268 }
6269 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_PREFIX, NULL, GFP_ATOMIC);
6270 return;
6271 errout:
6272 rtnl_set_sk_err(net, RTNLGRP_IPV6_PREFIX, err);
6273 }
6274
__ipv6_ifa_notify(int event,struct inet6_ifaddr * ifp)6275 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
6276 {
6277 struct net *net = dev_net(ifp->idev->dev);
6278
6279 if (event)
6280 ASSERT_RTNL();
6281
6282 inet6_ifa_notify(event ? : RTM_NEWADDR, ifp);
6283
6284 switch (event) {
6285 case RTM_NEWADDR:
6286 /*
6287 * If the address was optimistic we inserted the route at the
6288 * start of our DAD process, so we don't need to do it again.
6289 * If the device was taken down in the middle of the DAD
6290 * cycle there is a race where we could get here without a
6291 * host route, so nothing to insert. That will be fixed when
6292 * the device is brought up.
6293 */
6294 if (ifp->rt && !rcu_access_pointer(ifp->rt->fib6_node)) {
6295 ip6_ins_rt(net, ifp->rt);
6296 } else if (!ifp->rt && (ifp->idev->dev->flags & IFF_UP)) {
6297 pr_warn("BUG: Address %pI6c on device %s is missing its host route.\n",
6298 &ifp->addr, ifp->idev->dev->name);
6299 }
6300
6301 if (ifp->idev->cnf.forwarding)
6302 addrconf_join_anycast(ifp);
6303 if (!ipv6_addr_any(&ifp->peer_addr))
6304 addrconf_prefix_route(&ifp->peer_addr, 128,
6305 ifp->rt_priority, ifp->idev->dev,
6306 0, 0, GFP_ATOMIC);
6307 break;
6308 case RTM_DELADDR:
6309 if (ifp->idev->cnf.forwarding)
6310 addrconf_leave_anycast(ifp);
6311 addrconf_leave_solict(ifp->idev, &ifp->addr);
6312 if (!ipv6_addr_any(&ifp->peer_addr)) {
6313 struct fib6_info *rt;
6314
6315 rt = addrconf_get_prefix_route(&ifp->peer_addr, 128,
6316 ifp->idev->dev, 0, 0,
6317 false);
6318 if (rt)
6319 ip6_del_rt(net, rt, false);
6320 }
6321 if (ifp->rt) {
6322 ip6_del_rt(net, ifp->rt, false);
6323 ifp->rt = NULL;
6324 }
6325 rt_genid_bump_ipv6(net);
6326 break;
6327 }
6328 atomic_inc(&net->ipv6.dev_addr_genid);
6329 }
6330
ipv6_ifa_notify(int event,struct inet6_ifaddr * ifp)6331 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
6332 {
6333 if (likely(ifp->idev->dead == 0))
6334 __ipv6_ifa_notify(event, ifp);
6335 }
6336
6337 #ifdef CONFIG_SYSCTL
6338
addrconf_sysctl_forward(const struct ctl_table * ctl,int write,void * buffer,size_t * lenp,loff_t * ppos)6339 static int addrconf_sysctl_forward(const struct ctl_table *ctl, int write,
6340 void *buffer, size_t *lenp, loff_t *ppos)
6341 {
6342 int *valp = ctl->data;
6343 int val = *valp;
6344 loff_t pos = *ppos;
6345 struct ctl_table lctl;
6346 int ret;
6347
6348 /*
6349 * ctl->data points to idev->cnf.forwarding, we should
6350 * not modify it until we get the rtnl lock.
6351 */
6352 lctl = *ctl;
6353 lctl.data = &val;
6354
6355 ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
6356
6357 if (write)
6358 ret = addrconf_fixup_forwarding(ctl, valp, val);
6359 if (ret)
6360 *ppos = pos;
6361 return ret;
6362 }
6363
addrconf_sysctl_mtu(const struct ctl_table * ctl,int write,void * buffer,size_t * lenp,loff_t * ppos)6364 static int addrconf_sysctl_mtu(const struct ctl_table *ctl, int write,
6365 void *buffer, size_t *lenp, loff_t *ppos)
6366 {
6367 struct inet6_dev *idev = ctl->extra1;
6368 int min_mtu = IPV6_MIN_MTU;
6369 struct ctl_table lctl;
6370
6371 lctl = *ctl;
6372 lctl.extra1 = &min_mtu;
6373 lctl.extra2 = idev ? &idev->dev->mtu : NULL;
6374
6375 return proc_dointvec_minmax(&lctl, write, buffer, lenp, ppos);
6376 }
6377
dev_disable_change(struct inet6_dev * idev)6378 static void dev_disable_change(struct inet6_dev *idev)
6379 {
6380 struct netdev_notifier_info info;
6381
6382 if (!idev || !idev->dev)
6383 return;
6384
6385 netdev_notifier_info_init(&info, idev->dev);
6386 if (idev->cnf.disable_ipv6)
6387 addrconf_notify(NULL, NETDEV_DOWN, &info);
6388 else
6389 addrconf_notify(NULL, NETDEV_UP, &info);
6390 }
6391
addrconf_disable_change(struct net * net,__s32 newf)6392 static void addrconf_disable_change(struct net *net, __s32 newf)
6393 {
6394 struct net_device *dev;
6395 struct inet6_dev *idev;
6396
6397 for_each_netdev(net, dev) {
6398 idev = __in6_dev_get_rtnl_net(dev);
6399 if (idev) {
6400 int changed = (!idev->cnf.disable_ipv6) ^ (!newf);
6401
6402 WRITE_ONCE(idev->cnf.disable_ipv6, newf);
6403 if (changed)
6404 dev_disable_change(idev);
6405 }
6406 }
6407 }
6408
addrconf_disable_ipv6(const struct ctl_table * table,int * p,int newf)6409 static int addrconf_disable_ipv6(const struct ctl_table *table, int *p, int newf)
6410 {
6411 struct net *net = (struct net *)table->extra2;
6412 int old;
6413
6414 if (p == &net->ipv6.devconf_dflt->disable_ipv6) {
6415 WRITE_ONCE(*p, newf);
6416 return 0;
6417 }
6418
6419 if (!rtnl_net_trylock(net))
6420 return restart_syscall();
6421
6422 old = *p;
6423 WRITE_ONCE(*p, newf);
6424
6425 if (p == &net->ipv6.devconf_all->disable_ipv6) {
6426 WRITE_ONCE(net->ipv6.devconf_dflt->disable_ipv6, newf);
6427 addrconf_disable_change(net, newf);
6428 } else if ((!newf) ^ (!old)) {
6429 dev_disable_change((struct inet6_dev *)table->extra1);
6430 }
6431
6432 rtnl_net_unlock(net);
6433 return 0;
6434 }
6435
addrconf_sysctl_disable(const struct ctl_table * ctl,int write,void * buffer,size_t * lenp,loff_t * ppos)6436 static int addrconf_sysctl_disable(const struct ctl_table *ctl, int write,
6437 void *buffer, size_t *lenp, loff_t *ppos)
6438 {
6439 int *valp = ctl->data;
6440 int val = *valp;
6441 loff_t pos = *ppos;
6442 struct ctl_table lctl;
6443 int ret;
6444
6445 /*
6446 * ctl->data points to idev->cnf.disable_ipv6, we should
6447 * not modify it until we get the rtnl lock.
6448 */
6449 lctl = *ctl;
6450 lctl.data = &val;
6451
6452 ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
6453
6454 if (write)
6455 ret = addrconf_disable_ipv6(ctl, valp, val);
6456 if (ret)
6457 *ppos = pos;
6458 return ret;
6459 }
6460
addrconf_sysctl_proxy_ndp(const struct ctl_table * ctl,int write,void * buffer,size_t * lenp,loff_t * ppos)6461 static int addrconf_sysctl_proxy_ndp(const struct ctl_table *ctl, int write,
6462 void *buffer, size_t *lenp, loff_t *ppos)
6463 {
6464 int *valp = ctl->data;
6465 int ret;
6466 int old, new;
6467
6468 old = *valp;
6469 ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
6470 new = *valp;
6471
6472 if (write && old != new) {
6473 struct net *net = ctl->extra2;
6474
6475 if (!rtnl_net_trylock(net))
6476 return restart_syscall();
6477
6478 if (valp == &net->ipv6.devconf_dflt->proxy_ndp) {
6479 inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
6480 NETCONFA_PROXY_NEIGH,
6481 NETCONFA_IFINDEX_DEFAULT,
6482 net->ipv6.devconf_dflt);
6483 } else if (valp == &net->ipv6.devconf_all->proxy_ndp) {
6484 inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
6485 NETCONFA_PROXY_NEIGH,
6486 NETCONFA_IFINDEX_ALL,
6487 net->ipv6.devconf_all);
6488 } else {
6489 struct inet6_dev *idev = ctl->extra1;
6490
6491 inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
6492 NETCONFA_PROXY_NEIGH,
6493 idev->dev->ifindex,
6494 &idev->cnf);
6495 }
6496 rtnl_net_unlock(net);
6497 }
6498
6499 return ret;
6500 }
6501
addrconf_sysctl_addr_gen_mode(const struct ctl_table * ctl,int write,void * buffer,size_t * lenp,loff_t * ppos)6502 static int addrconf_sysctl_addr_gen_mode(const struct ctl_table *ctl, int write,
6503 void *buffer, size_t *lenp,
6504 loff_t *ppos)
6505 {
6506 int ret = 0;
6507 u32 new_val;
6508 struct inet6_dev *idev = (struct inet6_dev *)ctl->extra1;
6509 struct net *net = (struct net *)ctl->extra2;
6510 struct ctl_table tmp = {
6511 .data = &new_val,
6512 .maxlen = sizeof(new_val),
6513 .mode = ctl->mode,
6514 };
6515
6516 if (!rtnl_net_trylock(net))
6517 return restart_syscall();
6518
6519 new_val = *((u32 *)ctl->data);
6520
6521 ret = proc_douintvec(&tmp, write, buffer, lenp, ppos);
6522 if (ret != 0)
6523 goto out;
6524
6525 if (write) {
6526 if (check_addr_gen_mode(new_val) < 0) {
6527 ret = -EINVAL;
6528 goto out;
6529 }
6530
6531 if (idev) {
6532 if (check_stable_privacy(idev, net, new_val) < 0) {
6533 ret = -EINVAL;
6534 goto out;
6535 }
6536
6537 if (idev->cnf.addr_gen_mode != new_val) {
6538 WRITE_ONCE(idev->cnf.addr_gen_mode, new_val);
6539 netdev_lock_ops(idev->dev);
6540 addrconf_init_auto_addrs(idev->dev);
6541 netdev_unlock_ops(idev->dev);
6542 }
6543 } else if (&net->ipv6.devconf_all->addr_gen_mode == ctl->data) {
6544 struct net_device *dev;
6545
6546 WRITE_ONCE(net->ipv6.devconf_dflt->addr_gen_mode, new_val);
6547 for_each_netdev(net, dev) {
6548 idev = __in6_dev_get_rtnl_net(dev);
6549 if (idev &&
6550 idev->cnf.addr_gen_mode != new_val) {
6551 WRITE_ONCE(idev->cnf.addr_gen_mode,
6552 new_val);
6553 netdev_lock_ops(idev->dev);
6554 addrconf_init_auto_addrs(idev->dev);
6555 netdev_unlock_ops(idev->dev);
6556 }
6557 }
6558 }
6559
6560 WRITE_ONCE(*((u32 *)ctl->data), new_val);
6561 }
6562
6563 out:
6564 rtnl_net_unlock(net);
6565
6566 return ret;
6567 }
6568
addrconf_sysctl_stable_secret(const struct ctl_table * ctl,int write,void * buffer,size_t * lenp,loff_t * ppos)6569 static int addrconf_sysctl_stable_secret(const struct ctl_table *ctl, int write,
6570 void *buffer, size_t *lenp,
6571 loff_t *ppos)
6572 {
6573 int err;
6574 struct in6_addr addr;
6575 char str[IPV6_MAX_STRLEN];
6576 struct ctl_table lctl = *ctl;
6577 struct net *net = ctl->extra2;
6578 struct ipv6_stable_secret *secret = ctl->data;
6579
6580 if (&net->ipv6.devconf_all->stable_secret == ctl->data)
6581 return -EIO;
6582
6583 lctl.maxlen = IPV6_MAX_STRLEN;
6584 lctl.data = str;
6585
6586 if (!rtnl_net_trylock(net))
6587 return restart_syscall();
6588
6589 if (!write && !secret->initialized) {
6590 err = -EIO;
6591 goto out;
6592 }
6593
6594 err = snprintf(str, sizeof(str), "%pI6", &secret->secret);
6595 if (err >= sizeof(str)) {
6596 err = -EIO;
6597 goto out;
6598 }
6599
6600 err = proc_dostring(&lctl, write, buffer, lenp, ppos);
6601 if (err || !write)
6602 goto out;
6603
6604 if (in6_pton(str, -1, addr.in6_u.u6_addr8, -1, NULL) != 1) {
6605 err = -EIO;
6606 goto out;
6607 }
6608
6609 secret->initialized = true;
6610 secret->secret = addr;
6611
6612 if (&net->ipv6.devconf_dflt->stable_secret == ctl->data) {
6613 struct net_device *dev;
6614
6615 for_each_netdev(net, dev) {
6616 struct inet6_dev *idev = __in6_dev_get_rtnl_net(dev);
6617
6618 if (idev) {
6619 WRITE_ONCE(idev->cnf.addr_gen_mode,
6620 IN6_ADDR_GEN_MODE_STABLE_PRIVACY);
6621 }
6622 }
6623 } else {
6624 struct inet6_dev *idev = ctl->extra1;
6625
6626 WRITE_ONCE(idev->cnf.addr_gen_mode,
6627 IN6_ADDR_GEN_MODE_STABLE_PRIVACY);
6628 }
6629
6630 out:
6631 rtnl_net_unlock(net);
6632
6633 return err;
6634 }
6635
6636 static
addrconf_sysctl_ignore_routes_with_linkdown(const struct ctl_table * ctl,int write,void * buffer,size_t * lenp,loff_t * ppos)6637 int addrconf_sysctl_ignore_routes_with_linkdown(const struct ctl_table *ctl,
6638 int write, void *buffer,
6639 size_t *lenp,
6640 loff_t *ppos)
6641 {
6642 int *valp = ctl->data;
6643 int val = *valp;
6644 loff_t pos = *ppos;
6645 struct ctl_table lctl;
6646 int ret;
6647
6648 /* ctl->data points to idev->cnf.ignore_routes_when_linkdown
6649 * we should not modify it until we get the rtnl lock.
6650 */
6651 lctl = *ctl;
6652 lctl.data = &val;
6653
6654 ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
6655
6656 if (write)
6657 ret = addrconf_fixup_linkdown(ctl, valp, val);
6658 if (ret)
6659 *ppos = pos;
6660 return ret;
6661 }
6662
6663 static
addrconf_set_nopolicy(struct rt6_info * rt,int action)6664 void addrconf_set_nopolicy(struct rt6_info *rt, int action)
6665 {
6666 if (rt) {
6667 if (action)
6668 rt->dst.flags |= DST_NOPOLICY;
6669 else
6670 rt->dst.flags &= ~DST_NOPOLICY;
6671 }
6672 }
6673
6674 static
addrconf_disable_policy_idev(struct inet6_dev * idev,int val)6675 void addrconf_disable_policy_idev(struct inet6_dev *idev, int val)
6676 {
6677 struct inet6_ifaddr *ifa;
6678
6679 read_lock_bh(&idev->lock);
6680 list_for_each_entry(ifa, &idev->addr_list, if_list) {
6681 spin_lock(&ifa->lock);
6682 if (ifa->rt) {
6683 /* host routes only use builtin fib6_nh */
6684 struct fib6_nh *nh = ifa->rt->fib6_nh;
6685 int cpu;
6686
6687 rcu_read_lock();
6688 ifa->rt->dst_nopolicy = val ? true : false;
6689 if (nh->rt6i_pcpu) {
6690 for_each_possible_cpu(cpu) {
6691 struct rt6_info **rtp;
6692
6693 rtp = per_cpu_ptr(nh->rt6i_pcpu, cpu);
6694 addrconf_set_nopolicy(*rtp, val);
6695 }
6696 }
6697 rcu_read_unlock();
6698 }
6699 spin_unlock(&ifa->lock);
6700 }
6701 read_unlock_bh(&idev->lock);
6702 }
6703
6704 static
addrconf_disable_policy(const struct ctl_table * ctl,int * valp,int val)6705 int addrconf_disable_policy(const struct ctl_table *ctl, int *valp, int val)
6706 {
6707 struct net *net = (struct net *)ctl->extra2;
6708 struct inet6_dev *idev;
6709
6710 if (valp == &net->ipv6.devconf_dflt->disable_policy) {
6711 WRITE_ONCE(*valp, val);
6712 return 0;
6713 }
6714
6715 if (!rtnl_net_trylock(net))
6716 return restart_syscall();
6717
6718 WRITE_ONCE(*valp, val);
6719
6720 if (valp == &net->ipv6.devconf_all->disable_policy) {
6721 struct net_device *dev;
6722
6723 for_each_netdev(net, dev) {
6724 idev = __in6_dev_get_rtnl_net(dev);
6725 if (idev)
6726 addrconf_disable_policy_idev(idev, val);
6727 }
6728 } else {
6729 idev = (struct inet6_dev *)ctl->extra1;
6730 addrconf_disable_policy_idev(idev, val);
6731 }
6732
6733 rtnl_net_unlock(net);
6734 return 0;
6735 }
6736
addrconf_sysctl_disable_policy(const struct ctl_table * ctl,int write,void * buffer,size_t * lenp,loff_t * ppos)6737 static int addrconf_sysctl_disable_policy(const struct ctl_table *ctl, int write,
6738 void *buffer, size_t *lenp, loff_t *ppos)
6739 {
6740 int *valp = ctl->data;
6741 int val = *valp;
6742 loff_t pos = *ppos;
6743 struct ctl_table lctl;
6744 int ret;
6745
6746 lctl = *ctl;
6747 lctl.data = &val;
6748 ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
6749
6750 if (write && (*valp != val))
6751 ret = addrconf_disable_policy(ctl, valp, val);
6752
6753 if (ret)
6754 *ppos = pos;
6755
6756 return ret;
6757 }
6758
addrconf_force_forward_change(struct net * net,__s32 newf)6759 static void addrconf_force_forward_change(struct net *net, __s32 newf)
6760 {
6761 struct net_device *dev;
6762 struct inet6_dev *idev;
6763
6764 for_each_netdev(net, dev) {
6765 idev = __in6_dev_get_rtnl_net(dev);
6766 if (idev) {
6767 int changed = (!idev->cnf.force_forwarding) ^ (!newf);
6768
6769 WRITE_ONCE(idev->cnf.force_forwarding, newf);
6770 if (changed)
6771 inet6_netconf_notify_devconf(dev_net(dev), RTM_NEWNETCONF,
6772 NETCONFA_FORCE_FORWARDING,
6773 dev->ifindex, &idev->cnf);
6774 }
6775 }
6776 }
6777
addrconf_sysctl_force_forwarding(const struct ctl_table * ctl,int write,void * buffer,size_t * lenp,loff_t * ppos)6778 static int addrconf_sysctl_force_forwarding(const struct ctl_table *ctl, int write,
6779 void *buffer, size_t *lenp, loff_t *ppos)
6780 {
6781 struct inet6_dev *idev = ctl->extra1;
6782 struct ctl_table tmp_ctl = *ctl;
6783 struct net *net = ctl->extra2;
6784 int *valp = ctl->data;
6785 int new_val = *valp;
6786 int old_val = *valp;
6787 loff_t pos = *ppos;
6788 int ret;
6789
6790 tmp_ctl.extra1 = SYSCTL_ZERO;
6791 tmp_ctl.extra2 = SYSCTL_ONE;
6792 tmp_ctl.data = &new_val;
6793
6794 ret = proc_douintvec_minmax(&tmp_ctl, write, buffer, lenp, ppos);
6795
6796 if (write && old_val != new_val) {
6797 if (!rtnl_net_trylock(net))
6798 return restart_syscall();
6799
6800 WRITE_ONCE(*valp, new_val);
6801
6802 if (valp == &net->ipv6.devconf_dflt->force_forwarding) {
6803 inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
6804 NETCONFA_FORCE_FORWARDING,
6805 NETCONFA_IFINDEX_DEFAULT,
6806 net->ipv6.devconf_dflt);
6807 } else if (valp == &net->ipv6.devconf_all->force_forwarding) {
6808 inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
6809 NETCONFA_FORCE_FORWARDING,
6810 NETCONFA_IFINDEX_ALL,
6811 net->ipv6.devconf_all);
6812
6813 addrconf_force_forward_change(net, new_val);
6814 } else {
6815 inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
6816 NETCONFA_FORCE_FORWARDING,
6817 idev->dev->ifindex,
6818 &idev->cnf);
6819 }
6820 rtnl_net_unlock(net);
6821 }
6822
6823 if (ret)
6824 *ppos = pos;
6825 return ret;
6826 }
6827
6828 static int minus_one = -1;
6829 static const int two_five_five = 255;
6830 static u32 ioam6_if_id_max = U16_MAX;
6831
6832 static const struct ctl_table addrconf_sysctl[] = {
6833 {
6834 .procname = "forwarding",
6835 .data = &ipv6_devconf.forwarding,
6836 .maxlen = sizeof(int),
6837 .mode = 0644,
6838 .proc_handler = addrconf_sysctl_forward,
6839 },
6840 {
6841 .procname = "hop_limit",
6842 .data = &ipv6_devconf.hop_limit,
6843 .maxlen = sizeof(int),
6844 .mode = 0644,
6845 .proc_handler = proc_dointvec_minmax,
6846 .extra1 = (void *)SYSCTL_ONE,
6847 .extra2 = (void *)&two_five_five,
6848 },
6849 {
6850 .procname = "mtu",
6851 .data = &ipv6_devconf.mtu6,
6852 .maxlen = sizeof(int),
6853 .mode = 0644,
6854 .proc_handler = addrconf_sysctl_mtu,
6855 },
6856 {
6857 .procname = "accept_ra",
6858 .data = &ipv6_devconf.accept_ra,
6859 .maxlen = sizeof(int),
6860 .mode = 0644,
6861 .proc_handler = proc_dointvec,
6862 },
6863 {
6864 .procname = "accept_redirects",
6865 .data = &ipv6_devconf.accept_redirects,
6866 .maxlen = sizeof(int),
6867 .mode = 0644,
6868 .proc_handler = proc_dointvec,
6869 },
6870 {
6871 .procname = "autoconf",
6872 .data = &ipv6_devconf.autoconf,
6873 .maxlen = sizeof(int),
6874 .mode = 0644,
6875 .proc_handler = proc_dointvec,
6876 },
6877 {
6878 .procname = "dad_transmits",
6879 .data = &ipv6_devconf.dad_transmits,
6880 .maxlen = sizeof(int),
6881 .mode = 0644,
6882 .proc_handler = proc_dointvec,
6883 },
6884 {
6885 .procname = "router_solicitations",
6886 .data = &ipv6_devconf.rtr_solicits,
6887 .maxlen = sizeof(int),
6888 .mode = 0644,
6889 .proc_handler = proc_dointvec_minmax,
6890 .extra1 = &minus_one,
6891 },
6892 {
6893 .procname = "router_solicitation_interval",
6894 .data = &ipv6_devconf.rtr_solicit_interval,
6895 .maxlen = sizeof(int),
6896 .mode = 0644,
6897 .proc_handler = proc_dointvec_jiffies,
6898 },
6899 {
6900 .procname = "router_solicitation_max_interval",
6901 .data = &ipv6_devconf.rtr_solicit_max_interval,
6902 .maxlen = sizeof(int),
6903 .mode = 0644,
6904 .proc_handler = proc_dointvec_jiffies,
6905 },
6906 {
6907 .procname = "router_solicitation_delay",
6908 .data = &ipv6_devconf.rtr_solicit_delay,
6909 .maxlen = sizeof(int),
6910 .mode = 0644,
6911 .proc_handler = proc_dointvec_jiffies,
6912 },
6913 {
6914 .procname = "force_mld_version",
6915 .data = &ipv6_devconf.force_mld_version,
6916 .maxlen = sizeof(int),
6917 .mode = 0644,
6918 .proc_handler = proc_dointvec,
6919 },
6920 {
6921 .procname = "mldv1_unsolicited_report_interval",
6922 .data =
6923 &ipv6_devconf.mldv1_unsolicited_report_interval,
6924 .maxlen = sizeof(int),
6925 .mode = 0644,
6926 .proc_handler = proc_dointvec_ms_jiffies,
6927 },
6928 {
6929 .procname = "mldv2_unsolicited_report_interval",
6930 .data =
6931 &ipv6_devconf.mldv2_unsolicited_report_interval,
6932 .maxlen = sizeof(int),
6933 .mode = 0644,
6934 .proc_handler = proc_dointvec_ms_jiffies,
6935 },
6936 {
6937 .procname = "use_tempaddr",
6938 .data = &ipv6_devconf.use_tempaddr,
6939 .maxlen = sizeof(int),
6940 .mode = 0644,
6941 .proc_handler = proc_dointvec,
6942 },
6943 {
6944 .procname = "temp_valid_lft",
6945 .data = &ipv6_devconf.temp_valid_lft,
6946 .maxlen = sizeof(int),
6947 .mode = 0644,
6948 .proc_handler = proc_dointvec,
6949 },
6950 {
6951 .procname = "temp_prefered_lft",
6952 .data = &ipv6_devconf.temp_prefered_lft,
6953 .maxlen = sizeof(int),
6954 .mode = 0644,
6955 .proc_handler = proc_dointvec,
6956 },
6957 {
6958 .procname = "regen_min_advance",
6959 .data = &ipv6_devconf.regen_min_advance,
6960 .maxlen = sizeof(int),
6961 .mode = 0644,
6962 .proc_handler = proc_dointvec,
6963 },
6964 {
6965 .procname = "regen_max_retry",
6966 .data = &ipv6_devconf.regen_max_retry,
6967 .maxlen = sizeof(int),
6968 .mode = 0644,
6969 .proc_handler = proc_dointvec,
6970 },
6971 {
6972 .procname = "max_desync_factor",
6973 .data = &ipv6_devconf.max_desync_factor,
6974 .maxlen = sizeof(int),
6975 .mode = 0644,
6976 .proc_handler = proc_dointvec,
6977 },
6978 {
6979 .procname = "max_addresses",
6980 .data = &ipv6_devconf.max_addresses,
6981 .maxlen = sizeof(int),
6982 .mode = 0644,
6983 .proc_handler = proc_dointvec,
6984 },
6985 {
6986 .procname = "accept_ra_defrtr",
6987 .data = &ipv6_devconf.accept_ra_defrtr,
6988 .maxlen = sizeof(int),
6989 .mode = 0644,
6990 .proc_handler = proc_dointvec,
6991 },
6992 {
6993 .procname = "ra_defrtr_metric",
6994 .data = &ipv6_devconf.ra_defrtr_metric,
6995 .maxlen = sizeof(u32),
6996 .mode = 0644,
6997 .proc_handler = proc_douintvec_minmax,
6998 .extra1 = (void *)SYSCTL_ONE,
6999 },
7000 {
7001 .procname = "accept_ra_min_hop_limit",
7002 .data = &ipv6_devconf.accept_ra_min_hop_limit,
7003 .maxlen = sizeof(int),
7004 .mode = 0644,
7005 .proc_handler = proc_dointvec,
7006 },
7007 {
7008 .procname = "accept_ra_min_lft",
7009 .data = &ipv6_devconf.accept_ra_min_lft,
7010 .maxlen = sizeof(int),
7011 .mode = 0644,
7012 .proc_handler = proc_dointvec,
7013 },
7014 {
7015 .procname = "accept_ra_pinfo",
7016 .data = &ipv6_devconf.accept_ra_pinfo,
7017 .maxlen = sizeof(int),
7018 .mode = 0644,
7019 .proc_handler = proc_dointvec,
7020 },
7021 {
7022 .procname = "ra_honor_pio_life",
7023 .data = &ipv6_devconf.ra_honor_pio_life,
7024 .maxlen = sizeof(u8),
7025 .mode = 0644,
7026 .proc_handler = proc_dou8vec_minmax,
7027 .extra1 = SYSCTL_ZERO,
7028 .extra2 = SYSCTL_ONE,
7029 },
7030 {
7031 .procname = "ra_honor_pio_pflag",
7032 .data = &ipv6_devconf.ra_honor_pio_pflag,
7033 .maxlen = sizeof(u8),
7034 .mode = 0644,
7035 .proc_handler = proc_dou8vec_minmax,
7036 .extra1 = SYSCTL_ZERO,
7037 .extra2 = SYSCTL_ONE,
7038 },
7039 #ifdef CONFIG_IPV6_ROUTER_PREF
7040 {
7041 .procname = "accept_ra_rtr_pref",
7042 .data = &ipv6_devconf.accept_ra_rtr_pref,
7043 .maxlen = sizeof(int),
7044 .mode = 0644,
7045 .proc_handler = proc_dointvec,
7046 },
7047 {
7048 .procname = "router_probe_interval",
7049 .data = &ipv6_devconf.rtr_probe_interval,
7050 .maxlen = sizeof(int),
7051 .mode = 0644,
7052 .proc_handler = proc_dointvec_jiffies,
7053 },
7054 #ifdef CONFIG_IPV6_ROUTE_INFO
7055 {
7056 .procname = "accept_ra_rt_info_min_plen",
7057 .data = &ipv6_devconf.accept_ra_rt_info_min_plen,
7058 .maxlen = sizeof(int),
7059 .mode = 0644,
7060 .proc_handler = proc_dointvec,
7061 },
7062 {
7063 .procname = "accept_ra_rt_info_max_plen",
7064 .data = &ipv6_devconf.accept_ra_rt_info_max_plen,
7065 .maxlen = sizeof(int),
7066 .mode = 0644,
7067 .proc_handler = proc_dointvec,
7068 },
7069 #endif
7070 #endif
7071 {
7072 .procname = "proxy_ndp",
7073 .data = &ipv6_devconf.proxy_ndp,
7074 .maxlen = sizeof(int),
7075 .mode = 0644,
7076 .proc_handler = addrconf_sysctl_proxy_ndp,
7077 },
7078 {
7079 .procname = "accept_source_route",
7080 .data = &ipv6_devconf.accept_source_route,
7081 .maxlen = sizeof(int),
7082 .mode = 0644,
7083 .proc_handler = proc_dointvec,
7084 },
7085 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
7086 {
7087 .procname = "optimistic_dad",
7088 .data = &ipv6_devconf.optimistic_dad,
7089 .maxlen = sizeof(int),
7090 .mode = 0644,
7091 .proc_handler = proc_dointvec,
7092 },
7093 {
7094 .procname = "use_optimistic",
7095 .data = &ipv6_devconf.use_optimistic,
7096 .maxlen = sizeof(int),
7097 .mode = 0644,
7098 .proc_handler = proc_dointvec,
7099 },
7100 #endif
7101 #ifdef CONFIG_IPV6_MROUTE
7102 {
7103 .procname = "mc_forwarding",
7104 .data = &ipv6_devconf.mc_forwarding,
7105 .maxlen = sizeof(int),
7106 .mode = 0444,
7107 .proc_handler = proc_dointvec,
7108 },
7109 #endif
7110 {
7111 .procname = "disable_ipv6",
7112 .data = &ipv6_devconf.disable_ipv6,
7113 .maxlen = sizeof(int),
7114 .mode = 0644,
7115 .proc_handler = addrconf_sysctl_disable,
7116 },
7117 {
7118 .procname = "accept_dad",
7119 .data = &ipv6_devconf.accept_dad,
7120 .maxlen = sizeof(int),
7121 .mode = 0644,
7122 .proc_handler = proc_dointvec,
7123 },
7124 {
7125 .procname = "force_tllao",
7126 .data = &ipv6_devconf.force_tllao,
7127 .maxlen = sizeof(int),
7128 .mode = 0644,
7129 .proc_handler = proc_dointvec
7130 },
7131 {
7132 .procname = "ndisc_notify",
7133 .data = &ipv6_devconf.ndisc_notify,
7134 .maxlen = sizeof(int),
7135 .mode = 0644,
7136 .proc_handler = proc_dointvec
7137 },
7138 {
7139 .procname = "suppress_frag_ndisc",
7140 .data = &ipv6_devconf.suppress_frag_ndisc,
7141 .maxlen = sizeof(int),
7142 .mode = 0644,
7143 .proc_handler = proc_dointvec
7144 },
7145 {
7146 .procname = "accept_ra_from_local",
7147 .data = &ipv6_devconf.accept_ra_from_local,
7148 .maxlen = sizeof(int),
7149 .mode = 0644,
7150 .proc_handler = proc_dointvec,
7151 },
7152 {
7153 .procname = "accept_ra_mtu",
7154 .data = &ipv6_devconf.accept_ra_mtu,
7155 .maxlen = sizeof(int),
7156 .mode = 0644,
7157 .proc_handler = proc_dointvec,
7158 },
7159 {
7160 .procname = "stable_secret",
7161 .data = &ipv6_devconf.stable_secret,
7162 .maxlen = IPV6_MAX_STRLEN,
7163 .mode = 0600,
7164 .proc_handler = addrconf_sysctl_stable_secret,
7165 },
7166 {
7167 .procname = "use_oif_addrs_only",
7168 .data = &ipv6_devconf.use_oif_addrs_only,
7169 .maxlen = sizeof(int),
7170 .mode = 0644,
7171 .proc_handler = proc_dointvec,
7172 },
7173 {
7174 .procname = "ignore_routes_with_linkdown",
7175 .data = &ipv6_devconf.ignore_routes_with_linkdown,
7176 .maxlen = sizeof(int),
7177 .mode = 0644,
7178 .proc_handler = addrconf_sysctl_ignore_routes_with_linkdown,
7179 },
7180 {
7181 .procname = "drop_unicast_in_l2_multicast",
7182 .data = &ipv6_devconf.drop_unicast_in_l2_multicast,
7183 .maxlen = sizeof(int),
7184 .mode = 0644,
7185 .proc_handler = proc_dointvec,
7186 },
7187 {
7188 .procname = "drop_unsolicited_na",
7189 .data = &ipv6_devconf.drop_unsolicited_na,
7190 .maxlen = sizeof(int),
7191 .mode = 0644,
7192 .proc_handler = proc_dointvec,
7193 },
7194 {
7195 .procname = "keep_addr_on_down",
7196 .data = &ipv6_devconf.keep_addr_on_down,
7197 .maxlen = sizeof(int),
7198 .mode = 0644,
7199 .proc_handler = proc_dointvec,
7200
7201 },
7202 {
7203 .procname = "seg6_enabled",
7204 .data = &ipv6_devconf.seg6_enabled,
7205 .maxlen = sizeof(int),
7206 .mode = 0644,
7207 .proc_handler = proc_dointvec,
7208 },
7209 #ifdef CONFIG_IPV6_SEG6_HMAC
7210 {
7211 .procname = "seg6_require_hmac",
7212 .data = &ipv6_devconf.seg6_require_hmac,
7213 .maxlen = sizeof(int),
7214 .mode = 0644,
7215 .proc_handler = proc_dointvec,
7216 },
7217 #endif
7218 {
7219 .procname = "enhanced_dad",
7220 .data = &ipv6_devconf.enhanced_dad,
7221 .maxlen = sizeof(int),
7222 .mode = 0644,
7223 .proc_handler = proc_dointvec,
7224 },
7225 {
7226 .procname = "addr_gen_mode",
7227 .data = &ipv6_devconf.addr_gen_mode,
7228 .maxlen = sizeof(int),
7229 .mode = 0644,
7230 .proc_handler = addrconf_sysctl_addr_gen_mode,
7231 },
7232 {
7233 .procname = "disable_policy",
7234 .data = &ipv6_devconf.disable_policy,
7235 .maxlen = sizeof(int),
7236 .mode = 0644,
7237 .proc_handler = addrconf_sysctl_disable_policy,
7238 },
7239 {
7240 .procname = "ndisc_tclass",
7241 .data = &ipv6_devconf.ndisc_tclass,
7242 .maxlen = sizeof(int),
7243 .mode = 0644,
7244 .proc_handler = proc_dointvec_minmax,
7245 .extra1 = (void *)SYSCTL_ZERO,
7246 .extra2 = (void *)&two_five_five,
7247 },
7248 {
7249 .procname = "rpl_seg_enabled",
7250 .data = &ipv6_devconf.rpl_seg_enabled,
7251 .maxlen = sizeof(int),
7252 .mode = 0644,
7253 .proc_handler = proc_dointvec_minmax,
7254 .extra1 = SYSCTL_ZERO,
7255 .extra2 = SYSCTL_ONE,
7256 },
7257 {
7258 .procname = "ioam6_enabled",
7259 .data = &ipv6_devconf.ioam6_enabled,
7260 .maxlen = sizeof(u8),
7261 .mode = 0644,
7262 .proc_handler = proc_dou8vec_minmax,
7263 .extra1 = (void *)SYSCTL_ZERO,
7264 .extra2 = (void *)SYSCTL_ONE,
7265 },
7266 {
7267 .procname = "ioam6_id",
7268 .data = &ipv6_devconf.ioam6_id,
7269 .maxlen = sizeof(u32),
7270 .mode = 0644,
7271 .proc_handler = proc_douintvec_minmax,
7272 .extra1 = (void *)SYSCTL_ZERO,
7273 .extra2 = (void *)&ioam6_if_id_max,
7274 },
7275 {
7276 .procname = "ioam6_id_wide",
7277 .data = &ipv6_devconf.ioam6_id_wide,
7278 .maxlen = sizeof(u32),
7279 .mode = 0644,
7280 .proc_handler = proc_douintvec,
7281 },
7282 {
7283 .procname = "ndisc_evict_nocarrier",
7284 .data = &ipv6_devconf.ndisc_evict_nocarrier,
7285 .maxlen = sizeof(u8),
7286 .mode = 0644,
7287 .proc_handler = proc_dou8vec_minmax,
7288 .extra1 = (void *)SYSCTL_ZERO,
7289 .extra2 = (void *)SYSCTL_ONE,
7290 },
7291 {
7292 .procname = "accept_untracked_na",
7293 .data = &ipv6_devconf.accept_untracked_na,
7294 .maxlen = sizeof(int),
7295 .mode = 0644,
7296 .proc_handler = proc_dointvec_minmax,
7297 .extra1 = SYSCTL_ZERO,
7298 .extra2 = SYSCTL_TWO,
7299 },
7300 {
7301 .procname = "force_forwarding",
7302 .data = &ipv6_devconf.force_forwarding,
7303 .maxlen = sizeof(int),
7304 .mode = 0644,
7305 .proc_handler = addrconf_sysctl_force_forwarding,
7306 },
7307 };
7308
__addrconf_sysctl_register(struct net * net,char * dev_name,struct inet6_dev * idev,struct ipv6_devconf * p)7309 static int __addrconf_sysctl_register(struct net *net, char *dev_name,
7310 struct inet6_dev *idev, struct ipv6_devconf *p)
7311 {
7312 size_t table_size = ARRAY_SIZE(addrconf_sysctl);
7313 int i, ifindex;
7314 struct ctl_table *table;
7315 char path[sizeof("net/ipv6/conf/") + IFNAMSIZ];
7316
7317 table = kmemdup(addrconf_sysctl, sizeof(addrconf_sysctl), GFP_KERNEL_ACCOUNT);
7318 if (!table)
7319 goto out;
7320
7321 for (i = 0; i < table_size; i++) {
7322 table[i].data += (char *)p - (char *)&ipv6_devconf;
7323 /* If one of these is already set, then it is not safe to
7324 * overwrite either of them: this makes proc_dointvec_minmax
7325 * usable.
7326 */
7327 if (!table[i].extra1 && !table[i].extra2) {
7328 table[i].extra1 = idev; /* embedded; no ref */
7329 table[i].extra2 = net;
7330 }
7331 }
7332
7333 snprintf(path, sizeof(path), "net/ipv6/conf/%s", dev_name);
7334
7335 p->sysctl_header = register_net_sysctl_sz(net, path, table,
7336 table_size);
7337 if (!p->sysctl_header)
7338 goto free;
7339
7340 if (!strcmp(dev_name, "all"))
7341 ifindex = NETCONFA_IFINDEX_ALL;
7342 else if (!strcmp(dev_name, "default"))
7343 ifindex = NETCONFA_IFINDEX_DEFAULT;
7344 else
7345 ifindex = idev->dev->ifindex;
7346 inet6_netconf_notify_devconf(net, RTM_NEWNETCONF, NETCONFA_ALL,
7347 ifindex, p);
7348 return 0;
7349
7350 free:
7351 kfree(table);
7352 out:
7353 return -ENOBUFS;
7354 }
7355
__addrconf_sysctl_unregister(struct net * net,struct ipv6_devconf * p,int ifindex)7356 static void __addrconf_sysctl_unregister(struct net *net,
7357 struct ipv6_devconf *p, int ifindex)
7358 {
7359 const struct ctl_table *table;
7360
7361 if (!p->sysctl_header)
7362 return;
7363
7364 table = p->sysctl_header->ctl_table_arg;
7365 unregister_net_sysctl_table(p->sysctl_header);
7366 p->sysctl_header = NULL;
7367 kfree(table);
7368
7369 inet6_netconf_notify_devconf(net, RTM_DELNETCONF, 0, ifindex, NULL);
7370 }
7371
addrconf_sysctl_register(struct inet6_dev * idev)7372 static int addrconf_sysctl_register(struct inet6_dev *idev)
7373 {
7374 int err;
7375
7376 if (!sysctl_dev_name_is_allowed(idev->dev->name))
7377 return -EINVAL;
7378
7379 err = neigh_sysctl_register(idev->dev, idev->nd_parms,
7380 &ndisc_ifinfo_sysctl_change);
7381 if (err)
7382 return err;
7383 err = __addrconf_sysctl_register(dev_net(idev->dev), idev->dev->name,
7384 idev, &idev->cnf);
7385 if (err)
7386 neigh_sysctl_unregister(idev->nd_parms);
7387
7388 return err;
7389 }
7390
addrconf_sysctl_unregister(struct inet6_dev * idev)7391 static void addrconf_sysctl_unregister(struct inet6_dev *idev)
7392 {
7393 __addrconf_sysctl_unregister(dev_net(idev->dev), &idev->cnf,
7394 idev->dev->ifindex);
7395 neigh_sysctl_unregister(idev->nd_parms);
7396 }
7397
7398
7399 #endif
7400
addrconf_init_net(struct net * net)7401 static int __net_init addrconf_init_net(struct net *net)
7402 {
7403 int err = -ENOMEM;
7404 struct ipv6_devconf *all, *dflt;
7405
7406 spin_lock_init(&net->ipv6.addrconf_hash_lock);
7407 INIT_DEFERRABLE_WORK(&net->ipv6.addr_chk_work, addrconf_verify_work);
7408 net->ipv6.inet6_addr_lst = kzalloc_objs(struct hlist_head,
7409 IN6_ADDR_HSIZE);
7410 if (!net->ipv6.inet6_addr_lst)
7411 goto err_alloc_addr;
7412
7413 all = kmemdup(&ipv6_devconf, sizeof(ipv6_devconf), GFP_KERNEL);
7414 if (!all)
7415 goto err_alloc_all;
7416
7417 dflt = kmemdup(&ipv6_devconf_dflt, sizeof(ipv6_devconf_dflt), GFP_KERNEL);
7418 if (!dflt)
7419 goto err_alloc_dflt;
7420
7421 if (!net_eq(net, &init_net)) {
7422 switch (net_inherit_devconf()) {
7423 case 1: /* copy from init_net */
7424 memcpy(all, init_net.ipv6.devconf_all,
7425 sizeof(ipv6_devconf));
7426 memcpy(dflt, init_net.ipv6.devconf_dflt,
7427 sizeof(ipv6_devconf_dflt));
7428 break;
7429 case 3: /* copy from the current netns */
7430 memcpy(all, current->nsproxy->net_ns->ipv6.devconf_all,
7431 sizeof(ipv6_devconf));
7432 memcpy(dflt,
7433 current->nsproxy->net_ns->ipv6.devconf_dflt,
7434 sizeof(ipv6_devconf_dflt));
7435 break;
7436 case 0:
7437 case 2:
7438 /* use compiled values */
7439 break;
7440 }
7441 }
7442
7443 /* these will be inherited by all namespaces */
7444 dflt->autoconf = ipv6_defaults.autoconf;
7445 dflt->disable_ipv6 = ipv6_defaults.disable_ipv6;
7446
7447 dflt->stable_secret.initialized = false;
7448 all->stable_secret.initialized = false;
7449
7450 net->ipv6.devconf_all = all;
7451 net->ipv6.devconf_dflt = dflt;
7452
7453 #ifdef CONFIG_SYSCTL
7454 err = __addrconf_sysctl_register(net, "all", NULL, all);
7455 if (err < 0)
7456 goto err_reg_all;
7457
7458 err = __addrconf_sysctl_register(net, "default", NULL, dflt);
7459 if (err < 0)
7460 goto err_reg_dflt;
7461 #endif
7462 return 0;
7463
7464 #ifdef CONFIG_SYSCTL
7465 err_reg_dflt:
7466 __addrconf_sysctl_unregister(net, all, NETCONFA_IFINDEX_ALL);
7467 err_reg_all:
7468 kfree(dflt);
7469 net->ipv6.devconf_dflt = NULL;
7470 #endif
7471 err_alloc_dflt:
7472 kfree(all);
7473 net->ipv6.devconf_all = NULL;
7474 err_alloc_all:
7475 kfree(net->ipv6.inet6_addr_lst);
7476 err_alloc_addr:
7477 return err;
7478 }
7479
addrconf_exit_net(struct net * net)7480 static void __net_exit addrconf_exit_net(struct net *net)
7481 {
7482 int i;
7483
7484 #ifdef CONFIG_SYSCTL
7485 __addrconf_sysctl_unregister(net, net->ipv6.devconf_dflt,
7486 NETCONFA_IFINDEX_DEFAULT);
7487 __addrconf_sysctl_unregister(net, net->ipv6.devconf_all,
7488 NETCONFA_IFINDEX_ALL);
7489 #endif
7490 kfree(net->ipv6.devconf_dflt);
7491 net->ipv6.devconf_dflt = NULL;
7492 kfree(net->ipv6.devconf_all);
7493 net->ipv6.devconf_all = NULL;
7494
7495 cancel_delayed_work_sync(&net->ipv6.addr_chk_work);
7496 /*
7497 * Check hash table, then free it.
7498 */
7499 for (i = 0; i < IN6_ADDR_HSIZE; i++)
7500 WARN_ON_ONCE(!hlist_empty(&net->ipv6.inet6_addr_lst[i]));
7501
7502 kfree(net->ipv6.inet6_addr_lst);
7503 net->ipv6.inet6_addr_lst = NULL;
7504 }
7505
7506 static struct pernet_operations addrconf_ops = {
7507 .init = addrconf_init_net,
7508 .exit = addrconf_exit_net,
7509 };
7510
7511 static struct rtnl_af_ops inet6_ops __read_mostly = {
7512 .family = AF_INET6,
7513 .fill_link_af = inet6_fill_link_af,
7514 .get_link_af_size = inet6_get_link_af_size,
7515 .validate_link_af = inet6_validate_link_af,
7516 .set_link_af = inet6_set_link_af,
7517 };
7518
7519 static const struct rtnl_msg_handler addrconf_rtnl_msg_handlers[] __initconst_or_module = {
7520 {.owner = THIS_MODULE, .protocol = PF_INET6, .msgtype = RTM_GETLINK,
7521 .dumpit = inet6_dump_ifinfo, .flags = RTNL_FLAG_DUMP_UNLOCKED},
7522 {.owner = THIS_MODULE, .protocol = PF_INET6, .msgtype = RTM_NEWADDR,
7523 .doit = inet6_rtm_newaddr, .flags = RTNL_FLAG_DOIT_PERNET},
7524 {.owner = THIS_MODULE, .protocol = PF_INET6, .msgtype = RTM_DELADDR,
7525 .doit = inet6_rtm_deladdr, .flags = RTNL_FLAG_DOIT_PERNET},
7526 {.owner = THIS_MODULE, .protocol = PF_INET6, .msgtype = RTM_GETADDR,
7527 .doit = inet6_rtm_getaddr, .dumpit = inet6_dump_ifaddr,
7528 .flags = RTNL_FLAG_DOIT_UNLOCKED | RTNL_FLAG_DUMP_UNLOCKED},
7529 {.owner = THIS_MODULE, .protocol = PF_INET6, .msgtype = RTM_GETMULTICAST,
7530 .dumpit = inet6_dump_ifmcaddr,
7531 .flags = RTNL_FLAG_DUMP_UNLOCKED},
7532 {.owner = THIS_MODULE, .protocol = PF_INET6, .msgtype = RTM_GETANYCAST,
7533 .dumpit = inet6_dump_ifacaddr,
7534 .flags = RTNL_FLAG_DUMP_UNLOCKED},
7535 {.owner = THIS_MODULE, .protocol = PF_INET6, .msgtype = RTM_GETNETCONF,
7536 .doit = inet6_netconf_get_devconf, .dumpit = inet6_netconf_dump_devconf,
7537 .flags = RTNL_FLAG_DOIT_UNLOCKED | RTNL_FLAG_DUMP_UNLOCKED},
7538 };
7539
7540 /*
7541 * Init / cleanup code
7542 */
7543
addrconf_init(void)7544 int __init addrconf_init(void)
7545 {
7546 struct inet6_dev *idev;
7547 int err;
7548
7549 err = ipv6_addr_label_init();
7550 if (err < 0) {
7551 pr_crit("%s: cannot initialize default policy table: %d\n",
7552 __func__, err);
7553 goto out;
7554 }
7555
7556 err = register_pernet_subsys(&addrconf_ops);
7557 if (err < 0)
7558 goto out_addrlabel;
7559
7560 /* All works using addrconf_wq need to lock rtnl. */
7561 addrconf_wq = create_singlethread_workqueue("ipv6_addrconf");
7562 if (!addrconf_wq) {
7563 err = -ENOMEM;
7564 goto out_nowq;
7565 }
7566
7567 rtnl_net_lock(&init_net);
7568 idev = ipv6_add_dev(blackhole_netdev);
7569 rtnl_net_unlock(&init_net);
7570 if (IS_ERR(idev)) {
7571 err = PTR_ERR(idev);
7572 goto errlo;
7573 }
7574
7575 ip6_route_init_special_entries();
7576
7577 register_netdevice_notifier(&ipv6_dev_notf);
7578
7579 addrconf_verify(&init_net);
7580
7581 err = rtnl_af_register(&inet6_ops);
7582 if (err)
7583 goto erraf;
7584
7585 err = rtnl_register_many(addrconf_rtnl_msg_handlers);
7586 if (err)
7587 goto errout;
7588
7589 err = ipv6_addr_label_rtnl_register();
7590 if (err < 0)
7591 goto errout;
7592
7593 return 0;
7594 errout:
7595 rtnl_unregister_all(PF_INET6);
7596 rtnl_af_unregister(&inet6_ops);
7597 erraf:
7598 unregister_netdevice_notifier(&ipv6_dev_notf);
7599 errlo:
7600 destroy_workqueue(addrconf_wq);
7601 out_nowq:
7602 unregister_pernet_subsys(&addrconf_ops);
7603 out_addrlabel:
7604 ipv6_addr_label_cleanup();
7605 out:
7606 return err;
7607 }
7608
addrconf_cleanup(void)7609 void addrconf_cleanup(void)
7610 {
7611 struct net_device *dev;
7612
7613 unregister_netdevice_notifier(&ipv6_dev_notf);
7614 unregister_pernet_subsys(&addrconf_ops);
7615 ipv6_addr_label_cleanup();
7616
7617 rtnl_af_unregister(&inet6_ops);
7618
7619 rtnl_net_lock(&init_net);
7620
7621 /* clean dev list */
7622 for_each_netdev(&init_net, dev) {
7623 if (!__in6_dev_get_rtnl_net(dev))
7624 continue;
7625 addrconf_ifdown(dev, true);
7626 }
7627 addrconf_ifdown(init_net.loopback_dev, true);
7628
7629 rtnl_net_unlock(&init_net);
7630
7631 destroy_workqueue(addrconf_wq);
7632 }
7633