1 // SPDX-License-Identifier: GPL-2.0-or-later
2 /*
3 * IPv6 Address [auto]configuration
4 * Linux INET6 implementation
5 *
6 * Authors:
7 * Pedro Roque <roque@di.fc.ul.pt>
8 * Alexey Kuznetsov <kuznet@ms2.inr.ac.ru>
9 */
10
11 /*
12 * Changes:
13 *
14 * Janos Farkas : delete timer on ifdown
15 * <chexum@bankinf.banki.hu>
16 * Andi Kleen : kill double kfree on module
17 * unload.
18 * Maciej W. Rozycki : FDDI support
19 * sekiya@USAGI : Don't send too many RS
20 * packets.
21 * yoshfuji@USAGI : Fixed interval between DAD
22 * packets.
23 * YOSHIFUJI Hideaki @USAGI : improved accuracy of
24 * address validation timer.
25 * YOSHIFUJI Hideaki @USAGI : Privacy Extensions (RFC3041)
26 * support.
27 * Yuji SEKIYA @USAGI : Don't assign a same IPv6
28 * address on a same interface.
29 * YOSHIFUJI Hideaki @USAGI : ARCnet support
30 * YOSHIFUJI Hideaki @USAGI : convert /proc/net/if_inet6 to
31 * seq_file.
32 * YOSHIFUJI Hideaki @USAGI : improved source address
33 * selection; consider scope,
34 * status etc.
35 */
36
37 #define pr_fmt(fmt) "IPv6: " fmt
38
39 #include <linux/errno.h>
40 #include <linux/types.h>
41 #include <linux/kernel.h>
42 #include <linux/sched/signal.h>
43 #include <linux/socket.h>
44 #include <linux/sockios.h>
45 #include <linux/net.h>
46 #include <linux/inet.h>
47 #include <linux/in6.h>
48 #include <linux/netdevice.h>
49 #include <linux/if_addr.h>
50 #include <linux/if_arp.h>
51 #include <linux/if_arcnet.h>
52 #include <linux/if_infiniband.h>
53 #include <linux/route.h>
54 #include <linux/inetdevice.h>
55 #include <linux/init.h>
56 #include <linux/slab.h>
57 #ifdef CONFIG_SYSCTL
58 #include <linux/sysctl.h>
59 #endif
60 #include <linux/capability.h>
61 #include <linux/delay.h>
62 #include <linux/notifier.h>
63 #include <linux/string.h>
64 #include <linux/hash.h>
65
66 #include <net/ip_tunnels.h>
67 #include <net/net_namespace.h>
68 #include <net/sock.h>
69 #include <net/snmp.h>
70
71 #include <net/6lowpan.h>
72 #include <net/firewire.h>
73 #include <net/ipv6.h>
74 #include <net/protocol.h>
75 #include <net/ndisc.h>
76 #include <net/ip6_route.h>
77 #include <net/addrconf.h>
78 #include <net/tcp.h>
79 #include <net/ip.h>
80 #include <net/netlink.h>
81 #include <net/pkt_sched.h>
82 #include <net/l3mdev.h>
83 #include <net/netdev_lock.h>
84 #include <linux/if_tunnel.h>
85 #include <linux/rtnetlink.h>
86 #include <linux/netconf.h>
87 #include <linux/random.h>
88 #include <linux/uaccess.h>
89 #include <linux/unaligned.h>
90
91 #include <linux/proc_fs.h>
92 #include <linux/seq_file.h>
93 #include <linux/export.h>
94 #include <linux/ioam6.h>
95
96 #define IPV6_MAX_STRLEN \
97 sizeof("ffff:ffff:ffff:ffff:ffff:ffff:255.255.255.255")
98
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_min_advance = REGEN_MIN_ADVANCE,
199 .regen_max_retry = REGEN_MAX_RETRY,
200 .max_desync_factor = MAX_DESYNC_FACTOR,
201 .max_addresses = IPV6_MAX_ADDRESSES,
202 .accept_ra_defrtr = 1,
203 .ra_defrtr_metric = IP6_RT_PRIO_USER,
204 .accept_ra_from_local = 0,
205 .accept_ra_min_hop_limit= 1,
206 .accept_ra_min_lft = 0,
207 .accept_ra_pinfo = 1,
208 #ifdef CONFIG_IPV6_ROUTER_PREF
209 .accept_ra_rtr_pref = 1,
210 .rtr_probe_interval = 60 * HZ,
211 #ifdef CONFIG_IPV6_ROUTE_INFO
212 .accept_ra_rt_info_min_plen = 0,
213 .accept_ra_rt_info_max_plen = 0,
214 #endif
215 #endif
216 .proxy_ndp = 0,
217 .accept_source_route = 0, /* we do not accept RH0 by default. */
218 .disable_ipv6 = 0,
219 .accept_dad = 0,
220 .suppress_frag_ndisc = 1,
221 .accept_ra_mtu = 1,
222 .stable_secret = {
223 .initialized = false,
224 },
225 .use_oif_addrs_only = 0,
226 .ignore_routes_with_linkdown = 0,
227 .keep_addr_on_down = 0,
228 .seg6_enabled = 0,
229 #ifdef CONFIG_IPV6_SEG6_HMAC
230 .seg6_require_hmac = 0,
231 #endif
232 .enhanced_dad = 1,
233 .addr_gen_mode = IN6_ADDR_GEN_MODE_EUI64,
234 .disable_policy = 0,
235 .rpl_seg_enabled = 0,
236 .ioam6_enabled = 0,
237 .ioam6_id = IOAM6_DEFAULT_IF_ID,
238 .ioam6_id_wide = IOAM6_DEFAULT_IF_ID_WIDE,
239 .ndisc_evict_nocarrier = 1,
240 .ra_honor_pio_life = 0,
241 .ra_honor_pio_pflag = 0,
242 .force_forwarding = 0,
243 };
244
245 static struct ipv6_devconf ipv6_devconf_dflt __read_mostly = {
246 .forwarding = 0,
247 .hop_limit = IPV6_DEFAULT_HOPLIMIT,
248 .mtu6 = IPV6_MIN_MTU,
249 .accept_ra = 1,
250 .accept_redirects = 1,
251 .autoconf = 1,
252 .force_mld_version = 0,
253 .mldv1_unsolicited_report_interval = 10 * HZ,
254 .mldv2_unsolicited_report_interval = HZ,
255 .dad_transmits = 1,
256 .rtr_solicits = MAX_RTR_SOLICITATIONS,
257 .rtr_solicit_interval = RTR_SOLICITATION_INTERVAL,
258 .rtr_solicit_max_interval = RTR_SOLICITATION_MAX_INTERVAL,
259 .rtr_solicit_delay = MAX_RTR_SOLICITATION_DELAY,
260 .use_tempaddr = 0,
261 .temp_valid_lft = TEMP_VALID_LIFETIME,
262 .temp_prefered_lft = TEMP_PREFERRED_LIFETIME,
263 .regen_min_advance = REGEN_MIN_ADVANCE,
264 .regen_max_retry = REGEN_MAX_RETRY,
265 .max_desync_factor = MAX_DESYNC_FACTOR,
266 .max_addresses = IPV6_MAX_ADDRESSES,
267 .accept_ra_defrtr = 1,
268 .ra_defrtr_metric = IP6_RT_PRIO_USER,
269 .accept_ra_from_local = 0,
270 .accept_ra_min_hop_limit= 1,
271 .accept_ra_min_lft = 0,
272 .accept_ra_pinfo = 1,
273 #ifdef CONFIG_IPV6_ROUTER_PREF
274 .accept_ra_rtr_pref = 1,
275 .rtr_probe_interval = 60 * HZ,
276 #ifdef CONFIG_IPV6_ROUTE_INFO
277 .accept_ra_rt_info_min_plen = 0,
278 .accept_ra_rt_info_max_plen = 0,
279 #endif
280 #endif
281 .proxy_ndp = 0,
282 .accept_source_route = 0, /* we do not accept RH0 by default. */
283 .disable_ipv6 = 0,
284 .accept_dad = 1,
285 .suppress_frag_ndisc = 1,
286 .accept_ra_mtu = 1,
287 .stable_secret = {
288 .initialized = false,
289 },
290 .use_oif_addrs_only = 0,
291 .ignore_routes_with_linkdown = 0,
292 .keep_addr_on_down = 0,
293 .seg6_enabled = 0,
294 #ifdef CONFIG_IPV6_SEG6_HMAC
295 .seg6_require_hmac = 0,
296 #endif
297 .enhanced_dad = 1,
298 .addr_gen_mode = IN6_ADDR_GEN_MODE_EUI64,
299 .disable_policy = 0,
300 .rpl_seg_enabled = 0,
301 .ioam6_enabled = 0,
302 .ioam6_id = IOAM6_DEFAULT_IF_ID,
303 .ioam6_id_wide = IOAM6_DEFAULT_IF_ID_WIDE,
304 .ndisc_evict_nocarrier = 1,
305 .ra_honor_pio_life = 0,
306 .ra_honor_pio_pflag = 0,
307 .force_forwarding = 0,
308 };
309
310 /* Check if link is ready: is it up and is a valid qdisc available */
addrconf_link_ready(const struct net_device * dev)311 static inline bool addrconf_link_ready(const struct net_device *dev)
312 {
313 return netif_oper_up(dev) && !qdisc_tx_is_noop(dev);
314 }
315
addrconf_del_rs_timer(struct inet6_dev * idev)316 static void addrconf_del_rs_timer(struct inet6_dev *idev)
317 {
318 if (timer_delete(&idev->rs_timer))
319 __in6_dev_put(idev);
320 }
321
addrconf_del_dad_work(struct inet6_ifaddr * ifp)322 static void addrconf_del_dad_work(struct inet6_ifaddr *ifp)
323 {
324 if (cancel_delayed_work(&ifp->dad_work))
325 __in6_ifa_put(ifp);
326 }
327
addrconf_mod_rs_timer(struct inet6_dev * idev,unsigned long when)328 static void addrconf_mod_rs_timer(struct inet6_dev *idev,
329 unsigned long when)
330 {
331 if (!mod_timer(&idev->rs_timer, jiffies + when))
332 in6_dev_hold(idev);
333 }
334
addrconf_mod_dad_work(struct inet6_ifaddr * ifp,unsigned long delay)335 static void addrconf_mod_dad_work(struct inet6_ifaddr *ifp,
336 unsigned long delay)
337 {
338 in6_ifa_hold(ifp);
339 if (mod_delayed_work(addrconf_wq, &ifp->dad_work, delay))
340 in6_ifa_put(ifp);
341 }
342
snmp6_alloc_dev(struct inet6_dev * idev)343 static int snmp6_alloc_dev(struct inet6_dev *idev)
344 {
345 int i;
346
347 idev->stats.ipv6 = alloc_percpu_gfp(struct ipstats_mib, GFP_KERNEL_ACCOUNT);
348 if (!idev->stats.ipv6)
349 goto err_ip;
350
351 for_each_possible_cpu(i) {
352 struct ipstats_mib *addrconf_stats;
353 addrconf_stats = per_cpu_ptr(idev->stats.ipv6, i);
354 u64_stats_init(&addrconf_stats->syncp);
355 }
356
357
358 idev->stats.icmpv6dev = kzalloc(sizeof(struct icmpv6_mib_device),
359 GFP_KERNEL);
360 if (!idev->stats.icmpv6dev)
361 goto err_icmp;
362 idev->stats.icmpv6msgdev = kzalloc(sizeof(struct icmpv6msg_mib_device),
363 GFP_KERNEL_ACCOUNT);
364 if (!idev->stats.icmpv6msgdev)
365 goto err_icmpmsg;
366
367 return 0;
368
369 err_icmpmsg:
370 kfree(idev->stats.icmpv6dev);
371 err_icmp:
372 free_percpu(idev->stats.ipv6);
373 err_ip:
374 return -ENOMEM;
375 }
376
ipv6_add_dev(struct net_device * dev)377 static struct inet6_dev *ipv6_add_dev(struct net_device *dev)
378 {
379 struct inet6_dev *ndev;
380 int err = -ENOMEM;
381
382 ASSERT_RTNL();
383 netdev_ops_assert_locked(dev);
384
385 if (dev->mtu < IPV6_MIN_MTU && dev != blackhole_netdev)
386 return ERR_PTR(-EINVAL);
387
388 ndev = kzalloc(sizeof(*ndev), GFP_KERNEL_ACCOUNT);
389 if (!ndev)
390 return ERR_PTR(err);
391
392 rwlock_init(&ndev->lock);
393 ndev->dev = dev;
394 INIT_LIST_HEAD(&ndev->addr_list);
395 timer_setup(&ndev->rs_timer, addrconf_rs_timer, 0);
396 memcpy(&ndev->cnf, dev_net(dev)->ipv6.devconf_dflt, sizeof(ndev->cnf));
397
398 if (ndev->cnf.stable_secret.initialized)
399 ndev->cnf.addr_gen_mode = IN6_ADDR_GEN_MODE_STABLE_PRIVACY;
400
401 ndev->cnf.mtu6 = dev->mtu;
402 ndev->ra_mtu = 0;
403 ndev->nd_parms = neigh_parms_alloc(dev, &nd_tbl);
404 if (!ndev->nd_parms) {
405 kfree(ndev);
406 return ERR_PTR(err);
407 }
408 if (ndev->cnf.forwarding)
409 netif_disable_lro(dev);
410 /* We refer to the device */
411 netdev_hold(dev, &ndev->dev_tracker, GFP_KERNEL);
412
413 if (snmp6_alloc_dev(ndev) < 0) {
414 netdev_dbg(dev, "%s: cannot allocate memory for statistics\n",
415 __func__);
416 neigh_parms_release(&nd_tbl, ndev->nd_parms);
417 netdev_put(dev, &ndev->dev_tracker);
418 kfree(ndev);
419 return ERR_PTR(err);
420 }
421
422 if (dev != blackhole_netdev) {
423 if (snmp6_register_dev(ndev) < 0) {
424 netdev_dbg(dev, "%s: cannot create /proc/net/dev_snmp6/%s\n",
425 __func__, dev->name);
426 goto err_release;
427 }
428 }
429 /* One reference from device. */
430 refcount_set(&ndev->refcnt, 1);
431
432 if (dev->flags & (IFF_NOARP | IFF_LOOPBACK))
433 ndev->cnf.accept_dad = -1;
434
435 #if IS_ENABLED(CONFIG_IPV6_SIT)
436 if (dev->type == ARPHRD_SIT && (dev->priv_flags & IFF_ISATAP)) {
437 pr_info("%s: Disabled Multicast RS\n", dev->name);
438 ndev->cnf.rtr_solicits = 0;
439 }
440 #endif
441
442 INIT_LIST_HEAD(&ndev->tempaddr_list);
443 ndev->desync_factor = U32_MAX;
444 if ((dev->flags&IFF_LOOPBACK) ||
445 dev->type == ARPHRD_TUNNEL ||
446 dev->type == ARPHRD_TUNNEL6 ||
447 dev->type == ARPHRD_SIT ||
448 dev->type == ARPHRD_NONE) {
449 ndev->cnf.use_tempaddr = -1;
450 }
451
452 ndev->token = in6addr_any;
453
454 if (netif_running(dev) && addrconf_link_ready(dev))
455 ndev->if_flags |= IF_READY;
456
457 ipv6_mc_init_dev(ndev);
458 ndev->tstamp = jiffies;
459 if (dev != blackhole_netdev) {
460 err = addrconf_sysctl_register(ndev);
461 if (err) {
462 ipv6_mc_destroy_dev(ndev);
463 snmp6_unregister_dev(ndev);
464 goto err_release;
465 }
466 }
467 /* protected by rtnl_lock */
468 rcu_assign_pointer(dev->ip6_ptr, ndev);
469
470 if (dev != blackhole_netdev) {
471 /* Join interface-local all-node multicast group */
472 ipv6_dev_mc_inc(dev, &in6addr_interfacelocal_allnodes);
473
474 /* Join all-node multicast group */
475 ipv6_dev_mc_inc(dev, &in6addr_linklocal_allnodes);
476
477 /* Join all-router multicast group if forwarding is set */
478 if (ndev->cnf.forwarding && (dev->flags & IFF_MULTICAST))
479 ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
480 }
481 return ndev;
482
483 err_release:
484 neigh_parms_release(&nd_tbl, ndev->nd_parms);
485 ndev->dead = 1;
486 in6_dev_finish_destroy(ndev);
487 return ERR_PTR(err);
488 }
489
ipv6_find_idev(struct net_device * dev)490 static struct inet6_dev *ipv6_find_idev(struct net_device *dev)
491 {
492 struct inet6_dev *idev;
493
494 ASSERT_RTNL();
495
496 idev = __in6_dev_get(dev);
497 if (!idev) {
498 idev = ipv6_add_dev(dev);
499 if (IS_ERR(idev))
500 return idev;
501 }
502
503 if (dev->flags&IFF_UP)
504 ipv6_mc_up(idev);
505 return idev;
506 }
507
inet6_netconf_msgsize_devconf(int type)508 static int inet6_netconf_msgsize_devconf(int type)
509 {
510 int size = NLMSG_ALIGN(sizeof(struct netconfmsg))
511 + nla_total_size(4); /* NETCONFA_IFINDEX */
512 bool all = false;
513
514 if (type == NETCONFA_ALL)
515 all = true;
516
517 if (all || type == NETCONFA_FORWARDING)
518 size += nla_total_size(4);
519 #ifdef CONFIG_IPV6_MROUTE
520 if (all || type == NETCONFA_MC_FORWARDING)
521 size += nla_total_size(4);
522 #endif
523 if (all || type == NETCONFA_PROXY_NEIGH)
524 size += nla_total_size(4);
525
526 if (all || type == NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN)
527 size += nla_total_size(4);
528
529 return size;
530 }
531
inet6_netconf_fill_devconf(struct sk_buff * skb,int ifindex,struct ipv6_devconf * devconf,u32 portid,u32 seq,int event,unsigned int flags,int type)532 static int inet6_netconf_fill_devconf(struct sk_buff *skb, int ifindex,
533 struct ipv6_devconf *devconf, u32 portid,
534 u32 seq, int event, unsigned int flags,
535 int type)
536 {
537 struct nlmsghdr *nlh;
538 struct netconfmsg *ncm;
539 bool all = false;
540
541 nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct netconfmsg),
542 flags);
543 if (!nlh)
544 return -EMSGSIZE;
545
546 if (type == NETCONFA_ALL)
547 all = true;
548
549 ncm = nlmsg_data(nlh);
550 ncm->ncm_family = AF_INET6;
551
552 if (nla_put_s32(skb, NETCONFA_IFINDEX, ifindex) < 0)
553 goto nla_put_failure;
554
555 if (!devconf)
556 goto out;
557
558 if ((all || type == NETCONFA_FORWARDING) &&
559 nla_put_s32(skb, NETCONFA_FORWARDING,
560 READ_ONCE(devconf->forwarding)) < 0)
561 goto nla_put_failure;
562 #ifdef CONFIG_IPV6_MROUTE
563 if ((all || type == NETCONFA_MC_FORWARDING) &&
564 nla_put_s32(skb, NETCONFA_MC_FORWARDING,
565 atomic_read(&devconf->mc_forwarding)) < 0)
566 goto nla_put_failure;
567 #endif
568 if ((all || type == NETCONFA_PROXY_NEIGH) &&
569 nla_put_s32(skb, NETCONFA_PROXY_NEIGH,
570 READ_ONCE(devconf->proxy_ndp)) < 0)
571 goto nla_put_failure;
572
573 if ((all || type == NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN) &&
574 nla_put_s32(skb, NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
575 READ_ONCE(devconf->ignore_routes_with_linkdown)) < 0)
576 goto nla_put_failure;
577
578 out:
579 nlmsg_end(skb, nlh);
580 return 0;
581
582 nla_put_failure:
583 nlmsg_cancel(skb, nlh);
584 return -EMSGSIZE;
585 }
586
inet6_netconf_notify_devconf(struct net * net,int event,int type,int ifindex,struct ipv6_devconf * devconf)587 void inet6_netconf_notify_devconf(struct net *net, int event, int type,
588 int ifindex, struct ipv6_devconf *devconf)
589 {
590 struct sk_buff *skb;
591 int err = -ENOBUFS;
592
593 skb = nlmsg_new(inet6_netconf_msgsize_devconf(type), GFP_KERNEL);
594 if (!skb)
595 goto errout;
596
597 err = inet6_netconf_fill_devconf(skb, ifindex, devconf, 0, 0,
598 event, 0, type);
599 if (err < 0) {
600 /* -EMSGSIZE implies BUG in inet6_netconf_msgsize_devconf() */
601 WARN_ON(err == -EMSGSIZE);
602 kfree_skb(skb);
603 goto errout;
604 }
605 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_NETCONF, NULL, GFP_KERNEL);
606 return;
607 errout:
608 rtnl_set_sk_err(net, RTNLGRP_IPV6_NETCONF, err);
609 }
610
611 static const struct nla_policy devconf_ipv6_policy[NETCONFA_MAX+1] = {
612 [NETCONFA_IFINDEX] = { .len = sizeof(int) },
613 [NETCONFA_FORWARDING] = { .len = sizeof(int) },
614 [NETCONFA_PROXY_NEIGH] = { .len = sizeof(int) },
615 [NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN] = { .len = sizeof(int) },
616 };
617
inet6_netconf_valid_get_req(struct sk_buff * skb,const struct nlmsghdr * nlh,struct nlattr ** tb,struct netlink_ext_ack * extack)618 static int inet6_netconf_valid_get_req(struct sk_buff *skb,
619 const struct nlmsghdr *nlh,
620 struct nlattr **tb,
621 struct netlink_ext_ack *extack)
622 {
623 int i, err;
624
625 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(struct netconfmsg))) {
626 NL_SET_ERR_MSG_MOD(extack, "Invalid header for netconf get request");
627 return -EINVAL;
628 }
629
630 if (!netlink_strict_get_check(skb))
631 return nlmsg_parse_deprecated(nlh, sizeof(struct netconfmsg),
632 tb, NETCONFA_MAX,
633 devconf_ipv6_policy, extack);
634
635 err = nlmsg_parse_deprecated_strict(nlh, sizeof(struct netconfmsg),
636 tb, NETCONFA_MAX,
637 devconf_ipv6_policy, extack);
638 if (err)
639 return err;
640
641 for (i = 0; i <= NETCONFA_MAX; i++) {
642 if (!tb[i])
643 continue;
644
645 switch (i) {
646 case NETCONFA_IFINDEX:
647 break;
648 default:
649 NL_SET_ERR_MSG_MOD(extack, "Unsupported attribute in netconf get request");
650 return -EINVAL;
651 }
652 }
653
654 return 0;
655 }
656
inet6_netconf_get_devconf(struct sk_buff * in_skb,struct nlmsghdr * nlh,struct netlink_ext_ack * extack)657 static int inet6_netconf_get_devconf(struct sk_buff *in_skb,
658 struct nlmsghdr *nlh,
659 struct netlink_ext_ack *extack)
660 {
661 struct net *net = sock_net(in_skb->sk);
662 struct nlattr *tb[NETCONFA_MAX+1];
663 struct inet6_dev *in6_dev = NULL;
664 struct net_device *dev = NULL;
665 struct sk_buff *skb;
666 struct ipv6_devconf *devconf;
667 int ifindex;
668 int err;
669
670 err = inet6_netconf_valid_get_req(in_skb, nlh, tb, extack);
671 if (err < 0)
672 return err;
673
674 if (!tb[NETCONFA_IFINDEX])
675 return -EINVAL;
676
677 err = -EINVAL;
678 ifindex = nla_get_s32(tb[NETCONFA_IFINDEX]);
679 switch (ifindex) {
680 case NETCONFA_IFINDEX_ALL:
681 devconf = net->ipv6.devconf_all;
682 break;
683 case NETCONFA_IFINDEX_DEFAULT:
684 devconf = net->ipv6.devconf_dflt;
685 break;
686 default:
687 dev = dev_get_by_index(net, ifindex);
688 if (!dev)
689 return -EINVAL;
690 in6_dev = in6_dev_get(dev);
691 if (!in6_dev)
692 goto errout;
693 devconf = &in6_dev->cnf;
694 break;
695 }
696
697 err = -ENOBUFS;
698 skb = nlmsg_new(inet6_netconf_msgsize_devconf(NETCONFA_ALL), GFP_KERNEL);
699 if (!skb)
700 goto errout;
701
702 err = inet6_netconf_fill_devconf(skb, ifindex, devconf,
703 NETLINK_CB(in_skb).portid,
704 nlh->nlmsg_seq, RTM_NEWNETCONF, 0,
705 NETCONFA_ALL);
706 if (err < 0) {
707 /* -EMSGSIZE implies BUG in inet6_netconf_msgsize_devconf() */
708 WARN_ON(err == -EMSGSIZE);
709 kfree_skb(skb);
710 goto errout;
711 }
712 err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
713 errout:
714 if (in6_dev)
715 in6_dev_put(in6_dev);
716 dev_put(dev);
717 return err;
718 }
719
720 /* Combine dev_addr_genid and dev_base_seq to detect changes.
721 */
inet6_base_seq(const struct net * net)722 static u32 inet6_base_seq(const struct net *net)
723 {
724 u32 res = atomic_read(&net->ipv6.dev_addr_genid) +
725 READ_ONCE(net->dev_base_seq);
726
727 /* Must not return 0 (see nl_dump_check_consistent()).
728 * Chose a value far away from 0.
729 */
730 if (!res)
731 res = 0x80000000;
732 return res;
733 }
734
inet6_netconf_dump_devconf(struct sk_buff * skb,struct netlink_callback * cb)735 static int inet6_netconf_dump_devconf(struct sk_buff *skb,
736 struct netlink_callback *cb)
737 {
738 const struct nlmsghdr *nlh = cb->nlh;
739 struct net *net = sock_net(skb->sk);
740 struct {
741 unsigned long ifindex;
742 unsigned int all_default;
743 } *ctx = (void *)cb->ctx;
744 struct net_device *dev;
745 struct inet6_dev *idev;
746 int err = 0;
747
748 if (cb->strict_check) {
749 struct netlink_ext_ack *extack = cb->extack;
750 struct netconfmsg *ncm;
751
752 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ncm))) {
753 NL_SET_ERR_MSG_MOD(extack, "Invalid header for netconf dump request");
754 return -EINVAL;
755 }
756
757 if (nlmsg_attrlen(nlh, sizeof(*ncm))) {
758 NL_SET_ERR_MSG_MOD(extack, "Invalid data after header in netconf dump request");
759 return -EINVAL;
760 }
761 }
762
763 rcu_read_lock();
764 for_each_netdev_dump(net, dev, ctx->ifindex) {
765 idev = __in6_dev_get(dev);
766 if (!idev)
767 continue;
768 err = inet6_netconf_fill_devconf(skb, dev->ifindex,
769 &idev->cnf,
770 NETLINK_CB(cb->skb).portid,
771 nlh->nlmsg_seq,
772 RTM_NEWNETCONF,
773 NLM_F_MULTI,
774 NETCONFA_ALL);
775 if (err < 0)
776 goto done;
777 }
778 if (ctx->all_default == 0) {
779 err = inet6_netconf_fill_devconf(skb, NETCONFA_IFINDEX_ALL,
780 net->ipv6.devconf_all,
781 NETLINK_CB(cb->skb).portid,
782 nlh->nlmsg_seq,
783 RTM_NEWNETCONF, NLM_F_MULTI,
784 NETCONFA_ALL);
785 if (err < 0)
786 goto done;
787 ctx->all_default++;
788 }
789 if (ctx->all_default == 1) {
790 err = inet6_netconf_fill_devconf(skb, NETCONFA_IFINDEX_DEFAULT,
791 net->ipv6.devconf_dflt,
792 NETLINK_CB(cb->skb).portid,
793 nlh->nlmsg_seq,
794 RTM_NEWNETCONF, NLM_F_MULTI,
795 NETCONFA_ALL);
796 if (err < 0)
797 goto done;
798 ctx->all_default++;
799 }
800 done:
801 rcu_read_unlock();
802 return err;
803 }
804
805 #ifdef CONFIG_SYSCTL
dev_forward_change(struct inet6_dev * idev)806 static void dev_forward_change(struct inet6_dev *idev)
807 {
808 struct net_device *dev;
809 struct inet6_ifaddr *ifa;
810 LIST_HEAD(tmp_addr_list);
811
812 if (!idev)
813 return;
814 dev = idev->dev;
815 if (idev->cnf.forwarding)
816 dev_disable_lro(dev);
817 if (dev->flags & IFF_MULTICAST) {
818 if (idev->cnf.forwarding) {
819 ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
820 ipv6_dev_mc_inc(dev, &in6addr_interfacelocal_allrouters);
821 ipv6_dev_mc_inc(dev, &in6addr_sitelocal_allrouters);
822 } else {
823 ipv6_dev_mc_dec(dev, &in6addr_linklocal_allrouters);
824 ipv6_dev_mc_dec(dev, &in6addr_interfacelocal_allrouters);
825 ipv6_dev_mc_dec(dev, &in6addr_sitelocal_allrouters);
826 }
827 }
828
829 read_lock_bh(&idev->lock);
830 list_for_each_entry(ifa, &idev->addr_list, if_list) {
831 if (ifa->flags&IFA_F_TENTATIVE)
832 continue;
833 list_add_tail(&ifa->if_list_aux, &tmp_addr_list);
834 }
835 read_unlock_bh(&idev->lock);
836
837 while (!list_empty(&tmp_addr_list)) {
838 ifa = list_first_entry(&tmp_addr_list,
839 struct inet6_ifaddr, if_list_aux);
840 list_del(&ifa->if_list_aux);
841 if (idev->cnf.forwarding)
842 addrconf_join_anycast(ifa);
843 else
844 addrconf_leave_anycast(ifa);
845 }
846
847 inet6_netconf_notify_devconf(dev_net(dev), RTM_NEWNETCONF,
848 NETCONFA_FORWARDING,
849 dev->ifindex, &idev->cnf);
850 }
851
852
addrconf_forward_change(struct net * net,__s32 newf)853 static void addrconf_forward_change(struct net *net, __s32 newf)
854 {
855 struct net_device *dev;
856 struct inet6_dev *idev;
857
858 for_each_netdev(net, dev) {
859 idev = __in6_dev_get_rtnl_net(dev);
860 if (idev) {
861 int changed = (!idev->cnf.forwarding) ^ (!newf);
862 /* Disabling all.forwarding sets 0 to force_forwarding for all interfaces */
863 if (newf == 0)
864 WRITE_ONCE(idev->cnf.force_forwarding, 0);
865
866 WRITE_ONCE(idev->cnf.forwarding, newf);
867 if (changed)
868 dev_forward_change(idev);
869 }
870 }
871 }
872
addrconf_fixup_forwarding(const struct ctl_table * table,int * p,int newf)873 static int addrconf_fixup_forwarding(const struct ctl_table *table, int *p, int newf)
874 {
875 struct net *net = (struct net *)table->extra2;
876 int old;
877
878 if (!rtnl_net_trylock(net))
879 return restart_syscall();
880
881 old = *p;
882 WRITE_ONCE(*p, newf);
883
884 if (p == &net->ipv6.devconf_dflt->forwarding) {
885 if ((!newf) ^ (!old))
886 inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
887 NETCONFA_FORWARDING,
888 NETCONFA_IFINDEX_DEFAULT,
889 net->ipv6.devconf_dflt);
890 rtnl_net_unlock(net);
891 return 0;
892 }
893
894 if (p == &net->ipv6.devconf_all->forwarding) {
895 int old_dflt = net->ipv6.devconf_dflt->forwarding;
896
897 WRITE_ONCE(net->ipv6.devconf_dflt->forwarding, newf);
898 if ((!newf) ^ (!old_dflt))
899 inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
900 NETCONFA_FORWARDING,
901 NETCONFA_IFINDEX_DEFAULT,
902 net->ipv6.devconf_dflt);
903
904 addrconf_forward_change(net, newf);
905 if ((!newf) ^ (!old))
906 inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
907 NETCONFA_FORWARDING,
908 NETCONFA_IFINDEX_ALL,
909 net->ipv6.devconf_all);
910 } else if ((!newf) ^ (!old))
911 dev_forward_change((struct inet6_dev *)table->extra1);
912 rtnl_net_unlock(net);
913
914 if (newf)
915 rt6_purge_dflt_routers(net);
916 return 1;
917 }
918
addrconf_linkdown_change(struct net * net,__s32 newf)919 static void addrconf_linkdown_change(struct net *net, __s32 newf)
920 {
921 struct net_device *dev;
922 struct inet6_dev *idev;
923
924 for_each_netdev(net, dev) {
925 idev = __in6_dev_get_rtnl_net(dev);
926 if (idev) {
927 int changed = (!idev->cnf.ignore_routes_with_linkdown) ^ (!newf);
928
929 WRITE_ONCE(idev->cnf.ignore_routes_with_linkdown, newf);
930 if (changed)
931 inet6_netconf_notify_devconf(dev_net(dev),
932 RTM_NEWNETCONF,
933 NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
934 dev->ifindex,
935 &idev->cnf);
936 }
937 }
938 }
939
addrconf_fixup_linkdown(const struct ctl_table * table,int * p,int newf)940 static int addrconf_fixup_linkdown(const struct ctl_table *table, int *p, int newf)
941 {
942 struct net *net = (struct net *)table->extra2;
943 int old;
944
945 if (!rtnl_net_trylock(net))
946 return restart_syscall();
947
948 old = *p;
949 WRITE_ONCE(*p, newf);
950
951 if (p == &net->ipv6.devconf_dflt->ignore_routes_with_linkdown) {
952 if ((!newf) ^ (!old))
953 inet6_netconf_notify_devconf(net,
954 RTM_NEWNETCONF,
955 NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
956 NETCONFA_IFINDEX_DEFAULT,
957 net->ipv6.devconf_dflt);
958 rtnl_net_unlock(net);
959 return 0;
960 }
961
962 if (p == &net->ipv6.devconf_all->ignore_routes_with_linkdown) {
963 WRITE_ONCE(net->ipv6.devconf_dflt->ignore_routes_with_linkdown, newf);
964 addrconf_linkdown_change(net, newf);
965 if ((!newf) ^ (!old))
966 inet6_netconf_notify_devconf(net,
967 RTM_NEWNETCONF,
968 NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
969 NETCONFA_IFINDEX_ALL,
970 net->ipv6.devconf_all);
971 }
972
973 rtnl_net_unlock(net);
974
975 return 1;
976 }
977
978 #endif
979
980 /* Nobody refers to this ifaddr, destroy it */
inet6_ifa_finish_destroy(struct inet6_ifaddr * ifp)981 void inet6_ifa_finish_destroy(struct inet6_ifaddr *ifp)
982 {
983 WARN_ON(!hlist_unhashed(&ifp->addr_lst));
984
985 #ifdef NET_REFCNT_DEBUG
986 pr_debug("%s\n", __func__);
987 #endif
988
989 in6_dev_put(ifp->idev);
990
991 if (cancel_delayed_work(&ifp->dad_work))
992 pr_notice("delayed DAD work was pending while freeing ifa=%p\n",
993 ifp);
994
995 if (ifp->state != INET6_IFADDR_STATE_DEAD) {
996 pr_warn("Freeing alive inet6 address %p\n", ifp);
997 return;
998 }
999
1000 kfree_rcu(ifp, rcu);
1001 }
1002
1003 static void
ipv6_link_dev_addr(struct inet6_dev * idev,struct inet6_ifaddr * ifp)1004 ipv6_link_dev_addr(struct inet6_dev *idev, struct inet6_ifaddr *ifp)
1005 {
1006 struct list_head *p;
1007 int ifp_scope = ipv6_addr_src_scope(&ifp->addr);
1008
1009 /*
1010 * Each device address list is sorted in order of scope -
1011 * global before linklocal.
1012 */
1013 list_for_each(p, &idev->addr_list) {
1014 struct inet6_ifaddr *ifa
1015 = list_entry(p, struct inet6_ifaddr, if_list);
1016 if (ifp_scope >= ipv6_addr_src_scope(&ifa->addr))
1017 break;
1018 }
1019
1020 list_add_tail_rcu(&ifp->if_list, p);
1021 }
1022
inet6_addr_hash(const struct net * net,const struct in6_addr * addr)1023 static u32 inet6_addr_hash(const struct net *net, const struct in6_addr *addr)
1024 {
1025 u32 val = __ipv6_addr_jhash(addr, net_hash_mix(net));
1026
1027 return hash_32(val, IN6_ADDR_HSIZE_SHIFT);
1028 }
1029
ipv6_chk_same_addr(struct net * net,const struct in6_addr * addr,struct net_device * dev,unsigned int hash)1030 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
1031 struct net_device *dev, unsigned int hash)
1032 {
1033 struct inet6_ifaddr *ifp;
1034
1035 hlist_for_each_entry(ifp, &net->ipv6.inet6_addr_lst[hash], addr_lst) {
1036 if (ipv6_addr_equal(&ifp->addr, addr)) {
1037 if (!dev || ifp->idev->dev == dev)
1038 return true;
1039 }
1040 }
1041 return false;
1042 }
1043
ipv6_add_addr_hash(struct net_device * dev,struct inet6_ifaddr * ifa)1044 static int ipv6_add_addr_hash(struct net_device *dev, struct inet6_ifaddr *ifa)
1045 {
1046 struct net *net = dev_net(dev);
1047 unsigned int hash = inet6_addr_hash(net, &ifa->addr);
1048 int err = 0;
1049
1050 spin_lock_bh(&net->ipv6.addrconf_hash_lock);
1051
1052 /* Ignore adding duplicate addresses on an interface */
1053 if (ipv6_chk_same_addr(net, &ifa->addr, dev, hash)) {
1054 netdev_dbg(dev, "ipv6_add_addr: already assigned\n");
1055 err = -EEXIST;
1056 } else {
1057 hlist_add_head_rcu(&ifa->addr_lst, &net->ipv6.inet6_addr_lst[hash]);
1058 }
1059
1060 spin_unlock_bh(&net->ipv6.addrconf_hash_lock);
1061
1062 return err;
1063 }
1064
1065 /* On success it returns ifp with increased reference count */
1066
1067 static struct inet6_ifaddr *
ipv6_add_addr(struct inet6_dev * idev,struct ifa6_config * cfg,bool can_block,struct netlink_ext_ack * extack)1068 ipv6_add_addr(struct inet6_dev *idev, struct ifa6_config *cfg,
1069 bool can_block, struct netlink_ext_ack *extack)
1070 {
1071 gfp_t gfp_flags = can_block ? GFP_KERNEL : GFP_ATOMIC;
1072 int addr_type = ipv6_addr_type(cfg->pfx);
1073 struct net *net = dev_net(idev->dev);
1074 struct inet6_ifaddr *ifa = NULL;
1075 struct fib6_info *f6i = NULL;
1076 int err = 0;
1077
1078 if (addr_type == IPV6_ADDR_ANY) {
1079 NL_SET_ERR_MSG_MOD(extack, "Invalid address");
1080 return ERR_PTR(-EADDRNOTAVAIL);
1081 } else if (addr_type & IPV6_ADDR_MULTICAST &&
1082 !(cfg->ifa_flags & IFA_F_MCAUTOJOIN)) {
1083 NL_SET_ERR_MSG_MOD(extack, "Cannot assign multicast address without \"IFA_F_MCAUTOJOIN\" flag");
1084 return ERR_PTR(-EADDRNOTAVAIL);
1085 } else if (!(idev->dev->flags & IFF_LOOPBACK) &&
1086 !netif_is_l3_master(idev->dev) &&
1087 addr_type & IPV6_ADDR_LOOPBACK) {
1088 NL_SET_ERR_MSG_MOD(extack, "Cannot assign loopback address on this device");
1089 return ERR_PTR(-EADDRNOTAVAIL);
1090 }
1091
1092 if (idev->dead) {
1093 NL_SET_ERR_MSG_MOD(extack, "device is going away");
1094 err = -ENODEV;
1095 goto out;
1096 }
1097
1098 if (idev->cnf.disable_ipv6) {
1099 NL_SET_ERR_MSG_MOD(extack, "IPv6 is disabled on this device");
1100 err = -EACCES;
1101 goto out;
1102 }
1103
1104 /* validator notifier needs to be blocking;
1105 * do not call in atomic context
1106 */
1107 if (can_block) {
1108 struct in6_validator_info i6vi = {
1109 .i6vi_addr = *cfg->pfx,
1110 .i6vi_dev = idev,
1111 .extack = extack,
1112 };
1113
1114 err = inet6addr_validator_notifier_call_chain(NETDEV_UP, &i6vi);
1115 err = notifier_to_errno(err);
1116 if (err < 0)
1117 goto out;
1118 }
1119
1120 ifa = kzalloc(sizeof(*ifa), gfp_flags | __GFP_ACCOUNT);
1121 if (!ifa) {
1122 err = -ENOBUFS;
1123 goto out;
1124 }
1125
1126 f6i = addrconf_f6i_alloc(net, idev, cfg->pfx, false, gfp_flags, extack);
1127 if (IS_ERR(f6i)) {
1128 err = PTR_ERR(f6i);
1129 f6i = NULL;
1130 goto out;
1131 }
1132
1133 neigh_parms_data_state_setall(idev->nd_parms);
1134
1135 ifa->addr = *cfg->pfx;
1136 if (cfg->peer_pfx)
1137 ifa->peer_addr = *cfg->peer_pfx;
1138
1139 spin_lock_init(&ifa->lock);
1140 INIT_DELAYED_WORK(&ifa->dad_work, addrconf_dad_work);
1141 INIT_HLIST_NODE(&ifa->addr_lst);
1142 ifa->scope = cfg->scope;
1143 ifa->prefix_len = cfg->plen;
1144 ifa->rt_priority = cfg->rt_priority;
1145 ifa->flags = cfg->ifa_flags;
1146 ifa->ifa_proto = cfg->ifa_proto;
1147 /* No need to add the TENTATIVE flag for addresses with NODAD */
1148 if (!(cfg->ifa_flags & IFA_F_NODAD))
1149 ifa->flags |= IFA_F_TENTATIVE;
1150 ifa->valid_lft = cfg->valid_lft;
1151 ifa->prefered_lft = cfg->preferred_lft;
1152 ifa->cstamp = ifa->tstamp = jiffies;
1153 ifa->tokenized = false;
1154
1155 ifa->rt = f6i;
1156
1157 ifa->idev = idev;
1158 in6_dev_hold(idev);
1159
1160 /* For caller */
1161 refcount_set(&ifa->refcnt, 1);
1162
1163 rcu_read_lock();
1164
1165 err = ipv6_add_addr_hash(idev->dev, ifa);
1166 if (err < 0) {
1167 rcu_read_unlock();
1168 goto out;
1169 }
1170
1171 write_lock_bh(&idev->lock);
1172
1173 /* Add to inet6_dev unicast addr list. */
1174 ipv6_link_dev_addr(idev, ifa);
1175
1176 if (ifa->flags&IFA_F_TEMPORARY) {
1177 list_add(&ifa->tmp_list, &idev->tempaddr_list);
1178 in6_ifa_hold(ifa);
1179 }
1180
1181 in6_ifa_hold(ifa);
1182 write_unlock_bh(&idev->lock);
1183
1184 rcu_read_unlock();
1185
1186 inet6addr_notifier_call_chain(NETDEV_UP, ifa);
1187 out:
1188 if (unlikely(err < 0)) {
1189 fib6_info_release(f6i);
1190
1191 if (ifa) {
1192 if (ifa->idev)
1193 in6_dev_put(ifa->idev);
1194 kfree(ifa);
1195 }
1196 ifa = ERR_PTR(err);
1197 }
1198
1199 return ifa;
1200 }
1201
1202 enum cleanup_prefix_rt_t {
1203 CLEANUP_PREFIX_RT_NOP, /* no cleanup action for prefix route */
1204 CLEANUP_PREFIX_RT_DEL, /* delete the prefix route */
1205 CLEANUP_PREFIX_RT_EXPIRE, /* update the lifetime of the prefix route */
1206 };
1207
1208 /*
1209 * Check, whether the prefix for ifp would still need a prefix route
1210 * after deleting ifp. The function returns one of the CLEANUP_PREFIX_RT_*
1211 * constants.
1212 *
1213 * 1) we don't purge prefix if address was not permanent.
1214 * prefix is managed by its own lifetime.
1215 * 2) we also don't purge, if the address was IFA_F_NOPREFIXROUTE.
1216 * 3) if there are no addresses, delete prefix.
1217 * 4) if there are still other permanent address(es),
1218 * corresponding prefix is still permanent.
1219 * 5) if there are still other addresses with IFA_F_NOPREFIXROUTE,
1220 * don't purge the prefix, assume user space is managing it.
1221 * 6) otherwise, update prefix lifetime to the
1222 * longest valid lifetime among the corresponding
1223 * addresses on the device.
1224 * Note: subsequent RA will update lifetime.
1225 **/
1226 static enum cleanup_prefix_rt_t
check_cleanup_prefix_route(struct inet6_ifaddr * ifp,unsigned long * expires)1227 check_cleanup_prefix_route(struct inet6_ifaddr *ifp, unsigned long *expires)
1228 {
1229 struct inet6_ifaddr *ifa;
1230 struct inet6_dev *idev = ifp->idev;
1231 unsigned long lifetime;
1232 enum cleanup_prefix_rt_t action = CLEANUP_PREFIX_RT_DEL;
1233
1234 *expires = jiffies;
1235
1236 list_for_each_entry(ifa, &idev->addr_list, if_list) {
1237 if (ifa == ifp)
1238 continue;
1239 if (ifa->prefix_len != ifp->prefix_len ||
1240 !ipv6_prefix_equal(&ifa->addr, &ifp->addr,
1241 ifp->prefix_len))
1242 continue;
1243 if (ifa->flags & (IFA_F_PERMANENT | IFA_F_NOPREFIXROUTE))
1244 return CLEANUP_PREFIX_RT_NOP;
1245
1246 action = CLEANUP_PREFIX_RT_EXPIRE;
1247
1248 spin_lock(&ifa->lock);
1249
1250 lifetime = addrconf_timeout_fixup(ifa->valid_lft, HZ);
1251 /*
1252 * Note: Because this address is
1253 * not permanent, lifetime <
1254 * LONG_MAX / HZ here.
1255 */
1256 if (time_before(*expires, ifa->tstamp + lifetime * HZ))
1257 *expires = ifa->tstamp + lifetime * HZ;
1258 spin_unlock(&ifa->lock);
1259 }
1260
1261 return action;
1262 }
1263
1264 static void
cleanup_prefix_route(struct inet6_ifaddr * ifp,unsigned long expires,bool del_rt,bool del_peer)1265 cleanup_prefix_route(struct inet6_ifaddr *ifp, unsigned long expires,
1266 bool del_rt, bool del_peer)
1267 {
1268 struct fib6_table *table;
1269 struct fib6_info *f6i;
1270
1271 f6i = addrconf_get_prefix_route(del_peer ? &ifp->peer_addr : &ifp->addr,
1272 ifp->prefix_len,
1273 ifp->idev->dev, 0, RTF_DEFAULT, true);
1274 if (f6i) {
1275 if (del_rt)
1276 ip6_del_rt(dev_net(ifp->idev->dev), f6i, false);
1277 else {
1278 if (!(f6i->fib6_flags & RTF_EXPIRES)) {
1279 table = f6i->fib6_table;
1280 spin_lock_bh(&table->tb6_lock);
1281
1282 fib6_set_expires(f6i, expires);
1283 fib6_add_gc_list(f6i);
1284
1285 spin_unlock_bh(&table->tb6_lock);
1286 }
1287 fib6_info_release(f6i);
1288 }
1289 }
1290 }
1291
1292
1293 /* This function wants to get referenced ifp and releases it before return */
1294
ipv6_del_addr(struct inet6_ifaddr * ifp)1295 static void ipv6_del_addr(struct inet6_ifaddr *ifp)
1296 {
1297 enum cleanup_prefix_rt_t action = CLEANUP_PREFIX_RT_NOP;
1298 struct net *net = dev_net(ifp->idev->dev);
1299 unsigned long expires;
1300 int state;
1301
1302 ASSERT_RTNL();
1303
1304 spin_lock_bh(&ifp->lock);
1305 state = ifp->state;
1306 ifp->state = INET6_IFADDR_STATE_DEAD;
1307 spin_unlock_bh(&ifp->lock);
1308
1309 if (state == INET6_IFADDR_STATE_DEAD)
1310 goto out;
1311
1312 spin_lock_bh(&net->ipv6.addrconf_hash_lock);
1313 hlist_del_init_rcu(&ifp->addr_lst);
1314 spin_unlock_bh(&net->ipv6.addrconf_hash_lock);
1315
1316 write_lock_bh(&ifp->idev->lock);
1317
1318 if (ifp->flags&IFA_F_TEMPORARY) {
1319 list_del(&ifp->tmp_list);
1320 if (ifp->ifpub) {
1321 in6_ifa_put(ifp->ifpub);
1322 ifp->ifpub = NULL;
1323 }
1324 __in6_ifa_put(ifp);
1325 }
1326
1327 if (ifp->flags & IFA_F_PERMANENT && !(ifp->flags & IFA_F_NOPREFIXROUTE))
1328 action = check_cleanup_prefix_route(ifp, &expires);
1329
1330 list_del_rcu(&ifp->if_list);
1331 __in6_ifa_put(ifp);
1332
1333 write_unlock_bh(&ifp->idev->lock);
1334
1335 addrconf_del_dad_work(ifp);
1336
1337 ipv6_ifa_notify(RTM_DELADDR, ifp);
1338
1339 inet6addr_notifier_call_chain(NETDEV_DOWN, ifp);
1340
1341 if (action != CLEANUP_PREFIX_RT_NOP) {
1342 cleanup_prefix_route(ifp, expires,
1343 action == CLEANUP_PREFIX_RT_DEL, false);
1344 }
1345
1346 /* clean up prefsrc entries */
1347 rt6_remove_prefsrc(ifp);
1348 out:
1349 in6_ifa_put(ifp);
1350 }
1351
ipv6_get_regen_advance(const struct inet6_dev * idev)1352 static unsigned long ipv6_get_regen_advance(const struct inet6_dev *idev)
1353 {
1354 return READ_ONCE(idev->cnf.regen_min_advance) +
1355 READ_ONCE(idev->cnf.regen_max_retry) *
1356 READ_ONCE(idev->cnf.dad_transmits) *
1357 max(NEIGH_VAR(idev->nd_parms, RETRANS_TIME), HZ/100) / HZ;
1358 }
1359
ipv6_create_tempaddr(struct inet6_ifaddr * ifp,bool block)1360 static int ipv6_create_tempaddr(struct inet6_ifaddr *ifp, bool block)
1361 {
1362 struct inet6_dev *idev = ifp->idev;
1363 unsigned long tmp_tstamp, age;
1364 unsigned long regen_advance;
1365 unsigned long now = jiffies;
1366 u32 if_public_preferred_lft;
1367 s32 cnf_temp_preferred_lft;
1368 struct inet6_ifaddr *ift;
1369 struct ifa6_config cfg;
1370 long max_desync_factor;
1371 struct in6_addr addr;
1372 int ret = 0;
1373
1374 write_lock_bh(&idev->lock);
1375
1376 retry:
1377 in6_dev_hold(idev);
1378 if (READ_ONCE(idev->cnf.use_tempaddr) <= 0) {
1379 write_unlock_bh(&idev->lock);
1380 pr_info("%s: use_tempaddr is disabled\n", __func__);
1381 in6_dev_put(idev);
1382 ret = -1;
1383 goto out;
1384 }
1385 spin_lock_bh(&ifp->lock);
1386 if (ifp->regen_count++ >= READ_ONCE(idev->cnf.regen_max_retry)) {
1387 WRITE_ONCE(idev->cnf.use_tempaddr, -1); /*XXX*/
1388 spin_unlock_bh(&ifp->lock);
1389 write_unlock_bh(&idev->lock);
1390 pr_warn("%s: regeneration time exceeded - disabled temporary address support\n",
1391 __func__);
1392 in6_dev_put(idev);
1393 ret = -1;
1394 goto out;
1395 }
1396 in6_ifa_hold(ifp);
1397 memcpy(addr.s6_addr, ifp->addr.s6_addr, 8);
1398 ipv6_gen_rnd_iid(&addr);
1399
1400 age = (now - ifp->tstamp) / HZ;
1401
1402 regen_advance = ipv6_get_regen_advance(idev);
1403
1404 /* recalculate max_desync_factor each time and update
1405 * idev->desync_factor if it's larger
1406 */
1407 cnf_temp_preferred_lft = READ_ONCE(idev->cnf.temp_prefered_lft);
1408 max_desync_factor = min_t(long,
1409 READ_ONCE(idev->cnf.max_desync_factor),
1410 cnf_temp_preferred_lft - regen_advance);
1411
1412 if (unlikely(idev->desync_factor > max_desync_factor)) {
1413 if (max_desync_factor > 0) {
1414 get_random_bytes(&idev->desync_factor,
1415 sizeof(idev->desync_factor));
1416 idev->desync_factor %= max_desync_factor;
1417 } else {
1418 idev->desync_factor = 0;
1419 }
1420 }
1421
1422 if_public_preferred_lft = ifp->prefered_lft;
1423
1424 memset(&cfg, 0, sizeof(cfg));
1425 cfg.valid_lft = min_t(__u32, ifp->valid_lft,
1426 READ_ONCE(idev->cnf.temp_valid_lft) + age);
1427 cfg.preferred_lft = cnf_temp_preferred_lft + age - idev->desync_factor;
1428 cfg.preferred_lft = min_t(__u32, if_public_preferred_lft, cfg.preferred_lft);
1429 cfg.preferred_lft = min_t(__u32, cfg.valid_lft, cfg.preferred_lft);
1430
1431 cfg.plen = ifp->prefix_len;
1432 tmp_tstamp = ifp->tstamp;
1433 spin_unlock_bh(&ifp->lock);
1434
1435 write_unlock_bh(&idev->lock);
1436
1437 /* From RFC 4941:
1438 *
1439 * A temporary address is created only if this calculated Preferred
1440 * Lifetime is greater than REGEN_ADVANCE time units. In
1441 * particular, an implementation must not create a temporary address
1442 * with a zero Preferred Lifetime.
1443 *
1444 * ...
1445 *
1446 * When creating a temporary address, the lifetime values MUST be
1447 * derived from the corresponding prefix as follows:
1448 *
1449 * ...
1450 *
1451 * * Its Preferred Lifetime is the lower of the Preferred Lifetime
1452 * of the public address or TEMP_PREFERRED_LIFETIME -
1453 * DESYNC_FACTOR.
1454 *
1455 * To comply with the RFC's requirements, clamp the preferred lifetime
1456 * to a minimum of regen_advance, unless that would exceed valid_lft or
1457 * ifp->prefered_lft.
1458 *
1459 * Use age calculation as in addrconf_verify to avoid unnecessary
1460 * temporary addresses being generated.
1461 */
1462 age = (now - tmp_tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
1463 if (cfg.preferred_lft <= regen_advance + age) {
1464 cfg.preferred_lft = regen_advance + age + 1;
1465 if (cfg.preferred_lft > cfg.valid_lft ||
1466 cfg.preferred_lft > if_public_preferred_lft) {
1467 in6_ifa_put(ifp);
1468 in6_dev_put(idev);
1469 ret = -1;
1470 goto out;
1471 }
1472 }
1473
1474 cfg.ifa_flags = IFA_F_TEMPORARY;
1475 /* set in addrconf_prefix_rcv() */
1476 if (ifp->flags & IFA_F_OPTIMISTIC)
1477 cfg.ifa_flags |= IFA_F_OPTIMISTIC;
1478
1479 cfg.pfx = &addr;
1480 cfg.scope = ipv6_addr_scope(cfg.pfx);
1481
1482 ift = ipv6_add_addr(idev, &cfg, block, NULL);
1483 if (IS_ERR(ift)) {
1484 in6_ifa_put(ifp);
1485 in6_dev_put(idev);
1486 pr_info("%s: retry temporary address regeneration\n", __func__);
1487 write_lock_bh(&idev->lock);
1488 goto retry;
1489 }
1490
1491 spin_lock_bh(&ift->lock);
1492 ift->ifpub = ifp;
1493 ift->cstamp = now;
1494 ift->tstamp = tmp_tstamp;
1495 spin_unlock_bh(&ift->lock);
1496
1497 addrconf_dad_start(ift);
1498 in6_ifa_put(ift);
1499 in6_dev_put(idev);
1500 out:
1501 return ret;
1502 }
1503
1504 /*
1505 * Choose an appropriate source address (RFC3484)
1506 */
1507 enum {
1508 IPV6_SADDR_RULE_INIT = 0,
1509 IPV6_SADDR_RULE_LOCAL,
1510 IPV6_SADDR_RULE_SCOPE,
1511 IPV6_SADDR_RULE_PREFERRED,
1512 #ifdef CONFIG_IPV6_MIP6
1513 IPV6_SADDR_RULE_HOA,
1514 #endif
1515 IPV6_SADDR_RULE_OIF,
1516 IPV6_SADDR_RULE_LABEL,
1517 IPV6_SADDR_RULE_PRIVACY,
1518 IPV6_SADDR_RULE_ORCHID,
1519 IPV6_SADDR_RULE_PREFIX,
1520 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1521 IPV6_SADDR_RULE_NOT_OPTIMISTIC,
1522 #endif
1523 IPV6_SADDR_RULE_MAX
1524 };
1525
1526 struct ipv6_saddr_score {
1527 int rule;
1528 int addr_type;
1529 struct inet6_ifaddr *ifa;
1530 DECLARE_BITMAP(scorebits, IPV6_SADDR_RULE_MAX);
1531 int scopedist;
1532 int matchlen;
1533 };
1534
1535 struct ipv6_saddr_dst {
1536 const struct in6_addr *addr;
1537 int ifindex;
1538 int scope;
1539 int label;
1540 unsigned int prefs;
1541 };
1542
ipv6_saddr_preferred(int type)1543 static inline int ipv6_saddr_preferred(int type)
1544 {
1545 if (type & (IPV6_ADDR_MAPPED|IPV6_ADDR_COMPATv4|IPV6_ADDR_LOOPBACK))
1546 return 1;
1547 return 0;
1548 }
1549
ipv6_use_optimistic_addr(const struct net * net,const struct inet6_dev * idev)1550 static bool ipv6_use_optimistic_addr(const struct net *net,
1551 const struct inet6_dev *idev)
1552 {
1553 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1554 if (!idev)
1555 return false;
1556 if (!READ_ONCE(net->ipv6.devconf_all->optimistic_dad) &&
1557 !READ_ONCE(idev->cnf.optimistic_dad))
1558 return false;
1559 if (!READ_ONCE(net->ipv6.devconf_all->use_optimistic) &&
1560 !READ_ONCE(idev->cnf.use_optimistic))
1561 return false;
1562
1563 return true;
1564 #else
1565 return false;
1566 #endif
1567 }
1568
ipv6_allow_optimistic_dad(const struct net * net,const struct inet6_dev * idev)1569 static bool ipv6_allow_optimistic_dad(const struct net *net,
1570 const struct inet6_dev *idev)
1571 {
1572 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1573 if (!idev)
1574 return false;
1575 if (!READ_ONCE(net->ipv6.devconf_all->optimistic_dad) &&
1576 !READ_ONCE(idev->cnf.optimistic_dad))
1577 return false;
1578
1579 return true;
1580 #else
1581 return false;
1582 #endif
1583 }
1584
ipv6_get_saddr_eval(struct net * net,struct ipv6_saddr_score * score,struct ipv6_saddr_dst * dst,int i)1585 static int ipv6_get_saddr_eval(struct net *net,
1586 struct ipv6_saddr_score *score,
1587 struct ipv6_saddr_dst *dst,
1588 int i)
1589 {
1590 int ret;
1591
1592 if (i <= score->rule) {
1593 switch (i) {
1594 case IPV6_SADDR_RULE_SCOPE:
1595 ret = score->scopedist;
1596 break;
1597 case IPV6_SADDR_RULE_PREFIX:
1598 ret = score->matchlen;
1599 break;
1600 default:
1601 ret = !!test_bit(i, score->scorebits);
1602 }
1603 goto out;
1604 }
1605
1606 switch (i) {
1607 case IPV6_SADDR_RULE_INIT:
1608 /* Rule 0: remember if hiscore is not ready yet */
1609 ret = !!score->ifa;
1610 break;
1611 case IPV6_SADDR_RULE_LOCAL:
1612 /* Rule 1: Prefer same address */
1613 ret = ipv6_addr_equal(&score->ifa->addr, dst->addr);
1614 break;
1615 case IPV6_SADDR_RULE_SCOPE:
1616 /* Rule 2: Prefer appropriate scope
1617 *
1618 * ret
1619 * ^
1620 * -1 | d 15
1621 * ---+--+-+---> scope
1622 * |
1623 * | d is scope of the destination.
1624 * B-d | \
1625 * | \ <- smaller scope is better if
1626 * B-15 | \ if scope is enough for destination.
1627 * | ret = B - scope (-1 <= scope >= d <= 15).
1628 * d-C-1 | /
1629 * |/ <- greater is better
1630 * -C / if scope is not enough for destination.
1631 * /| ret = scope - C (-1 <= d < scope <= 15).
1632 *
1633 * d - C - 1 < B -15 (for all -1 <= d <= 15).
1634 * C > d + 14 - B >= 15 + 14 - B = 29 - B.
1635 * Assume B = 0 and we get C > 29.
1636 */
1637 ret = __ipv6_addr_src_scope(score->addr_type);
1638 if (ret >= dst->scope)
1639 ret = -ret;
1640 else
1641 ret -= 128; /* 30 is enough */
1642 score->scopedist = ret;
1643 break;
1644 case IPV6_SADDR_RULE_PREFERRED:
1645 {
1646 /* Rule 3: Avoid deprecated and optimistic addresses */
1647 u8 avoid = IFA_F_DEPRECATED;
1648
1649 if (!ipv6_use_optimistic_addr(net, score->ifa->idev))
1650 avoid |= IFA_F_OPTIMISTIC;
1651 ret = ipv6_saddr_preferred(score->addr_type) ||
1652 !(score->ifa->flags & avoid);
1653 break;
1654 }
1655 #ifdef CONFIG_IPV6_MIP6
1656 case IPV6_SADDR_RULE_HOA:
1657 {
1658 /* Rule 4: Prefer home address */
1659 int prefhome = !(dst->prefs & IPV6_PREFER_SRC_COA);
1660 ret = !(score->ifa->flags & IFA_F_HOMEADDRESS) ^ prefhome;
1661 break;
1662 }
1663 #endif
1664 case IPV6_SADDR_RULE_OIF:
1665 /* Rule 5: Prefer outgoing interface */
1666 ret = (!dst->ifindex ||
1667 dst->ifindex == score->ifa->idev->dev->ifindex);
1668 break;
1669 case IPV6_SADDR_RULE_LABEL:
1670 /* Rule 6: Prefer matching label */
1671 ret = ipv6_addr_label(net,
1672 &score->ifa->addr, score->addr_type,
1673 score->ifa->idev->dev->ifindex) == dst->label;
1674 break;
1675 case IPV6_SADDR_RULE_PRIVACY:
1676 {
1677 /* Rule 7: Prefer public address
1678 * Note: prefer temporary address if use_tempaddr >= 2
1679 */
1680 int preftmp = dst->prefs & (IPV6_PREFER_SRC_PUBLIC|IPV6_PREFER_SRC_TMP) ?
1681 !!(dst->prefs & IPV6_PREFER_SRC_TMP) :
1682 READ_ONCE(score->ifa->idev->cnf.use_tempaddr) >= 2;
1683 ret = (!(score->ifa->flags & IFA_F_TEMPORARY)) ^ preftmp;
1684 break;
1685 }
1686 case IPV6_SADDR_RULE_ORCHID:
1687 /* Rule 8-: Prefer ORCHID vs ORCHID or
1688 * non-ORCHID vs non-ORCHID
1689 */
1690 ret = !(ipv6_addr_orchid(&score->ifa->addr) ^
1691 ipv6_addr_orchid(dst->addr));
1692 break;
1693 case IPV6_SADDR_RULE_PREFIX:
1694 /* Rule 8: Use longest matching prefix */
1695 ret = ipv6_addr_diff(&score->ifa->addr, dst->addr);
1696 if (ret > score->ifa->prefix_len)
1697 ret = score->ifa->prefix_len;
1698 score->matchlen = ret;
1699 break;
1700 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1701 case IPV6_SADDR_RULE_NOT_OPTIMISTIC:
1702 /* Optimistic addresses still have lower precedence than other
1703 * preferred addresses.
1704 */
1705 ret = !(score->ifa->flags & IFA_F_OPTIMISTIC);
1706 break;
1707 #endif
1708 default:
1709 ret = 0;
1710 }
1711
1712 if (ret)
1713 __set_bit(i, score->scorebits);
1714 score->rule = i;
1715 out:
1716 return ret;
1717 }
1718
__ipv6_dev_get_saddr(struct net * net,struct ipv6_saddr_dst * dst,struct inet6_dev * idev,struct ipv6_saddr_score * scores,int hiscore_idx)1719 static int __ipv6_dev_get_saddr(struct net *net,
1720 struct ipv6_saddr_dst *dst,
1721 struct inet6_dev *idev,
1722 struct ipv6_saddr_score *scores,
1723 int hiscore_idx)
1724 {
1725 struct ipv6_saddr_score *score = &scores[1 - hiscore_idx], *hiscore = &scores[hiscore_idx];
1726
1727 list_for_each_entry_rcu(score->ifa, &idev->addr_list, if_list) {
1728 int i;
1729
1730 /*
1731 * - Tentative Address (RFC2462 section 5.4)
1732 * - A tentative address is not considered
1733 * "assigned to an interface" in the traditional
1734 * sense, unless it is also flagged as optimistic.
1735 * - Candidate Source Address (section 4)
1736 * - In any case, anycast addresses, multicast
1737 * addresses, and the unspecified address MUST
1738 * NOT be included in a candidate set.
1739 */
1740 if ((score->ifa->flags & IFA_F_TENTATIVE) &&
1741 (!(score->ifa->flags & IFA_F_OPTIMISTIC)))
1742 continue;
1743
1744 score->addr_type = __ipv6_addr_type(&score->ifa->addr);
1745
1746 if (unlikely(score->addr_type == IPV6_ADDR_ANY ||
1747 score->addr_type & IPV6_ADDR_MULTICAST)) {
1748 net_dbg_ratelimited("ADDRCONF: unspecified / multicast address assigned as unicast address on %s",
1749 idev->dev->name);
1750 continue;
1751 }
1752
1753 score->rule = -1;
1754 bitmap_zero(score->scorebits, IPV6_SADDR_RULE_MAX);
1755
1756 for (i = 0; i < IPV6_SADDR_RULE_MAX; i++) {
1757 int minihiscore, miniscore;
1758
1759 minihiscore = ipv6_get_saddr_eval(net, hiscore, dst, i);
1760 miniscore = ipv6_get_saddr_eval(net, score, dst, i);
1761
1762 if (minihiscore > miniscore) {
1763 if (i == IPV6_SADDR_RULE_SCOPE &&
1764 score->scopedist > 0) {
1765 /*
1766 * special case:
1767 * each remaining entry
1768 * has too small (not enough)
1769 * scope, because ifa entries
1770 * are sorted by their scope
1771 * values.
1772 */
1773 goto out;
1774 }
1775 break;
1776 } else if (minihiscore < miniscore) {
1777 swap(hiscore, score);
1778 hiscore_idx = 1 - hiscore_idx;
1779
1780 /* restore our iterator */
1781 score->ifa = hiscore->ifa;
1782
1783 break;
1784 }
1785 }
1786 }
1787 out:
1788 return hiscore_idx;
1789 }
1790
ipv6_get_saddr_master(struct net * net,const struct net_device * dst_dev,const struct net_device * master,struct ipv6_saddr_dst * dst,struct ipv6_saddr_score * scores,int hiscore_idx)1791 static int ipv6_get_saddr_master(struct net *net,
1792 const struct net_device *dst_dev,
1793 const struct net_device *master,
1794 struct ipv6_saddr_dst *dst,
1795 struct ipv6_saddr_score *scores,
1796 int hiscore_idx)
1797 {
1798 struct inet6_dev *idev;
1799
1800 idev = __in6_dev_get(dst_dev);
1801 if (idev)
1802 hiscore_idx = __ipv6_dev_get_saddr(net, dst, idev,
1803 scores, hiscore_idx);
1804
1805 idev = __in6_dev_get(master);
1806 if (idev)
1807 hiscore_idx = __ipv6_dev_get_saddr(net, dst, idev,
1808 scores, hiscore_idx);
1809
1810 return hiscore_idx;
1811 }
1812
ipv6_dev_get_saddr(struct net * net,const struct net_device * dst_dev,const struct in6_addr * daddr,unsigned int prefs,struct in6_addr * saddr)1813 int ipv6_dev_get_saddr(struct net *net, const struct net_device *dst_dev,
1814 const struct in6_addr *daddr, unsigned int prefs,
1815 struct in6_addr *saddr)
1816 {
1817 struct ipv6_saddr_score scores[2], *hiscore;
1818 struct ipv6_saddr_dst dst;
1819 struct inet6_dev *idev;
1820 struct net_device *dev;
1821 int dst_type;
1822 bool use_oif_addr = false;
1823 int hiscore_idx = 0;
1824 int ret = 0;
1825
1826 dst_type = __ipv6_addr_type(daddr);
1827 dst.addr = daddr;
1828 dst.ifindex = dst_dev ? dst_dev->ifindex : 0;
1829 dst.scope = __ipv6_addr_src_scope(dst_type);
1830 dst.label = ipv6_addr_label(net, daddr, dst_type, dst.ifindex);
1831 dst.prefs = prefs;
1832
1833 scores[hiscore_idx].rule = -1;
1834 scores[hiscore_idx].ifa = NULL;
1835
1836 rcu_read_lock();
1837
1838 /* Candidate Source Address (section 4)
1839 * - multicast and link-local destination address,
1840 * the set of candidate source address MUST only
1841 * include addresses assigned to interfaces
1842 * belonging to the same link as the outgoing
1843 * interface.
1844 * (- For site-local destination addresses, the
1845 * set of candidate source addresses MUST only
1846 * include addresses assigned to interfaces
1847 * belonging to the same site as the outgoing
1848 * interface.)
1849 * - "It is RECOMMENDED that the candidate source addresses
1850 * be the set of unicast addresses assigned to the
1851 * interface that will be used to send to the destination
1852 * (the 'outgoing' interface)." (RFC 6724)
1853 */
1854 if (dst_dev) {
1855 idev = __in6_dev_get(dst_dev);
1856 if ((dst_type & IPV6_ADDR_MULTICAST) ||
1857 dst.scope <= IPV6_ADDR_SCOPE_LINKLOCAL ||
1858 (idev && READ_ONCE(idev->cnf.use_oif_addrs_only))) {
1859 use_oif_addr = true;
1860 }
1861 }
1862
1863 if (use_oif_addr) {
1864 if (idev)
1865 hiscore_idx = __ipv6_dev_get_saddr(net, &dst, idev, scores, hiscore_idx);
1866 } else {
1867 const struct net_device *master;
1868 int master_idx = 0;
1869
1870 /* if dst_dev exists and is enslaved to an L3 device, then
1871 * prefer addresses from dst_dev and then the master over
1872 * any other enslaved devices in the L3 domain.
1873 */
1874 master = l3mdev_master_dev_rcu(dst_dev);
1875 if (master) {
1876 master_idx = master->ifindex;
1877
1878 hiscore_idx = ipv6_get_saddr_master(net, dst_dev,
1879 master, &dst,
1880 scores, hiscore_idx);
1881
1882 if (scores[hiscore_idx].ifa &&
1883 scores[hiscore_idx].scopedist >= 0)
1884 goto out;
1885 }
1886
1887 for_each_netdev_rcu(net, dev) {
1888 /* only consider addresses on devices in the
1889 * same L3 domain
1890 */
1891 if (l3mdev_master_ifindex_rcu(dev) != master_idx)
1892 continue;
1893 idev = __in6_dev_get(dev);
1894 if (!idev)
1895 continue;
1896 hiscore_idx = __ipv6_dev_get_saddr(net, &dst, idev, scores, hiscore_idx);
1897 }
1898 }
1899
1900 out:
1901 hiscore = &scores[hiscore_idx];
1902 if (!hiscore->ifa)
1903 ret = -EADDRNOTAVAIL;
1904 else
1905 *saddr = hiscore->ifa->addr;
1906
1907 rcu_read_unlock();
1908 return ret;
1909 }
1910 EXPORT_SYMBOL(ipv6_dev_get_saddr);
1911
__ipv6_get_lladdr(struct inet6_dev * idev,struct in6_addr * addr,u32 banned_flags)1912 static int __ipv6_get_lladdr(struct inet6_dev *idev, struct in6_addr *addr,
1913 u32 banned_flags)
1914 {
1915 struct inet6_ifaddr *ifp;
1916 int err = -EADDRNOTAVAIL;
1917
1918 list_for_each_entry_reverse(ifp, &idev->addr_list, if_list) {
1919 if (ifp->scope > IFA_LINK)
1920 break;
1921 if (ifp->scope == IFA_LINK &&
1922 !(ifp->flags & banned_flags)) {
1923 *addr = ifp->addr;
1924 err = 0;
1925 break;
1926 }
1927 }
1928 return err;
1929 }
1930
ipv6_get_lladdr(struct net_device * dev,struct in6_addr * addr,u32 banned_flags)1931 int ipv6_get_lladdr(struct net_device *dev, struct in6_addr *addr,
1932 u32 banned_flags)
1933 {
1934 struct inet6_dev *idev;
1935 int err = -EADDRNOTAVAIL;
1936
1937 rcu_read_lock();
1938 idev = __in6_dev_get(dev);
1939 if (idev) {
1940 read_lock_bh(&idev->lock);
1941 err = __ipv6_get_lladdr(idev, addr, banned_flags);
1942 read_unlock_bh(&idev->lock);
1943 }
1944 rcu_read_unlock();
1945 return err;
1946 }
1947
ipv6_count_addresses(const struct inet6_dev * idev)1948 static int ipv6_count_addresses(const struct inet6_dev *idev)
1949 {
1950 const struct inet6_ifaddr *ifp;
1951 int cnt = 0;
1952
1953 rcu_read_lock();
1954 list_for_each_entry_rcu(ifp, &idev->addr_list, if_list)
1955 cnt++;
1956 rcu_read_unlock();
1957 return cnt;
1958 }
1959
ipv6_chk_addr(struct net * net,const struct in6_addr * addr,const struct net_device * dev,int strict)1960 int ipv6_chk_addr(struct net *net, const struct in6_addr *addr,
1961 const struct net_device *dev, int strict)
1962 {
1963 return ipv6_chk_addr_and_flags(net, addr, dev, !dev,
1964 strict, IFA_F_TENTATIVE);
1965 }
1966 EXPORT_SYMBOL(ipv6_chk_addr);
1967
1968 /* device argument is used to find the L3 domain of interest. If
1969 * skip_dev_check is set, then the ifp device is not checked against
1970 * the passed in dev argument. So the 2 cases for addresses checks are:
1971 * 1. does the address exist in the L3 domain that dev is part of
1972 * (skip_dev_check = true), or
1973 *
1974 * 2. does the address exist on the specific device
1975 * (skip_dev_check = false)
1976 */
1977 static struct net_device *
__ipv6_chk_addr_and_flags(struct net * net,const struct in6_addr * addr,const struct net_device * dev,bool skip_dev_check,int strict,u32 banned_flags)1978 __ipv6_chk_addr_and_flags(struct net *net, const struct in6_addr *addr,
1979 const struct net_device *dev, bool skip_dev_check,
1980 int strict, u32 banned_flags)
1981 {
1982 unsigned int hash = inet6_addr_hash(net, addr);
1983 struct net_device *l3mdev, *ndev;
1984 struct inet6_ifaddr *ifp;
1985 u32 ifp_flags;
1986
1987 rcu_read_lock();
1988
1989 l3mdev = l3mdev_master_dev_rcu(dev);
1990 if (skip_dev_check)
1991 dev = NULL;
1992
1993 hlist_for_each_entry_rcu(ifp, &net->ipv6.inet6_addr_lst[hash], addr_lst) {
1994 ndev = ifp->idev->dev;
1995
1996 if (l3mdev_master_dev_rcu(ndev) != l3mdev)
1997 continue;
1998
1999 /* Decouple optimistic from tentative for evaluation here.
2000 * Ban optimistic addresses explicitly, when required.
2001 */
2002 ifp_flags = (ifp->flags&IFA_F_OPTIMISTIC)
2003 ? (ifp->flags&~IFA_F_TENTATIVE)
2004 : ifp->flags;
2005 if (ipv6_addr_equal(&ifp->addr, addr) &&
2006 !(ifp_flags&banned_flags) &&
2007 (!dev || ndev == dev ||
2008 !(ifp->scope&(IFA_LINK|IFA_HOST) || strict))) {
2009 rcu_read_unlock();
2010 return ndev;
2011 }
2012 }
2013
2014 rcu_read_unlock();
2015 return NULL;
2016 }
2017
ipv6_chk_addr_and_flags(struct net * net,const struct in6_addr * addr,const struct net_device * dev,bool skip_dev_check,int strict,u32 banned_flags)2018 int ipv6_chk_addr_and_flags(struct net *net, const struct in6_addr *addr,
2019 const struct net_device *dev, bool skip_dev_check,
2020 int strict, u32 banned_flags)
2021 {
2022 return __ipv6_chk_addr_and_flags(net, addr, dev, skip_dev_check,
2023 strict, banned_flags) ? 1 : 0;
2024 }
2025 EXPORT_SYMBOL(ipv6_chk_addr_and_flags);
2026
2027
2028 /* Compares an address/prefix_len with addresses on device @dev.
2029 * If one is found it returns true.
2030 */
ipv6_chk_custom_prefix(const struct in6_addr * addr,const unsigned int prefix_len,struct net_device * dev)2031 bool ipv6_chk_custom_prefix(const struct in6_addr *addr,
2032 const unsigned int prefix_len, struct net_device *dev)
2033 {
2034 const struct inet6_ifaddr *ifa;
2035 const struct inet6_dev *idev;
2036 bool ret = false;
2037
2038 rcu_read_lock();
2039 idev = __in6_dev_get(dev);
2040 if (idev) {
2041 list_for_each_entry_rcu(ifa, &idev->addr_list, if_list) {
2042 ret = ipv6_prefix_equal(addr, &ifa->addr, prefix_len);
2043 if (ret)
2044 break;
2045 }
2046 }
2047 rcu_read_unlock();
2048
2049 return ret;
2050 }
2051 EXPORT_SYMBOL(ipv6_chk_custom_prefix);
2052
ipv6_chk_prefix(const struct in6_addr * addr,struct net_device * dev)2053 int ipv6_chk_prefix(const struct in6_addr *addr, struct net_device *dev)
2054 {
2055 const struct inet6_ifaddr *ifa;
2056 const struct inet6_dev *idev;
2057 int onlink;
2058
2059 onlink = 0;
2060 rcu_read_lock();
2061 idev = __in6_dev_get(dev);
2062 if (idev) {
2063 list_for_each_entry_rcu(ifa, &idev->addr_list, if_list) {
2064 onlink = ipv6_prefix_equal(addr, &ifa->addr,
2065 ifa->prefix_len);
2066 if (onlink)
2067 break;
2068 }
2069 }
2070 rcu_read_unlock();
2071 return onlink;
2072 }
2073 EXPORT_SYMBOL(ipv6_chk_prefix);
2074
2075 /**
2076 * ipv6_dev_find - find the first device with a given source address.
2077 * @net: the net namespace
2078 * @addr: the source address
2079 * @dev: used to find the L3 domain of interest
2080 *
2081 * The caller should be protected by RCU, or RTNL.
2082 */
ipv6_dev_find(struct net * net,const struct in6_addr * addr,struct net_device * dev)2083 struct net_device *ipv6_dev_find(struct net *net, const struct in6_addr *addr,
2084 struct net_device *dev)
2085 {
2086 return __ipv6_chk_addr_and_flags(net, addr, dev, !dev, 1,
2087 IFA_F_TENTATIVE);
2088 }
2089 EXPORT_SYMBOL(ipv6_dev_find);
2090
ipv6_get_ifaddr(struct net * net,const struct in6_addr * addr,struct net_device * dev,int strict)2091 struct inet6_ifaddr *ipv6_get_ifaddr(struct net *net, const struct in6_addr *addr,
2092 struct net_device *dev, int strict)
2093 {
2094 unsigned int hash = inet6_addr_hash(net, addr);
2095 struct inet6_ifaddr *ifp, *result = NULL;
2096
2097 rcu_read_lock();
2098 hlist_for_each_entry_rcu(ifp, &net->ipv6.inet6_addr_lst[hash], addr_lst) {
2099 if (ipv6_addr_equal(&ifp->addr, addr)) {
2100 if (!dev || ifp->idev->dev == dev ||
2101 !(ifp->scope&(IFA_LINK|IFA_HOST) || strict)) {
2102 if (in6_ifa_hold_safe(ifp)) {
2103 result = ifp;
2104 break;
2105 }
2106 }
2107 }
2108 }
2109 rcu_read_unlock();
2110
2111 return result;
2112 }
2113
2114 /* Gets referenced address, destroys ifaddr */
2115
addrconf_dad_stop(struct inet6_ifaddr * ifp,int dad_failed)2116 static void addrconf_dad_stop(struct inet6_ifaddr *ifp, int dad_failed)
2117 {
2118 if (dad_failed)
2119 ifp->flags |= IFA_F_DADFAILED;
2120
2121 if (ifp->flags&IFA_F_TEMPORARY) {
2122 struct inet6_ifaddr *ifpub;
2123 spin_lock_bh(&ifp->lock);
2124 ifpub = ifp->ifpub;
2125 if (ifpub) {
2126 in6_ifa_hold(ifpub);
2127 spin_unlock_bh(&ifp->lock);
2128 ipv6_create_tempaddr(ifpub, true);
2129 in6_ifa_put(ifpub);
2130 } else {
2131 spin_unlock_bh(&ifp->lock);
2132 }
2133 ipv6_del_addr(ifp);
2134 } else if (ifp->flags&IFA_F_PERMANENT || !dad_failed) {
2135 spin_lock_bh(&ifp->lock);
2136 addrconf_del_dad_work(ifp);
2137 ifp->flags |= IFA_F_TENTATIVE;
2138 if (dad_failed)
2139 ifp->flags &= ~IFA_F_OPTIMISTIC;
2140 spin_unlock_bh(&ifp->lock);
2141 if (dad_failed)
2142 ipv6_ifa_notify(0, ifp);
2143 in6_ifa_put(ifp);
2144 } else {
2145 ipv6_del_addr(ifp);
2146 }
2147 }
2148
addrconf_dad_end(struct inet6_ifaddr * ifp)2149 static int addrconf_dad_end(struct inet6_ifaddr *ifp)
2150 {
2151 int err = -ENOENT;
2152
2153 spin_lock_bh(&ifp->lock);
2154 if (ifp->state == INET6_IFADDR_STATE_DAD) {
2155 ifp->state = INET6_IFADDR_STATE_POSTDAD;
2156 err = 0;
2157 }
2158 spin_unlock_bh(&ifp->lock);
2159
2160 return err;
2161 }
2162
addrconf_dad_failure(struct sk_buff * skb,struct inet6_ifaddr * ifp)2163 void addrconf_dad_failure(struct sk_buff *skb, struct inet6_ifaddr *ifp)
2164 {
2165 struct inet6_dev *idev = ifp->idev;
2166 struct net *net = dev_net(idev->dev);
2167 int max_addresses;
2168
2169 if (addrconf_dad_end(ifp)) {
2170 in6_ifa_put(ifp);
2171 return;
2172 }
2173
2174 net_info_ratelimited("%s: IPv6 duplicate address %pI6c used by %pM detected!\n",
2175 ifp->idev->dev->name, &ifp->addr, eth_hdr(skb)->h_source);
2176
2177 spin_lock_bh(&ifp->lock);
2178
2179 if (ifp->flags & IFA_F_STABLE_PRIVACY) {
2180 struct in6_addr new_addr;
2181 struct inet6_ifaddr *ifp2;
2182 int retries = ifp->stable_privacy_retry + 1;
2183 struct ifa6_config cfg = {
2184 .pfx = &new_addr,
2185 .plen = ifp->prefix_len,
2186 .ifa_flags = ifp->flags,
2187 .valid_lft = ifp->valid_lft,
2188 .preferred_lft = ifp->prefered_lft,
2189 .scope = ifp->scope,
2190 };
2191
2192 if (retries > net->ipv6.sysctl.idgen_retries) {
2193 net_info_ratelimited("%s: privacy stable address generation failed because of DAD conflicts!\n",
2194 ifp->idev->dev->name);
2195 goto errdad;
2196 }
2197
2198 new_addr = ifp->addr;
2199 if (ipv6_generate_stable_address(&new_addr, retries,
2200 idev))
2201 goto errdad;
2202
2203 spin_unlock_bh(&ifp->lock);
2204
2205 max_addresses = READ_ONCE(idev->cnf.max_addresses);
2206 if (max_addresses &&
2207 ipv6_count_addresses(idev) >= max_addresses)
2208 goto lock_errdad;
2209
2210 net_info_ratelimited("%s: generating new stable privacy address because of DAD conflict\n",
2211 ifp->idev->dev->name);
2212
2213 ifp2 = ipv6_add_addr(idev, &cfg, false, NULL);
2214 if (IS_ERR(ifp2))
2215 goto lock_errdad;
2216
2217 spin_lock_bh(&ifp2->lock);
2218 ifp2->stable_privacy_retry = retries;
2219 ifp2->state = INET6_IFADDR_STATE_PREDAD;
2220 spin_unlock_bh(&ifp2->lock);
2221
2222 addrconf_mod_dad_work(ifp2, net->ipv6.sysctl.idgen_delay);
2223 in6_ifa_put(ifp2);
2224 lock_errdad:
2225 spin_lock_bh(&ifp->lock);
2226 }
2227
2228 errdad:
2229 /* transition from _POSTDAD to _ERRDAD */
2230 ifp->state = INET6_IFADDR_STATE_ERRDAD;
2231 spin_unlock_bh(&ifp->lock);
2232
2233 addrconf_mod_dad_work(ifp, 0);
2234 in6_ifa_put(ifp);
2235 }
2236
2237 /* Join to solicited addr multicast group. */
addrconf_join_solict(struct net_device * dev,const struct in6_addr * addr)2238 void addrconf_join_solict(struct net_device *dev, const struct in6_addr *addr)
2239 {
2240 struct in6_addr maddr;
2241
2242 if (READ_ONCE(dev->flags) & (IFF_LOOPBACK | IFF_NOARP))
2243 return;
2244
2245 addrconf_addr_solict_mult(addr, &maddr);
2246 ipv6_dev_mc_inc(dev, &maddr);
2247 }
2248
addrconf_leave_solict(struct inet6_dev * idev,const struct in6_addr * addr)2249 void addrconf_leave_solict(struct inet6_dev *idev, const struct in6_addr *addr)
2250 {
2251 struct in6_addr maddr;
2252
2253 if (READ_ONCE(idev->dev->flags) & (IFF_LOOPBACK | IFF_NOARP))
2254 return;
2255
2256 addrconf_addr_solict_mult(addr, &maddr);
2257 __ipv6_dev_mc_dec(idev, &maddr);
2258 }
2259
addrconf_join_anycast(struct inet6_ifaddr * ifp)2260 static void addrconf_join_anycast(struct inet6_ifaddr *ifp)
2261 {
2262 struct in6_addr addr;
2263
2264 if (ifp->prefix_len >= 127) /* RFC 6164 */
2265 return;
2266 ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
2267 if (ipv6_addr_any(&addr))
2268 return;
2269 __ipv6_dev_ac_inc(ifp->idev, &addr);
2270 }
2271
addrconf_leave_anycast(struct inet6_ifaddr * ifp)2272 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp)
2273 {
2274 struct in6_addr addr;
2275
2276 if (ifp->prefix_len >= 127) /* RFC 6164 */
2277 return;
2278 ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
2279 if (ipv6_addr_any(&addr))
2280 return;
2281 __ipv6_dev_ac_dec(ifp->idev, &addr);
2282 }
2283
addrconf_ifid_6lowpan(u8 * eui,struct net_device * dev)2284 static int addrconf_ifid_6lowpan(u8 *eui, struct net_device *dev)
2285 {
2286 switch (dev->addr_len) {
2287 case ETH_ALEN:
2288 memcpy(eui, dev->dev_addr, 3);
2289 eui[3] = 0xFF;
2290 eui[4] = 0xFE;
2291 memcpy(eui + 5, dev->dev_addr + 3, 3);
2292 break;
2293 case EUI64_ADDR_LEN:
2294 memcpy(eui, dev->dev_addr, EUI64_ADDR_LEN);
2295 eui[0] ^= 2;
2296 break;
2297 default:
2298 return -1;
2299 }
2300
2301 return 0;
2302 }
2303
addrconf_ifid_ieee1394(u8 * eui,struct net_device * dev)2304 static int addrconf_ifid_ieee1394(u8 *eui, struct net_device *dev)
2305 {
2306 const union fwnet_hwaddr *ha;
2307
2308 if (dev->addr_len != FWNET_ALEN)
2309 return -1;
2310
2311 ha = (const union fwnet_hwaddr *)dev->dev_addr;
2312
2313 memcpy(eui, &ha->uc.uniq_id, sizeof(ha->uc.uniq_id));
2314 eui[0] ^= 2;
2315 return 0;
2316 }
2317
addrconf_ifid_arcnet(u8 * eui,struct net_device * dev)2318 static int addrconf_ifid_arcnet(u8 *eui, struct net_device *dev)
2319 {
2320 /* XXX: inherit EUI-64 from other interface -- yoshfuji */
2321 if (dev->addr_len != ARCNET_ALEN)
2322 return -1;
2323 memset(eui, 0, 7);
2324 eui[7] = *(u8 *)dev->dev_addr;
2325 return 0;
2326 }
2327
addrconf_ifid_infiniband(u8 * eui,struct net_device * dev)2328 static int addrconf_ifid_infiniband(u8 *eui, struct net_device *dev)
2329 {
2330 if (dev->addr_len != INFINIBAND_ALEN)
2331 return -1;
2332 memcpy(eui, dev->dev_addr + 12, 8);
2333 eui[0] |= 2;
2334 return 0;
2335 }
2336
__ipv6_isatap_ifid(u8 * eui,__be32 addr)2337 static int __ipv6_isatap_ifid(u8 *eui, __be32 addr)
2338 {
2339 if (addr == 0)
2340 return -1;
2341 eui[0] = (ipv4_is_zeronet(addr) || ipv4_is_private_10(addr) ||
2342 ipv4_is_loopback(addr) || ipv4_is_linklocal_169(addr) ||
2343 ipv4_is_private_172(addr) || ipv4_is_test_192(addr) ||
2344 ipv4_is_anycast_6to4(addr) || ipv4_is_private_192(addr) ||
2345 ipv4_is_test_198(addr) || ipv4_is_multicast(addr) ||
2346 ipv4_is_lbcast(addr)) ? 0x00 : 0x02;
2347 eui[1] = 0;
2348 eui[2] = 0x5E;
2349 eui[3] = 0xFE;
2350 memcpy(eui + 4, &addr, 4);
2351 return 0;
2352 }
2353
addrconf_ifid_sit(u8 * eui,struct net_device * dev)2354 static int addrconf_ifid_sit(u8 *eui, struct net_device *dev)
2355 {
2356 if (dev->priv_flags & IFF_ISATAP)
2357 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
2358 return -1;
2359 }
2360
addrconf_ifid_gre(u8 * eui,struct net_device * dev)2361 static int addrconf_ifid_gre(u8 *eui, struct net_device *dev)
2362 {
2363 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
2364 }
2365
addrconf_ifid_ip6tnl(u8 * eui,struct net_device * dev)2366 static int addrconf_ifid_ip6tnl(u8 *eui, struct net_device *dev)
2367 {
2368 memcpy(eui, dev->perm_addr, 3);
2369 memcpy(eui + 5, dev->perm_addr + 3, 3);
2370 eui[3] = 0xFF;
2371 eui[4] = 0xFE;
2372 eui[0] ^= 2;
2373 return 0;
2374 }
2375
ipv6_generate_eui64(u8 * eui,struct net_device * dev)2376 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev)
2377 {
2378 switch (dev->type) {
2379 case ARPHRD_ETHER:
2380 case ARPHRD_FDDI:
2381 return addrconf_ifid_eui48(eui, dev);
2382 case ARPHRD_ARCNET:
2383 return addrconf_ifid_arcnet(eui, dev);
2384 case ARPHRD_INFINIBAND:
2385 return addrconf_ifid_infiniband(eui, dev);
2386 case ARPHRD_SIT:
2387 return addrconf_ifid_sit(eui, dev);
2388 case ARPHRD_IPGRE:
2389 case ARPHRD_TUNNEL:
2390 return addrconf_ifid_gre(eui, dev);
2391 case ARPHRD_6LOWPAN:
2392 return addrconf_ifid_6lowpan(eui, dev);
2393 case ARPHRD_IEEE1394:
2394 return addrconf_ifid_ieee1394(eui, dev);
2395 case ARPHRD_TUNNEL6:
2396 case ARPHRD_IP6GRE:
2397 case ARPHRD_RAWIP:
2398 return addrconf_ifid_ip6tnl(eui, dev);
2399 }
2400 return -1;
2401 }
2402
ipv6_inherit_eui64(u8 * eui,struct inet6_dev * idev)2403 static int ipv6_inherit_eui64(u8 *eui, struct inet6_dev *idev)
2404 {
2405 int err = -1;
2406 struct inet6_ifaddr *ifp;
2407
2408 read_lock_bh(&idev->lock);
2409 list_for_each_entry_reverse(ifp, &idev->addr_list, if_list) {
2410 if (ifp->scope > IFA_LINK)
2411 break;
2412 if (ifp->scope == IFA_LINK && !(ifp->flags&IFA_F_TENTATIVE)) {
2413 memcpy(eui, ifp->addr.s6_addr+8, 8);
2414 err = 0;
2415 break;
2416 }
2417 }
2418 read_unlock_bh(&idev->lock);
2419 return err;
2420 }
2421
2422 /* Generation of a randomized Interface Identifier
2423 * draft-ietf-6man-rfc4941bis, Section 3.3.1
2424 */
2425
ipv6_gen_rnd_iid(struct in6_addr * addr)2426 static void ipv6_gen_rnd_iid(struct in6_addr *addr)
2427 {
2428 regen:
2429 get_random_bytes(&addr->s6_addr[8], 8);
2430
2431 /* <draft-ietf-6man-rfc4941bis-08.txt>, Section 3.3.1:
2432 * check if generated address is not inappropriate:
2433 *
2434 * - Reserved IPv6 Interface Identifiers
2435 * - XXX: already assigned to an address on the device
2436 */
2437
2438 /* Subnet-router anycast: 0000:0000:0000:0000 */
2439 if (!(addr->s6_addr32[2] | addr->s6_addr32[3]))
2440 goto regen;
2441
2442 /* IANA Ethernet block: 0200:5EFF:FE00:0000-0200:5EFF:FE00:5212
2443 * Proxy Mobile IPv6: 0200:5EFF:FE00:5213
2444 * IANA Ethernet block: 0200:5EFF:FE00:5214-0200:5EFF:FEFF:FFFF
2445 */
2446 if (ntohl(addr->s6_addr32[2]) == 0x02005eff &&
2447 (ntohl(addr->s6_addr32[3]) & 0Xff000000) == 0xfe000000)
2448 goto regen;
2449
2450 /* Reserved subnet anycast addresses */
2451 if (ntohl(addr->s6_addr32[2]) == 0xfdffffff &&
2452 ntohl(addr->s6_addr32[3]) >= 0Xffffff80)
2453 goto regen;
2454 }
2455
2456 /*
2457 * Add prefix route.
2458 */
2459
2460 static void
addrconf_prefix_route(struct in6_addr * pfx,int plen,u32 metric,struct net_device * dev,unsigned long expires,u32 flags,gfp_t gfp_flags)2461 addrconf_prefix_route(struct in6_addr *pfx, int plen, u32 metric,
2462 struct net_device *dev, unsigned long expires,
2463 u32 flags, gfp_t gfp_flags)
2464 {
2465 struct fib6_config cfg = {
2466 .fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_PREFIX,
2467 .fc_metric = metric ? : IP6_RT_PRIO_ADDRCONF,
2468 .fc_ifindex = dev->ifindex,
2469 .fc_expires = expires,
2470 .fc_dst_len = plen,
2471 .fc_flags = RTF_UP | flags,
2472 .fc_nlinfo.nl_net = dev_net(dev),
2473 .fc_protocol = RTPROT_KERNEL,
2474 .fc_type = RTN_UNICAST,
2475 };
2476
2477 cfg.fc_dst = *pfx;
2478
2479 /* Prevent useless cloning on PtP SIT.
2480 This thing is done here expecting that the whole
2481 class of non-broadcast devices need not cloning.
2482 */
2483 #if IS_ENABLED(CONFIG_IPV6_SIT)
2484 if (dev->type == ARPHRD_SIT && (dev->flags & IFF_POINTOPOINT))
2485 cfg.fc_flags |= RTF_NONEXTHOP;
2486 #endif
2487
2488 ip6_route_add(&cfg, gfp_flags, NULL);
2489 }
2490
2491
addrconf_get_prefix_route(const struct in6_addr * pfx,int plen,const struct net_device * dev,u32 flags,u32 noflags,bool no_gw)2492 static struct fib6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
2493 int plen,
2494 const struct net_device *dev,
2495 u32 flags, u32 noflags,
2496 bool no_gw)
2497 {
2498 struct fib6_node *fn;
2499 struct fib6_info *rt = NULL;
2500 struct fib6_table *table;
2501 u32 tb_id = l3mdev_fib_table(dev) ? : RT6_TABLE_PREFIX;
2502
2503 table = fib6_get_table(dev_net(dev), tb_id);
2504 if (!table)
2505 return NULL;
2506
2507 rcu_read_lock();
2508 fn = fib6_locate(&table->tb6_root, pfx, plen, NULL, 0, true);
2509 if (!fn)
2510 goto out;
2511
2512 for_each_fib6_node_rt_rcu(fn) {
2513 /* prefix routes only use builtin fib6_nh */
2514 if (rt->nh)
2515 continue;
2516
2517 if (rt->fib6_nh->fib_nh_dev->ifindex != dev->ifindex)
2518 continue;
2519 if (no_gw && rt->fib6_nh->fib_nh_gw_family)
2520 continue;
2521 if ((rt->fib6_flags & flags) != flags)
2522 continue;
2523 if ((rt->fib6_flags & noflags) != 0)
2524 continue;
2525 if (!fib6_info_hold_safe(rt))
2526 continue;
2527 break;
2528 }
2529 out:
2530 rcu_read_unlock();
2531 return rt;
2532 }
2533
2534
2535 /* Create "default" multicast route to the interface */
2536
addrconf_add_mroute(struct net_device * dev)2537 static void addrconf_add_mroute(struct net_device *dev)
2538 {
2539 struct fib6_config cfg = {
2540 .fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_LOCAL,
2541 .fc_metric = IP6_RT_PRIO_ADDRCONF,
2542 .fc_ifindex = dev->ifindex,
2543 .fc_dst_len = 8,
2544 .fc_flags = RTF_UP,
2545 .fc_type = RTN_MULTICAST,
2546 .fc_nlinfo.nl_net = dev_net(dev),
2547 .fc_protocol = RTPROT_KERNEL,
2548 };
2549
2550 ipv6_addr_set(&cfg.fc_dst, htonl(0xFF000000), 0, 0, 0);
2551
2552 ip6_route_add(&cfg, GFP_KERNEL, NULL);
2553 }
2554
addrconf_add_dev(struct net_device * dev)2555 static struct inet6_dev *addrconf_add_dev(struct net_device *dev)
2556 {
2557 struct inet6_dev *idev;
2558
2559 ASSERT_RTNL();
2560
2561 idev = ipv6_find_idev(dev);
2562 if (IS_ERR(idev))
2563 return idev;
2564
2565 if (idev->cnf.disable_ipv6)
2566 return ERR_PTR(-EACCES);
2567
2568 /* Add default multicast route */
2569 if (!(dev->flags & IFF_LOOPBACK) && !netif_is_l3_master(dev))
2570 addrconf_add_mroute(dev);
2571
2572 return idev;
2573 }
2574
delete_tempaddrs(struct inet6_dev * idev,struct inet6_ifaddr * ifp)2575 static void delete_tempaddrs(struct inet6_dev *idev,
2576 struct inet6_ifaddr *ifp)
2577 {
2578 struct inet6_ifaddr *ift, *tmp;
2579
2580 write_lock_bh(&idev->lock);
2581 list_for_each_entry_safe(ift, tmp, &idev->tempaddr_list, tmp_list) {
2582 if (ift->ifpub != ifp)
2583 continue;
2584
2585 in6_ifa_hold(ift);
2586 write_unlock_bh(&idev->lock);
2587 ipv6_del_addr(ift);
2588 write_lock_bh(&idev->lock);
2589 }
2590 write_unlock_bh(&idev->lock);
2591 }
2592
manage_tempaddrs(struct inet6_dev * idev,struct inet6_ifaddr * ifp,__u32 valid_lft,__u32 prefered_lft,bool create,unsigned long now)2593 static void manage_tempaddrs(struct inet6_dev *idev,
2594 struct inet6_ifaddr *ifp,
2595 __u32 valid_lft, __u32 prefered_lft,
2596 bool create, unsigned long now)
2597 {
2598 u32 flags;
2599 struct inet6_ifaddr *ift;
2600
2601 read_lock_bh(&idev->lock);
2602 /* update all temporary addresses in the list */
2603 list_for_each_entry(ift, &idev->tempaddr_list, tmp_list) {
2604 int age, max_valid, max_prefered;
2605
2606 if (ifp != ift->ifpub)
2607 continue;
2608
2609 /* RFC 4941 section 3.3:
2610 * If a received option will extend the lifetime of a public
2611 * address, the lifetimes of temporary addresses should
2612 * be extended, subject to the overall constraint that no
2613 * temporary addresses should ever remain "valid" or "preferred"
2614 * for a time longer than (TEMP_VALID_LIFETIME) or
2615 * (TEMP_PREFERRED_LIFETIME - DESYNC_FACTOR), respectively.
2616 */
2617 age = (now - ift->cstamp) / HZ;
2618 max_valid = READ_ONCE(idev->cnf.temp_valid_lft) - age;
2619 if (max_valid < 0)
2620 max_valid = 0;
2621
2622 max_prefered = READ_ONCE(idev->cnf.temp_prefered_lft) -
2623 idev->desync_factor - age;
2624 if (max_prefered < 0)
2625 max_prefered = 0;
2626
2627 if (valid_lft > max_valid)
2628 valid_lft = max_valid;
2629
2630 if (prefered_lft > max_prefered)
2631 prefered_lft = max_prefered;
2632
2633 spin_lock(&ift->lock);
2634 flags = ift->flags;
2635 ift->valid_lft = valid_lft;
2636 ift->prefered_lft = prefered_lft;
2637 ift->tstamp = now;
2638 if (prefered_lft > 0)
2639 ift->flags &= ~IFA_F_DEPRECATED;
2640
2641 spin_unlock(&ift->lock);
2642 if (!(flags&IFA_F_TENTATIVE))
2643 ipv6_ifa_notify(0, ift);
2644 }
2645
2646 /* Also create a temporary address if it's enabled but no temporary
2647 * address currently exists.
2648 * However, we get called with valid_lft == 0, prefered_lft == 0, create == false
2649 * as part of cleanup (ie. deleting the mngtmpaddr).
2650 * We don't want that to result in creating a new temporary ip address.
2651 */
2652 if (list_empty(&idev->tempaddr_list) && (valid_lft || prefered_lft))
2653 create = true;
2654
2655 if (create && READ_ONCE(idev->cnf.use_tempaddr) > 0) {
2656 /* When a new public address is created as described
2657 * in [ADDRCONF], also create a new temporary address.
2658 */
2659 read_unlock_bh(&idev->lock);
2660 ipv6_create_tempaddr(ifp, false);
2661 } else {
2662 read_unlock_bh(&idev->lock);
2663 }
2664 }
2665
is_addr_mode_generate_stable(struct inet6_dev * idev)2666 static bool is_addr_mode_generate_stable(struct inet6_dev *idev)
2667 {
2668 return idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_STABLE_PRIVACY ||
2669 idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_RANDOM;
2670 }
2671
addrconf_prefix_rcv_add_addr(struct net * net,struct net_device * dev,const struct prefix_info * pinfo,struct inet6_dev * in6_dev,const struct in6_addr * addr,int addr_type,u32 addr_flags,bool sllao,bool tokenized,__u32 valid_lft,u32 prefered_lft)2672 int addrconf_prefix_rcv_add_addr(struct net *net, struct net_device *dev,
2673 const struct prefix_info *pinfo,
2674 struct inet6_dev *in6_dev,
2675 const struct in6_addr *addr, int addr_type,
2676 u32 addr_flags, bool sllao, bool tokenized,
2677 __u32 valid_lft, u32 prefered_lft)
2678 {
2679 struct inet6_ifaddr *ifp = ipv6_get_ifaddr(net, addr, dev, 1);
2680 int create = 0, update_lft = 0;
2681
2682 if (!ifp && valid_lft) {
2683 int max_addresses = READ_ONCE(in6_dev->cnf.max_addresses);
2684 struct ifa6_config cfg = {
2685 .pfx = addr,
2686 .plen = pinfo->prefix_len,
2687 .ifa_flags = addr_flags,
2688 .valid_lft = valid_lft,
2689 .preferred_lft = prefered_lft,
2690 .scope = addr_type & IPV6_ADDR_SCOPE_MASK,
2691 .ifa_proto = IFAPROT_KERNEL_RA
2692 };
2693
2694 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2695 if ((READ_ONCE(net->ipv6.devconf_all->optimistic_dad) ||
2696 READ_ONCE(in6_dev->cnf.optimistic_dad)) &&
2697 !net->ipv6.devconf_all->forwarding && sllao)
2698 cfg.ifa_flags |= IFA_F_OPTIMISTIC;
2699 #endif
2700
2701 /* Do not allow to create too much of autoconfigured
2702 * addresses; this would be too easy way to crash kernel.
2703 */
2704 if (!max_addresses ||
2705 ipv6_count_addresses(in6_dev) < max_addresses)
2706 ifp = ipv6_add_addr(in6_dev, &cfg, false, NULL);
2707
2708 if (IS_ERR_OR_NULL(ifp))
2709 return -1;
2710
2711 create = 1;
2712 spin_lock_bh(&ifp->lock);
2713 ifp->flags |= IFA_F_MANAGETEMPADDR;
2714 ifp->cstamp = jiffies;
2715 ifp->tokenized = tokenized;
2716 spin_unlock_bh(&ifp->lock);
2717 addrconf_dad_start(ifp);
2718 }
2719
2720 if (ifp) {
2721 u32 flags;
2722 unsigned long now;
2723 u32 stored_lft;
2724
2725 /* update lifetime (RFC2462 5.5.3 e) */
2726 spin_lock_bh(&ifp->lock);
2727 now = jiffies;
2728 if (ifp->valid_lft > (now - ifp->tstamp) / HZ)
2729 stored_lft = ifp->valid_lft - (now - ifp->tstamp) / HZ;
2730 else
2731 stored_lft = 0;
2732
2733 /* RFC4862 Section 5.5.3e:
2734 * "Note that the preferred lifetime of the
2735 * corresponding address is always reset to
2736 * the Preferred Lifetime in the received
2737 * Prefix Information option, regardless of
2738 * whether the valid lifetime is also reset or
2739 * ignored."
2740 *
2741 * So we should always update prefered_lft here.
2742 */
2743 update_lft = !create && stored_lft;
2744
2745 if (update_lft && !READ_ONCE(in6_dev->cnf.ra_honor_pio_life)) {
2746 const u32 minimum_lft = min_t(u32,
2747 stored_lft, MIN_VALID_LIFETIME);
2748 valid_lft = max(valid_lft, minimum_lft);
2749 }
2750
2751 if (update_lft) {
2752 ifp->valid_lft = valid_lft;
2753 ifp->prefered_lft = prefered_lft;
2754 WRITE_ONCE(ifp->tstamp, now);
2755 flags = ifp->flags;
2756 ifp->flags &= ~IFA_F_DEPRECATED;
2757 spin_unlock_bh(&ifp->lock);
2758
2759 if (!(flags&IFA_F_TENTATIVE))
2760 ipv6_ifa_notify(0, ifp);
2761 } else
2762 spin_unlock_bh(&ifp->lock);
2763
2764 manage_tempaddrs(in6_dev, ifp, valid_lft, prefered_lft,
2765 create, now);
2766
2767 in6_ifa_put(ifp);
2768 addrconf_verify(net);
2769 }
2770
2771 return 0;
2772 }
2773 EXPORT_SYMBOL_GPL(addrconf_prefix_rcv_add_addr);
2774
addrconf_prefix_rcv(struct net_device * dev,u8 * opt,int len,bool sllao)2775 void addrconf_prefix_rcv(struct net_device *dev, u8 *opt, int len, bool sllao)
2776 {
2777 struct prefix_info *pinfo;
2778 struct fib6_table *table;
2779 __u32 valid_lft;
2780 __u32 prefered_lft;
2781 int addr_type, err;
2782 u32 addr_flags = 0;
2783 struct inet6_dev *in6_dev;
2784 struct net *net = dev_net(dev);
2785 bool ignore_autoconf = false;
2786
2787 pinfo = (struct prefix_info *) opt;
2788
2789 if (len < sizeof(struct prefix_info)) {
2790 netdev_dbg(dev, "addrconf: prefix option too short\n");
2791 return;
2792 }
2793
2794 /*
2795 * Validation checks ([ADDRCONF], page 19)
2796 */
2797
2798 addr_type = ipv6_addr_type(&pinfo->prefix);
2799
2800 if (addr_type & (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL))
2801 return;
2802
2803 valid_lft = ntohl(pinfo->valid);
2804 prefered_lft = ntohl(pinfo->prefered);
2805
2806 if (prefered_lft > valid_lft) {
2807 net_warn_ratelimited("addrconf: prefix option has invalid lifetime\n");
2808 return;
2809 }
2810
2811 in6_dev = in6_dev_get(dev);
2812
2813 if (!in6_dev) {
2814 net_dbg_ratelimited("addrconf: device %s not configured\n",
2815 dev->name);
2816 return;
2817 }
2818
2819 if (valid_lft != 0 && valid_lft < in6_dev->cnf.accept_ra_min_lft)
2820 goto put;
2821
2822 /*
2823 * Two things going on here:
2824 * 1) Add routes for on-link prefixes
2825 * 2) Configure prefixes with the auto flag set
2826 */
2827
2828 if (pinfo->onlink) {
2829 struct fib6_info *rt;
2830 unsigned long rt_expires;
2831
2832 /* Avoid arithmetic overflow. Really, we could
2833 * save rt_expires in seconds, likely valid_lft,
2834 * but it would require division in fib gc, that it
2835 * not good.
2836 */
2837 if (HZ > USER_HZ)
2838 rt_expires = addrconf_timeout_fixup(valid_lft, HZ);
2839 else
2840 rt_expires = addrconf_timeout_fixup(valid_lft, USER_HZ);
2841
2842 if (addrconf_finite_timeout(rt_expires))
2843 rt_expires *= HZ;
2844
2845 rt = addrconf_get_prefix_route(&pinfo->prefix,
2846 pinfo->prefix_len,
2847 dev,
2848 RTF_ADDRCONF | RTF_PREFIX_RT,
2849 RTF_DEFAULT, true);
2850
2851 if (rt) {
2852 /* Autoconf prefix route */
2853 if (valid_lft == 0) {
2854 ip6_del_rt(net, rt, false);
2855 rt = NULL;
2856 } else {
2857 table = rt->fib6_table;
2858 spin_lock_bh(&table->tb6_lock);
2859
2860 if (addrconf_finite_timeout(rt_expires)) {
2861 /* not infinity */
2862 fib6_set_expires(rt, jiffies + rt_expires);
2863 fib6_add_gc_list(rt);
2864 } else {
2865 fib6_clean_expires(rt);
2866 fib6_remove_gc_list(rt);
2867 }
2868
2869 spin_unlock_bh(&table->tb6_lock);
2870 }
2871 } else if (valid_lft) {
2872 clock_t expires = 0;
2873 int flags = RTF_ADDRCONF | RTF_PREFIX_RT;
2874 if (addrconf_finite_timeout(rt_expires)) {
2875 /* not infinity */
2876 flags |= RTF_EXPIRES;
2877 expires = jiffies_to_clock_t(rt_expires);
2878 }
2879 addrconf_prefix_route(&pinfo->prefix, pinfo->prefix_len,
2880 0, dev, expires, flags,
2881 GFP_ATOMIC);
2882 }
2883 fib6_info_release(rt);
2884 }
2885
2886 /* Try to figure out our local address for this prefix */
2887
2888 ignore_autoconf = READ_ONCE(in6_dev->cnf.ra_honor_pio_pflag) && pinfo->preferpd;
2889 if (pinfo->autoconf && in6_dev->cnf.autoconf && !ignore_autoconf) {
2890 struct in6_addr addr;
2891 bool tokenized = false, dev_addr_generated = false;
2892
2893 if (pinfo->prefix_len == 64) {
2894 memcpy(&addr, &pinfo->prefix, 8);
2895
2896 if (!ipv6_addr_any(&in6_dev->token)) {
2897 read_lock_bh(&in6_dev->lock);
2898 memcpy(addr.s6_addr + 8,
2899 in6_dev->token.s6_addr + 8, 8);
2900 read_unlock_bh(&in6_dev->lock);
2901 tokenized = true;
2902 } else if (is_addr_mode_generate_stable(in6_dev) &&
2903 !ipv6_generate_stable_address(&addr, 0,
2904 in6_dev)) {
2905 addr_flags |= IFA_F_STABLE_PRIVACY;
2906 goto ok;
2907 } else if (ipv6_generate_eui64(addr.s6_addr + 8, dev) &&
2908 ipv6_inherit_eui64(addr.s6_addr + 8, in6_dev)) {
2909 goto put;
2910 } else {
2911 dev_addr_generated = true;
2912 }
2913 goto ok;
2914 }
2915 net_dbg_ratelimited("IPv6 addrconf: prefix with wrong length %d\n",
2916 pinfo->prefix_len);
2917 goto put;
2918
2919 ok:
2920 err = addrconf_prefix_rcv_add_addr(net, dev, pinfo, in6_dev,
2921 &addr, addr_type,
2922 addr_flags, sllao,
2923 tokenized, valid_lft,
2924 prefered_lft);
2925 if (err)
2926 goto put;
2927
2928 /* Ignore error case here because previous prefix add addr was
2929 * successful which will be notified.
2930 */
2931 ndisc_ops_prefix_rcv_add_addr(net, dev, pinfo, in6_dev, &addr,
2932 addr_type, addr_flags, sllao,
2933 tokenized, valid_lft,
2934 prefered_lft,
2935 dev_addr_generated);
2936 }
2937 inet6_prefix_notify(RTM_NEWPREFIX, in6_dev, pinfo);
2938 put:
2939 in6_dev_put(in6_dev);
2940 }
2941
addrconf_set_sit_dstaddr(struct net * net,struct net_device * dev,struct in6_ifreq * ireq)2942 static int addrconf_set_sit_dstaddr(struct net *net, struct net_device *dev,
2943 struct in6_ifreq *ireq)
2944 {
2945 struct ip_tunnel_parm_kern p = { };
2946 int err;
2947
2948 if (!(ipv6_addr_type(&ireq->ifr6_addr) & IPV6_ADDR_COMPATv4))
2949 return -EADDRNOTAVAIL;
2950
2951 p.iph.daddr = ireq->ifr6_addr.s6_addr32[3];
2952 p.iph.version = 4;
2953 p.iph.ihl = 5;
2954 p.iph.protocol = IPPROTO_IPV6;
2955 p.iph.ttl = 64;
2956
2957 if (!dev->netdev_ops->ndo_tunnel_ctl)
2958 return -EOPNOTSUPP;
2959 err = dev->netdev_ops->ndo_tunnel_ctl(dev, &p, SIOCADDTUNNEL);
2960 if (err)
2961 return err;
2962
2963 dev = __dev_get_by_name(net, p.name);
2964 if (!dev)
2965 return -ENOBUFS;
2966 return dev_open(dev, NULL);
2967 }
2968
2969 /*
2970 * Set destination address.
2971 * Special case for SIT interfaces where we create a new "virtual"
2972 * device.
2973 */
addrconf_set_dstaddr(struct net * net,void __user * arg)2974 int addrconf_set_dstaddr(struct net *net, void __user *arg)
2975 {
2976 struct net_device *dev;
2977 struct in6_ifreq ireq;
2978 int err = -ENODEV;
2979
2980 if (!IS_ENABLED(CONFIG_IPV6_SIT))
2981 return -ENODEV;
2982 if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2983 return -EFAULT;
2984
2985 rtnl_net_lock(net);
2986 dev = __dev_get_by_index(net, ireq.ifr6_ifindex);
2987 if (dev && dev->type == ARPHRD_SIT)
2988 err = addrconf_set_sit_dstaddr(net, dev, &ireq);
2989 rtnl_net_unlock(net);
2990 return err;
2991 }
2992
ipv6_mc_config(struct sock * sk,bool join,const struct in6_addr * addr,int ifindex)2993 static int ipv6_mc_config(struct sock *sk, bool join,
2994 const struct in6_addr *addr, int ifindex)
2995 {
2996 int ret;
2997
2998 ASSERT_RTNL();
2999
3000 lock_sock(sk);
3001 if (join)
3002 ret = ipv6_sock_mc_join(sk, ifindex, addr);
3003 else
3004 ret = ipv6_sock_mc_drop(sk, ifindex, addr);
3005 release_sock(sk);
3006
3007 return ret;
3008 }
3009
3010 /*
3011 * Manual configuration of address on an interface
3012 */
inet6_addr_add(struct net * net,struct net_device * dev,struct ifa6_config * cfg,clock_t expires,u32 flags,struct netlink_ext_ack * extack)3013 static int inet6_addr_add(struct net *net, struct net_device *dev,
3014 struct ifa6_config *cfg, clock_t expires, u32 flags,
3015 struct netlink_ext_ack *extack)
3016 {
3017 struct inet6_ifaddr *ifp;
3018 struct inet6_dev *idev;
3019
3020 ASSERT_RTNL_NET(net);
3021
3022 if (cfg->plen > 128) {
3023 NL_SET_ERR_MSG_MOD(extack, "Invalid prefix length");
3024 return -EINVAL;
3025 }
3026
3027 if (cfg->ifa_flags & IFA_F_MANAGETEMPADDR && cfg->plen != 64) {
3028 NL_SET_ERR_MSG_MOD(extack, "address with \"mngtmpaddr\" flag must have a prefix length of 64");
3029 return -EINVAL;
3030 }
3031
3032 idev = addrconf_add_dev(dev);
3033 if (IS_ERR(idev)) {
3034 NL_SET_ERR_MSG_MOD(extack, "IPv6 is disabled on this device");
3035 return PTR_ERR(idev);
3036 }
3037
3038 if (cfg->ifa_flags & IFA_F_MCAUTOJOIN) {
3039 int ret = ipv6_mc_config(net->ipv6.mc_autojoin_sk,
3040 true, cfg->pfx, dev->ifindex);
3041
3042 if (ret < 0) {
3043 NL_SET_ERR_MSG_MOD(extack, "Multicast auto join failed");
3044 return ret;
3045 }
3046 }
3047
3048 cfg->scope = ipv6_addr_scope(cfg->pfx);
3049
3050 ifp = ipv6_add_addr(idev, cfg, true, extack);
3051 if (!IS_ERR(ifp)) {
3052 if (!(cfg->ifa_flags & IFA_F_NOPREFIXROUTE)) {
3053 addrconf_prefix_route(&ifp->addr, ifp->prefix_len,
3054 ifp->rt_priority, dev, expires,
3055 flags, GFP_KERNEL);
3056 }
3057
3058 /* Send a netlink notification if DAD is enabled and
3059 * optimistic flag is not set
3060 */
3061 if (!(ifp->flags & (IFA_F_OPTIMISTIC | IFA_F_NODAD)))
3062 ipv6_ifa_notify(0, ifp);
3063 /*
3064 * Note that section 3.1 of RFC 4429 indicates
3065 * that the Optimistic flag should not be set for
3066 * manually configured addresses
3067 */
3068 addrconf_dad_start(ifp);
3069 if (cfg->ifa_flags & IFA_F_MANAGETEMPADDR)
3070 manage_tempaddrs(idev, ifp, cfg->valid_lft,
3071 cfg->preferred_lft, true, jiffies);
3072 in6_ifa_put(ifp);
3073 addrconf_verify_rtnl(net);
3074 return 0;
3075 } else if (cfg->ifa_flags & IFA_F_MCAUTOJOIN) {
3076 ipv6_mc_config(net->ipv6.mc_autojoin_sk, false,
3077 cfg->pfx, dev->ifindex);
3078 }
3079
3080 return PTR_ERR(ifp);
3081 }
3082
inet6_addr_del(struct net * net,int ifindex,u32 ifa_flags,const struct in6_addr * pfx,unsigned int plen,struct netlink_ext_ack * extack)3083 static int inet6_addr_del(struct net *net, int ifindex, u32 ifa_flags,
3084 const struct in6_addr *pfx, unsigned int plen,
3085 struct netlink_ext_ack *extack)
3086 {
3087 struct inet6_ifaddr *ifp;
3088 struct inet6_dev *idev;
3089 struct net_device *dev;
3090
3091 if (plen > 128) {
3092 NL_SET_ERR_MSG_MOD(extack, "Invalid prefix length");
3093 return -EINVAL;
3094 }
3095
3096 dev = __dev_get_by_index(net, ifindex);
3097 if (!dev) {
3098 NL_SET_ERR_MSG_MOD(extack, "Unable to find the interface");
3099 return -ENODEV;
3100 }
3101
3102 idev = __in6_dev_get_rtnl_net(dev);
3103 if (!idev) {
3104 NL_SET_ERR_MSG_MOD(extack, "IPv6 is disabled on this device");
3105 return -ENXIO;
3106 }
3107
3108 read_lock_bh(&idev->lock);
3109 list_for_each_entry(ifp, &idev->addr_list, if_list) {
3110 if (ifp->prefix_len == plen &&
3111 ipv6_addr_equal(pfx, &ifp->addr)) {
3112 in6_ifa_hold(ifp);
3113 read_unlock_bh(&idev->lock);
3114
3115 ipv6_del_addr(ifp);
3116
3117 if (!(ifp->flags & IFA_F_TEMPORARY) &&
3118 (ifp->flags & IFA_F_MANAGETEMPADDR))
3119 delete_tempaddrs(idev, ifp);
3120
3121 addrconf_verify_rtnl(net);
3122 if (ipv6_addr_is_multicast(pfx)) {
3123 ipv6_mc_config(net->ipv6.mc_autojoin_sk,
3124 false, pfx, dev->ifindex);
3125 }
3126 return 0;
3127 }
3128 }
3129 read_unlock_bh(&idev->lock);
3130
3131 NL_SET_ERR_MSG_MOD(extack, "address not found");
3132 return -EADDRNOTAVAIL;
3133 }
3134
3135
addrconf_add_ifaddr(struct net * net,void __user * arg)3136 int addrconf_add_ifaddr(struct net *net, void __user *arg)
3137 {
3138 struct ifa6_config cfg = {
3139 .ifa_flags = IFA_F_PERMANENT,
3140 .preferred_lft = INFINITY_LIFE_TIME,
3141 .valid_lft = INFINITY_LIFE_TIME,
3142 };
3143 struct net_device *dev;
3144 struct in6_ifreq ireq;
3145 int err;
3146
3147 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
3148 return -EPERM;
3149
3150 if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
3151 return -EFAULT;
3152
3153 cfg.pfx = &ireq.ifr6_addr;
3154 cfg.plen = ireq.ifr6_prefixlen;
3155
3156 rtnl_net_lock(net);
3157 dev = __dev_get_by_index(net, ireq.ifr6_ifindex);
3158 if (dev) {
3159 netdev_lock_ops(dev);
3160 err = inet6_addr_add(net, dev, &cfg, 0, 0, NULL);
3161 netdev_unlock_ops(dev);
3162 } else {
3163 err = -ENODEV;
3164 }
3165 rtnl_net_unlock(net);
3166 return err;
3167 }
3168
addrconf_del_ifaddr(struct net * net,void __user * arg)3169 int addrconf_del_ifaddr(struct net *net, void __user *arg)
3170 {
3171 struct in6_ifreq ireq;
3172 int err;
3173
3174 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
3175 return -EPERM;
3176
3177 if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
3178 return -EFAULT;
3179
3180 rtnl_net_lock(net);
3181 err = inet6_addr_del(net, ireq.ifr6_ifindex, 0, &ireq.ifr6_addr,
3182 ireq.ifr6_prefixlen, NULL);
3183 rtnl_net_unlock(net);
3184 return err;
3185 }
3186
add_addr(struct inet6_dev * idev,const struct in6_addr * addr,int plen,int scope,u8 proto)3187 static void add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
3188 int plen, int scope, u8 proto)
3189 {
3190 struct inet6_ifaddr *ifp;
3191 struct ifa6_config cfg = {
3192 .pfx = addr,
3193 .plen = plen,
3194 .ifa_flags = IFA_F_PERMANENT,
3195 .valid_lft = INFINITY_LIFE_TIME,
3196 .preferred_lft = INFINITY_LIFE_TIME,
3197 .scope = scope,
3198 .ifa_proto = proto
3199 };
3200
3201 ifp = ipv6_add_addr(idev, &cfg, true, NULL);
3202 if (!IS_ERR(ifp)) {
3203 spin_lock_bh(&ifp->lock);
3204 ifp->flags &= ~IFA_F_TENTATIVE;
3205 spin_unlock_bh(&ifp->lock);
3206 rt_genid_bump_ipv6(dev_net(idev->dev));
3207 ipv6_ifa_notify(RTM_NEWADDR, ifp);
3208 in6_ifa_put(ifp);
3209 }
3210 }
3211
3212 #if IS_ENABLED(CONFIG_IPV6_SIT) || IS_ENABLED(CONFIG_NET_IPGRE)
add_v4_addrs(struct inet6_dev * idev)3213 static void add_v4_addrs(struct inet6_dev *idev)
3214 {
3215 struct in6_addr addr;
3216 struct net_device *dev;
3217 struct net *net = dev_net(idev->dev);
3218 int scope, plen;
3219 u32 pflags = 0;
3220
3221 ASSERT_RTNL();
3222
3223 memset(&addr, 0, sizeof(struct in6_addr));
3224 memcpy(&addr.s6_addr32[3], idev->dev->dev_addr, 4);
3225
3226 if (!(idev->dev->flags & IFF_POINTOPOINT) && idev->dev->type == ARPHRD_SIT) {
3227 scope = IPV6_ADDR_COMPATv4;
3228 plen = 96;
3229 pflags |= RTF_NONEXTHOP;
3230 } else {
3231 if (idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_NONE)
3232 return;
3233
3234 addr.s6_addr32[0] = htonl(0xfe800000);
3235 scope = IFA_LINK;
3236 plen = 64;
3237 }
3238
3239 if (addr.s6_addr32[3]) {
3240 add_addr(idev, &addr, plen, scope, IFAPROT_UNSPEC);
3241 addrconf_prefix_route(&addr, plen, 0, idev->dev, 0, pflags,
3242 GFP_KERNEL);
3243 return;
3244 }
3245
3246 for_each_netdev(net, dev) {
3247 struct in_device *in_dev = __in_dev_get_rtnl(dev);
3248 if (in_dev && (dev->flags & IFF_UP)) {
3249 struct in_ifaddr *ifa;
3250 int flag = scope;
3251
3252 in_dev_for_each_ifa_rtnl(ifa, in_dev) {
3253 addr.s6_addr32[3] = ifa->ifa_local;
3254
3255 if (ifa->ifa_scope == RT_SCOPE_LINK)
3256 continue;
3257 if (ifa->ifa_scope >= RT_SCOPE_HOST) {
3258 if (idev->dev->flags&IFF_POINTOPOINT)
3259 continue;
3260 flag |= IFA_HOST;
3261 }
3262
3263 add_addr(idev, &addr, plen, flag,
3264 IFAPROT_UNSPEC);
3265 addrconf_prefix_route(&addr, plen, 0, idev->dev,
3266 0, pflags, GFP_KERNEL);
3267 }
3268 }
3269 }
3270 }
3271 #endif
3272
init_loopback(struct net_device * dev)3273 static void init_loopback(struct net_device *dev)
3274 {
3275 struct inet6_dev *idev;
3276
3277 /* ::1 */
3278
3279 ASSERT_RTNL();
3280
3281 idev = ipv6_find_idev(dev);
3282 if (IS_ERR(idev)) {
3283 pr_debug("%s: add_dev failed\n", __func__);
3284 return;
3285 }
3286
3287 add_addr(idev, &in6addr_loopback, 128, IFA_HOST, IFAPROT_KERNEL_LO);
3288 }
3289
addrconf_add_linklocal(struct inet6_dev * idev,const struct in6_addr * addr,u32 flags)3290 void addrconf_add_linklocal(struct inet6_dev *idev,
3291 const struct in6_addr *addr, u32 flags)
3292 {
3293 struct ifa6_config cfg = {
3294 .pfx = addr,
3295 .plen = 64,
3296 .ifa_flags = flags | IFA_F_PERMANENT,
3297 .valid_lft = INFINITY_LIFE_TIME,
3298 .preferred_lft = INFINITY_LIFE_TIME,
3299 .scope = IFA_LINK,
3300 .ifa_proto = IFAPROT_KERNEL_LL
3301 };
3302 struct inet6_ifaddr *ifp;
3303
3304 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
3305 if ((READ_ONCE(dev_net(idev->dev)->ipv6.devconf_all->optimistic_dad) ||
3306 READ_ONCE(idev->cnf.optimistic_dad)) &&
3307 !dev_net(idev->dev)->ipv6.devconf_all->forwarding)
3308 cfg.ifa_flags |= IFA_F_OPTIMISTIC;
3309 #endif
3310
3311 ifp = ipv6_add_addr(idev, &cfg, true, NULL);
3312 if (!IS_ERR(ifp)) {
3313 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, 0, idev->dev,
3314 0, 0, GFP_ATOMIC);
3315 addrconf_dad_start(ifp);
3316 in6_ifa_put(ifp);
3317 }
3318 }
3319 EXPORT_SYMBOL_GPL(addrconf_add_linklocal);
3320
ipv6_reserved_interfaceid(struct in6_addr address)3321 static bool ipv6_reserved_interfaceid(struct in6_addr address)
3322 {
3323 if ((address.s6_addr32[2] | address.s6_addr32[3]) == 0)
3324 return true;
3325
3326 if (address.s6_addr32[2] == htonl(0x02005eff) &&
3327 ((address.s6_addr32[3] & htonl(0xfe000000)) == htonl(0xfe000000)))
3328 return true;
3329
3330 if (address.s6_addr32[2] == htonl(0xfdffffff) &&
3331 ((address.s6_addr32[3] & htonl(0xffffff80)) == htonl(0xffffff80)))
3332 return true;
3333
3334 return false;
3335 }
3336
ipv6_generate_stable_address(struct in6_addr * address,u8 dad_count,const struct inet6_dev * idev)3337 static int ipv6_generate_stable_address(struct in6_addr *address,
3338 u8 dad_count,
3339 const struct inet6_dev *idev)
3340 {
3341 static DEFINE_SPINLOCK(lock);
3342 static __u32 digest[SHA1_DIGEST_WORDS];
3343 static __u32 workspace[SHA1_WORKSPACE_WORDS];
3344
3345 static union {
3346 char __data[SHA1_BLOCK_SIZE];
3347 struct {
3348 struct in6_addr secret;
3349 __be32 prefix[2];
3350 unsigned char hwaddr[MAX_ADDR_LEN];
3351 u8 dad_count;
3352 } __packed;
3353 } data;
3354
3355 struct in6_addr secret;
3356 struct in6_addr temp;
3357 struct net *net = dev_net(idev->dev);
3358
3359 BUILD_BUG_ON(sizeof(data.__data) != sizeof(data));
3360
3361 if (idev->cnf.stable_secret.initialized)
3362 secret = idev->cnf.stable_secret.secret;
3363 else if (net->ipv6.devconf_dflt->stable_secret.initialized)
3364 secret = net->ipv6.devconf_dflt->stable_secret.secret;
3365 else
3366 return -1;
3367
3368 retry:
3369 spin_lock_bh(&lock);
3370
3371 sha1_init_raw(digest);
3372 memset(&data, 0, sizeof(data));
3373 memset(workspace, 0, sizeof(workspace));
3374 memcpy(data.hwaddr, idev->dev->perm_addr, idev->dev->addr_len);
3375 data.prefix[0] = address->s6_addr32[0];
3376 data.prefix[1] = address->s6_addr32[1];
3377 data.secret = secret;
3378 data.dad_count = dad_count;
3379
3380 sha1_transform(digest, data.__data, workspace);
3381
3382 temp = *address;
3383 temp.s6_addr32[2] = (__force __be32)digest[0];
3384 temp.s6_addr32[3] = (__force __be32)digest[1];
3385
3386 spin_unlock_bh(&lock);
3387
3388 if (ipv6_reserved_interfaceid(temp)) {
3389 dad_count++;
3390 if (dad_count > dev_net(idev->dev)->ipv6.sysctl.idgen_retries)
3391 return -1;
3392 goto retry;
3393 }
3394
3395 *address = temp;
3396 return 0;
3397 }
3398
ipv6_gen_mode_random_init(struct inet6_dev * idev)3399 static void ipv6_gen_mode_random_init(struct inet6_dev *idev)
3400 {
3401 struct ipv6_stable_secret *s = &idev->cnf.stable_secret;
3402
3403 if (s->initialized)
3404 return;
3405 s = &idev->cnf.stable_secret;
3406 get_random_bytes(&s->secret, sizeof(s->secret));
3407 s->initialized = true;
3408 }
3409
addrconf_addr_gen(struct inet6_dev * idev,bool prefix_route)3410 static void addrconf_addr_gen(struct inet6_dev *idev, bool prefix_route)
3411 {
3412 struct in6_addr addr;
3413
3414 /* no link local addresses on L3 master devices */
3415 if (netif_is_l3_master(idev->dev))
3416 return;
3417
3418 /* no link local addresses on devices flagged as slaves */
3419 if (idev->dev->priv_flags & IFF_NO_ADDRCONF)
3420 return;
3421
3422 ipv6_addr_set(&addr, htonl(0xFE800000), 0, 0, 0);
3423
3424 switch (idev->cnf.addr_gen_mode) {
3425 case IN6_ADDR_GEN_MODE_RANDOM:
3426 ipv6_gen_mode_random_init(idev);
3427 fallthrough;
3428 case IN6_ADDR_GEN_MODE_STABLE_PRIVACY:
3429 if (!ipv6_generate_stable_address(&addr, 0, idev))
3430 addrconf_add_linklocal(idev, &addr,
3431 IFA_F_STABLE_PRIVACY);
3432 else if (prefix_route)
3433 addrconf_prefix_route(&addr, 64, 0, idev->dev,
3434 0, 0, GFP_KERNEL);
3435 break;
3436 case IN6_ADDR_GEN_MODE_EUI64:
3437 /* addrconf_add_linklocal also adds a prefix_route and we
3438 * only need to care about prefix routes if ipv6_generate_eui64
3439 * couldn't generate one.
3440 */
3441 if (ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) == 0)
3442 addrconf_add_linklocal(idev, &addr, 0);
3443 else if (prefix_route)
3444 addrconf_prefix_route(&addr, 64, 0, idev->dev,
3445 0, 0, GFP_KERNEL);
3446 break;
3447 case IN6_ADDR_GEN_MODE_NONE:
3448 default:
3449 /* will not add any link local address */
3450 break;
3451 }
3452 }
3453
addrconf_dev_config(struct net_device * dev)3454 static void addrconf_dev_config(struct net_device *dev)
3455 {
3456 struct inet6_dev *idev;
3457
3458 ASSERT_RTNL();
3459
3460 if ((dev->type != ARPHRD_ETHER) &&
3461 (dev->type != ARPHRD_FDDI) &&
3462 (dev->type != ARPHRD_ARCNET) &&
3463 (dev->type != ARPHRD_INFINIBAND) &&
3464 (dev->type != ARPHRD_IEEE1394) &&
3465 (dev->type != ARPHRD_TUNNEL6) &&
3466 (dev->type != ARPHRD_6LOWPAN) &&
3467 (dev->type != ARPHRD_IP6GRE) &&
3468 (dev->type != ARPHRD_TUNNEL) &&
3469 (dev->type != ARPHRD_NONE) &&
3470 (dev->type != ARPHRD_RAWIP)) {
3471 /* Alas, we support only Ethernet autoconfiguration. */
3472 idev = __in6_dev_get(dev);
3473 if (!IS_ERR_OR_NULL(idev) && dev->flags & IFF_UP &&
3474 dev->flags & IFF_MULTICAST)
3475 ipv6_mc_up(idev);
3476 return;
3477 }
3478
3479 idev = addrconf_add_dev(dev);
3480 if (IS_ERR(idev))
3481 return;
3482
3483 /* this device type has no EUI support */
3484 if (dev->type == ARPHRD_NONE &&
3485 idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_EUI64)
3486 WRITE_ONCE(idev->cnf.addr_gen_mode,
3487 IN6_ADDR_GEN_MODE_RANDOM);
3488
3489 addrconf_addr_gen(idev, false);
3490 }
3491
3492 #if IS_ENABLED(CONFIG_IPV6_SIT)
addrconf_sit_config(struct net_device * dev)3493 static void addrconf_sit_config(struct net_device *dev)
3494 {
3495 struct inet6_dev *idev;
3496
3497 ASSERT_RTNL();
3498
3499 /*
3500 * Configure the tunnel with one of our IPv4
3501 * addresses... we should configure all of
3502 * our v4 addrs in the tunnel
3503 */
3504
3505 idev = ipv6_find_idev(dev);
3506 if (IS_ERR(idev)) {
3507 pr_debug("%s: add_dev failed\n", __func__);
3508 return;
3509 }
3510
3511 if (dev->priv_flags & IFF_ISATAP) {
3512 addrconf_addr_gen(idev, false);
3513 return;
3514 }
3515
3516 add_v4_addrs(idev);
3517
3518 if (dev->flags&IFF_POINTOPOINT)
3519 addrconf_add_mroute(dev);
3520 }
3521 #endif
3522
3523 #if IS_ENABLED(CONFIG_NET_IPGRE)
addrconf_gre_config(struct net_device * dev)3524 static void addrconf_gre_config(struct net_device *dev)
3525 {
3526 struct inet6_dev *idev;
3527
3528 ASSERT_RTNL();
3529
3530 idev = addrconf_add_dev(dev);
3531 if (IS_ERR(idev))
3532 return;
3533
3534 /* Generate the IPv6 link-local address using addrconf_addr_gen(),
3535 * unless we have an IPv4 GRE device not bound to an IP address and
3536 * which is in EUI64 mode (as __ipv6_isatap_ifid() would fail in this
3537 * case). Such devices fall back to add_v4_addrs() instead.
3538 */
3539 if (!(*(__be32 *)dev->dev_addr == 0 &&
3540 idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_EUI64)) {
3541 addrconf_addr_gen(idev, true);
3542 return;
3543 }
3544
3545 add_v4_addrs(idev);
3546 }
3547 #endif
3548
addrconf_init_auto_addrs(struct net_device * dev)3549 static void addrconf_init_auto_addrs(struct net_device *dev)
3550 {
3551 switch (dev->type) {
3552 #if IS_ENABLED(CONFIG_IPV6_SIT)
3553 case ARPHRD_SIT:
3554 addrconf_sit_config(dev);
3555 break;
3556 #endif
3557 #if IS_ENABLED(CONFIG_NET_IPGRE)
3558 case ARPHRD_IPGRE:
3559 addrconf_gre_config(dev);
3560 break;
3561 #endif
3562 case ARPHRD_LOOPBACK:
3563 init_loopback(dev);
3564 break;
3565
3566 default:
3567 addrconf_dev_config(dev);
3568 break;
3569 }
3570 }
3571
fixup_permanent_addr(struct net * net,struct inet6_dev * idev,struct inet6_ifaddr * ifp)3572 static int fixup_permanent_addr(struct net *net,
3573 struct inet6_dev *idev,
3574 struct inet6_ifaddr *ifp)
3575 {
3576 /* !fib6_node means the host route was removed from the
3577 * FIB, for example, if 'lo' device is taken down. In that
3578 * case regenerate the host route.
3579 */
3580 if (!ifp->rt || !ifp->rt->fib6_node) {
3581 struct fib6_info *f6i, *prev;
3582
3583 f6i = addrconf_f6i_alloc(net, idev, &ifp->addr, false,
3584 GFP_ATOMIC, NULL);
3585 if (IS_ERR(f6i))
3586 return PTR_ERR(f6i);
3587
3588 /* ifp->rt can be accessed outside of rtnl */
3589 spin_lock(&ifp->lock);
3590 prev = ifp->rt;
3591 ifp->rt = f6i;
3592 spin_unlock(&ifp->lock);
3593
3594 fib6_info_release(prev);
3595 }
3596
3597 if (!(ifp->flags & IFA_F_NOPREFIXROUTE)) {
3598 addrconf_prefix_route(&ifp->addr, ifp->prefix_len,
3599 ifp->rt_priority, idev->dev, 0, 0,
3600 GFP_ATOMIC);
3601 }
3602
3603 if (ifp->state == INET6_IFADDR_STATE_PREDAD)
3604 addrconf_dad_start(ifp);
3605
3606 return 0;
3607 }
3608
addrconf_permanent_addr(struct net * net,struct net_device * dev)3609 static void addrconf_permanent_addr(struct net *net, struct net_device *dev)
3610 {
3611 struct inet6_ifaddr *ifp, *tmp;
3612 struct inet6_dev *idev;
3613
3614 idev = __in6_dev_get(dev);
3615 if (!idev)
3616 return;
3617
3618 write_lock_bh(&idev->lock);
3619
3620 list_for_each_entry_safe(ifp, tmp, &idev->addr_list, if_list) {
3621 if ((ifp->flags & IFA_F_PERMANENT) &&
3622 fixup_permanent_addr(net, idev, ifp) < 0) {
3623 write_unlock_bh(&idev->lock);
3624 in6_ifa_hold(ifp);
3625 ipv6_del_addr(ifp);
3626 write_lock_bh(&idev->lock);
3627
3628 net_info_ratelimited("%s: Failed to add prefix route for address %pI6c; dropping\n",
3629 idev->dev->name, &ifp->addr);
3630 }
3631 }
3632
3633 write_unlock_bh(&idev->lock);
3634 }
3635
addrconf_notify(struct notifier_block * this,unsigned long event,void * ptr)3636 static int addrconf_notify(struct notifier_block *this, unsigned long event,
3637 void *ptr)
3638 {
3639 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
3640 struct netdev_notifier_change_info *change_info;
3641 struct netdev_notifier_changeupper_info *info;
3642 struct inet6_dev *idev = __in6_dev_get(dev);
3643 struct net *net = dev_net(dev);
3644 int run_pending = 0;
3645 int err;
3646
3647 switch (event) {
3648 case NETDEV_REGISTER:
3649 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
3650 idev = ipv6_add_dev(dev);
3651 if (IS_ERR(idev))
3652 return notifier_from_errno(PTR_ERR(idev));
3653 }
3654 break;
3655
3656 case NETDEV_CHANGEMTU:
3657 /* if MTU under IPV6_MIN_MTU stop IPv6 on this interface. */
3658 if (dev->mtu < IPV6_MIN_MTU) {
3659 addrconf_ifdown(dev, dev != net->loopback_dev);
3660 break;
3661 }
3662
3663 if (idev) {
3664 rt6_mtu_change(dev, dev->mtu);
3665 WRITE_ONCE(idev->cnf.mtu6, dev->mtu);
3666 break;
3667 }
3668
3669 /* allocate new idev */
3670 idev = ipv6_add_dev(dev);
3671 if (IS_ERR(idev))
3672 break;
3673
3674 /* device is still not ready */
3675 if (!(idev->if_flags & IF_READY))
3676 break;
3677
3678 run_pending = 1;
3679 fallthrough;
3680 case NETDEV_UP:
3681 case NETDEV_CHANGE:
3682 if (idev && idev->cnf.disable_ipv6)
3683 break;
3684
3685 if (dev->priv_flags & IFF_NO_ADDRCONF) {
3686 if (event == NETDEV_UP && !IS_ERR_OR_NULL(idev) &&
3687 dev->flags & IFF_UP && dev->flags & IFF_MULTICAST)
3688 ipv6_mc_up(idev);
3689 break;
3690 }
3691
3692 if (event == NETDEV_UP) {
3693 /* restore routes for permanent addresses */
3694 addrconf_permanent_addr(net, dev);
3695
3696 if (!addrconf_link_ready(dev)) {
3697 /* device is not ready yet. */
3698 pr_debug("ADDRCONF(NETDEV_UP): %s: link is not ready\n",
3699 dev->name);
3700 break;
3701 }
3702
3703 if (!idev && dev->mtu >= IPV6_MIN_MTU)
3704 idev = ipv6_add_dev(dev);
3705
3706 if (!IS_ERR_OR_NULL(idev)) {
3707 idev->if_flags |= IF_READY;
3708 run_pending = 1;
3709 }
3710 } else if (event == NETDEV_CHANGE) {
3711 if (!addrconf_link_ready(dev)) {
3712 /* device is still not ready. */
3713 rt6_sync_down_dev(dev, event);
3714 break;
3715 }
3716
3717 if (!IS_ERR_OR_NULL(idev)) {
3718 if (idev->if_flags & IF_READY) {
3719 /* device is already configured -
3720 * but resend MLD reports, we might
3721 * have roamed and need to update
3722 * multicast snooping switches
3723 */
3724 ipv6_mc_up(idev);
3725 change_info = ptr;
3726 if (change_info->flags_changed & IFF_NOARP)
3727 addrconf_dad_run(idev, true);
3728 rt6_sync_up(dev, RTNH_F_LINKDOWN);
3729 break;
3730 }
3731 idev->if_flags |= IF_READY;
3732 }
3733
3734 pr_debug("ADDRCONF(NETDEV_CHANGE): %s: link becomes ready\n",
3735 dev->name);
3736
3737 run_pending = 1;
3738 }
3739
3740 addrconf_init_auto_addrs(dev);
3741
3742 if (!IS_ERR_OR_NULL(idev)) {
3743 if (run_pending)
3744 addrconf_dad_run(idev, false);
3745
3746 /* Device has an address by now */
3747 rt6_sync_up(dev, RTNH_F_DEAD);
3748
3749 /*
3750 * If the MTU changed during the interface down,
3751 * when the interface up, the changed MTU must be
3752 * reflected in the idev as well as routers.
3753 */
3754 if (idev->cnf.mtu6 != dev->mtu &&
3755 dev->mtu >= IPV6_MIN_MTU) {
3756 rt6_mtu_change(dev, dev->mtu);
3757 WRITE_ONCE(idev->cnf.mtu6, dev->mtu);
3758 }
3759 WRITE_ONCE(idev->tstamp, jiffies);
3760 inet6_ifinfo_notify(RTM_NEWLINK, idev);
3761
3762 /*
3763 * If the changed mtu during down is lower than
3764 * IPV6_MIN_MTU stop IPv6 on this interface.
3765 */
3766 if (dev->mtu < IPV6_MIN_MTU)
3767 addrconf_ifdown(dev, dev != net->loopback_dev);
3768 }
3769 break;
3770
3771 case NETDEV_DOWN:
3772 case NETDEV_UNREGISTER:
3773 /*
3774 * Remove all addresses from this interface.
3775 */
3776 addrconf_ifdown(dev, event != NETDEV_DOWN);
3777 break;
3778
3779 case NETDEV_CHANGENAME:
3780 if (idev) {
3781 snmp6_unregister_dev(idev);
3782 addrconf_sysctl_unregister(idev);
3783 err = addrconf_sysctl_register(idev);
3784 if (err)
3785 return notifier_from_errno(err);
3786 err = snmp6_register_dev(idev);
3787 if (err) {
3788 addrconf_sysctl_unregister(idev);
3789 return notifier_from_errno(err);
3790 }
3791 }
3792 break;
3793
3794 case NETDEV_PRE_TYPE_CHANGE:
3795 case NETDEV_POST_TYPE_CHANGE:
3796 if (idev)
3797 addrconf_type_change(dev, event);
3798 break;
3799
3800 case NETDEV_CHANGEUPPER:
3801 info = ptr;
3802
3803 /* flush all routes if dev is linked to or unlinked from
3804 * an L3 master device (e.g., VRF)
3805 */
3806 if (info->upper_dev && netif_is_l3_master(info->upper_dev))
3807 addrconf_ifdown(dev, false);
3808 }
3809
3810 return NOTIFY_OK;
3811 }
3812
3813 /*
3814 * addrconf module should be notified of a device going up
3815 */
3816 static struct notifier_block ipv6_dev_notf = {
3817 .notifier_call = addrconf_notify,
3818 .priority = ADDRCONF_NOTIFY_PRIORITY,
3819 };
3820
addrconf_type_change(struct net_device * dev,unsigned long event)3821 static void addrconf_type_change(struct net_device *dev, unsigned long event)
3822 {
3823 struct inet6_dev *idev;
3824 ASSERT_RTNL();
3825
3826 idev = __in6_dev_get(dev);
3827
3828 if (event == NETDEV_POST_TYPE_CHANGE)
3829 ipv6_mc_remap(idev);
3830 else if (event == NETDEV_PRE_TYPE_CHANGE)
3831 ipv6_mc_unmap(idev);
3832 }
3833
addr_is_local(const struct in6_addr * addr)3834 static bool addr_is_local(const struct in6_addr *addr)
3835 {
3836 return ipv6_addr_type(addr) &
3837 (IPV6_ADDR_LINKLOCAL | IPV6_ADDR_LOOPBACK);
3838 }
3839
addrconf_ifdown(struct net_device * dev,bool unregister)3840 static int addrconf_ifdown(struct net_device *dev, bool unregister)
3841 {
3842 unsigned long event = unregister ? NETDEV_UNREGISTER : NETDEV_DOWN;
3843 struct net *net = dev_net(dev);
3844 struct inet6_dev *idev;
3845 struct inet6_ifaddr *ifa;
3846 LIST_HEAD(tmp_addr_list);
3847 bool keep_addr = false;
3848 bool was_ready;
3849 int state, i;
3850
3851 ASSERT_RTNL();
3852
3853 rt6_disable_ip(dev, event);
3854
3855 idev = __in6_dev_get(dev);
3856 if (!idev)
3857 return -ENODEV;
3858
3859 /*
3860 * Step 1: remove reference to ipv6 device from parent device.
3861 * Do not dev_put!
3862 */
3863 if (unregister) {
3864 WRITE_ONCE(idev->dead, 1);
3865
3866 /* protected by rtnl_lock */
3867 RCU_INIT_POINTER(dev->ip6_ptr, NULL);
3868
3869 /* Step 1.5: remove snmp6 entry */
3870 snmp6_unregister_dev(idev);
3871
3872 }
3873
3874 /* combine the user config with event to determine if permanent
3875 * addresses are to be removed from address hash table
3876 */
3877 if (!unregister && !idev->cnf.disable_ipv6) {
3878 /* aggregate the system setting and interface setting */
3879 int _keep_addr = READ_ONCE(net->ipv6.devconf_all->keep_addr_on_down);
3880
3881 if (!_keep_addr)
3882 _keep_addr = READ_ONCE(idev->cnf.keep_addr_on_down);
3883
3884 keep_addr = (_keep_addr > 0);
3885 }
3886
3887 /* Step 2: clear hash table */
3888 for (i = 0; i < IN6_ADDR_HSIZE; i++) {
3889 struct hlist_head *h = &net->ipv6.inet6_addr_lst[i];
3890
3891 spin_lock_bh(&net->ipv6.addrconf_hash_lock);
3892 restart:
3893 hlist_for_each_entry_rcu(ifa, h, addr_lst) {
3894 if (ifa->idev == idev) {
3895 addrconf_del_dad_work(ifa);
3896 /* combined flag + permanent flag decide if
3897 * address is retained on a down event
3898 */
3899 if (!keep_addr ||
3900 !(ifa->flags & IFA_F_PERMANENT) ||
3901 addr_is_local(&ifa->addr)) {
3902 hlist_del_init_rcu(&ifa->addr_lst);
3903 goto restart;
3904 }
3905 }
3906 }
3907 spin_unlock_bh(&net->ipv6.addrconf_hash_lock);
3908 }
3909
3910 write_lock_bh(&idev->lock);
3911
3912 addrconf_del_rs_timer(idev);
3913
3914 /* Step 2: clear flags for stateless addrconf, repeated down
3915 * detection
3916 */
3917 was_ready = idev->if_flags & IF_READY;
3918 if (!unregister)
3919 idev->if_flags &= ~(IF_RS_SENT|IF_RA_RCVD|IF_READY);
3920
3921 /* Step 3: clear tempaddr list */
3922 while (!list_empty(&idev->tempaddr_list)) {
3923 ifa = list_first_entry(&idev->tempaddr_list,
3924 struct inet6_ifaddr, tmp_list);
3925 list_del(&ifa->tmp_list);
3926 write_unlock_bh(&idev->lock);
3927 spin_lock_bh(&ifa->lock);
3928
3929 if (ifa->ifpub) {
3930 in6_ifa_put(ifa->ifpub);
3931 ifa->ifpub = NULL;
3932 }
3933 spin_unlock_bh(&ifa->lock);
3934 in6_ifa_put(ifa);
3935 write_lock_bh(&idev->lock);
3936 }
3937
3938 list_for_each_entry(ifa, &idev->addr_list, if_list)
3939 list_add_tail(&ifa->if_list_aux, &tmp_addr_list);
3940 write_unlock_bh(&idev->lock);
3941
3942 while (!list_empty(&tmp_addr_list)) {
3943 struct fib6_info *rt = NULL;
3944 bool keep;
3945
3946 ifa = list_first_entry(&tmp_addr_list,
3947 struct inet6_ifaddr, if_list_aux);
3948 list_del(&ifa->if_list_aux);
3949
3950 addrconf_del_dad_work(ifa);
3951
3952 keep = keep_addr && (ifa->flags & IFA_F_PERMANENT) &&
3953 !addr_is_local(&ifa->addr);
3954
3955 spin_lock_bh(&ifa->lock);
3956
3957 if (keep) {
3958 /* set state to skip the notifier below */
3959 state = INET6_IFADDR_STATE_DEAD;
3960 ifa->state = INET6_IFADDR_STATE_PREDAD;
3961 if (!(ifa->flags & IFA_F_NODAD))
3962 ifa->flags |= IFA_F_TENTATIVE;
3963
3964 rt = ifa->rt;
3965 ifa->rt = NULL;
3966 } else {
3967 state = ifa->state;
3968 ifa->state = INET6_IFADDR_STATE_DEAD;
3969 }
3970
3971 spin_unlock_bh(&ifa->lock);
3972
3973 if (rt)
3974 ip6_del_rt(net, rt, false);
3975
3976 if (state != INET6_IFADDR_STATE_DEAD) {
3977 __ipv6_ifa_notify(RTM_DELADDR, ifa);
3978 inet6addr_notifier_call_chain(NETDEV_DOWN, ifa);
3979 } else {
3980 if (idev->cnf.forwarding)
3981 addrconf_leave_anycast(ifa);
3982 addrconf_leave_solict(ifa->idev, &ifa->addr);
3983 }
3984
3985 if (!keep) {
3986 write_lock_bh(&idev->lock);
3987 list_del_rcu(&ifa->if_list);
3988 write_unlock_bh(&idev->lock);
3989 in6_ifa_put(ifa);
3990 }
3991 }
3992
3993 /* Step 5: Discard anycast and multicast list */
3994 if (unregister) {
3995 ipv6_ac_destroy_dev(idev);
3996 ipv6_mc_destroy_dev(idev);
3997 } else if (was_ready) {
3998 ipv6_mc_down(idev);
3999 }
4000
4001 WRITE_ONCE(idev->tstamp, jiffies);
4002 idev->ra_mtu = 0;
4003
4004 /* Last: Shot the device (if unregistered) */
4005 if (unregister) {
4006 addrconf_sysctl_unregister(idev);
4007 neigh_parms_release(&nd_tbl, idev->nd_parms);
4008 neigh_ifdown(&nd_tbl, dev);
4009 in6_dev_put(idev);
4010 }
4011 return 0;
4012 }
4013
addrconf_rs_timer(struct timer_list * t)4014 static void addrconf_rs_timer(struct timer_list *t)
4015 {
4016 struct inet6_dev *idev = timer_container_of(idev, t, rs_timer);
4017 struct net_device *dev = idev->dev;
4018 struct in6_addr lladdr;
4019 int rtr_solicits;
4020
4021 write_lock(&idev->lock);
4022 if (idev->dead || !(idev->if_flags & IF_READY))
4023 goto out;
4024
4025 if (!ipv6_accept_ra(idev))
4026 goto out;
4027
4028 /* Announcement received after solicitation was sent */
4029 if (idev->if_flags & IF_RA_RCVD)
4030 goto out;
4031
4032 rtr_solicits = READ_ONCE(idev->cnf.rtr_solicits);
4033
4034 if (idev->rs_probes++ < rtr_solicits || rtr_solicits < 0) {
4035 write_unlock(&idev->lock);
4036 if (!ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE))
4037 ndisc_send_rs(dev, &lladdr,
4038 &in6addr_linklocal_allrouters);
4039 else
4040 goto put;
4041
4042 write_lock(&idev->lock);
4043 idev->rs_interval = rfc3315_s14_backoff_update(
4044 idev->rs_interval,
4045 READ_ONCE(idev->cnf.rtr_solicit_max_interval));
4046 /* The wait after the last probe can be shorter */
4047 addrconf_mod_rs_timer(idev, (idev->rs_probes ==
4048 READ_ONCE(idev->cnf.rtr_solicits)) ?
4049 READ_ONCE(idev->cnf.rtr_solicit_delay) :
4050 idev->rs_interval);
4051 } else {
4052 /*
4053 * Note: we do not support deprecated "all on-link"
4054 * assumption any longer.
4055 */
4056 pr_debug("%s: no IPv6 routers present\n", idev->dev->name);
4057 }
4058
4059 out:
4060 write_unlock(&idev->lock);
4061 put:
4062 in6_dev_put(idev);
4063 }
4064
4065 /*
4066 * Duplicate Address Detection
4067 */
addrconf_dad_kick(struct inet6_ifaddr * ifp)4068 static void addrconf_dad_kick(struct inet6_ifaddr *ifp)
4069 {
4070 struct inet6_dev *idev = ifp->idev;
4071 unsigned long rand_num;
4072 u64 nonce;
4073
4074 if (ifp->flags & IFA_F_OPTIMISTIC)
4075 rand_num = 0;
4076 else
4077 rand_num = get_random_u32_below(
4078 READ_ONCE(idev->cnf.rtr_solicit_delay) ? : 1);
4079
4080 nonce = 0;
4081 if (READ_ONCE(idev->cnf.enhanced_dad) ||
4082 READ_ONCE(dev_net(idev->dev)->ipv6.devconf_all->enhanced_dad)) {
4083 do
4084 get_random_bytes(&nonce, 6);
4085 while (nonce == 0);
4086 }
4087 ifp->dad_nonce = nonce;
4088 ifp->dad_probes = READ_ONCE(idev->cnf.dad_transmits);
4089 addrconf_mod_dad_work(ifp, rand_num);
4090 }
4091
addrconf_dad_begin(struct inet6_ifaddr * ifp)4092 static void addrconf_dad_begin(struct inet6_ifaddr *ifp)
4093 {
4094 struct inet6_dev *idev = ifp->idev;
4095 struct net_device *dev = idev->dev;
4096 bool bump_id, notify = false;
4097 struct net *net;
4098
4099 addrconf_join_solict(dev, &ifp->addr);
4100
4101 read_lock_bh(&idev->lock);
4102 spin_lock(&ifp->lock);
4103 if (ifp->state == INET6_IFADDR_STATE_DEAD)
4104 goto out;
4105
4106 net = dev_net(dev);
4107 if (dev->flags&(IFF_NOARP|IFF_LOOPBACK) ||
4108 (READ_ONCE(net->ipv6.devconf_all->accept_dad) < 1 &&
4109 READ_ONCE(idev->cnf.accept_dad) < 1) ||
4110 !(ifp->flags&IFA_F_TENTATIVE) ||
4111 ifp->flags & IFA_F_NODAD) {
4112 bool send_na = false;
4113
4114 if (ifp->flags & IFA_F_TENTATIVE &&
4115 !(ifp->flags & IFA_F_OPTIMISTIC))
4116 send_na = true;
4117 bump_id = ifp->flags & IFA_F_TENTATIVE;
4118 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
4119 spin_unlock(&ifp->lock);
4120 read_unlock_bh(&idev->lock);
4121
4122 addrconf_dad_completed(ifp, bump_id, send_na);
4123 return;
4124 }
4125
4126 if (!(idev->if_flags & IF_READY)) {
4127 spin_unlock(&ifp->lock);
4128 read_unlock_bh(&idev->lock);
4129 /*
4130 * If the device is not ready:
4131 * - keep it tentative if it is a permanent address.
4132 * - otherwise, kill it.
4133 */
4134 in6_ifa_hold(ifp);
4135 addrconf_dad_stop(ifp, 0);
4136 return;
4137 }
4138
4139 /*
4140 * Optimistic nodes can start receiving
4141 * Frames right away
4142 */
4143 if (ifp->flags & IFA_F_OPTIMISTIC) {
4144 ip6_ins_rt(net, ifp->rt);
4145 if (ipv6_use_optimistic_addr(net, idev)) {
4146 /* Because optimistic nodes can use this address,
4147 * notify listeners. If DAD fails, RTM_DELADDR is sent.
4148 */
4149 notify = true;
4150 }
4151 }
4152
4153 addrconf_dad_kick(ifp);
4154 out:
4155 spin_unlock(&ifp->lock);
4156 read_unlock_bh(&idev->lock);
4157 if (notify)
4158 ipv6_ifa_notify(RTM_NEWADDR, ifp);
4159 }
4160
addrconf_dad_start(struct inet6_ifaddr * ifp)4161 static void addrconf_dad_start(struct inet6_ifaddr *ifp)
4162 {
4163 bool begin_dad = false;
4164
4165 spin_lock_bh(&ifp->lock);
4166 if (ifp->state != INET6_IFADDR_STATE_DEAD) {
4167 ifp->state = INET6_IFADDR_STATE_PREDAD;
4168 begin_dad = true;
4169 }
4170 spin_unlock_bh(&ifp->lock);
4171
4172 if (begin_dad)
4173 addrconf_mod_dad_work(ifp, 0);
4174 }
4175
addrconf_dad_work(struct work_struct * w)4176 static void addrconf_dad_work(struct work_struct *w)
4177 {
4178 struct inet6_ifaddr *ifp = container_of(to_delayed_work(w),
4179 struct inet6_ifaddr,
4180 dad_work);
4181 struct inet6_dev *idev = ifp->idev;
4182 bool bump_id, disable_ipv6 = false;
4183 struct in6_addr mcaddr;
4184 struct net *net;
4185
4186 enum {
4187 DAD_PROCESS,
4188 DAD_BEGIN,
4189 DAD_ABORT,
4190 } action = DAD_PROCESS;
4191
4192 net = dev_net(idev->dev);
4193
4194 rtnl_net_lock(net);
4195
4196 spin_lock_bh(&ifp->lock);
4197 if (ifp->state == INET6_IFADDR_STATE_PREDAD) {
4198 action = DAD_BEGIN;
4199 ifp->state = INET6_IFADDR_STATE_DAD;
4200 } else if (ifp->state == INET6_IFADDR_STATE_ERRDAD) {
4201 action = DAD_ABORT;
4202 ifp->state = INET6_IFADDR_STATE_POSTDAD;
4203
4204 if ((READ_ONCE(net->ipv6.devconf_all->accept_dad) > 1 ||
4205 READ_ONCE(idev->cnf.accept_dad) > 1) &&
4206 !idev->cnf.disable_ipv6 &&
4207 !(ifp->flags & IFA_F_STABLE_PRIVACY)) {
4208 struct in6_addr addr;
4209
4210 addr.s6_addr32[0] = htonl(0xfe800000);
4211 addr.s6_addr32[1] = 0;
4212
4213 if (!ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) &&
4214 ipv6_addr_equal(&ifp->addr, &addr)) {
4215 /* DAD failed for link-local based on MAC */
4216 WRITE_ONCE(idev->cnf.disable_ipv6, 1);
4217
4218 pr_info("%s: IPv6 being disabled!\n",
4219 ifp->idev->dev->name);
4220 disable_ipv6 = true;
4221 }
4222 }
4223 }
4224 spin_unlock_bh(&ifp->lock);
4225
4226 if (action == DAD_BEGIN) {
4227 addrconf_dad_begin(ifp);
4228 goto out;
4229 } else if (action == DAD_ABORT) {
4230 in6_ifa_hold(ifp);
4231 addrconf_dad_stop(ifp, 1);
4232 if (disable_ipv6)
4233 addrconf_ifdown(idev->dev, false);
4234 goto out;
4235 }
4236
4237 if (!ifp->dad_probes && addrconf_dad_end(ifp))
4238 goto out;
4239
4240 write_lock_bh(&idev->lock);
4241 if (idev->dead || !(idev->if_flags & IF_READY)) {
4242 write_unlock_bh(&idev->lock);
4243 goto out;
4244 }
4245
4246 spin_lock(&ifp->lock);
4247 if (ifp->state == INET6_IFADDR_STATE_DEAD) {
4248 spin_unlock(&ifp->lock);
4249 write_unlock_bh(&idev->lock);
4250 goto out;
4251 }
4252
4253 if (ifp->dad_probes == 0) {
4254 bool send_na = false;
4255
4256 /*
4257 * DAD was successful
4258 */
4259
4260 if (ifp->flags & IFA_F_TENTATIVE &&
4261 !(ifp->flags & IFA_F_OPTIMISTIC))
4262 send_na = true;
4263 bump_id = ifp->flags & IFA_F_TENTATIVE;
4264 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
4265 spin_unlock(&ifp->lock);
4266 write_unlock_bh(&idev->lock);
4267
4268 addrconf_dad_completed(ifp, bump_id, send_na);
4269
4270 goto out;
4271 }
4272
4273 ifp->dad_probes--;
4274 addrconf_mod_dad_work(ifp,
4275 max(NEIGH_VAR(ifp->idev->nd_parms, RETRANS_TIME),
4276 HZ/100));
4277 spin_unlock(&ifp->lock);
4278 write_unlock_bh(&idev->lock);
4279
4280 /* send a neighbour solicitation for our addr */
4281 addrconf_addr_solict_mult(&ifp->addr, &mcaddr);
4282 ndisc_send_ns(ifp->idev->dev, &ifp->addr, &mcaddr, &in6addr_any,
4283 ifp->dad_nonce);
4284 out:
4285 in6_ifa_put(ifp);
4286 rtnl_net_unlock(net);
4287 }
4288
4289 /* ifp->idev must be at least read locked */
ipv6_lonely_lladdr(struct inet6_ifaddr * ifp)4290 static bool ipv6_lonely_lladdr(struct inet6_ifaddr *ifp)
4291 {
4292 struct inet6_ifaddr *ifpiter;
4293 struct inet6_dev *idev = ifp->idev;
4294
4295 list_for_each_entry_reverse(ifpiter, &idev->addr_list, if_list) {
4296 if (ifpiter->scope > IFA_LINK)
4297 break;
4298 if (ifp != ifpiter && ifpiter->scope == IFA_LINK &&
4299 (ifpiter->flags & (IFA_F_PERMANENT|IFA_F_TENTATIVE|
4300 IFA_F_OPTIMISTIC|IFA_F_DADFAILED)) ==
4301 IFA_F_PERMANENT)
4302 return false;
4303 }
4304 return true;
4305 }
4306
addrconf_dad_completed(struct inet6_ifaddr * ifp,bool bump_id,bool send_na)4307 static void addrconf_dad_completed(struct inet6_ifaddr *ifp, bool bump_id,
4308 bool send_na)
4309 {
4310 struct net_device *dev = ifp->idev->dev;
4311 struct in6_addr lladdr;
4312 bool send_rs, send_mld;
4313
4314 addrconf_del_dad_work(ifp);
4315
4316 /*
4317 * Configure the address for reception. Now it is valid.
4318 */
4319
4320 ipv6_ifa_notify(RTM_NEWADDR, ifp);
4321
4322 /* If added prefix is link local and we are prepared to process
4323 router advertisements, start sending router solicitations.
4324 */
4325
4326 read_lock_bh(&ifp->idev->lock);
4327 send_mld = ifp->scope == IFA_LINK && ipv6_lonely_lladdr(ifp);
4328 send_rs = send_mld &&
4329 ipv6_accept_ra(ifp->idev) &&
4330 READ_ONCE(ifp->idev->cnf.rtr_solicits) != 0 &&
4331 (dev->flags & IFF_LOOPBACK) == 0 &&
4332 (dev->type != ARPHRD_TUNNEL) &&
4333 !netif_is_team_port(dev);
4334 read_unlock_bh(&ifp->idev->lock);
4335
4336 /* While dad is in progress mld report's source address is in6_addrany.
4337 * Resend with proper ll now.
4338 */
4339 if (send_mld)
4340 ipv6_mc_dad_complete(ifp->idev);
4341
4342 /* send unsolicited NA if enabled */
4343 if (send_na &&
4344 (READ_ONCE(ifp->idev->cnf.ndisc_notify) ||
4345 READ_ONCE(dev_net(dev)->ipv6.devconf_all->ndisc_notify))) {
4346 ndisc_send_na(dev, &in6addr_linklocal_allnodes, &ifp->addr,
4347 /*router=*/ !!ifp->idev->cnf.forwarding,
4348 /*solicited=*/ false, /*override=*/ true,
4349 /*inc_opt=*/ true);
4350 }
4351
4352 if (send_rs) {
4353 /*
4354 * If a host as already performed a random delay
4355 * [...] as part of DAD [...] there is no need
4356 * to delay again before sending the first RS
4357 */
4358 if (ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE))
4359 return;
4360 ndisc_send_rs(dev, &lladdr, &in6addr_linklocal_allrouters);
4361
4362 write_lock_bh(&ifp->idev->lock);
4363 spin_lock(&ifp->lock);
4364 ifp->idev->rs_interval = rfc3315_s14_backoff_init(
4365 READ_ONCE(ifp->idev->cnf.rtr_solicit_interval));
4366 ifp->idev->rs_probes = 1;
4367 ifp->idev->if_flags |= IF_RS_SENT;
4368 addrconf_mod_rs_timer(ifp->idev, ifp->idev->rs_interval);
4369 spin_unlock(&ifp->lock);
4370 write_unlock_bh(&ifp->idev->lock);
4371 }
4372
4373 if (bump_id)
4374 rt_genid_bump_ipv6(dev_net(dev));
4375
4376 /* Make sure that a new temporary address will be created
4377 * before this temporary address becomes deprecated.
4378 */
4379 if (ifp->flags & IFA_F_TEMPORARY)
4380 addrconf_verify_rtnl(dev_net(dev));
4381 }
4382
addrconf_dad_run(struct inet6_dev * idev,bool restart)4383 static void addrconf_dad_run(struct inet6_dev *idev, bool restart)
4384 {
4385 struct inet6_ifaddr *ifp;
4386
4387 read_lock_bh(&idev->lock);
4388 list_for_each_entry(ifp, &idev->addr_list, if_list) {
4389 spin_lock(&ifp->lock);
4390 if ((ifp->flags & IFA_F_TENTATIVE &&
4391 ifp->state == INET6_IFADDR_STATE_DAD) || restart) {
4392 if (restart)
4393 ifp->state = INET6_IFADDR_STATE_PREDAD;
4394 addrconf_dad_kick(ifp);
4395 }
4396 spin_unlock(&ifp->lock);
4397 }
4398 read_unlock_bh(&idev->lock);
4399 }
4400
4401 #ifdef CONFIG_PROC_FS
4402 struct if6_iter_state {
4403 struct seq_net_private p;
4404 int bucket;
4405 int offset;
4406 };
4407
if6_get_first(struct seq_file * seq,loff_t pos)4408 static struct inet6_ifaddr *if6_get_first(struct seq_file *seq, loff_t pos)
4409 {
4410 struct if6_iter_state *state = seq->private;
4411 struct net *net = seq_file_net(seq);
4412 struct inet6_ifaddr *ifa = NULL;
4413 int p = 0;
4414
4415 /* initial bucket if pos is 0 */
4416 if (pos == 0) {
4417 state->bucket = 0;
4418 state->offset = 0;
4419 }
4420
4421 for (; state->bucket < IN6_ADDR_HSIZE; ++state->bucket) {
4422 hlist_for_each_entry_rcu(ifa, &net->ipv6.inet6_addr_lst[state->bucket],
4423 addr_lst) {
4424 /* sync with offset */
4425 if (p < state->offset) {
4426 p++;
4427 continue;
4428 }
4429 return ifa;
4430 }
4431
4432 /* prepare for next bucket */
4433 state->offset = 0;
4434 p = 0;
4435 }
4436 return NULL;
4437 }
4438
if6_get_next(struct seq_file * seq,struct inet6_ifaddr * ifa)4439 static struct inet6_ifaddr *if6_get_next(struct seq_file *seq,
4440 struct inet6_ifaddr *ifa)
4441 {
4442 struct if6_iter_state *state = seq->private;
4443 struct net *net = seq_file_net(seq);
4444
4445 hlist_for_each_entry_continue_rcu(ifa, addr_lst) {
4446 state->offset++;
4447 return ifa;
4448 }
4449
4450 state->offset = 0;
4451 while (++state->bucket < IN6_ADDR_HSIZE) {
4452 hlist_for_each_entry_rcu(ifa,
4453 &net->ipv6.inet6_addr_lst[state->bucket], addr_lst) {
4454 return ifa;
4455 }
4456 }
4457
4458 return NULL;
4459 }
4460
if6_seq_start(struct seq_file * seq,loff_t * pos)4461 static void *if6_seq_start(struct seq_file *seq, loff_t *pos)
4462 __acquires(rcu)
4463 {
4464 rcu_read_lock();
4465 return if6_get_first(seq, *pos);
4466 }
4467
if6_seq_next(struct seq_file * seq,void * v,loff_t * pos)4468 static void *if6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
4469 {
4470 struct inet6_ifaddr *ifa;
4471
4472 ifa = if6_get_next(seq, v);
4473 ++*pos;
4474 return ifa;
4475 }
4476
if6_seq_stop(struct seq_file * seq,void * v)4477 static void if6_seq_stop(struct seq_file *seq, void *v)
4478 __releases(rcu)
4479 {
4480 rcu_read_unlock();
4481 }
4482
if6_seq_show(struct seq_file * seq,void * v)4483 static int if6_seq_show(struct seq_file *seq, void *v)
4484 {
4485 struct inet6_ifaddr *ifp = (struct inet6_ifaddr *)v;
4486 seq_printf(seq, "%pi6 %02x %02x %02x %02x %8s\n",
4487 &ifp->addr,
4488 ifp->idev->dev->ifindex,
4489 ifp->prefix_len,
4490 ifp->scope,
4491 (u8) ifp->flags,
4492 ifp->idev->dev->name);
4493 return 0;
4494 }
4495
4496 static const struct seq_operations if6_seq_ops = {
4497 .start = if6_seq_start,
4498 .next = if6_seq_next,
4499 .show = if6_seq_show,
4500 .stop = if6_seq_stop,
4501 };
4502
if6_proc_net_init(struct net * net)4503 static int __net_init if6_proc_net_init(struct net *net)
4504 {
4505 if (!proc_create_net("if_inet6", 0444, net->proc_net, &if6_seq_ops,
4506 sizeof(struct if6_iter_state)))
4507 return -ENOMEM;
4508 return 0;
4509 }
4510
if6_proc_net_exit(struct net * net)4511 static void __net_exit if6_proc_net_exit(struct net *net)
4512 {
4513 remove_proc_entry("if_inet6", net->proc_net);
4514 }
4515
4516 static struct pernet_operations if6_proc_net_ops = {
4517 .init = if6_proc_net_init,
4518 .exit = if6_proc_net_exit,
4519 };
4520
if6_proc_init(void)4521 int __init if6_proc_init(void)
4522 {
4523 return register_pernet_subsys(&if6_proc_net_ops);
4524 }
4525
if6_proc_exit(void)4526 void if6_proc_exit(void)
4527 {
4528 unregister_pernet_subsys(&if6_proc_net_ops);
4529 }
4530 #endif /* CONFIG_PROC_FS */
4531
4532 #if IS_ENABLED(CONFIG_IPV6_MIP6)
4533 /* Check if address is a home address configured on any interface. */
ipv6_chk_home_addr(struct net * net,const struct in6_addr * addr)4534 int ipv6_chk_home_addr(struct net *net, const struct in6_addr *addr)
4535 {
4536 unsigned int hash = inet6_addr_hash(net, addr);
4537 struct inet6_ifaddr *ifp = NULL;
4538 int ret = 0;
4539
4540 rcu_read_lock();
4541 hlist_for_each_entry_rcu(ifp, &net->ipv6.inet6_addr_lst[hash], addr_lst) {
4542 if (ipv6_addr_equal(&ifp->addr, addr) &&
4543 (ifp->flags & IFA_F_HOMEADDRESS)) {
4544 ret = 1;
4545 break;
4546 }
4547 }
4548 rcu_read_unlock();
4549 return ret;
4550 }
4551 #endif
4552
4553 /* RFC6554 has some algorithm to avoid loops in segment routing by
4554 * checking if the segments contains any of a local interface address.
4555 *
4556 * Quote:
4557 *
4558 * To detect loops in the SRH, a router MUST determine if the SRH
4559 * includes multiple addresses assigned to any interface on that router.
4560 * If such addresses appear more than once and are separated by at least
4561 * one address not assigned to that router.
4562 */
ipv6_chk_rpl_srh_loop(struct net * net,const struct in6_addr * segs,unsigned char nsegs)4563 int ipv6_chk_rpl_srh_loop(struct net *net, const struct in6_addr *segs,
4564 unsigned char nsegs)
4565 {
4566 const struct in6_addr *addr;
4567 int i, ret = 0, found = 0;
4568 struct inet6_ifaddr *ifp;
4569 bool separated = false;
4570 unsigned int hash;
4571 bool hash_found;
4572
4573 rcu_read_lock();
4574 for (i = 0; i < nsegs; i++) {
4575 addr = &segs[i];
4576 hash = inet6_addr_hash(net, addr);
4577
4578 hash_found = false;
4579 hlist_for_each_entry_rcu(ifp, &net->ipv6.inet6_addr_lst[hash], addr_lst) {
4580
4581 if (ipv6_addr_equal(&ifp->addr, addr)) {
4582 hash_found = true;
4583 break;
4584 }
4585 }
4586
4587 if (hash_found) {
4588 if (found > 1 && separated) {
4589 ret = 1;
4590 break;
4591 }
4592
4593 separated = false;
4594 found++;
4595 } else {
4596 separated = true;
4597 }
4598 }
4599 rcu_read_unlock();
4600
4601 return ret;
4602 }
4603
4604 /*
4605 * Periodic address status verification
4606 */
4607
addrconf_verify_rtnl(struct net * net)4608 static void addrconf_verify_rtnl(struct net *net)
4609 {
4610 unsigned long now, next, next_sec, next_sched;
4611 struct inet6_ifaddr *ifp;
4612 int i;
4613
4614 ASSERT_RTNL();
4615
4616 rcu_read_lock_bh();
4617 now = jiffies;
4618 next = round_jiffies_up(now + ADDR_CHECK_FREQUENCY);
4619
4620 cancel_delayed_work(&net->ipv6.addr_chk_work);
4621
4622 for (i = 0; i < IN6_ADDR_HSIZE; i++) {
4623 restart:
4624 hlist_for_each_entry_rcu_bh(ifp, &net->ipv6.inet6_addr_lst[i], addr_lst) {
4625 unsigned long age;
4626
4627 /* When setting preferred_lft to a value not zero or
4628 * infinity, while valid_lft is infinity
4629 * IFA_F_PERMANENT has a non-infinity life time.
4630 */
4631 if ((ifp->flags & IFA_F_PERMANENT) &&
4632 (ifp->prefered_lft == INFINITY_LIFE_TIME))
4633 continue;
4634
4635 spin_lock(&ifp->lock);
4636 /* We try to batch several events at once. */
4637 age = (now - ifp->tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
4638
4639 if ((ifp->flags&IFA_F_TEMPORARY) &&
4640 !(ifp->flags&IFA_F_TENTATIVE) &&
4641 ifp->prefered_lft != INFINITY_LIFE_TIME &&
4642 !ifp->regen_count && ifp->ifpub) {
4643 /* This is a non-regenerated temporary addr. */
4644
4645 unsigned long regen_advance = ipv6_get_regen_advance(ifp->idev);
4646
4647 if (age + regen_advance >= ifp->prefered_lft) {
4648 struct inet6_ifaddr *ifpub = ifp->ifpub;
4649 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
4650 next = ifp->tstamp + ifp->prefered_lft * HZ;
4651
4652 ifp->regen_count++;
4653 in6_ifa_hold(ifp);
4654 in6_ifa_hold(ifpub);
4655 spin_unlock(&ifp->lock);
4656
4657 spin_lock(&ifpub->lock);
4658 ifpub->regen_count = 0;
4659 spin_unlock(&ifpub->lock);
4660 rcu_read_unlock_bh();
4661 ipv6_create_tempaddr(ifpub, true);
4662 in6_ifa_put(ifpub);
4663 in6_ifa_put(ifp);
4664 rcu_read_lock_bh();
4665 goto restart;
4666 } else if (time_before(ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ, next))
4667 next = ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ;
4668 }
4669
4670 if (ifp->valid_lft != INFINITY_LIFE_TIME &&
4671 age >= ifp->valid_lft) {
4672 spin_unlock(&ifp->lock);
4673 in6_ifa_hold(ifp);
4674 rcu_read_unlock_bh();
4675 ipv6_del_addr(ifp);
4676 rcu_read_lock_bh();
4677 goto restart;
4678 } else if (ifp->prefered_lft == INFINITY_LIFE_TIME) {
4679 spin_unlock(&ifp->lock);
4680 continue;
4681 } else if (age >= ifp->prefered_lft) {
4682 /* jiffies - ifp->tstamp > age >= ifp->prefered_lft */
4683 int deprecate = 0;
4684
4685 if (!(ifp->flags&IFA_F_DEPRECATED)) {
4686 deprecate = 1;
4687 ifp->flags |= IFA_F_DEPRECATED;
4688 }
4689
4690 if ((ifp->valid_lft != INFINITY_LIFE_TIME) &&
4691 (time_before(ifp->tstamp + ifp->valid_lft * HZ, next)))
4692 next = ifp->tstamp + ifp->valid_lft * HZ;
4693
4694 spin_unlock(&ifp->lock);
4695
4696 if (deprecate) {
4697 in6_ifa_hold(ifp);
4698
4699 ipv6_ifa_notify(0, ifp);
4700 in6_ifa_put(ifp);
4701 goto restart;
4702 }
4703 } else {
4704 /* ifp->prefered_lft <= ifp->valid_lft */
4705 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
4706 next = ifp->tstamp + ifp->prefered_lft * HZ;
4707 spin_unlock(&ifp->lock);
4708 }
4709 }
4710 }
4711
4712 next_sec = round_jiffies_up(next);
4713 next_sched = next;
4714
4715 /* If rounded timeout is accurate enough, accept it. */
4716 if (time_before(next_sec, next + ADDRCONF_TIMER_FUZZ))
4717 next_sched = next_sec;
4718
4719 /* And minimum interval is ADDRCONF_TIMER_FUZZ_MAX. */
4720 if (time_before(next_sched, jiffies + ADDRCONF_TIMER_FUZZ_MAX))
4721 next_sched = jiffies + ADDRCONF_TIMER_FUZZ_MAX;
4722
4723 pr_debug("now = %lu, schedule = %lu, rounded schedule = %lu => %lu\n",
4724 now, next, next_sec, next_sched);
4725 mod_delayed_work(addrconf_wq, &net->ipv6.addr_chk_work, next_sched - now);
4726 rcu_read_unlock_bh();
4727 }
4728
addrconf_verify_work(struct work_struct * w)4729 static void addrconf_verify_work(struct work_struct *w)
4730 {
4731 struct net *net = container_of(to_delayed_work(w), struct net,
4732 ipv6.addr_chk_work);
4733
4734 rtnl_net_lock(net);
4735 addrconf_verify_rtnl(net);
4736 rtnl_net_unlock(net);
4737 }
4738
addrconf_verify(struct net * net)4739 static void addrconf_verify(struct net *net)
4740 {
4741 mod_delayed_work(addrconf_wq, &net->ipv6.addr_chk_work, 0);
4742 }
4743
extract_addr(struct nlattr * addr,struct nlattr * local,struct in6_addr ** peer_pfx)4744 static struct in6_addr *extract_addr(struct nlattr *addr, struct nlattr *local,
4745 struct in6_addr **peer_pfx)
4746 {
4747 struct in6_addr *pfx = NULL;
4748
4749 *peer_pfx = NULL;
4750
4751 if (addr)
4752 pfx = nla_data(addr);
4753
4754 if (local) {
4755 if (pfx && nla_memcmp(local, pfx, sizeof(*pfx)))
4756 *peer_pfx = pfx;
4757 pfx = nla_data(local);
4758 }
4759
4760 return pfx;
4761 }
4762
4763 static const struct nla_policy ifa_ipv6_policy[IFA_MAX+1] = {
4764 [IFA_ADDRESS] = { .len = sizeof(struct in6_addr) },
4765 [IFA_LOCAL] = { .len = sizeof(struct in6_addr) },
4766 [IFA_CACHEINFO] = { .len = sizeof(struct ifa_cacheinfo) },
4767 [IFA_FLAGS] = { .len = sizeof(u32) },
4768 [IFA_RT_PRIORITY] = { .len = sizeof(u32) },
4769 [IFA_TARGET_NETNSID] = { .type = NLA_S32 },
4770 [IFA_PROTO] = { .type = NLA_U8 },
4771 };
4772
4773 static int
inet6_rtm_deladdr(struct sk_buff * skb,struct nlmsghdr * nlh,struct netlink_ext_ack * extack)4774 inet6_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh,
4775 struct netlink_ext_ack *extack)
4776 {
4777 struct net *net = sock_net(skb->sk);
4778 struct ifaddrmsg *ifm;
4779 struct nlattr *tb[IFA_MAX+1];
4780 struct in6_addr *pfx, *peer_pfx;
4781 u32 ifa_flags;
4782 int err;
4783
4784 err = nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFA_MAX,
4785 ifa_ipv6_policy, extack);
4786 if (err < 0)
4787 return err;
4788
4789 ifm = nlmsg_data(nlh);
4790 pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
4791 if (!pfx)
4792 return -EINVAL;
4793
4794 ifa_flags = nla_get_u32_default(tb[IFA_FLAGS], ifm->ifa_flags);
4795
4796 /* We ignore other flags so far. */
4797 ifa_flags &= IFA_F_MANAGETEMPADDR;
4798
4799 rtnl_net_lock(net);
4800 err = inet6_addr_del(net, ifm->ifa_index, ifa_flags, pfx,
4801 ifm->ifa_prefixlen, extack);
4802 rtnl_net_unlock(net);
4803
4804 return err;
4805 }
4806
modify_prefix_route(struct net * net,struct inet6_ifaddr * ifp,unsigned long expires,u32 flags,bool modify_peer)4807 static int modify_prefix_route(struct net *net, struct inet6_ifaddr *ifp,
4808 unsigned long expires, u32 flags,
4809 bool modify_peer)
4810 {
4811 struct fib6_table *table;
4812 struct fib6_info *f6i;
4813 u32 prio;
4814
4815 f6i = addrconf_get_prefix_route(modify_peer ? &ifp->peer_addr : &ifp->addr,
4816 ifp->prefix_len,
4817 ifp->idev->dev, 0, RTF_DEFAULT, true);
4818 if (!f6i)
4819 return -ENOENT;
4820
4821 prio = ifp->rt_priority ? : IP6_RT_PRIO_ADDRCONF;
4822 if (f6i->fib6_metric != prio) {
4823 /* delete old one */
4824 ip6_del_rt(dev_net(ifp->idev->dev), f6i, false);
4825
4826 /* add new one */
4827 addrconf_prefix_route(modify_peer ? &ifp->peer_addr : &ifp->addr,
4828 ifp->prefix_len,
4829 ifp->rt_priority, ifp->idev->dev,
4830 expires, flags, GFP_KERNEL);
4831 return 0;
4832 }
4833 if (f6i != net->ipv6.fib6_null_entry) {
4834 table = f6i->fib6_table;
4835 spin_lock_bh(&table->tb6_lock);
4836
4837 if (!(flags & RTF_EXPIRES)) {
4838 fib6_clean_expires(f6i);
4839 fib6_remove_gc_list(f6i);
4840 } else {
4841 fib6_set_expires(f6i, expires);
4842 fib6_add_gc_list(f6i);
4843 }
4844
4845 spin_unlock_bh(&table->tb6_lock);
4846 }
4847 fib6_info_release(f6i);
4848
4849 return 0;
4850 }
4851
inet6_addr_modify(struct net * net,struct inet6_ifaddr * ifp,struct ifa6_config * cfg,clock_t expires,u32 flags)4852 static int inet6_addr_modify(struct net *net, struct inet6_ifaddr *ifp,
4853 struct ifa6_config *cfg, clock_t expires,
4854 u32 flags)
4855 {
4856 bool was_managetempaddr;
4857 bool new_peer = false;
4858 bool had_prefixroute;
4859
4860 ASSERT_RTNL_NET(net);
4861
4862 if (cfg->ifa_flags & IFA_F_MANAGETEMPADDR &&
4863 (ifp->flags & IFA_F_TEMPORARY || ifp->prefix_len != 64))
4864 return -EINVAL;
4865
4866 if (!(ifp->flags & IFA_F_TENTATIVE) || ifp->flags & IFA_F_DADFAILED)
4867 cfg->ifa_flags &= ~IFA_F_OPTIMISTIC;
4868
4869 if (cfg->peer_pfx &&
4870 memcmp(&ifp->peer_addr, cfg->peer_pfx, sizeof(struct in6_addr))) {
4871 if (!ipv6_addr_any(&ifp->peer_addr))
4872 cleanup_prefix_route(ifp, expires, true, true);
4873 new_peer = true;
4874 }
4875
4876 spin_lock_bh(&ifp->lock);
4877 was_managetempaddr = ifp->flags & IFA_F_MANAGETEMPADDR;
4878 had_prefixroute = ifp->flags & IFA_F_PERMANENT &&
4879 !(ifp->flags & IFA_F_NOPREFIXROUTE);
4880 ifp->flags &= ~(IFA_F_DEPRECATED | IFA_F_PERMANENT | IFA_F_NODAD |
4881 IFA_F_HOMEADDRESS | IFA_F_MANAGETEMPADDR |
4882 IFA_F_NOPREFIXROUTE);
4883 ifp->flags |= cfg->ifa_flags;
4884 WRITE_ONCE(ifp->tstamp, jiffies);
4885 WRITE_ONCE(ifp->valid_lft, cfg->valid_lft);
4886 WRITE_ONCE(ifp->prefered_lft, cfg->preferred_lft);
4887 WRITE_ONCE(ifp->ifa_proto, cfg->ifa_proto);
4888
4889 if (cfg->rt_priority && cfg->rt_priority != ifp->rt_priority)
4890 WRITE_ONCE(ifp->rt_priority, cfg->rt_priority);
4891
4892 if (new_peer)
4893 ifp->peer_addr = *cfg->peer_pfx;
4894
4895 spin_unlock_bh(&ifp->lock);
4896 if (!(ifp->flags&IFA_F_TENTATIVE))
4897 ipv6_ifa_notify(0, ifp);
4898
4899 if (!(cfg->ifa_flags & IFA_F_NOPREFIXROUTE)) {
4900 int rc = -ENOENT;
4901
4902 if (had_prefixroute)
4903 rc = modify_prefix_route(net, ifp, expires, flags, false);
4904
4905 /* prefix route could have been deleted; if so restore it */
4906 if (rc == -ENOENT) {
4907 addrconf_prefix_route(&ifp->addr, ifp->prefix_len,
4908 ifp->rt_priority, ifp->idev->dev,
4909 expires, flags, GFP_KERNEL);
4910 }
4911
4912 if (had_prefixroute && !ipv6_addr_any(&ifp->peer_addr))
4913 rc = modify_prefix_route(net, ifp, expires, flags, true);
4914
4915 if (rc == -ENOENT && !ipv6_addr_any(&ifp->peer_addr)) {
4916 addrconf_prefix_route(&ifp->peer_addr, ifp->prefix_len,
4917 ifp->rt_priority, ifp->idev->dev,
4918 expires, flags, GFP_KERNEL);
4919 }
4920 } else if (had_prefixroute) {
4921 enum cleanup_prefix_rt_t action;
4922 unsigned long rt_expires;
4923
4924 write_lock_bh(&ifp->idev->lock);
4925 action = check_cleanup_prefix_route(ifp, &rt_expires);
4926 write_unlock_bh(&ifp->idev->lock);
4927
4928 if (action != CLEANUP_PREFIX_RT_NOP) {
4929 cleanup_prefix_route(ifp, rt_expires,
4930 action == CLEANUP_PREFIX_RT_DEL, false);
4931 }
4932 }
4933
4934 if (was_managetempaddr || ifp->flags & IFA_F_MANAGETEMPADDR) {
4935 if (was_managetempaddr && !(ifp->flags & IFA_F_MANAGETEMPADDR))
4936 delete_tempaddrs(ifp->idev, ifp);
4937 else
4938 manage_tempaddrs(ifp->idev, ifp, cfg->valid_lft,
4939 cfg->preferred_lft, !was_managetempaddr,
4940 jiffies);
4941 }
4942
4943 addrconf_verify_rtnl(net);
4944
4945 return 0;
4946 }
4947
4948 static int
inet6_rtm_newaddr(struct sk_buff * skb,struct nlmsghdr * nlh,struct netlink_ext_ack * extack)4949 inet6_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh,
4950 struct netlink_ext_ack *extack)
4951 {
4952 struct net *net = sock_net(skb->sk);
4953 struct nlattr *tb[IFA_MAX+1];
4954 struct in6_addr *peer_pfx;
4955 struct inet6_ifaddr *ifa;
4956 struct net_device *dev;
4957 struct inet6_dev *idev;
4958 struct ifa6_config cfg;
4959 struct ifaddrmsg *ifm;
4960 unsigned long timeout;
4961 clock_t expires;
4962 u32 flags;
4963 int err;
4964
4965 err = nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFA_MAX,
4966 ifa_ipv6_policy, extack);
4967 if (err < 0)
4968 return err;
4969
4970 memset(&cfg, 0, sizeof(cfg));
4971
4972 ifm = nlmsg_data(nlh);
4973 cfg.pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
4974 if (!cfg.pfx)
4975 return -EINVAL;
4976
4977 cfg.peer_pfx = peer_pfx;
4978 cfg.plen = ifm->ifa_prefixlen;
4979 if (tb[IFA_RT_PRIORITY])
4980 cfg.rt_priority = nla_get_u32(tb[IFA_RT_PRIORITY]);
4981
4982 if (tb[IFA_PROTO])
4983 cfg.ifa_proto = nla_get_u8(tb[IFA_PROTO]);
4984
4985 cfg.ifa_flags = nla_get_u32_default(tb[IFA_FLAGS], ifm->ifa_flags);
4986
4987 /* We ignore other flags so far. */
4988 cfg.ifa_flags &= IFA_F_NODAD | IFA_F_HOMEADDRESS |
4989 IFA_F_MANAGETEMPADDR | IFA_F_NOPREFIXROUTE |
4990 IFA_F_MCAUTOJOIN | IFA_F_OPTIMISTIC;
4991
4992 cfg.ifa_flags |= IFA_F_PERMANENT;
4993 cfg.valid_lft = INFINITY_LIFE_TIME;
4994 cfg.preferred_lft = INFINITY_LIFE_TIME;
4995 expires = 0;
4996 flags = 0;
4997
4998 if (tb[IFA_CACHEINFO]) {
4999 struct ifa_cacheinfo *ci;
5000
5001 ci = nla_data(tb[IFA_CACHEINFO]);
5002 cfg.valid_lft = ci->ifa_valid;
5003 cfg.preferred_lft = ci->ifa_prefered;
5004
5005 if (!cfg.valid_lft || cfg.preferred_lft > cfg.valid_lft) {
5006 NL_SET_ERR_MSG_MOD(extack, "address lifetime invalid");
5007 return -EINVAL;
5008 }
5009
5010 timeout = addrconf_timeout_fixup(cfg.valid_lft, HZ);
5011 if (addrconf_finite_timeout(timeout)) {
5012 cfg.ifa_flags &= ~IFA_F_PERMANENT;
5013 cfg.valid_lft = timeout;
5014 expires = jiffies_to_clock_t(timeout * HZ);
5015 flags = RTF_EXPIRES;
5016 }
5017
5018 timeout = addrconf_timeout_fixup(cfg.preferred_lft, HZ);
5019 if (addrconf_finite_timeout(timeout)) {
5020 if (timeout == 0)
5021 cfg.ifa_flags |= IFA_F_DEPRECATED;
5022
5023 cfg.preferred_lft = timeout;
5024 }
5025 }
5026
5027 rtnl_net_lock(net);
5028
5029 dev = __dev_get_by_index(net, ifm->ifa_index);
5030 if (!dev) {
5031 NL_SET_ERR_MSG_MOD(extack, "Unable to find the interface");
5032 err = -ENODEV;
5033 goto unlock_rtnl;
5034 }
5035
5036 netdev_lock_ops(dev);
5037 idev = ipv6_find_idev(dev);
5038 if (IS_ERR(idev)) {
5039 err = PTR_ERR(idev);
5040 goto unlock;
5041 }
5042
5043 if (!ipv6_allow_optimistic_dad(net, idev))
5044 cfg.ifa_flags &= ~IFA_F_OPTIMISTIC;
5045
5046 if (cfg.ifa_flags & IFA_F_NODAD &&
5047 cfg.ifa_flags & IFA_F_OPTIMISTIC) {
5048 NL_SET_ERR_MSG(extack, "IFA_F_NODAD and IFA_F_OPTIMISTIC are mutually exclusive");
5049 err = -EINVAL;
5050 goto unlock;
5051 }
5052
5053 ifa = ipv6_get_ifaddr(net, cfg.pfx, dev, 1);
5054 if (!ifa) {
5055 /*
5056 * It would be best to check for !NLM_F_CREATE here but
5057 * userspace already relies on not having to provide this.
5058 */
5059 err = inet6_addr_add(net, dev, &cfg, expires, flags, extack);
5060 goto unlock;
5061 }
5062
5063 if (nlh->nlmsg_flags & NLM_F_EXCL ||
5064 !(nlh->nlmsg_flags & NLM_F_REPLACE)) {
5065 NL_SET_ERR_MSG_MOD(extack, "address already assigned");
5066 err = -EEXIST;
5067 } else {
5068 err = inet6_addr_modify(net, ifa, &cfg, expires, flags);
5069 }
5070
5071 in6_ifa_put(ifa);
5072 unlock:
5073 netdev_unlock_ops(dev);
5074 unlock_rtnl:
5075 rtnl_net_unlock(net);
5076
5077 return err;
5078 }
5079
put_ifaddrmsg(struct nlmsghdr * nlh,u8 prefixlen,u32 flags,u8 scope,int ifindex)5080 static void put_ifaddrmsg(struct nlmsghdr *nlh, u8 prefixlen, u32 flags,
5081 u8 scope, int ifindex)
5082 {
5083 struct ifaddrmsg *ifm;
5084
5085 ifm = nlmsg_data(nlh);
5086 ifm->ifa_family = AF_INET6;
5087 ifm->ifa_prefixlen = prefixlen;
5088 ifm->ifa_flags = flags;
5089 ifm->ifa_scope = scope;
5090 ifm->ifa_index = ifindex;
5091 }
5092
put_cacheinfo(struct sk_buff * skb,unsigned long cstamp,unsigned long tstamp,u32 preferred,u32 valid)5093 static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
5094 unsigned long tstamp, u32 preferred, u32 valid)
5095 {
5096 struct ifa_cacheinfo ci;
5097
5098 ci.cstamp = cstamp_delta(cstamp);
5099 ci.tstamp = cstamp_delta(tstamp);
5100 ci.ifa_prefered = preferred;
5101 ci.ifa_valid = valid;
5102
5103 return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
5104 }
5105
rt_scope(int ifa_scope)5106 static inline int rt_scope(int ifa_scope)
5107 {
5108 if (ifa_scope & IFA_HOST)
5109 return RT_SCOPE_HOST;
5110 else if (ifa_scope & IFA_LINK)
5111 return RT_SCOPE_LINK;
5112 else if (ifa_scope & IFA_SITE)
5113 return RT_SCOPE_SITE;
5114 else
5115 return RT_SCOPE_UNIVERSE;
5116 }
5117
inet6_ifaddr_msgsize(void)5118 static inline int inet6_ifaddr_msgsize(void)
5119 {
5120 return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
5121 + nla_total_size(16) /* IFA_LOCAL */
5122 + nla_total_size(16) /* IFA_ADDRESS */
5123 + nla_total_size(sizeof(struct ifa_cacheinfo))
5124 + nla_total_size(4) /* IFA_FLAGS */
5125 + nla_total_size(1) /* IFA_PROTO */
5126 + nla_total_size(4) /* IFA_RT_PRIORITY */;
5127 }
5128
inet6_fill_ifaddr(struct sk_buff * skb,const struct inet6_ifaddr * ifa,struct inet6_fill_args * args)5129 static int inet6_fill_ifaddr(struct sk_buff *skb,
5130 const struct inet6_ifaddr *ifa,
5131 struct inet6_fill_args *args)
5132 {
5133 struct nlmsghdr *nlh;
5134 u32 preferred, valid;
5135 u32 flags, priority;
5136 u8 proto;
5137
5138 nlh = nlmsg_put(skb, args->portid, args->seq, args->event,
5139 sizeof(struct ifaddrmsg), args->flags);
5140 if (!nlh)
5141 return -EMSGSIZE;
5142
5143 flags = READ_ONCE(ifa->flags);
5144 put_ifaddrmsg(nlh, ifa->prefix_len, ifa->flags, rt_scope(ifa->scope),
5145 ifa->idev->dev->ifindex);
5146
5147 if (args->netnsid >= 0 &&
5148 nla_put_s32(skb, IFA_TARGET_NETNSID, args->netnsid))
5149 goto error;
5150
5151 preferred = READ_ONCE(ifa->prefered_lft);
5152 valid = READ_ONCE(ifa->valid_lft);
5153
5154 if (!((flags & IFA_F_PERMANENT) &&
5155 (preferred == INFINITY_LIFE_TIME))) {
5156 if (preferred != INFINITY_LIFE_TIME) {
5157 long tval = (jiffies - READ_ONCE(ifa->tstamp)) / HZ;
5158
5159 if (preferred > tval)
5160 preferred -= tval;
5161 else
5162 preferred = 0;
5163 if (valid != INFINITY_LIFE_TIME) {
5164 if (valid > tval)
5165 valid -= tval;
5166 else
5167 valid = 0;
5168 }
5169 }
5170 } else {
5171 preferred = INFINITY_LIFE_TIME;
5172 valid = INFINITY_LIFE_TIME;
5173 }
5174
5175 if (!ipv6_addr_any(&ifa->peer_addr)) {
5176 if (nla_put_in6_addr(skb, IFA_LOCAL, &ifa->addr) < 0 ||
5177 nla_put_in6_addr(skb, IFA_ADDRESS, &ifa->peer_addr) < 0)
5178 goto error;
5179 } else {
5180 if (nla_put_in6_addr(skb, IFA_ADDRESS, &ifa->addr) < 0)
5181 goto error;
5182 }
5183
5184 priority = READ_ONCE(ifa->rt_priority);
5185 if (priority && nla_put_u32(skb, IFA_RT_PRIORITY, priority))
5186 goto error;
5187
5188 if (put_cacheinfo(skb, ifa->cstamp, READ_ONCE(ifa->tstamp),
5189 preferred, valid) < 0)
5190 goto error;
5191
5192 if (nla_put_u32(skb, IFA_FLAGS, flags) < 0)
5193 goto error;
5194
5195 proto = READ_ONCE(ifa->ifa_proto);
5196 if (proto && nla_put_u8(skb, IFA_PROTO, proto))
5197 goto error;
5198
5199 nlmsg_end(skb, nlh);
5200 return 0;
5201
5202 error:
5203 nlmsg_cancel(skb, nlh);
5204 return -EMSGSIZE;
5205 }
5206
inet6_fill_ifmcaddr(struct sk_buff * skb,const struct ifmcaddr6 * ifmca,struct inet6_fill_args * args)5207 int inet6_fill_ifmcaddr(struct sk_buff *skb,
5208 const struct ifmcaddr6 *ifmca,
5209 struct inet6_fill_args *args)
5210 {
5211 int ifindex = ifmca->idev->dev->ifindex;
5212 u8 scope = RT_SCOPE_UNIVERSE;
5213 struct nlmsghdr *nlh;
5214
5215 if (!args->force_rt_scope_universe &&
5216 ipv6_addr_scope(&ifmca->mca_addr) & IFA_SITE)
5217 scope = RT_SCOPE_SITE;
5218
5219 nlh = nlmsg_put(skb, args->portid, args->seq, args->event,
5220 sizeof(struct ifaddrmsg), args->flags);
5221 if (!nlh)
5222 return -EMSGSIZE;
5223
5224 if (args->netnsid >= 0 &&
5225 nla_put_s32(skb, IFA_TARGET_NETNSID, args->netnsid)) {
5226 nlmsg_cancel(skb, nlh);
5227 return -EMSGSIZE;
5228 }
5229
5230 put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
5231 if (nla_put_in6_addr(skb, IFA_MULTICAST, &ifmca->mca_addr) < 0 ||
5232 put_cacheinfo(skb, ifmca->mca_cstamp, READ_ONCE(ifmca->mca_tstamp),
5233 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
5234 nlmsg_cancel(skb, nlh);
5235 return -EMSGSIZE;
5236 }
5237
5238 nlmsg_end(skb, nlh);
5239 return 0;
5240 }
5241
inet6_fill_ifacaddr(struct sk_buff * skb,const struct ifacaddr6 * ifaca,struct inet6_fill_args * args)5242 int inet6_fill_ifacaddr(struct sk_buff *skb,
5243 const struct ifacaddr6 *ifaca,
5244 struct inet6_fill_args *args)
5245 {
5246 struct net_device *dev = fib6_info_nh_dev(ifaca->aca_rt);
5247 int ifindex = dev ? dev->ifindex : 1;
5248 u8 scope = RT_SCOPE_UNIVERSE;
5249 struct nlmsghdr *nlh;
5250
5251 if (ipv6_addr_scope(&ifaca->aca_addr) & IFA_SITE)
5252 scope = RT_SCOPE_SITE;
5253
5254 nlh = nlmsg_put(skb, args->portid, args->seq, args->event,
5255 sizeof(struct ifaddrmsg), args->flags);
5256 if (!nlh)
5257 return -EMSGSIZE;
5258
5259 if (args->netnsid >= 0 &&
5260 nla_put_s32(skb, IFA_TARGET_NETNSID, args->netnsid)) {
5261 nlmsg_cancel(skb, nlh);
5262 return -EMSGSIZE;
5263 }
5264
5265 put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
5266 if (nla_put_in6_addr(skb, IFA_ANYCAST, &ifaca->aca_addr) < 0 ||
5267 put_cacheinfo(skb, ifaca->aca_cstamp, READ_ONCE(ifaca->aca_tstamp),
5268 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
5269 nlmsg_cancel(skb, nlh);
5270 return -EMSGSIZE;
5271 }
5272
5273 nlmsg_end(skb, nlh);
5274 return 0;
5275 }
5276
5277 /* called with rcu_read_lock() */
in6_dump_addrs(const struct inet6_dev * idev,struct sk_buff * skb,struct netlink_callback * cb,int * s_ip_idx,struct inet6_fill_args * fillargs)5278 static int in6_dump_addrs(const struct inet6_dev *idev, struct sk_buff *skb,
5279 struct netlink_callback *cb, int *s_ip_idx,
5280 struct inet6_fill_args *fillargs)
5281 {
5282 const struct ifmcaddr6 *ifmca;
5283 const struct ifacaddr6 *ifaca;
5284 int ip_idx = 0;
5285 int err = 0;
5286
5287 switch (fillargs->type) {
5288 case UNICAST_ADDR: {
5289 const struct inet6_ifaddr *ifa;
5290 fillargs->event = RTM_NEWADDR;
5291
5292 /* unicast address incl. temp addr */
5293 list_for_each_entry_rcu(ifa, &idev->addr_list, if_list) {
5294 if (ip_idx < *s_ip_idx)
5295 goto next;
5296 err = inet6_fill_ifaddr(skb, ifa, fillargs);
5297 if (err < 0)
5298 break;
5299 nl_dump_check_consistent(cb, nlmsg_hdr(skb));
5300 next:
5301 ip_idx++;
5302 }
5303 break;
5304 }
5305 case MULTICAST_ADDR:
5306 fillargs->event = RTM_GETMULTICAST;
5307
5308 /* multicast address */
5309 for (ifmca = rcu_dereference(idev->mc_list);
5310 ifmca;
5311 ifmca = rcu_dereference(ifmca->next), ip_idx++) {
5312 if (ip_idx < *s_ip_idx)
5313 continue;
5314 err = inet6_fill_ifmcaddr(skb, ifmca, fillargs);
5315 if (err < 0)
5316 break;
5317 }
5318 break;
5319 case ANYCAST_ADDR:
5320 fillargs->event = RTM_GETANYCAST;
5321 /* anycast address */
5322 for (ifaca = rcu_dereference(idev->ac_list); ifaca;
5323 ifaca = rcu_dereference(ifaca->aca_next), ip_idx++) {
5324 if (ip_idx < *s_ip_idx)
5325 continue;
5326 err = inet6_fill_ifacaddr(skb, ifaca, fillargs);
5327 if (err < 0)
5328 break;
5329 }
5330 break;
5331 default:
5332 break;
5333 }
5334 *s_ip_idx = err ? ip_idx : 0;
5335 return err;
5336 }
5337
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)5338 static int inet6_valid_dump_ifaddr_req(const struct nlmsghdr *nlh,
5339 struct inet6_fill_args *fillargs,
5340 struct net **tgt_net, struct sock *sk,
5341 struct netlink_callback *cb)
5342 {
5343 struct netlink_ext_ack *extack = cb->extack;
5344 struct nlattr *tb[IFA_MAX+1];
5345 struct ifaddrmsg *ifm;
5346 int err, i;
5347
5348 ifm = nlmsg_payload(nlh, sizeof(*ifm));
5349 if (!ifm) {
5350 NL_SET_ERR_MSG_MOD(extack, "Invalid header for address dump request");
5351 return -EINVAL;
5352 }
5353
5354 if (ifm->ifa_prefixlen || ifm->ifa_flags || ifm->ifa_scope) {
5355 NL_SET_ERR_MSG_MOD(extack, "Invalid values in header for address dump request");
5356 return -EINVAL;
5357 }
5358
5359 fillargs->ifindex = ifm->ifa_index;
5360 if (fillargs->ifindex) {
5361 cb->answer_flags |= NLM_F_DUMP_FILTERED;
5362 fillargs->flags |= NLM_F_DUMP_FILTERED;
5363 }
5364
5365 err = nlmsg_parse_deprecated_strict(nlh, sizeof(*ifm), tb, IFA_MAX,
5366 ifa_ipv6_policy, extack);
5367 if (err < 0)
5368 return err;
5369
5370 for (i = 0; i <= IFA_MAX; ++i) {
5371 if (!tb[i])
5372 continue;
5373
5374 if (i == IFA_TARGET_NETNSID) {
5375 struct net *net;
5376
5377 fillargs->netnsid = nla_get_s32(tb[i]);
5378 net = rtnl_get_net_ns_capable(sk, fillargs->netnsid);
5379 if (IS_ERR(net)) {
5380 fillargs->netnsid = -1;
5381 NL_SET_ERR_MSG_MOD(extack, "Invalid target network namespace id");
5382 return PTR_ERR(net);
5383 }
5384 *tgt_net = net;
5385 } else {
5386 NL_SET_ERR_MSG_MOD(extack, "Unsupported attribute in dump request");
5387 return -EINVAL;
5388 }
5389 }
5390
5391 return 0;
5392 }
5393
inet6_dump_addr(struct sk_buff * skb,struct netlink_callback * cb,enum addr_type_t type)5394 static int inet6_dump_addr(struct sk_buff *skb, struct netlink_callback *cb,
5395 enum addr_type_t type)
5396 {
5397 struct net *tgt_net = sock_net(skb->sk);
5398 const struct nlmsghdr *nlh = cb->nlh;
5399 struct inet6_fill_args fillargs = {
5400 .portid = NETLINK_CB(cb->skb).portid,
5401 .seq = cb->nlh->nlmsg_seq,
5402 .flags = NLM_F_MULTI,
5403 .netnsid = -1,
5404 .type = type,
5405 .force_rt_scope_universe = false,
5406 };
5407 struct {
5408 unsigned long ifindex;
5409 int ip_idx;
5410 } *ctx = (void *)cb->ctx;
5411 struct net_device *dev;
5412 struct inet6_dev *idev;
5413 int err = 0;
5414
5415 rcu_read_lock();
5416 if (cb->strict_check) {
5417 err = inet6_valid_dump_ifaddr_req(nlh, &fillargs, &tgt_net,
5418 skb->sk, cb);
5419 if (err < 0)
5420 goto done;
5421
5422 err = 0;
5423 if (fillargs.ifindex) {
5424 dev = dev_get_by_index_rcu(tgt_net, fillargs.ifindex);
5425 if (!dev) {
5426 err = -ENODEV;
5427 goto done;
5428 }
5429 idev = __in6_dev_get(dev);
5430 if (idev)
5431 err = in6_dump_addrs(idev, skb, cb,
5432 &ctx->ip_idx,
5433 &fillargs);
5434 goto done;
5435 }
5436 }
5437
5438 cb->seq = inet6_base_seq(tgt_net);
5439 for_each_netdev_dump(tgt_net, dev, ctx->ifindex) {
5440 idev = __in6_dev_get(dev);
5441 if (!idev)
5442 continue;
5443 err = in6_dump_addrs(idev, skb, cb, &ctx->ip_idx,
5444 &fillargs);
5445 if (err < 0)
5446 goto done;
5447 }
5448 done:
5449 rcu_read_unlock();
5450 if (fillargs.netnsid >= 0)
5451 put_net(tgt_net);
5452
5453 return err;
5454 }
5455
inet6_dump_ifaddr(struct sk_buff * skb,struct netlink_callback * cb)5456 static int inet6_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
5457 {
5458 enum addr_type_t type = UNICAST_ADDR;
5459
5460 return inet6_dump_addr(skb, cb, type);
5461 }
5462
inet6_dump_ifmcaddr(struct sk_buff * skb,struct netlink_callback * cb)5463 static int inet6_dump_ifmcaddr(struct sk_buff *skb, struct netlink_callback *cb)
5464 {
5465 enum addr_type_t type = MULTICAST_ADDR;
5466
5467 return inet6_dump_addr(skb, cb, type);
5468 }
5469
5470
inet6_dump_ifacaddr(struct sk_buff * skb,struct netlink_callback * cb)5471 static int inet6_dump_ifacaddr(struct sk_buff *skb, struct netlink_callback *cb)
5472 {
5473 enum addr_type_t type = ANYCAST_ADDR;
5474
5475 return inet6_dump_addr(skb, cb, type);
5476 }
5477
inet6_rtm_valid_getaddr_req(struct sk_buff * skb,const struct nlmsghdr * nlh,struct nlattr ** tb,struct netlink_ext_ack * extack)5478 static int inet6_rtm_valid_getaddr_req(struct sk_buff *skb,
5479 const struct nlmsghdr *nlh,
5480 struct nlattr **tb,
5481 struct netlink_ext_ack *extack)
5482 {
5483 struct ifaddrmsg *ifm;
5484 int i, err;
5485
5486 ifm = nlmsg_payload(nlh, sizeof(*ifm));
5487 if (!ifm) {
5488 NL_SET_ERR_MSG_MOD(extack, "Invalid header for get address request");
5489 return -EINVAL;
5490 }
5491
5492 if (!netlink_strict_get_check(skb))
5493 return nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFA_MAX,
5494 ifa_ipv6_policy, extack);
5495
5496 if (ifm->ifa_prefixlen || ifm->ifa_flags || ifm->ifa_scope) {
5497 NL_SET_ERR_MSG_MOD(extack, "Invalid values in header for get address request");
5498 return -EINVAL;
5499 }
5500
5501 err = nlmsg_parse_deprecated_strict(nlh, sizeof(*ifm), tb, IFA_MAX,
5502 ifa_ipv6_policy, extack);
5503 if (err)
5504 return err;
5505
5506 for (i = 0; i <= IFA_MAX; i++) {
5507 if (!tb[i])
5508 continue;
5509
5510 switch (i) {
5511 case IFA_TARGET_NETNSID:
5512 case IFA_ADDRESS:
5513 case IFA_LOCAL:
5514 break;
5515 default:
5516 NL_SET_ERR_MSG_MOD(extack, "Unsupported attribute in get address request");
5517 return -EINVAL;
5518 }
5519 }
5520
5521 return 0;
5522 }
5523
inet6_rtm_getaddr(struct sk_buff * in_skb,struct nlmsghdr * nlh,struct netlink_ext_ack * extack)5524 static int inet6_rtm_getaddr(struct sk_buff *in_skb, struct nlmsghdr *nlh,
5525 struct netlink_ext_ack *extack)
5526 {
5527 struct net *tgt_net = sock_net(in_skb->sk);
5528 struct inet6_fill_args fillargs = {
5529 .portid = NETLINK_CB(in_skb).portid,
5530 .seq = nlh->nlmsg_seq,
5531 .event = RTM_NEWADDR,
5532 .flags = 0,
5533 .netnsid = -1,
5534 .force_rt_scope_universe = false,
5535 };
5536 struct ifaddrmsg *ifm;
5537 struct nlattr *tb[IFA_MAX+1];
5538 struct in6_addr *addr = NULL, *peer;
5539 struct net_device *dev = NULL;
5540 struct inet6_ifaddr *ifa;
5541 struct sk_buff *skb;
5542 int err;
5543
5544 err = inet6_rtm_valid_getaddr_req(in_skb, nlh, tb, extack);
5545 if (err < 0)
5546 return err;
5547
5548 if (tb[IFA_TARGET_NETNSID]) {
5549 fillargs.netnsid = nla_get_s32(tb[IFA_TARGET_NETNSID]);
5550
5551 tgt_net = rtnl_get_net_ns_capable(NETLINK_CB(in_skb).sk,
5552 fillargs.netnsid);
5553 if (IS_ERR(tgt_net))
5554 return PTR_ERR(tgt_net);
5555 }
5556
5557 addr = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer);
5558 if (!addr) {
5559 err = -EINVAL;
5560 goto errout;
5561 }
5562 ifm = nlmsg_data(nlh);
5563 if (ifm->ifa_index)
5564 dev = dev_get_by_index(tgt_net, ifm->ifa_index);
5565
5566 ifa = ipv6_get_ifaddr(tgt_net, addr, dev, 1);
5567 if (!ifa) {
5568 err = -EADDRNOTAVAIL;
5569 goto errout;
5570 }
5571
5572 skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_KERNEL);
5573 if (!skb) {
5574 err = -ENOBUFS;
5575 goto errout_ifa;
5576 }
5577
5578 err = inet6_fill_ifaddr(skb, ifa, &fillargs);
5579 if (err < 0) {
5580 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
5581 WARN_ON(err == -EMSGSIZE);
5582 kfree_skb(skb);
5583 goto errout_ifa;
5584 }
5585 err = rtnl_unicast(skb, tgt_net, NETLINK_CB(in_skb).portid);
5586 errout_ifa:
5587 in6_ifa_put(ifa);
5588 errout:
5589 dev_put(dev);
5590 if (fillargs.netnsid >= 0)
5591 put_net(tgt_net);
5592
5593 return err;
5594 }
5595
inet6_ifa_notify(int event,struct inet6_ifaddr * ifa)5596 static void inet6_ifa_notify(int event, struct inet6_ifaddr *ifa)
5597 {
5598 struct sk_buff *skb;
5599 struct net *net = dev_net(ifa->idev->dev);
5600 struct inet6_fill_args fillargs = {
5601 .portid = 0,
5602 .seq = 0,
5603 .event = event,
5604 .flags = 0,
5605 .netnsid = -1,
5606 .force_rt_scope_universe = false,
5607 };
5608 int err = -ENOBUFS;
5609
5610 skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_ATOMIC);
5611 if (!skb)
5612 goto errout;
5613
5614 err = inet6_fill_ifaddr(skb, ifa, &fillargs);
5615 if (err < 0) {
5616 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
5617 WARN_ON(err == -EMSGSIZE);
5618 kfree_skb(skb);
5619 goto errout;
5620 }
5621 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
5622 return;
5623 errout:
5624 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
5625 }
5626
ipv6_store_devconf(const struct ipv6_devconf * cnf,__s32 * array,int bytes)5627 static void ipv6_store_devconf(const struct ipv6_devconf *cnf,
5628 __s32 *array, int bytes)
5629 {
5630 BUG_ON(bytes < (DEVCONF_MAX * 4));
5631
5632 memset(array, 0, bytes);
5633 array[DEVCONF_FORWARDING] = READ_ONCE(cnf->forwarding);
5634 array[DEVCONF_HOPLIMIT] = READ_ONCE(cnf->hop_limit);
5635 array[DEVCONF_MTU6] = READ_ONCE(cnf->mtu6);
5636 array[DEVCONF_ACCEPT_RA] = READ_ONCE(cnf->accept_ra);
5637 array[DEVCONF_ACCEPT_REDIRECTS] = READ_ONCE(cnf->accept_redirects);
5638 array[DEVCONF_AUTOCONF] = READ_ONCE(cnf->autoconf);
5639 array[DEVCONF_DAD_TRANSMITS] = READ_ONCE(cnf->dad_transmits);
5640 array[DEVCONF_RTR_SOLICITS] = READ_ONCE(cnf->rtr_solicits);
5641 array[DEVCONF_RTR_SOLICIT_INTERVAL] =
5642 jiffies_to_msecs(READ_ONCE(cnf->rtr_solicit_interval));
5643 array[DEVCONF_RTR_SOLICIT_MAX_INTERVAL] =
5644 jiffies_to_msecs(READ_ONCE(cnf->rtr_solicit_max_interval));
5645 array[DEVCONF_RTR_SOLICIT_DELAY] =
5646 jiffies_to_msecs(READ_ONCE(cnf->rtr_solicit_delay));
5647 array[DEVCONF_FORCE_MLD_VERSION] = READ_ONCE(cnf->force_mld_version);
5648 array[DEVCONF_MLDV1_UNSOLICITED_REPORT_INTERVAL] =
5649 jiffies_to_msecs(READ_ONCE(cnf->mldv1_unsolicited_report_interval));
5650 array[DEVCONF_MLDV2_UNSOLICITED_REPORT_INTERVAL] =
5651 jiffies_to_msecs(READ_ONCE(cnf->mldv2_unsolicited_report_interval));
5652 array[DEVCONF_USE_TEMPADDR] = READ_ONCE(cnf->use_tempaddr);
5653 array[DEVCONF_TEMP_VALID_LFT] = READ_ONCE(cnf->temp_valid_lft);
5654 array[DEVCONF_TEMP_PREFERED_LFT] = READ_ONCE(cnf->temp_prefered_lft);
5655 array[DEVCONF_REGEN_MAX_RETRY] = READ_ONCE(cnf->regen_max_retry);
5656 array[DEVCONF_MAX_DESYNC_FACTOR] = READ_ONCE(cnf->max_desync_factor);
5657 array[DEVCONF_MAX_ADDRESSES] = READ_ONCE(cnf->max_addresses);
5658 array[DEVCONF_ACCEPT_RA_DEFRTR] = READ_ONCE(cnf->accept_ra_defrtr);
5659 array[DEVCONF_RA_DEFRTR_METRIC] = READ_ONCE(cnf->ra_defrtr_metric);
5660 array[DEVCONF_ACCEPT_RA_MIN_HOP_LIMIT] =
5661 READ_ONCE(cnf->accept_ra_min_hop_limit);
5662 array[DEVCONF_ACCEPT_RA_PINFO] = READ_ONCE(cnf->accept_ra_pinfo);
5663 #ifdef CONFIG_IPV6_ROUTER_PREF
5664 array[DEVCONF_ACCEPT_RA_RTR_PREF] = READ_ONCE(cnf->accept_ra_rtr_pref);
5665 array[DEVCONF_RTR_PROBE_INTERVAL] =
5666 jiffies_to_msecs(READ_ONCE(cnf->rtr_probe_interval));
5667 #ifdef CONFIG_IPV6_ROUTE_INFO
5668 array[DEVCONF_ACCEPT_RA_RT_INFO_MIN_PLEN] =
5669 READ_ONCE(cnf->accept_ra_rt_info_min_plen);
5670 array[DEVCONF_ACCEPT_RA_RT_INFO_MAX_PLEN] =
5671 READ_ONCE(cnf->accept_ra_rt_info_max_plen);
5672 #endif
5673 #endif
5674 array[DEVCONF_PROXY_NDP] = READ_ONCE(cnf->proxy_ndp);
5675 array[DEVCONF_ACCEPT_SOURCE_ROUTE] =
5676 READ_ONCE(cnf->accept_source_route);
5677 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
5678 array[DEVCONF_OPTIMISTIC_DAD] = READ_ONCE(cnf->optimistic_dad);
5679 array[DEVCONF_USE_OPTIMISTIC] = READ_ONCE(cnf->use_optimistic);
5680 #endif
5681 #ifdef CONFIG_IPV6_MROUTE
5682 array[DEVCONF_MC_FORWARDING] = atomic_read(&cnf->mc_forwarding);
5683 #endif
5684 array[DEVCONF_DISABLE_IPV6] = READ_ONCE(cnf->disable_ipv6);
5685 array[DEVCONF_ACCEPT_DAD] = READ_ONCE(cnf->accept_dad);
5686 array[DEVCONF_FORCE_TLLAO] = READ_ONCE(cnf->force_tllao);
5687 array[DEVCONF_NDISC_NOTIFY] = READ_ONCE(cnf->ndisc_notify);
5688 array[DEVCONF_SUPPRESS_FRAG_NDISC] =
5689 READ_ONCE(cnf->suppress_frag_ndisc);
5690 array[DEVCONF_ACCEPT_RA_FROM_LOCAL] =
5691 READ_ONCE(cnf->accept_ra_from_local);
5692 array[DEVCONF_ACCEPT_RA_MTU] = READ_ONCE(cnf->accept_ra_mtu);
5693 array[DEVCONF_IGNORE_ROUTES_WITH_LINKDOWN] =
5694 READ_ONCE(cnf->ignore_routes_with_linkdown);
5695 /* we omit DEVCONF_STABLE_SECRET for now */
5696 array[DEVCONF_USE_OIF_ADDRS_ONLY] = READ_ONCE(cnf->use_oif_addrs_only);
5697 array[DEVCONF_DROP_UNICAST_IN_L2_MULTICAST] =
5698 READ_ONCE(cnf->drop_unicast_in_l2_multicast);
5699 array[DEVCONF_DROP_UNSOLICITED_NA] = READ_ONCE(cnf->drop_unsolicited_na);
5700 array[DEVCONF_KEEP_ADDR_ON_DOWN] = READ_ONCE(cnf->keep_addr_on_down);
5701 array[DEVCONF_SEG6_ENABLED] = READ_ONCE(cnf->seg6_enabled);
5702 #ifdef CONFIG_IPV6_SEG6_HMAC
5703 array[DEVCONF_SEG6_REQUIRE_HMAC] = READ_ONCE(cnf->seg6_require_hmac);
5704 #endif
5705 array[DEVCONF_ENHANCED_DAD] = READ_ONCE(cnf->enhanced_dad);
5706 array[DEVCONF_ADDR_GEN_MODE] = READ_ONCE(cnf->addr_gen_mode);
5707 array[DEVCONF_DISABLE_POLICY] = READ_ONCE(cnf->disable_policy);
5708 array[DEVCONF_NDISC_TCLASS] = READ_ONCE(cnf->ndisc_tclass);
5709 array[DEVCONF_RPL_SEG_ENABLED] = READ_ONCE(cnf->rpl_seg_enabled);
5710 array[DEVCONF_IOAM6_ENABLED] = READ_ONCE(cnf->ioam6_enabled);
5711 array[DEVCONF_IOAM6_ID] = READ_ONCE(cnf->ioam6_id);
5712 array[DEVCONF_IOAM6_ID_WIDE] = READ_ONCE(cnf->ioam6_id_wide);
5713 array[DEVCONF_NDISC_EVICT_NOCARRIER] =
5714 READ_ONCE(cnf->ndisc_evict_nocarrier);
5715 array[DEVCONF_ACCEPT_UNTRACKED_NA] =
5716 READ_ONCE(cnf->accept_untracked_na);
5717 array[DEVCONF_ACCEPT_RA_MIN_LFT] = READ_ONCE(cnf->accept_ra_min_lft);
5718 array[DEVCONF_FORCE_FORWARDING] = READ_ONCE(cnf->force_forwarding);
5719 }
5720
inet6_ifla6_size(void)5721 static inline size_t inet6_ifla6_size(void)
5722 {
5723 return nla_total_size(4) /* IFLA_INET6_FLAGS */
5724 + nla_total_size(sizeof(struct ifla_cacheinfo))
5725 + nla_total_size(DEVCONF_MAX * 4) /* IFLA_INET6_CONF */
5726 + nla_total_size(IPSTATS_MIB_MAX * 8) /* IFLA_INET6_STATS */
5727 + nla_total_size(ICMP6_MIB_MAX * 8) /* IFLA_INET6_ICMP6STATS */
5728 + nla_total_size(sizeof(struct in6_addr)) /* IFLA_INET6_TOKEN */
5729 + nla_total_size(1) /* IFLA_INET6_ADDR_GEN_MODE */
5730 + nla_total_size(4) /* IFLA_INET6_RA_MTU */
5731 + 0;
5732 }
5733
inet6_if_nlmsg_size(void)5734 static inline size_t inet6_if_nlmsg_size(void)
5735 {
5736 return NLMSG_ALIGN(sizeof(struct ifinfomsg))
5737 + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
5738 + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
5739 + nla_total_size(4) /* IFLA_MTU */
5740 + nla_total_size(4) /* IFLA_LINK */
5741 + nla_total_size(1) /* IFLA_OPERSTATE */
5742 + nla_total_size(inet6_ifla6_size()); /* IFLA_PROTINFO */
5743 }
5744
__snmp6_fill_statsdev(u64 * stats,atomic_long_t * mib,int bytes)5745 static inline void __snmp6_fill_statsdev(u64 *stats, atomic_long_t *mib,
5746 int bytes)
5747 {
5748 int i;
5749 int pad = bytes - sizeof(u64) * ICMP6_MIB_MAX;
5750 BUG_ON(pad < 0);
5751
5752 /* Use put_unaligned() because stats may not be aligned for u64. */
5753 put_unaligned(ICMP6_MIB_MAX, &stats[0]);
5754 for (i = 1; i < ICMP6_MIB_MAX; i++)
5755 put_unaligned(atomic_long_read(&mib[i]), &stats[i]);
5756
5757 memset(&stats[ICMP6_MIB_MAX], 0, pad);
5758 }
5759
__snmp6_fill_stats64(u64 * stats,void __percpu * mib,int bytes,size_t syncpoff)5760 static inline void __snmp6_fill_stats64(u64 *stats, void __percpu *mib,
5761 int bytes, size_t syncpoff)
5762 {
5763 int i, c;
5764 u64 buff[IPSTATS_MIB_MAX];
5765 int pad = bytes - sizeof(u64) * IPSTATS_MIB_MAX;
5766
5767 BUG_ON(pad < 0);
5768
5769 memset(buff, 0, sizeof(buff));
5770 buff[0] = IPSTATS_MIB_MAX;
5771
5772 for_each_possible_cpu(c) {
5773 for (i = 1; i < IPSTATS_MIB_MAX; i++)
5774 buff[i] += snmp_get_cpu_field64(mib, c, i, syncpoff);
5775 }
5776
5777 memcpy(stats, buff, IPSTATS_MIB_MAX * sizeof(u64));
5778 memset(&stats[IPSTATS_MIB_MAX], 0, pad);
5779 }
5780
snmp6_fill_stats(u64 * stats,struct inet6_dev * idev,int attrtype,int bytes)5781 static void snmp6_fill_stats(u64 *stats, struct inet6_dev *idev, int attrtype,
5782 int bytes)
5783 {
5784 switch (attrtype) {
5785 case IFLA_INET6_STATS:
5786 __snmp6_fill_stats64(stats, idev->stats.ipv6, bytes,
5787 offsetof(struct ipstats_mib, syncp));
5788 break;
5789 case IFLA_INET6_ICMP6STATS:
5790 __snmp6_fill_statsdev(stats, idev->stats.icmpv6dev->mibs, bytes);
5791 break;
5792 }
5793 }
5794
inet6_fill_ifla6_stats_attrs(struct sk_buff * skb,struct inet6_dev * idev)5795 static int inet6_fill_ifla6_stats_attrs(struct sk_buff *skb,
5796 struct inet6_dev *idev)
5797 {
5798 struct nlattr *nla;
5799
5800 nla = nla_reserve(skb, IFLA_INET6_STATS, IPSTATS_MIB_MAX * sizeof(u64));
5801 if (!nla)
5802 goto nla_put_failure;
5803 snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_STATS, nla_len(nla));
5804
5805 nla = nla_reserve(skb, IFLA_INET6_ICMP6STATS, ICMP6_MIB_MAX * sizeof(u64));
5806 if (!nla)
5807 goto nla_put_failure;
5808 snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_ICMP6STATS, nla_len(nla));
5809
5810 return 0;
5811
5812 nla_put_failure:
5813 return -EMSGSIZE;
5814 }
5815
inet6_fill_ifla6_attrs(struct sk_buff * skb,struct inet6_dev * idev,u32 ext_filter_mask)5816 static int inet6_fill_ifla6_attrs(struct sk_buff *skb, struct inet6_dev *idev,
5817 u32 ext_filter_mask)
5818 {
5819 struct ifla_cacheinfo ci;
5820 struct nlattr *nla;
5821 u32 ra_mtu;
5822
5823 if (nla_put_u32(skb, IFLA_INET6_FLAGS, READ_ONCE(idev->if_flags)))
5824 goto nla_put_failure;
5825 ci.max_reasm_len = IPV6_MAXPLEN;
5826 ci.tstamp = cstamp_delta(READ_ONCE(idev->tstamp));
5827 ci.reachable_time = jiffies_to_msecs(idev->nd_parms->reachable_time);
5828 ci.retrans_time = jiffies_to_msecs(NEIGH_VAR(idev->nd_parms, RETRANS_TIME));
5829 if (nla_put(skb, IFLA_INET6_CACHEINFO, sizeof(ci), &ci))
5830 goto nla_put_failure;
5831 nla = nla_reserve(skb, IFLA_INET6_CONF, DEVCONF_MAX * sizeof(s32));
5832 if (!nla)
5833 goto nla_put_failure;
5834 ipv6_store_devconf(&idev->cnf, nla_data(nla), nla_len(nla));
5835
5836 /* XXX - MC not implemented */
5837
5838 if (!(ext_filter_mask & RTEXT_FILTER_SKIP_STATS)) {
5839 if (inet6_fill_ifla6_stats_attrs(skb, idev) < 0)
5840 goto nla_put_failure;
5841 }
5842
5843 nla = nla_reserve(skb, IFLA_INET6_TOKEN, sizeof(struct in6_addr));
5844 if (!nla)
5845 goto nla_put_failure;
5846 read_lock_bh(&idev->lock);
5847 memcpy(nla_data(nla), idev->token.s6_addr, nla_len(nla));
5848 read_unlock_bh(&idev->lock);
5849
5850 if (nla_put_u8(skb, IFLA_INET6_ADDR_GEN_MODE,
5851 READ_ONCE(idev->cnf.addr_gen_mode)))
5852 goto nla_put_failure;
5853
5854 ra_mtu = READ_ONCE(idev->ra_mtu);
5855 if (ra_mtu && nla_put_u32(skb, IFLA_INET6_RA_MTU, ra_mtu))
5856 goto nla_put_failure;
5857
5858 return 0;
5859
5860 nla_put_failure:
5861 return -EMSGSIZE;
5862 }
5863
inet6_get_link_af_size(const struct net_device * dev,u32 ext_filter_mask)5864 static size_t inet6_get_link_af_size(const struct net_device *dev,
5865 u32 ext_filter_mask)
5866 {
5867 if (!__in6_dev_get(dev))
5868 return 0;
5869
5870 return inet6_ifla6_size();
5871 }
5872
inet6_fill_link_af(struct sk_buff * skb,const struct net_device * dev,u32 ext_filter_mask)5873 static int inet6_fill_link_af(struct sk_buff *skb, const struct net_device *dev,
5874 u32 ext_filter_mask)
5875 {
5876 struct inet6_dev *idev = __in6_dev_get(dev);
5877
5878 if (!idev)
5879 return -ENODATA;
5880
5881 if (inet6_fill_ifla6_attrs(skb, idev, ext_filter_mask) < 0)
5882 return -EMSGSIZE;
5883
5884 return 0;
5885 }
5886
inet6_set_iftoken(struct inet6_dev * idev,struct in6_addr * token,struct netlink_ext_ack * extack)5887 static int inet6_set_iftoken(struct inet6_dev *idev, struct in6_addr *token,
5888 struct netlink_ext_ack *extack)
5889 {
5890 struct inet6_ifaddr *ifp;
5891 struct net_device *dev = idev->dev;
5892 bool clear_token, update_rs = false;
5893 struct in6_addr ll_addr;
5894
5895 ASSERT_RTNL();
5896
5897 if (!token)
5898 return -EINVAL;
5899
5900 if (dev->flags & IFF_LOOPBACK) {
5901 NL_SET_ERR_MSG_MOD(extack, "Device is loopback");
5902 return -EINVAL;
5903 }
5904
5905 if (dev->flags & IFF_NOARP) {
5906 NL_SET_ERR_MSG_MOD(extack,
5907 "Device does not do neighbour discovery");
5908 return -EINVAL;
5909 }
5910
5911 if (!ipv6_accept_ra(idev)) {
5912 NL_SET_ERR_MSG_MOD(extack,
5913 "Router advertisement is disabled on device");
5914 return -EINVAL;
5915 }
5916
5917 if (READ_ONCE(idev->cnf.rtr_solicits) == 0) {
5918 NL_SET_ERR_MSG(extack,
5919 "Router solicitation is disabled on device");
5920 return -EINVAL;
5921 }
5922
5923 write_lock_bh(&idev->lock);
5924
5925 BUILD_BUG_ON(sizeof(token->s6_addr) != 16);
5926 memcpy(idev->token.s6_addr + 8, token->s6_addr + 8, 8);
5927
5928 write_unlock_bh(&idev->lock);
5929
5930 clear_token = ipv6_addr_any(token);
5931 if (clear_token)
5932 goto update_lft;
5933
5934 if (!idev->dead && (idev->if_flags & IF_READY) &&
5935 !ipv6_get_lladdr(dev, &ll_addr, IFA_F_TENTATIVE |
5936 IFA_F_OPTIMISTIC)) {
5937 /* If we're not ready, then normal ifup will take care
5938 * of this. Otherwise, we need to request our rs here.
5939 */
5940 ndisc_send_rs(dev, &ll_addr, &in6addr_linklocal_allrouters);
5941 update_rs = true;
5942 }
5943
5944 update_lft:
5945 write_lock_bh(&idev->lock);
5946
5947 if (update_rs) {
5948 idev->if_flags |= IF_RS_SENT;
5949 idev->rs_interval = rfc3315_s14_backoff_init(
5950 READ_ONCE(idev->cnf.rtr_solicit_interval));
5951 idev->rs_probes = 1;
5952 addrconf_mod_rs_timer(idev, idev->rs_interval);
5953 }
5954
5955 /* Well, that's kinda nasty ... */
5956 list_for_each_entry(ifp, &idev->addr_list, if_list) {
5957 spin_lock(&ifp->lock);
5958 if (ifp->tokenized) {
5959 ifp->valid_lft = 0;
5960 ifp->prefered_lft = 0;
5961 }
5962 spin_unlock(&ifp->lock);
5963 }
5964
5965 write_unlock_bh(&idev->lock);
5966 inet6_ifinfo_notify(RTM_NEWLINK, idev);
5967 addrconf_verify_rtnl(dev_net(dev));
5968 return 0;
5969 }
5970
5971 static const struct nla_policy inet6_af_policy[IFLA_INET6_MAX + 1] = {
5972 [IFLA_INET6_ADDR_GEN_MODE] = { .type = NLA_U8 },
5973 [IFLA_INET6_TOKEN] = { .len = sizeof(struct in6_addr) },
5974 [IFLA_INET6_RA_MTU] = { .type = NLA_REJECT,
5975 .reject_message =
5976 "IFLA_INET6_RA_MTU can not be set" },
5977 };
5978
check_addr_gen_mode(int mode)5979 static int check_addr_gen_mode(int mode)
5980 {
5981 if (mode != IN6_ADDR_GEN_MODE_EUI64 &&
5982 mode != IN6_ADDR_GEN_MODE_NONE &&
5983 mode != IN6_ADDR_GEN_MODE_STABLE_PRIVACY &&
5984 mode != IN6_ADDR_GEN_MODE_RANDOM)
5985 return -EINVAL;
5986 return 1;
5987 }
5988
check_stable_privacy(struct inet6_dev * idev,struct net * net,int mode)5989 static int check_stable_privacy(struct inet6_dev *idev, struct net *net,
5990 int mode)
5991 {
5992 if (mode == IN6_ADDR_GEN_MODE_STABLE_PRIVACY &&
5993 !idev->cnf.stable_secret.initialized &&
5994 !net->ipv6.devconf_dflt->stable_secret.initialized)
5995 return -EINVAL;
5996 return 1;
5997 }
5998
inet6_validate_link_af(const struct net_device * dev,const struct nlattr * nla,struct netlink_ext_ack * extack)5999 static int inet6_validate_link_af(const struct net_device *dev,
6000 const struct nlattr *nla,
6001 struct netlink_ext_ack *extack)
6002 {
6003 struct nlattr *tb[IFLA_INET6_MAX + 1];
6004 struct inet6_dev *idev = NULL;
6005 int err;
6006
6007 if (dev) {
6008 idev = __in6_dev_get(dev);
6009 if (!idev)
6010 return -EAFNOSUPPORT;
6011 }
6012
6013 err = nla_parse_nested_deprecated(tb, IFLA_INET6_MAX, nla,
6014 inet6_af_policy, extack);
6015 if (err)
6016 return err;
6017
6018 if (!tb[IFLA_INET6_TOKEN] && !tb[IFLA_INET6_ADDR_GEN_MODE])
6019 return -EINVAL;
6020
6021 if (tb[IFLA_INET6_ADDR_GEN_MODE]) {
6022 u8 mode = nla_get_u8(tb[IFLA_INET6_ADDR_GEN_MODE]);
6023
6024 if (check_addr_gen_mode(mode) < 0)
6025 return -EINVAL;
6026 if (dev && check_stable_privacy(idev, dev_net(dev), mode) < 0)
6027 return -EINVAL;
6028 }
6029
6030 return 0;
6031 }
6032
inet6_set_link_af(struct net_device * dev,const struct nlattr * nla,struct netlink_ext_ack * extack)6033 static int inet6_set_link_af(struct net_device *dev, const struct nlattr *nla,
6034 struct netlink_ext_ack *extack)
6035 {
6036 struct inet6_dev *idev = __in6_dev_get(dev);
6037 struct nlattr *tb[IFLA_INET6_MAX + 1];
6038 int err;
6039
6040 if (!idev)
6041 return -EAFNOSUPPORT;
6042
6043 if (nla_parse_nested_deprecated(tb, IFLA_INET6_MAX, nla, NULL, NULL) < 0)
6044 return -EINVAL;
6045
6046 if (tb[IFLA_INET6_TOKEN]) {
6047 err = inet6_set_iftoken(idev, nla_data(tb[IFLA_INET6_TOKEN]),
6048 extack);
6049 if (err)
6050 return err;
6051 }
6052
6053 if (tb[IFLA_INET6_ADDR_GEN_MODE]) {
6054 u8 mode = nla_get_u8(tb[IFLA_INET6_ADDR_GEN_MODE]);
6055
6056 WRITE_ONCE(idev->cnf.addr_gen_mode, mode);
6057 }
6058
6059 return 0;
6060 }
6061
inet6_fill_ifinfo(struct sk_buff * skb,struct inet6_dev * idev,u32 portid,u32 seq,int event,unsigned int flags)6062 static int inet6_fill_ifinfo(struct sk_buff *skb, struct inet6_dev *idev,
6063 u32 portid, u32 seq, int event, unsigned int flags)
6064 {
6065 struct net_device *dev = idev->dev;
6066 struct ifinfomsg *hdr;
6067 struct nlmsghdr *nlh;
6068 int ifindex, iflink;
6069 void *protoinfo;
6070
6071 nlh = nlmsg_put(skb, portid, seq, event, sizeof(*hdr), flags);
6072 if (!nlh)
6073 return -EMSGSIZE;
6074
6075 hdr = nlmsg_data(nlh);
6076 hdr->ifi_family = AF_INET6;
6077 hdr->__ifi_pad = 0;
6078 hdr->ifi_type = dev->type;
6079 ifindex = READ_ONCE(dev->ifindex);
6080 hdr->ifi_index = ifindex;
6081 hdr->ifi_flags = netif_get_flags(dev);
6082 hdr->ifi_change = 0;
6083
6084 iflink = dev_get_iflink(dev);
6085 if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
6086 (dev->addr_len &&
6087 nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
6088 nla_put_u32(skb, IFLA_MTU, READ_ONCE(dev->mtu)) ||
6089 (ifindex != iflink &&
6090 nla_put_u32(skb, IFLA_LINK, iflink)) ||
6091 nla_put_u8(skb, IFLA_OPERSTATE,
6092 netif_running(dev) ? READ_ONCE(dev->operstate) : IF_OPER_DOWN))
6093 goto nla_put_failure;
6094 protoinfo = nla_nest_start_noflag(skb, IFLA_PROTINFO);
6095 if (!protoinfo)
6096 goto nla_put_failure;
6097
6098 if (inet6_fill_ifla6_attrs(skb, idev, 0) < 0)
6099 goto nla_put_failure;
6100
6101 nla_nest_end(skb, protoinfo);
6102 nlmsg_end(skb, nlh);
6103 return 0;
6104
6105 nla_put_failure:
6106 nlmsg_cancel(skb, nlh);
6107 return -EMSGSIZE;
6108 }
6109
inet6_valid_dump_ifinfo(const struct nlmsghdr * nlh,struct netlink_ext_ack * extack)6110 static int inet6_valid_dump_ifinfo(const struct nlmsghdr *nlh,
6111 struct netlink_ext_ack *extack)
6112 {
6113 struct ifinfomsg *ifm;
6114
6115 ifm = nlmsg_payload(nlh, sizeof(*ifm));
6116 if (!ifm) {
6117 NL_SET_ERR_MSG_MOD(extack, "Invalid header for link dump request");
6118 return -EINVAL;
6119 }
6120
6121 if (nlmsg_attrlen(nlh, sizeof(*ifm))) {
6122 NL_SET_ERR_MSG_MOD(extack, "Invalid data after header");
6123 return -EINVAL;
6124 }
6125
6126 if (ifm->__ifi_pad || ifm->ifi_type || ifm->ifi_flags ||
6127 ifm->ifi_change || ifm->ifi_index) {
6128 NL_SET_ERR_MSG_MOD(extack, "Invalid values in header for dump request");
6129 return -EINVAL;
6130 }
6131
6132 return 0;
6133 }
6134
inet6_dump_ifinfo(struct sk_buff * skb,struct netlink_callback * cb)6135 static int inet6_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
6136 {
6137 struct net *net = sock_net(skb->sk);
6138 struct {
6139 unsigned long ifindex;
6140 } *ctx = (void *)cb->ctx;
6141 struct net_device *dev;
6142 struct inet6_dev *idev;
6143 int err;
6144
6145 /* only requests using strict checking can pass data to
6146 * influence the dump
6147 */
6148 if (cb->strict_check) {
6149 err = inet6_valid_dump_ifinfo(cb->nlh, cb->extack);
6150
6151 if (err < 0)
6152 return err;
6153 }
6154
6155 err = 0;
6156 rcu_read_lock();
6157 for_each_netdev_dump(net, dev, ctx->ifindex) {
6158 idev = __in6_dev_get(dev);
6159 if (!idev)
6160 continue;
6161 err = inet6_fill_ifinfo(skb, idev,
6162 NETLINK_CB(cb->skb).portid,
6163 cb->nlh->nlmsg_seq,
6164 RTM_NEWLINK, NLM_F_MULTI);
6165 if (err < 0)
6166 break;
6167 }
6168 rcu_read_unlock();
6169
6170 return err;
6171 }
6172
inet6_ifinfo_notify(int event,struct inet6_dev * idev)6173 void inet6_ifinfo_notify(int event, struct inet6_dev *idev)
6174 {
6175 struct sk_buff *skb;
6176 struct net *net = dev_net(idev->dev);
6177 int err = -ENOBUFS;
6178
6179 skb = nlmsg_new(inet6_if_nlmsg_size(), GFP_ATOMIC);
6180 if (!skb)
6181 goto errout;
6182
6183 err = inet6_fill_ifinfo(skb, idev, 0, 0, event, 0);
6184 if (err < 0) {
6185 /* -EMSGSIZE implies BUG in inet6_if_nlmsg_size() */
6186 WARN_ON(err == -EMSGSIZE);
6187 kfree_skb(skb);
6188 goto errout;
6189 }
6190 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFINFO, NULL, GFP_ATOMIC);
6191 return;
6192 errout:
6193 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFINFO, err);
6194 }
6195
inet6_prefix_nlmsg_size(void)6196 static inline size_t inet6_prefix_nlmsg_size(void)
6197 {
6198 return NLMSG_ALIGN(sizeof(struct prefixmsg))
6199 + nla_total_size(sizeof(struct in6_addr))
6200 + nla_total_size(sizeof(struct prefix_cacheinfo));
6201 }
6202
inet6_fill_prefix(struct sk_buff * skb,struct inet6_dev * idev,struct prefix_info * pinfo,u32 portid,u32 seq,int event,unsigned int flags)6203 static int inet6_fill_prefix(struct sk_buff *skb, struct inet6_dev *idev,
6204 struct prefix_info *pinfo, u32 portid, u32 seq,
6205 int event, unsigned int flags)
6206 {
6207 struct prefixmsg *pmsg;
6208 struct nlmsghdr *nlh;
6209 struct prefix_cacheinfo ci;
6210
6211 nlh = nlmsg_put(skb, portid, seq, event, sizeof(*pmsg), flags);
6212 if (!nlh)
6213 return -EMSGSIZE;
6214
6215 pmsg = nlmsg_data(nlh);
6216 pmsg->prefix_family = AF_INET6;
6217 pmsg->prefix_pad1 = 0;
6218 pmsg->prefix_pad2 = 0;
6219 pmsg->prefix_ifindex = idev->dev->ifindex;
6220 pmsg->prefix_len = pinfo->prefix_len;
6221 pmsg->prefix_type = pinfo->type;
6222 pmsg->prefix_pad3 = 0;
6223 pmsg->prefix_flags = pinfo->flags;
6224
6225 if (nla_put(skb, PREFIX_ADDRESS, sizeof(pinfo->prefix), &pinfo->prefix))
6226 goto nla_put_failure;
6227 ci.preferred_time = ntohl(pinfo->prefered);
6228 ci.valid_time = ntohl(pinfo->valid);
6229 if (nla_put(skb, PREFIX_CACHEINFO, sizeof(ci), &ci))
6230 goto nla_put_failure;
6231 nlmsg_end(skb, nlh);
6232 return 0;
6233
6234 nla_put_failure:
6235 nlmsg_cancel(skb, nlh);
6236 return -EMSGSIZE;
6237 }
6238
inet6_prefix_notify(int event,struct inet6_dev * idev,struct prefix_info * pinfo)6239 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
6240 struct prefix_info *pinfo)
6241 {
6242 struct sk_buff *skb;
6243 struct net *net = dev_net(idev->dev);
6244 int err = -ENOBUFS;
6245
6246 skb = nlmsg_new(inet6_prefix_nlmsg_size(), GFP_ATOMIC);
6247 if (!skb)
6248 goto errout;
6249
6250 err = inet6_fill_prefix(skb, idev, pinfo, 0, 0, event, 0);
6251 if (err < 0) {
6252 /* -EMSGSIZE implies BUG in inet6_prefix_nlmsg_size() */
6253 WARN_ON(err == -EMSGSIZE);
6254 kfree_skb(skb);
6255 goto errout;
6256 }
6257 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_PREFIX, NULL, GFP_ATOMIC);
6258 return;
6259 errout:
6260 rtnl_set_sk_err(net, RTNLGRP_IPV6_PREFIX, err);
6261 }
6262
__ipv6_ifa_notify(int event,struct inet6_ifaddr * ifp)6263 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
6264 {
6265 struct net *net = dev_net(ifp->idev->dev);
6266
6267 if (event)
6268 ASSERT_RTNL();
6269
6270 inet6_ifa_notify(event ? : RTM_NEWADDR, ifp);
6271
6272 switch (event) {
6273 case RTM_NEWADDR:
6274 /*
6275 * If the address was optimistic we inserted the route at the
6276 * start of our DAD process, so we don't need to do it again.
6277 * If the device was taken down in the middle of the DAD
6278 * cycle there is a race where we could get here without a
6279 * host route, so nothing to insert. That will be fixed when
6280 * the device is brought up.
6281 */
6282 if (ifp->rt && !rcu_access_pointer(ifp->rt->fib6_node)) {
6283 ip6_ins_rt(net, ifp->rt);
6284 } else if (!ifp->rt && (ifp->idev->dev->flags & IFF_UP)) {
6285 pr_warn("BUG: Address %pI6c on device %s is missing its host route.\n",
6286 &ifp->addr, ifp->idev->dev->name);
6287 }
6288
6289 if (ifp->idev->cnf.forwarding)
6290 addrconf_join_anycast(ifp);
6291 if (!ipv6_addr_any(&ifp->peer_addr))
6292 addrconf_prefix_route(&ifp->peer_addr, 128,
6293 ifp->rt_priority, ifp->idev->dev,
6294 0, 0, GFP_ATOMIC);
6295 break;
6296 case RTM_DELADDR:
6297 if (ifp->idev->cnf.forwarding)
6298 addrconf_leave_anycast(ifp);
6299 addrconf_leave_solict(ifp->idev, &ifp->addr);
6300 if (!ipv6_addr_any(&ifp->peer_addr)) {
6301 struct fib6_info *rt;
6302
6303 rt = addrconf_get_prefix_route(&ifp->peer_addr, 128,
6304 ifp->idev->dev, 0, 0,
6305 false);
6306 if (rt)
6307 ip6_del_rt(net, rt, false);
6308 }
6309 if (ifp->rt) {
6310 ip6_del_rt(net, ifp->rt, false);
6311 ifp->rt = NULL;
6312 }
6313 rt_genid_bump_ipv6(net);
6314 break;
6315 }
6316 atomic_inc(&net->ipv6.dev_addr_genid);
6317 }
6318
ipv6_ifa_notify(int event,struct inet6_ifaddr * ifp)6319 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
6320 {
6321 if (likely(ifp->idev->dead == 0))
6322 __ipv6_ifa_notify(event, ifp);
6323 }
6324
6325 #ifdef CONFIG_SYSCTL
6326
addrconf_sysctl_forward(const struct ctl_table * ctl,int write,void * buffer,size_t * lenp,loff_t * ppos)6327 static int addrconf_sysctl_forward(const struct ctl_table *ctl, int write,
6328 void *buffer, size_t *lenp, loff_t *ppos)
6329 {
6330 int *valp = ctl->data;
6331 int val = *valp;
6332 loff_t pos = *ppos;
6333 struct ctl_table lctl;
6334 int ret;
6335
6336 /*
6337 * ctl->data points to idev->cnf.forwarding, we should
6338 * not modify it until we get the rtnl lock.
6339 */
6340 lctl = *ctl;
6341 lctl.data = &val;
6342
6343 ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
6344
6345 if (write)
6346 ret = addrconf_fixup_forwarding(ctl, valp, val);
6347 if (ret)
6348 *ppos = pos;
6349 return ret;
6350 }
6351
addrconf_sysctl_mtu(const struct ctl_table * ctl,int write,void * buffer,size_t * lenp,loff_t * ppos)6352 static int addrconf_sysctl_mtu(const struct ctl_table *ctl, int write,
6353 void *buffer, size_t *lenp, loff_t *ppos)
6354 {
6355 struct inet6_dev *idev = ctl->extra1;
6356 int min_mtu = IPV6_MIN_MTU;
6357 struct ctl_table lctl;
6358
6359 lctl = *ctl;
6360 lctl.extra1 = &min_mtu;
6361 lctl.extra2 = idev ? &idev->dev->mtu : NULL;
6362
6363 return proc_dointvec_minmax(&lctl, write, buffer, lenp, ppos);
6364 }
6365
dev_disable_change(struct inet6_dev * idev)6366 static void dev_disable_change(struct inet6_dev *idev)
6367 {
6368 struct netdev_notifier_info info;
6369
6370 if (!idev || !idev->dev)
6371 return;
6372
6373 netdev_notifier_info_init(&info, idev->dev);
6374 if (idev->cnf.disable_ipv6)
6375 addrconf_notify(NULL, NETDEV_DOWN, &info);
6376 else
6377 addrconf_notify(NULL, NETDEV_UP, &info);
6378 }
6379
addrconf_disable_change(struct net * net,__s32 newf)6380 static void addrconf_disable_change(struct net *net, __s32 newf)
6381 {
6382 struct net_device *dev;
6383 struct inet6_dev *idev;
6384
6385 for_each_netdev(net, dev) {
6386 idev = __in6_dev_get_rtnl_net(dev);
6387 if (idev) {
6388 int changed = (!idev->cnf.disable_ipv6) ^ (!newf);
6389
6390 WRITE_ONCE(idev->cnf.disable_ipv6, newf);
6391 if (changed)
6392 dev_disable_change(idev);
6393 }
6394 }
6395 }
6396
addrconf_disable_ipv6(const struct ctl_table * table,int * p,int newf)6397 static int addrconf_disable_ipv6(const struct ctl_table *table, int *p, int newf)
6398 {
6399 struct net *net = (struct net *)table->extra2;
6400 int old;
6401
6402 if (p == &net->ipv6.devconf_dflt->disable_ipv6) {
6403 WRITE_ONCE(*p, newf);
6404 return 0;
6405 }
6406
6407 if (!rtnl_net_trylock(net))
6408 return restart_syscall();
6409
6410 old = *p;
6411 WRITE_ONCE(*p, newf);
6412
6413 if (p == &net->ipv6.devconf_all->disable_ipv6) {
6414 WRITE_ONCE(net->ipv6.devconf_dflt->disable_ipv6, newf);
6415 addrconf_disable_change(net, newf);
6416 } else if ((!newf) ^ (!old)) {
6417 dev_disable_change((struct inet6_dev *)table->extra1);
6418 }
6419
6420 rtnl_net_unlock(net);
6421 return 0;
6422 }
6423
addrconf_sysctl_disable(const struct ctl_table * ctl,int write,void * buffer,size_t * lenp,loff_t * ppos)6424 static int addrconf_sysctl_disable(const struct ctl_table *ctl, int write,
6425 void *buffer, size_t *lenp, loff_t *ppos)
6426 {
6427 int *valp = ctl->data;
6428 int val = *valp;
6429 loff_t pos = *ppos;
6430 struct ctl_table lctl;
6431 int ret;
6432
6433 /*
6434 * ctl->data points to idev->cnf.disable_ipv6, we should
6435 * not modify it until we get the rtnl lock.
6436 */
6437 lctl = *ctl;
6438 lctl.data = &val;
6439
6440 ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
6441
6442 if (write)
6443 ret = addrconf_disable_ipv6(ctl, valp, val);
6444 if (ret)
6445 *ppos = pos;
6446 return ret;
6447 }
6448
addrconf_sysctl_proxy_ndp(const struct ctl_table * ctl,int write,void * buffer,size_t * lenp,loff_t * ppos)6449 static int addrconf_sysctl_proxy_ndp(const struct ctl_table *ctl, int write,
6450 void *buffer, size_t *lenp, loff_t *ppos)
6451 {
6452 int *valp = ctl->data;
6453 int ret;
6454 int old, new;
6455
6456 old = *valp;
6457 ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
6458 new = *valp;
6459
6460 if (write && old != new) {
6461 struct net *net = ctl->extra2;
6462
6463 if (!rtnl_net_trylock(net))
6464 return restart_syscall();
6465
6466 if (valp == &net->ipv6.devconf_dflt->proxy_ndp) {
6467 inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
6468 NETCONFA_PROXY_NEIGH,
6469 NETCONFA_IFINDEX_DEFAULT,
6470 net->ipv6.devconf_dflt);
6471 } else if (valp == &net->ipv6.devconf_all->proxy_ndp) {
6472 inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
6473 NETCONFA_PROXY_NEIGH,
6474 NETCONFA_IFINDEX_ALL,
6475 net->ipv6.devconf_all);
6476 } else {
6477 struct inet6_dev *idev = ctl->extra1;
6478
6479 inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
6480 NETCONFA_PROXY_NEIGH,
6481 idev->dev->ifindex,
6482 &idev->cnf);
6483 }
6484 rtnl_net_unlock(net);
6485 }
6486
6487 return ret;
6488 }
6489
addrconf_sysctl_addr_gen_mode(const struct ctl_table * ctl,int write,void * buffer,size_t * lenp,loff_t * ppos)6490 static int addrconf_sysctl_addr_gen_mode(const struct ctl_table *ctl, int write,
6491 void *buffer, size_t *lenp,
6492 loff_t *ppos)
6493 {
6494 int ret = 0;
6495 u32 new_val;
6496 struct inet6_dev *idev = (struct inet6_dev *)ctl->extra1;
6497 struct net *net = (struct net *)ctl->extra2;
6498 struct ctl_table tmp = {
6499 .data = &new_val,
6500 .maxlen = sizeof(new_val),
6501 .mode = ctl->mode,
6502 };
6503
6504 if (!rtnl_net_trylock(net))
6505 return restart_syscall();
6506
6507 new_val = *((u32 *)ctl->data);
6508
6509 ret = proc_douintvec(&tmp, write, buffer, lenp, ppos);
6510 if (ret != 0)
6511 goto out;
6512
6513 if (write) {
6514 if (check_addr_gen_mode(new_val) < 0) {
6515 ret = -EINVAL;
6516 goto out;
6517 }
6518
6519 if (idev) {
6520 if (check_stable_privacy(idev, net, new_val) < 0) {
6521 ret = -EINVAL;
6522 goto out;
6523 }
6524
6525 if (idev->cnf.addr_gen_mode != new_val) {
6526 WRITE_ONCE(idev->cnf.addr_gen_mode, new_val);
6527 netdev_lock_ops(idev->dev);
6528 addrconf_init_auto_addrs(idev->dev);
6529 netdev_unlock_ops(idev->dev);
6530 }
6531 } else if (&net->ipv6.devconf_all->addr_gen_mode == ctl->data) {
6532 struct net_device *dev;
6533
6534 WRITE_ONCE(net->ipv6.devconf_dflt->addr_gen_mode, new_val);
6535 for_each_netdev(net, dev) {
6536 idev = __in6_dev_get_rtnl_net(dev);
6537 if (idev &&
6538 idev->cnf.addr_gen_mode != new_val) {
6539 WRITE_ONCE(idev->cnf.addr_gen_mode,
6540 new_val);
6541 netdev_lock_ops(idev->dev);
6542 addrconf_init_auto_addrs(idev->dev);
6543 netdev_unlock_ops(idev->dev);
6544 }
6545 }
6546 }
6547
6548 WRITE_ONCE(*((u32 *)ctl->data), new_val);
6549 }
6550
6551 out:
6552 rtnl_net_unlock(net);
6553
6554 return ret;
6555 }
6556
addrconf_sysctl_stable_secret(const struct ctl_table * ctl,int write,void * buffer,size_t * lenp,loff_t * ppos)6557 static int addrconf_sysctl_stable_secret(const struct ctl_table *ctl, int write,
6558 void *buffer, size_t *lenp,
6559 loff_t *ppos)
6560 {
6561 int err;
6562 struct in6_addr addr;
6563 char str[IPV6_MAX_STRLEN];
6564 struct ctl_table lctl = *ctl;
6565 struct net *net = ctl->extra2;
6566 struct ipv6_stable_secret *secret = ctl->data;
6567
6568 if (&net->ipv6.devconf_all->stable_secret == ctl->data)
6569 return -EIO;
6570
6571 lctl.maxlen = IPV6_MAX_STRLEN;
6572 lctl.data = str;
6573
6574 if (!rtnl_net_trylock(net))
6575 return restart_syscall();
6576
6577 if (!write && !secret->initialized) {
6578 err = -EIO;
6579 goto out;
6580 }
6581
6582 err = snprintf(str, sizeof(str), "%pI6", &secret->secret);
6583 if (err >= sizeof(str)) {
6584 err = -EIO;
6585 goto out;
6586 }
6587
6588 err = proc_dostring(&lctl, write, buffer, lenp, ppos);
6589 if (err || !write)
6590 goto out;
6591
6592 if (in6_pton(str, -1, addr.in6_u.u6_addr8, -1, NULL) != 1) {
6593 err = -EIO;
6594 goto out;
6595 }
6596
6597 secret->initialized = true;
6598 secret->secret = addr;
6599
6600 if (&net->ipv6.devconf_dflt->stable_secret == ctl->data) {
6601 struct net_device *dev;
6602
6603 for_each_netdev(net, dev) {
6604 struct inet6_dev *idev = __in6_dev_get_rtnl_net(dev);
6605
6606 if (idev) {
6607 WRITE_ONCE(idev->cnf.addr_gen_mode,
6608 IN6_ADDR_GEN_MODE_STABLE_PRIVACY);
6609 }
6610 }
6611 } else {
6612 struct inet6_dev *idev = ctl->extra1;
6613
6614 WRITE_ONCE(idev->cnf.addr_gen_mode,
6615 IN6_ADDR_GEN_MODE_STABLE_PRIVACY);
6616 }
6617
6618 out:
6619 rtnl_net_unlock(net);
6620
6621 return err;
6622 }
6623
6624 static
addrconf_sysctl_ignore_routes_with_linkdown(const struct ctl_table * ctl,int write,void * buffer,size_t * lenp,loff_t * ppos)6625 int addrconf_sysctl_ignore_routes_with_linkdown(const struct ctl_table *ctl,
6626 int write, void *buffer,
6627 size_t *lenp,
6628 loff_t *ppos)
6629 {
6630 int *valp = ctl->data;
6631 int val = *valp;
6632 loff_t pos = *ppos;
6633 struct ctl_table lctl;
6634 int ret;
6635
6636 /* ctl->data points to idev->cnf.ignore_routes_when_linkdown
6637 * we should not modify it until we get the rtnl lock.
6638 */
6639 lctl = *ctl;
6640 lctl.data = &val;
6641
6642 ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
6643
6644 if (write)
6645 ret = addrconf_fixup_linkdown(ctl, valp, val);
6646 if (ret)
6647 *ppos = pos;
6648 return ret;
6649 }
6650
6651 static
addrconf_set_nopolicy(struct rt6_info * rt,int action)6652 void addrconf_set_nopolicy(struct rt6_info *rt, int action)
6653 {
6654 if (rt) {
6655 if (action)
6656 rt->dst.flags |= DST_NOPOLICY;
6657 else
6658 rt->dst.flags &= ~DST_NOPOLICY;
6659 }
6660 }
6661
6662 static
addrconf_disable_policy_idev(struct inet6_dev * idev,int val)6663 void addrconf_disable_policy_idev(struct inet6_dev *idev, int val)
6664 {
6665 struct inet6_ifaddr *ifa;
6666
6667 read_lock_bh(&idev->lock);
6668 list_for_each_entry(ifa, &idev->addr_list, if_list) {
6669 spin_lock(&ifa->lock);
6670 if (ifa->rt) {
6671 /* host routes only use builtin fib6_nh */
6672 struct fib6_nh *nh = ifa->rt->fib6_nh;
6673 int cpu;
6674
6675 rcu_read_lock();
6676 ifa->rt->dst_nopolicy = val ? true : false;
6677 if (nh->rt6i_pcpu) {
6678 for_each_possible_cpu(cpu) {
6679 struct rt6_info **rtp;
6680
6681 rtp = per_cpu_ptr(nh->rt6i_pcpu, cpu);
6682 addrconf_set_nopolicy(*rtp, val);
6683 }
6684 }
6685 rcu_read_unlock();
6686 }
6687 spin_unlock(&ifa->lock);
6688 }
6689 read_unlock_bh(&idev->lock);
6690 }
6691
6692 static
addrconf_disable_policy(const struct ctl_table * ctl,int * valp,int val)6693 int addrconf_disable_policy(const struct ctl_table *ctl, int *valp, int val)
6694 {
6695 struct net *net = (struct net *)ctl->extra2;
6696 struct inet6_dev *idev;
6697
6698 if (valp == &net->ipv6.devconf_dflt->disable_policy) {
6699 WRITE_ONCE(*valp, val);
6700 return 0;
6701 }
6702
6703 if (!rtnl_net_trylock(net))
6704 return restart_syscall();
6705
6706 WRITE_ONCE(*valp, val);
6707
6708 if (valp == &net->ipv6.devconf_all->disable_policy) {
6709 struct net_device *dev;
6710
6711 for_each_netdev(net, dev) {
6712 idev = __in6_dev_get_rtnl_net(dev);
6713 if (idev)
6714 addrconf_disable_policy_idev(idev, val);
6715 }
6716 } else {
6717 idev = (struct inet6_dev *)ctl->extra1;
6718 addrconf_disable_policy_idev(idev, val);
6719 }
6720
6721 rtnl_net_unlock(net);
6722 return 0;
6723 }
6724
addrconf_sysctl_disable_policy(const struct ctl_table * ctl,int write,void * buffer,size_t * lenp,loff_t * ppos)6725 static int addrconf_sysctl_disable_policy(const struct ctl_table *ctl, int write,
6726 void *buffer, size_t *lenp, loff_t *ppos)
6727 {
6728 int *valp = ctl->data;
6729 int val = *valp;
6730 loff_t pos = *ppos;
6731 struct ctl_table lctl;
6732 int ret;
6733
6734 lctl = *ctl;
6735 lctl.data = &val;
6736 ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
6737
6738 if (write && (*valp != val))
6739 ret = addrconf_disable_policy(ctl, valp, val);
6740
6741 if (ret)
6742 *ppos = pos;
6743
6744 return ret;
6745 }
6746
addrconf_force_forward_change(struct net * net,__s32 newf)6747 static void addrconf_force_forward_change(struct net *net, __s32 newf)
6748 {
6749 struct net_device *dev;
6750 struct inet6_dev *idev;
6751
6752 for_each_netdev(net, dev) {
6753 idev = __in6_dev_get_rtnl_net(dev);
6754 if (idev) {
6755 int changed = (!idev->cnf.force_forwarding) ^ (!newf);
6756
6757 WRITE_ONCE(idev->cnf.force_forwarding, newf);
6758 if (changed)
6759 inet6_netconf_notify_devconf(dev_net(dev), RTM_NEWNETCONF,
6760 NETCONFA_FORCE_FORWARDING,
6761 dev->ifindex, &idev->cnf);
6762 }
6763 }
6764 }
6765
addrconf_sysctl_force_forwarding(const struct ctl_table * ctl,int write,void * buffer,size_t * lenp,loff_t * ppos)6766 static int addrconf_sysctl_force_forwarding(const struct ctl_table *ctl, int write,
6767 void *buffer, size_t *lenp, loff_t *ppos)
6768 {
6769 struct inet6_dev *idev = ctl->extra1;
6770 struct ctl_table tmp_ctl = *ctl;
6771 struct net *net = ctl->extra2;
6772 int *valp = ctl->data;
6773 int new_val = *valp;
6774 int old_val = *valp;
6775 loff_t pos = *ppos;
6776 int ret;
6777
6778 tmp_ctl.extra1 = SYSCTL_ZERO;
6779 tmp_ctl.extra2 = SYSCTL_ONE;
6780 tmp_ctl.data = &new_val;
6781
6782 ret = proc_douintvec_minmax(&tmp_ctl, write, buffer, lenp, ppos);
6783
6784 if (write && old_val != new_val) {
6785 if (!rtnl_net_trylock(net))
6786 return restart_syscall();
6787
6788 WRITE_ONCE(*valp, new_val);
6789
6790 if (valp == &net->ipv6.devconf_dflt->force_forwarding) {
6791 inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
6792 NETCONFA_FORCE_FORWARDING,
6793 NETCONFA_IFINDEX_DEFAULT,
6794 net->ipv6.devconf_dflt);
6795 } else if (valp == &net->ipv6.devconf_all->force_forwarding) {
6796 inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
6797 NETCONFA_FORCE_FORWARDING,
6798 NETCONFA_IFINDEX_ALL,
6799 net->ipv6.devconf_all);
6800
6801 addrconf_force_forward_change(net, new_val);
6802 } else {
6803 inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
6804 NETCONFA_FORCE_FORWARDING,
6805 idev->dev->ifindex,
6806 &idev->cnf);
6807 }
6808 rtnl_net_unlock(net);
6809 }
6810
6811 if (ret)
6812 *ppos = pos;
6813 return ret;
6814 }
6815
6816 static int minus_one = -1;
6817 static const int two_five_five = 255;
6818 static u32 ioam6_if_id_max = U16_MAX;
6819
6820 static const struct ctl_table addrconf_sysctl[] = {
6821 {
6822 .procname = "forwarding",
6823 .data = &ipv6_devconf.forwarding,
6824 .maxlen = sizeof(int),
6825 .mode = 0644,
6826 .proc_handler = addrconf_sysctl_forward,
6827 },
6828 {
6829 .procname = "hop_limit",
6830 .data = &ipv6_devconf.hop_limit,
6831 .maxlen = sizeof(int),
6832 .mode = 0644,
6833 .proc_handler = proc_dointvec_minmax,
6834 .extra1 = (void *)SYSCTL_ONE,
6835 .extra2 = (void *)&two_five_five,
6836 },
6837 {
6838 .procname = "mtu",
6839 .data = &ipv6_devconf.mtu6,
6840 .maxlen = sizeof(int),
6841 .mode = 0644,
6842 .proc_handler = addrconf_sysctl_mtu,
6843 },
6844 {
6845 .procname = "accept_ra",
6846 .data = &ipv6_devconf.accept_ra,
6847 .maxlen = sizeof(int),
6848 .mode = 0644,
6849 .proc_handler = proc_dointvec,
6850 },
6851 {
6852 .procname = "accept_redirects",
6853 .data = &ipv6_devconf.accept_redirects,
6854 .maxlen = sizeof(int),
6855 .mode = 0644,
6856 .proc_handler = proc_dointvec,
6857 },
6858 {
6859 .procname = "autoconf",
6860 .data = &ipv6_devconf.autoconf,
6861 .maxlen = sizeof(int),
6862 .mode = 0644,
6863 .proc_handler = proc_dointvec,
6864 },
6865 {
6866 .procname = "dad_transmits",
6867 .data = &ipv6_devconf.dad_transmits,
6868 .maxlen = sizeof(int),
6869 .mode = 0644,
6870 .proc_handler = proc_dointvec,
6871 },
6872 {
6873 .procname = "router_solicitations",
6874 .data = &ipv6_devconf.rtr_solicits,
6875 .maxlen = sizeof(int),
6876 .mode = 0644,
6877 .proc_handler = proc_dointvec_minmax,
6878 .extra1 = &minus_one,
6879 },
6880 {
6881 .procname = "router_solicitation_interval",
6882 .data = &ipv6_devconf.rtr_solicit_interval,
6883 .maxlen = sizeof(int),
6884 .mode = 0644,
6885 .proc_handler = proc_dointvec_jiffies,
6886 },
6887 {
6888 .procname = "router_solicitation_max_interval",
6889 .data = &ipv6_devconf.rtr_solicit_max_interval,
6890 .maxlen = sizeof(int),
6891 .mode = 0644,
6892 .proc_handler = proc_dointvec_jiffies,
6893 },
6894 {
6895 .procname = "router_solicitation_delay",
6896 .data = &ipv6_devconf.rtr_solicit_delay,
6897 .maxlen = sizeof(int),
6898 .mode = 0644,
6899 .proc_handler = proc_dointvec_jiffies,
6900 },
6901 {
6902 .procname = "force_mld_version",
6903 .data = &ipv6_devconf.force_mld_version,
6904 .maxlen = sizeof(int),
6905 .mode = 0644,
6906 .proc_handler = proc_dointvec,
6907 },
6908 {
6909 .procname = "mldv1_unsolicited_report_interval",
6910 .data =
6911 &ipv6_devconf.mldv1_unsolicited_report_interval,
6912 .maxlen = sizeof(int),
6913 .mode = 0644,
6914 .proc_handler = proc_dointvec_ms_jiffies,
6915 },
6916 {
6917 .procname = "mldv2_unsolicited_report_interval",
6918 .data =
6919 &ipv6_devconf.mldv2_unsolicited_report_interval,
6920 .maxlen = sizeof(int),
6921 .mode = 0644,
6922 .proc_handler = proc_dointvec_ms_jiffies,
6923 },
6924 {
6925 .procname = "use_tempaddr",
6926 .data = &ipv6_devconf.use_tempaddr,
6927 .maxlen = sizeof(int),
6928 .mode = 0644,
6929 .proc_handler = proc_dointvec,
6930 },
6931 {
6932 .procname = "temp_valid_lft",
6933 .data = &ipv6_devconf.temp_valid_lft,
6934 .maxlen = sizeof(int),
6935 .mode = 0644,
6936 .proc_handler = proc_dointvec,
6937 },
6938 {
6939 .procname = "temp_prefered_lft",
6940 .data = &ipv6_devconf.temp_prefered_lft,
6941 .maxlen = sizeof(int),
6942 .mode = 0644,
6943 .proc_handler = proc_dointvec,
6944 },
6945 {
6946 .procname = "regen_min_advance",
6947 .data = &ipv6_devconf.regen_min_advance,
6948 .maxlen = sizeof(int),
6949 .mode = 0644,
6950 .proc_handler = proc_dointvec,
6951 },
6952 {
6953 .procname = "regen_max_retry",
6954 .data = &ipv6_devconf.regen_max_retry,
6955 .maxlen = sizeof(int),
6956 .mode = 0644,
6957 .proc_handler = proc_dointvec,
6958 },
6959 {
6960 .procname = "max_desync_factor",
6961 .data = &ipv6_devconf.max_desync_factor,
6962 .maxlen = sizeof(int),
6963 .mode = 0644,
6964 .proc_handler = proc_dointvec,
6965 },
6966 {
6967 .procname = "max_addresses",
6968 .data = &ipv6_devconf.max_addresses,
6969 .maxlen = sizeof(int),
6970 .mode = 0644,
6971 .proc_handler = proc_dointvec,
6972 },
6973 {
6974 .procname = "accept_ra_defrtr",
6975 .data = &ipv6_devconf.accept_ra_defrtr,
6976 .maxlen = sizeof(int),
6977 .mode = 0644,
6978 .proc_handler = proc_dointvec,
6979 },
6980 {
6981 .procname = "ra_defrtr_metric",
6982 .data = &ipv6_devconf.ra_defrtr_metric,
6983 .maxlen = sizeof(u32),
6984 .mode = 0644,
6985 .proc_handler = proc_douintvec_minmax,
6986 .extra1 = (void *)SYSCTL_ONE,
6987 },
6988 {
6989 .procname = "accept_ra_min_hop_limit",
6990 .data = &ipv6_devconf.accept_ra_min_hop_limit,
6991 .maxlen = sizeof(int),
6992 .mode = 0644,
6993 .proc_handler = proc_dointvec,
6994 },
6995 {
6996 .procname = "accept_ra_min_lft",
6997 .data = &ipv6_devconf.accept_ra_min_lft,
6998 .maxlen = sizeof(int),
6999 .mode = 0644,
7000 .proc_handler = proc_dointvec,
7001 },
7002 {
7003 .procname = "accept_ra_pinfo",
7004 .data = &ipv6_devconf.accept_ra_pinfo,
7005 .maxlen = sizeof(int),
7006 .mode = 0644,
7007 .proc_handler = proc_dointvec,
7008 },
7009 {
7010 .procname = "ra_honor_pio_life",
7011 .data = &ipv6_devconf.ra_honor_pio_life,
7012 .maxlen = sizeof(u8),
7013 .mode = 0644,
7014 .proc_handler = proc_dou8vec_minmax,
7015 .extra1 = SYSCTL_ZERO,
7016 .extra2 = SYSCTL_ONE,
7017 },
7018 {
7019 .procname = "ra_honor_pio_pflag",
7020 .data = &ipv6_devconf.ra_honor_pio_pflag,
7021 .maxlen = sizeof(u8),
7022 .mode = 0644,
7023 .proc_handler = proc_dou8vec_minmax,
7024 .extra1 = SYSCTL_ZERO,
7025 .extra2 = SYSCTL_ONE,
7026 },
7027 #ifdef CONFIG_IPV6_ROUTER_PREF
7028 {
7029 .procname = "accept_ra_rtr_pref",
7030 .data = &ipv6_devconf.accept_ra_rtr_pref,
7031 .maxlen = sizeof(int),
7032 .mode = 0644,
7033 .proc_handler = proc_dointvec,
7034 },
7035 {
7036 .procname = "router_probe_interval",
7037 .data = &ipv6_devconf.rtr_probe_interval,
7038 .maxlen = sizeof(int),
7039 .mode = 0644,
7040 .proc_handler = proc_dointvec_jiffies,
7041 },
7042 #ifdef CONFIG_IPV6_ROUTE_INFO
7043 {
7044 .procname = "accept_ra_rt_info_min_plen",
7045 .data = &ipv6_devconf.accept_ra_rt_info_min_plen,
7046 .maxlen = sizeof(int),
7047 .mode = 0644,
7048 .proc_handler = proc_dointvec,
7049 },
7050 {
7051 .procname = "accept_ra_rt_info_max_plen",
7052 .data = &ipv6_devconf.accept_ra_rt_info_max_plen,
7053 .maxlen = sizeof(int),
7054 .mode = 0644,
7055 .proc_handler = proc_dointvec,
7056 },
7057 #endif
7058 #endif
7059 {
7060 .procname = "proxy_ndp",
7061 .data = &ipv6_devconf.proxy_ndp,
7062 .maxlen = sizeof(int),
7063 .mode = 0644,
7064 .proc_handler = addrconf_sysctl_proxy_ndp,
7065 },
7066 {
7067 .procname = "accept_source_route",
7068 .data = &ipv6_devconf.accept_source_route,
7069 .maxlen = sizeof(int),
7070 .mode = 0644,
7071 .proc_handler = proc_dointvec,
7072 },
7073 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
7074 {
7075 .procname = "optimistic_dad",
7076 .data = &ipv6_devconf.optimistic_dad,
7077 .maxlen = sizeof(int),
7078 .mode = 0644,
7079 .proc_handler = proc_dointvec,
7080 },
7081 {
7082 .procname = "use_optimistic",
7083 .data = &ipv6_devconf.use_optimistic,
7084 .maxlen = sizeof(int),
7085 .mode = 0644,
7086 .proc_handler = proc_dointvec,
7087 },
7088 #endif
7089 #ifdef CONFIG_IPV6_MROUTE
7090 {
7091 .procname = "mc_forwarding",
7092 .data = &ipv6_devconf.mc_forwarding,
7093 .maxlen = sizeof(int),
7094 .mode = 0444,
7095 .proc_handler = proc_dointvec,
7096 },
7097 #endif
7098 {
7099 .procname = "disable_ipv6",
7100 .data = &ipv6_devconf.disable_ipv6,
7101 .maxlen = sizeof(int),
7102 .mode = 0644,
7103 .proc_handler = addrconf_sysctl_disable,
7104 },
7105 {
7106 .procname = "accept_dad",
7107 .data = &ipv6_devconf.accept_dad,
7108 .maxlen = sizeof(int),
7109 .mode = 0644,
7110 .proc_handler = proc_dointvec,
7111 },
7112 {
7113 .procname = "force_tllao",
7114 .data = &ipv6_devconf.force_tllao,
7115 .maxlen = sizeof(int),
7116 .mode = 0644,
7117 .proc_handler = proc_dointvec
7118 },
7119 {
7120 .procname = "ndisc_notify",
7121 .data = &ipv6_devconf.ndisc_notify,
7122 .maxlen = sizeof(int),
7123 .mode = 0644,
7124 .proc_handler = proc_dointvec
7125 },
7126 {
7127 .procname = "suppress_frag_ndisc",
7128 .data = &ipv6_devconf.suppress_frag_ndisc,
7129 .maxlen = sizeof(int),
7130 .mode = 0644,
7131 .proc_handler = proc_dointvec
7132 },
7133 {
7134 .procname = "accept_ra_from_local",
7135 .data = &ipv6_devconf.accept_ra_from_local,
7136 .maxlen = sizeof(int),
7137 .mode = 0644,
7138 .proc_handler = proc_dointvec,
7139 },
7140 {
7141 .procname = "accept_ra_mtu",
7142 .data = &ipv6_devconf.accept_ra_mtu,
7143 .maxlen = sizeof(int),
7144 .mode = 0644,
7145 .proc_handler = proc_dointvec,
7146 },
7147 {
7148 .procname = "stable_secret",
7149 .data = &ipv6_devconf.stable_secret,
7150 .maxlen = IPV6_MAX_STRLEN,
7151 .mode = 0600,
7152 .proc_handler = addrconf_sysctl_stable_secret,
7153 },
7154 {
7155 .procname = "use_oif_addrs_only",
7156 .data = &ipv6_devconf.use_oif_addrs_only,
7157 .maxlen = sizeof(int),
7158 .mode = 0644,
7159 .proc_handler = proc_dointvec,
7160 },
7161 {
7162 .procname = "ignore_routes_with_linkdown",
7163 .data = &ipv6_devconf.ignore_routes_with_linkdown,
7164 .maxlen = sizeof(int),
7165 .mode = 0644,
7166 .proc_handler = addrconf_sysctl_ignore_routes_with_linkdown,
7167 },
7168 {
7169 .procname = "drop_unicast_in_l2_multicast",
7170 .data = &ipv6_devconf.drop_unicast_in_l2_multicast,
7171 .maxlen = sizeof(int),
7172 .mode = 0644,
7173 .proc_handler = proc_dointvec,
7174 },
7175 {
7176 .procname = "drop_unsolicited_na",
7177 .data = &ipv6_devconf.drop_unsolicited_na,
7178 .maxlen = sizeof(int),
7179 .mode = 0644,
7180 .proc_handler = proc_dointvec,
7181 },
7182 {
7183 .procname = "keep_addr_on_down",
7184 .data = &ipv6_devconf.keep_addr_on_down,
7185 .maxlen = sizeof(int),
7186 .mode = 0644,
7187 .proc_handler = proc_dointvec,
7188
7189 },
7190 {
7191 .procname = "seg6_enabled",
7192 .data = &ipv6_devconf.seg6_enabled,
7193 .maxlen = sizeof(int),
7194 .mode = 0644,
7195 .proc_handler = proc_dointvec,
7196 },
7197 #ifdef CONFIG_IPV6_SEG6_HMAC
7198 {
7199 .procname = "seg6_require_hmac",
7200 .data = &ipv6_devconf.seg6_require_hmac,
7201 .maxlen = sizeof(int),
7202 .mode = 0644,
7203 .proc_handler = proc_dointvec,
7204 },
7205 #endif
7206 {
7207 .procname = "enhanced_dad",
7208 .data = &ipv6_devconf.enhanced_dad,
7209 .maxlen = sizeof(int),
7210 .mode = 0644,
7211 .proc_handler = proc_dointvec,
7212 },
7213 {
7214 .procname = "addr_gen_mode",
7215 .data = &ipv6_devconf.addr_gen_mode,
7216 .maxlen = sizeof(int),
7217 .mode = 0644,
7218 .proc_handler = addrconf_sysctl_addr_gen_mode,
7219 },
7220 {
7221 .procname = "disable_policy",
7222 .data = &ipv6_devconf.disable_policy,
7223 .maxlen = sizeof(int),
7224 .mode = 0644,
7225 .proc_handler = addrconf_sysctl_disable_policy,
7226 },
7227 {
7228 .procname = "ndisc_tclass",
7229 .data = &ipv6_devconf.ndisc_tclass,
7230 .maxlen = sizeof(int),
7231 .mode = 0644,
7232 .proc_handler = proc_dointvec_minmax,
7233 .extra1 = (void *)SYSCTL_ZERO,
7234 .extra2 = (void *)&two_five_five,
7235 },
7236 {
7237 .procname = "rpl_seg_enabled",
7238 .data = &ipv6_devconf.rpl_seg_enabled,
7239 .maxlen = sizeof(int),
7240 .mode = 0644,
7241 .proc_handler = proc_dointvec,
7242 },
7243 {
7244 .procname = "ioam6_enabled",
7245 .data = &ipv6_devconf.ioam6_enabled,
7246 .maxlen = sizeof(u8),
7247 .mode = 0644,
7248 .proc_handler = proc_dou8vec_minmax,
7249 .extra1 = (void *)SYSCTL_ZERO,
7250 .extra2 = (void *)SYSCTL_ONE,
7251 },
7252 {
7253 .procname = "ioam6_id",
7254 .data = &ipv6_devconf.ioam6_id,
7255 .maxlen = sizeof(u32),
7256 .mode = 0644,
7257 .proc_handler = proc_douintvec_minmax,
7258 .extra1 = (void *)SYSCTL_ZERO,
7259 .extra2 = (void *)&ioam6_if_id_max,
7260 },
7261 {
7262 .procname = "ioam6_id_wide",
7263 .data = &ipv6_devconf.ioam6_id_wide,
7264 .maxlen = sizeof(u32),
7265 .mode = 0644,
7266 .proc_handler = proc_douintvec,
7267 },
7268 {
7269 .procname = "ndisc_evict_nocarrier",
7270 .data = &ipv6_devconf.ndisc_evict_nocarrier,
7271 .maxlen = sizeof(u8),
7272 .mode = 0644,
7273 .proc_handler = proc_dou8vec_minmax,
7274 .extra1 = (void *)SYSCTL_ZERO,
7275 .extra2 = (void *)SYSCTL_ONE,
7276 },
7277 {
7278 .procname = "accept_untracked_na",
7279 .data = &ipv6_devconf.accept_untracked_na,
7280 .maxlen = sizeof(int),
7281 .mode = 0644,
7282 .proc_handler = proc_dointvec_minmax,
7283 .extra1 = SYSCTL_ZERO,
7284 .extra2 = SYSCTL_TWO,
7285 },
7286 {
7287 .procname = "force_forwarding",
7288 .data = &ipv6_devconf.force_forwarding,
7289 .maxlen = sizeof(int),
7290 .mode = 0644,
7291 .proc_handler = addrconf_sysctl_force_forwarding,
7292 },
7293 };
7294
__addrconf_sysctl_register(struct net * net,char * dev_name,struct inet6_dev * idev,struct ipv6_devconf * p)7295 static int __addrconf_sysctl_register(struct net *net, char *dev_name,
7296 struct inet6_dev *idev, struct ipv6_devconf *p)
7297 {
7298 size_t table_size = ARRAY_SIZE(addrconf_sysctl);
7299 int i, ifindex;
7300 struct ctl_table *table;
7301 char path[sizeof("net/ipv6/conf/") + IFNAMSIZ];
7302
7303 table = kmemdup(addrconf_sysctl, sizeof(addrconf_sysctl), GFP_KERNEL_ACCOUNT);
7304 if (!table)
7305 goto out;
7306
7307 for (i = 0; i < table_size; i++) {
7308 table[i].data += (char *)p - (char *)&ipv6_devconf;
7309 /* If one of these is already set, then it is not safe to
7310 * overwrite either of them: this makes proc_dointvec_minmax
7311 * usable.
7312 */
7313 if (!table[i].extra1 && !table[i].extra2) {
7314 table[i].extra1 = idev; /* embedded; no ref */
7315 table[i].extra2 = net;
7316 }
7317 }
7318
7319 snprintf(path, sizeof(path), "net/ipv6/conf/%s", dev_name);
7320
7321 p->sysctl_header = register_net_sysctl_sz(net, path, table,
7322 table_size);
7323 if (!p->sysctl_header)
7324 goto free;
7325
7326 if (!strcmp(dev_name, "all"))
7327 ifindex = NETCONFA_IFINDEX_ALL;
7328 else if (!strcmp(dev_name, "default"))
7329 ifindex = NETCONFA_IFINDEX_DEFAULT;
7330 else
7331 ifindex = idev->dev->ifindex;
7332 inet6_netconf_notify_devconf(net, RTM_NEWNETCONF, NETCONFA_ALL,
7333 ifindex, p);
7334 return 0;
7335
7336 free:
7337 kfree(table);
7338 out:
7339 return -ENOBUFS;
7340 }
7341
__addrconf_sysctl_unregister(struct net * net,struct ipv6_devconf * p,int ifindex)7342 static void __addrconf_sysctl_unregister(struct net *net,
7343 struct ipv6_devconf *p, int ifindex)
7344 {
7345 const struct ctl_table *table;
7346
7347 if (!p->sysctl_header)
7348 return;
7349
7350 table = p->sysctl_header->ctl_table_arg;
7351 unregister_net_sysctl_table(p->sysctl_header);
7352 p->sysctl_header = NULL;
7353 kfree(table);
7354
7355 inet6_netconf_notify_devconf(net, RTM_DELNETCONF, 0, ifindex, NULL);
7356 }
7357
addrconf_sysctl_register(struct inet6_dev * idev)7358 static int addrconf_sysctl_register(struct inet6_dev *idev)
7359 {
7360 int err;
7361
7362 if (!sysctl_dev_name_is_allowed(idev->dev->name))
7363 return -EINVAL;
7364
7365 err = neigh_sysctl_register(idev->dev, idev->nd_parms,
7366 &ndisc_ifinfo_sysctl_change);
7367 if (err)
7368 return err;
7369 err = __addrconf_sysctl_register(dev_net(idev->dev), idev->dev->name,
7370 idev, &idev->cnf);
7371 if (err)
7372 neigh_sysctl_unregister(idev->nd_parms);
7373
7374 return err;
7375 }
7376
addrconf_sysctl_unregister(struct inet6_dev * idev)7377 static void addrconf_sysctl_unregister(struct inet6_dev *idev)
7378 {
7379 __addrconf_sysctl_unregister(dev_net(idev->dev), &idev->cnf,
7380 idev->dev->ifindex);
7381 neigh_sysctl_unregister(idev->nd_parms);
7382 }
7383
7384
7385 #endif
7386
addrconf_init_net(struct net * net)7387 static int __net_init addrconf_init_net(struct net *net)
7388 {
7389 int err = -ENOMEM;
7390 struct ipv6_devconf *all, *dflt;
7391
7392 spin_lock_init(&net->ipv6.addrconf_hash_lock);
7393 INIT_DEFERRABLE_WORK(&net->ipv6.addr_chk_work, addrconf_verify_work);
7394 net->ipv6.inet6_addr_lst = kcalloc(IN6_ADDR_HSIZE,
7395 sizeof(struct hlist_head),
7396 GFP_KERNEL);
7397 if (!net->ipv6.inet6_addr_lst)
7398 goto err_alloc_addr;
7399
7400 all = kmemdup(&ipv6_devconf, sizeof(ipv6_devconf), GFP_KERNEL);
7401 if (!all)
7402 goto err_alloc_all;
7403
7404 dflt = kmemdup(&ipv6_devconf_dflt, sizeof(ipv6_devconf_dflt), GFP_KERNEL);
7405 if (!dflt)
7406 goto err_alloc_dflt;
7407
7408 if (!net_eq(net, &init_net)) {
7409 switch (net_inherit_devconf()) {
7410 case 1: /* copy from init_net */
7411 memcpy(all, init_net.ipv6.devconf_all,
7412 sizeof(ipv6_devconf));
7413 memcpy(dflt, init_net.ipv6.devconf_dflt,
7414 sizeof(ipv6_devconf_dflt));
7415 break;
7416 case 3: /* copy from the current netns */
7417 memcpy(all, current->nsproxy->net_ns->ipv6.devconf_all,
7418 sizeof(ipv6_devconf));
7419 memcpy(dflt,
7420 current->nsproxy->net_ns->ipv6.devconf_dflt,
7421 sizeof(ipv6_devconf_dflt));
7422 break;
7423 case 0:
7424 case 2:
7425 /* use compiled values */
7426 break;
7427 }
7428 }
7429
7430 /* these will be inherited by all namespaces */
7431 dflt->autoconf = ipv6_defaults.autoconf;
7432 dflt->disable_ipv6 = ipv6_defaults.disable_ipv6;
7433
7434 dflt->stable_secret.initialized = false;
7435 all->stable_secret.initialized = false;
7436
7437 net->ipv6.devconf_all = all;
7438 net->ipv6.devconf_dflt = dflt;
7439
7440 #ifdef CONFIG_SYSCTL
7441 err = __addrconf_sysctl_register(net, "all", NULL, all);
7442 if (err < 0)
7443 goto err_reg_all;
7444
7445 err = __addrconf_sysctl_register(net, "default", NULL, dflt);
7446 if (err < 0)
7447 goto err_reg_dflt;
7448 #endif
7449 return 0;
7450
7451 #ifdef CONFIG_SYSCTL
7452 err_reg_dflt:
7453 __addrconf_sysctl_unregister(net, all, NETCONFA_IFINDEX_ALL);
7454 err_reg_all:
7455 kfree(dflt);
7456 net->ipv6.devconf_dflt = NULL;
7457 #endif
7458 err_alloc_dflt:
7459 kfree(all);
7460 net->ipv6.devconf_all = NULL;
7461 err_alloc_all:
7462 kfree(net->ipv6.inet6_addr_lst);
7463 err_alloc_addr:
7464 return err;
7465 }
7466
addrconf_exit_net(struct net * net)7467 static void __net_exit addrconf_exit_net(struct net *net)
7468 {
7469 int i;
7470
7471 #ifdef CONFIG_SYSCTL
7472 __addrconf_sysctl_unregister(net, net->ipv6.devconf_dflt,
7473 NETCONFA_IFINDEX_DEFAULT);
7474 __addrconf_sysctl_unregister(net, net->ipv6.devconf_all,
7475 NETCONFA_IFINDEX_ALL);
7476 #endif
7477 kfree(net->ipv6.devconf_dflt);
7478 net->ipv6.devconf_dflt = NULL;
7479 kfree(net->ipv6.devconf_all);
7480 net->ipv6.devconf_all = NULL;
7481
7482 cancel_delayed_work_sync(&net->ipv6.addr_chk_work);
7483 /*
7484 * Check hash table, then free it.
7485 */
7486 for (i = 0; i < IN6_ADDR_HSIZE; i++)
7487 WARN_ON_ONCE(!hlist_empty(&net->ipv6.inet6_addr_lst[i]));
7488
7489 kfree(net->ipv6.inet6_addr_lst);
7490 net->ipv6.inet6_addr_lst = NULL;
7491 }
7492
7493 static struct pernet_operations addrconf_ops = {
7494 .init = addrconf_init_net,
7495 .exit = addrconf_exit_net,
7496 };
7497
7498 static struct rtnl_af_ops inet6_ops __read_mostly = {
7499 .family = AF_INET6,
7500 .fill_link_af = inet6_fill_link_af,
7501 .get_link_af_size = inet6_get_link_af_size,
7502 .validate_link_af = inet6_validate_link_af,
7503 .set_link_af = inet6_set_link_af,
7504 };
7505
7506 static const struct rtnl_msg_handler addrconf_rtnl_msg_handlers[] __initconst_or_module = {
7507 {.owner = THIS_MODULE, .protocol = PF_INET6, .msgtype = RTM_GETLINK,
7508 .dumpit = inet6_dump_ifinfo, .flags = RTNL_FLAG_DUMP_UNLOCKED},
7509 {.owner = THIS_MODULE, .protocol = PF_INET6, .msgtype = RTM_NEWADDR,
7510 .doit = inet6_rtm_newaddr, .flags = RTNL_FLAG_DOIT_PERNET},
7511 {.owner = THIS_MODULE, .protocol = PF_INET6, .msgtype = RTM_DELADDR,
7512 .doit = inet6_rtm_deladdr, .flags = RTNL_FLAG_DOIT_PERNET},
7513 {.owner = THIS_MODULE, .protocol = PF_INET6, .msgtype = RTM_GETADDR,
7514 .doit = inet6_rtm_getaddr, .dumpit = inet6_dump_ifaddr,
7515 .flags = RTNL_FLAG_DOIT_UNLOCKED | RTNL_FLAG_DUMP_UNLOCKED},
7516 {.owner = THIS_MODULE, .protocol = PF_INET6, .msgtype = RTM_GETMULTICAST,
7517 .dumpit = inet6_dump_ifmcaddr,
7518 .flags = RTNL_FLAG_DUMP_UNLOCKED},
7519 {.owner = THIS_MODULE, .protocol = PF_INET6, .msgtype = RTM_GETANYCAST,
7520 .dumpit = inet6_dump_ifacaddr,
7521 .flags = RTNL_FLAG_DUMP_UNLOCKED},
7522 {.owner = THIS_MODULE, .protocol = PF_INET6, .msgtype = RTM_GETNETCONF,
7523 .doit = inet6_netconf_get_devconf, .dumpit = inet6_netconf_dump_devconf,
7524 .flags = RTNL_FLAG_DOIT_UNLOCKED | RTNL_FLAG_DUMP_UNLOCKED},
7525 };
7526
7527 /*
7528 * Init / cleanup code
7529 */
7530
addrconf_init(void)7531 int __init addrconf_init(void)
7532 {
7533 struct inet6_dev *idev;
7534 int err;
7535
7536 err = ipv6_addr_label_init();
7537 if (err < 0) {
7538 pr_crit("%s: cannot initialize default policy table: %d\n",
7539 __func__, err);
7540 goto out;
7541 }
7542
7543 err = register_pernet_subsys(&addrconf_ops);
7544 if (err < 0)
7545 goto out_addrlabel;
7546
7547 /* All works using addrconf_wq need to lock rtnl. */
7548 addrconf_wq = create_singlethread_workqueue("ipv6_addrconf");
7549 if (!addrconf_wq) {
7550 err = -ENOMEM;
7551 goto out_nowq;
7552 }
7553
7554 rtnl_net_lock(&init_net);
7555 idev = ipv6_add_dev(blackhole_netdev);
7556 rtnl_net_unlock(&init_net);
7557 if (IS_ERR(idev)) {
7558 err = PTR_ERR(idev);
7559 goto errlo;
7560 }
7561
7562 ip6_route_init_special_entries();
7563
7564 register_netdevice_notifier(&ipv6_dev_notf);
7565
7566 addrconf_verify(&init_net);
7567
7568 err = rtnl_af_register(&inet6_ops);
7569 if (err)
7570 goto erraf;
7571
7572 err = rtnl_register_many(addrconf_rtnl_msg_handlers);
7573 if (err)
7574 goto errout;
7575
7576 err = ipv6_addr_label_rtnl_register();
7577 if (err < 0)
7578 goto errout;
7579
7580 return 0;
7581 errout:
7582 rtnl_unregister_all(PF_INET6);
7583 rtnl_af_unregister(&inet6_ops);
7584 erraf:
7585 unregister_netdevice_notifier(&ipv6_dev_notf);
7586 errlo:
7587 destroy_workqueue(addrconf_wq);
7588 out_nowq:
7589 unregister_pernet_subsys(&addrconf_ops);
7590 out_addrlabel:
7591 ipv6_addr_label_cleanup();
7592 out:
7593 return err;
7594 }
7595
addrconf_cleanup(void)7596 void addrconf_cleanup(void)
7597 {
7598 struct net_device *dev;
7599
7600 unregister_netdevice_notifier(&ipv6_dev_notf);
7601 unregister_pernet_subsys(&addrconf_ops);
7602 ipv6_addr_label_cleanup();
7603
7604 rtnl_af_unregister(&inet6_ops);
7605
7606 rtnl_net_lock(&init_net);
7607
7608 /* clean dev list */
7609 for_each_netdev(&init_net, dev) {
7610 if (!__in6_dev_get_rtnl_net(dev))
7611 continue;
7612 addrconf_ifdown(dev, true);
7613 }
7614 addrconf_ifdown(init_net.loopback_dev, true);
7615
7616 rtnl_net_unlock(&init_net);
7617
7618 destroy_workqueue(addrconf_wq);
7619 }
7620