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 | ||
b6459415 | 23 | #include <linux/bpf.h> |
4fc268d2 | 24 | #include <linux/capability.h> |
1da177e4 | 25 | #include <linux/kernel.h> |
b6459415 | 26 | #include <linux/filter.h> |
1da177e4 | 27 | #include <linux/init.h> |
1da177e4 LT |
28 | #include <linux/signal.h> |
29 | #include <linux/sched.h> | |
30 | #include <linux/errno.h> | |
31 | #include <linux/string.h> | |
32 | #include <linux/stat.h> | |
33 | #include <linux/socket.h> | |
34 | #include <linux/un.h> | |
35 | #include <linux/fcntl.h> | |
36 | #include <linux/termios.h> | |
37 | #include <linux/sockios.h> | |
38 | #include <linux/net.h> | |
39 | #include <linux/fs.h> | |
40 | #include <linux/slab.h> | |
7c0f6ba6 | 41 | #include <linux/uaccess.h> |
1da177e4 LT |
42 | #include <linux/skbuff.h> |
43 | #include <linux/netdevice.h> | |
44 | #include <linux/rtnetlink.h> | |
45 | #include <linux/proc_fs.h> | |
46 | #include <linux/seq_file.h> | |
1da177e4 LT |
47 | #include <linux/notifier.h> |
48 | #include <linux/security.h> | |
49 | #include <linux/jhash.h> | |
50 | #include <linux/jiffies.h> | |
51 | #include <linux/random.h> | |
52 | #include <linux/bitops.h> | |
53 | #include <linux/mm.h> | |
54 | #include <linux/types.h> | |
54e0f520 | 55 | #include <linux/audit.h> |
af65bdfc | 56 | #include <linux/mutex.h> |
ccdfcc39 | 57 | #include <linux/vmalloc.h> |
bcbde0d4 | 58 | #include <linux/if_arp.h> |
e341694e | 59 | #include <linux/rhashtable.h> |
9652e931 | 60 | #include <asm/cacheflush.h> |
e341694e | 61 | #include <linux/hash.h> |
ee1c2442 | 62 | #include <linux/genetlink.h> |
7212462f | 63 | #include <linux/net_namespace.h> |
bc5b6c0b | 64 | #include <linux/nospec.h> |
951cf368 | 65 | #include <linux/btf_ids.h> |
54e0f520 | 66 | |
457c4cbc | 67 | #include <net/net_namespace.h> |
25e3f70f | 68 | #include <net/netns/generic.h> |
1da177e4 LT |
69 | #include <net/sock.h> |
70 | #include <net/scm.h> | |
82ace47a | 71 | #include <net/netlink.h> |
7e3ce05e MRL |
72 | #define CREATE_TRACE_POINTS |
73 | #include <trace/events/netlink.h> | |
1da177e4 | 74 | |
0f29c768 | 75 | #include "af_netlink.h" |
1da177e4 | 76 | |
5c398dc8 ED |
77 | struct listeners { |
78 | struct rcu_head rcu; | |
2b738124 | 79 | unsigned long masks[]; |
6c04bb18 JB |
80 | }; |
81 | ||
cd967e05 | 82 | /* state bits */ |
cc3a572f | 83 | #define NETLINK_S_CONGESTED 0x0 |
cd967e05 | 84 | |
035c4c16 | 85 | static inline int netlink_is_kernel(struct sock *sk) |
aed81560 | 86 | { |
cc3a572f | 87 | return nlk_sk(sk)->flags & NETLINK_F_KERNEL_SOCKET; |
aed81560 DL |
88 | } |
89 | ||
91dd93f9 | 90 | struct netlink_table *nl_table __read_mostly; |
0f29c768 | 91 | EXPORT_SYMBOL_GPL(nl_table); |
1da177e4 LT |
92 | |
93 | static DECLARE_WAIT_QUEUE_HEAD(nl_table_wait); | |
94 | ||
8a0f5ccf HX |
95 | static struct lock_class_key nlk_cb_mutex_keys[MAX_LINKS]; |
96 | ||
97 | static const char *const nlk_cb_mutex_key_strings[MAX_LINKS + 1] = { | |
98 | "nlk_cb_mutex-ROUTE", | |
99 | "nlk_cb_mutex-1", | |
100 | "nlk_cb_mutex-USERSOCK", | |
101 | "nlk_cb_mutex-FIREWALL", | |
102 | "nlk_cb_mutex-SOCK_DIAG", | |
103 | "nlk_cb_mutex-NFLOG", | |
104 | "nlk_cb_mutex-XFRM", | |
105 | "nlk_cb_mutex-SELINUX", | |
106 | "nlk_cb_mutex-ISCSI", | |
107 | "nlk_cb_mutex-AUDIT", | |
108 | "nlk_cb_mutex-FIB_LOOKUP", | |
109 | "nlk_cb_mutex-CONNECTOR", | |
110 | "nlk_cb_mutex-NETFILTER", | |
111 | "nlk_cb_mutex-IP6_FW", | |
112 | "nlk_cb_mutex-DNRTMSG", | |
113 | "nlk_cb_mutex-KOBJECT_UEVENT", | |
114 | "nlk_cb_mutex-GENERIC", | |
115 | "nlk_cb_mutex-17", | |
116 | "nlk_cb_mutex-SCSITRANSPORT", | |
117 | "nlk_cb_mutex-ECRYPTFS", | |
118 | "nlk_cb_mutex-RDMA", | |
119 | "nlk_cb_mutex-CRYPTO", | |
120 | "nlk_cb_mutex-SMC", | |
121 | "nlk_cb_mutex-23", | |
122 | "nlk_cb_mutex-24", | |
123 | "nlk_cb_mutex-25", | |
124 | "nlk_cb_mutex-26", | |
125 | "nlk_cb_mutex-27", | |
126 | "nlk_cb_mutex-28", | |
127 | "nlk_cb_mutex-29", | |
128 | "nlk_cb_mutex-30", | |
129 | "nlk_cb_mutex-31", | |
130 | "nlk_cb_mutex-MAX_LINKS" | |
131 | }; | |
132 | ||
1da177e4 | 133 | static int netlink_dump(struct sock *sk); |
1da177e4 | 134 | |
78fd1d0a | 135 | /* nl_table locking explained: |
21e4902a | 136 | * Lookup and traversal are protected with an RCU read-side lock. Insertion |
c5adde94 | 137 | * and removal are protected with per bucket lock while using RCU list |
21e4902a TG |
138 | * modification primitives and may run in parallel to RCU protected lookups. |
139 | * Destruction of the Netlink socket may only occur *after* nl_table_lock has | |
140 | * been acquired * either during or after the socket has been removed from | |
141 | * the list and after an RCU grace period. | |
78fd1d0a | 142 | */ |
0f29c768 AV |
143 | DEFINE_RWLOCK(nl_table_lock); |
144 | EXPORT_SYMBOL_GPL(nl_table_lock); | |
1da177e4 LT |
145 | static atomic_t nl_table_users = ATOMIC_INIT(0); |
146 | ||
6d772ac5 ED |
147 | #define nl_deref_protected(X) rcu_dereference_protected(X, lockdep_is_held(&nl_table_lock)); |
148 | ||
efa172f4 | 149 | static BLOCKING_NOTIFIER_HEAD(netlink_chain); |
1da177e4 | 150 | |
bcbde0d4 | 151 | |
c428ecd1 HX |
152 | static const struct rhashtable_params netlink_rhashtable_params; |
153 | ||
7e3ce05e MRL |
154 | void do_trace_netlink_extack(const char *msg) |
155 | { | |
156 | trace_netlink_extack(msg); | |
157 | } | |
158 | EXPORT_SYMBOL(do_trace_netlink_extack); | |
159 | ||
b57ef81f | 160 | static inline u32 netlink_group_mask(u32 group) |
d629b836 | 161 | { |
0caf6d99 PM |
162 | if (group > 32) |
163 | return 0; | |
d629b836 PM |
164 | return group ? 1 << (group - 1) : 0; |
165 | } | |
166 | ||
1853c949 DB |
167 | static struct sk_buff *netlink_to_full_skb(const struct sk_buff *skb, |
168 | gfp_t gfp_mask) | |
169 | { | |
170 | unsigned int len = skb_end_offset(skb); | |
171 | struct sk_buff *new; | |
172 | ||
173 | new = alloc_skb(len, gfp_mask); | |
174 | if (new == NULL) | |
175 | return NULL; | |
176 | ||
177 | NETLINK_CB(new).portid = NETLINK_CB(skb).portid; | |
178 | NETLINK_CB(new).dst_group = NETLINK_CB(skb).dst_group; | |
179 | NETLINK_CB(new).creds = NETLINK_CB(skb).creds; | |
180 | ||
59ae1d12 | 181 | skb_put_data(new, skb->data, len); |
1853c949 DB |
182 | return new; |
183 | } | |
184 | ||
25e3f70f CW |
185 | static unsigned int netlink_tap_net_id; |
186 | ||
187 | struct netlink_tap_net { | |
188 | struct list_head netlink_tap_all; | |
b1042d35 | 189 | struct mutex netlink_tap_lock; |
25e3f70f CW |
190 | }; |
191 | ||
bcbde0d4 DB |
192 | int netlink_add_tap(struct netlink_tap *nt) |
193 | { | |
25e3f70f CW |
194 | struct net *net = dev_net(nt->dev); |
195 | struct netlink_tap_net *nn = net_generic(net, netlink_tap_net_id); | |
196 | ||
bcbde0d4 DB |
197 | if (unlikely(nt->dev->type != ARPHRD_NETLINK)) |
198 | return -EINVAL; | |
199 | ||
b1042d35 | 200 | mutex_lock(&nn->netlink_tap_lock); |
25e3f70f | 201 | list_add_rcu(&nt->list, &nn->netlink_tap_all); |
b1042d35 | 202 | mutex_unlock(&nn->netlink_tap_lock); |
bcbde0d4 | 203 | |
fcd4d35e | 204 | __module_get(nt->module); |
bcbde0d4 DB |
205 | |
206 | return 0; | |
207 | } | |
208 | EXPORT_SYMBOL_GPL(netlink_add_tap); | |
209 | ||
2173f8d9 | 210 | static int __netlink_remove_tap(struct netlink_tap *nt) |
bcbde0d4 | 211 | { |
25e3f70f CW |
212 | struct net *net = dev_net(nt->dev); |
213 | struct netlink_tap_net *nn = net_generic(net, netlink_tap_net_id); | |
bcbde0d4 DB |
214 | bool found = false; |
215 | struct netlink_tap *tmp; | |
216 | ||
b1042d35 | 217 | mutex_lock(&nn->netlink_tap_lock); |
bcbde0d4 | 218 | |
25e3f70f | 219 | list_for_each_entry(tmp, &nn->netlink_tap_all, list) { |
bcbde0d4 DB |
220 | if (nt == tmp) { |
221 | list_del_rcu(&nt->list); | |
222 | found = true; | |
223 | goto out; | |
224 | } | |
225 | } | |
226 | ||
227 | pr_warn("__netlink_remove_tap: %p not found\n", nt); | |
228 | out: | |
b1042d35 | 229 | mutex_unlock(&nn->netlink_tap_lock); |
bcbde0d4 | 230 | |
92b80eb3 | 231 | if (found) |
bcbde0d4 DB |
232 | module_put(nt->module); |
233 | ||
234 | return found ? 0 : -ENODEV; | |
235 | } | |
bcbde0d4 DB |
236 | |
237 | int netlink_remove_tap(struct netlink_tap *nt) | |
238 | { | |
239 | int ret; | |
240 | ||
241 | ret = __netlink_remove_tap(nt); | |
242 | synchronize_net(); | |
243 | ||
244 | return ret; | |
245 | } | |
246 | EXPORT_SYMBOL_GPL(netlink_remove_tap); | |
247 | ||
25e3f70f CW |
248 | static __net_init int netlink_tap_init_net(struct net *net) |
249 | { | |
250 | struct netlink_tap_net *nn = net_generic(net, netlink_tap_net_id); | |
251 | ||
252 | INIT_LIST_HEAD(&nn->netlink_tap_all); | |
b1042d35 | 253 | mutex_init(&nn->netlink_tap_lock); |
25e3f70f CW |
254 | return 0; |
255 | } | |
256 | ||
25e3f70f CW |
257 | static struct pernet_operations netlink_tap_net_ops = { |
258 | .init = netlink_tap_init_net, | |
25e3f70f CW |
259 | .id = &netlink_tap_net_id, |
260 | .size = sizeof(struct netlink_tap_net), | |
261 | }; | |
262 | ||
5ffd5cdd DB |
263 | static bool netlink_filter_tap(const struct sk_buff *skb) |
264 | { | |
265 | struct sock *sk = skb->sk; | |
5ffd5cdd DB |
266 | |
267 | /* We take the more conservative approach and | |
268 | * whitelist socket protocols that may pass. | |
269 | */ | |
270 | switch (sk->sk_protocol) { | |
271 | case NETLINK_ROUTE: | |
272 | case NETLINK_USERSOCK: | |
273 | case NETLINK_SOCK_DIAG: | |
274 | case NETLINK_NFLOG: | |
275 | case NETLINK_XFRM: | |
276 | case NETLINK_FIB_LOOKUP: | |
277 | case NETLINK_NETFILTER: | |
278 | case NETLINK_GENERIC: | |
498044bb | 279 | return true; |
5ffd5cdd DB |
280 | } |
281 | ||
498044bb | 282 | return false; |
5ffd5cdd DB |
283 | } |
284 | ||
bcbde0d4 DB |
285 | static int __netlink_deliver_tap_skb(struct sk_buff *skb, |
286 | struct net_device *dev) | |
287 | { | |
288 | struct sk_buff *nskb; | |
5ffd5cdd | 289 | struct sock *sk = skb->sk; |
bcbde0d4 DB |
290 | int ret = -ENOMEM; |
291 | ||
93c64764 KC |
292 | if (!net_eq(dev_net(dev), sock_net(sk))) |
293 | return 0; | |
294 | ||
bcbde0d4 | 295 | dev_hold(dev); |
1853c949 | 296 | |
d1b4c689 | 297 | if (is_vmalloc_addr(skb->head)) |
1853c949 DB |
298 | nskb = netlink_to_full_skb(skb, GFP_ATOMIC); |
299 | else | |
300 | nskb = skb_clone(skb, GFP_ATOMIC); | |
bcbde0d4 DB |
301 | if (nskb) { |
302 | nskb->dev = dev; | |
5ffd5cdd | 303 | nskb->protocol = htons((u16) sk->sk_protocol); |
604d13c9 DB |
304 | nskb->pkt_type = netlink_is_kernel(sk) ? |
305 | PACKET_KERNEL : PACKET_USER; | |
4e48ed88 | 306 | skb_reset_network_header(nskb); |
bcbde0d4 DB |
307 | ret = dev_queue_xmit(nskb); |
308 | if (unlikely(ret > 0)) | |
309 | ret = net_xmit_errno(ret); | |
310 | } | |
311 | ||
312 | dev_put(dev); | |
313 | return ret; | |
314 | } | |
315 | ||
25e3f70f | 316 | static void __netlink_deliver_tap(struct sk_buff *skb, struct netlink_tap_net *nn) |
bcbde0d4 DB |
317 | { |
318 | int ret; | |
319 | struct netlink_tap *tmp; | |
320 | ||
5ffd5cdd DB |
321 | if (!netlink_filter_tap(skb)) |
322 | return; | |
323 | ||
25e3f70f | 324 | list_for_each_entry_rcu(tmp, &nn->netlink_tap_all, list) { |
bcbde0d4 DB |
325 | ret = __netlink_deliver_tap_skb(skb, tmp->dev); |
326 | if (unlikely(ret)) | |
327 | break; | |
328 | } | |
329 | } | |
330 | ||
25e3f70f | 331 | static void netlink_deliver_tap(struct net *net, struct sk_buff *skb) |
bcbde0d4 | 332 | { |
25e3f70f CW |
333 | struct netlink_tap_net *nn = net_generic(net, netlink_tap_net_id); |
334 | ||
bcbde0d4 DB |
335 | rcu_read_lock(); |
336 | ||
25e3f70f CW |
337 | if (unlikely(!list_empty(&nn->netlink_tap_all))) |
338 | __netlink_deliver_tap(skb, nn); | |
bcbde0d4 DB |
339 | |
340 | rcu_read_unlock(); | |
341 | } | |
342 | ||
73bfd370 DB |
343 | static void netlink_deliver_tap_kernel(struct sock *dst, struct sock *src, |
344 | struct sk_buff *skb) | |
345 | { | |
346 | if (!(netlink_is_kernel(dst) && netlink_is_kernel(src))) | |
25e3f70f | 347 | netlink_deliver_tap(sock_net(dst), skb); |
73bfd370 DB |
348 | } |
349 | ||
cd1df525 PM |
350 | static void netlink_overrun(struct sock *sk) |
351 | { | |
352 | struct netlink_sock *nlk = nlk_sk(sk); | |
353 | ||
cc3a572f ND |
354 | if (!(nlk->flags & NETLINK_F_RECV_NO_ENOBUFS)) { |
355 | if (!test_and_set_bit(NETLINK_S_CONGESTED, | |
356 | &nlk_sk(sk)->state)) { | |
cd1df525 | 357 | sk->sk_err = ENOBUFS; |
e3ae2365 | 358 | sk_error_report(sk); |
cd1df525 PM |
359 | } |
360 | } | |
361 | atomic_inc(&sk->sk_drops); | |
362 | } | |
363 | ||
364 | static void netlink_rcv_wake(struct sock *sk) | |
365 | { | |
366 | struct netlink_sock *nlk = nlk_sk(sk); | |
367 | ||
174bce38 | 368 | if (skb_queue_empty_lockless(&sk->sk_receive_queue)) |
cc3a572f ND |
369 | clear_bit(NETLINK_S_CONGESTED, &nlk->state); |
370 | if (!test_bit(NETLINK_S_CONGESTED, &nlk->state)) | |
cd1df525 PM |
371 | wake_up_interruptible(&nlk->wait); |
372 | } | |
373 | ||
cf0a018a PM |
374 | static void netlink_skb_destructor(struct sk_buff *skb) |
375 | { | |
c05cdb1b | 376 | if (is_vmalloc_addr(skb->head)) { |
3a36515f PN |
377 | if (!skb->cloned || |
378 | !atomic_dec_return(&(skb_shinfo(skb)->dataref))) | |
379 | vfree(skb->head); | |
380 | ||
c05cdb1b PNA |
381 | skb->head = NULL; |
382 | } | |
9652e931 PM |
383 | if (skb->sk != NULL) |
384 | sock_rfree(skb); | |
cf0a018a PM |
385 | } |
386 | ||
387 | static void netlink_skb_set_owner_r(struct sk_buff *skb, struct sock *sk) | |
388 | { | |
389 | WARN_ON(skb->sk != NULL); | |
390 | skb->sk = sk; | |
391 | skb->destructor = netlink_skb_destructor; | |
392 | atomic_add(skb->truesize, &sk->sk_rmem_alloc); | |
393 | sk_mem_charge(sk, skb->truesize); | |
394 | } | |
395 | ||
ed5d7788 | 396 | static void netlink_sock_destruct(struct sock *sk) |
1da177e4 | 397 | { |
3f660d66 HX |
398 | struct netlink_sock *nlk = nlk_sk(sk); |
399 | ||
16b304f3 | 400 | if (nlk->cb_running) { |
ed5d7788 HX |
401 | if (nlk->cb.done) |
402 | nlk->cb.done(&nlk->cb); | |
16b304f3 PS |
403 | module_put(nlk->cb.module); |
404 | kfree_skb(nlk->cb.skb); | |
3f660d66 HX |
405 | } |
406 | ||
1da177e4 LT |
407 | skb_queue_purge(&sk->sk_receive_queue); |
408 | ||
409 | if (!sock_flag(sk, SOCK_DEAD)) { | |
6ac552fd | 410 | printk(KERN_ERR "Freeing alive netlink socket %p\n", sk); |
1da177e4 LT |
411 | return; |
412 | } | |
547b792c IJ |
413 | |
414 | WARN_ON(atomic_read(&sk->sk_rmem_alloc)); | |
14afee4b | 415 | WARN_ON(refcount_read(&sk->sk_wmem_alloc)); |
547b792c | 416 | WARN_ON(nlk_sk(sk)->groups); |
1da177e4 LT |
417 | } |
418 | ||
707693c8 HX |
419 | static void netlink_sock_destruct_work(struct work_struct *work) |
420 | { | |
421 | struct netlink_sock *nlk = container_of(work, struct netlink_sock, | |
422 | work); | |
423 | ||
ed5d7788 | 424 | sk_free(&nlk->sk); |
707693c8 HX |
425 | } |
426 | ||
6ac552fd PM |
427 | /* This lock without WQ_FLAG_EXCLUSIVE is good on UP and it is _very_ bad on |
428 | * SMP. Look, when several writers sleep and reader wakes them up, all but one | |
1da177e4 LT |
429 | * immediately hit write lock and grab all the cpus. Exclusive sleep solves |
430 | * this, _but_ remember, it adds useless work on UP machines. | |
431 | */ | |
432 | ||
d136f1bd | 433 | void netlink_table_grab(void) |
9a429c49 | 434 | __acquires(nl_table_lock) |
1da177e4 | 435 | { |
d136f1bd JB |
436 | might_sleep(); |
437 | ||
6abd219c | 438 | write_lock_irq(&nl_table_lock); |
1da177e4 LT |
439 | |
440 | if (atomic_read(&nl_table_users)) { | |
441 | DECLARE_WAITQUEUE(wait, current); | |
442 | ||
443 | add_wait_queue_exclusive(&nl_table_wait, &wait); | |
6ac552fd | 444 | for (;;) { |
1da177e4 LT |
445 | set_current_state(TASK_UNINTERRUPTIBLE); |
446 | if (atomic_read(&nl_table_users) == 0) | |
447 | break; | |
6abd219c | 448 | write_unlock_irq(&nl_table_lock); |
1da177e4 | 449 | schedule(); |
6abd219c | 450 | write_lock_irq(&nl_table_lock); |
1da177e4 LT |
451 | } |
452 | ||
453 | __set_current_state(TASK_RUNNING); | |
454 | remove_wait_queue(&nl_table_wait, &wait); | |
455 | } | |
456 | } | |
457 | ||
d136f1bd | 458 | void netlink_table_ungrab(void) |
9a429c49 | 459 | __releases(nl_table_lock) |
1da177e4 | 460 | { |
6abd219c | 461 | write_unlock_irq(&nl_table_lock); |
1da177e4 LT |
462 | wake_up(&nl_table_wait); |
463 | } | |
464 | ||
6ac552fd | 465 | static inline void |
1da177e4 LT |
466 | netlink_lock_table(void) |
467 | { | |
1d482e66 JB |
468 | unsigned long flags; |
469 | ||
1da177e4 LT |
470 | /* read_lock() synchronizes us to netlink_table_grab */ |
471 | ||
1d482e66 | 472 | read_lock_irqsave(&nl_table_lock, flags); |
1da177e4 | 473 | atomic_inc(&nl_table_users); |
1d482e66 | 474 | read_unlock_irqrestore(&nl_table_lock, flags); |
1da177e4 LT |
475 | } |
476 | ||
6ac552fd | 477 | static inline void |
1da177e4 LT |
478 | netlink_unlock_table(void) |
479 | { | |
480 | if (atomic_dec_and_test(&nl_table_users)) | |
481 | wake_up(&nl_table_wait); | |
482 | } | |
483 | ||
e341694e | 484 | struct netlink_compare_arg |
1da177e4 | 485 | { |
c428ecd1 | 486 | possible_net_t pnet; |
e341694e TG |
487 | u32 portid; |
488 | }; | |
1da177e4 | 489 | |
8f2ddaac HX |
490 | /* Doing sizeof directly may yield 4 extra bytes on 64-bit. */ |
491 | #define netlink_compare_arg_len \ | |
492 | (offsetof(struct netlink_compare_arg, portid) + sizeof(u32)) | |
c428ecd1 HX |
493 | |
494 | static inline int netlink_compare(struct rhashtable_compare_arg *arg, | |
495 | const void *ptr) | |
1da177e4 | 496 | { |
c428ecd1 HX |
497 | const struct netlink_compare_arg *x = arg->key; |
498 | const struct netlink_sock *nlk = ptr; | |
1da177e4 | 499 | |
da314c99 | 500 | return nlk->portid != x->portid || |
c428ecd1 HX |
501 | !net_eq(sock_net(&nlk->sk), read_pnet(&x->pnet)); |
502 | } | |
503 | ||
504 | static void netlink_compare_arg_init(struct netlink_compare_arg *arg, | |
505 | struct net *net, u32 portid) | |
506 | { | |
507 | memset(arg, 0, sizeof(*arg)); | |
508 | write_pnet(&arg->pnet, net); | |
509 | arg->portid = portid; | |
1da177e4 LT |
510 | } |
511 | ||
e341694e TG |
512 | static struct sock *__netlink_lookup(struct netlink_table *table, u32 portid, |
513 | struct net *net) | |
1da177e4 | 514 | { |
c428ecd1 | 515 | struct netlink_compare_arg arg; |
1da177e4 | 516 | |
c428ecd1 HX |
517 | netlink_compare_arg_init(&arg, net, portid); |
518 | return rhashtable_lookup_fast(&table->hash, &arg, | |
519 | netlink_rhashtable_params); | |
1da177e4 LT |
520 | } |
521 | ||
c428ecd1 | 522 | static int __netlink_insert(struct netlink_table *table, struct sock *sk) |
c5adde94 | 523 | { |
c428ecd1 | 524 | struct netlink_compare_arg arg; |
c5adde94 | 525 | |
da314c99 | 526 | netlink_compare_arg_init(&arg, sock_net(sk), nlk_sk(sk)->portid); |
c428ecd1 HX |
527 | return rhashtable_lookup_insert_key(&table->hash, &arg, |
528 | &nlk_sk(sk)->node, | |
529 | netlink_rhashtable_params); | |
c5adde94 YX |
530 | } |
531 | ||
e341694e | 532 | static struct sock *netlink_lookup(struct net *net, int protocol, u32 portid) |
1da177e4 | 533 | { |
e341694e TG |
534 | struct netlink_table *table = &nl_table[protocol]; |
535 | struct sock *sk; | |
1da177e4 | 536 | |
e341694e TG |
537 | rcu_read_lock(); |
538 | sk = __netlink_lookup(table, portid, net); | |
539 | if (sk) | |
540 | sock_hold(sk); | |
541 | rcu_read_unlock(); | |
1da177e4 | 542 | |
e341694e | 543 | return sk; |
1da177e4 LT |
544 | } |
545 | ||
90ddc4f0 | 546 | static const struct proto_ops netlink_ops; |
1da177e4 | 547 | |
4277a083 PM |
548 | static void |
549 | netlink_update_listeners(struct sock *sk) | |
550 | { | |
551 | struct netlink_table *tbl = &nl_table[sk->sk_protocol]; | |
4277a083 PM |
552 | unsigned long mask; |
553 | unsigned int i; | |
6d772ac5 ED |
554 | struct listeners *listeners; |
555 | ||
556 | listeners = nl_deref_protected(tbl->listeners); | |
557 | if (!listeners) | |
558 | return; | |
4277a083 | 559 | |
b4ff4f04 | 560 | for (i = 0; i < NLGRPLONGS(tbl->groups); i++) { |
4277a083 | 561 | mask = 0; |
b67bfe0d | 562 | sk_for_each_bound(sk, &tbl->mc_list) { |
b4ff4f04 JB |
563 | if (i < NLGRPLONGS(nlk_sk(sk)->ngroups)) |
564 | mask |= nlk_sk(sk)->groups[i]; | |
565 | } | |
6d772ac5 | 566 | listeners->masks[i] = mask; |
4277a083 PM |
567 | } |
568 | /* this function is only called with the netlink table "grabbed", which | |
569 | * makes sure updates are visible before bind or setsockopt return. */ | |
570 | } | |
571 | ||
8ea65f4a | 572 | static int netlink_insert(struct sock *sk, u32 portid) |
1da177e4 | 573 | { |
da12c90e | 574 | struct netlink_table *table = &nl_table[sk->sk_protocol]; |
919d9db9 | 575 | int err; |
1da177e4 | 576 | |
c5adde94 | 577 | lock_sock(sk); |
1da177e4 | 578 | |
da314c99 HX |
579 | err = nlk_sk(sk)->portid == portid ? 0 : -EBUSY; |
580 | if (nlk_sk(sk)->bound) | |
1da177e4 LT |
581 | goto err; |
582 | ||
c1bb9484 ED |
583 | /* portid can be read locklessly from netlink_getname(). */ |
584 | WRITE_ONCE(nlk_sk(sk)->portid, portid); | |
585 | ||
e341694e | 586 | sock_hold(sk); |
919d9db9 | 587 | |
c428ecd1 HX |
588 | err = __netlink_insert(table, sk); |
589 | if (err) { | |
4e7c1330 DB |
590 | /* In case the hashtable backend returns with -EBUSY |
591 | * from here, it must not escape to the caller. | |
592 | */ | |
593 | if (unlikely(err == -EBUSY)) | |
594 | err = -EOVERFLOW; | |
c428ecd1 HX |
595 | if (err == -EEXIST) |
596 | err = -EADDRINUSE; | |
c5adde94 | 597 | sock_put(sk); |
1f770c0a | 598 | goto err; |
919d9db9 HX |
599 | } |
600 | ||
da314c99 HX |
601 | /* We need to ensure that the socket is hashed and visible. */ |
602 | smp_wmb(); | |
7707a4d0 ED |
603 | /* Paired with lockless reads from netlink_bind(), |
604 | * netlink_connect() and netlink_sendmsg(). | |
605 | */ | |
606 | WRITE_ONCE(nlk_sk(sk)->bound, portid); | |
1f770c0a | 607 | |
1da177e4 | 608 | err: |
c5adde94 | 609 | release_sock(sk); |
1da177e4 LT |
610 | return err; |
611 | } | |
612 | ||
613 | static void netlink_remove(struct sock *sk) | |
614 | { | |
e341694e TG |
615 | struct netlink_table *table; |
616 | ||
e341694e | 617 | table = &nl_table[sk->sk_protocol]; |
c428ecd1 HX |
618 | if (!rhashtable_remove_fast(&table->hash, &nlk_sk(sk)->node, |
619 | netlink_rhashtable_params)) { | |
41c6d650 | 620 | WARN_ON(refcount_read(&sk->sk_refcnt) == 1); |
e341694e TG |
621 | __sock_put(sk); |
622 | } | |
e341694e | 623 | |
1da177e4 | 624 | netlink_table_grab(); |
b10dcb3b | 625 | if (nlk_sk(sk)->subscriptions) { |
1da177e4 | 626 | __sk_del_bind_node(sk); |
b10dcb3b JB |
627 | netlink_update_listeners(sk); |
628 | } | |
ee1c2442 JB |
629 | if (sk->sk_protocol == NETLINK_GENERIC) |
630 | atomic_inc(&genl_sk_destructing_cnt); | |
1da177e4 LT |
631 | netlink_table_ungrab(); |
632 | } | |
633 | ||
634 | static struct proto netlink_proto = { | |
635 | .name = "NETLINK", | |
636 | .owner = THIS_MODULE, | |
637 | .obj_size = sizeof(struct netlink_sock), | |
638 | }; | |
639 | ||
1b8d7ae4 | 640 | static int __netlink_create(struct net *net, struct socket *sock, |
11aa9c28 EB |
641 | struct mutex *cb_mutex, int protocol, |
642 | int kern) | |
1da177e4 LT |
643 | { |
644 | struct sock *sk; | |
645 | struct netlink_sock *nlk; | |
ab33a171 PM |
646 | |
647 | sock->ops = &netlink_ops; | |
648 | ||
11aa9c28 | 649 | sk = sk_alloc(net, PF_NETLINK, GFP_KERNEL, &netlink_proto, kern); |
ab33a171 PM |
650 | if (!sk) |
651 | return -ENOMEM; | |
652 | ||
653 | sock_init_data(sock, sk); | |
654 | ||
655 | nlk = nlk_sk(sk); | |
658cb354 | 656 | if (cb_mutex) { |
ffa4d721 | 657 | nlk->cb_mutex = cb_mutex; |
658cb354 | 658 | } else { |
ffa4d721 PM |
659 | nlk->cb_mutex = &nlk->cb_def_mutex; |
660 | mutex_init(nlk->cb_mutex); | |
8a0f5ccf HX |
661 | lockdep_set_class_and_name(nlk->cb_mutex, |
662 | nlk_cb_mutex_keys + protocol, | |
663 | nlk_cb_mutex_key_strings[protocol]); | |
ffa4d721 | 664 | } |
ab33a171 PM |
665 | init_waitqueue_head(&nlk->wait); |
666 | ||
667 | sk->sk_destruct = netlink_sock_destruct; | |
668 | sk->sk_protocol = protocol; | |
669 | return 0; | |
670 | } | |
671 | ||
3f378b68 EP |
672 | static int netlink_create(struct net *net, struct socket *sock, int protocol, |
673 | int kern) | |
ab33a171 PM |
674 | { |
675 | struct module *module = NULL; | |
af65bdfc | 676 | struct mutex *cb_mutex; |
f7fa9b10 | 677 | struct netlink_sock *nlk; |
023e2cfa JB |
678 | int (*bind)(struct net *net, int group); |
679 | void (*unbind)(struct net *net, int group); | |
ab33a171 | 680 | int err = 0; |
1da177e4 LT |
681 | |
682 | sock->state = SS_UNCONNECTED; | |
683 | ||
684 | if (sock->type != SOCK_RAW && sock->type != SOCK_DGRAM) | |
685 | return -ESOCKTNOSUPPORT; | |
686 | ||
6ac552fd | 687 | if (protocol < 0 || protocol >= MAX_LINKS) |
1da177e4 | 688 | return -EPROTONOSUPPORT; |
bc5b6c0b | 689 | protocol = array_index_nospec(protocol, MAX_LINKS); |
1da177e4 | 690 | |
77247bbb | 691 | netlink_lock_table(); |
95a5afca | 692 | #ifdef CONFIG_MODULES |
ab33a171 | 693 | if (!nl_table[protocol].registered) { |
77247bbb | 694 | netlink_unlock_table(); |
4fdb3bb7 | 695 | request_module("net-pf-%d-proto-%d", PF_NETLINK, protocol); |
77247bbb | 696 | netlink_lock_table(); |
4fdb3bb7 | 697 | } |
ab33a171 PM |
698 | #endif |
699 | if (nl_table[protocol].registered && | |
700 | try_module_get(nl_table[protocol].module)) | |
701 | module = nl_table[protocol].module; | |
974c37e9 AD |
702 | else |
703 | err = -EPROTONOSUPPORT; | |
af65bdfc | 704 | cb_mutex = nl_table[protocol].cb_mutex; |
03292745 | 705 | bind = nl_table[protocol].bind; |
4f520900 | 706 | unbind = nl_table[protocol].unbind; |
77247bbb | 707 | netlink_unlock_table(); |
4fdb3bb7 | 708 | |
974c37e9 AD |
709 | if (err < 0) |
710 | goto out; | |
711 | ||
11aa9c28 | 712 | err = __netlink_create(net, sock, cb_mutex, protocol, kern); |
6ac552fd | 713 | if (err < 0) |
f7fa9b10 PM |
714 | goto out_module; |
715 | ||
c1fd3b94 | 716 | sock_prot_inuse_add(net, &netlink_proto, 1); |
6f756a8c | 717 | |
f7fa9b10 | 718 | nlk = nlk_sk(sock->sk); |
f7fa9b10 | 719 | nlk->module = module; |
03292745 | 720 | nlk->netlink_bind = bind; |
4f520900 | 721 | nlk->netlink_unbind = unbind; |
ab33a171 PM |
722 | out: |
723 | return err; | |
1da177e4 | 724 | |
ab33a171 PM |
725 | out_module: |
726 | module_put(module); | |
727 | goto out; | |
1da177e4 LT |
728 | } |
729 | ||
21e4902a TG |
730 | static void deferred_put_nlk_sk(struct rcu_head *head) |
731 | { | |
732 | struct netlink_sock *nlk = container_of(head, struct netlink_sock, rcu); | |
ed5d7788 HX |
733 | struct sock *sk = &nlk->sk; |
734 | ||
be82485f XL |
735 | kfree(nlk->groups); |
736 | nlk->groups = NULL; | |
737 | ||
41c6d650 | 738 | if (!refcount_dec_and_test(&sk->sk_refcnt)) |
ed5d7788 HX |
739 | return; |
740 | ||
741 | if (nlk->cb_running && nlk->cb.done) { | |
742 | INIT_WORK(&nlk->work, netlink_sock_destruct_work); | |
743 | schedule_work(&nlk->work); | |
744 | return; | |
745 | } | |
21e4902a | 746 | |
ed5d7788 | 747 | sk_free(sk); |
21e4902a TG |
748 | } |
749 | ||
1da177e4 LT |
750 | static int netlink_release(struct socket *sock) |
751 | { | |
752 | struct sock *sk = sock->sk; | |
753 | struct netlink_sock *nlk; | |
754 | ||
755 | if (!sk) | |
756 | return 0; | |
757 | ||
758 | netlink_remove(sk); | |
ac57b3a9 | 759 | sock_orphan(sk); |
1da177e4 LT |
760 | nlk = nlk_sk(sk); |
761 | ||
3f660d66 HX |
762 | /* |
763 | * OK. Socket is unlinked, any packets that arrive now | |
764 | * will be purged. | |
765 | */ | |
1da177e4 | 766 | |
ee1c2442 JB |
767 | /* must not acquire netlink_table_lock in any way again before unbind |
768 | * and notifying genetlink is done as otherwise it might deadlock | |
769 | */ | |
770 | if (nlk->netlink_unbind) { | |
771 | int i; | |
772 | ||
773 | for (i = 0; i < nlk->ngroups; i++) | |
774 | if (test_bit(i, nlk->groups)) | |
775 | nlk->netlink_unbind(sock_net(sk), i + 1); | |
776 | } | |
777 | if (sk->sk_protocol == NETLINK_GENERIC && | |
778 | atomic_dec_return(&genl_sk_destructing_cnt) == 0) | |
779 | wake_up(&genl_sk_destructing_waitq); | |
780 | ||
1da177e4 LT |
781 | sock->sk = NULL; |
782 | wake_up_interruptible_all(&nlk->wait); | |
783 | ||
784 | skb_queue_purge(&sk->sk_write_queue); | |
785 | ||
e2726020 | 786 | if (nlk->portid && nlk->bound) { |
1da177e4 | 787 | struct netlink_notify n = { |
3b1e0a65 | 788 | .net = sock_net(sk), |
1da177e4 | 789 | .protocol = sk->sk_protocol, |
15e47304 | 790 | .portid = nlk->portid, |
1da177e4 | 791 | }; |
efa172f4 | 792 | blocking_notifier_call_chain(&netlink_chain, |
e041c683 | 793 | NETLINK_URELEASE, &n); |
746fac4d | 794 | } |
4fdb3bb7 | 795 | |
5e7c001c | 796 | module_put(nlk->module); |
4fdb3bb7 | 797 | |
aed81560 | 798 | if (netlink_is_kernel(sk)) { |
b10dcb3b | 799 | netlink_table_grab(); |
869e58f8 DL |
800 | BUG_ON(nl_table[sk->sk_protocol].registered == 0); |
801 | if (--nl_table[sk->sk_protocol].registered == 0) { | |
6d772ac5 ED |
802 | struct listeners *old; |
803 | ||
804 | old = nl_deref_protected(nl_table[sk->sk_protocol].listeners); | |
805 | RCU_INIT_POINTER(nl_table[sk->sk_protocol].listeners, NULL); | |
806 | kfree_rcu(old, rcu); | |
869e58f8 | 807 | nl_table[sk->sk_protocol].module = NULL; |
9785e10a | 808 | nl_table[sk->sk_protocol].bind = NULL; |
4f520900 | 809 | nl_table[sk->sk_protocol].unbind = NULL; |
9785e10a | 810 | nl_table[sk->sk_protocol].flags = 0; |
869e58f8 DL |
811 | nl_table[sk->sk_protocol].registered = 0; |
812 | } | |
b10dcb3b | 813 | netlink_table_ungrab(); |
658cb354 | 814 | } |
77247bbb | 815 | |
c1fd3b94 | 816 | sock_prot_inuse_add(sock_net(sk), &netlink_proto, -1); |
0cafd77d ED |
817 | |
818 | /* Because struct net might disappear soon, do not keep a pointer. */ | |
819 | if (!sk->sk_net_refcnt && sock_net(sk) != &init_net) { | |
820 | __netns_tracker_free(sock_net(sk), &sk->ns_tracker, false); | |
821 | /* Because of deferred_put_nlk_sk and use of work queue, | |
822 | * it is possible netns will be freed before this socket. | |
823 | */ | |
824 | sock_net_set(sk, &init_net); | |
825 | __netns_tracker_alloc(&init_net, &sk->ns_tracker, | |
826 | false, GFP_KERNEL); | |
827 | } | |
21e4902a | 828 | call_rcu(&nlk->rcu, deferred_put_nlk_sk); |
1da177e4 LT |
829 | return 0; |
830 | } | |
831 | ||
832 | static int netlink_autobind(struct socket *sock) | |
833 | { | |
834 | struct sock *sk = sock->sk; | |
3b1e0a65 | 835 | struct net *net = sock_net(sk); |
da12c90e | 836 | struct netlink_table *table = &nl_table[sk->sk_protocol]; |
15e47304 | 837 | s32 portid = task_tgid_vnr(current); |
1da177e4 | 838 | int err; |
b9fbe709 HX |
839 | s32 rover = -4096; |
840 | bool ok; | |
1da177e4 LT |
841 | |
842 | retry: | |
843 | cond_resched(); | |
e341694e | 844 | rcu_read_lock(); |
b9fbe709 HX |
845 | ok = !__netlink_lookup(table, portid, net); |
846 | rcu_read_unlock(); | |
847 | if (!ok) { | |
e341694e | 848 | /* Bind collision, search negative portid values. */ |
b9fbe709 HX |
849 | if (rover == -4096) |
850 | /* rover will be in range [S32_MIN, -4097] */ | |
8032bf12 | 851 | rover = S32_MIN + get_random_u32_below(-4096 - S32_MIN); |
b9fbe709 | 852 | else if (rover >= -4096) |
e341694e | 853 | rover = -4097; |
b9fbe709 | 854 | portid = rover--; |
e341694e | 855 | goto retry; |
1da177e4 | 856 | } |
1da177e4 | 857 | |
8ea65f4a | 858 | err = netlink_insert(sk, portid); |
1da177e4 LT |
859 | if (err == -EADDRINUSE) |
860 | goto retry; | |
d470e3b4 DM |
861 | |
862 | /* If 2 threads race to autobind, that is fine. */ | |
863 | if (err == -EBUSY) | |
864 | err = 0; | |
865 | ||
866 | return err; | |
1da177e4 LT |
867 | } |
868 | ||
aa4cf945 EB |
869 | /** |
870 | * __netlink_ns_capable - General netlink message capability test | |
871 | * @nsp: NETLINK_CB of the socket buffer holding a netlink command from userspace. | |
872 | * @user_ns: The user namespace of the capability to use | |
873 | * @cap: The capability to use | |
874 | * | |
875 | * Test to see if the opener of the socket we received the message | |
876 | * from had when the netlink socket was created and the sender of the | |
85405918 | 877 | * message has the capability @cap in the user namespace @user_ns. |
aa4cf945 EB |
878 | */ |
879 | bool __netlink_ns_capable(const struct netlink_skb_parms *nsp, | |
880 | struct user_namespace *user_ns, int cap) | |
881 | { | |
2d7a85f4 EB |
882 | return ((nsp->flags & NETLINK_SKB_DST) || |
883 | file_ns_capable(nsp->sk->sk_socket->file, user_ns, cap)) && | |
884 | ns_capable(user_ns, cap); | |
aa4cf945 EB |
885 | } |
886 | EXPORT_SYMBOL(__netlink_ns_capable); | |
887 | ||
888 | /** | |
889 | * netlink_ns_capable - General netlink message capability test | |
890 | * @skb: socket buffer holding a netlink command from userspace | |
891 | * @user_ns: The user namespace of the capability to use | |
892 | * @cap: The capability to use | |
893 | * | |
894 | * Test to see if the opener of the socket we received the message | |
895 | * from had when the netlink socket was created and the sender of the | |
85405918 | 896 | * message has the capability @cap in the user namespace @user_ns. |
aa4cf945 EB |
897 | */ |
898 | bool netlink_ns_capable(const struct sk_buff *skb, | |
899 | struct user_namespace *user_ns, int cap) | |
900 | { | |
901 | return __netlink_ns_capable(&NETLINK_CB(skb), user_ns, cap); | |
902 | } | |
903 | EXPORT_SYMBOL(netlink_ns_capable); | |
904 | ||
905 | /** | |
906 | * netlink_capable - Netlink global message capability test | |
907 | * @skb: socket buffer holding a netlink command from userspace | |
908 | * @cap: The capability to use | |
909 | * | |
910 | * Test to see if the opener of the socket we received the message | |
911 | * from had when the netlink socket was created and the sender of the | |
85405918 | 912 | * message has the capability @cap in all user namespaces. |
aa4cf945 EB |
913 | */ |
914 | bool netlink_capable(const struct sk_buff *skb, int cap) | |
915 | { | |
916 | return netlink_ns_capable(skb, &init_user_ns, cap); | |
917 | } | |
918 | EXPORT_SYMBOL(netlink_capable); | |
919 | ||
920 | /** | |
921 | * netlink_net_capable - Netlink network namespace message capability test | |
922 | * @skb: socket buffer holding a netlink command from userspace | |
923 | * @cap: The capability to use | |
924 | * | |
925 | * Test to see if the opener of the socket we received the message | |
926 | * from had when the netlink socket was created and the sender of the | |
85405918 | 927 | * message has the capability @cap over the network namespace of |
aa4cf945 EB |
928 | * the socket we received the message from. |
929 | */ | |
930 | bool netlink_net_capable(const struct sk_buff *skb, int cap) | |
931 | { | |
932 | return netlink_ns_capable(skb, sock_net(skb->sk)->user_ns, cap); | |
933 | } | |
934 | EXPORT_SYMBOL(netlink_net_capable); | |
935 | ||
5187cd05 | 936 | static inline int netlink_allowed(const struct socket *sock, unsigned int flag) |
746fac4d | 937 | { |
9785e10a | 938 | return (nl_table[sock->sk->sk_protocol].flags & flag) || |
df008c91 | 939 | ns_capable(sock_net(sock->sk)->user_ns, CAP_NET_ADMIN); |
746fac4d | 940 | } |
1da177e4 | 941 | |
f7fa9b10 PM |
942 | static void |
943 | netlink_update_subscriptions(struct sock *sk, unsigned int subscriptions) | |
944 | { | |
945 | struct netlink_sock *nlk = nlk_sk(sk); | |
946 | ||
947 | if (nlk->subscriptions && !subscriptions) | |
948 | __sk_del_bind_node(sk); | |
949 | else if (!nlk->subscriptions && subscriptions) | |
950 | sk_add_bind_node(sk, &nl_table[sk->sk_protocol].mc_list); | |
951 | nlk->subscriptions = subscriptions; | |
952 | } | |
953 | ||
b4ff4f04 | 954 | static int netlink_realloc_groups(struct sock *sk) |
513c2500 PM |
955 | { |
956 | struct netlink_sock *nlk = nlk_sk(sk); | |
957 | unsigned int groups; | |
b4ff4f04 | 958 | unsigned long *new_groups; |
513c2500 PM |
959 | int err = 0; |
960 | ||
b4ff4f04 JB |
961 | netlink_table_grab(); |
962 | ||
513c2500 | 963 | groups = nl_table[sk->sk_protocol].groups; |
b4ff4f04 | 964 | if (!nl_table[sk->sk_protocol].registered) { |
513c2500 | 965 | err = -ENOENT; |
b4ff4f04 JB |
966 | goto out_unlock; |
967 | } | |
513c2500 | 968 | |
b4ff4f04 JB |
969 | if (nlk->ngroups >= groups) |
970 | goto out_unlock; | |
513c2500 | 971 | |
b4ff4f04 JB |
972 | new_groups = krealloc(nlk->groups, NLGRPSZ(groups), GFP_ATOMIC); |
973 | if (new_groups == NULL) { | |
974 | err = -ENOMEM; | |
975 | goto out_unlock; | |
976 | } | |
6ac552fd | 977 | memset((char *)new_groups + NLGRPSZ(nlk->ngroups), 0, |
b4ff4f04 JB |
978 | NLGRPSZ(groups) - NLGRPSZ(nlk->ngroups)); |
979 | ||
980 | nlk->groups = new_groups; | |
513c2500 | 981 | nlk->ngroups = groups; |
b4ff4f04 JB |
982 | out_unlock: |
983 | netlink_table_ungrab(); | |
984 | return err; | |
513c2500 PM |
985 | } |
986 | ||
02c81ab9 | 987 | static void netlink_undo_bind(int group, long unsigned int groups, |
023e2cfa | 988 | struct sock *sk) |
4f520900 | 989 | { |
023e2cfa | 990 | struct netlink_sock *nlk = nlk_sk(sk); |
4f520900 RGB |
991 | int undo; |
992 | ||
993 | if (!nlk->netlink_unbind) | |
994 | return; | |
995 | ||
996 | for (undo = 0; undo < group; undo++) | |
6251edd9 | 997 | if (test_bit(undo, &groups)) |
8b7c36d8 | 998 | nlk->netlink_unbind(sock_net(sk), undo + 1); |
4f520900 RGB |
999 | } |
1000 | ||
6ac552fd PM |
1001 | static int netlink_bind(struct socket *sock, struct sockaddr *addr, |
1002 | int addr_len) | |
1da177e4 LT |
1003 | { |
1004 | struct sock *sk = sock->sk; | |
3b1e0a65 | 1005 | struct net *net = sock_net(sk); |
1da177e4 LT |
1006 | struct netlink_sock *nlk = nlk_sk(sk); |
1007 | struct sockaddr_nl *nladdr = (struct sockaddr_nl *)addr; | |
f7736080 | 1008 | int err = 0; |
d852be84 | 1009 | unsigned long groups; |
da314c99 | 1010 | bool bound; |
746fac4d | 1011 | |
4e4b5376 HFS |
1012 | if (addr_len < sizeof(struct sockaddr_nl)) |
1013 | return -EINVAL; | |
1014 | ||
1da177e4 LT |
1015 | if (nladdr->nl_family != AF_NETLINK) |
1016 | return -EINVAL; | |
d852be84 | 1017 | groups = nladdr->nl_groups; |
1da177e4 LT |
1018 | |
1019 | /* Only superuser is allowed to listen multicasts */ | |
4f520900 | 1020 | if (groups) { |
5187cd05 | 1021 | if (!netlink_allowed(sock, NL_CFG_F_NONROOT_RECV)) |
513c2500 | 1022 | return -EPERM; |
b4ff4f04 JB |
1023 | err = netlink_realloc_groups(sk); |
1024 | if (err) | |
1025 | return err; | |
513c2500 | 1026 | } |
1da177e4 | 1027 | |
428f944b | 1028 | if (nlk->ngroups < BITS_PER_LONG) |
91874ecf | 1029 | groups &= (1UL << nlk->ngroups) - 1; |
1da177e4 | 1030 | |
7707a4d0 ED |
1031 | /* Paired with WRITE_ONCE() in netlink_insert() */ |
1032 | bound = READ_ONCE(nlk->bound); | |
da314c99 HX |
1033 | if (bound) { |
1034 | /* Ensure nlk->portid is up-to-date. */ | |
1035 | smp_rmb(); | |
1036 | ||
15e47304 | 1037 | if (nladdr->nl_pid != nlk->portid) |
1da177e4 | 1038 | return -EINVAL; |
da314c99 | 1039 | } |
4f520900 RGB |
1040 | |
1041 | if (nlk->netlink_bind && groups) { | |
1042 | int group; | |
1043 | ||
3a20773b NA |
1044 | /* nl_groups is a u32, so cap the maximum groups we can bind */ |
1045 | for (group = 0; group < BITS_PER_TYPE(u32); group++) { | |
4f520900 RGB |
1046 | if (!test_bit(group, &groups)) |
1047 | continue; | |
8b7c36d8 | 1048 | err = nlk->netlink_bind(net, group + 1); |
4f520900 RGB |
1049 | if (!err) |
1050 | continue; | |
023e2cfa | 1051 | netlink_undo_bind(group, groups, sk); |
f2764bd4 | 1052 | return err; |
4f520900 RGB |
1053 | } |
1054 | } | |
1055 | ||
da314c99 HX |
1056 | /* No need for barriers here as we return to user-space without |
1057 | * using any of the bound attributes. | |
1058 | */ | |
f2764bd4 | 1059 | netlink_lock_table(); |
da314c99 | 1060 | if (!bound) { |
1da177e4 | 1061 | err = nladdr->nl_pid ? |
8ea65f4a | 1062 | netlink_insert(sk, nladdr->nl_pid) : |
1da177e4 | 1063 | netlink_autobind(sock); |
4f520900 | 1064 | if (err) { |
3a20773b | 1065 | netlink_undo_bind(BITS_PER_TYPE(u32), groups, sk); |
f7736080 | 1066 | goto unlock; |
4f520900 | 1067 | } |
1da177e4 LT |
1068 | } |
1069 | ||
4f520900 | 1070 | if (!groups && (nlk->groups == NULL || !(u32)nlk->groups[0])) |
f7736080 XL |
1071 | goto unlock; |
1072 | netlink_unlock_table(); | |
1da177e4 LT |
1073 | |
1074 | netlink_table_grab(); | |
f7fa9b10 | 1075 | netlink_update_subscriptions(sk, nlk->subscriptions + |
4f520900 | 1076 | hweight32(groups) - |
746fac4d | 1077 | hweight32(nlk->groups[0])); |
4f520900 | 1078 | nlk->groups[0] = (nlk->groups[0] & ~0xffffffffUL) | groups; |
4277a083 | 1079 | netlink_update_listeners(sk); |
1da177e4 LT |
1080 | netlink_table_ungrab(); |
1081 | ||
1082 | return 0; | |
f7736080 XL |
1083 | |
1084 | unlock: | |
1085 | netlink_unlock_table(); | |
1086 | return err; | |
1da177e4 LT |
1087 | } |
1088 | ||
1089 | static int netlink_connect(struct socket *sock, struct sockaddr *addr, | |
1090 | int alen, int flags) | |
1091 | { | |
1092 | int err = 0; | |
1093 | struct sock *sk = sock->sk; | |
1094 | struct netlink_sock *nlk = nlk_sk(sk); | |
6ac552fd | 1095 | struct sockaddr_nl *nladdr = (struct sockaddr_nl *)addr; |
1da177e4 | 1096 | |
6503d961 CG |
1097 | if (alen < sizeof(addr->sa_family)) |
1098 | return -EINVAL; | |
1099 | ||
1da177e4 | 1100 | if (addr->sa_family == AF_UNSPEC) { |
9b663b5c ED |
1101 | /* paired with READ_ONCE() in netlink_getsockbyportid() */ |
1102 | WRITE_ONCE(sk->sk_state, NETLINK_UNCONNECTED); | |
004db64d ED |
1103 | /* dst_portid and dst_group can be read locklessly */ |
1104 | WRITE_ONCE(nlk->dst_portid, 0); | |
1105 | WRITE_ONCE(nlk->dst_group, 0); | |
1da177e4 LT |
1106 | return 0; |
1107 | } | |
1108 | if (addr->sa_family != AF_NETLINK) | |
1109 | return -EINVAL; | |
1110 | ||
78802879 AP |
1111 | if (alen < sizeof(struct sockaddr_nl)) |
1112 | return -EINVAL; | |
1113 | ||
46833a86 | 1114 | if ((nladdr->nl_groups || nladdr->nl_pid) && |
5187cd05 | 1115 | !netlink_allowed(sock, NL_CFG_F_NONROOT_SEND)) |
1da177e4 LT |
1116 | return -EPERM; |
1117 | ||
da314c99 HX |
1118 | /* No need for barriers here as we return to user-space without |
1119 | * using any of the bound attributes. | |
7707a4d0 | 1120 | * Paired with WRITE_ONCE() in netlink_insert(). |
da314c99 | 1121 | */ |
7707a4d0 | 1122 | if (!READ_ONCE(nlk->bound)) |
1da177e4 LT |
1123 | err = netlink_autobind(sock); |
1124 | ||
1125 | if (err == 0) { | |
9b663b5c ED |
1126 | /* paired with READ_ONCE() in netlink_getsockbyportid() */ |
1127 | WRITE_ONCE(sk->sk_state, NETLINK_CONNECTED); | |
004db64d ED |
1128 | /* dst_portid and dst_group can be read locklessly */ |
1129 | WRITE_ONCE(nlk->dst_portid, nladdr->nl_pid); | |
1130 | WRITE_ONCE(nlk->dst_group, ffs(nladdr->nl_groups)); | |
1da177e4 LT |
1131 | } |
1132 | ||
1133 | return err; | |
1134 | } | |
1135 | ||
6ac552fd | 1136 | static int netlink_getname(struct socket *sock, struct sockaddr *addr, |
9b2c45d4 | 1137 | int peer) |
1da177e4 LT |
1138 | { |
1139 | struct sock *sk = sock->sk; | |
1140 | struct netlink_sock *nlk = nlk_sk(sk); | |
13cfa97b | 1141 | DECLARE_SOCKADDR(struct sockaddr_nl *, nladdr, addr); |
746fac4d | 1142 | |
1da177e4 LT |
1143 | nladdr->nl_family = AF_NETLINK; |
1144 | nladdr->nl_pad = 0; | |
1da177e4 LT |
1145 | |
1146 | if (peer) { | |
004db64d ED |
1147 | /* Paired with WRITE_ONCE() in netlink_connect() */ |
1148 | nladdr->nl_pid = READ_ONCE(nlk->dst_portid); | |
1149 | nladdr->nl_groups = netlink_group_mask(READ_ONCE(nlk->dst_group)); | |
1da177e4 | 1150 | } else { |
c1bb9484 ED |
1151 | /* Paired with WRITE_ONCE() in netlink_insert() */ |
1152 | nladdr->nl_pid = READ_ONCE(nlk->portid); | |
f7736080 | 1153 | netlink_lock_table(); |
513c2500 | 1154 | nladdr->nl_groups = nlk->groups ? nlk->groups[0] : 0; |
f7736080 | 1155 | netlink_unlock_table(); |
1da177e4 | 1156 | } |
9b2c45d4 | 1157 | return sizeof(*nladdr); |
1da177e4 LT |
1158 | } |
1159 | ||
025c6818 DD |
1160 | static int netlink_ioctl(struct socket *sock, unsigned int cmd, |
1161 | unsigned long arg) | |
1162 | { | |
1163 | /* try to hand this ioctl down to the NIC drivers. | |
1164 | */ | |
1165 | return -ENOIOCTLCMD; | |
1166 | } | |
1167 | ||
15e47304 | 1168 | static struct sock *netlink_getsockbyportid(struct sock *ssk, u32 portid) |
1da177e4 | 1169 | { |
1da177e4 LT |
1170 | struct sock *sock; |
1171 | struct netlink_sock *nlk; | |
1172 | ||
15e47304 | 1173 | sock = netlink_lookup(sock_net(ssk), ssk->sk_protocol, portid); |
1da177e4 LT |
1174 | if (!sock) |
1175 | return ERR_PTR(-ECONNREFUSED); | |
1176 | ||
1177 | /* Don't bother queuing skb if kernel socket has no input function */ | |
1178 | nlk = nlk_sk(sock); | |
9b663b5c ED |
1179 | /* dst_portid and sk_state can be changed in netlink_connect() */ |
1180 | if (READ_ONCE(sock->sk_state) == NETLINK_CONNECTED && | |
004db64d | 1181 | READ_ONCE(nlk->dst_portid) != nlk_sk(ssk)->portid) { |
1da177e4 LT |
1182 | sock_put(sock); |
1183 | return ERR_PTR(-ECONNREFUSED); | |
1184 | } | |
1185 | return sock; | |
1186 | } | |
1187 | ||
1188 | struct sock *netlink_getsockbyfilp(struct file *filp) | |
1189 | { | |
496ad9aa | 1190 | struct inode *inode = file_inode(filp); |
1da177e4 LT |
1191 | struct sock *sock; |
1192 | ||
1193 | if (!S_ISSOCK(inode->i_mode)) | |
1194 | return ERR_PTR(-ENOTSOCK); | |
1195 | ||
1196 | sock = SOCKET_I(inode)->sk; | |
1197 | if (sock->sk_family != AF_NETLINK) | |
1198 | return ERR_PTR(-EINVAL); | |
1199 | ||
1200 | sock_hold(sock); | |
1201 | return sock; | |
1202 | } | |
1203 | ||
3a36515f PN |
1204 | static struct sk_buff *netlink_alloc_large_skb(unsigned int size, |
1205 | int broadcast) | |
c05cdb1b PNA |
1206 | { |
1207 | struct sk_buff *skb; | |
1208 | void *data; | |
1209 | ||
3a36515f | 1210 | if (size <= NLMSG_GOODSIZE || broadcast) |
c05cdb1b PNA |
1211 | return alloc_skb(size, GFP_KERNEL); |
1212 | ||
3a36515f PN |
1213 | size = SKB_DATA_ALIGN(size) + |
1214 | SKB_DATA_ALIGN(sizeof(struct skb_shared_info)); | |
c05cdb1b PNA |
1215 | |
1216 | data = vmalloc(size); | |
1217 | if (data == NULL) | |
3a36515f | 1218 | return NULL; |
c05cdb1b | 1219 | |
2ea2f62c | 1220 | skb = __build_skb(data, size); |
3a36515f PN |
1221 | if (skb == NULL) |
1222 | vfree(data); | |
2ea2f62c | 1223 | else |
3a36515f | 1224 | skb->destructor = netlink_skb_destructor; |
c05cdb1b PNA |
1225 | |
1226 | return skb; | |
c05cdb1b PNA |
1227 | } |
1228 | ||
1da177e4 LT |
1229 | /* |
1230 | * Attach a skb to a netlink socket. | |
1231 | * The caller must hold a reference to the destination socket. On error, the | |
1232 | * reference is dropped. The skb is not send to the destination, just all | |
1233 | * all error checks are performed and memory in the queue is reserved. | |
1234 | * Return values: | |
1235 | * < 0: error. skb freed, reference to sock dropped. | |
1236 | * 0: continue | |
1237 | * 1: repeat lookup - reference dropped while waiting for socket memory. | |
1238 | */ | |
9457afee | 1239 | int netlink_attachskb(struct sock *sk, struct sk_buff *skb, |
c3d8d1e3 | 1240 | long *timeo, struct sock *ssk) |
1da177e4 LT |
1241 | { |
1242 | struct netlink_sock *nlk; | |
1243 | ||
1244 | nlk = nlk_sk(sk); | |
1245 | ||
5fd96123 | 1246 | if ((atomic_read(&sk->sk_rmem_alloc) > sk->sk_rcvbuf || |
d1b4c689 | 1247 | test_bit(NETLINK_S_CONGESTED, &nlk->state))) { |
1da177e4 | 1248 | DECLARE_WAITQUEUE(wait, current); |
c3d8d1e3 | 1249 | if (!*timeo) { |
aed81560 | 1250 | if (!ssk || netlink_is_kernel(ssk)) |
1da177e4 LT |
1251 | netlink_overrun(sk); |
1252 | sock_put(sk); | |
1253 | kfree_skb(skb); | |
1254 | return -EAGAIN; | |
1255 | } | |
1256 | ||
1257 | __set_current_state(TASK_INTERRUPTIBLE); | |
1258 | add_wait_queue(&nlk->wait, &wait); | |
1259 | ||
1260 | if ((atomic_read(&sk->sk_rmem_alloc) > sk->sk_rcvbuf || | |
cc3a572f | 1261 | test_bit(NETLINK_S_CONGESTED, &nlk->state)) && |
1da177e4 | 1262 | !sock_flag(sk, SOCK_DEAD)) |
c3d8d1e3 | 1263 | *timeo = schedule_timeout(*timeo); |
1da177e4 LT |
1264 | |
1265 | __set_current_state(TASK_RUNNING); | |
1266 | remove_wait_queue(&nlk->wait, &wait); | |
1267 | sock_put(sk); | |
1268 | ||
1269 | if (signal_pending(current)) { | |
1270 | kfree_skb(skb); | |
c3d8d1e3 | 1271 | return sock_intr_errno(*timeo); |
1da177e4 LT |
1272 | } |
1273 | return 1; | |
1274 | } | |
cf0a018a | 1275 | netlink_skb_set_owner_r(skb, sk); |
1da177e4 LT |
1276 | return 0; |
1277 | } | |
1278 | ||
4a7e7c2a | 1279 | static int __netlink_sendskb(struct sock *sk, struct sk_buff *skb) |
1da177e4 | 1280 | { |
1da177e4 LT |
1281 | int len = skb->len; |
1282 | ||
25e3f70f | 1283 | netlink_deliver_tap(sock_net(sk), skb); |
bcbde0d4 | 1284 | |
d1b4c689 | 1285 | skb_queue_tail(&sk->sk_receive_queue, skb); |
676d2369 | 1286 | sk->sk_data_ready(sk); |
4a7e7c2a ED |
1287 | return len; |
1288 | } | |
1289 | ||
1290 | int netlink_sendskb(struct sock *sk, struct sk_buff *skb) | |
1291 | { | |
1292 | int len = __netlink_sendskb(sk, skb); | |
1293 | ||
1da177e4 LT |
1294 | sock_put(sk); |
1295 | return len; | |
1296 | } | |
1297 | ||
1298 | void netlink_detachskb(struct sock *sk, struct sk_buff *skb) | |
1299 | { | |
1300 | kfree_skb(skb); | |
1301 | sock_put(sk); | |
1302 | } | |
1303 | ||
b57ef81f | 1304 | static struct sk_buff *netlink_trim(struct sk_buff *skb, gfp_t allocation) |
1da177e4 LT |
1305 | { |
1306 | int delta; | |
1307 | ||
1298ca46 | 1308 | WARN_ON(skb->sk != NULL); |
4305b541 | 1309 | delta = skb->end - skb->tail; |
c05cdb1b | 1310 | if (is_vmalloc_addr(skb->head) || delta * 2 < skb->truesize) |
1da177e4 LT |
1311 | return skb; |
1312 | ||
1313 | if (skb_shared(skb)) { | |
1314 | struct sk_buff *nskb = skb_clone(skb, allocation); | |
1315 | if (!nskb) | |
1316 | return skb; | |
8460c00f | 1317 | consume_skb(skb); |
1da177e4 LT |
1318 | skb = nskb; |
1319 | } | |
1320 | ||
158f323b ED |
1321 | pskb_expand_head(skb, 0, -delta, |
1322 | (allocation & ~__GFP_DIRECT_RECLAIM) | | |
1323 | __GFP_NOWARN | __GFP_NORETRY); | |
1da177e4 LT |
1324 | return skb; |
1325 | } | |
1326 | ||
3fbc2905 EB |
1327 | static int netlink_unicast_kernel(struct sock *sk, struct sk_buff *skb, |
1328 | struct sock *ssk) | |
cd40b7d3 DL |
1329 | { |
1330 | int ret; | |
1331 | struct netlink_sock *nlk = nlk_sk(sk); | |
1332 | ||
1333 | ret = -ECONNREFUSED; | |
1334 | if (nlk->netlink_rcv != NULL) { | |
1335 | ret = skb->len; | |
cf0a018a | 1336 | netlink_skb_set_owner_r(skb, sk); |
e32123e5 | 1337 | NETLINK_CB(skb).sk = ssk; |
73bfd370 | 1338 | netlink_deliver_tap_kernel(sk, ssk, skb); |
cd40b7d3 | 1339 | nlk->netlink_rcv(skb); |
bfb253c9 ED |
1340 | consume_skb(skb); |
1341 | } else { | |
1342 | kfree_skb(skb); | |
cd40b7d3 | 1343 | } |
cd40b7d3 DL |
1344 | sock_put(sk); |
1345 | return ret; | |
1346 | } | |
1347 | ||
1348 | int netlink_unicast(struct sock *ssk, struct sk_buff *skb, | |
15e47304 | 1349 | u32 portid, int nonblock) |
1da177e4 LT |
1350 | { |
1351 | struct sock *sk; | |
1352 | int err; | |
1353 | long timeo; | |
1354 | ||
1355 | skb = netlink_trim(skb, gfp_any()); | |
1356 | ||
1357 | timeo = sock_sndtimeo(ssk, nonblock); | |
1358 | retry: | |
15e47304 | 1359 | sk = netlink_getsockbyportid(ssk, portid); |
1da177e4 LT |
1360 | if (IS_ERR(sk)) { |
1361 | kfree_skb(skb); | |
1362 | return PTR_ERR(sk); | |
1363 | } | |
cd40b7d3 | 1364 | if (netlink_is_kernel(sk)) |
3fbc2905 | 1365 | return netlink_unicast_kernel(sk, skb, ssk); |
cd40b7d3 | 1366 | |
b1153f29 | 1367 | if (sk_filter(sk, skb)) { |
84874607 | 1368 | err = skb->len; |
b1153f29 SH |
1369 | kfree_skb(skb); |
1370 | sock_put(sk); | |
1371 | return err; | |
1372 | } | |
1373 | ||
9457afee | 1374 | err = netlink_attachskb(sk, skb, &timeo, ssk); |
1da177e4 LT |
1375 | if (err == 1) |
1376 | goto retry; | |
1377 | if (err) | |
1378 | return err; | |
1379 | ||
7ee015e0 | 1380 | return netlink_sendskb(sk, skb); |
1da177e4 | 1381 | } |
6ac552fd | 1382 | EXPORT_SYMBOL(netlink_unicast); |
1da177e4 | 1383 | |
4277a083 PM |
1384 | int netlink_has_listeners(struct sock *sk, unsigned int group) |
1385 | { | |
1386 | int res = 0; | |
5c398dc8 | 1387 | struct listeners *listeners; |
4277a083 | 1388 | |
aed81560 | 1389 | BUG_ON(!netlink_is_kernel(sk)); |
b4ff4f04 JB |
1390 | |
1391 | rcu_read_lock(); | |
1392 | listeners = rcu_dereference(nl_table[sk->sk_protocol].listeners); | |
1393 | ||
6d772ac5 | 1394 | if (listeners && group - 1 < nl_table[sk->sk_protocol].groups) |
5c398dc8 | 1395 | res = test_bit(group - 1, listeners->masks); |
b4ff4f04 JB |
1396 | |
1397 | rcu_read_unlock(); | |
1398 | ||
4277a083 PM |
1399 | return res; |
1400 | } | |
1401 | EXPORT_SYMBOL_GPL(netlink_has_listeners); | |
1402 | ||
59c28058 JK |
1403 | bool netlink_strict_get_check(struct sk_buff *skb) |
1404 | { | |
1405 | const struct netlink_sock *nlk = nlk_sk(NETLINK_CB(skb).sk); | |
1406 | ||
1407 | return nlk->flags & NETLINK_F_STRICT_CHK; | |
1408 | } | |
1409 | EXPORT_SYMBOL_GPL(netlink_strict_get_check); | |
1410 | ||
b57ef81f | 1411 | static int netlink_broadcast_deliver(struct sock *sk, struct sk_buff *skb) |
1da177e4 LT |
1412 | { |
1413 | struct netlink_sock *nlk = nlk_sk(sk); | |
1414 | ||
1415 | if (atomic_read(&sk->sk_rmem_alloc) <= sk->sk_rcvbuf && | |
cc3a572f | 1416 | !test_bit(NETLINK_S_CONGESTED, &nlk->state)) { |
cf0a018a | 1417 | netlink_skb_set_owner_r(skb, sk); |
4a7e7c2a | 1418 | __netlink_sendskb(sk, skb); |
2c645800 | 1419 | return atomic_read(&sk->sk_rmem_alloc) > (sk->sk_rcvbuf >> 1); |
1da177e4 LT |
1420 | } |
1421 | return -1; | |
1422 | } | |
1423 | ||
1424 | struct netlink_broadcast_data { | |
1425 | struct sock *exclude_sk; | |
b4b51029 | 1426 | struct net *net; |
15e47304 | 1427 | u32 portid; |
1da177e4 LT |
1428 | u32 group; |
1429 | int failure; | |
ff491a73 | 1430 | int delivery_failure; |
1da177e4 LT |
1431 | int congested; |
1432 | int delivered; | |
7d877f3b | 1433 | gfp_t allocation; |
1da177e4 LT |
1434 | struct sk_buff *skb, *skb2; |
1435 | }; | |
1436 | ||
46c9521f RR |
1437 | static void do_one_broadcast(struct sock *sk, |
1438 | struct netlink_broadcast_data *p) | |
1da177e4 LT |
1439 | { |
1440 | struct netlink_sock *nlk = nlk_sk(sk); | |
1441 | int val; | |
1442 | ||
1443 | if (p->exclude_sk == sk) | |
46c9521f | 1444 | return; |
1da177e4 | 1445 | |
15e47304 | 1446 | if (nlk->portid == p->portid || p->group - 1 >= nlk->ngroups || |
f7fa9b10 | 1447 | !test_bit(p->group - 1, nlk->groups)) |
46c9521f | 1448 | return; |
1da177e4 | 1449 | |
59324cf3 ND |
1450 | if (!net_eq(sock_net(sk), p->net)) { |
1451 | if (!(nlk->flags & NETLINK_F_LISTEN_ALL_NSID)) | |
1452 | return; | |
1453 | ||
1454 | if (!peernet_has_id(sock_net(sk), p->net)) | |
1455 | return; | |
1456 | ||
1457 | if (!file_ns_capable(sk->sk_socket->file, p->net->user_ns, | |
1458 | CAP_NET_BROADCAST)) | |
1459 | return; | |
1460 | } | |
b4b51029 | 1461 | |
1da177e4 LT |
1462 | if (p->failure) { |
1463 | netlink_overrun(sk); | |
46c9521f | 1464 | return; |
1da177e4 LT |
1465 | } |
1466 | ||
1467 | sock_hold(sk); | |
1468 | if (p->skb2 == NULL) { | |
68acc024 | 1469 | if (skb_shared(p->skb)) { |
1da177e4 LT |
1470 | p->skb2 = skb_clone(p->skb, p->allocation); |
1471 | } else { | |
68acc024 TC |
1472 | p->skb2 = skb_get(p->skb); |
1473 | /* | |
1474 | * skb ownership may have been set when | |
1475 | * delivered to a previous socket. | |
1476 | */ | |
1477 | skb_orphan(p->skb2); | |
1da177e4 LT |
1478 | } |
1479 | } | |
1480 | if (p->skb2 == NULL) { | |
1481 | netlink_overrun(sk); | |
1482 | /* Clone failed. Notify ALL listeners. */ | |
1483 | p->failure = 1; | |
cc3a572f | 1484 | if (nlk->flags & NETLINK_F_BROADCAST_SEND_ERROR) |
be0c22a4 | 1485 | p->delivery_failure = 1; |
59324cf3 ND |
1486 | goto out; |
1487 | } | |
59324cf3 | 1488 | if (sk_filter(sk, p->skb2)) { |
b1153f29 SH |
1489 | kfree_skb(p->skb2); |
1490 | p->skb2 = NULL; | |
59324cf3 ND |
1491 | goto out; |
1492 | } | |
1493 | NETLINK_CB(p->skb2).nsid = peernet2id(sock_net(sk), p->net); | |
7212462f ND |
1494 | if (NETLINK_CB(p->skb2).nsid != NETNSA_NSID_NOT_ASSIGNED) |
1495 | NETLINK_CB(p->skb2).nsid_is_set = true; | |
59324cf3 ND |
1496 | val = netlink_broadcast_deliver(sk, p->skb2); |
1497 | if (val < 0) { | |
1da177e4 | 1498 | netlink_overrun(sk); |
cc3a572f | 1499 | if (nlk->flags & NETLINK_F_BROADCAST_SEND_ERROR) |
be0c22a4 | 1500 | p->delivery_failure = 1; |
1da177e4 LT |
1501 | } else { |
1502 | p->congested |= val; | |
1503 | p->delivered = 1; | |
1504 | p->skb2 = NULL; | |
1505 | } | |
59324cf3 | 1506 | out: |
1da177e4 | 1507 | sock_put(sk); |
1da177e4 LT |
1508 | } |
1509 | ||
549017aa FW |
1510 | int netlink_broadcast(struct sock *ssk, struct sk_buff *skb, u32 portid, |
1511 | u32 group, gfp_t allocation) | |
1da177e4 | 1512 | { |
3b1e0a65 | 1513 | struct net *net = sock_net(ssk); |
1da177e4 | 1514 | struct netlink_broadcast_data info; |
1da177e4 LT |
1515 | struct sock *sk; |
1516 | ||
1517 | skb = netlink_trim(skb, allocation); | |
1518 | ||
1519 | info.exclude_sk = ssk; | |
b4b51029 | 1520 | info.net = net; |
15e47304 | 1521 | info.portid = portid; |
1da177e4 LT |
1522 | info.group = group; |
1523 | info.failure = 0; | |
ff491a73 | 1524 | info.delivery_failure = 0; |
1da177e4 LT |
1525 | info.congested = 0; |
1526 | info.delivered = 0; | |
1527 | info.allocation = allocation; | |
1528 | info.skb = skb; | |
1529 | info.skb2 = NULL; | |
1530 | ||
1531 | /* While we sleep in clone, do not allow to change socket list */ | |
1532 | ||
1533 | netlink_lock_table(); | |
1534 | ||
b67bfe0d | 1535 | sk_for_each_bound(sk, &nl_table[ssk->sk_protocol].mc_list) |
1da177e4 LT |
1536 | do_one_broadcast(sk, &info); |
1537 | ||
70d4bf6d | 1538 | consume_skb(skb); |
aa1c6a6f | 1539 | |
1da177e4 LT |
1540 | netlink_unlock_table(); |
1541 | ||
70d4bf6d NH |
1542 | if (info.delivery_failure) { |
1543 | kfree_skb(info.skb2); | |
ff491a73 | 1544 | return -ENOBUFS; |
658cb354 ED |
1545 | } |
1546 | consume_skb(info.skb2); | |
ff491a73 | 1547 | |
1da177e4 | 1548 | if (info.delivered) { |
d0164adc | 1549 | if (info.congested && gfpflags_allow_blocking(allocation)) |
1da177e4 LT |
1550 | yield(); |
1551 | return 0; | |
1552 | } | |
1da177e4 LT |
1553 | return -ESRCH; |
1554 | } | |
6ac552fd | 1555 | EXPORT_SYMBOL(netlink_broadcast); |
1da177e4 LT |
1556 | |
1557 | struct netlink_set_err_data { | |
1558 | struct sock *exclude_sk; | |
15e47304 | 1559 | u32 portid; |
1da177e4 LT |
1560 | u32 group; |
1561 | int code; | |
1562 | }; | |
1563 | ||
b57ef81f | 1564 | static int do_one_set_err(struct sock *sk, struct netlink_set_err_data *p) |
1da177e4 LT |
1565 | { |
1566 | struct netlink_sock *nlk = nlk_sk(sk); | |
1a50307b | 1567 | int ret = 0; |
1da177e4 LT |
1568 | |
1569 | if (sk == p->exclude_sk) | |
1570 | goto out; | |
1571 | ||
09ad9bc7 | 1572 | if (!net_eq(sock_net(sk), sock_net(p->exclude_sk))) |
b4b51029 EB |
1573 | goto out; |
1574 | ||
15e47304 | 1575 | if (nlk->portid == p->portid || p->group - 1 >= nlk->ngroups || |
f7fa9b10 | 1576 | !test_bit(p->group - 1, nlk->groups)) |
1da177e4 LT |
1577 | goto out; |
1578 | ||
cc3a572f | 1579 | if (p->code == ENOBUFS && nlk->flags & NETLINK_F_RECV_NO_ENOBUFS) { |
1a50307b PNA |
1580 | ret = 1; |
1581 | goto out; | |
1582 | } | |
1583 | ||
1da177e4 | 1584 | sk->sk_err = p->code; |
e3ae2365 | 1585 | sk_error_report(sk); |
1da177e4 | 1586 | out: |
1a50307b | 1587 | return ret; |
1da177e4 LT |
1588 | } |
1589 | ||
4843b93c PNA |
1590 | /** |
1591 | * netlink_set_err - report error to broadcast listeners | |
1592 | * @ssk: the kernel netlink socket, as returned by netlink_kernel_create() | |
15e47304 | 1593 | * @portid: the PORTID of a process that we want to skip (if any) |
840e93f2 | 1594 | * @group: the broadcast group that will notice the error |
4843b93c | 1595 | * @code: error code, must be negative (as usual in kernelspace) |
1a50307b PNA |
1596 | * |
1597 | * This function returns the number of broadcast listeners that have set the | |
cc3a572f | 1598 | * NETLINK_NO_ENOBUFS socket option. |
4843b93c | 1599 | */ |
15e47304 | 1600 | int netlink_set_err(struct sock *ssk, u32 portid, u32 group, int code) |
1da177e4 LT |
1601 | { |
1602 | struct netlink_set_err_data info; | |
1da177e4 | 1603 | struct sock *sk; |
1a50307b | 1604 | int ret = 0; |
1da177e4 LT |
1605 | |
1606 | info.exclude_sk = ssk; | |
15e47304 | 1607 | info.portid = portid; |
1da177e4 | 1608 | info.group = group; |
4843b93c PNA |
1609 | /* sk->sk_err wants a positive error value */ |
1610 | info.code = -code; | |
1da177e4 LT |
1611 | |
1612 | read_lock(&nl_table_lock); | |
1613 | ||
b67bfe0d | 1614 | sk_for_each_bound(sk, &nl_table[ssk->sk_protocol].mc_list) |
1a50307b | 1615 | ret += do_one_set_err(sk, &info); |
1da177e4 LT |
1616 | |
1617 | read_unlock(&nl_table_lock); | |
1a50307b | 1618 | return ret; |
1da177e4 | 1619 | } |
dd5b6ce6 | 1620 | EXPORT_SYMBOL(netlink_set_err); |
1da177e4 | 1621 | |
84659eb5 JB |
1622 | /* must be called with netlink table grabbed */ |
1623 | static void netlink_update_socket_mc(struct netlink_sock *nlk, | |
1624 | unsigned int group, | |
1625 | int is_new) | |
1626 | { | |
1627 | int old, new = !!is_new, subscriptions; | |
1628 | ||
1629 | old = test_bit(group - 1, nlk->groups); | |
1630 | subscriptions = nlk->subscriptions - old + new; | |
1631 | if (new) | |
1632 | __set_bit(group - 1, nlk->groups); | |
1633 | else | |
1634 | __clear_bit(group - 1, nlk->groups); | |
1635 | netlink_update_subscriptions(&nlk->sk, subscriptions); | |
1636 | netlink_update_listeners(&nlk->sk); | |
1637 | } | |
1638 | ||
9a4595bc | 1639 | static int netlink_setsockopt(struct socket *sock, int level, int optname, |
a7b75c5a | 1640 | sockptr_t optval, unsigned int optlen) |
9a4595bc PM |
1641 | { |
1642 | struct sock *sk = sock->sk; | |
1643 | struct netlink_sock *nlk = nlk_sk(sk); | |
eb496534 JB |
1644 | unsigned int val = 0; |
1645 | int err; | |
9a4595bc PM |
1646 | |
1647 | if (level != SOL_NETLINK) | |
1648 | return -ENOPROTOOPT; | |
1649 | ||
d1b4c689 | 1650 | if (optlen >= sizeof(int) && |
a7b75c5a | 1651 | copy_from_sockptr(&val, optval, sizeof(val))) |
9a4595bc PM |
1652 | return -EFAULT; |
1653 | ||
1654 | switch (optname) { | |
1655 | case NETLINK_PKTINFO: | |
1656 | if (val) | |
cc3a572f | 1657 | nlk->flags |= NETLINK_F_RECV_PKTINFO; |
9a4595bc | 1658 | else |
cc3a572f | 1659 | nlk->flags &= ~NETLINK_F_RECV_PKTINFO; |
9a4595bc PM |
1660 | err = 0; |
1661 | break; | |
1662 | case NETLINK_ADD_MEMBERSHIP: | |
1663 | case NETLINK_DROP_MEMBERSHIP: { | |
5187cd05 | 1664 | if (!netlink_allowed(sock, NL_CFG_F_NONROOT_RECV)) |
9a4595bc | 1665 | return -EPERM; |
b4ff4f04 JB |
1666 | err = netlink_realloc_groups(sk); |
1667 | if (err) | |
1668 | return err; | |
9a4595bc PM |
1669 | if (!val || val - 1 >= nlk->ngroups) |
1670 | return -EINVAL; | |
7774d5e0 | 1671 | if (optname == NETLINK_ADD_MEMBERSHIP && nlk->netlink_bind) { |
023e2cfa | 1672 | err = nlk->netlink_bind(sock_net(sk), val); |
4f520900 RGB |
1673 | if (err) |
1674 | return err; | |
1675 | } | |
9a4595bc | 1676 | netlink_table_grab(); |
84659eb5 JB |
1677 | netlink_update_socket_mc(nlk, val, |
1678 | optname == NETLINK_ADD_MEMBERSHIP); | |
9a4595bc | 1679 | netlink_table_ungrab(); |
7774d5e0 | 1680 | if (optname == NETLINK_DROP_MEMBERSHIP && nlk->netlink_unbind) |
023e2cfa | 1681 | nlk->netlink_unbind(sock_net(sk), val); |
03292745 | 1682 | |
9a4595bc PM |
1683 | err = 0; |
1684 | break; | |
1685 | } | |
be0c22a4 PNA |
1686 | case NETLINK_BROADCAST_ERROR: |
1687 | if (val) | |
cc3a572f | 1688 | nlk->flags |= NETLINK_F_BROADCAST_SEND_ERROR; |
be0c22a4 | 1689 | else |
cc3a572f | 1690 | nlk->flags &= ~NETLINK_F_BROADCAST_SEND_ERROR; |
be0c22a4 PNA |
1691 | err = 0; |
1692 | break; | |
38938bfe PNA |
1693 | case NETLINK_NO_ENOBUFS: |
1694 | if (val) { | |
cc3a572f ND |
1695 | nlk->flags |= NETLINK_F_RECV_NO_ENOBUFS; |
1696 | clear_bit(NETLINK_S_CONGESTED, &nlk->state); | |
38938bfe | 1697 | wake_up_interruptible(&nlk->wait); |
658cb354 | 1698 | } else { |
cc3a572f | 1699 | nlk->flags &= ~NETLINK_F_RECV_NO_ENOBUFS; |
658cb354 | 1700 | } |
38938bfe PNA |
1701 | err = 0; |
1702 | break; | |
59324cf3 ND |
1703 | case NETLINK_LISTEN_ALL_NSID: |
1704 | if (!ns_capable(sock_net(sk)->user_ns, CAP_NET_BROADCAST)) | |
1705 | return -EPERM; | |
1706 | ||
1707 | if (val) | |
1708 | nlk->flags |= NETLINK_F_LISTEN_ALL_NSID; | |
1709 | else | |
1710 | nlk->flags &= ~NETLINK_F_LISTEN_ALL_NSID; | |
1711 | err = 0; | |
1712 | break; | |
0a6a3a23 CR |
1713 | case NETLINK_CAP_ACK: |
1714 | if (val) | |
1715 | nlk->flags |= NETLINK_F_CAP_ACK; | |
1716 | else | |
1717 | nlk->flags &= ~NETLINK_F_CAP_ACK; | |
1718 | err = 0; | |
1719 | break; | |
2d4bc933 JB |
1720 | case NETLINK_EXT_ACK: |
1721 | if (val) | |
1722 | nlk->flags |= NETLINK_F_EXT_ACK; | |
1723 | else | |
1724 | nlk->flags &= ~NETLINK_F_EXT_ACK; | |
1725 | err = 0; | |
1726 | break; | |
d3e8869e | 1727 | case NETLINK_GET_STRICT_CHK: |
89d35528 DA |
1728 | if (val) |
1729 | nlk->flags |= NETLINK_F_STRICT_CHK; | |
1730 | else | |
1731 | nlk->flags &= ~NETLINK_F_STRICT_CHK; | |
1732 | err = 0; | |
1733 | break; | |
9a4595bc PM |
1734 | default: |
1735 | err = -ENOPROTOOPT; | |
1736 | } | |
1737 | return err; | |
1738 | } | |
1739 | ||
1740 | static int netlink_getsockopt(struct socket *sock, int level, int optname, | |
746fac4d | 1741 | char __user *optval, int __user *optlen) |
9a4595bc PM |
1742 | { |
1743 | struct sock *sk = sock->sk; | |
1744 | struct netlink_sock *nlk = nlk_sk(sk); | |
d913d32c KI |
1745 | unsigned int flag; |
1746 | int len, val; | |
9a4595bc PM |
1747 | |
1748 | if (level != SOL_NETLINK) | |
1749 | return -ENOPROTOOPT; | |
1750 | ||
1751 | if (get_user(len, optlen)) | |
1752 | return -EFAULT; | |
1753 | if (len < 0) | |
1754 | return -EINVAL; | |
1755 | ||
1756 | switch (optname) { | |
1757 | case NETLINK_PKTINFO: | |
d913d32c | 1758 | flag = NETLINK_F_RECV_PKTINFO; |
9a4595bc | 1759 | break; |
be0c22a4 | 1760 | case NETLINK_BROADCAST_ERROR: |
d913d32c | 1761 | flag = NETLINK_F_BROADCAST_SEND_ERROR; |
be0c22a4 | 1762 | break; |
38938bfe | 1763 | case NETLINK_NO_ENOBUFS: |
d913d32c | 1764 | flag = NETLINK_F_RECV_NO_ENOBUFS; |
38938bfe | 1765 | break; |
b42be38b | 1766 | case NETLINK_LIST_MEMBERSHIPS: { |
d913d32c | 1767 | int pos, idx, shift, err = 0; |
b42be38b | 1768 | |
47191d65 | 1769 | netlink_lock_table(); |
b42be38b DH |
1770 | for (pos = 0; pos * 8 < nlk->ngroups; pos += sizeof(u32)) { |
1771 | if (len - pos < sizeof(u32)) | |
1772 | break; | |
1773 | ||
1774 | idx = pos / sizeof(unsigned long); | |
1775 | shift = (pos % sizeof(unsigned long)) * 8; | |
1776 | if (put_user((u32)(nlk->groups[idx] >> shift), | |
1777 | (u32 __user *)(optval + pos))) { | |
1778 | err = -EFAULT; | |
1779 | break; | |
1780 | } | |
1781 | } | |
1782 | if (put_user(ALIGN(nlk->ngroups / 8, sizeof(u32)), optlen)) | |
1783 | err = -EFAULT; | |
47191d65 | 1784 | netlink_unlock_table(); |
d913d32c | 1785 | return err; |
b42be38b | 1786 | } |
0a6a3a23 | 1787 | case NETLINK_CAP_ACK: |
d913d32c | 1788 | flag = NETLINK_F_CAP_ACK; |
0a6a3a23 | 1789 | break; |
2d4bc933 | 1790 | case NETLINK_EXT_ACK: |
d913d32c | 1791 | flag = NETLINK_F_EXT_ACK; |
2d4bc933 | 1792 | break; |
d3e8869e | 1793 | case NETLINK_GET_STRICT_CHK: |
d913d32c | 1794 | flag = NETLINK_F_STRICT_CHK; |
89d35528 | 1795 | break; |
9a4595bc | 1796 | default: |
d913d32c | 1797 | return -ENOPROTOOPT; |
9a4595bc | 1798 | } |
d913d32c KI |
1799 | |
1800 | if (len < sizeof(int)) | |
1801 | return -EINVAL; | |
1802 | ||
1803 | len = sizeof(int); | |
1804 | val = nlk->flags & flag ? 1 : 0; | |
1805 | ||
1806 | if (put_user(len, optlen) || | |
1807 | copy_to_user(optval, &val, len)) | |
1808 | return -EFAULT; | |
1809 | ||
1810 | return 0; | |
9a4595bc PM |
1811 | } |
1812 | ||
1813 | static void netlink_cmsg_recv_pktinfo(struct msghdr *msg, struct sk_buff *skb) | |
1814 | { | |
1815 | struct nl_pktinfo info; | |
1816 | ||
1817 | info.group = NETLINK_CB(skb).dst_group; | |
1818 | put_cmsg(msg, SOL_NETLINK, NETLINK_PKTINFO, sizeof(info), &info); | |
1819 | } | |
1820 | ||
59324cf3 ND |
1821 | static void netlink_cmsg_listen_all_nsid(struct sock *sk, struct msghdr *msg, |
1822 | struct sk_buff *skb) | |
1823 | { | |
1824 | if (!NETLINK_CB(skb).nsid_is_set) | |
1825 | return; | |
1826 | ||
1827 | put_cmsg(msg, SOL_NETLINK, NETLINK_LISTEN_ALL_NSID, sizeof(int), | |
1828 | &NETLINK_CB(skb).nsid); | |
1829 | } | |
1830 | ||
1b784140 | 1831 | static int netlink_sendmsg(struct socket *sock, struct msghdr *msg, size_t len) |
1da177e4 | 1832 | { |
1da177e4 LT |
1833 | struct sock *sk = sock->sk; |
1834 | struct netlink_sock *nlk = nlk_sk(sk); | |
342dfc30 | 1835 | DECLARE_SOCKADDR(struct sockaddr_nl *, addr, msg->msg_name); |
15e47304 | 1836 | u32 dst_portid; |
d629b836 | 1837 | u32 dst_group; |
1da177e4 LT |
1838 | struct sk_buff *skb; |
1839 | int err; | |
1840 | struct scm_cookie scm; | |
2d7a85f4 | 1841 | u32 netlink_skb_flags = 0; |
1da177e4 | 1842 | |
4d11af5d | 1843 | if (msg->msg_flags & MSG_OOB) |
1da177e4 LT |
1844 | return -EOPNOTSUPP; |
1845 | ||
f123cffd HM |
1846 | if (len == 0) { |
1847 | pr_warn_once("Zero length message leads to an empty skb\n"); | |
1848 | return -ENODATA; | |
1849 | } | |
1850 | ||
7cc05662 | 1851 | err = scm_send(sock, msg, &scm, true); |
1da177e4 LT |
1852 | if (err < 0) |
1853 | return err; | |
1854 | ||
1855 | if (msg->msg_namelen) { | |
b47030c7 | 1856 | err = -EINVAL; |
6091f09c ED |
1857 | if (msg->msg_namelen < sizeof(struct sockaddr_nl)) |
1858 | goto out; | |
1da177e4 | 1859 | if (addr->nl_family != AF_NETLINK) |
b47030c7 | 1860 | goto out; |
15e47304 | 1861 | dst_portid = addr->nl_pid; |
d629b836 | 1862 | dst_group = ffs(addr->nl_groups); |
b47030c7 | 1863 | err = -EPERM; |
15e47304 | 1864 | if ((dst_group || dst_portid) && |
5187cd05 | 1865 | !netlink_allowed(sock, NL_CFG_F_NONROOT_SEND)) |
b47030c7 | 1866 | goto out; |
2d7a85f4 | 1867 | netlink_skb_flags |= NETLINK_SKB_DST; |
1da177e4 | 1868 | } else { |
004db64d ED |
1869 | /* Paired with WRITE_ONCE() in netlink_connect() */ |
1870 | dst_portid = READ_ONCE(nlk->dst_portid); | |
1871 | dst_group = READ_ONCE(nlk->dst_group); | |
1da177e4 LT |
1872 | } |
1873 | ||
7707a4d0 ED |
1874 | /* Paired with WRITE_ONCE() in netlink_insert() */ |
1875 | if (!READ_ONCE(nlk->bound)) { | |
1da177e4 LT |
1876 | err = netlink_autobind(sock); |
1877 | if (err) | |
1878 | goto out; | |
da314c99 HX |
1879 | } else { |
1880 | /* Ensure nlk is hashed and visible. */ | |
1881 | smp_rmb(); | |
1da177e4 LT |
1882 | } |
1883 | ||
1884 | err = -EMSGSIZE; | |
1885 | if (len > sk->sk_sndbuf - 32) | |
1886 | goto out; | |
1887 | err = -ENOBUFS; | |
3a36515f | 1888 | skb = netlink_alloc_large_skb(len, dst_group); |
6ac552fd | 1889 | if (skb == NULL) |
1da177e4 LT |
1890 | goto out; |
1891 | ||
15e47304 | 1892 | NETLINK_CB(skb).portid = nlk->portid; |
d629b836 | 1893 | NETLINK_CB(skb).dst_group = dst_group; |
7cc05662 | 1894 | NETLINK_CB(skb).creds = scm.creds; |
2d7a85f4 | 1895 | NETLINK_CB(skb).flags = netlink_skb_flags; |
1da177e4 | 1896 | |
1da177e4 | 1897 | err = -EFAULT; |
6ce8e9ce | 1898 | if (memcpy_from_msg(skb_put(skb, len), msg, len)) { |
1da177e4 LT |
1899 | kfree_skb(skb); |
1900 | goto out; | |
1901 | } | |
1902 | ||
1903 | err = security_netlink_send(sk, skb); | |
1904 | if (err) { | |
1905 | kfree_skb(skb); | |
1906 | goto out; | |
1907 | } | |
1908 | ||
d629b836 | 1909 | if (dst_group) { |
63354797 | 1910 | refcount_inc(&skb->users); |
15e47304 | 1911 | netlink_broadcast(sk, skb, dst_portid, dst_group, GFP_KERNEL); |
1da177e4 | 1912 | } |
4d11af5d | 1913 | err = netlink_unicast(sk, skb, dst_portid, msg->msg_flags & MSG_DONTWAIT); |
1da177e4 LT |
1914 | |
1915 | out: | |
7cc05662 | 1916 | scm_destroy(&scm); |
1da177e4 LT |
1917 | return err; |
1918 | } | |
1919 | ||
1b784140 | 1920 | static int netlink_recvmsg(struct socket *sock, struct msghdr *msg, size_t len, |
1da177e4 LT |
1921 | int flags) |
1922 | { | |
1da177e4 LT |
1923 | struct scm_cookie scm; |
1924 | struct sock *sk = sock->sk; | |
1925 | struct netlink_sock *nlk = nlk_sk(sk); | |
a1865f2e | 1926 | size_t copied, max_recvmsg_len; |
68d6ac6d | 1927 | struct sk_buff *skb, *data_skb; |
b44d211e | 1928 | int err, ret; |
1da177e4 | 1929 | |
4d11af5d | 1930 | if (flags & MSG_OOB) |
1da177e4 LT |
1931 | return -EOPNOTSUPP; |
1932 | ||
1933 | copied = 0; | |
1934 | ||
f4b41f06 | 1935 | skb = skb_recv_datagram(sk, flags, &err); |
6ac552fd | 1936 | if (skb == NULL) |
1da177e4 LT |
1937 | goto out; |
1938 | ||
68d6ac6d JB |
1939 | data_skb = skb; |
1940 | ||
1dacc76d JB |
1941 | #ifdef CONFIG_COMPAT_NETLINK_MESSAGES |
1942 | if (unlikely(skb_shinfo(skb)->frag_list)) { | |
1dacc76d | 1943 | /* |
68d6ac6d JB |
1944 | * If this skb has a frag_list, then here that means that we |
1945 | * will have to use the frag_list skb's data for compat tasks | |
1946 | * and the regular skb's data for normal (non-compat) tasks. | |
1dacc76d | 1947 | * |
68d6ac6d JB |
1948 | * If we need to send the compat skb, assign it to the |
1949 | * 'data_skb' variable so that it will be used below for data | |
1950 | * copying. We keep 'skb' for everything else, including | |
1951 | * freeing both later. | |
1dacc76d | 1952 | */ |
68d6ac6d JB |
1953 | if (flags & MSG_CMSG_COMPAT) |
1954 | data_skb = skb_shinfo(skb)->frag_list; | |
1dacc76d JB |
1955 | } |
1956 | #endif | |
1957 | ||
9063e21f | 1958 | /* Record the max length of recvmsg() calls for future allocations */ |
a1865f2e ED |
1959 | max_recvmsg_len = max(READ_ONCE(nlk->max_recvmsg_len), len); |
1960 | max_recvmsg_len = min_t(size_t, max_recvmsg_len, | |
1961 | SKB_WITH_OVERHEAD(32768)); | |
1962 | WRITE_ONCE(nlk->max_recvmsg_len, max_recvmsg_len); | |
9063e21f | 1963 | |
68d6ac6d | 1964 | copied = data_skb->len; |
1da177e4 LT |
1965 | if (len < copied) { |
1966 | msg->msg_flags |= MSG_TRUNC; | |
1967 | copied = len; | |
1968 | } | |
1969 | ||
51f3d02b | 1970 | err = skb_copy_datagram_msg(data_skb, 0, msg, copied); |
1da177e4 LT |
1971 | |
1972 | if (msg->msg_name) { | |
342dfc30 | 1973 | DECLARE_SOCKADDR(struct sockaddr_nl *, addr, msg->msg_name); |
1da177e4 LT |
1974 | addr->nl_family = AF_NETLINK; |
1975 | addr->nl_pad = 0; | |
15e47304 | 1976 | addr->nl_pid = NETLINK_CB(skb).portid; |
d629b836 | 1977 | addr->nl_groups = netlink_group_mask(NETLINK_CB(skb).dst_group); |
1da177e4 LT |
1978 | msg->msg_namelen = sizeof(*addr); |
1979 | } | |
1980 | ||
cc3a572f | 1981 | if (nlk->flags & NETLINK_F_RECV_PKTINFO) |
cc9a06cd | 1982 | netlink_cmsg_recv_pktinfo(msg, skb); |
59324cf3 ND |
1983 | if (nlk->flags & NETLINK_F_LISTEN_ALL_NSID) |
1984 | netlink_cmsg_listen_all_nsid(sk, msg, skb); | |
cc9a06cd | 1985 | |
7cc05662 CH |
1986 | memset(&scm, 0, sizeof(scm)); |
1987 | scm.creds = *NETLINK_CREDS(skb); | |
188ccb55 | 1988 | if (flags & MSG_TRUNC) |
68d6ac6d | 1989 | copied = data_skb->len; |
daa3766e | 1990 | |
1da177e4 LT |
1991 | skb_free_datagram(sk, skb); |
1992 | ||
16b304f3 PS |
1993 | if (nlk->cb_running && |
1994 | atomic_read(&sk->sk_rmem_alloc) <= sk->sk_rcvbuf / 2) { | |
b44d211e AV |
1995 | ret = netlink_dump(sk); |
1996 | if (ret) { | |
ac30ef83 | 1997 | sk->sk_err = -ret; |
e3ae2365 | 1998 | sk_error_report(sk); |
b44d211e AV |
1999 | } |
2000 | } | |
1da177e4 | 2001 | |
7cc05662 | 2002 | scm_recv(sock, msg, &scm, flags); |
1da177e4 LT |
2003 | out: |
2004 | netlink_rcv_wake(sk); | |
2005 | return err ? : copied; | |
2006 | } | |
2007 | ||
676d2369 | 2008 | static void netlink_data_ready(struct sock *sk) |
1da177e4 | 2009 | { |
cd40b7d3 | 2010 | BUG(); |
1da177e4 LT |
2011 | } |
2012 | ||
2013 | /* | |
746fac4d | 2014 | * We export these functions to other modules. They provide a |
1da177e4 LT |
2015 | * complete set of kernel non-blocking support for message |
2016 | * queueing. | |
2017 | */ | |
2018 | ||
2019 | struct sock * | |
9f00d977 PNA |
2020 | __netlink_kernel_create(struct net *net, int unit, struct module *module, |
2021 | struct netlink_kernel_cfg *cfg) | |
1da177e4 LT |
2022 | { |
2023 | struct socket *sock; | |
2024 | struct sock *sk; | |
77247bbb | 2025 | struct netlink_sock *nlk; |
5c398dc8 | 2026 | struct listeners *listeners = NULL; |
a31f2d17 PNA |
2027 | struct mutex *cb_mutex = cfg ? cfg->cb_mutex : NULL; |
2028 | unsigned int groups; | |
1da177e4 | 2029 | |
fab2caf6 | 2030 | BUG_ON(!nl_table); |
1da177e4 | 2031 | |
6ac552fd | 2032 | if (unit < 0 || unit >= MAX_LINKS) |
1da177e4 LT |
2033 | return NULL; |
2034 | ||
2035 | if (sock_create_lite(PF_NETLINK, SOCK_DGRAM, unit, &sock)) | |
2036 | return NULL; | |
13d3078e EB |
2037 | |
2038 | if (__netlink_create(net, sock, cb_mutex, unit, 1) < 0) | |
23fe1866 PE |
2039 | goto out_sock_release_nosk; |
2040 | ||
2041 | sk = sock->sk; | |
4fdb3bb7 | 2042 | |
a31f2d17 | 2043 | if (!cfg || cfg->groups < 32) |
4277a083 | 2044 | groups = 32; |
a31f2d17 PNA |
2045 | else |
2046 | groups = cfg->groups; | |
4277a083 | 2047 | |
5c398dc8 | 2048 | listeners = kzalloc(sizeof(*listeners) + NLGRPSZ(groups), GFP_KERNEL); |
4277a083 PM |
2049 | if (!listeners) |
2050 | goto out_sock_release; | |
2051 | ||
1da177e4 | 2052 | sk->sk_data_ready = netlink_data_ready; |
a31f2d17 PNA |
2053 | if (cfg && cfg->input) |
2054 | nlk_sk(sk)->netlink_rcv = cfg->input; | |
1da177e4 | 2055 | |
8ea65f4a | 2056 | if (netlink_insert(sk, 0)) |
77247bbb | 2057 | goto out_sock_release; |
4fdb3bb7 | 2058 | |
77247bbb | 2059 | nlk = nlk_sk(sk); |
cc3a572f | 2060 | nlk->flags |= NETLINK_F_KERNEL_SOCKET; |
4fdb3bb7 | 2061 | |
4fdb3bb7 | 2062 | netlink_table_grab(); |
b4b51029 EB |
2063 | if (!nl_table[unit].registered) { |
2064 | nl_table[unit].groups = groups; | |
5c398dc8 | 2065 | rcu_assign_pointer(nl_table[unit].listeners, listeners); |
b4b51029 EB |
2066 | nl_table[unit].cb_mutex = cb_mutex; |
2067 | nl_table[unit].module = module; | |
9785e10a PNA |
2068 | if (cfg) { |
2069 | nl_table[unit].bind = cfg->bind; | |
6251edd9 | 2070 | nl_table[unit].unbind = cfg->unbind; |
9785e10a PNA |
2071 | nl_table[unit].flags = cfg->flags; |
2072 | } | |
b4b51029 | 2073 | nl_table[unit].registered = 1; |
f937f1f4 JJ |
2074 | } else { |
2075 | kfree(listeners); | |
869e58f8 | 2076 | nl_table[unit].registered++; |
b4b51029 | 2077 | } |
4fdb3bb7 | 2078 | netlink_table_ungrab(); |
77247bbb PM |
2079 | return sk; |
2080 | ||
4fdb3bb7 | 2081 | out_sock_release: |
4277a083 | 2082 | kfree(listeners); |
9dfbec1f | 2083 | netlink_kernel_release(sk); |
23fe1866 PE |
2084 | return NULL; |
2085 | ||
2086 | out_sock_release_nosk: | |
4fdb3bb7 | 2087 | sock_release(sock); |
77247bbb | 2088 | return NULL; |
1da177e4 | 2089 | } |
9f00d977 | 2090 | EXPORT_SYMBOL(__netlink_kernel_create); |
b7c6ba6e DL |
2091 | |
2092 | void | |
2093 | netlink_kernel_release(struct sock *sk) | |
2094 | { | |
13d3078e EB |
2095 | if (sk == NULL || sk->sk_socket == NULL) |
2096 | return; | |
2097 | ||
2098 | sock_release(sk->sk_socket); | |
b7c6ba6e DL |
2099 | } |
2100 | EXPORT_SYMBOL(netlink_kernel_release); | |
2101 | ||
d136f1bd | 2102 | int __netlink_change_ngroups(struct sock *sk, unsigned int groups) |
b4ff4f04 | 2103 | { |
5c398dc8 | 2104 | struct listeners *new, *old; |
b4ff4f04 | 2105 | struct netlink_table *tbl = &nl_table[sk->sk_protocol]; |
b4ff4f04 JB |
2106 | |
2107 | if (groups < 32) | |
2108 | groups = 32; | |
2109 | ||
b4ff4f04 | 2110 | if (NLGRPSZ(tbl->groups) < NLGRPSZ(groups)) { |
5c398dc8 ED |
2111 | new = kzalloc(sizeof(*new) + NLGRPSZ(groups), GFP_ATOMIC); |
2112 | if (!new) | |
d136f1bd | 2113 | return -ENOMEM; |
6d772ac5 | 2114 | old = nl_deref_protected(tbl->listeners); |
5c398dc8 ED |
2115 | memcpy(new->masks, old->masks, NLGRPSZ(tbl->groups)); |
2116 | rcu_assign_pointer(tbl->listeners, new); | |
2117 | ||
37b6b935 | 2118 | kfree_rcu(old, rcu); |
b4ff4f04 JB |
2119 | } |
2120 | tbl->groups = groups; | |
2121 | ||
d136f1bd JB |
2122 | return 0; |
2123 | } | |
2124 | ||
2125 | /** | |
2126 | * netlink_change_ngroups - change number of multicast groups | |
2127 | * | |
2128 | * This changes the number of multicast groups that are available | |
2129 | * on a certain netlink family. Note that it is not possible to | |
2130 | * change the number of groups to below 32. Also note that it does | |
2131 | * not implicitly call netlink_clear_multicast_users() when the | |
2132 | * number of groups is reduced. | |
2133 | * | |
2134 | * @sk: The kernel netlink socket, as returned by netlink_kernel_create(). | |
2135 | * @groups: The new number of groups. | |
2136 | */ | |
2137 | int netlink_change_ngroups(struct sock *sk, unsigned int groups) | |
2138 | { | |
2139 | int err; | |
2140 | ||
2141 | netlink_table_grab(); | |
2142 | err = __netlink_change_ngroups(sk, groups); | |
b4ff4f04 | 2143 | netlink_table_ungrab(); |
d136f1bd | 2144 | |
b4ff4f04 JB |
2145 | return err; |
2146 | } | |
b4ff4f04 | 2147 | |
b8273570 JB |
2148 | void __netlink_clear_multicast_users(struct sock *ksk, unsigned int group) |
2149 | { | |
2150 | struct sock *sk; | |
b8273570 JB |
2151 | struct netlink_table *tbl = &nl_table[ksk->sk_protocol]; |
2152 | ||
b67bfe0d | 2153 | sk_for_each_bound(sk, &tbl->mc_list) |
b8273570 JB |
2154 | netlink_update_socket_mc(nlk_sk(sk), group, 0); |
2155 | } | |
2156 | ||
a46621a3 | 2157 | struct nlmsghdr * |
15e47304 | 2158 | __nlmsg_put(struct sk_buff *skb, u32 portid, u32 seq, int type, int len, int flags) |
a46621a3 DV |
2159 | { |
2160 | struct nlmsghdr *nlh; | |
573ce260 | 2161 | int size = nlmsg_msg_size(len); |
a46621a3 | 2162 | |
4df864c1 | 2163 | nlh = skb_put(skb, NLMSG_ALIGN(size)); |
a46621a3 DV |
2164 | nlh->nlmsg_type = type; |
2165 | nlh->nlmsg_len = size; | |
2166 | nlh->nlmsg_flags = flags; | |
15e47304 | 2167 | nlh->nlmsg_pid = portid; |
a46621a3 DV |
2168 | nlh->nlmsg_seq = seq; |
2169 | if (!__builtin_constant_p(size) || NLMSG_ALIGN(size) - size != 0) | |
573ce260 | 2170 | memset(nlmsg_data(nlh) + len, 0, NLMSG_ALIGN(size) - size); |
a46621a3 DV |
2171 | return nlh; |
2172 | } | |
2173 | EXPORT_SYMBOL(__nlmsg_put); | |
2174 | ||
1da177e4 LT |
2175 | /* |
2176 | * It looks a bit ugly. | |
2177 | * It would be better to create kernel thread. | |
2178 | */ | |
2179 | ||
e11eb32d DS |
2180 | static int netlink_dump_done(struct netlink_sock *nlk, struct sk_buff *skb, |
2181 | struct netlink_callback *cb, | |
2182 | struct netlink_ext_ack *extack) | |
2183 | { | |
2184 | struct nlmsghdr *nlh; | |
2185 | ||
2186 | nlh = nlmsg_put_answer(skb, cb, NLMSG_DONE, sizeof(nlk->dump_done_errno), | |
2187 | NLM_F_MULTI | cb->answer_flags); | |
2188 | if (WARN_ON(!nlh)) | |
2189 | return -ENOBUFS; | |
2190 | ||
2191 | nl_dump_check_consistent(cb, nlh); | |
2192 | memcpy(nlmsg_data(nlh), &nlk->dump_done_errno, sizeof(nlk->dump_done_errno)); | |
2193 | ||
2194 | if (extack->_msg && nlk->flags & NETLINK_F_EXT_ACK) { | |
2195 | nlh->nlmsg_flags |= NLM_F_ACK_TLVS; | |
2196 | if (!nla_put_string(skb, NLMSGERR_ATTR_MSG, extack->_msg)) | |
2197 | nlmsg_end(skb, nlh); | |
2198 | } | |
2199 | ||
2200 | return 0; | |
2201 | } | |
2202 | ||
1da177e4 LT |
2203 | static int netlink_dump(struct sock *sk) |
2204 | { | |
2205 | struct netlink_sock *nlk = nlk_sk(sk); | |
4a19edb6 | 2206 | struct netlink_ext_ack extack = {}; |
1da177e4 | 2207 | struct netlink_callback *cb; |
c7ac8679 | 2208 | struct sk_buff *skb = NULL; |
a1865f2e | 2209 | size_t max_recvmsg_len; |
92964c79 | 2210 | struct module *module; |
0642840b | 2211 | int err = -ENOBUFS; |
db65a3aa | 2212 | int alloc_min_size; |
c7ac8679 | 2213 | int alloc_size; |
1da177e4 | 2214 | |
af65bdfc | 2215 | mutex_lock(nlk->cb_mutex); |
16b304f3 | 2216 | if (!nlk->cb_running) { |
bf8b79e4 TG |
2217 | err = -EINVAL; |
2218 | goto errout_skb; | |
1da177e4 LT |
2219 | } |
2220 | ||
d1b4c689 | 2221 | if (atomic_read(&sk->sk_rmem_alloc) >= sk->sk_rcvbuf) |
f9c22888 | 2222 | goto errout_skb; |
9063e21f ED |
2223 | |
2224 | /* NLMSG_GOODSIZE is small to avoid high order allocations being | |
2225 | * required, but it makes sense to _attempt_ a 16K bytes allocation | |
2226 | * to reduce number of system calls on dump operations, if user | |
2227 | * ever provided a big enough buffer. | |
2228 | */ | |
db65a3aa AR |
2229 | cb = &nlk->cb; |
2230 | alloc_min_size = max_t(int, cb->min_dump_alloc, NLMSG_GOODSIZE); | |
2231 | ||
a1865f2e ED |
2232 | max_recvmsg_len = READ_ONCE(nlk->max_recvmsg_len); |
2233 | if (alloc_min_size < max_recvmsg_len) { | |
2234 | alloc_size = max_recvmsg_len; | |
d35c99ff ED |
2235 | skb = alloc_skb(alloc_size, |
2236 | (GFP_KERNEL & ~__GFP_DIRECT_RECLAIM) | | |
2237 | __GFP_NOWARN | __GFP_NORETRY); | |
9063e21f | 2238 | } |
db65a3aa AR |
2239 | if (!skb) { |
2240 | alloc_size = alloc_min_size; | |
c5b0db32 | 2241 | skb = alloc_skb(alloc_size, GFP_KERNEL); |
db65a3aa | 2242 | } |
c7ac8679 | 2243 | if (!skb) |
c63d6ea3 | 2244 | goto errout_skb; |
db65a3aa AR |
2245 | |
2246 | /* Trim skb to allocated size. User is expected to provide buffer as | |
2247 | * large as max(min_dump_alloc, 16KiB (mac_recvmsg_len capped at | |
2248 | * netlink_recvmsg())). dump will pack as many smaller messages as | |
2249 | * could fit within the allocated skb. skb is typically allocated | |
2250 | * with larger space than required (could be as much as near 2x the | |
2251 | * requested size with align to next power of 2 approach). Allowing | |
2252 | * dump to use the excess space makes it difficult for a user to have a | |
2253 | * reasonable static buffer based on the expected largest dump of a | |
2254 | * single netdev. The outcome is MSG_TRUNC error. | |
2255 | */ | |
d1b4c689 | 2256 | skb_reserve(skb, skb_tailroom(skb) - alloc_size); |
99c07327 ED |
2257 | |
2258 | /* Make sure malicious BPF programs can not read unitialized memory | |
2259 | * from skb->head -> skb->data | |
2260 | */ | |
2261 | skb_reset_network_header(skb); | |
2262 | skb_reset_mac_header(skb); | |
2263 | ||
f9c22888 | 2264 | netlink_skb_set_owner_r(skb, sk); |
c7ac8679 | 2265 | |
4a19edb6 DA |
2266 | if (nlk->dump_done_errno > 0) { |
2267 | cb->extack = &extack; | |
0642840b | 2268 | nlk->dump_done_errno = cb->dump(skb, cb); |
4a19edb6 DA |
2269 | cb->extack = NULL; |
2270 | } | |
1da177e4 | 2271 | |
0642840b JD |
2272 | if (nlk->dump_done_errno > 0 || |
2273 | skb_tailroom(skb) < nlmsg_total_size(sizeof(nlk->dump_done_errno))) { | |
af65bdfc | 2274 | mutex_unlock(nlk->cb_mutex); |
b1153f29 SH |
2275 | |
2276 | if (sk_filter(sk, skb)) | |
2277 | kfree_skb(skb); | |
4a7e7c2a ED |
2278 | else |
2279 | __netlink_sendskb(sk, skb); | |
1da177e4 LT |
2280 | return 0; |
2281 | } | |
2282 | ||
e11eb32d | 2283 | if (netlink_dump_done(nlk, skb, cb, &extack)) |
bf8b79e4 TG |
2284 | goto errout_skb; |
2285 | ||
e11eb32d DS |
2286 | #ifdef CONFIG_COMPAT_NETLINK_MESSAGES |
2287 | /* frag_list skb's data is used for compat tasks | |
2288 | * and the regular skb's data for normal (non-compat) tasks. | |
2289 | * See netlink_recvmsg(). | |
2290 | */ | |
2291 | if (unlikely(skb_shinfo(skb)->frag_list)) { | |
2292 | if (netlink_dump_done(nlk, skb_shinfo(skb)->frag_list, cb, &extack)) | |
2293 | goto errout_skb; | |
4a19edb6 | 2294 | } |
e11eb32d | 2295 | #endif |
4a19edb6 | 2296 | |
b1153f29 SH |
2297 | if (sk_filter(sk, skb)) |
2298 | kfree_skb(skb); | |
4a7e7c2a ED |
2299 | else |
2300 | __netlink_sendskb(sk, skb); | |
1da177e4 | 2301 | |
a8f74b22 TG |
2302 | if (cb->done) |
2303 | cb->done(cb); | |
1da177e4 | 2304 | |
16b304f3 | 2305 | nlk->cb_running = false; |
92964c79 HX |
2306 | module = cb->module; |
2307 | skb = cb->skb; | |
16b304f3 | 2308 | mutex_unlock(nlk->cb_mutex); |
92964c79 HX |
2309 | module_put(module); |
2310 | consume_skb(skb); | |
1da177e4 | 2311 | return 0; |
1797754e | 2312 | |
bf8b79e4 | 2313 | errout_skb: |
af65bdfc | 2314 | mutex_unlock(nlk->cb_mutex); |
bf8b79e4 | 2315 | kfree_skb(skb); |
bf8b79e4 | 2316 | return err; |
1da177e4 LT |
2317 | } |
2318 | ||
6dc878a8 G |
2319 | int __netlink_dump_start(struct sock *ssk, struct sk_buff *skb, |
2320 | const struct nlmsghdr *nlh, | |
2321 | struct netlink_dump_control *control) | |
1da177e4 | 2322 | { |
89d35528 | 2323 | struct netlink_sock *nlk, *nlk2; |
1da177e4 LT |
2324 | struct netlink_callback *cb; |
2325 | struct sock *sk; | |
b44d211e | 2326 | int ret; |
1da177e4 | 2327 | |
63354797 | 2328 | refcount_inc(&skb->users); |
f9c22888 | 2329 | |
15e47304 | 2330 | sk = netlink_lookup(sock_net(ssk), ssk->sk_protocol, NETLINK_CB(skb).portid); |
1da177e4 | 2331 | if (sk == NULL) { |
16b304f3 PS |
2332 | ret = -ECONNREFUSED; |
2333 | goto error_free; | |
1da177e4 | 2334 | } |
6dc878a8 | 2335 | |
16b304f3 | 2336 | nlk = nlk_sk(sk); |
af65bdfc | 2337 | mutex_lock(nlk->cb_mutex); |
6dc878a8 | 2338 | /* A dump is in progress... */ |
16b304f3 | 2339 | if (nlk->cb_running) { |
6dc878a8 | 2340 | ret = -EBUSY; |
16b304f3 | 2341 | goto error_unlock; |
1da177e4 | 2342 | } |
6dc878a8 | 2343 | /* add reference of module which cb->dump belongs to */ |
16b304f3 | 2344 | if (!try_module_get(control->module)) { |
6dc878a8 | 2345 | ret = -EPROTONOSUPPORT; |
16b304f3 | 2346 | goto error_unlock; |
6dc878a8 G |
2347 | } |
2348 | ||
16b304f3 PS |
2349 | cb = &nlk->cb; |
2350 | memset(cb, 0, sizeof(*cb)); | |
2351 | cb->dump = control->dump; | |
2352 | cb->done = control->done; | |
2353 | cb->nlh = nlh; | |
2354 | cb->data = control->data; | |
2355 | cb->module = control->module; | |
2356 | cb->min_dump_alloc = control->min_dump_alloc; | |
2357 | cb->skb = skb; | |
2358 | ||
89d35528 DA |
2359 | nlk2 = nlk_sk(NETLINK_CB(skb).sk); |
2360 | cb->strict_check = !!(nlk2->flags & NETLINK_F_STRICT_CHK); | |
2361 | ||
3730cf4d FW |
2362 | if (control->start) { |
2363 | ret = control->start(cb); | |
41c87425 | 2364 | if (ret) |
b87b6194 | 2365 | goto error_put; |
41c87425 JD |
2366 | } |
2367 | ||
16b304f3 | 2368 | nlk->cb_running = true; |
0642840b | 2369 | nlk->dump_done_errno = INT_MAX; |
16b304f3 | 2370 | |
af65bdfc | 2371 | mutex_unlock(nlk->cb_mutex); |
1da177e4 | 2372 | |
41c87425 | 2373 | ret = netlink_dump(sk); |
fc9e50f5 | 2374 | |
1da177e4 | 2375 | sock_put(sk); |
5c58298c | 2376 | |
b44d211e AV |
2377 | if (ret) |
2378 | return ret; | |
2379 | ||
5c58298c DL |
2380 | /* We successfully started a dump, by returning -EINTR we |
2381 | * signal not to send ACK even if it was requested. | |
2382 | */ | |
2383 | return -EINTR; | |
16b304f3 | 2384 | |
b87b6194 JD |
2385 | error_put: |
2386 | module_put(control->module); | |
16b304f3 PS |
2387 | error_unlock: |
2388 | sock_put(sk); | |
2389 | mutex_unlock(nlk->cb_mutex); | |
2390 | error_free: | |
2391 | kfree_skb(skb); | |
2392 | return ret; | |
1da177e4 | 2393 | } |
6dc878a8 | 2394 | EXPORT_SYMBOL(__netlink_dump_start); |
1da177e4 | 2395 | |
0c95cea2 JK |
2396 | static size_t |
2397 | netlink_ack_tlv_len(struct netlink_sock *nlk, int err, | |
2398 | const struct netlink_ext_ack *extack) | |
2399 | { | |
2400 | size_t tlvlen; | |
2401 | ||
2402 | if (!extack || !(nlk->flags & NETLINK_F_EXT_ACK)) | |
2403 | return 0; | |
2404 | ||
2405 | tlvlen = 0; | |
2406 | if (extack->_msg) | |
2407 | tlvlen += nla_total_size(strlen(extack->_msg) + 1); | |
2408 | if (extack->cookie_len) | |
2409 | tlvlen += nla_total_size(extack->cookie_len); | |
2410 | ||
2411 | /* Following attributes are only reported as error (not warning) */ | |
2412 | if (!err) | |
2413 | return tlvlen; | |
2414 | ||
2415 | if (extack->bad_attr) | |
2416 | tlvlen += nla_total_size(sizeof(u32)); | |
2417 | if (extack->policy) | |
2418 | tlvlen += netlink_policy_dump_attr_size_estimate(extack->policy); | |
690252f1 JK |
2419 | if (extack->miss_type) |
2420 | tlvlen += nla_total_size(sizeof(u32)); | |
2421 | if (extack->miss_nest) | |
2422 | tlvlen += nla_total_size(sizeof(u32)); | |
0c95cea2 JK |
2423 | |
2424 | return tlvlen; | |
2425 | } | |
2426 | ||
2427 | static void | |
2428 | netlink_ack_tlv_fill(struct sk_buff *in_skb, struct sk_buff *skb, | |
2429 | struct nlmsghdr *nlh, int err, | |
2430 | const struct netlink_ext_ack *extack) | |
2431 | { | |
2432 | if (extack->_msg) | |
2433 | WARN_ON(nla_put_string(skb, NLMSGERR_ATTR_MSG, extack->_msg)); | |
2434 | if (extack->cookie_len) | |
2435 | WARN_ON(nla_put(skb, NLMSGERR_ATTR_COOKIE, | |
2436 | extack->cookie_len, extack->cookie)); | |
2437 | ||
2438 | if (!err) | |
2439 | return; | |
2440 | ||
2441 | if (extack->bad_attr && | |
2442 | !WARN_ON((u8 *)extack->bad_attr < in_skb->data || | |
2443 | (u8 *)extack->bad_attr >= in_skb->data + in_skb->len)) | |
2444 | WARN_ON(nla_put_u32(skb, NLMSGERR_ATTR_OFFS, | |
2445 | (u8 *)extack->bad_attr - (u8 *)nlh)); | |
2446 | if (extack->policy) | |
2447 | netlink_policy_dump_write_attr(skb, extack->policy, | |
2448 | NLMSGERR_ATTR_POLICY); | |
690252f1 JK |
2449 | if (extack->miss_type) |
2450 | WARN_ON(nla_put_u32(skb, NLMSGERR_ATTR_MISS_TYPE, | |
2451 | extack->miss_type)); | |
2452 | if (extack->miss_nest && | |
2453 | !WARN_ON((u8 *)extack->miss_nest < in_skb->data || | |
2454 | (u8 *)extack->miss_nest > in_skb->data + in_skb->len)) | |
2455 | WARN_ON(nla_put_u32(skb, NLMSGERR_ATTR_MISS_NEST, | |
2456 | (u8 *)extack->miss_nest - (u8 *)nlh)); | |
0c95cea2 JK |
2457 | } |
2458 | ||
2d4bc933 JB |
2459 | void netlink_ack(struct sk_buff *in_skb, struct nlmsghdr *nlh, int err, |
2460 | const struct netlink_ext_ack *extack) | |
1da177e4 LT |
2461 | { |
2462 | struct sk_buff *skb; | |
2463 | struct nlmsghdr *rep; | |
2464 | struct nlmsgerr *errmsg; | |
339bf98f | 2465 | size_t payload = sizeof(*errmsg); |
0a6a3a23 | 2466 | struct netlink_sock *nlk = nlk_sk(NETLINK_CB(in_skb).sk); |
2d4bc933 | 2467 | unsigned int flags = 0; |
0c95cea2 | 2468 | size_t tlvlen; |
1da177e4 | 2469 | |
0a6a3a23 | 2470 | /* Error messages get the original request appened, unless the user |
2d4bc933 JB |
2471 | * requests to cap the error message, and get extra error data if |
2472 | * requested. | |
0a6a3a23 | 2473 | */ |
fe2a31d7 MK |
2474 | if (err && !(nlk->flags & NETLINK_F_CAP_ACK)) |
2475 | payload += nlmsg_len(nlh); | |
2476 | else | |
2d4bc933 | 2477 | flags |= NLM_F_CAPPED; |
1da177e4 | 2478 | |
0c95cea2 | 2479 | tlvlen = netlink_ack_tlv_len(nlk, err, extack); |
2d4bc933 JB |
2480 | if (tlvlen) |
2481 | flags |= NLM_F_ACK_TLVS; | |
2482 | ||
2483 | skb = nlmsg_new(payload + tlvlen, GFP_KERNEL); | |
738136a0 | 2484 | if (!skb) |
e6976148 | 2485 | goto err_skb; |
1da177e4 | 2486 | |
710d21fd | 2487 | rep = nlmsg_put(skb, NETLINK_CB(in_skb).portid, nlh->nlmsg_seq, |
738136a0 JK |
2488 | NLMSG_ERROR, sizeof(*errmsg), flags); |
2489 | if (!rep) | |
2490 | goto err_bad_put; | |
bf8b79e4 | 2491 | errmsg = nlmsg_data(rep); |
1da177e4 | 2492 | errmsg->error = err; |
738136a0 JK |
2493 | errmsg->msg = *nlh; |
2494 | ||
2495 | if (!(flags & NLM_F_CAPPED)) { | |
2496 | if (!nlmsg_append(skb, nlmsg_len(nlh))) | |
2497 | goto err_bad_put; | |
2498 | ||
c73a72f4 | 2499 | memcpy(nlmsg_data(&errmsg->msg), nlmsg_data(nlh), |
738136a0 JK |
2500 | nlmsg_len(nlh)); |
2501 | } | |
2d4bc933 | 2502 | |
0c95cea2 JK |
2503 | if (tlvlen) |
2504 | netlink_ack_tlv_fill(in_skb, skb, nlh, err, extack); | |
2d4bc933 JB |
2505 | |
2506 | nlmsg_end(skb, rep); | |
2507 | ||
01757f53 | 2508 | nlmsg_unicast(in_skb->sk, skb, NETLINK_CB(in_skb).portid); |
738136a0 JK |
2509 | |
2510 | return; | |
2511 | ||
2512 | err_bad_put: | |
e6976148 TC |
2513 | nlmsg_free(skb); |
2514 | err_skb: | |
738136a0 JK |
2515 | NETLINK_CB(in_skb).sk->sk_err = ENOBUFS; |
2516 | sk_error_report(NETLINK_CB(in_skb).sk); | |
1da177e4 | 2517 | } |
6ac552fd | 2518 | EXPORT_SYMBOL(netlink_ack); |
1da177e4 | 2519 | |
cd40b7d3 | 2520 | int netlink_rcv_skb(struct sk_buff *skb, int (*cb)(struct sk_buff *, |
2d4bc933 JB |
2521 | struct nlmsghdr *, |
2522 | struct netlink_ext_ack *)) | |
82ace47a | 2523 | { |
cbbdf843 | 2524 | struct netlink_ext_ack extack; |
82ace47a TG |
2525 | struct nlmsghdr *nlh; |
2526 | int err; | |
2527 | ||
2528 | while (skb->len >= nlmsg_total_size(0)) { | |
cd40b7d3 DL |
2529 | int msglen; |
2530 | ||
cd443f1e | 2531 | memset(&extack, 0, sizeof(extack)); |
b529ccf2 | 2532 | nlh = nlmsg_hdr(skb); |
d35b6856 | 2533 | err = 0; |
82ace47a | 2534 | |
ad8e4b75 | 2535 | if (nlh->nlmsg_len < NLMSG_HDRLEN || skb->len < nlh->nlmsg_len) |
82ace47a TG |
2536 | return 0; |
2537 | ||
d35b6856 TG |
2538 | /* Only requests are handled by the kernel */ |
2539 | if (!(nlh->nlmsg_flags & NLM_F_REQUEST)) | |
5c58298c | 2540 | goto ack; |
45e7ae7f TG |
2541 | |
2542 | /* Skip control messages */ | |
2543 | if (nlh->nlmsg_type < NLMSG_MIN_TYPE) | |
5c58298c | 2544 | goto ack; |
d35b6856 | 2545 | |
2d4bc933 | 2546 | err = cb(skb, nlh, &extack); |
5c58298c DL |
2547 | if (err == -EINTR) |
2548 | goto skip; | |
2549 | ||
2550 | ack: | |
d35b6856 | 2551 | if (nlh->nlmsg_flags & NLM_F_ACK || err) |
2d4bc933 | 2552 | netlink_ack(skb, nlh, err, &extack); |
82ace47a | 2553 | |
5c58298c | 2554 | skip: |
6ac552fd | 2555 | msglen = NLMSG_ALIGN(nlh->nlmsg_len); |
cd40b7d3 DL |
2556 | if (msglen > skb->len) |
2557 | msglen = skb->len; | |
2558 | skb_pull(skb, msglen); | |
82ace47a TG |
2559 | } |
2560 | ||
2561 | return 0; | |
2562 | } | |
6ac552fd | 2563 | EXPORT_SYMBOL(netlink_rcv_skb); |
82ace47a | 2564 | |
d387f6ad TG |
2565 | /** |
2566 | * nlmsg_notify - send a notification netlink message | |
2567 | * @sk: netlink socket to use | |
2568 | * @skb: notification message | |
15e47304 | 2569 | * @portid: destination netlink portid for reports or 0 |
d387f6ad TG |
2570 | * @group: destination multicast group or 0 |
2571 | * @report: 1 to report back, 0 to disable | |
2572 | * @flags: allocation flags | |
2573 | */ | |
15e47304 | 2574 | int nlmsg_notify(struct sock *sk, struct sk_buff *skb, u32 portid, |
d387f6ad TG |
2575 | unsigned int group, int report, gfp_t flags) |
2576 | { | |
2577 | int err = 0; | |
2578 | ||
2579 | if (group) { | |
15e47304 | 2580 | int exclude_portid = 0; |
d387f6ad TG |
2581 | |
2582 | if (report) { | |
63354797 | 2583 | refcount_inc(&skb->users); |
15e47304 | 2584 | exclude_portid = portid; |
d387f6ad TG |
2585 | } |
2586 | ||
1ce85fe4 PNA |
2587 | /* errors reported via destination sk->sk_err, but propagate |
2588 | * delivery errors if NETLINK_BROADCAST_ERROR flag is set */ | |
15e47304 | 2589 | err = nlmsg_multicast(sk, skb, exclude_portid, group, flags); |
fef773fc YD |
2590 | if (err == -ESRCH) |
2591 | err = 0; | |
d387f6ad TG |
2592 | } |
2593 | ||
1ce85fe4 PNA |
2594 | if (report) { |
2595 | int err2; | |
2596 | ||
15e47304 | 2597 | err2 = nlmsg_unicast(sk, skb, portid); |
fef773fc | 2598 | if (!err) |
1ce85fe4 PNA |
2599 | err = err2; |
2600 | } | |
d387f6ad TG |
2601 | |
2602 | return err; | |
2603 | } | |
6ac552fd | 2604 | EXPORT_SYMBOL(nlmsg_notify); |
d387f6ad | 2605 | |
1da177e4 LT |
2606 | #ifdef CONFIG_PROC_FS |
2607 | struct nl_seq_iter { | |
e372c414 | 2608 | struct seq_net_private p; |
56d28b1e | 2609 | struct rhashtable_iter hti; |
1da177e4 | 2610 | int link; |
1da177e4 LT |
2611 | }; |
2612 | ||
abf9979f | 2613 | static void netlink_walk_start(struct nl_seq_iter *iter) |
1da177e4 | 2614 | { |
6c4128f6 | 2615 | rhashtable_walk_enter(&nl_table[iter->link].hash, &iter->hti); |
97a6ec4a | 2616 | rhashtable_walk_start(&iter->hti); |
1da177e4 LT |
2617 | } |
2618 | ||
56d28b1e | 2619 | static void netlink_walk_stop(struct nl_seq_iter *iter) |
1da177e4 | 2620 | { |
56d28b1e HX |
2621 | rhashtable_walk_stop(&iter->hti); |
2622 | rhashtable_walk_exit(&iter->hti); | |
1da177e4 LT |
2623 | } |
2624 | ||
56d28b1e | 2625 | static void *__netlink_seq_next(struct seq_file *seq) |
1da177e4 | 2626 | { |
56d28b1e | 2627 | struct nl_seq_iter *iter = seq->private; |
e341694e | 2628 | struct netlink_sock *nlk; |
1da177e4 | 2629 | |
56d28b1e HX |
2630 | do { |
2631 | for (;;) { | |
56d28b1e | 2632 | nlk = rhashtable_walk_next(&iter->hti); |
746fac4d | 2633 | |
56d28b1e HX |
2634 | if (IS_ERR(nlk)) { |
2635 | if (PTR_ERR(nlk) == -EAGAIN) | |
2636 | continue; | |
e341694e | 2637 | |
56d28b1e HX |
2638 | return nlk; |
2639 | } | |
1da177e4 | 2640 | |
56d28b1e HX |
2641 | if (nlk) |
2642 | break; | |
1da177e4 | 2643 | |
56d28b1e HX |
2644 | netlink_walk_stop(iter); |
2645 | if (++iter->link >= MAX_LINKS) | |
2646 | return NULL; | |
da12c90e | 2647 | |
abf9979f | 2648 | netlink_walk_start(iter); |
1da177e4 | 2649 | } |
56d28b1e | 2650 | } while (sock_net(&nlk->sk) != seq_file_net(seq)); |
1da177e4 | 2651 | |
56d28b1e HX |
2652 | return nlk; |
2653 | } | |
1da177e4 | 2654 | |
56d28b1e | 2655 | static void *netlink_seq_start(struct seq_file *seq, loff_t *posp) |
64fbca01 | 2656 | __acquires(RCU) |
56d28b1e HX |
2657 | { |
2658 | struct nl_seq_iter *iter = seq->private; | |
2659 | void *obj = SEQ_START_TOKEN; | |
2660 | loff_t pos; | |
56d28b1e HX |
2661 | |
2662 | iter->link = 0; | |
2663 | ||
abf9979f | 2664 | netlink_walk_start(iter); |
56d28b1e HX |
2665 | |
2666 | for (pos = *posp; pos && obj && !IS_ERR(obj); pos--) | |
2667 | obj = __netlink_seq_next(seq); | |
2668 | ||
2669 | return obj; | |
2670 | } | |
2671 | ||
2672 | static void *netlink_seq_next(struct seq_file *seq, void *v, loff_t *pos) | |
2673 | { | |
2674 | ++*pos; | |
2675 | return __netlink_seq_next(seq); | |
1da177e4 LT |
2676 | } |
2677 | ||
138d0be3 | 2678 | static void netlink_native_seq_stop(struct seq_file *seq, void *v) |
1da177e4 | 2679 | { |
56d28b1e HX |
2680 | struct nl_seq_iter *iter = seq->private; |
2681 | ||
2682 | if (iter->link >= MAX_LINKS) | |
2683 | return; | |
2684 | ||
2685 | netlink_walk_stop(iter); | |
1da177e4 LT |
2686 | } |
2687 | ||
2688 | ||
138d0be3 | 2689 | static int netlink_native_seq_show(struct seq_file *seq, void *v) |
1da177e4 | 2690 | { |
658cb354 | 2691 | if (v == SEQ_START_TOKEN) { |
1da177e4 | 2692 | seq_puts(seq, |
ae552ac2 YB |
2693 | "sk Eth Pid Groups " |
2694 | "Rmem Wmem Dump Locks Drops Inode\n"); | |
658cb354 | 2695 | } else { |
1da177e4 LT |
2696 | struct sock *s = v; |
2697 | struct netlink_sock *nlk = nlk_sk(s); | |
2698 | ||
ea9a0379 | 2699 | seq_printf(seq, "%pK %-3d %-10u %08x %-8d %-8d %-5d %-8d %-8u %-8lu\n", |
1da177e4 LT |
2700 | s, |
2701 | s->sk_protocol, | |
15e47304 | 2702 | nlk->portid, |
513c2500 | 2703 | nlk->groups ? (u32)nlk->groups[0] : 0, |
31e6d363 ED |
2704 | sk_rmem_alloc_get(s), |
2705 | sk_wmem_alloc_get(s), | |
16b304f3 | 2706 | nlk->cb_running, |
41c6d650 | 2707 | refcount_read(&s->sk_refcnt), |
cf0aa4e0 MY |
2708 | atomic_read(&s->sk_drops), |
2709 | sock_i_ino(s) | |
1da177e4 LT |
2710 | ); |
2711 | ||
2712 | } | |
2713 | return 0; | |
2714 | } | |
2715 | ||
138d0be3 YS |
2716 | #ifdef CONFIG_BPF_SYSCALL |
2717 | struct bpf_iter__netlink { | |
2718 | __bpf_md_ptr(struct bpf_iter_meta *, meta); | |
2719 | __bpf_md_ptr(struct netlink_sock *, sk); | |
2720 | }; | |
2721 | ||
2722 | DEFINE_BPF_ITER_FUNC(netlink, struct bpf_iter_meta *meta, struct netlink_sock *sk) | |
2723 | ||
2724 | static int netlink_prog_seq_show(struct bpf_prog *prog, | |
2725 | struct bpf_iter_meta *meta, | |
2726 | void *v) | |
2727 | { | |
2728 | struct bpf_iter__netlink ctx; | |
2729 | ||
2730 | meta->seq_num--; /* skip SEQ_START_TOKEN */ | |
2731 | ctx.meta = meta; | |
2732 | ctx.sk = nlk_sk((struct sock *)v); | |
2733 | return bpf_iter_run_prog(prog, &ctx); | |
2734 | } | |
2735 | ||
2736 | static int netlink_seq_show(struct seq_file *seq, void *v) | |
2737 | { | |
2738 | struct bpf_iter_meta meta; | |
2739 | struct bpf_prog *prog; | |
2740 | ||
2741 | meta.seq = seq; | |
2742 | prog = bpf_iter_get_info(&meta, false); | |
2743 | if (!prog) | |
2744 | return netlink_native_seq_show(seq, v); | |
2745 | ||
2746 | if (v != SEQ_START_TOKEN) | |
2747 | return netlink_prog_seq_show(prog, &meta, v); | |
2748 | ||
2749 | return 0; | |
2750 | } | |
2751 | ||
2752 | static void netlink_seq_stop(struct seq_file *seq, void *v) | |
2753 | { | |
2754 | struct bpf_iter_meta meta; | |
2755 | struct bpf_prog *prog; | |
2756 | ||
2757 | if (!v) { | |
2758 | meta.seq = seq; | |
2759 | prog = bpf_iter_get_info(&meta, true); | |
2760 | if (prog) | |
2761 | (void)netlink_prog_seq_show(prog, &meta, v); | |
2762 | } | |
2763 | ||
2764 | netlink_native_seq_stop(seq, v); | |
2765 | } | |
2766 | #else | |
2767 | static int netlink_seq_show(struct seq_file *seq, void *v) | |
2768 | { | |
2769 | return netlink_native_seq_show(seq, v); | |
2770 | } | |
2771 | ||
2772 | static void netlink_seq_stop(struct seq_file *seq, void *v) | |
2773 | { | |
2774 | netlink_native_seq_stop(seq, v); | |
2775 | } | |
2776 | #endif | |
2777 | ||
56b3d975 | 2778 | static const struct seq_operations netlink_seq_ops = { |
1da177e4 LT |
2779 | .start = netlink_seq_start, |
2780 | .next = netlink_seq_next, | |
2781 | .stop = netlink_seq_stop, | |
2782 | .show = netlink_seq_show, | |
2783 | }; | |
1da177e4 LT |
2784 | #endif |
2785 | ||
2786 | int netlink_register_notifier(struct notifier_block *nb) | |
2787 | { | |
efa172f4 | 2788 | return blocking_notifier_chain_register(&netlink_chain, nb); |
1da177e4 | 2789 | } |
6ac552fd | 2790 | EXPORT_SYMBOL(netlink_register_notifier); |
1da177e4 LT |
2791 | |
2792 | int netlink_unregister_notifier(struct notifier_block *nb) | |
2793 | { | |
efa172f4 | 2794 | return blocking_notifier_chain_unregister(&netlink_chain, nb); |
1da177e4 | 2795 | } |
6ac552fd | 2796 | EXPORT_SYMBOL(netlink_unregister_notifier); |
746fac4d | 2797 | |
90ddc4f0 | 2798 | static const struct proto_ops netlink_ops = { |
1da177e4 LT |
2799 | .family = PF_NETLINK, |
2800 | .owner = THIS_MODULE, | |
2801 | .release = netlink_release, | |
2802 | .bind = netlink_bind, | |
2803 | .connect = netlink_connect, | |
2804 | .socketpair = sock_no_socketpair, | |
2805 | .accept = sock_no_accept, | |
2806 | .getname = netlink_getname, | |
a11e1d43 | 2807 | .poll = datagram_poll, |
025c6818 | 2808 | .ioctl = netlink_ioctl, |
1da177e4 LT |
2809 | .listen = sock_no_listen, |
2810 | .shutdown = sock_no_shutdown, | |
9a4595bc PM |
2811 | .setsockopt = netlink_setsockopt, |
2812 | .getsockopt = netlink_getsockopt, | |
1da177e4 LT |
2813 | .sendmsg = netlink_sendmsg, |
2814 | .recvmsg = netlink_recvmsg, | |
d1b4c689 | 2815 | .mmap = sock_no_mmap, |
1da177e4 LT |
2816 | .sendpage = sock_no_sendpage, |
2817 | }; | |
2818 | ||
ec1b4cf7 | 2819 | static const struct net_proto_family netlink_family_ops = { |
1da177e4 LT |
2820 | .family = PF_NETLINK, |
2821 | .create = netlink_create, | |
2822 | .owner = THIS_MODULE, /* for consistency 8) */ | |
2823 | }; | |
2824 | ||
4665079c | 2825 | static int __net_init netlink_net_init(struct net *net) |
b4b51029 EB |
2826 | { |
2827 | #ifdef CONFIG_PROC_FS | |
c3506372 CH |
2828 | if (!proc_create_net("netlink", 0, net->proc_net, &netlink_seq_ops, |
2829 | sizeof(struct nl_seq_iter))) | |
b4b51029 EB |
2830 | return -ENOMEM; |
2831 | #endif | |
2832 | return 0; | |
2833 | } | |
2834 | ||
4665079c | 2835 | static void __net_exit netlink_net_exit(struct net *net) |
b4b51029 EB |
2836 | { |
2837 | #ifdef CONFIG_PROC_FS | |
ece31ffd | 2838 | remove_proc_entry("netlink", net->proc_net); |
b4b51029 EB |
2839 | #endif |
2840 | } | |
2841 | ||
b963ea89 DM |
2842 | static void __init netlink_add_usersock_entry(void) |
2843 | { | |
5c398dc8 | 2844 | struct listeners *listeners; |
b963ea89 DM |
2845 | int groups = 32; |
2846 | ||
5c398dc8 | 2847 | listeners = kzalloc(sizeof(*listeners) + NLGRPSZ(groups), GFP_KERNEL); |
b963ea89 | 2848 | if (!listeners) |
5c398dc8 | 2849 | panic("netlink_add_usersock_entry: Cannot allocate listeners\n"); |
b963ea89 DM |
2850 | |
2851 | netlink_table_grab(); | |
2852 | ||
2853 | nl_table[NETLINK_USERSOCK].groups = groups; | |
5c398dc8 | 2854 | rcu_assign_pointer(nl_table[NETLINK_USERSOCK].listeners, listeners); |
b963ea89 DM |
2855 | nl_table[NETLINK_USERSOCK].module = THIS_MODULE; |
2856 | nl_table[NETLINK_USERSOCK].registered = 1; | |
9785e10a | 2857 | nl_table[NETLINK_USERSOCK].flags = NL_CFG_F_NONROOT_SEND; |
b963ea89 DM |
2858 | |
2859 | netlink_table_ungrab(); | |
2860 | } | |
2861 | ||
022cbae6 | 2862 | static struct pernet_operations __net_initdata netlink_net_ops = { |
b4b51029 EB |
2863 | .init = netlink_net_init, |
2864 | .exit = netlink_net_exit, | |
2865 | }; | |
2866 | ||
49f7b33e | 2867 | static inline u32 netlink_hash(const void *data, u32 len, u32 seed) |
c428ecd1 HX |
2868 | { |
2869 | const struct netlink_sock *nlk = data; | |
2870 | struct netlink_compare_arg arg; | |
2871 | ||
da314c99 | 2872 | netlink_compare_arg_init(&arg, sock_net(&nlk->sk), nlk->portid); |
11b58ba1 | 2873 | return jhash2((u32 *)&arg, netlink_compare_arg_len / sizeof(u32), seed); |
c428ecd1 HX |
2874 | } |
2875 | ||
2876 | static const struct rhashtable_params netlink_rhashtable_params = { | |
2877 | .head_offset = offsetof(struct netlink_sock, node), | |
2878 | .key_len = netlink_compare_arg_len, | |
c428ecd1 HX |
2879 | .obj_hashfn = netlink_hash, |
2880 | .obj_cmpfn = netlink_compare, | |
b5e2c150 | 2881 | .automatic_shrinking = true, |
c428ecd1 HX |
2882 | }; |
2883 | ||
138d0be3 | 2884 | #if defined(CONFIG_BPF_SYSCALL) && defined(CONFIG_PROC_FS) |
951cf368 YS |
2885 | BTF_ID_LIST(btf_netlink_sock_id) |
2886 | BTF_ID(struct, netlink_sock) | |
2887 | ||
14fc6bd6 | 2888 | static const struct bpf_iter_seq_info netlink_seq_info = { |
15172a46 YS |
2889 | .seq_ops = &netlink_seq_ops, |
2890 | .init_seq_private = bpf_iter_init_seq_net, | |
2891 | .fini_seq_private = bpf_iter_fini_seq_net, | |
2892 | .seq_priv_size = sizeof(struct nl_seq_iter), | |
14fc6bd6 YS |
2893 | }; |
2894 | ||
2895 | static struct bpf_iter_reg netlink_reg_info = { | |
2896 | .target = "netlink", | |
3c32cc1b YS |
2897 | .ctx_arg_info_size = 1, |
2898 | .ctx_arg_info = { | |
2899 | { offsetof(struct bpf_iter__netlink, sk), | |
2900 | PTR_TO_BTF_ID_OR_NULL }, | |
2901 | }, | |
14fc6bd6 | 2902 | .seq_info = &netlink_seq_info, |
15172a46 YS |
2903 | }; |
2904 | ||
138d0be3 YS |
2905 | static int __init bpf_iter_register(void) |
2906 | { | |
951cf368 | 2907 | netlink_reg_info.ctx_arg_info[0].btf_id = *btf_netlink_sock_id; |
15172a46 | 2908 | return bpf_iter_reg_target(&netlink_reg_info); |
138d0be3 YS |
2909 | } |
2910 | #endif | |
2911 | ||
1da177e4 LT |
2912 | static int __init netlink_proto_init(void) |
2913 | { | |
1da177e4 | 2914 | int i; |
1da177e4 LT |
2915 | int err = proto_register(&netlink_proto, 0); |
2916 | ||
2917 | if (err != 0) | |
2918 | goto out; | |
2919 | ||
138d0be3 YS |
2920 | #if defined(CONFIG_BPF_SYSCALL) && defined(CONFIG_PROC_FS) |
2921 | err = bpf_iter_register(); | |
2922 | if (err) | |
2923 | goto out; | |
2924 | #endif | |
2925 | ||
c593642c | 2926 | BUILD_BUG_ON(sizeof(struct netlink_skb_parms) > sizeof_field(struct sk_buff, cb)); |
1da177e4 | 2927 | |
0da974f4 | 2928 | nl_table = kcalloc(MAX_LINKS, sizeof(*nl_table), GFP_KERNEL); |
fab2caf6 AM |
2929 | if (!nl_table) |
2930 | goto panic; | |
1da177e4 | 2931 | |
1da177e4 | 2932 | for (i = 0; i < MAX_LINKS; i++) { |
c428ecd1 HX |
2933 | if (rhashtable_init(&nl_table[i].hash, |
2934 | &netlink_rhashtable_params) < 0) { | |
e341694e TG |
2935 | while (--i > 0) |
2936 | rhashtable_destroy(&nl_table[i].hash); | |
1da177e4 | 2937 | kfree(nl_table); |
fab2caf6 | 2938 | goto panic; |
1da177e4 | 2939 | } |
1da177e4 LT |
2940 | } |
2941 | ||
b963ea89 DM |
2942 | netlink_add_usersock_entry(); |
2943 | ||
1da177e4 | 2944 | sock_register(&netlink_family_ops); |
b4b51029 | 2945 | register_pernet_subsys(&netlink_net_ops); |
25e3f70f | 2946 | register_pernet_subsys(&netlink_tap_net_ops); |
746fac4d | 2947 | /* The netlink device handler may be needed early. */ |
1da177e4 LT |
2948 | rtnetlink_init(); |
2949 | out: | |
2950 | return err; | |
fab2caf6 AM |
2951 | panic: |
2952 | panic("netlink_init: Cannot allocate nl_table\n"); | |
1da177e4 LT |
2953 | } |
2954 | ||
1da177e4 | 2955 | core_initcall(netlink_proto_init); |