Commit | Line | Data |
---|---|---|
1da177e4 LT |
1 | /* |
2 | * NETLINK Kernel-user communication protocol. | |
3 | * | |
113aa838 | 4 | * Authors: Alan Cox <alan@lxorguk.ukuu.org.uk> |
1da177e4 | 5 | * Alexey Kuznetsov <kuznet@ms2.inr.ac.ru> |
cd1df525 | 6 | * Patrick McHardy <kaber@trash.net> |
1da177e4 LT |
7 | * |
8 | * This program is free software; you can redistribute it and/or | |
9 | * modify it under the terms of the GNU General Public License | |
10 | * as published by the Free Software Foundation; either version | |
11 | * 2 of the License, or (at your option) any later version. | |
746fac4d | 12 | * |
1da177e4 LT |
13 | * Tue Jun 26 14:36:48 MEST 2001 Herbert "herp" Rosmanith |
14 | * added netlink_proto_exit | |
15 | * Tue Jan 22 18:32:44 BRST 2002 Arnaldo C. de Melo <acme@conectiva.com.br> | |
16 | * use nlk_sk, as sk->protinfo is on a diet 8) | |
4fdb3bb7 HW |
17 | * Fri Jul 22 19:51:12 MEST 2005 Harald Welte <laforge@gnumonks.org> |
18 | * - inc module use count of module that owns | |
19 | * the kernel socket in case userspace opens | |
20 | * socket of same protocol | |
21 | * - remove all module support, since netlink is | |
22 | * mandatory if CONFIG_NET=y these days | |
1da177e4 LT |
23 | */ |
24 | ||
1da177e4 LT |
25 | #include <linux/module.h> |
26 | ||
4fc268d2 | 27 | #include <linux/capability.h> |
1da177e4 LT |
28 | #include <linux/kernel.h> |
29 | #include <linux/init.h> | |
1da177e4 LT |
30 | #include <linux/signal.h> |
31 | #include <linux/sched.h> | |
32 | #include <linux/errno.h> | |
33 | #include <linux/string.h> | |
34 | #include <linux/stat.h> | |
35 | #include <linux/socket.h> | |
36 | #include <linux/un.h> | |
37 | #include <linux/fcntl.h> | |
38 | #include <linux/termios.h> | |
39 | #include <linux/sockios.h> | |
40 | #include <linux/net.h> | |
41 | #include <linux/fs.h> | |
42 | #include <linux/slab.h> | |
43 | #include <asm/uaccess.h> | |
44 | #include <linux/skbuff.h> | |
45 | #include <linux/netdevice.h> | |
46 | #include <linux/rtnetlink.h> | |
47 | #include <linux/proc_fs.h> | |
48 | #include <linux/seq_file.h> | |
1da177e4 LT |
49 | #include <linux/notifier.h> |
50 | #include <linux/security.h> | |
51 | #include <linux/jhash.h> | |
52 | #include <linux/jiffies.h> | |
53 | #include <linux/random.h> | |
54 | #include <linux/bitops.h> | |
55 | #include <linux/mm.h> | |
56 | #include <linux/types.h> | |
54e0f520 | 57 | #include <linux/audit.h> |
af65bdfc | 58 | #include <linux/mutex.h> |
ccdfcc39 | 59 | #include <linux/vmalloc.h> |
bcbde0d4 | 60 | #include <linux/if_arp.h> |
e341694e | 61 | #include <linux/rhashtable.h> |
9652e931 | 62 | #include <asm/cacheflush.h> |
e341694e | 63 | #include <linux/hash.h> |
ee1c2442 | 64 | #include <linux/genetlink.h> |
54e0f520 | 65 | |
457c4cbc | 66 | #include <net/net_namespace.h> |
1da177e4 LT |
67 | #include <net/sock.h> |
68 | #include <net/scm.h> | |
82ace47a | 69 | #include <net/netlink.h> |
1da177e4 | 70 | |
0f29c768 | 71 | #include "af_netlink.h" |
1da177e4 | 72 | |
5c398dc8 ED |
73 | struct listeners { |
74 | struct rcu_head rcu; | |
75 | unsigned long masks[0]; | |
6c04bb18 JB |
76 | }; |
77 | ||
cd967e05 | 78 | /* state bits */ |
cc3a572f | 79 | #define NETLINK_S_CONGESTED 0x0 |
cd967e05 PM |
80 | |
81 | /* flags */ | |
cc3a572f ND |
82 | #define NETLINK_F_KERNEL_SOCKET 0x1 |
83 | #define NETLINK_F_RECV_PKTINFO 0x2 | |
84 | #define NETLINK_F_BROADCAST_SEND_ERROR 0x4 | |
85 | #define NETLINK_F_RECV_NO_ENOBUFS 0x8 | |
59324cf3 | 86 | #define NETLINK_F_LISTEN_ALL_NSID 0x10 |
0a6a3a23 | 87 | #define NETLINK_F_CAP_ACK 0x20 |
77247bbb | 88 | |
035c4c16 | 89 | static inline int netlink_is_kernel(struct sock *sk) |
aed81560 | 90 | { |
cc3a572f | 91 | return nlk_sk(sk)->flags & NETLINK_F_KERNEL_SOCKET; |
aed81560 DL |
92 | } |
93 | ||
91dd93f9 | 94 | struct netlink_table *nl_table __read_mostly; |
0f29c768 | 95 | EXPORT_SYMBOL_GPL(nl_table); |
1da177e4 LT |
96 | |
97 | static DECLARE_WAIT_QUEUE_HEAD(nl_table_wait); | |
98 | ||
99 | static int netlink_dump(struct sock *sk); | |
9652e931 | 100 | static void netlink_skb_destructor(struct sk_buff *skb); |
1da177e4 | 101 | |
78fd1d0a | 102 | /* nl_table locking explained: |
21e4902a | 103 | * Lookup and traversal are protected with an RCU read-side lock. Insertion |
c5adde94 | 104 | * and removal are protected with per bucket lock while using RCU list |
21e4902a TG |
105 | * modification primitives and may run in parallel to RCU protected lookups. |
106 | * Destruction of the Netlink socket may only occur *after* nl_table_lock has | |
107 | * been acquired * either during or after the socket has been removed from | |
108 | * the list and after an RCU grace period. | |
78fd1d0a | 109 | */ |
0f29c768 AV |
110 | DEFINE_RWLOCK(nl_table_lock); |
111 | EXPORT_SYMBOL_GPL(nl_table_lock); | |
1da177e4 LT |
112 | static atomic_t nl_table_users = ATOMIC_INIT(0); |
113 | ||
6d772ac5 ED |
114 | #define nl_deref_protected(X) rcu_dereference_protected(X, lockdep_is_held(&nl_table_lock)); |
115 | ||
e041c683 | 116 | static ATOMIC_NOTIFIER_HEAD(netlink_chain); |
1da177e4 | 117 | |
bcbde0d4 DB |
118 | static DEFINE_SPINLOCK(netlink_tap_lock); |
119 | static struct list_head netlink_tap_all __read_mostly; | |
120 | ||
c428ecd1 HX |
121 | static const struct rhashtable_params netlink_rhashtable_params; |
122 | ||
b57ef81f | 123 | static inline u32 netlink_group_mask(u32 group) |
d629b836 PM |
124 | { |
125 | return group ? 1 << (group - 1) : 0; | |
126 | } | |
127 | ||
bcbde0d4 DB |
128 | int netlink_add_tap(struct netlink_tap *nt) |
129 | { | |
130 | if (unlikely(nt->dev->type != ARPHRD_NETLINK)) | |
131 | return -EINVAL; | |
132 | ||
133 | spin_lock(&netlink_tap_lock); | |
134 | list_add_rcu(&nt->list, &netlink_tap_all); | |
135 | spin_unlock(&netlink_tap_lock); | |
136 | ||
fcd4d35e | 137 | __module_get(nt->module); |
bcbde0d4 DB |
138 | |
139 | return 0; | |
140 | } | |
141 | EXPORT_SYMBOL_GPL(netlink_add_tap); | |
142 | ||
2173f8d9 | 143 | static int __netlink_remove_tap(struct netlink_tap *nt) |
bcbde0d4 DB |
144 | { |
145 | bool found = false; | |
146 | struct netlink_tap *tmp; | |
147 | ||
148 | spin_lock(&netlink_tap_lock); | |
149 | ||
150 | list_for_each_entry(tmp, &netlink_tap_all, list) { | |
151 | if (nt == tmp) { | |
152 | list_del_rcu(&nt->list); | |
153 | found = true; | |
154 | goto out; | |
155 | } | |
156 | } | |
157 | ||
158 | pr_warn("__netlink_remove_tap: %p not found\n", nt); | |
159 | out: | |
160 | spin_unlock(&netlink_tap_lock); | |
161 | ||
92b80eb3 | 162 | if (found) |
bcbde0d4 DB |
163 | module_put(nt->module); |
164 | ||
165 | return found ? 0 : -ENODEV; | |
166 | } | |
bcbde0d4 DB |
167 | |
168 | int netlink_remove_tap(struct netlink_tap *nt) | |
169 | { | |
170 | int ret; | |
171 | ||
172 | ret = __netlink_remove_tap(nt); | |
173 | synchronize_net(); | |
174 | ||
175 | return ret; | |
176 | } | |
177 | EXPORT_SYMBOL_GPL(netlink_remove_tap); | |
178 | ||
5ffd5cdd DB |
179 | static bool netlink_filter_tap(const struct sk_buff *skb) |
180 | { | |
181 | struct sock *sk = skb->sk; | |
5ffd5cdd DB |
182 | |
183 | /* We take the more conservative approach and | |
184 | * whitelist socket protocols that may pass. | |
185 | */ | |
186 | switch (sk->sk_protocol) { | |
187 | case NETLINK_ROUTE: | |
188 | case NETLINK_USERSOCK: | |
189 | case NETLINK_SOCK_DIAG: | |
190 | case NETLINK_NFLOG: | |
191 | case NETLINK_XFRM: | |
192 | case NETLINK_FIB_LOOKUP: | |
193 | case NETLINK_NETFILTER: | |
194 | case NETLINK_GENERIC: | |
498044bb | 195 | return true; |
5ffd5cdd DB |
196 | } |
197 | ||
498044bb | 198 | return false; |
5ffd5cdd DB |
199 | } |
200 | ||
bcbde0d4 DB |
201 | static int __netlink_deliver_tap_skb(struct sk_buff *skb, |
202 | struct net_device *dev) | |
203 | { | |
204 | struct sk_buff *nskb; | |
5ffd5cdd | 205 | struct sock *sk = skb->sk; |
bcbde0d4 DB |
206 | int ret = -ENOMEM; |
207 | ||
208 | dev_hold(dev); | |
209 | nskb = skb_clone(skb, GFP_ATOMIC); | |
210 | if (nskb) { | |
211 | nskb->dev = dev; | |
5ffd5cdd | 212 | nskb->protocol = htons((u16) sk->sk_protocol); |
604d13c9 DB |
213 | nskb->pkt_type = netlink_is_kernel(sk) ? |
214 | PACKET_KERNEL : PACKET_USER; | |
4e48ed88 | 215 | skb_reset_network_header(nskb); |
bcbde0d4 DB |
216 | ret = dev_queue_xmit(nskb); |
217 | if (unlikely(ret > 0)) | |
218 | ret = net_xmit_errno(ret); | |
219 | } | |
220 | ||
221 | dev_put(dev); | |
222 | return ret; | |
223 | } | |
224 | ||
225 | static void __netlink_deliver_tap(struct sk_buff *skb) | |
226 | { | |
227 | int ret; | |
228 | struct netlink_tap *tmp; | |
229 | ||
5ffd5cdd DB |
230 | if (!netlink_filter_tap(skb)) |
231 | return; | |
232 | ||
bcbde0d4 DB |
233 | list_for_each_entry_rcu(tmp, &netlink_tap_all, list) { |
234 | ret = __netlink_deliver_tap_skb(skb, tmp->dev); | |
235 | if (unlikely(ret)) | |
236 | break; | |
237 | } | |
238 | } | |
239 | ||
240 | static void netlink_deliver_tap(struct sk_buff *skb) | |
241 | { | |
242 | rcu_read_lock(); | |
243 | ||
244 | if (unlikely(!list_empty(&netlink_tap_all))) | |
245 | __netlink_deliver_tap(skb); | |
246 | ||
247 | rcu_read_unlock(); | |
248 | } | |
249 | ||
73bfd370 DB |
250 | static void netlink_deliver_tap_kernel(struct sock *dst, struct sock *src, |
251 | struct sk_buff *skb) | |
252 | { | |
253 | if (!(netlink_is_kernel(dst) && netlink_is_kernel(src))) | |
254 | netlink_deliver_tap(skb); | |
255 | } | |
256 | ||
cd1df525 PM |
257 | static void netlink_overrun(struct sock *sk) |
258 | { | |
259 | struct netlink_sock *nlk = nlk_sk(sk); | |
260 | ||
cc3a572f ND |
261 | if (!(nlk->flags & NETLINK_F_RECV_NO_ENOBUFS)) { |
262 | if (!test_and_set_bit(NETLINK_S_CONGESTED, | |
263 | &nlk_sk(sk)->state)) { | |
cd1df525 PM |
264 | sk->sk_err = ENOBUFS; |
265 | sk->sk_error_report(sk); | |
266 | } | |
267 | } | |
268 | atomic_inc(&sk->sk_drops); | |
269 | } | |
270 | ||
271 | static void netlink_rcv_wake(struct sock *sk) | |
272 | { | |
273 | struct netlink_sock *nlk = nlk_sk(sk); | |
274 | ||
275 | if (skb_queue_empty(&sk->sk_receive_queue)) | |
cc3a572f ND |
276 | clear_bit(NETLINK_S_CONGESTED, &nlk->state); |
277 | if (!test_bit(NETLINK_S_CONGESTED, &nlk->state)) | |
cd1df525 PM |
278 | wake_up_interruptible(&nlk->wait); |
279 | } | |
280 | ||
ccdfcc39 | 281 | #ifdef CONFIG_NETLINK_MMAP |
9652e931 PM |
282 | static bool netlink_skb_is_mmaped(const struct sk_buff *skb) |
283 | { | |
284 | return NETLINK_CB(skb).flags & NETLINK_SKB_MMAPED; | |
285 | } | |
286 | ||
f9c22888 PM |
287 | static bool netlink_rx_is_mmaped(struct sock *sk) |
288 | { | |
289 | return nlk_sk(sk)->rx_ring.pg_vec != NULL; | |
290 | } | |
291 | ||
5fd96123 PM |
292 | static bool netlink_tx_is_mmaped(struct sock *sk) |
293 | { | |
294 | return nlk_sk(sk)->tx_ring.pg_vec != NULL; | |
295 | } | |
296 | ||
ccdfcc39 PM |
297 | static __pure struct page *pgvec_to_page(const void *addr) |
298 | { | |
299 | if (is_vmalloc_addr(addr)) | |
300 | return vmalloc_to_page(addr); | |
301 | else | |
302 | return virt_to_page(addr); | |
303 | } | |
304 | ||
305 | static void free_pg_vec(void **pg_vec, unsigned int order, unsigned int len) | |
306 | { | |
307 | unsigned int i; | |
308 | ||
309 | for (i = 0; i < len; i++) { | |
310 | if (pg_vec[i] != NULL) { | |
311 | if (is_vmalloc_addr(pg_vec[i])) | |
312 | vfree(pg_vec[i]); | |
313 | else | |
314 | free_pages((unsigned long)pg_vec[i], order); | |
315 | } | |
316 | } | |
317 | kfree(pg_vec); | |
318 | } | |
319 | ||
320 | static void *alloc_one_pg_vec_page(unsigned long order) | |
321 | { | |
322 | void *buffer; | |
323 | gfp_t gfp_flags = GFP_KERNEL | __GFP_COMP | __GFP_ZERO | | |
324 | __GFP_NOWARN | __GFP_NORETRY; | |
325 | ||
326 | buffer = (void *)__get_free_pages(gfp_flags, order); | |
327 | if (buffer != NULL) | |
328 | return buffer; | |
329 | ||
330 | buffer = vzalloc((1 << order) * PAGE_SIZE); | |
331 | if (buffer != NULL) | |
332 | return buffer; | |
333 | ||
334 | gfp_flags &= ~__GFP_NORETRY; | |
335 | return (void *)__get_free_pages(gfp_flags, order); | |
336 | } | |
337 | ||
338 | static void **alloc_pg_vec(struct netlink_sock *nlk, | |
339 | struct nl_mmap_req *req, unsigned int order) | |
340 | { | |
341 | unsigned int block_nr = req->nm_block_nr; | |
342 | unsigned int i; | |
8a849bb7 | 343 | void **pg_vec; |
ccdfcc39 PM |
344 | |
345 | pg_vec = kcalloc(block_nr, sizeof(void *), GFP_KERNEL); | |
346 | if (pg_vec == NULL) | |
347 | return NULL; | |
348 | ||
349 | for (i = 0; i < block_nr; i++) { | |
8a849bb7 | 350 | pg_vec[i] = alloc_one_pg_vec_page(order); |
ccdfcc39 PM |
351 | if (pg_vec[i] == NULL) |
352 | goto err1; | |
353 | } | |
354 | ||
355 | return pg_vec; | |
356 | err1: | |
357 | free_pg_vec(pg_vec, order, block_nr); | |
358 | return NULL; | |
359 | } | |
360 | ||
0470eb99 FW |
361 | |
362 | static void | |
363 | __netlink_set_ring(struct sock *sk, struct nl_mmap_req *req, bool tx_ring, void **pg_vec, | |
364 | unsigned int order) | |
365 | { | |
366 | struct netlink_sock *nlk = nlk_sk(sk); | |
367 | struct sk_buff_head *queue; | |
368 | struct netlink_ring *ring; | |
369 | ||
370 | queue = tx_ring ? &sk->sk_write_queue : &sk->sk_receive_queue; | |
371 | ring = tx_ring ? &nlk->tx_ring : &nlk->rx_ring; | |
372 | ||
373 | spin_lock_bh(&queue->lock); | |
374 | ||
375 | ring->frame_max = req->nm_frame_nr - 1; | |
376 | ring->head = 0; | |
377 | ring->frame_size = req->nm_frame_size; | |
378 | ring->pg_vec_pages = req->nm_block_size / PAGE_SIZE; | |
379 | ||
380 | swap(ring->pg_vec_len, req->nm_block_nr); | |
381 | swap(ring->pg_vec_order, order); | |
382 | swap(ring->pg_vec, pg_vec); | |
383 | ||
384 | __skb_queue_purge(queue); | |
385 | spin_unlock_bh(&queue->lock); | |
386 | ||
387 | WARN_ON(atomic_read(&nlk->mapped)); | |
388 | ||
389 | if (pg_vec) | |
390 | free_pg_vec(pg_vec, order, req->nm_block_nr); | |
391 | } | |
392 | ||
ccdfcc39 | 393 | static int netlink_set_ring(struct sock *sk, struct nl_mmap_req *req, |
0470eb99 | 394 | bool tx_ring) |
ccdfcc39 PM |
395 | { |
396 | struct netlink_sock *nlk = nlk_sk(sk); | |
397 | struct netlink_ring *ring; | |
ccdfcc39 PM |
398 | void **pg_vec = NULL; |
399 | unsigned int order = 0; | |
ccdfcc39 PM |
400 | |
401 | ring = tx_ring ? &nlk->tx_ring : &nlk->rx_ring; | |
ccdfcc39 | 402 | |
0470eb99 FW |
403 | if (atomic_read(&nlk->mapped)) |
404 | return -EBUSY; | |
405 | if (atomic_read(&ring->pending)) | |
406 | return -EBUSY; | |
ccdfcc39 PM |
407 | |
408 | if (req->nm_block_nr) { | |
409 | if (ring->pg_vec != NULL) | |
410 | return -EBUSY; | |
411 | ||
412 | if ((int)req->nm_block_size <= 0) | |
413 | return -EINVAL; | |
74e83b23 | 414 | if (!PAGE_ALIGNED(req->nm_block_size)) |
ccdfcc39 PM |
415 | return -EINVAL; |
416 | if (req->nm_frame_size < NL_MMAP_HDRLEN) | |
417 | return -EINVAL; | |
418 | if (!IS_ALIGNED(req->nm_frame_size, NL_MMAP_MSG_ALIGNMENT)) | |
419 | return -EINVAL; | |
420 | ||
421 | ring->frames_per_block = req->nm_block_size / | |
422 | req->nm_frame_size; | |
423 | if (ring->frames_per_block == 0) | |
424 | return -EINVAL; | |
425 | if (ring->frames_per_block * req->nm_block_nr != | |
426 | req->nm_frame_nr) | |
427 | return -EINVAL; | |
428 | ||
429 | order = get_order(req->nm_block_size); | |
430 | pg_vec = alloc_pg_vec(nlk, req, order); | |
431 | if (pg_vec == NULL) | |
432 | return -ENOMEM; | |
433 | } else { | |
434 | if (req->nm_frame_nr) | |
435 | return -EINVAL; | |
436 | } | |
437 | ||
ccdfcc39 | 438 | mutex_lock(&nlk->pg_vec_lock); |
0470eb99 FW |
439 | if (atomic_read(&nlk->mapped) == 0) { |
440 | __netlink_set_ring(sk, req, tx_ring, pg_vec, order); | |
441 | mutex_unlock(&nlk->pg_vec_lock); | |
442 | return 0; | |
ccdfcc39 | 443 | } |
0470eb99 | 444 | |
ccdfcc39 PM |
445 | mutex_unlock(&nlk->pg_vec_lock); |
446 | ||
447 | if (pg_vec) | |
448 | free_pg_vec(pg_vec, order, req->nm_block_nr); | |
0470eb99 FW |
449 | |
450 | return -EBUSY; | |
ccdfcc39 PM |
451 | } |
452 | ||
453 | static void netlink_mm_open(struct vm_area_struct *vma) | |
454 | { | |
455 | struct file *file = vma->vm_file; | |
456 | struct socket *sock = file->private_data; | |
457 | struct sock *sk = sock->sk; | |
458 | ||
459 | if (sk) | |
460 | atomic_inc(&nlk_sk(sk)->mapped); | |
461 | } | |
462 | ||
463 | static void netlink_mm_close(struct vm_area_struct *vma) | |
464 | { | |
465 | struct file *file = vma->vm_file; | |
466 | struct socket *sock = file->private_data; | |
467 | struct sock *sk = sock->sk; | |
468 | ||
469 | if (sk) | |
470 | atomic_dec(&nlk_sk(sk)->mapped); | |
471 | } | |
472 | ||
473 | static const struct vm_operations_struct netlink_mmap_ops = { | |
474 | .open = netlink_mm_open, | |
475 | .close = netlink_mm_close, | |
476 | }; | |
477 | ||
478 | static int netlink_mmap(struct file *file, struct socket *sock, | |
479 | struct vm_area_struct *vma) | |
480 | { | |
481 | struct sock *sk = sock->sk; | |
482 | struct netlink_sock *nlk = nlk_sk(sk); | |
483 | struct netlink_ring *ring; | |
484 | unsigned long start, size, expected; | |
485 | unsigned int i; | |
486 | int err = -EINVAL; | |
487 | ||
488 | if (vma->vm_pgoff) | |
489 | return -EINVAL; | |
490 | ||
491 | mutex_lock(&nlk->pg_vec_lock); | |
492 | ||
493 | expected = 0; | |
494 | for (ring = &nlk->rx_ring; ring <= &nlk->tx_ring; ring++) { | |
495 | if (ring->pg_vec == NULL) | |
496 | continue; | |
497 | expected += ring->pg_vec_len * ring->pg_vec_pages * PAGE_SIZE; | |
498 | } | |
499 | ||
500 | if (expected == 0) | |
501 | goto out; | |
502 | ||
503 | size = vma->vm_end - vma->vm_start; | |
504 | if (size != expected) | |
505 | goto out; | |
506 | ||
507 | start = vma->vm_start; | |
508 | for (ring = &nlk->rx_ring; ring <= &nlk->tx_ring; ring++) { | |
509 | if (ring->pg_vec == NULL) | |
510 | continue; | |
511 | ||
512 | for (i = 0; i < ring->pg_vec_len; i++) { | |
513 | struct page *page; | |
514 | void *kaddr = ring->pg_vec[i]; | |
515 | unsigned int pg_num; | |
516 | ||
517 | for (pg_num = 0; pg_num < ring->pg_vec_pages; pg_num++) { | |
518 | page = pgvec_to_page(kaddr); | |
519 | err = vm_insert_page(vma, start, page); | |
520 | if (err < 0) | |
521 | goto out; | |
522 | start += PAGE_SIZE; | |
523 | kaddr += PAGE_SIZE; | |
524 | } | |
525 | } | |
526 | } | |
527 | ||
528 | atomic_inc(&nlk->mapped); | |
529 | vma->vm_ops = &netlink_mmap_ops; | |
530 | err = 0; | |
531 | out: | |
532 | mutex_unlock(&nlk->pg_vec_lock); | |
7cdbac71 | 533 | return err; |
ccdfcc39 | 534 | } |
9652e931 | 535 | |
4682a035 | 536 | static void netlink_frame_flush_dcache(const struct nl_mmap_hdr *hdr, unsigned int nm_len) |
9652e931 PM |
537 | { |
538 | #if ARCH_IMPLEMENTS_FLUSH_DCACHE_PAGE == 1 | |
539 | struct page *p_start, *p_end; | |
540 | ||
541 | /* First page is flushed through netlink_{get,set}_status */ | |
542 | p_start = pgvec_to_page(hdr + PAGE_SIZE); | |
4682a035 | 543 | p_end = pgvec_to_page((void *)hdr + NL_MMAP_HDRLEN + nm_len - 1); |
9652e931 PM |
544 | while (p_start <= p_end) { |
545 | flush_dcache_page(p_start); | |
546 | p_start++; | |
547 | } | |
548 | #endif | |
549 | } | |
550 | ||
551 | static enum nl_mmap_status netlink_get_status(const struct nl_mmap_hdr *hdr) | |
552 | { | |
553 | smp_rmb(); | |
554 | flush_dcache_page(pgvec_to_page(hdr)); | |
555 | return hdr->nm_status; | |
556 | } | |
557 | ||
558 | static void netlink_set_status(struct nl_mmap_hdr *hdr, | |
559 | enum nl_mmap_status status) | |
560 | { | |
a18e6a18 | 561 | smp_mb(); |
9652e931 PM |
562 | hdr->nm_status = status; |
563 | flush_dcache_page(pgvec_to_page(hdr)); | |
9652e931 PM |
564 | } |
565 | ||
566 | static struct nl_mmap_hdr * | |
567 | __netlink_lookup_frame(const struct netlink_ring *ring, unsigned int pos) | |
568 | { | |
569 | unsigned int pg_vec_pos, frame_off; | |
570 | ||
571 | pg_vec_pos = pos / ring->frames_per_block; | |
572 | frame_off = pos % ring->frames_per_block; | |
573 | ||
574 | return ring->pg_vec[pg_vec_pos] + (frame_off * ring->frame_size); | |
575 | } | |
576 | ||
577 | static struct nl_mmap_hdr * | |
578 | netlink_lookup_frame(const struct netlink_ring *ring, unsigned int pos, | |
579 | enum nl_mmap_status status) | |
580 | { | |
581 | struct nl_mmap_hdr *hdr; | |
582 | ||
583 | hdr = __netlink_lookup_frame(ring, pos); | |
584 | if (netlink_get_status(hdr) != status) | |
585 | return NULL; | |
586 | ||
587 | return hdr; | |
588 | } | |
589 | ||
590 | static struct nl_mmap_hdr * | |
591 | netlink_current_frame(const struct netlink_ring *ring, | |
592 | enum nl_mmap_status status) | |
593 | { | |
594 | return netlink_lookup_frame(ring, ring->head, status); | |
595 | } | |
596 | ||
597 | static struct nl_mmap_hdr * | |
598 | netlink_previous_frame(const struct netlink_ring *ring, | |
599 | enum nl_mmap_status status) | |
600 | { | |
601 | unsigned int prev; | |
602 | ||
603 | prev = ring->head ? ring->head - 1 : ring->frame_max; | |
604 | return netlink_lookup_frame(ring, prev, status); | |
605 | } | |
606 | ||
607 | static void netlink_increment_head(struct netlink_ring *ring) | |
608 | { | |
609 | ring->head = ring->head != ring->frame_max ? ring->head + 1 : 0; | |
610 | } | |
611 | ||
612 | static void netlink_forward_ring(struct netlink_ring *ring) | |
613 | { | |
614 | unsigned int head = ring->head, pos = head; | |
615 | const struct nl_mmap_hdr *hdr; | |
616 | ||
617 | do { | |
618 | hdr = __netlink_lookup_frame(ring, pos); | |
619 | if (hdr->nm_status == NL_MMAP_STATUS_UNUSED) | |
620 | break; | |
621 | if (hdr->nm_status != NL_MMAP_STATUS_SKIP) | |
622 | break; | |
623 | netlink_increment_head(ring); | |
624 | } while (ring->head != head); | |
625 | } | |
626 | ||
cd1df525 PM |
627 | static bool netlink_dump_space(struct netlink_sock *nlk) |
628 | { | |
629 | struct netlink_ring *ring = &nlk->rx_ring; | |
630 | struct nl_mmap_hdr *hdr; | |
631 | unsigned int n; | |
632 | ||
633 | hdr = netlink_current_frame(ring, NL_MMAP_STATUS_UNUSED); | |
634 | if (hdr == NULL) | |
635 | return false; | |
636 | ||
637 | n = ring->head + ring->frame_max / 2; | |
638 | if (n > ring->frame_max) | |
639 | n -= ring->frame_max; | |
640 | ||
641 | hdr = __netlink_lookup_frame(ring, n); | |
642 | ||
643 | return hdr->nm_status == NL_MMAP_STATUS_UNUSED; | |
644 | } | |
645 | ||
9652e931 PM |
646 | static unsigned int netlink_poll(struct file *file, struct socket *sock, |
647 | poll_table *wait) | |
648 | { | |
649 | struct sock *sk = sock->sk; | |
650 | struct netlink_sock *nlk = nlk_sk(sk); | |
651 | unsigned int mask; | |
cd1df525 | 652 | int err; |
9652e931 | 653 | |
cd1df525 PM |
654 | if (nlk->rx_ring.pg_vec != NULL) { |
655 | /* Memory mapped sockets don't call recvmsg(), so flow control | |
656 | * for dumps is performed here. A dump is allowed to continue | |
657 | * if at least half the ring is unused. | |
658 | */ | |
16b304f3 | 659 | while (nlk->cb_running && netlink_dump_space(nlk)) { |
cd1df525 PM |
660 | err = netlink_dump(sk); |
661 | if (err < 0) { | |
ac30ef83 | 662 | sk->sk_err = -err; |
cd1df525 PM |
663 | sk->sk_error_report(sk); |
664 | break; | |
665 | } | |
666 | } | |
667 | netlink_rcv_wake(sk); | |
668 | } | |
5fd96123 | 669 | |
9652e931 PM |
670 | mask = datagram_poll(file, sock, wait); |
671 | ||
672 | spin_lock_bh(&sk->sk_receive_queue.lock); | |
673 | if (nlk->rx_ring.pg_vec) { | |
674 | netlink_forward_ring(&nlk->rx_ring); | |
675 | if (!netlink_previous_frame(&nlk->rx_ring, NL_MMAP_STATUS_UNUSED)) | |
676 | mask |= POLLIN | POLLRDNORM; | |
677 | } | |
678 | spin_unlock_bh(&sk->sk_receive_queue.lock); | |
679 | ||
680 | spin_lock_bh(&sk->sk_write_queue.lock); | |
681 | if (nlk->tx_ring.pg_vec) { | |
682 | if (netlink_current_frame(&nlk->tx_ring, NL_MMAP_STATUS_UNUSED)) | |
683 | mask |= POLLOUT | POLLWRNORM; | |
684 | } | |
685 | spin_unlock_bh(&sk->sk_write_queue.lock); | |
686 | ||
687 | return mask; | |
688 | } | |
689 | ||
690 | static struct nl_mmap_hdr *netlink_mmap_hdr(struct sk_buff *skb) | |
691 | { | |
692 | return (struct nl_mmap_hdr *)(skb->head - NL_MMAP_HDRLEN); | |
693 | } | |
694 | ||
695 | static void netlink_ring_setup_skb(struct sk_buff *skb, struct sock *sk, | |
696 | struct netlink_ring *ring, | |
697 | struct nl_mmap_hdr *hdr) | |
698 | { | |
699 | unsigned int size; | |
700 | void *data; | |
701 | ||
702 | size = ring->frame_size - NL_MMAP_HDRLEN; | |
703 | data = (void *)hdr + NL_MMAP_HDRLEN; | |
704 | ||
705 | skb->head = data; | |
706 | skb->data = data; | |
707 | skb_reset_tail_pointer(skb); | |
708 | skb->end = skb->tail + size; | |
709 | skb->len = 0; | |
710 | ||
711 | skb->destructor = netlink_skb_destructor; | |
712 | NETLINK_CB(skb).flags |= NETLINK_SKB_MMAPED; | |
713 | NETLINK_CB(skb).sk = sk; | |
714 | } | |
5fd96123 PM |
715 | |
716 | static int netlink_mmap_sendmsg(struct sock *sk, struct msghdr *msg, | |
717 | u32 dst_portid, u32 dst_group, | |
7cc05662 | 718 | struct scm_cookie *scm) |
5fd96123 PM |
719 | { |
720 | struct netlink_sock *nlk = nlk_sk(sk); | |
721 | struct netlink_ring *ring; | |
722 | struct nl_mmap_hdr *hdr; | |
723 | struct sk_buff *skb; | |
724 | unsigned int maxlen; | |
5fd96123 PM |
725 | int err = 0, len = 0; |
726 | ||
5fd96123 PM |
727 | mutex_lock(&nlk->pg_vec_lock); |
728 | ||
729 | ring = &nlk->tx_ring; | |
730 | maxlen = ring->frame_size - NL_MMAP_HDRLEN; | |
731 | ||
732 | do { | |
4682a035 DM |
733 | unsigned int nm_len; |
734 | ||
5fd96123 PM |
735 | hdr = netlink_current_frame(ring, NL_MMAP_STATUS_VALID); |
736 | if (hdr == NULL) { | |
737 | if (!(msg->msg_flags & MSG_DONTWAIT) && | |
738 | atomic_read(&nlk->tx_ring.pending)) | |
739 | schedule(); | |
740 | continue; | |
741 | } | |
4682a035 DM |
742 | |
743 | nm_len = ACCESS_ONCE(hdr->nm_len); | |
744 | if (nm_len > maxlen) { | |
5fd96123 PM |
745 | err = -EINVAL; |
746 | goto out; | |
747 | } | |
748 | ||
4682a035 | 749 | netlink_frame_flush_dcache(hdr, nm_len); |
5fd96123 | 750 | |
4682a035 DM |
751 | skb = alloc_skb(nm_len, GFP_KERNEL); |
752 | if (skb == NULL) { | |
753 | err = -ENOBUFS; | |
754 | goto out; | |
5fd96123 | 755 | } |
4682a035 DM |
756 | __skb_put(skb, nm_len); |
757 | memcpy(skb->data, (void *)hdr + NL_MMAP_HDRLEN, nm_len); | |
758 | netlink_set_status(hdr, NL_MMAP_STATUS_UNUSED); | |
5fd96123 PM |
759 | |
760 | netlink_increment_head(ring); | |
761 | ||
762 | NETLINK_CB(skb).portid = nlk->portid; | |
763 | NETLINK_CB(skb).dst_group = dst_group; | |
7cc05662 | 764 | NETLINK_CB(skb).creds = scm->creds; |
5fd96123 PM |
765 | |
766 | err = security_netlink_send(sk, skb); | |
767 | if (err) { | |
768 | kfree_skb(skb); | |
769 | goto out; | |
770 | } | |
771 | ||
772 | if (unlikely(dst_group)) { | |
773 | atomic_inc(&skb->users); | |
774 | netlink_broadcast(sk, skb, dst_portid, dst_group, | |
775 | GFP_KERNEL); | |
776 | } | |
777 | err = netlink_unicast(sk, skb, dst_portid, | |
778 | msg->msg_flags & MSG_DONTWAIT); | |
779 | if (err < 0) | |
780 | goto out; | |
781 | len += err; | |
782 | ||
783 | } while (hdr != NULL || | |
784 | (!(msg->msg_flags & MSG_DONTWAIT) && | |
785 | atomic_read(&nlk->tx_ring.pending))); | |
786 | ||
787 | if (len > 0) | |
788 | err = len; | |
789 | out: | |
790 | mutex_unlock(&nlk->pg_vec_lock); | |
791 | return err; | |
792 | } | |
f9c22888 PM |
793 | |
794 | static void netlink_queue_mmaped_skb(struct sock *sk, struct sk_buff *skb) | |
795 | { | |
796 | struct nl_mmap_hdr *hdr; | |
797 | ||
798 | hdr = netlink_mmap_hdr(skb); | |
799 | hdr->nm_len = skb->len; | |
800 | hdr->nm_group = NETLINK_CB(skb).dst_group; | |
801 | hdr->nm_pid = NETLINK_CB(skb).creds.pid; | |
1bf9310a ND |
802 | hdr->nm_uid = from_kuid(sk_user_ns(sk), NETLINK_CB(skb).creds.uid); |
803 | hdr->nm_gid = from_kgid(sk_user_ns(sk), NETLINK_CB(skb).creds.gid); | |
4682a035 | 804 | netlink_frame_flush_dcache(hdr, hdr->nm_len); |
f9c22888 PM |
805 | netlink_set_status(hdr, NL_MMAP_STATUS_VALID); |
806 | ||
807 | NETLINK_CB(skb).flags |= NETLINK_SKB_DELIVERED; | |
808 | kfree_skb(skb); | |
809 | } | |
810 | ||
811 | static void netlink_ring_set_copied(struct sock *sk, struct sk_buff *skb) | |
812 | { | |
813 | struct netlink_sock *nlk = nlk_sk(sk); | |
814 | struct netlink_ring *ring = &nlk->rx_ring; | |
815 | struct nl_mmap_hdr *hdr; | |
816 | ||
817 | spin_lock_bh(&sk->sk_receive_queue.lock); | |
818 | hdr = netlink_current_frame(ring, NL_MMAP_STATUS_UNUSED); | |
819 | if (hdr == NULL) { | |
820 | spin_unlock_bh(&sk->sk_receive_queue.lock); | |
821 | kfree_skb(skb); | |
cd1df525 | 822 | netlink_overrun(sk); |
f9c22888 PM |
823 | return; |
824 | } | |
825 | netlink_increment_head(ring); | |
826 | __skb_queue_tail(&sk->sk_receive_queue, skb); | |
827 | spin_unlock_bh(&sk->sk_receive_queue.lock); | |
828 | ||
829 | hdr->nm_len = skb->len; | |
830 | hdr->nm_group = NETLINK_CB(skb).dst_group; | |
831 | hdr->nm_pid = NETLINK_CB(skb).creds.pid; | |
1bf9310a ND |
832 | hdr->nm_uid = from_kuid(sk_user_ns(sk), NETLINK_CB(skb).creds.uid); |
833 | hdr->nm_gid = from_kgid(sk_user_ns(sk), NETLINK_CB(skb).creds.gid); | |
f9c22888 PM |
834 | netlink_set_status(hdr, NL_MMAP_STATUS_COPY); |
835 | } | |
836 | ||
ccdfcc39 | 837 | #else /* CONFIG_NETLINK_MMAP */ |
9652e931 | 838 | #define netlink_skb_is_mmaped(skb) false |
f9c22888 | 839 | #define netlink_rx_is_mmaped(sk) false |
5fd96123 | 840 | #define netlink_tx_is_mmaped(sk) false |
ccdfcc39 | 841 | #define netlink_mmap sock_no_mmap |
9652e931 | 842 | #define netlink_poll datagram_poll |
7cc05662 | 843 | #define netlink_mmap_sendmsg(sk, msg, dst_portid, dst_group, scm) 0 |
ccdfcc39 PM |
844 | #endif /* CONFIG_NETLINK_MMAP */ |
845 | ||
cf0a018a PM |
846 | static void netlink_skb_destructor(struct sk_buff *skb) |
847 | { | |
9652e931 PM |
848 | #ifdef CONFIG_NETLINK_MMAP |
849 | struct nl_mmap_hdr *hdr; | |
850 | struct netlink_ring *ring; | |
851 | struct sock *sk; | |
852 | ||
853 | /* If a packet from the kernel to userspace was freed because of an | |
854 | * error without being delivered to userspace, the kernel must reset | |
855 | * the status. In the direction userspace to kernel, the status is | |
856 | * always reset here after the packet was processed and freed. | |
857 | */ | |
858 | if (netlink_skb_is_mmaped(skb)) { | |
859 | hdr = netlink_mmap_hdr(skb); | |
860 | sk = NETLINK_CB(skb).sk; | |
861 | ||
5fd96123 PM |
862 | if (NETLINK_CB(skb).flags & NETLINK_SKB_TX) { |
863 | netlink_set_status(hdr, NL_MMAP_STATUS_UNUSED); | |
864 | ring = &nlk_sk(sk)->tx_ring; | |
865 | } else { | |
866 | if (!(NETLINK_CB(skb).flags & NETLINK_SKB_DELIVERED)) { | |
867 | hdr->nm_len = 0; | |
868 | netlink_set_status(hdr, NL_MMAP_STATUS_VALID); | |
869 | } | |
870 | ring = &nlk_sk(sk)->rx_ring; | |
9652e931 | 871 | } |
9652e931 PM |
872 | |
873 | WARN_ON(atomic_read(&ring->pending) == 0); | |
874 | atomic_dec(&ring->pending); | |
875 | sock_put(sk); | |
876 | ||
5e71d9d7 | 877 | skb->head = NULL; |
9652e931 PM |
878 | } |
879 | #endif | |
c05cdb1b | 880 | if (is_vmalloc_addr(skb->head)) { |
3a36515f PN |
881 | if (!skb->cloned || |
882 | !atomic_dec_return(&(skb_shinfo(skb)->dataref))) | |
883 | vfree(skb->head); | |
884 | ||
c05cdb1b PNA |
885 | skb->head = NULL; |
886 | } | |
9652e931 PM |
887 | if (skb->sk != NULL) |
888 | sock_rfree(skb); | |
cf0a018a PM |
889 | } |
890 | ||
891 | static void netlink_skb_set_owner_r(struct sk_buff *skb, struct sock *sk) | |
892 | { | |
893 | WARN_ON(skb->sk != NULL); | |
894 | skb->sk = sk; | |
895 | skb->destructor = netlink_skb_destructor; | |
896 | atomic_add(skb->truesize, &sk->sk_rmem_alloc); | |
897 | sk_mem_charge(sk, skb->truesize); | |
898 | } | |
899 | ||
1da177e4 LT |
900 | static void netlink_sock_destruct(struct sock *sk) |
901 | { | |
3f660d66 HX |
902 | struct netlink_sock *nlk = nlk_sk(sk); |
903 | ||
16b304f3 PS |
904 | if (nlk->cb_running) { |
905 | if (nlk->cb.done) | |
906 | nlk->cb.done(&nlk->cb); | |
6dc878a8 | 907 | |
16b304f3 PS |
908 | module_put(nlk->cb.module); |
909 | kfree_skb(nlk->cb.skb); | |
3f660d66 HX |
910 | } |
911 | ||
1da177e4 | 912 | skb_queue_purge(&sk->sk_receive_queue); |
ccdfcc39 PM |
913 | #ifdef CONFIG_NETLINK_MMAP |
914 | if (1) { | |
915 | struct nl_mmap_req req; | |
916 | ||
917 | memset(&req, 0, sizeof(req)); | |
918 | if (nlk->rx_ring.pg_vec) | |
0470eb99 | 919 | __netlink_set_ring(sk, &req, false, NULL, 0); |
ccdfcc39 PM |
920 | memset(&req, 0, sizeof(req)); |
921 | if (nlk->tx_ring.pg_vec) | |
0470eb99 | 922 | __netlink_set_ring(sk, &req, true, NULL, 0); |
ccdfcc39 PM |
923 | } |
924 | #endif /* CONFIG_NETLINK_MMAP */ | |
1da177e4 LT |
925 | |
926 | if (!sock_flag(sk, SOCK_DEAD)) { | |
6ac552fd | 927 | printk(KERN_ERR "Freeing alive netlink socket %p\n", sk); |
1da177e4 LT |
928 | return; |
929 | } | |
547b792c IJ |
930 | |
931 | WARN_ON(atomic_read(&sk->sk_rmem_alloc)); | |
932 | WARN_ON(atomic_read(&sk->sk_wmem_alloc)); | |
933 | WARN_ON(nlk_sk(sk)->groups); | |
1da177e4 LT |
934 | } |
935 | ||
6ac552fd PM |
936 | /* This lock without WQ_FLAG_EXCLUSIVE is good on UP and it is _very_ bad on |
937 | * SMP. Look, when several writers sleep and reader wakes them up, all but one | |
1da177e4 LT |
938 | * immediately hit write lock and grab all the cpus. Exclusive sleep solves |
939 | * this, _but_ remember, it adds useless work on UP machines. | |
940 | */ | |
941 | ||
d136f1bd | 942 | void netlink_table_grab(void) |
9a429c49 | 943 | __acquires(nl_table_lock) |
1da177e4 | 944 | { |
d136f1bd JB |
945 | might_sleep(); |
946 | ||
6abd219c | 947 | write_lock_irq(&nl_table_lock); |
1da177e4 LT |
948 | |
949 | if (atomic_read(&nl_table_users)) { | |
950 | DECLARE_WAITQUEUE(wait, current); | |
951 | ||
952 | add_wait_queue_exclusive(&nl_table_wait, &wait); | |
6ac552fd | 953 | for (;;) { |
1da177e4 LT |
954 | set_current_state(TASK_UNINTERRUPTIBLE); |
955 | if (atomic_read(&nl_table_users) == 0) | |
956 | break; | |
6abd219c | 957 | write_unlock_irq(&nl_table_lock); |
1da177e4 | 958 | schedule(); |
6abd219c | 959 | write_lock_irq(&nl_table_lock); |
1da177e4 LT |
960 | } |
961 | ||
962 | __set_current_state(TASK_RUNNING); | |
963 | remove_wait_queue(&nl_table_wait, &wait); | |
964 | } | |
965 | } | |
966 | ||
d136f1bd | 967 | void netlink_table_ungrab(void) |
9a429c49 | 968 | __releases(nl_table_lock) |
1da177e4 | 969 | { |
6abd219c | 970 | write_unlock_irq(&nl_table_lock); |
1da177e4 LT |
971 | wake_up(&nl_table_wait); |
972 | } | |
973 | ||
6ac552fd | 974 | static inline void |
1da177e4 LT |
975 | netlink_lock_table(void) |
976 | { | |
977 | /* read_lock() synchronizes us to netlink_table_grab */ | |
978 | ||
979 | read_lock(&nl_table_lock); | |
980 | atomic_inc(&nl_table_users); | |
981 | read_unlock(&nl_table_lock); | |
982 | } | |
983 | ||
6ac552fd | 984 | static inline void |
1da177e4 LT |
985 | netlink_unlock_table(void) |
986 | { | |
987 | if (atomic_dec_and_test(&nl_table_users)) | |
988 | wake_up(&nl_table_wait); | |
989 | } | |
990 | ||
e341694e | 991 | struct netlink_compare_arg |
1da177e4 | 992 | { |
c428ecd1 | 993 | possible_net_t pnet; |
e341694e TG |
994 | u32 portid; |
995 | }; | |
1da177e4 | 996 | |
8f2ddaac HX |
997 | /* Doing sizeof directly may yield 4 extra bytes on 64-bit. */ |
998 | #define netlink_compare_arg_len \ | |
999 | (offsetof(struct netlink_compare_arg, portid) + sizeof(u32)) | |
c428ecd1 HX |
1000 | |
1001 | static inline int netlink_compare(struct rhashtable_compare_arg *arg, | |
1002 | const void *ptr) | |
1da177e4 | 1003 | { |
c428ecd1 HX |
1004 | const struct netlink_compare_arg *x = arg->key; |
1005 | const struct netlink_sock *nlk = ptr; | |
1da177e4 | 1006 | |
c428ecd1 HX |
1007 | return nlk->portid != x->portid || |
1008 | !net_eq(sock_net(&nlk->sk), read_pnet(&x->pnet)); | |
1009 | } | |
1010 | ||
1011 | static void netlink_compare_arg_init(struct netlink_compare_arg *arg, | |
1012 | struct net *net, u32 portid) | |
1013 | { | |
1014 | memset(arg, 0, sizeof(*arg)); | |
1015 | write_pnet(&arg->pnet, net); | |
1016 | arg->portid = portid; | |
1da177e4 LT |
1017 | } |
1018 | ||
e341694e TG |
1019 | static struct sock *__netlink_lookup(struct netlink_table *table, u32 portid, |
1020 | struct net *net) | |
1da177e4 | 1021 | { |
c428ecd1 | 1022 | struct netlink_compare_arg arg; |
1da177e4 | 1023 | |
c428ecd1 HX |
1024 | netlink_compare_arg_init(&arg, net, portid); |
1025 | return rhashtable_lookup_fast(&table->hash, &arg, | |
1026 | netlink_rhashtable_params); | |
1da177e4 LT |
1027 | } |
1028 | ||
c428ecd1 | 1029 | static int __netlink_insert(struct netlink_table *table, struct sock *sk) |
c5adde94 | 1030 | { |
c428ecd1 | 1031 | struct netlink_compare_arg arg; |
c5adde94 | 1032 | |
c428ecd1 HX |
1033 | netlink_compare_arg_init(&arg, sock_net(sk), nlk_sk(sk)->portid); |
1034 | return rhashtable_lookup_insert_key(&table->hash, &arg, | |
1035 | &nlk_sk(sk)->node, | |
1036 | netlink_rhashtable_params); | |
c5adde94 YX |
1037 | } |
1038 | ||
e341694e | 1039 | static struct sock *netlink_lookup(struct net *net, int protocol, u32 portid) |
1da177e4 | 1040 | { |
e341694e TG |
1041 | struct netlink_table *table = &nl_table[protocol]; |
1042 | struct sock *sk; | |
1da177e4 | 1043 | |
e341694e TG |
1044 | rcu_read_lock(); |
1045 | sk = __netlink_lookup(table, portid, net); | |
1046 | if (sk) | |
1047 | sock_hold(sk); | |
1048 | rcu_read_unlock(); | |
1da177e4 | 1049 | |
e341694e | 1050 | return sk; |
1da177e4 LT |
1051 | } |
1052 | ||
90ddc4f0 | 1053 | static const struct proto_ops netlink_ops; |
1da177e4 | 1054 | |
4277a083 PM |
1055 | static void |
1056 | netlink_update_listeners(struct sock *sk) | |
1057 | { | |
1058 | struct netlink_table *tbl = &nl_table[sk->sk_protocol]; | |
4277a083 PM |
1059 | unsigned long mask; |
1060 | unsigned int i; | |
6d772ac5 ED |
1061 | struct listeners *listeners; |
1062 | ||
1063 | listeners = nl_deref_protected(tbl->listeners); | |
1064 | if (!listeners) | |
1065 | return; | |
4277a083 | 1066 | |
b4ff4f04 | 1067 | for (i = 0; i < NLGRPLONGS(tbl->groups); i++) { |
4277a083 | 1068 | mask = 0; |
b67bfe0d | 1069 | sk_for_each_bound(sk, &tbl->mc_list) { |
b4ff4f04 JB |
1070 | if (i < NLGRPLONGS(nlk_sk(sk)->ngroups)) |
1071 | mask |= nlk_sk(sk)->groups[i]; | |
1072 | } | |
6d772ac5 | 1073 | listeners->masks[i] = mask; |
4277a083 PM |
1074 | } |
1075 | /* this function is only called with the netlink table "grabbed", which | |
1076 | * makes sure updates are visible before bind or setsockopt return. */ | |
1077 | } | |
1078 | ||
8ea65f4a | 1079 | static int netlink_insert(struct sock *sk, u32 portid) |
1da177e4 | 1080 | { |
da12c90e | 1081 | struct netlink_table *table = &nl_table[sk->sk_protocol]; |
919d9db9 | 1082 | int err; |
1da177e4 | 1083 | |
c5adde94 | 1084 | lock_sock(sk); |
1da177e4 LT |
1085 | |
1086 | err = -EBUSY; | |
15e47304 | 1087 | if (nlk_sk(sk)->portid) |
1da177e4 LT |
1088 | goto err; |
1089 | ||
1090 | err = -ENOMEM; | |
97defe1e TG |
1091 | if (BITS_PER_LONG > 32 && |
1092 | unlikely(atomic_read(&table->hash.nelems) >= UINT_MAX)) | |
1da177e4 LT |
1093 | goto err; |
1094 | ||
15e47304 | 1095 | nlk_sk(sk)->portid = portid; |
e341694e | 1096 | sock_hold(sk); |
919d9db9 | 1097 | |
c428ecd1 HX |
1098 | err = __netlink_insert(table, sk); |
1099 | if (err) { | |
4e7c1330 DB |
1100 | /* In case the hashtable backend returns with -EBUSY |
1101 | * from here, it must not escape to the caller. | |
1102 | */ | |
1103 | if (unlikely(err == -EBUSY)) | |
1104 | err = -EOVERFLOW; | |
c428ecd1 HX |
1105 | if (err == -EEXIST) |
1106 | err = -EADDRINUSE; | |
c0bb07df | 1107 | nlk_sk(sk)->portid = 0; |
c5adde94 | 1108 | sock_put(sk); |
919d9db9 HX |
1109 | } |
1110 | ||
1da177e4 | 1111 | err: |
c5adde94 | 1112 | release_sock(sk); |
1da177e4 LT |
1113 | return err; |
1114 | } | |
1115 | ||
1116 | static void netlink_remove(struct sock *sk) | |
1117 | { | |
e341694e TG |
1118 | struct netlink_table *table; |
1119 | ||
e341694e | 1120 | table = &nl_table[sk->sk_protocol]; |
c428ecd1 HX |
1121 | if (!rhashtable_remove_fast(&table->hash, &nlk_sk(sk)->node, |
1122 | netlink_rhashtable_params)) { | |
e341694e TG |
1123 | WARN_ON(atomic_read(&sk->sk_refcnt) == 1); |
1124 | __sock_put(sk); | |
1125 | } | |
e341694e | 1126 | |
1da177e4 | 1127 | netlink_table_grab(); |
b10dcb3b | 1128 | if (nlk_sk(sk)->subscriptions) { |
1da177e4 | 1129 | __sk_del_bind_node(sk); |
b10dcb3b JB |
1130 | netlink_update_listeners(sk); |
1131 | } | |
ee1c2442 JB |
1132 | if (sk->sk_protocol == NETLINK_GENERIC) |
1133 | atomic_inc(&genl_sk_destructing_cnt); | |
1da177e4 LT |
1134 | netlink_table_ungrab(); |
1135 | } | |
1136 | ||
1137 | static struct proto netlink_proto = { | |
1138 | .name = "NETLINK", | |
1139 | .owner = THIS_MODULE, | |
1140 | .obj_size = sizeof(struct netlink_sock), | |
1141 | }; | |
1142 | ||
1b8d7ae4 | 1143 | static int __netlink_create(struct net *net, struct socket *sock, |
11aa9c28 EB |
1144 | struct mutex *cb_mutex, int protocol, |
1145 | int kern) | |
1da177e4 LT |
1146 | { |
1147 | struct sock *sk; | |
1148 | struct netlink_sock *nlk; | |
ab33a171 PM |
1149 | |
1150 | sock->ops = &netlink_ops; | |
1151 | ||
11aa9c28 | 1152 | sk = sk_alloc(net, PF_NETLINK, GFP_KERNEL, &netlink_proto, kern); |
ab33a171 PM |
1153 | if (!sk) |
1154 | return -ENOMEM; | |
1155 | ||
1156 | sock_init_data(sock, sk); | |
1157 | ||
1158 | nlk = nlk_sk(sk); | |
658cb354 | 1159 | if (cb_mutex) { |
ffa4d721 | 1160 | nlk->cb_mutex = cb_mutex; |
658cb354 | 1161 | } else { |
ffa4d721 PM |
1162 | nlk->cb_mutex = &nlk->cb_def_mutex; |
1163 | mutex_init(nlk->cb_mutex); | |
1164 | } | |
ab33a171 | 1165 | init_waitqueue_head(&nlk->wait); |
ccdfcc39 PM |
1166 | #ifdef CONFIG_NETLINK_MMAP |
1167 | mutex_init(&nlk->pg_vec_lock); | |
1168 | #endif | |
ab33a171 PM |
1169 | |
1170 | sk->sk_destruct = netlink_sock_destruct; | |
1171 | sk->sk_protocol = protocol; | |
1172 | return 0; | |
1173 | } | |
1174 | ||
3f378b68 EP |
1175 | static int netlink_create(struct net *net, struct socket *sock, int protocol, |
1176 | int kern) | |
ab33a171 PM |
1177 | { |
1178 | struct module *module = NULL; | |
af65bdfc | 1179 | struct mutex *cb_mutex; |
f7fa9b10 | 1180 | struct netlink_sock *nlk; |
023e2cfa JB |
1181 | int (*bind)(struct net *net, int group); |
1182 | void (*unbind)(struct net *net, int group); | |
ab33a171 | 1183 | int err = 0; |
1da177e4 LT |
1184 | |
1185 | sock->state = SS_UNCONNECTED; | |
1186 | ||
1187 | if (sock->type != SOCK_RAW && sock->type != SOCK_DGRAM) | |
1188 | return -ESOCKTNOSUPPORT; | |
1189 | ||
6ac552fd | 1190 | if (protocol < 0 || protocol >= MAX_LINKS) |
1da177e4 LT |
1191 | return -EPROTONOSUPPORT; |
1192 | ||
77247bbb | 1193 | netlink_lock_table(); |
95a5afca | 1194 | #ifdef CONFIG_MODULES |
ab33a171 | 1195 | if (!nl_table[protocol].registered) { |
77247bbb | 1196 | netlink_unlock_table(); |
4fdb3bb7 | 1197 | request_module("net-pf-%d-proto-%d", PF_NETLINK, protocol); |
77247bbb | 1198 | netlink_lock_table(); |
4fdb3bb7 | 1199 | } |
ab33a171 PM |
1200 | #endif |
1201 | if (nl_table[protocol].registered && | |
1202 | try_module_get(nl_table[protocol].module)) | |
1203 | module = nl_table[protocol].module; | |
974c37e9 AD |
1204 | else |
1205 | err = -EPROTONOSUPPORT; | |
af65bdfc | 1206 | cb_mutex = nl_table[protocol].cb_mutex; |
03292745 | 1207 | bind = nl_table[protocol].bind; |
4f520900 | 1208 | unbind = nl_table[protocol].unbind; |
77247bbb | 1209 | netlink_unlock_table(); |
4fdb3bb7 | 1210 | |
974c37e9 AD |
1211 | if (err < 0) |
1212 | goto out; | |
1213 | ||
11aa9c28 | 1214 | err = __netlink_create(net, sock, cb_mutex, protocol, kern); |
6ac552fd | 1215 | if (err < 0) |
f7fa9b10 PM |
1216 | goto out_module; |
1217 | ||
6f756a8c | 1218 | local_bh_disable(); |
c1fd3b94 | 1219 | sock_prot_inuse_add(net, &netlink_proto, 1); |
6f756a8c DM |
1220 | local_bh_enable(); |
1221 | ||
f7fa9b10 | 1222 | nlk = nlk_sk(sock->sk); |
f7fa9b10 | 1223 | nlk->module = module; |
03292745 | 1224 | nlk->netlink_bind = bind; |
4f520900 | 1225 | nlk->netlink_unbind = unbind; |
ab33a171 PM |
1226 | out: |
1227 | return err; | |
1da177e4 | 1228 | |
ab33a171 PM |
1229 | out_module: |
1230 | module_put(module); | |
1231 | goto out; | |
1da177e4 LT |
1232 | } |
1233 | ||
21e4902a TG |
1234 | static void deferred_put_nlk_sk(struct rcu_head *head) |
1235 | { | |
1236 | struct netlink_sock *nlk = container_of(head, struct netlink_sock, rcu); | |
1237 | ||
1238 | sock_put(&nlk->sk); | |
1239 | } | |
1240 | ||
1da177e4 LT |
1241 | static int netlink_release(struct socket *sock) |
1242 | { | |
1243 | struct sock *sk = sock->sk; | |
1244 | struct netlink_sock *nlk; | |
1245 | ||
1246 | if (!sk) | |
1247 | return 0; | |
1248 | ||
1249 | netlink_remove(sk); | |
ac57b3a9 | 1250 | sock_orphan(sk); |
1da177e4 LT |
1251 | nlk = nlk_sk(sk); |
1252 | ||
3f660d66 HX |
1253 | /* |
1254 | * OK. Socket is unlinked, any packets that arrive now | |
1255 | * will be purged. | |
1256 | */ | |
1da177e4 | 1257 | |
ee1c2442 JB |
1258 | /* must not acquire netlink_table_lock in any way again before unbind |
1259 | * and notifying genetlink is done as otherwise it might deadlock | |
1260 | */ | |
1261 | if (nlk->netlink_unbind) { | |
1262 | int i; | |
1263 | ||
1264 | for (i = 0; i < nlk->ngroups; i++) | |
1265 | if (test_bit(i, nlk->groups)) | |
1266 | nlk->netlink_unbind(sock_net(sk), i + 1); | |
1267 | } | |
1268 | if (sk->sk_protocol == NETLINK_GENERIC && | |
1269 | atomic_dec_return(&genl_sk_destructing_cnt) == 0) | |
1270 | wake_up(&genl_sk_destructing_waitq); | |
1271 | ||
1da177e4 LT |
1272 | sock->sk = NULL; |
1273 | wake_up_interruptible_all(&nlk->wait); | |
1274 | ||
1275 | skb_queue_purge(&sk->sk_write_queue); | |
1276 | ||
15e47304 | 1277 | if (nlk->portid) { |
1da177e4 | 1278 | struct netlink_notify n = { |
3b1e0a65 | 1279 | .net = sock_net(sk), |
1da177e4 | 1280 | .protocol = sk->sk_protocol, |
15e47304 | 1281 | .portid = nlk->portid, |
1da177e4 | 1282 | }; |
e041c683 AS |
1283 | atomic_notifier_call_chain(&netlink_chain, |
1284 | NETLINK_URELEASE, &n); | |
746fac4d | 1285 | } |
4fdb3bb7 | 1286 | |
5e7c001c | 1287 | module_put(nlk->module); |
4fdb3bb7 | 1288 | |
aed81560 | 1289 | if (netlink_is_kernel(sk)) { |
b10dcb3b | 1290 | netlink_table_grab(); |
869e58f8 DL |
1291 | BUG_ON(nl_table[sk->sk_protocol].registered == 0); |
1292 | if (--nl_table[sk->sk_protocol].registered == 0) { | |
6d772ac5 ED |
1293 | struct listeners *old; |
1294 | ||
1295 | old = nl_deref_protected(nl_table[sk->sk_protocol].listeners); | |
1296 | RCU_INIT_POINTER(nl_table[sk->sk_protocol].listeners, NULL); | |
1297 | kfree_rcu(old, rcu); | |
869e58f8 | 1298 | nl_table[sk->sk_protocol].module = NULL; |
9785e10a | 1299 | nl_table[sk->sk_protocol].bind = NULL; |
4f520900 | 1300 | nl_table[sk->sk_protocol].unbind = NULL; |
9785e10a | 1301 | nl_table[sk->sk_protocol].flags = 0; |
869e58f8 DL |
1302 | nl_table[sk->sk_protocol].registered = 0; |
1303 | } | |
b10dcb3b | 1304 | netlink_table_ungrab(); |
658cb354 | 1305 | } |
77247bbb | 1306 | |
f7fa9b10 PM |
1307 | kfree(nlk->groups); |
1308 | nlk->groups = NULL; | |
1309 | ||
3755810c | 1310 | local_bh_disable(); |
c1fd3b94 | 1311 | sock_prot_inuse_add(sock_net(sk), &netlink_proto, -1); |
3755810c | 1312 | local_bh_enable(); |
21e4902a | 1313 | call_rcu(&nlk->rcu, deferred_put_nlk_sk); |
1da177e4 LT |
1314 | return 0; |
1315 | } | |
1316 | ||
1317 | static int netlink_autobind(struct socket *sock) | |
1318 | { | |
1319 | struct sock *sk = sock->sk; | |
3b1e0a65 | 1320 | struct net *net = sock_net(sk); |
da12c90e | 1321 | struct netlink_table *table = &nl_table[sk->sk_protocol]; |
15e47304 | 1322 | s32 portid = task_tgid_vnr(current); |
1da177e4 | 1323 | int err; |
b9fbe709 HX |
1324 | s32 rover = -4096; |
1325 | bool ok; | |
1da177e4 LT |
1326 | |
1327 | retry: | |
1328 | cond_resched(); | |
e341694e | 1329 | rcu_read_lock(); |
b9fbe709 HX |
1330 | ok = !__netlink_lookup(table, portid, net); |
1331 | rcu_read_unlock(); | |
1332 | if (!ok) { | |
e341694e | 1333 | /* Bind collision, search negative portid values. */ |
b9fbe709 HX |
1334 | if (rover == -4096) |
1335 | /* rover will be in range [S32_MIN, -4097] */ | |
1336 | rover = S32_MIN + prandom_u32_max(-4096 - S32_MIN); | |
1337 | else if (rover >= -4096) | |
e341694e | 1338 | rover = -4097; |
b9fbe709 | 1339 | portid = rover--; |
e341694e | 1340 | goto retry; |
1da177e4 | 1341 | } |
1da177e4 | 1342 | |
8ea65f4a | 1343 | err = netlink_insert(sk, portid); |
1da177e4 LT |
1344 | if (err == -EADDRINUSE) |
1345 | goto retry; | |
d470e3b4 DM |
1346 | |
1347 | /* If 2 threads race to autobind, that is fine. */ | |
1348 | if (err == -EBUSY) | |
1349 | err = 0; | |
1350 | ||
1351 | return err; | |
1da177e4 LT |
1352 | } |
1353 | ||
aa4cf945 EB |
1354 | /** |
1355 | * __netlink_ns_capable - General netlink message capability test | |
1356 | * @nsp: NETLINK_CB of the socket buffer holding a netlink command from userspace. | |
1357 | * @user_ns: The user namespace of the capability to use | |
1358 | * @cap: The capability to use | |
1359 | * | |
1360 | * Test to see if the opener of the socket we received the message | |
1361 | * from had when the netlink socket was created and the sender of the | |
1362 | * message has has the capability @cap in the user namespace @user_ns. | |
1363 | */ | |
1364 | bool __netlink_ns_capable(const struct netlink_skb_parms *nsp, | |
1365 | struct user_namespace *user_ns, int cap) | |
1366 | { | |
2d7a85f4 EB |
1367 | return ((nsp->flags & NETLINK_SKB_DST) || |
1368 | file_ns_capable(nsp->sk->sk_socket->file, user_ns, cap)) && | |
1369 | ns_capable(user_ns, cap); | |
aa4cf945 EB |
1370 | } |
1371 | EXPORT_SYMBOL(__netlink_ns_capable); | |
1372 | ||
1373 | /** | |
1374 | * netlink_ns_capable - General netlink message capability test | |
1375 | * @skb: socket buffer holding a netlink command from userspace | |
1376 | * @user_ns: The user namespace of the capability to use | |
1377 | * @cap: The capability to use | |
1378 | * | |
1379 | * Test to see if the opener of the socket we received the message | |
1380 | * from had when the netlink socket was created and the sender of the | |
1381 | * message has has the capability @cap in the user namespace @user_ns. | |
1382 | */ | |
1383 | bool netlink_ns_capable(const struct sk_buff *skb, | |
1384 | struct user_namespace *user_ns, int cap) | |
1385 | { | |
1386 | return __netlink_ns_capable(&NETLINK_CB(skb), user_ns, cap); | |
1387 | } | |
1388 | EXPORT_SYMBOL(netlink_ns_capable); | |
1389 | ||
1390 | /** | |
1391 | * netlink_capable - Netlink global message capability test | |
1392 | * @skb: socket buffer holding a netlink command from userspace | |
1393 | * @cap: The capability to use | |
1394 | * | |
1395 | * Test to see if the opener of the socket we received the message | |
1396 | * from had when the netlink socket was created and the sender of the | |
1397 | * message has has the capability @cap in all user namespaces. | |
1398 | */ | |
1399 | bool netlink_capable(const struct sk_buff *skb, int cap) | |
1400 | { | |
1401 | return netlink_ns_capable(skb, &init_user_ns, cap); | |
1402 | } | |
1403 | EXPORT_SYMBOL(netlink_capable); | |
1404 | ||
1405 | /** | |
1406 | * netlink_net_capable - Netlink network namespace message capability test | |
1407 | * @skb: socket buffer holding a netlink command from userspace | |
1408 | * @cap: The capability to use | |
1409 | * | |
1410 | * Test to see if the opener of the socket we received the message | |
1411 | * from had when the netlink socket was created and the sender of the | |
1412 | * message has has the capability @cap over the network namespace of | |
1413 | * the socket we received the message from. | |
1414 | */ | |
1415 | bool netlink_net_capable(const struct sk_buff *skb, int cap) | |
1416 | { | |
1417 | return netlink_ns_capable(skb, sock_net(skb->sk)->user_ns, cap); | |
1418 | } | |
1419 | EXPORT_SYMBOL(netlink_net_capable); | |
1420 | ||
5187cd05 | 1421 | static inline int netlink_allowed(const struct socket *sock, unsigned int flag) |
746fac4d | 1422 | { |
9785e10a | 1423 | return (nl_table[sock->sk->sk_protocol].flags & flag) || |
df008c91 | 1424 | ns_capable(sock_net(sock->sk)->user_ns, CAP_NET_ADMIN); |
746fac4d | 1425 | } |
1da177e4 | 1426 | |
f7fa9b10 PM |
1427 | static void |
1428 | netlink_update_subscriptions(struct sock *sk, unsigned int subscriptions) | |
1429 | { | |
1430 | struct netlink_sock *nlk = nlk_sk(sk); | |
1431 | ||
1432 | if (nlk->subscriptions && !subscriptions) | |
1433 | __sk_del_bind_node(sk); | |
1434 | else if (!nlk->subscriptions && subscriptions) | |
1435 | sk_add_bind_node(sk, &nl_table[sk->sk_protocol].mc_list); | |
1436 | nlk->subscriptions = subscriptions; | |
1437 | } | |
1438 | ||
b4ff4f04 | 1439 | static int netlink_realloc_groups(struct sock *sk) |
513c2500 PM |
1440 | { |
1441 | struct netlink_sock *nlk = nlk_sk(sk); | |
1442 | unsigned int groups; | |
b4ff4f04 | 1443 | unsigned long *new_groups; |
513c2500 PM |
1444 | int err = 0; |
1445 | ||
b4ff4f04 JB |
1446 | netlink_table_grab(); |
1447 | ||
513c2500 | 1448 | groups = nl_table[sk->sk_protocol].groups; |
b4ff4f04 | 1449 | if (!nl_table[sk->sk_protocol].registered) { |
513c2500 | 1450 | err = -ENOENT; |
b4ff4f04 JB |
1451 | goto out_unlock; |
1452 | } | |
513c2500 | 1453 | |
b4ff4f04 JB |
1454 | if (nlk->ngroups >= groups) |
1455 | goto out_unlock; | |
513c2500 | 1456 | |
b4ff4f04 JB |
1457 | new_groups = krealloc(nlk->groups, NLGRPSZ(groups), GFP_ATOMIC); |
1458 | if (new_groups == NULL) { | |
1459 | err = -ENOMEM; | |
1460 | goto out_unlock; | |
1461 | } | |
6ac552fd | 1462 | memset((char *)new_groups + NLGRPSZ(nlk->ngroups), 0, |
b4ff4f04 JB |
1463 | NLGRPSZ(groups) - NLGRPSZ(nlk->ngroups)); |
1464 | ||
1465 | nlk->groups = new_groups; | |
513c2500 | 1466 | nlk->ngroups = groups; |
b4ff4f04 JB |
1467 | out_unlock: |
1468 | netlink_table_ungrab(); | |
1469 | return err; | |
513c2500 PM |
1470 | } |
1471 | ||
02c81ab9 | 1472 | static void netlink_undo_bind(int group, long unsigned int groups, |
023e2cfa | 1473 | struct sock *sk) |
4f520900 | 1474 | { |
023e2cfa | 1475 | struct netlink_sock *nlk = nlk_sk(sk); |
4f520900 RGB |
1476 | int undo; |
1477 | ||
1478 | if (!nlk->netlink_unbind) | |
1479 | return; | |
1480 | ||
1481 | for (undo = 0; undo < group; undo++) | |
6251edd9 | 1482 | if (test_bit(undo, &groups)) |
8b7c36d8 | 1483 | nlk->netlink_unbind(sock_net(sk), undo + 1); |
4f520900 RGB |
1484 | } |
1485 | ||
6ac552fd PM |
1486 | static int netlink_bind(struct socket *sock, struct sockaddr *addr, |
1487 | int addr_len) | |
1da177e4 LT |
1488 | { |
1489 | struct sock *sk = sock->sk; | |
3b1e0a65 | 1490 | struct net *net = sock_net(sk); |
1da177e4 LT |
1491 | struct netlink_sock *nlk = nlk_sk(sk); |
1492 | struct sockaddr_nl *nladdr = (struct sockaddr_nl *)addr; | |
1493 | int err; | |
4f520900 | 1494 | long unsigned int groups = nladdr->nl_groups; |
746fac4d | 1495 | |
4e4b5376 HFS |
1496 | if (addr_len < sizeof(struct sockaddr_nl)) |
1497 | return -EINVAL; | |
1498 | ||
1da177e4 LT |
1499 | if (nladdr->nl_family != AF_NETLINK) |
1500 | return -EINVAL; | |
1501 | ||
1502 | /* Only superuser is allowed to listen multicasts */ | |
4f520900 | 1503 | if (groups) { |
5187cd05 | 1504 | if (!netlink_allowed(sock, NL_CFG_F_NONROOT_RECV)) |
513c2500 | 1505 | return -EPERM; |
b4ff4f04 JB |
1506 | err = netlink_realloc_groups(sk); |
1507 | if (err) | |
1508 | return err; | |
513c2500 | 1509 | } |
1da177e4 | 1510 | |
4f520900 | 1511 | if (nlk->portid) |
15e47304 | 1512 | if (nladdr->nl_pid != nlk->portid) |
1da177e4 | 1513 | return -EINVAL; |
4f520900 RGB |
1514 | |
1515 | if (nlk->netlink_bind && groups) { | |
1516 | int group; | |
1517 | ||
1518 | for (group = 0; group < nlk->ngroups; group++) { | |
1519 | if (!test_bit(group, &groups)) | |
1520 | continue; | |
8b7c36d8 | 1521 | err = nlk->netlink_bind(net, group + 1); |
4f520900 RGB |
1522 | if (!err) |
1523 | continue; | |
023e2cfa | 1524 | netlink_undo_bind(group, groups, sk); |
4f520900 RGB |
1525 | return err; |
1526 | } | |
1527 | } | |
1528 | ||
1529 | if (!nlk->portid) { | |
1da177e4 | 1530 | err = nladdr->nl_pid ? |
8ea65f4a | 1531 | netlink_insert(sk, nladdr->nl_pid) : |
1da177e4 | 1532 | netlink_autobind(sock); |
4f520900 | 1533 | if (err) { |
023e2cfa | 1534 | netlink_undo_bind(nlk->ngroups, groups, sk); |
1da177e4 | 1535 | return err; |
4f520900 | 1536 | } |
1da177e4 LT |
1537 | } |
1538 | ||
4f520900 | 1539 | if (!groups && (nlk->groups == NULL || !(u32)nlk->groups[0])) |
1da177e4 LT |
1540 | return 0; |
1541 | ||
1542 | netlink_table_grab(); | |
f7fa9b10 | 1543 | netlink_update_subscriptions(sk, nlk->subscriptions + |
4f520900 | 1544 | hweight32(groups) - |
746fac4d | 1545 | hweight32(nlk->groups[0])); |
4f520900 | 1546 | nlk->groups[0] = (nlk->groups[0] & ~0xffffffffUL) | groups; |
4277a083 | 1547 | netlink_update_listeners(sk); |
1da177e4 LT |
1548 | netlink_table_ungrab(); |
1549 | ||
1550 | return 0; | |
1551 | } | |
1552 | ||
1553 | static int netlink_connect(struct socket *sock, struct sockaddr *addr, | |
1554 | int alen, int flags) | |
1555 | { | |
1556 | int err = 0; | |
1557 | struct sock *sk = sock->sk; | |
1558 | struct netlink_sock *nlk = nlk_sk(sk); | |
6ac552fd | 1559 | struct sockaddr_nl *nladdr = (struct sockaddr_nl *)addr; |
1da177e4 | 1560 | |
6503d961 CG |
1561 | if (alen < sizeof(addr->sa_family)) |
1562 | return -EINVAL; | |
1563 | ||
1da177e4 LT |
1564 | if (addr->sa_family == AF_UNSPEC) { |
1565 | sk->sk_state = NETLINK_UNCONNECTED; | |
15e47304 | 1566 | nlk->dst_portid = 0; |
d629b836 | 1567 | nlk->dst_group = 0; |
1da177e4 LT |
1568 | return 0; |
1569 | } | |
1570 | if (addr->sa_family != AF_NETLINK) | |
1571 | return -EINVAL; | |
1572 | ||
46833a86 | 1573 | if ((nladdr->nl_groups || nladdr->nl_pid) && |
5187cd05 | 1574 | !netlink_allowed(sock, NL_CFG_F_NONROOT_SEND)) |
1da177e4 LT |
1575 | return -EPERM; |
1576 | ||
15e47304 | 1577 | if (!nlk->portid) |
1da177e4 LT |
1578 | err = netlink_autobind(sock); |
1579 | ||
1580 | if (err == 0) { | |
1581 | sk->sk_state = NETLINK_CONNECTED; | |
15e47304 | 1582 | nlk->dst_portid = nladdr->nl_pid; |
d629b836 | 1583 | nlk->dst_group = ffs(nladdr->nl_groups); |
1da177e4 LT |
1584 | } |
1585 | ||
1586 | return err; | |
1587 | } | |
1588 | ||
6ac552fd PM |
1589 | static int netlink_getname(struct socket *sock, struct sockaddr *addr, |
1590 | int *addr_len, int peer) | |
1da177e4 LT |
1591 | { |
1592 | struct sock *sk = sock->sk; | |
1593 | struct netlink_sock *nlk = nlk_sk(sk); | |
13cfa97b | 1594 | DECLARE_SOCKADDR(struct sockaddr_nl *, nladdr, addr); |
746fac4d | 1595 | |
1da177e4 LT |
1596 | nladdr->nl_family = AF_NETLINK; |
1597 | nladdr->nl_pad = 0; | |
1598 | *addr_len = sizeof(*nladdr); | |
1599 | ||
1600 | if (peer) { | |
15e47304 | 1601 | nladdr->nl_pid = nlk->dst_portid; |
d629b836 | 1602 | nladdr->nl_groups = netlink_group_mask(nlk->dst_group); |
1da177e4 | 1603 | } else { |
15e47304 | 1604 | nladdr->nl_pid = nlk->portid; |
513c2500 | 1605 | nladdr->nl_groups = nlk->groups ? nlk->groups[0] : 0; |
1da177e4 LT |
1606 | } |
1607 | return 0; | |
1608 | } | |
1609 | ||
15e47304 | 1610 | static struct sock *netlink_getsockbyportid(struct sock *ssk, u32 portid) |
1da177e4 | 1611 | { |
1da177e4 LT |
1612 | struct sock *sock; |
1613 | struct netlink_sock *nlk; | |
1614 | ||
15e47304 | 1615 | sock = netlink_lookup(sock_net(ssk), ssk->sk_protocol, portid); |
1da177e4 LT |
1616 | if (!sock) |
1617 | return ERR_PTR(-ECONNREFUSED); | |
1618 | ||
1619 | /* Don't bother queuing skb if kernel socket has no input function */ | |
1620 | nlk = nlk_sk(sock); | |
cd40b7d3 | 1621 | if (sock->sk_state == NETLINK_CONNECTED && |
15e47304 | 1622 | nlk->dst_portid != nlk_sk(ssk)->portid) { |
1da177e4 LT |
1623 | sock_put(sock); |
1624 | return ERR_PTR(-ECONNREFUSED); | |
1625 | } | |
1626 | return sock; | |
1627 | } | |
1628 | ||
1629 | struct sock *netlink_getsockbyfilp(struct file *filp) | |
1630 | { | |
496ad9aa | 1631 | struct inode *inode = file_inode(filp); |
1da177e4 LT |
1632 | struct sock *sock; |
1633 | ||
1634 | if (!S_ISSOCK(inode->i_mode)) | |
1635 | return ERR_PTR(-ENOTSOCK); | |
1636 | ||
1637 | sock = SOCKET_I(inode)->sk; | |
1638 | if (sock->sk_family != AF_NETLINK) | |
1639 | return ERR_PTR(-EINVAL); | |
1640 | ||
1641 | sock_hold(sock); | |
1642 | return sock; | |
1643 | } | |
1644 | ||
3a36515f PN |
1645 | static struct sk_buff *netlink_alloc_large_skb(unsigned int size, |
1646 | int broadcast) | |
c05cdb1b PNA |
1647 | { |
1648 | struct sk_buff *skb; | |
1649 | void *data; | |
1650 | ||
3a36515f | 1651 | if (size <= NLMSG_GOODSIZE || broadcast) |
c05cdb1b PNA |
1652 | return alloc_skb(size, GFP_KERNEL); |
1653 | ||
3a36515f PN |
1654 | size = SKB_DATA_ALIGN(size) + |
1655 | SKB_DATA_ALIGN(sizeof(struct skb_shared_info)); | |
c05cdb1b PNA |
1656 | |
1657 | data = vmalloc(size); | |
1658 | if (data == NULL) | |
3a36515f | 1659 | return NULL; |
c05cdb1b | 1660 | |
2ea2f62c | 1661 | skb = __build_skb(data, size); |
3a36515f PN |
1662 | if (skb == NULL) |
1663 | vfree(data); | |
2ea2f62c | 1664 | else |
3a36515f | 1665 | skb->destructor = netlink_skb_destructor; |
c05cdb1b PNA |
1666 | |
1667 | return skb; | |
c05cdb1b PNA |
1668 | } |
1669 | ||
1da177e4 LT |
1670 | /* |
1671 | * Attach a skb to a netlink socket. | |
1672 | * The caller must hold a reference to the destination socket. On error, the | |
1673 | * reference is dropped. The skb is not send to the destination, just all | |
1674 | * all error checks are performed and memory in the queue is reserved. | |
1675 | * Return values: | |
1676 | * < 0: error. skb freed, reference to sock dropped. | |
1677 | * 0: continue | |
1678 | * 1: repeat lookup - reference dropped while waiting for socket memory. | |
1679 | */ | |
9457afee | 1680 | int netlink_attachskb(struct sock *sk, struct sk_buff *skb, |
c3d8d1e3 | 1681 | long *timeo, struct sock *ssk) |
1da177e4 LT |
1682 | { |
1683 | struct netlink_sock *nlk; | |
1684 | ||
1685 | nlk = nlk_sk(sk); | |
1686 | ||
5fd96123 | 1687 | if ((atomic_read(&sk->sk_rmem_alloc) > sk->sk_rcvbuf || |
cc3a572f | 1688 | test_bit(NETLINK_S_CONGESTED, &nlk->state)) && |
5fd96123 | 1689 | !netlink_skb_is_mmaped(skb)) { |
1da177e4 | 1690 | DECLARE_WAITQUEUE(wait, current); |
c3d8d1e3 | 1691 | if (!*timeo) { |
aed81560 | 1692 | if (!ssk || netlink_is_kernel(ssk)) |
1da177e4 LT |
1693 | netlink_overrun(sk); |
1694 | sock_put(sk); | |
1695 | kfree_skb(skb); | |
1696 | return -EAGAIN; | |
1697 | } | |
1698 | ||
1699 | __set_current_state(TASK_INTERRUPTIBLE); | |
1700 | add_wait_queue(&nlk->wait, &wait); | |
1701 | ||
1702 | if ((atomic_read(&sk->sk_rmem_alloc) > sk->sk_rcvbuf || | |
cc3a572f | 1703 | test_bit(NETLINK_S_CONGESTED, &nlk->state)) && |
1da177e4 | 1704 | !sock_flag(sk, SOCK_DEAD)) |
c3d8d1e3 | 1705 | *timeo = schedule_timeout(*timeo); |
1da177e4 LT |
1706 | |
1707 | __set_current_state(TASK_RUNNING); | |
1708 | remove_wait_queue(&nlk->wait, &wait); | |
1709 | sock_put(sk); | |
1710 | ||
1711 | if (signal_pending(current)) { | |
1712 | kfree_skb(skb); | |
c3d8d1e3 | 1713 | return sock_intr_errno(*timeo); |
1da177e4 LT |
1714 | } |
1715 | return 1; | |
1716 | } | |
cf0a018a | 1717 | netlink_skb_set_owner_r(skb, sk); |
1da177e4 LT |
1718 | return 0; |
1719 | } | |
1720 | ||
4a7e7c2a | 1721 | static int __netlink_sendskb(struct sock *sk, struct sk_buff *skb) |
1da177e4 | 1722 | { |
1da177e4 LT |
1723 | int len = skb->len; |
1724 | ||
bcbde0d4 DB |
1725 | netlink_deliver_tap(skb); |
1726 | ||
f9c22888 PM |
1727 | #ifdef CONFIG_NETLINK_MMAP |
1728 | if (netlink_skb_is_mmaped(skb)) | |
1729 | netlink_queue_mmaped_skb(sk, skb); | |
1730 | else if (netlink_rx_is_mmaped(sk)) | |
1731 | netlink_ring_set_copied(sk, skb); | |
1732 | else | |
1733 | #endif /* CONFIG_NETLINK_MMAP */ | |
1734 | skb_queue_tail(&sk->sk_receive_queue, skb); | |
676d2369 | 1735 | sk->sk_data_ready(sk); |
4a7e7c2a ED |
1736 | return len; |
1737 | } | |
1738 | ||
1739 | int netlink_sendskb(struct sock *sk, struct sk_buff *skb) | |
1740 | { | |
1741 | int len = __netlink_sendskb(sk, skb); | |
1742 | ||
1da177e4 LT |
1743 | sock_put(sk); |
1744 | return len; | |
1745 | } | |
1746 | ||
1747 | void netlink_detachskb(struct sock *sk, struct sk_buff *skb) | |
1748 | { | |
1749 | kfree_skb(skb); | |
1750 | sock_put(sk); | |
1751 | } | |
1752 | ||
b57ef81f | 1753 | static struct sk_buff *netlink_trim(struct sk_buff *skb, gfp_t allocation) |
1da177e4 LT |
1754 | { |
1755 | int delta; | |
1756 | ||
1298ca46 | 1757 | WARN_ON(skb->sk != NULL); |
5fd96123 PM |
1758 | if (netlink_skb_is_mmaped(skb)) |
1759 | return skb; | |
1da177e4 | 1760 | |
4305b541 | 1761 | delta = skb->end - skb->tail; |
c05cdb1b | 1762 | if (is_vmalloc_addr(skb->head) || delta * 2 < skb->truesize) |
1da177e4 LT |
1763 | return skb; |
1764 | ||
1765 | if (skb_shared(skb)) { | |
1766 | struct sk_buff *nskb = skb_clone(skb, allocation); | |
1767 | if (!nskb) | |
1768 | return skb; | |
8460c00f | 1769 | consume_skb(skb); |
1da177e4 LT |
1770 | skb = nskb; |
1771 | } | |
1772 | ||
1773 | if (!pskb_expand_head(skb, 0, -delta, allocation)) | |
1774 | skb->truesize -= delta; | |
1775 | ||
1776 | return skb; | |
1777 | } | |
1778 | ||
3fbc2905 EB |
1779 | static int netlink_unicast_kernel(struct sock *sk, struct sk_buff *skb, |
1780 | struct sock *ssk) | |
cd40b7d3 DL |
1781 | { |
1782 | int ret; | |
1783 | struct netlink_sock *nlk = nlk_sk(sk); | |
1784 | ||
1785 | ret = -ECONNREFUSED; | |
1786 | if (nlk->netlink_rcv != NULL) { | |
1787 | ret = skb->len; | |
cf0a018a | 1788 | netlink_skb_set_owner_r(skb, sk); |
e32123e5 | 1789 | NETLINK_CB(skb).sk = ssk; |
73bfd370 | 1790 | netlink_deliver_tap_kernel(sk, ssk, skb); |
cd40b7d3 | 1791 | nlk->netlink_rcv(skb); |
bfb253c9 ED |
1792 | consume_skb(skb); |
1793 | } else { | |
1794 | kfree_skb(skb); | |
cd40b7d3 | 1795 | } |
cd40b7d3 DL |
1796 | sock_put(sk); |
1797 | return ret; | |
1798 | } | |
1799 | ||
1800 | int netlink_unicast(struct sock *ssk, struct sk_buff *skb, | |
15e47304 | 1801 | u32 portid, int nonblock) |
1da177e4 LT |
1802 | { |
1803 | struct sock *sk; | |
1804 | int err; | |
1805 | long timeo; | |
1806 | ||
1807 | skb = netlink_trim(skb, gfp_any()); | |
1808 | ||
1809 | timeo = sock_sndtimeo(ssk, nonblock); | |
1810 | retry: | |
15e47304 | 1811 | sk = netlink_getsockbyportid(ssk, portid); |
1da177e4 LT |
1812 | if (IS_ERR(sk)) { |
1813 | kfree_skb(skb); | |
1814 | return PTR_ERR(sk); | |
1815 | } | |
cd40b7d3 | 1816 | if (netlink_is_kernel(sk)) |
3fbc2905 | 1817 | return netlink_unicast_kernel(sk, skb, ssk); |
cd40b7d3 | 1818 | |
b1153f29 | 1819 | if (sk_filter(sk, skb)) { |
84874607 | 1820 | err = skb->len; |
b1153f29 SH |
1821 | kfree_skb(skb); |
1822 | sock_put(sk); | |
1823 | return err; | |
1824 | } | |
1825 | ||
9457afee | 1826 | err = netlink_attachskb(sk, skb, &timeo, ssk); |
1da177e4 LT |
1827 | if (err == 1) |
1828 | goto retry; | |
1829 | if (err) | |
1830 | return err; | |
1831 | ||
7ee015e0 | 1832 | return netlink_sendskb(sk, skb); |
1da177e4 | 1833 | } |
6ac552fd | 1834 | EXPORT_SYMBOL(netlink_unicast); |
1da177e4 | 1835 | |
f9c22888 PM |
1836 | struct sk_buff *netlink_alloc_skb(struct sock *ssk, unsigned int size, |
1837 | u32 dst_portid, gfp_t gfp_mask) | |
1838 | { | |
1839 | #ifdef CONFIG_NETLINK_MMAP | |
1840 | struct sock *sk = NULL; | |
1841 | struct sk_buff *skb; | |
1842 | struct netlink_ring *ring; | |
1843 | struct nl_mmap_hdr *hdr; | |
1844 | unsigned int maxlen; | |
1845 | ||
1846 | sk = netlink_getsockbyportid(ssk, dst_portid); | |
1847 | if (IS_ERR(sk)) | |
1848 | goto out; | |
1849 | ||
1850 | ring = &nlk_sk(sk)->rx_ring; | |
1851 | /* fast-path without atomic ops for common case: non-mmaped receiver */ | |
1852 | if (ring->pg_vec == NULL) | |
1853 | goto out_put; | |
1854 | ||
aae9f0e2 TG |
1855 | if (ring->frame_size - NL_MMAP_HDRLEN < size) |
1856 | goto out_put; | |
1857 | ||
f9c22888 PM |
1858 | skb = alloc_skb_head(gfp_mask); |
1859 | if (skb == NULL) | |
1860 | goto err1; | |
1861 | ||
1862 | spin_lock_bh(&sk->sk_receive_queue.lock); | |
1863 | /* check again under lock */ | |
1864 | if (ring->pg_vec == NULL) | |
1865 | goto out_free; | |
1866 | ||
aae9f0e2 | 1867 | /* check again under lock */ |
f9c22888 PM |
1868 | maxlen = ring->frame_size - NL_MMAP_HDRLEN; |
1869 | if (maxlen < size) | |
1870 | goto out_free; | |
1871 | ||
1872 | netlink_forward_ring(ring); | |
1873 | hdr = netlink_current_frame(ring, NL_MMAP_STATUS_UNUSED); | |
1874 | if (hdr == NULL) | |
1875 | goto err2; | |
1876 | netlink_ring_setup_skb(skb, sk, ring, hdr); | |
1877 | netlink_set_status(hdr, NL_MMAP_STATUS_RESERVED); | |
1878 | atomic_inc(&ring->pending); | |
1879 | netlink_increment_head(ring); | |
1880 | ||
1881 | spin_unlock_bh(&sk->sk_receive_queue.lock); | |
1882 | return skb; | |
1883 | ||
1884 | err2: | |
1885 | kfree_skb(skb); | |
1886 | spin_unlock_bh(&sk->sk_receive_queue.lock); | |
cd1df525 | 1887 | netlink_overrun(sk); |
f9c22888 PM |
1888 | err1: |
1889 | sock_put(sk); | |
1890 | return NULL; | |
1891 | ||
1892 | out_free: | |
1893 | kfree_skb(skb); | |
1894 | spin_unlock_bh(&sk->sk_receive_queue.lock); | |
1895 | out_put: | |
1896 | sock_put(sk); | |
1897 | out: | |
1898 | #endif | |
1899 | return alloc_skb(size, gfp_mask); | |
1900 | } | |
1901 | EXPORT_SYMBOL_GPL(netlink_alloc_skb); | |
1902 | ||
4277a083 PM |
1903 | int netlink_has_listeners(struct sock *sk, unsigned int group) |
1904 | { | |
1905 | int res = 0; | |
5c398dc8 | 1906 | struct listeners *listeners; |
4277a083 | 1907 | |
aed81560 | 1908 | BUG_ON(!netlink_is_kernel(sk)); |
b4ff4f04 JB |
1909 | |
1910 | rcu_read_lock(); | |
1911 | listeners = rcu_dereference(nl_table[sk->sk_protocol].listeners); | |
1912 | ||
6d772ac5 | 1913 | if (listeners && group - 1 < nl_table[sk->sk_protocol].groups) |
5c398dc8 | 1914 | res = test_bit(group - 1, listeners->masks); |
b4ff4f04 JB |
1915 | |
1916 | rcu_read_unlock(); | |
1917 | ||
4277a083 PM |
1918 | return res; |
1919 | } | |
1920 | EXPORT_SYMBOL_GPL(netlink_has_listeners); | |
1921 | ||
b57ef81f | 1922 | static int netlink_broadcast_deliver(struct sock *sk, struct sk_buff *skb) |
1da177e4 LT |
1923 | { |
1924 | struct netlink_sock *nlk = nlk_sk(sk); | |
1925 | ||
1926 | if (atomic_read(&sk->sk_rmem_alloc) <= sk->sk_rcvbuf && | |
cc3a572f | 1927 | !test_bit(NETLINK_S_CONGESTED, &nlk->state)) { |
cf0a018a | 1928 | netlink_skb_set_owner_r(skb, sk); |
4a7e7c2a | 1929 | __netlink_sendskb(sk, skb); |
2c645800 | 1930 | return atomic_read(&sk->sk_rmem_alloc) > (sk->sk_rcvbuf >> 1); |
1da177e4 LT |
1931 | } |
1932 | return -1; | |
1933 | } | |
1934 | ||
1935 | struct netlink_broadcast_data { | |
1936 | struct sock *exclude_sk; | |
b4b51029 | 1937 | struct net *net; |
15e47304 | 1938 | u32 portid; |
1da177e4 LT |
1939 | u32 group; |
1940 | int failure; | |
ff491a73 | 1941 | int delivery_failure; |
1da177e4 LT |
1942 | int congested; |
1943 | int delivered; | |
7d877f3b | 1944 | gfp_t allocation; |
1da177e4 | 1945 | struct sk_buff *skb, *skb2; |
910a7e90 EB |
1946 | int (*tx_filter)(struct sock *dsk, struct sk_buff *skb, void *data); |
1947 | void *tx_data; | |
1da177e4 LT |
1948 | }; |
1949 | ||
46c9521f RR |
1950 | static void do_one_broadcast(struct sock *sk, |
1951 | struct netlink_broadcast_data *p) | |
1da177e4 LT |
1952 | { |
1953 | struct netlink_sock *nlk = nlk_sk(sk); | |
1954 | int val; | |
1955 | ||
1956 | if (p->exclude_sk == sk) | |
46c9521f | 1957 | return; |
1da177e4 | 1958 | |
15e47304 | 1959 | if (nlk->portid == p->portid || p->group - 1 >= nlk->ngroups || |
f7fa9b10 | 1960 | !test_bit(p->group - 1, nlk->groups)) |
46c9521f | 1961 | return; |
1da177e4 | 1962 | |
59324cf3 ND |
1963 | if (!net_eq(sock_net(sk), p->net)) { |
1964 | if (!(nlk->flags & NETLINK_F_LISTEN_ALL_NSID)) | |
1965 | return; | |
1966 | ||
1967 | if (!peernet_has_id(sock_net(sk), p->net)) | |
1968 | return; | |
1969 | ||
1970 | if (!file_ns_capable(sk->sk_socket->file, p->net->user_ns, | |
1971 | CAP_NET_BROADCAST)) | |
1972 | return; | |
1973 | } | |
b4b51029 | 1974 | |
1da177e4 LT |
1975 | if (p->failure) { |
1976 | netlink_overrun(sk); | |
46c9521f | 1977 | return; |
1da177e4 LT |
1978 | } |
1979 | ||
1980 | sock_hold(sk); | |
1981 | if (p->skb2 == NULL) { | |
68acc024 | 1982 | if (skb_shared(p->skb)) { |
1da177e4 LT |
1983 | p->skb2 = skb_clone(p->skb, p->allocation); |
1984 | } else { | |
68acc024 TC |
1985 | p->skb2 = skb_get(p->skb); |
1986 | /* | |
1987 | * skb ownership may have been set when | |
1988 | * delivered to a previous socket. | |
1989 | */ | |
1990 | skb_orphan(p->skb2); | |
1da177e4 LT |
1991 | } |
1992 | } | |
1993 | if (p->skb2 == NULL) { | |
1994 | netlink_overrun(sk); | |
1995 | /* Clone failed. Notify ALL listeners. */ | |
1996 | p->failure = 1; | |
cc3a572f | 1997 | if (nlk->flags & NETLINK_F_BROADCAST_SEND_ERROR) |
be0c22a4 | 1998 | p->delivery_failure = 1; |
59324cf3 ND |
1999 | goto out; |
2000 | } | |
2001 | if (p->tx_filter && p->tx_filter(sk, p->skb2, p->tx_data)) { | |
910a7e90 EB |
2002 | kfree_skb(p->skb2); |
2003 | p->skb2 = NULL; | |
59324cf3 ND |
2004 | goto out; |
2005 | } | |
2006 | if (sk_filter(sk, p->skb2)) { | |
b1153f29 SH |
2007 | kfree_skb(p->skb2); |
2008 | p->skb2 = NULL; | |
59324cf3 ND |
2009 | goto out; |
2010 | } | |
2011 | NETLINK_CB(p->skb2).nsid = peernet2id(sock_net(sk), p->net); | |
2012 | NETLINK_CB(p->skb2).nsid_is_set = true; | |
2013 | val = netlink_broadcast_deliver(sk, p->skb2); | |
2014 | if (val < 0) { | |
1da177e4 | 2015 | netlink_overrun(sk); |
cc3a572f | 2016 | if (nlk->flags & NETLINK_F_BROADCAST_SEND_ERROR) |
be0c22a4 | 2017 | p->delivery_failure = 1; |
1da177e4 LT |
2018 | } else { |
2019 | p->congested |= val; | |
2020 | p->delivered = 1; | |
2021 | p->skb2 = NULL; | |
2022 | } | |
59324cf3 | 2023 | out: |
1da177e4 | 2024 | sock_put(sk); |
1da177e4 LT |
2025 | } |
2026 | ||
15e47304 | 2027 | int netlink_broadcast_filtered(struct sock *ssk, struct sk_buff *skb, u32 portid, |
910a7e90 EB |
2028 | u32 group, gfp_t allocation, |
2029 | int (*filter)(struct sock *dsk, struct sk_buff *skb, void *data), | |
2030 | void *filter_data) | |
1da177e4 | 2031 | { |
3b1e0a65 | 2032 | struct net *net = sock_net(ssk); |
1da177e4 | 2033 | struct netlink_broadcast_data info; |
1da177e4 LT |
2034 | struct sock *sk; |
2035 | ||
2036 | skb = netlink_trim(skb, allocation); | |
2037 | ||
2038 | info.exclude_sk = ssk; | |
b4b51029 | 2039 | info.net = net; |
15e47304 | 2040 | info.portid = portid; |
1da177e4 LT |
2041 | info.group = group; |
2042 | info.failure = 0; | |
ff491a73 | 2043 | info.delivery_failure = 0; |
1da177e4 LT |
2044 | info.congested = 0; |
2045 | info.delivered = 0; | |
2046 | info.allocation = allocation; | |
2047 | info.skb = skb; | |
2048 | info.skb2 = NULL; | |
910a7e90 EB |
2049 | info.tx_filter = filter; |
2050 | info.tx_data = filter_data; | |
1da177e4 LT |
2051 | |
2052 | /* While we sleep in clone, do not allow to change socket list */ | |
2053 | ||
2054 | netlink_lock_table(); | |
2055 | ||
b67bfe0d | 2056 | sk_for_each_bound(sk, &nl_table[ssk->sk_protocol].mc_list) |
1da177e4 LT |
2057 | do_one_broadcast(sk, &info); |
2058 | ||
70d4bf6d | 2059 | consume_skb(skb); |
aa1c6a6f | 2060 | |
1da177e4 LT |
2061 | netlink_unlock_table(); |
2062 | ||
70d4bf6d NH |
2063 | if (info.delivery_failure) { |
2064 | kfree_skb(info.skb2); | |
ff491a73 | 2065 | return -ENOBUFS; |
658cb354 ED |
2066 | } |
2067 | consume_skb(info.skb2); | |
ff491a73 | 2068 | |
1da177e4 LT |
2069 | if (info.delivered) { |
2070 | if (info.congested && (allocation & __GFP_WAIT)) | |
2071 | yield(); | |
2072 | return 0; | |
2073 | } | |
1da177e4 LT |
2074 | return -ESRCH; |
2075 | } | |
910a7e90 EB |
2076 | EXPORT_SYMBOL(netlink_broadcast_filtered); |
2077 | ||
15e47304 | 2078 | int netlink_broadcast(struct sock *ssk, struct sk_buff *skb, u32 portid, |
910a7e90 EB |
2079 | u32 group, gfp_t allocation) |
2080 | { | |
15e47304 | 2081 | return netlink_broadcast_filtered(ssk, skb, portid, group, allocation, |
910a7e90 EB |
2082 | NULL, NULL); |
2083 | } | |
6ac552fd | 2084 | EXPORT_SYMBOL(netlink_broadcast); |
1da177e4 LT |
2085 | |
2086 | struct netlink_set_err_data { | |
2087 | struct sock *exclude_sk; | |
15e47304 | 2088 | u32 portid; |
1da177e4 LT |
2089 | u32 group; |
2090 | int code; | |
2091 | }; | |
2092 | ||
b57ef81f | 2093 | static int do_one_set_err(struct sock *sk, struct netlink_set_err_data *p) |
1da177e4 LT |
2094 | { |
2095 | struct netlink_sock *nlk = nlk_sk(sk); | |
1a50307b | 2096 | int ret = 0; |
1da177e4 LT |
2097 | |
2098 | if (sk == p->exclude_sk) | |
2099 | goto out; | |
2100 | ||
09ad9bc7 | 2101 | if (!net_eq(sock_net(sk), sock_net(p->exclude_sk))) |
b4b51029 EB |
2102 | goto out; |
2103 | ||
15e47304 | 2104 | if (nlk->portid == p->portid || p->group - 1 >= nlk->ngroups || |
f7fa9b10 | 2105 | !test_bit(p->group - 1, nlk->groups)) |
1da177e4 LT |
2106 | goto out; |
2107 | ||
cc3a572f | 2108 | if (p->code == ENOBUFS && nlk->flags & NETLINK_F_RECV_NO_ENOBUFS) { |
1a50307b PNA |
2109 | ret = 1; |
2110 | goto out; | |
2111 | } | |
2112 | ||
1da177e4 LT |
2113 | sk->sk_err = p->code; |
2114 | sk->sk_error_report(sk); | |
2115 | out: | |
1a50307b | 2116 | return ret; |
1da177e4 LT |
2117 | } |
2118 | ||
4843b93c PNA |
2119 | /** |
2120 | * netlink_set_err - report error to broadcast listeners | |
2121 | * @ssk: the kernel netlink socket, as returned by netlink_kernel_create() | |
15e47304 | 2122 | * @portid: the PORTID of a process that we want to skip (if any) |
840e93f2 | 2123 | * @group: the broadcast group that will notice the error |
4843b93c | 2124 | * @code: error code, must be negative (as usual in kernelspace) |
1a50307b PNA |
2125 | * |
2126 | * This function returns the number of broadcast listeners that have set the | |
cc3a572f | 2127 | * NETLINK_NO_ENOBUFS socket option. |
4843b93c | 2128 | */ |
15e47304 | 2129 | int netlink_set_err(struct sock *ssk, u32 portid, u32 group, int code) |
1da177e4 LT |
2130 | { |
2131 | struct netlink_set_err_data info; | |
1da177e4 | 2132 | struct sock *sk; |
1a50307b | 2133 | int ret = 0; |
1da177e4 LT |
2134 | |
2135 | info.exclude_sk = ssk; | |
15e47304 | 2136 | info.portid = portid; |
1da177e4 | 2137 | info.group = group; |
4843b93c PNA |
2138 | /* sk->sk_err wants a positive error value */ |
2139 | info.code = -code; | |
1da177e4 LT |
2140 | |
2141 | read_lock(&nl_table_lock); | |
2142 | ||
b67bfe0d | 2143 | sk_for_each_bound(sk, &nl_table[ssk->sk_protocol].mc_list) |
1a50307b | 2144 | ret += do_one_set_err(sk, &info); |
1da177e4 LT |
2145 | |
2146 | read_unlock(&nl_table_lock); | |
1a50307b | 2147 | return ret; |
1da177e4 | 2148 | } |
dd5b6ce6 | 2149 | EXPORT_SYMBOL(netlink_set_err); |
1da177e4 | 2150 | |
84659eb5 JB |
2151 | /* must be called with netlink table grabbed */ |
2152 | static void netlink_update_socket_mc(struct netlink_sock *nlk, | |
2153 | unsigned int group, | |
2154 | int is_new) | |
2155 | { | |
2156 | int old, new = !!is_new, subscriptions; | |
2157 | ||
2158 | old = test_bit(group - 1, nlk->groups); | |
2159 | subscriptions = nlk->subscriptions - old + new; | |
2160 | if (new) | |
2161 | __set_bit(group - 1, nlk->groups); | |
2162 | else | |
2163 | __clear_bit(group - 1, nlk->groups); | |
2164 | netlink_update_subscriptions(&nlk->sk, subscriptions); | |
2165 | netlink_update_listeners(&nlk->sk); | |
2166 | } | |
2167 | ||
9a4595bc | 2168 | static int netlink_setsockopt(struct socket *sock, int level, int optname, |
b7058842 | 2169 | char __user *optval, unsigned int optlen) |
9a4595bc PM |
2170 | { |
2171 | struct sock *sk = sock->sk; | |
2172 | struct netlink_sock *nlk = nlk_sk(sk); | |
eb496534 JB |
2173 | unsigned int val = 0; |
2174 | int err; | |
9a4595bc PM |
2175 | |
2176 | if (level != SOL_NETLINK) | |
2177 | return -ENOPROTOOPT; | |
2178 | ||
ccdfcc39 PM |
2179 | if (optname != NETLINK_RX_RING && optname != NETLINK_TX_RING && |
2180 | optlen >= sizeof(int) && | |
eb496534 | 2181 | get_user(val, (unsigned int __user *)optval)) |
9a4595bc PM |
2182 | return -EFAULT; |
2183 | ||
2184 | switch (optname) { | |
2185 | case NETLINK_PKTINFO: | |
2186 | if (val) | |
cc3a572f | 2187 | nlk->flags |= NETLINK_F_RECV_PKTINFO; |
9a4595bc | 2188 | else |
cc3a572f | 2189 | nlk->flags &= ~NETLINK_F_RECV_PKTINFO; |
9a4595bc PM |
2190 | err = 0; |
2191 | break; | |
2192 | case NETLINK_ADD_MEMBERSHIP: | |
2193 | case NETLINK_DROP_MEMBERSHIP: { | |
5187cd05 | 2194 | if (!netlink_allowed(sock, NL_CFG_F_NONROOT_RECV)) |
9a4595bc | 2195 | return -EPERM; |
b4ff4f04 JB |
2196 | err = netlink_realloc_groups(sk); |
2197 | if (err) | |
2198 | return err; | |
9a4595bc PM |
2199 | if (!val || val - 1 >= nlk->ngroups) |
2200 | return -EINVAL; | |
7774d5e0 | 2201 | if (optname == NETLINK_ADD_MEMBERSHIP && nlk->netlink_bind) { |
023e2cfa | 2202 | err = nlk->netlink_bind(sock_net(sk), val); |
4f520900 RGB |
2203 | if (err) |
2204 | return err; | |
2205 | } | |
9a4595bc | 2206 | netlink_table_grab(); |
84659eb5 JB |
2207 | netlink_update_socket_mc(nlk, val, |
2208 | optname == NETLINK_ADD_MEMBERSHIP); | |
9a4595bc | 2209 | netlink_table_ungrab(); |
7774d5e0 | 2210 | if (optname == NETLINK_DROP_MEMBERSHIP && nlk->netlink_unbind) |
023e2cfa | 2211 | nlk->netlink_unbind(sock_net(sk), val); |
03292745 | 2212 | |
9a4595bc PM |
2213 | err = 0; |
2214 | break; | |
2215 | } | |
be0c22a4 PNA |
2216 | case NETLINK_BROADCAST_ERROR: |
2217 | if (val) | |
cc3a572f | 2218 | nlk->flags |= NETLINK_F_BROADCAST_SEND_ERROR; |
be0c22a4 | 2219 | else |
cc3a572f | 2220 | nlk->flags &= ~NETLINK_F_BROADCAST_SEND_ERROR; |
be0c22a4 PNA |
2221 | err = 0; |
2222 | break; | |
38938bfe PNA |
2223 | case NETLINK_NO_ENOBUFS: |
2224 | if (val) { | |
cc3a572f ND |
2225 | nlk->flags |= NETLINK_F_RECV_NO_ENOBUFS; |
2226 | clear_bit(NETLINK_S_CONGESTED, &nlk->state); | |
38938bfe | 2227 | wake_up_interruptible(&nlk->wait); |
658cb354 | 2228 | } else { |
cc3a572f | 2229 | nlk->flags &= ~NETLINK_F_RECV_NO_ENOBUFS; |
658cb354 | 2230 | } |
38938bfe PNA |
2231 | err = 0; |
2232 | break; | |
ccdfcc39 PM |
2233 | #ifdef CONFIG_NETLINK_MMAP |
2234 | case NETLINK_RX_RING: | |
2235 | case NETLINK_TX_RING: { | |
2236 | struct nl_mmap_req req; | |
2237 | ||
2238 | /* Rings might consume more memory than queue limits, require | |
2239 | * CAP_NET_ADMIN. | |
2240 | */ | |
2241 | if (!capable(CAP_NET_ADMIN)) | |
2242 | return -EPERM; | |
2243 | if (optlen < sizeof(req)) | |
2244 | return -EINVAL; | |
2245 | if (copy_from_user(&req, optval, sizeof(req))) | |
2246 | return -EFAULT; | |
0470eb99 | 2247 | err = netlink_set_ring(sk, &req, |
ccdfcc39 PM |
2248 | optname == NETLINK_TX_RING); |
2249 | break; | |
2250 | } | |
2251 | #endif /* CONFIG_NETLINK_MMAP */ | |
59324cf3 ND |
2252 | case NETLINK_LISTEN_ALL_NSID: |
2253 | if (!ns_capable(sock_net(sk)->user_ns, CAP_NET_BROADCAST)) | |
2254 | return -EPERM; | |
2255 | ||
2256 | if (val) | |
2257 | nlk->flags |= NETLINK_F_LISTEN_ALL_NSID; | |
2258 | else | |
2259 | nlk->flags &= ~NETLINK_F_LISTEN_ALL_NSID; | |
2260 | err = 0; | |
2261 | break; | |
0a6a3a23 CR |
2262 | case NETLINK_CAP_ACK: |
2263 | if (val) | |
2264 | nlk->flags |= NETLINK_F_CAP_ACK; | |
2265 | else | |
2266 | nlk->flags &= ~NETLINK_F_CAP_ACK; | |
2267 | err = 0; | |
2268 | break; | |
9a4595bc PM |
2269 | default: |
2270 | err = -ENOPROTOOPT; | |
2271 | } | |
2272 | return err; | |
2273 | } | |
2274 | ||
2275 | static int netlink_getsockopt(struct socket *sock, int level, int optname, | |
746fac4d | 2276 | char __user *optval, int __user *optlen) |
9a4595bc PM |
2277 | { |
2278 | struct sock *sk = sock->sk; | |
2279 | struct netlink_sock *nlk = nlk_sk(sk); | |
2280 | int len, val, err; | |
2281 | ||
2282 | if (level != SOL_NETLINK) | |
2283 | return -ENOPROTOOPT; | |
2284 | ||
2285 | if (get_user(len, optlen)) | |
2286 | return -EFAULT; | |
2287 | if (len < 0) | |
2288 | return -EINVAL; | |
2289 | ||
2290 | switch (optname) { | |
2291 | case NETLINK_PKTINFO: | |
2292 | if (len < sizeof(int)) | |
2293 | return -EINVAL; | |
2294 | len = sizeof(int); | |
cc3a572f | 2295 | val = nlk->flags & NETLINK_F_RECV_PKTINFO ? 1 : 0; |
a27b58fe HC |
2296 | if (put_user(len, optlen) || |
2297 | put_user(val, optval)) | |
2298 | return -EFAULT; | |
9a4595bc PM |
2299 | err = 0; |
2300 | break; | |
be0c22a4 PNA |
2301 | case NETLINK_BROADCAST_ERROR: |
2302 | if (len < sizeof(int)) | |
2303 | return -EINVAL; | |
2304 | len = sizeof(int); | |
cc3a572f | 2305 | val = nlk->flags & NETLINK_F_BROADCAST_SEND_ERROR ? 1 : 0; |
be0c22a4 PNA |
2306 | if (put_user(len, optlen) || |
2307 | put_user(val, optval)) | |
2308 | return -EFAULT; | |
2309 | err = 0; | |
2310 | break; | |
38938bfe PNA |
2311 | case NETLINK_NO_ENOBUFS: |
2312 | if (len < sizeof(int)) | |
2313 | return -EINVAL; | |
2314 | len = sizeof(int); | |
cc3a572f | 2315 | val = nlk->flags & NETLINK_F_RECV_NO_ENOBUFS ? 1 : 0; |
38938bfe PNA |
2316 | if (put_user(len, optlen) || |
2317 | put_user(val, optval)) | |
2318 | return -EFAULT; | |
2319 | err = 0; | |
2320 | break; | |
b42be38b DH |
2321 | case NETLINK_LIST_MEMBERSHIPS: { |
2322 | int pos, idx, shift; | |
2323 | ||
2324 | err = 0; | |
2325 | netlink_table_grab(); | |
2326 | for (pos = 0; pos * 8 < nlk->ngroups; pos += sizeof(u32)) { | |
2327 | if (len - pos < sizeof(u32)) | |
2328 | break; | |
2329 | ||
2330 | idx = pos / sizeof(unsigned long); | |
2331 | shift = (pos % sizeof(unsigned long)) * 8; | |
2332 | if (put_user((u32)(nlk->groups[idx] >> shift), | |
2333 | (u32 __user *)(optval + pos))) { | |
2334 | err = -EFAULT; | |
2335 | break; | |
2336 | } | |
2337 | } | |
2338 | if (put_user(ALIGN(nlk->ngroups / 8, sizeof(u32)), optlen)) | |
2339 | err = -EFAULT; | |
2340 | netlink_table_ungrab(); | |
2341 | break; | |
2342 | } | |
0a6a3a23 CR |
2343 | case NETLINK_CAP_ACK: |
2344 | if (len < sizeof(int)) | |
2345 | return -EINVAL; | |
2346 | len = sizeof(int); | |
2347 | val = nlk->flags & NETLINK_F_CAP_ACK ? 1 : 0; | |
2348 | if (put_user(len, optlen) || | |
2349 | put_user(val, optval)) | |
2350 | return -EFAULT; | |
2351 | err = 0; | |
2352 | break; | |
9a4595bc PM |
2353 | default: |
2354 | err = -ENOPROTOOPT; | |
2355 | } | |
2356 | return err; | |
2357 | } | |
2358 | ||
2359 | static void netlink_cmsg_recv_pktinfo(struct msghdr *msg, struct sk_buff *skb) | |
2360 | { | |
2361 | struct nl_pktinfo info; | |
2362 | ||
2363 | info.group = NETLINK_CB(skb).dst_group; | |
2364 | put_cmsg(msg, SOL_NETLINK, NETLINK_PKTINFO, sizeof(info), &info); | |
2365 | } | |
2366 | ||
59324cf3 ND |
2367 | static void netlink_cmsg_listen_all_nsid(struct sock *sk, struct msghdr *msg, |
2368 | struct sk_buff *skb) | |
2369 | { | |
2370 | if (!NETLINK_CB(skb).nsid_is_set) | |
2371 | return; | |
2372 | ||
2373 | put_cmsg(msg, SOL_NETLINK, NETLINK_LISTEN_ALL_NSID, sizeof(int), | |
2374 | &NETLINK_CB(skb).nsid); | |
2375 | } | |
2376 | ||
1b784140 | 2377 | static int netlink_sendmsg(struct socket *sock, struct msghdr *msg, size_t len) |
1da177e4 | 2378 | { |
1da177e4 LT |
2379 | struct sock *sk = sock->sk; |
2380 | struct netlink_sock *nlk = nlk_sk(sk); | |
342dfc30 | 2381 | DECLARE_SOCKADDR(struct sockaddr_nl *, addr, msg->msg_name); |
15e47304 | 2382 | u32 dst_portid; |
d629b836 | 2383 | u32 dst_group; |
1da177e4 LT |
2384 | struct sk_buff *skb; |
2385 | int err; | |
2386 | struct scm_cookie scm; | |
2d7a85f4 | 2387 | u32 netlink_skb_flags = 0; |
1da177e4 LT |
2388 | |
2389 | if (msg->msg_flags&MSG_OOB) | |
2390 | return -EOPNOTSUPP; | |
2391 | ||
7cc05662 | 2392 | err = scm_send(sock, msg, &scm, true); |
1da177e4 LT |
2393 | if (err < 0) |
2394 | return err; | |
2395 | ||
2396 | if (msg->msg_namelen) { | |
b47030c7 | 2397 | err = -EINVAL; |
1da177e4 | 2398 | if (addr->nl_family != AF_NETLINK) |
b47030c7 | 2399 | goto out; |
15e47304 | 2400 | dst_portid = addr->nl_pid; |
d629b836 | 2401 | dst_group = ffs(addr->nl_groups); |
b47030c7 | 2402 | err = -EPERM; |
15e47304 | 2403 | if ((dst_group || dst_portid) && |
5187cd05 | 2404 | !netlink_allowed(sock, NL_CFG_F_NONROOT_SEND)) |
b47030c7 | 2405 | goto out; |
2d7a85f4 | 2406 | netlink_skb_flags |= NETLINK_SKB_DST; |
1da177e4 | 2407 | } else { |
15e47304 | 2408 | dst_portid = nlk->dst_portid; |
d629b836 | 2409 | dst_group = nlk->dst_group; |
1da177e4 LT |
2410 | } |
2411 | ||
15e47304 | 2412 | if (!nlk->portid) { |
1da177e4 LT |
2413 | err = netlink_autobind(sock); |
2414 | if (err) | |
2415 | goto out; | |
2416 | } | |
2417 | ||
a8866ff6 AV |
2418 | /* It's a really convoluted way for userland to ask for mmaped |
2419 | * sendmsg(), but that's what we've got... | |
2420 | */ | |
5fd96123 | 2421 | if (netlink_tx_is_mmaped(sk) && |
c953e239 | 2422 | iter_is_iovec(&msg->msg_iter) && |
a8866ff6 | 2423 | msg->msg_iter.nr_segs == 1 && |
c0371da6 | 2424 | msg->msg_iter.iov->iov_base == NULL) { |
5fd96123 | 2425 | err = netlink_mmap_sendmsg(sk, msg, dst_portid, dst_group, |
7cc05662 | 2426 | &scm); |
5fd96123 PM |
2427 | goto out; |
2428 | } | |
2429 | ||
1da177e4 LT |
2430 | err = -EMSGSIZE; |
2431 | if (len > sk->sk_sndbuf - 32) | |
2432 | goto out; | |
2433 | err = -ENOBUFS; | |
3a36515f | 2434 | skb = netlink_alloc_large_skb(len, dst_group); |
6ac552fd | 2435 | if (skb == NULL) |
1da177e4 LT |
2436 | goto out; |
2437 | ||
15e47304 | 2438 | NETLINK_CB(skb).portid = nlk->portid; |
d629b836 | 2439 | NETLINK_CB(skb).dst_group = dst_group; |
7cc05662 | 2440 | NETLINK_CB(skb).creds = scm.creds; |
2d7a85f4 | 2441 | NETLINK_CB(skb).flags = netlink_skb_flags; |
1da177e4 | 2442 | |
1da177e4 | 2443 | err = -EFAULT; |
6ce8e9ce | 2444 | if (memcpy_from_msg(skb_put(skb, len), msg, len)) { |
1da177e4 LT |
2445 | kfree_skb(skb); |
2446 | goto out; | |
2447 | } | |
2448 | ||
2449 | err = security_netlink_send(sk, skb); | |
2450 | if (err) { | |
2451 | kfree_skb(skb); | |
2452 | goto out; | |
2453 | } | |
2454 | ||
d629b836 | 2455 | if (dst_group) { |
1da177e4 | 2456 | atomic_inc(&skb->users); |
15e47304 | 2457 | netlink_broadcast(sk, skb, dst_portid, dst_group, GFP_KERNEL); |
1da177e4 | 2458 | } |
15e47304 | 2459 | err = netlink_unicast(sk, skb, dst_portid, msg->msg_flags&MSG_DONTWAIT); |
1da177e4 LT |
2460 | |
2461 | out: | |
7cc05662 | 2462 | scm_destroy(&scm); |
1da177e4 LT |
2463 | return err; |
2464 | } | |
2465 | ||
1b784140 | 2466 | static int netlink_recvmsg(struct socket *sock, struct msghdr *msg, size_t len, |
1da177e4 LT |
2467 | int flags) |
2468 | { | |
1da177e4 LT |
2469 | struct scm_cookie scm; |
2470 | struct sock *sk = sock->sk; | |
2471 | struct netlink_sock *nlk = nlk_sk(sk); | |
2472 | int noblock = flags&MSG_DONTWAIT; | |
2473 | size_t copied; | |
68d6ac6d | 2474 | struct sk_buff *skb, *data_skb; |
b44d211e | 2475 | int err, ret; |
1da177e4 LT |
2476 | |
2477 | if (flags&MSG_OOB) | |
2478 | return -EOPNOTSUPP; | |
2479 | ||
2480 | copied = 0; | |
2481 | ||
6ac552fd PM |
2482 | skb = skb_recv_datagram(sk, flags, noblock, &err); |
2483 | if (skb == NULL) | |
1da177e4 LT |
2484 | goto out; |
2485 | ||
68d6ac6d JB |
2486 | data_skb = skb; |
2487 | ||
1dacc76d JB |
2488 | #ifdef CONFIG_COMPAT_NETLINK_MESSAGES |
2489 | if (unlikely(skb_shinfo(skb)->frag_list)) { | |
1dacc76d | 2490 | /* |
68d6ac6d JB |
2491 | * If this skb has a frag_list, then here that means that we |
2492 | * will have to use the frag_list skb's data for compat tasks | |
2493 | * and the regular skb's data for normal (non-compat) tasks. | |
1dacc76d | 2494 | * |
68d6ac6d JB |
2495 | * If we need to send the compat skb, assign it to the |
2496 | * 'data_skb' variable so that it will be used below for data | |
2497 | * copying. We keep 'skb' for everything else, including | |
2498 | * freeing both later. | |
1dacc76d | 2499 | */ |
68d6ac6d JB |
2500 | if (flags & MSG_CMSG_COMPAT) |
2501 | data_skb = skb_shinfo(skb)->frag_list; | |
1dacc76d JB |
2502 | } |
2503 | #endif | |
2504 | ||
9063e21f ED |
2505 | /* Record the max length of recvmsg() calls for future allocations */ |
2506 | nlk->max_recvmsg_len = max(nlk->max_recvmsg_len, len); | |
2507 | nlk->max_recvmsg_len = min_t(size_t, nlk->max_recvmsg_len, | |
2508 | 16384); | |
2509 | ||
68d6ac6d | 2510 | copied = data_skb->len; |
1da177e4 LT |
2511 | if (len < copied) { |
2512 | msg->msg_flags |= MSG_TRUNC; | |
2513 | copied = len; | |
2514 | } | |
2515 | ||
68d6ac6d | 2516 | skb_reset_transport_header(data_skb); |
51f3d02b | 2517 | err = skb_copy_datagram_msg(data_skb, 0, msg, copied); |
1da177e4 LT |
2518 | |
2519 | if (msg->msg_name) { | |
342dfc30 | 2520 | DECLARE_SOCKADDR(struct sockaddr_nl *, addr, msg->msg_name); |
1da177e4 LT |
2521 | addr->nl_family = AF_NETLINK; |
2522 | addr->nl_pad = 0; | |
15e47304 | 2523 | addr->nl_pid = NETLINK_CB(skb).portid; |
d629b836 | 2524 | addr->nl_groups = netlink_group_mask(NETLINK_CB(skb).dst_group); |
1da177e4 LT |
2525 | msg->msg_namelen = sizeof(*addr); |
2526 | } | |
2527 | ||
cc3a572f | 2528 | if (nlk->flags & NETLINK_F_RECV_PKTINFO) |
cc9a06cd | 2529 | netlink_cmsg_recv_pktinfo(msg, skb); |
59324cf3 ND |
2530 | if (nlk->flags & NETLINK_F_LISTEN_ALL_NSID) |
2531 | netlink_cmsg_listen_all_nsid(sk, msg, skb); | |
cc9a06cd | 2532 | |
7cc05662 CH |
2533 | memset(&scm, 0, sizeof(scm)); |
2534 | scm.creds = *NETLINK_CREDS(skb); | |
188ccb55 | 2535 | if (flags & MSG_TRUNC) |
68d6ac6d | 2536 | copied = data_skb->len; |
daa3766e | 2537 | |
1da177e4 LT |
2538 | skb_free_datagram(sk, skb); |
2539 | ||
16b304f3 PS |
2540 | if (nlk->cb_running && |
2541 | atomic_read(&sk->sk_rmem_alloc) <= sk->sk_rcvbuf / 2) { | |
b44d211e AV |
2542 | ret = netlink_dump(sk); |
2543 | if (ret) { | |
ac30ef83 | 2544 | sk->sk_err = -ret; |
b44d211e AV |
2545 | sk->sk_error_report(sk); |
2546 | } | |
2547 | } | |
1da177e4 | 2548 | |
7cc05662 | 2549 | scm_recv(sock, msg, &scm, flags); |
1da177e4 LT |
2550 | out: |
2551 | netlink_rcv_wake(sk); | |
2552 | return err ? : copied; | |
2553 | } | |
2554 | ||
676d2369 | 2555 | static void netlink_data_ready(struct sock *sk) |
1da177e4 | 2556 | { |
cd40b7d3 | 2557 | BUG(); |
1da177e4 LT |
2558 | } |
2559 | ||
2560 | /* | |
746fac4d | 2561 | * We export these functions to other modules. They provide a |
1da177e4 LT |
2562 | * complete set of kernel non-blocking support for message |
2563 | * queueing. | |
2564 | */ | |
2565 | ||
2566 | struct sock * | |
9f00d977 PNA |
2567 | __netlink_kernel_create(struct net *net, int unit, struct module *module, |
2568 | struct netlink_kernel_cfg *cfg) | |
1da177e4 LT |
2569 | { |
2570 | struct socket *sock; | |
2571 | struct sock *sk; | |
77247bbb | 2572 | struct netlink_sock *nlk; |
5c398dc8 | 2573 | struct listeners *listeners = NULL; |
a31f2d17 PNA |
2574 | struct mutex *cb_mutex = cfg ? cfg->cb_mutex : NULL; |
2575 | unsigned int groups; | |
1da177e4 | 2576 | |
fab2caf6 | 2577 | BUG_ON(!nl_table); |
1da177e4 | 2578 | |
6ac552fd | 2579 | if (unit < 0 || unit >= MAX_LINKS) |
1da177e4 LT |
2580 | return NULL; |
2581 | ||
2582 | if (sock_create_lite(PF_NETLINK, SOCK_DGRAM, unit, &sock)) | |
2583 | return NULL; | |
13d3078e EB |
2584 | |
2585 | if (__netlink_create(net, sock, cb_mutex, unit, 1) < 0) | |
23fe1866 PE |
2586 | goto out_sock_release_nosk; |
2587 | ||
2588 | sk = sock->sk; | |
4fdb3bb7 | 2589 | |
a31f2d17 | 2590 | if (!cfg || cfg->groups < 32) |
4277a083 | 2591 | groups = 32; |
a31f2d17 PNA |
2592 | else |
2593 | groups = cfg->groups; | |
4277a083 | 2594 | |
5c398dc8 | 2595 | listeners = kzalloc(sizeof(*listeners) + NLGRPSZ(groups), GFP_KERNEL); |
4277a083 PM |
2596 | if (!listeners) |
2597 | goto out_sock_release; | |
2598 | ||
1da177e4 | 2599 | sk->sk_data_ready = netlink_data_ready; |
a31f2d17 PNA |
2600 | if (cfg && cfg->input) |
2601 | nlk_sk(sk)->netlink_rcv = cfg->input; | |
1da177e4 | 2602 | |
8ea65f4a | 2603 | if (netlink_insert(sk, 0)) |
77247bbb | 2604 | goto out_sock_release; |
4fdb3bb7 | 2605 | |
77247bbb | 2606 | nlk = nlk_sk(sk); |
cc3a572f | 2607 | nlk->flags |= NETLINK_F_KERNEL_SOCKET; |
4fdb3bb7 | 2608 | |
4fdb3bb7 | 2609 | netlink_table_grab(); |
b4b51029 EB |
2610 | if (!nl_table[unit].registered) { |
2611 | nl_table[unit].groups = groups; | |
5c398dc8 | 2612 | rcu_assign_pointer(nl_table[unit].listeners, listeners); |
b4b51029 EB |
2613 | nl_table[unit].cb_mutex = cb_mutex; |
2614 | nl_table[unit].module = module; | |
9785e10a PNA |
2615 | if (cfg) { |
2616 | nl_table[unit].bind = cfg->bind; | |
6251edd9 | 2617 | nl_table[unit].unbind = cfg->unbind; |
9785e10a | 2618 | nl_table[unit].flags = cfg->flags; |
da12c90e G |
2619 | if (cfg->compare) |
2620 | nl_table[unit].compare = cfg->compare; | |
9785e10a | 2621 | } |
b4b51029 | 2622 | nl_table[unit].registered = 1; |
f937f1f4 JJ |
2623 | } else { |
2624 | kfree(listeners); | |
869e58f8 | 2625 | nl_table[unit].registered++; |
b4b51029 | 2626 | } |
4fdb3bb7 | 2627 | netlink_table_ungrab(); |
77247bbb PM |
2628 | return sk; |
2629 | ||
4fdb3bb7 | 2630 | out_sock_release: |
4277a083 | 2631 | kfree(listeners); |
9dfbec1f | 2632 | netlink_kernel_release(sk); |
23fe1866 PE |
2633 | return NULL; |
2634 | ||
2635 | out_sock_release_nosk: | |
4fdb3bb7 | 2636 | sock_release(sock); |
77247bbb | 2637 | return NULL; |
1da177e4 | 2638 | } |
9f00d977 | 2639 | EXPORT_SYMBOL(__netlink_kernel_create); |
b7c6ba6e DL |
2640 | |
2641 | void | |
2642 | netlink_kernel_release(struct sock *sk) | |
2643 | { | |
13d3078e EB |
2644 | if (sk == NULL || sk->sk_socket == NULL) |
2645 | return; | |
2646 | ||
2647 | sock_release(sk->sk_socket); | |
b7c6ba6e DL |
2648 | } |
2649 | EXPORT_SYMBOL(netlink_kernel_release); | |
2650 | ||
d136f1bd | 2651 | int __netlink_change_ngroups(struct sock *sk, unsigned int groups) |
b4ff4f04 | 2652 | { |
5c398dc8 | 2653 | struct listeners *new, *old; |
b4ff4f04 | 2654 | struct netlink_table *tbl = &nl_table[sk->sk_protocol]; |
b4ff4f04 JB |
2655 | |
2656 | if (groups < 32) | |
2657 | groups = 32; | |
2658 | ||
b4ff4f04 | 2659 | if (NLGRPSZ(tbl->groups) < NLGRPSZ(groups)) { |
5c398dc8 ED |
2660 | new = kzalloc(sizeof(*new) + NLGRPSZ(groups), GFP_ATOMIC); |
2661 | if (!new) | |
d136f1bd | 2662 | return -ENOMEM; |
6d772ac5 | 2663 | old = nl_deref_protected(tbl->listeners); |
5c398dc8 ED |
2664 | memcpy(new->masks, old->masks, NLGRPSZ(tbl->groups)); |
2665 | rcu_assign_pointer(tbl->listeners, new); | |
2666 | ||
37b6b935 | 2667 | kfree_rcu(old, rcu); |
b4ff4f04 JB |
2668 | } |
2669 | tbl->groups = groups; | |
2670 | ||
d136f1bd JB |
2671 | return 0; |
2672 | } | |
2673 | ||
2674 | /** | |
2675 | * netlink_change_ngroups - change number of multicast groups | |
2676 | * | |
2677 | * This changes the number of multicast groups that are available | |
2678 | * on a certain netlink family. Note that it is not possible to | |
2679 | * change the number of groups to below 32. Also note that it does | |
2680 | * not implicitly call netlink_clear_multicast_users() when the | |
2681 | * number of groups is reduced. | |
2682 | * | |
2683 | * @sk: The kernel netlink socket, as returned by netlink_kernel_create(). | |
2684 | * @groups: The new number of groups. | |
2685 | */ | |
2686 | int netlink_change_ngroups(struct sock *sk, unsigned int groups) | |
2687 | { | |
2688 | int err; | |
2689 | ||
2690 | netlink_table_grab(); | |
2691 | err = __netlink_change_ngroups(sk, groups); | |
b4ff4f04 | 2692 | netlink_table_ungrab(); |
d136f1bd | 2693 | |
b4ff4f04 JB |
2694 | return err; |
2695 | } | |
b4ff4f04 | 2696 | |
b8273570 JB |
2697 | void __netlink_clear_multicast_users(struct sock *ksk, unsigned int group) |
2698 | { | |
2699 | struct sock *sk; | |
b8273570 JB |
2700 | struct netlink_table *tbl = &nl_table[ksk->sk_protocol]; |
2701 | ||
b67bfe0d | 2702 | sk_for_each_bound(sk, &tbl->mc_list) |
b8273570 JB |
2703 | netlink_update_socket_mc(nlk_sk(sk), group, 0); |
2704 | } | |
2705 | ||
a46621a3 | 2706 | struct nlmsghdr * |
15e47304 | 2707 | __nlmsg_put(struct sk_buff *skb, u32 portid, u32 seq, int type, int len, int flags) |
a46621a3 DV |
2708 | { |
2709 | struct nlmsghdr *nlh; | |
573ce260 | 2710 | int size = nlmsg_msg_size(len); |
a46621a3 | 2711 | |
23b45672 | 2712 | nlh = (struct nlmsghdr *)skb_put(skb, NLMSG_ALIGN(size)); |
a46621a3 DV |
2713 | nlh->nlmsg_type = type; |
2714 | nlh->nlmsg_len = size; | |
2715 | nlh->nlmsg_flags = flags; | |
15e47304 | 2716 | nlh->nlmsg_pid = portid; |
a46621a3 DV |
2717 | nlh->nlmsg_seq = seq; |
2718 | if (!__builtin_constant_p(size) || NLMSG_ALIGN(size) - size != 0) | |
573ce260 | 2719 | memset(nlmsg_data(nlh) + len, 0, NLMSG_ALIGN(size) - size); |
a46621a3 DV |
2720 | return nlh; |
2721 | } | |
2722 | EXPORT_SYMBOL(__nlmsg_put); | |
2723 | ||
1da177e4 LT |
2724 | /* |
2725 | * It looks a bit ugly. | |
2726 | * It would be better to create kernel thread. | |
2727 | */ | |
2728 | ||
2729 | static int netlink_dump(struct sock *sk) | |
2730 | { | |
2731 | struct netlink_sock *nlk = nlk_sk(sk); | |
2732 | struct netlink_callback *cb; | |
c7ac8679 | 2733 | struct sk_buff *skb = NULL; |
1da177e4 | 2734 | struct nlmsghdr *nlh; |
bf8b79e4 | 2735 | int len, err = -ENOBUFS; |
c7ac8679 | 2736 | int alloc_size; |
1da177e4 | 2737 | |
af65bdfc | 2738 | mutex_lock(nlk->cb_mutex); |
16b304f3 | 2739 | if (!nlk->cb_running) { |
bf8b79e4 TG |
2740 | err = -EINVAL; |
2741 | goto errout_skb; | |
1da177e4 LT |
2742 | } |
2743 | ||
16b304f3 | 2744 | cb = &nlk->cb; |
c7ac8679 GR |
2745 | alloc_size = max_t(int, cb->min_dump_alloc, NLMSG_GOODSIZE); |
2746 | ||
f9c22888 PM |
2747 | if (!netlink_rx_is_mmaped(sk) && |
2748 | atomic_read(&sk->sk_rmem_alloc) >= sk->sk_rcvbuf) | |
2749 | goto errout_skb; | |
9063e21f ED |
2750 | |
2751 | /* NLMSG_GOODSIZE is small to avoid high order allocations being | |
2752 | * required, but it makes sense to _attempt_ a 16K bytes allocation | |
2753 | * to reduce number of system calls on dump operations, if user | |
2754 | * ever provided a big enough buffer. | |
2755 | */ | |
2756 | if (alloc_size < nlk->max_recvmsg_len) { | |
2757 | skb = netlink_alloc_skb(sk, | |
2758 | nlk->max_recvmsg_len, | |
2759 | nlk->portid, | |
2760 | GFP_KERNEL | | |
2761 | __GFP_NOWARN | | |
2762 | __GFP_NORETRY); | |
2763 | /* available room should be exact amount to avoid MSG_TRUNC */ | |
2764 | if (skb) | |
2765 | skb_reserve(skb, skb_tailroom(skb) - | |
2766 | nlk->max_recvmsg_len); | |
2767 | } | |
2768 | if (!skb) | |
2769 | skb = netlink_alloc_skb(sk, alloc_size, nlk->portid, | |
2770 | GFP_KERNEL); | |
c7ac8679 | 2771 | if (!skb) |
c63d6ea3 | 2772 | goto errout_skb; |
f9c22888 | 2773 | netlink_skb_set_owner_r(skb, sk); |
c7ac8679 | 2774 | |
1da177e4 LT |
2775 | len = cb->dump(skb, cb); |
2776 | ||
2777 | if (len > 0) { | |
af65bdfc | 2778 | mutex_unlock(nlk->cb_mutex); |
b1153f29 SH |
2779 | |
2780 | if (sk_filter(sk, skb)) | |
2781 | kfree_skb(skb); | |
4a7e7c2a ED |
2782 | else |
2783 | __netlink_sendskb(sk, skb); | |
1da177e4 LT |
2784 | return 0; |
2785 | } | |
2786 | ||
bf8b79e4 TG |
2787 | nlh = nlmsg_put_answer(skb, cb, NLMSG_DONE, sizeof(len), NLM_F_MULTI); |
2788 | if (!nlh) | |
2789 | goto errout_skb; | |
2790 | ||
670dc283 JB |
2791 | nl_dump_check_consistent(cb, nlh); |
2792 | ||
bf8b79e4 TG |
2793 | memcpy(nlmsg_data(nlh), &len, sizeof(len)); |
2794 | ||
b1153f29 SH |
2795 | if (sk_filter(sk, skb)) |
2796 | kfree_skb(skb); | |
4a7e7c2a ED |
2797 | else |
2798 | __netlink_sendskb(sk, skb); | |
1da177e4 | 2799 | |
a8f74b22 TG |
2800 | if (cb->done) |
2801 | cb->done(cb); | |
1da177e4 | 2802 | |
16b304f3 PS |
2803 | nlk->cb_running = false; |
2804 | mutex_unlock(nlk->cb_mutex); | |
6dc878a8 | 2805 | module_put(cb->module); |
16b304f3 | 2806 | consume_skb(cb->skb); |
1da177e4 | 2807 | return 0; |
1797754e | 2808 | |
bf8b79e4 | 2809 | errout_skb: |
af65bdfc | 2810 | mutex_unlock(nlk->cb_mutex); |
bf8b79e4 | 2811 | kfree_skb(skb); |
bf8b79e4 | 2812 | return err; |
1da177e4 LT |
2813 | } |
2814 | ||
6dc878a8 G |
2815 | int __netlink_dump_start(struct sock *ssk, struct sk_buff *skb, |
2816 | const struct nlmsghdr *nlh, | |
2817 | struct netlink_dump_control *control) | |
1da177e4 LT |
2818 | { |
2819 | struct netlink_callback *cb; | |
2820 | struct sock *sk; | |
2821 | struct netlink_sock *nlk; | |
b44d211e | 2822 | int ret; |
1da177e4 | 2823 | |
f9c22888 PM |
2824 | /* Memory mapped dump requests need to be copied to avoid looping |
2825 | * on the pending state in netlink_mmap_sendmsg() while the CB hold | |
2826 | * a reference to the skb. | |
2827 | */ | |
2828 | if (netlink_skb_is_mmaped(skb)) { | |
2829 | skb = skb_copy(skb, GFP_KERNEL); | |
16b304f3 | 2830 | if (skb == NULL) |
f9c22888 | 2831 | return -ENOBUFS; |
f9c22888 PM |
2832 | } else |
2833 | atomic_inc(&skb->users); | |
2834 | ||
15e47304 | 2835 | sk = netlink_lookup(sock_net(ssk), ssk->sk_protocol, NETLINK_CB(skb).portid); |
1da177e4 | 2836 | if (sk == NULL) { |
16b304f3 PS |
2837 | ret = -ECONNREFUSED; |
2838 | goto error_free; | |
1da177e4 | 2839 | } |
6dc878a8 | 2840 | |
16b304f3 | 2841 | nlk = nlk_sk(sk); |
af65bdfc | 2842 | mutex_lock(nlk->cb_mutex); |
6dc878a8 | 2843 | /* A dump is in progress... */ |
16b304f3 | 2844 | if (nlk->cb_running) { |
6dc878a8 | 2845 | ret = -EBUSY; |
16b304f3 | 2846 | goto error_unlock; |
1da177e4 | 2847 | } |
6dc878a8 | 2848 | /* add reference of module which cb->dump belongs to */ |
16b304f3 | 2849 | if (!try_module_get(control->module)) { |
6dc878a8 | 2850 | ret = -EPROTONOSUPPORT; |
16b304f3 | 2851 | goto error_unlock; |
6dc878a8 G |
2852 | } |
2853 | ||
16b304f3 PS |
2854 | cb = &nlk->cb; |
2855 | memset(cb, 0, sizeof(*cb)); | |
2856 | cb->dump = control->dump; | |
2857 | cb->done = control->done; | |
2858 | cb->nlh = nlh; | |
2859 | cb->data = control->data; | |
2860 | cb->module = control->module; | |
2861 | cb->min_dump_alloc = control->min_dump_alloc; | |
2862 | cb->skb = skb; | |
2863 | ||
2864 | nlk->cb_running = true; | |
2865 | ||
af65bdfc | 2866 | mutex_unlock(nlk->cb_mutex); |
1da177e4 | 2867 | |
b44d211e | 2868 | ret = netlink_dump(sk); |
1da177e4 | 2869 | sock_put(sk); |
5c58298c | 2870 | |
b44d211e AV |
2871 | if (ret) |
2872 | return ret; | |
2873 | ||
5c58298c DL |
2874 | /* We successfully started a dump, by returning -EINTR we |
2875 | * signal not to send ACK even if it was requested. | |
2876 | */ | |
2877 | return -EINTR; | |
16b304f3 PS |
2878 | |
2879 | error_unlock: | |
2880 | sock_put(sk); | |
2881 | mutex_unlock(nlk->cb_mutex); | |
2882 | error_free: | |
2883 | kfree_skb(skb); | |
2884 | return ret; | |
1da177e4 | 2885 | } |
6dc878a8 | 2886 | EXPORT_SYMBOL(__netlink_dump_start); |
1da177e4 LT |
2887 | |
2888 | void netlink_ack(struct sk_buff *in_skb, struct nlmsghdr *nlh, int err) | |
2889 | { | |
2890 | struct sk_buff *skb; | |
2891 | struct nlmsghdr *rep; | |
2892 | struct nlmsgerr *errmsg; | |
339bf98f | 2893 | size_t payload = sizeof(*errmsg); |
0a6a3a23 | 2894 | struct netlink_sock *nlk = nlk_sk(NETLINK_CB(in_skb).sk); |
1da177e4 | 2895 | |
0a6a3a23 CR |
2896 | /* Error messages get the original request appened, unless the user |
2897 | * requests to cap the error message. | |
2898 | */ | |
2899 | if (!(nlk->flags & NETLINK_F_CAP_ACK) && err) | |
339bf98f | 2900 | payload += nlmsg_len(nlh); |
1da177e4 | 2901 | |
f9c22888 PM |
2902 | skb = netlink_alloc_skb(in_skb->sk, nlmsg_total_size(payload), |
2903 | NETLINK_CB(in_skb).portid, GFP_KERNEL); | |
1da177e4 LT |
2904 | if (!skb) { |
2905 | struct sock *sk; | |
2906 | ||
3b1e0a65 | 2907 | sk = netlink_lookup(sock_net(in_skb->sk), |
b4b51029 | 2908 | in_skb->sk->sk_protocol, |
15e47304 | 2909 | NETLINK_CB(in_skb).portid); |
1da177e4 LT |
2910 | if (sk) { |
2911 | sk->sk_err = ENOBUFS; | |
2912 | sk->sk_error_report(sk); | |
2913 | sock_put(sk); | |
2914 | } | |
2915 | return; | |
2916 | } | |
2917 | ||
15e47304 | 2918 | rep = __nlmsg_put(skb, NETLINK_CB(in_skb).portid, nlh->nlmsg_seq, |
5dba93ae | 2919 | NLMSG_ERROR, payload, 0); |
bf8b79e4 | 2920 | errmsg = nlmsg_data(rep); |
1da177e4 | 2921 | errmsg->error = err; |
0a6a3a23 | 2922 | memcpy(&errmsg->msg, nlh, payload > sizeof(*errmsg) ? nlh->nlmsg_len : sizeof(*nlh)); |
15e47304 | 2923 | netlink_unicast(in_skb->sk, skb, NETLINK_CB(in_skb).portid, MSG_DONTWAIT); |
1da177e4 | 2924 | } |
6ac552fd | 2925 | EXPORT_SYMBOL(netlink_ack); |
1da177e4 | 2926 | |
cd40b7d3 | 2927 | int netlink_rcv_skb(struct sk_buff *skb, int (*cb)(struct sk_buff *, |
1d00a4eb | 2928 | struct nlmsghdr *)) |
82ace47a | 2929 | { |
82ace47a TG |
2930 | struct nlmsghdr *nlh; |
2931 | int err; | |
2932 | ||
2933 | while (skb->len >= nlmsg_total_size(0)) { | |
cd40b7d3 DL |
2934 | int msglen; |
2935 | ||
b529ccf2 | 2936 | nlh = nlmsg_hdr(skb); |
d35b6856 | 2937 | err = 0; |
82ace47a | 2938 | |
ad8e4b75 | 2939 | if (nlh->nlmsg_len < NLMSG_HDRLEN || skb->len < nlh->nlmsg_len) |
82ace47a TG |
2940 | return 0; |
2941 | ||
d35b6856 TG |
2942 | /* Only requests are handled by the kernel */ |
2943 | if (!(nlh->nlmsg_flags & NLM_F_REQUEST)) | |
5c58298c | 2944 | goto ack; |
45e7ae7f TG |
2945 | |
2946 | /* Skip control messages */ | |
2947 | if (nlh->nlmsg_type < NLMSG_MIN_TYPE) | |
5c58298c | 2948 | goto ack; |
d35b6856 | 2949 | |
1d00a4eb | 2950 | err = cb(skb, nlh); |
5c58298c DL |
2951 | if (err == -EINTR) |
2952 | goto skip; | |
2953 | ||
2954 | ack: | |
d35b6856 | 2955 | if (nlh->nlmsg_flags & NLM_F_ACK || err) |
82ace47a | 2956 | netlink_ack(skb, nlh, err); |
82ace47a | 2957 | |
5c58298c | 2958 | skip: |
6ac552fd | 2959 | msglen = NLMSG_ALIGN(nlh->nlmsg_len); |
cd40b7d3 DL |
2960 | if (msglen > skb->len) |
2961 | msglen = skb->len; | |
2962 | skb_pull(skb, msglen); | |
82ace47a TG |
2963 | } |
2964 | ||
2965 | return 0; | |
2966 | } | |
6ac552fd | 2967 | EXPORT_SYMBOL(netlink_rcv_skb); |
82ace47a | 2968 | |
d387f6ad TG |
2969 | /** |
2970 | * nlmsg_notify - send a notification netlink message | |
2971 | * @sk: netlink socket to use | |
2972 | * @skb: notification message | |
15e47304 | 2973 | * @portid: destination netlink portid for reports or 0 |
d387f6ad TG |
2974 | * @group: destination multicast group or 0 |
2975 | * @report: 1 to report back, 0 to disable | |
2976 | * @flags: allocation flags | |
2977 | */ | |
15e47304 | 2978 | int nlmsg_notify(struct sock *sk, struct sk_buff *skb, u32 portid, |
d387f6ad TG |
2979 | unsigned int group, int report, gfp_t flags) |
2980 | { | |
2981 | int err = 0; | |
2982 | ||
2983 | if (group) { | |
15e47304 | 2984 | int exclude_portid = 0; |
d387f6ad TG |
2985 | |
2986 | if (report) { | |
2987 | atomic_inc(&skb->users); | |
15e47304 | 2988 | exclude_portid = portid; |
d387f6ad TG |
2989 | } |
2990 | ||
1ce85fe4 PNA |
2991 | /* errors reported via destination sk->sk_err, but propagate |
2992 | * delivery errors if NETLINK_BROADCAST_ERROR flag is set */ | |
15e47304 | 2993 | err = nlmsg_multicast(sk, skb, exclude_portid, group, flags); |
d387f6ad TG |
2994 | } |
2995 | ||
1ce85fe4 PNA |
2996 | if (report) { |
2997 | int err2; | |
2998 | ||
15e47304 | 2999 | err2 = nlmsg_unicast(sk, skb, portid); |
1ce85fe4 PNA |
3000 | if (!err || err == -ESRCH) |
3001 | err = err2; | |
3002 | } | |
d387f6ad TG |
3003 | |
3004 | return err; | |
3005 | } | |
6ac552fd | 3006 | EXPORT_SYMBOL(nlmsg_notify); |
d387f6ad | 3007 | |
1da177e4 LT |
3008 | #ifdef CONFIG_PROC_FS |
3009 | struct nl_seq_iter { | |
e372c414 | 3010 | struct seq_net_private p; |
56d28b1e | 3011 | struct rhashtable_iter hti; |
1da177e4 | 3012 | int link; |
1da177e4 LT |
3013 | }; |
3014 | ||
56d28b1e | 3015 | static int netlink_walk_start(struct nl_seq_iter *iter) |
1da177e4 | 3016 | { |
56d28b1e | 3017 | int err; |
1da177e4 | 3018 | |
56d28b1e HX |
3019 | err = rhashtable_walk_init(&nl_table[iter->link].hash, &iter->hti); |
3020 | if (err) { | |
3021 | iter->link = MAX_LINKS; | |
3022 | return err; | |
1da177e4 | 3023 | } |
56d28b1e HX |
3024 | |
3025 | err = rhashtable_walk_start(&iter->hti); | |
3026 | return err == -EAGAIN ? 0 : err; | |
1da177e4 LT |
3027 | } |
3028 | ||
56d28b1e | 3029 | static void netlink_walk_stop(struct nl_seq_iter *iter) |
1da177e4 | 3030 | { |
56d28b1e HX |
3031 | rhashtable_walk_stop(&iter->hti); |
3032 | rhashtable_walk_exit(&iter->hti); | |
1da177e4 LT |
3033 | } |
3034 | ||
56d28b1e | 3035 | static void *__netlink_seq_next(struct seq_file *seq) |
1da177e4 | 3036 | { |
56d28b1e | 3037 | struct nl_seq_iter *iter = seq->private; |
e341694e | 3038 | struct netlink_sock *nlk; |
1da177e4 | 3039 | |
56d28b1e HX |
3040 | do { |
3041 | for (;;) { | |
3042 | int err; | |
1da177e4 | 3043 | |
56d28b1e | 3044 | nlk = rhashtable_walk_next(&iter->hti); |
746fac4d | 3045 | |
56d28b1e HX |
3046 | if (IS_ERR(nlk)) { |
3047 | if (PTR_ERR(nlk) == -EAGAIN) | |
3048 | continue; | |
e341694e | 3049 | |
56d28b1e HX |
3050 | return nlk; |
3051 | } | |
1da177e4 | 3052 | |
56d28b1e HX |
3053 | if (nlk) |
3054 | break; | |
1da177e4 | 3055 | |
56d28b1e HX |
3056 | netlink_walk_stop(iter); |
3057 | if (++iter->link >= MAX_LINKS) | |
3058 | return NULL; | |
da12c90e | 3059 | |
56d28b1e HX |
3060 | err = netlink_walk_start(iter); |
3061 | if (err) | |
3062 | return ERR_PTR(err); | |
1da177e4 | 3063 | } |
56d28b1e | 3064 | } while (sock_net(&nlk->sk) != seq_file_net(seq)); |
1da177e4 | 3065 | |
56d28b1e HX |
3066 | return nlk; |
3067 | } | |
1da177e4 | 3068 | |
56d28b1e HX |
3069 | static void *netlink_seq_start(struct seq_file *seq, loff_t *posp) |
3070 | { | |
3071 | struct nl_seq_iter *iter = seq->private; | |
3072 | void *obj = SEQ_START_TOKEN; | |
3073 | loff_t pos; | |
3074 | int err; | |
3075 | ||
3076 | iter->link = 0; | |
3077 | ||
3078 | err = netlink_walk_start(iter); | |
3079 | if (err) | |
3080 | return ERR_PTR(err); | |
3081 | ||
3082 | for (pos = *posp; pos && obj && !IS_ERR(obj); pos--) | |
3083 | obj = __netlink_seq_next(seq); | |
3084 | ||
3085 | return obj; | |
3086 | } | |
3087 | ||
3088 | static void *netlink_seq_next(struct seq_file *seq, void *v, loff_t *pos) | |
3089 | { | |
3090 | ++*pos; | |
3091 | return __netlink_seq_next(seq); | |
1da177e4 LT |
3092 | } |
3093 | ||
3094 | static void netlink_seq_stop(struct seq_file *seq, void *v) | |
3095 | { | |
56d28b1e HX |
3096 | struct nl_seq_iter *iter = seq->private; |
3097 | ||
3098 | if (iter->link >= MAX_LINKS) | |
3099 | return; | |
3100 | ||
3101 | netlink_walk_stop(iter); | |
1da177e4 LT |
3102 | } |
3103 | ||
3104 | ||
3105 | static int netlink_seq_show(struct seq_file *seq, void *v) | |
3106 | { | |
658cb354 | 3107 | if (v == SEQ_START_TOKEN) { |
1da177e4 LT |
3108 | seq_puts(seq, |
3109 | "sk Eth Pid Groups " | |
cf0aa4e0 | 3110 | "Rmem Wmem Dump Locks Drops Inode\n"); |
658cb354 | 3111 | } else { |
1da177e4 LT |
3112 | struct sock *s = v; |
3113 | struct netlink_sock *nlk = nlk_sk(s); | |
3114 | ||
16b304f3 | 3115 | seq_printf(seq, "%pK %-3d %-6u %08x %-8d %-8d %d %-8d %-8d %-8lu\n", |
1da177e4 LT |
3116 | s, |
3117 | s->sk_protocol, | |
15e47304 | 3118 | nlk->portid, |
513c2500 | 3119 | nlk->groups ? (u32)nlk->groups[0] : 0, |
31e6d363 ED |
3120 | sk_rmem_alloc_get(s), |
3121 | sk_wmem_alloc_get(s), | |
16b304f3 | 3122 | nlk->cb_running, |
38938bfe | 3123 | atomic_read(&s->sk_refcnt), |
cf0aa4e0 MY |
3124 | atomic_read(&s->sk_drops), |
3125 | sock_i_ino(s) | |
1da177e4 LT |
3126 | ); |
3127 | ||
3128 | } | |
3129 | return 0; | |
3130 | } | |
3131 | ||
56b3d975 | 3132 | static const struct seq_operations netlink_seq_ops = { |
1da177e4 LT |
3133 | .start = netlink_seq_start, |
3134 | .next = netlink_seq_next, | |
3135 | .stop = netlink_seq_stop, | |
3136 | .show = netlink_seq_show, | |
3137 | }; | |
3138 | ||
3139 | ||
3140 | static int netlink_seq_open(struct inode *inode, struct file *file) | |
3141 | { | |
e372c414 DL |
3142 | return seq_open_net(inode, file, &netlink_seq_ops, |
3143 | sizeof(struct nl_seq_iter)); | |
b4b51029 EB |
3144 | } |
3145 | ||
da7071d7 | 3146 | static const struct file_operations netlink_seq_fops = { |
1da177e4 LT |
3147 | .owner = THIS_MODULE, |
3148 | .open = netlink_seq_open, | |
3149 | .read = seq_read, | |
3150 | .llseek = seq_lseek, | |
e372c414 | 3151 | .release = seq_release_net, |
1da177e4 LT |
3152 | }; |
3153 | ||
3154 | #endif | |
3155 | ||
3156 | int netlink_register_notifier(struct notifier_block *nb) | |
3157 | { | |
e041c683 | 3158 | return atomic_notifier_chain_register(&netlink_chain, nb); |
1da177e4 | 3159 | } |
6ac552fd | 3160 | EXPORT_SYMBOL(netlink_register_notifier); |
1da177e4 LT |
3161 | |
3162 | int netlink_unregister_notifier(struct notifier_block *nb) | |
3163 | { | |
e041c683 | 3164 | return atomic_notifier_chain_unregister(&netlink_chain, nb); |
1da177e4 | 3165 | } |
6ac552fd | 3166 | EXPORT_SYMBOL(netlink_unregister_notifier); |
746fac4d | 3167 | |
90ddc4f0 | 3168 | static const struct proto_ops netlink_ops = { |
1da177e4 LT |
3169 | .family = PF_NETLINK, |
3170 | .owner = THIS_MODULE, | |
3171 | .release = netlink_release, | |
3172 | .bind = netlink_bind, | |
3173 | .connect = netlink_connect, | |
3174 | .socketpair = sock_no_socketpair, | |
3175 | .accept = sock_no_accept, | |
3176 | .getname = netlink_getname, | |
9652e931 | 3177 | .poll = netlink_poll, |
1da177e4 LT |
3178 | .ioctl = sock_no_ioctl, |
3179 | .listen = sock_no_listen, | |
3180 | .shutdown = sock_no_shutdown, | |
9a4595bc PM |
3181 | .setsockopt = netlink_setsockopt, |
3182 | .getsockopt = netlink_getsockopt, | |
1da177e4 LT |
3183 | .sendmsg = netlink_sendmsg, |
3184 | .recvmsg = netlink_recvmsg, | |
ccdfcc39 | 3185 | .mmap = netlink_mmap, |
1da177e4 LT |
3186 | .sendpage = sock_no_sendpage, |
3187 | }; | |
3188 | ||
ec1b4cf7 | 3189 | static const struct net_proto_family netlink_family_ops = { |
1da177e4 LT |
3190 | .family = PF_NETLINK, |
3191 | .create = netlink_create, | |
3192 | .owner = THIS_MODULE, /* for consistency 8) */ | |
3193 | }; | |
3194 | ||
4665079c | 3195 | static int __net_init netlink_net_init(struct net *net) |
b4b51029 EB |
3196 | { |
3197 | #ifdef CONFIG_PROC_FS | |
d4beaa66 | 3198 | if (!proc_create("netlink", 0, net->proc_net, &netlink_seq_fops)) |
b4b51029 EB |
3199 | return -ENOMEM; |
3200 | #endif | |
3201 | return 0; | |
3202 | } | |
3203 | ||
4665079c | 3204 | static void __net_exit netlink_net_exit(struct net *net) |
b4b51029 EB |
3205 | { |
3206 | #ifdef CONFIG_PROC_FS | |
ece31ffd | 3207 | remove_proc_entry("netlink", net->proc_net); |
b4b51029 EB |
3208 | #endif |
3209 | } | |
3210 | ||
b963ea89 DM |
3211 | static void __init netlink_add_usersock_entry(void) |
3212 | { | |
5c398dc8 | 3213 | struct listeners *listeners; |
b963ea89 DM |
3214 | int groups = 32; |
3215 | ||
5c398dc8 | 3216 | listeners = kzalloc(sizeof(*listeners) + NLGRPSZ(groups), GFP_KERNEL); |
b963ea89 | 3217 | if (!listeners) |
5c398dc8 | 3218 | panic("netlink_add_usersock_entry: Cannot allocate listeners\n"); |
b963ea89 DM |
3219 | |
3220 | netlink_table_grab(); | |
3221 | ||
3222 | nl_table[NETLINK_USERSOCK].groups = groups; | |
5c398dc8 | 3223 | rcu_assign_pointer(nl_table[NETLINK_USERSOCK].listeners, listeners); |
b963ea89 DM |
3224 | nl_table[NETLINK_USERSOCK].module = THIS_MODULE; |
3225 | nl_table[NETLINK_USERSOCK].registered = 1; | |
9785e10a | 3226 | nl_table[NETLINK_USERSOCK].flags = NL_CFG_F_NONROOT_SEND; |
b963ea89 DM |
3227 | |
3228 | netlink_table_ungrab(); | |
3229 | } | |
3230 | ||
022cbae6 | 3231 | static struct pernet_operations __net_initdata netlink_net_ops = { |
b4b51029 EB |
3232 | .init = netlink_net_init, |
3233 | .exit = netlink_net_exit, | |
3234 | }; | |
3235 | ||
49f7b33e | 3236 | static inline u32 netlink_hash(const void *data, u32 len, u32 seed) |
c428ecd1 HX |
3237 | { |
3238 | const struct netlink_sock *nlk = data; | |
3239 | struct netlink_compare_arg arg; | |
3240 | ||
3241 | netlink_compare_arg_init(&arg, sock_net(&nlk->sk), nlk->portid); | |
11b58ba1 | 3242 | return jhash2((u32 *)&arg, netlink_compare_arg_len / sizeof(u32), seed); |
c428ecd1 HX |
3243 | } |
3244 | ||
3245 | static const struct rhashtable_params netlink_rhashtable_params = { | |
3246 | .head_offset = offsetof(struct netlink_sock, node), | |
3247 | .key_len = netlink_compare_arg_len, | |
c428ecd1 HX |
3248 | .obj_hashfn = netlink_hash, |
3249 | .obj_cmpfn = netlink_compare, | |
b5e2c150 | 3250 | .automatic_shrinking = true, |
c428ecd1 HX |
3251 | }; |
3252 | ||
1da177e4 LT |
3253 | static int __init netlink_proto_init(void) |
3254 | { | |
1da177e4 | 3255 | int i; |
1da177e4 LT |
3256 | int err = proto_register(&netlink_proto, 0); |
3257 | ||
3258 | if (err != 0) | |
3259 | goto out; | |
3260 | ||
fab25745 | 3261 | BUILD_BUG_ON(sizeof(struct netlink_skb_parms) > FIELD_SIZEOF(struct sk_buff, cb)); |
1da177e4 | 3262 | |
0da974f4 | 3263 | nl_table = kcalloc(MAX_LINKS, sizeof(*nl_table), GFP_KERNEL); |
fab2caf6 AM |
3264 | if (!nl_table) |
3265 | goto panic; | |
1da177e4 | 3266 | |
1da177e4 | 3267 | for (i = 0; i < MAX_LINKS; i++) { |
c428ecd1 HX |
3268 | if (rhashtable_init(&nl_table[i].hash, |
3269 | &netlink_rhashtable_params) < 0) { | |
e341694e TG |
3270 | while (--i > 0) |
3271 | rhashtable_destroy(&nl_table[i].hash); | |
1da177e4 | 3272 | kfree(nl_table); |
fab2caf6 | 3273 | goto panic; |
1da177e4 | 3274 | } |
1da177e4 LT |
3275 | } |
3276 | ||
bcbde0d4 DB |
3277 | INIT_LIST_HEAD(&netlink_tap_all); |
3278 | ||
b963ea89 DM |
3279 | netlink_add_usersock_entry(); |
3280 | ||
1da177e4 | 3281 | sock_register(&netlink_family_ops); |
b4b51029 | 3282 | register_pernet_subsys(&netlink_net_ops); |
746fac4d | 3283 | /* The netlink device handler may be needed early. */ |
1da177e4 LT |
3284 | rtnetlink_init(); |
3285 | out: | |
3286 | return err; | |
fab2caf6 AM |
3287 | panic: |
3288 | panic("netlink_init: Cannot allocate nl_table\n"); | |
1da177e4 LT |
3289 | } |
3290 | ||
1da177e4 | 3291 | core_initcall(netlink_proto_init); |