1 // SPDX-License-Identifier: GPL-2.0-or-later
2 /*
3 * IPv6 Address [auto]configuration
4 * Linux INET6 implementation
5 *
6 * Authors:
7 * Pedro Roque <roque@di.fc.ul.pt>
8 * Alexey Kuznetsov <kuznet@ms2.inr.ac.ru>
9 */
10
11 /*
12 * Changes:
13 *
14 * Janos Farkas : delete timer on ifdown
15 * <chexum@bankinf.banki.hu>
16 * Andi Kleen : kill double kfree on module
17 * unload.
18 * Maciej W. Rozycki : FDDI support
19 * sekiya@USAGI : Don't send too many RS
20 * packets.
21 * yoshfuji@USAGI : Fixed interval between DAD
22 * packets.
23 * YOSHIFUJI Hideaki @USAGI : improved accuracy of
24 * address validation timer.
25 * YOSHIFUJI Hideaki @USAGI : Privacy Extensions (RFC3041)
26 * support.
27 * Yuji SEKIYA @USAGI : Don't assign a same IPv6
28 * address on a same interface.
29 * YOSHIFUJI Hideaki @USAGI : ARCnet support
30 * YOSHIFUJI Hideaki @USAGI : convert /proc/net/if_inet6 to
31 * seq_file.
32 * YOSHIFUJI Hideaki @USAGI : improved source address
33 * selection; consider scope,
34 * status etc.
35 */
36
37 #define pr_fmt(fmt) "IPv6: " fmt
38
39 #include <linux/errno.h>
40 #include <linux/types.h>
41 #include <linux/kernel.h>
42 #include <linux/sched/signal.h>
43 #include <linux/socket.h>
44 #include <linux/sockios.h>
45 #include <linux/net.h>
46 #include <linux/inet.h>
47 #include <linux/in6.h>
48 #include <linux/netdevice.h>
49 #include <linux/if_addr.h>
50 #include <linux/if_arp.h>
51 #include <linux/if_arcnet.h>
52 #include <linux/if_infiniband.h>
53 #include <linux/route.h>
54 #include <linux/inetdevice.h>
55 #include <linux/init.h>
56 #include <linux/slab.h>
57 #ifdef CONFIG_SYSCTL
58 #include <linux/sysctl.h>
59 #endif
60 #include <linux/capability.h>
61 #include <linux/delay.h>
62 #include <linux/notifier.h>
63 #include <linux/string.h>
64 #include <linux/hash.h>
65
66 #include <net/net_namespace.h>
67 #include <net/sock.h>
68 #include <net/snmp.h>
69
70 #include <net/6lowpan.h>
71 #include <net/firewire.h>
72 #include <net/ipv6.h>
73 #include <net/protocol.h>
74 #include <net/ndisc.h>
75 #include <net/ip6_route.h>
76 #include <net/addrconf.h>
77 #include <net/tcp.h>
78 #include <net/ip.h>
79 #include <net/netlink.h>
80 #include <net/pkt_sched.h>
81 #include <net/l3mdev.h>
82 #include <linux/if_tunnel.h>
83 #include <linux/rtnetlink.h>
84 #include <linux/netconf.h>
85 #include <linux/random.h>
86 #include <linux/uaccess.h>
87 #include <asm/unaligned.h>
88
89 #include <linux/proc_fs.h>
90 #include <linux/seq_file.h>
91 #include <linux/export.h>
92 #include <linux/ioam6.h>
93
94 #define INFINITY_LIFE_TIME 0xFFFFFFFF
95
96 #define IPV6_MAX_STRLEN \
97 sizeof("ffff:ffff:ffff:ffff:ffff:ffff:255.255.255.255")
98
cstamp_delta(unsigned long cstamp)99 static inline u32 cstamp_delta(unsigned long cstamp)
100 {
101 return (cstamp - INITIAL_JIFFIES) * 100UL / HZ;
102 }
103
rfc3315_s14_backoff_init(s32 irt)104 static inline s32 rfc3315_s14_backoff_init(s32 irt)
105 {
106 /* multiply 'initial retransmission time' by 0.9 .. 1.1 */
107 u64 tmp = get_random_u32_inclusive(900000, 1100000) * (u64)irt;
108 do_div(tmp, 1000000);
109 return (s32)tmp;
110 }
111
rfc3315_s14_backoff_update(s32 rt,s32 mrt)112 static inline s32 rfc3315_s14_backoff_update(s32 rt, s32 mrt)
113 {
114 /* multiply 'retransmission timeout' by 1.9 .. 2.1 */
115 u64 tmp = get_random_u32_inclusive(1900000, 2100000) * (u64)rt;
116 do_div(tmp, 1000000);
117 if ((s32)tmp > mrt) {
118 /* multiply 'maximum retransmission time' by 0.9 .. 1.1 */
119 tmp = get_random_u32_inclusive(900000, 1100000) * (u64)mrt;
120 do_div(tmp, 1000000);
121 }
122 return (s32)tmp;
123 }
124
125 #ifdef CONFIG_SYSCTL
126 static int addrconf_sysctl_register(struct inet6_dev *idev);
127 static void addrconf_sysctl_unregister(struct inet6_dev *idev);
128 #else
addrconf_sysctl_register(struct inet6_dev * idev)129 static inline int addrconf_sysctl_register(struct inet6_dev *idev)
130 {
131 return 0;
132 }
133
addrconf_sysctl_unregister(struct inet6_dev * idev)134 static inline void addrconf_sysctl_unregister(struct inet6_dev *idev)
135 {
136 }
137 #endif
138
139 static void ipv6_gen_rnd_iid(struct in6_addr *addr);
140
141 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev);
142 static int ipv6_count_addresses(const struct inet6_dev *idev);
143 static int ipv6_generate_stable_address(struct in6_addr *addr,
144 u8 dad_count,
145 const struct inet6_dev *idev);
146
147 #define IN6_ADDR_HSIZE_SHIFT 8
148 #define IN6_ADDR_HSIZE (1 << IN6_ADDR_HSIZE_SHIFT)
149
150 static void addrconf_verify(struct net *net);
151 static void addrconf_verify_rtnl(struct net *net);
152
153 static struct workqueue_struct *addrconf_wq;
154
155 static void addrconf_join_anycast(struct inet6_ifaddr *ifp);
156 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp);
157
158 static void addrconf_type_change(struct net_device *dev,
159 unsigned long event);
160 static int addrconf_ifdown(struct net_device *dev, bool unregister);
161
162 static struct fib6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
163 int plen,
164 const struct net_device *dev,
165 u32 flags, u32 noflags,
166 bool no_gw);
167
168 static void addrconf_dad_start(struct inet6_ifaddr *ifp);
169 static void addrconf_dad_work(struct work_struct *w);
170 static void addrconf_dad_completed(struct inet6_ifaddr *ifp, bool bump_id,
171 bool send_na);
172 static void addrconf_dad_run(struct inet6_dev *idev, bool restart);
173 static void addrconf_rs_timer(struct timer_list *t);
174 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
175 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
176
177 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
178 struct prefix_info *pinfo);
179
180 static struct ipv6_devconf ipv6_devconf __read_mostly = {
181 .forwarding = 0,
182 .hop_limit = IPV6_DEFAULT_HOPLIMIT,
183 .mtu6 = IPV6_MIN_MTU,
184 .accept_ra = 1,
185 .accept_redirects = 1,
186 .autoconf = 1,
187 .force_mld_version = 0,
188 .mldv1_unsolicited_report_interval = 10 * HZ,
189 .mldv2_unsolicited_report_interval = HZ,
190 .dad_transmits = 1,
191 .rtr_solicits = MAX_RTR_SOLICITATIONS,
192 .rtr_solicit_interval = RTR_SOLICITATION_INTERVAL,
193 .rtr_solicit_max_interval = RTR_SOLICITATION_MAX_INTERVAL,
194 .rtr_solicit_delay = MAX_RTR_SOLICITATION_DELAY,
195 .use_tempaddr = 0,
196 .temp_valid_lft = TEMP_VALID_LIFETIME,
197 .temp_prefered_lft = TEMP_PREFERRED_LIFETIME,
198 .regen_max_retry = REGEN_MAX_RETRY,
199 .max_desync_factor = MAX_DESYNC_FACTOR,
200 .max_addresses = IPV6_MAX_ADDRESSES,
201 .accept_ra_defrtr = 1,
202 .ra_defrtr_metric = IP6_RT_PRIO_USER,
203 .accept_ra_from_local = 0,
204 .accept_ra_min_hop_limit= 1,
205 .accept_ra_min_lft = 0,
206 .accept_ra_pinfo = 1,
207 #ifdef CONFIG_IPV6_ROUTER_PREF
208 .accept_ra_rtr_pref = 1,
209 .rtr_probe_interval = 60 * HZ,
210 #ifdef CONFIG_IPV6_ROUTE_INFO
211 .accept_ra_rt_info_min_plen = 0,
212 .accept_ra_rt_info_max_plen = 0,
213 #endif
214 #endif
215 .proxy_ndp = 0,
216 .accept_source_route = 0, /* we do not accept RH0 by default. */
217 .disable_ipv6 = 0,
218 .accept_dad = 0,
219 .suppress_frag_ndisc = 1,
220 .accept_ra_mtu = 1,
221 .stable_secret = {
222 .initialized = false,
223 },
224 .use_oif_addrs_only = 0,
225 .ignore_routes_with_linkdown = 0,
226 .keep_addr_on_down = 0,
227 .seg6_enabled = 0,
228 #ifdef CONFIG_IPV6_SEG6_HMAC
229 .seg6_require_hmac = 0,
230 #endif
231 .enhanced_dad = 1,
232 .addr_gen_mode = IN6_ADDR_GEN_MODE_EUI64,
233 .disable_policy = 0,
234 .rpl_seg_enabled = 0,
235 .ioam6_enabled = 0,
236 .ioam6_id = IOAM6_DEFAULT_IF_ID,
237 .ioam6_id_wide = IOAM6_DEFAULT_IF_ID_WIDE,
238 .ndisc_evict_nocarrier = 1,
239 };
240
241 static struct ipv6_devconf ipv6_devconf_dflt __read_mostly = {
242 .forwarding = 0,
243 .hop_limit = IPV6_DEFAULT_HOPLIMIT,
244 .mtu6 = IPV6_MIN_MTU,
245 .accept_ra = 1,
246 .accept_redirects = 1,
247 .autoconf = 1,
248 .force_mld_version = 0,
249 .mldv1_unsolicited_report_interval = 10 * HZ,
250 .mldv2_unsolicited_report_interval = HZ,
251 .dad_transmits = 1,
252 .rtr_solicits = MAX_RTR_SOLICITATIONS,
253 .rtr_solicit_interval = RTR_SOLICITATION_INTERVAL,
254 .rtr_solicit_max_interval = RTR_SOLICITATION_MAX_INTERVAL,
255 .rtr_solicit_delay = MAX_RTR_SOLICITATION_DELAY,
256 .use_tempaddr = 0,
257 .temp_valid_lft = TEMP_VALID_LIFETIME,
258 .temp_prefered_lft = TEMP_PREFERRED_LIFETIME,
259 .regen_max_retry = REGEN_MAX_RETRY,
260 .max_desync_factor = MAX_DESYNC_FACTOR,
261 .max_addresses = IPV6_MAX_ADDRESSES,
262 .accept_ra_defrtr = 1,
263 .ra_defrtr_metric = IP6_RT_PRIO_USER,
264 .accept_ra_from_local = 0,
265 .accept_ra_min_hop_limit= 1,
266 .accept_ra_min_lft = 0,
267 .accept_ra_pinfo = 1,
268 #ifdef CONFIG_IPV6_ROUTER_PREF
269 .accept_ra_rtr_pref = 1,
270 .rtr_probe_interval = 60 * HZ,
271 #ifdef CONFIG_IPV6_ROUTE_INFO
272 .accept_ra_rt_info_min_plen = 0,
273 .accept_ra_rt_info_max_plen = 0,
274 #endif
275 #endif
276 .proxy_ndp = 0,
277 .accept_source_route = 0, /* we do not accept RH0 by default. */
278 .disable_ipv6 = 0,
279 .accept_dad = 1,
280 .suppress_frag_ndisc = 1,
281 .accept_ra_mtu = 1,
282 .stable_secret = {
283 .initialized = false,
284 },
285 .use_oif_addrs_only = 0,
286 .ignore_routes_with_linkdown = 0,
287 .keep_addr_on_down = 0,
288 .seg6_enabled = 0,
289 #ifdef CONFIG_IPV6_SEG6_HMAC
290 .seg6_require_hmac = 0,
291 #endif
292 .enhanced_dad = 1,
293 .addr_gen_mode = IN6_ADDR_GEN_MODE_EUI64,
294 .disable_policy = 0,
295 .rpl_seg_enabled = 0,
296 .ioam6_enabled = 0,
297 .ioam6_id = IOAM6_DEFAULT_IF_ID,
298 .ioam6_id_wide = IOAM6_DEFAULT_IF_ID_WIDE,
299 .ndisc_evict_nocarrier = 1,
300 };
301
302 /* Check if link is ready: is it up and is a valid qdisc available */
addrconf_link_ready(const struct net_device * dev)303 static inline bool addrconf_link_ready(const struct net_device *dev)
304 {
305 return netif_oper_up(dev) && !qdisc_tx_is_noop(dev);
306 }
307
addrconf_del_rs_timer(struct inet6_dev * idev)308 static void addrconf_del_rs_timer(struct inet6_dev *idev)
309 {
310 if (del_timer(&idev->rs_timer))
311 __in6_dev_put(idev);
312 }
313
addrconf_del_dad_work(struct inet6_ifaddr * ifp)314 static void addrconf_del_dad_work(struct inet6_ifaddr *ifp)
315 {
316 if (cancel_delayed_work(&ifp->dad_work))
317 __in6_ifa_put(ifp);
318 }
319
addrconf_mod_rs_timer(struct inet6_dev * idev,unsigned long when)320 static void addrconf_mod_rs_timer(struct inet6_dev *idev,
321 unsigned long when)
322 {
323 if (!mod_timer(&idev->rs_timer, jiffies + when))
324 in6_dev_hold(idev);
325 }
326
addrconf_mod_dad_work(struct inet6_ifaddr * ifp,unsigned long delay)327 static void addrconf_mod_dad_work(struct inet6_ifaddr *ifp,
328 unsigned long delay)
329 {
330 in6_ifa_hold(ifp);
331 if (mod_delayed_work(addrconf_wq, &ifp->dad_work, delay))
332 in6_ifa_put(ifp);
333 }
334
snmp6_alloc_dev(struct inet6_dev * idev)335 static int snmp6_alloc_dev(struct inet6_dev *idev)
336 {
337 int i;
338
339 idev->stats.ipv6 = alloc_percpu_gfp(struct ipstats_mib, GFP_KERNEL_ACCOUNT);
340 if (!idev->stats.ipv6)
341 goto err_ip;
342
343 for_each_possible_cpu(i) {
344 struct ipstats_mib *addrconf_stats;
345 addrconf_stats = per_cpu_ptr(idev->stats.ipv6, i);
346 u64_stats_init(&addrconf_stats->syncp);
347 }
348
349
350 idev->stats.icmpv6dev = kzalloc(sizeof(struct icmpv6_mib_device),
351 GFP_KERNEL);
352 if (!idev->stats.icmpv6dev)
353 goto err_icmp;
354 idev->stats.icmpv6msgdev = kzalloc(sizeof(struct icmpv6msg_mib_device),
355 GFP_KERNEL_ACCOUNT);
356 if (!idev->stats.icmpv6msgdev)
357 goto err_icmpmsg;
358
359 return 0;
360
361 err_icmpmsg:
362 kfree(idev->stats.icmpv6dev);
363 err_icmp:
364 free_percpu(idev->stats.ipv6);
365 err_ip:
366 return -ENOMEM;
367 }
368
ipv6_add_dev(struct net_device * dev)369 static struct inet6_dev *ipv6_add_dev(struct net_device *dev)
370 {
371 struct inet6_dev *ndev;
372 int err = -ENOMEM;
373
374 ASSERT_RTNL();
375
376 if (dev->mtu < IPV6_MIN_MTU && dev != blackhole_netdev)
377 return ERR_PTR(-EINVAL);
378
379 ndev = kzalloc(sizeof(*ndev), GFP_KERNEL_ACCOUNT);
380 if (!ndev)
381 return ERR_PTR(err);
382
383 rwlock_init(&ndev->lock);
384 ndev->dev = dev;
385 INIT_LIST_HEAD(&ndev->addr_list);
386 timer_setup(&ndev->rs_timer, addrconf_rs_timer, 0);
387 memcpy(&ndev->cnf, dev_net(dev)->ipv6.devconf_dflt, sizeof(ndev->cnf));
388
389 if (ndev->cnf.stable_secret.initialized)
390 ndev->cnf.addr_gen_mode = IN6_ADDR_GEN_MODE_STABLE_PRIVACY;
391
392 ndev->cnf.mtu6 = dev->mtu;
393 ndev->ra_mtu = 0;
394 ndev->nd_parms = neigh_parms_alloc(dev, &nd_tbl);
395 if (!ndev->nd_parms) {
396 kfree(ndev);
397 return ERR_PTR(err);
398 }
399 if (ndev->cnf.forwarding)
400 dev_disable_lro(dev);
401 /* We refer to the device */
402 netdev_hold(dev, &ndev->dev_tracker, GFP_KERNEL);
403
404 if (snmp6_alloc_dev(ndev) < 0) {
405 netdev_dbg(dev, "%s: cannot allocate memory for statistics\n",
406 __func__);
407 neigh_parms_release(&nd_tbl, ndev->nd_parms);
408 netdev_put(dev, &ndev->dev_tracker);
409 kfree(ndev);
410 return ERR_PTR(err);
411 }
412
413 if (dev != blackhole_netdev) {
414 if (snmp6_register_dev(ndev) < 0) {
415 netdev_dbg(dev, "%s: cannot create /proc/net/dev_snmp6/%s\n",
416 __func__, dev->name);
417 goto err_release;
418 }
419 }
420 /* One reference from device. */
421 refcount_set(&ndev->refcnt, 1);
422
423 if (dev->flags & (IFF_NOARP | IFF_LOOPBACK))
424 ndev->cnf.accept_dad = -1;
425
426 #if IS_ENABLED(CONFIG_IPV6_SIT)
427 if (dev->type == ARPHRD_SIT && (dev->priv_flags & IFF_ISATAP)) {
428 pr_info("%s: Disabled Multicast RS\n", dev->name);
429 ndev->cnf.rtr_solicits = 0;
430 }
431 #endif
432
433 INIT_LIST_HEAD(&ndev->tempaddr_list);
434 ndev->desync_factor = U32_MAX;
435 if ((dev->flags&IFF_LOOPBACK) ||
436 dev->type == ARPHRD_TUNNEL ||
437 dev->type == ARPHRD_TUNNEL6 ||
438 dev->type == ARPHRD_SIT ||
439 dev->type == ARPHRD_NONE) {
440 ndev->cnf.use_tempaddr = -1;
441 }
442
443 ndev->token = in6addr_any;
444
445 if (netif_running(dev) && addrconf_link_ready(dev))
446 ndev->if_flags |= IF_READY;
447
448 ipv6_mc_init_dev(ndev);
449 ndev->tstamp = jiffies;
450 if (dev != blackhole_netdev) {
451 err = addrconf_sysctl_register(ndev);
452 if (err) {
453 ipv6_mc_destroy_dev(ndev);
454 snmp6_unregister_dev(ndev);
455 goto err_release;
456 }
457 }
458 /* protected by rtnl_lock */
459 rcu_assign_pointer(dev->ip6_ptr, ndev);
460
461 if (dev != blackhole_netdev) {
462 /* Join interface-local all-node multicast group */
463 ipv6_dev_mc_inc(dev, &in6addr_interfacelocal_allnodes);
464
465 /* Join all-node multicast group */
466 ipv6_dev_mc_inc(dev, &in6addr_linklocal_allnodes);
467
468 /* Join all-router multicast group if forwarding is set */
469 if (ndev->cnf.forwarding && (dev->flags & IFF_MULTICAST))
470 ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
471 }
472 return ndev;
473
474 err_release:
475 neigh_parms_release(&nd_tbl, ndev->nd_parms);
476 ndev->dead = 1;
477 in6_dev_finish_destroy(ndev);
478 return ERR_PTR(err);
479 }
480
ipv6_find_idev(struct net_device * dev)481 static struct inet6_dev *ipv6_find_idev(struct net_device *dev)
482 {
483 struct inet6_dev *idev;
484
485 ASSERT_RTNL();
486
487 idev = __in6_dev_get(dev);
488 if (!idev) {
489 idev = ipv6_add_dev(dev);
490 if (IS_ERR(idev))
491 return idev;
492 }
493
494 if (dev->flags&IFF_UP)
495 ipv6_mc_up(idev);
496 return idev;
497 }
498
inet6_netconf_msgsize_devconf(int type)499 static int inet6_netconf_msgsize_devconf(int type)
500 {
501 int size = NLMSG_ALIGN(sizeof(struct netconfmsg))
502 + nla_total_size(4); /* NETCONFA_IFINDEX */
503 bool all = false;
504
505 if (type == NETCONFA_ALL)
506 all = true;
507
508 if (all || type == NETCONFA_FORWARDING)
509 size += nla_total_size(4);
510 #ifdef CONFIG_IPV6_MROUTE
511 if (all || type == NETCONFA_MC_FORWARDING)
512 size += nla_total_size(4);
513 #endif
514 if (all || type == NETCONFA_PROXY_NEIGH)
515 size += nla_total_size(4);
516
517 if (all || type == NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN)
518 size += nla_total_size(4);
519
520 return size;
521 }
522
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)523 static int inet6_netconf_fill_devconf(struct sk_buff *skb, int ifindex,
524 struct ipv6_devconf *devconf, u32 portid,
525 u32 seq, int event, unsigned int flags,
526 int type)
527 {
528 struct nlmsghdr *nlh;
529 struct netconfmsg *ncm;
530 bool all = false;
531
532 nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct netconfmsg),
533 flags);
534 if (!nlh)
535 return -EMSGSIZE;
536
537 if (type == NETCONFA_ALL)
538 all = true;
539
540 ncm = nlmsg_data(nlh);
541 ncm->ncm_family = AF_INET6;
542
543 if (nla_put_s32(skb, NETCONFA_IFINDEX, ifindex) < 0)
544 goto nla_put_failure;
545
546 if (!devconf)
547 goto out;
548
549 if ((all || type == NETCONFA_FORWARDING) &&
550 nla_put_s32(skb, NETCONFA_FORWARDING, devconf->forwarding) < 0)
551 goto nla_put_failure;
552 #ifdef CONFIG_IPV6_MROUTE
553 if ((all || type == NETCONFA_MC_FORWARDING) &&
554 nla_put_s32(skb, NETCONFA_MC_FORWARDING,
555 atomic_read(&devconf->mc_forwarding)) < 0)
556 goto nla_put_failure;
557 #endif
558 if ((all || type == NETCONFA_PROXY_NEIGH) &&
559 nla_put_s32(skb, NETCONFA_PROXY_NEIGH, devconf->proxy_ndp) < 0)
560 goto nla_put_failure;
561
562 if ((all || type == NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN) &&
563 nla_put_s32(skb, NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
564 devconf->ignore_routes_with_linkdown) < 0)
565 goto nla_put_failure;
566
567 out:
568 nlmsg_end(skb, nlh);
569 return 0;
570
571 nla_put_failure:
572 nlmsg_cancel(skb, nlh);
573 return -EMSGSIZE;
574 }
575
inet6_netconf_notify_devconf(struct net * net,int event,int type,int ifindex,struct ipv6_devconf * devconf)576 void inet6_netconf_notify_devconf(struct net *net, int event, int type,
577 int ifindex, struct ipv6_devconf *devconf)
578 {
579 struct sk_buff *skb;
580 int err = -ENOBUFS;
581
582 skb = nlmsg_new(inet6_netconf_msgsize_devconf(type), GFP_KERNEL);
583 if (!skb)
584 goto errout;
585
586 err = inet6_netconf_fill_devconf(skb, ifindex, devconf, 0, 0,
587 event, 0, type);
588 if (err < 0) {
589 /* -EMSGSIZE implies BUG in inet6_netconf_msgsize_devconf() */
590 WARN_ON(err == -EMSGSIZE);
591 kfree_skb(skb);
592 goto errout;
593 }
594 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_NETCONF, NULL, GFP_KERNEL);
595 return;
596 errout:
597 rtnl_set_sk_err(net, RTNLGRP_IPV6_NETCONF, err);
598 }
599
600 static const struct nla_policy devconf_ipv6_policy[NETCONFA_MAX+1] = {
601 [NETCONFA_IFINDEX] = { .len = sizeof(int) },
602 [NETCONFA_FORWARDING] = { .len = sizeof(int) },
603 [NETCONFA_PROXY_NEIGH] = { .len = sizeof(int) },
604 [NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN] = { .len = sizeof(int) },
605 };
606
inet6_netconf_valid_get_req(struct sk_buff * skb,const struct nlmsghdr * nlh,struct nlattr ** tb,struct netlink_ext_ack * extack)607 static int inet6_netconf_valid_get_req(struct sk_buff *skb,
608 const struct nlmsghdr *nlh,
609 struct nlattr **tb,
610 struct netlink_ext_ack *extack)
611 {
612 int i, err;
613
614 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(struct netconfmsg))) {
615 NL_SET_ERR_MSG_MOD(extack, "Invalid header for netconf get request");
616 return -EINVAL;
617 }
618
619 if (!netlink_strict_get_check(skb))
620 return nlmsg_parse_deprecated(nlh, sizeof(struct netconfmsg),
621 tb, NETCONFA_MAX,
622 devconf_ipv6_policy, extack);
623
624 err = nlmsg_parse_deprecated_strict(nlh, sizeof(struct netconfmsg),
625 tb, NETCONFA_MAX,
626 devconf_ipv6_policy, extack);
627 if (err)
628 return err;
629
630 for (i = 0; i <= NETCONFA_MAX; i++) {
631 if (!tb[i])
632 continue;
633
634 switch (i) {
635 case NETCONFA_IFINDEX:
636 break;
637 default:
638 NL_SET_ERR_MSG_MOD(extack, "Unsupported attribute in netconf get request");
639 return -EINVAL;
640 }
641 }
642
643 return 0;
644 }
645
inet6_netconf_get_devconf(struct sk_buff * in_skb,struct nlmsghdr * nlh,struct netlink_ext_ack * extack)646 static int inet6_netconf_get_devconf(struct sk_buff *in_skb,
647 struct nlmsghdr *nlh,
648 struct netlink_ext_ack *extack)
649 {
650 struct net *net = sock_net(in_skb->sk);
651 struct nlattr *tb[NETCONFA_MAX+1];
652 struct inet6_dev *in6_dev = NULL;
653 struct net_device *dev = NULL;
654 struct sk_buff *skb;
655 struct ipv6_devconf *devconf;
656 int ifindex;
657 int err;
658
659 err = inet6_netconf_valid_get_req(in_skb, nlh, tb, extack);
660 if (err < 0)
661 return err;
662
663 if (!tb[NETCONFA_IFINDEX])
664 return -EINVAL;
665
666 err = -EINVAL;
667 ifindex = nla_get_s32(tb[NETCONFA_IFINDEX]);
668 switch (ifindex) {
669 case NETCONFA_IFINDEX_ALL:
670 devconf = net->ipv6.devconf_all;
671 break;
672 case NETCONFA_IFINDEX_DEFAULT:
673 devconf = net->ipv6.devconf_dflt;
674 break;
675 default:
676 dev = dev_get_by_index(net, ifindex);
677 if (!dev)
678 return -EINVAL;
679 in6_dev = in6_dev_get(dev);
680 if (!in6_dev)
681 goto errout;
682 devconf = &in6_dev->cnf;
683 break;
684 }
685
686 err = -ENOBUFS;
687 skb = nlmsg_new(inet6_netconf_msgsize_devconf(NETCONFA_ALL), GFP_KERNEL);
688 if (!skb)
689 goto errout;
690
691 err = inet6_netconf_fill_devconf(skb, ifindex, devconf,
692 NETLINK_CB(in_skb).portid,
693 nlh->nlmsg_seq, RTM_NEWNETCONF, 0,
694 NETCONFA_ALL);
695 if (err < 0) {
696 /* -EMSGSIZE implies BUG in inet6_netconf_msgsize_devconf() */
697 WARN_ON(err == -EMSGSIZE);
698 kfree_skb(skb);
699 goto errout;
700 }
701 err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
702 errout:
703 if (in6_dev)
704 in6_dev_put(in6_dev);
705 dev_put(dev);
706 return err;
707 }
708
709 /* Combine dev_addr_genid and dev_base_seq to detect changes.
710 */
inet6_base_seq(const struct net * net)711 static u32 inet6_base_seq(const struct net *net)
712 {
713 u32 res = atomic_read(&net->ipv6.dev_addr_genid) +
714 net->dev_base_seq;
715
716 /* Must not return 0 (see nl_dump_check_consistent()).
717 * Chose a value far away from 0.
718 */
719 if (!res)
720 res = 0x80000000;
721 return res;
722 }
723
724
inet6_netconf_dump_devconf(struct sk_buff * skb,struct netlink_callback * cb)725 static int inet6_netconf_dump_devconf(struct sk_buff *skb,
726 struct netlink_callback *cb)
727 {
728 const struct nlmsghdr *nlh = cb->nlh;
729 struct net *net = sock_net(skb->sk);
730 int h, s_h;
731 int idx, s_idx;
732 struct net_device *dev;
733 struct inet6_dev *idev;
734 struct hlist_head *head;
735
736 if (cb->strict_check) {
737 struct netlink_ext_ack *extack = cb->extack;
738 struct netconfmsg *ncm;
739
740 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ncm))) {
741 NL_SET_ERR_MSG_MOD(extack, "Invalid header for netconf dump request");
742 return -EINVAL;
743 }
744
745 if (nlmsg_attrlen(nlh, sizeof(*ncm))) {
746 NL_SET_ERR_MSG_MOD(extack, "Invalid data after header in netconf dump request");
747 return -EINVAL;
748 }
749 }
750
751 s_h = cb->args[0];
752 s_idx = idx = cb->args[1];
753
754 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
755 idx = 0;
756 head = &net->dev_index_head[h];
757 rcu_read_lock();
758 cb->seq = inet6_base_seq(net);
759 hlist_for_each_entry_rcu(dev, head, index_hlist) {
760 if (idx < s_idx)
761 goto cont;
762 idev = __in6_dev_get(dev);
763 if (!idev)
764 goto cont;
765
766 if (inet6_netconf_fill_devconf(skb, dev->ifindex,
767 &idev->cnf,
768 NETLINK_CB(cb->skb).portid,
769 nlh->nlmsg_seq,
770 RTM_NEWNETCONF,
771 NLM_F_MULTI,
772 NETCONFA_ALL) < 0) {
773 rcu_read_unlock();
774 goto done;
775 }
776 nl_dump_check_consistent(cb, nlmsg_hdr(skb));
777 cont:
778 idx++;
779 }
780 rcu_read_unlock();
781 }
782 if (h == NETDEV_HASHENTRIES) {
783 if (inet6_netconf_fill_devconf(skb, NETCONFA_IFINDEX_ALL,
784 net->ipv6.devconf_all,
785 NETLINK_CB(cb->skb).portid,
786 nlh->nlmsg_seq,
787 RTM_NEWNETCONF, NLM_F_MULTI,
788 NETCONFA_ALL) < 0)
789 goto done;
790 else
791 h++;
792 }
793 if (h == NETDEV_HASHENTRIES + 1) {
794 if (inet6_netconf_fill_devconf(skb, NETCONFA_IFINDEX_DEFAULT,
795 net->ipv6.devconf_dflt,
796 NETLINK_CB(cb->skb).portid,
797 nlh->nlmsg_seq,
798 RTM_NEWNETCONF, NLM_F_MULTI,
799 NETCONFA_ALL) < 0)
800 goto done;
801 else
802 h++;
803 }
804 done:
805 cb->args[0] = h;
806 cb->args[1] = idx;
807
808 return skb->len;
809 }
810
811 #ifdef CONFIG_SYSCTL
dev_forward_change(struct inet6_dev * idev)812 static void dev_forward_change(struct inet6_dev *idev)
813 {
814 struct net_device *dev;
815 struct inet6_ifaddr *ifa;
816 LIST_HEAD(tmp_addr_list);
817
818 if (!idev)
819 return;
820 dev = idev->dev;
821 if (idev->cnf.forwarding)
822 dev_disable_lro(dev);
823 if (dev->flags & IFF_MULTICAST) {
824 if (idev->cnf.forwarding) {
825 ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
826 ipv6_dev_mc_inc(dev, &in6addr_interfacelocal_allrouters);
827 ipv6_dev_mc_inc(dev, &in6addr_sitelocal_allrouters);
828 } else {
829 ipv6_dev_mc_dec(dev, &in6addr_linklocal_allrouters);
830 ipv6_dev_mc_dec(dev, &in6addr_interfacelocal_allrouters);
831 ipv6_dev_mc_dec(dev, &in6addr_sitelocal_allrouters);
832 }
833 }
834
835 read_lock_bh(&idev->lock);
836 list_for_each_entry(ifa, &idev->addr_list, if_list) {
837 if (ifa->flags&IFA_F_TENTATIVE)
838 continue;
839 list_add_tail(&ifa->if_list_aux, &tmp_addr_list);
840 }
841 read_unlock_bh(&idev->lock);
842
843 while (!list_empty(&tmp_addr_list)) {
844 ifa = list_first_entry(&tmp_addr_list,
845 struct inet6_ifaddr, if_list_aux);
846 list_del(&ifa->if_list_aux);
847 if (idev->cnf.forwarding)
848 addrconf_join_anycast(ifa);
849 else
850 addrconf_leave_anycast(ifa);
851 }
852
853 inet6_netconf_notify_devconf(dev_net(dev), RTM_NEWNETCONF,
854 NETCONFA_FORWARDING,
855 dev->ifindex, &idev->cnf);
856 }
857
858
addrconf_forward_change(struct net * net,__s32 newf)859 static void addrconf_forward_change(struct net *net, __s32 newf)
860 {
861 struct net_device *dev;
862 struct inet6_dev *idev;
863
864 for_each_netdev(net, dev) {
865 idev = __in6_dev_get(dev);
866 if (idev) {
867 int changed = (!idev->cnf.forwarding) ^ (!newf);
868 idev->cnf.forwarding = newf;
869 if (changed)
870 dev_forward_change(idev);
871 }
872 }
873 }
874
addrconf_fixup_forwarding(struct ctl_table * table,int * p,int newf)875 static int addrconf_fixup_forwarding(struct ctl_table *table, int *p, int newf)
876 {
877 struct net *net;
878 int old;
879
880 if (!rtnl_trylock())
881 return restart_syscall();
882
883 net = (struct net *)table->extra2;
884 old = *p;
885 *p = newf;
886
887 if (p == &net->ipv6.devconf_dflt->forwarding) {
888 if ((!newf) ^ (!old))
889 inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
890 NETCONFA_FORWARDING,
891 NETCONFA_IFINDEX_DEFAULT,
892 net->ipv6.devconf_dflt);
893 rtnl_unlock();
894 return 0;
895 }
896
897 if (p == &net->ipv6.devconf_all->forwarding) {
898 int old_dflt = net->ipv6.devconf_dflt->forwarding;
899
900 net->ipv6.devconf_dflt->forwarding = newf;
901 if ((!newf) ^ (!old_dflt))
902 inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
903 NETCONFA_FORWARDING,
904 NETCONFA_IFINDEX_DEFAULT,
905 net->ipv6.devconf_dflt);
906
907 addrconf_forward_change(net, newf);
908 if ((!newf) ^ (!old))
909 inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
910 NETCONFA_FORWARDING,
911 NETCONFA_IFINDEX_ALL,
912 net->ipv6.devconf_all);
913 } else if ((!newf) ^ (!old))
914 dev_forward_change((struct inet6_dev *)table->extra1);
915 rtnl_unlock();
916
917 if (newf)
918 rt6_purge_dflt_routers(net);
919 return 1;
920 }
921
addrconf_linkdown_change(struct net * net,__s32 newf)922 static void addrconf_linkdown_change(struct net *net, __s32 newf)
923 {
924 struct net_device *dev;
925 struct inet6_dev *idev;
926
927 for_each_netdev(net, dev) {
928 idev = __in6_dev_get(dev);
929 if (idev) {
930 int changed = (!idev->cnf.ignore_routes_with_linkdown) ^ (!newf);
931
932 idev->cnf.ignore_routes_with_linkdown = newf;
933 if (changed)
934 inet6_netconf_notify_devconf(dev_net(dev),
935 RTM_NEWNETCONF,
936 NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
937 dev->ifindex,
938 &idev->cnf);
939 }
940 }
941 }
942
addrconf_fixup_linkdown(struct ctl_table * table,int * p,int newf)943 static int addrconf_fixup_linkdown(struct ctl_table *table, int *p, int newf)
944 {
945 struct net *net;
946 int old;
947
948 if (!rtnl_trylock())
949 return restart_syscall();
950
951 net = (struct net *)table->extra2;
952 old = *p;
953 *p = newf;
954
955 if (p == &net->ipv6.devconf_dflt->ignore_routes_with_linkdown) {
956 if ((!newf) ^ (!old))
957 inet6_netconf_notify_devconf(net,
958 RTM_NEWNETCONF,
959 NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
960 NETCONFA_IFINDEX_DEFAULT,
961 net->ipv6.devconf_dflt);
962 rtnl_unlock();
963 return 0;
964 }
965
966 if (p == &net->ipv6.devconf_all->ignore_routes_with_linkdown) {
967 net->ipv6.devconf_dflt->ignore_routes_with_linkdown = newf;
968 addrconf_linkdown_change(net, newf);
969 if ((!newf) ^ (!old))
970 inet6_netconf_notify_devconf(net,
971 RTM_NEWNETCONF,
972 NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
973 NETCONFA_IFINDEX_ALL,
974 net->ipv6.devconf_all);
975 }
976 rtnl_unlock();
977
978 return 1;
979 }
980
981 #endif
982
983 /* Nobody refers to this ifaddr, destroy it */
inet6_ifa_finish_destroy(struct inet6_ifaddr * ifp)984 void inet6_ifa_finish_destroy(struct inet6_ifaddr *ifp)
985 {
986 WARN_ON(!hlist_unhashed(&ifp->addr_lst));
987
988 #ifdef NET_REFCNT_DEBUG
989 pr_debug("%s\n", __func__);
990 #endif
991
992 in6_dev_put(ifp->idev);
993
994 if (cancel_delayed_work(&ifp->dad_work))
995 pr_notice("delayed DAD work was pending while freeing ifa=%p\n",
996 ifp);
997
998 if (ifp->state != INET6_IFADDR_STATE_DEAD) {
999 pr_warn("Freeing alive inet6 address %p\n", ifp);
1000 return;
1001 }
1002
1003 kfree_rcu(ifp, rcu);
1004 }
1005
1006 static void
ipv6_link_dev_addr(struct inet6_dev * idev,struct inet6_ifaddr * ifp)1007 ipv6_link_dev_addr(struct inet6_dev *idev, struct inet6_ifaddr *ifp)
1008 {
1009 struct list_head *p;
1010 int ifp_scope = ipv6_addr_src_scope(&ifp->addr);
1011
1012 /*
1013 * Each device address list is sorted in order of scope -
1014 * global before linklocal.
1015 */
1016 list_for_each(p, &idev->addr_list) {
1017 struct inet6_ifaddr *ifa
1018 = list_entry(p, struct inet6_ifaddr, if_list);
1019 if (ifp_scope >= ipv6_addr_src_scope(&ifa->addr))
1020 break;
1021 }
1022
1023 list_add_tail_rcu(&ifp->if_list, p);
1024 }
1025
inet6_addr_hash(const struct net * net,const struct in6_addr * addr)1026 static u32 inet6_addr_hash(const struct net *net, const struct in6_addr *addr)
1027 {
1028 u32 val = ipv6_addr_hash(addr) ^ net_hash_mix(net);
1029
1030 return hash_32(val, IN6_ADDR_HSIZE_SHIFT);
1031 }
1032
ipv6_chk_same_addr(struct net * net,const struct in6_addr * addr,struct net_device * dev,unsigned int hash)1033 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
1034 struct net_device *dev, unsigned int hash)
1035 {
1036 struct inet6_ifaddr *ifp;
1037
1038 hlist_for_each_entry(ifp, &net->ipv6.inet6_addr_lst[hash], addr_lst) {
1039 if (ipv6_addr_equal(&ifp->addr, addr)) {
1040 if (!dev || ifp->idev->dev == dev)
1041 return true;
1042 }
1043 }
1044 return false;
1045 }
1046
ipv6_add_addr_hash(struct net_device * dev,struct inet6_ifaddr * ifa)1047 static int ipv6_add_addr_hash(struct net_device *dev, struct inet6_ifaddr *ifa)
1048 {
1049 struct net *net = dev_net(dev);
1050 unsigned int hash = inet6_addr_hash(net, &ifa->addr);
1051 int err = 0;
1052
1053 spin_lock_bh(&net->ipv6.addrconf_hash_lock);
1054
1055 /* Ignore adding duplicate addresses on an interface */
1056 if (ipv6_chk_same_addr(net, &ifa->addr, dev, hash)) {
1057 netdev_dbg(dev, "ipv6_add_addr: already assigned\n");
1058 err = -EEXIST;
1059 } else {
1060 hlist_add_head_rcu(&ifa->addr_lst, &net->ipv6.inet6_addr_lst[hash]);
1061 }
1062
1063 spin_unlock_bh(&net->ipv6.addrconf_hash_lock);
1064
1065 return err;
1066 }
1067
1068 /* On success it returns ifp with increased reference count */
1069
1070 static struct inet6_ifaddr *
ipv6_add_addr(struct inet6_dev * idev,struct ifa6_config * cfg,bool can_block,struct netlink_ext_ack * extack)1071 ipv6_add_addr(struct inet6_dev *idev, struct ifa6_config *cfg,
1072 bool can_block, struct netlink_ext_ack *extack)
1073 {
1074 gfp_t gfp_flags = can_block ? GFP_KERNEL : GFP_ATOMIC;
1075 int addr_type = ipv6_addr_type(cfg->pfx);
1076 struct net *net = dev_net(idev->dev);
1077 struct inet6_ifaddr *ifa = NULL;
1078 struct fib6_info *f6i = NULL;
1079 int err = 0;
1080
1081 if (addr_type == IPV6_ADDR_ANY) {
1082 NL_SET_ERR_MSG_MOD(extack, "Invalid address");
1083 return ERR_PTR(-EADDRNOTAVAIL);
1084 } else if (addr_type & IPV6_ADDR_MULTICAST &&
1085 !(cfg->ifa_flags & IFA_F_MCAUTOJOIN)) {
1086 NL_SET_ERR_MSG_MOD(extack, "Cannot assign multicast address without \"IFA_F_MCAUTOJOIN\" flag");
1087 return ERR_PTR(-EADDRNOTAVAIL);
1088 } else if (!(idev->dev->flags & IFF_LOOPBACK) &&
1089 !netif_is_l3_master(idev->dev) &&
1090 addr_type & IPV6_ADDR_LOOPBACK) {
1091 NL_SET_ERR_MSG_MOD(extack, "Cannot assign loopback address on this device");
1092 return ERR_PTR(-EADDRNOTAVAIL);
1093 }
1094
1095 if (idev->dead) {
1096 NL_SET_ERR_MSG_MOD(extack, "device is going away");
1097 err = -ENODEV;
1098 goto out;
1099 }
1100
1101 if (idev->cnf.disable_ipv6) {
1102 NL_SET_ERR_MSG_MOD(extack, "IPv6 is disabled on this device");
1103 err = -EACCES;
1104 goto out;
1105 }
1106
1107 /* validator notifier needs to be blocking;
1108 * do not call in atomic context
1109 */
1110 if (can_block) {
1111 struct in6_validator_info i6vi = {
1112 .i6vi_addr = *cfg->pfx,
1113 .i6vi_dev = idev,
1114 .extack = extack,
1115 };
1116
1117 err = inet6addr_validator_notifier_call_chain(NETDEV_UP, &i6vi);
1118 err = notifier_to_errno(err);
1119 if (err < 0)
1120 goto out;
1121 }
1122
1123 ifa = kzalloc(sizeof(*ifa), gfp_flags | __GFP_ACCOUNT);
1124 if (!ifa) {
1125 err = -ENOBUFS;
1126 goto out;
1127 }
1128
1129 f6i = addrconf_f6i_alloc(net, idev, cfg->pfx, false, gfp_flags, extack);
1130 if (IS_ERR(f6i)) {
1131 err = PTR_ERR(f6i);
1132 f6i = NULL;
1133 goto out;
1134 }
1135
1136 neigh_parms_data_state_setall(idev->nd_parms);
1137
1138 ifa->addr = *cfg->pfx;
1139 if (cfg->peer_pfx)
1140 ifa->peer_addr = *cfg->peer_pfx;
1141
1142 spin_lock_init(&ifa->lock);
1143 INIT_DELAYED_WORK(&ifa->dad_work, addrconf_dad_work);
1144 INIT_HLIST_NODE(&ifa->addr_lst);
1145 ifa->scope = cfg->scope;
1146 ifa->prefix_len = cfg->plen;
1147 ifa->rt_priority = cfg->rt_priority;
1148 ifa->flags = cfg->ifa_flags;
1149 ifa->ifa_proto = cfg->ifa_proto;
1150 /* No need to add the TENTATIVE flag for addresses with NODAD */
1151 if (!(cfg->ifa_flags & IFA_F_NODAD))
1152 ifa->flags |= IFA_F_TENTATIVE;
1153 ifa->valid_lft = cfg->valid_lft;
1154 ifa->prefered_lft = cfg->preferred_lft;
1155 ifa->cstamp = ifa->tstamp = jiffies;
1156 ifa->tokenized = false;
1157
1158 ifa->rt = f6i;
1159
1160 ifa->idev = idev;
1161 in6_dev_hold(idev);
1162
1163 /* For caller */
1164 refcount_set(&ifa->refcnt, 1);
1165
1166 rcu_read_lock();
1167
1168 err = ipv6_add_addr_hash(idev->dev, ifa);
1169 if (err < 0) {
1170 rcu_read_unlock();
1171 goto out;
1172 }
1173
1174 write_lock_bh(&idev->lock);
1175
1176 /* Add to inet6_dev unicast addr list. */
1177 ipv6_link_dev_addr(idev, ifa);
1178
1179 if (ifa->flags&IFA_F_TEMPORARY) {
1180 list_add(&ifa->tmp_list, &idev->tempaddr_list);
1181 in6_ifa_hold(ifa);
1182 }
1183
1184 in6_ifa_hold(ifa);
1185 write_unlock_bh(&idev->lock);
1186
1187 rcu_read_unlock();
1188
1189 inet6addr_notifier_call_chain(NETDEV_UP, ifa);
1190 out:
1191 if (unlikely(err < 0)) {
1192 fib6_info_release(f6i);
1193
1194 if (ifa) {
1195 if (ifa->idev)
1196 in6_dev_put(ifa->idev);
1197 kfree(ifa);
1198 }
1199 ifa = ERR_PTR(err);
1200 }
1201
1202 return ifa;
1203 }
1204
1205 enum cleanup_prefix_rt_t {
1206 CLEANUP_PREFIX_RT_NOP, /* no cleanup action for prefix route */
1207 CLEANUP_PREFIX_RT_DEL, /* delete the prefix route */
1208 CLEANUP_PREFIX_RT_EXPIRE, /* update the lifetime of the prefix route */
1209 };
1210
1211 /*
1212 * Check, whether the prefix for ifp would still need a prefix route
1213 * after deleting ifp. The function returns one of the CLEANUP_PREFIX_RT_*
1214 * constants.
1215 *
1216 * 1) we don't purge prefix if address was not permanent.
1217 * prefix is managed by its own lifetime.
1218 * 2) we also don't purge, if the address was IFA_F_NOPREFIXROUTE.
1219 * 3) if there are no addresses, delete prefix.
1220 * 4) if there are still other permanent address(es),
1221 * corresponding prefix is still permanent.
1222 * 5) if there are still other addresses with IFA_F_NOPREFIXROUTE,
1223 * don't purge the prefix, assume user space is managing it.
1224 * 6) otherwise, update prefix lifetime to the
1225 * longest valid lifetime among the corresponding
1226 * addresses on the device.
1227 * Note: subsequent RA will update lifetime.
1228 **/
1229 static enum cleanup_prefix_rt_t
check_cleanup_prefix_route(struct inet6_ifaddr * ifp,unsigned long * expires)1230 check_cleanup_prefix_route(struct inet6_ifaddr *ifp, unsigned long *expires)
1231 {
1232 struct inet6_ifaddr *ifa;
1233 struct inet6_dev *idev = ifp->idev;
1234 unsigned long lifetime;
1235 enum cleanup_prefix_rt_t action = CLEANUP_PREFIX_RT_DEL;
1236
1237 *expires = jiffies;
1238
1239 list_for_each_entry(ifa, &idev->addr_list, if_list) {
1240 if (ifa == ifp)
1241 continue;
1242 if (ifa->prefix_len != ifp->prefix_len ||
1243 !ipv6_prefix_equal(&ifa->addr, &ifp->addr,
1244 ifp->prefix_len))
1245 continue;
1246 if (ifa->flags & (IFA_F_PERMANENT | IFA_F_NOPREFIXROUTE))
1247 return CLEANUP_PREFIX_RT_NOP;
1248
1249 action = CLEANUP_PREFIX_RT_EXPIRE;
1250
1251 spin_lock(&ifa->lock);
1252
1253 lifetime = addrconf_timeout_fixup(ifa->valid_lft, HZ);
1254 /*
1255 * Note: Because this address is
1256 * not permanent, lifetime <
1257 * LONG_MAX / HZ here.
1258 */
1259 if (time_before(*expires, ifa->tstamp + lifetime * HZ))
1260 *expires = ifa->tstamp + lifetime * HZ;
1261 spin_unlock(&ifa->lock);
1262 }
1263
1264 return action;
1265 }
1266
1267 static void
cleanup_prefix_route(struct inet6_ifaddr * ifp,unsigned long expires,bool del_rt,bool del_peer)1268 cleanup_prefix_route(struct inet6_ifaddr *ifp, unsigned long expires,
1269 bool del_rt, bool del_peer)
1270 {
1271 struct fib6_info *f6i;
1272
1273 f6i = addrconf_get_prefix_route(del_peer ? &ifp->peer_addr : &ifp->addr,
1274 ifp->prefix_len,
1275 ifp->idev->dev, 0, RTF_DEFAULT, true);
1276 if (f6i) {
1277 if (del_rt)
1278 ip6_del_rt(dev_net(ifp->idev->dev), f6i, false);
1279 else {
1280 if (!(f6i->fib6_flags & RTF_EXPIRES))
1281 fib6_set_expires(f6i, expires);
1282 fib6_info_release(f6i);
1283 }
1284 }
1285 }
1286
1287
1288 /* This function wants to get referenced ifp and releases it before return */
1289
ipv6_del_addr(struct inet6_ifaddr * ifp)1290 static void ipv6_del_addr(struct inet6_ifaddr *ifp)
1291 {
1292 enum cleanup_prefix_rt_t action = CLEANUP_PREFIX_RT_NOP;
1293 struct net *net = dev_net(ifp->idev->dev);
1294 unsigned long expires;
1295 int state;
1296
1297 ASSERT_RTNL();
1298
1299 spin_lock_bh(&ifp->lock);
1300 state = ifp->state;
1301 ifp->state = INET6_IFADDR_STATE_DEAD;
1302 spin_unlock_bh(&ifp->lock);
1303
1304 if (state == INET6_IFADDR_STATE_DEAD)
1305 goto out;
1306
1307 spin_lock_bh(&net->ipv6.addrconf_hash_lock);
1308 hlist_del_init_rcu(&ifp->addr_lst);
1309 spin_unlock_bh(&net->ipv6.addrconf_hash_lock);
1310
1311 write_lock_bh(&ifp->idev->lock);
1312
1313 if (ifp->flags&IFA_F_TEMPORARY) {
1314 list_del(&ifp->tmp_list);
1315 if (ifp->ifpub) {
1316 in6_ifa_put(ifp->ifpub);
1317 ifp->ifpub = NULL;
1318 }
1319 __in6_ifa_put(ifp);
1320 }
1321
1322 if (ifp->flags & IFA_F_PERMANENT && !(ifp->flags & IFA_F_NOPREFIXROUTE))
1323 action = check_cleanup_prefix_route(ifp, &expires);
1324
1325 list_del_rcu(&ifp->if_list);
1326 __in6_ifa_put(ifp);
1327
1328 write_unlock_bh(&ifp->idev->lock);
1329
1330 addrconf_del_dad_work(ifp);
1331
1332 ipv6_ifa_notify(RTM_DELADDR, ifp);
1333
1334 inet6addr_notifier_call_chain(NETDEV_DOWN, ifp);
1335
1336 if (action != CLEANUP_PREFIX_RT_NOP) {
1337 cleanup_prefix_route(ifp, expires,
1338 action == CLEANUP_PREFIX_RT_DEL, false);
1339 }
1340
1341 /* clean up prefsrc entries */
1342 rt6_remove_prefsrc(ifp);
1343 out:
1344 in6_ifa_put(ifp);
1345 }
1346
ipv6_create_tempaddr(struct inet6_ifaddr * ifp,bool block)1347 static int ipv6_create_tempaddr(struct inet6_ifaddr *ifp, bool block)
1348 {
1349 struct inet6_dev *idev = ifp->idev;
1350 unsigned long tmp_tstamp, age;
1351 unsigned long regen_advance;
1352 unsigned long now = jiffies;
1353 s32 cnf_temp_preferred_lft;
1354 struct inet6_ifaddr *ift;
1355 struct ifa6_config cfg;
1356 long max_desync_factor;
1357 struct in6_addr addr;
1358 int ret = 0;
1359
1360 write_lock_bh(&idev->lock);
1361
1362 retry:
1363 in6_dev_hold(idev);
1364 if (idev->cnf.use_tempaddr <= 0) {
1365 write_unlock_bh(&idev->lock);
1366 pr_info("%s: use_tempaddr is disabled\n", __func__);
1367 in6_dev_put(idev);
1368 ret = -1;
1369 goto out;
1370 }
1371 spin_lock_bh(&ifp->lock);
1372 if (ifp->regen_count++ >= idev->cnf.regen_max_retry) {
1373 idev->cnf.use_tempaddr = -1; /*XXX*/
1374 spin_unlock_bh(&ifp->lock);
1375 write_unlock_bh(&idev->lock);
1376 pr_warn("%s: regeneration time exceeded - disabled temporary address support\n",
1377 __func__);
1378 in6_dev_put(idev);
1379 ret = -1;
1380 goto out;
1381 }
1382 in6_ifa_hold(ifp);
1383 memcpy(addr.s6_addr, ifp->addr.s6_addr, 8);
1384 ipv6_gen_rnd_iid(&addr);
1385
1386 age = (now - ifp->tstamp) / HZ;
1387
1388 regen_advance = idev->cnf.regen_max_retry *
1389 idev->cnf.dad_transmits *
1390 max(NEIGH_VAR(idev->nd_parms, RETRANS_TIME), HZ/100) / HZ;
1391
1392 /* recalculate max_desync_factor each time and update
1393 * idev->desync_factor if it's larger
1394 */
1395 cnf_temp_preferred_lft = READ_ONCE(idev->cnf.temp_prefered_lft);
1396 max_desync_factor = min_t(long,
1397 idev->cnf.max_desync_factor,
1398 cnf_temp_preferred_lft - regen_advance);
1399
1400 if (unlikely(idev->desync_factor > max_desync_factor)) {
1401 if (max_desync_factor > 0) {
1402 get_random_bytes(&idev->desync_factor,
1403 sizeof(idev->desync_factor));
1404 idev->desync_factor %= max_desync_factor;
1405 } else {
1406 idev->desync_factor = 0;
1407 }
1408 }
1409
1410 memset(&cfg, 0, sizeof(cfg));
1411 cfg.valid_lft = min_t(__u32, ifp->valid_lft,
1412 idev->cnf.temp_valid_lft + age);
1413 cfg.preferred_lft = cnf_temp_preferred_lft + age - idev->desync_factor;
1414 cfg.preferred_lft = min_t(__u32, ifp->prefered_lft, cfg.preferred_lft);
1415
1416 cfg.plen = ifp->prefix_len;
1417 tmp_tstamp = ifp->tstamp;
1418 spin_unlock_bh(&ifp->lock);
1419
1420 write_unlock_bh(&idev->lock);
1421
1422 /* A temporary address is created only if this calculated Preferred
1423 * Lifetime is greater than REGEN_ADVANCE time units. In particular,
1424 * an implementation must not create a temporary address with a zero
1425 * Preferred Lifetime.
1426 * Use age calculation as in addrconf_verify to avoid unnecessary
1427 * temporary addresses being generated.
1428 */
1429 age = (now - tmp_tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
1430 if (cfg.preferred_lft <= regen_advance + age) {
1431 in6_ifa_put(ifp);
1432 in6_dev_put(idev);
1433 ret = -1;
1434 goto out;
1435 }
1436
1437 cfg.ifa_flags = IFA_F_TEMPORARY;
1438 /* set in addrconf_prefix_rcv() */
1439 if (ifp->flags & IFA_F_OPTIMISTIC)
1440 cfg.ifa_flags |= IFA_F_OPTIMISTIC;
1441
1442 cfg.pfx = &addr;
1443 cfg.scope = ipv6_addr_scope(cfg.pfx);
1444
1445 ift = ipv6_add_addr(idev, &cfg, block, NULL);
1446 if (IS_ERR(ift)) {
1447 in6_ifa_put(ifp);
1448 in6_dev_put(idev);
1449 pr_info("%s: retry temporary address regeneration\n", __func__);
1450 write_lock_bh(&idev->lock);
1451 goto retry;
1452 }
1453
1454 spin_lock_bh(&ift->lock);
1455 ift->ifpub = ifp;
1456 ift->cstamp = now;
1457 ift->tstamp = tmp_tstamp;
1458 spin_unlock_bh(&ift->lock);
1459
1460 addrconf_dad_start(ift);
1461 in6_ifa_put(ift);
1462 in6_dev_put(idev);
1463 out:
1464 return ret;
1465 }
1466
1467 /*
1468 * Choose an appropriate source address (RFC3484)
1469 */
1470 enum {
1471 IPV6_SADDR_RULE_INIT = 0,
1472 IPV6_SADDR_RULE_LOCAL,
1473 IPV6_SADDR_RULE_SCOPE,
1474 IPV6_SADDR_RULE_PREFERRED,
1475 #ifdef CONFIG_IPV6_MIP6
1476 IPV6_SADDR_RULE_HOA,
1477 #endif
1478 IPV6_SADDR_RULE_OIF,
1479 IPV6_SADDR_RULE_LABEL,
1480 IPV6_SADDR_RULE_PRIVACY,
1481 IPV6_SADDR_RULE_ORCHID,
1482 IPV6_SADDR_RULE_PREFIX,
1483 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1484 IPV6_SADDR_RULE_NOT_OPTIMISTIC,
1485 #endif
1486 IPV6_SADDR_RULE_MAX
1487 };
1488
1489 struct ipv6_saddr_score {
1490 int rule;
1491 int addr_type;
1492 struct inet6_ifaddr *ifa;
1493 DECLARE_BITMAP(scorebits, IPV6_SADDR_RULE_MAX);
1494 int scopedist;
1495 int matchlen;
1496 };
1497
1498 struct ipv6_saddr_dst {
1499 const struct in6_addr *addr;
1500 int ifindex;
1501 int scope;
1502 int label;
1503 unsigned int prefs;
1504 };
1505
ipv6_saddr_preferred(int type)1506 static inline int ipv6_saddr_preferred(int type)
1507 {
1508 if (type & (IPV6_ADDR_MAPPED|IPV6_ADDR_COMPATv4|IPV6_ADDR_LOOPBACK))
1509 return 1;
1510 return 0;
1511 }
1512
ipv6_use_optimistic_addr(struct net * net,struct inet6_dev * idev)1513 static bool ipv6_use_optimistic_addr(struct net *net,
1514 struct inet6_dev *idev)
1515 {
1516 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1517 if (!idev)
1518 return false;
1519 if (!net->ipv6.devconf_all->optimistic_dad && !idev->cnf.optimistic_dad)
1520 return false;
1521 if (!net->ipv6.devconf_all->use_optimistic && !idev->cnf.use_optimistic)
1522 return false;
1523
1524 return true;
1525 #else
1526 return false;
1527 #endif
1528 }
1529
ipv6_allow_optimistic_dad(struct net * net,struct inet6_dev * idev)1530 static bool ipv6_allow_optimistic_dad(struct net *net,
1531 struct inet6_dev *idev)
1532 {
1533 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1534 if (!idev)
1535 return false;
1536 if (!net->ipv6.devconf_all->optimistic_dad && !idev->cnf.optimistic_dad)
1537 return false;
1538
1539 return true;
1540 #else
1541 return false;
1542 #endif
1543 }
1544
ipv6_get_saddr_eval(struct net * net,struct ipv6_saddr_score * score,struct ipv6_saddr_dst * dst,int i)1545 static int ipv6_get_saddr_eval(struct net *net,
1546 struct ipv6_saddr_score *score,
1547 struct ipv6_saddr_dst *dst,
1548 int i)
1549 {
1550 int ret;
1551
1552 if (i <= score->rule) {
1553 switch (i) {
1554 case IPV6_SADDR_RULE_SCOPE:
1555 ret = score->scopedist;
1556 break;
1557 case IPV6_SADDR_RULE_PREFIX:
1558 ret = score->matchlen;
1559 break;
1560 default:
1561 ret = !!test_bit(i, score->scorebits);
1562 }
1563 goto out;
1564 }
1565
1566 switch (i) {
1567 case IPV6_SADDR_RULE_INIT:
1568 /* Rule 0: remember if hiscore is not ready yet */
1569 ret = !!score->ifa;
1570 break;
1571 case IPV6_SADDR_RULE_LOCAL:
1572 /* Rule 1: Prefer same address */
1573 ret = ipv6_addr_equal(&score->ifa->addr, dst->addr);
1574 break;
1575 case IPV6_SADDR_RULE_SCOPE:
1576 /* Rule 2: Prefer appropriate scope
1577 *
1578 * ret
1579 * ^
1580 * -1 | d 15
1581 * ---+--+-+---> scope
1582 * |
1583 * | d is scope of the destination.
1584 * B-d | \
1585 * | \ <- smaller scope is better if
1586 * B-15 | \ if scope is enough for destination.
1587 * | ret = B - scope (-1 <= scope >= d <= 15).
1588 * d-C-1 | /
1589 * |/ <- greater is better
1590 * -C / if scope is not enough for destination.
1591 * /| ret = scope - C (-1 <= d < scope <= 15).
1592 *
1593 * d - C - 1 < B -15 (for all -1 <= d <= 15).
1594 * C > d + 14 - B >= 15 + 14 - B = 29 - B.
1595 * Assume B = 0 and we get C > 29.
1596 */
1597 ret = __ipv6_addr_src_scope(score->addr_type);
1598 if (ret >= dst->scope)
1599 ret = -ret;
1600 else
1601 ret -= 128; /* 30 is enough */
1602 score->scopedist = ret;
1603 break;
1604 case IPV6_SADDR_RULE_PREFERRED:
1605 {
1606 /* Rule 3: Avoid deprecated and optimistic addresses */
1607 u8 avoid = IFA_F_DEPRECATED;
1608
1609 if (!ipv6_use_optimistic_addr(net, score->ifa->idev))
1610 avoid |= IFA_F_OPTIMISTIC;
1611 ret = ipv6_saddr_preferred(score->addr_type) ||
1612 !(score->ifa->flags & avoid);
1613 break;
1614 }
1615 #ifdef CONFIG_IPV6_MIP6
1616 case IPV6_SADDR_RULE_HOA:
1617 {
1618 /* Rule 4: Prefer home address */
1619 int prefhome = !(dst->prefs & IPV6_PREFER_SRC_COA);
1620 ret = !(score->ifa->flags & IFA_F_HOMEADDRESS) ^ prefhome;
1621 break;
1622 }
1623 #endif
1624 case IPV6_SADDR_RULE_OIF:
1625 /* Rule 5: Prefer outgoing interface */
1626 ret = (!dst->ifindex ||
1627 dst->ifindex == score->ifa->idev->dev->ifindex);
1628 break;
1629 case IPV6_SADDR_RULE_LABEL:
1630 /* Rule 6: Prefer matching label */
1631 ret = ipv6_addr_label(net,
1632 &score->ifa->addr, score->addr_type,
1633 score->ifa->idev->dev->ifindex) == dst->label;
1634 break;
1635 case IPV6_SADDR_RULE_PRIVACY:
1636 {
1637 /* Rule 7: Prefer public address
1638 * Note: prefer temporary address if use_tempaddr >= 2
1639 */
1640 int preftmp = dst->prefs & (IPV6_PREFER_SRC_PUBLIC|IPV6_PREFER_SRC_TMP) ?
1641 !!(dst->prefs & IPV6_PREFER_SRC_TMP) :
1642 score->ifa->idev->cnf.use_tempaddr >= 2;
1643 ret = (!(score->ifa->flags & IFA_F_TEMPORARY)) ^ preftmp;
1644 break;
1645 }
1646 case IPV6_SADDR_RULE_ORCHID:
1647 /* Rule 8-: Prefer ORCHID vs ORCHID or
1648 * non-ORCHID vs non-ORCHID
1649 */
1650 ret = !(ipv6_addr_orchid(&score->ifa->addr) ^
1651 ipv6_addr_orchid(dst->addr));
1652 break;
1653 case IPV6_SADDR_RULE_PREFIX:
1654 /* Rule 8: Use longest matching prefix */
1655 ret = ipv6_addr_diff(&score->ifa->addr, dst->addr);
1656 if (ret > score->ifa->prefix_len)
1657 ret = score->ifa->prefix_len;
1658 score->matchlen = ret;
1659 break;
1660 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1661 case IPV6_SADDR_RULE_NOT_OPTIMISTIC:
1662 /* Optimistic addresses still have lower precedence than other
1663 * preferred addresses.
1664 */
1665 ret = !(score->ifa->flags & IFA_F_OPTIMISTIC);
1666 break;
1667 #endif
1668 default:
1669 ret = 0;
1670 }
1671
1672 if (ret)
1673 __set_bit(i, score->scorebits);
1674 score->rule = i;
1675 out:
1676 return ret;
1677 }
1678
__ipv6_dev_get_saddr(struct net * net,struct ipv6_saddr_dst * dst,struct inet6_dev * idev,struct ipv6_saddr_score * scores,int hiscore_idx)1679 static int __ipv6_dev_get_saddr(struct net *net,
1680 struct ipv6_saddr_dst *dst,
1681 struct inet6_dev *idev,
1682 struct ipv6_saddr_score *scores,
1683 int hiscore_idx)
1684 {
1685 struct ipv6_saddr_score *score = &scores[1 - hiscore_idx], *hiscore = &scores[hiscore_idx];
1686
1687 list_for_each_entry_rcu(score->ifa, &idev->addr_list, if_list) {
1688 int i;
1689
1690 /*
1691 * - Tentative Address (RFC2462 section 5.4)
1692 * - A tentative address is not considered
1693 * "assigned to an interface" in the traditional
1694 * sense, unless it is also flagged as optimistic.
1695 * - Candidate Source Address (section 4)
1696 * - In any case, anycast addresses, multicast
1697 * addresses, and the unspecified address MUST
1698 * NOT be included in a candidate set.
1699 */
1700 if ((score->ifa->flags & IFA_F_TENTATIVE) &&
1701 (!(score->ifa->flags & IFA_F_OPTIMISTIC)))
1702 continue;
1703
1704 score->addr_type = __ipv6_addr_type(&score->ifa->addr);
1705
1706 if (unlikely(score->addr_type == IPV6_ADDR_ANY ||
1707 score->addr_type & IPV6_ADDR_MULTICAST)) {
1708 net_dbg_ratelimited("ADDRCONF: unspecified / multicast address assigned as unicast address on %s",
1709 idev->dev->name);
1710 continue;
1711 }
1712
1713 score->rule = -1;
1714 bitmap_zero(score->scorebits, IPV6_SADDR_RULE_MAX);
1715
1716 for (i = 0; i < IPV6_SADDR_RULE_MAX; i++) {
1717 int minihiscore, miniscore;
1718
1719 minihiscore = ipv6_get_saddr_eval(net, hiscore, dst, i);
1720 miniscore = ipv6_get_saddr_eval(net, score, dst, i);
1721
1722 if (minihiscore > miniscore) {
1723 if (i == IPV6_SADDR_RULE_SCOPE &&
1724 score->scopedist > 0) {
1725 /*
1726 * special case:
1727 * each remaining entry
1728 * has too small (not enough)
1729 * scope, because ifa entries
1730 * are sorted by their scope
1731 * values.
1732 */
1733 goto out;
1734 }
1735 break;
1736 } else if (minihiscore < miniscore) {
1737 swap(hiscore, score);
1738 hiscore_idx = 1 - hiscore_idx;
1739
1740 /* restore our iterator */
1741 score->ifa = hiscore->ifa;
1742
1743 break;
1744 }
1745 }
1746 }
1747 out:
1748 return hiscore_idx;
1749 }
1750
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)1751 static int ipv6_get_saddr_master(struct net *net,
1752 const struct net_device *dst_dev,
1753 const struct net_device *master,
1754 struct ipv6_saddr_dst *dst,
1755 struct ipv6_saddr_score *scores,
1756 int hiscore_idx)
1757 {
1758 struct inet6_dev *idev;
1759
1760 idev = __in6_dev_get(dst_dev);
1761 if (idev)
1762 hiscore_idx = __ipv6_dev_get_saddr(net, dst, idev,
1763 scores, hiscore_idx);
1764
1765 idev = __in6_dev_get(master);
1766 if (idev)
1767 hiscore_idx = __ipv6_dev_get_saddr(net, dst, idev,
1768 scores, hiscore_idx);
1769
1770 return hiscore_idx;
1771 }
1772
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)1773 int ipv6_dev_get_saddr(struct net *net, const struct net_device *dst_dev,
1774 const struct in6_addr *daddr, unsigned int prefs,
1775 struct in6_addr *saddr)
1776 {
1777 struct ipv6_saddr_score scores[2], *hiscore;
1778 struct ipv6_saddr_dst dst;
1779 struct inet6_dev *idev;
1780 struct net_device *dev;
1781 int dst_type;
1782 bool use_oif_addr = false;
1783 int hiscore_idx = 0;
1784 int ret = 0;
1785
1786 dst_type = __ipv6_addr_type(daddr);
1787 dst.addr = daddr;
1788 dst.ifindex = dst_dev ? dst_dev->ifindex : 0;
1789 dst.scope = __ipv6_addr_src_scope(dst_type);
1790 dst.label = ipv6_addr_label(net, daddr, dst_type, dst.ifindex);
1791 dst.prefs = prefs;
1792
1793 scores[hiscore_idx].rule = -1;
1794 scores[hiscore_idx].ifa = NULL;
1795
1796 rcu_read_lock();
1797
1798 /* Candidate Source Address (section 4)
1799 * - multicast and link-local destination address,
1800 * the set of candidate source address MUST only
1801 * include addresses assigned to interfaces
1802 * belonging to the same link as the outgoing
1803 * interface.
1804 * (- For site-local destination addresses, the
1805 * set of candidate source addresses MUST only
1806 * include addresses assigned to interfaces
1807 * belonging to the same site as the outgoing
1808 * interface.)
1809 * - "It is RECOMMENDED that the candidate source addresses
1810 * be the set of unicast addresses assigned to the
1811 * interface that will be used to send to the destination
1812 * (the 'outgoing' interface)." (RFC 6724)
1813 */
1814 if (dst_dev) {
1815 idev = __in6_dev_get(dst_dev);
1816 if ((dst_type & IPV6_ADDR_MULTICAST) ||
1817 dst.scope <= IPV6_ADDR_SCOPE_LINKLOCAL ||
1818 (idev && idev->cnf.use_oif_addrs_only)) {
1819 use_oif_addr = true;
1820 }
1821 }
1822
1823 if (use_oif_addr) {
1824 if (idev)
1825 hiscore_idx = __ipv6_dev_get_saddr(net, &dst, idev, scores, hiscore_idx);
1826 } else {
1827 const struct net_device *master;
1828 int master_idx = 0;
1829
1830 /* if dst_dev exists and is enslaved to an L3 device, then
1831 * prefer addresses from dst_dev and then the master over
1832 * any other enslaved devices in the L3 domain.
1833 */
1834 master = l3mdev_master_dev_rcu(dst_dev);
1835 if (master) {
1836 master_idx = master->ifindex;
1837
1838 hiscore_idx = ipv6_get_saddr_master(net, dst_dev,
1839 master, &dst,
1840 scores, hiscore_idx);
1841
1842 if (scores[hiscore_idx].ifa)
1843 goto out;
1844 }
1845
1846 for_each_netdev_rcu(net, dev) {
1847 /* only consider addresses on devices in the
1848 * same L3 domain
1849 */
1850 if (l3mdev_master_ifindex_rcu(dev) != master_idx)
1851 continue;
1852 idev = __in6_dev_get(dev);
1853 if (!idev)
1854 continue;
1855 hiscore_idx = __ipv6_dev_get_saddr(net, &dst, idev, scores, hiscore_idx);
1856 }
1857 }
1858
1859 out:
1860 hiscore = &scores[hiscore_idx];
1861 if (!hiscore->ifa)
1862 ret = -EADDRNOTAVAIL;
1863 else
1864 *saddr = hiscore->ifa->addr;
1865
1866 rcu_read_unlock();
1867 return ret;
1868 }
1869 EXPORT_SYMBOL(ipv6_dev_get_saddr);
1870
__ipv6_get_lladdr(struct inet6_dev * idev,struct in6_addr * addr,u32 banned_flags)1871 static int __ipv6_get_lladdr(struct inet6_dev *idev, struct in6_addr *addr,
1872 u32 banned_flags)
1873 {
1874 struct inet6_ifaddr *ifp;
1875 int err = -EADDRNOTAVAIL;
1876
1877 list_for_each_entry_reverse(ifp, &idev->addr_list, if_list) {
1878 if (ifp->scope > IFA_LINK)
1879 break;
1880 if (ifp->scope == IFA_LINK &&
1881 !(ifp->flags & banned_flags)) {
1882 *addr = ifp->addr;
1883 err = 0;
1884 break;
1885 }
1886 }
1887 return err;
1888 }
1889
ipv6_get_lladdr(struct net_device * dev,struct in6_addr * addr,u32 banned_flags)1890 int ipv6_get_lladdr(struct net_device *dev, struct in6_addr *addr,
1891 u32 banned_flags)
1892 {
1893 struct inet6_dev *idev;
1894 int err = -EADDRNOTAVAIL;
1895
1896 rcu_read_lock();
1897 idev = __in6_dev_get(dev);
1898 if (idev) {
1899 read_lock_bh(&idev->lock);
1900 err = __ipv6_get_lladdr(idev, addr, banned_flags);
1901 read_unlock_bh(&idev->lock);
1902 }
1903 rcu_read_unlock();
1904 return err;
1905 }
1906
ipv6_count_addresses(const struct inet6_dev * idev)1907 static int ipv6_count_addresses(const struct inet6_dev *idev)
1908 {
1909 const struct inet6_ifaddr *ifp;
1910 int cnt = 0;
1911
1912 rcu_read_lock();
1913 list_for_each_entry_rcu(ifp, &idev->addr_list, if_list)
1914 cnt++;
1915 rcu_read_unlock();
1916 return cnt;
1917 }
1918
ipv6_chk_addr(struct net * net,const struct in6_addr * addr,const struct net_device * dev,int strict)1919 int ipv6_chk_addr(struct net *net, const struct in6_addr *addr,
1920 const struct net_device *dev, int strict)
1921 {
1922 return ipv6_chk_addr_and_flags(net, addr, dev, !dev,
1923 strict, IFA_F_TENTATIVE);
1924 }
1925 EXPORT_SYMBOL(ipv6_chk_addr);
1926
1927 /* device argument is used to find the L3 domain of interest. If
1928 * skip_dev_check is set, then the ifp device is not checked against
1929 * the passed in dev argument. So the 2 cases for addresses checks are:
1930 * 1. does the address exist in the L3 domain that dev is part of
1931 * (skip_dev_check = true), or
1932 *
1933 * 2. does the address exist on the specific device
1934 * (skip_dev_check = false)
1935 */
1936 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)1937 __ipv6_chk_addr_and_flags(struct net *net, const struct in6_addr *addr,
1938 const struct net_device *dev, bool skip_dev_check,
1939 int strict, u32 banned_flags)
1940 {
1941 unsigned int hash = inet6_addr_hash(net, addr);
1942 struct net_device *l3mdev, *ndev;
1943 struct inet6_ifaddr *ifp;
1944 u32 ifp_flags;
1945
1946 rcu_read_lock();
1947
1948 l3mdev = l3mdev_master_dev_rcu(dev);
1949 if (skip_dev_check)
1950 dev = NULL;
1951
1952 hlist_for_each_entry_rcu(ifp, &net->ipv6.inet6_addr_lst[hash], addr_lst) {
1953 ndev = ifp->idev->dev;
1954
1955 if (l3mdev_master_dev_rcu(ndev) != l3mdev)
1956 continue;
1957
1958 /* Decouple optimistic from tentative for evaluation here.
1959 * Ban optimistic addresses explicitly, when required.
1960 */
1961 ifp_flags = (ifp->flags&IFA_F_OPTIMISTIC)
1962 ? (ifp->flags&~IFA_F_TENTATIVE)
1963 : ifp->flags;
1964 if (ipv6_addr_equal(&ifp->addr, addr) &&
1965 !(ifp_flags&banned_flags) &&
1966 (!dev || ndev == dev ||
1967 !(ifp->scope&(IFA_LINK|IFA_HOST) || strict))) {
1968 rcu_read_unlock();
1969 return ndev;
1970 }
1971 }
1972
1973 rcu_read_unlock();
1974 return NULL;
1975 }
1976
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)1977 int ipv6_chk_addr_and_flags(struct net *net, const struct in6_addr *addr,
1978 const struct net_device *dev, bool skip_dev_check,
1979 int strict, u32 banned_flags)
1980 {
1981 return __ipv6_chk_addr_and_flags(net, addr, dev, skip_dev_check,
1982 strict, banned_flags) ? 1 : 0;
1983 }
1984 EXPORT_SYMBOL(ipv6_chk_addr_and_flags);
1985
1986
1987 /* Compares an address/prefix_len with addresses on device @dev.
1988 * If one is found it returns true.
1989 */
ipv6_chk_custom_prefix(const struct in6_addr * addr,const unsigned int prefix_len,struct net_device * dev)1990 bool ipv6_chk_custom_prefix(const struct in6_addr *addr,
1991 const unsigned int prefix_len, struct net_device *dev)
1992 {
1993 const struct inet6_ifaddr *ifa;
1994 const struct inet6_dev *idev;
1995 bool ret = false;
1996
1997 rcu_read_lock();
1998 idev = __in6_dev_get(dev);
1999 if (idev) {
2000 list_for_each_entry_rcu(ifa, &idev->addr_list, if_list) {
2001 ret = ipv6_prefix_equal(addr, &ifa->addr, prefix_len);
2002 if (ret)
2003 break;
2004 }
2005 }
2006 rcu_read_unlock();
2007
2008 return ret;
2009 }
2010 EXPORT_SYMBOL(ipv6_chk_custom_prefix);
2011
ipv6_chk_prefix(const struct in6_addr * addr,struct net_device * dev)2012 int ipv6_chk_prefix(const struct in6_addr *addr, struct net_device *dev)
2013 {
2014 const struct inet6_ifaddr *ifa;
2015 const struct inet6_dev *idev;
2016 int onlink;
2017
2018 onlink = 0;
2019 rcu_read_lock();
2020 idev = __in6_dev_get(dev);
2021 if (idev) {
2022 list_for_each_entry_rcu(ifa, &idev->addr_list, if_list) {
2023 onlink = ipv6_prefix_equal(addr, &ifa->addr,
2024 ifa->prefix_len);
2025 if (onlink)
2026 break;
2027 }
2028 }
2029 rcu_read_unlock();
2030 return onlink;
2031 }
2032 EXPORT_SYMBOL(ipv6_chk_prefix);
2033
2034 /**
2035 * ipv6_dev_find - find the first device with a given source address.
2036 * @net: the net namespace
2037 * @addr: the source address
2038 * @dev: used to find the L3 domain of interest
2039 *
2040 * The caller should be protected by RCU, or RTNL.
2041 */
ipv6_dev_find(struct net * net,const struct in6_addr * addr,struct net_device * dev)2042 struct net_device *ipv6_dev_find(struct net *net, const struct in6_addr *addr,
2043 struct net_device *dev)
2044 {
2045 return __ipv6_chk_addr_and_flags(net, addr, dev, !dev, 1,
2046 IFA_F_TENTATIVE);
2047 }
2048 EXPORT_SYMBOL(ipv6_dev_find);
2049
ipv6_get_ifaddr(struct net * net,const struct in6_addr * addr,struct net_device * dev,int strict)2050 struct inet6_ifaddr *ipv6_get_ifaddr(struct net *net, const struct in6_addr *addr,
2051 struct net_device *dev, int strict)
2052 {
2053 unsigned int hash = inet6_addr_hash(net, addr);
2054 struct inet6_ifaddr *ifp, *result = NULL;
2055
2056 rcu_read_lock();
2057 hlist_for_each_entry_rcu(ifp, &net->ipv6.inet6_addr_lst[hash], addr_lst) {
2058 if (ipv6_addr_equal(&ifp->addr, addr)) {
2059 if (!dev || ifp->idev->dev == dev ||
2060 !(ifp->scope&(IFA_LINK|IFA_HOST) || strict)) {
2061 if (in6_ifa_hold_safe(ifp)) {
2062 result = ifp;
2063 break;
2064 }
2065 }
2066 }
2067 }
2068 rcu_read_unlock();
2069
2070 return result;
2071 }
2072
2073 /* Gets referenced address, destroys ifaddr */
2074
addrconf_dad_stop(struct inet6_ifaddr * ifp,int dad_failed)2075 static void addrconf_dad_stop(struct inet6_ifaddr *ifp, int dad_failed)
2076 {
2077 if (dad_failed)
2078 ifp->flags |= IFA_F_DADFAILED;
2079
2080 if (ifp->flags&IFA_F_TEMPORARY) {
2081 struct inet6_ifaddr *ifpub;
2082 spin_lock_bh(&ifp->lock);
2083 ifpub = ifp->ifpub;
2084 if (ifpub) {
2085 in6_ifa_hold(ifpub);
2086 spin_unlock_bh(&ifp->lock);
2087 ipv6_create_tempaddr(ifpub, true);
2088 in6_ifa_put(ifpub);
2089 } else {
2090 spin_unlock_bh(&ifp->lock);
2091 }
2092 ipv6_del_addr(ifp);
2093 } else if (ifp->flags&IFA_F_PERMANENT || !dad_failed) {
2094 spin_lock_bh(&ifp->lock);
2095 addrconf_del_dad_work(ifp);
2096 ifp->flags |= IFA_F_TENTATIVE;
2097 if (dad_failed)
2098 ifp->flags &= ~IFA_F_OPTIMISTIC;
2099 spin_unlock_bh(&ifp->lock);
2100 if (dad_failed)
2101 ipv6_ifa_notify(0, ifp);
2102 in6_ifa_put(ifp);
2103 } else {
2104 ipv6_del_addr(ifp);
2105 }
2106 }
2107
addrconf_dad_end(struct inet6_ifaddr * ifp)2108 static int addrconf_dad_end(struct inet6_ifaddr *ifp)
2109 {
2110 int err = -ENOENT;
2111
2112 spin_lock_bh(&ifp->lock);
2113 if (ifp->state == INET6_IFADDR_STATE_DAD) {
2114 ifp->state = INET6_IFADDR_STATE_POSTDAD;
2115 err = 0;
2116 }
2117 spin_unlock_bh(&ifp->lock);
2118
2119 return err;
2120 }
2121
addrconf_dad_failure(struct sk_buff * skb,struct inet6_ifaddr * ifp)2122 void addrconf_dad_failure(struct sk_buff *skb, struct inet6_ifaddr *ifp)
2123 {
2124 struct inet6_dev *idev = ifp->idev;
2125 struct net *net = dev_net(idev->dev);
2126
2127 if (addrconf_dad_end(ifp)) {
2128 in6_ifa_put(ifp);
2129 return;
2130 }
2131
2132 net_info_ratelimited("%s: IPv6 duplicate address %pI6c used by %pM detected!\n",
2133 ifp->idev->dev->name, &ifp->addr, eth_hdr(skb)->h_source);
2134
2135 spin_lock_bh(&ifp->lock);
2136
2137 if (ifp->flags & IFA_F_STABLE_PRIVACY) {
2138 struct in6_addr new_addr;
2139 struct inet6_ifaddr *ifp2;
2140 int retries = ifp->stable_privacy_retry + 1;
2141 struct ifa6_config cfg = {
2142 .pfx = &new_addr,
2143 .plen = ifp->prefix_len,
2144 .ifa_flags = ifp->flags,
2145 .valid_lft = ifp->valid_lft,
2146 .preferred_lft = ifp->prefered_lft,
2147 .scope = ifp->scope,
2148 };
2149
2150 if (retries > net->ipv6.sysctl.idgen_retries) {
2151 net_info_ratelimited("%s: privacy stable address generation failed because of DAD conflicts!\n",
2152 ifp->idev->dev->name);
2153 goto errdad;
2154 }
2155
2156 new_addr = ifp->addr;
2157 if (ipv6_generate_stable_address(&new_addr, retries,
2158 idev))
2159 goto errdad;
2160
2161 spin_unlock_bh(&ifp->lock);
2162
2163 if (idev->cnf.max_addresses &&
2164 ipv6_count_addresses(idev) >=
2165 idev->cnf.max_addresses)
2166 goto lock_errdad;
2167
2168 net_info_ratelimited("%s: generating new stable privacy address because of DAD conflict\n",
2169 ifp->idev->dev->name);
2170
2171 ifp2 = ipv6_add_addr(idev, &cfg, false, NULL);
2172 if (IS_ERR(ifp2))
2173 goto lock_errdad;
2174
2175 spin_lock_bh(&ifp2->lock);
2176 ifp2->stable_privacy_retry = retries;
2177 ifp2->state = INET6_IFADDR_STATE_PREDAD;
2178 spin_unlock_bh(&ifp2->lock);
2179
2180 addrconf_mod_dad_work(ifp2, net->ipv6.sysctl.idgen_delay);
2181 in6_ifa_put(ifp2);
2182 lock_errdad:
2183 spin_lock_bh(&ifp->lock);
2184 }
2185
2186 errdad:
2187 /* transition from _POSTDAD to _ERRDAD */
2188 ifp->state = INET6_IFADDR_STATE_ERRDAD;
2189 spin_unlock_bh(&ifp->lock);
2190
2191 addrconf_mod_dad_work(ifp, 0);
2192 in6_ifa_put(ifp);
2193 }
2194
2195 /* Join to solicited addr multicast group.
2196 * caller must hold RTNL */
addrconf_join_solict(struct net_device * dev,const struct in6_addr * addr)2197 void addrconf_join_solict(struct net_device *dev, const struct in6_addr *addr)
2198 {
2199 struct in6_addr maddr;
2200
2201 if (dev->flags&(IFF_LOOPBACK|IFF_NOARP))
2202 return;
2203
2204 addrconf_addr_solict_mult(addr, &maddr);
2205 ipv6_dev_mc_inc(dev, &maddr);
2206 }
2207
2208 /* caller must hold RTNL */
addrconf_leave_solict(struct inet6_dev * idev,const struct in6_addr * addr)2209 void addrconf_leave_solict(struct inet6_dev *idev, const struct in6_addr *addr)
2210 {
2211 struct in6_addr maddr;
2212
2213 if (idev->dev->flags&(IFF_LOOPBACK|IFF_NOARP))
2214 return;
2215
2216 addrconf_addr_solict_mult(addr, &maddr);
2217 __ipv6_dev_mc_dec(idev, &maddr);
2218 }
2219
2220 /* caller must hold RTNL */
addrconf_join_anycast(struct inet6_ifaddr * ifp)2221 static void addrconf_join_anycast(struct inet6_ifaddr *ifp)
2222 {
2223 struct in6_addr addr;
2224
2225 if (ifp->prefix_len >= 127) /* RFC 6164 */
2226 return;
2227 ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
2228 if (ipv6_addr_any(&addr))
2229 return;
2230 __ipv6_dev_ac_inc(ifp->idev, &addr);
2231 }
2232
2233 /* caller must hold RTNL */
addrconf_leave_anycast(struct inet6_ifaddr * ifp)2234 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp)
2235 {
2236 struct in6_addr addr;
2237
2238 if (ifp->prefix_len >= 127) /* RFC 6164 */
2239 return;
2240 ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
2241 if (ipv6_addr_any(&addr))
2242 return;
2243 __ipv6_dev_ac_dec(ifp->idev, &addr);
2244 }
2245
addrconf_ifid_6lowpan(u8 * eui,struct net_device * dev)2246 static int addrconf_ifid_6lowpan(u8 *eui, struct net_device *dev)
2247 {
2248 switch (dev->addr_len) {
2249 case ETH_ALEN:
2250 memcpy(eui, dev->dev_addr, 3);
2251 eui[3] = 0xFF;
2252 eui[4] = 0xFE;
2253 memcpy(eui + 5, dev->dev_addr + 3, 3);
2254 break;
2255 case EUI64_ADDR_LEN:
2256 memcpy(eui, dev->dev_addr, EUI64_ADDR_LEN);
2257 eui[0] ^= 2;
2258 break;
2259 default:
2260 return -1;
2261 }
2262
2263 return 0;
2264 }
2265
addrconf_ifid_ieee1394(u8 * eui,struct net_device * dev)2266 static int addrconf_ifid_ieee1394(u8 *eui, struct net_device *dev)
2267 {
2268 const union fwnet_hwaddr *ha;
2269
2270 if (dev->addr_len != FWNET_ALEN)
2271 return -1;
2272
2273 ha = (const union fwnet_hwaddr *)dev->dev_addr;
2274
2275 memcpy(eui, &ha->uc.uniq_id, sizeof(ha->uc.uniq_id));
2276 eui[0] ^= 2;
2277 return 0;
2278 }
2279
addrconf_ifid_arcnet(u8 * eui,struct net_device * dev)2280 static int addrconf_ifid_arcnet(u8 *eui, struct net_device *dev)
2281 {
2282 /* XXX: inherit EUI-64 from other interface -- yoshfuji */
2283 if (dev->addr_len != ARCNET_ALEN)
2284 return -1;
2285 memset(eui, 0, 7);
2286 eui[7] = *(u8 *)dev->dev_addr;
2287 return 0;
2288 }
2289
addrconf_ifid_infiniband(u8 * eui,struct net_device * dev)2290 static int addrconf_ifid_infiniband(u8 *eui, struct net_device *dev)
2291 {
2292 if (dev->addr_len != INFINIBAND_ALEN)
2293 return -1;
2294 memcpy(eui, dev->dev_addr + 12, 8);
2295 eui[0] |= 2;
2296 return 0;
2297 }
2298
__ipv6_isatap_ifid(u8 * eui,__be32 addr)2299 static int __ipv6_isatap_ifid(u8 *eui, __be32 addr)
2300 {
2301 if (addr == 0)
2302 return -1;
2303 eui[0] = (ipv4_is_zeronet(addr) || ipv4_is_private_10(addr) ||
2304 ipv4_is_loopback(addr) || ipv4_is_linklocal_169(addr) ||
2305 ipv4_is_private_172(addr) || ipv4_is_test_192(addr) ||
2306 ipv4_is_anycast_6to4(addr) || ipv4_is_private_192(addr) ||
2307 ipv4_is_test_198(addr) || ipv4_is_multicast(addr) ||
2308 ipv4_is_lbcast(addr)) ? 0x00 : 0x02;
2309 eui[1] = 0;
2310 eui[2] = 0x5E;
2311 eui[3] = 0xFE;
2312 memcpy(eui + 4, &addr, 4);
2313 return 0;
2314 }
2315
addrconf_ifid_sit(u8 * eui,struct net_device * dev)2316 static int addrconf_ifid_sit(u8 *eui, struct net_device *dev)
2317 {
2318 if (dev->priv_flags & IFF_ISATAP)
2319 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
2320 return -1;
2321 }
2322
addrconf_ifid_gre(u8 * eui,struct net_device * dev)2323 static int addrconf_ifid_gre(u8 *eui, struct net_device *dev)
2324 {
2325 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
2326 }
2327
addrconf_ifid_ip6tnl(u8 * eui,struct net_device * dev)2328 static int addrconf_ifid_ip6tnl(u8 *eui, struct net_device *dev)
2329 {
2330 memcpy(eui, dev->perm_addr, 3);
2331 memcpy(eui + 5, dev->perm_addr + 3, 3);
2332 eui[3] = 0xFF;
2333 eui[4] = 0xFE;
2334 eui[0] ^= 2;
2335 return 0;
2336 }
2337
ipv6_generate_eui64(u8 * eui,struct net_device * dev)2338 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev)
2339 {
2340 switch (dev->type) {
2341 case ARPHRD_ETHER:
2342 case ARPHRD_FDDI:
2343 return addrconf_ifid_eui48(eui, dev);
2344 case ARPHRD_ARCNET:
2345 return addrconf_ifid_arcnet(eui, dev);
2346 case ARPHRD_INFINIBAND:
2347 return addrconf_ifid_infiniband(eui, dev);
2348 case ARPHRD_SIT:
2349 return addrconf_ifid_sit(eui, dev);
2350 case ARPHRD_IPGRE:
2351 case ARPHRD_TUNNEL:
2352 return addrconf_ifid_gre(eui, dev);
2353 case ARPHRD_6LOWPAN:
2354 return addrconf_ifid_6lowpan(eui, dev);
2355 case ARPHRD_IEEE1394:
2356 return addrconf_ifid_ieee1394(eui, dev);
2357 case ARPHRD_TUNNEL6:
2358 case ARPHRD_IP6GRE:
2359 case ARPHRD_RAWIP:
2360 return addrconf_ifid_ip6tnl(eui, dev);
2361 }
2362 return -1;
2363 }
2364
ipv6_inherit_eui64(u8 * eui,struct inet6_dev * idev)2365 static int ipv6_inherit_eui64(u8 *eui, struct inet6_dev *idev)
2366 {
2367 int err = -1;
2368 struct inet6_ifaddr *ifp;
2369
2370 read_lock_bh(&idev->lock);
2371 list_for_each_entry_reverse(ifp, &idev->addr_list, if_list) {
2372 if (ifp->scope > IFA_LINK)
2373 break;
2374 if (ifp->scope == IFA_LINK && !(ifp->flags&IFA_F_TENTATIVE)) {
2375 memcpy(eui, ifp->addr.s6_addr+8, 8);
2376 err = 0;
2377 break;
2378 }
2379 }
2380 read_unlock_bh(&idev->lock);
2381 return err;
2382 }
2383
2384 /* Generation of a randomized Interface Identifier
2385 * draft-ietf-6man-rfc4941bis, Section 3.3.1
2386 */
2387
ipv6_gen_rnd_iid(struct in6_addr * addr)2388 static void ipv6_gen_rnd_iid(struct in6_addr *addr)
2389 {
2390 regen:
2391 get_random_bytes(&addr->s6_addr[8], 8);
2392
2393 /* <draft-ietf-6man-rfc4941bis-08.txt>, Section 3.3.1:
2394 * check if generated address is not inappropriate:
2395 *
2396 * - Reserved IPv6 Interface Identifiers
2397 * - XXX: already assigned to an address on the device
2398 */
2399
2400 /* Subnet-router anycast: 0000:0000:0000:0000 */
2401 if (!(addr->s6_addr32[2] | addr->s6_addr32[3]))
2402 goto regen;
2403
2404 /* IANA Ethernet block: 0200:5EFF:FE00:0000-0200:5EFF:FE00:5212
2405 * Proxy Mobile IPv6: 0200:5EFF:FE00:5213
2406 * IANA Ethernet block: 0200:5EFF:FE00:5214-0200:5EFF:FEFF:FFFF
2407 */
2408 if (ntohl(addr->s6_addr32[2]) == 0x02005eff &&
2409 (ntohl(addr->s6_addr32[3]) & 0Xff000000) == 0xfe000000)
2410 goto regen;
2411
2412 /* Reserved subnet anycast addresses */
2413 if (ntohl(addr->s6_addr32[2]) == 0xfdffffff &&
2414 ntohl(addr->s6_addr32[3]) >= 0Xffffff80)
2415 goto regen;
2416 }
2417
2418 /*
2419 * Add prefix route.
2420 */
2421
2422 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)2423 addrconf_prefix_route(struct in6_addr *pfx, int plen, u32 metric,
2424 struct net_device *dev, unsigned long expires,
2425 u32 flags, gfp_t gfp_flags)
2426 {
2427 struct fib6_config cfg = {
2428 .fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_PREFIX,
2429 .fc_metric = metric ? : IP6_RT_PRIO_ADDRCONF,
2430 .fc_ifindex = dev->ifindex,
2431 .fc_expires = expires,
2432 .fc_dst_len = plen,
2433 .fc_flags = RTF_UP | flags,
2434 .fc_nlinfo.nl_net = dev_net(dev),
2435 .fc_protocol = RTPROT_KERNEL,
2436 .fc_type = RTN_UNICAST,
2437 };
2438
2439 cfg.fc_dst = *pfx;
2440
2441 /* Prevent useless cloning on PtP SIT.
2442 This thing is done here expecting that the whole
2443 class of non-broadcast devices need not cloning.
2444 */
2445 #if IS_ENABLED(CONFIG_IPV6_SIT)
2446 if (dev->type == ARPHRD_SIT && (dev->flags & IFF_POINTOPOINT))
2447 cfg.fc_flags |= RTF_NONEXTHOP;
2448 #endif
2449
2450 ip6_route_add(&cfg, gfp_flags, NULL);
2451 }
2452
2453
addrconf_get_prefix_route(const struct in6_addr * pfx,int plen,const struct net_device * dev,u32 flags,u32 noflags,bool no_gw)2454 static struct fib6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
2455 int plen,
2456 const struct net_device *dev,
2457 u32 flags, u32 noflags,
2458 bool no_gw)
2459 {
2460 struct fib6_node *fn;
2461 struct fib6_info *rt = NULL;
2462 struct fib6_table *table;
2463 u32 tb_id = l3mdev_fib_table(dev) ? : RT6_TABLE_PREFIX;
2464
2465 table = fib6_get_table(dev_net(dev), tb_id);
2466 if (!table)
2467 return NULL;
2468
2469 rcu_read_lock();
2470 fn = fib6_locate(&table->tb6_root, pfx, plen, NULL, 0, true);
2471 if (!fn)
2472 goto out;
2473
2474 for_each_fib6_node_rt_rcu(fn) {
2475 /* prefix routes only use builtin fib6_nh */
2476 if (rt->nh)
2477 continue;
2478
2479 if (rt->fib6_nh->fib_nh_dev->ifindex != dev->ifindex)
2480 continue;
2481 if (no_gw && rt->fib6_nh->fib_nh_gw_family)
2482 continue;
2483 if ((rt->fib6_flags & flags) != flags)
2484 continue;
2485 if ((rt->fib6_flags & noflags) != 0)
2486 continue;
2487 if (!fib6_info_hold_safe(rt))
2488 continue;
2489 break;
2490 }
2491 out:
2492 rcu_read_unlock();
2493 return rt;
2494 }
2495
2496
2497 /* Create "default" multicast route to the interface */
2498
addrconf_add_mroute(struct net_device * dev)2499 static void addrconf_add_mroute(struct net_device *dev)
2500 {
2501 struct fib6_config cfg = {
2502 .fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_LOCAL,
2503 .fc_metric = IP6_RT_PRIO_ADDRCONF,
2504 .fc_ifindex = dev->ifindex,
2505 .fc_dst_len = 8,
2506 .fc_flags = RTF_UP,
2507 .fc_type = RTN_MULTICAST,
2508 .fc_nlinfo.nl_net = dev_net(dev),
2509 .fc_protocol = RTPROT_KERNEL,
2510 };
2511
2512 ipv6_addr_set(&cfg.fc_dst, htonl(0xFF000000), 0, 0, 0);
2513
2514 ip6_route_add(&cfg, GFP_KERNEL, NULL);
2515 }
2516
addrconf_add_dev(struct net_device * dev)2517 static struct inet6_dev *addrconf_add_dev(struct net_device *dev)
2518 {
2519 struct inet6_dev *idev;
2520
2521 ASSERT_RTNL();
2522
2523 idev = ipv6_find_idev(dev);
2524 if (IS_ERR(idev))
2525 return idev;
2526
2527 if (idev->cnf.disable_ipv6)
2528 return ERR_PTR(-EACCES);
2529
2530 /* Add default multicast route */
2531 if (!(dev->flags & IFF_LOOPBACK) && !netif_is_l3_master(dev))
2532 addrconf_add_mroute(dev);
2533
2534 return idev;
2535 }
2536
manage_tempaddrs(struct inet6_dev * idev,struct inet6_ifaddr * ifp,__u32 valid_lft,__u32 prefered_lft,bool create,unsigned long now)2537 static void manage_tempaddrs(struct inet6_dev *idev,
2538 struct inet6_ifaddr *ifp,
2539 __u32 valid_lft, __u32 prefered_lft,
2540 bool create, unsigned long now)
2541 {
2542 u32 flags;
2543 struct inet6_ifaddr *ift;
2544
2545 read_lock_bh(&idev->lock);
2546 /* update all temporary addresses in the list */
2547 list_for_each_entry(ift, &idev->tempaddr_list, tmp_list) {
2548 int age, max_valid, max_prefered;
2549
2550 if (ifp != ift->ifpub)
2551 continue;
2552
2553 /* RFC 4941 section 3.3:
2554 * If a received option will extend the lifetime of a public
2555 * address, the lifetimes of temporary addresses should
2556 * be extended, subject to the overall constraint that no
2557 * temporary addresses should ever remain "valid" or "preferred"
2558 * for a time longer than (TEMP_VALID_LIFETIME) or
2559 * (TEMP_PREFERRED_LIFETIME - DESYNC_FACTOR), respectively.
2560 */
2561 age = (now - ift->cstamp) / HZ;
2562 max_valid = idev->cnf.temp_valid_lft - age;
2563 if (max_valid < 0)
2564 max_valid = 0;
2565
2566 max_prefered = idev->cnf.temp_prefered_lft -
2567 idev->desync_factor - age;
2568 if (max_prefered < 0)
2569 max_prefered = 0;
2570
2571 if (valid_lft > max_valid)
2572 valid_lft = max_valid;
2573
2574 if (prefered_lft > max_prefered)
2575 prefered_lft = max_prefered;
2576
2577 spin_lock(&ift->lock);
2578 flags = ift->flags;
2579 ift->valid_lft = valid_lft;
2580 ift->prefered_lft = prefered_lft;
2581 ift->tstamp = now;
2582 if (prefered_lft > 0)
2583 ift->flags &= ~IFA_F_DEPRECATED;
2584
2585 spin_unlock(&ift->lock);
2586 if (!(flags&IFA_F_TENTATIVE))
2587 ipv6_ifa_notify(0, ift);
2588 }
2589
2590 /* Also create a temporary address if it's enabled but no temporary
2591 * address currently exists.
2592 * However, we get called with valid_lft == 0, prefered_lft == 0, create == false
2593 * as part of cleanup (ie. deleting the mngtmpaddr).
2594 * We don't want that to result in creating a new temporary ip address.
2595 */
2596 if (list_empty(&idev->tempaddr_list) && (valid_lft || prefered_lft))
2597 create = true;
2598
2599 if (create && idev->cnf.use_tempaddr > 0) {
2600 /* When a new public address is created as described
2601 * in [ADDRCONF], also create a new temporary address.
2602 */
2603 read_unlock_bh(&idev->lock);
2604 ipv6_create_tempaddr(ifp, false);
2605 } else {
2606 read_unlock_bh(&idev->lock);
2607 }
2608 }
2609
is_addr_mode_generate_stable(struct inet6_dev * idev)2610 static bool is_addr_mode_generate_stable(struct inet6_dev *idev)
2611 {
2612 return idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_STABLE_PRIVACY ||
2613 idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_RANDOM;
2614 }
2615
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)2616 int addrconf_prefix_rcv_add_addr(struct net *net, struct net_device *dev,
2617 const struct prefix_info *pinfo,
2618 struct inet6_dev *in6_dev,
2619 const struct in6_addr *addr, int addr_type,
2620 u32 addr_flags, bool sllao, bool tokenized,
2621 __u32 valid_lft, u32 prefered_lft)
2622 {
2623 struct inet6_ifaddr *ifp = ipv6_get_ifaddr(net, addr, dev, 1);
2624 int create = 0, update_lft = 0;
2625
2626 if (!ifp && valid_lft) {
2627 int max_addresses = in6_dev->cnf.max_addresses;
2628 struct ifa6_config cfg = {
2629 .pfx = addr,
2630 .plen = pinfo->prefix_len,
2631 .ifa_flags = addr_flags,
2632 .valid_lft = valid_lft,
2633 .preferred_lft = prefered_lft,
2634 .scope = addr_type & IPV6_ADDR_SCOPE_MASK,
2635 .ifa_proto = IFAPROT_KERNEL_RA
2636 };
2637
2638 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2639 if ((net->ipv6.devconf_all->optimistic_dad ||
2640 in6_dev->cnf.optimistic_dad) &&
2641 !net->ipv6.devconf_all->forwarding && sllao)
2642 cfg.ifa_flags |= IFA_F_OPTIMISTIC;
2643 #endif
2644
2645 /* Do not allow to create too much of autoconfigured
2646 * addresses; this would be too easy way to crash kernel.
2647 */
2648 if (!max_addresses ||
2649 ipv6_count_addresses(in6_dev) < max_addresses)
2650 ifp = ipv6_add_addr(in6_dev, &cfg, false, NULL);
2651
2652 if (IS_ERR_OR_NULL(ifp))
2653 return -1;
2654
2655 create = 1;
2656 spin_lock_bh(&ifp->lock);
2657 ifp->flags |= IFA_F_MANAGETEMPADDR;
2658 ifp->cstamp = jiffies;
2659 ifp->tokenized = tokenized;
2660 spin_unlock_bh(&ifp->lock);
2661 addrconf_dad_start(ifp);
2662 }
2663
2664 if (ifp) {
2665 u32 flags;
2666 unsigned long now;
2667 u32 stored_lft;
2668
2669 /* update lifetime (RFC2462 5.5.3 e) */
2670 spin_lock_bh(&ifp->lock);
2671 now = jiffies;
2672 if (ifp->valid_lft > (now - ifp->tstamp) / HZ)
2673 stored_lft = ifp->valid_lft - (now - ifp->tstamp) / HZ;
2674 else
2675 stored_lft = 0;
2676 if (!create && stored_lft) {
2677 const u32 minimum_lft = min_t(u32,
2678 stored_lft, MIN_VALID_LIFETIME);
2679 valid_lft = max(valid_lft, minimum_lft);
2680
2681 /* RFC4862 Section 5.5.3e:
2682 * "Note that the preferred lifetime of the
2683 * corresponding address is always reset to
2684 * the Preferred Lifetime in the received
2685 * Prefix Information option, regardless of
2686 * whether the valid lifetime is also reset or
2687 * ignored."
2688 *
2689 * So we should always update prefered_lft here.
2690 */
2691 update_lft = 1;
2692 }
2693
2694 if (update_lft) {
2695 ifp->valid_lft = valid_lft;
2696 ifp->prefered_lft = prefered_lft;
2697 ifp->tstamp = now;
2698 flags = ifp->flags;
2699 ifp->flags &= ~IFA_F_DEPRECATED;
2700 spin_unlock_bh(&ifp->lock);
2701
2702 if (!(flags&IFA_F_TENTATIVE))
2703 ipv6_ifa_notify(0, ifp);
2704 } else
2705 spin_unlock_bh(&ifp->lock);
2706
2707 manage_tempaddrs(in6_dev, ifp, valid_lft, prefered_lft,
2708 create, now);
2709
2710 in6_ifa_put(ifp);
2711 addrconf_verify(net);
2712 }
2713
2714 return 0;
2715 }
2716 EXPORT_SYMBOL_GPL(addrconf_prefix_rcv_add_addr);
2717
addrconf_prefix_rcv(struct net_device * dev,u8 * opt,int len,bool sllao)2718 void addrconf_prefix_rcv(struct net_device *dev, u8 *opt, int len, bool sllao)
2719 {
2720 struct prefix_info *pinfo;
2721 __u32 valid_lft;
2722 __u32 prefered_lft;
2723 int addr_type, err;
2724 u32 addr_flags = 0;
2725 struct inet6_dev *in6_dev;
2726 struct net *net = dev_net(dev);
2727
2728 pinfo = (struct prefix_info *) opt;
2729
2730 if (len < sizeof(struct prefix_info)) {
2731 netdev_dbg(dev, "addrconf: prefix option too short\n");
2732 return;
2733 }
2734
2735 /*
2736 * Validation checks ([ADDRCONF], page 19)
2737 */
2738
2739 addr_type = ipv6_addr_type(&pinfo->prefix);
2740
2741 if (addr_type & (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL))
2742 return;
2743
2744 valid_lft = ntohl(pinfo->valid);
2745 prefered_lft = ntohl(pinfo->prefered);
2746
2747 if (prefered_lft > valid_lft) {
2748 net_warn_ratelimited("addrconf: prefix option has invalid lifetime\n");
2749 return;
2750 }
2751
2752 in6_dev = in6_dev_get(dev);
2753
2754 if (!in6_dev) {
2755 net_dbg_ratelimited("addrconf: device %s not configured\n",
2756 dev->name);
2757 return;
2758 }
2759
2760 if (valid_lft != 0 && valid_lft < in6_dev->cnf.accept_ra_min_lft)
2761 goto put;
2762
2763 /*
2764 * Two things going on here:
2765 * 1) Add routes for on-link prefixes
2766 * 2) Configure prefixes with the auto flag set
2767 */
2768
2769 if (pinfo->onlink) {
2770 struct fib6_info *rt;
2771 unsigned long rt_expires;
2772
2773 /* Avoid arithmetic overflow. Really, we could
2774 * save rt_expires in seconds, likely valid_lft,
2775 * but it would require division in fib gc, that it
2776 * not good.
2777 */
2778 if (HZ > USER_HZ)
2779 rt_expires = addrconf_timeout_fixup(valid_lft, HZ);
2780 else
2781 rt_expires = addrconf_timeout_fixup(valid_lft, USER_HZ);
2782
2783 if (addrconf_finite_timeout(rt_expires))
2784 rt_expires *= HZ;
2785
2786 rt = addrconf_get_prefix_route(&pinfo->prefix,
2787 pinfo->prefix_len,
2788 dev,
2789 RTF_ADDRCONF | RTF_PREFIX_RT,
2790 RTF_DEFAULT, true);
2791
2792 if (rt) {
2793 /* Autoconf prefix route */
2794 if (valid_lft == 0) {
2795 ip6_del_rt(net, rt, false);
2796 rt = NULL;
2797 } else if (addrconf_finite_timeout(rt_expires)) {
2798 /* not infinity */
2799 fib6_set_expires(rt, jiffies + rt_expires);
2800 } else {
2801 fib6_clean_expires(rt);
2802 }
2803 } else if (valid_lft) {
2804 clock_t expires = 0;
2805 int flags = RTF_ADDRCONF | RTF_PREFIX_RT;
2806 if (addrconf_finite_timeout(rt_expires)) {
2807 /* not infinity */
2808 flags |= RTF_EXPIRES;
2809 expires = jiffies_to_clock_t(rt_expires);
2810 }
2811 addrconf_prefix_route(&pinfo->prefix, pinfo->prefix_len,
2812 0, dev, expires, flags,
2813 GFP_ATOMIC);
2814 }
2815 fib6_info_release(rt);
2816 }
2817
2818 /* Try to figure out our local address for this prefix */
2819
2820 if (pinfo->autoconf && in6_dev->cnf.autoconf) {
2821 struct in6_addr addr;
2822 bool tokenized = false, dev_addr_generated = false;
2823
2824 if (pinfo->prefix_len == 64) {
2825 memcpy(&addr, &pinfo->prefix, 8);
2826
2827 if (!ipv6_addr_any(&in6_dev->token)) {
2828 read_lock_bh(&in6_dev->lock);
2829 memcpy(addr.s6_addr + 8,
2830 in6_dev->token.s6_addr + 8, 8);
2831 read_unlock_bh(&in6_dev->lock);
2832 tokenized = true;
2833 } else if (is_addr_mode_generate_stable(in6_dev) &&
2834 !ipv6_generate_stable_address(&addr, 0,
2835 in6_dev)) {
2836 addr_flags |= IFA_F_STABLE_PRIVACY;
2837 goto ok;
2838 } else if (ipv6_generate_eui64(addr.s6_addr + 8, dev) &&
2839 ipv6_inherit_eui64(addr.s6_addr + 8, in6_dev)) {
2840 goto put;
2841 } else {
2842 dev_addr_generated = true;
2843 }
2844 goto ok;
2845 }
2846 net_dbg_ratelimited("IPv6 addrconf: prefix with wrong length %d\n",
2847 pinfo->prefix_len);
2848 goto put;
2849
2850 ok:
2851 err = addrconf_prefix_rcv_add_addr(net, dev, pinfo, in6_dev,
2852 &addr, addr_type,
2853 addr_flags, sllao,
2854 tokenized, valid_lft,
2855 prefered_lft);
2856 if (err)
2857 goto put;
2858
2859 /* Ignore error case here because previous prefix add addr was
2860 * successful which will be notified.
2861 */
2862 ndisc_ops_prefix_rcv_add_addr(net, dev, pinfo, in6_dev, &addr,
2863 addr_type, addr_flags, sllao,
2864 tokenized, valid_lft,
2865 prefered_lft,
2866 dev_addr_generated);
2867 }
2868 inet6_prefix_notify(RTM_NEWPREFIX, in6_dev, pinfo);
2869 put:
2870 in6_dev_put(in6_dev);
2871 }
2872
addrconf_set_sit_dstaddr(struct net * net,struct net_device * dev,struct in6_ifreq * ireq)2873 static int addrconf_set_sit_dstaddr(struct net *net, struct net_device *dev,
2874 struct in6_ifreq *ireq)
2875 {
2876 struct ip_tunnel_parm p = { };
2877 int err;
2878
2879 if (!(ipv6_addr_type(&ireq->ifr6_addr) & IPV6_ADDR_COMPATv4))
2880 return -EADDRNOTAVAIL;
2881
2882 p.iph.daddr = ireq->ifr6_addr.s6_addr32[3];
2883 p.iph.version = 4;
2884 p.iph.ihl = 5;
2885 p.iph.protocol = IPPROTO_IPV6;
2886 p.iph.ttl = 64;
2887
2888 if (!dev->netdev_ops->ndo_tunnel_ctl)
2889 return -EOPNOTSUPP;
2890 err = dev->netdev_ops->ndo_tunnel_ctl(dev, &p, SIOCADDTUNNEL);
2891 if (err)
2892 return err;
2893
2894 dev = __dev_get_by_name(net, p.name);
2895 if (!dev)
2896 return -ENOBUFS;
2897 return dev_open(dev, NULL);
2898 }
2899
2900 /*
2901 * Set destination address.
2902 * Special case for SIT interfaces where we create a new "virtual"
2903 * device.
2904 */
addrconf_set_dstaddr(struct net * net,void __user * arg)2905 int addrconf_set_dstaddr(struct net *net, void __user *arg)
2906 {
2907 struct net_device *dev;
2908 struct in6_ifreq ireq;
2909 int err = -ENODEV;
2910
2911 if (!IS_ENABLED(CONFIG_IPV6_SIT))
2912 return -ENODEV;
2913 if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2914 return -EFAULT;
2915
2916 rtnl_lock();
2917 dev = __dev_get_by_index(net, ireq.ifr6_ifindex);
2918 if (dev && dev->type == ARPHRD_SIT)
2919 err = addrconf_set_sit_dstaddr(net, dev, &ireq);
2920 rtnl_unlock();
2921 return err;
2922 }
2923
ipv6_mc_config(struct sock * sk,bool join,const struct in6_addr * addr,int ifindex)2924 static int ipv6_mc_config(struct sock *sk, bool join,
2925 const struct in6_addr *addr, int ifindex)
2926 {
2927 int ret;
2928
2929 ASSERT_RTNL();
2930
2931 lock_sock(sk);
2932 if (join)
2933 ret = ipv6_sock_mc_join(sk, ifindex, addr);
2934 else
2935 ret = ipv6_sock_mc_drop(sk, ifindex, addr);
2936 release_sock(sk);
2937
2938 return ret;
2939 }
2940
2941 /*
2942 * Manual configuration of address on an interface
2943 */
inet6_addr_add(struct net * net,int ifindex,struct ifa6_config * cfg,struct netlink_ext_ack * extack)2944 static int inet6_addr_add(struct net *net, int ifindex,
2945 struct ifa6_config *cfg,
2946 struct netlink_ext_ack *extack)
2947 {
2948 struct inet6_ifaddr *ifp;
2949 struct inet6_dev *idev;
2950 struct net_device *dev;
2951 unsigned long timeout;
2952 clock_t expires;
2953 u32 flags;
2954
2955 ASSERT_RTNL();
2956
2957 if (cfg->plen > 128) {
2958 NL_SET_ERR_MSG_MOD(extack, "Invalid prefix length");
2959 return -EINVAL;
2960 }
2961
2962 /* check the lifetime */
2963 if (!cfg->valid_lft || cfg->preferred_lft > cfg->valid_lft) {
2964 NL_SET_ERR_MSG_MOD(extack, "address lifetime invalid");
2965 return -EINVAL;
2966 }
2967
2968 if (cfg->ifa_flags & IFA_F_MANAGETEMPADDR && cfg->plen != 64) {
2969 NL_SET_ERR_MSG_MOD(extack, "address with \"mngtmpaddr\" flag must have a prefix length of 64");
2970 return -EINVAL;
2971 }
2972
2973 dev = __dev_get_by_index(net, ifindex);
2974 if (!dev)
2975 return -ENODEV;
2976
2977 idev = addrconf_add_dev(dev);
2978 if (IS_ERR(idev)) {
2979 NL_SET_ERR_MSG_MOD(extack, "IPv6 is disabled on this device");
2980 return PTR_ERR(idev);
2981 }
2982
2983 if (cfg->ifa_flags & IFA_F_MCAUTOJOIN) {
2984 int ret = ipv6_mc_config(net->ipv6.mc_autojoin_sk,
2985 true, cfg->pfx, ifindex);
2986
2987 if (ret < 0) {
2988 NL_SET_ERR_MSG_MOD(extack, "Multicast auto join failed");
2989 return ret;
2990 }
2991 }
2992
2993 cfg->scope = ipv6_addr_scope(cfg->pfx);
2994
2995 timeout = addrconf_timeout_fixup(cfg->valid_lft, HZ);
2996 if (addrconf_finite_timeout(timeout)) {
2997 expires = jiffies_to_clock_t(timeout * HZ);
2998 cfg->valid_lft = timeout;
2999 flags = RTF_EXPIRES;
3000 } else {
3001 expires = 0;
3002 flags = 0;
3003 cfg->ifa_flags |= IFA_F_PERMANENT;
3004 }
3005
3006 timeout = addrconf_timeout_fixup(cfg->preferred_lft, HZ);
3007 if (addrconf_finite_timeout(timeout)) {
3008 if (timeout == 0)
3009 cfg->ifa_flags |= IFA_F_DEPRECATED;
3010 cfg->preferred_lft = timeout;
3011 }
3012
3013 ifp = ipv6_add_addr(idev, cfg, true, extack);
3014 if (!IS_ERR(ifp)) {
3015 if (!(cfg->ifa_flags & IFA_F_NOPREFIXROUTE)) {
3016 addrconf_prefix_route(&ifp->addr, ifp->prefix_len,
3017 ifp->rt_priority, dev, expires,
3018 flags, GFP_KERNEL);
3019 }
3020
3021 /* Send a netlink notification if DAD is enabled and
3022 * optimistic flag is not set
3023 */
3024 if (!(ifp->flags & (IFA_F_OPTIMISTIC | IFA_F_NODAD)))
3025 ipv6_ifa_notify(0, ifp);
3026 /*
3027 * Note that section 3.1 of RFC 4429 indicates
3028 * that the Optimistic flag should not be set for
3029 * manually configured addresses
3030 */
3031 addrconf_dad_start(ifp);
3032 if (cfg->ifa_flags & IFA_F_MANAGETEMPADDR)
3033 manage_tempaddrs(idev, ifp, cfg->valid_lft,
3034 cfg->preferred_lft, true, jiffies);
3035 in6_ifa_put(ifp);
3036 addrconf_verify_rtnl(net);
3037 return 0;
3038 } else if (cfg->ifa_flags & IFA_F_MCAUTOJOIN) {
3039 ipv6_mc_config(net->ipv6.mc_autojoin_sk, false,
3040 cfg->pfx, ifindex);
3041 }
3042
3043 return PTR_ERR(ifp);
3044 }
3045
inet6_addr_del(struct net * net,int ifindex,u32 ifa_flags,const struct in6_addr * pfx,unsigned int plen,struct netlink_ext_ack * extack)3046 static int inet6_addr_del(struct net *net, int ifindex, u32 ifa_flags,
3047 const struct in6_addr *pfx, unsigned int plen,
3048 struct netlink_ext_ack *extack)
3049 {
3050 struct inet6_ifaddr *ifp;
3051 struct inet6_dev *idev;
3052 struct net_device *dev;
3053
3054 if (plen > 128) {
3055 NL_SET_ERR_MSG_MOD(extack, "Invalid prefix length");
3056 return -EINVAL;
3057 }
3058
3059 dev = __dev_get_by_index(net, ifindex);
3060 if (!dev) {
3061 NL_SET_ERR_MSG_MOD(extack, "Unable to find the interface");
3062 return -ENODEV;
3063 }
3064
3065 idev = __in6_dev_get(dev);
3066 if (!idev) {
3067 NL_SET_ERR_MSG_MOD(extack, "IPv6 is disabled on this device");
3068 return -ENXIO;
3069 }
3070
3071 read_lock_bh(&idev->lock);
3072 list_for_each_entry(ifp, &idev->addr_list, if_list) {
3073 if (ifp->prefix_len == plen &&
3074 ipv6_addr_equal(pfx, &ifp->addr)) {
3075 in6_ifa_hold(ifp);
3076 read_unlock_bh(&idev->lock);
3077
3078 if (!(ifp->flags & IFA_F_TEMPORARY) &&
3079 (ifa_flags & IFA_F_MANAGETEMPADDR))
3080 manage_tempaddrs(idev, ifp, 0, 0, false,
3081 jiffies);
3082 ipv6_del_addr(ifp);
3083 addrconf_verify_rtnl(net);
3084 if (ipv6_addr_is_multicast(pfx)) {
3085 ipv6_mc_config(net->ipv6.mc_autojoin_sk,
3086 false, pfx, dev->ifindex);
3087 }
3088 return 0;
3089 }
3090 }
3091 read_unlock_bh(&idev->lock);
3092
3093 NL_SET_ERR_MSG_MOD(extack, "address not found");
3094 return -EADDRNOTAVAIL;
3095 }
3096
3097
addrconf_add_ifaddr(struct net * net,void __user * arg)3098 int addrconf_add_ifaddr(struct net *net, void __user *arg)
3099 {
3100 struct ifa6_config cfg = {
3101 .ifa_flags = IFA_F_PERMANENT,
3102 .preferred_lft = INFINITY_LIFE_TIME,
3103 .valid_lft = INFINITY_LIFE_TIME,
3104 };
3105 struct in6_ifreq ireq;
3106 int err;
3107
3108 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
3109 return -EPERM;
3110
3111 if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
3112 return -EFAULT;
3113
3114 cfg.pfx = &ireq.ifr6_addr;
3115 cfg.plen = ireq.ifr6_prefixlen;
3116
3117 rtnl_lock();
3118 err = inet6_addr_add(net, ireq.ifr6_ifindex, &cfg, NULL);
3119 rtnl_unlock();
3120 return err;
3121 }
3122
addrconf_del_ifaddr(struct net * net,void __user * arg)3123 int addrconf_del_ifaddr(struct net *net, void __user *arg)
3124 {
3125 struct in6_ifreq ireq;
3126 int err;
3127
3128 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
3129 return -EPERM;
3130
3131 if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
3132 return -EFAULT;
3133
3134 rtnl_lock();
3135 err = inet6_addr_del(net, ireq.ifr6_ifindex, 0, &ireq.ifr6_addr,
3136 ireq.ifr6_prefixlen, NULL);
3137 rtnl_unlock();
3138 return err;
3139 }
3140
add_addr(struct inet6_dev * idev,const struct in6_addr * addr,int plen,int scope,u8 proto)3141 static void add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
3142 int plen, int scope, u8 proto)
3143 {
3144 struct inet6_ifaddr *ifp;
3145 struct ifa6_config cfg = {
3146 .pfx = addr,
3147 .plen = plen,
3148 .ifa_flags = IFA_F_PERMANENT,
3149 .valid_lft = INFINITY_LIFE_TIME,
3150 .preferred_lft = INFINITY_LIFE_TIME,
3151 .scope = scope,
3152 .ifa_proto = proto
3153 };
3154
3155 ifp = ipv6_add_addr(idev, &cfg, true, NULL);
3156 if (!IS_ERR(ifp)) {
3157 spin_lock_bh(&ifp->lock);
3158 ifp->flags &= ~IFA_F_TENTATIVE;
3159 spin_unlock_bh(&ifp->lock);
3160 rt_genid_bump_ipv6(dev_net(idev->dev));
3161 ipv6_ifa_notify(RTM_NEWADDR, ifp);
3162 in6_ifa_put(ifp);
3163 }
3164 }
3165
3166 #if IS_ENABLED(CONFIG_IPV6_SIT) || IS_ENABLED(CONFIG_NET_IPGRE) || IS_ENABLED(CONFIG_IPV6_GRE)
add_v4_addrs(struct inet6_dev * idev)3167 static void add_v4_addrs(struct inet6_dev *idev)
3168 {
3169 struct in6_addr addr;
3170 struct net_device *dev;
3171 struct net *net = dev_net(idev->dev);
3172 int scope, plen, offset = 0;
3173 u32 pflags = 0;
3174
3175 ASSERT_RTNL();
3176
3177 memset(&addr, 0, sizeof(struct in6_addr));
3178 /* in case of IP6GRE the dev_addr is an IPv6 and therefore we use only the last 4 bytes */
3179 if (idev->dev->addr_len == sizeof(struct in6_addr))
3180 offset = sizeof(struct in6_addr) - 4;
3181 memcpy(&addr.s6_addr32[3], idev->dev->dev_addr + offset, 4);
3182
3183 if (!(idev->dev->flags & IFF_POINTOPOINT) && idev->dev->type == ARPHRD_SIT) {
3184 scope = IPV6_ADDR_COMPATv4;
3185 plen = 96;
3186 pflags |= RTF_NONEXTHOP;
3187 } else {
3188 if (idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_NONE)
3189 return;
3190
3191 addr.s6_addr32[0] = htonl(0xfe800000);
3192 scope = IFA_LINK;
3193 plen = 64;
3194 }
3195
3196 if (addr.s6_addr32[3]) {
3197 add_addr(idev, &addr, plen, scope, IFAPROT_UNSPEC);
3198 addrconf_prefix_route(&addr, plen, 0, idev->dev, 0, pflags,
3199 GFP_KERNEL);
3200 return;
3201 }
3202
3203 for_each_netdev(net, dev) {
3204 struct in_device *in_dev = __in_dev_get_rtnl(dev);
3205 if (in_dev && (dev->flags & IFF_UP)) {
3206 struct in_ifaddr *ifa;
3207 int flag = scope;
3208
3209 in_dev_for_each_ifa_rtnl(ifa, in_dev) {
3210 addr.s6_addr32[3] = ifa->ifa_local;
3211
3212 if (ifa->ifa_scope == RT_SCOPE_LINK)
3213 continue;
3214 if (ifa->ifa_scope >= RT_SCOPE_HOST) {
3215 if (idev->dev->flags&IFF_POINTOPOINT)
3216 continue;
3217 flag |= IFA_HOST;
3218 }
3219
3220 add_addr(idev, &addr, plen, flag,
3221 IFAPROT_UNSPEC);
3222 addrconf_prefix_route(&addr, plen, 0, idev->dev,
3223 0, pflags, GFP_KERNEL);
3224 }
3225 }
3226 }
3227 }
3228 #endif
3229
init_loopback(struct net_device * dev)3230 static void init_loopback(struct net_device *dev)
3231 {
3232 struct inet6_dev *idev;
3233
3234 /* ::1 */
3235
3236 ASSERT_RTNL();
3237
3238 idev = ipv6_find_idev(dev);
3239 if (IS_ERR(idev)) {
3240 pr_debug("%s: add_dev failed\n", __func__);
3241 return;
3242 }
3243
3244 add_addr(idev, &in6addr_loopback, 128, IFA_HOST, IFAPROT_KERNEL_LO);
3245 }
3246
addrconf_add_linklocal(struct inet6_dev * idev,const struct in6_addr * addr,u32 flags)3247 void addrconf_add_linklocal(struct inet6_dev *idev,
3248 const struct in6_addr *addr, u32 flags)
3249 {
3250 struct ifa6_config cfg = {
3251 .pfx = addr,
3252 .plen = 64,
3253 .ifa_flags = flags | IFA_F_PERMANENT,
3254 .valid_lft = INFINITY_LIFE_TIME,
3255 .preferred_lft = INFINITY_LIFE_TIME,
3256 .scope = IFA_LINK,
3257 .ifa_proto = IFAPROT_KERNEL_LL
3258 };
3259 struct inet6_ifaddr *ifp;
3260
3261 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
3262 if ((dev_net(idev->dev)->ipv6.devconf_all->optimistic_dad ||
3263 idev->cnf.optimistic_dad) &&
3264 !dev_net(idev->dev)->ipv6.devconf_all->forwarding)
3265 cfg.ifa_flags |= IFA_F_OPTIMISTIC;
3266 #endif
3267
3268 ifp = ipv6_add_addr(idev, &cfg, true, NULL);
3269 if (!IS_ERR(ifp)) {
3270 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, 0, idev->dev,
3271 0, 0, GFP_ATOMIC);
3272 addrconf_dad_start(ifp);
3273 in6_ifa_put(ifp);
3274 }
3275 }
3276 EXPORT_SYMBOL_GPL(addrconf_add_linklocal);
3277
ipv6_reserved_interfaceid(struct in6_addr address)3278 static bool ipv6_reserved_interfaceid(struct in6_addr address)
3279 {
3280 if ((address.s6_addr32[2] | address.s6_addr32[3]) == 0)
3281 return true;
3282
3283 if (address.s6_addr32[2] == htonl(0x02005eff) &&
3284 ((address.s6_addr32[3] & htonl(0xfe000000)) == htonl(0xfe000000)))
3285 return true;
3286
3287 if (address.s6_addr32[2] == htonl(0xfdffffff) &&
3288 ((address.s6_addr32[3] & htonl(0xffffff80)) == htonl(0xffffff80)))
3289 return true;
3290
3291 return false;
3292 }
3293
ipv6_generate_stable_address(struct in6_addr * address,u8 dad_count,const struct inet6_dev * idev)3294 static int ipv6_generate_stable_address(struct in6_addr *address,
3295 u8 dad_count,
3296 const struct inet6_dev *idev)
3297 {
3298 static DEFINE_SPINLOCK(lock);
3299 static __u32 digest[SHA1_DIGEST_WORDS];
3300 static __u32 workspace[SHA1_WORKSPACE_WORDS];
3301
3302 static union {
3303 char __data[SHA1_BLOCK_SIZE];
3304 struct {
3305 struct in6_addr secret;
3306 __be32 prefix[2];
3307 unsigned char hwaddr[MAX_ADDR_LEN];
3308 u8 dad_count;
3309 } __packed;
3310 } data;
3311
3312 struct in6_addr secret;
3313 struct in6_addr temp;
3314 struct net *net = dev_net(idev->dev);
3315
3316 BUILD_BUG_ON(sizeof(data.__data) != sizeof(data));
3317
3318 if (idev->cnf.stable_secret.initialized)
3319 secret = idev->cnf.stable_secret.secret;
3320 else if (net->ipv6.devconf_dflt->stable_secret.initialized)
3321 secret = net->ipv6.devconf_dflt->stable_secret.secret;
3322 else
3323 return -1;
3324
3325 retry:
3326 spin_lock_bh(&lock);
3327
3328 sha1_init(digest);
3329 memset(&data, 0, sizeof(data));
3330 memset(workspace, 0, sizeof(workspace));
3331 memcpy(data.hwaddr, idev->dev->perm_addr, idev->dev->addr_len);
3332 data.prefix[0] = address->s6_addr32[0];
3333 data.prefix[1] = address->s6_addr32[1];
3334 data.secret = secret;
3335 data.dad_count = dad_count;
3336
3337 sha1_transform(digest, data.__data, workspace);
3338
3339 temp = *address;
3340 temp.s6_addr32[2] = (__force __be32)digest[0];
3341 temp.s6_addr32[3] = (__force __be32)digest[1];
3342
3343 spin_unlock_bh(&lock);
3344
3345 if (ipv6_reserved_interfaceid(temp)) {
3346 dad_count++;
3347 if (dad_count > dev_net(idev->dev)->ipv6.sysctl.idgen_retries)
3348 return -1;
3349 goto retry;
3350 }
3351
3352 *address = temp;
3353 return 0;
3354 }
3355
ipv6_gen_mode_random_init(struct inet6_dev * idev)3356 static void ipv6_gen_mode_random_init(struct inet6_dev *idev)
3357 {
3358 struct ipv6_stable_secret *s = &idev->cnf.stable_secret;
3359
3360 if (s->initialized)
3361 return;
3362 s = &idev->cnf.stable_secret;
3363 get_random_bytes(&s->secret, sizeof(s->secret));
3364 s->initialized = true;
3365 }
3366
addrconf_addr_gen(struct inet6_dev * idev,bool prefix_route)3367 static void addrconf_addr_gen(struct inet6_dev *idev, bool prefix_route)
3368 {
3369 struct in6_addr addr;
3370
3371 /* no link local addresses on L3 master devices */
3372 if (netif_is_l3_master(idev->dev))
3373 return;
3374
3375 /* no link local addresses on devices flagged as slaves */
3376 if (idev->dev->priv_flags & IFF_NO_ADDRCONF)
3377 return;
3378
3379 ipv6_addr_set(&addr, htonl(0xFE800000), 0, 0, 0);
3380
3381 switch (idev->cnf.addr_gen_mode) {
3382 case IN6_ADDR_GEN_MODE_RANDOM:
3383 ipv6_gen_mode_random_init(idev);
3384 fallthrough;
3385 case IN6_ADDR_GEN_MODE_STABLE_PRIVACY:
3386 if (!ipv6_generate_stable_address(&addr, 0, idev))
3387 addrconf_add_linklocal(idev, &addr,
3388 IFA_F_STABLE_PRIVACY);
3389 else if (prefix_route)
3390 addrconf_prefix_route(&addr, 64, 0, idev->dev,
3391 0, 0, GFP_KERNEL);
3392 break;
3393 case IN6_ADDR_GEN_MODE_EUI64:
3394 /* addrconf_add_linklocal also adds a prefix_route and we
3395 * only need to care about prefix routes if ipv6_generate_eui64
3396 * couldn't generate one.
3397 */
3398 if (ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) == 0)
3399 addrconf_add_linklocal(idev, &addr, 0);
3400 else if (prefix_route)
3401 addrconf_prefix_route(&addr, 64, 0, idev->dev,
3402 0, 0, GFP_KERNEL);
3403 break;
3404 case IN6_ADDR_GEN_MODE_NONE:
3405 default:
3406 /* will not add any link local address */
3407 break;
3408 }
3409 }
3410
addrconf_dev_config(struct net_device * dev)3411 static void addrconf_dev_config(struct net_device *dev)
3412 {
3413 struct inet6_dev *idev;
3414
3415 ASSERT_RTNL();
3416
3417 if ((dev->type != ARPHRD_ETHER) &&
3418 (dev->type != ARPHRD_FDDI) &&
3419 (dev->type != ARPHRD_ARCNET) &&
3420 (dev->type != ARPHRD_INFINIBAND) &&
3421 (dev->type != ARPHRD_IEEE1394) &&
3422 (dev->type != ARPHRD_TUNNEL6) &&
3423 (dev->type != ARPHRD_6LOWPAN) &&
3424 (dev->type != ARPHRD_TUNNEL) &&
3425 (dev->type != ARPHRD_NONE) &&
3426 (dev->type != ARPHRD_RAWIP)) {
3427 /* Alas, we support only Ethernet autoconfiguration. */
3428 idev = __in6_dev_get(dev);
3429 if (!IS_ERR_OR_NULL(idev) && dev->flags & IFF_UP &&
3430 dev->flags & IFF_MULTICAST)
3431 ipv6_mc_up(idev);
3432 return;
3433 }
3434
3435 idev = addrconf_add_dev(dev);
3436 if (IS_ERR(idev))
3437 return;
3438
3439 /* this device type has no EUI support */
3440 if (dev->type == ARPHRD_NONE &&
3441 idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_EUI64)
3442 idev->cnf.addr_gen_mode = IN6_ADDR_GEN_MODE_RANDOM;
3443
3444 addrconf_addr_gen(idev, false);
3445 }
3446
3447 #if IS_ENABLED(CONFIG_IPV6_SIT)
addrconf_sit_config(struct net_device * dev)3448 static void addrconf_sit_config(struct net_device *dev)
3449 {
3450 struct inet6_dev *idev;
3451
3452 ASSERT_RTNL();
3453
3454 /*
3455 * Configure the tunnel with one of our IPv4
3456 * addresses... we should configure all of
3457 * our v4 addrs in the tunnel
3458 */
3459
3460 idev = ipv6_find_idev(dev);
3461 if (IS_ERR(idev)) {
3462 pr_debug("%s: add_dev failed\n", __func__);
3463 return;
3464 }
3465
3466 if (dev->priv_flags & IFF_ISATAP) {
3467 addrconf_addr_gen(idev, false);
3468 return;
3469 }
3470
3471 add_v4_addrs(idev);
3472
3473 if (dev->flags&IFF_POINTOPOINT)
3474 addrconf_add_mroute(dev);
3475 }
3476 #endif
3477
3478 #if IS_ENABLED(CONFIG_NET_IPGRE) || IS_ENABLED(CONFIG_IPV6_GRE)
addrconf_gre_config(struct net_device * dev)3479 static void addrconf_gre_config(struct net_device *dev)
3480 {
3481 struct inet6_dev *idev;
3482
3483 ASSERT_RTNL();
3484
3485 idev = ipv6_find_idev(dev);
3486 if (IS_ERR(idev)) {
3487 pr_debug("%s: add_dev failed\n", __func__);
3488 return;
3489 }
3490
3491 if (dev->type == ARPHRD_ETHER) {
3492 addrconf_addr_gen(idev, true);
3493 return;
3494 }
3495
3496 add_v4_addrs(idev);
3497
3498 if (dev->flags & IFF_POINTOPOINT)
3499 addrconf_add_mroute(dev);
3500 }
3501 #endif
3502
addrconf_init_auto_addrs(struct net_device * dev)3503 static void addrconf_init_auto_addrs(struct net_device *dev)
3504 {
3505 switch (dev->type) {
3506 #if IS_ENABLED(CONFIG_IPV6_SIT)
3507 case ARPHRD_SIT:
3508 addrconf_sit_config(dev);
3509 break;
3510 #endif
3511 #if IS_ENABLED(CONFIG_NET_IPGRE) || IS_ENABLED(CONFIG_IPV6_GRE)
3512 case ARPHRD_IP6GRE:
3513 case ARPHRD_IPGRE:
3514 addrconf_gre_config(dev);
3515 break;
3516 #endif
3517 case ARPHRD_LOOPBACK:
3518 init_loopback(dev);
3519 break;
3520
3521 default:
3522 addrconf_dev_config(dev);
3523 break;
3524 }
3525 }
3526
fixup_permanent_addr(struct net * net,struct inet6_dev * idev,struct inet6_ifaddr * ifp)3527 static int fixup_permanent_addr(struct net *net,
3528 struct inet6_dev *idev,
3529 struct inet6_ifaddr *ifp)
3530 {
3531 /* !fib6_node means the host route was removed from the
3532 * FIB, for example, if 'lo' device is taken down. In that
3533 * case regenerate the host route.
3534 */
3535 if (!ifp->rt || !ifp->rt->fib6_node) {
3536 struct fib6_info *f6i, *prev;
3537
3538 f6i = addrconf_f6i_alloc(net, idev, &ifp->addr, false,
3539 GFP_ATOMIC, NULL);
3540 if (IS_ERR(f6i))
3541 return PTR_ERR(f6i);
3542
3543 /* ifp->rt can be accessed outside of rtnl */
3544 spin_lock(&ifp->lock);
3545 prev = ifp->rt;
3546 ifp->rt = f6i;
3547 spin_unlock(&ifp->lock);
3548
3549 fib6_info_release(prev);
3550 }
3551
3552 if (!(ifp->flags & IFA_F_NOPREFIXROUTE)) {
3553 addrconf_prefix_route(&ifp->addr, ifp->prefix_len,
3554 ifp->rt_priority, idev->dev, 0, 0,
3555 GFP_ATOMIC);
3556 }
3557
3558 if (ifp->state == INET6_IFADDR_STATE_PREDAD)
3559 addrconf_dad_start(ifp);
3560
3561 return 0;
3562 }
3563
addrconf_permanent_addr(struct net * net,struct net_device * dev)3564 static void addrconf_permanent_addr(struct net *net, struct net_device *dev)
3565 {
3566 struct inet6_ifaddr *ifp, *tmp;
3567 struct inet6_dev *idev;
3568
3569 idev = __in6_dev_get(dev);
3570 if (!idev)
3571 return;
3572
3573 write_lock_bh(&idev->lock);
3574
3575 list_for_each_entry_safe(ifp, tmp, &idev->addr_list, if_list) {
3576 if ((ifp->flags & IFA_F_PERMANENT) &&
3577 fixup_permanent_addr(net, idev, ifp) < 0) {
3578 write_unlock_bh(&idev->lock);
3579 in6_ifa_hold(ifp);
3580 ipv6_del_addr(ifp);
3581 write_lock_bh(&idev->lock);
3582
3583 net_info_ratelimited("%s: Failed to add prefix route for address %pI6c; dropping\n",
3584 idev->dev->name, &ifp->addr);
3585 }
3586 }
3587
3588 write_unlock_bh(&idev->lock);
3589 }
3590
addrconf_notify(struct notifier_block * this,unsigned long event,void * ptr)3591 static int addrconf_notify(struct notifier_block *this, unsigned long event,
3592 void *ptr)
3593 {
3594 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
3595 struct netdev_notifier_change_info *change_info;
3596 struct netdev_notifier_changeupper_info *info;
3597 struct inet6_dev *idev = __in6_dev_get(dev);
3598 struct net *net = dev_net(dev);
3599 int run_pending = 0;
3600 int err;
3601
3602 switch (event) {
3603 case NETDEV_REGISTER:
3604 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
3605 idev = ipv6_add_dev(dev);
3606 if (IS_ERR(idev))
3607 return notifier_from_errno(PTR_ERR(idev));
3608 }
3609 break;
3610
3611 case NETDEV_CHANGEMTU:
3612 /* if MTU under IPV6_MIN_MTU stop IPv6 on this interface. */
3613 if (dev->mtu < IPV6_MIN_MTU) {
3614 addrconf_ifdown(dev, dev != net->loopback_dev);
3615 break;
3616 }
3617
3618 if (idev) {
3619 rt6_mtu_change(dev, dev->mtu);
3620 idev->cnf.mtu6 = dev->mtu;
3621 break;
3622 }
3623
3624 /* allocate new idev */
3625 idev = ipv6_add_dev(dev);
3626 if (IS_ERR(idev))
3627 break;
3628
3629 /* device is still not ready */
3630 if (!(idev->if_flags & IF_READY))
3631 break;
3632
3633 run_pending = 1;
3634 fallthrough;
3635 case NETDEV_UP:
3636 case NETDEV_CHANGE:
3637 if (idev && idev->cnf.disable_ipv6)
3638 break;
3639
3640 if (dev->priv_flags & IFF_NO_ADDRCONF) {
3641 if (event == NETDEV_UP && !IS_ERR_OR_NULL(idev) &&
3642 dev->flags & IFF_UP && dev->flags & IFF_MULTICAST)
3643 ipv6_mc_up(idev);
3644 break;
3645 }
3646
3647 if (event == NETDEV_UP) {
3648 /* restore routes for permanent addresses */
3649 addrconf_permanent_addr(net, dev);
3650
3651 if (!addrconf_link_ready(dev)) {
3652 /* device is not ready yet. */
3653 pr_debug("ADDRCONF(NETDEV_UP): %s: link is not ready\n",
3654 dev->name);
3655 break;
3656 }
3657
3658 if (!idev && dev->mtu >= IPV6_MIN_MTU)
3659 idev = ipv6_add_dev(dev);
3660
3661 if (!IS_ERR_OR_NULL(idev)) {
3662 idev->if_flags |= IF_READY;
3663 run_pending = 1;
3664 }
3665 } else if (event == NETDEV_CHANGE) {
3666 if (!addrconf_link_ready(dev)) {
3667 /* device is still not ready. */
3668 rt6_sync_down_dev(dev, event);
3669 break;
3670 }
3671
3672 if (!IS_ERR_OR_NULL(idev)) {
3673 if (idev->if_flags & IF_READY) {
3674 /* device is already configured -
3675 * but resend MLD reports, we might
3676 * have roamed and need to update
3677 * multicast snooping switches
3678 */
3679 ipv6_mc_up(idev);
3680 change_info = ptr;
3681 if (change_info->flags_changed & IFF_NOARP)
3682 addrconf_dad_run(idev, true);
3683 rt6_sync_up(dev, RTNH_F_LINKDOWN);
3684 break;
3685 }
3686 idev->if_flags |= IF_READY;
3687 }
3688
3689 pr_debug("ADDRCONF(NETDEV_CHANGE): %s: link becomes ready\n",
3690 dev->name);
3691
3692 run_pending = 1;
3693 }
3694
3695 addrconf_init_auto_addrs(dev);
3696
3697 if (!IS_ERR_OR_NULL(idev)) {
3698 if (run_pending)
3699 addrconf_dad_run(idev, false);
3700
3701 /* Device has an address by now */
3702 rt6_sync_up(dev, RTNH_F_DEAD);
3703
3704 /*
3705 * If the MTU changed during the interface down,
3706 * when the interface up, the changed MTU must be
3707 * reflected in the idev as well as routers.
3708 */
3709 if (idev->cnf.mtu6 != dev->mtu &&
3710 dev->mtu >= IPV6_MIN_MTU) {
3711 rt6_mtu_change(dev, dev->mtu);
3712 idev->cnf.mtu6 = dev->mtu;
3713 }
3714 idev->tstamp = jiffies;
3715 inet6_ifinfo_notify(RTM_NEWLINK, idev);
3716
3717 /*
3718 * If the changed mtu during down is lower than
3719 * IPV6_MIN_MTU stop IPv6 on this interface.
3720 */
3721 if (dev->mtu < IPV6_MIN_MTU)
3722 addrconf_ifdown(dev, dev != net->loopback_dev);
3723 }
3724 break;
3725
3726 case NETDEV_DOWN:
3727 case NETDEV_UNREGISTER:
3728 /*
3729 * Remove all addresses from this interface.
3730 */
3731 addrconf_ifdown(dev, event != NETDEV_DOWN);
3732 break;
3733
3734 case NETDEV_CHANGENAME:
3735 if (idev) {
3736 snmp6_unregister_dev(idev);
3737 addrconf_sysctl_unregister(idev);
3738 err = addrconf_sysctl_register(idev);
3739 if (err)
3740 return notifier_from_errno(err);
3741 err = snmp6_register_dev(idev);
3742 if (err) {
3743 addrconf_sysctl_unregister(idev);
3744 return notifier_from_errno(err);
3745 }
3746 }
3747 break;
3748
3749 case NETDEV_PRE_TYPE_CHANGE:
3750 case NETDEV_POST_TYPE_CHANGE:
3751 if (idev)
3752 addrconf_type_change(dev, event);
3753 break;
3754
3755 case NETDEV_CHANGEUPPER:
3756 info = ptr;
3757
3758 /* flush all routes if dev is linked to or unlinked from
3759 * an L3 master device (e.g., VRF)
3760 */
3761 if (info->upper_dev && netif_is_l3_master(info->upper_dev))
3762 addrconf_ifdown(dev, false);
3763 }
3764
3765 return NOTIFY_OK;
3766 }
3767
3768 /*
3769 * addrconf module should be notified of a device going up
3770 */
3771 static struct notifier_block ipv6_dev_notf = {
3772 .notifier_call = addrconf_notify,
3773 .priority = ADDRCONF_NOTIFY_PRIORITY,
3774 };
3775
addrconf_type_change(struct net_device * dev,unsigned long event)3776 static void addrconf_type_change(struct net_device *dev, unsigned long event)
3777 {
3778 struct inet6_dev *idev;
3779 ASSERT_RTNL();
3780
3781 idev = __in6_dev_get(dev);
3782
3783 if (event == NETDEV_POST_TYPE_CHANGE)
3784 ipv6_mc_remap(idev);
3785 else if (event == NETDEV_PRE_TYPE_CHANGE)
3786 ipv6_mc_unmap(idev);
3787 }
3788
addr_is_local(const struct in6_addr * addr)3789 static bool addr_is_local(const struct in6_addr *addr)
3790 {
3791 return ipv6_addr_type(addr) &
3792 (IPV6_ADDR_LINKLOCAL | IPV6_ADDR_LOOPBACK);
3793 }
3794
addrconf_ifdown(struct net_device * dev,bool unregister)3795 static int addrconf_ifdown(struct net_device *dev, bool unregister)
3796 {
3797 unsigned long event = unregister ? NETDEV_UNREGISTER : NETDEV_DOWN;
3798 struct net *net = dev_net(dev);
3799 struct inet6_dev *idev;
3800 struct inet6_ifaddr *ifa;
3801 LIST_HEAD(tmp_addr_list);
3802 bool keep_addr = false;
3803 bool was_ready;
3804 int state, i;
3805
3806 ASSERT_RTNL();
3807
3808 rt6_disable_ip(dev, event);
3809
3810 idev = __in6_dev_get(dev);
3811 if (!idev)
3812 return -ENODEV;
3813
3814 /*
3815 * Step 1: remove reference to ipv6 device from parent device.
3816 * Do not dev_put!
3817 */
3818 if (unregister) {
3819 idev->dead = 1;
3820
3821 /* protected by rtnl_lock */
3822 RCU_INIT_POINTER(dev->ip6_ptr, NULL);
3823
3824 /* Step 1.5: remove snmp6 entry */
3825 snmp6_unregister_dev(idev);
3826
3827 }
3828
3829 /* combine the user config with event to determine if permanent
3830 * addresses are to be removed from address hash table
3831 */
3832 if (!unregister && !idev->cnf.disable_ipv6) {
3833 /* aggregate the system setting and interface setting */
3834 int _keep_addr = net->ipv6.devconf_all->keep_addr_on_down;
3835
3836 if (!_keep_addr)
3837 _keep_addr = idev->cnf.keep_addr_on_down;
3838
3839 keep_addr = (_keep_addr > 0);
3840 }
3841
3842 /* Step 2: clear hash table */
3843 for (i = 0; i < IN6_ADDR_HSIZE; i++) {
3844 struct hlist_head *h = &net->ipv6.inet6_addr_lst[i];
3845
3846 spin_lock_bh(&net->ipv6.addrconf_hash_lock);
3847 restart:
3848 hlist_for_each_entry_rcu(ifa, h, addr_lst) {
3849 if (ifa->idev == idev) {
3850 addrconf_del_dad_work(ifa);
3851 /* combined flag + permanent flag decide if
3852 * address is retained on a down event
3853 */
3854 if (!keep_addr ||
3855 !(ifa->flags & IFA_F_PERMANENT) ||
3856 addr_is_local(&ifa->addr)) {
3857 hlist_del_init_rcu(&ifa->addr_lst);
3858 goto restart;
3859 }
3860 }
3861 }
3862 spin_unlock_bh(&net->ipv6.addrconf_hash_lock);
3863 }
3864
3865 write_lock_bh(&idev->lock);
3866
3867 addrconf_del_rs_timer(idev);
3868
3869 /* Step 2: clear flags for stateless addrconf, repeated down
3870 * detection
3871 */
3872 was_ready = idev->if_flags & IF_READY;
3873 if (!unregister)
3874 idev->if_flags &= ~(IF_RS_SENT|IF_RA_RCVD|IF_READY);
3875
3876 /* Step 3: clear tempaddr list */
3877 while (!list_empty(&idev->tempaddr_list)) {
3878 ifa = list_first_entry(&idev->tempaddr_list,
3879 struct inet6_ifaddr, tmp_list);
3880 list_del(&ifa->tmp_list);
3881 write_unlock_bh(&idev->lock);
3882 spin_lock_bh(&ifa->lock);
3883
3884 if (ifa->ifpub) {
3885 in6_ifa_put(ifa->ifpub);
3886 ifa->ifpub = NULL;
3887 }
3888 spin_unlock_bh(&ifa->lock);
3889 in6_ifa_put(ifa);
3890 write_lock_bh(&idev->lock);
3891 }
3892
3893 list_for_each_entry(ifa, &idev->addr_list, if_list)
3894 list_add_tail(&ifa->if_list_aux, &tmp_addr_list);
3895 write_unlock_bh(&idev->lock);
3896
3897 while (!list_empty(&tmp_addr_list)) {
3898 struct fib6_info *rt = NULL;
3899 bool keep;
3900
3901 ifa = list_first_entry(&tmp_addr_list,
3902 struct inet6_ifaddr, if_list_aux);
3903 list_del(&ifa->if_list_aux);
3904
3905 addrconf_del_dad_work(ifa);
3906
3907 keep = keep_addr && (ifa->flags & IFA_F_PERMANENT) &&
3908 !addr_is_local(&ifa->addr);
3909
3910 spin_lock_bh(&ifa->lock);
3911
3912 if (keep) {
3913 /* set state to skip the notifier below */
3914 state = INET6_IFADDR_STATE_DEAD;
3915 ifa->state = INET6_IFADDR_STATE_PREDAD;
3916 if (!(ifa->flags & IFA_F_NODAD))
3917 ifa->flags |= IFA_F_TENTATIVE;
3918
3919 rt = ifa->rt;
3920 ifa->rt = NULL;
3921 } else {
3922 state = ifa->state;
3923 ifa->state = INET6_IFADDR_STATE_DEAD;
3924 }
3925
3926 spin_unlock_bh(&ifa->lock);
3927
3928 if (rt)
3929 ip6_del_rt(net, rt, false);
3930
3931 if (state != INET6_IFADDR_STATE_DEAD) {
3932 __ipv6_ifa_notify(RTM_DELADDR, ifa);
3933 inet6addr_notifier_call_chain(NETDEV_DOWN, ifa);
3934 } else {
3935 if (idev->cnf.forwarding)
3936 addrconf_leave_anycast(ifa);
3937 addrconf_leave_solict(ifa->idev, &ifa->addr);
3938 }
3939
3940 if (!keep) {
3941 write_lock_bh(&idev->lock);
3942 list_del_rcu(&ifa->if_list);
3943 write_unlock_bh(&idev->lock);
3944 in6_ifa_put(ifa);
3945 }
3946 }
3947
3948 /* Step 5: Discard anycast and multicast list */
3949 if (unregister) {
3950 ipv6_ac_destroy_dev(idev);
3951 ipv6_mc_destroy_dev(idev);
3952 } else if (was_ready) {
3953 ipv6_mc_down(idev);
3954 }
3955
3956 idev->tstamp = jiffies;
3957 idev->ra_mtu = 0;
3958
3959 /* Last: Shot the device (if unregistered) */
3960 if (unregister) {
3961 addrconf_sysctl_unregister(idev);
3962 neigh_parms_release(&nd_tbl, idev->nd_parms);
3963 neigh_ifdown(&nd_tbl, dev);
3964 in6_dev_put(idev);
3965 }
3966 return 0;
3967 }
3968
addrconf_rs_timer(struct timer_list * t)3969 static void addrconf_rs_timer(struct timer_list *t)
3970 {
3971 struct inet6_dev *idev = from_timer(idev, t, rs_timer);
3972 struct net_device *dev = idev->dev;
3973 struct in6_addr lladdr;
3974
3975 write_lock(&idev->lock);
3976 if (idev->dead || !(idev->if_flags & IF_READY))
3977 goto out;
3978
3979 if (!ipv6_accept_ra(idev))
3980 goto out;
3981
3982 /* Announcement received after solicitation was sent */
3983 if (idev->if_flags & IF_RA_RCVD)
3984 goto out;
3985
3986 if (idev->rs_probes++ < idev->cnf.rtr_solicits || idev->cnf.rtr_solicits < 0) {
3987 write_unlock(&idev->lock);
3988 if (!ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE))
3989 ndisc_send_rs(dev, &lladdr,
3990 &in6addr_linklocal_allrouters);
3991 else
3992 goto put;
3993
3994 write_lock(&idev->lock);
3995 idev->rs_interval = rfc3315_s14_backoff_update(
3996 idev->rs_interval, idev->cnf.rtr_solicit_max_interval);
3997 /* The wait after the last probe can be shorter */
3998 addrconf_mod_rs_timer(idev, (idev->rs_probes ==
3999 idev->cnf.rtr_solicits) ?
4000 idev->cnf.rtr_solicit_delay :
4001 idev->rs_interval);
4002 } else {
4003 /*
4004 * Note: we do not support deprecated "all on-link"
4005 * assumption any longer.
4006 */
4007 pr_debug("%s: no IPv6 routers present\n", idev->dev->name);
4008 }
4009
4010 out:
4011 write_unlock(&idev->lock);
4012 put:
4013 in6_dev_put(idev);
4014 }
4015
4016 /*
4017 * Duplicate Address Detection
4018 */
addrconf_dad_kick(struct inet6_ifaddr * ifp)4019 static void addrconf_dad_kick(struct inet6_ifaddr *ifp)
4020 {
4021 unsigned long rand_num;
4022 struct inet6_dev *idev = ifp->idev;
4023 u64 nonce;
4024
4025 if (ifp->flags & IFA_F_OPTIMISTIC)
4026 rand_num = 0;
4027 else
4028 rand_num = get_random_u32_below(idev->cnf.rtr_solicit_delay ? : 1);
4029
4030 nonce = 0;
4031 if (idev->cnf.enhanced_dad ||
4032 dev_net(idev->dev)->ipv6.devconf_all->enhanced_dad) {
4033 do
4034 get_random_bytes(&nonce, 6);
4035 while (nonce == 0);
4036 }
4037 ifp->dad_nonce = nonce;
4038 ifp->dad_probes = idev->cnf.dad_transmits;
4039 addrconf_mod_dad_work(ifp, rand_num);
4040 }
4041
addrconf_dad_begin(struct inet6_ifaddr * ifp)4042 static void addrconf_dad_begin(struct inet6_ifaddr *ifp)
4043 {
4044 struct inet6_dev *idev = ifp->idev;
4045 struct net_device *dev = idev->dev;
4046 bool bump_id, notify = false;
4047 struct net *net;
4048
4049 addrconf_join_solict(dev, &ifp->addr);
4050
4051 read_lock_bh(&idev->lock);
4052 spin_lock(&ifp->lock);
4053 if (ifp->state == INET6_IFADDR_STATE_DEAD)
4054 goto out;
4055
4056 net = dev_net(dev);
4057 if (dev->flags&(IFF_NOARP|IFF_LOOPBACK) ||
4058 (net->ipv6.devconf_all->accept_dad < 1 &&
4059 idev->cnf.accept_dad < 1) ||
4060 !(ifp->flags&IFA_F_TENTATIVE) ||
4061 ifp->flags & IFA_F_NODAD) {
4062 bool send_na = false;
4063
4064 if (ifp->flags & IFA_F_TENTATIVE &&
4065 !(ifp->flags & IFA_F_OPTIMISTIC))
4066 send_na = true;
4067 bump_id = ifp->flags & IFA_F_TENTATIVE;
4068 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
4069 spin_unlock(&ifp->lock);
4070 read_unlock_bh(&idev->lock);
4071
4072 addrconf_dad_completed(ifp, bump_id, send_na);
4073 return;
4074 }
4075
4076 if (!(idev->if_flags & IF_READY)) {
4077 spin_unlock(&ifp->lock);
4078 read_unlock_bh(&idev->lock);
4079 /*
4080 * If the device is not ready:
4081 * - keep it tentative if it is a permanent address.
4082 * - otherwise, kill it.
4083 */
4084 in6_ifa_hold(ifp);
4085 addrconf_dad_stop(ifp, 0);
4086 return;
4087 }
4088
4089 /*
4090 * Optimistic nodes can start receiving
4091 * Frames right away
4092 */
4093 if (ifp->flags & IFA_F_OPTIMISTIC) {
4094 ip6_ins_rt(net, ifp->rt);
4095 if (ipv6_use_optimistic_addr(net, idev)) {
4096 /* Because optimistic nodes can use this address,
4097 * notify listeners. If DAD fails, RTM_DELADDR is sent.
4098 */
4099 notify = true;
4100 }
4101 }
4102
4103 addrconf_dad_kick(ifp);
4104 out:
4105 spin_unlock(&ifp->lock);
4106 read_unlock_bh(&idev->lock);
4107 if (notify)
4108 ipv6_ifa_notify(RTM_NEWADDR, ifp);
4109 }
4110
addrconf_dad_start(struct inet6_ifaddr * ifp)4111 static void addrconf_dad_start(struct inet6_ifaddr *ifp)
4112 {
4113 bool begin_dad = false;
4114
4115 spin_lock_bh(&ifp->lock);
4116 if (ifp->state != INET6_IFADDR_STATE_DEAD) {
4117 ifp->state = INET6_IFADDR_STATE_PREDAD;
4118 begin_dad = true;
4119 }
4120 spin_unlock_bh(&ifp->lock);
4121
4122 if (begin_dad)
4123 addrconf_mod_dad_work(ifp, 0);
4124 }
4125
addrconf_dad_work(struct work_struct * w)4126 static void addrconf_dad_work(struct work_struct *w)
4127 {
4128 struct inet6_ifaddr *ifp = container_of(to_delayed_work(w),
4129 struct inet6_ifaddr,
4130 dad_work);
4131 struct inet6_dev *idev = ifp->idev;
4132 bool bump_id, disable_ipv6 = false;
4133 struct in6_addr mcaddr;
4134
4135 enum {
4136 DAD_PROCESS,
4137 DAD_BEGIN,
4138 DAD_ABORT,
4139 } action = DAD_PROCESS;
4140
4141 rtnl_lock();
4142
4143 spin_lock_bh(&ifp->lock);
4144 if (ifp->state == INET6_IFADDR_STATE_PREDAD) {
4145 action = DAD_BEGIN;
4146 ifp->state = INET6_IFADDR_STATE_DAD;
4147 } else if (ifp->state == INET6_IFADDR_STATE_ERRDAD) {
4148 action = DAD_ABORT;
4149 ifp->state = INET6_IFADDR_STATE_POSTDAD;
4150
4151 if ((dev_net(idev->dev)->ipv6.devconf_all->accept_dad > 1 ||
4152 idev->cnf.accept_dad > 1) &&
4153 !idev->cnf.disable_ipv6 &&
4154 !(ifp->flags & IFA_F_STABLE_PRIVACY)) {
4155 struct in6_addr addr;
4156
4157 addr.s6_addr32[0] = htonl(0xfe800000);
4158 addr.s6_addr32[1] = 0;
4159
4160 if (!ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) &&
4161 ipv6_addr_equal(&ifp->addr, &addr)) {
4162 /* DAD failed for link-local based on MAC */
4163 WRITE_ONCE(idev->cnf.disable_ipv6, 1);
4164
4165 pr_info("%s: IPv6 being disabled!\n",
4166 ifp->idev->dev->name);
4167 disable_ipv6 = true;
4168 }
4169 }
4170 }
4171 spin_unlock_bh(&ifp->lock);
4172
4173 if (action == DAD_BEGIN) {
4174 addrconf_dad_begin(ifp);
4175 goto out;
4176 } else if (action == DAD_ABORT) {
4177 in6_ifa_hold(ifp);
4178 addrconf_dad_stop(ifp, 1);
4179 if (disable_ipv6)
4180 addrconf_ifdown(idev->dev, false);
4181 goto out;
4182 }
4183
4184 if (!ifp->dad_probes && addrconf_dad_end(ifp))
4185 goto out;
4186
4187 write_lock_bh(&idev->lock);
4188 if (idev->dead || !(idev->if_flags & IF_READY)) {
4189 write_unlock_bh(&idev->lock);
4190 goto out;
4191 }
4192
4193 spin_lock(&ifp->lock);
4194 if (ifp->state == INET6_IFADDR_STATE_DEAD) {
4195 spin_unlock(&ifp->lock);
4196 write_unlock_bh(&idev->lock);
4197 goto out;
4198 }
4199
4200 if (ifp->dad_probes == 0) {
4201 bool send_na = false;
4202
4203 /*
4204 * DAD was successful
4205 */
4206
4207 if (ifp->flags & IFA_F_TENTATIVE &&
4208 !(ifp->flags & IFA_F_OPTIMISTIC))
4209 send_na = true;
4210 bump_id = ifp->flags & IFA_F_TENTATIVE;
4211 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
4212 spin_unlock(&ifp->lock);
4213 write_unlock_bh(&idev->lock);
4214
4215 addrconf_dad_completed(ifp, bump_id, send_na);
4216
4217 goto out;
4218 }
4219
4220 ifp->dad_probes--;
4221 addrconf_mod_dad_work(ifp,
4222 max(NEIGH_VAR(ifp->idev->nd_parms, RETRANS_TIME),
4223 HZ/100));
4224 spin_unlock(&ifp->lock);
4225 write_unlock_bh(&idev->lock);
4226
4227 /* send a neighbour solicitation for our addr */
4228 addrconf_addr_solict_mult(&ifp->addr, &mcaddr);
4229 ndisc_send_ns(ifp->idev->dev, &ifp->addr, &mcaddr, &in6addr_any,
4230 ifp->dad_nonce);
4231 out:
4232 in6_ifa_put(ifp);
4233 rtnl_unlock();
4234 }
4235
4236 /* ifp->idev must be at least read locked */
ipv6_lonely_lladdr(struct inet6_ifaddr * ifp)4237 static bool ipv6_lonely_lladdr(struct inet6_ifaddr *ifp)
4238 {
4239 struct inet6_ifaddr *ifpiter;
4240 struct inet6_dev *idev = ifp->idev;
4241
4242 list_for_each_entry_reverse(ifpiter, &idev->addr_list, if_list) {
4243 if (ifpiter->scope > IFA_LINK)
4244 break;
4245 if (ifp != ifpiter && ifpiter->scope == IFA_LINK &&
4246 (ifpiter->flags & (IFA_F_PERMANENT|IFA_F_TENTATIVE|
4247 IFA_F_OPTIMISTIC|IFA_F_DADFAILED)) ==
4248 IFA_F_PERMANENT)
4249 return false;
4250 }
4251 return true;
4252 }
4253
addrconf_dad_completed(struct inet6_ifaddr * ifp,bool bump_id,bool send_na)4254 static void addrconf_dad_completed(struct inet6_ifaddr *ifp, bool bump_id,
4255 bool send_na)
4256 {
4257 struct net_device *dev = ifp->idev->dev;
4258 struct in6_addr lladdr;
4259 bool send_rs, send_mld;
4260
4261 addrconf_del_dad_work(ifp);
4262
4263 /*
4264 * Configure the address for reception. Now it is valid.
4265 */
4266
4267 ipv6_ifa_notify(RTM_NEWADDR, ifp);
4268
4269 /* If added prefix is link local and we are prepared to process
4270 router advertisements, start sending router solicitations.
4271 */
4272
4273 read_lock_bh(&ifp->idev->lock);
4274 send_mld = ifp->scope == IFA_LINK && ipv6_lonely_lladdr(ifp);
4275 send_rs = send_mld &&
4276 ipv6_accept_ra(ifp->idev) &&
4277 ifp->idev->cnf.rtr_solicits != 0 &&
4278 (dev->flags & IFF_LOOPBACK) == 0 &&
4279 (dev->type != ARPHRD_TUNNEL) &&
4280 !netif_is_team_port(dev);
4281 read_unlock_bh(&ifp->idev->lock);
4282
4283 /* While dad is in progress mld report's source address is in6_addrany.
4284 * Resend with proper ll now.
4285 */
4286 if (send_mld)
4287 ipv6_mc_dad_complete(ifp->idev);
4288
4289 /* send unsolicited NA if enabled */
4290 if (send_na &&
4291 (ifp->idev->cnf.ndisc_notify ||
4292 dev_net(dev)->ipv6.devconf_all->ndisc_notify)) {
4293 ndisc_send_na(dev, &in6addr_linklocal_allnodes, &ifp->addr,
4294 /*router=*/ !!ifp->idev->cnf.forwarding,
4295 /*solicited=*/ false, /*override=*/ true,
4296 /*inc_opt=*/ true);
4297 }
4298
4299 if (send_rs) {
4300 /*
4301 * If a host as already performed a random delay
4302 * [...] as part of DAD [...] there is no need
4303 * to delay again before sending the first RS
4304 */
4305 if (ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE))
4306 return;
4307 ndisc_send_rs(dev, &lladdr, &in6addr_linklocal_allrouters);
4308
4309 write_lock_bh(&ifp->idev->lock);
4310 spin_lock(&ifp->lock);
4311 ifp->idev->rs_interval = rfc3315_s14_backoff_init(
4312 ifp->idev->cnf.rtr_solicit_interval);
4313 ifp->idev->rs_probes = 1;
4314 ifp->idev->if_flags |= IF_RS_SENT;
4315 addrconf_mod_rs_timer(ifp->idev, ifp->idev->rs_interval);
4316 spin_unlock(&ifp->lock);
4317 write_unlock_bh(&ifp->idev->lock);
4318 }
4319
4320 if (bump_id)
4321 rt_genid_bump_ipv6(dev_net(dev));
4322
4323 /* Make sure that a new temporary address will be created
4324 * before this temporary address becomes deprecated.
4325 */
4326 if (ifp->flags & IFA_F_TEMPORARY)
4327 addrconf_verify_rtnl(dev_net(dev));
4328 }
4329
addrconf_dad_run(struct inet6_dev * idev,bool restart)4330 static void addrconf_dad_run(struct inet6_dev *idev, bool restart)
4331 {
4332 struct inet6_ifaddr *ifp;
4333
4334 read_lock_bh(&idev->lock);
4335 list_for_each_entry(ifp, &idev->addr_list, if_list) {
4336 spin_lock(&ifp->lock);
4337 if ((ifp->flags & IFA_F_TENTATIVE &&
4338 ifp->state == INET6_IFADDR_STATE_DAD) || restart) {
4339 if (restart)
4340 ifp->state = INET6_IFADDR_STATE_PREDAD;
4341 addrconf_dad_kick(ifp);
4342 }
4343 spin_unlock(&ifp->lock);
4344 }
4345 read_unlock_bh(&idev->lock);
4346 }
4347
4348 #ifdef CONFIG_PROC_FS
4349 struct if6_iter_state {
4350 struct seq_net_private p;
4351 int bucket;
4352 int offset;
4353 };
4354
if6_get_first(struct seq_file * seq,loff_t pos)4355 static struct inet6_ifaddr *if6_get_first(struct seq_file *seq, loff_t pos)
4356 {
4357 struct if6_iter_state *state = seq->private;
4358 struct net *net = seq_file_net(seq);
4359 struct inet6_ifaddr *ifa = NULL;
4360 int p = 0;
4361
4362 /* initial bucket if pos is 0 */
4363 if (pos == 0) {
4364 state->bucket = 0;
4365 state->offset = 0;
4366 }
4367
4368 for (; state->bucket < IN6_ADDR_HSIZE; ++state->bucket) {
4369 hlist_for_each_entry_rcu(ifa, &net->ipv6.inet6_addr_lst[state->bucket],
4370 addr_lst) {
4371 /* sync with offset */
4372 if (p < state->offset) {
4373 p++;
4374 continue;
4375 }
4376 return ifa;
4377 }
4378
4379 /* prepare for next bucket */
4380 state->offset = 0;
4381 p = 0;
4382 }
4383 return NULL;
4384 }
4385
if6_get_next(struct seq_file * seq,struct inet6_ifaddr * ifa)4386 static struct inet6_ifaddr *if6_get_next(struct seq_file *seq,
4387 struct inet6_ifaddr *ifa)
4388 {
4389 struct if6_iter_state *state = seq->private;
4390 struct net *net = seq_file_net(seq);
4391
4392 hlist_for_each_entry_continue_rcu(ifa, addr_lst) {
4393 state->offset++;
4394 return ifa;
4395 }
4396
4397 state->offset = 0;
4398 while (++state->bucket < IN6_ADDR_HSIZE) {
4399 hlist_for_each_entry_rcu(ifa,
4400 &net->ipv6.inet6_addr_lst[state->bucket], addr_lst) {
4401 return ifa;
4402 }
4403 }
4404
4405 return NULL;
4406 }
4407
if6_seq_start(struct seq_file * seq,loff_t * pos)4408 static void *if6_seq_start(struct seq_file *seq, loff_t *pos)
4409 __acquires(rcu)
4410 {
4411 rcu_read_lock();
4412 return if6_get_first(seq, *pos);
4413 }
4414
if6_seq_next(struct seq_file * seq,void * v,loff_t * pos)4415 static void *if6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
4416 {
4417 struct inet6_ifaddr *ifa;
4418
4419 ifa = if6_get_next(seq, v);
4420 ++*pos;
4421 return ifa;
4422 }
4423
if6_seq_stop(struct seq_file * seq,void * v)4424 static void if6_seq_stop(struct seq_file *seq, void *v)
4425 __releases(rcu)
4426 {
4427 rcu_read_unlock();
4428 }
4429
if6_seq_show(struct seq_file * seq,void * v)4430 static int if6_seq_show(struct seq_file *seq, void *v)
4431 {
4432 struct inet6_ifaddr *ifp = (struct inet6_ifaddr *)v;
4433 seq_printf(seq, "%pi6 %02x %02x %02x %02x %8s\n",
4434 &ifp->addr,
4435 ifp->idev->dev->ifindex,
4436 ifp->prefix_len,
4437 ifp->scope,
4438 (u8) ifp->flags,
4439 ifp->idev->dev->name);
4440 return 0;
4441 }
4442
4443 static const struct seq_operations if6_seq_ops = {
4444 .start = if6_seq_start,
4445 .next = if6_seq_next,
4446 .show = if6_seq_show,
4447 .stop = if6_seq_stop,
4448 };
4449
if6_proc_net_init(struct net * net)4450 static int __net_init if6_proc_net_init(struct net *net)
4451 {
4452 if (!proc_create_net("if_inet6", 0444, net->proc_net, &if6_seq_ops,
4453 sizeof(struct if6_iter_state)))
4454 return -ENOMEM;
4455 return 0;
4456 }
4457
if6_proc_net_exit(struct net * net)4458 static void __net_exit if6_proc_net_exit(struct net *net)
4459 {
4460 remove_proc_entry("if_inet6", net->proc_net);
4461 }
4462
4463 static struct pernet_operations if6_proc_net_ops = {
4464 .init = if6_proc_net_init,
4465 .exit = if6_proc_net_exit,
4466 };
4467
if6_proc_init(void)4468 int __init if6_proc_init(void)
4469 {
4470 return register_pernet_subsys(&if6_proc_net_ops);
4471 }
4472
if6_proc_exit(void)4473 void if6_proc_exit(void)
4474 {
4475 unregister_pernet_subsys(&if6_proc_net_ops);
4476 }
4477 #endif /* CONFIG_PROC_FS */
4478
4479 #if IS_ENABLED(CONFIG_IPV6_MIP6)
4480 /* Check if address is a home address configured on any interface. */
ipv6_chk_home_addr(struct net * net,const struct in6_addr * addr)4481 int ipv6_chk_home_addr(struct net *net, const struct in6_addr *addr)
4482 {
4483 unsigned int hash = inet6_addr_hash(net, addr);
4484 struct inet6_ifaddr *ifp = NULL;
4485 int ret = 0;
4486
4487 rcu_read_lock();
4488 hlist_for_each_entry_rcu(ifp, &net->ipv6.inet6_addr_lst[hash], addr_lst) {
4489 if (ipv6_addr_equal(&ifp->addr, addr) &&
4490 (ifp->flags & IFA_F_HOMEADDRESS)) {
4491 ret = 1;
4492 break;
4493 }
4494 }
4495 rcu_read_unlock();
4496 return ret;
4497 }
4498 #endif
4499
4500 /* RFC6554 has some algorithm to avoid loops in segment routing by
4501 * checking if the segments contains any of a local interface address.
4502 *
4503 * Quote:
4504 *
4505 * To detect loops in the SRH, a router MUST determine if the SRH
4506 * includes multiple addresses assigned to any interface on that router.
4507 * If such addresses appear more than once and are separated by at least
4508 * one address not assigned to that router.
4509 */
ipv6_chk_rpl_srh_loop(struct net * net,const struct in6_addr * segs,unsigned char nsegs)4510 int ipv6_chk_rpl_srh_loop(struct net *net, const struct in6_addr *segs,
4511 unsigned char nsegs)
4512 {
4513 const struct in6_addr *addr;
4514 int i, ret = 0, found = 0;
4515 struct inet6_ifaddr *ifp;
4516 bool separated = false;
4517 unsigned int hash;
4518 bool hash_found;
4519
4520 rcu_read_lock();
4521 for (i = 0; i < nsegs; i++) {
4522 addr = &segs[i];
4523 hash = inet6_addr_hash(net, addr);
4524
4525 hash_found = false;
4526 hlist_for_each_entry_rcu(ifp, &net->ipv6.inet6_addr_lst[hash], addr_lst) {
4527
4528 if (ipv6_addr_equal(&ifp->addr, addr)) {
4529 hash_found = true;
4530 break;
4531 }
4532 }
4533
4534 if (hash_found) {
4535 if (found > 1 && separated) {
4536 ret = 1;
4537 break;
4538 }
4539
4540 separated = false;
4541 found++;
4542 } else {
4543 separated = true;
4544 }
4545 }
4546 rcu_read_unlock();
4547
4548 return ret;
4549 }
4550
4551 /*
4552 * Periodic address status verification
4553 */
4554
addrconf_verify_rtnl(struct net * net)4555 static void addrconf_verify_rtnl(struct net *net)
4556 {
4557 unsigned long now, next, next_sec, next_sched;
4558 struct inet6_ifaddr *ifp;
4559 int i;
4560
4561 ASSERT_RTNL();
4562
4563 rcu_read_lock_bh();
4564 now = jiffies;
4565 next = round_jiffies_up(now + ADDR_CHECK_FREQUENCY);
4566
4567 cancel_delayed_work(&net->ipv6.addr_chk_work);
4568
4569 for (i = 0; i < IN6_ADDR_HSIZE; i++) {
4570 restart:
4571 hlist_for_each_entry_rcu_bh(ifp, &net->ipv6.inet6_addr_lst[i], addr_lst) {
4572 unsigned long age;
4573
4574 /* When setting preferred_lft to a value not zero or
4575 * infinity, while valid_lft is infinity
4576 * IFA_F_PERMANENT has a non-infinity life time.
4577 */
4578 if ((ifp->flags & IFA_F_PERMANENT) &&
4579 (ifp->prefered_lft == INFINITY_LIFE_TIME))
4580 continue;
4581
4582 spin_lock(&ifp->lock);
4583 /* We try to batch several events at once. */
4584 age = (now - ifp->tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
4585
4586 if ((ifp->flags&IFA_F_TEMPORARY) &&
4587 !(ifp->flags&IFA_F_TENTATIVE) &&
4588 ifp->prefered_lft != INFINITY_LIFE_TIME &&
4589 !ifp->regen_count && ifp->ifpub) {
4590 /* This is a non-regenerated temporary addr. */
4591
4592 unsigned long regen_advance = ifp->idev->cnf.regen_max_retry *
4593 ifp->idev->cnf.dad_transmits *
4594 max(NEIGH_VAR(ifp->idev->nd_parms, RETRANS_TIME), HZ/100) / HZ;
4595
4596 if (age + regen_advance >= ifp->prefered_lft) {
4597 struct inet6_ifaddr *ifpub = ifp->ifpub;
4598 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
4599 next = ifp->tstamp + ifp->prefered_lft * HZ;
4600
4601 ifp->regen_count++;
4602 in6_ifa_hold(ifp);
4603 in6_ifa_hold(ifpub);
4604 spin_unlock(&ifp->lock);
4605
4606 spin_lock(&ifpub->lock);
4607 ifpub->regen_count = 0;
4608 spin_unlock(&ifpub->lock);
4609 rcu_read_unlock_bh();
4610 ipv6_create_tempaddr(ifpub, true);
4611 in6_ifa_put(ifpub);
4612 in6_ifa_put(ifp);
4613 rcu_read_lock_bh();
4614 goto restart;
4615 } else if (time_before(ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ, next))
4616 next = ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ;
4617 }
4618
4619 if (ifp->valid_lft != INFINITY_LIFE_TIME &&
4620 age >= ifp->valid_lft) {
4621 spin_unlock(&ifp->lock);
4622 in6_ifa_hold(ifp);
4623 rcu_read_unlock_bh();
4624 ipv6_del_addr(ifp);
4625 rcu_read_lock_bh();
4626 goto restart;
4627 } else if (ifp->prefered_lft == INFINITY_LIFE_TIME) {
4628 spin_unlock(&ifp->lock);
4629 continue;
4630 } else if (age >= ifp->prefered_lft) {
4631 /* jiffies - ifp->tstamp > age >= ifp->prefered_lft */
4632 int deprecate = 0;
4633
4634 if (!(ifp->flags&IFA_F_DEPRECATED)) {
4635 deprecate = 1;
4636 ifp->flags |= IFA_F_DEPRECATED;
4637 }
4638
4639 if ((ifp->valid_lft != INFINITY_LIFE_TIME) &&
4640 (time_before(ifp->tstamp + ifp->valid_lft * HZ, next)))
4641 next = ifp->tstamp + ifp->valid_lft * HZ;
4642
4643 spin_unlock(&ifp->lock);
4644
4645 if (deprecate) {
4646 in6_ifa_hold(ifp);
4647
4648 ipv6_ifa_notify(0, ifp);
4649 in6_ifa_put(ifp);
4650 goto restart;
4651 }
4652 } else {
4653 /* ifp->prefered_lft <= ifp->valid_lft */
4654 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
4655 next = ifp->tstamp + ifp->prefered_lft * HZ;
4656 spin_unlock(&ifp->lock);
4657 }
4658 }
4659 }
4660
4661 next_sec = round_jiffies_up(next);
4662 next_sched = next;
4663
4664 /* If rounded timeout is accurate enough, accept it. */
4665 if (time_before(next_sec, next + ADDRCONF_TIMER_FUZZ))
4666 next_sched = next_sec;
4667
4668 /* And minimum interval is ADDRCONF_TIMER_FUZZ_MAX. */
4669 if (time_before(next_sched, jiffies + ADDRCONF_TIMER_FUZZ_MAX))
4670 next_sched = jiffies + ADDRCONF_TIMER_FUZZ_MAX;
4671
4672 pr_debug("now = %lu, schedule = %lu, rounded schedule = %lu => %lu\n",
4673 now, next, next_sec, next_sched);
4674 mod_delayed_work(addrconf_wq, &net->ipv6.addr_chk_work, next_sched - now);
4675 rcu_read_unlock_bh();
4676 }
4677
addrconf_verify_work(struct work_struct * w)4678 static void addrconf_verify_work(struct work_struct *w)
4679 {
4680 struct net *net = container_of(to_delayed_work(w), struct net,
4681 ipv6.addr_chk_work);
4682
4683 rtnl_lock();
4684 addrconf_verify_rtnl(net);
4685 rtnl_unlock();
4686 }
4687
addrconf_verify(struct net * net)4688 static void addrconf_verify(struct net *net)
4689 {
4690 mod_delayed_work(addrconf_wq, &net->ipv6.addr_chk_work, 0);
4691 }
4692
extract_addr(struct nlattr * addr,struct nlattr * local,struct in6_addr ** peer_pfx)4693 static struct in6_addr *extract_addr(struct nlattr *addr, struct nlattr *local,
4694 struct in6_addr **peer_pfx)
4695 {
4696 struct in6_addr *pfx = NULL;
4697
4698 *peer_pfx = NULL;
4699
4700 if (addr)
4701 pfx = nla_data(addr);
4702
4703 if (local) {
4704 if (pfx && nla_memcmp(local, pfx, sizeof(*pfx)))
4705 *peer_pfx = pfx;
4706 pfx = nla_data(local);
4707 }
4708
4709 return pfx;
4710 }
4711
4712 static const struct nla_policy ifa_ipv6_policy[IFA_MAX+1] = {
4713 [IFA_ADDRESS] = { .len = sizeof(struct in6_addr) },
4714 [IFA_LOCAL] = { .len = sizeof(struct in6_addr) },
4715 [IFA_CACHEINFO] = { .len = sizeof(struct ifa_cacheinfo) },
4716 [IFA_FLAGS] = { .len = sizeof(u32) },
4717 [IFA_RT_PRIORITY] = { .len = sizeof(u32) },
4718 [IFA_TARGET_NETNSID] = { .type = NLA_S32 },
4719 [IFA_PROTO] = { .type = NLA_U8 },
4720 };
4721
4722 static int
inet6_rtm_deladdr(struct sk_buff * skb,struct nlmsghdr * nlh,struct netlink_ext_ack * extack)4723 inet6_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh,
4724 struct netlink_ext_ack *extack)
4725 {
4726 struct net *net = sock_net(skb->sk);
4727 struct ifaddrmsg *ifm;
4728 struct nlattr *tb[IFA_MAX+1];
4729 struct in6_addr *pfx, *peer_pfx;
4730 u32 ifa_flags;
4731 int err;
4732
4733 err = nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFA_MAX,
4734 ifa_ipv6_policy, extack);
4735 if (err < 0)
4736 return err;
4737
4738 ifm = nlmsg_data(nlh);
4739 pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
4740 if (!pfx)
4741 return -EINVAL;
4742
4743 ifa_flags = tb[IFA_FLAGS] ? nla_get_u32(tb[IFA_FLAGS]) : ifm->ifa_flags;
4744
4745 /* We ignore other flags so far. */
4746 ifa_flags &= IFA_F_MANAGETEMPADDR;
4747
4748 return inet6_addr_del(net, ifm->ifa_index, ifa_flags, pfx,
4749 ifm->ifa_prefixlen, extack);
4750 }
4751
modify_prefix_route(struct inet6_ifaddr * ifp,unsigned long expires,u32 flags,bool modify_peer)4752 static int modify_prefix_route(struct inet6_ifaddr *ifp,
4753 unsigned long expires, u32 flags,
4754 bool modify_peer)
4755 {
4756 struct fib6_info *f6i;
4757 u32 prio;
4758
4759 f6i = addrconf_get_prefix_route(modify_peer ? &ifp->peer_addr : &ifp->addr,
4760 ifp->prefix_len,
4761 ifp->idev->dev, 0, RTF_DEFAULT, true);
4762 if (!f6i)
4763 return -ENOENT;
4764
4765 prio = ifp->rt_priority ? : IP6_RT_PRIO_ADDRCONF;
4766 if (f6i->fib6_metric != prio) {
4767 /* delete old one */
4768 ip6_del_rt(dev_net(ifp->idev->dev), f6i, false);
4769
4770 /* add new one */
4771 addrconf_prefix_route(modify_peer ? &ifp->peer_addr : &ifp->addr,
4772 ifp->prefix_len,
4773 ifp->rt_priority, ifp->idev->dev,
4774 expires, flags, GFP_KERNEL);
4775 } else {
4776 if (!expires)
4777 fib6_clean_expires(f6i);
4778 else
4779 fib6_set_expires(f6i, expires);
4780
4781 fib6_info_release(f6i);
4782 }
4783
4784 return 0;
4785 }
4786
inet6_addr_modify(struct net * net,struct inet6_ifaddr * ifp,struct ifa6_config * cfg)4787 static int inet6_addr_modify(struct net *net, struct inet6_ifaddr *ifp,
4788 struct ifa6_config *cfg)
4789 {
4790 u32 flags;
4791 clock_t expires;
4792 unsigned long timeout;
4793 bool was_managetempaddr;
4794 bool had_prefixroute;
4795 bool new_peer = false;
4796
4797 ASSERT_RTNL();
4798
4799 if (!cfg->valid_lft || cfg->preferred_lft > cfg->valid_lft)
4800 return -EINVAL;
4801
4802 if (cfg->ifa_flags & IFA_F_MANAGETEMPADDR &&
4803 (ifp->flags & IFA_F_TEMPORARY || ifp->prefix_len != 64))
4804 return -EINVAL;
4805
4806 if (!(ifp->flags & IFA_F_TENTATIVE) || ifp->flags & IFA_F_DADFAILED)
4807 cfg->ifa_flags &= ~IFA_F_OPTIMISTIC;
4808
4809 timeout = addrconf_timeout_fixup(cfg->valid_lft, HZ);
4810 if (addrconf_finite_timeout(timeout)) {
4811 expires = jiffies_to_clock_t(timeout * HZ);
4812 cfg->valid_lft = timeout;
4813 flags = RTF_EXPIRES;
4814 } else {
4815 expires = 0;
4816 flags = 0;
4817 cfg->ifa_flags |= IFA_F_PERMANENT;
4818 }
4819
4820 timeout = addrconf_timeout_fixup(cfg->preferred_lft, HZ);
4821 if (addrconf_finite_timeout(timeout)) {
4822 if (timeout == 0)
4823 cfg->ifa_flags |= IFA_F_DEPRECATED;
4824 cfg->preferred_lft = timeout;
4825 }
4826
4827 if (cfg->peer_pfx &&
4828 memcmp(&ifp->peer_addr, cfg->peer_pfx, sizeof(struct in6_addr))) {
4829 if (!ipv6_addr_any(&ifp->peer_addr))
4830 cleanup_prefix_route(ifp, expires, true, true);
4831 new_peer = true;
4832 }
4833
4834 spin_lock_bh(&ifp->lock);
4835 was_managetempaddr = ifp->flags & IFA_F_MANAGETEMPADDR;
4836 had_prefixroute = ifp->flags & IFA_F_PERMANENT &&
4837 !(ifp->flags & IFA_F_NOPREFIXROUTE);
4838 ifp->flags &= ~(IFA_F_DEPRECATED | IFA_F_PERMANENT | IFA_F_NODAD |
4839 IFA_F_HOMEADDRESS | IFA_F_MANAGETEMPADDR |
4840 IFA_F_NOPREFIXROUTE);
4841 ifp->flags |= cfg->ifa_flags;
4842 ifp->tstamp = jiffies;
4843 ifp->valid_lft = cfg->valid_lft;
4844 ifp->prefered_lft = cfg->preferred_lft;
4845 ifp->ifa_proto = cfg->ifa_proto;
4846
4847 if (cfg->rt_priority && cfg->rt_priority != ifp->rt_priority)
4848 ifp->rt_priority = cfg->rt_priority;
4849
4850 if (new_peer)
4851 ifp->peer_addr = *cfg->peer_pfx;
4852
4853 spin_unlock_bh(&ifp->lock);
4854 if (!(ifp->flags&IFA_F_TENTATIVE))
4855 ipv6_ifa_notify(0, ifp);
4856
4857 if (!(cfg->ifa_flags & IFA_F_NOPREFIXROUTE)) {
4858 int rc = -ENOENT;
4859
4860 if (had_prefixroute)
4861 rc = modify_prefix_route(ifp, expires, flags, false);
4862
4863 /* prefix route could have been deleted; if so restore it */
4864 if (rc == -ENOENT) {
4865 addrconf_prefix_route(&ifp->addr, ifp->prefix_len,
4866 ifp->rt_priority, ifp->idev->dev,
4867 expires, flags, GFP_KERNEL);
4868 }
4869
4870 if (had_prefixroute && !ipv6_addr_any(&ifp->peer_addr))
4871 rc = modify_prefix_route(ifp, expires, flags, true);
4872
4873 if (rc == -ENOENT && !ipv6_addr_any(&ifp->peer_addr)) {
4874 addrconf_prefix_route(&ifp->peer_addr, ifp->prefix_len,
4875 ifp->rt_priority, ifp->idev->dev,
4876 expires, flags, GFP_KERNEL);
4877 }
4878 } else if (had_prefixroute) {
4879 enum cleanup_prefix_rt_t action;
4880 unsigned long rt_expires;
4881
4882 write_lock_bh(&ifp->idev->lock);
4883 action = check_cleanup_prefix_route(ifp, &rt_expires);
4884 write_unlock_bh(&ifp->idev->lock);
4885
4886 if (action != CLEANUP_PREFIX_RT_NOP) {
4887 cleanup_prefix_route(ifp, rt_expires,
4888 action == CLEANUP_PREFIX_RT_DEL, false);
4889 }
4890 }
4891
4892 if (was_managetempaddr || ifp->flags & IFA_F_MANAGETEMPADDR) {
4893 if (was_managetempaddr &&
4894 !(ifp->flags & IFA_F_MANAGETEMPADDR)) {
4895 cfg->valid_lft = 0;
4896 cfg->preferred_lft = 0;
4897 }
4898 manage_tempaddrs(ifp->idev, ifp, cfg->valid_lft,
4899 cfg->preferred_lft, !was_managetempaddr,
4900 jiffies);
4901 }
4902
4903 addrconf_verify_rtnl(net);
4904
4905 return 0;
4906 }
4907
4908 static int
inet6_rtm_newaddr(struct sk_buff * skb,struct nlmsghdr * nlh,struct netlink_ext_ack * extack)4909 inet6_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh,
4910 struct netlink_ext_ack *extack)
4911 {
4912 struct net *net = sock_net(skb->sk);
4913 struct ifaddrmsg *ifm;
4914 struct nlattr *tb[IFA_MAX+1];
4915 struct in6_addr *peer_pfx;
4916 struct inet6_ifaddr *ifa;
4917 struct net_device *dev;
4918 struct inet6_dev *idev;
4919 struct ifa6_config cfg;
4920 int err;
4921
4922 err = nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFA_MAX,
4923 ifa_ipv6_policy, extack);
4924 if (err < 0)
4925 return err;
4926
4927 memset(&cfg, 0, sizeof(cfg));
4928
4929 ifm = nlmsg_data(nlh);
4930 cfg.pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
4931 if (!cfg.pfx)
4932 return -EINVAL;
4933
4934 cfg.peer_pfx = peer_pfx;
4935 cfg.plen = ifm->ifa_prefixlen;
4936 if (tb[IFA_RT_PRIORITY])
4937 cfg.rt_priority = nla_get_u32(tb[IFA_RT_PRIORITY]);
4938
4939 if (tb[IFA_PROTO])
4940 cfg.ifa_proto = nla_get_u8(tb[IFA_PROTO]);
4941
4942 cfg.valid_lft = INFINITY_LIFE_TIME;
4943 cfg.preferred_lft = INFINITY_LIFE_TIME;
4944
4945 if (tb[IFA_CACHEINFO]) {
4946 struct ifa_cacheinfo *ci;
4947
4948 ci = nla_data(tb[IFA_CACHEINFO]);
4949 cfg.valid_lft = ci->ifa_valid;
4950 cfg.preferred_lft = ci->ifa_prefered;
4951 }
4952
4953 dev = __dev_get_by_index(net, ifm->ifa_index);
4954 if (!dev) {
4955 NL_SET_ERR_MSG_MOD(extack, "Unable to find the interface");
4956 return -ENODEV;
4957 }
4958
4959 if (tb[IFA_FLAGS])
4960 cfg.ifa_flags = nla_get_u32(tb[IFA_FLAGS]);
4961 else
4962 cfg.ifa_flags = ifm->ifa_flags;
4963
4964 /* We ignore other flags so far. */
4965 cfg.ifa_flags &= IFA_F_NODAD | IFA_F_HOMEADDRESS |
4966 IFA_F_MANAGETEMPADDR | IFA_F_NOPREFIXROUTE |
4967 IFA_F_MCAUTOJOIN | IFA_F_OPTIMISTIC;
4968
4969 idev = ipv6_find_idev(dev);
4970 if (IS_ERR(idev))
4971 return PTR_ERR(idev);
4972
4973 if (!ipv6_allow_optimistic_dad(net, idev))
4974 cfg.ifa_flags &= ~IFA_F_OPTIMISTIC;
4975
4976 if (cfg.ifa_flags & IFA_F_NODAD &&
4977 cfg.ifa_flags & IFA_F_OPTIMISTIC) {
4978 NL_SET_ERR_MSG(extack, "IFA_F_NODAD and IFA_F_OPTIMISTIC are mutually exclusive");
4979 return -EINVAL;
4980 }
4981
4982 ifa = ipv6_get_ifaddr(net, cfg.pfx, dev, 1);
4983 if (!ifa) {
4984 /*
4985 * It would be best to check for !NLM_F_CREATE here but
4986 * userspace already relies on not having to provide this.
4987 */
4988 return inet6_addr_add(net, ifm->ifa_index, &cfg, extack);
4989 }
4990
4991 if (nlh->nlmsg_flags & NLM_F_EXCL ||
4992 !(nlh->nlmsg_flags & NLM_F_REPLACE)) {
4993 NL_SET_ERR_MSG_MOD(extack, "address already assigned");
4994 err = -EEXIST;
4995 } else {
4996 err = inet6_addr_modify(net, ifa, &cfg);
4997 }
4998
4999 in6_ifa_put(ifa);
5000
5001 return err;
5002 }
5003
put_ifaddrmsg(struct nlmsghdr * nlh,u8 prefixlen,u32 flags,u8 scope,int ifindex)5004 static void put_ifaddrmsg(struct nlmsghdr *nlh, u8 prefixlen, u32 flags,
5005 u8 scope, int ifindex)
5006 {
5007 struct ifaddrmsg *ifm;
5008
5009 ifm = nlmsg_data(nlh);
5010 ifm->ifa_family = AF_INET6;
5011 ifm->ifa_prefixlen = prefixlen;
5012 ifm->ifa_flags = flags;
5013 ifm->ifa_scope = scope;
5014 ifm->ifa_index = ifindex;
5015 }
5016
put_cacheinfo(struct sk_buff * skb,unsigned long cstamp,unsigned long tstamp,u32 preferred,u32 valid)5017 static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
5018 unsigned long tstamp, u32 preferred, u32 valid)
5019 {
5020 struct ifa_cacheinfo ci;
5021
5022 ci.cstamp = cstamp_delta(cstamp);
5023 ci.tstamp = cstamp_delta(tstamp);
5024 ci.ifa_prefered = preferred;
5025 ci.ifa_valid = valid;
5026
5027 return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
5028 }
5029
rt_scope(int ifa_scope)5030 static inline int rt_scope(int ifa_scope)
5031 {
5032 if (ifa_scope & IFA_HOST)
5033 return RT_SCOPE_HOST;
5034 else if (ifa_scope & IFA_LINK)
5035 return RT_SCOPE_LINK;
5036 else if (ifa_scope & IFA_SITE)
5037 return RT_SCOPE_SITE;
5038 else
5039 return RT_SCOPE_UNIVERSE;
5040 }
5041
inet6_ifaddr_msgsize(void)5042 static inline int inet6_ifaddr_msgsize(void)
5043 {
5044 return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
5045 + nla_total_size(16) /* IFA_LOCAL */
5046 + nla_total_size(16) /* IFA_ADDRESS */
5047 + nla_total_size(sizeof(struct ifa_cacheinfo))
5048 + nla_total_size(4) /* IFA_FLAGS */
5049 + nla_total_size(1) /* IFA_PROTO */
5050 + nla_total_size(4) /* IFA_RT_PRIORITY */;
5051 }
5052
5053 enum addr_type_t {
5054 UNICAST_ADDR,
5055 MULTICAST_ADDR,
5056 ANYCAST_ADDR,
5057 };
5058
5059 struct inet6_fill_args {
5060 u32 portid;
5061 u32 seq;
5062 int event;
5063 unsigned int flags;
5064 int netnsid;
5065 int ifindex;
5066 enum addr_type_t type;
5067 };
5068
inet6_fill_ifaddr(struct sk_buff * skb,struct inet6_ifaddr * ifa,struct inet6_fill_args * args)5069 static int inet6_fill_ifaddr(struct sk_buff *skb, struct inet6_ifaddr *ifa,
5070 struct inet6_fill_args *args)
5071 {
5072 struct nlmsghdr *nlh;
5073 u32 preferred, valid;
5074
5075 nlh = nlmsg_put(skb, args->portid, args->seq, args->event,
5076 sizeof(struct ifaddrmsg), args->flags);
5077 if (!nlh)
5078 return -EMSGSIZE;
5079
5080 put_ifaddrmsg(nlh, ifa->prefix_len, ifa->flags, rt_scope(ifa->scope),
5081 ifa->idev->dev->ifindex);
5082
5083 if (args->netnsid >= 0 &&
5084 nla_put_s32(skb, IFA_TARGET_NETNSID, args->netnsid))
5085 goto error;
5086
5087 spin_lock_bh(&ifa->lock);
5088 if (!((ifa->flags&IFA_F_PERMANENT) &&
5089 (ifa->prefered_lft == INFINITY_LIFE_TIME))) {
5090 preferred = ifa->prefered_lft;
5091 valid = ifa->valid_lft;
5092 if (preferred != INFINITY_LIFE_TIME) {
5093 long tval = (jiffies - ifa->tstamp)/HZ;
5094 if (preferred > tval)
5095 preferred -= tval;
5096 else
5097 preferred = 0;
5098 if (valid != INFINITY_LIFE_TIME) {
5099 if (valid > tval)
5100 valid -= tval;
5101 else
5102 valid = 0;
5103 }
5104 }
5105 } else {
5106 preferred = INFINITY_LIFE_TIME;
5107 valid = INFINITY_LIFE_TIME;
5108 }
5109 spin_unlock_bh(&ifa->lock);
5110
5111 if (!ipv6_addr_any(&ifa->peer_addr)) {
5112 if (nla_put_in6_addr(skb, IFA_LOCAL, &ifa->addr) < 0 ||
5113 nla_put_in6_addr(skb, IFA_ADDRESS, &ifa->peer_addr) < 0)
5114 goto error;
5115 } else
5116 if (nla_put_in6_addr(skb, IFA_ADDRESS, &ifa->addr) < 0)
5117 goto error;
5118
5119 if (ifa->rt_priority &&
5120 nla_put_u32(skb, IFA_RT_PRIORITY, ifa->rt_priority))
5121 goto error;
5122
5123 if (put_cacheinfo(skb, ifa->cstamp, ifa->tstamp, preferred, valid) < 0)
5124 goto error;
5125
5126 if (nla_put_u32(skb, IFA_FLAGS, ifa->flags) < 0)
5127 goto error;
5128
5129 if (ifa->ifa_proto &&
5130 nla_put_u8(skb, IFA_PROTO, ifa->ifa_proto))
5131 goto error;
5132
5133 nlmsg_end(skb, nlh);
5134 return 0;
5135
5136 error:
5137 nlmsg_cancel(skb, nlh);
5138 return -EMSGSIZE;
5139 }
5140
inet6_fill_ifmcaddr(struct sk_buff * skb,struct ifmcaddr6 * ifmca,struct inet6_fill_args * args)5141 static int inet6_fill_ifmcaddr(struct sk_buff *skb, struct ifmcaddr6 *ifmca,
5142 struct inet6_fill_args *args)
5143 {
5144 struct nlmsghdr *nlh;
5145 u8 scope = RT_SCOPE_UNIVERSE;
5146 int ifindex = ifmca->idev->dev->ifindex;
5147
5148 if (ipv6_addr_scope(&ifmca->mca_addr) & IFA_SITE)
5149 scope = RT_SCOPE_SITE;
5150
5151 nlh = nlmsg_put(skb, args->portid, args->seq, args->event,
5152 sizeof(struct ifaddrmsg), args->flags);
5153 if (!nlh)
5154 return -EMSGSIZE;
5155
5156 if (args->netnsid >= 0 &&
5157 nla_put_s32(skb, IFA_TARGET_NETNSID, args->netnsid)) {
5158 nlmsg_cancel(skb, nlh);
5159 return -EMSGSIZE;
5160 }
5161
5162 put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
5163 if (nla_put_in6_addr(skb, IFA_MULTICAST, &ifmca->mca_addr) < 0 ||
5164 put_cacheinfo(skb, ifmca->mca_cstamp, ifmca->mca_tstamp,
5165 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
5166 nlmsg_cancel(skb, nlh);
5167 return -EMSGSIZE;
5168 }
5169
5170 nlmsg_end(skb, nlh);
5171 return 0;
5172 }
5173
inet6_fill_ifacaddr(struct sk_buff * skb,struct ifacaddr6 * ifaca,struct inet6_fill_args * args)5174 static int inet6_fill_ifacaddr(struct sk_buff *skb, struct ifacaddr6 *ifaca,
5175 struct inet6_fill_args *args)
5176 {
5177 struct net_device *dev = fib6_info_nh_dev(ifaca->aca_rt);
5178 int ifindex = dev ? dev->ifindex : 1;
5179 struct nlmsghdr *nlh;
5180 u8 scope = RT_SCOPE_UNIVERSE;
5181
5182 if (ipv6_addr_scope(&ifaca->aca_addr) & IFA_SITE)
5183 scope = RT_SCOPE_SITE;
5184
5185 nlh = nlmsg_put(skb, args->portid, args->seq, args->event,
5186 sizeof(struct ifaddrmsg), args->flags);
5187 if (!nlh)
5188 return -EMSGSIZE;
5189
5190 if (args->netnsid >= 0 &&
5191 nla_put_s32(skb, IFA_TARGET_NETNSID, args->netnsid)) {
5192 nlmsg_cancel(skb, nlh);
5193 return -EMSGSIZE;
5194 }
5195
5196 put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
5197 if (nla_put_in6_addr(skb, IFA_ANYCAST, &ifaca->aca_addr) < 0 ||
5198 put_cacheinfo(skb, ifaca->aca_cstamp, ifaca->aca_tstamp,
5199 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
5200 nlmsg_cancel(skb, nlh);
5201 return -EMSGSIZE;
5202 }
5203
5204 nlmsg_end(skb, nlh);
5205 return 0;
5206 }
5207
5208 /* called with rcu_read_lock() */
in6_dump_addrs(struct inet6_dev * idev,struct sk_buff * skb,struct netlink_callback * cb,int s_ip_idx,struct inet6_fill_args * fillargs)5209 static int in6_dump_addrs(struct inet6_dev *idev, struct sk_buff *skb,
5210 struct netlink_callback *cb, int s_ip_idx,
5211 struct inet6_fill_args *fillargs)
5212 {
5213 struct ifmcaddr6 *ifmca;
5214 struct ifacaddr6 *ifaca;
5215 int ip_idx = 0;
5216 int err = 1;
5217
5218 read_lock_bh(&idev->lock);
5219 switch (fillargs->type) {
5220 case UNICAST_ADDR: {
5221 struct inet6_ifaddr *ifa;
5222 fillargs->event = RTM_NEWADDR;
5223
5224 /* unicast address incl. temp addr */
5225 list_for_each_entry(ifa, &idev->addr_list, if_list) {
5226 if (ip_idx < s_ip_idx)
5227 goto next;
5228 err = inet6_fill_ifaddr(skb, ifa, fillargs);
5229 if (err < 0)
5230 break;
5231 nl_dump_check_consistent(cb, nlmsg_hdr(skb));
5232 next:
5233 ip_idx++;
5234 }
5235 break;
5236 }
5237 case MULTICAST_ADDR:
5238 read_unlock_bh(&idev->lock);
5239 fillargs->event = RTM_GETMULTICAST;
5240
5241 /* multicast address */
5242 for (ifmca = rtnl_dereference(idev->mc_list);
5243 ifmca;
5244 ifmca = rtnl_dereference(ifmca->next), ip_idx++) {
5245 if (ip_idx < s_ip_idx)
5246 continue;
5247 err = inet6_fill_ifmcaddr(skb, ifmca, fillargs);
5248 if (err < 0)
5249 break;
5250 }
5251 read_lock_bh(&idev->lock);
5252 break;
5253 case ANYCAST_ADDR:
5254 fillargs->event = RTM_GETANYCAST;
5255 /* anycast address */
5256 for (ifaca = idev->ac_list; ifaca;
5257 ifaca = ifaca->aca_next, ip_idx++) {
5258 if (ip_idx < s_ip_idx)
5259 continue;
5260 err = inet6_fill_ifacaddr(skb, ifaca, fillargs);
5261 if (err < 0)
5262 break;
5263 }
5264 break;
5265 default:
5266 break;
5267 }
5268 read_unlock_bh(&idev->lock);
5269 cb->args[2] = ip_idx;
5270 return err;
5271 }
5272
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)5273 static int inet6_valid_dump_ifaddr_req(const struct nlmsghdr *nlh,
5274 struct inet6_fill_args *fillargs,
5275 struct net **tgt_net, struct sock *sk,
5276 struct netlink_callback *cb)
5277 {
5278 struct netlink_ext_ack *extack = cb->extack;
5279 struct nlattr *tb[IFA_MAX+1];
5280 struct ifaddrmsg *ifm;
5281 int err, i;
5282
5283 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) {
5284 NL_SET_ERR_MSG_MOD(extack, "Invalid header for address dump request");
5285 return -EINVAL;
5286 }
5287
5288 ifm = nlmsg_data(nlh);
5289 if (ifm->ifa_prefixlen || ifm->ifa_flags || ifm->ifa_scope) {
5290 NL_SET_ERR_MSG_MOD(extack, "Invalid values in header for address dump request");
5291 return -EINVAL;
5292 }
5293
5294 fillargs->ifindex = ifm->ifa_index;
5295 if (fillargs->ifindex) {
5296 cb->answer_flags |= NLM_F_DUMP_FILTERED;
5297 fillargs->flags |= NLM_F_DUMP_FILTERED;
5298 }
5299
5300 err = nlmsg_parse_deprecated_strict(nlh, sizeof(*ifm), tb, IFA_MAX,
5301 ifa_ipv6_policy, extack);
5302 if (err < 0)
5303 return err;
5304
5305 for (i = 0; i <= IFA_MAX; ++i) {
5306 if (!tb[i])
5307 continue;
5308
5309 if (i == IFA_TARGET_NETNSID) {
5310 struct net *net;
5311
5312 fillargs->netnsid = nla_get_s32(tb[i]);
5313 net = rtnl_get_net_ns_capable(sk, fillargs->netnsid);
5314 if (IS_ERR(net)) {
5315 fillargs->netnsid = -1;
5316 NL_SET_ERR_MSG_MOD(extack, "Invalid target network namespace id");
5317 return PTR_ERR(net);
5318 }
5319 *tgt_net = net;
5320 } else {
5321 NL_SET_ERR_MSG_MOD(extack, "Unsupported attribute in dump request");
5322 return -EINVAL;
5323 }
5324 }
5325
5326 return 0;
5327 }
5328
inet6_dump_addr(struct sk_buff * skb,struct netlink_callback * cb,enum addr_type_t type)5329 static int inet6_dump_addr(struct sk_buff *skb, struct netlink_callback *cb,
5330 enum addr_type_t type)
5331 {
5332 const struct nlmsghdr *nlh = cb->nlh;
5333 struct inet6_fill_args fillargs = {
5334 .portid = NETLINK_CB(cb->skb).portid,
5335 .seq = cb->nlh->nlmsg_seq,
5336 .flags = NLM_F_MULTI,
5337 .netnsid = -1,
5338 .type = type,
5339 };
5340 struct net *tgt_net = sock_net(skb->sk);
5341 int idx, s_idx, s_ip_idx;
5342 int h, s_h;
5343 struct net_device *dev;
5344 struct inet6_dev *idev;
5345 struct hlist_head *head;
5346 int err = 0;
5347
5348 s_h = cb->args[0];
5349 s_idx = idx = cb->args[1];
5350 s_ip_idx = cb->args[2];
5351
5352 if (cb->strict_check) {
5353 err = inet6_valid_dump_ifaddr_req(nlh, &fillargs, &tgt_net,
5354 skb->sk, cb);
5355 if (err < 0)
5356 goto put_tgt_net;
5357
5358 err = 0;
5359 if (fillargs.ifindex) {
5360 dev = __dev_get_by_index(tgt_net, fillargs.ifindex);
5361 if (!dev) {
5362 err = -ENODEV;
5363 goto put_tgt_net;
5364 }
5365 idev = __in6_dev_get(dev);
5366 if (idev) {
5367 err = in6_dump_addrs(idev, skb, cb, s_ip_idx,
5368 &fillargs);
5369 if (err > 0)
5370 err = 0;
5371 }
5372 goto put_tgt_net;
5373 }
5374 }
5375
5376 rcu_read_lock();
5377 cb->seq = inet6_base_seq(tgt_net);
5378 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
5379 idx = 0;
5380 head = &tgt_net->dev_index_head[h];
5381 hlist_for_each_entry_rcu(dev, head, index_hlist) {
5382 if (idx < s_idx)
5383 goto cont;
5384 if (h > s_h || idx > s_idx)
5385 s_ip_idx = 0;
5386 idev = __in6_dev_get(dev);
5387 if (!idev)
5388 goto cont;
5389
5390 if (in6_dump_addrs(idev, skb, cb, s_ip_idx,
5391 &fillargs) < 0)
5392 goto done;
5393 cont:
5394 idx++;
5395 }
5396 }
5397 done:
5398 rcu_read_unlock();
5399 cb->args[0] = h;
5400 cb->args[1] = idx;
5401 put_tgt_net:
5402 if (fillargs.netnsid >= 0)
5403 put_net(tgt_net);
5404
5405 return skb->len ? : err;
5406 }
5407
inet6_dump_ifaddr(struct sk_buff * skb,struct netlink_callback * cb)5408 static int inet6_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
5409 {
5410 enum addr_type_t type = UNICAST_ADDR;
5411
5412 return inet6_dump_addr(skb, cb, type);
5413 }
5414
inet6_dump_ifmcaddr(struct sk_buff * skb,struct netlink_callback * cb)5415 static int inet6_dump_ifmcaddr(struct sk_buff *skb, struct netlink_callback *cb)
5416 {
5417 enum addr_type_t type = MULTICAST_ADDR;
5418
5419 return inet6_dump_addr(skb, cb, type);
5420 }
5421
5422
inet6_dump_ifacaddr(struct sk_buff * skb,struct netlink_callback * cb)5423 static int inet6_dump_ifacaddr(struct sk_buff *skb, struct netlink_callback *cb)
5424 {
5425 enum addr_type_t type = ANYCAST_ADDR;
5426
5427 return inet6_dump_addr(skb, cb, type);
5428 }
5429
inet6_rtm_valid_getaddr_req(struct sk_buff * skb,const struct nlmsghdr * nlh,struct nlattr ** tb,struct netlink_ext_ack * extack)5430 static int inet6_rtm_valid_getaddr_req(struct sk_buff *skb,
5431 const struct nlmsghdr *nlh,
5432 struct nlattr **tb,
5433 struct netlink_ext_ack *extack)
5434 {
5435 struct ifaddrmsg *ifm;
5436 int i, err;
5437
5438 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) {
5439 NL_SET_ERR_MSG_MOD(extack, "Invalid header for get address request");
5440 return -EINVAL;
5441 }
5442
5443 if (!netlink_strict_get_check(skb))
5444 return nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFA_MAX,
5445 ifa_ipv6_policy, extack);
5446
5447 ifm = nlmsg_data(nlh);
5448 if (ifm->ifa_prefixlen || ifm->ifa_flags || ifm->ifa_scope) {
5449 NL_SET_ERR_MSG_MOD(extack, "Invalid values in header for get address request");
5450 return -EINVAL;
5451 }
5452
5453 err = nlmsg_parse_deprecated_strict(nlh, sizeof(*ifm), tb, IFA_MAX,
5454 ifa_ipv6_policy, extack);
5455 if (err)
5456 return err;
5457
5458 for (i = 0; i <= IFA_MAX; i++) {
5459 if (!tb[i])
5460 continue;
5461
5462 switch (i) {
5463 case IFA_TARGET_NETNSID:
5464 case IFA_ADDRESS:
5465 case IFA_LOCAL:
5466 break;
5467 default:
5468 NL_SET_ERR_MSG_MOD(extack, "Unsupported attribute in get address request");
5469 return -EINVAL;
5470 }
5471 }
5472
5473 return 0;
5474 }
5475
inet6_rtm_getaddr(struct sk_buff * in_skb,struct nlmsghdr * nlh,struct netlink_ext_ack * extack)5476 static int inet6_rtm_getaddr(struct sk_buff *in_skb, struct nlmsghdr *nlh,
5477 struct netlink_ext_ack *extack)
5478 {
5479 struct net *tgt_net = sock_net(in_skb->sk);
5480 struct inet6_fill_args fillargs = {
5481 .portid = NETLINK_CB(in_skb).portid,
5482 .seq = nlh->nlmsg_seq,
5483 .event = RTM_NEWADDR,
5484 .flags = 0,
5485 .netnsid = -1,
5486 };
5487 struct ifaddrmsg *ifm;
5488 struct nlattr *tb[IFA_MAX+1];
5489 struct in6_addr *addr = NULL, *peer;
5490 struct net_device *dev = NULL;
5491 struct inet6_ifaddr *ifa;
5492 struct sk_buff *skb;
5493 int err;
5494
5495 err = inet6_rtm_valid_getaddr_req(in_skb, nlh, tb, extack);
5496 if (err < 0)
5497 return err;
5498
5499 if (tb[IFA_TARGET_NETNSID]) {
5500 fillargs.netnsid = nla_get_s32(tb[IFA_TARGET_NETNSID]);
5501
5502 tgt_net = rtnl_get_net_ns_capable(NETLINK_CB(in_skb).sk,
5503 fillargs.netnsid);
5504 if (IS_ERR(tgt_net))
5505 return PTR_ERR(tgt_net);
5506 }
5507
5508 addr = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer);
5509 if (!addr) {
5510 err = -EINVAL;
5511 goto errout;
5512 }
5513 ifm = nlmsg_data(nlh);
5514 if (ifm->ifa_index)
5515 dev = dev_get_by_index(tgt_net, ifm->ifa_index);
5516
5517 ifa = ipv6_get_ifaddr(tgt_net, addr, dev, 1);
5518 if (!ifa) {
5519 err = -EADDRNOTAVAIL;
5520 goto errout;
5521 }
5522
5523 skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_KERNEL);
5524 if (!skb) {
5525 err = -ENOBUFS;
5526 goto errout_ifa;
5527 }
5528
5529 err = inet6_fill_ifaddr(skb, ifa, &fillargs);
5530 if (err < 0) {
5531 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
5532 WARN_ON(err == -EMSGSIZE);
5533 kfree_skb(skb);
5534 goto errout_ifa;
5535 }
5536 err = rtnl_unicast(skb, tgt_net, NETLINK_CB(in_skb).portid);
5537 errout_ifa:
5538 in6_ifa_put(ifa);
5539 errout:
5540 dev_put(dev);
5541 if (fillargs.netnsid >= 0)
5542 put_net(tgt_net);
5543
5544 return err;
5545 }
5546
inet6_ifa_notify(int event,struct inet6_ifaddr * ifa)5547 static void inet6_ifa_notify(int event, struct inet6_ifaddr *ifa)
5548 {
5549 struct sk_buff *skb;
5550 struct net *net = dev_net(ifa->idev->dev);
5551 struct inet6_fill_args fillargs = {
5552 .portid = 0,
5553 .seq = 0,
5554 .event = event,
5555 .flags = 0,
5556 .netnsid = -1,
5557 };
5558 int err = -ENOBUFS;
5559
5560 skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_ATOMIC);
5561 if (!skb)
5562 goto errout;
5563
5564 err = inet6_fill_ifaddr(skb, ifa, &fillargs);
5565 if (err < 0) {
5566 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
5567 WARN_ON(err == -EMSGSIZE);
5568 kfree_skb(skb);
5569 goto errout;
5570 }
5571 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
5572 return;
5573 errout:
5574 if (err < 0)
5575 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
5576 }
5577
ipv6_store_devconf(struct ipv6_devconf * cnf,__s32 * array,int bytes)5578 static inline void ipv6_store_devconf(struct ipv6_devconf *cnf,
5579 __s32 *array, int bytes)
5580 {
5581 BUG_ON(bytes < (DEVCONF_MAX * 4));
5582
5583 memset(array, 0, bytes);
5584 array[DEVCONF_FORWARDING] = cnf->forwarding;
5585 array[DEVCONF_HOPLIMIT] = cnf->hop_limit;
5586 array[DEVCONF_MTU6] = cnf->mtu6;
5587 array[DEVCONF_ACCEPT_RA] = cnf->accept_ra;
5588 array[DEVCONF_ACCEPT_REDIRECTS] = cnf->accept_redirects;
5589 array[DEVCONF_AUTOCONF] = cnf->autoconf;
5590 array[DEVCONF_DAD_TRANSMITS] = cnf->dad_transmits;
5591 array[DEVCONF_RTR_SOLICITS] = cnf->rtr_solicits;
5592 array[DEVCONF_RTR_SOLICIT_INTERVAL] =
5593 jiffies_to_msecs(cnf->rtr_solicit_interval);
5594 array[DEVCONF_RTR_SOLICIT_MAX_INTERVAL] =
5595 jiffies_to_msecs(cnf->rtr_solicit_max_interval);
5596 array[DEVCONF_RTR_SOLICIT_DELAY] =
5597 jiffies_to_msecs(cnf->rtr_solicit_delay);
5598 array[DEVCONF_FORCE_MLD_VERSION] = cnf->force_mld_version;
5599 array[DEVCONF_MLDV1_UNSOLICITED_REPORT_INTERVAL] =
5600 jiffies_to_msecs(cnf->mldv1_unsolicited_report_interval);
5601 array[DEVCONF_MLDV2_UNSOLICITED_REPORT_INTERVAL] =
5602 jiffies_to_msecs(cnf->mldv2_unsolicited_report_interval);
5603 array[DEVCONF_USE_TEMPADDR] = cnf->use_tempaddr;
5604 array[DEVCONF_TEMP_VALID_LFT] = cnf->temp_valid_lft;
5605 array[DEVCONF_TEMP_PREFERED_LFT] = cnf->temp_prefered_lft;
5606 array[DEVCONF_REGEN_MAX_RETRY] = cnf->regen_max_retry;
5607 array[DEVCONF_MAX_DESYNC_FACTOR] = cnf->max_desync_factor;
5608 array[DEVCONF_MAX_ADDRESSES] = cnf->max_addresses;
5609 array[DEVCONF_ACCEPT_RA_DEFRTR] = cnf->accept_ra_defrtr;
5610 array[DEVCONF_RA_DEFRTR_METRIC] = cnf->ra_defrtr_metric;
5611 array[DEVCONF_ACCEPT_RA_MIN_HOP_LIMIT] = cnf->accept_ra_min_hop_limit;
5612 array[DEVCONF_ACCEPT_RA_PINFO] = cnf->accept_ra_pinfo;
5613 #ifdef CONFIG_IPV6_ROUTER_PREF
5614 array[DEVCONF_ACCEPT_RA_RTR_PREF] = cnf->accept_ra_rtr_pref;
5615 array[DEVCONF_RTR_PROBE_INTERVAL] =
5616 jiffies_to_msecs(cnf->rtr_probe_interval);
5617 #ifdef CONFIG_IPV6_ROUTE_INFO
5618 array[DEVCONF_ACCEPT_RA_RT_INFO_MIN_PLEN] = cnf->accept_ra_rt_info_min_plen;
5619 array[DEVCONF_ACCEPT_RA_RT_INFO_MAX_PLEN] = cnf->accept_ra_rt_info_max_plen;
5620 #endif
5621 #endif
5622 array[DEVCONF_PROXY_NDP] = cnf->proxy_ndp;
5623 array[DEVCONF_ACCEPT_SOURCE_ROUTE] = cnf->accept_source_route;
5624 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
5625 array[DEVCONF_OPTIMISTIC_DAD] = cnf->optimistic_dad;
5626 array[DEVCONF_USE_OPTIMISTIC] = cnf->use_optimistic;
5627 #endif
5628 #ifdef CONFIG_IPV6_MROUTE
5629 array[DEVCONF_MC_FORWARDING] = atomic_read(&cnf->mc_forwarding);
5630 #endif
5631 array[DEVCONF_DISABLE_IPV6] = cnf->disable_ipv6;
5632 array[DEVCONF_ACCEPT_DAD] = cnf->accept_dad;
5633 array[DEVCONF_FORCE_TLLAO] = cnf->force_tllao;
5634 array[DEVCONF_NDISC_NOTIFY] = cnf->ndisc_notify;
5635 array[DEVCONF_SUPPRESS_FRAG_NDISC] = cnf->suppress_frag_ndisc;
5636 array[DEVCONF_ACCEPT_RA_FROM_LOCAL] = cnf->accept_ra_from_local;
5637 array[DEVCONF_ACCEPT_RA_MTU] = cnf->accept_ra_mtu;
5638 array[DEVCONF_IGNORE_ROUTES_WITH_LINKDOWN] = cnf->ignore_routes_with_linkdown;
5639 /* we omit DEVCONF_STABLE_SECRET for now */
5640 array[DEVCONF_USE_OIF_ADDRS_ONLY] = cnf->use_oif_addrs_only;
5641 array[DEVCONF_DROP_UNICAST_IN_L2_MULTICAST] = cnf->drop_unicast_in_l2_multicast;
5642 array[DEVCONF_DROP_UNSOLICITED_NA] = cnf->drop_unsolicited_na;
5643 array[DEVCONF_KEEP_ADDR_ON_DOWN] = cnf->keep_addr_on_down;
5644 array[DEVCONF_SEG6_ENABLED] = cnf->seg6_enabled;
5645 #ifdef CONFIG_IPV6_SEG6_HMAC
5646 array[DEVCONF_SEG6_REQUIRE_HMAC] = cnf->seg6_require_hmac;
5647 #endif
5648 array[DEVCONF_ENHANCED_DAD] = cnf->enhanced_dad;
5649 array[DEVCONF_ADDR_GEN_MODE] = cnf->addr_gen_mode;
5650 array[DEVCONF_DISABLE_POLICY] = cnf->disable_policy;
5651 array[DEVCONF_NDISC_TCLASS] = cnf->ndisc_tclass;
5652 array[DEVCONF_RPL_SEG_ENABLED] = cnf->rpl_seg_enabled;
5653 array[DEVCONF_IOAM6_ENABLED] = cnf->ioam6_enabled;
5654 array[DEVCONF_IOAM6_ID] = cnf->ioam6_id;
5655 array[DEVCONF_IOAM6_ID_WIDE] = cnf->ioam6_id_wide;
5656 array[DEVCONF_NDISC_EVICT_NOCARRIER] = cnf->ndisc_evict_nocarrier;
5657 array[DEVCONF_ACCEPT_UNTRACKED_NA] = cnf->accept_untracked_na;
5658 array[DEVCONF_ACCEPT_RA_MIN_LFT] = cnf->accept_ra_min_lft;
5659 }
5660
inet6_ifla6_size(void)5661 static inline size_t inet6_ifla6_size(void)
5662 {
5663 return nla_total_size(4) /* IFLA_INET6_FLAGS */
5664 + nla_total_size(sizeof(struct ifla_cacheinfo))
5665 + nla_total_size(DEVCONF_MAX * 4) /* IFLA_INET6_CONF */
5666 + nla_total_size(IPSTATS_MIB_MAX * 8) /* IFLA_INET6_STATS */
5667 + nla_total_size(ICMP6_MIB_MAX * 8) /* IFLA_INET6_ICMP6STATS */
5668 + nla_total_size(sizeof(struct in6_addr)) /* IFLA_INET6_TOKEN */
5669 + nla_total_size(1) /* IFLA_INET6_ADDR_GEN_MODE */
5670 + nla_total_size(4) /* IFLA_INET6_RA_MTU */
5671 + 0;
5672 }
5673
inet6_if_nlmsg_size(void)5674 static inline size_t inet6_if_nlmsg_size(void)
5675 {
5676 return NLMSG_ALIGN(sizeof(struct ifinfomsg))
5677 + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
5678 + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
5679 + nla_total_size(4) /* IFLA_MTU */
5680 + nla_total_size(4) /* IFLA_LINK */
5681 + nla_total_size(1) /* IFLA_OPERSTATE */
5682 + nla_total_size(inet6_ifla6_size()); /* IFLA_PROTINFO */
5683 }
5684
__snmp6_fill_statsdev(u64 * stats,atomic_long_t * mib,int bytes)5685 static inline void __snmp6_fill_statsdev(u64 *stats, atomic_long_t *mib,
5686 int bytes)
5687 {
5688 int i;
5689 int pad = bytes - sizeof(u64) * ICMP6_MIB_MAX;
5690 BUG_ON(pad < 0);
5691
5692 /* Use put_unaligned() because stats may not be aligned for u64. */
5693 put_unaligned(ICMP6_MIB_MAX, &stats[0]);
5694 for (i = 1; i < ICMP6_MIB_MAX; i++)
5695 put_unaligned(atomic_long_read(&mib[i]), &stats[i]);
5696
5697 memset(&stats[ICMP6_MIB_MAX], 0, pad);
5698 }
5699
__snmp6_fill_stats64(u64 * stats,void __percpu * mib,int bytes,size_t syncpoff)5700 static inline void __snmp6_fill_stats64(u64 *stats, void __percpu *mib,
5701 int bytes, size_t syncpoff)
5702 {
5703 int i, c;
5704 u64 buff[IPSTATS_MIB_MAX];
5705 int pad = bytes - sizeof(u64) * IPSTATS_MIB_MAX;
5706
5707 BUG_ON(pad < 0);
5708
5709 memset(buff, 0, sizeof(buff));
5710 buff[0] = IPSTATS_MIB_MAX;
5711
5712 for_each_possible_cpu(c) {
5713 for (i = 1; i < IPSTATS_MIB_MAX; i++)
5714 buff[i] += snmp_get_cpu_field64(mib, c, i, syncpoff);
5715 }
5716
5717 memcpy(stats, buff, IPSTATS_MIB_MAX * sizeof(u64));
5718 memset(&stats[IPSTATS_MIB_MAX], 0, pad);
5719 }
5720
snmp6_fill_stats(u64 * stats,struct inet6_dev * idev,int attrtype,int bytes)5721 static void snmp6_fill_stats(u64 *stats, struct inet6_dev *idev, int attrtype,
5722 int bytes)
5723 {
5724 switch (attrtype) {
5725 case IFLA_INET6_STATS:
5726 __snmp6_fill_stats64(stats, idev->stats.ipv6, bytes,
5727 offsetof(struct ipstats_mib, syncp));
5728 break;
5729 case IFLA_INET6_ICMP6STATS:
5730 __snmp6_fill_statsdev(stats, idev->stats.icmpv6dev->mibs, bytes);
5731 break;
5732 }
5733 }
5734
inet6_fill_ifla6_attrs(struct sk_buff * skb,struct inet6_dev * idev,u32 ext_filter_mask)5735 static int inet6_fill_ifla6_attrs(struct sk_buff *skb, struct inet6_dev *idev,
5736 u32 ext_filter_mask)
5737 {
5738 struct nlattr *nla;
5739 struct ifla_cacheinfo ci;
5740
5741 if (nla_put_u32(skb, IFLA_INET6_FLAGS, idev->if_flags))
5742 goto nla_put_failure;
5743 ci.max_reasm_len = IPV6_MAXPLEN;
5744 ci.tstamp = cstamp_delta(idev->tstamp);
5745 ci.reachable_time = jiffies_to_msecs(idev->nd_parms->reachable_time);
5746 ci.retrans_time = jiffies_to_msecs(NEIGH_VAR(idev->nd_parms, RETRANS_TIME));
5747 if (nla_put(skb, IFLA_INET6_CACHEINFO, sizeof(ci), &ci))
5748 goto nla_put_failure;
5749 nla = nla_reserve(skb, IFLA_INET6_CONF, DEVCONF_MAX * sizeof(s32));
5750 if (!nla)
5751 goto nla_put_failure;
5752 ipv6_store_devconf(&idev->cnf, nla_data(nla), nla_len(nla));
5753
5754 /* XXX - MC not implemented */
5755
5756 if (ext_filter_mask & RTEXT_FILTER_SKIP_STATS)
5757 return 0;
5758
5759 nla = nla_reserve(skb, IFLA_INET6_STATS, IPSTATS_MIB_MAX * sizeof(u64));
5760 if (!nla)
5761 goto nla_put_failure;
5762 snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_STATS, nla_len(nla));
5763
5764 nla = nla_reserve(skb, IFLA_INET6_ICMP6STATS, ICMP6_MIB_MAX * sizeof(u64));
5765 if (!nla)
5766 goto nla_put_failure;
5767 snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_ICMP6STATS, nla_len(nla));
5768
5769 nla = nla_reserve(skb, IFLA_INET6_TOKEN, sizeof(struct in6_addr));
5770 if (!nla)
5771 goto nla_put_failure;
5772 read_lock_bh(&idev->lock);
5773 memcpy(nla_data(nla), idev->token.s6_addr, nla_len(nla));
5774 read_unlock_bh(&idev->lock);
5775
5776 if (nla_put_u8(skb, IFLA_INET6_ADDR_GEN_MODE, idev->cnf.addr_gen_mode))
5777 goto nla_put_failure;
5778
5779 if (idev->ra_mtu &&
5780 nla_put_u32(skb, IFLA_INET6_RA_MTU, idev->ra_mtu))
5781 goto nla_put_failure;
5782
5783 return 0;
5784
5785 nla_put_failure:
5786 return -EMSGSIZE;
5787 }
5788
inet6_get_link_af_size(const struct net_device * dev,u32 ext_filter_mask)5789 static size_t inet6_get_link_af_size(const struct net_device *dev,
5790 u32 ext_filter_mask)
5791 {
5792 if (!__in6_dev_get(dev))
5793 return 0;
5794
5795 return inet6_ifla6_size();
5796 }
5797
inet6_fill_link_af(struct sk_buff * skb,const struct net_device * dev,u32 ext_filter_mask)5798 static int inet6_fill_link_af(struct sk_buff *skb, const struct net_device *dev,
5799 u32 ext_filter_mask)
5800 {
5801 struct inet6_dev *idev = __in6_dev_get(dev);
5802
5803 if (!idev)
5804 return -ENODATA;
5805
5806 if (inet6_fill_ifla6_attrs(skb, idev, ext_filter_mask) < 0)
5807 return -EMSGSIZE;
5808
5809 return 0;
5810 }
5811
inet6_set_iftoken(struct inet6_dev * idev,struct in6_addr * token,struct netlink_ext_ack * extack)5812 static int inet6_set_iftoken(struct inet6_dev *idev, struct in6_addr *token,
5813 struct netlink_ext_ack *extack)
5814 {
5815 struct inet6_ifaddr *ifp;
5816 struct net_device *dev = idev->dev;
5817 bool clear_token, update_rs = false;
5818 struct in6_addr ll_addr;
5819
5820 ASSERT_RTNL();
5821
5822 if (!token)
5823 return -EINVAL;
5824
5825 if (dev->flags & IFF_LOOPBACK) {
5826 NL_SET_ERR_MSG_MOD(extack, "Device is loopback");
5827 return -EINVAL;
5828 }
5829
5830 if (dev->flags & IFF_NOARP) {
5831 NL_SET_ERR_MSG_MOD(extack,
5832 "Device does not do neighbour discovery");
5833 return -EINVAL;
5834 }
5835
5836 if (!ipv6_accept_ra(idev)) {
5837 NL_SET_ERR_MSG_MOD(extack,
5838 "Router advertisement is disabled on device");
5839 return -EINVAL;
5840 }
5841
5842 if (idev->cnf.rtr_solicits == 0) {
5843 NL_SET_ERR_MSG(extack,
5844 "Router solicitation is disabled on device");
5845 return -EINVAL;
5846 }
5847
5848 write_lock_bh(&idev->lock);
5849
5850 BUILD_BUG_ON(sizeof(token->s6_addr) != 16);
5851 memcpy(idev->token.s6_addr + 8, token->s6_addr + 8, 8);
5852
5853 write_unlock_bh(&idev->lock);
5854
5855 clear_token = ipv6_addr_any(token);
5856 if (clear_token)
5857 goto update_lft;
5858
5859 if (!idev->dead && (idev->if_flags & IF_READY) &&
5860 !ipv6_get_lladdr(dev, &ll_addr, IFA_F_TENTATIVE |
5861 IFA_F_OPTIMISTIC)) {
5862 /* If we're not ready, then normal ifup will take care
5863 * of this. Otherwise, we need to request our rs here.
5864 */
5865 ndisc_send_rs(dev, &ll_addr, &in6addr_linklocal_allrouters);
5866 update_rs = true;
5867 }
5868
5869 update_lft:
5870 write_lock_bh(&idev->lock);
5871
5872 if (update_rs) {
5873 idev->if_flags |= IF_RS_SENT;
5874 idev->rs_interval = rfc3315_s14_backoff_init(
5875 idev->cnf.rtr_solicit_interval);
5876 idev->rs_probes = 1;
5877 addrconf_mod_rs_timer(idev, idev->rs_interval);
5878 }
5879
5880 /* Well, that's kinda nasty ... */
5881 list_for_each_entry(ifp, &idev->addr_list, if_list) {
5882 spin_lock(&ifp->lock);
5883 if (ifp->tokenized) {
5884 ifp->valid_lft = 0;
5885 ifp->prefered_lft = 0;
5886 }
5887 spin_unlock(&ifp->lock);
5888 }
5889
5890 write_unlock_bh(&idev->lock);
5891 inet6_ifinfo_notify(RTM_NEWLINK, idev);
5892 addrconf_verify_rtnl(dev_net(dev));
5893 return 0;
5894 }
5895
5896 static const struct nla_policy inet6_af_policy[IFLA_INET6_MAX + 1] = {
5897 [IFLA_INET6_ADDR_GEN_MODE] = { .type = NLA_U8 },
5898 [IFLA_INET6_TOKEN] = { .len = sizeof(struct in6_addr) },
5899 [IFLA_INET6_RA_MTU] = { .type = NLA_REJECT,
5900 .reject_message =
5901 "IFLA_INET6_RA_MTU can not be set" },
5902 };
5903
check_addr_gen_mode(int mode)5904 static int check_addr_gen_mode(int mode)
5905 {
5906 if (mode != IN6_ADDR_GEN_MODE_EUI64 &&
5907 mode != IN6_ADDR_GEN_MODE_NONE &&
5908 mode != IN6_ADDR_GEN_MODE_STABLE_PRIVACY &&
5909 mode != IN6_ADDR_GEN_MODE_RANDOM)
5910 return -EINVAL;
5911 return 1;
5912 }
5913
check_stable_privacy(struct inet6_dev * idev,struct net * net,int mode)5914 static int check_stable_privacy(struct inet6_dev *idev, struct net *net,
5915 int mode)
5916 {
5917 if (mode == IN6_ADDR_GEN_MODE_STABLE_PRIVACY &&
5918 !idev->cnf.stable_secret.initialized &&
5919 !net->ipv6.devconf_dflt->stable_secret.initialized)
5920 return -EINVAL;
5921 return 1;
5922 }
5923
inet6_validate_link_af(const struct net_device * dev,const struct nlattr * nla,struct netlink_ext_ack * extack)5924 static int inet6_validate_link_af(const struct net_device *dev,
5925 const struct nlattr *nla,
5926 struct netlink_ext_ack *extack)
5927 {
5928 struct nlattr *tb[IFLA_INET6_MAX + 1];
5929 struct inet6_dev *idev = NULL;
5930 int err;
5931
5932 if (dev) {
5933 idev = __in6_dev_get(dev);
5934 if (!idev)
5935 return -EAFNOSUPPORT;
5936 }
5937
5938 err = nla_parse_nested_deprecated(tb, IFLA_INET6_MAX, nla,
5939 inet6_af_policy, extack);
5940 if (err)
5941 return err;
5942
5943 if (!tb[IFLA_INET6_TOKEN] && !tb[IFLA_INET6_ADDR_GEN_MODE])
5944 return -EINVAL;
5945
5946 if (tb[IFLA_INET6_ADDR_GEN_MODE]) {
5947 u8 mode = nla_get_u8(tb[IFLA_INET6_ADDR_GEN_MODE]);
5948
5949 if (check_addr_gen_mode(mode) < 0)
5950 return -EINVAL;
5951 if (dev && check_stable_privacy(idev, dev_net(dev), mode) < 0)
5952 return -EINVAL;
5953 }
5954
5955 return 0;
5956 }
5957
inet6_set_link_af(struct net_device * dev,const struct nlattr * nla,struct netlink_ext_ack * extack)5958 static int inet6_set_link_af(struct net_device *dev, const struct nlattr *nla,
5959 struct netlink_ext_ack *extack)
5960 {
5961 struct inet6_dev *idev = __in6_dev_get(dev);
5962 struct nlattr *tb[IFLA_INET6_MAX + 1];
5963 int err;
5964
5965 if (!idev)
5966 return -EAFNOSUPPORT;
5967
5968 if (nla_parse_nested_deprecated(tb, IFLA_INET6_MAX, nla, NULL, NULL) < 0)
5969 return -EINVAL;
5970
5971 if (tb[IFLA_INET6_TOKEN]) {
5972 err = inet6_set_iftoken(idev, nla_data(tb[IFLA_INET6_TOKEN]),
5973 extack);
5974 if (err)
5975 return err;
5976 }
5977
5978 if (tb[IFLA_INET6_ADDR_GEN_MODE]) {
5979 u8 mode = nla_get_u8(tb[IFLA_INET6_ADDR_GEN_MODE]);
5980
5981 idev->cnf.addr_gen_mode = mode;
5982 }
5983
5984 return 0;
5985 }
5986
inet6_fill_ifinfo(struct sk_buff * skb,struct inet6_dev * idev,u32 portid,u32 seq,int event,unsigned int flags)5987 static int inet6_fill_ifinfo(struct sk_buff *skb, struct inet6_dev *idev,
5988 u32 portid, u32 seq, int event, unsigned int flags)
5989 {
5990 struct net_device *dev = idev->dev;
5991 struct ifinfomsg *hdr;
5992 struct nlmsghdr *nlh;
5993 void *protoinfo;
5994
5995 nlh = nlmsg_put(skb, portid, seq, event, sizeof(*hdr), flags);
5996 if (!nlh)
5997 return -EMSGSIZE;
5998
5999 hdr = nlmsg_data(nlh);
6000 hdr->ifi_family = AF_INET6;
6001 hdr->__ifi_pad = 0;
6002 hdr->ifi_type = dev->type;
6003 hdr->ifi_index = dev->ifindex;
6004 hdr->ifi_flags = dev_get_flags(dev);
6005 hdr->ifi_change = 0;
6006
6007 if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
6008 (dev->addr_len &&
6009 nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
6010 nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
6011 (dev->ifindex != dev_get_iflink(dev) &&
6012 nla_put_u32(skb, IFLA_LINK, dev_get_iflink(dev))) ||
6013 nla_put_u8(skb, IFLA_OPERSTATE,
6014 netif_running(dev) ? READ_ONCE(dev->operstate) : IF_OPER_DOWN))
6015 goto nla_put_failure;
6016 protoinfo = nla_nest_start_noflag(skb, IFLA_PROTINFO);
6017 if (!protoinfo)
6018 goto nla_put_failure;
6019
6020 if (inet6_fill_ifla6_attrs(skb, idev, 0) < 0)
6021 goto nla_put_failure;
6022
6023 nla_nest_end(skb, protoinfo);
6024 nlmsg_end(skb, nlh);
6025 return 0;
6026
6027 nla_put_failure:
6028 nlmsg_cancel(skb, nlh);
6029 return -EMSGSIZE;
6030 }
6031
inet6_valid_dump_ifinfo(const struct nlmsghdr * nlh,struct netlink_ext_ack * extack)6032 static int inet6_valid_dump_ifinfo(const struct nlmsghdr *nlh,
6033 struct netlink_ext_ack *extack)
6034 {
6035 struct ifinfomsg *ifm;
6036
6037 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) {
6038 NL_SET_ERR_MSG_MOD(extack, "Invalid header for link dump request");
6039 return -EINVAL;
6040 }
6041
6042 if (nlmsg_attrlen(nlh, sizeof(*ifm))) {
6043 NL_SET_ERR_MSG_MOD(extack, "Invalid data after header");
6044 return -EINVAL;
6045 }
6046
6047 ifm = nlmsg_data(nlh);
6048 if (ifm->__ifi_pad || ifm->ifi_type || ifm->ifi_flags ||
6049 ifm->ifi_change || ifm->ifi_index) {
6050 NL_SET_ERR_MSG_MOD(extack, "Invalid values in header for dump request");
6051 return -EINVAL;
6052 }
6053
6054 return 0;
6055 }
6056
inet6_dump_ifinfo(struct sk_buff * skb,struct netlink_callback * cb)6057 static int inet6_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
6058 {
6059 struct net *net = sock_net(skb->sk);
6060 int h, s_h;
6061 int idx = 0, s_idx;
6062 struct net_device *dev;
6063 struct inet6_dev *idev;
6064 struct hlist_head *head;
6065
6066 /* only requests using strict checking can pass data to
6067 * influence the dump
6068 */
6069 if (cb->strict_check) {
6070 int err = inet6_valid_dump_ifinfo(cb->nlh, cb->extack);
6071
6072 if (err < 0)
6073 return err;
6074 }
6075
6076 s_h = cb->args[0];
6077 s_idx = cb->args[1];
6078
6079 rcu_read_lock();
6080 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
6081 idx = 0;
6082 head = &net->dev_index_head[h];
6083 hlist_for_each_entry_rcu(dev, head, index_hlist) {
6084 if (idx < s_idx)
6085 goto cont;
6086 idev = __in6_dev_get(dev);
6087 if (!idev)
6088 goto cont;
6089 if (inet6_fill_ifinfo(skb, idev,
6090 NETLINK_CB(cb->skb).portid,
6091 cb->nlh->nlmsg_seq,
6092 RTM_NEWLINK, NLM_F_MULTI) < 0)
6093 goto out;
6094 cont:
6095 idx++;
6096 }
6097 }
6098 out:
6099 rcu_read_unlock();
6100 cb->args[1] = idx;
6101 cb->args[0] = h;
6102
6103 return skb->len;
6104 }
6105
inet6_ifinfo_notify(int event,struct inet6_dev * idev)6106 void inet6_ifinfo_notify(int event, struct inet6_dev *idev)
6107 {
6108 struct sk_buff *skb;
6109 struct net *net = dev_net(idev->dev);
6110 int err = -ENOBUFS;
6111
6112 skb = nlmsg_new(inet6_if_nlmsg_size(), GFP_ATOMIC);
6113 if (!skb)
6114 goto errout;
6115
6116 err = inet6_fill_ifinfo(skb, idev, 0, 0, event, 0);
6117 if (err < 0) {
6118 /* -EMSGSIZE implies BUG in inet6_if_nlmsg_size() */
6119 WARN_ON(err == -EMSGSIZE);
6120 kfree_skb(skb);
6121 goto errout;
6122 }
6123 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFINFO, NULL, GFP_ATOMIC);
6124 return;
6125 errout:
6126 if (err < 0)
6127 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFINFO, err);
6128 }
6129
inet6_prefix_nlmsg_size(void)6130 static inline size_t inet6_prefix_nlmsg_size(void)
6131 {
6132 return NLMSG_ALIGN(sizeof(struct prefixmsg))
6133 + nla_total_size(sizeof(struct in6_addr))
6134 + nla_total_size(sizeof(struct prefix_cacheinfo));
6135 }
6136
inet6_fill_prefix(struct sk_buff * skb,struct inet6_dev * idev,struct prefix_info * pinfo,u32 portid,u32 seq,int event,unsigned int flags)6137 static int inet6_fill_prefix(struct sk_buff *skb, struct inet6_dev *idev,
6138 struct prefix_info *pinfo, u32 portid, u32 seq,
6139 int event, unsigned int flags)
6140 {
6141 struct prefixmsg *pmsg;
6142 struct nlmsghdr *nlh;
6143 struct prefix_cacheinfo ci;
6144
6145 nlh = nlmsg_put(skb, portid, seq, event, sizeof(*pmsg), flags);
6146 if (!nlh)
6147 return -EMSGSIZE;
6148
6149 pmsg = nlmsg_data(nlh);
6150 pmsg->prefix_family = AF_INET6;
6151 pmsg->prefix_pad1 = 0;
6152 pmsg->prefix_pad2 = 0;
6153 pmsg->prefix_ifindex = idev->dev->ifindex;
6154 pmsg->prefix_len = pinfo->prefix_len;
6155 pmsg->prefix_type = pinfo->type;
6156 pmsg->prefix_pad3 = 0;
6157 pmsg->prefix_flags = pinfo->flags;
6158
6159 if (nla_put(skb, PREFIX_ADDRESS, sizeof(pinfo->prefix), &pinfo->prefix))
6160 goto nla_put_failure;
6161 ci.preferred_time = ntohl(pinfo->prefered);
6162 ci.valid_time = ntohl(pinfo->valid);
6163 if (nla_put(skb, PREFIX_CACHEINFO, sizeof(ci), &ci))
6164 goto nla_put_failure;
6165 nlmsg_end(skb, nlh);
6166 return 0;
6167
6168 nla_put_failure:
6169 nlmsg_cancel(skb, nlh);
6170 return -EMSGSIZE;
6171 }
6172
inet6_prefix_notify(int event,struct inet6_dev * idev,struct prefix_info * pinfo)6173 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
6174 struct prefix_info *pinfo)
6175 {
6176 struct sk_buff *skb;
6177 struct net *net = dev_net(idev->dev);
6178 int err = -ENOBUFS;
6179
6180 skb = nlmsg_new(inet6_prefix_nlmsg_size(), GFP_ATOMIC);
6181 if (!skb)
6182 goto errout;
6183
6184 err = inet6_fill_prefix(skb, idev, pinfo, 0, 0, event, 0);
6185 if (err < 0) {
6186 /* -EMSGSIZE implies BUG in inet6_prefix_nlmsg_size() */
6187 WARN_ON(err == -EMSGSIZE);
6188 kfree_skb(skb);
6189 goto errout;
6190 }
6191 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_PREFIX, NULL, GFP_ATOMIC);
6192 return;
6193 errout:
6194 if (err < 0)
6195 rtnl_set_sk_err(net, RTNLGRP_IPV6_PREFIX, err);
6196 }
6197
__ipv6_ifa_notify(int event,struct inet6_ifaddr * ifp)6198 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
6199 {
6200 struct net *net = dev_net(ifp->idev->dev);
6201
6202 if (event)
6203 ASSERT_RTNL();
6204
6205 inet6_ifa_notify(event ? : RTM_NEWADDR, ifp);
6206
6207 switch (event) {
6208 case RTM_NEWADDR:
6209 /*
6210 * If the address was optimistic we inserted the route at the
6211 * start of our DAD process, so we don't need to do it again.
6212 * If the device was taken down in the middle of the DAD
6213 * cycle there is a race where we could get here without a
6214 * host route, so nothing to insert. That will be fixed when
6215 * the device is brought up.
6216 */
6217 if (ifp->rt && !rcu_access_pointer(ifp->rt->fib6_node)) {
6218 ip6_ins_rt(net, ifp->rt);
6219 } else if (!ifp->rt && (ifp->idev->dev->flags & IFF_UP)) {
6220 pr_warn("BUG: Address %pI6c on device %s is missing its host route.\n",
6221 &ifp->addr, ifp->idev->dev->name);
6222 }
6223
6224 if (ifp->idev->cnf.forwarding)
6225 addrconf_join_anycast(ifp);
6226 if (!ipv6_addr_any(&ifp->peer_addr))
6227 addrconf_prefix_route(&ifp->peer_addr, 128,
6228 ifp->rt_priority, ifp->idev->dev,
6229 0, 0, GFP_ATOMIC);
6230 break;
6231 case RTM_DELADDR:
6232 if (ifp->idev->cnf.forwarding)
6233 addrconf_leave_anycast(ifp);
6234 addrconf_leave_solict(ifp->idev, &ifp->addr);
6235 if (!ipv6_addr_any(&ifp->peer_addr)) {
6236 struct fib6_info *rt;
6237
6238 rt = addrconf_get_prefix_route(&ifp->peer_addr, 128,
6239 ifp->idev->dev, 0, 0,
6240 false);
6241 if (rt)
6242 ip6_del_rt(net, rt, false);
6243 }
6244 if (ifp->rt) {
6245 ip6_del_rt(net, ifp->rt, false);
6246 ifp->rt = NULL;
6247 }
6248 rt_genid_bump_ipv6(net);
6249 break;
6250 }
6251 atomic_inc(&net->ipv6.dev_addr_genid);
6252 }
6253
ipv6_ifa_notify(int event,struct inet6_ifaddr * ifp)6254 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
6255 {
6256 if (likely(ifp->idev->dead == 0))
6257 __ipv6_ifa_notify(event, ifp);
6258 }
6259
6260 #ifdef CONFIG_SYSCTL
6261
addrconf_sysctl_forward(struct ctl_table * ctl,int write,void * buffer,size_t * lenp,loff_t * ppos)6262 static int addrconf_sysctl_forward(struct ctl_table *ctl, int write,
6263 void *buffer, size_t *lenp, loff_t *ppos)
6264 {
6265 int *valp = ctl->data;
6266 int val = *valp;
6267 loff_t pos = *ppos;
6268 struct ctl_table lctl;
6269 int ret;
6270
6271 /*
6272 * ctl->data points to idev->cnf.forwarding, we should
6273 * not modify it until we get the rtnl lock.
6274 */
6275 lctl = *ctl;
6276 lctl.data = &val;
6277
6278 ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
6279
6280 if (write)
6281 ret = addrconf_fixup_forwarding(ctl, valp, val);
6282 if (ret)
6283 *ppos = pos;
6284 return ret;
6285 }
6286
addrconf_sysctl_mtu(struct ctl_table * ctl,int write,void * buffer,size_t * lenp,loff_t * ppos)6287 static int addrconf_sysctl_mtu(struct ctl_table *ctl, int write,
6288 void *buffer, size_t *lenp, loff_t *ppos)
6289 {
6290 struct inet6_dev *idev = ctl->extra1;
6291 int min_mtu = IPV6_MIN_MTU;
6292 struct ctl_table lctl;
6293
6294 lctl = *ctl;
6295 lctl.extra1 = &min_mtu;
6296 lctl.extra2 = idev ? &idev->dev->mtu : NULL;
6297
6298 return proc_dointvec_minmax(&lctl, write, buffer, lenp, ppos);
6299 }
6300
dev_disable_change(struct inet6_dev * idev)6301 static void dev_disable_change(struct inet6_dev *idev)
6302 {
6303 struct netdev_notifier_info info;
6304
6305 if (!idev || !idev->dev)
6306 return;
6307
6308 netdev_notifier_info_init(&info, idev->dev);
6309 if (idev->cnf.disable_ipv6)
6310 addrconf_notify(NULL, NETDEV_DOWN, &info);
6311 else
6312 addrconf_notify(NULL, NETDEV_UP, &info);
6313 }
6314
addrconf_disable_change(struct net * net,__s32 newf)6315 static void addrconf_disable_change(struct net *net, __s32 newf)
6316 {
6317 struct net_device *dev;
6318 struct inet6_dev *idev;
6319
6320 for_each_netdev(net, dev) {
6321 idev = __in6_dev_get(dev);
6322 if (idev) {
6323 int changed = (!idev->cnf.disable_ipv6) ^ (!newf);
6324
6325 WRITE_ONCE(idev->cnf.disable_ipv6, newf);
6326 if (changed)
6327 dev_disable_change(idev);
6328 }
6329 }
6330 }
6331
addrconf_disable_ipv6(struct ctl_table * table,int * p,int newf)6332 static int addrconf_disable_ipv6(struct ctl_table *table, int *p, int newf)
6333 {
6334 struct net *net;
6335 int old;
6336
6337 if (!rtnl_trylock())
6338 return restart_syscall();
6339
6340 net = (struct net *)table->extra2;
6341 old = *p;
6342 WRITE_ONCE(*p, newf);
6343
6344 if (p == &net->ipv6.devconf_dflt->disable_ipv6) {
6345 rtnl_unlock();
6346 return 0;
6347 }
6348
6349 if (p == &net->ipv6.devconf_all->disable_ipv6) {
6350 WRITE_ONCE(net->ipv6.devconf_dflt->disable_ipv6, newf);
6351 addrconf_disable_change(net, newf);
6352 } else if ((!newf) ^ (!old))
6353 dev_disable_change((struct inet6_dev *)table->extra1);
6354
6355 rtnl_unlock();
6356 return 0;
6357 }
6358
addrconf_sysctl_disable(struct ctl_table * ctl,int write,void * buffer,size_t * lenp,loff_t * ppos)6359 static int addrconf_sysctl_disable(struct ctl_table *ctl, int write,
6360 void *buffer, size_t *lenp, loff_t *ppos)
6361 {
6362 int *valp = ctl->data;
6363 int val = *valp;
6364 loff_t pos = *ppos;
6365 struct ctl_table lctl;
6366 int ret;
6367
6368 /*
6369 * ctl->data points to idev->cnf.disable_ipv6, we should
6370 * not modify it until we get the rtnl lock.
6371 */
6372 lctl = *ctl;
6373 lctl.data = &val;
6374
6375 ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
6376
6377 if (write)
6378 ret = addrconf_disable_ipv6(ctl, valp, val);
6379 if (ret)
6380 *ppos = pos;
6381 return ret;
6382 }
6383
addrconf_sysctl_proxy_ndp(struct ctl_table * ctl,int write,void * buffer,size_t * lenp,loff_t * ppos)6384 static int addrconf_sysctl_proxy_ndp(struct ctl_table *ctl, int write,
6385 void *buffer, size_t *lenp, loff_t *ppos)
6386 {
6387 int *valp = ctl->data;
6388 int ret;
6389 int old, new;
6390
6391 old = *valp;
6392 ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
6393 new = *valp;
6394
6395 if (write && old != new) {
6396 struct net *net = ctl->extra2;
6397
6398 if (!rtnl_trylock())
6399 return restart_syscall();
6400
6401 if (valp == &net->ipv6.devconf_dflt->proxy_ndp)
6402 inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
6403 NETCONFA_PROXY_NEIGH,
6404 NETCONFA_IFINDEX_DEFAULT,
6405 net->ipv6.devconf_dflt);
6406 else if (valp == &net->ipv6.devconf_all->proxy_ndp)
6407 inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
6408 NETCONFA_PROXY_NEIGH,
6409 NETCONFA_IFINDEX_ALL,
6410 net->ipv6.devconf_all);
6411 else {
6412 struct inet6_dev *idev = ctl->extra1;
6413
6414 inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
6415 NETCONFA_PROXY_NEIGH,
6416 idev->dev->ifindex,
6417 &idev->cnf);
6418 }
6419 rtnl_unlock();
6420 }
6421
6422 return ret;
6423 }
6424
addrconf_sysctl_addr_gen_mode(struct ctl_table * ctl,int write,void * buffer,size_t * lenp,loff_t * ppos)6425 static int addrconf_sysctl_addr_gen_mode(struct ctl_table *ctl, int write,
6426 void *buffer, size_t *lenp,
6427 loff_t *ppos)
6428 {
6429 int ret = 0;
6430 u32 new_val;
6431 struct inet6_dev *idev = (struct inet6_dev *)ctl->extra1;
6432 struct net *net = (struct net *)ctl->extra2;
6433 struct ctl_table tmp = {
6434 .data = &new_val,
6435 .maxlen = sizeof(new_val),
6436 .mode = ctl->mode,
6437 };
6438
6439 if (!rtnl_trylock())
6440 return restart_syscall();
6441
6442 new_val = *((u32 *)ctl->data);
6443
6444 ret = proc_douintvec(&tmp, write, buffer, lenp, ppos);
6445 if (ret != 0)
6446 goto out;
6447
6448 if (write) {
6449 if (check_addr_gen_mode(new_val) < 0) {
6450 ret = -EINVAL;
6451 goto out;
6452 }
6453
6454 if (idev) {
6455 if (check_stable_privacy(idev, net, new_val) < 0) {
6456 ret = -EINVAL;
6457 goto out;
6458 }
6459
6460 if (idev->cnf.addr_gen_mode != new_val) {
6461 idev->cnf.addr_gen_mode = new_val;
6462 addrconf_init_auto_addrs(idev->dev);
6463 }
6464 } else if (&net->ipv6.devconf_all->addr_gen_mode == ctl->data) {
6465 struct net_device *dev;
6466
6467 net->ipv6.devconf_dflt->addr_gen_mode = new_val;
6468 for_each_netdev(net, dev) {
6469 idev = __in6_dev_get(dev);
6470 if (idev &&
6471 idev->cnf.addr_gen_mode != new_val) {
6472 idev->cnf.addr_gen_mode = new_val;
6473 addrconf_init_auto_addrs(idev->dev);
6474 }
6475 }
6476 }
6477
6478 *((u32 *)ctl->data) = new_val;
6479 }
6480
6481 out:
6482 rtnl_unlock();
6483
6484 return ret;
6485 }
6486
addrconf_sysctl_stable_secret(struct ctl_table * ctl,int write,void * buffer,size_t * lenp,loff_t * ppos)6487 static int addrconf_sysctl_stable_secret(struct ctl_table *ctl, int write,
6488 void *buffer, size_t *lenp,
6489 loff_t *ppos)
6490 {
6491 int err;
6492 struct in6_addr addr;
6493 char str[IPV6_MAX_STRLEN];
6494 struct ctl_table lctl = *ctl;
6495 struct net *net = ctl->extra2;
6496 struct ipv6_stable_secret *secret = ctl->data;
6497
6498 if (&net->ipv6.devconf_all->stable_secret == ctl->data)
6499 return -EIO;
6500
6501 lctl.maxlen = IPV6_MAX_STRLEN;
6502 lctl.data = str;
6503
6504 if (!rtnl_trylock())
6505 return restart_syscall();
6506
6507 if (!write && !secret->initialized) {
6508 err = -EIO;
6509 goto out;
6510 }
6511
6512 err = snprintf(str, sizeof(str), "%pI6", &secret->secret);
6513 if (err >= sizeof(str)) {
6514 err = -EIO;
6515 goto out;
6516 }
6517
6518 err = proc_dostring(&lctl, write, buffer, lenp, ppos);
6519 if (err || !write)
6520 goto out;
6521
6522 if (in6_pton(str, -1, addr.in6_u.u6_addr8, -1, NULL) != 1) {
6523 err = -EIO;
6524 goto out;
6525 }
6526
6527 secret->initialized = true;
6528 secret->secret = addr;
6529
6530 if (&net->ipv6.devconf_dflt->stable_secret == ctl->data) {
6531 struct net_device *dev;
6532
6533 for_each_netdev(net, dev) {
6534 struct inet6_dev *idev = __in6_dev_get(dev);
6535
6536 if (idev) {
6537 idev->cnf.addr_gen_mode =
6538 IN6_ADDR_GEN_MODE_STABLE_PRIVACY;
6539 }
6540 }
6541 } else {
6542 struct inet6_dev *idev = ctl->extra1;
6543
6544 idev->cnf.addr_gen_mode = IN6_ADDR_GEN_MODE_STABLE_PRIVACY;
6545 }
6546
6547 out:
6548 rtnl_unlock();
6549
6550 return err;
6551 }
6552
6553 static
addrconf_sysctl_ignore_routes_with_linkdown(struct ctl_table * ctl,int write,void * buffer,size_t * lenp,loff_t * ppos)6554 int addrconf_sysctl_ignore_routes_with_linkdown(struct ctl_table *ctl,
6555 int write, void *buffer,
6556 size_t *lenp,
6557 loff_t *ppos)
6558 {
6559 int *valp = ctl->data;
6560 int val = *valp;
6561 loff_t pos = *ppos;
6562 struct ctl_table lctl;
6563 int ret;
6564
6565 /* ctl->data points to idev->cnf.ignore_routes_when_linkdown
6566 * we should not modify it until we get the rtnl lock.
6567 */
6568 lctl = *ctl;
6569 lctl.data = &val;
6570
6571 ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
6572
6573 if (write)
6574 ret = addrconf_fixup_linkdown(ctl, valp, val);
6575 if (ret)
6576 *ppos = pos;
6577 return ret;
6578 }
6579
6580 static
addrconf_set_nopolicy(struct rt6_info * rt,int action)6581 void addrconf_set_nopolicy(struct rt6_info *rt, int action)
6582 {
6583 if (rt) {
6584 if (action)
6585 rt->dst.flags |= DST_NOPOLICY;
6586 else
6587 rt->dst.flags &= ~DST_NOPOLICY;
6588 }
6589 }
6590
6591 static
addrconf_disable_policy_idev(struct inet6_dev * idev,int val)6592 void addrconf_disable_policy_idev(struct inet6_dev *idev, int val)
6593 {
6594 struct inet6_ifaddr *ifa;
6595
6596 read_lock_bh(&idev->lock);
6597 list_for_each_entry(ifa, &idev->addr_list, if_list) {
6598 spin_lock(&ifa->lock);
6599 if (ifa->rt) {
6600 /* host routes only use builtin fib6_nh */
6601 struct fib6_nh *nh = ifa->rt->fib6_nh;
6602 int cpu;
6603
6604 rcu_read_lock();
6605 ifa->rt->dst_nopolicy = val ? true : false;
6606 if (nh->rt6i_pcpu) {
6607 for_each_possible_cpu(cpu) {
6608 struct rt6_info **rtp;
6609
6610 rtp = per_cpu_ptr(nh->rt6i_pcpu, cpu);
6611 addrconf_set_nopolicy(*rtp, val);
6612 }
6613 }
6614 rcu_read_unlock();
6615 }
6616 spin_unlock(&ifa->lock);
6617 }
6618 read_unlock_bh(&idev->lock);
6619 }
6620
6621 static
addrconf_disable_policy(struct ctl_table * ctl,int * valp,int val)6622 int addrconf_disable_policy(struct ctl_table *ctl, int *valp, int val)
6623 {
6624 struct inet6_dev *idev;
6625 struct net *net;
6626
6627 if (!rtnl_trylock())
6628 return restart_syscall();
6629
6630 *valp = val;
6631
6632 net = (struct net *)ctl->extra2;
6633 if (valp == &net->ipv6.devconf_dflt->disable_policy) {
6634 rtnl_unlock();
6635 return 0;
6636 }
6637
6638 if (valp == &net->ipv6.devconf_all->disable_policy) {
6639 struct net_device *dev;
6640
6641 for_each_netdev(net, dev) {
6642 idev = __in6_dev_get(dev);
6643 if (idev)
6644 addrconf_disable_policy_idev(idev, val);
6645 }
6646 } else {
6647 idev = (struct inet6_dev *)ctl->extra1;
6648 addrconf_disable_policy_idev(idev, val);
6649 }
6650
6651 rtnl_unlock();
6652 return 0;
6653 }
6654
addrconf_sysctl_disable_policy(struct ctl_table * ctl,int write,void * buffer,size_t * lenp,loff_t * ppos)6655 static int addrconf_sysctl_disable_policy(struct ctl_table *ctl, int write,
6656 void *buffer, size_t *lenp, loff_t *ppos)
6657 {
6658 int *valp = ctl->data;
6659 int val = *valp;
6660 loff_t pos = *ppos;
6661 struct ctl_table lctl;
6662 int ret;
6663
6664 lctl = *ctl;
6665 lctl.data = &val;
6666 ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
6667
6668 if (write && (*valp != val))
6669 ret = addrconf_disable_policy(ctl, valp, val);
6670
6671 if (ret)
6672 *ppos = pos;
6673
6674 return ret;
6675 }
6676
6677 static int minus_one = -1;
6678 static const int two_five_five = 255;
6679 static u32 ioam6_if_id_max = U16_MAX;
6680
6681 static const struct ctl_table addrconf_sysctl[] = {
6682 {
6683 .procname = "forwarding",
6684 .data = &ipv6_devconf.forwarding,
6685 .maxlen = sizeof(int),
6686 .mode = 0644,
6687 .proc_handler = addrconf_sysctl_forward,
6688 },
6689 {
6690 .procname = "hop_limit",
6691 .data = &ipv6_devconf.hop_limit,
6692 .maxlen = sizeof(int),
6693 .mode = 0644,
6694 .proc_handler = proc_dointvec_minmax,
6695 .extra1 = (void *)SYSCTL_ONE,
6696 .extra2 = (void *)&two_five_five,
6697 },
6698 {
6699 .procname = "mtu",
6700 .data = &ipv6_devconf.mtu6,
6701 .maxlen = sizeof(int),
6702 .mode = 0644,
6703 .proc_handler = addrconf_sysctl_mtu,
6704 },
6705 {
6706 .procname = "accept_ra",
6707 .data = &ipv6_devconf.accept_ra,
6708 .maxlen = sizeof(int),
6709 .mode = 0644,
6710 .proc_handler = proc_dointvec,
6711 },
6712 {
6713 .procname = "accept_redirects",
6714 .data = &ipv6_devconf.accept_redirects,
6715 .maxlen = sizeof(int),
6716 .mode = 0644,
6717 .proc_handler = proc_dointvec,
6718 },
6719 {
6720 .procname = "autoconf",
6721 .data = &ipv6_devconf.autoconf,
6722 .maxlen = sizeof(int),
6723 .mode = 0644,
6724 .proc_handler = proc_dointvec,
6725 },
6726 {
6727 .procname = "dad_transmits",
6728 .data = &ipv6_devconf.dad_transmits,
6729 .maxlen = sizeof(int),
6730 .mode = 0644,
6731 .proc_handler = proc_dointvec,
6732 },
6733 {
6734 .procname = "router_solicitations",
6735 .data = &ipv6_devconf.rtr_solicits,
6736 .maxlen = sizeof(int),
6737 .mode = 0644,
6738 .proc_handler = proc_dointvec_minmax,
6739 .extra1 = &minus_one,
6740 },
6741 {
6742 .procname = "router_solicitation_interval",
6743 .data = &ipv6_devconf.rtr_solicit_interval,
6744 .maxlen = sizeof(int),
6745 .mode = 0644,
6746 .proc_handler = proc_dointvec_jiffies,
6747 },
6748 {
6749 .procname = "router_solicitation_max_interval",
6750 .data = &ipv6_devconf.rtr_solicit_max_interval,
6751 .maxlen = sizeof(int),
6752 .mode = 0644,
6753 .proc_handler = proc_dointvec_jiffies,
6754 },
6755 {
6756 .procname = "router_solicitation_delay",
6757 .data = &ipv6_devconf.rtr_solicit_delay,
6758 .maxlen = sizeof(int),
6759 .mode = 0644,
6760 .proc_handler = proc_dointvec_jiffies,
6761 },
6762 {
6763 .procname = "force_mld_version",
6764 .data = &ipv6_devconf.force_mld_version,
6765 .maxlen = sizeof(int),
6766 .mode = 0644,
6767 .proc_handler = proc_dointvec,
6768 },
6769 {
6770 .procname = "mldv1_unsolicited_report_interval",
6771 .data =
6772 &ipv6_devconf.mldv1_unsolicited_report_interval,
6773 .maxlen = sizeof(int),
6774 .mode = 0644,
6775 .proc_handler = proc_dointvec_ms_jiffies,
6776 },
6777 {
6778 .procname = "mldv2_unsolicited_report_interval",
6779 .data =
6780 &ipv6_devconf.mldv2_unsolicited_report_interval,
6781 .maxlen = sizeof(int),
6782 .mode = 0644,
6783 .proc_handler = proc_dointvec_ms_jiffies,
6784 },
6785 {
6786 .procname = "use_tempaddr",
6787 .data = &ipv6_devconf.use_tempaddr,
6788 .maxlen = sizeof(int),
6789 .mode = 0644,
6790 .proc_handler = proc_dointvec,
6791 },
6792 {
6793 .procname = "temp_valid_lft",
6794 .data = &ipv6_devconf.temp_valid_lft,
6795 .maxlen = sizeof(int),
6796 .mode = 0644,
6797 .proc_handler = proc_dointvec,
6798 },
6799 {
6800 .procname = "temp_prefered_lft",
6801 .data = &ipv6_devconf.temp_prefered_lft,
6802 .maxlen = sizeof(int),
6803 .mode = 0644,
6804 .proc_handler = proc_dointvec,
6805 },
6806 {
6807 .procname = "regen_max_retry",
6808 .data = &ipv6_devconf.regen_max_retry,
6809 .maxlen = sizeof(int),
6810 .mode = 0644,
6811 .proc_handler = proc_dointvec,
6812 },
6813 {
6814 .procname = "max_desync_factor",
6815 .data = &ipv6_devconf.max_desync_factor,
6816 .maxlen = sizeof(int),
6817 .mode = 0644,
6818 .proc_handler = proc_dointvec,
6819 },
6820 {
6821 .procname = "max_addresses",
6822 .data = &ipv6_devconf.max_addresses,
6823 .maxlen = sizeof(int),
6824 .mode = 0644,
6825 .proc_handler = proc_dointvec,
6826 },
6827 {
6828 .procname = "accept_ra_defrtr",
6829 .data = &ipv6_devconf.accept_ra_defrtr,
6830 .maxlen = sizeof(int),
6831 .mode = 0644,
6832 .proc_handler = proc_dointvec,
6833 },
6834 {
6835 .procname = "ra_defrtr_metric",
6836 .data = &ipv6_devconf.ra_defrtr_metric,
6837 .maxlen = sizeof(u32),
6838 .mode = 0644,
6839 .proc_handler = proc_douintvec_minmax,
6840 .extra1 = (void *)SYSCTL_ONE,
6841 },
6842 {
6843 .procname = "accept_ra_min_hop_limit",
6844 .data = &ipv6_devconf.accept_ra_min_hop_limit,
6845 .maxlen = sizeof(int),
6846 .mode = 0644,
6847 .proc_handler = proc_dointvec,
6848 },
6849 {
6850 .procname = "accept_ra_min_lft",
6851 .data = &ipv6_devconf.accept_ra_min_lft,
6852 .maxlen = sizeof(int),
6853 .mode = 0644,
6854 .proc_handler = proc_dointvec,
6855 },
6856 {
6857 .procname = "accept_ra_pinfo",
6858 .data = &ipv6_devconf.accept_ra_pinfo,
6859 .maxlen = sizeof(int),
6860 .mode = 0644,
6861 .proc_handler = proc_dointvec,
6862 },
6863 #ifdef CONFIG_IPV6_ROUTER_PREF
6864 {
6865 .procname = "accept_ra_rtr_pref",
6866 .data = &ipv6_devconf.accept_ra_rtr_pref,
6867 .maxlen = sizeof(int),
6868 .mode = 0644,
6869 .proc_handler = proc_dointvec,
6870 },
6871 {
6872 .procname = "router_probe_interval",
6873 .data = &ipv6_devconf.rtr_probe_interval,
6874 .maxlen = sizeof(int),
6875 .mode = 0644,
6876 .proc_handler = proc_dointvec_jiffies,
6877 },
6878 #ifdef CONFIG_IPV6_ROUTE_INFO
6879 {
6880 .procname = "accept_ra_rt_info_min_plen",
6881 .data = &ipv6_devconf.accept_ra_rt_info_min_plen,
6882 .maxlen = sizeof(int),
6883 .mode = 0644,
6884 .proc_handler = proc_dointvec,
6885 },
6886 {
6887 .procname = "accept_ra_rt_info_max_plen",
6888 .data = &ipv6_devconf.accept_ra_rt_info_max_plen,
6889 .maxlen = sizeof(int),
6890 .mode = 0644,
6891 .proc_handler = proc_dointvec,
6892 },
6893 #endif
6894 #endif
6895 {
6896 .procname = "proxy_ndp",
6897 .data = &ipv6_devconf.proxy_ndp,
6898 .maxlen = sizeof(int),
6899 .mode = 0644,
6900 .proc_handler = addrconf_sysctl_proxy_ndp,
6901 },
6902 {
6903 .procname = "accept_source_route",
6904 .data = &ipv6_devconf.accept_source_route,
6905 .maxlen = sizeof(int),
6906 .mode = 0644,
6907 .proc_handler = proc_dointvec,
6908 },
6909 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
6910 {
6911 .procname = "optimistic_dad",
6912 .data = &ipv6_devconf.optimistic_dad,
6913 .maxlen = sizeof(int),
6914 .mode = 0644,
6915 .proc_handler = proc_dointvec,
6916 },
6917 {
6918 .procname = "use_optimistic",
6919 .data = &ipv6_devconf.use_optimistic,
6920 .maxlen = sizeof(int),
6921 .mode = 0644,
6922 .proc_handler = proc_dointvec,
6923 },
6924 #endif
6925 #ifdef CONFIG_IPV6_MROUTE
6926 {
6927 .procname = "mc_forwarding",
6928 .data = &ipv6_devconf.mc_forwarding,
6929 .maxlen = sizeof(int),
6930 .mode = 0444,
6931 .proc_handler = proc_dointvec,
6932 },
6933 #endif
6934 {
6935 .procname = "disable_ipv6",
6936 .data = &ipv6_devconf.disable_ipv6,
6937 .maxlen = sizeof(int),
6938 .mode = 0644,
6939 .proc_handler = addrconf_sysctl_disable,
6940 },
6941 {
6942 .procname = "accept_dad",
6943 .data = &ipv6_devconf.accept_dad,
6944 .maxlen = sizeof(int),
6945 .mode = 0644,
6946 .proc_handler = proc_dointvec,
6947 },
6948 {
6949 .procname = "force_tllao",
6950 .data = &ipv6_devconf.force_tllao,
6951 .maxlen = sizeof(int),
6952 .mode = 0644,
6953 .proc_handler = proc_dointvec
6954 },
6955 {
6956 .procname = "ndisc_notify",
6957 .data = &ipv6_devconf.ndisc_notify,
6958 .maxlen = sizeof(int),
6959 .mode = 0644,
6960 .proc_handler = proc_dointvec
6961 },
6962 {
6963 .procname = "suppress_frag_ndisc",
6964 .data = &ipv6_devconf.suppress_frag_ndisc,
6965 .maxlen = sizeof(int),
6966 .mode = 0644,
6967 .proc_handler = proc_dointvec
6968 },
6969 {
6970 .procname = "accept_ra_from_local",
6971 .data = &ipv6_devconf.accept_ra_from_local,
6972 .maxlen = sizeof(int),
6973 .mode = 0644,
6974 .proc_handler = proc_dointvec,
6975 },
6976 {
6977 .procname = "accept_ra_mtu",
6978 .data = &ipv6_devconf.accept_ra_mtu,
6979 .maxlen = sizeof(int),
6980 .mode = 0644,
6981 .proc_handler = proc_dointvec,
6982 },
6983 {
6984 .procname = "stable_secret",
6985 .data = &ipv6_devconf.stable_secret,
6986 .maxlen = IPV6_MAX_STRLEN,
6987 .mode = 0600,
6988 .proc_handler = addrconf_sysctl_stable_secret,
6989 },
6990 {
6991 .procname = "use_oif_addrs_only",
6992 .data = &ipv6_devconf.use_oif_addrs_only,
6993 .maxlen = sizeof(int),
6994 .mode = 0644,
6995 .proc_handler = proc_dointvec,
6996 },
6997 {
6998 .procname = "ignore_routes_with_linkdown",
6999 .data = &ipv6_devconf.ignore_routes_with_linkdown,
7000 .maxlen = sizeof(int),
7001 .mode = 0644,
7002 .proc_handler = addrconf_sysctl_ignore_routes_with_linkdown,
7003 },
7004 {
7005 .procname = "drop_unicast_in_l2_multicast",
7006 .data = &ipv6_devconf.drop_unicast_in_l2_multicast,
7007 .maxlen = sizeof(int),
7008 .mode = 0644,
7009 .proc_handler = proc_dointvec,
7010 },
7011 {
7012 .procname = "drop_unsolicited_na",
7013 .data = &ipv6_devconf.drop_unsolicited_na,
7014 .maxlen = sizeof(int),
7015 .mode = 0644,
7016 .proc_handler = proc_dointvec,
7017 },
7018 {
7019 .procname = "keep_addr_on_down",
7020 .data = &ipv6_devconf.keep_addr_on_down,
7021 .maxlen = sizeof(int),
7022 .mode = 0644,
7023 .proc_handler = proc_dointvec,
7024
7025 },
7026 {
7027 .procname = "seg6_enabled",
7028 .data = &ipv6_devconf.seg6_enabled,
7029 .maxlen = sizeof(int),
7030 .mode = 0644,
7031 .proc_handler = proc_dointvec,
7032 },
7033 #ifdef CONFIG_IPV6_SEG6_HMAC
7034 {
7035 .procname = "seg6_require_hmac",
7036 .data = &ipv6_devconf.seg6_require_hmac,
7037 .maxlen = sizeof(int),
7038 .mode = 0644,
7039 .proc_handler = proc_dointvec,
7040 },
7041 #endif
7042 {
7043 .procname = "enhanced_dad",
7044 .data = &ipv6_devconf.enhanced_dad,
7045 .maxlen = sizeof(int),
7046 .mode = 0644,
7047 .proc_handler = proc_dointvec,
7048 },
7049 {
7050 .procname = "addr_gen_mode",
7051 .data = &ipv6_devconf.addr_gen_mode,
7052 .maxlen = sizeof(int),
7053 .mode = 0644,
7054 .proc_handler = addrconf_sysctl_addr_gen_mode,
7055 },
7056 {
7057 .procname = "disable_policy",
7058 .data = &ipv6_devconf.disable_policy,
7059 .maxlen = sizeof(int),
7060 .mode = 0644,
7061 .proc_handler = addrconf_sysctl_disable_policy,
7062 },
7063 {
7064 .procname = "ndisc_tclass",
7065 .data = &ipv6_devconf.ndisc_tclass,
7066 .maxlen = sizeof(int),
7067 .mode = 0644,
7068 .proc_handler = proc_dointvec_minmax,
7069 .extra1 = (void *)SYSCTL_ZERO,
7070 .extra2 = (void *)&two_five_five,
7071 },
7072 {
7073 .procname = "rpl_seg_enabled",
7074 .data = &ipv6_devconf.rpl_seg_enabled,
7075 .maxlen = sizeof(int),
7076 .mode = 0644,
7077 .proc_handler = proc_dointvec,
7078 },
7079 {
7080 .procname = "ioam6_enabled",
7081 .data = &ipv6_devconf.ioam6_enabled,
7082 .maxlen = sizeof(u8),
7083 .mode = 0644,
7084 .proc_handler = proc_dou8vec_minmax,
7085 .extra1 = (void *)SYSCTL_ZERO,
7086 .extra2 = (void *)SYSCTL_ONE,
7087 },
7088 {
7089 .procname = "ioam6_id",
7090 .data = &ipv6_devconf.ioam6_id,
7091 .maxlen = sizeof(u32),
7092 .mode = 0644,
7093 .proc_handler = proc_douintvec_minmax,
7094 .extra1 = (void *)SYSCTL_ZERO,
7095 .extra2 = (void *)&ioam6_if_id_max,
7096 },
7097 {
7098 .procname = "ioam6_id_wide",
7099 .data = &ipv6_devconf.ioam6_id_wide,
7100 .maxlen = sizeof(u32),
7101 .mode = 0644,
7102 .proc_handler = proc_douintvec,
7103 },
7104 {
7105 .procname = "ndisc_evict_nocarrier",
7106 .data = &ipv6_devconf.ndisc_evict_nocarrier,
7107 .maxlen = sizeof(u8),
7108 .mode = 0644,
7109 .proc_handler = proc_dou8vec_minmax,
7110 .extra1 = (void *)SYSCTL_ZERO,
7111 .extra2 = (void *)SYSCTL_ONE,
7112 },
7113 {
7114 .procname = "accept_untracked_na",
7115 .data = &ipv6_devconf.accept_untracked_na,
7116 .maxlen = sizeof(int),
7117 .mode = 0644,
7118 .proc_handler = proc_dointvec_minmax,
7119 .extra1 = SYSCTL_ZERO,
7120 .extra2 = SYSCTL_TWO,
7121 },
7122 {
7123 /* sentinel */
7124 }
7125 };
7126
__addrconf_sysctl_register(struct net * net,char * dev_name,struct inet6_dev * idev,struct ipv6_devconf * p)7127 static int __addrconf_sysctl_register(struct net *net, char *dev_name,
7128 struct inet6_dev *idev, struct ipv6_devconf *p)
7129 {
7130 int i, ifindex;
7131 struct ctl_table *table;
7132 char path[sizeof("net/ipv6/conf/") + IFNAMSIZ];
7133
7134 table = kmemdup(addrconf_sysctl, sizeof(addrconf_sysctl), GFP_KERNEL_ACCOUNT);
7135 if (!table)
7136 goto out;
7137
7138 for (i = 0; table[i].data; i++) {
7139 table[i].data += (char *)p - (char *)&ipv6_devconf;
7140 /* If one of these is already set, then it is not safe to
7141 * overwrite either of them: this makes proc_dointvec_minmax
7142 * usable.
7143 */
7144 if (!table[i].extra1 && !table[i].extra2) {
7145 table[i].extra1 = idev; /* embedded; no ref */
7146 table[i].extra2 = net;
7147 }
7148 }
7149
7150 snprintf(path, sizeof(path), "net/ipv6/conf/%s", dev_name);
7151
7152 p->sysctl_header = register_net_sysctl_sz(net, path, table,
7153 ARRAY_SIZE(addrconf_sysctl));
7154 if (!p->sysctl_header)
7155 goto free;
7156
7157 if (!strcmp(dev_name, "all"))
7158 ifindex = NETCONFA_IFINDEX_ALL;
7159 else if (!strcmp(dev_name, "default"))
7160 ifindex = NETCONFA_IFINDEX_DEFAULT;
7161 else
7162 ifindex = idev->dev->ifindex;
7163 inet6_netconf_notify_devconf(net, RTM_NEWNETCONF, NETCONFA_ALL,
7164 ifindex, p);
7165 return 0;
7166
7167 free:
7168 kfree(table);
7169 out:
7170 return -ENOBUFS;
7171 }
7172
__addrconf_sysctl_unregister(struct net * net,struct ipv6_devconf * p,int ifindex)7173 static void __addrconf_sysctl_unregister(struct net *net,
7174 struct ipv6_devconf *p, int ifindex)
7175 {
7176 struct ctl_table *table;
7177
7178 if (!p->sysctl_header)
7179 return;
7180
7181 table = p->sysctl_header->ctl_table_arg;
7182 unregister_net_sysctl_table(p->sysctl_header);
7183 p->sysctl_header = NULL;
7184 kfree(table);
7185
7186 inet6_netconf_notify_devconf(net, RTM_DELNETCONF, 0, ifindex, NULL);
7187 }
7188
addrconf_sysctl_register(struct inet6_dev * idev)7189 static int addrconf_sysctl_register(struct inet6_dev *idev)
7190 {
7191 int err;
7192
7193 if (!sysctl_dev_name_is_allowed(idev->dev->name))
7194 return -EINVAL;
7195
7196 err = neigh_sysctl_register(idev->dev, idev->nd_parms,
7197 &ndisc_ifinfo_sysctl_change);
7198 if (err)
7199 return err;
7200 err = __addrconf_sysctl_register(dev_net(idev->dev), idev->dev->name,
7201 idev, &idev->cnf);
7202 if (err)
7203 neigh_sysctl_unregister(idev->nd_parms);
7204
7205 return err;
7206 }
7207
addrconf_sysctl_unregister(struct inet6_dev * idev)7208 static void addrconf_sysctl_unregister(struct inet6_dev *idev)
7209 {
7210 __addrconf_sysctl_unregister(dev_net(idev->dev), &idev->cnf,
7211 idev->dev->ifindex);
7212 neigh_sysctl_unregister(idev->nd_parms);
7213 }
7214
7215
7216 #endif
7217
addrconf_init_net(struct net * net)7218 static int __net_init addrconf_init_net(struct net *net)
7219 {
7220 int err = -ENOMEM;
7221 struct ipv6_devconf *all, *dflt;
7222
7223 spin_lock_init(&net->ipv6.addrconf_hash_lock);
7224 INIT_DEFERRABLE_WORK(&net->ipv6.addr_chk_work, addrconf_verify_work);
7225 net->ipv6.inet6_addr_lst = kcalloc(IN6_ADDR_HSIZE,
7226 sizeof(struct hlist_head),
7227 GFP_KERNEL);
7228 if (!net->ipv6.inet6_addr_lst)
7229 goto err_alloc_addr;
7230
7231 all = kmemdup(&ipv6_devconf, sizeof(ipv6_devconf), GFP_KERNEL);
7232 if (!all)
7233 goto err_alloc_all;
7234
7235 dflt = kmemdup(&ipv6_devconf_dflt, sizeof(ipv6_devconf_dflt), GFP_KERNEL);
7236 if (!dflt)
7237 goto err_alloc_dflt;
7238
7239 if (!net_eq(net, &init_net)) {
7240 switch (net_inherit_devconf()) {
7241 case 1: /* copy from init_net */
7242 memcpy(all, init_net.ipv6.devconf_all,
7243 sizeof(ipv6_devconf));
7244 memcpy(dflt, init_net.ipv6.devconf_dflt,
7245 sizeof(ipv6_devconf_dflt));
7246 break;
7247 case 3: /* copy from the current netns */
7248 memcpy(all, current->nsproxy->net_ns->ipv6.devconf_all,
7249 sizeof(ipv6_devconf));
7250 memcpy(dflt,
7251 current->nsproxy->net_ns->ipv6.devconf_dflt,
7252 sizeof(ipv6_devconf_dflt));
7253 break;
7254 case 0:
7255 case 2:
7256 /* use compiled values */
7257 break;
7258 }
7259 }
7260
7261 /* these will be inherited by all namespaces */
7262 dflt->autoconf = ipv6_defaults.autoconf;
7263 dflt->disable_ipv6 = ipv6_defaults.disable_ipv6;
7264
7265 dflt->stable_secret.initialized = false;
7266 all->stable_secret.initialized = false;
7267
7268 net->ipv6.devconf_all = all;
7269 net->ipv6.devconf_dflt = dflt;
7270
7271 #ifdef CONFIG_SYSCTL
7272 err = __addrconf_sysctl_register(net, "all", NULL, all);
7273 if (err < 0)
7274 goto err_reg_all;
7275
7276 err = __addrconf_sysctl_register(net, "default", NULL, dflt);
7277 if (err < 0)
7278 goto err_reg_dflt;
7279 #endif
7280 return 0;
7281
7282 #ifdef CONFIG_SYSCTL
7283 err_reg_dflt:
7284 __addrconf_sysctl_unregister(net, all, NETCONFA_IFINDEX_ALL);
7285 err_reg_all:
7286 kfree(dflt);
7287 net->ipv6.devconf_dflt = NULL;
7288 #endif
7289 err_alloc_dflt:
7290 kfree(all);
7291 net->ipv6.devconf_all = NULL;
7292 err_alloc_all:
7293 kfree(net->ipv6.inet6_addr_lst);
7294 err_alloc_addr:
7295 return err;
7296 }
7297
addrconf_exit_net(struct net * net)7298 static void __net_exit addrconf_exit_net(struct net *net)
7299 {
7300 int i;
7301
7302 #ifdef CONFIG_SYSCTL
7303 __addrconf_sysctl_unregister(net, net->ipv6.devconf_dflt,
7304 NETCONFA_IFINDEX_DEFAULT);
7305 __addrconf_sysctl_unregister(net, net->ipv6.devconf_all,
7306 NETCONFA_IFINDEX_ALL);
7307 #endif
7308 kfree(net->ipv6.devconf_dflt);
7309 net->ipv6.devconf_dflt = NULL;
7310 kfree(net->ipv6.devconf_all);
7311 net->ipv6.devconf_all = NULL;
7312
7313 cancel_delayed_work_sync(&net->ipv6.addr_chk_work);
7314 /*
7315 * Check hash table, then free it.
7316 */
7317 for (i = 0; i < IN6_ADDR_HSIZE; i++)
7318 WARN_ON_ONCE(!hlist_empty(&net->ipv6.inet6_addr_lst[i]));
7319
7320 kfree(net->ipv6.inet6_addr_lst);
7321 net->ipv6.inet6_addr_lst = NULL;
7322 }
7323
7324 static struct pernet_operations addrconf_ops = {
7325 .init = addrconf_init_net,
7326 .exit = addrconf_exit_net,
7327 };
7328
7329 static struct rtnl_af_ops inet6_ops __read_mostly = {
7330 .family = AF_INET6,
7331 .fill_link_af = inet6_fill_link_af,
7332 .get_link_af_size = inet6_get_link_af_size,
7333 .validate_link_af = inet6_validate_link_af,
7334 .set_link_af = inet6_set_link_af,
7335 };
7336
7337 /*
7338 * Init / cleanup code
7339 */
7340
addrconf_init(void)7341 int __init addrconf_init(void)
7342 {
7343 struct inet6_dev *idev;
7344 int err;
7345
7346 err = ipv6_addr_label_init();
7347 if (err < 0) {
7348 pr_crit("%s: cannot initialize default policy table: %d\n",
7349 __func__, err);
7350 goto out;
7351 }
7352
7353 err = register_pernet_subsys(&addrconf_ops);
7354 if (err < 0)
7355 goto out_addrlabel;
7356
7357 addrconf_wq = create_workqueue("ipv6_addrconf");
7358 if (!addrconf_wq) {
7359 err = -ENOMEM;
7360 goto out_nowq;
7361 }
7362
7363 rtnl_lock();
7364 idev = ipv6_add_dev(blackhole_netdev);
7365 rtnl_unlock();
7366 if (IS_ERR(idev)) {
7367 err = PTR_ERR(idev);
7368 goto errlo;
7369 }
7370
7371 ip6_route_init_special_entries();
7372
7373 register_netdevice_notifier(&ipv6_dev_notf);
7374
7375 addrconf_verify(&init_net);
7376
7377 rtnl_af_register(&inet6_ops);
7378
7379 err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETLINK,
7380 NULL, inet6_dump_ifinfo, 0);
7381 if (err < 0)
7382 goto errout;
7383
7384 err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_NEWADDR,
7385 inet6_rtm_newaddr, NULL, 0);
7386 if (err < 0)
7387 goto errout;
7388 err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_DELADDR,
7389 inet6_rtm_deladdr, NULL, 0);
7390 if (err < 0)
7391 goto errout;
7392 err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETADDR,
7393 inet6_rtm_getaddr, inet6_dump_ifaddr,
7394 RTNL_FLAG_DOIT_UNLOCKED);
7395 if (err < 0)
7396 goto errout;
7397 err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETMULTICAST,
7398 NULL, inet6_dump_ifmcaddr, 0);
7399 if (err < 0)
7400 goto errout;
7401 err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETANYCAST,
7402 NULL, inet6_dump_ifacaddr, 0);
7403 if (err < 0)
7404 goto errout;
7405 err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETNETCONF,
7406 inet6_netconf_get_devconf,
7407 inet6_netconf_dump_devconf,
7408 RTNL_FLAG_DOIT_UNLOCKED);
7409 if (err < 0)
7410 goto errout;
7411 err = ipv6_addr_label_rtnl_register();
7412 if (err < 0)
7413 goto errout;
7414
7415 return 0;
7416 errout:
7417 rtnl_unregister_all(PF_INET6);
7418 rtnl_af_unregister(&inet6_ops);
7419 unregister_netdevice_notifier(&ipv6_dev_notf);
7420 errlo:
7421 destroy_workqueue(addrconf_wq);
7422 out_nowq:
7423 unregister_pernet_subsys(&addrconf_ops);
7424 out_addrlabel:
7425 ipv6_addr_label_cleanup();
7426 out:
7427 return err;
7428 }
7429
addrconf_cleanup(void)7430 void addrconf_cleanup(void)
7431 {
7432 struct net_device *dev;
7433
7434 unregister_netdevice_notifier(&ipv6_dev_notf);
7435 unregister_pernet_subsys(&addrconf_ops);
7436 ipv6_addr_label_cleanup();
7437
7438 rtnl_af_unregister(&inet6_ops);
7439
7440 rtnl_lock();
7441
7442 /* clean dev list */
7443 for_each_netdev(&init_net, dev) {
7444 if (__in6_dev_get(dev) == NULL)
7445 continue;
7446 addrconf_ifdown(dev, true);
7447 }
7448 addrconf_ifdown(init_net.loopback_dev, true);
7449
7450 rtnl_unlock();
7451
7452 destroy_workqueue(addrconf_wq);
7453 }
7454