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> |
9652e931 | 60 | #include <asm/cacheflush.h> |
54e0f520 | 61 | |
457c4cbc | 62 | #include <net/net_namespace.h> |
1da177e4 LT |
63 | #include <net/sock.h> |
64 | #include <net/scm.h> | |
82ace47a | 65 | #include <net/netlink.h> |
1da177e4 | 66 | |
0f29c768 | 67 | #include "af_netlink.h" |
1da177e4 | 68 | |
5c398dc8 ED |
69 | struct listeners { |
70 | struct rcu_head rcu; | |
71 | unsigned long masks[0]; | |
6c04bb18 JB |
72 | }; |
73 | ||
cd967e05 PM |
74 | /* state bits */ |
75 | #define NETLINK_CONGESTED 0x0 | |
76 | ||
77 | /* flags */ | |
77247bbb | 78 | #define NETLINK_KERNEL_SOCKET 0x1 |
9a4595bc | 79 | #define NETLINK_RECV_PKTINFO 0x2 |
be0c22a4 | 80 | #define NETLINK_BROADCAST_SEND_ERROR 0x4 |
38938bfe | 81 | #define NETLINK_RECV_NO_ENOBUFS 0x8 |
77247bbb | 82 | |
035c4c16 | 83 | static inline int netlink_is_kernel(struct sock *sk) |
aed81560 DL |
84 | { |
85 | return nlk_sk(sk)->flags & NETLINK_KERNEL_SOCKET; | |
86 | } | |
87 | ||
0f29c768 AV |
88 | struct netlink_table *nl_table; |
89 | EXPORT_SYMBOL_GPL(nl_table); | |
1da177e4 LT |
90 | |
91 | static DECLARE_WAIT_QUEUE_HEAD(nl_table_wait); | |
92 | ||
93 | static int netlink_dump(struct sock *sk); | |
9652e931 | 94 | static void netlink_skb_destructor(struct sk_buff *skb); |
1da177e4 | 95 | |
0f29c768 AV |
96 | DEFINE_RWLOCK(nl_table_lock); |
97 | EXPORT_SYMBOL_GPL(nl_table_lock); | |
1da177e4 LT |
98 | static atomic_t nl_table_users = ATOMIC_INIT(0); |
99 | ||
6d772ac5 ED |
100 | #define nl_deref_protected(X) rcu_dereference_protected(X, lockdep_is_held(&nl_table_lock)); |
101 | ||
e041c683 | 102 | static ATOMIC_NOTIFIER_HEAD(netlink_chain); |
1da177e4 | 103 | |
b57ef81f | 104 | static inline u32 netlink_group_mask(u32 group) |
d629b836 PM |
105 | { |
106 | return group ? 1 << (group - 1) : 0; | |
107 | } | |
108 | ||
15e47304 | 109 | static inline struct hlist_head *nl_portid_hashfn(struct nl_portid_hash *hash, u32 portid) |
1da177e4 | 110 | { |
15e47304 | 111 | return &hash->table[jhash_1word(portid, hash->rnd) & hash->mask]; |
1da177e4 LT |
112 | } |
113 | ||
cd1df525 PM |
114 | static void netlink_overrun(struct sock *sk) |
115 | { | |
116 | struct netlink_sock *nlk = nlk_sk(sk); | |
117 | ||
118 | if (!(nlk->flags & NETLINK_RECV_NO_ENOBUFS)) { | |
119 | if (!test_and_set_bit(NETLINK_CONGESTED, &nlk_sk(sk)->state)) { | |
120 | sk->sk_err = ENOBUFS; | |
121 | sk->sk_error_report(sk); | |
122 | } | |
123 | } | |
124 | atomic_inc(&sk->sk_drops); | |
125 | } | |
126 | ||
127 | static void netlink_rcv_wake(struct sock *sk) | |
128 | { | |
129 | struct netlink_sock *nlk = nlk_sk(sk); | |
130 | ||
131 | if (skb_queue_empty(&sk->sk_receive_queue)) | |
132 | clear_bit(NETLINK_CONGESTED, &nlk->state); | |
133 | if (!test_bit(NETLINK_CONGESTED, &nlk->state)) | |
134 | wake_up_interruptible(&nlk->wait); | |
135 | } | |
136 | ||
ccdfcc39 | 137 | #ifdef CONFIG_NETLINK_MMAP |
9652e931 PM |
138 | static bool netlink_skb_is_mmaped(const struct sk_buff *skb) |
139 | { | |
140 | return NETLINK_CB(skb).flags & NETLINK_SKB_MMAPED; | |
141 | } | |
142 | ||
f9c22888 PM |
143 | static bool netlink_rx_is_mmaped(struct sock *sk) |
144 | { | |
145 | return nlk_sk(sk)->rx_ring.pg_vec != NULL; | |
146 | } | |
147 | ||
5fd96123 PM |
148 | static bool netlink_tx_is_mmaped(struct sock *sk) |
149 | { | |
150 | return nlk_sk(sk)->tx_ring.pg_vec != NULL; | |
151 | } | |
152 | ||
ccdfcc39 PM |
153 | static __pure struct page *pgvec_to_page(const void *addr) |
154 | { | |
155 | if (is_vmalloc_addr(addr)) | |
156 | return vmalloc_to_page(addr); | |
157 | else | |
158 | return virt_to_page(addr); | |
159 | } | |
160 | ||
161 | static void free_pg_vec(void **pg_vec, unsigned int order, unsigned int len) | |
162 | { | |
163 | unsigned int i; | |
164 | ||
165 | for (i = 0; i < len; i++) { | |
166 | if (pg_vec[i] != NULL) { | |
167 | if (is_vmalloc_addr(pg_vec[i])) | |
168 | vfree(pg_vec[i]); | |
169 | else | |
170 | free_pages((unsigned long)pg_vec[i], order); | |
171 | } | |
172 | } | |
173 | kfree(pg_vec); | |
174 | } | |
175 | ||
176 | static void *alloc_one_pg_vec_page(unsigned long order) | |
177 | { | |
178 | void *buffer; | |
179 | gfp_t gfp_flags = GFP_KERNEL | __GFP_COMP | __GFP_ZERO | | |
180 | __GFP_NOWARN | __GFP_NORETRY; | |
181 | ||
182 | buffer = (void *)__get_free_pages(gfp_flags, order); | |
183 | if (buffer != NULL) | |
184 | return buffer; | |
185 | ||
186 | buffer = vzalloc((1 << order) * PAGE_SIZE); | |
187 | if (buffer != NULL) | |
188 | return buffer; | |
189 | ||
190 | gfp_flags &= ~__GFP_NORETRY; | |
191 | return (void *)__get_free_pages(gfp_flags, order); | |
192 | } | |
193 | ||
194 | static void **alloc_pg_vec(struct netlink_sock *nlk, | |
195 | struct nl_mmap_req *req, unsigned int order) | |
196 | { | |
197 | unsigned int block_nr = req->nm_block_nr; | |
198 | unsigned int i; | |
199 | void **pg_vec, *ptr; | |
200 | ||
201 | pg_vec = kcalloc(block_nr, sizeof(void *), GFP_KERNEL); | |
202 | if (pg_vec == NULL) | |
203 | return NULL; | |
204 | ||
205 | for (i = 0; i < block_nr; i++) { | |
206 | pg_vec[i] = ptr = alloc_one_pg_vec_page(order); | |
207 | if (pg_vec[i] == NULL) | |
208 | goto err1; | |
209 | } | |
210 | ||
211 | return pg_vec; | |
212 | err1: | |
213 | free_pg_vec(pg_vec, order, block_nr); | |
214 | return NULL; | |
215 | } | |
216 | ||
217 | static int netlink_set_ring(struct sock *sk, struct nl_mmap_req *req, | |
218 | bool closing, bool tx_ring) | |
219 | { | |
220 | struct netlink_sock *nlk = nlk_sk(sk); | |
221 | struct netlink_ring *ring; | |
222 | struct sk_buff_head *queue; | |
223 | void **pg_vec = NULL; | |
224 | unsigned int order = 0; | |
225 | int err; | |
226 | ||
227 | ring = tx_ring ? &nlk->tx_ring : &nlk->rx_ring; | |
228 | queue = tx_ring ? &sk->sk_write_queue : &sk->sk_receive_queue; | |
229 | ||
230 | if (!closing) { | |
231 | if (atomic_read(&nlk->mapped)) | |
232 | return -EBUSY; | |
233 | if (atomic_read(&ring->pending)) | |
234 | return -EBUSY; | |
235 | } | |
236 | ||
237 | if (req->nm_block_nr) { | |
238 | if (ring->pg_vec != NULL) | |
239 | return -EBUSY; | |
240 | ||
241 | if ((int)req->nm_block_size <= 0) | |
242 | return -EINVAL; | |
243 | if (!IS_ALIGNED(req->nm_block_size, PAGE_SIZE)) | |
244 | return -EINVAL; | |
245 | if (req->nm_frame_size < NL_MMAP_HDRLEN) | |
246 | return -EINVAL; | |
247 | if (!IS_ALIGNED(req->nm_frame_size, NL_MMAP_MSG_ALIGNMENT)) | |
248 | return -EINVAL; | |
249 | ||
250 | ring->frames_per_block = req->nm_block_size / | |
251 | req->nm_frame_size; | |
252 | if (ring->frames_per_block == 0) | |
253 | return -EINVAL; | |
254 | if (ring->frames_per_block * req->nm_block_nr != | |
255 | req->nm_frame_nr) | |
256 | return -EINVAL; | |
257 | ||
258 | order = get_order(req->nm_block_size); | |
259 | pg_vec = alloc_pg_vec(nlk, req, order); | |
260 | if (pg_vec == NULL) | |
261 | return -ENOMEM; | |
262 | } else { | |
263 | if (req->nm_frame_nr) | |
264 | return -EINVAL; | |
265 | } | |
266 | ||
267 | err = -EBUSY; | |
268 | mutex_lock(&nlk->pg_vec_lock); | |
269 | if (closing || atomic_read(&nlk->mapped) == 0) { | |
270 | err = 0; | |
271 | spin_lock_bh(&queue->lock); | |
272 | ||
273 | ring->frame_max = req->nm_frame_nr - 1; | |
274 | ring->head = 0; | |
275 | ring->frame_size = req->nm_frame_size; | |
276 | ring->pg_vec_pages = req->nm_block_size / PAGE_SIZE; | |
277 | ||
278 | swap(ring->pg_vec_len, req->nm_block_nr); | |
279 | swap(ring->pg_vec_order, order); | |
280 | swap(ring->pg_vec, pg_vec); | |
281 | ||
282 | __skb_queue_purge(queue); | |
283 | spin_unlock_bh(&queue->lock); | |
284 | ||
285 | WARN_ON(atomic_read(&nlk->mapped)); | |
286 | } | |
287 | mutex_unlock(&nlk->pg_vec_lock); | |
288 | ||
289 | if (pg_vec) | |
290 | free_pg_vec(pg_vec, order, req->nm_block_nr); | |
291 | return err; | |
292 | } | |
293 | ||
294 | static void netlink_mm_open(struct vm_area_struct *vma) | |
295 | { | |
296 | struct file *file = vma->vm_file; | |
297 | struct socket *sock = file->private_data; | |
298 | struct sock *sk = sock->sk; | |
299 | ||
300 | if (sk) | |
301 | atomic_inc(&nlk_sk(sk)->mapped); | |
302 | } | |
303 | ||
304 | static void netlink_mm_close(struct vm_area_struct *vma) | |
305 | { | |
306 | struct file *file = vma->vm_file; | |
307 | struct socket *sock = file->private_data; | |
308 | struct sock *sk = sock->sk; | |
309 | ||
310 | if (sk) | |
311 | atomic_dec(&nlk_sk(sk)->mapped); | |
312 | } | |
313 | ||
314 | static const struct vm_operations_struct netlink_mmap_ops = { | |
315 | .open = netlink_mm_open, | |
316 | .close = netlink_mm_close, | |
317 | }; | |
318 | ||
319 | static int netlink_mmap(struct file *file, struct socket *sock, | |
320 | struct vm_area_struct *vma) | |
321 | { | |
322 | struct sock *sk = sock->sk; | |
323 | struct netlink_sock *nlk = nlk_sk(sk); | |
324 | struct netlink_ring *ring; | |
325 | unsigned long start, size, expected; | |
326 | unsigned int i; | |
327 | int err = -EINVAL; | |
328 | ||
329 | if (vma->vm_pgoff) | |
330 | return -EINVAL; | |
331 | ||
332 | mutex_lock(&nlk->pg_vec_lock); | |
333 | ||
334 | expected = 0; | |
335 | for (ring = &nlk->rx_ring; ring <= &nlk->tx_ring; ring++) { | |
336 | if (ring->pg_vec == NULL) | |
337 | continue; | |
338 | expected += ring->pg_vec_len * ring->pg_vec_pages * PAGE_SIZE; | |
339 | } | |
340 | ||
341 | if (expected == 0) | |
342 | goto out; | |
343 | ||
344 | size = vma->vm_end - vma->vm_start; | |
345 | if (size != expected) | |
346 | goto out; | |
347 | ||
348 | start = vma->vm_start; | |
349 | for (ring = &nlk->rx_ring; ring <= &nlk->tx_ring; ring++) { | |
350 | if (ring->pg_vec == NULL) | |
351 | continue; | |
352 | ||
353 | for (i = 0; i < ring->pg_vec_len; i++) { | |
354 | struct page *page; | |
355 | void *kaddr = ring->pg_vec[i]; | |
356 | unsigned int pg_num; | |
357 | ||
358 | for (pg_num = 0; pg_num < ring->pg_vec_pages; pg_num++) { | |
359 | page = pgvec_to_page(kaddr); | |
360 | err = vm_insert_page(vma, start, page); | |
361 | if (err < 0) | |
362 | goto out; | |
363 | start += PAGE_SIZE; | |
364 | kaddr += PAGE_SIZE; | |
365 | } | |
366 | } | |
367 | } | |
368 | ||
369 | atomic_inc(&nlk->mapped); | |
370 | vma->vm_ops = &netlink_mmap_ops; | |
371 | err = 0; | |
372 | out: | |
373 | mutex_unlock(&nlk->pg_vec_lock); | |
374 | return 0; | |
375 | } | |
9652e931 PM |
376 | |
377 | static void netlink_frame_flush_dcache(const struct nl_mmap_hdr *hdr) | |
378 | { | |
379 | #if ARCH_IMPLEMENTS_FLUSH_DCACHE_PAGE == 1 | |
380 | struct page *p_start, *p_end; | |
381 | ||
382 | /* First page is flushed through netlink_{get,set}_status */ | |
383 | p_start = pgvec_to_page(hdr + PAGE_SIZE); | |
1d5085cb | 384 | p_end = pgvec_to_page((void *)hdr + NL_MMAP_HDRLEN + hdr->nm_len - 1); |
9652e931 PM |
385 | while (p_start <= p_end) { |
386 | flush_dcache_page(p_start); | |
387 | p_start++; | |
388 | } | |
389 | #endif | |
390 | } | |
391 | ||
392 | static enum nl_mmap_status netlink_get_status(const struct nl_mmap_hdr *hdr) | |
393 | { | |
394 | smp_rmb(); | |
395 | flush_dcache_page(pgvec_to_page(hdr)); | |
396 | return hdr->nm_status; | |
397 | } | |
398 | ||
399 | static void netlink_set_status(struct nl_mmap_hdr *hdr, | |
400 | enum nl_mmap_status status) | |
401 | { | |
402 | hdr->nm_status = status; | |
403 | flush_dcache_page(pgvec_to_page(hdr)); | |
404 | smp_wmb(); | |
405 | } | |
406 | ||
407 | static struct nl_mmap_hdr * | |
408 | __netlink_lookup_frame(const struct netlink_ring *ring, unsigned int pos) | |
409 | { | |
410 | unsigned int pg_vec_pos, frame_off; | |
411 | ||
412 | pg_vec_pos = pos / ring->frames_per_block; | |
413 | frame_off = pos % ring->frames_per_block; | |
414 | ||
415 | return ring->pg_vec[pg_vec_pos] + (frame_off * ring->frame_size); | |
416 | } | |
417 | ||
418 | static struct nl_mmap_hdr * | |
419 | netlink_lookup_frame(const struct netlink_ring *ring, unsigned int pos, | |
420 | enum nl_mmap_status status) | |
421 | { | |
422 | struct nl_mmap_hdr *hdr; | |
423 | ||
424 | hdr = __netlink_lookup_frame(ring, pos); | |
425 | if (netlink_get_status(hdr) != status) | |
426 | return NULL; | |
427 | ||
428 | return hdr; | |
429 | } | |
430 | ||
431 | static struct nl_mmap_hdr * | |
432 | netlink_current_frame(const struct netlink_ring *ring, | |
433 | enum nl_mmap_status status) | |
434 | { | |
435 | return netlink_lookup_frame(ring, ring->head, status); | |
436 | } | |
437 | ||
438 | static struct nl_mmap_hdr * | |
439 | netlink_previous_frame(const struct netlink_ring *ring, | |
440 | enum nl_mmap_status status) | |
441 | { | |
442 | unsigned int prev; | |
443 | ||
444 | prev = ring->head ? ring->head - 1 : ring->frame_max; | |
445 | return netlink_lookup_frame(ring, prev, status); | |
446 | } | |
447 | ||
448 | static void netlink_increment_head(struct netlink_ring *ring) | |
449 | { | |
450 | ring->head = ring->head != ring->frame_max ? ring->head + 1 : 0; | |
451 | } | |
452 | ||
453 | static void netlink_forward_ring(struct netlink_ring *ring) | |
454 | { | |
455 | unsigned int head = ring->head, pos = head; | |
456 | const struct nl_mmap_hdr *hdr; | |
457 | ||
458 | do { | |
459 | hdr = __netlink_lookup_frame(ring, pos); | |
460 | if (hdr->nm_status == NL_MMAP_STATUS_UNUSED) | |
461 | break; | |
462 | if (hdr->nm_status != NL_MMAP_STATUS_SKIP) | |
463 | break; | |
464 | netlink_increment_head(ring); | |
465 | } while (ring->head != head); | |
466 | } | |
467 | ||
cd1df525 PM |
468 | static bool netlink_dump_space(struct netlink_sock *nlk) |
469 | { | |
470 | struct netlink_ring *ring = &nlk->rx_ring; | |
471 | struct nl_mmap_hdr *hdr; | |
472 | unsigned int n; | |
473 | ||
474 | hdr = netlink_current_frame(ring, NL_MMAP_STATUS_UNUSED); | |
475 | if (hdr == NULL) | |
476 | return false; | |
477 | ||
478 | n = ring->head + ring->frame_max / 2; | |
479 | if (n > ring->frame_max) | |
480 | n -= ring->frame_max; | |
481 | ||
482 | hdr = __netlink_lookup_frame(ring, n); | |
483 | ||
484 | return hdr->nm_status == NL_MMAP_STATUS_UNUSED; | |
485 | } | |
486 | ||
9652e931 PM |
487 | static unsigned int netlink_poll(struct file *file, struct socket *sock, |
488 | poll_table *wait) | |
489 | { | |
490 | struct sock *sk = sock->sk; | |
491 | struct netlink_sock *nlk = nlk_sk(sk); | |
492 | unsigned int mask; | |
cd1df525 | 493 | int err; |
9652e931 | 494 | |
cd1df525 PM |
495 | if (nlk->rx_ring.pg_vec != NULL) { |
496 | /* Memory mapped sockets don't call recvmsg(), so flow control | |
497 | * for dumps is performed here. A dump is allowed to continue | |
498 | * if at least half the ring is unused. | |
499 | */ | |
500 | while (nlk->cb != NULL && netlink_dump_space(nlk)) { | |
501 | err = netlink_dump(sk); | |
502 | if (err < 0) { | |
503 | sk->sk_err = err; | |
504 | sk->sk_error_report(sk); | |
505 | break; | |
506 | } | |
507 | } | |
508 | netlink_rcv_wake(sk); | |
509 | } | |
5fd96123 | 510 | |
9652e931 PM |
511 | mask = datagram_poll(file, sock, wait); |
512 | ||
513 | spin_lock_bh(&sk->sk_receive_queue.lock); | |
514 | if (nlk->rx_ring.pg_vec) { | |
515 | netlink_forward_ring(&nlk->rx_ring); | |
516 | if (!netlink_previous_frame(&nlk->rx_ring, NL_MMAP_STATUS_UNUSED)) | |
517 | mask |= POLLIN | POLLRDNORM; | |
518 | } | |
519 | spin_unlock_bh(&sk->sk_receive_queue.lock); | |
520 | ||
521 | spin_lock_bh(&sk->sk_write_queue.lock); | |
522 | if (nlk->tx_ring.pg_vec) { | |
523 | if (netlink_current_frame(&nlk->tx_ring, NL_MMAP_STATUS_UNUSED)) | |
524 | mask |= POLLOUT | POLLWRNORM; | |
525 | } | |
526 | spin_unlock_bh(&sk->sk_write_queue.lock); | |
527 | ||
528 | return mask; | |
529 | } | |
530 | ||
531 | static struct nl_mmap_hdr *netlink_mmap_hdr(struct sk_buff *skb) | |
532 | { | |
533 | return (struct nl_mmap_hdr *)(skb->head - NL_MMAP_HDRLEN); | |
534 | } | |
535 | ||
536 | static void netlink_ring_setup_skb(struct sk_buff *skb, struct sock *sk, | |
537 | struct netlink_ring *ring, | |
538 | struct nl_mmap_hdr *hdr) | |
539 | { | |
540 | unsigned int size; | |
541 | void *data; | |
542 | ||
543 | size = ring->frame_size - NL_MMAP_HDRLEN; | |
544 | data = (void *)hdr + NL_MMAP_HDRLEN; | |
545 | ||
546 | skb->head = data; | |
547 | skb->data = data; | |
548 | skb_reset_tail_pointer(skb); | |
549 | skb->end = skb->tail + size; | |
550 | skb->len = 0; | |
551 | ||
552 | skb->destructor = netlink_skb_destructor; | |
553 | NETLINK_CB(skb).flags |= NETLINK_SKB_MMAPED; | |
554 | NETLINK_CB(skb).sk = sk; | |
555 | } | |
5fd96123 PM |
556 | |
557 | static int netlink_mmap_sendmsg(struct sock *sk, struct msghdr *msg, | |
558 | u32 dst_portid, u32 dst_group, | |
559 | struct sock_iocb *siocb) | |
560 | { | |
561 | struct netlink_sock *nlk = nlk_sk(sk); | |
562 | struct netlink_ring *ring; | |
563 | struct nl_mmap_hdr *hdr; | |
564 | struct sk_buff *skb; | |
565 | unsigned int maxlen; | |
566 | bool excl = true; | |
567 | int err = 0, len = 0; | |
568 | ||
569 | /* Netlink messages are validated by the receiver before processing. | |
570 | * In order to avoid userspace changing the contents of the message | |
571 | * after validation, the socket and the ring may only be used by a | |
572 | * single process, otherwise we fall back to copying. | |
573 | */ | |
574 | if (atomic_long_read(&sk->sk_socket->file->f_count) > 2 || | |
575 | atomic_read(&nlk->mapped) > 1) | |
576 | excl = false; | |
577 | ||
578 | mutex_lock(&nlk->pg_vec_lock); | |
579 | ||
580 | ring = &nlk->tx_ring; | |
581 | maxlen = ring->frame_size - NL_MMAP_HDRLEN; | |
582 | ||
583 | do { | |
584 | hdr = netlink_current_frame(ring, NL_MMAP_STATUS_VALID); | |
585 | if (hdr == NULL) { | |
586 | if (!(msg->msg_flags & MSG_DONTWAIT) && | |
587 | atomic_read(&nlk->tx_ring.pending)) | |
588 | schedule(); | |
589 | continue; | |
590 | } | |
591 | if (hdr->nm_len > maxlen) { | |
592 | err = -EINVAL; | |
593 | goto out; | |
594 | } | |
595 | ||
596 | netlink_frame_flush_dcache(hdr); | |
597 | ||
598 | if (likely(dst_portid == 0 && dst_group == 0 && excl)) { | |
599 | skb = alloc_skb_head(GFP_KERNEL); | |
600 | if (skb == NULL) { | |
601 | err = -ENOBUFS; | |
602 | goto out; | |
603 | } | |
604 | sock_hold(sk); | |
605 | netlink_ring_setup_skb(skb, sk, ring, hdr); | |
606 | NETLINK_CB(skb).flags |= NETLINK_SKB_TX; | |
607 | __skb_put(skb, hdr->nm_len); | |
608 | netlink_set_status(hdr, NL_MMAP_STATUS_RESERVED); | |
609 | atomic_inc(&ring->pending); | |
610 | } else { | |
611 | skb = alloc_skb(hdr->nm_len, GFP_KERNEL); | |
612 | if (skb == NULL) { | |
613 | err = -ENOBUFS; | |
614 | goto out; | |
615 | } | |
616 | __skb_put(skb, hdr->nm_len); | |
617 | memcpy(skb->data, (void *)hdr + NL_MMAP_HDRLEN, hdr->nm_len); | |
618 | netlink_set_status(hdr, NL_MMAP_STATUS_UNUSED); | |
619 | } | |
620 | ||
621 | netlink_increment_head(ring); | |
622 | ||
623 | NETLINK_CB(skb).portid = nlk->portid; | |
624 | NETLINK_CB(skb).dst_group = dst_group; | |
625 | NETLINK_CB(skb).creds = siocb->scm->creds; | |
626 | ||
627 | err = security_netlink_send(sk, skb); | |
628 | if (err) { | |
629 | kfree_skb(skb); | |
630 | goto out; | |
631 | } | |
632 | ||
633 | if (unlikely(dst_group)) { | |
634 | atomic_inc(&skb->users); | |
635 | netlink_broadcast(sk, skb, dst_portid, dst_group, | |
636 | GFP_KERNEL); | |
637 | } | |
638 | err = netlink_unicast(sk, skb, dst_portid, | |
639 | msg->msg_flags & MSG_DONTWAIT); | |
640 | if (err < 0) | |
641 | goto out; | |
642 | len += err; | |
643 | ||
644 | } while (hdr != NULL || | |
645 | (!(msg->msg_flags & MSG_DONTWAIT) && | |
646 | atomic_read(&nlk->tx_ring.pending))); | |
647 | ||
648 | if (len > 0) | |
649 | err = len; | |
650 | out: | |
651 | mutex_unlock(&nlk->pg_vec_lock); | |
652 | return err; | |
653 | } | |
f9c22888 PM |
654 | |
655 | static void netlink_queue_mmaped_skb(struct sock *sk, struct sk_buff *skb) | |
656 | { | |
657 | struct nl_mmap_hdr *hdr; | |
658 | ||
659 | hdr = netlink_mmap_hdr(skb); | |
660 | hdr->nm_len = skb->len; | |
661 | hdr->nm_group = NETLINK_CB(skb).dst_group; | |
662 | hdr->nm_pid = NETLINK_CB(skb).creds.pid; | |
1bf9310a ND |
663 | hdr->nm_uid = from_kuid(sk_user_ns(sk), NETLINK_CB(skb).creds.uid); |
664 | hdr->nm_gid = from_kgid(sk_user_ns(sk), NETLINK_CB(skb).creds.gid); | |
f9c22888 PM |
665 | netlink_frame_flush_dcache(hdr); |
666 | netlink_set_status(hdr, NL_MMAP_STATUS_VALID); | |
667 | ||
668 | NETLINK_CB(skb).flags |= NETLINK_SKB_DELIVERED; | |
669 | kfree_skb(skb); | |
670 | } | |
671 | ||
672 | static void netlink_ring_set_copied(struct sock *sk, struct sk_buff *skb) | |
673 | { | |
674 | struct netlink_sock *nlk = nlk_sk(sk); | |
675 | struct netlink_ring *ring = &nlk->rx_ring; | |
676 | struct nl_mmap_hdr *hdr; | |
677 | ||
678 | spin_lock_bh(&sk->sk_receive_queue.lock); | |
679 | hdr = netlink_current_frame(ring, NL_MMAP_STATUS_UNUSED); | |
680 | if (hdr == NULL) { | |
681 | spin_unlock_bh(&sk->sk_receive_queue.lock); | |
682 | kfree_skb(skb); | |
cd1df525 | 683 | netlink_overrun(sk); |
f9c22888 PM |
684 | return; |
685 | } | |
686 | netlink_increment_head(ring); | |
687 | __skb_queue_tail(&sk->sk_receive_queue, skb); | |
688 | spin_unlock_bh(&sk->sk_receive_queue.lock); | |
689 | ||
690 | hdr->nm_len = skb->len; | |
691 | hdr->nm_group = NETLINK_CB(skb).dst_group; | |
692 | hdr->nm_pid = NETLINK_CB(skb).creds.pid; | |
1bf9310a ND |
693 | hdr->nm_uid = from_kuid(sk_user_ns(sk), NETLINK_CB(skb).creds.uid); |
694 | hdr->nm_gid = from_kgid(sk_user_ns(sk), NETLINK_CB(skb).creds.gid); | |
f9c22888 PM |
695 | netlink_set_status(hdr, NL_MMAP_STATUS_COPY); |
696 | } | |
697 | ||
ccdfcc39 | 698 | #else /* CONFIG_NETLINK_MMAP */ |
9652e931 | 699 | #define netlink_skb_is_mmaped(skb) false |
f9c22888 | 700 | #define netlink_rx_is_mmaped(sk) false |
5fd96123 | 701 | #define netlink_tx_is_mmaped(sk) false |
ccdfcc39 | 702 | #define netlink_mmap sock_no_mmap |
9652e931 | 703 | #define netlink_poll datagram_poll |
5fd96123 | 704 | #define netlink_mmap_sendmsg(sk, msg, dst_portid, dst_group, siocb) 0 |
ccdfcc39 PM |
705 | #endif /* CONFIG_NETLINK_MMAP */ |
706 | ||
658cb354 ED |
707 | static void netlink_destroy_callback(struct netlink_callback *cb) |
708 | { | |
709 | kfree_skb(cb->skb); | |
710 | kfree(cb); | |
711 | } | |
712 | ||
bfb253c9 ED |
713 | static void netlink_consume_callback(struct netlink_callback *cb) |
714 | { | |
715 | consume_skb(cb->skb); | |
716 | kfree(cb); | |
717 | } | |
718 | ||
cf0a018a PM |
719 | static void netlink_skb_destructor(struct sk_buff *skb) |
720 | { | |
9652e931 PM |
721 | #ifdef CONFIG_NETLINK_MMAP |
722 | struct nl_mmap_hdr *hdr; | |
723 | struct netlink_ring *ring; | |
724 | struct sock *sk; | |
725 | ||
726 | /* If a packet from the kernel to userspace was freed because of an | |
727 | * error without being delivered to userspace, the kernel must reset | |
728 | * the status. In the direction userspace to kernel, the status is | |
729 | * always reset here after the packet was processed and freed. | |
730 | */ | |
731 | if (netlink_skb_is_mmaped(skb)) { | |
732 | hdr = netlink_mmap_hdr(skb); | |
733 | sk = NETLINK_CB(skb).sk; | |
734 | ||
5fd96123 PM |
735 | if (NETLINK_CB(skb).flags & NETLINK_SKB_TX) { |
736 | netlink_set_status(hdr, NL_MMAP_STATUS_UNUSED); | |
737 | ring = &nlk_sk(sk)->tx_ring; | |
738 | } else { | |
739 | if (!(NETLINK_CB(skb).flags & NETLINK_SKB_DELIVERED)) { | |
740 | hdr->nm_len = 0; | |
741 | netlink_set_status(hdr, NL_MMAP_STATUS_VALID); | |
742 | } | |
743 | ring = &nlk_sk(sk)->rx_ring; | |
9652e931 | 744 | } |
9652e931 PM |
745 | |
746 | WARN_ON(atomic_read(&ring->pending) == 0); | |
747 | atomic_dec(&ring->pending); | |
748 | sock_put(sk); | |
749 | ||
5e71d9d7 | 750 | skb->head = NULL; |
9652e931 PM |
751 | } |
752 | #endif | |
c05cdb1b PNA |
753 | if (is_vmalloc_addr(skb->head)) { |
754 | vfree(skb->head); | |
755 | skb->head = NULL; | |
756 | } | |
9652e931 PM |
757 | if (skb->sk != NULL) |
758 | sock_rfree(skb); | |
cf0a018a PM |
759 | } |
760 | ||
761 | static void netlink_skb_set_owner_r(struct sk_buff *skb, struct sock *sk) | |
762 | { | |
763 | WARN_ON(skb->sk != NULL); | |
764 | skb->sk = sk; | |
765 | skb->destructor = netlink_skb_destructor; | |
766 | atomic_add(skb->truesize, &sk->sk_rmem_alloc); | |
767 | sk_mem_charge(sk, skb->truesize); | |
768 | } | |
769 | ||
1da177e4 LT |
770 | static void netlink_sock_destruct(struct sock *sk) |
771 | { | |
3f660d66 HX |
772 | struct netlink_sock *nlk = nlk_sk(sk); |
773 | ||
3f660d66 HX |
774 | if (nlk->cb) { |
775 | if (nlk->cb->done) | |
776 | nlk->cb->done(nlk->cb); | |
6dc878a8 G |
777 | |
778 | module_put(nlk->cb->module); | |
3f660d66 HX |
779 | netlink_destroy_callback(nlk->cb); |
780 | } | |
781 | ||
1da177e4 | 782 | skb_queue_purge(&sk->sk_receive_queue); |
ccdfcc39 PM |
783 | #ifdef CONFIG_NETLINK_MMAP |
784 | if (1) { | |
785 | struct nl_mmap_req req; | |
786 | ||
787 | memset(&req, 0, sizeof(req)); | |
788 | if (nlk->rx_ring.pg_vec) | |
789 | netlink_set_ring(sk, &req, true, false); | |
790 | memset(&req, 0, sizeof(req)); | |
791 | if (nlk->tx_ring.pg_vec) | |
792 | netlink_set_ring(sk, &req, true, true); | |
793 | } | |
794 | #endif /* CONFIG_NETLINK_MMAP */ | |
1da177e4 LT |
795 | |
796 | if (!sock_flag(sk, SOCK_DEAD)) { | |
6ac552fd | 797 | printk(KERN_ERR "Freeing alive netlink socket %p\n", sk); |
1da177e4 LT |
798 | return; |
799 | } | |
547b792c IJ |
800 | |
801 | WARN_ON(atomic_read(&sk->sk_rmem_alloc)); | |
802 | WARN_ON(atomic_read(&sk->sk_wmem_alloc)); | |
803 | WARN_ON(nlk_sk(sk)->groups); | |
1da177e4 LT |
804 | } |
805 | ||
6ac552fd PM |
806 | /* This lock without WQ_FLAG_EXCLUSIVE is good on UP and it is _very_ bad on |
807 | * SMP. Look, when several writers sleep and reader wakes them up, all but one | |
1da177e4 LT |
808 | * immediately hit write lock and grab all the cpus. Exclusive sleep solves |
809 | * this, _but_ remember, it adds useless work on UP machines. | |
810 | */ | |
811 | ||
d136f1bd | 812 | void netlink_table_grab(void) |
9a429c49 | 813 | __acquires(nl_table_lock) |
1da177e4 | 814 | { |
d136f1bd JB |
815 | might_sleep(); |
816 | ||
6abd219c | 817 | write_lock_irq(&nl_table_lock); |
1da177e4 LT |
818 | |
819 | if (atomic_read(&nl_table_users)) { | |
820 | DECLARE_WAITQUEUE(wait, current); | |
821 | ||
822 | add_wait_queue_exclusive(&nl_table_wait, &wait); | |
6ac552fd | 823 | for (;;) { |
1da177e4 LT |
824 | set_current_state(TASK_UNINTERRUPTIBLE); |
825 | if (atomic_read(&nl_table_users) == 0) | |
826 | break; | |
6abd219c | 827 | write_unlock_irq(&nl_table_lock); |
1da177e4 | 828 | schedule(); |
6abd219c | 829 | write_lock_irq(&nl_table_lock); |
1da177e4 LT |
830 | } |
831 | ||
832 | __set_current_state(TASK_RUNNING); | |
833 | remove_wait_queue(&nl_table_wait, &wait); | |
834 | } | |
835 | } | |
836 | ||
d136f1bd | 837 | void netlink_table_ungrab(void) |
9a429c49 | 838 | __releases(nl_table_lock) |
1da177e4 | 839 | { |
6abd219c | 840 | write_unlock_irq(&nl_table_lock); |
1da177e4 LT |
841 | wake_up(&nl_table_wait); |
842 | } | |
843 | ||
6ac552fd | 844 | static inline void |
1da177e4 LT |
845 | netlink_lock_table(void) |
846 | { | |
847 | /* read_lock() synchronizes us to netlink_table_grab */ | |
848 | ||
849 | read_lock(&nl_table_lock); | |
850 | atomic_inc(&nl_table_users); | |
851 | read_unlock(&nl_table_lock); | |
852 | } | |
853 | ||
6ac552fd | 854 | static inline void |
1da177e4 LT |
855 | netlink_unlock_table(void) |
856 | { | |
857 | if (atomic_dec_and_test(&nl_table_users)) | |
858 | wake_up(&nl_table_wait); | |
859 | } | |
860 | ||
15e47304 | 861 | static struct sock *netlink_lookup(struct net *net, int protocol, u32 portid) |
1da177e4 | 862 | { |
15e47304 | 863 | struct nl_portid_hash *hash = &nl_table[protocol].hash; |
1da177e4 LT |
864 | struct hlist_head *head; |
865 | struct sock *sk; | |
1da177e4 LT |
866 | |
867 | read_lock(&nl_table_lock); | |
15e47304 | 868 | head = nl_portid_hashfn(hash, portid); |
b67bfe0d | 869 | sk_for_each(sk, head) { |
15e47304 | 870 | if (net_eq(sock_net(sk), net) && (nlk_sk(sk)->portid == portid)) { |
1da177e4 LT |
871 | sock_hold(sk); |
872 | goto found; | |
873 | } | |
874 | } | |
875 | sk = NULL; | |
876 | found: | |
877 | read_unlock(&nl_table_lock); | |
878 | return sk; | |
879 | } | |
880 | ||
15e47304 | 881 | static struct hlist_head *nl_portid_hash_zalloc(size_t size) |
1da177e4 LT |
882 | { |
883 | if (size <= PAGE_SIZE) | |
ea72912c | 884 | return kzalloc(size, GFP_ATOMIC); |
1da177e4 LT |
885 | else |
886 | return (struct hlist_head *) | |
ea72912c ED |
887 | __get_free_pages(GFP_ATOMIC | __GFP_ZERO, |
888 | get_order(size)); | |
1da177e4 LT |
889 | } |
890 | ||
15e47304 | 891 | static void nl_portid_hash_free(struct hlist_head *table, size_t size) |
1da177e4 LT |
892 | { |
893 | if (size <= PAGE_SIZE) | |
894 | kfree(table); | |
895 | else | |
896 | free_pages((unsigned long)table, get_order(size)); | |
897 | } | |
898 | ||
15e47304 | 899 | static int nl_portid_hash_rehash(struct nl_portid_hash *hash, int grow) |
1da177e4 LT |
900 | { |
901 | unsigned int omask, mask, shift; | |
902 | size_t osize, size; | |
903 | struct hlist_head *otable, *table; | |
904 | int i; | |
905 | ||
906 | omask = mask = hash->mask; | |
907 | osize = size = (mask + 1) * sizeof(*table); | |
908 | shift = hash->shift; | |
909 | ||
910 | if (grow) { | |
911 | if (++shift > hash->max_shift) | |
912 | return 0; | |
913 | mask = mask * 2 + 1; | |
914 | size *= 2; | |
915 | } | |
916 | ||
15e47304 | 917 | table = nl_portid_hash_zalloc(size); |
1da177e4 LT |
918 | if (!table) |
919 | return 0; | |
920 | ||
1da177e4 LT |
921 | otable = hash->table; |
922 | hash->table = table; | |
923 | hash->mask = mask; | |
924 | hash->shift = shift; | |
925 | get_random_bytes(&hash->rnd, sizeof(hash->rnd)); | |
926 | ||
927 | for (i = 0; i <= omask; i++) { | |
928 | struct sock *sk; | |
b67bfe0d | 929 | struct hlist_node *tmp; |
1da177e4 | 930 | |
b67bfe0d | 931 | sk_for_each_safe(sk, tmp, &otable[i]) |
15e47304 | 932 | __sk_add_node(sk, nl_portid_hashfn(hash, nlk_sk(sk)->portid)); |
1da177e4 LT |
933 | } |
934 | ||
15e47304 | 935 | nl_portid_hash_free(otable, osize); |
1da177e4 LT |
936 | hash->rehash_time = jiffies + 10 * 60 * HZ; |
937 | return 1; | |
938 | } | |
939 | ||
15e47304 | 940 | static inline int nl_portid_hash_dilute(struct nl_portid_hash *hash, int len) |
1da177e4 LT |
941 | { |
942 | int avg = hash->entries >> hash->shift; | |
943 | ||
15e47304 | 944 | if (unlikely(avg > 1) && nl_portid_hash_rehash(hash, 1)) |
1da177e4 LT |
945 | return 1; |
946 | ||
947 | if (unlikely(len > avg) && time_after(jiffies, hash->rehash_time)) { | |
15e47304 | 948 | nl_portid_hash_rehash(hash, 0); |
1da177e4 LT |
949 | return 1; |
950 | } | |
951 | ||
952 | return 0; | |
953 | } | |
954 | ||
90ddc4f0 | 955 | static const struct proto_ops netlink_ops; |
1da177e4 | 956 | |
4277a083 PM |
957 | static void |
958 | netlink_update_listeners(struct sock *sk) | |
959 | { | |
960 | struct netlink_table *tbl = &nl_table[sk->sk_protocol]; | |
4277a083 PM |
961 | unsigned long mask; |
962 | unsigned int i; | |
6d772ac5 ED |
963 | struct listeners *listeners; |
964 | ||
965 | listeners = nl_deref_protected(tbl->listeners); | |
966 | if (!listeners) | |
967 | return; | |
4277a083 | 968 | |
b4ff4f04 | 969 | for (i = 0; i < NLGRPLONGS(tbl->groups); i++) { |
4277a083 | 970 | mask = 0; |
b67bfe0d | 971 | sk_for_each_bound(sk, &tbl->mc_list) { |
b4ff4f04 JB |
972 | if (i < NLGRPLONGS(nlk_sk(sk)->ngroups)) |
973 | mask |= nlk_sk(sk)->groups[i]; | |
974 | } | |
6d772ac5 | 975 | listeners->masks[i] = mask; |
4277a083 PM |
976 | } |
977 | /* this function is only called with the netlink table "grabbed", which | |
978 | * makes sure updates are visible before bind or setsockopt return. */ | |
979 | } | |
980 | ||
15e47304 | 981 | static int netlink_insert(struct sock *sk, struct net *net, u32 portid) |
1da177e4 | 982 | { |
15e47304 | 983 | struct nl_portid_hash *hash = &nl_table[sk->sk_protocol].hash; |
1da177e4 LT |
984 | struct hlist_head *head; |
985 | int err = -EADDRINUSE; | |
986 | struct sock *osk; | |
1da177e4 LT |
987 | int len; |
988 | ||
989 | netlink_table_grab(); | |
15e47304 | 990 | head = nl_portid_hashfn(hash, portid); |
1da177e4 | 991 | len = 0; |
b67bfe0d | 992 | sk_for_each(osk, head) { |
15e47304 | 993 | if (net_eq(sock_net(osk), net) && (nlk_sk(osk)->portid == portid)) |
1da177e4 LT |
994 | break; |
995 | len++; | |
996 | } | |
b67bfe0d | 997 | if (osk) |
1da177e4 LT |
998 | goto err; |
999 | ||
1000 | err = -EBUSY; | |
15e47304 | 1001 | if (nlk_sk(sk)->portid) |
1da177e4 LT |
1002 | goto err; |
1003 | ||
1004 | err = -ENOMEM; | |
1005 | if (BITS_PER_LONG > 32 && unlikely(hash->entries >= UINT_MAX)) | |
1006 | goto err; | |
1007 | ||
15e47304 EB |
1008 | if (len && nl_portid_hash_dilute(hash, len)) |
1009 | head = nl_portid_hashfn(hash, portid); | |
1da177e4 | 1010 | hash->entries++; |
15e47304 | 1011 | nlk_sk(sk)->portid = portid; |
1da177e4 LT |
1012 | sk_add_node(sk, head); |
1013 | err = 0; | |
1014 | ||
1015 | err: | |
1016 | netlink_table_ungrab(); | |
1017 | return err; | |
1018 | } | |
1019 | ||
1020 | static void netlink_remove(struct sock *sk) | |
1021 | { | |
1022 | netlink_table_grab(); | |
d470e3b4 DM |
1023 | if (sk_del_node_init(sk)) |
1024 | nl_table[sk->sk_protocol].hash.entries--; | |
f7fa9b10 | 1025 | if (nlk_sk(sk)->subscriptions) |
1da177e4 LT |
1026 | __sk_del_bind_node(sk); |
1027 | netlink_table_ungrab(); | |
1028 | } | |
1029 | ||
1030 | static struct proto netlink_proto = { | |
1031 | .name = "NETLINK", | |
1032 | .owner = THIS_MODULE, | |
1033 | .obj_size = sizeof(struct netlink_sock), | |
1034 | }; | |
1035 | ||
1b8d7ae4 EB |
1036 | static int __netlink_create(struct net *net, struct socket *sock, |
1037 | struct mutex *cb_mutex, int protocol) | |
1da177e4 LT |
1038 | { |
1039 | struct sock *sk; | |
1040 | struct netlink_sock *nlk; | |
ab33a171 PM |
1041 | |
1042 | sock->ops = &netlink_ops; | |
1043 | ||
6257ff21 | 1044 | sk = sk_alloc(net, PF_NETLINK, GFP_KERNEL, &netlink_proto); |
ab33a171 PM |
1045 | if (!sk) |
1046 | return -ENOMEM; | |
1047 | ||
1048 | sock_init_data(sock, sk); | |
1049 | ||
1050 | nlk = nlk_sk(sk); | |
658cb354 | 1051 | if (cb_mutex) { |
ffa4d721 | 1052 | nlk->cb_mutex = cb_mutex; |
658cb354 | 1053 | } else { |
ffa4d721 PM |
1054 | nlk->cb_mutex = &nlk->cb_def_mutex; |
1055 | mutex_init(nlk->cb_mutex); | |
1056 | } | |
ab33a171 | 1057 | init_waitqueue_head(&nlk->wait); |
ccdfcc39 PM |
1058 | #ifdef CONFIG_NETLINK_MMAP |
1059 | mutex_init(&nlk->pg_vec_lock); | |
1060 | #endif | |
ab33a171 PM |
1061 | |
1062 | sk->sk_destruct = netlink_sock_destruct; | |
1063 | sk->sk_protocol = protocol; | |
1064 | return 0; | |
1065 | } | |
1066 | ||
3f378b68 EP |
1067 | static int netlink_create(struct net *net, struct socket *sock, int protocol, |
1068 | int kern) | |
ab33a171 PM |
1069 | { |
1070 | struct module *module = NULL; | |
af65bdfc | 1071 | struct mutex *cb_mutex; |
f7fa9b10 | 1072 | struct netlink_sock *nlk; |
03292745 | 1073 | void (*bind)(int group); |
ab33a171 | 1074 | int err = 0; |
1da177e4 LT |
1075 | |
1076 | sock->state = SS_UNCONNECTED; | |
1077 | ||
1078 | if (sock->type != SOCK_RAW && sock->type != SOCK_DGRAM) | |
1079 | return -ESOCKTNOSUPPORT; | |
1080 | ||
6ac552fd | 1081 | if (protocol < 0 || protocol >= MAX_LINKS) |
1da177e4 LT |
1082 | return -EPROTONOSUPPORT; |
1083 | ||
77247bbb | 1084 | netlink_lock_table(); |
95a5afca | 1085 | #ifdef CONFIG_MODULES |
ab33a171 | 1086 | if (!nl_table[protocol].registered) { |
77247bbb | 1087 | netlink_unlock_table(); |
4fdb3bb7 | 1088 | request_module("net-pf-%d-proto-%d", PF_NETLINK, protocol); |
77247bbb | 1089 | netlink_lock_table(); |
4fdb3bb7 | 1090 | } |
ab33a171 PM |
1091 | #endif |
1092 | if (nl_table[protocol].registered && | |
1093 | try_module_get(nl_table[protocol].module)) | |
1094 | module = nl_table[protocol].module; | |
974c37e9 AD |
1095 | else |
1096 | err = -EPROTONOSUPPORT; | |
af65bdfc | 1097 | cb_mutex = nl_table[protocol].cb_mutex; |
03292745 | 1098 | bind = nl_table[protocol].bind; |
77247bbb | 1099 | netlink_unlock_table(); |
4fdb3bb7 | 1100 | |
974c37e9 AD |
1101 | if (err < 0) |
1102 | goto out; | |
1103 | ||
6ac552fd PM |
1104 | err = __netlink_create(net, sock, cb_mutex, protocol); |
1105 | if (err < 0) | |
f7fa9b10 PM |
1106 | goto out_module; |
1107 | ||
6f756a8c | 1108 | local_bh_disable(); |
c1fd3b94 | 1109 | sock_prot_inuse_add(net, &netlink_proto, 1); |
6f756a8c DM |
1110 | local_bh_enable(); |
1111 | ||
f7fa9b10 | 1112 | nlk = nlk_sk(sock->sk); |
f7fa9b10 | 1113 | nlk->module = module; |
03292745 | 1114 | nlk->netlink_bind = bind; |
ab33a171 PM |
1115 | out: |
1116 | return err; | |
1da177e4 | 1117 | |
ab33a171 PM |
1118 | out_module: |
1119 | module_put(module); | |
1120 | goto out; | |
1da177e4 LT |
1121 | } |
1122 | ||
1123 | static int netlink_release(struct socket *sock) | |
1124 | { | |
1125 | struct sock *sk = sock->sk; | |
1126 | struct netlink_sock *nlk; | |
1127 | ||
1128 | if (!sk) | |
1129 | return 0; | |
1130 | ||
1131 | netlink_remove(sk); | |
ac57b3a9 | 1132 | sock_orphan(sk); |
1da177e4 LT |
1133 | nlk = nlk_sk(sk); |
1134 | ||
3f660d66 HX |
1135 | /* |
1136 | * OK. Socket is unlinked, any packets that arrive now | |
1137 | * will be purged. | |
1138 | */ | |
1da177e4 | 1139 | |
1da177e4 LT |
1140 | sock->sk = NULL; |
1141 | wake_up_interruptible_all(&nlk->wait); | |
1142 | ||
1143 | skb_queue_purge(&sk->sk_write_queue); | |
1144 | ||
15e47304 | 1145 | if (nlk->portid) { |
1da177e4 | 1146 | struct netlink_notify n = { |
3b1e0a65 | 1147 | .net = sock_net(sk), |
1da177e4 | 1148 | .protocol = sk->sk_protocol, |
15e47304 | 1149 | .portid = nlk->portid, |
1da177e4 | 1150 | }; |
e041c683 AS |
1151 | atomic_notifier_call_chain(&netlink_chain, |
1152 | NETLINK_URELEASE, &n); | |
746fac4d | 1153 | } |
4fdb3bb7 | 1154 | |
5e7c001c | 1155 | module_put(nlk->module); |
4fdb3bb7 | 1156 | |
4277a083 | 1157 | netlink_table_grab(); |
aed81560 | 1158 | if (netlink_is_kernel(sk)) { |
869e58f8 DL |
1159 | BUG_ON(nl_table[sk->sk_protocol].registered == 0); |
1160 | if (--nl_table[sk->sk_protocol].registered == 0) { | |
6d772ac5 ED |
1161 | struct listeners *old; |
1162 | ||
1163 | old = nl_deref_protected(nl_table[sk->sk_protocol].listeners); | |
1164 | RCU_INIT_POINTER(nl_table[sk->sk_protocol].listeners, NULL); | |
1165 | kfree_rcu(old, rcu); | |
869e58f8 | 1166 | nl_table[sk->sk_protocol].module = NULL; |
9785e10a PNA |
1167 | nl_table[sk->sk_protocol].bind = NULL; |
1168 | nl_table[sk->sk_protocol].flags = 0; | |
869e58f8 DL |
1169 | nl_table[sk->sk_protocol].registered = 0; |
1170 | } | |
658cb354 | 1171 | } else if (nlk->subscriptions) { |
4277a083 | 1172 | netlink_update_listeners(sk); |
658cb354 | 1173 | } |
4277a083 | 1174 | netlink_table_ungrab(); |
77247bbb | 1175 | |
f7fa9b10 PM |
1176 | kfree(nlk->groups); |
1177 | nlk->groups = NULL; | |
1178 | ||
3755810c | 1179 | local_bh_disable(); |
c1fd3b94 | 1180 | sock_prot_inuse_add(sock_net(sk), &netlink_proto, -1); |
3755810c | 1181 | local_bh_enable(); |
1da177e4 LT |
1182 | sock_put(sk); |
1183 | return 0; | |
1184 | } | |
1185 | ||
1186 | static int netlink_autobind(struct socket *sock) | |
1187 | { | |
1188 | struct sock *sk = sock->sk; | |
3b1e0a65 | 1189 | struct net *net = sock_net(sk); |
15e47304 | 1190 | struct nl_portid_hash *hash = &nl_table[sk->sk_protocol].hash; |
1da177e4 LT |
1191 | struct hlist_head *head; |
1192 | struct sock *osk; | |
15e47304 | 1193 | s32 portid = task_tgid_vnr(current); |
1da177e4 LT |
1194 | int err; |
1195 | static s32 rover = -4097; | |
1196 | ||
1197 | retry: | |
1198 | cond_resched(); | |
1199 | netlink_table_grab(); | |
15e47304 | 1200 | head = nl_portid_hashfn(hash, portid); |
b67bfe0d | 1201 | sk_for_each(osk, head) { |
878628fb | 1202 | if (!net_eq(sock_net(osk), net)) |
b4b51029 | 1203 | continue; |
15e47304 EB |
1204 | if (nlk_sk(osk)->portid == portid) { |
1205 | /* Bind collision, search negative portid values. */ | |
1206 | portid = rover--; | |
1da177e4 LT |
1207 | if (rover > -4097) |
1208 | rover = -4097; | |
1209 | netlink_table_ungrab(); | |
1210 | goto retry; | |
1211 | } | |
1212 | } | |
1213 | netlink_table_ungrab(); | |
1214 | ||
15e47304 | 1215 | err = netlink_insert(sk, net, portid); |
1da177e4 LT |
1216 | if (err == -EADDRINUSE) |
1217 | goto retry; | |
d470e3b4 DM |
1218 | |
1219 | /* If 2 threads race to autobind, that is fine. */ | |
1220 | if (err == -EBUSY) | |
1221 | err = 0; | |
1222 | ||
1223 | return err; | |
1da177e4 LT |
1224 | } |
1225 | ||
b57ef81f | 1226 | static inline int netlink_capable(const struct socket *sock, unsigned int flag) |
746fac4d | 1227 | { |
9785e10a | 1228 | return (nl_table[sock->sk->sk_protocol].flags & flag) || |
df008c91 | 1229 | ns_capable(sock_net(sock->sk)->user_ns, CAP_NET_ADMIN); |
746fac4d | 1230 | } |
1da177e4 | 1231 | |
f7fa9b10 PM |
1232 | static void |
1233 | netlink_update_subscriptions(struct sock *sk, unsigned int subscriptions) | |
1234 | { | |
1235 | struct netlink_sock *nlk = nlk_sk(sk); | |
1236 | ||
1237 | if (nlk->subscriptions && !subscriptions) | |
1238 | __sk_del_bind_node(sk); | |
1239 | else if (!nlk->subscriptions && subscriptions) | |
1240 | sk_add_bind_node(sk, &nl_table[sk->sk_protocol].mc_list); | |
1241 | nlk->subscriptions = subscriptions; | |
1242 | } | |
1243 | ||
b4ff4f04 | 1244 | static int netlink_realloc_groups(struct sock *sk) |
513c2500 PM |
1245 | { |
1246 | struct netlink_sock *nlk = nlk_sk(sk); | |
1247 | unsigned int groups; | |
b4ff4f04 | 1248 | unsigned long *new_groups; |
513c2500 PM |
1249 | int err = 0; |
1250 | ||
b4ff4f04 JB |
1251 | netlink_table_grab(); |
1252 | ||
513c2500 | 1253 | groups = nl_table[sk->sk_protocol].groups; |
b4ff4f04 | 1254 | if (!nl_table[sk->sk_protocol].registered) { |
513c2500 | 1255 | err = -ENOENT; |
b4ff4f04 JB |
1256 | goto out_unlock; |
1257 | } | |
513c2500 | 1258 | |
b4ff4f04 JB |
1259 | if (nlk->ngroups >= groups) |
1260 | goto out_unlock; | |
513c2500 | 1261 | |
b4ff4f04 JB |
1262 | new_groups = krealloc(nlk->groups, NLGRPSZ(groups), GFP_ATOMIC); |
1263 | if (new_groups == NULL) { | |
1264 | err = -ENOMEM; | |
1265 | goto out_unlock; | |
1266 | } | |
6ac552fd | 1267 | memset((char *)new_groups + NLGRPSZ(nlk->ngroups), 0, |
b4ff4f04 JB |
1268 | NLGRPSZ(groups) - NLGRPSZ(nlk->ngroups)); |
1269 | ||
1270 | nlk->groups = new_groups; | |
513c2500 | 1271 | nlk->ngroups = groups; |
b4ff4f04 JB |
1272 | out_unlock: |
1273 | netlink_table_ungrab(); | |
1274 | return err; | |
513c2500 PM |
1275 | } |
1276 | ||
6ac552fd PM |
1277 | static int netlink_bind(struct socket *sock, struct sockaddr *addr, |
1278 | int addr_len) | |
1da177e4 LT |
1279 | { |
1280 | struct sock *sk = sock->sk; | |
3b1e0a65 | 1281 | struct net *net = sock_net(sk); |
1da177e4 LT |
1282 | struct netlink_sock *nlk = nlk_sk(sk); |
1283 | struct sockaddr_nl *nladdr = (struct sockaddr_nl *)addr; | |
1284 | int err; | |
746fac4d | 1285 | |
4e4b5376 HFS |
1286 | if (addr_len < sizeof(struct sockaddr_nl)) |
1287 | return -EINVAL; | |
1288 | ||
1da177e4 LT |
1289 | if (nladdr->nl_family != AF_NETLINK) |
1290 | return -EINVAL; | |
1291 | ||
1292 | /* Only superuser is allowed to listen multicasts */ | |
513c2500 | 1293 | if (nladdr->nl_groups) { |
9785e10a | 1294 | if (!netlink_capable(sock, NL_CFG_F_NONROOT_RECV)) |
513c2500 | 1295 | return -EPERM; |
b4ff4f04 JB |
1296 | err = netlink_realloc_groups(sk); |
1297 | if (err) | |
1298 | return err; | |
513c2500 | 1299 | } |
1da177e4 | 1300 | |
15e47304 EB |
1301 | if (nlk->portid) { |
1302 | if (nladdr->nl_pid != nlk->portid) | |
1da177e4 LT |
1303 | return -EINVAL; |
1304 | } else { | |
1305 | err = nladdr->nl_pid ? | |
b4b51029 | 1306 | netlink_insert(sk, net, nladdr->nl_pid) : |
1da177e4 LT |
1307 | netlink_autobind(sock); |
1308 | if (err) | |
1309 | return err; | |
1310 | } | |
1311 | ||
513c2500 | 1312 | if (!nladdr->nl_groups && (nlk->groups == NULL || !(u32)nlk->groups[0])) |
1da177e4 LT |
1313 | return 0; |
1314 | ||
1315 | netlink_table_grab(); | |
f7fa9b10 | 1316 | netlink_update_subscriptions(sk, nlk->subscriptions + |
746fac4d YH |
1317 | hweight32(nladdr->nl_groups) - |
1318 | hweight32(nlk->groups[0])); | |
1319 | nlk->groups[0] = (nlk->groups[0] & ~0xffffffffUL) | nladdr->nl_groups; | |
4277a083 | 1320 | netlink_update_listeners(sk); |
1da177e4 LT |
1321 | netlink_table_ungrab(); |
1322 | ||
03292745 PNA |
1323 | if (nlk->netlink_bind && nlk->groups[0]) { |
1324 | int i; | |
1325 | ||
1326 | for (i=0; i<nlk->ngroups; i++) { | |
1327 | if (test_bit(i, nlk->groups)) | |
1328 | nlk->netlink_bind(i); | |
1329 | } | |
1330 | } | |
1331 | ||
1da177e4 LT |
1332 | return 0; |
1333 | } | |
1334 | ||
1335 | static int netlink_connect(struct socket *sock, struct sockaddr *addr, | |
1336 | int alen, int flags) | |
1337 | { | |
1338 | int err = 0; | |
1339 | struct sock *sk = sock->sk; | |
1340 | struct netlink_sock *nlk = nlk_sk(sk); | |
6ac552fd | 1341 | struct sockaddr_nl *nladdr = (struct sockaddr_nl *)addr; |
1da177e4 | 1342 | |
6503d961 CG |
1343 | if (alen < sizeof(addr->sa_family)) |
1344 | return -EINVAL; | |
1345 | ||
1da177e4 LT |
1346 | if (addr->sa_family == AF_UNSPEC) { |
1347 | sk->sk_state = NETLINK_UNCONNECTED; | |
15e47304 | 1348 | nlk->dst_portid = 0; |
d629b836 | 1349 | nlk->dst_group = 0; |
1da177e4 LT |
1350 | return 0; |
1351 | } | |
1352 | if (addr->sa_family != AF_NETLINK) | |
1353 | return -EINVAL; | |
1354 | ||
1355 | /* Only superuser is allowed to send multicasts */ | |
9785e10a | 1356 | if (nladdr->nl_groups && !netlink_capable(sock, NL_CFG_F_NONROOT_SEND)) |
1da177e4 LT |
1357 | return -EPERM; |
1358 | ||
15e47304 | 1359 | if (!nlk->portid) |
1da177e4 LT |
1360 | err = netlink_autobind(sock); |
1361 | ||
1362 | if (err == 0) { | |
1363 | sk->sk_state = NETLINK_CONNECTED; | |
15e47304 | 1364 | nlk->dst_portid = nladdr->nl_pid; |
d629b836 | 1365 | nlk->dst_group = ffs(nladdr->nl_groups); |
1da177e4 LT |
1366 | } |
1367 | ||
1368 | return err; | |
1369 | } | |
1370 | ||
6ac552fd PM |
1371 | static int netlink_getname(struct socket *sock, struct sockaddr *addr, |
1372 | int *addr_len, int peer) | |
1da177e4 LT |
1373 | { |
1374 | struct sock *sk = sock->sk; | |
1375 | struct netlink_sock *nlk = nlk_sk(sk); | |
13cfa97b | 1376 | DECLARE_SOCKADDR(struct sockaddr_nl *, nladdr, addr); |
746fac4d | 1377 | |
1da177e4 LT |
1378 | nladdr->nl_family = AF_NETLINK; |
1379 | nladdr->nl_pad = 0; | |
1380 | *addr_len = sizeof(*nladdr); | |
1381 | ||
1382 | if (peer) { | |
15e47304 | 1383 | nladdr->nl_pid = nlk->dst_portid; |
d629b836 | 1384 | nladdr->nl_groups = netlink_group_mask(nlk->dst_group); |
1da177e4 | 1385 | } else { |
15e47304 | 1386 | nladdr->nl_pid = nlk->portid; |
513c2500 | 1387 | nladdr->nl_groups = nlk->groups ? nlk->groups[0] : 0; |
1da177e4 LT |
1388 | } |
1389 | return 0; | |
1390 | } | |
1391 | ||
15e47304 | 1392 | static struct sock *netlink_getsockbyportid(struct sock *ssk, u32 portid) |
1da177e4 | 1393 | { |
1da177e4 LT |
1394 | struct sock *sock; |
1395 | struct netlink_sock *nlk; | |
1396 | ||
15e47304 | 1397 | sock = netlink_lookup(sock_net(ssk), ssk->sk_protocol, portid); |
1da177e4 LT |
1398 | if (!sock) |
1399 | return ERR_PTR(-ECONNREFUSED); | |
1400 | ||
1401 | /* Don't bother queuing skb if kernel socket has no input function */ | |
1402 | nlk = nlk_sk(sock); | |
cd40b7d3 | 1403 | if (sock->sk_state == NETLINK_CONNECTED && |
15e47304 | 1404 | nlk->dst_portid != nlk_sk(ssk)->portid) { |
1da177e4 LT |
1405 | sock_put(sock); |
1406 | return ERR_PTR(-ECONNREFUSED); | |
1407 | } | |
1408 | return sock; | |
1409 | } | |
1410 | ||
1411 | struct sock *netlink_getsockbyfilp(struct file *filp) | |
1412 | { | |
496ad9aa | 1413 | struct inode *inode = file_inode(filp); |
1da177e4 LT |
1414 | struct sock *sock; |
1415 | ||
1416 | if (!S_ISSOCK(inode->i_mode)) | |
1417 | return ERR_PTR(-ENOTSOCK); | |
1418 | ||
1419 | sock = SOCKET_I(inode)->sk; | |
1420 | if (sock->sk_family != AF_NETLINK) | |
1421 | return ERR_PTR(-EINVAL); | |
1422 | ||
1423 | sock_hold(sock); | |
1424 | return sock; | |
1425 | } | |
1426 | ||
c05cdb1b PNA |
1427 | static struct sk_buff *netlink_alloc_large_skb(unsigned int size) |
1428 | { | |
1429 | struct sk_buff *skb; | |
1430 | void *data; | |
1431 | ||
1432 | if (size <= NLMSG_GOODSIZE) | |
1433 | return alloc_skb(size, GFP_KERNEL); | |
1434 | ||
1435 | skb = alloc_skb_head(GFP_KERNEL); | |
1436 | if (skb == NULL) | |
1437 | return NULL; | |
1438 | ||
1439 | data = vmalloc(size); | |
1440 | if (data == NULL) | |
1441 | goto err; | |
1442 | ||
1443 | skb->head = data; | |
1444 | skb->data = data; | |
1445 | skb_reset_tail_pointer(skb); | |
1446 | skb->end = skb->tail + size; | |
1447 | skb->len = 0; | |
1448 | skb->destructor = netlink_skb_destructor; | |
1449 | ||
1450 | return skb; | |
1451 | err: | |
1452 | kfree_skb(skb); | |
1453 | return NULL; | |
1454 | } | |
1455 | ||
1da177e4 LT |
1456 | /* |
1457 | * Attach a skb to a netlink socket. | |
1458 | * The caller must hold a reference to the destination socket. On error, the | |
1459 | * reference is dropped. The skb is not send to the destination, just all | |
1460 | * all error checks are performed and memory in the queue is reserved. | |
1461 | * Return values: | |
1462 | * < 0: error. skb freed, reference to sock dropped. | |
1463 | * 0: continue | |
1464 | * 1: repeat lookup - reference dropped while waiting for socket memory. | |
1465 | */ | |
9457afee | 1466 | int netlink_attachskb(struct sock *sk, struct sk_buff *skb, |
c3d8d1e3 | 1467 | long *timeo, struct sock *ssk) |
1da177e4 LT |
1468 | { |
1469 | struct netlink_sock *nlk; | |
1470 | ||
1471 | nlk = nlk_sk(sk); | |
1472 | ||
5fd96123 PM |
1473 | if ((atomic_read(&sk->sk_rmem_alloc) > sk->sk_rcvbuf || |
1474 | test_bit(NETLINK_CONGESTED, &nlk->state)) && | |
1475 | !netlink_skb_is_mmaped(skb)) { | |
1da177e4 | 1476 | DECLARE_WAITQUEUE(wait, current); |
c3d8d1e3 | 1477 | if (!*timeo) { |
aed81560 | 1478 | if (!ssk || netlink_is_kernel(ssk)) |
1da177e4 LT |
1479 | netlink_overrun(sk); |
1480 | sock_put(sk); | |
1481 | kfree_skb(skb); | |
1482 | return -EAGAIN; | |
1483 | } | |
1484 | ||
1485 | __set_current_state(TASK_INTERRUPTIBLE); | |
1486 | add_wait_queue(&nlk->wait, &wait); | |
1487 | ||
1488 | if ((atomic_read(&sk->sk_rmem_alloc) > sk->sk_rcvbuf || | |
cd967e05 | 1489 | test_bit(NETLINK_CONGESTED, &nlk->state)) && |
1da177e4 | 1490 | !sock_flag(sk, SOCK_DEAD)) |
c3d8d1e3 | 1491 | *timeo = schedule_timeout(*timeo); |
1da177e4 LT |
1492 | |
1493 | __set_current_state(TASK_RUNNING); | |
1494 | remove_wait_queue(&nlk->wait, &wait); | |
1495 | sock_put(sk); | |
1496 | ||
1497 | if (signal_pending(current)) { | |
1498 | kfree_skb(skb); | |
c3d8d1e3 | 1499 | return sock_intr_errno(*timeo); |
1da177e4 LT |
1500 | } |
1501 | return 1; | |
1502 | } | |
cf0a018a | 1503 | netlink_skb_set_owner_r(skb, sk); |
1da177e4 LT |
1504 | return 0; |
1505 | } | |
1506 | ||
4a7e7c2a | 1507 | static int __netlink_sendskb(struct sock *sk, struct sk_buff *skb) |
1da177e4 | 1508 | { |
1da177e4 LT |
1509 | int len = skb->len; |
1510 | ||
f9c22888 PM |
1511 | #ifdef CONFIG_NETLINK_MMAP |
1512 | if (netlink_skb_is_mmaped(skb)) | |
1513 | netlink_queue_mmaped_skb(sk, skb); | |
1514 | else if (netlink_rx_is_mmaped(sk)) | |
1515 | netlink_ring_set_copied(sk, skb); | |
1516 | else | |
1517 | #endif /* CONFIG_NETLINK_MMAP */ | |
1518 | skb_queue_tail(&sk->sk_receive_queue, skb); | |
1da177e4 | 1519 | sk->sk_data_ready(sk, len); |
4a7e7c2a ED |
1520 | return len; |
1521 | } | |
1522 | ||
1523 | int netlink_sendskb(struct sock *sk, struct sk_buff *skb) | |
1524 | { | |
1525 | int len = __netlink_sendskb(sk, skb); | |
1526 | ||
1da177e4 LT |
1527 | sock_put(sk); |
1528 | return len; | |
1529 | } | |
1530 | ||
1531 | void netlink_detachskb(struct sock *sk, struct sk_buff *skb) | |
1532 | { | |
1533 | kfree_skb(skb); | |
1534 | sock_put(sk); | |
1535 | } | |
1536 | ||
b57ef81f | 1537 | static struct sk_buff *netlink_trim(struct sk_buff *skb, gfp_t allocation) |
1da177e4 LT |
1538 | { |
1539 | int delta; | |
1540 | ||
1298ca46 | 1541 | WARN_ON(skb->sk != NULL); |
5fd96123 PM |
1542 | if (netlink_skb_is_mmaped(skb)) |
1543 | return skb; | |
1da177e4 | 1544 | |
4305b541 | 1545 | delta = skb->end - skb->tail; |
c05cdb1b | 1546 | if (is_vmalloc_addr(skb->head) || delta * 2 < skb->truesize) |
1da177e4 LT |
1547 | return skb; |
1548 | ||
1549 | if (skb_shared(skb)) { | |
1550 | struct sk_buff *nskb = skb_clone(skb, allocation); | |
1551 | if (!nskb) | |
1552 | return skb; | |
8460c00f | 1553 | consume_skb(skb); |
1da177e4 LT |
1554 | skb = nskb; |
1555 | } | |
1556 | ||
1557 | if (!pskb_expand_head(skb, 0, -delta, allocation)) | |
1558 | skb->truesize -= delta; | |
1559 | ||
1560 | return skb; | |
1561 | } | |
1562 | ||
3fbc2905 EB |
1563 | static int netlink_unicast_kernel(struct sock *sk, struct sk_buff *skb, |
1564 | struct sock *ssk) | |
cd40b7d3 DL |
1565 | { |
1566 | int ret; | |
1567 | struct netlink_sock *nlk = nlk_sk(sk); | |
1568 | ||
1569 | ret = -ECONNREFUSED; | |
1570 | if (nlk->netlink_rcv != NULL) { | |
1571 | ret = skb->len; | |
cf0a018a | 1572 | netlink_skb_set_owner_r(skb, sk); |
e32123e5 | 1573 | NETLINK_CB(skb).sk = ssk; |
cd40b7d3 | 1574 | nlk->netlink_rcv(skb); |
bfb253c9 ED |
1575 | consume_skb(skb); |
1576 | } else { | |
1577 | kfree_skb(skb); | |
cd40b7d3 | 1578 | } |
cd40b7d3 DL |
1579 | sock_put(sk); |
1580 | return ret; | |
1581 | } | |
1582 | ||
1583 | int netlink_unicast(struct sock *ssk, struct sk_buff *skb, | |
15e47304 | 1584 | u32 portid, int nonblock) |
1da177e4 LT |
1585 | { |
1586 | struct sock *sk; | |
1587 | int err; | |
1588 | long timeo; | |
1589 | ||
1590 | skb = netlink_trim(skb, gfp_any()); | |
1591 | ||
1592 | timeo = sock_sndtimeo(ssk, nonblock); | |
1593 | retry: | |
15e47304 | 1594 | sk = netlink_getsockbyportid(ssk, portid); |
1da177e4 LT |
1595 | if (IS_ERR(sk)) { |
1596 | kfree_skb(skb); | |
1597 | return PTR_ERR(sk); | |
1598 | } | |
cd40b7d3 | 1599 | if (netlink_is_kernel(sk)) |
3fbc2905 | 1600 | return netlink_unicast_kernel(sk, skb, ssk); |
cd40b7d3 | 1601 | |
b1153f29 | 1602 | if (sk_filter(sk, skb)) { |
84874607 | 1603 | err = skb->len; |
b1153f29 SH |
1604 | kfree_skb(skb); |
1605 | sock_put(sk); | |
1606 | return err; | |
1607 | } | |
1608 | ||
9457afee | 1609 | err = netlink_attachskb(sk, skb, &timeo, ssk); |
1da177e4 LT |
1610 | if (err == 1) |
1611 | goto retry; | |
1612 | if (err) | |
1613 | return err; | |
1614 | ||
7ee015e0 | 1615 | return netlink_sendskb(sk, skb); |
1da177e4 | 1616 | } |
6ac552fd | 1617 | EXPORT_SYMBOL(netlink_unicast); |
1da177e4 | 1618 | |
f9c22888 PM |
1619 | struct sk_buff *netlink_alloc_skb(struct sock *ssk, unsigned int size, |
1620 | u32 dst_portid, gfp_t gfp_mask) | |
1621 | { | |
1622 | #ifdef CONFIG_NETLINK_MMAP | |
1623 | struct sock *sk = NULL; | |
1624 | struct sk_buff *skb; | |
1625 | struct netlink_ring *ring; | |
1626 | struct nl_mmap_hdr *hdr; | |
1627 | unsigned int maxlen; | |
1628 | ||
1629 | sk = netlink_getsockbyportid(ssk, dst_portid); | |
1630 | if (IS_ERR(sk)) | |
1631 | goto out; | |
1632 | ||
1633 | ring = &nlk_sk(sk)->rx_ring; | |
1634 | /* fast-path without atomic ops for common case: non-mmaped receiver */ | |
1635 | if (ring->pg_vec == NULL) | |
1636 | goto out_put; | |
1637 | ||
1638 | skb = alloc_skb_head(gfp_mask); | |
1639 | if (skb == NULL) | |
1640 | goto err1; | |
1641 | ||
1642 | spin_lock_bh(&sk->sk_receive_queue.lock); | |
1643 | /* check again under lock */ | |
1644 | if (ring->pg_vec == NULL) | |
1645 | goto out_free; | |
1646 | ||
1647 | maxlen = ring->frame_size - NL_MMAP_HDRLEN; | |
1648 | if (maxlen < size) | |
1649 | goto out_free; | |
1650 | ||
1651 | netlink_forward_ring(ring); | |
1652 | hdr = netlink_current_frame(ring, NL_MMAP_STATUS_UNUSED); | |
1653 | if (hdr == NULL) | |
1654 | goto err2; | |
1655 | netlink_ring_setup_skb(skb, sk, ring, hdr); | |
1656 | netlink_set_status(hdr, NL_MMAP_STATUS_RESERVED); | |
1657 | atomic_inc(&ring->pending); | |
1658 | netlink_increment_head(ring); | |
1659 | ||
1660 | spin_unlock_bh(&sk->sk_receive_queue.lock); | |
1661 | return skb; | |
1662 | ||
1663 | err2: | |
1664 | kfree_skb(skb); | |
1665 | spin_unlock_bh(&sk->sk_receive_queue.lock); | |
cd1df525 | 1666 | netlink_overrun(sk); |
f9c22888 PM |
1667 | err1: |
1668 | sock_put(sk); | |
1669 | return NULL; | |
1670 | ||
1671 | out_free: | |
1672 | kfree_skb(skb); | |
1673 | spin_unlock_bh(&sk->sk_receive_queue.lock); | |
1674 | out_put: | |
1675 | sock_put(sk); | |
1676 | out: | |
1677 | #endif | |
1678 | return alloc_skb(size, gfp_mask); | |
1679 | } | |
1680 | EXPORT_SYMBOL_GPL(netlink_alloc_skb); | |
1681 | ||
4277a083 PM |
1682 | int netlink_has_listeners(struct sock *sk, unsigned int group) |
1683 | { | |
1684 | int res = 0; | |
5c398dc8 | 1685 | struct listeners *listeners; |
4277a083 | 1686 | |
aed81560 | 1687 | BUG_ON(!netlink_is_kernel(sk)); |
b4ff4f04 JB |
1688 | |
1689 | rcu_read_lock(); | |
1690 | listeners = rcu_dereference(nl_table[sk->sk_protocol].listeners); | |
1691 | ||
6d772ac5 | 1692 | if (listeners && group - 1 < nl_table[sk->sk_protocol].groups) |
5c398dc8 | 1693 | res = test_bit(group - 1, listeners->masks); |
b4ff4f04 JB |
1694 | |
1695 | rcu_read_unlock(); | |
1696 | ||
4277a083 PM |
1697 | return res; |
1698 | } | |
1699 | EXPORT_SYMBOL_GPL(netlink_has_listeners); | |
1700 | ||
b57ef81f | 1701 | static int netlink_broadcast_deliver(struct sock *sk, struct sk_buff *skb) |
1da177e4 LT |
1702 | { |
1703 | struct netlink_sock *nlk = nlk_sk(sk); | |
1704 | ||
1705 | if (atomic_read(&sk->sk_rmem_alloc) <= sk->sk_rcvbuf && | |
cd967e05 | 1706 | !test_bit(NETLINK_CONGESTED, &nlk->state)) { |
cf0a018a | 1707 | netlink_skb_set_owner_r(skb, sk); |
4a7e7c2a | 1708 | __netlink_sendskb(sk, skb); |
2c645800 | 1709 | return atomic_read(&sk->sk_rmem_alloc) > (sk->sk_rcvbuf >> 1); |
1da177e4 LT |
1710 | } |
1711 | return -1; | |
1712 | } | |
1713 | ||
1714 | struct netlink_broadcast_data { | |
1715 | struct sock *exclude_sk; | |
b4b51029 | 1716 | struct net *net; |
15e47304 | 1717 | u32 portid; |
1da177e4 LT |
1718 | u32 group; |
1719 | int failure; | |
ff491a73 | 1720 | int delivery_failure; |
1da177e4 LT |
1721 | int congested; |
1722 | int delivered; | |
7d877f3b | 1723 | gfp_t allocation; |
1da177e4 | 1724 | struct sk_buff *skb, *skb2; |
910a7e90 EB |
1725 | int (*tx_filter)(struct sock *dsk, struct sk_buff *skb, void *data); |
1726 | void *tx_data; | |
1da177e4 LT |
1727 | }; |
1728 | ||
b57ef81f | 1729 | static int do_one_broadcast(struct sock *sk, |
1da177e4 LT |
1730 | struct netlink_broadcast_data *p) |
1731 | { | |
1732 | struct netlink_sock *nlk = nlk_sk(sk); | |
1733 | int val; | |
1734 | ||
1735 | if (p->exclude_sk == sk) | |
1736 | goto out; | |
1737 | ||
15e47304 | 1738 | if (nlk->portid == p->portid || p->group - 1 >= nlk->ngroups || |
f7fa9b10 | 1739 | !test_bit(p->group - 1, nlk->groups)) |
1da177e4 LT |
1740 | goto out; |
1741 | ||
878628fb | 1742 | if (!net_eq(sock_net(sk), p->net)) |
b4b51029 EB |
1743 | goto out; |
1744 | ||
1da177e4 LT |
1745 | if (p->failure) { |
1746 | netlink_overrun(sk); | |
1747 | goto out; | |
1748 | } | |
1749 | ||
1750 | sock_hold(sk); | |
1751 | if (p->skb2 == NULL) { | |
68acc024 | 1752 | if (skb_shared(p->skb)) { |
1da177e4 LT |
1753 | p->skb2 = skb_clone(p->skb, p->allocation); |
1754 | } else { | |
68acc024 TC |
1755 | p->skb2 = skb_get(p->skb); |
1756 | /* | |
1757 | * skb ownership may have been set when | |
1758 | * delivered to a previous socket. | |
1759 | */ | |
1760 | skb_orphan(p->skb2); | |
1da177e4 LT |
1761 | } |
1762 | } | |
1763 | if (p->skb2 == NULL) { | |
1764 | netlink_overrun(sk); | |
1765 | /* Clone failed. Notify ALL listeners. */ | |
1766 | p->failure = 1; | |
be0c22a4 PNA |
1767 | if (nlk->flags & NETLINK_BROADCAST_SEND_ERROR) |
1768 | p->delivery_failure = 1; | |
910a7e90 EB |
1769 | } else if (p->tx_filter && p->tx_filter(sk, p->skb2, p->tx_data)) { |
1770 | kfree_skb(p->skb2); | |
1771 | p->skb2 = NULL; | |
b1153f29 SH |
1772 | } else if (sk_filter(sk, p->skb2)) { |
1773 | kfree_skb(p->skb2); | |
1774 | p->skb2 = NULL; | |
1da177e4 LT |
1775 | } else if ((val = netlink_broadcast_deliver(sk, p->skb2)) < 0) { |
1776 | netlink_overrun(sk); | |
be0c22a4 PNA |
1777 | if (nlk->flags & NETLINK_BROADCAST_SEND_ERROR) |
1778 | p->delivery_failure = 1; | |
1da177e4 LT |
1779 | } else { |
1780 | p->congested |= val; | |
1781 | p->delivered = 1; | |
1782 | p->skb2 = NULL; | |
1783 | } | |
1784 | sock_put(sk); | |
1785 | ||
1786 | out: | |
1787 | return 0; | |
1788 | } | |
1789 | ||
15e47304 | 1790 | int netlink_broadcast_filtered(struct sock *ssk, struct sk_buff *skb, u32 portid, |
910a7e90 EB |
1791 | u32 group, gfp_t allocation, |
1792 | int (*filter)(struct sock *dsk, struct sk_buff *skb, void *data), | |
1793 | void *filter_data) | |
1da177e4 | 1794 | { |
3b1e0a65 | 1795 | struct net *net = sock_net(ssk); |
1da177e4 | 1796 | struct netlink_broadcast_data info; |
1da177e4 LT |
1797 | struct sock *sk; |
1798 | ||
1799 | skb = netlink_trim(skb, allocation); | |
1800 | ||
1801 | info.exclude_sk = ssk; | |
b4b51029 | 1802 | info.net = net; |
15e47304 | 1803 | info.portid = portid; |
1da177e4 LT |
1804 | info.group = group; |
1805 | info.failure = 0; | |
ff491a73 | 1806 | info.delivery_failure = 0; |
1da177e4 LT |
1807 | info.congested = 0; |
1808 | info.delivered = 0; | |
1809 | info.allocation = allocation; | |
1810 | info.skb = skb; | |
1811 | info.skb2 = NULL; | |
910a7e90 EB |
1812 | info.tx_filter = filter; |
1813 | info.tx_data = filter_data; | |
1da177e4 LT |
1814 | |
1815 | /* While we sleep in clone, do not allow to change socket list */ | |
1816 | ||
1817 | netlink_lock_table(); | |
1818 | ||
b67bfe0d | 1819 | sk_for_each_bound(sk, &nl_table[ssk->sk_protocol].mc_list) |
1da177e4 LT |
1820 | do_one_broadcast(sk, &info); |
1821 | ||
70d4bf6d | 1822 | consume_skb(skb); |
aa1c6a6f | 1823 | |
1da177e4 LT |
1824 | netlink_unlock_table(); |
1825 | ||
70d4bf6d NH |
1826 | if (info.delivery_failure) { |
1827 | kfree_skb(info.skb2); | |
ff491a73 | 1828 | return -ENOBUFS; |
658cb354 ED |
1829 | } |
1830 | consume_skb(info.skb2); | |
ff491a73 | 1831 | |
1da177e4 LT |
1832 | if (info.delivered) { |
1833 | if (info.congested && (allocation & __GFP_WAIT)) | |
1834 | yield(); | |
1835 | return 0; | |
1836 | } | |
1da177e4 LT |
1837 | return -ESRCH; |
1838 | } | |
910a7e90 EB |
1839 | EXPORT_SYMBOL(netlink_broadcast_filtered); |
1840 | ||
15e47304 | 1841 | int netlink_broadcast(struct sock *ssk, struct sk_buff *skb, u32 portid, |
910a7e90 EB |
1842 | u32 group, gfp_t allocation) |
1843 | { | |
15e47304 | 1844 | return netlink_broadcast_filtered(ssk, skb, portid, group, allocation, |
910a7e90 EB |
1845 | NULL, NULL); |
1846 | } | |
6ac552fd | 1847 | EXPORT_SYMBOL(netlink_broadcast); |
1da177e4 LT |
1848 | |
1849 | struct netlink_set_err_data { | |
1850 | struct sock *exclude_sk; | |
15e47304 | 1851 | u32 portid; |
1da177e4 LT |
1852 | u32 group; |
1853 | int code; | |
1854 | }; | |
1855 | ||
b57ef81f | 1856 | static int do_one_set_err(struct sock *sk, struct netlink_set_err_data *p) |
1da177e4 LT |
1857 | { |
1858 | struct netlink_sock *nlk = nlk_sk(sk); | |
1a50307b | 1859 | int ret = 0; |
1da177e4 LT |
1860 | |
1861 | if (sk == p->exclude_sk) | |
1862 | goto out; | |
1863 | ||
09ad9bc7 | 1864 | if (!net_eq(sock_net(sk), sock_net(p->exclude_sk))) |
b4b51029 EB |
1865 | goto out; |
1866 | ||
15e47304 | 1867 | if (nlk->portid == p->portid || p->group - 1 >= nlk->ngroups || |
f7fa9b10 | 1868 | !test_bit(p->group - 1, nlk->groups)) |
1da177e4 LT |
1869 | goto out; |
1870 | ||
1a50307b PNA |
1871 | if (p->code == ENOBUFS && nlk->flags & NETLINK_RECV_NO_ENOBUFS) { |
1872 | ret = 1; | |
1873 | goto out; | |
1874 | } | |
1875 | ||
1da177e4 LT |
1876 | sk->sk_err = p->code; |
1877 | sk->sk_error_report(sk); | |
1878 | out: | |
1a50307b | 1879 | return ret; |
1da177e4 LT |
1880 | } |
1881 | ||
4843b93c PNA |
1882 | /** |
1883 | * netlink_set_err - report error to broadcast listeners | |
1884 | * @ssk: the kernel netlink socket, as returned by netlink_kernel_create() | |
15e47304 | 1885 | * @portid: the PORTID of a process that we want to skip (if any) |
4843b93c PNA |
1886 | * @groups: the broadcast group that will notice the error |
1887 | * @code: error code, must be negative (as usual in kernelspace) | |
1a50307b PNA |
1888 | * |
1889 | * This function returns the number of broadcast listeners that have set the | |
1890 | * NETLINK_RECV_NO_ENOBUFS socket option. | |
4843b93c | 1891 | */ |
15e47304 | 1892 | int netlink_set_err(struct sock *ssk, u32 portid, u32 group, int code) |
1da177e4 LT |
1893 | { |
1894 | struct netlink_set_err_data info; | |
1da177e4 | 1895 | struct sock *sk; |
1a50307b | 1896 | int ret = 0; |
1da177e4 LT |
1897 | |
1898 | info.exclude_sk = ssk; | |
15e47304 | 1899 | info.portid = portid; |
1da177e4 | 1900 | info.group = group; |
4843b93c PNA |
1901 | /* sk->sk_err wants a positive error value */ |
1902 | info.code = -code; | |
1da177e4 LT |
1903 | |
1904 | read_lock(&nl_table_lock); | |
1905 | ||
b67bfe0d | 1906 | sk_for_each_bound(sk, &nl_table[ssk->sk_protocol].mc_list) |
1a50307b | 1907 | ret += do_one_set_err(sk, &info); |
1da177e4 LT |
1908 | |
1909 | read_unlock(&nl_table_lock); | |
1a50307b | 1910 | return ret; |
1da177e4 | 1911 | } |
dd5b6ce6 | 1912 | EXPORT_SYMBOL(netlink_set_err); |
1da177e4 | 1913 | |
84659eb5 JB |
1914 | /* must be called with netlink table grabbed */ |
1915 | static void netlink_update_socket_mc(struct netlink_sock *nlk, | |
1916 | unsigned int group, | |
1917 | int is_new) | |
1918 | { | |
1919 | int old, new = !!is_new, subscriptions; | |
1920 | ||
1921 | old = test_bit(group - 1, nlk->groups); | |
1922 | subscriptions = nlk->subscriptions - old + new; | |
1923 | if (new) | |
1924 | __set_bit(group - 1, nlk->groups); | |
1925 | else | |
1926 | __clear_bit(group - 1, nlk->groups); | |
1927 | netlink_update_subscriptions(&nlk->sk, subscriptions); | |
1928 | netlink_update_listeners(&nlk->sk); | |
1929 | } | |
1930 | ||
9a4595bc | 1931 | static int netlink_setsockopt(struct socket *sock, int level, int optname, |
b7058842 | 1932 | char __user *optval, unsigned int optlen) |
9a4595bc PM |
1933 | { |
1934 | struct sock *sk = sock->sk; | |
1935 | struct netlink_sock *nlk = nlk_sk(sk); | |
eb496534 JB |
1936 | unsigned int val = 0; |
1937 | int err; | |
9a4595bc PM |
1938 | |
1939 | if (level != SOL_NETLINK) | |
1940 | return -ENOPROTOOPT; | |
1941 | ||
ccdfcc39 PM |
1942 | if (optname != NETLINK_RX_RING && optname != NETLINK_TX_RING && |
1943 | optlen >= sizeof(int) && | |
eb496534 | 1944 | get_user(val, (unsigned int __user *)optval)) |
9a4595bc PM |
1945 | return -EFAULT; |
1946 | ||
1947 | switch (optname) { | |
1948 | case NETLINK_PKTINFO: | |
1949 | if (val) | |
1950 | nlk->flags |= NETLINK_RECV_PKTINFO; | |
1951 | else | |
1952 | nlk->flags &= ~NETLINK_RECV_PKTINFO; | |
1953 | err = 0; | |
1954 | break; | |
1955 | case NETLINK_ADD_MEMBERSHIP: | |
1956 | case NETLINK_DROP_MEMBERSHIP: { | |
9785e10a | 1957 | if (!netlink_capable(sock, NL_CFG_F_NONROOT_RECV)) |
9a4595bc | 1958 | return -EPERM; |
b4ff4f04 JB |
1959 | err = netlink_realloc_groups(sk); |
1960 | if (err) | |
1961 | return err; | |
9a4595bc PM |
1962 | if (!val || val - 1 >= nlk->ngroups) |
1963 | return -EINVAL; | |
1964 | netlink_table_grab(); | |
84659eb5 JB |
1965 | netlink_update_socket_mc(nlk, val, |
1966 | optname == NETLINK_ADD_MEMBERSHIP); | |
9a4595bc | 1967 | netlink_table_ungrab(); |
03292745 PNA |
1968 | |
1969 | if (nlk->netlink_bind) | |
1970 | nlk->netlink_bind(val); | |
1971 | ||
9a4595bc PM |
1972 | err = 0; |
1973 | break; | |
1974 | } | |
be0c22a4 PNA |
1975 | case NETLINK_BROADCAST_ERROR: |
1976 | if (val) | |
1977 | nlk->flags |= NETLINK_BROADCAST_SEND_ERROR; | |
1978 | else | |
1979 | nlk->flags &= ~NETLINK_BROADCAST_SEND_ERROR; | |
1980 | err = 0; | |
1981 | break; | |
38938bfe PNA |
1982 | case NETLINK_NO_ENOBUFS: |
1983 | if (val) { | |
1984 | nlk->flags |= NETLINK_RECV_NO_ENOBUFS; | |
cd967e05 | 1985 | clear_bit(NETLINK_CONGESTED, &nlk->state); |
38938bfe | 1986 | wake_up_interruptible(&nlk->wait); |
658cb354 | 1987 | } else { |
38938bfe | 1988 | nlk->flags &= ~NETLINK_RECV_NO_ENOBUFS; |
658cb354 | 1989 | } |
38938bfe PNA |
1990 | err = 0; |
1991 | break; | |
ccdfcc39 PM |
1992 | #ifdef CONFIG_NETLINK_MMAP |
1993 | case NETLINK_RX_RING: | |
1994 | case NETLINK_TX_RING: { | |
1995 | struct nl_mmap_req req; | |
1996 | ||
1997 | /* Rings might consume more memory than queue limits, require | |
1998 | * CAP_NET_ADMIN. | |
1999 | */ | |
2000 | if (!capable(CAP_NET_ADMIN)) | |
2001 | return -EPERM; | |
2002 | if (optlen < sizeof(req)) | |
2003 | return -EINVAL; | |
2004 | if (copy_from_user(&req, optval, sizeof(req))) | |
2005 | return -EFAULT; | |
2006 | err = netlink_set_ring(sk, &req, false, | |
2007 | optname == NETLINK_TX_RING); | |
2008 | break; | |
2009 | } | |
2010 | #endif /* CONFIG_NETLINK_MMAP */ | |
9a4595bc PM |
2011 | default: |
2012 | err = -ENOPROTOOPT; | |
2013 | } | |
2014 | return err; | |
2015 | } | |
2016 | ||
2017 | static int netlink_getsockopt(struct socket *sock, int level, int optname, | |
746fac4d | 2018 | char __user *optval, int __user *optlen) |
9a4595bc PM |
2019 | { |
2020 | struct sock *sk = sock->sk; | |
2021 | struct netlink_sock *nlk = nlk_sk(sk); | |
2022 | int len, val, err; | |
2023 | ||
2024 | if (level != SOL_NETLINK) | |
2025 | return -ENOPROTOOPT; | |
2026 | ||
2027 | if (get_user(len, optlen)) | |
2028 | return -EFAULT; | |
2029 | if (len < 0) | |
2030 | return -EINVAL; | |
2031 | ||
2032 | switch (optname) { | |
2033 | case NETLINK_PKTINFO: | |
2034 | if (len < sizeof(int)) | |
2035 | return -EINVAL; | |
2036 | len = sizeof(int); | |
2037 | val = nlk->flags & NETLINK_RECV_PKTINFO ? 1 : 0; | |
a27b58fe HC |
2038 | if (put_user(len, optlen) || |
2039 | put_user(val, optval)) | |
2040 | return -EFAULT; | |
9a4595bc PM |
2041 | err = 0; |
2042 | break; | |
be0c22a4 PNA |
2043 | case NETLINK_BROADCAST_ERROR: |
2044 | if (len < sizeof(int)) | |
2045 | return -EINVAL; | |
2046 | len = sizeof(int); | |
2047 | val = nlk->flags & NETLINK_BROADCAST_SEND_ERROR ? 1 : 0; | |
2048 | if (put_user(len, optlen) || | |
2049 | put_user(val, optval)) | |
2050 | return -EFAULT; | |
2051 | err = 0; | |
2052 | break; | |
38938bfe PNA |
2053 | case NETLINK_NO_ENOBUFS: |
2054 | if (len < sizeof(int)) | |
2055 | return -EINVAL; | |
2056 | len = sizeof(int); | |
2057 | val = nlk->flags & NETLINK_RECV_NO_ENOBUFS ? 1 : 0; | |
2058 | if (put_user(len, optlen) || | |
2059 | put_user(val, optval)) | |
2060 | return -EFAULT; | |
2061 | err = 0; | |
2062 | break; | |
9a4595bc PM |
2063 | default: |
2064 | err = -ENOPROTOOPT; | |
2065 | } | |
2066 | return err; | |
2067 | } | |
2068 | ||
2069 | static void netlink_cmsg_recv_pktinfo(struct msghdr *msg, struct sk_buff *skb) | |
2070 | { | |
2071 | struct nl_pktinfo info; | |
2072 | ||
2073 | info.group = NETLINK_CB(skb).dst_group; | |
2074 | put_cmsg(msg, SOL_NETLINK, NETLINK_PKTINFO, sizeof(info), &info); | |
2075 | } | |
2076 | ||
1da177e4 LT |
2077 | static int netlink_sendmsg(struct kiocb *kiocb, struct socket *sock, |
2078 | struct msghdr *msg, size_t len) | |
2079 | { | |
2080 | struct sock_iocb *siocb = kiocb_to_siocb(kiocb); | |
2081 | struct sock *sk = sock->sk; | |
2082 | struct netlink_sock *nlk = nlk_sk(sk); | |
6ac552fd | 2083 | struct sockaddr_nl *addr = msg->msg_name; |
15e47304 | 2084 | u32 dst_portid; |
d629b836 | 2085 | u32 dst_group; |
1da177e4 LT |
2086 | struct sk_buff *skb; |
2087 | int err; | |
2088 | struct scm_cookie scm; | |
2089 | ||
2090 | if (msg->msg_flags&MSG_OOB) | |
2091 | return -EOPNOTSUPP; | |
2092 | ||
16e57262 | 2093 | if (NULL == siocb->scm) |
1da177e4 | 2094 | siocb->scm = &scm; |
16e57262 | 2095 | |
e0e3cea4 | 2096 | err = scm_send(sock, msg, siocb->scm, true); |
1da177e4 LT |
2097 | if (err < 0) |
2098 | return err; | |
2099 | ||
2100 | if (msg->msg_namelen) { | |
b47030c7 | 2101 | err = -EINVAL; |
1da177e4 | 2102 | if (addr->nl_family != AF_NETLINK) |
b47030c7 | 2103 | goto out; |
15e47304 | 2104 | dst_portid = addr->nl_pid; |
d629b836 | 2105 | dst_group = ffs(addr->nl_groups); |
b47030c7 | 2106 | err = -EPERM; |
15e47304 | 2107 | if ((dst_group || dst_portid) && |
9785e10a | 2108 | !netlink_capable(sock, NL_CFG_F_NONROOT_SEND)) |
b47030c7 | 2109 | goto out; |
1da177e4 | 2110 | } else { |
15e47304 | 2111 | dst_portid = nlk->dst_portid; |
d629b836 | 2112 | dst_group = nlk->dst_group; |
1da177e4 LT |
2113 | } |
2114 | ||
15e47304 | 2115 | if (!nlk->portid) { |
1da177e4 LT |
2116 | err = netlink_autobind(sock); |
2117 | if (err) | |
2118 | goto out; | |
2119 | } | |
2120 | ||
5fd96123 PM |
2121 | if (netlink_tx_is_mmaped(sk) && |
2122 | msg->msg_iov->iov_base == NULL) { | |
2123 | err = netlink_mmap_sendmsg(sk, msg, dst_portid, dst_group, | |
2124 | siocb); | |
2125 | goto out; | |
2126 | } | |
2127 | ||
1da177e4 LT |
2128 | err = -EMSGSIZE; |
2129 | if (len > sk->sk_sndbuf - 32) | |
2130 | goto out; | |
2131 | err = -ENOBUFS; | |
c05cdb1b | 2132 | skb = netlink_alloc_large_skb(len); |
6ac552fd | 2133 | if (skb == NULL) |
1da177e4 LT |
2134 | goto out; |
2135 | ||
15e47304 | 2136 | NETLINK_CB(skb).portid = nlk->portid; |
d629b836 | 2137 | NETLINK_CB(skb).dst_group = dst_group; |
dbe9a417 | 2138 | NETLINK_CB(skb).creds = siocb->scm->creds; |
1da177e4 | 2139 | |
1da177e4 | 2140 | err = -EFAULT; |
6ac552fd | 2141 | if (memcpy_fromiovec(skb_put(skb, len), msg->msg_iov, len)) { |
1da177e4 LT |
2142 | kfree_skb(skb); |
2143 | goto out; | |
2144 | } | |
2145 | ||
2146 | err = security_netlink_send(sk, skb); | |
2147 | if (err) { | |
2148 | kfree_skb(skb); | |
2149 | goto out; | |
2150 | } | |
2151 | ||
d629b836 | 2152 | if (dst_group) { |
1da177e4 | 2153 | atomic_inc(&skb->users); |
15e47304 | 2154 | netlink_broadcast(sk, skb, dst_portid, dst_group, GFP_KERNEL); |
1da177e4 | 2155 | } |
15e47304 | 2156 | err = netlink_unicast(sk, skb, dst_portid, msg->msg_flags&MSG_DONTWAIT); |
1da177e4 LT |
2157 | |
2158 | out: | |
b47030c7 | 2159 | scm_destroy(siocb->scm); |
1da177e4 LT |
2160 | return err; |
2161 | } | |
2162 | ||
2163 | static int netlink_recvmsg(struct kiocb *kiocb, struct socket *sock, | |
2164 | struct msghdr *msg, size_t len, | |
2165 | int flags) | |
2166 | { | |
2167 | struct sock_iocb *siocb = kiocb_to_siocb(kiocb); | |
2168 | struct scm_cookie scm; | |
2169 | struct sock *sk = sock->sk; | |
2170 | struct netlink_sock *nlk = nlk_sk(sk); | |
2171 | int noblock = flags&MSG_DONTWAIT; | |
2172 | size_t copied; | |
68d6ac6d | 2173 | struct sk_buff *skb, *data_skb; |
b44d211e | 2174 | int err, ret; |
1da177e4 LT |
2175 | |
2176 | if (flags&MSG_OOB) | |
2177 | return -EOPNOTSUPP; | |
2178 | ||
2179 | copied = 0; | |
2180 | ||
6ac552fd PM |
2181 | skb = skb_recv_datagram(sk, flags, noblock, &err); |
2182 | if (skb == NULL) | |
1da177e4 LT |
2183 | goto out; |
2184 | ||
68d6ac6d JB |
2185 | data_skb = skb; |
2186 | ||
1dacc76d JB |
2187 | #ifdef CONFIG_COMPAT_NETLINK_MESSAGES |
2188 | if (unlikely(skb_shinfo(skb)->frag_list)) { | |
1dacc76d | 2189 | /* |
68d6ac6d JB |
2190 | * If this skb has a frag_list, then here that means that we |
2191 | * will have to use the frag_list skb's data for compat tasks | |
2192 | * and the regular skb's data for normal (non-compat) tasks. | |
1dacc76d | 2193 | * |
68d6ac6d JB |
2194 | * If we need to send the compat skb, assign it to the |
2195 | * 'data_skb' variable so that it will be used below for data | |
2196 | * copying. We keep 'skb' for everything else, including | |
2197 | * freeing both later. | |
1dacc76d | 2198 | */ |
68d6ac6d JB |
2199 | if (flags & MSG_CMSG_COMPAT) |
2200 | data_skb = skb_shinfo(skb)->frag_list; | |
1dacc76d JB |
2201 | } |
2202 | #endif | |
2203 | ||
1da177e4 LT |
2204 | msg->msg_namelen = 0; |
2205 | ||
68d6ac6d | 2206 | copied = data_skb->len; |
1da177e4 LT |
2207 | if (len < copied) { |
2208 | msg->msg_flags |= MSG_TRUNC; | |
2209 | copied = len; | |
2210 | } | |
2211 | ||
68d6ac6d JB |
2212 | skb_reset_transport_header(data_skb); |
2213 | err = skb_copy_datagram_iovec(data_skb, 0, msg->msg_iov, copied); | |
1da177e4 LT |
2214 | |
2215 | if (msg->msg_name) { | |
6ac552fd | 2216 | struct sockaddr_nl *addr = (struct sockaddr_nl *)msg->msg_name; |
1da177e4 LT |
2217 | addr->nl_family = AF_NETLINK; |
2218 | addr->nl_pad = 0; | |
15e47304 | 2219 | addr->nl_pid = NETLINK_CB(skb).portid; |
d629b836 | 2220 | addr->nl_groups = netlink_group_mask(NETLINK_CB(skb).dst_group); |
1da177e4 LT |
2221 | msg->msg_namelen = sizeof(*addr); |
2222 | } | |
2223 | ||
cc9a06cd PM |
2224 | if (nlk->flags & NETLINK_RECV_PKTINFO) |
2225 | netlink_cmsg_recv_pktinfo(msg, skb); | |
2226 | ||
1da177e4 LT |
2227 | if (NULL == siocb->scm) { |
2228 | memset(&scm, 0, sizeof(scm)); | |
2229 | siocb->scm = &scm; | |
2230 | } | |
2231 | siocb->scm->creds = *NETLINK_CREDS(skb); | |
188ccb55 | 2232 | if (flags & MSG_TRUNC) |
68d6ac6d | 2233 | copied = data_skb->len; |
daa3766e | 2234 | |
1da177e4 LT |
2235 | skb_free_datagram(sk, skb); |
2236 | ||
b44d211e AV |
2237 | if (nlk->cb && atomic_read(&sk->sk_rmem_alloc) <= sk->sk_rcvbuf / 2) { |
2238 | ret = netlink_dump(sk); | |
2239 | if (ret) { | |
2240 | sk->sk_err = ret; | |
2241 | sk->sk_error_report(sk); | |
2242 | } | |
2243 | } | |
1da177e4 LT |
2244 | |
2245 | scm_recv(sock, msg, siocb->scm, flags); | |
1da177e4 LT |
2246 | out: |
2247 | netlink_rcv_wake(sk); | |
2248 | return err ? : copied; | |
2249 | } | |
2250 | ||
2251 | static void netlink_data_ready(struct sock *sk, int len) | |
2252 | { | |
cd40b7d3 | 2253 | BUG(); |
1da177e4 LT |
2254 | } |
2255 | ||
2256 | /* | |
746fac4d | 2257 | * We export these functions to other modules. They provide a |
1da177e4 LT |
2258 | * complete set of kernel non-blocking support for message |
2259 | * queueing. | |
2260 | */ | |
2261 | ||
2262 | struct sock * | |
9f00d977 PNA |
2263 | __netlink_kernel_create(struct net *net, int unit, struct module *module, |
2264 | struct netlink_kernel_cfg *cfg) | |
1da177e4 LT |
2265 | { |
2266 | struct socket *sock; | |
2267 | struct sock *sk; | |
77247bbb | 2268 | struct netlink_sock *nlk; |
5c398dc8 | 2269 | struct listeners *listeners = NULL; |
a31f2d17 PNA |
2270 | struct mutex *cb_mutex = cfg ? cfg->cb_mutex : NULL; |
2271 | unsigned int groups; | |
1da177e4 | 2272 | |
fab2caf6 | 2273 | BUG_ON(!nl_table); |
1da177e4 | 2274 | |
6ac552fd | 2275 | if (unit < 0 || unit >= MAX_LINKS) |
1da177e4 LT |
2276 | return NULL; |
2277 | ||
2278 | if (sock_create_lite(PF_NETLINK, SOCK_DGRAM, unit, &sock)) | |
2279 | return NULL; | |
2280 | ||
23fe1866 PE |
2281 | /* |
2282 | * We have to just have a reference on the net from sk, but don't | |
2283 | * get_net it. Besides, we cannot get and then put the net here. | |
2284 | * So we create one inside init_net and the move it to net. | |
2285 | */ | |
2286 | ||
2287 | if (__netlink_create(&init_net, sock, cb_mutex, unit) < 0) | |
2288 | goto out_sock_release_nosk; | |
2289 | ||
2290 | sk = sock->sk; | |
edf02087 | 2291 | sk_change_net(sk, net); |
4fdb3bb7 | 2292 | |
a31f2d17 | 2293 | if (!cfg || cfg->groups < 32) |
4277a083 | 2294 | groups = 32; |
a31f2d17 PNA |
2295 | else |
2296 | groups = cfg->groups; | |
4277a083 | 2297 | |
5c398dc8 | 2298 | listeners = kzalloc(sizeof(*listeners) + NLGRPSZ(groups), GFP_KERNEL); |
4277a083 PM |
2299 | if (!listeners) |
2300 | goto out_sock_release; | |
2301 | ||
1da177e4 | 2302 | sk->sk_data_ready = netlink_data_ready; |
a31f2d17 PNA |
2303 | if (cfg && cfg->input) |
2304 | nlk_sk(sk)->netlink_rcv = cfg->input; | |
1da177e4 | 2305 | |
b4b51029 | 2306 | if (netlink_insert(sk, net, 0)) |
77247bbb | 2307 | goto out_sock_release; |
4fdb3bb7 | 2308 | |
77247bbb PM |
2309 | nlk = nlk_sk(sk); |
2310 | nlk->flags |= NETLINK_KERNEL_SOCKET; | |
4fdb3bb7 | 2311 | |
4fdb3bb7 | 2312 | netlink_table_grab(); |
b4b51029 EB |
2313 | if (!nl_table[unit].registered) { |
2314 | nl_table[unit].groups = groups; | |
5c398dc8 | 2315 | rcu_assign_pointer(nl_table[unit].listeners, listeners); |
b4b51029 EB |
2316 | nl_table[unit].cb_mutex = cb_mutex; |
2317 | nl_table[unit].module = module; | |
9785e10a PNA |
2318 | if (cfg) { |
2319 | nl_table[unit].bind = cfg->bind; | |
2320 | nl_table[unit].flags = cfg->flags; | |
2321 | } | |
b4b51029 | 2322 | nl_table[unit].registered = 1; |
f937f1f4 JJ |
2323 | } else { |
2324 | kfree(listeners); | |
869e58f8 | 2325 | nl_table[unit].registered++; |
b4b51029 | 2326 | } |
4fdb3bb7 | 2327 | netlink_table_ungrab(); |
77247bbb PM |
2328 | return sk; |
2329 | ||
4fdb3bb7 | 2330 | out_sock_release: |
4277a083 | 2331 | kfree(listeners); |
9dfbec1f | 2332 | netlink_kernel_release(sk); |
23fe1866 PE |
2333 | return NULL; |
2334 | ||
2335 | out_sock_release_nosk: | |
4fdb3bb7 | 2336 | sock_release(sock); |
77247bbb | 2337 | return NULL; |
1da177e4 | 2338 | } |
9f00d977 | 2339 | EXPORT_SYMBOL(__netlink_kernel_create); |
b7c6ba6e DL |
2340 | |
2341 | void | |
2342 | netlink_kernel_release(struct sock *sk) | |
2343 | { | |
edf02087 | 2344 | sk_release_kernel(sk); |
b7c6ba6e DL |
2345 | } |
2346 | EXPORT_SYMBOL(netlink_kernel_release); | |
2347 | ||
d136f1bd | 2348 | int __netlink_change_ngroups(struct sock *sk, unsigned int groups) |
b4ff4f04 | 2349 | { |
5c398dc8 | 2350 | struct listeners *new, *old; |
b4ff4f04 | 2351 | struct netlink_table *tbl = &nl_table[sk->sk_protocol]; |
b4ff4f04 JB |
2352 | |
2353 | if (groups < 32) | |
2354 | groups = 32; | |
2355 | ||
b4ff4f04 | 2356 | if (NLGRPSZ(tbl->groups) < NLGRPSZ(groups)) { |
5c398dc8 ED |
2357 | new = kzalloc(sizeof(*new) + NLGRPSZ(groups), GFP_ATOMIC); |
2358 | if (!new) | |
d136f1bd | 2359 | return -ENOMEM; |
6d772ac5 | 2360 | old = nl_deref_protected(tbl->listeners); |
5c398dc8 ED |
2361 | memcpy(new->masks, old->masks, NLGRPSZ(tbl->groups)); |
2362 | rcu_assign_pointer(tbl->listeners, new); | |
2363 | ||
37b6b935 | 2364 | kfree_rcu(old, rcu); |
b4ff4f04 JB |
2365 | } |
2366 | tbl->groups = groups; | |
2367 | ||
d136f1bd JB |
2368 | return 0; |
2369 | } | |
2370 | ||
2371 | /** | |
2372 | * netlink_change_ngroups - change number of multicast groups | |
2373 | * | |
2374 | * This changes the number of multicast groups that are available | |
2375 | * on a certain netlink family. Note that it is not possible to | |
2376 | * change the number of groups to below 32. Also note that it does | |
2377 | * not implicitly call netlink_clear_multicast_users() when the | |
2378 | * number of groups is reduced. | |
2379 | * | |
2380 | * @sk: The kernel netlink socket, as returned by netlink_kernel_create(). | |
2381 | * @groups: The new number of groups. | |
2382 | */ | |
2383 | int netlink_change_ngroups(struct sock *sk, unsigned int groups) | |
2384 | { | |
2385 | int err; | |
2386 | ||
2387 | netlink_table_grab(); | |
2388 | err = __netlink_change_ngroups(sk, groups); | |
b4ff4f04 | 2389 | netlink_table_ungrab(); |
d136f1bd | 2390 | |
b4ff4f04 JB |
2391 | return err; |
2392 | } | |
b4ff4f04 | 2393 | |
b8273570 JB |
2394 | void __netlink_clear_multicast_users(struct sock *ksk, unsigned int group) |
2395 | { | |
2396 | struct sock *sk; | |
b8273570 JB |
2397 | struct netlink_table *tbl = &nl_table[ksk->sk_protocol]; |
2398 | ||
b67bfe0d | 2399 | sk_for_each_bound(sk, &tbl->mc_list) |
b8273570 JB |
2400 | netlink_update_socket_mc(nlk_sk(sk), group, 0); |
2401 | } | |
2402 | ||
84659eb5 JB |
2403 | /** |
2404 | * netlink_clear_multicast_users - kick off multicast listeners | |
2405 | * | |
2406 | * This function removes all listeners from the given group. | |
2407 | * @ksk: The kernel netlink socket, as returned by | |
2408 | * netlink_kernel_create(). | |
2409 | * @group: The multicast group to clear. | |
2410 | */ | |
2411 | void netlink_clear_multicast_users(struct sock *ksk, unsigned int group) | |
2412 | { | |
84659eb5 | 2413 | netlink_table_grab(); |
b8273570 | 2414 | __netlink_clear_multicast_users(ksk, group); |
84659eb5 JB |
2415 | netlink_table_ungrab(); |
2416 | } | |
84659eb5 | 2417 | |
a46621a3 | 2418 | struct nlmsghdr * |
15e47304 | 2419 | __nlmsg_put(struct sk_buff *skb, u32 portid, u32 seq, int type, int len, int flags) |
a46621a3 DV |
2420 | { |
2421 | struct nlmsghdr *nlh; | |
573ce260 | 2422 | int size = nlmsg_msg_size(len); |
a46621a3 DV |
2423 | |
2424 | nlh = (struct nlmsghdr*)skb_put(skb, NLMSG_ALIGN(size)); | |
2425 | nlh->nlmsg_type = type; | |
2426 | nlh->nlmsg_len = size; | |
2427 | nlh->nlmsg_flags = flags; | |
15e47304 | 2428 | nlh->nlmsg_pid = portid; |
a46621a3 DV |
2429 | nlh->nlmsg_seq = seq; |
2430 | if (!__builtin_constant_p(size) || NLMSG_ALIGN(size) - size != 0) | |
573ce260 | 2431 | memset(nlmsg_data(nlh) + len, 0, NLMSG_ALIGN(size) - size); |
a46621a3 DV |
2432 | return nlh; |
2433 | } | |
2434 | EXPORT_SYMBOL(__nlmsg_put); | |
2435 | ||
1da177e4 LT |
2436 | /* |
2437 | * It looks a bit ugly. | |
2438 | * It would be better to create kernel thread. | |
2439 | */ | |
2440 | ||
2441 | static int netlink_dump(struct sock *sk) | |
2442 | { | |
2443 | struct netlink_sock *nlk = nlk_sk(sk); | |
2444 | struct netlink_callback *cb; | |
c7ac8679 | 2445 | struct sk_buff *skb = NULL; |
1da177e4 | 2446 | struct nlmsghdr *nlh; |
bf8b79e4 | 2447 | int len, err = -ENOBUFS; |
c7ac8679 | 2448 | int alloc_size; |
1da177e4 | 2449 | |
af65bdfc | 2450 | mutex_lock(nlk->cb_mutex); |
1da177e4 LT |
2451 | |
2452 | cb = nlk->cb; | |
2453 | if (cb == NULL) { | |
bf8b79e4 TG |
2454 | err = -EINVAL; |
2455 | goto errout_skb; | |
1da177e4 LT |
2456 | } |
2457 | ||
c7ac8679 GR |
2458 | alloc_size = max_t(int, cb->min_dump_alloc, NLMSG_GOODSIZE); |
2459 | ||
f9c22888 PM |
2460 | if (!netlink_rx_is_mmaped(sk) && |
2461 | atomic_read(&sk->sk_rmem_alloc) >= sk->sk_rcvbuf) | |
2462 | goto errout_skb; | |
2463 | skb = netlink_alloc_skb(sk, alloc_size, nlk->portid, GFP_KERNEL); | |
c7ac8679 | 2464 | if (!skb) |
c63d6ea3 | 2465 | goto errout_skb; |
f9c22888 | 2466 | netlink_skb_set_owner_r(skb, sk); |
c7ac8679 | 2467 | |
1da177e4 LT |
2468 | len = cb->dump(skb, cb); |
2469 | ||
2470 | if (len > 0) { | |
af65bdfc | 2471 | mutex_unlock(nlk->cb_mutex); |
b1153f29 SH |
2472 | |
2473 | if (sk_filter(sk, skb)) | |
2474 | kfree_skb(skb); | |
4a7e7c2a ED |
2475 | else |
2476 | __netlink_sendskb(sk, skb); | |
1da177e4 LT |
2477 | return 0; |
2478 | } | |
2479 | ||
bf8b79e4 TG |
2480 | nlh = nlmsg_put_answer(skb, cb, NLMSG_DONE, sizeof(len), NLM_F_MULTI); |
2481 | if (!nlh) | |
2482 | goto errout_skb; | |
2483 | ||
670dc283 JB |
2484 | nl_dump_check_consistent(cb, nlh); |
2485 | ||
bf8b79e4 TG |
2486 | memcpy(nlmsg_data(nlh), &len, sizeof(len)); |
2487 | ||
b1153f29 SH |
2488 | if (sk_filter(sk, skb)) |
2489 | kfree_skb(skb); | |
4a7e7c2a ED |
2490 | else |
2491 | __netlink_sendskb(sk, skb); | |
1da177e4 | 2492 | |
a8f74b22 TG |
2493 | if (cb->done) |
2494 | cb->done(cb); | |
1da177e4 | 2495 | nlk->cb = NULL; |
af65bdfc | 2496 | mutex_unlock(nlk->cb_mutex); |
1da177e4 | 2497 | |
6dc878a8 | 2498 | module_put(cb->module); |
bfb253c9 | 2499 | netlink_consume_callback(cb); |
1da177e4 | 2500 | return 0; |
1797754e | 2501 | |
bf8b79e4 | 2502 | errout_skb: |
af65bdfc | 2503 | mutex_unlock(nlk->cb_mutex); |
bf8b79e4 | 2504 | kfree_skb(skb); |
bf8b79e4 | 2505 | return err; |
1da177e4 LT |
2506 | } |
2507 | ||
6dc878a8 G |
2508 | int __netlink_dump_start(struct sock *ssk, struct sk_buff *skb, |
2509 | const struct nlmsghdr *nlh, | |
2510 | struct netlink_dump_control *control) | |
1da177e4 LT |
2511 | { |
2512 | struct netlink_callback *cb; | |
2513 | struct sock *sk; | |
2514 | struct netlink_sock *nlk; | |
b44d211e | 2515 | int ret; |
1da177e4 | 2516 | |
0da974f4 | 2517 | cb = kzalloc(sizeof(*cb), GFP_KERNEL); |
1da177e4 LT |
2518 | if (cb == NULL) |
2519 | return -ENOBUFS; | |
2520 | ||
f9c22888 PM |
2521 | /* Memory mapped dump requests need to be copied to avoid looping |
2522 | * on the pending state in netlink_mmap_sendmsg() while the CB hold | |
2523 | * a reference to the skb. | |
2524 | */ | |
2525 | if (netlink_skb_is_mmaped(skb)) { | |
2526 | skb = skb_copy(skb, GFP_KERNEL); | |
2527 | if (skb == NULL) { | |
2528 | kfree(cb); | |
2529 | return -ENOBUFS; | |
2530 | } | |
2531 | } else | |
2532 | atomic_inc(&skb->users); | |
2533 | ||
80d326fa PNA |
2534 | cb->dump = control->dump; |
2535 | cb->done = control->done; | |
1da177e4 | 2536 | cb->nlh = nlh; |
7175c883 | 2537 | cb->data = control->data; |
6dc878a8 | 2538 | cb->module = control->module; |
80d326fa | 2539 | cb->min_dump_alloc = control->min_dump_alloc; |
1da177e4 LT |
2540 | cb->skb = skb; |
2541 | ||
15e47304 | 2542 | sk = netlink_lookup(sock_net(ssk), ssk->sk_protocol, NETLINK_CB(skb).portid); |
1da177e4 LT |
2543 | if (sk == NULL) { |
2544 | netlink_destroy_callback(cb); | |
2545 | return -ECONNREFUSED; | |
2546 | } | |
2547 | nlk = nlk_sk(sk); | |
6dc878a8 | 2548 | |
af65bdfc | 2549 | mutex_lock(nlk->cb_mutex); |
6dc878a8 | 2550 | /* A dump is in progress... */ |
3f660d66 | 2551 | if (nlk->cb) { |
af65bdfc | 2552 | mutex_unlock(nlk->cb_mutex); |
1da177e4 | 2553 | netlink_destroy_callback(cb); |
6dc878a8 G |
2554 | ret = -EBUSY; |
2555 | goto out; | |
1da177e4 | 2556 | } |
6dc878a8 G |
2557 | /* add reference of module which cb->dump belongs to */ |
2558 | if (!try_module_get(cb->module)) { | |
2559 | mutex_unlock(nlk->cb_mutex); | |
2560 | netlink_destroy_callback(cb); | |
2561 | ret = -EPROTONOSUPPORT; | |
2562 | goto out; | |
2563 | } | |
2564 | ||
1da177e4 | 2565 | nlk->cb = cb; |
af65bdfc | 2566 | mutex_unlock(nlk->cb_mutex); |
1da177e4 | 2567 | |
b44d211e | 2568 | ret = netlink_dump(sk); |
6dc878a8 | 2569 | out: |
1da177e4 | 2570 | sock_put(sk); |
5c58298c | 2571 | |
b44d211e AV |
2572 | if (ret) |
2573 | return ret; | |
2574 | ||
5c58298c DL |
2575 | /* We successfully started a dump, by returning -EINTR we |
2576 | * signal not to send ACK even if it was requested. | |
2577 | */ | |
2578 | return -EINTR; | |
1da177e4 | 2579 | } |
6dc878a8 | 2580 | EXPORT_SYMBOL(__netlink_dump_start); |
1da177e4 LT |
2581 | |
2582 | void netlink_ack(struct sk_buff *in_skb, struct nlmsghdr *nlh, int err) | |
2583 | { | |
2584 | struct sk_buff *skb; | |
2585 | struct nlmsghdr *rep; | |
2586 | struct nlmsgerr *errmsg; | |
339bf98f | 2587 | size_t payload = sizeof(*errmsg); |
1da177e4 | 2588 | |
339bf98f TG |
2589 | /* error messages get the original request appened */ |
2590 | if (err) | |
2591 | payload += nlmsg_len(nlh); | |
1da177e4 | 2592 | |
f9c22888 PM |
2593 | skb = netlink_alloc_skb(in_skb->sk, nlmsg_total_size(payload), |
2594 | NETLINK_CB(in_skb).portid, GFP_KERNEL); | |
1da177e4 LT |
2595 | if (!skb) { |
2596 | struct sock *sk; | |
2597 | ||
3b1e0a65 | 2598 | sk = netlink_lookup(sock_net(in_skb->sk), |
b4b51029 | 2599 | in_skb->sk->sk_protocol, |
15e47304 | 2600 | NETLINK_CB(in_skb).portid); |
1da177e4 LT |
2601 | if (sk) { |
2602 | sk->sk_err = ENOBUFS; | |
2603 | sk->sk_error_report(sk); | |
2604 | sock_put(sk); | |
2605 | } | |
2606 | return; | |
2607 | } | |
2608 | ||
15e47304 | 2609 | rep = __nlmsg_put(skb, NETLINK_CB(in_skb).portid, nlh->nlmsg_seq, |
5dba93ae | 2610 | NLMSG_ERROR, payload, 0); |
bf8b79e4 | 2611 | errmsg = nlmsg_data(rep); |
1da177e4 | 2612 | errmsg->error = err; |
bf8b79e4 | 2613 | memcpy(&errmsg->msg, nlh, err ? nlh->nlmsg_len : sizeof(*nlh)); |
15e47304 | 2614 | netlink_unicast(in_skb->sk, skb, NETLINK_CB(in_skb).portid, MSG_DONTWAIT); |
1da177e4 | 2615 | } |
6ac552fd | 2616 | EXPORT_SYMBOL(netlink_ack); |
1da177e4 | 2617 | |
cd40b7d3 | 2618 | int netlink_rcv_skb(struct sk_buff *skb, int (*cb)(struct sk_buff *, |
1d00a4eb | 2619 | struct nlmsghdr *)) |
82ace47a | 2620 | { |
82ace47a TG |
2621 | struct nlmsghdr *nlh; |
2622 | int err; | |
2623 | ||
2624 | while (skb->len >= nlmsg_total_size(0)) { | |
cd40b7d3 DL |
2625 | int msglen; |
2626 | ||
b529ccf2 | 2627 | nlh = nlmsg_hdr(skb); |
d35b6856 | 2628 | err = 0; |
82ace47a | 2629 | |
ad8e4b75 | 2630 | if (nlh->nlmsg_len < NLMSG_HDRLEN || skb->len < nlh->nlmsg_len) |
82ace47a TG |
2631 | return 0; |
2632 | ||
d35b6856 TG |
2633 | /* Only requests are handled by the kernel */ |
2634 | if (!(nlh->nlmsg_flags & NLM_F_REQUEST)) | |
5c58298c | 2635 | goto ack; |
45e7ae7f TG |
2636 | |
2637 | /* Skip control messages */ | |
2638 | if (nlh->nlmsg_type < NLMSG_MIN_TYPE) | |
5c58298c | 2639 | goto ack; |
d35b6856 | 2640 | |
1d00a4eb | 2641 | err = cb(skb, nlh); |
5c58298c DL |
2642 | if (err == -EINTR) |
2643 | goto skip; | |
2644 | ||
2645 | ack: | |
d35b6856 | 2646 | if (nlh->nlmsg_flags & NLM_F_ACK || err) |
82ace47a | 2647 | netlink_ack(skb, nlh, err); |
82ace47a | 2648 | |
5c58298c | 2649 | skip: |
6ac552fd | 2650 | msglen = NLMSG_ALIGN(nlh->nlmsg_len); |
cd40b7d3 DL |
2651 | if (msglen > skb->len) |
2652 | msglen = skb->len; | |
2653 | skb_pull(skb, msglen); | |
82ace47a TG |
2654 | } |
2655 | ||
2656 | return 0; | |
2657 | } | |
6ac552fd | 2658 | EXPORT_SYMBOL(netlink_rcv_skb); |
82ace47a | 2659 | |
d387f6ad TG |
2660 | /** |
2661 | * nlmsg_notify - send a notification netlink message | |
2662 | * @sk: netlink socket to use | |
2663 | * @skb: notification message | |
15e47304 | 2664 | * @portid: destination netlink portid for reports or 0 |
d387f6ad TG |
2665 | * @group: destination multicast group or 0 |
2666 | * @report: 1 to report back, 0 to disable | |
2667 | * @flags: allocation flags | |
2668 | */ | |
15e47304 | 2669 | int nlmsg_notify(struct sock *sk, struct sk_buff *skb, u32 portid, |
d387f6ad TG |
2670 | unsigned int group, int report, gfp_t flags) |
2671 | { | |
2672 | int err = 0; | |
2673 | ||
2674 | if (group) { | |
15e47304 | 2675 | int exclude_portid = 0; |
d387f6ad TG |
2676 | |
2677 | if (report) { | |
2678 | atomic_inc(&skb->users); | |
15e47304 | 2679 | exclude_portid = portid; |
d387f6ad TG |
2680 | } |
2681 | ||
1ce85fe4 PNA |
2682 | /* errors reported via destination sk->sk_err, but propagate |
2683 | * delivery errors if NETLINK_BROADCAST_ERROR flag is set */ | |
15e47304 | 2684 | err = nlmsg_multicast(sk, skb, exclude_portid, group, flags); |
d387f6ad TG |
2685 | } |
2686 | ||
1ce85fe4 PNA |
2687 | if (report) { |
2688 | int err2; | |
2689 | ||
15e47304 | 2690 | err2 = nlmsg_unicast(sk, skb, portid); |
1ce85fe4 PNA |
2691 | if (!err || err == -ESRCH) |
2692 | err = err2; | |
2693 | } | |
d387f6ad TG |
2694 | |
2695 | return err; | |
2696 | } | |
6ac552fd | 2697 | EXPORT_SYMBOL(nlmsg_notify); |
d387f6ad | 2698 | |
1da177e4 LT |
2699 | #ifdef CONFIG_PROC_FS |
2700 | struct nl_seq_iter { | |
e372c414 | 2701 | struct seq_net_private p; |
1da177e4 LT |
2702 | int link; |
2703 | int hash_idx; | |
2704 | }; | |
2705 | ||
2706 | static struct sock *netlink_seq_socket_idx(struct seq_file *seq, loff_t pos) | |
2707 | { | |
2708 | struct nl_seq_iter *iter = seq->private; | |
2709 | int i, j; | |
2710 | struct sock *s; | |
1da177e4 LT |
2711 | loff_t off = 0; |
2712 | ||
6ac552fd | 2713 | for (i = 0; i < MAX_LINKS; i++) { |
15e47304 | 2714 | struct nl_portid_hash *hash = &nl_table[i].hash; |
1da177e4 LT |
2715 | |
2716 | for (j = 0; j <= hash->mask; j++) { | |
b67bfe0d | 2717 | sk_for_each(s, &hash->table[j]) { |
1218854a | 2718 | if (sock_net(s) != seq_file_net(seq)) |
b4b51029 | 2719 | continue; |
1da177e4 LT |
2720 | if (off == pos) { |
2721 | iter->link = i; | |
2722 | iter->hash_idx = j; | |
2723 | return s; | |
2724 | } | |
2725 | ++off; | |
2726 | } | |
2727 | } | |
2728 | } | |
2729 | return NULL; | |
2730 | } | |
2731 | ||
2732 | static void *netlink_seq_start(struct seq_file *seq, loff_t *pos) | |
9a429c49 | 2733 | __acquires(nl_table_lock) |
1da177e4 LT |
2734 | { |
2735 | read_lock(&nl_table_lock); | |
2736 | return *pos ? netlink_seq_socket_idx(seq, *pos - 1) : SEQ_START_TOKEN; | |
2737 | } | |
2738 | ||
2739 | static void *netlink_seq_next(struct seq_file *seq, void *v, loff_t *pos) | |
2740 | { | |
2741 | struct sock *s; | |
2742 | struct nl_seq_iter *iter; | |
2743 | int i, j; | |
2744 | ||
2745 | ++*pos; | |
2746 | ||
2747 | if (v == SEQ_START_TOKEN) | |
2748 | return netlink_seq_socket_idx(seq, 0); | |
746fac4d | 2749 | |
b4b51029 EB |
2750 | iter = seq->private; |
2751 | s = v; | |
2752 | do { | |
2753 | s = sk_next(s); | |
1218854a | 2754 | } while (s && sock_net(s) != seq_file_net(seq)); |
1da177e4 LT |
2755 | if (s) |
2756 | return s; | |
2757 | ||
1da177e4 LT |
2758 | i = iter->link; |
2759 | j = iter->hash_idx + 1; | |
2760 | ||
2761 | do { | |
15e47304 | 2762 | struct nl_portid_hash *hash = &nl_table[i].hash; |
1da177e4 LT |
2763 | |
2764 | for (; j <= hash->mask; j++) { | |
2765 | s = sk_head(&hash->table[j]); | |
1218854a | 2766 | while (s && sock_net(s) != seq_file_net(seq)) |
b4b51029 | 2767 | s = sk_next(s); |
1da177e4 LT |
2768 | if (s) { |
2769 | iter->link = i; | |
2770 | iter->hash_idx = j; | |
2771 | return s; | |
2772 | } | |
2773 | } | |
2774 | ||
2775 | j = 0; | |
2776 | } while (++i < MAX_LINKS); | |
2777 | ||
2778 | return NULL; | |
2779 | } | |
2780 | ||
2781 | static void netlink_seq_stop(struct seq_file *seq, void *v) | |
9a429c49 | 2782 | __releases(nl_table_lock) |
1da177e4 LT |
2783 | { |
2784 | read_unlock(&nl_table_lock); | |
2785 | } | |
2786 | ||
2787 | ||
2788 | static int netlink_seq_show(struct seq_file *seq, void *v) | |
2789 | { | |
658cb354 | 2790 | if (v == SEQ_START_TOKEN) { |
1da177e4 LT |
2791 | seq_puts(seq, |
2792 | "sk Eth Pid Groups " | |
cf0aa4e0 | 2793 | "Rmem Wmem Dump Locks Drops Inode\n"); |
658cb354 | 2794 | } else { |
1da177e4 LT |
2795 | struct sock *s = v; |
2796 | struct netlink_sock *nlk = nlk_sk(s); | |
2797 | ||
9f1e0ad0 | 2798 | seq_printf(seq, "%pK %-3d %-6u %08x %-8d %-8d %pK %-8d %-8d %-8lu\n", |
1da177e4 LT |
2799 | s, |
2800 | s->sk_protocol, | |
15e47304 | 2801 | nlk->portid, |
513c2500 | 2802 | nlk->groups ? (u32)nlk->groups[0] : 0, |
31e6d363 ED |
2803 | sk_rmem_alloc_get(s), |
2804 | sk_wmem_alloc_get(s), | |
1da177e4 | 2805 | nlk->cb, |
38938bfe | 2806 | atomic_read(&s->sk_refcnt), |
cf0aa4e0 MY |
2807 | atomic_read(&s->sk_drops), |
2808 | sock_i_ino(s) | |
1da177e4 LT |
2809 | ); |
2810 | ||
2811 | } | |
2812 | return 0; | |
2813 | } | |
2814 | ||
56b3d975 | 2815 | static const struct seq_operations netlink_seq_ops = { |
1da177e4 LT |
2816 | .start = netlink_seq_start, |
2817 | .next = netlink_seq_next, | |
2818 | .stop = netlink_seq_stop, | |
2819 | .show = netlink_seq_show, | |
2820 | }; | |
2821 | ||
2822 | ||
2823 | static int netlink_seq_open(struct inode *inode, struct file *file) | |
2824 | { | |
e372c414 DL |
2825 | return seq_open_net(inode, file, &netlink_seq_ops, |
2826 | sizeof(struct nl_seq_iter)); | |
b4b51029 EB |
2827 | } |
2828 | ||
da7071d7 | 2829 | static const struct file_operations netlink_seq_fops = { |
1da177e4 LT |
2830 | .owner = THIS_MODULE, |
2831 | .open = netlink_seq_open, | |
2832 | .read = seq_read, | |
2833 | .llseek = seq_lseek, | |
e372c414 | 2834 | .release = seq_release_net, |
1da177e4 LT |
2835 | }; |
2836 | ||
2837 | #endif | |
2838 | ||
2839 | int netlink_register_notifier(struct notifier_block *nb) | |
2840 | { | |
e041c683 | 2841 | return atomic_notifier_chain_register(&netlink_chain, nb); |
1da177e4 | 2842 | } |
6ac552fd | 2843 | EXPORT_SYMBOL(netlink_register_notifier); |
1da177e4 LT |
2844 | |
2845 | int netlink_unregister_notifier(struct notifier_block *nb) | |
2846 | { | |
e041c683 | 2847 | return atomic_notifier_chain_unregister(&netlink_chain, nb); |
1da177e4 | 2848 | } |
6ac552fd | 2849 | EXPORT_SYMBOL(netlink_unregister_notifier); |
746fac4d | 2850 | |
90ddc4f0 | 2851 | static const struct proto_ops netlink_ops = { |
1da177e4 LT |
2852 | .family = PF_NETLINK, |
2853 | .owner = THIS_MODULE, | |
2854 | .release = netlink_release, | |
2855 | .bind = netlink_bind, | |
2856 | .connect = netlink_connect, | |
2857 | .socketpair = sock_no_socketpair, | |
2858 | .accept = sock_no_accept, | |
2859 | .getname = netlink_getname, | |
9652e931 | 2860 | .poll = netlink_poll, |
1da177e4 LT |
2861 | .ioctl = sock_no_ioctl, |
2862 | .listen = sock_no_listen, | |
2863 | .shutdown = sock_no_shutdown, | |
9a4595bc PM |
2864 | .setsockopt = netlink_setsockopt, |
2865 | .getsockopt = netlink_getsockopt, | |
1da177e4 LT |
2866 | .sendmsg = netlink_sendmsg, |
2867 | .recvmsg = netlink_recvmsg, | |
ccdfcc39 | 2868 | .mmap = netlink_mmap, |
1da177e4 LT |
2869 | .sendpage = sock_no_sendpage, |
2870 | }; | |
2871 | ||
ec1b4cf7 | 2872 | static const struct net_proto_family netlink_family_ops = { |
1da177e4 LT |
2873 | .family = PF_NETLINK, |
2874 | .create = netlink_create, | |
2875 | .owner = THIS_MODULE, /* for consistency 8) */ | |
2876 | }; | |
2877 | ||
4665079c | 2878 | static int __net_init netlink_net_init(struct net *net) |
b4b51029 EB |
2879 | { |
2880 | #ifdef CONFIG_PROC_FS | |
d4beaa66 | 2881 | if (!proc_create("netlink", 0, net->proc_net, &netlink_seq_fops)) |
b4b51029 EB |
2882 | return -ENOMEM; |
2883 | #endif | |
2884 | return 0; | |
2885 | } | |
2886 | ||
4665079c | 2887 | static void __net_exit netlink_net_exit(struct net *net) |
b4b51029 EB |
2888 | { |
2889 | #ifdef CONFIG_PROC_FS | |
ece31ffd | 2890 | remove_proc_entry("netlink", net->proc_net); |
b4b51029 EB |
2891 | #endif |
2892 | } | |
2893 | ||
b963ea89 DM |
2894 | static void __init netlink_add_usersock_entry(void) |
2895 | { | |
5c398dc8 | 2896 | struct listeners *listeners; |
b963ea89 DM |
2897 | int groups = 32; |
2898 | ||
5c398dc8 | 2899 | listeners = kzalloc(sizeof(*listeners) + NLGRPSZ(groups), GFP_KERNEL); |
b963ea89 | 2900 | if (!listeners) |
5c398dc8 | 2901 | panic("netlink_add_usersock_entry: Cannot allocate listeners\n"); |
b963ea89 DM |
2902 | |
2903 | netlink_table_grab(); | |
2904 | ||
2905 | nl_table[NETLINK_USERSOCK].groups = groups; | |
5c398dc8 | 2906 | rcu_assign_pointer(nl_table[NETLINK_USERSOCK].listeners, listeners); |
b963ea89 DM |
2907 | nl_table[NETLINK_USERSOCK].module = THIS_MODULE; |
2908 | nl_table[NETLINK_USERSOCK].registered = 1; | |
9785e10a | 2909 | nl_table[NETLINK_USERSOCK].flags = NL_CFG_F_NONROOT_SEND; |
b963ea89 DM |
2910 | |
2911 | netlink_table_ungrab(); | |
2912 | } | |
2913 | ||
022cbae6 | 2914 | static struct pernet_operations __net_initdata netlink_net_ops = { |
b4b51029 EB |
2915 | .init = netlink_net_init, |
2916 | .exit = netlink_net_exit, | |
2917 | }; | |
2918 | ||
1da177e4 LT |
2919 | static int __init netlink_proto_init(void) |
2920 | { | |
1da177e4 | 2921 | int i; |
26ff5ddc | 2922 | unsigned long limit; |
1da177e4 LT |
2923 | unsigned int order; |
2924 | int err = proto_register(&netlink_proto, 0); | |
2925 | ||
2926 | if (err != 0) | |
2927 | goto out; | |
2928 | ||
fab25745 | 2929 | BUILD_BUG_ON(sizeof(struct netlink_skb_parms) > FIELD_SIZEOF(struct sk_buff, cb)); |
1da177e4 | 2930 | |
0da974f4 | 2931 | nl_table = kcalloc(MAX_LINKS, sizeof(*nl_table), GFP_KERNEL); |
fab2caf6 AM |
2932 | if (!nl_table) |
2933 | goto panic; | |
1da177e4 | 2934 | |
4481374c JB |
2935 | if (totalram_pages >= (128 * 1024)) |
2936 | limit = totalram_pages >> (21 - PAGE_SHIFT); | |
1da177e4 | 2937 | else |
4481374c | 2938 | limit = totalram_pages >> (23 - PAGE_SHIFT); |
1da177e4 | 2939 | |
26ff5ddc DC |
2940 | order = get_bitmask_order(limit) - 1 + PAGE_SHIFT; |
2941 | limit = (1UL << order) / sizeof(struct hlist_head); | |
2942 | order = get_bitmask_order(min(limit, (unsigned long)UINT_MAX)) - 1; | |
1da177e4 LT |
2943 | |
2944 | for (i = 0; i < MAX_LINKS; i++) { | |
15e47304 | 2945 | struct nl_portid_hash *hash = &nl_table[i].hash; |
1da177e4 | 2946 | |
15e47304 | 2947 | hash->table = nl_portid_hash_zalloc(1 * sizeof(*hash->table)); |
1da177e4 LT |
2948 | if (!hash->table) { |
2949 | while (i-- > 0) | |
15e47304 | 2950 | nl_portid_hash_free(nl_table[i].hash.table, |
1da177e4 LT |
2951 | 1 * sizeof(*hash->table)); |
2952 | kfree(nl_table); | |
fab2caf6 | 2953 | goto panic; |
1da177e4 | 2954 | } |
1da177e4 LT |
2955 | hash->max_shift = order; |
2956 | hash->shift = 0; | |
2957 | hash->mask = 0; | |
2958 | hash->rehash_time = jiffies; | |
2959 | } | |
2960 | ||
b963ea89 DM |
2961 | netlink_add_usersock_entry(); |
2962 | ||
1da177e4 | 2963 | sock_register(&netlink_family_ops); |
b4b51029 | 2964 | register_pernet_subsys(&netlink_net_ops); |
746fac4d | 2965 | /* The netlink device handler may be needed early. */ |
1da177e4 LT |
2966 | rtnetlink_init(); |
2967 | out: | |
2968 | return err; | |
fab2caf6 AM |
2969 | panic: |
2970 | panic("netlink_init: Cannot allocate nl_table\n"); | |
1da177e4 LT |
2971 | } |
2972 | ||
1da177e4 | 2973 | core_initcall(netlink_proto_init); |