1 // SPDX-License-Identifier: GPL-2.0-or-later
2 /*
3 * IPv6 Address [auto]configuration
4 * Linux INET6 implementation
5 *
6 * Authors:
7 * Pedro Roque <roque@di.fc.ul.pt>
8 * Alexey Kuznetsov <kuznet@ms2.inr.ac.ru>
9 */
10
11 /*
12 * Changes:
13 *
14 * Janos Farkas : delete timer on ifdown
15 * <chexum@bankinf.banki.hu>
16 * Andi Kleen : kill double kfree on module
17 * unload.
18 * Maciej W. Rozycki : FDDI support
19 * sekiya@USAGI : Don't send too many RS
20 * packets.
21 * yoshfuji@USAGI : Fixed interval between DAD
22 * packets.
23 * YOSHIFUJI Hideaki @USAGI : improved accuracy of
24 * address validation timer.
25 * YOSHIFUJI Hideaki @USAGI : Privacy Extensions (RFC3041)
26 * support.
27 * Yuji SEKIYA @USAGI : Don't assign a same IPv6
28 * address on a same interface.
29 * YOSHIFUJI Hideaki @USAGI : ARCnet support
30 * YOSHIFUJI Hideaki @USAGI : convert /proc/net/if_inet6 to
31 * seq_file.
32 * YOSHIFUJI Hideaki @USAGI : improved source address
33 * selection; consider scope,
34 * status etc.
35 */
36
37 #define pr_fmt(fmt) "IPv6: " fmt
38
39 #include <linux/errno.h>
40 #include <linux/types.h>
41 #include <linux/kernel.h>
42 #include <linux/sched/signal.h>
43 #include <linux/socket.h>
44 #include <linux/sockios.h>
45 #include <linux/net.h>
46 #include <linux/inet.h>
47 #include <linux/in6.h>
48 #include <linux/netdevice.h>
49 #include <linux/if_addr.h>
50 #include <linux/if_arp.h>
51 #include <linux/if_arcnet.h>
52 #include <linux/if_infiniband.h>
53 #include <linux/route.h>
54 #include <linux/inetdevice.h>
55 #include <linux/init.h>
56 #include <linux/slab.h>
57 #ifdef CONFIG_SYSCTL
58 #include <linux/sysctl.h>
59 #endif
60 #include <linux/capability.h>
61 #include <linux/delay.h>
62 #include <linux/notifier.h>
63 #include <linux/string.h>
64 #include <linux/hash.h>
65
66 #include <net/net_namespace.h>
67 #include <net/sock.h>
68 #include <net/snmp.h>
69
70 #include <net/6lowpan.h>
71 #include <net/firewire.h>
72 #include <net/ipv6.h>
73 #include <net/protocol.h>
74 #include <net/ndisc.h>
75 #include <net/ip6_route.h>
76 #include <net/addrconf.h>
77 #include <net/tcp.h>
78 #include <net/ip.h>
79 #include <net/netlink.h>
80 #include <net/pkt_sched.h>
81 #include <net/l3mdev.h>
82 #include <linux/if_tunnel.h>
83 #include <linux/rtnetlink.h>
84 #include <linux/netconf.h>
85 #include <linux/random.h>
86 #include <linux/uaccess.h>
87 #include <asm/unaligned.h>
88
89 #include <linux/proc_fs.h>
90 #include <linux/seq_file.h>
91 #include <linux/export.h>
92 #include <linux/ioam6.h>
93
94 #define INFINITY_LIFE_TIME 0xFFFFFFFF
95
96 #define IPV6_MAX_STRLEN \
97 sizeof("ffff:ffff:ffff:ffff:ffff:ffff:255.255.255.255")
98
cstamp_delta(unsigned long cstamp)99 static inline u32 cstamp_delta(unsigned long cstamp)
100 {
101 return (cstamp - INITIAL_JIFFIES) * 100UL / HZ;
102 }
103
rfc3315_s14_backoff_init(s32 irt)104 static inline s32 rfc3315_s14_backoff_init(s32 irt)
105 {
106 /* multiply 'initial retransmission time' by 0.9 .. 1.1 */
107 u64 tmp = get_random_u32_inclusive(900000, 1100000) * (u64)irt;
108 do_div(tmp, 1000000);
109 return (s32)tmp;
110 }
111
rfc3315_s14_backoff_update(s32 rt,s32 mrt)112 static inline s32 rfc3315_s14_backoff_update(s32 rt, s32 mrt)
113 {
114 /* multiply 'retransmission timeout' by 1.9 .. 2.1 */
115 u64 tmp = get_random_u32_inclusive(1900000, 2100000) * (u64)rt;
116 do_div(tmp, 1000000);
117 if ((s32)tmp > mrt) {
118 /* multiply 'maximum retransmission time' by 0.9 .. 1.1 */
119 tmp = get_random_u32_inclusive(900000, 1100000) * (u64)mrt;
120 do_div(tmp, 1000000);
121 }
122 return (s32)tmp;
123 }
124
125 #ifdef CONFIG_SYSCTL
126 static int addrconf_sysctl_register(struct inet6_dev *idev);
127 static void addrconf_sysctl_unregister(struct inet6_dev *idev);
128 #else
addrconf_sysctl_register(struct inet6_dev * idev)129 static inline int addrconf_sysctl_register(struct inet6_dev *idev)
130 {
131 return 0;
132 }
133
addrconf_sysctl_unregister(struct inet6_dev * idev)134 static inline void addrconf_sysctl_unregister(struct inet6_dev *idev)
135 {
136 }
137 #endif
138
139 static void ipv6_gen_rnd_iid(struct in6_addr *addr);
140
141 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev);
142 static int ipv6_count_addresses(const struct inet6_dev *idev);
143 static int ipv6_generate_stable_address(struct in6_addr *addr,
144 u8 dad_count,
145 const struct inet6_dev *idev);
146
147 #define IN6_ADDR_HSIZE_SHIFT 8
148 #define IN6_ADDR_HSIZE (1 << IN6_ADDR_HSIZE_SHIFT)
149
150 static void addrconf_verify(struct net *net);
151 static void addrconf_verify_rtnl(struct net *net);
152
153 static struct workqueue_struct *addrconf_wq;
154
155 static void addrconf_join_anycast(struct inet6_ifaddr *ifp);
156 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp);
157
158 static void addrconf_type_change(struct net_device *dev,
159 unsigned long event);
160 static int addrconf_ifdown(struct net_device *dev, bool unregister);
161
162 static struct fib6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
163 int plen,
164 const struct net_device *dev,
165 u32 flags, u32 noflags,
166 bool no_gw);
167
168 static void addrconf_dad_start(struct inet6_ifaddr *ifp);
169 static void addrconf_dad_work(struct work_struct *w);
170 static void addrconf_dad_completed(struct inet6_ifaddr *ifp, bool bump_id,
171 bool send_na);
172 static void addrconf_dad_run(struct inet6_dev *idev, bool restart);
173 static void addrconf_rs_timer(struct timer_list *t);
174 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
175 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
176
177 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
178 struct prefix_info *pinfo);
179
180 static struct ipv6_devconf ipv6_devconf __read_mostly = {
181 .forwarding = 0,
182 .hop_limit = IPV6_DEFAULT_HOPLIMIT,
183 .mtu6 = IPV6_MIN_MTU,
184 .accept_ra = 1,
185 .accept_redirects = 1,
186 .autoconf = 1,
187 .force_mld_version = 0,
188 .mldv1_unsolicited_report_interval = 10 * HZ,
189 .mldv2_unsolicited_report_interval = HZ,
190 .dad_transmits = 1,
191 .rtr_solicits = MAX_RTR_SOLICITATIONS,
192 .rtr_solicit_interval = RTR_SOLICITATION_INTERVAL,
193 .rtr_solicit_max_interval = RTR_SOLICITATION_MAX_INTERVAL,
194 .rtr_solicit_delay = MAX_RTR_SOLICITATION_DELAY,
195 .use_tempaddr = 0,
196 .temp_valid_lft = TEMP_VALID_LIFETIME,
197 .temp_prefered_lft = TEMP_PREFERRED_LIFETIME,
198 .regen_max_retry = REGEN_MAX_RETRY,
199 .max_desync_factor = MAX_DESYNC_FACTOR,
200 .max_addresses = IPV6_MAX_ADDRESSES,
201 .accept_ra_defrtr = 1,
202 .ra_defrtr_metric = IP6_RT_PRIO_USER,
203 .accept_ra_from_local = 0,
204 .accept_ra_min_hop_limit= 1,
205 .accept_ra_min_lft = 0,
206 .accept_ra_pinfo = 1,
207 #ifdef CONFIG_IPV6_ROUTER_PREF
208 .accept_ra_rtr_pref = 1,
209 .rtr_probe_interval = 60 * HZ,
210 #ifdef CONFIG_IPV6_ROUTE_INFO
211 .accept_ra_rt_info_min_plen = 0,
212 .accept_ra_rt_info_max_plen = 0,
213 #endif
214 #endif
215 .accept_ra_rt_table = 0,
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 };
241
242 static struct ipv6_devconf ipv6_devconf_dflt __read_mostly = {
243 .forwarding = 0,
244 .hop_limit = IPV6_DEFAULT_HOPLIMIT,
245 .mtu6 = IPV6_MIN_MTU,
246 .accept_ra = 1,
247 .accept_redirects = 1,
248 .autoconf = 1,
249 .force_mld_version = 0,
250 .mldv1_unsolicited_report_interval = 10 * HZ,
251 .mldv2_unsolicited_report_interval = HZ,
252 .dad_transmits = 1,
253 .rtr_solicits = MAX_RTR_SOLICITATIONS,
254 .rtr_solicit_interval = RTR_SOLICITATION_INTERVAL,
255 .rtr_solicit_max_interval = RTR_SOLICITATION_MAX_INTERVAL,
256 .rtr_solicit_delay = MAX_RTR_SOLICITATION_DELAY,
257 .use_tempaddr = 0,
258 .temp_valid_lft = TEMP_VALID_LIFETIME,
259 .temp_prefered_lft = TEMP_PREFERRED_LIFETIME,
260 .regen_max_retry = REGEN_MAX_RETRY,
261 .max_desync_factor = MAX_DESYNC_FACTOR,
262 .max_addresses = IPV6_MAX_ADDRESSES,
263 .accept_ra_defrtr = 1,
264 .ra_defrtr_metric = IP6_RT_PRIO_USER,
265 .accept_ra_from_local = 0,
266 .accept_ra_min_hop_limit= 1,
267 .accept_ra_min_lft = 0,
268 .accept_ra_pinfo = 1,
269 #ifdef CONFIG_IPV6_ROUTER_PREF
270 .accept_ra_rtr_pref = 1,
271 .rtr_probe_interval = 60 * HZ,
272 #ifdef CONFIG_IPV6_ROUTE_INFO
273 .accept_ra_rt_info_min_plen = 0,
274 .accept_ra_rt_info_max_plen = 0,
275 #endif
276 #endif
277 .accept_ra_rt_table = 0,
278 .proxy_ndp = 0,
279 .accept_source_route = 0, /* we do not accept RH0 by default. */
280 .disable_ipv6 = 0,
281 .accept_dad = 1,
282 .suppress_frag_ndisc = 1,
283 .accept_ra_mtu = 1,
284 .stable_secret = {
285 .initialized = false,
286 },
287 .use_oif_addrs_only = 0,
288 .ignore_routes_with_linkdown = 0,
289 .keep_addr_on_down = 0,
290 .seg6_enabled = 0,
291 #ifdef CONFIG_IPV6_SEG6_HMAC
292 .seg6_require_hmac = 0,
293 #endif
294 .enhanced_dad = 1,
295 .addr_gen_mode = IN6_ADDR_GEN_MODE_EUI64,
296 .disable_policy = 0,
297 .rpl_seg_enabled = 0,
298 .ioam6_enabled = 0,
299 .ioam6_id = IOAM6_DEFAULT_IF_ID,
300 .ioam6_id_wide = IOAM6_DEFAULT_IF_ID_WIDE,
301 .ndisc_evict_nocarrier = 1,
302 };
303
304 /* Check if link is ready: is it up and is a valid qdisc available */
addrconf_link_ready(const struct net_device * dev)305 static inline bool addrconf_link_ready(const struct net_device *dev)
306 {
307 return netif_oper_up(dev) && !qdisc_tx_is_noop(dev);
308 }
309
addrconf_del_rs_timer(struct inet6_dev * idev)310 static void addrconf_del_rs_timer(struct inet6_dev *idev)
311 {
312 if (del_timer(&idev->rs_timer))
313 __in6_dev_put(idev);
314 }
315
addrconf_del_dad_work(struct inet6_ifaddr * ifp)316 static void addrconf_del_dad_work(struct inet6_ifaddr *ifp)
317 {
318 if (cancel_delayed_work(&ifp->dad_work))
319 __in6_ifa_put(ifp);
320 }
321
addrconf_mod_rs_timer(struct inet6_dev * idev,unsigned long when)322 static void addrconf_mod_rs_timer(struct inet6_dev *idev,
323 unsigned long when)
324 {
325 if (!mod_timer(&idev->rs_timer, jiffies + when))
326 in6_dev_hold(idev);
327 }
328
addrconf_mod_dad_work(struct inet6_ifaddr * ifp,unsigned long delay)329 static void addrconf_mod_dad_work(struct inet6_ifaddr *ifp,
330 unsigned long delay)
331 {
332 in6_ifa_hold(ifp);
333 if (mod_delayed_work(addrconf_wq, &ifp->dad_work, delay))
334 in6_ifa_put(ifp);
335 }
336
snmp6_alloc_dev(struct inet6_dev * idev)337 static int snmp6_alloc_dev(struct inet6_dev *idev)
338 {
339 int i;
340
341 idev->stats.ipv6 = alloc_percpu_gfp(struct ipstats_mib, GFP_KERNEL_ACCOUNT);
342 if (!idev->stats.ipv6)
343 goto err_ip;
344
345 for_each_possible_cpu(i) {
346 struct ipstats_mib *addrconf_stats;
347 addrconf_stats = per_cpu_ptr(idev->stats.ipv6, i);
348 u64_stats_init(&addrconf_stats->syncp);
349 }
350
351
352 idev->stats.icmpv6dev = kzalloc(sizeof(struct icmpv6_mib_device),
353 GFP_KERNEL);
354 if (!idev->stats.icmpv6dev)
355 goto err_icmp;
356 idev->stats.icmpv6msgdev = kzalloc(sizeof(struct icmpv6msg_mib_device),
357 GFP_KERNEL_ACCOUNT);
358 if (!idev->stats.icmpv6msgdev)
359 goto err_icmpmsg;
360
361 return 0;
362
363 err_icmpmsg:
364 kfree(idev->stats.icmpv6dev);
365 err_icmp:
366 free_percpu(idev->stats.ipv6);
367 err_ip:
368 return -ENOMEM;
369 }
370
ipv6_add_dev(struct net_device * dev)371 static struct inet6_dev *ipv6_add_dev(struct net_device *dev)
372 {
373 struct inet6_dev *ndev;
374 int err = -ENOMEM;
375
376 ASSERT_RTNL();
377
378 if (dev->mtu < IPV6_MIN_MTU && dev != blackhole_netdev)
379 return ERR_PTR(-EINVAL);
380
381 ndev = kzalloc(sizeof(*ndev), GFP_KERNEL_ACCOUNT);
382 if (!ndev)
383 return ERR_PTR(err);
384
385 rwlock_init(&ndev->lock);
386 ndev->dev = dev;
387 INIT_LIST_HEAD(&ndev->addr_list);
388 timer_setup(&ndev->rs_timer, addrconf_rs_timer, 0);
389 memcpy(&ndev->cnf, dev_net(dev)->ipv6.devconf_dflt, sizeof(ndev->cnf));
390
391 if (ndev->cnf.stable_secret.initialized)
392 ndev->cnf.addr_gen_mode = IN6_ADDR_GEN_MODE_STABLE_PRIVACY;
393
394 ndev->cnf.mtu6 = dev->mtu;
395 ndev->ra_mtu = 0;
396 ndev->nd_parms = neigh_parms_alloc(dev, &nd_tbl);
397 if (!ndev->nd_parms) {
398 kfree(ndev);
399 return ERR_PTR(err);
400 }
401 if (ndev->cnf.forwarding)
402 dev_disable_lro(dev);
403 /* We refer to the device */
404 netdev_hold(dev, &ndev->dev_tracker, GFP_KERNEL);
405
406 if (snmp6_alloc_dev(ndev) < 0) {
407 netdev_dbg(dev, "%s: cannot allocate memory for statistics\n",
408 __func__);
409 neigh_parms_release(&nd_tbl, ndev->nd_parms);
410 netdev_put(dev, &ndev->dev_tracker);
411 kfree(ndev);
412 return ERR_PTR(err);
413 }
414
415 if (dev != blackhole_netdev) {
416 if (snmp6_register_dev(ndev) < 0) {
417 netdev_dbg(dev, "%s: cannot create /proc/net/dev_snmp6/%s\n",
418 __func__, dev->name);
419 goto err_release;
420 }
421 }
422 /* One reference from device. */
423 refcount_set(&ndev->refcnt, 1);
424
425 if (dev->flags & (IFF_NOARP | IFF_LOOPBACK))
426 ndev->cnf.accept_dad = -1;
427
428 #if IS_ENABLED(CONFIG_IPV6_SIT)
429 if (dev->type == ARPHRD_SIT && (dev->priv_flags & IFF_ISATAP)) {
430 pr_info("%s: Disabled Multicast RS\n", dev->name);
431 ndev->cnf.rtr_solicits = 0;
432 }
433 #endif
434
435 INIT_LIST_HEAD(&ndev->tempaddr_list);
436 ndev->desync_factor = U32_MAX;
437 if ((dev->flags&IFF_LOOPBACK) ||
438 dev->type == ARPHRD_TUNNEL ||
439 dev->type == ARPHRD_TUNNEL6 ||
440 dev->type == ARPHRD_SIT ||
441 dev->type == ARPHRD_NONE) {
442 ndev->cnf.use_tempaddr = -1;
443 }
444
445 ndev->token = in6addr_any;
446
447 if (netif_running(dev) && addrconf_link_ready(dev))
448 ndev->if_flags |= IF_READY;
449
450 ipv6_mc_init_dev(ndev);
451 ndev->tstamp = jiffies;
452 if (dev != blackhole_netdev) {
453 err = addrconf_sysctl_register(ndev);
454 if (err) {
455 ipv6_mc_destroy_dev(ndev);
456 snmp6_unregister_dev(ndev);
457 goto err_release;
458 }
459 }
460 /* protected by rtnl_lock */
461 rcu_assign_pointer(dev->ip6_ptr, ndev);
462
463 if (dev != blackhole_netdev) {
464 /* Join interface-local all-node multicast group */
465 ipv6_dev_mc_inc(dev, &in6addr_interfacelocal_allnodes);
466
467 /* Join all-node multicast group */
468 ipv6_dev_mc_inc(dev, &in6addr_linklocal_allnodes);
469
470 /* Join all-router multicast group if forwarding is set */
471 if (ndev->cnf.forwarding && (dev->flags & IFF_MULTICAST))
472 ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
473 }
474 return ndev;
475
476 err_release:
477 neigh_parms_release(&nd_tbl, ndev->nd_parms);
478 ndev->dead = 1;
479 in6_dev_finish_destroy(ndev);
480 return ERR_PTR(err);
481 }
482
ipv6_find_idev(struct net_device * dev)483 static struct inet6_dev *ipv6_find_idev(struct net_device *dev)
484 {
485 struct inet6_dev *idev;
486
487 ASSERT_RTNL();
488
489 idev = __in6_dev_get(dev);
490 if (!idev) {
491 idev = ipv6_add_dev(dev);
492 if (IS_ERR(idev))
493 return idev;
494 }
495
496 if (dev->flags&IFF_UP)
497 ipv6_mc_up(idev);
498 return idev;
499 }
500
inet6_netconf_msgsize_devconf(int type)501 static int inet6_netconf_msgsize_devconf(int type)
502 {
503 int size = NLMSG_ALIGN(sizeof(struct netconfmsg))
504 + nla_total_size(4); /* NETCONFA_IFINDEX */
505 bool all = false;
506
507 if (type == NETCONFA_ALL)
508 all = true;
509
510 if (all || type == NETCONFA_FORWARDING)
511 size += nla_total_size(4);
512 #ifdef CONFIG_IPV6_MROUTE
513 if (all || type == NETCONFA_MC_FORWARDING)
514 size += nla_total_size(4);
515 #endif
516 if (all || type == NETCONFA_PROXY_NEIGH)
517 size += nla_total_size(4);
518
519 if (all || type == NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN)
520 size += nla_total_size(4);
521
522 return size;
523 }
524
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)525 static int inet6_netconf_fill_devconf(struct sk_buff *skb, int ifindex,
526 struct ipv6_devconf *devconf, u32 portid,
527 u32 seq, int event, unsigned int flags,
528 int type)
529 {
530 struct nlmsghdr *nlh;
531 struct netconfmsg *ncm;
532 bool all = false;
533
534 nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct netconfmsg),
535 flags);
536 if (!nlh)
537 return -EMSGSIZE;
538
539 if (type == NETCONFA_ALL)
540 all = true;
541
542 ncm = nlmsg_data(nlh);
543 ncm->ncm_family = AF_INET6;
544
545 if (nla_put_s32(skb, NETCONFA_IFINDEX, ifindex) < 0)
546 goto nla_put_failure;
547
548 if (!devconf)
549 goto out;
550
551 if ((all || type == NETCONFA_FORWARDING) &&
552 nla_put_s32(skb, NETCONFA_FORWARDING, devconf->forwarding) < 0)
553 goto nla_put_failure;
554 #ifdef CONFIG_IPV6_MROUTE
555 if ((all || type == NETCONFA_MC_FORWARDING) &&
556 nla_put_s32(skb, NETCONFA_MC_FORWARDING,
557 atomic_read(&devconf->mc_forwarding)) < 0)
558 goto nla_put_failure;
559 #endif
560 if ((all || type == NETCONFA_PROXY_NEIGH) &&
561 nla_put_s32(skb, NETCONFA_PROXY_NEIGH, devconf->proxy_ndp) < 0)
562 goto nla_put_failure;
563
564 if ((all || type == NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN) &&
565 nla_put_s32(skb, NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
566 devconf->ignore_routes_with_linkdown) < 0)
567 goto nla_put_failure;
568
569 out:
570 nlmsg_end(skb, nlh);
571 return 0;
572
573 nla_put_failure:
574 nlmsg_cancel(skb, nlh);
575 return -EMSGSIZE;
576 }
577
inet6_netconf_notify_devconf(struct net * net,int event,int type,int ifindex,struct ipv6_devconf * devconf)578 void inet6_netconf_notify_devconf(struct net *net, int event, int type,
579 int ifindex, struct ipv6_devconf *devconf)
580 {
581 struct sk_buff *skb;
582 int err = -ENOBUFS;
583
584 skb = nlmsg_new(inet6_netconf_msgsize_devconf(type), GFP_KERNEL);
585 if (!skb)
586 goto errout;
587
588 err = inet6_netconf_fill_devconf(skb, ifindex, devconf, 0, 0,
589 event, 0, type);
590 if (err < 0) {
591 /* -EMSGSIZE implies BUG in inet6_netconf_msgsize_devconf() */
592 WARN_ON(err == -EMSGSIZE);
593 kfree_skb(skb);
594 goto errout;
595 }
596 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_NETCONF, NULL, GFP_KERNEL);
597 return;
598 errout:
599 rtnl_set_sk_err(net, RTNLGRP_IPV6_NETCONF, err);
600 }
601
602 static const struct nla_policy devconf_ipv6_policy[NETCONFA_MAX+1] = {
603 [NETCONFA_IFINDEX] = { .len = sizeof(int) },
604 [NETCONFA_FORWARDING] = { .len = sizeof(int) },
605 [NETCONFA_PROXY_NEIGH] = { .len = sizeof(int) },
606 [NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN] = { .len = sizeof(int) },
607 };
608
inet6_netconf_valid_get_req(struct sk_buff * skb,const struct nlmsghdr * nlh,struct nlattr ** tb,struct netlink_ext_ack * extack)609 static int inet6_netconf_valid_get_req(struct sk_buff *skb,
610 const struct nlmsghdr *nlh,
611 struct nlattr **tb,
612 struct netlink_ext_ack *extack)
613 {
614 int i, err;
615
616 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(struct netconfmsg))) {
617 NL_SET_ERR_MSG_MOD(extack, "Invalid header for netconf get request");
618 return -EINVAL;
619 }
620
621 if (!netlink_strict_get_check(skb))
622 return nlmsg_parse_deprecated(nlh, sizeof(struct netconfmsg),
623 tb, NETCONFA_MAX,
624 devconf_ipv6_policy, extack);
625
626 err = nlmsg_parse_deprecated_strict(nlh, sizeof(struct netconfmsg),
627 tb, NETCONFA_MAX,
628 devconf_ipv6_policy, extack);
629 if (err)
630 return err;
631
632 for (i = 0; i <= NETCONFA_MAX; i++) {
633 if (!tb[i])
634 continue;
635
636 switch (i) {
637 case NETCONFA_IFINDEX:
638 break;
639 default:
640 NL_SET_ERR_MSG_MOD(extack, "Unsupported attribute in netconf get request");
641 return -EINVAL;
642 }
643 }
644
645 return 0;
646 }
647
inet6_netconf_get_devconf(struct sk_buff * in_skb,struct nlmsghdr * nlh,struct netlink_ext_ack * extack)648 static int inet6_netconf_get_devconf(struct sk_buff *in_skb,
649 struct nlmsghdr *nlh,
650 struct netlink_ext_ack *extack)
651 {
652 struct net *net = sock_net(in_skb->sk);
653 struct nlattr *tb[NETCONFA_MAX+1];
654 struct inet6_dev *in6_dev = NULL;
655 struct net_device *dev = NULL;
656 struct sk_buff *skb;
657 struct ipv6_devconf *devconf;
658 int ifindex;
659 int err;
660
661 err = inet6_netconf_valid_get_req(in_skb, nlh, tb, extack);
662 if (err < 0)
663 return err;
664
665 if (!tb[NETCONFA_IFINDEX])
666 return -EINVAL;
667
668 err = -EINVAL;
669 ifindex = nla_get_s32(tb[NETCONFA_IFINDEX]);
670 switch (ifindex) {
671 case NETCONFA_IFINDEX_ALL:
672 devconf = net->ipv6.devconf_all;
673 break;
674 case NETCONFA_IFINDEX_DEFAULT:
675 devconf = net->ipv6.devconf_dflt;
676 break;
677 default:
678 dev = dev_get_by_index(net, ifindex);
679 if (!dev)
680 return -EINVAL;
681 in6_dev = in6_dev_get(dev);
682 if (!in6_dev)
683 goto errout;
684 devconf = &in6_dev->cnf;
685 break;
686 }
687
688 err = -ENOBUFS;
689 skb = nlmsg_new(inet6_netconf_msgsize_devconf(NETCONFA_ALL), GFP_KERNEL);
690 if (!skb)
691 goto errout;
692
693 err = inet6_netconf_fill_devconf(skb, ifindex, devconf,
694 NETLINK_CB(in_skb).portid,
695 nlh->nlmsg_seq, RTM_NEWNETCONF, 0,
696 NETCONFA_ALL);
697 if (err < 0) {
698 /* -EMSGSIZE implies BUG in inet6_netconf_msgsize_devconf() */
699 WARN_ON(err == -EMSGSIZE);
700 kfree_skb(skb);
701 goto errout;
702 }
703 err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
704 errout:
705 if (in6_dev)
706 in6_dev_put(in6_dev);
707 dev_put(dev);
708 return err;
709 }
710
711 /* Combine dev_addr_genid and dev_base_seq to detect changes.
712 */
inet6_base_seq(const struct net * net)713 static u32 inet6_base_seq(const struct net *net)
714 {
715 u32 res = atomic_read(&net->ipv6.dev_addr_genid) +
716 net->dev_base_seq;
717
718 /* Must not return 0 (see nl_dump_check_consistent()).
719 * Chose a value far away from 0.
720 */
721 if (!res)
722 res = 0x80000000;
723 return res;
724 }
725
726
inet6_netconf_dump_devconf(struct sk_buff * skb,struct netlink_callback * cb)727 static int inet6_netconf_dump_devconf(struct sk_buff *skb,
728 struct netlink_callback *cb)
729 {
730 const struct nlmsghdr *nlh = cb->nlh;
731 struct net *net = sock_net(skb->sk);
732 int h, s_h;
733 int idx, s_idx;
734 struct net_device *dev;
735 struct inet6_dev *idev;
736 struct hlist_head *head;
737
738 if (cb->strict_check) {
739 struct netlink_ext_ack *extack = cb->extack;
740 struct netconfmsg *ncm;
741
742 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ncm))) {
743 NL_SET_ERR_MSG_MOD(extack, "Invalid header for netconf dump request");
744 return -EINVAL;
745 }
746
747 if (nlmsg_attrlen(nlh, sizeof(*ncm))) {
748 NL_SET_ERR_MSG_MOD(extack, "Invalid data after header in netconf dump request");
749 return -EINVAL;
750 }
751 }
752
753 s_h = cb->args[0];
754 s_idx = idx = cb->args[1];
755
756 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
757 idx = 0;
758 head = &net->dev_index_head[h];
759 rcu_read_lock();
760 cb->seq = inet6_base_seq(net);
761 hlist_for_each_entry_rcu(dev, head, index_hlist) {
762 if (idx < s_idx)
763 goto cont;
764 idev = __in6_dev_get(dev);
765 if (!idev)
766 goto cont;
767
768 if (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) < 0) {
775 rcu_read_unlock();
776 goto done;
777 }
778 nl_dump_check_consistent(cb, nlmsg_hdr(skb));
779 cont:
780 idx++;
781 }
782 rcu_read_unlock();
783 }
784 if (h == NETDEV_HASHENTRIES) {
785 if (inet6_netconf_fill_devconf(skb, NETCONFA_IFINDEX_ALL,
786 net->ipv6.devconf_all,
787 NETLINK_CB(cb->skb).portid,
788 nlh->nlmsg_seq,
789 RTM_NEWNETCONF, NLM_F_MULTI,
790 NETCONFA_ALL) < 0)
791 goto done;
792 else
793 h++;
794 }
795 if (h == NETDEV_HASHENTRIES + 1) {
796 if (inet6_netconf_fill_devconf(skb, NETCONFA_IFINDEX_DEFAULT,
797 net->ipv6.devconf_dflt,
798 NETLINK_CB(cb->skb).portid,
799 nlh->nlmsg_seq,
800 RTM_NEWNETCONF, NLM_F_MULTI,
801 NETCONFA_ALL) < 0)
802 goto done;
803 else
804 h++;
805 }
806 done:
807 cb->args[0] = h;
808 cb->args[1] = idx;
809
810 return skb->len;
811 }
812
813 #ifdef CONFIG_SYSCTL
dev_forward_change(struct inet6_dev * idev)814 static void dev_forward_change(struct inet6_dev *idev)
815 {
816 struct net_device *dev;
817 struct inet6_ifaddr *ifa;
818 LIST_HEAD(tmp_addr_list);
819
820 if (!idev)
821 return;
822 dev = idev->dev;
823 if (idev->cnf.forwarding)
824 dev_disable_lro(dev);
825 if (dev->flags & IFF_MULTICAST) {
826 if (idev->cnf.forwarding) {
827 ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
828 ipv6_dev_mc_inc(dev, &in6addr_interfacelocal_allrouters);
829 ipv6_dev_mc_inc(dev, &in6addr_sitelocal_allrouters);
830 } else {
831 ipv6_dev_mc_dec(dev, &in6addr_linklocal_allrouters);
832 ipv6_dev_mc_dec(dev, &in6addr_interfacelocal_allrouters);
833 ipv6_dev_mc_dec(dev, &in6addr_sitelocal_allrouters);
834 }
835 }
836
837 read_lock_bh(&idev->lock);
838 list_for_each_entry(ifa, &idev->addr_list, if_list) {
839 if (ifa->flags&IFA_F_TENTATIVE)
840 continue;
841 list_add_tail(&ifa->if_list_aux, &tmp_addr_list);
842 }
843 read_unlock_bh(&idev->lock);
844
845 while (!list_empty(&tmp_addr_list)) {
846 ifa = list_first_entry(&tmp_addr_list,
847 struct inet6_ifaddr, if_list_aux);
848 list_del(&ifa->if_list_aux);
849 if (idev->cnf.forwarding)
850 addrconf_join_anycast(ifa);
851 else
852 addrconf_leave_anycast(ifa);
853 }
854
855 inet6_netconf_notify_devconf(dev_net(dev), RTM_NEWNETCONF,
856 NETCONFA_FORWARDING,
857 dev->ifindex, &idev->cnf);
858 }
859
860
addrconf_forward_change(struct net * net,__s32 newf)861 static void addrconf_forward_change(struct net *net, __s32 newf)
862 {
863 struct net_device *dev;
864 struct inet6_dev *idev;
865
866 for_each_netdev(net, dev) {
867 idev = __in6_dev_get(dev);
868 if (idev) {
869 int changed = (!idev->cnf.forwarding) ^ (!newf);
870 idev->cnf.forwarding = newf;
871 if (changed)
872 dev_forward_change(idev);
873 }
874 }
875 }
876
addrconf_fixup_forwarding(struct ctl_table * table,int * p,int newf)877 static int addrconf_fixup_forwarding(struct ctl_table *table, int *p, int newf)
878 {
879 struct net *net;
880 int old;
881
882 if (!rtnl_trylock())
883 return restart_syscall();
884
885 net = (struct net *)table->extra2;
886 old = *p;
887 *p = newf;
888
889 if (p == &net->ipv6.devconf_dflt->forwarding) {
890 if ((!newf) ^ (!old))
891 inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
892 NETCONFA_FORWARDING,
893 NETCONFA_IFINDEX_DEFAULT,
894 net->ipv6.devconf_dflt);
895 rtnl_unlock();
896 return 0;
897 }
898
899 if (p == &net->ipv6.devconf_all->forwarding) {
900 int old_dflt = net->ipv6.devconf_dflt->forwarding;
901
902 net->ipv6.devconf_dflt->forwarding = newf;
903 if ((!newf) ^ (!old_dflt))
904 inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
905 NETCONFA_FORWARDING,
906 NETCONFA_IFINDEX_DEFAULT,
907 net->ipv6.devconf_dflt);
908
909 addrconf_forward_change(net, newf);
910 if ((!newf) ^ (!old))
911 inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
912 NETCONFA_FORWARDING,
913 NETCONFA_IFINDEX_ALL,
914 net->ipv6.devconf_all);
915 } else if ((!newf) ^ (!old))
916 dev_forward_change((struct inet6_dev *)table->extra1);
917 rtnl_unlock();
918
919 if (newf)
920 rt6_purge_dflt_routers(net);
921 return 1;
922 }
923
addrconf_linkdown_change(struct net * net,__s32 newf)924 static void addrconf_linkdown_change(struct net *net, __s32 newf)
925 {
926 struct net_device *dev;
927 struct inet6_dev *idev;
928
929 for_each_netdev(net, dev) {
930 idev = __in6_dev_get(dev);
931 if (idev) {
932 int changed = (!idev->cnf.ignore_routes_with_linkdown) ^ (!newf);
933
934 idev->cnf.ignore_routes_with_linkdown = newf;
935 if (changed)
936 inet6_netconf_notify_devconf(dev_net(dev),
937 RTM_NEWNETCONF,
938 NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
939 dev->ifindex,
940 &idev->cnf);
941 }
942 }
943 }
944
addrconf_fixup_linkdown(struct ctl_table * table,int * p,int newf)945 static int addrconf_fixup_linkdown(struct ctl_table *table, int *p, int newf)
946 {
947 struct net *net;
948 int old;
949
950 if (!rtnl_trylock())
951 return restart_syscall();
952
953 net = (struct net *)table->extra2;
954 old = *p;
955 *p = newf;
956
957 if (p == &net->ipv6.devconf_dflt->ignore_routes_with_linkdown) {
958 if ((!newf) ^ (!old))
959 inet6_netconf_notify_devconf(net,
960 RTM_NEWNETCONF,
961 NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
962 NETCONFA_IFINDEX_DEFAULT,
963 net->ipv6.devconf_dflt);
964 rtnl_unlock();
965 return 0;
966 }
967
968 if (p == &net->ipv6.devconf_all->ignore_routes_with_linkdown) {
969 net->ipv6.devconf_dflt->ignore_routes_with_linkdown = newf;
970 addrconf_linkdown_change(net, newf);
971 if ((!newf) ^ (!old))
972 inet6_netconf_notify_devconf(net,
973 RTM_NEWNETCONF,
974 NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
975 NETCONFA_IFINDEX_ALL,
976 net->ipv6.devconf_all);
977 }
978 rtnl_unlock();
979
980 return 1;
981 }
982
983 #endif
984
985 /* Nobody refers to this ifaddr, destroy it */
inet6_ifa_finish_destroy(struct inet6_ifaddr * ifp)986 void inet6_ifa_finish_destroy(struct inet6_ifaddr *ifp)
987 {
988 WARN_ON(!hlist_unhashed(&ifp->addr_lst));
989
990 #ifdef NET_REFCNT_DEBUG
991 pr_debug("%s\n", __func__);
992 #endif
993
994 in6_dev_put(ifp->idev);
995
996 if (cancel_delayed_work(&ifp->dad_work))
997 pr_notice("delayed DAD work was pending while freeing ifa=%p\n",
998 ifp);
999
1000 if (ifp->state != INET6_IFADDR_STATE_DEAD) {
1001 pr_warn("Freeing alive inet6 address %p\n", ifp);
1002 return;
1003 }
1004
1005 kfree_rcu(ifp, rcu);
1006 }
1007
1008 static void
ipv6_link_dev_addr(struct inet6_dev * idev,struct inet6_ifaddr * ifp)1009 ipv6_link_dev_addr(struct inet6_dev *idev, struct inet6_ifaddr *ifp)
1010 {
1011 struct list_head *p;
1012 int ifp_scope = ipv6_addr_src_scope(&ifp->addr);
1013
1014 /*
1015 * Each device address list is sorted in order of scope -
1016 * global before linklocal.
1017 */
1018 list_for_each(p, &idev->addr_list) {
1019 struct inet6_ifaddr *ifa
1020 = list_entry(p, struct inet6_ifaddr, if_list);
1021 if (ifp_scope >= ipv6_addr_src_scope(&ifa->addr))
1022 break;
1023 }
1024
1025 list_add_tail_rcu(&ifp->if_list, p);
1026 }
1027
inet6_addr_hash(const struct net * net,const struct in6_addr * addr)1028 static u32 inet6_addr_hash(const struct net *net, const struct in6_addr *addr)
1029 {
1030 u32 val = ipv6_addr_hash(addr) ^ net_hash_mix(net);
1031
1032 return hash_32(val, IN6_ADDR_HSIZE_SHIFT);
1033 }
1034
ipv6_chk_same_addr(struct net * net,const struct in6_addr * addr,struct net_device * dev,unsigned int hash)1035 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
1036 struct net_device *dev, unsigned int hash)
1037 {
1038 struct inet6_ifaddr *ifp;
1039
1040 hlist_for_each_entry(ifp, &net->ipv6.inet6_addr_lst[hash], addr_lst) {
1041 if (ipv6_addr_equal(&ifp->addr, addr)) {
1042 if (!dev || ifp->idev->dev == dev)
1043 return true;
1044 }
1045 }
1046 return false;
1047 }
1048
ipv6_add_addr_hash(struct net_device * dev,struct inet6_ifaddr * ifa)1049 static int ipv6_add_addr_hash(struct net_device *dev, struct inet6_ifaddr *ifa)
1050 {
1051 struct net *net = dev_net(dev);
1052 unsigned int hash = inet6_addr_hash(net, &ifa->addr);
1053 int err = 0;
1054
1055 spin_lock_bh(&net->ipv6.addrconf_hash_lock);
1056
1057 /* Ignore adding duplicate addresses on an interface */
1058 if (ipv6_chk_same_addr(net, &ifa->addr, dev, hash)) {
1059 netdev_dbg(dev, "ipv6_add_addr: already assigned\n");
1060 err = -EEXIST;
1061 } else {
1062 hlist_add_head_rcu(&ifa->addr_lst, &net->ipv6.inet6_addr_lst[hash]);
1063 }
1064
1065 spin_unlock_bh(&net->ipv6.addrconf_hash_lock);
1066
1067 return err;
1068 }
1069
1070 /* On success it returns ifp with increased reference count */
1071
1072 static struct inet6_ifaddr *
ipv6_add_addr(struct inet6_dev * idev,struct ifa6_config * cfg,bool can_block,struct netlink_ext_ack * extack)1073 ipv6_add_addr(struct inet6_dev *idev, struct ifa6_config *cfg,
1074 bool can_block, struct netlink_ext_ack *extack)
1075 {
1076 gfp_t gfp_flags = can_block ? GFP_KERNEL : GFP_ATOMIC;
1077 int addr_type = ipv6_addr_type(cfg->pfx);
1078 struct net *net = dev_net(idev->dev);
1079 struct inet6_ifaddr *ifa = NULL;
1080 struct fib6_info *f6i = NULL;
1081 int err = 0;
1082
1083 if (addr_type == IPV6_ADDR_ANY) {
1084 NL_SET_ERR_MSG_MOD(extack, "Invalid address");
1085 return ERR_PTR(-EADDRNOTAVAIL);
1086 } else if (addr_type & IPV6_ADDR_MULTICAST &&
1087 !(cfg->ifa_flags & IFA_F_MCAUTOJOIN)) {
1088 NL_SET_ERR_MSG_MOD(extack, "Cannot assign multicast address without \"IFA_F_MCAUTOJOIN\" flag");
1089 return ERR_PTR(-EADDRNOTAVAIL);
1090 } else if (!(idev->dev->flags & IFF_LOOPBACK) &&
1091 !netif_is_l3_master(idev->dev) &&
1092 addr_type & IPV6_ADDR_LOOPBACK) {
1093 NL_SET_ERR_MSG_MOD(extack, "Cannot assign loopback address on this device");
1094 return ERR_PTR(-EADDRNOTAVAIL);
1095 }
1096
1097 if (idev->dead) {
1098 NL_SET_ERR_MSG_MOD(extack, "device is going away");
1099 err = -ENODEV;
1100 goto out;
1101 }
1102
1103 if (idev->cnf.disable_ipv6) {
1104 NL_SET_ERR_MSG_MOD(extack, "IPv6 is disabled on this device");
1105 err = -EACCES;
1106 goto out;
1107 }
1108
1109 /* validator notifier needs to be blocking;
1110 * do not call in atomic context
1111 */
1112 if (can_block) {
1113 struct in6_validator_info i6vi = {
1114 .i6vi_addr = *cfg->pfx,
1115 .i6vi_dev = idev,
1116 .extack = extack,
1117 };
1118
1119 err = inet6addr_validator_notifier_call_chain(NETDEV_UP, &i6vi);
1120 err = notifier_to_errno(err);
1121 if (err < 0)
1122 goto out;
1123 }
1124
1125 ifa = kzalloc(sizeof(*ifa), gfp_flags | __GFP_ACCOUNT);
1126 if (!ifa) {
1127 err = -ENOBUFS;
1128 goto out;
1129 }
1130
1131 f6i = addrconf_f6i_alloc(net, idev, cfg->pfx, false, gfp_flags, extack);
1132 if (IS_ERR(f6i)) {
1133 err = PTR_ERR(f6i);
1134 f6i = NULL;
1135 goto out;
1136 }
1137
1138 neigh_parms_data_state_setall(idev->nd_parms);
1139
1140 ifa->addr = *cfg->pfx;
1141 if (cfg->peer_pfx)
1142 ifa->peer_addr = *cfg->peer_pfx;
1143
1144 spin_lock_init(&ifa->lock);
1145 INIT_DELAYED_WORK(&ifa->dad_work, addrconf_dad_work);
1146 INIT_HLIST_NODE(&ifa->addr_lst);
1147 ifa->scope = cfg->scope;
1148 ifa->prefix_len = cfg->plen;
1149 ifa->rt_priority = cfg->rt_priority;
1150 ifa->flags = cfg->ifa_flags;
1151 ifa->ifa_proto = cfg->ifa_proto;
1152 /* No need to add the TENTATIVE flag for addresses with NODAD */
1153 if (!(cfg->ifa_flags & IFA_F_NODAD))
1154 ifa->flags |= IFA_F_TENTATIVE;
1155 ifa->valid_lft = cfg->valid_lft;
1156 ifa->prefered_lft = cfg->preferred_lft;
1157 ifa->cstamp = ifa->tstamp = jiffies;
1158 ifa->tokenized = false;
1159
1160 ifa->rt = f6i;
1161
1162 ifa->idev = idev;
1163 in6_dev_hold(idev);
1164
1165 /* For caller */
1166 refcount_set(&ifa->refcnt, 1);
1167
1168 rcu_read_lock();
1169
1170 err = ipv6_add_addr_hash(idev->dev, ifa);
1171 if (err < 0) {
1172 rcu_read_unlock();
1173 goto out;
1174 }
1175
1176 write_lock_bh(&idev->lock);
1177
1178 /* Add to inet6_dev unicast addr list. */
1179 ipv6_link_dev_addr(idev, ifa);
1180
1181 if (ifa->flags&IFA_F_TEMPORARY) {
1182 list_add(&ifa->tmp_list, &idev->tempaddr_list);
1183 in6_ifa_hold(ifa);
1184 }
1185
1186 in6_ifa_hold(ifa);
1187 write_unlock_bh(&idev->lock);
1188
1189 rcu_read_unlock();
1190
1191 inet6addr_notifier_call_chain(NETDEV_UP, ifa);
1192 out:
1193 if (unlikely(err < 0)) {
1194 fib6_info_release(f6i);
1195
1196 if (ifa) {
1197 if (ifa->idev)
1198 in6_dev_put(ifa->idev);
1199 kfree(ifa);
1200 }
1201 ifa = ERR_PTR(err);
1202 }
1203
1204 return ifa;
1205 }
1206
1207 enum cleanup_prefix_rt_t {
1208 CLEANUP_PREFIX_RT_NOP, /* no cleanup action for prefix route */
1209 CLEANUP_PREFIX_RT_DEL, /* delete the prefix route */
1210 CLEANUP_PREFIX_RT_EXPIRE, /* update the lifetime of the prefix route */
1211 };
1212
1213 /*
1214 * Check, whether the prefix for ifp would still need a prefix route
1215 * after deleting ifp. The function returns one of the CLEANUP_PREFIX_RT_*
1216 * constants.
1217 *
1218 * 1) we don't purge prefix if address was not permanent.
1219 * prefix is managed by its own lifetime.
1220 * 2) we also don't purge, if the address was IFA_F_NOPREFIXROUTE.
1221 * 3) if there are no addresses, delete prefix.
1222 * 4) if there are still other permanent address(es),
1223 * corresponding prefix is still permanent.
1224 * 5) if there are still other addresses with IFA_F_NOPREFIXROUTE,
1225 * don't purge the prefix, assume user space is managing it.
1226 * 6) otherwise, update prefix lifetime to the
1227 * longest valid lifetime among the corresponding
1228 * addresses on the device.
1229 * Note: subsequent RA will update lifetime.
1230 **/
1231 static enum cleanup_prefix_rt_t
check_cleanup_prefix_route(struct inet6_ifaddr * ifp,unsigned long * expires)1232 check_cleanup_prefix_route(struct inet6_ifaddr *ifp, unsigned long *expires)
1233 {
1234 struct inet6_ifaddr *ifa;
1235 struct inet6_dev *idev = ifp->idev;
1236 unsigned long lifetime;
1237 enum cleanup_prefix_rt_t action = CLEANUP_PREFIX_RT_DEL;
1238
1239 *expires = jiffies;
1240
1241 list_for_each_entry(ifa, &idev->addr_list, if_list) {
1242 if (ifa == ifp)
1243 continue;
1244 if (ifa->prefix_len != ifp->prefix_len ||
1245 !ipv6_prefix_equal(&ifa->addr, &ifp->addr,
1246 ifp->prefix_len))
1247 continue;
1248 if (ifa->flags & (IFA_F_PERMANENT | IFA_F_NOPREFIXROUTE))
1249 return CLEANUP_PREFIX_RT_NOP;
1250
1251 action = CLEANUP_PREFIX_RT_EXPIRE;
1252
1253 spin_lock(&ifa->lock);
1254
1255 lifetime = addrconf_timeout_fixup(ifa->valid_lft, HZ);
1256 /*
1257 * Note: Because this address is
1258 * not permanent, lifetime <
1259 * LONG_MAX / HZ here.
1260 */
1261 if (time_before(*expires, ifa->tstamp + lifetime * HZ))
1262 *expires = ifa->tstamp + lifetime * HZ;
1263 spin_unlock(&ifa->lock);
1264 }
1265
1266 return action;
1267 }
1268
1269 static void
cleanup_prefix_route(struct inet6_ifaddr * ifp,unsigned long expires,bool del_rt,bool del_peer)1270 cleanup_prefix_route(struct inet6_ifaddr *ifp, unsigned long expires,
1271 bool del_rt, bool del_peer)
1272 {
1273 struct fib6_info *f6i;
1274
1275 f6i = addrconf_get_prefix_route(del_peer ? &ifp->peer_addr : &ifp->addr,
1276 ifp->prefix_len,
1277 ifp->idev->dev, 0, RTF_DEFAULT, true);
1278 if (f6i) {
1279 if (del_rt)
1280 ip6_del_rt(dev_net(ifp->idev->dev), f6i, false);
1281 else {
1282 if (!(f6i->fib6_flags & RTF_EXPIRES))
1283 fib6_set_expires(f6i, expires);
1284 fib6_info_release(f6i);
1285 }
1286 }
1287 }
1288
1289
1290 /* This function wants to get referenced ifp and releases it before return */
1291
ipv6_del_addr(struct inet6_ifaddr * ifp)1292 static void ipv6_del_addr(struct inet6_ifaddr *ifp)
1293 {
1294 enum cleanup_prefix_rt_t action = CLEANUP_PREFIX_RT_NOP;
1295 struct net *net = dev_net(ifp->idev->dev);
1296 unsigned long expires;
1297 int state;
1298
1299 ASSERT_RTNL();
1300
1301 spin_lock_bh(&ifp->lock);
1302 state = ifp->state;
1303 ifp->state = INET6_IFADDR_STATE_DEAD;
1304 spin_unlock_bh(&ifp->lock);
1305
1306 if (state == INET6_IFADDR_STATE_DEAD)
1307 goto out;
1308
1309 spin_lock_bh(&net->ipv6.addrconf_hash_lock);
1310 hlist_del_init_rcu(&ifp->addr_lst);
1311 spin_unlock_bh(&net->ipv6.addrconf_hash_lock);
1312
1313 write_lock_bh(&ifp->idev->lock);
1314
1315 if (ifp->flags&IFA_F_TEMPORARY) {
1316 list_del(&ifp->tmp_list);
1317 if (ifp->ifpub) {
1318 in6_ifa_put(ifp->ifpub);
1319 ifp->ifpub = NULL;
1320 }
1321 __in6_ifa_put(ifp);
1322 }
1323
1324 if (ifp->flags & IFA_F_PERMANENT && !(ifp->flags & IFA_F_NOPREFIXROUTE))
1325 action = check_cleanup_prefix_route(ifp, &expires);
1326
1327 list_del_rcu(&ifp->if_list);
1328 __in6_ifa_put(ifp);
1329
1330 write_unlock_bh(&ifp->idev->lock);
1331
1332 addrconf_del_dad_work(ifp);
1333
1334 ipv6_ifa_notify(RTM_DELADDR, ifp);
1335
1336 inet6addr_notifier_call_chain(NETDEV_DOWN, ifp);
1337
1338 if (action != CLEANUP_PREFIX_RT_NOP) {
1339 cleanup_prefix_route(ifp, expires,
1340 action == CLEANUP_PREFIX_RT_DEL, false);
1341 }
1342
1343 /* clean up prefsrc entries */
1344 rt6_remove_prefsrc(ifp);
1345 out:
1346 in6_ifa_put(ifp);
1347 }
1348
ipv6_create_tempaddr(struct inet6_ifaddr * ifp,bool block)1349 static int ipv6_create_tempaddr(struct inet6_ifaddr *ifp, bool block)
1350 {
1351 struct inet6_dev *idev = ifp->idev;
1352 unsigned long tmp_tstamp, age;
1353 unsigned long regen_advance;
1354 unsigned long now = jiffies;
1355 s32 cnf_temp_preferred_lft;
1356 struct inet6_ifaddr *ift;
1357 struct ifa6_config cfg;
1358 long max_desync_factor;
1359 struct in6_addr addr;
1360 int ret = 0;
1361
1362 write_lock_bh(&idev->lock);
1363
1364 retry:
1365 in6_dev_hold(idev);
1366 if (idev->cnf.use_tempaddr <= 0) {
1367 write_unlock_bh(&idev->lock);
1368 pr_info("%s: use_tempaddr is disabled\n", __func__);
1369 in6_dev_put(idev);
1370 ret = -1;
1371 goto out;
1372 }
1373 spin_lock_bh(&ifp->lock);
1374 if (ifp->regen_count++ >= idev->cnf.regen_max_retry) {
1375 idev->cnf.use_tempaddr = -1; /*XXX*/
1376 spin_unlock_bh(&ifp->lock);
1377 write_unlock_bh(&idev->lock);
1378 pr_warn("%s: regeneration time exceeded - disabled temporary address support\n",
1379 __func__);
1380 in6_dev_put(idev);
1381 ret = -1;
1382 goto out;
1383 }
1384 in6_ifa_hold(ifp);
1385 memcpy(addr.s6_addr, ifp->addr.s6_addr, 8);
1386 ipv6_gen_rnd_iid(&addr);
1387
1388 age = (now - ifp->tstamp) / HZ;
1389
1390 regen_advance = idev->cnf.regen_max_retry *
1391 idev->cnf.dad_transmits *
1392 max(NEIGH_VAR(idev->nd_parms, RETRANS_TIME), HZ/100) / HZ;
1393
1394 /* recalculate max_desync_factor each time and update
1395 * idev->desync_factor if it's larger
1396 */
1397 cnf_temp_preferred_lft = READ_ONCE(idev->cnf.temp_prefered_lft);
1398 max_desync_factor = min_t(long,
1399 idev->cnf.max_desync_factor,
1400 cnf_temp_preferred_lft - regen_advance);
1401
1402 if (unlikely(idev->desync_factor > max_desync_factor)) {
1403 if (max_desync_factor > 0) {
1404 get_random_bytes(&idev->desync_factor,
1405 sizeof(idev->desync_factor));
1406 idev->desync_factor %= max_desync_factor;
1407 } else {
1408 idev->desync_factor = 0;
1409 }
1410 }
1411
1412 memset(&cfg, 0, sizeof(cfg));
1413 cfg.valid_lft = min_t(__u32, ifp->valid_lft,
1414 idev->cnf.temp_valid_lft + age);
1415 cfg.preferred_lft = cnf_temp_preferred_lft + age - idev->desync_factor;
1416 cfg.preferred_lft = min_t(__u32, ifp->prefered_lft, cfg.preferred_lft);
1417
1418 cfg.plen = ifp->prefix_len;
1419 tmp_tstamp = ifp->tstamp;
1420 spin_unlock_bh(&ifp->lock);
1421
1422 write_unlock_bh(&idev->lock);
1423
1424 /* A temporary address is created only if this calculated Preferred
1425 * Lifetime is greater than REGEN_ADVANCE time units. In particular,
1426 * an implementation must not create a temporary address with a zero
1427 * Preferred Lifetime.
1428 * Use age calculation as in addrconf_verify to avoid unnecessary
1429 * temporary addresses being generated.
1430 */
1431 age = (now - tmp_tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
1432 if (cfg.preferred_lft <= regen_advance + age) {
1433 in6_ifa_put(ifp);
1434 in6_dev_put(idev);
1435 ret = -1;
1436 goto out;
1437 }
1438
1439 cfg.ifa_flags = IFA_F_TEMPORARY;
1440 /* set in addrconf_prefix_rcv() */
1441 if (ifp->flags & IFA_F_OPTIMISTIC)
1442 cfg.ifa_flags |= IFA_F_OPTIMISTIC;
1443
1444 cfg.pfx = &addr;
1445 cfg.scope = ipv6_addr_scope(cfg.pfx);
1446
1447 ift = ipv6_add_addr(idev, &cfg, block, NULL);
1448 if (IS_ERR(ift)) {
1449 in6_ifa_put(ifp);
1450 in6_dev_put(idev);
1451 pr_info("%s: retry temporary address regeneration\n", __func__);
1452 write_lock_bh(&idev->lock);
1453 goto retry;
1454 }
1455
1456 spin_lock_bh(&ift->lock);
1457 ift->ifpub = ifp;
1458 ift->cstamp = now;
1459 ift->tstamp = tmp_tstamp;
1460 spin_unlock_bh(&ift->lock);
1461
1462 addrconf_dad_start(ift);
1463 in6_ifa_put(ift);
1464 in6_dev_put(idev);
1465 out:
1466 return ret;
1467 }
1468
1469 /*
1470 * Choose an appropriate source address (RFC3484)
1471 */
1472 enum {
1473 IPV6_SADDR_RULE_INIT = 0,
1474 IPV6_SADDR_RULE_LOCAL,
1475 IPV6_SADDR_RULE_SCOPE,
1476 IPV6_SADDR_RULE_PREFERRED,
1477 #ifdef CONFIG_IPV6_MIP6
1478 IPV6_SADDR_RULE_HOA,
1479 #endif
1480 IPV6_SADDR_RULE_OIF,
1481 IPV6_SADDR_RULE_LABEL,
1482 IPV6_SADDR_RULE_PRIVACY,
1483 IPV6_SADDR_RULE_ORCHID,
1484 IPV6_SADDR_RULE_PREFIX,
1485 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1486 IPV6_SADDR_RULE_NOT_OPTIMISTIC,
1487 #endif
1488 IPV6_SADDR_RULE_MAX
1489 };
1490
1491 struct ipv6_saddr_score {
1492 int rule;
1493 int addr_type;
1494 struct inet6_ifaddr *ifa;
1495 DECLARE_BITMAP(scorebits, IPV6_SADDR_RULE_MAX);
1496 int scopedist;
1497 int matchlen;
1498 };
1499
1500 struct ipv6_saddr_dst {
1501 const struct in6_addr *addr;
1502 int ifindex;
1503 int scope;
1504 int label;
1505 unsigned int prefs;
1506 };
1507
ipv6_saddr_preferred(int type)1508 static inline int ipv6_saddr_preferred(int type)
1509 {
1510 if (type & (IPV6_ADDR_MAPPED|IPV6_ADDR_COMPATv4|IPV6_ADDR_LOOPBACK))
1511 return 1;
1512 return 0;
1513 }
1514
ipv6_use_optimistic_addr(struct net * net,struct inet6_dev * idev)1515 static bool ipv6_use_optimistic_addr(struct net *net,
1516 struct inet6_dev *idev)
1517 {
1518 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1519 if (!idev)
1520 return false;
1521 if (!net->ipv6.devconf_all->optimistic_dad && !idev->cnf.optimistic_dad)
1522 return false;
1523 if (!net->ipv6.devconf_all->use_optimistic && !idev->cnf.use_optimistic)
1524 return false;
1525
1526 return true;
1527 #else
1528 return false;
1529 #endif
1530 }
1531
ipv6_allow_optimistic_dad(struct net * net,struct inet6_dev * idev)1532 static bool ipv6_allow_optimistic_dad(struct net *net,
1533 struct inet6_dev *idev)
1534 {
1535 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1536 if (!idev)
1537 return false;
1538 if (!net->ipv6.devconf_all->optimistic_dad && !idev->cnf.optimistic_dad)
1539 return false;
1540
1541 return true;
1542 #else
1543 return false;
1544 #endif
1545 }
1546
ipv6_get_saddr_eval(struct net * net,struct ipv6_saddr_score * score,struct ipv6_saddr_dst * dst,int i)1547 static int ipv6_get_saddr_eval(struct net *net,
1548 struct ipv6_saddr_score *score,
1549 struct ipv6_saddr_dst *dst,
1550 int i)
1551 {
1552 int ret;
1553
1554 if (i <= score->rule) {
1555 switch (i) {
1556 case IPV6_SADDR_RULE_SCOPE:
1557 ret = score->scopedist;
1558 break;
1559 case IPV6_SADDR_RULE_PREFIX:
1560 ret = score->matchlen;
1561 break;
1562 default:
1563 ret = !!test_bit(i, score->scorebits);
1564 }
1565 goto out;
1566 }
1567
1568 switch (i) {
1569 case IPV6_SADDR_RULE_INIT:
1570 /* Rule 0: remember if hiscore is not ready yet */
1571 ret = !!score->ifa;
1572 break;
1573 case IPV6_SADDR_RULE_LOCAL:
1574 /* Rule 1: Prefer same address */
1575 ret = ipv6_addr_equal(&score->ifa->addr, dst->addr);
1576 break;
1577 case IPV6_SADDR_RULE_SCOPE:
1578 /* Rule 2: Prefer appropriate scope
1579 *
1580 * ret
1581 * ^
1582 * -1 | d 15
1583 * ---+--+-+---> scope
1584 * |
1585 * | d is scope of the destination.
1586 * B-d | \
1587 * | \ <- smaller scope is better if
1588 * B-15 | \ if scope is enough for destination.
1589 * | ret = B - scope (-1 <= scope >= d <= 15).
1590 * d-C-1 | /
1591 * |/ <- greater is better
1592 * -C / if scope is not enough for destination.
1593 * /| ret = scope - C (-1 <= d < scope <= 15).
1594 *
1595 * d - C - 1 < B -15 (for all -1 <= d <= 15).
1596 * C > d + 14 - B >= 15 + 14 - B = 29 - B.
1597 * Assume B = 0 and we get C > 29.
1598 */
1599 ret = __ipv6_addr_src_scope(score->addr_type);
1600 if (ret >= dst->scope)
1601 ret = -ret;
1602 else
1603 ret -= 128; /* 30 is enough */
1604 score->scopedist = ret;
1605 break;
1606 case IPV6_SADDR_RULE_PREFERRED:
1607 {
1608 /* Rule 3: Avoid deprecated and optimistic addresses */
1609 u8 avoid = IFA_F_DEPRECATED;
1610
1611 if (!ipv6_use_optimistic_addr(net, score->ifa->idev))
1612 avoid |= IFA_F_OPTIMISTIC;
1613 ret = ipv6_saddr_preferred(score->addr_type) ||
1614 !(score->ifa->flags & avoid);
1615 break;
1616 }
1617 #ifdef CONFIG_IPV6_MIP6
1618 case IPV6_SADDR_RULE_HOA:
1619 {
1620 /* Rule 4: Prefer home address */
1621 int prefhome = !(dst->prefs & IPV6_PREFER_SRC_COA);
1622 ret = !(score->ifa->flags & IFA_F_HOMEADDRESS) ^ prefhome;
1623 break;
1624 }
1625 #endif
1626 case IPV6_SADDR_RULE_OIF:
1627 /* Rule 5: Prefer outgoing interface */
1628 ret = (!dst->ifindex ||
1629 dst->ifindex == score->ifa->idev->dev->ifindex);
1630 break;
1631 case IPV6_SADDR_RULE_LABEL:
1632 /* Rule 6: Prefer matching label */
1633 ret = ipv6_addr_label(net,
1634 &score->ifa->addr, score->addr_type,
1635 score->ifa->idev->dev->ifindex) == dst->label;
1636 break;
1637 case IPV6_SADDR_RULE_PRIVACY:
1638 {
1639 /* Rule 7: Prefer public address
1640 * Note: prefer temporary address if use_tempaddr >= 2
1641 */
1642 int preftmp = dst->prefs & (IPV6_PREFER_SRC_PUBLIC|IPV6_PREFER_SRC_TMP) ?
1643 !!(dst->prefs & IPV6_PREFER_SRC_TMP) :
1644 score->ifa->idev->cnf.use_tempaddr >= 2;
1645 ret = (!(score->ifa->flags & IFA_F_TEMPORARY)) ^ preftmp;
1646 break;
1647 }
1648 case IPV6_SADDR_RULE_ORCHID:
1649 /* Rule 8-: Prefer ORCHID vs ORCHID or
1650 * non-ORCHID vs non-ORCHID
1651 */
1652 ret = !(ipv6_addr_orchid(&score->ifa->addr) ^
1653 ipv6_addr_orchid(dst->addr));
1654 break;
1655 case IPV6_SADDR_RULE_PREFIX:
1656 /* Rule 8: Use longest matching prefix */
1657 ret = ipv6_addr_diff(&score->ifa->addr, dst->addr);
1658 if (ret > score->ifa->prefix_len)
1659 ret = score->ifa->prefix_len;
1660 score->matchlen = ret;
1661 break;
1662 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1663 case IPV6_SADDR_RULE_NOT_OPTIMISTIC:
1664 /* Optimistic addresses still have lower precedence than other
1665 * preferred addresses.
1666 */
1667 ret = !(score->ifa->flags & IFA_F_OPTIMISTIC);
1668 break;
1669 #endif
1670 default:
1671 ret = 0;
1672 }
1673
1674 if (ret)
1675 __set_bit(i, score->scorebits);
1676 score->rule = i;
1677 out:
1678 return ret;
1679 }
1680
__ipv6_dev_get_saddr(struct net * net,struct ipv6_saddr_dst * dst,struct inet6_dev * idev,struct ipv6_saddr_score * scores,int hiscore_idx)1681 static int __ipv6_dev_get_saddr(struct net *net,
1682 struct ipv6_saddr_dst *dst,
1683 struct inet6_dev *idev,
1684 struct ipv6_saddr_score *scores,
1685 int hiscore_idx)
1686 {
1687 struct ipv6_saddr_score *score = &scores[1 - hiscore_idx], *hiscore = &scores[hiscore_idx];
1688
1689 list_for_each_entry_rcu(score->ifa, &idev->addr_list, if_list) {
1690 int i;
1691
1692 /*
1693 * - Tentative Address (RFC2462 section 5.4)
1694 * - A tentative address is not considered
1695 * "assigned to an interface" in the traditional
1696 * sense, unless it is also flagged as optimistic.
1697 * - Candidate Source Address (section 4)
1698 * - In any case, anycast addresses, multicast
1699 * addresses, and the unspecified address MUST
1700 * NOT be included in a candidate set.
1701 */
1702 if ((score->ifa->flags & IFA_F_TENTATIVE) &&
1703 (!(score->ifa->flags & IFA_F_OPTIMISTIC)))
1704 continue;
1705
1706 score->addr_type = __ipv6_addr_type(&score->ifa->addr);
1707
1708 if (unlikely(score->addr_type == IPV6_ADDR_ANY ||
1709 score->addr_type & IPV6_ADDR_MULTICAST)) {
1710 net_dbg_ratelimited("ADDRCONF: unspecified / multicast address assigned as unicast address on %s",
1711 idev->dev->name);
1712 continue;
1713 }
1714
1715 score->rule = -1;
1716 bitmap_zero(score->scorebits, IPV6_SADDR_RULE_MAX);
1717
1718 for (i = 0; i < IPV6_SADDR_RULE_MAX; i++) {
1719 int minihiscore, miniscore;
1720
1721 minihiscore = ipv6_get_saddr_eval(net, hiscore, dst, i);
1722 miniscore = ipv6_get_saddr_eval(net, score, dst, i);
1723
1724 if (minihiscore > miniscore) {
1725 if (i == IPV6_SADDR_RULE_SCOPE &&
1726 score->scopedist > 0) {
1727 /*
1728 * special case:
1729 * each remaining entry
1730 * has too small (not enough)
1731 * scope, because ifa entries
1732 * are sorted by their scope
1733 * values.
1734 */
1735 goto out;
1736 }
1737 break;
1738 } else if (minihiscore < miniscore) {
1739 swap(hiscore, score);
1740 hiscore_idx = 1 - hiscore_idx;
1741
1742 /* restore our iterator */
1743 score->ifa = hiscore->ifa;
1744
1745 break;
1746 }
1747 }
1748 }
1749 out:
1750 return hiscore_idx;
1751 }
1752
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)1753 static int ipv6_get_saddr_master(struct net *net,
1754 const struct net_device *dst_dev,
1755 const struct net_device *master,
1756 struct ipv6_saddr_dst *dst,
1757 struct ipv6_saddr_score *scores,
1758 int hiscore_idx)
1759 {
1760 struct inet6_dev *idev;
1761
1762 idev = __in6_dev_get(dst_dev);
1763 if (idev)
1764 hiscore_idx = __ipv6_dev_get_saddr(net, dst, idev,
1765 scores, hiscore_idx);
1766
1767 idev = __in6_dev_get(master);
1768 if (idev)
1769 hiscore_idx = __ipv6_dev_get_saddr(net, dst, idev,
1770 scores, hiscore_idx);
1771
1772 return hiscore_idx;
1773 }
1774
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)1775 int ipv6_dev_get_saddr(struct net *net, const struct net_device *dst_dev,
1776 const struct in6_addr *daddr, unsigned int prefs,
1777 struct in6_addr *saddr)
1778 {
1779 struct ipv6_saddr_score scores[2], *hiscore;
1780 struct ipv6_saddr_dst dst;
1781 struct inet6_dev *idev;
1782 struct net_device *dev;
1783 int dst_type;
1784 bool use_oif_addr = false;
1785 int hiscore_idx = 0;
1786 int ret = 0;
1787
1788 dst_type = __ipv6_addr_type(daddr);
1789 dst.addr = daddr;
1790 dst.ifindex = dst_dev ? dst_dev->ifindex : 0;
1791 dst.scope = __ipv6_addr_src_scope(dst_type);
1792 dst.label = ipv6_addr_label(net, daddr, dst_type, dst.ifindex);
1793 dst.prefs = prefs;
1794
1795 scores[hiscore_idx].rule = -1;
1796 scores[hiscore_idx].ifa = NULL;
1797
1798 rcu_read_lock();
1799
1800 /* Candidate Source Address (section 4)
1801 * - multicast and link-local destination address,
1802 * the set of candidate source address MUST only
1803 * include addresses assigned to interfaces
1804 * belonging to the same link as the outgoing
1805 * interface.
1806 * (- For site-local destination addresses, the
1807 * set of candidate source addresses MUST only
1808 * include addresses assigned to interfaces
1809 * belonging to the same site as the outgoing
1810 * interface.)
1811 * - "It is RECOMMENDED that the candidate source addresses
1812 * be the set of unicast addresses assigned to the
1813 * interface that will be used to send to the destination
1814 * (the 'outgoing' interface)." (RFC 6724)
1815 */
1816 if (dst_dev) {
1817 idev = __in6_dev_get(dst_dev);
1818 if ((dst_type & IPV6_ADDR_MULTICAST) ||
1819 dst.scope <= IPV6_ADDR_SCOPE_LINKLOCAL ||
1820 (idev && idev->cnf.use_oif_addrs_only)) {
1821 use_oif_addr = true;
1822 }
1823 }
1824
1825 if (use_oif_addr) {
1826 if (idev)
1827 hiscore_idx = __ipv6_dev_get_saddr(net, &dst, idev, scores, hiscore_idx);
1828 } else {
1829 const struct net_device *master;
1830 int master_idx = 0;
1831
1832 /* if dst_dev exists and is enslaved to an L3 device, then
1833 * prefer addresses from dst_dev and then the master over
1834 * any other enslaved devices in the L3 domain.
1835 */
1836 master = l3mdev_master_dev_rcu(dst_dev);
1837 if (master) {
1838 master_idx = master->ifindex;
1839
1840 hiscore_idx = ipv6_get_saddr_master(net, dst_dev,
1841 master, &dst,
1842 scores, hiscore_idx);
1843
1844 if (scores[hiscore_idx].ifa &&
1845 scores[hiscore_idx].scopedist >= 0)
1846 goto out;
1847 }
1848
1849 for_each_netdev_rcu(net, dev) {
1850 /* only consider addresses on devices in the
1851 * same L3 domain
1852 */
1853 if (l3mdev_master_ifindex_rcu(dev) != master_idx)
1854 continue;
1855 idev = __in6_dev_get(dev);
1856 if (!idev)
1857 continue;
1858 hiscore_idx = __ipv6_dev_get_saddr(net, &dst, idev, scores, hiscore_idx);
1859 }
1860 }
1861
1862 out:
1863 hiscore = &scores[hiscore_idx];
1864 if (!hiscore->ifa)
1865 ret = -EADDRNOTAVAIL;
1866 else
1867 *saddr = hiscore->ifa->addr;
1868
1869 rcu_read_unlock();
1870 return ret;
1871 }
1872 EXPORT_SYMBOL(ipv6_dev_get_saddr);
1873
__ipv6_get_lladdr(struct inet6_dev * idev,struct in6_addr * addr,u32 banned_flags)1874 static int __ipv6_get_lladdr(struct inet6_dev *idev, struct in6_addr *addr,
1875 u32 banned_flags)
1876 {
1877 struct inet6_ifaddr *ifp;
1878 int err = -EADDRNOTAVAIL;
1879
1880 list_for_each_entry_reverse(ifp, &idev->addr_list, if_list) {
1881 if (ifp->scope > IFA_LINK)
1882 break;
1883 if (ifp->scope == IFA_LINK &&
1884 !(ifp->flags & banned_flags)) {
1885 *addr = ifp->addr;
1886 err = 0;
1887 break;
1888 }
1889 }
1890 return err;
1891 }
1892
ipv6_get_lladdr(struct net_device * dev,struct in6_addr * addr,u32 banned_flags)1893 int ipv6_get_lladdr(struct net_device *dev, struct in6_addr *addr,
1894 u32 banned_flags)
1895 {
1896 struct inet6_dev *idev;
1897 int err = -EADDRNOTAVAIL;
1898
1899 rcu_read_lock();
1900 idev = __in6_dev_get(dev);
1901 if (idev) {
1902 read_lock_bh(&idev->lock);
1903 err = __ipv6_get_lladdr(idev, addr, banned_flags);
1904 read_unlock_bh(&idev->lock);
1905 }
1906 rcu_read_unlock();
1907 return err;
1908 }
1909
ipv6_count_addresses(const struct inet6_dev * idev)1910 static int ipv6_count_addresses(const struct inet6_dev *idev)
1911 {
1912 const struct inet6_ifaddr *ifp;
1913 int cnt = 0;
1914
1915 rcu_read_lock();
1916 list_for_each_entry_rcu(ifp, &idev->addr_list, if_list)
1917 cnt++;
1918 rcu_read_unlock();
1919 return cnt;
1920 }
1921
ipv6_chk_addr(struct net * net,const struct in6_addr * addr,const struct net_device * dev,int strict)1922 int ipv6_chk_addr(struct net *net, const struct in6_addr *addr,
1923 const struct net_device *dev, int strict)
1924 {
1925 return ipv6_chk_addr_and_flags(net, addr, dev, !dev,
1926 strict, IFA_F_TENTATIVE);
1927 }
1928 EXPORT_SYMBOL(ipv6_chk_addr);
1929
1930 /* device argument is used to find the L3 domain of interest. If
1931 * skip_dev_check is set, then the ifp device is not checked against
1932 * the passed in dev argument. So the 2 cases for addresses checks are:
1933 * 1. does the address exist in the L3 domain that dev is part of
1934 * (skip_dev_check = true), or
1935 *
1936 * 2. does the address exist on the specific device
1937 * (skip_dev_check = false)
1938 */
1939 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)1940 __ipv6_chk_addr_and_flags(struct net *net, const struct in6_addr *addr,
1941 const struct net_device *dev, bool skip_dev_check,
1942 int strict, u32 banned_flags)
1943 {
1944 unsigned int hash = inet6_addr_hash(net, addr);
1945 struct net_device *l3mdev, *ndev;
1946 struct inet6_ifaddr *ifp;
1947 u32 ifp_flags;
1948
1949 rcu_read_lock();
1950
1951 l3mdev = l3mdev_master_dev_rcu(dev);
1952 if (skip_dev_check)
1953 dev = NULL;
1954
1955 hlist_for_each_entry_rcu(ifp, &net->ipv6.inet6_addr_lst[hash], addr_lst) {
1956 ndev = ifp->idev->dev;
1957
1958 if (l3mdev_master_dev_rcu(ndev) != l3mdev)
1959 continue;
1960
1961 /* Decouple optimistic from tentative for evaluation here.
1962 * Ban optimistic addresses explicitly, when required.
1963 */
1964 ifp_flags = (ifp->flags&IFA_F_OPTIMISTIC)
1965 ? (ifp->flags&~IFA_F_TENTATIVE)
1966 : ifp->flags;
1967 if (ipv6_addr_equal(&ifp->addr, addr) &&
1968 !(ifp_flags&banned_flags) &&
1969 (!dev || ndev == dev ||
1970 !(ifp->scope&(IFA_LINK|IFA_HOST) || strict))) {
1971 rcu_read_unlock();
1972 return ndev;
1973 }
1974 }
1975
1976 rcu_read_unlock();
1977 return NULL;
1978 }
1979
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)1980 int ipv6_chk_addr_and_flags(struct net *net, const struct in6_addr *addr,
1981 const struct net_device *dev, bool skip_dev_check,
1982 int strict, u32 banned_flags)
1983 {
1984 return __ipv6_chk_addr_and_flags(net, addr, dev, skip_dev_check,
1985 strict, banned_flags) ? 1 : 0;
1986 }
1987 EXPORT_SYMBOL(ipv6_chk_addr_and_flags);
1988
1989
1990 /* Compares an address/prefix_len with addresses on device @dev.
1991 * If one is found it returns true.
1992 */
ipv6_chk_custom_prefix(const struct in6_addr * addr,const unsigned int prefix_len,struct net_device * dev)1993 bool ipv6_chk_custom_prefix(const struct in6_addr *addr,
1994 const unsigned int prefix_len, struct net_device *dev)
1995 {
1996 const struct inet6_ifaddr *ifa;
1997 const struct inet6_dev *idev;
1998 bool ret = false;
1999
2000 rcu_read_lock();
2001 idev = __in6_dev_get(dev);
2002 if (idev) {
2003 list_for_each_entry_rcu(ifa, &idev->addr_list, if_list) {
2004 ret = ipv6_prefix_equal(addr, &ifa->addr, prefix_len);
2005 if (ret)
2006 break;
2007 }
2008 }
2009 rcu_read_unlock();
2010
2011 return ret;
2012 }
2013 EXPORT_SYMBOL(ipv6_chk_custom_prefix);
2014
ipv6_chk_prefix(const struct in6_addr * addr,struct net_device * dev)2015 int ipv6_chk_prefix(const struct in6_addr *addr, struct net_device *dev)
2016 {
2017 const struct inet6_ifaddr *ifa;
2018 const struct inet6_dev *idev;
2019 int onlink;
2020
2021 onlink = 0;
2022 rcu_read_lock();
2023 idev = __in6_dev_get(dev);
2024 if (idev) {
2025 list_for_each_entry_rcu(ifa, &idev->addr_list, if_list) {
2026 onlink = ipv6_prefix_equal(addr, &ifa->addr,
2027 ifa->prefix_len);
2028 if (onlink)
2029 break;
2030 }
2031 }
2032 rcu_read_unlock();
2033 return onlink;
2034 }
2035 EXPORT_SYMBOL(ipv6_chk_prefix);
2036
2037 /**
2038 * ipv6_dev_find - find the first device with a given source address.
2039 * @net: the net namespace
2040 * @addr: the source address
2041 * @dev: used to find the L3 domain of interest
2042 *
2043 * The caller should be protected by RCU, or RTNL.
2044 */
ipv6_dev_find(struct net * net,const struct in6_addr * addr,struct net_device * dev)2045 struct net_device *ipv6_dev_find(struct net *net, const struct in6_addr *addr,
2046 struct net_device *dev)
2047 {
2048 return __ipv6_chk_addr_and_flags(net, addr, dev, !dev, 1,
2049 IFA_F_TENTATIVE);
2050 }
2051 EXPORT_SYMBOL(ipv6_dev_find);
2052
ipv6_get_ifaddr(struct net * net,const struct in6_addr * addr,struct net_device * dev,int strict)2053 struct inet6_ifaddr *ipv6_get_ifaddr(struct net *net, const struct in6_addr *addr,
2054 struct net_device *dev, int strict)
2055 {
2056 unsigned int hash = inet6_addr_hash(net, addr);
2057 struct inet6_ifaddr *ifp, *result = NULL;
2058
2059 rcu_read_lock();
2060 hlist_for_each_entry_rcu(ifp, &net->ipv6.inet6_addr_lst[hash], addr_lst) {
2061 if (ipv6_addr_equal(&ifp->addr, addr)) {
2062 if (!dev || ifp->idev->dev == dev ||
2063 !(ifp->scope&(IFA_LINK|IFA_HOST) || strict)) {
2064 if (in6_ifa_hold_safe(ifp)) {
2065 result = ifp;
2066 break;
2067 }
2068 }
2069 }
2070 }
2071 rcu_read_unlock();
2072
2073 return result;
2074 }
2075
2076 /* Gets referenced address, destroys ifaddr */
2077
addrconf_dad_stop(struct inet6_ifaddr * ifp,int dad_failed)2078 static void addrconf_dad_stop(struct inet6_ifaddr *ifp, int dad_failed)
2079 {
2080 if (dad_failed)
2081 ifp->flags |= IFA_F_DADFAILED;
2082
2083 if (ifp->flags&IFA_F_TEMPORARY) {
2084 struct inet6_ifaddr *ifpub;
2085 spin_lock_bh(&ifp->lock);
2086 ifpub = ifp->ifpub;
2087 if (ifpub) {
2088 in6_ifa_hold(ifpub);
2089 spin_unlock_bh(&ifp->lock);
2090 ipv6_create_tempaddr(ifpub, true);
2091 in6_ifa_put(ifpub);
2092 } else {
2093 spin_unlock_bh(&ifp->lock);
2094 }
2095 ipv6_del_addr(ifp);
2096 } else if (ifp->flags&IFA_F_PERMANENT || !dad_failed) {
2097 spin_lock_bh(&ifp->lock);
2098 addrconf_del_dad_work(ifp);
2099 ifp->flags |= IFA_F_TENTATIVE;
2100 if (dad_failed)
2101 ifp->flags &= ~IFA_F_OPTIMISTIC;
2102 spin_unlock_bh(&ifp->lock);
2103 if (dad_failed)
2104 ipv6_ifa_notify(0, ifp);
2105 in6_ifa_put(ifp);
2106 } else {
2107 ipv6_del_addr(ifp);
2108 }
2109 }
2110
addrconf_dad_end(struct inet6_ifaddr * ifp)2111 static int addrconf_dad_end(struct inet6_ifaddr *ifp)
2112 {
2113 int err = -ENOENT;
2114
2115 spin_lock_bh(&ifp->lock);
2116 if (ifp->state == INET6_IFADDR_STATE_DAD) {
2117 ifp->state = INET6_IFADDR_STATE_POSTDAD;
2118 err = 0;
2119 }
2120 spin_unlock_bh(&ifp->lock);
2121
2122 return err;
2123 }
2124
addrconf_dad_failure(struct sk_buff * skb,struct inet6_ifaddr * ifp)2125 void addrconf_dad_failure(struct sk_buff *skb, struct inet6_ifaddr *ifp)
2126 {
2127 struct inet6_dev *idev = ifp->idev;
2128 struct net *net = dev_net(idev->dev);
2129
2130 if (addrconf_dad_end(ifp)) {
2131 in6_ifa_put(ifp);
2132 return;
2133 }
2134
2135 net_info_ratelimited("%s: IPv6 duplicate address %pI6c used by %pM detected!\n",
2136 ifp->idev->dev->name, &ifp->addr, eth_hdr(skb)->h_source);
2137
2138 spin_lock_bh(&ifp->lock);
2139
2140 if (ifp->flags & IFA_F_STABLE_PRIVACY) {
2141 struct in6_addr new_addr;
2142 struct inet6_ifaddr *ifp2;
2143 int retries = ifp->stable_privacy_retry + 1;
2144 struct ifa6_config cfg = {
2145 .pfx = &new_addr,
2146 .plen = ifp->prefix_len,
2147 .ifa_flags = ifp->flags,
2148 .valid_lft = ifp->valid_lft,
2149 .preferred_lft = ifp->prefered_lft,
2150 .scope = ifp->scope,
2151 };
2152
2153 if (retries > net->ipv6.sysctl.idgen_retries) {
2154 net_info_ratelimited("%s: privacy stable address generation failed because of DAD conflicts!\n",
2155 ifp->idev->dev->name);
2156 goto errdad;
2157 }
2158
2159 new_addr = ifp->addr;
2160 if (ipv6_generate_stable_address(&new_addr, retries,
2161 idev))
2162 goto errdad;
2163
2164 spin_unlock_bh(&ifp->lock);
2165
2166 if (idev->cnf.max_addresses &&
2167 ipv6_count_addresses(idev) >=
2168 idev->cnf.max_addresses)
2169 goto lock_errdad;
2170
2171 net_info_ratelimited("%s: generating new stable privacy address because of DAD conflict\n",
2172 ifp->idev->dev->name);
2173
2174 ifp2 = ipv6_add_addr(idev, &cfg, false, NULL);
2175 if (IS_ERR(ifp2))
2176 goto lock_errdad;
2177
2178 spin_lock_bh(&ifp2->lock);
2179 ifp2->stable_privacy_retry = retries;
2180 ifp2->state = INET6_IFADDR_STATE_PREDAD;
2181 spin_unlock_bh(&ifp2->lock);
2182
2183 addrconf_mod_dad_work(ifp2, net->ipv6.sysctl.idgen_delay);
2184 in6_ifa_put(ifp2);
2185 lock_errdad:
2186 spin_lock_bh(&ifp->lock);
2187 }
2188
2189 errdad:
2190 /* transition from _POSTDAD to _ERRDAD */
2191 ifp->state = INET6_IFADDR_STATE_ERRDAD;
2192 spin_unlock_bh(&ifp->lock);
2193
2194 addrconf_mod_dad_work(ifp, 0);
2195 in6_ifa_put(ifp);
2196 }
2197
2198 /* Join to solicited addr multicast group.
2199 * caller must hold RTNL */
addrconf_join_solict(struct net_device * dev,const struct in6_addr * addr)2200 void addrconf_join_solict(struct net_device *dev, const struct in6_addr *addr)
2201 {
2202 struct in6_addr maddr;
2203
2204 if (dev->flags&(IFF_LOOPBACK|IFF_NOARP))
2205 return;
2206
2207 addrconf_addr_solict_mult(addr, &maddr);
2208 ipv6_dev_mc_inc(dev, &maddr);
2209 }
2210
2211 /* caller must hold RTNL */
addrconf_leave_solict(struct inet6_dev * idev,const struct in6_addr * addr)2212 void addrconf_leave_solict(struct inet6_dev *idev, const struct in6_addr *addr)
2213 {
2214 struct in6_addr maddr;
2215
2216 if (idev->dev->flags&(IFF_LOOPBACK|IFF_NOARP))
2217 return;
2218
2219 addrconf_addr_solict_mult(addr, &maddr);
2220 __ipv6_dev_mc_dec(idev, &maddr);
2221 }
2222
2223 /* caller must hold RTNL */
addrconf_join_anycast(struct inet6_ifaddr * ifp)2224 static void addrconf_join_anycast(struct inet6_ifaddr *ifp)
2225 {
2226 struct in6_addr addr;
2227
2228 if (ifp->prefix_len >= 127) /* RFC 6164 */
2229 return;
2230 ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
2231 if (ipv6_addr_any(&addr))
2232 return;
2233 __ipv6_dev_ac_inc(ifp->idev, &addr);
2234 }
2235
2236 /* caller must hold RTNL */
addrconf_leave_anycast(struct inet6_ifaddr * ifp)2237 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp)
2238 {
2239 struct in6_addr addr;
2240
2241 if (ifp->prefix_len >= 127) /* RFC 6164 */
2242 return;
2243 ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
2244 if (ipv6_addr_any(&addr))
2245 return;
2246 __ipv6_dev_ac_dec(ifp->idev, &addr);
2247 }
2248
addrconf_ifid_6lowpan(u8 * eui,struct net_device * dev)2249 static int addrconf_ifid_6lowpan(u8 *eui, struct net_device *dev)
2250 {
2251 switch (dev->addr_len) {
2252 case ETH_ALEN:
2253 memcpy(eui, dev->dev_addr, 3);
2254 eui[3] = 0xFF;
2255 eui[4] = 0xFE;
2256 memcpy(eui + 5, dev->dev_addr + 3, 3);
2257 break;
2258 case EUI64_ADDR_LEN:
2259 memcpy(eui, dev->dev_addr, EUI64_ADDR_LEN);
2260 eui[0] ^= 2;
2261 break;
2262 default:
2263 return -1;
2264 }
2265
2266 return 0;
2267 }
2268
addrconf_ifid_ieee1394(u8 * eui,struct net_device * dev)2269 static int addrconf_ifid_ieee1394(u8 *eui, struct net_device *dev)
2270 {
2271 const union fwnet_hwaddr *ha;
2272
2273 if (dev->addr_len != FWNET_ALEN)
2274 return -1;
2275
2276 ha = (const union fwnet_hwaddr *)dev->dev_addr;
2277
2278 memcpy(eui, &ha->uc.uniq_id, sizeof(ha->uc.uniq_id));
2279 eui[0] ^= 2;
2280 return 0;
2281 }
2282
addrconf_ifid_arcnet(u8 * eui,struct net_device * dev)2283 static int addrconf_ifid_arcnet(u8 *eui, struct net_device *dev)
2284 {
2285 /* XXX: inherit EUI-64 from other interface -- yoshfuji */
2286 if (dev->addr_len != ARCNET_ALEN)
2287 return -1;
2288 memset(eui, 0, 7);
2289 eui[7] = *(u8 *)dev->dev_addr;
2290 return 0;
2291 }
2292
addrconf_ifid_infiniband(u8 * eui,struct net_device * dev)2293 static int addrconf_ifid_infiniband(u8 *eui, struct net_device *dev)
2294 {
2295 if (dev->addr_len != INFINIBAND_ALEN)
2296 return -1;
2297 memcpy(eui, dev->dev_addr + 12, 8);
2298 eui[0] |= 2;
2299 return 0;
2300 }
2301
__ipv6_isatap_ifid(u8 * eui,__be32 addr)2302 static int __ipv6_isatap_ifid(u8 *eui, __be32 addr)
2303 {
2304 if (addr == 0)
2305 return -1;
2306 eui[0] = (ipv4_is_zeronet(addr) || ipv4_is_private_10(addr) ||
2307 ipv4_is_loopback(addr) || ipv4_is_linklocal_169(addr) ||
2308 ipv4_is_private_172(addr) || ipv4_is_test_192(addr) ||
2309 ipv4_is_anycast_6to4(addr) || ipv4_is_private_192(addr) ||
2310 ipv4_is_test_198(addr) || ipv4_is_multicast(addr) ||
2311 ipv4_is_lbcast(addr)) ? 0x00 : 0x02;
2312 eui[1] = 0;
2313 eui[2] = 0x5E;
2314 eui[3] = 0xFE;
2315 memcpy(eui + 4, &addr, 4);
2316 return 0;
2317 }
2318
addrconf_ifid_sit(u8 * eui,struct net_device * dev)2319 static int addrconf_ifid_sit(u8 *eui, struct net_device *dev)
2320 {
2321 if (dev->priv_flags & IFF_ISATAP)
2322 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
2323 return -1;
2324 }
2325
addrconf_ifid_gre(u8 * eui,struct net_device * dev)2326 static int addrconf_ifid_gre(u8 *eui, struct net_device *dev)
2327 {
2328 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
2329 }
2330
addrconf_ifid_ip6tnl(u8 * eui,struct net_device * dev)2331 static int addrconf_ifid_ip6tnl(u8 *eui, struct net_device *dev)
2332 {
2333 memcpy(eui, dev->perm_addr, 3);
2334 memcpy(eui + 5, dev->perm_addr + 3, 3);
2335 eui[3] = 0xFF;
2336 eui[4] = 0xFE;
2337 eui[0] ^= 2;
2338 return 0;
2339 }
2340
ipv6_generate_eui64(u8 * eui,struct net_device * dev)2341 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev)
2342 {
2343 switch (dev->type) {
2344 case ARPHRD_ETHER:
2345 case ARPHRD_FDDI:
2346 return addrconf_ifid_eui48(eui, dev);
2347 case ARPHRD_ARCNET:
2348 return addrconf_ifid_arcnet(eui, dev);
2349 case ARPHRD_INFINIBAND:
2350 return addrconf_ifid_infiniband(eui, dev);
2351 case ARPHRD_SIT:
2352 return addrconf_ifid_sit(eui, dev);
2353 case ARPHRD_IPGRE:
2354 case ARPHRD_TUNNEL:
2355 return addrconf_ifid_gre(eui, dev);
2356 case ARPHRD_6LOWPAN:
2357 return addrconf_ifid_6lowpan(eui, dev);
2358 case ARPHRD_IEEE1394:
2359 return addrconf_ifid_ieee1394(eui, dev);
2360 case ARPHRD_TUNNEL6:
2361 case ARPHRD_IP6GRE:
2362 case ARPHRD_RAWIP:
2363 return addrconf_ifid_ip6tnl(eui, dev);
2364 }
2365 return -1;
2366 }
2367
ipv6_inherit_eui64(u8 * eui,struct inet6_dev * idev)2368 static int ipv6_inherit_eui64(u8 *eui, struct inet6_dev *idev)
2369 {
2370 int err = -1;
2371 struct inet6_ifaddr *ifp;
2372
2373 read_lock_bh(&idev->lock);
2374 list_for_each_entry_reverse(ifp, &idev->addr_list, if_list) {
2375 if (ifp->scope > IFA_LINK)
2376 break;
2377 if (ifp->scope == IFA_LINK && !(ifp->flags&IFA_F_TENTATIVE)) {
2378 memcpy(eui, ifp->addr.s6_addr+8, 8);
2379 err = 0;
2380 break;
2381 }
2382 }
2383 read_unlock_bh(&idev->lock);
2384 return err;
2385 }
2386
2387 /* Generation of a randomized Interface Identifier
2388 * draft-ietf-6man-rfc4941bis, Section 3.3.1
2389 */
2390
ipv6_gen_rnd_iid(struct in6_addr * addr)2391 static void ipv6_gen_rnd_iid(struct in6_addr *addr)
2392 {
2393 regen:
2394 get_random_bytes(&addr->s6_addr[8], 8);
2395
2396 /* <draft-ietf-6man-rfc4941bis-08.txt>, Section 3.3.1:
2397 * check if generated address is not inappropriate:
2398 *
2399 * - Reserved IPv6 Interface Identifiers
2400 * - XXX: already assigned to an address on the device
2401 */
2402
2403 /* Subnet-router anycast: 0000:0000:0000:0000 */
2404 if (!(addr->s6_addr32[2] | addr->s6_addr32[3]))
2405 goto regen;
2406
2407 /* IANA Ethernet block: 0200:5EFF:FE00:0000-0200:5EFF:FE00:5212
2408 * Proxy Mobile IPv6: 0200:5EFF:FE00:5213
2409 * IANA Ethernet block: 0200:5EFF:FE00:5214-0200:5EFF:FEFF:FFFF
2410 */
2411 if (ntohl(addr->s6_addr32[2]) == 0x02005eff &&
2412 (ntohl(addr->s6_addr32[3]) & 0Xff000000) == 0xfe000000)
2413 goto regen;
2414
2415 /* Reserved subnet anycast addresses */
2416 if (ntohl(addr->s6_addr32[2]) == 0xfdffffff &&
2417 ntohl(addr->s6_addr32[3]) >= 0Xffffff80)
2418 goto regen;
2419 }
2420
addrconf_rt_table(const struct net_device * dev,u32 default_table)2421 u32 addrconf_rt_table(const struct net_device *dev, u32 default_table)
2422 {
2423 struct inet6_dev *idev = in6_dev_get(dev);
2424 int sysctl;
2425 u32 table;
2426
2427 if (!idev)
2428 return default_table;
2429 sysctl = idev->cnf.accept_ra_rt_table;
2430 if (sysctl == 0) {
2431 table = default_table;
2432 } else if (sysctl > 0) {
2433 table = (u32) sysctl;
2434 } else {
2435 table = (unsigned) dev->ifindex + (-sysctl);
2436 }
2437 in6_dev_put(idev);
2438 return table;
2439 }
2440
2441 /*
2442 * Add prefix route.
2443 */
2444
2445 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)2446 addrconf_prefix_route(struct in6_addr *pfx, int plen, u32 metric,
2447 struct net_device *dev, unsigned long expires,
2448 u32 flags, gfp_t gfp_flags)
2449 {
2450 struct fib6_config cfg = {
2451 .fc_table = l3mdev_fib_table(dev) ? : addrconf_rt_table(dev, RT6_TABLE_PREFIX),
2452 .fc_metric = metric ? : IP6_RT_PRIO_ADDRCONF,
2453 .fc_ifindex = dev->ifindex,
2454 .fc_expires = expires,
2455 .fc_dst_len = plen,
2456 .fc_flags = RTF_UP | flags,
2457 .fc_nlinfo.nl_net = dev_net(dev),
2458 .fc_protocol = RTPROT_KERNEL,
2459 .fc_type = RTN_UNICAST,
2460 };
2461
2462 cfg.fc_dst = *pfx;
2463
2464 /* Prevent useless cloning on PtP SIT.
2465 This thing is done here expecting that the whole
2466 class of non-broadcast devices need not cloning.
2467 */
2468 #if IS_ENABLED(CONFIG_IPV6_SIT)
2469 if (dev->type == ARPHRD_SIT && (dev->flags & IFF_POINTOPOINT))
2470 cfg.fc_flags |= RTF_NONEXTHOP;
2471 #endif
2472
2473 ip6_route_add(&cfg, gfp_flags, NULL);
2474 }
2475
2476
addrconf_get_prefix_route(const struct in6_addr * pfx,int plen,const struct net_device * dev,u32 flags,u32 noflags,bool no_gw)2477 static struct fib6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
2478 int plen,
2479 const struct net_device *dev,
2480 u32 flags, u32 noflags,
2481 bool no_gw)
2482 {
2483 struct fib6_node *fn;
2484 struct fib6_info *rt = NULL;
2485 struct fib6_table *table;
2486 u32 tb_id = l3mdev_fib_table(dev) ? : addrconf_rt_table(dev, RT6_TABLE_PREFIX);
2487
2488 table = fib6_get_table(dev_net(dev), tb_id);
2489 if (!table)
2490 return NULL;
2491
2492 rcu_read_lock();
2493 fn = fib6_locate(&table->tb6_root, pfx, plen, NULL, 0, true);
2494 if (!fn)
2495 goto out;
2496
2497 for_each_fib6_node_rt_rcu(fn) {
2498 /* prefix routes only use builtin fib6_nh */
2499 if (rt->nh)
2500 continue;
2501
2502 if (rt->fib6_nh->fib_nh_dev->ifindex != dev->ifindex)
2503 continue;
2504 if (no_gw && rt->fib6_nh->fib_nh_gw_family)
2505 continue;
2506 if ((rt->fib6_flags & flags) != flags)
2507 continue;
2508 if ((rt->fib6_flags & noflags) != 0)
2509 continue;
2510 if (!fib6_info_hold_safe(rt))
2511 continue;
2512 break;
2513 }
2514 out:
2515 rcu_read_unlock();
2516 return rt;
2517 }
2518
2519
2520 /* Create "default" multicast route to the interface */
2521
addrconf_add_mroute(struct net_device * dev)2522 static void addrconf_add_mroute(struct net_device *dev)
2523 {
2524 struct fib6_config cfg = {
2525 .fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_LOCAL,
2526 .fc_metric = IP6_RT_PRIO_ADDRCONF,
2527 .fc_ifindex = dev->ifindex,
2528 .fc_dst_len = 8,
2529 .fc_flags = RTF_UP,
2530 .fc_type = RTN_MULTICAST,
2531 .fc_nlinfo.nl_net = dev_net(dev),
2532 .fc_protocol = RTPROT_KERNEL,
2533 };
2534
2535 ipv6_addr_set(&cfg.fc_dst, htonl(0xFF000000), 0, 0, 0);
2536
2537 ip6_route_add(&cfg, GFP_KERNEL, NULL);
2538 }
2539
addrconf_add_dev(struct net_device * dev)2540 static struct inet6_dev *addrconf_add_dev(struct net_device *dev)
2541 {
2542 struct inet6_dev *idev;
2543
2544 ASSERT_RTNL();
2545
2546 idev = ipv6_find_idev(dev);
2547 if (IS_ERR(idev))
2548 return idev;
2549
2550 if (idev->cnf.disable_ipv6)
2551 return ERR_PTR(-EACCES);
2552
2553 /* Add default multicast route */
2554 if (!(dev->flags & IFF_LOOPBACK) && !netif_is_l3_master(dev))
2555 addrconf_add_mroute(dev);
2556
2557 return idev;
2558 }
2559
manage_tempaddrs(struct inet6_dev * idev,struct inet6_ifaddr * ifp,__u32 valid_lft,__u32 prefered_lft,bool create,unsigned long now)2560 static void manage_tempaddrs(struct inet6_dev *idev,
2561 struct inet6_ifaddr *ifp,
2562 __u32 valid_lft, __u32 prefered_lft,
2563 bool create, unsigned long now)
2564 {
2565 u32 flags;
2566 struct inet6_ifaddr *ift;
2567
2568 read_lock_bh(&idev->lock);
2569 /* update all temporary addresses in the list */
2570 list_for_each_entry(ift, &idev->tempaddr_list, tmp_list) {
2571 int age, max_valid, max_prefered;
2572
2573 if (ifp != ift->ifpub)
2574 continue;
2575
2576 /* RFC 4941 section 3.3:
2577 * If a received option will extend the lifetime of a public
2578 * address, the lifetimes of temporary addresses should
2579 * be extended, subject to the overall constraint that no
2580 * temporary addresses should ever remain "valid" or "preferred"
2581 * for a time longer than (TEMP_VALID_LIFETIME) or
2582 * (TEMP_PREFERRED_LIFETIME - DESYNC_FACTOR), respectively.
2583 */
2584 age = (now - ift->cstamp) / HZ;
2585 max_valid = idev->cnf.temp_valid_lft - age;
2586 if (max_valid < 0)
2587 max_valid = 0;
2588
2589 max_prefered = idev->cnf.temp_prefered_lft -
2590 idev->desync_factor - age;
2591 if (max_prefered < 0)
2592 max_prefered = 0;
2593
2594 if (valid_lft > max_valid)
2595 valid_lft = max_valid;
2596
2597 if (prefered_lft > max_prefered)
2598 prefered_lft = max_prefered;
2599
2600 spin_lock(&ift->lock);
2601 flags = ift->flags;
2602 ift->valid_lft = valid_lft;
2603 ift->prefered_lft = prefered_lft;
2604 ift->tstamp = now;
2605 if (prefered_lft > 0)
2606 ift->flags &= ~IFA_F_DEPRECATED;
2607
2608 spin_unlock(&ift->lock);
2609 if (!(flags&IFA_F_TENTATIVE))
2610 ipv6_ifa_notify(0, ift);
2611 }
2612
2613 /* Also create a temporary address if it's enabled but no temporary
2614 * address currently exists.
2615 * However, we get called with valid_lft == 0, prefered_lft == 0, create == false
2616 * as part of cleanup (ie. deleting the mngtmpaddr).
2617 * We don't want that to result in creating a new temporary ip address.
2618 */
2619 if (list_empty(&idev->tempaddr_list) && (valid_lft || prefered_lft))
2620 create = true;
2621
2622 if (create && idev->cnf.use_tempaddr > 0) {
2623 /* When a new public address is created as described
2624 * in [ADDRCONF], also create a new temporary address.
2625 */
2626 read_unlock_bh(&idev->lock);
2627 ipv6_create_tempaddr(ifp, false);
2628 } else {
2629 read_unlock_bh(&idev->lock);
2630 }
2631 }
2632
is_addr_mode_generate_stable(struct inet6_dev * idev)2633 static bool is_addr_mode_generate_stable(struct inet6_dev *idev)
2634 {
2635 return idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_STABLE_PRIVACY ||
2636 idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_RANDOM;
2637 }
2638
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)2639 int addrconf_prefix_rcv_add_addr(struct net *net, struct net_device *dev,
2640 const struct prefix_info *pinfo,
2641 struct inet6_dev *in6_dev,
2642 const struct in6_addr *addr, int addr_type,
2643 u32 addr_flags, bool sllao, bool tokenized,
2644 __u32 valid_lft, u32 prefered_lft)
2645 {
2646 struct inet6_ifaddr *ifp = ipv6_get_ifaddr(net, addr, dev, 1);
2647 int create = 0, update_lft = 0;
2648
2649 if (!ifp && valid_lft) {
2650 int max_addresses = in6_dev->cnf.max_addresses;
2651 struct ifa6_config cfg = {
2652 .pfx = addr,
2653 .plen = pinfo->prefix_len,
2654 .ifa_flags = addr_flags,
2655 .valid_lft = valid_lft,
2656 .preferred_lft = prefered_lft,
2657 .scope = addr_type & IPV6_ADDR_SCOPE_MASK,
2658 .ifa_proto = IFAPROT_KERNEL_RA
2659 };
2660
2661 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2662 if ((net->ipv6.devconf_all->optimistic_dad ||
2663 in6_dev->cnf.optimistic_dad) &&
2664 !net->ipv6.devconf_all->forwarding && sllao)
2665 cfg.ifa_flags |= IFA_F_OPTIMISTIC;
2666 #endif
2667
2668 /* Do not allow to create too much of autoconfigured
2669 * addresses; this would be too easy way to crash kernel.
2670 */
2671 if (!max_addresses ||
2672 ipv6_count_addresses(in6_dev) < max_addresses)
2673 ifp = ipv6_add_addr(in6_dev, &cfg, false, NULL);
2674
2675 if (IS_ERR_OR_NULL(ifp))
2676 return -1;
2677
2678 create = 1;
2679 spin_lock_bh(&ifp->lock);
2680 ifp->flags |= IFA_F_MANAGETEMPADDR;
2681 ifp->cstamp = jiffies;
2682 ifp->tokenized = tokenized;
2683 spin_unlock_bh(&ifp->lock);
2684 addrconf_dad_start(ifp);
2685 }
2686
2687 if (ifp) {
2688 u32 flags;
2689 unsigned long now;
2690 u32 stored_lft;
2691
2692 /* update lifetime (RFC2462 5.5.3 e) */
2693 spin_lock_bh(&ifp->lock);
2694 now = jiffies;
2695 if (ifp->valid_lft > (now - ifp->tstamp) / HZ)
2696 stored_lft = ifp->valid_lft - (now - ifp->tstamp) / HZ;
2697 else
2698 stored_lft = 0;
2699 if (!create && stored_lft) {
2700 const u32 minimum_lft = min_t(u32,
2701 stored_lft, MIN_VALID_LIFETIME);
2702 valid_lft = max(valid_lft, minimum_lft);
2703
2704 /* RFC4862 Section 5.5.3e:
2705 * "Note that the preferred lifetime of the
2706 * corresponding address is always reset to
2707 * the Preferred Lifetime in the received
2708 * Prefix Information option, regardless of
2709 * whether the valid lifetime is also reset or
2710 * ignored."
2711 *
2712 * So we should always update prefered_lft here.
2713 */
2714 update_lft = 1;
2715 }
2716
2717 if (update_lft) {
2718 ifp->valid_lft = valid_lft;
2719 ifp->prefered_lft = prefered_lft;
2720 ifp->tstamp = now;
2721 flags = ifp->flags;
2722 ifp->flags &= ~IFA_F_DEPRECATED;
2723 spin_unlock_bh(&ifp->lock);
2724
2725 if (!(flags&IFA_F_TENTATIVE))
2726 ipv6_ifa_notify(0, ifp);
2727 } else
2728 spin_unlock_bh(&ifp->lock);
2729
2730 manage_tempaddrs(in6_dev, ifp, valid_lft, prefered_lft,
2731 create, now);
2732
2733 in6_ifa_put(ifp);
2734 addrconf_verify(net);
2735 }
2736
2737 return 0;
2738 }
2739 EXPORT_SYMBOL_GPL(addrconf_prefix_rcv_add_addr);
2740
addrconf_prefix_rcv(struct net_device * dev,u8 * opt,int len,bool sllao)2741 void addrconf_prefix_rcv(struct net_device *dev, u8 *opt, int len, bool sllao)
2742 {
2743 struct prefix_info *pinfo;
2744 __u32 valid_lft;
2745 __u32 prefered_lft;
2746 int addr_type, err;
2747 u32 addr_flags = 0;
2748 struct inet6_dev *in6_dev;
2749 struct net *net = dev_net(dev);
2750
2751 pinfo = (struct prefix_info *) opt;
2752
2753 if (len < sizeof(struct prefix_info)) {
2754 netdev_dbg(dev, "addrconf: prefix option too short\n");
2755 return;
2756 }
2757
2758 /*
2759 * Validation checks ([ADDRCONF], page 19)
2760 */
2761
2762 addr_type = ipv6_addr_type(&pinfo->prefix);
2763
2764 if (addr_type & (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL))
2765 return;
2766
2767 valid_lft = ntohl(pinfo->valid);
2768 prefered_lft = ntohl(pinfo->prefered);
2769
2770 if (prefered_lft > valid_lft) {
2771 net_warn_ratelimited("addrconf: prefix option has invalid lifetime\n");
2772 return;
2773 }
2774
2775 in6_dev = in6_dev_get(dev);
2776
2777 if (!in6_dev) {
2778 net_dbg_ratelimited("addrconf: device %s not configured\n",
2779 dev->name);
2780 return;
2781 }
2782
2783 if (valid_lft != 0 && valid_lft < in6_dev->cnf.accept_ra_min_lft)
2784 goto put;
2785
2786 /*
2787 * Two things going on here:
2788 * 1) Add routes for on-link prefixes
2789 * 2) Configure prefixes with the auto flag set
2790 */
2791
2792 if (pinfo->onlink) {
2793 struct fib6_info *rt;
2794 unsigned long rt_expires;
2795
2796 /* Avoid arithmetic overflow. Really, we could
2797 * save rt_expires in seconds, likely valid_lft,
2798 * but it would require division in fib gc, that it
2799 * not good.
2800 */
2801 if (HZ > USER_HZ)
2802 rt_expires = addrconf_timeout_fixup(valid_lft, HZ);
2803 else
2804 rt_expires = addrconf_timeout_fixup(valid_lft, USER_HZ);
2805
2806 if (addrconf_finite_timeout(rt_expires))
2807 rt_expires *= HZ;
2808
2809 rt = addrconf_get_prefix_route(&pinfo->prefix,
2810 pinfo->prefix_len,
2811 dev,
2812 RTF_ADDRCONF | RTF_PREFIX_RT,
2813 RTF_DEFAULT, true);
2814
2815 if (rt) {
2816 /* Autoconf prefix route */
2817 if (valid_lft == 0) {
2818 ip6_del_rt(net, rt, false);
2819 rt = NULL;
2820 } else if (addrconf_finite_timeout(rt_expires)) {
2821 /* not infinity */
2822 fib6_set_expires(rt, jiffies + rt_expires);
2823 } else {
2824 fib6_clean_expires(rt);
2825 }
2826 } else if (valid_lft) {
2827 clock_t expires = 0;
2828 int flags = RTF_ADDRCONF | RTF_PREFIX_RT;
2829 if (addrconf_finite_timeout(rt_expires)) {
2830 /* not infinity */
2831 flags |= RTF_EXPIRES;
2832 expires = jiffies_to_clock_t(rt_expires);
2833 }
2834 addrconf_prefix_route(&pinfo->prefix, pinfo->prefix_len,
2835 0, dev, expires, flags,
2836 GFP_ATOMIC);
2837 }
2838 fib6_info_release(rt);
2839 }
2840
2841 /* Try to figure out our local address for this prefix */
2842
2843 if (pinfo->autoconf && in6_dev->cnf.autoconf) {
2844 struct in6_addr addr;
2845 bool tokenized = false, dev_addr_generated = false;
2846
2847 if (pinfo->prefix_len == 64) {
2848 memcpy(&addr, &pinfo->prefix, 8);
2849
2850 if (!ipv6_addr_any(&in6_dev->token)) {
2851 read_lock_bh(&in6_dev->lock);
2852 memcpy(addr.s6_addr + 8,
2853 in6_dev->token.s6_addr + 8, 8);
2854 read_unlock_bh(&in6_dev->lock);
2855 tokenized = true;
2856 } else if (is_addr_mode_generate_stable(in6_dev) &&
2857 !ipv6_generate_stable_address(&addr, 0,
2858 in6_dev)) {
2859 addr_flags |= IFA_F_STABLE_PRIVACY;
2860 goto ok;
2861 } else if (ipv6_generate_eui64(addr.s6_addr + 8, dev) &&
2862 ipv6_inherit_eui64(addr.s6_addr + 8, in6_dev)) {
2863 goto put;
2864 } else {
2865 dev_addr_generated = true;
2866 }
2867 goto ok;
2868 }
2869 net_dbg_ratelimited("IPv6 addrconf: prefix with wrong length %d\n",
2870 pinfo->prefix_len);
2871 goto put;
2872
2873 ok:
2874 err = addrconf_prefix_rcv_add_addr(net, dev, pinfo, in6_dev,
2875 &addr, addr_type,
2876 addr_flags, sllao,
2877 tokenized, valid_lft,
2878 prefered_lft);
2879 if (err)
2880 goto put;
2881
2882 /* Ignore error case here because previous prefix add addr was
2883 * successful which will be notified.
2884 */
2885 ndisc_ops_prefix_rcv_add_addr(net, dev, pinfo, in6_dev, &addr,
2886 addr_type, addr_flags, sllao,
2887 tokenized, valid_lft,
2888 prefered_lft,
2889 dev_addr_generated);
2890 }
2891 inet6_prefix_notify(RTM_NEWPREFIX, in6_dev, pinfo);
2892 put:
2893 in6_dev_put(in6_dev);
2894 }
2895
addrconf_set_sit_dstaddr(struct net * net,struct net_device * dev,struct in6_ifreq * ireq)2896 static int addrconf_set_sit_dstaddr(struct net *net, struct net_device *dev,
2897 struct in6_ifreq *ireq)
2898 {
2899 struct ip_tunnel_parm p = { };
2900 int err;
2901
2902 if (!(ipv6_addr_type(&ireq->ifr6_addr) & IPV6_ADDR_COMPATv4))
2903 return -EADDRNOTAVAIL;
2904
2905 p.iph.daddr = ireq->ifr6_addr.s6_addr32[3];
2906 p.iph.version = 4;
2907 p.iph.ihl = 5;
2908 p.iph.protocol = IPPROTO_IPV6;
2909 p.iph.ttl = 64;
2910
2911 if (!dev->netdev_ops->ndo_tunnel_ctl)
2912 return -EOPNOTSUPP;
2913 err = dev->netdev_ops->ndo_tunnel_ctl(dev, &p, SIOCADDTUNNEL);
2914 if (err)
2915 return err;
2916
2917 dev = __dev_get_by_name(net, p.name);
2918 if (!dev)
2919 return -ENOBUFS;
2920 return dev_open(dev, NULL);
2921 }
2922
2923 /*
2924 * Set destination address.
2925 * Special case for SIT interfaces where we create a new "virtual"
2926 * device.
2927 */
addrconf_set_dstaddr(struct net * net,void __user * arg)2928 int addrconf_set_dstaddr(struct net *net, void __user *arg)
2929 {
2930 struct net_device *dev;
2931 struct in6_ifreq ireq;
2932 int err = -ENODEV;
2933
2934 if (!IS_ENABLED(CONFIG_IPV6_SIT))
2935 return -ENODEV;
2936 if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2937 return -EFAULT;
2938
2939 rtnl_lock();
2940 dev = __dev_get_by_index(net, ireq.ifr6_ifindex);
2941 if (dev && dev->type == ARPHRD_SIT)
2942 err = addrconf_set_sit_dstaddr(net, dev, &ireq);
2943 rtnl_unlock();
2944 return err;
2945 }
2946
ipv6_mc_config(struct sock * sk,bool join,const struct in6_addr * addr,int ifindex)2947 static int ipv6_mc_config(struct sock *sk, bool join,
2948 const struct in6_addr *addr, int ifindex)
2949 {
2950 int ret;
2951
2952 ASSERT_RTNL();
2953
2954 lock_sock(sk);
2955 if (join)
2956 ret = ipv6_sock_mc_join(sk, ifindex, addr);
2957 else
2958 ret = ipv6_sock_mc_drop(sk, ifindex, addr);
2959 release_sock(sk);
2960
2961 return ret;
2962 }
2963
2964 /*
2965 * Manual configuration of address on an interface
2966 */
inet6_addr_add(struct net * net,int ifindex,struct ifa6_config * cfg,struct netlink_ext_ack * extack)2967 static int inet6_addr_add(struct net *net, int ifindex,
2968 struct ifa6_config *cfg,
2969 struct netlink_ext_ack *extack)
2970 {
2971 struct inet6_ifaddr *ifp;
2972 struct inet6_dev *idev;
2973 struct net_device *dev;
2974 unsigned long timeout;
2975 clock_t expires;
2976 u32 flags;
2977
2978 ASSERT_RTNL();
2979
2980 if (cfg->plen > 128) {
2981 NL_SET_ERR_MSG_MOD(extack, "Invalid prefix length");
2982 return -EINVAL;
2983 }
2984
2985 /* check the lifetime */
2986 if (!cfg->valid_lft || cfg->preferred_lft > cfg->valid_lft) {
2987 NL_SET_ERR_MSG_MOD(extack, "address lifetime invalid");
2988 return -EINVAL;
2989 }
2990
2991 if (cfg->ifa_flags & IFA_F_MANAGETEMPADDR && cfg->plen != 64) {
2992 NL_SET_ERR_MSG_MOD(extack, "address with \"mngtmpaddr\" flag must have a prefix length of 64");
2993 return -EINVAL;
2994 }
2995
2996 dev = __dev_get_by_index(net, ifindex);
2997 if (!dev)
2998 return -ENODEV;
2999
3000 idev = addrconf_add_dev(dev);
3001 if (IS_ERR(idev)) {
3002 NL_SET_ERR_MSG_MOD(extack, "IPv6 is disabled on this device");
3003 return PTR_ERR(idev);
3004 }
3005
3006 if (cfg->ifa_flags & IFA_F_MCAUTOJOIN) {
3007 int ret = ipv6_mc_config(net->ipv6.mc_autojoin_sk,
3008 true, cfg->pfx, ifindex);
3009
3010 if (ret < 0) {
3011 NL_SET_ERR_MSG_MOD(extack, "Multicast auto join failed");
3012 return ret;
3013 }
3014 }
3015
3016 cfg->scope = ipv6_addr_scope(cfg->pfx);
3017
3018 timeout = addrconf_timeout_fixup(cfg->valid_lft, HZ);
3019 if (addrconf_finite_timeout(timeout)) {
3020 expires = jiffies_to_clock_t(timeout * HZ);
3021 cfg->valid_lft = timeout;
3022 flags = RTF_EXPIRES;
3023 } else {
3024 expires = 0;
3025 flags = 0;
3026 cfg->ifa_flags |= IFA_F_PERMANENT;
3027 }
3028
3029 timeout = addrconf_timeout_fixup(cfg->preferred_lft, HZ);
3030 if (addrconf_finite_timeout(timeout)) {
3031 if (timeout == 0)
3032 cfg->ifa_flags |= IFA_F_DEPRECATED;
3033 cfg->preferred_lft = timeout;
3034 }
3035
3036 ifp = ipv6_add_addr(idev, cfg, true, extack);
3037 if (!IS_ERR(ifp)) {
3038 if (!(cfg->ifa_flags & IFA_F_NOPREFIXROUTE)) {
3039 addrconf_prefix_route(&ifp->addr, ifp->prefix_len,
3040 ifp->rt_priority, dev, expires,
3041 flags, GFP_KERNEL);
3042 }
3043
3044 /* Send a netlink notification if DAD is enabled and
3045 * optimistic flag is not set
3046 */
3047 if (!(ifp->flags & (IFA_F_OPTIMISTIC | IFA_F_NODAD)))
3048 ipv6_ifa_notify(0, ifp);
3049 /*
3050 * Note that section 3.1 of RFC 4429 indicates
3051 * that the Optimistic flag should not be set for
3052 * manually configured addresses
3053 */
3054 addrconf_dad_start(ifp);
3055 if (cfg->ifa_flags & IFA_F_MANAGETEMPADDR)
3056 manage_tempaddrs(idev, ifp, cfg->valid_lft,
3057 cfg->preferred_lft, true, jiffies);
3058 in6_ifa_put(ifp);
3059 addrconf_verify_rtnl(net);
3060 return 0;
3061 } else if (cfg->ifa_flags & IFA_F_MCAUTOJOIN) {
3062 ipv6_mc_config(net->ipv6.mc_autojoin_sk, false,
3063 cfg->pfx, ifindex);
3064 }
3065
3066 return PTR_ERR(ifp);
3067 }
3068
inet6_addr_del(struct net * net,int ifindex,u32 ifa_flags,const struct in6_addr * pfx,unsigned int plen,struct netlink_ext_ack * extack)3069 static int inet6_addr_del(struct net *net, int ifindex, u32 ifa_flags,
3070 const struct in6_addr *pfx, unsigned int plen,
3071 struct netlink_ext_ack *extack)
3072 {
3073 struct inet6_ifaddr *ifp;
3074 struct inet6_dev *idev;
3075 struct net_device *dev;
3076
3077 if (plen > 128) {
3078 NL_SET_ERR_MSG_MOD(extack, "Invalid prefix length");
3079 return -EINVAL;
3080 }
3081
3082 dev = __dev_get_by_index(net, ifindex);
3083 if (!dev) {
3084 NL_SET_ERR_MSG_MOD(extack, "Unable to find the interface");
3085 return -ENODEV;
3086 }
3087
3088 idev = __in6_dev_get(dev);
3089 if (!idev) {
3090 NL_SET_ERR_MSG_MOD(extack, "IPv6 is disabled on this device");
3091 return -ENXIO;
3092 }
3093
3094 read_lock_bh(&idev->lock);
3095 list_for_each_entry(ifp, &idev->addr_list, if_list) {
3096 if (ifp->prefix_len == plen &&
3097 ipv6_addr_equal(pfx, &ifp->addr)) {
3098 in6_ifa_hold(ifp);
3099 read_unlock_bh(&idev->lock);
3100
3101 if (!(ifp->flags & IFA_F_TEMPORARY) &&
3102 (ifa_flags & IFA_F_MANAGETEMPADDR))
3103 manage_tempaddrs(idev, ifp, 0, 0, false,
3104 jiffies);
3105 ipv6_del_addr(ifp);
3106 addrconf_verify_rtnl(net);
3107 if (ipv6_addr_is_multicast(pfx)) {
3108 ipv6_mc_config(net->ipv6.mc_autojoin_sk,
3109 false, pfx, dev->ifindex);
3110 }
3111 return 0;
3112 }
3113 }
3114 read_unlock_bh(&idev->lock);
3115
3116 NL_SET_ERR_MSG_MOD(extack, "address not found");
3117 return -EADDRNOTAVAIL;
3118 }
3119
3120
addrconf_add_ifaddr(struct net * net,void __user * arg)3121 int addrconf_add_ifaddr(struct net *net, void __user *arg)
3122 {
3123 struct ifa6_config cfg = {
3124 .ifa_flags = IFA_F_PERMANENT,
3125 .preferred_lft = INFINITY_LIFE_TIME,
3126 .valid_lft = INFINITY_LIFE_TIME,
3127 };
3128 struct in6_ifreq ireq;
3129 int err;
3130
3131 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
3132 return -EPERM;
3133
3134 if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
3135 return -EFAULT;
3136
3137 cfg.pfx = &ireq.ifr6_addr;
3138 cfg.plen = ireq.ifr6_prefixlen;
3139
3140 rtnl_lock();
3141 err = inet6_addr_add(net, ireq.ifr6_ifindex, &cfg, NULL);
3142 rtnl_unlock();
3143 return err;
3144 }
3145
addrconf_del_ifaddr(struct net * net,void __user * arg)3146 int addrconf_del_ifaddr(struct net *net, void __user *arg)
3147 {
3148 struct in6_ifreq ireq;
3149 int err;
3150
3151 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
3152 return -EPERM;
3153
3154 if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
3155 return -EFAULT;
3156
3157 rtnl_lock();
3158 err = inet6_addr_del(net, ireq.ifr6_ifindex, 0, &ireq.ifr6_addr,
3159 ireq.ifr6_prefixlen, NULL);
3160 rtnl_unlock();
3161 return err;
3162 }
3163
add_addr(struct inet6_dev * idev,const struct in6_addr * addr,int plen,int scope,u8 proto)3164 static void add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
3165 int plen, int scope, u8 proto)
3166 {
3167 struct inet6_ifaddr *ifp;
3168 struct ifa6_config cfg = {
3169 .pfx = addr,
3170 .plen = plen,
3171 .ifa_flags = IFA_F_PERMANENT,
3172 .valid_lft = INFINITY_LIFE_TIME,
3173 .preferred_lft = INFINITY_LIFE_TIME,
3174 .scope = scope,
3175 .ifa_proto = proto
3176 };
3177
3178 ifp = ipv6_add_addr(idev, &cfg, true, NULL);
3179 if (!IS_ERR(ifp)) {
3180 spin_lock_bh(&ifp->lock);
3181 ifp->flags &= ~IFA_F_TENTATIVE;
3182 spin_unlock_bh(&ifp->lock);
3183 rt_genid_bump_ipv6(dev_net(idev->dev));
3184 ipv6_ifa_notify(RTM_NEWADDR, ifp);
3185 in6_ifa_put(ifp);
3186 }
3187 }
3188
3189 #if IS_ENABLED(CONFIG_IPV6_SIT) || IS_ENABLED(CONFIG_NET_IPGRE) || IS_ENABLED(CONFIG_IPV6_GRE)
add_v4_addrs(struct inet6_dev * idev)3190 static void add_v4_addrs(struct inet6_dev *idev)
3191 {
3192 struct in6_addr addr;
3193 struct net_device *dev;
3194 struct net *net = dev_net(idev->dev);
3195 int scope, plen, offset = 0;
3196 u32 pflags = 0;
3197
3198 ASSERT_RTNL();
3199
3200 memset(&addr, 0, sizeof(struct in6_addr));
3201 /* in case of IP6GRE the dev_addr is an IPv6 and therefore we use only the last 4 bytes */
3202 if (idev->dev->addr_len == sizeof(struct in6_addr))
3203 offset = sizeof(struct in6_addr) - 4;
3204 memcpy(&addr.s6_addr32[3], idev->dev->dev_addr + offset, 4);
3205
3206 if (!(idev->dev->flags & IFF_POINTOPOINT) && idev->dev->type == ARPHRD_SIT) {
3207 scope = IPV6_ADDR_COMPATv4;
3208 plen = 96;
3209 pflags |= RTF_NONEXTHOP;
3210 } else {
3211 if (idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_NONE)
3212 return;
3213
3214 addr.s6_addr32[0] = htonl(0xfe800000);
3215 scope = IFA_LINK;
3216 plen = 64;
3217 }
3218
3219 if (addr.s6_addr32[3]) {
3220 add_addr(idev, &addr, plen, scope, IFAPROT_UNSPEC);
3221 addrconf_prefix_route(&addr, plen, 0, idev->dev, 0, pflags,
3222 GFP_KERNEL);
3223 return;
3224 }
3225
3226 for_each_netdev(net, dev) {
3227 struct in_device *in_dev = __in_dev_get_rtnl(dev);
3228 if (in_dev && (dev->flags & IFF_UP)) {
3229 struct in_ifaddr *ifa;
3230 int flag = scope;
3231
3232 in_dev_for_each_ifa_rtnl(ifa, in_dev) {
3233 addr.s6_addr32[3] = ifa->ifa_local;
3234
3235 if (ifa->ifa_scope == RT_SCOPE_LINK)
3236 continue;
3237 if (ifa->ifa_scope >= RT_SCOPE_HOST) {
3238 if (idev->dev->flags&IFF_POINTOPOINT)
3239 continue;
3240 flag |= IFA_HOST;
3241 }
3242
3243 add_addr(idev, &addr, plen, flag,
3244 IFAPROT_UNSPEC);
3245 addrconf_prefix_route(&addr, plen, 0, idev->dev,
3246 0, pflags, GFP_KERNEL);
3247 }
3248 }
3249 }
3250 }
3251 #endif
3252
init_loopback(struct net_device * dev)3253 static void init_loopback(struct net_device *dev)
3254 {
3255 struct inet6_dev *idev;
3256
3257 /* ::1 */
3258
3259 ASSERT_RTNL();
3260
3261 idev = ipv6_find_idev(dev);
3262 if (IS_ERR(idev)) {
3263 pr_debug("%s: add_dev failed\n", __func__);
3264 return;
3265 }
3266
3267 add_addr(idev, &in6addr_loopback, 128, IFA_HOST, IFAPROT_KERNEL_LO);
3268 }
3269
addrconf_add_linklocal(struct inet6_dev * idev,const struct in6_addr * addr,u32 flags)3270 void addrconf_add_linklocal(struct inet6_dev *idev,
3271 const struct in6_addr *addr, u32 flags)
3272 {
3273 struct ifa6_config cfg = {
3274 .pfx = addr,
3275 .plen = 64,
3276 .ifa_flags = flags | IFA_F_PERMANENT,
3277 .valid_lft = INFINITY_LIFE_TIME,
3278 .preferred_lft = INFINITY_LIFE_TIME,
3279 .scope = IFA_LINK,
3280 .ifa_proto = IFAPROT_KERNEL_LL
3281 };
3282 struct inet6_ifaddr *ifp;
3283
3284 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
3285 if ((dev_net(idev->dev)->ipv6.devconf_all->optimistic_dad ||
3286 idev->cnf.optimistic_dad) &&
3287 !dev_net(idev->dev)->ipv6.devconf_all->forwarding)
3288 cfg.ifa_flags |= IFA_F_OPTIMISTIC;
3289 #endif
3290
3291 ifp = ipv6_add_addr(idev, &cfg, true, NULL);
3292 if (!IS_ERR(ifp)) {
3293 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, 0, idev->dev,
3294 0, 0, GFP_ATOMIC);
3295 addrconf_dad_start(ifp);
3296 in6_ifa_put(ifp);
3297 }
3298 }
3299 EXPORT_SYMBOL_GPL(addrconf_add_linklocal);
3300
ipv6_reserved_interfaceid(struct in6_addr address)3301 static bool ipv6_reserved_interfaceid(struct in6_addr address)
3302 {
3303 if ((address.s6_addr32[2] | address.s6_addr32[3]) == 0)
3304 return true;
3305
3306 if (address.s6_addr32[2] == htonl(0x02005eff) &&
3307 ((address.s6_addr32[3] & htonl(0xfe000000)) == htonl(0xfe000000)))
3308 return true;
3309
3310 if (address.s6_addr32[2] == htonl(0xfdffffff) &&
3311 ((address.s6_addr32[3] & htonl(0xffffff80)) == htonl(0xffffff80)))
3312 return true;
3313
3314 return false;
3315 }
3316
ipv6_generate_stable_address(struct in6_addr * address,u8 dad_count,const struct inet6_dev * idev)3317 static int ipv6_generate_stable_address(struct in6_addr *address,
3318 u8 dad_count,
3319 const struct inet6_dev *idev)
3320 {
3321 static DEFINE_SPINLOCK(lock);
3322 static __u32 digest[SHA1_DIGEST_WORDS];
3323 static __u32 workspace[SHA1_WORKSPACE_WORDS];
3324
3325 static union {
3326 char __data[SHA1_BLOCK_SIZE];
3327 struct {
3328 struct in6_addr secret;
3329 __be32 prefix[2];
3330 unsigned char hwaddr[MAX_ADDR_LEN];
3331 u8 dad_count;
3332 } __packed;
3333 } data;
3334
3335 struct in6_addr secret;
3336 struct in6_addr temp;
3337 struct net *net = dev_net(idev->dev);
3338
3339 BUILD_BUG_ON(sizeof(data.__data) != sizeof(data));
3340
3341 if (idev->cnf.stable_secret.initialized)
3342 secret = idev->cnf.stable_secret.secret;
3343 else if (net->ipv6.devconf_dflt->stable_secret.initialized)
3344 secret = net->ipv6.devconf_dflt->stable_secret.secret;
3345 else
3346 return -1;
3347
3348 retry:
3349 spin_lock_bh(&lock);
3350
3351 sha1_init(digest);
3352 memset(&data, 0, sizeof(data));
3353 memset(workspace, 0, sizeof(workspace));
3354 memcpy(data.hwaddr, idev->dev->perm_addr, idev->dev->addr_len);
3355 data.prefix[0] = address->s6_addr32[0];
3356 data.prefix[1] = address->s6_addr32[1];
3357 data.secret = secret;
3358 data.dad_count = dad_count;
3359
3360 sha1_transform(digest, data.__data, workspace);
3361
3362 temp = *address;
3363 temp.s6_addr32[2] = (__force __be32)digest[0];
3364 temp.s6_addr32[3] = (__force __be32)digest[1];
3365
3366 spin_unlock_bh(&lock);
3367
3368 if (ipv6_reserved_interfaceid(temp)) {
3369 dad_count++;
3370 if (dad_count > dev_net(idev->dev)->ipv6.sysctl.idgen_retries)
3371 return -1;
3372 goto retry;
3373 }
3374
3375 *address = temp;
3376 return 0;
3377 }
3378
ipv6_gen_mode_random_init(struct inet6_dev * idev)3379 static void ipv6_gen_mode_random_init(struct inet6_dev *idev)
3380 {
3381 struct ipv6_stable_secret *s = &idev->cnf.stable_secret;
3382
3383 if (s->initialized)
3384 return;
3385 s = &idev->cnf.stable_secret;
3386 get_random_bytes(&s->secret, sizeof(s->secret));
3387 s->initialized = true;
3388 }
3389
addrconf_addr_gen(struct inet6_dev * idev,bool prefix_route)3390 static void addrconf_addr_gen(struct inet6_dev *idev, bool prefix_route)
3391 {
3392 struct in6_addr addr;
3393
3394 /* no link local addresses on L3 master devices */
3395 if (netif_is_l3_master(idev->dev))
3396 return;
3397
3398 /* no link local addresses on devices flagged as slaves */
3399 if (idev->dev->priv_flags & IFF_NO_ADDRCONF)
3400 return;
3401
3402 ipv6_addr_set(&addr, htonl(0xFE800000), 0, 0, 0);
3403
3404 switch (idev->cnf.addr_gen_mode) {
3405 case IN6_ADDR_GEN_MODE_RANDOM:
3406 ipv6_gen_mode_random_init(idev);
3407 fallthrough;
3408 case IN6_ADDR_GEN_MODE_STABLE_PRIVACY:
3409 if (!ipv6_generate_stable_address(&addr, 0, idev))
3410 addrconf_add_linklocal(idev, &addr,
3411 IFA_F_STABLE_PRIVACY);
3412 else if (prefix_route)
3413 addrconf_prefix_route(&addr, 64, 0, idev->dev,
3414 0, 0, GFP_KERNEL);
3415 break;
3416 case IN6_ADDR_GEN_MODE_EUI64:
3417 /* addrconf_add_linklocal also adds a prefix_route and we
3418 * only need to care about prefix routes if ipv6_generate_eui64
3419 * couldn't generate one.
3420 */
3421 if (ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) == 0)
3422 addrconf_add_linklocal(idev, &addr, 0);
3423 else if (prefix_route)
3424 addrconf_prefix_route(&addr, 64, 0, idev->dev,
3425 0, 0, GFP_KERNEL);
3426 break;
3427 case IN6_ADDR_GEN_MODE_NONE:
3428 default:
3429 /* will not add any link local address */
3430 break;
3431 }
3432 }
3433
addrconf_dev_config(struct net_device * dev)3434 static void addrconf_dev_config(struct net_device *dev)
3435 {
3436 struct inet6_dev *idev;
3437
3438 ASSERT_RTNL();
3439
3440 if ((dev->type != ARPHRD_ETHER) &&
3441 (dev->type != ARPHRD_FDDI) &&
3442 (dev->type != ARPHRD_ARCNET) &&
3443 (dev->type != ARPHRD_INFINIBAND) &&
3444 (dev->type != ARPHRD_IEEE1394) &&
3445 (dev->type != ARPHRD_TUNNEL6) &&
3446 (dev->type != ARPHRD_6LOWPAN) &&
3447 (dev->type != ARPHRD_TUNNEL) &&
3448 (dev->type != ARPHRD_NONE) &&
3449 (dev->type != ARPHRD_RAWIP)) {
3450 /* Alas, we support only Ethernet autoconfiguration. */
3451 idev = __in6_dev_get(dev);
3452 if (!IS_ERR_OR_NULL(idev) && dev->flags & IFF_UP &&
3453 dev->flags & IFF_MULTICAST)
3454 ipv6_mc_up(idev);
3455 return;
3456 }
3457
3458 idev = addrconf_add_dev(dev);
3459 if (IS_ERR(idev))
3460 return;
3461
3462 /* this device type has no EUI support */
3463 if (dev->type == ARPHRD_NONE &&
3464 idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_EUI64)
3465 idev->cnf.addr_gen_mode = IN6_ADDR_GEN_MODE_RANDOM;
3466
3467 addrconf_addr_gen(idev, false);
3468 }
3469
3470 #if IS_ENABLED(CONFIG_IPV6_SIT)
addrconf_sit_config(struct net_device * dev)3471 static void addrconf_sit_config(struct net_device *dev)
3472 {
3473 struct inet6_dev *idev;
3474
3475 ASSERT_RTNL();
3476
3477 /*
3478 * Configure the tunnel with one of our IPv4
3479 * addresses... we should configure all of
3480 * our v4 addrs in the tunnel
3481 */
3482
3483 idev = ipv6_find_idev(dev);
3484 if (IS_ERR(idev)) {
3485 pr_debug("%s: add_dev failed\n", __func__);
3486 return;
3487 }
3488
3489 if (dev->priv_flags & IFF_ISATAP) {
3490 addrconf_addr_gen(idev, false);
3491 return;
3492 }
3493
3494 add_v4_addrs(idev);
3495
3496 if (dev->flags&IFF_POINTOPOINT)
3497 addrconf_add_mroute(dev);
3498 }
3499 #endif
3500
3501 #if IS_ENABLED(CONFIG_NET_IPGRE) || IS_ENABLED(CONFIG_IPV6_GRE)
addrconf_gre_config(struct net_device * dev)3502 static void addrconf_gre_config(struct net_device *dev)
3503 {
3504 struct inet6_dev *idev;
3505
3506 ASSERT_RTNL();
3507
3508 idev = ipv6_find_idev(dev);
3509 if (IS_ERR(idev)) {
3510 pr_debug("%s: add_dev failed\n", __func__);
3511 return;
3512 }
3513
3514 if (dev->type == ARPHRD_ETHER) {
3515 addrconf_addr_gen(idev, true);
3516 return;
3517 }
3518
3519 add_v4_addrs(idev);
3520
3521 if (dev->flags & IFF_POINTOPOINT)
3522 addrconf_add_mroute(dev);
3523 }
3524 #endif
3525
addrconf_init_auto_addrs(struct net_device * dev)3526 static void addrconf_init_auto_addrs(struct net_device *dev)
3527 {
3528 switch (dev->type) {
3529 #if IS_ENABLED(CONFIG_IPV6_SIT)
3530 case ARPHRD_SIT:
3531 addrconf_sit_config(dev);
3532 break;
3533 #endif
3534 #if IS_ENABLED(CONFIG_NET_IPGRE) || IS_ENABLED(CONFIG_IPV6_GRE)
3535 case ARPHRD_IP6GRE:
3536 case ARPHRD_IPGRE:
3537 addrconf_gre_config(dev);
3538 break;
3539 #endif
3540 case ARPHRD_LOOPBACK:
3541 init_loopback(dev);
3542 break;
3543
3544 default:
3545 addrconf_dev_config(dev);
3546 break;
3547 }
3548 }
3549
fixup_permanent_addr(struct net * net,struct inet6_dev * idev,struct inet6_ifaddr * ifp)3550 static int fixup_permanent_addr(struct net *net,
3551 struct inet6_dev *idev,
3552 struct inet6_ifaddr *ifp)
3553 {
3554 /* !fib6_node means the host route was removed from the
3555 * FIB, for example, if 'lo' device is taken down. In that
3556 * case regenerate the host route.
3557 */
3558 if (!ifp->rt || !ifp->rt->fib6_node) {
3559 struct fib6_info *f6i, *prev;
3560
3561 f6i = addrconf_f6i_alloc(net, idev, &ifp->addr, false,
3562 GFP_ATOMIC, NULL);
3563 if (IS_ERR(f6i))
3564 return PTR_ERR(f6i);
3565
3566 /* ifp->rt can be accessed outside of rtnl */
3567 spin_lock(&ifp->lock);
3568 prev = ifp->rt;
3569 ifp->rt = f6i;
3570 spin_unlock(&ifp->lock);
3571
3572 fib6_info_release(prev);
3573 }
3574
3575 if (!(ifp->flags & IFA_F_NOPREFIXROUTE)) {
3576 addrconf_prefix_route(&ifp->addr, ifp->prefix_len,
3577 ifp->rt_priority, idev->dev, 0, 0,
3578 GFP_ATOMIC);
3579 }
3580
3581 if (ifp->state == INET6_IFADDR_STATE_PREDAD)
3582 addrconf_dad_start(ifp);
3583
3584 return 0;
3585 }
3586
addrconf_permanent_addr(struct net * net,struct net_device * dev)3587 static void addrconf_permanent_addr(struct net *net, struct net_device *dev)
3588 {
3589 struct inet6_ifaddr *ifp, *tmp;
3590 struct inet6_dev *idev;
3591
3592 idev = __in6_dev_get(dev);
3593 if (!idev)
3594 return;
3595
3596 write_lock_bh(&idev->lock);
3597
3598 list_for_each_entry_safe(ifp, tmp, &idev->addr_list, if_list) {
3599 if ((ifp->flags & IFA_F_PERMANENT) &&
3600 fixup_permanent_addr(net, idev, ifp) < 0) {
3601 write_unlock_bh(&idev->lock);
3602 in6_ifa_hold(ifp);
3603 ipv6_del_addr(ifp);
3604 write_lock_bh(&idev->lock);
3605
3606 net_info_ratelimited("%s: Failed to add prefix route for address %pI6c; dropping\n",
3607 idev->dev->name, &ifp->addr);
3608 }
3609 }
3610
3611 write_unlock_bh(&idev->lock);
3612 }
3613
addrconf_notify(struct notifier_block * this,unsigned long event,void * ptr)3614 static int addrconf_notify(struct notifier_block *this, unsigned long event,
3615 void *ptr)
3616 {
3617 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
3618 struct netdev_notifier_change_info *change_info;
3619 struct netdev_notifier_changeupper_info *info;
3620 struct inet6_dev *idev = __in6_dev_get(dev);
3621 struct net *net = dev_net(dev);
3622 int run_pending = 0;
3623 int err;
3624
3625 switch (event) {
3626 case NETDEV_REGISTER:
3627 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
3628 idev = ipv6_add_dev(dev);
3629 if (IS_ERR(idev))
3630 return notifier_from_errno(PTR_ERR(idev));
3631 }
3632 break;
3633
3634 case NETDEV_CHANGEMTU:
3635 /* if MTU under IPV6_MIN_MTU stop IPv6 on this interface. */
3636 if (dev->mtu < IPV6_MIN_MTU) {
3637 addrconf_ifdown(dev, dev != net->loopback_dev);
3638 break;
3639 }
3640
3641 if (idev) {
3642 rt6_mtu_change(dev, dev->mtu);
3643 idev->cnf.mtu6 = dev->mtu;
3644 break;
3645 }
3646
3647 /* allocate new idev */
3648 idev = ipv6_add_dev(dev);
3649 if (IS_ERR(idev))
3650 break;
3651
3652 /* device is still not ready */
3653 if (!(idev->if_flags & IF_READY))
3654 break;
3655
3656 run_pending = 1;
3657 fallthrough;
3658 case NETDEV_UP:
3659 case NETDEV_CHANGE:
3660 if (idev && idev->cnf.disable_ipv6)
3661 break;
3662
3663 if (dev->priv_flags & IFF_NO_ADDRCONF) {
3664 if (event == NETDEV_UP && !IS_ERR_OR_NULL(idev) &&
3665 dev->flags & IFF_UP && dev->flags & IFF_MULTICAST)
3666 ipv6_mc_up(idev);
3667 break;
3668 }
3669
3670 if (event == NETDEV_UP) {
3671 /* restore routes for permanent addresses */
3672 addrconf_permanent_addr(net, dev);
3673
3674 if (!addrconf_link_ready(dev)) {
3675 /* device is not ready yet. */
3676 pr_debug("ADDRCONF(NETDEV_UP): %s: link is not ready\n",
3677 dev->name);
3678 break;
3679 }
3680
3681 if (!idev && dev->mtu >= IPV6_MIN_MTU)
3682 idev = ipv6_add_dev(dev);
3683
3684 if (!IS_ERR_OR_NULL(idev)) {
3685 idev->if_flags |= IF_READY;
3686 run_pending = 1;
3687 }
3688 } else if (event == NETDEV_CHANGE) {
3689 if (!addrconf_link_ready(dev)) {
3690 /* device is still not ready. */
3691 rt6_sync_down_dev(dev, event);
3692 break;
3693 }
3694
3695 if (!IS_ERR_OR_NULL(idev)) {
3696 if (idev->if_flags & IF_READY) {
3697 /* device is already configured -
3698 * but resend MLD reports, we might
3699 * have roamed and need to update
3700 * multicast snooping switches
3701 */
3702 ipv6_mc_up(idev);
3703 change_info = ptr;
3704 if (change_info->flags_changed & IFF_NOARP)
3705 addrconf_dad_run(idev, true);
3706 rt6_sync_up(dev, RTNH_F_LINKDOWN);
3707 break;
3708 }
3709 idev->if_flags |= IF_READY;
3710 }
3711
3712 pr_debug("ADDRCONF(NETDEV_CHANGE): %s: link becomes ready\n",
3713 dev->name);
3714
3715 run_pending = 1;
3716 }
3717
3718 addrconf_init_auto_addrs(dev);
3719
3720 if (!IS_ERR_OR_NULL(idev)) {
3721 if (run_pending)
3722 addrconf_dad_run(idev, false);
3723
3724 /* Device has an address by now */
3725 rt6_sync_up(dev, RTNH_F_DEAD);
3726
3727 /*
3728 * If the MTU changed during the interface down,
3729 * when the interface up, the changed MTU must be
3730 * reflected in the idev as well as routers.
3731 */
3732 if (idev->cnf.mtu6 != dev->mtu &&
3733 dev->mtu >= IPV6_MIN_MTU) {
3734 rt6_mtu_change(dev, dev->mtu);
3735 idev->cnf.mtu6 = dev->mtu;
3736 }
3737 idev->tstamp = jiffies;
3738 inet6_ifinfo_notify(RTM_NEWLINK, idev);
3739
3740 /*
3741 * If the changed mtu during down is lower than
3742 * IPV6_MIN_MTU stop IPv6 on this interface.
3743 */
3744 if (dev->mtu < IPV6_MIN_MTU)
3745 addrconf_ifdown(dev, dev != net->loopback_dev);
3746 }
3747 break;
3748
3749 case NETDEV_DOWN:
3750 case NETDEV_UNREGISTER:
3751 /*
3752 * Remove all addresses from this interface.
3753 */
3754 addrconf_ifdown(dev, event != NETDEV_DOWN);
3755 break;
3756
3757 case NETDEV_CHANGENAME:
3758 if (idev) {
3759 snmp6_unregister_dev(idev);
3760 addrconf_sysctl_unregister(idev);
3761 err = addrconf_sysctl_register(idev);
3762 if (err)
3763 return notifier_from_errno(err);
3764 err = snmp6_register_dev(idev);
3765 if (err) {
3766 addrconf_sysctl_unregister(idev);
3767 return notifier_from_errno(err);
3768 }
3769 }
3770 break;
3771
3772 case NETDEV_PRE_TYPE_CHANGE:
3773 case NETDEV_POST_TYPE_CHANGE:
3774 if (idev)
3775 addrconf_type_change(dev, event);
3776 break;
3777
3778 case NETDEV_CHANGEUPPER:
3779 info = ptr;
3780
3781 /* flush all routes if dev is linked to or unlinked from
3782 * an L3 master device (e.g., VRF)
3783 */
3784 if (info->upper_dev && netif_is_l3_master(info->upper_dev))
3785 addrconf_ifdown(dev, false);
3786 }
3787
3788 return NOTIFY_OK;
3789 }
3790
3791 /*
3792 * addrconf module should be notified of a device going up
3793 */
3794 static struct notifier_block ipv6_dev_notf = {
3795 .notifier_call = addrconf_notify,
3796 .priority = ADDRCONF_NOTIFY_PRIORITY,
3797 };
3798
addrconf_type_change(struct net_device * dev,unsigned long event)3799 static void addrconf_type_change(struct net_device *dev, unsigned long event)
3800 {
3801 struct inet6_dev *idev;
3802 ASSERT_RTNL();
3803
3804 idev = __in6_dev_get(dev);
3805
3806 if (event == NETDEV_POST_TYPE_CHANGE)
3807 ipv6_mc_remap(idev);
3808 else if (event == NETDEV_PRE_TYPE_CHANGE)
3809 ipv6_mc_unmap(idev);
3810 }
3811
addr_is_local(const struct in6_addr * addr)3812 static bool addr_is_local(const struct in6_addr *addr)
3813 {
3814 return ipv6_addr_type(addr) &
3815 (IPV6_ADDR_LINKLOCAL | IPV6_ADDR_LOOPBACK);
3816 }
3817
addrconf_ifdown(struct net_device * dev,bool unregister)3818 static int addrconf_ifdown(struct net_device *dev, bool unregister)
3819 {
3820 unsigned long event = unregister ? NETDEV_UNREGISTER : NETDEV_DOWN;
3821 struct net *net = dev_net(dev);
3822 struct inet6_dev *idev;
3823 struct inet6_ifaddr *ifa;
3824 LIST_HEAD(tmp_addr_list);
3825 bool keep_addr = false;
3826 bool was_ready;
3827 int state, i;
3828
3829 ASSERT_RTNL();
3830
3831 rt6_disable_ip(dev, event);
3832
3833 idev = __in6_dev_get(dev);
3834 if (!idev)
3835 return -ENODEV;
3836
3837 /*
3838 * Step 1: remove reference to ipv6 device from parent device.
3839 * Do not dev_put!
3840 */
3841 if (unregister) {
3842 idev->dead = 1;
3843
3844 /* protected by rtnl_lock */
3845 RCU_INIT_POINTER(dev->ip6_ptr, NULL);
3846
3847 /* Step 1.5: remove snmp6 entry */
3848 snmp6_unregister_dev(idev);
3849
3850 }
3851
3852 /* combine the user config with event to determine if permanent
3853 * addresses are to be removed from address hash table
3854 */
3855 if (!unregister && !idev->cnf.disable_ipv6) {
3856 /* aggregate the system setting and interface setting */
3857 int _keep_addr = net->ipv6.devconf_all->keep_addr_on_down;
3858
3859 if (!_keep_addr)
3860 _keep_addr = idev->cnf.keep_addr_on_down;
3861
3862 keep_addr = (_keep_addr > 0);
3863 }
3864
3865 /* Step 2: clear hash table */
3866 for (i = 0; i < IN6_ADDR_HSIZE; i++) {
3867 struct hlist_head *h = &net->ipv6.inet6_addr_lst[i];
3868
3869 spin_lock_bh(&net->ipv6.addrconf_hash_lock);
3870 restart:
3871 hlist_for_each_entry_rcu(ifa, h, addr_lst) {
3872 if (ifa->idev == idev) {
3873 addrconf_del_dad_work(ifa);
3874 /* combined flag + permanent flag decide if
3875 * address is retained on a down event
3876 */
3877 if (!keep_addr ||
3878 !(ifa->flags & IFA_F_PERMANENT) ||
3879 addr_is_local(&ifa->addr)) {
3880 hlist_del_init_rcu(&ifa->addr_lst);
3881 goto restart;
3882 }
3883 }
3884 }
3885 spin_unlock_bh(&net->ipv6.addrconf_hash_lock);
3886 }
3887
3888 write_lock_bh(&idev->lock);
3889
3890 addrconf_del_rs_timer(idev);
3891
3892 /* Step 2: clear flags for stateless addrconf, repeated down
3893 * detection
3894 */
3895 was_ready = idev->if_flags & IF_READY;
3896 if (!unregister)
3897 idev->if_flags &= ~(IF_RS_SENT|IF_RA_RCVD|IF_READY);
3898
3899 /* Step 3: clear tempaddr list */
3900 while (!list_empty(&idev->tempaddr_list)) {
3901 ifa = list_first_entry(&idev->tempaddr_list,
3902 struct inet6_ifaddr, tmp_list);
3903 list_del(&ifa->tmp_list);
3904 write_unlock_bh(&idev->lock);
3905 spin_lock_bh(&ifa->lock);
3906
3907 if (ifa->ifpub) {
3908 in6_ifa_put(ifa->ifpub);
3909 ifa->ifpub = NULL;
3910 }
3911 spin_unlock_bh(&ifa->lock);
3912 in6_ifa_put(ifa);
3913 write_lock_bh(&idev->lock);
3914 }
3915
3916 list_for_each_entry(ifa, &idev->addr_list, if_list)
3917 list_add_tail(&ifa->if_list_aux, &tmp_addr_list);
3918 write_unlock_bh(&idev->lock);
3919
3920 while (!list_empty(&tmp_addr_list)) {
3921 struct fib6_info *rt = NULL;
3922 bool keep;
3923
3924 ifa = list_first_entry(&tmp_addr_list,
3925 struct inet6_ifaddr, if_list_aux);
3926 list_del(&ifa->if_list_aux);
3927
3928 addrconf_del_dad_work(ifa);
3929
3930 keep = keep_addr && (ifa->flags & IFA_F_PERMANENT) &&
3931 !addr_is_local(&ifa->addr);
3932
3933 spin_lock_bh(&ifa->lock);
3934
3935 if (keep) {
3936 /* set state to skip the notifier below */
3937 state = INET6_IFADDR_STATE_DEAD;
3938 ifa->state = INET6_IFADDR_STATE_PREDAD;
3939 if (!(ifa->flags & IFA_F_NODAD))
3940 ifa->flags |= IFA_F_TENTATIVE;
3941
3942 rt = ifa->rt;
3943 ifa->rt = NULL;
3944 } else {
3945 state = ifa->state;
3946 ifa->state = INET6_IFADDR_STATE_DEAD;
3947 }
3948
3949 spin_unlock_bh(&ifa->lock);
3950
3951 if (rt)
3952 ip6_del_rt(net, rt, false);
3953
3954 if (state != INET6_IFADDR_STATE_DEAD) {
3955 __ipv6_ifa_notify(RTM_DELADDR, ifa);
3956 inet6addr_notifier_call_chain(NETDEV_DOWN, ifa);
3957 } else {
3958 if (idev->cnf.forwarding)
3959 addrconf_leave_anycast(ifa);
3960 addrconf_leave_solict(ifa->idev, &ifa->addr);
3961 }
3962
3963 if (!keep) {
3964 write_lock_bh(&idev->lock);
3965 list_del_rcu(&ifa->if_list);
3966 write_unlock_bh(&idev->lock);
3967 in6_ifa_put(ifa);
3968 }
3969 }
3970
3971 /* Step 5: Discard anycast and multicast list */
3972 if (unregister) {
3973 ipv6_ac_destroy_dev(idev);
3974 ipv6_mc_destroy_dev(idev);
3975 } else if (was_ready) {
3976 ipv6_mc_down(idev);
3977 }
3978
3979 idev->tstamp = jiffies;
3980 idev->ra_mtu = 0;
3981
3982 /* Last: Shot the device (if unregistered) */
3983 if (unregister) {
3984 addrconf_sysctl_unregister(idev);
3985 neigh_parms_release(&nd_tbl, idev->nd_parms);
3986 neigh_ifdown(&nd_tbl, dev);
3987 in6_dev_put(idev);
3988 }
3989 return 0;
3990 }
3991
addrconf_rs_timer(struct timer_list * t)3992 static void addrconf_rs_timer(struct timer_list *t)
3993 {
3994 struct inet6_dev *idev = from_timer(idev, t, rs_timer);
3995 struct net_device *dev = idev->dev;
3996 struct in6_addr lladdr;
3997
3998 write_lock(&idev->lock);
3999 if (idev->dead || !(idev->if_flags & IF_READY))
4000 goto out;
4001
4002 if (!ipv6_accept_ra(idev))
4003 goto out;
4004
4005 /* Announcement received after solicitation was sent */
4006 if (idev->if_flags & IF_RA_RCVD)
4007 goto out;
4008
4009 if (idev->rs_probes++ < idev->cnf.rtr_solicits || idev->cnf.rtr_solicits < 0) {
4010 write_unlock(&idev->lock);
4011 if (!ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE))
4012 ndisc_send_rs(dev, &lladdr,
4013 &in6addr_linklocal_allrouters);
4014 else
4015 goto put;
4016
4017 write_lock(&idev->lock);
4018 idev->rs_interval = rfc3315_s14_backoff_update(
4019 idev->rs_interval, idev->cnf.rtr_solicit_max_interval);
4020 /* The wait after the last probe can be shorter */
4021 addrconf_mod_rs_timer(idev, (idev->rs_probes ==
4022 idev->cnf.rtr_solicits) ?
4023 idev->cnf.rtr_solicit_delay :
4024 idev->rs_interval);
4025 } else {
4026 /*
4027 * Note: we do not support deprecated "all on-link"
4028 * assumption any longer.
4029 */
4030 pr_debug("%s: no IPv6 routers present\n", idev->dev->name);
4031 }
4032
4033 out:
4034 write_unlock(&idev->lock);
4035 put:
4036 in6_dev_put(idev);
4037 }
4038
4039 /*
4040 * Duplicate Address Detection
4041 */
addrconf_dad_kick(struct inet6_ifaddr * ifp)4042 static void addrconf_dad_kick(struct inet6_ifaddr *ifp)
4043 {
4044 unsigned long rand_num;
4045 struct inet6_dev *idev = ifp->idev;
4046 u64 nonce;
4047
4048 if (ifp->flags & IFA_F_OPTIMISTIC)
4049 rand_num = 0;
4050 else
4051 rand_num = get_random_u32_below(idev->cnf.rtr_solicit_delay ? : 1);
4052
4053 nonce = 0;
4054 if (idev->cnf.enhanced_dad ||
4055 dev_net(idev->dev)->ipv6.devconf_all->enhanced_dad) {
4056 do
4057 get_random_bytes(&nonce, 6);
4058 while (nonce == 0);
4059 }
4060 ifp->dad_nonce = nonce;
4061 ifp->dad_probes = idev->cnf.dad_transmits;
4062 addrconf_mod_dad_work(ifp, rand_num);
4063 }
4064
addrconf_dad_begin(struct inet6_ifaddr * ifp)4065 static void addrconf_dad_begin(struct inet6_ifaddr *ifp)
4066 {
4067 struct inet6_dev *idev = ifp->idev;
4068 struct net_device *dev = idev->dev;
4069 bool bump_id, notify = false;
4070 struct net *net;
4071
4072 addrconf_join_solict(dev, &ifp->addr);
4073
4074 read_lock_bh(&idev->lock);
4075 spin_lock(&ifp->lock);
4076 if (ifp->state == INET6_IFADDR_STATE_DEAD)
4077 goto out;
4078
4079 net = dev_net(dev);
4080 if (dev->flags&(IFF_NOARP|IFF_LOOPBACK) ||
4081 (net->ipv6.devconf_all->accept_dad < 1 &&
4082 idev->cnf.accept_dad < 1) ||
4083 !(ifp->flags&IFA_F_TENTATIVE) ||
4084 ifp->flags & IFA_F_NODAD) {
4085 bool send_na = false;
4086
4087 if (ifp->flags & IFA_F_TENTATIVE &&
4088 !(ifp->flags & IFA_F_OPTIMISTIC))
4089 send_na = true;
4090 bump_id = ifp->flags & IFA_F_TENTATIVE;
4091 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
4092 spin_unlock(&ifp->lock);
4093 read_unlock_bh(&idev->lock);
4094
4095 addrconf_dad_completed(ifp, bump_id, send_na);
4096 return;
4097 }
4098
4099 if (!(idev->if_flags & IF_READY)) {
4100 spin_unlock(&ifp->lock);
4101 read_unlock_bh(&idev->lock);
4102 /*
4103 * If the device is not ready:
4104 * - keep it tentative if it is a permanent address.
4105 * - otherwise, kill it.
4106 */
4107 in6_ifa_hold(ifp);
4108 addrconf_dad_stop(ifp, 0);
4109 return;
4110 }
4111
4112 /*
4113 * Optimistic nodes can start receiving
4114 * Frames right away
4115 */
4116 if (ifp->flags & IFA_F_OPTIMISTIC) {
4117 ip6_ins_rt(net, ifp->rt);
4118 if (ipv6_use_optimistic_addr(net, idev)) {
4119 /* Because optimistic nodes can use this address,
4120 * notify listeners. If DAD fails, RTM_DELADDR is sent.
4121 */
4122 notify = true;
4123 }
4124 }
4125
4126 addrconf_dad_kick(ifp);
4127 out:
4128 spin_unlock(&ifp->lock);
4129 read_unlock_bh(&idev->lock);
4130 if (notify)
4131 ipv6_ifa_notify(RTM_NEWADDR, ifp);
4132 }
4133
addrconf_dad_start(struct inet6_ifaddr * ifp)4134 static void addrconf_dad_start(struct inet6_ifaddr *ifp)
4135 {
4136 bool begin_dad = false;
4137
4138 spin_lock_bh(&ifp->lock);
4139 if (ifp->state != INET6_IFADDR_STATE_DEAD) {
4140 ifp->state = INET6_IFADDR_STATE_PREDAD;
4141 begin_dad = true;
4142 }
4143 spin_unlock_bh(&ifp->lock);
4144
4145 if (begin_dad)
4146 addrconf_mod_dad_work(ifp, 0);
4147 }
4148
addrconf_dad_work(struct work_struct * w)4149 static void addrconf_dad_work(struct work_struct *w)
4150 {
4151 struct inet6_ifaddr *ifp = container_of(to_delayed_work(w),
4152 struct inet6_ifaddr,
4153 dad_work);
4154 struct inet6_dev *idev = ifp->idev;
4155 bool bump_id, disable_ipv6 = false;
4156 struct in6_addr mcaddr;
4157
4158 enum {
4159 DAD_PROCESS,
4160 DAD_BEGIN,
4161 DAD_ABORT,
4162 } action = DAD_PROCESS;
4163
4164 rtnl_lock();
4165
4166 spin_lock_bh(&ifp->lock);
4167 if (ifp->state == INET6_IFADDR_STATE_PREDAD) {
4168 action = DAD_BEGIN;
4169 ifp->state = INET6_IFADDR_STATE_DAD;
4170 } else if (ifp->state == INET6_IFADDR_STATE_ERRDAD) {
4171 action = DAD_ABORT;
4172 ifp->state = INET6_IFADDR_STATE_POSTDAD;
4173
4174 if ((dev_net(idev->dev)->ipv6.devconf_all->accept_dad > 1 ||
4175 idev->cnf.accept_dad > 1) &&
4176 !idev->cnf.disable_ipv6 &&
4177 !(ifp->flags & IFA_F_STABLE_PRIVACY)) {
4178 struct in6_addr addr;
4179
4180 addr.s6_addr32[0] = htonl(0xfe800000);
4181 addr.s6_addr32[1] = 0;
4182
4183 if (!ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) &&
4184 ipv6_addr_equal(&ifp->addr, &addr)) {
4185 /* DAD failed for link-local based on MAC */
4186 WRITE_ONCE(idev->cnf.disable_ipv6, 1);
4187
4188 pr_info("%s: IPv6 being disabled!\n",
4189 ifp->idev->dev->name);
4190 disable_ipv6 = true;
4191 }
4192 }
4193 }
4194 spin_unlock_bh(&ifp->lock);
4195
4196 if (action == DAD_BEGIN) {
4197 addrconf_dad_begin(ifp);
4198 goto out;
4199 } else if (action == DAD_ABORT) {
4200 in6_ifa_hold(ifp);
4201 addrconf_dad_stop(ifp, 1);
4202 if (disable_ipv6)
4203 addrconf_ifdown(idev->dev, false);
4204 goto out;
4205 }
4206
4207 if (!ifp->dad_probes && addrconf_dad_end(ifp))
4208 goto out;
4209
4210 write_lock_bh(&idev->lock);
4211 if (idev->dead || !(idev->if_flags & IF_READY)) {
4212 write_unlock_bh(&idev->lock);
4213 goto out;
4214 }
4215
4216 spin_lock(&ifp->lock);
4217 if (ifp->state == INET6_IFADDR_STATE_DEAD) {
4218 spin_unlock(&ifp->lock);
4219 write_unlock_bh(&idev->lock);
4220 goto out;
4221 }
4222
4223 if (ifp->dad_probes == 0) {
4224 bool send_na = false;
4225
4226 /*
4227 * DAD was successful
4228 */
4229
4230 if (ifp->flags & IFA_F_TENTATIVE &&
4231 !(ifp->flags & IFA_F_OPTIMISTIC))
4232 send_na = true;
4233 bump_id = ifp->flags & IFA_F_TENTATIVE;
4234 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
4235 spin_unlock(&ifp->lock);
4236 write_unlock_bh(&idev->lock);
4237
4238 addrconf_dad_completed(ifp, bump_id, send_na);
4239
4240 goto out;
4241 }
4242
4243 ifp->dad_probes--;
4244 addrconf_mod_dad_work(ifp,
4245 max(NEIGH_VAR(ifp->idev->nd_parms, RETRANS_TIME),
4246 HZ/100));
4247 spin_unlock(&ifp->lock);
4248 write_unlock_bh(&idev->lock);
4249
4250 /* send a neighbour solicitation for our addr */
4251 addrconf_addr_solict_mult(&ifp->addr, &mcaddr);
4252 ndisc_send_ns(ifp->idev->dev, &ifp->addr, &mcaddr, &in6addr_any,
4253 ifp->dad_nonce);
4254 out:
4255 in6_ifa_put(ifp);
4256 rtnl_unlock();
4257 }
4258
4259 /* ifp->idev must be at least read locked */
ipv6_lonely_lladdr(struct inet6_ifaddr * ifp)4260 static bool ipv6_lonely_lladdr(struct inet6_ifaddr *ifp)
4261 {
4262 struct inet6_ifaddr *ifpiter;
4263 struct inet6_dev *idev = ifp->idev;
4264
4265 list_for_each_entry_reverse(ifpiter, &idev->addr_list, if_list) {
4266 if (ifpiter->scope > IFA_LINK)
4267 break;
4268 if (ifp != ifpiter && ifpiter->scope == IFA_LINK &&
4269 (ifpiter->flags & (IFA_F_PERMANENT|IFA_F_TENTATIVE|
4270 IFA_F_OPTIMISTIC|IFA_F_DADFAILED)) ==
4271 IFA_F_PERMANENT)
4272 return false;
4273 }
4274 return true;
4275 }
4276
addrconf_dad_completed(struct inet6_ifaddr * ifp,bool bump_id,bool send_na)4277 static void addrconf_dad_completed(struct inet6_ifaddr *ifp, bool bump_id,
4278 bool send_na)
4279 {
4280 struct net_device *dev = ifp->idev->dev;
4281 struct in6_addr lladdr;
4282 bool send_rs, send_mld;
4283
4284 addrconf_del_dad_work(ifp);
4285
4286 /*
4287 * Configure the address for reception. Now it is valid.
4288 */
4289
4290 ipv6_ifa_notify(RTM_NEWADDR, ifp);
4291
4292 /* If added prefix is link local and we are prepared to process
4293 router advertisements, start sending router solicitations.
4294 */
4295
4296 read_lock_bh(&ifp->idev->lock);
4297 send_mld = ifp->scope == IFA_LINK && ipv6_lonely_lladdr(ifp);
4298 send_rs = send_mld &&
4299 ipv6_accept_ra(ifp->idev) &&
4300 ifp->idev->cnf.rtr_solicits != 0 &&
4301 (dev->flags & IFF_LOOPBACK) == 0 &&
4302 (dev->type != ARPHRD_TUNNEL) &&
4303 !netif_is_team_port(dev);
4304 read_unlock_bh(&ifp->idev->lock);
4305
4306 /* While dad is in progress mld report's source address is in6_addrany.
4307 * Resend with proper ll now.
4308 */
4309 if (send_mld)
4310 ipv6_mc_dad_complete(ifp->idev);
4311
4312 /* send unsolicited NA if enabled */
4313 if (send_na &&
4314 (ifp->idev->cnf.ndisc_notify ||
4315 dev_net(dev)->ipv6.devconf_all->ndisc_notify)) {
4316 ndisc_send_na(dev, &in6addr_linklocal_allnodes, &ifp->addr,
4317 /*router=*/ !!ifp->idev->cnf.forwarding,
4318 /*solicited=*/ false, /*override=*/ true,
4319 /*inc_opt=*/ true);
4320 }
4321
4322 if (send_rs) {
4323 /*
4324 * If a host as already performed a random delay
4325 * [...] as part of DAD [...] there is no need
4326 * to delay again before sending the first RS
4327 */
4328 if (ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE))
4329 return;
4330 ndisc_send_rs(dev, &lladdr, &in6addr_linklocal_allrouters);
4331
4332 write_lock_bh(&ifp->idev->lock);
4333 spin_lock(&ifp->lock);
4334 ifp->idev->rs_interval = rfc3315_s14_backoff_init(
4335 ifp->idev->cnf.rtr_solicit_interval);
4336 ifp->idev->rs_probes = 1;
4337 ifp->idev->if_flags |= IF_RS_SENT;
4338 addrconf_mod_rs_timer(ifp->idev, ifp->idev->rs_interval);
4339 spin_unlock(&ifp->lock);
4340 write_unlock_bh(&ifp->idev->lock);
4341 }
4342
4343 if (bump_id)
4344 rt_genid_bump_ipv6(dev_net(dev));
4345
4346 /* Make sure that a new temporary address will be created
4347 * before this temporary address becomes deprecated.
4348 */
4349 if (ifp->flags & IFA_F_TEMPORARY)
4350 addrconf_verify_rtnl(dev_net(dev));
4351 }
4352
addrconf_dad_run(struct inet6_dev * idev,bool restart)4353 static void addrconf_dad_run(struct inet6_dev *idev, bool restart)
4354 {
4355 struct inet6_ifaddr *ifp;
4356
4357 read_lock_bh(&idev->lock);
4358 list_for_each_entry(ifp, &idev->addr_list, if_list) {
4359 spin_lock(&ifp->lock);
4360 if ((ifp->flags & IFA_F_TENTATIVE &&
4361 ifp->state == INET6_IFADDR_STATE_DAD) || restart) {
4362 if (restart)
4363 ifp->state = INET6_IFADDR_STATE_PREDAD;
4364 addrconf_dad_kick(ifp);
4365 }
4366 spin_unlock(&ifp->lock);
4367 }
4368 read_unlock_bh(&idev->lock);
4369 }
4370
4371 #ifdef CONFIG_PROC_FS
4372 struct if6_iter_state {
4373 struct seq_net_private p;
4374 int bucket;
4375 int offset;
4376 };
4377
if6_get_first(struct seq_file * seq,loff_t pos)4378 static struct inet6_ifaddr *if6_get_first(struct seq_file *seq, loff_t pos)
4379 {
4380 struct if6_iter_state *state = seq->private;
4381 struct net *net = seq_file_net(seq);
4382 struct inet6_ifaddr *ifa = NULL;
4383 int p = 0;
4384
4385 /* initial bucket if pos is 0 */
4386 if (pos == 0) {
4387 state->bucket = 0;
4388 state->offset = 0;
4389 }
4390
4391 for (; state->bucket < IN6_ADDR_HSIZE; ++state->bucket) {
4392 hlist_for_each_entry_rcu(ifa, &net->ipv6.inet6_addr_lst[state->bucket],
4393 addr_lst) {
4394 /* sync with offset */
4395 if (p < state->offset) {
4396 p++;
4397 continue;
4398 }
4399 return ifa;
4400 }
4401
4402 /* prepare for next bucket */
4403 state->offset = 0;
4404 p = 0;
4405 }
4406 return NULL;
4407 }
4408
if6_get_next(struct seq_file * seq,struct inet6_ifaddr * ifa)4409 static struct inet6_ifaddr *if6_get_next(struct seq_file *seq,
4410 struct inet6_ifaddr *ifa)
4411 {
4412 struct if6_iter_state *state = seq->private;
4413 struct net *net = seq_file_net(seq);
4414
4415 hlist_for_each_entry_continue_rcu(ifa, addr_lst) {
4416 state->offset++;
4417 return ifa;
4418 }
4419
4420 state->offset = 0;
4421 while (++state->bucket < IN6_ADDR_HSIZE) {
4422 hlist_for_each_entry_rcu(ifa,
4423 &net->ipv6.inet6_addr_lst[state->bucket], addr_lst) {
4424 return ifa;
4425 }
4426 }
4427
4428 return NULL;
4429 }
4430
if6_seq_start(struct seq_file * seq,loff_t * pos)4431 static void *if6_seq_start(struct seq_file *seq, loff_t *pos)
4432 __acquires(rcu)
4433 {
4434 rcu_read_lock();
4435 return if6_get_first(seq, *pos);
4436 }
4437
if6_seq_next(struct seq_file * seq,void * v,loff_t * pos)4438 static void *if6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
4439 {
4440 struct inet6_ifaddr *ifa;
4441
4442 ifa = if6_get_next(seq, v);
4443 ++*pos;
4444 return ifa;
4445 }
4446
if6_seq_stop(struct seq_file * seq,void * v)4447 static void if6_seq_stop(struct seq_file *seq, void *v)
4448 __releases(rcu)
4449 {
4450 rcu_read_unlock();
4451 }
4452
if6_seq_show(struct seq_file * seq,void * v)4453 static int if6_seq_show(struct seq_file *seq, void *v)
4454 {
4455 struct inet6_ifaddr *ifp = (struct inet6_ifaddr *)v;
4456 seq_printf(seq, "%pi6 %02x %02x %02x %02x %8s\n",
4457 &ifp->addr,
4458 ifp->idev->dev->ifindex,
4459 ifp->prefix_len,
4460 ifp->scope,
4461 (u8) ifp->flags,
4462 ifp->idev->dev->name);
4463 return 0;
4464 }
4465
4466 static const struct seq_operations if6_seq_ops = {
4467 .start = if6_seq_start,
4468 .next = if6_seq_next,
4469 .show = if6_seq_show,
4470 .stop = if6_seq_stop,
4471 };
4472
if6_proc_net_init(struct net * net)4473 static int __net_init if6_proc_net_init(struct net *net)
4474 {
4475 if (!proc_create_net("if_inet6", 0444, net->proc_net, &if6_seq_ops,
4476 sizeof(struct if6_iter_state)))
4477 return -ENOMEM;
4478 return 0;
4479 }
4480
if6_proc_net_exit(struct net * net)4481 static void __net_exit if6_proc_net_exit(struct net *net)
4482 {
4483 remove_proc_entry("if_inet6", net->proc_net);
4484 }
4485
4486 static struct pernet_operations if6_proc_net_ops = {
4487 .init = if6_proc_net_init,
4488 .exit = if6_proc_net_exit,
4489 };
4490
if6_proc_init(void)4491 int __init if6_proc_init(void)
4492 {
4493 return register_pernet_subsys(&if6_proc_net_ops);
4494 }
4495
if6_proc_exit(void)4496 void if6_proc_exit(void)
4497 {
4498 unregister_pernet_subsys(&if6_proc_net_ops);
4499 }
4500 #endif /* CONFIG_PROC_FS */
4501
4502 #if IS_ENABLED(CONFIG_IPV6_MIP6)
4503 /* Check if address is a home address configured on any interface. */
ipv6_chk_home_addr(struct net * net,const struct in6_addr * addr)4504 int ipv6_chk_home_addr(struct net *net, const struct in6_addr *addr)
4505 {
4506 unsigned int hash = inet6_addr_hash(net, addr);
4507 struct inet6_ifaddr *ifp = NULL;
4508 int ret = 0;
4509
4510 rcu_read_lock();
4511 hlist_for_each_entry_rcu(ifp, &net->ipv6.inet6_addr_lst[hash], addr_lst) {
4512 if (ipv6_addr_equal(&ifp->addr, addr) &&
4513 (ifp->flags & IFA_F_HOMEADDRESS)) {
4514 ret = 1;
4515 break;
4516 }
4517 }
4518 rcu_read_unlock();
4519 return ret;
4520 }
4521 #endif
4522
4523 /* RFC6554 has some algorithm to avoid loops in segment routing by
4524 * checking if the segments contains any of a local interface address.
4525 *
4526 * Quote:
4527 *
4528 * To detect loops in the SRH, a router MUST determine if the SRH
4529 * includes multiple addresses assigned to any interface on that router.
4530 * If such addresses appear more than once and are separated by at least
4531 * one address not assigned to that router.
4532 */
ipv6_chk_rpl_srh_loop(struct net * net,const struct in6_addr * segs,unsigned char nsegs)4533 int ipv6_chk_rpl_srh_loop(struct net *net, const struct in6_addr *segs,
4534 unsigned char nsegs)
4535 {
4536 const struct in6_addr *addr;
4537 int i, ret = 0, found = 0;
4538 struct inet6_ifaddr *ifp;
4539 bool separated = false;
4540 unsigned int hash;
4541 bool hash_found;
4542
4543 rcu_read_lock();
4544 for (i = 0; i < nsegs; i++) {
4545 addr = &segs[i];
4546 hash = inet6_addr_hash(net, addr);
4547
4548 hash_found = false;
4549 hlist_for_each_entry_rcu(ifp, &net->ipv6.inet6_addr_lst[hash], addr_lst) {
4550
4551 if (ipv6_addr_equal(&ifp->addr, addr)) {
4552 hash_found = true;
4553 break;
4554 }
4555 }
4556
4557 if (hash_found) {
4558 if (found > 1 && separated) {
4559 ret = 1;
4560 break;
4561 }
4562
4563 separated = false;
4564 found++;
4565 } else {
4566 separated = true;
4567 }
4568 }
4569 rcu_read_unlock();
4570
4571 return ret;
4572 }
4573
4574 /*
4575 * Periodic address status verification
4576 */
4577
addrconf_verify_rtnl(struct net * net)4578 static void addrconf_verify_rtnl(struct net *net)
4579 {
4580 unsigned long now, next, next_sec, next_sched;
4581 struct inet6_ifaddr *ifp;
4582 int i;
4583
4584 ASSERT_RTNL();
4585
4586 rcu_read_lock_bh();
4587 now = jiffies;
4588 next = round_jiffies_up(now + ADDR_CHECK_FREQUENCY);
4589
4590 cancel_delayed_work(&net->ipv6.addr_chk_work);
4591
4592 for (i = 0; i < IN6_ADDR_HSIZE; i++) {
4593 restart:
4594 hlist_for_each_entry_rcu_bh(ifp, &net->ipv6.inet6_addr_lst[i], addr_lst) {
4595 unsigned long age;
4596
4597 /* When setting preferred_lft to a value not zero or
4598 * infinity, while valid_lft is infinity
4599 * IFA_F_PERMANENT has a non-infinity life time.
4600 */
4601 if ((ifp->flags & IFA_F_PERMANENT) &&
4602 (ifp->prefered_lft == INFINITY_LIFE_TIME))
4603 continue;
4604
4605 spin_lock(&ifp->lock);
4606 /* We try to batch several events at once. */
4607 age = (now - ifp->tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
4608
4609 if ((ifp->flags&IFA_F_TEMPORARY) &&
4610 !(ifp->flags&IFA_F_TENTATIVE) &&
4611 ifp->prefered_lft != INFINITY_LIFE_TIME &&
4612 !ifp->regen_count && ifp->ifpub) {
4613 /* This is a non-regenerated temporary addr. */
4614
4615 unsigned long regen_advance = ifp->idev->cnf.regen_max_retry *
4616 ifp->idev->cnf.dad_transmits *
4617 max(NEIGH_VAR(ifp->idev->nd_parms, RETRANS_TIME), HZ/100) / HZ;
4618
4619 if (age + regen_advance >= ifp->prefered_lft) {
4620 struct inet6_ifaddr *ifpub = ifp->ifpub;
4621 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
4622 next = ifp->tstamp + ifp->prefered_lft * HZ;
4623
4624 ifp->regen_count++;
4625 in6_ifa_hold(ifp);
4626 in6_ifa_hold(ifpub);
4627 spin_unlock(&ifp->lock);
4628
4629 spin_lock(&ifpub->lock);
4630 ifpub->regen_count = 0;
4631 spin_unlock(&ifpub->lock);
4632 rcu_read_unlock_bh();
4633 ipv6_create_tempaddr(ifpub, true);
4634 in6_ifa_put(ifpub);
4635 in6_ifa_put(ifp);
4636 rcu_read_lock_bh();
4637 goto restart;
4638 } else if (time_before(ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ, next))
4639 next = ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ;
4640 }
4641
4642 if (ifp->valid_lft != INFINITY_LIFE_TIME &&
4643 age >= ifp->valid_lft) {
4644 spin_unlock(&ifp->lock);
4645 in6_ifa_hold(ifp);
4646 rcu_read_unlock_bh();
4647 ipv6_del_addr(ifp);
4648 rcu_read_lock_bh();
4649 goto restart;
4650 } else if (ifp->prefered_lft == INFINITY_LIFE_TIME) {
4651 spin_unlock(&ifp->lock);
4652 continue;
4653 } else if (age >= ifp->prefered_lft) {
4654 /* jiffies - ifp->tstamp > age >= ifp->prefered_lft */
4655 int deprecate = 0;
4656
4657 if (!(ifp->flags&IFA_F_DEPRECATED)) {
4658 deprecate = 1;
4659 ifp->flags |= IFA_F_DEPRECATED;
4660 }
4661
4662 if ((ifp->valid_lft != INFINITY_LIFE_TIME) &&
4663 (time_before(ifp->tstamp + ifp->valid_lft * HZ, next)))
4664 next = ifp->tstamp + ifp->valid_lft * HZ;
4665
4666 spin_unlock(&ifp->lock);
4667
4668 if (deprecate) {
4669 in6_ifa_hold(ifp);
4670
4671 ipv6_ifa_notify(0, ifp);
4672 in6_ifa_put(ifp);
4673 goto restart;
4674 }
4675 } else {
4676 /* ifp->prefered_lft <= ifp->valid_lft */
4677 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
4678 next = ifp->tstamp + ifp->prefered_lft * HZ;
4679 spin_unlock(&ifp->lock);
4680 }
4681 }
4682 }
4683
4684 next_sec = round_jiffies_up(next);
4685 next_sched = next;
4686
4687 /* If rounded timeout is accurate enough, accept it. */
4688 if (time_before(next_sec, next + ADDRCONF_TIMER_FUZZ))
4689 next_sched = next_sec;
4690
4691 /* And minimum interval is ADDRCONF_TIMER_FUZZ_MAX. */
4692 if (time_before(next_sched, jiffies + ADDRCONF_TIMER_FUZZ_MAX))
4693 next_sched = jiffies + ADDRCONF_TIMER_FUZZ_MAX;
4694
4695 pr_debug("now = %lu, schedule = %lu, rounded schedule = %lu => %lu\n",
4696 now, next, next_sec, next_sched);
4697 mod_delayed_work(addrconf_wq, &net->ipv6.addr_chk_work, next_sched - now);
4698 rcu_read_unlock_bh();
4699 }
4700
addrconf_verify_work(struct work_struct * w)4701 static void addrconf_verify_work(struct work_struct *w)
4702 {
4703 struct net *net = container_of(to_delayed_work(w), struct net,
4704 ipv6.addr_chk_work);
4705
4706 rtnl_lock();
4707 addrconf_verify_rtnl(net);
4708 rtnl_unlock();
4709 }
4710
addrconf_verify(struct net * net)4711 static void addrconf_verify(struct net *net)
4712 {
4713 mod_delayed_work(addrconf_wq, &net->ipv6.addr_chk_work, 0);
4714 }
4715
extract_addr(struct nlattr * addr,struct nlattr * local,struct in6_addr ** peer_pfx)4716 static struct in6_addr *extract_addr(struct nlattr *addr, struct nlattr *local,
4717 struct in6_addr **peer_pfx)
4718 {
4719 struct in6_addr *pfx = NULL;
4720
4721 *peer_pfx = NULL;
4722
4723 if (addr)
4724 pfx = nla_data(addr);
4725
4726 if (local) {
4727 if (pfx && nla_memcmp(local, pfx, sizeof(*pfx)))
4728 *peer_pfx = pfx;
4729 pfx = nla_data(local);
4730 }
4731
4732 return pfx;
4733 }
4734
4735 static const struct nla_policy ifa_ipv6_policy[IFA_MAX+1] = {
4736 [IFA_ADDRESS] = { .len = sizeof(struct in6_addr) },
4737 [IFA_LOCAL] = { .len = sizeof(struct in6_addr) },
4738 [IFA_CACHEINFO] = { .len = sizeof(struct ifa_cacheinfo) },
4739 [IFA_FLAGS] = { .len = sizeof(u32) },
4740 [IFA_RT_PRIORITY] = { .len = sizeof(u32) },
4741 [IFA_TARGET_NETNSID] = { .type = NLA_S32 },
4742 [IFA_PROTO] = { .type = NLA_U8 },
4743 };
4744
4745 static int
inet6_rtm_deladdr(struct sk_buff * skb,struct nlmsghdr * nlh,struct netlink_ext_ack * extack)4746 inet6_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh,
4747 struct netlink_ext_ack *extack)
4748 {
4749 struct net *net = sock_net(skb->sk);
4750 struct ifaddrmsg *ifm;
4751 struct nlattr *tb[IFA_MAX+1];
4752 struct in6_addr *pfx, *peer_pfx;
4753 u32 ifa_flags;
4754 int err;
4755
4756 err = nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFA_MAX,
4757 ifa_ipv6_policy, extack);
4758 if (err < 0)
4759 return err;
4760
4761 ifm = nlmsg_data(nlh);
4762 pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
4763 if (!pfx)
4764 return -EINVAL;
4765
4766 ifa_flags = tb[IFA_FLAGS] ? nla_get_u32(tb[IFA_FLAGS]) : ifm->ifa_flags;
4767
4768 /* We ignore other flags so far. */
4769 ifa_flags &= IFA_F_MANAGETEMPADDR;
4770
4771 return inet6_addr_del(net, ifm->ifa_index, ifa_flags, pfx,
4772 ifm->ifa_prefixlen, extack);
4773 }
4774
modify_prefix_route(struct inet6_ifaddr * ifp,unsigned long expires,u32 flags,bool modify_peer)4775 static int modify_prefix_route(struct inet6_ifaddr *ifp,
4776 unsigned long expires, u32 flags,
4777 bool modify_peer)
4778 {
4779 struct fib6_info *f6i;
4780 u32 prio;
4781
4782 f6i = addrconf_get_prefix_route(modify_peer ? &ifp->peer_addr : &ifp->addr,
4783 ifp->prefix_len,
4784 ifp->idev->dev, 0, RTF_DEFAULT, true);
4785 if (!f6i)
4786 return -ENOENT;
4787
4788 prio = ifp->rt_priority ? : IP6_RT_PRIO_ADDRCONF;
4789 if (f6i->fib6_metric != prio) {
4790 /* delete old one */
4791 ip6_del_rt(dev_net(ifp->idev->dev), f6i, false);
4792
4793 /* add new one */
4794 addrconf_prefix_route(modify_peer ? &ifp->peer_addr : &ifp->addr,
4795 ifp->prefix_len,
4796 ifp->rt_priority, ifp->idev->dev,
4797 expires, flags, GFP_KERNEL);
4798 } else {
4799 if (!expires)
4800 fib6_clean_expires(f6i);
4801 else
4802 fib6_set_expires(f6i, expires);
4803
4804 fib6_info_release(f6i);
4805 }
4806
4807 return 0;
4808 }
4809
inet6_addr_modify(struct net * net,struct inet6_ifaddr * ifp,struct ifa6_config * cfg)4810 static int inet6_addr_modify(struct net *net, struct inet6_ifaddr *ifp,
4811 struct ifa6_config *cfg)
4812 {
4813 u32 flags;
4814 clock_t expires;
4815 unsigned long timeout;
4816 bool was_managetempaddr;
4817 bool had_prefixroute;
4818 bool new_peer = false;
4819
4820 ASSERT_RTNL();
4821
4822 if (!cfg->valid_lft || cfg->preferred_lft > cfg->valid_lft)
4823 return -EINVAL;
4824
4825 if (cfg->ifa_flags & IFA_F_MANAGETEMPADDR &&
4826 (ifp->flags & IFA_F_TEMPORARY || ifp->prefix_len != 64))
4827 return -EINVAL;
4828
4829 if (!(ifp->flags & IFA_F_TENTATIVE) || ifp->flags & IFA_F_DADFAILED)
4830 cfg->ifa_flags &= ~IFA_F_OPTIMISTIC;
4831
4832 timeout = addrconf_timeout_fixup(cfg->valid_lft, HZ);
4833 if (addrconf_finite_timeout(timeout)) {
4834 expires = jiffies_to_clock_t(timeout * HZ);
4835 cfg->valid_lft = timeout;
4836 flags = RTF_EXPIRES;
4837 } else {
4838 expires = 0;
4839 flags = 0;
4840 cfg->ifa_flags |= IFA_F_PERMANENT;
4841 }
4842
4843 timeout = addrconf_timeout_fixup(cfg->preferred_lft, HZ);
4844 if (addrconf_finite_timeout(timeout)) {
4845 if (timeout == 0)
4846 cfg->ifa_flags |= IFA_F_DEPRECATED;
4847 cfg->preferred_lft = timeout;
4848 }
4849
4850 if (cfg->peer_pfx &&
4851 memcmp(&ifp->peer_addr, cfg->peer_pfx, sizeof(struct in6_addr))) {
4852 if (!ipv6_addr_any(&ifp->peer_addr))
4853 cleanup_prefix_route(ifp, expires, true, true);
4854 new_peer = true;
4855 }
4856
4857 spin_lock_bh(&ifp->lock);
4858 was_managetempaddr = ifp->flags & IFA_F_MANAGETEMPADDR;
4859 had_prefixroute = ifp->flags & IFA_F_PERMANENT &&
4860 !(ifp->flags & IFA_F_NOPREFIXROUTE);
4861 ifp->flags &= ~(IFA_F_DEPRECATED | IFA_F_PERMANENT | IFA_F_NODAD |
4862 IFA_F_HOMEADDRESS | IFA_F_MANAGETEMPADDR |
4863 IFA_F_NOPREFIXROUTE);
4864 ifp->flags |= cfg->ifa_flags;
4865 ifp->tstamp = jiffies;
4866 ifp->valid_lft = cfg->valid_lft;
4867 ifp->prefered_lft = cfg->preferred_lft;
4868 ifp->ifa_proto = cfg->ifa_proto;
4869
4870 if (cfg->rt_priority && cfg->rt_priority != ifp->rt_priority)
4871 ifp->rt_priority = cfg->rt_priority;
4872
4873 if (new_peer)
4874 ifp->peer_addr = *cfg->peer_pfx;
4875
4876 spin_unlock_bh(&ifp->lock);
4877 if (!(ifp->flags&IFA_F_TENTATIVE))
4878 ipv6_ifa_notify(0, ifp);
4879
4880 if (!(cfg->ifa_flags & IFA_F_NOPREFIXROUTE)) {
4881 int rc = -ENOENT;
4882
4883 if (had_prefixroute)
4884 rc = modify_prefix_route(ifp, expires, flags, false);
4885
4886 /* prefix route could have been deleted; if so restore it */
4887 if (rc == -ENOENT) {
4888 addrconf_prefix_route(&ifp->addr, ifp->prefix_len,
4889 ifp->rt_priority, ifp->idev->dev,
4890 expires, flags, GFP_KERNEL);
4891 }
4892
4893 if (had_prefixroute && !ipv6_addr_any(&ifp->peer_addr))
4894 rc = modify_prefix_route(ifp, expires, flags, true);
4895
4896 if (rc == -ENOENT && !ipv6_addr_any(&ifp->peer_addr)) {
4897 addrconf_prefix_route(&ifp->peer_addr, ifp->prefix_len,
4898 ifp->rt_priority, ifp->idev->dev,
4899 expires, flags, GFP_KERNEL);
4900 }
4901 } else if (had_prefixroute) {
4902 enum cleanup_prefix_rt_t action;
4903 unsigned long rt_expires;
4904
4905 write_lock_bh(&ifp->idev->lock);
4906 action = check_cleanup_prefix_route(ifp, &rt_expires);
4907 write_unlock_bh(&ifp->idev->lock);
4908
4909 if (action != CLEANUP_PREFIX_RT_NOP) {
4910 cleanup_prefix_route(ifp, rt_expires,
4911 action == CLEANUP_PREFIX_RT_DEL, false);
4912 }
4913 }
4914
4915 if (was_managetempaddr || ifp->flags & IFA_F_MANAGETEMPADDR) {
4916 if (was_managetempaddr &&
4917 !(ifp->flags & IFA_F_MANAGETEMPADDR)) {
4918 cfg->valid_lft = 0;
4919 cfg->preferred_lft = 0;
4920 }
4921 manage_tempaddrs(ifp->idev, ifp, cfg->valid_lft,
4922 cfg->preferred_lft, !was_managetempaddr,
4923 jiffies);
4924 }
4925
4926 addrconf_verify_rtnl(net);
4927
4928 return 0;
4929 }
4930
4931 static int
inet6_rtm_newaddr(struct sk_buff * skb,struct nlmsghdr * nlh,struct netlink_ext_ack * extack)4932 inet6_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh,
4933 struct netlink_ext_ack *extack)
4934 {
4935 struct net *net = sock_net(skb->sk);
4936 struct ifaddrmsg *ifm;
4937 struct nlattr *tb[IFA_MAX+1];
4938 struct in6_addr *peer_pfx;
4939 struct inet6_ifaddr *ifa;
4940 struct net_device *dev;
4941 struct inet6_dev *idev;
4942 struct ifa6_config cfg;
4943 int err;
4944
4945 err = nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFA_MAX,
4946 ifa_ipv6_policy, extack);
4947 if (err < 0)
4948 return err;
4949
4950 memset(&cfg, 0, sizeof(cfg));
4951
4952 ifm = nlmsg_data(nlh);
4953 cfg.pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
4954 if (!cfg.pfx)
4955 return -EINVAL;
4956
4957 cfg.peer_pfx = peer_pfx;
4958 cfg.plen = ifm->ifa_prefixlen;
4959 if (tb[IFA_RT_PRIORITY])
4960 cfg.rt_priority = nla_get_u32(tb[IFA_RT_PRIORITY]);
4961
4962 if (tb[IFA_PROTO])
4963 cfg.ifa_proto = nla_get_u8(tb[IFA_PROTO]);
4964
4965 cfg.valid_lft = INFINITY_LIFE_TIME;
4966 cfg.preferred_lft = INFINITY_LIFE_TIME;
4967
4968 if (tb[IFA_CACHEINFO]) {
4969 struct ifa_cacheinfo *ci;
4970
4971 ci = nla_data(tb[IFA_CACHEINFO]);
4972 cfg.valid_lft = ci->ifa_valid;
4973 cfg.preferred_lft = ci->ifa_prefered;
4974 }
4975
4976 dev = __dev_get_by_index(net, ifm->ifa_index);
4977 if (!dev) {
4978 NL_SET_ERR_MSG_MOD(extack, "Unable to find the interface");
4979 return -ENODEV;
4980 }
4981
4982 if (tb[IFA_FLAGS])
4983 cfg.ifa_flags = nla_get_u32(tb[IFA_FLAGS]);
4984 else
4985 cfg.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 idev = ipv6_find_idev(dev);
4993 if (IS_ERR(idev))
4994 return PTR_ERR(idev);
4995
4996 if (!ipv6_allow_optimistic_dad(net, idev))
4997 cfg.ifa_flags &= ~IFA_F_OPTIMISTIC;
4998
4999 if (cfg.ifa_flags & IFA_F_NODAD &&
5000 cfg.ifa_flags & IFA_F_OPTIMISTIC) {
5001 NL_SET_ERR_MSG(extack, "IFA_F_NODAD and IFA_F_OPTIMISTIC are mutually exclusive");
5002 return -EINVAL;
5003 }
5004
5005 ifa = ipv6_get_ifaddr(net, cfg.pfx, dev, 1);
5006 if (!ifa) {
5007 /*
5008 * It would be best to check for !NLM_F_CREATE here but
5009 * userspace already relies on not having to provide this.
5010 */
5011 return inet6_addr_add(net, ifm->ifa_index, &cfg, extack);
5012 }
5013
5014 if (nlh->nlmsg_flags & NLM_F_EXCL ||
5015 !(nlh->nlmsg_flags & NLM_F_REPLACE)) {
5016 NL_SET_ERR_MSG_MOD(extack, "address already assigned");
5017 err = -EEXIST;
5018 } else {
5019 err = inet6_addr_modify(net, ifa, &cfg);
5020 }
5021
5022 in6_ifa_put(ifa);
5023
5024 return err;
5025 }
5026
put_ifaddrmsg(struct nlmsghdr * nlh,u8 prefixlen,u32 flags,u8 scope,int ifindex)5027 static void put_ifaddrmsg(struct nlmsghdr *nlh, u8 prefixlen, u32 flags,
5028 u8 scope, int ifindex)
5029 {
5030 struct ifaddrmsg *ifm;
5031
5032 ifm = nlmsg_data(nlh);
5033 ifm->ifa_family = AF_INET6;
5034 ifm->ifa_prefixlen = prefixlen;
5035 ifm->ifa_flags = flags;
5036 ifm->ifa_scope = scope;
5037 ifm->ifa_index = ifindex;
5038 }
5039
put_cacheinfo(struct sk_buff * skb,unsigned long cstamp,unsigned long tstamp,u32 preferred,u32 valid)5040 static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
5041 unsigned long tstamp, u32 preferred, u32 valid)
5042 {
5043 struct ifa_cacheinfo ci;
5044
5045 ci.cstamp = cstamp_delta(cstamp);
5046 ci.tstamp = cstamp_delta(tstamp);
5047 ci.ifa_prefered = preferred;
5048 ci.ifa_valid = valid;
5049
5050 return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
5051 }
5052
rt_scope(int ifa_scope)5053 static inline int rt_scope(int ifa_scope)
5054 {
5055 if (ifa_scope & IFA_HOST)
5056 return RT_SCOPE_HOST;
5057 else if (ifa_scope & IFA_LINK)
5058 return RT_SCOPE_LINK;
5059 else if (ifa_scope & IFA_SITE)
5060 return RT_SCOPE_SITE;
5061 else
5062 return RT_SCOPE_UNIVERSE;
5063 }
5064
inet6_ifaddr_msgsize(void)5065 static inline int inet6_ifaddr_msgsize(void)
5066 {
5067 return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
5068 + nla_total_size(16) /* IFA_LOCAL */
5069 + nla_total_size(16) /* IFA_ADDRESS */
5070 + nla_total_size(sizeof(struct ifa_cacheinfo))
5071 + nla_total_size(4) /* IFA_FLAGS */
5072 + nla_total_size(1) /* IFA_PROTO */
5073 + nla_total_size(4) /* IFA_RT_PRIORITY */;
5074 }
5075
5076 enum addr_type_t {
5077 UNICAST_ADDR,
5078 MULTICAST_ADDR,
5079 ANYCAST_ADDR,
5080 };
5081
5082 struct inet6_fill_args {
5083 u32 portid;
5084 u32 seq;
5085 int event;
5086 unsigned int flags;
5087 int netnsid;
5088 int ifindex;
5089 enum addr_type_t type;
5090 };
5091
inet6_fill_ifaddr(struct sk_buff * skb,struct inet6_ifaddr * ifa,struct inet6_fill_args * args)5092 static int inet6_fill_ifaddr(struct sk_buff *skb, struct inet6_ifaddr *ifa,
5093 struct inet6_fill_args *args)
5094 {
5095 struct nlmsghdr *nlh;
5096 u32 preferred, valid;
5097
5098 nlh = nlmsg_put(skb, args->portid, args->seq, args->event,
5099 sizeof(struct ifaddrmsg), args->flags);
5100 if (!nlh)
5101 return -EMSGSIZE;
5102
5103 put_ifaddrmsg(nlh, ifa->prefix_len, ifa->flags, rt_scope(ifa->scope),
5104 ifa->idev->dev->ifindex);
5105
5106 if (args->netnsid >= 0 &&
5107 nla_put_s32(skb, IFA_TARGET_NETNSID, args->netnsid))
5108 goto error;
5109
5110 spin_lock_bh(&ifa->lock);
5111 if (!((ifa->flags&IFA_F_PERMANENT) &&
5112 (ifa->prefered_lft == INFINITY_LIFE_TIME))) {
5113 preferred = ifa->prefered_lft;
5114 valid = ifa->valid_lft;
5115 if (preferred != INFINITY_LIFE_TIME) {
5116 long tval = (jiffies - ifa->tstamp)/HZ;
5117 if (preferred > tval)
5118 preferred -= tval;
5119 else
5120 preferred = 0;
5121 if (valid != INFINITY_LIFE_TIME) {
5122 if (valid > tval)
5123 valid -= tval;
5124 else
5125 valid = 0;
5126 }
5127 }
5128 } else {
5129 preferred = INFINITY_LIFE_TIME;
5130 valid = INFINITY_LIFE_TIME;
5131 }
5132 spin_unlock_bh(&ifa->lock);
5133
5134 if (!ipv6_addr_any(&ifa->peer_addr)) {
5135 if (nla_put_in6_addr(skb, IFA_LOCAL, &ifa->addr) < 0 ||
5136 nla_put_in6_addr(skb, IFA_ADDRESS, &ifa->peer_addr) < 0)
5137 goto error;
5138 } else
5139 if (nla_put_in6_addr(skb, IFA_ADDRESS, &ifa->addr) < 0)
5140 goto error;
5141
5142 if (ifa->rt_priority &&
5143 nla_put_u32(skb, IFA_RT_PRIORITY, ifa->rt_priority))
5144 goto error;
5145
5146 if (put_cacheinfo(skb, ifa->cstamp, ifa->tstamp, preferred, valid) < 0)
5147 goto error;
5148
5149 if (nla_put_u32(skb, IFA_FLAGS, ifa->flags) < 0)
5150 goto error;
5151
5152 if (ifa->ifa_proto &&
5153 nla_put_u8(skb, IFA_PROTO, ifa->ifa_proto))
5154 goto error;
5155
5156 nlmsg_end(skb, nlh);
5157 return 0;
5158
5159 error:
5160 nlmsg_cancel(skb, nlh);
5161 return -EMSGSIZE;
5162 }
5163
inet6_fill_ifmcaddr(struct sk_buff * skb,struct ifmcaddr6 * ifmca,struct inet6_fill_args * args)5164 static int inet6_fill_ifmcaddr(struct sk_buff *skb, struct ifmcaddr6 *ifmca,
5165 struct inet6_fill_args *args)
5166 {
5167 struct nlmsghdr *nlh;
5168 u8 scope = RT_SCOPE_UNIVERSE;
5169 int ifindex = ifmca->idev->dev->ifindex;
5170
5171 if (ipv6_addr_scope(&ifmca->mca_addr) & IFA_SITE)
5172 scope = RT_SCOPE_SITE;
5173
5174 nlh = nlmsg_put(skb, args->portid, args->seq, args->event,
5175 sizeof(struct ifaddrmsg), args->flags);
5176 if (!nlh)
5177 return -EMSGSIZE;
5178
5179 if (args->netnsid >= 0 &&
5180 nla_put_s32(skb, IFA_TARGET_NETNSID, args->netnsid)) {
5181 nlmsg_cancel(skb, nlh);
5182 return -EMSGSIZE;
5183 }
5184
5185 put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
5186 if (nla_put_in6_addr(skb, IFA_MULTICAST, &ifmca->mca_addr) < 0 ||
5187 put_cacheinfo(skb, ifmca->mca_cstamp, ifmca->mca_tstamp,
5188 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
5189 nlmsg_cancel(skb, nlh);
5190 return -EMSGSIZE;
5191 }
5192
5193 nlmsg_end(skb, nlh);
5194 return 0;
5195 }
5196
inet6_fill_ifacaddr(struct sk_buff * skb,struct ifacaddr6 * ifaca,struct inet6_fill_args * args)5197 static int inet6_fill_ifacaddr(struct sk_buff *skb, struct ifacaddr6 *ifaca,
5198 struct inet6_fill_args *args)
5199 {
5200 struct net_device *dev = fib6_info_nh_dev(ifaca->aca_rt);
5201 int ifindex = dev ? dev->ifindex : 1;
5202 struct nlmsghdr *nlh;
5203 u8 scope = RT_SCOPE_UNIVERSE;
5204
5205 if (ipv6_addr_scope(&ifaca->aca_addr) & IFA_SITE)
5206 scope = RT_SCOPE_SITE;
5207
5208 nlh = nlmsg_put(skb, args->portid, args->seq, args->event,
5209 sizeof(struct ifaddrmsg), args->flags);
5210 if (!nlh)
5211 return -EMSGSIZE;
5212
5213 if (args->netnsid >= 0 &&
5214 nla_put_s32(skb, IFA_TARGET_NETNSID, args->netnsid)) {
5215 nlmsg_cancel(skb, nlh);
5216 return -EMSGSIZE;
5217 }
5218
5219 put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
5220 if (nla_put_in6_addr(skb, IFA_ANYCAST, &ifaca->aca_addr) < 0 ||
5221 put_cacheinfo(skb, ifaca->aca_cstamp, ifaca->aca_tstamp,
5222 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
5223 nlmsg_cancel(skb, nlh);
5224 return -EMSGSIZE;
5225 }
5226
5227 nlmsg_end(skb, nlh);
5228 return 0;
5229 }
5230
5231 /* called with rcu_read_lock() */
in6_dump_addrs(struct inet6_dev * idev,struct sk_buff * skb,struct netlink_callback * cb,int s_ip_idx,struct inet6_fill_args * fillargs)5232 static int in6_dump_addrs(struct inet6_dev *idev, struct sk_buff *skb,
5233 struct netlink_callback *cb, int s_ip_idx,
5234 struct inet6_fill_args *fillargs)
5235 {
5236 struct ifmcaddr6 *ifmca;
5237 struct ifacaddr6 *ifaca;
5238 int ip_idx = 0;
5239 int err = 1;
5240
5241 read_lock_bh(&idev->lock);
5242 switch (fillargs->type) {
5243 case UNICAST_ADDR: {
5244 struct inet6_ifaddr *ifa;
5245 fillargs->event = RTM_NEWADDR;
5246
5247 /* unicast address incl. temp addr */
5248 list_for_each_entry(ifa, &idev->addr_list, if_list) {
5249 if (ip_idx < s_ip_idx)
5250 goto next;
5251 err = inet6_fill_ifaddr(skb, ifa, fillargs);
5252 if (err < 0)
5253 break;
5254 nl_dump_check_consistent(cb, nlmsg_hdr(skb));
5255 next:
5256 ip_idx++;
5257 }
5258 break;
5259 }
5260 case MULTICAST_ADDR:
5261 read_unlock_bh(&idev->lock);
5262 fillargs->event = RTM_GETMULTICAST;
5263
5264 /* multicast address */
5265 for (ifmca = rtnl_dereference(idev->mc_list);
5266 ifmca;
5267 ifmca = rtnl_dereference(ifmca->next), ip_idx++) {
5268 if (ip_idx < s_ip_idx)
5269 continue;
5270 err = inet6_fill_ifmcaddr(skb, ifmca, fillargs);
5271 if (err < 0)
5272 break;
5273 }
5274 read_lock_bh(&idev->lock);
5275 break;
5276 case ANYCAST_ADDR:
5277 fillargs->event = RTM_GETANYCAST;
5278 /* anycast address */
5279 for (ifaca = idev->ac_list; ifaca;
5280 ifaca = ifaca->aca_next, ip_idx++) {
5281 if (ip_idx < s_ip_idx)
5282 continue;
5283 err = inet6_fill_ifacaddr(skb, ifaca, fillargs);
5284 if (err < 0)
5285 break;
5286 }
5287 break;
5288 default:
5289 break;
5290 }
5291 read_unlock_bh(&idev->lock);
5292 cb->args[2] = ip_idx;
5293 return err;
5294 }
5295
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)5296 static int inet6_valid_dump_ifaddr_req(const struct nlmsghdr *nlh,
5297 struct inet6_fill_args *fillargs,
5298 struct net **tgt_net, struct sock *sk,
5299 struct netlink_callback *cb)
5300 {
5301 struct netlink_ext_ack *extack = cb->extack;
5302 struct nlattr *tb[IFA_MAX+1];
5303 struct ifaddrmsg *ifm;
5304 int err, i;
5305
5306 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) {
5307 NL_SET_ERR_MSG_MOD(extack, "Invalid header for address dump request");
5308 return -EINVAL;
5309 }
5310
5311 ifm = nlmsg_data(nlh);
5312 if (ifm->ifa_prefixlen || ifm->ifa_flags || ifm->ifa_scope) {
5313 NL_SET_ERR_MSG_MOD(extack, "Invalid values in header for address dump request");
5314 return -EINVAL;
5315 }
5316
5317 fillargs->ifindex = ifm->ifa_index;
5318 if (fillargs->ifindex) {
5319 cb->answer_flags |= NLM_F_DUMP_FILTERED;
5320 fillargs->flags |= NLM_F_DUMP_FILTERED;
5321 }
5322
5323 err = nlmsg_parse_deprecated_strict(nlh, sizeof(*ifm), tb, IFA_MAX,
5324 ifa_ipv6_policy, extack);
5325 if (err < 0)
5326 return err;
5327
5328 for (i = 0; i <= IFA_MAX; ++i) {
5329 if (!tb[i])
5330 continue;
5331
5332 if (i == IFA_TARGET_NETNSID) {
5333 struct net *net;
5334
5335 fillargs->netnsid = nla_get_s32(tb[i]);
5336 net = rtnl_get_net_ns_capable(sk, fillargs->netnsid);
5337 if (IS_ERR(net)) {
5338 fillargs->netnsid = -1;
5339 NL_SET_ERR_MSG_MOD(extack, "Invalid target network namespace id");
5340 return PTR_ERR(net);
5341 }
5342 *tgt_net = net;
5343 } else {
5344 NL_SET_ERR_MSG_MOD(extack, "Unsupported attribute in dump request");
5345 return -EINVAL;
5346 }
5347 }
5348
5349 return 0;
5350 }
5351
inet6_dump_addr(struct sk_buff * skb,struct netlink_callback * cb,enum addr_type_t type)5352 static int inet6_dump_addr(struct sk_buff *skb, struct netlink_callback *cb,
5353 enum addr_type_t type)
5354 {
5355 const struct nlmsghdr *nlh = cb->nlh;
5356 struct inet6_fill_args fillargs = {
5357 .portid = NETLINK_CB(cb->skb).portid,
5358 .seq = cb->nlh->nlmsg_seq,
5359 .flags = NLM_F_MULTI,
5360 .netnsid = -1,
5361 .type = type,
5362 };
5363 struct net *tgt_net = sock_net(skb->sk);
5364 int idx, s_idx, s_ip_idx;
5365 int h, s_h;
5366 struct net_device *dev;
5367 struct inet6_dev *idev;
5368 struct hlist_head *head;
5369 int err = 0;
5370
5371 s_h = cb->args[0];
5372 s_idx = idx = cb->args[1];
5373 s_ip_idx = cb->args[2];
5374
5375 if (cb->strict_check) {
5376 err = inet6_valid_dump_ifaddr_req(nlh, &fillargs, &tgt_net,
5377 skb->sk, cb);
5378 if (err < 0)
5379 goto put_tgt_net;
5380
5381 err = 0;
5382 if (fillargs.ifindex) {
5383 dev = __dev_get_by_index(tgt_net, fillargs.ifindex);
5384 if (!dev) {
5385 err = -ENODEV;
5386 goto put_tgt_net;
5387 }
5388 idev = __in6_dev_get(dev);
5389 if (idev) {
5390 err = in6_dump_addrs(idev, skb, cb, s_ip_idx,
5391 &fillargs);
5392 if (err > 0)
5393 err = 0;
5394 }
5395 goto put_tgt_net;
5396 }
5397 }
5398
5399 rcu_read_lock();
5400 cb->seq = inet6_base_seq(tgt_net);
5401 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
5402 idx = 0;
5403 head = &tgt_net->dev_index_head[h];
5404 hlist_for_each_entry_rcu(dev, head, index_hlist) {
5405 if (idx < s_idx)
5406 goto cont;
5407 if (h > s_h || idx > s_idx)
5408 s_ip_idx = 0;
5409 idev = __in6_dev_get(dev);
5410 if (!idev)
5411 goto cont;
5412
5413 if (in6_dump_addrs(idev, skb, cb, s_ip_idx,
5414 &fillargs) < 0)
5415 goto done;
5416 cont:
5417 idx++;
5418 }
5419 }
5420 done:
5421 rcu_read_unlock();
5422 cb->args[0] = h;
5423 cb->args[1] = idx;
5424 put_tgt_net:
5425 if (fillargs.netnsid >= 0)
5426 put_net(tgt_net);
5427
5428 return skb->len ? : err;
5429 }
5430
inet6_dump_ifaddr(struct sk_buff * skb,struct netlink_callback * cb)5431 static int inet6_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
5432 {
5433 enum addr_type_t type = UNICAST_ADDR;
5434
5435 return inet6_dump_addr(skb, cb, type);
5436 }
5437
inet6_dump_ifmcaddr(struct sk_buff * skb,struct netlink_callback * cb)5438 static int inet6_dump_ifmcaddr(struct sk_buff *skb, struct netlink_callback *cb)
5439 {
5440 enum addr_type_t type = MULTICAST_ADDR;
5441
5442 return inet6_dump_addr(skb, cb, type);
5443 }
5444
5445
inet6_dump_ifacaddr(struct sk_buff * skb,struct netlink_callback * cb)5446 static int inet6_dump_ifacaddr(struct sk_buff *skb, struct netlink_callback *cb)
5447 {
5448 enum addr_type_t type = ANYCAST_ADDR;
5449
5450 return inet6_dump_addr(skb, cb, type);
5451 }
5452
inet6_rtm_valid_getaddr_req(struct sk_buff * skb,const struct nlmsghdr * nlh,struct nlattr ** tb,struct netlink_ext_ack * extack)5453 static int inet6_rtm_valid_getaddr_req(struct sk_buff *skb,
5454 const struct nlmsghdr *nlh,
5455 struct nlattr **tb,
5456 struct netlink_ext_ack *extack)
5457 {
5458 struct ifaddrmsg *ifm;
5459 int i, err;
5460
5461 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) {
5462 NL_SET_ERR_MSG_MOD(extack, "Invalid header for get address request");
5463 return -EINVAL;
5464 }
5465
5466 if (!netlink_strict_get_check(skb))
5467 return nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFA_MAX,
5468 ifa_ipv6_policy, extack);
5469
5470 ifm = nlmsg_data(nlh);
5471 if (ifm->ifa_prefixlen || ifm->ifa_flags || ifm->ifa_scope) {
5472 NL_SET_ERR_MSG_MOD(extack, "Invalid values in header for get address request");
5473 return -EINVAL;
5474 }
5475
5476 err = nlmsg_parse_deprecated_strict(nlh, sizeof(*ifm), tb, IFA_MAX,
5477 ifa_ipv6_policy, extack);
5478 if (err)
5479 return err;
5480
5481 for (i = 0; i <= IFA_MAX; i++) {
5482 if (!tb[i])
5483 continue;
5484
5485 switch (i) {
5486 case IFA_TARGET_NETNSID:
5487 case IFA_ADDRESS:
5488 case IFA_LOCAL:
5489 break;
5490 default:
5491 NL_SET_ERR_MSG_MOD(extack, "Unsupported attribute in get address request");
5492 return -EINVAL;
5493 }
5494 }
5495
5496 return 0;
5497 }
5498
inet6_rtm_getaddr(struct sk_buff * in_skb,struct nlmsghdr * nlh,struct netlink_ext_ack * extack)5499 static int inet6_rtm_getaddr(struct sk_buff *in_skb, struct nlmsghdr *nlh,
5500 struct netlink_ext_ack *extack)
5501 {
5502 struct net *tgt_net = sock_net(in_skb->sk);
5503 struct inet6_fill_args fillargs = {
5504 .portid = NETLINK_CB(in_skb).portid,
5505 .seq = nlh->nlmsg_seq,
5506 .event = RTM_NEWADDR,
5507 .flags = 0,
5508 .netnsid = -1,
5509 };
5510 struct ifaddrmsg *ifm;
5511 struct nlattr *tb[IFA_MAX+1];
5512 struct in6_addr *addr = NULL, *peer;
5513 struct net_device *dev = NULL;
5514 struct inet6_ifaddr *ifa;
5515 struct sk_buff *skb;
5516 int err;
5517
5518 err = inet6_rtm_valid_getaddr_req(in_skb, nlh, tb, extack);
5519 if (err < 0)
5520 return err;
5521
5522 if (tb[IFA_TARGET_NETNSID]) {
5523 fillargs.netnsid = nla_get_s32(tb[IFA_TARGET_NETNSID]);
5524
5525 tgt_net = rtnl_get_net_ns_capable(NETLINK_CB(in_skb).sk,
5526 fillargs.netnsid);
5527 if (IS_ERR(tgt_net))
5528 return PTR_ERR(tgt_net);
5529 }
5530
5531 addr = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer);
5532 if (!addr) {
5533 err = -EINVAL;
5534 goto errout;
5535 }
5536 ifm = nlmsg_data(nlh);
5537 if (ifm->ifa_index)
5538 dev = dev_get_by_index(tgt_net, ifm->ifa_index);
5539
5540 ifa = ipv6_get_ifaddr(tgt_net, addr, dev, 1);
5541 if (!ifa) {
5542 err = -EADDRNOTAVAIL;
5543 goto errout;
5544 }
5545
5546 skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_KERNEL);
5547 if (!skb) {
5548 err = -ENOBUFS;
5549 goto errout_ifa;
5550 }
5551
5552 err = inet6_fill_ifaddr(skb, ifa, &fillargs);
5553 if (err < 0) {
5554 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
5555 WARN_ON(err == -EMSGSIZE);
5556 kfree_skb(skb);
5557 goto errout_ifa;
5558 }
5559 err = rtnl_unicast(skb, tgt_net, NETLINK_CB(in_skb).portid);
5560 errout_ifa:
5561 in6_ifa_put(ifa);
5562 errout:
5563 dev_put(dev);
5564 if (fillargs.netnsid >= 0)
5565 put_net(tgt_net);
5566
5567 return err;
5568 }
5569
inet6_ifa_notify(int event,struct inet6_ifaddr * ifa)5570 static void inet6_ifa_notify(int event, struct inet6_ifaddr *ifa)
5571 {
5572 struct sk_buff *skb;
5573 struct net *net = dev_net(ifa->idev->dev);
5574 struct inet6_fill_args fillargs = {
5575 .portid = 0,
5576 .seq = 0,
5577 .event = event,
5578 .flags = 0,
5579 .netnsid = -1,
5580 };
5581 int err = -ENOBUFS;
5582
5583 skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_ATOMIC);
5584 if (!skb)
5585 goto errout;
5586
5587 err = inet6_fill_ifaddr(skb, ifa, &fillargs);
5588 if (err < 0) {
5589 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
5590 WARN_ON(err == -EMSGSIZE);
5591 kfree_skb(skb);
5592 goto errout;
5593 }
5594 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
5595 return;
5596 errout:
5597 if (err < 0)
5598 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
5599 }
5600
ipv6_store_devconf(struct ipv6_devconf * cnf,__s32 * array,int bytes)5601 static inline void ipv6_store_devconf(struct ipv6_devconf *cnf,
5602 __s32 *array, int bytes)
5603 {
5604 BUG_ON(bytes < (DEVCONF_MAX * 4));
5605
5606 memset(array, 0, bytes);
5607 array[DEVCONF_FORWARDING] = cnf->forwarding;
5608 array[DEVCONF_HOPLIMIT] = cnf->hop_limit;
5609 array[DEVCONF_MTU6] = cnf->mtu6;
5610 array[DEVCONF_ACCEPT_RA] = cnf->accept_ra;
5611 array[DEVCONF_ACCEPT_REDIRECTS] = cnf->accept_redirects;
5612 array[DEVCONF_AUTOCONF] = cnf->autoconf;
5613 array[DEVCONF_DAD_TRANSMITS] = cnf->dad_transmits;
5614 array[DEVCONF_RTR_SOLICITS] = cnf->rtr_solicits;
5615 array[DEVCONF_RTR_SOLICIT_INTERVAL] =
5616 jiffies_to_msecs(cnf->rtr_solicit_interval);
5617 array[DEVCONF_RTR_SOLICIT_MAX_INTERVAL] =
5618 jiffies_to_msecs(cnf->rtr_solicit_max_interval);
5619 array[DEVCONF_RTR_SOLICIT_DELAY] =
5620 jiffies_to_msecs(cnf->rtr_solicit_delay);
5621 array[DEVCONF_FORCE_MLD_VERSION] = cnf->force_mld_version;
5622 array[DEVCONF_MLDV1_UNSOLICITED_REPORT_INTERVAL] =
5623 jiffies_to_msecs(cnf->mldv1_unsolicited_report_interval);
5624 array[DEVCONF_MLDV2_UNSOLICITED_REPORT_INTERVAL] =
5625 jiffies_to_msecs(cnf->mldv2_unsolicited_report_interval);
5626 array[DEVCONF_USE_TEMPADDR] = cnf->use_tempaddr;
5627 array[DEVCONF_TEMP_VALID_LFT] = cnf->temp_valid_lft;
5628 array[DEVCONF_TEMP_PREFERED_LFT] = cnf->temp_prefered_lft;
5629 array[DEVCONF_REGEN_MAX_RETRY] = cnf->regen_max_retry;
5630 array[DEVCONF_MAX_DESYNC_FACTOR] = cnf->max_desync_factor;
5631 array[DEVCONF_MAX_ADDRESSES] = cnf->max_addresses;
5632 array[DEVCONF_ACCEPT_RA_DEFRTR] = cnf->accept_ra_defrtr;
5633 array[DEVCONF_RA_DEFRTR_METRIC] = cnf->ra_defrtr_metric;
5634 array[DEVCONF_ACCEPT_RA_MIN_HOP_LIMIT] = cnf->accept_ra_min_hop_limit;
5635 array[DEVCONF_ACCEPT_RA_PINFO] = cnf->accept_ra_pinfo;
5636 #ifdef CONFIG_IPV6_ROUTER_PREF
5637 array[DEVCONF_ACCEPT_RA_RTR_PREF] = cnf->accept_ra_rtr_pref;
5638 array[DEVCONF_RTR_PROBE_INTERVAL] =
5639 jiffies_to_msecs(cnf->rtr_probe_interval);
5640 #ifdef CONFIG_IPV6_ROUTE_INFO
5641 array[DEVCONF_ACCEPT_RA_RT_INFO_MIN_PLEN] = cnf->accept_ra_rt_info_min_plen;
5642 array[DEVCONF_ACCEPT_RA_RT_INFO_MAX_PLEN] = cnf->accept_ra_rt_info_max_plen;
5643 #endif
5644 #endif
5645 array[DEVCONF_PROXY_NDP] = cnf->proxy_ndp;
5646 array[DEVCONF_ACCEPT_SOURCE_ROUTE] = cnf->accept_source_route;
5647 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
5648 array[DEVCONF_OPTIMISTIC_DAD] = cnf->optimistic_dad;
5649 array[DEVCONF_USE_OPTIMISTIC] = cnf->use_optimistic;
5650 #endif
5651 #ifdef CONFIG_IPV6_MROUTE
5652 array[DEVCONF_MC_FORWARDING] = atomic_read(&cnf->mc_forwarding);
5653 #endif
5654 array[DEVCONF_DISABLE_IPV6] = cnf->disable_ipv6;
5655 array[DEVCONF_ACCEPT_DAD] = cnf->accept_dad;
5656 array[DEVCONF_FORCE_TLLAO] = cnf->force_tllao;
5657 array[DEVCONF_NDISC_NOTIFY] = cnf->ndisc_notify;
5658 array[DEVCONF_SUPPRESS_FRAG_NDISC] = cnf->suppress_frag_ndisc;
5659 array[DEVCONF_ACCEPT_RA_FROM_LOCAL] = cnf->accept_ra_from_local;
5660 array[DEVCONF_ACCEPT_RA_MTU] = cnf->accept_ra_mtu;
5661 array[DEVCONF_IGNORE_ROUTES_WITH_LINKDOWN] = cnf->ignore_routes_with_linkdown;
5662 /* we omit DEVCONF_STABLE_SECRET for now */
5663 array[DEVCONF_USE_OIF_ADDRS_ONLY] = cnf->use_oif_addrs_only;
5664 array[DEVCONF_DROP_UNICAST_IN_L2_MULTICAST] = cnf->drop_unicast_in_l2_multicast;
5665 array[DEVCONF_DROP_UNSOLICITED_NA] = cnf->drop_unsolicited_na;
5666 array[DEVCONF_KEEP_ADDR_ON_DOWN] = cnf->keep_addr_on_down;
5667 array[DEVCONF_SEG6_ENABLED] = cnf->seg6_enabled;
5668 #ifdef CONFIG_IPV6_SEG6_HMAC
5669 array[DEVCONF_SEG6_REQUIRE_HMAC] = cnf->seg6_require_hmac;
5670 #endif
5671 array[DEVCONF_ENHANCED_DAD] = cnf->enhanced_dad;
5672 array[DEVCONF_ADDR_GEN_MODE] = cnf->addr_gen_mode;
5673 array[DEVCONF_DISABLE_POLICY] = cnf->disable_policy;
5674 array[DEVCONF_NDISC_TCLASS] = cnf->ndisc_tclass;
5675 array[DEVCONF_RPL_SEG_ENABLED] = cnf->rpl_seg_enabled;
5676 array[DEVCONF_IOAM6_ENABLED] = cnf->ioam6_enabled;
5677 array[DEVCONF_IOAM6_ID] = cnf->ioam6_id;
5678 array[DEVCONF_IOAM6_ID_WIDE] = cnf->ioam6_id_wide;
5679 array[DEVCONF_NDISC_EVICT_NOCARRIER] = cnf->ndisc_evict_nocarrier;
5680 array[DEVCONF_ACCEPT_UNTRACKED_NA] = cnf->accept_untracked_na;
5681 array[DEVCONF_ACCEPT_RA_MIN_LFT] = cnf->accept_ra_min_lft;
5682 }
5683
inet6_ifla6_size(void)5684 static inline size_t inet6_ifla6_size(void)
5685 {
5686 return nla_total_size(4) /* IFLA_INET6_FLAGS */
5687 + nla_total_size(sizeof(struct ifla_cacheinfo))
5688 + nla_total_size(DEVCONF_MAX * 4) /* IFLA_INET6_CONF */
5689 + nla_total_size(IPSTATS_MIB_MAX * 8) /* IFLA_INET6_STATS */
5690 + nla_total_size(ICMP6_MIB_MAX * 8) /* IFLA_INET6_ICMP6STATS */
5691 + nla_total_size(sizeof(struct in6_addr)) /* IFLA_INET6_TOKEN */
5692 + nla_total_size(1) /* IFLA_INET6_ADDR_GEN_MODE */
5693 + nla_total_size(4) /* IFLA_INET6_RA_MTU */
5694 + 0;
5695 }
5696
inet6_if_nlmsg_size(void)5697 static inline size_t inet6_if_nlmsg_size(void)
5698 {
5699 return NLMSG_ALIGN(sizeof(struct ifinfomsg))
5700 + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
5701 + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
5702 + nla_total_size(4) /* IFLA_MTU */
5703 + nla_total_size(4) /* IFLA_LINK */
5704 + nla_total_size(1) /* IFLA_OPERSTATE */
5705 + nla_total_size(inet6_ifla6_size()); /* IFLA_PROTINFO */
5706 }
5707
__snmp6_fill_statsdev(u64 * stats,atomic_long_t * mib,int bytes)5708 static inline void __snmp6_fill_statsdev(u64 *stats, atomic_long_t *mib,
5709 int bytes)
5710 {
5711 int i;
5712 int pad = bytes - sizeof(u64) * ICMP6_MIB_MAX;
5713 BUG_ON(pad < 0);
5714
5715 /* Use put_unaligned() because stats may not be aligned for u64. */
5716 put_unaligned(ICMP6_MIB_MAX, &stats[0]);
5717 for (i = 1; i < ICMP6_MIB_MAX; i++)
5718 put_unaligned(atomic_long_read(&mib[i]), &stats[i]);
5719
5720 memset(&stats[ICMP6_MIB_MAX], 0, pad);
5721 }
5722
__snmp6_fill_stats64(u64 * stats,void __percpu * mib,int bytes,size_t syncpoff)5723 static inline void __snmp6_fill_stats64(u64 *stats, void __percpu *mib,
5724 int bytes, size_t syncpoff)
5725 {
5726 int i, c;
5727 u64 buff[IPSTATS_MIB_MAX];
5728 int pad = bytes - sizeof(u64) * IPSTATS_MIB_MAX;
5729
5730 BUG_ON(pad < 0);
5731
5732 memset(buff, 0, sizeof(buff));
5733 buff[0] = IPSTATS_MIB_MAX;
5734
5735 for_each_possible_cpu(c) {
5736 for (i = 1; i < IPSTATS_MIB_MAX; i++)
5737 buff[i] += snmp_get_cpu_field64(mib, c, i, syncpoff);
5738 }
5739
5740 memcpy(stats, buff, IPSTATS_MIB_MAX * sizeof(u64));
5741 memset(&stats[IPSTATS_MIB_MAX], 0, pad);
5742 }
5743
snmp6_fill_stats(u64 * stats,struct inet6_dev * idev,int attrtype,int bytes)5744 static void snmp6_fill_stats(u64 *stats, struct inet6_dev *idev, int attrtype,
5745 int bytes)
5746 {
5747 switch (attrtype) {
5748 case IFLA_INET6_STATS:
5749 __snmp6_fill_stats64(stats, idev->stats.ipv6, bytes,
5750 offsetof(struct ipstats_mib, syncp));
5751 break;
5752 case IFLA_INET6_ICMP6STATS:
5753 __snmp6_fill_statsdev(stats, idev->stats.icmpv6dev->mibs, bytes);
5754 break;
5755 }
5756 }
5757
inet6_fill_ifla6_attrs(struct sk_buff * skb,struct inet6_dev * idev,u32 ext_filter_mask)5758 static int inet6_fill_ifla6_attrs(struct sk_buff *skb, struct inet6_dev *idev,
5759 u32 ext_filter_mask)
5760 {
5761 struct nlattr *nla;
5762 struct ifla_cacheinfo ci;
5763
5764 if (nla_put_u32(skb, IFLA_INET6_FLAGS, idev->if_flags))
5765 goto nla_put_failure;
5766 ci.max_reasm_len = IPV6_MAXPLEN;
5767 ci.tstamp = cstamp_delta(idev->tstamp);
5768 ci.reachable_time = jiffies_to_msecs(idev->nd_parms->reachable_time);
5769 ci.retrans_time = jiffies_to_msecs(NEIGH_VAR(idev->nd_parms, RETRANS_TIME));
5770 if (nla_put(skb, IFLA_INET6_CACHEINFO, sizeof(ci), &ci))
5771 goto nla_put_failure;
5772 nla = nla_reserve(skb, IFLA_INET6_CONF, DEVCONF_MAX * sizeof(s32));
5773 if (!nla)
5774 goto nla_put_failure;
5775 ipv6_store_devconf(&idev->cnf, nla_data(nla), nla_len(nla));
5776
5777 /* XXX - MC not implemented */
5778
5779 if (ext_filter_mask & RTEXT_FILTER_SKIP_STATS)
5780 return 0;
5781
5782 nla = nla_reserve(skb, IFLA_INET6_STATS, IPSTATS_MIB_MAX * sizeof(u64));
5783 if (!nla)
5784 goto nla_put_failure;
5785 snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_STATS, nla_len(nla));
5786
5787 nla = nla_reserve(skb, IFLA_INET6_ICMP6STATS, ICMP6_MIB_MAX * sizeof(u64));
5788 if (!nla)
5789 goto nla_put_failure;
5790 snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_ICMP6STATS, nla_len(nla));
5791
5792 nla = nla_reserve(skb, IFLA_INET6_TOKEN, sizeof(struct in6_addr));
5793 if (!nla)
5794 goto nla_put_failure;
5795 read_lock_bh(&idev->lock);
5796 memcpy(nla_data(nla), idev->token.s6_addr, nla_len(nla));
5797 read_unlock_bh(&idev->lock);
5798
5799 if (nla_put_u8(skb, IFLA_INET6_ADDR_GEN_MODE, idev->cnf.addr_gen_mode))
5800 goto nla_put_failure;
5801
5802 if (idev->ra_mtu &&
5803 nla_put_u32(skb, IFLA_INET6_RA_MTU, idev->ra_mtu))
5804 goto nla_put_failure;
5805
5806 return 0;
5807
5808 nla_put_failure:
5809 return -EMSGSIZE;
5810 }
5811
inet6_get_link_af_size(const struct net_device * dev,u32 ext_filter_mask)5812 static size_t inet6_get_link_af_size(const struct net_device *dev,
5813 u32 ext_filter_mask)
5814 {
5815 if (!__in6_dev_get(dev))
5816 return 0;
5817
5818 return inet6_ifla6_size();
5819 }
5820
inet6_fill_link_af(struct sk_buff * skb,const struct net_device * dev,u32 ext_filter_mask)5821 static int inet6_fill_link_af(struct sk_buff *skb, const struct net_device *dev,
5822 u32 ext_filter_mask)
5823 {
5824 struct inet6_dev *idev = __in6_dev_get(dev);
5825
5826 if (!idev)
5827 return -ENODATA;
5828
5829 if (inet6_fill_ifla6_attrs(skb, idev, ext_filter_mask) < 0)
5830 return -EMSGSIZE;
5831
5832 return 0;
5833 }
5834
inet6_set_iftoken(struct inet6_dev * idev,struct in6_addr * token,struct netlink_ext_ack * extack)5835 static int inet6_set_iftoken(struct inet6_dev *idev, struct in6_addr *token,
5836 struct netlink_ext_ack *extack)
5837 {
5838 struct inet6_ifaddr *ifp;
5839 struct net_device *dev = idev->dev;
5840 bool clear_token, update_rs = false;
5841 struct in6_addr ll_addr;
5842
5843 ASSERT_RTNL();
5844
5845 if (!token)
5846 return -EINVAL;
5847
5848 if (dev->flags & IFF_LOOPBACK) {
5849 NL_SET_ERR_MSG_MOD(extack, "Device is loopback");
5850 return -EINVAL;
5851 }
5852
5853 if (dev->flags & IFF_NOARP) {
5854 NL_SET_ERR_MSG_MOD(extack,
5855 "Device does not do neighbour discovery");
5856 return -EINVAL;
5857 }
5858
5859 if (!ipv6_accept_ra(idev)) {
5860 NL_SET_ERR_MSG_MOD(extack,
5861 "Router advertisement is disabled on device");
5862 return -EINVAL;
5863 }
5864
5865 if (idev->cnf.rtr_solicits == 0) {
5866 NL_SET_ERR_MSG(extack,
5867 "Router solicitation is disabled on device");
5868 return -EINVAL;
5869 }
5870
5871 write_lock_bh(&idev->lock);
5872
5873 BUILD_BUG_ON(sizeof(token->s6_addr) != 16);
5874 memcpy(idev->token.s6_addr + 8, token->s6_addr + 8, 8);
5875
5876 write_unlock_bh(&idev->lock);
5877
5878 clear_token = ipv6_addr_any(token);
5879 if (clear_token)
5880 goto update_lft;
5881
5882 if (!idev->dead && (idev->if_flags & IF_READY) &&
5883 !ipv6_get_lladdr(dev, &ll_addr, IFA_F_TENTATIVE |
5884 IFA_F_OPTIMISTIC)) {
5885 /* If we're not ready, then normal ifup will take care
5886 * of this. Otherwise, we need to request our rs here.
5887 */
5888 ndisc_send_rs(dev, &ll_addr, &in6addr_linklocal_allrouters);
5889 update_rs = true;
5890 }
5891
5892 update_lft:
5893 write_lock_bh(&idev->lock);
5894
5895 if (update_rs) {
5896 idev->if_flags |= IF_RS_SENT;
5897 idev->rs_interval = rfc3315_s14_backoff_init(
5898 idev->cnf.rtr_solicit_interval);
5899 idev->rs_probes = 1;
5900 addrconf_mod_rs_timer(idev, idev->rs_interval);
5901 }
5902
5903 /* Well, that's kinda nasty ... */
5904 list_for_each_entry(ifp, &idev->addr_list, if_list) {
5905 spin_lock(&ifp->lock);
5906 if (ifp->tokenized) {
5907 ifp->valid_lft = 0;
5908 ifp->prefered_lft = 0;
5909 }
5910 spin_unlock(&ifp->lock);
5911 }
5912
5913 write_unlock_bh(&idev->lock);
5914 inet6_ifinfo_notify(RTM_NEWLINK, idev);
5915 addrconf_verify_rtnl(dev_net(dev));
5916 return 0;
5917 }
5918
5919 static const struct nla_policy inet6_af_policy[IFLA_INET6_MAX + 1] = {
5920 [IFLA_INET6_ADDR_GEN_MODE] = { .type = NLA_U8 },
5921 [IFLA_INET6_TOKEN] = { .len = sizeof(struct in6_addr) },
5922 [IFLA_INET6_RA_MTU] = { .type = NLA_REJECT,
5923 .reject_message =
5924 "IFLA_INET6_RA_MTU can not be set" },
5925 };
5926
check_addr_gen_mode(int mode)5927 static int check_addr_gen_mode(int mode)
5928 {
5929 if (mode != IN6_ADDR_GEN_MODE_EUI64 &&
5930 mode != IN6_ADDR_GEN_MODE_NONE &&
5931 mode != IN6_ADDR_GEN_MODE_STABLE_PRIVACY &&
5932 mode != IN6_ADDR_GEN_MODE_RANDOM)
5933 return -EINVAL;
5934 return 1;
5935 }
5936
check_stable_privacy(struct inet6_dev * idev,struct net * net,int mode)5937 static int check_stable_privacy(struct inet6_dev *idev, struct net *net,
5938 int mode)
5939 {
5940 if (mode == IN6_ADDR_GEN_MODE_STABLE_PRIVACY &&
5941 !idev->cnf.stable_secret.initialized &&
5942 !net->ipv6.devconf_dflt->stable_secret.initialized)
5943 return -EINVAL;
5944 return 1;
5945 }
5946
inet6_validate_link_af(const struct net_device * dev,const struct nlattr * nla,struct netlink_ext_ack * extack)5947 static int inet6_validate_link_af(const struct net_device *dev,
5948 const struct nlattr *nla,
5949 struct netlink_ext_ack *extack)
5950 {
5951 struct nlattr *tb[IFLA_INET6_MAX + 1];
5952 struct inet6_dev *idev = NULL;
5953 int err;
5954
5955 if (dev) {
5956 idev = __in6_dev_get(dev);
5957 if (!idev)
5958 return -EAFNOSUPPORT;
5959 }
5960
5961 err = nla_parse_nested_deprecated(tb, IFLA_INET6_MAX, nla,
5962 inet6_af_policy, extack);
5963 if (err)
5964 return err;
5965
5966 if (!tb[IFLA_INET6_TOKEN] && !tb[IFLA_INET6_ADDR_GEN_MODE])
5967 return -EINVAL;
5968
5969 if (tb[IFLA_INET6_ADDR_GEN_MODE]) {
5970 u8 mode = nla_get_u8(tb[IFLA_INET6_ADDR_GEN_MODE]);
5971
5972 if (check_addr_gen_mode(mode) < 0)
5973 return -EINVAL;
5974 if (dev && check_stable_privacy(idev, dev_net(dev), mode) < 0)
5975 return -EINVAL;
5976 }
5977
5978 return 0;
5979 }
5980
inet6_set_link_af(struct net_device * dev,const struct nlattr * nla,struct netlink_ext_ack * extack)5981 static int inet6_set_link_af(struct net_device *dev, const struct nlattr *nla,
5982 struct netlink_ext_ack *extack)
5983 {
5984 struct inet6_dev *idev = __in6_dev_get(dev);
5985 struct nlattr *tb[IFLA_INET6_MAX + 1];
5986 int err;
5987
5988 if (!idev)
5989 return -EAFNOSUPPORT;
5990
5991 if (nla_parse_nested_deprecated(tb, IFLA_INET6_MAX, nla, NULL, NULL) < 0)
5992 return -EINVAL;
5993
5994 if (tb[IFLA_INET6_TOKEN]) {
5995 err = inet6_set_iftoken(idev, nla_data(tb[IFLA_INET6_TOKEN]),
5996 extack);
5997 if (err)
5998 return err;
5999 }
6000
6001 if (tb[IFLA_INET6_ADDR_GEN_MODE]) {
6002 u8 mode = nla_get_u8(tb[IFLA_INET6_ADDR_GEN_MODE]);
6003
6004 idev->cnf.addr_gen_mode = mode;
6005 }
6006
6007 return 0;
6008 }
6009
inet6_fill_ifinfo(struct sk_buff * skb,struct inet6_dev * idev,u32 portid,u32 seq,int event,unsigned int flags)6010 static int inet6_fill_ifinfo(struct sk_buff *skb, struct inet6_dev *idev,
6011 u32 portid, u32 seq, int event, unsigned int flags)
6012 {
6013 struct net_device *dev = idev->dev;
6014 struct ifinfomsg *hdr;
6015 struct nlmsghdr *nlh;
6016 void *protoinfo;
6017
6018 nlh = nlmsg_put(skb, portid, seq, event, sizeof(*hdr), flags);
6019 if (!nlh)
6020 return -EMSGSIZE;
6021
6022 hdr = nlmsg_data(nlh);
6023 hdr->ifi_family = AF_INET6;
6024 hdr->__ifi_pad = 0;
6025 hdr->ifi_type = dev->type;
6026 hdr->ifi_index = dev->ifindex;
6027 hdr->ifi_flags = dev_get_flags(dev);
6028 hdr->ifi_change = 0;
6029
6030 if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
6031 (dev->addr_len &&
6032 nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
6033 nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
6034 (dev->ifindex != dev_get_iflink(dev) &&
6035 nla_put_u32(skb, IFLA_LINK, dev_get_iflink(dev))) ||
6036 nla_put_u8(skb, IFLA_OPERSTATE,
6037 netif_running(dev) ? READ_ONCE(dev->operstate) : IF_OPER_DOWN))
6038 goto nla_put_failure;
6039 protoinfo = nla_nest_start_noflag(skb, IFLA_PROTINFO);
6040 if (!protoinfo)
6041 goto nla_put_failure;
6042
6043 if (inet6_fill_ifla6_attrs(skb, idev, 0) < 0)
6044 goto nla_put_failure;
6045
6046 nla_nest_end(skb, protoinfo);
6047 nlmsg_end(skb, nlh);
6048 return 0;
6049
6050 nla_put_failure:
6051 nlmsg_cancel(skb, nlh);
6052 return -EMSGSIZE;
6053 }
6054
inet6_valid_dump_ifinfo(const struct nlmsghdr * nlh,struct netlink_ext_ack * extack)6055 static int inet6_valid_dump_ifinfo(const struct nlmsghdr *nlh,
6056 struct netlink_ext_ack *extack)
6057 {
6058 struct ifinfomsg *ifm;
6059
6060 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) {
6061 NL_SET_ERR_MSG_MOD(extack, "Invalid header for link dump request");
6062 return -EINVAL;
6063 }
6064
6065 if (nlmsg_attrlen(nlh, sizeof(*ifm))) {
6066 NL_SET_ERR_MSG_MOD(extack, "Invalid data after header");
6067 return -EINVAL;
6068 }
6069
6070 ifm = nlmsg_data(nlh);
6071 if (ifm->__ifi_pad || ifm->ifi_type || ifm->ifi_flags ||
6072 ifm->ifi_change || ifm->ifi_index) {
6073 NL_SET_ERR_MSG_MOD(extack, "Invalid values in header for dump request");
6074 return -EINVAL;
6075 }
6076
6077 return 0;
6078 }
6079
inet6_dump_ifinfo(struct sk_buff * skb,struct netlink_callback * cb)6080 static int inet6_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
6081 {
6082 struct net *net = sock_net(skb->sk);
6083 int h, s_h;
6084 int idx = 0, s_idx;
6085 struct net_device *dev;
6086 struct inet6_dev *idev;
6087 struct hlist_head *head;
6088
6089 /* only requests using strict checking can pass data to
6090 * influence the dump
6091 */
6092 if (cb->strict_check) {
6093 int err = inet6_valid_dump_ifinfo(cb->nlh, cb->extack);
6094
6095 if (err < 0)
6096 return err;
6097 }
6098
6099 s_h = cb->args[0];
6100 s_idx = cb->args[1];
6101
6102 rcu_read_lock();
6103 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
6104 idx = 0;
6105 head = &net->dev_index_head[h];
6106 hlist_for_each_entry_rcu(dev, head, index_hlist) {
6107 if (idx < s_idx)
6108 goto cont;
6109 idev = __in6_dev_get(dev);
6110 if (!idev)
6111 goto cont;
6112 if (inet6_fill_ifinfo(skb, idev,
6113 NETLINK_CB(cb->skb).portid,
6114 cb->nlh->nlmsg_seq,
6115 RTM_NEWLINK, NLM_F_MULTI) < 0)
6116 goto out;
6117 cont:
6118 idx++;
6119 }
6120 }
6121 out:
6122 rcu_read_unlock();
6123 cb->args[1] = idx;
6124 cb->args[0] = h;
6125
6126 return skb->len;
6127 }
6128
inet6_ifinfo_notify(int event,struct inet6_dev * idev)6129 void inet6_ifinfo_notify(int event, struct inet6_dev *idev)
6130 {
6131 struct sk_buff *skb;
6132 struct net *net = dev_net(idev->dev);
6133 int err = -ENOBUFS;
6134
6135 skb = nlmsg_new(inet6_if_nlmsg_size(), GFP_ATOMIC);
6136 if (!skb)
6137 goto errout;
6138
6139 err = inet6_fill_ifinfo(skb, idev, 0, 0, event, 0);
6140 if (err < 0) {
6141 /* -EMSGSIZE implies BUG in inet6_if_nlmsg_size() */
6142 WARN_ON(err == -EMSGSIZE);
6143 kfree_skb(skb);
6144 goto errout;
6145 }
6146 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFINFO, NULL, GFP_ATOMIC);
6147 return;
6148 errout:
6149 if (err < 0)
6150 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFINFO, err);
6151 }
6152
inet6_prefix_nlmsg_size(void)6153 static inline size_t inet6_prefix_nlmsg_size(void)
6154 {
6155 return NLMSG_ALIGN(sizeof(struct prefixmsg))
6156 + nla_total_size(sizeof(struct in6_addr))
6157 + nla_total_size(sizeof(struct prefix_cacheinfo));
6158 }
6159
inet6_fill_prefix(struct sk_buff * skb,struct inet6_dev * idev,struct prefix_info * pinfo,u32 portid,u32 seq,int event,unsigned int flags)6160 static int inet6_fill_prefix(struct sk_buff *skb, struct inet6_dev *idev,
6161 struct prefix_info *pinfo, u32 portid, u32 seq,
6162 int event, unsigned int flags)
6163 {
6164 struct prefixmsg *pmsg;
6165 struct nlmsghdr *nlh;
6166 struct prefix_cacheinfo ci;
6167
6168 nlh = nlmsg_put(skb, portid, seq, event, sizeof(*pmsg), flags);
6169 if (!nlh)
6170 return -EMSGSIZE;
6171
6172 pmsg = nlmsg_data(nlh);
6173 pmsg->prefix_family = AF_INET6;
6174 pmsg->prefix_pad1 = 0;
6175 pmsg->prefix_pad2 = 0;
6176 pmsg->prefix_ifindex = idev->dev->ifindex;
6177 pmsg->prefix_len = pinfo->prefix_len;
6178 pmsg->prefix_type = pinfo->type;
6179 pmsg->prefix_pad3 = 0;
6180 pmsg->prefix_flags = pinfo->flags;
6181
6182 if (nla_put(skb, PREFIX_ADDRESS, sizeof(pinfo->prefix), &pinfo->prefix))
6183 goto nla_put_failure;
6184 ci.preferred_time = ntohl(pinfo->prefered);
6185 ci.valid_time = ntohl(pinfo->valid);
6186 if (nla_put(skb, PREFIX_CACHEINFO, sizeof(ci), &ci))
6187 goto nla_put_failure;
6188 nlmsg_end(skb, nlh);
6189 return 0;
6190
6191 nla_put_failure:
6192 nlmsg_cancel(skb, nlh);
6193 return -EMSGSIZE;
6194 }
6195
inet6_prefix_notify(int event,struct inet6_dev * idev,struct prefix_info * pinfo)6196 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
6197 struct prefix_info *pinfo)
6198 {
6199 struct sk_buff *skb;
6200 struct net *net = dev_net(idev->dev);
6201 int err = -ENOBUFS;
6202
6203 skb = nlmsg_new(inet6_prefix_nlmsg_size(), GFP_ATOMIC);
6204 if (!skb)
6205 goto errout;
6206
6207 err = inet6_fill_prefix(skb, idev, pinfo, 0, 0, event, 0);
6208 if (err < 0) {
6209 /* -EMSGSIZE implies BUG in inet6_prefix_nlmsg_size() */
6210 WARN_ON(err == -EMSGSIZE);
6211 kfree_skb(skb);
6212 goto errout;
6213 }
6214 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_PREFIX, NULL, GFP_ATOMIC);
6215 return;
6216 errout:
6217 if (err < 0)
6218 rtnl_set_sk_err(net, RTNLGRP_IPV6_PREFIX, err);
6219 }
6220
__ipv6_ifa_notify(int event,struct inet6_ifaddr * ifp)6221 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
6222 {
6223 struct net *net = dev_net(ifp->idev->dev);
6224
6225 if (event)
6226 ASSERT_RTNL();
6227
6228 inet6_ifa_notify(event ? : RTM_NEWADDR, ifp);
6229
6230 switch (event) {
6231 case RTM_NEWADDR:
6232 /*
6233 * If the address was optimistic we inserted the route at the
6234 * start of our DAD process, so we don't need to do it again.
6235 * If the device was taken down in the middle of the DAD
6236 * cycle there is a race where we could get here without a
6237 * host route, so nothing to insert. That will be fixed when
6238 * the device is brought up.
6239 */
6240 if (ifp->rt && !rcu_access_pointer(ifp->rt->fib6_node)) {
6241 ip6_ins_rt(net, ifp->rt);
6242 } else if (!ifp->rt && (ifp->idev->dev->flags & IFF_UP)) {
6243 pr_warn("BUG: Address %pI6c on device %s is missing its host route.\n",
6244 &ifp->addr, ifp->idev->dev->name);
6245 }
6246
6247 if (ifp->idev->cnf.forwarding)
6248 addrconf_join_anycast(ifp);
6249 if (!ipv6_addr_any(&ifp->peer_addr))
6250 addrconf_prefix_route(&ifp->peer_addr, 128,
6251 ifp->rt_priority, ifp->idev->dev,
6252 0, 0, GFP_ATOMIC);
6253 break;
6254 case RTM_DELADDR:
6255 if (ifp->idev->cnf.forwarding)
6256 addrconf_leave_anycast(ifp);
6257 addrconf_leave_solict(ifp->idev, &ifp->addr);
6258 if (!ipv6_addr_any(&ifp->peer_addr)) {
6259 struct fib6_info *rt;
6260
6261 rt = addrconf_get_prefix_route(&ifp->peer_addr, 128,
6262 ifp->idev->dev, 0, 0,
6263 false);
6264 if (rt)
6265 ip6_del_rt(net, rt, false);
6266 }
6267 if (ifp->rt) {
6268 ip6_del_rt(net, ifp->rt, false);
6269 ifp->rt = NULL;
6270 }
6271 rt_genid_bump_ipv6(net);
6272 break;
6273 }
6274 atomic_inc(&net->ipv6.dev_addr_genid);
6275 }
6276
ipv6_ifa_notify(int event,struct inet6_ifaddr * ifp)6277 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
6278 {
6279 if (likely(ifp->idev->dead == 0))
6280 __ipv6_ifa_notify(event, ifp);
6281 }
6282
6283 #ifdef CONFIG_SYSCTL
6284
addrconf_sysctl_forward(struct ctl_table * ctl,int write,void * buffer,size_t * lenp,loff_t * ppos)6285 static int addrconf_sysctl_forward(struct ctl_table *ctl, int write,
6286 void *buffer, size_t *lenp, loff_t *ppos)
6287 {
6288 int *valp = ctl->data;
6289 int val = *valp;
6290 loff_t pos = *ppos;
6291 struct ctl_table lctl;
6292 int ret;
6293
6294 /*
6295 * ctl->data points to idev->cnf.forwarding, we should
6296 * not modify it until we get the rtnl lock.
6297 */
6298 lctl = *ctl;
6299 lctl.data = &val;
6300
6301 ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
6302
6303 if (write)
6304 ret = addrconf_fixup_forwarding(ctl, valp, val);
6305 if (ret)
6306 *ppos = pos;
6307 return ret;
6308 }
6309
addrconf_sysctl_mtu(struct ctl_table * ctl,int write,void * buffer,size_t * lenp,loff_t * ppos)6310 static int addrconf_sysctl_mtu(struct ctl_table *ctl, int write,
6311 void *buffer, size_t *lenp, loff_t *ppos)
6312 {
6313 struct inet6_dev *idev = ctl->extra1;
6314 int min_mtu = IPV6_MIN_MTU;
6315 struct ctl_table lctl;
6316
6317 lctl = *ctl;
6318 lctl.extra1 = &min_mtu;
6319 lctl.extra2 = idev ? &idev->dev->mtu : NULL;
6320
6321 return proc_dointvec_minmax(&lctl, write, buffer, lenp, ppos);
6322 }
6323
dev_disable_change(struct inet6_dev * idev)6324 static void dev_disable_change(struct inet6_dev *idev)
6325 {
6326 struct netdev_notifier_info info;
6327
6328 if (!idev || !idev->dev)
6329 return;
6330
6331 netdev_notifier_info_init(&info, idev->dev);
6332 if (idev->cnf.disable_ipv6)
6333 addrconf_notify(NULL, NETDEV_DOWN, &info);
6334 else
6335 addrconf_notify(NULL, NETDEV_UP, &info);
6336 }
6337
addrconf_disable_change(struct net * net,__s32 newf)6338 static void addrconf_disable_change(struct net *net, __s32 newf)
6339 {
6340 struct net_device *dev;
6341 struct inet6_dev *idev;
6342
6343 for_each_netdev(net, dev) {
6344 idev = __in6_dev_get(dev);
6345 if (idev) {
6346 int changed = (!idev->cnf.disable_ipv6) ^ (!newf);
6347
6348 WRITE_ONCE(idev->cnf.disable_ipv6, newf);
6349 if (changed)
6350 dev_disable_change(idev);
6351 }
6352 }
6353 }
6354
addrconf_disable_ipv6(struct ctl_table * table,int * p,int newf)6355 static int addrconf_disable_ipv6(struct ctl_table *table, int *p, int newf)
6356 {
6357 struct net *net;
6358 int old;
6359
6360 if (!rtnl_trylock())
6361 return restart_syscall();
6362
6363 net = (struct net *)table->extra2;
6364 old = *p;
6365 WRITE_ONCE(*p, newf);
6366
6367 if (p == &net->ipv6.devconf_dflt->disable_ipv6) {
6368 rtnl_unlock();
6369 return 0;
6370 }
6371
6372 if (p == &net->ipv6.devconf_all->disable_ipv6) {
6373 WRITE_ONCE(net->ipv6.devconf_dflt->disable_ipv6, newf);
6374 addrconf_disable_change(net, newf);
6375 } else if ((!newf) ^ (!old))
6376 dev_disable_change((struct inet6_dev *)table->extra1);
6377
6378 rtnl_unlock();
6379 return 0;
6380 }
6381
addrconf_sysctl_disable(struct ctl_table * ctl,int write,void * buffer,size_t * lenp,loff_t * ppos)6382 static int addrconf_sysctl_disable(struct ctl_table *ctl, int write,
6383 void *buffer, size_t *lenp, loff_t *ppos)
6384 {
6385 int *valp = ctl->data;
6386 int val = *valp;
6387 loff_t pos = *ppos;
6388 struct ctl_table lctl;
6389 int ret;
6390
6391 /*
6392 * ctl->data points to idev->cnf.disable_ipv6, we should
6393 * not modify it until we get the rtnl lock.
6394 */
6395 lctl = *ctl;
6396 lctl.data = &val;
6397
6398 ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
6399
6400 if (write)
6401 ret = addrconf_disable_ipv6(ctl, valp, val);
6402 if (ret)
6403 *ppos = pos;
6404 return ret;
6405 }
6406
addrconf_sysctl_proxy_ndp(struct ctl_table * ctl,int write,void * buffer,size_t * lenp,loff_t * ppos)6407 static int addrconf_sysctl_proxy_ndp(struct ctl_table *ctl, int write,
6408 void *buffer, size_t *lenp, loff_t *ppos)
6409 {
6410 int *valp = ctl->data;
6411 int ret;
6412 int old, new;
6413
6414 old = *valp;
6415 ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
6416 new = *valp;
6417
6418 if (write && old != new) {
6419 struct net *net = ctl->extra2;
6420
6421 if (!rtnl_trylock())
6422 return restart_syscall();
6423
6424 if (valp == &net->ipv6.devconf_dflt->proxy_ndp)
6425 inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
6426 NETCONFA_PROXY_NEIGH,
6427 NETCONFA_IFINDEX_DEFAULT,
6428 net->ipv6.devconf_dflt);
6429 else if (valp == &net->ipv6.devconf_all->proxy_ndp)
6430 inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
6431 NETCONFA_PROXY_NEIGH,
6432 NETCONFA_IFINDEX_ALL,
6433 net->ipv6.devconf_all);
6434 else {
6435 struct inet6_dev *idev = ctl->extra1;
6436
6437 inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
6438 NETCONFA_PROXY_NEIGH,
6439 idev->dev->ifindex,
6440 &idev->cnf);
6441 }
6442 rtnl_unlock();
6443 }
6444
6445 return ret;
6446 }
6447
addrconf_sysctl_addr_gen_mode(struct ctl_table * ctl,int write,void * buffer,size_t * lenp,loff_t * ppos)6448 static int addrconf_sysctl_addr_gen_mode(struct ctl_table *ctl, int write,
6449 void *buffer, size_t *lenp,
6450 loff_t *ppos)
6451 {
6452 int ret = 0;
6453 u32 new_val;
6454 struct inet6_dev *idev = (struct inet6_dev *)ctl->extra1;
6455 struct net *net = (struct net *)ctl->extra2;
6456 struct ctl_table tmp = {
6457 .data = &new_val,
6458 .maxlen = sizeof(new_val),
6459 .mode = ctl->mode,
6460 };
6461
6462 if (!rtnl_trylock())
6463 return restart_syscall();
6464
6465 new_val = *((u32 *)ctl->data);
6466
6467 ret = proc_douintvec(&tmp, write, buffer, lenp, ppos);
6468 if (ret != 0)
6469 goto out;
6470
6471 if (write) {
6472 if (check_addr_gen_mode(new_val) < 0) {
6473 ret = -EINVAL;
6474 goto out;
6475 }
6476
6477 if (idev) {
6478 if (check_stable_privacy(idev, net, new_val) < 0) {
6479 ret = -EINVAL;
6480 goto out;
6481 }
6482
6483 if (idev->cnf.addr_gen_mode != new_val) {
6484 idev->cnf.addr_gen_mode = new_val;
6485 addrconf_init_auto_addrs(idev->dev);
6486 }
6487 } else if (&net->ipv6.devconf_all->addr_gen_mode == ctl->data) {
6488 struct net_device *dev;
6489
6490 net->ipv6.devconf_dflt->addr_gen_mode = new_val;
6491 for_each_netdev(net, dev) {
6492 idev = __in6_dev_get(dev);
6493 if (idev &&
6494 idev->cnf.addr_gen_mode != new_val) {
6495 idev->cnf.addr_gen_mode = new_val;
6496 addrconf_init_auto_addrs(idev->dev);
6497 }
6498 }
6499 }
6500
6501 *((u32 *)ctl->data) = new_val;
6502 }
6503
6504 out:
6505 rtnl_unlock();
6506
6507 return ret;
6508 }
6509
addrconf_sysctl_stable_secret(struct ctl_table * ctl,int write,void * buffer,size_t * lenp,loff_t * ppos)6510 static int addrconf_sysctl_stable_secret(struct ctl_table *ctl, int write,
6511 void *buffer, size_t *lenp,
6512 loff_t *ppos)
6513 {
6514 int err;
6515 struct in6_addr addr;
6516 char str[IPV6_MAX_STRLEN];
6517 struct ctl_table lctl = *ctl;
6518 struct net *net = ctl->extra2;
6519 struct ipv6_stable_secret *secret = ctl->data;
6520
6521 if (&net->ipv6.devconf_all->stable_secret == ctl->data)
6522 return -EIO;
6523
6524 lctl.maxlen = IPV6_MAX_STRLEN;
6525 lctl.data = str;
6526
6527 if (!rtnl_trylock())
6528 return restart_syscall();
6529
6530 if (!write && !secret->initialized) {
6531 err = -EIO;
6532 goto out;
6533 }
6534
6535 err = snprintf(str, sizeof(str), "%pI6", &secret->secret);
6536 if (err >= sizeof(str)) {
6537 err = -EIO;
6538 goto out;
6539 }
6540
6541 err = proc_dostring(&lctl, write, buffer, lenp, ppos);
6542 if (err || !write)
6543 goto out;
6544
6545 if (in6_pton(str, -1, addr.in6_u.u6_addr8, -1, NULL) != 1) {
6546 err = -EIO;
6547 goto out;
6548 }
6549
6550 secret->initialized = true;
6551 secret->secret = addr;
6552
6553 if (&net->ipv6.devconf_dflt->stable_secret == ctl->data) {
6554 struct net_device *dev;
6555
6556 for_each_netdev(net, dev) {
6557 struct inet6_dev *idev = __in6_dev_get(dev);
6558
6559 if (idev) {
6560 idev->cnf.addr_gen_mode =
6561 IN6_ADDR_GEN_MODE_STABLE_PRIVACY;
6562 }
6563 }
6564 } else {
6565 struct inet6_dev *idev = ctl->extra1;
6566
6567 idev->cnf.addr_gen_mode = IN6_ADDR_GEN_MODE_STABLE_PRIVACY;
6568 }
6569
6570 out:
6571 rtnl_unlock();
6572
6573 return err;
6574 }
6575
6576 static
addrconf_sysctl_ignore_routes_with_linkdown(struct ctl_table * ctl,int write,void * buffer,size_t * lenp,loff_t * ppos)6577 int addrconf_sysctl_ignore_routes_with_linkdown(struct ctl_table *ctl,
6578 int write, void *buffer,
6579 size_t *lenp,
6580 loff_t *ppos)
6581 {
6582 int *valp = ctl->data;
6583 int val = *valp;
6584 loff_t pos = *ppos;
6585 struct ctl_table lctl;
6586 int ret;
6587
6588 /* ctl->data points to idev->cnf.ignore_routes_when_linkdown
6589 * we should not modify it until we get the rtnl lock.
6590 */
6591 lctl = *ctl;
6592 lctl.data = &val;
6593
6594 ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
6595
6596 if (write)
6597 ret = addrconf_fixup_linkdown(ctl, valp, val);
6598 if (ret)
6599 *ppos = pos;
6600 return ret;
6601 }
6602
6603 static
addrconf_set_nopolicy(struct rt6_info * rt,int action)6604 void addrconf_set_nopolicy(struct rt6_info *rt, int action)
6605 {
6606 if (rt) {
6607 if (action)
6608 rt->dst.flags |= DST_NOPOLICY;
6609 else
6610 rt->dst.flags &= ~DST_NOPOLICY;
6611 }
6612 }
6613
6614 static
addrconf_disable_policy_idev(struct inet6_dev * idev,int val)6615 void addrconf_disable_policy_idev(struct inet6_dev *idev, int val)
6616 {
6617 struct inet6_ifaddr *ifa;
6618
6619 read_lock_bh(&idev->lock);
6620 list_for_each_entry(ifa, &idev->addr_list, if_list) {
6621 spin_lock(&ifa->lock);
6622 if (ifa->rt) {
6623 /* host routes only use builtin fib6_nh */
6624 struct fib6_nh *nh = ifa->rt->fib6_nh;
6625 int cpu;
6626
6627 rcu_read_lock();
6628 ifa->rt->dst_nopolicy = val ? true : false;
6629 if (nh->rt6i_pcpu) {
6630 for_each_possible_cpu(cpu) {
6631 struct rt6_info **rtp;
6632
6633 rtp = per_cpu_ptr(nh->rt6i_pcpu, cpu);
6634 addrconf_set_nopolicy(*rtp, val);
6635 }
6636 }
6637 rcu_read_unlock();
6638 }
6639 spin_unlock(&ifa->lock);
6640 }
6641 read_unlock_bh(&idev->lock);
6642 }
6643
6644 static
addrconf_disable_policy(struct ctl_table * ctl,int * valp,int val)6645 int addrconf_disable_policy(struct ctl_table *ctl, int *valp, int val)
6646 {
6647 struct inet6_dev *idev;
6648 struct net *net;
6649
6650 if (!rtnl_trylock())
6651 return restart_syscall();
6652
6653 *valp = val;
6654
6655 net = (struct net *)ctl->extra2;
6656 if (valp == &net->ipv6.devconf_dflt->disable_policy) {
6657 rtnl_unlock();
6658 return 0;
6659 }
6660
6661 if (valp == &net->ipv6.devconf_all->disable_policy) {
6662 struct net_device *dev;
6663
6664 for_each_netdev(net, dev) {
6665 idev = __in6_dev_get(dev);
6666 if (idev)
6667 addrconf_disable_policy_idev(idev, val);
6668 }
6669 } else {
6670 idev = (struct inet6_dev *)ctl->extra1;
6671 addrconf_disable_policy_idev(idev, val);
6672 }
6673
6674 rtnl_unlock();
6675 return 0;
6676 }
6677
addrconf_sysctl_disable_policy(struct ctl_table * ctl,int write,void * buffer,size_t * lenp,loff_t * ppos)6678 static int addrconf_sysctl_disable_policy(struct ctl_table *ctl, int write,
6679 void *buffer, size_t *lenp, loff_t *ppos)
6680 {
6681 int *valp = ctl->data;
6682 int val = *valp;
6683 loff_t pos = *ppos;
6684 struct ctl_table lctl;
6685 int ret;
6686
6687 lctl = *ctl;
6688 lctl.data = &val;
6689 ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
6690
6691 if (write && (*valp != val))
6692 ret = addrconf_disable_policy(ctl, valp, val);
6693
6694 if (ret)
6695 *ppos = pos;
6696
6697 return ret;
6698 }
6699
6700 static int minus_one = -1;
6701 static const int two_five_five = 255;
6702 static u32 ioam6_if_id_max = U16_MAX;
6703
6704 static const struct ctl_table addrconf_sysctl[] = {
6705 {
6706 .procname = "forwarding",
6707 .data = &ipv6_devconf.forwarding,
6708 .maxlen = sizeof(int),
6709 .mode = 0644,
6710 .proc_handler = addrconf_sysctl_forward,
6711 },
6712 {
6713 .procname = "hop_limit",
6714 .data = &ipv6_devconf.hop_limit,
6715 .maxlen = sizeof(int),
6716 .mode = 0644,
6717 .proc_handler = proc_dointvec_minmax,
6718 .extra1 = (void *)SYSCTL_ONE,
6719 .extra2 = (void *)&two_five_five,
6720 },
6721 {
6722 .procname = "mtu",
6723 .data = &ipv6_devconf.mtu6,
6724 .maxlen = sizeof(int),
6725 .mode = 0644,
6726 .proc_handler = addrconf_sysctl_mtu,
6727 },
6728 {
6729 .procname = "accept_ra",
6730 .data = &ipv6_devconf.accept_ra,
6731 .maxlen = sizeof(int),
6732 .mode = 0644,
6733 .proc_handler = proc_dointvec,
6734 },
6735 {
6736 .procname = "accept_redirects",
6737 .data = &ipv6_devconf.accept_redirects,
6738 .maxlen = sizeof(int),
6739 .mode = 0644,
6740 .proc_handler = proc_dointvec,
6741 },
6742 {
6743 .procname = "autoconf",
6744 .data = &ipv6_devconf.autoconf,
6745 .maxlen = sizeof(int),
6746 .mode = 0644,
6747 .proc_handler = proc_dointvec,
6748 },
6749 {
6750 .procname = "dad_transmits",
6751 .data = &ipv6_devconf.dad_transmits,
6752 .maxlen = sizeof(int),
6753 .mode = 0644,
6754 .proc_handler = proc_dointvec,
6755 },
6756 {
6757 .procname = "router_solicitations",
6758 .data = &ipv6_devconf.rtr_solicits,
6759 .maxlen = sizeof(int),
6760 .mode = 0644,
6761 .proc_handler = proc_dointvec_minmax,
6762 .extra1 = &minus_one,
6763 },
6764 {
6765 .procname = "router_solicitation_interval",
6766 .data = &ipv6_devconf.rtr_solicit_interval,
6767 .maxlen = sizeof(int),
6768 .mode = 0644,
6769 .proc_handler = proc_dointvec_jiffies,
6770 },
6771 {
6772 .procname = "router_solicitation_max_interval",
6773 .data = &ipv6_devconf.rtr_solicit_max_interval,
6774 .maxlen = sizeof(int),
6775 .mode = 0644,
6776 .proc_handler = proc_dointvec_jiffies,
6777 },
6778 {
6779 .procname = "router_solicitation_delay",
6780 .data = &ipv6_devconf.rtr_solicit_delay,
6781 .maxlen = sizeof(int),
6782 .mode = 0644,
6783 .proc_handler = proc_dointvec_jiffies,
6784 },
6785 {
6786 .procname = "force_mld_version",
6787 .data = &ipv6_devconf.force_mld_version,
6788 .maxlen = sizeof(int),
6789 .mode = 0644,
6790 .proc_handler = proc_dointvec,
6791 },
6792 {
6793 .procname = "mldv1_unsolicited_report_interval",
6794 .data =
6795 &ipv6_devconf.mldv1_unsolicited_report_interval,
6796 .maxlen = sizeof(int),
6797 .mode = 0644,
6798 .proc_handler = proc_dointvec_ms_jiffies,
6799 },
6800 {
6801 .procname = "mldv2_unsolicited_report_interval",
6802 .data =
6803 &ipv6_devconf.mldv2_unsolicited_report_interval,
6804 .maxlen = sizeof(int),
6805 .mode = 0644,
6806 .proc_handler = proc_dointvec_ms_jiffies,
6807 },
6808 {
6809 .procname = "use_tempaddr",
6810 .data = &ipv6_devconf.use_tempaddr,
6811 .maxlen = sizeof(int),
6812 .mode = 0644,
6813 .proc_handler = proc_dointvec,
6814 },
6815 {
6816 .procname = "temp_valid_lft",
6817 .data = &ipv6_devconf.temp_valid_lft,
6818 .maxlen = sizeof(int),
6819 .mode = 0644,
6820 .proc_handler = proc_dointvec,
6821 },
6822 {
6823 .procname = "temp_prefered_lft",
6824 .data = &ipv6_devconf.temp_prefered_lft,
6825 .maxlen = sizeof(int),
6826 .mode = 0644,
6827 .proc_handler = proc_dointvec,
6828 },
6829 {
6830 .procname = "regen_max_retry",
6831 .data = &ipv6_devconf.regen_max_retry,
6832 .maxlen = sizeof(int),
6833 .mode = 0644,
6834 .proc_handler = proc_dointvec,
6835 },
6836 {
6837 .procname = "max_desync_factor",
6838 .data = &ipv6_devconf.max_desync_factor,
6839 .maxlen = sizeof(int),
6840 .mode = 0644,
6841 .proc_handler = proc_dointvec,
6842 },
6843 {
6844 .procname = "max_addresses",
6845 .data = &ipv6_devconf.max_addresses,
6846 .maxlen = sizeof(int),
6847 .mode = 0644,
6848 .proc_handler = proc_dointvec,
6849 },
6850 {
6851 .procname = "accept_ra_defrtr",
6852 .data = &ipv6_devconf.accept_ra_defrtr,
6853 .maxlen = sizeof(int),
6854 .mode = 0644,
6855 .proc_handler = proc_dointvec,
6856 },
6857 {
6858 .procname = "ra_defrtr_metric",
6859 .data = &ipv6_devconf.ra_defrtr_metric,
6860 .maxlen = sizeof(u32),
6861 .mode = 0644,
6862 .proc_handler = proc_douintvec_minmax,
6863 .extra1 = (void *)SYSCTL_ONE,
6864 },
6865 {
6866 .procname = "accept_ra_min_hop_limit",
6867 .data = &ipv6_devconf.accept_ra_min_hop_limit,
6868 .maxlen = sizeof(int),
6869 .mode = 0644,
6870 .proc_handler = proc_dointvec,
6871 },
6872 {
6873 .procname = "accept_ra_min_lft",
6874 .data = &ipv6_devconf.accept_ra_min_lft,
6875 .maxlen = sizeof(int),
6876 .mode = 0644,
6877 .proc_handler = proc_dointvec,
6878 },
6879 {
6880 .procname = "accept_ra_pinfo",
6881 .data = &ipv6_devconf.accept_ra_pinfo,
6882 .maxlen = sizeof(int),
6883 .mode = 0644,
6884 .proc_handler = proc_dointvec,
6885 },
6886 #ifdef CONFIG_IPV6_ROUTER_PREF
6887 {
6888 .procname = "accept_ra_rtr_pref",
6889 .data = &ipv6_devconf.accept_ra_rtr_pref,
6890 .maxlen = sizeof(int),
6891 .mode = 0644,
6892 .proc_handler = proc_dointvec,
6893 },
6894 {
6895 .procname = "router_probe_interval",
6896 .data = &ipv6_devconf.rtr_probe_interval,
6897 .maxlen = sizeof(int),
6898 .mode = 0644,
6899 .proc_handler = proc_dointvec_jiffies,
6900 },
6901 #ifdef CONFIG_IPV6_ROUTE_INFO
6902 {
6903 .procname = "accept_ra_rt_info_min_plen",
6904 .data = &ipv6_devconf.accept_ra_rt_info_min_plen,
6905 .maxlen = sizeof(int),
6906 .mode = 0644,
6907 .proc_handler = proc_dointvec,
6908 },
6909 {
6910 .procname = "accept_ra_rt_info_max_plen",
6911 .data = &ipv6_devconf.accept_ra_rt_info_max_plen,
6912 .maxlen = sizeof(int),
6913 .mode = 0644,
6914 .proc_handler = proc_dointvec,
6915 },
6916 #endif
6917 #endif
6918 {
6919 .procname = "accept_ra_rt_table",
6920 .data = &ipv6_devconf.accept_ra_rt_table,
6921 .maxlen = sizeof(int),
6922 .mode = 0644,
6923 .proc_handler = proc_dointvec,
6924 },
6925 {
6926 .procname = "proxy_ndp",
6927 .data = &ipv6_devconf.proxy_ndp,
6928 .maxlen = sizeof(int),
6929 .mode = 0644,
6930 .proc_handler = addrconf_sysctl_proxy_ndp,
6931 },
6932 {
6933 .procname = "accept_source_route",
6934 .data = &ipv6_devconf.accept_source_route,
6935 .maxlen = sizeof(int),
6936 .mode = 0644,
6937 .proc_handler = proc_dointvec,
6938 },
6939 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
6940 {
6941 .procname = "optimistic_dad",
6942 .data = &ipv6_devconf.optimistic_dad,
6943 .maxlen = sizeof(int),
6944 .mode = 0644,
6945 .proc_handler = proc_dointvec,
6946 },
6947 {
6948 .procname = "use_optimistic",
6949 .data = &ipv6_devconf.use_optimistic,
6950 .maxlen = sizeof(int),
6951 .mode = 0644,
6952 .proc_handler = proc_dointvec,
6953 },
6954 #endif
6955 #ifdef CONFIG_IPV6_MROUTE
6956 {
6957 .procname = "mc_forwarding",
6958 .data = &ipv6_devconf.mc_forwarding,
6959 .maxlen = sizeof(int),
6960 .mode = 0444,
6961 .proc_handler = proc_dointvec,
6962 },
6963 #endif
6964 {
6965 .procname = "disable_ipv6",
6966 .data = &ipv6_devconf.disable_ipv6,
6967 .maxlen = sizeof(int),
6968 .mode = 0644,
6969 .proc_handler = addrconf_sysctl_disable,
6970 },
6971 {
6972 .procname = "accept_dad",
6973 .data = &ipv6_devconf.accept_dad,
6974 .maxlen = sizeof(int),
6975 .mode = 0644,
6976 .proc_handler = proc_dointvec,
6977 },
6978 {
6979 .procname = "force_tllao",
6980 .data = &ipv6_devconf.force_tllao,
6981 .maxlen = sizeof(int),
6982 .mode = 0644,
6983 .proc_handler = proc_dointvec
6984 },
6985 {
6986 .procname = "ndisc_notify",
6987 .data = &ipv6_devconf.ndisc_notify,
6988 .maxlen = sizeof(int),
6989 .mode = 0644,
6990 .proc_handler = proc_dointvec
6991 },
6992 {
6993 .procname = "suppress_frag_ndisc",
6994 .data = &ipv6_devconf.suppress_frag_ndisc,
6995 .maxlen = sizeof(int),
6996 .mode = 0644,
6997 .proc_handler = proc_dointvec
6998 },
6999 {
7000 .procname = "accept_ra_from_local",
7001 .data = &ipv6_devconf.accept_ra_from_local,
7002 .maxlen = sizeof(int),
7003 .mode = 0644,
7004 .proc_handler = proc_dointvec,
7005 },
7006 {
7007 .procname = "accept_ra_mtu",
7008 .data = &ipv6_devconf.accept_ra_mtu,
7009 .maxlen = sizeof(int),
7010 .mode = 0644,
7011 .proc_handler = proc_dointvec,
7012 },
7013 {
7014 .procname = "stable_secret",
7015 .data = &ipv6_devconf.stable_secret,
7016 .maxlen = IPV6_MAX_STRLEN,
7017 .mode = 0600,
7018 .proc_handler = addrconf_sysctl_stable_secret,
7019 },
7020 {
7021 .procname = "use_oif_addrs_only",
7022 .data = &ipv6_devconf.use_oif_addrs_only,
7023 .maxlen = sizeof(int),
7024 .mode = 0644,
7025 .proc_handler = proc_dointvec,
7026 },
7027 {
7028 .procname = "ignore_routes_with_linkdown",
7029 .data = &ipv6_devconf.ignore_routes_with_linkdown,
7030 .maxlen = sizeof(int),
7031 .mode = 0644,
7032 .proc_handler = addrconf_sysctl_ignore_routes_with_linkdown,
7033 },
7034 {
7035 .procname = "drop_unicast_in_l2_multicast",
7036 .data = &ipv6_devconf.drop_unicast_in_l2_multicast,
7037 .maxlen = sizeof(int),
7038 .mode = 0644,
7039 .proc_handler = proc_dointvec,
7040 },
7041 {
7042 .procname = "drop_unsolicited_na",
7043 .data = &ipv6_devconf.drop_unsolicited_na,
7044 .maxlen = sizeof(int),
7045 .mode = 0644,
7046 .proc_handler = proc_dointvec,
7047 },
7048 {
7049 .procname = "keep_addr_on_down",
7050 .data = &ipv6_devconf.keep_addr_on_down,
7051 .maxlen = sizeof(int),
7052 .mode = 0644,
7053 .proc_handler = proc_dointvec,
7054
7055 },
7056 {
7057 .procname = "seg6_enabled",
7058 .data = &ipv6_devconf.seg6_enabled,
7059 .maxlen = sizeof(int),
7060 .mode = 0644,
7061 .proc_handler = proc_dointvec,
7062 },
7063 #ifdef CONFIG_IPV6_SEG6_HMAC
7064 {
7065 .procname = "seg6_require_hmac",
7066 .data = &ipv6_devconf.seg6_require_hmac,
7067 .maxlen = sizeof(int),
7068 .mode = 0644,
7069 .proc_handler = proc_dointvec,
7070 },
7071 #endif
7072 {
7073 .procname = "enhanced_dad",
7074 .data = &ipv6_devconf.enhanced_dad,
7075 .maxlen = sizeof(int),
7076 .mode = 0644,
7077 .proc_handler = proc_dointvec,
7078 },
7079 {
7080 .procname = "addr_gen_mode",
7081 .data = &ipv6_devconf.addr_gen_mode,
7082 .maxlen = sizeof(int),
7083 .mode = 0644,
7084 .proc_handler = addrconf_sysctl_addr_gen_mode,
7085 },
7086 {
7087 .procname = "disable_policy",
7088 .data = &ipv6_devconf.disable_policy,
7089 .maxlen = sizeof(int),
7090 .mode = 0644,
7091 .proc_handler = addrconf_sysctl_disable_policy,
7092 },
7093 {
7094 .procname = "ndisc_tclass",
7095 .data = &ipv6_devconf.ndisc_tclass,
7096 .maxlen = sizeof(int),
7097 .mode = 0644,
7098 .proc_handler = proc_dointvec_minmax,
7099 .extra1 = (void *)SYSCTL_ZERO,
7100 .extra2 = (void *)&two_five_five,
7101 },
7102 {
7103 .procname = "rpl_seg_enabled",
7104 .data = &ipv6_devconf.rpl_seg_enabled,
7105 .maxlen = sizeof(int),
7106 .mode = 0644,
7107 .proc_handler = proc_dointvec,
7108 },
7109 {
7110 .procname = "ioam6_enabled",
7111 .data = &ipv6_devconf.ioam6_enabled,
7112 .maxlen = sizeof(u8),
7113 .mode = 0644,
7114 .proc_handler = proc_dou8vec_minmax,
7115 .extra1 = (void *)SYSCTL_ZERO,
7116 .extra2 = (void *)SYSCTL_ONE,
7117 },
7118 {
7119 .procname = "ioam6_id",
7120 .data = &ipv6_devconf.ioam6_id,
7121 .maxlen = sizeof(u32),
7122 .mode = 0644,
7123 .proc_handler = proc_douintvec_minmax,
7124 .extra1 = (void *)SYSCTL_ZERO,
7125 .extra2 = (void *)&ioam6_if_id_max,
7126 },
7127 {
7128 .procname = "ioam6_id_wide",
7129 .data = &ipv6_devconf.ioam6_id_wide,
7130 .maxlen = sizeof(u32),
7131 .mode = 0644,
7132 .proc_handler = proc_douintvec,
7133 },
7134 {
7135 .procname = "ndisc_evict_nocarrier",
7136 .data = &ipv6_devconf.ndisc_evict_nocarrier,
7137 .maxlen = sizeof(u8),
7138 .mode = 0644,
7139 .proc_handler = proc_dou8vec_minmax,
7140 .extra1 = (void *)SYSCTL_ZERO,
7141 .extra2 = (void *)SYSCTL_ONE,
7142 },
7143 {
7144 .procname = "accept_untracked_na",
7145 .data = &ipv6_devconf.accept_untracked_na,
7146 .maxlen = sizeof(int),
7147 .mode = 0644,
7148 .proc_handler = proc_dointvec_minmax,
7149 .extra1 = SYSCTL_ZERO,
7150 .extra2 = SYSCTL_TWO,
7151 },
7152 {
7153 /* sentinel */
7154 }
7155 };
7156
__addrconf_sysctl_register(struct net * net,char * dev_name,struct inet6_dev * idev,struct ipv6_devconf * p)7157 static int __addrconf_sysctl_register(struct net *net, char *dev_name,
7158 struct inet6_dev *idev, struct ipv6_devconf *p)
7159 {
7160 int i, ifindex;
7161 struct ctl_table *table;
7162 char path[sizeof("net/ipv6/conf/") + IFNAMSIZ];
7163
7164 table = kmemdup(addrconf_sysctl, sizeof(addrconf_sysctl), GFP_KERNEL_ACCOUNT);
7165 if (!table)
7166 goto out;
7167
7168 for (i = 0; table[i].data; i++) {
7169 table[i].data += (char *)p - (char *)&ipv6_devconf;
7170 /* If one of these is already set, then it is not safe to
7171 * overwrite either of them: this makes proc_dointvec_minmax
7172 * usable.
7173 */
7174 if (!table[i].extra1 && !table[i].extra2) {
7175 table[i].extra1 = idev; /* embedded; no ref */
7176 table[i].extra2 = net;
7177 }
7178 }
7179
7180 snprintf(path, sizeof(path), "net/ipv6/conf/%s", dev_name);
7181
7182 p->sysctl_header = register_net_sysctl_sz(net, path, table,
7183 ARRAY_SIZE(addrconf_sysctl));
7184 if (!p->sysctl_header)
7185 goto free;
7186
7187 if (!strcmp(dev_name, "all"))
7188 ifindex = NETCONFA_IFINDEX_ALL;
7189 else if (!strcmp(dev_name, "default"))
7190 ifindex = NETCONFA_IFINDEX_DEFAULT;
7191 else
7192 ifindex = idev->dev->ifindex;
7193 inet6_netconf_notify_devconf(net, RTM_NEWNETCONF, NETCONFA_ALL,
7194 ifindex, p);
7195 return 0;
7196
7197 free:
7198 kfree(table);
7199 out:
7200 return -ENOBUFS;
7201 }
7202
__addrconf_sysctl_unregister(struct net * net,struct ipv6_devconf * p,int ifindex)7203 static void __addrconf_sysctl_unregister(struct net *net,
7204 struct ipv6_devconf *p, int ifindex)
7205 {
7206 struct ctl_table *table;
7207
7208 if (!p->sysctl_header)
7209 return;
7210
7211 table = p->sysctl_header->ctl_table_arg;
7212 unregister_net_sysctl_table(p->sysctl_header);
7213 p->sysctl_header = NULL;
7214 kfree(table);
7215
7216 inet6_netconf_notify_devconf(net, RTM_DELNETCONF, 0, ifindex, NULL);
7217 }
7218
addrconf_sysctl_register(struct inet6_dev * idev)7219 static int addrconf_sysctl_register(struct inet6_dev *idev)
7220 {
7221 int err;
7222
7223 if (!sysctl_dev_name_is_allowed(idev->dev->name))
7224 return -EINVAL;
7225
7226 err = neigh_sysctl_register(idev->dev, idev->nd_parms,
7227 &ndisc_ifinfo_sysctl_change);
7228 if (err)
7229 return err;
7230 err = __addrconf_sysctl_register(dev_net(idev->dev), idev->dev->name,
7231 idev, &idev->cnf);
7232 if (err)
7233 neigh_sysctl_unregister(idev->nd_parms);
7234
7235 return err;
7236 }
7237
addrconf_sysctl_unregister(struct inet6_dev * idev)7238 static void addrconf_sysctl_unregister(struct inet6_dev *idev)
7239 {
7240 __addrconf_sysctl_unregister(dev_net(idev->dev), &idev->cnf,
7241 idev->dev->ifindex);
7242 neigh_sysctl_unregister(idev->nd_parms);
7243 }
7244
7245
7246 #endif
7247
addrconf_init_net(struct net * net)7248 static int __net_init addrconf_init_net(struct net *net)
7249 {
7250 int err = -ENOMEM;
7251 struct ipv6_devconf *all, *dflt;
7252
7253 spin_lock_init(&net->ipv6.addrconf_hash_lock);
7254 INIT_DEFERRABLE_WORK(&net->ipv6.addr_chk_work, addrconf_verify_work);
7255 net->ipv6.inet6_addr_lst = kcalloc(IN6_ADDR_HSIZE,
7256 sizeof(struct hlist_head),
7257 GFP_KERNEL);
7258 if (!net->ipv6.inet6_addr_lst)
7259 goto err_alloc_addr;
7260
7261 all = kmemdup(&ipv6_devconf, sizeof(ipv6_devconf), GFP_KERNEL);
7262 if (!all)
7263 goto err_alloc_all;
7264
7265 dflt = kmemdup(&ipv6_devconf_dflt, sizeof(ipv6_devconf_dflt), GFP_KERNEL);
7266 if (!dflt)
7267 goto err_alloc_dflt;
7268
7269 if (!net_eq(net, &init_net)) {
7270 switch (net_inherit_devconf()) {
7271 case 1: /* copy from init_net */
7272 memcpy(all, init_net.ipv6.devconf_all,
7273 sizeof(ipv6_devconf));
7274 memcpy(dflt, init_net.ipv6.devconf_dflt,
7275 sizeof(ipv6_devconf_dflt));
7276 break;
7277 case 3: /* copy from the current netns */
7278 memcpy(all, current->nsproxy->net_ns->ipv6.devconf_all,
7279 sizeof(ipv6_devconf));
7280 memcpy(dflt,
7281 current->nsproxy->net_ns->ipv6.devconf_dflt,
7282 sizeof(ipv6_devconf_dflt));
7283 break;
7284 case 0:
7285 case 2:
7286 /* use compiled values */
7287 break;
7288 }
7289 }
7290
7291 /* these will be inherited by all namespaces */
7292 dflt->autoconf = ipv6_defaults.autoconf;
7293 dflt->disable_ipv6 = ipv6_defaults.disable_ipv6;
7294
7295 dflt->stable_secret.initialized = false;
7296 all->stable_secret.initialized = false;
7297
7298 net->ipv6.devconf_all = all;
7299 net->ipv6.devconf_dflt = dflt;
7300
7301 #ifdef CONFIG_SYSCTL
7302 err = __addrconf_sysctl_register(net, "all", NULL, all);
7303 if (err < 0)
7304 goto err_reg_all;
7305
7306 err = __addrconf_sysctl_register(net, "default", NULL, dflt);
7307 if (err < 0)
7308 goto err_reg_dflt;
7309 #endif
7310 return 0;
7311
7312 #ifdef CONFIG_SYSCTL
7313 err_reg_dflt:
7314 __addrconf_sysctl_unregister(net, all, NETCONFA_IFINDEX_ALL);
7315 err_reg_all:
7316 kfree(dflt);
7317 net->ipv6.devconf_dflt = NULL;
7318 #endif
7319 err_alloc_dflt:
7320 kfree(all);
7321 net->ipv6.devconf_all = NULL;
7322 err_alloc_all:
7323 kfree(net->ipv6.inet6_addr_lst);
7324 err_alloc_addr:
7325 return err;
7326 }
7327
addrconf_exit_net(struct net * net)7328 static void __net_exit addrconf_exit_net(struct net *net)
7329 {
7330 int i;
7331
7332 #ifdef CONFIG_SYSCTL
7333 __addrconf_sysctl_unregister(net, net->ipv6.devconf_dflt,
7334 NETCONFA_IFINDEX_DEFAULT);
7335 __addrconf_sysctl_unregister(net, net->ipv6.devconf_all,
7336 NETCONFA_IFINDEX_ALL);
7337 #endif
7338 kfree(net->ipv6.devconf_dflt);
7339 net->ipv6.devconf_dflt = NULL;
7340 kfree(net->ipv6.devconf_all);
7341 net->ipv6.devconf_all = NULL;
7342
7343 cancel_delayed_work_sync(&net->ipv6.addr_chk_work);
7344 /*
7345 * Check hash table, then free it.
7346 */
7347 for (i = 0; i < IN6_ADDR_HSIZE; i++)
7348 WARN_ON_ONCE(!hlist_empty(&net->ipv6.inet6_addr_lst[i]));
7349
7350 kfree(net->ipv6.inet6_addr_lst);
7351 net->ipv6.inet6_addr_lst = NULL;
7352 }
7353
7354 static struct pernet_operations addrconf_ops = {
7355 .init = addrconf_init_net,
7356 .exit = addrconf_exit_net,
7357 };
7358
7359 static struct rtnl_af_ops inet6_ops __read_mostly = {
7360 .family = AF_INET6,
7361 .fill_link_af = inet6_fill_link_af,
7362 .get_link_af_size = inet6_get_link_af_size,
7363 .validate_link_af = inet6_validate_link_af,
7364 .set_link_af = inet6_set_link_af,
7365 };
7366
7367 /*
7368 * Init / cleanup code
7369 */
7370
addrconf_init(void)7371 int __init addrconf_init(void)
7372 {
7373 struct inet6_dev *idev;
7374 int err;
7375
7376 err = ipv6_addr_label_init();
7377 if (err < 0) {
7378 pr_crit("%s: cannot initialize default policy table: %d\n",
7379 __func__, err);
7380 goto out;
7381 }
7382
7383 err = register_pernet_subsys(&addrconf_ops);
7384 if (err < 0)
7385 goto out_addrlabel;
7386
7387 addrconf_wq = create_workqueue("ipv6_addrconf");
7388 if (!addrconf_wq) {
7389 err = -ENOMEM;
7390 goto out_nowq;
7391 }
7392
7393 rtnl_lock();
7394 idev = ipv6_add_dev(blackhole_netdev);
7395 rtnl_unlock();
7396 if (IS_ERR(idev)) {
7397 err = PTR_ERR(idev);
7398 goto errlo;
7399 }
7400
7401 ip6_route_init_special_entries();
7402
7403 register_netdevice_notifier(&ipv6_dev_notf);
7404
7405 addrconf_verify(&init_net);
7406
7407 rtnl_af_register(&inet6_ops);
7408
7409 err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETLINK,
7410 NULL, inet6_dump_ifinfo, 0);
7411 if (err < 0)
7412 goto errout;
7413
7414 err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_NEWADDR,
7415 inet6_rtm_newaddr, NULL, 0);
7416 if (err < 0)
7417 goto errout;
7418 err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_DELADDR,
7419 inet6_rtm_deladdr, NULL, 0);
7420 if (err < 0)
7421 goto errout;
7422 err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETADDR,
7423 inet6_rtm_getaddr, inet6_dump_ifaddr,
7424 RTNL_FLAG_DOIT_UNLOCKED);
7425 if (err < 0)
7426 goto errout;
7427 err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETMULTICAST,
7428 NULL, inet6_dump_ifmcaddr, 0);
7429 if (err < 0)
7430 goto errout;
7431 err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETANYCAST,
7432 NULL, inet6_dump_ifacaddr, 0);
7433 if (err < 0)
7434 goto errout;
7435 err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETNETCONF,
7436 inet6_netconf_get_devconf,
7437 inet6_netconf_dump_devconf,
7438 RTNL_FLAG_DOIT_UNLOCKED);
7439 if (err < 0)
7440 goto errout;
7441 err = ipv6_addr_label_rtnl_register();
7442 if (err < 0)
7443 goto errout;
7444
7445 return 0;
7446 errout:
7447 rtnl_unregister_all(PF_INET6);
7448 rtnl_af_unregister(&inet6_ops);
7449 unregister_netdevice_notifier(&ipv6_dev_notf);
7450 errlo:
7451 destroy_workqueue(addrconf_wq);
7452 out_nowq:
7453 unregister_pernet_subsys(&addrconf_ops);
7454 out_addrlabel:
7455 ipv6_addr_label_cleanup();
7456 out:
7457 return err;
7458 }
7459
addrconf_cleanup(void)7460 void addrconf_cleanup(void)
7461 {
7462 struct net_device *dev;
7463
7464 unregister_netdevice_notifier(&ipv6_dev_notf);
7465 unregister_pernet_subsys(&addrconf_ops);
7466 ipv6_addr_label_cleanup();
7467
7468 rtnl_af_unregister(&inet6_ops);
7469
7470 rtnl_lock();
7471
7472 /* clean dev list */
7473 for_each_netdev(&init_net, dev) {
7474 if (__in6_dev_get(dev) == NULL)
7475 continue;
7476 addrconf_ifdown(dev, true);
7477 }
7478 addrconf_ifdown(init_net.loopback_dev, true);
7479
7480 rtnl_unlock();
7481
7482 destroy_workqueue(addrconf_wq);
7483 }
7484