~ [ source navigation ] ~ [ diff markup ] ~ [ identifier search ] ~

TOMOYO Linux Cross Reference
Linux/net/ipv6/addrconf.c

Version: ~ [ linux-5.15-rc7 ] ~ [ linux-5.14.14 ] ~ [ linux-5.13.19 ] ~ [ linux-5.12.19 ] ~ [ linux-5.11.22 ] ~ [ linux-5.10.75 ] ~ [ linux-5.9.16 ] ~ [ linux-5.8.18 ] ~ [ linux-5.7.19 ] ~ [ linux-5.6.19 ] ~ [ linux-5.5.19 ] ~ [ linux-5.4.155 ] ~ [ linux-5.3.18 ] ~ [ linux-5.2.21 ] ~ [ linux-5.1.21 ] ~ [ linux-5.0.21 ] ~ [ linux-4.20.17 ] ~ [ linux-4.19.213 ] ~ [ linux-4.18.20 ] ~ [ linux-4.17.19 ] ~ [ linux-4.16.18 ] ~ [ linux-4.15.18 ] ~ [ linux-4.14.252 ] ~ [ linux-4.13.16 ] ~ [ linux-4.12.14 ] ~ [ linux-4.11.12 ] ~ [ linux-4.10.17 ] ~ [ linux-4.9.287 ] ~ [ linux-4.8.17 ] ~ [ linux-4.7.10 ] ~ [ linux-4.6.7 ] ~ [ linux-4.5.7 ] ~ [ linux-4.4.289 ] ~ [ linux-4.3.6 ] ~ [ linux-4.2.8 ] ~ [ linux-4.1.52 ] ~ [ linux-4.0.9 ] ~ [ linux-3.18.140 ] ~ [ linux-3.16.85 ] ~ [ linux-3.14.79 ] ~ [ linux-3.12.74 ] ~ [ linux-3.10.108 ] ~ [ linux-2.6.32.71 ] ~ [ linux-2.6.0 ] ~ [ linux-2.4.37.11 ] ~ [ unix-v6-master ] ~ [ ccs-tools-1.8.5 ] ~ [ policy-sample ] ~
Architecture: ~ [ i386 ] ~ [ alpha ] ~ [ m68k ] ~ [ mips ] ~ [ ppc ] ~ [ sparc ] ~ [ sparc64 ] ~

  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 
 93 #define INFINITY_LIFE_TIME      0xFFFFFFFF
 94 
 95 #define IPV6_MAX_STRLEN \
 96         sizeof("ffff:ffff:ffff:ffff:ffff:ffff:255.255.255.255")
 97 
 98 static inline u32 cstamp_delta(unsigned long cstamp)
 99 {
100         return (cstamp - INITIAL_JIFFIES) * 100UL / HZ;
101 }
102 
103 static inline s32 rfc3315_s14_backoff_init(s32 irt)
104 {
105         /* multiply 'initial retransmission time' by 0.9 .. 1.1 */
106         u64 tmp = (900000 + prandom_u32() % 200001) * (u64)irt;
107         do_div(tmp, 1000000);
108         return (s32)tmp;
109 }
110 
111 static inline s32 rfc3315_s14_backoff_update(s32 rt, s32 mrt)
112 {
113         /* multiply 'retransmission timeout' by 1.9 .. 2.1 */
114         u64 tmp = (1900000 + prandom_u32() % 200001) * (u64)rt;
115         do_div(tmp, 1000000);
116         if ((s32)tmp > mrt) {
117                 /* multiply 'maximum retransmission time' by 0.9 .. 1.1 */
118                 tmp = (900000 + prandom_u32() % 200001) * (u64)mrt;
119                 do_div(tmp, 1000000);
120         }
121         return (s32)tmp;
122 }
123 
124 #ifdef CONFIG_SYSCTL
125 static int addrconf_sysctl_register(struct inet6_dev *idev);
126 static void addrconf_sysctl_unregister(struct inet6_dev *idev);
127 #else
128 static inline int addrconf_sysctl_register(struct inet6_dev *idev)
129 {
130         return 0;
131 }
132 
133 static inline void addrconf_sysctl_unregister(struct inet6_dev *idev)
134 {
135 }
136 #endif
137 
138 static void ipv6_regen_rndid(struct inet6_dev *idev);
139 static void ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr);
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  *      Configured unicast address hash table
151  */
152 static struct hlist_head inet6_addr_lst[IN6_ADDR_HSIZE];
153 static DEFINE_SPINLOCK(addrconf_hash_lock);
154 
155 static void addrconf_verify(void);
156 static void addrconf_verify_rtnl(void);
157 static void addrconf_verify_work(struct work_struct *);
158 
159 static struct workqueue_struct *addrconf_wq;
160 static DECLARE_DELAYED_WORK(addr_chk_work, addrconf_verify_work);
161 
162 static void addrconf_join_anycast(struct inet6_ifaddr *ifp);
163 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp);
164 
165 static void addrconf_type_change(struct net_device *dev,
166                                  unsigned long event);
167 static int addrconf_ifdown(struct net_device *dev, int how);
168 
169 static struct fib6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
170                                                   int plen,
171                                                   const struct net_device *dev,
172                                                   u32 flags, u32 noflags,
173                                                   bool no_gw);
174 
175 static void addrconf_dad_start(struct inet6_ifaddr *ifp);
176 static void addrconf_dad_work(struct work_struct *w);
177 static void addrconf_dad_completed(struct inet6_ifaddr *ifp, bool bump_id,
178                                    bool send_na);
179 static void addrconf_dad_run(struct inet6_dev *idev, bool restart);
180 static void addrconf_rs_timer(struct timer_list *t);
181 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
182 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
183 
184 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
185                                 struct prefix_info *pinfo);
186 
187 static struct ipv6_devconf ipv6_devconf __read_mostly = {
188         .forwarding             = 0,
189         .hop_limit              = IPV6_DEFAULT_HOPLIMIT,
190         .mtu6                   = IPV6_MIN_MTU,
191         .accept_ra              = 1,
192         .accept_redirects       = 1,
193         .autoconf               = 1,
194         .force_mld_version      = 0,
195         .mldv1_unsolicited_report_interval = 10 * HZ,
196         .mldv2_unsolicited_report_interval = HZ,
197         .dad_transmits          = 1,
198         .rtr_solicits           = MAX_RTR_SOLICITATIONS,
199         .rtr_solicit_interval   = RTR_SOLICITATION_INTERVAL,
200         .rtr_solicit_max_interval = RTR_SOLICITATION_MAX_INTERVAL,
201         .rtr_solicit_delay      = MAX_RTR_SOLICITATION_DELAY,
202         .use_tempaddr           = 0,
203         .temp_valid_lft         = TEMP_VALID_LIFETIME,
204         .temp_prefered_lft      = TEMP_PREFERRED_LIFETIME,
205         .regen_max_retry        = REGEN_MAX_RETRY,
206         .max_desync_factor      = MAX_DESYNC_FACTOR,
207         .max_addresses          = IPV6_MAX_ADDRESSES,
208         .accept_ra_defrtr       = 1,
209         .accept_ra_from_local   = 0,
210         .accept_ra_min_hop_limit= 1,
211         .accept_ra_pinfo        = 1,
212 #ifdef CONFIG_IPV6_ROUTER_PREF
213         .accept_ra_rtr_pref     = 1,
214         .rtr_probe_interval     = 60 * HZ,
215 #ifdef CONFIG_IPV6_ROUTE_INFO
216         .accept_ra_rt_info_min_plen = 0,
217         .accept_ra_rt_info_max_plen = 0,
218 #endif
219 #endif
220         .proxy_ndp              = 0,
221         .accept_source_route    = 0,    /* we do not accept RH0 by default. */
222         .disable_ipv6           = 0,
223         .accept_dad             = 0,
224         .suppress_frag_ndisc    = 1,
225         .accept_ra_mtu          = 1,
226         .stable_secret          = {
227                 .initialized = false,
228         },
229         .use_oif_addrs_only     = 0,
230         .ignore_routes_with_linkdown = 0,
231         .keep_addr_on_down      = 0,
232         .seg6_enabled           = 0,
233 #ifdef CONFIG_IPV6_SEG6_HMAC
234         .seg6_require_hmac      = 0,
235 #endif
236         .enhanced_dad           = 1,
237         .addr_gen_mode          = IN6_ADDR_GEN_MODE_EUI64,
238         .disable_policy         = 0,
239 };
240 
241 static struct ipv6_devconf ipv6_devconf_dflt __read_mostly = {
242         .forwarding             = 0,
243         .hop_limit              = IPV6_DEFAULT_HOPLIMIT,
244         .mtu6                   = IPV6_MIN_MTU,
245         .accept_ra              = 1,
246         .accept_redirects       = 1,
247         .autoconf               = 1,
248         .force_mld_version      = 0,
249         .mldv1_unsolicited_report_interval = 10 * HZ,
250         .mldv2_unsolicited_report_interval = HZ,
251         .dad_transmits          = 1,
252         .rtr_solicits           = MAX_RTR_SOLICITATIONS,
253         .rtr_solicit_interval   = RTR_SOLICITATION_INTERVAL,
254         .rtr_solicit_max_interval = RTR_SOLICITATION_MAX_INTERVAL,
255         .rtr_solicit_delay      = MAX_RTR_SOLICITATION_DELAY,
256         .use_tempaddr           = 0,
257         .temp_valid_lft         = TEMP_VALID_LIFETIME,
258         .temp_prefered_lft      = TEMP_PREFERRED_LIFETIME,
259         .regen_max_retry        = REGEN_MAX_RETRY,
260         .max_desync_factor      = MAX_DESYNC_FACTOR,
261         .max_addresses          = IPV6_MAX_ADDRESSES,
262         .accept_ra_defrtr       = 1,
263         .accept_ra_from_local   = 0,
264         .accept_ra_min_hop_limit= 1,
265         .accept_ra_pinfo        = 1,
266 #ifdef CONFIG_IPV6_ROUTER_PREF
267         .accept_ra_rtr_pref     = 1,
268         .rtr_probe_interval     = 60 * HZ,
269 #ifdef CONFIG_IPV6_ROUTE_INFO
270         .accept_ra_rt_info_min_plen = 0,
271         .accept_ra_rt_info_max_plen = 0,
272 #endif
273 #endif
274         .proxy_ndp              = 0,
275         .accept_source_route    = 0,    /* we do not accept RH0 by default. */
276         .disable_ipv6           = 0,
277         .accept_dad             = 1,
278         .suppress_frag_ndisc    = 1,
279         .accept_ra_mtu          = 1,
280         .stable_secret          = {
281                 .initialized = false,
282         },
283         .use_oif_addrs_only     = 0,
284         .ignore_routes_with_linkdown = 0,
285         .keep_addr_on_down      = 0,
286         .seg6_enabled           = 0,
287 #ifdef CONFIG_IPV6_SEG6_HMAC
288         .seg6_require_hmac      = 0,
289 #endif
290         .enhanced_dad           = 1,
291         .addr_gen_mode          = IN6_ADDR_GEN_MODE_EUI64,
292         .disable_policy         = 0,
293 };
294 
295 /* Check if link is ready: is it up and is a valid qdisc available */
296 static inline bool addrconf_link_ready(const struct net_device *dev)
297 {
298         return netif_oper_up(dev) && !qdisc_tx_is_noop(dev);
299 }
300 
301 static void addrconf_del_rs_timer(struct inet6_dev *idev)
302 {
303         if (del_timer(&idev->rs_timer))
304                 __in6_dev_put(idev);
305 }
306 
307 static void addrconf_del_dad_work(struct inet6_ifaddr *ifp)
308 {
309         if (cancel_delayed_work(&ifp->dad_work))
310                 __in6_ifa_put(ifp);
311 }
312 
313 static void addrconf_mod_rs_timer(struct inet6_dev *idev,
314                                   unsigned long when)
315 {
316         if (!timer_pending(&idev->rs_timer))
317                 in6_dev_hold(idev);
318         mod_timer(&idev->rs_timer, jiffies + when);
319 }
320 
321 static void addrconf_mod_dad_work(struct inet6_ifaddr *ifp,
322                                    unsigned long delay)
323 {
324         in6_ifa_hold(ifp);
325         if (mod_delayed_work(addrconf_wq, &ifp->dad_work, delay))
326                 in6_ifa_put(ifp);
327 }
328 
329 static int snmp6_alloc_dev(struct inet6_dev *idev)
330 {
331         int i;
332 
333         idev->stats.ipv6 = alloc_percpu(struct ipstats_mib);
334         if (!idev->stats.ipv6)
335                 goto err_ip;
336 
337         for_each_possible_cpu(i) {
338                 struct ipstats_mib *addrconf_stats;
339                 addrconf_stats = per_cpu_ptr(idev->stats.ipv6, i);
340                 u64_stats_init(&addrconf_stats->syncp);
341         }
342 
343 
344         idev->stats.icmpv6dev = kzalloc(sizeof(struct icmpv6_mib_device),
345                                         GFP_KERNEL);
346         if (!idev->stats.icmpv6dev)
347                 goto err_icmp;
348         idev->stats.icmpv6msgdev = kzalloc(sizeof(struct icmpv6msg_mib_device),
349                                            GFP_KERNEL);
350         if (!idev->stats.icmpv6msgdev)
351                 goto err_icmpmsg;
352 
353         return 0;
354 
355 err_icmpmsg:
356         kfree(idev->stats.icmpv6dev);
357 err_icmp:
358         free_percpu(idev->stats.ipv6);
359 err_ip:
360         return -ENOMEM;
361 }
362 
363 static struct inet6_dev *ipv6_add_dev(struct net_device *dev)
364 {
365         struct inet6_dev *ndev;
366         int err = -ENOMEM;
367 
368         ASSERT_RTNL();
369 
370         if (dev->mtu < IPV6_MIN_MTU)
371                 return ERR_PTR(-EINVAL);
372 
373         ndev = kzalloc(sizeof(struct inet6_dev), GFP_KERNEL);
374         if (!ndev)
375                 return ERR_PTR(err);
376 
377         rwlock_init(&ndev->lock);
378         ndev->dev = dev;
379         INIT_LIST_HEAD(&ndev->addr_list);
380         timer_setup(&ndev->rs_timer, addrconf_rs_timer, 0);
381         memcpy(&ndev->cnf, dev_net(dev)->ipv6.devconf_dflt, sizeof(ndev->cnf));
382 
383         if (ndev->cnf.stable_secret.initialized)
384                 ndev->cnf.addr_gen_mode = IN6_ADDR_GEN_MODE_STABLE_PRIVACY;
385 
386         ndev->cnf.mtu6 = dev->mtu;
387         ndev->nd_parms = neigh_parms_alloc(dev, &nd_tbl);
388         if (!ndev->nd_parms) {
389                 kfree(ndev);
390                 return ERR_PTR(err);
391         }
392         if (ndev->cnf.forwarding)
393                 dev_disable_lro(dev);
394         /* We refer to the device */
395         dev_hold(dev);
396 
397         if (snmp6_alloc_dev(ndev) < 0) {
398                 netdev_dbg(dev, "%s: cannot allocate memory for statistics\n",
399                            __func__);
400                 neigh_parms_release(&nd_tbl, ndev->nd_parms);
401                 dev_put(dev);
402                 kfree(ndev);
403                 return ERR_PTR(err);
404         }
405 
406         if (snmp6_register_dev(ndev) < 0) {
407                 netdev_dbg(dev, "%s: cannot create /proc/net/dev_snmp6/%s\n",
408                            __func__, dev->name);
409                 goto err_release;
410         }
411 
412         /* One reference from device. */
413         refcount_set(&ndev->refcnt, 1);
414 
415         if (dev->flags & (IFF_NOARP | IFF_LOOPBACK))
416                 ndev->cnf.accept_dad = -1;
417 
418 #if IS_ENABLED(CONFIG_IPV6_SIT)
419         if (dev->type == ARPHRD_SIT && (dev->priv_flags & IFF_ISATAP)) {
420                 pr_info("%s: Disabled Multicast RS\n", dev->name);
421                 ndev->cnf.rtr_solicits = 0;
422         }
423 #endif
424 
425         INIT_LIST_HEAD(&ndev->tempaddr_list);
426         ndev->desync_factor = U32_MAX;
427         if ((dev->flags&IFF_LOOPBACK) ||
428             dev->type == ARPHRD_TUNNEL ||
429             dev->type == ARPHRD_TUNNEL6 ||
430             dev->type == ARPHRD_SIT ||
431             dev->type == ARPHRD_NONE) {
432                 ndev->cnf.use_tempaddr = -1;
433         } else
434                 ipv6_regen_rndid(ndev);
435 
436         ndev->token = in6addr_any;
437 
438         if (netif_running(dev) && addrconf_link_ready(dev))
439                 ndev->if_flags |= IF_READY;
440 
441         ipv6_mc_init_dev(ndev);
442         ndev->tstamp = jiffies;
443         err = addrconf_sysctl_register(ndev);
444         if (err) {
445                 ipv6_mc_destroy_dev(ndev);
446                 snmp6_unregister_dev(ndev);
447                 goto err_release;
448         }
449         /* protected by rtnl_lock */
450         rcu_assign_pointer(dev->ip6_ptr, ndev);
451 
452         /* Join interface-local all-node multicast group */
453         ipv6_dev_mc_inc(dev, &in6addr_interfacelocal_allnodes);
454 
455         /* Join all-node multicast group */
456         ipv6_dev_mc_inc(dev, &in6addr_linklocal_allnodes);
457 
458         /* Join all-router multicast group if forwarding is set */
459         if (ndev->cnf.forwarding && (dev->flags & IFF_MULTICAST))
460                 ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
461 
462         return ndev;
463 
464 err_release:
465         neigh_parms_release(&nd_tbl, ndev->nd_parms);
466         ndev->dead = 1;
467         in6_dev_finish_destroy(ndev);
468         return ERR_PTR(err);
469 }
470 
471 static struct inet6_dev *ipv6_find_idev(struct net_device *dev)
472 {
473         struct inet6_dev *idev;
474 
475         ASSERT_RTNL();
476 
477         idev = __in6_dev_get(dev);
478         if (!idev) {
479                 idev = ipv6_add_dev(dev);
480                 if (IS_ERR(idev))
481                         return idev;
482         }
483 
484         if (dev->flags&IFF_UP)
485                 ipv6_mc_up(idev);
486         return idev;
487 }
488 
489 static int inet6_netconf_msgsize_devconf(int type)
490 {
491         int size =  NLMSG_ALIGN(sizeof(struct netconfmsg))
492                     + nla_total_size(4);        /* NETCONFA_IFINDEX */
493         bool all = false;
494 
495         if (type == NETCONFA_ALL)
496                 all = true;
497 
498         if (all || type == NETCONFA_FORWARDING)
499                 size += nla_total_size(4);
500 #ifdef CONFIG_IPV6_MROUTE
501         if (all || type == NETCONFA_MC_FORWARDING)
502                 size += nla_total_size(4);
503 #endif
504         if (all || type == NETCONFA_PROXY_NEIGH)
505                 size += nla_total_size(4);
506 
507         if (all || type == NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN)
508                 size += nla_total_size(4);
509 
510         return size;
511 }
512 
513 static int inet6_netconf_fill_devconf(struct sk_buff *skb, int ifindex,
514                                       struct ipv6_devconf *devconf, u32 portid,
515                                       u32 seq, int event, unsigned int flags,
516                                       int type)
517 {
518         struct nlmsghdr  *nlh;
519         struct netconfmsg *ncm;
520         bool all = false;
521 
522         nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct netconfmsg),
523                         flags);
524         if (!nlh)
525                 return -EMSGSIZE;
526 
527         if (type == NETCONFA_ALL)
528                 all = true;
529 
530         ncm = nlmsg_data(nlh);
531         ncm->ncm_family = AF_INET6;
532 
533         if (nla_put_s32(skb, NETCONFA_IFINDEX, ifindex) < 0)
534                 goto nla_put_failure;
535 
536         if (!devconf)
537                 goto out;
538 
539         if ((all || type == NETCONFA_FORWARDING) &&
540             nla_put_s32(skb, NETCONFA_FORWARDING, devconf->forwarding) < 0)
541                 goto nla_put_failure;
542 #ifdef CONFIG_IPV6_MROUTE
543         if ((all || type == NETCONFA_MC_FORWARDING) &&
544             nla_put_s32(skb, NETCONFA_MC_FORWARDING,
545                         devconf->mc_forwarding) < 0)
546                 goto nla_put_failure;
547 #endif
548         if ((all || type == NETCONFA_PROXY_NEIGH) &&
549             nla_put_s32(skb, NETCONFA_PROXY_NEIGH, devconf->proxy_ndp) < 0)
550                 goto nla_put_failure;
551 
552         if ((all || type == NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN) &&
553             nla_put_s32(skb, NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
554                         devconf->ignore_routes_with_linkdown) < 0)
555                 goto nla_put_failure;
556 
557 out:
558         nlmsg_end(skb, nlh);
559         return 0;
560 
561 nla_put_failure:
562         nlmsg_cancel(skb, nlh);
563         return -EMSGSIZE;
564 }
565 
566 void inet6_netconf_notify_devconf(struct net *net, int event, int type,
567                                   int ifindex, struct ipv6_devconf *devconf)
568 {
569         struct sk_buff *skb;
570         int err = -ENOBUFS;
571 
572         skb = nlmsg_new(inet6_netconf_msgsize_devconf(type), GFP_KERNEL);
573         if (!skb)
574                 goto errout;
575 
576         err = inet6_netconf_fill_devconf(skb, ifindex, devconf, 0, 0,
577                                          event, 0, type);
578         if (err < 0) {
579                 /* -EMSGSIZE implies BUG in inet6_netconf_msgsize_devconf() */
580                 WARN_ON(err == -EMSGSIZE);
581                 kfree_skb(skb);
582                 goto errout;
583         }
584         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_NETCONF, NULL, GFP_KERNEL);
585         return;
586 errout:
587         rtnl_set_sk_err(net, RTNLGRP_IPV6_NETCONF, err);
588 }
589 
590 static const struct nla_policy devconf_ipv6_policy[NETCONFA_MAX+1] = {
591         [NETCONFA_IFINDEX]      = { .len = sizeof(int) },
592         [NETCONFA_FORWARDING]   = { .len = sizeof(int) },
593         [NETCONFA_PROXY_NEIGH]  = { .len = sizeof(int) },
594         [NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN]  = { .len = sizeof(int) },
595 };
596 
597 static int inet6_netconf_valid_get_req(struct sk_buff *skb,
598                                        const struct nlmsghdr *nlh,
599                                        struct nlattr **tb,
600                                        struct netlink_ext_ack *extack)
601 {
602         int i, err;
603 
604         if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(struct netconfmsg))) {
605                 NL_SET_ERR_MSG_MOD(extack, "Invalid header for netconf get request");
606                 return -EINVAL;
607         }
608 
609         if (!netlink_strict_get_check(skb))
610                 return nlmsg_parse_deprecated(nlh, sizeof(struct netconfmsg),
611                                               tb, NETCONFA_MAX,
612                                               devconf_ipv6_policy, extack);
613 
614         err = nlmsg_parse_deprecated_strict(nlh, sizeof(struct netconfmsg),
615                                             tb, NETCONFA_MAX,
616                                             devconf_ipv6_policy, extack);
617         if (err)
618                 return err;
619 
620         for (i = 0; i <= NETCONFA_MAX; i++) {
621                 if (!tb[i])
622                         continue;
623 
624                 switch (i) {
625                 case NETCONFA_IFINDEX:
626                         break;
627                 default:
628                         NL_SET_ERR_MSG_MOD(extack, "Unsupported attribute in netconf get request");
629                         return -EINVAL;
630                 }
631         }
632 
633         return 0;
634 }
635 
636 static int inet6_netconf_get_devconf(struct sk_buff *in_skb,
637                                      struct nlmsghdr *nlh,
638                                      struct netlink_ext_ack *extack)
639 {
640         struct net *net = sock_net(in_skb->sk);
641         struct nlattr *tb[NETCONFA_MAX+1];
642         struct inet6_dev *in6_dev = NULL;
643         struct net_device *dev = NULL;
644         struct sk_buff *skb;
645         struct ipv6_devconf *devconf;
646         int ifindex;
647         int err;
648 
649         err = inet6_netconf_valid_get_req(in_skb, nlh, tb, extack);
650         if (err < 0)
651                 return err;
652 
653         if (!tb[NETCONFA_IFINDEX])
654                 return -EINVAL;
655 
656         err = -EINVAL;
657         ifindex = nla_get_s32(tb[NETCONFA_IFINDEX]);
658         switch (ifindex) {
659         case NETCONFA_IFINDEX_ALL:
660                 devconf = net->ipv6.devconf_all;
661                 break;
662         case NETCONFA_IFINDEX_DEFAULT:
663                 devconf = net->ipv6.devconf_dflt;
664                 break;
665         default:
666                 dev = dev_get_by_index(net, ifindex);
667                 if (!dev)
668                         return -EINVAL;
669                 in6_dev = in6_dev_get(dev);
670                 if (!in6_dev)
671                         goto errout;
672                 devconf = &in6_dev->cnf;
673                 break;
674         }
675 
676         err = -ENOBUFS;
677         skb = nlmsg_new(inet6_netconf_msgsize_devconf(NETCONFA_ALL), GFP_KERNEL);
678         if (!skb)
679                 goto errout;
680 
681         err = inet6_netconf_fill_devconf(skb, ifindex, devconf,
682                                          NETLINK_CB(in_skb).portid,
683                                          nlh->nlmsg_seq, RTM_NEWNETCONF, 0,
684                                          NETCONFA_ALL);
685         if (err < 0) {
686                 /* -EMSGSIZE implies BUG in inet6_netconf_msgsize_devconf() */
687                 WARN_ON(err == -EMSGSIZE);
688                 kfree_skb(skb);
689                 goto errout;
690         }
691         err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
692 errout:
693         if (in6_dev)
694                 in6_dev_put(in6_dev);
695         if (dev)
696                 dev_put(dev);
697         return err;
698 }
699 
700 static int inet6_netconf_dump_devconf(struct sk_buff *skb,
701                                       struct netlink_callback *cb)
702 {
703         const struct nlmsghdr *nlh = cb->nlh;
704         struct net *net = sock_net(skb->sk);
705         int h, s_h;
706         int idx, s_idx;
707         struct net_device *dev;
708         struct inet6_dev *idev;
709         struct hlist_head *head;
710 
711         if (cb->strict_check) {
712                 struct netlink_ext_ack *extack = cb->extack;
713                 struct netconfmsg *ncm;
714 
715                 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ncm))) {
716                         NL_SET_ERR_MSG_MOD(extack, "Invalid header for netconf dump request");
717                         return -EINVAL;
718                 }
719 
720                 if (nlmsg_attrlen(nlh, sizeof(*ncm))) {
721                         NL_SET_ERR_MSG_MOD(extack, "Invalid data after header in netconf dump request");
722                         return -EINVAL;
723                 }
724         }
725 
726         s_h = cb->args[0];
727         s_idx = idx = cb->args[1];
728 
729         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
730                 idx = 0;
731                 head = &net->dev_index_head[h];
732                 rcu_read_lock();
733                 cb->seq = atomic_read(&net->ipv6.dev_addr_genid) ^
734                           net->dev_base_seq;
735                 hlist_for_each_entry_rcu(dev, head, index_hlist) {
736                         if (idx < s_idx)
737                                 goto cont;
738                         idev = __in6_dev_get(dev);
739                         if (!idev)
740                                 goto cont;
741 
742                         if (inet6_netconf_fill_devconf(skb, dev->ifindex,
743                                                        &idev->cnf,
744                                                        NETLINK_CB(cb->skb).portid,
745                                                        nlh->nlmsg_seq,
746                                                        RTM_NEWNETCONF,
747                                                        NLM_F_MULTI,
748                                                        NETCONFA_ALL) < 0) {
749                                 rcu_read_unlock();
750                                 goto done;
751                         }
752                         nl_dump_check_consistent(cb, nlmsg_hdr(skb));
753 cont:
754                         idx++;
755                 }
756                 rcu_read_unlock();
757         }
758         if (h == NETDEV_HASHENTRIES) {
759                 if (inet6_netconf_fill_devconf(skb, NETCONFA_IFINDEX_ALL,
760                                                net->ipv6.devconf_all,
761                                                NETLINK_CB(cb->skb).portid,
762                                                nlh->nlmsg_seq,
763                                                RTM_NEWNETCONF, NLM_F_MULTI,
764                                                NETCONFA_ALL) < 0)
765                         goto done;
766                 else
767                         h++;
768         }
769         if (h == NETDEV_HASHENTRIES + 1) {
770                 if (inet6_netconf_fill_devconf(skb, NETCONFA_IFINDEX_DEFAULT,
771                                                net->ipv6.devconf_dflt,
772                                                NETLINK_CB(cb->skb).portid,
773                                                nlh->nlmsg_seq,
774                                                RTM_NEWNETCONF, NLM_F_MULTI,
775                                                NETCONFA_ALL) < 0)
776                         goto done;
777                 else
778                         h++;
779         }
780 done:
781         cb->args[0] = h;
782         cb->args[1] = idx;
783 
784         return skb->len;
785 }
786 
787 #ifdef CONFIG_SYSCTL
788 static void dev_forward_change(struct inet6_dev *idev)
789 {
790         struct net_device *dev;
791         struct inet6_ifaddr *ifa;
792 
793         if (!idev)
794                 return;
795         dev = idev->dev;
796         if (idev->cnf.forwarding)
797                 dev_disable_lro(dev);
798         if (dev->flags & IFF_MULTICAST) {
799                 if (idev->cnf.forwarding) {
800                         ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
801                         ipv6_dev_mc_inc(dev, &in6addr_interfacelocal_allrouters);
802                         ipv6_dev_mc_inc(dev, &in6addr_sitelocal_allrouters);
803                 } else {
804                         ipv6_dev_mc_dec(dev, &in6addr_linklocal_allrouters);
805                         ipv6_dev_mc_dec(dev, &in6addr_interfacelocal_allrouters);
806                         ipv6_dev_mc_dec(dev, &in6addr_sitelocal_allrouters);
807                 }
808         }
809 
810         list_for_each_entry(ifa, &idev->addr_list, if_list) {
811                 if (ifa->flags&IFA_F_TENTATIVE)
812                         continue;
813                 if (idev->cnf.forwarding)
814                         addrconf_join_anycast(ifa);
815                 else
816                         addrconf_leave_anycast(ifa);
817         }
818         inet6_netconf_notify_devconf(dev_net(dev), RTM_NEWNETCONF,
819                                      NETCONFA_FORWARDING,
820                                      dev->ifindex, &idev->cnf);
821 }
822 
823 
824 static void addrconf_forward_change(struct net *net, __s32 newf)
825 {
826         struct net_device *dev;
827         struct inet6_dev *idev;
828 
829         for_each_netdev(net, dev) {
830                 idev = __in6_dev_get(dev);
831                 if (idev) {
832                         int changed = (!idev->cnf.forwarding) ^ (!newf);
833                         idev->cnf.forwarding = newf;
834                         if (changed)
835                                 dev_forward_change(idev);
836                 }
837         }
838 }
839 
840 static int addrconf_fixup_forwarding(struct ctl_table *table, int *p, int newf)
841 {
842         struct net *net;
843         int old;
844 
845         if (!rtnl_trylock())
846                 return restart_syscall();
847 
848         net = (struct net *)table->extra2;
849         old = *p;
850         *p = newf;
851 
852         if (p == &net->ipv6.devconf_dflt->forwarding) {
853                 if ((!newf) ^ (!old))
854                         inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
855                                                      NETCONFA_FORWARDING,
856                                                      NETCONFA_IFINDEX_DEFAULT,
857                                                      net->ipv6.devconf_dflt);
858                 rtnl_unlock();
859                 return 0;
860         }
861 
862         if (p == &net->ipv6.devconf_all->forwarding) {
863                 int old_dflt = net->ipv6.devconf_dflt->forwarding;
864 
865                 net->ipv6.devconf_dflt->forwarding = newf;
866                 if ((!newf) ^ (!old_dflt))
867                         inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
868                                                      NETCONFA_FORWARDING,
869                                                      NETCONFA_IFINDEX_DEFAULT,
870                                                      net->ipv6.devconf_dflt);
871 
872                 addrconf_forward_change(net, newf);
873                 if ((!newf) ^ (!old))
874                         inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
875                                                      NETCONFA_FORWARDING,
876                                                      NETCONFA_IFINDEX_ALL,
877                                                      net->ipv6.devconf_all);
878         } else if ((!newf) ^ (!old))
879                 dev_forward_change((struct inet6_dev *)table->extra1);
880         rtnl_unlock();
881 
882         if (newf)
883                 rt6_purge_dflt_routers(net);
884         return 1;
885 }
886 
887 static void addrconf_linkdown_change(struct net *net, __s32 newf)
888 {
889         struct net_device *dev;
890         struct inet6_dev *idev;
891 
892         for_each_netdev(net, dev) {
893                 idev = __in6_dev_get(dev);
894                 if (idev) {
895                         int changed = (!idev->cnf.ignore_routes_with_linkdown) ^ (!newf);
896 
897                         idev->cnf.ignore_routes_with_linkdown = newf;
898                         if (changed)
899                                 inet6_netconf_notify_devconf(dev_net(dev),
900                                                              RTM_NEWNETCONF,
901                                                              NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
902                                                              dev->ifindex,
903                                                              &idev->cnf);
904                 }
905         }
906 }
907 
908 static int addrconf_fixup_linkdown(struct ctl_table *table, int *p, int newf)
909 {
910         struct net *net;
911         int old;
912 
913         if (!rtnl_trylock())
914                 return restart_syscall();
915 
916         net = (struct net *)table->extra2;
917         old = *p;
918         *p = newf;
919 
920         if (p == &net->ipv6.devconf_dflt->ignore_routes_with_linkdown) {
921                 if ((!newf) ^ (!old))
922                         inet6_netconf_notify_devconf(net,
923                                                      RTM_NEWNETCONF,
924                                                      NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
925                                                      NETCONFA_IFINDEX_DEFAULT,
926                                                      net->ipv6.devconf_dflt);
927                 rtnl_unlock();
928                 return 0;
929         }
930 
931         if (p == &net->ipv6.devconf_all->ignore_routes_with_linkdown) {
932                 net->ipv6.devconf_dflt->ignore_routes_with_linkdown = newf;
933                 addrconf_linkdown_change(net, newf);
934                 if ((!newf) ^ (!old))
935                         inet6_netconf_notify_devconf(net,
936                                                      RTM_NEWNETCONF,
937                                                      NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
938                                                      NETCONFA_IFINDEX_ALL,
939                                                      net->ipv6.devconf_all);
940         }
941         rtnl_unlock();
942 
943         return 1;
944 }
945 
946 #endif
947 
948 /* Nobody refers to this ifaddr, destroy it */
949 void inet6_ifa_finish_destroy(struct inet6_ifaddr *ifp)
950 {
951         WARN_ON(!hlist_unhashed(&ifp->addr_lst));
952 
953 #ifdef NET_REFCNT_DEBUG
954         pr_debug("%s\n", __func__);
955 #endif
956 
957         in6_dev_put(ifp->idev);
958 
959         if (cancel_delayed_work(&ifp->dad_work))
960                 pr_notice("delayed DAD work was pending while freeing ifa=%p\n",
961                           ifp);
962 
963         if (ifp->state != INET6_IFADDR_STATE_DEAD) {
964                 pr_warn("Freeing alive inet6 address %p\n", ifp);
965                 return;
966         }
967 
968         kfree_rcu(ifp, rcu);
969 }
970 
971 static void
972 ipv6_link_dev_addr(struct inet6_dev *idev, struct inet6_ifaddr *ifp)
973 {
974         struct list_head *p;
975         int ifp_scope = ipv6_addr_src_scope(&ifp->addr);
976 
977         /*
978          * Each device address list is sorted in order of scope -
979          * global before linklocal.
980          */
981         list_for_each(p, &idev->addr_list) {
982                 struct inet6_ifaddr *ifa
983                         = list_entry(p, struct inet6_ifaddr, if_list);
984                 if (ifp_scope >= ipv6_addr_src_scope(&ifa->addr))
985                         break;
986         }
987 
988         list_add_tail_rcu(&ifp->if_list, p);
989 }
990 
991 static u32 inet6_addr_hash(const struct net *net, const struct in6_addr *addr)
992 {
993         u32 val = ipv6_addr_hash(addr) ^ net_hash_mix(net);
994 
995         return hash_32(val, IN6_ADDR_HSIZE_SHIFT);
996 }
997 
998 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
999                                struct net_device *dev, unsigned int hash)
1000 {
1001         struct inet6_ifaddr *ifp;
1002 
1003         hlist_for_each_entry(ifp, &inet6_addr_lst[hash], addr_lst) {
1004                 if (!net_eq(dev_net(ifp->idev->dev), net))
1005                         continue;
1006                 if (ipv6_addr_equal(&ifp->addr, addr)) {
1007                         if (!dev || ifp->idev->dev == dev)
1008                                 return true;
1009                 }
1010         }
1011         return false;
1012 }
1013 
1014 static int ipv6_add_addr_hash(struct net_device *dev, struct inet6_ifaddr *ifa)
1015 {
1016         unsigned int hash = inet6_addr_hash(dev_net(dev), &ifa->addr);
1017         int err = 0;
1018 
1019         spin_lock(&addrconf_hash_lock);
1020 
1021         /* Ignore adding duplicate addresses on an interface */
1022         if (ipv6_chk_same_addr(dev_net(dev), &ifa->addr, dev, hash)) {
1023                 netdev_dbg(dev, "ipv6_add_addr: already assigned\n");
1024                 err = -EEXIST;
1025         } else {
1026                 hlist_add_head_rcu(&ifa->addr_lst, &inet6_addr_lst[hash]);
1027         }
1028 
1029         spin_unlock(&addrconf_hash_lock);
1030 
1031         return err;
1032 }
1033 
1034 /* On success it returns ifp with increased reference count */
1035 
1036 static struct inet6_ifaddr *
1037 ipv6_add_addr(struct inet6_dev *idev, struct ifa6_config *cfg,
1038               bool can_block, struct netlink_ext_ack *extack)
1039 {
1040         gfp_t gfp_flags = can_block ? GFP_KERNEL : GFP_ATOMIC;
1041         int addr_type = ipv6_addr_type(cfg->pfx);
1042         struct net *net = dev_net(idev->dev);
1043         struct inet6_ifaddr *ifa = NULL;
1044         struct fib6_info *f6i = NULL;
1045         int err = 0;
1046 
1047         if (addr_type == IPV6_ADDR_ANY ||
1048             (addr_type & IPV6_ADDR_MULTICAST &&
1049              !(cfg->ifa_flags & IFA_F_MCAUTOJOIN)) ||
1050             (!(idev->dev->flags & IFF_LOOPBACK) &&
1051              !netif_is_l3_master(idev->dev) &&
1052              addr_type & IPV6_ADDR_LOOPBACK))
1053                 return ERR_PTR(-EADDRNOTAVAIL);
1054 
1055         if (idev->dead) {
1056                 err = -ENODEV;                  /*XXX*/
1057                 goto out;
1058         }
1059 
1060         if (idev->cnf.disable_ipv6) {
1061                 err = -EACCES;
1062                 goto out;
1063         }
1064 
1065         /* validator notifier needs to be blocking;
1066          * do not call in atomic context
1067          */
1068         if (can_block) {
1069                 struct in6_validator_info i6vi = {
1070                         .i6vi_addr = *cfg->pfx,
1071                         .i6vi_dev = idev,
1072                         .extack = extack,
1073                 };
1074 
1075                 err = inet6addr_validator_notifier_call_chain(NETDEV_UP, &i6vi);
1076                 err = notifier_to_errno(err);
1077                 if (err < 0)
1078                         goto out;
1079         }
1080 
1081         ifa = kzalloc(sizeof(*ifa), gfp_flags);
1082         if (!ifa) {
1083                 err = -ENOBUFS;
1084                 goto out;
1085         }
1086 
1087         f6i = addrconf_f6i_alloc(net, idev, cfg->pfx, false, gfp_flags);
1088         if (IS_ERR(f6i)) {
1089                 err = PTR_ERR(f6i);
1090                 f6i = NULL;
1091                 goto out;
1092         }
1093 
1094         if (net->ipv6.devconf_all->disable_policy ||
1095             idev->cnf.disable_policy)
1096                 f6i->dst_nopolicy = true;
1097 
1098         neigh_parms_data_state_setall(idev->nd_parms);
1099 
1100         ifa->addr = *cfg->pfx;
1101         if (cfg->peer_pfx)
1102                 ifa->peer_addr = *cfg->peer_pfx;
1103 
1104         spin_lock_init(&ifa->lock);
1105         INIT_DELAYED_WORK(&ifa->dad_work, addrconf_dad_work);
1106         INIT_HLIST_NODE(&ifa->addr_lst);
1107         ifa->scope = cfg->scope;
1108         ifa->prefix_len = cfg->plen;
1109         ifa->rt_priority = cfg->rt_priority;
1110         ifa->flags = cfg->ifa_flags;
1111         /* No need to add the TENTATIVE flag for addresses with NODAD */
1112         if (!(cfg->ifa_flags & IFA_F_NODAD))
1113                 ifa->flags |= IFA_F_TENTATIVE;
1114         ifa->valid_lft = cfg->valid_lft;
1115         ifa->prefered_lft = cfg->preferred_lft;
1116         ifa->cstamp = ifa->tstamp = jiffies;
1117         ifa->tokenized = false;
1118 
1119         ifa->rt = f6i;
1120 
1121         ifa->idev = idev;
1122         in6_dev_hold(idev);
1123 
1124         /* For caller */
1125         refcount_set(&ifa->refcnt, 1);
1126 
1127         rcu_read_lock_bh();
1128 
1129         err = ipv6_add_addr_hash(idev->dev, ifa);
1130         if (err < 0) {
1131                 rcu_read_unlock_bh();
1132                 goto out;
1133         }
1134 
1135         write_lock(&idev->lock);
1136 
1137         /* Add to inet6_dev unicast addr list. */
1138         ipv6_link_dev_addr(idev, ifa);
1139 
1140         if (ifa->flags&IFA_F_TEMPORARY) {
1141                 list_add(&ifa->tmp_list, &idev->tempaddr_list);
1142                 in6_ifa_hold(ifa);
1143         }
1144 
1145         in6_ifa_hold(ifa);
1146         write_unlock(&idev->lock);
1147 
1148         rcu_read_unlock_bh();
1149 
1150         inet6addr_notifier_call_chain(NETDEV_UP, ifa);
1151 out:
1152         if (unlikely(err < 0)) {
1153                 fib6_info_release(f6i);
1154 
1155                 if (ifa) {
1156                         if (ifa->idev)
1157                                 in6_dev_put(ifa->idev);
1158                         kfree(ifa);
1159                 }
1160                 ifa = ERR_PTR(err);
1161         }
1162 
1163         return ifa;
1164 }
1165 
1166 enum cleanup_prefix_rt_t {
1167         CLEANUP_PREFIX_RT_NOP,    /* no cleanup action for prefix route */
1168         CLEANUP_PREFIX_RT_DEL,    /* delete the prefix route */
1169         CLEANUP_PREFIX_RT_EXPIRE, /* update the lifetime of the prefix route */
1170 };
1171 
1172 /*
1173  * Check, whether the prefix for ifp would still need a prefix route
1174  * after deleting ifp. The function returns one of the CLEANUP_PREFIX_RT_*
1175  * constants.
1176  *
1177  * 1) we don't purge prefix if address was not permanent.
1178  *    prefix is managed by its own lifetime.
1179  * 2) we also don't purge, if the address was IFA_F_NOPREFIXROUTE.
1180  * 3) if there are no addresses, delete prefix.
1181  * 4) if there are still other permanent address(es),
1182  *    corresponding prefix is still permanent.
1183  * 5) if there are still other addresses with IFA_F_NOPREFIXROUTE,
1184  *    don't purge the prefix, assume user space is managing it.
1185  * 6) otherwise, update prefix lifetime to the
1186  *    longest valid lifetime among the corresponding
1187  *    addresses on the device.
1188  *    Note: subsequent RA will update lifetime.
1189  **/
1190 static enum cleanup_prefix_rt_t
1191 check_cleanup_prefix_route(struct inet6_ifaddr *ifp, unsigned long *expires)
1192 {
1193         struct inet6_ifaddr *ifa;
1194         struct inet6_dev *idev = ifp->idev;
1195         unsigned long lifetime;
1196         enum cleanup_prefix_rt_t action = CLEANUP_PREFIX_RT_DEL;
1197 
1198         *expires = jiffies;
1199 
1200         list_for_each_entry(ifa, &idev->addr_list, if_list) {
1201                 if (ifa == ifp)
1202                         continue;
1203                 if (ifa->prefix_len != ifp->prefix_len ||
1204                     !ipv6_prefix_equal(&ifa->addr, &ifp->addr,
1205                                        ifp->prefix_len))
1206                         continue;
1207                 if (ifa->flags & (IFA_F_PERMANENT | IFA_F_NOPREFIXROUTE))
1208                         return CLEANUP_PREFIX_RT_NOP;
1209 
1210                 action = CLEANUP_PREFIX_RT_EXPIRE;
1211 
1212                 spin_lock(&ifa->lock);
1213 
1214                 lifetime = addrconf_timeout_fixup(ifa->valid_lft, HZ);
1215                 /*
1216                  * Note: Because this address is
1217                  * not permanent, lifetime <
1218                  * LONG_MAX / HZ here.
1219                  */
1220                 if (time_before(*expires, ifa->tstamp + lifetime * HZ))
1221                         *expires = ifa->tstamp + lifetime * HZ;
1222                 spin_unlock(&ifa->lock);
1223         }
1224 
1225         return action;
1226 }
1227 
1228 static void
1229 cleanup_prefix_route(struct inet6_ifaddr *ifp, unsigned long expires,
1230                      bool del_rt, bool del_peer)
1231 {
1232         struct fib6_info *f6i;
1233 
1234         f6i = addrconf_get_prefix_route(del_peer ? &ifp->peer_addr : &ifp->addr,
1235                                         ifp->prefix_len,
1236                                         ifp->idev->dev, 0, RTF_DEFAULT, true);
1237         if (f6i) {
1238                 if (del_rt)
1239                         ip6_del_rt(dev_net(ifp->idev->dev), f6i);
1240                 else {
1241                         if (!(f6i->fib6_flags & RTF_EXPIRES))
1242                                 fib6_set_expires(f6i, expires);
1243                         fib6_info_release(f6i);
1244                 }
1245         }
1246 }
1247 
1248 
1249 /* This function wants to get referenced ifp and releases it before return */
1250 
1251 static void ipv6_del_addr(struct inet6_ifaddr *ifp)
1252 {
1253         int state;
1254         enum cleanup_prefix_rt_t action = CLEANUP_PREFIX_RT_NOP;
1255         unsigned long expires;
1256 
1257         ASSERT_RTNL();
1258 
1259         spin_lock_bh(&ifp->lock);
1260         state = ifp->state;
1261         ifp->state = INET6_IFADDR_STATE_DEAD;
1262         spin_unlock_bh(&ifp->lock);
1263 
1264         if (state == INET6_IFADDR_STATE_DEAD)
1265                 goto out;
1266 
1267         spin_lock_bh(&addrconf_hash_lock);
1268         hlist_del_init_rcu(&ifp->addr_lst);
1269         spin_unlock_bh(&addrconf_hash_lock);
1270 
1271         write_lock_bh(&ifp->idev->lock);
1272 
1273         if (ifp->flags&IFA_F_TEMPORARY) {
1274                 list_del(&ifp->tmp_list);
1275                 if (ifp->ifpub) {
1276                         in6_ifa_put(ifp->ifpub);
1277                         ifp->ifpub = NULL;
1278                 }
1279                 __in6_ifa_put(ifp);
1280         }
1281 
1282         if (ifp->flags & IFA_F_PERMANENT && !(ifp->flags & IFA_F_NOPREFIXROUTE))
1283                 action = check_cleanup_prefix_route(ifp, &expires);
1284 
1285         list_del_rcu(&ifp->if_list);
1286         __in6_ifa_put(ifp);
1287 
1288         write_unlock_bh(&ifp->idev->lock);
1289 
1290         addrconf_del_dad_work(ifp);
1291 
1292         ipv6_ifa_notify(RTM_DELADDR, ifp);
1293 
1294         inet6addr_notifier_call_chain(NETDEV_DOWN, ifp);
1295 
1296         if (action != CLEANUP_PREFIX_RT_NOP) {
1297                 cleanup_prefix_route(ifp, expires,
1298                         action == CLEANUP_PREFIX_RT_DEL, false);
1299         }
1300 
1301         /* clean up prefsrc entries */
1302         rt6_remove_prefsrc(ifp);
1303 out:
1304         in6_ifa_put(ifp);
1305 }
1306 
1307 static int ipv6_create_tempaddr(struct inet6_ifaddr *ifp,
1308                                 struct inet6_ifaddr *ift,
1309                                 bool block)
1310 {
1311         struct inet6_dev *idev = ifp->idev;
1312         struct in6_addr addr, *tmpaddr;
1313         unsigned long tmp_tstamp, age;
1314         unsigned long regen_advance;
1315         struct ifa6_config cfg;
1316         int ret = 0;
1317         unsigned long now = jiffies;
1318         long max_desync_factor;
1319         s32 cnf_temp_preferred_lft;
1320 
1321         write_lock_bh(&idev->lock);
1322         if (ift) {
1323                 spin_lock_bh(&ift->lock);
1324                 memcpy(&addr.s6_addr[8], &ift->addr.s6_addr[8], 8);
1325                 spin_unlock_bh(&ift->lock);
1326                 tmpaddr = &addr;
1327         } else {
1328                 tmpaddr = NULL;
1329         }
1330 retry:
1331         in6_dev_hold(idev);
1332         if (idev->cnf.use_tempaddr <= 0) {
1333                 write_unlock_bh(&idev->lock);
1334                 pr_info("%s: use_tempaddr is disabled\n", __func__);
1335                 in6_dev_put(idev);
1336                 ret = -1;
1337                 goto out;
1338         }
1339         spin_lock_bh(&ifp->lock);
1340         if (ifp->regen_count++ >= idev->cnf.regen_max_retry) {
1341                 idev->cnf.use_tempaddr = -1;    /*XXX*/
1342                 spin_unlock_bh(&ifp->lock);
1343                 write_unlock_bh(&idev->lock);
1344                 pr_warn("%s: regeneration time exceeded - disabled temporary address support\n",
1345                         __func__);
1346                 in6_dev_put(idev);
1347                 ret = -1;
1348                 goto out;
1349         }
1350         in6_ifa_hold(ifp);
1351         memcpy(addr.s6_addr, ifp->addr.s6_addr, 8);
1352         ipv6_try_regen_rndid(idev, tmpaddr);
1353         memcpy(&addr.s6_addr[8], idev->rndid, 8);
1354         age = (now - ifp->tstamp) / HZ;
1355 
1356         regen_advance = idev->cnf.regen_max_retry *
1357                         idev->cnf.dad_transmits *
1358                         NEIGH_VAR(idev->nd_parms, RETRANS_TIME) / HZ;
1359 
1360         /* recalculate max_desync_factor each time and update
1361          * idev->desync_factor if it's larger
1362          */
1363         cnf_temp_preferred_lft = READ_ONCE(idev->cnf.temp_prefered_lft);
1364         max_desync_factor = min_t(__u32,
1365                                   idev->cnf.max_desync_factor,
1366                                   cnf_temp_preferred_lft - regen_advance);
1367 
1368         if (unlikely(idev->desync_factor > max_desync_factor)) {
1369                 if (max_desync_factor > 0) {
1370                         get_random_bytes(&idev->desync_factor,
1371                                          sizeof(idev->desync_factor));
1372                         idev->desync_factor %= max_desync_factor;
1373                 } else {
1374                         idev->desync_factor = 0;
1375                 }
1376         }
1377 
1378         memset(&cfg, 0, sizeof(cfg));
1379         cfg.valid_lft = min_t(__u32, ifp->valid_lft,
1380                               idev->cnf.temp_valid_lft + age);
1381         cfg.preferred_lft = cnf_temp_preferred_lft + age - idev->desync_factor;
1382         cfg.preferred_lft = min_t(__u32, ifp->prefered_lft, cfg.preferred_lft);
1383 
1384         cfg.plen = ifp->prefix_len;
1385         tmp_tstamp = ifp->tstamp;
1386         spin_unlock_bh(&ifp->lock);
1387 
1388         write_unlock_bh(&idev->lock);
1389 
1390         /* A temporary address is created only if this calculated Preferred
1391          * Lifetime is greater than REGEN_ADVANCE time units.  In particular,
1392          * an implementation must not create a temporary address with a zero
1393          * Preferred Lifetime.
1394          * Use age calculation as in addrconf_verify to avoid unnecessary
1395          * temporary addresses being generated.
1396          */
1397         age = (now - tmp_tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
1398         if (cfg.preferred_lft <= regen_advance + age) {
1399                 in6_ifa_put(ifp);
1400                 in6_dev_put(idev);
1401                 ret = -1;
1402                 goto out;
1403         }
1404 
1405         cfg.ifa_flags = IFA_F_TEMPORARY;
1406         /* set in addrconf_prefix_rcv() */
1407         if (ifp->flags & IFA_F_OPTIMISTIC)
1408                 cfg.ifa_flags |= IFA_F_OPTIMISTIC;
1409 
1410         cfg.pfx = &addr;
1411         cfg.scope = ipv6_addr_scope(cfg.pfx);
1412 
1413         ift = ipv6_add_addr(idev, &cfg, block, NULL);
1414         if (IS_ERR(ift)) {
1415                 in6_ifa_put(ifp);
1416                 in6_dev_put(idev);
1417                 pr_info("%s: retry temporary address regeneration\n", __func__);
1418                 tmpaddr = &addr;
1419                 write_lock_bh(&idev->lock);
1420                 goto retry;
1421         }
1422 
1423         spin_lock_bh(&ift->lock);
1424         ift->ifpub = ifp;
1425         ift->cstamp = now;
1426         ift->tstamp = tmp_tstamp;
1427         spin_unlock_bh(&ift->lock);
1428 
1429         addrconf_dad_start(ift);
1430         in6_ifa_put(ift);
1431         in6_dev_put(idev);
1432 out:
1433         return ret;
1434 }
1435 
1436 /*
1437  *      Choose an appropriate source address (RFC3484)
1438  */
1439 enum {
1440         IPV6_SADDR_RULE_INIT = 0,
1441         IPV6_SADDR_RULE_LOCAL,
1442         IPV6_SADDR_RULE_SCOPE,
1443         IPV6_SADDR_RULE_PREFERRED,
1444 #ifdef CONFIG_IPV6_MIP6
1445         IPV6_SADDR_RULE_HOA,
1446 #endif
1447         IPV6_SADDR_RULE_OIF,
1448         IPV6_SADDR_RULE_LABEL,
1449         IPV6_SADDR_RULE_PRIVACY,
1450         IPV6_SADDR_RULE_ORCHID,
1451         IPV6_SADDR_RULE_PREFIX,
1452 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1453         IPV6_SADDR_RULE_NOT_OPTIMISTIC,
1454 #endif
1455         IPV6_SADDR_RULE_MAX
1456 };
1457 
1458 struct ipv6_saddr_score {
1459         int                     rule;
1460         int                     addr_type;
1461         struct inet6_ifaddr     *ifa;
1462         DECLARE_BITMAP(scorebits, IPV6_SADDR_RULE_MAX);
1463         int                     scopedist;
1464         int                     matchlen;
1465 };
1466 
1467 struct ipv6_saddr_dst {
1468         const struct in6_addr *addr;
1469         int ifindex;
1470         int scope;
1471         int label;
1472         unsigned int prefs;
1473 };
1474 
1475 static inline int ipv6_saddr_preferred(int type)
1476 {
1477         if (type & (IPV6_ADDR_MAPPED|IPV6_ADDR_COMPATv4|IPV6_ADDR_LOOPBACK))
1478                 return 1;
1479         return 0;
1480 }
1481 
1482 static bool ipv6_use_optimistic_addr(struct net *net,
1483                                      struct inet6_dev *idev)
1484 {
1485 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1486         if (!idev)
1487                 return false;
1488         if (!net->ipv6.devconf_all->optimistic_dad && !idev->cnf.optimistic_dad)
1489                 return false;
1490         if (!net->ipv6.devconf_all->use_optimistic && !idev->cnf.use_optimistic)
1491                 return false;
1492 
1493         return true;
1494 #else
1495         return false;
1496 #endif
1497 }
1498 
1499 static bool ipv6_allow_optimistic_dad(struct net *net,
1500                                       struct inet6_dev *idev)
1501 {
1502 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1503         if (!idev)
1504                 return false;
1505         if (!net->ipv6.devconf_all->optimistic_dad && !idev->cnf.optimistic_dad)
1506                 return false;
1507 
1508         return true;
1509 #else
1510         return false;
1511 #endif
1512 }
1513 
1514 static int ipv6_get_saddr_eval(struct net *net,
1515                                struct ipv6_saddr_score *score,
1516                                struct ipv6_saddr_dst *dst,
1517                                int i)
1518 {
1519         int ret;
1520 
1521         if (i <= score->rule) {
1522                 switch (i) {
1523                 case IPV6_SADDR_RULE_SCOPE:
1524                         ret = score->scopedist;
1525                         break;
1526                 case IPV6_SADDR_RULE_PREFIX:
1527                         ret = score->matchlen;
1528                         break;
1529                 default:
1530                         ret = !!test_bit(i, score->scorebits);
1531                 }
1532                 goto out;
1533         }
1534 
1535         switch (i) {
1536         case IPV6_SADDR_RULE_INIT:
1537                 /* Rule 0: remember if hiscore is not ready yet */
1538                 ret = !!score->ifa;
1539                 break;
1540         case IPV6_SADDR_RULE_LOCAL:
1541                 /* Rule 1: Prefer same address */
1542                 ret = ipv6_addr_equal(&score->ifa->addr, dst->addr);
1543                 break;
1544         case IPV6_SADDR_RULE_SCOPE:
1545                 /* Rule 2: Prefer appropriate scope
1546                  *
1547                  *      ret
1548                  *       ^
1549                  *    -1 |  d 15
1550                  *    ---+--+-+---> scope
1551                  *       |
1552                  *       |             d is scope of the destination.
1553                  *  B-d  |  \
1554                  *       |   \      <- smaller scope is better if
1555                  *  B-15 |    \        if scope is enough for destination.
1556                  *       |             ret = B - scope (-1 <= scope >= d <= 15).
1557                  * d-C-1 | /
1558                  *       |/         <- greater is better
1559                  *   -C  /             if scope is not enough for destination.
1560                  *      /|             ret = scope - C (-1 <= d < scope <= 15).
1561                  *
1562                  * d - C - 1 < B -15 (for all -1 <= d <= 15).
1563                  * C > d + 14 - B >= 15 + 14 - B = 29 - B.
1564                  * Assume B = 0 and we get C > 29.
1565                  */
1566                 ret = __ipv6_addr_src_scope(score->addr_type);
1567                 if (ret >= dst->scope)
1568                         ret = -ret;
1569                 else
1570                         ret -= 128;     /* 30 is enough */
1571                 score->scopedist = ret;
1572                 break;
1573         case IPV6_SADDR_RULE_PREFERRED:
1574             {
1575                 /* Rule 3: Avoid deprecated and optimistic addresses */
1576                 u8 avoid = IFA_F_DEPRECATED;
1577 
1578                 if (!ipv6_use_optimistic_addr(net, score->ifa->idev))
1579                         avoid |= IFA_F_OPTIMISTIC;
1580                 ret = ipv6_saddr_preferred(score->addr_type) ||
1581                       !(score->ifa->flags & avoid);
1582                 break;
1583             }
1584 #ifdef CONFIG_IPV6_MIP6
1585         case IPV6_SADDR_RULE_HOA:
1586             {
1587                 /* Rule 4: Prefer home address */
1588                 int prefhome = !(dst->prefs & IPV6_PREFER_SRC_COA);
1589                 ret = !(score->ifa->flags & IFA_F_HOMEADDRESS) ^ prefhome;
1590                 break;
1591             }
1592 #endif
1593         case IPV6_SADDR_RULE_OIF:
1594                 /* Rule 5: Prefer outgoing interface */
1595                 ret = (!dst->ifindex ||
1596                        dst->ifindex == score->ifa->idev->dev->ifindex);
1597                 break;
1598         case IPV6_SADDR_RULE_LABEL:
1599                 /* Rule 6: Prefer matching label */
1600                 ret = ipv6_addr_label(net,
1601                                       &score->ifa->addr, score->addr_type,
1602                                       score->ifa->idev->dev->ifindex) == dst->label;
1603                 break;
1604         case IPV6_SADDR_RULE_PRIVACY:
1605             {
1606                 /* Rule 7: Prefer public address
1607                  * Note: prefer temporary address if use_tempaddr >= 2
1608                  */
1609                 int preftmp = dst->prefs & (IPV6_PREFER_SRC_PUBLIC|IPV6_PREFER_SRC_TMP) ?
1610                                 !!(dst->prefs & IPV6_PREFER_SRC_TMP) :
1611                                 score->ifa->idev->cnf.use_tempaddr >= 2;
1612                 ret = (!(score->ifa->flags & IFA_F_TEMPORARY)) ^ preftmp;
1613                 break;
1614             }
1615         case IPV6_SADDR_RULE_ORCHID:
1616                 /* Rule 8-: Prefer ORCHID vs ORCHID or
1617                  *          non-ORCHID vs non-ORCHID
1618                  */
1619                 ret = !(ipv6_addr_orchid(&score->ifa->addr) ^
1620                         ipv6_addr_orchid(dst->addr));
1621                 break;
1622         case IPV6_SADDR_RULE_PREFIX:
1623                 /* Rule 8: Use longest matching prefix */
1624                 ret = ipv6_addr_diff(&score->ifa->addr, dst->addr);
1625                 if (ret > score->ifa->prefix_len)
1626                         ret = score->ifa->prefix_len;
1627                 score->matchlen = ret;
1628                 break;
1629 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1630         case IPV6_SADDR_RULE_NOT_OPTIMISTIC:
1631                 /* Optimistic addresses still have lower precedence than other
1632                  * preferred addresses.
1633                  */
1634                 ret = !(score->ifa->flags & IFA_F_OPTIMISTIC);
1635                 break;
1636 #endif
1637         default:
1638                 ret = 0;
1639         }
1640 
1641         if (ret)
1642                 __set_bit(i, score->scorebits);
1643         score->rule = i;
1644 out:
1645         return ret;
1646 }
1647 
1648 static int __ipv6_dev_get_saddr(struct net *net,
1649                                 struct ipv6_saddr_dst *dst,
1650                                 struct inet6_dev *idev,
1651                                 struct ipv6_saddr_score *scores,
1652                                 int hiscore_idx)
1653 {
1654         struct ipv6_saddr_score *score = &scores[1 - hiscore_idx], *hiscore = &scores[hiscore_idx];
1655 
1656         list_for_each_entry_rcu(score->ifa, &idev->addr_list, if_list) {
1657                 int i;
1658 
1659                 /*
1660                  * - Tentative Address (RFC2462 section 5.4)
1661                  *  - A tentative address is not considered
1662                  *    "assigned to an interface" in the traditional
1663                  *    sense, unless it is also flagged as optimistic.
1664                  * - Candidate Source Address (section 4)
1665                  *  - In any case, anycast addresses, multicast
1666                  *    addresses, and the unspecified address MUST
1667                  *    NOT be included in a candidate set.
1668                  */
1669                 if ((score->ifa->flags & IFA_F_TENTATIVE) &&
1670                     (!(score->ifa->flags & IFA_F_OPTIMISTIC)))
1671                         continue;
1672 
1673                 score->addr_type = __ipv6_addr_type(&score->ifa->addr);
1674 
1675                 if (unlikely(score->addr_type == IPV6_ADDR_ANY ||
1676                              score->addr_type & IPV6_ADDR_MULTICAST)) {
1677                         net_dbg_ratelimited("ADDRCONF: unspecified / multicast address assigned as unicast address on %s",
1678                                             idev->dev->name);
1679                         continue;
1680                 }
1681 
1682                 score->rule = -1;
1683                 bitmap_zero(score->scorebits, IPV6_SADDR_RULE_MAX);
1684 
1685                 for (i = 0; i < IPV6_SADDR_RULE_MAX; i++) {
1686                         int minihiscore, miniscore;
1687 
1688                         minihiscore = ipv6_get_saddr_eval(net, hiscore, dst, i);
1689                         miniscore = ipv6_get_saddr_eval(net, score, dst, i);
1690 
1691                         if (minihiscore > miniscore) {
1692                                 if (i == IPV6_SADDR_RULE_SCOPE &&
1693                                     score->scopedist > 0) {
1694                                         /*
1695                                          * special case:
1696                                          * each remaining entry
1697                                          * has too small (not enough)
1698                                          * scope, because ifa entries
1699                                          * are sorted by their scope
1700                                          * values.
1701                                          */
1702                                         goto out;
1703                                 }
1704                                 break;
1705                         } else if (minihiscore < miniscore) {
1706                                 swap(hiscore, score);
1707                                 hiscore_idx = 1 - hiscore_idx;
1708 
1709                                 /* restore our iterator */
1710                                 score->ifa = hiscore->ifa;
1711 
1712                                 break;
1713                         }
1714                 }
1715         }
1716 out:
1717         return hiscore_idx;
1718 }
1719 
1720 static int ipv6_get_saddr_master(struct net *net,
1721                                  const struct net_device *dst_dev,
1722                                  const struct net_device *master,
1723                                  struct ipv6_saddr_dst *dst,
1724                                  struct ipv6_saddr_score *scores,
1725                                  int hiscore_idx)
1726 {
1727         struct inet6_dev *idev;
1728 
1729         idev = __in6_dev_get(dst_dev);
1730         if (idev)
1731                 hiscore_idx = __ipv6_dev_get_saddr(net, dst, idev,
1732                                                    scores, hiscore_idx);
1733 
1734         idev = __in6_dev_get(master);
1735         if (idev)
1736                 hiscore_idx = __ipv6_dev_get_saddr(net, dst, idev,
1737                                                    scores, hiscore_idx);
1738 
1739         return hiscore_idx;
1740 }
1741 
1742 int ipv6_dev_get_saddr(struct net *net, const struct net_device *dst_dev,
1743                        const struct in6_addr *daddr, unsigned int prefs,
1744                        struct in6_addr *saddr)
1745 {
1746         struct ipv6_saddr_score scores[2], *hiscore;
1747         struct ipv6_saddr_dst dst;
1748         struct inet6_dev *idev;
1749         struct net_device *dev;
1750         int dst_type;
1751         bool use_oif_addr = false;
1752         int hiscore_idx = 0;
1753         int ret = 0;
1754 
1755         dst_type = __ipv6_addr_type(daddr);
1756         dst.addr = daddr;
1757         dst.ifindex = dst_dev ? dst_dev->ifindex : 0;
1758         dst.scope = __ipv6_addr_src_scope(dst_type);
1759         dst.label = ipv6_addr_label(net, daddr, dst_type, dst.ifindex);
1760         dst.prefs = prefs;
1761 
1762         scores[hiscore_idx].rule = -1;
1763         scores[hiscore_idx].ifa = NULL;
1764 
1765         rcu_read_lock();
1766 
1767         /* Candidate Source Address (section 4)
1768          *  - multicast and link-local destination address,
1769          *    the set of candidate source address MUST only
1770          *    include addresses assigned to interfaces
1771          *    belonging to the same link as the outgoing
1772          *    interface.
1773          * (- For site-local destination addresses, the
1774          *    set of candidate source addresses MUST only
1775          *    include addresses assigned to interfaces
1776          *    belonging to the same site as the outgoing
1777          *    interface.)
1778          *  - "It is RECOMMENDED that the candidate source addresses
1779          *    be the set of unicast addresses assigned to the
1780          *    interface that will be used to send to the destination
1781          *    (the 'outgoing' interface)." (RFC 6724)
1782          */
1783         if (dst_dev) {
1784                 idev = __in6_dev_get(dst_dev);
1785                 if ((dst_type & IPV6_ADDR_MULTICAST) ||
1786                     dst.scope <= IPV6_ADDR_SCOPE_LINKLOCAL ||
1787                     (idev && idev->cnf.use_oif_addrs_only)) {
1788                         use_oif_addr = true;
1789                 }
1790         }
1791 
1792         if (use_oif_addr) {
1793                 if (idev)
1794                         hiscore_idx = __ipv6_dev_get_saddr(net, &dst, idev, scores, hiscore_idx);
1795         } else {
1796                 const struct net_device *master;
1797                 int master_idx = 0;
1798 
1799                 /* if dst_dev exists and is enslaved to an L3 device, then
1800                  * prefer addresses from dst_dev and then the master over
1801                  * any other enslaved devices in the L3 domain.
1802                  */
1803                 master = l3mdev_master_dev_rcu(dst_dev);
1804                 if (master) {
1805                         master_idx = master->ifindex;
1806 
1807                         hiscore_idx = ipv6_get_saddr_master(net, dst_dev,
1808                                                             master, &dst,
1809                                                             scores, hiscore_idx);
1810 
1811                         if (scores[hiscore_idx].ifa)
1812                                 goto out;
1813                 }
1814 
1815                 for_each_netdev_rcu(net, dev) {
1816                         /* only consider addresses on devices in the
1817                          * same L3 domain
1818                          */
1819                         if (l3mdev_master_ifindex_rcu(dev) != master_idx)
1820                                 continue;
1821                         idev = __in6_dev_get(dev);
1822                         if (!idev)
1823                                 continue;
1824                         hiscore_idx = __ipv6_dev_get_saddr(net, &dst, idev, scores, hiscore_idx);
1825                 }
1826         }
1827 
1828 out:
1829         hiscore = &scores[hiscore_idx];
1830         if (!hiscore->ifa)
1831                 ret = -EADDRNOTAVAIL;
1832         else
1833                 *saddr = hiscore->ifa->addr;
1834 
1835         rcu_read_unlock();
1836         return ret;
1837 }
1838 EXPORT_SYMBOL(ipv6_dev_get_saddr);
1839 
1840 int __ipv6_get_lladdr(struct inet6_dev *idev, struct in6_addr *addr,
1841                       u32 banned_flags)
1842 {
1843         struct inet6_ifaddr *ifp;
1844         int err = -EADDRNOTAVAIL;
1845 
1846         list_for_each_entry_reverse(ifp, &idev->addr_list, if_list) {
1847                 if (ifp->scope > IFA_LINK)
1848                         break;
1849                 if (ifp->scope == IFA_LINK &&
1850                     !(ifp->flags & banned_flags)) {
1851                         *addr = ifp->addr;
1852                         err = 0;
1853                         break;
1854                 }
1855         }
1856         return err;
1857 }
1858 
1859 int ipv6_get_lladdr(struct net_device *dev, struct in6_addr *addr,
1860                     u32 banned_flags)
1861 {
1862         struct inet6_dev *idev;
1863         int err = -EADDRNOTAVAIL;
1864 
1865         rcu_read_lock();
1866         idev = __in6_dev_get(dev);
1867         if (idev) {
1868                 read_lock_bh(&idev->lock);
1869                 err = __ipv6_get_lladdr(idev, addr, banned_flags);
1870                 read_unlock_bh(&idev->lock);
1871         }
1872         rcu_read_unlock();
1873         return err;
1874 }
1875 
1876 static int ipv6_count_addresses(const struct inet6_dev *idev)
1877 {
1878         const struct inet6_ifaddr *ifp;
1879         int cnt = 0;
1880 
1881         rcu_read_lock();
1882         list_for_each_entry_rcu(ifp, &idev->addr_list, if_list)
1883                 cnt++;
1884         rcu_read_unlock();
1885         return cnt;
1886 }
1887 
1888 int ipv6_chk_addr(struct net *net, const struct in6_addr *addr,
1889                   const struct net_device *dev, int strict)
1890 {
1891         return ipv6_chk_addr_and_flags(net, addr, dev, !dev,
1892                                        strict, IFA_F_TENTATIVE);
1893 }
1894 EXPORT_SYMBOL(ipv6_chk_addr);
1895 
1896 /* device argument is used to find the L3 domain of interest. If
1897  * skip_dev_check is set, then the ifp device is not checked against
1898  * the passed in dev argument. So the 2 cases for addresses checks are:
1899  *   1. does the address exist in the L3 domain that dev is part of
1900  *      (skip_dev_check = true), or
1901  *
1902  *   2. does the address exist on the specific device
1903  *      (skip_dev_check = false)
1904  */
1905 int ipv6_chk_addr_and_flags(struct net *net, const struct in6_addr *addr,
1906                             const struct net_device *dev, bool skip_dev_check,
1907                             int strict, u32 banned_flags)
1908 {
1909         unsigned int hash = inet6_addr_hash(net, addr);
1910         const struct net_device *l3mdev;
1911         struct inet6_ifaddr *ifp;
1912         u32 ifp_flags;
1913 
1914         rcu_read_lock();
1915 
1916         l3mdev = l3mdev_master_dev_rcu(dev);
1917         if (skip_dev_check)
1918                 dev = NULL;
1919 
1920         hlist_for_each_entry_rcu(ifp, &inet6_addr_lst[hash], addr_lst) {
1921                 if (!net_eq(dev_net(ifp->idev->dev), net))
1922                         continue;
1923 
1924                 if (l3mdev_master_dev_rcu(ifp->idev->dev) != l3mdev)
1925                         continue;
1926 
1927                 /* Decouple optimistic from tentative for evaluation here.
1928                  * Ban optimistic addresses explicitly, when required.
1929                  */
1930                 ifp_flags = (ifp->flags&IFA_F_OPTIMISTIC)
1931                             ? (ifp->flags&~IFA_F_TENTATIVE)
1932                             : ifp->flags;
1933                 if (ipv6_addr_equal(&ifp->addr, addr) &&
1934                     !(ifp_flags&banned_flags) &&
1935                     (!dev || ifp->idev->dev == dev ||
1936                      !(ifp->scope&(IFA_LINK|IFA_HOST) || strict))) {
1937                         rcu_read_unlock();
1938                         return 1;
1939                 }
1940         }
1941 
1942         rcu_read_unlock();
1943         return 0;
1944 }
1945 EXPORT_SYMBOL(ipv6_chk_addr_and_flags);
1946 
1947 
1948 /* Compares an address/prefix_len with addresses on device @dev.
1949  * If one is found it returns true.
1950  */
1951 bool ipv6_chk_custom_prefix(const struct in6_addr *addr,
1952         const unsigned int prefix_len, struct net_device *dev)
1953 {
1954         const struct inet6_ifaddr *ifa;
1955         const struct inet6_dev *idev;
1956         bool ret = false;
1957 
1958         rcu_read_lock();
1959         idev = __in6_dev_get(dev);
1960         if (idev) {
1961                 list_for_each_entry_rcu(ifa, &idev->addr_list, if_list) {
1962                         ret = ipv6_prefix_equal(addr, &ifa->addr, prefix_len);
1963                         if (ret)
1964                                 break;
1965                 }
1966         }
1967         rcu_read_unlock();
1968 
1969         return ret;
1970 }
1971 EXPORT_SYMBOL(ipv6_chk_custom_prefix);
1972 
1973 int ipv6_chk_prefix(const struct in6_addr *addr, struct net_device *dev)
1974 {
1975         const struct inet6_ifaddr *ifa;
1976         const struct inet6_dev *idev;
1977         int     onlink;
1978 
1979         onlink = 0;
1980         rcu_read_lock();
1981         idev = __in6_dev_get(dev);
1982         if (idev) {
1983                 list_for_each_entry_rcu(ifa, &idev->addr_list, if_list) {
1984                         onlink = ipv6_prefix_equal(addr, &ifa->addr,
1985                                                    ifa->prefix_len);
1986                         if (onlink)
1987                                 break;
1988                 }
1989         }
1990         rcu_read_unlock();
1991         return onlink;
1992 }
1993 EXPORT_SYMBOL(ipv6_chk_prefix);
1994 
1995 struct inet6_ifaddr *ipv6_get_ifaddr(struct net *net, const struct in6_addr *addr,
1996                                      struct net_device *dev, int strict)
1997 {
1998         unsigned int hash = inet6_addr_hash(net, addr);
1999         struct inet6_ifaddr *ifp, *result = NULL;
2000 
2001         rcu_read_lock();
2002         hlist_for_each_entry_rcu(ifp, &inet6_addr_lst[hash], addr_lst) {
2003                 if (!net_eq(dev_net(ifp->idev->dev), net))
2004                         continue;
2005                 if (ipv6_addr_equal(&ifp->addr, addr)) {
2006                         if (!dev || ifp->idev->dev == dev ||
2007                             !(ifp->scope&(IFA_LINK|IFA_HOST) || strict)) {
2008                                 result = ifp;
2009                                 in6_ifa_hold(ifp);
2010                                 break;
2011                         }
2012                 }
2013         }
2014         rcu_read_unlock();
2015 
2016         return result;
2017 }
2018 
2019 /* Gets referenced address, destroys ifaddr */
2020 
2021 static void addrconf_dad_stop(struct inet6_ifaddr *ifp, int dad_failed)
2022 {
2023         if (dad_failed)
2024                 ifp->flags |= IFA_F_DADFAILED;
2025 
2026         if (ifp->flags&IFA_F_TEMPORARY) {
2027                 struct inet6_ifaddr *ifpub;
2028                 spin_lock_bh(&ifp->lock);
2029                 ifpub = ifp->ifpub;
2030                 if (ifpub) {
2031                         in6_ifa_hold(ifpub);
2032                         spin_unlock_bh(&ifp->lock);
2033                         ipv6_create_tempaddr(ifpub, ifp, true);
2034                         in6_ifa_put(ifpub);
2035                 } else {
2036                         spin_unlock_bh(&ifp->lock);
2037                 }
2038                 ipv6_del_addr(ifp);
2039         } else if (ifp->flags&IFA_F_PERMANENT || !dad_failed) {
2040                 spin_lock_bh(&ifp->lock);
2041                 addrconf_del_dad_work(ifp);
2042                 ifp->flags |= IFA_F_TENTATIVE;
2043                 if (dad_failed)
2044                         ifp->flags &= ~IFA_F_OPTIMISTIC;
2045                 spin_unlock_bh(&ifp->lock);
2046                 if (dad_failed)
2047                         ipv6_ifa_notify(0, ifp);
2048                 in6_ifa_put(ifp);
2049         } else {
2050                 ipv6_del_addr(ifp);
2051         }
2052 }
2053 
2054 static int addrconf_dad_end(struct inet6_ifaddr *ifp)
2055 {
2056         int err = -ENOENT;
2057 
2058         spin_lock_bh(&ifp->lock);
2059         if (ifp->state == INET6_IFADDR_STATE_DAD) {
2060                 ifp->state = INET6_IFADDR_STATE_POSTDAD;
2061                 err = 0;
2062         }
2063         spin_unlock_bh(&ifp->lock);
2064 
2065         return err;
2066 }
2067 
2068 void addrconf_dad_failure(struct sk_buff *skb, struct inet6_ifaddr *ifp)
2069 {
2070         struct inet6_dev *idev = ifp->idev;
2071         struct net *net = dev_net(ifp->idev->dev);
2072 
2073         if (addrconf_dad_end(ifp)) {
2074                 in6_ifa_put(ifp);
2075                 return;
2076         }
2077 
2078         net_info_ratelimited("%s: IPv6 duplicate address %pI6c used by %pM detected!\n",
2079                              ifp->idev->dev->name, &ifp->addr, eth_hdr(skb)->h_source);
2080 
2081         spin_lock_bh(&ifp->lock);
2082 
2083         if (ifp->flags & IFA_F_STABLE_PRIVACY) {
2084                 struct in6_addr new_addr;
2085                 struct inet6_ifaddr *ifp2;
2086                 int retries = ifp->stable_privacy_retry + 1;
2087                 struct ifa6_config cfg = {
2088                         .pfx = &new_addr,
2089                         .plen = ifp->prefix_len,
2090                         .ifa_flags = ifp->flags,
2091                         .valid_lft = ifp->valid_lft,
2092                         .preferred_lft = ifp->prefered_lft,
2093                         .scope = ifp->scope,
2094                 };
2095 
2096                 if (retries > net->ipv6.sysctl.idgen_retries) {
2097                         net_info_ratelimited("%s: privacy stable address generation failed because of DAD conflicts!\n",
2098                                              ifp->idev->dev->name);
2099                         goto errdad;
2100                 }
2101 
2102                 new_addr = ifp->addr;
2103                 if (ipv6_generate_stable_address(&new_addr, retries,
2104                                                  idev))
2105                         goto errdad;
2106 
2107                 spin_unlock_bh(&ifp->lock);
2108 
2109                 if (idev->cnf.max_addresses &&
2110                     ipv6_count_addresses(idev) >=
2111                     idev->cnf.max_addresses)
2112                         goto lock_errdad;
2113 
2114                 net_info_ratelimited("%s: generating new stable privacy address because of DAD conflict\n",
2115                                      ifp->idev->dev->name);
2116 
2117                 ifp2 = ipv6_add_addr(idev, &cfg, false, NULL);
2118                 if (IS_ERR(ifp2))
2119                         goto lock_errdad;
2120 
2121                 spin_lock_bh(&ifp2->lock);
2122                 ifp2->stable_privacy_retry = retries;
2123                 ifp2->state = INET6_IFADDR_STATE_PREDAD;
2124                 spin_unlock_bh(&ifp2->lock);
2125 
2126                 addrconf_mod_dad_work(ifp2, net->ipv6.sysctl.idgen_delay);
2127                 in6_ifa_put(ifp2);
2128 lock_errdad:
2129                 spin_lock_bh(&ifp->lock);
2130         }
2131 
2132 errdad:
2133         /* transition from _POSTDAD to _ERRDAD */
2134         ifp->state = INET6_IFADDR_STATE_ERRDAD;
2135         spin_unlock_bh(&ifp->lock);
2136 
2137         addrconf_mod_dad_work(ifp, 0);
2138         in6_ifa_put(ifp);
2139 }
2140 
2141 /* Join to solicited addr multicast group.
2142  * caller must hold RTNL */
2143 void addrconf_join_solict(struct net_device *dev, const struct in6_addr *addr)
2144 {
2145         struct in6_addr maddr;
2146 
2147         if (dev->flags&(IFF_LOOPBACK|IFF_NOARP))
2148                 return;
2149 
2150         addrconf_addr_solict_mult(addr, &maddr);
2151         ipv6_dev_mc_inc(dev, &maddr);
2152 }
2153 
2154 /* caller must hold RTNL */
2155 void addrconf_leave_solict(struct inet6_dev *idev, const struct in6_addr *addr)
2156 {
2157         struct in6_addr maddr;
2158 
2159         if (idev->dev->flags&(IFF_LOOPBACK|IFF_NOARP))
2160                 return;
2161 
2162         addrconf_addr_solict_mult(addr, &maddr);
2163         __ipv6_dev_mc_dec(idev, &maddr);
2164 }
2165 
2166 /* caller must hold RTNL */
2167 static void addrconf_join_anycast(struct inet6_ifaddr *ifp)
2168 {
2169         struct in6_addr addr;
2170 
2171         if (ifp->prefix_len >= 127) /* RFC 6164 */
2172                 return;
2173         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
2174         if (ipv6_addr_any(&addr))
2175                 return;
2176         __ipv6_dev_ac_inc(ifp->idev, &addr);
2177 }
2178 
2179 /* caller must hold RTNL */
2180 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp)
2181 {
2182         struct in6_addr addr;
2183 
2184         if (ifp->prefix_len >= 127) /* RFC 6164 */
2185                 return;
2186         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
2187         if (ipv6_addr_any(&addr))
2188                 return;
2189         __ipv6_dev_ac_dec(ifp->idev, &addr);
2190 }
2191 
2192 static int addrconf_ifid_6lowpan(u8 *eui, struct net_device *dev)
2193 {
2194         switch (dev->addr_len) {
2195         case ETH_ALEN:
2196                 memcpy(eui, dev->dev_addr, 3);
2197                 eui[3] = 0xFF;
2198                 eui[4] = 0xFE;
2199                 memcpy(eui + 5, dev->dev_addr + 3, 3);
2200                 break;
2201         case EUI64_ADDR_LEN:
2202                 memcpy(eui, dev->dev_addr, EUI64_ADDR_LEN);
2203                 eui[0] ^= 2;
2204                 break;
2205         default:
2206                 return -1;
2207         }
2208 
2209         return 0;
2210 }
2211 
2212 static int addrconf_ifid_ieee1394(u8 *eui, struct net_device *dev)
2213 {
2214         union fwnet_hwaddr *ha;
2215 
2216         if (dev->addr_len != FWNET_ALEN)
2217                 return -1;
2218 
2219         ha = (union fwnet_hwaddr *)dev->dev_addr;
2220 
2221         memcpy(eui, &ha->uc.uniq_id, sizeof(ha->uc.uniq_id));
2222         eui[0] ^= 2;
2223         return 0;
2224 }
2225 
2226 static int addrconf_ifid_arcnet(u8 *eui, struct net_device *dev)
2227 {
2228         /* XXX: inherit EUI-64 from other interface -- yoshfuji */
2229         if (dev->addr_len != ARCNET_ALEN)
2230                 return -1;
2231         memset(eui, 0, 7);
2232         eui[7] = *(u8 *)dev->dev_addr;
2233         return 0;
2234 }
2235 
2236 static int addrconf_ifid_infiniband(u8 *eui, struct net_device *dev)
2237 {
2238         if (dev->addr_len != INFINIBAND_ALEN)
2239                 return -1;
2240         memcpy(eui, dev->dev_addr + 12, 8);
2241         eui[0] |= 2;
2242         return 0;
2243 }
2244 
2245 static int __ipv6_isatap_ifid(u8 *eui, __be32 addr)
2246 {
2247         if (addr == 0)
2248                 return -1;
2249         eui[0] = (ipv4_is_zeronet(addr) || ipv4_is_private_10(addr) ||
2250                   ipv4_is_loopback(addr) || ipv4_is_linklocal_169(addr) ||
2251                   ipv4_is_private_172(addr) || ipv4_is_test_192(addr) ||
2252                   ipv4_is_anycast_6to4(addr) || ipv4_is_private_192(addr) ||
2253                   ipv4_is_test_198(addr) || ipv4_is_multicast(addr) ||
2254                   ipv4_is_lbcast(addr)) ? 0x00 : 0x02;
2255         eui[1] = 0;
2256         eui[2] = 0x5E;
2257         eui[3] = 0xFE;
2258         memcpy(eui + 4, &addr, 4);
2259         return 0;
2260 }
2261 
2262 static int addrconf_ifid_sit(u8 *eui, struct net_device *dev)
2263 {
2264         if (dev->priv_flags & IFF_ISATAP)
2265                 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
2266         return -1;
2267 }
2268 
2269 static int addrconf_ifid_gre(u8 *eui, struct net_device *dev)
2270 {
2271         return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
2272 }
2273 
2274 static int addrconf_ifid_ip6tnl(u8 *eui, struct net_device *dev)
2275 {
2276         memcpy(eui, dev->perm_addr, 3);
2277         memcpy(eui + 5, dev->perm_addr + 3, 3);
2278         eui[3] = 0xFF;
2279         eui[4] = 0xFE;
2280         eui[0] ^= 2;
2281         return 0;
2282 }
2283 
2284 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev)
2285 {
2286         switch (dev->type) {
2287         case ARPHRD_ETHER:
2288         case ARPHRD_FDDI:
2289                 return addrconf_ifid_eui48(eui, dev);
2290         case ARPHRD_ARCNET:
2291                 return addrconf_ifid_arcnet(eui, dev);
2292         case ARPHRD_INFINIBAND:
2293                 return addrconf_ifid_infiniband(eui, dev);
2294         case ARPHRD_SIT:
2295                 return addrconf_ifid_sit(eui, dev);
2296         case ARPHRD_IPGRE:
2297         case ARPHRD_TUNNEL:
2298                 return addrconf_ifid_gre(eui, dev);
2299         case ARPHRD_6LOWPAN:
2300                 return addrconf_ifid_6lowpan(eui, dev);
2301         case ARPHRD_IEEE1394:
2302                 return addrconf_ifid_ieee1394(eui, dev);
2303         case ARPHRD_TUNNEL6:
2304         case ARPHRD_IP6GRE:
2305         case ARPHRD_RAWIP:
2306                 return addrconf_ifid_ip6tnl(eui, dev);
2307         }
2308         return -1;
2309 }
2310 
2311 static int ipv6_inherit_eui64(u8 *eui, struct inet6_dev *idev)
2312 {
2313         int err = -1;
2314         struct inet6_ifaddr *ifp;
2315 
2316         read_lock_bh(&idev->lock);
2317         list_for_each_entry_reverse(ifp, &idev->addr_list, if_list) {
2318                 if (ifp->scope > IFA_LINK)
2319                         break;
2320                 if (ifp->scope == IFA_LINK && !(ifp->flags&IFA_F_TENTATIVE)) {
2321                         memcpy(eui, ifp->addr.s6_addr+8, 8);
2322                         err = 0;
2323                         break;
2324                 }
2325         }
2326         read_unlock_bh(&idev->lock);
2327         return err;
2328 }
2329 
2330 /* (re)generation of randomized interface identifier (RFC 3041 3.2, 3.5) */
2331 static void ipv6_regen_rndid(struct inet6_dev *idev)
2332 {
2333 regen:
2334         get_random_bytes(idev->rndid, sizeof(idev->rndid));
2335         idev->rndid[0] &= ~0x02;
2336 
2337         /*
2338          * <draft-ietf-ipngwg-temp-addresses-v2-00.txt>:
2339          * check if generated address is not inappropriate
2340          *
2341          *  - Reserved subnet anycast (RFC 2526)
2342          *      11111101 11....11 1xxxxxxx
2343          *  - ISATAP (RFC4214) 6.1
2344          *      00-00-5E-FE-xx-xx-xx-xx
2345          *  - value 0
2346          *  - XXX: already assigned to an address on the device
2347          */
2348         if (idev->rndid[0] == 0xfd &&
2349             (idev->rndid[1]&idev->rndid[2]&idev->rndid[3]&idev->rndid[4]&idev->rndid[5]&idev->rndid[6]) == 0xff &&
2350             (idev->rndid[7]&0x80))
2351                 goto regen;
2352         if ((idev->rndid[0]|idev->rndid[1]) == 0) {
2353                 if (idev->rndid[2] == 0x5e && idev->rndid[3] == 0xfe)
2354                         goto regen;
2355                 if ((idev->rndid[2]|idev->rndid[3]|idev->rndid[4]|idev->rndid[5]|idev->rndid[6]|idev->rndid[7]) == 0x00)
2356                         goto regen;
2357         }
2358 }
2359 
2360 static void  ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr)
2361 {
2362         if (tmpaddr && memcmp(idev->rndid, &tmpaddr->s6_addr[8], 8) == 0)
2363                 ipv6_regen_rndid(idev);
2364 }
2365 
2366 /*
2367  *      Add prefix route.
2368  */
2369 
2370 static void
2371 addrconf_prefix_route(struct in6_addr *pfx, int plen, u32 metric,
2372                       struct net_device *dev, unsigned long expires,
2373                       u32 flags, gfp_t gfp_flags)
2374 {
2375         struct fib6_config cfg = {
2376                 .fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_PREFIX,
2377                 .fc_metric = metric ? : IP6_RT_PRIO_ADDRCONF,
2378                 .fc_ifindex = dev->ifindex,
2379                 .fc_expires = expires,
2380                 .fc_dst_len = plen,
2381                 .fc_flags = RTF_UP | flags,
2382                 .fc_nlinfo.nl_net = dev_net(dev),
2383                 .fc_protocol = RTPROT_KERNEL,
2384                 .fc_type = RTN_UNICAST,
2385         };
2386 
2387         cfg.fc_dst = *pfx;
2388 
2389         /* Prevent useless cloning on PtP SIT.
2390            This thing is done here expecting that the whole
2391            class of non-broadcast devices need not cloning.
2392          */
2393 #if IS_ENABLED(CONFIG_IPV6_SIT)
2394         if (dev->type == ARPHRD_SIT && (dev->flags & IFF_POINTOPOINT))
2395                 cfg.fc_flags |= RTF_NONEXTHOP;
2396 #endif
2397 
2398         ip6_route_add(&cfg, gfp_flags, NULL);
2399 }
2400 
2401 
2402 static struct fib6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
2403                                                   int plen,
2404                                                   const struct net_device *dev,
2405                                                   u32 flags, u32 noflags,
2406                                                   bool no_gw)
2407 {
2408         struct fib6_node *fn;
2409         struct fib6_info *rt = NULL;
2410         struct fib6_table *table;
2411         u32 tb_id = l3mdev_fib_table(dev) ? : RT6_TABLE_PREFIX;
2412 
2413         table = fib6_get_table(dev_net(dev), tb_id);
2414         if (!table)
2415                 return NULL;
2416 
2417         rcu_read_lock();
2418         fn = fib6_locate(&table->tb6_root, pfx, plen, NULL, 0, true);
2419         if (!fn)
2420                 goto out;
2421 
2422         for_each_fib6_node_rt_rcu(fn) {
2423                 /* prefix routes only use builtin fib6_nh */
2424                 if (rt->nh)
2425                         continue;
2426 
2427                 if (rt->fib6_nh->fib_nh_dev->ifindex != dev->ifindex)
2428                         continue;
2429                 if (no_gw && rt->fib6_nh->fib_nh_gw_family)
2430                         continue;
2431                 if ((rt->fib6_flags & flags) != flags)
2432                         continue;
2433                 if ((rt->fib6_flags & noflags) != 0)
2434                         continue;
2435                 if (!fib6_info_hold_safe(rt))
2436                         continue;
2437                 break;
2438         }
2439 out:
2440         rcu_read_unlock();
2441         return rt;
2442 }
2443 
2444 
2445 /* Create "default" multicast route to the interface */
2446 
2447 static void addrconf_add_mroute(struct net_device *dev)
2448 {
2449         struct fib6_config cfg = {
2450                 .fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_LOCAL,
2451                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
2452                 .fc_ifindex = dev->ifindex,
2453                 .fc_dst_len = 8,
2454                 .fc_flags = RTF_UP,
2455                 .fc_type = RTN_UNICAST,
2456                 .fc_nlinfo.nl_net = dev_net(dev),
2457         };
2458 
2459         ipv6_addr_set(&cfg.fc_dst, htonl(0xFF000000), 0, 0, 0);
2460 
2461         ip6_route_add(&cfg, GFP_KERNEL, NULL);
2462 }
2463 
2464 static struct inet6_dev *addrconf_add_dev(struct net_device *dev)
2465 {
2466         struct inet6_dev *idev;
2467 
2468         ASSERT_RTNL();
2469 
2470         idev = ipv6_find_idev(dev);
2471         if (IS_ERR(idev))
2472                 return idev;
2473 
2474         if (idev->cnf.disable_ipv6)
2475                 return ERR_PTR(-EACCES);
2476 
2477         /* Add default multicast route */
2478         if (!(dev->flags & IFF_LOOPBACK) && !netif_is_l3_master(dev))
2479                 addrconf_add_mroute(dev);
2480 
2481         return idev;
2482 }
2483 
2484 static void manage_tempaddrs(struct inet6_dev *idev,
2485                              struct inet6_ifaddr *ifp,
2486                              __u32 valid_lft, __u32 prefered_lft,
2487                              bool create, unsigned long now)
2488 {
2489         u32 flags;
2490         struct inet6_ifaddr *ift;
2491 
2492         read_lock_bh(&idev->lock);
2493         /* update all temporary addresses in the list */
2494         list_for_each_entry(ift, &idev->tempaddr_list, tmp_list) {
2495                 int age, max_valid, max_prefered;
2496 
2497                 if (ifp != ift->ifpub)
2498                         continue;
2499 
2500                 /* RFC 4941 section 3.3:
2501                  * If a received option will extend the lifetime of a public
2502                  * address, the lifetimes of temporary addresses should
2503                  * be extended, subject to the overall constraint that no
2504                  * temporary addresses should ever remain "valid" or "preferred"
2505                  * for a time longer than (TEMP_VALID_LIFETIME) or
2506                  * (TEMP_PREFERRED_LIFETIME - DESYNC_FACTOR), respectively.
2507                  */
2508                 age = (now - ift->cstamp) / HZ;
2509                 max_valid = idev->cnf.temp_valid_lft - age;
2510                 if (max_valid < 0)
2511                         max_valid = 0;
2512 
2513                 max_prefered = idev->cnf.temp_prefered_lft -
2514                                idev->desync_factor - age;
2515                 if (max_prefered < 0)
2516                         max_prefered = 0;
2517 
2518                 if (valid_lft > max_valid)
2519                         valid_lft = max_valid;
2520 
2521                 if (prefered_lft > max_prefered)
2522                         prefered_lft = max_prefered;
2523 
2524                 spin_lock(&ift->lock);
2525                 flags = ift->flags;
2526                 ift->valid_lft = valid_lft;
2527                 ift->prefered_lft = prefered_lft;
2528                 ift->tstamp = now;
2529                 if (prefered_lft > 0)
2530                         ift->flags &= ~IFA_F_DEPRECATED;
2531 
2532                 spin_unlock(&ift->lock);
2533                 if (!(flags&IFA_F_TENTATIVE))
2534                         ipv6_ifa_notify(0, ift);
2535         }
2536 
2537         if ((create || list_empty(&idev->tempaddr_list)) &&
2538             idev->cnf.use_tempaddr > 0) {
2539                 /* When a new public address is created as described
2540                  * in [ADDRCONF], also create a new temporary address.
2541                  * Also create a temporary address if it's enabled but
2542                  * no temporary address currently exists.
2543                  */
2544                 read_unlock_bh(&idev->lock);
2545                 ipv6_create_tempaddr(ifp, NULL, false);
2546         } else {
2547                 read_unlock_bh(&idev->lock);
2548         }
2549 }
2550 
2551 static bool is_addr_mode_generate_stable(struct inet6_dev *idev)
2552 {
2553         return idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_STABLE_PRIVACY ||
2554                idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_RANDOM;
2555 }
2556 
2557 int addrconf_prefix_rcv_add_addr(struct net *net, struct net_device *dev,
2558                                  const struct prefix_info *pinfo,
2559                                  struct inet6_dev *in6_dev,
2560                                  const struct in6_addr *addr, int addr_type,
2561                                  u32 addr_flags, bool sllao, bool tokenized,
2562                                  __u32 valid_lft, u32 prefered_lft)
2563 {
2564         struct inet6_ifaddr *ifp = ipv6_get_ifaddr(net, addr, dev, 1);
2565         int create = 0, update_lft = 0;
2566 
2567         if (!ifp && valid_lft) {
2568                 int max_addresses = in6_dev->cnf.max_addresses;
2569                 struct ifa6_config cfg = {
2570                         .pfx = addr,
2571                         .plen = pinfo->prefix_len,
2572                         .ifa_flags = addr_flags,
2573                         .valid_lft = valid_lft,
2574                         .preferred_lft = prefered_lft,
2575                         .scope = addr_type & IPV6_ADDR_SCOPE_MASK,
2576                 };
2577 
2578 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2579                 if ((net->ipv6.devconf_all->optimistic_dad ||
2580                      in6_dev->cnf.optimistic_dad) &&
2581                     !net->ipv6.devconf_all->forwarding && sllao)
2582                         cfg.ifa_flags |= IFA_F_OPTIMISTIC;
2583 #endif
2584 
2585                 /* Do not allow to create too much of autoconfigured
2586                  * addresses; this would be too easy way to crash kernel.
2587                  */
2588                 if (!max_addresses ||
2589                     ipv6_count_addresses(in6_dev) < max_addresses)
2590                         ifp = ipv6_add_addr(in6_dev, &cfg, false, NULL);
2591 
2592                 if (IS_ERR_OR_NULL(ifp))
2593                         return -1;
2594 
2595                 create = 1;
2596                 spin_lock_bh(&ifp->lock);
2597                 ifp->flags |= IFA_F_MANAGETEMPADDR;
2598                 ifp->cstamp = jiffies;
2599                 ifp->tokenized = tokenized;
2600                 spin_unlock_bh(&ifp->lock);
2601                 addrconf_dad_start(ifp);
2602         }
2603 
2604         if (ifp) {
2605                 u32 flags;
2606                 unsigned long now;
2607                 u32 stored_lft;
2608 
2609                 /* update lifetime (RFC2462 5.5.3 e) */
2610                 spin_lock_bh(&ifp->lock);
2611                 now = jiffies;
2612                 if (ifp->valid_lft > (now - ifp->tstamp) / HZ)
2613                         stored_lft = ifp->valid_lft - (now - ifp->tstamp) / HZ;
2614                 else
2615                         stored_lft = 0;
2616                 if (!create && stored_lft) {
2617                         const u32 minimum_lft = min_t(u32,
2618                                 stored_lft, MIN_VALID_LIFETIME);
2619                         valid_lft = max(valid_lft, minimum_lft);
2620 
2621                         /* RFC4862 Section 5.5.3e:
2622                          * "Note that the preferred lifetime of the
2623                          *  corresponding address is always reset to
2624                          *  the Preferred Lifetime in the received
2625                          *  Prefix Information option, regardless of
2626                          *  whether the valid lifetime is also reset or
2627                          *  ignored."
2628                          *
2629                          * So we should always update prefered_lft here.
2630                          */
2631                         update_lft = 1;
2632                 }
2633 
2634                 if (update_lft) {
2635                         ifp->valid_lft = valid_lft;
2636                         ifp->prefered_lft = prefered_lft;
2637                         ifp->tstamp = now;
2638                         flags = ifp->flags;
2639                         ifp->flags &= ~IFA_F_DEPRECATED;
2640                         spin_unlock_bh(&ifp->lock);
2641 
2642                         if (!(flags&IFA_F_TENTATIVE))
2643                                 ipv6_ifa_notify(0, ifp);
2644                 } else
2645                         spin_unlock_bh(&ifp->lock);
2646 
2647                 manage_tempaddrs(in6_dev, ifp, valid_lft, prefered_lft,
2648                                  create, now);
2649 
2650                 in6_ifa_put(ifp);
2651                 addrconf_verify();
2652         }
2653 
2654         return 0;
2655 }
2656 EXPORT_SYMBOL_GPL(addrconf_prefix_rcv_add_addr);
2657 
2658 void addrconf_prefix_rcv(struct net_device *dev, u8 *opt, int len, bool sllao)
2659 {
2660         struct prefix_info *pinfo;
2661         __u32 valid_lft;
2662         __u32 prefered_lft;
2663         int addr_type, err;
2664         u32 addr_flags = 0;
2665         struct inet6_dev *in6_dev;
2666         struct net *net = dev_net(dev);
2667 
2668         pinfo = (struct prefix_info *) opt;
2669 
2670         if (len < sizeof(struct prefix_info)) {
2671                 netdev_dbg(dev, "addrconf: prefix option too short\n");
2672                 return;
2673         }
2674 
2675         /*
2676          *      Validation checks ([ADDRCONF], page 19)
2677          */
2678 
2679         addr_type = ipv6_addr_type(&pinfo->prefix);
2680 
2681         if (addr_type & (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL))
2682                 return;
2683 
2684         valid_lft = ntohl(pinfo->valid);
2685         prefered_lft = ntohl(pinfo->prefered);
2686 
2687         if (prefered_lft > valid_lft) {
2688                 net_warn_ratelimited("addrconf: prefix option has invalid lifetime\n");
2689                 return;
2690         }
2691 
2692         in6_dev = in6_dev_get(dev);
2693 
2694         if (!in6_dev) {
2695                 net_dbg_ratelimited("addrconf: device %s not configured\n",
2696                                     dev->name);
2697                 return;
2698         }
2699 
2700         /*
2701          *      Two things going on here:
2702          *      1) Add routes for on-link prefixes
2703          *      2) Configure prefixes with the auto flag set
2704          */
2705 
2706         if (pinfo->onlink) {
2707                 struct fib6_info *rt;
2708                 unsigned long rt_expires;
2709 
2710                 /* Avoid arithmetic overflow. Really, we could
2711                  * save rt_expires in seconds, likely valid_lft,
2712                  * but it would require division in fib gc, that it
2713                  * not good.
2714                  */
2715                 if (HZ > USER_HZ)
2716                         rt_expires = addrconf_timeout_fixup(valid_lft, HZ);
2717                 else
2718                         rt_expires = addrconf_timeout_fixup(valid_lft, USER_HZ);
2719 
2720                 if (addrconf_finite_timeout(rt_expires))
2721                         rt_expires *= HZ;
2722 
2723                 rt = addrconf_get_prefix_route(&pinfo->prefix,
2724                                                pinfo->prefix_len,
2725                                                dev,
2726                                                RTF_ADDRCONF | RTF_PREFIX_RT,
2727                                                RTF_DEFAULT, true);
2728 
2729                 if (rt) {
2730                         /* Autoconf prefix route */
2731                         if (valid_lft == 0) {
2732                                 ip6_del_rt(net, rt);
2733                                 rt = NULL;
2734                         } else if (addrconf_finite_timeout(rt_expires)) {
2735                                 /* not infinity */
2736                                 fib6_set_expires(rt, jiffies + rt_expires);
2737                         } else {
2738                                 fib6_clean_expires(rt);
2739                         }
2740                 } else if (valid_lft) {
2741                         clock_t expires = 0;
2742                         int flags = RTF_ADDRCONF | RTF_PREFIX_RT;
2743                         if (addrconf_finite_timeout(rt_expires)) {
2744                                 /* not infinity */
2745                                 flags |= RTF_EXPIRES;
2746                                 expires = jiffies_to_clock_t(rt_expires);
2747                         }
2748                         addrconf_prefix_route(&pinfo->prefix, pinfo->prefix_len,
2749                                               0, dev, expires, flags,
2750                                               GFP_ATOMIC);
2751                 }
2752                 fib6_info_release(rt);
2753         }
2754 
2755         /* Try to figure out our local address for this prefix */
2756 
2757         if (pinfo->autoconf && in6_dev->cnf.autoconf) {
2758                 struct in6_addr addr;
2759                 bool tokenized = false, dev_addr_generated = false;
2760 
2761                 if (pinfo->prefix_len == 64) {
2762                         memcpy(&addr, &pinfo->prefix, 8);
2763 
2764                         if (!ipv6_addr_any(&in6_dev->token)) {
2765                                 read_lock_bh(&in6_dev->lock);
2766                                 memcpy(addr.s6_addr + 8,
2767                                        in6_dev->token.s6_addr + 8, 8);
2768                                 read_unlock_bh(&in6_dev->lock);
2769                                 tokenized = true;
2770                         } else if (is_addr_mode_generate_stable(in6_dev) &&
2771                                    !ipv6_generate_stable_address(&addr, 0,
2772                                                                  in6_dev)) {
2773                                 addr_flags |= IFA_F_STABLE_PRIVACY;
2774                                 goto ok;
2775                         } else if (ipv6_generate_eui64(addr.s6_addr + 8, dev) &&
2776                                    ipv6_inherit_eui64(addr.s6_addr + 8, in6_dev)) {
2777                                 goto put;
2778                         } else {
2779                                 dev_addr_generated = true;
2780                         }
2781                         goto ok;
2782                 }
2783                 net_dbg_ratelimited("IPv6 addrconf: prefix with wrong length %d\n",
2784                                     pinfo->prefix_len);
2785                 goto put;
2786 
2787 ok:
2788                 err = addrconf_prefix_rcv_add_addr(net, dev, pinfo, in6_dev,
2789                                                    &addr, addr_type,
2790                                                    addr_flags, sllao,
2791                                                    tokenized, valid_lft,
2792                                                    prefered_lft);
2793                 if (err)
2794                         goto put;
2795 
2796                 /* Ignore error case here because previous prefix add addr was
2797                  * successful which will be notified.
2798                  */
2799                 ndisc_ops_prefix_rcv_add_addr(net, dev, pinfo, in6_dev, &addr,
2800                                               addr_type, addr_flags, sllao,
2801                                               tokenized, valid_lft,
2802                                               prefered_lft,
2803                                               dev_addr_generated);
2804         }
2805         inet6_prefix_notify(RTM_NEWPREFIX, in6_dev, pinfo);
2806 put:
2807         in6_dev_put(in6_dev);
2808 }
2809 
2810 /*
2811  *      Set destination address.
2812  *      Special case for SIT interfaces where we create a new "virtual"
2813  *      device.
2814  */
2815 int addrconf_set_dstaddr(struct net *net, void __user *arg)
2816 {
2817         struct in6_ifreq ireq;
2818         struct net_device *dev;
2819         int err = -EINVAL;
2820 
2821         rtnl_lock();
2822 
2823         err = -EFAULT;
2824         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2825                 goto err_exit;
2826 
2827         dev = __dev_get_by_index(net, ireq.ifr6_ifindex);
2828 
2829         err = -ENODEV;
2830         if (!dev)
2831                 goto err_exit;
2832 
2833 #if IS_ENABLED(CONFIG_IPV6_SIT)
2834         if (dev->type == ARPHRD_SIT) {
2835                 const struct net_device_ops *ops = dev->netdev_ops;
2836                 struct ifreq ifr;
2837                 struct ip_tunnel_parm p;
2838 
2839                 err = -EADDRNOTAVAIL;
2840                 if (!(ipv6_addr_type(&ireq.ifr6_addr) & IPV6_ADDR_COMPATv4))
2841                         goto err_exit;
2842 
2843                 memset(&p, 0, sizeof(p));
2844                 p.iph.daddr = ireq.ifr6_addr.s6_addr32[3];
2845                 p.iph.saddr = 0;
2846                 p.iph.version = 4;
2847                 p.iph.ihl = 5;
2848                 p.iph.protocol = IPPROTO_IPV6;
2849                 p.iph.ttl = 64;
2850                 ifr.ifr_ifru.ifru_data = (__force void __user *)&p;
2851 
2852                 if (ops->ndo_do_ioctl) {
2853                         mm_segment_t oldfs = get_fs();
2854 
2855                         set_fs(KERNEL_DS);
2856                         err = ops->ndo_do_ioctl(dev, &ifr, SIOCADDTUNNEL);
2857                         set_fs(oldfs);
2858                 } else
2859                         err = -EOPNOTSUPP;
2860 
2861                 if (err == 0) {
2862                         err = -ENOBUFS;
2863                         dev = __dev_get_by_name(net, p.name);
2864                         if (!dev)
2865                                 goto err_exit;
2866                         err = dev_open(dev, NULL);
2867                 }
2868         }
2869 #endif
2870 
2871 err_exit:
2872         rtnl_unlock();
2873         return err;
2874 }
2875 
2876 static int ipv6_mc_config(struct sock *sk, bool join,
2877                           const struct in6_addr *addr, int ifindex)
2878 {
2879         int ret;
2880 
2881         ASSERT_RTNL();
2882 
2883         lock_sock(sk);
2884         if (join)
2885                 ret = ipv6_sock_mc_join(sk, ifindex, addr);
2886         else
2887                 ret = ipv6_sock_mc_drop(sk, ifindex, addr);
2888         release_sock(sk);
2889 
2890         return ret;
2891 }
2892 
2893 /*
2894  *      Manual configuration of address on an interface
2895  */
2896 static int inet6_addr_add(struct net *net, int ifindex,
2897                           struct ifa6_config *cfg,
2898                           struct netlink_ext_ack *extack)
2899 {
2900         struct inet6_ifaddr *ifp;
2901         struct inet6_dev *idev;
2902         struct net_device *dev;
2903         unsigned long timeout;
2904         clock_t expires;
2905         u32 flags;
2906 
2907         ASSERT_RTNL();
2908 
2909         if (cfg->plen > 128)
2910                 return -EINVAL;
2911 
2912         /* check the lifetime */
2913         if (!cfg->valid_lft || cfg->preferred_lft > cfg->valid_lft)
2914                 return -EINVAL;
2915 
2916         if (cfg->ifa_flags & IFA_F_MANAGETEMPADDR && cfg->plen != 64)
2917                 return -EINVAL;
2918 
2919         dev = __dev_get_by_index(net, ifindex);
2920         if (!dev)
2921                 return -ENODEV;
2922 
2923         idev = addrconf_add_dev(dev);
2924         if (IS_ERR(idev))
2925                 return PTR_ERR(idev);
2926 
2927         if (cfg->ifa_flags & IFA_F_MCAUTOJOIN) {
2928                 int ret = ipv6_mc_config(net->ipv6.mc_autojoin_sk,
2929                                          true, cfg->pfx, ifindex);
2930 
2931                 if (ret < 0)
2932                         return ret;
2933         }
2934 
2935         cfg->scope = ipv6_addr_scope(cfg->pfx);
2936 
2937         timeout = addrconf_timeout_fixup(cfg->valid_lft, HZ);
2938         if (addrconf_finite_timeout(timeout)) {
2939                 expires = jiffies_to_clock_t(timeout * HZ);
2940                 cfg->valid_lft = timeout;
2941                 flags = RTF_EXPIRES;
2942         } else {
2943                 expires = 0;
2944                 flags = 0;
2945                 cfg->ifa_flags |= IFA_F_PERMANENT;
2946         }
2947 
2948         timeout = addrconf_timeout_fixup(cfg->preferred_lft, HZ);
2949         if (addrconf_finite_timeout(timeout)) {
2950                 if (timeout == 0)
2951                         cfg->ifa_flags |= IFA_F_DEPRECATED;
2952                 cfg->preferred_lft = timeout;
2953         }
2954 
2955         ifp = ipv6_add_addr(idev, cfg, true, extack);
2956         if (!IS_ERR(ifp)) {
2957                 if (!(cfg->ifa_flags & IFA_F_NOPREFIXROUTE)) {
2958                         addrconf_prefix_route(&ifp->addr, ifp->prefix_len,
2959                                               ifp->rt_priority, dev, expires,
2960                                               flags, GFP_KERNEL);
2961                 }
2962 
2963                 /* Send a netlink notification if DAD is enabled and
2964                  * optimistic flag is not set
2965                  */
2966                 if (!(ifp->flags & (IFA_F_OPTIMISTIC | IFA_F_NODAD)))
2967                         ipv6_ifa_notify(0, ifp);
2968                 /*
2969                  * Note that section 3.1 of RFC 4429 indicates
2970                  * that the Optimistic flag should not be set for
2971                  * manually configured addresses
2972                  */
2973                 addrconf_dad_start(ifp);
2974                 if (cfg->ifa_flags & IFA_F_MANAGETEMPADDR)
2975                         manage_tempaddrs(idev, ifp, cfg->valid_lft,
2976                                          cfg->preferred_lft, true, jiffies);
2977                 in6_ifa_put(ifp);
2978                 addrconf_verify_rtnl();
2979                 return 0;
2980         } else if (cfg->ifa_flags & IFA_F_MCAUTOJOIN) {
2981                 ipv6_mc_config(net->ipv6.mc_autojoin_sk, false,
2982                                cfg->pfx, ifindex);
2983         }
2984 
2985         return PTR_ERR(ifp);
2986 }
2987 
2988 static int inet6_addr_del(struct net *net, int ifindex, u32 ifa_flags,
2989                           const struct in6_addr *pfx, unsigned int plen)
2990 {
2991         struct inet6_ifaddr *ifp;
2992         struct inet6_dev *idev;
2993         struct net_device *dev;
2994 
2995         if (plen > 128)
2996                 return -EINVAL;
2997 
2998         dev = __dev_get_by_index(net, ifindex);
2999         if (!dev)
3000                 return -ENODEV;
3001 
3002         idev = __in6_dev_get(dev);
3003         if (!idev)
3004                 return -ENXIO;
3005 
3006         read_lock_bh(&idev->lock);
3007         list_for_each_entry(ifp, &idev->addr_list, if_list) {
3008                 if (ifp->prefix_len == plen &&
3009                     ipv6_addr_equal(pfx, &ifp->addr)) {
3010                         in6_ifa_hold(ifp);
3011                         read_unlock_bh(&idev->lock);
3012 
3013                         if (!(ifp->flags & IFA_F_TEMPORARY) &&
3014                             (ifa_flags & IFA_F_MANAGETEMPADDR))
3015                                 manage_tempaddrs(idev, ifp, 0, 0, false,
3016                                                  jiffies);
3017                         ipv6_del_addr(ifp);
3018                         addrconf_verify_rtnl();
3019                         if (ipv6_addr_is_multicast(pfx)) {
3020                                 ipv6_mc_config(net->ipv6.mc_autojoin_sk,
3021                                                false, pfx, dev->ifindex);
3022                         }
3023                         return 0;
3024                 }
3025         }
3026         read_unlock_bh(&idev->lock);
3027         return -EADDRNOTAVAIL;
3028 }
3029 
3030 
3031 int addrconf_add_ifaddr(struct net *net, void __user *arg)
3032 {
3033         struct ifa6_config cfg = {
3034                 .ifa_flags = IFA_F_PERMANENT,
3035                 .preferred_lft = INFINITY_LIFE_TIME,
3036                 .valid_lft = INFINITY_LIFE_TIME,
3037         };
3038         struct in6_ifreq ireq;
3039         int err;
3040 
3041         if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
3042                 return -EPERM;
3043 
3044         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
3045                 return -EFAULT;
3046 
3047         cfg.pfx = &ireq.ifr6_addr;
3048         cfg.plen = ireq.ifr6_prefixlen;
3049 
3050         rtnl_lock();
3051         err = inet6_addr_add(net, ireq.ifr6_ifindex, &cfg, NULL);
3052         rtnl_unlock();
3053         return err;
3054 }
3055 
3056 int addrconf_del_ifaddr(struct net *net, void __user *arg)
3057 {
3058         struct in6_ifreq ireq;
3059         int err;
3060 
3061         if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
3062                 return -EPERM;
3063 
3064         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
3065                 return -EFAULT;
3066 
3067         rtnl_lock();
3068         err = inet6_addr_del(net, ireq.ifr6_ifindex, 0, &ireq.ifr6_addr,
3069                              ireq.ifr6_prefixlen);
3070         rtnl_unlock();
3071         return err;
3072 }
3073 
3074 static void add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
3075                      int plen, int scope)
3076 {
3077         struct inet6_ifaddr *ifp;
3078         struct ifa6_config cfg = {
3079                 .pfx = addr,
3080                 .plen = plen,
3081                 .ifa_flags = IFA_F_PERMANENT,
3082                 .valid_lft = INFINITY_LIFE_TIME,
3083                 .preferred_lft = INFINITY_LIFE_TIME,
3084                 .scope = scope
3085         };
3086 
3087         ifp = ipv6_add_addr(idev, &cfg, true, NULL);
3088         if (!IS_ERR(ifp)) {
3089                 spin_lock_bh(&ifp->lock);
3090                 ifp->flags &= ~IFA_F_TENTATIVE;
3091                 spin_unlock_bh(&ifp->lock);
3092                 rt_genid_bump_ipv6(dev_net(idev->dev));
3093                 ipv6_ifa_notify(RTM_NEWADDR, ifp);
3094                 in6_ifa_put(ifp);
3095         }
3096 }
3097 
3098 #if IS_ENABLED(CONFIG_IPV6_SIT)
3099 static void sit_add_v4_addrs(struct inet6_dev *idev)
3100 {
3101         struct in6_addr addr;
3102         struct net_device *dev;
3103         struct net *net = dev_net(idev->dev);
3104         int scope, plen;
3105         u32 pflags = 0;
3106 
3107         ASSERT_RTNL();
3108 
3109         memset(&addr, 0, sizeof(struct in6_addr));
3110         memcpy(&addr.s6_addr32[3], idev->dev->dev_addr, 4);
3111 
3112         if (idev->dev->flags&IFF_POINTOPOINT) {
3113                 addr.s6_addr32[0] = htonl(0xfe800000);
3114                 scope = IFA_LINK;
3115                 plen = 64;
3116         } else {
3117                 scope = IPV6_ADDR_COMPATv4;
3118                 plen = 96;
3119                 pflags |= RTF_NONEXTHOP;
3120         }
3121 
3122         if (addr.s6_addr32[3]) {
3123                 add_addr(idev, &addr, plen, scope);
3124                 addrconf_prefix_route(&addr, plen, 0, idev->dev, 0, pflags,
3125                                       GFP_KERNEL);
3126                 return;
3127         }
3128 
3129         for_each_netdev(net, dev) {
3130                 struct in_device *in_dev = __in_dev_get_rtnl(dev);
3131                 if (in_dev && (dev->flags & IFF_UP)) {
3132                         struct in_ifaddr *ifa;
3133                         int flag = scope;
3134 
3135                         in_dev_for_each_ifa_rtnl(ifa, in_dev) {
3136                                 addr.s6_addr32[3] = ifa->ifa_local;
3137 
3138                                 if (ifa->ifa_scope == RT_SCOPE_LINK)
3139                                         continue;
3140                                 if (ifa->ifa_scope >= RT_SCOPE_HOST) {
3141                                         if (idev->dev->flags&IFF_POINTOPOINT)
3142                                                 continue;
3143                                         flag |= IFA_HOST;
3144                                 }
3145 
3146                                 add_addr(idev, &addr, plen, flag);
3147                                 addrconf_prefix_route(&addr, plen, 0, idev->dev,
3148                                                       0, pflags, GFP_KERNEL);
3149                         }
3150                 }
3151         }
3152 }
3153 #endif
3154 
3155 static void init_loopback(struct net_device *dev)
3156 {
3157         struct inet6_dev  *idev;
3158 
3159         /* ::1 */
3160 
3161         ASSERT_RTNL();
3162 
3163         idev = ipv6_find_idev(dev);
3164         if (IS_ERR(idev)) {
3165                 pr_debug("%s: add_dev failed\n", __func__);
3166                 return;
3167         }
3168 
3169         add_addr(idev, &in6addr_loopback, 128, IFA_HOST);
3170 }
3171 
3172 void addrconf_add_linklocal(struct inet6_dev *idev,
3173                             const struct in6_addr *addr, u32 flags)
3174 {
3175         struct ifa6_config cfg = {
3176                 .pfx = addr,
3177                 .plen = 64,
3178                 .ifa_flags = flags | IFA_F_PERMANENT,
3179                 .valid_lft = INFINITY_LIFE_TIME,
3180                 .preferred_lft = INFINITY_LIFE_TIME,
3181                 .scope = IFA_LINK
3182         };
3183         struct inet6_ifaddr *ifp;
3184 
3185 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
3186         if ((dev_net(idev->dev)->ipv6.devconf_all->optimistic_dad ||
3187              idev->cnf.optimistic_dad) &&
3188             !dev_net(idev->dev)->ipv6.devconf_all->forwarding)
3189                 cfg.ifa_flags |= IFA_F_OPTIMISTIC;
3190 #endif
3191 
3192         ifp = ipv6_add_addr(idev, &cfg, true, NULL);
3193         if (!IS_ERR(ifp)) {
3194                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, 0, idev->dev,
3195                                       0, 0, GFP_ATOMIC);
3196                 addrconf_dad_start(ifp);
3197                 in6_ifa_put(ifp);
3198         }
3199 }
3200 EXPORT_SYMBOL_GPL(addrconf_add_linklocal);
3201 
3202 static bool ipv6_reserved_interfaceid(struct in6_addr address)
3203 {
3204         if ((address.s6_addr32[2] | address.s6_addr32[3]) == 0)
3205                 return true;
3206 
3207         if (address.s6_addr32[2] == htonl(0x02005eff) &&
3208             ((address.s6_addr32[3] & htonl(0xfe000000)) == htonl(0xfe000000)))
3209                 return true;
3210 
3211         if (address.s6_addr32[2] == htonl(0xfdffffff) &&
3212             ((address.s6_addr32[3] & htonl(0xffffff80)) == htonl(0xffffff80)))
3213                 return true;
3214 
3215         return false;
3216 }
3217 
3218 static int ipv6_generate_stable_address(struct in6_addr *address,
3219                                         u8 dad_count,
3220                                         const struct inet6_dev *idev)
3221 {
3222         static DEFINE_SPINLOCK(lock);
3223         static __u32 digest[SHA_DIGEST_WORDS];
3224         static __u32 workspace[SHA_WORKSPACE_WORDS];
3225 
3226         static union {
3227                 char __data[SHA_MESSAGE_BYTES];
3228                 struct {
3229                         struct in6_addr secret;
3230                         __be32 prefix[2];
3231                         unsigned char hwaddr[MAX_ADDR_LEN];
3232                         u8 dad_count;
3233                 } __packed;
3234         } data;
3235 
3236         struct in6_addr secret;
3237         struct in6_addr temp;
3238         struct net *net = dev_net(idev->dev);
3239 
3240         BUILD_BUG_ON(sizeof(data.__data) != sizeof(data));
3241 
3242         if (idev->cnf.stable_secret.initialized)
3243                 secret = idev->cnf.stable_secret.secret;
3244         else if (net->ipv6.devconf_dflt->stable_secret.initialized)
3245                 secret = net->ipv6.devconf_dflt->stable_secret.secret;
3246         else
3247                 return -1;
3248 
3249 retry:
3250         spin_lock_bh(&lock);
3251 
3252         sha_init(digest);
3253         memset(&data, 0, sizeof(data));
3254         memset(workspace, 0, sizeof(workspace));
3255         memcpy(data.hwaddr, idev->dev->perm_addr, idev->dev->addr_len);
3256         data.prefix[0] = address->s6_addr32[0];
3257         data.prefix[1] = address->s6_addr32[1];
3258         data.secret = secret;
3259         data.dad_count = dad_count;
3260 
3261         sha_transform(digest, data.__data, workspace);
3262 
3263         temp = *address;
3264         temp.s6_addr32[2] = (__force __be32)digest[0];
3265         temp.s6_addr32[3] = (__force __be32)digest[1];
3266 
3267         spin_unlock_bh(&lock);
3268 
3269         if (ipv6_reserved_interfaceid(temp)) {
3270                 dad_count++;
3271                 if (dad_count > dev_net(idev->dev)->ipv6.sysctl.idgen_retries)
3272                         return -1;
3273                 goto retry;
3274         }
3275 
3276         *address = temp;
3277         return 0;
3278 }
3279 
3280 static void ipv6_gen_mode_random_init(struct inet6_dev *idev)
3281 {
3282         struct ipv6_stable_secret *s = &idev->cnf.stable_secret;
3283 
3284         if (s->initialized)
3285                 return;
3286         s = &idev->cnf.stable_secret;
3287         get_random_bytes(&s->secret, sizeof(s->secret));
3288         s->initialized = true;
3289 }
3290 
3291 static void addrconf_addr_gen(struct inet6_dev *idev, bool prefix_route)
3292 {
3293         struct in6_addr addr;
3294 
3295         /* no link local addresses on L3 master devices */
3296         if (netif_is_l3_master(idev->dev))
3297                 return;
3298 
3299         /* no link local addresses on devices flagged as slaves */
3300         if (idev->dev->flags & IFF_SLAVE)
3301                 return;
3302 
3303         ipv6_addr_set(&addr, htonl(0xFE800000), 0, 0, 0);
3304 
3305         switch (idev->cnf.addr_gen_mode) {
3306         case IN6_ADDR_GEN_MODE_RANDOM:
3307                 ipv6_gen_mode_random_init(idev);
3308                 /* fallthrough */
3309         case IN6_ADDR_GEN_MODE_STABLE_PRIVACY:
3310                 if (!ipv6_generate_stable_address(&addr, 0, idev))
3311                         addrconf_add_linklocal(idev, &addr,
3312                                                IFA_F_STABLE_PRIVACY);
3313                 else if (prefix_route)
3314                         addrconf_prefix_route(&addr, 64, 0, idev->dev,
3315                                               0, 0, GFP_KERNEL);
3316                 break;
3317         case IN6_ADDR_GEN_MODE_EUI64:
3318                 /* addrconf_add_linklocal also adds a prefix_route and we
3319                  * only need to care about prefix routes if ipv6_generate_eui64
3320                  * couldn't generate one.
3321                  */
3322                 if (ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) == 0)
3323                         addrconf_add_linklocal(idev, &addr, 0);
3324                 else if (prefix_route)
3325                         addrconf_prefix_route(&addr, 64, 0, idev->dev,
3326                                               0, 0, GFP_KERNEL);
3327                 break;
3328         case IN6_ADDR_GEN_MODE_NONE:
3329         default:
3330                 /* will not add any link local address */
3331                 break;
3332         }
3333 }
3334 
3335 static void addrconf_dev_config(struct net_device *dev)
3336 {
3337         struct inet6_dev *idev;
3338 
3339         ASSERT_RTNL();
3340 
3341         if ((dev->type != ARPHRD_ETHER) &&
3342             (dev->type != ARPHRD_FDDI) &&
3343             (dev->type != ARPHRD_ARCNET) &&
3344             (dev->type != ARPHRD_INFINIBAND) &&
3345             (dev->type != ARPHRD_IEEE1394) &&
3346             (dev->type != ARPHRD_TUNNEL6) &&
3347             (dev->type != ARPHRD_6LOWPAN) &&
3348             (dev->type != ARPHRD_IP6GRE) &&
3349             (dev->type != ARPHRD_IPGRE) &&
3350             (dev->type != ARPHRD_TUNNEL) &&
3351             (dev->type != ARPHRD_NONE) &&
3352             (dev->type != ARPHRD_RAWIP)) {
3353                 /* Alas, we support only Ethernet autoconfiguration. */
3354                 idev = __in6_dev_get(dev);
3355                 if (!IS_ERR_OR_NULL(idev) && dev->flags & IFF_UP &&
3356                     dev->flags & IFF_MULTICAST)
3357                         ipv6_mc_up(idev);
3358                 return;
3359         }
3360 
3361         idev = addrconf_add_dev(dev);
3362         if (IS_ERR(idev))
3363                 return;
3364 
3365         /* this device type has no EUI support */
3366         if (dev->type == ARPHRD_NONE &&
3367             idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_EUI64)
3368                 idev->cnf.addr_gen_mode = IN6_ADDR_GEN_MODE_RANDOM;
3369 
3370         addrconf_addr_gen(idev, false);
3371 }
3372 
3373 #if IS_ENABLED(CONFIG_IPV6_SIT)
3374 static void addrconf_sit_config(struct net_device *dev)
3375 {
3376         struct inet6_dev *idev;
3377 
3378         ASSERT_RTNL();
3379 
3380         /*
3381          * Configure the tunnel with one of our IPv4
3382          * addresses... we should configure all of
3383          * our v4 addrs in the tunnel
3384          */
3385 
3386         idev = ipv6_find_idev(dev);
3387         if (IS_ERR(idev)) {
3388                 pr_debug("%s: add_dev failed\n", __func__);
3389                 return;
3390         }
3391 
3392         if (dev->priv_flags & IFF_ISATAP) {
3393                 addrconf_addr_gen(idev, false);
3394                 return;
3395         }
3396 
3397         sit_add_v4_addrs(idev);
3398 
3399         if (dev->flags&IFF_POINTOPOINT)
3400                 addrconf_add_mroute(dev);
3401 }
3402 #endif
3403 
3404 #if IS_ENABLED(CONFIG_NET_IPGRE)
3405 static void addrconf_gre_config(struct net_device *dev)
3406 {
3407         struct inet6_dev *idev;
3408 
3409         ASSERT_RTNL();
3410 
3411         idev = ipv6_find_idev(dev);
3412         if (IS_ERR(idev)) {
3413                 pr_debug("%s: add_dev failed\n", __func__);
3414                 return;
3415         }
3416 
3417         addrconf_addr_gen(idev, true);
3418         if (dev->flags & IFF_POINTOPOINT)
3419                 addrconf_add_mroute(dev);
3420 }
3421 #endif
3422 
3423 static int fixup_permanent_addr(struct net *net,
3424                                 struct inet6_dev *idev,
3425                                 struct inet6_ifaddr *ifp)
3426 {
3427         /* !fib6_node means the host route was removed from the
3428          * FIB, for example, if 'lo' device is taken down. In that
3429          * case regenerate the host route.
3430          */
3431         if (!ifp->rt || !ifp->rt->fib6_node) {
3432                 struct fib6_info *f6i, *prev;
3433 
3434                 f6i = addrconf_f6i_alloc(net, idev, &ifp->addr, false,
3435                                          GFP_ATOMIC);
3436                 if (IS_ERR(f6i))
3437                         return PTR_ERR(f6i);
3438 
3439                 /* ifp->rt can be accessed outside of rtnl */
3440                 spin_lock(&ifp->lock);
3441                 prev = ifp->rt;
3442                 ifp->rt = f6i;
3443                 spin_unlock(&ifp->lock);
3444 
3445                 fib6_info_release(prev);
3446         }
3447 
3448         if (!(ifp->flags & IFA_F_NOPREFIXROUTE)) {
3449                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len,
3450                                       ifp->rt_priority, idev->dev, 0, 0,
3451                                       GFP_ATOMIC);
3452         }
3453 
3454         if (ifp->state == INET6_IFADDR_STATE_PREDAD)
3455                 addrconf_dad_start(ifp);
3456 
3457         return 0;
3458 }
3459 
3460 static void addrconf_permanent_addr(struct net *net, struct net_device *dev)
3461 {
3462         struct inet6_ifaddr *ifp, *tmp;
3463         struct inet6_dev *idev;
3464 
3465         idev = __in6_dev_get(dev);
3466         if (!idev)
3467                 return;
3468 
3469         write_lock_bh(&idev->lock);
3470 
3471         list_for_each_entry_safe(ifp, tmp, &idev->addr_list, if_list) {
3472                 if ((ifp->flags & IFA_F_PERMANENT) &&
3473                     fixup_permanent_addr(net, idev, ifp) < 0) {
3474                         write_unlock_bh(&idev->lock);
3475                         in6_ifa_hold(ifp);
3476                         ipv6_del_addr(ifp);
3477                         write_lock_bh(&idev->lock);
3478 
3479                         net_info_ratelimited("%s: Failed to add prefix route for address %pI6c; dropping\n",
3480                                              idev->dev->name, &ifp->addr);
3481                 }
3482         }
3483 
3484         write_unlock_bh(&idev->lock);
3485 }
3486 
3487 static int addrconf_notify(struct notifier_block *this, unsigned long event,
3488                            void *ptr)
3489 {
3490         struct net_device *dev = netdev_notifier_info_to_dev(ptr);
3491         struct netdev_notifier_change_info *change_info;
3492         struct netdev_notifier_changeupper_info *info;
3493         struct inet6_dev *idev = __in6_dev_get(dev);
3494         struct net *net = dev_net(dev);
3495         int run_pending = 0;
3496         int err;
3497 
3498         switch (event) {
3499         case NETDEV_REGISTER:
3500                 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
3501                         idev = ipv6_add_dev(dev);
3502                         if (IS_ERR(idev))
3503                                 return notifier_from_errno(PTR_ERR(idev));
3504                 }
3505                 break;
3506 
3507         case NETDEV_CHANGEMTU:
3508                 /* if MTU under IPV6_MIN_MTU stop IPv6 on this interface. */
3509                 if (dev->mtu < IPV6_MIN_MTU) {
3510                         addrconf_ifdown(dev, dev != net->loopback_dev);
3511                         break;
3512                 }
3513 
3514                 if (idev) {
3515                         rt6_mtu_change(dev, dev->mtu);
3516                         idev->cnf.mtu6 = dev->mtu;
3517                         break;
3518                 }
3519 
3520                 /* allocate new idev */
3521                 idev = ipv6_add_dev(dev);
3522                 if (IS_ERR(idev))
3523                         break;
3524 
3525                 /* device is still not ready */
3526                 if (!(idev->if_flags & IF_READY))
3527                         break;
3528 
3529                 run_pending = 1;
3530 
3531                 /* fall through */
3532 
3533         case NETDEV_UP:
3534         case NETDEV_CHANGE:
3535                 if (dev->flags & IFF_SLAVE)
3536                         break;
3537 
3538                 if (idev && idev->cnf.disable_ipv6)
3539                         break;
3540 
3541                 if (event == NETDEV_UP) {
3542                         /* restore routes for permanent addresses */
3543                         addrconf_permanent_addr(net, dev);
3544 
3545                         if (!addrconf_link_ready(dev)) {
3546                                 /* device is not ready yet. */
3547                                 pr_debug("ADDRCONF(NETDEV_UP): %s: link is not ready\n",
3548                                          dev->name);
3549                                 break;
3550                         }
3551 
3552                         if (!idev && dev->mtu >= IPV6_MIN_MTU)
3553                                 idev = ipv6_add_dev(dev);
3554 
3555                         if (!IS_ERR_OR_NULL(idev)) {
3556                                 idev->if_flags |= IF_READY;
3557                                 run_pending = 1;
3558                         }
3559                 } else if (event == NETDEV_CHANGE) {
3560                         if (!addrconf_link_ready(dev)) {
3561                                 /* device is still not ready. */
3562                                 rt6_sync_down_dev(dev, event);
3563                                 break;
3564                         }
3565 
3566                         if (!IS_ERR_OR_NULL(idev)) {
3567                                 if (idev->if_flags & IF_READY) {
3568                                         /* device is already configured -
3569                                          * but resend MLD reports, we might
3570                                          * have roamed and need to update
3571                                          * multicast snooping switches
3572                                          */
3573                                         ipv6_mc_up(idev);
3574                                         change_info = ptr;
3575                                         if (change_info->flags_changed & IFF_NOARP)
3576                                                 addrconf_dad_run(idev, true);
3577                                         rt6_sync_up(dev, RTNH_F_LINKDOWN);
3578                                         break;
3579                                 }
3580                                 idev->if_flags |= IF_READY;
3581                         }
3582 
3583                         pr_info("ADDRCONF(NETDEV_CHANGE): %s: link becomes ready\n",
3584                                 dev->name);
3585 
3586                         run_pending = 1;
3587                 }
3588 
3589                 switch (dev->type) {
3590 #if IS_ENABLED(CONFIG_IPV6_SIT)
3591                 case ARPHRD_SIT:
3592                         addrconf_sit_config(dev);
3593                         break;
3594 #endif
3595 #if IS_ENABLED(CONFIG_NET_IPGRE)
3596                 case ARPHRD_IPGRE:
3597                         addrconf_gre_config(dev);
3598                         break;
3599 #endif
3600                 case ARPHRD_LOOPBACK:
3601                         init_loopback(dev);
3602                         break;
3603 
3604                 default:
3605                         addrconf_dev_config(dev);
3606                         break;
3607                 }
3608 
3609                 if (!IS_ERR_OR_NULL(idev)) {
3610                         if (run_pending)
3611                                 addrconf_dad_run(idev, false);
3612 
3613                         /* Device has an address by now */
3614                         rt6_sync_up(dev, RTNH_F_DEAD);
3615 
3616                         /*
3617                          * If the MTU changed during the interface down,
3618                          * when the interface up, the changed MTU must be
3619                          * reflected in the idev as well as routers.
3620                          */
3621                         if (idev->cnf.mtu6 != dev->mtu &&
3622                             dev->mtu >= IPV6_MIN_MTU) {
3623                                 rt6_mtu_change(dev, dev->mtu);
3624                                 idev->cnf.mtu6 = dev->mtu;
3625                         }
3626                         idev->tstamp = jiffies;
3627                         inet6_ifinfo_notify(RTM_NEWLINK, idev);
3628 
3629                         /*
3630                          * If the changed mtu during down is lower than
3631                          * IPV6_MIN_MTU stop IPv6 on this interface.
3632                          */
3633                         if (dev->mtu < IPV6_MIN_MTU)
3634                                 addrconf_ifdown(dev, dev != net->loopback_dev);
3635                 }
3636                 break;
3637 
3638         case NETDEV_DOWN:
3639         case NETDEV_UNREGISTER:
3640                 /*
3641                  *      Remove all addresses from this interface.
3642                  */
3643                 addrconf_ifdown(dev, event != NETDEV_DOWN);
3644                 break;
3645 
3646         case NETDEV_CHANGENAME:
3647                 if (idev) {
3648                         snmp6_unregister_dev(idev);
3649                         addrconf_sysctl_unregister(idev);
3650                         err = addrconf_sysctl_register(idev);
3651                         if (err)
3652                                 return notifier_from_errno(err);
3653                         err = snmp6_register_dev(idev);
3654                         if (err) {
3655                                 addrconf_sysctl_unregister(idev);
3656                                 return notifier_from_errno(err);
3657                         }
3658                 }
3659                 break;
3660 
3661         case NETDEV_PRE_TYPE_CHANGE:
3662         case NETDEV_POST_TYPE_CHANGE:
3663                 if (idev)
3664                         addrconf_type_change(dev, event);
3665                 break;
3666 
3667         case NETDEV_CHANGEUPPER:
3668                 info = ptr;
3669 
3670                 /* flush all routes if dev is linked to or unlinked from
3671                  * an L3 master device (e.g., VRF)
3672                  */
3673                 if (info->upper_dev && netif_is_l3_master(info->upper_dev))
3674                         addrconf_ifdown(dev, 0);
3675         }
3676 
3677         return NOTIFY_OK;
3678 }
3679 
3680 /*
3681  *      addrconf module should be notified of a device going up
3682  */
3683 static struct notifier_block ipv6_dev_notf = {
3684         .notifier_call = addrconf_notify,
3685         .priority = ADDRCONF_NOTIFY_PRIORITY,
3686 };
3687 
3688 static void addrconf_type_change(struct net_device *dev, unsigned long event)
3689 {
3690         struct inet6_dev *idev;
3691         ASSERT_RTNL();
3692 
3693         idev = __in6_dev_get(dev);
3694 
3695         if (event == NETDEV_POST_TYPE_CHANGE)
3696                 ipv6_mc_remap(idev);
3697         else if (event == NETDEV_PRE_TYPE_CHANGE)
3698                 ipv6_mc_unmap(idev);
3699 }
3700 
3701 static bool addr_is_local(const struct in6_addr *addr)
3702 {
3703         return ipv6_addr_type(addr) &
3704                 (IPV6_ADDR_LINKLOCAL | IPV6_ADDR_LOOPBACK);
3705 }
3706 
3707 static int addrconf_ifdown(struct net_device *dev, int how)
3708 {
3709         unsigned long event = how ? NETDEV_UNREGISTER : NETDEV_DOWN;
3710         struct net *net = dev_net(dev);
3711         struct inet6_dev *idev;
3712         struct inet6_ifaddr *ifa, *tmp;
3713         bool keep_addr = false;
3714         int state, i;
3715 
3716         ASSERT_RTNL();
3717 
3718         rt6_disable_ip(dev, event);
3719 
3720         idev = __in6_dev_get(dev);
3721         if (!idev)
3722                 return -ENODEV;
3723 
3724         /*
3725          * Step 1: remove reference to ipv6 device from parent device.
3726          *         Do not dev_put!
3727          */
3728         if (how) {
3729                 idev->dead = 1;
3730 
3731                 /* protected by rtnl_lock */
3732                 RCU_INIT_POINTER(dev->ip6_ptr, NULL);
3733 
3734                 /* Step 1.5: remove snmp6 entry */
3735                 snmp6_unregister_dev(idev);
3736 
3737         }
3738 
3739         /* combine the user config with event to determine if permanent
3740          * addresses are to be removed from address hash table
3741          */
3742         if (!how && !idev->cnf.disable_ipv6) {
3743                 /* aggregate the system setting and interface setting */
3744                 int _keep_addr = net->ipv6.devconf_all->keep_addr_on_down;
3745 
3746                 if (!_keep_addr)
3747                         _keep_addr = idev->cnf.keep_addr_on_down;
3748 
3749                 keep_addr = (_keep_addr > 0);
3750         }
3751 
3752         /* Step 2: clear hash table */
3753         for (i = 0; i < IN6_ADDR_HSIZE; i++) {
3754                 struct hlist_head *h = &inet6_addr_lst[i];
3755 
3756                 spin_lock_bh(&addrconf_hash_lock);
3757 restart:
3758                 hlist_for_each_entry_rcu(ifa, h, addr_lst) {
3759                         if (ifa->idev == idev) {
3760                                 addrconf_del_dad_work(ifa);
3761                                 /* combined flag + permanent flag decide if
3762                                  * address is retained on a down event
3763                                  */
3764                                 if (!keep_addr ||
3765                                     !(ifa->flags & IFA_F_PERMANENT) ||
3766                                     addr_is_local(&ifa->addr)) {
3767                                         hlist_del_init_rcu(&ifa->addr_lst);
3768                                         goto restart;
3769                                 }
3770                         }
3771                 }
3772                 spin_unlock_bh(&addrconf_hash_lock);
3773         }
3774 
3775         write_lock_bh(&idev->lock);
3776 
3777         addrconf_del_rs_timer(idev);
3778 
3779         /* Step 2: clear flags for stateless addrconf */
3780         if (!how)
3781                 idev->if_flags &= ~(IF_RS_SENT|IF_RA_RCVD|IF_READY);
3782 
3783         /* Step 3: clear tempaddr list */
3784         while (!list_empty(&idev->tempaddr_list)) {
3785                 ifa = list_first_entry(&idev->tempaddr_list,
3786                                        struct inet6_ifaddr, tmp_list);
3787                 list_del(&ifa->tmp_list);
3788                 write_unlock_bh(&idev->lock);
3789                 spin_lock_bh(&ifa->lock);
3790 
3791                 if (ifa->ifpub) {
3792                         in6_ifa_put(ifa->ifpub);
3793                         ifa->ifpub = NULL;
3794                 }
3795                 spin_unlock_bh(&ifa->lock);
3796                 in6_ifa_put(ifa);
3797                 write_lock_bh(&idev->lock);
3798         }
3799 
3800         list_for_each_entry_safe(ifa, tmp, &idev->addr_list, if_list) {
3801                 struct fib6_info *rt = NULL;
3802                 bool keep;
3803 
3804                 addrconf_del_dad_work(ifa);
3805 
3806                 keep = keep_addr && (ifa->flags & IFA_F_PERMANENT) &&
3807                         !addr_is_local(&ifa->addr);
3808 
3809                 write_unlock_bh(&idev->lock);
3810                 spin_lock_bh(&ifa->lock);
3811 
3812                 if (keep) {
3813                         /* set state to skip the notifier below */
3814                         state = INET6_IFADDR_STATE_DEAD;
3815                         ifa->state = INET6_IFADDR_STATE_PREDAD;
3816                         if (!(ifa->flags & IFA_F_NODAD))
3817                                 ifa->flags |= IFA_F_TENTATIVE;
3818 
3819                         rt = ifa->rt;
3820                         ifa->rt = NULL;
3821                 } else {
3822                         state = ifa->state;
3823                         ifa->state = INET6_IFADDR_STATE_DEAD;
3824                 }
3825 
3826                 spin_unlock_bh(&ifa->lock);
3827 
3828                 if (rt)
3829                         ip6_del_rt(net, rt);
3830 
3831                 if (state != INET6_IFADDR_STATE_DEAD) {
3832                         __ipv6_ifa_notify(RTM_DELADDR, ifa);
3833                         inet6addr_notifier_call_chain(NETDEV_DOWN, ifa);
3834                 } else {
3835                         if (idev->cnf.forwarding)
3836                                 addrconf_leave_anycast(ifa);
3837                         addrconf_leave_solict(ifa->idev, &ifa->addr);
3838                 }
3839 
3840                 write_lock_bh(&idev->lock);
3841                 if (!keep) {
3842                         list_del_rcu(&ifa->if_list);
3843                         in6_ifa_put(ifa);
3844                 }
3845         }
3846 
3847         write_unlock_bh(&idev->lock);
3848 
3849         /* Step 5: Discard anycast and multicast list */
3850         if (how) {
3851                 ipv6_ac_destroy_dev(idev);
3852                 ipv6_mc_destroy_dev(idev);
3853         } else {
3854                 ipv6_mc_down(idev);
3855         }
3856 
3857         idev->tstamp = jiffies;
3858 
3859         /* Last: Shot the device (if unregistered) */
3860         if (how) {
3861                 addrconf_sysctl_unregister(idev);
3862                 neigh_parms_release(&nd_tbl, idev->nd_parms);
3863                 neigh_ifdown(&nd_tbl, dev);
3864                 in6_dev_put(idev);
3865         }
3866         return 0;
3867 }
3868 
3869 static void addrconf_rs_timer(struct timer_list *t)
3870 {
3871         struct inet6_dev *idev = from_timer(idev, t, rs_timer);
3872         struct net_device *dev = idev->dev;
3873         struct in6_addr lladdr;
3874 
3875         write_lock(&idev->lock);
3876         if (idev->dead || !(idev->if_flags & IF_READY))
3877                 goto out;
3878 
3879         if (!ipv6_accept_ra(idev))
3880                 goto out;
3881 
3882         /* Announcement received after solicitation was sent */
3883         if (idev->if_flags & IF_RA_RCVD)
3884                 goto out;
3885 
3886         if (idev->rs_probes++ < idev->cnf.rtr_solicits || idev->cnf.rtr_solicits < 0) {
3887                 write_unlock(&idev->lock);
3888                 if (!ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE))
3889                         ndisc_send_rs(dev, &lladdr,
3890                                       &in6addr_linklocal_allrouters);
3891                 else
3892                         goto put;
3893 
3894                 write_lock(&idev->lock);
3895                 idev->rs_interval = rfc3315_s14_backoff_update(
3896                         idev->rs_interval, idev->cnf.rtr_solicit_max_interval);
3897                 /* The wait after the last probe can be shorter */
3898                 addrconf_mod_rs_timer(idev, (idev->rs_probes ==
3899                                              idev->cnf.rtr_solicits) ?
3900                                       idev->cnf.rtr_solicit_delay :
3901                                       idev->rs_interval);
3902         } else {
3903                 /*
3904                  * Note: we do not support deprecated "all on-link"
3905                  * assumption any longer.
3906                  */
3907                 pr_debug("%s: no IPv6 routers present\n", idev->dev->name);
3908         }
3909 
3910 out:
3911         write_unlock(&idev->lock);
3912 put:
3913         in6_dev_put(idev);
3914 }
3915 
3916 /*
3917  *      Duplicate Address Detection
3918  */
3919 static void addrconf_dad_kick(struct inet6_ifaddr *ifp)
3920 {
3921         unsigned long rand_num;
3922         struct inet6_dev *idev = ifp->idev;
3923         u64 nonce;
3924 
3925         if (ifp->flags & IFA_F_OPTIMISTIC)
3926                 rand_num = 0;
3927         else
3928                 rand_num = prandom_u32() % (idev->cnf.rtr_solicit_delay ? : 1);
3929 
3930         nonce = 0;
3931         if (idev->cnf.enhanced_dad ||
3932             dev_net(idev->dev)->ipv6.devconf_all->enhanced_dad) {
3933                 do
3934                         get_random_bytes(&nonce, 6);
3935                 while (nonce == 0);
3936         }
3937         ifp->dad_nonce = nonce;
3938         ifp->dad_probes = idev->cnf.dad_transmits;
3939         addrconf_mod_dad_work(ifp, rand_num);
3940 }
3941 
3942 static void addrconf_dad_begin(struct inet6_ifaddr *ifp)
3943 {
3944         struct inet6_dev *idev = ifp->idev;
3945         struct net_device *dev = idev->dev;
3946         bool bump_id, notify = false;
3947         struct net *net;
3948 
3949         addrconf_join_solict(dev, &ifp->addr);
3950 
3951         prandom_seed((__force u32) ifp->addr.s6_addr32[3]);
3952 
3953         read_lock_bh(&idev->lock);
3954         spin_lock(&ifp->lock);
3955         if (ifp->state == INET6_IFADDR_STATE_DEAD)
3956                 goto out;
3957 
3958         net = dev_net(dev);
3959         if (dev->flags&(IFF_NOARP|IFF_LOOPBACK) ||
3960             (net->ipv6.devconf_all->accept_dad < 1 &&
3961              idev->cnf.accept_dad < 1) ||
3962             !(ifp->flags&IFA_F_TENTATIVE) ||
3963             ifp->flags & IFA_F_NODAD) {
3964                 bool send_na = false;
3965 
3966                 if (ifp->flags & IFA_F_TENTATIVE &&
3967                     !(ifp->flags & IFA_F_OPTIMISTIC))
3968                         send_na = true;
3969                 bump_id = ifp->flags & IFA_F_TENTATIVE;
3970                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
3971                 spin_unlock(&ifp->lock);
3972                 read_unlock_bh(&idev->lock);
3973 
3974                 addrconf_dad_completed(ifp, bump_id, send_na);
3975                 return;
3976         }
3977 
3978         if (!(idev->if_flags & IF_READY)) {
3979                 spin_unlock(&ifp->lock);
3980                 read_unlock_bh(&idev->lock);
3981                 /*
3982                  * If the device is not ready:
3983                  * - keep it tentative if it is a permanent address.
3984                  * - otherwise, kill it.
3985                  */
3986                 in6_ifa_hold(ifp);
3987                 addrconf_dad_stop(ifp, 0);
3988                 return;
3989         }
3990 
3991         /*
3992          * Optimistic nodes can start receiving
3993          * Frames right away
3994          */
3995         if (ifp->flags & IFA_F_OPTIMISTIC) {
3996                 ip6_ins_rt(net, ifp->rt);
3997                 if (ipv6_use_optimistic_addr(net, idev)) {
3998                         /* Because optimistic nodes can use this address,
3999                          * notify listeners. If DAD fails, RTM_DELADDR is sent.
4000                          */
4001                         notify = true;
4002                 }
4003         }
4004 
4005         addrconf_dad_kick(ifp);
4006 out:
4007         spin_unlock(&ifp->lock);
4008         read_unlock_bh(&idev->lock);
4009         if (notify)
4010                 ipv6_ifa_notify(RTM_NEWADDR, ifp);
4011 }
4012 
4013 static void addrconf_dad_start(struct inet6_ifaddr *ifp)
4014 {
4015         bool begin_dad = false;
4016 
4017         spin_lock_bh(&ifp->lock);
4018         if (ifp->state != INET6_IFADDR_STATE_DEAD) {
4019                 ifp->state = INET6_IFADDR_STATE_PREDAD;
4020                 begin_dad = true;
4021         }
4022         spin_unlock_bh(&ifp->lock);
4023 
4024         if (begin_dad)
4025                 addrconf_mod_dad_work(ifp, 0);
4026 }
4027 
4028 static void addrconf_dad_work(struct work_struct *w)
4029 {
4030         struct inet6_ifaddr *ifp = container_of(to_delayed_work(w),
4031                                                 struct inet6_ifaddr,
4032                                                 dad_work);
4033         struct inet6_dev *idev = ifp->idev;
4034         bool bump_id, disable_ipv6 = false;
4035         struct in6_addr mcaddr;
4036 
4037         enum {
4038                 DAD_PROCESS,
4039                 DAD_BEGIN,
4040                 DAD_ABORT,
4041         } action = DAD_PROCESS;
4042 
4043         rtnl_lock();
4044 
4045         spin_lock_bh(&ifp->lock);
4046         if (ifp->state == INET6_IFADDR_STATE_PREDAD) {
4047                 action = DAD_BEGIN;
4048                 ifp->state = INET6_IFADDR_STATE_DAD;
4049         } else if (ifp->state == INET6_IFADDR_STATE_ERRDAD) {
4050                 action = DAD_ABORT;
4051                 ifp->state = INET6_IFADDR_STATE_POSTDAD;
4052 
4053                 if ((dev_net(idev->dev)->ipv6.devconf_all->accept_dad > 1 ||
4054                      idev->cnf.accept_dad > 1) &&
4055                     !idev->cnf.disable_ipv6 &&
4056                     !(ifp->flags & IFA_F_STABLE_PRIVACY)) {
4057                         struct in6_addr addr;
4058 
4059                         addr.s6_addr32[0] = htonl(0xfe800000);
4060                         addr.s6_addr32[1] = 0;
4061 
4062                         if (!ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) &&
4063                             ipv6_addr_equal(&ifp->addr, &addr)) {
4064                                 /* DAD failed for link-local based on MAC */
4065                                 idev->cnf.disable_ipv6 = 1;
4066 
4067                                 pr_info("%s: IPv6 being disabled!\n",
4068                                         ifp->idev->dev->name);
4069                                 disable_ipv6 = true;
4070                         }
4071                 }
4072         }
4073         spin_unlock_bh(&ifp->lock);
4074 
4075         if (action == DAD_BEGIN) {
4076                 addrconf_dad_begin(ifp);
4077                 goto out;
4078         } else if (action == DAD_ABORT) {
4079                 in6_ifa_hold(ifp);
4080                 addrconf_dad_stop(ifp, 1);
4081                 if (disable_ipv6)
4082                         addrconf_ifdown(idev->dev, 0);
4083                 goto out;
4084         }
4085 
4086         if (!ifp->dad_probes && addrconf_dad_end(ifp))
4087                 goto out;
4088 
4089         write_lock_bh(&idev->lock);
4090         if (idev->dead || !(idev->if_flags & IF_READY)) {
4091                 write_unlock_bh(&idev->lock);
4092                 goto out;
4093         }
4094 
4095         spin_lock(&ifp->lock);
4096         if (ifp->state == INET6_IFADDR_STATE_DEAD) {
4097                 spin_unlock(&ifp->lock);
4098                 write_unlock_bh(&idev->lock);
4099                 goto out;
4100         }
4101 
4102         if (ifp->dad_probes == 0) {
4103                 bool send_na = false;
4104 
4105                 /*
4106                  * DAD was successful
4107                  */
4108 
4109                 if (ifp->flags & IFA_F_TENTATIVE &&
4110                     !(ifp->flags & IFA_F_OPTIMISTIC))
4111                         send_na = true;
4112                 bump_id = ifp->flags & IFA_F_TENTATIVE;
4113                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
4114                 spin_unlock(&ifp->lock);
4115                 write_unlock_bh(&idev->lock);
4116 
4117                 addrconf_dad_completed(ifp, bump_id, send_na);
4118 
4119                 goto out;
4120         }
4121 
4122         ifp->dad_probes--;
4123         addrconf_mod_dad_work(ifp,
4124                               NEIGH_VAR(ifp->idev->nd_parms, RETRANS_TIME));
4125         spin_unlock(&ifp->lock);
4126         write_unlock_bh(&idev->lock);
4127 
4128         /* send a neighbour solicitation for our addr */
4129         addrconf_addr_solict_mult(&ifp->addr, &mcaddr);
4130         ndisc_send_ns(ifp->idev->dev, &ifp->addr, &mcaddr, &in6addr_any,
4131                       ifp->dad_nonce);
4132 out:
4133         in6_ifa_put(ifp);
4134         rtnl_unlock();
4135 }
4136 
4137 /* ifp->idev must be at least read locked */
4138 static bool ipv6_lonely_lladdr(struct inet6_ifaddr *ifp)
4139 {
4140         struct inet6_ifaddr *ifpiter;
4141         struct inet6_dev *idev = ifp->idev;
4142 
4143         list_for_each_entry_reverse(ifpiter, &idev->addr_list, if_list) {
4144                 if (ifpiter->scope > IFA_LINK)
4145                         break;
4146                 if (ifp != ifpiter && ifpiter->scope == IFA_LINK &&
4147                     (ifpiter->flags & (IFA_F_PERMANENT|IFA_F_TENTATIVE|
4148                                        IFA_F_OPTIMISTIC|IFA_F_DADFAILED)) ==
4149                     IFA_F_PERMANENT)
4150                         return false;
4151         }
4152         return true;
4153 }
4154 
4155 static void addrconf_dad_completed(struct inet6_ifaddr *ifp, bool bump_id,
4156                                    bool send_na)
4157 {
4158         struct net_device *dev = ifp->idev->dev;
4159         struct in6_addr lladdr;
4160         bool send_rs, send_mld;
4161 
4162         addrconf_del_dad_work(ifp);
4163 
4164         /*
4165          *      Configure the address for reception. Now it is valid.
4166          */
4167 
4168         ipv6_ifa_notify(RTM_NEWADDR, ifp);
4169 
4170         /* If added prefix is link local and we are prepared to process
4171            router advertisements, start sending router solicitations.
4172          */
4173 
4174         read_lock_bh(&ifp->idev->lock);
4175         send_mld = ifp->scope == IFA_LINK && ipv6_lonely_lladdr(ifp);
4176         send_rs = send_mld &&
4177                   ipv6_accept_ra(ifp->idev) &&
4178                   ifp->idev->cnf.rtr_solicits != 0 &&
4179                   (dev->flags&IFF_LOOPBACK) == 0;
4180         read_unlock_bh(&ifp->idev->lock);
4181 
4182         /* While dad is in progress mld report's source address is in6_addrany.
4183          * Resend with proper ll now.
4184          */
4185         if (send_mld)
4186                 ipv6_mc_dad_complete(ifp->idev);
4187 
4188         /* send unsolicited NA if enabled */
4189         if (send_na &&
4190             (ifp->idev->cnf.ndisc_notify ||
4191              dev_net(dev)->ipv6.devconf_all->ndisc_notify)) {
4192                 ndisc_send_na(dev, &in6addr_linklocal_allnodes, &ifp->addr,
4193                               /*router=*/ !!ifp->idev->cnf.forwarding,
4194                               /*solicited=*/ false, /*override=*/ true,
4195                               /*inc_opt=*/ true);
4196         }
4197 
4198         if (send_rs) {
4199                 /*
4200                  *      If a host as already performed a random delay
4201                  *      [...] as part of DAD [...] there is no need
4202                  *      to delay again before sending the first RS
4203                  */
4204                 if (ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE))
4205                         return;
4206                 ndisc_send_rs(dev, &lladdr, &in6addr_linklocal_allrouters);
4207 
4208                 write_lock_bh(&ifp->idev->lock);
4209                 spin_lock(&ifp->lock);
4210                 ifp->idev->rs_interval = rfc3315_s14_backoff_init(
4211                         ifp->idev->cnf.rtr_solicit_interval);
4212                 ifp->idev->rs_probes = 1;
4213                 ifp->idev->if_flags |= IF_RS_SENT;
4214                 addrconf_mod_rs_timer(ifp->idev, ifp->idev->rs_interval);
4215                 spin_unlock(&ifp->lock);
4216                 write_unlock_bh(&ifp->idev->lock);
4217         }
4218 
4219         if (bump_id)
4220                 rt_genid_bump_ipv6(dev_net(dev));
4221 
4222         /* Make sure that a new temporary address will be created
4223          * before this temporary address becomes deprecated.
4224          */
4225         if (ifp->flags & IFA_F_TEMPORARY)
4226                 addrconf_verify_rtnl();
4227 }
4228 
4229 static void addrconf_dad_run(struct inet6_dev *idev, bool restart)
4230 {
4231         struct inet6_ifaddr *ifp;
4232 
4233         read_lock_bh(&idev->lock);
4234         list_for_each_entry(ifp, &idev->addr_list, if_list) {
4235                 spin_lock(&ifp->lock);
4236                 if ((ifp->flags & IFA_F_TENTATIVE &&
4237                      ifp->state == INET6_IFADDR_STATE_DAD) || restart) {
4238                         if (restart)
4239                                 ifp->state = INET6_IFADDR_STATE_PREDAD;
4240                         addrconf_dad_kick(ifp);
4241                 }
4242                 spin_unlock(&ifp->lock);
4243         }
4244         read_unlock_bh(&idev->lock);
4245 }
4246 
4247 #ifdef CONFIG_PROC_FS
4248 struct if6_iter_state {
4249         struct seq_net_private p;
4250         int bucket;
4251         int offset;
4252 };
4253 
4254 static struct inet6_ifaddr *if6_get_first(struct seq_file *seq, loff_t pos)
4255 {
4256         struct if6_iter_state *state = seq->private;
4257         struct net *net = seq_file_net(seq);
4258         struct inet6_ifaddr *ifa = NULL;
4259         int p = 0;
4260 
4261         /* initial bucket if pos is 0 */
4262         if (pos == 0) {
4263                 state->bucket = 0;
4264                 state->offset = 0;
4265         }
4266 
4267         for (; state->bucket < IN6_ADDR_HSIZE; ++state->bucket) {
4268                 hlist_for_each_entry_rcu(ifa, &inet6_addr_lst[state->bucket],
4269                                          addr_lst) {
4270                         if (!net_eq(dev_net(ifa->idev->dev), net))
4271                                 continue;
4272                         /* sync with offset */
4273                         if (p < state->offset) {
4274                                 p++;
4275                                 continue;
4276                         }
4277                         return ifa;
4278                 }
4279 
4280                 /* prepare for next bucket */
4281                 state->offset = 0;
4282                 p = 0;
4283         }
4284         return NULL;
4285 }
4286 
4287 static struct inet6_ifaddr *if6_get_next(struct seq_file *seq,
4288                                          struct inet6_ifaddr *ifa)
4289 {
4290         struct if6_iter_state *state = seq->private;
4291         struct net *net = seq_file_net(seq);
4292 
4293         hlist_for_each_entry_continue_rcu(ifa, addr_lst) {
4294                 if (!net_eq(dev_net(ifa->idev->dev), net))
4295                         continue;
4296                 state->offset++;
4297                 return ifa;
4298         }
4299 
4300         state->offset = 0;
4301         while (++state->bucket < IN6_ADDR_HSIZE) {
4302                 hlist_for_each_entry_rcu(ifa,
4303                                      &inet6_addr_lst[state->bucket], addr_lst) {
4304                         if (!net_eq(dev_net(ifa->idev->dev), net))
4305                                 continue;
4306                         return ifa;
4307                 }
4308         }
4309 
4310         return NULL;
4311 }
4312 
4313 static void *if6_seq_start(struct seq_file *seq, loff_t *pos)
4314         __acquires(rcu)
4315 {
4316         rcu_read_lock();
4317         return if6_get_first(seq, *pos);
4318 }
4319 
4320 static void *if6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
4321 {
4322         struct inet6_ifaddr *ifa;
4323 
4324         ifa = if6_get_next(seq, v);
4325         ++*pos;
4326         return ifa;
4327 }
4328 
4329 static void if6_seq_stop(struct seq_file *seq, void *v)
4330         __releases(rcu)
4331 {
4332         rcu_read_unlock();
4333 }
4334 
4335 static int if6_seq_show(struct seq_file *seq, void *v)
4336 {
4337         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *)v;
4338         seq_printf(seq, "%pi6 %02x %02x %02x %02x %8s\n",
4339                    &ifp->addr,
4340                    ifp->idev->dev->ifindex,
4341                    ifp->prefix_len,
4342                    ifp->scope,
4343                    (u8) ifp->flags,
4344                    ifp->idev->dev->name);
4345         return 0;
4346 }
4347 
4348 static const struct seq_operations if6_seq_ops = {
4349         .start  = if6_seq_start,
4350         .next   = if6_seq_next,
4351         .show   = if6_seq_show,
4352         .stop   = if6_seq_stop,
4353 };
4354 
4355 static int __net_init if6_proc_net_init(struct net *net)
4356 {
4357         if (!proc_create_net("if_inet6", 0444, net->proc_net, &if6_seq_ops,
4358                         sizeof(struct if6_iter_state)))
4359                 return -ENOMEM;
4360         return 0;
4361 }
4362 
4363 static void __net_exit if6_proc_net_exit(struct net *net)
4364 {
4365         remove_proc_entry("if_inet6", net->proc_net);
4366 }
4367 
4368 static struct pernet_operations if6_proc_net_ops = {
4369         .init = if6_proc_net_init,
4370         .exit = if6_proc_net_exit,
4371 };
4372 
4373 int __init if6_proc_init(void)
4374 {
4375         return register_pernet_subsys(&if6_proc_net_ops);
4376 }
4377 
4378 void if6_proc_exit(void)
4379 {
4380         unregister_pernet_subsys(&if6_proc_net_ops);
4381 }
4382 #endif  /* CONFIG_PROC_FS */
4383 
4384 #if IS_ENABLED(CONFIG_IPV6_MIP6)
4385 /* Check if address is a home address configured on any interface. */
4386 int ipv6_chk_home_addr(struct net *net, const struct in6_addr *addr)
4387 {
4388         unsigned int hash = inet6_addr_hash(net, addr);
4389         struct inet6_ifaddr *ifp = NULL;
4390         int ret = 0;
4391 
4392         rcu_read_lock();
4393         hlist_for_each_entry_rcu(ifp, &inet6_addr_lst[hash], addr_lst) {
4394                 if (!net_eq(dev_net(ifp->idev->dev), net))
4395                         continue;
4396                 if (ipv6_addr_equal(&ifp->addr, addr) &&
4397                     (ifp->flags & IFA_F_HOMEADDRESS)) {
4398                         ret = 1;
4399                         break;
4400                 }
4401         }
4402         rcu_read_unlock();
4403         return ret;
4404 }
4405 #endif
4406 
4407 /*
4408  *      Periodic address status verification
4409  */
4410 
4411 static void addrconf_verify_rtnl(void)
4412 {
4413         unsigned long now, next, next_sec, next_sched;
4414         struct inet6_ifaddr *ifp;
4415         int i;
4416 
4417         ASSERT_RTNL();
4418 
4419         rcu_read_lock_bh();
4420         now = jiffies;
4421         next = round_jiffies_up(now + ADDR_CHECK_FREQUENCY);
4422 
4423         cancel_delayed_work(&addr_chk_work);
4424 
4425         for (i = 0; i < IN6_ADDR_HSIZE; i++) {
4426 restart:
4427                 hlist_for_each_entry_rcu_bh(ifp, &inet6_addr_lst[i], addr_lst) {
4428                         unsigned long age;
4429 
4430                         /* When setting preferred_lft to a value not zero or
4431                          * infinity, while valid_lft is infinity
4432                          * IFA_F_PERMANENT has a non-infinity life time.
4433                          */
4434                         if ((ifp->flags & IFA_F_PERMANENT) &&
4435                             (ifp->prefered_lft == INFINITY_LIFE_TIME))
4436                                 continue;
4437 
4438                         spin_lock(&ifp->lock);
4439                         /* We try to batch several events at once. */
4440                         age = (now - ifp->tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
4441 
4442                         if (ifp->valid_lft != INFINITY_LIFE_TIME &&
4443                             age >= ifp->valid_lft) {
4444                                 spin_unlock(&ifp->lock);
4445                                 in6_ifa_hold(ifp);
4446                                 ipv6_del_addr(ifp);
4447                                 goto restart;
4448                         } else if (ifp->prefered_lft == INFINITY_LIFE_TIME) {
4449                                 spin_unlock(&ifp->lock);
4450                                 continue;
4451                         } else if (age >= ifp->prefered_lft) {
4452                                 /* jiffies - ifp->tstamp > age >= ifp->prefered_lft */
4453                                 int deprecate = 0;
4454 
4455                                 if (!(ifp->flags&IFA_F_DEPRECATED)) {
4456                                         deprecate = 1;
4457                                         ifp->flags |= IFA_F_DEPRECATED;
4458                                 }
4459 
4460                                 if ((ifp->valid_lft != INFINITY_LIFE_TIME) &&
4461                                     (time_before(ifp->tstamp + ifp->valid_lft * HZ, next)))
4462                                         next = ifp->tstamp + ifp->valid_lft * HZ;
4463 
4464                                 spin_unlock(&ifp->lock);
4465 
4466                                 if (deprecate) {
4467                                         in6_ifa_hold(ifp);
4468 
4469                                         ipv6_ifa_notify(0, ifp);
4470                                         in6_ifa_put(ifp);
4471                                         goto restart;
4472                                 }
4473                         } else if ((ifp->flags&IFA_F_TEMPORARY) &&
4474                                    !(ifp->flags&IFA_F_TENTATIVE)) {
4475                                 unsigned long regen_advance = ifp->idev->cnf.regen_max_retry *
4476                                         ifp->idev->cnf.dad_transmits *
4477                                         NEIGH_VAR(ifp->idev->nd_parms, RETRANS_TIME) / HZ;
4478 
4479                                 if (age >= ifp->prefered_lft - regen_advance) {
4480                                         struct inet6_ifaddr *ifpub = ifp->ifpub;
4481                                         if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
4482                                                 next = ifp->tstamp + ifp->prefered_lft * HZ;
4483                                         if (!ifp->regen_count && ifpub) {
4484                                                 ifp->regen_count++;
4485                                                 in6_ifa_hold(ifp);
4486                                                 in6_ifa_hold(ifpub);
4487                                                 spin_unlock(&ifp->lock);
4488 
4489                                                 spin_lock(&ifpub->lock);
4490                                                 ifpub->regen_count = 0;
4491                                                 spin_unlock(&ifpub->lock);
4492                                                 rcu_read_unlock_bh();
4493                                                 ipv6_create_tempaddr(ifpub, ifp, true);
4494                                                 in6_ifa_put(ifpub);
4495                                                 in6_ifa_put(ifp);
4496                                                 rcu_read_lock_bh();
4497                                                 goto restart;
4498                                         }
4499                                 } else if (time_before(ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ, next))
4500                                         next = ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ;
4501                                 spin_unlock(&ifp->lock);
4502                         } else {
4503                                 /* ifp->prefered_lft <= ifp->valid_lft */
4504                                 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
4505                                         next = ifp->tstamp + ifp->prefered_lft * HZ;
4506                                 spin_unlock(&ifp->lock);
4507                         }
4508                 }
4509         }
4510 
4511         next_sec = round_jiffies_up(next);
4512         next_sched = next;
4513 
4514         /* If rounded timeout is accurate enough, accept it. */
4515         if (time_before(next_sec, next + ADDRCONF_TIMER_FUZZ))
4516                 next_sched = next_sec;
4517 
4518         /* And minimum interval is ADDRCONF_TIMER_FUZZ_MAX. */
4519         if (time_before(next_sched, jiffies + ADDRCONF_TIMER_FUZZ_MAX))
4520                 next_sched = jiffies + ADDRCONF_TIMER_FUZZ_MAX;
4521 
4522         pr_debug("now = %lu, schedule = %lu, rounded schedule = %lu => %lu\n",
4523                  now, next, next_sec, next_sched);
4524         mod_delayed_work(addrconf_wq, &addr_chk_work, next_sched - now);
4525         rcu_read_unlock_bh();
4526 }
4527 
4528 static void addrconf_verify_work(struct work_struct *w)
4529 {
4530         rtnl_lock();
4531         addrconf_verify_rtnl();
4532         rtnl_unlock();
4533 }
4534 
4535 static void addrconf_verify(void)
4536 {
4537         mod_delayed_work(addrconf_wq, &addr_chk_work, 0);
4538 }
4539 
4540 static struct in6_addr *extract_addr(struct nlattr *addr, struct nlattr *local,
4541                                      struct in6_addr **peer_pfx)
4542 {
4543         struct in6_addr *pfx = NULL;
4544 
4545         *peer_pfx = NULL;
4546 
4547         if (addr)
4548                 pfx = nla_data(addr);
4549 
4550         if (local) {
4551                 if (pfx && nla_memcmp(local, pfx, sizeof(*pfx)))
4552                         *peer_pfx = pfx;
4553                 pfx = nla_data(local);
4554         }
4555 
4556         return pfx;
4557 }
4558 
4559 static const struct nla_policy ifa_ipv6_policy[IFA_MAX+1] = {
4560         [IFA_ADDRESS]           = { .len = sizeof(struct in6_addr) },
4561         [IFA_LOCAL]             = { .len = sizeof(struct in6_addr) },
4562         [IFA_CACHEINFO]         = { .len = sizeof(struct ifa_cacheinfo) },
4563         [IFA_FLAGS]             = { .len = sizeof(u32) },
4564         [IFA_RT_PRIORITY]       = { .len = sizeof(u32) },
4565         [IFA_TARGET_NETNSID]    = { .type = NLA_S32 },
4566 };
4567 
4568 static int
4569 inet6_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh,
4570                   struct netlink_ext_ack *extack)
4571 {
4572         struct net *net = sock_net(skb->sk);
4573         struct ifaddrmsg *ifm;
4574         struct nlattr *tb[IFA_MAX+1];
4575         struct in6_addr *pfx, *peer_pfx;
4576         u32 ifa_flags;
4577         int err;
4578 
4579         err = nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFA_MAX,
4580                                      ifa_ipv6_policy, extack);
4581         if (err < 0)
4582                 return err;
4583 
4584         ifm = nlmsg_data(nlh);
4585         pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
4586         if (!pfx)
4587                 return -EINVAL;
4588 
4589         ifa_flags = tb[IFA_FLAGS] ? nla_get_u32(tb[IFA_FLAGS]) : ifm->ifa_flags;
4590 
4591         /* We ignore other flags so far. */
4592         ifa_flags &= IFA_F_MANAGETEMPADDR;
4593 
4594         return inet6_addr_del(net, ifm->ifa_index, ifa_flags, pfx,
4595                               ifm->ifa_prefixlen);
4596 }
4597 
4598 static int modify_prefix_route(struct inet6_ifaddr *ifp,
4599                                unsigned long expires, u32 flags,
4600                                bool modify_peer)
4601 {
4602         struct fib6_info *f6i;
4603         u32 prio;
4604 
4605         f6i = addrconf_get_prefix_route(modify_peer ? &ifp->peer_addr : &ifp->addr,
4606                                         ifp->prefix_len,
4607                                         ifp->idev->dev, 0, RTF_DEFAULT, true);
4608         if (!f6i)
4609                 return -ENOENT;
4610 
4611         prio = ifp->rt_priority ? : IP6_RT_PRIO_ADDRCONF;
4612         if (f6i->fib6_metric != prio) {
4613                 /* delete old one */
4614                 ip6_del_rt(dev_net(ifp->idev->dev), f6i);
4615 
4616                 /* add new one */
4617                 addrconf_prefix_route(modify_peer ? &ifp->peer_addr : &ifp->addr,
4618                                       ifp->prefix_len,
4619                                       ifp->rt_priority, ifp->idev->dev,
4620                                       expires, flags, GFP_KERNEL);
4621         } else {
4622                 if (!expires)
4623                         fib6_clean_expires(f6i);
4624                 else
4625                         fib6_set_expires(f6i, expires);
4626 
4627                 fib6_info_release(f6i);
4628         }
4629 
4630         return 0;
4631 }
4632 
4633 static int inet6_addr_modify(struct inet6_ifaddr *ifp, struct ifa6_config *cfg)
4634 {
4635         u32 flags;
4636         clock_t expires;
4637         unsigned long timeout;
4638         bool was_managetempaddr;
4639         bool had_prefixroute;
4640         bool new_peer = false;
4641 
4642         ASSERT_RTNL();
4643 
4644         if (!cfg->valid_lft || cfg->preferred_lft > cfg->valid_lft)
4645                 return -EINVAL;
4646 
4647         if (cfg->ifa_flags & IFA_F_MANAGETEMPADDR &&
4648             (ifp->flags & IFA_F_TEMPORARY || ifp->prefix_len != 64))
4649                 return -EINVAL;
4650 
4651         if (!(ifp->flags & IFA_F_TENTATIVE) || ifp->flags & IFA_F_DADFAILED)
4652                 cfg->ifa_flags &= ~IFA_F_OPTIMISTIC;
4653 
4654         timeout = addrconf_timeout_fixup(cfg->valid_lft, HZ);
4655         if (addrconf_finite_timeout(timeout)) {
4656                 expires = jiffies_to_clock_t(timeout * HZ);
4657                 cfg->valid_lft = timeout;
4658                 flags = RTF_EXPIRES;
4659         } else {
4660                 expires = 0;
4661                 flags = 0;
4662                 cfg->ifa_flags |= IFA_F_PERMANENT;
4663         }
4664 
4665         timeout = addrconf_timeout_fixup(cfg->preferred_lft, HZ);
4666         if (addrconf_finite_timeout(timeout)) {
4667                 if (timeout == 0)
4668                         cfg->ifa_flags |= IFA_F_DEPRECATED;
4669                 cfg->preferred_lft = timeout;
4670         }
4671 
4672         if (cfg->peer_pfx &&
4673             memcmp(&ifp->peer_addr, cfg->peer_pfx, sizeof(struct in6_addr))) {
4674                 if (!ipv6_addr_any(&ifp->peer_addr))
4675                         cleanup_prefix_route(ifp, expires, true, true);
4676                 new_peer = true;
4677         }
4678 
4679         spin_lock_bh(&ifp->lock);
4680         was_managetempaddr = ifp->flags & IFA_F_MANAGETEMPADDR;
4681         had_prefixroute = ifp->flags & IFA_F_PERMANENT &&
4682                           !(ifp->flags & IFA_F_NOPREFIXROUTE);
4683         ifp->flags &= ~(IFA_F_DEPRECATED | IFA_F_PERMANENT | IFA_F_NODAD |
4684                         IFA_F_HOMEADDRESS | IFA_F_MANAGETEMPADDR |
4685                         IFA_F_NOPREFIXROUTE);
4686         ifp->flags |= cfg->ifa_flags;
4687         ifp->tstamp = jiffies;
4688         ifp->valid_lft = cfg->valid_lft;
4689         ifp->prefered_lft = cfg->preferred_lft;
4690 
4691         if (cfg->rt_priority && cfg->rt_priority != ifp->rt_priority)
4692                 ifp->rt_priority = cfg->rt_priority;
4693 
4694         if (new_peer)
4695                 ifp->peer_addr = *cfg->peer_pfx;
4696 
4697         spin_unlock_bh(&ifp->lock);
4698         if (!(ifp->flags&IFA_F_TENTATIVE))
4699                 ipv6_ifa_notify(0, ifp);
4700 
4701         if (!(cfg->ifa_flags & IFA_F_NOPREFIXROUTE)) {
4702                 int rc = -ENOENT;
4703 
4704                 if (had_prefixroute)
4705                         rc = modify_prefix_route(ifp, expires, flags, false);
4706 
4707                 /* prefix route could have been deleted; if so restore it */
4708                 if (rc == -ENOENT) {
4709                         addrconf_prefix_route(&ifp->addr, ifp->prefix_len,
4710                                               ifp->rt_priority, ifp->idev->dev,
4711                                               expires, flags, GFP_KERNEL);
4712                 }
4713 
4714                 if (had_prefixroute && !ipv6_addr_any(&ifp->peer_addr))
4715                         rc = modify_prefix_route(ifp, expires, flags, true);
4716 
4717                 if (rc == -ENOENT && !ipv6_addr_any(&ifp->peer_addr)) {
4718                         addrconf_prefix_route(&ifp->peer_addr, ifp->prefix_len,
4719                                               ifp->rt_priority, ifp->idev->dev,
4720                                               expires, flags, GFP_KERNEL);
4721                 }
4722         } else if (had_prefixroute) {
4723                 enum cleanup_prefix_rt_t action;
4724                 unsigned long rt_expires;
4725 
4726                 write_lock_bh(&ifp->idev->lock);
4727                 action = check_cleanup_prefix_route(ifp, &rt_expires);
4728                 write_unlock_bh(&ifp->idev->lock);
4729 
4730                 if (action != CLEANUP_PREFIX_RT_NOP) {
4731                         cleanup_prefix_route(ifp, rt_expires,
4732                                 action == CLEANUP_PREFIX_RT_DEL, false);
4733                 }
4734         }
4735 
4736         if (was_managetempaddr || ifp->flags & IFA_F_MANAGETEMPADDR) {
4737                 if (was_managetempaddr &&
4738                     !(ifp->flags & IFA_F_MANAGETEMPADDR)) {
4739                         cfg->valid_lft = 0;
4740                         cfg->preferred_lft = 0;
4741                 }
4742                 manage_tempaddrs(ifp->idev, ifp, cfg->valid_lft,
4743                                  cfg->preferred_lft, !was_managetempaddr,
4744                                  jiffies);
4745         }
4746 
4747         addrconf_verify_rtnl();
4748 
4749         return 0;
4750 }
4751 
4752 static int
4753 inet6_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh,
4754                   struct netlink_ext_ack *extack)
4755 {
4756         struct net *net = sock_net(skb->sk);
4757         struct ifaddrmsg *ifm;
4758         struct nlattr *tb[IFA_MAX+1];
4759         struct in6_addr *peer_pfx;
4760         struct inet6_ifaddr *ifa;
4761         struct net_device *dev;
4762         struct inet6_dev *idev;
4763         struct ifa6_config cfg;
4764         int err;
4765 
4766         err = nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFA_MAX,
4767                                      ifa_ipv6_policy, extack);
4768         if (err < 0)
4769                 return err;
4770 
4771         memset(&cfg, 0, sizeof(cfg));
4772 
4773         ifm = nlmsg_data(nlh);
4774         cfg.pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
4775         if (!cfg.pfx)
4776                 return -EINVAL;
4777 
4778         cfg.peer_pfx = peer_pfx;
4779         cfg.plen = ifm->ifa_prefixlen;
4780         if (tb[IFA_RT_PRIORITY])
4781                 cfg.rt_priority = nla_get_u32(tb[IFA_RT_PRIORITY]);
4782 
4783         cfg.valid_lft = INFINITY_LIFE_TIME;
4784         cfg.preferred_lft = INFINITY_LIFE_TIME;
4785 
4786         if (tb[IFA_CACHEINFO]) {
4787                 struct ifa_cacheinfo *ci;
4788 
4789                 ci = nla_data(tb[IFA_CACHEINFO]);
4790                 cfg.valid_lft = ci->ifa_valid;
4791                 cfg.preferred_lft = ci->ifa_prefered;
4792         }
4793 
4794         dev =  __dev_get_by_index(net, ifm->ifa_index);
4795         if (!dev)
4796                 return -ENODEV;
4797 
4798         if (tb[IFA_FLAGS])
4799                 cfg.ifa_flags = nla_get_u32(tb[IFA_FLAGS]);
4800         else
4801                 cfg.ifa_flags = ifm->ifa_flags;
4802 
4803         /* We ignore other flags so far. */
4804         cfg.ifa_flags &= IFA_F_NODAD | IFA_F_HOMEADDRESS |
4805                          IFA_F_MANAGETEMPADDR | IFA_F_NOPREFIXROUTE |
4806                          IFA_F_MCAUTOJOIN | IFA_F_OPTIMISTIC;
4807 
4808         idev = ipv6_find_idev(dev);
4809         if (IS_ERR(idev))
4810                 return PTR_ERR(idev);
4811 
4812         if (!ipv6_allow_optimistic_dad(net, idev))
4813                 cfg.ifa_flags &= ~IFA_F_OPTIMISTIC;
4814 
4815         if (cfg.ifa_flags & IFA_F_NODAD &&
4816             cfg.ifa_flags & IFA_F_OPTIMISTIC) {
4817                 NL_SET_ERR_MSG(extack, "IFA_F_NODAD and IFA_F_OPTIMISTIC are mutually exclusive");
4818                 return -EINVAL;
4819         }
4820 
4821         ifa = ipv6_get_ifaddr(net, cfg.pfx, dev, 1);
4822         if (!ifa) {
4823                 /*
4824                  * It would be best to check for !NLM_F_CREATE here but
4825                  * userspace already relies on not having to provide this.
4826                  */
4827                 return inet6_addr_add(net, ifm->ifa_index, &cfg, extack);
4828         }
4829 
4830         if (nlh->nlmsg_flags & NLM_F_EXCL ||
4831             !(nlh->nlmsg_flags & NLM_F_REPLACE))
4832                 err = -EEXIST;
4833         else
4834                 err = inet6_addr_modify(ifa, &cfg);
4835 
4836         in6_ifa_put(ifa);
4837 
4838         return err;
4839 }
4840 
4841 static void put_ifaddrmsg(struct nlmsghdr *nlh, u8 prefixlen, u32 flags,
4842                           u8 scope, int ifindex)
4843 {
4844         struct ifaddrmsg *ifm;
4845 
4846         ifm = nlmsg_data(nlh);
4847         ifm->ifa_family = AF_INET6;
4848         ifm->ifa_prefixlen = prefixlen;
4849         ifm->ifa_flags = flags;
4850         ifm->ifa_scope = scope;
4851         ifm->ifa_index = ifindex;
4852 }
4853 
4854 static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
4855                          unsigned long tstamp, u32 preferred, u32 valid)
4856 {
4857         struct ifa_cacheinfo ci;
4858 
4859         ci.cstamp = cstamp_delta(cstamp);
4860         ci.tstamp = cstamp_delta(tstamp);
4861         ci.ifa_prefered = preferred;
4862         ci.ifa_valid = valid;
4863 
4864         return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
4865 }
4866 
4867 static inline int rt_scope(int ifa_scope)
4868 {
4869         if (ifa_scope & IFA_HOST)
4870                 return RT_SCOPE_HOST;
4871         else if (ifa_scope & IFA_LINK)
4872                 return RT_SCOPE_LINK;
4873         else if (ifa_scope & IFA_SITE)
4874                 return RT_SCOPE_SITE;
4875         else
4876                 return RT_SCOPE_UNIVERSE;
4877 }
4878 
4879 static inline int inet6_ifaddr_msgsize(void)
4880 {
4881         return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
4882                + nla_total_size(16) /* IFA_LOCAL */
4883                + nla_total_size(16) /* IFA_ADDRESS */
4884                + nla_total_size(sizeof(struct ifa_cacheinfo))
4885                + nla_total_size(4)  /* IFA_FLAGS */
4886                + nla_total_size(4)  /* IFA_RT_PRIORITY */;
4887 }
4888 
4889 enum addr_type_t {
4890         UNICAST_ADDR,
4891         MULTICAST_ADDR,
4892         ANYCAST_ADDR,
4893 };
4894 
4895 struct inet6_fill_args {
4896         u32 portid;
4897         u32 seq;
4898         int event;
4899         unsigned int flags;
4900         int netnsid;
4901         int ifindex;
4902         enum addr_type_t type;
4903 };
4904 
4905 static int inet6_fill_ifaddr(struct sk_buff *skb, struct inet6_ifaddr *ifa,
4906                              struct inet6_fill_args *args)
4907 {
4908         struct nlmsghdr  *nlh;
4909         u32 preferred, valid;
4910 
4911         nlh = nlmsg_put(skb, args->portid, args->seq, args->event,
4912                         sizeof(struct ifaddrmsg), args->flags);
4913         if (!nlh)
4914                 return -EMSGSIZE;
4915 
4916         put_ifaddrmsg(nlh, ifa->prefix_len, ifa->flags, rt_scope(ifa->scope),
4917                       ifa->idev->dev->ifindex);
4918 
4919         if (args->netnsid >= 0 &&
4920             nla_put_s32(skb, IFA_TARGET_NETNSID, args->netnsid))
4921                 goto error;
4922 
4923         if (!((ifa->flags&IFA_F_PERMANENT) &&
4924               (ifa->prefered_lft == INFINITY_LIFE_TIME))) {
4925                 preferred = ifa->prefered_lft;
4926                 valid = ifa->valid_lft;
4927                 if (preferred != INFINITY_LIFE_TIME) {
4928                         long tval = (jiffies - ifa->tstamp)/HZ;
4929                         if (preferred > tval)
4930                                 preferred -= tval;
4931                         else
4932                                 preferred = 0;
4933                         if (valid != INFINITY_LIFE_TIME) {
4934                                 if (valid > tval)
4935                                         valid -= tval;
4936                                 else
4937                                         valid = 0;
4938                         }
4939                 }
4940         } else {
4941                 preferred = INFINITY_LIFE_TIME;
4942                 valid = INFINITY_LIFE_TIME;
4943         }
4944 
4945         if (!ipv6_addr_any(&ifa->peer_addr)) {
4946                 if (nla_put_in6_addr(skb, IFA_LOCAL, &ifa->addr) < 0 ||
4947                     nla_put_in6_addr(skb, IFA_ADDRESS, &ifa->peer_addr) < 0)
4948                         goto error;
4949         } else
4950                 if (nla_put_in6_addr(skb, IFA_ADDRESS, &ifa->addr) < 0)
4951                         goto error;
4952 
4953         if (ifa->rt_priority &&
4954             nla_put_u32(skb, IFA_RT_PRIORITY, ifa->rt_priority))
4955                 goto error;
4956 
4957         if (put_cacheinfo(skb, ifa->cstamp, ifa->tstamp, preferred, valid) < 0)
4958                 goto error;
4959 
4960         if (nla_put_u32(skb, IFA_FLAGS, ifa->flags) < 0)
4961                 goto error;
4962 
4963         nlmsg_end(skb, nlh);
4964         return 0;
4965 
4966 error:
4967         nlmsg_cancel(skb, nlh);
4968         return -EMSGSIZE;
4969 }
4970 
4971 static int inet6_fill_ifmcaddr(struct sk_buff *skb, struct ifmcaddr6 *ifmca,
4972                                struct inet6_fill_args *args)
4973 {
4974         struct nlmsghdr  *nlh;
4975         u8 scope = RT_SCOPE_UNIVERSE;
4976         int ifindex = ifmca->idev->dev->ifindex;
4977 
4978         if (ipv6_addr_scope(&ifmca->mca_addr) & IFA_SITE)
4979                 scope = RT_SCOPE_SITE;
4980 
4981         nlh = nlmsg_put(skb, args->portid, args->seq, args->event,
4982                         sizeof(struct ifaddrmsg), args->flags);
4983         if (!nlh)
4984                 return -EMSGSIZE;
4985 
4986         if (args->netnsid >= 0 &&
4987             nla_put_s32(skb, IFA_TARGET_NETNSID, args->netnsid))
4988                 return -EMSGSIZE;
4989 
4990         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
4991         if (nla_put_in6_addr(skb, IFA_MULTICAST, &ifmca->mca_addr) < 0 ||
4992             put_cacheinfo(skb, ifmca->mca_cstamp, ifmca->mca_tstamp,
4993                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
4994                 nlmsg_cancel(skb, nlh);
4995                 return -EMSGSIZE;
4996         }
4997 
4998         nlmsg_end(skb, nlh);
4999         return 0;
5000 }
5001 
5002 static int inet6_fill_ifacaddr(struct sk_buff *skb, struct ifacaddr6 *ifaca,
5003                                struct inet6_fill_args *args)
5004 {
5005         struct net_device *dev = fib6_info_nh_dev(ifaca->aca_rt);
5006         int ifindex = dev ? dev->ifindex : 1;
5007         struct nlmsghdr  *nlh;
5008         u8 scope = RT_SCOPE_UNIVERSE;
5009 
5010         if (ipv6_addr_scope(&ifaca->aca_addr) & IFA_SITE)
5011                 scope = RT_SCOPE_SITE;
5012 
5013         nlh = nlmsg_put(skb, args->portid, args->seq, args->event,
5014                         sizeof(struct ifaddrmsg), args->flags);
5015         if (!nlh)
5016                 return -EMSGSIZE;
5017 
5018         if (args->netnsid >= 0 &&
5019             nla_put_s32(skb, IFA_TARGET_NETNSID, args->netnsid))
5020                 return -EMSGSIZE;
5021 
5022         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
5023         if (nla_put_in6_addr(skb, IFA_ANYCAST, &ifaca->aca_addr) < 0 ||
5024             put_cacheinfo(skb, ifaca->aca_cstamp, ifaca->aca_tstamp,
5025                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
5026                 nlmsg_cancel(skb, nlh);
5027                 return -EMSGSIZE;
5028         }
5029 
5030         nlmsg_end(skb, nlh);
5031         return 0;
5032 }
5033 
5034 /* called with rcu_read_lock() */
5035 static int in6_dump_addrs(struct inet6_dev *idev, struct sk_buff *skb,
5036                           struct netlink_callback *cb, int s_ip_idx,
5037                           struct inet6_fill_args *fillargs)
5038 {
5039         struct ifmcaddr6 *ifmca;
5040         struct ifacaddr6 *ifaca;
5041         int ip_idx = 0;
5042         int err = 1;
5043 
5044         read_lock_bh(&idev->lock);
5045         switch (fillargs->type) {
5046         case UNICAST_ADDR: {
5047                 struct inet6_ifaddr *ifa;
5048                 fillargs->event = RTM_NEWADDR;
5049 
5050                 /* unicast address incl. temp addr */
5051                 list_for_each_entry(ifa, &idev->addr_list, if_list) {
5052                         if (ip_idx < s_ip_idx)
5053                                 goto next;
5054                         err = inet6_fill_ifaddr(skb, ifa, fillargs);
5055                         if (err < 0)
5056                                 break;
5057                         nl_dump_check_consistent(cb, nlmsg_hdr(skb));
5058 next:
5059                         ip_idx++;
5060                 }
5061                 break;
5062         }
5063         case MULTICAST_ADDR:
5064                 fillargs->event = RTM_GETMULTICAST;
5065 
5066                 /* multicast address */
5067                 for (ifmca = idev->mc_list; ifmca;
5068                      ifmca = ifmca->next, ip_idx++) {
5069                         if (ip_idx < s_ip_idx)
5070                                 continue;
5071                         err = inet6_fill_ifmcaddr(skb, ifmca, fillargs);
5072                         if (err < 0)
5073                                 break;
5074                 }
5075                 break;
5076         case ANYCAST_ADDR:
5077                 fillargs->event = RTM_GETANYCAST;
5078                 /* anycast address */
5079                 for (ifaca = idev->ac_list; ifaca;
5080                      ifaca = ifaca->aca_next, ip_idx++) {
5081                         if (ip_idx < s_ip_idx)
5082                                 continue;
5083                         err = inet6_fill_ifacaddr(skb, ifaca, fillargs);
5084                         if (err < 0)
5085                                 break;
5086                 }
5087                 break;
5088         default:
5089                 break;
5090         }
5091         read_unlock_bh(&idev->lock);
5092         cb->args[2] = ip_idx;
5093         return err;
5094 }
5095 
5096 static int inet6_valid_dump_ifaddr_req(const struct nlmsghdr *nlh,
5097                                        struct inet6_fill_args *fillargs,
5098                                        struct net **tgt_net, struct sock *sk,
5099                                        struct netlink_callback *cb)
5100 {
5101         struct netlink_ext_ack *extack = cb->extack;
5102         struct nlattr *tb[IFA_MAX+1];
5103         struct ifaddrmsg *ifm;
5104         int err, i;
5105 
5106         if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) {
5107                 NL_SET_ERR_MSG_MOD(extack, "Invalid header for address dump request");
5108                 return -EINVAL;
5109         }
5110 
5111         ifm = nlmsg_data(nlh);
5112         if (ifm->ifa_prefixlen || ifm->ifa_flags || ifm->ifa_scope) {
5113                 NL_SET_ERR_MSG_MOD(extack, "Invalid values in header for address dump request");
5114                 return -EINVAL;
5115         }
5116 
5117         fillargs->ifindex = ifm->ifa_index;
5118         if (fillargs->ifindex) {
5119                 cb->answer_flags |= NLM_F_DUMP_FILTERED;
5120                 fillargs->flags |= NLM_F_DUMP_FILTERED;
5121         }
5122 
5123         err = nlmsg_parse_deprecated_strict(nlh, sizeof(*ifm), tb, IFA_MAX,
5124                                             ifa_ipv6_policy, extack);
5125         if (err < 0)
5126                 return err;
5127 
5128         for (i = 0; i <= IFA_MAX; ++i) {
5129                 if (!tb[i])
5130                         continue;
5131 
5132                 if (i == IFA_TARGET_NETNSID) {
5133                         struct net *net;
5134 
5135                         fillargs->netnsid = nla_get_s32(tb[i]);
5136                         net = rtnl_get_net_ns_capable(sk, fillargs->netnsid);
5137                         if (IS_ERR(net)) {
5138                                 fillargs->netnsid = -1;
5139                                 NL_SET_ERR_MSG_MOD(extack, "Invalid target network namespace id");
5140                                 return PTR_ERR(net);
5141                         }
5142                         *tgt_net = net;
5143                 } else {
5144                         NL_SET_ERR_MSG_MOD(extack, "Unsupported attribute in dump request");
5145                         return -EINVAL;
5146                 }
5147         }
5148 
5149         return 0;
5150 }
5151 
5152 static int inet6_dump_addr(struct sk_buff *skb, struct netlink_callback *cb,
5153                            enum addr_type_t type)
5154 {
5155         const struct nlmsghdr *nlh = cb->nlh;
5156         struct inet6_fill_args fillargs = {
5157                 .portid = NETLINK_CB(cb->skb).portid,
5158                 .seq = cb->nlh->nlmsg_seq,
5159                 .flags = NLM_F_MULTI,
5160                 .netnsid = -1,
5161                 .type = type,
5162         };
5163         struct net *net = sock_net(skb->sk);
5164         struct net *tgt_net = net;
5165         int idx, s_idx, s_ip_idx;
5166         int h, s_h;
5167         struct net_device *dev;
5168         struct inet6_dev *idev;
5169         struct hlist_head *head;
5170         int err = 0;
5171 
5172         s_h = cb->args[0];
5173         s_idx = idx = cb->args[1];
5174         s_ip_idx = cb->args[2];
5175 
5176         if (cb->strict_check) {
5177                 err = inet6_valid_dump_ifaddr_req(nlh, &fillargs, &tgt_net,
5178                                                   skb->sk, cb);
5179                 if (err < 0)
5180                         goto put_tgt_net;
5181 
5182                 err = 0;
5183                 if (fillargs.ifindex) {
5184                         dev = __dev_get_by_index(tgt_net, fillargs.ifindex);
5185                         if (!dev) {
5186                                 err = -ENODEV;
5187                                 goto put_tgt_net;
5188                         }
5189                         idev = __in6_dev_get(dev);
5190                         if (idev) {
5191                                 err = in6_dump_addrs(idev, skb, cb, s_ip_idx,
5192                                                      &fillargs);
5193                                 if (err > 0)
5194                                         err = 0;
5195                         }
5196                         goto put_tgt_net;
5197                 }
5198         }
5199 
5200         rcu_read_lock();
5201         cb->seq = atomic_read(&tgt_net->ipv6.dev_addr_genid) ^ tgt_net->dev_base_seq;
5202         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
5203                 idx = 0;
5204                 head = &tgt_net->dev_index_head[h];
5205                 hlist_for_each_entry_rcu(dev, head, index_hlist) {
5206                         if (idx < s_idx)
5207                                 goto cont;
5208                         if (h > s_h || idx > s_idx)
5209                                 s_ip_idx = 0;
5210                         idev = __in6_dev_get(dev);
5211                         if (!idev)
5212                                 goto cont;
5213 
5214                         if (in6_dump_addrs(idev, skb, cb, s_ip_idx,
5215                                            &fillargs) < 0)
5216                                 goto done;
5217 cont:
5218                         idx++;
5219                 }
5220         }
5221 done:
5222         rcu_read_unlock();
5223         cb->args[0] = h;
5224         cb->args[1] = idx;
5225 put_tgt_net:
5226         if (fillargs.netnsid >= 0)
5227                 put_net(tgt_net);
5228 
5229         return skb->len ? : err;
5230 }
5231 
5232 static int inet6_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
5233 {
5234         enum addr_type_t type = UNICAST_ADDR;
5235 
5236         return inet6_dump_addr(skb, cb, type);
5237 }
5238 
5239 static int inet6_dump_ifmcaddr(struct sk_buff *skb, struct netlink_callback *cb)
5240 {
5241         enum addr_type_t type = MULTICAST_ADDR;
5242 
5243         return inet6_dump_addr(skb, cb, type);
5244 }
5245 
5246 
5247 static int inet6_dump_ifacaddr(struct sk_buff *skb, struct netlink_callback *cb)
5248 {
5249         enum addr_type_t type = ANYCAST_ADDR;
5250 
5251         return inet6_dump_addr(skb, cb, type);
5252 }
5253 
5254 static int inet6_rtm_valid_getaddr_req(struct sk_buff *skb,
5255                                        const struct nlmsghdr *nlh,
5256                                        struct nlattr **tb,
5257                                        struct netlink_ext_ack *extack)
5258 {
5259         struct ifaddrmsg *ifm;
5260         int i, err;
5261 
5262         if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) {
5263                 NL_SET_ERR_MSG_MOD(extack, "Invalid header for get address request");
5264                 return -EINVAL;
5265         }
5266 
5267         if (!netlink_strict_get_check(skb))
5268                 return nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFA_MAX,
5269                                               ifa_ipv6_policy, extack);
5270 
5271         ifm = nlmsg_data(nlh);
5272         if (ifm->ifa_prefixlen || ifm->ifa_flags || ifm->ifa_scope) {
5273                 NL_SET_ERR_MSG_MOD(extack, "Invalid values in header for get address request");
5274                 return -EINVAL;
5275         }
5276 
5277         err = nlmsg_parse_deprecated_strict(nlh, sizeof(*ifm), tb, IFA_MAX,
5278                                             ifa_ipv6_policy, extack);
5279         if (err)
5280                 return err;
5281 
5282         for (i = 0; i <= IFA_MAX; i++) {
5283                 if (!tb[i])
5284                         continue;
5285 
5286                 switch (i) {
5287                 case IFA_TARGET_NETNSID:
5288                 case IFA_ADDRESS:
5289                 case IFA_LOCAL:
5290                         break;
5291                 default:
5292                         NL_SET_ERR_MSG_MOD(extack, "Unsupported attribute in get address request");
5293                         return -EINVAL;
5294                 }
5295         }
5296 
5297         return 0;
5298 }
5299 
5300 static int inet6_rtm_getaddr(struct sk_buff *in_skb, struct nlmsghdr *nlh,
5301                              struct netlink_ext_ack *extack)
5302 {
5303         struct net *net = sock_net(in_skb->sk);
5304         struct inet6_fill_args fillargs = {
5305                 .portid = NETLINK_CB(in_skb).portid,
5306                 .seq = nlh->nlmsg_seq,
5307                 .event = RTM_NEWADDR,
5308                 .flags = 0,
5309                 .netnsid = -1,
5310         };
5311         struct net *tgt_net = net;
5312         struct ifaddrmsg *ifm;
5313         struct nlattr *tb[IFA_MAX+1];
5314         struct in6_addr *addr = NULL, *peer;
5315         struct net_device *dev = NULL;
5316         struct inet6_ifaddr *ifa;
5317         struct sk_buff *skb;
5318         int err;
5319 
5320         err = inet6_rtm_valid_getaddr_req(in_skb, nlh, tb, extack);
5321         if (err < 0)
5322                 return err;
5323 
5324         if (tb[IFA_TARGET_NETNSID]) {
5325                 fillargs.netnsid = nla_get_s32(tb[IFA_TARGET_NETNSID]);
5326 
5327                 tgt_net = rtnl_get_net_ns_capable(NETLINK_CB(in_skb).sk,
5328                                                   fillargs.netnsid);
5329                 if (IS_ERR(tgt_net))
5330                         return PTR_ERR(tgt_net);
5331         }
5332 
5333         addr = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer);
5334         if (!addr)
5335                 return -EINVAL;
5336 
5337         ifm = nlmsg_data(nlh);
5338         if (ifm->ifa_index)
5339                 dev = dev_get_by_index(tgt_net, ifm->ifa_index);
5340 
5341         ifa = ipv6_get_ifaddr(tgt_net, addr, dev, 1);
5342         if (!ifa) {
5343                 err = -EADDRNOTAVAIL;
5344                 goto errout;
5345         }
5346 
5347         skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_KERNEL);
5348         if (!skb) {
5349                 err = -ENOBUFS;
5350                 goto errout_ifa;
5351         }
5352 
5353         err = inet6_fill_ifaddr(skb, ifa, &fillargs);
5354         if (err < 0) {
5355                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
5356                 WARN_ON(err == -EMSGSIZE);
5357                 kfree_skb(skb);
5358                 goto errout_ifa;
5359         }
5360         err = rtnl_unicast(skb, tgt_net, NETLINK_CB(in_skb).portid);
5361 errout_ifa:
5362         in6_ifa_put(ifa);
5363 errout:
5364         if (dev)
5365                 dev_put(dev);
5366         if (fillargs.netnsid >= 0)
5367                 put_net(tgt_net);
5368 
5369         return err;
5370 }
5371 
5372 static void inet6_ifa_notify(int event, struct inet6_ifaddr *ifa)
5373 {
5374         struct sk_buff *skb;
5375         struct net *net = dev_net(ifa->idev->dev);
5376         struct inet6_fill_args fillargs = {
5377                 .portid = 0,
5378                 .seq = 0,
5379                 .event = event,
5380                 .flags = 0,
5381                 .netnsid = -1,
5382         };
5383         int err = -ENOBUFS;
5384 
5385         skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_ATOMIC);
5386         if (!skb)
5387                 goto errout;
5388 
5389         err = inet6_fill_ifaddr(skb, ifa, &fillargs);
5390         if (err < 0) {
5391                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
5392                 WARN_ON(err == -EMSGSIZE);
5393                 kfree_skb(skb);
5394                 goto errout;
5395         }
5396         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
5397         return;
5398 errout:
5399         if (err < 0)
5400                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
5401 }
5402 
5403 static inline void ipv6_store_devconf(struct ipv6_devconf *cnf,
5404                                 __s32 *array, int bytes)
5405 {
5406         BUG_ON(bytes < (DEVCONF_MAX * 4));
5407 
5408         memset(array, 0, bytes);
5409         array[DEVCONF_FORWARDING] = cnf->forwarding;
5410         array[DEVCONF_HOPLIMIT] = cnf->hop_limit;
5411         array[DEVCONF_MTU6] = cnf->mtu6;
5412         array[DEVCONF_ACCEPT_RA] = cnf->accept_ra;
5413         array[DEVCONF_ACCEPT_REDIRECTS] = cnf->accept_redirects;
5414         array[DEVCONF_AUTOCONF] = cnf->autoconf;
5415         array[DEVCONF_DAD_TRANSMITS] = cnf->dad_transmits;
5416         array[DEVCONF_RTR_SOLICITS] = cnf->rtr_solicits;
5417         array[DEVCONF_RTR_SOLICIT_INTERVAL] =
5418                 jiffies_to_msecs(cnf->rtr_solicit_interval);
5419         array[DEVCONF_RTR_SOLICIT_MAX_INTERVAL] =
5420                 jiffies_to_msecs(cnf->rtr_solicit_max_interval);
5421         array[DEVCONF_RTR_SOLICIT_DELAY] =
5422                 jiffies_to_msecs(cnf->rtr_solicit_delay);
5423         array[DEVCONF_FORCE_MLD_VERSION] = cnf->force_mld_version;
5424         array[DEVCONF_MLDV1_UNSOLICITED_REPORT_INTERVAL] =
5425                 jiffies_to_msecs(cnf->mldv1_unsolicited_report_interval);
5426         array[DEVCONF_MLDV2_UNSOLICITED_REPORT_INTERVAL] =
5427                 jiffies_to_msecs(cnf->mldv2_unsolicited_report_interval);
5428         array[DEVCONF_USE_TEMPADDR] = cnf->use_tempaddr;
5429         array[DEVCONF_TEMP_VALID_LFT] = cnf->temp_valid_lft;
5430         array[DEVCONF_TEMP_PREFERED_LFT] = cnf->temp_prefered_lft;
5431         array[DEVCONF_REGEN_MAX_RETRY] = cnf->regen_max_retry;
5432         array[DEVCONF_MAX_DESYNC_FACTOR] = cnf->max_desync_factor;
5433         array[DEVCONF_MAX_ADDRESSES] = cnf->max_addresses;
5434         array[DEVCONF_ACCEPT_RA_DEFRTR] = cnf->accept_ra_defrtr;
5435         array[DEVCONF_ACCEPT_RA_MIN_HOP_LIMIT] = cnf->accept_ra_min_hop_limit;
5436         array[DEVCONF_ACCEPT_RA_PINFO] = cnf->accept_ra_pinfo;
5437 #ifdef CONFIG_IPV6_ROUTER_PREF
5438         array[DEVCONF_ACCEPT_RA_RTR_PREF] = cnf->accept_ra_rtr_pref;
5439         array[DEVCONF_RTR_PROBE_INTERVAL] =
5440                 jiffies_to_msecs(cnf->rtr_probe_interval);
5441 #ifdef CONFIG_IPV6_ROUTE_INFO
5442         array[DEVCONF_ACCEPT_RA_RT_INFO_MIN_PLEN] = cnf->accept_ra_rt_info_min_plen;
5443         array[DEVCONF_ACCEPT_RA_RT_INFO_MAX_PLEN] = cnf->accept_ra_rt_info_max_plen;
5444 #endif
5445 #endif
5446         array[DEVCONF_PROXY_NDP] = cnf->proxy_ndp;
5447         array[DEVCONF_ACCEPT_SOURCE_ROUTE] = cnf->accept_source_route;
5448 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
5449         array[DEVCONF_OPTIMISTIC_DAD] = cnf->optimistic_dad;
5450         array[DEVCONF_USE_OPTIMISTIC] = cnf->use_optimistic;
5451 #endif
5452 #ifdef CONFIG_IPV6_MROUTE
5453         array[DEVCONF_MC_FORWARDING] = cnf->mc_forwarding;
5454 #endif
5455         array[DEVCONF_DISABLE_IPV6] = cnf->disable_ipv6;
5456         array[DEVCONF_ACCEPT_DAD] = cnf->accept_dad;
5457         array[DEVCONF_FORCE_TLLAO] = cnf->force_tllao;
5458         array[DEVCONF_NDISC_NOTIFY] = cnf->ndisc_notify;
5459         array[DEVCONF_SUPPRESS_FRAG_NDISC] = cnf->suppress_frag_ndisc;
5460         array[DEVCONF_ACCEPT_RA_FROM_LOCAL] = cnf->accept_ra_from_local;
5461         array[DEVCONF_ACCEPT_RA_MTU] = cnf->accept_ra_mtu;
5462         array[DEVCONF_IGNORE_ROUTES_WITH_LINKDOWN] = cnf->ignore_routes_with_linkdown;
5463         /* we omit DEVCONF_STABLE_SECRET for now */
5464         array[DEVCONF_USE_OIF_ADDRS_ONLY] = cnf->use_oif_addrs_only;
5465         array[DEVCONF_DROP_UNICAST_IN_L2_MULTICAST] = cnf->drop_unicast_in_l2_multicast;
5466         array[DEVCONF_DROP_UNSOLICITED_NA] = cnf->drop_unsolicited_na;
5467         array[DEVCONF_KEEP_ADDR_ON_DOWN] = cnf->keep_addr_on_down;
5468         array[DEVCONF_SEG6_ENABLED] = cnf->seg6_enabled;
5469 #ifdef CONFIG_IPV6_SEG6_HMAC
5470         array[DEVCONF_SEG6_REQUIRE_HMAC] = cnf->seg6_require_hmac;
5471 #endif
5472         array[DEVCONF_ENHANCED_DAD] = cnf->enhanced_dad;
5473         array[DEVCONF_ADDR_GEN_MODE] = cnf->addr_gen_mode;
5474         array[DEVCONF_DISABLE_POLICY] = cnf->disable_policy;
5475         array[DEVCONF_NDISC_TCLASS] = cnf->ndisc_tclass;
5476 }
5477 
5478 static inline size_t inet6_ifla6_size(void)
5479 {
5480         return nla_total_size(4) /* IFLA_INET6_FLAGS */
5481              + nla_total_size(sizeof(struct ifla_cacheinfo))
5482              + nla_total_size(DEVCONF_MAX * 4) /* IFLA_INET6_CONF */
5483              + nla_total_size(IPSTATS_MIB_MAX * 8) /* IFLA_INET6_STATS */
5484              + nla_total_size(ICMP6_MIB_MAX * 8) /* IFLA_INET6_ICMP6STATS */
5485              + nla_total_size(sizeof(struct in6_addr)) /* IFLA_INET6_TOKEN */
5486              + nla_total_size(1) /* IFLA_INET6_ADDR_GEN_MODE */
5487              + 0;
5488 }
5489 
5490 static inline size_t inet6_if_nlmsg_size(void)
5491 {
5492         return NLMSG_ALIGN(sizeof(struct ifinfomsg))
5493                + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
5494                + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
5495                + nla_total_size(4) /* IFLA_MTU */
5496                + nla_total_size(4) /* IFLA_LINK */
5497                + nla_total_size(1) /* IFLA_OPERSTATE */
5498                + nla_total_size(inet6_ifla6_size()); /* IFLA_PROTINFO */
5499 }
5500 
5501 static inline void __snmp6_fill_statsdev(u64 *stats, atomic_long_t *mib,
5502                                         int bytes)
5503 {
5504         int i;
5505         int pad = bytes - sizeof(u64) * ICMP6_MIB_MAX;
5506         BUG_ON(pad < 0);
5507 
5508         /* Use put_unaligned() because stats may not be aligned for u64. */
5509         put_unaligned(ICMP6_MIB_MAX, &stats[0]);
5510         for (i = 1; i < ICMP6_MIB_MAX; i++)
5511                 put_unaligned(atomic_long_read(&mib[i]), &stats[i]);
5512 
5513         memset(&stats[ICMP6_MIB_MAX], 0, pad);
5514 }
5515 
5516 static inline void __snmp6_fill_stats64(u64 *stats, void __percpu *mib,
5517                                         int bytes, size_t syncpoff)
5518 {
5519         int i, c;
5520         u64 buff[IPSTATS_MIB_MAX];
5521         int pad = bytes - sizeof(u64) * IPSTATS_MIB_MAX;
5522 
5523         BUG_ON(pad < 0);
5524 
5525         memset(buff, 0, sizeof(buff));
5526         buff[0] = IPSTATS_MIB_MAX;
5527 
5528         for_each_possible_cpu(c) {
5529                 for (i = 1; i < IPSTATS_MIB_MAX; i++)
5530                         buff[i] += snmp_get_cpu_field64(mib, c, i, syncpoff);
5531         }
5532 
5533         memcpy(stats, buff, IPSTATS_MIB_MAX * sizeof(u64));
5534         memset(&stats[IPSTATS_MIB_MAX], 0, pad);
5535 }
5536 
5537 static void snmp6_fill_stats(u64 *stats, struct inet6_dev *idev, int attrtype,
5538                              int bytes)
5539 {
5540         switch (attrtype) {
5541         case IFLA_INET6_STATS:
5542                 __snmp6_fill_stats64(stats, idev->stats.ipv6, bytes,
5543                                      offsetof(struct ipstats_mib, syncp));
5544                 break;
5545         case IFLA_INET6_ICMP6STATS:
5546                 __snmp6_fill_statsdev(stats, idev->stats.icmpv6dev->mibs, bytes);
5547                 break;
5548         }
5549 }
5550 
5551 static int inet6_fill_ifla6_attrs(struct sk_buff *skb, struct inet6_dev *idev,
5552                                   u32 ext_filter_mask)
5553 {
5554         struct nlattr *nla;
5555         struct ifla_cacheinfo ci;
5556 
5557         if (nla_put_u32(skb, IFLA_INET6_FLAGS, idev->if_flags))
5558                 goto nla_put_failure;
5559         ci.max_reasm_len = IPV6_MAXPLEN;
5560         ci.tstamp = cstamp_delta(idev->tstamp);
5561         ci.reachable_time = jiffies_to_msecs(idev->nd_parms->reachable_time);
5562         ci.retrans_time = jiffies_to_msecs(NEIGH_VAR(idev->nd_parms, RETRANS_TIME));
5563         if (nla_put(skb, IFLA_INET6_CACHEINFO, sizeof(ci), &ci))
5564                 goto nla_put_failure;
5565         nla = nla_reserve(skb, IFLA_INET6_CONF, DEVCONF_MAX * sizeof(s32));
5566         if (!nla)
5567                 goto nla_put_failure;
5568         ipv6_store_devconf(&idev->cnf, nla_data(nla), nla_len(nla));
5569 
5570         /* XXX - MC not implemented */
5571 
5572         if (ext_filter_mask & RTEXT_FILTER_SKIP_STATS)
5573                 return 0;
5574 
5575         nla = nla_reserve(skb, IFLA_INET6_STATS, IPSTATS_MIB_MAX * sizeof(u64));
5576         if (!nla)
5577                 goto nla_put_failure;
5578         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_STATS, nla_len(nla));
5579 
5580         nla = nla_reserve(skb, IFLA_INET6_ICMP6STATS, ICMP6_MIB_MAX * sizeof(u64));
5581         if (!nla)
5582                 goto nla_put_failure;
5583         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_ICMP6STATS, nla_len(nla));
5584 
5585         nla = nla_reserve(skb, IFLA_INET6_TOKEN, sizeof(struct in6_addr));
5586         if (!nla)
5587                 goto nla_put_failure;
5588         read_lock_bh(&idev->lock);
5589         memcpy(nla_data(nla), idev->token.s6_addr, nla_len(nla));
5590         read_unlock_bh(&idev->lock);
5591 
5592         if (nla_put_u8(skb, IFLA_INET6_ADDR_GEN_MODE, idev->cnf.addr_gen_mode))
5593                 goto nla_put_failure;
5594 
5595         return 0;
5596 
5597 nla_put_failure:
5598         return -EMSGSIZE;
5599 }
5600 
5601 static size_t inet6_get_link_af_size(const struct net_device *dev,
5602                                      u32 ext_filter_mask)
5603 {
5604         if (!__in6_dev_get(dev))
5605                 return 0;
5606 
5607         return inet6_ifla6_size();
5608 }
5609 
5610 static int inet6_fill_link_af(struct sk_buff *skb, const struct net_device *dev,
5611                               u32 ext_filter_mask)
5612 {
5613         struct inet6_dev *idev = __in6_dev_get(dev);
5614 
5615         if (!idev)
5616                 return -ENODATA;
5617 
5618         if (inet6_fill_ifla6_attrs(skb, idev, ext_filter_mask) < 0)
5619                 return -EMSGSIZE;
5620 
5621         return 0;
5622 }
5623 
5624 static int inet6_set_iftoken(struct inet6_dev *idev, struct in6_addr *token)
5625 {
5626         struct inet6_ifaddr *ifp;
5627         struct net_device *dev = idev->dev;
5628         bool clear_token, update_rs = false;
5629         struct in6_addr ll_addr;
5630 
5631         ASSERT_RTNL();
5632 
5633         if (!token)
5634                 return -EINVAL;
5635         if (dev->flags & (IFF_LOOPBACK | IFF_NOARP))
5636                 return -EINVAL;
5637         if (!ipv6_accept_ra(idev))
5638                 return -EINVAL;
5639         if (idev->cnf.rtr_solicits == 0)
5640                 return -EINVAL;
5641 
5642         write_lock_bh(&idev->lock);
5643 
5644         BUILD_BUG_ON(sizeof(token->s6_addr) != 16);
5645         memcpy(idev->token.s6_addr + 8, token->s6_addr + 8, 8);
5646 
5647         write_unlock_bh(&idev->lock);
5648 
5649         clear_token = ipv6_addr_any(token);
5650         if (clear_token)
5651                 goto update_lft;
5652 
5653         if (!idev->dead && (idev->if_flags & IF_READY) &&
5654             !ipv6_get_lladdr(dev, &ll_addr, IFA_F_TENTATIVE |
5655                              IFA_F_OPTIMISTIC)) {
5656                 /* If we're not ready, then normal ifup will take care
5657                  * of this. Otherwise, we need to request our rs here.
5658                  */
5659                 ndisc_send_rs(dev, &ll_addr, &in6addr_linklocal_allrouters);
5660                 update_rs = true;
5661         }
5662 
5663 update_lft:
5664         write_lock_bh(&idev->lock);
5665 
5666         if (update_rs) {
5667                 idev->if_flags |= IF_RS_SENT;
5668                 idev->rs_interval = rfc3315_s14_backoff_init(
5669                         idev->cnf.rtr_solicit_interval);
5670                 idev->rs_probes = 1;
5671                 addrconf_mod_rs_timer(idev, idev->rs_interval);
5672         }
5673 
5674         /* Well, that's kinda nasty ... */
5675         list_for_each_entry(ifp, &idev->addr_list, if_list) {
5676                 spin_lock(&ifp->lock);
5677                 if (ifp->tokenized) {
5678                         ifp->valid_lft = 0;
5679                         ifp->prefered_lft = 0;
5680                 }
5681                 spin_unlock(&ifp->lock);
5682         }
5683 
5684         write_unlock_bh(&idev->lock);
5685         inet6_ifinfo_notify(RTM_NEWLINK, idev);
5686         addrconf_verify_rtnl();
5687         return 0;
5688 }
5689 
5690 static const struct nla_policy inet6_af_policy[IFLA_INET6_MAX + 1] = {
5691         [IFLA_INET6_ADDR_GEN_MODE]      = { .type = NLA_U8 },
5692         [IFLA_INET6_TOKEN]              = { .len = sizeof(struct in6_addr) },
5693 };
5694 
5695 static int check_addr_gen_mode(int mode)
5696 {
5697         if (mode != IN6_ADDR_GEN_MODE_EUI64 &&
5698             mode != IN6_ADDR_GEN_MODE_NONE &&
5699             mode != IN6_ADDR_GEN_MODE_STABLE_PRIVACY &&
5700             mode != IN6_ADDR_GEN_MODE_RANDOM)
5701                 return -EINVAL;
5702         return 1;
5703 }
5704 
5705 static int check_stable_privacy(struct inet6_dev *idev, struct net *net,
5706                                 int mode)
5707 {
5708         if (mode == IN6_ADDR_GEN_MODE_STABLE_PRIVACY &&
5709             !idev->cnf.stable_secret.initialized &&
5710             !net->ipv6.devconf_dflt->stable_secret.initialized)
5711                 return -EINVAL;
5712         return 1;
5713 }
5714 
5715 static int inet6_validate_link_af(const struct net_device *dev,
5716                                   const struct nlattr *nla)
5717 {
5718         struct nlattr *tb[IFLA_INET6_MAX + 1];
5719         struct inet6_dev *idev = NULL;
5720         int err;
5721 
5722         if (dev) {
5723                 idev = __in6_dev_get(dev);
5724                 if (!idev)
5725                         return -EAFNOSUPPORT;
5726         }
5727 
5728         err = nla_parse_nested_deprecated(tb, IFLA_INET6_MAX, nla,
5729                                           inet6_af_policy, NULL);
5730         if (err)
5731                 return err;
5732 
5733         if (!tb[IFLA_INET6_TOKEN] && !tb[IFLA_INET6_ADDR_GEN_MODE])
5734                 return -EINVAL;
5735 
5736         if (tb[IFLA_INET6_ADDR_GEN_MODE]) {
5737                 u8 mode = nla_get_u8(tb[IFLA_INET6_ADDR_GEN_MODE]);
5738 
5739                 if (check_addr_gen_mode(mode) < 0)
5740                         return -EINVAL;
5741                 if (dev && check_stable_privacy(idev, dev_net(dev), mode) < 0)
5742                         return -EINVAL;
5743         }
5744 
5745         return 0;
5746 }
5747 
5748 static int inet6_set_link_af(struct net_device *dev, const struct nlattr *nla)
5749 {
5750         struct inet6_dev *idev = __in6_dev_get(dev);
5751         struct nlattr *tb[IFLA_INET6_MAX + 1];
5752         int err;
5753 
5754         if (!idev)
5755                 return -EAFNOSUPPORT;
5756 
5757         if (nla_parse_nested_deprecated(tb, IFLA_INET6_MAX, nla, NULL, NULL) < 0)
5758                 BUG();
5759 
5760         if (tb[IFLA_INET6_TOKEN]) {
5761                 err = inet6_set_iftoken(idev, nla_data(tb[IFLA_INET6_TOKEN]));
5762                 if (err)
5763                         return err;
5764         }
5765 
5766         if (tb[IFLA_INET6_ADDR_GEN_MODE]) {
5767                 u8 mode = nla_get_u8(tb[IFLA_INET6_ADDR_GEN_MODE]);
5768 
5769                 idev->cnf.addr_gen_mode = mode;
5770         }
5771 
5772         return 0;
5773 }
5774 
5775 static int inet6_fill_ifinfo(struct sk_buff *skb, struct inet6_dev *idev,
5776                              u32 portid, u32 seq, int event, unsigned int flags)
5777 {
5778         struct net_device *dev = idev->dev;
5779         struct ifinfomsg *hdr;
5780         struct nlmsghdr *nlh;
5781         void *protoinfo;
5782 
5783         nlh = nlmsg_put(skb, portid, seq, event, sizeof(*hdr), flags);
5784         if (!nlh)
5785                 return -EMSGSIZE;
5786 
5787         hdr = nlmsg_data(nlh);
5788         hdr->ifi_family = AF_INET6;
5789         hdr->__ifi_pad = 0;
5790         hdr->ifi_type = dev->type;
5791         hdr->ifi_index = dev->ifindex;
5792         hdr->ifi_flags = dev_get_flags(dev);
5793         hdr->ifi_change = 0;
5794 
5795         if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
5796             (dev->addr_len &&
5797              nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
5798             nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
5799             (dev->ifindex != dev_get_iflink(dev) &&
5800              nla_put_u32(skb, IFLA_LINK, dev_get_iflink(dev))) ||
5801             nla_put_u8(skb, IFLA_OPERSTATE,
5802                        netif_running(dev) ? dev->operstate : IF_OPER_DOWN))
5803                 goto nla_put_failure;
5804         protoinfo = nla_nest_start_noflag(skb, IFLA_PROTINFO);
5805         if (!protoinfo)
5806                 goto nla_put_failure;
5807 
5808         if (inet6_fill_ifla6_attrs(skb, idev, 0) < 0)
5809                 goto nla_put_failure;
5810 
5811         nla_nest_end(skb, protoinfo);
5812         nlmsg_end(skb, nlh);
5813         return 0;
5814 
5815 nla_put_failure:
5816         nlmsg_cancel(skb, nlh);
5817         return -EMSGSIZE;
5818 }
5819 
5820 static int inet6_valid_dump_ifinfo(const struct nlmsghdr *nlh,
5821                                    struct netlink_ext_ack *extack)
5822 {
5823         struct ifinfomsg *ifm;
5824 
5825         if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) {
5826                 NL_SET_ERR_MSG_MOD(extack, "Invalid header for link dump request");
5827                 return -EINVAL;
5828         }
5829 
5830         if (nlmsg_attrlen(nlh, sizeof(*ifm))) {
5831                 NL_SET_ERR_MSG_MOD(extack, "Invalid data after header");
5832                 return -EINVAL;
5833         }
5834 
5835         ifm = nlmsg_data(nlh);
5836         if (ifm->__ifi_pad || ifm->ifi_type || ifm->ifi_flags ||
5837             ifm->ifi_change || ifm->ifi_index) {
5838                 NL_SET_ERR_MSG_MOD(extack, "Invalid values in header for dump request");
5839                 return -EINVAL;
5840         }
5841 
5842         return 0;
5843 }
5844 
5845 static int inet6_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
5846 {
5847         struct net *net = sock_net(skb->sk);
5848         int h, s_h;
5849         int idx = 0, s_idx;
5850         struct net_device *dev;
5851         struct inet6_dev *idev;
5852         struct hlist_head *head;
5853 
5854         /* only requests using strict checking can pass data to
5855          * influence the dump
5856          */
5857         if (cb->strict_check) {
5858                 int err = inet6_valid_dump_ifinfo(cb->nlh, cb->extack);
5859 
5860                 if (err < 0)
5861                         return err;
5862         }
5863 
5864         s_h = cb->args[0];
5865         s_idx = cb->args[1];
5866 
5867         rcu_read_lock();
5868         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
5869                 idx = 0;
5870                 head = &net->dev_index_head[h];
5871                 hlist_for_each_entry_rcu(dev, head, index_hlist) {
5872                         if (idx < s_idx)
5873                                 goto cont;
5874                         idev = __in6_dev_get(dev);
5875                         if (!idev)
5876                                 goto cont;
5877                         if (inet6_fill_ifinfo(skb, idev,
5878                                               NETLINK_CB(cb->skb).portid,
5879                                               cb->nlh->nlmsg_seq,
5880                                               RTM_NEWLINK, NLM_F_MULTI) < 0)
5881                                 goto out;
5882 cont:
5883                         idx++;
5884                 }
5885         }
5886 out:
5887         rcu_read_unlock();
5888         cb->args[1] = idx;
5889         cb->args[0] = h;
5890 
5891         return skb->len;
5892 }
5893 
5894 void inet6_ifinfo_notify(int event, struct inet6_dev *idev)
5895 {
5896         struct sk_buff *skb;
5897         struct net *net = dev_net(idev->dev);
5898         int err = -ENOBUFS;
5899 
5900         skb = nlmsg_new(inet6_if_nlmsg_size(), GFP_ATOMIC);
5901         if (!skb)
5902                 goto errout;
5903 
5904         err = inet6_fill_ifinfo(skb, idev, 0, 0, event, 0);
5905         if (err < 0) {
5906                 /* -EMSGSIZE implies BUG in inet6_if_nlmsg_size() */
5907                 WARN_ON(err == -EMSGSIZE);
5908                 kfree_skb(skb);
5909                 goto errout;
5910         }
5911         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFINFO, NULL, GFP_ATOMIC);
5912         return;
5913 errout:
5914         if (err < 0)
5915                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFINFO, err);
5916 }
5917 
5918 static inline size_t inet6_prefix_nlmsg_size(void)
5919 {
5920         return NLMSG_ALIGN(sizeof(struct prefixmsg))
5921                + nla_total_size(sizeof(struct in6_addr))
5922                + nla_total_size(sizeof(struct prefix_cacheinfo));
5923 }
5924 
5925 static int inet6_fill_prefix(struct sk_buff *skb, struct inet6_dev *idev,
5926                              struct prefix_info *pinfo, u32 portid, u32 seq,
5927                              int event, unsigned int flags)
5928 {
5929         struct prefixmsg *pmsg;
5930         struct nlmsghdr *nlh;
5931         struct prefix_cacheinfo ci;
5932 
5933         nlh = nlmsg_put(skb, portid, seq, event, sizeof(*pmsg), flags);
5934         if (!nlh)
5935                 return -EMSGSIZE;
5936 
5937         pmsg = nlmsg_data(nlh);
5938         pmsg->prefix_family = AF_INET6;
5939         pmsg->prefix_pad1 = 0;
5940         pmsg->prefix_pad2 = 0;
5941         pmsg->prefix_ifindex = idev->dev->ifindex;
5942         pmsg->prefix_len = pinfo->prefix_len;
5943         pmsg->prefix_type = pinfo->type;
5944         pmsg->prefix_pad3 = 0;
5945         pmsg->prefix_flags = 0;
5946         if (pinfo->onlink)
5947                 pmsg->prefix_flags |= IF_PREFIX_ONLINK;
5948         if (pinfo->autoconf)
5949                 pmsg->prefix_flags |= IF_PREFIX_AUTOCONF;
5950 
5951         if (nla_put(skb, PREFIX_ADDRESS, sizeof(pinfo->prefix), &pinfo->prefix))
5952                 goto nla_put_failure;
5953         ci.preferred_time = ntohl(pinfo->prefered);
5954         ci.valid_time = ntohl(pinfo->valid);
5955         if (nla_put(skb, PREFIX_CACHEINFO, sizeof(ci), &ci))
5956                 goto nla_put_failure;
5957         nlmsg_end(skb, nlh);
5958         return 0;
5959 
5960 nla_put_failure:
5961         nlmsg_cancel(skb, nlh);
5962         return -EMSGSIZE;
5963 }
5964 
5965 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
5966                          struct prefix_info *pinfo)
5967 {
5968         struct sk_buff *skb;
5969         struct net *net = dev_net(idev->dev);
5970         int err = -ENOBUFS;
5971 
5972         skb = nlmsg_new(inet6_prefix_nlmsg_size(), GFP_ATOMIC);
5973         if (!skb)
5974                 goto errout;
5975 
5976         err = inet6_fill_prefix(skb, idev, pinfo, 0, 0, event, 0);
5977         if (err < 0) {
5978                 /* -EMSGSIZE implies BUG in inet6_prefix_nlmsg_size() */
5979                 WARN_ON(err == -EMSGSIZE);
5980                 kfree_skb(skb);
5981                 goto errout;
5982         }
5983         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_PREFIX, NULL, GFP_ATOMIC);
5984         return;
5985 errout:
5986         if (err < 0)
5987                 rtnl_set_sk_err(net, RTNLGRP_IPV6_PREFIX, err);
5988 }
5989 
5990 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
5991 {
5992         struct net *net = dev_net(ifp->idev->dev);
5993 
5994         if (event)
5995                 ASSERT_RTNL();
5996 
5997         inet6_ifa_notify(event ? : RTM_NEWADDR, ifp);
5998 
5999         switch (event) {
6000         case RTM_NEWADDR:
6001                 /*
6002                  * If the address was optimistic we inserted the route at the
6003                  * start of our DAD process, so we don't need to do it again.
6004                  * If the device was taken down in the middle of the DAD
6005                  * cycle there is a race where we could get here without a
6006                  * host route, so nothing to insert. That will be fixed when
6007                  * the device is brought up.
6008                  */
6009                 if (ifp->rt && !rcu_access_pointer(ifp->rt->fib6_node)) {
6010                         ip6_ins_rt(net, ifp->rt);
6011                 } else if (!ifp->rt && (ifp->idev->dev->flags & IFF_UP)) {
6012                         pr_warn("BUG: Address %pI6c on device %s is missing its host route.\n",
6013                                 &ifp->addr, ifp->idev->dev->name);
6014                 }
6015 
6016                 if (ifp->idev->cnf.forwarding)
6017                         addrconf_join_anycast(ifp);
6018                 if (!ipv6_addr_any(&ifp->peer_addr))
6019                         addrconf_prefix_route(&ifp->peer_addr, 128,
6020                                               ifp->rt_priority, ifp->idev->dev,
6021                                               0, 0, GFP_ATOMIC);
6022                 break;
6023         case RTM_DELADDR:
6024                 if (ifp->idev->cnf.forwarding)
6025                         addrconf_leave_anycast(ifp);
6026                 addrconf_leave_solict(ifp->idev, &ifp->addr);
6027                 if (!ipv6_addr_any(&ifp->peer_addr)) {
6028                         struct fib6_info *rt;
6029 
6030                         rt = addrconf_get_prefix_route(&ifp->peer_addr, 128,
6031                                                        ifp->idev->dev, 0, 0,
6032                                                        false);
6033                         if (rt)
6034                                 ip6_del_rt(net, rt);
6035                 }
6036                 if (ifp->rt) {
6037                         ip6_del_rt(net, ifp->rt);
6038                         ifp->rt = NULL;
6039                 }
6040                 rt_genid_bump_ipv6(net);
6041                 break;
6042         }
6043         atomic_inc(&net->ipv6.dev_addr_genid);
6044 }
6045 
6046 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
6047 {
6048         rcu_read_lock_bh();
6049         if (likely(ifp->idev->dead == 0))
6050                 __ipv6_ifa_notify(event, ifp);
6051         rcu_read_unlock_bh();
6052 }
6053 
6054 #ifdef CONFIG_SYSCTL
6055 
6056 static
6057 int addrconf_sysctl_forward(struct ctl_table *ctl, int write,
6058                            void __user *buffer, size_t *lenp, loff_t *ppos)
6059 {
6060         int *valp = ctl->data;
6061         int val = *valp;
6062         loff_t pos = *ppos;
6063         struct ctl_table lctl;
6064         int ret;
6065 
6066         /*
6067          * ctl->data points to idev->cnf.forwarding, we should
6068          * not modify it until we get the rtnl lock.
6069          */
6070         lctl = *ctl;
6071         lctl.data = &val;
6072 
6073         ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
6074 
6075         if (write)
6076                 ret = addrconf_fixup_forwarding(ctl, valp, val);
6077         if (ret)
6078                 *ppos = pos;
6079         return ret;
6080 }
6081 
6082 static
6083 int addrconf_sysctl_mtu(struct ctl_table *ctl, int write,
6084                         void __user *buffer, size_t *lenp, loff_t *ppos)
6085 {
6086         struct inet6_dev *idev = ctl->extra1;
6087         int min_mtu = IPV6_MIN_MTU;
6088         struct ctl_table lctl;
6089 
6090         lctl = *ctl;
6091         lctl.extra1 = &min_mtu;
6092         lctl.extra2 = idev ? &idev->dev->mtu : NULL;
6093 
6094         return proc_dointvec_minmax(&lctl, write, buffer, lenp, ppos);
6095 }
6096 
6097 static void dev_disable_change(struct inet6_dev *idev)
6098 {
6099         struct netdev_notifier_info info;
6100 
6101         if (!idev || !idev->dev)
6102                 return;
6103 
6104         netdev_notifier_info_init(&info, idev->dev);
6105         if (idev->cnf.disable_ipv6)
6106                 addrconf_notify(NULL, NETDEV_DOWN, &info);
6107         else
6108                 addrconf_notify(NULL, NETDEV_UP, &info);
6109 }
6110 
6111 static void addrconf_disable_change(struct net *net, __s32 newf)
6112 {
6113         struct net_device *dev;
6114         struct inet6_dev *idev;
6115 
6116         for_each_netdev(net, dev) {
6117                 idev = __in6_dev_get(dev);
6118                 if (idev) {
6119                         int changed = (!idev->cnf.disable_ipv6) ^ (!newf);
6120                         idev->cnf.disable_ipv6 = newf;
6121                         if (changed)
6122                                 dev_disable_change(idev);
6123                 }
6124         }
6125 }
6126 
6127 static int addrconf_disable_ipv6(struct ctl_table *table, int *p, int newf)
6128 {
6129         struct net *net;
6130         int old;
6131 
6132         if (!rtnl_trylock())
6133                 return restart_syscall();
6134 
6135         net = (struct net *)table->extra2;
6136         old = *p;
6137         *p = newf;
6138 
6139         if (p == &net->ipv6.devconf_dflt->disable_ipv6) {
6140                 rtnl_unlock();
6141                 return 0;
6142         }
6143 
6144         if (p == &net->ipv6.devconf_all->disable_ipv6) {
6145                 net->ipv6.devconf_dflt->disable_ipv6 = newf;
6146                 addrconf_disable_change(net, newf);
6147         } else if ((!newf) ^ (!old))
6148                 dev_disable_change((struct inet6_dev *)table->extra1);
6149 
6150         rtnl_unlock();
6151         return 0;
6152 }
6153 
6154 static
6155 int addrconf_sysctl_disable(struct ctl_table *ctl, int write,
6156                             void __user *buffer, size_t *lenp, loff_t *ppos)
6157 {
6158         int *valp = ctl->data;
6159         int val = *valp;
6160         loff_t pos = *ppos;
6161         struct ctl_table lctl;
6162         int ret;
6163 
6164         /*
6165          * ctl->data points to idev->cnf.disable_ipv6, we should
6166          * not modify it until we get the rtnl lock.
6167          */
6168         lctl = *ctl;
6169         lctl.data = &val;
6170 
6171         ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
6172 
6173         if (write)
6174                 ret = addrconf_disable_ipv6(ctl, valp, val);
6175         if (ret)
6176                 *ppos = pos;
6177         return ret;
6178 }
6179 
6180 static
6181 int addrconf_sysctl_proxy_ndp(struct ctl_table *ctl, int write,
6182                               void __user *buffer, size_t *lenp, loff_t *ppos)
6183 {
6184         int *valp = ctl->data;
6185         int ret;
6186         int old, new;
6187 
6188         old = *valp;
6189         ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
6190         new = *valp;
6191 
6192         if (write && old != new) {
6193                 struct net *net = ctl->extra2;
6194 
6195                 if (!rtnl_trylock())
6196                         return restart_syscall();
6197 
6198                 if (valp == &net->ipv6.devconf_dflt->proxy_ndp)
6199                         inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
6200                                                      NETCONFA_PROXY_NEIGH,
6201                                                      NETCONFA_IFINDEX_DEFAULT,
6202                                                      net->ipv6.devconf_dflt);
6203                 else if (valp == &net->ipv6.devconf_all->proxy_ndp)
6204                         inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
6205                                                      NETCONFA_PROXY_NEIGH,
6206                                                      NETCONFA_IFINDEX_ALL,
6207                                                      net->ipv6.devconf_all);
6208                 else {
6209                         struct inet6_dev *idev = ctl->extra1;
6210 
6211                         inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
6212                                                      NETCONFA_PROXY_NEIGH,
6213                                                      idev->dev->ifindex,
6214                                                      &idev->cnf);
6215                 }
6216                 rtnl_unlock();
6217         }
6218 
6219         return ret;
6220 }
6221 
6222 static int addrconf_sysctl_addr_gen_mode(struct ctl_table *ctl, int write,
6223                                          void __user *buffer, size_t *lenp,
6224                                          loff_t *ppos)
6225 {
6226         int ret = 0;
6227         u32 new_val;
6228         struct inet6_dev *idev = (struct inet6_dev *)ctl->extra1;
6229         struct net *net = (struct net *)ctl->extra2;
6230         struct ctl_table tmp = {
6231                 .data = &new_val,
6232                 .maxlen = sizeof(new_val),
6233                 .mode = ctl->mode,
6234         };
6235 
6236         if (!rtnl_trylock())
6237                 return restart_syscall();
6238 
6239         new_val = *((u32 *)ctl->data);
6240 
6241         ret = proc_douintvec(&tmp, write, buffer, lenp, ppos);
6242         if (ret != 0)
6243                 goto out;
6244 
6245         if (write) {
6246                 if (check_addr_gen_mode(new_val) < 0) {
6247                         ret = -EINVAL;
6248                         goto out;
6249                 }
6250 
6251                 if (idev) {
6252                         if (check_stable_privacy(idev, net, new_val) < 0) {
6253                                 ret = -EINVAL;
6254                                 goto out;
6255                         }
6256 
6257                         if (idev->cnf.addr_gen_mode != new_val) {
6258                                 idev->cnf.addr_gen_mode = new_val;
6259                                 addrconf_dev_config(idev->dev);
6260                         }
6261                 } else if (&net->ipv6.devconf_all->addr_gen_mode == ctl->data) {
6262                         struct net_device *dev;
6263 
6264                         net->ipv6.devconf_dflt->addr_gen_mode = new_val;
6265                         for_each_netdev(net, dev) {
6266                                 idev = __in6_dev_get(dev);
6267                                 if (idev &&
6268                                     idev->cnf.addr_gen_mode != new_val) {
6269                                         idev->cnf.addr_gen_mode = new_val;
6270                                         addrconf_dev_config(idev->dev);
6271                                 }
6272                         }
6273                 }
6274 
6275                 *((u32 *)ctl->data) = new_val;
6276         }
6277 
6278 out:
6279         rtnl_unlock();
6280 
6281         return ret;
6282 }
6283 
6284 static int addrconf_sysctl_stable_secret(struct ctl_table *ctl, int write,
6285                                          void __user *buffer, size_t *lenp,
6286                                          loff_t *ppos)
6287 {
6288         int err;
6289         struct in6_addr addr;
6290         char str[IPV6_MAX_STRLEN];
6291         struct ctl_table lctl = *ctl;
6292         struct net *net = ctl->extra2;
6293         struct ipv6_stable_secret *secret = ctl->data;
6294 
6295         if (&net->ipv6.devconf_all->stable_secret == ctl->data)
6296                 return -EIO;
6297 
6298         lctl.maxlen = IPV6_MAX_STRLEN;
6299         lctl.data = str;
6300 
6301         if (!rtnl_trylock())
6302                 return restart_syscall();
6303 
6304         if (!write && !secret->initialized) {
6305                 err = -EIO;
6306                 goto out;
6307         }
6308 
6309         err = snprintf(str, sizeof(str), "%pI6", &secret->secret);
6310         if (err >= sizeof(str)) {
6311                 err = -EIO;
6312                 goto out;
6313         }
6314 
6315         err = proc_dostring(&lctl, write, buffer, lenp, ppos);
6316         if (err || !write)
6317                 goto out;
6318 
6319         if (in6_pton(str, -1, addr.in6_u.u6_addr8, -1, NULL) != 1) {
6320                 err = -EIO;
6321                 goto out;
6322         }
6323 
6324         secret->initialized = true;
6325         secret->secret = addr;
6326 
6327         if (&net->ipv6.devconf_dflt->stable_secret == ctl->data) {
6328                 struct net_device *dev;
6329 
6330                 for_each_netdev(net, dev) {
6331                         struct inet6_dev *idev = __in6_dev_get(dev);
6332 
6333                         if (idev) {
6334                                 idev->cnf.addr_gen_mode =
6335                                         IN6_ADDR_GEN_MODE_STABLE_PRIVACY;
6336                         }
6337                 }
6338         } else {
6339                 struct inet6_dev *idev = ctl->extra1;
6340 
6341                 idev->cnf.addr_gen_mode = IN6_ADDR_GEN_MODE_STABLE_PRIVACY;
6342         }
6343 
6344 out:
6345         rtnl_unlock();
6346 
6347         return err;
6348 }
6349 
6350 static
6351 int addrconf_sysctl_ignore_routes_with_linkdown(struct ctl_table *ctl,
6352                                                 int write,
6353                                                 void __user *buffer,
6354                                                 size_t *lenp,
6355                                                 loff_t *ppos)
6356 {
6357         int *valp = ctl->data;
6358         int val = *valp;
6359         loff_t pos = *ppos;
6360         struct ctl_table lctl;
6361         int ret;
6362 
6363         /* ctl->data points to idev->cnf.ignore_routes_when_linkdown
6364          * we should not modify it until we get the rtnl lock.
6365          */
6366         lctl = *ctl;
6367         lctl.data = &val;
6368 
6369         ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
6370 
6371         if (write)
6372                 ret = addrconf_fixup_linkdown(ctl, valp, val);
6373         if (ret)
6374                 *ppos = pos;
6375         return ret;
6376 }
6377 
6378 static
6379 void addrconf_set_nopolicy(struct rt6_info *rt, int action)
6380 {
6381         if (rt) {
6382                 if (action)
6383                         rt->dst.flags |= DST_NOPOLICY;
6384                 else
6385                         rt->dst.flags &= ~DST_NOPOLICY;
6386         }
6387 }
6388 
6389 static
6390 void addrconf_disable_policy_idev(struct inet6_dev *idev, int val)
6391 {
6392         struct inet6_ifaddr *ifa;
6393 
6394         read_lock_bh(&idev->lock);
6395         list_for_each_entry(ifa, &idev->addr_list, if_list) {
6396                 spin_lock(&ifa->lock);
6397                 if (ifa->rt) {
6398                         /* host routes only use builtin fib6_nh */
6399                         struct fib6_nh *nh = ifa->rt->fib6_nh;
6400                         int cpu;
6401 
6402                         rcu_read_lock();
6403                         ifa->rt->dst_nopolicy = val ? true : false;
6404                         if (nh->rt6i_pcpu) {
6405                                 for_each_possible_cpu(cpu) {
6406                                         struct rt6_info **rtp;
6407 
6408                                         rtp = per_cpu_ptr(nh->rt6i_pcpu, cpu);
6409                                         addrconf_set_nopolicy(*rtp, val);
6410                                 }
6411                         }
6412                         rcu_read_unlock();
6413                 }
6414                 spin_unlock(&ifa->lock);
6415         }
6416         read_unlock_bh(&idev->lock);
6417 }
6418 
6419 static
6420 int addrconf_disable_policy(struct ctl_table *ctl, int *valp, int val)
6421 {
6422         struct inet6_dev *idev;
6423         struct net *net;
6424 
6425         if (!rtnl_trylock())
6426                 return restart_syscall();
6427 
6428         *valp = val;
6429 
6430         net = (struct net *)ctl->extra2;
6431         if (valp == &net->ipv6.devconf_dflt->disable_policy) {
6432                 rtnl_unlock();
6433                 return 0;
6434         }
6435 
6436         if (valp == &net->ipv6.devconf_all->disable_policy)  {
6437                 struct net_device *dev;
6438 
6439                 for_each_netdev(net, dev) {
6440                         idev = __in6_dev_get(dev);
6441                         if (idev)
6442                                 addrconf_disable_policy_idev(idev, val);
6443                 }
6444         } else {
6445                 idev = (struct inet6_dev *)ctl->extra1;
6446                 addrconf_disable_policy_idev(idev, val);
6447         }
6448 
6449         rtnl_unlock();
6450         return 0;
6451 }
6452 
6453 static
6454 int addrconf_sysctl_disable_policy(struct ctl_table *ctl, int write,
6455                                    void __user *buffer, size_t *lenp,
6456                                    loff_t *ppos)
6457 {
6458         int *valp = ctl->data;
6459         int val = *valp;
6460         loff_t pos = *ppos;
6461         struct ctl_table lctl;
6462         int ret;
6463 
6464         lctl = *ctl;
6465         lctl.data = &val;
6466         ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
6467 
6468         if (write && (*valp != val))
6469                 ret = addrconf_disable_policy(ctl, valp, val);
6470 
6471         if (ret)
6472                 *ppos = pos;
6473 
6474         return ret;
6475 }
6476 
6477 static int minus_one = -1;
6478 static const int two_five_five = 255;
6479 
6480 static const struct ctl_table addrconf_sysctl[] = {
6481         {
6482                 .procname       = "forwarding",
6483                 .data           = &ipv6_devconf.forwarding,
6484                 .maxlen         = sizeof(int),
6485                 .mode           = 0644,
6486                 .proc_handler   = addrconf_sysctl_forward,
6487         },
6488         {
6489                 .procname       = "hop_limit",
6490                 .data           = &ipv6_devconf.hop_limit,
6491                 .maxlen         = sizeof(int),
6492                 .mode           = 0644,
6493                 .proc_handler   = proc_dointvec_minmax,
6494                 .extra1         = (void *)SYSCTL_ONE,
6495                 .extra2         = (void *)&two_five_five,
6496         },
6497         {
6498                 .procname       = "mtu",
6499                 .data           = &ipv6_devconf.mtu6,
6500                 .maxlen         = sizeof(int),
6501                 .mode           = 0644,
6502                 .proc_handler   = addrconf_sysctl_mtu,
6503         },
6504         {
6505                 .procname       = "accept_ra",
6506                 .data           = &ipv6_devconf.accept_ra,
6507                 .maxlen         = sizeof(int),
6508                 .mode           = 0644,
6509                 .proc_handler   = proc_dointvec,
6510         },
6511         {
6512                 .procname       = "accept_redirects",
6513                 .data           = &ipv6_devconf.accept_redirects,
6514                 .maxlen         = sizeof(int),
6515                 .mode           = 0644,
6516                 .proc_handler   = proc_dointvec,
6517         },
6518         {
6519                 .procname       = "autoconf",
6520                 .data           = &ipv6_devconf.autoconf,
6521                 .maxlen         = sizeof(int),
6522                 .mode           = 0644,
6523                 .proc_handler   = proc_dointvec,
6524         },
6525         {
6526                 .procname       = "dad_transmits",
6527                 .data           = &ipv6_devconf.dad_transmits,
6528                 .maxlen         = sizeof(int),
6529                 .mode           = 0644,
6530                 .proc_handler   = proc_dointvec,
6531         },
6532         {
6533                 .procname       = "router_solicitations",
6534                 .data           = &ipv6_devconf.rtr_solicits,
6535                 .maxlen         = sizeof(int),
6536                 .mode           = 0644,
6537                 .proc_handler   = proc_dointvec_minmax,
6538                 .extra1         = &minus_one,
6539         },
6540         {
6541                 .procname       = "router_solicitation_interval",
6542                 .data           = &ipv6_devconf.rtr_solicit_interval,
6543                 .maxlen         = sizeof(int),
6544                 .mode           = 0644,
6545                 .proc_handler   = proc_dointvec_jiffies,