]>
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 | ||
41c6d650 | 694 | if (!refcount_dec_and_test(&sk->sk_refcnt)) |
ed5d7788 HX |
695 | return; |
696 | ||
697 | if (nlk->cb_running && nlk->cb.done) { | |
698 | INIT_WORK(&nlk->work, netlink_sock_destruct_work); | |
699 | schedule_work(&nlk->work); | |
700 | return; | |
701 | } | |
21e4902a | 702 | |
ed5d7788 | 703 | sk_free(sk); |
21e4902a TG |
704 | } |
705 | ||
1da177e4 LT |
706 | static int netlink_release(struct socket *sock) |
707 | { | |
708 | struct sock *sk = sock->sk; | |
709 | struct netlink_sock *nlk; | |
710 | ||
711 | if (!sk) | |
712 | return 0; | |
713 | ||
714 | netlink_remove(sk); | |
ac57b3a9 | 715 | sock_orphan(sk); |
1da177e4 LT |
716 | nlk = nlk_sk(sk); |
717 | ||
3f660d66 HX |
718 | /* |
719 | * OK. Socket is unlinked, any packets that arrive now | |
720 | * will be purged. | |
721 | */ | |
1da177e4 | 722 | |
ee1c2442 JB |
723 | /* must not acquire netlink_table_lock in any way again before unbind |
724 | * and notifying genetlink is done as otherwise it might deadlock | |
725 | */ | |
726 | if (nlk->netlink_unbind) { | |
727 | int i; | |
728 | ||
729 | for (i = 0; i < nlk->ngroups; i++) | |
730 | if (test_bit(i, nlk->groups)) | |
731 | nlk->netlink_unbind(sock_net(sk), i + 1); | |
732 | } | |
733 | if (sk->sk_protocol == NETLINK_GENERIC && | |
734 | atomic_dec_return(&genl_sk_destructing_cnt) == 0) | |
735 | wake_up(&genl_sk_destructing_waitq); | |
736 | ||
1da177e4 LT |
737 | sock->sk = NULL; |
738 | wake_up_interruptible_all(&nlk->wait); | |
739 | ||
740 | skb_queue_purge(&sk->sk_write_queue); | |
741 | ||
e2726020 | 742 | if (nlk->portid && nlk->bound) { |
1da177e4 | 743 | struct netlink_notify n = { |
3b1e0a65 | 744 | .net = sock_net(sk), |
1da177e4 | 745 | .protocol = sk->sk_protocol, |
15e47304 | 746 | .portid = nlk->portid, |
1da177e4 | 747 | }; |
efa172f4 | 748 | blocking_notifier_call_chain(&netlink_chain, |
e041c683 | 749 | NETLINK_URELEASE, &n); |
746fac4d | 750 | } |
4fdb3bb7 | 751 | |
5e7c001c | 752 | module_put(nlk->module); |
4fdb3bb7 | 753 | |
aed81560 | 754 | if (netlink_is_kernel(sk)) { |
b10dcb3b | 755 | netlink_table_grab(); |
869e58f8 DL |
756 | BUG_ON(nl_table[sk->sk_protocol].registered == 0); |
757 | if (--nl_table[sk->sk_protocol].registered == 0) { | |
6d772ac5 ED |
758 | struct listeners *old; |
759 | ||
760 | old = nl_deref_protected(nl_table[sk->sk_protocol].listeners); | |
761 | RCU_INIT_POINTER(nl_table[sk->sk_protocol].listeners, NULL); | |
762 | kfree_rcu(old, rcu); | |
869e58f8 | 763 | nl_table[sk->sk_protocol].module = NULL; |
9785e10a | 764 | nl_table[sk->sk_protocol].bind = NULL; |
4f520900 | 765 | nl_table[sk->sk_protocol].unbind = NULL; |
9785e10a | 766 | nl_table[sk->sk_protocol].flags = 0; |
869e58f8 DL |
767 | nl_table[sk->sk_protocol].registered = 0; |
768 | } | |
b10dcb3b | 769 | netlink_table_ungrab(); |
658cb354 | 770 | } |
77247bbb | 771 | |
f7fa9b10 PM |
772 | kfree(nlk->groups); |
773 | nlk->groups = NULL; | |
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; | |
958 | int err; | |
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 RGB |
985 | |
986 | if (nlk->netlink_bind && groups) { | |
987 | int group; | |
988 | ||
989 | for (group = 0; group < nlk->ngroups; group++) { | |
990 | if (!test_bit(group, &groups)) | |
991 | continue; | |
8b7c36d8 | 992 | err = nlk->netlink_bind(net, group + 1); |
4f520900 RGB |
993 | if (!err) |
994 | continue; | |
023e2cfa | 995 | netlink_undo_bind(group, groups, sk); |
4f520900 RGB |
996 | return err; |
997 | } | |
998 | } | |
999 | ||
da314c99 HX |
1000 | /* No need for barriers here as we return to user-space without |
1001 | * using any of the bound attributes. | |
1002 | */ | |
1003 | if (!bound) { | |
1da177e4 | 1004 | err = nladdr->nl_pid ? |
8ea65f4a | 1005 | netlink_insert(sk, nladdr->nl_pid) : |
1da177e4 | 1006 | netlink_autobind(sock); |
4f520900 | 1007 | if (err) { |
023e2cfa | 1008 | netlink_undo_bind(nlk->ngroups, groups, sk); |
1da177e4 | 1009 | return err; |
4f520900 | 1010 | } |
1da177e4 LT |
1011 | } |
1012 | ||
4f520900 | 1013 | if (!groups && (nlk->groups == NULL || !(u32)nlk->groups[0])) |
1da177e4 LT |
1014 | return 0; |
1015 | ||
1016 | netlink_table_grab(); | |
f7fa9b10 | 1017 | netlink_update_subscriptions(sk, nlk->subscriptions + |
4f520900 | 1018 | hweight32(groups) - |
746fac4d | 1019 | hweight32(nlk->groups[0])); |
4f520900 | 1020 | nlk->groups[0] = (nlk->groups[0] & ~0xffffffffUL) | groups; |
4277a083 | 1021 | netlink_update_listeners(sk); |
1da177e4 LT |
1022 | netlink_table_ungrab(); |
1023 | ||
1024 | return 0; | |
1025 | } | |
1026 | ||
1027 | static int netlink_connect(struct socket *sock, struct sockaddr *addr, | |
1028 | int alen, int flags) | |
1029 | { | |
1030 | int err = 0; | |
1031 | struct sock *sk = sock->sk; | |
1032 | struct netlink_sock *nlk = nlk_sk(sk); | |
6ac552fd | 1033 | struct sockaddr_nl *nladdr = (struct sockaddr_nl *)addr; |
1da177e4 | 1034 | |
6503d961 CG |
1035 | if (alen < sizeof(addr->sa_family)) |
1036 | return -EINVAL; | |
1037 | ||
1da177e4 LT |
1038 | if (addr->sa_family == AF_UNSPEC) { |
1039 | sk->sk_state = NETLINK_UNCONNECTED; | |
15e47304 | 1040 | nlk->dst_portid = 0; |
d629b836 | 1041 | nlk->dst_group = 0; |
1da177e4 LT |
1042 | return 0; |
1043 | } | |
1044 | if (addr->sa_family != AF_NETLINK) | |
1045 | return -EINVAL; | |
1046 | ||
46833a86 | 1047 | if ((nladdr->nl_groups || nladdr->nl_pid) && |
5187cd05 | 1048 | !netlink_allowed(sock, NL_CFG_F_NONROOT_SEND)) |
1da177e4 LT |
1049 | return -EPERM; |
1050 | ||
da314c99 HX |
1051 | /* No need for barriers here as we return to user-space without |
1052 | * using any of the bound attributes. | |
1053 | */ | |
1054 | if (!nlk->bound) | |
1da177e4 LT |
1055 | err = netlink_autobind(sock); |
1056 | ||
1057 | if (err == 0) { | |
1058 | sk->sk_state = NETLINK_CONNECTED; | |
15e47304 | 1059 | nlk->dst_portid = nladdr->nl_pid; |
d629b836 | 1060 | nlk->dst_group = ffs(nladdr->nl_groups); |
1da177e4 LT |
1061 | } |
1062 | ||
1063 | return err; | |
1064 | } | |
1065 | ||
6ac552fd PM |
1066 | static int netlink_getname(struct socket *sock, struct sockaddr *addr, |
1067 | int *addr_len, int peer) | |
1da177e4 LT |
1068 | { |
1069 | struct sock *sk = sock->sk; | |
1070 | struct netlink_sock *nlk = nlk_sk(sk); | |
13cfa97b | 1071 | DECLARE_SOCKADDR(struct sockaddr_nl *, nladdr, addr); |
746fac4d | 1072 | |
1da177e4 LT |
1073 | nladdr->nl_family = AF_NETLINK; |
1074 | nladdr->nl_pad = 0; | |
1075 | *addr_len = sizeof(*nladdr); | |
1076 | ||
1077 | if (peer) { | |
15e47304 | 1078 | nladdr->nl_pid = nlk->dst_portid; |
d629b836 | 1079 | nladdr->nl_groups = netlink_group_mask(nlk->dst_group); |
1da177e4 | 1080 | } else { |
15e47304 | 1081 | nladdr->nl_pid = nlk->portid; |
513c2500 | 1082 | nladdr->nl_groups = nlk->groups ? nlk->groups[0] : 0; |
1da177e4 LT |
1083 | } |
1084 | return 0; | |
1085 | } | |
1086 | ||
025c6818 DD |
1087 | static int netlink_ioctl(struct socket *sock, unsigned int cmd, |
1088 | unsigned long arg) | |
1089 | { | |
1090 | /* try to hand this ioctl down to the NIC drivers. | |
1091 | */ | |
1092 | return -ENOIOCTLCMD; | |
1093 | } | |
1094 | ||
15e47304 | 1095 | static struct sock *netlink_getsockbyportid(struct sock *ssk, u32 portid) |
1da177e4 | 1096 | { |
1da177e4 LT |
1097 | struct sock *sock; |
1098 | struct netlink_sock *nlk; | |
1099 | ||
15e47304 | 1100 | sock = netlink_lookup(sock_net(ssk), ssk->sk_protocol, portid); |
1da177e4 LT |
1101 | if (!sock) |
1102 | return ERR_PTR(-ECONNREFUSED); | |
1103 | ||
1104 | /* Don't bother queuing skb if kernel socket has no input function */ | |
1105 | nlk = nlk_sk(sock); | |
cd40b7d3 | 1106 | if (sock->sk_state == NETLINK_CONNECTED && |
15e47304 | 1107 | nlk->dst_portid != nlk_sk(ssk)->portid) { |
1da177e4 LT |
1108 | sock_put(sock); |
1109 | return ERR_PTR(-ECONNREFUSED); | |
1110 | } | |
1111 | return sock; | |
1112 | } | |
1113 | ||
1114 | struct sock *netlink_getsockbyfilp(struct file *filp) | |
1115 | { | |
496ad9aa | 1116 | struct inode *inode = file_inode(filp); |
1da177e4 LT |
1117 | struct sock *sock; |
1118 | ||
1119 | if (!S_ISSOCK(inode->i_mode)) | |
1120 | return ERR_PTR(-ENOTSOCK); | |
1121 | ||
1122 | sock = SOCKET_I(inode)->sk; | |
1123 | if (sock->sk_family != AF_NETLINK) | |
1124 | return ERR_PTR(-EINVAL); | |
1125 | ||
1126 | sock_hold(sock); | |
1127 | return sock; | |
1128 | } | |
1129 | ||
3a36515f PN |
1130 | static struct sk_buff *netlink_alloc_large_skb(unsigned int size, |
1131 | int broadcast) | |
c05cdb1b PNA |
1132 | { |
1133 | struct sk_buff *skb; | |
1134 | void *data; | |
1135 | ||
3a36515f | 1136 | if (size <= NLMSG_GOODSIZE || broadcast) |
c05cdb1b PNA |
1137 | return alloc_skb(size, GFP_KERNEL); |
1138 | ||
3a36515f PN |
1139 | size = SKB_DATA_ALIGN(size) + |
1140 | SKB_DATA_ALIGN(sizeof(struct skb_shared_info)); | |
c05cdb1b PNA |
1141 | |
1142 | data = vmalloc(size); | |
1143 | if (data == NULL) | |
3a36515f | 1144 | return NULL; |
c05cdb1b | 1145 | |
2ea2f62c | 1146 | skb = __build_skb(data, size); |
3a36515f PN |
1147 | if (skb == NULL) |
1148 | vfree(data); | |
2ea2f62c | 1149 | else |
3a36515f | 1150 | skb->destructor = netlink_skb_destructor; |
c05cdb1b PNA |
1151 | |
1152 | return skb; | |
c05cdb1b PNA |
1153 | } |
1154 | ||
1da177e4 LT |
1155 | /* |
1156 | * Attach a skb to a netlink socket. | |
1157 | * The caller must hold a reference to the destination socket. On error, the | |
1158 | * reference is dropped. The skb is not send to the destination, just all | |
1159 | * all error checks are performed and memory in the queue is reserved. | |
1160 | * Return values: | |
1161 | * < 0: error. skb freed, reference to sock dropped. | |
1162 | * 0: continue | |
1163 | * 1: repeat lookup - reference dropped while waiting for socket memory. | |
1164 | */ | |
9457afee | 1165 | int netlink_attachskb(struct sock *sk, struct sk_buff *skb, |
c3d8d1e3 | 1166 | long *timeo, struct sock *ssk) |
1da177e4 LT |
1167 | { |
1168 | struct netlink_sock *nlk; | |
1169 | ||
1170 | nlk = nlk_sk(sk); | |
1171 | ||
5fd96123 | 1172 | if ((atomic_read(&sk->sk_rmem_alloc) > sk->sk_rcvbuf || |
d1b4c689 | 1173 | test_bit(NETLINK_S_CONGESTED, &nlk->state))) { |
1da177e4 | 1174 | DECLARE_WAITQUEUE(wait, current); |
c3d8d1e3 | 1175 | if (!*timeo) { |
aed81560 | 1176 | if (!ssk || netlink_is_kernel(ssk)) |
1da177e4 LT |
1177 | netlink_overrun(sk); |
1178 | sock_put(sk); | |
1179 | kfree_skb(skb); | |
1180 | return -EAGAIN; | |
1181 | } | |
1182 | ||
1183 | __set_current_state(TASK_INTERRUPTIBLE); | |
1184 | add_wait_queue(&nlk->wait, &wait); | |
1185 | ||
1186 | if ((atomic_read(&sk->sk_rmem_alloc) > sk->sk_rcvbuf || | |
cc3a572f | 1187 | test_bit(NETLINK_S_CONGESTED, &nlk->state)) && |
1da177e4 | 1188 | !sock_flag(sk, SOCK_DEAD)) |
c3d8d1e3 | 1189 | *timeo = schedule_timeout(*timeo); |
1da177e4 LT |
1190 | |
1191 | __set_current_state(TASK_RUNNING); | |
1192 | remove_wait_queue(&nlk->wait, &wait); | |
1193 | sock_put(sk); | |
1194 | ||
1195 | if (signal_pending(current)) { | |
1196 | kfree_skb(skb); | |
c3d8d1e3 | 1197 | return sock_intr_errno(*timeo); |
1da177e4 LT |
1198 | } |
1199 | return 1; | |
1200 | } | |
cf0a018a | 1201 | netlink_skb_set_owner_r(skb, sk); |
1da177e4 LT |
1202 | return 0; |
1203 | } | |
1204 | ||
4a7e7c2a | 1205 | static int __netlink_sendskb(struct sock *sk, struct sk_buff *skb) |
1da177e4 | 1206 | { |
1da177e4 LT |
1207 | int len = skb->len; |
1208 | ||
bcbde0d4 DB |
1209 | netlink_deliver_tap(skb); |
1210 | ||
d1b4c689 | 1211 | skb_queue_tail(&sk->sk_receive_queue, skb); |
676d2369 | 1212 | sk->sk_data_ready(sk); |
4a7e7c2a ED |
1213 | return len; |
1214 | } | |
1215 | ||
1216 | int netlink_sendskb(struct sock *sk, struct sk_buff *skb) | |
1217 | { | |
1218 | int len = __netlink_sendskb(sk, skb); | |
1219 | ||
1da177e4 LT |
1220 | sock_put(sk); |
1221 | return len; | |
1222 | } | |
1223 | ||
1224 | void netlink_detachskb(struct sock *sk, struct sk_buff *skb) | |
1225 | { | |
1226 | kfree_skb(skb); | |
1227 | sock_put(sk); | |
1228 | } | |
1229 | ||
b57ef81f | 1230 | static struct sk_buff *netlink_trim(struct sk_buff *skb, gfp_t allocation) |
1da177e4 LT |
1231 | { |
1232 | int delta; | |
1233 | ||
1298ca46 | 1234 | WARN_ON(skb->sk != NULL); |
4305b541 | 1235 | delta = skb->end - skb->tail; |
c05cdb1b | 1236 | if (is_vmalloc_addr(skb->head) || delta * 2 < skb->truesize) |
1da177e4 LT |
1237 | return skb; |
1238 | ||
1239 | if (skb_shared(skb)) { | |
1240 | struct sk_buff *nskb = skb_clone(skb, allocation); | |
1241 | if (!nskb) | |
1242 | return skb; | |
8460c00f | 1243 | consume_skb(skb); |
1da177e4 LT |
1244 | skb = nskb; |
1245 | } | |
1246 | ||
158f323b ED |
1247 | pskb_expand_head(skb, 0, -delta, |
1248 | (allocation & ~__GFP_DIRECT_RECLAIM) | | |
1249 | __GFP_NOWARN | __GFP_NORETRY); | |
1da177e4 LT |
1250 | return skb; |
1251 | } | |
1252 | ||
3fbc2905 EB |
1253 | static int netlink_unicast_kernel(struct sock *sk, struct sk_buff *skb, |
1254 | struct sock *ssk) | |
cd40b7d3 DL |
1255 | { |
1256 | int ret; | |
1257 | struct netlink_sock *nlk = nlk_sk(sk); | |
1258 | ||
1259 | ret = -ECONNREFUSED; | |
1260 | if (nlk->netlink_rcv != NULL) { | |
1261 | ret = skb->len; | |
cf0a018a | 1262 | netlink_skb_set_owner_r(skb, sk); |
e32123e5 | 1263 | NETLINK_CB(skb).sk = ssk; |
73bfd370 | 1264 | netlink_deliver_tap_kernel(sk, ssk, skb); |
cd40b7d3 | 1265 | nlk->netlink_rcv(skb); |
bfb253c9 ED |
1266 | consume_skb(skb); |
1267 | } else { | |
1268 | kfree_skb(skb); | |
cd40b7d3 | 1269 | } |
cd40b7d3 DL |
1270 | sock_put(sk); |
1271 | return ret; | |
1272 | } | |
1273 | ||
1274 | int netlink_unicast(struct sock *ssk, struct sk_buff *skb, | |
15e47304 | 1275 | u32 portid, int nonblock) |
1da177e4 LT |
1276 | { |
1277 | struct sock *sk; | |
1278 | int err; | |
1279 | long timeo; | |
1280 | ||
1281 | skb = netlink_trim(skb, gfp_any()); | |
1282 | ||
1283 | timeo = sock_sndtimeo(ssk, nonblock); | |
1284 | retry: | |
15e47304 | 1285 | sk = netlink_getsockbyportid(ssk, portid); |
1da177e4 LT |
1286 | if (IS_ERR(sk)) { |
1287 | kfree_skb(skb); | |
1288 | return PTR_ERR(sk); | |
1289 | } | |
cd40b7d3 | 1290 | if (netlink_is_kernel(sk)) |
3fbc2905 | 1291 | return netlink_unicast_kernel(sk, skb, ssk); |
cd40b7d3 | 1292 | |
b1153f29 | 1293 | if (sk_filter(sk, skb)) { |
84874607 | 1294 | err = skb->len; |
b1153f29 SH |
1295 | kfree_skb(skb); |
1296 | sock_put(sk); | |
1297 | return err; | |
1298 | } | |
1299 | ||
9457afee | 1300 | err = netlink_attachskb(sk, skb, &timeo, ssk); |
1da177e4 LT |
1301 | if (err == 1) |
1302 | goto retry; | |
1303 | if (err) | |
1304 | return err; | |
1305 | ||
7ee015e0 | 1306 | return netlink_sendskb(sk, skb); |
1da177e4 | 1307 | } |
6ac552fd | 1308 | EXPORT_SYMBOL(netlink_unicast); |
1da177e4 | 1309 | |
4277a083 PM |
1310 | int netlink_has_listeners(struct sock *sk, unsigned int group) |
1311 | { | |
1312 | int res = 0; | |
5c398dc8 | 1313 | struct listeners *listeners; |
4277a083 | 1314 | |
aed81560 | 1315 | BUG_ON(!netlink_is_kernel(sk)); |
b4ff4f04 JB |
1316 | |
1317 | rcu_read_lock(); | |
1318 | listeners = rcu_dereference(nl_table[sk->sk_protocol].listeners); | |
1319 | ||
6d772ac5 | 1320 | if (listeners && group - 1 < nl_table[sk->sk_protocol].groups) |
5c398dc8 | 1321 | res = test_bit(group - 1, listeners->masks); |
b4ff4f04 JB |
1322 | |
1323 | rcu_read_unlock(); | |
1324 | ||
4277a083 PM |
1325 | return res; |
1326 | } | |
1327 | EXPORT_SYMBOL_GPL(netlink_has_listeners); | |
1328 | ||
b57ef81f | 1329 | static int netlink_broadcast_deliver(struct sock *sk, struct sk_buff *skb) |
1da177e4 LT |
1330 | { |
1331 | struct netlink_sock *nlk = nlk_sk(sk); | |
1332 | ||
1333 | if (atomic_read(&sk->sk_rmem_alloc) <= sk->sk_rcvbuf && | |
cc3a572f | 1334 | !test_bit(NETLINK_S_CONGESTED, &nlk->state)) { |
cf0a018a | 1335 | netlink_skb_set_owner_r(skb, sk); |
4a7e7c2a | 1336 | __netlink_sendskb(sk, skb); |
2c645800 | 1337 | return atomic_read(&sk->sk_rmem_alloc) > (sk->sk_rcvbuf >> 1); |
1da177e4 LT |
1338 | } |
1339 | return -1; | |
1340 | } | |
1341 | ||
1342 | struct netlink_broadcast_data { | |
1343 | struct sock *exclude_sk; | |
b4b51029 | 1344 | struct net *net; |
15e47304 | 1345 | u32 portid; |
1da177e4 LT |
1346 | u32 group; |
1347 | int failure; | |
ff491a73 | 1348 | int delivery_failure; |
1da177e4 LT |
1349 | int congested; |
1350 | int delivered; | |
7d877f3b | 1351 | gfp_t allocation; |
1da177e4 | 1352 | struct sk_buff *skb, *skb2; |
910a7e90 EB |
1353 | int (*tx_filter)(struct sock *dsk, struct sk_buff *skb, void *data); |
1354 | void *tx_data; | |
1da177e4 LT |
1355 | }; |
1356 | ||
46c9521f RR |
1357 | static void do_one_broadcast(struct sock *sk, |
1358 | struct netlink_broadcast_data *p) | |
1da177e4 LT |
1359 | { |
1360 | struct netlink_sock *nlk = nlk_sk(sk); | |
1361 | int val; | |
1362 | ||
1363 | if (p->exclude_sk == sk) | |
46c9521f | 1364 | return; |
1da177e4 | 1365 | |
15e47304 | 1366 | if (nlk->portid == p->portid || p->group - 1 >= nlk->ngroups || |
f7fa9b10 | 1367 | !test_bit(p->group - 1, nlk->groups)) |
46c9521f | 1368 | return; |
1da177e4 | 1369 | |
59324cf3 ND |
1370 | if (!net_eq(sock_net(sk), p->net)) { |
1371 | if (!(nlk->flags & NETLINK_F_LISTEN_ALL_NSID)) | |
1372 | return; | |
1373 | ||
1374 | if (!peernet_has_id(sock_net(sk), p->net)) | |
1375 | return; | |
1376 | ||
1377 | if (!file_ns_capable(sk->sk_socket->file, p->net->user_ns, | |
1378 | CAP_NET_BROADCAST)) | |
1379 | return; | |
1380 | } | |
b4b51029 | 1381 | |
1da177e4 LT |
1382 | if (p->failure) { |
1383 | netlink_overrun(sk); | |
46c9521f | 1384 | return; |
1da177e4 LT |
1385 | } |
1386 | ||
1387 | sock_hold(sk); | |
1388 | if (p->skb2 == NULL) { | |
68acc024 | 1389 | if (skb_shared(p->skb)) { |
1da177e4 LT |
1390 | p->skb2 = skb_clone(p->skb, p->allocation); |
1391 | } else { | |
68acc024 TC |
1392 | p->skb2 = skb_get(p->skb); |
1393 | /* | |
1394 | * skb ownership may have been set when | |
1395 | * delivered to a previous socket. | |
1396 | */ | |
1397 | skb_orphan(p->skb2); | |
1da177e4 LT |
1398 | } |
1399 | } | |
1400 | if (p->skb2 == NULL) { | |
1401 | netlink_overrun(sk); | |
1402 | /* Clone failed. Notify ALL listeners. */ | |
1403 | p->failure = 1; | |
cc3a572f | 1404 | if (nlk->flags & NETLINK_F_BROADCAST_SEND_ERROR) |
be0c22a4 | 1405 | p->delivery_failure = 1; |
59324cf3 ND |
1406 | goto out; |
1407 | } | |
1408 | if (p->tx_filter && p->tx_filter(sk, p->skb2, p->tx_data)) { | |
910a7e90 EB |
1409 | kfree_skb(p->skb2); |
1410 | p->skb2 = NULL; | |
59324cf3 ND |
1411 | goto out; |
1412 | } | |
1413 | if (sk_filter(sk, p->skb2)) { | |
b1153f29 SH |
1414 | kfree_skb(p->skb2); |
1415 | p->skb2 = NULL; | |
59324cf3 ND |
1416 | goto out; |
1417 | } | |
1418 | NETLINK_CB(p->skb2).nsid = peernet2id(sock_net(sk), p->net); | |
7212462f ND |
1419 | if (NETLINK_CB(p->skb2).nsid != NETNSA_NSID_NOT_ASSIGNED) |
1420 | NETLINK_CB(p->skb2).nsid_is_set = true; | |
59324cf3 ND |
1421 | val = netlink_broadcast_deliver(sk, p->skb2); |
1422 | if (val < 0) { | |
1da177e4 | 1423 | netlink_overrun(sk); |
cc3a572f | 1424 | if (nlk->flags & NETLINK_F_BROADCAST_SEND_ERROR) |
be0c22a4 | 1425 | p->delivery_failure = 1; |
1da177e4 LT |
1426 | } else { |
1427 | p->congested |= val; | |
1428 | p->delivered = 1; | |
1429 | p->skb2 = NULL; | |
1430 | } | |
59324cf3 | 1431 | out: |
1da177e4 | 1432 | sock_put(sk); |
1da177e4 LT |
1433 | } |
1434 | ||
15e47304 | 1435 | int netlink_broadcast_filtered(struct sock *ssk, struct sk_buff *skb, u32 portid, |
910a7e90 EB |
1436 | u32 group, gfp_t allocation, |
1437 | int (*filter)(struct sock *dsk, struct sk_buff *skb, void *data), | |
1438 | void *filter_data) | |
1da177e4 | 1439 | { |
3b1e0a65 | 1440 | struct net *net = sock_net(ssk); |
1da177e4 | 1441 | struct netlink_broadcast_data info; |
1da177e4 LT |
1442 | struct sock *sk; |
1443 | ||
1444 | skb = netlink_trim(skb, allocation); | |
1445 | ||
1446 | info.exclude_sk = ssk; | |
b4b51029 | 1447 | info.net = net; |
15e47304 | 1448 | info.portid = portid; |
1da177e4 LT |
1449 | info.group = group; |
1450 | info.failure = 0; | |
ff491a73 | 1451 | info.delivery_failure = 0; |
1da177e4 LT |
1452 | info.congested = 0; |
1453 | info.delivered = 0; | |
1454 | info.allocation = allocation; | |
1455 | info.skb = skb; | |
1456 | info.skb2 = NULL; | |
910a7e90 EB |
1457 | info.tx_filter = filter; |
1458 | info.tx_data = filter_data; | |
1da177e4 LT |
1459 | |
1460 | /* While we sleep in clone, do not allow to change socket list */ | |
1461 | ||
1462 | netlink_lock_table(); | |
1463 | ||
b67bfe0d | 1464 | sk_for_each_bound(sk, &nl_table[ssk->sk_protocol].mc_list) |
1da177e4 LT |
1465 | do_one_broadcast(sk, &info); |
1466 | ||
70d4bf6d | 1467 | consume_skb(skb); |
aa1c6a6f | 1468 | |
1da177e4 LT |
1469 | netlink_unlock_table(); |
1470 | ||
70d4bf6d NH |
1471 | if (info.delivery_failure) { |
1472 | kfree_skb(info.skb2); | |
ff491a73 | 1473 | return -ENOBUFS; |
658cb354 ED |
1474 | } |
1475 | consume_skb(info.skb2); | |
ff491a73 | 1476 | |
1da177e4 | 1477 | if (info.delivered) { |
d0164adc | 1478 | if (info.congested && gfpflags_allow_blocking(allocation)) |
1da177e4 LT |
1479 | yield(); |
1480 | return 0; | |
1481 | } | |
1da177e4 LT |
1482 | return -ESRCH; |
1483 | } | |
910a7e90 EB |
1484 | EXPORT_SYMBOL(netlink_broadcast_filtered); |
1485 | ||
15e47304 | 1486 | int netlink_broadcast(struct sock *ssk, struct sk_buff *skb, u32 portid, |
910a7e90 EB |
1487 | u32 group, gfp_t allocation) |
1488 | { | |
15e47304 | 1489 | return netlink_broadcast_filtered(ssk, skb, portid, group, allocation, |
910a7e90 EB |
1490 | NULL, NULL); |
1491 | } | |
6ac552fd | 1492 | EXPORT_SYMBOL(netlink_broadcast); |
1da177e4 LT |
1493 | |
1494 | struct netlink_set_err_data { | |
1495 | struct sock *exclude_sk; | |
15e47304 | 1496 | u32 portid; |
1da177e4 LT |
1497 | u32 group; |
1498 | int code; | |
1499 | }; | |
1500 | ||
b57ef81f | 1501 | static int do_one_set_err(struct sock *sk, struct netlink_set_err_data *p) |
1da177e4 LT |
1502 | { |
1503 | struct netlink_sock *nlk = nlk_sk(sk); | |
1a50307b | 1504 | int ret = 0; |
1da177e4 LT |
1505 | |
1506 | if (sk == p->exclude_sk) | |
1507 | goto out; | |
1508 | ||
09ad9bc7 | 1509 | if (!net_eq(sock_net(sk), sock_net(p->exclude_sk))) |
b4b51029 EB |
1510 | goto out; |
1511 | ||
15e47304 | 1512 | if (nlk->portid == p->portid || p->group - 1 >= nlk->ngroups || |
f7fa9b10 | 1513 | !test_bit(p->group - 1, nlk->groups)) |
1da177e4 LT |
1514 | goto out; |
1515 | ||
cc3a572f | 1516 | if (p->code == ENOBUFS && nlk->flags & NETLINK_F_RECV_NO_ENOBUFS) { |
1a50307b PNA |
1517 | ret = 1; |
1518 | goto out; | |
1519 | } | |
1520 | ||
1da177e4 LT |
1521 | sk->sk_err = p->code; |
1522 | sk->sk_error_report(sk); | |
1523 | out: | |
1a50307b | 1524 | return ret; |
1da177e4 LT |
1525 | } |
1526 | ||
4843b93c PNA |
1527 | /** |
1528 | * netlink_set_err - report error to broadcast listeners | |
1529 | * @ssk: the kernel netlink socket, as returned by netlink_kernel_create() | |
15e47304 | 1530 | * @portid: the PORTID of a process that we want to skip (if any) |
840e93f2 | 1531 | * @group: the broadcast group that will notice the error |
4843b93c | 1532 | * @code: error code, must be negative (as usual in kernelspace) |
1a50307b PNA |
1533 | * |
1534 | * This function returns the number of broadcast listeners that have set the | |
cc3a572f | 1535 | * NETLINK_NO_ENOBUFS socket option. |
4843b93c | 1536 | */ |
15e47304 | 1537 | int netlink_set_err(struct sock *ssk, u32 portid, u32 group, int code) |
1da177e4 LT |
1538 | { |
1539 | struct netlink_set_err_data info; | |
1da177e4 | 1540 | struct sock *sk; |
1a50307b | 1541 | int ret = 0; |
1da177e4 LT |
1542 | |
1543 | info.exclude_sk = ssk; | |
15e47304 | 1544 | info.portid = portid; |
1da177e4 | 1545 | info.group = group; |
4843b93c PNA |
1546 | /* sk->sk_err wants a positive error value */ |
1547 | info.code = -code; | |
1da177e4 LT |
1548 | |
1549 | read_lock(&nl_table_lock); | |
1550 | ||
b67bfe0d | 1551 | sk_for_each_bound(sk, &nl_table[ssk->sk_protocol].mc_list) |
1a50307b | 1552 | ret += do_one_set_err(sk, &info); |
1da177e4 LT |
1553 | |
1554 | read_unlock(&nl_table_lock); | |
1a50307b | 1555 | return ret; |
1da177e4 | 1556 | } |
dd5b6ce6 | 1557 | EXPORT_SYMBOL(netlink_set_err); |
1da177e4 | 1558 | |
84659eb5 JB |
1559 | /* must be called with netlink table grabbed */ |
1560 | static void netlink_update_socket_mc(struct netlink_sock *nlk, | |
1561 | unsigned int group, | |
1562 | int is_new) | |
1563 | { | |
1564 | int old, new = !!is_new, subscriptions; | |
1565 | ||
1566 | old = test_bit(group - 1, nlk->groups); | |
1567 | subscriptions = nlk->subscriptions - old + new; | |
1568 | if (new) | |
1569 | __set_bit(group - 1, nlk->groups); | |
1570 | else | |
1571 | __clear_bit(group - 1, nlk->groups); | |
1572 | netlink_update_subscriptions(&nlk->sk, subscriptions); | |
1573 | netlink_update_listeners(&nlk->sk); | |
1574 | } | |
1575 | ||
9a4595bc | 1576 | static int netlink_setsockopt(struct socket *sock, int level, int optname, |
b7058842 | 1577 | char __user *optval, unsigned int optlen) |
9a4595bc PM |
1578 | { |
1579 | struct sock *sk = sock->sk; | |
1580 | struct netlink_sock *nlk = nlk_sk(sk); | |
eb496534 JB |
1581 | unsigned int val = 0; |
1582 | int err; | |
9a4595bc PM |
1583 | |
1584 | if (level != SOL_NETLINK) | |
1585 | return -ENOPROTOOPT; | |
1586 | ||
d1b4c689 | 1587 | if (optlen >= sizeof(int) && |
eb496534 | 1588 | get_user(val, (unsigned int __user *)optval)) |
9a4595bc PM |
1589 | return -EFAULT; |
1590 | ||
1591 | switch (optname) { | |
1592 | case NETLINK_PKTINFO: | |
1593 | if (val) | |
cc3a572f | 1594 | nlk->flags |= NETLINK_F_RECV_PKTINFO; |
9a4595bc | 1595 | else |
cc3a572f | 1596 | nlk->flags &= ~NETLINK_F_RECV_PKTINFO; |
9a4595bc PM |
1597 | err = 0; |
1598 | break; | |
1599 | case NETLINK_ADD_MEMBERSHIP: | |
1600 | case NETLINK_DROP_MEMBERSHIP: { | |
5187cd05 | 1601 | if (!netlink_allowed(sock, NL_CFG_F_NONROOT_RECV)) |
9a4595bc | 1602 | return -EPERM; |
b4ff4f04 JB |
1603 | err = netlink_realloc_groups(sk); |
1604 | if (err) | |
1605 | return err; | |
9a4595bc PM |
1606 | if (!val || val - 1 >= nlk->ngroups) |
1607 | return -EINVAL; | |
7774d5e0 | 1608 | if (optname == NETLINK_ADD_MEMBERSHIP && nlk->netlink_bind) { |
023e2cfa | 1609 | err = nlk->netlink_bind(sock_net(sk), val); |
4f520900 RGB |
1610 | if (err) |
1611 | return err; | |
1612 | } | |
9a4595bc | 1613 | netlink_table_grab(); |
84659eb5 JB |
1614 | netlink_update_socket_mc(nlk, val, |
1615 | optname == NETLINK_ADD_MEMBERSHIP); | |
9a4595bc | 1616 | netlink_table_ungrab(); |
7774d5e0 | 1617 | if (optname == NETLINK_DROP_MEMBERSHIP && nlk->netlink_unbind) |
023e2cfa | 1618 | nlk->netlink_unbind(sock_net(sk), val); |
03292745 | 1619 | |
9a4595bc PM |
1620 | err = 0; |
1621 | break; | |
1622 | } | |
be0c22a4 PNA |
1623 | case NETLINK_BROADCAST_ERROR: |
1624 | if (val) | |
cc3a572f | 1625 | nlk->flags |= NETLINK_F_BROADCAST_SEND_ERROR; |
be0c22a4 | 1626 | else |
cc3a572f | 1627 | nlk->flags &= ~NETLINK_F_BROADCAST_SEND_ERROR; |
be0c22a4 PNA |
1628 | err = 0; |
1629 | break; | |
38938bfe PNA |
1630 | case NETLINK_NO_ENOBUFS: |
1631 | if (val) { | |
cc3a572f ND |
1632 | nlk->flags |= NETLINK_F_RECV_NO_ENOBUFS; |
1633 | clear_bit(NETLINK_S_CONGESTED, &nlk->state); | |
38938bfe | 1634 | wake_up_interruptible(&nlk->wait); |
658cb354 | 1635 | } else { |
cc3a572f | 1636 | nlk->flags &= ~NETLINK_F_RECV_NO_ENOBUFS; |
658cb354 | 1637 | } |
38938bfe PNA |
1638 | err = 0; |
1639 | break; | |
59324cf3 ND |
1640 | case NETLINK_LISTEN_ALL_NSID: |
1641 | if (!ns_capable(sock_net(sk)->user_ns, CAP_NET_BROADCAST)) | |
1642 | return -EPERM; | |
1643 | ||
1644 | if (val) | |
1645 | nlk->flags |= NETLINK_F_LISTEN_ALL_NSID; | |
1646 | else | |
1647 | nlk->flags &= ~NETLINK_F_LISTEN_ALL_NSID; | |
1648 | err = 0; | |
1649 | break; | |
0a6a3a23 CR |
1650 | case NETLINK_CAP_ACK: |
1651 | if (val) | |
1652 | nlk->flags |= NETLINK_F_CAP_ACK; | |
1653 | else | |
1654 | nlk->flags &= ~NETLINK_F_CAP_ACK; | |
1655 | err = 0; | |
1656 | break; | |
2d4bc933 JB |
1657 | case NETLINK_EXT_ACK: |
1658 | if (val) | |
1659 | nlk->flags |= NETLINK_F_EXT_ACK; | |
1660 | else | |
1661 | nlk->flags &= ~NETLINK_F_EXT_ACK; | |
1662 | err = 0; | |
1663 | break; | |
9a4595bc PM |
1664 | default: |
1665 | err = -ENOPROTOOPT; | |
1666 | } | |
1667 | return err; | |
1668 | } | |
1669 | ||
1670 | static int netlink_getsockopt(struct socket *sock, int level, int optname, | |
746fac4d | 1671 | char __user *optval, int __user *optlen) |
9a4595bc PM |
1672 | { |
1673 | struct sock *sk = sock->sk; | |
1674 | struct netlink_sock *nlk = nlk_sk(sk); | |
1675 | int len, val, err; | |
1676 | ||
1677 | if (level != SOL_NETLINK) | |
1678 | return -ENOPROTOOPT; | |
1679 | ||
1680 | if (get_user(len, optlen)) | |
1681 | return -EFAULT; | |
1682 | if (len < 0) | |
1683 | return -EINVAL; | |
1684 | ||
1685 | switch (optname) { | |
1686 | case NETLINK_PKTINFO: | |
1687 | if (len < sizeof(int)) | |
1688 | return -EINVAL; | |
1689 | len = sizeof(int); | |
cc3a572f | 1690 | val = nlk->flags & NETLINK_F_RECV_PKTINFO ? 1 : 0; |
a27b58fe HC |
1691 | if (put_user(len, optlen) || |
1692 | put_user(val, optval)) | |
1693 | return -EFAULT; | |
9a4595bc PM |
1694 | err = 0; |
1695 | break; | |
be0c22a4 PNA |
1696 | case NETLINK_BROADCAST_ERROR: |
1697 | if (len < sizeof(int)) | |
1698 | return -EINVAL; | |
1699 | len = sizeof(int); | |
cc3a572f | 1700 | val = nlk->flags & NETLINK_F_BROADCAST_SEND_ERROR ? 1 : 0; |
be0c22a4 PNA |
1701 | if (put_user(len, optlen) || |
1702 | put_user(val, optval)) | |
1703 | return -EFAULT; | |
1704 | err = 0; | |
1705 | break; | |
38938bfe PNA |
1706 | case NETLINK_NO_ENOBUFS: |
1707 | if (len < sizeof(int)) | |
1708 | return -EINVAL; | |
1709 | len = sizeof(int); | |
cc3a572f | 1710 | val = nlk->flags & NETLINK_F_RECV_NO_ENOBUFS ? 1 : 0; |
38938bfe PNA |
1711 | if (put_user(len, optlen) || |
1712 | put_user(val, optval)) | |
1713 | return -EFAULT; | |
1714 | err = 0; | |
1715 | break; | |
b42be38b DH |
1716 | case NETLINK_LIST_MEMBERSHIPS: { |
1717 | int pos, idx, shift; | |
1718 | ||
1719 | err = 0; | |
47191d65 | 1720 | netlink_lock_table(); |
b42be38b DH |
1721 | for (pos = 0; pos * 8 < nlk->ngroups; pos += sizeof(u32)) { |
1722 | if (len - pos < sizeof(u32)) | |
1723 | break; | |
1724 | ||
1725 | idx = pos / sizeof(unsigned long); | |
1726 | shift = (pos % sizeof(unsigned long)) * 8; | |
1727 | if (put_user((u32)(nlk->groups[idx] >> shift), | |
1728 | (u32 __user *)(optval + pos))) { | |
1729 | err = -EFAULT; | |
1730 | break; | |
1731 | } | |
1732 | } | |
1733 | if (put_user(ALIGN(nlk->ngroups / 8, sizeof(u32)), optlen)) | |
1734 | err = -EFAULT; | |
47191d65 | 1735 | netlink_unlock_table(); |
b42be38b DH |
1736 | break; |
1737 | } | |
0a6a3a23 CR |
1738 | case NETLINK_CAP_ACK: |
1739 | if (len < sizeof(int)) | |
1740 | return -EINVAL; | |
1741 | len = sizeof(int); | |
1742 | val = nlk->flags & NETLINK_F_CAP_ACK ? 1 : 0; | |
1743 | if (put_user(len, optlen) || | |
1744 | put_user(val, optval)) | |
1745 | return -EFAULT; | |
1746 | err = 0; | |
1747 | break; | |
2d4bc933 JB |
1748 | case NETLINK_EXT_ACK: |
1749 | if (len < sizeof(int)) | |
1750 | return -EINVAL; | |
1751 | len = sizeof(int); | |
1752 | val = nlk->flags & NETLINK_F_EXT_ACK ? 1 : 0; | |
1753 | if (put_user(len, optlen) || put_user(val, optval)) | |
1754 | return -EFAULT; | |
1755 | err = 0; | |
1756 | break; | |
9a4595bc PM |
1757 | default: |
1758 | err = -ENOPROTOOPT; | |
1759 | } | |
1760 | return err; | |
1761 | } | |
1762 | ||
1763 | static void netlink_cmsg_recv_pktinfo(struct msghdr *msg, struct sk_buff *skb) | |
1764 | { | |
1765 | struct nl_pktinfo info; | |
1766 | ||
1767 | info.group = NETLINK_CB(skb).dst_group; | |
1768 | put_cmsg(msg, SOL_NETLINK, NETLINK_PKTINFO, sizeof(info), &info); | |
1769 | } | |
1770 | ||
59324cf3 ND |
1771 | static void netlink_cmsg_listen_all_nsid(struct sock *sk, struct msghdr *msg, |
1772 | struct sk_buff *skb) | |
1773 | { | |
1774 | if (!NETLINK_CB(skb).nsid_is_set) | |
1775 | return; | |
1776 | ||
1777 | put_cmsg(msg, SOL_NETLINK, NETLINK_LISTEN_ALL_NSID, sizeof(int), | |
1778 | &NETLINK_CB(skb).nsid); | |
1779 | } | |
1780 | ||
1b784140 | 1781 | static int netlink_sendmsg(struct socket *sock, struct msghdr *msg, size_t len) |
1da177e4 | 1782 | { |
1da177e4 LT |
1783 | struct sock *sk = sock->sk; |
1784 | struct netlink_sock *nlk = nlk_sk(sk); | |
342dfc30 | 1785 | DECLARE_SOCKADDR(struct sockaddr_nl *, addr, msg->msg_name); |
15e47304 | 1786 | u32 dst_portid; |
d629b836 | 1787 | u32 dst_group; |
1da177e4 LT |
1788 | struct sk_buff *skb; |
1789 | int err; | |
1790 | struct scm_cookie scm; | |
2d7a85f4 | 1791 | u32 netlink_skb_flags = 0; |
1da177e4 LT |
1792 | |
1793 | if (msg->msg_flags&MSG_OOB) | |
1794 | return -EOPNOTSUPP; | |
1795 | ||
7cc05662 | 1796 | err = scm_send(sock, msg, &scm, true); |
1da177e4 LT |
1797 | if (err < 0) |
1798 | return err; | |
1799 | ||
1800 | if (msg->msg_namelen) { | |
b47030c7 | 1801 | err = -EINVAL; |
1da177e4 | 1802 | if (addr->nl_family != AF_NETLINK) |
b47030c7 | 1803 | goto out; |
15e47304 | 1804 | dst_portid = addr->nl_pid; |
d629b836 | 1805 | dst_group = ffs(addr->nl_groups); |
b47030c7 | 1806 | err = -EPERM; |
15e47304 | 1807 | if ((dst_group || dst_portid) && |
5187cd05 | 1808 | !netlink_allowed(sock, NL_CFG_F_NONROOT_SEND)) |
b47030c7 | 1809 | goto out; |
2d7a85f4 | 1810 | netlink_skb_flags |= NETLINK_SKB_DST; |
1da177e4 | 1811 | } else { |
15e47304 | 1812 | dst_portid = nlk->dst_portid; |
d629b836 | 1813 | dst_group = nlk->dst_group; |
1da177e4 LT |
1814 | } |
1815 | ||
da314c99 | 1816 | if (!nlk->bound) { |
1da177e4 LT |
1817 | err = netlink_autobind(sock); |
1818 | if (err) | |
1819 | goto out; | |
da314c99 HX |
1820 | } else { |
1821 | /* Ensure nlk is hashed and visible. */ | |
1822 | smp_rmb(); | |
1da177e4 LT |
1823 | } |
1824 | ||
1825 | err = -EMSGSIZE; | |
1826 | if (len > sk->sk_sndbuf - 32) | |
1827 | goto out; | |
1828 | err = -ENOBUFS; | |
3a36515f | 1829 | skb = netlink_alloc_large_skb(len, dst_group); |
6ac552fd | 1830 | if (skb == NULL) |
1da177e4 LT |
1831 | goto out; |
1832 | ||
15e47304 | 1833 | NETLINK_CB(skb).portid = nlk->portid; |
d629b836 | 1834 | NETLINK_CB(skb).dst_group = dst_group; |
7cc05662 | 1835 | NETLINK_CB(skb).creds = scm.creds; |
2d7a85f4 | 1836 | NETLINK_CB(skb).flags = netlink_skb_flags; |
1da177e4 | 1837 | |
1da177e4 | 1838 | err = -EFAULT; |
6ce8e9ce | 1839 | if (memcpy_from_msg(skb_put(skb, len), msg, len)) { |
1da177e4 LT |
1840 | kfree_skb(skb); |
1841 | goto out; | |
1842 | } | |
1843 | ||
1844 | err = security_netlink_send(sk, skb); | |
1845 | if (err) { | |
1846 | kfree_skb(skb); | |
1847 | goto out; | |
1848 | } | |
1849 | ||
d629b836 | 1850 | if (dst_group) { |
63354797 | 1851 | refcount_inc(&skb->users); |
15e47304 | 1852 | netlink_broadcast(sk, skb, dst_portid, dst_group, GFP_KERNEL); |
1da177e4 | 1853 | } |
15e47304 | 1854 | err = netlink_unicast(sk, skb, dst_portid, msg->msg_flags&MSG_DONTWAIT); |
1da177e4 LT |
1855 | |
1856 | out: | |
7cc05662 | 1857 | scm_destroy(&scm); |
1da177e4 LT |
1858 | return err; |
1859 | } | |
1860 | ||
1b784140 | 1861 | static int netlink_recvmsg(struct socket *sock, struct msghdr *msg, size_t len, |
1da177e4 LT |
1862 | int flags) |
1863 | { | |
1da177e4 LT |
1864 | struct scm_cookie scm; |
1865 | struct sock *sk = sock->sk; | |
1866 | struct netlink_sock *nlk = nlk_sk(sk); | |
1867 | int noblock = flags&MSG_DONTWAIT; | |
1868 | size_t copied; | |
68d6ac6d | 1869 | struct sk_buff *skb, *data_skb; |
b44d211e | 1870 | int err, ret; |
1da177e4 LT |
1871 | |
1872 | if (flags&MSG_OOB) | |
1873 | return -EOPNOTSUPP; | |
1874 | ||
1875 | copied = 0; | |
1876 | ||
6ac552fd PM |
1877 | skb = skb_recv_datagram(sk, flags, noblock, &err); |
1878 | if (skb == NULL) | |
1da177e4 LT |
1879 | goto out; |
1880 | ||
68d6ac6d JB |
1881 | data_skb = skb; |
1882 | ||
1dacc76d JB |
1883 | #ifdef CONFIG_COMPAT_NETLINK_MESSAGES |
1884 | if (unlikely(skb_shinfo(skb)->frag_list)) { | |
1dacc76d | 1885 | /* |
68d6ac6d JB |
1886 | * If this skb has a frag_list, then here that means that we |
1887 | * will have to use the frag_list skb's data for compat tasks | |
1888 | * and the regular skb's data for normal (non-compat) tasks. | |
1dacc76d | 1889 | * |
68d6ac6d JB |
1890 | * If we need to send the compat skb, assign it to the |
1891 | * 'data_skb' variable so that it will be used below for data | |
1892 | * copying. We keep 'skb' for everything else, including | |
1893 | * freeing both later. | |
1dacc76d | 1894 | */ |
68d6ac6d JB |
1895 | if (flags & MSG_CMSG_COMPAT) |
1896 | data_skb = skb_shinfo(skb)->frag_list; | |
1dacc76d JB |
1897 | } |
1898 | #endif | |
1899 | ||
9063e21f ED |
1900 | /* Record the max length of recvmsg() calls for future allocations */ |
1901 | nlk->max_recvmsg_len = max(nlk->max_recvmsg_len, len); | |
1902 | nlk->max_recvmsg_len = min_t(size_t, nlk->max_recvmsg_len, | |
d35c99ff | 1903 | SKB_WITH_OVERHEAD(32768)); |
9063e21f | 1904 | |
68d6ac6d | 1905 | copied = data_skb->len; |
1da177e4 LT |
1906 | if (len < copied) { |
1907 | msg->msg_flags |= MSG_TRUNC; | |
1908 | copied = len; | |
1909 | } | |
1910 | ||
68d6ac6d | 1911 | skb_reset_transport_header(data_skb); |
51f3d02b | 1912 | err = skb_copy_datagram_msg(data_skb, 0, msg, copied); |
1da177e4 LT |
1913 | |
1914 | if (msg->msg_name) { | |
342dfc30 | 1915 | DECLARE_SOCKADDR(struct sockaddr_nl *, addr, msg->msg_name); |
1da177e4 LT |
1916 | addr->nl_family = AF_NETLINK; |
1917 | addr->nl_pad = 0; | |
15e47304 | 1918 | addr->nl_pid = NETLINK_CB(skb).portid; |
d629b836 | 1919 | addr->nl_groups = netlink_group_mask(NETLINK_CB(skb).dst_group); |
1da177e4 LT |
1920 | msg->msg_namelen = sizeof(*addr); |
1921 | } | |
1922 | ||
cc3a572f | 1923 | if (nlk->flags & NETLINK_F_RECV_PKTINFO) |
cc9a06cd | 1924 | netlink_cmsg_recv_pktinfo(msg, skb); |
59324cf3 ND |
1925 | if (nlk->flags & NETLINK_F_LISTEN_ALL_NSID) |
1926 | netlink_cmsg_listen_all_nsid(sk, msg, skb); | |
cc9a06cd | 1927 | |
7cc05662 CH |
1928 | memset(&scm, 0, sizeof(scm)); |
1929 | scm.creds = *NETLINK_CREDS(skb); | |
188ccb55 | 1930 | if (flags & MSG_TRUNC) |
68d6ac6d | 1931 | copied = data_skb->len; |
daa3766e | 1932 | |
1da177e4 LT |
1933 | skb_free_datagram(sk, skb); |
1934 | ||
16b304f3 PS |
1935 | if (nlk->cb_running && |
1936 | atomic_read(&sk->sk_rmem_alloc) <= sk->sk_rcvbuf / 2) { | |
b44d211e AV |
1937 | ret = netlink_dump(sk); |
1938 | if (ret) { | |
ac30ef83 | 1939 | sk->sk_err = -ret; |
b44d211e AV |
1940 | sk->sk_error_report(sk); |
1941 | } | |
1942 | } | |
1da177e4 | 1943 | |
7cc05662 | 1944 | scm_recv(sock, msg, &scm, flags); |
1da177e4 LT |
1945 | out: |
1946 | netlink_rcv_wake(sk); | |
1947 | return err ? : copied; | |
1948 | } | |
1949 | ||
676d2369 | 1950 | static void netlink_data_ready(struct sock *sk) |
1da177e4 | 1951 | { |
cd40b7d3 | 1952 | BUG(); |
1da177e4 LT |
1953 | } |
1954 | ||
1955 | /* | |
746fac4d | 1956 | * We export these functions to other modules. They provide a |
1da177e4 LT |
1957 | * complete set of kernel non-blocking support for message |
1958 | * queueing. | |
1959 | */ | |
1960 | ||
1961 | struct sock * | |
9f00d977 PNA |
1962 | __netlink_kernel_create(struct net *net, int unit, struct module *module, |
1963 | struct netlink_kernel_cfg *cfg) | |
1da177e4 LT |
1964 | { |
1965 | struct socket *sock; | |
1966 | struct sock *sk; | |
77247bbb | 1967 | struct netlink_sock *nlk; |
5c398dc8 | 1968 | struct listeners *listeners = NULL; |
a31f2d17 PNA |
1969 | struct mutex *cb_mutex = cfg ? cfg->cb_mutex : NULL; |
1970 | unsigned int groups; | |
1da177e4 | 1971 | |
fab2caf6 | 1972 | BUG_ON(!nl_table); |
1da177e4 | 1973 | |
6ac552fd | 1974 | if (unit < 0 || unit >= MAX_LINKS) |
1da177e4 LT |
1975 | return NULL; |
1976 | ||
1977 | if (sock_create_lite(PF_NETLINK, SOCK_DGRAM, unit, &sock)) | |
1978 | return NULL; | |
13d3078e EB |
1979 | |
1980 | if (__netlink_create(net, sock, cb_mutex, unit, 1) < 0) | |
23fe1866 PE |
1981 | goto out_sock_release_nosk; |
1982 | ||
1983 | sk = sock->sk; | |
4fdb3bb7 | 1984 | |
a31f2d17 | 1985 | if (!cfg || cfg->groups < 32) |
4277a083 | 1986 | groups = 32; |
a31f2d17 PNA |
1987 | else |
1988 | groups = cfg->groups; | |
4277a083 | 1989 | |
5c398dc8 | 1990 | listeners = kzalloc(sizeof(*listeners) + NLGRPSZ(groups), GFP_KERNEL); |
4277a083 PM |
1991 | if (!listeners) |
1992 | goto out_sock_release; | |
1993 | ||
1da177e4 | 1994 | sk->sk_data_ready = netlink_data_ready; |
a31f2d17 PNA |
1995 | if (cfg && cfg->input) |
1996 | nlk_sk(sk)->netlink_rcv = cfg->input; | |
1da177e4 | 1997 | |
8ea65f4a | 1998 | if (netlink_insert(sk, 0)) |
77247bbb | 1999 | goto out_sock_release; |
4fdb3bb7 | 2000 | |
77247bbb | 2001 | nlk = nlk_sk(sk); |
cc3a572f | 2002 | nlk->flags |= NETLINK_F_KERNEL_SOCKET; |
4fdb3bb7 | 2003 | |
4fdb3bb7 | 2004 | netlink_table_grab(); |
b4b51029 EB |
2005 | if (!nl_table[unit].registered) { |
2006 | nl_table[unit].groups = groups; | |
5c398dc8 | 2007 | rcu_assign_pointer(nl_table[unit].listeners, listeners); |
b4b51029 EB |
2008 | nl_table[unit].cb_mutex = cb_mutex; |
2009 | nl_table[unit].module = module; | |
9785e10a PNA |
2010 | if (cfg) { |
2011 | nl_table[unit].bind = cfg->bind; | |
6251edd9 | 2012 | nl_table[unit].unbind = cfg->unbind; |
9785e10a | 2013 | nl_table[unit].flags = cfg->flags; |
da12c90e G |
2014 | if (cfg->compare) |
2015 | nl_table[unit].compare = cfg->compare; | |
9785e10a | 2016 | } |
b4b51029 | 2017 | nl_table[unit].registered = 1; |
f937f1f4 JJ |
2018 | } else { |
2019 | kfree(listeners); | |
869e58f8 | 2020 | nl_table[unit].registered++; |
b4b51029 | 2021 | } |
4fdb3bb7 | 2022 | netlink_table_ungrab(); |
77247bbb PM |
2023 | return sk; |
2024 | ||
4fdb3bb7 | 2025 | out_sock_release: |
4277a083 | 2026 | kfree(listeners); |
9dfbec1f | 2027 | netlink_kernel_release(sk); |
23fe1866 PE |
2028 | return NULL; |
2029 | ||
2030 | out_sock_release_nosk: | |
4fdb3bb7 | 2031 | sock_release(sock); |
77247bbb | 2032 | return NULL; |
1da177e4 | 2033 | } |
9f00d977 | 2034 | EXPORT_SYMBOL(__netlink_kernel_create); |
b7c6ba6e DL |
2035 | |
2036 | void | |
2037 | netlink_kernel_release(struct sock *sk) | |
2038 | { | |
13d3078e EB |
2039 | if (sk == NULL || sk->sk_socket == NULL) |
2040 | return; | |
2041 | ||
2042 | sock_release(sk->sk_socket); | |
b7c6ba6e DL |
2043 | } |
2044 | EXPORT_SYMBOL(netlink_kernel_release); | |
2045 | ||
d136f1bd | 2046 | int __netlink_change_ngroups(struct sock *sk, unsigned int groups) |
b4ff4f04 | 2047 | { |
5c398dc8 | 2048 | struct listeners *new, *old; |
b4ff4f04 | 2049 | struct netlink_table *tbl = &nl_table[sk->sk_protocol]; |
b4ff4f04 JB |
2050 | |
2051 | if (groups < 32) | |
2052 | groups = 32; | |
2053 | ||
b4ff4f04 | 2054 | if (NLGRPSZ(tbl->groups) < NLGRPSZ(groups)) { |
5c398dc8 ED |
2055 | new = kzalloc(sizeof(*new) + NLGRPSZ(groups), GFP_ATOMIC); |
2056 | if (!new) | |
d136f1bd | 2057 | return -ENOMEM; |
6d772ac5 | 2058 | old = nl_deref_protected(tbl->listeners); |
5c398dc8 ED |
2059 | memcpy(new->masks, old->masks, NLGRPSZ(tbl->groups)); |
2060 | rcu_assign_pointer(tbl->listeners, new); | |
2061 | ||
37b6b935 | 2062 | kfree_rcu(old, rcu); |
b4ff4f04 JB |
2063 | } |
2064 | tbl->groups = groups; | |
2065 | ||
d136f1bd JB |
2066 | return 0; |
2067 | } | |
2068 | ||
2069 | /** | |
2070 | * netlink_change_ngroups - change number of multicast groups | |
2071 | * | |
2072 | * This changes the number of multicast groups that are available | |
2073 | * on a certain netlink family. Note that it is not possible to | |
2074 | * change the number of groups to below 32. Also note that it does | |
2075 | * not implicitly call netlink_clear_multicast_users() when the | |
2076 | * number of groups is reduced. | |
2077 | * | |
2078 | * @sk: The kernel netlink socket, as returned by netlink_kernel_create(). | |
2079 | * @groups: The new number of groups. | |
2080 | */ | |
2081 | int netlink_change_ngroups(struct sock *sk, unsigned int groups) | |
2082 | { | |
2083 | int err; | |
2084 | ||
2085 | netlink_table_grab(); | |
2086 | err = __netlink_change_ngroups(sk, groups); | |
b4ff4f04 | 2087 | netlink_table_ungrab(); |
d136f1bd | 2088 | |
b4ff4f04 JB |
2089 | return err; |
2090 | } | |
b4ff4f04 | 2091 | |
b8273570 JB |
2092 | void __netlink_clear_multicast_users(struct sock *ksk, unsigned int group) |
2093 | { | |
2094 | struct sock *sk; | |
b8273570 JB |
2095 | struct netlink_table *tbl = &nl_table[ksk->sk_protocol]; |
2096 | ||
b67bfe0d | 2097 | sk_for_each_bound(sk, &tbl->mc_list) |
b8273570 JB |
2098 | netlink_update_socket_mc(nlk_sk(sk), group, 0); |
2099 | } | |
2100 | ||
a46621a3 | 2101 | struct nlmsghdr * |
15e47304 | 2102 | __nlmsg_put(struct sk_buff *skb, u32 portid, u32 seq, int type, int len, int flags) |
a46621a3 DV |
2103 | { |
2104 | struct nlmsghdr *nlh; | |
573ce260 | 2105 | int size = nlmsg_msg_size(len); |
a46621a3 | 2106 | |
4df864c1 | 2107 | nlh = skb_put(skb, NLMSG_ALIGN(size)); |
a46621a3 DV |
2108 | nlh->nlmsg_type = type; |
2109 | nlh->nlmsg_len = size; | |
2110 | nlh->nlmsg_flags = flags; | |
15e47304 | 2111 | nlh->nlmsg_pid = portid; |
a46621a3 DV |
2112 | nlh->nlmsg_seq = seq; |
2113 | if (!__builtin_constant_p(size) || NLMSG_ALIGN(size) - size != 0) | |
573ce260 | 2114 | memset(nlmsg_data(nlh) + len, 0, NLMSG_ALIGN(size) - size); |
a46621a3 DV |
2115 | return nlh; |
2116 | } | |
2117 | EXPORT_SYMBOL(__nlmsg_put); | |
2118 | ||
1da177e4 LT |
2119 | /* |
2120 | * It looks a bit ugly. | |
2121 | * It would be better to create kernel thread. | |
2122 | */ | |
2123 | ||
2124 | static int netlink_dump(struct sock *sk) | |
2125 | { | |
2126 | struct netlink_sock *nlk = nlk_sk(sk); | |
2127 | struct netlink_callback *cb; | |
c7ac8679 | 2128 | struct sk_buff *skb = NULL; |
1da177e4 | 2129 | struct nlmsghdr *nlh; |
92964c79 | 2130 | struct module *module; |
bf8b79e4 | 2131 | int len, err = -ENOBUFS; |
db65a3aa | 2132 | int alloc_min_size; |
c7ac8679 | 2133 | int alloc_size; |
1da177e4 | 2134 | |
af65bdfc | 2135 | mutex_lock(nlk->cb_mutex); |
16b304f3 | 2136 | if (!nlk->cb_running) { |
bf8b79e4 TG |
2137 | err = -EINVAL; |
2138 | goto errout_skb; | |
1da177e4 LT |
2139 | } |
2140 | ||
d1b4c689 | 2141 | if (atomic_read(&sk->sk_rmem_alloc) >= sk->sk_rcvbuf) |
f9c22888 | 2142 | goto errout_skb; |
9063e21f ED |
2143 | |
2144 | /* NLMSG_GOODSIZE is small to avoid high order allocations being | |
2145 | * required, but it makes sense to _attempt_ a 16K bytes allocation | |
2146 | * to reduce number of system calls on dump operations, if user | |
2147 | * ever provided a big enough buffer. | |
2148 | */ | |
db65a3aa AR |
2149 | cb = &nlk->cb; |
2150 | alloc_min_size = max_t(int, cb->min_dump_alloc, NLMSG_GOODSIZE); | |
2151 | ||
2152 | if (alloc_min_size < nlk->max_recvmsg_len) { | |
2153 | alloc_size = nlk->max_recvmsg_len; | |
d35c99ff ED |
2154 | skb = alloc_skb(alloc_size, |
2155 | (GFP_KERNEL & ~__GFP_DIRECT_RECLAIM) | | |
2156 | __GFP_NOWARN | __GFP_NORETRY); | |
9063e21f | 2157 | } |
db65a3aa AR |
2158 | if (!skb) { |
2159 | alloc_size = alloc_min_size; | |
c5b0db32 | 2160 | skb = alloc_skb(alloc_size, GFP_KERNEL); |
db65a3aa | 2161 | } |
c7ac8679 | 2162 | if (!skb) |
c63d6ea3 | 2163 | goto errout_skb; |
db65a3aa AR |
2164 | |
2165 | /* Trim skb to allocated size. User is expected to provide buffer as | |
2166 | * large as max(min_dump_alloc, 16KiB (mac_recvmsg_len capped at | |
2167 | * netlink_recvmsg())). dump will pack as many smaller messages as | |
2168 | * could fit within the allocated skb. skb is typically allocated | |
2169 | * with larger space than required (could be as much as near 2x the | |
2170 | * requested size with align to next power of 2 approach). Allowing | |
2171 | * dump to use the excess space makes it difficult for a user to have a | |
2172 | * reasonable static buffer based on the expected largest dump of a | |
2173 | * single netdev. The outcome is MSG_TRUNC error. | |
2174 | */ | |
d1b4c689 | 2175 | skb_reserve(skb, skb_tailroom(skb) - alloc_size); |
f9c22888 | 2176 | netlink_skb_set_owner_r(skb, sk); |
c7ac8679 | 2177 | |
1da177e4 LT |
2178 | len = cb->dump(skb, cb); |
2179 | ||
2180 | if (len > 0) { | |
af65bdfc | 2181 | mutex_unlock(nlk->cb_mutex); |
b1153f29 SH |
2182 | |
2183 | if (sk_filter(sk, skb)) | |
2184 | kfree_skb(skb); | |
4a7e7c2a ED |
2185 | else |
2186 | __netlink_sendskb(sk, skb); | |
1da177e4 LT |
2187 | return 0; |
2188 | } | |
2189 | ||
bf8b79e4 TG |
2190 | nlh = nlmsg_put_answer(skb, cb, NLMSG_DONE, sizeof(len), NLM_F_MULTI); |
2191 | if (!nlh) | |
2192 | goto errout_skb; | |
2193 | ||
670dc283 JB |
2194 | nl_dump_check_consistent(cb, nlh); |
2195 | ||
bf8b79e4 TG |
2196 | memcpy(nlmsg_data(nlh), &len, sizeof(len)); |
2197 | ||
b1153f29 SH |
2198 | if (sk_filter(sk, skb)) |
2199 | kfree_skb(skb); | |
4a7e7c2a ED |
2200 | else |
2201 | __netlink_sendskb(sk, skb); | |
1da177e4 | 2202 | |
a8f74b22 TG |
2203 | if (cb->done) |
2204 | cb->done(cb); | |
1da177e4 | 2205 | |
16b304f3 | 2206 | nlk->cb_running = false; |
92964c79 HX |
2207 | module = cb->module; |
2208 | skb = cb->skb; | |
16b304f3 | 2209 | mutex_unlock(nlk->cb_mutex); |
92964c79 HX |
2210 | module_put(module); |
2211 | consume_skb(skb); | |
1da177e4 | 2212 | return 0; |
1797754e | 2213 | |
bf8b79e4 | 2214 | errout_skb: |
af65bdfc | 2215 | mutex_unlock(nlk->cb_mutex); |
bf8b79e4 | 2216 | kfree_skb(skb); |
bf8b79e4 | 2217 | return err; |
1da177e4 LT |
2218 | } |
2219 | ||
6dc878a8 G |
2220 | int __netlink_dump_start(struct sock *ssk, struct sk_buff *skb, |
2221 | const struct nlmsghdr *nlh, | |
2222 | struct netlink_dump_control *control) | |
1da177e4 LT |
2223 | { |
2224 | struct netlink_callback *cb; | |
2225 | struct sock *sk; | |
2226 | struct netlink_sock *nlk; | |
b44d211e | 2227 | int ret; |
1da177e4 | 2228 | |
63354797 | 2229 | refcount_inc(&skb->users); |
f9c22888 | 2230 | |
15e47304 | 2231 | sk = netlink_lookup(sock_net(ssk), ssk->sk_protocol, NETLINK_CB(skb).portid); |
1da177e4 | 2232 | if (sk == NULL) { |
16b304f3 PS |
2233 | ret = -ECONNREFUSED; |
2234 | goto error_free; | |
1da177e4 | 2235 | } |
6dc878a8 | 2236 | |
16b304f3 | 2237 | nlk = nlk_sk(sk); |
af65bdfc | 2238 | mutex_lock(nlk->cb_mutex); |
6dc878a8 | 2239 | /* A dump is in progress... */ |
16b304f3 | 2240 | if (nlk->cb_running) { |
6dc878a8 | 2241 | ret = -EBUSY; |
16b304f3 | 2242 | goto error_unlock; |
1da177e4 | 2243 | } |
6dc878a8 | 2244 | /* add reference of module which cb->dump belongs to */ |
16b304f3 | 2245 | if (!try_module_get(control->module)) { |
6dc878a8 | 2246 | ret = -EPROTONOSUPPORT; |
16b304f3 | 2247 | goto error_unlock; |
6dc878a8 G |
2248 | } |
2249 | ||
16b304f3 PS |
2250 | cb = &nlk->cb; |
2251 | memset(cb, 0, sizeof(*cb)); | |
fc9e50f5 | 2252 | cb->start = control->start; |
16b304f3 PS |
2253 | cb->dump = control->dump; |
2254 | cb->done = control->done; | |
2255 | cb->nlh = nlh; | |
2256 | cb->data = control->data; | |
2257 | cb->module = control->module; | |
2258 | cb->min_dump_alloc = control->min_dump_alloc; | |
2259 | cb->skb = skb; | |
2260 | ||
2261 | nlk->cb_running = true; | |
2262 | ||
af65bdfc | 2263 | mutex_unlock(nlk->cb_mutex); |
1da177e4 | 2264 | |
fc9e50f5 TH |
2265 | if (cb->start) |
2266 | cb->start(cb); | |
2267 | ||
b44d211e | 2268 | ret = netlink_dump(sk); |
1da177e4 | 2269 | sock_put(sk); |
5c58298c | 2270 | |
b44d211e AV |
2271 | if (ret) |
2272 | return ret; | |
2273 | ||
5c58298c DL |
2274 | /* We successfully started a dump, by returning -EINTR we |
2275 | * signal not to send ACK even if it was requested. | |
2276 | */ | |
2277 | return -EINTR; | |
16b304f3 PS |
2278 | |
2279 | error_unlock: | |
2280 | sock_put(sk); | |
2281 | mutex_unlock(nlk->cb_mutex); | |
2282 | error_free: | |
2283 | kfree_skb(skb); | |
2284 | return ret; | |
1da177e4 | 2285 | } |
6dc878a8 | 2286 | EXPORT_SYMBOL(__netlink_dump_start); |
1da177e4 | 2287 | |
2d4bc933 JB |
2288 | void netlink_ack(struct sk_buff *in_skb, struct nlmsghdr *nlh, int err, |
2289 | const struct netlink_ext_ack *extack) | |
1da177e4 LT |
2290 | { |
2291 | struct sk_buff *skb; | |
2292 | struct nlmsghdr *rep; | |
2293 | struct nlmsgerr *errmsg; | |
339bf98f | 2294 | size_t payload = sizeof(*errmsg); |
2d4bc933 | 2295 | size_t tlvlen = 0; |
0a6a3a23 | 2296 | struct netlink_sock *nlk = nlk_sk(NETLINK_CB(in_skb).sk); |
2d4bc933 | 2297 | unsigned int flags = 0; |
1da177e4 | 2298 | |
0a6a3a23 | 2299 | /* Error messages get the original request appened, unless the user |
2d4bc933 JB |
2300 | * requests to cap the error message, and get extra error data if |
2301 | * requested. | |
0a6a3a23 | 2302 | */ |
2d4bc933 JB |
2303 | if (err) { |
2304 | if (!(nlk->flags & NETLINK_F_CAP_ACK)) | |
2305 | payload += nlmsg_len(nlh); | |
2306 | else | |
2307 | flags |= NLM_F_CAPPED; | |
2308 | if (nlk->flags & NETLINK_F_EXT_ACK && extack) { | |
2309 | if (extack->_msg) | |
2310 | tlvlen += nla_total_size(strlen(extack->_msg) + 1); | |
2311 | if (extack->bad_attr) | |
2312 | tlvlen += nla_total_size(sizeof(u32)); | |
2313 | } | |
2314 | } else { | |
2315 | flags |= NLM_F_CAPPED; | |
ba0dc5f6 JB |
2316 | |
2317 | if (nlk->flags & NETLINK_F_EXT_ACK && | |
2318 | extack && extack->cookie_len) | |
2319 | tlvlen += nla_total_size(extack->cookie_len); | |
2d4bc933 | 2320 | } |
1da177e4 | 2321 | |
2d4bc933 JB |
2322 | if (tlvlen) |
2323 | flags |= NLM_F_ACK_TLVS; | |
2324 | ||
2325 | skb = nlmsg_new(payload + tlvlen, GFP_KERNEL); | |
1da177e4 LT |
2326 | if (!skb) { |
2327 | struct sock *sk; | |
2328 | ||
3b1e0a65 | 2329 | sk = netlink_lookup(sock_net(in_skb->sk), |
b4b51029 | 2330 | in_skb->sk->sk_protocol, |
15e47304 | 2331 | NETLINK_CB(in_skb).portid); |
1da177e4 LT |
2332 | if (sk) { |
2333 | sk->sk_err = ENOBUFS; | |
2334 | sk->sk_error_report(sk); | |
2335 | sock_put(sk); | |
2336 | } | |
2337 | return; | |
2338 | } | |
2339 | ||
15e47304 | 2340 | rep = __nlmsg_put(skb, NETLINK_CB(in_skb).portid, nlh->nlmsg_seq, |
2d4bc933 | 2341 | NLMSG_ERROR, payload, flags); |
bf8b79e4 | 2342 | errmsg = nlmsg_data(rep); |
1da177e4 | 2343 | errmsg->error = err; |
0a6a3a23 | 2344 | memcpy(&errmsg->msg, nlh, payload > sizeof(*errmsg) ? nlh->nlmsg_len : sizeof(*nlh)); |
2d4bc933 | 2345 | |
ba0dc5f6 JB |
2346 | if (nlk->flags & NETLINK_F_EXT_ACK && extack) { |
2347 | if (err) { | |
2348 | if (extack->_msg) | |
2349 | WARN_ON(nla_put_string(skb, NLMSGERR_ATTR_MSG, | |
2350 | extack->_msg)); | |
2351 | if (extack->bad_attr && | |
2352 | !WARN_ON((u8 *)extack->bad_attr < in_skb->data || | |
2353 | (u8 *)extack->bad_attr >= in_skb->data + | |
2354 | in_skb->len)) | |
2355 | WARN_ON(nla_put_u32(skb, NLMSGERR_ATTR_OFFS, | |
2356 | (u8 *)extack->bad_attr - | |
2357 | in_skb->data)); | |
2358 | } else { | |
2359 | if (extack->cookie_len) | |
2360 | WARN_ON(nla_put(skb, NLMSGERR_ATTR_COOKIE, | |
2361 | extack->cookie_len, | |
2362 | extack->cookie)); | |
2363 | } | |
2d4bc933 JB |
2364 | } |
2365 | ||
2366 | nlmsg_end(skb, rep); | |
2367 | ||
15e47304 | 2368 | netlink_unicast(in_skb->sk, skb, NETLINK_CB(in_skb).portid, MSG_DONTWAIT); |
1da177e4 | 2369 | } |
6ac552fd | 2370 | EXPORT_SYMBOL(netlink_ack); |
1da177e4 | 2371 | |
cd40b7d3 | 2372 | int netlink_rcv_skb(struct sk_buff *skb, int (*cb)(struct sk_buff *, |
2d4bc933 JB |
2373 | struct nlmsghdr *, |
2374 | struct netlink_ext_ack *)) | |
82ace47a | 2375 | { |
2d4bc933 | 2376 | struct netlink_ext_ack extack = {}; |
82ace47a TG |
2377 | struct nlmsghdr *nlh; |
2378 | int err; | |
2379 | ||
2380 | while (skb->len >= nlmsg_total_size(0)) { | |
cd40b7d3 DL |
2381 | int msglen; |
2382 | ||
b529ccf2 | 2383 | nlh = nlmsg_hdr(skb); |
d35b6856 | 2384 | err = 0; |
82ace47a | 2385 | |
ad8e4b75 | 2386 | if (nlh->nlmsg_len < NLMSG_HDRLEN || skb->len < nlh->nlmsg_len) |
82ace47a TG |
2387 | return 0; |
2388 | ||
d35b6856 TG |
2389 | /* Only requests are handled by the kernel */ |
2390 | if (!(nlh->nlmsg_flags & NLM_F_REQUEST)) | |
5c58298c | 2391 | goto ack; |
45e7ae7f TG |
2392 | |
2393 | /* Skip control messages */ | |
2394 | if (nlh->nlmsg_type < NLMSG_MIN_TYPE) | |
5c58298c | 2395 | goto ack; |
d35b6856 | 2396 | |
2d4bc933 | 2397 | err = cb(skb, nlh, &extack); |
5c58298c DL |
2398 | if (err == -EINTR) |
2399 | goto skip; | |
2400 | ||
2401 | ack: | |
d35b6856 | 2402 | if (nlh->nlmsg_flags & NLM_F_ACK || err) |
2d4bc933 | 2403 | netlink_ack(skb, nlh, err, &extack); |
82ace47a | 2404 | |
5c58298c | 2405 | skip: |
6ac552fd | 2406 | msglen = NLMSG_ALIGN(nlh->nlmsg_len); |
cd40b7d3 DL |
2407 | if (msglen > skb->len) |
2408 | msglen = skb->len; | |
2409 | skb_pull(skb, msglen); | |
82ace47a TG |
2410 | } |
2411 | ||
2412 | return 0; | |
2413 | } | |
6ac552fd | 2414 | EXPORT_SYMBOL(netlink_rcv_skb); |
82ace47a | 2415 | |
d387f6ad TG |
2416 | /** |
2417 | * nlmsg_notify - send a notification netlink message | |
2418 | * @sk: netlink socket to use | |
2419 | * @skb: notification message | |
15e47304 | 2420 | * @portid: destination netlink portid for reports or 0 |
d387f6ad TG |
2421 | * @group: destination multicast group or 0 |
2422 | * @report: 1 to report back, 0 to disable | |
2423 | * @flags: allocation flags | |
2424 | */ | |
15e47304 | 2425 | int nlmsg_notify(struct sock *sk, struct sk_buff *skb, u32 portid, |
d387f6ad TG |
2426 | unsigned int group, int report, gfp_t flags) |
2427 | { | |
2428 | int err = 0; | |
2429 | ||
2430 | if (group) { | |
15e47304 | 2431 | int exclude_portid = 0; |
d387f6ad TG |
2432 | |
2433 | if (report) { | |
63354797 | 2434 | refcount_inc(&skb->users); |
15e47304 | 2435 | exclude_portid = portid; |
d387f6ad TG |
2436 | } |
2437 | ||
1ce85fe4 PNA |
2438 | /* errors reported via destination sk->sk_err, but propagate |
2439 | * delivery errors if NETLINK_BROADCAST_ERROR flag is set */ | |
15e47304 | 2440 | err = nlmsg_multicast(sk, skb, exclude_portid, group, flags); |
d387f6ad TG |
2441 | } |
2442 | ||
1ce85fe4 PNA |
2443 | if (report) { |
2444 | int err2; | |
2445 | ||
15e47304 | 2446 | err2 = nlmsg_unicast(sk, skb, portid); |
1ce85fe4 PNA |
2447 | if (!err || err == -ESRCH) |
2448 | err = err2; | |
2449 | } | |
d387f6ad TG |
2450 | |
2451 | return err; | |
2452 | } | |
6ac552fd | 2453 | EXPORT_SYMBOL(nlmsg_notify); |
d387f6ad | 2454 | |
1da177e4 LT |
2455 | #ifdef CONFIG_PROC_FS |
2456 | struct nl_seq_iter { | |
e372c414 | 2457 | struct seq_net_private p; |
56d28b1e | 2458 | struct rhashtable_iter hti; |
1da177e4 | 2459 | int link; |
1da177e4 LT |
2460 | }; |
2461 | ||
56d28b1e | 2462 | static int netlink_walk_start(struct nl_seq_iter *iter) |
1da177e4 | 2463 | { |
56d28b1e | 2464 | int err; |
1da177e4 | 2465 | |
8f6fd83c BC |
2466 | err = rhashtable_walk_init(&nl_table[iter->link].hash, &iter->hti, |
2467 | GFP_KERNEL); | |
56d28b1e HX |
2468 | if (err) { |
2469 | iter->link = MAX_LINKS; | |
2470 | return err; | |
1da177e4 | 2471 | } |
56d28b1e HX |
2472 | |
2473 | err = rhashtable_walk_start(&iter->hti); | |
2474 | return err == -EAGAIN ? 0 : err; | |
1da177e4 LT |
2475 | } |
2476 | ||
56d28b1e | 2477 | static void netlink_walk_stop(struct nl_seq_iter *iter) |
1da177e4 | 2478 | { |
56d28b1e HX |
2479 | rhashtable_walk_stop(&iter->hti); |
2480 | rhashtable_walk_exit(&iter->hti); | |
1da177e4 LT |
2481 | } |
2482 | ||
56d28b1e | 2483 | static void *__netlink_seq_next(struct seq_file *seq) |
1da177e4 | 2484 | { |
56d28b1e | 2485 | struct nl_seq_iter *iter = seq->private; |
e341694e | 2486 | struct netlink_sock *nlk; |
1da177e4 | 2487 | |
56d28b1e HX |
2488 | do { |
2489 | for (;;) { | |
2490 | int err; | |
1da177e4 | 2491 | |
56d28b1e | 2492 | nlk = rhashtable_walk_next(&iter->hti); |
746fac4d | 2493 | |
56d28b1e HX |
2494 | if (IS_ERR(nlk)) { |
2495 | if (PTR_ERR(nlk) == -EAGAIN) | |
2496 | continue; | |
e341694e | 2497 | |
56d28b1e HX |
2498 | return nlk; |
2499 | } | |
1da177e4 | 2500 | |
56d28b1e HX |
2501 | if (nlk) |
2502 | break; | |
1da177e4 | 2503 | |
56d28b1e HX |
2504 | netlink_walk_stop(iter); |
2505 | if (++iter->link >= MAX_LINKS) | |
2506 | return NULL; | |
da12c90e | 2507 | |
56d28b1e HX |
2508 | err = netlink_walk_start(iter); |
2509 | if (err) | |
2510 | return ERR_PTR(err); | |
1da177e4 | 2511 | } |
56d28b1e | 2512 | } while (sock_net(&nlk->sk) != seq_file_net(seq)); |
1da177e4 | 2513 | |
56d28b1e HX |
2514 | return nlk; |
2515 | } | |
1da177e4 | 2516 | |
56d28b1e HX |
2517 | static void *netlink_seq_start(struct seq_file *seq, loff_t *posp) |
2518 | { | |
2519 | struct nl_seq_iter *iter = seq->private; | |
2520 | void *obj = SEQ_START_TOKEN; | |
2521 | loff_t pos; | |
2522 | int err; | |
2523 | ||
2524 | iter->link = 0; | |
2525 | ||
2526 | err = netlink_walk_start(iter); | |
2527 | if (err) | |
2528 | return ERR_PTR(err); | |
2529 | ||
2530 | for (pos = *posp; pos && obj && !IS_ERR(obj); pos--) | |
2531 | obj = __netlink_seq_next(seq); | |
2532 | ||
2533 | return obj; | |
2534 | } | |
2535 | ||
2536 | static void *netlink_seq_next(struct seq_file *seq, void *v, loff_t *pos) | |
2537 | { | |
2538 | ++*pos; | |
2539 | return __netlink_seq_next(seq); | |
1da177e4 LT |
2540 | } |
2541 | ||
2542 | static void netlink_seq_stop(struct seq_file *seq, void *v) | |
2543 | { | |
56d28b1e HX |
2544 | struct nl_seq_iter *iter = seq->private; |
2545 | ||
2546 | if (iter->link >= MAX_LINKS) | |
2547 | return; | |
2548 | ||
2549 | netlink_walk_stop(iter); | |
1da177e4 LT |
2550 | } |
2551 | ||
2552 | ||
2553 | static int netlink_seq_show(struct seq_file *seq, void *v) | |
2554 | { | |
658cb354 | 2555 | if (v == SEQ_START_TOKEN) { |
1da177e4 LT |
2556 | seq_puts(seq, |
2557 | "sk Eth Pid Groups " | |
cf0aa4e0 | 2558 | "Rmem Wmem Dump Locks Drops Inode\n"); |
658cb354 | 2559 | } else { |
1da177e4 LT |
2560 | struct sock *s = v; |
2561 | struct netlink_sock *nlk = nlk_sk(s); | |
2562 | ||
16b304f3 | 2563 | seq_printf(seq, "%pK %-3d %-6u %08x %-8d %-8d %d %-8d %-8d %-8lu\n", |
1da177e4 LT |
2564 | s, |
2565 | s->sk_protocol, | |
15e47304 | 2566 | nlk->portid, |
513c2500 | 2567 | nlk->groups ? (u32)nlk->groups[0] : 0, |
31e6d363 ED |
2568 | sk_rmem_alloc_get(s), |
2569 | sk_wmem_alloc_get(s), | |
16b304f3 | 2570 | nlk->cb_running, |
41c6d650 | 2571 | refcount_read(&s->sk_refcnt), |
cf0aa4e0 MY |
2572 | atomic_read(&s->sk_drops), |
2573 | sock_i_ino(s) | |
1da177e4 LT |
2574 | ); |
2575 | ||
2576 | } | |
2577 | return 0; | |
2578 | } | |
2579 | ||
56b3d975 | 2580 | static const struct seq_operations netlink_seq_ops = { |
1da177e4 LT |
2581 | .start = netlink_seq_start, |
2582 | .next = netlink_seq_next, | |
2583 | .stop = netlink_seq_stop, | |
2584 | .show = netlink_seq_show, | |
2585 | }; | |
2586 | ||
2587 | ||
2588 | static int netlink_seq_open(struct inode *inode, struct file *file) | |
2589 | { | |
e372c414 DL |
2590 | return seq_open_net(inode, file, &netlink_seq_ops, |
2591 | sizeof(struct nl_seq_iter)); | |
b4b51029 EB |
2592 | } |
2593 | ||
da7071d7 | 2594 | static const struct file_operations netlink_seq_fops = { |
1da177e4 LT |
2595 | .owner = THIS_MODULE, |
2596 | .open = netlink_seq_open, | |
2597 | .read = seq_read, | |
2598 | .llseek = seq_lseek, | |
e372c414 | 2599 | .release = seq_release_net, |
1da177e4 LT |
2600 | }; |
2601 | ||
2602 | #endif | |
2603 | ||
2604 | int netlink_register_notifier(struct notifier_block *nb) | |
2605 | { | |
efa172f4 | 2606 | return blocking_notifier_chain_register(&netlink_chain, nb); |
1da177e4 | 2607 | } |
6ac552fd | 2608 | EXPORT_SYMBOL(netlink_register_notifier); |
1da177e4 LT |
2609 | |
2610 | int netlink_unregister_notifier(struct notifier_block *nb) | |
2611 | { | |
efa172f4 | 2612 | return blocking_notifier_chain_unregister(&netlink_chain, nb); |
1da177e4 | 2613 | } |
6ac552fd | 2614 | EXPORT_SYMBOL(netlink_unregister_notifier); |
746fac4d | 2615 | |
90ddc4f0 | 2616 | static const struct proto_ops netlink_ops = { |
1da177e4 LT |
2617 | .family = PF_NETLINK, |
2618 | .owner = THIS_MODULE, | |
2619 | .release = netlink_release, | |
2620 | .bind = netlink_bind, | |
2621 | .connect = netlink_connect, | |
2622 | .socketpair = sock_no_socketpair, | |
2623 | .accept = sock_no_accept, | |
2624 | .getname = netlink_getname, | |
d1b4c689 | 2625 | .poll = datagram_poll, |
025c6818 | 2626 | .ioctl = netlink_ioctl, |
1da177e4 LT |
2627 | .listen = sock_no_listen, |
2628 | .shutdown = sock_no_shutdown, | |
9a4595bc PM |
2629 | .setsockopt = netlink_setsockopt, |
2630 | .getsockopt = netlink_getsockopt, | |
1da177e4 LT |
2631 | .sendmsg = netlink_sendmsg, |
2632 | .recvmsg = netlink_recvmsg, | |
d1b4c689 | 2633 | .mmap = sock_no_mmap, |
1da177e4 LT |
2634 | .sendpage = sock_no_sendpage, |
2635 | }; | |
2636 | ||
ec1b4cf7 | 2637 | static const struct net_proto_family netlink_family_ops = { |
1da177e4 LT |
2638 | .family = PF_NETLINK, |
2639 | .create = netlink_create, | |
2640 | .owner = THIS_MODULE, /* for consistency 8) */ | |
2641 | }; | |
2642 | ||
4665079c | 2643 | static int __net_init netlink_net_init(struct net *net) |
b4b51029 EB |
2644 | { |
2645 | #ifdef CONFIG_PROC_FS | |
d4beaa66 | 2646 | if (!proc_create("netlink", 0, net->proc_net, &netlink_seq_fops)) |
b4b51029 EB |
2647 | return -ENOMEM; |
2648 | #endif | |
2649 | return 0; | |
2650 | } | |
2651 | ||
4665079c | 2652 | static void __net_exit netlink_net_exit(struct net *net) |
b4b51029 EB |
2653 | { |
2654 | #ifdef CONFIG_PROC_FS | |
ece31ffd | 2655 | remove_proc_entry("netlink", net->proc_net); |
b4b51029 EB |
2656 | #endif |
2657 | } | |
2658 | ||
b963ea89 DM |
2659 | static void __init netlink_add_usersock_entry(void) |
2660 | { | |
5c398dc8 | 2661 | struct listeners *listeners; |
b963ea89 DM |
2662 | int groups = 32; |
2663 | ||
5c398dc8 | 2664 | listeners = kzalloc(sizeof(*listeners) + NLGRPSZ(groups), GFP_KERNEL); |
b963ea89 | 2665 | if (!listeners) |
5c398dc8 | 2666 | panic("netlink_add_usersock_entry: Cannot allocate listeners\n"); |
b963ea89 DM |
2667 | |
2668 | netlink_table_grab(); | |
2669 | ||
2670 | nl_table[NETLINK_USERSOCK].groups = groups; | |
5c398dc8 | 2671 | rcu_assign_pointer(nl_table[NETLINK_USERSOCK].listeners, listeners); |
b963ea89 DM |
2672 | nl_table[NETLINK_USERSOCK].module = THIS_MODULE; |
2673 | nl_table[NETLINK_USERSOCK].registered = 1; | |
9785e10a | 2674 | nl_table[NETLINK_USERSOCK].flags = NL_CFG_F_NONROOT_SEND; |
b963ea89 DM |
2675 | |
2676 | netlink_table_ungrab(); | |
2677 | } | |
2678 | ||
022cbae6 | 2679 | static struct pernet_operations __net_initdata netlink_net_ops = { |
b4b51029 EB |
2680 | .init = netlink_net_init, |
2681 | .exit = netlink_net_exit, | |
2682 | }; | |
2683 | ||
49f7b33e | 2684 | static inline u32 netlink_hash(const void *data, u32 len, u32 seed) |
c428ecd1 HX |
2685 | { |
2686 | const struct netlink_sock *nlk = data; | |
2687 | struct netlink_compare_arg arg; | |
2688 | ||
da314c99 | 2689 | netlink_compare_arg_init(&arg, sock_net(&nlk->sk), nlk->portid); |
11b58ba1 | 2690 | return jhash2((u32 *)&arg, netlink_compare_arg_len / sizeof(u32), seed); |
c428ecd1 HX |
2691 | } |
2692 | ||
2693 | static const struct rhashtable_params netlink_rhashtable_params = { | |
2694 | .head_offset = offsetof(struct netlink_sock, node), | |
2695 | .key_len = netlink_compare_arg_len, | |
c428ecd1 HX |
2696 | .obj_hashfn = netlink_hash, |
2697 | .obj_cmpfn = netlink_compare, | |
b5e2c150 | 2698 | .automatic_shrinking = true, |
c428ecd1 HX |
2699 | }; |
2700 | ||
1da177e4 LT |
2701 | static int __init netlink_proto_init(void) |
2702 | { | |
1da177e4 | 2703 | int i; |
1da177e4 LT |
2704 | int err = proto_register(&netlink_proto, 0); |
2705 | ||
2706 | if (err != 0) | |
2707 | goto out; | |
2708 | ||
fab25745 | 2709 | BUILD_BUG_ON(sizeof(struct netlink_skb_parms) > FIELD_SIZEOF(struct sk_buff, cb)); |
1da177e4 | 2710 | |
0da974f4 | 2711 | nl_table = kcalloc(MAX_LINKS, sizeof(*nl_table), GFP_KERNEL); |
fab2caf6 AM |
2712 | if (!nl_table) |
2713 | goto panic; | |
1da177e4 | 2714 | |
1da177e4 | 2715 | for (i = 0; i < MAX_LINKS; i++) { |
c428ecd1 HX |
2716 | if (rhashtable_init(&nl_table[i].hash, |
2717 | &netlink_rhashtable_params) < 0) { | |
e341694e TG |
2718 | while (--i > 0) |
2719 | rhashtable_destroy(&nl_table[i].hash); | |
1da177e4 | 2720 | kfree(nl_table); |
fab2caf6 | 2721 | goto panic; |
1da177e4 | 2722 | } |
1da177e4 LT |
2723 | } |
2724 | ||
bcbde0d4 DB |
2725 | INIT_LIST_HEAD(&netlink_tap_all); |
2726 | ||
b963ea89 DM |
2727 | netlink_add_usersock_entry(); |
2728 | ||
1da177e4 | 2729 | sock_register(&netlink_family_ops); |
b4b51029 | 2730 | register_pernet_subsys(&netlink_net_ops); |
746fac4d | 2731 | /* The netlink device handler may be needed early. */ |
1da177e4 LT |
2732 | rtnetlink_init(); |
2733 | out: | |
2734 | return err; | |
fab2caf6 AM |
2735 | panic: |
2736 | panic("netlink_init: Cannot allocate nl_table\n"); | |
1da177e4 LT |
2737 | } |
2738 | ||
1da177e4 | 2739 | core_initcall(netlink_proto_init); |