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_assert_locked_ops_compat(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 0;
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 } else if (p == &net->ipv6.devconf_all->ignore_routes_with_linkdown) {
959 WRITE_ONCE(net->ipv6.devconf_dflt->ignore_routes_with_linkdown, newf);
960 addrconf_linkdown_change(net, newf);
961 if ((!newf) ^ (!old))
962 inet6_netconf_notify_devconf(net,
963 RTM_NEWNETCONF,
964 NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
965 NETCONFA_IFINDEX_ALL,
966 net->ipv6.devconf_all);
967 } else {
968 if (!newf ^ !old) {
969 struct inet6_dev *idev = table->extra1;
970
971 inet6_netconf_notify_devconf(net,
972 RTM_NEWNETCONF,
973 NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
974 idev->dev->ifindex,
975 &idev->cnf);
976 }
977 }
978
979 rtnl_net_unlock(net);
980
981 return 0;
982 }
983
984 #endif
985
986 /* Nobody refers to this ifaddr, destroy it */
inet6_ifa_finish_destroy(struct inet6_ifaddr * ifp)987 void inet6_ifa_finish_destroy(struct inet6_ifaddr *ifp)
988 {
989 WARN_ON(!hlist_unhashed(&ifp->addr_lst));
990
991 #ifdef NET_REFCNT_DEBUG
992 pr_debug("%s\n", __func__);
993 #endif
994
995 in6_dev_put(ifp->idev);
996
997 if (cancel_delayed_work(&ifp->dad_work))
998 pr_notice("delayed DAD work was pending while freeing ifa=%p\n",
999 ifp);
1000
1001 if (ifp->state != INET6_IFADDR_STATE_DEAD) {
1002 pr_warn("Freeing alive inet6 address %p\n", ifp);
1003 return;
1004 }
1005
1006 kfree_rcu(ifp, rcu);
1007 }
1008
1009 static void
ipv6_link_dev_addr(struct inet6_dev * idev,struct inet6_ifaddr * ifp)1010 ipv6_link_dev_addr(struct inet6_dev *idev, struct inet6_ifaddr *ifp)
1011 {
1012 struct list_head *p;
1013 int ifp_scope = ipv6_addr_src_scope(&ifp->addr);
1014
1015 /*
1016 * Each device address list is sorted in order of scope -
1017 * global before linklocal.
1018 */
1019 list_for_each(p, &idev->addr_list) {
1020 struct inet6_ifaddr *ifa
1021 = list_entry(p, struct inet6_ifaddr, if_list);
1022 if (ifp_scope >= ipv6_addr_src_scope(&ifa->addr))
1023 break;
1024 }
1025
1026 list_add_tail_rcu(&ifp->if_list, p);
1027 }
1028
inet6_addr_hash(const struct net * net,const struct in6_addr * addr)1029 static u32 inet6_addr_hash(const struct net *net, const struct in6_addr *addr)
1030 {
1031 u32 val = __ipv6_addr_jhash(addr, net_hash_mix(net));
1032
1033 return hash_32(val, IN6_ADDR_HSIZE_SHIFT);
1034 }
1035
ipv6_chk_same_addr(struct net * net,const struct in6_addr * addr,struct net_device * dev,unsigned int hash)1036 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
1037 struct net_device *dev, unsigned int hash)
1038 {
1039 struct inet6_ifaddr *ifp;
1040
1041 hlist_for_each_entry(ifp, &net->ipv6.inet6_addr_lst[hash], addr_lst) {
1042 if (ipv6_addr_equal(&ifp->addr, addr)) {
1043 if (!dev || ifp->idev->dev == dev)
1044 return true;
1045 }
1046 }
1047 return false;
1048 }
1049
ipv6_add_addr_hash(struct net_device * dev,struct inet6_ifaddr * ifa)1050 static int ipv6_add_addr_hash(struct net_device *dev, struct inet6_ifaddr *ifa)
1051 {
1052 struct net *net = dev_net(dev);
1053 unsigned int hash = inet6_addr_hash(net, &ifa->addr);
1054 int err = 0;
1055
1056 spin_lock_bh(&net->ipv6.addrconf_hash_lock);
1057
1058 /* Ignore adding duplicate addresses on an interface */
1059 if (ipv6_chk_same_addr(net, &ifa->addr, dev, hash)) {
1060 netdev_dbg(dev, "ipv6_add_addr: already assigned\n");
1061 err = -EEXIST;
1062 } else {
1063 hlist_add_head_rcu(&ifa->addr_lst, &net->ipv6.inet6_addr_lst[hash]);
1064 }
1065
1066 spin_unlock_bh(&net->ipv6.addrconf_hash_lock);
1067
1068 return err;
1069 }
1070
1071 /* On success it returns ifp with increased reference count */
1072
1073 static struct inet6_ifaddr *
ipv6_add_addr(struct inet6_dev * idev,struct ifa6_config * cfg,bool can_block,struct netlink_ext_ack * extack)1074 ipv6_add_addr(struct inet6_dev *idev, struct ifa6_config *cfg,
1075 bool can_block, struct netlink_ext_ack *extack)
1076 {
1077 gfp_t gfp_flags = can_block ? GFP_KERNEL : GFP_ATOMIC;
1078 int addr_type = ipv6_addr_type(cfg->pfx);
1079 struct net *net = dev_net(idev->dev);
1080 struct inet6_ifaddr *ifa = NULL;
1081 struct fib6_info *f6i = NULL;
1082 int err = 0;
1083
1084 if (addr_type == IPV6_ADDR_ANY) {
1085 NL_SET_ERR_MSG_MOD(extack, "Invalid address");
1086 return ERR_PTR(-EADDRNOTAVAIL);
1087 } else if (addr_type & IPV6_ADDR_MULTICAST &&
1088 !(cfg->ifa_flags & IFA_F_MCAUTOJOIN)) {
1089 NL_SET_ERR_MSG_MOD(extack, "Cannot assign multicast address without \"IFA_F_MCAUTOJOIN\" flag");
1090 return ERR_PTR(-EADDRNOTAVAIL);
1091 } else if (!(idev->dev->flags & IFF_LOOPBACK) &&
1092 !netif_is_l3_master(idev->dev) &&
1093 addr_type & IPV6_ADDR_LOOPBACK) {
1094 NL_SET_ERR_MSG_MOD(extack, "Cannot assign loopback address on this device");
1095 return ERR_PTR(-EADDRNOTAVAIL);
1096 }
1097
1098 if (idev->dead) {
1099 NL_SET_ERR_MSG_MOD(extack, "device is going away");
1100 err = -ENODEV;
1101 goto out;
1102 }
1103
1104 if (idev->cnf.disable_ipv6) {
1105 NL_SET_ERR_MSG_MOD(extack, "IPv6 is disabled on this device");
1106 err = -EACCES;
1107 goto out;
1108 }
1109
1110 /* validator notifier needs to be blocking;
1111 * do not call in atomic context
1112 */
1113 if (can_block) {
1114 struct in6_validator_info i6vi = {
1115 .i6vi_addr = *cfg->pfx,
1116 .i6vi_dev = idev,
1117 .extack = extack,
1118 };
1119
1120 err = inet6addr_validator_notifier_call_chain(NETDEV_UP, &i6vi);
1121 err = notifier_to_errno(err);
1122 if (err < 0)
1123 goto out;
1124 }
1125
1126 ifa = kzalloc_obj(*ifa, gfp_flags | __GFP_ACCOUNT);
1127 if (!ifa) {
1128 err = -ENOBUFS;
1129 goto out;
1130 }
1131
1132 f6i = addrconf_f6i_alloc(net, idev, cfg->pfx, false, gfp_flags, extack);
1133 if (IS_ERR(f6i)) {
1134 err = PTR_ERR(f6i);
1135 f6i = NULL;
1136 goto out;
1137 }
1138
1139 neigh_parms_data_state_setall(idev->nd_parms);
1140
1141 ifa->addr = *cfg->pfx;
1142 if (cfg->peer_pfx)
1143 ifa->peer_addr = *cfg->peer_pfx;
1144
1145 spin_lock_init(&ifa->lock);
1146 INIT_DELAYED_WORK(&ifa->dad_work, addrconf_dad_work);
1147 INIT_HLIST_NODE(&ifa->addr_lst);
1148 ifa->scope = cfg->scope;
1149 ifa->prefix_len = cfg->plen;
1150 ifa->rt_priority = cfg->rt_priority;
1151 ifa->flags = cfg->ifa_flags;
1152 ifa->ifa_proto = cfg->ifa_proto;
1153 /* No need to add the TENTATIVE flag for addresses with NODAD */
1154 if (!(cfg->ifa_flags & IFA_F_NODAD))
1155 ifa->flags |= IFA_F_TENTATIVE;
1156 ifa->valid_lft = cfg->valid_lft;
1157 ifa->prefered_lft = cfg->preferred_lft;
1158 ifa->cstamp = ifa->tstamp = jiffies;
1159 ifa->tokenized = false;
1160
1161 ifa->rt = f6i;
1162
1163 ifa->idev = idev;
1164 in6_dev_hold(idev);
1165
1166 /* For caller */
1167 refcount_set(&ifa->refcnt, 1);
1168
1169 rcu_read_lock();
1170
1171 err = ipv6_add_addr_hash(idev->dev, ifa);
1172 if (err < 0) {
1173 rcu_read_unlock();
1174 goto out;
1175 }
1176
1177 write_lock_bh(&idev->lock);
1178
1179 /* Add to inet6_dev unicast addr list. */
1180 ipv6_link_dev_addr(idev, ifa);
1181
1182 if (ifa->flags&IFA_F_TEMPORARY) {
1183 /* manage_tempaddrs() relies on addresses being added to the head */
1184 list_add(&ifa->tmp_list, &idev->tempaddr_list);
1185 in6_ifa_hold(ifa);
1186 }
1187
1188 in6_ifa_hold(ifa);
1189 write_unlock_bh(&idev->lock);
1190
1191 rcu_read_unlock();
1192
1193 inet6addr_notifier_call_chain(NETDEV_UP, ifa);
1194 out:
1195 if (unlikely(err < 0)) {
1196 fib6_info_release(f6i);
1197
1198 if (ifa) {
1199 if (ifa->idev)
1200 in6_dev_put(ifa->idev);
1201 kfree(ifa);
1202 }
1203 ifa = ERR_PTR(err);
1204 }
1205
1206 return ifa;
1207 }
1208
1209 enum cleanup_prefix_rt_t {
1210 CLEANUP_PREFIX_RT_NOP, /* no cleanup action for prefix route */
1211 CLEANUP_PREFIX_RT_DEL, /* delete the prefix route */
1212 CLEANUP_PREFIX_RT_EXPIRE, /* update the lifetime of the prefix route */
1213 };
1214
1215 /*
1216 * Check, whether the prefix for ifp would still need a prefix route
1217 * after deleting ifp. The function returns one of the CLEANUP_PREFIX_RT_*
1218 * constants.
1219 *
1220 * 1) we don't purge prefix if address was not permanent.
1221 * prefix is managed by its own lifetime.
1222 * 2) we also don't purge, if the address was IFA_F_NOPREFIXROUTE.
1223 * 3) if there are no addresses, delete prefix.
1224 * 4) if there are still other permanent address(es),
1225 * corresponding prefix is still permanent.
1226 * 5) if there are still other addresses with IFA_F_NOPREFIXROUTE,
1227 * don't purge the prefix, assume user space is managing it.
1228 * 6) otherwise, update prefix lifetime to the
1229 * longest valid lifetime among the corresponding
1230 * addresses on the device.
1231 * Note: subsequent RA will update lifetime.
1232 **/
1233 static enum cleanup_prefix_rt_t
check_cleanup_prefix_route(struct inet6_ifaddr * ifp,unsigned long * expires)1234 check_cleanup_prefix_route(struct inet6_ifaddr *ifp, unsigned long *expires)
1235 {
1236 struct inet6_ifaddr *ifa;
1237 struct inet6_dev *idev = ifp->idev;
1238 unsigned long lifetime;
1239 enum cleanup_prefix_rt_t action = CLEANUP_PREFIX_RT_DEL;
1240
1241 *expires = jiffies;
1242
1243 list_for_each_entry(ifa, &idev->addr_list, if_list) {
1244 if (ifa == ifp)
1245 continue;
1246 if (ifa->prefix_len != ifp->prefix_len ||
1247 !ipv6_prefix_equal(&ifa->addr, &ifp->addr,
1248 ifp->prefix_len))
1249 continue;
1250 if (ifa->flags & (IFA_F_PERMANENT | IFA_F_NOPREFIXROUTE))
1251 return CLEANUP_PREFIX_RT_NOP;
1252
1253 action = CLEANUP_PREFIX_RT_EXPIRE;
1254
1255 spin_lock(&ifa->lock);
1256
1257 lifetime = addrconf_timeout_fixup(ifa->valid_lft, HZ);
1258 /*
1259 * Note: Because this address is
1260 * not permanent, lifetime <
1261 * LONG_MAX / HZ here.
1262 */
1263 if (time_before(*expires, ifa->tstamp + lifetime * HZ))
1264 *expires = ifa->tstamp + lifetime * HZ;
1265 spin_unlock(&ifa->lock);
1266 }
1267
1268 return action;
1269 }
1270
1271 static void
cleanup_prefix_route(struct inet6_ifaddr * ifp,unsigned long expires,bool del_rt,bool del_peer)1272 cleanup_prefix_route(struct inet6_ifaddr *ifp, unsigned long expires,
1273 bool del_rt, bool del_peer)
1274 {
1275 struct net *net = dev_net(ifp->idev->dev);
1276 struct fib6_table *table;
1277 struct fib6_info *f6i;
1278
1279 f6i = addrconf_get_prefix_route(del_peer ? &ifp->peer_addr : &ifp->addr,
1280 ifp->prefix_len,
1281 ifp->idev->dev, 0, RTF_DEFAULT, true);
1282 if (f6i) {
1283 if (del_rt)
1284 ip6_del_rt(net, f6i, false);
1285 else {
1286 if (f6i != net->ipv6.fib6_null_entry &&
1287 !(f6i->fib6_flags & RTF_EXPIRES)) {
1288 table = f6i->fib6_table;
1289 spin_lock_bh(&table->tb6_lock);
1290
1291 fib6_set_expires(f6i, expires);
1292 fib6_add_gc_list(f6i);
1293
1294 spin_unlock_bh(&table->tb6_lock);
1295 }
1296 fib6_info_release(f6i);
1297 }
1298 }
1299 }
1300
1301
1302 /* This function wants to get referenced ifp and releases it before return */
1303
ipv6_del_addr(struct inet6_ifaddr * ifp)1304 static void ipv6_del_addr(struct inet6_ifaddr *ifp)
1305 {
1306 enum cleanup_prefix_rt_t action = CLEANUP_PREFIX_RT_NOP;
1307 struct net *net = dev_net(ifp->idev->dev);
1308 unsigned long expires;
1309 int state;
1310
1311 ASSERT_RTNL();
1312
1313 spin_lock_bh(&ifp->lock);
1314 state = ifp->state;
1315 ifp->state = INET6_IFADDR_STATE_DEAD;
1316 spin_unlock_bh(&ifp->lock);
1317
1318 if (state == INET6_IFADDR_STATE_DEAD)
1319 goto out;
1320
1321 spin_lock_bh(&net->ipv6.addrconf_hash_lock);
1322 hlist_del_init_rcu(&ifp->addr_lst);
1323 spin_unlock_bh(&net->ipv6.addrconf_hash_lock);
1324
1325 write_lock_bh(&ifp->idev->lock);
1326
1327 if (ifp->flags&IFA_F_TEMPORARY) {
1328 list_del(&ifp->tmp_list);
1329 if (ifp->ifpub) {
1330 in6_ifa_put(ifp->ifpub);
1331 ifp->ifpub = NULL;
1332 }
1333 __in6_ifa_put(ifp);
1334 }
1335
1336 if (!(ifp->flags & IFA_F_NOPREFIXROUTE))
1337 action = check_cleanup_prefix_route(ifp, &expires);
1338
1339 list_del_rcu(&ifp->if_list);
1340 __in6_ifa_put(ifp);
1341
1342 write_unlock_bh(&ifp->idev->lock);
1343
1344 addrconf_del_dad_work(ifp);
1345
1346 ipv6_ifa_notify(RTM_DELADDR, ifp);
1347
1348 inet6addr_notifier_call_chain(NETDEV_DOWN, ifp);
1349
1350 if (action != CLEANUP_PREFIX_RT_NOP) {
1351 cleanup_prefix_route(ifp, expires,
1352 action == CLEANUP_PREFIX_RT_DEL, false);
1353 }
1354
1355 /* clean up prefsrc entries */
1356 rt6_remove_prefsrc(ifp);
1357 out:
1358 in6_ifa_put(ifp);
1359 }
1360
ipv6_get_regen_advance(const struct inet6_dev * idev)1361 static unsigned long ipv6_get_regen_advance(const struct inet6_dev *idev)
1362 {
1363 return READ_ONCE(idev->cnf.regen_min_advance) +
1364 READ_ONCE(idev->cnf.regen_max_retry) *
1365 READ_ONCE(idev->cnf.dad_transmits) *
1366 max(NEIGH_VAR(idev->nd_parms, RETRANS_TIME), HZ/100) / HZ;
1367 }
1368
ipv6_create_tempaddr(struct inet6_ifaddr * ifp,bool block)1369 static int ipv6_create_tempaddr(struct inet6_ifaddr *ifp, bool block)
1370 {
1371 struct inet6_dev *idev = ifp->idev;
1372 unsigned long tmp_tstamp, age;
1373 unsigned long regen_advance;
1374 unsigned long now = jiffies;
1375 u32 if_public_preferred_lft;
1376 s32 cnf_temp_preferred_lft;
1377 struct inet6_ifaddr *ift;
1378 struct ifa6_config cfg;
1379 long max_desync_factor;
1380 struct in6_addr addr;
1381 int ret = 0;
1382
1383 write_lock_bh(&idev->lock);
1384
1385 retry:
1386 in6_dev_hold(idev);
1387 if (READ_ONCE(idev->cnf.use_tempaddr) <= 0) {
1388 write_unlock_bh(&idev->lock);
1389 pr_info("%s: use_tempaddr is disabled\n", __func__);
1390 in6_dev_put(idev);
1391 ret = -1;
1392 goto out;
1393 }
1394 spin_lock_bh(&ifp->lock);
1395 if (ifp->regen_count++ >= READ_ONCE(idev->cnf.regen_max_retry)) {
1396 WRITE_ONCE(idev->cnf.use_tempaddr, -1); /*XXX*/
1397 spin_unlock_bh(&ifp->lock);
1398 write_unlock_bh(&idev->lock);
1399 pr_warn("%s: regeneration time exceeded - disabled temporary address support\n",
1400 __func__);
1401 in6_dev_put(idev);
1402 ret = -1;
1403 goto out;
1404 }
1405 in6_ifa_hold(ifp);
1406 memcpy(addr.s6_addr, ifp->addr.s6_addr, 8);
1407 ipv6_gen_rnd_iid(&addr);
1408
1409 age = (now - ifp->tstamp) / HZ;
1410
1411 regen_advance = ipv6_get_regen_advance(idev);
1412
1413 /* recalculate max_desync_factor each time and update
1414 * idev->desync_factor if it's larger
1415 */
1416 cnf_temp_preferred_lft = READ_ONCE(idev->cnf.temp_prefered_lft);
1417 max_desync_factor = min_t(long,
1418 READ_ONCE(idev->cnf.max_desync_factor),
1419 cnf_temp_preferred_lft - regen_advance);
1420
1421 if (unlikely(idev->desync_factor > max_desync_factor)) {
1422 if (max_desync_factor > 0) {
1423 get_random_bytes(&idev->desync_factor,
1424 sizeof(idev->desync_factor));
1425 idev->desync_factor %= max_desync_factor;
1426 } else {
1427 idev->desync_factor = 0;
1428 }
1429 }
1430
1431 if_public_preferred_lft = ifp->prefered_lft;
1432
1433 memset(&cfg, 0, sizeof(cfg));
1434 cfg.valid_lft = min_t(__u32, ifp->valid_lft,
1435 READ_ONCE(idev->cnf.temp_valid_lft) + age);
1436 cfg.preferred_lft = cnf_temp_preferred_lft + age - idev->desync_factor;
1437 cfg.preferred_lft = min_t(__u32, if_public_preferred_lft, cfg.preferred_lft);
1438 cfg.preferred_lft = min_t(__u32, cfg.valid_lft, cfg.preferred_lft);
1439
1440 cfg.plen = ifp->prefix_len;
1441 tmp_tstamp = ifp->tstamp;
1442 spin_unlock_bh(&ifp->lock);
1443
1444 write_unlock_bh(&idev->lock);
1445
1446 /* From RFC 4941:
1447 *
1448 * A temporary address is created only if this calculated Preferred
1449 * Lifetime is greater than REGEN_ADVANCE time units. In
1450 * particular, an implementation must not create a temporary address
1451 * with a zero Preferred Lifetime.
1452 *
1453 * ...
1454 *
1455 * When creating a temporary address, the lifetime values MUST be
1456 * derived from the corresponding prefix as follows:
1457 *
1458 * ...
1459 *
1460 * * Its Preferred Lifetime is the lower of the Preferred Lifetime
1461 * of the public address or TEMP_PREFERRED_LIFETIME -
1462 * DESYNC_FACTOR.
1463 *
1464 * To comply with the RFC's requirements, clamp the preferred lifetime
1465 * to a minimum of regen_advance, unless that would exceed valid_lft or
1466 * ifp->prefered_lft.
1467 *
1468 * Use age calculation as in addrconf_verify to avoid unnecessary
1469 * temporary addresses being generated.
1470 */
1471 age = (now - tmp_tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
1472 if (cfg.preferred_lft <= regen_advance + age) {
1473 cfg.preferred_lft = regen_advance + age + 1;
1474 if (cfg.preferred_lft > cfg.valid_lft ||
1475 cfg.preferred_lft > if_public_preferred_lft) {
1476 in6_ifa_put(ifp);
1477 in6_dev_put(idev);
1478 ret = -1;
1479 goto out;
1480 }
1481 }
1482
1483 cfg.ifa_flags = IFA_F_TEMPORARY;
1484 /* set in addrconf_prefix_rcv() */
1485 if (ifp->flags & IFA_F_OPTIMISTIC)
1486 cfg.ifa_flags |= IFA_F_OPTIMISTIC;
1487
1488 cfg.pfx = &addr;
1489 cfg.scope = ipv6_addr_scope(cfg.pfx);
1490
1491 ift = ipv6_add_addr(idev, &cfg, block, NULL);
1492 if (IS_ERR(ift)) {
1493 in6_ifa_put(ifp);
1494 in6_dev_put(idev);
1495 pr_info("%s: retry temporary address regeneration\n", __func__);
1496 write_lock_bh(&idev->lock);
1497 goto retry;
1498 }
1499
1500 spin_lock_bh(&ift->lock);
1501 ift->ifpub = ifp;
1502 ift->cstamp = now;
1503 ift->tstamp = tmp_tstamp;
1504 spin_unlock_bh(&ift->lock);
1505
1506 addrconf_dad_start(ift);
1507 in6_ifa_put(ift);
1508 in6_dev_put(idev);
1509 out:
1510 return ret;
1511 }
1512
1513 /*
1514 * Choose an appropriate source address (RFC3484)
1515 */
1516 enum {
1517 IPV6_SADDR_RULE_INIT = 0,
1518 IPV6_SADDR_RULE_LOCAL,
1519 IPV6_SADDR_RULE_SCOPE,
1520 IPV6_SADDR_RULE_PREFERRED,
1521 #ifdef CONFIG_IPV6_MIP6
1522 IPV6_SADDR_RULE_HOA,
1523 #endif
1524 IPV6_SADDR_RULE_OIF,
1525 IPV6_SADDR_RULE_LABEL,
1526 IPV6_SADDR_RULE_PRIVACY,
1527 IPV6_SADDR_RULE_ORCHID,
1528 IPV6_SADDR_RULE_PREFIX,
1529 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1530 IPV6_SADDR_RULE_NOT_OPTIMISTIC,
1531 #endif
1532 IPV6_SADDR_RULE_MAX
1533 };
1534
1535 struct ipv6_saddr_score {
1536 int rule;
1537 int addr_type;
1538 struct inet6_ifaddr *ifa;
1539 DECLARE_BITMAP(scorebits, IPV6_SADDR_RULE_MAX);
1540 int scopedist;
1541 int matchlen;
1542 };
1543
1544 struct ipv6_saddr_dst {
1545 const struct in6_addr *addr;
1546 int ifindex;
1547 int scope;
1548 int label;
1549 unsigned int prefs;
1550 };
1551
ipv6_saddr_preferred(int type)1552 static inline int ipv6_saddr_preferred(int type)
1553 {
1554 if (type & (IPV6_ADDR_MAPPED|IPV6_ADDR_COMPATv4|IPV6_ADDR_LOOPBACK))
1555 return 1;
1556 return 0;
1557 }
1558
ipv6_use_optimistic_addr(const struct net * net,const struct inet6_dev * idev)1559 static bool ipv6_use_optimistic_addr(const struct net *net,
1560 const struct inet6_dev *idev)
1561 {
1562 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1563 if (!idev)
1564 return false;
1565 if (!READ_ONCE(net->ipv6.devconf_all->optimistic_dad) &&
1566 !READ_ONCE(idev->cnf.optimistic_dad))
1567 return false;
1568 if (!READ_ONCE(net->ipv6.devconf_all->use_optimistic) &&
1569 !READ_ONCE(idev->cnf.use_optimistic))
1570 return false;
1571
1572 return true;
1573 #else
1574 return false;
1575 #endif
1576 }
1577
ipv6_allow_optimistic_dad(const struct net * net,const struct inet6_dev * idev)1578 static bool ipv6_allow_optimistic_dad(const struct net *net,
1579 const struct inet6_dev *idev)
1580 {
1581 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1582 if (!idev)
1583 return false;
1584 if (!READ_ONCE(net->ipv6.devconf_all->optimistic_dad) &&
1585 !READ_ONCE(idev->cnf.optimistic_dad))
1586 return false;
1587
1588 return true;
1589 #else
1590 return false;
1591 #endif
1592 }
1593
ipv6_get_saddr_eval(struct net * net,struct ipv6_saddr_score * score,struct ipv6_saddr_dst * dst,int i)1594 static int ipv6_get_saddr_eval(struct net *net,
1595 struct ipv6_saddr_score *score,
1596 struct ipv6_saddr_dst *dst,
1597 int i)
1598 {
1599 int ret;
1600
1601 if (i <= score->rule) {
1602 switch (i) {
1603 case IPV6_SADDR_RULE_SCOPE:
1604 ret = score->scopedist;
1605 break;
1606 case IPV6_SADDR_RULE_PREFIX:
1607 ret = score->matchlen;
1608 break;
1609 default:
1610 ret = !!test_bit(i, score->scorebits);
1611 }
1612 goto out;
1613 }
1614
1615 switch (i) {
1616 case IPV6_SADDR_RULE_INIT:
1617 /* Rule 0: remember if hiscore is not ready yet */
1618 ret = !!score->ifa;
1619 break;
1620 case IPV6_SADDR_RULE_LOCAL:
1621 /* Rule 1: Prefer same address */
1622 ret = ipv6_addr_equal(&score->ifa->addr, dst->addr);
1623 break;
1624 case IPV6_SADDR_RULE_SCOPE:
1625 /* Rule 2: Prefer appropriate scope
1626 *
1627 * ret
1628 * ^
1629 * -1 | d 15
1630 * ---+--+-+---> scope
1631 * |
1632 * | d is scope of the destination.
1633 * B-d | \
1634 * | \ <- smaller scope is better if
1635 * B-15 | \ if scope is enough for destination.
1636 * | ret = B - scope (-1 <= scope >= d <= 15).
1637 * d-C-1 | /
1638 * |/ <- greater is better
1639 * -C / if scope is not enough for destination.
1640 * /| ret = scope - C (-1 <= d < scope <= 15).
1641 *
1642 * d - C - 1 < B -15 (for all -1 <= d <= 15).
1643 * C > d + 14 - B >= 15 + 14 - B = 29 - B.
1644 * Assume B = 0 and we get C > 29.
1645 */
1646 ret = __ipv6_addr_src_scope(score->addr_type);
1647 if (ret >= dst->scope)
1648 ret = -ret;
1649 else
1650 ret -= 128; /* 30 is enough */
1651 score->scopedist = ret;
1652 break;
1653 case IPV6_SADDR_RULE_PREFERRED:
1654 {
1655 /* Rule 3: Avoid deprecated and optimistic addresses */
1656 u8 avoid = IFA_F_DEPRECATED;
1657
1658 if (!ipv6_use_optimistic_addr(net, score->ifa->idev))
1659 avoid |= IFA_F_OPTIMISTIC;
1660 ret = ipv6_saddr_preferred(score->addr_type) ||
1661 !(score->ifa->flags & avoid);
1662 break;
1663 }
1664 #ifdef CONFIG_IPV6_MIP6
1665 case IPV6_SADDR_RULE_HOA:
1666 {
1667 /* Rule 4: Prefer home address */
1668 int prefhome = !(dst->prefs & IPV6_PREFER_SRC_COA);
1669 ret = !(score->ifa->flags & IFA_F_HOMEADDRESS) ^ prefhome;
1670 break;
1671 }
1672 #endif
1673 case IPV6_SADDR_RULE_OIF:
1674 /* Rule 5: Prefer outgoing interface */
1675 ret = (!dst->ifindex ||
1676 dst->ifindex == score->ifa->idev->dev->ifindex);
1677 break;
1678 case IPV6_SADDR_RULE_LABEL:
1679 /* Rule 6: Prefer matching label */
1680 ret = ipv6_addr_label(net,
1681 &score->ifa->addr, score->addr_type,
1682 score->ifa->idev->dev->ifindex) == dst->label;
1683 break;
1684 case IPV6_SADDR_RULE_PRIVACY:
1685 {
1686 /* Rule 7: Prefer public address
1687 * Note: prefer temporary address if use_tempaddr >= 2
1688 */
1689 int preftmp = dst->prefs & (IPV6_PREFER_SRC_PUBLIC|IPV6_PREFER_SRC_TMP) ?
1690 !!(dst->prefs & IPV6_PREFER_SRC_TMP) :
1691 READ_ONCE(score->ifa->idev->cnf.use_tempaddr) >= 2;
1692 ret = (!(score->ifa->flags & IFA_F_TEMPORARY)) ^ preftmp;
1693 break;
1694 }
1695 case IPV6_SADDR_RULE_ORCHID:
1696 /* Rule 8-: Prefer ORCHID vs ORCHID or
1697 * non-ORCHID vs non-ORCHID
1698 */
1699 ret = !(ipv6_addr_orchid(&score->ifa->addr) ^
1700 ipv6_addr_orchid(dst->addr));
1701 break;
1702 case IPV6_SADDR_RULE_PREFIX:
1703 /* Rule 8: Use longest matching prefix */
1704 ret = ipv6_addr_diff(&score->ifa->addr, dst->addr);
1705 if (ret > score->ifa->prefix_len)
1706 ret = score->ifa->prefix_len;
1707 score->matchlen = ret;
1708 break;
1709 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1710 case IPV6_SADDR_RULE_NOT_OPTIMISTIC:
1711 /* Optimistic addresses still have lower precedence than other
1712 * preferred addresses.
1713 */
1714 ret = !(score->ifa->flags & IFA_F_OPTIMISTIC);
1715 break;
1716 #endif
1717 default:
1718 ret = 0;
1719 }
1720
1721 if (ret)
1722 __set_bit(i, score->scorebits);
1723 score->rule = i;
1724 out:
1725 return ret;
1726 }
1727
__ipv6_dev_get_saddr(struct net * net,struct ipv6_saddr_dst * dst,struct inet6_dev * idev,struct ipv6_saddr_score * scores,int hiscore_idx)1728 static int __ipv6_dev_get_saddr(struct net *net,
1729 struct ipv6_saddr_dst *dst,
1730 struct inet6_dev *idev,
1731 struct ipv6_saddr_score *scores,
1732 int hiscore_idx)
1733 {
1734 struct ipv6_saddr_score *score = &scores[1 - hiscore_idx], *hiscore = &scores[hiscore_idx];
1735
1736 list_for_each_entry_rcu(score->ifa, &idev->addr_list, if_list) {
1737 int i;
1738
1739 /*
1740 * - Tentative Address (RFC2462 section 5.4)
1741 * - A tentative address is not considered
1742 * "assigned to an interface" in the traditional
1743 * sense, unless it is also flagged as optimistic.
1744 * - Candidate Source Address (section 4)
1745 * - In any case, anycast addresses, multicast
1746 * addresses, and the unspecified address MUST
1747 * NOT be included in a candidate set.
1748 */
1749 if ((score->ifa->flags & IFA_F_TENTATIVE) &&
1750 (!(score->ifa->flags & IFA_F_OPTIMISTIC)))
1751 continue;
1752
1753 score->addr_type = __ipv6_addr_type(&score->ifa->addr);
1754
1755 if (unlikely(score->addr_type == IPV6_ADDR_ANY ||
1756 score->addr_type & IPV6_ADDR_MULTICAST)) {
1757 net_dbg_ratelimited("ADDRCONF: unspecified / multicast address assigned as unicast address on %s",
1758 idev->dev->name);
1759 continue;
1760 }
1761
1762 score->rule = -1;
1763 bitmap_zero(score->scorebits, IPV6_SADDR_RULE_MAX);
1764
1765 for (i = 0; i < IPV6_SADDR_RULE_MAX; i++) {
1766 int minihiscore, miniscore;
1767
1768 minihiscore = ipv6_get_saddr_eval(net, hiscore, dst, i);
1769 miniscore = ipv6_get_saddr_eval(net, score, dst, i);
1770
1771 if (minihiscore > miniscore) {
1772 if (i == IPV6_SADDR_RULE_SCOPE &&
1773 score->scopedist > 0) {
1774 /*
1775 * special case:
1776 * each remaining entry
1777 * has too small (not enough)
1778 * scope, because ifa entries
1779 * are sorted by their scope
1780 * values.
1781 */
1782 goto out;
1783 }
1784 break;
1785 } else if (minihiscore < miniscore) {
1786 swap(hiscore, score);
1787 hiscore_idx = 1 - hiscore_idx;
1788
1789 /* restore our iterator */
1790 score->ifa = hiscore->ifa;
1791
1792 break;
1793 }
1794 }
1795 }
1796 out:
1797 return hiscore_idx;
1798 }
1799
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)1800 static int ipv6_get_saddr_master(struct net *net,
1801 const struct net_device *dst_dev,
1802 const struct net_device *master,
1803 struct ipv6_saddr_dst *dst,
1804 struct ipv6_saddr_score *scores,
1805 int hiscore_idx)
1806 {
1807 struct inet6_dev *idev;
1808
1809 idev = __in6_dev_get(dst_dev);
1810 if (idev)
1811 hiscore_idx = __ipv6_dev_get_saddr(net, dst, idev,
1812 scores, hiscore_idx);
1813
1814 idev = __in6_dev_get(master);
1815 if (idev)
1816 hiscore_idx = __ipv6_dev_get_saddr(net, dst, idev,
1817 scores, hiscore_idx);
1818
1819 return hiscore_idx;
1820 }
1821
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)1822 int ipv6_dev_get_saddr(struct net *net, const struct net_device *dst_dev,
1823 const struct in6_addr *daddr, unsigned int prefs,
1824 struct in6_addr *saddr)
1825 {
1826 struct ipv6_saddr_score scores[2], *hiscore;
1827 struct ipv6_saddr_dst dst;
1828 struct inet6_dev *idev;
1829 struct net_device *dev;
1830 int dst_type;
1831 bool use_oif_addr = false;
1832 int hiscore_idx = 0;
1833 int ret = 0;
1834
1835 dst_type = __ipv6_addr_type(daddr);
1836 dst.addr = daddr;
1837 dst.ifindex = dst_dev ? dst_dev->ifindex : 0;
1838 dst.scope = __ipv6_addr_src_scope(dst_type);
1839 dst.label = ipv6_addr_label(net, daddr, dst_type, dst.ifindex);
1840 dst.prefs = prefs;
1841
1842 scores[hiscore_idx].rule = -1;
1843 scores[hiscore_idx].ifa = NULL;
1844
1845 rcu_read_lock();
1846
1847 /* Candidate Source Address (section 4)
1848 * - multicast and link-local destination address,
1849 * the set of candidate source address MUST only
1850 * include addresses assigned to interfaces
1851 * belonging to the same link as the outgoing
1852 * interface.
1853 * (- For site-local destination addresses, the
1854 * set of candidate source addresses MUST only
1855 * include addresses assigned to interfaces
1856 * belonging to the same site as the outgoing
1857 * interface.)
1858 * - "It is RECOMMENDED that the candidate source addresses
1859 * be the set of unicast addresses assigned to the
1860 * interface that will be used to send to the destination
1861 * (the 'outgoing' interface)." (RFC 6724)
1862 */
1863 if (dst_dev) {
1864 idev = __in6_dev_get(dst_dev);
1865 if ((dst_type & IPV6_ADDR_MULTICAST) ||
1866 dst.scope <= IPV6_ADDR_SCOPE_LINKLOCAL ||
1867 (idev && READ_ONCE(idev->cnf.use_oif_addrs_only))) {
1868 use_oif_addr = true;
1869 }
1870 }
1871
1872 if (use_oif_addr) {
1873 if (idev)
1874 hiscore_idx = __ipv6_dev_get_saddr(net, &dst, idev, scores, hiscore_idx);
1875 } else {
1876 const struct net_device *master;
1877 int master_idx = 0;
1878
1879 /* if dst_dev exists and is enslaved to an L3 device, then
1880 * prefer addresses from dst_dev and then the master over
1881 * any other enslaved devices in the L3 domain.
1882 */
1883 master = l3mdev_master_dev_rcu(dst_dev);
1884 if (master) {
1885 master_idx = master->ifindex;
1886
1887 hiscore_idx = ipv6_get_saddr_master(net, dst_dev,
1888 master, &dst,
1889 scores, hiscore_idx);
1890
1891 if (scores[hiscore_idx].ifa &&
1892 scores[hiscore_idx].scopedist >= 0)
1893 goto out;
1894 }
1895
1896 for_each_netdev_rcu(net, dev) {
1897 /* only consider addresses on devices in the
1898 * same L3 domain
1899 */
1900 if (l3mdev_master_ifindex_rcu(dev) != master_idx)
1901 continue;
1902 idev = __in6_dev_get(dev);
1903 if (!idev)
1904 continue;
1905 hiscore_idx = __ipv6_dev_get_saddr(net, &dst, idev, scores, hiscore_idx);
1906 }
1907 }
1908
1909 out:
1910 hiscore = &scores[hiscore_idx];
1911 if (!hiscore->ifa)
1912 ret = -EADDRNOTAVAIL;
1913 else
1914 *saddr = hiscore->ifa->addr;
1915
1916 rcu_read_unlock();
1917 return ret;
1918 }
1919 EXPORT_SYMBOL(ipv6_dev_get_saddr);
1920
__ipv6_get_lladdr(struct inet6_dev * idev,struct in6_addr * addr,u32 banned_flags)1921 static int __ipv6_get_lladdr(struct inet6_dev *idev, struct in6_addr *addr,
1922 u32 banned_flags)
1923 {
1924 struct inet6_ifaddr *ifp;
1925 int err = -EADDRNOTAVAIL;
1926
1927 list_for_each_entry_reverse(ifp, &idev->addr_list, if_list) {
1928 if (ifp->scope > IFA_LINK)
1929 break;
1930 if (ifp->scope == IFA_LINK &&
1931 !(ifp->flags & banned_flags)) {
1932 *addr = ifp->addr;
1933 err = 0;
1934 break;
1935 }
1936 }
1937 return err;
1938 }
1939
ipv6_get_lladdr(struct net_device * dev,struct in6_addr * addr,u32 banned_flags)1940 int ipv6_get_lladdr(struct net_device *dev, struct in6_addr *addr,
1941 u32 banned_flags)
1942 {
1943 struct inet6_dev *idev;
1944 int err = -EADDRNOTAVAIL;
1945
1946 rcu_read_lock();
1947 idev = __in6_dev_get(dev);
1948 if (idev) {
1949 read_lock_bh(&idev->lock);
1950 err = __ipv6_get_lladdr(idev, addr, banned_flags);
1951 read_unlock_bh(&idev->lock);
1952 }
1953 rcu_read_unlock();
1954 return err;
1955 }
1956
ipv6_count_addresses(const struct inet6_dev * idev)1957 static int ipv6_count_addresses(const struct inet6_dev *idev)
1958 {
1959 const struct inet6_ifaddr *ifp;
1960 int cnt = 0;
1961
1962 rcu_read_lock();
1963 list_for_each_entry_rcu(ifp, &idev->addr_list, if_list)
1964 cnt++;
1965 rcu_read_unlock();
1966 return cnt;
1967 }
1968
ipv6_chk_addr(struct net * net,const struct in6_addr * addr,const struct net_device * dev,int strict)1969 int ipv6_chk_addr(struct net *net, const struct in6_addr *addr,
1970 const struct net_device *dev, int strict)
1971 {
1972 return ipv6_chk_addr_and_flags(net, addr, dev, !dev,
1973 strict, IFA_F_TENTATIVE);
1974 }
1975 EXPORT_SYMBOL(ipv6_chk_addr);
1976
1977 /* device argument is used to find the L3 domain of interest. If
1978 * skip_dev_check is set, then the ifp device is not checked against
1979 * the passed in dev argument. So the 2 cases for addresses checks are:
1980 * 1. does the address exist in the L3 domain that dev is part of
1981 * (skip_dev_check = true), or
1982 *
1983 * 2. does the address exist on the specific device
1984 * (skip_dev_check = false)
1985 */
1986 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)1987 __ipv6_chk_addr_and_flags(struct net *net, const struct in6_addr *addr,
1988 const struct net_device *dev, bool skip_dev_check,
1989 int strict, u32 banned_flags)
1990 {
1991 unsigned int hash = inet6_addr_hash(net, addr);
1992 struct net_device *l3mdev, *ndev;
1993 struct inet6_ifaddr *ifp;
1994 u32 ifp_flags;
1995
1996 rcu_read_lock();
1997
1998 l3mdev = l3mdev_master_dev_rcu(dev);
1999 if (skip_dev_check)
2000 dev = NULL;
2001
2002 hlist_for_each_entry_rcu(ifp, &net->ipv6.inet6_addr_lst[hash], addr_lst) {
2003 ndev = ifp->idev->dev;
2004
2005 if (l3mdev_master_dev_rcu(ndev) != l3mdev)
2006 continue;
2007
2008 /* Decouple optimistic from tentative for evaluation here.
2009 * Ban optimistic addresses explicitly, when required.
2010 */
2011 ifp_flags = (ifp->flags&IFA_F_OPTIMISTIC)
2012 ? (ifp->flags&~IFA_F_TENTATIVE)
2013 : ifp->flags;
2014 if (ipv6_addr_equal(&ifp->addr, addr) &&
2015 !(ifp_flags&banned_flags) &&
2016 (!dev || ndev == dev ||
2017 !(ifp->scope&(IFA_LINK|IFA_HOST) || strict))) {
2018 rcu_read_unlock();
2019 return ndev;
2020 }
2021 }
2022
2023 rcu_read_unlock();
2024 return NULL;
2025 }
2026
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)2027 int ipv6_chk_addr_and_flags(struct net *net, const struct in6_addr *addr,
2028 const struct net_device *dev, bool skip_dev_check,
2029 int strict, u32 banned_flags)
2030 {
2031 return __ipv6_chk_addr_and_flags(net, addr, dev, skip_dev_check,
2032 strict, banned_flags) ? 1 : 0;
2033 }
2034 EXPORT_SYMBOL(ipv6_chk_addr_and_flags);
2035
2036
2037 /* Compares an address/prefix_len with addresses on device @dev.
2038 * If one is found it returns true.
2039 */
ipv6_chk_custom_prefix(const struct in6_addr * addr,const unsigned int prefix_len,struct net_device * dev)2040 bool ipv6_chk_custom_prefix(const struct in6_addr *addr,
2041 const unsigned int prefix_len, struct net_device *dev)
2042 {
2043 const struct inet6_ifaddr *ifa;
2044 const struct inet6_dev *idev;
2045 bool ret = false;
2046
2047 rcu_read_lock();
2048 idev = __in6_dev_get(dev);
2049 if (idev) {
2050 list_for_each_entry_rcu(ifa, &idev->addr_list, if_list) {
2051 ret = ipv6_prefix_equal(addr, &ifa->addr, prefix_len);
2052 if (ret)
2053 break;
2054 }
2055 }
2056 rcu_read_unlock();
2057
2058 return ret;
2059 }
2060 EXPORT_SYMBOL(ipv6_chk_custom_prefix);
2061
ipv6_chk_prefix(const struct in6_addr * addr,struct net_device * dev)2062 int ipv6_chk_prefix(const struct in6_addr *addr, struct net_device *dev)
2063 {
2064 const struct inet6_ifaddr *ifa;
2065 const struct inet6_dev *idev;
2066 int onlink;
2067
2068 onlink = 0;
2069 rcu_read_lock();
2070 idev = __in6_dev_get(dev);
2071 if (idev) {
2072 list_for_each_entry_rcu(ifa, &idev->addr_list, if_list) {
2073 onlink = ipv6_prefix_equal(addr, &ifa->addr,
2074 ifa->prefix_len);
2075 if (onlink)
2076 break;
2077 }
2078 }
2079 rcu_read_unlock();
2080 return onlink;
2081 }
2082 EXPORT_SYMBOL(ipv6_chk_prefix);
2083
2084 /**
2085 * ipv6_dev_find - find the first device with a given source address.
2086 * @net: the net namespace
2087 * @addr: the source address
2088 * @dev: used to find the L3 domain of interest
2089 *
2090 * The caller should be protected by RCU, or RTNL.
2091 */
ipv6_dev_find(struct net * net,const struct in6_addr * addr,struct net_device * dev)2092 struct net_device *ipv6_dev_find(struct net *net, const struct in6_addr *addr,
2093 struct net_device *dev)
2094 {
2095 return __ipv6_chk_addr_and_flags(net, addr, dev, !dev, 1,
2096 IFA_F_TENTATIVE);
2097 }
2098 EXPORT_SYMBOL(ipv6_dev_find);
2099
ipv6_get_ifaddr(struct net * net,const struct in6_addr * addr,struct net_device * dev,int strict)2100 struct inet6_ifaddr *ipv6_get_ifaddr(struct net *net, const struct in6_addr *addr,
2101 struct net_device *dev, int strict)
2102 {
2103 unsigned int hash = inet6_addr_hash(net, addr);
2104 struct inet6_ifaddr *ifp, *result = NULL;
2105
2106 rcu_read_lock();
2107 hlist_for_each_entry_rcu(ifp, &net->ipv6.inet6_addr_lst[hash], addr_lst) {
2108 if (ipv6_addr_equal(&ifp->addr, addr)) {
2109 if (!dev || ifp->idev->dev == dev ||
2110 !(ifp->scope&(IFA_LINK|IFA_HOST) || strict)) {
2111 if (in6_ifa_hold_safe(ifp)) {
2112 result = ifp;
2113 break;
2114 }
2115 }
2116 }
2117 }
2118 rcu_read_unlock();
2119
2120 return result;
2121 }
2122
2123 /* Gets referenced address, destroys ifaddr */
2124
addrconf_dad_stop(struct inet6_ifaddr * ifp,int dad_failed)2125 static void addrconf_dad_stop(struct inet6_ifaddr *ifp, int dad_failed)
2126 {
2127 if (dad_failed)
2128 ifp->flags |= IFA_F_DADFAILED;
2129
2130 if (ifp->flags&IFA_F_TEMPORARY) {
2131 struct inet6_ifaddr *ifpub;
2132 spin_lock_bh(&ifp->lock);
2133 ifpub = ifp->ifpub;
2134 if (ifpub) {
2135 in6_ifa_hold(ifpub);
2136 spin_unlock_bh(&ifp->lock);
2137 ipv6_create_tempaddr(ifpub, true);
2138 in6_ifa_put(ifpub);
2139 } else {
2140 spin_unlock_bh(&ifp->lock);
2141 }
2142 ipv6_del_addr(ifp);
2143 } else if (ifp->flags&IFA_F_PERMANENT || !dad_failed) {
2144 spin_lock_bh(&ifp->lock);
2145 addrconf_del_dad_work(ifp);
2146 ifp->flags |= IFA_F_TENTATIVE;
2147 if (dad_failed)
2148 ifp->flags &= ~IFA_F_OPTIMISTIC;
2149 spin_unlock_bh(&ifp->lock);
2150 if (dad_failed)
2151 ipv6_ifa_notify(0, ifp);
2152 in6_ifa_put(ifp);
2153 } else {
2154 ipv6_del_addr(ifp);
2155 }
2156 }
2157
addrconf_dad_end(struct inet6_ifaddr * ifp)2158 static int addrconf_dad_end(struct inet6_ifaddr *ifp)
2159 {
2160 int err = -ENOENT;
2161
2162 spin_lock_bh(&ifp->lock);
2163 if (ifp->state == INET6_IFADDR_STATE_DAD) {
2164 ifp->state = INET6_IFADDR_STATE_POSTDAD;
2165 err = 0;
2166 }
2167 spin_unlock_bh(&ifp->lock);
2168
2169 return err;
2170 }
2171
addrconf_dad_failure(struct sk_buff * skb,struct inet6_ifaddr * ifp)2172 void addrconf_dad_failure(struct sk_buff *skb, struct inet6_ifaddr *ifp)
2173 {
2174 struct inet6_dev *idev = ifp->idev;
2175 struct net *net = dev_net(idev->dev);
2176 int max_addresses;
2177
2178 spin_lock_bh(&ifp->lock);
2179
2180 if (ifp->state != INET6_IFADDR_STATE_DAD) {
2181 spin_unlock_bh(&ifp->lock);
2182 in6_ifa_put(ifp);
2183 return;
2184 }
2185 ifp->state = INET6_IFADDR_STATE_POSTDAD;
2186
2187 net_info_ratelimited("%s: IPv6 duplicate address %pI6c used by %pM detected!\n",
2188 ifp->idev->dev->name, &ifp->addr, eth_hdr(skb)->h_source);
2189
2190 if (ifp->flags & IFA_F_STABLE_PRIVACY) {
2191 struct in6_addr new_addr;
2192 struct inet6_ifaddr *ifp2;
2193 int retries = ifp->stable_privacy_retry + 1;
2194 struct ifa6_config cfg = {
2195 .pfx = &new_addr,
2196 .plen = ifp->prefix_len,
2197 .ifa_flags = ifp->flags,
2198 .valid_lft = ifp->valid_lft,
2199 .preferred_lft = ifp->prefered_lft,
2200 .scope = ifp->scope,
2201 };
2202
2203 if (retries > net->ipv6.sysctl.idgen_retries) {
2204 net_info_ratelimited("%s: privacy stable address generation failed because of DAD conflicts!\n",
2205 ifp->idev->dev->name);
2206 goto errdad;
2207 }
2208
2209 new_addr = ifp->addr;
2210 if (ipv6_generate_stable_address(&new_addr, retries,
2211 idev))
2212 goto errdad;
2213
2214 spin_unlock_bh(&ifp->lock);
2215
2216 max_addresses = READ_ONCE(idev->cnf.max_addresses);
2217 if (max_addresses &&
2218 ipv6_count_addresses(idev) >= max_addresses)
2219 goto lock_errdad;
2220
2221 net_info_ratelimited("%s: generating new stable privacy address because of DAD conflict\n",
2222 ifp->idev->dev->name);
2223
2224 ifp2 = ipv6_add_addr(idev, &cfg, false, NULL);
2225 if (IS_ERR(ifp2))
2226 goto lock_errdad;
2227
2228 spin_lock_bh(&ifp2->lock);
2229 ifp2->stable_privacy_retry = retries;
2230 ifp2->state = INET6_IFADDR_STATE_PREDAD;
2231 spin_unlock_bh(&ifp2->lock);
2232
2233 addrconf_mod_dad_work(ifp2, net->ipv6.sysctl.idgen_delay);
2234 in6_ifa_put(ifp2);
2235 lock_errdad:
2236 spin_lock_bh(&ifp->lock);
2237 if (ifp->state != INET6_IFADDR_STATE_POSTDAD) {
2238 spin_unlock_bh(&ifp->lock);
2239 in6_ifa_put(ifp);
2240 return;
2241 }
2242 }
2243
2244 errdad:
2245 /* transition from _POSTDAD to _ERRDAD */
2246 ifp->state = INET6_IFADDR_STATE_ERRDAD;
2247 spin_unlock_bh(&ifp->lock);
2248
2249 addrconf_mod_dad_work(ifp, 0);
2250 in6_ifa_put(ifp);
2251 }
2252
2253 /* Join to solicited addr multicast group. */
addrconf_join_solict(struct net_device * dev,const struct in6_addr * addr)2254 void addrconf_join_solict(struct net_device *dev, const struct in6_addr *addr)
2255 {
2256 struct in6_addr maddr;
2257
2258 if (READ_ONCE(dev->flags) & (IFF_LOOPBACK | IFF_NOARP))
2259 return;
2260
2261 addrconf_addr_solict_mult(addr, &maddr);
2262 ipv6_dev_mc_inc(dev, &maddr);
2263 }
2264
addrconf_leave_solict(struct inet6_dev * idev,const struct in6_addr * addr)2265 void addrconf_leave_solict(struct inet6_dev *idev, const struct in6_addr *addr)
2266 {
2267 struct in6_addr maddr;
2268
2269 if (READ_ONCE(idev->dev->flags) & (IFF_LOOPBACK | IFF_NOARP))
2270 return;
2271
2272 addrconf_addr_solict_mult(addr, &maddr);
2273 __ipv6_dev_mc_dec(idev, &maddr);
2274 }
2275
addrconf_join_anycast(struct inet6_ifaddr * ifp)2276 static void addrconf_join_anycast(struct inet6_ifaddr *ifp)
2277 {
2278 struct in6_addr addr;
2279
2280 if (ifp->prefix_len >= 127) /* RFC 6164 */
2281 return;
2282 ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
2283 if (ipv6_addr_any(&addr))
2284 return;
2285 __ipv6_dev_ac_inc(ifp->idev, &addr);
2286 }
2287
addrconf_leave_anycast(struct inet6_ifaddr * ifp)2288 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp)
2289 {
2290 struct in6_addr addr;
2291
2292 if (ifp->prefix_len >= 127) /* RFC 6164 */
2293 return;
2294 ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
2295 if (ipv6_addr_any(&addr))
2296 return;
2297 __ipv6_dev_ac_dec(ifp->idev, &addr);
2298 }
2299
addrconf_ifid_6lowpan(u8 * eui,struct net_device * dev)2300 static int addrconf_ifid_6lowpan(u8 *eui, struct net_device *dev)
2301 {
2302 switch (dev->addr_len) {
2303 case ETH_ALEN:
2304 memcpy(eui, dev->dev_addr, 3);
2305 eui[3] = 0xFF;
2306 eui[4] = 0xFE;
2307 memcpy(eui + 5, dev->dev_addr + 3, 3);
2308 break;
2309 case EUI64_ADDR_LEN:
2310 memcpy(eui, dev->dev_addr, EUI64_ADDR_LEN);
2311 eui[0] ^= 2;
2312 break;
2313 default:
2314 return -1;
2315 }
2316
2317 return 0;
2318 }
2319
addrconf_ifid_ieee1394(u8 * eui,struct net_device * dev)2320 static int addrconf_ifid_ieee1394(u8 *eui, struct net_device *dev)
2321 {
2322 const union fwnet_hwaddr *ha;
2323
2324 if (dev->addr_len != FWNET_ALEN)
2325 return -1;
2326
2327 ha = (const union fwnet_hwaddr *)dev->dev_addr;
2328
2329 memcpy(eui, &ha->uc.uniq_id, sizeof(ha->uc.uniq_id));
2330 eui[0] ^= 2;
2331 return 0;
2332 }
2333
addrconf_ifid_arcnet(u8 * eui,struct net_device * dev)2334 static int addrconf_ifid_arcnet(u8 *eui, struct net_device *dev)
2335 {
2336 /* XXX: inherit EUI-64 from other interface -- yoshfuji */
2337 if (dev->addr_len != ARCNET_ALEN)
2338 return -1;
2339 memset(eui, 0, 7);
2340 eui[7] = *(u8 *)dev->dev_addr;
2341 return 0;
2342 }
2343
addrconf_ifid_infiniband(u8 * eui,struct net_device * dev)2344 static int addrconf_ifid_infiniband(u8 *eui, struct net_device *dev)
2345 {
2346 if (dev->addr_len != INFINIBAND_ALEN)
2347 return -1;
2348 memcpy(eui, dev->dev_addr + 12, 8);
2349 eui[0] |= 2;
2350 return 0;
2351 }
2352
__ipv6_isatap_ifid(u8 * eui,__be32 addr)2353 static int __ipv6_isatap_ifid(u8 *eui, __be32 addr)
2354 {
2355 if (addr == 0)
2356 return -1;
2357 eui[0] = (ipv4_is_zeronet(addr) || ipv4_is_private_10(addr) ||
2358 ipv4_is_loopback(addr) || ipv4_is_linklocal_169(addr) ||
2359 ipv4_is_private_172(addr) || ipv4_is_test_192(addr) ||
2360 ipv4_is_anycast_6to4(addr) || ipv4_is_private_192(addr) ||
2361 ipv4_is_test_198(addr) || ipv4_is_multicast(addr) ||
2362 ipv4_is_lbcast(addr)) ? 0x00 : 0x02;
2363 eui[1] = 0;
2364 eui[2] = 0x5E;
2365 eui[3] = 0xFE;
2366 memcpy(eui + 4, &addr, 4);
2367 return 0;
2368 }
2369
addrconf_ifid_sit(u8 * eui,struct net_device * dev)2370 static int addrconf_ifid_sit(u8 *eui, struct net_device *dev)
2371 {
2372 if (dev->priv_flags & IFF_ISATAP)
2373 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
2374 return -1;
2375 }
2376
addrconf_ifid_gre(u8 * eui,struct net_device * dev)2377 static int addrconf_ifid_gre(u8 *eui, struct net_device *dev)
2378 {
2379 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
2380 }
2381
addrconf_ifid_ip6tnl(u8 * eui,struct net_device * dev)2382 static int addrconf_ifid_ip6tnl(u8 *eui, struct net_device *dev)
2383 {
2384 memcpy(eui, dev->perm_addr, 3);
2385 memcpy(eui + 5, dev->perm_addr + 3, 3);
2386 eui[3] = 0xFF;
2387 eui[4] = 0xFE;
2388 eui[0] ^= 2;
2389 return 0;
2390 }
2391
ipv6_generate_eui64(u8 * eui,struct net_device * dev)2392 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev)
2393 {
2394 switch (dev->type) {
2395 case ARPHRD_ETHER:
2396 case ARPHRD_FDDI:
2397 return addrconf_ifid_eui48(eui, dev);
2398 case ARPHRD_ARCNET:
2399 return addrconf_ifid_arcnet(eui, dev);
2400 case ARPHRD_INFINIBAND:
2401 return addrconf_ifid_infiniband(eui, dev);
2402 case ARPHRD_SIT:
2403 return addrconf_ifid_sit(eui, dev);
2404 case ARPHRD_IPGRE:
2405 case ARPHRD_TUNNEL:
2406 return addrconf_ifid_gre(eui, dev);
2407 case ARPHRD_6LOWPAN:
2408 return addrconf_ifid_6lowpan(eui, dev);
2409 case ARPHRD_IEEE1394:
2410 return addrconf_ifid_ieee1394(eui, dev);
2411 case ARPHRD_TUNNEL6:
2412 case ARPHRD_IP6GRE:
2413 case ARPHRD_RAWIP:
2414 return addrconf_ifid_ip6tnl(eui, dev);
2415 }
2416 return -1;
2417 }
2418
ipv6_inherit_eui64(u8 * eui,struct inet6_dev * idev)2419 static int ipv6_inherit_eui64(u8 *eui, struct inet6_dev *idev)
2420 {
2421 int err = -1;
2422 struct inet6_ifaddr *ifp;
2423
2424 read_lock_bh(&idev->lock);
2425 list_for_each_entry_reverse(ifp, &idev->addr_list, if_list) {
2426 if (ifp->scope > IFA_LINK)
2427 break;
2428 if (ifp->scope == IFA_LINK && !(ifp->flags&IFA_F_TENTATIVE)) {
2429 memcpy(eui, ifp->addr.s6_addr+8, 8);
2430 err = 0;
2431 break;
2432 }
2433 }
2434 read_unlock_bh(&idev->lock);
2435 return err;
2436 }
2437
2438 /* Generation of a randomized Interface Identifier
2439 * draft-ietf-6man-rfc4941bis, Section 3.3.1
2440 */
2441
ipv6_gen_rnd_iid(struct in6_addr * addr)2442 static void ipv6_gen_rnd_iid(struct in6_addr *addr)
2443 {
2444 regen:
2445 get_random_bytes(&addr->s6_addr[8], 8);
2446
2447 /* <draft-ietf-6man-rfc4941bis-08.txt>, Section 3.3.1:
2448 * check if generated address is not inappropriate:
2449 *
2450 * - Reserved IPv6 Interface Identifiers
2451 * - XXX: already assigned to an address on the device
2452 */
2453
2454 /* Subnet-router anycast: 0000:0000:0000:0000 */
2455 if (!(addr->s6_addr32[2] | addr->s6_addr32[3]))
2456 goto regen;
2457
2458 /* IANA Ethernet block: 0200:5EFF:FE00:0000-0200:5EFF:FE00:5212
2459 * Proxy Mobile IPv6: 0200:5EFF:FE00:5213
2460 * IANA Ethernet block: 0200:5EFF:FE00:5214-0200:5EFF:FEFF:FFFF
2461 */
2462 if (ntohl(addr->s6_addr32[2]) == 0x02005eff &&
2463 (ntohl(addr->s6_addr32[3]) & 0Xff000000) == 0xfe000000)
2464 goto regen;
2465
2466 /* Reserved subnet anycast addresses */
2467 if (ntohl(addr->s6_addr32[2]) == 0xfdffffff &&
2468 ntohl(addr->s6_addr32[3]) >= 0Xffffff80)
2469 goto regen;
2470 }
2471
2472 /*
2473 * Add prefix route.
2474 */
2475
2476 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)2477 addrconf_prefix_route(struct in6_addr *pfx, int plen, u32 metric,
2478 struct net_device *dev, unsigned long expires,
2479 u32 flags, gfp_t gfp_flags)
2480 {
2481 struct fib6_config cfg = {
2482 .fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_PREFIX,
2483 .fc_metric = metric ? : IP6_RT_PRIO_ADDRCONF,
2484 .fc_ifindex = dev->ifindex,
2485 .fc_expires = expires,
2486 .fc_dst_len = plen,
2487 .fc_flags = RTF_UP | flags,
2488 .fc_nlinfo.nl_net = dev_net(dev),
2489 .fc_protocol = RTPROT_KERNEL,
2490 .fc_type = RTN_UNICAST,
2491 };
2492
2493 cfg.fc_dst = *pfx;
2494
2495 /* Prevent useless cloning on PtP SIT.
2496 This thing is done here expecting that the whole
2497 class of non-broadcast devices need not cloning.
2498 */
2499 #if IS_ENABLED(CONFIG_IPV6_SIT)
2500 if (dev->type == ARPHRD_SIT && (dev->flags & IFF_POINTOPOINT))
2501 cfg.fc_flags |= RTF_NONEXTHOP;
2502 #endif
2503
2504 ip6_route_add(&cfg, gfp_flags, NULL);
2505 }
2506
2507
addrconf_get_prefix_route(const struct in6_addr * pfx,int plen,const struct net_device * dev,u32 flags,u32 noflags,bool no_gw)2508 static struct fib6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
2509 int plen,
2510 const struct net_device *dev,
2511 u32 flags, u32 noflags,
2512 bool no_gw)
2513 {
2514 struct fib6_node *fn;
2515 struct fib6_info *rt = NULL;
2516 struct fib6_table *table;
2517 u32 tb_id = l3mdev_fib_table(dev) ? : RT6_TABLE_PREFIX;
2518
2519 table = fib6_get_table(dev_net(dev), tb_id);
2520 if (!table)
2521 return NULL;
2522
2523 rcu_read_lock();
2524 fn = fib6_locate(&table->tb6_root, pfx, plen, NULL, 0, true);
2525 if (!fn)
2526 goto out;
2527
2528 for_each_fib6_node_rt_rcu(fn) {
2529 /* prefix routes only use builtin fib6_nh */
2530 if (rt->nh)
2531 continue;
2532
2533 if (rt->fib6_nh->fib_nh_dev->ifindex != dev->ifindex)
2534 continue;
2535 if (no_gw && rt->fib6_nh->fib_nh_gw_family)
2536 continue;
2537 if ((rt->fib6_flags & flags) != flags)
2538 continue;
2539 if ((rt->fib6_flags & noflags) != 0)
2540 continue;
2541 if (!fib6_info_hold_safe(rt))
2542 continue;
2543 break;
2544 }
2545 out:
2546 rcu_read_unlock();
2547 return rt;
2548 }
2549
2550
2551 /* Create "default" multicast route to the interface */
2552
addrconf_add_mroute(struct net_device * dev)2553 static void addrconf_add_mroute(struct net_device *dev)
2554 {
2555 struct fib6_config cfg = {
2556 .fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_LOCAL,
2557 .fc_metric = IP6_RT_PRIO_ADDRCONF,
2558 .fc_ifindex = dev->ifindex,
2559 .fc_dst_len = 8,
2560 .fc_flags = RTF_UP,
2561 .fc_type = RTN_MULTICAST,
2562 .fc_nlinfo.nl_net = dev_net(dev),
2563 .fc_protocol = RTPROT_KERNEL,
2564 };
2565
2566 ipv6_addr_set(&cfg.fc_dst, htonl(0xFF000000), 0, 0, 0);
2567
2568 ip6_route_add(&cfg, GFP_KERNEL, NULL);
2569 }
2570
addrconf_add_dev(struct net_device * dev)2571 static struct inet6_dev *addrconf_add_dev(struct net_device *dev)
2572 {
2573 struct inet6_dev *idev;
2574
2575 ASSERT_RTNL();
2576
2577 idev = ipv6_find_idev(dev);
2578 if (IS_ERR(idev))
2579 return idev;
2580
2581 if (idev->cnf.disable_ipv6)
2582 return ERR_PTR(-EACCES);
2583
2584 /* Add default multicast route */
2585 if (!(dev->flags & IFF_LOOPBACK) && !netif_is_l3_master(dev))
2586 addrconf_add_mroute(dev);
2587
2588 return idev;
2589 }
2590
delete_tempaddrs(struct inet6_dev * idev,struct inet6_ifaddr * ifp)2591 static void delete_tempaddrs(struct inet6_dev *idev,
2592 struct inet6_ifaddr *ifp)
2593 {
2594 struct inet6_ifaddr *ift, *tmp;
2595
2596 write_lock_bh(&idev->lock);
2597 list_for_each_entry_safe(ift, tmp, &idev->tempaddr_list, tmp_list) {
2598 if (ift->ifpub != ifp)
2599 continue;
2600
2601 in6_ifa_hold(ift);
2602 write_unlock_bh(&idev->lock);
2603 ipv6_del_addr(ift);
2604 write_lock_bh(&idev->lock);
2605 }
2606 write_unlock_bh(&idev->lock);
2607 }
2608
manage_tempaddrs(struct inet6_dev * idev,struct inet6_ifaddr * ifp,__u32 valid_lft,__u32 prefered_lft,bool create,unsigned long now)2609 static void manage_tempaddrs(struct inet6_dev *idev,
2610 struct inet6_ifaddr *ifp,
2611 __u32 valid_lft, __u32 prefered_lft,
2612 bool create, unsigned long now)
2613 {
2614 u32 orig_prefered_lft = prefered_lft;
2615 struct inet6_ifaddr *ift;
2616 bool reset_done = false;
2617 u32 flags;
2618
2619 read_lock_bh(&idev->lock);
2620 /* update all temporary addresses in the list */
2621 list_for_each_entry(ift, &idev->tempaddr_list, tmp_list) {
2622 int age, max_valid, max_prefered;
2623
2624 if (ifp != ift->ifpub)
2625 continue;
2626
2627 /* RFC 4941 section 3.3:
2628 * If a received option will extend the lifetime of a public
2629 * address, the lifetimes of temporary addresses should
2630 * be extended, subject to the overall constraint that no
2631 * temporary addresses should ever remain "valid" or "preferred"
2632 * for a time longer than (TEMP_VALID_LIFETIME) or
2633 * (TEMP_PREFERRED_LIFETIME - DESYNC_FACTOR), respectively.
2634 */
2635 age = (now - ift->cstamp) / HZ;
2636 max_valid = READ_ONCE(idev->cnf.temp_valid_lft) - age;
2637 if (max_valid < 0)
2638 max_valid = 0;
2639
2640 max_prefered = READ_ONCE(idev->cnf.temp_prefered_lft) -
2641 idev->desync_factor - age;
2642 if (max_prefered < 0)
2643 max_prefered = 0;
2644
2645 if (valid_lft > max_valid)
2646 valid_lft = max_valid;
2647
2648 if (prefered_lft > max_prefered)
2649 prefered_lft = max_prefered;
2650
2651 spin_lock(&ift->lock);
2652 /* the first match is the most recent temp address */
2653 if (!reset_done && orig_prefered_lft > 0) {
2654 ift->regen_count = 0;
2655 reset_done = true;
2656 }
2657 flags = ift->flags;
2658 ift->valid_lft = valid_lft;
2659 ift->prefered_lft = prefered_lft;
2660 ift->tstamp = now;
2661 if (prefered_lft > 0)
2662 ift->flags &= ~IFA_F_DEPRECATED;
2663
2664 spin_unlock(&ift->lock);
2665 if (!(flags&IFA_F_TENTATIVE))
2666 ipv6_ifa_notify(0, ift);
2667 }
2668
2669 /* Also create a temporary address if it's enabled but no temporary
2670 * address currently exists.
2671 * However, we get called with valid_lft == 0, prefered_lft == 0, create == false
2672 * as part of cleanup (ie. deleting the mngtmpaddr).
2673 * We don't want that to result in creating a new temporary ip address.
2674 */
2675 if (list_empty(&idev->tempaddr_list) && (valid_lft || prefered_lft))
2676 create = true;
2677
2678 if (create && READ_ONCE(idev->cnf.use_tempaddr) > 0) {
2679 /* When a new public address is created as described
2680 * in [ADDRCONF], also create a new temporary address.
2681 */
2682 read_unlock_bh(&idev->lock);
2683 ipv6_create_tempaddr(ifp, false);
2684 } else {
2685 read_unlock_bh(&idev->lock);
2686 }
2687 }
2688
is_addr_mode_generate_stable(struct inet6_dev * idev)2689 static bool is_addr_mode_generate_stable(struct inet6_dev *idev)
2690 {
2691 return idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_STABLE_PRIVACY ||
2692 idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_RANDOM;
2693 }
2694
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)2695 int addrconf_prefix_rcv_add_addr(struct net *net, struct net_device *dev,
2696 const struct prefix_info *pinfo,
2697 struct inet6_dev *in6_dev,
2698 const struct in6_addr *addr, int addr_type,
2699 u32 addr_flags, bool sllao, bool tokenized,
2700 __u32 valid_lft, u32 prefered_lft)
2701 {
2702 struct inet6_ifaddr *ifp = ipv6_get_ifaddr(net, addr, dev, 1);
2703 int create = 0, update_lft = 0;
2704
2705 if (!ifp && valid_lft) {
2706 int max_addresses = READ_ONCE(in6_dev->cnf.max_addresses);
2707 struct ifa6_config cfg = {
2708 .pfx = addr,
2709 .plen = pinfo->prefix_len,
2710 .ifa_flags = addr_flags,
2711 .valid_lft = valid_lft,
2712 .preferred_lft = prefered_lft,
2713 .scope = addr_type & IPV6_ADDR_SCOPE_MASK,
2714 .ifa_proto = IFAPROT_KERNEL_RA
2715 };
2716
2717 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2718 if ((READ_ONCE(net->ipv6.devconf_all->optimistic_dad) ||
2719 READ_ONCE(in6_dev->cnf.optimistic_dad)) &&
2720 !net->ipv6.devconf_all->forwarding && sllao)
2721 cfg.ifa_flags |= IFA_F_OPTIMISTIC;
2722 #endif
2723
2724 /* Do not allow to create too much of autoconfigured
2725 * addresses; this would be too easy way to crash kernel.
2726 */
2727 if (!max_addresses ||
2728 ipv6_count_addresses(in6_dev) < max_addresses)
2729 ifp = ipv6_add_addr(in6_dev, &cfg, false, NULL);
2730
2731 if (IS_ERR_OR_NULL(ifp))
2732 return -1;
2733
2734 create = 1;
2735 spin_lock_bh(&ifp->lock);
2736 ifp->flags |= IFA_F_MANAGETEMPADDR;
2737 ifp->cstamp = jiffies;
2738 ifp->tokenized = tokenized;
2739 spin_unlock_bh(&ifp->lock);
2740 addrconf_dad_start(ifp);
2741 }
2742
2743 if (ifp) {
2744 u32 flags;
2745 unsigned long now;
2746 u32 stored_lft;
2747
2748 /* update lifetime (RFC2462 5.5.3 e) */
2749 spin_lock_bh(&ifp->lock);
2750 now = jiffies;
2751 if (ifp->valid_lft > (now - ifp->tstamp) / HZ)
2752 stored_lft = ifp->valid_lft - (now - ifp->tstamp) / HZ;
2753 else
2754 stored_lft = 0;
2755
2756 /* RFC4862 Section 5.5.3e:
2757 * "Note that the preferred lifetime of the
2758 * corresponding address is always reset to
2759 * the Preferred Lifetime in the received
2760 * Prefix Information option, regardless of
2761 * whether the valid lifetime is also reset or
2762 * ignored."
2763 *
2764 * So we should always update prefered_lft here.
2765 */
2766 update_lft = !create && stored_lft;
2767
2768 if (update_lft && !READ_ONCE(in6_dev->cnf.ra_honor_pio_life)) {
2769 const u32 minimum_lft = min_t(u32,
2770 stored_lft, MIN_VALID_LIFETIME);
2771 valid_lft = max(valid_lft, minimum_lft);
2772 }
2773
2774 if (update_lft) {
2775 ifp->valid_lft = valid_lft;
2776 ifp->prefered_lft = prefered_lft;
2777 WRITE_ONCE(ifp->tstamp, now);
2778 flags = ifp->flags;
2779 ifp->flags &= ~IFA_F_DEPRECATED;
2780 spin_unlock_bh(&ifp->lock);
2781
2782 if (!(flags&IFA_F_TENTATIVE))
2783 ipv6_ifa_notify(0, ifp);
2784 } else
2785 spin_unlock_bh(&ifp->lock);
2786
2787 manage_tempaddrs(in6_dev, ifp, valid_lft, prefered_lft,
2788 create, now);
2789
2790 in6_ifa_put(ifp);
2791 addrconf_verify(net);
2792 }
2793
2794 return 0;
2795 }
2796 EXPORT_SYMBOL_GPL(addrconf_prefix_rcv_add_addr);
2797
addrconf_prefix_rcv(struct net_device * dev,u8 * opt,int len,bool sllao)2798 void addrconf_prefix_rcv(struct net_device *dev, u8 *opt, int len, bool sllao)
2799 {
2800 struct prefix_info *pinfo;
2801 struct fib6_table *table;
2802 __u32 valid_lft;
2803 __u32 prefered_lft;
2804 int addr_type, err;
2805 u32 addr_flags = 0;
2806 struct inet6_dev *in6_dev;
2807 struct net *net = dev_net(dev);
2808 bool ignore_autoconf = false;
2809
2810 pinfo = (struct prefix_info *) opt;
2811
2812 if (len < sizeof(struct prefix_info)) {
2813 netdev_dbg(dev, "addrconf: prefix option too short\n");
2814 return;
2815 }
2816
2817 /*
2818 * Validation checks ([ADDRCONF], page 19)
2819 */
2820
2821 addr_type = ipv6_addr_type(&pinfo->prefix);
2822
2823 if (addr_type & (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL))
2824 return;
2825
2826 valid_lft = ntohl(pinfo->valid);
2827 prefered_lft = ntohl(pinfo->prefered);
2828
2829 if (prefered_lft > valid_lft) {
2830 net_warn_ratelimited("addrconf: prefix option has invalid lifetime\n");
2831 return;
2832 }
2833
2834 in6_dev = in6_dev_get(dev);
2835
2836 if (!in6_dev) {
2837 net_dbg_ratelimited("addrconf: device %s not configured\n",
2838 dev->name);
2839 return;
2840 }
2841
2842 if (valid_lft != 0 && valid_lft < in6_dev->cnf.accept_ra_min_lft)
2843 goto put;
2844
2845 /*
2846 * Two things going on here:
2847 * 1) Add routes for on-link prefixes
2848 * 2) Configure prefixes with the auto flag set
2849 */
2850
2851 if (pinfo->onlink) {
2852 struct fib6_info *rt;
2853 unsigned long rt_expires;
2854
2855 /* Avoid arithmetic overflow. Really, we could
2856 * save rt_expires in seconds, likely valid_lft,
2857 * but it would require division in fib gc, that it
2858 * not good.
2859 */
2860 if (HZ > USER_HZ)
2861 rt_expires = addrconf_timeout_fixup(valid_lft, HZ);
2862 else
2863 rt_expires = addrconf_timeout_fixup(valid_lft, USER_HZ);
2864
2865 if (addrconf_finite_timeout(rt_expires))
2866 rt_expires *= HZ;
2867
2868 rt = addrconf_get_prefix_route(&pinfo->prefix,
2869 pinfo->prefix_len,
2870 dev,
2871 RTF_ADDRCONF | RTF_PREFIX_RT,
2872 RTF_DEFAULT, true);
2873
2874 if (rt) {
2875 /* Autoconf prefix route */
2876 if (valid_lft == 0) {
2877 ip6_del_rt_reason(net, rt,
2878 RT_DEL_REASON_RA_WITHDRAWN);
2879 rt = NULL;
2880 } else {
2881 table = rt->fib6_table;
2882 spin_lock_bh(&table->tb6_lock);
2883
2884 if (addrconf_finite_timeout(rt_expires)) {
2885 /* not infinity */
2886 fib6_set_expires(rt, jiffies + rt_expires);
2887 fib6_add_gc_list(rt);
2888 } else {
2889 fib6_clean_expires(rt);
2890 fib6_may_remove_gc_list(net, rt);
2891 }
2892
2893 spin_unlock_bh(&table->tb6_lock);
2894 }
2895 } else if (valid_lft) {
2896 clock_t expires = 0;
2897 int flags = RTF_ADDRCONF | RTF_PREFIX_RT;
2898 if (addrconf_finite_timeout(rt_expires)) {
2899 /* not infinity */
2900 flags |= RTF_EXPIRES;
2901 expires = jiffies_to_clock_t(rt_expires);
2902 }
2903 addrconf_prefix_route(&pinfo->prefix, pinfo->prefix_len,
2904 0, dev, expires, flags,
2905 GFP_ATOMIC);
2906 }
2907 fib6_info_release(rt);
2908 }
2909
2910 /* Try to figure out our local address for this prefix */
2911
2912 ignore_autoconf = READ_ONCE(in6_dev->cnf.ra_honor_pio_pflag) && pinfo->preferpd;
2913 if (pinfo->autoconf && in6_dev->cnf.autoconf && !ignore_autoconf) {
2914 struct in6_addr addr;
2915 bool tokenized = false, dev_addr_generated = false;
2916
2917 if (pinfo->prefix_len == 64) {
2918 memcpy(&addr, &pinfo->prefix, 8);
2919
2920 if (!ipv6_addr_any(&in6_dev->token)) {
2921 read_lock_bh(&in6_dev->lock);
2922 memcpy(addr.s6_addr + 8,
2923 in6_dev->token.s6_addr + 8, 8);
2924 read_unlock_bh(&in6_dev->lock);
2925 tokenized = true;
2926 } else if (is_addr_mode_generate_stable(in6_dev) &&
2927 !ipv6_generate_stable_address(&addr, 0,
2928 in6_dev)) {
2929 addr_flags |= IFA_F_STABLE_PRIVACY;
2930 goto ok;
2931 } else if (ipv6_generate_eui64(addr.s6_addr + 8, dev) &&
2932 ipv6_inherit_eui64(addr.s6_addr + 8, in6_dev)) {
2933 goto put;
2934 } else {
2935 dev_addr_generated = true;
2936 }
2937 goto ok;
2938 }
2939 net_dbg_ratelimited("IPv6 addrconf: prefix with wrong length %d\n",
2940 pinfo->prefix_len);
2941 goto put;
2942
2943 ok:
2944 err = addrconf_prefix_rcv_add_addr(net, dev, pinfo, in6_dev,
2945 &addr, addr_type,
2946 addr_flags, sllao,
2947 tokenized, valid_lft,
2948 prefered_lft);
2949 if (err)
2950 goto put;
2951
2952 /* Ignore error case here because previous prefix add addr was
2953 * successful which will be notified.
2954 */
2955 ndisc_ops_prefix_rcv_add_addr(net, dev, pinfo, in6_dev, &addr,
2956 addr_type, addr_flags, sllao,
2957 tokenized, valid_lft,
2958 prefered_lft,
2959 dev_addr_generated);
2960 }
2961 inet6_prefix_notify(RTM_NEWPREFIX, in6_dev, pinfo);
2962 put:
2963 in6_dev_put(in6_dev);
2964 }
2965
addrconf_set_sit_dstaddr(struct net * net,struct net_device * dev,struct in6_ifreq * ireq)2966 static int addrconf_set_sit_dstaddr(struct net *net, struct net_device *dev,
2967 struct in6_ifreq *ireq)
2968 {
2969 struct ip_tunnel_parm_kern p = { };
2970 int err;
2971
2972 if (!(ipv6_addr_type(&ireq->ifr6_addr) & IPV6_ADDR_COMPATv4))
2973 return -EADDRNOTAVAIL;
2974
2975 p.iph.daddr = ireq->ifr6_addr.s6_addr32[3];
2976 p.iph.version = 4;
2977 p.iph.ihl = 5;
2978 p.iph.protocol = IPPROTO_IPV6;
2979 p.iph.ttl = 64;
2980
2981 if (!dev->netdev_ops->ndo_tunnel_ctl)
2982 return -EOPNOTSUPP;
2983 err = dev->netdev_ops->ndo_tunnel_ctl(dev, &p, SIOCADDTUNNEL);
2984 if (err)
2985 return err;
2986
2987 dev = __dev_get_by_name(net, p.name);
2988 if (!dev)
2989 return -ENOBUFS;
2990 return dev_open(dev, NULL);
2991 }
2992
2993 /*
2994 * Set destination address.
2995 * Special case for SIT interfaces where we create a new "virtual"
2996 * device.
2997 */
addrconf_set_dstaddr(struct net * net,void __user * arg)2998 int addrconf_set_dstaddr(struct net *net, void __user *arg)
2999 {
3000 struct net_device *dev;
3001 struct in6_ifreq ireq;
3002 int err = -ENODEV;
3003
3004 if (!IS_ENABLED(CONFIG_IPV6_SIT))
3005 return -ENODEV;
3006 if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
3007 return -EFAULT;
3008
3009 rtnl_net_lock(net);
3010 dev = __dev_get_by_index(net, ireq.ifr6_ifindex);
3011 if (dev && dev->type == ARPHRD_SIT)
3012 err = addrconf_set_sit_dstaddr(net, dev, &ireq);
3013 rtnl_net_unlock(net);
3014 return err;
3015 }
3016
ipv6_mc_config(struct sock * sk,bool join,const struct in6_addr * addr,int ifindex)3017 static int ipv6_mc_config(struct sock *sk, bool join,
3018 const struct in6_addr *addr, int ifindex)
3019 {
3020 int ret;
3021
3022 ASSERT_RTNL();
3023
3024 lock_sock(sk);
3025 if (join)
3026 ret = ipv6_sock_mc_join(sk, ifindex, addr);
3027 else
3028 ret = ipv6_sock_mc_drop(sk, ifindex, addr);
3029 release_sock(sk);
3030
3031 return ret;
3032 }
3033
3034 /*
3035 * Manual configuration of address on an interface
3036 */
inet6_addr_add(struct net * net,struct net_device * dev,struct ifa6_config * cfg,clock_t expires,u32 flags,struct netlink_ext_ack * extack)3037 static int inet6_addr_add(struct net *net, struct net_device *dev,
3038 struct ifa6_config *cfg, clock_t expires, u32 flags,
3039 struct netlink_ext_ack *extack)
3040 {
3041 struct inet6_ifaddr *ifp;
3042 struct inet6_dev *idev;
3043
3044 ASSERT_RTNL_NET(net);
3045
3046 if (cfg->plen > 128) {
3047 NL_SET_ERR_MSG_MOD(extack, "Invalid prefix length");
3048 return -EINVAL;
3049 }
3050
3051 if (cfg->ifa_flags & IFA_F_MANAGETEMPADDR && cfg->plen != 64) {
3052 NL_SET_ERR_MSG_MOD(extack, "address with \"mngtmpaddr\" flag must have a prefix length of 64");
3053 return -EINVAL;
3054 }
3055
3056 idev = addrconf_add_dev(dev);
3057 if (IS_ERR(idev)) {
3058 NL_SET_ERR_MSG_MOD(extack, "IPv6 is disabled on this device");
3059 return PTR_ERR(idev);
3060 }
3061
3062 if (cfg->ifa_flags & IFA_F_MCAUTOJOIN) {
3063 int ret = ipv6_mc_config(net->ipv6.mc_autojoin_sk,
3064 true, cfg->pfx, dev->ifindex);
3065
3066 if (ret < 0) {
3067 NL_SET_ERR_MSG_MOD(extack, "Multicast auto join failed");
3068 return ret;
3069 }
3070 }
3071
3072 cfg->scope = ipv6_addr_scope(cfg->pfx);
3073
3074 ifp = ipv6_add_addr(idev, cfg, true, extack);
3075 if (!IS_ERR(ifp)) {
3076 if (!(cfg->ifa_flags & IFA_F_NOPREFIXROUTE)) {
3077 addrconf_prefix_route(&ifp->addr, ifp->prefix_len,
3078 ifp->rt_priority, dev, expires,
3079 flags, GFP_KERNEL);
3080 }
3081
3082 /* Send a netlink notification if DAD is enabled and
3083 * optimistic flag is not set
3084 */
3085 if (!(ifp->flags & (IFA_F_OPTIMISTIC | IFA_F_NODAD)))
3086 ipv6_ifa_notify(0, ifp);
3087 /*
3088 * Note that section 3.1 of RFC 4429 indicates
3089 * that the Optimistic flag should not be set for
3090 * manually configured addresses
3091 */
3092 addrconf_dad_start(ifp);
3093 if (cfg->ifa_flags & IFA_F_MANAGETEMPADDR)
3094 manage_tempaddrs(idev, ifp, cfg->valid_lft,
3095 cfg->preferred_lft, true, jiffies);
3096 in6_ifa_put(ifp);
3097 addrconf_verify_rtnl(net);
3098 return 0;
3099 } else if (cfg->ifa_flags & IFA_F_MCAUTOJOIN) {
3100 ipv6_mc_config(net->ipv6.mc_autojoin_sk, false,
3101 cfg->pfx, dev->ifindex);
3102 }
3103
3104 return PTR_ERR(ifp);
3105 }
3106
inet6_addr_del(struct net * net,int ifindex,u32 ifa_flags,const struct in6_addr * pfx,unsigned int plen,struct netlink_ext_ack * extack)3107 static int inet6_addr_del(struct net *net, int ifindex, u32 ifa_flags,
3108 const struct in6_addr *pfx, unsigned int plen,
3109 struct netlink_ext_ack *extack)
3110 {
3111 struct inet6_ifaddr *ifp;
3112 struct inet6_dev *idev;
3113 struct net_device *dev;
3114
3115 if (plen > 128) {
3116 NL_SET_ERR_MSG_MOD(extack, "Invalid prefix length");
3117 return -EINVAL;
3118 }
3119
3120 dev = __dev_get_by_index(net, ifindex);
3121 if (!dev) {
3122 NL_SET_ERR_MSG_MOD(extack, "Unable to find the interface");
3123 return -ENODEV;
3124 }
3125
3126 idev = __in6_dev_get_rtnl_net(dev);
3127 if (!idev) {
3128 NL_SET_ERR_MSG_MOD(extack, "IPv6 is disabled on this device");
3129 return -ENXIO;
3130 }
3131
3132 read_lock_bh(&idev->lock);
3133 list_for_each_entry(ifp, &idev->addr_list, if_list) {
3134 if (ifp->prefix_len == plen &&
3135 ipv6_addr_equal(pfx, &ifp->addr)) {
3136 in6_ifa_hold(ifp);
3137 read_unlock_bh(&idev->lock);
3138
3139 if (!(ifp->flags & IFA_F_TEMPORARY) &&
3140 (ifp->flags & IFA_F_MANAGETEMPADDR))
3141 delete_tempaddrs(idev, ifp);
3142
3143 ipv6_del_addr(ifp);
3144
3145 addrconf_verify_rtnl(net);
3146 if (ipv6_addr_is_multicast(pfx)) {
3147 ipv6_mc_config(net->ipv6.mc_autojoin_sk,
3148 false, pfx, dev->ifindex);
3149 }
3150 return 0;
3151 }
3152 }
3153 read_unlock_bh(&idev->lock);
3154
3155 NL_SET_ERR_MSG_MOD(extack, "address not found");
3156 return -EADDRNOTAVAIL;
3157 }
3158
3159
addrconf_add_ifaddr(struct net * net,void __user * arg)3160 int addrconf_add_ifaddr(struct net *net, void __user *arg)
3161 {
3162 struct ifa6_config cfg = {
3163 .ifa_flags = IFA_F_PERMANENT,
3164 .preferred_lft = INFINITY_LIFE_TIME,
3165 .valid_lft = INFINITY_LIFE_TIME,
3166 };
3167 struct net_device *dev;
3168 struct in6_ifreq ireq;
3169 int err;
3170
3171 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
3172 return -EPERM;
3173
3174 if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
3175 return -EFAULT;
3176
3177 cfg.pfx = &ireq.ifr6_addr;
3178 cfg.plen = ireq.ifr6_prefixlen;
3179
3180 rtnl_net_lock(net);
3181 dev = __dev_get_by_index(net, ireq.ifr6_ifindex);
3182 if (dev) {
3183 netdev_lock_ops(dev);
3184 err = inet6_addr_add(net, dev, &cfg, 0, 0, NULL);
3185 netdev_unlock_ops(dev);
3186 } else {
3187 err = -ENODEV;
3188 }
3189 rtnl_net_unlock(net);
3190 return err;
3191 }
3192
addrconf_del_ifaddr(struct net * net,void __user * arg)3193 int addrconf_del_ifaddr(struct net *net, void __user *arg)
3194 {
3195 struct in6_ifreq ireq;
3196 int err;
3197
3198 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
3199 return -EPERM;
3200
3201 if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
3202 return -EFAULT;
3203
3204 rtnl_net_lock(net);
3205 err = inet6_addr_del(net, ireq.ifr6_ifindex, 0, &ireq.ifr6_addr,
3206 ireq.ifr6_prefixlen, NULL);
3207 rtnl_net_unlock(net);
3208 return err;
3209 }
3210
add_addr(struct inet6_dev * idev,const struct in6_addr * addr,int plen,int scope,u8 proto)3211 static void add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
3212 int plen, int scope, u8 proto)
3213 {
3214 struct inet6_ifaddr *ifp;
3215 struct ifa6_config cfg = {
3216 .pfx = addr,
3217 .plen = plen,
3218 .ifa_flags = IFA_F_PERMANENT,
3219 .valid_lft = INFINITY_LIFE_TIME,
3220 .preferred_lft = INFINITY_LIFE_TIME,
3221 .scope = scope,
3222 .ifa_proto = proto
3223 };
3224
3225 ifp = ipv6_add_addr(idev, &cfg, true, NULL);
3226 if (!IS_ERR(ifp)) {
3227 spin_lock_bh(&ifp->lock);
3228 ifp->flags &= ~IFA_F_TENTATIVE;
3229 spin_unlock_bh(&ifp->lock);
3230 rt_genid_bump_ipv6(dev_net(idev->dev));
3231 ipv6_ifa_notify(RTM_NEWADDR, ifp);
3232 in6_ifa_put(ifp);
3233 }
3234 }
3235
3236 #if IS_ENABLED(CONFIG_IPV6_SIT) || IS_ENABLED(CONFIG_NET_IPGRE)
add_v4_addrs(struct inet6_dev * idev)3237 static void add_v4_addrs(struct inet6_dev *idev)
3238 {
3239 struct in6_addr addr;
3240 struct net_device *dev;
3241 struct net *net = dev_net(idev->dev);
3242 int scope, plen;
3243 u32 pflags = 0;
3244
3245 ASSERT_RTNL();
3246
3247 memset(&addr, 0, sizeof(struct in6_addr));
3248 memcpy(&addr.s6_addr32[3], idev->dev->dev_addr, 4);
3249
3250 if (!(idev->dev->flags & IFF_POINTOPOINT) && idev->dev->type == ARPHRD_SIT) {
3251 scope = IPV6_ADDR_COMPATv4;
3252 plen = 96;
3253 pflags |= RTF_NONEXTHOP;
3254 } else {
3255 if (idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_NONE)
3256 return;
3257
3258 addr.s6_addr32[0] = htonl(0xfe800000);
3259 scope = IFA_LINK;
3260 plen = 64;
3261 }
3262
3263 if (addr.s6_addr32[3]) {
3264 add_addr(idev, &addr, plen, scope, IFAPROT_UNSPEC);
3265 addrconf_prefix_route(&addr, plen, 0, idev->dev, 0, pflags,
3266 GFP_KERNEL);
3267 return;
3268 }
3269
3270 for_each_netdev(net, dev) {
3271 struct in_device *in_dev = __in_dev_get_rtnl(dev);
3272 if (in_dev && (dev->flags & IFF_UP)) {
3273 struct in_ifaddr *ifa;
3274 int flag = scope;
3275
3276 in_dev_for_each_ifa_rtnl(ifa, in_dev) {
3277 addr.s6_addr32[3] = ifa->ifa_local;
3278
3279 if (ifa->ifa_scope == RT_SCOPE_LINK)
3280 continue;
3281 if (ifa->ifa_scope >= RT_SCOPE_HOST) {
3282 if (idev->dev->flags&IFF_POINTOPOINT)
3283 continue;
3284 flag |= IFA_HOST;
3285 }
3286
3287 add_addr(idev, &addr, plen, flag,
3288 IFAPROT_UNSPEC);
3289 addrconf_prefix_route(&addr, plen, 0, idev->dev,
3290 0, pflags, GFP_KERNEL);
3291 }
3292 }
3293 }
3294 }
3295 #endif
3296
init_loopback(struct net_device * dev)3297 static void init_loopback(struct net_device *dev)
3298 {
3299 struct inet6_dev *idev;
3300
3301 /* ::1 */
3302
3303 ASSERT_RTNL();
3304
3305 idev = ipv6_find_idev(dev);
3306 if (IS_ERR(idev)) {
3307 pr_debug("%s: add_dev failed\n", __func__);
3308 return;
3309 }
3310
3311 add_addr(idev, &in6addr_loopback, 128, IFA_HOST, IFAPROT_KERNEL_LO);
3312 }
3313
addrconf_add_linklocal(struct inet6_dev * idev,const struct in6_addr * addr,u32 flags)3314 void addrconf_add_linklocal(struct inet6_dev *idev,
3315 const struct in6_addr *addr, u32 flags)
3316 {
3317 struct ifa6_config cfg = {
3318 .pfx = addr,
3319 .plen = 64,
3320 .ifa_flags = flags | IFA_F_PERMANENT,
3321 .valid_lft = INFINITY_LIFE_TIME,
3322 .preferred_lft = INFINITY_LIFE_TIME,
3323 .scope = IFA_LINK,
3324 .ifa_proto = IFAPROT_KERNEL_LL
3325 };
3326 struct inet6_ifaddr *ifp;
3327
3328 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
3329 if ((READ_ONCE(dev_net(idev->dev)->ipv6.devconf_all->optimistic_dad) ||
3330 READ_ONCE(idev->cnf.optimistic_dad)) &&
3331 !dev_net(idev->dev)->ipv6.devconf_all->forwarding)
3332 cfg.ifa_flags |= IFA_F_OPTIMISTIC;
3333 #endif
3334
3335 ifp = ipv6_add_addr(idev, &cfg, true, NULL);
3336 if (!IS_ERR(ifp)) {
3337 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, 0, idev->dev,
3338 0, 0, GFP_ATOMIC);
3339 addrconf_dad_start(ifp);
3340 in6_ifa_put(ifp);
3341 }
3342 }
3343 EXPORT_SYMBOL_GPL(addrconf_add_linklocal);
3344
ipv6_reserved_interfaceid(struct in6_addr address)3345 static bool ipv6_reserved_interfaceid(struct in6_addr address)
3346 {
3347 if ((address.s6_addr32[2] | address.s6_addr32[3]) == 0)
3348 return true;
3349
3350 if (address.s6_addr32[2] == htonl(0x02005eff) &&
3351 ((address.s6_addr32[3] & htonl(0xfe000000)) == htonl(0xfe000000)))
3352 return true;
3353
3354 if (address.s6_addr32[2] == htonl(0xfdffffff) &&
3355 ((address.s6_addr32[3] & htonl(0xffffff80)) == htonl(0xffffff80)))
3356 return true;
3357
3358 return false;
3359 }
3360
ipv6_generate_stable_address(struct in6_addr * address,u8 dad_count,const struct inet6_dev * idev)3361 static int ipv6_generate_stable_address(struct in6_addr *address,
3362 u8 dad_count,
3363 const struct inet6_dev *idev)
3364 {
3365 static DEFINE_SPINLOCK(lock);
3366 static struct sha1_ctx sha_ctx;
3367
3368 static union {
3369 u8 __data[SHA1_BLOCK_SIZE];
3370 struct {
3371 struct in6_addr secret;
3372 __be32 prefix[2];
3373 unsigned char hwaddr[MAX_ADDR_LEN];
3374 u8 dad_count;
3375 } __packed;
3376 } data;
3377
3378 struct in6_addr secret;
3379 struct in6_addr temp;
3380 struct net *net = dev_net(idev->dev);
3381
3382 BUILD_BUG_ON(sizeof(data.__data) != sizeof(data));
3383
3384 if (idev->cnf.stable_secret.initialized)
3385 secret = idev->cnf.stable_secret.secret;
3386 else if (net->ipv6.devconf_dflt->stable_secret.initialized)
3387 secret = net->ipv6.devconf_dflt->stable_secret.secret;
3388 else
3389 return -1;
3390
3391 retry:
3392 spin_lock_bh(&lock);
3393
3394 sha1_init(&sha_ctx);
3395
3396 memset(&data, 0, sizeof(data));
3397 memcpy(data.hwaddr, idev->dev->perm_addr, idev->dev->addr_len);
3398 data.prefix[0] = address->s6_addr32[0];
3399 data.prefix[1] = address->s6_addr32[1];
3400 data.secret = secret;
3401 data.dad_count = dad_count;
3402
3403 sha1_update(&sha_ctx, data.__data, sizeof(data));
3404
3405 /*
3406 * Note that the SHA-1 finalization is omitted here, and the digest is
3407 * pulled directly from the internal SHA-1 state (making it incompatible
3408 * with standard SHA-1). Unusual, but technically okay since the data
3409 * length is fixed and is a multiple of the SHA-1 block size.
3410 */
3411 temp = *address;
3412 temp.s6_addr32[2] = (__force __be32)sha_ctx.state.h[0];
3413 temp.s6_addr32[3] = (__force __be32)sha_ctx.state.h[1];
3414
3415 spin_unlock_bh(&lock);
3416
3417 if (ipv6_reserved_interfaceid(temp)) {
3418 dad_count++;
3419 if (dad_count > dev_net(idev->dev)->ipv6.sysctl.idgen_retries)
3420 return -1;
3421 goto retry;
3422 }
3423
3424 *address = temp;
3425 return 0;
3426 }
3427
ipv6_gen_mode_random_init(struct inet6_dev * idev)3428 static void ipv6_gen_mode_random_init(struct inet6_dev *idev)
3429 {
3430 struct ipv6_stable_secret *s = &idev->cnf.stable_secret;
3431
3432 if (s->initialized)
3433 return;
3434 s = &idev->cnf.stable_secret;
3435 get_random_bytes(&s->secret, sizeof(s->secret));
3436 s->initialized = true;
3437 }
3438
addrconf_addr_gen(struct inet6_dev * idev,bool prefix_route)3439 static void addrconf_addr_gen(struct inet6_dev *idev, bool prefix_route)
3440 {
3441 struct in6_addr addr;
3442
3443 /* no link local addresses on L3 master devices */
3444 if (netif_is_l3_master(idev->dev))
3445 return;
3446
3447 /* no link local addresses on devices flagged as slaves */
3448 if (idev->dev->priv_flags & IFF_NO_ADDRCONF)
3449 return;
3450
3451 ipv6_addr_set(&addr, htonl(0xFE800000), 0, 0, 0);
3452
3453 switch (idev->cnf.addr_gen_mode) {
3454 case IN6_ADDR_GEN_MODE_RANDOM:
3455 ipv6_gen_mode_random_init(idev);
3456 fallthrough;
3457 case IN6_ADDR_GEN_MODE_STABLE_PRIVACY:
3458 if (!ipv6_generate_stable_address(&addr, 0, idev))
3459 addrconf_add_linklocal(idev, &addr,
3460 IFA_F_STABLE_PRIVACY);
3461 else if (prefix_route)
3462 addrconf_prefix_route(&addr, 64, 0, idev->dev,
3463 0, 0, GFP_KERNEL);
3464 break;
3465 case IN6_ADDR_GEN_MODE_EUI64:
3466 /* addrconf_add_linklocal also adds a prefix_route and we
3467 * only need to care about prefix routes if ipv6_generate_eui64
3468 * couldn't generate one.
3469 */
3470 if (ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) == 0)
3471 addrconf_add_linklocal(idev, &addr, 0);
3472 else if (prefix_route)
3473 addrconf_prefix_route(&addr, 64, 0, idev->dev,
3474 0, 0, GFP_KERNEL);
3475 break;
3476 case IN6_ADDR_GEN_MODE_NONE:
3477 default:
3478 /* will not add any link local address */
3479 break;
3480 }
3481 }
3482
addrconf_dev_config(struct net_device * dev)3483 static void addrconf_dev_config(struct net_device *dev)
3484 {
3485 struct inet6_dev *idev;
3486
3487 ASSERT_RTNL();
3488
3489 if ((dev->type != ARPHRD_ETHER) &&
3490 (dev->type != ARPHRD_FDDI) &&
3491 (dev->type != ARPHRD_ARCNET) &&
3492 (dev->type != ARPHRD_INFINIBAND) &&
3493 (dev->type != ARPHRD_IEEE1394) &&
3494 (dev->type != ARPHRD_TUNNEL6) &&
3495 (dev->type != ARPHRD_6LOWPAN) &&
3496 (dev->type != ARPHRD_IP6GRE) &&
3497 (dev->type != ARPHRD_TUNNEL) &&
3498 (dev->type != ARPHRD_NONE) &&
3499 (dev->type != ARPHRD_RAWIP)) {
3500 /* Alas, we support only Ethernet autoconfiguration. */
3501 idev = __in6_dev_get(dev);
3502 if (!IS_ERR_OR_NULL(idev) && dev->flags & IFF_UP &&
3503 dev->flags & IFF_MULTICAST)
3504 ipv6_mc_up(idev);
3505 return;
3506 }
3507
3508 idev = addrconf_add_dev(dev);
3509 if (IS_ERR(idev))
3510 return;
3511
3512 /* this device type has no EUI support */
3513 if (dev->type == ARPHRD_NONE &&
3514 idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_EUI64)
3515 WRITE_ONCE(idev->cnf.addr_gen_mode,
3516 IN6_ADDR_GEN_MODE_RANDOM);
3517
3518 addrconf_addr_gen(idev, false);
3519 }
3520
3521 #if IS_ENABLED(CONFIG_IPV6_SIT)
addrconf_sit_config(struct net_device * dev)3522 static void addrconf_sit_config(struct net_device *dev)
3523 {
3524 struct inet6_dev *idev;
3525
3526 ASSERT_RTNL();
3527
3528 /*
3529 * Configure the tunnel with one of our IPv4
3530 * addresses... we should configure all of
3531 * our v4 addrs in the tunnel
3532 */
3533
3534 idev = ipv6_find_idev(dev);
3535 if (IS_ERR(idev)) {
3536 pr_debug("%s: add_dev failed\n", __func__);
3537 return;
3538 }
3539
3540 if (dev->priv_flags & IFF_ISATAP) {
3541 addrconf_addr_gen(idev, false);
3542 return;
3543 }
3544
3545 add_v4_addrs(idev);
3546
3547 if (dev->flags&IFF_POINTOPOINT)
3548 addrconf_add_mroute(dev);
3549 }
3550 #endif
3551
3552 #if IS_ENABLED(CONFIG_NET_IPGRE)
addrconf_gre_config(struct net_device * dev)3553 static void addrconf_gre_config(struct net_device *dev)
3554 {
3555 struct inet6_dev *idev;
3556
3557 ASSERT_RTNL();
3558
3559 idev = addrconf_add_dev(dev);
3560 if (IS_ERR(idev))
3561 return;
3562
3563 /* Generate the IPv6 link-local address using addrconf_addr_gen(),
3564 * unless we have an IPv4 GRE device not bound to an IP address and
3565 * which is in EUI64 mode (as __ipv6_isatap_ifid() would fail in this
3566 * case). Such devices fall back to add_v4_addrs() instead.
3567 */
3568 if (!(*(__be32 *)dev->dev_addr == 0 &&
3569 idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_EUI64)) {
3570 addrconf_addr_gen(idev, true);
3571 return;
3572 }
3573
3574 add_v4_addrs(idev);
3575 }
3576 #endif
3577
addrconf_init_auto_addrs(struct net_device * dev)3578 static void addrconf_init_auto_addrs(struct net_device *dev)
3579 {
3580 switch (dev->type) {
3581 #if IS_ENABLED(CONFIG_IPV6_SIT)
3582 case ARPHRD_SIT:
3583 addrconf_sit_config(dev);
3584 break;
3585 #endif
3586 #if IS_ENABLED(CONFIG_NET_IPGRE)
3587 case ARPHRD_IPGRE:
3588 addrconf_gre_config(dev);
3589 break;
3590 #endif
3591 case ARPHRD_LOOPBACK:
3592 init_loopback(dev);
3593 break;
3594
3595 default:
3596 addrconf_dev_config(dev);
3597 break;
3598 }
3599 }
3600
fixup_permanent_addr(struct net * net,struct inet6_dev * idev,struct inet6_ifaddr * ifp)3601 static int fixup_permanent_addr(struct net *net,
3602 struct inet6_dev *idev,
3603 struct inet6_ifaddr *ifp)
3604 {
3605 /* !fib6_node means the host route was removed from the
3606 * FIB, for example, if 'lo' device is taken down. In that
3607 * case regenerate the host route.
3608 */
3609 if (!ifp->rt || !ifp->rt->fib6_node) {
3610 struct fib6_info *f6i, *prev;
3611
3612 f6i = addrconf_f6i_alloc(net, idev, &ifp->addr, false,
3613 GFP_KERNEL, NULL);
3614 if (IS_ERR(f6i))
3615 return PTR_ERR(f6i);
3616
3617 /* ifp->rt can be accessed outside of rtnl */
3618 spin_lock_bh(&ifp->lock);
3619 prev = ifp->rt;
3620 ifp->rt = f6i;
3621 spin_unlock_bh(&ifp->lock);
3622
3623 fib6_info_release(prev);
3624 }
3625
3626 if (!(ifp->flags & IFA_F_NOPREFIXROUTE)) {
3627 addrconf_prefix_route(&ifp->addr, ifp->prefix_len,
3628 ifp->rt_priority, idev->dev, 0, 0,
3629 GFP_KERNEL);
3630 }
3631
3632 if (ifp->state == INET6_IFADDR_STATE_PREDAD)
3633 addrconf_dad_start(ifp);
3634
3635 return 0;
3636 }
3637
addrconf_permanent_addr(struct net * net,struct net_device * dev)3638 static void addrconf_permanent_addr(struct net *net, struct net_device *dev)
3639 {
3640 struct inet6_ifaddr *ifp;
3641 LIST_HEAD(tmp_addr_list);
3642 struct inet6_dev *idev;
3643
3644 /* Mutual exclusion with other if_list_aux users. */
3645 ASSERT_RTNL();
3646
3647 idev = __in6_dev_get(dev);
3648 if (!idev)
3649 return;
3650
3651 write_lock_bh(&idev->lock);
3652 list_for_each_entry(ifp, &idev->addr_list, if_list) {
3653 if (ifp->flags & IFA_F_PERMANENT)
3654 list_add_tail(&ifp->if_list_aux, &tmp_addr_list);
3655 }
3656 write_unlock_bh(&idev->lock);
3657
3658 while (!list_empty(&tmp_addr_list)) {
3659 ifp = list_first_entry(&tmp_addr_list,
3660 struct inet6_ifaddr, if_list_aux);
3661 list_del(&ifp->if_list_aux);
3662
3663 if (fixup_permanent_addr(net, idev, ifp) < 0) {
3664 net_info_ratelimited("%s: Failed to add prefix route for address %pI6c; dropping\n",
3665 idev->dev->name, &ifp->addr);
3666 in6_ifa_hold(ifp);
3667 ipv6_del_addr(ifp);
3668 }
3669 }
3670 }
3671
addrconf_notify(struct notifier_block * this,unsigned long event,void * ptr)3672 static int addrconf_notify(struct notifier_block *this, unsigned long event,
3673 void *ptr)
3674 {
3675 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
3676 struct netdev_notifier_change_info *change_info;
3677 struct netdev_notifier_changeupper_info *info;
3678 struct inet6_dev *idev = __in6_dev_get(dev);
3679 struct net *net = dev_net(dev);
3680 int run_pending = 0;
3681 int err;
3682
3683 switch (event) {
3684 case NETDEV_REGISTER:
3685 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
3686 idev = ipv6_add_dev(dev);
3687 if (IS_ERR(idev))
3688 return notifier_from_errno(PTR_ERR(idev));
3689 }
3690 break;
3691
3692 case NETDEV_CHANGEMTU:
3693 /* if MTU under IPV6_MIN_MTU stop IPv6 on this interface. */
3694 if (dev->mtu < IPV6_MIN_MTU) {
3695 addrconf_ifdown(dev, dev != net->loopback_dev);
3696 break;
3697 }
3698
3699 if (idev) {
3700 rt6_mtu_change(dev, dev->mtu);
3701 WRITE_ONCE(idev->cnf.mtu6, dev->mtu);
3702 break;
3703 }
3704
3705 /* allocate new idev */
3706 idev = ipv6_add_dev(dev);
3707 if (IS_ERR(idev))
3708 break;
3709
3710 /* device is still not ready */
3711 if (!(idev->if_flags & IF_READY))
3712 break;
3713
3714 run_pending = 1;
3715 fallthrough;
3716 case NETDEV_UP:
3717 case NETDEV_CHANGE:
3718 if (idev && idev->cnf.disable_ipv6)
3719 break;
3720
3721 if (dev->priv_flags & IFF_NO_ADDRCONF) {
3722 if (event == NETDEV_UP && !IS_ERR_OR_NULL(idev) &&
3723 dev->flags & IFF_UP && dev->flags & IFF_MULTICAST)
3724 ipv6_mc_up(idev);
3725 break;
3726 }
3727
3728 if (event == NETDEV_UP) {
3729 /* restore routes for permanent addresses */
3730 addrconf_permanent_addr(net, dev);
3731
3732 if (!addrconf_link_ready(dev)) {
3733 /* device is not ready yet. */
3734 pr_debug("ADDRCONF(NETDEV_UP): %s: link is not ready\n",
3735 dev->name);
3736 break;
3737 }
3738
3739 if (!idev && dev->mtu >= IPV6_MIN_MTU)
3740 idev = ipv6_add_dev(dev);
3741
3742 if (!IS_ERR_OR_NULL(idev)) {
3743 idev->if_flags |= IF_READY;
3744 run_pending = 1;
3745 }
3746 } else if (event == NETDEV_CHANGE) {
3747 if (!addrconf_link_ready(dev)) {
3748 /* device is still not ready. */
3749 rt6_sync_down_dev(dev, event);
3750 break;
3751 }
3752
3753 if (!IS_ERR_OR_NULL(idev)) {
3754 if (idev->if_flags & IF_READY) {
3755 /* device is already configured -
3756 * but resend MLD reports, we might
3757 * have roamed and need to update
3758 * multicast snooping switches
3759 */
3760 ipv6_mc_up(idev);
3761 change_info = ptr;
3762 if (change_info->flags_changed & IFF_NOARP)
3763 addrconf_dad_run(idev, true);
3764 rt6_sync_up(dev, RTNH_F_LINKDOWN);
3765 break;
3766 }
3767 idev->if_flags |= IF_READY;
3768 }
3769
3770 pr_debug("ADDRCONF(NETDEV_CHANGE): %s: link becomes ready\n",
3771 dev->name);
3772
3773 run_pending = 1;
3774 }
3775
3776 addrconf_init_auto_addrs(dev);
3777
3778 if (!IS_ERR_OR_NULL(idev)) {
3779 if (run_pending)
3780 addrconf_dad_run(idev, false);
3781
3782 /* Device has an address by now */
3783 rt6_sync_up(dev, RTNH_F_DEAD);
3784
3785 /*
3786 * If the MTU changed during the interface down,
3787 * when the interface up, the changed MTU must be
3788 * reflected in the idev as well as routers.
3789 */
3790 if (idev->cnf.mtu6 != dev->mtu &&
3791 dev->mtu >= IPV6_MIN_MTU) {
3792 rt6_mtu_change(dev, dev->mtu);
3793 WRITE_ONCE(idev->cnf.mtu6, dev->mtu);
3794 }
3795 WRITE_ONCE(idev->tstamp, jiffies);
3796 inet6_ifinfo_notify(RTM_NEWLINK, idev);
3797
3798 /*
3799 * If the changed mtu during down is lower than
3800 * IPV6_MIN_MTU stop IPv6 on this interface.
3801 */
3802 if (dev->mtu < IPV6_MIN_MTU)
3803 addrconf_ifdown(dev, dev != net->loopback_dev);
3804 }
3805 break;
3806
3807 case NETDEV_DOWN:
3808 case NETDEV_UNREGISTER:
3809 /*
3810 * Remove all addresses from this interface.
3811 */
3812 addrconf_ifdown(dev, event != NETDEV_DOWN);
3813 break;
3814
3815 case NETDEV_CHANGENAME:
3816 if (idev) {
3817 snmp6_unregister_dev(idev);
3818 addrconf_sysctl_unregister(idev);
3819 err = addrconf_sysctl_register(idev);
3820 if (err)
3821 return notifier_from_errno(err);
3822 err = snmp6_register_dev(idev);
3823 if (err) {
3824 addrconf_sysctl_unregister(idev);
3825 return notifier_from_errno(err);
3826 }
3827 }
3828 break;
3829
3830 case NETDEV_PRE_TYPE_CHANGE:
3831 case NETDEV_POST_TYPE_CHANGE:
3832 if (idev)
3833 addrconf_type_change(dev, event);
3834 break;
3835
3836 case NETDEV_CHANGEUPPER:
3837 info = ptr;
3838
3839 /* flush all routes if dev is linked to or unlinked from
3840 * an L3 master device (e.g., VRF)
3841 */
3842 if (info->upper_dev && netif_is_l3_master(info->upper_dev))
3843 addrconf_ifdown(dev, false);
3844 }
3845
3846 return NOTIFY_OK;
3847 }
3848
3849 /*
3850 * addrconf module should be notified of a device going up
3851 */
3852 static struct notifier_block ipv6_dev_notf = {
3853 .notifier_call = addrconf_notify,
3854 .priority = ADDRCONF_NOTIFY_PRIORITY,
3855 };
3856
addrconf_type_change(struct net_device * dev,unsigned long event)3857 static void addrconf_type_change(struct net_device *dev, unsigned long event)
3858 {
3859 struct inet6_dev *idev;
3860 ASSERT_RTNL();
3861
3862 idev = __in6_dev_get(dev);
3863
3864 if (event == NETDEV_POST_TYPE_CHANGE)
3865 ipv6_mc_remap(idev);
3866 else if (event == NETDEV_PRE_TYPE_CHANGE)
3867 ipv6_mc_unmap(idev);
3868 }
3869
addr_is_local(const struct in6_addr * addr)3870 static bool addr_is_local(const struct in6_addr *addr)
3871 {
3872 return ipv6_addr_type(addr) &
3873 (IPV6_ADDR_LINKLOCAL | IPV6_ADDR_LOOPBACK);
3874 }
3875
addrconf_ifdown(struct net_device * dev,bool unregister)3876 static int addrconf_ifdown(struct net_device *dev, bool unregister)
3877 {
3878 unsigned long event = unregister ? NETDEV_UNREGISTER : NETDEV_DOWN;
3879 struct net *net = dev_net(dev);
3880 struct inet6_dev *idev;
3881 struct inet6_ifaddr *ifa;
3882 LIST_HEAD(tmp_addr_list);
3883 bool keep_addr = false;
3884 bool was_ready;
3885 int state, i;
3886
3887 ASSERT_RTNL();
3888
3889 rt6_disable_ip(dev, event);
3890
3891 idev = __in6_dev_get(dev);
3892 if (!idev)
3893 return -ENODEV;
3894
3895 /*
3896 * Step 1: remove reference to ipv6 device from parent device.
3897 * Do not dev_put!
3898 */
3899 if (unregister) {
3900 WRITE_ONCE(idev->dead, 1);
3901
3902 /* protected by rtnl_lock */
3903 RCU_INIT_POINTER(dev->ip6_ptr, NULL);
3904
3905 /* Step 1.5: remove snmp6 entry */
3906 snmp6_unregister_dev(idev);
3907
3908 }
3909
3910 /* combine the user config with event to determine if permanent
3911 * addresses are to be removed from address hash table
3912 */
3913 if (!unregister && !idev->cnf.disable_ipv6) {
3914 /* aggregate the system setting and interface setting */
3915 int _keep_addr = READ_ONCE(net->ipv6.devconf_all->keep_addr_on_down);
3916
3917 if (!_keep_addr)
3918 _keep_addr = READ_ONCE(idev->cnf.keep_addr_on_down);
3919
3920 keep_addr = (_keep_addr > 0);
3921 }
3922
3923 /* Step 2: clear hash table */
3924 for (i = 0; i < IN6_ADDR_HSIZE; i++) {
3925 struct hlist_head *h = &net->ipv6.inet6_addr_lst[i];
3926
3927 spin_lock_bh(&net->ipv6.addrconf_hash_lock);
3928 restart:
3929 hlist_for_each_entry_rcu(ifa, h, addr_lst) {
3930 if (ifa->idev == idev) {
3931 addrconf_del_dad_work(ifa);
3932 /* combined flag + permanent flag decide if
3933 * address is retained on a down event
3934 */
3935 if (!keep_addr ||
3936 !(ifa->flags & IFA_F_PERMANENT) ||
3937 addr_is_local(&ifa->addr)) {
3938 hlist_del_init_rcu(&ifa->addr_lst);
3939 goto restart;
3940 }
3941 }
3942 }
3943 spin_unlock_bh(&net->ipv6.addrconf_hash_lock);
3944 }
3945
3946 write_lock_bh(&idev->lock);
3947
3948 addrconf_del_rs_timer(idev);
3949
3950 /* Step 2: clear flags for stateless addrconf, repeated down
3951 * detection
3952 */
3953 was_ready = idev->if_flags & IF_READY;
3954 if (!unregister)
3955 idev->if_flags &= ~(IF_RS_SENT|IF_RA_RCVD|IF_READY);
3956
3957 /* Step 3: clear tempaddr list */
3958 while (!list_empty(&idev->tempaddr_list)) {
3959 ifa = list_first_entry(&idev->tempaddr_list,
3960 struct inet6_ifaddr, tmp_list);
3961 list_del(&ifa->tmp_list);
3962 write_unlock_bh(&idev->lock);
3963 spin_lock_bh(&ifa->lock);
3964
3965 if (ifa->ifpub) {
3966 in6_ifa_put(ifa->ifpub);
3967 ifa->ifpub = NULL;
3968 }
3969 spin_unlock_bh(&ifa->lock);
3970 in6_ifa_put(ifa);
3971 write_lock_bh(&idev->lock);
3972 }
3973
3974 list_for_each_entry(ifa, &idev->addr_list, if_list)
3975 list_add_tail(&ifa->if_list_aux, &tmp_addr_list);
3976 write_unlock_bh(&idev->lock);
3977
3978 while (!list_empty(&tmp_addr_list)) {
3979 struct fib6_info *rt = NULL;
3980 bool keep;
3981
3982 ifa = list_first_entry(&tmp_addr_list,
3983 struct inet6_ifaddr, if_list_aux);
3984 list_del(&ifa->if_list_aux);
3985
3986 addrconf_del_dad_work(ifa);
3987
3988 keep = keep_addr && (ifa->flags & IFA_F_PERMANENT) &&
3989 !addr_is_local(&ifa->addr);
3990
3991 spin_lock_bh(&ifa->lock);
3992
3993 if (keep) {
3994 /* set state to skip the notifier below */
3995 state = INET6_IFADDR_STATE_DEAD;
3996 ifa->state = INET6_IFADDR_STATE_PREDAD;
3997 if (!(ifa->flags & IFA_F_NODAD))
3998 ifa->flags |= IFA_F_TENTATIVE;
3999
4000 rt = ifa->rt;
4001 ifa->rt = NULL;
4002 } else {
4003 state = ifa->state;
4004 ifa->state = INET6_IFADDR_STATE_DEAD;
4005 }
4006
4007 spin_unlock_bh(&ifa->lock);
4008
4009 if (rt)
4010 ip6_del_rt(net, rt, false);
4011
4012 if (state != INET6_IFADDR_STATE_DEAD) {
4013 __ipv6_ifa_notify(RTM_DELADDR, ifa);
4014 inet6addr_notifier_call_chain(NETDEV_DOWN, ifa);
4015 } else {
4016 if (idev->cnf.forwarding)
4017 addrconf_leave_anycast(ifa);
4018 addrconf_leave_solict(ifa->idev, &ifa->addr);
4019 }
4020
4021 if (!keep) {
4022 write_lock_bh(&idev->lock);
4023 list_del_rcu(&ifa->if_list);
4024 write_unlock_bh(&idev->lock);
4025 in6_ifa_put(ifa);
4026 }
4027 }
4028
4029 /* Step 5: Discard anycast and multicast list */
4030 if (unregister) {
4031 ipv6_ac_destroy_dev(idev);
4032 ipv6_mc_destroy_dev(idev);
4033 } else if (was_ready) {
4034 ipv6_mc_down(idev);
4035 }
4036
4037 WRITE_ONCE(idev->tstamp, jiffies);
4038 idev->ra_mtu = 0;
4039
4040 /* Last: Shot the device (if unregistered) */
4041 if (unregister) {
4042 addrconf_sysctl_unregister(idev);
4043 neigh_parms_release(&nd_tbl, idev->nd_parms);
4044 neigh_ifdown(&nd_tbl, dev);
4045 in6_dev_put(idev);
4046 }
4047 return 0;
4048 }
4049
addrconf_rs_timer(struct timer_list * t)4050 static void addrconf_rs_timer(struct timer_list *t)
4051 {
4052 struct inet6_dev *idev = timer_container_of(idev, t, rs_timer);
4053 struct net_device *dev = idev->dev;
4054 struct in6_addr lladdr;
4055 int rtr_solicits;
4056
4057 write_lock(&idev->lock);
4058 if (idev->dead || !(idev->if_flags & IF_READY))
4059 goto out;
4060
4061 if (!ipv6_accept_ra(idev))
4062 goto out;
4063
4064 /* Announcement received after solicitation was sent */
4065 if (idev->if_flags & IF_RA_RCVD)
4066 goto out;
4067
4068 rtr_solicits = READ_ONCE(idev->cnf.rtr_solicits);
4069
4070 if (idev->rs_probes++ < rtr_solicits || rtr_solicits < 0) {
4071 write_unlock(&idev->lock);
4072 if (!ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE))
4073 ndisc_send_rs(dev, &lladdr,
4074 &in6addr_linklocal_allrouters);
4075 else
4076 goto put;
4077
4078 write_lock(&idev->lock);
4079 idev->rs_interval = rfc3315_s14_backoff_update(
4080 idev->rs_interval,
4081 READ_ONCE(idev->cnf.rtr_solicit_max_interval));
4082 /* The wait after the last probe can be shorter */
4083 addrconf_mod_rs_timer(idev, (idev->rs_probes ==
4084 READ_ONCE(idev->cnf.rtr_solicits)) ?
4085 READ_ONCE(idev->cnf.rtr_solicit_delay) :
4086 idev->rs_interval);
4087 } else {
4088 /*
4089 * Note: we do not support deprecated "all on-link"
4090 * assumption any longer.
4091 */
4092 pr_debug("%s: no IPv6 routers present\n", idev->dev->name);
4093 }
4094
4095 out:
4096 write_unlock(&idev->lock);
4097 put:
4098 in6_dev_put(idev);
4099 }
4100
4101 /*
4102 * Duplicate Address Detection
4103 */
addrconf_dad_kick(struct inet6_ifaddr * ifp)4104 static void addrconf_dad_kick(struct inet6_ifaddr *ifp)
4105 {
4106 struct inet6_dev *idev = ifp->idev;
4107 unsigned long rand_num;
4108 u64 nonce;
4109
4110 if (ifp->flags & IFA_F_OPTIMISTIC)
4111 rand_num = 0;
4112 else
4113 rand_num = get_random_u32_below(
4114 READ_ONCE(idev->cnf.rtr_solicit_delay) ? : 1);
4115
4116 nonce = 0;
4117 if (READ_ONCE(idev->cnf.enhanced_dad) ||
4118 READ_ONCE(dev_net(idev->dev)->ipv6.devconf_all->enhanced_dad)) {
4119 do
4120 get_random_bytes(&nonce, 6);
4121 while (nonce == 0);
4122 }
4123 ifp->dad_nonce = nonce;
4124 ifp->dad_probes = READ_ONCE(idev->cnf.dad_transmits);
4125 addrconf_mod_dad_work(ifp, rand_num);
4126 }
4127
addrconf_dad_begin(struct inet6_ifaddr * ifp)4128 static void addrconf_dad_begin(struct inet6_ifaddr *ifp)
4129 {
4130 struct inet6_dev *idev = ifp->idev;
4131 struct net_device *dev = idev->dev;
4132 bool bump_id, notify = false;
4133 struct net *net;
4134
4135 addrconf_join_solict(dev, &ifp->addr);
4136
4137 read_lock_bh(&idev->lock);
4138 spin_lock(&ifp->lock);
4139 if (ifp->state == INET6_IFADDR_STATE_DEAD)
4140 goto out;
4141
4142 net = dev_net(dev);
4143 if (dev->flags&(IFF_NOARP|IFF_LOOPBACK) ||
4144 (READ_ONCE(net->ipv6.devconf_all->accept_dad) < 1 &&
4145 READ_ONCE(idev->cnf.accept_dad) < 1) ||
4146 !(ifp->flags&IFA_F_TENTATIVE) ||
4147 ifp->flags & IFA_F_NODAD) {
4148 bool send_na = false;
4149
4150 if (ifp->flags & IFA_F_TENTATIVE &&
4151 !(ifp->flags & IFA_F_OPTIMISTIC))
4152 send_na = true;
4153 bump_id = ifp->flags & IFA_F_TENTATIVE;
4154 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
4155 spin_unlock(&ifp->lock);
4156 read_unlock_bh(&idev->lock);
4157
4158 addrconf_dad_completed(ifp, bump_id, send_na);
4159 return;
4160 }
4161
4162 if (!(idev->if_flags & IF_READY)) {
4163 spin_unlock(&ifp->lock);
4164 read_unlock_bh(&idev->lock);
4165 /*
4166 * If the device is not ready:
4167 * - keep it tentative if it is a permanent address.
4168 * - otherwise, kill it.
4169 */
4170 in6_ifa_hold(ifp);
4171 addrconf_dad_stop(ifp, 0);
4172 return;
4173 }
4174
4175 /*
4176 * Optimistic nodes can start receiving
4177 * Frames right away
4178 */
4179 if (ifp->flags & IFA_F_OPTIMISTIC) {
4180 ip6_ins_rt(net, ifp->rt);
4181 if (ipv6_use_optimistic_addr(net, idev)) {
4182 /* Because optimistic nodes can use this address,
4183 * notify listeners. If DAD fails, RTM_DELADDR is sent.
4184 */
4185 notify = true;
4186 }
4187 }
4188
4189 addrconf_dad_kick(ifp);
4190 out:
4191 spin_unlock(&ifp->lock);
4192 read_unlock_bh(&idev->lock);
4193 if (notify)
4194 ipv6_ifa_notify(RTM_NEWADDR, ifp);
4195 }
4196
addrconf_dad_start(struct inet6_ifaddr * ifp)4197 static void addrconf_dad_start(struct inet6_ifaddr *ifp)
4198 {
4199 bool begin_dad = false;
4200
4201 spin_lock_bh(&ifp->lock);
4202 if (ifp->state != INET6_IFADDR_STATE_DEAD) {
4203 ifp->state = INET6_IFADDR_STATE_PREDAD;
4204 begin_dad = true;
4205 }
4206 spin_unlock_bh(&ifp->lock);
4207
4208 if (begin_dad)
4209 addrconf_mod_dad_work(ifp, 0);
4210 }
4211
addrconf_dad_work(struct work_struct * w)4212 static void addrconf_dad_work(struct work_struct *w)
4213 {
4214 struct inet6_ifaddr *ifp = container_of(to_delayed_work(w),
4215 struct inet6_ifaddr,
4216 dad_work);
4217 struct inet6_dev *idev = ifp->idev;
4218 bool bump_id, disable_ipv6 = false;
4219 struct in6_addr mcaddr;
4220 struct net *net;
4221
4222 enum {
4223 DAD_PROCESS,
4224 DAD_BEGIN,
4225 DAD_ABORT,
4226 } action = DAD_PROCESS;
4227
4228 net = dev_net(idev->dev);
4229
4230 rtnl_net_lock(net);
4231
4232 spin_lock_bh(&ifp->lock);
4233 if (ifp->state == INET6_IFADDR_STATE_PREDAD) {
4234 action = DAD_BEGIN;
4235 ifp->state = INET6_IFADDR_STATE_DAD;
4236 } else if (ifp->state == INET6_IFADDR_STATE_ERRDAD) {
4237 action = DAD_ABORT;
4238 ifp->state = INET6_IFADDR_STATE_POSTDAD;
4239
4240 if ((READ_ONCE(net->ipv6.devconf_all->accept_dad) > 1 ||
4241 READ_ONCE(idev->cnf.accept_dad) > 1) &&
4242 !idev->cnf.disable_ipv6 &&
4243 !(ifp->flags & IFA_F_STABLE_PRIVACY)) {
4244 struct in6_addr addr;
4245
4246 addr.s6_addr32[0] = htonl(0xfe800000);
4247 addr.s6_addr32[1] = 0;
4248
4249 if (!ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) &&
4250 ipv6_addr_equal(&ifp->addr, &addr)) {
4251 /* DAD failed for link-local based on MAC */
4252 WRITE_ONCE(idev->cnf.disable_ipv6, 1);
4253
4254 pr_info("%s: IPv6 being disabled!\n",
4255 ifp->idev->dev->name);
4256 disable_ipv6 = true;
4257 }
4258 }
4259 }
4260 spin_unlock_bh(&ifp->lock);
4261
4262 if (action == DAD_BEGIN) {
4263 addrconf_dad_begin(ifp);
4264 goto out;
4265 } else if (action == DAD_ABORT) {
4266 in6_ifa_hold(ifp);
4267 addrconf_dad_stop(ifp, 1);
4268 if (disable_ipv6)
4269 addrconf_ifdown(idev->dev, false);
4270 goto out;
4271 }
4272
4273 if (!ifp->dad_probes && addrconf_dad_end(ifp))
4274 goto out;
4275
4276 write_lock_bh(&idev->lock);
4277 if (idev->dead || !(idev->if_flags & IF_READY)) {
4278 write_unlock_bh(&idev->lock);
4279 goto out;
4280 }
4281
4282 spin_lock(&ifp->lock);
4283 if (ifp->state == INET6_IFADDR_STATE_DEAD) {
4284 spin_unlock(&ifp->lock);
4285 write_unlock_bh(&idev->lock);
4286 goto out;
4287 }
4288
4289 if (ifp->dad_probes == 0) {
4290 bool send_na = false;
4291
4292 /*
4293 * DAD was successful
4294 */
4295
4296 if (ifp->flags & IFA_F_TENTATIVE &&
4297 !(ifp->flags & IFA_F_OPTIMISTIC))
4298 send_na = true;
4299 bump_id = ifp->flags & IFA_F_TENTATIVE;
4300 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
4301 spin_unlock(&ifp->lock);
4302 write_unlock_bh(&idev->lock);
4303
4304 addrconf_dad_completed(ifp, bump_id, send_na);
4305
4306 goto out;
4307 }
4308
4309 ifp->dad_probes--;
4310 addrconf_mod_dad_work(ifp,
4311 max(NEIGH_VAR(ifp->idev->nd_parms, RETRANS_TIME),
4312 HZ/100));
4313 spin_unlock(&ifp->lock);
4314 write_unlock_bh(&idev->lock);
4315
4316 /* send a neighbour solicitation for our addr */
4317 addrconf_addr_solict_mult(&ifp->addr, &mcaddr);
4318 ndisc_send_ns(ifp->idev->dev, &ifp->addr, &mcaddr, &in6addr_any,
4319 ifp->dad_nonce);
4320 out:
4321 in6_ifa_put(ifp);
4322 rtnl_net_unlock(net);
4323 }
4324
4325 /* ifp->idev must be at least read locked */
ipv6_lonely_lladdr(struct inet6_ifaddr * ifp)4326 static bool ipv6_lonely_lladdr(struct inet6_ifaddr *ifp)
4327 {
4328 struct inet6_ifaddr *ifpiter;
4329 struct inet6_dev *idev = ifp->idev;
4330
4331 list_for_each_entry_reverse(ifpiter, &idev->addr_list, if_list) {
4332 if (ifpiter->scope > IFA_LINK)
4333 break;
4334 if (ifp != ifpiter && ifpiter->scope == IFA_LINK &&
4335 (ifpiter->flags & (IFA_F_PERMANENT|IFA_F_TENTATIVE|
4336 IFA_F_OPTIMISTIC|IFA_F_DADFAILED)) ==
4337 IFA_F_PERMANENT)
4338 return false;
4339 }
4340 return true;
4341 }
4342
addrconf_dad_completed(struct inet6_ifaddr * ifp,bool bump_id,bool send_na)4343 static void addrconf_dad_completed(struct inet6_ifaddr *ifp, bool bump_id,
4344 bool send_na)
4345 {
4346 struct net_device *dev = ifp->idev->dev;
4347 struct in6_addr lladdr;
4348 bool send_rs, send_mld;
4349
4350 addrconf_del_dad_work(ifp);
4351
4352 /*
4353 * Configure the address for reception. Now it is valid.
4354 */
4355
4356 ipv6_ifa_notify(RTM_NEWADDR, ifp);
4357
4358 /* If added prefix is link local and we are prepared to process
4359 router advertisements, start sending router solicitations.
4360 */
4361
4362 read_lock_bh(&ifp->idev->lock);
4363 send_mld = ifp->scope == IFA_LINK && ipv6_lonely_lladdr(ifp);
4364 send_rs = send_mld &&
4365 ipv6_accept_ra(ifp->idev) &&
4366 READ_ONCE(ifp->idev->cnf.rtr_solicits) != 0 &&
4367 (dev->flags & IFF_LOOPBACK) == 0 &&
4368 (dev->type != ARPHRD_TUNNEL) &&
4369 !netif_is_team_port(dev);
4370 read_unlock_bh(&ifp->idev->lock);
4371
4372 /* While dad is in progress mld report's source address is in6_addrany.
4373 * Resend with proper ll now.
4374 */
4375 if (send_mld)
4376 ipv6_mc_dad_complete(ifp->idev);
4377
4378 /* send unsolicited NA if enabled */
4379 if (send_na &&
4380 (READ_ONCE(ifp->idev->cnf.ndisc_notify) ||
4381 READ_ONCE(dev_net(dev)->ipv6.devconf_all->ndisc_notify))) {
4382 ndisc_send_na(dev, &in6addr_linklocal_allnodes, &ifp->addr,
4383 /*router=*/ !!ifp->idev->cnf.forwarding,
4384 /*solicited=*/ false, /*override=*/ true,
4385 /*inc_opt=*/ true);
4386 }
4387
4388 if (send_rs) {
4389 /*
4390 * If a host as already performed a random delay
4391 * [...] as part of DAD [...] there is no need
4392 * to delay again before sending the first RS
4393 */
4394 if (ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE))
4395 return;
4396 ndisc_send_rs(dev, &lladdr, &in6addr_linklocal_allrouters);
4397
4398 write_lock_bh(&ifp->idev->lock);
4399 spin_lock(&ifp->lock);
4400 ifp->idev->rs_interval = rfc3315_s14_backoff_init(
4401 READ_ONCE(ifp->idev->cnf.rtr_solicit_interval));
4402 ifp->idev->rs_probes = 1;
4403 ifp->idev->if_flags |= IF_RS_SENT;
4404 addrconf_mod_rs_timer(ifp->idev, ifp->idev->rs_interval);
4405 spin_unlock(&ifp->lock);
4406 write_unlock_bh(&ifp->idev->lock);
4407 }
4408
4409 if (bump_id)
4410 rt_genid_bump_ipv6(dev_net(dev));
4411
4412 /* Make sure that a new temporary address will be created
4413 * before this temporary address becomes deprecated.
4414 */
4415 if (ifp->flags & IFA_F_TEMPORARY)
4416 addrconf_verify_rtnl(dev_net(dev));
4417 }
4418
addrconf_dad_run(struct inet6_dev * idev,bool restart)4419 static void addrconf_dad_run(struct inet6_dev *idev, bool restart)
4420 {
4421 struct inet6_ifaddr *ifp;
4422
4423 read_lock_bh(&idev->lock);
4424 list_for_each_entry(ifp, &idev->addr_list, if_list) {
4425 spin_lock(&ifp->lock);
4426 if ((ifp->flags & IFA_F_TENTATIVE &&
4427 ifp->state == INET6_IFADDR_STATE_DAD) || restart) {
4428 if (restart)
4429 ifp->state = INET6_IFADDR_STATE_PREDAD;
4430 addrconf_dad_kick(ifp);
4431 }
4432 spin_unlock(&ifp->lock);
4433 }
4434 read_unlock_bh(&idev->lock);
4435 }
4436
4437 #ifdef CONFIG_PROC_FS
4438 struct if6_iter_state {
4439 struct seq_net_private p;
4440 int bucket;
4441 int offset;
4442 };
4443
if6_get_first(struct seq_file * seq,loff_t pos)4444 static struct inet6_ifaddr *if6_get_first(struct seq_file *seq, loff_t pos)
4445 {
4446 struct if6_iter_state *state = seq->private;
4447 struct net *net = seq_file_net(seq);
4448 struct inet6_ifaddr *ifa = NULL;
4449 int p = 0;
4450
4451 /* initial bucket if pos is 0 */
4452 if (pos == 0) {
4453 state->bucket = 0;
4454 state->offset = 0;
4455 }
4456
4457 for (; state->bucket < IN6_ADDR_HSIZE; ++state->bucket) {
4458 hlist_for_each_entry_rcu(ifa, &net->ipv6.inet6_addr_lst[state->bucket],
4459 addr_lst) {
4460 /* sync with offset */
4461 if (p < state->offset) {
4462 p++;
4463 continue;
4464 }
4465 return ifa;
4466 }
4467
4468 /* prepare for next bucket */
4469 state->offset = 0;
4470 p = 0;
4471 }
4472 return NULL;
4473 }
4474
if6_get_next(struct seq_file * seq,struct inet6_ifaddr * ifa)4475 static struct inet6_ifaddr *if6_get_next(struct seq_file *seq,
4476 struct inet6_ifaddr *ifa)
4477 {
4478 struct if6_iter_state *state = seq->private;
4479 struct net *net = seq_file_net(seq);
4480
4481 hlist_for_each_entry_continue_rcu(ifa, addr_lst) {
4482 state->offset++;
4483 return ifa;
4484 }
4485
4486 state->offset = 0;
4487 while (++state->bucket < IN6_ADDR_HSIZE) {
4488 hlist_for_each_entry_rcu(ifa,
4489 &net->ipv6.inet6_addr_lst[state->bucket], addr_lst) {
4490 return ifa;
4491 }
4492 }
4493
4494 return NULL;
4495 }
4496
if6_seq_start(struct seq_file * seq,loff_t * pos)4497 static void *if6_seq_start(struct seq_file *seq, loff_t *pos)
4498 __acquires(rcu)
4499 {
4500 rcu_read_lock();
4501 return if6_get_first(seq, *pos);
4502 }
4503
if6_seq_next(struct seq_file * seq,void * v,loff_t * pos)4504 static void *if6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
4505 {
4506 struct inet6_ifaddr *ifa;
4507
4508 ifa = if6_get_next(seq, v);
4509 ++*pos;
4510 return ifa;
4511 }
4512
if6_seq_stop(struct seq_file * seq,void * v)4513 static void if6_seq_stop(struct seq_file *seq, void *v)
4514 __releases(rcu)
4515 {
4516 rcu_read_unlock();
4517 }
4518
if6_seq_show(struct seq_file * seq,void * v)4519 static int if6_seq_show(struct seq_file *seq, void *v)
4520 {
4521 struct inet6_ifaddr *ifp = (struct inet6_ifaddr *)v;
4522 seq_printf(seq, "%pi6 %02x %02x %02x %02x %8s\n",
4523 &ifp->addr,
4524 ifp->idev->dev->ifindex,
4525 ifp->prefix_len,
4526 ifp->scope,
4527 (u8) ifp->flags,
4528 ifp->idev->dev->name);
4529 return 0;
4530 }
4531
4532 static const struct seq_operations if6_seq_ops = {
4533 .start = if6_seq_start,
4534 .next = if6_seq_next,
4535 .show = if6_seq_show,
4536 .stop = if6_seq_stop,
4537 };
4538
if6_proc_net_init(struct net * net)4539 static int __net_init if6_proc_net_init(struct net *net)
4540 {
4541 if (!proc_create_net("if_inet6", 0444, net->proc_net, &if6_seq_ops,
4542 sizeof(struct if6_iter_state)))
4543 return -ENOMEM;
4544 return 0;
4545 }
4546
if6_proc_net_exit(struct net * net)4547 static void __net_exit if6_proc_net_exit(struct net *net)
4548 {
4549 remove_proc_entry("if_inet6", net->proc_net);
4550 }
4551
4552 static struct pernet_operations if6_proc_net_ops = {
4553 .init = if6_proc_net_init,
4554 .exit = if6_proc_net_exit,
4555 };
4556
if6_proc_init(void)4557 int __init if6_proc_init(void)
4558 {
4559 return register_pernet_subsys(&if6_proc_net_ops);
4560 }
4561
if6_proc_exit(void)4562 void if6_proc_exit(void)
4563 {
4564 unregister_pernet_subsys(&if6_proc_net_ops);
4565 }
4566 #endif /* CONFIG_PROC_FS */
4567
4568 #if IS_ENABLED(CONFIG_IPV6_MIP6)
4569 /* Check if address is a home address configured on any interface. */
ipv6_chk_home_addr(struct net * net,const struct in6_addr * addr)4570 int ipv6_chk_home_addr(struct net *net, const struct in6_addr *addr)
4571 {
4572 unsigned int hash = inet6_addr_hash(net, addr);
4573 struct inet6_ifaddr *ifp = NULL;
4574 int ret = 0;
4575
4576 rcu_read_lock();
4577 hlist_for_each_entry_rcu(ifp, &net->ipv6.inet6_addr_lst[hash], addr_lst) {
4578 if (ipv6_addr_equal(&ifp->addr, addr) &&
4579 (ifp->flags & IFA_F_HOMEADDRESS)) {
4580 ret = 1;
4581 break;
4582 }
4583 }
4584 rcu_read_unlock();
4585 return ret;
4586 }
4587 #endif
4588
4589 /* RFC6554 has some algorithm to avoid loops in segment routing by
4590 * checking if the segments contains any of a local interface address.
4591 *
4592 * Quote:
4593 *
4594 * To detect loops in the SRH, a router MUST determine if the SRH
4595 * includes multiple addresses assigned to any interface on that router.
4596 * If such addresses appear more than once and are separated by at least
4597 * one address not assigned to that router.
4598 */
ipv6_chk_rpl_srh_loop(struct net * net,const struct in6_addr * segs,unsigned char nsegs)4599 int ipv6_chk_rpl_srh_loop(struct net *net, const struct in6_addr *segs,
4600 unsigned char nsegs)
4601 {
4602 const struct in6_addr *addr;
4603 int i, ret = 0, found = 0;
4604 struct inet6_ifaddr *ifp;
4605 bool separated = false;
4606 unsigned int hash;
4607 bool hash_found;
4608
4609 rcu_read_lock();
4610 for (i = 0; i < nsegs; i++) {
4611 addr = &segs[i];
4612 hash = inet6_addr_hash(net, addr);
4613
4614 hash_found = false;
4615 hlist_for_each_entry_rcu(ifp, &net->ipv6.inet6_addr_lst[hash], addr_lst) {
4616
4617 if (ipv6_addr_equal(&ifp->addr, addr)) {
4618 hash_found = true;
4619 break;
4620 }
4621 }
4622
4623 if (hash_found) {
4624 if (found > 1 && separated) {
4625 ret = 1;
4626 break;
4627 }
4628
4629 separated = false;
4630 found++;
4631 } else {
4632 separated = true;
4633 }
4634 }
4635 rcu_read_unlock();
4636
4637 return ret;
4638 }
4639
4640 /*
4641 * Periodic address status verification
4642 */
4643
addrconf_verify_rtnl(struct net * net)4644 static void addrconf_verify_rtnl(struct net *net)
4645 {
4646 unsigned long now, next, next_sec, next_sched;
4647 struct inet6_ifaddr *ifp;
4648 int i;
4649
4650 ASSERT_RTNL();
4651
4652 rcu_read_lock_bh();
4653 now = jiffies;
4654 next = round_jiffies_up(now + ADDR_CHECK_FREQUENCY);
4655
4656 cancel_delayed_work(&net->ipv6.addr_chk_work);
4657
4658 for (i = 0; i < IN6_ADDR_HSIZE; i++) {
4659 restart:
4660 hlist_for_each_entry_rcu_bh(ifp, &net->ipv6.inet6_addr_lst[i], addr_lst) {
4661 unsigned long age;
4662
4663 /* When setting preferred_lft to a value not zero or
4664 * infinity, while valid_lft is infinity
4665 * IFA_F_PERMANENT has a non-infinity life time.
4666 */
4667 if ((ifp->flags & IFA_F_PERMANENT) &&
4668 (ifp->prefered_lft == INFINITY_LIFE_TIME))
4669 continue;
4670
4671 spin_lock(&ifp->lock);
4672 /* We try to batch several events at once. */
4673 age = (now - ifp->tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
4674
4675 if ((ifp->flags&IFA_F_TEMPORARY) &&
4676 !(ifp->flags&IFA_F_TENTATIVE) &&
4677 ifp->prefered_lft != INFINITY_LIFE_TIME &&
4678 !ifp->regen_count && ifp->ifpub) {
4679 /* This is a non-regenerated temporary addr. */
4680
4681 unsigned long regen_advance = ipv6_get_regen_advance(ifp->idev);
4682
4683 if (age + regen_advance >= ifp->prefered_lft) {
4684 struct inet6_ifaddr *ifpub = ifp->ifpub;
4685 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
4686 next = ifp->tstamp + ifp->prefered_lft * HZ;
4687
4688 ifp->regen_count++;
4689 in6_ifa_hold(ifp);
4690 in6_ifa_hold(ifpub);
4691 spin_unlock(&ifp->lock);
4692
4693 spin_lock(&ifpub->lock);
4694 ifpub->regen_count = 0;
4695 spin_unlock(&ifpub->lock);
4696 rcu_read_unlock_bh();
4697 ipv6_create_tempaddr(ifpub, true);
4698 in6_ifa_put(ifpub);
4699 in6_ifa_put(ifp);
4700 rcu_read_lock_bh();
4701 goto restart;
4702 } else if (time_before(ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ, next))
4703 next = ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ;
4704 }
4705
4706 if (ifp->valid_lft != INFINITY_LIFE_TIME &&
4707 age >= ifp->valid_lft) {
4708 spin_unlock(&ifp->lock);
4709 in6_ifa_hold(ifp);
4710 rcu_read_unlock_bh();
4711 ipv6_del_addr(ifp);
4712 rcu_read_lock_bh();
4713 goto restart;
4714 } else if (ifp->prefered_lft == INFINITY_LIFE_TIME) {
4715 spin_unlock(&ifp->lock);
4716 continue;
4717 } else if (age >= ifp->prefered_lft) {
4718 /* jiffies - ifp->tstamp > age >= ifp->prefered_lft */
4719 int deprecate = 0;
4720
4721 if (!(ifp->flags&IFA_F_DEPRECATED)) {
4722 deprecate = 1;
4723 ifp->flags |= IFA_F_DEPRECATED;
4724 }
4725
4726 if ((ifp->valid_lft != INFINITY_LIFE_TIME) &&
4727 (time_before(ifp->tstamp + ifp->valid_lft * HZ, next)))
4728 next = ifp->tstamp + ifp->valid_lft * HZ;
4729
4730 spin_unlock(&ifp->lock);
4731
4732 if (deprecate) {
4733 in6_ifa_hold(ifp);
4734
4735 ipv6_ifa_notify(0, ifp);
4736 in6_ifa_put(ifp);
4737 goto restart;
4738 }
4739 } else {
4740 /* ifp->prefered_lft <= ifp->valid_lft */
4741 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
4742 next = ifp->tstamp + ifp->prefered_lft * HZ;
4743 spin_unlock(&ifp->lock);
4744 }
4745 }
4746 }
4747
4748 next_sec = round_jiffies_up(next);
4749 next_sched = next;
4750
4751 /* If rounded timeout is accurate enough, accept it. */
4752 if (time_before(next_sec, next + ADDRCONF_TIMER_FUZZ))
4753 next_sched = next_sec;
4754
4755 /* And minimum interval is ADDRCONF_TIMER_FUZZ_MAX. */
4756 if (time_before(next_sched, jiffies + ADDRCONF_TIMER_FUZZ_MAX))
4757 next_sched = jiffies + ADDRCONF_TIMER_FUZZ_MAX;
4758
4759 pr_debug("now = %lu, schedule = %lu, rounded schedule = %lu => %lu\n",
4760 now, next, next_sec, next_sched);
4761 mod_delayed_work(addrconf_wq, &net->ipv6.addr_chk_work, next_sched - now);
4762 rcu_read_unlock_bh();
4763 }
4764
addrconf_verify_work(struct work_struct * w)4765 static void addrconf_verify_work(struct work_struct *w)
4766 {
4767 struct net *net = container_of(to_delayed_work(w), struct net,
4768 ipv6.addr_chk_work);
4769
4770 rtnl_net_lock(net);
4771 addrconf_verify_rtnl(net);
4772 rtnl_net_unlock(net);
4773 }
4774
addrconf_verify(struct net * net)4775 static void addrconf_verify(struct net *net)
4776 {
4777 mod_delayed_work(addrconf_wq, &net->ipv6.addr_chk_work, 0);
4778 }
4779
extract_addr(struct nlattr * addr,struct nlattr * local,struct in6_addr ** peer_pfx)4780 static struct in6_addr *extract_addr(struct nlattr *addr, struct nlattr *local,
4781 struct in6_addr **peer_pfx)
4782 {
4783 struct in6_addr *pfx = NULL;
4784
4785 *peer_pfx = NULL;
4786
4787 if (addr)
4788 pfx = nla_data(addr);
4789
4790 if (local) {
4791 if (pfx && nla_memcmp(local, pfx, sizeof(*pfx)))
4792 *peer_pfx = pfx;
4793 pfx = nla_data(local);
4794 }
4795
4796 return pfx;
4797 }
4798
4799 static const struct nla_policy ifa_ipv6_policy[IFA_MAX+1] = {
4800 [IFA_ADDRESS] = { .len = sizeof(struct in6_addr) },
4801 [IFA_LOCAL] = { .len = sizeof(struct in6_addr) },
4802 [IFA_CACHEINFO] = { .len = sizeof(struct ifa_cacheinfo) },
4803 [IFA_FLAGS] = { .len = sizeof(u32) },
4804 [IFA_RT_PRIORITY] = { .len = sizeof(u32) },
4805 [IFA_TARGET_NETNSID] = { .type = NLA_S32 },
4806 [IFA_PROTO] = { .type = NLA_U8 },
4807 };
4808
4809 static int
inet6_rtm_deladdr(struct sk_buff * skb,struct nlmsghdr * nlh,struct netlink_ext_ack * extack)4810 inet6_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh,
4811 struct netlink_ext_ack *extack)
4812 {
4813 struct net *net = sock_net(skb->sk);
4814 struct ifaddrmsg *ifm;
4815 struct nlattr *tb[IFA_MAX+1];
4816 struct in6_addr *pfx, *peer_pfx;
4817 u32 ifa_flags;
4818 int err;
4819
4820 err = nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFA_MAX,
4821 ifa_ipv6_policy, extack);
4822 if (err < 0)
4823 return err;
4824
4825 ifm = nlmsg_data(nlh);
4826 pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
4827 if (!pfx)
4828 return -EINVAL;
4829
4830 ifa_flags = nla_get_u32_default(tb[IFA_FLAGS], ifm->ifa_flags);
4831
4832 /* We ignore other flags so far. */
4833 ifa_flags &= IFA_F_MANAGETEMPADDR;
4834
4835 rtnl_net_lock(net);
4836 err = inet6_addr_del(net, ifm->ifa_index, ifa_flags, pfx,
4837 ifm->ifa_prefixlen, extack);
4838 rtnl_net_unlock(net);
4839
4840 return err;
4841 }
4842
modify_prefix_route(struct net * net,struct inet6_ifaddr * ifp,unsigned long expires,u32 flags,bool modify_peer)4843 static int modify_prefix_route(struct net *net, struct inet6_ifaddr *ifp,
4844 unsigned long expires, u32 flags,
4845 bool modify_peer)
4846 {
4847 struct fib6_table *table;
4848 struct fib6_info *f6i;
4849 u32 prio;
4850
4851 f6i = addrconf_get_prefix_route(modify_peer ? &ifp->peer_addr : &ifp->addr,
4852 ifp->prefix_len,
4853 ifp->idev->dev, 0, RTF_DEFAULT, true);
4854 if (!f6i)
4855 return -ENOENT;
4856
4857 prio = ifp->rt_priority ? : IP6_RT_PRIO_ADDRCONF;
4858 if (f6i->fib6_metric != prio) {
4859 /* delete old one */
4860 ip6_del_rt(dev_net(ifp->idev->dev), f6i, false);
4861
4862 /* add new one */
4863 addrconf_prefix_route(modify_peer ? &ifp->peer_addr : &ifp->addr,
4864 ifp->prefix_len,
4865 ifp->rt_priority, ifp->idev->dev,
4866 expires, flags, GFP_KERNEL);
4867 return 0;
4868 }
4869 if (f6i != net->ipv6.fib6_null_entry) {
4870 table = f6i->fib6_table;
4871 spin_lock_bh(&table->tb6_lock);
4872
4873 if (!(flags & RTF_EXPIRES)) {
4874 fib6_clean_expires(f6i);
4875 fib6_may_remove_gc_list(net, f6i);
4876 } else {
4877 fib6_set_expires(f6i, expires);
4878 fib6_add_gc_list(f6i);
4879 }
4880
4881 spin_unlock_bh(&table->tb6_lock);
4882 }
4883 fib6_info_release(f6i);
4884
4885 return 0;
4886 }
4887
inet6_addr_modify(struct net * net,struct inet6_ifaddr * ifp,struct ifa6_config * cfg,clock_t expires,u32 flags)4888 static int inet6_addr_modify(struct net *net, struct inet6_ifaddr *ifp,
4889 struct ifa6_config *cfg, clock_t expires,
4890 u32 flags)
4891 {
4892 bool was_managetempaddr;
4893 bool new_peer = false;
4894 bool had_prefixroute;
4895
4896 ASSERT_RTNL_NET(net);
4897
4898 if (cfg->ifa_flags & IFA_F_MANAGETEMPADDR &&
4899 (ifp->flags & IFA_F_TEMPORARY || ifp->prefix_len != 64))
4900 return -EINVAL;
4901
4902 if (!(ifp->flags & IFA_F_TENTATIVE) || ifp->flags & IFA_F_DADFAILED)
4903 cfg->ifa_flags &= ~IFA_F_OPTIMISTIC;
4904
4905 if (cfg->peer_pfx &&
4906 memcmp(&ifp->peer_addr, cfg->peer_pfx, sizeof(struct in6_addr))) {
4907 if (!ipv6_addr_any(&ifp->peer_addr))
4908 cleanup_prefix_route(ifp, expires, true, true);
4909 new_peer = true;
4910 }
4911
4912 spin_lock_bh(&ifp->lock);
4913 was_managetempaddr = ifp->flags & IFA_F_MANAGETEMPADDR;
4914 had_prefixroute = ifp->flags & IFA_F_PERMANENT &&
4915 !(ifp->flags & IFA_F_NOPREFIXROUTE);
4916 ifp->flags &= ~(IFA_F_DEPRECATED | IFA_F_PERMANENT | IFA_F_NODAD |
4917 IFA_F_HOMEADDRESS | IFA_F_MANAGETEMPADDR |
4918 IFA_F_NOPREFIXROUTE);
4919 ifp->flags |= cfg->ifa_flags;
4920 WRITE_ONCE(ifp->tstamp, jiffies);
4921 WRITE_ONCE(ifp->valid_lft, cfg->valid_lft);
4922 WRITE_ONCE(ifp->prefered_lft, cfg->preferred_lft);
4923 WRITE_ONCE(ifp->ifa_proto, cfg->ifa_proto);
4924
4925 if (cfg->rt_priority && cfg->rt_priority != ifp->rt_priority)
4926 WRITE_ONCE(ifp->rt_priority, cfg->rt_priority);
4927
4928 if (new_peer)
4929 ifp->peer_addr = *cfg->peer_pfx;
4930
4931 spin_unlock_bh(&ifp->lock);
4932 if (!(ifp->flags&IFA_F_TENTATIVE))
4933 ipv6_ifa_notify(0, ifp);
4934
4935 if (!(cfg->ifa_flags & IFA_F_NOPREFIXROUTE)) {
4936 int rc = -ENOENT;
4937
4938 if (had_prefixroute)
4939 rc = modify_prefix_route(net, ifp, expires, flags, false);
4940
4941 /* prefix route could have been deleted; if so restore it */
4942 if (rc == -ENOENT) {
4943 addrconf_prefix_route(&ifp->addr, ifp->prefix_len,
4944 ifp->rt_priority, ifp->idev->dev,
4945 expires, flags, GFP_KERNEL);
4946 }
4947
4948 if (had_prefixroute && !ipv6_addr_any(&ifp->peer_addr))
4949 rc = modify_prefix_route(net, ifp, expires, flags, true);
4950
4951 if (rc == -ENOENT && !ipv6_addr_any(&ifp->peer_addr)) {
4952 addrconf_prefix_route(&ifp->peer_addr, ifp->prefix_len,
4953 ifp->rt_priority, ifp->idev->dev,
4954 expires, flags, GFP_KERNEL);
4955 }
4956 } else if (had_prefixroute) {
4957 enum cleanup_prefix_rt_t action;
4958 unsigned long rt_expires;
4959
4960 write_lock_bh(&ifp->idev->lock);
4961 action = check_cleanup_prefix_route(ifp, &rt_expires);
4962 write_unlock_bh(&ifp->idev->lock);
4963
4964 if (action != CLEANUP_PREFIX_RT_NOP) {
4965 cleanup_prefix_route(ifp, rt_expires,
4966 action == CLEANUP_PREFIX_RT_DEL, false);
4967 }
4968 }
4969
4970 if (was_managetempaddr || ifp->flags & IFA_F_MANAGETEMPADDR) {
4971 if (was_managetempaddr && !(ifp->flags & IFA_F_MANAGETEMPADDR))
4972 delete_tempaddrs(ifp->idev, ifp);
4973 else
4974 manage_tempaddrs(ifp->idev, ifp, cfg->valid_lft,
4975 cfg->preferred_lft, !was_managetempaddr,
4976 jiffies);
4977 }
4978
4979 addrconf_verify_rtnl(net);
4980
4981 return 0;
4982 }
4983
4984 static int
inet6_rtm_newaddr(struct sk_buff * skb,struct nlmsghdr * nlh,struct netlink_ext_ack * extack)4985 inet6_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh,
4986 struct netlink_ext_ack *extack)
4987 {
4988 struct net *net = sock_net(skb->sk);
4989 struct nlattr *tb[IFA_MAX+1];
4990 struct in6_addr *peer_pfx;
4991 struct inet6_ifaddr *ifa;
4992 struct net_device *dev;
4993 struct inet6_dev *idev;
4994 struct ifa6_config cfg;
4995 struct ifaddrmsg *ifm;
4996 unsigned long timeout;
4997 clock_t expires;
4998 u32 flags;
4999 int err;
5000
5001 err = nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFA_MAX,
5002 ifa_ipv6_policy, extack);
5003 if (err < 0)
5004 return err;
5005
5006 memset(&cfg, 0, sizeof(cfg));
5007
5008 ifm = nlmsg_data(nlh);
5009 cfg.pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
5010 if (!cfg.pfx)
5011 return -EINVAL;
5012
5013 cfg.peer_pfx = peer_pfx;
5014 cfg.plen = ifm->ifa_prefixlen;
5015 if (tb[IFA_RT_PRIORITY])
5016 cfg.rt_priority = nla_get_u32(tb[IFA_RT_PRIORITY]);
5017
5018 if (tb[IFA_PROTO])
5019 cfg.ifa_proto = nla_get_u8(tb[IFA_PROTO]);
5020
5021 cfg.ifa_flags = nla_get_u32_default(tb[IFA_FLAGS], ifm->ifa_flags);
5022
5023 /* We ignore other flags so far. */
5024 cfg.ifa_flags &= IFA_F_NODAD | IFA_F_HOMEADDRESS |
5025 IFA_F_MANAGETEMPADDR | IFA_F_NOPREFIXROUTE |
5026 IFA_F_MCAUTOJOIN | IFA_F_OPTIMISTIC;
5027
5028 cfg.ifa_flags |= IFA_F_PERMANENT;
5029 cfg.valid_lft = INFINITY_LIFE_TIME;
5030 cfg.preferred_lft = INFINITY_LIFE_TIME;
5031 expires = 0;
5032 flags = 0;
5033
5034 if (tb[IFA_CACHEINFO]) {
5035 struct ifa_cacheinfo *ci;
5036
5037 ci = nla_data(tb[IFA_CACHEINFO]);
5038 cfg.valid_lft = ci->ifa_valid;
5039 cfg.preferred_lft = ci->ifa_prefered;
5040
5041 if (!cfg.valid_lft || cfg.preferred_lft > cfg.valid_lft) {
5042 NL_SET_ERR_MSG_MOD(extack, "address lifetime invalid");
5043 return -EINVAL;
5044 }
5045
5046 timeout = addrconf_timeout_fixup(cfg.valid_lft, HZ);
5047 if (addrconf_finite_timeout(timeout)) {
5048 cfg.ifa_flags &= ~IFA_F_PERMANENT;
5049 cfg.valid_lft = timeout;
5050 expires = jiffies_to_clock_t(timeout * HZ);
5051 flags = RTF_EXPIRES;
5052 }
5053
5054 timeout = addrconf_timeout_fixup(cfg.preferred_lft, HZ);
5055 if (addrconf_finite_timeout(timeout)) {
5056 if (timeout == 0)
5057 cfg.ifa_flags |= IFA_F_DEPRECATED;
5058
5059 cfg.preferred_lft = timeout;
5060 }
5061 }
5062
5063 rtnl_net_lock(net);
5064
5065 dev = __dev_get_by_index(net, ifm->ifa_index);
5066 if (!dev) {
5067 NL_SET_ERR_MSG_MOD(extack, "Unable to find the interface");
5068 err = -ENODEV;
5069 goto unlock_rtnl;
5070 }
5071
5072 netdev_lock_ops(dev);
5073 idev = ipv6_find_idev(dev);
5074 if (IS_ERR(idev)) {
5075 err = PTR_ERR(idev);
5076 goto unlock;
5077 }
5078
5079 if (!ipv6_allow_optimistic_dad(net, idev))
5080 cfg.ifa_flags &= ~IFA_F_OPTIMISTIC;
5081
5082 if (cfg.ifa_flags & IFA_F_NODAD &&
5083 cfg.ifa_flags & IFA_F_OPTIMISTIC) {
5084 NL_SET_ERR_MSG(extack, "IFA_F_NODAD and IFA_F_OPTIMISTIC are mutually exclusive");
5085 err = -EINVAL;
5086 goto unlock;
5087 }
5088
5089 ifa = ipv6_get_ifaddr(net, cfg.pfx, dev, 1);
5090 if (!ifa) {
5091 /*
5092 * It would be best to check for !NLM_F_CREATE here but
5093 * userspace already relies on not having to provide this.
5094 */
5095 err = inet6_addr_add(net, dev, &cfg, expires, flags, extack);
5096 goto unlock;
5097 }
5098
5099 if (nlh->nlmsg_flags & NLM_F_EXCL ||
5100 !(nlh->nlmsg_flags & NLM_F_REPLACE)) {
5101 NL_SET_ERR_MSG_MOD(extack, "address already assigned");
5102 err = -EEXIST;
5103 } else {
5104 err = inet6_addr_modify(net, ifa, &cfg, expires, flags);
5105 }
5106
5107 in6_ifa_put(ifa);
5108 unlock:
5109 netdev_unlock_ops(dev);
5110 unlock_rtnl:
5111 rtnl_net_unlock(net);
5112
5113 return err;
5114 }
5115
put_ifaddrmsg(struct nlmsghdr * nlh,u8 prefixlen,u32 flags,u8 scope,int ifindex)5116 static void put_ifaddrmsg(struct nlmsghdr *nlh, u8 prefixlen, u32 flags,
5117 u8 scope, int ifindex)
5118 {
5119 struct ifaddrmsg *ifm;
5120
5121 ifm = nlmsg_data(nlh);
5122 ifm->ifa_family = AF_INET6;
5123 ifm->ifa_prefixlen = prefixlen;
5124 ifm->ifa_flags = flags;
5125 ifm->ifa_scope = scope;
5126 ifm->ifa_index = ifindex;
5127 }
5128
put_cacheinfo(struct sk_buff * skb,unsigned long cstamp,unsigned long tstamp,u32 preferred,u32 valid)5129 static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
5130 unsigned long tstamp, u32 preferred, u32 valid)
5131 {
5132 struct ifa_cacheinfo ci;
5133
5134 ci.cstamp = cstamp_delta(cstamp);
5135 ci.tstamp = cstamp_delta(tstamp);
5136 ci.ifa_prefered = preferred;
5137 ci.ifa_valid = valid;
5138
5139 return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
5140 }
5141
rt_scope(int ifa_scope)5142 static inline int rt_scope(int ifa_scope)
5143 {
5144 if (ifa_scope & IFA_HOST)
5145 return RT_SCOPE_HOST;
5146 else if (ifa_scope & IFA_LINK)
5147 return RT_SCOPE_LINK;
5148 else if (ifa_scope & IFA_SITE)
5149 return RT_SCOPE_SITE;
5150 else
5151 return RT_SCOPE_UNIVERSE;
5152 }
5153
inet6_ifaddr_msgsize(void)5154 static inline int inet6_ifaddr_msgsize(void)
5155 {
5156 return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
5157 + nla_total_size(16) /* IFA_LOCAL */
5158 + nla_total_size(16) /* IFA_ADDRESS */
5159 + nla_total_size(sizeof(struct ifa_cacheinfo))
5160 + nla_total_size(4) /* IFA_FLAGS */
5161 + nla_total_size(1) /* IFA_PROTO */
5162 + nla_total_size(4) /* IFA_RT_PRIORITY */;
5163 }
5164
inet6_fill_ifaddr(struct sk_buff * skb,const struct inet6_ifaddr * ifa,struct inet6_fill_args * args)5165 static int inet6_fill_ifaddr(struct sk_buff *skb,
5166 const struct inet6_ifaddr *ifa,
5167 struct inet6_fill_args *args)
5168 {
5169 struct nlmsghdr *nlh;
5170 u32 preferred, valid;
5171 u32 flags, priority;
5172 u8 proto;
5173
5174 nlh = nlmsg_put(skb, args->portid, args->seq, args->event,
5175 sizeof(struct ifaddrmsg), args->flags);
5176 if (!nlh)
5177 return -EMSGSIZE;
5178
5179 flags = READ_ONCE(ifa->flags);
5180 put_ifaddrmsg(nlh, ifa->prefix_len, ifa->flags, rt_scope(ifa->scope),
5181 ifa->idev->dev->ifindex);
5182
5183 if (args->netnsid >= 0 &&
5184 nla_put_s32(skb, IFA_TARGET_NETNSID, args->netnsid))
5185 goto error;
5186
5187 preferred = READ_ONCE(ifa->prefered_lft);
5188 valid = READ_ONCE(ifa->valid_lft);
5189
5190 if (!((flags & IFA_F_PERMANENT) &&
5191 (preferred == INFINITY_LIFE_TIME))) {
5192 if (preferred != INFINITY_LIFE_TIME) {
5193 long tval = (jiffies - READ_ONCE(ifa->tstamp)) / HZ;
5194
5195 if (preferred > tval)
5196 preferred -= tval;
5197 else
5198 preferred = 0;
5199 if (valid != INFINITY_LIFE_TIME) {
5200 if (valid > tval)
5201 valid -= tval;
5202 else
5203 valid = 0;
5204 }
5205 }
5206 } else {
5207 preferred = INFINITY_LIFE_TIME;
5208 valid = INFINITY_LIFE_TIME;
5209 }
5210
5211 if (!ipv6_addr_any(&ifa->peer_addr)) {
5212 if (nla_put_in6_addr(skb, IFA_LOCAL, &ifa->addr) < 0 ||
5213 nla_put_in6_addr(skb, IFA_ADDRESS, &ifa->peer_addr) < 0)
5214 goto error;
5215 } else {
5216 if (nla_put_in6_addr(skb, IFA_ADDRESS, &ifa->addr) < 0)
5217 goto error;
5218 }
5219
5220 priority = READ_ONCE(ifa->rt_priority);
5221 if (priority && nla_put_u32(skb, IFA_RT_PRIORITY, priority))
5222 goto error;
5223
5224 if (put_cacheinfo(skb, ifa->cstamp, READ_ONCE(ifa->tstamp),
5225 preferred, valid) < 0)
5226 goto error;
5227
5228 if (nla_put_u32(skb, IFA_FLAGS, flags) < 0)
5229 goto error;
5230
5231 proto = READ_ONCE(ifa->ifa_proto);
5232 if (proto && nla_put_u8(skb, IFA_PROTO, proto))
5233 goto error;
5234
5235 nlmsg_end(skb, nlh);
5236 return 0;
5237
5238 error:
5239 nlmsg_cancel(skb, nlh);
5240 return -EMSGSIZE;
5241 }
5242
inet6_fill_ifmcaddr(struct sk_buff * skb,const struct ifmcaddr6 * ifmca,struct inet6_fill_args * args)5243 int inet6_fill_ifmcaddr(struct sk_buff *skb,
5244 const struct ifmcaddr6 *ifmca,
5245 struct inet6_fill_args *args)
5246 {
5247 int ifindex = ifmca->idev->dev->ifindex;
5248 u8 scope = RT_SCOPE_UNIVERSE;
5249 struct nlmsghdr *nlh;
5250
5251 if (!args->force_rt_scope_universe &&
5252 ipv6_addr_scope(&ifmca->mca_addr) & IFA_SITE)
5253 scope = RT_SCOPE_SITE;
5254
5255 nlh = nlmsg_put(skb, args->portid, args->seq, args->event,
5256 sizeof(struct ifaddrmsg), args->flags);
5257 if (!nlh)
5258 return -EMSGSIZE;
5259
5260 if (args->netnsid >= 0 &&
5261 nla_put_s32(skb, IFA_TARGET_NETNSID, args->netnsid)) {
5262 nlmsg_cancel(skb, nlh);
5263 return -EMSGSIZE;
5264 }
5265
5266 put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
5267 if (nla_put_in6_addr(skb, IFA_MULTICAST, &ifmca->mca_addr) < 0 ||
5268 nla_put_u32(skb, IFA_MC_USERS, READ_ONCE(ifmca->mca_users)) < 0 ||
5269 put_cacheinfo(skb, ifmca->mca_cstamp, READ_ONCE(ifmca->mca_tstamp),
5270 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
5271 nlmsg_cancel(skb, nlh);
5272 return -EMSGSIZE;
5273 }
5274
5275 nlmsg_end(skb, nlh);
5276 return 0;
5277 }
5278
inet6_fill_ifacaddr(struct sk_buff * skb,const struct ifacaddr6 * ifaca,struct inet6_fill_args * args)5279 int inet6_fill_ifacaddr(struct sk_buff *skb,
5280 const struct ifacaddr6 *ifaca,
5281 struct inet6_fill_args *args)
5282 {
5283 struct net_device *dev = fib6_info_nh_dev(ifaca->aca_rt);
5284 int ifindex = dev ? dev->ifindex : 1;
5285 u8 scope = RT_SCOPE_UNIVERSE;
5286 struct nlmsghdr *nlh;
5287
5288 if (ipv6_addr_scope(&ifaca->aca_addr) & IFA_SITE)
5289 scope = RT_SCOPE_SITE;
5290
5291 nlh = nlmsg_put(skb, args->portid, args->seq, args->event,
5292 sizeof(struct ifaddrmsg), args->flags);
5293 if (!nlh)
5294 return -EMSGSIZE;
5295
5296 if (args->netnsid >= 0 &&
5297 nla_put_s32(skb, IFA_TARGET_NETNSID, args->netnsid)) {
5298 nlmsg_cancel(skb, nlh);
5299 return -EMSGSIZE;
5300 }
5301
5302 put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
5303 if (nla_put_in6_addr(skb, IFA_ANYCAST, &ifaca->aca_addr) < 0 ||
5304 put_cacheinfo(skb, ifaca->aca_cstamp, READ_ONCE(ifaca->aca_tstamp),
5305 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
5306 nlmsg_cancel(skb, nlh);
5307 return -EMSGSIZE;
5308 }
5309
5310 nlmsg_end(skb, nlh);
5311 return 0;
5312 }
5313
5314 /* 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)5315 static int in6_dump_addrs(const struct inet6_dev *idev, struct sk_buff *skb,
5316 struct netlink_callback *cb, int *s_ip_idx,
5317 struct inet6_fill_args *fillargs)
5318 {
5319 const struct ifmcaddr6 *ifmca;
5320 const struct ifacaddr6 *ifaca;
5321 int ip_idx = 0;
5322 int err = 0;
5323
5324 switch (fillargs->type) {
5325 case UNICAST_ADDR: {
5326 const struct inet6_ifaddr *ifa;
5327 fillargs->event = RTM_NEWADDR;
5328
5329 /* unicast address incl. temp addr */
5330 list_for_each_entry_rcu(ifa, &idev->addr_list, if_list) {
5331 if (ip_idx < *s_ip_idx)
5332 goto next;
5333 err = inet6_fill_ifaddr(skb, ifa, fillargs);
5334 if (err < 0)
5335 break;
5336 nl_dump_check_consistent(cb, nlmsg_hdr(skb));
5337 next:
5338 ip_idx++;
5339 }
5340 break;
5341 }
5342 case MULTICAST_ADDR:
5343 fillargs->event = RTM_GETMULTICAST;
5344
5345 /* multicast address */
5346 for (ifmca = rcu_dereference(idev->mc_list);
5347 ifmca;
5348 ifmca = rcu_dereference(ifmca->next), ip_idx++) {
5349 if (ip_idx < *s_ip_idx)
5350 continue;
5351 err = inet6_fill_ifmcaddr(skb, ifmca, fillargs);
5352 if (err < 0)
5353 break;
5354 }
5355 break;
5356 case ANYCAST_ADDR:
5357 fillargs->event = RTM_GETANYCAST;
5358 /* anycast address */
5359 for (ifaca = rcu_dereference(idev->ac_list); ifaca;
5360 ifaca = rcu_dereference(ifaca->aca_next), ip_idx++) {
5361 if (ip_idx < *s_ip_idx)
5362 continue;
5363 err = inet6_fill_ifacaddr(skb, ifaca, fillargs);
5364 if (err < 0)
5365 break;
5366 }
5367 break;
5368 default:
5369 break;
5370 }
5371 *s_ip_idx = err ? ip_idx : 0;
5372 return err;
5373 }
5374
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)5375 static int inet6_valid_dump_ifaddr_req(const struct nlmsghdr *nlh,
5376 struct inet6_fill_args *fillargs,
5377 struct net **tgt_net, struct sock *sk,
5378 struct netlink_callback *cb)
5379 {
5380 struct netlink_ext_ack *extack = cb->extack;
5381 struct nlattr *tb[IFA_MAX+1];
5382 struct ifaddrmsg *ifm;
5383 int err, i;
5384
5385 ifm = nlmsg_payload(nlh, sizeof(*ifm));
5386 if (!ifm) {
5387 NL_SET_ERR_MSG_MOD(extack, "Invalid header for address dump request");
5388 return -EINVAL;
5389 }
5390
5391 if (ifm->ifa_prefixlen || ifm->ifa_flags || ifm->ifa_scope) {
5392 NL_SET_ERR_MSG_MOD(extack, "Invalid values in header for address dump request");
5393 return -EINVAL;
5394 }
5395
5396 fillargs->ifindex = ifm->ifa_index;
5397 if (fillargs->ifindex) {
5398 cb->answer_flags |= NLM_F_DUMP_FILTERED;
5399 fillargs->flags |= NLM_F_DUMP_FILTERED;
5400 }
5401
5402 err = nlmsg_parse_deprecated_strict(nlh, sizeof(*ifm), tb, IFA_MAX,
5403 ifa_ipv6_policy, extack);
5404 if (err < 0)
5405 return err;
5406
5407 for (i = 0; i <= IFA_MAX; ++i) {
5408 if (!tb[i])
5409 continue;
5410
5411 if (i == IFA_TARGET_NETNSID) {
5412 struct net *net;
5413
5414 fillargs->netnsid = nla_get_s32(tb[i]);
5415 net = rtnl_get_net_ns_capable(sk, fillargs->netnsid);
5416 if (IS_ERR(net)) {
5417 fillargs->netnsid = -1;
5418 NL_SET_ERR_MSG_MOD(extack, "Invalid target network namespace id");
5419 return PTR_ERR(net);
5420 }
5421 *tgt_net = net;
5422 } else {
5423 NL_SET_ERR_MSG_MOD(extack, "Unsupported attribute in dump request");
5424 return -EINVAL;
5425 }
5426 }
5427
5428 return 0;
5429 }
5430
inet6_dump_addr(struct sk_buff * skb,struct netlink_callback * cb,enum addr_type_t type)5431 static int inet6_dump_addr(struct sk_buff *skb, struct netlink_callback *cb,
5432 enum addr_type_t type)
5433 {
5434 struct net *tgt_net = sock_net(skb->sk);
5435 const struct nlmsghdr *nlh = cb->nlh;
5436 struct inet6_fill_args fillargs = {
5437 .portid = NETLINK_CB(cb->skb).portid,
5438 .seq = cb->nlh->nlmsg_seq,
5439 .flags = NLM_F_MULTI,
5440 .netnsid = -1,
5441 .type = type,
5442 .force_rt_scope_universe = false,
5443 };
5444 struct {
5445 unsigned long ifindex;
5446 int ip_idx;
5447 } *ctx = (void *)cb->ctx;
5448 struct net_device *dev;
5449 struct inet6_dev *idev;
5450 int err = 0;
5451
5452 rcu_read_lock();
5453 if (cb->strict_check) {
5454 err = inet6_valid_dump_ifaddr_req(nlh, &fillargs, &tgt_net,
5455 skb->sk, cb);
5456 if (err < 0)
5457 goto done;
5458
5459 err = 0;
5460 if (fillargs.ifindex) {
5461 dev = dev_get_by_index_rcu(tgt_net, fillargs.ifindex);
5462 if (!dev) {
5463 err = -ENODEV;
5464 goto done;
5465 }
5466 idev = __in6_dev_get(dev);
5467 if (idev)
5468 err = in6_dump_addrs(idev, skb, cb,
5469 &ctx->ip_idx,
5470 &fillargs);
5471 goto done;
5472 }
5473 }
5474
5475 cb->seq = inet6_base_seq(tgt_net);
5476 for_each_netdev_dump(tgt_net, dev, ctx->ifindex) {
5477 idev = __in6_dev_get(dev);
5478 if (!idev)
5479 continue;
5480 err = in6_dump_addrs(idev, skb, cb, &ctx->ip_idx,
5481 &fillargs);
5482 if (err < 0)
5483 goto done;
5484 }
5485 done:
5486 rcu_read_unlock();
5487 if (fillargs.netnsid >= 0)
5488 put_net(tgt_net);
5489
5490 return err;
5491 }
5492
inet6_dump_ifaddr(struct sk_buff * skb,struct netlink_callback * cb)5493 static int inet6_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
5494 {
5495 enum addr_type_t type = UNICAST_ADDR;
5496
5497 return inet6_dump_addr(skb, cb, type);
5498 }
5499
inet6_dump_ifmcaddr(struct sk_buff * skb,struct netlink_callback * cb)5500 static int inet6_dump_ifmcaddr(struct sk_buff *skb, struct netlink_callback *cb)
5501 {
5502 enum addr_type_t type = MULTICAST_ADDR;
5503
5504 return inet6_dump_addr(skb, cb, type);
5505 }
5506
5507
inet6_dump_ifacaddr(struct sk_buff * skb,struct netlink_callback * cb)5508 static int inet6_dump_ifacaddr(struct sk_buff *skb, struct netlink_callback *cb)
5509 {
5510 enum addr_type_t type = ANYCAST_ADDR;
5511
5512 return inet6_dump_addr(skb, cb, type);
5513 }
5514
inet6_rtm_valid_getaddr_req(struct sk_buff * skb,const struct nlmsghdr * nlh,struct nlattr ** tb,struct netlink_ext_ack * extack)5515 static int inet6_rtm_valid_getaddr_req(struct sk_buff *skb,
5516 const struct nlmsghdr *nlh,
5517 struct nlattr **tb,
5518 struct netlink_ext_ack *extack)
5519 {
5520 struct ifaddrmsg *ifm;
5521 int i, err;
5522
5523 ifm = nlmsg_payload(nlh, sizeof(*ifm));
5524 if (!ifm) {
5525 NL_SET_ERR_MSG_MOD(extack, "Invalid header for get address request");
5526 return -EINVAL;
5527 }
5528
5529 if (!netlink_strict_get_check(skb))
5530 return nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFA_MAX,
5531 ifa_ipv6_policy, extack);
5532
5533 if (ifm->ifa_prefixlen || ifm->ifa_flags || ifm->ifa_scope) {
5534 NL_SET_ERR_MSG_MOD(extack, "Invalid values in header for get address request");
5535 return -EINVAL;
5536 }
5537
5538 err = nlmsg_parse_deprecated_strict(nlh, sizeof(*ifm), tb, IFA_MAX,
5539 ifa_ipv6_policy, extack);
5540 if (err)
5541 return err;
5542
5543 for (i = 0; i <= IFA_MAX; i++) {
5544 if (!tb[i])
5545 continue;
5546
5547 switch (i) {
5548 case IFA_TARGET_NETNSID:
5549 case IFA_ADDRESS:
5550 case IFA_LOCAL:
5551 break;
5552 default:
5553 NL_SET_ERR_MSG_MOD(extack, "Unsupported attribute in get address request");
5554 return -EINVAL;
5555 }
5556 }
5557
5558 return 0;
5559 }
5560
inet6_rtm_getaddr(struct sk_buff * in_skb,struct nlmsghdr * nlh,struct netlink_ext_ack * extack)5561 static int inet6_rtm_getaddr(struct sk_buff *in_skb, struct nlmsghdr *nlh,
5562 struct netlink_ext_ack *extack)
5563 {
5564 struct net *tgt_net = sock_net(in_skb->sk);
5565 struct inet6_fill_args fillargs = {
5566 .portid = NETLINK_CB(in_skb).portid,
5567 .seq = nlh->nlmsg_seq,
5568 .event = RTM_NEWADDR,
5569 .flags = 0,
5570 .netnsid = -1,
5571 .force_rt_scope_universe = false,
5572 };
5573 struct ifaddrmsg *ifm;
5574 struct nlattr *tb[IFA_MAX+1];
5575 struct in6_addr *addr = NULL, *peer;
5576 struct net_device *dev = NULL;
5577 struct inet6_ifaddr *ifa;
5578 struct sk_buff *skb;
5579 int err;
5580
5581 err = inet6_rtm_valid_getaddr_req(in_skb, nlh, tb, extack);
5582 if (err < 0)
5583 return err;
5584
5585 if (tb[IFA_TARGET_NETNSID]) {
5586 fillargs.netnsid = nla_get_s32(tb[IFA_TARGET_NETNSID]);
5587
5588 tgt_net = rtnl_get_net_ns_capable(NETLINK_CB(in_skb).sk,
5589 fillargs.netnsid);
5590 if (IS_ERR(tgt_net))
5591 return PTR_ERR(tgt_net);
5592 }
5593
5594 addr = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer);
5595 if (!addr) {
5596 err = -EINVAL;
5597 goto errout;
5598 }
5599 ifm = nlmsg_data(nlh);
5600 if (ifm->ifa_index)
5601 dev = dev_get_by_index(tgt_net, ifm->ifa_index);
5602
5603 ifa = ipv6_get_ifaddr(tgt_net, addr, dev, 1);
5604 if (!ifa) {
5605 err = -EADDRNOTAVAIL;
5606 goto errout;
5607 }
5608
5609 skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_KERNEL);
5610 if (!skb) {
5611 err = -ENOBUFS;
5612 goto errout_ifa;
5613 }
5614
5615 err = inet6_fill_ifaddr(skb, ifa, &fillargs);
5616 if (err < 0) {
5617 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
5618 WARN_ON(err == -EMSGSIZE);
5619 kfree_skb(skb);
5620 goto errout_ifa;
5621 }
5622 err = rtnl_unicast(skb, tgt_net, NETLINK_CB(in_skb).portid);
5623 errout_ifa:
5624 in6_ifa_put(ifa);
5625 errout:
5626 dev_put(dev);
5627 if (fillargs.netnsid >= 0)
5628 put_net(tgt_net);
5629
5630 return err;
5631 }
5632
inet6_ifa_notify(int event,struct inet6_ifaddr * ifa)5633 static void inet6_ifa_notify(int event, struct inet6_ifaddr *ifa)
5634 {
5635 struct sk_buff *skb;
5636 struct net *net = dev_net(ifa->idev->dev);
5637 struct inet6_fill_args fillargs = {
5638 .portid = 0,
5639 .seq = 0,
5640 .event = event,
5641 .flags = 0,
5642 .netnsid = -1,
5643 .force_rt_scope_universe = false,
5644 };
5645 int err = -ENOBUFS;
5646
5647 skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_ATOMIC);
5648 if (!skb)
5649 goto errout;
5650
5651 err = inet6_fill_ifaddr(skb, ifa, &fillargs);
5652 if (err < 0) {
5653 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
5654 WARN_ON(err == -EMSGSIZE);
5655 kfree_skb(skb);
5656 goto errout;
5657 }
5658 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
5659 return;
5660 errout:
5661 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
5662 }
5663
ipv6_store_devconf(const struct ipv6_devconf * cnf,__s32 * array,int bytes)5664 static void ipv6_store_devconf(const struct ipv6_devconf *cnf,
5665 __s32 *array, int bytes)
5666 {
5667 BUG_ON(bytes < (DEVCONF_MAX * 4));
5668
5669 memset(array, 0, bytes);
5670 array[DEVCONF_FORWARDING] = READ_ONCE(cnf->forwarding);
5671 array[DEVCONF_HOPLIMIT] = READ_ONCE(cnf->hop_limit);
5672 array[DEVCONF_MTU6] = READ_ONCE(cnf->mtu6);
5673 array[DEVCONF_ACCEPT_RA] = READ_ONCE(cnf->accept_ra);
5674 array[DEVCONF_ACCEPT_REDIRECTS] = READ_ONCE(cnf->accept_redirects);
5675 array[DEVCONF_AUTOCONF] = READ_ONCE(cnf->autoconf);
5676 array[DEVCONF_DAD_TRANSMITS] = READ_ONCE(cnf->dad_transmits);
5677 array[DEVCONF_RTR_SOLICITS] = READ_ONCE(cnf->rtr_solicits);
5678 array[DEVCONF_RTR_SOLICIT_INTERVAL] =
5679 jiffies_to_msecs(READ_ONCE(cnf->rtr_solicit_interval));
5680 array[DEVCONF_RTR_SOLICIT_MAX_INTERVAL] =
5681 jiffies_to_msecs(READ_ONCE(cnf->rtr_solicit_max_interval));
5682 array[DEVCONF_RTR_SOLICIT_DELAY] =
5683 jiffies_to_msecs(READ_ONCE(cnf->rtr_solicit_delay));
5684 array[DEVCONF_FORCE_MLD_VERSION] = READ_ONCE(cnf->force_mld_version);
5685 array[DEVCONF_MLDV1_UNSOLICITED_REPORT_INTERVAL] =
5686 jiffies_to_msecs(READ_ONCE(cnf->mldv1_unsolicited_report_interval));
5687 array[DEVCONF_MLDV2_UNSOLICITED_REPORT_INTERVAL] =
5688 jiffies_to_msecs(READ_ONCE(cnf->mldv2_unsolicited_report_interval));
5689 array[DEVCONF_USE_TEMPADDR] = READ_ONCE(cnf->use_tempaddr);
5690 array[DEVCONF_TEMP_VALID_LFT] = READ_ONCE(cnf->temp_valid_lft);
5691 array[DEVCONF_TEMP_PREFERED_LFT] = READ_ONCE(cnf->temp_prefered_lft);
5692 array[DEVCONF_REGEN_MAX_RETRY] = READ_ONCE(cnf->regen_max_retry);
5693 array[DEVCONF_MAX_DESYNC_FACTOR] = READ_ONCE(cnf->max_desync_factor);
5694 array[DEVCONF_MAX_ADDRESSES] = READ_ONCE(cnf->max_addresses);
5695 array[DEVCONF_ACCEPT_RA_DEFRTR] = READ_ONCE(cnf->accept_ra_defrtr);
5696 array[DEVCONF_RA_DEFRTR_METRIC] = READ_ONCE(cnf->ra_defrtr_metric);
5697 array[DEVCONF_ACCEPT_RA_MIN_HOP_LIMIT] =
5698 READ_ONCE(cnf->accept_ra_min_hop_limit);
5699 array[DEVCONF_ACCEPT_RA_PINFO] = READ_ONCE(cnf->accept_ra_pinfo);
5700 #ifdef CONFIG_IPV6_ROUTER_PREF
5701 array[DEVCONF_ACCEPT_RA_RTR_PREF] = READ_ONCE(cnf->accept_ra_rtr_pref);
5702 array[DEVCONF_RTR_PROBE_INTERVAL] =
5703 jiffies_to_msecs(READ_ONCE(cnf->rtr_probe_interval));
5704 #ifdef CONFIG_IPV6_ROUTE_INFO
5705 array[DEVCONF_ACCEPT_RA_RT_INFO_MIN_PLEN] =
5706 READ_ONCE(cnf->accept_ra_rt_info_min_plen);
5707 array[DEVCONF_ACCEPT_RA_RT_INFO_MAX_PLEN] =
5708 READ_ONCE(cnf->accept_ra_rt_info_max_plen);
5709 #endif
5710 #endif
5711 array[DEVCONF_PROXY_NDP] = READ_ONCE(cnf->proxy_ndp);
5712 array[DEVCONF_ACCEPT_SOURCE_ROUTE] =
5713 READ_ONCE(cnf->accept_source_route);
5714 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
5715 array[DEVCONF_OPTIMISTIC_DAD] = READ_ONCE(cnf->optimistic_dad);
5716 array[DEVCONF_USE_OPTIMISTIC] = READ_ONCE(cnf->use_optimistic);
5717 #endif
5718 #ifdef CONFIG_IPV6_MROUTE
5719 array[DEVCONF_MC_FORWARDING] = atomic_read(&cnf->mc_forwarding);
5720 #endif
5721 array[DEVCONF_DISABLE_IPV6] = READ_ONCE(cnf->disable_ipv6);
5722 array[DEVCONF_ACCEPT_DAD] = READ_ONCE(cnf->accept_dad);
5723 array[DEVCONF_FORCE_TLLAO] = READ_ONCE(cnf->force_tllao);
5724 array[DEVCONF_NDISC_NOTIFY] = READ_ONCE(cnf->ndisc_notify);
5725 array[DEVCONF_SUPPRESS_FRAG_NDISC] =
5726 READ_ONCE(cnf->suppress_frag_ndisc);
5727 array[DEVCONF_ACCEPT_RA_FROM_LOCAL] =
5728 READ_ONCE(cnf->accept_ra_from_local);
5729 array[DEVCONF_ACCEPT_RA_MTU] = READ_ONCE(cnf->accept_ra_mtu);
5730 array[DEVCONF_IGNORE_ROUTES_WITH_LINKDOWN] =
5731 READ_ONCE(cnf->ignore_routes_with_linkdown);
5732 /* we omit DEVCONF_STABLE_SECRET for now */
5733 array[DEVCONF_USE_OIF_ADDRS_ONLY] = READ_ONCE(cnf->use_oif_addrs_only);
5734 array[DEVCONF_DROP_UNICAST_IN_L2_MULTICAST] =
5735 READ_ONCE(cnf->drop_unicast_in_l2_multicast);
5736 array[DEVCONF_DROP_UNSOLICITED_NA] = READ_ONCE(cnf->drop_unsolicited_na);
5737 array[DEVCONF_KEEP_ADDR_ON_DOWN] = READ_ONCE(cnf->keep_addr_on_down);
5738 array[DEVCONF_SEG6_ENABLED] = READ_ONCE(cnf->seg6_enabled);
5739 #ifdef CONFIG_IPV6_SEG6_HMAC
5740 array[DEVCONF_SEG6_REQUIRE_HMAC] = READ_ONCE(cnf->seg6_require_hmac);
5741 #endif
5742 array[DEVCONF_ENHANCED_DAD] = READ_ONCE(cnf->enhanced_dad);
5743 array[DEVCONF_ADDR_GEN_MODE] = READ_ONCE(cnf->addr_gen_mode);
5744 array[DEVCONF_DISABLE_POLICY] = READ_ONCE(cnf->disable_policy);
5745 array[DEVCONF_NDISC_TCLASS] = READ_ONCE(cnf->ndisc_tclass);
5746 array[DEVCONF_RPL_SEG_ENABLED] = READ_ONCE(cnf->rpl_seg_enabled);
5747 array[DEVCONF_IOAM6_ENABLED] = READ_ONCE(cnf->ioam6_enabled);
5748 array[DEVCONF_IOAM6_ID] = READ_ONCE(cnf->ioam6_id);
5749 array[DEVCONF_IOAM6_ID_WIDE] = READ_ONCE(cnf->ioam6_id_wide);
5750 array[DEVCONF_NDISC_EVICT_NOCARRIER] =
5751 READ_ONCE(cnf->ndisc_evict_nocarrier);
5752 array[DEVCONF_ACCEPT_UNTRACKED_NA] =
5753 READ_ONCE(cnf->accept_untracked_na);
5754 array[DEVCONF_ACCEPT_RA_MIN_LFT] = READ_ONCE(cnf->accept_ra_min_lft);
5755 array[DEVCONF_FORCE_FORWARDING] = READ_ONCE(cnf->force_forwarding);
5756 }
5757
inet6_ifla6_size(void)5758 static inline size_t inet6_ifla6_size(void)
5759 {
5760 return nla_total_size(4) /* IFLA_INET6_FLAGS */
5761 + nla_total_size(sizeof(struct ifla_cacheinfo))
5762 + nla_total_size(DEVCONF_MAX * 4) /* IFLA_INET6_CONF */
5763 + nla_total_size(IPSTATS_MIB_MAX * 8) /* IFLA_INET6_STATS */
5764 + nla_total_size(ICMP6_MIB_MAX * 8) /* IFLA_INET6_ICMP6STATS */
5765 + nla_total_size(sizeof(struct in6_addr)) /* IFLA_INET6_TOKEN */
5766 + nla_total_size(1) /* IFLA_INET6_ADDR_GEN_MODE */
5767 + nla_total_size(4) /* IFLA_INET6_RA_MTU */
5768 + 0;
5769 }
5770
inet6_if_nlmsg_size(void)5771 static inline size_t inet6_if_nlmsg_size(void)
5772 {
5773 return NLMSG_ALIGN(sizeof(struct ifinfomsg))
5774 + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
5775 + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
5776 + nla_total_size(4) /* IFLA_MTU */
5777 + nla_total_size(4) /* IFLA_LINK */
5778 + nla_total_size(1) /* IFLA_OPERSTATE */
5779 + nla_total_size(inet6_ifla6_size()); /* IFLA_PROTINFO */
5780 }
5781
__snmp6_fill_statsdev(u64 * stats,atomic_long_t * mib,int bytes)5782 static inline void __snmp6_fill_statsdev(u64 *stats, atomic_long_t *mib,
5783 int bytes)
5784 {
5785 int i;
5786 int pad = bytes - sizeof(u64) * ICMP6_MIB_MAX;
5787 BUG_ON(pad < 0);
5788
5789 /* Use put_unaligned() because stats may not be aligned for u64. */
5790 put_unaligned(ICMP6_MIB_MAX, &stats[0]);
5791 for (i = 1; i < ICMP6_MIB_MAX; i++)
5792 put_unaligned(atomic_long_read(&mib[i]), &stats[i]);
5793
5794 memset(&stats[ICMP6_MIB_MAX], 0, pad);
5795 }
5796
__snmp6_fill_stats64(u64 * stats,void __percpu * mib,int bytes,size_t syncpoff)5797 static inline void __snmp6_fill_stats64(u64 *stats, void __percpu *mib,
5798 int bytes, size_t syncpoff)
5799 {
5800 int i, c;
5801 u64 buff[IPSTATS_MIB_MAX];
5802 int pad = bytes - sizeof(u64) * IPSTATS_MIB_MAX;
5803
5804 BUG_ON(pad < 0);
5805
5806 memset(buff, 0, sizeof(buff));
5807 buff[0] = IPSTATS_MIB_MAX;
5808
5809 for_each_possible_cpu(c) {
5810 for (i = 1; i < IPSTATS_MIB_MAX; i++)
5811 buff[i] += snmp_get_cpu_field64(mib, c, i, syncpoff);
5812 }
5813
5814 memcpy(stats, buff, IPSTATS_MIB_MAX * sizeof(u64));
5815 memset(&stats[IPSTATS_MIB_MAX], 0, pad);
5816 }
5817
snmp6_fill_stats(u64 * stats,struct inet6_dev * idev,int attrtype,int bytes)5818 static void snmp6_fill_stats(u64 *stats, struct inet6_dev *idev, int attrtype,
5819 int bytes)
5820 {
5821 switch (attrtype) {
5822 case IFLA_INET6_STATS:
5823 __snmp6_fill_stats64(stats, idev->stats.ipv6, bytes,
5824 offsetof(struct ipstats_mib, syncp));
5825 break;
5826 case IFLA_INET6_ICMP6STATS:
5827 __snmp6_fill_statsdev(stats, idev->stats.icmpv6dev->mibs, bytes);
5828 break;
5829 }
5830 }
5831
inet6_fill_ifla6_stats_attrs(struct sk_buff * skb,struct inet6_dev * idev)5832 static int inet6_fill_ifla6_stats_attrs(struct sk_buff *skb,
5833 struct inet6_dev *idev)
5834 {
5835 struct nlattr *nla;
5836
5837 nla = nla_reserve(skb, IFLA_INET6_STATS, IPSTATS_MIB_MAX * sizeof(u64));
5838 if (!nla)
5839 goto nla_put_failure;
5840 snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_STATS, nla_len(nla));
5841
5842 nla = nla_reserve(skb, IFLA_INET6_ICMP6STATS, ICMP6_MIB_MAX * sizeof(u64));
5843 if (!nla)
5844 goto nla_put_failure;
5845 snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_ICMP6STATS, nla_len(nla));
5846
5847 return 0;
5848
5849 nla_put_failure:
5850 return -EMSGSIZE;
5851 }
5852
inet6_fill_ifla6_attrs(struct sk_buff * skb,struct inet6_dev * idev,u32 ext_filter_mask)5853 static int inet6_fill_ifla6_attrs(struct sk_buff *skb, struct inet6_dev *idev,
5854 u32 ext_filter_mask)
5855 {
5856 struct ifla_cacheinfo ci;
5857 struct nlattr *nla;
5858 u32 ra_mtu;
5859
5860 if (nla_put_u32(skb, IFLA_INET6_FLAGS, READ_ONCE(idev->if_flags)))
5861 goto nla_put_failure;
5862 ci.max_reasm_len = IPV6_MAXPLEN;
5863 ci.tstamp = cstamp_delta(READ_ONCE(idev->tstamp));
5864 ci.reachable_time = jiffies_to_msecs(idev->nd_parms->reachable_time);
5865 ci.retrans_time = jiffies_to_msecs(NEIGH_VAR(idev->nd_parms, RETRANS_TIME));
5866 if (nla_put(skb, IFLA_INET6_CACHEINFO, sizeof(ci), &ci))
5867 goto nla_put_failure;
5868 nla = nla_reserve(skb, IFLA_INET6_CONF, DEVCONF_MAX * sizeof(s32));
5869 if (!nla)
5870 goto nla_put_failure;
5871 ipv6_store_devconf(&idev->cnf, nla_data(nla), nla_len(nla));
5872
5873 /* XXX - MC not implemented */
5874
5875 if (!(ext_filter_mask & RTEXT_FILTER_SKIP_STATS)) {
5876 if (inet6_fill_ifla6_stats_attrs(skb, idev) < 0)
5877 goto nla_put_failure;
5878 }
5879
5880 nla = nla_reserve(skb, IFLA_INET6_TOKEN, sizeof(struct in6_addr));
5881 if (!nla)
5882 goto nla_put_failure;
5883 read_lock_bh(&idev->lock);
5884 memcpy(nla_data(nla), idev->token.s6_addr, nla_len(nla));
5885 read_unlock_bh(&idev->lock);
5886
5887 if (nla_put_u8(skb, IFLA_INET6_ADDR_GEN_MODE,
5888 READ_ONCE(idev->cnf.addr_gen_mode)))
5889 goto nla_put_failure;
5890
5891 ra_mtu = READ_ONCE(idev->ra_mtu);
5892 if (ra_mtu && nla_put_u32(skb, IFLA_INET6_RA_MTU, ra_mtu))
5893 goto nla_put_failure;
5894
5895 return 0;
5896
5897 nla_put_failure:
5898 return -EMSGSIZE;
5899 }
5900
inet6_get_link_af_size(const struct net_device * dev,u32 ext_filter_mask)5901 static size_t inet6_get_link_af_size(const struct net_device *dev,
5902 u32 ext_filter_mask)
5903 {
5904 if (!__in6_dev_get(dev))
5905 return 0;
5906
5907 return inet6_ifla6_size();
5908 }
5909
inet6_fill_link_af(struct sk_buff * skb,const struct net_device * dev,u32 ext_filter_mask)5910 static int inet6_fill_link_af(struct sk_buff *skb, const struct net_device *dev,
5911 u32 ext_filter_mask)
5912 {
5913 struct inet6_dev *idev = __in6_dev_get(dev);
5914
5915 if (!idev)
5916 return -ENODATA;
5917
5918 if (inet6_fill_ifla6_attrs(skb, idev, ext_filter_mask) < 0)
5919 return -EMSGSIZE;
5920
5921 return 0;
5922 }
5923
inet6_set_iftoken(struct inet6_dev * idev,struct in6_addr * token,struct netlink_ext_ack * extack)5924 static int inet6_set_iftoken(struct inet6_dev *idev, struct in6_addr *token,
5925 struct netlink_ext_ack *extack)
5926 {
5927 struct inet6_ifaddr *ifp;
5928 struct net_device *dev = idev->dev;
5929 bool clear_token, update_rs = false;
5930 struct in6_addr ll_addr;
5931
5932 ASSERT_RTNL();
5933
5934 if (!token)
5935 return -EINVAL;
5936
5937 if (dev->flags & IFF_LOOPBACK) {
5938 NL_SET_ERR_MSG_MOD(extack, "Device is loopback");
5939 return -EINVAL;
5940 }
5941
5942 if (dev->flags & IFF_NOARP) {
5943 NL_SET_ERR_MSG_MOD(extack,
5944 "Device does not do neighbour discovery");
5945 return -EINVAL;
5946 }
5947
5948 if (!ipv6_accept_ra(idev)) {
5949 NL_SET_ERR_MSG_MOD(extack,
5950 "Router advertisement is disabled on device");
5951 return -EINVAL;
5952 }
5953
5954 if (READ_ONCE(idev->cnf.rtr_solicits) == 0) {
5955 NL_SET_ERR_MSG(extack,
5956 "Router solicitation is disabled on device");
5957 return -EINVAL;
5958 }
5959
5960 write_lock_bh(&idev->lock);
5961
5962 BUILD_BUG_ON(sizeof(token->s6_addr) != 16);
5963 memcpy(idev->token.s6_addr + 8, token->s6_addr + 8, 8);
5964
5965 write_unlock_bh(&idev->lock);
5966
5967 clear_token = ipv6_addr_any(token);
5968 if (clear_token)
5969 goto update_lft;
5970
5971 if (!idev->dead && (idev->if_flags & IF_READY) &&
5972 !ipv6_get_lladdr(dev, &ll_addr, IFA_F_TENTATIVE |
5973 IFA_F_OPTIMISTIC)) {
5974 /* If we're not ready, then normal ifup will take care
5975 * of this. Otherwise, we need to request our rs here.
5976 */
5977 ndisc_send_rs(dev, &ll_addr, &in6addr_linklocal_allrouters);
5978 update_rs = true;
5979 }
5980
5981 update_lft:
5982 write_lock_bh(&idev->lock);
5983
5984 if (update_rs) {
5985 idev->if_flags |= IF_RS_SENT;
5986 idev->rs_interval = rfc3315_s14_backoff_init(
5987 READ_ONCE(idev->cnf.rtr_solicit_interval));
5988 idev->rs_probes = 1;
5989 addrconf_mod_rs_timer(idev, idev->rs_interval);
5990 }
5991
5992 /* Well, that's kinda nasty ... */
5993 list_for_each_entry(ifp, &idev->addr_list, if_list) {
5994 spin_lock(&ifp->lock);
5995 if (ifp->tokenized) {
5996 ifp->valid_lft = 0;
5997 ifp->prefered_lft = 0;
5998 }
5999 spin_unlock(&ifp->lock);
6000 }
6001
6002 write_unlock_bh(&idev->lock);
6003 inet6_ifinfo_notify(RTM_NEWLINK, idev);
6004 addrconf_verify_rtnl(dev_net(dev));
6005 return 0;
6006 }
6007
6008 static const struct nla_policy inet6_af_policy[IFLA_INET6_MAX + 1] = {
6009 [IFLA_INET6_ADDR_GEN_MODE] = { .type = NLA_U8 },
6010 [IFLA_INET6_TOKEN] = { .len = sizeof(struct in6_addr) },
6011 [IFLA_INET6_RA_MTU] = { .type = NLA_REJECT,
6012 .reject_message =
6013 "IFLA_INET6_RA_MTU can not be set" },
6014 };
6015
check_addr_gen_mode(int mode)6016 static int check_addr_gen_mode(int mode)
6017 {
6018 if (mode != IN6_ADDR_GEN_MODE_EUI64 &&
6019 mode != IN6_ADDR_GEN_MODE_NONE &&
6020 mode != IN6_ADDR_GEN_MODE_STABLE_PRIVACY &&
6021 mode != IN6_ADDR_GEN_MODE_RANDOM)
6022 return -EINVAL;
6023 return 1;
6024 }
6025
check_stable_privacy(struct inet6_dev * idev,struct net * net,int mode)6026 static int check_stable_privacy(struct inet6_dev *idev, struct net *net,
6027 int mode)
6028 {
6029 if (mode == IN6_ADDR_GEN_MODE_STABLE_PRIVACY &&
6030 !idev->cnf.stable_secret.initialized &&
6031 !net->ipv6.devconf_dflt->stable_secret.initialized)
6032 return -EINVAL;
6033 return 1;
6034 }
6035
inet6_validate_link_af(const struct net_device * dev,const struct nlattr * nla,struct netlink_ext_ack * extack)6036 static int inet6_validate_link_af(const struct net_device *dev,
6037 const struct nlattr *nla,
6038 struct netlink_ext_ack *extack)
6039 {
6040 struct nlattr *tb[IFLA_INET6_MAX + 1];
6041 struct inet6_dev *idev = NULL;
6042 int err;
6043
6044 if (dev) {
6045 idev = __in6_dev_get(dev);
6046 if (!idev)
6047 return -EAFNOSUPPORT;
6048 }
6049
6050 err = nla_parse_nested_deprecated(tb, IFLA_INET6_MAX, nla,
6051 inet6_af_policy, extack);
6052 if (err)
6053 return err;
6054
6055 if (!tb[IFLA_INET6_TOKEN] && !tb[IFLA_INET6_ADDR_GEN_MODE])
6056 return -EINVAL;
6057
6058 if (tb[IFLA_INET6_ADDR_GEN_MODE]) {
6059 u8 mode = nla_get_u8(tb[IFLA_INET6_ADDR_GEN_MODE]);
6060
6061 if (check_addr_gen_mode(mode) < 0)
6062 return -EINVAL;
6063 if (dev && check_stable_privacy(idev, dev_net(dev), mode) < 0)
6064 return -EINVAL;
6065 }
6066
6067 return 0;
6068 }
6069
inet6_set_link_af(struct net_device * dev,const struct nlattr * nla,struct netlink_ext_ack * extack)6070 static int inet6_set_link_af(struct net_device *dev, const struct nlattr *nla,
6071 struct netlink_ext_ack *extack)
6072 {
6073 struct inet6_dev *idev = __in6_dev_get(dev);
6074 struct nlattr *tb[IFLA_INET6_MAX + 1];
6075 int err;
6076
6077 if (!idev)
6078 return -EAFNOSUPPORT;
6079
6080 if (nla_parse_nested_deprecated(tb, IFLA_INET6_MAX, nla, NULL, NULL) < 0)
6081 return -EINVAL;
6082
6083 if (tb[IFLA_INET6_TOKEN]) {
6084 err = inet6_set_iftoken(idev, nla_data(tb[IFLA_INET6_TOKEN]),
6085 extack);
6086 if (err)
6087 return err;
6088 }
6089
6090 if (tb[IFLA_INET6_ADDR_GEN_MODE]) {
6091 u8 mode = nla_get_u8(tb[IFLA_INET6_ADDR_GEN_MODE]);
6092
6093 WRITE_ONCE(idev->cnf.addr_gen_mode, mode);
6094 }
6095
6096 return 0;
6097 }
6098
inet6_fill_ifinfo(struct sk_buff * skb,struct inet6_dev * idev,u32 portid,u32 seq,int event,unsigned int flags)6099 static int inet6_fill_ifinfo(struct sk_buff *skb, struct inet6_dev *idev,
6100 u32 portid, u32 seq, int event, unsigned int flags)
6101 {
6102 struct net_device *dev = idev->dev;
6103 struct ifinfomsg *hdr;
6104 struct nlmsghdr *nlh;
6105 int ifindex, iflink;
6106 void *protoinfo;
6107
6108 nlh = nlmsg_put(skb, portid, seq, event, sizeof(*hdr), flags);
6109 if (!nlh)
6110 return -EMSGSIZE;
6111
6112 hdr = nlmsg_data(nlh);
6113 hdr->ifi_family = AF_INET6;
6114 hdr->__ifi_pad = 0;
6115 hdr->ifi_type = dev->type;
6116 ifindex = READ_ONCE(dev->ifindex);
6117 hdr->ifi_index = ifindex;
6118 hdr->ifi_flags = netif_get_flags(dev);
6119 hdr->ifi_change = 0;
6120
6121 iflink = dev_get_iflink(dev);
6122 if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
6123 (dev->addr_len &&
6124 nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
6125 nla_put_u32(skb, IFLA_MTU, READ_ONCE(dev->mtu)) ||
6126 (ifindex != iflink &&
6127 nla_put_u32(skb, IFLA_LINK, iflink)) ||
6128 nla_put_u8(skb, IFLA_OPERSTATE,
6129 netif_running(dev) ? READ_ONCE(dev->operstate) : IF_OPER_DOWN))
6130 goto nla_put_failure;
6131 protoinfo = nla_nest_start_noflag(skb, IFLA_PROTINFO);
6132 if (!protoinfo)
6133 goto nla_put_failure;
6134
6135 if (inet6_fill_ifla6_attrs(skb, idev, 0) < 0)
6136 goto nla_put_failure;
6137
6138 nla_nest_end(skb, protoinfo);
6139 nlmsg_end(skb, nlh);
6140 return 0;
6141
6142 nla_put_failure:
6143 nlmsg_cancel(skb, nlh);
6144 return -EMSGSIZE;
6145 }
6146
inet6_valid_dump_ifinfo(const struct nlmsghdr * nlh,struct netlink_ext_ack * extack)6147 static int inet6_valid_dump_ifinfo(const struct nlmsghdr *nlh,
6148 struct netlink_ext_ack *extack)
6149 {
6150 struct ifinfomsg *ifm;
6151
6152 ifm = nlmsg_payload(nlh, sizeof(*ifm));
6153 if (!ifm) {
6154 NL_SET_ERR_MSG_MOD(extack, "Invalid header for link dump request");
6155 return -EINVAL;
6156 }
6157
6158 if (nlmsg_attrlen(nlh, sizeof(*ifm))) {
6159 NL_SET_ERR_MSG_MOD(extack, "Invalid data after header");
6160 return -EINVAL;
6161 }
6162
6163 if (ifm->__ifi_pad || ifm->ifi_type || ifm->ifi_flags ||
6164 ifm->ifi_change || ifm->ifi_index) {
6165 NL_SET_ERR_MSG_MOD(extack, "Invalid values in header for dump request");
6166 return -EINVAL;
6167 }
6168
6169 return 0;
6170 }
6171
inet6_dump_ifinfo(struct sk_buff * skb,struct netlink_callback * cb)6172 static int inet6_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
6173 {
6174 struct net *net = sock_net(skb->sk);
6175 struct {
6176 unsigned long ifindex;
6177 } *ctx = (void *)cb->ctx;
6178 struct net_device *dev;
6179 struct inet6_dev *idev;
6180 int err;
6181
6182 /* only requests using strict checking can pass data to
6183 * influence the dump
6184 */
6185 if (cb->strict_check) {
6186 err = inet6_valid_dump_ifinfo(cb->nlh, cb->extack);
6187
6188 if (err < 0)
6189 return err;
6190 }
6191
6192 err = 0;
6193 rcu_read_lock();
6194 for_each_netdev_dump(net, dev, ctx->ifindex) {
6195 idev = __in6_dev_get(dev);
6196 if (!idev)
6197 continue;
6198 err = inet6_fill_ifinfo(skb, idev,
6199 NETLINK_CB(cb->skb).portid,
6200 cb->nlh->nlmsg_seq,
6201 RTM_NEWLINK, NLM_F_MULTI);
6202 if (err < 0)
6203 break;
6204 }
6205 rcu_read_unlock();
6206
6207 return err;
6208 }
6209
inet6_ifinfo_notify(int event,struct inet6_dev * idev)6210 void inet6_ifinfo_notify(int event, struct inet6_dev *idev)
6211 {
6212 struct sk_buff *skb;
6213 struct net *net = dev_net(idev->dev);
6214 int err = -ENOBUFS;
6215
6216 skb = nlmsg_new(inet6_if_nlmsg_size(), GFP_ATOMIC);
6217 if (!skb)
6218 goto errout;
6219
6220 err = inet6_fill_ifinfo(skb, idev, 0, 0, event, 0);
6221 if (err < 0) {
6222 /* -EMSGSIZE implies BUG in inet6_if_nlmsg_size() */
6223 WARN_ON(err == -EMSGSIZE);
6224 kfree_skb(skb);
6225 goto errout;
6226 }
6227 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFINFO, NULL, GFP_ATOMIC);
6228 return;
6229 errout:
6230 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFINFO, err);
6231 }
6232
inet6_prefix_nlmsg_size(void)6233 static inline size_t inet6_prefix_nlmsg_size(void)
6234 {
6235 return NLMSG_ALIGN(sizeof(struct prefixmsg))
6236 + nla_total_size(sizeof(struct in6_addr))
6237 + nla_total_size(sizeof(struct prefix_cacheinfo));
6238 }
6239
inet6_fill_prefix(struct sk_buff * skb,struct inet6_dev * idev,struct prefix_info * pinfo,u32 portid,u32 seq,int event,unsigned int flags)6240 static int inet6_fill_prefix(struct sk_buff *skb, struct inet6_dev *idev,
6241 struct prefix_info *pinfo, u32 portid, u32 seq,
6242 int event, unsigned int flags)
6243 {
6244 struct prefixmsg *pmsg;
6245 struct nlmsghdr *nlh;
6246 struct prefix_cacheinfo ci;
6247
6248 nlh = nlmsg_put(skb, portid, seq, event, sizeof(*pmsg), flags);
6249 if (!nlh)
6250 return -EMSGSIZE;
6251
6252 pmsg = nlmsg_data(nlh);
6253 pmsg->prefix_family = AF_INET6;
6254 pmsg->prefix_pad1 = 0;
6255 pmsg->prefix_pad2 = 0;
6256 pmsg->prefix_ifindex = idev->dev->ifindex;
6257 pmsg->prefix_len = pinfo->prefix_len;
6258 pmsg->prefix_type = pinfo->type;
6259 pmsg->prefix_pad3 = 0;
6260 pmsg->prefix_flags = pinfo->flags;
6261
6262 if (nla_put(skb, PREFIX_ADDRESS, sizeof(pinfo->prefix), &pinfo->prefix))
6263 goto nla_put_failure;
6264 ci.preferred_time = ntohl(pinfo->prefered);
6265 ci.valid_time = ntohl(pinfo->valid);
6266 if (nla_put(skb, PREFIX_CACHEINFO, sizeof(ci), &ci))
6267 goto nla_put_failure;
6268 nlmsg_end(skb, nlh);
6269 return 0;
6270
6271 nla_put_failure:
6272 nlmsg_cancel(skb, nlh);
6273 return -EMSGSIZE;
6274 }
6275
inet6_prefix_notify(int event,struct inet6_dev * idev,struct prefix_info * pinfo)6276 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
6277 struct prefix_info *pinfo)
6278 {
6279 struct sk_buff *skb;
6280 struct net *net = dev_net(idev->dev);
6281 int err = -ENOBUFS;
6282
6283 skb = nlmsg_new(inet6_prefix_nlmsg_size(), GFP_ATOMIC);
6284 if (!skb)
6285 goto errout;
6286
6287 err = inet6_fill_prefix(skb, idev, pinfo, 0, 0, event, 0);
6288 if (err < 0) {
6289 /* -EMSGSIZE implies BUG in inet6_prefix_nlmsg_size() */
6290 WARN_ON(err == -EMSGSIZE);
6291 kfree_skb(skb);
6292 goto errout;
6293 }
6294 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_PREFIX, NULL, GFP_ATOMIC);
6295 return;
6296 errout:
6297 rtnl_set_sk_err(net, RTNLGRP_IPV6_PREFIX, err);
6298 }
6299
__ipv6_ifa_notify(int event,struct inet6_ifaddr * ifp)6300 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
6301 {
6302 struct net *net = dev_net(ifp->idev->dev);
6303
6304 if (event)
6305 ASSERT_RTNL();
6306
6307 inet6_ifa_notify(event ? : RTM_NEWADDR, ifp);
6308
6309 switch (event) {
6310 case RTM_NEWADDR:
6311 /*
6312 * If the address was optimistic we inserted the route at the
6313 * start of our DAD process, so we don't need to do it again.
6314 * If the device was taken down in the middle of the DAD
6315 * cycle there is a race where we could get here without a
6316 * host route, so nothing to insert. That will be fixed when
6317 * the device is brought up.
6318 */
6319 if (ifp->rt && !rcu_access_pointer(ifp->rt->fib6_node)) {
6320 ip6_ins_rt(net, ifp->rt);
6321 } else if (!ifp->rt && (ifp->idev->dev->flags & IFF_UP)) {
6322 pr_warn("BUG: Address %pI6c on device %s is missing its host route.\n",
6323 &ifp->addr, ifp->idev->dev->name);
6324 }
6325
6326 if (ifp->idev->cnf.forwarding)
6327 addrconf_join_anycast(ifp);
6328 if (!ipv6_addr_any(&ifp->peer_addr))
6329 addrconf_prefix_route(&ifp->peer_addr, 128,
6330 ifp->rt_priority, ifp->idev->dev,
6331 0, 0, GFP_ATOMIC);
6332 break;
6333 case RTM_DELADDR:
6334 if (ifp->idev->cnf.forwarding)
6335 addrconf_leave_anycast(ifp);
6336 addrconf_leave_solict(ifp->idev, &ifp->addr);
6337 if (!ipv6_addr_any(&ifp->peer_addr)) {
6338 struct fib6_info *rt;
6339
6340 rt = addrconf_get_prefix_route(&ifp->peer_addr, 128,
6341 ifp->idev->dev, 0, 0,
6342 false);
6343 if (rt)
6344 ip6_del_rt(net, rt, false);
6345 }
6346 if (ifp->rt) {
6347 ip6_del_rt(net, ifp->rt, false);
6348 ifp->rt = NULL;
6349 }
6350 rt_genid_bump_ipv6(net);
6351 break;
6352 }
6353 atomic_inc(&net->ipv6.dev_addr_genid);
6354 }
6355
ipv6_ifa_notify(int event,struct inet6_ifaddr * ifp)6356 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
6357 {
6358 if (likely(ifp->idev->dead == 0))
6359 __ipv6_ifa_notify(event, ifp);
6360 }
6361
6362 #ifdef CONFIG_SYSCTL
6363
addrconf_sysctl_forward(const struct ctl_table * ctl,int write,void * buffer,size_t * lenp,loff_t * ppos)6364 static int addrconf_sysctl_forward(const struct ctl_table *ctl, int write,
6365 void *buffer, size_t *lenp, loff_t *ppos)
6366 {
6367 struct ctl_table lctl;
6368 int *valp = ctl->data;
6369 int val = *valp;
6370 int ret;
6371
6372 /*
6373 * ctl->data points to idev->cnf.forwarding, we should
6374 * not modify it until we get the rtnl lock.
6375 */
6376 lctl = *ctl;
6377 lctl.data = &val;
6378
6379 ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
6380 if (ret)
6381 return ret;
6382
6383 if (write)
6384 ret = addrconf_fixup_forwarding(ctl, valp, val);
6385 return ret;
6386 }
6387
addrconf_sysctl_mtu(const struct ctl_table * ctl,int write,void * buffer,size_t * lenp,loff_t * ppos)6388 static int addrconf_sysctl_mtu(const struct ctl_table *ctl, int write,
6389 void *buffer, size_t *lenp, loff_t *ppos)
6390 {
6391 struct inet6_dev *idev = ctl->extra1;
6392 int min_mtu = IPV6_MIN_MTU;
6393 struct ctl_table lctl;
6394
6395 lctl = *ctl;
6396 lctl.extra1 = &min_mtu;
6397 lctl.extra2 = idev ? &idev->dev->mtu : NULL;
6398
6399 return proc_dointvec_minmax(&lctl, write, buffer, lenp, ppos);
6400 }
6401
dev_disable_change(struct inet6_dev * idev)6402 static void dev_disable_change(struct inet6_dev *idev)
6403 {
6404 struct netdev_notifier_info info;
6405
6406 if (!idev || !idev->dev)
6407 return;
6408
6409 netdev_notifier_info_init(&info, idev->dev);
6410 if (idev->cnf.disable_ipv6)
6411 addrconf_notify(NULL, NETDEV_DOWN, &info);
6412 else
6413 addrconf_notify(NULL, NETDEV_UP, &info);
6414 }
6415
addrconf_disable_change(struct net * net,__s32 newf)6416 static void addrconf_disable_change(struct net *net, __s32 newf)
6417 {
6418 struct net_device *dev;
6419 struct inet6_dev *idev;
6420
6421 for_each_netdev(net, dev) {
6422 idev = __in6_dev_get_rtnl_net(dev);
6423 if (idev) {
6424 int changed = (!idev->cnf.disable_ipv6) ^ (!newf);
6425
6426 WRITE_ONCE(idev->cnf.disable_ipv6, newf);
6427 if (changed)
6428 dev_disable_change(idev);
6429 }
6430 }
6431 }
6432
addrconf_disable_ipv6(const struct ctl_table * table,int * p,int newf)6433 static int addrconf_disable_ipv6(const struct ctl_table *table, int *p, int newf)
6434 {
6435 struct net *net = (struct net *)table->extra2;
6436 int old;
6437
6438 if (p == &net->ipv6.devconf_dflt->disable_ipv6) {
6439 WRITE_ONCE(*p, newf);
6440 return 0;
6441 }
6442
6443 if (!rtnl_net_trylock(net))
6444 return restart_syscall();
6445
6446 old = *p;
6447 WRITE_ONCE(*p, newf);
6448
6449 if (p == &net->ipv6.devconf_all->disable_ipv6) {
6450 WRITE_ONCE(net->ipv6.devconf_dflt->disable_ipv6, newf);
6451 addrconf_disable_change(net, newf);
6452 } else if ((!newf) ^ (!old)) {
6453 dev_disable_change((struct inet6_dev *)table->extra1);
6454 }
6455
6456 rtnl_net_unlock(net);
6457 return 0;
6458 }
6459
addrconf_sysctl_disable(const struct ctl_table * ctl,int write,void * buffer,size_t * lenp,loff_t * ppos)6460 static int addrconf_sysctl_disable(const struct ctl_table *ctl, int write,
6461 void *buffer, size_t *lenp, loff_t *ppos)
6462 {
6463 struct ctl_table lctl;
6464 int *valp = ctl->data;
6465 int val = *valp;
6466 int ret;
6467
6468 /*
6469 * ctl->data points to idev->cnf.disable_ipv6, we should
6470 * not modify it until we get the rtnl lock.
6471 */
6472 lctl = *ctl;
6473 lctl.data = &val;
6474
6475 ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
6476 if (ret)
6477 return ret;
6478
6479 if (write)
6480 ret = addrconf_disable_ipv6(ctl, valp, val);
6481 return ret;
6482 }
6483
addrconf_sysctl_proxy_ndp(const struct ctl_table * ctl,int write,void * buffer,size_t * lenp,loff_t * ppos)6484 static int addrconf_sysctl_proxy_ndp(const struct ctl_table *ctl, int write,
6485 void *buffer, size_t *lenp, loff_t *ppos)
6486 {
6487 struct net *net = ctl->extra2;
6488 int *valp = ctl->data;
6489 int old, new;
6490 int ret;
6491
6492 if (write && !rtnl_net_trylock(net))
6493 return restart_syscall();
6494
6495 old = *valp;
6496 ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
6497 new = *valp;
6498
6499 if (write && old != new) {
6500 if (valp == &net->ipv6.devconf_dflt->proxy_ndp) {
6501 inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
6502 NETCONFA_PROXY_NEIGH,
6503 NETCONFA_IFINDEX_DEFAULT,
6504 net->ipv6.devconf_dflt);
6505 } else if (valp == &net->ipv6.devconf_all->proxy_ndp) {
6506 inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
6507 NETCONFA_PROXY_NEIGH,
6508 NETCONFA_IFINDEX_ALL,
6509 net->ipv6.devconf_all);
6510 } else {
6511 struct inet6_dev *idev = ctl->extra1;
6512
6513 inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
6514 NETCONFA_PROXY_NEIGH,
6515 idev->dev->ifindex,
6516 &idev->cnf);
6517 }
6518 }
6519 if (write)
6520 rtnl_net_unlock(net);
6521
6522 return ret;
6523 }
6524
addrconf_sysctl_addr_gen_mode(const struct ctl_table * ctl,int write,void * buffer,size_t * lenp,loff_t * ppos)6525 static int addrconf_sysctl_addr_gen_mode(const struct ctl_table *ctl, int write,
6526 void *buffer, size_t *lenp,
6527 loff_t *ppos)
6528 {
6529 int ret = 0;
6530 u32 new_val;
6531 struct inet6_dev *idev = (struct inet6_dev *)ctl->extra1;
6532 struct net *net = (struct net *)ctl->extra2;
6533 struct ctl_table tmp = {
6534 .data = &new_val,
6535 .maxlen = sizeof(new_val),
6536 .mode = ctl->mode,
6537 };
6538
6539 if (!rtnl_net_trylock(net))
6540 return restart_syscall();
6541
6542 new_val = *((u32 *)ctl->data);
6543
6544 ret = proc_douintvec(&tmp, write, buffer, lenp, ppos);
6545 if (ret != 0)
6546 goto out;
6547
6548 if (write) {
6549 if (check_addr_gen_mode(new_val) < 0) {
6550 ret = -EINVAL;
6551 goto out;
6552 }
6553
6554 if (idev) {
6555 if (check_stable_privacy(idev, net, new_val) < 0) {
6556 ret = -EINVAL;
6557 goto out;
6558 }
6559
6560 if (idev->cnf.addr_gen_mode != new_val) {
6561 WRITE_ONCE(idev->cnf.addr_gen_mode, new_val);
6562 netdev_lock_ops(idev->dev);
6563 addrconf_init_auto_addrs(idev->dev);
6564 netdev_unlock_ops(idev->dev);
6565 }
6566 } else if (&net->ipv6.devconf_all->addr_gen_mode == ctl->data) {
6567 struct net_device *dev;
6568
6569 WRITE_ONCE(net->ipv6.devconf_dflt->addr_gen_mode, new_val);
6570 for_each_netdev(net, dev) {
6571 idev = __in6_dev_get_rtnl_net(dev);
6572 if (idev &&
6573 idev->cnf.addr_gen_mode != new_val) {
6574 WRITE_ONCE(idev->cnf.addr_gen_mode,
6575 new_val);
6576 netdev_lock_ops(idev->dev);
6577 addrconf_init_auto_addrs(idev->dev);
6578 netdev_unlock_ops(idev->dev);
6579 }
6580 }
6581 }
6582
6583 WRITE_ONCE(*((u32 *)ctl->data), new_val);
6584 }
6585
6586 out:
6587 rtnl_net_unlock(net);
6588
6589 return ret;
6590 }
6591
addrconf_sysctl_stable_secret(const struct ctl_table * ctl,int write,void * buffer,size_t * lenp,loff_t * ppos)6592 static int addrconf_sysctl_stable_secret(const struct ctl_table *ctl, int write,
6593 void *buffer, size_t *lenp,
6594 loff_t *ppos)
6595 {
6596 int err;
6597 struct in6_addr addr;
6598 char str[IPV6_MAX_STRLEN];
6599 struct ctl_table lctl = *ctl;
6600 struct net *net = ctl->extra2;
6601 struct ipv6_stable_secret *secret = ctl->data;
6602
6603 if (&net->ipv6.devconf_all->stable_secret == ctl->data)
6604 return -EIO;
6605
6606 lctl.maxlen = IPV6_MAX_STRLEN;
6607 lctl.data = str;
6608
6609 if (!rtnl_net_trylock(net))
6610 return restart_syscall();
6611
6612 if (!write && !secret->initialized) {
6613 err = -EIO;
6614 goto out;
6615 }
6616
6617 err = snprintf(str, sizeof(str), "%pI6", &secret->secret);
6618 if (err >= sizeof(str)) {
6619 err = -EIO;
6620 goto out;
6621 }
6622
6623 err = proc_dostring(&lctl, write, buffer, lenp, ppos);
6624 if (err || !write)
6625 goto out;
6626
6627 if (in6_pton(str, -1, addr.in6_u.u6_addr8, -1, NULL) != 1) {
6628 err = -EIO;
6629 goto out;
6630 }
6631
6632 secret->initialized = true;
6633 secret->secret = addr;
6634
6635 if (&net->ipv6.devconf_dflt->stable_secret == ctl->data) {
6636 struct net_device *dev;
6637
6638 for_each_netdev(net, dev) {
6639 struct inet6_dev *idev = __in6_dev_get_rtnl_net(dev);
6640
6641 if (idev) {
6642 WRITE_ONCE(idev->cnf.addr_gen_mode,
6643 IN6_ADDR_GEN_MODE_STABLE_PRIVACY);
6644 }
6645 }
6646 } else {
6647 struct inet6_dev *idev = ctl->extra1;
6648
6649 WRITE_ONCE(idev->cnf.addr_gen_mode,
6650 IN6_ADDR_GEN_MODE_STABLE_PRIVACY);
6651 }
6652
6653 out:
6654 rtnl_net_unlock(net);
6655
6656 return err;
6657 }
6658
6659 static
addrconf_sysctl_ignore_routes_with_linkdown(const struct ctl_table * ctl,int write,void * buffer,size_t * lenp,loff_t * ppos)6660 int addrconf_sysctl_ignore_routes_with_linkdown(const struct ctl_table *ctl,
6661 int write, void *buffer,
6662 size_t *lenp,
6663 loff_t *ppos)
6664 {
6665 struct ctl_table lctl;
6666 int *valp = ctl->data;
6667 int val = *valp;
6668 int ret;
6669
6670 /* ctl->data points to idev->cnf.ignore_routes_when_linkdown
6671 * we should not modify it until we get the rtnl lock.
6672 */
6673 lctl = *ctl;
6674 lctl.data = &val;
6675
6676 ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
6677 if (ret)
6678 return ret;
6679
6680 if (write)
6681 ret = addrconf_fixup_linkdown(ctl, valp, val);
6682 return ret;
6683 }
6684
6685 static
addrconf_set_nopolicy(struct rt6_info * rt,int action)6686 void addrconf_set_nopolicy(struct rt6_info *rt, int action)
6687 {
6688 if (rt) {
6689 if (action)
6690 rt->dst.flags |= DST_NOPOLICY;
6691 else
6692 rt->dst.flags &= ~DST_NOPOLICY;
6693 }
6694 }
6695
6696 static
addrconf_disable_policy_idev(struct inet6_dev * idev,int val)6697 void addrconf_disable_policy_idev(struct inet6_dev *idev, int val)
6698 {
6699 struct inet6_ifaddr *ifa;
6700
6701 read_lock_bh(&idev->lock);
6702 list_for_each_entry(ifa, &idev->addr_list, if_list) {
6703 spin_lock(&ifa->lock);
6704 if (ifa->rt) {
6705 /* host routes only use builtin fib6_nh */
6706 struct fib6_nh *nh = ifa->rt->fib6_nh;
6707 int cpu;
6708
6709 rcu_read_lock();
6710 ifa->rt->dst_nopolicy = val ? true : false;
6711 if (nh->rt6i_pcpu) {
6712 for_each_possible_cpu(cpu) {
6713 struct rt6_info **rtp;
6714
6715 rtp = per_cpu_ptr(nh->rt6i_pcpu, cpu);
6716 addrconf_set_nopolicy(*rtp, val);
6717 }
6718 }
6719 rcu_read_unlock();
6720 }
6721 spin_unlock(&ifa->lock);
6722 }
6723 read_unlock_bh(&idev->lock);
6724 }
6725
6726 static
addrconf_disable_policy(const struct ctl_table * ctl,int * valp,int val)6727 int addrconf_disable_policy(const struct ctl_table *ctl, int *valp, int val)
6728 {
6729 struct net *net = (struct net *)ctl->extra2;
6730 struct inet6_dev *idev;
6731
6732 if (valp == &net->ipv6.devconf_dflt->disable_policy) {
6733 WRITE_ONCE(*valp, val);
6734 return 0;
6735 }
6736
6737 if (!rtnl_net_trylock(net))
6738 return restart_syscall();
6739
6740 WRITE_ONCE(*valp, val);
6741
6742 if (valp == &net->ipv6.devconf_all->disable_policy) {
6743 struct net_device *dev;
6744
6745 for_each_netdev(net, dev) {
6746 idev = __in6_dev_get_rtnl_net(dev);
6747 if (idev)
6748 addrconf_disable_policy_idev(idev, val);
6749 }
6750 } else {
6751 idev = (struct inet6_dev *)ctl->extra1;
6752 addrconf_disable_policy_idev(idev, val);
6753 }
6754
6755 rtnl_net_unlock(net);
6756 return 0;
6757 }
6758
addrconf_sysctl_disable_policy(const struct ctl_table * ctl,int write,void * buffer,size_t * lenp,loff_t * ppos)6759 static int addrconf_sysctl_disable_policy(const struct ctl_table *ctl, int write,
6760 void *buffer, size_t *lenp, loff_t *ppos)
6761 {
6762 struct ctl_table lctl;
6763 int *valp = ctl->data;
6764 int val = *valp;
6765 int ret;
6766
6767 lctl = *ctl;
6768 lctl.data = &val;
6769 ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
6770 if (ret)
6771 return ret;
6772
6773 if (write && (*valp != val))
6774 ret = addrconf_disable_policy(ctl, valp, val);
6775
6776 return ret;
6777 }
6778
addrconf_force_forward_change(struct net * net,__s32 newf)6779 static void addrconf_force_forward_change(struct net *net, __s32 newf)
6780 {
6781 struct net_device *dev;
6782 struct inet6_dev *idev;
6783
6784 for_each_netdev(net, dev) {
6785 idev = __in6_dev_get_rtnl_net(dev);
6786 if (idev) {
6787 int changed = (!idev->cnf.force_forwarding) ^ (!newf);
6788
6789 WRITE_ONCE(idev->cnf.force_forwarding, newf);
6790 if (changed)
6791 inet6_netconf_notify_devconf(dev_net(dev), RTM_NEWNETCONF,
6792 NETCONFA_FORCE_FORWARDING,
6793 dev->ifindex, &idev->cnf);
6794 }
6795 }
6796 }
6797
addrconf_sysctl_force_forwarding(const struct ctl_table * ctl,int write,void * buffer,size_t * lenp,loff_t * ppos)6798 static int addrconf_sysctl_force_forwarding(const struct ctl_table *ctl, int write,
6799 void *buffer, size_t *lenp, loff_t *ppos)
6800 {
6801 struct inet6_dev *idev = ctl->extra1;
6802 struct ctl_table tmp_ctl = *ctl;
6803 struct net *net = ctl->extra2;
6804 int *valp = ctl->data;
6805 int new_val = *valp;
6806 int old_val = *valp;
6807 int ret;
6808
6809 tmp_ctl.extra1 = SYSCTL_ZERO;
6810 tmp_ctl.extra2 = SYSCTL_ONE;
6811 tmp_ctl.data = &new_val;
6812
6813 ret = proc_douintvec_minmax(&tmp_ctl, write, buffer, lenp, ppos);
6814
6815 if (write && old_val != new_val) {
6816 if (!rtnl_net_trylock(net))
6817 return restart_syscall();
6818
6819 WRITE_ONCE(*valp, new_val);
6820
6821 if (valp == &net->ipv6.devconf_dflt->force_forwarding) {
6822 inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
6823 NETCONFA_FORCE_FORWARDING,
6824 NETCONFA_IFINDEX_DEFAULT,
6825 net->ipv6.devconf_dflt);
6826 } else if (valp == &net->ipv6.devconf_all->force_forwarding) {
6827 inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
6828 NETCONFA_FORCE_FORWARDING,
6829 NETCONFA_IFINDEX_ALL,
6830 net->ipv6.devconf_all);
6831
6832 addrconf_force_forward_change(net, new_val);
6833 } else {
6834 inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
6835 NETCONFA_FORCE_FORWARDING,
6836 idev->dev->ifindex,
6837 &idev->cnf);
6838 }
6839 rtnl_net_unlock(net);
6840 }
6841
6842 return ret;
6843 }
6844
6845 static int minus_one = -1;
6846 static const int two_five_five = 255;
6847 static u32 ioam6_if_id_max = U16_MAX;
6848
6849 static const struct ctl_table addrconf_sysctl[] = {
6850 {
6851 .procname = "forwarding",
6852 .data = &ipv6_devconf.forwarding,
6853 .maxlen = sizeof(int),
6854 .mode = 0644,
6855 .proc_handler = addrconf_sysctl_forward,
6856 },
6857 {
6858 .procname = "hop_limit",
6859 .data = &ipv6_devconf.hop_limit,
6860 .maxlen = sizeof(int),
6861 .mode = 0644,
6862 .proc_handler = proc_dointvec_minmax,
6863 .extra1 = (void *)SYSCTL_ONE,
6864 .extra2 = (void *)&two_five_five,
6865 },
6866 {
6867 .procname = "mtu",
6868 .data = &ipv6_devconf.mtu6,
6869 .maxlen = sizeof(int),
6870 .mode = 0644,
6871 .proc_handler = addrconf_sysctl_mtu,
6872 },
6873 {
6874 .procname = "accept_ra",
6875 .data = &ipv6_devconf.accept_ra,
6876 .maxlen = sizeof(int),
6877 .mode = 0644,
6878 .proc_handler = proc_dointvec,
6879 },
6880 {
6881 .procname = "accept_redirects",
6882 .data = &ipv6_devconf.accept_redirects,
6883 .maxlen = sizeof(int),
6884 .mode = 0644,
6885 .proc_handler = proc_dointvec,
6886 },
6887 {
6888 .procname = "autoconf",
6889 .data = &ipv6_devconf.autoconf,
6890 .maxlen = sizeof(int),
6891 .mode = 0644,
6892 .proc_handler = proc_dointvec,
6893 },
6894 {
6895 .procname = "dad_transmits",
6896 .data = &ipv6_devconf.dad_transmits,
6897 .maxlen = sizeof(int),
6898 .mode = 0644,
6899 .proc_handler = proc_dointvec,
6900 },
6901 {
6902 .procname = "router_solicitations",
6903 .data = &ipv6_devconf.rtr_solicits,
6904 .maxlen = sizeof(int),
6905 .mode = 0644,
6906 .proc_handler = proc_dointvec_minmax,
6907 .extra1 = &minus_one,
6908 },
6909 {
6910 .procname = "router_solicitation_interval",
6911 .data = &ipv6_devconf.rtr_solicit_interval,
6912 .maxlen = sizeof(int),
6913 .mode = 0644,
6914 .proc_handler = proc_dointvec_jiffies,
6915 },
6916 {
6917 .procname = "router_solicitation_max_interval",
6918 .data = &ipv6_devconf.rtr_solicit_max_interval,
6919 .maxlen = sizeof(int),
6920 .mode = 0644,
6921 .proc_handler = proc_dointvec_jiffies,
6922 },
6923 {
6924 .procname = "router_solicitation_delay",
6925 .data = &ipv6_devconf.rtr_solicit_delay,
6926 .maxlen = sizeof(int),
6927 .mode = 0644,
6928 .proc_handler = proc_dointvec_jiffies,
6929 },
6930 {
6931 .procname = "force_mld_version",
6932 .data = &ipv6_devconf.force_mld_version,
6933 .maxlen = sizeof(int),
6934 .mode = 0644,
6935 .proc_handler = proc_dointvec,
6936 },
6937 {
6938 .procname = "mldv1_unsolicited_report_interval",
6939 .data =
6940 &ipv6_devconf.mldv1_unsolicited_report_interval,
6941 .maxlen = sizeof(int),
6942 .mode = 0644,
6943 .proc_handler = proc_dointvec_ms_jiffies,
6944 },
6945 {
6946 .procname = "mldv2_unsolicited_report_interval",
6947 .data =
6948 &ipv6_devconf.mldv2_unsolicited_report_interval,
6949 .maxlen = sizeof(int),
6950 .mode = 0644,
6951 .proc_handler = proc_dointvec_ms_jiffies,
6952 },
6953 {
6954 .procname = "use_tempaddr",
6955 .data = &ipv6_devconf.use_tempaddr,
6956 .maxlen = sizeof(int),
6957 .mode = 0644,
6958 .proc_handler = proc_dointvec,
6959 },
6960 {
6961 .procname = "temp_valid_lft",
6962 .data = &ipv6_devconf.temp_valid_lft,
6963 .maxlen = sizeof(int),
6964 .mode = 0644,
6965 .proc_handler = proc_dointvec,
6966 },
6967 {
6968 .procname = "temp_prefered_lft",
6969 .data = &ipv6_devconf.temp_prefered_lft,
6970 .maxlen = sizeof(int),
6971 .mode = 0644,
6972 .proc_handler = proc_dointvec,
6973 },
6974 {
6975 .procname = "regen_min_advance",
6976 .data = &ipv6_devconf.regen_min_advance,
6977 .maxlen = sizeof(int),
6978 .mode = 0644,
6979 .proc_handler = proc_dointvec,
6980 },
6981 {
6982 .procname = "regen_max_retry",
6983 .data = &ipv6_devconf.regen_max_retry,
6984 .maxlen = sizeof(int),
6985 .mode = 0644,
6986 .proc_handler = proc_dointvec,
6987 },
6988 {
6989 .procname = "max_desync_factor",
6990 .data = &ipv6_devconf.max_desync_factor,
6991 .maxlen = sizeof(int),
6992 .mode = 0644,
6993 .proc_handler = proc_dointvec,
6994 },
6995 {
6996 .procname = "max_addresses",
6997 .data = &ipv6_devconf.max_addresses,
6998 .maxlen = sizeof(int),
6999 .mode = 0644,
7000 .proc_handler = proc_dointvec,
7001 },
7002 {
7003 .procname = "accept_ra_defrtr",
7004 .data = &ipv6_devconf.accept_ra_defrtr,
7005 .maxlen = sizeof(int),
7006 .mode = 0644,
7007 .proc_handler = proc_dointvec,
7008 },
7009 {
7010 .procname = "ra_defrtr_metric",
7011 .data = &ipv6_devconf.ra_defrtr_metric,
7012 .maxlen = sizeof(u32),
7013 .mode = 0644,
7014 .proc_handler = proc_douintvec_minmax,
7015 .extra1 = (void *)SYSCTL_ONE,
7016 },
7017 {
7018 .procname = "accept_ra_min_hop_limit",
7019 .data = &ipv6_devconf.accept_ra_min_hop_limit,
7020 .maxlen = sizeof(int),
7021 .mode = 0644,
7022 .proc_handler = proc_dointvec,
7023 },
7024 {
7025 .procname = "accept_ra_min_lft",
7026 .data = &ipv6_devconf.accept_ra_min_lft,
7027 .maxlen = sizeof(int),
7028 .mode = 0644,
7029 .proc_handler = proc_dointvec,
7030 },
7031 {
7032 .procname = "accept_ra_pinfo",
7033 .data = &ipv6_devconf.accept_ra_pinfo,
7034 .maxlen = sizeof(int),
7035 .mode = 0644,
7036 .proc_handler = proc_dointvec,
7037 },
7038 {
7039 .procname = "ra_honor_pio_life",
7040 .data = &ipv6_devconf.ra_honor_pio_life,
7041 .maxlen = sizeof(u8),
7042 .mode = 0644,
7043 .proc_handler = proc_dou8vec_minmax,
7044 .extra1 = SYSCTL_ZERO,
7045 .extra2 = SYSCTL_ONE,
7046 },
7047 {
7048 .procname = "ra_honor_pio_pflag",
7049 .data = &ipv6_devconf.ra_honor_pio_pflag,
7050 .maxlen = sizeof(u8),
7051 .mode = 0644,
7052 .proc_handler = proc_dou8vec_minmax,
7053 .extra1 = SYSCTL_ZERO,
7054 .extra2 = SYSCTL_ONE,
7055 },
7056 #ifdef CONFIG_IPV6_ROUTER_PREF
7057 {
7058 .procname = "accept_ra_rtr_pref",
7059 .data = &ipv6_devconf.accept_ra_rtr_pref,
7060 .maxlen = sizeof(int),
7061 .mode = 0644,
7062 .proc_handler = proc_dointvec,
7063 },
7064 {
7065 .procname = "router_probe_interval",
7066 .data = &ipv6_devconf.rtr_probe_interval,
7067 .maxlen = sizeof(int),
7068 .mode = 0644,
7069 .proc_handler = proc_dointvec_jiffies,
7070 },
7071 #ifdef CONFIG_IPV6_ROUTE_INFO
7072 {
7073 .procname = "accept_ra_rt_info_min_plen",
7074 .data = &ipv6_devconf.accept_ra_rt_info_min_plen,
7075 .maxlen = sizeof(int),
7076 .mode = 0644,
7077 .proc_handler = proc_dointvec,
7078 },
7079 {
7080 .procname = "accept_ra_rt_info_max_plen",
7081 .data = &ipv6_devconf.accept_ra_rt_info_max_plen,
7082 .maxlen = sizeof(int),
7083 .mode = 0644,
7084 .proc_handler = proc_dointvec,
7085 },
7086 #endif
7087 #endif
7088 {
7089 .procname = "proxy_ndp",
7090 .data = &ipv6_devconf.proxy_ndp,
7091 .maxlen = sizeof(int),
7092 .mode = 0644,
7093 .proc_handler = addrconf_sysctl_proxy_ndp,
7094 },
7095 {
7096 .procname = "accept_source_route",
7097 .data = &ipv6_devconf.accept_source_route,
7098 .maxlen = sizeof(int),
7099 .mode = 0644,
7100 .proc_handler = proc_dointvec,
7101 },
7102 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
7103 {
7104 .procname = "optimistic_dad",
7105 .data = &ipv6_devconf.optimistic_dad,
7106 .maxlen = sizeof(int),
7107 .mode = 0644,
7108 .proc_handler = proc_dointvec,
7109 },
7110 {
7111 .procname = "use_optimistic",
7112 .data = &ipv6_devconf.use_optimistic,
7113 .maxlen = sizeof(int),
7114 .mode = 0644,
7115 .proc_handler = proc_dointvec,
7116 },
7117 #endif
7118 #ifdef CONFIG_IPV6_MROUTE
7119 {
7120 .procname = "mc_forwarding",
7121 .data = &ipv6_devconf.mc_forwarding,
7122 .maxlen = sizeof(int),
7123 .mode = 0444,
7124 .proc_handler = proc_dointvec,
7125 },
7126 #endif
7127 {
7128 .procname = "disable_ipv6",
7129 .data = &ipv6_devconf.disable_ipv6,
7130 .maxlen = sizeof(int),
7131 .mode = 0644,
7132 .proc_handler = addrconf_sysctl_disable,
7133 },
7134 {
7135 .procname = "accept_dad",
7136 .data = &ipv6_devconf.accept_dad,
7137 .maxlen = sizeof(int),
7138 .mode = 0644,
7139 .proc_handler = proc_dointvec,
7140 },
7141 {
7142 .procname = "force_tllao",
7143 .data = &ipv6_devconf.force_tllao,
7144 .maxlen = sizeof(int),
7145 .mode = 0644,
7146 .proc_handler = proc_dointvec
7147 },
7148 {
7149 .procname = "ndisc_notify",
7150 .data = &ipv6_devconf.ndisc_notify,
7151 .maxlen = sizeof(int),
7152 .mode = 0644,
7153 .proc_handler = proc_dointvec
7154 },
7155 {
7156 .procname = "suppress_frag_ndisc",
7157 .data = &ipv6_devconf.suppress_frag_ndisc,
7158 .maxlen = sizeof(int),
7159 .mode = 0644,
7160 .proc_handler = proc_dointvec
7161 },
7162 {
7163 .procname = "accept_ra_from_local",
7164 .data = &ipv6_devconf.accept_ra_from_local,
7165 .maxlen = sizeof(int),
7166 .mode = 0644,
7167 .proc_handler = proc_dointvec,
7168 },
7169 {
7170 .procname = "accept_ra_mtu",
7171 .data = &ipv6_devconf.accept_ra_mtu,
7172 .maxlen = sizeof(int),
7173 .mode = 0644,
7174 .proc_handler = proc_dointvec,
7175 },
7176 {
7177 .procname = "stable_secret",
7178 .data = &ipv6_devconf.stable_secret,
7179 .maxlen = IPV6_MAX_STRLEN,
7180 .mode = 0600,
7181 .proc_handler = addrconf_sysctl_stable_secret,
7182 },
7183 {
7184 .procname = "use_oif_addrs_only",
7185 .data = &ipv6_devconf.use_oif_addrs_only,
7186 .maxlen = sizeof(int),
7187 .mode = 0644,
7188 .proc_handler = proc_dointvec,
7189 },
7190 {
7191 .procname = "ignore_routes_with_linkdown",
7192 .data = &ipv6_devconf.ignore_routes_with_linkdown,
7193 .maxlen = sizeof(int),
7194 .mode = 0644,
7195 .proc_handler = addrconf_sysctl_ignore_routes_with_linkdown,
7196 },
7197 {
7198 .procname = "drop_unicast_in_l2_multicast",
7199 .data = &ipv6_devconf.drop_unicast_in_l2_multicast,
7200 .maxlen = sizeof(int),
7201 .mode = 0644,
7202 .proc_handler = proc_dointvec,
7203 },
7204 {
7205 .procname = "drop_unsolicited_na",
7206 .data = &ipv6_devconf.drop_unsolicited_na,
7207 .maxlen = sizeof(int),
7208 .mode = 0644,
7209 .proc_handler = proc_dointvec,
7210 },
7211 {
7212 .procname = "keep_addr_on_down",
7213 .data = &ipv6_devconf.keep_addr_on_down,
7214 .maxlen = sizeof(int),
7215 .mode = 0644,
7216 .proc_handler = proc_dointvec,
7217
7218 },
7219 {
7220 .procname = "seg6_enabled",
7221 .data = &ipv6_devconf.seg6_enabled,
7222 .maxlen = sizeof(int),
7223 .mode = 0644,
7224 .proc_handler = proc_dointvec,
7225 },
7226 #ifdef CONFIG_IPV6_SEG6_HMAC
7227 {
7228 .procname = "seg6_require_hmac",
7229 .data = &ipv6_devconf.seg6_require_hmac,
7230 .maxlen = sizeof(int),
7231 .mode = 0644,
7232 .proc_handler = proc_dointvec,
7233 },
7234 #endif
7235 {
7236 .procname = "enhanced_dad",
7237 .data = &ipv6_devconf.enhanced_dad,
7238 .maxlen = sizeof(int),
7239 .mode = 0644,
7240 .proc_handler = proc_dointvec,
7241 },
7242 {
7243 .procname = "addr_gen_mode",
7244 .data = &ipv6_devconf.addr_gen_mode,
7245 .maxlen = sizeof(int),
7246 .mode = 0644,
7247 .proc_handler = addrconf_sysctl_addr_gen_mode,
7248 },
7249 {
7250 .procname = "disable_policy",
7251 .data = &ipv6_devconf.disable_policy,
7252 .maxlen = sizeof(int),
7253 .mode = 0644,
7254 .proc_handler = addrconf_sysctl_disable_policy,
7255 },
7256 {
7257 .procname = "ndisc_tclass",
7258 .data = &ipv6_devconf.ndisc_tclass,
7259 .maxlen = sizeof(int),
7260 .mode = 0644,
7261 .proc_handler = proc_dointvec_minmax,
7262 .extra1 = (void *)SYSCTL_ZERO,
7263 .extra2 = (void *)&two_five_five,
7264 },
7265 {
7266 .procname = "rpl_seg_enabled",
7267 .data = &ipv6_devconf.rpl_seg_enabled,
7268 .maxlen = sizeof(int),
7269 .mode = 0644,
7270 .proc_handler = proc_dointvec_minmax,
7271 .extra1 = SYSCTL_ZERO,
7272 .extra2 = SYSCTL_ONE,
7273 },
7274 {
7275 .procname = "ioam6_enabled",
7276 .data = &ipv6_devconf.ioam6_enabled,
7277 .maxlen = sizeof(u8),
7278 .mode = 0644,
7279 .proc_handler = proc_dou8vec_minmax,
7280 .extra1 = (void *)SYSCTL_ZERO,
7281 .extra2 = (void *)SYSCTL_ONE,
7282 },
7283 {
7284 .procname = "ioam6_id",
7285 .data = &ipv6_devconf.ioam6_id,
7286 .maxlen = sizeof(u32),
7287 .mode = 0644,
7288 .proc_handler = proc_douintvec_minmax,
7289 .extra1 = (void *)SYSCTL_ZERO,
7290 .extra2 = (void *)&ioam6_if_id_max,
7291 },
7292 {
7293 .procname = "ioam6_id_wide",
7294 .data = &ipv6_devconf.ioam6_id_wide,
7295 .maxlen = sizeof(u32),
7296 .mode = 0644,
7297 .proc_handler = proc_douintvec,
7298 },
7299 {
7300 .procname = "ndisc_evict_nocarrier",
7301 .data = &ipv6_devconf.ndisc_evict_nocarrier,
7302 .maxlen = sizeof(u8),
7303 .mode = 0644,
7304 .proc_handler = proc_dou8vec_minmax,
7305 .extra1 = (void *)SYSCTL_ZERO,
7306 .extra2 = (void *)SYSCTL_ONE,
7307 },
7308 {
7309 .procname = "accept_untracked_na",
7310 .data = &ipv6_devconf.accept_untracked_na,
7311 .maxlen = sizeof(int),
7312 .mode = 0644,
7313 .proc_handler = proc_dointvec_minmax,
7314 .extra1 = SYSCTL_ZERO,
7315 .extra2 = SYSCTL_TWO,
7316 },
7317 {
7318 .procname = "force_forwarding",
7319 .data = &ipv6_devconf.force_forwarding,
7320 .maxlen = sizeof(int),
7321 .mode = 0644,
7322 .proc_handler = addrconf_sysctl_force_forwarding,
7323 },
7324 };
7325
__addrconf_sysctl_register(struct net * net,char * dev_name,struct inet6_dev * idev,struct ipv6_devconf * p)7326 static int __addrconf_sysctl_register(struct net *net, char *dev_name,
7327 struct inet6_dev *idev, struct ipv6_devconf *p)
7328 {
7329 size_t table_size = ARRAY_SIZE(addrconf_sysctl);
7330 int i, ifindex;
7331 struct ctl_table *table;
7332 char path[sizeof("net/ipv6/conf/") + IFNAMSIZ];
7333
7334 table = kmemdup(addrconf_sysctl, sizeof(addrconf_sysctl), GFP_KERNEL_ACCOUNT);
7335 if (!table)
7336 goto out;
7337
7338 for (i = 0; i < table_size; i++) {
7339 table[i].data += (char *)p - (char *)&ipv6_devconf;
7340 /* If one of these is already set, then it is not safe to
7341 * overwrite either of them: this makes proc_dointvec_minmax
7342 * usable.
7343 */
7344 if (!table[i].extra1 && !table[i].extra2) {
7345 table[i].extra1 = idev; /* embedded; no ref */
7346 table[i].extra2 = net;
7347 }
7348 }
7349
7350 snprintf(path, sizeof(path), "net/ipv6/conf/%s", dev_name);
7351
7352 p->sysctl_header = register_net_sysctl_sz(net, path, table,
7353 table_size);
7354 if (!p->sysctl_header)
7355 goto free;
7356
7357 if (!strcmp(dev_name, "all"))
7358 ifindex = NETCONFA_IFINDEX_ALL;
7359 else if (!strcmp(dev_name, "default"))
7360 ifindex = NETCONFA_IFINDEX_DEFAULT;
7361 else
7362 ifindex = idev->dev->ifindex;
7363 inet6_netconf_notify_devconf(net, RTM_NEWNETCONF, NETCONFA_ALL,
7364 ifindex, p);
7365 return 0;
7366
7367 free:
7368 kfree(table);
7369 out:
7370 return -ENOBUFS;
7371 }
7372
__addrconf_sysctl_unregister(struct net * net,struct ipv6_devconf * p,int ifindex)7373 static void __addrconf_sysctl_unregister(struct net *net,
7374 struct ipv6_devconf *p, int ifindex)
7375 {
7376 const struct ctl_table *table;
7377
7378 if (!p->sysctl_header)
7379 return;
7380
7381 table = p->sysctl_header->ctl_table_arg;
7382 unregister_net_sysctl_table(p->sysctl_header);
7383 p->sysctl_header = NULL;
7384 kfree(table);
7385
7386 inet6_netconf_notify_devconf(net, RTM_DELNETCONF, 0, ifindex, NULL);
7387 }
7388
addrconf_sysctl_register(struct inet6_dev * idev)7389 static int addrconf_sysctl_register(struct inet6_dev *idev)
7390 {
7391 int err;
7392
7393 if (!sysctl_dev_name_is_allowed(idev->dev->name))
7394 return -EINVAL;
7395
7396 err = neigh_sysctl_register(idev->dev, idev->nd_parms,
7397 &ndisc_ifinfo_sysctl_change);
7398 if (err)
7399 return err;
7400 err = __addrconf_sysctl_register(dev_net(idev->dev), idev->dev->name,
7401 idev, &idev->cnf);
7402 if (err)
7403 neigh_sysctl_unregister(idev->nd_parms);
7404
7405 return err;
7406 }
7407
addrconf_sysctl_unregister(struct inet6_dev * idev)7408 static void addrconf_sysctl_unregister(struct inet6_dev *idev)
7409 {
7410 __addrconf_sysctl_unregister(dev_net(idev->dev), &idev->cnf,
7411 idev->dev->ifindex);
7412 neigh_sysctl_unregister(idev->nd_parms);
7413 }
7414
7415
7416 #endif
7417
addrconf_init_net(struct net * net)7418 static int __net_init addrconf_init_net(struct net *net)
7419 {
7420 int err = -ENOMEM;
7421 struct ipv6_devconf *all, *dflt;
7422
7423 spin_lock_init(&net->ipv6.addrconf_hash_lock);
7424 INIT_DEFERRABLE_WORK(&net->ipv6.addr_chk_work, addrconf_verify_work);
7425 net->ipv6.inet6_addr_lst = kzalloc_objs(struct hlist_head,
7426 IN6_ADDR_HSIZE);
7427 if (!net->ipv6.inet6_addr_lst)
7428 goto err_alloc_addr;
7429
7430 all = kmemdup(&ipv6_devconf, sizeof(ipv6_devconf), GFP_KERNEL);
7431 if (!all)
7432 goto err_alloc_all;
7433
7434 dflt = kmemdup(&ipv6_devconf_dflt, sizeof(ipv6_devconf_dflt), GFP_KERNEL);
7435 if (!dflt)
7436 goto err_alloc_dflt;
7437
7438 if (!net_eq(net, &init_net)) {
7439 switch (net_inherit_devconf()) {
7440 case 1: /* copy from init_net */
7441 memcpy(all, init_net.ipv6.devconf_all,
7442 sizeof(ipv6_devconf));
7443 memcpy(dflt, init_net.ipv6.devconf_dflt,
7444 sizeof(ipv6_devconf_dflt));
7445 break;
7446 case 3: /* copy from the current netns */
7447 memcpy(all, current->nsproxy->net_ns->ipv6.devconf_all,
7448 sizeof(ipv6_devconf));
7449 memcpy(dflt,
7450 current->nsproxy->net_ns->ipv6.devconf_dflt,
7451 sizeof(ipv6_devconf_dflt));
7452 break;
7453 case 0:
7454 case 2:
7455 /* use compiled values */
7456 break;
7457 }
7458 }
7459
7460 /* these will be inherited by all namespaces */
7461 dflt->autoconf = ipv6_defaults.autoconf;
7462 dflt->disable_ipv6 = ipv6_defaults.disable_ipv6;
7463
7464 dflt->stable_secret.initialized = false;
7465 all->stable_secret.initialized = false;
7466
7467 net->ipv6.devconf_all = all;
7468 net->ipv6.devconf_dflt = dflt;
7469
7470 #ifdef CONFIG_SYSCTL
7471 err = __addrconf_sysctl_register(net, "all", NULL, all);
7472 if (err < 0)
7473 goto err_reg_all;
7474
7475 err = __addrconf_sysctl_register(net, "default", NULL, dflt);
7476 if (err < 0)
7477 goto err_reg_dflt;
7478 #endif
7479 return 0;
7480
7481 #ifdef CONFIG_SYSCTL
7482 err_reg_dflt:
7483 __addrconf_sysctl_unregister(net, all, NETCONFA_IFINDEX_ALL);
7484 err_reg_all:
7485 kfree(dflt);
7486 net->ipv6.devconf_dflt = NULL;
7487 #endif
7488 err_alloc_dflt:
7489 kfree(all);
7490 net->ipv6.devconf_all = NULL;
7491 err_alloc_all:
7492 kfree(net->ipv6.inet6_addr_lst);
7493 err_alloc_addr:
7494 return err;
7495 }
7496
addrconf_exit_net(struct net * net)7497 static void __net_exit addrconf_exit_net(struct net *net)
7498 {
7499 int i;
7500
7501 #ifdef CONFIG_SYSCTL
7502 __addrconf_sysctl_unregister(net, net->ipv6.devconf_dflt,
7503 NETCONFA_IFINDEX_DEFAULT);
7504 __addrconf_sysctl_unregister(net, net->ipv6.devconf_all,
7505 NETCONFA_IFINDEX_ALL);
7506 #endif
7507 kfree(net->ipv6.devconf_dflt);
7508 net->ipv6.devconf_dflt = NULL;
7509 kfree(net->ipv6.devconf_all);
7510 net->ipv6.devconf_all = NULL;
7511
7512 cancel_delayed_work_sync(&net->ipv6.addr_chk_work);
7513 /*
7514 * Check hash table, then free it.
7515 */
7516 for (i = 0; i < IN6_ADDR_HSIZE; i++)
7517 WARN_ON_ONCE(!hlist_empty(&net->ipv6.inet6_addr_lst[i]));
7518
7519 kfree(net->ipv6.inet6_addr_lst);
7520 net->ipv6.inet6_addr_lst = NULL;
7521 }
7522
7523 static struct pernet_operations addrconf_ops = {
7524 .init = addrconf_init_net,
7525 .exit = addrconf_exit_net,
7526 };
7527
7528 static struct rtnl_af_ops inet6_ops __read_mostly = {
7529 .family = AF_INET6,
7530 .fill_link_af = inet6_fill_link_af,
7531 .get_link_af_size = inet6_get_link_af_size,
7532 .validate_link_af = inet6_validate_link_af,
7533 .set_link_af = inet6_set_link_af,
7534 };
7535
7536 static const struct rtnl_msg_handler addrconf_rtnl_msg_handlers[] __initconst_or_module = {
7537 {.owner = THIS_MODULE, .protocol = PF_INET6, .msgtype = RTM_GETLINK,
7538 .dumpit = inet6_dump_ifinfo, .flags = RTNL_FLAG_DUMP_UNLOCKED},
7539 {.owner = THIS_MODULE, .protocol = PF_INET6, .msgtype = RTM_NEWADDR,
7540 .doit = inet6_rtm_newaddr, .flags = RTNL_FLAG_DOIT_PERNET},
7541 {.owner = THIS_MODULE, .protocol = PF_INET6, .msgtype = RTM_DELADDR,
7542 .doit = inet6_rtm_deladdr, .flags = RTNL_FLAG_DOIT_PERNET},
7543 {.owner = THIS_MODULE, .protocol = PF_INET6, .msgtype = RTM_GETADDR,
7544 .doit = inet6_rtm_getaddr, .dumpit = inet6_dump_ifaddr,
7545 .flags = RTNL_FLAG_DOIT_UNLOCKED | RTNL_FLAG_DUMP_UNLOCKED},
7546 {.owner = THIS_MODULE, .protocol = PF_INET6, .msgtype = RTM_GETMULTICAST,
7547 .dumpit = inet6_dump_ifmcaddr,
7548 .flags = RTNL_FLAG_DUMP_UNLOCKED},
7549 {.owner = THIS_MODULE, .protocol = PF_INET6, .msgtype = RTM_GETANYCAST,
7550 .dumpit = inet6_dump_ifacaddr,
7551 .flags = RTNL_FLAG_DUMP_UNLOCKED},
7552 {.owner = THIS_MODULE, .protocol = PF_INET6, .msgtype = RTM_GETNETCONF,
7553 .doit = inet6_netconf_get_devconf, .dumpit = inet6_netconf_dump_devconf,
7554 .flags = RTNL_FLAG_DOIT_UNLOCKED | RTNL_FLAG_DUMP_UNLOCKED},
7555 };
7556
7557 /*
7558 * Init / cleanup code
7559 */
7560
addrconf_init(void)7561 int __init addrconf_init(void)
7562 {
7563 struct inet6_dev *idev;
7564 int err;
7565
7566 err = ipv6_addr_label_init();
7567 if (err < 0) {
7568 pr_crit("%s: cannot initialize default policy table: %d\n",
7569 __func__, err);
7570 goto out;
7571 }
7572
7573 err = register_pernet_subsys(&addrconf_ops);
7574 if (err < 0)
7575 goto out_addrlabel;
7576
7577 /* All works using addrconf_wq need to lock rtnl. */
7578 addrconf_wq = create_singlethread_workqueue("ipv6_addrconf");
7579 if (!addrconf_wq) {
7580 err = -ENOMEM;
7581 goto out_nowq;
7582 }
7583
7584 rtnl_net_lock(&init_net);
7585 idev = ipv6_add_dev(blackhole_netdev);
7586 rtnl_net_unlock(&init_net);
7587 if (IS_ERR(idev)) {
7588 err = PTR_ERR(idev);
7589 goto errlo;
7590 }
7591
7592 ip6_route_init_special_entries();
7593
7594 register_netdevice_notifier(&ipv6_dev_notf);
7595
7596 addrconf_verify(&init_net);
7597
7598 err = rtnl_af_register(&inet6_ops);
7599 if (err)
7600 goto erraf;
7601
7602 err = rtnl_register_many(addrconf_rtnl_msg_handlers);
7603 if (err)
7604 goto errout;
7605
7606 err = ipv6_addr_label_rtnl_register();
7607 if (err < 0)
7608 goto errout;
7609
7610 return 0;
7611 errout:
7612 rtnl_unregister_all(PF_INET6);
7613 rtnl_af_unregister(&inet6_ops);
7614 erraf:
7615 unregister_netdevice_notifier(&ipv6_dev_notf);
7616 errlo:
7617 destroy_workqueue(addrconf_wq);
7618 out_nowq:
7619 unregister_pernet_subsys(&addrconf_ops);
7620 out_addrlabel:
7621 ipv6_addr_label_cleanup();
7622 out:
7623 return err;
7624 }
7625
addrconf_cleanup(void)7626 void addrconf_cleanup(void)
7627 {
7628 struct net_device *dev;
7629
7630 unregister_netdevice_notifier(&ipv6_dev_notf);
7631 unregister_pernet_subsys(&addrconf_ops);
7632 ipv6_addr_label_cleanup();
7633
7634 rtnl_af_unregister(&inet6_ops);
7635
7636 rtnl_net_lock(&init_net);
7637
7638 /* clean dev list */
7639 for_each_netdev(&init_net, dev) {
7640 if (!__in6_dev_get_rtnl_net(dev))
7641 continue;
7642 addrconf_ifdown(dev, true);
7643 }
7644 addrconf_ifdown(init_net.loopback_dev, true);
7645
7646 rtnl_net_unlock(&init_net);
7647
7648 destroy_workqueue(addrconf_wq);
7649 }
7650