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); | |
7cdbac71 | 374 | return err; |
ccdfcc39 | 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 | |
753 | if (skb->sk != NULL) | |
754 | sock_rfree(skb); | |
cf0a018a PM |
755 | } |
756 | ||
757 | static void netlink_skb_set_owner_r(struct sk_buff *skb, struct sock *sk) | |
758 | { | |
759 | WARN_ON(skb->sk != NULL); | |
760 | skb->sk = sk; | |
761 | skb->destructor = netlink_skb_destructor; | |
762 | atomic_add(skb->truesize, &sk->sk_rmem_alloc); | |
763 | sk_mem_charge(sk, skb->truesize); | |
764 | } | |
765 | ||
1da177e4 LT |
766 | static void netlink_sock_destruct(struct sock *sk) |
767 | { | |
3f660d66 HX |
768 | struct netlink_sock *nlk = nlk_sk(sk); |
769 | ||
3f660d66 HX |
770 | if (nlk->cb) { |
771 | if (nlk->cb->done) | |
772 | nlk->cb->done(nlk->cb); | |
6dc878a8 G |
773 | |
774 | module_put(nlk->cb->module); | |
3f660d66 HX |
775 | netlink_destroy_callback(nlk->cb); |
776 | } | |
777 | ||
1da177e4 | 778 | skb_queue_purge(&sk->sk_receive_queue); |
ccdfcc39 PM |
779 | #ifdef CONFIG_NETLINK_MMAP |
780 | if (1) { | |
781 | struct nl_mmap_req req; | |
782 | ||
783 | memset(&req, 0, sizeof(req)); | |
784 | if (nlk->rx_ring.pg_vec) | |
785 | netlink_set_ring(sk, &req, true, false); | |
786 | memset(&req, 0, sizeof(req)); | |
787 | if (nlk->tx_ring.pg_vec) | |
788 | netlink_set_ring(sk, &req, true, true); | |
789 | } | |
790 | #endif /* CONFIG_NETLINK_MMAP */ | |
1da177e4 LT |
791 | |
792 | if (!sock_flag(sk, SOCK_DEAD)) { | |
6ac552fd | 793 | printk(KERN_ERR "Freeing alive netlink socket %p\n", sk); |
1da177e4 LT |
794 | return; |
795 | } | |
547b792c IJ |
796 | |
797 | WARN_ON(atomic_read(&sk->sk_rmem_alloc)); | |
798 | WARN_ON(atomic_read(&sk->sk_wmem_alloc)); | |
799 | WARN_ON(nlk_sk(sk)->groups); | |
1da177e4 LT |
800 | } |
801 | ||
6ac552fd PM |
802 | /* This lock without WQ_FLAG_EXCLUSIVE is good on UP and it is _very_ bad on |
803 | * SMP. Look, when several writers sleep and reader wakes them up, all but one | |
1da177e4 LT |
804 | * immediately hit write lock and grab all the cpus. Exclusive sleep solves |
805 | * this, _but_ remember, it adds useless work on UP machines. | |
806 | */ | |
807 | ||
d136f1bd | 808 | void netlink_table_grab(void) |
9a429c49 | 809 | __acquires(nl_table_lock) |
1da177e4 | 810 | { |
d136f1bd JB |
811 | might_sleep(); |
812 | ||
6abd219c | 813 | write_lock_irq(&nl_table_lock); |
1da177e4 LT |
814 | |
815 | if (atomic_read(&nl_table_users)) { | |
816 | DECLARE_WAITQUEUE(wait, current); | |
817 | ||
818 | add_wait_queue_exclusive(&nl_table_wait, &wait); | |
6ac552fd | 819 | for (;;) { |
1da177e4 LT |
820 | set_current_state(TASK_UNINTERRUPTIBLE); |
821 | if (atomic_read(&nl_table_users) == 0) | |
822 | break; | |
6abd219c | 823 | write_unlock_irq(&nl_table_lock); |
1da177e4 | 824 | schedule(); |
6abd219c | 825 | write_lock_irq(&nl_table_lock); |
1da177e4 LT |
826 | } |
827 | ||
828 | __set_current_state(TASK_RUNNING); | |
829 | remove_wait_queue(&nl_table_wait, &wait); | |
830 | } | |
831 | } | |
832 | ||
d136f1bd | 833 | void netlink_table_ungrab(void) |
9a429c49 | 834 | __releases(nl_table_lock) |
1da177e4 | 835 | { |
6abd219c | 836 | write_unlock_irq(&nl_table_lock); |
1da177e4 LT |
837 | wake_up(&nl_table_wait); |
838 | } | |
839 | ||
6ac552fd | 840 | static inline void |
1da177e4 LT |
841 | netlink_lock_table(void) |
842 | { | |
843 | /* read_lock() synchronizes us to netlink_table_grab */ | |
844 | ||
845 | read_lock(&nl_table_lock); | |
846 | atomic_inc(&nl_table_users); | |
847 | read_unlock(&nl_table_lock); | |
848 | } | |
849 | ||
6ac552fd | 850 | static inline void |
1da177e4 LT |
851 | netlink_unlock_table(void) |
852 | { | |
853 | if (atomic_dec_and_test(&nl_table_users)) | |
854 | wake_up(&nl_table_wait); | |
855 | } | |
856 | ||
15e47304 | 857 | static struct sock *netlink_lookup(struct net *net, int protocol, u32 portid) |
1da177e4 | 858 | { |
15e47304 | 859 | struct nl_portid_hash *hash = &nl_table[protocol].hash; |
1da177e4 LT |
860 | struct hlist_head *head; |
861 | struct sock *sk; | |
1da177e4 LT |
862 | |
863 | read_lock(&nl_table_lock); | |
15e47304 | 864 | head = nl_portid_hashfn(hash, portid); |
b67bfe0d | 865 | sk_for_each(sk, head) { |
15e47304 | 866 | if (net_eq(sock_net(sk), net) && (nlk_sk(sk)->portid == portid)) { |
1da177e4 LT |
867 | sock_hold(sk); |
868 | goto found; | |
869 | } | |
870 | } | |
871 | sk = NULL; | |
872 | found: | |
873 | read_unlock(&nl_table_lock); | |
874 | return sk; | |
875 | } | |
876 | ||
15e47304 | 877 | static struct hlist_head *nl_portid_hash_zalloc(size_t size) |
1da177e4 LT |
878 | { |
879 | if (size <= PAGE_SIZE) | |
ea72912c | 880 | return kzalloc(size, GFP_ATOMIC); |
1da177e4 LT |
881 | else |
882 | return (struct hlist_head *) | |
ea72912c ED |
883 | __get_free_pages(GFP_ATOMIC | __GFP_ZERO, |
884 | get_order(size)); | |
1da177e4 LT |
885 | } |
886 | ||
15e47304 | 887 | static void nl_portid_hash_free(struct hlist_head *table, size_t size) |
1da177e4 LT |
888 | { |
889 | if (size <= PAGE_SIZE) | |
890 | kfree(table); | |
891 | else | |
892 | free_pages((unsigned long)table, get_order(size)); | |
893 | } | |
894 | ||
15e47304 | 895 | static int nl_portid_hash_rehash(struct nl_portid_hash *hash, int grow) |
1da177e4 LT |
896 | { |
897 | unsigned int omask, mask, shift; | |
898 | size_t osize, size; | |
899 | struct hlist_head *otable, *table; | |
900 | int i; | |
901 | ||
902 | omask = mask = hash->mask; | |
903 | osize = size = (mask + 1) * sizeof(*table); | |
904 | shift = hash->shift; | |
905 | ||
906 | if (grow) { | |
907 | if (++shift > hash->max_shift) | |
908 | return 0; | |
909 | mask = mask * 2 + 1; | |
910 | size *= 2; | |
911 | } | |
912 | ||
15e47304 | 913 | table = nl_portid_hash_zalloc(size); |
1da177e4 LT |
914 | if (!table) |
915 | return 0; | |
916 | ||
1da177e4 LT |
917 | otable = hash->table; |
918 | hash->table = table; | |
919 | hash->mask = mask; | |
920 | hash->shift = shift; | |
921 | get_random_bytes(&hash->rnd, sizeof(hash->rnd)); | |
922 | ||
923 | for (i = 0; i <= omask; i++) { | |
924 | struct sock *sk; | |
b67bfe0d | 925 | struct hlist_node *tmp; |
1da177e4 | 926 | |
b67bfe0d | 927 | sk_for_each_safe(sk, tmp, &otable[i]) |
15e47304 | 928 | __sk_add_node(sk, nl_portid_hashfn(hash, nlk_sk(sk)->portid)); |
1da177e4 LT |
929 | } |
930 | ||
15e47304 | 931 | nl_portid_hash_free(otable, osize); |
1da177e4 LT |
932 | hash->rehash_time = jiffies + 10 * 60 * HZ; |
933 | return 1; | |
934 | } | |
935 | ||
15e47304 | 936 | static inline int nl_portid_hash_dilute(struct nl_portid_hash *hash, int len) |
1da177e4 LT |
937 | { |
938 | int avg = hash->entries >> hash->shift; | |
939 | ||
15e47304 | 940 | if (unlikely(avg > 1) && nl_portid_hash_rehash(hash, 1)) |
1da177e4 LT |
941 | return 1; |
942 | ||
943 | if (unlikely(len > avg) && time_after(jiffies, hash->rehash_time)) { | |
15e47304 | 944 | nl_portid_hash_rehash(hash, 0); |
1da177e4 LT |
945 | return 1; |
946 | } | |
947 | ||
948 | return 0; | |
949 | } | |
950 | ||
90ddc4f0 | 951 | static const struct proto_ops netlink_ops; |
1da177e4 | 952 | |
4277a083 PM |
953 | static void |
954 | netlink_update_listeners(struct sock *sk) | |
955 | { | |
956 | struct netlink_table *tbl = &nl_table[sk->sk_protocol]; | |
4277a083 PM |
957 | unsigned long mask; |
958 | unsigned int i; | |
6d772ac5 ED |
959 | struct listeners *listeners; |
960 | ||
961 | listeners = nl_deref_protected(tbl->listeners); | |
962 | if (!listeners) | |
963 | return; | |
4277a083 | 964 | |
b4ff4f04 | 965 | for (i = 0; i < NLGRPLONGS(tbl->groups); i++) { |
4277a083 | 966 | mask = 0; |
b67bfe0d | 967 | sk_for_each_bound(sk, &tbl->mc_list) { |
b4ff4f04 JB |
968 | if (i < NLGRPLONGS(nlk_sk(sk)->ngroups)) |
969 | mask |= nlk_sk(sk)->groups[i]; | |
970 | } | |
6d772ac5 | 971 | listeners->masks[i] = mask; |
4277a083 PM |
972 | } |
973 | /* this function is only called with the netlink table "grabbed", which | |
974 | * makes sure updates are visible before bind or setsockopt return. */ | |
975 | } | |
976 | ||
15e47304 | 977 | static int netlink_insert(struct sock *sk, struct net *net, u32 portid) |
1da177e4 | 978 | { |
15e47304 | 979 | struct nl_portid_hash *hash = &nl_table[sk->sk_protocol].hash; |
1da177e4 LT |
980 | struct hlist_head *head; |
981 | int err = -EADDRINUSE; | |
982 | struct sock *osk; | |
1da177e4 LT |
983 | int len; |
984 | ||
985 | netlink_table_grab(); | |
15e47304 | 986 | head = nl_portid_hashfn(hash, portid); |
1da177e4 | 987 | len = 0; |
b67bfe0d | 988 | sk_for_each(osk, head) { |
15e47304 | 989 | if (net_eq(sock_net(osk), net) && (nlk_sk(osk)->portid == portid)) |
1da177e4 LT |
990 | break; |
991 | len++; | |
992 | } | |
b67bfe0d | 993 | if (osk) |
1da177e4 LT |
994 | goto err; |
995 | ||
996 | err = -EBUSY; | |
15e47304 | 997 | if (nlk_sk(sk)->portid) |
1da177e4 LT |
998 | goto err; |
999 | ||
1000 | err = -ENOMEM; | |
1001 | if (BITS_PER_LONG > 32 && unlikely(hash->entries >= UINT_MAX)) | |
1002 | goto err; | |
1003 | ||
15e47304 EB |
1004 | if (len && nl_portid_hash_dilute(hash, len)) |
1005 | head = nl_portid_hashfn(hash, portid); | |
1da177e4 | 1006 | hash->entries++; |
15e47304 | 1007 | nlk_sk(sk)->portid = portid; |
1da177e4 LT |
1008 | sk_add_node(sk, head); |
1009 | err = 0; | |
1010 | ||
1011 | err: | |
1012 | netlink_table_ungrab(); | |
1013 | return err; | |
1014 | } | |
1015 | ||
1016 | static void netlink_remove(struct sock *sk) | |
1017 | { | |
1018 | netlink_table_grab(); | |
d470e3b4 DM |
1019 | if (sk_del_node_init(sk)) |
1020 | nl_table[sk->sk_protocol].hash.entries--; | |
f7fa9b10 | 1021 | if (nlk_sk(sk)->subscriptions) |
1da177e4 LT |
1022 | __sk_del_bind_node(sk); |
1023 | netlink_table_ungrab(); | |
1024 | } | |
1025 | ||
1026 | static struct proto netlink_proto = { | |
1027 | .name = "NETLINK", | |
1028 | .owner = THIS_MODULE, | |
1029 | .obj_size = sizeof(struct netlink_sock), | |
1030 | }; | |
1031 | ||
1b8d7ae4 EB |
1032 | static int __netlink_create(struct net *net, struct socket *sock, |
1033 | struct mutex *cb_mutex, int protocol) | |
1da177e4 LT |
1034 | { |
1035 | struct sock *sk; | |
1036 | struct netlink_sock *nlk; | |
ab33a171 PM |
1037 | |
1038 | sock->ops = &netlink_ops; | |
1039 | ||
6257ff21 | 1040 | sk = sk_alloc(net, PF_NETLINK, GFP_KERNEL, &netlink_proto); |
ab33a171 PM |
1041 | if (!sk) |
1042 | return -ENOMEM; | |
1043 | ||
1044 | sock_init_data(sock, sk); | |
1045 | ||
1046 | nlk = nlk_sk(sk); | |
658cb354 | 1047 | if (cb_mutex) { |
ffa4d721 | 1048 | nlk->cb_mutex = cb_mutex; |
658cb354 | 1049 | } else { |
ffa4d721 PM |
1050 | nlk->cb_mutex = &nlk->cb_def_mutex; |
1051 | mutex_init(nlk->cb_mutex); | |
1052 | } | |
ab33a171 | 1053 | init_waitqueue_head(&nlk->wait); |
ccdfcc39 PM |
1054 | #ifdef CONFIG_NETLINK_MMAP |
1055 | mutex_init(&nlk->pg_vec_lock); | |
1056 | #endif | |
ab33a171 PM |
1057 | |
1058 | sk->sk_destruct = netlink_sock_destruct; | |
1059 | sk->sk_protocol = protocol; | |
1060 | return 0; | |
1061 | } | |
1062 | ||
3f378b68 EP |
1063 | static int netlink_create(struct net *net, struct socket *sock, int protocol, |
1064 | int kern) | |
ab33a171 PM |
1065 | { |
1066 | struct module *module = NULL; | |
af65bdfc | 1067 | struct mutex *cb_mutex; |
f7fa9b10 | 1068 | struct netlink_sock *nlk; |
03292745 | 1069 | void (*bind)(int group); |
ab33a171 | 1070 | int err = 0; |
1da177e4 LT |
1071 | |
1072 | sock->state = SS_UNCONNECTED; | |
1073 | ||
1074 | if (sock->type != SOCK_RAW && sock->type != SOCK_DGRAM) | |
1075 | return -ESOCKTNOSUPPORT; | |
1076 | ||
6ac552fd | 1077 | if (protocol < 0 || protocol >= MAX_LINKS) |
1da177e4 LT |
1078 | return -EPROTONOSUPPORT; |
1079 | ||
77247bbb | 1080 | netlink_lock_table(); |
95a5afca | 1081 | #ifdef CONFIG_MODULES |
ab33a171 | 1082 | if (!nl_table[protocol].registered) { |
77247bbb | 1083 | netlink_unlock_table(); |
4fdb3bb7 | 1084 | request_module("net-pf-%d-proto-%d", PF_NETLINK, protocol); |
77247bbb | 1085 | netlink_lock_table(); |
4fdb3bb7 | 1086 | } |
ab33a171 PM |
1087 | #endif |
1088 | if (nl_table[protocol].registered && | |
1089 | try_module_get(nl_table[protocol].module)) | |
1090 | module = nl_table[protocol].module; | |
974c37e9 AD |
1091 | else |
1092 | err = -EPROTONOSUPPORT; | |
af65bdfc | 1093 | cb_mutex = nl_table[protocol].cb_mutex; |
03292745 | 1094 | bind = nl_table[protocol].bind; |
77247bbb | 1095 | netlink_unlock_table(); |
4fdb3bb7 | 1096 | |
974c37e9 AD |
1097 | if (err < 0) |
1098 | goto out; | |
1099 | ||
6ac552fd PM |
1100 | err = __netlink_create(net, sock, cb_mutex, protocol); |
1101 | if (err < 0) | |
f7fa9b10 PM |
1102 | goto out_module; |
1103 | ||
6f756a8c | 1104 | local_bh_disable(); |
c1fd3b94 | 1105 | sock_prot_inuse_add(net, &netlink_proto, 1); |
6f756a8c DM |
1106 | local_bh_enable(); |
1107 | ||
f7fa9b10 | 1108 | nlk = nlk_sk(sock->sk); |
f7fa9b10 | 1109 | nlk->module = module; |
03292745 | 1110 | nlk->netlink_bind = bind; |
ab33a171 PM |
1111 | out: |
1112 | return err; | |
1da177e4 | 1113 | |
ab33a171 PM |
1114 | out_module: |
1115 | module_put(module); | |
1116 | goto out; | |
1da177e4 LT |
1117 | } |
1118 | ||
1119 | static int netlink_release(struct socket *sock) | |
1120 | { | |
1121 | struct sock *sk = sock->sk; | |
1122 | struct netlink_sock *nlk; | |
1123 | ||
1124 | if (!sk) | |
1125 | return 0; | |
1126 | ||
1127 | netlink_remove(sk); | |
ac57b3a9 | 1128 | sock_orphan(sk); |
1da177e4 LT |
1129 | nlk = nlk_sk(sk); |
1130 | ||
3f660d66 HX |
1131 | /* |
1132 | * OK. Socket is unlinked, any packets that arrive now | |
1133 | * will be purged. | |
1134 | */ | |
1da177e4 | 1135 | |
1da177e4 LT |
1136 | sock->sk = NULL; |
1137 | wake_up_interruptible_all(&nlk->wait); | |
1138 | ||
1139 | skb_queue_purge(&sk->sk_write_queue); | |
1140 | ||
15e47304 | 1141 | if (nlk->portid) { |
1da177e4 | 1142 | struct netlink_notify n = { |
3b1e0a65 | 1143 | .net = sock_net(sk), |
1da177e4 | 1144 | .protocol = sk->sk_protocol, |
15e47304 | 1145 | .portid = nlk->portid, |
1da177e4 | 1146 | }; |
e041c683 AS |
1147 | atomic_notifier_call_chain(&netlink_chain, |
1148 | NETLINK_URELEASE, &n); | |
746fac4d | 1149 | } |
4fdb3bb7 | 1150 | |
5e7c001c | 1151 | module_put(nlk->module); |
4fdb3bb7 | 1152 | |
4277a083 | 1153 | netlink_table_grab(); |
aed81560 | 1154 | if (netlink_is_kernel(sk)) { |
869e58f8 DL |
1155 | BUG_ON(nl_table[sk->sk_protocol].registered == 0); |
1156 | if (--nl_table[sk->sk_protocol].registered == 0) { | |
6d772ac5 ED |
1157 | struct listeners *old; |
1158 | ||
1159 | old = nl_deref_protected(nl_table[sk->sk_protocol].listeners); | |
1160 | RCU_INIT_POINTER(nl_table[sk->sk_protocol].listeners, NULL); | |
1161 | kfree_rcu(old, rcu); | |
869e58f8 | 1162 | nl_table[sk->sk_protocol].module = NULL; |
9785e10a PNA |
1163 | nl_table[sk->sk_protocol].bind = NULL; |
1164 | nl_table[sk->sk_protocol].flags = 0; | |
869e58f8 DL |
1165 | nl_table[sk->sk_protocol].registered = 0; |
1166 | } | |
658cb354 | 1167 | } else if (nlk->subscriptions) { |
4277a083 | 1168 | netlink_update_listeners(sk); |
658cb354 | 1169 | } |
4277a083 | 1170 | netlink_table_ungrab(); |
77247bbb | 1171 | |
f7fa9b10 PM |
1172 | kfree(nlk->groups); |
1173 | nlk->groups = NULL; | |
1174 | ||
3755810c | 1175 | local_bh_disable(); |
c1fd3b94 | 1176 | sock_prot_inuse_add(sock_net(sk), &netlink_proto, -1); |
3755810c | 1177 | local_bh_enable(); |
1da177e4 LT |
1178 | sock_put(sk); |
1179 | return 0; | |
1180 | } | |
1181 | ||
1182 | static int netlink_autobind(struct socket *sock) | |
1183 | { | |
1184 | struct sock *sk = sock->sk; | |
3b1e0a65 | 1185 | struct net *net = sock_net(sk); |
15e47304 | 1186 | struct nl_portid_hash *hash = &nl_table[sk->sk_protocol].hash; |
1da177e4 LT |
1187 | struct hlist_head *head; |
1188 | struct sock *osk; | |
15e47304 | 1189 | s32 portid = task_tgid_vnr(current); |
1da177e4 LT |
1190 | int err; |
1191 | static s32 rover = -4097; | |
1192 | ||
1193 | retry: | |
1194 | cond_resched(); | |
1195 | netlink_table_grab(); | |
15e47304 | 1196 | head = nl_portid_hashfn(hash, portid); |
b67bfe0d | 1197 | sk_for_each(osk, head) { |
878628fb | 1198 | if (!net_eq(sock_net(osk), net)) |
b4b51029 | 1199 | continue; |
15e47304 EB |
1200 | if (nlk_sk(osk)->portid == portid) { |
1201 | /* Bind collision, search negative portid values. */ | |
1202 | portid = rover--; | |
1da177e4 LT |
1203 | if (rover > -4097) |
1204 | rover = -4097; | |
1205 | netlink_table_ungrab(); | |
1206 | goto retry; | |
1207 | } | |
1208 | } | |
1209 | netlink_table_ungrab(); | |
1210 | ||
15e47304 | 1211 | err = netlink_insert(sk, net, portid); |
1da177e4 LT |
1212 | if (err == -EADDRINUSE) |
1213 | goto retry; | |
d470e3b4 DM |
1214 | |
1215 | /* If 2 threads race to autobind, that is fine. */ | |
1216 | if (err == -EBUSY) | |
1217 | err = 0; | |
1218 | ||
1219 | return err; | |
1da177e4 LT |
1220 | } |
1221 | ||
b57ef81f | 1222 | static inline int netlink_capable(const struct socket *sock, unsigned int flag) |
746fac4d | 1223 | { |
9785e10a | 1224 | return (nl_table[sock->sk->sk_protocol].flags & flag) || |
df008c91 | 1225 | ns_capable(sock_net(sock->sk)->user_ns, CAP_NET_ADMIN); |
746fac4d | 1226 | } |
1da177e4 | 1227 | |
f7fa9b10 PM |
1228 | static void |
1229 | netlink_update_subscriptions(struct sock *sk, unsigned int subscriptions) | |
1230 | { | |
1231 | struct netlink_sock *nlk = nlk_sk(sk); | |
1232 | ||
1233 | if (nlk->subscriptions && !subscriptions) | |
1234 | __sk_del_bind_node(sk); | |
1235 | else if (!nlk->subscriptions && subscriptions) | |
1236 | sk_add_bind_node(sk, &nl_table[sk->sk_protocol].mc_list); | |
1237 | nlk->subscriptions = subscriptions; | |
1238 | } | |
1239 | ||
b4ff4f04 | 1240 | static int netlink_realloc_groups(struct sock *sk) |
513c2500 PM |
1241 | { |
1242 | struct netlink_sock *nlk = nlk_sk(sk); | |
1243 | unsigned int groups; | |
b4ff4f04 | 1244 | unsigned long *new_groups; |
513c2500 PM |
1245 | int err = 0; |
1246 | ||
b4ff4f04 JB |
1247 | netlink_table_grab(); |
1248 | ||
513c2500 | 1249 | groups = nl_table[sk->sk_protocol].groups; |
b4ff4f04 | 1250 | if (!nl_table[sk->sk_protocol].registered) { |
513c2500 | 1251 | err = -ENOENT; |
b4ff4f04 JB |
1252 | goto out_unlock; |
1253 | } | |
513c2500 | 1254 | |
b4ff4f04 JB |
1255 | if (nlk->ngroups >= groups) |
1256 | goto out_unlock; | |
513c2500 | 1257 | |
b4ff4f04 JB |
1258 | new_groups = krealloc(nlk->groups, NLGRPSZ(groups), GFP_ATOMIC); |
1259 | if (new_groups == NULL) { | |
1260 | err = -ENOMEM; | |
1261 | goto out_unlock; | |
1262 | } | |
6ac552fd | 1263 | memset((char *)new_groups + NLGRPSZ(nlk->ngroups), 0, |
b4ff4f04 JB |
1264 | NLGRPSZ(groups) - NLGRPSZ(nlk->ngroups)); |
1265 | ||
1266 | nlk->groups = new_groups; | |
513c2500 | 1267 | nlk->ngroups = groups; |
b4ff4f04 JB |
1268 | out_unlock: |
1269 | netlink_table_ungrab(); | |
1270 | return err; | |
513c2500 PM |
1271 | } |
1272 | ||
6ac552fd PM |
1273 | static int netlink_bind(struct socket *sock, struct sockaddr *addr, |
1274 | int addr_len) | |
1da177e4 LT |
1275 | { |
1276 | struct sock *sk = sock->sk; | |
3b1e0a65 | 1277 | struct net *net = sock_net(sk); |
1da177e4 LT |
1278 | struct netlink_sock *nlk = nlk_sk(sk); |
1279 | struct sockaddr_nl *nladdr = (struct sockaddr_nl *)addr; | |
1280 | int err; | |
746fac4d | 1281 | |
4e4b5376 HFS |
1282 | if (addr_len < sizeof(struct sockaddr_nl)) |
1283 | return -EINVAL; | |
1284 | ||
1da177e4 LT |
1285 | if (nladdr->nl_family != AF_NETLINK) |
1286 | return -EINVAL; | |
1287 | ||
1288 | /* Only superuser is allowed to listen multicasts */ | |
513c2500 | 1289 | if (nladdr->nl_groups) { |
9785e10a | 1290 | if (!netlink_capable(sock, NL_CFG_F_NONROOT_RECV)) |
513c2500 | 1291 | return -EPERM; |
b4ff4f04 JB |
1292 | err = netlink_realloc_groups(sk); |
1293 | if (err) | |
1294 | return err; | |
513c2500 | 1295 | } |
1da177e4 | 1296 | |
15e47304 EB |
1297 | if (nlk->portid) { |
1298 | if (nladdr->nl_pid != nlk->portid) | |
1da177e4 LT |
1299 | return -EINVAL; |
1300 | } else { | |
1301 | err = nladdr->nl_pid ? | |
b4b51029 | 1302 | netlink_insert(sk, net, nladdr->nl_pid) : |
1da177e4 LT |
1303 | netlink_autobind(sock); |
1304 | if (err) | |
1305 | return err; | |
1306 | } | |
1307 | ||
513c2500 | 1308 | if (!nladdr->nl_groups && (nlk->groups == NULL || !(u32)nlk->groups[0])) |
1da177e4 LT |
1309 | return 0; |
1310 | ||
1311 | netlink_table_grab(); | |
f7fa9b10 | 1312 | netlink_update_subscriptions(sk, nlk->subscriptions + |
746fac4d YH |
1313 | hweight32(nladdr->nl_groups) - |
1314 | hweight32(nlk->groups[0])); | |
1315 | nlk->groups[0] = (nlk->groups[0] & ~0xffffffffUL) | nladdr->nl_groups; | |
4277a083 | 1316 | netlink_update_listeners(sk); |
1da177e4 LT |
1317 | netlink_table_ungrab(); |
1318 | ||
03292745 PNA |
1319 | if (nlk->netlink_bind && nlk->groups[0]) { |
1320 | int i; | |
1321 | ||
1322 | for (i=0; i<nlk->ngroups; i++) { | |
1323 | if (test_bit(i, nlk->groups)) | |
1324 | nlk->netlink_bind(i); | |
1325 | } | |
1326 | } | |
1327 | ||
1da177e4 LT |
1328 | return 0; |
1329 | } | |
1330 | ||
1331 | static int netlink_connect(struct socket *sock, struct sockaddr *addr, | |
1332 | int alen, int flags) | |
1333 | { | |
1334 | int err = 0; | |
1335 | struct sock *sk = sock->sk; | |
1336 | struct netlink_sock *nlk = nlk_sk(sk); | |
6ac552fd | 1337 | struct sockaddr_nl *nladdr = (struct sockaddr_nl *)addr; |
1da177e4 | 1338 | |
6503d961 CG |
1339 | if (alen < sizeof(addr->sa_family)) |
1340 | return -EINVAL; | |
1341 | ||
1da177e4 LT |
1342 | if (addr->sa_family == AF_UNSPEC) { |
1343 | sk->sk_state = NETLINK_UNCONNECTED; | |
15e47304 | 1344 | nlk->dst_portid = 0; |
d629b836 | 1345 | nlk->dst_group = 0; |
1da177e4 LT |
1346 | return 0; |
1347 | } | |
1348 | if (addr->sa_family != AF_NETLINK) | |
1349 | return -EINVAL; | |
1350 | ||
1351 | /* Only superuser is allowed to send multicasts */ | |
9785e10a | 1352 | if (nladdr->nl_groups && !netlink_capable(sock, NL_CFG_F_NONROOT_SEND)) |
1da177e4 LT |
1353 | return -EPERM; |
1354 | ||
15e47304 | 1355 | if (!nlk->portid) |
1da177e4 LT |
1356 | err = netlink_autobind(sock); |
1357 | ||
1358 | if (err == 0) { | |
1359 | sk->sk_state = NETLINK_CONNECTED; | |
15e47304 | 1360 | nlk->dst_portid = nladdr->nl_pid; |
d629b836 | 1361 | nlk->dst_group = ffs(nladdr->nl_groups); |
1da177e4 LT |
1362 | } |
1363 | ||
1364 | return err; | |
1365 | } | |
1366 | ||
6ac552fd PM |
1367 | static int netlink_getname(struct socket *sock, struct sockaddr *addr, |
1368 | int *addr_len, int peer) | |
1da177e4 LT |
1369 | { |
1370 | struct sock *sk = sock->sk; | |
1371 | struct netlink_sock *nlk = nlk_sk(sk); | |
13cfa97b | 1372 | DECLARE_SOCKADDR(struct sockaddr_nl *, nladdr, addr); |
746fac4d | 1373 | |
1da177e4 LT |
1374 | nladdr->nl_family = AF_NETLINK; |
1375 | nladdr->nl_pad = 0; | |
1376 | *addr_len = sizeof(*nladdr); | |
1377 | ||
1378 | if (peer) { | |
15e47304 | 1379 | nladdr->nl_pid = nlk->dst_portid; |
d629b836 | 1380 | nladdr->nl_groups = netlink_group_mask(nlk->dst_group); |
1da177e4 | 1381 | } else { |
15e47304 | 1382 | nladdr->nl_pid = nlk->portid; |
513c2500 | 1383 | nladdr->nl_groups = nlk->groups ? nlk->groups[0] : 0; |
1da177e4 LT |
1384 | } |
1385 | return 0; | |
1386 | } | |
1387 | ||
15e47304 | 1388 | static struct sock *netlink_getsockbyportid(struct sock *ssk, u32 portid) |
1da177e4 | 1389 | { |
1da177e4 LT |
1390 | struct sock *sock; |
1391 | struct netlink_sock *nlk; | |
1392 | ||
15e47304 | 1393 | sock = netlink_lookup(sock_net(ssk), ssk->sk_protocol, portid); |
1da177e4 LT |
1394 | if (!sock) |
1395 | return ERR_PTR(-ECONNREFUSED); | |
1396 | ||
1397 | /* Don't bother queuing skb if kernel socket has no input function */ | |
1398 | nlk = nlk_sk(sock); | |
cd40b7d3 | 1399 | if (sock->sk_state == NETLINK_CONNECTED && |
15e47304 | 1400 | nlk->dst_portid != nlk_sk(ssk)->portid) { |
1da177e4 LT |
1401 | sock_put(sock); |
1402 | return ERR_PTR(-ECONNREFUSED); | |
1403 | } | |
1404 | return sock; | |
1405 | } | |
1406 | ||
1407 | struct sock *netlink_getsockbyfilp(struct file *filp) | |
1408 | { | |
496ad9aa | 1409 | struct inode *inode = file_inode(filp); |
1da177e4 LT |
1410 | struct sock *sock; |
1411 | ||
1412 | if (!S_ISSOCK(inode->i_mode)) | |
1413 | return ERR_PTR(-ENOTSOCK); | |
1414 | ||
1415 | sock = SOCKET_I(inode)->sk; | |
1416 | if (sock->sk_family != AF_NETLINK) | |
1417 | return ERR_PTR(-EINVAL); | |
1418 | ||
1419 | sock_hold(sock); | |
1420 | return sock; | |
1421 | } | |
1422 | ||
1423 | /* | |
1424 | * Attach a skb to a netlink socket. | |
1425 | * The caller must hold a reference to the destination socket. On error, the | |
1426 | * reference is dropped. The skb is not send to the destination, just all | |
1427 | * all error checks are performed and memory in the queue is reserved. | |
1428 | * Return values: | |
1429 | * < 0: error. skb freed, reference to sock dropped. | |
1430 | * 0: continue | |
1431 | * 1: repeat lookup - reference dropped while waiting for socket memory. | |
1432 | */ | |
9457afee | 1433 | int netlink_attachskb(struct sock *sk, struct sk_buff *skb, |
c3d8d1e3 | 1434 | long *timeo, struct sock *ssk) |
1da177e4 LT |
1435 | { |
1436 | struct netlink_sock *nlk; | |
1437 | ||
1438 | nlk = nlk_sk(sk); | |
1439 | ||
5fd96123 PM |
1440 | if ((atomic_read(&sk->sk_rmem_alloc) > sk->sk_rcvbuf || |
1441 | test_bit(NETLINK_CONGESTED, &nlk->state)) && | |
1442 | !netlink_skb_is_mmaped(skb)) { | |
1da177e4 | 1443 | DECLARE_WAITQUEUE(wait, current); |
c3d8d1e3 | 1444 | if (!*timeo) { |
aed81560 | 1445 | if (!ssk || netlink_is_kernel(ssk)) |
1da177e4 LT |
1446 | netlink_overrun(sk); |
1447 | sock_put(sk); | |
1448 | kfree_skb(skb); | |
1449 | return -EAGAIN; | |
1450 | } | |
1451 | ||
1452 | __set_current_state(TASK_INTERRUPTIBLE); | |
1453 | add_wait_queue(&nlk->wait, &wait); | |
1454 | ||
1455 | if ((atomic_read(&sk->sk_rmem_alloc) > sk->sk_rcvbuf || | |
cd967e05 | 1456 | test_bit(NETLINK_CONGESTED, &nlk->state)) && |
1da177e4 | 1457 | !sock_flag(sk, SOCK_DEAD)) |
c3d8d1e3 | 1458 | *timeo = schedule_timeout(*timeo); |
1da177e4 LT |
1459 | |
1460 | __set_current_state(TASK_RUNNING); | |
1461 | remove_wait_queue(&nlk->wait, &wait); | |
1462 | sock_put(sk); | |
1463 | ||
1464 | if (signal_pending(current)) { | |
1465 | kfree_skb(skb); | |
c3d8d1e3 | 1466 | return sock_intr_errno(*timeo); |
1da177e4 LT |
1467 | } |
1468 | return 1; | |
1469 | } | |
cf0a018a | 1470 | netlink_skb_set_owner_r(skb, sk); |
1da177e4 LT |
1471 | return 0; |
1472 | } | |
1473 | ||
4a7e7c2a | 1474 | static int __netlink_sendskb(struct sock *sk, struct sk_buff *skb) |
1da177e4 | 1475 | { |
1da177e4 LT |
1476 | int len = skb->len; |
1477 | ||
f9c22888 PM |
1478 | #ifdef CONFIG_NETLINK_MMAP |
1479 | if (netlink_skb_is_mmaped(skb)) | |
1480 | netlink_queue_mmaped_skb(sk, skb); | |
1481 | else if (netlink_rx_is_mmaped(sk)) | |
1482 | netlink_ring_set_copied(sk, skb); | |
1483 | else | |
1484 | #endif /* CONFIG_NETLINK_MMAP */ | |
1485 | skb_queue_tail(&sk->sk_receive_queue, skb); | |
1da177e4 | 1486 | sk->sk_data_ready(sk, len); |
4a7e7c2a ED |
1487 | return len; |
1488 | } | |
1489 | ||
1490 | int netlink_sendskb(struct sock *sk, struct sk_buff *skb) | |
1491 | { | |
1492 | int len = __netlink_sendskb(sk, skb); | |
1493 | ||
1da177e4 LT |
1494 | sock_put(sk); |
1495 | return len; | |
1496 | } | |
1497 | ||
1498 | void netlink_detachskb(struct sock *sk, struct sk_buff *skb) | |
1499 | { | |
1500 | kfree_skb(skb); | |
1501 | sock_put(sk); | |
1502 | } | |
1503 | ||
b57ef81f | 1504 | static struct sk_buff *netlink_trim(struct sk_buff *skb, gfp_t allocation) |
1da177e4 LT |
1505 | { |
1506 | int delta; | |
1507 | ||
1298ca46 | 1508 | WARN_ON(skb->sk != NULL); |
5fd96123 PM |
1509 | if (netlink_skb_is_mmaped(skb)) |
1510 | return skb; | |
1da177e4 | 1511 | |
4305b541 | 1512 | delta = skb->end - skb->tail; |
1da177e4 LT |
1513 | if (delta * 2 < skb->truesize) |
1514 | return skb; | |
1515 | ||
1516 | if (skb_shared(skb)) { | |
1517 | struct sk_buff *nskb = skb_clone(skb, allocation); | |
1518 | if (!nskb) | |
1519 | return skb; | |
8460c00f | 1520 | consume_skb(skb); |
1da177e4 LT |
1521 | skb = nskb; |
1522 | } | |
1523 | ||
1524 | if (!pskb_expand_head(skb, 0, -delta, allocation)) | |
1525 | skb->truesize -= delta; | |
1526 | ||
1527 | return skb; | |
1528 | } | |
1529 | ||
3fbc2905 EB |
1530 | static int netlink_unicast_kernel(struct sock *sk, struct sk_buff *skb, |
1531 | struct sock *ssk) | |
cd40b7d3 DL |
1532 | { |
1533 | int ret; | |
1534 | struct netlink_sock *nlk = nlk_sk(sk); | |
1535 | ||
1536 | ret = -ECONNREFUSED; | |
1537 | if (nlk->netlink_rcv != NULL) { | |
1538 | ret = skb->len; | |
cf0a018a | 1539 | netlink_skb_set_owner_r(skb, sk); |
e32123e5 | 1540 | NETLINK_CB(skb).sk = ssk; |
cd40b7d3 | 1541 | nlk->netlink_rcv(skb); |
bfb253c9 ED |
1542 | consume_skb(skb); |
1543 | } else { | |
1544 | kfree_skb(skb); | |
cd40b7d3 | 1545 | } |
cd40b7d3 DL |
1546 | sock_put(sk); |
1547 | return ret; | |
1548 | } | |
1549 | ||
1550 | int netlink_unicast(struct sock *ssk, struct sk_buff *skb, | |
15e47304 | 1551 | u32 portid, int nonblock) |
1da177e4 LT |
1552 | { |
1553 | struct sock *sk; | |
1554 | int err; | |
1555 | long timeo; | |
1556 | ||
1557 | skb = netlink_trim(skb, gfp_any()); | |
1558 | ||
1559 | timeo = sock_sndtimeo(ssk, nonblock); | |
1560 | retry: | |
15e47304 | 1561 | sk = netlink_getsockbyportid(ssk, portid); |
1da177e4 LT |
1562 | if (IS_ERR(sk)) { |
1563 | kfree_skb(skb); | |
1564 | return PTR_ERR(sk); | |
1565 | } | |
cd40b7d3 | 1566 | if (netlink_is_kernel(sk)) |
3fbc2905 | 1567 | return netlink_unicast_kernel(sk, skb, ssk); |
cd40b7d3 | 1568 | |
b1153f29 | 1569 | if (sk_filter(sk, skb)) { |
84874607 | 1570 | err = skb->len; |
b1153f29 SH |
1571 | kfree_skb(skb); |
1572 | sock_put(sk); | |
1573 | return err; | |
1574 | } | |
1575 | ||
9457afee | 1576 | err = netlink_attachskb(sk, skb, &timeo, ssk); |
1da177e4 LT |
1577 | if (err == 1) |
1578 | goto retry; | |
1579 | if (err) | |
1580 | return err; | |
1581 | ||
7ee015e0 | 1582 | return netlink_sendskb(sk, skb); |
1da177e4 | 1583 | } |
6ac552fd | 1584 | EXPORT_SYMBOL(netlink_unicast); |
1da177e4 | 1585 | |
f9c22888 PM |
1586 | struct sk_buff *netlink_alloc_skb(struct sock *ssk, unsigned int size, |
1587 | u32 dst_portid, gfp_t gfp_mask) | |
1588 | { | |
1589 | #ifdef CONFIG_NETLINK_MMAP | |
1590 | struct sock *sk = NULL; | |
1591 | struct sk_buff *skb; | |
1592 | struct netlink_ring *ring; | |
1593 | struct nl_mmap_hdr *hdr; | |
1594 | unsigned int maxlen; | |
1595 | ||
1596 | sk = netlink_getsockbyportid(ssk, dst_portid); | |
1597 | if (IS_ERR(sk)) | |
1598 | goto out; | |
1599 | ||
1600 | ring = &nlk_sk(sk)->rx_ring; | |
1601 | /* fast-path without atomic ops for common case: non-mmaped receiver */ | |
1602 | if (ring->pg_vec == NULL) | |
1603 | goto out_put; | |
1604 | ||
1605 | skb = alloc_skb_head(gfp_mask); | |
1606 | if (skb == NULL) | |
1607 | goto err1; | |
1608 | ||
1609 | spin_lock_bh(&sk->sk_receive_queue.lock); | |
1610 | /* check again under lock */ | |
1611 | if (ring->pg_vec == NULL) | |
1612 | goto out_free; | |
1613 | ||
1614 | maxlen = ring->frame_size - NL_MMAP_HDRLEN; | |
1615 | if (maxlen < size) | |
1616 | goto out_free; | |
1617 | ||
1618 | netlink_forward_ring(ring); | |
1619 | hdr = netlink_current_frame(ring, NL_MMAP_STATUS_UNUSED); | |
1620 | if (hdr == NULL) | |
1621 | goto err2; | |
1622 | netlink_ring_setup_skb(skb, sk, ring, hdr); | |
1623 | netlink_set_status(hdr, NL_MMAP_STATUS_RESERVED); | |
1624 | atomic_inc(&ring->pending); | |
1625 | netlink_increment_head(ring); | |
1626 | ||
1627 | spin_unlock_bh(&sk->sk_receive_queue.lock); | |
1628 | return skb; | |
1629 | ||
1630 | err2: | |
1631 | kfree_skb(skb); | |
1632 | spin_unlock_bh(&sk->sk_receive_queue.lock); | |
cd1df525 | 1633 | netlink_overrun(sk); |
f9c22888 PM |
1634 | err1: |
1635 | sock_put(sk); | |
1636 | return NULL; | |
1637 | ||
1638 | out_free: | |
1639 | kfree_skb(skb); | |
1640 | spin_unlock_bh(&sk->sk_receive_queue.lock); | |
1641 | out_put: | |
1642 | sock_put(sk); | |
1643 | out: | |
1644 | #endif | |
1645 | return alloc_skb(size, gfp_mask); | |
1646 | } | |
1647 | EXPORT_SYMBOL_GPL(netlink_alloc_skb); | |
1648 | ||
4277a083 PM |
1649 | int netlink_has_listeners(struct sock *sk, unsigned int group) |
1650 | { | |
1651 | int res = 0; | |
5c398dc8 | 1652 | struct listeners *listeners; |
4277a083 | 1653 | |
aed81560 | 1654 | BUG_ON(!netlink_is_kernel(sk)); |
b4ff4f04 JB |
1655 | |
1656 | rcu_read_lock(); | |
1657 | listeners = rcu_dereference(nl_table[sk->sk_protocol].listeners); | |
1658 | ||
6d772ac5 | 1659 | if (listeners && group - 1 < nl_table[sk->sk_protocol].groups) |
5c398dc8 | 1660 | res = test_bit(group - 1, listeners->masks); |
b4ff4f04 JB |
1661 | |
1662 | rcu_read_unlock(); | |
1663 | ||
4277a083 PM |
1664 | return res; |
1665 | } | |
1666 | EXPORT_SYMBOL_GPL(netlink_has_listeners); | |
1667 | ||
b57ef81f | 1668 | static int netlink_broadcast_deliver(struct sock *sk, struct sk_buff *skb) |
1da177e4 LT |
1669 | { |
1670 | struct netlink_sock *nlk = nlk_sk(sk); | |
1671 | ||
1672 | if (atomic_read(&sk->sk_rmem_alloc) <= sk->sk_rcvbuf && | |
cd967e05 | 1673 | !test_bit(NETLINK_CONGESTED, &nlk->state)) { |
cf0a018a | 1674 | netlink_skb_set_owner_r(skb, sk); |
4a7e7c2a | 1675 | __netlink_sendskb(sk, skb); |
2c645800 | 1676 | return atomic_read(&sk->sk_rmem_alloc) > (sk->sk_rcvbuf >> 1); |
1da177e4 LT |
1677 | } |
1678 | return -1; | |
1679 | } | |
1680 | ||
1681 | struct netlink_broadcast_data { | |
1682 | struct sock *exclude_sk; | |
b4b51029 | 1683 | struct net *net; |
15e47304 | 1684 | u32 portid; |
1da177e4 LT |
1685 | u32 group; |
1686 | int failure; | |
ff491a73 | 1687 | int delivery_failure; |
1da177e4 LT |
1688 | int congested; |
1689 | int delivered; | |
7d877f3b | 1690 | gfp_t allocation; |
1da177e4 | 1691 | struct sk_buff *skb, *skb2; |
910a7e90 EB |
1692 | int (*tx_filter)(struct sock *dsk, struct sk_buff *skb, void *data); |
1693 | void *tx_data; | |
1da177e4 LT |
1694 | }; |
1695 | ||
b57ef81f | 1696 | static int do_one_broadcast(struct sock *sk, |
1da177e4 LT |
1697 | struct netlink_broadcast_data *p) |
1698 | { | |
1699 | struct netlink_sock *nlk = nlk_sk(sk); | |
1700 | int val; | |
1701 | ||
1702 | if (p->exclude_sk == sk) | |
1703 | goto out; | |
1704 | ||
15e47304 | 1705 | if (nlk->portid == p->portid || p->group - 1 >= nlk->ngroups || |
f7fa9b10 | 1706 | !test_bit(p->group - 1, nlk->groups)) |
1da177e4 LT |
1707 | goto out; |
1708 | ||
878628fb | 1709 | if (!net_eq(sock_net(sk), p->net)) |
b4b51029 EB |
1710 | goto out; |
1711 | ||
1da177e4 LT |
1712 | if (p->failure) { |
1713 | netlink_overrun(sk); | |
1714 | goto out; | |
1715 | } | |
1716 | ||
1717 | sock_hold(sk); | |
1718 | if (p->skb2 == NULL) { | |
68acc024 | 1719 | if (skb_shared(p->skb)) { |
1da177e4 LT |
1720 | p->skb2 = skb_clone(p->skb, p->allocation); |
1721 | } else { | |
68acc024 TC |
1722 | p->skb2 = skb_get(p->skb); |
1723 | /* | |
1724 | * skb ownership may have been set when | |
1725 | * delivered to a previous socket. | |
1726 | */ | |
1727 | skb_orphan(p->skb2); | |
1da177e4 LT |
1728 | } |
1729 | } | |
1730 | if (p->skb2 == NULL) { | |
1731 | netlink_overrun(sk); | |
1732 | /* Clone failed. Notify ALL listeners. */ | |
1733 | p->failure = 1; | |
be0c22a4 PNA |
1734 | if (nlk->flags & NETLINK_BROADCAST_SEND_ERROR) |
1735 | p->delivery_failure = 1; | |
910a7e90 EB |
1736 | } else if (p->tx_filter && p->tx_filter(sk, p->skb2, p->tx_data)) { |
1737 | kfree_skb(p->skb2); | |
1738 | p->skb2 = NULL; | |
b1153f29 SH |
1739 | } else if (sk_filter(sk, p->skb2)) { |
1740 | kfree_skb(p->skb2); | |
1741 | p->skb2 = NULL; | |
1da177e4 LT |
1742 | } else if ((val = netlink_broadcast_deliver(sk, p->skb2)) < 0) { |
1743 | netlink_overrun(sk); | |
be0c22a4 PNA |
1744 | if (nlk->flags & NETLINK_BROADCAST_SEND_ERROR) |
1745 | p->delivery_failure = 1; | |
1da177e4 LT |
1746 | } else { |
1747 | p->congested |= val; | |
1748 | p->delivered = 1; | |
1749 | p->skb2 = NULL; | |
1750 | } | |
1751 | sock_put(sk); | |
1752 | ||
1753 | out: | |
1754 | return 0; | |
1755 | } | |
1756 | ||
15e47304 | 1757 | int netlink_broadcast_filtered(struct sock *ssk, struct sk_buff *skb, u32 portid, |
910a7e90 EB |
1758 | u32 group, gfp_t allocation, |
1759 | int (*filter)(struct sock *dsk, struct sk_buff *skb, void *data), | |
1760 | void *filter_data) | |
1da177e4 | 1761 | { |
3b1e0a65 | 1762 | struct net *net = sock_net(ssk); |
1da177e4 | 1763 | struct netlink_broadcast_data info; |
1da177e4 LT |
1764 | struct sock *sk; |
1765 | ||
1766 | skb = netlink_trim(skb, allocation); | |
1767 | ||
1768 | info.exclude_sk = ssk; | |
b4b51029 | 1769 | info.net = net; |
15e47304 | 1770 | info.portid = portid; |
1da177e4 LT |
1771 | info.group = group; |
1772 | info.failure = 0; | |
ff491a73 | 1773 | info.delivery_failure = 0; |
1da177e4 LT |
1774 | info.congested = 0; |
1775 | info.delivered = 0; | |
1776 | info.allocation = allocation; | |
1777 | info.skb = skb; | |
1778 | info.skb2 = NULL; | |
910a7e90 EB |
1779 | info.tx_filter = filter; |
1780 | info.tx_data = filter_data; | |
1da177e4 LT |
1781 | |
1782 | /* While we sleep in clone, do not allow to change socket list */ | |
1783 | ||
1784 | netlink_lock_table(); | |
1785 | ||
b67bfe0d | 1786 | sk_for_each_bound(sk, &nl_table[ssk->sk_protocol].mc_list) |
1da177e4 LT |
1787 | do_one_broadcast(sk, &info); |
1788 | ||
70d4bf6d | 1789 | consume_skb(skb); |
aa1c6a6f | 1790 | |
1da177e4 LT |
1791 | netlink_unlock_table(); |
1792 | ||
70d4bf6d NH |
1793 | if (info.delivery_failure) { |
1794 | kfree_skb(info.skb2); | |
ff491a73 | 1795 | return -ENOBUFS; |
658cb354 ED |
1796 | } |
1797 | consume_skb(info.skb2); | |
ff491a73 | 1798 | |
1da177e4 LT |
1799 | if (info.delivered) { |
1800 | if (info.congested && (allocation & __GFP_WAIT)) | |
1801 | yield(); | |
1802 | return 0; | |
1803 | } | |
1da177e4 LT |
1804 | return -ESRCH; |
1805 | } | |
910a7e90 EB |
1806 | EXPORT_SYMBOL(netlink_broadcast_filtered); |
1807 | ||
15e47304 | 1808 | int netlink_broadcast(struct sock *ssk, struct sk_buff *skb, u32 portid, |
910a7e90 EB |
1809 | u32 group, gfp_t allocation) |
1810 | { | |
15e47304 | 1811 | return netlink_broadcast_filtered(ssk, skb, portid, group, allocation, |
910a7e90 EB |
1812 | NULL, NULL); |
1813 | } | |
6ac552fd | 1814 | EXPORT_SYMBOL(netlink_broadcast); |
1da177e4 LT |
1815 | |
1816 | struct netlink_set_err_data { | |
1817 | struct sock *exclude_sk; | |
15e47304 | 1818 | u32 portid; |
1da177e4 LT |
1819 | u32 group; |
1820 | int code; | |
1821 | }; | |
1822 | ||
b57ef81f | 1823 | static int do_one_set_err(struct sock *sk, struct netlink_set_err_data *p) |
1da177e4 LT |
1824 | { |
1825 | struct netlink_sock *nlk = nlk_sk(sk); | |
1a50307b | 1826 | int ret = 0; |
1da177e4 LT |
1827 | |
1828 | if (sk == p->exclude_sk) | |
1829 | goto out; | |
1830 | ||
09ad9bc7 | 1831 | if (!net_eq(sock_net(sk), sock_net(p->exclude_sk))) |
b4b51029 EB |
1832 | goto out; |
1833 | ||
15e47304 | 1834 | if (nlk->portid == p->portid || p->group - 1 >= nlk->ngroups || |
f7fa9b10 | 1835 | !test_bit(p->group - 1, nlk->groups)) |
1da177e4 LT |
1836 | goto out; |
1837 | ||
1a50307b PNA |
1838 | if (p->code == ENOBUFS && nlk->flags & NETLINK_RECV_NO_ENOBUFS) { |
1839 | ret = 1; | |
1840 | goto out; | |
1841 | } | |
1842 | ||
1da177e4 LT |
1843 | sk->sk_err = p->code; |
1844 | sk->sk_error_report(sk); | |
1845 | out: | |
1a50307b | 1846 | return ret; |
1da177e4 LT |
1847 | } |
1848 | ||
4843b93c PNA |
1849 | /** |
1850 | * netlink_set_err - report error to broadcast listeners | |
1851 | * @ssk: the kernel netlink socket, as returned by netlink_kernel_create() | |
15e47304 | 1852 | * @portid: the PORTID of a process that we want to skip (if any) |
4843b93c PNA |
1853 | * @groups: the broadcast group that will notice the error |
1854 | * @code: error code, must be negative (as usual in kernelspace) | |
1a50307b PNA |
1855 | * |
1856 | * This function returns the number of broadcast listeners that have set the | |
1857 | * NETLINK_RECV_NO_ENOBUFS socket option. | |
4843b93c | 1858 | */ |
15e47304 | 1859 | int netlink_set_err(struct sock *ssk, u32 portid, u32 group, int code) |
1da177e4 LT |
1860 | { |
1861 | struct netlink_set_err_data info; | |
1da177e4 | 1862 | struct sock *sk; |
1a50307b | 1863 | int ret = 0; |
1da177e4 LT |
1864 | |
1865 | info.exclude_sk = ssk; | |
15e47304 | 1866 | info.portid = portid; |
1da177e4 | 1867 | info.group = group; |
4843b93c PNA |
1868 | /* sk->sk_err wants a positive error value */ |
1869 | info.code = -code; | |
1da177e4 LT |
1870 | |
1871 | read_lock(&nl_table_lock); | |
1872 | ||
b67bfe0d | 1873 | sk_for_each_bound(sk, &nl_table[ssk->sk_protocol].mc_list) |
1a50307b | 1874 | ret += do_one_set_err(sk, &info); |
1da177e4 LT |
1875 | |
1876 | read_unlock(&nl_table_lock); | |
1a50307b | 1877 | return ret; |
1da177e4 | 1878 | } |
dd5b6ce6 | 1879 | EXPORT_SYMBOL(netlink_set_err); |
1da177e4 | 1880 | |
84659eb5 JB |
1881 | /* must be called with netlink table grabbed */ |
1882 | static void netlink_update_socket_mc(struct netlink_sock *nlk, | |
1883 | unsigned int group, | |
1884 | int is_new) | |
1885 | { | |
1886 | int old, new = !!is_new, subscriptions; | |
1887 | ||
1888 | old = test_bit(group - 1, nlk->groups); | |
1889 | subscriptions = nlk->subscriptions - old + new; | |
1890 | if (new) | |
1891 | __set_bit(group - 1, nlk->groups); | |
1892 | else | |
1893 | __clear_bit(group - 1, nlk->groups); | |
1894 | netlink_update_subscriptions(&nlk->sk, subscriptions); | |
1895 | netlink_update_listeners(&nlk->sk); | |
1896 | } | |
1897 | ||
9a4595bc | 1898 | static int netlink_setsockopt(struct socket *sock, int level, int optname, |
b7058842 | 1899 | char __user *optval, unsigned int optlen) |
9a4595bc PM |
1900 | { |
1901 | struct sock *sk = sock->sk; | |
1902 | struct netlink_sock *nlk = nlk_sk(sk); | |
eb496534 JB |
1903 | unsigned int val = 0; |
1904 | int err; | |
9a4595bc PM |
1905 | |
1906 | if (level != SOL_NETLINK) | |
1907 | return -ENOPROTOOPT; | |
1908 | ||
ccdfcc39 PM |
1909 | if (optname != NETLINK_RX_RING && optname != NETLINK_TX_RING && |
1910 | optlen >= sizeof(int) && | |
eb496534 | 1911 | get_user(val, (unsigned int __user *)optval)) |
9a4595bc PM |
1912 | return -EFAULT; |
1913 | ||
1914 | switch (optname) { | |
1915 | case NETLINK_PKTINFO: | |
1916 | if (val) | |
1917 | nlk->flags |= NETLINK_RECV_PKTINFO; | |
1918 | else | |
1919 | nlk->flags &= ~NETLINK_RECV_PKTINFO; | |
1920 | err = 0; | |
1921 | break; | |
1922 | case NETLINK_ADD_MEMBERSHIP: | |
1923 | case NETLINK_DROP_MEMBERSHIP: { | |
9785e10a | 1924 | if (!netlink_capable(sock, NL_CFG_F_NONROOT_RECV)) |
9a4595bc | 1925 | return -EPERM; |
b4ff4f04 JB |
1926 | err = netlink_realloc_groups(sk); |
1927 | if (err) | |
1928 | return err; | |
9a4595bc PM |
1929 | if (!val || val - 1 >= nlk->ngroups) |
1930 | return -EINVAL; | |
1931 | netlink_table_grab(); | |
84659eb5 JB |
1932 | netlink_update_socket_mc(nlk, val, |
1933 | optname == NETLINK_ADD_MEMBERSHIP); | |
9a4595bc | 1934 | netlink_table_ungrab(); |
03292745 PNA |
1935 | |
1936 | if (nlk->netlink_bind) | |
1937 | nlk->netlink_bind(val); | |
1938 | ||
9a4595bc PM |
1939 | err = 0; |
1940 | break; | |
1941 | } | |
be0c22a4 PNA |
1942 | case NETLINK_BROADCAST_ERROR: |
1943 | if (val) | |
1944 | nlk->flags |= NETLINK_BROADCAST_SEND_ERROR; | |
1945 | else | |
1946 | nlk->flags &= ~NETLINK_BROADCAST_SEND_ERROR; | |
1947 | err = 0; | |
1948 | break; | |
38938bfe PNA |
1949 | case NETLINK_NO_ENOBUFS: |
1950 | if (val) { | |
1951 | nlk->flags |= NETLINK_RECV_NO_ENOBUFS; | |
cd967e05 | 1952 | clear_bit(NETLINK_CONGESTED, &nlk->state); |
38938bfe | 1953 | wake_up_interruptible(&nlk->wait); |
658cb354 | 1954 | } else { |
38938bfe | 1955 | nlk->flags &= ~NETLINK_RECV_NO_ENOBUFS; |
658cb354 | 1956 | } |
38938bfe PNA |
1957 | err = 0; |
1958 | break; | |
ccdfcc39 PM |
1959 | #ifdef CONFIG_NETLINK_MMAP |
1960 | case NETLINK_RX_RING: | |
1961 | case NETLINK_TX_RING: { | |
1962 | struct nl_mmap_req req; | |
1963 | ||
1964 | /* Rings might consume more memory than queue limits, require | |
1965 | * CAP_NET_ADMIN. | |
1966 | */ | |
1967 | if (!capable(CAP_NET_ADMIN)) | |
1968 | return -EPERM; | |
1969 | if (optlen < sizeof(req)) | |
1970 | return -EINVAL; | |
1971 | if (copy_from_user(&req, optval, sizeof(req))) | |
1972 | return -EFAULT; | |
1973 | err = netlink_set_ring(sk, &req, false, | |
1974 | optname == NETLINK_TX_RING); | |
1975 | break; | |
1976 | } | |
1977 | #endif /* CONFIG_NETLINK_MMAP */ | |
9a4595bc PM |
1978 | default: |
1979 | err = -ENOPROTOOPT; | |
1980 | } | |
1981 | return err; | |
1982 | } | |
1983 | ||
1984 | static int netlink_getsockopt(struct socket *sock, int level, int optname, | |
746fac4d | 1985 | char __user *optval, int __user *optlen) |
9a4595bc PM |
1986 | { |
1987 | struct sock *sk = sock->sk; | |
1988 | struct netlink_sock *nlk = nlk_sk(sk); | |
1989 | int len, val, err; | |
1990 | ||
1991 | if (level != SOL_NETLINK) | |
1992 | return -ENOPROTOOPT; | |
1993 | ||
1994 | if (get_user(len, optlen)) | |
1995 | return -EFAULT; | |
1996 | if (len < 0) | |
1997 | return -EINVAL; | |
1998 | ||
1999 | switch (optname) { | |
2000 | case NETLINK_PKTINFO: | |
2001 | if (len < sizeof(int)) | |
2002 | return -EINVAL; | |
2003 | len = sizeof(int); | |
2004 | val = nlk->flags & NETLINK_RECV_PKTINFO ? 1 : 0; | |
a27b58fe HC |
2005 | if (put_user(len, optlen) || |
2006 | put_user(val, optval)) | |
2007 | return -EFAULT; | |
9a4595bc PM |
2008 | err = 0; |
2009 | break; | |
be0c22a4 PNA |
2010 | case NETLINK_BROADCAST_ERROR: |
2011 | if (len < sizeof(int)) | |
2012 | return -EINVAL; | |
2013 | len = sizeof(int); | |
2014 | val = nlk->flags & NETLINK_BROADCAST_SEND_ERROR ? 1 : 0; | |
2015 | if (put_user(len, optlen) || | |
2016 | put_user(val, optval)) | |
2017 | return -EFAULT; | |
2018 | err = 0; | |
2019 | break; | |
38938bfe PNA |
2020 | case NETLINK_NO_ENOBUFS: |
2021 | if (len < sizeof(int)) | |
2022 | return -EINVAL; | |
2023 | len = sizeof(int); | |
2024 | val = nlk->flags & NETLINK_RECV_NO_ENOBUFS ? 1 : 0; | |
2025 | if (put_user(len, optlen) || | |
2026 | put_user(val, optval)) | |
2027 | return -EFAULT; | |
2028 | err = 0; | |
2029 | break; | |
9a4595bc PM |
2030 | default: |
2031 | err = -ENOPROTOOPT; | |
2032 | } | |
2033 | return err; | |
2034 | } | |
2035 | ||
2036 | static void netlink_cmsg_recv_pktinfo(struct msghdr *msg, struct sk_buff *skb) | |
2037 | { | |
2038 | struct nl_pktinfo info; | |
2039 | ||
2040 | info.group = NETLINK_CB(skb).dst_group; | |
2041 | put_cmsg(msg, SOL_NETLINK, NETLINK_PKTINFO, sizeof(info), &info); | |
2042 | } | |
2043 | ||
1da177e4 LT |
2044 | static int netlink_sendmsg(struct kiocb *kiocb, struct socket *sock, |
2045 | struct msghdr *msg, size_t len) | |
2046 | { | |
2047 | struct sock_iocb *siocb = kiocb_to_siocb(kiocb); | |
2048 | struct sock *sk = sock->sk; | |
2049 | struct netlink_sock *nlk = nlk_sk(sk); | |
6ac552fd | 2050 | struct sockaddr_nl *addr = msg->msg_name; |
15e47304 | 2051 | u32 dst_portid; |
d629b836 | 2052 | u32 dst_group; |
1da177e4 LT |
2053 | struct sk_buff *skb; |
2054 | int err; | |
2055 | struct scm_cookie scm; | |
2056 | ||
2057 | if (msg->msg_flags&MSG_OOB) | |
2058 | return -EOPNOTSUPP; | |
2059 | ||
16e57262 | 2060 | if (NULL == siocb->scm) |
1da177e4 | 2061 | siocb->scm = &scm; |
16e57262 | 2062 | |
e0e3cea4 | 2063 | err = scm_send(sock, msg, siocb->scm, true); |
1da177e4 LT |
2064 | if (err < 0) |
2065 | return err; | |
2066 | ||
2067 | if (msg->msg_namelen) { | |
b47030c7 | 2068 | err = -EINVAL; |
1da177e4 | 2069 | if (addr->nl_family != AF_NETLINK) |
b47030c7 | 2070 | goto out; |
15e47304 | 2071 | dst_portid = addr->nl_pid; |
d629b836 | 2072 | dst_group = ffs(addr->nl_groups); |
b47030c7 | 2073 | err = -EPERM; |
15e47304 | 2074 | if ((dst_group || dst_portid) && |
9785e10a | 2075 | !netlink_capable(sock, NL_CFG_F_NONROOT_SEND)) |
b47030c7 | 2076 | goto out; |
1da177e4 | 2077 | } else { |
15e47304 | 2078 | dst_portid = nlk->dst_portid; |
d629b836 | 2079 | dst_group = nlk->dst_group; |
1da177e4 LT |
2080 | } |
2081 | ||
15e47304 | 2082 | if (!nlk->portid) { |
1da177e4 LT |
2083 | err = netlink_autobind(sock); |
2084 | if (err) | |
2085 | goto out; | |
2086 | } | |
2087 | ||
5fd96123 PM |
2088 | if (netlink_tx_is_mmaped(sk) && |
2089 | msg->msg_iov->iov_base == NULL) { | |
2090 | err = netlink_mmap_sendmsg(sk, msg, dst_portid, dst_group, | |
2091 | siocb); | |
2092 | goto out; | |
2093 | } | |
2094 | ||
1da177e4 LT |
2095 | err = -EMSGSIZE; |
2096 | if (len > sk->sk_sndbuf - 32) | |
2097 | goto out; | |
2098 | err = -ENOBUFS; | |
339bf98f | 2099 | skb = alloc_skb(len, GFP_KERNEL); |
6ac552fd | 2100 | if (skb == NULL) |
1da177e4 LT |
2101 | goto out; |
2102 | ||
15e47304 | 2103 | NETLINK_CB(skb).portid = nlk->portid; |
d629b836 | 2104 | NETLINK_CB(skb).dst_group = dst_group; |
dbe9a417 | 2105 | NETLINK_CB(skb).creds = siocb->scm->creds; |
1da177e4 | 2106 | |
1da177e4 | 2107 | err = -EFAULT; |
6ac552fd | 2108 | if (memcpy_fromiovec(skb_put(skb, len), msg->msg_iov, len)) { |
1da177e4 LT |
2109 | kfree_skb(skb); |
2110 | goto out; | |
2111 | } | |
2112 | ||
2113 | err = security_netlink_send(sk, skb); | |
2114 | if (err) { | |
2115 | kfree_skb(skb); | |
2116 | goto out; | |
2117 | } | |
2118 | ||
d629b836 | 2119 | if (dst_group) { |
1da177e4 | 2120 | atomic_inc(&skb->users); |
15e47304 | 2121 | netlink_broadcast(sk, skb, dst_portid, dst_group, GFP_KERNEL); |
1da177e4 | 2122 | } |
15e47304 | 2123 | err = netlink_unicast(sk, skb, dst_portid, msg->msg_flags&MSG_DONTWAIT); |
1da177e4 LT |
2124 | |
2125 | out: | |
b47030c7 | 2126 | scm_destroy(siocb->scm); |
1da177e4 LT |
2127 | return err; |
2128 | } | |
2129 | ||
2130 | static int netlink_recvmsg(struct kiocb *kiocb, struct socket *sock, | |
2131 | struct msghdr *msg, size_t len, | |
2132 | int flags) | |
2133 | { | |
2134 | struct sock_iocb *siocb = kiocb_to_siocb(kiocb); | |
2135 | struct scm_cookie scm; | |
2136 | struct sock *sk = sock->sk; | |
2137 | struct netlink_sock *nlk = nlk_sk(sk); | |
2138 | int noblock = flags&MSG_DONTWAIT; | |
2139 | size_t copied; | |
68d6ac6d | 2140 | struct sk_buff *skb, *data_skb; |
b44d211e | 2141 | int err, ret; |
1da177e4 LT |
2142 | |
2143 | if (flags&MSG_OOB) | |
2144 | return -EOPNOTSUPP; | |
2145 | ||
2146 | copied = 0; | |
2147 | ||
6ac552fd PM |
2148 | skb = skb_recv_datagram(sk, flags, noblock, &err); |
2149 | if (skb == NULL) | |
1da177e4 LT |
2150 | goto out; |
2151 | ||
68d6ac6d JB |
2152 | data_skb = skb; |
2153 | ||
1dacc76d JB |
2154 | #ifdef CONFIG_COMPAT_NETLINK_MESSAGES |
2155 | if (unlikely(skb_shinfo(skb)->frag_list)) { | |
1dacc76d | 2156 | /* |
68d6ac6d JB |
2157 | * If this skb has a frag_list, then here that means that we |
2158 | * will have to use the frag_list skb's data for compat tasks | |
2159 | * and the regular skb's data for normal (non-compat) tasks. | |
1dacc76d | 2160 | * |
68d6ac6d JB |
2161 | * If we need to send the compat skb, assign it to the |
2162 | * 'data_skb' variable so that it will be used below for data | |
2163 | * copying. We keep 'skb' for everything else, including | |
2164 | * freeing both later. | |
1dacc76d | 2165 | */ |
68d6ac6d JB |
2166 | if (flags & MSG_CMSG_COMPAT) |
2167 | data_skb = skb_shinfo(skb)->frag_list; | |
1dacc76d JB |
2168 | } |
2169 | #endif | |
2170 | ||
1da177e4 LT |
2171 | msg->msg_namelen = 0; |
2172 | ||
68d6ac6d | 2173 | copied = data_skb->len; |
1da177e4 LT |
2174 | if (len < copied) { |
2175 | msg->msg_flags |= MSG_TRUNC; | |
2176 | copied = len; | |
2177 | } | |
2178 | ||
68d6ac6d JB |
2179 | skb_reset_transport_header(data_skb); |
2180 | err = skb_copy_datagram_iovec(data_skb, 0, msg->msg_iov, copied); | |
1da177e4 LT |
2181 | |
2182 | if (msg->msg_name) { | |
6ac552fd | 2183 | struct sockaddr_nl *addr = (struct sockaddr_nl *)msg->msg_name; |
1da177e4 LT |
2184 | addr->nl_family = AF_NETLINK; |
2185 | addr->nl_pad = 0; | |
15e47304 | 2186 | addr->nl_pid = NETLINK_CB(skb).portid; |
d629b836 | 2187 | addr->nl_groups = netlink_group_mask(NETLINK_CB(skb).dst_group); |
1da177e4 LT |
2188 | msg->msg_namelen = sizeof(*addr); |
2189 | } | |
2190 | ||
cc9a06cd PM |
2191 | if (nlk->flags & NETLINK_RECV_PKTINFO) |
2192 | netlink_cmsg_recv_pktinfo(msg, skb); | |
2193 | ||
1da177e4 LT |
2194 | if (NULL == siocb->scm) { |
2195 | memset(&scm, 0, sizeof(scm)); | |
2196 | siocb->scm = &scm; | |
2197 | } | |
2198 | siocb->scm->creds = *NETLINK_CREDS(skb); | |
188ccb55 | 2199 | if (flags & MSG_TRUNC) |
68d6ac6d | 2200 | copied = data_skb->len; |
daa3766e | 2201 | |
1da177e4 LT |
2202 | skb_free_datagram(sk, skb); |
2203 | ||
b44d211e AV |
2204 | if (nlk->cb && atomic_read(&sk->sk_rmem_alloc) <= sk->sk_rcvbuf / 2) { |
2205 | ret = netlink_dump(sk); | |
2206 | if (ret) { | |
2207 | sk->sk_err = ret; | |
2208 | sk->sk_error_report(sk); | |
2209 | } | |
2210 | } | |
1da177e4 LT |
2211 | |
2212 | scm_recv(sock, msg, siocb->scm, flags); | |
1da177e4 LT |
2213 | out: |
2214 | netlink_rcv_wake(sk); | |
2215 | return err ? : copied; | |
2216 | } | |
2217 | ||
2218 | static void netlink_data_ready(struct sock *sk, int len) | |
2219 | { | |
cd40b7d3 | 2220 | BUG(); |
1da177e4 LT |
2221 | } |
2222 | ||
2223 | /* | |
746fac4d | 2224 | * We export these functions to other modules. They provide a |
1da177e4 LT |
2225 | * complete set of kernel non-blocking support for message |
2226 | * queueing. | |
2227 | */ | |
2228 | ||
2229 | struct sock * | |
9f00d977 PNA |
2230 | __netlink_kernel_create(struct net *net, int unit, struct module *module, |
2231 | struct netlink_kernel_cfg *cfg) | |
1da177e4 LT |
2232 | { |
2233 | struct socket *sock; | |
2234 | struct sock *sk; | |
77247bbb | 2235 | struct netlink_sock *nlk; |
5c398dc8 | 2236 | struct listeners *listeners = NULL; |
a31f2d17 PNA |
2237 | struct mutex *cb_mutex = cfg ? cfg->cb_mutex : NULL; |
2238 | unsigned int groups; | |
1da177e4 | 2239 | |
fab2caf6 | 2240 | BUG_ON(!nl_table); |
1da177e4 | 2241 | |
6ac552fd | 2242 | if (unit < 0 || unit >= MAX_LINKS) |
1da177e4 LT |
2243 | return NULL; |
2244 | ||
2245 | if (sock_create_lite(PF_NETLINK, SOCK_DGRAM, unit, &sock)) | |
2246 | return NULL; | |
2247 | ||
23fe1866 PE |
2248 | /* |
2249 | * We have to just have a reference on the net from sk, but don't | |
2250 | * get_net it. Besides, we cannot get and then put the net here. | |
2251 | * So we create one inside init_net and the move it to net. | |
2252 | */ | |
2253 | ||
2254 | if (__netlink_create(&init_net, sock, cb_mutex, unit) < 0) | |
2255 | goto out_sock_release_nosk; | |
2256 | ||
2257 | sk = sock->sk; | |
edf02087 | 2258 | sk_change_net(sk, net); |
4fdb3bb7 | 2259 | |
a31f2d17 | 2260 | if (!cfg || cfg->groups < 32) |
4277a083 | 2261 | groups = 32; |
a31f2d17 PNA |
2262 | else |
2263 | groups = cfg->groups; | |
4277a083 | 2264 | |
5c398dc8 | 2265 | listeners = kzalloc(sizeof(*listeners) + NLGRPSZ(groups), GFP_KERNEL); |
4277a083 PM |
2266 | if (!listeners) |
2267 | goto out_sock_release; | |
2268 | ||
1da177e4 | 2269 | sk->sk_data_ready = netlink_data_ready; |
a31f2d17 PNA |
2270 | if (cfg && cfg->input) |
2271 | nlk_sk(sk)->netlink_rcv = cfg->input; | |
1da177e4 | 2272 | |
b4b51029 | 2273 | if (netlink_insert(sk, net, 0)) |
77247bbb | 2274 | goto out_sock_release; |
4fdb3bb7 | 2275 | |
77247bbb PM |
2276 | nlk = nlk_sk(sk); |
2277 | nlk->flags |= NETLINK_KERNEL_SOCKET; | |
4fdb3bb7 | 2278 | |
4fdb3bb7 | 2279 | netlink_table_grab(); |
b4b51029 EB |
2280 | if (!nl_table[unit].registered) { |
2281 | nl_table[unit].groups = groups; | |
5c398dc8 | 2282 | rcu_assign_pointer(nl_table[unit].listeners, listeners); |
b4b51029 EB |
2283 | nl_table[unit].cb_mutex = cb_mutex; |
2284 | nl_table[unit].module = module; | |
9785e10a PNA |
2285 | if (cfg) { |
2286 | nl_table[unit].bind = cfg->bind; | |
2287 | nl_table[unit].flags = cfg->flags; | |
2288 | } | |
b4b51029 | 2289 | nl_table[unit].registered = 1; |
f937f1f4 JJ |
2290 | } else { |
2291 | kfree(listeners); | |
869e58f8 | 2292 | nl_table[unit].registered++; |
b4b51029 | 2293 | } |
4fdb3bb7 | 2294 | netlink_table_ungrab(); |
77247bbb PM |
2295 | return sk; |
2296 | ||
4fdb3bb7 | 2297 | out_sock_release: |
4277a083 | 2298 | kfree(listeners); |
9dfbec1f | 2299 | netlink_kernel_release(sk); |
23fe1866 PE |
2300 | return NULL; |
2301 | ||
2302 | out_sock_release_nosk: | |
4fdb3bb7 | 2303 | sock_release(sock); |
77247bbb | 2304 | return NULL; |
1da177e4 | 2305 | } |
9f00d977 | 2306 | EXPORT_SYMBOL(__netlink_kernel_create); |
b7c6ba6e DL |
2307 | |
2308 | void | |
2309 | netlink_kernel_release(struct sock *sk) | |
2310 | { | |
edf02087 | 2311 | sk_release_kernel(sk); |
b7c6ba6e DL |
2312 | } |
2313 | EXPORT_SYMBOL(netlink_kernel_release); | |
2314 | ||
d136f1bd | 2315 | int __netlink_change_ngroups(struct sock *sk, unsigned int groups) |
b4ff4f04 | 2316 | { |
5c398dc8 | 2317 | struct listeners *new, *old; |
b4ff4f04 | 2318 | struct netlink_table *tbl = &nl_table[sk->sk_protocol]; |
b4ff4f04 JB |
2319 | |
2320 | if (groups < 32) | |
2321 | groups = 32; | |
2322 | ||
b4ff4f04 | 2323 | if (NLGRPSZ(tbl->groups) < NLGRPSZ(groups)) { |
5c398dc8 ED |
2324 | new = kzalloc(sizeof(*new) + NLGRPSZ(groups), GFP_ATOMIC); |
2325 | if (!new) | |
d136f1bd | 2326 | return -ENOMEM; |
6d772ac5 | 2327 | old = nl_deref_protected(tbl->listeners); |
5c398dc8 ED |
2328 | memcpy(new->masks, old->masks, NLGRPSZ(tbl->groups)); |
2329 | rcu_assign_pointer(tbl->listeners, new); | |
2330 | ||
37b6b935 | 2331 | kfree_rcu(old, rcu); |
b4ff4f04 JB |
2332 | } |
2333 | tbl->groups = groups; | |
2334 | ||
d136f1bd JB |
2335 | return 0; |
2336 | } | |
2337 | ||
2338 | /** | |
2339 | * netlink_change_ngroups - change number of multicast groups | |
2340 | * | |
2341 | * This changes the number of multicast groups that are available | |
2342 | * on a certain netlink family. Note that it is not possible to | |
2343 | * change the number of groups to below 32. Also note that it does | |
2344 | * not implicitly call netlink_clear_multicast_users() when the | |
2345 | * number of groups is reduced. | |
2346 | * | |
2347 | * @sk: The kernel netlink socket, as returned by netlink_kernel_create(). | |
2348 | * @groups: The new number of groups. | |
2349 | */ | |
2350 | int netlink_change_ngroups(struct sock *sk, unsigned int groups) | |
2351 | { | |
2352 | int err; | |
2353 | ||
2354 | netlink_table_grab(); | |
2355 | err = __netlink_change_ngroups(sk, groups); | |
b4ff4f04 | 2356 | netlink_table_ungrab(); |
d136f1bd | 2357 | |
b4ff4f04 JB |
2358 | return err; |
2359 | } | |
b4ff4f04 | 2360 | |
b8273570 JB |
2361 | void __netlink_clear_multicast_users(struct sock *ksk, unsigned int group) |
2362 | { | |
2363 | struct sock *sk; | |
b8273570 JB |
2364 | struct netlink_table *tbl = &nl_table[ksk->sk_protocol]; |
2365 | ||
b67bfe0d | 2366 | sk_for_each_bound(sk, &tbl->mc_list) |
b8273570 JB |
2367 | netlink_update_socket_mc(nlk_sk(sk), group, 0); |
2368 | } | |
2369 | ||
84659eb5 JB |
2370 | /** |
2371 | * netlink_clear_multicast_users - kick off multicast listeners | |
2372 | * | |
2373 | * This function removes all listeners from the given group. | |
2374 | * @ksk: The kernel netlink socket, as returned by | |
2375 | * netlink_kernel_create(). | |
2376 | * @group: The multicast group to clear. | |
2377 | */ | |
2378 | void netlink_clear_multicast_users(struct sock *ksk, unsigned int group) | |
2379 | { | |
84659eb5 | 2380 | netlink_table_grab(); |
b8273570 | 2381 | __netlink_clear_multicast_users(ksk, group); |
84659eb5 JB |
2382 | netlink_table_ungrab(); |
2383 | } | |
84659eb5 | 2384 | |
a46621a3 | 2385 | struct nlmsghdr * |
15e47304 | 2386 | __nlmsg_put(struct sk_buff *skb, u32 portid, u32 seq, int type, int len, int flags) |
a46621a3 DV |
2387 | { |
2388 | struct nlmsghdr *nlh; | |
573ce260 | 2389 | int size = nlmsg_msg_size(len); |
a46621a3 DV |
2390 | |
2391 | nlh = (struct nlmsghdr*)skb_put(skb, NLMSG_ALIGN(size)); | |
2392 | nlh->nlmsg_type = type; | |
2393 | nlh->nlmsg_len = size; | |
2394 | nlh->nlmsg_flags = flags; | |
15e47304 | 2395 | nlh->nlmsg_pid = portid; |
a46621a3 DV |
2396 | nlh->nlmsg_seq = seq; |
2397 | if (!__builtin_constant_p(size) || NLMSG_ALIGN(size) - size != 0) | |
573ce260 | 2398 | memset(nlmsg_data(nlh) + len, 0, NLMSG_ALIGN(size) - size); |
a46621a3 DV |
2399 | return nlh; |
2400 | } | |
2401 | EXPORT_SYMBOL(__nlmsg_put); | |
2402 | ||
1da177e4 LT |
2403 | /* |
2404 | * It looks a bit ugly. | |
2405 | * It would be better to create kernel thread. | |
2406 | */ | |
2407 | ||
2408 | static int netlink_dump(struct sock *sk) | |
2409 | { | |
2410 | struct netlink_sock *nlk = nlk_sk(sk); | |
2411 | struct netlink_callback *cb; | |
c7ac8679 | 2412 | struct sk_buff *skb = NULL; |
1da177e4 | 2413 | struct nlmsghdr *nlh; |
bf8b79e4 | 2414 | int len, err = -ENOBUFS; |
c7ac8679 | 2415 | int alloc_size; |
1da177e4 | 2416 | |
af65bdfc | 2417 | mutex_lock(nlk->cb_mutex); |
1da177e4 LT |
2418 | |
2419 | cb = nlk->cb; | |
2420 | if (cb == NULL) { | |
bf8b79e4 TG |
2421 | err = -EINVAL; |
2422 | goto errout_skb; | |
1da177e4 LT |
2423 | } |
2424 | ||
c7ac8679 GR |
2425 | alloc_size = max_t(int, cb->min_dump_alloc, NLMSG_GOODSIZE); |
2426 | ||
f9c22888 PM |
2427 | if (!netlink_rx_is_mmaped(sk) && |
2428 | atomic_read(&sk->sk_rmem_alloc) >= sk->sk_rcvbuf) | |
2429 | goto errout_skb; | |
2430 | skb = netlink_alloc_skb(sk, alloc_size, nlk->portid, GFP_KERNEL); | |
c7ac8679 | 2431 | if (!skb) |
c63d6ea3 | 2432 | goto errout_skb; |
f9c22888 | 2433 | netlink_skb_set_owner_r(skb, sk); |
c7ac8679 | 2434 | |
1da177e4 LT |
2435 | len = cb->dump(skb, cb); |
2436 | ||
2437 | if (len > 0) { | |
af65bdfc | 2438 | mutex_unlock(nlk->cb_mutex); |
b1153f29 SH |
2439 | |
2440 | if (sk_filter(sk, skb)) | |
2441 | kfree_skb(skb); | |
4a7e7c2a ED |
2442 | else |
2443 | __netlink_sendskb(sk, skb); | |
1da177e4 LT |
2444 | return 0; |
2445 | } | |
2446 | ||
bf8b79e4 TG |
2447 | nlh = nlmsg_put_answer(skb, cb, NLMSG_DONE, sizeof(len), NLM_F_MULTI); |
2448 | if (!nlh) | |
2449 | goto errout_skb; | |
2450 | ||
670dc283 JB |
2451 | nl_dump_check_consistent(cb, nlh); |
2452 | ||
bf8b79e4 TG |
2453 | memcpy(nlmsg_data(nlh), &len, sizeof(len)); |
2454 | ||
b1153f29 SH |
2455 | if (sk_filter(sk, skb)) |
2456 | kfree_skb(skb); | |
4a7e7c2a ED |
2457 | else |
2458 | __netlink_sendskb(sk, skb); | |
1da177e4 | 2459 | |
a8f74b22 TG |
2460 | if (cb->done) |
2461 | cb->done(cb); | |
1da177e4 | 2462 | nlk->cb = NULL; |
af65bdfc | 2463 | mutex_unlock(nlk->cb_mutex); |
1da177e4 | 2464 | |
6dc878a8 | 2465 | module_put(cb->module); |
bfb253c9 | 2466 | netlink_consume_callback(cb); |
1da177e4 | 2467 | return 0; |
1797754e | 2468 | |
bf8b79e4 | 2469 | errout_skb: |
af65bdfc | 2470 | mutex_unlock(nlk->cb_mutex); |
bf8b79e4 | 2471 | kfree_skb(skb); |
bf8b79e4 | 2472 | return err; |
1da177e4 LT |
2473 | } |
2474 | ||
6dc878a8 G |
2475 | int __netlink_dump_start(struct sock *ssk, struct sk_buff *skb, |
2476 | const struct nlmsghdr *nlh, | |
2477 | struct netlink_dump_control *control) | |
1da177e4 LT |
2478 | { |
2479 | struct netlink_callback *cb; | |
2480 | struct sock *sk; | |
2481 | struct netlink_sock *nlk; | |
b44d211e | 2482 | int ret; |
1da177e4 | 2483 | |
0da974f4 | 2484 | cb = kzalloc(sizeof(*cb), GFP_KERNEL); |
1da177e4 LT |
2485 | if (cb == NULL) |
2486 | return -ENOBUFS; | |
2487 | ||
f9c22888 PM |
2488 | /* Memory mapped dump requests need to be copied to avoid looping |
2489 | * on the pending state in netlink_mmap_sendmsg() while the CB hold | |
2490 | * a reference to the skb. | |
2491 | */ | |
2492 | if (netlink_skb_is_mmaped(skb)) { | |
2493 | skb = skb_copy(skb, GFP_KERNEL); | |
2494 | if (skb == NULL) { | |
2495 | kfree(cb); | |
2496 | return -ENOBUFS; | |
2497 | } | |
2498 | } else | |
2499 | atomic_inc(&skb->users); | |
2500 | ||
80d326fa PNA |
2501 | cb->dump = control->dump; |
2502 | cb->done = control->done; | |
1da177e4 | 2503 | cb->nlh = nlh; |
7175c883 | 2504 | cb->data = control->data; |
6dc878a8 | 2505 | cb->module = control->module; |
80d326fa | 2506 | cb->min_dump_alloc = control->min_dump_alloc; |
1da177e4 LT |
2507 | cb->skb = skb; |
2508 | ||
15e47304 | 2509 | sk = netlink_lookup(sock_net(ssk), ssk->sk_protocol, NETLINK_CB(skb).portid); |
1da177e4 LT |
2510 | if (sk == NULL) { |
2511 | netlink_destroy_callback(cb); | |
2512 | return -ECONNREFUSED; | |
2513 | } | |
2514 | nlk = nlk_sk(sk); | |
6dc878a8 | 2515 | |
af65bdfc | 2516 | mutex_lock(nlk->cb_mutex); |
6dc878a8 | 2517 | /* A dump is in progress... */ |
3f660d66 | 2518 | if (nlk->cb) { |
af65bdfc | 2519 | mutex_unlock(nlk->cb_mutex); |
1da177e4 | 2520 | netlink_destroy_callback(cb); |
6dc878a8 G |
2521 | ret = -EBUSY; |
2522 | goto out; | |
1da177e4 | 2523 | } |
6dc878a8 G |
2524 | /* add reference of module which cb->dump belongs to */ |
2525 | if (!try_module_get(cb->module)) { | |
2526 | mutex_unlock(nlk->cb_mutex); | |
2527 | netlink_destroy_callback(cb); | |
2528 | ret = -EPROTONOSUPPORT; | |
2529 | goto out; | |
2530 | } | |
2531 | ||
1da177e4 | 2532 | nlk->cb = cb; |
af65bdfc | 2533 | mutex_unlock(nlk->cb_mutex); |
1da177e4 | 2534 | |
b44d211e | 2535 | ret = netlink_dump(sk); |
6dc878a8 | 2536 | out: |
1da177e4 | 2537 | sock_put(sk); |
5c58298c | 2538 | |
b44d211e AV |
2539 | if (ret) |
2540 | return ret; | |
2541 | ||
5c58298c DL |
2542 | /* We successfully started a dump, by returning -EINTR we |
2543 | * signal not to send ACK even if it was requested. | |
2544 | */ | |
2545 | return -EINTR; | |
1da177e4 | 2546 | } |
6dc878a8 | 2547 | EXPORT_SYMBOL(__netlink_dump_start); |
1da177e4 LT |
2548 | |
2549 | void netlink_ack(struct sk_buff *in_skb, struct nlmsghdr *nlh, int err) | |
2550 | { | |
2551 | struct sk_buff *skb; | |
2552 | struct nlmsghdr *rep; | |
2553 | struct nlmsgerr *errmsg; | |
339bf98f | 2554 | size_t payload = sizeof(*errmsg); |
1da177e4 | 2555 | |
339bf98f TG |
2556 | /* error messages get the original request appened */ |
2557 | if (err) | |
2558 | payload += nlmsg_len(nlh); | |
1da177e4 | 2559 | |
f9c22888 PM |
2560 | skb = netlink_alloc_skb(in_skb->sk, nlmsg_total_size(payload), |
2561 | NETLINK_CB(in_skb).portid, GFP_KERNEL); | |
1da177e4 LT |
2562 | if (!skb) { |
2563 | struct sock *sk; | |
2564 | ||
3b1e0a65 | 2565 | sk = netlink_lookup(sock_net(in_skb->sk), |
b4b51029 | 2566 | in_skb->sk->sk_protocol, |
15e47304 | 2567 | NETLINK_CB(in_skb).portid); |
1da177e4 LT |
2568 | if (sk) { |
2569 | sk->sk_err = ENOBUFS; | |
2570 | sk->sk_error_report(sk); | |
2571 | sock_put(sk); | |
2572 | } | |
2573 | return; | |
2574 | } | |
2575 | ||
15e47304 | 2576 | rep = __nlmsg_put(skb, NETLINK_CB(in_skb).portid, nlh->nlmsg_seq, |
5dba93ae | 2577 | NLMSG_ERROR, payload, 0); |
bf8b79e4 | 2578 | errmsg = nlmsg_data(rep); |
1da177e4 | 2579 | errmsg->error = err; |
bf8b79e4 | 2580 | memcpy(&errmsg->msg, nlh, err ? nlh->nlmsg_len : sizeof(*nlh)); |
15e47304 | 2581 | netlink_unicast(in_skb->sk, skb, NETLINK_CB(in_skb).portid, MSG_DONTWAIT); |
1da177e4 | 2582 | } |
6ac552fd | 2583 | EXPORT_SYMBOL(netlink_ack); |
1da177e4 | 2584 | |
cd40b7d3 | 2585 | int netlink_rcv_skb(struct sk_buff *skb, int (*cb)(struct sk_buff *, |
1d00a4eb | 2586 | struct nlmsghdr *)) |
82ace47a | 2587 | { |
82ace47a TG |
2588 | struct nlmsghdr *nlh; |
2589 | int err; | |
2590 | ||
2591 | while (skb->len >= nlmsg_total_size(0)) { | |
cd40b7d3 DL |
2592 | int msglen; |
2593 | ||
b529ccf2 | 2594 | nlh = nlmsg_hdr(skb); |
d35b6856 | 2595 | err = 0; |
82ace47a | 2596 | |
ad8e4b75 | 2597 | if (nlh->nlmsg_len < NLMSG_HDRLEN || skb->len < nlh->nlmsg_len) |
82ace47a TG |
2598 | return 0; |
2599 | ||
d35b6856 TG |
2600 | /* Only requests are handled by the kernel */ |
2601 | if (!(nlh->nlmsg_flags & NLM_F_REQUEST)) | |
5c58298c | 2602 | goto ack; |
45e7ae7f TG |
2603 | |
2604 | /* Skip control messages */ | |
2605 | if (nlh->nlmsg_type < NLMSG_MIN_TYPE) | |
5c58298c | 2606 | goto ack; |
d35b6856 | 2607 | |
1d00a4eb | 2608 | err = cb(skb, nlh); |
5c58298c DL |
2609 | if (err == -EINTR) |
2610 | goto skip; | |
2611 | ||
2612 | ack: | |
d35b6856 | 2613 | if (nlh->nlmsg_flags & NLM_F_ACK || err) |
82ace47a | 2614 | netlink_ack(skb, nlh, err); |
82ace47a | 2615 | |
5c58298c | 2616 | skip: |
6ac552fd | 2617 | msglen = NLMSG_ALIGN(nlh->nlmsg_len); |
cd40b7d3 DL |
2618 | if (msglen > skb->len) |
2619 | msglen = skb->len; | |
2620 | skb_pull(skb, msglen); | |
82ace47a TG |
2621 | } |
2622 | ||
2623 | return 0; | |
2624 | } | |
6ac552fd | 2625 | EXPORT_SYMBOL(netlink_rcv_skb); |
82ace47a | 2626 | |
d387f6ad TG |
2627 | /** |
2628 | * nlmsg_notify - send a notification netlink message | |
2629 | * @sk: netlink socket to use | |
2630 | * @skb: notification message | |
15e47304 | 2631 | * @portid: destination netlink portid for reports or 0 |
d387f6ad TG |
2632 | * @group: destination multicast group or 0 |
2633 | * @report: 1 to report back, 0 to disable | |
2634 | * @flags: allocation flags | |
2635 | */ | |
15e47304 | 2636 | int nlmsg_notify(struct sock *sk, struct sk_buff *skb, u32 portid, |
d387f6ad TG |
2637 | unsigned int group, int report, gfp_t flags) |
2638 | { | |
2639 | int err = 0; | |
2640 | ||
2641 | if (group) { | |
15e47304 | 2642 | int exclude_portid = 0; |
d387f6ad TG |
2643 | |
2644 | if (report) { | |
2645 | atomic_inc(&skb->users); | |
15e47304 | 2646 | exclude_portid = portid; |
d387f6ad TG |
2647 | } |
2648 | ||
1ce85fe4 PNA |
2649 | /* errors reported via destination sk->sk_err, but propagate |
2650 | * delivery errors if NETLINK_BROADCAST_ERROR flag is set */ | |
15e47304 | 2651 | err = nlmsg_multicast(sk, skb, exclude_portid, group, flags); |
d387f6ad TG |
2652 | } |
2653 | ||
1ce85fe4 PNA |
2654 | if (report) { |
2655 | int err2; | |
2656 | ||
15e47304 | 2657 | err2 = nlmsg_unicast(sk, skb, portid); |
1ce85fe4 PNA |
2658 | if (!err || err == -ESRCH) |
2659 | err = err2; | |
2660 | } | |
d387f6ad TG |
2661 | |
2662 | return err; | |
2663 | } | |
6ac552fd | 2664 | EXPORT_SYMBOL(nlmsg_notify); |
d387f6ad | 2665 | |
1da177e4 LT |
2666 | #ifdef CONFIG_PROC_FS |
2667 | struct nl_seq_iter { | |
e372c414 | 2668 | struct seq_net_private p; |
1da177e4 LT |
2669 | int link; |
2670 | int hash_idx; | |
2671 | }; | |
2672 | ||
2673 | static struct sock *netlink_seq_socket_idx(struct seq_file *seq, loff_t pos) | |
2674 | { | |
2675 | struct nl_seq_iter *iter = seq->private; | |
2676 | int i, j; | |
2677 | struct sock *s; | |
1da177e4 LT |
2678 | loff_t off = 0; |
2679 | ||
6ac552fd | 2680 | for (i = 0; i < MAX_LINKS; i++) { |
15e47304 | 2681 | struct nl_portid_hash *hash = &nl_table[i].hash; |
1da177e4 LT |
2682 | |
2683 | for (j = 0; j <= hash->mask; j++) { | |
b67bfe0d | 2684 | sk_for_each(s, &hash->table[j]) { |
1218854a | 2685 | if (sock_net(s) != seq_file_net(seq)) |
b4b51029 | 2686 | continue; |
1da177e4 LT |
2687 | if (off == pos) { |
2688 | iter->link = i; | |
2689 | iter->hash_idx = j; | |
2690 | return s; | |
2691 | } | |
2692 | ++off; | |
2693 | } | |
2694 | } | |
2695 | } | |
2696 | return NULL; | |
2697 | } | |
2698 | ||
2699 | static void *netlink_seq_start(struct seq_file *seq, loff_t *pos) | |
9a429c49 | 2700 | __acquires(nl_table_lock) |
1da177e4 LT |
2701 | { |
2702 | read_lock(&nl_table_lock); | |
2703 | return *pos ? netlink_seq_socket_idx(seq, *pos - 1) : SEQ_START_TOKEN; | |
2704 | } | |
2705 | ||
2706 | static void *netlink_seq_next(struct seq_file *seq, void *v, loff_t *pos) | |
2707 | { | |
2708 | struct sock *s; | |
2709 | struct nl_seq_iter *iter; | |
2710 | int i, j; | |
2711 | ||
2712 | ++*pos; | |
2713 | ||
2714 | if (v == SEQ_START_TOKEN) | |
2715 | return netlink_seq_socket_idx(seq, 0); | |
746fac4d | 2716 | |
b4b51029 EB |
2717 | iter = seq->private; |
2718 | s = v; | |
2719 | do { | |
2720 | s = sk_next(s); | |
1218854a | 2721 | } while (s && sock_net(s) != seq_file_net(seq)); |
1da177e4 LT |
2722 | if (s) |
2723 | return s; | |
2724 | ||
1da177e4 LT |
2725 | i = iter->link; |
2726 | j = iter->hash_idx + 1; | |
2727 | ||
2728 | do { | |
15e47304 | 2729 | struct nl_portid_hash *hash = &nl_table[i].hash; |
1da177e4 LT |
2730 | |
2731 | for (; j <= hash->mask; j++) { | |
2732 | s = sk_head(&hash->table[j]); | |
1218854a | 2733 | while (s && sock_net(s) != seq_file_net(seq)) |
b4b51029 | 2734 | s = sk_next(s); |
1da177e4 LT |
2735 | if (s) { |
2736 | iter->link = i; | |
2737 | iter->hash_idx = j; | |
2738 | return s; | |
2739 | } | |
2740 | } | |
2741 | ||
2742 | j = 0; | |
2743 | } while (++i < MAX_LINKS); | |
2744 | ||
2745 | return NULL; | |
2746 | } | |
2747 | ||
2748 | static void netlink_seq_stop(struct seq_file *seq, void *v) | |
9a429c49 | 2749 | __releases(nl_table_lock) |
1da177e4 LT |
2750 | { |
2751 | read_unlock(&nl_table_lock); | |
2752 | } | |
2753 | ||
2754 | ||
2755 | static int netlink_seq_show(struct seq_file *seq, void *v) | |
2756 | { | |
658cb354 | 2757 | if (v == SEQ_START_TOKEN) { |
1da177e4 LT |
2758 | seq_puts(seq, |
2759 | "sk Eth Pid Groups " | |
cf0aa4e0 | 2760 | "Rmem Wmem Dump Locks Drops Inode\n"); |
658cb354 | 2761 | } else { |
1da177e4 LT |
2762 | struct sock *s = v; |
2763 | struct netlink_sock *nlk = nlk_sk(s); | |
2764 | ||
9f1e0ad0 | 2765 | seq_printf(seq, "%pK %-3d %-6u %08x %-8d %-8d %pK %-8d %-8d %-8lu\n", |
1da177e4 LT |
2766 | s, |
2767 | s->sk_protocol, | |
15e47304 | 2768 | nlk->portid, |
513c2500 | 2769 | nlk->groups ? (u32)nlk->groups[0] : 0, |
31e6d363 ED |
2770 | sk_rmem_alloc_get(s), |
2771 | sk_wmem_alloc_get(s), | |
1da177e4 | 2772 | nlk->cb, |
38938bfe | 2773 | atomic_read(&s->sk_refcnt), |
cf0aa4e0 MY |
2774 | atomic_read(&s->sk_drops), |
2775 | sock_i_ino(s) | |
1da177e4 LT |
2776 | ); |
2777 | ||
2778 | } | |
2779 | return 0; | |
2780 | } | |
2781 | ||
56b3d975 | 2782 | static const struct seq_operations netlink_seq_ops = { |
1da177e4 LT |
2783 | .start = netlink_seq_start, |
2784 | .next = netlink_seq_next, | |
2785 | .stop = netlink_seq_stop, | |
2786 | .show = netlink_seq_show, | |
2787 | }; | |
2788 | ||
2789 | ||
2790 | static int netlink_seq_open(struct inode *inode, struct file *file) | |
2791 | { | |
e372c414 DL |
2792 | return seq_open_net(inode, file, &netlink_seq_ops, |
2793 | sizeof(struct nl_seq_iter)); | |
b4b51029 EB |
2794 | } |
2795 | ||
da7071d7 | 2796 | static const struct file_operations netlink_seq_fops = { |
1da177e4 LT |
2797 | .owner = THIS_MODULE, |
2798 | .open = netlink_seq_open, | |
2799 | .read = seq_read, | |
2800 | .llseek = seq_lseek, | |
e372c414 | 2801 | .release = seq_release_net, |
1da177e4 LT |
2802 | }; |
2803 | ||
2804 | #endif | |
2805 | ||
2806 | int netlink_register_notifier(struct notifier_block *nb) | |
2807 | { | |
e041c683 | 2808 | return atomic_notifier_chain_register(&netlink_chain, nb); |
1da177e4 | 2809 | } |
6ac552fd | 2810 | EXPORT_SYMBOL(netlink_register_notifier); |
1da177e4 LT |
2811 | |
2812 | int netlink_unregister_notifier(struct notifier_block *nb) | |
2813 | { | |
e041c683 | 2814 | return atomic_notifier_chain_unregister(&netlink_chain, nb); |
1da177e4 | 2815 | } |
6ac552fd | 2816 | EXPORT_SYMBOL(netlink_unregister_notifier); |
746fac4d | 2817 | |
90ddc4f0 | 2818 | static const struct proto_ops netlink_ops = { |
1da177e4 LT |
2819 | .family = PF_NETLINK, |
2820 | .owner = THIS_MODULE, | |
2821 | .release = netlink_release, | |
2822 | .bind = netlink_bind, | |
2823 | .connect = netlink_connect, | |
2824 | .socketpair = sock_no_socketpair, | |
2825 | .accept = sock_no_accept, | |
2826 | .getname = netlink_getname, | |
9652e931 | 2827 | .poll = netlink_poll, |
1da177e4 LT |
2828 | .ioctl = sock_no_ioctl, |
2829 | .listen = sock_no_listen, | |
2830 | .shutdown = sock_no_shutdown, | |
9a4595bc PM |
2831 | .setsockopt = netlink_setsockopt, |
2832 | .getsockopt = netlink_getsockopt, | |
1da177e4 LT |
2833 | .sendmsg = netlink_sendmsg, |
2834 | .recvmsg = netlink_recvmsg, | |
ccdfcc39 | 2835 | .mmap = netlink_mmap, |
1da177e4 LT |
2836 | .sendpage = sock_no_sendpage, |
2837 | }; | |
2838 | ||
ec1b4cf7 | 2839 | static const struct net_proto_family netlink_family_ops = { |
1da177e4 LT |
2840 | .family = PF_NETLINK, |
2841 | .create = netlink_create, | |
2842 | .owner = THIS_MODULE, /* for consistency 8) */ | |
2843 | }; | |
2844 | ||
4665079c | 2845 | static int __net_init netlink_net_init(struct net *net) |
b4b51029 EB |
2846 | { |
2847 | #ifdef CONFIG_PROC_FS | |
d4beaa66 | 2848 | if (!proc_create("netlink", 0, net->proc_net, &netlink_seq_fops)) |
b4b51029 EB |
2849 | return -ENOMEM; |
2850 | #endif | |
2851 | return 0; | |
2852 | } | |
2853 | ||
4665079c | 2854 | static void __net_exit netlink_net_exit(struct net *net) |
b4b51029 EB |
2855 | { |
2856 | #ifdef CONFIG_PROC_FS | |
ece31ffd | 2857 | remove_proc_entry("netlink", net->proc_net); |
b4b51029 EB |
2858 | #endif |
2859 | } | |
2860 | ||
b963ea89 DM |
2861 | static void __init netlink_add_usersock_entry(void) |
2862 | { | |
5c398dc8 | 2863 | struct listeners *listeners; |
b963ea89 DM |
2864 | int groups = 32; |
2865 | ||
5c398dc8 | 2866 | listeners = kzalloc(sizeof(*listeners) + NLGRPSZ(groups), GFP_KERNEL); |
b963ea89 | 2867 | if (!listeners) |
5c398dc8 | 2868 | panic("netlink_add_usersock_entry: Cannot allocate listeners\n"); |
b963ea89 DM |
2869 | |
2870 | netlink_table_grab(); | |
2871 | ||
2872 | nl_table[NETLINK_USERSOCK].groups = groups; | |
5c398dc8 | 2873 | rcu_assign_pointer(nl_table[NETLINK_USERSOCK].listeners, listeners); |
b963ea89 DM |
2874 | nl_table[NETLINK_USERSOCK].module = THIS_MODULE; |
2875 | nl_table[NETLINK_USERSOCK].registered = 1; | |
9785e10a | 2876 | nl_table[NETLINK_USERSOCK].flags = NL_CFG_F_NONROOT_SEND; |
b963ea89 DM |
2877 | |
2878 | netlink_table_ungrab(); | |
2879 | } | |
2880 | ||
022cbae6 | 2881 | static struct pernet_operations __net_initdata netlink_net_ops = { |
b4b51029 EB |
2882 | .init = netlink_net_init, |
2883 | .exit = netlink_net_exit, | |
2884 | }; | |
2885 | ||
1da177e4 LT |
2886 | static int __init netlink_proto_init(void) |
2887 | { | |
1da177e4 | 2888 | int i; |
26ff5ddc | 2889 | unsigned long limit; |
1da177e4 LT |
2890 | unsigned int order; |
2891 | int err = proto_register(&netlink_proto, 0); | |
2892 | ||
2893 | if (err != 0) | |
2894 | goto out; | |
2895 | ||
fab25745 | 2896 | BUILD_BUG_ON(sizeof(struct netlink_skb_parms) > FIELD_SIZEOF(struct sk_buff, cb)); |
1da177e4 | 2897 | |
0da974f4 | 2898 | nl_table = kcalloc(MAX_LINKS, sizeof(*nl_table), GFP_KERNEL); |
fab2caf6 AM |
2899 | if (!nl_table) |
2900 | goto panic; | |
1da177e4 | 2901 | |
4481374c JB |
2902 | if (totalram_pages >= (128 * 1024)) |
2903 | limit = totalram_pages >> (21 - PAGE_SHIFT); | |
1da177e4 | 2904 | else |
4481374c | 2905 | limit = totalram_pages >> (23 - PAGE_SHIFT); |
1da177e4 | 2906 | |
26ff5ddc DC |
2907 | order = get_bitmask_order(limit) - 1 + PAGE_SHIFT; |
2908 | limit = (1UL << order) / sizeof(struct hlist_head); | |
2909 | order = get_bitmask_order(min(limit, (unsigned long)UINT_MAX)) - 1; | |
1da177e4 LT |
2910 | |
2911 | for (i = 0; i < MAX_LINKS; i++) { | |
15e47304 | 2912 | struct nl_portid_hash *hash = &nl_table[i].hash; |
1da177e4 | 2913 | |
15e47304 | 2914 | hash->table = nl_portid_hash_zalloc(1 * sizeof(*hash->table)); |
1da177e4 LT |
2915 | if (!hash->table) { |
2916 | while (i-- > 0) | |
15e47304 | 2917 | nl_portid_hash_free(nl_table[i].hash.table, |
1da177e4 LT |
2918 | 1 * sizeof(*hash->table)); |
2919 | kfree(nl_table); | |
fab2caf6 | 2920 | goto panic; |
1da177e4 | 2921 | } |
1da177e4 LT |
2922 | hash->max_shift = order; |
2923 | hash->shift = 0; | |
2924 | hash->mask = 0; | |
2925 | hash->rehash_time = jiffies; | |
2926 | } | |
2927 | ||
b963ea89 DM |
2928 | netlink_add_usersock_entry(); |
2929 | ||
1da177e4 | 2930 | sock_register(&netlink_family_ops); |
b4b51029 | 2931 | register_pernet_subsys(&netlink_net_ops); |
746fac4d | 2932 | /* The netlink device handler may be needed early. */ |
1da177e4 LT |
2933 | rtnetlink_init(); |
2934 | out: | |
2935 | return err; | |
fab2caf6 AM |
2936 | panic: |
2937 | panic("netlink_init: Cannot allocate nl_table\n"); | |
1da177e4 LT |
2938 | } |
2939 | ||
1da177e4 | 2940 | core_initcall(netlink_proto_init); |