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 scores[hiscore_idx].scopedist >= 0)
1844 goto out;
1845 }
1846
1847 for_each_netdev_rcu(net, dev) {
1848 /* only consider addresses on devices in the
1849 * same L3 domain
1850 */
1851 if (l3mdev_master_ifindex_rcu(dev) != master_idx)
1852 continue;
1853 idev = __in6_dev_get(dev);
1854 if (!idev)
1855 continue;
1856 hiscore_idx = __ipv6_dev_get_saddr(net, &dst, idev, scores, hiscore_idx);
1857 }
1858 }
1859
1860 out:
1861 hiscore = &scores[hiscore_idx];
1862 if (!hiscore->ifa)
1863 ret = -EADDRNOTAVAIL;
1864 else
1865 *saddr = hiscore->ifa->addr;
1866
1867 rcu_read_unlock();
1868 return ret;
1869 }
1870 EXPORT_SYMBOL(ipv6_dev_get_saddr);
1871
__ipv6_get_lladdr(struct inet6_dev * idev,struct in6_addr * addr,u32 banned_flags)1872 static int __ipv6_get_lladdr(struct inet6_dev *idev, struct in6_addr *addr,
1873 u32 banned_flags)
1874 {
1875 struct inet6_ifaddr *ifp;
1876 int err = -EADDRNOTAVAIL;
1877
1878 list_for_each_entry_reverse(ifp, &idev->addr_list, if_list) {
1879 if (ifp->scope > IFA_LINK)
1880 break;
1881 if (ifp->scope == IFA_LINK &&
1882 !(ifp->flags & banned_flags)) {
1883 *addr = ifp->addr;
1884 err = 0;
1885 break;
1886 }
1887 }
1888 return err;
1889 }
1890
ipv6_get_lladdr(struct net_device * dev,struct in6_addr * addr,u32 banned_flags)1891 int ipv6_get_lladdr(struct net_device *dev, struct in6_addr *addr,
1892 u32 banned_flags)
1893 {
1894 struct inet6_dev *idev;
1895 int err = -EADDRNOTAVAIL;
1896
1897 rcu_read_lock();
1898 idev = __in6_dev_get(dev);
1899 if (idev) {
1900 read_lock_bh(&idev->lock);
1901 err = __ipv6_get_lladdr(idev, addr, banned_flags);
1902 read_unlock_bh(&idev->lock);
1903 }
1904 rcu_read_unlock();
1905 return err;
1906 }
1907
ipv6_count_addresses(const struct inet6_dev * idev)1908 static int ipv6_count_addresses(const struct inet6_dev *idev)
1909 {
1910 const struct inet6_ifaddr *ifp;
1911 int cnt = 0;
1912
1913 rcu_read_lock();
1914 list_for_each_entry_rcu(ifp, &idev->addr_list, if_list)
1915 cnt++;
1916 rcu_read_unlock();
1917 return cnt;
1918 }
1919
ipv6_chk_addr(struct net * net,const struct in6_addr * addr,const struct net_device * dev,int strict)1920 int ipv6_chk_addr(struct net *net, const struct in6_addr *addr,
1921 const struct net_device *dev, int strict)
1922 {
1923 return ipv6_chk_addr_and_flags(net, addr, dev, !dev,
1924 strict, IFA_F_TENTATIVE);
1925 }
1926 EXPORT_SYMBOL(ipv6_chk_addr);
1927
1928 /* device argument is used to find the L3 domain of interest. If
1929 * skip_dev_check is set, then the ifp device is not checked against
1930 * the passed in dev argument. So the 2 cases for addresses checks are:
1931 * 1. does the address exist in the L3 domain that dev is part of
1932 * (skip_dev_check = true), or
1933 *
1934 * 2. does the address exist on the specific device
1935 * (skip_dev_check = false)
1936 */
1937 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)1938 __ipv6_chk_addr_and_flags(struct net *net, const struct in6_addr *addr,
1939 const struct net_device *dev, bool skip_dev_check,
1940 int strict, u32 banned_flags)
1941 {
1942 unsigned int hash = inet6_addr_hash(net, addr);
1943 struct net_device *l3mdev, *ndev;
1944 struct inet6_ifaddr *ifp;
1945 u32 ifp_flags;
1946
1947 rcu_read_lock();
1948
1949 l3mdev = l3mdev_master_dev_rcu(dev);
1950 if (skip_dev_check)
1951 dev = NULL;
1952
1953 hlist_for_each_entry_rcu(ifp, &net->ipv6.inet6_addr_lst[hash], addr_lst) {
1954 ndev = ifp->idev->dev;
1955
1956 if (l3mdev_master_dev_rcu(ndev) != l3mdev)
1957 continue;
1958
1959 /* Decouple optimistic from tentative for evaluation here.
1960 * Ban optimistic addresses explicitly, when required.
1961 */
1962 ifp_flags = (ifp->flags&IFA_F_OPTIMISTIC)
1963 ? (ifp->flags&~IFA_F_TENTATIVE)
1964 : ifp->flags;
1965 if (ipv6_addr_equal(&ifp->addr, addr) &&
1966 !(ifp_flags&banned_flags) &&
1967 (!dev || ndev == dev ||
1968 !(ifp->scope&(IFA_LINK|IFA_HOST) || strict))) {
1969 rcu_read_unlock();
1970 return ndev;
1971 }
1972 }
1973
1974 rcu_read_unlock();
1975 return NULL;
1976 }
1977
ipv6_chk_addr_and_flags(struct net * net,const struct in6_addr * addr,const struct net_device * dev,bool skip_dev_check,int strict,u32 banned_flags)1978 int ipv6_chk_addr_and_flags(struct net *net, const struct in6_addr *addr,
1979 const struct net_device *dev, bool skip_dev_check,
1980 int strict, u32 banned_flags)
1981 {
1982 return __ipv6_chk_addr_and_flags(net, addr, dev, skip_dev_check,
1983 strict, banned_flags) ? 1 : 0;
1984 }
1985 EXPORT_SYMBOL(ipv6_chk_addr_and_flags);
1986
1987
1988 /* Compares an address/prefix_len with addresses on device @dev.
1989 * If one is found it returns true.
1990 */
ipv6_chk_custom_prefix(const struct in6_addr * addr,const unsigned int prefix_len,struct net_device * dev)1991 bool ipv6_chk_custom_prefix(const struct in6_addr *addr,
1992 const unsigned int prefix_len, struct net_device *dev)
1993 {
1994 const struct inet6_ifaddr *ifa;
1995 const struct inet6_dev *idev;
1996 bool ret = false;
1997
1998 rcu_read_lock();
1999 idev = __in6_dev_get(dev);
2000 if (idev) {
2001 list_for_each_entry_rcu(ifa, &idev->addr_list, if_list) {
2002 ret = ipv6_prefix_equal(addr, &ifa->addr, prefix_len);
2003 if (ret)
2004 break;
2005 }
2006 }
2007 rcu_read_unlock();
2008
2009 return ret;
2010 }
2011 EXPORT_SYMBOL(ipv6_chk_custom_prefix);
2012
ipv6_chk_prefix(const struct in6_addr * addr,struct net_device * dev)2013 int ipv6_chk_prefix(const struct in6_addr *addr, struct net_device *dev)
2014 {
2015 const struct inet6_ifaddr *ifa;
2016 const struct inet6_dev *idev;
2017 int onlink;
2018
2019 onlink = 0;
2020 rcu_read_lock();
2021 idev = __in6_dev_get(dev);
2022 if (idev) {
2023 list_for_each_entry_rcu(ifa, &idev->addr_list, if_list) {
2024 onlink = ipv6_prefix_equal(addr, &ifa->addr,
2025 ifa->prefix_len);
2026 if (onlink)
2027 break;
2028 }
2029 }
2030 rcu_read_unlock();
2031 return onlink;
2032 }
2033 EXPORT_SYMBOL(ipv6_chk_prefix);
2034
2035 /**
2036 * ipv6_dev_find - find the first device with a given source address.
2037 * @net: the net namespace
2038 * @addr: the source address
2039 * @dev: used to find the L3 domain of interest
2040 *
2041 * The caller should be protected by RCU, or RTNL.
2042 */
ipv6_dev_find(struct net * net,const struct in6_addr * addr,struct net_device * dev)2043 struct net_device *ipv6_dev_find(struct net *net, const struct in6_addr *addr,
2044 struct net_device *dev)
2045 {
2046 return __ipv6_chk_addr_and_flags(net, addr, dev, !dev, 1,
2047 IFA_F_TENTATIVE);
2048 }
2049 EXPORT_SYMBOL(ipv6_dev_find);
2050
ipv6_get_ifaddr(struct net * net,const struct in6_addr * addr,struct net_device * dev,int strict)2051 struct inet6_ifaddr *ipv6_get_ifaddr(struct net *net, const struct in6_addr *addr,
2052 struct net_device *dev, int strict)
2053 {
2054 unsigned int hash = inet6_addr_hash(net, addr);
2055 struct inet6_ifaddr *ifp, *result = NULL;
2056
2057 rcu_read_lock();
2058 hlist_for_each_entry_rcu(ifp, &net->ipv6.inet6_addr_lst[hash], addr_lst) {
2059 if (ipv6_addr_equal(&ifp->addr, addr)) {
2060 if (!dev || ifp->idev->dev == dev ||
2061 !(ifp->scope&(IFA_LINK|IFA_HOST) || strict)) {
2062 if (in6_ifa_hold_safe(ifp)) {
2063 result = ifp;
2064 break;
2065 }
2066 }
2067 }
2068 }
2069 rcu_read_unlock();
2070
2071 return result;
2072 }
2073
2074 /* Gets referenced address, destroys ifaddr */
2075
addrconf_dad_stop(struct inet6_ifaddr * ifp,int dad_failed)2076 static void addrconf_dad_stop(struct inet6_ifaddr *ifp, int dad_failed)
2077 {
2078 if (dad_failed)
2079 ifp->flags |= IFA_F_DADFAILED;
2080
2081 if (ifp->flags&IFA_F_TEMPORARY) {
2082 struct inet6_ifaddr *ifpub;
2083 spin_lock_bh(&ifp->lock);
2084 ifpub = ifp->ifpub;
2085 if (ifpub) {
2086 in6_ifa_hold(ifpub);
2087 spin_unlock_bh(&ifp->lock);
2088 ipv6_create_tempaddr(ifpub, true);
2089 in6_ifa_put(ifpub);
2090 } else {
2091 spin_unlock_bh(&ifp->lock);
2092 }
2093 ipv6_del_addr(ifp);
2094 } else if (ifp->flags&IFA_F_PERMANENT || !dad_failed) {
2095 spin_lock_bh(&ifp->lock);
2096 addrconf_del_dad_work(ifp);
2097 ifp->flags |= IFA_F_TENTATIVE;
2098 if (dad_failed)
2099 ifp->flags &= ~IFA_F_OPTIMISTIC;
2100 spin_unlock_bh(&ifp->lock);
2101 if (dad_failed)
2102 ipv6_ifa_notify(0, ifp);
2103 in6_ifa_put(ifp);
2104 } else {
2105 ipv6_del_addr(ifp);
2106 }
2107 }
2108
addrconf_dad_end(struct inet6_ifaddr * ifp)2109 static int addrconf_dad_end(struct inet6_ifaddr *ifp)
2110 {
2111 int err = -ENOENT;
2112
2113 spin_lock_bh(&ifp->lock);
2114 if (ifp->state == INET6_IFADDR_STATE_DAD) {
2115 ifp->state = INET6_IFADDR_STATE_POSTDAD;
2116 err = 0;
2117 }
2118 spin_unlock_bh(&ifp->lock);
2119
2120 return err;
2121 }
2122
addrconf_dad_failure(struct sk_buff * skb,struct inet6_ifaddr * ifp)2123 void addrconf_dad_failure(struct sk_buff *skb, struct inet6_ifaddr *ifp)
2124 {
2125 struct inet6_dev *idev = ifp->idev;
2126 struct net *net = dev_net(idev->dev);
2127
2128 if (addrconf_dad_end(ifp)) {
2129 in6_ifa_put(ifp);
2130 return;
2131 }
2132
2133 net_info_ratelimited("%s: IPv6 duplicate address %pI6c used by %pM detected!\n",
2134 ifp->idev->dev->name, &ifp->addr, eth_hdr(skb)->h_source);
2135
2136 spin_lock_bh(&ifp->lock);
2137
2138 if (ifp->flags & IFA_F_STABLE_PRIVACY) {
2139 struct in6_addr new_addr;
2140 struct inet6_ifaddr *ifp2;
2141 int retries = ifp->stable_privacy_retry + 1;
2142 struct ifa6_config cfg = {
2143 .pfx = &new_addr,
2144 .plen = ifp->prefix_len,
2145 .ifa_flags = ifp->flags,
2146 .valid_lft = ifp->valid_lft,
2147 .preferred_lft = ifp->prefered_lft,
2148 .scope = ifp->scope,
2149 };
2150
2151 if (retries > net->ipv6.sysctl.idgen_retries) {
2152 net_info_ratelimited("%s: privacy stable address generation failed because of DAD conflicts!\n",
2153 ifp->idev->dev->name);
2154 goto errdad;
2155 }
2156
2157 new_addr = ifp->addr;
2158 if (ipv6_generate_stable_address(&new_addr, retries,
2159 idev))
2160 goto errdad;
2161
2162 spin_unlock_bh(&ifp->lock);
2163
2164 if (idev->cnf.max_addresses &&
2165 ipv6_count_addresses(idev) >=
2166 idev->cnf.max_addresses)
2167 goto lock_errdad;
2168
2169 net_info_ratelimited("%s: generating new stable privacy address because of DAD conflict\n",
2170 ifp->idev->dev->name);
2171
2172 ifp2 = ipv6_add_addr(idev, &cfg, false, NULL);
2173 if (IS_ERR(ifp2))
2174 goto lock_errdad;
2175
2176 spin_lock_bh(&ifp2->lock);
2177 ifp2->stable_privacy_retry = retries;
2178 ifp2->state = INET6_IFADDR_STATE_PREDAD;
2179 spin_unlock_bh(&ifp2->lock);
2180
2181 addrconf_mod_dad_work(ifp2, net->ipv6.sysctl.idgen_delay);
2182 in6_ifa_put(ifp2);
2183 lock_errdad:
2184 spin_lock_bh(&ifp->lock);
2185 }
2186
2187 errdad:
2188 /* transition from _POSTDAD to _ERRDAD */
2189 ifp->state = INET6_IFADDR_STATE_ERRDAD;
2190 spin_unlock_bh(&ifp->lock);
2191
2192 addrconf_mod_dad_work(ifp, 0);
2193 in6_ifa_put(ifp);
2194 }
2195
2196 /* Join to solicited addr multicast group.
2197 * caller must hold RTNL */
addrconf_join_solict(struct net_device * dev,const struct in6_addr * addr)2198 void addrconf_join_solict(struct net_device *dev, const struct in6_addr *addr)
2199 {
2200 struct in6_addr maddr;
2201
2202 if (dev->flags&(IFF_LOOPBACK|IFF_NOARP))
2203 return;
2204
2205 addrconf_addr_solict_mult(addr, &maddr);
2206 ipv6_dev_mc_inc(dev, &maddr);
2207 }
2208
2209 /* caller must hold RTNL */
addrconf_leave_solict(struct inet6_dev * idev,const struct in6_addr * addr)2210 void addrconf_leave_solict(struct inet6_dev *idev, const struct in6_addr *addr)
2211 {
2212 struct in6_addr maddr;
2213
2214 if (idev->dev->flags&(IFF_LOOPBACK|IFF_NOARP))
2215 return;
2216
2217 addrconf_addr_solict_mult(addr, &maddr);
2218 __ipv6_dev_mc_dec(idev, &maddr);
2219 }
2220
2221 /* caller must hold RTNL */
addrconf_join_anycast(struct inet6_ifaddr * ifp)2222 static void addrconf_join_anycast(struct inet6_ifaddr *ifp)
2223 {
2224 struct in6_addr addr;
2225
2226 if (ifp->prefix_len >= 127) /* RFC 6164 */
2227 return;
2228 ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
2229 if (ipv6_addr_any(&addr))
2230 return;
2231 __ipv6_dev_ac_inc(ifp->idev, &addr);
2232 }
2233
2234 /* caller must hold RTNL */
addrconf_leave_anycast(struct inet6_ifaddr * ifp)2235 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp)
2236 {
2237 struct in6_addr addr;
2238
2239 if (ifp->prefix_len >= 127) /* RFC 6164 */
2240 return;
2241 ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
2242 if (ipv6_addr_any(&addr))
2243 return;
2244 __ipv6_dev_ac_dec(ifp->idev, &addr);
2245 }
2246
addrconf_ifid_6lowpan(u8 * eui,struct net_device * dev)2247 static int addrconf_ifid_6lowpan(u8 *eui, struct net_device *dev)
2248 {
2249 switch (dev->addr_len) {
2250 case ETH_ALEN:
2251 memcpy(eui, dev->dev_addr, 3);
2252 eui[3] = 0xFF;
2253 eui[4] = 0xFE;
2254 memcpy(eui + 5, dev->dev_addr + 3, 3);
2255 break;
2256 case EUI64_ADDR_LEN:
2257 memcpy(eui, dev->dev_addr, EUI64_ADDR_LEN);
2258 eui[0] ^= 2;
2259 break;
2260 default:
2261 return -1;
2262 }
2263
2264 return 0;
2265 }
2266
addrconf_ifid_ieee1394(u8 * eui,struct net_device * dev)2267 static int addrconf_ifid_ieee1394(u8 *eui, struct net_device *dev)
2268 {
2269 const union fwnet_hwaddr *ha;
2270
2271 if (dev->addr_len != FWNET_ALEN)
2272 return -1;
2273
2274 ha = (const union fwnet_hwaddr *)dev->dev_addr;
2275
2276 memcpy(eui, &ha->uc.uniq_id, sizeof(ha->uc.uniq_id));
2277 eui[0] ^= 2;
2278 return 0;
2279 }
2280
addrconf_ifid_arcnet(u8 * eui,struct net_device * dev)2281 static int addrconf_ifid_arcnet(u8 *eui, struct net_device *dev)
2282 {
2283 /* XXX: inherit EUI-64 from other interface -- yoshfuji */
2284 if (dev->addr_len != ARCNET_ALEN)
2285 return -1;
2286 memset(eui, 0, 7);
2287 eui[7] = *(u8 *)dev->dev_addr;
2288 return 0;
2289 }
2290
addrconf_ifid_infiniband(u8 * eui,struct net_device * dev)2291 static int addrconf_ifid_infiniband(u8 *eui, struct net_device *dev)
2292 {
2293 if (dev->addr_len != INFINIBAND_ALEN)
2294 return -1;
2295 memcpy(eui, dev->dev_addr + 12, 8);
2296 eui[0] |= 2;
2297 return 0;
2298 }
2299
__ipv6_isatap_ifid(u8 * eui,__be32 addr)2300 static int __ipv6_isatap_ifid(u8 *eui, __be32 addr)
2301 {
2302 if (addr == 0)
2303 return -1;
2304 eui[0] = (ipv4_is_zeronet(addr) || ipv4_is_private_10(addr) ||
2305 ipv4_is_loopback(addr) || ipv4_is_linklocal_169(addr) ||
2306 ipv4_is_private_172(addr) || ipv4_is_test_192(addr) ||
2307 ipv4_is_anycast_6to4(addr) || ipv4_is_private_192(addr) ||
2308 ipv4_is_test_198(addr) || ipv4_is_multicast(addr) ||
2309 ipv4_is_lbcast(addr)) ? 0x00 : 0x02;
2310 eui[1] = 0;
2311 eui[2] = 0x5E;
2312 eui[3] = 0xFE;
2313 memcpy(eui + 4, &addr, 4);
2314 return 0;
2315 }
2316
addrconf_ifid_sit(u8 * eui,struct net_device * dev)2317 static int addrconf_ifid_sit(u8 *eui, struct net_device *dev)
2318 {
2319 if (dev->priv_flags & IFF_ISATAP)
2320 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
2321 return -1;
2322 }
2323
addrconf_ifid_gre(u8 * eui,struct net_device * dev)2324 static int addrconf_ifid_gre(u8 *eui, struct net_device *dev)
2325 {
2326 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
2327 }
2328
addrconf_ifid_ip6tnl(u8 * eui,struct net_device * dev)2329 static int addrconf_ifid_ip6tnl(u8 *eui, struct net_device *dev)
2330 {
2331 memcpy(eui, dev->perm_addr, 3);
2332 memcpy(eui + 5, dev->perm_addr + 3, 3);
2333 eui[3] = 0xFF;
2334 eui[4] = 0xFE;
2335 eui[0] ^= 2;
2336 return 0;
2337 }
2338
ipv6_generate_eui64(u8 * eui,struct net_device * dev)2339 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev)
2340 {
2341 switch (dev->type) {
2342 case ARPHRD_ETHER:
2343 case ARPHRD_FDDI:
2344 return addrconf_ifid_eui48(eui, dev);
2345 case ARPHRD_ARCNET:
2346 return addrconf_ifid_arcnet(eui, dev);
2347 case ARPHRD_INFINIBAND:
2348 return addrconf_ifid_infiniband(eui, dev);
2349 case ARPHRD_SIT:
2350 return addrconf_ifid_sit(eui, dev);
2351 case ARPHRD_IPGRE:
2352 case ARPHRD_TUNNEL:
2353 return addrconf_ifid_gre(eui, dev);
2354 case ARPHRD_6LOWPAN:
2355 return addrconf_ifid_6lowpan(eui, dev);
2356 case ARPHRD_IEEE1394:
2357 return addrconf_ifid_ieee1394(eui, dev);
2358 case ARPHRD_TUNNEL6:
2359 case ARPHRD_IP6GRE:
2360 case ARPHRD_RAWIP:
2361 return addrconf_ifid_ip6tnl(eui, dev);
2362 }
2363 return -1;
2364 }
2365
ipv6_inherit_eui64(u8 * eui,struct inet6_dev * idev)2366 static int ipv6_inherit_eui64(u8 *eui, struct inet6_dev *idev)
2367 {
2368 int err = -1;
2369 struct inet6_ifaddr *ifp;
2370
2371 read_lock_bh(&idev->lock);
2372 list_for_each_entry_reverse(ifp, &idev->addr_list, if_list) {
2373 if (ifp->scope > IFA_LINK)
2374 break;
2375 if (ifp->scope == IFA_LINK && !(ifp->flags&IFA_F_TENTATIVE)) {
2376 memcpy(eui, ifp->addr.s6_addr+8, 8);
2377 err = 0;
2378 break;
2379 }
2380 }
2381 read_unlock_bh(&idev->lock);
2382 return err;
2383 }
2384
2385 /* Generation of a randomized Interface Identifier
2386 * draft-ietf-6man-rfc4941bis, Section 3.3.1
2387 */
2388
ipv6_gen_rnd_iid(struct in6_addr * addr)2389 static void ipv6_gen_rnd_iid(struct in6_addr *addr)
2390 {
2391 regen:
2392 get_random_bytes(&addr->s6_addr[8], 8);
2393
2394 /* <draft-ietf-6man-rfc4941bis-08.txt>, Section 3.3.1:
2395 * check if generated address is not inappropriate:
2396 *
2397 * - Reserved IPv6 Interface Identifiers
2398 * - XXX: already assigned to an address on the device
2399 */
2400
2401 /* Subnet-router anycast: 0000:0000:0000:0000 */
2402 if (!(addr->s6_addr32[2] | addr->s6_addr32[3]))
2403 goto regen;
2404
2405 /* IANA Ethernet block: 0200:5EFF:FE00:0000-0200:5EFF:FE00:5212
2406 * Proxy Mobile IPv6: 0200:5EFF:FE00:5213
2407 * IANA Ethernet block: 0200:5EFF:FE00:5214-0200:5EFF:FEFF:FFFF
2408 */
2409 if (ntohl(addr->s6_addr32[2]) == 0x02005eff &&
2410 (ntohl(addr->s6_addr32[3]) & 0Xff000000) == 0xfe000000)
2411 goto regen;
2412
2413 /* Reserved subnet anycast addresses */
2414 if (ntohl(addr->s6_addr32[2]) == 0xfdffffff &&
2415 ntohl(addr->s6_addr32[3]) >= 0Xffffff80)
2416 goto regen;
2417 }
2418
2419 /*
2420 * Add prefix route.
2421 */
2422
2423 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)2424 addrconf_prefix_route(struct in6_addr *pfx, int plen, u32 metric,
2425 struct net_device *dev, unsigned long expires,
2426 u32 flags, gfp_t gfp_flags)
2427 {
2428 struct fib6_config cfg = {
2429 .fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_PREFIX,
2430 .fc_metric = metric ? : IP6_RT_PRIO_ADDRCONF,
2431 .fc_ifindex = dev->ifindex,
2432 .fc_expires = expires,
2433 .fc_dst_len = plen,
2434 .fc_flags = RTF_UP | flags,
2435 .fc_nlinfo.nl_net = dev_net(dev),
2436 .fc_protocol = RTPROT_KERNEL,
2437 .fc_type = RTN_UNICAST,
2438 };
2439
2440 cfg.fc_dst = *pfx;
2441
2442 /* Prevent useless cloning on PtP SIT.
2443 This thing is done here expecting that the whole
2444 class of non-broadcast devices need not cloning.
2445 */
2446 #if IS_ENABLED(CONFIG_IPV6_SIT)
2447 if (dev->type == ARPHRD_SIT && (dev->flags & IFF_POINTOPOINT))
2448 cfg.fc_flags |= RTF_NONEXTHOP;
2449 #endif
2450
2451 ip6_route_add(&cfg, gfp_flags, NULL);
2452 }
2453
2454
addrconf_get_prefix_route(const struct in6_addr * pfx,int plen,const struct net_device * dev,u32 flags,u32 noflags,bool no_gw)2455 static struct fib6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
2456 int plen,
2457 const struct net_device *dev,
2458 u32 flags, u32 noflags,
2459 bool no_gw)
2460 {
2461 struct fib6_node *fn;
2462 struct fib6_info *rt = NULL;
2463 struct fib6_table *table;
2464 u32 tb_id = l3mdev_fib_table(dev) ? : RT6_TABLE_PREFIX;
2465
2466 table = fib6_get_table(dev_net(dev), tb_id);
2467 if (!table)
2468 return NULL;
2469
2470 rcu_read_lock();
2471 fn = fib6_locate(&table->tb6_root, pfx, plen, NULL, 0, true);
2472 if (!fn)
2473 goto out;
2474
2475 for_each_fib6_node_rt_rcu(fn) {
2476 /* prefix routes only use builtin fib6_nh */
2477 if (rt->nh)
2478 continue;
2479
2480 if (rt->fib6_nh->fib_nh_dev->ifindex != dev->ifindex)
2481 continue;
2482 if (no_gw && rt->fib6_nh->fib_nh_gw_family)
2483 continue;
2484 if ((rt->fib6_flags & flags) != flags)
2485 continue;
2486 if ((rt->fib6_flags & noflags) != 0)
2487 continue;
2488 if (!fib6_info_hold_safe(rt))
2489 continue;
2490 break;
2491 }
2492 out:
2493 rcu_read_unlock();
2494 return rt;
2495 }
2496
2497
2498 /* Create "default" multicast route to the interface */
2499
addrconf_add_mroute(struct net_device * dev)2500 static void addrconf_add_mroute(struct net_device *dev)
2501 {
2502 struct fib6_config cfg = {
2503 .fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_LOCAL,
2504 .fc_metric = IP6_RT_PRIO_ADDRCONF,
2505 .fc_ifindex = dev->ifindex,
2506 .fc_dst_len = 8,
2507 .fc_flags = RTF_UP,
2508 .fc_type = RTN_MULTICAST,
2509 .fc_nlinfo.nl_net = dev_net(dev),
2510 .fc_protocol = RTPROT_KERNEL,
2511 };
2512
2513 ipv6_addr_set(&cfg.fc_dst, htonl(0xFF000000), 0, 0, 0);
2514
2515 ip6_route_add(&cfg, GFP_KERNEL, NULL);
2516 }
2517
addrconf_add_dev(struct net_device * dev)2518 static struct inet6_dev *addrconf_add_dev(struct net_device *dev)
2519 {
2520 struct inet6_dev *idev;
2521
2522 ASSERT_RTNL();
2523
2524 idev = ipv6_find_idev(dev);
2525 if (IS_ERR(idev))
2526 return idev;
2527
2528 if (idev->cnf.disable_ipv6)
2529 return ERR_PTR(-EACCES);
2530
2531 /* Add default multicast route */
2532 if (!(dev->flags & IFF_LOOPBACK) && !netif_is_l3_master(dev))
2533 addrconf_add_mroute(dev);
2534
2535 return idev;
2536 }
2537
manage_tempaddrs(struct inet6_dev * idev,struct inet6_ifaddr * ifp,__u32 valid_lft,__u32 prefered_lft,bool create,unsigned long now)2538 static void manage_tempaddrs(struct inet6_dev *idev,
2539 struct inet6_ifaddr *ifp,
2540 __u32 valid_lft, __u32 prefered_lft,
2541 bool create, unsigned long now)
2542 {
2543 u32 flags;
2544 struct inet6_ifaddr *ift;
2545
2546 read_lock_bh(&idev->lock);
2547 /* update all temporary addresses in the list */
2548 list_for_each_entry(ift, &idev->tempaddr_list, tmp_list) {
2549 int age, max_valid, max_prefered;
2550
2551 if (ifp != ift->ifpub)
2552 continue;
2553
2554 /* RFC 4941 section 3.3:
2555 * If a received option will extend the lifetime of a public
2556 * address, the lifetimes of temporary addresses should
2557 * be extended, subject to the overall constraint that no
2558 * temporary addresses should ever remain "valid" or "preferred"
2559 * for a time longer than (TEMP_VALID_LIFETIME) or
2560 * (TEMP_PREFERRED_LIFETIME - DESYNC_FACTOR), respectively.
2561 */
2562 age = (now - ift->cstamp) / HZ;
2563 max_valid = idev->cnf.temp_valid_lft - age;
2564 if (max_valid < 0)
2565 max_valid = 0;
2566
2567 max_prefered = idev->cnf.temp_prefered_lft -
2568 idev->desync_factor - age;
2569 if (max_prefered < 0)
2570 max_prefered = 0;
2571
2572 if (valid_lft > max_valid)
2573 valid_lft = max_valid;
2574
2575 if (prefered_lft > max_prefered)
2576 prefered_lft = max_prefered;
2577
2578 spin_lock(&ift->lock);
2579 flags = ift->flags;
2580 ift->valid_lft = valid_lft;
2581 ift->prefered_lft = prefered_lft;
2582 ift->tstamp = now;
2583 if (prefered_lft > 0)
2584 ift->flags &= ~IFA_F_DEPRECATED;
2585
2586 spin_unlock(&ift->lock);
2587 if (!(flags&IFA_F_TENTATIVE))
2588 ipv6_ifa_notify(0, ift);
2589 }
2590
2591 /* Also create a temporary address if it's enabled but no temporary
2592 * address currently exists.
2593 * However, we get called with valid_lft == 0, prefered_lft == 0, create == false
2594 * as part of cleanup (ie. deleting the mngtmpaddr).
2595 * We don't want that to result in creating a new temporary ip address.
2596 */
2597 if (list_empty(&idev->tempaddr_list) && (valid_lft || prefered_lft))
2598 create = true;
2599
2600 if (create && idev->cnf.use_tempaddr > 0) {
2601 /* When a new public address is created as described
2602 * in [ADDRCONF], also create a new temporary address.
2603 */
2604 read_unlock_bh(&idev->lock);
2605 ipv6_create_tempaddr(ifp, false);
2606 } else {
2607 read_unlock_bh(&idev->lock);
2608 }
2609 }
2610
is_addr_mode_generate_stable(struct inet6_dev * idev)2611 static bool is_addr_mode_generate_stable(struct inet6_dev *idev)
2612 {
2613 return idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_STABLE_PRIVACY ||
2614 idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_RANDOM;
2615 }
2616
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)2617 int addrconf_prefix_rcv_add_addr(struct net *net, struct net_device *dev,
2618 const struct prefix_info *pinfo,
2619 struct inet6_dev *in6_dev,
2620 const struct in6_addr *addr, int addr_type,
2621 u32 addr_flags, bool sllao, bool tokenized,
2622 __u32 valid_lft, u32 prefered_lft)
2623 {
2624 struct inet6_ifaddr *ifp = ipv6_get_ifaddr(net, addr, dev, 1);
2625 int create = 0, update_lft = 0;
2626
2627 if (!ifp && valid_lft) {
2628 int max_addresses = in6_dev->cnf.max_addresses;
2629 struct ifa6_config cfg = {
2630 .pfx = addr,
2631 .plen = pinfo->prefix_len,
2632 .ifa_flags = addr_flags,
2633 .valid_lft = valid_lft,
2634 .preferred_lft = prefered_lft,
2635 .scope = addr_type & IPV6_ADDR_SCOPE_MASK,
2636 .ifa_proto = IFAPROT_KERNEL_RA
2637 };
2638
2639 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2640 if ((net->ipv6.devconf_all->optimistic_dad ||
2641 in6_dev->cnf.optimistic_dad) &&
2642 !net->ipv6.devconf_all->forwarding && sllao)
2643 cfg.ifa_flags |= IFA_F_OPTIMISTIC;
2644 #endif
2645
2646 /* Do not allow to create too much of autoconfigured
2647 * addresses; this would be too easy way to crash kernel.
2648 */
2649 if (!max_addresses ||
2650 ipv6_count_addresses(in6_dev) < max_addresses)
2651 ifp = ipv6_add_addr(in6_dev, &cfg, false, NULL);
2652
2653 if (IS_ERR_OR_NULL(ifp))
2654 return -1;
2655
2656 create = 1;
2657 spin_lock_bh(&ifp->lock);
2658 ifp->flags |= IFA_F_MANAGETEMPADDR;
2659 ifp->cstamp = jiffies;
2660 ifp->tokenized = tokenized;
2661 spin_unlock_bh(&ifp->lock);
2662 addrconf_dad_start(ifp);
2663 }
2664
2665 if (ifp) {
2666 u32 flags;
2667 unsigned long now;
2668 u32 stored_lft;
2669
2670 /* update lifetime (RFC2462 5.5.3 e) */
2671 spin_lock_bh(&ifp->lock);
2672 now = jiffies;
2673 if (ifp->valid_lft > (now - ifp->tstamp) / HZ)
2674 stored_lft = ifp->valid_lft - (now - ifp->tstamp) / HZ;
2675 else
2676 stored_lft = 0;
2677 if (!create && stored_lft) {
2678 const u32 minimum_lft = min_t(u32,
2679 stored_lft, MIN_VALID_LIFETIME);
2680 valid_lft = max(valid_lft, minimum_lft);
2681
2682 /* RFC4862 Section 5.5.3e:
2683 * "Note that the preferred lifetime of the
2684 * corresponding address is always reset to
2685 * the Preferred Lifetime in the received
2686 * Prefix Information option, regardless of
2687 * whether the valid lifetime is also reset or
2688 * ignored."
2689 *
2690 * So we should always update prefered_lft here.
2691 */
2692 update_lft = 1;
2693 }
2694
2695 if (update_lft) {
2696 ifp->valid_lft = valid_lft;
2697 ifp->prefered_lft = prefered_lft;
2698 ifp->tstamp = now;
2699 flags = ifp->flags;
2700 ifp->flags &= ~IFA_F_DEPRECATED;
2701 spin_unlock_bh(&ifp->lock);
2702
2703 if (!(flags&IFA_F_TENTATIVE))
2704 ipv6_ifa_notify(0, ifp);
2705 } else
2706 spin_unlock_bh(&ifp->lock);
2707
2708 manage_tempaddrs(in6_dev, ifp, valid_lft, prefered_lft,
2709 create, now);
2710
2711 in6_ifa_put(ifp);
2712 addrconf_verify(net);
2713 }
2714
2715 return 0;
2716 }
2717 EXPORT_SYMBOL_GPL(addrconf_prefix_rcv_add_addr);
2718
addrconf_prefix_rcv(struct net_device * dev,u8 * opt,int len,bool sllao)2719 void addrconf_prefix_rcv(struct net_device *dev, u8 *opt, int len, bool sllao)
2720 {
2721 struct prefix_info *pinfo;
2722 __u32 valid_lft;
2723 __u32 prefered_lft;
2724 int addr_type, err;
2725 u32 addr_flags = 0;
2726 struct inet6_dev *in6_dev;
2727 struct net *net = dev_net(dev);
2728
2729 pinfo = (struct prefix_info *) opt;
2730
2731 if (len < sizeof(struct prefix_info)) {
2732 netdev_dbg(dev, "addrconf: prefix option too short\n");
2733 return;
2734 }
2735
2736 /*
2737 * Validation checks ([ADDRCONF], page 19)
2738 */
2739
2740 addr_type = ipv6_addr_type(&pinfo->prefix);
2741
2742 if (addr_type & (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL))
2743 return;
2744
2745 valid_lft = ntohl(pinfo->valid);
2746 prefered_lft = ntohl(pinfo->prefered);
2747
2748 if (prefered_lft > valid_lft) {
2749 net_warn_ratelimited("addrconf: prefix option has invalid lifetime\n");
2750 return;
2751 }
2752
2753 in6_dev = in6_dev_get(dev);
2754
2755 if (!in6_dev) {
2756 net_dbg_ratelimited("addrconf: device %s not configured\n",
2757 dev->name);
2758 return;
2759 }
2760
2761 if (valid_lft != 0 && valid_lft < in6_dev->cnf.accept_ra_min_lft)
2762 goto put;
2763
2764 /*
2765 * Two things going on here:
2766 * 1) Add routes for on-link prefixes
2767 * 2) Configure prefixes with the auto flag set
2768 */
2769
2770 if (pinfo->onlink) {
2771 struct fib6_info *rt;
2772 unsigned long rt_expires;
2773
2774 /* Avoid arithmetic overflow. Really, we could
2775 * save rt_expires in seconds, likely valid_lft,
2776 * but it would require division in fib gc, that it
2777 * not good.
2778 */
2779 if (HZ > USER_HZ)
2780 rt_expires = addrconf_timeout_fixup(valid_lft, HZ);
2781 else
2782 rt_expires = addrconf_timeout_fixup(valid_lft, USER_HZ);
2783
2784 if (addrconf_finite_timeout(rt_expires))
2785 rt_expires *= HZ;
2786
2787 rt = addrconf_get_prefix_route(&pinfo->prefix,
2788 pinfo->prefix_len,
2789 dev,
2790 RTF_ADDRCONF | RTF_PREFIX_RT,
2791 RTF_DEFAULT, true);
2792
2793 if (rt) {
2794 /* Autoconf prefix route */
2795 if (valid_lft == 0) {
2796 ip6_del_rt(net, rt, false);
2797 rt = NULL;
2798 } else if (addrconf_finite_timeout(rt_expires)) {
2799 /* not infinity */
2800 fib6_set_expires(rt, jiffies + rt_expires);
2801 } else {
2802 fib6_clean_expires(rt);
2803 }
2804 } else if (valid_lft) {
2805 clock_t expires = 0;
2806 int flags = RTF_ADDRCONF | RTF_PREFIX_RT;
2807 if (addrconf_finite_timeout(rt_expires)) {
2808 /* not infinity */
2809 flags |= RTF_EXPIRES;
2810 expires = jiffies_to_clock_t(rt_expires);
2811 }
2812 addrconf_prefix_route(&pinfo->prefix, pinfo->prefix_len,
2813 0, dev, expires, flags,
2814 GFP_ATOMIC);
2815 }
2816 fib6_info_release(rt);
2817 }
2818
2819 /* Try to figure out our local address for this prefix */
2820
2821 if (pinfo->autoconf && in6_dev->cnf.autoconf) {
2822 struct in6_addr addr;
2823 bool tokenized = false, dev_addr_generated = false;
2824
2825 if (pinfo->prefix_len == 64) {
2826 memcpy(&addr, &pinfo->prefix, 8);
2827
2828 if (!ipv6_addr_any(&in6_dev->token)) {
2829 read_lock_bh(&in6_dev->lock);
2830 memcpy(addr.s6_addr + 8,
2831 in6_dev->token.s6_addr + 8, 8);
2832 read_unlock_bh(&in6_dev->lock);
2833 tokenized = true;
2834 } else if (is_addr_mode_generate_stable(in6_dev) &&
2835 !ipv6_generate_stable_address(&addr, 0,
2836 in6_dev)) {
2837 addr_flags |= IFA_F_STABLE_PRIVACY;
2838 goto ok;
2839 } else if (ipv6_generate_eui64(addr.s6_addr + 8, dev) &&
2840 ipv6_inherit_eui64(addr.s6_addr + 8, in6_dev)) {
2841 goto put;
2842 } else {
2843 dev_addr_generated = true;
2844 }
2845 goto ok;
2846 }
2847 net_dbg_ratelimited("IPv6 addrconf: prefix with wrong length %d\n",
2848 pinfo->prefix_len);
2849 goto put;
2850
2851 ok:
2852 err = addrconf_prefix_rcv_add_addr(net, dev, pinfo, in6_dev,
2853 &addr, addr_type,
2854 addr_flags, sllao,
2855 tokenized, valid_lft,
2856 prefered_lft);
2857 if (err)
2858 goto put;
2859
2860 /* Ignore error case here because previous prefix add addr was
2861 * successful which will be notified.
2862 */
2863 ndisc_ops_prefix_rcv_add_addr(net, dev, pinfo, in6_dev, &addr,
2864 addr_type, addr_flags, sllao,
2865 tokenized, valid_lft,
2866 prefered_lft,
2867 dev_addr_generated);
2868 }
2869 inet6_prefix_notify(RTM_NEWPREFIX, in6_dev, pinfo);
2870 put:
2871 in6_dev_put(in6_dev);
2872 }
2873
addrconf_set_sit_dstaddr(struct net * net,struct net_device * dev,struct in6_ifreq * ireq)2874 static int addrconf_set_sit_dstaddr(struct net *net, struct net_device *dev,
2875 struct in6_ifreq *ireq)
2876 {
2877 struct ip_tunnel_parm p = { };
2878 int err;
2879
2880 if (!(ipv6_addr_type(&ireq->ifr6_addr) & IPV6_ADDR_COMPATv4))
2881 return -EADDRNOTAVAIL;
2882
2883 p.iph.daddr = ireq->ifr6_addr.s6_addr32[3];
2884 p.iph.version = 4;
2885 p.iph.ihl = 5;
2886 p.iph.protocol = IPPROTO_IPV6;
2887 p.iph.ttl = 64;
2888
2889 if (!dev->netdev_ops->ndo_tunnel_ctl)
2890 return -EOPNOTSUPP;
2891 err = dev->netdev_ops->ndo_tunnel_ctl(dev, &p, SIOCADDTUNNEL);
2892 if (err)
2893 return err;
2894
2895 dev = __dev_get_by_name(net, p.name);
2896 if (!dev)
2897 return -ENOBUFS;
2898 return dev_open(dev, NULL);
2899 }
2900
2901 /*
2902 * Set destination address.
2903 * Special case for SIT interfaces where we create a new "virtual"
2904 * device.
2905 */
addrconf_set_dstaddr(struct net * net,void __user * arg)2906 int addrconf_set_dstaddr(struct net *net, void __user *arg)
2907 {
2908 struct net_device *dev;
2909 struct in6_ifreq ireq;
2910 int err = -ENODEV;
2911
2912 if (!IS_ENABLED(CONFIG_IPV6_SIT))
2913 return -ENODEV;
2914 if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2915 return -EFAULT;
2916
2917 rtnl_lock();
2918 dev = __dev_get_by_index(net, ireq.ifr6_ifindex);
2919 if (dev && dev->type == ARPHRD_SIT)
2920 err = addrconf_set_sit_dstaddr(net, dev, &ireq);
2921 rtnl_unlock();
2922 return err;
2923 }
2924
ipv6_mc_config(struct sock * sk,bool join,const struct in6_addr * addr,int ifindex)2925 static int ipv6_mc_config(struct sock *sk, bool join,
2926 const struct in6_addr *addr, int ifindex)
2927 {
2928 int ret;
2929
2930 ASSERT_RTNL();
2931
2932 lock_sock(sk);
2933 if (join)
2934 ret = ipv6_sock_mc_join(sk, ifindex, addr);
2935 else
2936 ret = ipv6_sock_mc_drop(sk, ifindex, addr);
2937 release_sock(sk);
2938
2939 return ret;
2940 }
2941
2942 /*
2943 * Manual configuration of address on an interface
2944 */
inet6_addr_add(struct net * net,int ifindex,struct ifa6_config * cfg,struct netlink_ext_ack * extack)2945 static int inet6_addr_add(struct net *net, int ifindex,
2946 struct ifa6_config *cfg,
2947 struct netlink_ext_ack *extack)
2948 {
2949 struct inet6_ifaddr *ifp;
2950 struct inet6_dev *idev;
2951 struct net_device *dev;
2952 unsigned long timeout;
2953 clock_t expires;
2954 u32 flags;
2955
2956 ASSERT_RTNL();
2957
2958 if (cfg->plen > 128) {
2959 NL_SET_ERR_MSG_MOD(extack, "Invalid prefix length");
2960 return -EINVAL;
2961 }
2962
2963 /* check the lifetime */
2964 if (!cfg->valid_lft || cfg->preferred_lft > cfg->valid_lft) {
2965 NL_SET_ERR_MSG_MOD(extack, "address lifetime invalid");
2966 return -EINVAL;
2967 }
2968
2969 if (cfg->ifa_flags & IFA_F_MANAGETEMPADDR && cfg->plen != 64) {
2970 NL_SET_ERR_MSG_MOD(extack, "address with \"mngtmpaddr\" flag must have a prefix length of 64");
2971 return -EINVAL;
2972 }
2973
2974 dev = __dev_get_by_index(net, ifindex);
2975 if (!dev)
2976 return -ENODEV;
2977
2978 idev = addrconf_add_dev(dev);
2979 if (IS_ERR(idev)) {
2980 NL_SET_ERR_MSG_MOD(extack, "IPv6 is disabled on this device");
2981 return PTR_ERR(idev);
2982 }
2983
2984 if (cfg->ifa_flags & IFA_F_MCAUTOJOIN) {
2985 int ret = ipv6_mc_config(net->ipv6.mc_autojoin_sk,
2986 true, cfg->pfx, ifindex);
2987
2988 if (ret < 0) {
2989 NL_SET_ERR_MSG_MOD(extack, "Multicast auto join failed");
2990 return ret;
2991 }
2992 }
2993
2994 cfg->scope = ipv6_addr_scope(cfg->pfx);
2995
2996 timeout = addrconf_timeout_fixup(cfg->valid_lft, HZ);
2997 if (addrconf_finite_timeout(timeout)) {
2998 expires = jiffies_to_clock_t(timeout * HZ);
2999 cfg->valid_lft = timeout;
3000 flags = RTF_EXPIRES;
3001 } else {
3002 expires = 0;
3003 flags = 0;
3004 cfg->ifa_flags |= IFA_F_PERMANENT;
3005 }
3006
3007 timeout = addrconf_timeout_fixup(cfg->preferred_lft, HZ);
3008 if (addrconf_finite_timeout(timeout)) {
3009 if (timeout == 0)
3010 cfg->ifa_flags |= IFA_F_DEPRECATED;
3011 cfg->preferred_lft = timeout;
3012 }
3013
3014 ifp = ipv6_add_addr(idev, cfg, true, extack);
3015 if (!IS_ERR(ifp)) {
3016 if (!(cfg->ifa_flags & IFA_F_NOPREFIXROUTE)) {
3017 addrconf_prefix_route(&ifp->addr, ifp->prefix_len,
3018 ifp->rt_priority, dev, expires,
3019 flags, GFP_KERNEL);
3020 }
3021
3022 /* Send a netlink notification if DAD is enabled and
3023 * optimistic flag is not set
3024 */
3025 if (!(ifp->flags & (IFA_F_OPTIMISTIC | IFA_F_NODAD)))
3026 ipv6_ifa_notify(0, ifp);
3027 /*
3028 * Note that section 3.1 of RFC 4429 indicates
3029 * that the Optimistic flag should not be set for
3030 * manually configured addresses
3031 */
3032 addrconf_dad_start(ifp);
3033 if (cfg->ifa_flags & IFA_F_MANAGETEMPADDR)
3034 manage_tempaddrs(idev, ifp, cfg->valid_lft,
3035 cfg->preferred_lft, true, jiffies);
3036 in6_ifa_put(ifp);
3037 addrconf_verify_rtnl(net);
3038 return 0;
3039 } else if (cfg->ifa_flags & IFA_F_MCAUTOJOIN) {
3040 ipv6_mc_config(net->ipv6.mc_autojoin_sk, false,
3041 cfg->pfx, ifindex);
3042 }
3043
3044 return PTR_ERR(ifp);
3045 }
3046
inet6_addr_del(struct net * net,int ifindex,u32 ifa_flags,const struct in6_addr * pfx,unsigned int plen,struct netlink_ext_ack * extack)3047 static int inet6_addr_del(struct net *net, int ifindex, u32 ifa_flags,
3048 const struct in6_addr *pfx, unsigned int plen,
3049 struct netlink_ext_ack *extack)
3050 {
3051 struct inet6_ifaddr *ifp;
3052 struct inet6_dev *idev;
3053 struct net_device *dev;
3054
3055 if (plen > 128) {
3056 NL_SET_ERR_MSG_MOD(extack, "Invalid prefix length");
3057 return -EINVAL;
3058 }
3059
3060 dev = __dev_get_by_index(net, ifindex);
3061 if (!dev) {
3062 NL_SET_ERR_MSG_MOD(extack, "Unable to find the interface");
3063 return -ENODEV;
3064 }
3065
3066 idev = __in6_dev_get(dev);
3067 if (!idev) {
3068 NL_SET_ERR_MSG_MOD(extack, "IPv6 is disabled on this device");
3069 return -ENXIO;
3070 }
3071
3072 read_lock_bh(&idev->lock);
3073 list_for_each_entry(ifp, &idev->addr_list, if_list) {
3074 if (ifp->prefix_len == plen &&
3075 ipv6_addr_equal(pfx, &ifp->addr)) {
3076 in6_ifa_hold(ifp);
3077 read_unlock_bh(&idev->lock);
3078
3079 if (!(ifp->flags & IFA_F_TEMPORARY) &&
3080 (ifa_flags & IFA_F_MANAGETEMPADDR))
3081 manage_tempaddrs(idev, ifp, 0, 0, false,
3082 jiffies);
3083 ipv6_del_addr(ifp);
3084 addrconf_verify_rtnl(net);
3085 if (ipv6_addr_is_multicast(pfx)) {
3086 ipv6_mc_config(net->ipv6.mc_autojoin_sk,
3087 false, pfx, dev->ifindex);
3088 }
3089 return 0;
3090 }
3091 }
3092 read_unlock_bh(&idev->lock);
3093
3094 NL_SET_ERR_MSG_MOD(extack, "address not found");
3095 return -EADDRNOTAVAIL;
3096 }
3097
3098
addrconf_add_ifaddr(struct net * net,void __user * arg)3099 int addrconf_add_ifaddr(struct net *net, void __user *arg)
3100 {
3101 struct ifa6_config cfg = {
3102 .ifa_flags = IFA_F_PERMANENT,
3103 .preferred_lft = INFINITY_LIFE_TIME,
3104 .valid_lft = INFINITY_LIFE_TIME,
3105 };
3106 struct in6_ifreq ireq;
3107 int err;
3108
3109 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
3110 return -EPERM;
3111
3112 if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
3113 return -EFAULT;
3114
3115 cfg.pfx = &ireq.ifr6_addr;
3116 cfg.plen = ireq.ifr6_prefixlen;
3117
3118 rtnl_lock();
3119 err = inet6_addr_add(net, ireq.ifr6_ifindex, &cfg, NULL);
3120 rtnl_unlock();
3121 return err;
3122 }
3123
addrconf_del_ifaddr(struct net * net,void __user * arg)3124 int addrconf_del_ifaddr(struct net *net, void __user *arg)
3125 {
3126 struct in6_ifreq ireq;
3127 int err;
3128
3129 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
3130 return -EPERM;
3131
3132 if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
3133 return -EFAULT;
3134
3135 rtnl_lock();
3136 err = inet6_addr_del(net, ireq.ifr6_ifindex, 0, &ireq.ifr6_addr,
3137 ireq.ifr6_prefixlen, NULL);
3138 rtnl_unlock();
3139 return err;
3140 }
3141
add_addr(struct inet6_dev * idev,const struct in6_addr * addr,int plen,int scope,u8 proto)3142 static void add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
3143 int plen, int scope, u8 proto)
3144 {
3145 struct inet6_ifaddr *ifp;
3146 struct ifa6_config cfg = {
3147 .pfx = addr,
3148 .plen = plen,
3149 .ifa_flags = IFA_F_PERMANENT,
3150 .valid_lft = INFINITY_LIFE_TIME,
3151 .preferred_lft = INFINITY_LIFE_TIME,
3152 .scope = scope,
3153 .ifa_proto = proto
3154 };
3155
3156 ifp = ipv6_add_addr(idev, &cfg, true, NULL);
3157 if (!IS_ERR(ifp)) {
3158 spin_lock_bh(&ifp->lock);
3159 ifp->flags &= ~IFA_F_TENTATIVE;
3160 spin_unlock_bh(&ifp->lock);
3161 rt_genid_bump_ipv6(dev_net(idev->dev));
3162 ipv6_ifa_notify(RTM_NEWADDR, ifp);
3163 in6_ifa_put(ifp);
3164 }
3165 }
3166
3167 #if IS_ENABLED(CONFIG_IPV6_SIT) || IS_ENABLED(CONFIG_NET_IPGRE) || IS_ENABLED(CONFIG_IPV6_GRE)
add_v4_addrs(struct inet6_dev * idev)3168 static void add_v4_addrs(struct inet6_dev *idev)
3169 {
3170 struct in6_addr addr;
3171 struct net_device *dev;
3172 struct net *net = dev_net(idev->dev);
3173 int scope, plen, offset = 0;
3174 u32 pflags = 0;
3175
3176 ASSERT_RTNL();
3177
3178 memset(&addr, 0, sizeof(struct in6_addr));
3179 /* in case of IP6GRE the dev_addr is an IPv6 and therefore we use only the last 4 bytes */
3180 if (idev->dev->addr_len == sizeof(struct in6_addr))
3181 offset = sizeof(struct in6_addr) - 4;
3182 memcpy(&addr.s6_addr32[3], idev->dev->dev_addr + offset, 4);
3183
3184 if (!(idev->dev->flags & IFF_POINTOPOINT) && idev->dev->type == ARPHRD_SIT) {
3185 scope = IPV6_ADDR_COMPATv4;
3186 plen = 96;
3187 pflags |= RTF_NONEXTHOP;
3188 } else {
3189 if (idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_NONE)
3190 return;
3191
3192 addr.s6_addr32[0] = htonl(0xfe800000);
3193 scope = IFA_LINK;
3194 plen = 64;
3195 }
3196
3197 if (addr.s6_addr32[3]) {
3198 add_addr(idev, &addr, plen, scope, IFAPROT_UNSPEC);
3199 addrconf_prefix_route(&addr, plen, 0, idev->dev, 0, pflags,
3200 GFP_KERNEL);
3201 return;
3202 }
3203
3204 for_each_netdev(net, dev) {
3205 struct in_device *in_dev = __in_dev_get_rtnl(dev);
3206 if (in_dev && (dev->flags & IFF_UP)) {
3207 struct in_ifaddr *ifa;
3208 int flag = scope;
3209
3210 in_dev_for_each_ifa_rtnl(ifa, in_dev) {
3211 addr.s6_addr32[3] = ifa->ifa_local;
3212
3213 if (ifa->ifa_scope == RT_SCOPE_LINK)
3214 continue;
3215 if (ifa->ifa_scope >= RT_SCOPE_HOST) {
3216 if (idev->dev->flags&IFF_POINTOPOINT)
3217 continue;
3218 flag |= IFA_HOST;
3219 }
3220
3221 add_addr(idev, &addr, plen, flag,
3222 IFAPROT_UNSPEC);
3223 addrconf_prefix_route(&addr, plen, 0, idev->dev,
3224 0, pflags, GFP_KERNEL);
3225 }
3226 }
3227 }
3228 }
3229 #endif
3230
init_loopback(struct net_device * dev)3231 static void init_loopback(struct net_device *dev)
3232 {
3233 struct inet6_dev *idev;
3234
3235 /* ::1 */
3236
3237 ASSERT_RTNL();
3238
3239 idev = ipv6_find_idev(dev);
3240 if (IS_ERR(idev)) {
3241 pr_debug("%s: add_dev failed\n", __func__);
3242 return;
3243 }
3244
3245 add_addr(idev, &in6addr_loopback, 128, IFA_HOST, IFAPROT_KERNEL_LO);
3246 }
3247
addrconf_add_linklocal(struct inet6_dev * idev,const struct in6_addr * addr,u32 flags)3248 void addrconf_add_linklocal(struct inet6_dev *idev,
3249 const struct in6_addr *addr, u32 flags)
3250 {
3251 struct ifa6_config cfg = {
3252 .pfx = addr,
3253 .plen = 64,
3254 .ifa_flags = flags | IFA_F_PERMANENT,
3255 .valid_lft = INFINITY_LIFE_TIME,
3256 .preferred_lft = INFINITY_LIFE_TIME,
3257 .scope = IFA_LINK,
3258 .ifa_proto = IFAPROT_KERNEL_LL
3259 };
3260 struct inet6_ifaddr *ifp;
3261
3262 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
3263 if ((dev_net(idev->dev)->ipv6.devconf_all->optimistic_dad ||
3264 idev->cnf.optimistic_dad) &&
3265 !dev_net(idev->dev)->ipv6.devconf_all->forwarding)
3266 cfg.ifa_flags |= IFA_F_OPTIMISTIC;
3267 #endif
3268
3269 ifp = ipv6_add_addr(idev, &cfg, true, NULL);
3270 if (!IS_ERR(ifp)) {
3271 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, 0, idev->dev,
3272 0, 0, GFP_ATOMIC);
3273 addrconf_dad_start(ifp);
3274 in6_ifa_put(ifp);
3275 }
3276 }
3277 EXPORT_SYMBOL_GPL(addrconf_add_linklocal);
3278
ipv6_reserved_interfaceid(struct in6_addr address)3279 static bool ipv6_reserved_interfaceid(struct in6_addr address)
3280 {
3281 if ((address.s6_addr32[2] | address.s6_addr32[3]) == 0)
3282 return true;
3283
3284 if (address.s6_addr32[2] == htonl(0x02005eff) &&
3285 ((address.s6_addr32[3] & htonl(0xfe000000)) == htonl(0xfe000000)))
3286 return true;
3287
3288 if (address.s6_addr32[2] == htonl(0xfdffffff) &&
3289 ((address.s6_addr32[3] & htonl(0xffffff80)) == htonl(0xffffff80)))
3290 return true;
3291
3292 return false;
3293 }
3294
ipv6_generate_stable_address(struct in6_addr * address,u8 dad_count,const struct inet6_dev * idev)3295 static int ipv6_generate_stable_address(struct in6_addr *address,
3296 u8 dad_count,
3297 const struct inet6_dev *idev)
3298 {
3299 static DEFINE_SPINLOCK(lock);
3300 static __u32 digest[SHA1_DIGEST_WORDS];
3301 static __u32 workspace[SHA1_WORKSPACE_WORDS];
3302
3303 static union {
3304 char __data[SHA1_BLOCK_SIZE];
3305 struct {
3306 struct in6_addr secret;
3307 __be32 prefix[2];
3308 unsigned char hwaddr[MAX_ADDR_LEN];
3309 u8 dad_count;
3310 } __packed;
3311 } data;
3312
3313 struct in6_addr secret;
3314 struct in6_addr temp;
3315 struct net *net = dev_net(idev->dev);
3316
3317 BUILD_BUG_ON(sizeof(data.__data) != sizeof(data));
3318
3319 if (idev->cnf.stable_secret.initialized)
3320 secret = idev->cnf.stable_secret.secret;
3321 else if (net->ipv6.devconf_dflt->stable_secret.initialized)
3322 secret = net->ipv6.devconf_dflt->stable_secret.secret;
3323 else
3324 return -1;
3325
3326 retry:
3327 spin_lock_bh(&lock);
3328
3329 sha1_init(digest);
3330 memset(&data, 0, sizeof(data));
3331 memset(workspace, 0, sizeof(workspace));
3332 memcpy(data.hwaddr, idev->dev->perm_addr, idev->dev->addr_len);
3333 data.prefix[0] = address->s6_addr32[0];
3334 data.prefix[1] = address->s6_addr32[1];
3335 data.secret = secret;
3336 data.dad_count = dad_count;
3337
3338 sha1_transform(digest, data.__data, workspace);
3339
3340 temp = *address;
3341 temp.s6_addr32[2] = (__force __be32)digest[0];
3342 temp.s6_addr32[3] = (__force __be32)digest[1];
3343
3344 spin_unlock_bh(&lock);
3345
3346 if (ipv6_reserved_interfaceid(temp)) {
3347 dad_count++;
3348 if (dad_count > dev_net(idev->dev)->ipv6.sysctl.idgen_retries)
3349 return -1;
3350 goto retry;
3351 }
3352
3353 *address = temp;
3354 return 0;
3355 }
3356
ipv6_gen_mode_random_init(struct inet6_dev * idev)3357 static void ipv6_gen_mode_random_init(struct inet6_dev *idev)
3358 {
3359 struct ipv6_stable_secret *s = &idev->cnf.stable_secret;
3360
3361 if (s->initialized)
3362 return;
3363 s = &idev->cnf.stable_secret;
3364 get_random_bytes(&s->secret, sizeof(s->secret));
3365 s->initialized = true;
3366 }
3367
addrconf_addr_gen(struct inet6_dev * idev,bool prefix_route)3368 static void addrconf_addr_gen(struct inet6_dev *idev, bool prefix_route)
3369 {
3370 struct in6_addr addr;
3371
3372 /* no link local addresses on L3 master devices */
3373 if (netif_is_l3_master(idev->dev))
3374 return;
3375
3376 /* no link local addresses on devices flagged as slaves */
3377 if (idev->dev->priv_flags & IFF_NO_ADDRCONF)
3378 return;
3379
3380 ipv6_addr_set(&addr, htonl(0xFE800000), 0, 0, 0);
3381
3382 switch (idev->cnf.addr_gen_mode) {
3383 case IN6_ADDR_GEN_MODE_RANDOM:
3384 ipv6_gen_mode_random_init(idev);
3385 fallthrough;
3386 case IN6_ADDR_GEN_MODE_STABLE_PRIVACY:
3387 if (!ipv6_generate_stable_address(&addr, 0, idev))
3388 addrconf_add_linklocal(idev, &addr,
3389 IFA_F_STABLE_PRIVACY);
3390 else if (prefix_route)
3391 addrconf_prefix_route(&addr, 64, 0, idev->dev,
3392 0, 0, GFP_KERNEL);
3393 break;
3394 case IN6_ADDR_GEN_MODE_EUI64:
3395 /* addrconf_add_linklocal also adds a prefix_route and we
3396 * only need to care about prefix routes if ipv6_generate_eui64
3397 * couldn't generate one.
3398 */
3399 if (ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) == 0)
3400 addrconf_add_linklocal(idev, &addr, 0);
3401 else if (prefix_route)
3402 addrconf_prefix_route(&addr, 64, 0, idev->dev,
3403 0, 0, GFP_KERNEL);
3404 break;
3405 case IN6_ADDR_GEN_MODE_NONE:
3406 default:
3407 /* will not add any link local address */
3408 break;
3409 }
3410 }
3411
addrconf_dev_config(struct net_device * dev)3412 static void addrconf_dev_config(struct net_device *dev)
3413 {
3414 struct inet6_dev *idev;
3415
3416 ASSERT_RTNL();
3417
3418 if ((dev->type != ARPHRD_ETHER) &&
3419 (dev->type != ARPHRD_FDDI) &&
3420 (dev->type != ARPHRD_ARCNET) &&
3421 (dev->type != ARPHRD_INFINIBAND) &&
3422 (dev->type != ARPHRD_IEEE1394) &&
3423 (dev->type != ARPHRD_TUNNEL6) &&
3424 (dev->type != ARPHRD_6LOWPAN) &&
3425 (dev->type != ARPHRD_TUNNEL) &&
3426 (dev->type != ARPHRD_NONE) &&
3427 (dev->type != ARPHRD_RAWIP)) {
3428 /* Alas, we support only Ethernet autoconfiguration. */
3429 idev = __in6_dev_get(dev);
3430 if (!IS_ERR_OR_NULL(idev) && dev->flags & IFF_UP &&
3431 dev->flags & IFF_MULTICAST)
3432 ipv6_mc_up(idev);
3433 return;
3434 }
3435
3436 idev = addrconf_add_dev(dev);
3437 if (IS_ERR(idev))
3438 return;
3439
3440 /* this device type has no EUI support */
3441 if (dev->type == ARPHRD_NONE &&
3442 idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_EUI64)
3443 idev->cnf.addr_gen_mode = IN6_ADDR_GEN_MODE_RANDOM;
3444
3445 addrconf_addr_gen(idev, false);
3446 }
3447
3448 #if IS_ENABLED(CONFIG_IPV6_SIT)
addrconf_sit_config(struct net_device * dev)3449 static void addrconf_sit_config(struct net_device *dev)
3450 {
3451 struct inet6_dev *idev;
3452
3453 ASSERT_RTNL();
3454
3455 /*
3456 * Configure the tunnel with one of our IPv4
3457 * addresses... we should configure all of
3458 * our v4 addrs in the tunnel
3459 */
3460
3461 idev = ipv6_find_idev(dev);
3462 if (IS_ERR(idev)) {
3463 pr_debug("%s: add_dev failed\n", __func__);
3464 return;
3465 }
3466
3467 if (dev->priv_flags & IFF_ISATAP) {
3468 addrconf_addr_gen(idev, false);
3469 return;
3470 }
3471
3472 add_v4_addrs(idev);
3473
3474 if (dev->flags&IFF_POINTOPOINT)
3475 addrconf_add_mroute(dev);
3476 }
3477 #endif
3478
3479 #if IS_ENABLED(CONFIG_NET_IPGRE) || IS_ENABLED(CONFIG_IPV6_GRE)
addrconf_gre_config(struct net_device * dev)3480 static void addrconf_gre_config(struct net_device *dev)
3481 {
3482 struct inet6_dev *idev;
3483
3484 ASSERT_RTNL();
3485
3486 idev = ipv6_find_idev(dev);
3487 if (IS_ERR(idev)) {
3488 pr_debug("%s: add_dev failed\n", __func__);
3489 return;
3490 }
3491
3492 if (dev->type == ARPHRD_ETHER) {
3493 addrconf_addr_gen(idev, true);
3494 return;
3495 }
3496
3497 add_v4_addrs(idev);
3498
3499 if (dev->flags & IFF_POINTOPOINT)
3500 addrconf_add_mroute(dev);
3501 }
3502 #endif
3503
addrconf_init_auto_addrs(struct net_device * dev)3504 static void addrconf_init_auto_addrs(struct net_device *dev)
3505 {
3506 switch (dev->type) {
3507 #if IS_ENABLED(CONFIG_IPV6_SIT)
3508 case ARPHRD_SIT:
3509 addrconf_sit_config(dev);
3510 break;
3511 #endif
3512 #if IS_ENABLED(CONFIG_NET_IPGRE) || IS_ENABLED(CONFIG_IPV6_GRE)
3513 case ARPHRD_IP6GRE:
3514 case ARPHRD_IPGRE:
3515 addrconf_gre_config(dev);
3516 break;
3517 #endif
3518 case ARPHRD_LOOPBACK:
3519 init_loopback(dev);
3520 break;
3521
3522 default:
3523 addrconf_dev_config(dev);
3524 break;
3525 }
3526 }
3527
fixup_permanent_addr(struct net * net,struct inet6_dev * idev,struct inet6_ifaddr * ifp)3528 static int fixup_permanent_addr(struct net *net,
3529 struct inet6_dev *idev,
3530 struct inet6_ifaddr *ifp)
3531 {
3532 /* !fib6_node means the host route was removed from the
3533 * FIB, for example, if 'lo' device is taken down. In that
3534 * case regenerate the host route.
3535 */
3536 if (!ifp->rt || !ifp->rt->fib6_node) {
3537 struct fib6_info *f6i, *prev;
3538
3539 f6i = addrconf_f6i_alloc(net, idev, &ifp->addr, false,
3540 GFP_ATOMIC, NULL);
3541 if (IS_ERR(f6i))
3542 return PTR_ERR(f6i);
3543
3544 /* ifp->rt can be accessed outside of rtnl */
3545 spin_lock(&ifp->lock);
3546 prev = ifp->rt;
3547 ifp->rt = f6i;
3548 spin_unlock(&ifp->lock);
3549
3550 fib6_info_release(prev);
3551 }
3552
3553 if (!(ifp->flags & IFA_F_NOPREFIXROUTE)) {
3554 addrconf_prefix_route(&ifp->addr, ifp->prefix_len,
3555 ifp->rt_priority, idev->dev, 0, 0,
3556 GFP_ATOMIC);
3557 }
3558
3559 if (ifp->state == INET6_IFADDR_STATE_PREDAD)
3560 addrconf_dad_start(ifp);
3561
3562 return 0;
3563 }
3564
addrconf_permanent_addr(struct net * net,struct net_device * dev)3565 static void addrconf_permanent_addr(struct net *net, struct net_device *dev)
3566 {
3567 struct inet6_ifaddr *ifp, *tmp;
3568 struct inet6_dev *idev;
3569
3570 idev = __in6_dev_get(dev);
3571 if (!idev)
3572 return;
3573
3574 write_lock_bh(&idev->lock);
3575
3576 list_for_each_entry_safe(ifp, tmp, &idev->addr_list, if_list) {
3577 if ((ifp->flags & IFA_F_PERMANENT) &&
3578 fixup_permanent_addr(net, idev, ifp) < 0) {
3579 write_unlock_bh(&idev->lock);
3580 in6_ifa_hold(ifp);
3581 ipv6_del_addr(ifp);
3582 write_lock_bh(&idev->lock);
3583
3584 net_info_ratelimited("%s: Failed to add prefix route for address %pI6c; dropping\n",
3585 idev->dev->name, &ifp->addr);
3586 }
3587 }
3588
3589 write_unlock_bh(&idev->lock);
3590 }
3591
addrconf_notify(struct notifier_block * this,unsigned long event,void * ptr)3592 static int addrconf_notify(struct notifier_block *this, unsigned long event,
3593 void *ptr)
3594 {
3595 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
3596 struct netdev_notifier_change_info *change_info;
3597 struct netdev_notifier_changeupper_info *info;
3598 struct inet6_dev *idev = __in6_dev_get(dev);
3599 struct net *net = dev_net(dev);
3600 int run_pending = 0;
3601 int err;
3602
3603 switch (event) {
3604 case NETDEV_REGISTER:
3605 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
3606 idev = ipv6_add_dev(dev);
3607 if (IS_ERR(idev))
3608 return notifier_from_errno(PTR_ERR(idev));
3609 }
3610 break;
3611
3612 case NETDEV_CHANGEMTU:
3613 /* if MTU under IPV6_MIN_MTU stop IPv6 on this interface. */
3614 if (dev->mtu < IPV6_MIN_MTU) {
3615 addrconf_ifdown(dev, dev != net->loopback_dev);
3616 break;
3617 }
3618
3619 if (idev) {
3620 rt6_mtu_change(dev, dev->mtu);
3621 idev->cnf.mtu6 = dev->mtu;
3622 break;
3623 }
3624
3625 /* allocate new idev */
3626 idev = ipv6_add_dev(dev);
3627 if (IS_ERR(idev))
3628 break;
3629
3630 /* device is still not ready */
3631 if (!(idev->if_flags & IF_READY))
3632 break;
3633
3634 run_pending = 1;
3635 fallthrough;
3636 case NETDEV_UP:
3637 case NETDEV_CHANGE:
3638 if (idev && idev->cnf.disable_ipv6)
3639 break;
3640
3641 if (dev->priv_flags & IFF_NO_ADDRCONF) {
3642 if (event == NETDEV_UP && !IS_ERR_OR_NULL(idev) &&
3643 dev->flags & IFF_UP && dev->flags & IFF_MULTICAST)
3644 ipv6_mc_up(idev);
3645 break;
3646 }
3647
3648 if (event == NETDEV_UP) {
3649 /* restore routes for permanent addresses */
3650 addrconf_permanent_addr(net, dev);
3651
3652 if (!addrconf_link_ready(dev)) {
3653 /* device is not ready yet. */
3654 pr_debug("ADDRCONF(NETDEV_UP): %s: link is not ready\n",
3655 dev->name);
3656 break;
3657 }
3658
3659 if (!idev && dev->mtu >= IPV6_MIN_MTU)
3660 idev = ipv6_add_dev(dev);
3661
3662 if (!IS_ERR_OR_NULL(idev)) {
3663 idev->if_flags |= IF_READY;
3664 run_pending = 1;
3665 }
3666 } else if (event == NETDEV_CHANGE) {
3667 if (!addrconf_link_ready(dev)) {
3668 /* device is still not ready. */
3669 rt6_sync_down_dev(dev, event);
3670 break;
3671 }
3672
3673 if (!IS_ERR_OR_NULL(idev)) {
3674 if (idev->if_flags & IF_READY) {
3675 /* device is already configured -
3676 * but resend MLD reports, we might
3677 * have roamed and need to update
3678 * multicast snooping switches
3679 */
3680 ipv6_mc_up(idev);
3681 change_info = ptr;
3682 if (change_info->flags_changed & IFF_NOARP)
3683 addrconf_dad_run(idev, true);
3684 rt6_sync_up(dev, RTNH_F_LINKDOWN);
3685 break;
3686 }
3687 idev->if_flags |= IF_READY;
3688 }
3689
3690 pr_debug("ADDRCONF(NETDEV_CHANGE): %s: link becomes ready\n",
3691 dev->name);
3692
3693 run_pending = 1;
3694 }
3695
3696 addrconf_init_auto_addrs(dev);
3697
3698 if (!IS_ERR_OR_NULL(idev)) {
3699 if (run_pending)
3700 addrconf_dad_run(idev, false);
3701
3702 /* Device has an address by now */
3703 rt6_sync_up(dev, RTNH_F_DEAD);
3704
3705 /*
3706 * If the MTU changed during the interface down,
3707 * when the interface up, the changed MTU must be
3708 * reflected in the idev as well as routers.
3709 */
3710 if (idev->cnf.mtu6 != dev->mtu &&
3711 dev->mtu >= IPV6_MIN_MTU) {
3712 rt6_mtu_change(dev, dev->mtu);
3713 idev->cnf.mtu6 = dev->mtu;
3714 }
3715 idev->tstamp = jiffies;
3716 inet6_ifinfo_notify(RTM_NEWLINK, idev);
3717
3718 /*
3719 * If the changed mtu during down is lower than
3720 * IPV6_MIN_MTU stop IPv6 on this interface.
3721 */
3722 if (dev->mtu < IPV6_MIN_MTU)
3723 addrconf_ifdown(dev, dev != net->loopback_dev);
3724 }
3725 break;
3726
3727 case NETDEV_DOWN:
3728 case NETDEV_UNREGISTER:
3729 /*
3730 * Remove all addresses from this interface.
3731 */
3732 addrconf_ifdown(dev, event != NETDEV_DOWN);
3733 break;
3734
3735 case NETDEV_CHANGENAME:
3736 if (idev) {
3737 snmp6_unregister_dev(idev);
3738 addrconf_sysctl_unregister(idev);
3739 err = addrconf_sysctl_register(idev);
3740 if (err)
3741 return notifier_from_errno(err);
3742 err = snmp6_register_dev(idev);
3743 if (err) {
3744 addrconf_sysctl_unregister(idev);
3745 return notifier_from_errno(err);
3746 }
3747 }
3748 break;
3749
3750 case NETDEV_PRE_TYPE_CHANGE:
3751 case NETDEV_POST_TYPE_CHANGE:
3752 if (idev)
3753 addrconf_type_change(dev, event);
3754 break;
3755
3756 case NETDEV_CHANGEUPPER:
3757 info = ptr;
3758
3759 /* flush all routes if dev is linked to or unlinked from
3760 * an L3 master device (e.g., VRF)
3761 */
3762 if (info->upper_dev && netif_is_l3_master(info->upper_dev))
3763 addrconf_ifdown(dev, false);
3764 }
3765
3766 return NOTIFY_OK;
3767 }
3768
3769 /*
3770 * addrconf module should be notified of a device going up
3771 */
3772 static struct notifier_block ipv6_dev_notf = {
3773 .notifier_call = addrconf_notify,
3774 .priority = ADDRCONF_NOTIFY_PRIORITY,
3775 };
3776
addrconf_type_change(struct net_device * dev,unsigned long event)3777 static void addrconf_type_change(struct net_device *dev, unsigned long event)
3778 {
3779 struct inet6_dev *idev;
3780 ASSERT_RTNL();
3781
3782 idev = __in6_dev_get(dev);
3783
3784 if (event == NETDEV_POST_TYPE_CHANGE)
3785 ipv6_mc_remap(idev);
3786 else if (event == NETDEV_PRE_TYPE_CHANGE)
3787 ipv6_mc_unmap(idev);
3788 }
3789
addr_is_local(const struct in6_addr * addr)3790 static bool addr_is_local(const struct in6_addr *addr)
3791 {
3792 return ipv6_addr_type(addr) &
3793 (IPV6_ADDR_LINKLOCAL | IPV6_ADDR_LOOPBACK);
3794 }
3795
addrconf_ifdown(struct net_device * dev,bool unregister)3796 static int addrconf_ifdown(struct net_device *dev, bool unregister)
3797 {
3798 unsigned long event = unregister ? NETDEV_UNREGISTER : NETDEV_DOWN;
3799 struct net *net = dev_net(dev);
3800 struct inet6_dev *idev;
3801 struct inet6_ifaddr *ifa;
3802 LIST_HEAD(tmp_addr_list);
3803 bool keep_addr = false;
3804 bool was_ready;
3805 int state, i;
3806
3807 ASSERT_RTNL();
3808
3809 rt6_disable_ip(dev, event);
3810
3811 idev = __in6_dev_get(dev);
3812 if (!idev)
3813 return -ENODEV;
3814
3815 /*
3816 * Step 1: remove reference to ipv6 device from parent device.
3817 * Do not dev_put!
3818 */
3819 if (unregister) {
3820 idev->dead = 1;
3821
3822 /* protected by rtnl_lock */
3823 RCU_INIT_POINTER(dev->ip6_ptr, NULL);
3824
3825 /* Step 1.5: remove snmp6 entry */
3826 snmp6_unregister_dev(idev);
3827
3828 }
3829
3830 /* combine the user config with event to determine if permanent
3831 * addresses are to be removed from address hash table
3832 */
3833 if (!unregister && !idev->cnf.disable_ipv6) {
3834 /* aggregate the system setting and interface setting */
3835 int _keep_addr = net->ipv6.devconf_all->keep_addr_on_down;
3836
3837 if (!_keep_addr)
3838 _keep_addr = idev->cnf.keep_addr_on_down;
3839
3840 keep_addr = (_keep_addr > 0);
3841 }
3842
3843 /* Step 2: clear hash table */
3844 for (i = 0; i < IN6_ADDR_HSIZE; i++) {
3845 struct hlist_head *h = &net->ipv6.inet6_addr_lst[i];
3846
3847 spin_lock_bh(&net->ipv6.addrconf_hash_lock);
3848 restart:
3849 hlist_for_each_entry_rcu(ifa, h, addr_lst) {
3850 if (ifa->idev == idev) {
3851 addrconf_del_dad_work(ifa);
3852 /* combined flag + permanent flag decide if
3853 * address is retained on a down event
3854 */
3855 if (!keep_addr ||
3856 !(ifa->flags & IFA_F_PERMANENT) ||
3857 addr_is_local(&ifa->addr)) {
3858 hlist_del_init_rcu(&ifa->addr_lst);
3859 goto restart;
3860 }
3861 }
3862 }
3863 spin_unlock_bh(&net->ipv6.addrconf_hash_lock);
3864 }
3865
3866 write_lock_bh(&idev->lock);
3867
3868 addrconf_del_rs_timer(idev);
3869
3870 /* Step 2: clear flags for stateless addrconf, repeated down
3871 * detection
3872 */
3873 was_ready = idev->if_flags & IF_READY;
3874 if (!unregister)
3875 idev->if_flags &= ~(IF_RS_SENT|IF_RA_RCVD|IF_READY);
3876
3877 /* Step 3: clear tempaddr list */
3878 while (!list_empty(&idev->tempaddr_list)) {
3879 ifa = list_first_entry(&idev->tempaddr_list,
3880 struct inet6_ifaddr, tmp_list);
3881 list_del(&ifa->tmp_list);
3882 write_unlock_bh(&idev->lock);
3883 spin_lock_bh(&ifa->lock);
3884
3885 if (ifa->ifpub) {
3886 in6_ifa_put(ifa->ifpub);
3887 ifa->ifpub = NULL;
3888 }
3889 spin_unlock_bh(&ifa->lock);
3890 in6_ifa_put(ifa);
3891 write_lock_bh(&idev->lock);
3892 }
3893
3894 list_for_each_entry(ifa, &idev->addr_list, if_list)
3895 list_add_tail(&ifa->if_list_aux, &tmp_addr_list);
3896 write_unlock_bh(&idev->lock);
3897
3898 while (!list_empty(&tmp_addr_list)) {
3899 struct fib6_info *rt = NULL;
3900 bool keep;
3901
3902 ifa = list_first_entry(&tmp_addr_list,
3903 struct inet6_ifaddr, if_list_aux);
3904 list_del(&ifa->if_list_aux);
3905
3906 addrconf_del_dad_work(ifa);
3907
3908 keep = keep_addr && (ifa->flags & IFA_F_PERMANENT) &&
3909 !addr_is_local(&ifa->addr);
3910
3911 spin_lock_bh(&ifa->lock);
3912
3913 if (keep) {
3914 /* set state to skip the notifier below */
3915 state = INET6_IFADDR_STATE_DEAD;
3916 ifa->state = INET6_IFADDR_STATE_PREDAD;
3917 if (!(ifa->flags & IFA_F_NODAD))
3918 ifa->flags |= IFA_F_TENTATIVE;
3919
3920 rt = ifa->rt;
3921 ifa->rt = NULL;
3922 } else {
3923 state = ifa->state;
3924 ifa->state = INET6_IFADDR_STATE_DEAD;
3925 }
3926
3927 spin_unlock_bh(&ifa->lock);
3928
3929 if (rt)
3930 ip6_del_rt(net, rt, false);
3931
3932 if (state != INET6_IFADDR_STATE_DEAD) {
3933 __ipv6_ifa_notify(RTM_DELADDR, ifa);
3934 inet6addr_notifier_call_chain(NETDEV_DOWN, ifa);
3935 } else {
3936 if (idev->cnf.forwarding)
3937 addrconf_leave_anycast(ifa);
3938 addrconf_leave_solict(ifa->idev, &ifa->addr);
3939 }
3940
3941 if (!keep) {
3942 write_lock_bh(&idev->lock);
3943 list_del_rcu(&ifa->if_list);
3944 write_unlock_bh(&idev->lock);
3945 in6_ifa_put(ifa);
3946 }
3947 }
3948
3949 /* Step 5: Discard anycast and multicast list */
3950 if (unregister) {
3951 ipv6_ac_destroy_dev(idev);
3952 ipv6_mc_destroy_dev(idev);
3953 } else if (was_ready) {
3954 ipv6_mc_down(idev);
3955 }
3956
3957 idev->tstamp = jiffies;
3958 idev->ra_mtu = 0;
3959
3960 /* Last: Shot the device (if unregistered) */
3961 if (unregister) {
3962 addrconf_sysctl_unregister(idev);
3963 neigh_parms_release(&nd_tbl, idev->nd_parms);
3964 neigh_ifdown(&nd_tbl, dev);
3965 in6_dev_put(idev);
3966 }
3967 return 0;
3968 }
3969
addrconf_rs_timer(struct timer_list * t)3970 static void addrconf_rs_timer(struct timer_list *t)
3971 {
3972 struct inet6_dev *idev = from_timer(idev, t, rs_timer);
3973 struct net_device *dev = idev->dev;
3974 struct in6_addr lladdr;
3975
3976 write_lock(&idev->lock);
3977 if (idev->dead || !(idev->if_flags & IF_READY))
3978 goto out;
3979
3980 if (!ipv6_accept_ra(idev))
3981 goto out;
3982
3983 /* Announcement received after solicitation was sent */
3984 if (idev->if_flags & IF_RA_RCVD)
3985 goto out;
3986
3987 if (idev->rs_probes++ < idev->cnf.rtr_solicits || idev->cnf.rtr_solicits < 0) {
3988 write_unlock(&idev->lock);
3989 if (!ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE))
3990 ndisc_send_rs(dev, &lladdr,
3991 &in6addr_linklocal_allrouters);
3992 else
3993 goto put;
3994
3995 write_lock(&idev->lock);
3996 idev->rs_interval = rfc3315_s14_backoff_update(
3997 idev->rs_interval, idev->cnf.rtr_solicit_max_interval);
3998 /* The wait after the last probe can be shorter */
3999 addrconf_mod_rs_timer(idev, (idev->rs_probes ==
4000 idev->cnf.rtr_solicits) ?
4001 idev->cnf.rtr_solicit_delay :
4002 idev->rs_interval);
4003 } else {
4004 /*
4005 * Note: we do not support deprecated "all on-link"
4006 * assumption any longer.
4007 */
4008 pr_debug("%s: no IPv6 routers present\n", idev->dev->name);
4009 }
4010
4011 out:
4012 write_unlock(&idev->lock);
4013 put:
4014 in6_dev_put(idev);
4015 }
4016
4017 /*
4018 * Duplicate Address Detection
4019 */
addrconf_dad_kick(struct inet6_ifaddr * ifp)4020 static void addrconf_dad_kick(struct inet6_ifaddr *ifp)
4021 {
4022 unsigned long rand_num;
4023 struct inet6_dev *idev = ifp->idev;
4024 u64 nonce;
4025
4026 if (ifp->flags & IFA_F_OPTIMISTIC)
4027 rand_num = 0;
4028 else
4029 rand_num = get_random_u32_below(idev->cnf.rtr_solicit_delay ? : 1);
4030
4031 nonce = 0;
4032 if (idev->cnf.enhanced_dad ||
4033 dev_net(idev->dev)->ipv6.devconf_all->enhanced_dad) {
4034 do
4035 get_random_bytes(&nonce, 6);
4036 while (nonce == 0);
4037 }
4038 ifp->dad_nonce = nonce;
4039 ifp->dad_probes = idev->cnf.dad_transmits;
4040 addrconf_mod_dad_work(ifp, rand_num);
4041 }
4042
addrconf_dad_begin(struct inet6_ifaddr * ifp)4043 static void addrconf_dad_begin(struct inet6_ifaddr *ifp)
4044 {
4045 struct inet6_dev *idev = ifp->idev;
4046 struct net_device *dev = idev->dev;
4047 bool bump_id, notify = false;
4048 struct net *net;
4049
4050 addrconf_join_solict(dev, &ifp->addr);
4051
4052 read_lock_bh(&idev->lock);
4053 spin_lock(&ifp->lock);
4054 if (ifp->state == INET6_IFADDR_STATE_DEAD)
4055 goto out;
4056
4057 net = dev_net(dev);
4058 if (dev->flags&(IFF_NOARP|IFF_LOOPBACK) ||
4059 (net->ipv6.devconf_all->accept_dad < 1 &&
4060 idev->cnf.accept_dad < 1) ||
4061 !(ifp->flags&IFA_F_TENTATIVE) ||
4062 ifp->flags & IFA_F_NODAD) {
4063 bool send_na = false;
4064
4065 if (ifp->flags & IFA_F_TENTATIVE &&
4066 !(ifp->flags & IFA_F_OPTIMISTIC))
4067 send_na = true;
4068 bump_id = ifp->flags & IFA_F_TENTATIVE;
4069 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
4070 spin_unlock(&ifp->lock);
4071 read_unlock_bh(&idev->lock);
4072
4073 addrconf_dad_completed(ifp, bump_id, send_na);
4074 return;
4075 }
4076
4077 if (!(idev->if_flags & IF_READY)) {
4078 spin_unlock(&ifp->lock);
4079 read_unlock_bh(&idev->lock);
4080 /*
4081 * If the device is not ready:
4082 * - keep it tentative if it is a permanent address.
4083 * - otherwise, kill it.
4084 */
4085 in6_ifa_hold(ifp);
4086 addrconf_dad_stop(ifp, 0);
4087 return;
4088 }
4089
4090 /*
4091 * Optimistic nodes can start receiving
4092 * Frames right away
4093 */
4094 if (ifp->flags & IFA_F_OPTIMISTIC) {
4095 ip6_ins_rt(net, ifp->rt);
4096 if (ipv6_use_optimistic_addr(net, idev)) {
4097 /* Because optimistic nodes can use this address,
4098 * notify listeners. If DAD fails, RTM_DELADDR is sent.
4099 */
4100 notify = true;
4101 }
4102 }
4103
4104 addrconf_dad_kick(ifp);
4105 out:
4106 spin_unlock(&ifp->lock);
4107 read_unlock_bh(&idev->lock);
4108 if (notify)
4109 ipv6_ifa_notify(RTM_NEWADDR, ifp);
4110 }
4111
addrconf_dad_start(struct inet6_ifaddr * ifp)4112 static void addrconf_dad_start(struct inet6_ifaddr *ifp)
4113 {
4114 bool begin_dad = false;
4115
4116 spin_lock_bh(&ifp->lock);
4117 if (ifp->state != INET6_IFADDR_STATE_DEAD) {
4118 ifp->state = INET6_IFADDR_STATE_PREDAD;
4119 begin_dad = true;
4120 }
4121 spin_unlock_bh(&ifp->lock);
4122
4123 if (begin_dad)
4124 addrconf_mod_dad_work(ifp, 0);
4125 }
4126
addrconf_dad_work(struct work_struct * w)4127 static void addrconf_dad_work(struct work_struct *w)
4128 {
4129 struct inet6_ifaddr *ifp = container_of(to_delayed_work(w),
4130 struct inet6_ifaddr,
4131 dad_work);
4132 struct inet6_dev *idev = ifp->idev;
4133 bool bump_id, disable_ipv6 = false;
4134 struct in6_addr mcaddr;
4135
4136 enum {
4137 DAD_PROCESS,
4138 DAD_BEGIN,
4139 DAD_ABORT,
4140 } action = DAD_PROCESS;
4141
4142 rtnl_lock();
4143
4144 spin_lock_bh(&ifp->lock);
4145 if (ifp->state == INET6_IFADDR_STATE_PREDAD) {
4146 action = DAD_BEGIN;
4147 ifp->state = INET6_IFADDR_STATE_DAD;
4148 } else if (ifp->state == INET6_IFADDR_STATE_ERRDAD) {
4149 action = DAD_ABORT;
4150 ifp->state = INET6_IFADDR_STATE_POSTDAD;
4151
4152 if ((dev_net(idev->dev)->ipv6.devconf_all->accept_dad > 1 ||
4153 idev->cnf.accept_dad > 1) &&
4154 !idev->cnf.disable_ipv6 &&
4155 !(ifp->flags & IFA_F_STABLE_PRIVACY)) {
4156 struct in6_addr addr;
4157
4158 addr.s6_addr32[0] = htonl(0xfe800000);
4159 addr.s6_addr32[1] = 0;
4160
4161 if (!ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) &&
4162 ipv6_addr_equal(&ifp->addr, &addr)) {
4163 /* DAD failed for link-local based on MAC */
4164 WRITE_ONCE(idev->cnf.disable_ipv6, 1);
4165
4166 pr_info("%s: IPv6 being disabled!\n",
4167 ifp->idev->dev->name);
4168 disable_ipv6 = true;
4169 }
4170 }
4171 }
4172 spin_unlock_bh(&ifp->lock);
4173
4174 if (action == DAD_BEGIN) {
4175 addrconf_dad_begin(ifp);
4176 goto out;
4177 } else if (action == DAD_ABORT) {
4178 in6_ifa_hold(ifp);
4179 addrconf_dad_stop(ifp, 1);
4180 if (disable_ipv6)
4181 addrconf_ifdown(idev->dev, false);
4182 goto out;
4183 }
4184
4185 if (!ifp->dad_probes && addrconf_dad_end(ifp))
4186 goto out;
4187
4188 write_lock_bh(&idev->lock);
4189 if (idev->dead || !(idev->if_flags & IF_READY)) {
4190 write_unlock_bh(&idev->lock);
4191 goto out;
4192 }
4193
4194 spin_lock(&ifp->lock);
4195 if (ifp->state == INET6_IFADDR_STATE_DEAD) {
4196 spin_unlock(&ifp->lock);
4197 write_unlock_bh(&idev->lock);
4198 goto out;
4199 }
4200
4201 if (ifp->dad_probes == 0) {
4202 bool send_na = false;
4203
4204 /*
4205 * DAD was successful
4206 */
4207
4208 if (ifp->flags & IFA_F_TENTATIVE &&
4209 !(ifp->flags & IFA_F_OPTIMISTIC))
4210 send_na = true;
4211 bump_id = ifp->flags & IFA_F_TENTATIVE;
4212 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
4213 spin_unlock(&ifp->lock);
4214 write_unlock_bh(&idev->lock);
4215
4216 addrconf_dad_completed(ifp, bump_id, send_na);
4217
4218 goto out;
4219 }
4220
4221 ifp->dad_probes--;
4222 addrconf_mod_dad_work(ifp,
4223 max(NEIGH_VAR(ifp->idev->nd_parms, RETRANS_TIME),
4224 HZ/100));
4225 spin_unlock(&ifp->lock);
4226 write_unlock_bh(&idev->lock);
4227
4228 /* send a neighbour solicitation for our addr */
4229 addrconf_addr_solict_mult(&ifp->addr, &mcaddr);
4230 ndisc_send_ns(ifp->idev->dev, &ifp->addr, &mcaddr, &in6addr_any,
4231 ifp->dad_nonce);
4232 out:
4233 in6_ifa_put(ifp);
4234 rtnl_unlock();
4235 }
4236
4237 /* ifp->idev must be at least read locked */
ipv6_lonely_lladdr(struct inet6_ifaddr * ifp)4238 static bool ipv6_lonely_lladdr(struct inet6_ifaddr *ifp)
4239 {
4240 struct inet6_ifaddr *ifpiter;
4241 struct inet6_dev *idev = ifp->idev;
4242
4243 list_for_each_entry_reverse(ifpiter, &idev->addr_list, if_list) {
4244 if (ifpiter->scope > IFA_LINK)
4245 break;
4246 if (ifp != ifpiter && ifpiter->scope == IFA_LINK &&
4247 (ifpiter->flags & (IFA_F_PERMANENT|IFA_F_TENTATIVE|
4248 IFA_F_OPTIMISTIC|IFA_F_DADFAILED)) ==
4249 IFA_F_PERMANENT)
4250 return false;
4251 }
4252 return true;
4253 }
4254
addrconf_dad_completed(struct inet6_ifaddr * ifp,bool bump_id,bool send_na)4255 static void addrconf_dad_completed(struct inet6_ifaddr *ifp, bool bump_id,
4256 bool send_na)
4257 {
4258 struct net_device *dev = ifp->idev->dev;
4259 struct in6_addr lladdr;
4260 bool send_rs, send_mld;
4261
4262 addrconf_del_dad_work(ifp);
4263
4264 /*
4265 * Configure the address for reception. Now it is valid.
4266 */
4267
4268 ipv6_ifa_notify(RTM_NEWADDR, ifp);
4269
4270 /* If added prefix is link local and we are prepared to process
4271 router advertisements, start sending router solicitations.
4272 */
4273
4274 read_lock_bh(&ifp->idev->lock);
4275 send_mld = ifp->scope == IFA_LINK && ipv6_lonely_lladdr(ifp);
4276 send_rs = send_mld &&
4277 ipv6_accept_ra(ifp->idev) &&
4278 ifp->idev->cnf.rtr_solicits != 0 &&
4279 (dev->flags & IFF_LOOPBACK) == 0 &&
4280 (dev->type != ARPHRD_TUNNEL) &&
4281 !netif_is_team_port(dev);
4282 read_unlock_bh(&ifp->idev->lock);
4283
4284 /* While dad is in progress mld report's source address is in6_addrany.
4285 * Resend with proper ll now.
4286 */
4287 if (send_mld)
4288 ipv6_mc_dad_complete(ifp->idev);
4289
4290 /* send unsolicited NA if enabled */
4291 if (send_na &&
4292 (ifp->idev->cnf.ndisc_notify ||
4293 dev_net(dev)->ipv6.devconf_all->ndisc_notify)) {
4294 ndisc_send_na(dev, &in6addr_linklocal_allnodes, &ifp->addr,
4295 /*router=*/ !!ifp->idev->cnf.forwarding,
4296 /*solicited=*/ false, /*override=*/ true,
4297 /*inc_opt=*/ true);
4298 }
4299
4300 if (send_rs) {
4301 /*
4302 * If a host as already performed a random delay
4303 * [...] as part of DAD [...] there is no need
4304 * to delay again before sending the first RS
4305 */
4306 if (ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE))
4307 return;
4308 ndisc_send_rs(dev, &lladdr, &in6addr_linklocal_allrouters);
4309
4310 write_lock_bh(&ifp->idev->lock);
4311 spin_lock(&ifp->lock);
4312 ifp->idev->rs_interval = rfc3315_s14_backoff_init(
4313 ifp->idev->cnf.rtr_solicit_interval);
4314 ifp->idev->rs_probes = 1;
4315 ifp->idev->if_flags |= IF_RS_SENT;
4316 addrconf_mod_rs_timer(ifp->idev, ifp->idev->rs_interval);
4317 spin_unlock(&ifp->lock);
4318 write_unlock_bh(&ifp->idev->lock);
4319 }
4320
4321 if (bump_id)
4322 rt_genid_bump_ipv6(dev_net(dev));
4323
4324 /* Make sure that a new temporary address will be created
4325 * before this temporary address becomes deprecated.
4326 */
4327 if (ifp->flags & IFA_F_TEMPORARY)
4328 addrconf_verify_rtnl(dev_net(dev));
4329 }
4330
addrconf_dad_run(struct inet6_dev * idev,bool restart)4331 static void addrconf_dad_run(struct inet6_dev *idev, bool restart)
4332 {
4333 struct inet6_ifaddr *ifp;
4334
4335 read_lock_bh(&idev->lock);
4336 list_for_each_entry(ifp, &idev->addr_list, if_list) {
4337 spin_lock(&ifp->lock);
4338 if ((ifp->flags & IFA_F_TENTATIVE &&
4339 ifp->state == INET6_IFADDR_STATE_DAD) || restart) {
4340 if (restart)
4341 ifp->state = INET6_IFADDR_STATE_PREDAD;
4342 addrconf_dad_kick(ifp);
4343 }
4344 spin_unlock(&ifp->lock);
4345 }
4346 read_unlock_bh(&idev->lock);
4347 }
4348
4349 #ifdef CONFIG_PROC_FS
4350 struct if6_iter_state {
4351 struct seq_net_private p;
4352 int bucket;
4353 int offset;
4354 };
4355
if6_get_first(struct seq_file * seq,loff_t pos)4356 static struct inet6_ifaddr *if6_get_first(struct seq_file *seq, loff_t pos)
4357 {
4358 struct if6_iter_state *state = seq->private;
4359 struct net *net = seq_file_net(seq);
4360 struct inet6_ifaddr *ifa = NULL;
4361 int p = 0;
4362
4363 /* initial bucket if pos is 0 */
4364 if (pos == 0) {
4365 state->bucket = 0;
4366 state->offset = 0;
4367 }
4368
4369 for (; state->bucket < IN6_ADDR_HSIZE; ++state->bucket) {
4370 hlist_for_each_entry_rcu(ifa, &net->ipv6.inet6_addr_lst[state->bucket],
4371 addr_lst) {
4372 /* sync with offset */
4373 if (p < state->offset) {
4374 p++;
4375 continue;
4376 }
4377 return ifa;
4378 }
4379
4380 /* prepare for next bucket */
4381 state->offset = 0;
4382 p = 0;
4383 }
4384 return NULL;
4385 }
4386
if6_get_next(struct seq_file * seq,struct inet6_ifaddr * ifa)4387 static struct inet6_ifaddr *if6_get_next(struct seq_file *seq,
4388 struct inet6_ifaddr *ifa)
4389 {
4390 struct if6_iter_state *state = seq->private;
4391 struct net *net = seq_file_net(seq);
4392
4393 hlist_for_each_entry_continue_rcu(ifa, addr_lst) {
4394 state->offset++;
4395 return ifa;
4396 }
4397
4398 state->offset = 0;
4399 while (++state->bucket < IN6_ADDR_HSIZE) {
4400 hlist_for_each_entry_rcu(ifa,
4401 &net->ipv6.inet6_addr_lst[state->bucket], addr_lst) {
4402 return ifa;
4403 }
4404 }
4405
4406 return NULL;
4407 }
4408
if6_seq_start(struct seq_file * seq,loff_t * pos)4409 static void *if6_seq_start(struct seq_file *seq, loff_t *pos)
4410 __acquires(rcu)
4411 {
4412 rcu_read_lock();
4413 return if6_get_first(seq, *pos);
4414 }
4415
if6_seq_next(struct seq_file * seq,void * v,loff_t * pos)4416 static void *if6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
4417 {
4418 struct inet6_ifaddr *ifa;
4419
4420 ifa = if6_get_next(seq, v);
4421 ++*pos;
4422 return ifa;
4423 }
4424
if6_seq_stop(struct seq_file * seq,void * v)4425 static void if6_seq_stop(struct seq_file *seq, void *v)
4426 __releases(rcu)
4427 {
4428 rcu_read_unlock();
4429 }
4430
if6_seq_show(struct seq_file * seq,void * v)4431 static int if6_seq_show(struct seq_file *seq, void *v)
4432 {
4433 struct inet6_ifaddr *ifp = (struct inet6_ifaddr *)v;
4434 seq_printf(seq, "%pi6 %02x %02x %02x %02x %8s\n",
4435 &ifp->addr,
4436 ifp->idev->dev->ifindex,
4437 ifp->prefix_len,
4438 ifp->scope,
4439 (u8) ifp->flags,
4440 ifp->idev->dev->name);
4441 return 0;
4442 }
4443
4444 static const struct seq_operations if6_seq_ops = {
4445 .start = if6_seq_start,
4446 .next = if6_seq_next,
4447 .show = if6_seq_show,
4448 .stop = if6_seq_stop,
4449 };
4450
if6_proc_net_init(struct net * net)4451 static int __net_init if6_proc_net_init(struct net *net)
4452 {
4453 if (!proc_create_net("if_inet6", 0444, net->proc_net, &if6_seq_ops,
4454 sizeof(struct if6_iter_state)))
4455 return -ENOMEM;
4456 return 0;
4457 }
4458
if6_proc_net_exit(struct net * net)4459 static void __net_exit if6_proc_net_exit(struct net *net)
4460 {
4461 remove_proc_entry("if_inet6", net->proc_net);
4462 }
4463
4464 static struct pernet_operations if6_proc_net_ops = {
4465 .init = if6_proc_net_init,
4466 .exit = if6_proc_net_exit,
4467 };
4468
if6_proc_init(void)4469 int __init if6_proc_init(void)
4470 {
4471 return register_pernet_subsys(&if6_proc_net_ops);
4472 }
4473
if6_proc_exit(void)4474 void if6_proc_exit(void)
4475 {
4476 unregister_pernet_subsys(&if6_proc_net_ops);
4477 }
4478 #endif /* CONFIG_PROC_FS */
4479
4480 #if IS_ENABLED(CONFIG_IPV6_MIP6)
4481 /* Check if address is a home address configured on any interface. */
ipv6_chk_home_addr(struct net * net,const struct in6_addr * addr)4482 int ipv6_chk_home_addr(struct net *net, const struct in6_addr *addr)
4483 {
4484 unsigned int hash = inet6_addr_hash(net, addr);
4485 struct inet6_ifaddr *ifp = NULL;
4486 int ret = 0;
4487
4488 rcu_read_lock();
4489 hlist_for_each_entry_rcu(ifp, &net->ipv6.inet6_addr_lst[hash], addr_lst) {
4490 if (ipv6_addr_equal(&ifp->addr, addr) &&
4491 (ifp->flags & IFA_F_HOMEADDRESS)) {
4492 ret = 1;
4493 break;
4494 }
4495 }
4496 rcu_read_unlock();
4497 return ret;
4498 }
4499 #endif
4500
4501 /* RFC6554 has some algorithm to avoid loops in segment routing by
4502 * checking if the segments contains any of a local interface address.
4503 *
4504 * Quote:
4505 *
4506 * To detect loops in the SRH, a router MUST determine if the SRH
4507 * includes multiple addresses assigned to any interface on that router.
4508 * If such addresses appear more than once and are separated by at least
4509 * one address not assigned to that router.
4510 */
ipv6_chk_rpl_srh_loop(struct net * net,const struct in6_addr * segs,unsigned char nsegs)4511 int ipv6_chk_rpl_srh_loop(struct net *net, const struct in6_addr *segs,
4512 unsigned char nsegs)
4513 {
4514 const struct in6_addr *addr;
4515 int i, ret = 0, found = 0;
4516 struct inet6_ifaddr *ifp;
4517 bool separated = false;
4518 unsigned int hash;
4519 bool hash_found;
4520
4521 rcu_read_lock();
4522 for (i = 0; i < nsegs; i++) {
4523 addr = &segs[i];
4524 hash = inet6_addr_hash(net, addr);
4525
4526 hash_found = false;
4527 hlist_for_each_entry_rcu(ifp, &net->ipv6.inet6_addr_lst[hash], addr_lst) {
4528
4529 if (ipv6_addr_equal(&ifp->addr, addr)) {
4530 hash_found = true;
4531 break;
4532 }
4533 }
4534
4535 if (hash_found) {
4536 if (found > 1 && separated) {
4537 ret = 1;
4538 break;
4539 }
4540
4541 separated = false;
4542 found++;
4543 } else {
4544 separated = true;
4545 }
4546 }
4547 rcu_read_unlock();
4548
4549 return ret;
4550 }
4551
4552 /*
4553 * Periodic address status verification
4554 */
4555
addrconf_verify_rtnl(struct net * net)4556 static void addrconf_verify_rtnl(struct net *net)
4557 {
4558 unsigned long now, next, next_sec, next_sched;
4559 struct inet6_ifaddr *ifp;
4560 int i;
4561
4562 ASSERT_RTNL();
4563
4564 rcu_read_lock_bh();
4565 now = jiffies;
4566 next = round_jiffies_up(now + ADDR_CHECK_FREQUENCY);
4567
4568 cancel_delayed_work(&net->ipv6.addr_chk_work);
4569
4570 for (i = 0; i < IN6_ADDR_HSIZE; i++) {
4571 restart:
4572 hlist_for_each_entry_rcu_bh(ifp, &net->ipv6.inet6_addr_lst[i], addr_lst) {
4573 unsigned long age;
4574
4575 /* When setting preferred_lft to a value not zero or
4576 * infinity, while valid_lft is infinity
4577 * IFA_F_PERMANENT has a non-infinity life time.
4578 */
4579 if ((ifp->flags & IFA_F_PERMANENT) &&
4580 (ifp->prefered_lft == INFINITY_LIFE_TIME))
4581 continue;
4582
4583 spin_lock(&ifp->lock);
4584 /* We try to batch several events at once. */
4585 age = (now - ifp->tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
4586
4587 if ((ifp->flags&IFA_F_TEMPORARY) &&
4588 !(ifp->flags&IFA_F_TENTATIVE) &&
4589 ifp->prefered_lft != INFINITY_LIFE_TIME &&
4590 !ifp->regen_count && ifp->ifpub) {
4591 /* This is a non-regenerated temporary addr. */
4592
4593 unsigned long regen_advance = ifp->idev->cnf.regen_max_retry *
4594 ifp->idev->cnf.dad_transmits *
4595 max(NEIGH_VAR(ifp->idev->nd_parms, RETRANS_TIME), HZ/100) / HZ;
4596
4597 if (age + regen_advance >= ifp->prefered_lft) {
4598 struct inet6_ifaddr *ifpub = ifp->ifpub;
4599 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
4600 next = ifp->tstamp + ifp->prefered_lft * HZ;
4601
4602 ifp->regen_count++;
4603 in6_ifa_hold(ifp);
4604 in6_ifa_hold(ifpub);
4605 spin_unlock(&ifp->lock);
4606
4607 spin_lock(&ifpub->lock);
4608 ifpub->regen_count = 0;
4609 spin_unlock(&ifpub->lock);
4610 rcu_read_unlock_bh();
4611 ipv6_create_tempaddr(ifpub, true);
4612 in6_ifa_put(ifpub);
4613 in6_ifa_put(ifp);
4614 rcu_read_lock_bh();
4615 goto restart;
4616 } else if (time_before(ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ, next))
4617 next = ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ;
4618 }
4619
4620 if (ifp->valid_lft != INFINITY_LIFE_TIME &&
4621 age >= ifp->valid_lft) {
4622 spin_unlock(&ifp->lock);
4623 in6_ifa_hold(ifp);
4624 rcu_read_unlock_bh();
4625 ipv6_del_addr(ifp);
4626 rcu_read_lock_bh();
4627 goto restart;
4628 } else if (ifp->prefered_lft == INFINITY_LIFE_TIME) {
4629 spin_unlock(&ifp->lock);
4630 continue;
4631 } else if (age >= ifp->prefered_lft) {
4632 /* jiffies - ifp->tstamp > age >= ifp->prefered_lft */
4633 int deprecate = 0;
4634
4635 if (!(ifp->flags&IFA_F_DEPRECATED)) {
4636 deprecate = 1;
4637 ifp->flags |= IFA_F_DEPRECATED;
4638 }
4639
4640 if ((ifp->valid_lft != INFINITY_LIFE_TIME) &&
4641 (time_before(ifp->tstamp + ifp->valid_lft * HZ, next)))
4642 next = ifp->tstamp + ifp->valid_lft * HZ;
4643
4644 spin_unlock(&ifp->lock);
4645
4646 if (deprecate) {
4647 in6_ifa_hold(ifp);
4648
4649 ipv6_ifa_notify(0, ifp);
4650 in6_ifa_put(ifp);
4651 goto restart;
4652 }
4653 } else {
4654 /* ifp->prefered_lft <= ifp->valid_lft */
4655 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
4656 next = ifp->tstamp + ifp->prefered_lft * HZ;
4657 spin_unlock(&ifp->lock);
4658 }
4659 }
4660 }
4661
4662 next_sec = round_jiffies_up(next);
4663 next_sched = next;
4664
4665 /* If rounded timeout is accurate enough, accept it. */
4666 if (time_before(next_sec, next + ADDRCONF_TIMER_FUZZ))
4667 next_sched = next_sec;
4668
4669 /* And minimum interval is ADDRCONF_TIMER_FUZZ_MAX. */
4670 if (time_before(next_sched, jiffies + ADDRCONF_TIMER_FUZZ_MAX))
4671 next_sched = jiffies + ADDRCONF_TIMER_FUZZ_MAX;
4672
4673 pr_debug("now = %lu, schedule = %lu, rounded schedule = %lu => %lu\n",
4674 now, next, next_sec, next_sched);
4675 mod_delayed_work(addrconf_wq, &net->ipv6.addr_chk_work, next_sched - now);
4676 rcu_read_unlock_bh();
4677 }
4678
addrconf_verify_work(struct work_struct * w)4679 static void addrconf_verify_work(struct work_struct *w)
4680 {
4681 struct net *net = container_of(to_delayed_work(w), struct net,
4682 ipv6.addr_chk_work);
4683
4684 rtnl_lock();
4685 addrconf_verify_rtnl(net);
4686 rtnl_unlock();
4687 }
4688
addrconf_verify(struct net * net)4689 static void addrconf_verify(struct net *net)
4690 {
4691 mod_delayed_work(addrconf_wq, &net->ipv6.addr_chk_work, 0);
4692 }
4693
extract_addr(struct nlattr * addr,struct nlattr * local,struct in6_addr ** peer_pfx)4694 static struct in6_addr *extract_addr(struct nlattr *addr, struct nlattr *local,
4695 struct in6_addr **peer_pfx)
4696 {
4697 struct in6_addr *pfx = NULL;
4698
4699 *peer_pfx = NULL;
4700
4701 if (addr)
4702 pfx = nla_data(addr);
4703
4704 if (local) {
4705 if (pfx && nla_memcmp(local, pfx, sizeof(*pfx)))
4706 *peer_pfx = pfx;
4707 pfx = nla_data(local);
4708 }
4709
4710 return pfx;
4711 }
4712
4713 static const struct nla_policy ifa_ipv6_policy[IFA_MAX+1] = {
4714 [IFA_ADDRESS] = { .len = sizeof(struct in6_addr) },
4715 [IFA_LOCAL] = { .len = sizeof(struct in6_addr) },
4716 [IFA_CACHEINFO] = { .len = sizeof(struct ifa_cacheinfo) },
4717 [IFA_FLAGS] = { .len = sizeof(u32) },
4718 [IFA_RT_PRIORITY] = { .len = sizeof(u32) },
4719 [IFA_TARGET_NETNSID] = { .type = NLA_S32 },
4720 [IFA_PROTO] = { .type = NLA_U8 },
4721 };
4722
4723 static int
inet6_rtm_deladdr(struct sk_buff * skb,struct nlmsghdr * nlh,struct netlink_ext_ack * extack)4724 inet6_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh,
4725 struct netlink_ext_ack *extack)
4726 {
4727 struct net *net = sock_net(skb->sk);
4728 struct ifaddrmsg *ifm;
4729 struct nlattr *tb[IFA_MAX+1];
4730 struct in6_addr *pfx, *peer_pfx;
4731 u32 ifa_flags;
4732 int err;
4733
4734 err = nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFA_MAX,
4735 ifa_ipv6_policy, extack);
4736 if (err < 0)
4737 return err;
4738
4739 ifm = nlmsg_data(nlh);
4740 pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
4741 if (!pfx)
4742 return -EINVAL;
4743
4744 ifa_flags = tb[IFA_FLAGS] ? nla_get_u32(tb[IFA_FLAGS]) : ifm->ifa_flags;
4745
4746 /* We ignore other flags so far. */
4747 ifa_flags &= IFA_F_MANAGETEMPADDR;
4748
4749 return inet6_addr_del(net, ifm->ifa_index, ifa_flags, pfx,
4750 ifm->ifa_prefixlen, extack);
4751 }
4752
modify_prefix_route(struct inet6_ifaddr * ifp,unsigned long expires,u32 flags,bool modify_peer)4753 static int modify_prefix_route(struct inet6_ifaddr *ifp,
4754 unsigned long expires, u32 flags,
4755 bool modify_peer)
4756 {
4757 struct fib6_info *f6i;
4758 u32 prio;
4759
4760 f6i = addrconf_get_prefix_route(modify_peer ? &ifp->peer_addr : &ifp->addr,
4761 ifp->prefix_len,
4762 ifp->idev->dev, 0, RTF_DEFAULT, true);
4763 if (!f6i)
4764 return -ENOENT;
4765
4766 prio = ifp->rt_priority ? : IP6_RT_PRIO_ADDRCONF;
4767 if (f6i->fib6_metric != prio) {
4768 /* delete old one */
4769 ip6_del_rt(dev_net(ifp->idev->dev), f6i, false);
4770
4771 /* add new one */
4772 addrconf_prefix_route(modify_peer ? &ifp->peer_addr : &ifp->addr,
4773 ifp->prefix_len,
4774 ifp->rt_priority, ifp->idev->dev,
4775 expires, flags, GFP_KERNEL);
4776 } else {
4777 if (!expires)
4778 fib6_clean_expires(f6i);
4779 else
4780 fib6_set_expires(f6i, expires);
4781
4782 fib6_info_release(f6i);
4783 }
4784
4785 return 0;
4786 }
4787
inet6_addr_modify(struct net * net,struct inet6_ifaddr * ifp,struct ifa6_config * cfg)4788 static int inet6_addr_modify(struct net *net, struct inet6_ifaddr *ifp,
4789 struct ifa6_config *cfg)
4790 {
4791 u32 flags;
4792 clock_t expires;
4793 unsigned long timeout;
4794 bool was_managetempaddr;
4795 bool had_prefixroute;
4796 bool new_peer = false;
4797
4798 ASSERT_RTNL();
4799
4800 if (!cfg->valid_lft || cfg->preferred_lft > cfg->valid_lft)
4801 return -EINVAL;
4802
4803 if (cfg->ifa_flags & IFA_F_MANAGETEMPADDR &&
4804 (ifp->flags & IFA_F_TEMPORARY || ifp->prefix_len != 64))
4805 return -EINVAL;
4806
4807 if (!(ifp->flags & IFA_F_TENTATIVE) || ifp->flags & IFA_F_DADFAILED)
4808 cfg->ifa_flags &= ~IFA_F_OPTIMISTIC;
4809
4810 timeout = addrconf_timeout_fixup(cfg->valid_lft, HZ);
4811 if (addrconf_finite_timeout(timeout)) {
4812 expires = jiffies_to_clock_t(timeout * HZ);
4813 cfg->valid_lft = timeout;
4814 flags = RTF_EXPIRES;
4815 } else {
4816 expires = 0;
4817 flags = 0;
4818 cfg->ifa_flags |= IFA_F_PERMANENT;
4819 }
4820
4821 timeout = addrconf_timeout_fixup(cfg->preferred_lft, HZ);
4822 if (addrconf_finite_timeout(timeout)) {
4823 if (timeout == 0)
4824 cfg->ifa_flags |= IFA_F_DEPRECATED;
4825 cfg->preferred_lft = timeout;
4826 }
4827
4828 if (cfg->peer_pfx &&
4829 memcmp(&ifp->peer_addr, cfg->peer_pfx, sizeof(struct in6_addr))) {
4830 if (!ipv6_addr_any(&ifp->peer_addr))
4831 cleanup_prefix_route(ifp, expires, true, true);
4832 new_peer = true;
4833 }
4834
4835 spin_lock_bh(&ifp->lock);
4836 was_managetempaddr = ifp->flags & IFA_F_MANAGETEMPADDR;
4837 had_prefixroute = ifp->flags & IFA_F_PERMANENT &&
4838 !(ifp->flags & IFA_F_NOPREFIXROUTE);
4839 ifp->flags &= ~(IFA_F_DEPRECATED | IFA_F_PERMANENT | IFA_F_NODAD |
4840 IFA_F_HOMEADDRESS | IFA_F_MANAGETEMPADDR |
4841 IFA_F_NOPREFIXROUTE);
4842 ifp->flags |= cfg->ifa_flags;
4843 ifp->tstamp = jiffies;
4844 ifp->valid_lft = cfg->valid_lft;
4845 ifp->prefered_lft = cfg->preferred_lft;
4846 ifp->ifa_proto = cfg->ifa_proto;
4847
4848 if (cfg->rt_priority && cfg->rt_priority != ifp->rt_priority)
4849 ifp->rt_priority = cfg->rt_priority;
4850
4851 if (new_peer)
4852 ifp->peer_addr = *cfg->peer_pfx;
4853
4854 spin_unlock_bh(&ifp->lock);
4855 if (!(ifp->flags&IFA_F_TENTATIVE))
4856 ipv6_ifa_notify(0, ifp);
4857
4858 if (!(cfg->ifa_flags & IFA_F_NOPREFIXROUTE)) {
4859 int rc = -ENOENT;
4860
4861 if (had_prefixroute)
4862 rc = modify_prefix_route(ifp, expires, flags, false);
4863
4864 /* prefix route could have been deleted; if so restore it */
4865 if (rc == -ENOENT) {
4866 addrconf_prefix_route(&ifp->addr, ifp->prefix_len,
4867 ifp->rt_priority, ifp->idev->dev,
4868 expires, flags, GFP_KERNEL);
4869 }
4870
4871 if (had_prefixroute && !ipv6_addr_any(&ifp->peer_addr))
4872 rc = modify_prefix_route(ifp, expires, flags, true);
4873
4874 if (rc == -ENOENT && !ipv6_addr_any(&ifp->peer_addr)) {
4875 addrconf_prefix_route(&ifp->peer_addr, ifp->prefix_len,
4876 ifp->rt_priority, ifp->idev->dev,
4877 expires, flags, GFP_KERNEL);
4878 }
4879 } else if (had_prefixroute) {
4880 enum cleanup_prefix_rt_t action;
4881 unsigned long rt_expires;
4882
4883 write_lock_bh(&ifp->idev->lock);
4884 action = check_cleanup_prefix_route(ifp, &rt_expires);
4885 write_unlock_bh(&ifp->idev->lock);
4886
4887 if (action != CLEANUP_PREFIX_RT_NOP) {
4888 cleanup_prefix_route(ifp, rt_expires,
4889 action == CLEANUP_PREFIX_RT_DEL, false);
4890 }
4891 }
4892
4893 if (was_managetempaddr || ifp->flags & IFA_F_MANAGETEMPADDR) {
4894 if (was_managetempaddr &&
4895 !(ifp->flags & IFA_F_MANAGETEMPADDR)) {
4896 cfg->valid_lft = 0;
4897 cfg->preferred_lft = 0;
4898 }
4899 manage_tempaddrs(ifp->idev, ifp, cfg->valid_lft,
4900 cfg->preferred_lft, !was_managetempaddr,
4901 jiffies);
4902 }
4903
4904 addrconf_verify_rtnl(net);
4905
4906 return 0;
4907 }
4908
4909 static int
inet6_rtm_newaddr(struct sk_buff * skb,struct nlmsghdr * nlh,struct netlink_ext_ack * extack)4910 inet6_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh,
4911 struct netlink_ext_ack *extack)
4912 {
4913 struct net *net = sock_net(skb->sk);
4914 struct ifaddrmsg *ifm;
4915 struct nlattr *tb[IFA_MAX+1];
4916 struct in6_addr *peer_pfx;
4917 struct inet6_ifaddr *ifa;
4918 struct net_device *dev;
4919 struct inet6_dev *idev;
4920 struct ifa6_config cfg;
4921 int err;
4922
4923 err = nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFA_MAX,
4924 ifa_ipv6_policy, extack);
4925 if (err < 0)
4926 return err;
4927
4928 memset(&cfg, 0, sizeof(cfg));
4929
4930 ifm = nlmsg_data(nlh);
4931 cfg.pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
4932 if (!cfg.pfx)
4933 return -EINVAL;
4934
4935 cfg.peer_pfx = peer_pfx;
4936 cfg.plen = ifm->ifa_prefixlen;
4937 if (tb[IFA_RT_PRIORITY])
4938 cfg.rt_priority = nla_get_u32(tb[IFA_RT_PRIORITY]);
4939
4940 if (tb[IFA_PROTO])
4941 cfg.ifa_proto = nla_get_u8(tb[IFA_PROTO]);
4942
4943 cfg.valid_lft = INFINITY_LIFE_TIME;
4944 cfg.preferred_lft = INFINITY_LIFE_TIME;
4945
4946 if (tb[IFA_CACHEINFO]) {
4947 struct ifa_cacheinfo *ci;
4948
4949 ci = nla_data(tb[IFA_CACHEINFO]);
4950 cfg.valid_lft = ci->ifa_valid;
4951 cfg.preferred_lft = ci->ifa_prefered;
4952 }
4953
4954 dev = __dev_get_by_index(net, ifm->ifa_index);
4955 if (!dev) {
4956 NL_SET_ERR_MSG_MOD(extack, "Unable to find the interface");
4957 return -ENODEV;
4958 }
4959
4960 if (tb[IFA_FLAGS])
4961 cfg.ifa_flags = nla_get_u32(tb[IFA_FLAGS]);
4962 else
4963 cfg.ifa_flags = ifm->ifa_flags;
4964
4965 /* We ignore other flags so far. */
4966 cfg.ifa_flags &= IFA_F_NODAD | IFA_F_HOMEADDRESS |
4967 IFA_F_MANAGETEMPADDR | IFA_F_NOPREFIXROUTE |
4968 IFA_F_MCAUTOJOIN | IFA_F_OPTIMISTIC;
4969
4970 idev = ipv6_find_idev(dev);
4971 if (IS_ERR(idev))
4972 return PTR_ERR(idev);
4973
4974 if (!ipv6_allow_optimistic_dad(net, idev))
4975 cfg.ifa_flags &= ~IFA_F_OPTIMISTIC;
4976
4977 if (cfg.ifa_flags & IFA_F_NODAD &&
4978 cfg.ifa_flags & IFA_F_OPTIMISTIC) {
4979 NL_SET_ERR_MSG(extack, "IFA_F_NODAD and IFA_F_OPTIMISTIC are mutually exclusive");
4980 return -EINVAL;
4981 }
4982
4983 ifa = ipv6_get_ifaddr(net, cfg.pfx, dev, 1);
4984 if (!ifa) {
4985 /*
4986 * It would be best to check for !NLM_F_CREATE here but
4987 * userspace already relies on not having to provide this.
4988 */
4989 return inet6_addr_add(net, ifm->ifa_index, &cfg, extack);
4990 }
4991
4992 if (nlh->nlmsg_flags & NLM_F_EXCL ||
4993 !(nlh->nlmsg_flags & NLM_F_REPLACE)) {
4994 NL_SET_ERR_MSG_MOD(extack, "address already assigned");
4995 err = -EEXIST;
4996 } else {
4997 err = inet6_addr_modify(net, ifa, &cfg);
4998 }
4999
5000 in6_ifa_put(ifa);
5001
5002 return err;
5003 }
5004
put_ifaddrmsg(struct nlmsghdr * nlh,u8 prefixlen,u32 flags,u8 scope,int ifindex)5005 static void put_ifaddrmsg(struct nlmsghdr *nlh, u8 prefixlen, u32 flags,
5006 u8 scope, int ifindex)
5007 {
5008 struct ifaddrmsg *ifm;
5009
5010 ifm = nlmsg_data(nlh);
5011 ifm->ifa_family = AF_INET6;
5012 ifm->ifa_prefixlen = prefixlen;
5013 ifm->ifa_flags = flags;
5014 ifm->ifa_scope = scope;
5015 ifm->ifa_index = ifindex;
5016 }
5017
put_cacheinfo(struct sk_buff * skb,unsigned long cstamp,unsigned long tstamp,u32 preferred,u32 valid)5018 static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
5019 unsigned long tstamp, u32 preferred, u32 valid)
5020 {
5021 struct ifa_cacheinfo ci;
5022
5023 ci.cstamp = cstamp_delta(cstamp);
5024 ci.tstamp = cstamp_delta(tstamp);
5025 ci.ifa_prefered = preferred;
5026 ci.ifa_valid = valid;
5027
5028 return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
5029 }
5030
rt_scope(int ifa_scope)5031 static inline int rt_scope(int ifa_scope)
5032 {
5033 if (ifa_scope & IFA_HOST)
5034 return RT_SCOPE_HOST;
5035 else if (ifa_scope & IFA_LINK)
5036 return RT_SCOPE_LINK;
5037 else if (ifa_scope & IFA_SITE)
5038 return RT_SCOPE_SITE;
5039 else
5040 return RT_SCOPE_UNIVERSE;
5041 }
5042
inet6_ifaddr_msgsize(void)5043 static inline int inet6_ifaddr_msgsize(void)
5044 {
5045 return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
5046 + nla_total_size(16) /* IFA_LOCAL */
5047 + nla_total_size(16) /* IFA_ADDRESS */
5048 + nla_total_size(sizeof(struct ifa_cacheinfo))
5049 + nla_total_size(4) /* IFA_FLAGS */
5050 + nla_total_size(1) /* IFA_PROTO */
5051 + nla_total_size(4) /* IFA_RT_PRIORITY */;
5052 }
5053
5054 enum addr_type_t {
5055 UNICAST_ADDR,
5056 MULTICAST_ADDR,
5057 ANYCAST_ADDR,
5058 };
5059
5060 struct inet6_fill_args {
5061 u32 portid;
5062 u32 seq;
5063 int event;
5064 unsigned int flags;
5065 int netnsid;
5066 int ifindex;
5067 enum addr_type_t type;
5068 };
5069
inet6_fill_ifaddr(struct sk_buff * skb,struct inet6_ifaddr * ifa,struct inet6_fill_args * args)5070 static int inet6_fill_ifaddr(struct sk_buff *skb, struct inet6_ifaddr *ifa,
5071 struct inet6_fill_args *args)
5072 {
5073 struct nlmsghdr *nlh;
5074 u32 preferred, valid;
5075
5076 nlh = nlmsg_put(skb, args->portid, args->seq, args->event,
5077 sizeof(struct ifaddrmsg), args->flags);
5078 if (!nlh)
5079 return -EMSGSIZE;
5080
5081 put_ifaddrmsg(nlh, ifa->prefix_len, ifa->flags, rt_scope(ifa->scope),
5082 ifa->idev->dev->ifindex);
5083
5084 if (args->netnsid >= 0 &&
5085 nla_put_s32(skb, IFA_TARGET_NETNSID, args->netnsid))
5086 goto error;
5087
5088 spin_lock_bh(&ifa->lock);
5089 if (!((ifa->flags&IFA_F_PERMANENT) &&
5090 (ifa->prefered_lft == INFINITY_LIFE_TIME))) {
5091 preferred = ifa->prefered_lft;
5092 valid = ifa->valid_lft;
5093 if (preferred != INFINITY_LIFE_TIME) {
5094 long tval = (jiffies - ifa->tstamp)/HZ;
5095 if (preferred > tval)
5096 preferred -= tval;
5097 else
5098 preferred = 0;
5099 if (valid != INFINITY_LIFE_TIME) {
5100 if (valid > tval)
5101 valid -= tval;
5102 else
5103 valid = 0;
5104 }
5105 }
5106 } else {
5107 preferred = INFINITY_LIFE_TIME;
5108 valid = INFINITY_LIFE_TIME;
5109 }
5110 spin_unlock_bh(&ifa->lock);
5111
5112 if (!ipv6_addr_any(&ifa->peer_addr)) {
5113 if (nla_put_in6_addr(skb, IFA_LOCAL, &ifa->addr) < 0 ||
5114 nla_put_in6_addr(skb, IFA_ADDRESS, &ifa->peer_addr) < 0)
5115 goto error;
5116 } else
5117 if (nla_put_in6_addr(skb, IFA_ADDRESS, &ifa->addr) < 0)
5118 goto error;
5119
5120 if (ifa->rt_priority &&
5121 nla_put_u32(skb, IFA_RT_PRIORITY, ifa->rt_priority))
5122 goto error;
5123
5124 if (put_cacheinfo(skb, ifa->cstamp, ifa->tstamp, preferred, valid) < 0)
5125 goto error;
5126
5127 if (nla_put_u32(skb, IFA_FLAGS, ifa->flags) < 0)
5128 goto error;
5129
5130 if (ifa->ifa_proto &&
5131 nla_put_u8(skb, IFA_PROTO, ifa->ifa_proto))
5132 goto error;
5133
5134 nlmsg_end(skb, nlh);
5135 return 0;
5136
5137 error:
5138 nlmsg_cancel(skb, nlh);
5139 return -EMSGSIZE;
5140 }
5141
inet6_fill_ifmcaddr(struct sk_buff * skb,struct ifmcaddr6 * ifmca,struct inet6_fill_args * args)5142 static int inet6_fill_ifmcaddr(struct sk_buff *skb, struct ifmcaddr6 *ifmca,
5143 struct inet6_fill_args *args)
5144 {
5145 struct nlmsghdr *nlh;
5146 u8 scope = RT_SCOPE_UNIVERSE;
5147 int ifindex = ifmca->idev->dev->ifindex;
5148
5149 if (ipv6_addr_scope(&ifmca->mca_addr) & IFA_SITE)
5150 scope = RT_SCOPE_SITE;
5151
5152 nlh = nlmsg_put(skb, args->portid, args->seq, args->event,
5153 sizeof(struct ifaddrmsg), args->flags);
5154 if (!nlh)
5155 return -EMSGSIZE;
5156
5157 if (args->netnsid >= 0 &&
5158 nla_put_s32(skb, IFA_TARGET_NETNSID, args->netnsid)) {
5159 nlmsg_cancel(skb, nlh);
5160 return -EMSGSIZE;
5161 }
5162
5163 put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
5164 if (nla_put_in6_addr(skb, IFA_MULTICAST, &ifmca->mca_addr) < 0 ||
5165 put_cacheinfo(skb, ifmca->mca_cstamp, ifmca->mca_tstamp,
5166 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
5167 nlmsg_cancel(skb, nlh);
5168 return -EMSGSIZE;
5169 }
5170
5171 nlmsg_end(skb, nlh);
5172 return 0;
5173 }
5174
inet6_fill_ifacaddr(struct sk_buff * skb,struct ifacaddr6 * ifaca,struct inet6_fill_args * args)5175 static int inet6_fill_ifacaddr(struct sk_buff *skb, struct ifacaddr6 *ifaca,
5176 struct inet6_fill_args *args)
5177 {
5178 struct net_device *dev = fib6_info_nh_dev(ifaca->aca_rt);
5179 int ifindex = dev ? dev->ifindex : 1;
5180 struct nlmsghdr *nlh;
5181 u8 scope = RT_SCOPE_UNIVERSE;
5182
5183 if (ipv6_addr_scope(&ifaca->aca_addr) & IFA_SITE)
5184 scope = RT_SCOPE_SITE;
5185
5186 nlh = nlmsg_put(skb, args->portid, args->seq, args->event,
5187 sizeof(struct ifaddrmsg), args->flags);
5188 if (!nlh)
5189 return -EMSGSIZE;
5190
5191 if (args->netnsid >= 0 &&
5192 nla_put_s32(skb, IFA_TARGET_NETNSID, args->netnsid)) {
5193 nlmsg_cancel(skb, nlh);
5194 return -EMSGSIZE;
5195 }
5196
5197 put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
5198 if (nla_put_in6_addr(skb, IFA_ANYCAST, &ifaca->aca_addr) < 0 ||
5199 put_cacheinfo(skb, ifaca->aca_cstamp, ifaca->aca_tstamp,
5200 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
5201 nlmsg_cancel(skb, nlh);
5202 return -EMSGSIZE;
5203 }
5204
5205 nlmsg_end(skb, nlh);
5206 return 0;
5207 }
5208
5209 /* 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)5210 static int in6_dump_addrs(struct inet6_dev *idev, struct sk_buff *skb,
5211 struct netlink_callback *cb, int s_ip_idx,
5212 struct inet6_fill_args *fillargs)
5213 {
5214 struct ifmcaddr6 *ifmca;
5215 struct ifacaddr6 *ifaca;
5216 int ip_idx = 0;
5217 int err = 1;
5218
5219 read_lock_bh(&idev->lock);
5220 switch (fillargs->type) {
5221 case UNICAST_ADDR: {
5222 struct inet6_ifaddr *ifa;
5223 fillargs->event = RTM_NEWADDR;
5224
5225 /* unicast address incl. temp addr */
5226 list_for_each_entry(ifa, &idev->addr_list, if_list) {
5227 if (ip_idx < s_ip_idx)
5228 goto next;
5229 err = inet6_fill_ifaddr(skb, ifa, fillargs);
5230 if (err < 0)
5231 break;
5232 nl_dump_check_consistent(cb, nlmsg_hdr(skb));
5233 next:
5234 ip_idx++;
5235 }
5236 break;
5237 }
5238 case MULTICAST_ADDR:
5239 read_unlock_bh(&idev->lock);
5240 fillargs->event = RTM_GETMULTICAST;
5241
5242 /* multicast address */
5243 for (ifmca = rtnl_dereference(idev->mc_list);
5244 ifmca;
5245 ifmca = rtnl_dereference(ifmca->next), ip_idx++) {
5246 if (ip_idx < s_ip_idx)
5247 continue;
5248 err = inet6_fill_ifmcaddr(skb, ifmca, fillargs);
5249 if (err < 0)
5250 break;
5251 }
5252 read_lock_bh(&idev->lock);
5253 break;
5254 case ANYCAST_ADDR:
5255 fillargs->event = RTM_GETANYCAST;
5256 /* anycast address */
5257 for (ifaca = idev->ac_list; ifaca;
5258 ifaca = ifaca->aca_next, ip_idx++) {
5259 if (ip_idx < s_ip_idx)
5260 continue;
5261 err = inet6_fill_ifacaddr(skb, ifaca, fillargs);
5262 if (err < 0)
5263 break;
5264 }
5265 break;
5266 default:
5267 break;
5268 }
5269 read_unlock_bh(&idev->lock);
5270 cb->args[2] = ip_idx;
5271 return err;
5272 }
5273
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)5274 static int inet6_valid_dump_ifaddr_req(const struct nlmsghdr *nlh,
5275 struct inet6_fill_args *fillargs,
5276 struct net **tgt_net, struct sock *sk,
5277 struct netlink_callback *cb)
5278 {
5279 struct netlink_ext_ack *extack = cb->extack;
5280 struct nlattr *tb[IFA_MAX+1];
5281 struct ifaddrmsg *ifm;
5282 int err, i;
5283
5284 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) {
5285 NL_SET_ERR_MSG_MOD(extack, "Invalid header for address dump request");
5286 return -EINVAL;
5287 }
5288
5289 ifm = nlmsg_data(nlh);
5290 if (ifm->ifa_prefixlen || ifm->ifa_flags || ifm->ifa_scope) {
5291 NL_SET_ERR_MSG_MOD(extack, "Invalid values in header for address dump request");
5292 return -EINVAL;
5293 }
5294
5295 fillargs->ifindex = ifm->ifa_index;
5296 if (fillargs->ifindex) {
5297 cb->answer_flags |= NLM_F_DUMP_FILTERED;
5298 fillargs->flags |= NLM_F_DUMP_FILTERED;
5299 }
5300
5301 err = nlmsg_parse_deprecated_strict(nlh, sizeof(*ifm), tb, IFA_MAX,
5302 ifa_ipv6_policy, extack);
5303 if (err < 0)
5304 return err;
5305
5306 for (i = 0; i <= IFA_MAX; ++i) {
5307 if (!tb[i])
5308 continue;
5309
5310 if (i == IFA_TARGET_NETNSID) {
5311 struct net *net;
5312
5313 fillargs->netnsid = nla_get_s32(tb[i]);
5314 net = rtnl_get_net_ns_capable(sk, fillargs->netnsid);
5315 if (IS_ERR(net)) {
5316 fillargs->netnsid = -1;
5317 NL_SET_ERR_MSG_MOD(extack, "Invalid target network namespace id");
5318 return PTR_ERR(net);
5319 }
5320 *tgt_net = net;
5321 } else {
5322 NL_SET_ERR_MSG_MOD(extack, "Unsupported attribute in dump request");
5323 return -EINVAL;
5324 }
5325 }
5326
5327 return 0;
5328 }
5329
inet6_dump_addr(struct sk_buff * skb,struct netlink_callback * cb,enum addr_type_t type)5330 static int inet6_dump_addr(struct sk_buff *skb, struct netlink_callback *cb,
5331 enum addr_type_t type)
5332 {
5333 const struct nlmsghdr *nlh = cb->nlh;
5334 struct inet6_fill_args fillargs = {
5335 .portid = NETLINK_CB(cb->skb).portid,
5336 .seq = cb->nlh->nlmsg_seq,
5337 .flags = NLM_F_MULTI,
5338 .netnsid = -1,
5339 .type = type,
5340 };
5341 struct net *tgt_net = sock_net(skb->sk);
5342 int idx, s_idx, s_ip_idx;
5343 int h, s_h;
5344 struct net_device *dev;
5345 struct inet6_dev *idev;
5346 struct hlist_head *head;
5347 int err = 0;
5348
5349 s_h = cb->args[0];
5350 s_idx = idx = cb->args[1];
5351 s_ip_idx = cb->args[2];
5352
5353 if (cb->strict_check) {
5354 err = inet6_valid_dump_ifaddr_req(nlh, &fillargs, &tgt_net,
5355 skb->sk, cb);
5356 if (err < 0)
5357 goto put_tgt_net;
5358
5359 err = 0;
5360 if (fillargs.ifindex) {
5361 dev = __dev_get_by_index(tgt_net, fillargs.ifindex);
5362 if (!dev) {
5363 err = -ENODEV;
5364 goto put_tgt_net;
5365 }
5366 idev = __in6_dev_get(dev);
5367 if (idev) {
5368 err = in6_dump_addrs(idev, skb, cb, s_ip_idx,
5369 &fillargs);
5370 if (err > 0)
5371 err = 0;
5372 }
5373 goto put_tgt_net;
5374 }
5375 }
5376
5377 rcu_read_lock();
5378 cb->seq = inet6_base_seq(tgt_net);
5379 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
5380 idx = 0;
5381 head = &tgt_net->dev_index_head[h];
5382 hlist_for_each_entry_rcu(dev, head, index_hlist) {
5383 if (idx < s_idx)
5384 goto cont;
5385 if (h > s_h || idx > s_idx)
5386 s_ip_idx = 0;
5387 idev = __in6_dev_get(dev);
5388 if (!idev)
5389 goto cont;
5390
5391 if (in6_dump_addrs(idev, skb, cb, s_ip_idx,
5392 &fillargs) < 0)
5393 goto done;
5394 cont:
5395 idx++;
5396 }
5397 }
5398 done:
5399 rcu_read_unlock();
5400 cb->args[0] = h;
5401 cb->args[1] = idx;
5402 put_tgt_net:
5403 if (fillargs.netnsid >= 0)
5404 put_net(tgt_net);
5405
5406 return skb->len ? : err;
5407 }
5408
inet6_dump_ifaddr(struct sk_buff * skb,struct netlink_callback * cb)5409 static int inet6_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
5410 {
5411 enum addr_type_t type = UNICAST_ADDR;
5412
5413 return inet6_dump_addr(skb, cb, type);
5414 }
5415
inet6_dump_ifmcaddr(struct sk_buff * skb,struct netlink_callback * cb)5416 static int inet6_dump_ifmcaddr(struct sk_buff *skb, struct netlink_callback *cb)
5417 {
5418 enum addr_type_t type = MULTICAST_ADDR;
5419
5420 return inet6_dump_addr(skb, cb, type);
5421 }
5422
5423
inet6_dump_ifacaddr(struct sk_buff * skb,struct netlink_callback * cb)5424 static int inet6_dump_ifacaddr(struct sk_buff *skb, struct netlink_callback *cb)
5425 {
5426 enum addr_type_t type = ANYCAST_ADDR;
5427
5428 return inet6_dump_addr(skb, cb, type);
5429 }
5430
inet6_rtm_valid_getaddr_req(struct sk_buff * skb,const struct nlmsghdr * nlh,struct nlattr ** tb,struct netlink_ext_ack * extack)5431 static int inet6_rtm_valid_getaddr_req(struct sk_buff *skb,
5432 const struct nlmsghdr *nlh,
5433 struct nlattr **tb,
5434 struct netlink_ext_ack *extack)
5435 {
5436 struct ifaddrmsg *ifm;
5437 int i, err;
5438
5439 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) {
5440 NL_SET_ERR_MSG_MOD(extack, "Invalid header for get address request");
5441 return -EINVAL;
5442 }
5443
5444 if (!netlink_strict_get_check(skb))
5445 return nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFA_MAX,
5446 ifa_ipv6_policy, extack);
5447
5448 ifm = nlmsg_data(nlh);
5449 if (ifm->ifa_prefixlen || ifm->ifa_flags || ifm->ifa_scope) {
5450 NL_SET_ERR_MSG_MOD(extack, "Invalid values in header for get address request");
5451 return -EINVAL;
5452 }
5453
5454 err = nlmsg_parse_deprecated_strict(nlh, sizeof(*ifm), tb, IFA_MAX,
5455 ifa_ipv6_policy, extack);
5456 if (err)
5457 return err;
5458
5459 for (i = 0; i <= IFA_MAX; i++) {
5460 if (!tb[i])
5461 continue;
5462
5463 switch (i) {
5464 case IFA_TARGET_NETNSID:
5465 case IFA_ADDRESS:
5466 case IFA_LOCAL:
5467 break;
5468 default:
5469 NL_SET_ERR_MSG_MOD(extack, "Unsupported attribute in get address request");
5470 return -EINVAL;
5471 }
5472 }
5473
5474 return 0;
5475 }
5476
inet6_rtm_getaddr(struct sk_buff * in_skb,struct nlmsghdr * nlh,struct netlink_ext_ack * extack)5477 static int inet6_rtm_getaddr(struct sk_buff *in_skb, struct nlmsghdr *nlh,
5478 struct netlink_ext_ack *extack)
5479 {
5480 struct net *tgt_net = sock_net(in_skb->sk);
5481 struct inet6_fill_args fillargs = {
5482 .portid = NETLINK_CB(in_skb).portid,
5483 .seq = nlh->nlmsg_seq,
5484 .event = RTM_NEWADDR,
5485 .flags = 0,
5486 .netnsid = -1,
5487 };
5488 struct ifaddrmsg *ifm;
5489 struct nlattr *tb[IFA_MAX+1];
5490 struct in6_addr *addr = NULL, *peer;
5491 struct net_device *dev = NULL;
5492 struct inet6_ifaddr *ifa;
5493 struct sk_buff *skb;
5494 int err;
5495
5496 err = inet6_rtm_valid_getaddr_req(in_skb, nlh, tb, extack);
5497 if (err < 0)
5498 return err;
5499
5500 if (tb[IFA_TARGET_NETNSID]) {
5501 fillargs.netnsid = nla_get_s32(tb[IFA_TARGET_NETNSID]);
5502
5503 tgt_net = rtnl_get_net_ns_capable(NETLINK_CB(in_skb).sk,
5504 fillargs.netnsid);
5505 if (IS_ERR(tgt_net))
5506 return PTR_ERR(tgt_net);
5507 }
5508
5509 addr = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer);
5510 if (!addr) {
5511 err = -EINVAL;
5512 goto errout;
5513 }
5514 ifm = nlmsg_data(nlh);
5515 if (ifm->ifa_index)
5516 dev = dev_get_by_index(tgt_net, ifm->ifa_index);
5517
5518 ifa = ipv6_get_ifaddr(tgt_net, addr, dev, 1);
5519 if (!ifa) {
5520 err = -EADDRNOTAVAIL;
5521 goto errout;
5522 }
5523
5524 skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_KERNEL);
5525 if (!skb) {
5526 err = -ENOBUFS;
5527 goto errout_ifa;
5528 }
5529
5530 err = inet6_fill_ifaddr(skb, ifa, &fillargs);
5531 if (err < 0) {
5532 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
5533 WARN_ON(err == -EMSGSIZE);
5534 kfree_skb(skb);
5535 goto errout_ifa;
5536 }
5537 err = rtnl_unicast(skb, tgt_net, NETLINK_CB(in_skb).portid);
5538 errout_ifa:
5539 in6_ifa_put(ifa);
5540 errout:
5541 dev_put(dev);
5542 if (fillargs.netnsid >= 0)
5543 put_net(tgt_net);
5544
5545 return err;
5546 }
5547
inet6_ifa_notify(int event,struct inet6_ifaddr * ifa)5548 static void inet6_ifa_notify(int event, struct inet6_ifaddr *ifa)
5549 {
5550 struct sk_buff *skb;
5551 struct net *net = dev_net(ifa->idev->dev);
5552 struct inet6_fill_args fillargs = {
5553 .portid = 0,
5554 .seq = 0,
5555 .event = event,
5556 .flags = 0,
5557 .netnsid = -1,
5558 };
5559 int err = -ENOBUFS;
5560
5561 skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_ATOMIC);
5562 if (!skb)
5563 goto errout;
5564
5565 err = inet6_fill_ifaddr(skb, ifa, &fillargs);
5566 if (err < 0) {
5567 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
5568 WARN_ON(err == -EMSGSIZE);
5569 kfree_skb(skb);
5570 goto errout;
5571 }
5572 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
5573 return;
5574 errout:
5575 if (err < 0)
5576 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
5577 }
5578
ipv6_store_devconf(struct ipv6_devconf * cnf,__s32 * array,int bytes)5579 static inline void ipv6_store_devconf(struct ipv6_devconf *cnf,
5580 __s32 *array, int bytes)
5581 {
5582 BUG_ON(bytes < (DEVCONF_MAX * 4));
5583
5584 memset(array, 0, bytes);
5585 array[DEVCONF_FORWARDING] = cnf->forwarding;
5586 array[DEVCONF_HOPLIMIT] = cnf->hop_limit;
5587 array[DEVCONF_MTU6] = cnf->mtu6;
5588 array[DEVCONF_ACCEPT_RA] = cnf->accept_ra;
5589 array[DEVCONF_ACCEPT_REDIRECTS] = cnf->accept_redirects;
5590 array[DEVCONF_AUTOCONF] = cnf->autoconf;
5591 array[DEVCONF_DAD_TRANSMITS] = cnf->dad_transmits;
5592 array[DEVCONF_RTR_SOLICITS] = cnf->rtr_solicits;
5593 array[DEVCONF_RTR_SOLICIT_INTERVAL] =
5594 jiffies_to_msecs(cnf->rtr_solicit_interval);
5595 array[DEVCONF_RTR_SOLICIT_MAX_INTERVAL] =
5596 jiffies_to_msecs(cnf->rtr_solicit_max_interval);
5597 array[DEVCONF_RTR_SOLICIT_DELAY] =
5598 jiffies_to_msecs(cnf->rtr_solicit_delay);
5599 array[DEVCONF_FORCE_MLD_VERSION] = cnf->force_mld_version;
5600 array[DEVCONF_MLDV1_UNSOLICITED_REPORT_INTERVAL] =
5601 jiffies_to_msecs(cnf->mldv1_unsolicited_report_interval);
5602 array[DEVCONF_MLDV2_UNSOLICITED_REPORT_INTERVAL] =
5603 jiffies_to_msecs(cnf->mldv2_unsolicited_report_interval);
5604 array[DEVCONF_USE_TEMPADDR] = cnf->use_tempaddr;
5605 array[DEVCONF_TEMP_VALID_LFT] = cnf->temp_valid_lft;
5606 array[DEVCONF_TEMP_PREFERED_LFT] = cnf->temp_prefered_lft;
5607 array[DEVCONF_REGEN_MAX_RETRY] = cnf->regen_max_retry;
5608 array[DEVCONF_MAX_DESYNC_FACTOR] = cnf->max_desync_factor;
5609 array[DEVCONF_MAX_ADDRESSES] = cnf->max_addresses;
5610 array[DEVCONF_ACCEPT_RA_DEFRTR] = cnf->accept_ra_defrtr;
5611 array[DEVCONF_RA_DEFRTR_METRIC] = cnf->ra_defrtr_metric;
5612 array[DEVCONF_ACCEPT_RA_MIN_HOP_LIMIT] = cnf->accept_ra_min_hop_limit;
5613 array[DEVCONF_ACCEPT_RA_PINFO] = cnf->accept_ra_pinfo;
5614 #ifdef CONFIG_IPV6_ROUTER_PREF
5615 array[DEVCONF_ACCEPT_RA_RTR_PREF] = cnf->accept_ra_rtr_pref;
5616 array[DEVCONF_RTR_PROBE_INTERVAL] =
5617 jiffies_to_msecs(cnf->rtr_probe_interval);
5618 #ifdef CONFIG_IPV6_ROUTE_INFO
5619 array[DEVCONF_ACCEPT_RA_RT_INFO_MIN_PLEN] = cnf->accept_ra_rt_info_min_plen;
5620 array[DEVCONF_ACCEPT_RA_RT_INFO_MAX_PLEN] = cnf->accept_ra_rt_info_max_plen;
5621 #endif
5622 #endif
5623 array[DEVCONF_PROXY_NDP] = cnf->proxy_ndp;
5624 array[DEVCONF_ACCEPT_SOURCE_ROUTE] = cnf->accept_source_route;
5625 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
5626 array[DEVCONF_OPTIMISTIC_DAD] = cnf->optimistic_dad;
5627 array[DEVCONF_USE_OPTIMISTIC] = cnf->use_optimistic;
5628 #endif
5629 #ifdef CONFIG_IPV6_MROUTE
5630 array[DEVCONF_MC_FORWARDING] = atomic_read(&cnf->mc_forwarding);
5631 #endif
5632 array[DEVCONF_DISABLE_IPV6] = cnf->disable_ipv6;
5633 array[DEVCONF_ACCEPT_DAD] = cnf->accept_dad;
5634 array[DEVCONF_FORCE_TLLAO] = cnf->force_tllao;
5635 array[DEVCONF_NDISC_NOTIFY] = cnf->ndisc_notify;
5636 array[DEVCONF_SUPPRESS_FRAG_NDISC] = cnf->suppress_frag_ndisc;
5637 array[DEVCONF_ACCEPT_RA_FROM_LOCAL] = cnf->accept_ra_from_local;
5638 array[DEVCONF_ACCEPT_RA_MTU] = cnf->accept_ra_mtu;
5639 array[DEVCONF_IGNORE_ROUTES_WITH_LINKDOWN] = cnf->ignore_routes_with_linkdown;
5640 /* we omit DEVCONF_STABLE_SECRET for now */
5641 array[DEVCONF_USE_OIF_ADDRS_ONLY] = cnf->use_oif_addrs_only;
5642 array[DEVCONF_DROP_UNICAST_IN_L2_MULTICAST] = cnf->drop_unicast_in_l2_multicast;
5643 array[DEVCONF_DROP_UNSOLICITED_NA] = cnf->drop_unsolicited_na;
5644 array[DEVCONF_KEEP_ADDR_ON_DOWN] = cnf->keep_addr_on_down;
5645 array[DEVCONF_SEG6_ENABLED] = cnf->seg6_enabled;
5646 #ifdef CONFIG_IPV6_SEG6_HMAC
5647 array[DEVCONF_SEG6_REQUIRE_HMAC] = cnf->seg6_require_hmac;
5648 #endif
5649 array[DEVCONF_ENHANCED_DAD] = cnf->enhanced_dad;
5650 array[DEVCONF_ADDR_GEN_MODE] = cnf->addr_gen_mode;
5651 array[DEVCONF_DISABLE_POLICY] = cnf->disable_policy;
5652 array[DEVCONF_NDISC_TCLASS] = cnf->ndisc_tclass;
5653 array[DEVCONF_RPL_SEG_ENABLED] = cnf->rpl_seg_enabled;
5654 array[DEVCONF_IOAM6_ENABLED] = cnf->ioam6_enabled;
5655 array[DEVCONF_IOAM6_ID] = cnf->ioam6_id;
5656 array[DEVCONF_IOAM6_ID_WIDE] = cnf->ioam6_id_wide;
5657 array[DEVCONF_NDISC_EVICT_NOCARRIER] = cnf->ndisc_evict_nocarrier;
5658 array[DEVCONF_ACCEPT_UNTRACKED_NA] = cnf->accept_untracked_na;
5659 array[DEVCONF_ACCEPT_RA_MIN_LFT] = cnf->accept_ra_min_lft;
5660 }
5661
inet6_ifla6_size(void)5662 static inline size_t inet6_ifla6_size(void)
5663 {
5664 return nla_total_size(4) /* IFLA_INET6_FLAGS */
5665 + nla_total_size(sizeof(struct ifla_cacheinfo))
5666 + nla_total_size(DEVCONF_MAX * 4) /* IFLA_INET6_CONF */
5667 + nla_total_size(IPSTATS_MIB_MAX * 8) /* IFLA_INET6_STATS */
5668 + nla_total_size(ICMP6_MIB_MAX * 8) /* IFLA_INET6_ICMP6STATS */
5669 + nla_total_size(sizeof(struct in6_addr)) /* IFLA_INET6_TOKEN */
5670 + nla_total_size(1) /* IFLA_INET6_ADDR_GEN_MODE */
5671 + nla_total_size(4) /* IFLA_INET6_RA_MTU */
5672 + 0;
5673 }
5674
inet6_if_nlmsg_size(void)5675 static inline size_t inet6_if_nlmsg_size(void)
5676 {
5677 return NLMSG_ALIGN(sizeof(struct ifinfomsg))
5678 + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
5679 + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
5680 + nla_total_size(4) /* IFLA_MTU */
5681 + nla_total_size(4) /* IFLA_LINK */
5682 + nla_total_size(1) /* IFLA_OPERSTATE */
5683 + nla_total_size(inet6_ifla6_size()); /* IFLA_PROTINFO */
5684 }
5685
__snmp6_fill_statsdev(u64 * stats,atomic_long_t * mib,int bytes)5686 static inline void __snmp6_fill_statsdev(u64 *stats, atomic_long_t *mib,
5687 int bytes)
5688 {
5689 int i;
5690 int pad = bytes - sizeof(u64) * ICMP6_MIB_MAX;
5691 BUG_ON(pad < 0);
5692
5693 /* Use put_unaligned() because stats may not be aligned for u64. */
5694 put_unaligned(ICMP6_MIB_MAX, &stats[0]);
5695 for (i = 1; i < ICMP6_MIB_MAX; i++)
5696 put_unaligned(atomic_long_read(&mib[i]), &stats[i]);
5697
5698 memset(&stats[ICMP6_MIB_MAX], 0, pad);
5699 }
5700
__snmp6_fill_stats64(u64 * stats,void __percpu * mib,int bytes,size_t syncpoff)5701 static inline void __snmp6_fill_stats64(u64 *stats, void __percpu *mib,
5702 int bytes, size_t syncpoff)
5703 {
5704 int i, c;
5705 u64 buff[IPSTATS_MIB_MAX];
5706 int pad = bytes - sizeof(u64) * IPSTATS_MIB_MAX;
5707
5708 BUG_ON(pad < 0);
5709
5710 memset(buff, 0, sizeof(buff));
5711 buff[0] = IPSTATS_MIB_MAX;
5712
5713 for_each_possible_cpu(c) {
5714 for (i = 1; i < IPSTATS_MIB_MAX; i++)
5715 buff[i] += snmp_get_cpu_field64(mib, c, i, syncpoff);
5716 }
5717
5718 memcpy(stats, buff, IPSTATS_MIB_MAX * sizeof(u64));
5719 memset(&stats[IPSTATS_MIB_MAX], 0, pad);
5720 }
5721
snmp6_fill_stats(u64 * stats,struct inet6_dev * idev,int attrtype,int bytes)5722 static void snmp6_fill_stats(u64 *stats, struct inet6_dev *idev, int attrtype,
5723 int bytes)
5724 {
5725 switch (attrtype) {
5726 case IFLA_INET6_STATS:
5727 __snmp6_fill_stats64(stats, idev->stats.ipv6, bytes,
5728 offsetof(struct ipstats_mib, syncp));
5729 break;
5730 case IFLA_INET6_ICMP6STATS:
5731 __snmp6_fill_statsdev(stats, idev->stats.icmpv6dev->mibs, bytes);
5732 break;
5733 }
5734 }
5735
inet6_fill_ifla6_attrs(struct sk_buff * skb,struct inet6_dev * idev,u32 ext_filter_mask)5736 static int inet6_fill_ifla6_attrs(struct sk_buff *skb, struct inet6_dev *idev,
5737 u32 ext_filter_mask)
5738 {
5739 struct nlattr *nla;
5740 struct ifla_cacheinfo ci;
5741
5742 if (nla_put_u32(skb, IFLA_INET6_FLAGS, idev->if_flags))
5743 goto nla_put_failure;
5744 ci.max_reasm_len = IPV6_MAXPLEN;
5745 ci.tstamp = cstamp_delta(idev->tstamp);
5746 ci.reachable_time = jiffies_to_msecs(idev->nd_parms->reachable_time);
5747 ci.retrans_time = jiffies_to_msecs(NEIGH_VAR(idev->nd_parms, RETRANS_TIME));
5748 if (nla_put(skb, IFLA_INET6_CACHEINFO, sizeof(ci), &ci))
5749 goto nla_put_failure;
5750 nla = nla_reserve(skb, IFLA_INET6_CONF, DEVCONF_MAX * sizeof(s32));
5751 if (!nla)
5752 goto nla_put_failure;
5753 ipv6_store_devconf(&idev->cnf, nla_data(nla), nla_len(nla));
5754
5755 /* XXX - MC not implemented */
5756
5757 if (ext_filter_mask & RTEXT_FILTER_SKIP_STATS)
5758 return 0;
5759
5760 nla = nla_reserve(skb, IFLA_INET6_STATS, IPSTATS_MIB_MAX * sizeof(u64));
5761 if (!nla)
5762 goto nla_put_failure;
5763 snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_STATS, nla_len(nla));
5764
5765 nla = nla_reserve(skb, IFLA_INET6_ICMP6STATS, ICMP6_MIB_MAX * sizeof(u64));
5766 if (!nla)
5767 goto nla_put_failure;
5768 snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_ICMP6STATS, nla_len(nla));
5769
5770 nla = nla_reserve(skb, IFLA_INET6_TOKEN, sizeof(struct in6_addr));
5771 if (!nla)
5772 goto nla_put_failure;
5773 read_lock_bh(&idev->lock);
5774 memcpy(nla_data(nla), idev->token.s6_addr, nla_len(nla));
5775 read_unlock_bh(&idev->lock);
5776
5777 if (nla_put_u8(skb, IFLA_INET6_ADDR_GEN_MODE, idev->cnf.addr_gen_mode))
5778 goto nla_put_failure;
5779
5780 if (idev->ra_mtu &&
5781 nla_put_u32(skb, IFLA_INET6_RA_MTU, idev->ra_mtu))
5782 goto nla_put_failure;
5783
5784 return 0;
5785
5786 nla_put_failure:
5787 return -EMSGSIZE;
5788 }
5789
inet6_get_link_af_size(const struct net_device * dev,u32 ext_filter_mask)5790 static size_t inet6_get_link_af_size(const struct net_device *dev,
5791 u32 ext_filter_mask)
5792 {
5793 if (!__in6_dev_get(dev))
5794 return 0;
5795
5796 return inet6_ifla6_size();
5797 }
5798
inet6_fill_link_af(struct sk_buff * skb,const struct net_device * dev,u32 ext_filter_mask)5799 static int inet6_fill_link_af(struct sk_buff *skb, const struct net_device *dev,
5800 u32 ext_filter_mask)
5801 {
5802 struct inet6_dev *idev = __in6_dev_get(dev);
5803
5804 if (!idev)
5805 return -ENODATA;
5806
5807 if (inet6_fill_ifla6_attrs(skb, idev, ext_filter_mask) < 0)
5808 return -EMSGSIZE;
5809
5810 return 0;
5811 }
5812
inet6_set_iftoken(struct inet6_dev * idev,struct in6_addr * token,struct netlink_ext_ack * extack)5813 static int inet6_set_iftoken(struct inet6_dev *idev, struct in6_addr *token,
5814 struct netlink_ext_ack *extack)
5815 {
5816 struct inet6_ifaddr *ifp;
5817 struct net_device *dev = idev->dev;
5818 bool clear_token, update_rs = false;
5819 struct in6_addr ll_addr;
5820
5821 ASSERT_RTNL();
5822
5823 if (!token)
5824 return -EINVAL;
5825
5826 if (dev->flags & IFF_LOOPBACK) {
5827 NL_SET_ERR_MSG_MOD(extack, "Device is loopback");
5828 return -EINVAL;
5829 }
5830
5831 if (dev->flags & IFF_NOARP) {
5832 NL_SET_ERR_MSG_MOD(extack,
5833 "Device does not do neighbour discovery");
5834 return -EINVAL;
5835 }
5836
5837 if (!ipv6_accept_ra(idev)) {
5838 NL_SET_ERR_MSG_MOD(extack,
5839 "Router advertisement is disabled on device");
5840 return -EINVAL;
5841 }
5842
5843 if (idev->cnf.rtr_solicits == 0) {
5844 NL_SET_ERR_MSG(extack,
5845 "Router solicitation is disabled on device");
5846 return -EINVAL;
5847 }
5848
5849 write_lock_bh(&idev->lock);
5850
5851 BUILD_BUG_ON(sizeof(token->s6_addr) != 16);
5852 memcpy(idev->token.s6_addr + 8, token->s6_addr + 8, 8);
5853
5854 write_unlock_bh(&idev->lock);
5855
5856 clear_token = ipv6_addr_any(token);
5857 if (clear_token)
5858 goto update_lft;
5859
5860 if (!idev->dead && (idev->if_flags & IF_READY) &&
5861 !ipv6_get_lladdr(dev, &ll_addr, IFA_F_TENTATIVE |
5862 IFA_F_OPTIMISTIC)) {
5863 /* If we're not ready, then normal ifup will take care
5864 * of this. Otherwise, we need to request our rs here.
5865 */
5866 ndisc_send_rs(dev, &ll_addr, &in6addr_linklocal_allrouters);
5867 update_rs = true;
5868 }
5869
5870 update_lft:
5871 write_lock_bh(&idev->lock);
5872
5873 if (update_rs) {
5874 idev->if_flags |= IF_RS_SENT;
5875 idev->rs_interval = rfc3315_s14_backoff_init(
5876 idev->cnf.rtr_solicit_interval);
5877 idev->rs_probes = 1;
5878 addrconf_mod_rs_timer(idev, idev->rs_interval);
5879 }
5880
5881 /* Well, that's kinda nasty ... */
5882 list_for_each_entry(ifp, &idev->addr_list, if_list) {
5883 spin_lock(&ifp->lock);
5884 if (ifp->tokenized) {
5885 ifp->valid_lft = 0;
5886 ifp->prefered_lft = 0;
5887 }
5888 spin_unlock(&ifp->lock);
5889 }
5890
5891 write_unlock_bh(&idev->lock);
5892 inet6_ifinfo_notify(RTM_NEWLINK, idev);
5893 addrconf_verify_rtnl(dev_net(dev));
5894 return 0;
5895 }
5896
5897 static const struct nla_policy inet6_af_policy[IFLA_INET6_MAX + 1] = {
5898 [IFLA_INET6_ADDR_GEN_MODE] = { .type = NLA_U8 },
5899 [IFLA_INET6_TOKEN] = { .len = sizeof(struct in6_addr) },
5900 [IFLA_INET6_RA_MTU] = { .type = NLA_REJECT,
5901 .reject_message =
5902 "IFLA_INET6_RA_MTU can not be set" },
5903 };
5904
check_addr_gen_mode(int mode)5905 static int check_addr_gen_mode(int mode)
5906 {
5907 if (mode != IN6_ADDR_GEN_MODE_EUI64 &&
5908 mode != IN6_ADDR_GEN_MODE_NONE &&
5909 mode != IN6_ADDR_GEN_MODE_STABLE_PRIVACY &&
5910 mode != IN6_ADDR_GEN_MODE_RANDOM)
5911 return -EINVAL;
5912 return 1;
5913 }
5914
check_stable_privacy(struct inet6_dev * idev,struct net * net,int mode)5915 static int check_stable_privacy(struct inet6_dev *idev, struct net *net,
5916 int mode)
5917 {
5918 if (mode == IN6_ADDR_GEN_MODE_STABLE_PRIVACY &&
5919 !idev->cnf.stable_secret.initialized &&
5920 !net->ipv6.devconf_dflt->stable_secret.initialized)
5921 return -EINVAL;
5922 return 1;
5923 }
5924
inet6_validate_link_af(const struct net_device * dev,const struct nlattr * nla,struct netlink_ext_ack * extack)5925 static int inet6_validate_link_af(const struct net_device *dev,
5926 const struct nlattr *nla,
5927 struct netlink_ext_ack *extack)
5928 {
5929 struct nlattr *tb[IFLA_INET6_MAX + 1];
5930 struct inet6_dev *idev = NULL;
5931 int err;
5932
5933 if (dev) {
5934 idev = __in6_dev_get(dev);
5935 if (!idev)
5936 return -EAFNOSUPPORT;
5937 }
5938
5939 err = nla_parse_nested_deprecated(tb, IFLA_INET6_MAX, nla,
5940 inet6_af_policy, extack);
5941 if (err)
5942 return err;
5943
5944 if (!tb[IFLA_INET6_TOKEN] && !tb[IFLA_INET6_ADDR_GEN_MODE])
5945 return -EINVAL;
5946
5947 if (tb[IFLA_INET6_ADDR_GEN_MODE]) {
5948 u8 mode = nla_get_u8(tb[IFLA_INET6_ADDR_GEN_MODE]);
5949
5950 if (check_addr_gen_mode(mode) < 0)
5951 return -EINVAL;
5952 if (dev && check_stable_privacy(idev, dev_net(dev), mode) < 0)
5953 return -EINVAL;
5954 }
5955
5956 return 0;
5957 }
5958
inet6_set_link_af(struct net_device * dev,const struct nlattr * nla,struct netlink_ext_ack * extack)5959 static int inet6_set_link_af(struct net_device *dev, const struct nlattr *nla,
5960 struct netlink_ext_ack *extack)
5961 {
5962 struct inet6_dev *idev = __in6_dev_get(dev);
5963 struct nlattr *tb[IFLA_INET6_MAX + 1];
5964 int err;
5965
5966 if (!idev)
5967 return -EAFNOSUPPORT;
5968
5969 if (nla_parse_nested_deprecated(tb, IFLA_INET6_MAX, nla, NULL, NULL) < 0)
5970 return -EINVAL;
5971
5972 if (tb[IFLA_INET6_TOKEN]) {
5973 err = inet6_set_iftoken(idev, nla_data(tb[IFLA_INET6_TOKEN]),
5974 extack);
5975 if (err)
5976 return err;
5977 }
5978
5979 if (tb[IFLA_INET6_ADDR_GEN_MODE]) {
5980 u8 mode = nla_get_u8(tb[IFLA_INET6_ADDR_GEN_MODE]);
5981
5982 idev->cnf.addr_gen_mode = mode;
5983 }
5984
5985 return 0;
5986 }
5987
inet6_fill_ifinfo(struct sk_buff * skb,struct inet6_dev * idev,u32 portid,u32 seq,int event,unsigned int flags)5988 static int inet6_fill_ifinfo(struct sk_buff *skb, struct inet6_dev *idev,
5989 u32 portid, u32 seq, int event, unsigned int flags)
5990 {
5991 struct net_device *dev = idev->dev;
5992 struct ifinfomsg *hdr;
5993 struct nlmsghdr *nlh;
5994 void *protoinfo;
5995
5996 nlh = nlmsg_put(skb, portid, seq, event, sizeof(*hdr), flags);
5997 if (!nlh)
5998 return -EMSGSIZE;
5999
6000 hdr = nlmsg_data(nlh);
6001 hdr->ifi_family = AF_INET6;
6002 hdr->__ifi_pad = 0;
6003 hdr->ifi_type = dev->type;
6004 hdr->ifi_index = dev->ifindex;
6005 hdr->ifi_flags = dev_get_flags(dev);
6006 hdr->ifi_change = 0;
6007
6008 if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
6009 (dev->addr_len &&
6010 nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
6011 nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
6012 (dev->ifindex != dev_get_iflink(dev) &&
6013 nla_put_u32(skb, IFLA_LINK, dev_get_iflink(dev))) ||
6014 nla_put_u8(skb, IFLA_OPERSTATE,
6015 netif_running(dev) ? READ_ONCE(dev->operstate) : IF_OPER_DOWN))
6016 goto nla_put_failure;
6017 protoinfo = nla_nest_start_noflag(skb, IFLA_PROTINFO);
6018 if (!protoinfo)
6019 goto nla_put_failure;
6020
6021 if (inet6_fill_ifla6_attrs(skb, idev, 0) < 0)
6022 goto nla_put_failure;
6023
6024 nla_nest_end(skb, protoinfo);
6025 nlmsg_end(skb, nlh);
6026 return 0;
6027
6028 nla_put_failure:
6029 nlmsg_cancel(skb, nlh);
6030 return -EMSGSIZE;
6031 }
6032
inet6_valid_dump_ifinfo(const struct nlmsghdr * nlh,struct netlink_ext_ack * extack)6033 static int inet6_valid_dump_ifinfo(const struct nlmsghdr *nlh,
6034 struct netlink_ext_ack *extack)
6035 {
6036 struct ifinfomsg *ifm;
6037
6038 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) {
6039 NL_SET_ERR_MSG_MOD(extack, "Invalid header for link dump request");
6040 return -EINVAL;
6041 }
6042
6043 if (nlmsg_attrlen(nlh, sizeof(*ifm))) {
6044 NL_SET_ERR_MSG_MOD(extack, "Invalid data after header");
6045 return -EINVAL;
6046 }
6047
6048 ifm = nlmsg_data(nlh);
6049 if (ifm->__ifi_pad || ifm->ifi_type || ifm->ifi_flags ||
6050 ifm->ifi_change || ifm->ifi_index) {
6051 NL_SET_ERR_MSG_MOD(extack, "Invalid values in header for dump request");
6052 return -EINVAL;
6053 }
6054
6055 return 0;
6056 }
6057
inet6_dump_ifinfo(struct sk_buff * skb,struct netlink_callback * cb)6058 static int inet6_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
6059 {
6060 struct net *net = sock_net(skb->sk);
6061 int h, s_h;
6062 int idx = 0, s_idx;
6063 struct net_device *dev;
6064 struct inet6_dev *idev;
6065 struct hlist_head *head;
6066
6067 /* only requests using strict checking can pass data to
6068 * influence the dump
6069 */
6070 if (cb->strict_check) {
6071 int err = inet6_valid_dump_ifinfo(cb->nlh, cb->extack);
6072
6073 if (err < 0)
6074 return err;
6075 }
6076
6077 s_h = cb->args[0];
6078 s_idx = cb->args[1];
6079
6080 rcu_read_lock();
6081 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
6082 idx = 0;
6083 head = &net->dev_index_head[h];
6084 hlist_for_each_entry_rcu(dev, head, index_hlist) {
6085 if (idx < s_idx)
6086 goto cont;
6087 idev = __in6_dev_get(dev);
6088 if (!idev)
6089 goto cont;
6090 if (inet6_fill_ifinfo(skb, idev,
6091 NETLINK_CB(cb->skb).portid,
6092 cb->nlh->nlmsg_seq,
6093 RTM_NEWLINK, NLM_F_MULTI) < 0)
6094 goto out;
6095 cont:
6096 idx++;
6097 }
6098 }
6099 out:
6100 rcu_read_unlock();
6101 cb->args[1] = idx;
6102 cb->args[0] = h;
6103
6104 return skb->len;
6105 }
6106
inet6_ifinfo_notify(int event,struct inet6_dev * idev)6107 void inet6_ifinfo_notify(int event, struct inet6_dev *idev)
6108 {
6109 struct sk_buff *skb;
6110 struct net *net = dev_net(idev->dev);
6111 int err = -ENOBUFS;
6112
6113 skb = nlmsg_new(inet6_if_nlmsg_size(), GFP_ATOMIC);
6114 if (!skb)
6115 goto errout;
6116
6117 err = inet6_fill_ifinfo(skb, idev, 0, 0, event, 0);
6118 if (err < 0) {
6119 /* -EMSGSIZE implies BUG in inet6_if_nlmsg_size() */
6120 WARN_ON(err == -EMSGSIZE);
6121 kfree_skb(skb);
6122 goto errout;
6123 }
6124 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFINFO, NULL, GFP_ATOMIC);
6125 return;
6126 errout:
6127 if (err < 0)
6128 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFINFO, err);
6129 }
6130
inet6_prefix_nlmsg_size(void)6131 static inline size_t inet6_prefix_nlmsg_size(void)
6132 {
6133 return NLMSG_ALIGN(sizeof(struct prefixmsg))
6134 + nla_total_size(sizeof(struct in6_addr))
6135 + nla_total_size(sizeof(struct prefix_cacheinfo));
6136 }
6137
inet6_fill_prefix(struct sk_buff * skb,struct inet6_dev * idev,struct prefix_info * pinfo,u32 portid,u32 seq,int event,unsigned int flags)6138 static int inet6_fill_prefix(struct sk_buff *skb, struct inet6_dev *idev,
6139 struct prefix_info *pinfo, u32 portid, u32 seq,
6140 int event, unsigned int flags)
6141 {
6142 struct prefixmsg *pmsg;
6143 struct nlmsghdr *nlh;
6144 struct prefix_cacheinfo ci;
6145
6146 nlh = nlmsg_put(skb, portid, seq, event, sizeof(*pmsg), flags);
6147 if (!nlh)
6148 return -EMSGSIZE;
6149
6150 pmsg = nlmsg_data(nlh);
6151 pmsg->prefix_family = AF_INET6;
6152 pmsg->prefix_pad1 = 0;
6153 pmsg->prefix_pad2 = 0;
6154 pmsg->prefix_ifindex = idev->dev->ifindex;
6155 pmsg->prefix_len = pinfo->prefix_len;
6156 pmsg->prefix_type = pinfo->type;
6157 pmsg->prefix_pad3 = 0;
6158 pmsg->prefix_flags = pinfo->flags;
6159
6160 if (nla_put(skb, PREFIX_ADDRESS, sizeof(pinfo->prefix), &pinfo->prefix))
6161 goto nla_put_failure;
6162 ci.preferred_time = ntohl(pinfo->prefered);
6163 ci.valid_time = ntohl(pinfo->valid);
6164 if (nla_put(skb, PREFIX_CACHEINFO, sizeof(ci), &ci))
6165 goto nla_put_failure;
6166 nlmsg_end(skb, nlh);
6167 return 0;
6168
6169 nla_put_failure:
6170 nlmsg_cancel(skb, nlh);
6171 return -EMSGSIZE;
6172 }
6173
inet6_prefix_notify(int event,struct inet6_dev * idev,struct prefix_info * pinfo)6174 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
6175 struct prefix_info *pinfo)
6176 {
6177 struct sk_buff *skb;
6178 struct net *net = dev_net(idev->dev);
6179 int err = -ENOBUFS;
6180
6181 skb = nlmsg_new(inet6_prefix_nlmsg_size(), GFP_ATOMIC);
6182 if (!skb)
6183 goto errout;
6184
6185 err = inet6_fill_prefix(skb, idev, pinfo, 0, 0, event, 0);
6186 if (err < 0) {
6187 /* -EMSGSIZE implies BUG in inet6_prefix_nlmsg_size() */
6188 WARN_ON(err == -EMSGSIZE);
6189 kfree_skb(skb);
6190 goto errout;
6191 }
6192 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_PREFIX, NULL, GFP_ATOMIC);
6193 return;
6194 errout:
6195 if (err < 0)
6196 rtnl_set_sk_err(net, RTNLGRP_IPV6_PREFIX, err);
6197 }
6198
__ipv6_ifa_notify(int event,struct inet6_ifaddr * ifp)6199 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
6200 {
6201 struct net *net = dev_net(ifp->idev->dev);
6202
6203 if (event)
6204 ASSERT_RTNL();
6205
6206 inet6_ifa_notify(event ? : RTM_NEWADDR, ifp);
6207
6208 switch (event) {
6209 case RTM_NEWADDR:
6210 /*
6211 * If the address was optimistic we inserted the route at the
6212 * start of our DAD process, so we don't need to do it again.
6213 * If the device was taken down in the middle of the DAD
6214 * cycle there is a race where we could get here without a
6215 * host route, so nothing to insert. That will be fixed when
6216 * the device is brought up.
6217 */
6218 if (ifp->rt && !rcu_access_pointer(ifp->rt->fib6_node)) {
6219 ip6_ins_rt(net, ifp->rt);
6220 } else if (!ifp->rt && (ifp->idev->dev->flags & IFF_UP)) {
6221 pr_warn("BUG: Address %pI6c on device %s is missing its host route.\n",
6222 &ifp->addr, ifp->idev->dev->name);
6223 }
6224
6225 if (ifp->idev->cnf.forwarding)
6226 addrconf_join_anycast(ifp);
6227 if (!ipv6_addr_any(&ifp->peer_addr))
6228 addrconf_prefix_route(&ifp->peer_addr, 128,
6229 ifp->rt_priority, ifp->idev->dev,
6230 0, 0, GFP_ATOMIC);
6231 break;
6232 case RTM_DELADDR:
6233 if (ifp->idev->cnf.forwarding)
6234 addrconf_leave_anycast(ifp);
6235 addrconf_leave_solict(ifp->idev, &ifp->addr);
6236 if (!ipv6_addr_any(&ifp->peer_addr)) {
6237 struct fib6_info *rt;
6238
6239 rt = addrconf_get_prefix_route(&ifp->peer_addr, 128,
6240 ifp->idev->dev, 0, 0,
6241 false);
6242 if (rt)
6243 ip6_del_rt(net, rt, false);
6244 }
6245 if (ifp->rt) {
6246 ip6_del_rt(net, ifp->rt, false);
6247 ifp->rt = NULL;
6248 }
6249 rt_genid_bump_ipv6(net);
6250 break;
6251 }
6252 atomic_inc(&net->ipv6.dev_addr_genid);
6253 }
6254
ipv6_ifa_notify(int event,struct inet6_ifaddr * ifp)6255 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
6256 {
6257 if (likely(ifp->idev->dead == 0))
6258 __ipv6_ifa_notify(event, ifp);
6259 }
6260
6261 #ifdef CONFIG_SYSCTL
6262
addrconf_sysctl_forward(struct ctl_table * ctl,int write,void * buffer,size_t * lenp,loff_t * ppos)6263 static int addrconf_sysctl_forward(struct ctl_table *ctl, int write,
6264 void *buffer, size_t *lenp, loff_t *ppos)
6265 {
6266 int *valp = ctl->data;
6267 int val = *valp;
6268 loff_t pos = *ppos;
6269 struct ctl_table lctl;
6270 int ret;
6271
6272 /*
6273 * ctl->data points to idev->cnf.forwarding, we should
6274 * not modify it until we get the rtnl lock.
6275 */
6276 lctl = *ctl;
6277 lctl.data = &val;
6278
6279 ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
6280
6281 if (write)
6282 ret = addrconf_fixup_forwarding(ctl, valp, val);
6283 if (ret)
6284 *ppos = pos;
6285 return ret;
6286 }
6287
addrconf_sysctl_mtu(struct ctl_table * ctl,int write,void * buffer,size_t * lenp,loff_t * ppos)6288 static int addrconf_sysctl_mtu(struct ctl_table *ctl, int write,
6289 void *buffer, size_t *lenp, loff_t *ppos)
6290 {
6291 struct inet6_dev *idev = ctl->extra1;
6292 int min_mtu = IPV6_MIN_MTU;
6293 struct ctl_table lctl;
6294
6295 lctl = *ctl;
6296 lctl.extra1 = &min_mtu;
6297 lctl.extra2 = idev ? &idev->dev->mtu : NULL;
6298
6299 return proc_dointvec_minmax(&lctl, write, buffer, lenp, ppos);
6300 }
6301
dev_disable_change(struct inet6_dev * idev)6302 static void dev_disable_change(struct inet6_dev *idev)
6303 {
6304 struct netdev_notifier_info info;
6305
6306 if (!idev || !idev->dev)
6307 return;
6308
6309 netdev_notifier_info_init(&info, idev->dev);
6310 if (idev->cnf.disable_ipv6)
6311 addrconf_notify(NULL, NETDEV_DOWN, &info);
6312 else
6313 addrconf_notify(NULL, NETDEV_UP, &info);
6314 }
6315
addrconf_disable_change(struct net * net,__s32 newf)6316 static void addrconf_disable_change(struct net *net, __s32 newf)
6317 {
6318 struct net_device *dev;
6319 struct inet6_dev *idev;
6320
6321 for_each_netdev(net, dev) {
6322 idev = __in6_dev_get(dev);
6323 if (idev) {
6324 int changed = (!idev->cnf.disable_ipv6) ^ (!newf);
6325
6326 WRITE_ONCE(idev->cnf.disable_ipv6, newf);
6327 if (changed)
6328 dev_disable_change(idev);
6329 }
6330 }
6331 }
6332
addrconf_disable_ipv6(struct ctl_table * table,int * p,int newf)6333 static int addrconf_disable_ipv6(struct ctl_table *table, int *p, int newf)
6334 {
6335 struct net *net;
6336 int old;
6337
6338 if (!rtnl_trylock())
6339 return restart_syscall();
6340
6341 net = (struct net *)table->extra2;
6342 old = *p;
6343 WRITE_ONCE(*p, newf);
6344
6345 if (p == &net->ipv6.devconf_dflt->disable_ipv6) {
6346 rtnl_unlock();
6347 return 0;
6348 }
6349
6350 if (p == &net->ipv6.devconf_all->disable_ipv6) {
6351 WRITE_ONCE(net->ipv6.devconf_dflt->disable_ipv6, newf);
6352 addrconf_disable_change(net, newf);
6353 } else if ((!newf) ^ (!old))
6354 dev_disable_change((struct inet6_dev *)table->extra1);
6355
6356 rtnl_unlock();
6357 return 0;
6358 }
6359
addrconf_sysctl_disable(struct ctl_table * ctl,int write,void * buffer,size_t * lenp,loff_t * ppos)6360 static int addrconf_sysctl_disable(struct ctl_table *ctl, int write,
6361 void *buffer, size_t *lenp, loff_t *ppos)
6362 {
6363 int *valp = ctl->data;
6364 int val = *valp;
6365 loff_t pos = *ppos;
6366 struct ctl_table lctl;
6367 int ret;
6368
6369 /*
6370 * ctl->data points to idev->cnf.disable_ipv6, we should
6371 * not modify it until we get the rtnl lock.
6372 */
6373 lctl = *ctl;
6374 lctl.data = &val;
6375
6376 ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
6377
6378 if (write)
6379 ret = addrconf_disable_ipv6(ctl, valp, val);
6380 if (ret)
6381 *ppos = pos;
6382 return ret;
6383 }
6384
addrconf_sysctl_proxy_ndp(struct ctl_table * ctl,int write,void * buffer,size_t * lenp,loff_t * ppos)6385 static int addrconf_sysctl_proxy_ndp(struct ctl_table *ctl, int write,
6386 void *buffer, size_t *lenp, loff_t *ppos)
6387 {
6388 int *valp = ctl->data;
6389 int ret;
6390 int old, new;
6391
6392 old = *valp;
6393 ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
6394 new = *valp;
6395
6396 if (write && old != new) {
6397 struct net *net = ctl->extra2;
6398
6399 if (!rtnl_trylock())
6400 return restart_syscall();
6401
6402 if (valp == &net->ipv6.devconf_dflt->proxy_ndp)
6403 inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
6404 NETCONFA_PROXY_NEIGH,
6405 NETCONFA_IFINDEX_DEFAULT,
6406 net->ipv6.devconf_dflt);
6407 else if (valp == &net->ipv6.devconf_all->proxy_ndp)
6408 inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
6409 NETCONFA_PROXY_NEIGH,
6410 NETCONFA_IFINDEX_ALL,
6411 net->ipv6.devconf_all);
6412 else {
6413 struct inet6_dev *idev = ctl->extra1;
6414
6415 inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
6416 NETCONFA_PROXY_NEIGH,
6417 idev->dev->ifindex,
6418 &idev->cnf);
6419 }
6420 rtnl_unlock();
6421 }
6422
6423 return ret;
6424 }
6425
addrconf_sysctl_addr_gen_mode(struct ctl_table * ctl,int write,void * buffer,size_t * lenp,loff_t * ppos)6426 static int addrconf_sysctl_addr_gen_mode(struct ctl_table *ctl, int write,
6427 void *buffer, size_t *lenp,
6428 loff_t *ppos)
6429 {
6430 int ret = 0;
6431 u32 new_val;
6432 struct inet6_dev *idev = (struct inet6_dev *)ctl->extra1;
6433 struct net *net = (struct net *)ctl->extra2;
6434 struct ctl_table tmp = {
6435 .data = &new_val,
6436 .maxlen = sizeof(new_val),
6437 .mode = ctl->mode,
6438 };
6439
6440 if (!rtnl_trylock())
6441 return restart_syscall();
6442
6443 new_val = *((u32 *)ctl->data);
6444
6445 ret = proc_douintvec(&tmp, write, buffer, lenp, ppos);
6446 if (ret != 0)
6447 goto out;
6448
6449 if (write) {
6450 if (check_addr_gen_mode(new_val) < 0) {
6451 ret = -EINVAL;
6452 goto out;
6453 }
6454
6455 if (idev) {
6456 if (check_stable_privacy(idev, net, new_val) < 0) {
6457 ret = -EINVAL;
6458 goto out;
6459 }
6460
6461 if (idev->cnf.addr_gen_mode != new_val) {
6462 idev->cnf.addr_gen_mode = new_val;
6463 addrconf_init_auto_addrs(idev->dev);
6464 }
6465 } else if (&net->ipv6.devconf_all->addr_gen_mode == ctl->data) {
6466 struct net_device *dev;
6467
6468 net->ipv6.devconf_dflt->addr_gen_mode = new_val;
6469 for_each_netdev(net, dev) {
6470 idev = __in6_dev_get(dev);
6471 if (idev &&
6472 idev->cnf.addr_gen_mode != new_val) {
6473 idev->cnf.addr_gen_mode = new_val;
6474 addrconf_init_auto_addrs(idev->dev);
6475 }
6476 }
6477 }
6478
6479 *((u32 *)ctl->data) = new_val;
6480 }
6481
6482 out:
6483 rtnl_unlock();
6484
6485 return ret;
6486 }
6487
addrconf_sysctl_stable_secret(struct ctl_table * ctl,int write,void * buffer,size_t * lenp,loff_t * ppos)6488 static int addrconf_sysctl_stable_secret(struct ctl_table *ctl, int write,
6489 void *buffer, size_t *lenp,
6490 loff_t *ppos)
6491 {
6492 int err;
6493 struct in6_addr addr;
6494 char str[IPV6_MAX_STRLEN];
6495 struct ctl_table lctl = *ctl;
6496 struct net *net = ctl->extra2;
6497 struct ipv6_stable_secret *secret = ctl->data;
6498
6499 if (&net->ipv6.devconf_all->stable_secret == ctl->data)
6500 return -EIO;
6501
6502 lctl.maxlen = IPV6_MAX_STRLEN;
6503 lctl.data = str;
6504
6505 if (!rtnl_trylock())
6506 return restart_syscall();
6507
6508 if (!write && !secret->initialized) {
6509 err = -EIO;
6510 goto out;
6511 }
6512
6513 err = snprintf(str, sizeof(str), "%pI6", &secret->secret);
6514 if (err >= sizeof(str)) {
6515 err = -EIO;
6516 goto out;
6517 }
6518
6519 err = proc_dostring(&lctl, write, buffer, lenp, ppos);
6520 if (err || !write)
6521 goto out;
6522
6523 if (in6_pton(str, -1, addr.in6_u.u6_addr8, -1, NULL) != 1) {
6524 err = -EIO;
6525 goto out;
6526 }
6527
6528 secret->initialized = true;
6529 secret->secret = addr;
6530
6531 if (&net->ipv6.devconf_dflt->stable_secret == ctl->data) {
6532 struct net_device *dev;
6533
6534 for_each_netdev(net, dev) {
6535 struct inet6_dev *idev = __in6_dev_get(dev);
6536
6537 if (idev) {
6538 idev->cnf.addr_gen_mode =
6539 IN6_ADDR_GEN_MODE_STABLE_PRIVACY;
6540 }
6541 }
6542 } else {
6543 struct inet6_dev *idev = ctl->extra1;
6544
6545 idev->cnf.addr_gen_mode = IN6_ADDR_GEN_MODE_STABLE_PRIVACY;
6546 }
6547
6548 out:
6549 rtnl_unlock();
6550
6551 return err;
6552 }
6553
6554 static
addrconf_sysctl_ignore_routes_with_linkdown(struct ctl_table * ctl,int write,void * buffer,size_t * lenp,loff_t * ppos)6555 int addrconf_sysctl_ignore_routes_with_linkdown(struct ctl_table *ctl,
6556 int write, void *buffer,
6557 size_t *lenp,
6558 loff_t *ppos)
6559 {
6560 int *valp = ctl->data;
6561 int val = *valp;
6562 loff_t pos = *ppos;
6563 struct ctl_table lctl;
6564 int ret;
6565
6566 /* ctl->data points to idev->cnf.ignore_routes_when_linkdown
6567 * we should not modify it until we get the rtnl lock.
6568 */
6569 lctl = *ctl;
6570 lctl.data = &val;
6571
6572 ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
6573
6574 if (write)
6575 ret = addrconf_fixup_linkdown(ctl, valp, val);
6576 if (ret)
6577 *ppos = pos;
6578 return ret;
6579 }
6580
6581 static
addrconf_set_nopolicy(struct rt6_info * rt,int action)6582 void addrconf_set_nopolicy(struct rt6_info *rt, int action)
6583 {
6584 if (rt) {
6585 if (action)
6586 rt->dst.flags |= DST_NOPOLICY;
6587 else
6588 rt->dst.flags &= ~DST_NOPOLICY;
6589 }
6590 }
6591
6592 static
addrconf_disable_policy_idev(struct inet6_dev * idev,int val)6593 void addrconf_disable_policy_idev(struct inet6_dev *idev, int val)
6594 {
6595 struct inet6_ifaddr *ifa;
6596
6597 read_lock_bh(&idev->lock);
6598 list_for_each_entry(ifa, &idev->addr_list, if_list) {
6599 spin_lock(&ifa->lock);
6600 if (ifa->rt) {
6601 /* host routes only use builtin fib6_nh */
6602 struct fib6_nh *nh = ifa->rt->fib6_nh;
6603 int cpu;
6604
6605 rcu_read_lock();
6606 ifa->rt->dst_nopolicy = val ? true : false;
6607 if (nh->rt6i_pcpu) {
6608 for_each_possible_cpu(cpu) {
6609 struct rt6_info **rtp;
6610
6611 rtp = per_cpu_ptr(nh->rt6i_pcpu, cpu);
6612 addrconf_set_nopolicy(*rtp, val);
6613 }
6614 }
6615 rcu_read_unlock();
6616 }
6617 spin_unlock(&ifa->lock);
6618 }
6619 read_unlock_bh(&idev->lock);
6620 }
6621
6622 static
addrconf_disable_policy(struct ctl_table * ctl,int * valp,int val)6623 int addrconf_disable_policy(struct ctl_table *ctl, int *valp, int val)
6624 {
6625 struct inet6_dev *idev;
6626 struct net *net;
6627
6628 if (!rtnl_trylock())
6629 return restart_syscall();
6630
6631 *valp = val;
6632
6633 net = (struct net *)ctl->extra2;
6634 if (valp == &net->ipv6.devconf_dflt->disable_policy) {
6635 rtnl_unlock();
6636 return 0;
6637 }
6638
6639 if (valp == &net->ipv6.devconf_all->disable_policy) {
6640 struct net_device *dev;
6641
6642 for_each_netdev(net, dev) {
6643 idev = __in6_dev_get(dev);
6644 if (idev)
6645 addrconf_disable_policy_idev(idev, val);
6646 }
6647 } else {
6648 idev = (struct inet6_dev *)ctl->extra1;
6649 addrconf_disable_policy_idev(idev, val);
6650 }
6651
6652 rtnl_unlock();
6653 return 0;
6654 }
6655
addrconf_sysctl_disable_policy(struct ctl_table * ctl,int write,void * buffer,size_t * lenp,loff_t * ppos)6656 static int addrconf_sysctl_disable_policy(struct ctl_table *ctl, int write,
6657 void *buffer, size_t *lenp, loff_t *ppos)
6658 {
6659 int *valp = ctl->data;
6660 int val = *valp;
6661 loff_t pos = *ppos;
6662 struct ctl_table lctl;
6663 int ret;
6664
6665 lctl = *ctl;
6666 lctl.data = &val;
6667 ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
6668
6669 if (write && (*valp != val))
6670 ret = addrconf_disable_policy(ctl, valp, val);
6671
6672 if (ret)
6673 *ppos = pos;
6674
6675 return ret;
6676 }
6677
6678 static int minus_one = -1;
6679 static const int two_five_five = 255;
6680 static u32 ioam6_if_id_max = U16_MAX;
6681
6682 static const struct ctl_table addrconf_sysctl[] = {
6683 {
6684 .procname = "forwarding",
6685 .data = &ipv6_devconf.forwarding,
6686 .maxlen = sizeof(int),
6687 .mode = 0644,
6688 .proc_handler = addrconf_sysctl_forward,
6689 },
6690 {
6691 .procname = "hop_limit",
6692 .data = &ipv6_devconf.hop_limit,
6693 .maxlen = sizeof(int),
6694 .mode = 0644,
6695 .proc_handler = proc_dointvec_minmax,
6696 .extra1 = (void *)SYSCTL_ONE,
6697 .extra2 = (void *)&two_five_five,
6698 },
6699 {
6700 .procname = "mtu",
6701 .data = &ipv6_devconf.mtu6,
6702 .maxlen = sizeof(int),
6703 .mode = 0644,
6704 .proc_handler = addrconf_sysctl_mtu,
6705 },
6706 {
6707 .procname = "accept_ra",
6708 .data = &ipv6_devconf.accept_ra,
6709 .maxlen = sizeof(int),
6710 .mode = 0644,
6711 .proc_handler = proc_dointvec,
6712 },
6713 {
6714 .procname = "accept_redirects",
6715 .data = &ipv6_devconf.accept_redirects,
6716 .maxlen = sizeof(int),
6717 .mode = 0644,
6718 .proc_handler = proc_dointvec,
6719 },
6720 {
6721 .procname = "autoconf",
6722 .data = &ipv6_devconf.autoconf,
6723 .maxlen = sizeof(int),
6724 .mode = 0644,
6725 .proc_handler = proc_dointvec,
6726 },
6727 {
6728 .procname = "dad_transmits",
6729 .data = &ipv6_devconf.dad_transmits,
6730 .maxlen = sizeof(int),
6731 .mode = 0644,
6732 .proc_handler = proc_dointvec,
6733 },
6734 {
6735 .procname = "router_solicitations",
6736 .data = &ipv6_devconf.rtr_solicits,
6737 .maxlen = sizeof(int),
6738 .mode = 0644,
6739 .proc_handler = proc_dointvec_minmax,
6740 .extra1 = &minus_one,
6741 },
6742 {
6743 .procname = "router_solicitation_interval",
6744 .data = &ipv6_devconf.rtr_solicit_interval,
6745 .maxlen = sizeof(int),
6746 .mode = 0644,
6747 .proc_handler = proc_dointvec_jiffies,
6748 },
6749 {
6750 .procname = "router_solicitation_max_interval",
6751 .data = &ipv6_devconf.rtr_solicit_max_interval,
6752 .maxlen = sizeof(int),
6753 .mode = 0644,
6754 .proc_handler = proc_dointvec_jiffies,
6755 },
6756 {
6757 .procname = "router_solicitation_delay",
6758 .data = &ipv6_devconf.rtr_solicit_delay,
6759 .maxlen = sizeof(int),
6760 .mode = 0644,
6761 .proc_handler = proc_dointvec_jiffies,
6762 },
6763 {
6764 .procname = "force_mld_version",
6765 .data = &ipv6_devconf.force_mld_version,
6766 .maxlen = sizeof(int),
6767 .mode = 0644,
6768 .proc_handler = proc_dointvec,
6769 },
6770 {
6771 .procname = "mldv1_unsolicited_report_interval",
6772 .data =
6773 &ipv6_devconf.mldv1_unsolicited_report_interval,
6774 .maxlen = sizeof(int),
6775 .mode = 0644,
6776 .proc_handler = proc_dointvec_ms_jiffies,
6777 },
6778 {
6779 .procname = "mldv2_unsolicited_report_interval",
6780 .data =
6781 &ipv6_devconf.mldv2_unsolicited_report_interval,
6782 .maxlen = sizeof(int),
6783 .mode = 0644,
6784 .proc_handler = proc_dointvec_ms_jiffies,
6785 },
6786 {
6787 .procname = "use_tempaddr",
6788 .data = &ipv6_devconf.use_tempaddr,
6789 .maxlen = sizeof(int),
6790 .mode = 0644,
6791 .proc_handler = proc_dointvec,
6792 },
6793 {
6794 .procname = "temp_valid_lft",
6795 .data = &ipv6_devconf.temp_valid_lft,
6796 .maxlen = sizeof(int),
6797 .mode = 0644,
6798 .proc_handler = proc_dointvec,
6799 },
6800 {
6801 .procname = "temp_prefered_lft",
6802 .data = &ipv6_devconf.temp_prefered_lft,
6803 .maxlen = sizeof(int),
6804 .mode = 0644,
6805 .proc_handler = proc_dointvec,
6806 },
6807 {
6808 .procname = "regen_max_retry",
6809 .data = &ipv6_devconf.regen_max_retry,
6810 .maxlen = sizeof(int),
6811 .mode = 0644,
6812 .proc_handler = proc_dointvec,
6813 },
6814 {
6815 .procname = "max_desync_factor",
6816 .data = &ipv6_devconf.max_desync_factor,
6817 .maxlen = sizeof(int),
6818 .mode = 0644,
6819 .proc_handler = proc_dointvec,
6820 },
6821 {
6822 .procname = "max_addresses",
6823 .data = &ipv6_devconf.max_addresses,
6824 .maxlen = sizeof(int),
6825 .mode = 0644,
6826 .proc_handler = proc_dointvec,
6827 },
6828 {
6829 .procname = "accept_ra_defrtr",
6830 .data = &ipv6_devconf.accept_ra_defrtr,
6831 .maxlen = sizeof(int),
6832 .mode = 0644,
6833 .proc_handler = proc_dointvec,
6834 },
6835 {
6836 .procname = "ra_defrtr_metric",
6837 .data = &ipv6_devconf.ra_defrtr_metric,
6838 .maxlen = sizeof(u32),
6839 .mode = 0644,
6840 .proc_handler = proc_douintvec_minmax,
6841 .extra1 = (void *)SYSCTL_ONE,
6842 },
6843 {
6844 .procname = "accept_ra_min_hop_limit",
6845 .data = &ipv6_devconf.accept_ra_min_hop_limit,
6846 .maxlen = sizeof(int),
6847 .mode = 0644,
6848 .proc_handler = proc_dointvec,
6849 },
6850 {
6851 .procname = "accept_ra_min_lft",
6852 .data = &ipv6_devconf.accept_ra_min_lft,
6853 .maxlen = sizeof(int),
6854 .mode = 0644,
6855 .proc_handler = proc_dointvec,
6856 },
6857 {
6858 .procname = "accept_ra_pinfo",
6859 .data = &ipv6_devconf.accept_ra_pinfo,
6860 .maxlen = sizeof(int),
6861 .mode = 0644,
6862 .proc_handler = proc_dointvec,
6863 },
6864 #ifdef CONFIG_IPV6_ROUTER_PREF
6865 {
6866 .procname = "accept_ra_rtr_pref",
6867 .data = &ipv6_devconf.accept_ra_rtr_pref,
6868 .maxlen = sizeof(int),
6869 .mode = 0644,
6870 .proc_handler = proc_dointvec,
6871 },
6872 {
6873 .procname = "router_probe_interval",
6874 .data = &ipv6_devconf.rtr_probe_interval,
6875 .maxlen = sizeof(int),
6876 .mode = 0644,
6877 .proc_handler = proc_dointvec_jiffies,
6878 },
6879 #ifdef CONFIG_IPV6_ROUTE_INFO
6880 {
6881 .procname = "accept_ra_rt_info_min_plen",
6882 .data = &ipv6_devconf.accept_ra_rt_info_min_plen,
6883 .maxlen = sizeof(int),
6884 .mode = 0644,
6885 .proc_handler = proc_dointvec,
6886 },
6887 {
6888 .procname = "accept_ra_rt_info_max_plen",
6889 .data = &ipv6_devconf.accept_ra_rt_info_max_plen,
6890 .maxlen = sizeof(int),
6891 .mode = 0644,
6892 .proc_handler = proc_dointvec,
6893 },
6894 #endif
6895 #endif
6896 {
6897 .procname = "proxy_ndp",
6898 .data = &ipv6_devconf.proxy_ndp,
6899 .maxlen = sizeof(int),
6900 .mode = 0644,
6901 .proc_handler = addrconf_sysctl_proxy_ndp,
6902 },
6903 {
6904 .procname = "accept_source_route",
6905 .data = &ipv6_devconf.accept_source_route,
6906 .maxlen = sizeof(int),
6907 .mode = 0644,
6908 .proc_handler = proc_dointvec,
6909 },
6910 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
6911 {
6912 .procname = "optimistic_dad",
6913 .data = &ipv6_devconf.optimistic_dad,
6914 .maxlen = sizeof(int),
6915 .mode = 0644,
6916 .proc_handler = proc_dointvec,
6917 },
6918 {
6919 .procname = "use_optimistic",
6920 .data = &ipv6_devconf.use_optimistic,
6921 .maxlen = sizeof(int),
6922 .mode = 0644,
6923 .proc_handler = proc_dointvec,
6924 },
6925 #endif
6926 #ifdef CONFIG_IPV6_MROUTE
6927 {
6928 .procname = "mc_forwarding",
6929 .data = &ipv6_devconf.mc_forwarding,
6930 .maxlen = sizeof(int),
6931 .mode = 0444,
6932 .proc_handler = proc_dointvec,
6933 },
6934 #endif
6935 {
6936 .procname = "disable_ipv6",
6937 .data = &ipv6_devconf.disable_ipv6,
6938 .maxlen = sizeof(int),
6939 .mode = 0644,
6940 .proc_handler = addrconf_sysctl_disable,
6941 },
6942 {
6943 .procname = "accept_dad",
6944 .data = &ipv6_devconf.accept_dad,
6945 .maxlen = sizeof(int),
6946 .mode = 0644,
6947 .proc_handler = proc_dointvec,
6948 },
6949 {
6950 .procname = "force_tllao",
6951 .data = &ipv6_devconf.force_tllao,
6952 .maxlen = sizeof(int),
6953 .mode = 0644,
6954 .proc_handler = proc_dointvec
6955 },
6956 {
6957 .procname = "ndisc_notify",
6958 .data = &ipv6_devconf.ndisc_notify,
6959 .maxlen = sizeof(int),
6960 .mode = 0644,
6961 .proc_handler = proc_dointvec
6962 },
6963 {
6964 .procname = "suppress_frag_ndisc",
6965 .data = &ipv6_devconf.suppress_frag_ndisc,
6966 .maxlen = sizeof(int),
6967 .mode = 0644,
6968 .proc_handler = proc_dointvec
6969 },
6970 {
6971 .procname = "accept_ra_from_local",
6972 .data = &ipv6_devconf.accept_ra_from_local,
6973 .maxlen = sizeof(int),
6974 .mode = 0644,
6975 .proc_handler = proc_dointvec,
6976 },
6977 {
6978 .procname = "accept_ra_mtu",
6979 .data = &ipv6_devconf.accept_ra_mtu,
6980 .maxlen = sizeof(int),
6981 .mode = 0644,
6982 .proc_handler = proc_dointvec,
6983 },
6984 {
6985 .procname = "stable_secret",
6986 .data = &ipv6_devconf.stable_secret,
6987 .maxlen = IPV6_MAX_STRLEN,
6988 .mode = 0600,
6989 .proc_handler = addrconf_sysctl_stable_secret,
6990 },
6991 {
6992 .procname = "use_oif_addrs_only",
6993 .data = &ipv6_devconf.use_oif_addrs_only,
6994 .maxlen = sizeof(int),
6995 .mode = 0644,
6996 .proc_handler = proc_dointvec,
6997 },
6998 {
6999 .procname = "ignore_routes_with_linkdown",
7000 .data = &ipv6_devconf.ignore_routes_with_linkdown,
7001 .maxlen = sizeof(int),
7002 .mode = 0644,
7003 .proc_handler = addrconf_sysctl_ignore_routes_with_linkdown,
7004 },
7005 {
7006 .procname = "drop_unicast_in_l2_multicast",
7007 .data = &ipv6_devconf.drop_unicast_in_l2_multicast,
7008 .maxlen = sizeof(int),
7009 .mode = 0644,
7010 .proc_handler = proc_dointvec,
7011 },
7012 {
7013 .procname = "drop_unsolicited_na",
7014 .data = &ipv6_devconf.drop_unsolicited_na,
7015 .maxlen = sizeof(int),
7016 .mode = 0644,
7017 .proc_handler = proc_dointvec,
7018 },
7019 {
7020 .procname = "keep_addr_on_down",
7021 .data = &ipv6_devconf.keep_addr_on_down,
7022 .maxlen = sizeof(int),
7023 .mode = 0644,
7024 .proc_handler = proc_dointvec,
7025
7026 },
7027 {
7028 .procname = "seg6_enabled",
7029 .data = &ipv6_devconf.seg6_enabled,
7030 .maxlen = sizeof(int),
7031 .mode = 0644,
7032 .proc_handler = proc_dointvec,
7033 },
7034 #ifdef CONFIG_IPV6_SEG6_HMAC
7035 {
7036 .procname = "seg6_require_hmac",
7037 .data = &ipv6_devconf.seg6_require_hmac,
7038 .maxlen = sizeof(int),
7039 .mode = 0644,
7040 .proc_handler = proc_dointvec,
7041 },
7042 #endif
7043 {
7044 .procname = "enhanced_dad",
7045 .data = &ipv6_devconf.enhanced_dad,
7046 .maxlen = sizeof(int),
7047 .mode = 0644,
7048 .proc_handler = proc_dointvec,
7049 },
7050 {
7051 .procname = "addr_gen_mode",
7052 .data = &ipv6_devconf.addr_gen_mode,
7053 .maxlen = sizeof(int),
7054 .mode = 0644,
7055 .proc_handler = addrconf_sysctl_addr_gen_mode,
7056 },
7057 {
7058 .procname = "disable_policy",
7059 .data = &ipv6_devconf.disable_policy,
7060 .maxlen = sizeof(int),
7061 .mode = 0644,
7062 .proc_handler = addrconf_sysctl_disable_policy,
7063 },
7064 {
7065 .procname = "ndisc_tclass",
7066 .data = &ipv6_devconf.ndisc_tclass,
7067 .maxlen = sizeof(int),
7068 .mode = 0644,
7069 .proc_handler = proc_dointvec_minmax,
7070 .extra1 = (void *)SYSCTL_ZERO,
7071 .extra2 = (void *)&two_five_five,
7072 },
7073 {
7074 .procname = "rpl_seg_enabled",
7075 .data = &ipv6_devconf.rpl_seg_enabled,
7076 .maxlen = sizeof(int),
7077 .mode = 0644,
7078 .proc_handler = proc_dointvec,
7079 },
7080 {
7081 .procname = "ioam6_enabled",
7082 .data = &ipv6_devconf.ioam6_enabled,
7083 .maxlen = sizeof(u8),
7084 .mode = 0644,
7085 .proc_handler = proc_dou8vec_minmax,
7086 .extra1 = (void *)SYSCTL_ZERO,
7087 .extra2 = (void *)SYSCTL_ONE,
7088 },
7089 {
7090 .procname = "ioam6_id",
7091 .data = &ipv6_devconf.ioam6_id,
7092 .maxlen = sizeof(u32),
7093 .mode = 0644,
7094 .proc_handler = proc_douintvec_minmax,
7095 .extra1 = (void *)SYSCTL_ZERO,
7096 .extra2 = (void *)&ioam6_if_id_max,
7097 },
7098 {
7099 .procname = "ioam6_id_wide",
7100 .data = &ipv6_devconf.ioam6_id_wide,
7101 .maxlen = sizeof(u32),
7102 .mode = 0644,
7103 .proc_handler = proc_douintvec,
7104 },
7105 {
7106 .procname = "ndisc_evict_nocarrier",
7107 .data = &ipv6_devconf.ndisc_evict_nocarrier,
7108 .maxlen = sizeof(u8),
7109 .mode = 0644,
7110 .proc_handler = proc_dou8vec_minmax,
7111 .extra1 = (void *)SYSCTL_ZERO,
7112 .extra2 = (void *)SYSCTL_ONE,
7113 },
7114 {
7115 .procname = "accept_untracked_na",
7116 .data = &ipv6_devconf.accept_untracked_na,
7117 .maxlen = sizeof(int),
7118 .mode = 0644,
7119 .proc_handler = proc_dointvec_minmax,
7120 .extra1 = SYSCTL_ZERO,
7121 .extra2 = SYSCTL_TWO,
7122 },
7123 {
7124 /* sentinel */
7125 }
7126 };
7127
__addrconf_sysctl_register(struct net * net,char * dev_name,struct inet6_dev * idev,struct ipv6_devconf * p)7128 static int __addrconf_sysctl_register(struct net *net, char *dev_name,
7129 struct inet6_dev *idev, struct ipv6_devconf *p)
7130 {
7131 int i, ifindex;
7132 struct ctl_table *table;
7133 char path[sizeof("net/ipv6/conf/") + IFNAMSIZ];
7134
7135 table = kmemdup(addrconf_sysctl, sizeof(addrconf_sysctl), GFP_KERNEL_ACCOUNT);
7136 if (!table)
7137 goto out;
7138
7139 for (i = 0; table[i].data; i++) {
7140 table[i].data += (char *)p - (char *)&ipv6_devconf;
7141 /* If one of these is already set, then it is not safe to
7142 * overwrite either of them: this makes proc_dointvec_minmax
7143 * usable.
7144 */
7145 if (!table[i].extra1 && !table[i].extra2) {
7146 table[i].extra1 = idev; /* embedded; no ref */
7147 table[i].extra2 = net;
7148 }
7149 }
7150
7151 snprintf(path, sizeof(path), "net/ipv6/conf/%s", dev_name);
7152
7153 p->sysctl_header = register_net_sysctl_sz(net, path, table,
7154 ARRAY_SIZE(addrconf_sysctl));
7155 if (!p->sysctl_header)
7156 goto free;
7157
7158 if (!strcmp(dev_name, "all"))
7159 ifindex = NETCONFA_IFINDEX_ALL;
7160 else if (!strcmp(dev_name, "default"))
7161 ifindex = NETCONFA_IFINDEX_DEFAULT;
7162 else
7163 ifindex = idev->dev->ifindex;
7164 inet6_netconf_notify_devconf(net, RTM_NEWNETCONF, NETCONFA_ALL,
7165 ifindex, p);
7166 return 0;
7167
7168 free:
7169 kfree(table);
7170 out:
7171 return -ENOBUFS;
7172 }
7173
__addrconf_sysctl_unregister(struct net * net,struct ipv6_devconf * p,int ifindex)7174 static void __addrconf_sysctl_unregister(struct net *net,
7175 struct ipv6_devconf *p, int ifindex)
7176 {
7177 struct ctl_table *table;
7178
7179 if (!p->sysctl_header)
7180 return;
7181
7182 table = p->sysctl_header->ctl_table_arg;
7183 unregister_net_sysctl_table(p->sysctl_header);
7184 p->sysctl_header = NULL;
7185 kfree(table);
7186
7187 inet6_netconf_notify_devconf(net, RTM_DELNETCONF, 0, ifindex, NULL);
7188 }
7189
addrconf_sysctl_register(struct inet6_dev * idev)7190 static int addrconf_sysctl_register(struct inet6_dev *idev)
7191 {
7192 int err;
7193
7194 if (!sysctl_dev_name_is_allowed(idev->dev->name))
7195 return -EINVAL;
7196
7197 err = neigh_sysctl_register(idev->dev, idev->nd_parms,
7198 &ndisc_ifinfo_sysctl_change);
7199 if (err)
7200 return err;
7201 err = __addrconf_sysctl_register(dev_net(idev->dev), idev->dev->name,
7202 idev, &idev->cnf);
7203 if (err)
7204 neigh_sysctl_unregister(idev->nd_parms);
7205
7206 return err;
7207 }
7208
addrconf_sysctl_unregister(struct inet6_dev * idev)7209 static void addrconf_sysctl_unregister(struct inet6_dev *idev)
7210 {
7211 __addrconf_sysctl_unregister(dev_net(idev->dev), &idev->cnf,
7212 idev->dev->ifindex);
7213 neigh_sysctl_unregister(idev->nd_parms);
7214 }
7215
7216
7217 #endif
7218
addrconf_init_net(struct net * net)7219 static int __net_init addrconf_init_net(struct net *net)
7220 {
7221 int err = -ENOMEM;
7222 struct ipv6_devconf *all, *dflt;
7223
7224 spin_lock_init(&net->ipv6.addrconf_hash_lock);
7225 INIT_DEFERRABLE_WORK(&net->ipv6.addr_chk_work, addrconf_verify_work);
7226 net->ipv6.inet6_addr_lst = kcalloc(IN6_ADDR_HSIZE,
7227 sizeof(struct hlist_head),
7228 GFP_KERNEL);
7229 if (!net->ipv6.inet6_addr_lst)
7230 goto err_alloc_addr;
7231
7232 all = kmemdup(&ipv6_devconf, sizeof(ipv6_devconf), GFP_KERNEL);
7233 if (!all)
7234 goto err_alloc_all;
7235
7236 dflt = kmemdup(&ipv6_devconf_dflt, sizeof(ipv6_devconf_dflt), GFP_KERNEL);
7237 if (!dflt)
7238 goto err_alloc_dflt;
7239
7240 if (!net_eq(net, &init_net)) {
7241 switch (net_inherit_devconf()) {
7242 case 1: /* copy from init_net */
7243 memcpy(all, init_net.ipv6.devconf_all,
7244 sizeof(ipv6_devconf));
7245 memcpy(dflt, init_net.ipv6.devconf_dflt,
7246 sizeof(ipv6_devconf_dflt));
7247 break;
7248 case 3: /* copy from the current netns */
7249 memcpy(all, current->nsproxy->net_ns->ipv6.devconf_all,
7250 sizeof(ipv6_devconf));
7251 memcpy(dflt,
7252 current->nsproxy->net_ns->ipv6.devconf_dflt,
7253 sizeof(ipv6_devconf_dflt));
7254 break;
7255 case 0:
7256 case 2:
7257 /* use compiled values */
7258 break;
7259 }
7260 }
7261
7262 /* these will be inherited by all namespaces */
7263 dflt->autoconf = ipv6_defaults.autoconf;
7264 dflt->disable_ipv6 = ipv6_defaults.disable_ipv6;
7265
7266 dflt->stable_secret.initialized = false;
7267 all->stable_secret.initialized = false;
7268
7269 net->ipv6.devconf_all = all;
7270 net->ipv6.devconf_dflt = dflt;
7271
7272 #ifdef CONFIG_SYSCTL
7273 err = __addrconf_sysctl_register(net, "all", NULL, all);
7274 if (err < 0)
7275 goto err_reg_all;
7276
7277 err = __addrconf_sysctl_register(net, "default", NULL, dflt);
7278 if (err < 0)
7279 goto err_reg_dflt;
7280 #endif
7281 return 0;
7282
7283 #ifdef CONFIG_SYSCTL
7284 err_reg_dflt:
7285 __addrconf_sysctl_unregister(net, all, NETCONFA_IFINDEX_ALL);
7286 err_reg_all:
7287 kfree(dflt);
7288 net->ipv6.devconf_dflt = NULL;
7289 #endif
7290 err_alloc_dflt:
7291 kfree(all);
7292 net->ipv6.devconf_all = NULL;
7293 err_alloc_all:
7294 kfree(net->ipv6.inet6_addr_lst);
7295 err_alloc_addr:
7296 return err;
7297 }
7298
addrconf_exit_net(struct net * net)7299 static void __net_exit addrconf_exit_net(struct net *net)
7300 {
7301 int i;
7302
7303 #ifdef CONFIG_SYSCTL
7304 __addrconf_sysctl_unregister(net, net->ipv6.devconf_dflt,
7305 NETCONFA_IFINDEX_DEFAULT);
7306 __addrconf_sysctl_unregister(net, net->ipv6.devconf_all,
7307 NETCONFA_IFINDEX_ALL);
7308 #endif
7309 kfree(net->ipv6.devconf_dflt);
7310 net->ipv6.devconf_dflt = NULL;
7311 kfree(net->ipv6.devconf_all);
7312 net->ipv6.devconf_all = NULL;
7313
7314 cancel_delayed_work_sync(&net->ipv6.addr_chk_work);
7315 /*
7316 * Check hash table, then free it.
7317 */
7318 for (i = 0; i < IN6_ADDR_HSIZE; i++)
7319 WARN_ON_ONCE(!hlist_empty(&net->ipv6.inet6_addr_lst[i]));
7320
7321 kfree(net->ipv6.inet6_addr_lst);
7322 net->ipv6.inet6_addr_lst = NULL;
7323 }
7324
7325 static struct pernet_operations addrconf_ops = {
7326 .init = addrconf_init_net,
7327 .exit = addrconf_exit_net,
7328 };
7329
7330 static struct rtnl_af_ops inet6_ops __read_mostly = {
7331 .family = AF_INET6,
7332 .fill_link_af = inet6_fill_link_af,
7333 .get_link_af_size = inet6_get_link_af_size,
7334 .validate_link_af = inet6_validate_link_af,
7335 .set_link_af = inet6_set_link_af,
7336 };
7337
7338 /*
7339 * Init / cleanup code
7340 */
7341
addrconf_init(void)7342 int __init addrconf_init(void)
7343 {
7344 struct inet6_dev *idev;
7345 int err;
7346
7347 err = ipv6_addr_label_init();
7348 if (err < 0) {
7349 pr_crit("%s: cannot initialize default policy table: %d\n",
7350 __func__, err);
7351 goto out;
7352 }
7353
7354 err = register_pernet_subsys(&addrconf_ops);
7355 if (err < 0)
7356 goto out_addrlabel;
7357
7358 addrconf_wq = create_workqueue("ipv6_addrconf");
7359 if (!addrconf_wq) {
7360 err = -ENOMEM;
7361 goto out_nowq;
7362 }
7363
7364 rtnl_lock();
7365 idev = ipv6_add_dev(blackhole_netdev);
7366 rtnl_unlock();
7367 if (IS_ERR(idev)) {
7368 err = PTR_ERR(idev);
7369 goto errlo;
7370 }
7371
7372 ip6_route_init_special_entries();
7373
7374 register_netdevice_notifier(&ipv6_dev_notf);
7375
7376 addrconf_verify(&init_net);
7377
7378 rtnl_af_register(&inet6_ops);
7379
7380 err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETLINK,
7381 NULL, inet6_dump_ifinfo, 0);
7382 if (err < 0)
7383 goto errout;
7384
7385 err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_NEWADDR,
7386 inet6_rtm_newaddr, NULL, 0);
7387 if (err < 0)
7388 goto errout;
7389 err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_DELADDR,
7390 inet6_rtm_deladdr, NULL, 0);
7391 if (err < 0)
7392 goto errout;
7393 err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETADDR,
7394 inet6_rtm_getaddr, inet6_dump_ifaddr,
7395 RTNL_FLAG_DOIT_UNLOCKED);
7396 if (err < 0)
7397 goto errout;
7398 err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETMULTICAST,
7399 NULL, inet6_dump_ifmcaddr, 0);
7400 if (err < 0)
7401 goto errout;
7402 err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETANYCAST,
7403 NULL, inet6_dump_ifacaddr, 0);
7404 if (err < 0)
7405 goto errout;
7406 err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETNETCONF,
7407 inet6_netconf_get_devconf,
7408 inet6_netconf_dump_devconf,
7409 RTNL_FLAG_DOIT_UNLOCKED);
7410 if (err < 0)
7411 goto errout;
7412 err = ipv6_addr_label_rtnl_register();
7413 if (err < 0)
7414 goto errout;
7415
7416 return 0;
7417 errout:
7418 rtnl_unregister_all(PF_INET6);
7419 rtnl_af_unregister(&inet6_ops);
7420 unregister_netdevice_notifier(&ipv6_dev_notf);
7421 errlo:
7422 destroy_workqueue(addrconf_wq);
7423 out_nowq:
7424 unregister_pernet_subsys(&addrconf_ops);
7425 out_addrlabel:
7426 ipv6_addr_label_cleanup();
7427 out:
7428 return err;
7429 }
7430
addrconf_cleanup(void)7431 void addrconf_cleanup(void)
7432 {
7433 struct net_device *dev;
7434
7435 unregister_netdevice_notifier(&ipv6_dev_notf);
7436 unregister_pernet_subsys(&addrconf_ops);
7437 ipv6_addr_label_cleanup();
7438
7439 rtnl_af_unregister(&inet6_ops);
7440
7441 rtnl_lock();
7442
7443 /* clean dev list */
7444 for_each_netdev(&init_net, dev) {
7445 if (__in6_dev_get(dev) == NULL)
7446 continue;
7447 addrconf_ifdown(dev, true);
7448 }
7449 addrconf_ifdown(init_net.loopback_dev, true);
7450
7451 rtnl_unlock();
7452
7453 destroy_workqueue(addrconf_wq);
7454 }
7455