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

TOMOYO Linux Cross Reference
Linux/net/netlink/af_netlink.c

Version: ~ [ linux-5.19-rc3 ] ~ [ linux-5.18.5 ] ~ [ linux-5.17.15 ] ~ [ linux-5.16.20 ] ~ [ linux-5.15.48 ] ~ [ linux-5.14.21 ] ~ [ linux-5.13.19 ] ~ [ linux-5.12.19 ] ~ [ linux-5.11.22 ] ~ [ linux-5.10.123 ] ~ [ linux-5.9.16 ] ~ [ linux-5.8.18 ] ~ [ linux-5.7.19 ] ~ [ linux-5.6.19 ] ~ [ linux-5.5.19 ] ~ [ linux-5.4.199 ] ~ [ linux-5.3.18 ] ~ [ linux-5.2.21 ] ~ [ linux-5.1.21 ] ~ [ linux-5.0.21 ] ~ [ linux-4.20.17 ] ~ [ linux-4.19.248 ] ~ [ linux-4.18.20 ] ~ [ linux-4.17.19 ] ~ [ linux-4.16.18 ] ~ [ linux-4.15.18 ] ~ [ linux-4.14.284 ] ~ [ linux-4.13.16 ] ~ [ linux-4.12.14 ] ~ [ linux-4.11.12 ] ~ [ linux-4.10.17 ] ~ [ linux-4.9.319 ] ~ [ linux-4.8.17 ] ~ [ linux-4.7.10 ] ~ [ linux-4.6.7 ] ~ [ linux-4.5.7 ] ~ [ linux-4.4.302 ] ~ [ linux-4.3.6 ] ~ [ linux-4.2.8 ] ~ [ linux-4.1.52 ] ~ [ linux-4.0.9 ] ~ [ linux-3.10.108 ] ~ [ linux-2.6.32.71 ] ~ [ linux-2.6.0 ] ~ [ linux-2.4.37.11 ] ~ [ unix-v6-master ] ~ [ ccs-tools-1.8.9 ] ~ [ policy-sample ] ~
Architecture: ~ [ i386 ] ~ [ alpha ] ~ [ m68k ] ~ [ mips ] ~ [ ppc ] ~ [ sparc ] ~ [ sparc64 ] ~

  1 // SPDX-License-Identifier: GPL-2.0-or-later
  2 /*
  3  * NETLINK      Kernel-user communication protocol.
  4  *
  5  *              Authors:        Alan Cox <alan@lxorguk.ukuu.org.uk>
  6  *                              Alexey Kuznetsov <kuznet@ms2.inr.ac.ru>
  7  *                              Patrick McHardy <kaber@trash.net>
  8  *
  9  * Tue Jun 26 14:36:48 MEST 2001 Herbert "herp" Rosmanith
 10  *                               added netlink_proto_exit
 11  * Tue Jan 22 18:32:44 BRST 2002 Arnaldo C. de Melo <acme@conectiva.com.br>
 12  *                               use nlk_sk, as sk->protinfo is on a diet 8)
 13  * Fri Jul 22 19:51:12 MEST 2005 Harald Welte <laforge@gnumonks.org>
 14  *                               - inc module use count of module that owns
 15  *                                 the kernel socket in case userspace opens
 16  *                                 socket of same protocol
 17  *                               - remove all module support, since netlink is
 18  *                                 mandatory if CONFIG_NET=y these days
 19  */
 20 
 21 #include <linux/module.h>
 22 
 23 #include <linux/capability.h>
 24 #include <linux/kernel.h>
 25 #include <linux/init.h>
 26 #include <linux/signal.h>
 27 #include <linux/sched.h>
 28 #include <linux/errno.h>
 29 #include <linux/string.h>
 30 #include <linux/stat.h>
 31 #include <linux/socket.h>
 32 #include <linux/un.h>
 33 #include <linux/fcntl.h>
 34 #include <linux/termios.h>
 35 #include <linux/sockios.h>
 36 #include <linux/net.h>
 37 #include <linux/fs.h>
 38 #include <linux/slab.h>
 39 #include <linux/uaccess.h>
 40 #include <linux/skbuff.h>
 41 #include <linux/netdevice.h>
 42 #include <linux/rtnetlink.h>
 43 #include <linux/proc_fs.h>
 44 #include <linux/seq_file.h>
 45 #include <linux/notifier.h>
 46 #include <linux/security.h>
 47 #include <linux/jhash.h>
 48 #include <linux/jiffies.h>
 49 #include <linux/random.h>
 50 #include <linux/bitops.h>
 51 #include <linux/mm.h>
 52 #include <linux/types.h>
 53 #include <linux/audit.h>
 54 #include <linux/mutex.h>
 55 #include <linux/vmalloc.h>
 56 #include <linux/if_arp.h>
 57 #include <linux/rhashtable.h>
 58 #include <asm/cacheflush.h>
 59 #include <linux/hash.h>
 60 #include <linux/genetlink.h>
 61 #include <linux/net_namespace.h>
 62 #include <linux/nospec.h>
 63 #include <linux/btf_ids.h>
 64 
 65 #include <net/net_namespace.h>
 66 #include <net/netns/generic.h>
 67 #include <net/sock.h>
 68 #include <net/scm.h>
 69 #include <net/netlink.h>
 70 #define CREATE_TRACE_POINTS
 71 #include <trace/events/netlink.h>
 72 
 73 #include "af_netlink.h"
 74 
 75 struct listeners {
 76         struct rcu_head         rcu;
 77         unsigned long           masks[];
 78 };
 79 
 80 /* state bits */
 81 #define NETLINK_S_CONGESTED             0x0
 82 
 83 static inline int netlink_is_kernel(struct sock *sk)
 84 {
 85         return nlk_sk(sk)->flags & NETLINK_F_KERNEL_SOCKET;
 86 }
 87 
 88 struct netlink_table *nl_table __read_mostly;
 89 EXPORT_SYMBOL_GPL(nl_table);
 90 
 91 static DECLARE_WAIT_QUEUE_HEAD(nl_table_wait);
 92 
 93 static struct lock_class_key nlk_cb_mutex_keys[MAX_LINKS];
 94 
 95 static const char *const nlk_cb_mutex_key_strings[MAX_LINKS + 1] = {
 96         "nlk_cb_mutex-ROUTE",
 97         "nlk_cb_mutex-1",
 98         "nlk_cb_mutex-USERSOCK",
 99         "nlk_cb_mutex-FIREWALL",
100         "nlk_cb_mutex-SOCK_DIAG",
101         "nlk_cb_mutex-NFLOG",
102         "nlk_cb_mutex-XFRM",
103         "nlk_cb_mutex-SELINUX",
104         "nlk_cb_mutex-ISCSI",
105         "nlk_cb_mutex-AUDIT",
106         "nlk_cb_mutex-FIB_LOOKUP",
107         "nlk_cb_mutex-CONNECTOR",
108         "nlk_cb_mutex-NETFILTER",
109         "nlk_cb_mutex-IP6_FW",
110         "nlk_cb_mutex-DNRTMSG",
111         "nlk_cb_mutex-KOBJECT_UEVENT",
112         "nlk_cb_mutex-GENERIC",
113         "nlk_cb_mutex-17",
114         "nlk_cb_mutex-SCSITRANSPORT",
115         "nlk_cb_mutex-ECRYPTFS",
116         "nlk_cb_mutex-RDMA",
117         "nlk_cb_mutex-CRYPTO",
118         "nlk_cb_mutex-SMC",
119         "nlk_cb_mutex-23",
120         "nlk_cb_mutex-24",
121         "nlk_cb_mutex-25",
122         "nlk_cb_mutex-26",
123         "nlk_cb_mutex-27",
124         "nlk_cb_mutex-28",
125         "nlk_cb_mutex-29",
126         "nlk_cb_mutex-30",
127         "nlk_cb_mutex-31",
128         "nlk_cb_mutex-MAX_LINKS"
129 };
130 
131 static int netlink_dump(struct sock *sk);
132 
133 /* nl_table locking explained:
134  * Lookup and traversal are protected with an RCU read-side lock. Insertion
135  * and removal are protected with per bucket lock while using RCU list
136  * modification primitives and may run in parallel to RCU protected lookups.
137  * Destruction of the Netlink socket may only occur *after* nl_table_lock has
138  * been acquired * either during or after the socket has been removed from
139  * the list and after an RCU grace period.
140  */
141 DEFINE_RWLOCK(nl_table_lock);
142 EXPORT_SYMBOL_GPL(nl_table_lock);
143 static atomic_t nl_table_users = ATOMIC_INIT(0);
144 
145 #define nl_deref_protected(X) rcu_dereference_protected(X, lockdep_is_held(&nl_table_lock));
146 
147 static BLOCKING_NOTIFIER_HEAD(netlink_chain);
148 
149 
150 static const struct rhashtable_params netlink_rhashtable_params;
151 
152 void do_trace_netlink_extack(const char *msg)
153 {
154         trace_netlink_extack(msg);
155 }
156 EXPORT_SYMBOL(do_trace_netlink_extack);
157 
158 static inline u32 netlink_group_mask(u32 group)
159 {
160         return group ? 1 << (group - 1) : 0;
161 }
162 
163 static struct sk_buff *netlink_to_full_skb(const struct sk_buff *skb,
164                                            gfp_t gfp_mask)
165 {
166         unsigned int len = skb_end_offset(skb);
167         struct sk_buff *new;
168 
169         new = alloc_skb(len, gfp_mask);
170         if (new == NULL)
171                 return NULL;
172 
173         NETLINK_CB(new).portid = NETLINK_CB(skb).portid;
174         NETLINK_CB(new).dst_group = NETLINK_CB(skb).dst_group;
175         NETLINK_CB(new).creds = NETLINK_CB(skb).creds;
176 
177         skb_put_data(new, skb->data, len);
178         return new;
179 }
180 
181 static unsigned int netlink_tap_net_id;
182 
183 struct netlink_tap_net {
184         struct list_head netlink_tap_all;
185         struct mutex netlink_tap_lock;
186 };
187 
188 int netlink_add_tap(struct netlink_tap *nt)
189 {
190         struct net *net = dev_net(nt->dev);
191         struct netlink_tap_net *nn = net_generic(net, netlink_tap_net_id);
192 
193         if (unlikely(nt->dev->type != ARPHRD_NETLINK))
194                 return -EINVAL;
195 
196         mutex_lock(&nn->netlink_tap_lock);
197         list_add_rcu(&nt->list, &nn->netlink_tap_all);
198         mutex_unlock(&nn->netlink_tap_lock);
199 
200         __module_get(nt->module);
201 
202         return 0;
203 }
204 EXPORT_SYMBOL_GPL(netlink_add_tap);
205 
206 static int __netlink_remove_tap(struct netlink_tap *nt)
207 {
208         struct net *net = dev_net(nt->dev);
209         struct netlink_tap_net *nn = net_generic(net, netlink_tap_net_id);
210         bool found = false;
211         struct netlink_tap *tmp;
212 
213         mutex_lock(&nn->netlink_tap_lock);
214 
215         list_for_each_entry(tmp, &nn->netlink_tap_all, list) {
216                 if (nt == tmp) {
217                         list_del_rcu(&nt->list);
218                         found = true;
219                         goto out;
220                 }
221         }
222 
223         pr_warn("__netlink_remove_tap: %p not found\n", nt);
224 out:
225         mutex_unlock(&nn->netlink_tap_lock);
226 
227         if (found)
228                 module_put(nt->module);
229 
230         return found ? 0 : -ENODEV;
231 }
232 
233 int netlink_remove_tap(struct netlink_tap *nt)
234 {
235         int ret;
236 
237         ret = __netlink_remove_tap(nt);
238         synchronize_net();
239 
240         return ret;
241 }
242 EXPORT_SYMBOL_GPL(netlink_remove_tap);
243 
244 static __net_init int netlink_tap_init_net(struct net *net)
245 {
246         struct netlink_tap_net *nn = net_generic(net, netlink_tap_net_id);
247 
248         INIT_LIST_HEAD(&nn->netlink_tap_all);
249         mutex_init(&nn->netlink_tap_lock);
250         return 0;
251 }
252 
253 static struct pernet_operations netlink_tap_net_ops = {
254         .init = netlink_tap_init_net,
255         .id   = &netlink_tap_net_id,
256         .size = sizeof(struct netlink_tap_net),
257 };
258 
259 static bool netlink_filter_tap(const struct sk_buff *skb)
260 {
261         struct sock *sk = skb->sk;
262 
263         /* We take the more conservative approach and
264          * whitelist socket protocols that may pass.
265          */
266         switch (sk->sk_protocol) {
267         case NETLINK_ROUTE:
268         case NETLINK_USERSOCK:
269         case NETLINK_SOCK_DIAG:
270         case NETLINK_NFLOG:
271         case NETLINK_XFRM:
272         case NETLINK_FIB_LOOKUP:
273         case NETLINK_NETFILTER:
274         case NETLINK_GENERIC:
275                 return true;
276         }
277 
278         return false;
279 }
280 
281 static int __netlink_deliver_tap_skb(struct sk_buff *skb,
282                                      struct net_device *dev)
283 {
284         struct sk_buff *nskb;
285         struct sock *sk = skb->sk;
286         int ret = -ENOMEM;
287 
288         if (!net_eq(dev_net(dev), sock_net(sk)))
289                 return 0;
290 
291         dev_hold(dev);
292 
293         if (is_vmalloc_addr(skb->head))
294                 nskb = netlink_to_full_skb(skb, GFP_ATOMIC);
295         else
296                 nskb = skb_clone(skb, GFP_ATOMIC);
297         if (nskb) {
298                 nskb->dev = dev;
299                 nskb->protocol = htons((u16) sk->sk_protocol);
300                 nskb->pkt_type = netlink_is_kernel(sk) ?
301                                  PACKET_KERNEL : PACKET_USER;
302                 skb_reset_network_header(nskb);
303                 ret = dev_queue_xmit(nskb);
304                 if (unlikely(ret > 0))
305                         ret = net_xmit_errno(ret);
306         }
307 
308         dev_put(dev);
309         return ret;
310 }
311 
312 static void __netlink_deliver_tap(struct sk_buff *skb, struct netlink_tap_net *nn)
313 {
314         int ret;
315         struct netlink_tap *tmp;
316 
317         if (!netlink_filter_tap(skb))
318                 return;
319 
320         list_for_each_entry_rcu(tmp, &nn->netlink_tap_all, list) {
321                 ret = __netlink_deliver_tap_skb(skb, tmp->dev);
322                 if (unlikely(ret))
323                         break;
324         }
325 }
326 
327 static void netlink_deliver_tap(struct net *net, struct sk_buff *skb)
328 {
329         struct netlink_tap_net *nn = net_generic(net, netlink_tap_net_id);
330 
331         rcu_read_lock();
332 
333         if (unlikely(!list_empty(&nn->netlink_tap_all)))
334                 __netlink_deliver_tap(skb, nn);
335 
336         rcu_read_unlock();
337 }
338 
339 static void netlink_deliver_tap_kernel(struct sock *dst, struct sock *src,
340                                        struct sk_buff *skb)
341 {
342         if (!(netlink_is_kernel(dst) && netlink_is_kernel(src)))
343                 netlink_deliver_tap(sock_net(dst), skb);
344 }
345 
346 static void netlink_overrun(struct sock *sk)
347 {
348         struct netlink_sock *nlk = nlk_sk(sk);
349 
350         if (!(nlk->flags & NETLINK_F_RECV_NO_ENOBUFS)) {
351                 if (!test_and_set_bit(NETLINK_S_CONGESTED,
352                                       &nlk_sk(sk)->state)) {
353                         sk->sk_err = ENOBUFS;
354                         sk->sk_error_report(sk);
355                 }
356         }
357         atomic_inc(&sk->sk_drops);
358 }
359 
360 static void netlink_rcv_wake(struct sock *sk)
361 {
362         struct netlink_sock *nlk = nlk_sk(sk);
363 
364         if (skb_queue_empty_lockless(&sk->sk_receive_queue))
365                 clear_bit(NETLINK_S_CONGESTED, &nlk->state);
366         if (!test_bit(NETLINK_S_CONGESTED, &nlk->state))
367                 wake_up_interruptible(&nlk->wait);
368 }
369 
370 static void netlink_skb_destructor(struct sk_buff *skb)
371 {
372         if (is_vmalloc_addr(skb->head)) {
373                 if (!skb->cloned ||
374                     !atomic_dec_return(&(skb_shinfo(skb)->dataref)))
375                         vfree(skb->head);
376 
377                 skb->head = NULL;
378         }
379         if (skb->sk != NULL)
380                 sock_rfree(skb);
381 }
382 
383 static void netlink_skb_set_owner_r(struct sk_buff *skb, struct sock *sk)
384 {
385         WARN_ON(skb->sk != NULL);
386         skb->sk = sk;
387         skb->destructor = netlink_skb_destructor;
388         atomic_add(skb->truesize, &sk->sk_rmem_alloc);
389         sk_mem_charge(sk, skb->truesize);
390 }
391 
392 static void netlink_sock_destruct(struct sock *sk)
393 {
394         struct netlink_sock *nlk = nlk_sk(sk);
395 
396         if (nlk->cb_running) {
397                 if (nlk->cb.done)
398                         nlk->cb.done(&nlk->cb);
399                 module_put(nlk->cb.module);
400                 kfree_skb(nlk->cb.skb);
401         }
402 
403         skb_queue_purge(&sk->sk_receive_queue);
404 
405         if (!sock_flag(sk, SOCK_DEAD)) {
406                 printk(KERN_ERR "Freeing alive netlink socket %p\n", sk);
407                 return;
408         }
409 
410         WARN_ON(atomic_read(&sk->sk_rmem_alloc));
411         WARN_ON(refcount_read(&sk->sk_wmem_alloc));
412         WARN_ON(nlk_sk(sk)->groups);
413 }
414 
415 static void netlink_sock_destruct_work(struct work_struct *work)
416 {
417         struct netlink_sock *nlk = container_of(work, struct netlink_sock,
418                                                 work);
419 
420         sk_free(&nlk->sk);
421 }
422 
423 /* This lock without WQ_FLAG_EXCLUSIVE is good on UP and it is _very_ bad on
424  * SMP. Look, when several writers sleep and reader wakes them up, all but one
425  * immediately hit write lock and grab all the cpus. Exclusive sleep solves
426  * this, _but_ remember, it adds useless work on UP machines.
427  */
428 
429 void netlink_table_grab(void)
430         __acquires(nl_table_lock)
431 {
432         might_sleep();
433 
434         write_lock_irq(&nl_table_lock);
435 
436         if (atomic_read(&nl_table_users)) {
437                 DECLARE_WAITQUEUE(wait, current);
438 
439                 add_wait_queue_exclusive(&nl_table_wait, &wait);
440                 for (;;) {
441                         set_current_state(TASK_UNINTERRUPTIBLE);
442                         if (atomic_read(&nl_table_users) == 0)
443                                 break;
444                         write_unlock_irq(&nl_table_lock);
445                         schedule();
446                         write_lock_irq(&nl_table_lock);
447                 }
448 
449                 __set_current_state(TASK_RUNNING);
450                 remove_wait_queue(&nl_table_wait, &wait);
451         }
452 }
453 
454 void netlink_table_ungrab(void)
455         __releases(nl_table_lock)
456 {
457         write_unlock_irq(&nl_table_lock);
458         wake_up(&nl_table_wait);
459 }
460 
461 static inline void
462 netlink_lock_table(void)
463 {
464         unsigned long flags;
465 
466         /* read_lock() synchronizes us to netlink_table_grab */
467 
468         read_lock_irqsave(&nl_table_lock, flags);
469         atomic_inc(&nl_table_users);
470         read_unlock_irqrestore(&nl_table_lock, flags);
471 }
472 
473 static inline void
474 netlink_unlock_table(void)
475 {
476         if (atomic_dec_and_test(&nl_table_users))
477                 wake_up(&nl_table_wait);
478 }
479 
480 struct netlink_compare_arg
481 {
482         possible_net_t pnet;
483         u32 portid;
484 };
485 
486 /* Doing sizeof directly may yield 4 extra bytes on 64-bit. */
487 #define netlink_compare_arg_len \
488         (offsetof(struct netlink_compare_arg, portid) + sizeof(u32))
489 
490 static inline int netlink_compare(struct rhashtable_compare_arg *arg,
491                                   const void *ptr)
492 {
493         const struct netlink_compare_arg *x = arg->key;
494         const struct netlink_sock *nlk = ptr;
495 
496         return nlk->portid != x->portid ||
497                !net_eq(sock_net(&nlk->sk), read_pnet(&x->pnet));
498 }
499 
500 static void netlink_compare_arg_init(struct netlink_compare_arg *arg,
501                                      struct net *net, u32 portid)
502 {
503         memset(arg, 0, sizeof(*arg));
504         write_pnet(&arg->pnet, net);
505         arg->portid = portid;
506 }
507 
508 static struct sock *__netlink_lookup(struct netlink_table *table, u32 portid,
509                                      struct net *net)
510 {
511         struct netlink_compare_arg arg;
512 
513         netlink_compare_arg_init(&arg, net, portid);
514         return rhashtable_lookup_fast(&table->hash, &arg,
515                                       netlink_rhashtable_params);
516 }
517 
518 static int __netlink_insert(struct netlink_table *table, struct sock *sk)
519 {
520         struct netlink_compare_arg arg;
521 
522         netlink_compare_arg_init(&arg, sock_net(sk), nlk_sk(sk)->portid);
523         return rhashtable_lookup_insert_key(&table->hash, &arg,
524                                             &nlk_sk(sk)->node,
525                                             netlink_rhashtable_params);
526 }
527 
528 static struct sock *netlink_lookup(struct net *net, int protocol, u32 portid)
529 {
530         struct netlink_table *table = &nl_table[protocol];
531         struct sock *sk;
532 
533         rcu_read_lock();
534         sk = __netlink_lookup(table, portid, net);
535         if (sk)
536                 sock_hold(sk);
537         rcu_read_unlock();
538 
539         return sk;
540 }
541 
542 static const struct proto_ops netlink_ops;
543 
544 static void
545 netlink_update_listeners(struct sock *sk)
546 {
547         struct netlink_table *tbl = &nl_table[sk->sk_protocol];
548         unsigned long mask;
549         unsigned int i;
550         struct listeners *listeners;
551 
552         listeners = nl_deref_protected(tbl->listeners);
553         if (!listeners)
554                 return;
555 
556         for (i = 0; i < NLGRPLONGS(tbl->groups); i++) {
557                 mask = 0;
558                 sk_for_each_bound(sk, &tbl->mc_list) {
559                         if (i < NLGRPLONGS(nlk_sk(sk)->ngroups))
560                                 mask |= nlk_sk(sk)->groups[i];
561                 }
562                 listeners->masks[i] = mask;
563         }
564         /* this function is only called with the netlink table "grabbed", which
565          * makes sure updates are visible before bind or setsockopt return. */
566 }
567 
568 static int netlink_insert(struct sock *sk, u32 portid)
569 {
570         struct netlink_table *table = &nl_table[sk->sk_protocol];
571         int err;
572 
573         lock_sock(sk);
574 
575         err = nlk_sk(sk)->portid == portid ? 0 : -EBUSY;
576         if (nlk_sk(sk)->bound)
577                 goto err;
578 
579         nlk_sk(sk)->portid = portid;
580         sock_hold(sk);
581 
582         err = __netlink_insert(table, sk);
583         if (err) {
584                 /* In case the hashtable backend returns with -EBUSY
585                  * from here, it must not escape to the caller.
586                  */
587                 if (unlikely(err == -EBUSY))
588                         err = -EOVERFLOW;
589                 if (err == -EEXIST)
590                         err = -EADDRINUSE;
591                 sock_put(sk);
592                 goto err;
593         }
594 
595         /* We need to ensure that the socket is hashed and visible. */
596         smp_wmb();
597         nlk_sk(sk)->bound = portid;
598 
599 err:
600         release_sock(sk);
601         return err;
602 }
603 
604 static void netlink_remove(struct sock *sk)
605 {
606         struct netlink_table *table;
607 
608         table = &nl_table[sk->sk_protocol];
609         if (!rhashtable_remove_fast(&table->hash, &nlk_sk(sk)->node,
610                                     netlink_rhashtable_params)) {
611                 WARN_ON(refcount_read(&sk->sk_refcnt) == 1);
612                 __sock_put(sk);
613         }
614 
615         netlink_table_grab();
616         if (nlk_sk(sk)->subscriptions) {
617                 __sk_del_bind_node(sk);
618                 netlink_update_listeners(sk);
619         }
620         if (sk->sk_protocol == NETLINK_GENERIC)
621                 atomic_inc(&genl_sk_destructing_cnt);
622         netlink_table_ungrab();
623 }
624 
625 static struct proto netlink_proto = {
626         .name     = "NETLINK",
627         .owner    = THIS_MODULE,
628         .obj_size = sizeof(struct netlink_sock),
629 };
630 
631 static int __netlink_create(struct net *net, struct socket *sock,
632                             struct mutex *cb_mutex, int protocol,
633                             int kern)
634 {
635         struct sock *sk;
636         struct netlink_sock *nlk;
637 
638         sock->ops = &netlink_ops;
639 
640         sk = sk_alloc(net, PF_NETLINK, GFP_KERNEL, &netlink_proto, kern);
641         if (!sk)
642                 return -ENOMEM;
643 
644         sock_init_data(sock, sk);
645 
646         nlk = nlk_sk(sk);
647         if (cb_mutex) {
648                 nlk->cb_mutex = cb_mutex;
649         } else {
650                 nlk->cb_mutex = &nlk->cb_def_mutex;
651                 mutex_init(nlk->cb_mutex);
652                 lockdep_set_class_and_name(nlk->cb_mutex,
653                                            nlk_cb_mutex_keys + protocol,
654                                            nlk_cb_mutex_key_strings[protocol]);
655         }
656         init_waitqueue_head(&nlk->wait);
657 
658         sk->sk_destruct = netlink_sock_destruct;
659         sk->sk_protocol = protocol;
660         return 0;
661 }
662 
663 static int netlink_create(struct net *net, struct socket *sock, int protocol,
664                           int kern)
665 {
666         struct module *module = NULL;
667         struct mutex *cb_mutex;
668         struct netlink_sock *nlk;
669         int (*bind)(struct net *net, int group);
670         void (*unbind)(struct net *net, int group);
671         int err = 0;
672 
673         sock->state = SS_UNCONNECTED;
674 
675         if (sock->type != SOCK_RAW && sock->type != SOCK_DGRAM)
676                 return -ESOCKTNOSUPPORT;
677 
678         if (protocol < 0 || protocol >= MAX_LINKS)
679                 return -EPROTONOSUPPORT;
680         protocol = array_index_nospec(protocol, MAX_LINKS);
681 
682         netlink_lock_table();
683 #ifdef CONFIG_MODULES
684         if (!nl_table[protocol].registered) {
685                 netlink_unlock_table();
686                 request_module("net-pf-%d-proto-%d", PF_NETLINK, protocol);
687                 netlink_lock_table();
688         }
689 #endif
690         if (nl_table[protocol].registered &&
691             try_module_get(nl_table[protocol].module))
692                 module = nl_table[protocol].module;
693         else
694                 err = -EPROTONOSUPPORT;
695         cb_mutex = nl_table[protocol].cb_mutex;
696         bind = nl_table[protocol].bind;
697         unbind = nl_table[protocol].unbind;
698         netlink_unlock_table();
699 
700         if (err < 0)
701                 goto out;
702 
703         err = __netlink_create(net, sock, cb_mutex, protocol, kern);
704         if (err < 0)
705                 goto out_module;
706 
707         local_bh_disable();
708         sock_prot_inuse_add(net, &netlink_proto, 1);
709         local_bh_enable();
710 
711         nlk = nlk_sk(sock->sk);
712         nlk->module = module;
713         nlk->netlink_bind = bind;
714         nlk->netlink_unbind = unbind;
715 out:
716         return err;
717 
718 out_module:
719         module_put(module);
720         goto out;
721 }
722 
723 static void deferred_put_nlk_sk(struct rcu_head *head)
724 {
725         struct netlink_sock *nlk = container_of(head, struct netlink_sock, rcu);
726         struct sock *sk = &nlk->sk;
727 
728         kfree(nlk->groups);
729         nlk->groups = NULL;
730 
731         if (!refcount_dec_and_test(&sk->sk_refcnt))
732                 return;
733 
734         if (nlk->cb_running && nlk->cb.done) {
735                 INIT_WORK(&nlk->work, netlink_sock_destruct_work);
736                 schedule_work(&nlk->work);
737                 return;
738         }
739 
740         sk_free(sk);
741 }
742 
743 static int netlink_release(struct socket *sock)
744 {
745         struct sock *sk = sock->sk;
746         struct netlink_sock *nlk;
747 
748         if (!sk)
749                 return 0;
750 
751         netlink_remove(sk);
752         sock_orphan(sk);
753         nlk = nlk_sk(sk);
754 
755         /*
756          * OK. Socket is unlinked, any packets that arrive now
757          * will be purged.
758          */
759 
760         /* must not acquire netlink_table_lock in any way again before unbind
761          * and notifying genetlink is done as otherwise it might deadlock
762          */
763         if (nlk->netlink_unbind) {
764                 int i;
765 
766                 for (i = 0; i < nlk->ngroups; i++)
767                         if (test_bit(i, nlk->groups))
768                                 nlk->netlink_unbind(sock_net(sk), i + 1);
769         }
770         if (sk->sk_protocol == NETLINK_GENERIC &&
771             atomic_dec_return(&genl_sk_destructing_cnt) == 0)
772                 wake_up(&genl_sk_destructing_waitq);
773 
774         sock->sk = NULL;
775         wake_up_interruptible_all(&nlk->wait);
776 
777         skb_queue_purge(&sk->sk_write_queue);
778 
779         if (nlk->portid && nlk->bound) {
780                 struct netlink_notify n = {
781                                                 .net = sock_net(sk),
782                                                 .protocol = sk->sk_protocol,
783                                                 .portid = nlk->portid,
784                                           };
785                 blocking_notifier_call_chain(&netlink_chain,
786                                 NETLINK_URELEASE, &n);
787         }
788 
789         module_put(nlk->module);
790 
791         if (netlink_is_kernel(sk)) {
792                 netlink_table_grab();
793                 BUG_ON(nl_table[sk->sk_protocol].registered == 0);
794                 if (--nl_table[sk->sk_protocol].registered == 0) {
795                         struct listeners *old;
796 
797                         old = nl_deref_protected(nl_table[sk->sk_protocol].listeners);
798                         RCU_INIT_POINTER(nl_table[sk->sk_protocol].listeners, NULL);
799                         kfree_rcu(old, rcu);
800                         nl_table[sk->sk_protocol].module = NULL;
801                         nl_table[sk->sk_protocol].bind = NULL;
802                         nl_table[sk->sk_protocol].unbind = NULL;
803                         nl_table[sk->sk_protocol].flags = 0;
804                         nl_table[sk->sk_protocol].registered = 0;
805                 }
806                 netlink_table_ungrab();
807         }
808 
809         local_bh_disable();
810         sock_prot_inuse_add(sock_net(sk), &netlink_proto, -1);
811         local_bh_enable();
812         call_rcu(&nlk->rcu, deferred_put_nlk_sk);
813         return 0;
814 }
815 
816 static int netlink_autobind(struct socket *sock)
817 {
818         struct sock *sk = sock->sk;
819         struct net *net = sock_net(sk);
820         struct netlink_table *table = &nl_table[sk->sk_protocol];
821         s32 portid = task_tgid_vnr(current);
822         int err;
823         s32 rover = -4096;
824         bool ok;
825 
826 retry:
827         cond_resched();
828         rcu_read_lock();
829         ok = !__netlink_lookup(table, portid, net);
830         rcu_read_unlock();
831         if (!ok) {
832                 /* Bind collision, search negative portid values. */
833                 if (rover == -4096)
834                         /* rover will be in range [S32_MIN, -4097] */
835                         rover = S32_MIN + prandom_u32_max(-4096 - S32_MIN);
836                 else if (rover >= -4096)
837                         rover = -4097;
838                 portid = rover--;
839                 goto retry;
840         }
841 
842         err = netlink_insert(sk, portid);
843         if (err == -EADDRINUSE)
844                 goto retry;
845 
846         /* If 2 threads race to autobind, that is fine.  */
847         if (err == -EBUSY)
848                 err = 0;
849 
850         return err;
851 }
852 
853 /**
854  * __netlink_ns_capable - General netlink message capability test
855  * @nsp: NETLINK_CB of the socket buffer holding a netlink command from userspace.
856  * @user_ns: The user namespace of the capability to use
857  * @cap: The capability to use
858  *
859  * Test to see if the opener of the socket we received the message
860  * from had when the netlink socket was created and the sender of the
861  * message has the capability @cap in the user namespace @user_ns.
862  */
863 bool __netlink_ns_capable(const struct netlink_skb_parms *nsp,
864                         struct user_namespace *user_ns, int cap)
865 {
866         return ((nsp->flags & NETLINK_SKB_DST) ||
867                 file_ns_capable(nsp->sk->sk_socket->file, user_ns, cap)) &&
868                 ns_capable(user_ns, cap);
869 }
870 EXPORT_SYMBOL(__netlink_ns_capable);
871 
872 /**
873  * netlink_ns_capable - General netlink message capability test
874  * @skb: socket buffer holding a netlink command from userspace
875  * @user_ns: The user namespace of the capability to use
876  * @cap: The capability to use
877  *
878  * Test to see if the opener of the socket we received the message
879  * from had when the netlink socket was created and the sender of the
880  * message has the capability @cap in the user namespace @user_ns.
881  */
882 bool netlink_ns_capable(const struct sk_buff *skb,
883                         struct user_namespace *user_ns, int cap)
884 {
885         return __netlink_ns_capable(&NETLINK_CB(skb), user_ns, cap);
886 }
887 EXPORT_SYMBOL(netlink_ns_capable);
888 
889 /**
890  * netlink_capable - Netlink global message capability test
891  * @skb: socket buffer holding a netlink command from userspace
892  * @cap: The capability to use
893  *
894  * Test to see if the opener of the socket we received the message
895  * from had when the netlink socket was created and the sender of the
896  * message has the capability @cap in all user namespaces.
897  */
898 bool netlink_capable(const struct sk_buff *skb, int cap)
899 {
900         return netlink_ns_capable(skb, &init_user_ns, cap);
901 }
902 EXPORT_SYMBOL(netlink_capable);
903 
904 /**
905  * netlink_net_capable - Netlink network namespace message capability test
906  * @skb: socket buffer holding a netlink command from userspace
907  * @cap: The capability to use
908  *
909  * Test to see if the opener of the socket we received the message
910  * from had when the netlink socket was created and the sender of the
911  * message has the capability @cap over the network namespace of
912  * the socket we received the message from.
913  */
914 bool netlink_net_capable(const struct sk_buff *skb, int cap)
915 {
916         return netlink_ns_capable(skb, sock_net(skb->sk)->user_ns, cap);
917 }
918 EXPORT_SYMBOL(netlink_net_capable);
919 
920 static inline int netlink_allowed(const struct socket *sock, unsigned int flag)
921 {
922         return (nl_table[sock->sk->sk_protocol].flags & flag) ||
923                 ns_capable(sock_net(sock->sk)->user_ns, CAP_NET_ADMIN);
924 }
925 
926 static void
927 netlink_update_subscriptions(struct sock *sk, unsigned int subscriptions)
928 {
929         struct netlink_sock *nlk = nlk_sk(sk);
930 
931         if (nlk->subscriptions && !subscriptions)
932                 __sk_del_bind_node(sk);
933         else if (!nlk->subscriptions && subscriptions)
934                 sk_add_bind_node(sk, &nl_table[sk->sk_protocol].mc_list);
935         nlk->subscriptions = subscriptions;
936 }
937 
938 static int netlink_realloc_groups(struct sock *sk)
939 {
940         struct netlink_sock *nlk = nlk_sk(sk);
941         unsigned int groups;
942         unsigned long *new_groups;
943         int err = 0;
944 
945         netlink_table_grab();
946 
947         groups = nl_table[sk->sk_protocol].groups;
948         if (!nl_table[sk->sk_protocol].registered) {
949                 err = -ENOENT;
950                 goto out_unlock;
951         }
952 
953         if (nlk->ngroups >= groups)
954                 goto out_unlock;
955 
956         new_groups = krealloc(nlk->groups, NLGRPSZ(groups), GFP_ATOMIC);
957         if (new_groups == NULL) {
958                 err = -ENOMEM;
959                 goto out_unlock;
960         }
961         memset((char *)new_groups + NLGRPSZ(nlk->ngroups), 0,
962                NLGRPSZ(groups) - NLGRPSZ(nlk->ngroups));
963 
964         nlk->groups = new_groups;
965         nlk->ngroups = groups;
966  out_unlock:
967         netlink_table_ungrab();
968         return err;
969 }
970 
971 static void netlink_undo_bind(int group, long unsigned int groups,
972                               struct sock *sk)
973 {
974         struct netlink_sock *nlk = nlk_sk(sk);
975         int undo;
976 
977         if (!nlk->netlink_unbind)
978                 return;
979 
980         for (undo = 0; undo < group; undo++)
981                 if (test_bit(undo, &groups))
982                         nlk->netlink_unbind(sock_net(sk), undo + 1);
983 }
984 
985 static int netlink_bind(struct socket *sock, struct sockaddr *addr,
986                         int addr_len)
987 {
988         struct sock *sk = sock->sk;
989         struct net *net = sock_net(sk);
990         struct netlink_sock *nlk = nlk_sk(sk);
991         struct sockaddr_nl *nladdr = (struct sockaddr_nl *)addr;
992         int err = 0;
993         unsigned long groups;
994         bool bound;
995 
996         if (addr_len < sizeof(struct sockaddr_nl))
997                 return -EINVAL;
998 
999         if (nladdr->nl_family != AF_NETLINK)
1000                 return -EINVAL;
1001         groups = nladdr->nl_groups;
1002 
1003         /* Only superuser is allowed to listen multicasts */
1004         if (groups) {
1005                 if (!netlink_allowed(sock, NL_CFG_F_NONROOT_RECV))
1006                         return -EPERM;
1007                 err = netlink_realloc_groups(sk);
1008                 if (err)
1009                         return err;
1010         }
1011 
1012         if (nlk->ngroups < BITS_PER_LONG)
1013                 groups &= (1UL << nlk->ngroups) - 1;
1014 
1015         bound = nlk->bound;
1016         if (bound) {
1017                 /* Ensure nlk->portid is up-to-date. */
1018                 smp_rmb();
1019 
1020                 if (nladdr->nl_pid != nlk->portid)
1021                         return -EINVAL;
1022         }
1023 
1024         if (nlk->netlink_bind && groups) {
1025                 int group;
1026 
1027                 /* nl_groups is a u32, so cap the maximum groups we can bind */
1028                 for (group = 0; group < BITS_PER_TYPE(u32); group++) {
1029                         if (!test_bit(group, &groups))
1030                                 continue;
1031                         err = nlk->netlink_bind(net, group + 1);
1032                         if (!err)
1033                                 continue;
1034                         netlink_undo_bind(group, groups, sk);
1035                         return err;
1036                 }
1037         }
1038 
1039         /* No need for barriers here as we return to user-space without
1040          * using any of the bound attributes.
1041          */
1042         netlink_lock_table();
1043         if (!bound) {
1044                 err = nladdr->nl_pid ?
1045                         netlink_insert(sk, nladdr->nl_pid) :
1046                         netlink_autobind(sock);
1047                 if (err) {
1048                         netlink_undo_bind(BITS_PER_TYPE(u32), groups, sk);
1049                         goto unlock;
1050                 }
1051         }
1052 
1053         if (!groups && (nlk->groups == NULL || !(u32)nlk->groups[0]))
1054                 goto unlock;
1055         netlink_unlock_table();
1056 
1057         netlink_table_grab();
1058         netlink_update_subscriptions(sk, nlk->subscriptions +
1059                                          hweight32(groups) -
1060                                          hweight32(nlk->groups[0]));
1061         nlk->groups[0] = (nlk->groups[0] & ~0xffffffffUL) | groups;
1062         netlink_update_listeners(sk);
1063         netlink_table_ungrab();
1064 
1065         return 0;
1066 
1067 unlock:
1068         netlink_unlock_table();
1069         return err;
1070 }
1071 
1072 static int netlink_connect(struct socket *sock, struct sockaddr *addr,
1073                            int alen, int flags)
1074 {
1075         int err = 0;
1076         struct sock *sk = sock->sk;
1077         struct netlink_sock *nlk = nlk_sk(sk);
1078         struct sockaddr_nl *nladdr = (struct sockaddr_nl *)addr;
1079 
1080         if (alen < sizeof(addr->sa_family))
1081                 return -EINVAL;
1082 
1083         if (addr->sa_family == AF_UNSPEC) {
1084                 sk->sk_state    = NETLINK_UNCONNECTED;
1085                 nlk->dst_portid = 0;
1086                 nlk->dst_group  = 0;
1087                 return 0;
1088         }
1089         if (addr->sa_family != AF_NETLINK)
1090                 return -EINVAL;
1091 
1092         if (alen < sizeof(struct sockaddr_nl))
1093                 return -EINVAL;
1094 
1095         if ((nladdr->nl_groups || nladdr->nl_pid) &&
1096             !netlink_allowed(sock, NL_CFG_F_NONROOT_SEND))
1097                 return -EPERM;
1098 
1099         /* No need for barriers here as we return to user-space without
1100          * using any of the bound attributes.
1101          */
1102         if (!nlk->bound)
1103                 err = netlink_autobind(sock);
1104 
1105         if (err == 0) {
1106                 sk->sk_state    = NETLINK_CONNECTED;
1107                 nlk->dst_portid = nladdr->nl_pid;
1108                 nlk->dst_group  = ffs(nladdr->nl_groups);
1109         }
1110 
1111         return err;
1112 }
1113 
1114 static int netlink_getname(struct socket *sock, struct sockaddr *addr,
1115                            int peer)
1116 {
1117         struct sock *sk = sock->sk;
1118         struct netlink_sock *nlk = nlk_sk(sk);
1119         DECLARE_SOCKADDR(struct sockaddr_nl *, nladdr, addr);
1120 
1121         nladdr->nl_family = AF_NETLINK;
1122         nladdr->nl_pad = 0;
1123 
1124         if (peer) {
1125                 nladdr->nl_pid = nlk->dst_portid;
1126                 nladdr->nl_groups = netlink_group_mask(nlk->dst_group);
1127         } else {
1128                 nladdr->nl_pid = nlk->portid;
1129                 netlink_lock_table();
1130                 nladdr->nl_groups = nlk->groups ? nlk->groups[0] : 0;
1131                 netlink_unlock_table();
1132         }
1133         return sizeof(*nladdr);
1134 }
1135 
1136 static int netlink_ioctl(struct socket *sock, unsigned int cmd,
1137                          unsigned long arg)
1138 {
1139         /* try to hand this ioctl down to the NIC drivers.
1140          */
1141         return -ENOIOCTLCMD;
1142 }
1143 
1144 static struct sock *netlink_getsockbyportid(struct sock *ssk, u32 portid)
1145 {
1146         struct sock *sock;
1147         struct netlink_sock *nlk;
1148 
1149         sock = netlink_lookup(sock_net(ssk), ssk->sk_protocol, portid);
1150         if (!sock)
1151                 return ERR_PTR(-ECONNREFUSED);
1152 
1153         /* Don't bother queuing skb if kernel socket has no input function */
1154         nlk = nlk_sk(sock);
1155         if (sock->sk_state == NETLINK_CONNECTED &&
1156             nlk->dst_portid != nlk_sk(ssk)->portid) {
1157                 sock_put(sock);
1158                 return ERR_PTR(-ECONNREFUSED);
1159         }
1160         return sock;
1161 }
1162 
1163 struct sock *netlink_getsockbyfilp(struct file *filp)
1164 {
1165         struct inode *inode = file_inode(filp);
1166         struct sock *sock;
1167 
1168         if (!S_ISSOCK(inode->i_mode))
1169                 return ERR_PTR(-ENOTSOCK);
1170 
1171         sock = SOCKET_I(inode)->sk;
1172         if (sock->sk_family != AF_NETLINK)
1173                 return ERR_PTR(-EINVAL);
1174 
1175         sock_hold(sock);
1176         return sock;
1177 }
1178 
1179 static struct sk_buff *netlink_alloc_large_skb(unsigned int size,
1180                                                int broadcast)
1181 {
1182         struct sk_buff *skb;
1183         void *data;
1184 
1185         if (size <= NLMSG_GOODSIZE || broadcast)
1186                 return alloc_skb(size, GFP_KERNEL);
1187 
1188         size = SKB_DATA_ALIGN(size) +
1189                SKB_DATA_ALIGN(sizeof(struct skb_shared_info));
1190 
1191         data = vmalloc(size);
1192         if (data == NULL)
1193                 return NULL;
1194 
1195         skb = __build_skb(data, size);
1196         if (skb == NULL)
1197                 vfree(data);
1198         else
1199                 skb->destructor = netlink_skb_destructor;
1200 
1201         return skb;
1202 }
1203 
1204 /*
1205  * Attach a skb to a netlink socket.
1206  * The caller must hold a reference to the destination socket. On error, the
1207  * reference is dropped. The skb is not send to the destination, just all
1208  * all error checks are performed and memory in the queue is reserved.
1209  * Return values:
1210  * < 0: error. skb freed, reference to sock dropped.
1211  * 0: continue
1212  * 1: repeat lookup - reference dropped while waiting for socket memory.
1213  */
1214 int netlink_attachskb(struct sock *sk, struct sk_buff *skb,
1215                       long *timeo, struct sock *ssk)
1216 {
1217         struct netlink_sock *nlk;
1218 
1219         nlk = nlk_sk(sk);
1220 
1221         if ((atomic_read(&sk->sk_rmem_alloc) > sk->sk_rcvbuf ||
1222              test_bit(NETLINK_S_CONGESTED, &nlk->state))) {
1223                 DECLARE_WAITQUEUE(wait, current);
1224                 if (!*timeo) {
1225                         if (!ssk || netlink_is_kernel(ssk))
1226                                 netlink_overrun(sk);
1227                         sock_put(sk);
1228                         kfree_skb(skb);
1229                         return -EAGAIN;
1230                 }
1231 
1232                 __set_current_state(TASK_INTERRUPTIBLE);
1233                 add_wait_queue(&nlk->wait, &wait);
1234 
1235                 if ((atomic_read(&sk->sk_rmem_alloc) > sk->sk_rcvbuf ||
1236                      test_bit(NETLINK_S_CONGESTED, &nlk->state)) &&
1237                     !sock_flag(sk, SOCK_DEAD))
1238                         *timeo = schedule_timeout(*timeo);
1239 
1240                 __set_current_state(TASK_RUNNING);
1241                 remove_wait_queue(&nlk->wait, &wait);
1242                 sock_put(sk);
1243 
1244                 if (signal_pending(current)) {
1245                         kfree_skb(skb);
1246                         return sock_intr_errno(*timeo);
1247                 }
1248                 return 1;
1249         }
1250         netlink_skb_set_owner_r(skb, sk);
1251         return 0;
1252 }
1253 
1254 static int __netlink_sendskb(struct sock *sk, struct sk_buff *skb)
1255 {
1256         int len = skb->len;
1257 
1258         netlink_deliver_tap(sock_net(sk), skb);
1259 
1260         skb_queue_tail(&sk->sk_receive_queue, skb);
1261         sk->sk_data_ready(sk);
1262         return len;
1263 }
1264 
1265 int netlink_sendskb(struct sock *sk, struct sk_buff *skb)
1266 {
1267         int len = __netlink_sendskb(sk, skb);
1268 
1269         sock_put(sk);
1270         return len;
1271 }
1272 
1273 void netlink_detachskb(struct sock *sk, struct sk_buff *skb)
1274 {
1275         kfree_skb(skb);
1276         sock_put(sk);
1277 }
1278 
1279 static struct sk_buff *netlink_trim(struct sk_buff *skb, gfp_t allocation)
1280 {
1281         int delta;
1282 
1283         WARN_ON(skb->sk != NULL);
1284         delta = skb->end - skb->tail;
1285         if (is_vmalloc_addr(skb->head) || delta * 2 < skb->truesize)
1286                 return skb;
1287 
1288         if (skb_shared(skb)) {
1289                 struct sk_buff *nskb = skb_clone(skb, allocation);
1290                 if (!nskb)
1291                         return skb;
1292                 consume_skb(skb);
1293                 skb = nskb;
1294         }
1295 
1296         pskb_expand_head(skb, 0, -delta,
1297                          (allocation & ~__GFP_DIRECT_RECLAIM) |
1298                          __GFP_NOWARN | __GFP_NORETRY);
1299         return skb;
1300 }
1301 
1302 static int netlink_unicast_kernel(struct sock *sk, struct sk_buff *skb,
1303                                   struct sock *ssk)
1304 {
1305         int ret;
1306         struct netlink_sock *nlk = nlk_sk(sk);
1307 
1308         ret = -ECONNREFUSED;
1309         if (nlk->netlink_rcv != NULL) {
1310                 ret = skb->len;
1311                 netlink_skb_set_owner_r(skb, sk);
1312                 NETLINK_CB(skb).sk = ssk;
1313                 netlink_deliver_tap_kernel(sk, ssk, skb);
1314                 nlk->netlink_rcv(skb);
1315                 consume_skb(skb);
1316         } else {
1317                 kfree_skb(skb);
1318         }
1319         sock_put(sk);
1320         return ret;
1321 }
1322 
1323 int netlink_unicast(struct sock *ssk, struct sk_buff *skb,
1324                     u32 portid, int nonblock)
1325 {
1326         struct sock *sk;
1327         int err;
1328         long timeo;
1329 
1330         skb = netlink_trim(skb, gfp_any());
1331 
1332         timeo = sock_sndtimeo(ssk, nonblock);
1333 retry:
1334         sk = netlink_getsockbyportid(ssk, portid);
1335         if (IS_ERR(sk)) {
1336                 kfree_skb(skb);
1337                 return PTR_ERR(sk);
1338         }
1339         if (netlink_is_kernel(sk))
1340                 return netlink_unicast_kernel(sk, skb, ssk);
1341 
1342         if (sk_filter(sk, skb)) {
1343                 err = skb->len;
1344                 kfree_skb(skb);
1345                 sock_put(sk);
1346                 return err;
1347         }
1348 
1349         err = netlink_attachskb(sk, skb, &timeo, ssk);
1350         if (err == 1)
1351                 goto retry;
1352         if (err)
1353                 return err;
1354 
1355         return netlink_sendskb(sk, skb);
1356 }
1357 EXPORT_SYMBOL(netlink_unicast);
1358 
1359 int netlink_has_listeners(struct sock *sk, unsigned int group)
1360 {
1361         int res = 0;
1362         struct listeners *listeners;
1363 
1364         BUG_ON(!netlink_is_kernel(sk));
1365 
1366         rcu_read_lock();
1367         listeners = rcu_dereference(nl_table[sk->sk_protocol].listeners);
1368 
1369         if (listeners && group - 1 < nl_table[sk->sk_protocol].groups)
1370                 res = test_bit(group - 1, listeners->masks);
1371 
1372         rcu_read_unlock();
1373 
1374         return res;
1375 }
1376 EXPORT_SYMBOL_GPL(netlink_has_listeners);
1377 
1378 bool netlink_strict_get_check(struct sk_buff *skb)
1379 {
1380         const struct netlink_sock *nlk = nlk_sk(NETLINK_CB(skb).sk);
1381 
1382         return nlk->flags & NETLINK_F_STRICT_CHK;
1383 }
1384 EXPORT_SYMBOL_GPL(netlink_strict_get_check);
1385 
1386 static int netlink_broadcast_deliver(struct sock *sk, struct sk_buff *skb)
1387 {
1388         struct netlink_sock *nlk = nlk_sk(sk);
1389 
1390         if (atomic_read(&sk->sk_rmem_alloc) <= sk->sk_rcvbuf &&
1391             !test_bit(NETLINK_S_CONGESTED, &nlk->state)) {
1392                 netlink_skb_set_owner_r(skb, sk);
1393                 __netlink_sendskb(sk, skb);
1394                 return atomic_read(&sk->sk_rmem_alloc) > (sk->sk_rcvbuf >> 1);
1395         }
1396         return -1;
1397 }
1398 
1399 struct netlink_broadcast_data {
1400         struct sock *exclude_sk;
1401         struct net *net;
1402         u32 portid;
1403         u32 group;
1404         int failure;
1405         int delivery_failure;
1406         int congested;
1407         int delivered;
1408         gfp_t allocation;
1409         struct sk_buff *skb, *skb2;
1410         int (*tx_filter)(struct sock *dsk, struct sk_buff *skb, void *data);
1411         void *tx_data;
1412 };
1413 
1414 static void do_one_broadcast(struct sock *sk,
1415                                     struct netlink_broadcast_data *p)
1416 {
1417         struct netlink_sock *nlk = nlk_sk(sk);
1418         int val;
1419 
1420         if (p->exclude_sk == sk)
1421                 return;
1422 
1423         if (nlk->portid == p->portid || p->group - 1 >= nlk->ngroups ||
1424             !test_bit(p->group - 1, nlk->groups))
1425                 return;
1426 
1427         if (!net_eq(sock_net(sk), p->net)) {
1428                 if (!(nlk->flags & NETLINK_F_LISTEN_ALL_NSID))
1429                         return;
1430 
1431                 if (!peernet_has_id(sock_net(sk), p->net))
1432                         return;
1433 
1434                 if (!file_ns_capable(sk->sk_socket->file, p->net->user_ns,
1435                                      CAP_NET_BROADCAST))
1436                         return;
1437         }
1438 
1439         if (p->failure) {
1440                 netlink_overrun(sk);
1441                 return;
1442         }
1443 
1444         sock_hold(sk);
1445         if (p->skb2 == NULL) {
1446                 if (skb_shared(p->skb)) {
1447                         p->skb2 = skb_clone(p->skb, p->allocation);
1448                 } else {
1449                         p->skb2 = skb_get(p->skb);
1450                         /*
1451                          * skb ownership may have been set when
1452                          * delivered to a previous socket.
1453                          */
1454                         skb_orphan(p->skb2);
1455                 }
1456         }
1457         if (p->skb2 == NULL) {
1458                 netlink_overrun(sk);
1459                 /* Clone failed. Notify ALL listeners. */
1460                 p->failure = 1;
1461                 if (nlk->flags & NETLINK_F_BROADCAST_SEND_ERROR)
1462                         p->delivery_failure = 1;
1463                 goto out;
1464         }
1465         if (p->tx_filter && p->tx_filter(sk, p->skb2, p->tx_data)) {
1466                 kfree_skb(p->skb2);
1467                 p->skb2 = NULL;
1468                 goto out;
1469         }
1470         if (sk_filter(sk, p->skb2)) {
1471                 kfree_skb(p->skb2);
1472                 p->skb2 = NULL;
1473                 goto out;
1474         }
1475         NETLINK_CB(p->skb2).nsid = peernet2id(sock_net(sk), p->net);
1476         if (NETLINK_CB(p->skb2).nsid != NETNSA_NSID_NOT_ASSIGNED)
1477                 NETLINK_CB(p->skb2).nsid_is_set = true;
1478         val = netlink_broadcast_deliver(sk, p->skb2);
1479         if (val < 0) {
1480                 netlink_overrun(sk);
1481                 if (nlk->flags & NETLINK_F_BROADCAST_SEND_ERROR)
1482                         p->delivery_failure = 1;
1483         } else {
1484                 p->congested |= val;
1485                 p->delivered = 1;
1486                 p->skb2 = NULL;
1487         }
1488 out:
1489         sock_put(sk);
1490 }
1491 
1492 int netlink_broadcast_filtered(struct sock *ssk, struct sk_buff *skb, u32 portid,
1493         u32 group, gfp_t allocation,
1494         int (*filter)(struct sock *dsk, struct sk_buff *skb, void *data),
1495         void *filter_data)
1496 {
1497         struct net *net = sock_net(ssk);
1498         struct netlink_broadcast_data info;
1499         struct sock *sk;
1500 
1501         skb = netlink_trim(skb, allocation);
1502 
1503         info.exclude_sk = ssk;
1504         info.net = net;
1505         info.portid = portid;
1506         info.group = group;
1507         info.failure = 0;
1508         info.delivery_failure = 0;
1509         info.congested = 0;
1510         info.delivered = 0;
1511         info.allocation = allocation;
1512         info.skb = skb;
1513         info.skb2 = NULL;
1514         info.tx_filter = filter;
1515         info.tx_data = filter_data;
1516 
1517         /* While we sleep in clone, do not allow to change socket list */
1518 
1519         netlink_lock_table();
1520 
1521         sk_for_each_bound(sk, &nl_table[ssk->sk_protocol].mc_list)
1522                 do_one_broadcast(sk, &info);
1523 
1524         consume_skb(skb);
1525 
1526         netlink_unlock_table();
1527 
1528         if (info.delivery_failure) {
1529                 kfree_skb(info.skb2);
1530                 return -ENOBUFS;
1531         }
1532         consume_skb(info.skb2);
1533 
1534         if (info.delivered) {
1535                 if (info.congested && gfpflags_allow_blocking(allocation))
1536                         yield();
1537                 return 0;
1538         }
1539         return -ESRCH;
1540 }
1541 EXPORT_SYMBOL(netlink_broadcast_filtered);
1542 
1543 int netlink_broadcast(struct sock *ssk, struct sk_buff *skb, u32 portid,
1544                       u32 group, gfp_t allocation)
1545 {
1546         return netlink_broadcast_filtered(ssk, skb, portid, group, allocation,
1547                 NULL, NULL);
1548 }
1549 EXPORT_SYMBOL(netlink_broadcast);
1550 
1551 struct netlink_set_err_data {
1552         struct sock *exclude_sk;
1553         u32 portid;
1554         u32 group;
1555         int code;
1556 };
1557 
1558 static int do_one_set_err(struct sock *sk, struct netlink_set_err_data *p)
1559 {
1560         struct netlink_sock *nlk = nlk_sk(sk);
1561         int ret = 0;
1562 
1563         if (sk == p->exclude_sk)
1564                 goto out;
1565 
1566         if (!net_eq(sock_net(sk), sock_net(p->exclude_sk)))
1567                 goto out;
1568 
1569         if (nlk->portid == p->portid || p->group - 1 >= nlk->ngroups ||
1570             !test_bit(p->group - 1, nlk->groups))
1571                 goto out;
1572 
1573         if (p->code == ENOBUFS && nlk->flags & NETLINK_F_RECV_NO_ENOBUFS) {
1574                 ret = 1;
1575                 goto out;
1576         }
1577 
1578         sk->sk_err = p->code;
1579         sk->sk_error_report(sk);
1580 out:
1581         return ret;
1582 }
1583 
1584 /**
1585  * netlink_set_err - report error to broadcast listeners
1586  * @ssk: the kernel netlink socket, as returned by netlink_kernel_create()
1587  * @portid: the PORTID of a process that we want to skip (if any)
1588  * @group: the broadcast group that will notice the error
1589  * @code: error code, must be negative (as usual in kernelspace)
1590  *
1591  * This function returns the number of broadcast listeners that have set the
1592  * NETLINK_NO_ENOBUFS socket option.
1593  */
1594 int netlink_set_err(struct sock *ssk, u32 portid, u32 group, int code)
1595 {
1596         struct netlink_set_err_data info;
1597         struct sock *sk;
1598         int ret = 0;
1599 
1600         info.exclude_sk = ssk;
1601         info.portid = portid;
1602         info.group = group;
1603         /* sk->sk_err wants a positive error value */
1604         info.code = -code;
1605 
1606         read_lock(&nl_table_lock);
1607 
1608         sk_for_each_bound(sk, &nl_table[ssk->sk_protocol].mc_list)
1609                 ret += do_one_set_err(sk, &info);
1610 
1611         read_unlock(&nl_table_lock);
1612         return ret;
1613 }
1614 EXPORT_SYMBOL(netlink_set_err);
1615 
1616 /* must be called with netlink table grabbed */
1617 static void netlink_update_socket_mc(struct netlink_sock *nlk,
1618                                      unsigned int group,
1619                                      int is_new)
1620 {
1621         int old, new = !!is_new, subscriptions;
1622 
1623         old = test_bit(group - 1, nlk->groups);
1624         subscriptions = nlk->subscriptions - old + new;
1625         if (new)
1626                 __set_bit(group - 1, nlk->groups);
1627         else
1628                 __clear_bit(group - 1, nlk->groups);
1629         netlink_update_subscriptions(&nlk->sk, subscriptions);
1630         netlink_update_listeners(&nlk->sk);
1631 }
1632 
1633 static int netlink_setsockopt(struct socket *sock, int level, int optname,
1634                               sockptr_t optval, unsigned int optlen)
1635 {
1636         struct sock *sk = sock->sk;
1637         struct netlink_sock *nlk = nlk_sk(sk);
1638         unsigned int val = 0;
1639         int err;
1640 
1641         if (level != SOL_NETLINK)
1642                 return -ENOPROTOOPT;
1643 
1644         if (optlen >= sizeof(int) &&
1645             copy_from_sockptr(&val, optval, sizeof(val)))
1646                 return -EFAULT;
1647 
1648         switch (optname) {
1649         case NETLINK_PKTINFO:
1650                 if (val)
1651                         nlk->flags |= NETLINK_F_RECV_PKTINFO;
1652                 else
1653                         nlk->flags &= ~NETLINK_F_RECV_PKTINFO;
1654                 err = 0;
1655                 break;
1656         case NETLINK_ADD_MEMBERSHIP:
1657         case NETLINK_DROP_MEMBERSHIP: {
1658                 if (!netlink_allowed(sock, NL_CFG_F_NONROOT_RECV))
1659                         return -EPERM;
1660                 err = netlink_realloc_groups(sk);
1661                 if (err)
1662                         return err;
1663                 if (!val || val - 1 >= nlk->ngroups)
1664                         return -EINVAL;
1665                 if (optname == NETLINK_ADD_MEMBERSHIP && nlk->netlink_bind) {
1666                         err = nlk->netlink_bind(sock_net(sk), val);
1667                         if (err)
1668                                 return err;
1669                 }
1670                 netlink_table_grab();
1671                 netlink_update_socket_mc(nlk, val,
1672                                          optname == NETLINK_ADD_MEMBERSHIP);
1673                 netlink_table_ungrab();
1674                 if (optname == NETLINK_DROP_MEMBERSHIP && nlk->netlink_unbind)
1675                         nlk->netlink_unbind(sock_net(sk), val);
1676 
1677                 err = 0;
1678                 break;
1679         }
1680         case NETLINK_BROADCAST_ERROR:
1681                 if (val)
1682                         nlk->flags |= NETLINK_F_BROADCAST_SEND_ERROR;
1683                 else
1684                         nlk->flags &= ~NETLINK_F_BROADCAST_SEND_ERROR;
1685                 err = 0;
1686                 break;
1687         case NETLINK_NO_ENOBUFS:
1688                 if (val) {
1689                         nlk->flags |= NETLINK_F_RECV_NO_ENOBUFS;
1690                         clear_bit(NETLINK_S_CONGESTED, &nlk->state);
1691                         wake_up_interruptible(&nlk->wait);
1692                 } else {
1693                         nlk->flags &= ~NETLINK_F_RECV_NO_ENOBUFS;
1694                 }
1695                 err = 0;
1696                 break;
1697         case NETLINK_LISTEN_ALL_NSID:
1698                 if (!ns_capable(sock_net(sk)->user_ns, CAP_NET_BROADCAST))
1699                         return -EPERM;
1700 
1701                 if (val)
1702                         nlk->flags |= NETLINK_F_LISTEN_ALL_NSID;
1703                 else
1704                         nlk->flags &= ~NETLINK_F_LISTEN_ALL_NSID;
1705                 err = 0;
1706                 break;
1707         case NETLINK_CAP_ACK:
1708                 if (val)
1709                         nlk->flags |= NETLINK_F_CAP_ACK;
1710                 else
1711                         nlk->flags &= ~NETLINK_F_CAP_ACK;
1712                 err = 0;
1713                 break;
1714         case NETLINK_EXT_ACK:
1715                 if (val)
1716                         nlk->flags |= NETLINK_F_EXT_ACK;
1717                 else
1718                         nlk->flags &= ~NETLINK_F_EXT_ACK;
1719                 err = 0;
1720                 break;
1721         case NETLINK_GET_STRICT_CHK:
1722                 if (val)
1723                         nlk->flags |= NETLINK_F_STRICT_CHK;
1724                 else
1725                         nlk->flags &= ~NETLINK_F_STRICT_CHK;
1726                 err = 0;
1727                 break;
1728         default:
1729                 err = -ENOPROTOOPT;
1730         }
1731         return err;
1732 }
1733 
1734 static int netlink_getsockopt(struct socket *sock, int level, int optname,
1735                               char __user *optval, int __user *optlen)
1736 {
1737         struct sock *sk = sock->sk;
1738         struct netlink_sock *nlk = nlk_sk(sk);
1739         int len, val, err;
1740 
1741         if (level != SOL_NETLINK)
1742                 return -ENOPROTOOPT;
1743 
1744         if (get_user(len, optlen))
1745                 return -EFAULT;
1746         if (len < 0)
1747                 return -EINVAL;
1748 
1749         switch (optname) {
1750         case NETLINK_PKTINFO:
1751                 if (len < sizeof(int))
1752                         return -EINVAL;
1753                 len = sizeof(int);
1754                 val = nlk->flags & NETLINK_F_RECV_PKTINFO ? 1 : 0;
1755                 if (put_user(len, optlen) ||
1756                     put_user(val, optval))
1757                         return -EFAULT;
1758                 err = 0;
1759                 break;
1760         case NETLINK_BROADCAST_ERROR:
1761                 if (len < sizeof(int))
1762                         return -EINVAL;
1763                 len = sizeof(int);
1764                 val = nlk->flags & NETLINK_F_BROADCAST_SEND_ERROR ? 1 : 0;
1765                 if (put_user(len, optlen) ||
1766                     put_user(val, optval))
1767                         return -EFAULT;
1768                 err = 0;
1769                 break;
1770         case NETLINK_NO_ENOBUFS:
1771                 if (len < sizeof(int))
1772                         return -EINVAL;
1773                 len = sizeof(int);
1774                 val = nlk->flags & NETLINK_F_RECV_NO_ENOBUFS ? 1 : 0;
1775                 if (put_user(len, optlen) ||
1776                     put_user(val, optval))
1777                         return -EFAULT;
1778                 err = 0;
1779                 break;
1780         case NETLINK_LIST_MEMBERSHIPS: {
1781                 int pos, idx, shift;
1782 
1783                 err = 0;
1784                 netlink_lock_table();
1785                 for (pos = 0; pos * 8 < nlk->ngroups; pos += sizeof(u32)) {
1786                         if (len - pos < sizeof(u32))
1787                                 break;
1788 
1789                         idx = pos / sizeof(unsigned long);
1790                         shift = (pos % sizeof(unsigned long)) * 8;
1791                         if (put_user((u32)(nlk->groups[idx] >> shift),
1792                                      (u32 __user *)(optval + pos))) {
1793                                 err = -EFAULT;
1794                                 break;
1795                         }
1796                 }
1797                 if (put_user(ALIGN(nlk->ngroups / 8, sizeof(u32)), optlen))
1798                         err = -EFAULT;
1799                 netlink_unlock_table();
1800                 break;
1801         }
1802         case NETLINK_CAP_ACK:
1803                 if (len < sizeof(int))
1804                         return -EINVAL;
1805                 len = sizeof(int);
1806                 val = nlk->flags & NETLINK_F_CAP_ACK ? 1 : 0;
1807                 if (put_user(len, optlen) ||
1808                     put_user(val, optval))
1809                         return -EFAULT;
1810                 err = 0;
1811                 break;
1812         case NETLINK_EXT_ACK:
1813                 if (len < sizeof(int))
1814                         return -EINVAL;
1815                 len = sizeof(int);
1816                 val = nlk->flags & NETLINK_F_EXT_ACK ? 1 : 0;
1817                 if (put_user(len, optlen) || put_user(val, optval))
1818                         return -EFAULT;
1819                 err = 0;
1820                 break;
1821         case NETLINK_GET_STRICT_CHK:
1822                 if (len < sizeof(int))
1823                         return -EINVAL;
1824                 len = sizeof(int);
1825                 val = nlk->flags & NETLINK_F_STRICT_CHK ? 1 : 0;
1826                 if (put_user(len, optlen) || put_user(val, optval))
1827                         return -EFAULT;
1828                 err = 0;
1829                 break;
1830         default:
1831                 err = -ENOPROTOOPT;
1832         }
1833         return err;
1834 }
1835 
1836 static void netlink_cmsg_recv_pktinfo(struct msghdr *msg, struct sk_buff *skb)
1837 {
1838         struct nl_pktinfo info;
1839 
1840         info.group = NETLINK_CB(skb).dst_group;
1841         put_cmsg(msg, SOL_NETLINK, NETLINK_PKTINFO, sizeof(info), &info);
1842 }
1843 
1844 static void netlink_cmsg_listen_all_nsid(struct sock *sk, struct msghdr *msg,
1845                                          struct sk_buff *skb)
1846 {
1847         if (!NETLINK_CB(skb).nsid_is_set)
1848                 return;
1849 
1850         put_cmsg(msg, SOL_NETLINK, NETLINK_LISTEN_ALL_NSID, sizeof(int),
1851                  &NETLINK_CB(skb).nsid);
1852 }
1853 
1854 static int netlink_sendmsg(struct socket *sock, struct msghdr *msg, size_t len)
1855 {
1856         struct sock *sk = sock->sk;
1857         struct netlink_sock *nlk = nlk_sk(sk);
1858         DECLARE_SOCKADDR(struct sockaddr_nl *, addr, msg->msg_name);
1859         u32 dst_portid;
1860         u32 dst_group;
1861         struct sk_buff *skb;
1862         int err;
1863         struct scm_cookie scm;
1864         u32 netlink_skb_flags = 0;
1865 
1866         if (msg->msg_flags & MSG_OOB)
1867                 return -EOPNOTSUPP;
1868 
1869         err = scm_send(sock, msg, &scm, true);
1870         if (err < 0)
1871                 return err;
1872 
1873         if (msg->msg_namelen) {
1874                 err = -EINVAL;
1875                 if (msg->msg_namelen < sizeof(struct sockaddr_nl))
1876                         goto out;
1877                 if (addr->nl_family != AF_NETLINK)
1878                         goto out;
1879                 dst_portid = addr->nl_pid;
1880                 dst_group = ffs(addr->nl_groups);
1881                 err =  -EPERM;
1882                 if ((dst_group || dst_portid) &&
1883                     !netlink_allowed(sock, NL_CFG_F_NONROOT_SEND))
1884                         goto out;
1885                 netlink_skb_flags |= NETLINK_SKB_DST;
1886         } else {
1887                 dst_portid = nlk->dst_portid;
1888                 dst_group = nlk->dst_group;
1889         }
1890 
1891         if (!nlk->bound) {
1892                 err = netlink_autobind(sock);
1893                 if (err)
1894                         goto out;
1895         } else {
1896                 /* Ensure nlk is hashed and visible. */
1897                 smp_rmb();
1898         }
1899 
1900         err = -EMSGSIZE;
1901         if (len > sk->sk_sndbuf - 32)
1902                 goto out;
1903         err = -ENOBUFS;
1904         skb = netlink_alloc_large_skb(len, dst_group);
1905         if (skb == NULL)
1906                 goto out;
1907 
1908         NETLINK_CB(skb).portid  = nlk->portid;
1909         NETLINK_CB(skb).dst_group = dst_group;
1910         NETLINK_CB(skb).creds   = scm.creds;
1911         NETLINK_CB(skb).flags   = netlink_skb_flags;
1912 
1913         err = -EFAULT;
1914         if (memcpy_from_msg(skb_put(skb, len), msg, len)) {
1915                 kfree_skb(skb);
1916                 goto out;
1917         }
1918 
1919         err = security_netlink_send(sk, skb);
1920         if (err) {
1921                 kfree_skb(skb);
1922                 goto out;
1923         }
1924 
1925         if (dst_group) {
1926                 refcount_inc(&skb->users);
1927                 netlink_broadcast(sk, skb, dst_portid, dst_group, GFP_KERNEL);
1928         }
1929         err = netlink_unicast(sk, skb, dst_portid, msg->msg_flags & MSG_DONTWAIT);
1930 
1931 out:
1932         scm_destroy(&scm);
1933         return err;
1934 }
1935 
1936 static int netlink_recvmsg(struct socket *sock, struct msghdr *msg, size_t len,
1937                            int flags)
1938 {
1939         struct scm_cookie scm;
1940         struct sock *sk = sock->sk;
1941         struct netlink_sock *nlk = nlk_sk(sk);
1942         int noblock = flags & MSG_DONTWAIT;
1943         size_t copied;
1944         struct sk_buff *skb, *data_skb;
1945         int err, ret;
1946 
1947         if (flags & MSG_OOB)
1948                 return -EOPNOTSUPP;
1949 
1950         copied = 0;
1951 
1952         skb = skb_recv_datagram(sk, flags, noblock, &err);
1953         if (skb == NULL)
1954                 goto out;
1955 
1956         data_skb = skb;
1957 
1958 #ifdef CONFIG_COMPAT_NETLINK_MESSAGES
1959         if (unlikely(skb_shinfo(skb)->frag_list)) {
1960                 /*
1961                  * If this skb has a frag_list, then here that means that we
1962                  * will have to use the frag_list skb's data for compat tasks
1963                  * and the regular skb's data for normal (non-compat) tasks.
1964                  *
1965                  * If we need to send the compat skb, assign it to the
1966                  * 'data_skb' variable so that it will be used below for data
1967                  * copying. We keep 'skb' for everything else, including
1968                  * freeing both later.
1969                  */
1970                 if (flags & MSG_CMSG_COMPAT)
1971                         data_skb = skb_shinfo(skb)->frag_list;
1972         }
1973 #endif
1974 
1975         /* Record the max length of recvmsg() calls for future allocations */
1976         nlk->max_recvmsg_len = max(nlk->max_recvmsg_len, len);
1977         nlk->max_recvmsg_len = min_t(size_t, nlk->max_recvmsg_len,
1978                                      SKB_WITH_OVERHEAD(32768));
1979 
1980         copied = data_skb->len;
1981         if (len < copied) {
1982                 msg->msg_flags |= MSG_TRUNC;
1983                 copied = len;
1984         }
1985 
1986         skb_reset_transport_header(data_skb);
1987         err = skb_copy_datagram_msg(data_skb, 0, msg, copied);
1988 
1989         if (msg->msg_name) {
1990                 DECLARE_SOCKADDR(struct sockaddr_nl *, addr, msg->msg_name);
1991                 addr->nl_family = AF_NETLINK;
1992                 addr->nl_pad    = 0;
1993                 addr->nl_pid    = NETLINK_CB(skb).portid;
1994                 addr->nl_groups = netlink_group_mask(NETLINK_CB(skb).dst_group);
1995                 msg->msg_namelen = sizeof(*addr);
1996         }
1997 
1998         if (nlk->flags & NETLINK_F_RECV_PKTINFO)
1999                 netlink_cmsg_recv_pktinfo(msg, skb);
2000         if (nlk->flags & NETLINK_F_LISTEN_ALL_NSID)
2001                 netlink_cmsg_listen_all_nsid(sk, msg, skb);
2002 
2003         memset(&scm, 0, sizeof(scm));
2004         scm.creds = *NETLINK_CREDS(skb);
2005         if (flags & MSG_TRUNC)
2006                 copied = data_skb->len;
2007 
2008         skb_free_datagram(sk, skb);
2009 
2010         if (nlk->cb_running &&
2011             atomic_read(&sk->sk_rmem_alloc) <= sk->sk_rcvbuf / 2) {
2012                 ret = netlink_dump(sk);
2013                 if (ret) {
2014                         sk->sk_err = -ret;
2015                         sk->sk_error_report(sk);
2016                 }
2017         }
2018 
2019         scm_recv(sock, msg, &scm, flags);
2020 out:
2021         netlink_rcv_wake(sk);
2022         return err ? : copied;
2023 }
2024 
2025 static void netlink_data_ready(struct sock *sk)
2026 {
2027         BUG();
2028 }
2029 
2030 /*
2031  *      We export these functions to other modules. They provide a
2032  *      complete set of kernel non-blocking support for message
2033  *      queueing.
2034  */
2035 
2036 struct sock *
2037 __netlink_kernel_create(struct net *net, int unit, struct module *module,
2038                         struct netlink_kernel_cfg *cfg)
2039 {
2040         struct socket *sock;
2041         struct sock *sk;
2042         struct netlink_sock *nlk;
2043         struct listeners *listeners = NULL;
2044         struct mutex *cb_mutex = cfg ? cfg->cb_mutex : NULL;
2045         unsigned int groups;
2046 
2047         BUG_ON(!nl_table);
2048 
2049         if (unit < 0 || unit >= MAX_LINKS)
2050                 return NULL;
2051 
2052         if (sock_create_lite(PF_NETLINK, SOCK_DGRAM, unit, &sock))
2053                 return NULL;
2054 
2055         if (__netlink_create(net, sock, cb_mutex, unit, 1) < 0)
2056                 goto out_sock_release_nosk;
2057 
2058         sk = sock->sk;
2059 
2060         if (!cfg || cfg->groups < 32)
2061                 groups = 32;
2062         else
2063                 groups = cfg->groups;
2064 
2065         listeners = kzalloc(sizeof(*listeners) + NLGRPSZ(groups), GFP_KERNEL);
2066         if (!listeners)
2067                 goto out_sock_release;
2068 
2069         sk->sk_data_ready = netlink_data_ready;
2070         if (cfg && cfg->input)
2071                 nlk_sk(sk)->netlink_rcv = cfg->input;
2072 
2073         if (netlink_insert(sk, 0))
2074                 goto out_sock_release;
2075 
2076         nlk = nlk_sk(sk);
2077         nlk->flags |= NETLINK_F_KERNEL_SOCKET;
2078 
2079         netlink_table_grab();
2080         if (!nl_table[unit].registered) {
2081                 nl_table[unit].groups = groups;
2082                 rcu_assign_pointer(nl_table[unit].listeners, listeners);
2083                 nl_table[unit].cb_mutex = cb_mutex;
2084                 nl_table[unit].module = module;
2085                 if (cfg) {
2086                         nl_table[unit].bind = cfg->bind;
2087                         nl_table[unit].unbind = cfg->unbind;
2088                         nl_table[unit].flags = cfg->flags;
2089                         if (cfg->compare)
2090                                 nl_table[unit].compare = cfg->compare;
2091                 }
2092                 nl_table[unit].registered = 1;
2093         } else {
2094                 kfree(listeners);
2095                 nl_table[unit].registered++;
2096         }
2097         netlink_table_ungrab();
2098         return sk;
2099 
2100 out_sock_release:
2101         kfree(listeners);
2102         netlink_kernel_release(sk);
2103         return NULL;
2104 
2105 out_sock_release_nosk:
2106         sock_release(sock);
2107         return NULL;
2108 }
2109 EXPORT_SYMBOL(__netlink_kernel_create);
2110 
2111 void
2112 netlink_kernel_release(struct sock *sk)
2113 {
2114         if (sk == NULL || sk->sk_socket == NULL)
2115                 return;
2116 
2117         sock_release(sk->sk_socket);
2118 }
2119 EXPORT_SYMBOL(netlink_kernel_release);
2120 
2121 int __netlink_change_ngroups(struct sock *sk, unsigned int groups)
2122 {
2123         struct listeners *new, *old;
2124         struct netlink_table *tbl = &nl_table[sk->sk_protocol];
2125 
2126         if (groups < 32)
2127                 groups = 32;
2128 
2129         if (NLGRPSZ(tbl->groups) < NLGRPSZ(groups)) {
2130                 new = kzalloc(sizeof(*new) + NLGRPSZ(groups), GFP_ATOMIC);
2131                 if (!new)
2132                         return -ENOMEM;
2133                 old = nl_deref_protected(tbl->listeners);
2134                 memcpy(new->masks, old->masks, NLGRPSZ(tbl->groups));
2135                 rcu_assign_pointer(tbl->listeners, new);
2136 
2137                 kfree_rcu(old, rcu);
2138         }
2139         tbl->groups = groups;
2140 
2141         return 0;
2142 }
2143 
2144 /**
2145  * netlink_change_ngroups - change number of multicast groups
2146  *
2147  * This changes the number of multicast groups that are available
2148  * on a certain netlink family. Note that it is not possible to
2149  * change the number of groups to below 32. Also note that it does
2150  * not implicitly call netlink_clear_multicast_users() when the
2151  * number of groups is reduced.
2152  *
2153  * @sk: The kernel netlink socket, as returned by netlink_kernel_create().
2154  * @groups: The new number of groups.
2155  */
2156 int netlink_change_ngroups(struct sock *sk, unsigned int groups)
2157 {
2158         int err;
2159 
2160         netlink_table_grab();
2161         err = __netlink_change_ngroups(sk, groups);
2162         netlink_table_ungrab();
2163 
2164         return err;
2165 }
2166 
2167 void __netlink_clear_multicast_users(struct sock *ksk, unsigned int group)
2168 {
2169         struct sock *sk;
2170         struct netlink_table *tbl = &nl_table[ksk->sk_protocol];
2171 
2172         sk_for_each_bound(sk, &tbl->mc_list)
2173                 netlink_update_socket_mc(nlk_sk(sk), group, 0);
2174 }
2175 
2176 struct nlmsghdr *
2177 __nlmsg_put(struct sk_buff *skb, u32 portid, u32 seq, int type, int len, int flags)
2178 {
2179         struct nlmsghdr *nlh;
2180         int size = nlmsg_msg_size(len);
2181 
2182         nlh = skb_put(skb, NLMSG_ALIGN(size));
2183         nlh->nlmsg_type = type;
2184         nlh->nlmsg_len = size;
2185         nlh->nlmsg_flags = flags;
2186         nlh->nlmsg_pid = portid;
2187         nlh->nlmsg_seq = seq;
2188         if (!__builtin_constant_p(size) || NLMSG_ALIGN(size) - size != 0)
2189                 memset(nlmsg_data(nlh) + len, 0, NLMSG_ALIGN(size) - size);
2190         return nlh;
2191 }
2192 EXPORT_SYMBOL(__nlmsg_put);
2193 
2194 /*
2195  * It looks a bit ugly.
2196  * It would be better to create kernel thread.
2197  */
2198 
2199 static int netlink_dump_done(struct netlink_sock *nlk, struct sk_buff *skb,
2200                              struct netlink_callback *cb,
2201                              struct netlink_ext_ack *extack)
2202 {
2203         struct nlmsghdr *nlh;
2204 
2205         nlh = nlmsg_put_answer(skb, cb, NLMSG_DONE, sizeof(nlk->dump_done_errno),
2206                                NLM_F_MULTI | cb->answer_flags);
2207         if (WARN_ON(!nlh))
2208                 return -ENOBUFS;
2209 
2210         nl_dump_check_consistent(cb, nlh);
2211         memcpy(nlmsg_data(nlh), &nlk->dump_done_errno, sizeof(nlk->dump_done_errno));
2212 
2213         if (extack->_msg && nlk->flags & NETLINK_F_EXT_ACK) {
2214                 nlh->nlmsg_flags |= NLM_F_ACK_TLVS;
2215                 if (!nla_put_string(skb, NLMSGERR_ATTR_MSG, extack->_msg))
2216                         nlmsg_end(skb, nlh);
2217         }
2218 
2219         return 0;
2220 }
2221 
2222 static int netlink_dump(struct sock *sk)
2223 {
2224         struct netlink_sock *nlk = nlk_sk(sk);
2225         struct netlink_ext_ack extack = {};
2226         struct netlink_callback *cb;
2227         struct sk_buff *skb = NULL;
2228         struct module *module;
2229         int err = -ENOBUFS;
2230         int alloc_min_size;
2231         int alloc_size;
2232 
2233         mutex_lock(nlk->cb_mutex);
2234         if (!nlk->cb_running) {
2235                 err = -EINVAL;
2236                 goto errout_skb;
2237         }
2238 
2239         if (atomic_read(&sk->sk_rmem_alloc) >= sk->sk_rcvbuf)
2240                 goto errout_skb;
2241 
2242         /* NLMSG_GOODSIZE is small to avoid high order allocations being
2243          * required, but it makes sense to _attempt_ a 16K bytes allocation
2244          * to reduce number of system calls on dump operations, if user
2245          * ever provided a big enough buffer.
2246          */
2247         cb = &nlk->cb;
2248         alloc_min_size = max_t(int, cb->min_dump_alloc, NLMSG_GOODSIZE);
2249 
2250         if (alloc_min_size < nlk->max_recvmsg_len) {
2251                 alloc_size = nlk->max_recvmsg_len;
2252                 skb = alloc_skb(alloc_size,
2253                                 (GFP_KERNEL & ~__GFP_DIRECT_RECLAIM) |
2254                                 __GFP_NOWARN | __GFP_NORETRY);
2255         }
2256         if (!skb) {
2257                 alloc_size = alloc_min_size;
2258                 skb = alloc_skb(alloc_size, GFP_KERNEL);
2259         }
2260         if (!skb)
2261                 goto errout_skb;
2262 
2263         /* Trim skb to allocated size. User is expected to provide buffer as
2264          * large as max(min_dump_alloc, 16KiB (mac_recvmsg_len capped at
2265          * netlink_recvmsg())). dump will pack as many smaller messages as
2266          * could fit within the allocated skb. skb is typically allocated
2267          * with larger space than required (could be as much as near 2x the
2268          * requested size with align to next power of 2 approach). Allowing
2269          * dump to use the excess space makes it difficult for a user to have a
2270          * reasonable static buffer based on the expected largest dump of a
2271          * single netdev. The outcome is MSG_TRUNC error.
2272          */
2273         skb_reserve(skb, skb_tailroom(skb) - alloc_size);
2274         netlink_skb_set_owner_r(skb, sk);
2275 
2276         if (nlk->dump_done_errno > 0) {
2277                 cb->extack = &extack;
2278                 nlk->dump_done_errno = cb->dump(skb, cb);
2279                 cb->extack = NULL;
2280         }
2281 
2282         if (nlk->dump_done_errno > 0 ||
2283             skb_tailroom(skb) < nlmsg_total_size(sizeof(nlk->dump_done_errno))) {
2284                 mutex_unlock(nlk->cb_mutex);
2285 
2286                 if (sk_filter(sk, skb))
2287                         kfree_skb(skb);
2288                 else
2289                         __netlink_sendskb(sk, skb);
2290                 return 0;
2291         }
2292 
2293         if (netlink_dump_done(nlk, skb, cb, &extack))
2294                 goto errout_skb;
2295 
2296 #ifdef CONFIG_COMPAT_NETLINK_MESSAGES
2297         /* frag_list skb's data is used for compat tasks
2298          * and the regular skb's data for normal (non-compat) tasks.
2299          * See netlink_recvmsg().
2300          */
2301         if (unlikely(skb_shinfo(skb)->frag_list)) {
2302                 if (netlink_dump_done(nlk, skb_shinfo(skb)->frag_list, cb, &extack))
2303                         goto errout_skb;
2304         }
2305 #endif
2306 
2307         if (sk_filter(sk, skb))
2308                 kfree_skb(skb);
2309         else
2310                 __netlink_sendskb(sk, skb);
2311 
2312         if (cb->done)
2313                 cb->done(cb);
2314 
2315         nlk->cb_running = false;
2316         module = cb->module;
2317         skb = cb->skb;
2318         mutex_unlock(nlk->cb_mutex);
2319         module_put(module);
2320         consume_skb(skb);
2321         return 0;
2322 
2323 errout_skb:
2324         mutex_unlock(nlk->cb_mutex);
2325         kfree_skb(skb);
2326         return err;
2327 }
2328 
2329 int __netlink_dump_start(struct sock *ssk, struct sk_buff *skb,
2330                          const struct nlmsghdr *nlh,
2331                          struct netlink_dump_control *control)
2332 {
2333         struct netlink_sock *nlk, *nlk2;
2334         struct netlink_callback *cb;
2335         struct sock *sk;
2336         int ret;
2337 
2338         refcount_inc(&skb->users);
2339 
2340         sk = netlink_lookup(sock_net(ssk), ssk->sk_protocol, NETLINK_CB(skb).portid);
2341         if (sk == NULL) {
2342                 ret = -ECONNREFUSED;
2343                 goto error_free;
2344         }
2345 
2346         nlk = nlk_sk(sk);
2347         mutex_lock(nlk->cb_mutex);
2348         /* A dump is in progress... */
2349         if (nlk->cb_running) {
2350                 ret = -EBUSY;
2351                 goto error_unlock;
2352         }
2353         /* add reference of module which cb->dump belongs to */
2354         if (!try_module_get(control->module)) {
2355                 ret = -EPROTONOSUPPORT;
2356                 goto error_unlock;
2357         }
2358 
2359         cb = &nlk->cb;
2360         memset(cb, 0, sizeof(*cb));
2361         cb->dump = control->dump;
2362         cb->done = control->done;
2363         cb->nlh = nlh;
2364         cb->data = control->data;
2365         cb->module = control->module;
2366         cb->min_dump_alloc = control->min_dump_alloc;
2367         cb->skb = skb;
2368 
2369         nlk2 = nlk_sk(NETLINK_CB(skb).sk);
2370         cb->strict_check = !!(nlk2->flags & NETLINK_F_STRICT_CHK);
2371 
2372         if (control->start) {
2373                 ret = control->start(cb);
2374                 if (ret)
2375                         goto error_put;
2376         }
2377 
2378         nlk->cb_running = true;
2379         nlk->dump_done_errno = INT_MAX;
2380 
2381         mutex_unlock(nlk->cb_mutex);
2382 
2383         ret = netlink_dump(sk);
2384 
2385         sock_put(sk);
2386 
2387         if (ret)
2388                 return ret;
2389 
2390         /* We successfully started a dump, by returning -EINTR we
2391          * signal not to send ACK even if it was requested.
2392          */
2393         return -EINTR;
2394 
2395 error_put:
2396         module_put(control->module);
2397 error_unlock:
2398         sock_put(sk);
2399         mutex_unlock(nlk->cb_mutex);
2400 error_free:
2401         kfree_skb(skb);
2402         return ret;
2403 }
2404 EXPORT_SYMBOL(__netlink_dump_start);
2405 
2406 void netlink_ack(struct sk_buff *in_skb, struct nlmsghdr *nlh, int err,
2407                  const struct netlink_ext_ack *extack)
2408 {
2409         struct sk_buff *skb;
2410         struct nlmsghdr *rep;
2411         struct nlmsgerr *errmsg;
2412         size_t payload = sizeof(*errmsg);
2413         size_t tlvlen = 0;
2414         struct netlink_sock *nlk = nlk_sk(NETLINK_CB(in_skb).sk);
2415         unsigned int flags = 0;
2416         bool nlk_has_extack = nlk->flags & NETLINK_F_EXT_ACK;
2417 
2418         /* Error messages get the original request appened, unless the user
2419          * requests to cap the error message, and get extra error data if
2420          * requested.
2421          */
2422         if (nlk_has_extack && extack && extack->_msg)
2423                 tlvlen += nla_total_size(strlen(extack->_msg) + 1);
2424 
2425         if (err && !(nlk->flags & NETLINK_F_CAP_ACK))
2426                 payload += nlmsg_len(nlh);
2427         else
2428                 flags |= NLM_F_CAPPED;
2429         if (err && nlk_has_extack && extack && extack->bad_attr)
2430                 tlvlen += nla_total_size(sizeof(u32));
2431         if (nlk_has_extack && extack && extack->cookie_len)
2432                 tlvlen += nla_total_size(extack->cookie_len);
2433         if (err && nlk_has_extack && extack && extack->policy)
2434                 tlvlen += netlink_policy_dump_attr_size_estimate(extack->policy);
2435 
2436         if (tlvlen)
2437                 flags |= NLM_F_ACK_TLVS;
2438 
2439         skb = nlmsg_new(payload + tlvlen, GFP_KERNEL);
2440         if (!skb) {
2441                 NETLINK_CB(in_skb).sk->sk_err = ENOBUFS;
2442                 NETLINK_CB(in_skb).sk->sk_error_report(NETLINK_CB(in_skb).sk);
2443                 return;
2444         }
2445 
2446         rep = __nlmsg_put(skb, NETLINK_CB(in_skb).portid, nlh->nlmsg_seq,
2447                           NLMSG_ERROR, payload, flags);
2448         errmsg = nlmsg_data(rep);
2449         errmsg->error = err;
2450         memcpy(&errmsg->msg, nlh, payload > sizeof(*errmsg) ? nlh->nlmsg_len : sizeof(*nlh));
2451 
2452         if (nlk_has_extack && extack) {
2453                 if (extack->_msg) {
2454                         WARN_ON(nla_put_string(skb, NLMSGERR_ATTR_MSG,
2455                                                extack->_msg));
2456                 }
2457                 if (err && extack->bad_attr &&
2458                     !WARN_ON((u8 *)extack->bad_attr < in_skb->data ||
2459                              (u8 *)extack->bad_attr >= in_skb->data +
2460                                                        in_skb->len))
2461                         WARN_ON(nla_put_u32(skb, NLMSGERR_ATTR_OFFS,
2462                                             (u8 *)extack->bad_attr -
2463                                             (u8 *)nlh));
2464                 if (extack->cookie_len)
2465                         WARN_ON(nla_put(skb, NLMSGERR_ATTR_COOKIE,
2466                                         extack->cookie_len, extack->cookie));
2467                 if (extack->policy)
2468                         netlink_policy_dump_write_attr(skb, extack->policy,
2469                                                        NLMSGERR_ATTR_POLICY);
2470         }
2471 
2472         nlmsg_end(skb, rep);
2473 
2474         netlink_unicast(in_skb->sk, skb, NETLINK_CB(in_skb).portid, MSG_DONTWAIT);
2475 }
2476 EXPORT_SYMBOL(netlink_ack);
2477 
2478 int netlink_rcv_skb(struct sk_buff *skb, int (*cb)(struct sk_buff *,
2479                                                    struct nlmsghdr *,
2480                                                    struct netlink_ext_ack *))
2481 {
2482         struct netlink_ext_ack extack;
2483         struct nlmsghdr *nlh;
2484         int err;
2485 
2486         while (skb->len >= nlmsg_total_size(0)) {
2487                 int msglen;
2488 
2489                 memset(&extack, 0, sizeof(extack));
2490                 nlh = nlmsg_hdr(skb);
2491                 err = 0;
2492 
2493                 if (nlh->nlmsg_len < NLMSG_HDRLEN || skb->len < nlh->nlmsg_len)
2494                         return 0;
2495 
2496                 /* Only requests are handled by the kernel */
2497                 if (!(nlh->nlmsg_flags & NLM_F_REQUEST))
2498                         goto ack;
2499 
2500                 /* Skip control messages */
2501                 if (nlh->nlmsg_type < NLMSG_MIN_TYPE)
2502                         goto ack;
2503 
2504                 err = cb(skb, nlh, &extack);
2505                 if (err == -EINTR)
2506                         goto skip;
2507 
2508 ack:
2509                 if (nlh->nlmsg_flags & NLM_F_ACK || err)
2510                         netlink_ack(skb, nlh, err, &extack);
2511 
2512 skip:
2513                 msglen = NLMSG_ALIGN(nlh->nlmsg_len);
2514                 if (msglen > skb->len)
2515                         msglen = skb->len;
2516                 skb_pull(skb, msglen);
2517         }
2518 
2519         return 0;
2520 }
2521 EXPORT_SYMBOL(netlink_rcv_skb);
2522 
2523 /**
2524  * nlmsg_notify - send a notification netlink message
2525  * @sk: netlink socket to use
2526  * @skb: notification message
2527  * @portid: destination netlink portid for reports or 0
2528  * @group: destination multicast group or 0
2529  * @report: 1 to report back, 0 to disable
2530  * @flags: allocation flags
2531  */
2532 int nlmsg_notify(struct sock *sk, struct sk_buff *skb, u32 portid,
2533                  unsigned int group, int report, gfp_t flags)
2534 {
2535         int err = 0;
2536 
2537         if (group) {
2538                 int exclude_portid = 0;
2539 
2540                 if (report) {
2541                         refcount_inc(&skb->users);
2542                         exclude_portid = portid;
2543                 }
2544 
2545                 /* errors reported via destination sk->sk_err, but propagate
2546                  * delivery errors if NETLINK_BROADCAST_ERROR flag is set */
2547                 err = nlmsg_multicast(sk, skb, exclude_portid, group, flags);
2548                 if (err == -ESRCH)
2549                         err = 0;
2550         }
2551 
2552         if (report) {
2553                 int err2;
2554 
2555                 err2 = nlmsg_unicast(sk, skb, portid);
2556                 if (!err)
2557                         err = err2;
2558         }
2559 
2560         return err;
2561 }
2562 EXPORT_SYMBOL(nlmsg_notify);
2563 
2564 #ifdef CONFIG_PROC_FS
2565 struct nl_seq_iter {
2566         struct seq_net_private p;
2567         struct rhashtable_iter hti;
2568         int link;
2569 };
2570 
2571 static void netlink_walk_start(struct nl_seq_iter *iter)
2572 {
2573         rhashtable_walk_enter(&nl_table[iter->link].hash, &iter->hti);
2574         rhashtable_walk_start(&iter->hti);
2575 }
2576 
2577 static void netlink_walk_stop(struct nl_seq_iter *iter)
2578 {
2579         rhashtable_walk_stop(&iter->hti);
2580         rhashtable_walk_exit(&iter->hti);
2581 }
2582 
2583 static void *__netlink_seq_next(struct seq_file *seq)
2584 {
2585         struct nl_seq_iter *iter = seq->private;
2586         struct netlink_sock *nlk;
2587 
2588         do {
2589                 for (;;) {
2590                         nlk = rhashtable_walk_next(&iter->hti);
2591 
2592                         if (IS_ERR(nlk)) {
2593                                 if (PTR_ERR(nlk) == -EAGAIN)
2594                                         continue;
2595 
2596                                 return nlk;
2597                         }
2598 
2599                         if (nlk)
2600                                 break;
2601 
2602                         netlink_walk_stop(iter);
2603                         if (++iter->link >= MAX_LINKS)
2604                                 return NULL;
2605 
2606                         netlink_walk_start(iter);
2607                 }
2608         } while (sock_net(&nlk->sk) != seq_file_net(seq));
2609 
2610         return nlk;
2611 }
2612 
2613 static void *netlink_seq_start(struct seq_file *seq, loff_t *posp)
2614         __acquires(RCU)
2615 {
2616         struct nl_seq_iter *iter = seq->private;
2617         void *obj = SEQ_START_TOKEN;
2618         loff_t pos;
2619 
2620         iter->link = 0;
2621 
2622         netlink_walk_start(iter);
2623 
2624         for (pos = *posp; pos && obj && !IS_ERR(obj); pos--)
2625                 obj = __netlink_seq_next(seq);
2626 
2627         return obj;
2628 }
2629 
2630 static void *netlink_seq_next(struct seq_file *seq, void *v, loff_t *pos)
2631 {
2632         ++*pos;
2633         return __netlink_seq_next(seq);
2634 }
2635 
2636 static void netlink_native_seq_stop(struct seq_file *seq, void *v)
2637 {
2638         struct nl_seq_iter *iter = seq->private;
2639 
2640         if (iter->link >= MAX_LINKS)
2641                 return;
2642 
2643         netlink_walk_stop(iter);
2644 }
2645 
2646 
2647 static int netlink_native_seq_show(struct seq_file *seq, void *v)
2648 {
2649         if (v == SEQ_START_TOKEN) {
2650                 seq_puts(seq,
2651                          "sk               Eth Pid        Groups   "
2652                          "Rmem     Wmem     Dump  Locks    Drops    Inode\n");
2653         } else {
2654                 struct sock *s = v;
2655                 struct netlink_sock *nlk = nlk_sk(s);
2656 
2657                 seq_printf(seq, "%pK %-3d %-10u %08x %-8d %-8d %-5d %-8d %-8u %-8lu\n",
2658                            s,
2659                            s->sk_protocol,
2660                            nlk->portid,
2661                            nlk->groups ? (u32)nlk->groups[0] : 0,
2662                            sk_rmem_alloc_get(s),
2663                            sk_wmem_alloc_get(s),
2664                            nlk->cb_running,
2665                            refcount_read(&s->sk_refcnt),
2666                            atomic_read(&s->sk_drops),
2667                            sock_i_ino(s)
2668                         );
2669 
2670         }
2671         return 0;
2672 }
2673 
2674 #ifdef CONFIG_BPF_SYSCALL
2675 struct bpf_iter__netlink {
2676         __bpf_md_ptr(struct bpf_iter_meta *, meta);
2677         __bpf_md_ptr(struct netlink_sock *, sk);
2678 };
2679 
2680 DEFINE_BPF_ITER_FUNC(netlink, struct bpf_iter_meta *meta, struct netlink_sock *sk)
2681 
2682 static int netlink_prog_seq_show(struct bpf_prog *prog,
2683                                   struct bpf_iter_meta *meta,
2684                                   void *v)
2685 {
2686         struct bpf_iter__netlink ctx;
2687 
2688         meta->seq_num--;  /* skip SEQ_START_TOKEN */
2689         ctx.meta = meta;
2690         ctx.sk = nlk_sk((struct sock *)v);
2691         return bpf_iter_run_prog(prog, &ctx);
2692 }
2693 
2694 static int netlink_seq_show(struct seq_file *seq, void *v)
2695 {
2696         struct bpf_iter_meta meta;
2697         struct bpf_prog *prog;
2698 
2699         meta.seq = seq;
2700         prog = bpf_iter_get_info(&meta, false);
2701         if (!prog)
2702                 return netlink_native_seq_show(seq, v);
2703 
2704         if (v != SEQ_START_TOKEN)
2705                 return netlink_prog_seq_show(prog, &meta, v);
2706 
2707         return 0;
2708 }
2709 
2710 static void netlink_seq_stop(struct seq_file *seq, void *v)
2711 {
2712         struct bpf_iter_meta meta;
2713         struct bpf_prog *prog;
2714 
2715         if (!v) {
2716                 meta.seq = seq;
2717                 prog = bpf_iter_get_info(&meta, true);
2718                 if (prog)
2719                         (void)netlink_prog_seq_show(prog, &meta, v);
2720         }
2721 
2722         netlink_native_seq_stop(seq, v);
2723 }
2724 #else
2725 static int netlink_seq_show(struct seq_file *seq, void *v)
2726 {
2727         return netlink_native_seq_show(seq, v);
2728 }
2729 
2730 static void netlink_seq_stop(struct seq_file *seq, void *v)
2731 {
2732         netlink_native_seq_stop(seq, v);
2733 }
2734 #endif
2735 
2736 static const struct seq_operations netlink_seq_ops = {
2737         .start  = netlink_seq_start,
2738         .next   = netlink_seq_next,
2739         .stop   = netlink_seq_stop,
2740         .show   = netlink_seq_show,
2741 };
2742 #endif
2743 
2744 int netlink_register_notifier(struct notifier_block *nb)
2745 {
2746         return blocking_notifier_chain_register(&netlink_chain, nb);
2747 }
2748 EXPORT_SYMBOL(netlink_register_notifier);
2749 
2750 int netlink_unregister_notifier(struct notifier_block *nb)
2751 {
2752         return blocking_notifier_chain_unregister(&netlink_chain, nb);
2753 }
2754 EXPORT_SYMBOL(netlink_unregister_notifier);
2755 
2756 static const struct proto_ops netlink_ops = {
2757         .family =       PF_NETLINK,
2758         .owner =        THIS_MODULE,
2759         .release =      netlink_release,
2760         .bind =         netlink_bind,
2761         .connect =      netlink_connect,
2762         .socketpair =   sock_no_socketpair,
2763         .accept =       sock_no_accept,
2764         .getname =      netlink_getname,
2765         .poll =         datagram_poll,
2766         .ioctl =        netlink_ioctl,
2767         .listen =       sock_no_listen,
2768         .shutdown =     sock_no_shutdown,
2769         .setsockopt =   netlink_setsockopt,
2770         .getsockopt =   netlink_getsockopt,
2771         .sendmsg =      netlink_sendmsg,
2772         .recvmsg =      netlink_recvmsg,
2773         .mmap =         sock_no_mmap,
2774         .sendpage =     sock_no_sendpage,
2775 };
2776 
2777 static const struct net_proto_family netlink_family_ops = {
2778         .family = PF_NETLINK,
2779         .create = netlink_create,
2780         .owner  = THIS_MODULE,  /* for consistency 8) */
2781 };
2782 
2783 static int __net_init netlink_net_init(struct net *net)
2784 {
2785 #ifdef CONFIG_PROC_FS
2786         if (!proc_create_net("netlink", 0, net->proc_net, &netlink_seq_ops,
2787                         sizeof(struct nl_seq_iter)))
2788                 return -ENOMEM;
2789 #endif
2790         return 0;
2791 }
2792 
2793 static void __net_exit netlink_net_exit(struct net *net)
2794 {
2795 #ifdef CONFIG_PROC_FS
2796         remove_proc_entry("netlink", net->proc_net);
2797 #endif
2798 }
2799 
2800 static void __init netlink_add_usersock_entry(void)
2801 {
2802         struct listeners *listeners;
2803         int groups = 32;
2804 
2805         listeners = kzalloc(sizeof(*listeners) + NLGRPSZ(groups), GFP_KERNEL);
2806         if (!listeners)
2807                 panic("netlink_add_usersock_entry: Cannot allocate listeners\n");
2808 
2809         netlink_table_grab();
2810 
2811         nl_table[NETLINK_USERSOCK].groups = groups;
2812         rcu_assign_pointer(nl_table[NETLINK_USERSOCK].listeners, listeners);
2813         nl_table[NETLINK_USERSOCK].module = THIS_MODULE;
2814         nl_table[NETLINK_USERSOCK].registered = 1;
2815         nl_table[NETLINK_USERSOCK].flags = NL_CFG_F_NONROOT_SEND;
2816 
2817         netlink_table_ungrab();
2818 }
2819 
2820 static struct pernet_operations __net_initdata netlink_net_ops = {
2821         .init = netlink_net_init,
2822         .exit = netlink_net_exit,
2823 };
2824 
2825 static inline u32 netlink_hash(const void *data, u32 len, u32 seed)
2826 {
2827         const struct netlink_sock *nlk = data;
2828         struct netlink_compare_arg arg;
2829 
2830         netlink_compare_arg_init(&arg, sock_net(&nlk->sk), nlk->portid);
2831         return jhash2((u32 *)&arg, netlink_compare_arg_len / sizeof(u32), seed);
2832 }
2833 
2834 static const struct rhashtable_params netlink_rhashtable_params = {
2835         .head_offset = offsetof(struct netlink_sock, node),
2836         .key_len = netlink_compare_arg_len,
2837         .obj_hashfn = netlink_hash,
2838         .obj_cmpfn = netlink_compare,
2839         .automatic_shrinking = true,
2840 };
2841 
2842 #if defined(CONFIG_BPF_SYSCALL) && defined(CONFIG_PROC_FS)
2843 BTF_ID_LIST(btf_netlink_sock_id)
2844 BTF_ID(struct, netlink_sock)
2845 
2846 static const struct bpf_iter_seq_info netlink_seq_info = {
2847         .seq_ops                = &netlink_seq_ops,
2848         .init_seq_private       = bpf_iter_init_seq_net,
2849         .fini_seq_private       = bpf_iter_fini_seq_net,
2850         .seq_priv_size          = sizeof(struct nl_seq_iter),
2851 };
2852 
2853 static struct bpf_iter_reg netlink_reg_info = {
2854         .target                 = "netlink",
2855         .ctx_arg_info_size      = 1,
2856         .ctx_arg_info           = {
2857                 { offsetof(struct bpf_iter__netlink, sk),
2858                   PTR_TO_BTF_ID_OR_NULL },
2859         },
2860         .seq_info               = &netlink_seq_info,
2861 };
2862 
2863 static int __init bpf_iter_register(void)
2864 {
2865         netlink_reg_info.ctx_arg_info[0].btf_id = *btf_netlink_sock_id;
2866         return bpf_iter_reg_target(&netlink_reg_info);
2867 }
2868 #endif
2869 
2870 static int __init netlink_proto_init(void)
2871 {
2872         int i;
2873         int err = proto_register(&netlink_proto, 0);
2874 
2875         if (err != 0)
2876                 goto out;
2877 
2878 #if defined(CONFIG_BPF_SYSCALL) && defined(CONFIG_PROC_FS)
2879         err = bpf_iter_register();
2880         if (err)
2881                 goto out;
2882 #endif
2883 
2884         BUILD_BUG_ON(sizeof(struct netlink_skb_parms) > sizeof_field(struct sk_buff, cb));
2885 
2886         nl_table = kcalloc(MAX_LINKS, sizeof(*nl_table), GFP_KERNEL);
2887         if (!nl_table)
2888                 goto panic;
2889 
2890         for (i = 0; i < MAX_LINKS; i++) {
2891                 if (rhashtable_init(&nl_table[i].hash,
2892                                     &netlink_rhashtable_params) < 0) {
2893                         while (--i > 0)
2894                                 rhashtable_destroy(&nl_table[i].hash);
2895                         kfree(nl_table);
2896                         goto panic;
2897                 }
2898         }
2899 
2900         netlink_add_usersock_entry();
2901 
2902         sock_register(&netlink_family_ops);
2903         register_pernet_subsys(&netlink_net_ops);
2904         register_pernet_subsys(&netlink_tap_net_ops);
2905         /* The netlink device handler may be needed early. */
2906         rtnetlink_init();
2907 out:
2908         return err;
2909 panic:
2910         panic("netlink_init: Cannot allocate nl_table\n");
2911 }
2912 
2913 core_initcall(netlink_proto_init);
2914 

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

kernel.org | git.kernel.org | LWN.net | Project Home | Wiki (Japanese) | Wiki (English) | SVN repository | Mail admin

Linux® is a registered trademark of Linus Torvalds in the United States and other countries.
TOMOYO® is a registered trademark of NTT DATA CORPORATION.

osdn.jp