]>
Commit | Line | Data |
---|---|---|
1da177e4 LT |
1 | /* |
2 | * NETLINK Kernel-user communication protocol. | |
3 | * | |
113aa838 | 4 | * Authors: Alan Cox <alan@lxorguk.ukuu.org.uk> |
1da177e4 | 5 | * Alexey Kuznetsov <kuznet@ms2.inr.ac.ru> |
cd1df525 | 6 | * Patrick McHardy <kaber@trash.net> |
1da177e4 LT |
7 | * |
8 | * This program is free software; you can redistribute it and/or | |
9 | * modify it under the terms of the GNU General Public License | |
10 | * as published by the Free Software Foundation; either version | |
11 | * 2 of the License, or (at your option) any later version. | |
746fac4d | 12 | * |
1da177e4 LT |
13 | * Tue Jun 26 14:36:48 MEST 2001 Herbert "herp" Rosmanith |
14 | * added netlink_proto_exit | |
15 | * Tue Jan 22 18:32:44 BRST 2002 Arnaldo C. de Melo <acme@conectiva.com.br> | |
16 | * use nlk_sk, as sk->protinfo is on a diet 8) | |
4fdb3bb7 HW |
17 | * Fri Jul 22 19:51:12 MEST 2005 Harald Welte <laforge@gnumonks.org> |
18 | * - inc module use count of module that owns | |
19 | * the kernel socket in case userspace opens | |
20 | * socket of same protocol | |
21 | * - remove all module support, since netlink is | |
22 | * mandatory if CONFIG_NET=y these days | |
1da177e4 LT |
23 | */ |
24 | ||
1da177e4 LT |
25 | #include <linux/module.h> |
26 | ||
4fc268d2 | 27 | #include <linux/capability.h> |
1da177e4 LT |
28 | #include <linux/kernel.h> |
29 | #include <linux/init.h> | |
1da177e4 LT |
30 | #include <linux/signal.h> |
31 | #include <linux/sched.h> | |
32 | #include <linux/errno.h> | |
33 | #include <linux/string.h> | |
34 | #include <linux/stat.h> | |
35 | #include <linux/socket.h> | |
36 | #include <linux/un.h> | |
37 | #include <linux/fcntl.h> | |
38 | #include <linux/termios.h> | |
39 | #include <linux/sockios.h> | |
40 | #include <linux/net.h> | |
41 | #include <linux/fs.h> | |
42 | #include <linux/slab.h> | |
43 | #include <asm/uaccess.h> | |
44 | #include <linux/skbuff.h> | |
45 | #include <linux/netdevice.h> | |
46 | #include <linux/rtnetlink.h> | |
47 | #include <linux/proc_fs.h> | |
48 | #include <linux/seq_file.h> | |
1da177e4 LT |
49 | #include <linux/notifier.h> |
50 | #include <linux/security.h> | |
51 | #include <linux/jhash.h> | |
52 | #include <linux/jiffies.h> | |
53 | #include <linux/random.h> | |
54 | #include <linux/bitops.h> | |
55 | #include <linux/mm.h> | |
56 | #include <linux/types.h> | |
54e0f520 | 57 | #include <linux/audit.h> |
af65bdfc | 58 | #include <linux/mutex.h> |
ccdfcc39 | 59 | #include <linux/vmalloc.h> |
bcbde0d4 | 60 | #include <linux/if_arp.h> |
9652e931 | 61 | #include <asm/cacheflush.h> |
54e0f520 | 62 | |
457c4cbc | 63 | #include <net/net_namespace.h> |
1da177e4 LT |
64 | #include <net/sock.h> |
65 | #include <net/scm.h> | |
82ace47a | 66 | #include <net/netlink.h> |
1da177e4 | 67 | |
0f29c768 | 68 | #include "af_netlink.h" |
1da177e4 | 69 | |
5c398dc8 ED |
70 | struct listeners { |
71 | struct rcu_head rcu; | |
72 | unsigned long masks[0]; | |
6c04bb18 JB |
73 | }; |
74 | ||
cd967e05 PM |
75 | /* state bits */ |
76 | #define NETLINK_CONGESTED 0x0 | |
77 | ||
78 | /* flags */ | |
77247bbb | 79 | #define NETLINK_KERNEL_SOCKET 0x1 |
9a4595bc | 80 | #define NETLINK_RECV_PKTINFO 0x2 |
be0c22a4 | 81 | #define NETLINK_BROADCAST_SEND_ERROR 0x4 |
38938bfe | 82 | #define NETLINK_RECV_NO_ENOBUFS 0x8 |
77247bbb | 83 | |
035c4c16 | 84 | static inline int netlink_is_kernel(struct sock *sk) |
aed81560 DL |
85 | { |
86 | return nlk_sk(sk)->flags & NETLINK_KERNEL_SOCKET; | |
87 | } | |
88 | ||
0f29c768 AV |
89 | struct netlink_table *nl_table; |
90 | EXPORT_SYMBOL_GPL(nl_table); | |
1da177e4 LT |
91 | |
92 | static DECLARE_WAIT_QUEUE_HEAD(nl_table_wait); | |
93 | ||
94 | static int netlink_dump(struct sock *sk); | |
9652e931 | 95 | static void netlink_skb_destructor(struct sk_buff *skb); |
1da177e4 | 96 | |
0f29c768 AV |
97 | DEFINE_RWLOCK(nl_table_lock); |
98 | EXPORT_SYMBOL_GPL(nl_table_lock); | |
1da177e4 LT |
99 | static atomic_t nl_table_users = ATOMIC_INIT(0); |
100 | ||
6d772ac5 ED |
101 | #define nl_deref_protected(X) rcu_dereference_protected(X, lockdep_is_held(&nl_table_lock)); |
102 | ||
e041c683 | 103 | static ATOMIC_NOTIFIER_HEAD(netlink_chain); |
1da177e4 | 104 | |
bcbde0d4 DB |
105 | static DEFINE_SPINLOCK(netlink_tap_lock); |
106 | static struct list_head netlink_tap_all __read_mostly; | |
107 | ||
b57ef81f | 108 | static inline u32 netlink_group_mask(u32 group) |
d629b836 PM |
109 | { |
110 | return group ? 1 << (group - 1) : 0; | |
111 | } | |
112 | ||
15e47304 | 113 | static inline struct hlist_head *nl_portid_hashfn(struct nl_portid_hash *hash, u32 portid) |
1da177e4 | 114 | { |
15e47304 | 115 | return &hash->table[jhash_1word(portid, hash->rnd) & hash->mask]; |
1da177e4 LT |
116 | } |
117 | ||
bcbde0d4 DB |
118 | int netlink_add_tap(struct netlink_tap *nt) |
119 | { | |
120 | if (unlikely(nt->dev->type != ARPHRD_NETLINK)) | |
121 | return -EINVAL; | |
122 | ||
123 | spin_lock(&netlink_tap_lock); | |
124 | list_add_rcu(&nt->list, &netlink_tap_all); | |
125 | spin_unlock(&netlink_tap_lock); | |
126 | ||
127 | if (nt->module) | |
128 | __module_get(nt->module); | |
129 | ||
130 | return 0; | |
131 | } | |
132 | EXPORT_SYMBOL_GPL(netlink_add_tap); | |
133 | ||
134 | int __netlink_remove_tap(struct netlink_tap *nt) | |
135 | { | |
136 | bool found = false; | |
137 | struct netlink_tap *tmp; | |
138 | ||
139 | spin_lock(&netlink_tap_lock); | |
140 | ||
141 | list_for_each_entry(tmp, &netlink_tap_all, list) { | |
142 | if (nt == tmp) { | |
143 | list_del_rcu(&nt->list); | |
144 | found = true; | |
145 | goto out; | |
146 | } | |
147 | } | |
148 | ||
149 | pr_warn("__netlink_remove_tap: %p not found\n", nt); | |
150 | out: | |
151 | spin_unlock(&netlink_tap_lock); | |
152 | ||
153 | if (found && nt->module) | |
154 | module_put(nt->module); | |
155 | ||
156 | return found ? 0 : -ENODEV; | |
157 | } | |
158 | EXPORT_SYMBOL_GPL(__netlink_remove_tap); | |
159 | ||
160 | int netlink_remove_tap(struct netlink_tap *nt) | |
161 | { | |
162 | int ret; | |
163 | ||
164 | ret = __netlink_remove_tap(nt); | |
165 | synchronize_net(); | |
166 | ||
167 | return ret; | |
168 | } | |
169 | EXPORT_SYMBOL_GPL(netlink_remove_tap); | |
170 | ||
171 | static int __netlink_deliver_tap_skb(struct sk_buff *skb, | |
172 | struct net_device *dev) | |
173 | { | |
174 | struct sk_buff *nskb; | |
175 | int ret = -ENOMEM; | |
176 | ||
177 | dev_hold(dev); | |
178 | nskb = skb_clone(skb, GFP_ATOMIC); | |
179 | if (nskb) { | |
180 | nskb->dev = dev; | |
181 | ret = dev_queue_xmit(nskb); | |
182 | if (unlikely(ret > 0)) | |
183 | ret = net_xmit_errno(ret); | |
184 | } | |
185 | ||
186 | dev_put(dev); | |
187 | return ret; | |
188 | } | |
189 | ||
190 | static void __netlink_deliver_tap(struct sk_buff *skb) | |
191 | { | |
192 | int ret; | |
193 | struct netlink_tap *tmp; | |
194 | ||
195 | list_for_each_entry_rcu(tmp, &netlink_tap_all, list) { | |
196 | ret = __netlink_deliver_tap_skb(skb, tmp->dev); | |
197 | if (unlikely(ret)) | |
198 | break; | |
199 | } | |
200 | } | |
201 | ||
202 | static void netlink_deliver_tap(struct sk_buff *skb) | |
203 | { | |
204 | rcu_read_lock(); | |
205 | ||
206 | if (unlikely(!list_empty(&netlink_tap_all))) | |
207 | __netlink_deliver_tap(skb); | |
208 | ||
209 | rcu_read_unlock(); | |
210 | } | |
211 | ||
cd1df525 PM |
212 | static void netlink_overrun(struct sock *sk) |
213 | { | |
214 | struct netlink_sock *nlk = nlk_sk(sk); | |
215 | ||
216 | if (!(nlk->flags & NETLINK_RECV_NO_ENOBUFS)) { | |
217 | if (!test_and_set_bit(NETLINK_CONGESTED, &nlk_sk(sk)->state)) { | |
218 | sk->sk_err = ENOBUFS; | |
219 | sk->sk_error_report(sk); | |
220 | } | |
221 | } | |
222 | atomic_inc(&sk->sk_drops); | |
223 | } | |
224 | ||
225 | static void netlink_rcv_wake(struct sock *sk) | |
226 | { | |
227 | struct netlink_sock *nlk = nlk_sk(sk); | |
228 | ||
229 | if (skb_queue_empty(&sk->sk_receive_queue)) | |
230 | clear_bit(NETLINK_CONGESTED, &nlk->state); | |
231 | if (!test_bit(NETLINK_CONGESTED, &nlk->state)) | |
232 | wake_up_interruptible(&nlk->wait); | |
233 | } | |
234 | ||
ccdfcc39 | 235 | #ifdef CONFIG_NETLINK_MMAP |
9652e931 PM |
236 | static bool netlink_skb_is_mmaped(const struct sk_buff *skb) |
237 | { | |
238 | return NETLINK_CB(skb).flags & NETLINK_SKB_MMAPED; | |
239 | } | |
240 | ||
f9c22888 PM |
241 | static bool netlink_rx_is_mmaped(struct sock *sk) |
242 | { | |
243 | return nlk_sk(sk)->rx_ring.pg_vec != NULL; | |
244 | } | |
245 | ||
5fd96123 PM |
246 | static bool netlink_tx_is_mmaped(struct sock *sk) |
247 | { | |
248 | return nlk_sk(sk)->tx_ring.pg_vec != NULL; | |
249 | } | |
250 | ||
ccdfcc39 PM |
251 | static __pure struct page *pgvec_to_page(const void *addr) |
252 | { | |
253 | if (is_vmalloc_addr(addr)) | |
254 | return vmalloc_to_page(addr); | |
255 | else | |
256 | return virt_to_page(addr); | |
257 | } | |
258 | ||
259 | static void free_pg_vec(void **pg_vec, unsigned int order, unsigned int len) | |
260 | { | |
261 | unsigned int i; | |
262 | ||
263 | for (i = 0; i < len; i++) { | |
264 | if (pg_vec[i] != NULL) { | |
265 | if (is_vmalloc_addr(pg_vec[i])) | |
266 | vfree(pg_vec[i]); | |
267 | else | |
268 | free_pages((unsigned long)pg_vec[i], order); | |
269 | } | |
270 | } | |
271 | kfree(pg_vec); | |
272 | } | |
273 | ||
274 | static void *alloc_one_pg_vec_page(unsigned long order) | |
275 | { | |
276 | void *buffer; | |
277 | gfp_t gfp_flags = GFP_KERNEL | __GFP_COMP | __GFP_ZERO | | |
278 | __GFP_NOWARN | __GFP_NORETRY; | |
279 | ||
280 | buffer = (void *)__get_free_pages(gfp_flags, order); | |
281 | if (buffer != NULL) | |
282 | return buffer; | |
283 | ||
284 | buffer = vzalloc((1 << order) * PAGE_SIZE); | |
285 | if (buffer != NULL) | |
286 | return buffer; | |
287 | ||
288 | gfp_flags &= ~__GFP_NORETRY; | |
289 | return (void *)__get_free_pages(gfp_flags, order); | |
290 | } | |
291 | ||
292 | static void **alloc_pg_vec(struct netlink_sock *nlk, | |
293 | struct nl_mmap_req *req, unsigned int order) | |
294 | { | |
295 | unsigned int block_nr = req->nm_block_nr; | |
296 | unsigned int i; | |
8a849bb7 | 297 | void **pg_vec; |
ccdfcc39 PM |
298 | |
299 | pg_vec = kcalloc(block_nr, sizeof(void *), GFP_KERNEL); | |
300 | if (pg_vec == NULL) | |
301 | return NULL; | |
302 | ||
303 | for (i = 0; i < block_nr; i++) { | |
8a849bb7 | 304 | pg_vec[i] = alloc_one_pg_vec_page(order); |
ccdfcc39 PM |
305 | if (pg_vec[i] == NULL) |
306 | goto err1; | |
307 | } | |
308 | ||
309 | return pg_vec; | |
310 | err1: | |
311 | free_pg_vec(pg_vec, order, block_nr); | |
312 | return NULL; | |
313 | } | |
314 | ||
315 | static int netlink_set_ring(struct sock *sk, struct nl_mmap_req *req, | |
316 | bool closing, bool tx_ring) | |
317 | { | |
318 | struct netlink_sock *nlk = nlk_sk(sk); | |
319 | struct netlink_ring *ring; | |
320 | struct sk_buff_head *queue; | |
321 | void **pg_vec = NULL; | |
322 | unsigned int order = 0; | |
323 | int err; | |
324 | ||
325 | ring = tx_ring ? &nlk->tx_ring : &nlk->rx_ring; | |
326 | queue = tx_ring ? &sk->sk_write_queue : &sk->sk_receive_queue; | |
327 | ||
328 | if (!closing) { | |
329 | if (atomic_read(&nlk->mapped)) | |
330 | return -EBUSY; | |
331 | if (atomic_read(&ring->pending)) | |
332 | return -EBUSY; | |
333 | } | |
334 | ||
335 | if (req->nm_block_nr) { | |
336 | if (ring->pg_vec != NULL) | |
337 | return -EBUSY; | |
338 | ||
339 | if ((int)req->nm_block_size <= 0) | |
340 | return -EINVAL; | |
341 | if (!IS_ALIGNED(req->nm_block_size, PAGE_SIZE)) | |
342 | return -EINVAL; | |
343 | if (req->nm_frame_size < NL_MMAP_HDRLEN) | |
344 | return -EINVAL; | |
345 | if (!IS_ALIGNED(req->nm_frame_size, NL_MMAP_MSG_ALIGNMENT)) | |
346 | return -EINVAL; | |
347 | ||
348 | ring->frames_per_block = req->nm_block_size / | |
349 | req->nm_frame_size; | |
350 | if (ring->frames_per_block == 0) | |
351 | return -EINVAL; | |
352 | if (ring->frames_per_block * req->nm_block_nr != | |
353 | req->nm_frame_nr) | |
354 | return -EINVAL; | |
355 | ||
356 | order = get_order(req->nm_block_size); | |
357 | pg_vec = alloc_pg_vec(nlk, req, order); | |
358 | if (pg_vec == NULL) | |
359 | return -ENOMEM; | |
360 | } else { | |
361 | if (req->nm_frame_nr) | |
362 | return -EINVAL; | |
363 | } | |
364 | ||
365 | err = -EBUSY; | |
366 | mutex_lock(&nlk->pg_vec_lock); | |
367 | if (closing || atomic_read(&nlk->mapped) == 0) { | |
368 | err = 0; | |
369 | spin_lock_bh(&queue->lock); | |
370 | ||
371 | ring->frame_max = req->nm_frame_nr - 1; | |
372 | ring->head = 0; | |
373 | ring->frame_size = req->nm_frame_size; | |
374 | ring->pg_vec_pages = req->nm_block_size / PAGE_SIZE; | |
375 | ||
376 | swap(ring->pg_vec_len, req->nm_block_nr); | |
377 | swap(ring->pg_vec_order, order); | |
378 | swap(ring->pg_vec, pg_vec); | |
379 | ||
380 | __skb_queue_purge(queue); | |
381 | spin_unlock_bh(&queue->lock); | |
382 | ||
383 | WARN_ON(atomic_read(&nlk->mapped)); | |
384 | } | |
385 | mutex_unlock(&nlk->pg_vec_lock); | |
386 | ||
387 | if (pg_vec) | |
388 | free_pg_vec(pg_vec, order, req->nm_block_nr); | |
389 | return err; | |
390 | } | |
391 | ||
392 | static void netlink_mm_open(struct vm_area_struct *vma) | |
393 | { | |
394 | struct file *file = vma->vm_file; | |
395 | struct socket *sock = file->private_data; | |
396 | struct sock *sk = sock->sk; | |
397 | ||
398 | if (sk) | |
399 | atomic_inc(&nlk_sk(sk)->mapped); | |
400 | } | |
401 | ||
402 | static void netlink_mm_close(struct vm_area_struct *vma) | |
403 | { | |
404 | struct file *file = vma->vm_file; | |
405 | struct socket *sock = file->private_data; | |
406 | struct sock *sk = sock->sk; | |
407 | ||
408 | if (sk) | |
409 | atomic_dec(&nlk_sk(sk)->mapped); | |
410 | } | |
411 | ||
412 | static const struct vm_operations_struct netlink_mmap_ops = { | |
413 | .open = netlink_mm_open, | |
414 | .close = netlink_mm_close, | |
415 | }; | |
416 | ||
417 | static int netlink_mmap(struct file *file, struct socket *sock, | |
418 | struct vm_area_struct *vma) | |
419 | { | |
420 | struct sock *sk = sock->sk; | |
421 | struct netlink_sock *nlk = nlk_sk(sk); | |
422 | struct netlink_ring *ring; | |
423 | unsigned long start, size, expected; | |
424 | unsigned int i; | |
425 | int err = -EINVAL; | |
426 | ||
427 | if (vma->vm_pgoff) | |
428 | return -EINVAL; | |
429 | ||
430 | mutex_lock(&nlk->pg_vec_lock); | |
431 | ||
432 | expected = 0; | |
433 | for (ring = &nlk->rx_ring; ring <= &nlk->tx_ring; ring++) { | |
434 | if (ring->pg_vec == NULL) | |
435 | continue; | |
436 | expected += ring->pg_vec_len * ring->pg_vec_pages * PAGE_SIZE; | |
437 | } | |
438 | ||
439 | if (expected == 0) | |
440 | goto out; | |
441 | ||
442 | size = vma->vm_end - vma->vm_start; | |
443 | if (size != expected) | |
444 | goto out; | |
445 | ||
446 | start = vma->vm_start; | |
447 | for (ring = &nlk->rx_ring; ring <= &nlk->tx_ring; ring++) { | |
448 | if (ring->pg_vec == NULL) | |
449 | continue; | |
450 | ||
451 | for (i = 0; i < ring->pg_vec_len; i++) { | |
452 | struct page *page; | |
453 | void *kaddr = ring->pg_vec[i]; | |
454 | unsigned int pg_num; | |
455 | ||
456 | for (pg_num = 0; pg_num < ring->pg_vec_pages; pg_num++) { | |
457 | page = pgvec_to_page(kaddr); | |
458 | err = vm_insert_page(vma, start, page); | |
459 | if (err < 0) | |
460 | goto out; | |
461 | start += PAGE_SIZE; | |
462 | kaddr += PAGE_SIZE; | |
463 | } | |
464 | } | |
465 | } | |
466 | ||
467 | atomic_inc(&nlk->mapped); | |
468 | vma->vm_ops = &netlink_mmap_ops; | |
469 | err = 0; | |
470 | out: | |
471 | mutex_unlock(&nlk->pg_vec_lock); | |
7cdbac71 | 472 | return err; |
ccdfcc39 | 473 | } |
9652e931 PM |
474 | |
475 | static void netlink_frame_flush_dcache(const struct nl_mmap_hdr *hdr) | |
476 | { | |
477 | #if ARCH_IMPLEMENTS_FLUSH_DCACHE_PAGE == 1 | |
478 | struct page *p_start, *p_end; | |
479 | ||
480 | /* First page is flushed through netlink_{get,set}_status */ | |
481 | p_start = pgvec_to_page(hdr + PAGE_SIZE); | |
1d5085cb | 482 | p_end = pgvec_to_page((void *)hdr + NL_MMAP_HDRLEN + hdr->nm_len - 1); |
9652e931 PM |
483 | while (p_start <= p_end) { |
484 | flush_dcache_page(p_start); | |
485 | p_start++; | |
486 | } | |
487 | #endif | |
488 | } | |
489 | ||
490 | static enum nl_mmap_status netlink_get_status(const struct nl_mmap_hdr *hdr) | |
491 | { | |
492 | smp_rmb(); | |
493 | flush_dcache_page(pgvec_to_page(hdr)); | |
494 | return hdr->nm_status; | |
495 | } | |
496 | ||
497 | static void netlink_set_status(struct nl_mmap_hdr *hdr, | |
498 | enum nl_mmap_status status) | |
499 | { | |
500 | hdr->nm_status = status; | |
501 | flush_dcache_page(pgvec_to_page(hdr)); | |
502 | smp_wmb(); | |
503 | } | |
504 | ||
505 | static struct nl_mmap_hdr * | |
506 | __netlink_lookup_frame(const struct netlink_ring *ring, unsigned int pos) | |
507 | { | |
508 | unsigned int pg_vec_pos, frame_off; | |
509 | ||
510 | pg_vec_pos = pos / ring->frames_per_block; | |
511 | frame_off = pos % ring->frames_per_block; | |
512 | ||
513 | return ring->pg_vec[pg_vec_pos] + (frame_off * ring->frame_size); | |
514 | } | |
515 | ||
516 | static struct nl_mmap_hdr * | |
517 | netlink_lookup_frame(const struct netlink_ring *ring, unsigned int pos, | |
518 | enum nl_mmap_status status) | |
519 | { | |
520 | struct nl_mmap_hdr *hdr; | |
521 | ||
522 | hdr = __netlink_lookup_frame(ring, pos); | |
523 | if (netlink_get_status(hdr) != status) | |
524 | return NULL; | |
525 | ||
526 | return hdr; | |
527 | } | |
528 | ||
529 | static struct nl_mmap_hdr * | |
530 | netlink_current_frame(const struct netlink_ring *ring, | |
531 | enum nl_mmap_status status) | |
532 | { | |
533 | return netlink_lookup_frame(ring, ring->head, status); | |
534 | } | |
535 | ||
536 | static struct nl_mmap_hdr * | |
537 | netlink_previous_frame(const struct netlink_ring *ring, | |
538 | enum nl_mmap_status status) | |
539 | { | |
540 | unsigned int prev; | |
541 | ||
542 | prev = ring->head ? ring->head - 1 : ring->frame_max; | |
543 | return netlink_lookup_frame(ring, prev, status); | |
544 | } | |
545 | ||
546 | static void netlink_increment_head(struct netlink_ring *ring) | |
547 | { | |
548 | ring->head = ring->head != ring->frame_max ? ring->head + 1 : 0; | |
549 | } | |
550 | ||
551 | static void netlink_forward_ring(struct netlink_ring *ring) | |
552 | { | |
553 | unsigned int head = ring->head, pos = head; | |
554 | const struct nl_mmap_hdr *hdr; | |
555 | ||
556 | do { | |
557 | hdr = __netlink_lookup_frame(ring, pos); | |
558 | if (hdr->nm_status == NL_MMAP_STATUS_UNUSED) | |
559 | break; | |
560 | if (hdr->nm_status != NL_MMAP_STATUS_SKIP) | |
561 | break; | |
562 | netlink_increment_head(ring); | |
563 | } while (ring->head != head); | |
564 | } | |
565 | ||
cd1df525 PM |
566 | static bool netlink_dump_space(struct netlink_sock *nlk) |
567 | { | |
568 | struct netlink_ring *ring = &nlk->rx_ring; | |
569 | struct nl_mmap_hdr *hdr; | |
570 | unsigned int n; | |
571 | ||
572 | hdr = netlink_current_frame(ring, NL_MMAP_STATUS_UNUSED); | |
573 | if (hdr == NULL) | |
574 | return false; | |
575 | ||
576 | n = ring->head + ring->frame_max / 2; | |
577 | if (n > ring->frame_max) | |
578 | n -= ring->frame_max; | |
579 | ||
580 | hdr = __netlink_lookup_frame(ring, n); | |
581 | ||
582 | return hdr->nm_status == NL_MMAP_STATUS_UNUSED; | |
583 | } | |
584 | ||
9652e931 PM |
585 | static unsigned int netlink_poll(struct file *file, struct socket *sock, |
586 | poll_table *wait) | |
587 | { | |
588 | struct sock *sk = sock->sk; | |
589 | struct netlink_sock *nlk = nlk_sk(sk); | |
590 | unsigned int mask; | |
cd1df525 | 591 | int err; |
9652e931 | 592 | |
cd1df525 PM |
593 | if (nlk->rx_ring.pg_vec != NULL) { |
594 | /* Memory mapped sockets don't call recvmsg(), so flow control | |
595 | * for dumps is performed here. A dump is allowed to continue | |
596 | * if at least half the ring is unused. | |
597 | */ | |
16b304f3 | 598 | while (nlk->cb_running && netlink_dump_space(nlk)) { |
cd1df525 PM |
599 | err = netlink_dump(sk); |
600 | if (err < 0) { | |
601 | sk->sk_err = err; | |
602 | sk->sk_error_report(sk); | |
603 | break; | |
604 | } | |
605 | } | |
606 | netlink_rcv_wake(sk); | |
607 | } | |
5fd96123 | 608 | |
9652e931 PM |
609 | mask = datagram_poll(file, sock, wait); |
610 | ||
611 | spin_lock_bh(&sk->sk_receive_queue.lock); | |
612 | if (nlk->rx_ring.pg_vec) { | |
613 | netlink_forward_ring(&nlk->rx_ring); | |
614 | if (!netlink_previous_frame(&nlk->rx_ring, NL_MMAP_STATUS_UNUSED)) | |
615 | mask |= POLLIN | POLLRDNORM; | |
616 | } | |
617 | spin_unlock_bh(&sk->sk_receive_queue.lock); | |
618 | ||
619 | spin_lock_bh(&sk->sk_write_queue.lock); | |
620 | if (nlk->tx_ring.pg_vec) { | |
621 | if (netlink_current_frame(&nlk->tx_ring, NL_MMAP_STATUS_UNUSED)) | |
622 | mask |= POLLOUT | POLLWRNORM; | |
623 | } | |
624 | spin_unlock_bh(&sk->sk_write_queue.lock); | |
625 | ||
626 | return mask; | |
627 | } | |
628 | ||
629 | static struct nl_mmap_hdr *netlink_mmap_hdr(struct sk_buff *skb) | |
630 | { | |
631 | return (struct nl_mmap_hdr *)(skb->head - NL_MMAP_HDRLEN); | |
632 | } | |
633 | ||
634 | static void netlink_ring_setup_skb(struct sk_buff *skb, struct sock *sk, | |
635 | struct netlink_ring *ring, | |
636 | struct nl_mmap_hdr *hdr) | |
637 | { | |
638 | unsigned int size; | |
639 | void *data; | |
640 | ||
641 | size = ring->frame_size - NL_MMAP_HDRLEN; | |
642 | data = (void *)hdr + NL_MMAP_HDRLEN; | |
643 | ||
644 | skb->head = data; | |
645 | skb->data = data; | |
646 | skb_reset_tail_pointer(skb); | |
647 | skb->end = skb->tail + size; | |
648 | skb->len = 0; | |
649 | ||
650 | skb->destructor = netlink_skb_destructor; | |
651 | NETLINK_CB(skb).flags |= NETLINK_SKB_MMAPED; | |
652 | NETLINK_CB(skb).sk = sk; | |
653 | } | |
5fd96123 PM |
654 | |
655 | static int netlink_mmap_sendmsg(struct sock *sk, struct msghdr *msg, | |
656 | u32 dst_portid, u32 dst_group, | |
657 | struct sock_iocb *siocb) | |
658 | { | |
659 | struct netlink_sock *nlk = nlk_sk(sk); | |
660 | struct netlink_ring *ring; | |
661 | struct nl_mmap_hdr *hdr; | |
662 | struct sk_buff *skb; | |
663 | unsigned int maxlen; | |
664 | bool excl = true; | |
665 | int err = 0, len = 0; | |
666 | ||
667 | /* Netlink messages are validated by the receiver before processing. | |
668 | * In order to avoid userspace changing the contents of the message | |
669 | * after validation, the socket and the ring may only be used by a | |
670 | * single process, otherwise we fall back to copying. | |
671 | */ | |
672 | if (atomic_long_read(&sk->sk_socket->file->f_count) > 2 || | |
673 | atomic_read(&nlk->mapped) > 1) | |
674 | excl = false; | |
675 | ||
676 | mutex_lock(&nlk->pg_vec_lock); | |
677 | ||
678 | ring = &nlk->tx_ring; | |
679 | maxlen = ring->frame_size - NL_MMAP_HDRLEN; | |
680 | ||
681 | do { | |
682 | hdr = netlink_current_frame(ring, NL_MMAP_STATUS_VALID); | |
683 | if (hdr == NULL) { | |
684 | if (!(msg->msg_flags & MSG_DONTWAIT) && | |
685 | atomic_read(&nlk->tx_ring.pending)) | |
686 | schedule(); | |
687 | continue; | |
688 | } | |
689 | if (hdr->nm_len > maxlen) { | |
690 | err = -EINVAL; | |
691 | goto out; | |
692 | } | |
693 | ||
694 | netlink_frame_flush_dcache(hdr); | |
695 | ||
696 | if (likely(dst_portid == 0 && dst_group == 0 && excl)) { | |
697 | skb = alloc_skb_head(GFP_KERNEL); | |
698 | if (skb == NULL) { | |
699 | err = -ENOBUFS; | |
700 | goto out; | |
701 | } | |
702 | sock_hold(sk); | |
703 | netlink_ring_setup_skb(skb, sk, ring, hdr); | |
704 | NETLINK_CB(skb).flags |= NETLINK_SKB_TX; | |
705 | __skb_put(skb, hdr->nm_len); | |
706 | netlink_set_status(hdr, NL_MMAP_STATUS_RESERVED); | |
707 | atomic_inc(&ring->pending); | |
708 | } else { | |
709 | skb = alloc_skb(hdr->nm_len, GFP_KERNEL); | |
710 | if (skb == NULL) { | |
711 | err = -ENOBUFS; | |
712 | goto out; | |
713 | } | |
714 | __skb_put(skb, hdr->nm_len); | |
715 | memcpy(skb->data, (void *)hdr + NL_MMAP_HDRLEN, hdr->nm_len); | |
716 | netlink_set_status(hdr, NL_MMAP_STATUS_UNUSED); | |
717 | } | |
718 | ||
719 | netlink_increment_head(ring); | |
720 | ||
721 | NETLINK_CB(skb).portid = nlk->portid; | |
722 | NETLINK_CB(skb).dst_group = dst_group; | |
723 | NETLINK_CB(skb).creds = siocb->scm->creds; | |
724 | ||
725 | err = security_netlink_send(sk, skb); | |
726 | if (err) { | |
727 | kfree_skb(skb); | |
728 | goto out; | |
729 | } | |
730 | ||
731 | if (unlikely(dst_group)) { | |
732 | atomic_inc(&skb->users); | |
733 | netlink_broadcast(sk, skb, dst_portid, dst_group, | |
734 | GFP_KERNEL); | |
735 | } | |
736 | err = netlink_unicast(sk, skb, dst_portid, | |
737 | msg->msg_flags & MSG_DONTWAIT); | |
738 | if (err < 0) | |
739 | goto out; | |
740 | len += err; | |
741 | ||
742 | } while (hdr != NULL || | |
743 | (!(msg->msg_flags & MSG_DONTWAIT) && | |
744 | atomic_read(&nlk->tx_ring.pending))); | |
745 | ||
746 | if (len > 0) | |
747 | err = len; | |
748 | out: | |
749 | mutex_unlock(&nlk->pg_vec_lock); | |
750 | return err; | |
751 | } | |
f9c22888 PM |
752 | |
753 | static void netlink_queue_mmaped_skb(struct sock *sk, struct sk_buff *skb) | |
754 | { | |
755 | struct nl_mmap_hdr *hdr; | |
756 | ||
757 | hdr = netlink_mmap_hdr(skb); | |
758 | hdr->nm_len = skb->len; | |
759 | hdr->nm_group = NETLINK_CB(skb).dst_group; | |
760 | hdr->nm_pid = NETLINK_CB(skb).creds.pid; | |
1bf9310a ND |
761 | hdr->nm_uid = from_kuid(sk_user_ns(sk), NETLINK_CB(skb).creds.uid); |
762 | hdr->nm_gid = from_kgid(sk_user_ns(sk), NETLINK_CB(skb).creds.gid); | |
f9c22888 PM |
763 | netlink_frame_flush_dcache(hdr); |
764 | netlink_set_status(hdr, NL_MMAP_STATUS_VALID); | |
765 | ||
766 | NETLINK_CB(skb).flags |= NETLINK_SKB_DELIVERED; | |
767 | kfree_skb(skb); | |
768 | } | |
769 | ||
770 | static void netlink_ring_set_copied(struct sock *sk, struct sk_buff *skb) | |
771 | { | |
772 | struct netlink_sock *nlk = nlk_sk(sk); | |
773 | struct netlink_ring *ring = &nlk->rx_ring; | |
774 | struct nl_mmap_hdr *hdr; | |
775 | ||
776 | spin_lock_bh(&sk->sk_receive_queue.lock); | |
777 | hdr = netlink_current_frame(ring, NL_MMAP_STATUS_UNUSED); | |
778 | if (hdr == NULL) { | |
779 | spin_unlock_bh(&sk->sk_receive_queue.lock); | |
780 | kfree_skb(skb); | |
cd1df525 | 781 | netlink_overrun(sk); |
f9c22888 PM |
782 | return; |
783 | } | |
784 | netlink_increment_head(ring); | |
785 | __skb_queue_tail(&sk->sk_receive_queue, skb); | |
786 | spin_unlock_bh(&sk->sk_receive_queue.lock); | |
787 | ||
788 | hdr->nm_len = skb->len; | |
789 | hdr->nm_group = NETLINK_CB(skb).dst_group; | |
790 | hdr->nm_pid = NETLINK_CB(skb).creds.pid; | |
1bf9310a ND |
791 | hdr->nm_uid = from_kuid(sk_user_ns(sk), NETLINK_CB(skb).creds.uid); |
792 | hdr->nm_gid = from_kgid(sk_user_ns(sk), NETLINK_CB(skb).creds.gid); | |
f9c22888 PM |
793 | netlink_set_status(hdr, NL_MMAP_STATUS_COPY); |
794 | } | |
795 | ||
ccdfcc39 | 796 | #else /* CONFIG_NETLINK_MMAP */ |
9652e931 | 797 | #define netlink_skb_is_mmaped(skb) false |
f9c22888 | 798 | #define netlink_rx_is_mmaped(sk) false |
5fd96123 | 799 | #define netlink_tx_is_mmaped(sk) false |
ccdfcc39 | 800 | #define netlink_mmap sock_no_mmap |
9652e931 | 801 | #define netlink_poll datagram_poll |
5fd96123 | 802 | #define netlink_mmap_sendmsg(sk, msg, dst_portid, dst_group, siocb) 0 |
ccdfcc39 PM |
803 | #endif /* CONFIG_NETLINK_MMAP */ |
804 | ||
cf0a018a PM |
805 | static void netlink_skb_destructor(struct sk_buff *skb) |
806 | { | |
9652e931 PM |
807 | #ifdef CONFIG_NETLINK_MMAP |
808 | struct nl_mmap_hdr *hdr; | |
809 | struct netlink_ring *ring; | |
810 | struct sock *sk; | |
811 | ||
812 | /* If a packet from the kernel to userspace was freed because of an | |
813 | * error without being delivered to userspace, the kernel must reset | |
814 | * the status. In the direction userspace to kernel, the status is | |
815 | * always reset here after the packet was processed and freed. | |
816 | */ | |
817 | if (netlink_skb_is_mmaped(skb)) { | |
818 | hdr = netlink_mmap_hdr(skb); | |
819 | sk = NETLINK_CB(skb).sk; | |
820 | ||
5fd96123 PM |
821 | if (NETLINK_CB(skb).flags & NETLINK_SKB_TX) { |
822 | netlink_set_status(hdr, NL_MMAP_STATUS_UNUSED); | |
823 | ring = &nlk_sk(sk)->tx_ring; | |
824 | } else { | |
825 | if (!(NETLINK_CB(skb).flags & NETLINK_SKB_DELIVERED)) { | |
826 | hdr->nm_len = 0; | |
827 | netlink_set_status(hdr, NL_MMAP_STATUS_VALID); | |
828 | } | |
829 | ring = &nlk_sk(sk)->rx_ring; | |
9652e931 | 830 | } |
9652e931 PM |
831 | |
832 | WARN_ON(atomic_read(&ring->pending) == 0); | |
833 | atomic_dec(&ring->pending); | |
834 | sock_put(sk); | |
835 | ||
5e71d9d7 | 836 | skb->head = NULL; |
9652e931 PM |
837 | } |
838 | #endif | |
c05cdb1b | 839 | if (is_vmalloc_addr(skb->head)) { |
3a36515f PN |
840 | if (!skb->cloned || |
841 | !atomic_dec_return(&(skb_shinfo(skb)->dataref))) | |
842 | vfree(skb->head); | |
843 | ||
c05cdb1b PNA |
844 | skb->head = NULL; |
845 | } | |
9652e931 PM |
846 | if (skb->sk != NULL) |
847 | sock_rfree(skb); | |
cf0a018a PM |
848 | } |
849 | ||
850 | static void netlink_skb_set_owner_r(struct sk_buff *skb, struct sock *sk) | |
851 | { | |
852 | WARN_ON(skb->sk != NULL); | |
853 | skb->sk = sk; | |
854 | skb->destructor = netlink_skb_destructor; | |
855 | atomic_add(skb->truesize, &sk->sk_rmem_alloc); | |
856 | sk_mem_charge(sk, skb->truesize); | |
857 | } | |
858 | ||
1da177e4 LT |
859 | static void netlink_sock_destruct(struct sock *sk) |
860 | { | |
3f660d66 HX |
861 | struct netlink_sock *nlk = nlk_sk(sk); |
862 | ||
16b304f3 PS |
863 | if (nlk->cb_running) { |
864 | if (nlk->cb.done) | |
865 | nlk->cb.done(&nlk->cb); | |
6dc878a8 | 866 | |
16b304f3 PS |
867 | module_put(nlk->cb.module); |
868 | kfree_skb(nlk->cb.skb); | |
3f660d66 HX |
869 | } |
870 | ||
1da177e4 | 871 | skb_queue_purge(&sk->sk_receive_queue); |
ccdfcc39 PM |
872 | #ifdef CONFIG_NETLINK_MMAP |
873 | if (1) { | |
874 | struct nl_mmap_req req; | |
875 | ||
876 | memset(&req, 0, sizeof(req)); | |
877 | if (nlk->rx_ring.pg_vec) | |
878 | netlink_set_ring(sk, &req, true, false); | |
879 | memset(&req, 0, sizeof(req)); | |
880 | if (nlk->tx_ring.pg_vec) | |
881 | netlink_set_ring(sk, &req, true, true); | |
882 | } | |
883 | #endif /* CONFIG_NETLINK_MMAP */ | |
1da177e4 LT |
884 | |
885 | if (!sock_flag(sk, SOCK_DEAD)) { | |
6ac552fd | 886 | printk(KERN_ERR "Freeing alive netlink socket %p\n", sk); |
1da177e4 LT |
887 | return; |
888 | } | |
547b792c IJ |
889 | |
890 | WARN_ON(atomic_read(&sk->sk_rmem_alloc)); | |
891 | WARN_ON(atomic_read(&sk->sk_wmem_alloc)); | |
892 | WARN_ON(nlk_sk(sk)->groups); | |
1da177e4 LT |
893 | } |
894 | ||
6ac552fd PM |
895 | /* This lock without WQ_FLAG_EXCLUSIVE is good on UP and it is _very_ bad on |
896 | * SMP. Look, when several writers sleep and reader wakes them up, all but one | |
1da177e4 LT |
897 | * immediately hit write lock and grab all the cpus. Exclusive sleep solves |
898 | * this, _but_ remember, it adds useless work on UP machines. | |
899 | */ | |
900 | ||
d136f1bd | 901 | void netlink_table_grab(void) |
9a429c49 | 902 | __acquires(nl_table_lock) |
1da177e4 | 903 | { |
d136f1bd JB |
904 | might_sleep(); |
905 | ||
6abd219c | 906 | write_lock_irq(&nl_table_lock); |
1da177e4 LT |
907 | |
908 | if (atomic_read(&nl_table_users)) { | |
909 | DECLARE_WAITQUEUE(wait, current); | |
910 | ||
911 | add_wait_queue_exclusive(&nl_table_wait, &wait); | |
6ac552fd | 912 | for (;;) { |
1da177e4 LT |
913 | set_current_state(TASK_UNINTERRUPTIBLE); |
914 | if (atomic_read(&nl_table_users) == 0) | |
915 | break; | |
6abd219c | 916 | write_unlock_irq(&nl_table_lock); |
1da177e4 | 917 | schedule(); |
6abd219c | 918 | write_lock_irq(&nl_table_lock); |
1da177e4 LT |
919 | } |
920 | ||
921 | __set_current_state(TASK_RUNNING); | |
922 | remove_wait_queue(&nl_table_wait, &wait); | |
923 | } | |
924 | } | |
925 | ||
d136f1bd | 926 | void netlink_table_ungrab(void) |
9a429c49 | 927 | __releases(nl_table_lock) |
1da177e4 | 928 | { |
6abd219c | 929 | write_unlock_irq(&nl_table_lock); |
1da177e4 LT |
930 | wake_up(&nl_table_wait); |
931 | } | |
932 | ||
6ac552fd | 933 | static inline void |
1da177e4 LT |
934 | netlink_lock_table(void) |
935 | { | |
936 | /* read_lock() synchronizes us to netlink_table_grab */ | |
937 | ||
938 | read_lock(&nl_table_lock); | |
939 | atomic_inc(&nl_table_users); | |
940 | read_unlock(&nl_table_lock); | |
941 | } | |
942 | ||
6ac552fd | 943 | static inline void |
1da177e4 LT |
944 | netlink_unlock_table(void) |
945 | { | |
946 | if (atomic_dec_and_test(&nl_table_users)) | |
947 | wake_up(&nl_table_wait); | |
948 | } | |
949 | ||
da12c90e G |
950 | static bool netlink_compare(struct net *net, struct sock *sk) |
951 | { | |
952 | return net_eq(sock_net(sk), net); | |
953 | } | |
954 | ||
15e47304 | 955 | static struct sock *netlink_lookup(struct net *net, int protocol, u32 portid) |
1da177e4 | 956 | { |
da12c90e G |
957 | struct netlink_table *table = &nl_table[protocol]; |
958 | struct nl_portid_hash *hash = &table->hash; | |
1da177e4 LT |
959 | struct hlist_head *head; |
960 | struct sock *sk; | |
1da177e4 LT |
961 | |
962 | read_lock(&nl_table_lock); | |
15e47304 | 963 | head = nl_portid_hashfn(hash, portid); |
b67bfe0d | 964 | sk_for_each(sk, head) { |
da12c90e G |
965 | if (table->compare(net, sk) && |
966 | (nlk_sk(sk)->portid == portid)) { | |
1da177e4 LT |
967 | sock_hold(sk); |
968 | goto found; | |
969 | } | |
970 | } | |
971 | sk = NULL; | |
972 | found: | |
973 | read_unlock(&nl_table_lock); | |
974 | return sk; | |
975 | } | |
976 | ||
15e47304 | 977 | static struct hlist_head *nl_portid_hash_zalloc(size_t size) |
1da177e4 LT |
978 | { |
979 | if (size <= PAGE_SIZE) | |
ea72912c | 980 | return kzalloc(size, GFP_ATOMIC); |
1da177e4 LT |
981 | else |
982 | return (struct hlist_head *) | |
ea72912c ED |
983 | __get_free_pages(GFP_ATOMIC | __GFP_ZERO, |
984 | get_order(size)); | |
1da177e4 LT |
985 | } |
986 | ||
15e47304 | 987 | static void nl_portid_hash_free(struct hlist_head *table, size_t size) |
1da177e4 LT |
988 | { |
989 | if (size <= PAGE_SIZE) | |
990 | kfree(table); | |
991 | else | |
992 | free_pages((unsigned long)table, get_order(size)); | |
993 | } | |
994 | ||
15e47304 | 995 | static int nl_portid_hash_rehash(struct nl_portid_hash *hash, int grow) |
1da177e4 LT |
996 | { |
997 | unsigned int omask, mask, shift; | |
998 | size_t osize, size; | |
999 | struct hlist_head *otable, *table; | |
1000 | int i; | |
1001 | ||
1002 | omask = mask = hash->mask; | |
1003 | osize = size = (mask + 1) * sizeof(*table); | |
1004 | shift = hash->shift; | |
1005 | ||
1006 | if (grow) { | |
1007 | if (++shift > hash->max_shift) | |
1008 | return 0; | |
1009 | mask = mask * 2 + 1; | |
1010 | size *= 2; | |
1011 | } | |
1012 | ||
15e47304 | 1013 | table = nl_portid_hash_zalloc(size); |
1da177e4 LT |
1014 | if (!table) |
1015 | return 0; | |
1016 | ||
1da177e4 LT |
1017 | otable = hash->table; |
1018 | hash->table = table; | |
1019 | hash->mask = mask; | |
1020 | hash->shift = shift; | |
1021 | get_random_bytes(&hash->rnd, sizeof(hash->rnd)); | |
1022 | ||
1023 | for (i = 0; i <= omask; i++) { | |
1024 | struct sock *sk; | |
b67bfe0d | 1025 | struct hlist_node *tmp; |
1da177e4 | 1026 | |
b67bfe0d | 1027 | sk_for_each_safe(sk, tmp, &otable[i]) |
15e47304 | 1028 | __sk_add_node(sk, nl_portid_hashfn(hash, nlk_sk(sk)->portid)); |
1da177e4 LT |
1029 | } |
1030 | ||
15e47304 | 1031 | nl_portid_hash_free(otable, osize); |
1da177e4 LT |
1032 | hash->rehash_time = jiffies + 10 * 60 * HZ; |
1033 | return 1; | |
1034 | } | |
1035 | ||
15e47304 | 1036 | static inline int nl_portid_hash_dilute(struct nl_portid_hash *hash, int len) |
1da177e4 LT |
1037 | { |
1038 | int avg = hash->entries >> hash->shift; | |
1039 | ||
15e47304 | 1040 | if (unlikely(avg > 1) && nl_portid_hash_rehash(hash, 1)) |
1da177e4 LT |
1041 | return 1; |
1042 | ||
1043 | if (unlikely(len > avg) && time_after(jiffies, hash->rehash_time)) { | |
15e47304 | 1044 | nl_portid_hash_rehash(hash, 0); |
1da177e4 LT |
1045 | return 1; |
1046 | } | |
1047 | ||
1048 | return 0; | |
1049 | } | |
1050 | ||
90ddc4f0 | 1051 | static const struct proto_ops netlink_ops; |
1da177e4 | 1052 | |
4277a083 PM |
1053 | static void |
1054 | netlink_update_listeners(struct sock *sk) | |
1055 | { | |
1056 | struct netlink_table *tbl = &nl_table[sk->sk_protocol]; | |
4277a083 PM |
1057 | unsigned long mask; |
1058 | unsigned int i; | |
6d772ac5 ED |
1059 | struct listeners *listeners; |
1060 | ||
1061 | listeners = nl_deref_protected(tbl->listeners); | |
1062 | if (!listeners) | |
1063 | return; | |
4277a083 | 1064 | |
b4ff4f04 | 1065 | for (i = 0; i < NLGRPLONGS(tbl->groups); i++) { |
4277a083 | 1066 | mask = 0; |
b67bfe0d | 1067 | sk_for_each_bound(sk, &tbl->mc_list) { |
b4ff4f04 JB |
1068 | if (i < NLGRPLONGS(nlk_sk(sk)->ngroups)) |
1069 | mask |= nlk_sk(sk)->groups[i]; | |
1070 | } | |
6d772ac5 | 1071 | listeners->masks[i] = mask; |
4277a083 PM |
1072 | } |
1073 | /* this function is only called with the netlink table "grabbed", which | |
1074 | * makes sure updates are visible before bind or setsockopt return. */ | |
1075 | } | |
1076 | ||
15e47304 | 1077 | static int netlink_insert(struct sock *sk, struct net *net, u32 portid) |
1da177e4 | 1078 | { |
da12c90e G |
1079 | struct netlink_table *table = &nl_table[sk->sk_protocol]; |
1080 | struct nl_portid_hash *hash = &table->hash; | |
1da177e4 LT |
1081 | struct hlist_head *head; |
1082 | int err = -EADDRINUSE; | |
1083 | struct sock *osk; | |
1da177e4 LT |
1084 | int len; |
1085 | ||
1086 | netlink_table_grab(); | |
15e47304 | 1087 | head = nl_portid_hashfn(hash, portid); |
1da177e4 | 1088 | len = 0; |
b67bfe0d | 1089 | sk_for_each(osk, head) { |
da12c90e G |
1090 | if (table->compare(net, osk) && |
1091 | (nlk_sk(osk)->portid == portid)) | |
1da177e4 LT |
1092 | break; |
1093 | len++; | |
1094 | } | |
b67bfe0d | 1095 | if (osk) |
1da177e4 LT |
1096 | goto err; |
1097 | ||
1098 | err = -EBUSY; | |
15e47304 | 1099 | if (nlk_sk(sk)->portid) |
1da177e4 LT |
1100 | goto err; |
1101 | ||
1102 | err = -ENOMEM; | |
1103 | if (BITS_PER_LONG > 32 && unlikely(hash->entries >= UINT_MAX)) | |
1104 | goto err; | |
1105 | ||
15e47304 EB |
1106 | if (len && nl_portid_hash_dilute(hash, len)) |
1107 | head = nl_portid_hashfn(hash, portid); | |
1da177e4 | 1108 | hash->entries++; |
15e47304 | 1109 | nlk_sk(sk)->portid = portid; |
1da177e4 LT |
1110 | sk_add_node(sk, head); |
1111 | err = 0; | |
1112 | ||
1113 | err: | |
1114 | netlink_table_ungrab(); | |
1115 | return err; | |
1116 | } | |
1117 | ||
1118 | static void netlink_remove(struct sock *sk) | |
1119 | { | |
1120 | netlink_table_grab(); | |
d470e3b4 DM |
1121 | if (sk_del_node_init(sk)) |
1122 | nl_table[sk->sk_protocol].hash.entries--; | |
f7fa9b10 | 1123 | if (nlk_sk(sk)->subscriptions) |
1da177e4 LT |
1124 | __sk_del_bind_node(sk); |
1125 | netlink_table_ungrab(); | |
1126 | } | |
1127 | ||
1128 | static struct proto netlink_proto = { | |
1129 | .name = "NETLINK", | |
1130 | .owner = THIS_MODULE, | |
1131 | .obj_size = sizeof(struct netlink_sock), | |
1132 | }; | |
1133 | ||
1b8d7ae4 EB |
1134 | static int __netlink_create(struct net *net, struct socket *sock, |
1135 | struct mutex *cb_mutex, int protocol) | |
1da177e4 LT |
1136 | { |
1137 | struct sock *sk; | |
1138 | struct netlink_sock *nlk; | |
ab33a171 PM |
1139 | |
1140 | sock->ops = &netlink_ops; | |
1141 | ||
6257ff21 | 1142 | sk = sk_alloc(net, PF_NETLINK, GFP_KERNEL, &netlink_proto); |
ab33a171 PM |
1143 | if (!sk) |
1144 | return -ENOMEM; | |
1145 | ||
1146 | sock_init_data(sock, sk); | |
1147 | ||
1148 | nlk = nlk_sk(sk); | |
658cb354 | 1149 | if (cb_mutex) { |
ffa4d721 | 1150 | nlk->cb_mutex = cb_mutex; |
658cb354 | 1151 | } else { |
ffa4d721 PM |
1152 | nlk->cb_mutex = &nlk->cb_def_mutex; |
1153 | mutex_init(nlk->cb_mutex); | |
1154 | } | |
ab33a171 | 1155 | init_waitqueue_head(&nlk->wait); |
ccdfcc39 PM |
1156 | #ifdef CONFIG_NETLINK_MMAP |
1157 | mutex_init(&nlk->pg_vec_lock); | |
1158 | #endif | |
ab33a171 PM |
1159 | |
1160 | sk->sk_destruct = netlink_sock_destruct; | |
1161 | sk->sk_protocol = protocol; | |
1162 | return 0; | |
1163 | } | |
1164 | ||
3f378b68 EP |
1165 | static int netlink_create(struct net *net, struct socket *sock, int protocol, |
1166 | int kern) | |
ab33a171 PM |
1167 | { |
1168 | struct module *module = NULL; | |
af65bdfc | 1169 | struct mutex *cb_mutex; |
f7fa9b10 | 1170 | struct netlink_sock *nlk; |
03292745 | 1171 | void (*bind)(int group); |
ab33a171 | 1172 | int err = 0; |
1da177e4 LT |
1173 | |
1174 | sock->state = SS_UNCONNECTED; | |
1175 | ||
1176 | if (sock->type != SOCK_RAW && sock->type != SOCK_DGRAM) | |
1177 | return -ESOCKTNOSUPPORT; | |
1178 | ||
6ac552fd | 1179 | if (protocol < 0 || protocol >= MAX_LINKS) |
1da177e4 LT |
1180 | return -EPROTONOSUPPORT; |
1181 | ||
77247bbb | 1182 | netlink_lock_table(); |
95a5afca | 1183 | #ifdef CONFIG_MODULES |
ab33a171 | 1184 | if (!nl_table[protocol].registered) { |
77247bbb | 1185 | netlink_unlock_table(); |
4fdb3bb7 | 1186 | request_module("net-pf-%d-proto-%d", PF_NETLINK, protocol); |
77247bbb | 1187 | netlink_lock_table(); |
4fdb3bb7 | 1188 | } |
ab33a171 PM |
1189 | #endif |
1190 | if (nl_table[protocol].registered && | |
1191 | try_module_get(nl_table[protocol].module)) | |
1192 | module = nl_table[protocol].module; | |
974c37e9 AD |
1193 | else |
1194 | err = -EPROTONOSUPPORT; | |
af65bdfc | 1195 | cb_mutex = nl_table[protocol].cb_mutex; |
03292745 | 1196 | bind = nl_table[protocol].bind; |
77247bbb | 1197 | netlink_unlock_table(); |
4fdb3bb7 | 1198 | |
974c37e9 AD |
1199 | if (err < 0) |
1200 | goto out; | |
1201 | ||
6ac552fd PM |
1202 | err = __netlink_create(net, sock, cb_mutex, protocol); |
1203 | if (err < 0) | |
f7fa9b10 PM |
1204 | goto out_module; |
1205 | ||
6f756a8c | 1206 | local_bh_disable(); |
c1fd3b94 | 1207 | sock_prot_inuse_add(net, &netlink_proto, 1); |
6f756a8c DM |
1208 | local_bh_enable(); |
1209 | ||
f7fa9b10 | 1210 | nlk = nlk_sk(sock->sk); |
f7fa9b10 | 1211 | nlk->module = module; |
03292745 | 1212 | nlk->netlink_bind = bind; |
ab33a171 PM |
1213 | out: |
1214 | return err; | |
1da177e4 | 1215 | |
ab33a171 PM |
1216 | out_module: |
1217 | module_put(module); | |
1218 | goto out; | |
1da177e4 LT |
1219 | } |
1220 | ||
1221 | static int netlink_release(struct socket *sock) | |
1222 | { | |
1223 | struct sock *sk = sock->sk; | |
1224 | struct netlink_sock *nlk; | |
1225 | ||
1226 | if (!sk) | |
1227 | return 0; | |
1228 | ||
1229 | netlink_remove(sk); | |
ac57b3a9 | 1230 | sock_orphan(sk); |
1da177e4 LT |
1231 | nlk = nlk_sk(sk); |
1232 | ||
3f660d66 HX |
1233 | /* |
1234 | * OK. Socket is unlinked, any packets that arrive now | |
1235 | * will be purged. | |
1236 | */ | |
1da177e4 | 1237 | |
1da177e4 LT |
1238 | sock->sk = NULL; |
1239 | wake_up_interruptible_all(&nlk->wait); | |
1240 | ||
1241 | skb_queue_purge(&sk->sk_write_queue); | |
1242 | ||
15e47304 | 1243 | if (nlk->portid) { |
1da177e4 | 1244 | struct netlink_notify n = { |
3b1e0a65 | 1245 | .net = sock_net(sk), |
1da177e4 | 1246 | .protocol = sk->sk_protocol, |
15e47304 | 1247 | .portid = nlk->portid, |
1da177e4 | 1248 | }; |
e041c683 AS |
1249 | atomic_notifier_call_chain(&netlink_chain, |
1250 | NETLINK_URELEASE, &n); | |
746fac4d | 1251 | } |
4fdb3bb7 | 1252 | |
5e7c001c | 1253 | module_put(nlk->module); |
4fdb3bb7 | 1254 | |
4277a083 | 1255 | netlink_table_grab(); |
aed81560 | 1256 | if (netlink_is_kernel(sk)) { |
869e58f8 DL |
1257 | BUG_ON(nl_table[sk->sk_protocol].registered == 0); |
1258 | if (--nl_table[sk->sk_protocol].registered == 0) { | |
6d772ac5 ED |
1259 | struct listeners *old; |
1260 | ||
1261 | old = nl_deref_protected(nl_table[sk->sk_protocol].listeners); | |
1262 | RCU_INIT_POINTER(nl_table[sk->sk_protocol].listeners, NULL); | |
1263 | kfree_rcu(old, rcu); | |
869e58f8 | 1264 | nl_table[sk->sk_protocol].module = NULL; |
9785e10a PNA |
1265 | nl_table[sk->sk_protocol].bind = NULL; |
1266 | nl_table[sk->sk_protocol].flags = 0; | |
869e58f8 DL |
1267 | nl_table[sk->sk_protocol].registered = 0; |
1268 | } | |
658cb354 | 1269 | } else if (nlk->subscriptions) { |
4277a083 | 1270 | netlink_update_listeners(sk); |
658cb354 | 1271 | } |
4277a083 | 1272 | netlink_table_ungrab(); |
77247bbb | 1273 | |
f7fa9b10 PM |
1274 | kfree(nlk->groups); |
1275 | nlk->groups = NULL; | |
1276 | ||
3755810c | 1277 | local_bh_disable(); |
c1fd3b94 | 1278 | sock_prot_inuse_add(sock_net(sk), &netlink_proto, -1); |
3755810c | 1279 | local_bh_enable(); |
1da177e4 LT |
1280 | sock_put(sk); |
1281 | return 0; | |
1282 | } | |
1283 | ||
1284 | static int netlink_autobind(struct socket *sock) | |
1285 | { | |
1286 | struct sock *sk = sock->sk; | |
3b1e0a65 | 1287 | struct net *net = sock_net(sk); |
da12c90e G |
1288 | struct netlink_table *table = &nl_table[sk->sk_protocol]; |
1289 | struct nl_portid_hash *hash = &table->hash; | |
1da177e4 LT |
1290 | struct hlist_head *head; |
1291 | struct sock *osk; | |
15e47304 | 1292 | s32 portid = task_tgid_vnr(current); |
1da177e4 LT |
1293 | int err; |
1294 | static s32 rover = -4097; | |
1295 | ||
1296 | retry: | |
1297 | cond_resched(); | |
1298 | netlink_table_grab(); | |
15e47304 | 1299 | head = nl_portid_hashfn(hash, portid); |
b67bfe0d | 1300 | sk_for_each(osk, head) { |
da12c90e | 1301 | if (!table->compare(net, osk)) |
b4b51029 | 1302 | continue; |
15e47304 EB |
1303 | if (nlk_sk(osk)->portid == portid) { |
1304 | /* Bind collision, search negative portid values. */ | |
1305 | portid = rover--; | |
1da177e4 LT |
1306 | if (rover > -4097) |
1307 | rover = -4097; | |
1308 | netlink_table_ungrab(); | |
1309 | goto retry; | |
1310 | } | |
1311 | } | |
1312 | netlink_table_ungrab(); | |
1313 | ||
15e47304 | 1314 | err = netlink_insert(sk, net, portid); |
1da177e4 LT |
1315 | if (err == -EADDRINUSE) |
1316 | goto retry; | |
d470e3b4 DM |
1317 | |
1318 | /* If 2 threads race to autobind, that is fine. */ | |
1319 | if (err == -EBUSY) | |
1320 | err = 0; | |
1321 | ||
1322 | return err; | |
1da177e4 LT |
1323 | } |
1324 | ||
b57ef81f | 1325 | static inline int netlink_capable(const struct socket *sock, unsigned int flag) |
746fac4d | 1326 | { |
9785e10a | 1327 | return (nl_table[sock->sk->sk_protocol].flags & flag) || |
df008c91 | 1328 | ns_capable(sock_net(sock->sk)->user_ns, CAP_NET_ADMIN); |
746fac4d | 1329 | } |
1da177e4 | 1330 | |
f7fa9b10 PM |
1331 | static void |
1332 | netlink_update_subscriptions(struct sock *sk, unsigned int subscriptions) | |
1333 | { | |
1334 | struct netlink_sock *nlk = nlk_sk(sk); | |
1335 | ||
1336 | if (nlk->subscriptions && !subscriptions) | |
1337 | __sk_del_bind_node(sk); | |
1338 | else if (!nlk->subscriptions && subscriptions) | |
1339 | sk_add_bind_node(sk, &nl_table[sk->sk_protocol].mc_list); | |
1340 | nlk->subscriptions = subscriptions; | |
1341 | } | |
1342 | ||
b4ff4f04 | 1343 | static int netlink_realloc_groups(struct sock *sk) |
513c2500 PM |
1344 | { |
1345 | struct netlink_sock *nlk = nlk_sk(sk); | |
1346 | unsigned int groups; | |
b4ff4f04 | 1347 | unsigned long *new_groups; |
513c2500 PM |
1348 | int err = 0; |
1349 | ||
b4ff4f04 JB |
1350 | netlink_table_grab(); |
1351 | ||
513c2500 | 1352 | groups = nl_table[sk->sk_protocol].groups; |
b4ff4f04 | 1353 | if (!nl_table[sk->sk_protocol].registered) { |
513c2500 | 1354 | err = -ENOENT; |
b4ff4f04 JB |
1355 | goto out_unlock; |
1356 | } | |
513c2500 | 1357 | |
b4ff4f04 JB |
1358 | if (nlk->ngroups >= groups) |
1359 | goto out_unlock; | |
513c2500 | 1360 | |
b4ff4f04 JB |
1361 | new_groups = krealloc(nlk->groups, NLGRPSZ(groups), GFP_ATOMIC); |
1362 | if (new_groups == NULL) { | |
1363 | err = -ENOMEM; | |
1364 | goto out_unlock; | |
1365 | } | |
6ac552fd | 1366 | memset((char *)new_groups + NLGRPSZ(nlk->ngroups), 0, |
b4ff4f04 JB |
1367 | NLGRPSZ(groups) - NLGRPSZ(nlk->ngroups)); |
1368 | ||
1369 | nlk->groups = new_groups; | |
513c2500 | 1370 | nlk->ngroups = groups; |
b4ff4f04 JB |
1371 | out_unlock: |
1372 | netlink_table_ungrab(); | |
1373 | return err; | |
513c2500 PM |
1374 | } |
1375 | ||
6ac552fd PM |
1376 | static int netlink_bind(struct socket *sock, struct sockaddr *addr, |
1377 | int addr_len) | |
1da177e4 LT |
1378 | { |
1379 | struct sock *sk = sock->sk; | |
3b1e0a65 | 1380 | struct net *net = sock_net(sk); |
1da177e4 LT |
1381 | struct netlink_sock *nlk = nlk_sk(sk); |
1382 | struct sockaddr_nl *nladdr = (struct sockaddr_nl *)addr; | |
1383 | int err; | |
746fac4d | 1384 | |
4e4b5376 HFS |
1385 | if (addr_len < sizeof(struct sockaddr_nl)) |
1386 | return -EINVAL; | |
1387 | ||
1da177e4 LT |
1388 | if (nladdr->nl_family != AF_NETLINK) |
1389 | return -EINVAL; | |
1390 | ||
1391 | /* Only superuser is allowed to listen multicasts */ | |
513c2500 | 1392 | if (nladdr->nl_groups) { |
9785e10a | 1393 | if (!netlink_capable(sock, NL_CFG_F_NONROOT_RECV)) |
513c2500 | 1394 | return -EPERM; |
b4ff4f04 JB |
1395 | err = netlink_realloc_groups(sk); |
1396 | if (err) | |
1397 | return err; | |
513c2500 | 1398 | } |
1da177e4 | 1399 | |
15e47304 EB |
1400 | if (nlk->portid) { |
1401 | if (nladdr->nl_pid != nlk->portid) | |
1da177e4 LT |
1402 | return -EINVAL; |
1403 | } else { | |
1404 | err = nladdr->nl_pid ? | |
b4b51029 | 1405 | netlink_insert(sk, net, nladdr->nl_pid) : |
1da177e4 LT |
1406 | netlink_autobind(sock); |
1407 | if (err) | |
1408 | return err; | |
1409 | } | |
1410 | ||
513c2500 | 1411 | if (!nladdr->nl_groups && (nlk->groups == NULL || !(u32)nlk->groups[0])) |
1da177e4 LT |
1412 | return 0; |
1413 | ||
1414 | netlink_table_grab(); | |
f7fa9b10 | 1415 | netlink_update_subscriptions(sk, nlk->subscriptions + |
746fac4d YH |
1416 | hweight32(nladdr->nl_groups) - |
1417 | hweight32(nlk->groups[0])); | |
1418 | nlk->groups[0] = (nlk->groups[0] & ~0xffffffffUL) | nladdr->nl_groups; | |
4277a083 | 1419 | netlink_update_listeners(sk); |
1da177e4 LT |
1420 | netlink_table_ungrab(); |
1421 | ||
03292745 PNA |
1422 | if (nlk->netlink_bind && nlk->groups[0]) { |
1423 | int i; | |
1424 | ||
1425 | for (i=0; i<nlk->ngroups; i++) { | |
1426 | if (test_bit(i, nlk->groups)) | |
1427 | nlk->netlink_bind(i); | |
1428 | } | |
1429 | } | |
1430 | ||
1da177e4 LT |
1431 | return 0; |
1432 | } | |
1433 | ||
1434 | static int netlink_connect(struct socket *sock, struct sockaddr *addr, | |
1435 | int alen, int flags) | |
1436 | { | |
1437 | int err = 0; | |
1438 | struct sock *sk = sock->sk; | |
1439 | struct netlink_sock *nlk = nlk_sk(sk); | |
6ac552fd | 1440 | struct sockaddr_nl *nladdr = (struct sockaddr_nl *)addr; |
1da177e4 | 1441 | |
6503d961 CG |
1442 | if (alen < sizeof(addr->sa_family)) |
1443 | return -EINVAL; | |
1444 | ||
1da177e4 LT |
1445 | if (addr->sa_family == AF_UNSPEC) { |
1446 | sk->sk_state = NETLINK_UNCONNECTED; | |
15e47304 | 1447 | nlk->dst_portid = 0; |
d629b836 | 1448 | nlk->dst_group = 0; |
1da177e4 LT |
1449 | return 0; |
1450 | } | |
1451 | if (addr->sa_family != AF_NETLINK) | |
1452 | return -EINVAL; | |
1453 | ||
1454 | /* Only superuser is allowed to send multicasts */ | |
9785e10a | 1455 | if (nladdr->nl_groups && !netlink_capable(sock, NL_CFG_F_NONROOT_SEND)) |
1da177e4 LT |
1456 | return -EPERM; |
1457 | ||
15e47304 | 1458 | if (!nlk->portid) |
1da177e4 LT |
1459 | err = netlink_autobind(sock); |
1460 | ||
1461 | if (err == 0) { | |
1462 | sk->sk_state = NETLINK_CONNECTED; | |
15e47304 | 1463 | nlk->dst_portid = nladdr->nl_pid; |
d629b836 | 1464 | nlk->dst_group = ffs(nladdr->nl_groups); |
1da177e4 LT |
1465 | } |
1466 | ||
1467 | return err; | |
1468 | } | |
1469 | ||
6ac552fd PM |
1470 | static int netlink_getname(struct socket *sock, struct sockaddr *addr, |
1471 | int *addr_len, int peer) | |
1da177e4 LT |
1472 | { |
1473 | struct sock *sk = sock->sk; | |
1474 | struct netlink_sock *nlk = nlk_sk(sk); | |
13cfa97b | 1475 | DECLARE_SOCKADDR(struct sockaddr_nl *, nladdr, addr); |
746fac4d | 1476 | |
1da177e4 LT |
1477 | nladdr->nl_family = AF_NETLINK; |
1478 | nladdr->nl_pad = 0; | |
1479 | *addr_len = sizeof(*nladdr); | |
1480 | ||
1481 | if (peer) { | |
15e47304 | 1482 | nladdr->nl_pid = nlk->dst_portid; |
d629b836 | 1483 | nladdr->nl_groups = netlink_group_mask(nlk->dst_group); |
1da177e4 | 1484 | } else { |
15e47304 | 1485 | nladdr->nl_pid = nlk->portid; |
513c2500 | 1486 | nladdr->nl_groups = nlk->groups ? nlk->groups[0] : 0; |
1da177e4 LT |
1487 | } |
1488 | return 0; | |
1489 | } | |
1490 | ||
15e47304 | 1491 | static struct sock *netlink_getsockbyportid(struct sock *ssk, u32 portid) |
1da177e4 | 1492 | { |
1da177e4 LT |
1493 | struct sock *sock; |
1494 | struct netlink_sock *nlk; | |
1495 | ||
15e47304 | 1496 | sock = netlink_lookup(sock_net(ssk), ssk->sk_protocol, portid); |
1da177e4 LT |
1497 | if (!sock) |
1498 | return ERR_PTR(-ECONNREFUSED); | |
1499 | ||
1500 | /* Don't bother queuing skb if kernel socket has no input function */ | |
1501 | nlk = nlk_sk(sock); | |
cd40b7d3 | 1502 | if (sock->sk_state == NETLINK_CONNECTED && |
15e47304 | 1503 | nlk->dst_portid != nlk_sk(ssk)->portid) { |
1da177e4 LT |
1504 | sock_put(sock); |
1505 | return ERR_PTR(-ECONNREFUSED); | |
1506 | } | |
1507 | return sock; | |
1508 | } | |
1509 | ||
1510 | struct sock *netlink_getsockbyfilp(struct file *filp) | |
1511 | { | |
496ad9aa | 1512 | struct inode *inode = file_inode(filp); |
1da177e4 LT |
1513 | struct sock *sock; |
1514 | ||
1515 | if (!S_ISSOCK(inode->i_mode)) | |
1516 | return ERR_PTR(-ENOTSOCK); | |
1517 | ||
1518 | sock = SOCKET_I(inode)->sk; | |
1519 | if (sock->sk_family != AF_NETLINK) | |
1520 | return ERR_PTR(-EINVAL); | |
1521 | ||
1522 | sock_hold(sock); | |
1523 | return sock; | |
1524 | } | |
1525 | ||
3a36515f PN |
1526 | static struct sk_buff *netlink_alloc_large_skb(unsigned int size, |
1527 | int broadcast) | |
c05cdb1b PNA |
1528 | { |
1529 | struct sk_buff *skb; | |
1530 | void *data; | |
1531 | ||
3a36515f | 1532 | if (size <= NLMSG_GOODSIZE || broadcast) |
c05cdb1b PNA |
1533 | return alloc_skb(size, GFP_KERNEL); |
1534 | ||
3a36515f PN |
1535 | size = SKB_DATA_ALIGN(size) + |
1536 | SKB_DATA_ALIGN(sizeof(struct skb_shared_info)); | |
c05cdb1b PNA |
1537 | |
1538 | data = vmalloc(size); | |
1539 | if (data == NULL) | |
3a36515f | 1540 | return NULL; |
c05cdb1b | 1541 | |
3a36515f PN |
1542 | skb = build_skb(data, size); |
1543 | if (skb == NULL) | |
1544 | vfree(data); | |
1545 | else { | |
1546 | skb->head_frag = 0; | |
1547 | skb->destructor = netlink_skb_destructor; | |
1548 | } | |
c05cdb1b PNA |
1549 | |
1550 | return skb; | |
c05cdb1b PNA |
1551 | } |
1552 | ||
1da177e4 LT |
1553 | /* |
1554 | * Attach a skb to a netlink socket. | |
1555 | * The caller must hold a reference to the destination socket. On error, the | |
1556 | * reference is dropped. The skb is not send to the destination, just all | |
1557 | * all error checks are performed and memory in the queue is reserved. | |
1558 | * Return values: | |
1559 | * < 0: error. skb freed, reference to sock dropped. | |
1560 | * 0: continue | |
1561 | * 1: repeat lookup - reference dropped while waiting for socket memory. | |
1562 | */ | |
9457afee | 1563 | int netlink_attachskb(struct sock *sk, struct sk_buff *skb, |
c3d8d1e3 | 1564 | long *timeo, struct sock *ssk) |
1da177e4 LT |
1565 | { |
1566 | struct netlink_sock *nlk; | |
1567 | ||
1568 | nlk = nlk_sk(sk); | |
1569 | ||
5fd96123 PM |
1570 | if ((atomic_read(&sk->sk_rmem_alloc) > sk->sk_rcvbuf || |
1571 | test_bit(NETLINK_CONGESTED, &nlk->state)) && | |
1572 | !netlink_skb_is_mmaped(skb)) { | |
1da177e4 | 1573 | DECLARE_WAITQUEUE(wait, current); |
c3d8d1e3 | 1574 | if (!*timeo) { |
aed81560 | 1575 | if (!ssk || netlink_is_kernel(ssk)) |
1da177e4 LT |
1576 | netlink_overrun(sk); |
1577 | sock_put(sk); | |
1578 | kfree_skb(skb); | |
1579 | return -EAGAIN; | |
1580 | } | |
1581 | ||
1582 | __set_current_state(TASK_INTERRUPTIBLE); | |
1583 | add_wait_queue(&nlk->wait, &wait); | |
1584 | ||
1585 | if ((atomic_read(&sk->sk_rmem_alloc) > sk->sk_rcvbuf || | |
cd967e05 | 1586 | test_bit(NETLINK_CONGESTED, &nlk->state)) && |
1da177e4 | 1587 | !sock_flag(sk, SOCK_DEAD)) |
c3d8d1e3 | 1588 | *timeo = schedule_timeout(*timeo); |
1da177e4 LT |
1589 | |
1590 | __set_current_state(TASK_RUNNING); | |
1591 | remove_wait_queue(&nlk->wait, &wait); | |
1592 | sock_put(sk); | |
1593 | ||
1594 | if (signal_pending(current)) { | |
1595 | kfree_skb(skb); | |
c3d8d1e3 | 1596 | return sock_intr_errno(*timeo); |
1da177e4 LT |
1597 | } |
1598 | return 1; | |
1599 | } | |
cf0a018a | 1600 | netlink_skb_set_owner_r(skb, sk); |
1da177e4 LT |
1601 | return 0; |
1602 | } | |
1603 | ||
4a7e7c2a | 1604 | static int __netlink_sendskb(struct sock *sk, struct sk_buff *skb) |
1da177e4 | 1605 | { |
1da177e4 LT |
1606 | int len = skb->len; |
1607 | ||
bcbde0d4 DB |
1608 | netlink_deliver_tap(skb); |
1609 | ||
f9c22888 PM |
1610 | #ifdef CONFIG_NETLINK_MMAP |
1611 | if (netlink_skb_is_mmaped(skb)) | |
1612 | netlink_queue_mmaped_skb(sk, skb); | |
1613 | else if (netlink_rx_is_mmaped(sk)) | |
1614 | netlink_ring_set_copied(sk, skb); | |
1615 | else | |
1616 | #endif /* CONFIG_NETLINK_MMAP */ | |
1617 | skb_queue_tail(&sk->sk_receive_queue, skb); | |
1da177e4 | 1618 | sk->sk_data_ready(sk, len); |
4a7e7c2a ED |
1619 | return len; |
1620 | } | |
1621 | ||
1622 | int netlink_sendskb(struct sock *sk, struct sk_buff *skb) | |
1623 | { | |
1624 | int len = __netlink_sendskb(sk, skb); | |
1625 | ||
1da177e4 LT |
1626 | sock_put(sk); |
1627 | return len; | |
1628 | } | |
1629 | ||
1630 | void netlink_detachskb(struct sock *sk, struct sk_buff *skb) | |
1631 | { | |
1632 | kfree_skb(skb); | |
1633 | sock_put(sk); | |
1634 | } | |
1635 | ||
b57ef81f | 1636 | static struct sk_buff *netlink_trim(struct sk_buff *skb, gfp_t allocation) |
1da177e4 LT |
1637 | { |
1638 | int delta; | |
1639 | ||
1298ca46 | 1640 | WARN_ON(skb->sk != NULL); |
5fd96123 PM |
1641 | if (netlink_skb_is_mmaped(skb)) |
1642 | return skb; | |
1da177e4 | 1643 | |
4305b541 | 1644 | delta = skb->end - skb->tail; |
c05cdb1b | 1645 | if (is_vmalloc_addr(skb->head) || delta * 2 < skb->truesize) |
1da177e4 LT |
1646 | return skb; |
1647 | ||
1648 | if (skb_shared(skb)) { | |
1649 | struct sk_buff *nskb = skb_clone(skb, allocation); | |
1650 | if (!nskb) | |
1651 | return skb; | |
8460c00f | 1652 | consume_skb(skb); |
1da177e4 LT |
1653 | skb = nskb; |
1654 | } | |
1655 | ||
1656 | if (!pskb_expand_head(skb, 0, -delta, allocation)) | |
1657 | skb->truesize -= delta; | |
1658 | ||
1659 | return skb; | |
1660 | } | |
1661 | ||
3fbc2905 EB |
1662 | static int netlink_unicast_kernel(struct sock *sk, struct sk_buff *skb, |
1663 | struct sock *ssk) | |
cd40b7d3 DL |
1664 | { |
1665 | int ret; | |
1666 | struct netlink_sock *nlk = nlk_sk(sk); | |
1667 | ||
1668 | ret = -ECONNREFUSED; | |
1669 | if (nlk->netlink_rcv != NULL) { | |
bcbde0d4 DB |
1670 | /* We could do a netlink_deliver_tap(skb) here as well |
1671 | * but since this is intended for the kernel only, we | |
1672 | * should rather let it stay under the hood. | |
1673 | */ | |
1674 | ||
cd40b7d3 | 1675 | ret = skb->len; |
cf0a018a | 1676 | netlink_skb_set_owner_r(skb, sk); |
e32123e5 | 1677 | NETLINK_CB(skb).sk = ssk; |
cd40b7d3 | 1678 | nlk->netlink_rcv(skb); |
bfb253c9 ED |
1679 | consume_skb(skb); |
1680 | } else { | |
1681 | kfree_skb(skb); | |
cd40b7d3 | 1682 | } |
cd40b7d3 DL |
1683 | sock_put(sk); |
1684 | return ret; | |
1685 | } | |
1686 | ||
1687 | int netlink_unicast(struct sock *ssk, struct sk_buff *skb, | |
15e47304 | 1688 | u32 portid, int nonblock) |
1da177e4 LT |
1689 | { |
1690 | struct sock *sk; | |
1691 | int err; | |
1692 | long timeo; | |
1693 | ||
1694 | skb = netlink_trim(skb, gfp_any()); | |
1695 | ||
1696 | timeo = sock_sndtimeo(ssk, nonblock); | |
1697 | retry: | |
15e47304 | 1698 | sk = netlink_getsockbyportid(ssk, portid); |
1da177e4 LT |
1699 | if (IS_ERR(sk)) { |
1700 | kfree_skb(skb); | |
1701 | return PTR_ERR(sk); | |
1702 | } | |
cd40b7d3 | 1703 | if (netlink_is_kernel(sk)) |
3fbc2905 | 1704 | return netlink_unicast_kernel(sk, skb, ssk); |
cd40b7d3 | 1705 | |
b1153f29 | 1706 | if (sk_filter(sk, skb)) { |
84874607 | 1707 | err = skb->len; |
b1153f29 SH |
1708 | kfree_skb(skb); |
1709 | sock_put(sk); | |
1710 | return err; | |
1711 | } | |
1712 | ||
9457afee | 1713 | err = netlink_attachskb(sk, skb, &timeo, ssk); |
1da177e4 LT |
1714 | if (err == 1) |
1715 | goto retry; | |
1716 | if (err) | |
1717 | return err; | |
1718 | ||
7ee015e0 | 1719 | return netlink_sendskb(sk, skb); |
1da177e4 | 1720 | } |
6ac552fd | 1721 | EXPORT_SYMBOL(netlink_unicast); |
1da177e4 | 1722 | |
f9c22888 PM |
1723 | struct sk_buff *netlink_alloc_skb(struct sock *ssk, unsigned int size, |
1724 | u32 dst_portid, gfp_t gfp_mask) | |
1725 | { | |
1726 | #ifdef CONFIG_NETLINK_MMAP | |
1727 | struct sock *sk = NULL; | |
1728 | struct sk_buff *skb; | |
1729 | struct netlink_ring *ring; | |
1730 | struct nl_mmap_hdr *hdr; | |
1731 | unsigned int maxlen; | |
1732 | ||
1733 | sk = netlink_getsockbyportid(ssk, dst_portid); | |
1734 | if (IS_ERR(sk)) | |
1735 | goto out; | |
1736 | ||
1737 | ring = &nlk_sk(sk)->rx_ring; | |
1738 | /* fast-path without atomic ops for common case: non-mmaped receiver */ | |
1739 | if (ring->pg_vec == NULL) | |
1740 | goto out_put; | |
1741 | ||
1742 | skb = alloc_skb_head(gfp_mask); | |
1743 | if (skb == NULL) | |
1744 | goto err1; | |
1745 | ||
1746 | spin_lock_bh(&sk->sk_receive_queue.lock); | |
1747 | /* check again under lock */ | |
1748 | if (ring->pg_vec == NULL) | |
1749 | goto out_free; | |
1750 | ||
1751 | maxlen = ring->frame_size - NL_MMAP_HDRLEN; | |
1752 | if (maxlen < size) | |
1753 | goto out_free; | |
1754 | ||
1755 | netlink_forward_ring(ring); | |
1756 | hdr = netlink_current_frame(ring, NL_MMAP_STATUS_UNUSED); | |
1757 | if (hdr == NULL) | |
1758 | goto err2; | |
1759 | netlink_ring_setup_skb(skb, sk, ring, hdr); | |
1760 | netlink_set_status(hdr, NL_MMAP_STATUS_RESERVED); | |
1761 | atomic_inc(&ring->pending); | |
1762 | netlink_increment_head(ring); | |
1763 | ||
1764 | spin_unlock_bh(&sk->sk_receive_queue.lock); | |
1765 | return skb; | |
1766 | ||
1767 | err2: | |
1768 | kfree_skb(skb); | |
1769 | spin_unlock_bh(&sk->sk_receive_queue.lock); | |
cd1df525 | 1770 | netlink_overrun(sk); |
f9c22888 PM |
1771 | err1: |
1772 | sock_put(sk); | |
1773 | return NULL; | |
1774 | ||
1775 | out_free: | |
1776 | kfree_skb(skb); | |
1777 | spin_unlock_bh(&sk->sk_receive_queue.lock); | |
1778 | out_put: | |
1779 | sock_put(sk); | |
1780 | out: | |
1781 | #endif | |
1782 | return alloc_skb(size, gfp_mask); | |
1783 | } | |
1784 | EXPORT_SYMBOL_GPL(netlink_alloc_skb); | |
1785 | ||
4277a083 PM |
1786 | int netlink_has_listeners(struct sock *sk, unsigned int group) |
1787 | { | |
1788 | int res = 0; | |
5c398dc8 | 1789 | struct listeners *listeners; |
4277a083 | 1790 | |
aed81560 | 1791 | BUG_ON(!netlink_is_kernel(sk)); |
b4ff4f04 JB |
1792 | |
1793 | rcu_read_lock(); | |
1794 | listeners = rcu_dereference(nl_table[sk->sk_protocol].listeners); | |
1795 | ||
6d772ac5 | 1796 | if (listeners && group - 1 < nl_table[sk->sk_protocol].groups) |
5c398dc8 | 1797 | res = test_bit(group - 1, listeners->masks); |
b4ff4f04 JB |
1798 | |
1799 | rcu_read_unlock(); | |
1800 | ||
4277a083 PM |
1801 | return res; |
1802 | } | |
1803 | EXPORT_SYMBOL_GPL(netlink_has_listeners); | |
1804 | ||
b57ef81f | 1805 | static int netlink_broadcast_deliver(struct sock *sk, struct sk_buff *skb) |
1da177e4 LT |
1806 | { |
1807 | struct netlink_sock *nlk = nlk_sk(sk); | |
1808 | ||
1809 | if (atomic_read(&sk->sk_rmem_alloc) <= sk->sk_rcvbuf && | |
cd967e05 | 1810 | !test_bit(NETLINK_CONGESTED, &nlk->state)) { |
cf0a018a | 1811 | netlink_skb_set_owner_r(skb, sk); |
4a7e7c2a | 1812 | __netlink_sendskb(sk, skb); |
2c645800 | 1813 | return atomic_read(&sk->sk_rmem_alloc) > (sk->sk_rcvbuf >> 1); |
1da177e4 LT |
1814 | } |
1815 | return -1; | |
1816 | } | |
1817 | ||
1818 | struct netlink_broadcast_data { | |
1819 | struct sock *exclude_sk; | |
b4b51029 | 1820 | struct net *net; |
15e47304 | 1821 | u32 portid; |
1da177e4 LT |
1822 | u32 group; |
1823 | int failure; | |
ff491a73 | 1824 | int delivery_failure; |
1da177e4 LT |
1825 | int congested; |
1826 | int delivered; | |
7d877f3b | 1827 | gfp_t allocation; |
1da177e4 | 1828 | struct sk_buff *skb, *skb2; |
910a7e90 EB |
1829 | int (*tx_filter)(struct sock *dsk, struct sk_buff *skb, void *data); |
1830 | void *tx_data; | |
1da177e4 LT |
1831 | }; |
1832 | ||
b57ef81f | 1833 | static int do_one_broadcast(struct sock *sk, |
1da177e4 LT |
1834 | struct netlink_broadcast_data *p) |
1835 | { | |
1836 | struct netlink_sock *nlk = nlk_sk(sk); | |
1837 | int val; | |
1838 | ||
1839 | if (p->exclude_sk == sk) | |
1840 | goto out; | |
1841 | ||
15e47304 | 1842 | if (nlk->portid == p->portid || p->group - 1 >= nlk->ngroups || |
f7fa9b10 | 1843 | !test_bit(p->group - 1, nlk->groups)) |
1da177e4 LT |
1844 | goto out; |
1845 | ||
878628fb | 1846 | if (!net_eq(sock_net(sk), p->net)) |
b4b51029 EB |
1847 | goto out; |
1848 | ||
1da177e4 LT |
1849 | if (p->failure) { |
1850 | netlink_overrun(sk); | |
1851 | goto out; | |
1852 | } | |
1853 | ||
1854 | sock_hold(sk); | |
1855 | if (p->skb2 == NULL) { | |
68acc024 | 1856 | if (skb_shared(p->skb)) { |
1da177e4 LT |
1857 | p->skb2 = skb_clone(p->skb, p->allocation); |
1858 | } else { | |
68acc024 TC |
1859 | p->skb2 = skb_get(p->skb); |
1860 | /* | |
1861 | * skb ownership may have been set when | |
1862 | * delivered to a previous socket. | |
1863 | */ | |
1864 | skb_orphan(p->skb2); | |
1da177e4 LT |
1865 | } |
1866 | } | |
1867 | if (p->skb2 == NULL) { | |
1868 | netlink_overrun(sk); | |
1869 | /* Clone failed. Notify ALL listeners. */ | |
1870 | p->failure = 1; | |
be0c22a4 PNA |
1871 | if (nlk->flags & NETLINK_BROADCAST_SEND_ERROR) |
1872 | p->delivery_failure = 1; | |
910a7e90 EB |
1873 | } else if (p->tx_filter && p->tx_filter(sk, p->skb2, p->tx_data)) { |
1874 | kfree_skb(p->skb2); | |
1875 | p->skb2 = NULL; | |
b1153f29 SH |
1876 | } else if (sk_filter(sk, p->skb2)) { |
1877 | kfree_skb(p->skb2); | |
1878 | p->skb2 = NULL; | |
1da177e4 LT |
1879 | } else if ((val = netlink_broadcast_deliver(sk, p->skb2)) < 0) { |
1880 | netlink_overrun(sk); | |
be0c22a4 PNA |
1881 | if (nlk->flags & NETLINK_BROADCAST_SEND_ERROR) |
1882 | p->delivery_failure = 1; | |
1da177e4 LT |
1883 | } else { |
1884 | p->congested |= val; | |
1885 | p->delivered = 1; | |
1886 | p->skb2 = NULL; | |
1887 | } | |
1888 | sock_put(sk); | |
1889 | ||
1890 | out: | |
1891 | return 0; | |
1892 | } | |
1893 | ||
15e47304 | 1894 | int netlink_broadcast_filtered(struct sock *ssk, struct sk_buff *skb, u32 portid, |
910a7e90 EB |
1895 | u32 group, gfp_t allocation, |
1896 | int (*filter)(struct sock *dsk, struct sk_buff *skb, void *data), | |
1897 | void *filter_data) | |
1da177e4 | 1898 | { |
3b1e0a65 | 1899 | struct net *net = sock_net(ssk); |
1da177e4 | 1900 | struct netlink_broadcast_data info; |
1da177e4 LT |
1901 | struct sock *sk; |
1902 | ||
1903 | skb = netlink_trim(skb, allocation); | |
1904 | ||
1905 | info.exclude_sk = ssk; | |
b4b51029 | 1906 | info.net = net; |
15e47304 | 1907 | info.portid = portid; |
1da177e4 LT |
1908 | info.group = group; |
1909 | info.failure = 0; | |
ff491a73 | 1910 | info.delivery_failure = 0; |
1da177e4 LT |
1911 | info.congested = 0; |
1912 | info.delivered = 0; | |
1913 | info.allocation = allocation; | |
1914 | info.skb = skb; | |
1915 | info.skb2 = NULL; | |
910a7e90 EB |
1916 | info.tx_filter = filter; |
1917 | info.tx_data = filter_data; | |
1da177e4 LT |
1918 | |
1919 | /* While we sleep in clone, do not allow to change socket list */ | |
1920 | ||
1921 | netlink_lock_table(); | |
1922 | ||
b67bfe0d | 1923 | sk_for_each_bound(sk, &nl_table[ssk->sk_protocol].mc_list) |
1da177e4 LT |
1924 | do_one_broadcast(sk, &info); |
1925 | ||
70d4bf6d | 1926 | consume_skb(skb); |
aa1c6a6f | 1927 | |
1da177e4 LT |
1928 | netlink_unlock_table(); |
1929 | ||
70d4bf6d NH |
1930 | if (info.delivery_failure) { |
1931 | kfree_skb(info.skb2); | |
ff491a73 | 1932 | return -ENOBUFS; |
658cb354 ED |
1933 | } |
1934 | consume_skb(info.skb2); | |
ff491a73 | 1935 | |
1da177e4 LT |
1936 | if (info.delivered) { |
1937 | if (info.congested && (allocation & __GFP_WAIT)) | |
1938 | yield(); | |
1939 | return 0; | |
1940 | } | |
1da177e4 LT |
1941 | return -ESRCH; |
1942 | } | |
910a7e90 EB |
1943 | EXPORT_SYMBOL(netlink_broadcast_filtered); |
1944 | ||
15e47304 | 1945 | int netlink_broadcast(struct sock *ssk, struct sk_buff *skb, u32 portid, |
910a7e90 EB |
1946 | u32 group, gfp_t allocation) |
1947 | { | |
15e47304 | 1948 | return netlink_broadcast_filtered(ssk, skb, portid, group, allocation, |
910a7e90 EB |
1949 | NULL, NULL); |
1950 | } | |
6ac552fd | 1951 | EXPORT_SYMBOL(netlink_broadcast); |
1da177e4 LT |
1952 | |
1953 | struct netlink_set_err_data { | |
1954 | struct sock *exclude_sk; | |
15e47304 | 1955 | u32 portid; |
1da177e4 LT |
1956 | u32 group; |
1957 | int code; | |
1958 | }; | |
1959 | ||
b57ef81f | 1960 | static int do_one_set_err(struct sock *sk, struct netlink_set_err_data *p) |
1da177e4 LT |
1961 | { |
1962 | struct netlink_sock *nlk = nlk_sk(sk); | |
1a50307b | 1963 | int ret = 0; |
1da177e4 LT |
1964 | |
1965 | if (sk == p->exclude_sk) | |
1966 | goto out; | |
1967 | ||
09ad9bc7 | 1968 | if (!net_eq(sock_net(sk), sock_net(p->exclude_sk))) |
b4b51029 EB |
1969 | goto out; |
1970 | ||
15e47304 | 1971 | if (nlk->portid == p->portid || p->group - 1 >= nlk->ngroups || |
f7fa9b10 | 1972 | !test_bit(p->group - 1, nlk->groups)) |
1da177e4 LT |
1973 | goto out; |
1974 | ||
1a50307b PNA |
1975 | if (p->code == ENOBUFS && nlk->flags & NETLINK_RECV_NO_ENOBUFS) { |
1976 | ret = 1; | |
1977 | goto out; | |
1978 | } | |
1979 | ||
1da177e4 LT |
1980 | sk->sk_err = p->code; |
1981 | sk->sk_error_report(sk); | |
1982 | out: | |
1a50307b | 1983 | return ret; |
1da177e4 LT |
1984 | } |
1985 | ||
4843b93c PNA |
1986 | /** |
1987 | * netlink_set_err - report error to broadcast listeners | |
1988 | * @ssk: the kernel netlink socket, as returned by netlink_kernel_create() | |
15e47304 | 1989 | * @portid: the PORTID of a process that we want to skip (if any) |
4843b93c PNA |
1990 | * @groups: the broadcast group that will notice the error |
1991 | * @code: error code, must be negative (as usual in kernelspace) | |
1a50307b PNA |
1992 | * |
1993 | * This function returns the number of broadcast listeners that have set the | |
1994 | * NETLINK_RECV_NO_ENOBUFS socket option. | |
4843b93c | 1995 | */ |
15e47304 | 1996 | int netlink_set_err(struct sock *ssk, u32 portid, u32 group, int code) |
1da177e4 LT |
1997 | { |
1998 | struct netlink_set_err_data info; | |
1da177e4 | 1999 | struct sock *sk; |
1a50307b | 2000 | int ret = 0; |
1da177e4 LT |
2001 | |
2002 | info.exclude_sk = ssk; | |
15e47304 | 2003 | info.portid = portid; |
1da177e4 | 2004 | info.group = group; |
4843b93c PNA |
2005 | /* sk->sk_err wants a positive error value */ |
2006 | info.code = -code; | |
1da177e4 LT |
2007 | |
2008 | read_lock(&nl_table_lock); | |
2009 | ||
b67bfe0d | 2010 | sk_for_each_bound(sk, &nl_table[ssk->sk_protocol].mc_list) |
1a50307b | 2011 | ret += do_one_set_err(sk, &info); |
1da177e4 LT |
2012 | |
2013 | read_unlock(&nl_table_lock); | |
1a50307b | 2014 | return ret; |
1da177e4 | 2015 | } |
dd5b6ce6 | 2016 | EXPORT_SYMBOL(netlink_set_err); |
1da177e4 | 2017 | |
84659eb5 JB |
2018 | /* must be called with netlink table grabbed */ |
2019 | static void netlink_update_socket_mc(struct netlink_sock *nlk, | |
2020 | unsigned int group, | |
2021 | int is_new) | |
2022 | { | |
2023 | int old, new = !!is_new, subscriptions; | |
2024 | ||
2025 | old = test_bit(group - 1, nlk->groups); | |
2026 | subscriptions = nlk->subscriptions - old + new; | |
2027 | if (new) | |
2028 | __set_bit(group - 1, nlk->groups); | |
2029 | else | |
2030 | __clear_bit(group - 1, nlk->groups); | |
2031 | netlink_update_subscriptions(&nlk->sk, subscriptions); | |
2032 | netlink_update_listeners(&nlk->sk); | |
2033 | } | |
2034 | ||
9a4595bc | 2035 | static int netlink_setsockopt(struct socket *sock, int level, int optname, |
b7058842 | 2036 | char __user *optval, unsigned int optlen) |
9a4595bc PM |
2037 | { |
2038 | struct sock *sk = sock->sk; | |
2039 | struct netlink_sock *nlk = nlk_sk(sk); | |
eb496534 JB |
2040 | unsigned int val = 0; |
2041 | int err; | |
9a4595bc PM |
2042 | |
2043 | if (level != SOL_NETLINK) | |
2044 | return -ENOPROTOOPT; | |
2045 | ||
ccdfcc39 PM |
2046 | if (optname != NETLINK_RX_RING && optname != NETLINK_TX_RING && |
2047 | optlen >= sizeof(int) && | |
eb496534 | 2048 | get_user(val, (unsigned int __user *)optval)) |
9a4595bc PM |
2049 | return -EFAULT; |
2050 | ||
2051 | switch (optname) { | |
2052 | case NETLINK_PKTINFO: | |
2053 | if (val) | |
2054 | nlk->flags |= NETLINK_RECV_PKTINFO; | |
2055 | else | |
2056 | nlk->flags &= ~NETLINK_RECV_PKTINFO; | |
2057 | err = 0; | |
2058 | break; | |
2059 | case NETLINK_ADD_MEMBERSHIP: | |
2060 | case NETLINK_DROP_MEMBERSHIP: { | |
9785e10a | 2061 | if (!netlink_capable(sock, NL_CFG_F_NONROOT_RECV)) |
9a4595bc | 2062 | return -EPERM; |
b4ff4f04 JB |
2063 | err = netlink_realloc_groups(sk); |
2064 | if (err) | |
2065 | return err; | |
9a4595bc PM |
2066 | if (!val || val - 1 >= nlk->ngroups) |
2067 | return -EINVAL; | |
2068 | netlink_table_grab(); | |
84659eb5 JB |
2069 | netlink_update_socket_mc(nlk, val, |
2070 | optname == NETLINK_ADD_MEMBERSHIP); | |
9a4595bc | 2071 | netlink_table_ungrab(); |
03292745 PNA |
2072 | |
2073 | if (nlk->netlink_bind) | |
2074 | nlk->netlink_bind(val); | |
2075 | ||
9a4595bc PM |
2076 | err = 0; |
2077 | break; | |
2078 | } | |
be0c22a4 PNA |
2079 | case NETLINK_BROADCAST_ERROR: |
2080 | if (val) | |
2081 | nlk->flags |= NETLINK_BROADCAST_SEND_ERROR; | |
2082 | else | |
2083 | nlk->flags &= ~NETLINK_BROADCAST_SEND_ERROR; | |
2084 | err = 0; | |
2085 | break; | |
38938bfe PNA |
2086 | case NETLINK_NO_ENOBUFS: |
2087 | if (val) { | |
2088 | nlk->flags |= NETLINK_RECV_NO_ENOBUFS; | |
cd967e05 | 2089 | clear_bit(NETLINK_CONGESTED, &nlk->state); |
38938bfe | 2090 | wake_up_interruptible(&nlk->wait); |
658cb354 | 2091 | } else { |
38938bfe | 2092 | nlk->flags &= ~NETLINK_RECV_NO_ENOBUFS; |
658cb354 | 2093 | } |
38938bfe PNA |
2094 | err = 0; |
2095 | break; | |
ccdfcc39 PM |
2096 | #ifdef CONFIG_NETLINK_MMAP |
2097 | case NETLINK_RX_RING: | |
2098 | case NETLINK_TX_RING: { | |
2099 | struct nl_mmap_req req; | |
2100 | ||
2101 | /* Rings might consume more memory than queue limits, require | |
2102 | * CAP_NET_ADMIN. | |
2103 | */ | |
2104 | if (!capable(CAP_NET_ADMIN)) | |
2105 | return -EPERM; | |
2106 | if (optlen < sizeof(req)) | |
2107 | return -EINVAL; | |
2108 | if (copy_from_user(&req, optval, sizeof(req))) | |
2109 | return -EFAULT; | |
2110 | err = netlink_set_ring(sk, &req, false, | |
2111 | optname == NETLINK_TX_RING); | |
2112 | break; | |
2113 | } | |
2114 | #endif /* CONFIG_NETLINK_MMAP */ | |
9a4595bc PM |
2115 | default: |
2116 | err = -ENOPROTOOPT; | |
2117 | } | |
2118 | return err; | |
2119 | } | |
2120 | ||
2121 | static int netlink_getsockopt(struct socket *sock, int level, int optname, | |
746fac4d | 2122 | char __user *optval, int __user *optlen) |
9a4595bc PM |
2123 | { |
2124 | struct sock *sk = sock->sk; | |
2125 | struct netlink_sock *nlk = nlk_sk(sk); | |
2126 | int len, val, err; | |
2127 | ||
2128 | if (level != SOL_NETLINK) | |
2129 | return -ENOPROTOOPT; | |
2130 | ||
2131 | if (get_user(len, optlen)) | |
2132 | return -EFAULT; | |
2133 | if (len < 0) | |
2134 | return -EINVAL; | |
2135 | ||
2136 | switch (optname) { | |
2137 | case NETLINK_PKTINFO: | |
2138 | if (len < sizeof(int)) | |
2139 | return -EINVAL; | |
2140 | len = sizeof(int); | |
2141 | val = nlk->flags & NETLINK_RECV_PKTINFO ? 1 : 0; | |
a27b58fe HC |
2142 | if (put_user(len, optlen) || |
2143 | put_user(val, optval)) | |
2144 | return -EFAULT; | |
9a4595bc PM |
2145 | err = 0; |
2146 | break; | |
be0c22a4 PNA |
2147 | case NETLINK_BROADCAST_ERROR: |
2148 | if (len < sizeof(int)) | |
2149 | return -EINVAL; | |
2150 | len = sizeof(int); | |
2151 | val = nlk->flags & NETLINK_BROADCAST_SEND_ERROR ? 1 : 0; | |
2152 | if (put_user(len, optlen) || | |
2153 | put_user(val, optval)) | |
2154 | return -EFAULT; | |
2155 | err = 0; | |
2156 | break; | |
38938bfe PNA |
2157 | case NETLINK_NO_ENOBUFS: |
2158 | if (len < sizeof(int)) | |
2159 | return -EINVAL; | |
2160 | len = sizeof(int); | |
2161 | val = nlk->flags & NETLINK_RECV_NO_ENOBUFS ? 1 : 0; | |
2162 | if (put_user(len, optlen) || | |
2163 | put_user(val, optval)) | |
2164 | return -EFAULT; | |
2165 | err = 0; | |
2166 | break; | |
9a4595bc PM |
2167 | default: |
2168 | err = -ENOPROTOOPT; | |
2169 | } | |
2170 | return err; | |
2171 | } | |
2172 | ||
2173 | static void netlink_cmsg_recv_pktinfo(struct msghdr *msg, struct sk_buff *skb) | |
2174 | { | |
2175 | struct nl_pktinfo info; | |
2176 | ||
2177 | info.group = NETLINK_CB(skb).dst_group; | |
2178 | put_cmsg(msg, SOL_NETLINK, NETLINK_PKTINFO, sizeof(info), &info); | |
2179 | } | |
2180 | ||
1da177e4 LT |
2181 | static int netlink_sendmsg(struct kiocb *kiocb, struct socket *sock, |
2182 | struct msghdr *msg, size_t len) | |
2183 | { | |
2184 | struct sock_iocb *siocb = kiocb_to_siocb(kiocb); | |
2185 | struct sock *sk = sock->sk; | |
2186 | struct netlink_sock *nlk = nlk_sk(sk); | |
6ac552fd | 2187 | struct sockaddr_nl *addr = msg->msg_name; |
15e47304 | 2188 | u32 dst_portid; |
d629b836 | 2189 | u32 dst_group; |
1da177e4 LT |
2190 | struct sk_buff *skb; |
2191 | int err; | |
2192 | struct scm_cookie scm; | |
2193 | ||
2194 | if (msg->msg_flags&MSG_OOB) | |
2195 | return -EOPNOTSUPP; | |
2196 | ||
16e57262 | 2197 | if (NULL == siocb->scm) |
1da177e4 | 2198 | siocb->scm = &scm; |
16e57262 | 2199 | |
e0e3cea4 | 2200 | err = scm_send(sock, msg, siocb->scm, true); |
1da177e4 LT |
2201 | if (err < 0) |
2202 | return err; | |
2203 | ||
2204 | if (msg->msg_namelen) { | |
b47030c7 | 2205 | err = -EINVAL; |
1da177e4 | 2206 | if (addr->nl_family != AF_NETLINK) |
b47030c7 | 2207 | goto out; |
15e47304 | 2208 | dst_portid = addr->nl_pid; |
d629b836 | 2209 | dst_group = ffs(addr->nl_groups); |
b47030c7 | 2210 | err = -EPERM; |
15e47304 | 2211 | if ((dst_group || dst_portid) && |
9785e10a | 2212 | !netlink_capable(sock, NL_CFG_F_NONROOT_SEND)) |
b47030c7 | 2213 | goto out; |
1da177e4 | 2214 | } else { |
15e47304 | 2215 | dst_portid = nlk->dst_portid; |
d629b836 | 2216 | dst_group = nlk->dst_group; |
1da177e4 LT |
2217 | } |
2218 | ||
15e47304 | 2219 | if (!nlk->portid) { |
1da177e4 LT |
2220 | err = netlink_autobind(sock); |
2221 | if (err) | |
2222 | goto out; | |
2223 | } | |
2224 | ||
5fd96123 PM |
2225 | if (netlink_tx_is_mmaped(sk) && |
2226 | msg->msg_iov->iov_base == NULL) { | |
2227 | err = netlink_mmap_sendmsg(sk, msg, dst_portid, dst_group, | |
2228 | siocb); | |
2229 | goto out; | |
2230 | } | |
2231 | ||
1da177e4 LT |
2232 | err = -EMSGSIZE; |
2233 | if (len > sk->sk_sndbuf - 32) | |
2234 | goto out; | |
2235 | err = -ENOBUFS; | |
3a36515f | 2236 | skb = netlink_alloc_large_skb(len, dst_group); |
6ac552fd | 2237 | if (skb == NULL) |
1da177e4 LT |
2238 | goto out; |
2239 | ||
15e47304 | 2240 | NETLINK_CB(skb).portid = nlk->portid; |
d629b836 | 2241 | NETLINK_CB(skb).dst_group = dst_group; |
dbe9a417 | 2242 | NETLINK_CB(skb).creds = siocb->scm->creds; |
1da177e4 | 2243 | |
1da177e4 | 2244 | err = -EFAULT; |
6ac552fd | 2245 | if (memcpy_fromiovec(skb_put(skb, len), msg->msg_iov, len)) { |
1da177e4 LT |
2246 | kfree_skb(skb); |
2247 | goto out; | |
2248 | } | |
2249 | ||
2250 | err = security_netlink_send(sk, skb); | |
2251 | if (err) { | |
2252 | kfree_skb(skb); | |
2253 | goto out; | |
2254 | } | |
2255 | ||
d629b836 | 2256 | if (dst_group) { |
1da177e4 | 2257 | atomic_inc(&skb->users); |
15e47304 | 2258 | netlink_broadcast(sk, skb, dst_portid, dst_group, GFP_KERNEL); |
1da177e4 | 2259 | } |
15e47304 | 2260 | err = netlink_unicast(sk, skb, dst_portid, msg->msg_flags&MSG_DONTWAIT); |
1da177e4 LT |
2261 | |
2262 | out: | |
b47030c7 | 2263 | scm_destroy(siocb->scm); |
1da177e4 LT |
2264 | return err; |
2265 | } | |
2266 | ||
2267 | static int netlink_recvmsg(struct kiocb *kiocb, struct socket *sock, | |
2268 | struct msghdr *msg, size_t len, | |
2269 | int flags) | |
2270 | { | |
2271 | struct sock_iocb *siocb = kiocb_to_siocb(kiocb); | |
2272 | struct scm_cookie scm; | |
2273 | struct sock *sk = sock->sk; | |
2274 | struct netlink_sock *nlk = nlk_sk(sk); | |
2275 | int noblock = flags&MSG_DONTWAIT; | |
2276 | size_t copied; | |
68d6ac6d | 2277 | struct sk_buff *skb, *data_skb; |
b44d211e | 2278 | int err, ret; |
1da177e4 LT |
2279 | |
2280 | if (flags&MSG_OOB) | |
2281 | return -EOPNOTSUPP; | |
2282 | ||
2283 | copied = 0; | |
2284 | ||
6ac552fd PM |
2285 | skb = skb_recv_datagram(sk, flags, noblock, &err); |
2286 | if (skb == NULL) | |
1da177e4 LT |
2287 | goto out; |
2288 | ||
68d6ac6d JB |
2289 | data_skb = skb; |
2290 | ||
1dacc76d JB |
2291 | #ifdef CONFIG_COMPAT_NETLINK_MESSAGES |
2292 | if (unlikely(skb_shinfo(skb)->frag_list)) { | |
1dacc76d | 2293 | /* |
68d6ac6d JB |
2294 | * If this skb has a frag_list, then here that means that we |
2295 | * will have to use the frag_list skb's data for compat tasks | |
2296 | * and the regular skb's data for normal (non-compat) tasks. | |
1dacc76d | 2297 | * |
68d6ac6d JB |
2298 | * If we need to send the compat skb, assign it to the |
2299 | * 'data_skb' variable so that it will be used below for data | |
2300 | * copying. We keep 'skb' for everything else, including | |
2301 | * freeing both later. | |
1dacc76d | 2302 | */ |
68d6ac6d JB |
2303 | if (flags & MSG_CMSG_COMPAT) |
2304 | data_skb = skb_shinfo(skb)->frag_list; | |
1dacc76d JB |
2305 | } |
2306 | #endif | |
2307 | ||
1da177e4 LT |
2308 | msg->msg_namelen = 0; |
2309 | ||
68d6ac6d | 2310 | copied = data_skb->len; |
1da177e4 LT |
2311 | if (len < copied) { |
2312 | msg->msg_flags |= MSG_TRUNC; | |
2313 | copied = len; | |
2314 | } | |
2315 | ||
68d6ac6d JB |
2316 | skb_reset_transport_header(data_skb); |
2317 | err = skb_copy_datagram_iovec(data_skb, 0, msg->msg_iov, copied); | |
1da177e4 LT |
2318 | |
2319 | if (msg->msg_name) { | |
6ac552fd | 2320 | struct sockaddr_nl *addr = (struct sockaddr_nl *)msg->msg_name; |
1da177e4 LT |
2321 | addr->nl_family = AF_NETLINK; |
2322 | addr->nl_pad = 0; | |
15e47304 | 2323 | addr->nl_pid = NETLINK_CB(skb).portid; |
d629b836 | 2324 | addr->nl_groups = netlink_group_mask(NETLINK_CB(skb).dst_group); |
1da177e4 LT |
2325 | msg->msg_namelen = sizeof(*addr); |
2326 | } | |
2327 | ||
cc9a06cd PM |
2328 | if (nlk->flags & NETLINK_RECV_PKTINFO) |
2329 | netlink_cmsg_recv_pktinfo(msg, skb); | |
2330 | ||
1da177e4 LT |
2331 | if (NULL == siocb->scm) { |
2332 | memset(&scm, 0, sizeof(scm)); | |
2333 | siocb->scm = &scm; | |
2334 | } | |
2335 | siocb->scm->creds = *NETLINK_CREDS(skb); | |
188ccb55 | 2336 | if (flags & MSG_TRUNC) |
68d6ac6d | 2337 | copied = data_skb->len; |
daa3766e | 2338 | |
1da177e4 LT |
2339 | skb_free_datagram(sk, skb); |
2340 | ||
16b304f3 PS |
2341 | if (nlk->cb_running && |
2342 | atomic_read(&sk->sk_rmem_alloc) <= sk->sk_rcvbuf / 2) { | |
b44d211e AV |
2343 | ret = netlink_dump(sk); |
2344 | if (ret) { | |
2345 | sk->sk_err = ret; | |
2346 | sk->sk_error_report(sk); | |
2347 | } | |
2348 | } | |
1da177e4 LT |
2349 | |
2350 | scm_recv(sock, msg, siocb->scm, flags); | |
1da177e4 LT |
2351 | out: |
2352 | netlink_rcv_wake(sk); | |
2353 | return err ? : copied; | |
2354 | } | |
2355 | ||
2356 | static void netlink_data_ready(struct sock *sk, int len) | |
2357 | { | |
cd40b7d3 | 2358 | BUG(); |
1da177e4 LT |
2359 | } |
2360 | ||
2361 | /* | |
746fac4d | 2362 | * We export these functions to other modules. They provide a |
1da177e4 LT |
2363 | * complete set of kernel non-blocking support for message |
2364 | * queueing. | |
2365 | */ | |
2366 | ||
2367 | struct sock * | |
9f00d977 PNA |
2368 | __netlink_kernel_create(struct net *net, int unit, struct module *module, |
2369 | struct netlink_kernel_cfg *cfg) | |
1da177e4 LT |
2370 | { |
2371 | struct socket *sock; | |
2372 | struct sock *sk; | |
77247bbb | 2373 | struct netlink_sock *nlk; |
5c398dc8 | 2374 | struct listeners *listeners = NULL; |
a31f2d17 PNA |
2375 | struct mutex *cb_mutex = cfg ? cfg->cb_mutex : NULL; |
2376 | unsigned int groups; | |
1da177e4 | 2377 | |
fab2caf6 | 2378 | BUG_ON(!nl_table); |
1da177e4 | 2379 | |
6ac552fd | 2380 | if (unit < 0 || unit >= MAX_LINKS) |
1da177e4 LT |
2381 | return NULL; |
2382 | ||
2383 | if (sock_create_lite(PF_NETLINK, SOCK_DGRAM, unit, &sock)) | |
2384 | return NULL; | |
2385 | ||
23fe1866 PE |
2386 | /* |
2387 | * We have to just have a reference on the net from sk, but don't | |
2388 | * get_net it. Besides, we cannot get and then put the net here. | |
2389 | * So we create one inside init_net and the move it to net. | |
2390 | */ | |
2391 | ||
2392 | if (__netlink_create(&init_net, sock, cb_mutex, unit) < 0) | |
2393 | goto out_sock_release_nosk; | |
2394 | ||
2395 | sk = sock->sk; | |
edf02087 | 2396 | sk_change_net(sk, net); |
4fdb3bb7 | 2397 | |
a31f2d17 | 2398 | if (!cfg || cfg->groups < 32) |
4277a083 | 2399 | groups = 32; |
a31f2d17 PNA |
2400 | else |
2401 | groups = cfg->groups; | |
4277a083 | 2402 | |
5c398dc8 | 2403 | listeners = kzalloc(sizeof(*listeners) + NLGRPSZ(groups), GFP_KERNEL); |
4277a083 PM |
2404 | if (!listeners) |
2405 | goto out_sock_release; | |
2406 | ||
1da177e4 | 2407 | sk->sk_data_ready = netlink_data_ready; |
a31f2d17 PNA |
2408 | if (cfg && cfg->input) |
2409 | nlk_sk(sk)->netlink_rcv = cfg->input; | |
1da177e4 | 2410 | |
b4b51029 | 2411 | if (netlink_insert(sk, net, 0)) |
77247bbb | 2412 | goto out_sock_release; |
4fdb3bb7 | 2413 | |
77247bbb PM |
2414 | nlk = nlk_sk(sk); |
2415 | nlk->flags |= NETLINK_KERNEL_SOCKET; | |
4fdb3bb7 | 2416 | |
4fdb3bb7 | 2417 | netlink_table_grab(); |
b4b51029 EB |
2418 | if (!nl_table[unit].registered) { |
2419 | nl_table[unit].groups = groups; | |
5c398dc8 | 2420 | rcu_assign_pointer(nl_table[unit].listeners, listeners); |
b4b51029 EB |
2421 | nl_table[unit].cb_mutex = cb_mutex; |
2422 | nl_table[unit].module = module; | |
9785e10a PNA |
2423 | if (cfg) { |
2424 | nl_table[unit].bind = cfg->bind; | |
2425 | nl_table[unit].flags = cfg->flags; | |
da12c90e G |
2426 | if (cfg->compare) |
2427 | nl_table[unit].compare = cfg->compare; | |
9785e10a | 2428 | } |
b4b51029 | 2429 | nl_table[unit].registered = 1; |
f937f1f4 JJ |
2430 | } else { |
2431 | kfree(listeners); | |
869e58f8 | 2432 | nl_table[unit].registered++; |
b4b51029 | 2433 | } |
4fdb3bb7 | 2434 | netlink_table_ungrab(); |
77247bbb PM |
2435 | return sk; |
2436 | ||
4fdb3bb7 | 2437 | out_sock_release: |
4277a083 | 2438 | kfree(listeners); |
9dfbec1f | 2439 | netlink_kernel_release(sk); |
23fe1866 PE |
2440 | return NULL; |
2441 | ||
2442 | out_sock_release_nosk: | |
4fdb3bb7 | 2443 | sock_release(sock); |
77247bbb | 2444 | return NULL; |
1da177e4 | 2445 | } |
9f00d977 | 2446 | EXPORT_SYMBOL(__netlink_kernel_create); |
b7c6ba6e DL |
2447 | |
2448 | void | |
2449 | netlink_kernel_release(struct sock *sk) | |
2450 | { | |
edf02087 | 2451 | sk_release_kernel(sk); |
b7c6ba6e DL |
2452 | } |
2453 | EXPORT_SYMBOL(netlink_kernel_release); | |
2454 | ||
d136f1bd | 2455 | int __netlink_change_ngroups(struct sock *sk, unsigned int groups) |
b4ff4f04 | 2456 | { |
5c398dc8 | 2457 | struct listeners *new, *old; |
b4ff4f04 | 2458 | struct netlink_table *tbl = &nl_table[sk->sk_protocol]; |
b4ff4f04 JB |
2459 | |
2460 | if (groups < 32) | |
2461 | groups = 32; | |
2462 | ||
b4ff4f04 | 2463 | if (NLGRPSZ(tbl->groups) < NLGRPSZ(groups)) { |
5c398dc8 ED |
2464 | new = kzalloc(sizeof(*new) + NLGRPSZ(groups), GFP_ATOMIC); |
2465 | if (!new) | |
d136f1bd | 2466 | return -ENOMEM; |
6d772ac5 | 2467 | old = nl_deref_protected(tbl->listeners); |
5c398dc8 ED |
2468 | memcpy(new->masks, old->masks, NLGRPSZ(tbl->groups)); |
2469 | rcu_assign_pointer(tbl->listeners, new); | |
2470 | ||
37b6b935 | 2471 | kfree_rcu(old, rcu); |
b4ff4f04 JB |
2472 | } |
2473 | tbl->groups = groups; | |
2474 | ||
d136f1bd JB |
2475 | return 0; |
2476 | } | |
2477 | ||
2478 | /** | |
2479 | * netlink_change_ngroups - change number of multicast groups | |
2480 | * | |
2481 | * This changes the number of multicast groups that are available | |
2482 | * on a certain netlink family. Note that it is not possible to | |
2483 | * change the number of groups to below 32. Also note that it does | |
2484 | * not implicitly call netlink_clear_multicast_users() when the | |
2485 | * number of groups is reduced. | |
2486 | * | |
2487 | * @sk: The kernel netlink socket, as returned by netlink_kernel_create(). | |
2488 | * @groups: The new number of groups. | |
2489 | */ | |
2490 | int netlink_change_ngroups(struct sock *sk, unsigned int groups) | |
2491 | { | |
2492 | int err; | |
2493 | ||
2494 | netlink_table_grab(); | |
2495 | err = __netlink_change_ngroups(sk, groups); | |
b4ff4f04 | 2496 | netlink_table_ungrab(); |
d136f1bd | 2497 | |
b4ff4f04 JB |
2498 | return err; |
2499 | } | |
b4ff4f04 | 2500 | |
b8273570 JB |
2501 | void __netlink_clear_multicast_users(struct sock *ksk, unsigned int group) |
2502 | { | |
2503 | struct sock *sk; | |
b8273570 JB |
2504 | struct netlink_table *tbl = &nl_table[ksk->sk_protocol]; |
2505 | ||
b67bfe0d | 2506 | sk_for_each_bound(sk, &tbl->mc_list) |
b8273570 JB |
2507 | netlink_update_socket_mc(nlk_sk(sk), group, 0); |
2508 | } | |
2509 | ||
84659eb5 JB |
2510 | /** |
2511 | * netlink_clear_multicast_users - kick off multicast listeners | |
2512 | * | |
2513 | * This function removes all listeners from the given group. | |
2514 | * @ksk: The kernel netlink socket, as returned by | |
2515 | * netlink_kernel_create(). | |
2516 | * @group: The multicast group to clear. | |
2517 | */ | |
2518 | void netlink_clear_multicast_users(struct sock *ksk, unsigned int group) | |
2519 | { | |
84659eb5 | 2520 | netlink_table_grab(); |
b8273570 | 2521 | __netlink_clear_multicast_users(ksk, group); |
84659eb5 JB |
2522 | netlink_table_ungrab(); |
2523 | } | |
84659eb5 | 2524 | |
a46621a3 | 2525 | struct nlmsghdr * |
15e47304 | 2526 | __nlmsg_put(struct sk_buff *skb, u32 portid, u32 seq, int type, int len, int flags) |
a46621a3 DV |
2527 | { |
2528 | struct nlmsghdr *nlh; | |
573ce260 | 2529 | int size = nlmsg_msg_size(len); |
a46621a3 DV |
2530 | |
2531 | nlh = (struct nlmsghdr*)skb_put(skb, NLMSG_ALIGN(size)); | |
2532 | nlh->nlmsg_type = type; | |
2533 | nlh->nlmsg_len = size; | |
2534 | nlh->nlmsg_flags = flags; | |
15e47304 | 2535 | nlh->nlmsg_pid = portid; |
a46621a3 DV |
2536 | nlh->nlmsg_seq = seq; |
2537 | if (!__builtin_constant_p(size) || NLMSG_ALIGN(size) - size != 0) | |
573ce260 | 2538 | memset(nlmsg_data(nlh) + len, 0, NLMSG_ALIGN(size) - size); |
a46621a3 DV |
2539 | return nlh; |
2540 | } | |
2541 | EXPORT_SYMBOL(__nlmsg_put); | |
2542 | ||
1da177e4 LT |
2543 | /* |
2544 | * It looks a bit ugly. | |
2545 | * It would be better to create kernel thread. | |
2546 | */ | |
2547 | ||
2548 | static int netlink_dump(struct sock *sk) | |
2549 | { | |
2550 | struct netlink_sock *nlk = nlk_sk(sk); | |
2551 | struct netlink_callback *cb; | |
c7ac8679 | 2552 | struct sk_buff *skb = NULL; |
1da177e4 | 2553 | struct nlmsghdr *nlh; |
bf8b79e4 | 2554 | int len, err = -ENOBUFS; |
c7ac8679 | 2555 | int alloc_size; |
1da177e4 | 2556 | |
af65bdfc | 2557 | mutex_lock(nlk->cb_mutex); |
16b304f3 | 2558 | if (!nlk->cb_running) { |
bf8b79e4 TG |
2559 | err = -EINVAL; |
2560 | goto errout_skb; | |
1da177e4 LT |
2561 | } |
2562 | ||
16b304f3 | 2563 | cb = &nlk->cb; |
c7ac8679 GR |
2564 | alloc_size = max_t(int, cb->min_dump_alloc, NLMSG_GOODSIZE); |
2565 | ||
f9c22888 PM |
2566 | if (!netlink_rx_is_mmaped(sk) && |
2567 | atomic_read(&sk->sk_rmem_alloc) >= sk->sk_rcvbuf) | |
2568 | goto errout_skb; | |
2569 | skb = netlink_alloc_skb(sk, alloc_size, nlk->portid, GFP_KERNEL); | |
c7ac8679 | 2570 | if (!skb) |
c63d6ea3 | 2571 | goto errout_skb; |
f9c22888 | 2572 | netlink_skb_set_owner_r(skb, sk); |
c7ac8679 | 2573 | |
1da177e4 LT |
2574 | len = cb->dump(skb, cb); |
2575 | ||
2576 | if (len > 0) { | |
af65bdfc | 2577 | mutex_unlock(nlk->cb_mutex); |
b1153f29 SH |
2578 | |
2579 | if (sk_filter(sk, skb)) | |
2580 | kfree_skb(skb); | |
4a7e7c2a ED |
2581 | else |
2582 | __netlink_sendskb(sk, skb); | |
1da177e4 LT |
2583 | return 0; |
2584 | } | |
2585 | ||
bf8b79e4 TG |
2586 | nlh = nlmsg_put_answer(skb, cb, NLMSG_DONE, sizeof(len), NLM_F_MULTI); |
2587 | if (!nlh) | |
2588 | goto errout_skb; | |
2589 | ||
670dc283 JB |
2590 | nl_dump_check_consistent(cb, nlh); |
2591 | ||
bf8b79e4 TG |
2592 | memcpy(nlmsg_data(nlh), &len, sizeof(len)); |
2593 | ||
b1153f29 SH |
2594 | if (sk_filter(sk, skb)) |
2595 | kfree_skb(skb); | |
4a7e7c2a ED |
2596 | else |
2597 | __netlink_sendskb(sk, skb); | |
1da177e4 | 2598 | |
a8f74b22 TG |
2599 | if (cb->done) |
2600 | cb->done(cb); | |
1da177e4 | 2601 | |
16b304f3 PS |
2602 | nlk->cb_running = false; |
2603 | mutex_unlock(nlk->cb_mutex); | |
6dc878a8 | 2604 | module_put(cb->module); |
16b304f3 | 2605 | consume_skb(cb->skb); |
1da177e4 | 2606 | return 0; |
1797754e | 2607 | |
bf8b79e4 | 2608 | errout_skb: |
af65bdfc | 2609 | mutex_unlock(nlk->cb_mutex); |
bf8b79e4 | 2610 | kfree_skb(skb); |
bf8b79e4 | 2611 | return err; |
1da177e4 LT |
2612 | } |
2613 | ||
6dc878a8 G |
2614 | int __netlink_dump_start(struct sock *ssk, struct sk_buff *skb, |
2615 | const struct nlmsghdr *nlh, | |
2616 | struct netlink_dump_control *control) | |
1da177e4 LT |
2617 | { |
2618 | struct netlink_callback *cb; | |
2619 | struct sock *sk; | |
2620 | struct netlink_sock *nlk; | |
b44d211e | 2621 | int ret; |
1da177e4 | 2622 | |
f9c22888 PM |
2623 | /* Memory mapped dump requests need to be copied to avoid looping |
2624 | * on the pending state in netlink_mmap_sendmsg() while the CB hold | |
2625 | * a reference to the skb. | |
2626 | */ | |
2627 | if (netlink_skb_is_mmaped(skb)) { | |
2628 | skb = skb_copy(skb, GFP_KERNEL); | |
16b304f3 | 2629 | if (skb == NULL) |
f9c22888 | 2630 | return -ENOBUFS; |
f9c22888 PM |
2631 | } else |
2632 | atomic_inc(&skb->users); | |
2633 | ||
15e47304 | 2634 | sk = netlink_lookup(sock_net(ssk), ssk->sk_protocol, NETLINK_CB(skb).portid); |
1da177e4 | 2635 | if (sk == NULL) { |
16b304f3 PS |
2636 | ret = -ECONNREFUSED; |
2637 | goto error_free; | |
1da177e4 | 2638 | } |
6dc878a8 | 2639 | |
16b304f3 | 2640 | nlk = nlk_sk(sk); |
af65bdfc | 2641 | mutex_lock(nlk->cb_mutex); |
6dc878a8 | 2642 | /* A dump is in progress... */ |
16b304f3 | 2643 | if (nlk->cb_running) { |
6dc878a8 | 2644 | ret = -EBUSY; |
16b304f3 | 2645 | goto error_unlock; |
1da177e4 | 2646 | } |
6dc878a8 | 2647 | /* add reference of module which cb->dump belongs to */ |
16b304f3 | 2648 | if (!try_module_get(control->module)) { |
6dc878a8 | 2649 | ret = -EPROTONOSUPPORT; |
16b304f3 | 2650 | goto error_unlock; |
6dc878a8 G |
2651 | } |
2652 | ||
16b304f3 PS |
2653 | cb = &nlk->cb; |
2654 | memset(cb, 0, sizeof(*cb)); | |
2655 | cb->dump = control->dump; | |
2656 | cb->done = control->done; | |
2657 | cb->nlh = nlh; | |
2658 | cb->data = control->data; | |
2659 | cb->module = control->module; | |
2660 | cb->min_dump_alloc = control->min_dump_alloc; | |
2661 | cb->skb = skb; | |
2662 | ||
2663 | nlk->cb_running = true; | |
2664 | ||
af65bdfc | 2665 | mutex_unlock(nlk->cb_mutex); |
1da177e4 | 2666 | |
b44d211e | 2667 | ret = netlink_dump(sk); |
1da177e4 | 2668 | sock_put(sk); |
5c58298c | 2669 | |
b44d211e AV |
2670 | if (ret) |
2671 | return ret; | |
2672 | ||
5c58298c DL |
2673 | /* We successfully started a dump, by returning -EINTR we |
2674 | * signal not to send ACK even if it was requested. | |
2675 | */ | |
2676 | return -EINTR; | |
16b304f3 PS |
2677 | |
2678 | error_unlock: | |
2679 | sock_put(sk); | |
2680 | mutex_unlock(nlk->cb_mutex); | |
2681 | error_free: | |
2682 | kfree_skb(skb); | |
2683 | return ret; | |
1da177e4 | 2684 | } |
6dc878a8 | 2685 | EXPORT_SYMBOL(__netlink_dump_start); |
1da177e4 LT |
2686 | |
2687 | void netlink_ack(struct sk_buff *in_skb, struct nlmsghdr *nlh, int err) | |
2688 | { | |
2689 | struct sk_buff *skb; | |
2690 | struct nlmsghdr *rep; | |
2691 | struct nlmsgerr *errmsg; | |
339bf98f | 2692 | size_t payload = sizeof(*errmsg); |
1da177e4 | 2693 | |
339bf98f TG |
2694 | /* error messages get the original request appened */ |
2695 | if (err) | |
2696 | payload += nlmsg_len(nlh); | |
1da177e4 | 2697 | |
f9c22888 PM |
2698 | skb = netlink_alloc_skb(in_skb->sk, nlmsg_total_size(payload), |
2699 | NETLINK_CB(in_skb).portid, GFP_KERNEL); | |
1da177e4 LT |
2700 | if (!skb) { |
2701 | struct sock *sk; | |
2702 | ||
3b1e0a65 | 2703 | sk = netlink_lookup(sock_net(in_skb->sk), |
b4b51029 | 2704 | in_skb->sk->sk_protocol, |
15e47304 | 2705 | NETLINK_CB(in_skb).portid); |
1da177e4 LT |
2706 | if (sk) { |
2707 | sk->sk_err = ENOBUFS; | |
2708 | sk->sk_error_report(sk); | |
2709 | sock_put(sk); | |
2710 | } | |
2711 | return; | |
2712 | } | |
2713 | ||
15e47304 | 2714 | rep = __nlmsg_put(skb, NETLINK_CB(in_skb).portid, nlh->nlmsg_seq, |
5dba93ae | 2715 | NLMSG_ERROR, payload, 0); |
bf8b79e4 | 2716 | errmsg = nlmsg_data(rep); |
1da177e4 | 2717 | errmsg->error = err; |
bf8b79e4 | 2718 | memcpy(&errmsg->msg, nlh, err ? nlh->nlmsg_len : sizeof(*nlh)); |
15e47304 | 2719 | netlink_unicast(in_skb->sk, skb, NETLINK_CB(in_skb).portid, MSG_DONTWAIT); |
1da177e4 | 2720 | } |
6ac552fd | 2721 | EXPORT_SYMBOL(netlink_ack); |
1da177e4 | 2722 | |
cd40b7d3 | 2723 | int netlink_rcv_skb(struct sk_buff *skb, int (*cb)(struct sk_buff *, |
1d00a4eb | 2724 | struct nlmsghdr *)) |
82ace47a | 2725 | { |
82ace47a TG |
2726 | struct nlmsghdr *nlh; |
2727 | int err; | |
2728 | ||
2729 | while (skb->len >= nlmsg_total_size(0)) { | |
cd40b7d3 DL |
2730 | int msglen; |
2731 | ||
b529ccf2 | 2732 | nlh = nlmsg_hdr(skb); |
d35b6856 | 2733 | err = 0; |
82ace47a | 2734 | |
ad8e4b75 | 2735 | if (nlh->nlmsg_len < NLMSG_HDRLEN || skb->len < nlh->nlmsg_len) |
82ace47a TG |
2736 | return 0; |
2737 | ||
d35b6856 TG |
2738 | /* Only requests are handled by the kernel */ |
2739 | if (!(nlh->nlmsg_flags & NLM_F_REQUEST)) | |
5c58298c | 2740 | goto ack; |
45e7ae7f TG |
2741 | |
2742 | /* Skip control messages */ | |
2743 | if (nlh->nlmsg_type < NLMSG_MIN_TYPE) | |
5c58298c | 2744 | goto ack; |
d35b6856 | 2745 | |
1d00a4eb | 2746 | err = cb(skb, nlh); |
5c58298c DL |
2747 | if (err == -EINTR) |
2748 | goto skip; | |
2749 | ||
2750 | ack: | |
d35b6856 | 2751 | if (nlh->nlmsg_flags & NLM_F_ACK || err) |
82ace47a | 2752 | netlink_ack(skb, nlh, err); |
82ace47a | 2753 | |
5c58298c | 2754 | skip: |
6ac552fd | 2755 | msglen = NLMSG_ALIGN(nlh->nlmsg_len); |
cd40b7d3 DL |
2756 | if (msglen > skb->len) |
2757 | msglen = skb->len; | |
2758 | skb_pull(skb, msglen); | |
82ace47a TG |
2759 | } |
2760 | ||
2761 | return 0; | |
2762 | } | |
6ac552fd | 2763 | EXPORT_SYMBOL(netlink_rcv_skb); |
82ace47a | 2764 | |
d387f6ad TG |
2765 | /** |
2766 | * nlmsg_notify - send a notification netlink message | |
2767 | * @sk: netlink socket to use | |
2768 | * @skb: notification message | |
15e47304 | 2769 | * @portid: destination netlink portid for reports or 0 |
d387f6ad TG |
2770 | * @group: destination multicast group or 0 |
2771 | * @report: 1 to report back, 0 to disable | |
2772 | * @flags: allocation flags | |
2773 | */ | |
15e47304 | 2774 | int nlmsg_notify(struct sock *sk, struct sk_buff *skb, u32 portid, |
d387f6ad TG |
2775 | unsigned int group, int report, gfp_t flags) |
2776 | { | |
2777 | int err = 0; | |
2778 | ||
2779 | if (group) { | |
15e47304 | 2780 | int exclude_portid = 0; |
d387f6ad TG |
2781 | |
2782 | if (report) { | |
2783 | atomic_inc(&skb->users); | |
15e47304 | 2784 | exclude_portid = portid; |
d387f6ad TG |
2785 | } |
2786 | ||
1ce85fe4 PNA |
2787 | /* errors reported via destination sk->sk_err, but propagate |
2788 | * delivery errors if NETLINK_BROADCAST_ERROR flag is set */ | |
15e47304 | 2789 | err = nlmsg_multicast(sk, skb, exclude_portid, group, flags); |
d387f6ad TG |
2790 | } |
2791 | ||
1ce85fe4 PNA |
2792 | if (report) { |
2793 | int err2; | |
2794 | ||
15e47304 | 2795 | err2 = nlmsg_unicast(sk, skb, portid); |
1ce85fe4 PNA |
2796 | if (!err || err == -ESRCH) |
2797 | err = err2; | |
2798 | } | |
d387f6ad TG |
2799 | |
2800 | return err; | |
2801 | } | |
6ac552fd | 2802 | EXPORT_SYMBOL(nlmsg_notify); |
d387f6ad | 2803 | |
1da177e4 LT |
2804 | #ifdef CONFIG_PROC_FS |
2805 | struct nl_seq_iter { | |
e372c414 | 2806 | struct seq_net_private p; |
1da177e4 LT |
2807 | int link; |
2808 | int hash_idx; | |
2809 | }; | |
2810 | ||
2811 | static struct sock *netlink_seq_socket_idx(struct seq_file *seq, loff_t pos) | |
2812 | { | |
2813 | struct nl_seq_iter *iter = seq->private; | |
2814 | int i, j; | |
2815 | struct sock *s; | |
1da177e4 LT |
2816 | loff_t off = 0; |
2817 | ||
6ac552fd | 2818 | for (i = 0; i < MAX_LINKS; i++) { |
15e47304 | 2819 | struct nl_portid_hash *hash = &nl_table[i].hash; |
1da177e4 LT |
2820 | |
2821 | for (j = 0; j <= hash->mask; j++) { | |
b67bfe0d | 2822 | sk_for_each(s, &hash->table[j]) { |
1218854a | 2823 | if (sock_net(s) != seq_file_net(seq)) |
b4b51029 | 2824 | continue; |
1da177e4 LT |
2825 | if (off == pos) { |
2826 | iter->link = i; | |
2827 | iter->hash_idx = j; | |
2828 | return s; | |
2829 | } | |
2830 | ++off; | |
2831 | } | |
2832 | } | |
2833 | } | |
2834 | return NULL; | |
2835 | } | |
2836 | ||
2837 | static void *netlink_seq_start(struct seq_file *seq, loff_t *pos) | |
9a429c49 | 2838 | __acquires(nl_table_lock) |
1da177e4 LT |
2839 | { |
2840 | read_lock(&nl_table_lock); | |
2841 | return *pos ? netlink_seq_socket_idx(seq, *pos - 1) : SEQ_START_TOKEN; | |
2842 | } | |
2843 | ||
2844 | static void *netlink_seq_next(struct seq_file *seq, void *v, loff_t *pos) | |
2845 | { | |
2846 | struct sock *s; | |
2847 | struct nl_seq_iter *iter; | |
da12c90e | 2848 | struct net *net; |
1da177e4 LT |
2849 | int i, j; |
2850 | ||
2851 | ++*pos; | |
2852 | ||
2853 | if (v == SEQ_START_TOKEN) | |
2854 | return netlink_seq_socket_idx(seq, 0); | |
746fac4d | 2855 | |
da12c90e | 2856 | net = seq_file_net(seq); |
b4b51029 EB |
2857 | iter = seq->private; |
2858 | s = v; | |
2859 | do { | |
2860 | s = sk_next(s); | |
da12c90e | 2861 | } while (s && !nl_table[s->sk_protocol].compare(net, s)); |
1da177e4 LT |
2862 | if (s) |
2863 | return s; | |
2864 | ||
1da177e4 LT |
2865 | i = iter->link; |
2866 | j = iter->hash_idx + 1; | |
2867 | ||
2868 | do { | |
15e47304 | 2869 | struct nl_portid_hash *hash = &nl_table[i].hash; |
1da177e4 LT |
2870 | |
2871 | for (; j <= hash->mask; j++) { | |
2872 | s = sk_head(&hash->table[j]); | |
da12c90e G |
2873 | |
2874 | while (s && !nl_table[s->sk_protocol].compare(net, s)) | |
b4b51029 | 2875 | s = sk_next(s); |
1da177e4 LT |
2876 | if (s) { |
2877 | iter->link = i; | |
2878 | iter->hash_idx = j; | |
2879 | return s; | |
2880 | } | |
2881 | } | |
2882 | ||
2883 | j = 0; | |
2884 | } while (++i < MAX_LINKS); | |
2885 | ||
2886 | return NULL; | |
2887 | } | |
2888 | ||
2889 | static void netlink_seq_stop(struct seq_file *seq, void *v) | |
9a429c49 | 2890 | __releases(nl_table_lock) |
1da177e4 LT |
2891 | { |
2892 | read_unlock(&nl_table_lock); | |
2893 | } | |
2894 | ||
2895 | ||
2896 | static int netlink_seq_show(struct seq_file *seq, void *v) | |
2897 | { | |
658cb354 | 2898 | if (v == SEQ_START_TOKEN) { |
1da177e4 LT |
2899 | seq_puts(seq, |
2900 | "sk Eth Pid Groups " | |
cf0aa4e0 | 2901 | "Rmem Wmem Dump Locks Drops Inode\n"); |
658cb354 | 2902 | } else { |
1da177e4 LT |
2903 | struct sock *s = v; |
2904 | struct netlink_sock *nlk = nlk_sk(s); | |
2905 | ||
16b304f3 | 2906 | seq_printf(seq, "%pK %-3d %-6u %08x %-8d %-8d %d %-8d %-8d %-8lu\n", |
1da177e4 LT |
2907 | s, |
2908 | s->sk_protocol, | |
15e47304 | 2909 | nlk->portid, |
513c2500 | 2910 | nlk->groups ? (u32)nlk->groups[0] : 0, |
31e6d363 ED |
2911 | sk_rmem_alloc_get(s), |
2912 | sk_wmem_alloc_get(s), | |
16b304f3 | 2913 | nlk->cb_running, |
38938bfe | 2914 | atomic_read(&s->sk_refcnt), |
cf0aa4e0 MY |
2915 | atomic_read(&s->sk_drops), |
2916 | sock_i_ino(s) | |
1da177e4 LT |
2917 | ); |
2918 | ||
2919 | } | |
2920 | return 0; | |
2921 | } | |
2922 | ||
56b3d975 | 2923 | static const struct seq_operations netlink_seq_ops = { |
1da177e4 LT |
2924 | .start = netlink_seq_start, |
2925 | .next = netlink_seq_next, | |
2926 | .stop = netlink_seq_stop, | |
2927 | .show = netlink_seq_show, | |
2928 | }; | |
2929 | ||
2930 | ||
2931 | static int netlink_seq_open(struct inode *inode, struct file *file) | |
2932 | { | |
e372c414 DL |
2933 | return seq_open_net(inode, file, &netlink_seq_ops, |
2934 | sizeof(struct nl_seq_iter)); | |
b4b51029 EB |
2935 | } |
2936 | ||
da7071d7 | 2937 | static const struct file_operations netlink_seq_fops = { |
1da177e4 LT |
2938 | .owner = THIS_MODULE, |
2939 | .open = netlink_seq_open, | |
2940 | .read = seq_read, | |
2941 | .llseek = seq_lseek, | |
e372c414 | 2942 | .release = seq_release_net, |
1da177e4 LT |
2943 | }; |
2944 | ||
2945 | #endif | |
2946 | ||
2947 | int netlink_register_notifier(struct notifier_block *nb) | |
2948 | { | |
e041c683 | 2949 | return atomic_notifier_chain_register(&netlink_chain, nb); |
1da177e4 | 2950 | } |
6ac552fd | 2951 | EXPORT_SYMBOL(netlink_register_notifier); |
1da177e4 LT |
2952 | |
2953 | int netlink_unregister_notifier(struct notifier_block *nb) | |
2954 | { | |
e041c683 | 2955 | return atomic_notifier_chain_unregister(&netlink_chain, nb); |
1da177e4 | 2956 | } |
6ac552fd | 2957 | EXPORT_SYMBOL(netlink_unregister_notifier); |
746fac4d | 2958 | |
90ddc4f0 | 2959 | static const struct proto_ops netlink_ops = { |
1da177e4 LT |
2960 | .family = PF_NETLINK, |
2961 | .owner = THIS_MODULE, | |
2962 | .release = netlink_release, | |
2963 | .bind = netlink_bind, | |
2964 | .connect = netlink_connect, | |
2965 | .socketpair = sock_no_socketpair, | |
2966 | .accept = sock_no_accept, | |
2967 | .getname = netlink_getname, | |
9652e931 | 2968 | .poll = netlink_poll, |
1da177e4 LT |
2969 | .ioctl = sock_no_ioctl, |
2970 | .listen = sock_no_listen, | |
2971 | .shutdown = sock_no_shutdown, | |
9a4595bc PM |
2972 | .setsockopt = netlink_setsockopt, |
2973 | .getsockopt = netlink_getsockopt, | |
1da177e4 LT |
2974 | .sendmsg = netlink_sendmsg, |
2975 | .recvmsg = netlink_recvmsg, | |
ccdfcc39 | 2976 | .mmap = netlink_mmap, |
1da177e4 LT |
2977 | .sendpage = sock_no_sendpage, |
2978 | }; | |
2979 | ||
ec1b4cf7 | 2980 | static const struct net_proto_family netlink_family_ops = { |
1da177e4 LT |
2981 | .family = PF_NETLINK, |
2982 | .create = netlink_create, | |
2983 | .owner = THIS_MODULE, /* for consistency 8) */ | |
2984 | }; | |
2985 | ||
4665079c | 2986 | static int __net_init netlink_net_init(struct net *net) |
b4b51029 EB |
2987 | { |
2988 | #ifdef CONFIG_PROC_FS | |
d4beaa66 | 2989 | if (!proc_create("netlink", 0, net->proc_net, &netlink_seq_fops)) |
b4b51029 EB |
2990 | return -ENOMEM; |
2991 | #endif | |
2992 | return 0; | |
2993 | } | |
2994 | ||
4665079c | 2995 | static void __net_exit netlink_net_exit(struct net *net) |
b4b51029 EB |
2996 | { |
2997 | #ifdef CONFIG_PROC_FS | |
ece31ffd | 2998 | remove_proc_entry("netlink", net->proc_net); |
b4b51029 EB |
2999 | #endif |
3000 | } | |
3001 | ||
b963ea89 DM |
3002 | static void __init netlink_add_usersock_entry(void) |
3003 | { | |
5c398dc8 | 3004 | struct listeners *listeners; |
b963ea89 DM |
3005 | int groups = 32; |
3006 | ||
5c398dc8 | 3007 | listeners = kzalloc(sizeof(*listeners) + NLGRPSZ(groups), GFP_KERNEL); |
b963ea89 | 3008 | if (!listeners) |
5c398dc8 | 3009 | panic("netlink_add_usersock_entry: Cannot allocate listeners\n"); |
b963ea89 DM |
3010 | |
3011 | netlink_table_grab(); | |
3012 | ||
3013 | nl_table[NETLINK_USERSOCK].groups = groups; | |
5c398dc8 | 3014 | rcu_assign_pointer(nl_table[NETLINK_USERSOCK].listeners, listeners); |
b963ea89 DM |
3015 | nl_table[NETLINK_USERSOCK].module = THIS_MODULE; |
3016 | nl_table[NETLINK_USERSOCK].registered = 1; | |
9785e10a | 3017 | nl_table[NETLINK_USERSOCK].flags = NL_CFG_F_NONROOT_SEND; |
b963ea89 DM |
3018 | |
3019 | netlink_table_ungrab(); | |
3020 | } | |
3021 | ||
022cbae6 | 3022 | static struct pernet_operations __net_initdata netlink_net_ops = { |
b4b51029 EB |
3023 | .init = netlink_net_init, |
3024 | .exit = netlink_net_exit, | |
3025 | }; | |
3026 | ||
1da177e4 LT |
3027 | static int __init netlink_proto_init(void) |
3028 | { | |
1da177e4 | 3029 | int i; |
26ff5ddc | 3030 | unsigned long limit; |
1da177e4 LT |
3031 | unsigned int order; |
3032 | int err = proto_register(&netlink_proto, 0); | |
3033 | ||
3034 | if (err != 0) | |
3035 | goto out; | |
3036 | ||
fab25745 | 3037 | BUILD_BUG_ON(sizeof(struct netlink_skb_parms) > FIELD_SIZEOF(struct sk_buff, cb)); |
1da177e4 | 3038 | |
0da974f4 | 3039 | nl_table = kcalloc(MAX_LINKS, sizeof(*nl_table), GFP_KERNEL); |
fab2caf6 AM |
3040 | if (!nl_table) |
3041 | goto panic; | |
1da177e4 | 3042 | |
4481374c JB |
3043 | if (totalram_pages >= (128 * 1024)) |
3044 | limit = totalram_pages >> (21 - PAGE_SHIFT); | |
1da177e4 | 3045 | else |
4481374c | 3046 | limit = totalram_pages >> (23 - PAGE_SHIFT); |
1da177e4 | 3047 | |
26ff5ddc DC |
3048 | order = get_bitmask_order(limit) - 1 + PAGE_SHIFT; |
3049 | limit = (1UL << order) / sizeof(struct hlist_head); | |
3050 | order = get_bitmask_order(min(limit, (unsigned long)UINT_MAX)) - 1; | |
1da177e4 LT |
3051 | |
3052 | for (i = 0; i < MAX_LINKS; i++) { | |
15e47304 | 3053 | struct nl_portid_hash *hash = &nl_table[i].hash; |
1da177e4 | 3054 | |
15e47304 | 3055 | hash->table = nl_portid_hash_zalloc(1 * sizeof(*hash->table)); |
1da177e4 LT |
3056 | if (!hash->table) { |
3057 | while (i-- > 0) | |
15e47304 | 3058 | nl_portid_hash_free(nl_table[i].hash.table, |
1da177e4 LT |
3059 | 1 * sizeof(*hash->table)); |
3060 | kfree(nl_table); | |
fab2caf6 | 3061 | goto panic; |
1da177e4 | 3062 | } |
1da177e4 LT |
3063 | hash->max_shift = order; |
3064 | hash->shift = 0; | |
3065 | hash->mask = 0; | |
3066 | hash->rehash_time = jiffies; | |
ca15febf G |
3067 | |
3068 | nl_table[i].compare = netlink_compare; | |
1da177e4 LT |
3069 | } |
3070 | ||
bcbde0d4 DB |
3071 | INIT_LIST_HEAD(&netlink_tap_all); |
3072 | ||
b963ea89 DM |
3073 | netlink_add_usersock_entry(); |
3074 | ||
1da177e4 | 3075 | sock_register(&netlink_family_ops); |
b4b51029 | 3076 | register_pernet_subsys(&netlink_net_ops); |
746fac4d | 3077 | /* The netlink device handler may be needed early. */ |
1da177e4 LT |
3078 | rtnetlink_init(); |
3079 | out: | |
3080 | return err; | |
fab2caf6 AM |
3081 | panic: |
3082 | panic("netlink_init: Cannot allocate nl_table\n"); | |
1da177e4 LT |
3083 | } |
3084 | ||
1da177e4 | 3085 | core_initcall(netlink_proto_init); |