Commit | Line | Data |
---|---|---|
1da177e4 LT |
1 | /* |
2 | * NETLINK Kernel-user communication protocol. | |
3 | * | |
113aa838 | 4 | * Authors: Alan Cox <alan@lxorguk.ukuu.org.uk> |
1da177e4 | 5 | * Alexey Kuznetsov <kuznet@ms2.inr.ac.ru> |
cd1df525 | 6 | * Patrick McHardy <kaber@trash.net> |
1da177e4 LT |
7 | * |
8 | * This program is free software; you can redistribute it and/or | |
9 | * modify it under the terms of the GNU General Public License | |
10 | * as published by the Free Software Foundation; either version | |
11 | * 2 of the License, or (at your option) any later version. | |
746fac4d | 12 | * |
1da177e4 LT |
13 | * Tue Jun 26 14:36:48 MEST 2001 Herbert "herp" Rosmanith |
14 | * added netlink_proto_exit | |
15 | * Tue Jan 22 18:32:44 BRST 2002 Arnaldo C. de Melo <acme@conectiva.com.br> | |
16 | * use nlk_sk, as sk->protinfo is on a diet 8) | |
4fdb3bb7 HW |
17 | * Fri Jul 22 19:51:12 MEST 2005 Harald Welte <laforge@gnumonks.org> |
18 | * - inc module use count of module that owns | |
19 | * the kernel socket in case userspace opens | |
20 | * socket of same protocol | |
21 | * - remove all module support, since netlink is | |
22 | * mandatory if CONFIG_NET=y these days | |
1da177e4 LT |
23 | */ |
24 | ||
1da177e4 LT |
25 | #include <linux/module.h> |
26 | ||
4fc268d2 | 27 | #include <linux/capability.h> |
1da177e4 LT |
28 | #include <linux/kernel.h> |
29 | #include <linux/init.h> | |
1da177e4 LT |
30 | #include <linux/signal.h> |
31 | #include <linux/sched.h> | |
32 | #include <linux/errno.h> | |
33 | #include <linux/string.h> | |
34 | #include <linux/stat.h> | |
35 | #include <linux/socket.h> | |
36 | #include <linux/un.h> | |
37 | #include <linux/fcntl.h> | |
38 | #include <linux/termios.h> | |
39 | #include <linux/sockios.h> | |
40 | #include <linux/net.h> | |
41 | #include <linux/fs.h> | |
42 | #include <linux/slab.h> | |
7c0f6ba6 | 43 | #include <linux/uaccess.h> |
1da177e4 LT |
44 | #include <linux/skbuff.h> |
45 | #include <linux/netdevice.h> | |
46 | #include <linux/rtnetlink.h> | |
47 | #include <linux/proc_fs.h> | |
48 | #include <linux/seq_file.h> | |
1da177e4 LT |
49 | #include <linux/notifier.h> |
50 | #include <linux/security.h> | |
51 | #include <linux/jhash.h> | |
52 | #include <linux/jiffies.h> | |
53 | #include <linux/random.h> | |
54 | #include <linux/bitops.h> | |
55 | #include <linux/mm.h> | |
56 | #include <linux/types.h> | |
54e0f520 | 57 | #include <linux/audit.h> |
af65bdfc | 58 | #include <linux/mutex.h> |
ccdfcc39 | 59 | #include <linux/vmalloc.h> |
bcbde0d4 | 60 | #include <linux/if_arp.h> |
e341694e | 61 | #include <linux/rhashtable.h> |
9652e931 | 62 | #include <asm/cacheflush.h> |
e341694e | 63 | #include <linux/hash.h> |
ee1c2442 | 64 | #include <linux/genetlink.h> |
7212462f | 65 | #include <linux/net_namespace.h> |
54e0f520 | 66 | |
457c4cbc | 67 | #include <net/net_namespace.h> |
1da177e4 LT |
68 | #include <net/sock.h> |
69 | #include <net/scm.h> | |
82ace47a | 70 | #include <net/netlink.h> |
1da177e4 | 71 | |
0f29c768 | 72 | #include "af_netlink.h" |
1da177e4 | 73 | |
5c398dc8 ED |
74 | struct listeners { |
75 | struct rcu_head rcu; | |
76 | unsigned long masks[0]; | |
6c04bb18 JB |
77 | }; |
78 | ||
cd967e05 | 79 | /* state bits */ |
cc3a572f | 80 | #define NETLINK_S_CONGESTED 0x0 |
cd967e05 | 81 | |
035c4c16 | 82 | static inline int netlink_is_kernel(struct sock *sk) |
aed81560 | 83 | { |
cc3a572f | 84 | return nlk_sk(sk)->flags & NETLINK_F_KERNEL_SOCKET; |
aed81560 DL |
85 | } |
86 | ||
91dd93f9 | 87 | struct netlink_table *nl_table __read_mostly; |
0f29c768 | 88 | EXPORT_SYMBOL_GPL(nl_table); |
1da177e4 LT |
89 | |
90 | static DECLARE_WAIT_QUEUE_HEAD(nl_table_wait); | |
91 | ||
8a0f5ccf HX |
92 | static struct lock_class_key nlk_cb_mutex_keys[MAX_LINKS]; |
93 | ||
94 | static const char *const nlk_cb_mutex_key_strings[MAX_LINKS + 1] = { | |
95 | "nlk_cb_mutex-ROUTE", | |
96 | "nlk_cb_mutex-1", | |
97 | "nlk_cb_mutex-USERSOCK", | |
98 | "nlk_cb_mutex-FIREWALL", | |
99 | "nlk_cb_mutex-SOCK_DIAG", | |
100 | "nlk_cb_mutex-NFLOG", | |
101 | "nlk_cb_mutex-XFRM", | |
102 | "nlk_cb_mutex-SELINUX", | |
103 | "nlk_cb_mutex-ISCSI", | |
104 | "nlk_cb_mutex-AUDIT", | |
105 | "nlk_cb_mutex-FIB_LOOKUP", | |
106 | "nlk_cb_mutex-CONNECTOR", | |
107 | "nlk_cb_mutex-NETFILTER", | |
108 | "nlk_cb_mutex-IP6_FW", | |
109 | "nlk_cb_mutex-DNRTMSG", | |
110 | "nlk_cb_mutex-KOBJECT_UEVENT", | |
111 | "nlk_cb_mutex-GENERIC", | |
112 | "nlk_cb_mutex-17", | |
113 | "nlk_cb_mutex-SCSITRANSPORT", | |
114 | "nlk_cb_mutex-ECRYPTFS", | |
115 | "nlk_cb_mutex-RDMA", | |
116 | "nlk_cb_mutex-CRYPTO", | |
117 | "nlk_cb_mutex-SMC", | |
118 | "nlk_cb_mutex-23", | |
119 | "nlk_cb_mutex-24", | |
120 | "nlk_cb_mutex-25", | |
121 | "nlk_cb_mutex-26", | |
122 | "nlk_cb_mutex-27", | |
123 | "nlk_cb_mutex-28", | |
124 | "nlk_cb_mutex-29", | |
125 | "nlk_cb_mutex-30", | |
126 | "nlk_cb_mutex-31", | |
127 | "nlk_cb_mutex-MAX_LINKS" | |
128 | }; | |
129 | ||
1da177e4 | 130 | static int netlink_dump(struct sock *sk); |
9652e931 | 131 | static void netlink_skb_destructor(struct sk_buff *skb); |
1da177e4 | 132 | |
78fd1d0a | 133 | /* nl_table locking explained: |
21e4902a | 134 | * Lookup and traversal are protected with an RCU read-side lock. Insertion |
c5adde94 | 135 | * and removal are protected with per bucket lock while using RCU list |
21e4902a TG |
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. | |
78fd1d0a | 140 | */ |
0f29c768 AV |
141 | DEFINE_RWLOCK(nl_table_lock); |
142 | EXPORT_SYMBOL_GPL(nl_table_lock); | |
1da177e4 LT |
143 | static atomic_t nl_table_users = ATOMIC_INIT(0); |
144 | ||
6d772ac5 ED |
145 | #define nl_deref_protected(X) rcu_dereference_protected(X, lockdep_is_held(&nl_table_lock)); |
146 | ||
efa172f4 | 147 | static BLOCKING_NOTIFIER_HEAD(netlink_chain); |
1da177e4 | 148 | |
bcbde0d4 DB |
149 | static DEFINE_SPINLOCK(netlink_tap_lock); |
150 | static struct list_head netlink_tap_all __read_mostly; | |
151 | ||
c428ecd1 HX |
152 | static const struct rhashtable_params netlink_rhashtable_params; |
153 | ||
b57ef81f | 154 | static inline u32 netlink_group_mask(u32 group) |
d629b836 PM |
155 | { |
156 | return group ? 1 << (group - 1) : 0; | |
157 | } | |
158 | ||
1853c949 DB |
159 | static struct sk_buff *netlink_to_full_skb(const struct sk_buff *skb, |
160 | gfp_t gfp_mask) | |
161 | { | |
162 | unsigned int len = skb_end_offset(skb); | |
163 | struct sk_buff *new; | |
164 | ||
165 | new = alloc_skb(len, gfp_mask); | |
166 | if (new == NULL) | |
167 | return NULL; | |
168 | ||
169 | NETLINK_CB(new).portid = NETLINK_CB(skb).portid; | |
170 | NETLINK_CB(new).dst_group = NETLINK_CB(skb).dst_group; | |
171 | NETLINK_CB(new).creds = NETLINK_CB(skb).creds; | |
172 | ||
59ae1d12 | 173 | skb_put_data(new, skb->data, len); |
1853c949 DB |
174 | return new; |
175 | } | |
176 | ||
bcbde0d4 DB |
177 | int netlink_add_tap(struct netlink_tap *nt) |
178 | { | |
179 | if (unlikely(nt->dev->type != ARPHRD_NETLINK)) | |
180 | return -EINVAL; | |
181 | ||
182 | spin_lock(&netlink_tap_lock); | |
183 | list_add_rcu(&nt->list, &netlink_tap_all); | |
184 | spin_unlock(&netlink_tap_lock); | |
185 | ||
fcd4d35e | 186 | __module_get(nt->module); |
bcbde0d4 DB |
187 | |
188 | return 0; | |
189 | } | |
190 | EXPORT_SYMBOL_GPL(netlink_add_tap); | |
191 | ||
2173f8d9 | 192 | static int __netlink_remove_tap(struct netlink_tap *nt) |
bcbde0d4 DB |
193 | { |
194 | bool found = false; | |
195 | struct netlink_tap *tmp; | |
196 | ||
197 | spin_lock(&netlink_tap_lock); | |
198 | ||
199 | list_for_each_entry(tmp, &netlink_tap_all, list) { | |
200 | if (nt == tmp) { | |
201 | list_del_rcu(&nt->list); | |
202 | found = true; | |
203 | goto out; | |
204 | } | |
205 | } | |
206 | ||
207 | pr_warn("__netlink_remove_tap: %p not found\n", nt); | |
208 | out: | |
209 | spin_unlock(&netlink_tap_lock); | |
210 | ||
92b80eb3 | 211 | if (found) |
bcbde0d4 DB |
212 | module_put(nt->module); |
213 | ||
214 | return found ? 0 : -ENODEV; | |
215 | } | |
bcbde0d4 DB |
216 | |
217 | int netlink_remove_tap(struct netlink_tap *nt) | |
218 | { | |
219 | int ret; | |
220 | ||
221 | ret = __netlink_remove_tap(nt); | |
222 | synchronize_net(); | |
223 | ||
224 | return ret; | |
225 | } | |
226 | EXPORT_SYMBOL_GPL(netlink_remove_tap); | |
227 | ||
5ffd5cdd DB |
228 | static bool netlink_filter_tap(const struct sk_buff *skb) |
229 | { | |
230 | struct sock *sk = skb->sk; | |
5ffd5cdd DB |
231 | |
232 | /* We take the more conservative approach and | |
233 | * whitelist socket protocols that may pass. | |
234 | */ | |
235 | switch (sk->sk_protocol) { | |
236 | case NETLINK_ROUTE: | |
237 | case NETLINK_USERSOCK: | |
238 | case NETLINK_SOCK_DIAG: | |
239 | case NETLINK_NFLOG: | |
240 | case NETLINK_XFRM: | |
241 | case NETLINK_FIB_LOOKUP: | |
242 | case NETLINK_NETFILTER: | |
243 | case NETLINK_GENERIC: | |
498044bb | 244 | return true; |
5ffd5cdd DB |
245 | } |
246 | ||
498044bb | 247 | return false; |
5ffd5cdd DB |
248 | } |
249 | ||
bcbde0d4 DB |
250 | static int __netlink_deliver_tap_skb(struct sk_buff *skb, |
251 | struct net_device *dev) | |
252 | { | |
253 | struct sk_buff *nskb; | |
5ffd5cdd | 254 | struct sock *sk = skb->sk; |
bcbde0d4 DB |
255 | int ret = -ENOMEM; |
256 | ||
257 | dev_hold(dev); | |
1853c949 | 258 | |
d1b4c689 | 259 | if (is_vmalloc_addr(skb->head)) |
1853c949 DB |
260 | nskb = netlink_to_full_skb(skb, GFP_ATOMIC); |
261 | else | |
262 | nskb = skb_clone(skb, GFP_ATOMIC); | |
bcbde0d4 DB |
263 | if (nskb) { |
264 | nskb->dev = dev; | |
5ffd5cdd | 265 | nskb->protocol = htons((u16) sk->sk_protocol); |
604d13c9 DB |
266 | nskb->pkt_type = netlink_is_kernel(sk) ? |
267 | PACKET_KERNEL : PACKET_USER; | |
4e48ed88 | 268 | skb_reset_network_header(nskb); |
bcbde0d4 DB |
269 | ret = dev_queue_xmit(nskb); |
270 | if (unlikely(ret > 0)) | |
271 | ret = net_xmit_errno(ret); | |
272 | } | |
273 | ||
274 | dev_put(dev); | |
275 | return ret; | |
276 | } | |
277 | ||
278 | static void __netlink_deliver_tap(struct sk_buff *skb) | |
279 | { | |
280 | int ret; | |
281 | struct netlink_tap *tmp; | |
282 | ||
5ffd5cdd DB |
283 | if (!netlink_filter_tap(skb)) |
284 | return; | |
285 | ||
bcbde0d4 DB |
286 | list_for_each_entry_rcu(tmp, &netlink_tap_all, list) { |
287 | ret = __netlink_deliver_tap_skb(skb, tmp->dev); | |
288 | if (unlikely(ret)) | |
289 | break; | |
290 | } | |
291 | } | |
292 | ||
293 | static void netlink_deliver_tap(struct sk_buff *skb) | |
294 | { | |
295 | rcu_read_lock(); | |
296 | ||
297 | if (unlikely(!list_empty(&netlink_tap_all))) | |
298 | __netlink_deliver_tap(skb); | |
299 | ||
300 | rcu_read_unlock(); | |
301 | } | |
302 | ||
73bfd370 DB |
303 | static void netlink_deliver_tap_kernel(struct sock *dst, struct sock *src, |
304 | struct sk_buff *skb) | |
305 | { | |
306 | if (!(netlink_is_kernel(dst) && netlink_is_kernel(src))) | |
307 | netlink_deliver_tap(skb); | |
308 | } | |
309 | ||
cd1df525 PM |
310 | static void netlink_overrun(struct sock *sk) |
311 | { | |
312 | struct netlink_sock *nlk = nlk_sk(sk); | |
313 | ||
cc3a572f ND |
314 | if (!(nlk->flags & NETLINK_F_RECV_NO_ENOBUFS)) { |
315 | if (!test_and_set_bit(NETLINK_S_CONGESTED, | |
316 | &nlk_sk(sk)->state)) { | |
cd1df525 PM |
317 | sk->sk_err = ENOBUFS; |
318 | sk->sk_error_report(sk); | |
319 | } | |
320 | } | |
321 | atomic_inc(&sk->sk_drops); | |
322 | } | |
323 | ||
324 | static void netlink_rcv_wake(struct sock *sk) | |
325 | { | |
326 | struct netlink_sock *nlk = nlk_sk(sk); | |
327 | ||
328 | if (skb_queue_empty(&sk->sk_receive_queue)) | |
cc3a572f ND |
329 | clear_bit(NETLINK_S_CONGESTED, &nlk->state); |
330 | if (!test_bit(NETLINK_S_CONGESTED, &nlk->state)) | |
cd1df525 PM |
331 | wake_up_interruptible(&nlk->wait); |
332 | } | |
333 | ||
cf0a018a PM |
334 | static void netlink_skb_destructor(struct sk_buff *skb) |
335 | { | |
c05cdb1b | 336 | if (is_vmalloc_addr(skb->head)) { |
3a36515f PN |
337 | if (!skb->cloned || |
338 | !atomic_dec_return(&(skb_shinfo(skb)->dataref))) | |
339 | vfree(skb->head); | |
340 | ||
c05cdb1b PNA |
341 | skb->head = NULL; |
342 | } | |
9652e931 PM |
343 | if (skb->sk != NULL) |
344 | sock_rfree(skb); | |
cf0a018a PM |
345 | } |
346 | ||
347 | static void netlink_skb_set_owner_r(struct sk_buff *skb, struct sock *sk) | |
348 | { | |
349 | WARN_ON(skb->sk != NULL); | |
350 | skb->sk = sk; | |
351 | skb->destructor = netlink_skb_destructor; | |
352 | atomic_add(skb->truesize, &sk->sk_rmem_alloc); | |
353 | sk_mem_charge(sk, skb->truesize); | |
354 | } | |
355 | ||
ed5d7788 | 356 | static void netlink_sock_destruct(struct sock *sk) |
1da177e4 | 357 | { |
3f660d66 HX |
358 | struct netlink_sock *nlk = nlk_sk(sk); |
359 | ||
16b304f3 | 360 | if (nlk->cb_running) { |
ed5d7788 HX |
361 | if (nlk->cb.done) |
362 | nlk->cb.done(&nlk->cb); | |
16b304f3 PS |
363 | module_put(nlk->cb.module); |
364 | kfree_skb(nlk->cb.skb); | |
3f660d66 HX |
365 | } |
366 | ||
1da177e4 LT |
367 | skb_queue_purge(&sk->sk_receive_queue); |
368 | ||
369 | if (!sock_flag(sk, SOCK_DEAD)) { | |
6ac552fd | 370 | printk(KERN_ERR "Freeing alive netlink socket %p\n", sk); |
1da177e4 LT |
371 | return; |
372 | } | |
547b792c IJ |
373 | |
374 | WARN_ON(atomic_read(&sk->sk_rmem_alloc)); | |
14afee4b | 375 | WARN_ON(refcount_read(&sk->sk_wmem_alloc)); |
547b792c | 376 | WARN_ON(nlk_sk(sk)->groups); |
1da177e4 LT |
377 | } |
378 | ||
707693c8 HX |
379 | static void netlink_sock_destruct_work(struct work_struct *work) |
380 | { | |
381 | struct netlink_sock *nlk = container_of(work, struct netlink_sock, | |
382 | work); | |
383 | ||
ed5d7788 | 384 | sk_free(&nlk->sk); |
707693c8 HX |
385 | } |
386 | ||
6ac552fd PM |
387 | /* This lock without WQ_FLAG_EXCLUSIVE is good on UP and it is _very_ bad on |
388 | * SMP. Look, when several writers sleep and reader wakes them up, all but one | |
1da177e4 LT |
389 | * immediately hit write lock and grab all the cpus. Exclusive sleep solves |
390 | * this, _but_ remember, it adds useless work on UP machines. | |
391 | */ | |
392 | ||
d136f1bd | 393 | void netlink_table_grab(void) |
9a429c49 | 394 | __acquires(nl_table_lock) |
1da177e4 | 395 | { |
d136f1bd JB |
396 | might_sleep(); |
397 | ||
6abd219c | 398 | write_lock_irq(&nl_table_lock); |
1da177e4 LT |
399 | |
400 | if (atomic_read(&nl_table_users)) { | |
401 | DECLARE_WAITQUEUE(wait, current); | |
402 | ||
403 | add_wait_queue_exclusive(&nl_table_wait, &wait); | |
6ac552fd | 404 | for (;;) { |
1da177e4 LT |
405 | set_current_state(TASK_UNINTERRUPTIBLE); |
406 | if (atomic_read(&nl_table_users) == 0) | |
407 | break; | |
6abd219c | 408 | write_unlock_irq(&nl_table_lock); |
1da177e4 | 409 | schedule(); |
6abd219c | 410 | write_lock_irq(&nl_table_lock); |
1da177e4 LT |
411 | } |
412 | ||
413 | __set_current_state(TASK_RUNNING); | |
414 | remove_wait_queue(&nl_table_wait, &wait); | |
415 | } | |
416 | } | |
417 | ||
d136f1bd | 418 | void netlink_table_ungrab(void) |
9a429c49 | 419 | __releases(nl_table_lock) |
1da177e4 | 420 | { |
6abd219c | 421 | write_unlock_irq(&nl_table_lock); |
1da177e4 LT |
422 | wake_up(&nl_table_wait); |
423 | } | |
424 | ||
6ac552fd | 425 | static inline void |
1da177e4 LT |
426 | netlink_lock_table(void) |
427 | { | |
428 | /* read_lock() synchronizes us to netlink_table_grab */ | |
429 | ||
430 | read_lock(&nl_table_lock); | |
431 | atomic_inc(&nl_table_users); | |
432 | read_unlock(&nl_table_lock); | |
433 | } | |
434 | ||
6ac552fd | 435 | static inline void |
1da177e4 LT |
436 | netlink_unlock_table(void) |
437 | { | |
438 | if (atomic_dec_and_test(&nl_table_users)) | |
439 | wake_up(&nl_table_wait); | |
440 | } | |
441 | ||
e341694e | 442 | struct netlink_compare_arg |
1da177e4 | 443 | { |
c428ecd1 | 444 | possible_net_t pnet; |
e341694e TG |
445 | u32 portid; |
446 | }; | |
1da177e4 | 447 | |
8f2ddaac HX |
448 | /* Doing sizeof directly may yield 4 extra bytes on 64-bit. */ |
449 | #define netlink_compare_arg_len \ | |
450 | (offsetof(struct netlink_compare_arg, portid) + sizeof(u32)) | |
c428ecd1 HX |
451 | |
452 | static inline int netlink_compare(struct rhashtable_compare_arg *arg, | |
453 | const void *ptr) | |
1da177e4 | 454 | { |
c428ecd1 HX |
455 | const struct netlink_compare_arg *x = arg->key; |
456 | const struct netlink_sock *nlk = ptr; | |
1da177e4 | 457 | |
da314c99 | 458 | return nlk->portid != x->portid || |
c428ecd1 HX |
459 | !net_eq(sock_net(&nlk->sk), read_pnet(&x->pnet)); |
460 | } | |
461 | ||
462 | static void netlink_compare_arg_init(struct netlink_compare_arg *arg, | |
463 | struct net *net, u32 portid) | |
464 | { | |
465 | memset(arg, 0, sizeof(*arg)); | |
466 | write_pnet(&arg->pnet, net); | |
467 | arg->portid = portid; | |
1da177e4 LT |
468 | } |
469 | ||
e341694e TG |
470 | static struct sock *__netlink_lookup(struct netlink_table *table, u32 portid, |
471 | struct net *net) | |
1da177e4 | 472 | { |
c428ecd1 | 473 | struct netlink_compare_arg arg; |
1da177e4 | 474 | |
c428ecd1 HX |
475 | netlink_compare_arg_init(&arg, net, portid); |
476 | return rhashtable_lookup_fast(&table->hash, &arg, | |
477 | netlink_rhashtable_params); | |
1da177e4 LT |
478 | } |
479 | ||
c428ecd1 | 480 | static int __netlink_insert(struct netlink_table *table, struct sock *sk) |
c5adde94 | 481 | { |
c428ecd1 | 482 | struct netlink_compare_arg arg; |
c5adde94 | 483 | |
da314c99 | 484 | netlink_compare_arg_init(&arg, sock_net(sk), nlk_sk(sk)->portid); |
c428ecd1 HX |
485 | return rhashtable_lookup_insert_key(&table->hash, &arg, |
486 | &nlk_sk(sk)->node, | |
487 | netlink_rhashtable_params); | |
c5adde94 YX |
488 | } |
489 | ||
e341694e | 490 | static struct sock *netlink_lookup(struct net *net, int protocol, u32 portid) |
1da177e4 | 491 | { |
e341694e TG |
492 | struct netlink_table *table = &nl_table[protocol]; |
493 | struct sock *sk; | |
1da177e4 | 494 | |
e341694e TG |
495 | rcu_read_lock(); |
496 | sk = __netlink_lookup(table, portid, net); | |
497 | if (sk) | |
498 | sock_hold(sk); | |
499 | rcu_read_unlock(); | |
1da177e4 | 500 | |
e341694e | 501 | return sk; |
1da177e4 LT |
502 | } |
503 | ||
90ddc4f0 | 504 | static const struct proto_ops netlink_ops; |
1da177e4 | 505 | |
4277a083 PM |
506 | static void |
507 | netlink_update_listeners(struct sock *sk) | |
508 | { | |
509 | struct netlink_table *tbl = &nl_table[sk->sk_protocol]; | |
4277a083 PM |
510 | unsigned long mask; |
511 | unsigned int i; | |
6d772ac5 ED |
512 | struct listeners *listeners; |
513 | ||
514 | listeners = nl_deref_protected(tbl->listeners); | |
515 | if (!listeners) | |
516 | return; | |
4277a083 | 517 | |
b4ff4f04 | 518 | for (i = 0; i < NLGRPLONGS(tbl->groups); i++) { |
4277a083 | 519 | mask = 0; |
b67bfe0d | 520 | sk_for_each_bound(sk, &tbl->mc_list) { |
b4ff4f04 JB |
521 | if (i < NLGRPLONGS(nlk_sk(sk)->ngroups)) |
522 | mask |= nlk_sk(sk)->groups[i]; | |
523 | } | |
6d772ac5 | 524 | listeners->masks[i] = mask; |
4277a083 PM |
525 | } |
526 | /* this function is only called with the netlink table "grabbed", which | |
527 | * makes sure updates are visible before bind or setsockopt return. */ | |
528 | } | |
529 | ||
8ea65f4a | 530 | static int netlink_insert(struct sock *sk, u32 portid) |
1da177e4 | 531 | { |
da12c90e | 532 | struct netlink_table *table = &nl_table[sk->sk_protocol]; |
919d9db9 | 533 | int err; |
1da177e4 | 534 | |
c5adde94 | 535 | lock_sock(sk); |
1da177e4 | 536 | |
da314c99 HX |
537 | err = nlk_sk(sk)->portid == portid ? 0 : -EBUSY; |
538 | if (nlk_sk(sk)->bound) | |
1da177e4 LT |
539 | goto err; |
540 | ||
541 | err = -ENOMEM; | |
97defe1e TG |
542 | if (BITS_PER_LONG > 32 && |
543 | unlikely(atomic_read(&table->hash.nelems) >= UINT_MAX)) | |
1da177e4 LT |
544 | goto err; |
545 | ||
da314c99 | 546 | nlk_sk(sk)->portid = portid; |
e341694e | 547 | sock_hold(sk); |
919d9db9 | 548 | |
c428ecd1 HX |
549 | err = __netlink_insert(table, sk); |
550 | if (err) { | |
4e7c1330 DB |
551 | /* In case the hashtable backend returns with -EBUSY |
552 | * from here, it must not escape to the caller. | |
553 | */ | |
554 | if (unlikely(err == -EBUSY)) | |
555 | err = -EOVERFLOW; | |
c428ecd1 HX |
556 | if (err == -EEXIST) |
557 | err = -EADDRINUSE; | |
c5adde94 | 558 | sock_put(sk); |
1f770c0a | 559 | goto err; |
919d9db9 HX |
560 | } |
561 | ||
da314c99 HX |
562 | /* We need to ensure that the socket is hashed and visible. */ |
563 | smp_wmb(); | |
564 | nlk_sk(sk)->bound = portid; | |
1f770c0a | 565 | |
1da177e4 | 566 | err: |
c5adde94 | 567 | release_sock(sk); |
1da177e4 LT |
568 | return err; |
569 | } | |
570 | ||
571 | static void netlink_remove(struct sock *sk) | |
572 | { | |
e341694e TG |
573 | struct netlink_table *table; |
574 | ||
e341694e | 575 | table = &nl_table[sk->sk_protocol]; |
c428ecd1 HX |
576 | if (!rhashtable_remove_fast(&table->hash, &nlk_sk(sk)->node, |
577 | netlink_rhashtable_params)) { | |
41c6d650 | 578 | WARN_ON(refcount_read(&sk->sk_refcnt) == 1); |
e341694e TG |
579 | __sock_put(sk); |
580 | } | |
e341694e | 581 | |
1da177e4 | 582 | netlink_table_grab(); |
b10dcb3b | 583 | if (nlk_sk(sk)->subscriptions) { |
1da177e4 | 584 | __sk_del_bind_node(sk); |
b10dcb3b JB |
585 | netlink_update_listeners(sk); |
586 | } | |
ee1c2442 JB |
587 | if (sk->sk_protocol == NETLINK_GENERIC) |
588 | atomic_inc(&genl_sk_destructing_cnt); | |
1da177e4 LT |
589 | netlink_table_ungrab(); |
590 | } | |
591 | ||
592 | static struct proto netlink_proto = { | |
593 | .name = "NETLINK", | |
594 | .owner = THIS_MODULE, | |
595 | .obj_size = sizeof(struct netlink_sock), | |
596 | }; | |
597 | ||
1b8d7ae4 | 598 | static int __netlink_create(struct net *net, struct socket *sock, |
11aa9c28 EB |
599 | struct mutex *cb_mutex, int protocol, |
600 | int kern) | |
1da177e4 LT |
601 | { |
602 | struct sock *sk; | |
603 | struct netlink_sock *nlk; | |
ab33a171 PM |
604 | |
605 | sock->ops = &netlink_ops; | |
606 | ||
11aa9c28 | 607 | sk = sk_alloc(net, PF_NETLINK, GFP_KERNEL, &netlink_proto, kern); |
ab33a171 PM |
608 | if (!sk) |
609 | return -ENOMEM; | |
610 | ||
611 | sock_init_data(sock, sk); | |
612 | ||
613 | nlk = nlk_sk(sk); | |
658cb354 | 614 | if (cb_mutex) { |
ffa4d721 | 615 | nlk->cb_mutex = cb_mutex; |
658cb354 | 616 | } else { |
ffa4d721 PM |
617 | nlk->cb_mutex = &nlk->cb_def_mutex; |
618 | mutex_init(nlk->cb_mutex); | |
8a0f5ccf HX |
619 | lockdep_set_class_and_name(nlk->cb_mutex, |
620 | nlk_cb_mutex_keys + protocol, | |
621 | nlk_cb_mutex_key_strings[protocol]); | |
ffa4d721 | 622 | } |
ab33a171 PM |
623 | init_waitqueue_head(&nlk->wait); |
624 | ||
625 | sk->sk_destruct = netlink_sock_destruct; | |
626 | sk->sk_protocol = protocol; | |
627 | return 0; | |
628 | } | |
629 | ||
3f378b68 EP |
630 | static int netlink_create(struct net *net, struct socket *sock, int protocol, |
631 | int kern) | |
ab33a171 PM |
632 | { |
633 | struct module *module = NULL; | |
af65bdfc | 634 | struct mutex *cb_mutex; |
f7fa9b10 | 635 | struct netlink_sock *nlk; |
023e2cfa JB |
636 | int (*bind)(struct net *net, int group); |
637 | void (*unbind)(struct net *net, int group); | |
ab33a171 | 638 | int err = 0; |
1da177e4 LT |
639 | |
640 | sock->state = SS_UNCONNECTED; | |
641 | ||
642 | if (sock->type != SOCK_RAW && sock->type != SOCK_DGRAM) | |
643 | return -ESOCKTNOSUPPORT; | |
644 | ||
6ac552fd | 645 | if (protocol < 0 || protocol >= MAX_LINKS) |
1da177e4 LT |
646 | return -EPROTONOSUPPORT; |
647 | ||
77247bbb | 648 | netlink_lock_table(); |
95a5afca | 649 | #ifdef CONFIG_MODULES |
ab33a171 | 650 | if (!nl_table[protocol].registered) { |
77247bbb | 651 | netlink_unlock_table(); |
4fdb3bb7 | 652 | request_module("net-pf-%d-proto-%d", PF_NETLINK, protocol); |
77247bbb | 653 | netlink_lock_table(); |
4fdb3bb7 | 654 | } |
ab33a171 PM |
655 | #endif |
656 | if (nl_table[protocol].registered && | |
657 | try_module_get(nl_table[protocol].module)) | |
658 | module = nl_table[protocol].module; | |
974c37e9 AD |
659 | else |
660 | err = -EPROTONOSUPPORT; | |
af65bdfc | 661 | cb_mutex = nl_table[protocol].cb_mutex; |
03292745 | 662 | bind = nl_table[protocol].bind; |
4f520900 | 663 | unbind = nl_table[protocol].unbind; |
77247bbb | 664 | netlink_unlock_table(); |
4fdb3bb7 | 665 | |
974c37e9 AD |
666 | if (err < 0) |
667 | goto out; | |
668 | ||
11aa9c28 | 669 | err = __netlink_create(net, sock, cb_mutex, protocol, kern); |
6ac552fd | 670 | if (err < 0) |
f7fa9b10 PM |
671 | goto out_module; |
672 | ||
6f756a8c | 673 | local_bh_disable(); |
c1fd3b94 | 674 | sock_prot_inuse_add(net, &netlink_proto, 1); |
6f756a8c DM |
675 | local_bh_enable(); |
676 | ||
f7fa9b10 | 677 | nlk = nlk_sk(sock->sk); |
f7fa9b10 | 678 | nlk->module = module; |
03292745 | 679 | nlk->netlink_bind = bind; |
4f520900 | 680 | nlk->netlink_unbind = unbind; |
ab33a171 PM |
681 | out: |
682 | return err; | |
1da177e4 | 683 | |
ab33a171 PM |
684 | out_module: |
685 | module_put(module); | |
686 | goto out; | |
1da177e4 LT |
687 | } |
688 | ||
21e4902a TG |
689 | static void deferred_put_nlk_sk(struct rcu_head *head) |
690 | { | |
691 | struct netlink_sock *nlk = container_of(head, struct netlink_sock, rcu); | |
ed5d7788 HX |
692 | struct sock *sk = &nlk->sk; |
693 | ||
be82485f XL |
694 | kfree(nlk->groups); |
695 | nlk->groups = NULL; | |
696 | ||
41c6d650 | 697 | if (!refcount_dec_and_test(&sk->sk_refcnt)) |
ed5d7788 HX |
698 | return; |
699 | ||
700 | if (nlk->cb_running && nlk->cb.done) { | |
701 | INIT_WORK(&nlk->work, netlink_sock_destruct_work); | |
702 | schedule_work(&nlk->work); | |
703 | return; | |
704 | } | |
21e4902a | 705 | |
ed5d7788 | 706 | sk_free(sk); |
21e4902a TG |
707 | } |
708 | ||
1da177e4 LT |
709 | static int netlink_release(struct socket *sock) |
710 | { | |
711 | struct sock *sk = sock->sk; | |
712 | struct netlink_sock *nlk; | |
713 | ||
714 | if (!sk) | |
715 | return 0; | |
716 | ||
717 | netlink_remove(sk); | |
ac57b3a9 | 718 | sock_orphan(sk); |
1da177e4 LT |
719 | nlk = nlk_sk(sk); |
720 | ||
3f660d66 HX |
721 | /* |
722 | * OK. Socket is unlinked, any packets that arrive now | |
723 | * will be purged. | |
724 | */ | |
1da177e4 | 725 | |
ee1c2442 JB |
726 | /* must not acquire netlink_table_lock in any way again before unbind |
727 | * and notifying genetlink is done as otherwise it might deadlock | |
728 | */ | |
729 | if (nlk->netlink_unbind) { | |
730 | int i; | |
731 | ||
732 | for (i = 0; i < nlk->ngroups; i++) | |
733 | if (test_bit(i, nlk->groups)) | |
734 | nlk->netlink_unbind(sock_net(sk), i + 1); | |
735 | } | |
736 | if (sk->sk_protocol == NETLINK_GENERIC && | |
737 | atomic_dec_return(&genl_sk_destructing_cnt) == 0) | |
738 | wake_up(&genl_sk_destructing_waitq); | |
739 | ||
1da177e4 LT |
740 | sock->sk = NULL; |
741 | wake_up_interruptible_all(&nlk->wait); | |
742 | ||
743 | skb_queue_purge(&sk->sk_write_queue); | |
744 | ||
e2726020 | 745 | if (nlk->portid && nlk->bound) { |
1da177e4 | 746 | struct netlink_notify n = { |
3b1e0a65 | 747 | .net = sock_net(sk), |
1da177e4 | 748 | .protocol = sk->sk_protocol, |
15e47304 | 749 | .portid = nlk->portid, |
1da177e4 | 750 | }; |
efa172f4 | 751 | blocking_notifier_call_chain(&netlink_chain, |
e041c683 | 752 | NETLINK_URELEASE, &n); |
746fac4d | 753 | } |
4fdb3bb7 | 754 | |
5e7c001c | 755 | module_put(nlk->module); |
4fdb3bb7 | 756 | |
aed81560 | 757 | if (netlink_is_kernel(sk)) { |
b10dcb3b | 758 | netlink_table_grab(); |
869e58f8 DL |
759 | BUG_ON(nl_table[sk->sk_protocol].registered == 0); |
760 | if (--nl_table[sk->sk_protocol].registered == 0) { | |
6d772ac5 ED |
761 | struct listeners *old; |
762 | ||
763 | old = nl_deref_protected(nl_table[sk->sk_protocol].listeners); | |
764 | RCU_INIT_POINTER(nl_table[sk->sk_protocol].listeners, NULL); | |
765 | kfree_rcu(old, rcu); | |
869e58f8 | 766 | nl_table[sk->sk_protocol].module = NULL; |
9785e10a | 767 | nl_table[sk->sk_protocol].bind = NULL; |
4f520900 | 768 | nl_table[sk->sk_protocol].unbind = NULL; |
9785e10a | 769 | nl_table[sk->sk_protocol].flags = 0; |
869e58f8 DL |
770 | nl_table[sk->sk_protocol].registered = 0; |
771 | } | |
b10dcb3b | 772 | netlink_table_ungrab(); |
658cb354 | 773 | } |
77247bbb | 774 | |
3755810c | 775 | local_bh_disable(); |
c1fd3b94 | 776 | sock_prot_inuse_add(sock_net(sk), &netlink_proto, -1); |
3755810c | 777 | local_bh_enable(); |
21e4902a | 778 | call_rcu(&nlk->rcu, deferred_put_nlk_sk); |
1da177e4 LT |
779 | return 0; |
780 | } | |
781 | ||
782 | static int netlink_autobind(struct socket *sock) | |
783 | { | |
784 | struct sock *sk = sock->sk; | |
3b1e0a65 | 785 | struct net *net = sock_net(sk); |
da12c90e | 786 | struct netlink_table *table = &nl_table[sk->sk_protocol]; |
15e47304 | 787 | s32 portid = task_tgid_vnr(current); |
1da177e4 | 788 | int err; |
b9fbe709 HX |
789 | s32 rover = -4096; |
790 | bool ok; | |
1da177e4 LT |
791 | |
792 | retry: | |
793 | cond_resched(); | |
e341694e | 794 | rcu_read_lock(); |
b9fbe709 HX |
795 | ok = !__netlink_lookup(table, portid, net); |
796 | rcu_read_unlock(); | |
797 | if (!ok) { | |
e341694e | 798 | /* Bind collision, search negative portid values. */ |
b9fbe709 HX |
799 | if (rover == -4096) |
800 | /* rover will be in range [S32_MIN, -4097] */ | |
801 | rover = S32_MIN + prandom_u32_max(-4096 - S32_MIN); | |
802 | else if (rover >= -4096) | |
e341694e | 803 | rover = -4097; |
b9fbe709 | 804 | portid = rover--; |
e341694e | 805 | goto retry; |
1da177e4 | 806 | } |
1da177e4 | 807 | |
8ea65f4a | 808 | err = netlink_insert(sk, portid); |
1da177e4 LT |
809 | if (err == -EADDRINUSE) |
810 | goto retry; | |
d470e3b4 DM |
811 | |
812 | /* If 2 threads race to autobind, that is fine. */ | |
813 | if (err == -EBUSY) | |
814 | err = 0; | |
815 | ||
816 | return err; | |
1da177e4 LT |
817 | } |
818 | ||
aa4cf945 EB |
819 | /** |
820 | * __netlink_ns_capable - General netlink message capability test | |
821 | * @nsp: NETLINK_CB of the socket buffer holding a netlink command from userspace. | |
822 | * @user_ns: The user namespace of the capability to use | |
823 | * @cap: The capability to use | |
824 | * | |
825 | * Test to see if the opener of the socket we received the message | |
826 | * from had when the netlink socket was created and the sender of the | |
827 | * message has has the capability @cap in the user namespace @user_ns. | |
828 | */ | |
829 | bool __netlink_ns_capable(const struct netlink_skb_parms *nsp, | |
830 | struct user_namespace *user_ns, int cap) | |
831 | { | |
2d7a85f4 EB |
832 | return ((nsp->flags & NETLINK_SKB_DST) || |
833 | file_ns_capable(nsp->sk->sk_socket->file, user_ns, cap)) && | |
834 | ns_capable(user_ns, cap); | |
aa4cf945 EB |
835 | } |
836 | EXPORT_SYMBOL(__netlink_ns_capable); | |
837 | ||
838 | /** | |
839 | * netlink_ns_capable - General netlink message capability test | |
840 | * @skb: socket buffer holding a netlink command from userspace | |
841 | * @user_ns: The user namespace of the capability to use | |
842 | * @cap: The capability to use | |
843 | * | |
844 | * Test to see if the opener of the socket we received the message | |
845 | * from had when the netlink socket was created and the sender of the | |
846 | * message has has the capability @cap in the user namespace @user_ns. | |
847 | */ | |
848 | bool netlink_ns_capable(const struct sk_buff *skb, | |
849 | struct user_namespace *user_ns, int cap) | |
850 | { | |
851 | return __netlink_ns_capable(&NETLINK_CB(skb), user_ns, cap); | |
852 | } | |
853 | EXPORT_SYMBOL(netlink_ns_capable); | |
854 | ||
855 | /** | |
856 | * netlink_capable - Netlink global message capability test | |
857 | * @skb: socket buffer holding a netlink command from userspace | |
858 | * @cap: The capability to use | |
859 | * | |
860 | * Test to see if the opener of the socket we received the message | |
861 | * from had when the netlink socket was created and the sender of the | |
862 | * message has has the capability @cap in all user namespaces. | |
863 | */ | |
864 | bool netlink_capable(const struct sk_buff *skb, int cap) | |
865 | { | |
866 | return netlink_ns_capable(skb, &init_user_ns, cap); | |
867 | } | |
868 | EXPORT_SYMBOL(netlink_capable); | |
869 | ||
870 | /** | |
871 | * netlink_net_capable - Netlink network namespace message capability test | |
872 | * @skb: socket buffer holding a netlink command from userspace | |
873 | * @cap: The capability to use | |
874 | * | |
875 | * Test to see if the opener of the socket we received the message | |
876 | * from had when the netlink socket was created and the sender of the | |
877 | * message has has the capability @cap over the network namespace of | |
878 | * the socket we received the message from. | |
879 | */ | |
880 | bool netlink_net_capable(const struct sk_buff *skb, int cap) | |
881 | { | |
882 | return netlink_ns_capable(skb, sock_net(skb->sk)->user_ns, cap); | |
883 | } | |
884 | EXPORT_SYMBOL(netlink_net_capable); | |
885 | ||
5187cd05 | 886 | static inline int netlink_allowed(const struct socket *sock, unsigned int flag) |
746fac4d | 887 | { |
9785e10a | 888 | return (nl_table[sock->sk->sk_protocol].flags & flag) || |
df008c91 | 889 | ns_capable(sock_net(sock->sk)->user_ns, CAP_NET_ADMIN); |
746fac4d | 890 | } |
1da177e4 | 891 | |
f7fa9b10 PM |
892 | static void |
893 | netlink_update_subscriptions(struct sock *sk, unsigned int subscriptions) | |
894 | { | |
895 | struct netlink_sock *nlk = nlk_sk(sk); | |
896 | ||
897 | if (nlk->subscriptions && !subscriptions) | |
898 | __sk_del_bind_node(sk); | |
899 | else if (!nlk->subscriptions && subscriptions) | |
900 | sk_add_bind_node(sk, &nl_table[sk->sk_protocol].mc_list); | |
901 | nlk->subscriptions = subscriptions; | |
902 | } | |
903 | ||
b4ff4f04 | 904 | static int netlink_realloc_groups(struct sock *sk) |
513c2500 PM |
905 | { |
906 | struct netlink_sock *nlk = nlk_sk(sk); | |
907 | unsigned int groups; | |
b4ff4f04 | 908 | unsigned long *new_groups; |
513c2500 PM |
909 | int err = 0; |
910 | ||
b4ff4f04 JB |
911 | netlink_table_grab(); |
912 | ||
513c2500 | 913 | groups = nl_table[sk->sk_protocol].groups; |
b4ff4f04 | 914 | if (!nl_table[sk->sk_protocol].registered) { |
513c2500 | 915 | err = -ENOENT; |
b4ff4f04 JB |
916 | goto out_unlock; |
917 | } | |
513c2500 | 918 | |
b4ff4f04 JB |
919 | if (nlk->ngroups >= groups) |
920 | goto out_unlock; | |
513c2500 | 921 | |
b4ff4f04 JB |
922 | new_groups = krealloc(nlk->groups, NLGRPSZ(groups), GFP_ATOMIC); |
923 | if (new_groups == NULL) { | |
924 | err = -ENOMEM; | |
925 | goto out_unlock; | |
926 | } | |
6ac552fd | 927 | memset((char *)new_groups + NLGRPSZ(nlk->ngroups), 0, |
b4ff4f04 JB |
928 | NLGRPSZ(groups) - NLGRPSZ(nlk->ngroups)); |
929 | ||
930 | nlk->groups = new_groups; | |
513c2500 | 931 | nlk->ngroups = groups; |
b4ff4f04 JB |
932 | out_unlock: |
933 | netlink_table_ungrab(); | |
934 | return err; | |
513c2500 PM |
935 | } |
936 | ||
02c81ab9 | 937 | static void netlink_undo_bind(int group, long unsigned int groups, |
023e2cfa | 938 | struct sock *sk) |
4f520900 | 939 | { |
023e2cfa | 940 | struct netlink_sock *nlk = nlk_sk(sk); |
4f520900 RGB |
941 | int undo; |
942 | ||
943 | if (!nlk->netlink_unbind) | |
944 | return; | |
945 | ||
946 | for (undo = 0; undo < group; undo++) | |
6251edd9 | 947 | if (test_bit(undo, &groups)) |
8b7c36d8 | 948 | nlk->netlink_unbind(sock_net(sk), undo + 1); |
4f520900 RGB |
949 | } |
950 | ||
6ac552fd PM |
951 | static int netlink_bind(struct socket *sock, struct sockaddr *addr, |
952 | int addr_len) | |
1da177e4 LT |
953 | { |
954 | struct sock *sk = sock->sk; | |
3b1e0a65 | 955 | struct net *net = sock_net(sk); |
1da177e4 LT |
956 | struct netlink_sock *nlk = nlk_sk(sk); |
957 | struct sockaddr_nl *nladdr = (struct sockaddr_nl *)addr; | |
f7736080 | 958 | int err = 0; |
4f520900 | 959 | long unsigned int groups = nladdr->nl_groups; |
da314c99 | 960 | bool bound; |
746fac4d | 961 | |
4e4b5376 HFS |
962 | if (addr_len < sizeof(struct sockaddr_nl)) |
963 | return -EINVAL; | |
964 | ||
1da177e4 LT |
965 | if (nladdr->nl_family != AF_NETLINK) |
966 | return -EINVAL; | |
967 | ||
968 | /* Only superuser is allowed to listen multicasts */ | |
4f520900 | 969 | if (groups) { |
5187cd05 | 970 | if (!netlink_allowed(sock, NL_CFG_F_NONROOT_RECV)) |
513c2500 | 971 | return -EPERM; |
b4ff4f04 JB |
972 | err = netlink_realloc_groups(sk); |
973 | if (err) | |
974 | return err; | |
513c2500 | 975 | } |
1da177e4 | 976 | |
da314c99 HX |
977 | bound = nlk->bound; |
978 | if (bound) { | |
979 | /* Ensure nlk->portid is up-to-date. */ | |
980 | smp_rmb(); | |
981 | ||
15e47304 | 982 | if (nladdr->nl_pid != nlk->portid) |
1da177e4 | 983 | return -EINVAL; |
da314c99 | 984 | } |
4f520900 | 985 | |
f7736080 | 986 | netlink_lock_table(); |
4f520900 RGB |
987 | if (nlk->netlink_bind && groups) { |
988 | int group; | |
989 | ||
990 | for (group = 0; group < nlk->ngroups; group++) { | |
991 | if (!test_bit(group, &groups)) | |
992 | continue; | |
8b7c36d8 | 993 | err = nlk->netlink_bind(net, group + 1); |
4f520900 RGB |
994 | if (!err) |
995 | continue; | |
023e2cfa | 996 | netlink_undo_bind(group, groups, sk); |
f7736080 | 997 | goto unlock; |
4f520900 RGB |
998 | } |
999 | } | |
1000 | ||
da314c99 HX |
1001 | /* No need for barriers here as we return to user-space without |
1002 | * using any of the bound attributes. | |
1003 | */ | |
1004 | if (!bound) { | |
1da177e4 | 1005 | err = nladdr->nl_pid ? |
8ea65f4a | 1006 | netlink_insert(sk, nladdr->nl_pid) : |
1da177e4 | 1007 | netlink_autobind(sock); |
4f520900 | 1008 | if (err) { |
023e2cfa | 1009 | netlink_undo_bind(nlk->ngroups, groups, sk); |
f7736080 | 1010 | goto unlock; |
4f520900 | 1011 | } |
1da177e4 LT |
1012 | } |
1013 | ||
4f520900 | 1014 | if (!groups && (nlk->groups == NULL || !(u32)nlk->groups[0])) |
f7736080 XL |
1015 | goto unlock; |
1016 | netlink_unlock_table(); | |
1da177e4 LT |
1017 | |
1018 | netlink_table_grab(); | |
f7fa9b10 | 1019 | netlink_update_subscriptions(sk, nlk->subscriptions + |
4f520900 | 1020 | hweight32(groups) - |
746fac4d | 1021 | hweight32(nlk->groups[0])); |
4f520900 | 1022 | nlk->groups[0] = (nlk->groups[0] & ~0xffffffffUL) | groups; |
4277a083 | 1023 | netlink_update_listeners(sk); |
1da177e4 LT |
1024 | netlink_table_ungrab(); |
1025 | ||
1026 | return 0; | |
f7736080 XL |
1027 | |
1028 | unlock: | |
1029 | netlink_unlock_table(); | |
1030 | return err; | |
1da177e4 LT |
1031 | } |
1032 | ||
1033 | static int netlink_connect(struct socket *sock, struct sockaddr *addr, | |
1034 | int alen, int flags) | |
1035 | { | |
1036 | int err = 0; | |
1037 | struct sock *sk = sock->sk; | |
1038 | struct netlink_sock *nlk = nlk_sk(sk); | |
6ac552fd | 1039 | struct sockaddr_nl *nladdr = (struct sockaddr_nl *)addr; |
1da177e4 | 1040 | |
6503d961 CG |
1041 | if (alen < sizeof(addr->sa_family)) |
1042 | return -EINVAL; | |
1043 | ||
1da177e4 LT |
1044 | if (addr->sa_family == AF_UNSPEC) { |
1045 | sk->sk_state = NETLINK_UNCONNECTED; | |
15e47304 | 1046 | nlk->dst_portid = 0; |
d629b836 | 1047 | nlk->dst_group = 0; |
1da177e4 LT |
1048 | return 0; |
1049 | } | |
1050 | if (addr->sa_family != AF_NETLINK) | |
1051 | return -EINVAL; | |
1052 | ||
46833a86 | 1053 | if ((nladdr->nl_groups || nladdr->nl_pid) && |
5187cd05 | 1054 | !netlink_allowed(sock, NL_CFG_F_NONROOT_SEND)) |
1da177e4 LT |
1055 | return -EPERM; |
1056 | ||
da314c99 HX |
1057 | /* No need for barriers here as we return to user-space without |
1058 | * using any of the bound attributes. | |
1059 | */ | |
1060 | if (!nlk->bound) | |
1da177e4 LT |
1061 | err = netlink_autobind(sock); |
1062 | ||
1063 | if (err == 0) { | |
1064 | sk->sk_state = NETLINK_CONNECTED; | |
15e47304 | 1065 | nlk->dst_portid = nladdr->nl_pid; |
d629b836 | 1066 | nlk->dst_group = ffs(nladdr->nl_groups); |
1da177e4 LT |
1067 | } |
1068 | ||
1069 | return err; | |
1070 | } | |
1071 | ||
6ac552fd PM |
1072 | static int netlink_getname(struct socket *sock, struct sockaddr *addr, |
1073 | int *addr_len, int peer) | |
1da177e4 LT |
1074 | { |
1075 | struct sock *sk = sock->sk; | |
1076 | struct netlink_sock *nlk = nlk_sk(sk); | |
13cfa97b | 1077 | DECLARE_SOCKADDR(struct sockaddr_nl *, nladdr, addr); |
746fac4d | 1078 | |
1da177e4 LT |
1079 | nladdr->nl_family = AF_NETLINK; |
1080 | nladdr->nl_pad = 0; | |
1081 | *addr_len = sizeof(*nladdr); | |
1082 | ||
1083 | if (peer) { | |
15e47304 | 1084 | nladdr->nl_pid = nlk->dst_portid; |
d629b836 | 1085 | nladdr->nl_groups = netlink_group_mask(nlk->dst_group); |
1da177e4 | 1086 | } else { |
15e47304 | 1087 | nladdr->nl_pid = nlk->portid; |
f7736080 | 1088 | netlink_lock_table(); |
513c2500 | 1089 | nladdr->nl_groups = nlk->groups ? nlk->groups[0] : 0; |
f7736080 | 1090 | netlink_unlock_table(); |
1da177e4 LT |
1091 | } |
1092 | return 0; | |
1093 | } | |
1094 | ||
025c6818 DD |
1095 | static int netlink_ioctl(struct socket *sock, unsigned int cmd, |
1096 | unsigned long arg) | |
1097 | { | |
1098 | /* try to hand this ioctl down to the NIC drivers. | |
1099 | */ | |
1100 | return -ENOIOCTLCMD; | |
1101 | } | |
1102 | ||
15e47304 | 1103 | static struct sock *netlink_getsockbyportid(struct sock *ssk, u32 portid) |
1da177e4 | 1104 | { |
1da177e4 LT |
1105 | struct sock *sock; |
1106 | struct netlink_sock *nlk; | |
1107 | ||
15e47304 | 1108 | sock = netlink_lookup(sock_net(ssk), ssk->sk_protocol, portid); |
1da177e4 LT |
1109 | if (!sock) |
1110 | return ERR_PTR(-ECONNREFUSED); | |
1111 | ||
1112 | /* Don't bother queuing skb if kernel socket has no input function */ | |
1113 | nlk = nlk_sk(sock); | |
cd40b7d3 | 1114 | if (sock->sk_state == NETLINK_CONNECTED && |
15e47304 | 1115 | nlk->dst_portid != nlk_sk(ssk)->portid) { |
1da177e4 LT |
1116 | sock_put(sock); |
1117 | return ERR_PTR(-ECONNREFUSED); | |
1118 | } | |
1119 | return sock; | |
1120 | } | |
1121 | ||
1122 | struct sock *netlink_getsockbyfilp(struct file *filp) | |
1123 | { | |
496ad9aa | 1124 | struct inode *inode = file_inode(filp); |
1da177e4 LT |
1125 | struct sock *sock; |
1126 | ||
1127 | if (!S_ISSOCK(inode->i_mode)) | |
1128 | return ERR_PTR(-ENOTSOCK); | |
1129 | ||
1130 | sock = SOCKET_I(inode)->sk; | |
1131 | if (sock->sk_family != AF_NETLINK) | |
1132 | return ERR_PTR(-EINVAL); | |
1133 | ||
1134 | sock_hold(sock); | |
1135 | return sock; | |
1136 | } | |
1137 | ||
3a36515f PN |
1138 | static struct sk_buff *netlink_alloc_large_skb(unsigned int size, |
1139 | int broadcast) | |
c05cdb1b PNA |
1140 | { |
1141 | struct sk_buff *skb; | |
1142 | void *data; | |
1143 | ||
3a36515f | 1144 | if (size <= NLMSG_GOODSIZE || broadcast) |
c05cdb1b PNA |
1145 | return alloc_skb(size, GFP_KERNEL); |
1146 | ||
3a36515f PN |
1147 | size = SKB_DATA_ALIGN(size) + |
1148 | SKB_DATA_ALIGN(sizeof(struct skb_shared_info)); | |
c05cdb1b PNA |
1149 | |
1150 | data = vmalloc(size); | |
1151 | if (data == NULL) | |
3a36515f | 1152 | return NULL; |
c05cdb1b | 1153 | |
2ea2f62c | 1154 | skb = __build_skb(data, size); |
3a36515f PN |
1155 | if (skb == NULL) |
1156 | vfree(data); | |
2ea2f62c | 1157 | else |
3a36515f | 1158 | skb->destructor = netlink_skb_destructor; |
c05cdb1b PNA |
1159 | |
1160 | return skb; | |
c05cdb1b PNA |
1161 | } |
1162 | ||
1da177e4 LT |
1163 | /* |
1164 | * Attach a skb to a netlink socket. | |
1165 | * The caller must hold a reference to the destination socket. On error, the | |
1166 | * reference is dropped. The skb is not send to the destination, just all | |
1167 | * all error checks are performed and memory in the queue is reserved. | |
1168 | * Return values: | |
1169 | * < 0: error. skb freed, reference to sock dropped. | |
1170 | * 0: continue | |
1171 | * 1: repeat lookup - reference dropped while waiting for socket memory. | |
1172 | */ | |
9457afee | 1173 | int netlink_attachskb(struct sock *sk, struct sk_buff *skb, |
c3d8d1e3 | 1174 | long *timeo, struct sock *ssk) |
1da177e4 LT |
1175 | { |
1176 | struct netlink_sock *nlk; | |
1177 | ||
1178 | nlk = nlk_sk(sk); | |
1179 | ||
5fd96123 | 1180 | if ((atomic_read(&sk->sk_rmem_alloc) > sk->sk_rcvbuf || |
d1b4c689 | 1181 | test_bit(NETLINK_S_CONGESTED, &nlk->state))) { |
1da177e4 | 1182 | DECLARE_WAITQUEUE(wait, current); |
c3d8d1e3 | 1183 | if (!*timeo) { |
aed81560 | 1184 | if (!ssk || netlink_is_kernel(ssk)) |
1da177e4 LT |
1185 | netlink_overrun(sk); |
1186 | sock_put(sk); | |
1187 | kfree_skb(skb); | |
1188 | return -EAGAIN; | |
1189 | } | |
1190 | ||
1191 | __set_current_state(TASK_INTERRUPTIBLE); | |
1192 | add_wait_queue(&nlk->wait, &wait); | |
1193 | ||
1194 | if ((atomic_read(&sk->sk_rmem_alloc) > sk->sk_rcvbuf || | |
cc3a572f | 1195 | test_bit(NETLINK_S_CONGESTED, &nlk->state)) && |
1da177e4 | 1196 | !sock_flag(sk, SOCK_DEAD)) |
c3d8d1e3 | 1197 | *timeo = schedule_timeout(*timeo); |
1da177e4 LT |
1198 | |
1199 | __set_current_state(TASK_RUNNING); | |
1200 | remove_wait_queue(&nlk->wait, &wait); | |
1201 | sock_put(sk); | |
1202 | ||
1203 | if (signal_pending(current)) { | |
1204 | kfree_skb(skb); | |
c3d8d1e3 | 1205 | return sock_intr_errno(*timeo); |
1da177e4 LT |
1206 | } |
1207 | return 1; | |
1208 | } | |
cf0a018a | 1209 | netlink_skb_set_owner_r(skb, sk); |
1da177e4 LT |
1210 | return 0; |
1211 | } | |
1212 | ||
4a7e7c2a | 1213 | static int __netlink_sendskb(struct sock *sk, struct sk_buff *skb) |
1da177e4 | 1214 | { |
1da177e4 LT |
1215 | int len = skb->len; |
1216 | ||
bcbde0d4 DB |
1217 | netlink_deliver_tap(skb); |
1218 | ||
d1b4c689 | 1219 | skb_queue_tail(&sk->sk_receive_queue, skb); |
676d2369 | 1220 | sk->sk_data_ready(sk); |
4a7e7c2a ED |
1221 | return len; |
1222 | } | |
1223 | ||
1224 | int netlink_sendskb(struct sock *sk, struct sk_buff *skb) | |
1225 | { | |
1226 | int len = __netlink_sendskb(sk, skb); | |
1227 | ||
1da177e4 LT |
1228 | sock_put(sk); |
1229 | return len; | |
1230 | } | |
1231 | ||
1232 | void netlink_detachskb(struct sock *sk, struct sk_buff *skb) | |
1233 | { | |
1234 | kfree_skb(skb); | |
1235 | sock_put(sk); | |
1236 | } | |
1237 | ||
b57ef81f | 1238 | static struct sk_buff *netlink_trim(struct sk_buff *skb, gfp_t allocation) |
1da177e4 LT |
1239 | { |
1240 | int delta; | |
1241 | ||
1298ca46 | 1242 | WARN_ON(skb->sk != NULL); |
4305b541 | 1243 | delta = skb->end - skb->tail; |
c05cdb1b | 1244 | if (is_vmalloc_addr(skb->head) || delta * 2 < skb->truesize) |
1da177e4 LT |
1245 | return skb; |
1246 | ||
1247 | if (skb_shared(skb)) { | |
1248 | struct sk_buff *nskb = skb_clone(skb, allocation); | |
1249 | if (!nskb) | |
1250 | return skb; | |
8460c00f | 1251 | consume_skb(skb); |
1da177e4 LT |
1252 | skb = nskb; |
1253 | } | |
1254 | ||
158f323b ED |
1255 | pskb_expand_head(skb, 0, -delta, |
1256 | (allocation & ~__GFP_DIRECT_RECLAIM) | | |
1257 | __GFP_NOWARN | __GFP_NORETRY); | |
1da177e4 LT |
1258 | return skb; |
1259 | } | |
1260 | ||
3fbc2905 EB |
1261 | static int netlink_unicast_kernel(struct sock *sk, struct sk_buff *skb, |
1262 | struct sock *ssk) | |
cd40b7d3 DL |
1263 | { |
1264 | int ret; | |
1265 | struct netlink_sock *nlk = nlk_sk(sk); | |
1266 | ||
1267 | ret = -ECONNREFUSED; | |
1268 | if (nlk->netlink_rcv != NULL) { | |
1269 | ret = skb->len; | |
cf0a018a | 1270 | netlink_skb_set_owner_r(skb, sk); |
e32123e5 | 1271 | NETLINK_CB(skb).sk = ssk; |
73bfd370 | 1272 | netlink_deliver_tap_kernel(sk, ssk, skb); |
cd40b7d3 | 1273 | nlk->netlink_rcv(skb); |
bfb253c9 ED |
1274 | consume_skb(skb); |
1275 | } else { | |
1276 | kfree_skb(skb); | |
cd40b7d3 | 1277 | } |
cd40b7d3 DL |
1278 | sock_put(sk); |
1279 | return ret; | |
1280 | } | |
1281 | ||
1282 | int netlink_unicast(struct sock *ssk, struct sk_buff *skb, | |
15e47304 | 1283 | u32 portid, int nonblock) |
1da177e4 LT |
1284 | { |
1285 | struct sock *sk; | |
1286 | int err; | |
1287 | long timeo; | |
1288 | ||
1289 | skb = netlink_trim(skb, gfp_any()); | |
1290 | ||
1291 | timeo = sock_sndtimeo(ssk, nonblock); | |
1292 | retry: | |
15e47304 | 1293 | sk = netlink_getsockbyportid(ssk, portid); |
1da177e4 LT |
1294 | if (IS_ERR(sk)) { |
1295 | kfree_skb(skb); | |
1296 | return PTR_ERR(sk); | |
1297 | } | |
cd40b7d3 | 1298 | if (netlink_is_kernel(sk)) |
3fbc2905 | 1299 | return netlink_unicast_kernel(sk, skb, ssk); |
cd40b7d3 | 1300 | |
b1153f29 | 1301 | if (sk_filter(sk, skb)) { |
84874607 | 1302 | err = skb->len; |
b1153f29 SH |
1303 | kfree_skb(skb); |
1304 | sock_put(sk); | |
1305 | return err; | |
1306 | } | |
1307 | ||
9457afee | 1308 | err = netlink_attachskb(sk, skb, &timeo, ssk); |
1da177e4 LT |
1309 | if (err == 1) |
1310 | goto retry; | |
1311 | if (err) | |
1312 | return err; | |
1313 | ||
7ee015e0 | 1314 | return netlink_sendskb(sk, skb); |
1da177e4 | 1315 | } |
6ac552fd | 1316 | EXPORT_SYMBOL(netlink_unicast); |
1da177e4 | 1317 | |
4277a083 PM |
1318 | int netlink_has_listeners(struct sock *sk, unsigned int group) |
1319 | { | |
1320 | int res = 0; | |
5c398dc8 | 1321 | struct listeners *listeners; |
4277a083 | 1322 | |
aed81560 | 1323 | BUG_ON(!netlink_is_kernel(sk)); |
b4ff4f04 JB |
1324 | |
1325 | rcu_read_lock(); | |
1326 | listeners = rcu_dereference(nl_table[sk->sk_protocol].listeners); | |
1327 | ||
6d772ac5 | 1328 | if (listeners && group - 1 < nl_table[sk->sk_protocol].groups) |
5c398dc8 | 1329 | res = test_bit(group - 1, listeners->masks); |
b4ff4f04 JB |
1330 | |
1331 | rcu_read_unlock(); | |
1332 | ||
4277a083 PM |
1333 | return res; |
1334 | } | |
1335 | EXPORT_SYMBOL_GPL(netlink_has_listeners); | |
1336 | ||
b57ef81f | 1337 | static int netlink_broadcast_deliver(struct sock *sk, struct sk_buff *skb) |
1da177e4 LT |
1338 | { |
1339 | struct netlink_sock *nlk = nlk_sk(sk); | |
1340 | ||
1341 | if (atomic_read(&sk->sk_rmem_alloc) <= sk->sk_rcvbuf && | |
cc3a572f | 1342 | !test_bit(NETLINK_S_CONGESTED, &nlk->state)) { |
cf0a018a | 1343 | netlink_skb_set_owner_r(skb, sk); |
4a7e7c2a | 1344 | __netlink_sendskb(sk, skb); |
2c645800 | 1345 | return atomic_read(&sk->sk_rmem_alloc) > (sk->sk_rcvbuf >> 1); |
1da177e4 LT |
1346 | } |
1347 | return -1; | |
1348 | } | |
1349 | ||
1350 | struct netlink_broadcast_data { | |
1351 | struct sock *exclude_sk; | |
b4b51029 | 1352 | struct net *net; |
15e47304 | 1353 | u32 portid; |
1da177e4 LT |
1354 | u32 group; |
1355 | int failure; | |
ff491a73 | 1356 | int delivery_failure; |
1da177e4 LT |
1357 | int congested; |
1358 | int delivered; | |
7d877f3b | 1359 | gfp_t allocation; |
1da177e4 | 1360 | struct sk_buff *skb, *skb2; |
910a7e90 EB |
1361 | int (*tx_filter)(struct sock *dsk, struct sk_buff *skb, void *data); |
1362 | void *tx_data; | |
1da177e4 LT |
1363 | }; |
1364 | ||
46c9521f RR |
1365 | static void do_one_broadcast(struct sock *sk, |
1366 | struct netlink_broadcast_data *p) | |
1da177e4 LT |
1367 | { |
1368 | struct netlink_sock *nlk = nlk_sk(sk); | |
1369 | int val; | |
1370 | ||
1371 | if (p->exclude_sk == sk) | |
46c9521f | 1372 | return; |
1da177e4 | 1373 | |
15e47304 | 1374 | if (nlk->portid == p->portid || p->group - 1 >= nlk->ngroups || |
f7fa9b10 | 1375 | !test_bit(p->group - 1, nlk->groups)) |
46c9521f | 1376 | return; |
1da177e4 | 1377 | |
59324cf3 ND |
1378 | if (!net_eq(sock_net(sk), p->net)) { |
1379 | if (!(nlk->flags & NETLINK_F_LISTEN_ALL_NSID)) | |
1380 | return; | |
1381 | ||
1382 | if (!peernet_has_id(sock_net(sk), p->net)) | |
1383 | return; | |
1384 | ||
1385 | if (!file_ns_capable(sk->sk_socket->file, p->net->user_ns, | |
1386 | CAP_NET_BROADCAST)) | |
1387 | return; | |
1388 | } | |
b4b51029 | 1389 | |
1da177e4 LT |
1390 | if (p->failure) { |
1391 | netlink_overrun(sk); | |
46c9521f | 1392 | return; |
1da177e4 LT |
1393 | } |
1394 | ||
1395 | sock_hold(sk); | |
1396 | if (p->skb2 == NULL) { | |
68acc024 | 1397 | if (skb_shared(p->skb)) { |
1da177e4 LT |
1398 | p->skb2 = skb_clone(p->skb, p->allocation); |
1399 | } else { | |
68acc024 TC |
1400 | p->skb2 = skb_get(p->skb); |
1401 | /* | |
1402 | * skb ownership may have been set when | |
1403 | * delivered to a previous socket. | |
1404 | */ | |
1405 | skb_orphan(p->skb2); | |
1da177e4 LT |
1406 | } |
1407 | } | |
1408 | if (p->skb2 == NULL) { | |
1409 | netlink_overrun(sk); | |
1410 | /* Clone failed. Notify ALL listeners. */ | |
1411 | p->failure = 1; | |
cc3a572f | 1412 | if (nlk->flags & NETLINK_F_BROADCAST_SEND_ERROR) |
be0c22a4 | 1413 | p->delivery_failure = 1; |
59324cf3 ND |
1414 | goto out; |
1415 | } | |
1416 | if (p->tx_filter && p->tx_filter(sk, p->skb2, p->tx_data)) { | |
910a7e90 EB |
1417 | kfree_skb(p->skb2); |
1418 | p->skb2 = NULL; | |
59324cf3 ND |
1419 | goto out; |
1420 | } | |
1421 | if (sk_filter(sk, p->skb2)) { | |
b1153f29 SH |
1422 | kfree_skb(p->skb2); |
1423 | p->skb2 = NULL; | |
59324cf3 ND |
1424 | goto out; |
1425 | } | |
1426 | NETLINK_CB(p->skb2).nsid = peernet2id(sock_net(sk), p->net); | |
7212462f ND |
1427 | if (NETLINK_CB(p->skb2).nsid != NETNSA_NSID_NOT_ASSIGNED) |
1428 | NETLINK_CB(p->skb2).nsid_is_set = true; | |
59324cf3 ND |
1429 | val = netlink_broadcast_deliver(sk, p->skb2); |
1430 | if (val < 0) { | |
1da177e4 | 1431 | netlink_overrun(sk); |
cc3a572f | 1432 | if (nlk->flags & NETLINK_F_BROADCAST_SEND_ERROR) |
be0c22a4 | 1433 | p->delivery_failure = 1; |
1da177e4 LT |
1434 | } else { |
1435 | p->congested |= val; | |
1436 | p->delivered = 1; | |
1437 | p->skb2 = NULL; | |
1438 | } | |
59324cf3 | 1439 | out: |
1da177e4 | 1440 | sock_put(sk); |
1da177e4 LT |
1441 | } |
1442 | ||
15e47304 | 1443 | int netlink_broadcast_filtered(struct sock *ssk, struct sk_buff *skb, u32 portid, |
910a7e90 EB |
1444 | u32 group, gfp_t allocation, |
1445 | int (*filter)(struct sock *dsk, struct sk_buff *skb, void *data), | |
1446 | void *filter_data) | |
1da177e4 | 1447 | { |
3b1e0a65 | 1448 | struct net *net = sock_net(ssk); |
1da177e4 | 1449 | struct netlink_broadcast_data info; |
1da177e4 LT |
1450 | struct sock *sk; |
1451 | ||
1452 | skb = netlink_trim(skb, allocation); | |
1453 | ||
1454 | info.exclude_sk = ssk; | |
b4b51029 | 1455 | info.net = net; |
15e47304 | 1456 | info.portid = portid; |
1da177e4 LT |
1457 | info.group = group; |
1458 | info.failure = 0; | |
ff491a73 | 1459 | info.delivery_failure = 0; |
1da177e4 LT |
1460 | info.congested = 0; |
1461 | info.delivered = 0; | |
1462 | info.allocation = allocation; | |
1463 | info.skb = skb; | |
1464 | info.skb2 = NULL; | |
910a7e90 EB |
1465 | info.tx_filter = filter; |
1466 | info.tx_data = filter_data; | |
1da177e4 LT |
1467 | |
1468 | /* While we sleep in clone, do not allow to change socket list */ | |
1469 | ||
1470 | netlink_lock_table(); | |
1471 | ||
b67bfe0d | 1472 | sk_for_each_bound(sk, &nl_table[ssk->sk_protocol].mc_list) |
1da177e4 LT |
1473 | do_one_broadcast(sk, &info); |
1474 | ||
70d4bf6d | 1475 | consume_skb(skb); |
aa1c6a6f | 1476 | |
1da177e4 LT |
1477 | netlink_unlock_table(); |
1478 | ||
70d4bf6d NH |
1479 | if (info.delivery_failure) { |
1480 | kfree_skb(info.skb2); | |
ff491a73 | 1481 | return -ENOBUFS; |
658cb354 ED |
1482 | } |
1483 | consume_skb(info.skb2); | |
ff491a73 | 1484 | |
1da177e4 | 1485 | if (info.delivered) { |
d0164adc | 1486 | if (info.congested && gfpflags_allow_blocking(allocation)) |
1da177e4 LT |
1487 | yield(); |
1488 | return 0; | |
1489 | } | |
1da177e4 LT |
1490 | return -ESRCH; |
1491 | } | |
910a7e90 EB |
1492 | EXPORT_SYMBOL(netlink_broadcast_filtered); |
1493 | ||
15e47304 | 1494 | int netlink_broadcast(struct sock *ssk, struct sk_buff *skb, u32 portid, |
910a7e90 EB |
1495 | u32 group, gfp_t allocation) |
1496 | { | |
15e47304 | 1497 | return netlink_broadcast_filtered(ssk, skb, portid, group, allocation, |
910a7e90 EB |
1498 | NULL, NULL); |
1499 | } | |
6ac552fd | 1500 | EXPORT_SYMBOL(netlink_broadcast); |
1da177e4 LT |
1501 | |
1502 | struct netlink_set_err_data { | |
1503 | struct sock *exclude_sk; | |
15e47304 | 1504 | u32 portid; |
1da177e4 LT |
1505 | u32 group; |
1506 | int code; | |
1507 | }; | |
1508 | ||
b57ef81f | 1509 | static int do_one_set_err(struct sock *sk, struct netlink_set_err_data *p) |
1da177e4 LT |
1510 | { |
1511 | struct netlink_sock *nlk = nlk_sk(sk); | |
1a50307b | 1512 | int ret = 0; |
1da177e4 LT |
1513 | |
1514 | if (sk == p->exclude_sk) | |
1515 | goto out; | |
1516 | ||
09ad9bc7 | 1517 | if (!net_eq(sock_net(sk), sock_net(p->exclude_sk))) |
b4b51029 EB |
1518 | goto out; |
1519 | ||
15e47304 | 1520 | if (nlk->portid == p->portid || p->group - 1 >= nlk->ngroups || |
f7fa9b10 | 1521 | !test_bit(p->group - 1, nlk->groups)) |
1da177e4 LT |
1522 | goto out; |
1523 | ||
cc3a572f | 1524 | if (p->code == ENOBUFS && nlk->flags & NETLINK_F_RECV_NO_ENOBUFS) { |
1a50307b PNA |
1525 | ret = 1; |
1526 | goto out; | |
1527 | } | |
1528 | ||
1da177e4 LT |
1529 | sk->sk_err = p->code; |
1530 | sk->sk_error_report(sk); | |
1531 | out: | |
1a50307b | 1532 | return ret; |
1da177e4 LT |
1533 | } |
1534 | ||
4843b93c PNA |
1535 | /** |
1536 | * netlink_set_err - report error to broadcast listeners | |
1537 | * @ssk: the kernel netlink socket, as returned by netlink_kernel_create() | |
15e47304 | 1538 | * @portid: the PORTID of a process that we want to skip (if any) |
840e93f2 | 1539 | * @group: the broadcast group that will notice the error |
4843b93c | 1540 | * @code: error code, must be negative (as usual in kernelspace) |
1a50307b PNA |
1541 | * |
1542 | * This function returns the number of broadcast listeners that have set the | |
cc3a572f | 1543 | * NETLINK_NO_ENOBUFS socket option. |
4843b93c | 1544 | */ |
15e47304 | 1545 | int netlink_set_err(struct sock *ssk, u32 portid, u32 group, int code) |
1da177e4 LT |
1546 | { |
1547 | struct netlink_set_err_data info; | |
1da177e4 | 1548 | struct sock *sk; |
1a50307b | 1549 | int ret = 0; |
1da177e4 LT |
1550 | |
1551 | info.exclude_sk = ssk; | |
15e47304 | 1552 | info.portid = portid; |
1da177e4 | 1553 | info.group = group; |
4843b93c PNA |
1554 | /* sk->sk_err wants a positive error value */ |
1555 | info.code = -code; | |
1da177e4 LT |
1556 | |
1557 | read_lock(&nl_table_lock); | |
1558 | ||
b67bfe0d | 1559 | sk_for_each_bound(sk, &nl_table[ssk->sk_protocol].mc_list) |
1a50307b | 1560 | ret += do_one_set_err(sk, &info); |
1da177e4 LT |
1561 | |
1562 | read_unlock(&nl_table_lock); | |
1a50307b | 1563 | return ret; |
1da177e4 | 1564 | } |
dd5b6ce6 | 1565 | EXPORT_SYMBOL(netlink_set_err); |
1da177e4 | 1566 | |
84659eb5 JB |
1567 | /* must be called with netlink table grabbed */ |
1568 | static void netlink_update_socket_mc(struct netlink_sock *nlk, | |
1569 | unsigned int group, | |
1570 | int is_new) | |
1571 | { | |
1572 | int old, new = !!is_new, subscriptions; | |
1573 | ||
1574 | old = test_bit(group - 1, nlk->groups); | |
1575 | subscriptions = nlk->subscriptions - old + new; | |
1576 | if (new) | |
1577 | __set_bit(group - 1, nlk->groups); | |
1578 | else | |
1579 | __clear_bit(group - 1, nlk->groups); | |
1580 | netlink_update_subscriptions(&nlk->sk, subscriptions); | |
1581 | netlink_update_listeners(&nlk->sk); | |
1582 | } | |
1583 | ||
9a4595bc | 1584 | static int netlink_setsockopt(struct socket *sock, int level, int optname, |
b7058842 | 1585 | char __user *optval, unsigned int optlen) |
9a4595bc PM |
1586 | { |
1587 | struct sock *sk = sock->sk; | |
1588 | struct netlink_sock *nlk = nlk_sk(sk); | |
eb496534 JB |
1589 | unsigned int val = 0; |
1590 | int err; | |
9a4595bc PM |
1591 | |
1592 | if (level != SOL_NETLINK) | |
1593 | return -ENOPROTOOPT; | |
1594 | ||
d1b4c689 | 1595 | if (optlen >= sizeof(int) && |
eb496534 | 1596 | get_user(val, (unsigned int __user *)optval)) |
9a4595bc PM |
1597 | return -EFAULT; |
1598 | ||
1599 | switch (optname) { | |
1600 | case NETLINK_PKTINFO: | |
1601 | if (val) | |
cc3a572f | 1602 | nlk->flags |= NETLINK_F_RECV_PKTINFO; |
9a4595bc | 1603 | else |
cc3a572f | 1604 | nlk->flags &= ~NETLINK_F_RECV_PKTINFO; |
9a4595bc PM |
1605 | err = 0; |
1606 | break; | |
1607 | case NETLINK_ADD_MEMBERSHIP: | |
1608 | case NETLINK_DROP_MEMBERSHIP: { | |
5187cd05 | 1609 | if (!netlink_allowed(sock, NL_CFG_F_NONROOT_RECV)) |
9a4595bc | 1610 | return -EPERM; |
b4ff4f04 JB |
1611 | err = netlink_realloc_groups(sk); |
1612 | if (err) | |
1613 | return err; | |
9a4595bc PM |
1614 | if (!val || val - 1 >= nlk->ngroups) |
1615 | return -EINVAL; | |
7774d5e0 | 1616 | if (optname == NETLINK_ADD_MEMBERSHIP && nlk->netlink_bind) { |
023e2cfa | 1617 | err = nlk->netlink_bind(sock_net(sk), val); |
4f520900 RGB |
1618 | if (err) |
1619 | return err; | |
1620 | } | |
9a4595bc | 1621 | netlink_table_grab(); |
84659eb5 JB |
1622 | netlink_update_socket_mc(nlk, val, |
1623 | optname == NETLINK_ADD_MEMBERSHIP); | |
9a4595bc | 1624 | netlink_table_ungrab(); |
7774d5e0 | 1625 | if (optname == NETLINK_DROP_MEMBERSHIP && nlk->netlink_unbind) |
023e2cfa | 1626 | nlk->netlink_unbind(sock_net(sk), val); |
03292745 | 1627 | |
9a4595bc PM |
1628 | err = 0; |
1629 | break; | |
1630 | } | |
be0c22a4 PNA |
1631 | case NETLINK_BROADCAST_ERROR: |
1632 | if (val) | |
cc3a572f | 1633 | nlk->flags |= NETLINK_F_BROADCAST_SEND_ERROR; |
be0c22a4 | 1634 | else |
cc3a572f | 1635 | nlk->flags &= ~NETLINK_F_BROADCAST_SEND_ERROR; |
be0c22a4 PNA |
1636 | err = 0; |
1637 | break; | |
38938bfe PNA |
1638 | case NETLINK_NO_ENOBUFS: |
1639 | if (val) { | |
cc3a572f ND |
1640 | nlk->flags |= NETLINK_F_RECV_NO_ENOBUFS; |
1641 | clear_bit(NETLINK_S_CONGESTED, &nlk->state); | |
38938bfe | 1642 | wake_up_interruptible(&nlk->wait); |
658cb354 | 1643 | } else { |
cc3a572f | 1644 | nlk->flags &= ~NETLINK_F_RECV_NO_ENOBUFS; |
658cb354 | 1645 | } |
38938bfe PNA |
1646 | err = 0; |
1647 | break; | |
59324cf3 ND |
1648 | case NETLINK_LISTEN_ALL_NSID: |
1649 | if (!ns_capable(sock_net(sk)->user_ns, CAP_NET_BROADCAST)) | |
1650 | return -EPERM; | |
1651 | ||
1652 | if (val) | |
1653 | nlk->flags |= NETLINK_F_LISTEN_ALL_NSID; | |
1654 | else | |
1655 | nlk->flags &= ~NETLINK_F_LISTEN_ALL_NSID; | |
1656 | err = 0; | |
1657 | break; | |
0a6a3a23 CR |
1658 | case NETLINK_CAP_ACK: |
1659 | if (val) | |
1660 | nlk->flags |= NETLINK_F_CAP_ACK; | |
1661 | else | |
1662 | nlk->flags &= ~NETLINK_F_CAP_ACK; | |
1663 | err = 0; | |
1664 | break; | |
2d4bc933 JB |
1665 | case NETLINK_EXT_ACK: |
1666 | if (val) | |
1667 | nlk->flags |= NETLINK_F_EXT_ACK; | |
1668 | else | |
1669 | nlk->flags &= ~NETLINK_F_EXT_ACK; | |
1670 | err = 0; | |
1671 | break; | |
9a4595bc PM |
1672 | default: |
1673 | err = -ENOPROTOOPT; | |
1674 | } | |
1675 | return err; | |
1676 | } | |
1677 | ||
1678 | static int netlink_getsockopt(struct socket *sock, int level, int optname, | |
746fac4d | 1679 | char __user *optval, int __user *optlen) |
9a4595bc PM |
1680 | { |
1681 | struct sock *sk = sock->sk; | |
1682 | struct netlink_sock *nlk = nlk_sk(sk); | |
1683 | int len, val, err; | |
1684 | ||
1685 | if (level != SOL_NETLINK) | |
1686 | return -ENOPROTOOPT; | |
1687 | ||
1688 | if (get_user(len, optlen)) | |
1689 | return -EFAULT; | |
1690 | if (len < 0) | |
1691 | return -EINVAL; | |
1692 | ||
1693 | switch (optname) { | |
1694 | case NETLINK_PKTINFO: | |
1695 | if (len < sizeof(int)) | |
1696 | return -EINVAL; | |
1697 | len = sizeof(int); | |
cc3a572f | 1698 | val = nlk->flags & NETLINK_F_RECV_PKTINFO ? 1 : 0; |
a27b58fe HC |
1699 | if (put_user(len, optlen) || |
1700 | put_user(val, optval)) | |
1701 | return -EFAULT; | |
9a4595bc PM |
1702 | err = 0; |
1703 | break; | |
be0c22a4 PNA |
1704 | case NETLINK_BROADCAST_ERROR: |
1705 | if (len < sizeof(int)) | |
1706 | return -EINVAL; | |
1707 | len = sizeof(int); | |
cc3a572f | 1708 | val = nlk->flags & NETLINK_F_BROADCAST_SEND_ERROR ? 1 : 0; |
be0c22a4 PNA |
1709 | if (put_user(len, optlen) || |
1710 | put_user(val, optval)) | |
1711 | return -EFAULT; | |
1712 | err = 0; | |
1713 | break; | |
38938bfe PNA |
1714 | case NETLINK_NO_ENOBUFS: |
1715 | if (len < sizeof(int)) | |
1716 | return -EINVAL; | |
1717 | len = sizeof(int); | |
cc3a572f | 1718 | val = nlk->flags & NETLINK_F_RECV_NO_ENOBUFS ? 1 : 0; |
38938bfe PNA |
1719 | if (put_user(len, optlen) || |
1720 | put_user(val, optval)) | |
1721 | return -EFAULT; | |
1722 | err = 0; | |
1723 | break; | |
b42be38b DH |
1724 | case NETLINK_LIST_MEMBERSHIPS: { |
1725 | int pos, idx, shift; | |
1726 | ||
1727 | err = 0; | |
47191d65 | 1728 | netlink_lock_table(); |
b42be38b DH |
1729 | for (pos = 0; pos * 8 < nlk->ngroups; pos += sizeof(u32)) { |
1730 | if (len - pos < sizeof(u32)) | |
1731 | break; | |
1732 | ||
1733 | idx = pos / sizeof(unsigned long); | |
1734 | shift = (pos % sizeof(unsigned long)) * 8; | |
1735 | if (put_user((u32)(nlk->groups[idx] >> shift), | |
1736 | (u32 __user *)(optval + pos))) { | |
1737 | err = -EFAULT; | |
1738 | break; | |
1739 | } | |
1740 | } | |
1741 | if (put_user(ALIGN(nlk->ngroups / 8, sizeof(u32)), optlen)) | |
1742 | err = -EFAULT; | |
47191d65 | 1743 | netlink_unlock_table(); |
b42be38b DH |
1744 | break; |
1745 | } | |
0a6a3a23 CR |
1746 | case NETLINK_CAP_ACK: |
1747 | if (len < sizeof(int)) | |
1748 | return -EINVAL; | |
1749 | len = sizeof(int); | |
1750 | val = nlk->flags & NETLINK_F_CAP_ACK ? 1 : 0; | |
1751 | if (put_user(len, optlen) || | |
1752 | put_user(val, optval)) | |
1753 | return -EFAULT; | |
1754 | err = 0; | |
1755 | break; | |
2d4bc933 JB |
1756 | case NETLINK_EXT_ACK: |
1757 | if (len < sizeof(int)) | |
1758 | return -EINVAL; | |
1759 | len = sizeof(int); | |
1760 | val = nlk->flags & NETLINK_F_EXT_ACK ? 1 : 0; | |
1761 | if (put_user(len, optlen) || put_user(val, optval)) | |
1762 | return -EFAULT; | |
1763 | err = 0; | |
1764 | break; | |
9a4595bc PM |
1765 | default: |
1766 | err = -ENOPROTOOPT; | |
1767 | } | |
1768 | return err; | |
1769 | } | |
1770 | ||
1771 | static void netlink_cmsg_recv_pktinfo(struct msghdr *msg, struct sk_buff *skb) | |
1772 | { | |
1773 | struct nl_pktinfo info; | |
1774 | ||
1775 | info.group = NETLINK_CB(skb).dst_group; | |
1776 | put_cmsg(msg, SOL_NETLINK, NETLINK_PKTINFO, sizeof(info), &info); | |
1777 | } | |
1778 | ||
59324cf3 ND |
1779 | static void netlink_cmsg_listen_all_nsid(struct sock *sk, struct msghdr *msg, |
1780 | struct sk_buff *skb) | |
1781 | { | |
1782 | if (!NETLINK_CB(skb).nsid_is_set) | |
1783 | return; | |
1784 | ||
1785 | put_cmsg(msg, SOL_NETLINK, NETLINK_LISTEN_ALL_NSID, sizeof(int), | |
1786 | &NETLINK_CB(skb).nsid); | |
1787 | } | |
1788 | ||
1b784140 | 1789 | static int netlink_sendmsg(struct socket *sock, struct msghdr *msg, size_t len) |
1da177e4 | 1790 | { |
1da177e4 LT |
1791 | struct sock *sk = sock->sk; |
1792 | struct netlink_sock *nlk = nlk_sk(sk); | |
342dfc30 | 1793 | DECLARE_SOCKADDR(struct sockaddr_nl *, addr, msg->msg_name); |
15e47304 | 1794 | u32 dst_portid; |
d629b836 | 1795 | u32 dst_group; |
1da177e4 LT |
1796 | struct sk_buff *skb; |
1797 | int err; | |
1798 | struct scm_cookie scm; | |
2d7a85f4 | 1799 | u32 netlink_skb_flags = 0; |
1da177e4 LT |
1800 | |
1801 | if (msg->msg_flags&MSG_OOB) | |
1802 | return -EOPNOTSUPP; | |
1803 | ||
7cc05662 | 1804 | err = scm_send(sock, msg, &scm, true); |
1da177e4 LT |
1805 | if (err < 0) |
1806 | return err; | |
1807 | ||
1808 | if (msg->msg_namelen) { | |
b47030c7 | 1809 | err = -EINVAL; |
1da177e4 | 1810 | if (addr->nl_family != AF_NETLINK) |
b47030c7 | 1811 | goto out; |
15e47304 | 1812 | dst_portid = addr->nl_pid; |
d629b836 | 1813 | dst_group = ffs(addr->nl_groups); |
b47030c7 | 1814 | err = -EPERM; |
15e47304 | 1815 | if ((dst_group || dst_portid) && |
5187cd05 | 1816 | !netlink_allowed(sock, NL_CFG_F_NONROOT_SEND)) |
b47030c7 | 1817 | goto out; |
2d7a85f4 | 1818 | netlink_skb_flags |= NETLINK_SKB_DST; |
1da177e4 | 1819 | } else { |
15e47304 | 1820 | dst_portid = nlk->dst_portid; |
d629b836 | 1821 | dst_group = nlk->dst_group; |
1da177e4 LT |
1822 | } |
1823 | ||
da314c99 | 1824 | if (!nlk->bound) { |
1da177e4 LT |
1825 | err = netlink_autobind(sock); |
1826 | if (err) | |
1827 | goto out; | |
da314c99 HX |
1828 | } else { |
1829 | /* Ensure nlk is hashed and visible. */ | |
1830 | smp_rmb(); | |
1da177e4 LT |
1831 | } |
1832 | ||
1833 | err = -EMSGSIZE; | |
1834 | if (len > sk->sk_sndbuf - 32) | |
1835 | goto out; | |
1836 | err = -ENOBUFS; | |
3a36515f | 1837 | skb = netlink_alloc_large_skb(len, dst_group); |
6ac552fd | 1838 | if (skb == NULL) |
1da177e4 LT |
1839 | goto out; |
1840 | ||
15e47304 | 1841 | NETLINK_CB(skb).portid = nlk->portid; |
d629b836 | 1842 | NETLINK_CB(skb).dst_group = dst_group; |
7cc05662 | 1843 | NETLINK_CB(skb).creds = scm.creds; |
2d7a85f4 | 1844 | NETLINK_CB(skb).flags = netlink_skb_flags; |
1da177e4 | 1845 | |
1da177e4 | 1846 | err = -EFAULT; |
6ce8e9ce | 1847 | if (memcpy_from_msg(skb_put(skb, len), msg, len)) { |
1da177e4 LT |
1848 | kfree_skb(skb); |
1849 | goto out; | |
1850 | } | |
1851 | ||
1852 | err = security_netlink_send(sk, skb); | |
1853 | if (err) { | |
1854 | kfree_skb(skb); | |
1855 | goto out; | |
1856 | } | |
1857 | ||
d629b836 | 1858 | if (dst_group) { |
63354797 | 1859 | refcount_inc(&skb->users); |
15e47304 | 1860 | netlink_broadcast(sk, skb, dst_portid, dst_group, GFP_KERNEL); |
1da177e4 | 1861 | } |
15e47304 | 1862 | err = netlink_unicast(sk, skb, dst_portid, msg->msg_flags&MSG_DONTWAIT); |
1da177e4 LT |
1863 | |
1864 | out: | |
7cc05662 | 1865 | scm_destroy(&scm); |
1da177e4 LT |
1866 | return err; |
1867 | } | |
1868 | ||
1b784140 | 1869 | static int netlink_recvmsg(struct socket *sock, struct msghdr *msg, size_t len, |
1da177e4 LT |
1870 | int flags) |
1871 | { | |
1da177e4 LT |
1872 | struct scm_cookie scm; |
1873 | struct sock *sk = sock->sk; | |
1874 | struct netlink_sock *nlk = nlk_sk(sk); | |
1875 | int noblock = flags&MSG_DONTWAIT; | |
1876 | size_t copied; | |
68d6ac6d | 1877 | struct sk_buff *skb, *data_skb; |
b44d211e | 1878 | int err, ret; |
1da177e4 LT |
1879 | |
1880 | if (flags&MSG_OOB) | |
1881 | return -EOPNOTSUPP; | |
1882 | ||
1883 | copied = 0; | |
1884 | ||
6ac552fd PM |
1885 | skb = skb_recv_datagram(sk, flags, noblock, &err); |
1886 | if (skb == NULL) | |
1da177e4 LT |
1887 | goto out; |
1888 | ||
68d6ac6d JB |
1889 | data_skb = skb; |
1890 | ||
1dacc76d JB |
1891 | #ifdef CONFIG_COMPAT_NETLINK_MESSAGES |
1892 | if (unlikely(skb_shinfo(skb)->frag_list)) { | |
1dacc76d | 1893 | /* |
68d6ac6d JB |
1894 | * If this skb has a frag_list, then here that means that we |
1895 | * will have to use the frag_list skb's data for compat tasks | |
1896 | * and the regular skb's data for normal (non-compat) tasks. | |
1dacc76d | 1897 | * |
68d6ac6d JB |
1898 | * If we need to send the compat skb, assign it to the |
1899 | * 'data_skb' variable so that it will be used below for data | |
1900 | * copying. We keep 'skb' for everything else, including | |
1901 | * freeing both later. | |
1dacc76d | 1902 | */ |
68d6ac6d JB |
1903 | if (flags & MSG_CMSG_COMPAT) |
1904 | data_skb = skb_shinfo(skb)->frag_list; | |
1dacc76d JB |
1905 | } |
1906 | #endif | |
1907 | ||
9063e21f ED |
1908 | /* Record the max length of recvmsg() calls for future allocations */ |
1909 | nlk->max_recvmsg_len = max(nlk->max_recvmsg_len, len); | |
1910 | nlk->max_recvmsg_len = min_t(size_t, nlk->max_recvmsg_len, | |
d35c99ff | 1911 | SKB_WITH_OVERHEAD(32768)); |
9063e21f | 1912 | |
68d6ac6d | 1913 | copied = data_skb->len; |
1da177e4 LT |
1914 | if (len < copied) { |
1915 | msg->msg_flags |= MSG_TRUNC; | |
1916 | copied = len; | |
1917 | } | |
1918 | ||
68d6ac6d | 1919 | skb_reset_transport_header(data_skb); |
51f3d02b | 1920 | err = skb_copy_datagram_msg(data_skb, 0, msg, copied); |
1da177e4 LT |
1921 | |
1922 | if (msg->msg_name) { | |
342dfc30 | 1923 | DECLARE_SOCKADDR(struct sockaddr_nl *, addr, msg->msg_name); |
1da177e4 LT |
1924 | addr->nl_family = AF_NETLINK; |
1925 | addr->nl_pad = 0; | |
15e47304 | 1926 | addr->nl_pid = NETLINK_CB(skb).portid; |
d629b836 | 1927 | addr->nl_groups = netlink_group_mask(NETLINK_CB(skb).dst_group); |
1da177e4 LT |
1928 | msg->msg_namelen = sizeof(*addr); |
1929 | } | |
1930 | ||
cc3a572f | 1931 | if (nlk->flags & NETLINK_F_RECV_PKTINFO) |
cc9a06cd | 1932 | netlink_cmsg_recv_pktinfo(msg, skb); |
59324cf3 ND |
1933 | if (nlk->flags & NETLINK_F_LISTEN_ALL_NSID) |
1934 | netlink_cmsg_listen_all_nsid(sk, msg, skb); | |
cc9a06cd | 1935 | |
7cc05662 CH |
1936 | memset(&scm, 0, sizeof(scm)); |
1937 | scm.creds = *NETLINK_CREDS(skb); | |
188ccb55 | 1938 | if (flags & MSG_TRUNC) |
68d6ac6d | 1939 | copied = data_skb->len; |
daa3766e | 1940 | |
1da177e4 LT |
1941 | skb_free_datagram(sk, skb); |
1942 | ||
16b304f3 PS |
1943 | if (nlk->cb_running && |
1944 | atomic_read(&sk->sk_rmem_alloc) <= sk->sk_rcvbuf / 2) { | |
b44d211e AV |
1945 | ret = netlink_dump(sk); |
1946 | if (ret) { | |
ac30ef83 | 1947 | sk->sk_err = -ret; |
b44d211e AV |
1948 | sk->sk_error_report(sk); |
1949 | } | |
1950 | } | |
1da177e4 | 1951 | |
7cc05662 | 1952 | scm_recv(sock, msg, &scm, flags); |
1da177e4 LT |
1953 | out: |
1954 | netlink_rcv_wake(sk); | |
1955 | return err ? : copied; | |
1956 | } | |
1957 | ||
676d2369 | 1958 | static void netlink_data_ready(struct sock *sk) |
1da177e4 | 1959 | { |
cd40b7d3 | 1960 | BUG(); |
1da177e4 LT |
1961 | } |
1962 | ||
1963 | /* | |
746fac4d | 1964 | * We export these functions to other modules. They provide a |
1da177e4 LT |
1965 | * complete set of kernel non-blocking support for message |
1966 | * queueing. | |
1967 | */ | |
1968 | ||
1969 | struct sock * | |
9f00d977 PNA |
1970 | __netlink_kernel_create(struct net *net, int unit, struct module *module, |
1971 | struct netlink_kernel_cfg *cfg) | |
1da177e4 LT |
1972 | { |
1973 | struct socket *sock; | |
1974 | struct sock *sk; | |
77247bbb | 1975 | struct netlink_sock *nlk; |
5c398dc8 | 1976 | struct listeners *listeners = NULL; |
a31f2d17 PNA |
1977 | struct mutex *cb_mutex = cfg ? cfg->cb_mutex : NULL; |
1978 | unsigned int groups; | |
1da177e4 | 1979 | |
fab2caf6 | 1980 | BUG_ON(!nl_table); |
1da177e4 | 1981 | |
6ac552fd | 1982 | if (unit < 0 || unit >= MAX_LINKS) |
1da177e4 LT |
1983 | return NULL; |
1984 | ||
1985 | if (sock_create_lite(PF_NETLINK, SOCK_DGRAM, unit, &sock)) | |
1986 | return NULL; | |
13d3078e EB |
1987 | |
1988 | if (__netlink_create(net, sock, cb_mutex, unit, 1) < 0) | |
23fe1866 PE |
1989 | goto out_sock_release_nosk; |
1990 | ||
1991 | sk = sock->sk; | |
4fdb3bb7 | 1992 | |
a31f2d17 | 1993 | if (!cfg || cfg->groups < 32) |
4277a083 | 1994 | groups = 32; |
a31f2d17 PNA |
1995 | else |
1996 | groups = cfg->groups; | |
4277a083 | 1997 | |
5c398dc8 | 1998 | listeners = kzalloc(sizeof(*listeners) + NLGRPSZ(groups), GFP_KERNEL); |
4277a083 PM |
1999 | if (!listeners) |
2000 | goto out_sock_release; | |
2001 | ||
1da177e4 | 2002 | sk->sk_data_ready = netlink_data_ready; |
a31f2d17 PNA |
2003 | if (cfg && cfg->input) |
2004 | nlk_sk(sk)->netlink_rcv = cfg->input; | |
1da177e4 | 2005 | |
8ea65f4a | 2006 | if (netlink_insert(sk, 0)) |
77247bbb | 2007 | goto out_sock_release; |
4fdb3bb7 | 2008 | |
77247bbb | 2009 | nlk = nlk_sk(sk); |
cc3a572f | 2010 | nlk->flags |= NETLINK_F_KERNEL_SOCKET; |
4fdb3bb7 | 2011 | |
4fdb3bb7 | 2012 | netlink_table_grab(); |
b4b51029 EB |
2013 | if (!nl_table[unit].registered) { |
2014 | nl_table[unit].groups = groups; | |
5c398dc8 | 2015 | rcu_assign_pointer(nl_table[unit].listeners, listeners); |
b4b51029 EB |
2016 | nl_table[unit].cb_mutex = cb_mutex; |
2017 | nl_table[unit].module = module; | |
9785e10a PNA |
2018 | if (cfg) { |
2019 | nl_table[unit].bind = cfg->bind; | |
6251edd9 | 2020 | nl_table[unit].unbind = cfg->unbind; |
9785e10a | 2021 | nl_table[unit].flags = cfg->flags; |
da12c90e G |
2022 | if (cfg->compare) |
2023 | nl_table[unit].compare = cfg->compare; | |
9785e10a | 2024 | } |
b4b51029 | 2025 | nl_table[unit].registered = 1; |
f937f1f4 JJ |
2026 | } else { |
2027 | kfree(listeners); | |
869e58f8 | 2028 | nl_table[unit].registered++; |
b4b51029 | 2029 | } |
4fdb3bb7 | 2030 | netlink_table_ungrab(); |
77247bbb PM |
2031 | return sk; |
2032 | ||
4fdb3bb7 | 2033 | out_sock_release: |
4277a083 | 2034 | kfree(listeners); |
9dfbec1f | 2035 | netlink_kernel_release(sk); |
23fe1866 PE |
2036 | return NULL; |
2037 | ||
2038 | out_sock_release_nosk: | |
4fdb3bb7 | 2039 | sock_release(sock); |
77247bbb | 2040 | return NULL; |
1da177e4 | 2041 | } |
9f00d977 | 2042 | EXPORT_SYMBOL(__netlink_kernel_create); |
b7c6ba6e DL |
2043 | |
2044 | void | |
2045 | netlink_kernel_release(struct sock *sk) | |
2046 | { | |
13d3078e EB |
2047 | if (sk == NULL || sk->sk_socket == NULL) |
2048 | return; | |
2049 | ||
2050 | sock_release(sk->sk_socket); | |
b7c6ba6e DL |
2051 | } |
2052 | EXPORT_SYMBOL(netlink_kernel_release); | |
2053 | ||
d136f1bd | 2054 | int __netlink_change_ngroups(struct sock *sk, unsigned int groups) |
b4ff4f04 | 2055 | { |
5c398dc8 | 2056 | struct listeners *new, *old; |
b4ff4f04 | 2057 | struct netlink_table *tbl = &nl_table[sk->sk_protocol]; |
b4ff4f04 JB |
2058 | |
2059 | if (groups < 32) | |
2060 | groups = 32; | |
2061 | ||
b4ff4f04 | 2062 | if (NLGRPSZ(tbl->groups) < NLGRPSZ(groups)) { |
5c398dc8 ED |
2063 | new = kzalloc(sizeof(*new) + NLGRPSZ(groups), GFP_ATOMIC); |
2064 | if (!new) | |
d136f1bd | 2065 | return -ENOMEM; |
6d772ac5 | 2066 | old = nl_deref_protected(tbl->listeners); |
5c398dc8 ED |
2067 | memcpy(new->masks, old->masks, NLGRPSZ(tbl->groups)); |
2068 | rcu_assign_pointer(tbl->listeners, new); | |
2069 | ||
37b6b935 | 2070 | kfree_rcu(old, rcu); |
b4ff4f04 JB |
2071 | } |
2072 | tbl->groups = groups; | |
2073 | ||
d136f1bd JB |
2074 | return 0; |
2075 | } | |
2076 | ||
2077 | /** | |
2078 | * netlink_change_ngroups - change number of multicast groups | |
2079 | * | |
2080 | * This changes the number of multicast groups that are available | |
2081 | * on a certain netlink family. Note that it is not possible to | |
2082 | * change the number of groups to below 32. Also note that it does | |
2083 | * not implicitly call netlink_clear_multicast_users() when the | |
2084 | * number of groups is reduced. | |
2085 | * | |
2086 | * @sk: The kernel netlink socket, as returned by netlink_kernel_create(). | |
2087 | * @groups: The new number of groups. | |
2088 | */ | |
2089 | int netlink_change_ngroups(struct sock *sk, unsigned int groups) | |
2090 | { | |
2091 | int err; | |
2092 | ||
2093 | netlink_table_grab(); | |
2094 | err = __netlink_change_ngroups(sk, groups); | |
b4ff4f04 | 2095 | netlink_table_ungrab(); |
d136f1bd | 2096 | |
b4ff4f04 JB |
2097 | return err; |
2098 | } | |
b4ff4f04 | 2099 | |
b8273570 JB |
2100 | void __netlink_clear_multicast_users(struct sock *ksk, unsigned int group) |
2101 | { | |
2102 | struct sock *sk; | |
b8273570 JB |
2103 | struct netlink_table *tbl = &nl_table[ksk->sk_protocol]; |
2104 | ||
b67bfe0d | 2105 | sk_for_each_bound(sk, &tbl->mc_list) |
b8273570 JB |
2106 | netlink_update_socket_mc(nlk_sk(sk), group, 0); |
2107 | } | |
2108 | ||
a46621a3 | 2109 | struct nlmsghdr * |
15e47304 | 2110 | __nlmsg_put(struct sk_buff *skb, u32 portid, u32 seq, int type, int len, int flags) |
a46621a3 DV |
2111 | { |
2112 | struct nlmsghdr *nlh; | |
573ce260 | 2113 | int size = nlmsg_msg_size(len); |
a46621a3 | 2114 | |
4df864c1 | 2115 | nlh = skb_put(skb, NLMSG_ALIGN(size)); |
a46621a3 DV |
2116 | nlh->nlmsg_type = type; |
2117 | nlh->nlmsg_len = size; | |
2118 | nlh->nlmsg_flags = flags; | |
15e47304 | 2119 | nlh->nlmsg_pid = portid; |
a46621a3 DV |
2120 | nlh->nlmsg_seq = seq; |
2121 | if (!__builtin_constant_p(size) || NLMSG_ALIGN(size) - size != 0) | |
573ce260 | 2122 | memset(nlmsg_data(nlh) + len, 0, NLMSG_ALIGN(size) - size); |
a46621a3 DV |
2123 | return nlh; |
2124 | } | |
2125 | EXPORT_SYMBOL(__nlmsg_put); | |
2126 | ||
1da177e4 LT |
2127 | /* |
2128 | * It looks a bit ugly. | |
2129 | * It would be better to create kernel thread. | |
2130 | */ | |
2131 | ||
2132 | static int netlink_dump(struct sock *sk) | |
2133 | { | |
2134 | struct netlink_sock *nlk = nlk_sk(sk); | |
2135 | struct netlink_callback *cb; | |
c7ac8679 | 2136 | struct sk_buff *skb = NULL; |
1da177e4 | 2137 | struct nlmsghdr *nlh; |
92964c79 | 2138 | struct module *module; |
bf8b79e4 | 2139 | int len, err = -ENOBUFS; |
db65a3aa | 2140 | int alloc_min_size; |
c7ac8679 | 2141 | int alloc_size; |
1da177e4 | 2142 | |
af65bdfc | 2143 | mutex_lock(nlk->cb_mutex); |
16b304f3 | 2144 | if (!nlk->cb_running) { |
bf8b79e4 TG |
2145 | err = -EINVAL; |
2146 | goto errout_skb; | |
1da177e4 LT |
2147 | } |
2148 | ||
d1b4c689 | 2149 | if (atomic_read(&sk->sk_rmem_alloc) >= sk->sk_rcvbuf) |
f9c22888 | 2150 | goto errout_skb; |
9063e21f ED |
2151 | |
2152 | /* NLMSG_GOODSIZE is small to avoid high order allocations being | |
2153 | * required, but it makes sense to _attempt_ a 16K bytes allocation | |
2154 | * to reduce number of system calls on dump operations, if user | |
2155 | * ever provided a big enough buffer. | |
2156 | */ | |
db65a3aa AR |
2157 | cb = &nlk->cb; |
2158 | alloc_min_size = max_t(int, cb->min_dump_alloc, NLMSG_GOODSIZE); | |
2159 | ||
2160 | if (alloc_min_size < nlk->max_recvmsg_len) { | |
2161 | alloc_size = nlk->max_recvmsg_len; | |
d35c99ff ED |
2162 | skb = alloc_skb(alloc_size, |
2163 | (GFP_KERNEL & ~__GFP_DIRECT_RECLAIM) | | |
2164 | __GFP_NOWARN | __GFP_NORETRY); | |
9063e21f | 2165 | } |
db65a3aa AR |
2166 | if (!skb) { |
2167 | alloc_size = alloc_min_size; | |
c5b0db32 | 2168 | skb = alloc_skb(alloc_size, GFP_KERNEL); |
db65a3aa | 2169 | } |
c7ac8679 | 2170 | if (!skb) |
c63d6ea3 | 2171 | goto errout_skb; |
db65a3aa AR |
2172 | |
2173 | /* Trim skb to allocated size. User is expected to provide buffer as | |
2174 | * large as max(min_dump_alloc, 16KiB (mac_recvmsg_len capped at | |
2175 | * netlink_recvmsg())). dump will pack as many smaller messages as | |
2176 | * could fit within the allocated skb. skb is typically allocated | |
2177 | * with larger space than required (could be as much as near 2x the | |
2178 | * requested size with align to next power of 2 approach). Allowing | |
2179 | * dump to use the excess space makes it difficult for a user to have a | |
2180 | * reasonable static buffer based on the expected largest dump of a | |
2181 | * single netdev. The outcome is MSG_TRUNC error. | |
2182 | */ | |
d1b4c689 | 2183 | skb_reserve(skb, skb_tailroom(skb) - alloc_size); |
f9c22888 | 2184 | netlink_skb_set_owner_r(skb, sk); |
c7ac8679 | 2185 | |
1da177e4 LT |
2186 | len = cb->dump(skb, cb); |
2187 | ||
2188 | if (len > 0) { | |
af65bdfc | 2189 | mutex_unlock(nlk->cb_mutex); |
b1153f29 SH |
2190 | |
2191 | if (sk_filter(sk, skb)) | |
2192 | kfree_skb(skb); | |
4a7e7c2a ED |
2193 | else |
2194 | __netlink_sendskb(sk, skb); | |
1da177e4 LT |
2195 | return 0; |
2196 | } | |
2197 | ||
bf8b79e4 TG |
2198 | nlh = nlmsg_put_answer(skb, cb, NLMSG_DONE, sizeof(len), NLM_F_MULTI); |
2199 | if (!nlh) | |
2200 | goto errout_skb; | |
2201 | ||
670dc283 JB |
2202 | nl_dump_check_consistent(cb, nlh); |
2203 | ||
bf8b79e4 TG |
2204 | memcpy(nlmsg_data(nlh), &len, sizeof(len)); |
2205 | ||
b1153f29 SH |
2206 | if (sk_filter(sk, skb)) |
2207 | kfree_skb(skb); | |
4a7e7c2a ED |
2208 | else |
2209 | __netlink_sendskb(sk, skb); | |
1da177e4 | 2210 | |
a8f74b22 TG |
2211 | if (cb->done) |
2212 | cb->done(cb); | |
1da177e4 | 2213 | |
16b304f3 | 2214 | nlk->cb_running = false; |
92964c79 HX |
2215 | module = cb->module; |
2216 | skb = cb->skb; | |
16b304f3 | 2217 | mutex_unlock(nlk->cb_mutex); |
92964c79 HX |
2218 | module_put(module); |
2219 | consume_skb(skb); | |
1da177e4 | 2220 | return 0; |
1797754e | 2221 | |
bf8b79e4 | 2222 | errout_skb: |
af65bdfc | 2223 | mutex_unlock(nlk->cb_mutex); |
bf8b79e4 | 2224 | kfree_skb(skb); |
bf8b79e4 | 2225 | return err; |
1da177e4 LT |
2226 | } |
2227 | ||
6dc878a8 G |
2228 | int __netlink_dump_start(struct sock *ssk, struct sk_buff *skb, |
2229 | const struct nlmsghdr *nlh, | |
2230 | struct netlink_dump_control *control) | |
1da177e4 LT |
2231 | { |
2232 | struct netlink_callback *cb; | |
2233 | struct sock *sk; | |
2234 | struct netlink_sock *nlk; | |
b44d211e | 2235 | int ret; |
1da177e4 | 2236 | |
63354797 | 2237 | refcount_inc(&skb->users); |
f9c22888 | 2238 | |
15e47304 | 2239 | sk = netlink_lookup(sock_net(ssk), ssk->sk_protocol, NETLINK_CB(skb).portid); |
1da177e4 | 2240 | if (sk == NULL) { |
16b304f3 PS |
2241 | ret = -ECONNREFUSED; |
2242 | goto error_free; | |
1da177e4 | 2243 | } |
6dc878a8 | 2244 | |
16b304f3 | 2245 | nlk = nlk_sk(sk); |
af65bdfc | 2246 | mutex_lock(nlk->cb_mutex); |
6dc878a8 | 2247 | /* A dump is in progress... */ |
16b304f3 | 2248 | if (nlk->cb_running) { |
6dc878a8 | 2249 | ret = -EBUSY; |
16b304f3 | 2250 | goto error_unlock; |
1da177e4 | 2251 | } |
6dc878a8 | 2252 | /* add reference of module which cb->dump belongs to */ |
16b304f3 | 2253 | if (!try_module_get(control->module)) { |
6dc878a8 | 2254 | ret = -EPROTONOSUPPORT; |
16b304f3 | 2255 | goto error_unlock; |
6dc878a8 G |
2256 | } |
2257 | ||
16b304f3 PS |
2258 | cb = &nlk->cb; |
2259 | memset(cb, 0, sizeof(*cb)); | |
fc9e50f5 | 2260 | cb->start = control->start; |
16b304f3 PS |
2261 | cb->dump = control->dump; |
2262 | cb->done = control->done; | |
2263 | cb->nlh = nlh; | |
2264 | cb->data = control->data; | |
2265 | cb->module = control->module; | |
2266 | cb->min_dump_alloc = control->min_dump_alloc; | |
2267 | cb->skb = skb; | |
2268 | ||
41c87425 JD |
2269 | if (cb->start) { |
2270 | ret = cb->start(cb); | |
2271 | if (ret) | |
2272 | goto error_unlock; | |
2273 | } | |
2274 | ||
16b304f3 PS |
2275 | nlk->cb_running = true; |
2276 | ||
af65bdfc | 2277 | mutex_unlock(nlk->cb_mutex); |
1da177e4 | 2278 | |
41c87425 | 2279 | ret = netlink_dump(sk); |
fc9e50f5 | 2280 | |
1da177e4 | 2281 | sock_put(sk); |
5c58298c | 2282 | |
b44d211e AV |
2283 | if (ret) |
2284 | return ret; | |
2285 | ||
5c58298c DL |
2286 | /* We successfully started a dump, by returning -EINTR we |
2287 | * signal not to send ACK even if it was requested. | |
2288 | */ | |
2289 | return -EINTR; | |
16b304f3 PS |
2290 | |
2291 | error_unlock: | |
2292 | sock_put(sk); | |
2293 | mutex_unlock(nlk->cb_mutex); | |
2294 | error_free: | |
2295 | kfree_skb(skb); | |
2296 | return ret; | |
1da177e4 | 2297 | } |
6dc878a8 | 2298 | EXPORT_SYMBOL(__netlink_dump_start); |
1da177e4 | 2299 | |
2d4bc933 JB |
2300 | void netlink_ack(struct sk_buff *in_skb, struct nlmsghdr *nlh, int err, |
2301 | const struct netlink_ext_ack *extack) | |
1da177e4 LT |
2302 | { |
2303 | struct sk_buff *skb; | |
2304 | struct nlmsghdr *rep; | |
2305 | struct nlmsgerr *errmsg; | |
339bf98f | 2306 | size_t payload = sizeof(*errmsg); |
2d4bc933 | 2307 | size_t tlvlen = 0; |
0a6a3a23 | 2308 | struct netlink_sock *nlk = nlk_sk(NETLINK_CB(in_skb).sk); |
2d4bc933 | 2309 | unsigned int flags = 0; |
48044eb4 | 2310 | bool nlk_has_extack = nlk->flags & NETLINK_F_EXT_ACK; |
1da177e4 | 2311 | |
0a6a3a23 | 2312 | /* Error messages get the original request appened, unless the user |
2d4bc933 JB |
2313 | * requests to cap the error message, and get extra error data if |
2314 | * requested. | |
0a6a3a23 | 2315 | */ |
2d4bc933 JB |
2316 | if (err) { |
2317 | if (!(nlk->flags & NETLINK_F_CAP_ACK)) | |
2318 | payload += nlmsg_len(nlh); | |
2319 | else | |
2320 | flags |= NLM_F_CAPPED; | |
48044eb4 | 2321 | if (nlk_has_extack && extack) { |
2d4bc933 JB |
2322 | if (extack->_msg) |
2323 | tlvlen += nla_total_size(strlen(extack->_msg) + 1); | |
2324 | if (extack->bad_attr) | |
2325 | tlvlen += nla_total_size(sizeof(u32)); | |
2326 | } | |
2327 | } else { | |
2328 | flags |= NLM_F_CAPPED; | |
ba0dc5f6 | 2329 | |
48044eb4 | 2330 | if (nlk_has_extack && extack && extack->cookie_len) |
ba0dc5f6 | 2331 | tlvlen += nla_total_size(extack->cookie_len); |
2d4bc933 | 2332 | } |
1da177e4 | 2333 | |
2d4bc933 JB |
2334 | if (tlvlen) |
2335 | flags |= NLM_F_ACK_TLVS; | |
2336 | ||
2337 | skb = nlmsg_new(payload + tlvlen, GFP_KERNEL); | |
1da177e4 LT |
2338 | if (!skb) { |
2339 | struct sock *sk; | |
2340 | ||
3b1e0a65 | 2341 | sk = netlink_lookup(sock_net(in_skb->sk), |
b4b51029 | 2342 | in_skb->sk->sk_protocol, |
15e47304 | 2343 | NETLINK_CB(in_skb).portid); |
1da177e4 LT |
2344 | if (sk) { |
2345 | sk->sk_err = ENOBUFS; | |
2346 | sk->sk_error_report(sk); | |
2347 | sock_put(sk); | |
2348 | } | |
2349 | return; | |
2350 | } | |
2351 | ||
15e47304 | 2352 | rep = __nlmsg_put(skb, NETLINK_CB(in_skb).portid, nlh->nlmsg_seq, |
2d4bc933 | 2353 | NLMSG_ERROR, payload, flags); |
bf8b79e4 | 2354 | errmsg = nlmsg_data(rep); |
1da177e4 | 2355 | errmsg->error = err; |
0a6a3a23 | 2356 | memcpy(&errmsg->msg, nlh, payload > sizeof(*errmsg) ? nlh->nlmsg_len : sizeof(*nlh)); |
2d4bc933 | 2357 | |
48044eb4 | 2358 | if (nlk_has_extack && extack) { |
ba0dc5f6 JB |
2359 | if (err) { |
2360 | if (extack->_msg) | |
2361 | WARN_ON(nla_put_string(skb, NLMSGERR_ATTR_MSG, | |
2362 | extack->_msg)); | |
2363 | if (extack->bad_attr && | |
2364 | !WARN_ON((u8 *)extack->bad_attr < in_skb->data || | |
2365 | (u8 *)extack->bad_attr >= in_skb->data + | |
2366 | in_skb->len)) | |
2367 | WARN_ON(nla_put_u32(skb, NLMSGERR_ATTR_OFFS, | |
2368 | (u8 *)extack->bad_attr - | |
2369 | in_skb->data)); | |
2370 | } else { | |
2371 | if (extack->cookie_len) | |
2372 | WARN_ON(nla_put(skb, NLMSGERR_ATTR_COOKIE, | |
2373 | extack->cookie_len, | |
2374 | extack->cookie)); | |
2375 | } | |
2d4bc933 JB |
2376 | } |
2377 | ||
2378 | nlmsg_end(skb, rep); | |
2379 | ||
15e47304 | 2380 | netlink_unicast(in_skb->sk, skb, NETLINK_CB(in_skb).portid, MSG_DONTWAIT); |
1da177e4 | 2381 | } |
6ac552fd | 2382 | EXPORT_SYMBOL(netlink_ack); |
1da177e4 | 2383 | |
cd40b7d3 | 2384 | int netlink_rcv_skb(struct sk_buff *skb, int (*cb)(struct sk_buff *, |
2d4bc933 JB |
2385 | struct nlmsghdr *, |
2386 | struct netlink_ext_ack *)) | |
82ace47a | 2387 | { |
2d4bc933 | 2388 | struct netlink_ext_ack extack = {}; |
82ace47a TG |
2389 | struct nlmsghdr *nlh; |
2390 | int err; | |
2391 | ||
2392 | while (skb->len >= nlmsg_total_size(0)) { | |
cd40b7d3 DL |
2393 | int msglen; |
2394 | ||
b529ccf2 | 2395 | nlh = nlmsg_hdr(skb); |
d35b6856 | 2396 | err = 0; |
82ace47a | 2397 | |
ad8e4b75 | 2398 | if (nlh->nlmsg_len < NLMSG_HDRLEN || skb->len < nlh->nlmsg_len) |
82ace47a TG |
2399 | return 0; |
2400 | ||
d35b6856 TG |
2401 | /* Only requests are handled by the kernel */ |
2402 | if (!(nlh->nlmsg_flags & NLM_F_REQUEST)) | |
5c58298c | 2403 | goto ack; |
45e7ae7f TG |
2404 | |
2405 | /* Skip control messages */ | |
2406 | if (nlh->nlmsg_type < NLMSG_MIN_TYPE) | |
5c58298c | 2407 | goto ack; |
d35b6856 | 2408 | |
2d4bc933 | 2409 | err = cb(skb, nlh, &extack); |
5c58298c DL |
2410 | if (err == -EINTR) |
2411 | goto skip; | |
2412 | ||
2413 | ack: | |
d35b6856 | 2414 | if (nlh->nlmsg_flags & NLM_F_ACK || err) |
2d4bc933 | 2415 | netlink_ack(skb, nlh, err, &extack); |
82ace47a | 2416 | |
5c58298c | 2417 | skip: |
6ac552fd | 2418 | msglen = NLMSG_ALIGN(nlh->nlmsg_len); |
cd40b7d3 DL |
2419 | if (msglen > skb->len) |
2420 | msglen = skb->len; | |
2421 | skb_pull(skb, msglen); | |
82ace47a TG |
2422 | } |
2423 | ||
2424 | return 0; | |
2425 | } | |
6ac552fd | 2426 | EXPORT_SYMBOL(netlink_rcv_skb); |
82ace47a | 2427 | |
d387f6ad TG |
2428 | /** |
2429 | * nlmsg_notify - send a notification netlink message | |
2430 | * @sk: netlink socket to use | |
2431 | * @skb: notification message | |
15e47304 | 2432 | * @portid: destination netlink portid for reports or 0 |
d387f6ad TG |
2433 | * @group: destination multicast group or 0 |
2434 | * @report: 1 to report back, 0 to disable | |
2435 | * @flags: allocation flags | |
2436 | */ | |
15e47304 | 2437 | int nlmsg_notify(struct sock *sk, struct sk_buff *skb, u32 portid, |
d387f6ad TG |
2438 | unsigned int group, int report, gfp_t flags) |
2439 | { | |
2440 | int err = 0; | |
2441 | ||
2442 | if (group) { | |
15e47304 | 2443 | int exclude_portid = 0; |
d387f6ad TG |
2444 | |
2445 | if (report) { | |
63354797 | 2446 | refcount_inc(&skb->users); |
15e47304 | 2447 | exclude_portid = portid; |
d387f6ad TG |
2448 | } |
2449 | ||
1ce85fe4 PNA |
2450 | /* errors reported via destination sk->sk_err, but propagate |
2451 | * delivery errors if NETLINK_BROADCAST_ERROR flag is set */ | |
15e47304 | 2452 | err = nlmsg_multicast(sk, skb, exclude_portid, group, flags); |
d387f6ad TG |
2453 | } |
2454 | ||
1ce85fe4 PNA |
2455 | if (report) { |
2456 | int err2; | |
2457 | ||
15e47304 | 2458 | err2 = nlmsg_unicast(sk, skb, portid); |
1ce85fe4 PNA |
2459 | if (!err || err == -ESRCH) |
2460 | err = err2; | |
2461 | } | |
d387f6ad TG |
2462 | |
2463 | return err; | |
2464 | } | |
6ac552fd | 2465 | EXPORT_SYMBOL(nlmsg_notify); |
d387f6ad | 2466 | |
1da177e4 LT |
2467 | #ifdef CONFIG_PROC_FS |
2468 | struct nl_seq_iter { | |
e372c414 | 2469 | struct seq_net_private p; |
56d28b1e | 2470 | struct rhashtable_iter hti; |
1da177e4 | 2471 | int link; |
1da177e4 LT |
2472 | }; |
2473 | ||
56d28b1e | 2474 | static int netlink_walk_start(struct nl_seq_iter *iter) |
1da177e4 | 2475 | { |
56d28b1e | 2476 | int err; |
1da177e4 | 2477 | |
8f6fd83c BC |
2478 | err = rhashtable_walk_init(&nl_table[iter->link].hash, &iter->hti, |
2479 | GFP_KERNEL); | |
56d28b1e HX |
2480 | if (err) { |
2481 | iter->link = MAX_LINKS; | |
2482 | return err; | |
1da177e4 | 2483 | } |
56d28b1e HX |
2484 | |
2485 | err = rhashtable_walk_start(&iter->hti); | |
2486 | return err == -EAGAIN ? 0 : err; | |
1da177e4 LT |
2487 | } |
2488 | ||
56d28b1e | 2489 | static void netlink_walk_stop(struct nl_seq_iter *iter) |
1da177e4 | 2490 | { |
56d28b1e HX |
2491 | rhashtable_walk_stop(&iter->hti); |
2492 | rhashtable_walk_exit(&iter->hti); | |
1da177e4 LT |
2493 | } |
2494 | ||
56d28b1e | 2495 | static void *__netlink_seq_next(struct seq_file *seq) |
1da177e4 | 2496 | { |
56d28b1e | 2497 | struct nl_seq_iter *iter = seq->private; |
e341694e | 2498 | struct netlink_sock *nlk; |
1da177e4 | 2499 | |
56d28b1e HX |
2500 | do { |
2501 | for (;;) { | |
2502 | int err; | |
1da177e4 | 2503 | |
56d28b1e | 2504 | nlk = rhashtable_walk_next(&iter->hti); |
746fac4d | 2505 | |
56d28b1e HX |
2506 | if (IS_ERR(nlk)) { |
2507 | if (PTR_ERR(nlk) == -EAGAIN) | |
2508 | continue; | |
e341694e | 2509 | |
56d28b1e HX |
2510 | return nlk; |
2511 | } | |
1da177e4 | 2512 | |
56d28b1e HX |
2513 | if (nlk) |
2514 | break; | |
1da177e4 | 2515 | |
56d28b1e HX |
2516 | netlink_walk_stop(iter); |
2517 | if (++iter->link >= MAX_LINKS) | |
2518 | return NULL; | |
da12c90e | 2519 | |
56d28b1e HX |
2520 | err = netlink_walk_start(iter); |
2521 | if (err) | |
2522 | return ERR_PTR(err); | |
1da177e4 | 2523 | } |
56d28b1e | 2524 | } while (sock_net(&nlk->sk) != seq_file_net(seq)); |
1da177e4 | 2525 | |
56d28b1e HX |
2526 | return nlk; |
2527 | } | |
1da177e4 | 2528 | |
56d28b1e HX |
2529 | static void *netlink_seq_start(struct seq_file *seq, loff_t *posp) |
2530 | { | |
2531 | struct nl_seq_iter *iter = seq->private; | |
2532 | void *obj = SEQ_START_TOKEN; | |
2533 | loff_t pos; | |
2534 | int err; | |
2535 | ||
2536 | iter->link = 0; | |
2537 | ||
2538 | err = netlink_walk_start(iter); | |
2539 | if (err) | |
2540 | return ERR_PTR(err); | |
2541 | ||
2542 | for (pos = *posp; pos && obj && !IS_ERR(obj); pos--) | |
2543 | obj = __netlink_seq_next(seq); | |
2544 | ||
2545 | return obj; | |
2546 | } | |
2547 | ||
2548 | static void *netlink_seq_next(struct seq_file *seq, void *v, loff_t *pos) | |
2549 | { | |
2550 | ++*pos; | |
2551 | return __netlink_seq_next(seq); | |
1da177e4 LT |
2552 | } |
2553 | ||
2554 | static void netlink_seq_stop(struct seq_file *seq, void *v) | |
2555 | { | |
56d28b1e HX |
2556 | struct nl_seq_iter *iter = seq->private; |
2557 | ||
2558 | if (iter->link >= MAX_LINKS) | |
2559 | return; | |
2560 | ||
2561 | netlink_walk_stop(iter); | |
1da177e4 LT |
2562 | } |
2563 | ||
2564 | ||
2565 | static int netlink_seq_show(struct seq_file *seq, void *v) | |
2566 | { | |
658cb354 | 2567 | if (v == SEQ_START_TOKEN) { |
1da177e4 LT |
2568 | seq_puts(seq, |
2569 | "sk Eth Pid Groups " | |
cf0aa4e0 | 2570 | "Rmem Wmem Dump Locks Drops Inode\n"); |
658cb354 | 2571 | } else { |
1da177e4 LT |
2572 | struct sock *s = v; |
2573 | struct netlink_sock *nlk = nlk_sk(s); | |
2574 | ||
16b304f3 | 2575 | seq_printf(seq, "%pK %-3d %-6u %08x %-8d %-8d %d %-8d %-8d %-8lu\n", |
1da177e4 LT |
2576 | s, |
2577 | s->sk_protocol, | |
15e47304 | 2578 | nlk->portid, |
513c2500 | 2579 | nlk->groups ? (u32)nlk->groups[0] : 0, |
31e6d363 ED |
2580 | sk_rmem_alloc_get(s), |
2581 | sk_wmem_alloc_get(s), | |
16b304f3 | 2582 | nlk->cb_running, |
41c6d650 | 2583 | refcount_read(&s->sk_refcnt), |
cf0aa4e0 MY |
2584 | atomic_read(&s->sk_drops), |
2585 | sock_i_ino(s) | |
1da177e4 LT |
2586 | ); |
2587 | ||
2588 | } | |
2589 | return 0; | |
2590 | } | |
2591 | ||
56b3d975 | 2592 | static const struct seq_operations netlink_seq_ops = { |
1da177e4 LT |
2593 | .start = netlink_seq_start, |
2594 | .next = netlink_seq_next, | |
2595 | .stop = netlink_seq_stop, | |
2596 | .show = netlink_seq_show, | |
2597 | }; | |
2598 | ||
2599 | ||
2600 | static int netlink_seq_open(struct inode *inode, struct file *file) | |
2601 | { | |
e372c414 DL |
2602 | return seq_open_net(inode, file, &netlink_seq_ops, |
2603 | sizeof(struct nl_seq_iter)); | |
b4b51029 EB |
2604 | } |
2605 | ||
da7071d7 | 2606 | static const struct file_operations netlink_seq_fops = { |
1da177e4 LT |
2607 | .owner = THIS_MODULE, |
2608 | .open = netlink_seq_open, | |
2609 | .read = seq_read, | |
2610 | .llseek = seq_lseek, | |
e372c414 | 2611 | .release = seq_release_net, |
1da177e4 LT |
2612 | }; |
2613 | ||
2614 | #endif | |
2615 | ||
2616 | int netlink_register_notifier(struct notifier_block *nb) | |
2617 | { | |
efa172f4 | 2618 | return blocking_notifier_chain_register(&netlink_chain, nb); |
1da177e4 | 2619 | } |
6ac552fd | 2620 | EXPORT_SYMBOL(netlink_register_notifier); |
1da177e4 LT |
2621 | |
2622 | int netlink_unregister_notifier(struct notifier_block *nb) | |
2623 | { | |
efa172f4 | 2624 | return blocking_notifier_chain_unregister(&netlink_chain, nb); |
1da177e4 | 2625 | } |
6ac552fd | 2626 | EXPORT_SYMBOL(netlink_unregister_notifier); |
746fac4d | 2627 | |
90ddc4f0 | 2628 | static const struct proto_ops netlink_ops = { |
1da177e4 LT |
2629 | .family = PF_NETLINK, |
2630 | .owner = THIS_MODULE, | |
2631 | .release = netlink_release, | |
2632 | .bind = netlink_bind, | |
2633 | .connect = netlink_connect, | |
2634 | .socketpair = sock_no_socketpair, | |
2635 | .accept = sock_no_accept, | |
2636 | .getname = netlink_getname, | |
d1b4c689 | 2637 | .poll = datagram_poll, |
025c6818 | 2638 | .ioctl = netlink_ioctl, |
1da177e4 LT |
2639 | .listen = sock_no_listen, |
2640 | .shutdown = sock_no_shutdown, | |
9a4595bc PM |
2641 | .setsockopt = netlink_setsockopt, |
2642 | .getsockopt = netlink_getsockopt, | |
1da177e4 LT |
2643 | .sendmsg = netlink_sendmsg, |
2644 | .recvmsg = netlink_recvmsg, | |
d1b4c689 | 2645 | .mmap = sock_no_mmap, |
1da177e4 LT |
2646 | .sendpage = sock_no_sendpage, |
2647 | }; | |
2648 | ||
ec1b4cf7 | 2649 | static const struct net_proto_family netlink_family_ops = { |
1da177e4 LT |
2650 | .family = PF_NETLINK, |
2651 | .create = netlink_create, | |
2652 | .owner = THIS_MODULE, /* for consistency 8) */ | |
2653 | }; | |
2654 | ||
4665079c | 2655 | static int __net_init netlink_net_init(struct net *net) |
b4b51029 EB |
2656 | { |
2657 | #ifdef CONFIG_PROC_FS | |
d4beaa66 | 2658 | if (!proc_create("netlink", 0, net->proc_net, &netlink_seq_fops)) |
b4b51029 EB |
2659 | return -ENOMEM; |
2660 | #endif | |
2661 | return 0; | |
2662 | } | |
2663 | ||
4665079c | 2664 | static void __net_exit netlink_net_exit(struct net *net) |
b4b51029 EB |
2665 | { |
2666 | #ifdef CONFIG_PROC_FS | |
ece31ffd | 2667 | remove_proc_entry("netlink", net->proc_net); |
b4b51029 EB |
2668 | #endif |
2669 | } | |
2670 | ||
b963ea89 DM |
2671 | static void __init netlink_add_usersock_entry(void) |
2672 | { | |
5c398dc8 | 2673 | struct listeners *listeners; |
b963ea89 DM |
2674 | int groups = 32; |
2675 | ||
5c398dc8 | 2676 | listeners = kzalloc(sizeof(*listeners) + NLGRPSZ(groups), GFP_KERNEL); |
b963ea89 | 2677 | if (!listeners) |
5c398dc8 | 2678 | panic("netlink_add_usersock_entry: Cannot allocate listeners\n"); |
b963ea89 DM |
2679 | |
2680 | netlink_table_grab(); | |
2681 | ||
2682 | nl_table[NETLINK_USERSOCK].groups = groups; | |
5c398dc8 | 2683 | rcu_assign_pointer(nl_table[NETLINK_USERSOCK].listeners, listeners); |
b963ea89 DM |
2684 | nl_table[NETLINK_USERSOCK].module = THIS_MODULE; |
2685 | nl_table[NETLINK_USERSOCK].registered = 1; | |
9785e10a | 2686 | nl_table[NETLINK_USERSOCK].flags = NL_CFG_F_NONROOT_SEND; |
b963ea89 DM |
2687 | |
2688 | netlink_table_ungrab(); | |
2689 | } | |
2690 | ||
022cbae6 | 2691 | static struct pernet_operations __net_initdata netlink_net_ops = { |
b4b51029 EB |
2692 | .init = netlink_net_init, |
2693 | .exit = netlink_net_exit, | |
2694 | }; | |
2695 | ||
49f7b33e | 2696 | static inline u32 netlink_hash(const void *data, u32 len, u32 seed) |
c428ecd1 HX |
2697 | { |
2698 | const struct netlink_sock *nlk = data; | |
2699 | struct netlink_compare_arg arg; | |
2700 | ||
da314c99 | 2701 | netlink_compare_arg_init(&arg, sock_net(&nlk->sk), nlk->portid); |
11b58ba1 | 2702 | return jhash2((u32 *)&arg, netlink_compare_arg_len / sizeof(u32), seed); |
c428ecd1 HX |
2703 | } |
2704 | ||
2705 | static const struct rhashtable_params netlink_rhashtable_params = { | |
2706 | .head_offset = offsetof(struct netlink_sock, node), | |
2707 | .key_len = netlink_compare_arg_len, | |
c428ecd1 HX |
2708 | .obj_hashfn = netlink_hash, |
2709 | .obj_cmpfn = netlink_compare, | |
b5e2c150 | 2710 | .automatic_shrinking = true, |
c428ecd1 HX |
2711 | }; |
2712 | ||
1da177e4 LT |
2713 | static int __init netlink_proto_init(void) |
2714 | { | |
1da177e4 | 2715 | int i; |
1da177e4 LT |
2716 | int err = proto_register(&netlink_proto, 0); |
2717 | ||
2718 | if (err != 0) | |
2719 | goto out; | |
2720 | ||
fab25745 | 2721 | BUILD_BUG_ON(sizeof(struct netlink_skb_parms) > FIELD_SIZEOF(struct sk_buff, cb)); |
1da177e4 | 2722 | |
0da974f4 | 2723 | nl_table = kcalloc(MAX_LINKS, sizeof(*nl_table), GFP_KERNEL); |
fab2caf6 AM |
2724 | if (!nl_table) |
2725 | goto panic; | |
1da177e4 | 2726 | |
1da177e4 | 2727 | for (i = 0; i < MAX_LINKS; i++) { |
c428ecd1 HX |
2728 | if (rhashtable_init(&nl_table[i].hash, |
2729 | &netlink_rhashtable_params) < 0) { | |
e341694e TG |
2730 | while (--i > 0) |
2731 | rhashtable_destroy(&nl_table[i].hash); | |
1da177e4 | 2732 | kfree(nl_table); |
fab2caf6 | 2733 | goto panic; |
1da177e4 | 2734 | } |
1da177e4 LT |
2735 | } |
2736 | ||
bcbde0d4 DB |
2737 | INIT_LIST_HEAD(&netlink_tap_all); |
2738 | ||
b963ea89 DM |
2739 | netlink_add_usersock_entry(); |
2740 | ||
1da177e4 | 2741 | sock_register(&netlink_family_ops); |
b4b51029 | 2742 | register_pernet_subsys(&netlink_net_ops); |
746fac4d | 2743 | /* The netlink device handler may be needed early. */ |
1da177e4 LT |
2744 | rtnetlink_init(); |
2745 | out: | |
2746 | return err; | |
fab2caf6 AM |
2747 | panic: |
2748 | panic("netlink_init: Cannot allocate nl_table\n"); | |
1da177e4 LT |
2749 | } |
2750 | ||
1da177e4 | 2751 | core_initcall(netlink_proto_init); |