Commit | Line | Data |
---|---|---|
2874c5fd | 1 | // SPDX-License-Identifier: GPL-2.0-or-later |
1da177e4 LT |
2 | /* |
3 | * NETLINK Kernel-user communication protocol. | |
4 | * | |
113aa838 | 5 | * Authors: Alan Cox <alan@lxorguk.ukuu.org.uk> |
1da177e4 | 6 | * Alexey Kuznetsov <kuznet@ms2.inr.ac.ru> |
cd1df525 | 7 | * Patrick McHardy <kaber@trash.net> |
1da177e4 | 8 | * |
1da177e4 LT |
9 | * Tue Jun 26 14:36:48 MEST 2001 Herbert "herp" Rosmanith |
10 | * added netlink_proto_exit | |
11 | * Tue Jan 22 18:32:44 BRST 2002 Arnaldo C. de Melo <acme@conectiva.com.br> | |
12 | * use nlk_sk, as sk->protinfo is on a diet 8) | |
4fdb3bb7 HW |
13 | * Fri Jul 22 19:51:12 MEST 2005 Harald Welte <laforge@gnumonks.org> |
14 | * - inc module use count of module that owns | |
15 | * the kernel socket in case userspace opens | |
16 | * socket of same protocol | |
17 | * - remove all module support, since netlink is | |
18 | * mandatory if CONFIG_NET=y these days | |
1da177e4 LT |
19 | */ |
20 | ||
1da177e4 LT |
21 | #include <linux/module.h> |
22 | ||
4fc268d2 | 23 | #include <linux/capability.h> |
1da177e4 LT |
24 | #include <linux/kernel.h> |
25 | #include <linux/init.h> | |
1da177e4 LT |
26 | #include <linux/signal.h> |
27 | #include <linux/sched.h> | |
28 | #include <linux/errno.h> | |
29 | #include <linux/string.h> | |
30 | #include <linux/stat.h> | |
31 | #include <linux/socket.h> | |
32 | #include <linux/un.h> | |
33 | #include <linux/fcntl.h> | |
34 | #include <linux/termios.h> | |
35 | #include <linux/sockios.h> | |
36 | #include <linux/net.h> | |
37 | #include <linux/fs.h> | |
38 | #include <linux/slab.h> | |
7c0f6ba6 | 39 | #include <linux/uaccess.h> |
1da177e4 LT |
40 | #include <linux/skbuff.h> |
41 | #include <linux/netdevice.h> | |
42 | #include <linux/rtnetlink.h> | |
43 | #include <linux/proc_fs.h> | |
44 | #include <linux/seq_file.h> | |
1da177e4 LT |
45 | #include <linux/notifier.h> |
46 | #include <linux/security.h> | |
47 | #include <linux/jhash.h> | |
48 | #include <linux/jiffies.h> | |
49 | #include <linux/random.h> | |
50 | #include <linux/bitops.h> | |
51 | #include <linux/mm.h> | |
52 | #include <linux/types.h> | |
54e0f520 | 53 | #include <linux/audit.h> |
af65bdfc | 54 | #include <linux/mutex.h> |
ccdfcc39 | 55 | #include <linux/vmalloc.h> |
bcbde0d4 | 56 | #include <linux/if_arp.h> |
e341694e | 57 | #include <linux/rhashtable.h> |
9652e931 | 58 | #include <asm/cacheflush.h> |
e341694e | 59 | #include <linux/hash.h> |
ee1c2442 | 60 | #include <linux/genetlink.h> |
7212462f | 61 | #include <linux/net_namespace.h> |
bc5b6c0b | 62 | #include <linux/nospec.h> |
54e0f520 | 63 | |
457c4cbc | 64 | #include <net/net_namespace.h> |
25e3f70f | 65 | #include <net/netns/generic.h> |
1da177e4 LT |
66 | #include <net/sock.h> |
67 | #include <net/scm.h> | |
82ace47a | 68 | #include <net/netlink.h> |
1da177e4 | 69 | |
0f29c768 | 70 | #include "af_netlink.h" |
1da177e4 | 71 | |
5c398dc8 ED |
72 | struct listeners { |
73 | struct rcu_head rcu; | |
74 | unsigned long masks[0]; | |
6c04bb18 JB |
75 | }; |
76 | ||
cd967e05 | 77 | /* state bits */ |
cc3a572f | 78 | #define NETLINK_S_CONGESTED 0x0 |
cd967e05 | 79 | |
035c4c16 | 80 | static inline int netlink_is_kernel(struct sock *sk) |
aed81560 | 81 | { |
cc3a572f | 82 | return nlk_sk(sk)->flags & NETLINK_F_KERNEL_SOCKET; |
aed81560 DL |
83 | } |
84 | ||
91dd93f9 | 85 | struct netlink_table *nl_table __read_mostly; |
0f29c768 | 86 | EXPORT_SYMBOL_GPL(nl_table); |
1da177e4 LT |
87 | |
88 | static DECLARE_WAIT_QUEUE_HEAD(nl_table_wait); | |
89 | ||
8a0f5ccf HX |
90 | static struct lock_class_key nlk_cb_mutex_keys[MAX_LINKS]; |
91 | ||
92 | static const char *const nlk_cb_mutex_key_strings[MAX_LINKS + 1] = { | |
93 | "nlk_cb_mutex-ROUTE", | |
94 | "nlk_cb_mutex-1", | |
95 | "nlk_cb_mutex-USERSOCK", | |
96 | "nlk_cb_mutex-FIREWALL", | |
97 | "nlk_cb_mutex-SOCK_DIAG", | |
98 | "nlk_cb_mutex-NFLOG", | |
99 | "nlk_cb_mutex-XFRM", | |
100 | "nlk_cb_mutex-SELINUX", | |
101 | "nlk_cb_mutex-ISCSI", | |
102 | "nlk_cb_mutex-AUDIT", | |
103 | "nlk_cb_mutex-FIB_LOOKUP", | |
104 | "nlk_cb_mutex-CONNECTOR", | |
105 | "nlk_cb_mutex-NETFILTER", | |
106 | "nlk_cb_mutex-IP6_FW", | |
107 | "nlk_cb_mutex-DNRTMSG", | |
108 | "nlk_cb_mutex-KOBJECT_UEVENT", | |
109 | "nlk_cb_mutex-GENERIC", | |
110 | "nlk_cb_mutex-17", | |
111 | "nlk_cb_mutex-SCSITRANSPORT", | |
112 | "nlk_cb_mutex-ECRYPTFS", | |
113 | "nlk_cb_mutex-RDMA", | |
114 | "nlk_cb_mutex-CRYPTO", | |
115 | "nlk_cb_mutex-SMC", | |
116 | "nlk_cb_mutex-23", | |
117 | "nlk_cb_mutex-24", | |
118 | "nlk_cb_mutex-25", | |
119 | "nlk_cb_mutex-26", | |
120 | "nlk_cb_mutex-27", | |
121 | "nlk_cb_mutex-28", | |
122 | "nlk_cb_mutex-29", | |
123 | "nlk_cb_mutex-30", | |
124 | "nlk_cb_mutex-31", | |
125 | "nlk_cb_mutex-MAX_LINKS" | |
126 | }; | |
127 | ||
1da177e4 | 128 | static int netlink_dump(struct sock *sk); |
1da177e4 | 129 | |
78fd1d0a | 130 | /* nl_table locking explained: |
21e4902a | 131 | * Lookup and traversal are protected with an RCU read-side lock. Insertion |
c5adde94 | 132 | * and removal are protected with per bucket lock while using RCU list |
21e4902a TG |
133 | * modification primitives and may run in parallel to RCU protected lookups. |
134 | * Destruction of the Netlink socket may only occur *after* nl_table_lock has | |
135 | * been acquired * either during or after the socket has been removed from | |
136 | * the list and after an RCU grace period. | |
78fd1d0a | 137 | */ |
0f29c768 AV |
138 | DEFINE_RWLOCK(nl_table_lock); |
139 | EXPORT_SYMBOL_GPL(nl_table_lock); | |
1da177e4 LT |
140 | static atomic_t nl_table_users = ATOMIC_INIT(0); |
141 | ||
6d772ac5 ED |
142 | #define nl_deref_protected(X) rcu_dereference_protected(X, lockdep_is_held(&nl_table_lock)); |
143 | ||
efa172f4 | 144 | static BLOCKING_NOTIFIER_HEAD(netlink_chain); |
1da177e4 | 145 | |
bcbde0d4 | 146 | |
c428ecd1 HX |
147 | static const struct rhashtable_params netlink_rhashtable_params; |
148 | ||
b57ef81f | 149 | static inline u32 netlink_group_mask(u32 group) |
d629b836 PM |
150 | { |
151 | return group ? 1 << (group - 1) : 0; | |
152 | } | |
153 | ||
1853c949 DB |
154 | static struct sk_buff *netlink_to_full_skb(const struct sk_buff *skb, |
155 | gfp_t gfp_mask) | |
156 | { | |
157 | unsigned int len = skb_end_offset(skb); | |
158 | struct sk_buff *new; | |
159 | ||
160 | new = alloc_skb(len, gfp_mask); | |
161 | if (new == NULL) | |
162 | return NULL; | |
163 | ||
164 | NETLINK_CB(new).portid = NETLINK_CB(skb).portid; | |
165 | NETLINK_CB(new).dst_group = NETLINK_CB(skb).dst_group; | |
166 | NETLINK_CB(new).creds = NETLINK_CB(skb).creds; | |
167 | ||
59ae1d12 | 168 | skb_put_data(new, skb->data, len); |
1853c949 DB |
169 | return new; |
170 | } | |
171 | ||
25e3f70f CW |
172 | static unsigned int netlink_tap_net_id; |
173 | ||
174 | struct netlink_tap_net { | |
175 | struct list_head netlink_tap_all; | |
b1042d35 | 176 | struct mutex netlink_tap_lock; |
25e3f70f CW |
177 | }; |
178 | ||
bcbde0d4 DB |
179 | int netlink_add_tap(struct netlink_tap *nt) |
180 | { | |
25e3f70f CW |
181 | struct net *net = dev_net(nt->dev); |
182 | struct netlink_tap_net *nn = net_generic(net, netlink_tap_net_id); | |
183 | ||
bcbde0d4 DB |
184 | if (unlikely(nt->dev->type != ARPHRD_NETLINK)) |
185 | return -EINVAL; | |
186 | ||
b1042d35 | 187 | mutex_lock(&nn->netlink_tap_lock); |
25e3f70f | 188 | list_add_rcu(&nt->list, &nn->netlink_tap_all); |
b1042d35 | 189 | mutex_unlock(&nn->netlink_tap_lock); |
bcbde0d4 | 190 | |
fcd4d35e | 191 | __module_get(nt->module); |
bcbde0d4 DB |
192 | |
193 | return 0; | |
194 | } | |
195 | EXPORT_SYMBOL_GPL(netlink_add_tap); | |
196 | ||
2173f8d9 | 197 | static int __netlink_remove_tap(struct netlink_tap *nt) |
bcbde0d4 | 198 | { |
25e3f70f CW |
199 | struct net *net = dev_net(nt->dev); |
200 | struct netlink_tap_net *nn = net_generic(net, netlink_tap_net_id); | |
bcbde0d4 DB |
201 | bool found = false; |
202 | struct netlink_tap *tmp; | |
203 | ||
b1042d35 | 204 | mutex_lock(&nn->netlink_tap_lock); |
bcbde0d4 | 205 | |
25e3f70f | 206 | list_for_each_entry(tmp, &nn->netlink_tap_all, list) { |
bcbde0d4 DB |
207 | if (nt == tmp) { |
208 | list_del_rcu(&nt->list); | |
209 | found = true; | |
210 | goto out; | |
211 | } | |
212 | } | |
213 | ||
214 | pr_warn("__netlink_remove_tap: %p not found\n", nt); | |
215 | out: | |
b1042d35 | 216 | mutex_unlock(&nn->netlink_tap_lock); |
bcbde0d4 | 217 | |
92b80eb3 | 218 | if (found) |
bcbde0d4 DB |
219 | module_put(nt->module); |
220 | ||
221 | return found ? 0 : -ENODEV; | |
222 | } | |
bcbde0d4 DB |
223 | |
224 | int netlink_remove_tap(struct netlink_tap *nt) | |
225 | { | |
226 | int ret; | |
227 | ||
228 | ret = __netlink_remove_tap(nt); | |
229 | synchronize_net(); | |
230 | ||
231 | return ret; | |
232 | } | |
233 | EXPORT_SYMBOL_GPL(netlink_remove_tap); | |
234 | ||
25e3f70f CW |
235 | static __net_init int netlink_tap_init_net(struct net *net) |
236 | { | |
237 | struct netlink_tap_net *nn = net_generic(net, netlink_tap_net_id); | |
238 | ||
239 | INIT_LIST_HEAD(&nn->netlink_tap_all); | |
b1042d35 | 240 | mutex_init(&nn->netlink_tap_lock); |
25e3f70f CW |
241 | return 0; |
242 | } | |
243 | ||
25e3f70f CW |
244 | static struct pernet_operations netlink_tap_net_ops = { |
245 | .init = netlink_tap_init_net, | |
25e3f70f CW |
246 | .id = &netlink_tap_net_id, |
247 | .size = sizeof(struct netlink_tap_net), | |
248 | }; | |
249 | ||
5ffd5cdd DB |
250 | static bool netlink_filter_tap(const struct sk_buff *skb) |
251 | { | |
252 | struct sock *sk = skb->sk; | |
5ffd5cdd DB |
253 | |
254 | /* We take the more conservative approach and | |
255 | * whitelist socket protocols that may pass. | |
256 | */ | |
257 | switch (sk->sk_protocol) { | |
258 | case NETLINK_ROUTE: | |
259 | case NETLINK_USERSOCK: | |
260 | case NETLINK_SOCK_DIAG: | |
261 | case NETLINK_NFLOG: | |
262 | case NETLINK_XFRM: | |
263 | case NETLINK_FIB_LOOKUP: | |
264 | case NETLINK_NETFILTER: | |
265 | case NETLINK_GENERIC: | |
498044bb | 266 | return true; |
5ffd5cdd DB |
267 | } |
268 | ||
498044bb | 269 | return false; |
5ffd5cdd DB |
270 | } |
271 | ||
bcbde0d4 DB |
272 | static int __netlink_deliver_tap_skb(struct sk_buff *skb, |
273 | struct net_device *dev) | |
274 | { | |
275 | struct sk_buff *nskb; | |
5ffd5cdd | 276 | struct sock *sk = skb->sk; |
bcbde0d4 DB |
277 | int ret = -ENOMEM; |
278 | ||
93c64764 KC |
279 | if (!net_eq(dev_net(dev), sock_net(sk))) |
280 | return 0; | |
281 | ||
bcbde0d4 | 282 | dev_hold(dev); |
1853c949 | 283 | |
d1b4c689 | 284 | if (is_vmalloc_addr(skb->head)) |
1853c949 DB |
285 | nskb = netlink_to_full_skb(skb, GFP_ATOMIC); |
286 | else | |
287 | nskb = skb_clone(skb, GFP_ATOMIC); | |
bcbde0d4 DB |
288 | if (nskb) { |
289 | nskb->dev = dev; | |
5ffd5cdd | 290 | nskb->protocol = htons((u16) sk->sk_protocol); |
604d13c9 DB |
291 | nskb->pkt_type = netlink_is_kernel(sk) ? |
292 | PACKET_KERNEL : PACKET_USER; | |
4e48ed88 | 293 | skb_reset_network_header(nskb); |
bcbde0d4 DB |
294 | ret = dev_queue_xmit(nskb); |
295 | if (unlikely(ret > 0)) | |
296 | ret = net_xmit_errno(ret); | |
297 | } | |
298 | ||
299 | dev_put(dev); | |
300 | return ret; | |
301 | } | |
302 | ||
25e3f70f | 303 | static void __netlink_deliver_tap(struct sk_buff *skb, struct netlink_tap_net *nn) |
bcbde0d4 DB |
304 | { |
305 | int ret; | |
306 | struct netlink_tap *tmp; | |
307 | ||
5ffd5cdd DB |
308 | if (!netlink_filter_tap(skb)) |
309 | return; | |
310 | ||
25e3f70f | 311 | list_for_each_entry_rcu(tmp, &nn->netlink_tap_all, list) { |
bcbde0d4 DB |
312 | ret = __netlink_deliver_tap_skb(skb, tmp->dev); |
313 | if (unlikely(ret)) | |
314 | break; | |
315 | } | |
316 | } | |
317 | ||
25e3f70f | 318 | static void netlink_deliver_tap(struct net *net, struct sk_buff *skb) |
bcbde0d4 | 319 | { |
25e3f70f CW |
320 | struct netlink_tap_net *nn = net_generic(net, netlink_tap_net_id); |
321 | ||
bcbde0d4 DB |
322 | rcu_read_lock(); |
323 | ||
25e3f70f CW |
324 | if (unlikely(!list_empty(&nn->netlink_tap_all))) |
325 | __netlink_deliver_tap(skb, nn); | |
bcbde0d4 DB |
326 | |
327 | rcu_read_unlock(); | |
328 | } | |
329 | ||
73bfd370 DB |
330 | static void netlink_deliver_tap_kernel(struct sock *dst, struct sock *src, |
331 | struct sk_buff *skb) | |
332 | { | |
333 | if (!(netlink_is_kernel(dst) && netlink_is_kernel(src))) | |
25e3f70f | 334 | netlink_deliver_tap(sock_net(dst), skb); |
73bfd370 DB |
335 | } |
336 | ||
cd1df525 PM |
337 | static void netlink_overrun(struct sock *sk) |
338 | { | |
339 | struct netlink_sock *nlk = nlk_sk(sk); | |
340 | ||
cc3a572f ND |
341 | if (!(nlk->flags & NETLINK_F_RECV_NO_ENOBUFS)) { |
342 | if (!test_and_set_bit(NETLINK_S_CONGESTED, | |
343 | &nlk_sk(sk)->state)) { | |
cd1df525 PM |
344 | sk->sk_err = ENOBUFS; |
345 | sk->sk_error_report(sk); | |
346 | } | |
347 | } | |
348 | atomic_inc(&sk->sk_drops); | |
349 | } | |
350 | ||
351 | static void netlink_rcv_wake(struct sock *sk) | |
352 | { | |
353 | struct netlink_sock *nlk = nlk_sk(sk); | |
354 | ||
355 | if (skb_queue_empty(&sk->sk_receive_queue)) | |
cc3a572f ND |
356 | clear_bit(NETLINK_S_CONGESTED, &nlk->state); |
357 | if (!test_bit(NETLINK_S_CONGESTED, &nlk->state)) | |
cd1df525 PM |
358 | wake_up_interruptible(&nlk->wait); |
359 | } | |
360 | ||
cf0a018a PM |
361 | static void netlink_skb_destructor(struct sk_buff *skb) |
362 | { | |
c05cdb1b | 363 | if (is_vmalloc_addr(skb->head)) { |
3a36515f PN |
364 | if (!skb->cloned || |
365 | !atomic_dec_return(&(skb_shinfo(skb)->dataref))) | |
366 | vfree(skb->head); | |
367 | ||
c05cdb1b PNA |
368 | skb->head = NULL; |
369 | } | |
9652e931 PM |
370 | if (skb->sk != NULL) |
371 | sock_rfree(skb); | |
cf0a018a PM |
372 | } |
373 | ||
374 | static void netlink_skb_set_owner_r(struct sk_buff *skb, struct sock *sk) | |
375 | { | |
376 | WARN_ON(skb->sk != NULL); | |
377 | skb->sk = sk; | |
378 | skb->destructor = netlink_skb_destructor; | |
379 | atomic_add(skb->truesize, &sk->sk_rmem_alloc); | |
380 | sk_mem_charge(sk, skb->truesize); | |
381 | } | |
382 | ||
ed5d7788 | 383 | static void netlink_sock_destruct(struct sock *sk) |
1da177e4 | 384 | { |
3f660d66 HX |
385 | struct netlink_sock *nlk = nlk_sk(sk); |
386 | ||
16b304f3 | 387 | if (nlk->cb_running) { |
ed5d7788 HX |
388 | if (nlk->cb.done) |
389 | nlk->cb.done(&nlk->cb); | |
16b304f3 PS |
390 | module_put(nlk->cb.module); |
391 | kfree_skb(nlk->cb.skb); | |
3f660d66 HX |
392 | } |
393 | ||
1da177e4 LT |
394 | skb_queue_purge(&sk->sk_receive_queue); |
395 | ||
396 | if (!sock_flag(sk, SOCK_DEAD)) { | |
6ac552fd | 397 | printk(KERN_ERR "Freeing alive netlink socket %p\n", sk); |
1da177e4 LT |
398 | return; |
399 | } | |
547b792c IJ |
400 | |
401 | WARN_ON(atomic_read(&sk->sk_rmem_alloc)); | |
14afee4b | 402 | WARN_ON(refcount_read(&sk->sk_wmem_alloc)); |
547b792c | 403 | WARN_ON(nlk_sk(sk)->groups); |
1da177e4 LT |
404 | } |
405 | ||
707693c8 HX |
406 | static void netlink_sock_destruct_work(struct work_struct *work) |
407 | { | |
408 | struct netlink_sock *nlk = container_of(work, struct netlink_sock, | |
409 | work); | |
410 | ||
ed5d7788 | 411 | sk_free(&nlk->sk); |
707693c8 HX |
412 | } |
413 | ||
6ac552fd PM |
414 | /* This lock without WQ_FLAG_EXCLUSIVE is good on UP and it is _very_ bad on |
415 | * SMP. Look, when several writers sleep and reader wakes them up, all but one | |
1da177e4 LT |
416 | * immediately hit write lock and grab all the cpus. Exclusive sleep solves |
417 | * this, _but_ remember, it adds useless work on UP machines. | |
418 | */ | |
419 | ||
d136f1bd | 420 | void netlink_table_grab(void) |
9a429c49 | 421 | __acquires(nl_table_lock) |
1da177e4 | 422 | { |
d136f1bd JB |
423 | might_sleep(); |
424 | ||
6abd219c | 425 | write_lock_irq(&nl_table_lock); |
1da177e4 LT |
426 | |
427 | if (atomic_read(&nl_table_users)) { | |
428 | DECLARE_WAITQUEUE(wait, current); | |
429 | ||
430 | add_wait_queue_exclusive(&nl_table_wait, &wait); | |
6ac552fd | 431 | for (;;) { |
1da177e4 LT |
432 | set_current_state(TASK_UNINTERRUPTIBLE); |
433 | if (atomic_read(&nl_table_users) == 0) | |
434 | break; | |
6abd219c | 435 | write_unlock_irq(&nl_table_lock); |
1da177e4 | 436 | schedule(); |
6abd219c | 437 | write_lock_irq(&nl_table_lock); |
1da177e4 LT |
438 | } |
439 | ||
440 | __set_current_state(TASK_RUNNING); | |
441 | remove_wait_queue(&nl_table_wait, &wait); | |
442 | } | |
443 | } | |
444 | ||
d136f1bd | 445 | void netlink_table_ungrab(void) |
9a429c49 | 446 | __releases(nl_table_lock) |
1da177e4 | 447 | { |
6abd219c | 448 | write_unlock_irq(&nl_table_lock); |
1da177e4 LT |
449 | wake_up(&nl_table_wait); |
450 | } | |
451 | ||
6ac552fd | 452 | static inline void |
1da177e4 LT |
453 | netlink_lock_table(void) |
454 | { | |
455 | /* read_lock() synchronizes us to netlink_table_grab */ | |
456 | ||
457 | read_lock(&nl_table_lock); | |
458 | atomic_inc(&nl_table_users); | |
459 | read_unlock(&nl_table_lock); | |
460 | } | |
461 | ||
6ac552fd | 462 | static inline void |
1da177e4 LT |
463 | netlink_unlock_table(void) |
464 | { | |
465 | if (atomic_dec_and_test(&nl_table_users)) | |
466 | wake_up(&nl_table_wait); | |
467 | } | |
468 | ||
e341694e | 469 | struct netlink_compare_arg |
1da177e4 | 470 | { |
c428ecd1 | 471 | possible_net_t pnet; |
e341694e TG |
472 | u32 portid; |
473 | }; | |
1da177e4 | 474 | |
8f2ddaac HX |
475 | /* Doing sizeof directly may yield 4 extra bytes on 64-bit. */ |
476 | #define netlink_compare_arg_len \ | |
477 | (offsetof(struct netlink_compare_arg, portid) + sizeof(u32)) | |
c428ecd1 HX |
478 | |
479 | static inline int netlink_compare(struct rhashtable_compare_arg *arg, | |
480 | const void *ptr) | |
1da177e4 | 481 | { |
c428ecd1 HX |
482 | const struct netlink_compare_arg *x = arg->key; |
483 | const struct netlink_sock *nlk = ptr; | |
1da177e4 | 484 | |
da314c99 | 485 | return nlk->portid != x->portid || |
c428ecd1 HX |
486 | !net_eq(sock_net(&nlk->sk), read_pnet(&x->pnet)); |
487 | } | |
488 | ||
489 | static void netlink_compare_arg_init(struct netlink_compare_arg *arg, | |
490 | struct net *net, u32 portid) | |
491 | { | |
492 | memset(arg, 0, sizeof(*arg)); | |
493 | write_pnet(&arg->pnet, net); | |
494 | arg->portid = portid; | |
1da177e4 LT |
495 | } |
496 | ||
e341694e TG |
497 | static struct sock *__netlink_lookup(struct netlink_table *table, u32 portid, |
498 | struct net *net) | |
1da177e4 | 499 | { |
c428ecd1 | 500 | struct netlink_compare_arg arg; |
1da177e4 | 501 | |
c428ecd1 HX |
502 | netlink_compare_arg_init(&arg, net, portid); |
503 | return rhashtable_lookup_fast(&table->hash, &arg, | |
504 | netlink_rhashtable_params); | |
1da177e4 LT |
505 | } |
506 | ||
c428ecd1 | 507 | static int __netlink_insert(struct netlink_table *table, struct sock *sk) |
c5adde94 | 508 | { |
c428ecd1 | 509 | struct netlink_compare_arg arg; |
c5adde94 | 510 | |
da314c99 | 511 | netlink_compare_arg_init(&arg, sock_net(sk), nlk_sk(sk)->portid); |
c428ecd1 HX |
512 | return rhashtable_lookup_insert_key(&table->hash, &arg, |
513 | &nlk_sk(sk)->node, | |
514 | netlink_rhashtable_params); | |
c5adde94 YX |
515 | } |
516 | ||
e341694e | 517 | static struct sock *netlink_lookup(struct net *net, int protocol, u32 portid) |
1da177e4 | 518 | { |
e341694e TG |
519 | struct netlink_table *table = &nl_table[protocol]; |
520 | struct sock *sk; | |
1da177e4 | 521 | |
e341694e TG |
522 | rcu_read_lock(); |
523 | sk = __netlink_lookup(table, portid, net); | |
524 | if (sk) | |
525 | sock_hold(sk); | |
526 | rcu_read_unlock(); | |
1da177e4 | 527 | |
e341694e | 528 | return sk; |
1da177e4 LT |
529 | } |
530 | ||
90ddc4f0 | 531 | static const struct proto_ops netlink_ops; |
1da177e4 | 532 | |
4277a083 PM |
533 | static void |
534 | netlink_update_listeners(struct sock *sk) | |
535 | { | |
536 | struct netlink_table *tbl = &nl_table[sk->sk_protocol]; | |
4277a083 PM |
537 | unsigned long mask; |
538 | unsigned int i; | |
6d772ac5 ED |
539 | struct listeners *listeners; |
540 | ||
541 | listeners = nl_deref_protected(tbl->listeners); | |
542 | if (!listeners) | |
543 | return; | |
4277a083 | 544 | |
b4ff4f04 | 545 | for (i = 0; i < NLGRPLONGS(tbl->groups); i++) { |
4277a083 | 546 | mask = 0; |
b67bfe0d | 547 | sk_for_each_bound(sk, &tbl->mc_list) { |
b4ff4f04 JB |
548 | if (i < NLGRPLONGS(nlk_sk(sk)->ngroups)) |
549 | mask |= nlk_sk(sk)->groups[i]; | |
550 | } | |
6d772ac5 | 551 | listeners->masks[i] = mask; |
4277a083 PM |
552 | } |
553 | /* this function is only called with the netlink table "grabbed", which | |
554 | * makes sure updates are visible before bind or setsockopt return. */ | |
555 | } | |
556 | ||
8ea65f4a | 557 | static int netlink_insert(struct sock *sk, u32 portid) |
1da177e4 | 558 | { |
da12c90e | 559 | struct netlink_table *table = &nl_table[sk->sk_protocol]; |
919d9db9 | 560 | int err; |
1da177e4 | 561 | |
c5adde94 | 562 | lock_sock(sk); |
1da177e4 | 563 | |
da314c99 HX |
564 | err = nlk_sk(sk)->portid == portid ? 0 : -EBUSY; |
565 | if (nlk_sk(sk)->bound) | |
1da177e4 LT |
566 | goto err; |
567 | ||
da314c99 | 568 | nlk_sk(sk)->portid = portid; |
e341694e | 569 | sock_hold(sk); |
919d9db9 | 570 | |
c428ecd1 HX |
571 | err = __netlink_insert(table, sk); |
572 | if (err) { | |
4e7c1330 DB |
573 | /* In case the hashtable backend returns with -EBUSY |
574 | * from here, it must not escape to the caller. | |
575 | */ | |
576 | if (unlikely(err == -EBUSY)) | |
577 | err = -EOVERFLOW; | |
c428ecd1 HX |
578 | if (err == -EEXIST) |
579 | err = -EADDRINUSE; | |
c5adde94 | 580 | sock_put(sk); |
1f770c0a | 581 | goto err; |
919d9db9 HX |
582 | } |
583 | ||
da314c99 HX |
584 | /* We need to ensure that the socket is hashed and visible. */ |
585 | smp_wmb(); | |
586 | nlk_sk(sk)->bound = portid; | |
1f770c0a | 587 | |
1da177e4 | 588 | err: |
c5adde94 | 589 | release_sock(sk); |
1da177e4 LT |
590 | return err; |
591 | } | |
592 | ||
593 | static void netlink_remove(struct sock *sk) | |
594 | { | |
e341694e TG |
595 | struct netlink_table *table; |
596 | ||
e341694e | 597 | table = &nl_table[sk->sk_protocol]; |
c428ecd1 HX |
598 | if (!rhashtable_remove_fast(&table->hash, &nlk_sk(sk)->node, |
599 | netlink_rhashtable_params)) { | |
41c6d650 | 600 | WARN_ON(refcount_read(&sk->sk_refcnt) == 1); |
e341694e TG |
601 | __sock_put(sk); |
602 | } | |
e341694e | 603 | |
1da177e4 | 604 | netlink_table_grab(); |
b10dcb3b | 605 | if (nlk_sk(sk)->subscriptions) { |
1da177e4 | 606 | __sk_del_bind_node(sk); |
b10dcb3b JB |
607 | netlink_update_listeners(sk); |
608 | } | |
ee1c2442 JB |
609 | if (sk->sk_protocol == NETLINK_GENERIC) |
610 | atomic_inc(&genl_sk_destructing_cnt); | |
1da177e4 LT |
611 | netlink_table_ungrab(); |
612 | } | |
613 | ||
614 | static struct proto netlink_proto = { | |
615 | .name = "NETLINK", | |
616 | .owner = THIS_MODULE, | |
617 | .obj_size = sizeof(struct netlink_sock), | |
618 | }; | |
619 | ||
1b8d7ae4 | 620 | static int __netlink_create(struct net *net, struct socket *sock, |
11aa9c28 EB |
621 | struct mutex *cb_mutex, int protocol, |
622 | int kern) | |
1da177e4 LT |
623 | { |
624 | struct sock *sk; | |
625 | struct netlink_sock *nlk; | |
ab33a171 PM |
626 | |
627 | sock->ops = &netlink_ops; | |
628 | ||
11aa9c28 | 629 | sk = sk_alloc(net, PF_NETLINK, GFP_KERNEL, &netlink_proto, kern); |
ab33a171 PM |
630 | if (!sk) |
631 | return -ENOMEM; | |
632 | ||
633 | sock_init_data(sock, sk); | |
634 | ||
635 | nlk = nlk_sk(sk); | |
658cb354 | 636 | if (cb_mutex) { |
ffa4d721 | 637 | nlk->cb_mutex = cb_mutex; |
658cb354 | 638 | } else { |
ffa4d721 PM |
639 | nlk->cb_mutex = &nlk->cb_def_mutex; |
640 | mutex_init(nlk->cb_mutex); | |
8a0f5ccf HX |
641 | lockdep_set_class_and_name(nlk->cb_mutex, |
642 | nlk_cb_mutex_keys + protocol, | |
643 | nlk_cb_mutex_key_strings[protocol]); | |
ffa4d721 | 644 | } |
ab33a171 PM |
645 | init_waitqueue_head(&nlk->wait); |
646 | ||
647 | sk->sk_destruct = netlink_sock_destruct; | |
648 | sk->sk_protocol = protocol; | |
649 | return 0; | |
650 | } | |
651 | ||
3f378b68 EP |
652 | static int netlink_create(struct net *net, struct socket *sock, int protocol, |
653 | int kern) | |
ab33a171 PM |
654 | { |
655 | struct module *module = NULL; | |
af65bdfc | 656 | struct mutex *cb_mutex; |
f7fa9b10 | 657 | struct netlink_sock *nlk; |
023e2cfa JB |
658 | int (*bind)(struct net *net, int group); |
659 | void (*unbind)(struct net *net, int group); | |
ab33a171 | 660 | int err = 0; |
1da177e4 LT |
661 | |
662 | sock->state = SS_UNCONNECTED; | |
663 | ||
664 | if (sock->type != SOCK_RAW && sock->type != SOCK_DGRAM) | |
665 | return -ESOCKTNOSUPPORT; | |
666 | ||
6ac552fd | 667 | if (protocol < 0 || protocol >= MAX_LINKS) |
1da177e4 | 668 | return -EPROTONOSUPPORT; |
bc5b6c0b | 669 | protocol = array_index_nospec(protocol, MAX_LINKS); |
1da177e4 | 670 | |
77247bbb | 671 | netlink_lock_table(); |
95a5afca | 672 | #ifdef CONFIG_MODULES |
ab33a171 | 673 | if (!nl_table[protocol].registered) { |
77247bbb | 674 | netlink_unlock_table(); |
4fdb3bb7 | 675 | request_module("net-pf-%d-proto-%d", PF_NETLINK, protocol); |
77247bbb | 676 | netlink_lock_table(); |
4fdb3bb7 | 677 | } |
ab33a171 PM |
678 | #endif |
679 | if (nl_table[protocol].registered && | |
680 | try_module_get(nl_table[protocol].module)) | |
681 | module = nl_table[protocol].module; | |
974c37e9 AD |
682 | else |
683 | err = -EPROTONOSUPPORT; | |
af65bdfc | 684 | cb_mutex = nl_table[protocol].cb_mutex; |
03292745 | 685 | bind = nl_table[protocol].bind; |
4f520900 | 686 | unbind = nl_table[protocol].unbind; |
77247bbb | 687 | netlink_unlock_table(); |
4fdb3bb7 | 688 | |
974c37e9 AD |
689 | if (err < 0) |
690 | goto out; | |
691 | ||
11aa9c28 | 692 | err = __netlink_create(net, sock, cb_mutex, protocol, kern); |
6ac552fd | 693 | if (err < 0) |
f7fa9b10 PM |
694 | goto out_module; |
695 | ||
6f756a8c | 696 | local_bh_disable(); |
c1fd3b94 | 697 | sock_prot_inuse_add(net, &netlink_proto, 1); |
6f756a8c DM |
698 | local_bh_enable(); |
699 | ||
f7fa9b10 | 700 | nlk = nlk_sk(sock->sk); |
f7fa9b10 | 701 | nlk->module = module; |
03292745 | 702 | nlk->netlink_bind = bind; |
4f520900 | 703 | nlk->netlink_unbind = unbind; |
ab33a171 PM |
704 | out: |
705 | return err; | |
1da177e4 | 706 | |
ab33a171 PM |
707 | out_module: |
708 | module_put(module); | |
709 | goto out; | |
1da177e4 LT |
710 | } |
711 | ||
21e4902a TG |
712 | static void deferred_put_nlk_sk(struct rcu_head *head) |
713 | { | |
714 | struct netlink_sock *nlk = container_of(head, struct netlink_sock, rcu); | |
ed5d7788 HX |
715 | struct sock *sk = &nlk->sk; |
716 | ||
be82485f XL |
717 | kfree(nlk->groups); |
718 | nlk->groups = NULL; | |
719 | ||
41c6d650 | 720 | if (!refcount_dec_and_test(&sk->sk_refcnt)) |
ed5d7788 HX |
721 | return; |
722 | ||
723 | if (nlk->cb_running && nlk->cb.done) { | |
724 | INIT_WORK(&nlk->work, netlink_sock_destruct_work); | |
725 | schedule_work(&nlk->work); | |
726 | return; | |
727 | } | |
21e4902a | 728 | |
ed5d7788 | 729 | sk_free(sk); |
21e4902a TG |
730 | } |
731 | ||
1da177e4 LT |
732 | static int netlink_release(struct socket *sock) |
733 | { | |
734 | struct sock *sk = sock->sk; | |
735 | struct netlink_sock *nlk; | |
736 | ||
737 | if (!sk) | |
738 | return 0; | |
739 | ||
740 | netlink_remove(sk); | |
ac57b3a9 | 741 | sock_orphan(sk); |
1da177e4 LT |
742 | nlk = nlk_sk(sk); |
743 | ||
3f660d66 HX |
744 | /* |
745 | * OK. Socket is unlinked, any packets that arrive now | |
746 | * will be purged. | |
747 | */ | |
1da177e4 | 748 | |
ee1c2442 JB |
749 | /* must not acquire netlink_table_lock in any way again before unbind |
750 | * and notifying genetlink is done as otherwise it might deadlock | |
751 | */ | |
752 | if (nlk->netlink_unbind) { | |
753 | int i; | |
754 | ||
755 | for (i = 0; i < nlk->ngroups; i++) | |
756 | if (test_bit(i, nlk->groups)) | |
757 | nlk->netlink_unbind(sock_net(sk), i + 1); | |
758 | } | |
759 | if (sk->sk_protocol == NETLINK_GENERIC && | |
760 | atomic_dec_return(&genl_sk_destructing_cnt) == 0) | |
761 | wake_up(&genl_sk_destructing_waitq); | |
762 | ||
1da177e4 LT |
763 | sock->sk = NULL; |
764 | wake_up_interruptible_all(&nlk->wait); | |
765 | ||
766 | skb_queue_purge(&sk->sk_write_queue); | |
767 | ||
e2726020 | 768 | if (nlk->portid && nlk->bound) { |
1da177e4 | 769 | struct netlink_notify n = { |
3b1e0a65 | 770 | .net = sock_net(sk), |
1da177e4 | 771 | .protocol = sk->sk_protocol, |
15e47304 | 772 | .portid = nlk->portid, |
1da177e4 | 773 | }; |
efa172f4 | 774 | blocking_notifier_call_chain(&netlink_chain, |
e041c683 | 775 | NETLINK_URELEASE, &n); |
746fac4d | 776 | } |
4fdb3bb7 | 777 | |
5e7c001c | 778 | module_put(nlk->module); |
4fdb3bb7 | 779 | |
aed81560 | 780 | if (netlink_is_kernel(sk)) { |
b10dcb3b | 781 | netlink_table_grab(); |
869e58f8 DL |
782 | BUG_ON(nl_table[sk->sk_protocol].registered == 0); |
783 | if (--nl_table[sk->sk_protocol].registered == 0) { | |
6d772ac5 ED |
784 | struct listeners *old; |
785 | ||
786 | old = nl_deref_protected(nl_table[sk->sk_protocol].listeners); | |
787 | RCU_INIT_POINTER(nl_table[sk->sk_protocol].listeners, NULL); | |
788 | kfree_rcu(old, rcu); | |
869e58f8 | 789 | nl_table[sk->sk_protocol].module = NULL; |
9785e10a | 790 | nl_table[sk->sk_protocol].bind = NULL; |
4f520900 | 791 | nl_table[sk->sk_protocol].unbind = NULL; |
9785e10a | 792 | nl_table[sk->sk_protocol].flags = 0; |
869e58f8 DL |
793 | nl_table[sk->sk_protocol].registered = 0; |
794 | } | |
b10dcb3b | 795 | netlink_table_ungrab(); |
658cb354 | 796 | } |
77247bbb | 797 | |
3755810c | 798 | local_bh_disable(); |
c1fd3b94 | 799 | sock_prot_inuse_add(sock_net(sk), &netlink_proto, -1); |
3755810c | 800 | local_bh_enable(); |
21e4902a | 801 | call_rcu(&nlk->rcu, deferred_put_nlk_sk); |
1da177e4 LT |
802 | return 0; |
803 | } | |
804 | ||
805 | static int netlink_autobind(struct socket *sock) | |
806 | { | |
807 | struct sock *sk = sock->sk; | |
3b1e0a65 | 808 | struct net *net = sock_net(sk); |
da12c90e | 809 | struct netlink_table *table = &nl_table[sk->sk_protocol]; |
15e47304 | 810 | s32 portid = task_tgid_vnr(current); |
1da177e4 | 811 | int err; |
b9fbe709 HX |
812 | s32 rover = -4096; |
813 | bool ok; | |
1da177e4 LT |
814 | |
815 | retry: | |
816 | cond_resched(); | |
e341694e | 817 | rcu_read_lock(); |
b9fbe709 HX |
818 | ok = !__netlink_lookup(table, portid, net); |
819 | rcu_read_unlock(); | |
820 | if (!ok) { | |
e341694e | 821 | /* Bind collision, search negative portid values. */ |
b9fbe709 HX |
822 | if (rover == -4096) |
823 | /* rover will be in range [S32_MIN, -4097] */ | |
824 | rover = S32_MIN + prandom_u32_max(-4096 - S32_MIN); | |
825 | else if (rover >= -4096) | |
e341694e | 826 | rover = -4097; |
b9fbe709 | 827 | portid = rover--; |
e341694e | 828 | goto retry; |
1da177e4 | 829 | } |
1da177e4 | 830 | |
8ea65f4a | 831 | err = netlink_insert(sk, portid); |
1da177e4 LT |
832 | if (err == -EADDRINUSE) |
833 | goto retry; | |
d470e3b4 DM |
834 | |
835 | /* If 2 threads race to autobind, that is fine. */ | |
836 | if (err == -EBUSY) | |
837 | err = 0; | |
838 | ||
839 | return err; | |
1da177e4 LT |
840 | } |
841 | ||
aa4cf945 EB |
842 | /** |
843 | * __netlink_ns_capable - General netlink message capability test | |
844 | * @nsp: NETLINK_CB of the socket buffer holding a netlink command from userspace. | |
845 | * @user_ns: The user namespace of the capability to use | |
846 | * @cap: The capability to use | |
847 | * | |
848 | * Test to see if the opener of the socket we received the message | |
849 | * from had when the netlink socket was created and the sender of the | |
850 | * message has has the capability @cap in the user namespace @user_ns. | |
851 | */ | |
852 | bool __netlink_ns_capable(const struct netlink_skb_parms *nsp, | |
853 | struct user_namespace *user_ns, int cap) | |
854 | { | |
2d7a85f4 EB |
855 | return ((nsp->flags & NETLINK_SKB_DST) || |
856 | file_ns_capable(nsp->sk->sk_socket->file, user_ns, cap)) && | |
857 | ns_capable(user_ns, cap); | |
aa4cf945 EB |
858 | } |
859 | EXPORT_SYMBOL(__netlink_ns_capable); | |
860 | ||
861 | /** | |
862 | * netlink_ns_capable - General netlink message capability test | |
863 | * @skb: socket buffer holding a netlink command from userspace | |
864 | * @user_ns: The user namespace of the capability to use | |
865 | * @cap: The capability to use | |
866 | * | |
867 | * Test to see if the opener of the socket we received the message | |
868 | * from had when the netlink socket was created and the sender of the | |
869 | * message has has the capability @cap in the user namespace @user_ns. | |
870 | */ | |
871 | bool netlink_ns_capable(const struct sk_buff *skb, | |
872 | struct user_namespace *user_ns, int cap) | |
873 | { | |
874 | return __netlink_ns_capable(&NETLINK_CB(skb), user_ns, cap); | |
875 | } | |
876 | EXPORT_SYMBOL(netlink_ns_capable); | |
877 | ||
878 | /** | |
879 | * netlink_capable - Netlink global message capability test | |
880 | * @skb: socket buffer holding a netlink command from userspace | |
881 | * @cap: The capability to use | |
882 | * | |
883 | * Test to see if the opener of the socket we received the message | |
884 | * from had when the netlink socket was created and the sender of the | |
885 | * message has has the capability @cap in all user namespaces. | |
886 | */ | |
887 | bool netlink_capable(const struct sk_buff *skb, int cap) | |
888 | { | |
889 | return netlink_ns_capable(skb, &init_user_ns, cap); | |
890 | } | |
891 | EXPORT_SYMBOL(netlink_capable); | |
892 | ||
893 | /** | |
894 | * netlink_net_capable - Netlink network namespace message capability test | |
895 | * @skb: socket buffer holding a netlink command from userspace | |
896 | * @cap: The capability to use | |
897 | * | |
898 | * Test to see if the opener of the socket we received the message | |
899 | * from had when the netlink socket was created and the sender of the | |
900 | * message has has the capability @cap over the network namespace of | |
901 | * the socket we received the message from. | |
902 | */ | |
903 | bool netlink_net_capable(const struct sk_buff *skb, int cap) | |
904 | { | |
905 | return netlink_ns_capable(skb, sock_net(skb->sk)->user_ns, cap); | |
906 | } | |
907 | EXPORT_SYMBOL(netlink_net_capable); | |
908 | ||
5187cd05 | 909 | static inline int netlink_allowed(const struct socket *sock, unsigned int flag) |
746fac4d | 910 | { |
9785e10a | 911 | return (nl_table[sock->sk->sk_protocol].flags & flag) || |
df008c91 | 912 | ns_capable(sock_net(sock->sk)->user_ns, CAP_NET_ADMIN); |
746fac4d | 913 | } |
1da177e4 | 914 | |
f7fa9b10 PM |
915 | static void |
916 | netlink_update_subscriptions(struct sock *sk, unsigned int subscriptions) | |
917 | { | |
918 | struct netlink_sock *nlk = nlk_sk(sk); | |
919 | ||
920 | if (nlk->subscriptions && !subscriptions) | |
921 | __sk_del_bind_node(sk); | |
922 | else if (!nlk->subscriptions && subscriptions) | |
923 | sk_add_bind_node(sk, &nl_table[sk->sk_protocol].mc_list); | |
924 | nlk->subscriptions = subscriptions; | |
925 | } | |
926 | ||
b4ff4f04 | 927 | static int netlink_realloc_groups(struct sock *sk) |
513c2500 PM |
928 | { |
929 | struct netlink_sock *nlk = nlk_sk(sk); | |
930 | unsigned int groups; | |
b4ff4f04 | 931 | unsigned long *new_groups; |
513c2500 PM |
932 | int err = 0; |
933 | ||
b4ff4f04 JB |
934 | netlink_table_grab(); |
935 | ||
513c2500 | 936 | groups = nl_table[sk->sk_protocol].groups; |
b4ff4f04 | 937 | if (!nl_table[sk->sk_protocol].registered) { |
513c2500 | 938 | err = -ENOENT; |
b4ff4f04 JB |
939 | goto out_unlock; |
940 | } | |
513c2500 | 941 | |
b4ff4f04 JB |
942 | if (nlk->ngroups >= groups) |
943 | goto out_unlock; | |
513c2500 | 944 | |
b4ff4f04 JB |
945 | new_groups = krealloc(nlk->groups, NLGRPSZ(groups), GFP_ATOMIC); |
946 | if (new_groups == NULL) { | |
947 | err = -ENOMEM; | |
948 | goto out_unlock; | |
949 | } | |
6ac552fd | 950 | memset((char *)new_groups + NLGRPSZ(nlk->ngroups), 0, |
b4ff4f04 JB |
951 | NLGRPSZ(groups) - NLGRPSZ(nlk->ngroups)); |
952 | ||
953 | nlk->groups = new_groups; | |
513c2500 | 954 | nlk->ngroups = groups; |
b4ff4f04 JB |
955 | out_unlock: |
956 | netlink_table_ungrab(); | |
957 | return err; | |
513c2500 PM |
958 | } |
959 | ||
02c81ab9 | 960 | static void netlink_undo_bind(int group, long unsigned int groups, |
023e2cfa | 961 | struct sock *sk) |
4f520900 | 962 | { |
023e2cfa | 963 | struct netlink_sock *nlk = nlk_sk(sk); |
4f520900 RGB |
964 | int undo; |
965 | ||
966 | if (!nlk->netlink_unbind) | |
967 | return; | |
968 | ||
969 | for (undo = 0; undo < group; undo++) | |
6251edd9 | 970 | if (test_bit(undo, &groups)) |
8b7c36d8 | 971 | nlk->netlink_unbind(sock_net(sk), undo + 1); |
4f520900 RGB |
972 | } |
973 | ||
6ac552fd PM |
974 | static int netlink_bind(struct socket *sock, struct sockaddr *addr, |
975 | int addr_len) | |
1da177e4 LT |
976 | { |
977 | struct sock *sk = sock->sk; | |
3b1e0a65 | 978 | struct net *net = sock_net(sk); |
1da177e4 LT |
979 | struct netlink_sock *nlk = nlk_sk(sk); |
980 | struct sockaddr_nl *nladdr = (struct sockaddr_nl *)addr; | |
f7736080 | 981 | int err = 0; |
d852be84 | 982 | unsigned long groups; |
da314c99 | 983 | bool bound; |
746fac4d | 984 | |
4e4b5376 HFS |
985 | if (addr_len < sizeof(struct sockaddr_nl)) |
986 | return -EINVAL; | |
987 | ||
1da177e4 LT |
988 | if (nladdr->nl_family != AF_NETLINK) |
989 | return -EINVAL; | |
d852be84 | 990 | groups = nladdr->nl_groups; |
1da177e4 LT |
991 | |
992 | /* Only superuser is allowed to listen multicasts */ | |
4f520900 | 993 | if (groups) { |
5187cd05 | 994 | if (!netlink_allowed(sock, NL_CFG_F_NONROOT_RECV)) |
513c2500 | 995 | return -EPERM; |
b4ff4f04 JB |
996 | err = netlink_realloc_groups(sk); |
997 | if (err) | |
998 | return err; | |
513c2500 | 999 | } |
1da177e4 | 1000 | |
428f944b | 1001 | if (nlk->ngroups < BITS_PER_LONG) |
91874ecf | 1002 | groups &= (1UL << nlk->ngroups) - 1; |
1da177e4 | 1003 | |
da314c99 HX |
1004 | bound = nlk->bound; |
1005 | if (bound) { | |
1006 | /* Ensure nlk->portid is up-to-date. */ | |
1007 | smp_rmb(); | |
1008 | ||
15e47304 | 1009 | if (nladdr->nl_pid != nlk->portid) |
1da177e4 | 1010 | return -EINVAL; |
da314c99 | 1011 | } |
4f520900 | 1012 | |
f7736080 | 1013 | netlink_lock_table(); |
4f520900 RGB |
1014 | if (nlk->netlink_bind && groups) { |
1015 | int group; | |
1016 | ||
1017 | for (group = 0; group < nlk->ngroups; group++) { | |
1018 | if (!test_bit(group, &groups)) | |
1019 | continue; | |
8b7c36d8 | 1020 | err = nlk->netlink_bind(net, group + 1); |
4f520900 RGB |
1021 | if (!err) |
1022 | continue; | |
023e2cfa | 1023 | netlink_undo_bind(group, groups, sk); |
f7736080 | 1024 | goto unlock; |
4f520900 RGB |
1025 | } |
1026 | } | |
1027 | ||
da314c99 HX |
1028 | /* No need for barriers here as we return to user-space without |
1029 | * using any of the bound attributes. | |
1030 | */ | |
1031 | if (!bound) { | |
1da177e4 | 1032 | err = nladdr->nl_pid ? |
8ea65f4a | 1033 | netlink_insert(sk, nladdr->nl_pid) : |
1da177e4 | 1034 | netlink_autobind(sock); |
4f520900 | 1035 | if (err) { |
023e2cfa | 1036 | netlink_undo_bind(nlk->ngroups, groups, sk); |
f7736080 | 1037 | goto unlock; |
4f520900 | 1038 | } |
1da177e4 LT |
1039 | } |
1040 | ||
4f520900 | 1041 | if (!groups && (nlk->groups == NULL || !(u32)nlk->groups[0])) |
f7736080 XL |
1042 | goto unlock; |
1043 | netlink_unlock_table(); | |
1da177e4 LT |
1044 | |
1045 | netlink_table_grab(); | |
f7fa9b10 | 1046 | netlink_update_subscriptions(sk, nlk->subscriptions + |
4f520900 | 1047 | hweight32(groups) - |
746fac4d | 1048 | hweight32(nlk->groups[0])); |
4f520900 | 1049 | nlk->groups[0] = (nlk->groups[0] & ~0xffffffffUL) | groups; |
4277a083 | 1050 | netlink_update_listeners(sk); |
1da177e4 LT |
1051 | netlink_table_ungrab(); |
1052 | ||
1053 | return 0; | |
f7736080 XL |
1054 | |
1055 | unlock: | |
1056 | netlink_unlock_table(); | |
1057 | return err; | |
1da177e4 LT |
1058 | } |
1059 | ||
1060 | static int netlink_connect(struct socket *sock, struct sockaddr *addr, | |
1061 | int alen, int flags) | |
1062 | { | |
1063 | int err = 0; | |
1064 | struct sock *sk = sock->sk; | |
1065 | struct netlink_sock *nlk = nlk_sk(sk); | |
6ac552fd | 1066 | struct sockaddr_nl *nladdr = (struct sockaddr_nl *)addr; |
1da177e4 | 1067 | |
6503d961 CG |
1068 | if (alen < sizeof(addr->sa_family)) |
1069 | return -EINVAL; | |
1070 | ||
1da177e4 LT |
1071 | if (addr->sa_family == AF_UNSPEC) { |
1072 | sk->sk_state = NETLINK_UNCONNECTED; | |
15e47304 | 1073 | nlk->dst_portid = 0; |
d629b836 | 1074 | nlk->dst_group = 0; |
1da177e4 LT |
1075 | return 0; |
1076 | } | |
1077 | if (addr->sa_family != AF_NETLINK) | |
1078 | return -EINVAL; | |
1079 | ||
78802879 AP |
1080 | if (alen < sizeof(struct sockaddr_nl)) |
1081 | return -EINVAL; | |
1082 | ||
46833a86 | 1083 | if ((nladdr->nl_groups || nladdr->nl_pid) && |
5187cd05 | 1084 | !netlink_allowed(sock, NL_CFG_F_NONROOT_SEND)) |
1da177e4 LT |
1085 | return -EPERM; |
1086 | ||
da314c99 HX |
1087 | /* No need for barriers here as we return to user-space without |
1088 | * using any of the bound attributes. | |
1089 | */ | |
1090 | if (!nlk->bound) | |
1da177e4 LT |
1091 | err = netlink_autobind(sock); |
1092 | ||
1093 | if (err == 0) { | |
1094 | sk->sk_state = NETLINK_CONNECTED; | |
15e47304 | 1095 | nlk->dst_portid = nladdr->nl_pid; |
d629b836 | 1096 | nlk->dst_group = ffs(nladdr->nl_groups); |
1da177e4 LT |
1097 | } |
1098 | ||
1099 | return err; | |
1100 | } | |
1101 | ||
6ac552fd | 1102 | static int netlink_getname(struct socket *sock, struct sockaddr *addr, |
9b2c45d4 | 1103 | int peer) |
1da177e4 LT |
1104 | { |
1105 | struct sock *sk = sock->sk; | |
1106 | struct netlink_sock *nlk = nlk_sk(sk); | |
13cfa97b | 1107 | DECLARE_SOCKADDR(struct sockaddr_nl *, nladdr, addr); |
746fac4d | 1108 | |
1da177e4 LT |
1109 | nladdr->nl_family = AF_NETLINK; |
1110 | nladdr->nl_pad = 0; | |
1da177e4 LT |
1111 | |
1112 | if (peer) { | |
15e47304 | 1113 | nladdr->nl_pid = nlk->dst_portid; |
d629b836 | 1114 | nladdr->nl_groups = netlink_group_mask(nlk->dst_group); |
1da177e4 | 1115 | } else { |
15e47304 | 1116 | nladdr->nl_pid = nlk->portid; |
f7736080 | 1117 | netlink_lock_table(); |
513c2500 | 1118 | nladdr->nl_groups = nlk->groups ? nlk->groups[0] : 0; |
f7736080 | 1119 | netlink_unlock_table(); |
1da177e4 | 1120 | } |
9b2c45d4 | 1121 | return sizeof(*nladdr); |
1da177e4 LT |
1122 | } |
1123 | ||
025c6818 DD |
1124 | static int netlink_ioctl(struct socket *sock, unsigned int cmd, |
1125 | unsigned long arg) | |
1126 | { | |
1127 | /* try to hand this ioctl down to the NIC drivers. | |
1128 | */ | |
1129 | return -ENOIOCTLCMD; | |
1130 | } | |
1131 | ||
15e47304 | 1132 | static struct sock *netlink_getsockbyportid(struct sock *ssk, u32 portid) |
1da177e4 | 1133 | { |
1da177e4 LT |
1134 | struct sock *sock; |
1135 | struct netlink_sock *nlk; | |
1136 | ||
15e47304 | 1137 | sock = netlink_lookup(sock_net(ssk), ssk->sk_protocol, portid); |
1da177e4 LT |
1138 | if (!sock) |
1139 | return ERR_PTR(-ECONNREFUSED); | |
1140 | ||
1141 | /* Don't bother queuing skb if kernel socket has no input function */ | |
1142 | nlk = nlk_sk(sock); | |
cd40b7d3 | 1143 | if (sock->sk_state == NETLINK_CONNECTED && |
15e47304 | 1144 | nlk->dst_portid != nlk_sk(ssk)->portid) { |
1da177e4 LT |
1145 | sock_put(sock); |
1146 | return ERR_PTR(-ECONNREFUSED); | |
1147 | } | |
1148 | return sock; | |
1149 | } | |
1150 | ||
1151 | struct sock *netlink_getsockbyfilp(struct file *filp) | |
1152 | { | |
496ad9aa | 1153 | struct inode *inode = file_inode(filp); |
1da177e4 LT |
1154 | struct sock *sock; |
1155 | ||
1156 | if (!S_ISSOCK(inode->i_mode)) | |
1157 | return ERR_PTR(-ENOTSOCK); | |
1158 | ||
1159 | sock = SOCKET_I(inode)->sk; | |
1160 | if (sock->sk_family != AF_NETLINK) | |
1161 | return ERR_PTR(-EINVAL); | |
1162 | ||
1163 | sock_hold(sock); | |
1164 | return sock; | |
1165 | } | |
1166 | ||
3a36515f PN |
1167 | static struct sk_buff *netlink_alloc_large_skb(unsigned int size, |
1168 | int broadcast) | |
c05cdb1b PNA |
1169 | { |
1170 | struct sk_buff *skb; | |
1171 | void *data; | |
1172 | ||
3a36515f | 1173 | if (size <= NLMSG_GOODSIZE || broadcast) |
c05cdb1b PNA |
1174 | return alloc_skb(size, GFP_KERNEL); |
1175 | ||
3a36515f PN |
1176 | size = SKB_DATA_ALIGN(size) + |
1177 | SKB_DATA_ALIGN(sizeof(struct skb_shared_info)); | |
c05cdb1b PNA |
1178 | |
1179 | data = vmalloc(size); | |
1180 | if (data == NULL) | |
3a36515f | 1181 | return NULL; |
c05cdb1b | 1182 | |
2ea2f62c | 1183 | skb = __build_skb(data, size); |
3a36515f PN |
1184 | if (skb == NULL) |
1185 | vfree(data); | |
2ea2f62c | 1186 | else |
3a36515f | 1187 | skb->destructor = netlink_skb_destructor; |
c05cdb1b PNA |
1188 | |
1189 | return skb; | |
c05cdb1b PNA |
1190 | } |
1191 | ||
1da177e4 LT |
1192 | /* |
1193 | * Attach a skb to a netlink socket. | |
1194 | * The caller must hold a reference to the destination socket. On error, the | |
1195 | * reference is dropped. The skb is not send to the destination, just all | |
1196 | * all error checks are performed and memory in the queue is reserved. | |
1197 | * Return values: | |
1198 | * < 0: error. skb freed, reference to sock dropped. | |
1199 | * 0: continue | |
1200 | * 1: repeat lookup - reference dropped while waiting for socket memory. | |
1201 | */ | |
9457afee | 1202 | int netlink_attachskb(struct sock *sk, struct sk_buff *skb, |
c3d8d1e3 | 1203 | long *timeo, struct sock *ssk) |
1da177e4 LT |
1204 | { |
1205 | struct netlink_sock *nlk; | |
1206 | ||
1207 | nlk = nlk_sk(sk); | |
1208 | ||
5fd96123 | 1209 | if ((atomic_read(&sk->sk_rmem_alloc) > sk->sk_rcvbuf || |
d1b4c689 | 1210 | test_bit(NETLINK_S_CONGESTED, &nlk->state))) { |
1da177e4 | 1211 | DECLARE_WAITQUEUE(wait, current); |
c3d8d1e3 | 1212 | if (!*timeo) { |
aed81560 | 1213 | if (!ssk || netlink_is_kernel(ssk)) |
1da177e4 LT |
1214 | netlink_overrun(sk); |
1215 | sock_put(sk); | |
1216 | kfree_skb(skb); | |
1217 | return -EAGAIN; | |
1218 | } | |
1219 | ||
1220 | __set_current_state(TASK_INTERRUPTIBLE); | |
1221 | add_wait_queue(&nlk->wait, &wait); | |
1222 | ||
1223 | if ((atomic_read(&sk->sk_rmem_alloc) > sk->sk_rcvbuf || | |
cc3a572f | 1224 | test_bit(NETLINK_S_CONGESTED, &nlk->state)) && |
1da177e4 | 1225 | !sock_flag(sk, SOCK_DEAD)) |
c3d8d1e3 | 1226 | *timeo = schedule_timeout(*timeo); |
1da177e4 LT |
1227 | |
1228 | __set_current_state(TASK_RUNNING); | |
1229 | remove_wait_queue(&nlk->wait, &wait); | |
1230 | sock_put(sk); | |
1231 | ||
1232 | if (signal_pending(current)) { | |
1233 | kfree_skb(skb); | |
c3d8d1e3 | 1234 | return sock_intr_errno(*timeo); |
1da177e4 LT |
1235 | } |
1236 | return 1; | |
1237 | } | |
cf0a018a | 1238 | netlink_skb_set_owner_r(skb, sk); |
1da177e4 LT |
1239 | return 0; |
1240 | } | |
1241 | ||
4a7e7c2a | 1242 | static int __netlink_sendskb(struct sock *sk, struct sk_buff *skb) |
1da177e4 | 1243 | { |
1da177e4 LT |
1244 | int len = skb->len; |
1245 | ||
25e3f70f | 1246 | netlink_deliver_tap(sock_net(sk), skb); |
bcbde0d4 | 1247 | |
d1b4c689 | 1248 | skb_queue_tail(&sk->sk_receive_queue, skb); |
676d2369 | 1249 | sk->sk_data_ready(sk); |
4a7e7c2a ED |
1250 | return len; |
1251 | } | |
1252 | ||
1253 | int netlink_sendskb(struct sock *sk, struct sk_buff *skb) | |
1254 | { | |
1255 | int len = __netlink_sendskb(sk, skb); | |
1256 | ||
1da177e4 LT |
1257 | sock_put(sk); |
1258 | return len; | |
1259 | } | |
1260 | ||
1261 | void netlink_detachskb(struct sock *sk, struct sk_buff *skb) | |
1262 | { | |
1263 | kfree_skb(skb); | |
1264 | sock_put(sk); | |
1265 | } | |
1266 | ||
b57ef81f | 1267 | static struct sk_buff *netlink_trim(struct sk_buff *skb, gfp_t allocation) |
1da177e4 LT |
1268 | { |
1269 | int delta; | |
1270 | ||
1298ca46 | 1271 | WARN_ON(skb->sk != NULL); |
4305b541 | 1272 | delta = skb->end - skb->tail; |
c05cdb1b | 1273 | if (is_vmalloc_addr(skb->head) || delta * 2 < skb->truesize) |
1da177e4 LT |
1274 | return skb; |
1275 | ||
1276 | if (skb_shared(skb)) { | |
1277 | struct sk_buff *nskb = skb_clone(skb, allocation); | |
1278 | if (!nskb) | |
1279 | return skb; | |
8460c00f | 1280 | consume_skb(skb); |
1da177e4 LT |
1281 | skb = nskb; |
1282 | } | |
1283 | ||
158f323b ED |
1284 | pskb_expand_head(skb, 0, -delta, |
1285 | (allocation & ~__GFP_DIRECT_RECLAIM) | | |
1286 | __GFP_NOWARN | __GFP_NORETRY); | |
1da177e4 LT |
1287 | return skb; |
1288 | } | |
1289 | ||
3fbc2905 EB |
1290 | static int netlink_unicast_kernel(struct sock *sk, struct sk_buff *skb, |
1291 | struct sock *ssk) | |
cd40b7d3 DL |
1292 | { |
1293 | int ret; | |
1294 | struct netlink_sock *nlk = nlk_sk(sk); | |
1295 | ||
1296 | ret = -ECONNREFUSED; | |
1297 | if (nlk->netlink_rcv != NULL) { | |
1298 | ret = skb->len; | |
cf0a018a | 1299 | netlink_skb_set_owner_r(skb, sk); |
e32123e5 | 1300 | NETLINK_CB(skb).sk = ssk; |
73bfd370 | 1301 | netlink_deliver_tap_kernel(sk, ssk, skb); |
cd40b7d3 | 1302 | nlk->netlink_rcv(skb); |
bfb253c9 ED |
1303 | consume_skb(skb); |
1304 | } else { | |
1305 | kfree_skb(skb); | |
cd40b7d3 | 1306 | } |
cd40b7d3 DL |
1307 | sock_put(sk); |
1308 | return ret; | |
1309 | } | |
1310 | ||
1311 | int netlink_unicast(struct sock *ssk, struct sk_buff *skb, | |
15e47304 | 1312 | u32 portid, int nonblock) |
1da177e4 LT |
1313 | { |
1314 | struct sock *sk; | |
1315 | int err; | |
1316 | long timeo; | |
1317 | ||
1318 | skb = netlink_trim(skb, gfp_any()); | |
1319 | ||
1320 | timeo = sock_sndtimeo(ssk, nonblock); | |
1321 | retry: | |
15e47304 | 1322 | sk = netlink_getsockbyportid(ssk, portid); |
1da177e4 LT |
1323 | if (IS_ERR(sk)) { |
1324 | kfree_skb(skb); | |
1325 | return PTR_ERR(sk); | |
1326 | } | |
cd40b7d3 | 1327 | if (netlink_is_kernel(sk)) |
3fbc2905 | 1328 | return netlink_unicast_kernel(sk, skb, ssk); |
cd40b7d3 | 1329 | |
b1153f29 | 1330 | if (sk_filter(sk, skb)) { |
84874607 | 1331 | err = skb->len; |
b1153f29 SH |
1332 | kfree_skb(skb); |
1333 | sock_put(sk); | |
1334 | return err; | |
1335 | } | |
1336 | ||
9457afee | 1337 | err = netlink_attachskb(sk, skb, &timeo, ssk); |
1da177e4 LT |
1338 | if (err == 1) |
1339 | goto retry; | |
1340 | if (err) | |
1341 | return err; | |
1342 | ||
7ee015e0 | 1343 | return netlink_sendskb(sk, skb); |
1da177e4 | 1344 | } |
6ac552fd | 1345 | EXPORT_SYMBOL(netlink_unicast); |
1da177e4 | 1346 | |
4277a083 PM |
1347 | int netlink_has_listeners(struct sock *sk, unsigned int group) |
1348 | { | |
1349 | int res = 0; | |
5c398dc8 | 1350 | struct listeners *listeners; |
4277a083 | 1351 | |
aed81560 | 1352 | BUG_ON(!netlink_is_kernel(sk)); |
b4ff4f04 JB |
1353 | |
1354 | rcu_read_lock(); | |
1355 | listeners = rcu_dereference(nl_table[sk->sk_protocol].listeners); | |
1356 | ||
6d772ac5 | 1357 | if (listeners && group - 1 < nl_table[sk->sk_protocol].groups) |
5c398dc8 | 1358 | res = test_bit(group - 1, listeners->masks); |
b4ff4f04 JB |
1359 | |
1360 | rcu_read_unlock(); | |
1361 | ||
4277a083 PM |
1362 | return res; |
1363 | } | |
1364 | EXPORT_SYMBOL_GPL(netlink_has_listeners); | |
1365 | ||
59c28058 JK |
1366 | bool netlink_strict_get_check(struct sk_buff *skb) |
1367 | { | |
1368 | const struct netlink_sock *nlk = nlk_sk(NETLINK_CB(skb).sk); | |
1369 | ||
1370 | return nlk->flags & NETLINK_F_STRICT_CHK; | |
1371 | } | |
1372 | EXPORT_SYMBOL_GPL(netlink_strict_get_check); | |
1373 | ||
b57ef81f | 1374 | static int netlink_broadcast_deliver(struct sock *sk, struct sk_buff *skb) |
1da177e4 LT |
1375 | { |
1376 | struct netlink_sock *nlk = nlk_sk(sk); | |
1377 | ||
1378 | if (atomic_read(&sk->sk_rmem_alloc) <= sk->sk_rcvbuf && | |
cc3a572f | 1379 | !test_bit(NETLINK_S_CONGESTED, &nlk->state)) { |
cf0a018a | 1380 | netlink_skb_set_owner_r(skb, sk); |
4a7e7c2a | 1381 | __netlink_sendskb(sk, skb); |
2c645800 | 1382 | return atomic_read(&sk->sk_rmem_alloc) > (sk->sk_rcvbuf >> 1); |
1da177e4 LT |
1383 | } |
1384 | return -1; | |
1385 | } | |
1386 | ||
1387 | struct netlink_broadcast_data { | |
1388 | struct sock *exclude_sk; | |
b4b51029 | 1389 | struct net *net; |
15e47304 | 1390 | u32 portid; |
1da177e4 LT |
1391 | u32 group; |
1392 | int failure; | |
ff491a73 | 1393 | int delivery_failure; |
1da177e4 LT |
1394 | int congested; |
1395 | int delivered; | |
7d877f3b | 1396 | gfp_t allocation; |
1da177e4 | 1397 | struct sk_buff *skb, *skb2; |
910a7e90 EB |
1398 | int (*tx_filter)(struct sock *dsk, struct sk_buff *skb, void *data); |
1399 | void *tx_data; | |
1da177e4 LT |
1400 | }; |
1401 | ||
46c9521f RR |
1402 | static void do_one_broadcast(struct sock *sk, |
1403 | struct netlink_broadcast_data *p) | |
1da177e4 LT |
1404 | { |
1405 | struct netlink_sock *nlk = nlk_sk(sk); | |
1406 | int val; | |
1407 | ||
1408 | if (p->exclude_sk == sk) | |
46c9521f | 1409 | return; |
1da177e4 | 1410 | |
15e47304 | 1411 | if (nlk->portid == p->portid || p->group - 1 >= nlk->ngroups || |
f7fa9b10 | 1412 | !test_bit(p->group - 1, nlk->groups)) |
46c9521f | 1413 | return; |
1da177e4 | 1414 | |
59324cf3 ND |
1415 | if (!net_eq(sock_net(sk), p->net)) { |
1416 | if (!(nlk->flags & NETLINK_F_LISTEN_ALL_NSID)) | |
1417 | return; | |
1418 | ||
1419 | if (!peernet_has_id(sock_net(sk), p->net)) | |
1420 | return; | |
1421 | ||
1422 | if (!file_ns_capable(sk->sk_socket->file, p->net->user_ns, | |
1423 | CAP_NET_BROADCAST)) | |
1424 | return; | |
1425 | } | |
b4b51029 | 1426 | |
1da177e4 LT |
1427 | if (p->failure) { |
1428 | netlink_overrun(sk); | |
46c9521f | 1429 | return; |
1da177e4 LT |
1430 | } |
1431 | ||
1432 | sock_hold(sk); | |
1433 | if (p->skb2 == NULL) { | |
68acc024 | 1434 | if (skb_shared(p->skb)) { |
1da177e4 LT |
1435 | p->skb2 = skb_clone(p->skb, p->allocation); |
1436 | } else { | |
68acc024 TC |
1437 | p->skb2 = skb_get(p->skb); |
1438 | /* | |
1439 | * skb ownership may have been set when | |
1440 | * delivered to a previous socket. | |
1441 | */ | |
1442 | skb_orphan(p->skb2); | |
1da177e4 LT |
1443 | } |
1444 | } | |
1445 | if (p->skb2 == NULL) { | |
1446 | netlink_overrun(sk); | |
1447 | /* Clone failed. Notify ALL listeners. */ | |
1448 | p->failure = 1; | |
cc3a572f | 1449 | if (nlk->flags & NETLINK_F_BROADCAST_SEND_ERROR) |
be0c22a4 | 1450 | p->delivery_failure = 1; |
59324cf3 ND |
1451 | goto out; |
1452 | } | |
1453 | if (p->tx_filter && p->tx_filter(sk, p->skb2, p->tx_data)) { | |
910a7e90 EB |
1454 | kfree_skb(p->skb2); |
1455 | p->skb2 = NULL; | |
59324cf3 ND |
1456 | goto out; |
1457 | } | |
1458 | if (sk_filter(sk, p->skb2)) { | |
b1153f29 SH |
1459 | kfree_skb(p->skb2); |
1460 | p->skb2 = NULL; | |
59324cf3 ND |
1461 | goto out; |
1462 | } | |
1463 | NETLINK_CB(p->skb2).nsid = peernet2id(sock_net(sk), p->net); | |
7212462f ND |
1464 | if (NETLINK_CB(p->skb2).nsid != NETNSA_NSID_NOT_ASSIGNED) |
1465 | NETLINK_CB(p->skb2).nsid_is_set = true; | |
59324cf3 ND |
1466 | val = netlink_broadcast_deliver(sk, p->skb2); |
1467 | if (val < 0) { | |
1da177e4 | 1468 | netlink_overrun(sk); |
cc3a572f | 1469 | if (nlk->flags & NETLINK_F_BROADCAST_SEND_ERROR) |
be0c22a4 | 1470 | p->delivery_failure = 1; |
1da177e4 LT |
1471 | } else { |
1472 | p->congested |= val; | |
1473 | p->delivered = 1; | |
1474 | p->skb2 = NULL; | |
1475 | } | |
59324cf3 | 1476 | out: |
1da177e4 | 1477 | sock_put(sk); |
1da177e4 LT |
1478 | } |
1479 | ||
15e47304 | 1480 | int netlink_broadcast_filtered(struct sock *ssk, struct sk_buff *skb, u32 portid, |
910a7e90 EB |
1481 | u32 group, gfp_t allocation, |
1482 | int (*filter)(struct sock *dsk, struct sk_buff *skb, void *data), | |
1483 | void *filter_data) | |
1da177e4 | 1484 | { |
3b1e0a65 | 1485 | struct net *net = sock_net(ssk); |
1da177e4 | 1486 | struct netlink_broadcast_data info; |
1da177e4 LT |
1487 | struct sock *sk; |
1488 | ||
1489 | skb = netlink_trim(skb, allocation); | |
1490 | ||
1491 | info.exclude_sk = ssk; | |
b4b51029 | 1492 | info.net = net; |
15e47304 | 1493 | info.portid = portid; |
1da177e4 LT |
1494 | info.group = group; |
1495 | info.failure = 0; | |
ff491a73 | 1496 | info.delivery_failure = 0; |
1da177e4 LT |
1497 | info.congested = 0; |
1498 | info.delivered = 0; | |
1499 | info.allocation = allocation; | |
1500 | info.skb = skb; | |
1501 | info.skb2 = NULL; | |
910a7e90 EB |
1502 | info.tx_filter = filter; |
1503 | info.tx_data = filter_data; | |
1da177e4 LT |
1504 | |
1505 | /* While we sleep in clone, do not allow to change socket list */ | |
1506 | ||
1507 | netlink_lock_table(); | |
1508 | ||
b67bfe0d | 1509 | sk_for_each_bound(sk, &nl_table[ssk->sk_protocol].mc_list) |
1da177e4 LT |
1510 | do_one_broadcast(sk, &info); |
1511 | ||
70d4bf6d | 1512 | consume_skb(skb); |
aa1c6a6f | 1513 | |
1da177e4 LT |
1514 | netlink_unlock_table(); |
1515 | ||
70d4bf6d NH |
1516 | if (info.delivery_failure) { |
1517 | kfree_skb(info.skb2); | |
ff491a73 | 1518 | return -ENOBUFS; |
658cb354 ED |
1519 | } |
1520 | consume_skb(info.skb2); | |
ff491a73 | 1521 | |
1da177e4 | 1522 | if (info.delivered) { |
d0164adc | 1523 | if (info.congested && gfpflags_allow_blocking(allocation)) |
1da177e4 LT |
1524 | yield(); |
1525 | return 0; | |
1526 | } | |
1da177e4 LT |
1527 | return -ESRCH; |
1528 | } | |
910a7e90 EB |
1529 | EXPORT_SYMBOL(netlink_broadcast_filtered); |
1530 | ||
15e47304 | 1531 | int netlink_broadcast(struct sock *ssk, struct sk_buff *skb, u32 portid, |
910a7e90 EB |
1532 | u32 group, gfp_t allocation) |
1533 | { | |
15e47304 | 1534 | return netlink_broadcast_filtered(ssk, skb, portid, group, allocation, |
910a7e90 EB |
1535 | NULL, NULL); |
1536 | } | |
6ac552fd | 1537 | EXPORT_SYMBOL(netlink_broadcast); |
1da177e4 LT |
1538 | |
1539 | struct netlink_set_err_data { | |
1540 | struct sock *exclude_sk; | |
15e47304 | 1541 | u32 portid; |
1da177e4 LT |
1542 | u32 group; |
1543 | int code; | |
1544 | }; | |
1545 | ||
b57ef81f | 1546 | static int do_one_set_err(struct sock *sk, struct netlink_set_err_data *p) |
1da177e4 LT |
1547 | { |
1548 | struct netlink_sock *nlk = nlk_sk(sk); | |
1a50307b | 1549 | int ret = 0; |
1da177e4 LT |
1550 | |
1551 | if (sk == p->exclude_sk) | |
1552 | goto out; | |
1553 | ||
09ad9bc7 | 1554 | if (!net_eq(sock_net(sk), sock_net(p->exclude_sk))) |
b4b51029 EB |
1555 | goto out; |
1556 | ||
15e47304 | 1557 | if (nlk->portid == p->portid || p->group - 1 >= nlk->ngroups || |
f7fa9b10 | 1558 | !test_bit(p->group - 1, nlk->groups)) |
1da177e4 LT |
1559 | goto out; |
1560 | ||
cc3a572f | 1561 | if (p->code == ENOBUFS && nlk->flags & NETLINK_F_RECV_NO_ENOBUFS) { |
1a50307b PNA |
1562 | ret = 1; |
1563 | goto out; | |
1564 | } | |
1565 | ||
1da177e4 LT |
1566 | sk->sk_err = p->code; |
1567 | sk->sk_error_report(sk); | |
1568 | out: | |
1a50307b | 1569 | return ret; |
1da177e4 LT |
1570 | } |
1571 | ||
4843b93c PNA |
1572 | /** |
1573 | * netlink_set_err - report error to broadcast listeners | |
1574 | * @ssk: the kernel netlink socket, as returned by netlink_kernel_create() | |
15e47304 | 1575 | * @portid: the PORTID of a process that we want to skip (if any) |
840e93f2 | 1576 | * @group: the broadcast group that will notice the error |
4843b93c | 1577 | * @code: error code, must be negative (as usual in kernelspace) |
1a50307b PNA |
1578 | * |
1579 | * This function returns the number of broadcast listeners that have set the | |
cc3a572f | 1580 | * NETLINK_NO_ENOBUFS socket option. |
4843b93c | 1581 | */ |
15e47304 | 1582 | int netlink_set_err(struct sock *ssk, u32 portid, u32 group, int code) |
1da177e4 LT |
1583 | { |
1584 | struct netlink_set_err_data info; | |
1da177e4 | 1585 | struct sock *sk; |
1a50307b | 1586 | int ret = 0; |
1da177e4 LT |
1587 | |
1588 | info.exclude_sk = ssk; | |
15e47304 | 1589 | info.portid = portid; |
1da177e4 | 1590 | info.group = group; |
4843b93c PNA |
1591 | /* sk->sk_err wants a positive error value */ |
1592 | info.code = -code; | |
1da177e4 LT |
1593 | |
1594 | read_lock(&nl_table_lock); | |
1595 | ||
b67bfe0d | 1596 | sk_for_each_bound(sk, &nl_table[ssk->sk_protocol].mc_list) |
1a50307b | 1597 | ret += do_one_set_err(sk, &info); |
1da177e4 LT |
1598 | |
1599 | read_unlock(&nl_table_lock); | |
1a50307b | 1600 | return ret; |
1da177e4 | 1601 | } |
dd5b6ce6 | 1602 | EXPORT_SYMBOL(netlink_set_err); |
1da177e4 | 1603 | |
84659eb5 JB |
1604 | /* must be called with netlink table grabbed */ |
1605 | static void netlink_update_socket_mc(struct netlink_sock *nlk, | |
1606 | unsigned int group, | |
1607 | int is_new) | |
1608 | { | |
1609 | int old, new = !!is_new, subscriptions; | |
1610 | ||
1611 | old = test_bit(group - 1, nlk->groups); | |
1612 | subscriptions = nlk->subscriptions - old + new; | |
1613 | if (new) | |
1614 | __set_bit(group - 1, nlk->groups); | |
1615 | else | |
1616 | __clear_bit(group - 1, nlk->groups); | |
1617 | netlink_update_subscriptions(&nlk->sk, subscriptions); | |
1618 | netlink_update_listeners(&nlk->sk); | |
1619 | } | |
1620 | ||
9a4595bc | 1621 | static int netlink_setsockopt(struct socket *sock, int level, int optname, |
b7058842 | 1622 | char __user *optval, unsigned int optlen) |
9a4595bc PM |
1623 | { |
1624 | struct sock *sk = sock->sk; | |
1625 | struct netlink_sock *nlk = nlk_sk(sk); | |
eb496534 JB |
1626 | unsigned int val = 0; |
1627 | int err; | |
9a4595bc PM |
1628 | |
1629 | if (level != SOL_NETLINK) | |
1630 | return -ENOPROTOOPT; | |
1631 | ||
d1b4c689 | 1632 | if (optlen >= sizeof(int) && |
eb496534 | 1633 | get_user(val, (unsigned int __user *)optval)) |
9a4595bc PM |
1634 | return -EFAULT; |
1635 | ||
1636 | switch (optname) { | |
1637 | case NETLINK_PKTINFO: | |
1638 | if (val) | |
cc3a572f | 1639 | nlk->flags |= NETLINK_F_RECV_PKTINFO; |
9a4595bc | 1640 | else |
cc3a572f | 1641 | nlk->flags &= ~NETLINK_F_RECV_PKTINFO; |
9a4595bc PM |
1642 | err = 0; |
1643 | break; | |
1644 | case NETLINK_ADD_MEMBERSHIP: | |
1645 | case NETLINK_DROP_MEMBERSHIP: { | |
5187cd05 | 1646 | if (!netlink_allowed(sock, NL_CFG_F_NONROOT_RECV)) |
9a4595bc | 1647 | return -EPERM; |
b4ff4f04 JB |
1648 | err = netlink_realloc_groups(sk); |
1649 | if (err) | |
1650 | return err; | |
9a4595bc PM |
1651 | if (!val || val - 1 >= nlk->ngroups) |
1652 | return -EINVAL; | |
7774d5e0 | 1653 | if (optname == NETLINK_ADD_MEMBERSHIP && nlk->netlink_bind) { |
023e2cfa | 1654 | err = nlk->netlink_bind(sock_net(sk), val); |
4f520900 RGB |
1655 | if (err) |
1656 | return err; | |
1657 | } | |
9a4595bc | 1658 | netlink_table_grab(); |
84659eb5 JB |
1659 | netlink_update_socket_mc(nlk, val, |
1660 | optname == NETLINK_ADD_MEMBERSHIP); | |
9a4595bc | 1661 | netlink_table_ungrab(); |
7774d5e0 | 1662 | if (optname == NETLINK_DROP_MEMBERSHIP && nlk->netlink_unbind) |
023e2cfa | 1663 | nlk->netlink_unbind(sock_net(sk), val); |
03292745 | 1664 | |
9a4595bc PM |
1665 | err = 0; |
1666 | break; | |
1667 | } | |
be0c22a4 PNA |
1668 | case NETLINK_BROADCAST_ERROR: |
1669 | if (val) | |
cc3a572f | 1670 | nlk->flags |= NETLINK_F_BROADCAST_SEND_ERROR; |
be0c22a4 | 1671 | else |
cc3a572f | 1672 | nlk->flags &= ~NETLINK_F_BROADCAST_SEND_ERROR; |
be0c22a4 PNA |
1673 | err = 0; |
1674 | break; | |
38938bfe PNA |
1675 | case NETLINK_NO_ENOBUFS: |
1676 | if (val) { | |
cc3a572f ND |
1677 | nlk->flags |= NETLINK_F_RECV_NO_ENOBUFS; |
1678 | clear_bit(NETLINK_S_CONGESTED, &nlk->state); | |
38938bfe | 1679 | wake_up_interruptible(&nlk->wait); |
658cb354 | 1680 | } else { |
cc3a572f | 1681 | nlk->flags &= ~NETLINK_F_RECV_NO_ENOBUFS; |
658cb354 | 1682 | } |
38938bfe PNA |
1683 | err = 0; |
1684 | break; | |
59324cf3 ND |
1685 | case NETLINK_LISTEN_ALL_NSID: |
1686 | if (!ns_capable(sock_net(sk)->user_ns, CAP_NET_BROADCAST)) | |
1687 | return -EPERM; | |
1688 | ||
1689 | if (val) | |
1690 | nlk->flags |= NETLINK_F_LISTEN_ALL_NSID; | |
1691 | else | |
1692 | nlk->flags &= ~NETLINK_F_LISTEN_ALL_NSID; | |
1693 | err = 0; | |
1694 | break; | |
0a6a3a23 CR |
1695 | case NETLINK_CAP_ACK: |
1696 | if (val) | |
1697 | nlk->flags |= NETLINK_F_CAP_ACK; | |
1698 | else | |
1699 | nlk->flags &= ~NETLINK_F_CAP_ACK; | |
1700 | err = 0; | |
1701 | break; | |
2d4bc933 JB |
1702 | case NETLINK_EXT_ACK: |
1703 | if (val) | |
1704 | nlk->flags |= NETLINK_F_EXT_ACK; | |
1705 | else | |
1706 | nlk->flags &= ~NETLINK_F_EXT_ACK; | |
1707 | err = 0; | |
1708 | break; | |
d3e8869e | 1709 | case NETLINK_GET_STRICT_CHK: |
89d35528 DA |
1710 | if (val) |
1711 | nlk->flags |= NETLINK_F_STRICT_CHK; | |
1712 | else | |
1713 | nlk->flags &= ~NETLINK_F_STRICT_CHK; | |
1714 | err = 0; | |
1715 | break; | |
9a4595bc PM |
1716 | default: |
1717 | err = -ENOPROTOOPT; | |
1718 | } | |
1719 | return err; | |
1720 | } | |
1721 | ||
1722 | static int netlink_getsockopt(struct socket *sock, int level, int optname, | |
746fac4d | 1723 | char __user *optval, int __user *optlen) |
9a4595bc PM |
1724 | { |
1725 | struct sock *sk = sock->sk; | |
1726 | struct netlink_sock *nlk = nlk_sk(sk); | |
1727 | int len, val, err; | |
1728 | ||
1729 | if (level != SOL_NETLINK) | |
1730 | return -ENOPROTOOPT; | |
1731 | ||
1732 | if (get_user(len, optlen)) | |
1733 | return -EFAULT; | |
1734 | if (len < 0) | |
1735 | return -EINVAL; | |
1736 | ||
1737 | switch (optname) { | |
1738 | case NETLINK_PKTINFO: | |
1739 | if (len < sizeof(int)) | |
1740 | return -EINVAL; | |
1741 | len = sizeof(int); | |
cc3a572f | 1742 | val = nlk->flags & NETLINK_F_RECV_PKTINFO ? 1 : 0; |
a27b58fe HC |
1743 | if (put_user(len, optlen) || |
1744 | put_user(val, optval)) | |
1745 | return -EFAULT; | |
9a4595bc PM |
1746 | err = 0; |
1747 | break; | |
be0c22a4 PNA |
1748 | case NETLINK_BROADCAST_ERROR: |
1749 | if (len < sizeof(int)) | |
1750 | return -EINVAL; | |
1751 | len = sizeof(int); | |
cc3a572f | 1752 | val = nlk->flags & NETLINK_F_BROADCAST_SEND_ERROR ? 1 : 0; |
be0c22a4 PNA |
1753 | if (put_user(len, optlen) || |
1754 | put_user(val, optval)) | |
1755 | return -EFAULT; | |
1756 | err = 0; | |
1757 | break; | |
38938bfe PNA |
1758 | case NETLINK_NO_ENOBUFS: |
1759 | if (len < sizeof(int)) | |
1760 | return -EINVAL; | |
1761 | len = sizeof(int); | |
cc3a572f | 1762 | val = nlk->flags & NETLINK_F_RECV_NO_ENOBUFS ? 1 : 0; |
38938bfe PNA |
1763 | if (put_user(len, optlen) || |
1764 | put_user(val, optval)) | |
1765 | return -EFAULT; | |
1766 | err = 0; | |
1767 | break; | |
b42be38b DH |
1768 | case NETLINK_LIST_MEMBERSHIPS: { |
1769 | int pos, idx, shift; | |
1770 | ||
1771 | err = 0; | |
47191d65 | 1772 | netlink_lock_table(); |
b42be38b DH |
1773 | for (pos = 0; pos * 8 < nlk->ngroups; pos += sizeof(u32)) { |
1774 | if (len - pos < sizeof(u32)) | |
1775 | break; | |
1776 | ||
1777 | idx = pos / sizeof(unsigned long); | |
1778 | shift = (pos % sizeof(unsigned long)) * 8; | |
1779 | if (put_user((u32)(nlk->groups[idx] >> shift), | |
1780 | (u32 __user *)(optval + pos))) { | |
1781 | err = -EFAULT; | |
1782 | break; | |
1783 | } | |
1784 | } | |
1785 | if (put_user(ALIGN(nlk->ngroups / 8, sizeof(u32)), optlen)) | |
1786 | err = -EFAULT; | |
47191d65 | 1787 | netlink_unlock_table(); |
b42be38b DH |
1788 | break; |
1789 | } | |
0a6a3a23 CR |
1790 | case NETLINK_CAP_ACK: |
1791 | if (len < sizeof(int)) | |
1792 | return -EINVAL; | |
1793 | len = sizeof(int); | |
1794 | val = nlk->flags & NETLINK_F_CAP_ACK ? 1 : 0; | |
1795 | if (put_user(len, optlen) || | |
1796 | put_user(val, optval)) | |
1797 | return -EFAULT; | |
1798 | err = 0; | |
1799 | break; | |
2d4bc933 JB |
1800 | case NETLINK_EXT_ACK: |
1801 | if (len < sizeof(int)) | |
1802 | return -EINVAL; | |
1803 | len = sizeof(int); | |
1804 | val = nlk->flags & NETLINK_F_EXT_ACK ? 1 : 0; | |
1805 | if (put_user(len, optlen) || put_user(val, optval)) | |
1806 | return -EFAULT; | |
1807 | err = 0; | |
1808 | break; | |
d3e8869e | 1809 | case NETLINK_GET_STRICT_CHK: |
89d35528 DA |
1810 | if (len < sizeof(int)) |
1811 | return -EINVAL; | |
1812 | len = sizeof(int); | |
1813 | val = nlk->flags & NETLINK_F_STRICT_CHK ? 1 : 0; | |
1814 | if (put_user(len, optlen) || put_user(val, optval)) | |
1815 | return -EFAULT; | |
1816 | err = 0; | |
1817 | break; | |
9a4595bc PM |
1818 | default: |
1819 | err = -ENOPROTOOPT; | |
1820 | } | |
1821 | return err; | |
1822 | } | |
1823 | ||
1824 | static void netlink_cmsg_recv_pktinfo(struct msghdr *msg, struct sk_buff *skb) | |
1825 | { | |
1826 | struct nl_pktinfo info; | |
1827 | ||
1828 | info.group = NETLINK_CB(skb).dst_group; | |
1829 | put_cmsg(msg, SOL_NETLINK, NETLINK_PKTINFO, sizeof(info), &info); | |
1830 | } | |
1831 | ||
59324cf3 ND |
1832 | static void netlink_cmsg_listen_all_nsid(struct sock *sk, struct msghdr *msg, |
1833 | struct sk_buff *skb) | |
1834 | { | |
1835 | if (!NETLINK_CB(skb).nsid_is_set) | |
1836 | return; | |
1837 | ||
1838 | put_cmsg(msg, SOL_NETLINK, NETLINK_LISTEN_ALL_NSID, sizeof(int), | |
1839 | &NETLINK_CB(skb).nsid); | |
1840 | } | |
1841 | ||
1b784140 | 1842 | static int netlink_sendmsg(struct socket *sock, struct msghdr *msg, size_t len) |
1da177e4 | 1843 | { |
1da177e4 LT |
1844 | struct sock *sk = sock->sk; |
1845 | struct netlink_sock *nlk = nlk_sk(sk); | |
342dfc30 | 1846 | DECLARE_SOCKADDR(struct sockaddr_nl *, addr, msg->msg_name); |
15e47304 | 1847 | u32 dst_portid; |
d629b836 | 1848 | u32 dst_group; |
1da177e4 LT |
1849 | struct sk_buff *skb; |
1850 | int err; | |
1851 | struct scm_cookie scm; | |
2d7a85f4 | 1852 | u32 netlink_skb_flags = 0; |
1da177e4 LT |
1853 | |
1854 | if (msg->msg_flags&MSG_OOB) | |
1855 | return -EOPNOTSUPP; | |
1856 | ||
7cc05662 | 1857 | err = scm_send(sock, msg, &scm, true); |
1da177e4 LT |
1858 | if (err < 0) |
1859 | return err; | |
1860 | ||
1861 | if (msg->msg_namelen) { | |
b47030c7 | 1862 | err = -EINVAL; |
6091f09c ED |
1863 | if (msg->msg_namelen < sizeof(struct sockaddr_nl)) |
1864 | goto out; | |
1da177e4 | 1865 | if (addr->nl_family != AF_NETLINK) |
b47030c7 | 1866 | goto out; |
15e47304 | 1867 | dst_portid = addr->nl_pid; |
d629b836 | 1868 | dst_group = ffs(addr->nl_groups); |
b47030c7 | 1869 | err = -EPERM; |
15e47304 | 1870 | if ((dst_group || dst_portid) && |
5187cd05 | 1871 | !netlink_allowed(sock, NL_CFG_F_NONROOT_SEND)) |
b47030c7 | 1872 | goto out; |
2d7a85f4 | 1873 | netlink_skb_flags |= NETLINK_SKB_DST; |
1da177e4 | 1874 | } else { |
15e47304 | 1875 | dst_portid = nlk->dst_portid; |
d629b836 | 1876 | dst_group = nlk->dst_group; |
1da177e4 LT |
1877 | } |
1878 | ||
da314c99 | 1879 | if (!nlk->bound) { |
1da177e4 LT |
1880 | err = netlink_autobind(sock); |
1881 | if (err) | |
1882 | goto out; | |
da314c99 HX |
1883 | } else { |
1884 | /* Ensure nlk is hashed and visible. */ | |
1885 | smp_rmb(); | |
1da177e4 LT |
1886 | } |
1887 | ||
1888 | err = -EMSGSIZE; | |
1889 | if (len > sk->sk_sndbuf - 32) | |
1890 | goto out; | |
1891 | err = -ENOBUFS; | |
3a36515f | 1892 | skb = netlink_alloc_large_skb(len, dst_group); |
6ac552fd | 1893 | if (skb == NULL) |
1da177e4 LT |
1894 | goto out; |
1895 | ||
15e47304 | 1896 | NETLINK_CB(skb).portid = nlk->portid; |
d629b836 | 1897 | NETLINK_CB(skb).dst_group = dst_group; |
7cc05662 | 1898 | NETLINK_CB(skb).creds = scm.creds; |
2d7a85f4 | 1899 | NETLINK_CB(skb).flags = netlink_skb_flags; |
1da177e4 | 1900 | |
1da177e4 | 1901 | err = -EFAULT; |
6ce8e9ce | 1902 | if (memcpy_from_msg(skb_put(skb, len), msg, len)) { |
1da177e4 LT |
1903 | kfree_skb(skb); |
1904 | goto out; | |
1905 | } | |
1906 | ||
1907 | err = security_netlink_send(sk, skb); | |
1908 | if (err) { | |
1909 | kfree_skb(skb); | |
1910 | goto out; | |
1911 | } | |
1912 | ||
d629b836 | 1913 | if (dst_group) { |
63354797 | 1914 | refcount_inc(&skb->users); |
15e47304 | 1915 | netlink_broadcast(sk, skb, dst_portid, dst_group, GFP_KERNEL); |
1da177e4 | 1916 | } |
15e47304 | 1917 | err = netlink_unicast(sk, skb, dst_portid, msg->msg_flags&MSG_DONTWAIT); |
1da177e4 LT |
1918 | |
1919 | out: | |
7cc05662 | 1920 | scm_destroy(&scm); |
1da177e4 LT |
1921 | return err; |
1922 | } | |
1923 | ||
1b784140 | 1924 | static int netlink_recvmsg(struct socket *sock, struct msghdr *msg, size_t len, |
1da177e4 LT |
1925 | int flags) |
1926 | { | |
1da177e4 LT |
1927 | struct scm_cookie scm; |
1928 | struct sock *sk = sock->sk; | |
1929 | struct netlink_sock *nlk = nlk_sk(sk); | |
1930 | int noblock = flags&MSG_DONTWAIT; | |
1931 | size_t copied; | |
68d6ac6d | 1932 | struct sk_buff *skb, *data_skb; |
b44d211e | 1933 | int err, ret; |
1da177e4 LT |
1934 | |
1935 | if (flags&MSG_OOB) | |
1936 | return -EOPNOTSUPP; | |
1937 | ||
1938 | copied = 0; | |
1939 | ||
6ac552fd PM |
1940 | skb = skb_recv_datagram(sk, flags, noblock, &err); |
1941 | if (skb == NULL) | |
1da177e4 LT |
1942 | goto out; |
1943 | ||
68d6ac6d JB |
1944 | data_skb = skb; |
1945 | ||
1dacc76d JB |
1946 | #ifdef CONFIG_COMPAT_NETLINK_MESSAGES |
1947 | if (unlikely(skb_shinfo(skb)->frag_list)) { | |
1dacc76d | 1948 | /* |
68d6ac6d JB |
1949 | * If this skb has a frag_list, then here that means that we |
1950 | * will have to use the frag_list skb's data for compat tasks | |
1951 | * and the regular skb's data for normal (non-compat) tasks. | |
1dacc76d | 1952 | * |
68d6ac6d JB |
1953 | * If we need to send the compat skb, assign it to the |
1954 | * 'data_skb' variable so that it will be used below for data | |
1955 | * copying. We keep 'skb' for everything else, including | |
1956 | * freeing both later. | |
1dacc76d | 1957 | */ |
68d6ac6d JB |
1958 | if (flags & MSG_CMSG_COMPAT) |
1959 | data_skb = skb_shinfo(skb)->frag_list; | |
1dacc76d JB |
1960 | } |
1961 | #endif | |
1962 | ||
9063e21f ED |
1963 | /* Record the max length of recvmsg() calls for future allocations */ |
1964 | nlk->max_recvmsg_len = max(nlk->max_recvmsg_len, len); | |
1965 | nlk->max_recvmsg_len = min_t(size_t, nlk->max_recvmsg_len, | |
d35c99ff | 1966 | SKB_WITH_OVERHEAD(32768)); |
9063e21f | 1967 | |
68d6ac6d | 1968 | copied = data_skb->len; |
1da177e4 LT |
1969 | if (len < copied) { |
1970 | msg->msg_flags |= MSG_TRUNC; | |
1971 | copied = len; | |
1972 | } | |
1973 | ||
68d6ac6d | 1974 | skb_reset_transport_header(data_skb); |
51f3d02b | 1975 | err = skb_copy_datagram_msg(data_skb, 0, msg, copied); |
1da177e4 LT |
1976 | |
1977 | if (msg->msg_name) { | |
342dfc30 | 1978 | DECLARE_SOCKADDR(struct sockaddr_nl *, addr, msg->msg_name); |
1da177e4 LT |
1979 | addr->nl_family = AF_NETLINK; |
1980 | addr->nl_pad = 0; | |
15e47304 | 1981 | addr->nl_pid = NETLINK_CB(skb).portid; |
d629b836 | 1982 | addr->nl_groups = netlink_group_mask(NETLINK_CB(skb).dst_group); |
1da177e4 LT |
1983 | msg->msg_namelen = sizeof(*addr); |
1984 | } | |
1985 | ||
cc3a572f | 1986 | if (nlk->flags & NETLINK_F_RECV_PKTINFO) |
cc9a06cd | 1987 | netlink_cmsg_recv_pktinfo(msg, skb); |
59324cf3 ND |
1988 | if (nlk->flags & NETLINK_F_LISTEN_ALL_NSID) |
1989 | netlink_cmsg_listen_all_nsid(sk, msg, skb); | |
cc9a06cd | 1990 | |
7cc05662 CH |
1991 | memset(&scm, 0, sizeof(scm)); |
1992 | scm.creds = *NETLINK_CREDS(skb); | |
188ccb55 | 1993 | if (flags & MSG_TRUNC) |
68d6ac6d | 1994 | copied = data_skb->len; |
daa3766e | 1995 | |
1da177e4 LT |
1996 | skb_free_datagram(sk, skb); |
1997 | ||
16b304f3 PS |
1998 | if (nlk->cb_running && |
1999 | atomic_read(&sk->sk_rmem_alloc) <= sk->sk_rcvbuf / 2) { | |
b44d211e AV |
2000 | ret = netlink_dump(sk); |
2001 | if (ret) { | |
ac30ef83 | 2002 | sk->sk_err = -ret; |
b44d211e AV |
2003 | sk->sk_error_report(sk); |
2004 | } | |
2005 | } | |
1da177e4 | 2006 | |
7cc05662 | 2007 | scm_recv(sock, msg, &scm, flags); |
1da177e4 LT |
2008 | out: |
2009 | netlink_rcv_wake(sk); | |
2010 | return err ? : copied; | |
2011 | } | |
2012 | ||
676d2369 | 2013 | static void netlink_data_ready(struct sock *sk) |
1da177e4 | 2014 | { |
cd40b7d3 | 2015 | BUG(); |
1da177e4 LT |
2016 | } |
2017 | ||
2018 | /* | |
746fac4d | 2019 | * We export these functions to other modules. They provide a |
1da177e4 LT |
2020 | * complete set of kernel non-blocking support for message |
2021 | * queueing. | |
2022 | */ | |
2023 | ||
2024 | struct sock * | |
9f00d977 PNA |
2025 | __netlink_kernel_create(struct net *net, int unit, struct module *module, |
2026 | struct netlink_kernel_cfg *cfg) | |
1da177e4 LT |
2027 | { |
2028 | struct socket *sock; | |
2029 | struct sock *sk; | |
77247bbb | 2030 | struct netlink_sock *nlk; |
5c398dc8 | 2031 | struct listeners *listeners = NULL; |
a31f2d17 PNA |
2032 | struct mutex *cb_mutex = cfg ? cfg->cb_mutex : NULL; |
2033 | unsigned int groups; | |
1da177e4 | 2034 | |
fab2caf6 | 2035 | BUG_ON(!nl_table); |
1da177e4 | 2036 | |
6ac552fd | 2037 | if (unit < 0 || unit >= MAX_LINKS) |
1da177e4 LT |
2038 | return NULL; |
2039 | ||
2040 | if (sock_create_lite(PF_NETLINK, SOCK_DGRAM, unit, &sock)) | |
2041 | return NULL; | |
13d3078e EB |
2042 | |
2043 | if (__netlink_create(net, sock, cb_mutex, unit, 1) < 0) | |
23fe1866 PE |
2044 | goto out_sock_release_nosk; |
2045 | ||
2046 | sk = sock->sk; | |
4fdb3bb7 | 2047 | |
a31f2d17 | 2048 | if (!cfg || cfg->groups < 32) |
4277a083 | 2049 | groups = 32; |
a31f2d17 PNA |
2050 | else |
2051 | groups = cfg->groups; | |
4277a083 | 2052 | |
5c398dc8 | 2053 | listeners = kzalloc(sizeof(*listeners) + NLGRPSZ(groups), GFP_KERNEL); |
4277a083 PM |
2054 | if (!listeners) |
2055 | goto out_sock_release; | |
2056 | ||
1da177e4 | 2057 | sk->sk_data_ready = netlink_data_ready; |
a31f2d17 PNA |
2058 | if (cfg && cfg->input) |
2059 | nlk_sk(sk)->netlink_rcv = cfg->input; | |
1da177e4 | 2060 | |
8ea65f4a | 2061 | if (netlink_insert(sk, 0)) |
77247bbb | 2062 | goto out_sock_release; |
4fdb3bb7 | 2063 | |
77247bbb | 2064 | nlk = nlk_sk(sk); |
cc3a572f | 2065 | nlk->flags |= NETLINK_F_KERNEL_SOCKET; |
4fdb3bb7 | 2066 | |
4fdb3bb7 | 2067 | netlink_table_grab(); |
b4b51029 EB |
2068 | if (!nl_table[unit].registered) { |
2069 | nl_table[unit].groups = groups; | |
5c398dc8 | 2070 | rcu_assign_pointer(nl_table[unit].listeners, listeners); |
b4b51029 EB |
2071 | nl_table[unit].cb_mutex = cb_mutex; |
2072 | nl_table[unit].module = module; | |
9785e10a PNA |
2073 | if (cfg) { |
2074 | nl_table[unit].bind = cfg->bind; | |
6251edd9 | 2075 | nl_table[unit].unbind = cfg->unbind; |
9785e10a | 2076 | nl_table[unit].flags = cfg->flags; |
da12c90e G |
2077 | if (cfg->compare) |
2078 | nl_table[unit].compare = cfg->compare; | |
9785e10a | 2079 | } |
b4b51029 | 2080 | nl_table[unit].registered = 1; |
f937f1f4 JJ |
2081 | } else { |
2082 | kfree(listeners); | |
869e58f8 | 2083 | nl_table[unit].registered++; |
b4b51029 | 2084 | } |
4fdb3bb7 | 2085 | netlink_table_ungrab(); |
77247bbb PM |
2086 | return sk; |
2087 | ||
4fdb3bb7 | 2088 | out_sock_release: |
4277a083 | 2089 | kfree(listeners); |
9dfbec1f | 2090 | netlink_kernel_release(sk); |
23fe1866 PE |
2091 | return NULL; |
2092 | ||
2093 | out_sock_release_nosk: | |
4fdb3bb7 | 2094 | sock_release(sock); |
77247bbb | 2095 | return NULL; |
1da177e4 | 2096 | } |
9f00d977 | 2097 | EXPORT_SYMBOL(__netlink_kernel_create); |
b7c6ba6e DL |
2098 | |
2099 | void | |
2100 | netlink_kernel_release(struct sock *sk) | |
2101 | { | |
13d3078e EB |
2102 | if (sk == NULL || sk->sk_socket == NULL) |
2103 | return; | |
2104 | ||
2105 | sock_release(sk->sk_socket); | |
b7c6ba6e DL |
2106 | } |
2107 | EXPORT_SYMBOL(netlink_kernel_release); | |
2108 | ||
d136f1bd | 2109 | int __netlink_change_ngroups(struct sock *sk, unsigned int groups) |
b4ff4f04 | 2110 | { |
5c398dc8 | 2111 | struct listeners *new, *old; |
b4ff4f04 | 2112 | struct netlink_table *tbl = &nl_table[sk->sk_protocol]; |
b4ff4f04 JB |
2113 | |
2114 | if (groups < 32) | |
2115 | groups = 32; | |
2116 | ||
b4ff4f04 | 2117 | if (NLGRPSZ(tbl->groups) < NLGRPSZ(groups)) { |
5c398dc8 ED |
2118 | new = kzalloc(sizeof(*new) + NLGRPSZ(groups), GFP_ATOMIC); |
2119 | if (!new) | |
d136f1bd | 2120 | return -ENOMEM; |
6d772ac5 | 2121 | old = nl_deref_protected(tbl->listeners); |
5c398dc8 ED |
2122 | memcpy(new->masks, old->masks, NLGRPSZ(tbl->groups)); |
2123 | rcu_assign_pointer(tbl->listeners, new); | |
2124 | ||
37b6b935 | 2125 | kfree_rcu(old, rcu); |
b4ff4f04 JB |
2126 | } |
2127 | tbl->groups = groups; | |
2128 | ||
d136f1bd JB |
2129 | return 0; |
2130 | } | |
2131 | ||
2132 | /** | |
2133 | * netlink_change_ngroups - change number of multicast groups | |
2134 | * | |
2135 | * This changes the number of multicast groups that are available | |
2136 | * on a certain netlink family. Note that it is not possible to | |
2137 | * change the number of groups to below 32. Also note that it does | |
2138 | * not implicitly call netlink_clear_multicast_users() when the | |
2139 | * number of groups is reduced. | |
2140 | * | |
2141 | * @sk: The kernel netlink socket, as returned by netlink_kernel_create(). | |
2142 | * @groups: The new number of groups. | |
2143 | */ | |
2144 | int netlink_change_ngroups(struct sock *sk, unsigned int groups) | |
2145 | { | |
2146 | int err; | |
2147 | ||
2148 | netlink_table_grab(); | |
2149 | err = __netlink_change_ngroups(sk, groups); | |
b4ff4f04 | 2150 | netlink_table_ungrab(); |
d136f1bd | 2151 | |
b4ff4f04 JB |
2152 | return err; |
2153 | } | |
b4ff4f04 | 2154 | |
b8273570 JB |
2155 | void __netlink_clear_multicast_users(struct sock *ksk, unsigned int group) |
2156 | { | |
2157 | struct sock *sk; | |
b8273570 JB |
2158 | struct netlink_table *tbl = &nl_table[ksk->sk_protocol]; |
2159 | ||
b67bfe0d | 2160 | sk_for_each_bound(sk, &tbl->mc_list) |
b8273570 JB |
2161 | netlink_update_socket_mc(nlk_sk(sk), group, 0); |
2162 | } | |
2163 | ||
a46621a3 | 2164 | struct nlmsghdr * |
15e47304 | 2165 | __nlmsg_put(struct sk_buff *skb, u32 portid, u32 seq, int type, int len, int flags) |
a46621a3 DV |
2166 | { |
2167 | struct nlmsghdr *nlh; | |
573ce260 | 2168 | int size = nlmsg_msg_size(len); |
a46621a3 | 2169 | |
4df864c1 | 2170 | nlh = skb_put(skb, NLMSG_ALIGN(size)); |
a46621a3 DV |
2171 | nlh->nlmsg_type = type; |
2172 | nlh->nlmsg_len = size; | |
2173 | nlh->nlmsg_flags = flags; | |
15e47304 | 2174 | nlh->nlmsg_pid = portid; |
a46621a3 DV |
2175 | nlh->nlmsg_seq = seq; |
2176 | if (!__builtin_constant_p(size) || NLMSG_ALIGN(size) - size != 0) | |
573ce260 | 2177 | memset(nlmsg_data(nlh) + len, 0, NLMSG_ALIGN(size) - size); |
a46621a3 DV |
2178 | return nlh; |
2179 | } | |
2180 | EXPORT_SYMBOL(__nlmsg_put); | |
2181 | ||
1da177e4 LT |
2182 | /* |
2183 | * It looks a bit ugly. | |
2184 | * It would be better to create kernel thread. | |
2185 | */ | |
2186 | ||
2187 | static int netlink_dump(struct sock *sk) | |
2188 | { | |
2189 | struct netlink_sock *nlk = nlk_sk(sk); | |
4a19edb6 | 2190 | struct netlink_ext_ack extack = {}; |
1da177e4 | 2191 | struct netlink_callback *cb; |
c7ac8679 | 2192 | struct sk_buff *skb = NULL; |
1da177e4 | 2193 | struct nlmsghdr *nlh; |
92964c79 | 2194 | struct module *module; |
0642840b | 2195 | int err = -ENOBUFS; |
db65a3aa | 2196 | int alloc_min_size; |
c7ac8679 | 2197 | int alloc_size; |
1da177e4 | 2198 | |
af65bdfc | 2199 | mutex_lock(nlk->cb_mutex); |
16b304f3 | 2200 | if (!nlk->cb_running) { |
bf8b79e4 TG |
2201 | err = -EINVAL; |
2202 | goto errout_skb; | |
1da177e4 LT |
2203 | } |
2204 | ||
d1b4c689 | 2205 | if (atomic_read(&sk->sk_rmem_alloc) >= sk->sk_rcvbuf) |
f9c22888 | 2206 | goto errout_skb; |
9063e21f ED |
2207 | |
2208 | /* NLMSG_GOODSIZE is small to avoid high order allocations being | |
2209 | * required, but it makes sense to _attempt_ a 16K bytes allocation | |
2210 | * to reduce number of system calls on dump operations, if user | |
2211 | * ever provided a big enough buffer. | |
2212 | */ | |
db65a3aa AR |
2213 | cb = &nlk->cb; |
2214 | alloc_min_size = max_t(int, cb->min_dump_alloc, NLMSG_GOODSIZE); | |
2215 | ||
2216 | if (alloc_min_size < nlk->max_recvmsg_len) { | |
2217 | alloc_size = nlk->max_recvmsg_len; | |
d35c99ff ED |
2218 | skb = alloc_skb(alloc_size, |
2219 | (GFP_KERNEL & ~__GFP_DIRECT_RECLAIM) | | |
2220 | __GFP_NOWARN | __GFP_NORETRY); | |
9063e21f | 2221 | } |
db65a3aa AR |
2222 | if (!skb) { |
2223 | alloc_size = alloc_min_size; | |
c5b0db32 | 2224 | skb = alloc_skb(alloc_size, GFP_KERNEL); |
db65a3aa | 2225 | } |
c7ac8679 | 2226 | if (!skb) |
c63d6ea3 | 2227 | goto errout_skb; |
db65a3aa AR |
2228 | |
2229 | /* Trim skb to allocated size. User is expected to provide buffer as | |
2230 | * large as max(min_dump_alloc, 16KiB (mac_recvmsg_len capped at | |
2231 | * netlink_recvmsg())). dump will pack as many smaller messages as | |
2232 | * could fit within the allocated skb. skb is typically allocated | |
2233 | * with larger space than required (could be as much as near 2x the | |
2234 | * requested size with align to next power of 2 approach). Allowing | |
2235 | * dump to use the excess space makes it difficult for a user to have a | |
2236 | * reasonable static buffer based on the expected largest dump of a | |
2237 | * single netdev. The outcome is MSG_TRUNC error. | |
2238 | */ | |
d1b4c689 | 2239 | skb_reserve(skb, skb_tailroom(skb) - alloc_size); |
f9c22888 | 2240 | netlink_skb_set_owner_r(skb, sk); |
c7ac8679 | 2241 | |
4a19edb6 DA |
2242 | if (nlk->dump_done_errno > 0) { |
2243 | cb->extack = &extack; | |
0642840b | 2244 | nlk->dump_done_errno = cb->dump(skb, cb); |
4a19edb6 DA |
2245 | cb->extack = NULL; |
2246 | } | |
1da177e4 | 2247 | |
0642840b JD |
2248 | if (nlk->dump_done_errno > 0 || |
2249 | skb_tailroom(skb) < nlmsg_total_size(sizeof(nlk->dump_done_errno))) { | |
af65bdfc | 2250 | mutex_unlock(nlk->cb_mutex); |
b1153f29 SH |
2251 | |
2252 | if (sk_filter(sk, skb)) | |
2253 | kfree_skb(skb); | |
4a7e7c2a ED |
2254 | else |
2255 | __netlink_sendskb(sk, skb); | |
1da177e4 LT |
2256 | return 0; |
2257 | } | |
2258 | ||
0642840b | 2259 | nlh = nlmsg_put_answer(skb, cb, NLMSG_DONE, |
22e6c58b DA |
2260 | sizeof(nlk->dump_done_errno), |
2261 | NLM_F_MULTI | cb->answer_flags); | |
0642840b | 2262 | if (WARN_ON(!nlh)) |
bf8b79e4 TG |
2263 | goto errout_skb; |
2264 | ||
670dc283 JB |
2265 | nl_dump_check_consistent(cb, nlh); |
2266 | ||
0642840b JD |
2267 | memcpy(nlmsg_data(nlh), &nlk->dump_done_errno, |
2268 | sizeof(nlk->dump_done_errno)); | |
bf8b79e4 | 2269 | |
4a19edb6 DA |
2270 | if (extack._msg && nlk->flags & NETLINK_F_EXT_ACK) { |
2271 | nlh->nlmsg_flags |= NLM_F_ACK_TLVS; | |
2272 | if (!nla_put_string(skb, NLMSGERR_ATTR_MSG, extack._msg)) | |
2273 | nlmsg_end(skb, nlh); | |
2274 | } | |
2275 | ||
b1153f29 SH |
2276 | if (sk_filter(sk, skb)) |
2277 | kfree_skb(skb); | |
4a7e7c2a ED |
2278 | else |
2279 | __netlink_sendskb(sk, skb); | |
1da177e4 | 2280 | |
a8f74b22 TG |
2281 | if (cb->done) |
2282 | cb->done(cb); | |
1da177e4 | 2283 | |
16b304f3 | 2284 | nlk->cb_running = false; |
92964c79 HX |
2285 | module = cb->module; |
2286 | skb = cb->skb; | |
16b304f3 | 2287 | mutex_unlock(nlk->cb_mutex); |
92964c79 HX |
2288 | module_put(module); |
2289 | consume_skb(skb); | |
1da177e4 | 2290 | return 0; |
1797754e | 2291 | |
bf8b79e4 | 2292 | errout_skb: |
af65bdfc | 2293 | mutex_unlock(nlk->cb_mutex); |
bf8b79e4 | 2294 | kfree_skb(skb); |
bf8b79e4 | 2295 | return err; |
1da177e4 LT |
2296 | } |
2297 | ||
6dc878a8 G |
2298 | int __netlink_dump_start(struct sock *ssk, struct sk_buff *skb, |
2299 | const struct nlmsghdr *nlh, | |
2300 | struct netlink_dump_control *control) | |
1da177e4 | 2301 | { |
89d35528 | 2302 | struct netlink_sock *nlk, *nlk2; |
1da177e4 LT |
2303 | struct netlink_callback *cb; |
2304 | struct sock *sk; | |
b44d211e | 2305 | int ret; |
1da177e4 | 2306 | |
63354797 | 2307 | refcount_inc(&skb->users); |
f9c22888 | 2308 | |
15e47304 | 2309 | sk = netlink_lookup(sock_net(ssk), ssk->sk_protocol, NETLINK_CB(skb).portid); |
1da177e4 | 2310 | if (sk == NULL) { |
16b304f3 PS |
2311 | ret = -ECONNREFUSED; |
2312 | goto error_free; | |
1da177e4 | 2313 | } |
6dc878a8 | 2314 | |
16b304f3 | 2315 | nlk = nlk_sk(sk); |
af65bdfc | 2316 | mutex_lock(nlk->cb_mutex); |
6dc878a8 | 2317 | /* A dump is in progress... */ |
16b304f3 | 2318 | if (nlk->cb_running) { |
6dc878a8 | 2319 | ret = -EBUSY; |
16b304f3 | 2320 | goto error_unlock; |
1da177e4 | 2321 | } |
6dc878a8 | 2322 | /* add reference of module which cb->dump belongs to */ |
16b304f3 | 2323 | if (!try_module_get(control->module)) { |
6dc878a8 | 2324 | ret = -EPROTONOSUPPORT; |
16b304f3 | 2325 | goto error_unlock; |
6dc878a8 G |
2326 | } |
2327 | ||
16b304f3 PS |
2328 | cb = &nlk->cb; |
2329 | memset(cb, 0, sizeof(*cb)); | |
2330 | cb->dump = control->dump; | |
2331 | cb->done = control->done; | |
2332 | cb->nlh = nlh; | |
2333 | cb->data = control->data; | |
2334 | cb->module = control->module; | |
2335 | cb->min_dump_alloc = control->min_dump_alloc; | |
2336 | cb->skb = skb; | |
2337 | ||
89d35528 DA |
2338 | nlk2 = nlk_sk(NETLINK_CB(skb).sk); |
2339 | cb->strict_check = !!(nlk2->flags & NETLINK_F_STRICT_CHK); | |
2340 | ||
3730cf4d FW |
2341 | if (control->start) { |
2342 | ret = control->start(cb); | |
41c87425 | 2343 | if (ret) |
b87b6194 | 2344 | goto error_put; |
41c87425 JD |
2345 | } |
2346 | ||
16b304f3 | 2347 | nlk->cb_running = true; |
0642840b | 2348 | nlk->dump_done_errno = INT_MAX; |
16b304f3 | 2349 | |
af65bdfc | 2350 | mutex_unlock(nlk->cb_mutex); |
1da177e4 | 2351 | |
41c87425 | 2352 | ret = netlink_dump(sk); |
fc9e50f5 | 2353 | |
1da177e4 | 2354 | sock_put(sk); |
5c58298c | 2355 | |
b44d211e AV |
2356 | if (ret) |
2357 | return ret; | |
2358 | ||
5c58298c DL |
2359 | /* We successfully started a dump, by returning -EINTR we |
2360 | * signal not to send ACK even if it was requested. | |
2361 | */ | |
2362 | return -EINTR; | |
16b304f3 | 2363 | |
b87b6194 JD |
2364 | error_put: |
2365 | module_put(control->module); | |
16b304f3 PS |
2366 | error_unlock: |
2367 | sock_put(sk); | |
2368 | mutex_unlock(nlk->cb_mutex); | |
2369 | error_free: | |
2370 | kfree_skb(skb); | |
2371 | return ret; | |
1da177e4 | 2372 | } |
6dc878a8 | 2373 | EXPORT_SYMBOL(__netlink_dump_start); |
1da177e4 | 2374 | |
2d4bc933 JB |
2375 | void netlink_ack(struct sk_buff *in_skb, struct nlmsghdr *nlh, int err, |
2376 | const struct netlink_ext_ack *extack) | |
1da177e4 LT |
2377 | { |
2378 | struct sk_buff *skb; | |
2379 | struct nlmsghdr *rep; | |
2380 | struct nlmsgerr *errmsg; | |
339bf98f | 2381 | size_t payload = sizeof(*errmsg); |
2d4bc933 | 2382 | size_t tlvlen = 0; |
0a6a3a23 | 2383 | struct netlink_sock *nlk = nlk_sk(NETLINK_CB(in_skb).sk); |
2d4bc933 | 2384 | unsigned int flags = 0; |
48044eb4 | 2385 | bool nlk_has_extack = nlk->flags & NETLINK_F_EXT_ACK; |
1da177e4 | 2386 | |
0a6a3a23 | 2387 | /* Error messages get the original request appened, unless the user |
2d4bc933 JB |
2388 | * requests to cap the error message, and get extra error data if |
2389 | * requested. | |
0a6a3a23 | 2390 | */ |
4f6265d4 DA |
2391 | if (nlk_has_extack && extack && extack->_msg) |
2392 | tlvlen += nla_total_size(strlen(extack->_msg) + 1); | |
2393 | ||
2d4bc933 JB |
2394 | if (err) { |
2395 | if (!(nlk->flags & NETLINK_F_CAP_ACK)) | |
2396 | payload += nlmsg_len(nlh); | |
2397 | else | |
2398 | flags |= NLM_F_CAPPED; | |
4f6265d4 DA |
2399 | if (nlk_has_extack && extack && extack->bad_attr) |
2400 | tlvlen += nla_total_size(sizeof(u32)); | |
2d4bc933 JB |
2401 | } else { |
2402 | flags |= NLM_F_CAPPED; | |
ba0dc5f6 | 2403 | |
48044eb4 | 2404 | if (nlk_has_extack && extack && extack->cookie_len) |
ba0dc5f6 | 2405 | tlvlen += nla_total_size(extack->cookie_len); |
2d4bc933 | 2406 | } |
1da177e4 | 2407 | |
2d4bc933 JB |
2408 | if (tlvlen) |
2409 | flags |= NLM_F_ACK_TLVS; | |
2410 | ||
2411 | skb = nlmsg_new(payload + tlvlen, GFP_KERNEL); | |
1da177e4 | 2412 | if (!skb) { |
a2084f56 JB |
2413 | NETLINK_CB(in_skb).sk->sk_err = ENOBUFS; |
2414 | NETLINK_CB(in_skb).sk->sk_error_report(NETLINK_CB(in_skb).sk); | |
1da177e4 LT |
2415 | return; |
2416 | } | |
2417 | ||
15e47304 | 2418 | rep = __nlmsg_put(skb, NETLINK_CB(in_skb).portid, nlh->nlmsg_seq, |
2d4bc933 | 2419 | NLMSG_ERROR, payload, flags); |
bf8b79e4 | 2420 | errmsg = nlmsg_data(rep); |
1da177e4 | 2421 | errmsg->error = err; |
0a6a3a23 | 2422 | memcpy(&errmsg->msg, nlh, payload > sizeof(*errmsg) ? nlh->nlmsg_len : sizeof(*nlh)); |
2d4bc933 | 2423 | |
48044eb4 | 2424 | if (nlk_has_extack && extack) { |
4f6265d4 DA |
2425 | if (extack->_msg) { |
2426 | WARN_ON(nla_put_string(skb, NLMSGERR_ATTR_MSG, | |
2427 | extack->_msg)); | |
2428 | } | |
ba0dc5f6 | 2429 | if (err) { |
ba0dc5f6 JB |
2430 | if (extack->bad_attr && |
2431 | !WARN_ON((u8 *)extack->bad_attr < in_skb->data || | |
2432 | (u8 *)extack->bad_attr >= in_skb->data + | |
2433 | in_skb->len)) | |
2434 | WARN_ON(nla_put_u32(skb, NLMSGERR_ATTR_OFFS, | |
2435 | (u8 *)extack->bad_attr - | |
2436 | in_skb->data)); | |
2437 | } else { | |
2438 | if (extack->cookie_len) | |
2439 | WARN_ON(nla_put(skb, NLMSGERR_ATTR_COOKIE, | |
2440 | extack->cookie_len, | |
2441 | extack->cookie)); | |
2442 | } | |
2d4bc933 JB |
2443 | } |
2444 | ||
2445 | nlmsg_end(skb, rep); | |
2446 | ||
15e47304 | 2447 | netlink_unicast(in_skb->sk, skb, NETLINK_CB(in_skb).portid, MSG_DONTWAIT); |
1da177e4 | 2448 | } |
6ac552fd | 2449 | EXPORT_SYMBOL(netlink_ack); |
1da177e4 | 2450 | |
cd40b7d3 | 2451 | int netlink_rcv_skb(struct sk_buff *skb, int (*cb)(struct sk_buff *, |
2d4bc933 JB |
2452 | struct nlmsghdr *, |
2453 | struct netlink_ext_ack *)) | |
82ace47a | 2454 | { |
cbbdf843 | 2455 | struct netlink_ext_ack extack; |
82ace47a TG |
2456 | struct nlmsghdr *nlh; |
2457 | int err; | |
2458 | ||
2459 | while (skb->len >= nlmsg_total_size(0)) { | |
cd40b7d3 DL |
2460 | int msglen; |
2461 | ||
cd443f1e | 2462 | memset(&extack, 0, sizeof(extack)); |
b529ccf2 | 2463 | nlh = nlmsg_hdr(skb); |
d35b6856 | 2464 | err = 0; |
82ace47a | 2465 | |
ad8e4b75 | 2466 | if (nlh->nlmsg_len < NLMSG_HDRLEN || skb->len < nlh->nlmsg_len) |
82ace47a TG |
2467 | return 0; |
2468 | ||
d35b6856 TG |
2469 | /* Only requests are handled by the kernel */ |
2470 | if (!(nlh->nlmsg_flags & NLM_F_REQUEST)) | |
5c58298c | 2471 | goto ack; |
45e7ae7f TG |
2472 | |
2473 | /* Skip control messages */ | |
2474 | if (nlh->nlmsg_type < NLMSG_MIN_TYPE) | |
5c58298c | 2475 | goto ack; |
d35b6856 | 2476 | |
2d4bc933 | 2477 | err = cb(skb, nlh, &extack); |
5c58298c DL |
2478 | if (err == -EINTR) |
2479 | goto skip; | |
2480 | ||
2481 | ack: | |
d35b6856 | 2482 | if (nlh->nlmsg_flags & NLM_F_ACK || err) |
2d4bc933 | 2483 | netlink_ack(skb, nlh, err, &extack); |
82ace47a | 2484 | |
5c58298c | 2485 | skip: |
6ac552fd | 2486 | msglen = NLMSG_ALIGN(nlh->nlmsg_len); |
cd40b7d3 DL |
2487 | if (msglen > skb->len) |
2488 | msglen = skb->len; | |
2489 | skb_pull(skb, msglen); | |
82ace47a TG |
2490 | } |
2491 | ||
2492 | return 0; | |
2493 | } | |
6ac552fd | 2494 | EXPORT_SYMBOL(netlink_rcv_skb); |
82ace47a | 2495 | |
d387f6ad TG |
2496 | /** |
2497 | * nlmsg_notify - send a notification netlink message | |
2498 | * @sk: netlink socket to use | |
2499 | * @skb: notification message | |
15e47304 | 2500 | * @portid: destination netlink portid for reports or 0 |
d387f6ad TG |
2501 | * @group: destination multicast group or 0 |
2502 | * @report: 1 to report back, 0 to disable | |
2503 | * @flags: allocation flags | |
2504 | */ | |
15e47304 | 2505 | int nlmsg_notify(struct sock *sk, struct sk_buff *skb, u32 portid, |
d387f6ad TG |
2506 | unsigned int group, int report, gfp_t flags) |
2507 | { | |
2508 | int err = 0; | |
2509 | ||
2510 | if (group) { | |
15e47304 | 2511 | int exclude_portid = 0; |
d387f6ad TG |
2512 | |
2513 | if (report) { | |
63354797 | 2514 | refcount_inc(&skb->users); |
15e47304 | 2515 | exclude_portid = portid; |
d387f6ad TG |
2516 | } |
2517 | ||
1ce85fe4 PNA |
2518 | /* errors reported via destination sk->sk_err, but propagate |
2519 | * delivery errors if NETLINK_BROADCAST_ERROR flag is set */ | |
15e47304 | 2520 | err = nlmsg_multicast(sk, skb, exclude_portid, group, flags); |
d387f6ad TG |
2521 | } |
2522 | ||
1ce85fe4 PNA |
2523 | if (report) { |
2524 | int err2; | |
2525 | ||
15e47304 | 2526 | err2 = nlmsg_unicast(sk, skb, portid); |
1ce85fe4 PNA |
2527 | if (!err || err == -ESRCH) |
2528 | err = err2; | |
2529 | } | |
d387f6ad TG |
2530 | |
2531 | return err; | |
2532 | } | |
6ac552fd | 2533 | EXPORT_SYMBOL(nlmsg_notify); |
d387f6ad | 2534 | |
1da177e4 LT |
2535 | #ifdef CONFIG_PROC_FS |
2536 | struct nl_seq_iter { | |
e372c414 | 2537 | struct seq_net_private p; |
56d28b1e | 2538 | struct rhashtable_iter hti; |
1da177e4 | 2539 | int link; |
1da177e4 LT |
2540 | }; |
2541 | ||
abf9979f | 2542 | static void netlink_walk_start(struct nl_seq_iter *iter) |
1da177e4 | 2543 | { |
6c4128f6 | 2544 | rhashtable_walk_enter(&nl_table[iter->link].hash, &iter->hti); |
97a6ec4a | 2545 | rhashtable_walk_start(&iter->hti); |
1da177e4 LT |
2546 | } |
2547 | ||
56d28b1e | 2548 | static void netlink_walk_stop(struct nl_seq_iter *iter) |
1da177e4 | 2549 | { |
56d28b1e HX |
2550 | rhashtable_walk_stop(&iter->hti); |
2551 | rhashtable_walk_exit(&iter->hti); | |
1da177e4 LT |
2552 | } |
2553 | ||
56d28b1e | 2554 | static void *__netlink_seq_next(struct seq_file *seq) |
1da177e4 | 2555 | { |
56d28b1e | 2556 | struct nl_seq_iter *iter = seq->private; |
e341694e | 2557 | struct netlink_sock *nlk; |
1da177e4 | 2558 | |
56d28b1e HX |
2559 | do { |
2560 | for (;;) { | |
56d28b1e | 2561 | nlk = rhashtable_walk_next(&iter->hti); |
746fac4d | 2562 | |
56d28b1e HX |
2563 | if (IS_ERR(nlk)) { |
2564 | if (PTR_ERR(nlk) == -EAGAIN) | |
2565 | continue; | |
e341694e | 2566 | |
56d28b1e HX |
2567 | return nlk; |
2568 | } | |
1da177e4 | 2569 | |
56d28b1e HX |
2570 | if (nlk) |
2571 | break; | |
1da177e4 | 2572 | |
56d28b1e HX |
2573 | netlink_walk_stop(iter); |
2574 | if (++iter->link >= MAX_LINKS) | |
2575 | return NULL; | |
da12c90e | 2576 | |
abf9979f | 2577 | netlink_walk_start(iter); |
1da177e4 | 2578 | } |
56d28b1e | 2579 | } while (sock_net(&nlk->sk) != seq_file_net(seq)); |
1da177e4 | 2580 | |
56d28b1e HX |
2581 | return nlk; |
2582 | } | |
1da177e4 | 2583 | |
56d28b1e HX |
2584 | static void *netlink_seq_start(struct seq_file *seq, loff_t *posp) |
2585 | { | |
2586 | struct nl_seq_iter *iter = seq->private; | |
2587 | void *obj = SEQ_START_TOKEN; | |
2588 | loff_t pos; | |
56d28b1e HX |
2589 | |
2590 | iter->link = 0; | |
2591 | ||
abf9979f | 2592 | netlink_walk_start(iter); |
56d28b1e HX |
2593 | |
2594 | for (pos = *posp; pos && obj && !IS_ERR(obj); pos--) | |
2595 | obj = __netlink_seq_next(seq); | |
2596 | ||
2597 | return obj; | |
2598 | } | |
2599 | ||
2600 | static void *netlink_seq_next(struct seq_file *seq, void *v, loff_t *pos) | |
2601 | { | |
2602 | ++*pos; | |
2603 | return __netlink_seq_next(seq); | |
1da177e4 LT |
2604 | } |
2605 | ||
2606 | static void netlink_seq_stop(struct seq_file *seq, void *v) | |
2607 | { | |
56d28b1e HX |
2608 | struct nl_seq_iter *iter = seq->private; |
2609 | ||
2610 | if (iter->link >= MAX_LINKS) | |
2611 | return; | |
2612 | ||
2613 | netlink_walk_stop(iter); | |
1da177e4 LT |
2614 | } |
2615 | ||
2616 | ||
2617 | static int netlink_seq_show(struct seq_file *seq, void *v) | |
2618 | { | |
658cb354 | 2619 | if (v == SEQ_START_TOKEN) { |
1da177e4 | 2620 | seq_puts(seq, |
ae552ac2 YB |
2621 | "sk Eth Pid Groups " |
2622 | "Rmem Wmem Dump Locks Drops Inode\n"); | |
658cb354 | 2623 | } else { |
1da177e4 LT |
2624 | struct sock *s = v; |
2625 | struct netlink_sock *nlk = nlk_sk(s); | |
2626 | ||
ea9a0379 | 2627 | seq_printf(seq, "%pK %-3d %-10u %08x %-8d %-8d %-5d %-8d %-8u %-8lu\n", |
1da177e4 LT |
2628 | s, |
2629 | s->sk_protocol, | |
15e47304 | 2630 | nlk->portid, |
513c2500 | 2631 | nlk->groups ? (u32)nlk->groups[0] : 0, |
31e6d363 ED |
2632 | sk_rmem_alloc_get(s), |
2633 | sk_wmem_alloc_get(s), | |
16b304f3 | 2634 | nlk->cb_running, |
41c6d650 | 2635 | refcount_read(&s->sk_refcnt), |
cf0aa4e0 MY |
2636 | atomic_read(&s->sk_drops), |
2637 | sock_i_ino(s) | |
1da177e4 LT |
2638 | ); |
2639 | ||
2640 | } | |
2641 | return 0; | |
2642 | } | |
2643 | ||
56b3d975 | 2644 | static const struct seq_operations netlink_seq_ops = { |
1da177e4 LT |
2645 | .start = netlink_seq_start, |
2646 | .next = netlink_seq_next, | |
2647 | .stop = netlink_seq_stop, | |
2648 | .show = netlink_seq_show, | |
2649 | }; | |
1da177e4 LT |
2650 | #endif |
2651 | ||
2652 | int netlink_register_notifier(struct notifier_block *nb) | |
2653 | { | |
efa172f4 | 2654 | return blocking_notifier_chain_register(&netlink_chain, nb); |
1da177e4 | 2655 | } |
6ac552fd | 2656 | EXPORT_SYMBOL(netlink_register_notifier); |
1da177e4 LT |
2657 | |
2658 | int netlink_unregister_notifier(struct notifier_block *nb) | |
2659 | { | |
efa172f4 | 2660 | return blocking_notifier_chain_unregister(&netlink_chain, nb); |
1da177e4 | 2661 | } |
6ac552fd | 2662 | EXPORT_SYMBOL(netlink_unregister_notifier); |
746fac4d | 2663 | |
90ddc4f0 | 2664 | static const struct proto_ops netlink_ops = { |
1da177e4 LT |
2665 | .family = PF_NETLINK, |
2666 | .owner = THIS_MODULE, | |
2667 | .release = netlink_release, | |
2668 | .bind = netlink_bind, | |
2669 | .connect = netlink_connect, | |
2670 | .socketpair = sock_no_socketpair, | |
2671 | .accept = sock_no_accept, | |
2672 | .getname = netlink_getname, | |
a11e1d43 | 2673 | .poll = datagram_poll, |
025c6818 | 2674 | .ioctl = netlink_ioctl, |
1da177e4 LT |
2675 | .listen = sock_no_listen, |
2676 | .shutdown = sock_no_shutdown, | |
9a4595bc PM |
2677 | .setsockopt = netlink_setsockopt, |
2678 | .getsockopt = netlink_getsockopt, | |
1da177e4 LT |
2679 | .sendmsg = netlink_sendmsg, |
2680 | .recvmsg = netlink_recvmsg, | |
d1b4c689 | 2681 | .mmap = sock_no_mmap, |
1da177e4 LT |
2682 | .sendpage = sock_no_sendpage, |
2683 | }; | |
2684 | ||
ec1b4cf7 | 2685 | static const struct net_proto_family netlink_family_ops = { |
1da177e4 LT |
2686 | .family = PF_NETLINK, |
2687 | .create = netlink_create, | |
2688 | .owner = THIS_MODULE, /* for consistency 8) */ | |
2689 | }; | |
2690 | ||
4665079c | 2691 | static int __net_init netlink_net_init(struct net *net) |
b4b51029 EB |
2692 | { |
2693 | #ifdef CONFIG_PROC_FS | |
c3506372 CH |
2694 | if (!proc_create_net("netlink", 0, net->proc_net, &netlink_seq_ops, |
2695 | sizeof(struct nl_seq_iter))) | |
b4b51029 EB |
2696 | return -ENOMEM; |
2697 | #endif | |
2698 | return 0; | |
2699 | } | |
2700 | ||
4665079c | 2701 | static void __net_exit netlink_net_exit(struct net *net) |
b4b51029 EB |
2702 | { |
2703 | #ifdef CONFIG_PROC_FS | |
ece31ffd | 2704 | remove_proc_entry("netlink", net->proc_net); |
b4b51029 EB |
2705 | #endif |
2706 | } | |
2707 | ||
b963ea89 DM |
2708 | static void __init netlink_add_usersock_entry(void) |
2709 | { | |
5c398dc8 | 2710 | struct listeners *listeners; |
b963ea89 DM |
2711 | int groups = 32; |
2712 | ||
5c398dc8 | 2713 | listeners = kzalloc(sizeof(*listeners) + NLGRPSZ(groups), GFP_KERNEL); |
b963ea89 | 2714 | if (!listeners) |
5c398dc8 | 2715 | panic("netlink_add_usersock_entry: Cannot allocate listeners\n"); |
b963ea89 DM |
2716 | |
2717 | netlink_table_grab(); | |
2718 | ||
2719 | nl_table[NETLINK_USERSOCK].groups = groups; | |
5c398dc8 | 2720 | rcu_assign_pointer(nl_table[NETLINK_USERSOCK].listeners, listeners); |
b963ea89 DM |
2721 | nl_table[NETLINK_USERSOCK].module = THIS_MODULE; |
2722 | nl_table[NETLINK_USERSOCK].registered = 1; | |
9785e10a | 2723 | nl_table[NETLINK_USERSOCK].flags = NL_CFG_F_NONROOT_SEND; |
b963ea89 DM |
2724 | |
2725 | netlink_table_ungrab(); | |
2726 | } | |
2727 | ||
022cbae6 | 2728 | static struct pernet_operations __net_initdata netlink_net_ops = { |
b4b51029 EB |
2729 | .init = netlink_net_init, |
2730 | .exit = netlink_net_exit, | |
2731 | }; | |
2732 | ||
49f7b33e | 2733 | static inline u32 netlink_hash(const void *data, u32 len, u32 seed) |
c428ecd1 HX |
2734 | { |
2735 | const struct netlink_sock *nlk = data; | |
2736 | struct netlink_compare_arg arg; | |
2737 | ||
da314c99 | 2738 | netlink_compare_arg_init(&arg, sock_net(&nlk->sk), nlk->portid); |
11b58ba1 | 2739 | return jhash2((u32 *)&arg, netlink_compare_arg_len / sizeof(u32), seed); |
c428ecd1 HX |
2740 | } |
2741 | ||
2742 | static const struct rhashtable_params netlink_rhashtable_params = { | |
2743 | .head_offset = offsetof(struct netlink_sock, node), | |
2744 | .key_len = netlink_compare_arg_len, | |
c428ecd1 HX |
2745 | .obj_hashfn = netlink_hash, |
2746 | .obj_cmpfn = netlink_compare, | |
b5e2c150 | 2747 | .automatic_shrinking = true, |
c428ecd1 HX |
2748 | }; |
2749 | ||
1da177e4 LT |
2750 | static int __init netlink_proto_init(void) |
2751 | { | |
1da177e4 | 2752 | int i; |
1da177e4 LT |
2753 | int err = proto_register(&netlink_proto, 0); |
2754 | ||
2755 | if (err != 0) | |
2756 | goto out; | |
2757 | ||
c593642c | 2758 | BUILD_BUG_ON(sizeof(struct netlink_skb_parms) > sizeof_field(struct sk_buff, cb)); |
1da177e4 | 2759 | |
0da974f4 | 2760 | nl_table = kcalloc(MAX_LINKS, sizeof(*nl_table), GFP_KERNEL); |
fab2caf6 AM |
2761 | if (!nl_table) |
2762 | goto panic; | |
1da177e4 | 2763 | |
1da177e4 | 2764 | for (i = 0; i < MAX_LINKS; i++) { |
c428ecd1 HX |
2765 | if (rhashtable_init(&nl_table[i].hash, |
2766 | &netlink_rhashtable_params) < 0) { | |
e341694e TG |
2767 | while (--i > 0) |
2768 | rhashtable_destroy(&nl_table[i].hash); | |
1da177e4 | 2769 | kfree(nl_table); |
fab2caf6 | 2770 | goto panic; |
1da177e4 | 2771 | } |
1da177e4 LT |
2772 | } |
2773 | ||
b963ea89 DM |
2774 | netlink_add_usersock_entry(); |
2775 | ||
1da177e4 | 2776 | sock_register(&netlink_family_ops); |
b4b51029 | 2777 | register_pernet_subsys(&netlink_net_ops); |
25e3f70f | 2778 | register_pernet_subsys(&netlink_tap_net_ops); |
746fac4d | 2779 | /* The netlink device handler may be needed early. */ |
1da177e4 LT |
2780 | rtnetlink_init(); |
2781 | out: | |
2782 | return err; | |
fab2caf6 AM |
2783 | panic: |
2784 | panic("netlink_init: Cannot allocate nl_table\n"); | |
1da177e4 LT |
2785 | } |
2786 | ||
1da177e4 | 2787 | core_initcall(netlink_proto_init); |