Commit | Line | Data |
---|---|---|
1da177e4 LT |
1 | /* |
2 | * NETLINK Kernel-user communication protocol. | |
3 | * | |
113aa838 | 4 | * Authors: Alan Cox <alan@lxorguk.ukuu.org.uk> |
1da177e4 | 5 | * Alexey Kuznetsov <kuznet@ms2.inr.ac.ru> |
cd1df525 | 6 | * Patrick McHardy <kaber@trash.net> |
1da177e4 LT |
7 | * |
8 | * This program is free software; you can redistribute it and/or | |
9 | * modify it under the terms of the GNU General Public License | |
10 | * as published by the Free Software Foundation; either version | |
11 | * 2 of the License, or (at your option) any later version. | |
746fac4d | 12 | * |
1da177e4 LT |
13 | * Tue Jun 26 14:36:48 MEST 2001 Herbert "herp" Rosmanith |
14 | * added netlink_proto_exit | |
15 | * Tue Jan 22 18:32:44 BRST 2002 Arnaldo C. de Melo <acme@conectiva.com.br> | |
16 | * use nlk_sk, as sk->protinfo is on a diet 8) | |
4fdb3bb7 HW |
17 | * Fri Jul 22 19:51:12 MEST 2005 Harald Welte <laforge@gnumonks.org> |
18 | * - inc module use count of module that owns | |
19 | * the kernel socket in case userspace opens | |
20 | * socket of same protocol | |
21 | * - remove all module support, since netlink is | |
22 | * mandatory if CONFIG_NET=y these days | |
1da177e4 LT |
23 | */ |
24 | ||
1da177e4 LT |
25 | #include <linux/module.h> |
26 | ||
4fc268d2 | 27 | #include <linux/capability.h> |
1da177e4 LT |
28 | #include <linux/kernel.h> |
29 | #include <linux/init.h> | |
1da177e4 LT |
30 | #include <linux/signal.h> |
31 | #include <linux/sched.h> | |
32 | #include <linux/errno.h> | |
33 | #include <linux/string.h> | |
34 | #include <linux/stat.h> | |
35 | #include <linux/socket.h> | |
36 | #include <linux/un.h> | |
37 | #include <linux/fcntl.h> | |
38 | #include <linux/termios.h> | |
39 | #include <linux/sockios.h> | |
40 | #include <linux/net.h> | |
41 | #include <linux/fs.h> | |
42 | #include <linux/slab.h> | |
43 | #include <asm/uaccess.h> | |
44 | #include <linux/skbuff.h> | |
45 | #include <linux/netdevice.h> | |
46 | #include <linux/rtnetlink.h> | |
47 | #include <linux/proc_fs.h> | |
48 | #include <linux/seq_file.h> | |
1da177e4 LT |
49 | #include <linux/notifier.h> |
50 | #include <linux/security.h> | |
51 | #include <linux/jhash.h> | |
52 | #include <linux/jiffies.h> | |
53 | #include <linux/random.h> | |
54 | #include <linux/bitops.h> | |
55 | #include <linux/mm.h> | |
56 | #include <linux/types.h> | |
54e0f520 | 57 | #include <linux/audit.h> |
af65bdfc | 58 | #include <linux/mutex.h> |
ccdfcc39 | 59 | #include <linux/vmalloc.h> |
bcbde0d4 | 60 | #include <linux/if_arp.h> |
e341694e | 61 | #include <linux/rhashtable.h> |
9652e931 | 62 | #include <asm/cacheflush.h> |
e341694e | 63 | #include <linux/hash.h> |
ee1c2442 | 64 | #include <linux/genetlink.h> |
54e0f520 | 65 | |
457c4cbc | 66 | #include <net/net_namespace.h> |
1da177e4 LT |
67 | #include <net/sock.h> |
68 | #include <net/scm.h> | |
82ace47a | 69 | #include <net/netlink.h> |
1da177e4 | 70 | |
0f29c768 | 71 | #include "af_netlink.h" |
1da177e4 | 72 | |
5c398dc8 ED |
73 | struct listeners { |
74 | struct rcu_head rcu; | |
75 | unsigned long masks[0]; | |
6c04bb18 JB |
76 | }; |
77 | ||
cd967e05 PM |
78 | /* state bits */ |
79 | #define NETLINK_CONGESTED 0x0 | |
80 | ||
81 | /* flags */ | |
77247bbb | 82 | #define NETLINK_KERNEL_SOCKET 0x1 |
9a4595bc | 83 | #define NETLINK_RECV_PKTINFO 0x2 |
be0c22a4 | 84 | #define NETLINK_BROADCAST_SEND_ERROR 0x4 |
38938bfe | 85 | #define NETLINK_RECV_NO_ENOBUFS 0x8 |
77247bbb | 86 | |
035c4c16 | 87 | static inline int netlink_is_kernel(struct sock *sk) |
aed81560 DL |
88 | { |
89 | return nlk_sk(sk)->flags & NETLINK_KERNEL_SOCKET; | |
90 | } | |
91 | ||
0f29c768 AV |
92 | struct netlink_table *nl_table; |
93 | EXPORT_SYMBOL_GPL(nl_table); | |
1da177e4 LT |
94 | |
95 | static DECLARE_WAIT_QUEUE_HEAD(nl_table_wait); | |
96 | ||
97 | static int netlink_dump(struct sock *sk); | |
9652e931 | 98 | static void netlink_skb_destructor(struct sk_buff *skb); |
1da177e4 | 99 | |
78fd1d0a | 100 | /* nl_table locking explained: |
21e4902a | 101 | * Lookup and traversal are protected with an RCU read-side lock. Insertion |
c5adde94 | 102 | * and removal are protected with per bucket lock while using RCU list |
21e4902a TG |
103 | * modification primitives and may run in parallel to RCU protected lookups. |
104 | * Destruction of the Netlink socket may only occur *after* nl_table_lock has | |
105 | * been acquired * either during or after the socket has been removed from | |
106 | * the list and after an RCU grace period. | |
78fd1d0a | 107 | */ |
0f29c768 AV |
108 | DEFINE_RWLOCK(nl_table_lock); |
109 | EXPORT_SYMBOL_GPL(nl_table_lock); | |
1da177e4 LT |
110 | static atomic_t nl_table_users = ATOMIC_INIT(0); |
111 | ||
6d772ac5 ED |
112 | #define nl_deref_protected(X) rcu_dereference_protected(X, lockdep_is_held(&nl_table_lock)); |
113 | ||
e041c683 | 114 | static ATOMIC_NOTIFIER_HEAD(netlink_chain); |
1da177e4 | 115 | |
bcbde0d4 DB |
116 | static DEFINE_SPINLOCK(netlink_tap_lock); |
117 | static struct list_head netlink_tap_all __read_mostly; | |
118 | ||
c428ecd1 HX |
119 | static const struct rhashtable_params netlink_rhashtable_params; |
120 | ||
b57ef81f | 121 | static inline u32 netlink_group_mask(u32 group) |
d629b836 PM |
122 | { |
123 | return group ? 1 << (group - 1) : 0; | |
124 | } | |
125 | ||
bcbde0d4 DB |
126 | int netlink_add_tap(struct netlink_tap *nt) |
127 | { | |
128 | if (unlikely(nt->dev->type != ARPHRD_NETLINK)) | |
129 | return -EINVAL; | |
130 | ||
131 | spin_lock(&netlink_tap_lock); | |
132 | list_add_rcu(&nt->list, &netlink_tap_all); | |
133 | spin_unlock(&netlink_tap_lock); | |
134 | ||
fcd4d35e | 135 | __module_get(nt->module); |
bcbde0d4 DB |
136 | |
137 | return 0; | |
138 | } | |
139 | EXPORT_SYMBOL_GPL(netlink_add_tap); | |
140 | ||
2173f8d9 | 141 | static int __netlink_remove_tap(struct netlink_tap *nt) |
bcbde0d4 DB |
142 | { |
143 | bool found = false; | |
144 | struct netlink_tap *tmp; | |
145 | ||
146 | spin_lock(&netlink_tap_lock); | |
147 | ||
148 | list_for_each_entry(tmp, &netlink_tap_all, list) { | |
149 | if (nt == tmp) { | |
150 | list_del_rcu(&nt->list); | |
151 | found = true; | |
152 | goto out; | |
153 | } | |
154 | } | |
155 | ||
156 | pr_warn("__netlink_remove_tap: %p not found\n", nt); | |
157 | out: | |
158 | spin_unlock(&netlink_tap_lock); | |
159 | ||
160 | if (found && nt->module) | |
161 | module_put(nt->module); | |
162 | ||
163 | return found ? 0 : -ENODEV; | |
164 | } | |
bcbde0d4 DB |
165 | |
166 | int netlink_remove_tap(struct netlink_tap *nt) | |
167 | { | |
168 | int ret; | |
169 | ||
170 | ret = __netlink_remove_tap(nt); | |
171 | synchronize_net(); | |
172 | ||
173 | return ret; | |
174 | } | |
175 | EXPORT_SYMBOL_GPL(netlink_remove_tap); | |
176 | ||
5ffd5cdd DB |
177 | static bool netlink_filter_tap(const struct sk_buff *skb) |
178 | { | |
179 | struct sock *sk = skb->sk; | |
5ffd5cdd DB |
180 | |
181 | /* We take the more conservative approach and | |
182 | * whitelist socket protocols that may pass. | |
183 | */ | |
184 | switch (sk->sk_protocol) { | |
185 | case NETLINK_ROUTE: | |
186 | case NETLINK_USERSOCK: | |
187 | case NETLINK_SOCK_DIAG: | |
188 | case NETLINK_NFLOG: | |
189 | case NETLINK_XFRM: | |
190 | case NETLINK_FIB_LOOKUP: | |
191 | case NETLINK_NETFILTER: | |
192 | case NETLINK_GENERIC: | |
498044bb | 193 | return true; |
5ffd5cdd DB |
194 | } |
195 | ||
498044bb | 196 | return false; |
5ffd5cdd DB |
197 | } |
198 | ||
bcbde0d4 DB |
199 | static int __netlink_deliver_tap_skb(struct sk_buff *skb, |
200 | struct net_device *dev) | |
201 | { | |
202 | struct sk_buff *nskb; | |
5ffd5cdd | 203 | struct sock *sk = skb->sk; |
bcbde0d4 DB |
204 | int ret = -ENOMEM; |
205 | ||
206 | dev_hold(dev); | |
207 | nskb = skb_clone(skb, GFP_ATOMIC); | |
208 | if (nskb) { | |
209 | nskb->dev = dev; | |
5ffd5cdd | 210 | nskb->protocol = htons((u16) sk->sk_protocol); |
604d13c9 DB |
211 | nskb->pkt_type = netlink_is_kernel(sk) ? |
212 | PACKET_KERNEL : PACKET_USER; | |
4e48ed88 | 213 | skb_reset_network_header(nskb); |
bcbde0d4 DB |
214 | ret = dev_queue_xmit(nskb); |
215 | if (unlikely(ret > 0)) | |
216 | ret = net_xmit_errno(ret); | |
217 | } | |
218 | ||
219 | dev_put(dev); | |
220 | return ret; | |
221 | } | |
222 | ||
223 | static void __netlink_deliver_tap(struct sk_buff *skb) | |
224 | { | |
225 | int ret; | |
226 | struct netlink_tap *tmp; | |
227 | ||
5ffd5cdd DB |
228 | if (!netlink_filter_tap(skb)) |
229 | return; | |
230 | ||
bcbde0d4 DB |
231 | list_for_each_entry_rcu(tmp, &netlink_tap_all, list) { |
232 | ret = __netlink_deliver_tap_skb(skb, tmp->dev); | |
233 | if (unlikely(ret)) | |
234 | break; | |
235 | } | |
236 | } | |
237 | ||
238 | static void netlink_deliver_tap(struct sk_buff *skb) | |
239 | { | |
240 | rcu_read_lock(); | |
241 | ||
242 | if (unlikely(!list_empty(&netlink_tap_all))) | |
243 | __netlink_deliver_tap(skb); | |
244 | ||
245 | rcu_read_unlock(); | |
246 | } | |
247 | ||
73bfd370 DB |
248 | static void netlink_deliver_tap_kernel(struct sock *dst, struct sock *src, |
249 | struct sk_buff *skb) | |
250 | { | |
251 | if (!(netlink_is_kernel(dst) && netlink_is_kernel(src))) | |
252 | netlink_deliver_tap(skb); | |
253 | } | |
254 | ||
cd1df525 PM |
255 | static void netlink_overrun(struct sock *sk) |
256 | { | |
257 | struct netlink_sock *nlk = nlk_sk(sk); | |
258 | ||
259 | if (!(nlk->flags & NETLINK_RECV_NO_ENOBUFS)) { | |
260 | if (!test_and_set_bit(NETLINK_CONGESTED, &nlk_sk(sk)->state)) { | |
261 | sk->sk_err = ENOBUFS; | |
262 | sk->sk_error_report(sk); | |
263 | } | |
264 | } | |
265 | atomic_inc(&sk->sk_drops); | |
266 | } | |
267 | ||
268 | static void netlink_rcv_wake(struct sock *sk) | |
269 | { | |
270 | struct netlink_sock *nlk = nlk_sk(sk); | |
271 | ||
272 | if (skb_queue_empty(&sk->sk_receive_queue)) | |
273 | clear_bit(NETLINK_CONGESTED, &nlk->state); | |
274 | if (!test_bit(NETLINK_CONGESTED, &nlk->state)) | |
275 | wake_up_interruptible(&nlk->wait); | |
276 | } | |
277 | ||
ccdfcc39 | 278 | #ifdef CONFIG_NETLINK_MMAP |
9652e931 PM |
279 | static bool netlink_skb_is_mmaped(const struct sk_buff *skb) |
280 | { | |
281 | return NETLINK_CB(skb).flags & NETLINK_SKB_MMAPED; | |
282 | } | |
283 | ||
f9c22888 PM |
284 | static bool netlink_rx_is_mmaped(struct sock *sk) |
285 | { | |
286 | return nlk_sk(sk)->rx_ring.pg_vec != NULL; | |
287 | } | |
288 | ||
5fd96123 PM |
289 | static bool netlink_tx_is_mmaped(struct sock *sk) |
290 | { | |
291 | return nlk_sk(sk)->tx_ring.pg_vec != NULL; | |
292 | } | |
293 | ||
ccdfcc39 PM |
294 | static __pure struct page *pgvec_to_page(const void *addr) |
295 | { | |
296 | if (is_vmalloc_addr(addr)) | |
297 | return vmalloc_to_page(addr); | |
298 | else | |
299 | return virt_to_page(addr); | |
300 | } | |
301 | ||
302 | static void free_pg_vec(void **pg_vec, unsigned int order, unsigned int len) | |
303 | { | |
304 | unsigned int i; | |
305 | ||
306 | for (i = 0; i < len; i++) { | |
307 | if (pg_vec[i] != NULL) { | |
308 | if (is_vmalloc_addr(pg_vec[i])) | |
309 | vfree(pg_vec[i]); | |
310 | else | |
311 | free_pages((unsigned long)pg_vec[i], order); | |
312 | } | |
313 | } | |
314 | kfree(pg_vec); | |
315 | } | |
316 | ||
317 | static void *alloc_one_pg_vec_page(unsigned long order) | |
318 | { | |
319 | void *buffer; | |
320 | gfp_t gfp_flags = GFP_KERNEL | __GFP_COMP | __GFP_ZERO | | |
321 | __GFP_NOWARN | __GFP_NORETRY; | |
322 | ||
323 | buffer = (void *)__get_free_pages(gfp_flags, order); | |
324 | if (buffer != NULL) | |
325 | return buffer; | |
326 | ||
327 | buffer = vzalloc((1 << order) * PAGE_SIZE); | |
328 | if (buffer != NULL) | |
329 | return buffer; | |
330 | ||
331 | gfp_flags &= ~__GFP_NORETRY; | |
332 | return (void *)__get_free_pages(gfp_flags, order); | |
333 | } | |
334 | ||
335 | static void **alloc_pg_vec(struct netlink_sock *nlk, | |
336 | struct nl_mmap_req *req, unsigned int order) | |
337 | { | |
338 | unsigned int block_nr = req->nm_block_nr; | |
339 | unsigned int i; | |
8a849bb7 | 340 | void **pg_vec; |
ccdfcc39 PM |
341 | |
342 | pg_vec = kcalloc(block_nr, sizeof(void *), GFP_KERNEL); | |
343 | if (pg_vec == NULL) | |
344 | return NULL; | |
345 | ||
346 | for (i = 0; i < block_nr; i++) { | |
8a849bb7 | 347 | pg_vec[i] = alloc_one_pg_vec_page(order); |
ccdfcc39 PM |
348 | if (pg_vec[i] == NULL) |
349 | goto err1; | |
350 | } | |
351 | ||
352 | return pg_vec; | |
353 | err1: | |
354 | free_pg_vec(pg_vec, order, block_nr); | |
355 | return NULL; | |
356 | } | |
357 | ||
358 | static int netlink_set_ring(struct sock *sk, struct nl_mmap_req *req, | |
359 | bool closing, bool tx_ring) | |
360 | { | |
361 | struct netlink_sock *nlk = nlk_sk(sk); | |
362 | struct netlink_ring *ring; | |
363 | struct sk_buff_head *queue; | |
364 | void **pg_vec = NULL; | |
365 | unsigned int order = 0; | |
366 | int err; | |
367 | ||
368 | ring = tx_ring ? &nlk->tx_ring : &nlk->rx_ring; | |
369 | queue = tx_ring ? &sk->sk_write_queue : &sk->sk_receive_queue; | |
370 | ||
371 | if (!closing) { | |
372 | if (atomic_read(&nlk->mapped)) | |
373 | return -EBUSY; | |
374 | if (atomic_read(&ring->pending)) | |
375 | return -EBUSY; | |
376 | } | |
377 | ||
378 | if (req->nm_block_nr) { | |
379 | if (ring->pg_vec != NULL) | |
380 | return -EBUSY; | |
381 | ||
382 | if ((int)req->nm_block_size <= 0) | |
383 | return -EINVAL; | |
74e83b23 | 384 | if (!PAGE_ALIGNED(req->nm_block_size)) |
ccdfcc39 PM |
385 | return -EINVAL; |
386 | if (req->nm_frame_size < NL_MMAP_HDRLEN) | |
387 | return -EINVAL; | |
388 | if (!IS_ALIGNED(req->nm_frame_size, NL_MMAP_MSG_ALIGNMENT)) | |
389 | return -EINVAL; | |
390 | ||
391 | ring->frames_per_block = req->nm_block_size / | |
392 | req->nm_frame_size; | |
393 | if (ring->frames_per_block == 0) | |
394 | return -EINVAL; | |
395 | if (ring->frames_per_block * req->nm_block_nr != | |
396 | req->nm_frame_nr) | |
397 | return -EINVAL; | |
398 | ||
399 | order = get_order(req->nm_block_size); | |
400 | pg_vec = alloc_pg_vec(nlk, req, order); | |
401 | if (pg_vec == NULL) | |
402 | return -ENOMEM; | |
403 | } else { | |
404 | if (req->nm_frame_nr) | |
405 | return -EINVAL; | |
406 | } | |
407 | ||
408 | err = -EBUSY; | |
409 | mutex_lock(&nlk->pg_vec_lock); | |
410 | if (closing || atomic_read(&nlk->mapped) == 0) { | |
411 | err = 0; | |
412 | spin_lock_bh(&queue->lock); | |
413 | ||
414 | ring->frame_max = req->nm_frame_nr - 1; | |
415 | ring->head = 0; | |
416 | ring->frame_size = req->nm_frame_size; | |
417 | ring->pg_vec_pages = req->nm_block_size / PAGE_SIZE; | |
418 | ||
419 | swap(ring->pg_vec_len, req->nm_block_nr); | |
420 | swap(ring->pg_vec_order, order); | |
421 | swap(ring->pg_vec, pg_vec); | |
422 | ||
423 | __skb_queue_purge(queue); | |
424 | spin_unlock_bh(&queue->lock); | |
425 | ||
426 | WARN_ON(atomic_read(&nlk->mapped)); | |
427 | } | |
428 | mutex_unlock(&nlk->pg_vec_lock); | |
429 | ||
430 | if (pg_vec) | |
431 | free_pg_vec(pg_vec, order, req->nm_block_nr); | |
432 | return err; | |
433 | } | |
434 | ||
435 | static void netlink_mm_open(struct vm_area_struct *vma) | |
436 | { | |
437 | struct file *file = vma->vm_file; | |
438 | struct socket *sock = file->private_data; | |
439 | struct sock *sk = sock->sk; | |
440 | ||
441 | if (sk) | |
442 | atomic_inc(&nlk_sk(sk)->mapped); | |
443 | } | |
444 | ||
445 | static void netlink_mm_close(struct vm_area_struct *vma) | |
446 | { | |
447 | struct file *file = vma->vm_file; | |
448 | struct socket *sock = file->private_data; | |
449 | struct sock *sk = sock->sk; | |
450 | ||
451 | if (sk) | |
452 | atomic_dec(&nlk_sk(sk)->mapped); | |
453 | } | |
454 | ||
455 | static const struct vm_operations_struct netlink_mmap_ops = { | |
456 | .open = netlink_mm_open, | |
457 | .close = netlink_mm_close, | |
458 | }; | |
459 | ||
460 | static int netlink_mmap(struct file *file, struct socket *sock, | |
461 | struct vm_area_struct *vma) | |
462 | { | |
463 | struct sock *sk = sock->sk; | |
464 | struct netlink_sock *nlk = nlk_sk(sk); | |
465 | struct netlink_ring *ring; | |
466 | unsigned long start, size, expected; | |
467 | unsigned int i; | |
468 | int err = -EINVAL; | |
469 | ||
470 | if (vma->vm_pgoff) | |
471 | return -EINVAL; | |
472 | ||
473 | mutex_lock(&nlk->pg_vec_lock); | |
474 | ||
475 | expected = 0; | |
476 | for (ring = &nlk->rx_ring; ring <= &nlk->tx_ring; ring++) { | |
477 | if (ring->pg_vec == NULL) | |
478 | continue; | |
479 | expected += ring->pg_vec_len * ring->pg_vec_pages * PAGE_SIZE; | |
480 | } | |
481 | ||
482 | if (expected == 0) | |
483 | goto out; | |
484 | ||
485 | size = vma->vm_end - vma->vm_start; | |
486 | if (size != expected) | |
487 | goto out; | |
488 | ||
489 | start = vma->vm_start; | |
490 | for (ring = &nlk->rx_ring; ring <= &nlk->tx_ring; ring++) { | |
491 | if (ring->pg_vec == NULL) | |
492 | continue; | |
493 | ||
494 | for (i = 0; i < ring->pg_vec_len; i++) { | |
495 | struct page *page; | |
496 | void *kaddr = ring->pg_vec[i]; | |
497 | unsigned int pg_num; | |
498 | ||
499 | for (pg_num = 0; pg_num < ring->pg_vec_pages; pg_num++) { | |
500 | page = pgvec_to_page(kaddr); | |
501 | err = vm_insert_page(vma, start, page); | |
502 | if (err < 0) | |
503 | goto out; | |
504 | start += PAGE_SIZE; | |
505 | kaddr += PAGE_SIZE; | |
506 | } | |
507 | } | |
508 | } | |
509 | ||
510 | atomic_inc(&nlk->mapped); | |
511 | vma->vm_ops = &netlink_mmap_ops; | |
512 | err = 0; | |
513 | out: | |
514 | mutex_unlock(&nlk->pg_vec_lock); | |
7cdbac71 | 515 | return err; |
ccdfcc39 | 516 | } |
9652e931 | 517 | |
4682a035 | 518 | static void netlink_frame_flush_dcache(const struct nl_mmap_hdr *hdr, unsigned int nm_len) |
9652e931 PM |
519 | { |
520 | #if ARCH_IMPLEMENTS_FLUSH_DCACHE_PAGE == 1 | |
521 | struct page *p_start, *p_end; | |
522 | ||
523 | /* First page is flushed through netlink_{get,set}_status */ | |
524 | p_start = pgvec_to_page(hdr + PAGE_SIZE); | |
4682a035 | 525 | p_end = pgvec_to_page((void *)hdr + NL_MMAP_HDRLEN + nm_len - 1); |
9652e931 PM |
526 | while (p_start <= p_end) { |
527 | flush_dcache_page(p_start); | |
528 | p_start++; | |
529 | } | |
530 | #endif | |
531 | } | |
532 | ||
533 | static enum nl_mmap_status netlink_get_status(const struct nl_mmap_hdr *hdr) | |
534 | { | |
535 | smp_rmb(); | |
536 | flush_dcache_page(pgvec_to_page(hdr)); | |
537 | return hdr->nm_status; | |
538 | } | |
539 | ||
540 | static void netlink_set_status(struct nl_mmap_hdr *hdr, | |
541 | enum nl_mmap_status status) | |
542 | { | |
a18e6a18 | 543 | smp_mb(); |
9652e931 PM |
544 | hdr->nm_status = status; |
545 | flush_dcache_page(pgvec_to_page(hdr)); | |
9652e931 PM |
546 | } |
547 | ||
548 | static struct nl_mmap_hdr * | |
549 | __netlink_lookup_frame(const struct netlink_ring *ring, unsigned int pos) | |
550 | { | |
551 | unsigned int pg_vec_pos, frame_off; | |
552 | ||
553 | pg_vec_pos = pos / ring->frames_per_block; | |
554 | frame_off = pos % ring->frames_per_block; | |
555 | ||
556 | return ring->pg_vec[pg_vec_pos] + (frame_off * ring->frame_size); | |
557 | } | |
558 | ||
559 | static struct nl_mmap_hdr * | |
560 | netlink_lookup_frame(const struct netlink_ring *ring, unsigned int pos, | |
561 | enum nl_mmap_status status) | |
562 | { | |
563 | struct nl_mmap_hdr *hdr; | |
564 | ||
565 | hdr = __netlink_lookup_frame(ring, pos); | |
566 | if (netlink_get_status(hdr) != status) | |
567 | return NULL; | |
568 | ||
569 | return hdr; | |
570 | } | |
571 | ||
572 | static struct nl_mmap_hdr * | |
573 | netlink_current_frame(const struct netlink_ring *ring, | |
574 | enum nl_mmap_status status) | |
575 | { | |
576 | return netlink_lookup_frame(ring, ring->head, status); | |
577 | } | |
578 | ||
579 | static struct nl_mmap_hdr * | |
580 | netlink_previous_frame(const struct netlink_ring *ring, | |
581 | enum nl_mmap_status status) | |
582 | { | |
583 | unsigned int prev; | |
584 | ||
585 | prev = ring->head ? ring->head - 1 : ring->frame_max; | |
586 | return netlink_lookup_frame(ring, prev, status); | |
587 | } | |
588 | ||
589 | static void netlink_increment_head(struct netlink_ring *ring) | |
590 | { | |
591 | ring->head = ring->head != ring->frame_max ? ring->head + 1 : 0; | |
592 | } | |
593 | ||
594 | static void netlink_forward_ring(struct netlink_ring *ring) | |
595 | { | |
596 | unsigned int head = ring->head, pos = head; | |
597 | const struct nl_mmap_hdr *hdr; | |
598 | ||
599 | do { | |
600 | hdr = __netlink_lookup_frame(ring, pos); | |
601 | if (hdr->nm_status == NL_MMAP_STATUS_UNUSED) | |
602 | break; | |
603 | if (hdr->nm_status != NL_MMAP_STATUS_SKIP) | |
604 | break; | |
605 | netlink_increment_head(ring); | |
606 | } while (ring->head != head); | |
607 | } | |
608 | ||
cd1df525 PM |
609 | static bool netlink_dump_space(struct netlink_sock *nlk) |
610 | { | |
611 | struct netlink_ring *ring = &nlk->rx_ring; | |
612 | struct nl_mmap_hdr *hdr; | |
613 | unsigned int n; | |
614 | ||
615 | hdr = netlink_current_frame(ring, NL_MMAP_STATUS_UNUSED); | |
616 | if (hdr == NULL) | |
617 | return false; | |
618 | ||
619 | n = ring->head + ring->frame_max / 2; | |
620 | if (n > ring->frame_max) | |
621 | n -= ring->frame_max; | |
622 | ||
623 | hdr = __netlink_lookup_frame(ring, n); | |
624 | ||
625 | return hdr->nm_status == NL_MMAP_STATUS_UNUSED; | |
626 | } | |
627 | ||
9652e931 PM |
628 | static unsigned int netlink_poll(struct file *file, struct socket *sock, |
629 | poll_table *wait) | |
630 | { | |
631 | struct sock *sk = sock->sk; | |
632 | struct netlink_sock *nlk = nlk_sk(sk); | |
633 | unsigned int mask; | |
cd1df525 | 634 | int err; |
9652e931 | 635 | |
cd1df525 PM |
636 | if (nlk->rx_ring.pg_vec != NULL) { |
637 | /* Memory mapped sockets don't call recvmsg(), so flow control | |
638 | * for dumps is performed here. A dump is allowed to continue | |
639 | * if at least half the ring is unused. | |
640 | */ | |
16b304f3 | 641 | while (nlk->cb_running && netlink_dump_space(nlk)) { |
cd1df525 PM |
642 | err = netlink_dump(sk); |
643 | if (err < 0) { | |
ac30ef83 | 644 | sk->sk_err = -err; |
cd1df525 PM |
645 | sk->sk_error_report(sk); |
646 | break; | |
647 | } | |
648 | } | |
649 | netlink_rcv_wake(sk); | |
650 | } | |
5fd96123 | 651 | |
9652e931 PM |
652 | mask = datagram_poll(file, sock, wait); |
653 | ||
654 | spin_lock_bh(&sk->sk_receive_queue.lock); | |
655 | if (nlk->rx_ring.pg_vec) { | |
656 | netlink_forward_ring(&nlk->rx_ring); | |
657 | if (!netlink_previous_frame(&nlk->rx_ring, NL_MMAP_STATUS_UNUSED)) | |
658 | mask |= POLLIN | POLLRDNORM; | |
659 | } | |
660 | spin_unlock_bh(&sk->sk_receive_queue.lock); | |
661 | ||
662 | spin_lock_bh(&sk->sk_write_queue.lock); | |
663 | if (nlk->tx_ring.pg_vec) { | |
664 | if (netlink_current_frame(&nlk->tx_ring, NL_MMAP_STATUS_UNUSED)) | |
665 | mask |= POLLOUT | POLLWRNORM; | |
666 | } | |
667 | spin_unlock_bh(&sk->sk_write_queue.lock); | |
668 | ||
669 | return mask; | |
670 | } | |
671 | ||
672 | static struct nl_mmap_hdr *netlink_mmap_hdr(struct sk_buff *skb) | |
673 | { | |
674 | return (struct nl_mmap_hdr *)(skb->head - NL_MMAP_HDRLEN); | |
675 | } | |
676 | ||
677 | static void netlink_ring_setup_skb(struct sk_buff *skb, struct sock *sk, | |
678 | struct netlink_ring *ring, | |
679 | struct nl_mmap_hdr *hdr) | |
680 | { | |
681 | unsigned int size; | |
682 | void *data; | |
683 | ||
684 | size = ring->frame_size - NL_MMAP_HDRLEN; | |
685 | data = (void *)hdr + NL_MMAP_HDRLEN; | |
686 | ||
687 | skb->head = data; | |
688 | skb->data = data; | |
689 | skb_reset_tail_pointer(skb); | |
690 | skb->end = skb->tail + size; | |
691 | skb->len = 0; | |
692 | ||
693 | skb->destructor = netlink_skb_destructor; | |
694 | NETLINK_CB(skb).flags |= NETLINK_SKB_MMAPED; | |
695 | NETLINK_CB(skb).sk = sk; | |
696 | } | |
5fd96123 PM |
697 | |
698 | static int netlink_mmap_sendmsg(struct sock *sk, struct msghdr *msg, | |
699 | u32 dst_portid, u32 dst_group, | |
7cc05662 | 700 | struct scm_cookie *scm) |
5fd96123 PM |
701 | { |
702 | struct netlink_sock *nlk = nlk_sk(sk); | |
703 | struct netlink_ring *ring; | |
704 | struct nl_mmap_hdr *hdr; | |
705 | struct sk_buff *skb; | |
706 | unsigned int maxlen; | |
5fd96123 PM |
707 | int err = 0, len = 0; |
708 | ||
5fd96123 PM |
709 | mutex_lock(&nlk->pg_vec_lock); |
710 | ||
711 | ring = &nlk->tx_ring; | |
712 | maxlen = ring->frame_size - NL_MMAP_HDRLEN; | |
713 | ||
714 | do { | |
4682a035 DM |
715 | unsigned int nm_len; |
716 | ||
5fd96123 PM |
717 | hdr = netlink_current_frame(ring, NL_MMAP_STATUS_VALID); |
718 | if (hdr == NULL) { | |
719 | if (!(msg->msg_flags & MSG_DONTWAIT) && | |
720 | atomic_read(&nlk->tx_ring.pending)) | |
721 | schedule(); | |
722 | continue; | |
723 | } | |
4682a035 DM |
724 | |
725 | nm_len = ACCESS_ONCE(hdr->nm_len); | |
726 | if (nm_len > maxlen) { | |
5fd96123 PM |
727 | err = -EINVAL; |
728 | goto out; | |
729 | } | |
730 | ||
4682a035 | 731 | netlink_frame_flush_dcache(hdr, nm_len); |
5fd96123 | 732 | |
4682a035 DM |
733 | skb = alloc_skb(nm_len, GFP_KERNEL); |
734 | if (skb == NULL) { | |
735 | err = -ENOBUFS; | |
736 | goto out; | |
5fd96123 | 737 | } |
4682a035 DM |
738 | __skb_put(skb, nm_len); |
739 | memcpy(skb->data, (void *)hdr + NL_MMAP_HDRLEN, nm_len); | |
740 | netlink_set_status(hdr, NL_MMAP_STATUS_UNUSED); | |
5fd96123 PM |
741 | |
742 | netlink_increment_head(ring); | |
743 | ||
744 | NETLINK_CB(skb).portid = nlk->portid; | |
745 | NETLINK_CB(skb).dst_group = dst_group; | |
7cc05662 | 746 | NETLINK_CB(skb).creds = scm->creds; |
5fd96123 PM |
747 | |
748 | err = security_netlink_send(sk, skb); | |
749 | if (err) { | |
750 | kfree_skb(skb); | |
751 | goto out; | |
752 | } | |
753 | ||
754 | if (unlikely(dst_group)) { | |
755 | atomic_inc(&skb->users); | |
756 | netlink_broadcast(sk, skb, dst_portid, dst_group, | |
757 | GFP_KERNEL); | |
758 | } | |
759 | err = netlink_unicast(sk, skb, dst_portid, | |
760 | msg->msg_flags & MSG_DONTWAIT); | |
761 | if (err < 0) | |
762 | goto out; | |
763 | len += err; | |
764 | ||
765 | } while (hdr != NULL || | |
766 | (!(msg->msg_flags & MSG_DONTWAIT) && | |
767 | atomic_read(&nlk->tx_ring.pending))); | |
768 | ||
769 | if (len > 0) | |
770 | err = len; | |
771 | out: | |
772 | mutex_unlock(&nlk->pg_vec_lock); | |
773 | return err; | |
774 | } | |
f9c22888 PM |
775 | |
776 | static void netlink_queue_mmaped_skb(struct sock *sk, struct sk_buff *skb) | |
777 | { | |
778 | struct nl_mmap_hdr *hdr; | |
779 | ||
780 | hdr = netlink_mmap_hdr(skb); | |
781 | hdr->nm_len = skb->len; | |
782 | hdr->nm_group = NETLINK_CB(skb).dst_group; | |
783 | hdr->nm_pid = NETLINK_CB(skb).creds.pid; | |
1bf9310a ND |
784 | hdr->nm_uid = from_kuid(sk_user_ns(sk), NETLINK_CB(skb).creds.uid); |
785 | hdr->nm_gid = from_kgid(sk_user_ns(sk), NETLINK_CB(skb).creds.gid); | |
4682a035 | 786 | netlink_frame_flush_dcache(hdr, hdr->nm_len); |
f9c22888 PM |
787 | netlink_set_status(hdr, NL_MMAP_STATUS_VALID); |
788 | ||
789 | NETLINK_CB(skb).flags |= NETLINK_SKB_DELIVERED; | |
790 | kfree_skb(skb); | |
791 | } | |
792 | ||
793 | static void netlink_ring_set_copied(struct sock *sk, struct sk_buff *skb) | |
794 | { | |
795 | struct netlink_sock *nlk = nlk_sk(sk); | |
796 | struct netlink_ring *ring = &nlk->rx_ring; | |
797 | struct nl_mmap_hdr *hdr; | |
798 | ||
799 | spin_lock_bh(&sk->sk_receive_queue.lock); | |
800 | hdr = netlink_current_frame(ring, NL_MMAP_STATUS_UNUSED); | |
801 | if (hdr == NULL) { | |
802 | spin_unlock_bh(&sk->sk_receive_queue.lock); | |
803 | kfree_skb(skb); | |
cd1df525 | 804 | netlink_overrun(sk); |
f9c22888 PM |
805 | return; |
806 | } | |
807 | netlink_increment_head(ring); | |
808 | __skb_queue_tail(&sk->sk_receive_queue, skb); | |
809 | spin_unlock_bh(&sk->sk_receive_queue.lock); | |
810 | ||
811 | hdr->nm_len = skb->len; | |
812 | hdr->nm_group = NETLINK_CB(skb).dst_group; | |
813 | hdr->nm_pid = NETLINK_CB(skb).creds.pid; | |
1bf9310a ND |
814 | hdr->nm_uid = from_kuid(sk_user_ns(sk), NETLINK_CB(skb).creds.uid); |
815 | hdr->nm_gid = from_kgid(sk_user_ns(sk), NETLINK_CB(skb).creds.gid); | |
f9c22888 PM |
816 | netlink_set_status(hdr, NL_MMAP_STATUS_COPY); |
817 | } | |
818 | ||
ccdfcc39 | 819 | #else /* CONFIG_NETLINK_MMAP */ |
9652e931 | 820 | #define netlink_skb_is_mmaped(skb) false |
f9c22888 | 821 | #define netlink_rx_is_mmaped(sk) false |
5fd96123 | 822 | #define netlink_tx_is_mmaped(sk) false |
ccdfcc39 | 823 | #define netlink_mmap sock_no_mmap |
9652e931 | 824 | #define netlink_poll datagram_poll |
7cc05662 | 825 | #define netlink_mmap_sendmsg(sk, msg, dst_portid, dst_group, scm) 0 |
ccdfcc39 PM |
826 | #endif /* CONFIG_NETLINK_MMAP */ |
827 | ||
cf0a018a PM |
828 | static void netlink_skb_destructor(struct sk_buff *skb) |
829 | { | |
9652e931 PM |
830 | #ifdef CONFIG_NETLINK_MMAP |
831 | struct nl_mmap_hdr *hdr; | |
832 | struct netlink_ring *ring; | |
833 | struct sock *sk; | |
834 | ||
835 | /* If a packet from the kernel to userspace was freed because of an | |
836 | * error without being delivered to userspace, the kernel must reset | |
837 | * the status. In the direction userspace to kernel, the status is | |
838 | * always reset here after the packet was processed and freed. | |
839 | */ | |
840 | if (netlink_skb_is_mmaped(skb)) { | |
841 | hdr = netlink_mmap_hdr(skb); | |
842 | sk = NETLINK_CB(skb).sk; | |
843 | ||
5fd96123 PM |
844 | if (NETLINK_CB(skb).flags & NETLINK_SKB_TX) { |
845 | netlink_set_status(hdr, NL_MMAP_STATUS_UNUSED); | |
846 | ring = &nlk_sk(sk)->tx_ring; | |
847 | } else { | |
848 | if (!(NETLINK_CB(skb).flags & NETLINK_SKB_DELIVERED)) { | |
849 | hdr->nm_len = 0; | |
850 | netlink_set_status(hdr, NL_MMAP_STATUS_VALID); | |
851 | } | |
852 | ring = &nlk_sk(sk)->rx_ring; | |
9652e931 | 853 | } |
9652e931 PM |
854 | |
855 | WARN_ON(atomic_read(&ring->pending) == 0); | |
856 | atomic_dec(&ring->pending); | |
857 | sock_put(sk); | |
858 | ||
5e71d9d7 | 859 | skb->head = NULL; |
9652e931 PM |
860 | } |
861 | #endif | |
c05cdb1b | 862 | if (is_vmalloc_addr(skb->head)) { |
3a36515f PN |
863 | if (!skb->cloned || |
864 | !atomic_dec_return(&(skb_shinfo(skb)->dataref))) | |
865 | vfree(skb->head); | |
866 | ||
c05cdb1b PNA |
867 | skb->head = NULL; |
868 | } | |
9652e931 PM |
869 | if (skb->sk != NULL) |
870 | sock_rfree(skb); | |
cf0a018a PM |
871 | } |
872 | ||
873 | static void netlink_skb_set_owner_r(struct sk_buff *skb, struct sock *sk) | |
874 | { | |
875 | WARN_ON(skb->sk != NULL); | |
876 | skb->sk = sk; | |
877 | skb->destructor = netlink_skb_destructor; | |
878 | atomic_add(skb->truesize, &sk->sk_rmem_alloc); | |
879 | sk_mem_charge(sk, skb->truesize); | |
880 | } | |
881 | ||
1da177e4 LT |
882 | static void netlink_sock_destruct(struct sock *sk) |
883 | { | |
3f660d66 HX |
884 | struct netlink_sock *nlk = nlk_sk(sk); |
885 | ||
16b304f3 PS |
886 | if (nlk->cb_running) { |
887 | if (nlk->cb.done) | |
888 | nlk->cb.done(&nlk->cb); | |
6dc878a8 | 889 | |
16b304f3 PS |
890 | module_put(nlk->cb.module); |
891 | kfree_skb(nlk->cb.skb); | |
3f660d66 HX |
892 | } |
893 | ||
1da177e4 | 894 | skb_queue_purge(&sk->sk_receive_queue); |
ccdfcc39 PM |
895 | #ifdef CONFIG_NETLINK_MMAP |
896 | if (1) { | |
897 | struct nl_mmap_req req; | |
898 | ||
899 | memset(&req, 0, sizeof(req)); | |
900 | if (nlk->rx_ring.pg_vec) | |
901 | netlink_set_ring(sk, &req, true, false); | |
902 | memset(&req, 0, sizeof(req)); | |
903 | if (nlk->tx_ring.pg_vec) | |
904 | netlink_set_ring(sk, &req, true, true); | |
905 | } | |
906 | #endif /* CONFIG_NETLINK_MMAP */ | |
1da177e4 LT |
907 | |
908 | if (!sock_flag(sk, SOCK_DEAD)) { | |
6ac552fd | 909 | printk(KERN_ERR "Freeing alive netlink socket %p\n", sk); |
1da177e4 LT |
910 | return; |
911 | } | |
547b792c IJ |
912 | |
913 | WARN_ON(atomic_read(&sk->sk_rmem_alloc)); | |
914 | WARN_ON(atomic_read(&sk->sk_wmem_alloc)); | |
915 | WARN_ON(nlk_sk(sk)->groups); | |
1da177e4 LT |
916 | } |
917 | ||
6ac552fd PM |
918 | /* This lock without WQ_FLAG_EXCLUSIVE is good on UP and it is _very_ bad on |
919 | * SMP. Look, when several writers sleep and reader wakes them up, all but one | |
1da177e4 LT |
920 | * immediately hit write lock and grab all the cpus. Exclusive sleep solves |
921 | * this, _but_ remember, it adds useless work on UP machines. | |
922 | */ | |
923 | ||
d136f1bd | 924 | void netlink_table_grab(void) |
9a429c49 | 925 | __acquires(nl_table_lock) |
1da177e4 | 926 | { |
d136f1bd JB |
927 | might_sleep(); |
928 | ||
6abd219c | 929 | write_lock_irq(&nl_table_lock); |
1da177e4 LT |
930 | |
931 | if (atomic_read(&nl_table_users)) { | |
932 | DECLARE_WAITQUEUE(wait, current); | |
933 | ||
934 | add_wait_queue_exclusive(&nl_table_wait, &wait); | |
6ac552fd | 935 | for (;;) { |
1da177e4 LT |
936 | set_current_state(TASK_UNINTERRUPTIBLE); |
937 | if (atomic_read(&nl_table_users) == 0) | |
938 | break; | |
6abd219c | 939 | write_unlock_irq(&nl_table_lock); |
1da177e4 | 940 | schedule(); |
6abd219c | 941 | write_lock_irq(&nl_table_lock); |
1da177e4 LT |
942 | } |
943 | ||
944 | __set_current_state(TASK_RUNNING); | |
945 | remove_wait_queue(&nl_table_wait, &wait); | |
946 | } | |
947 | } | |
948 | ||
d136f1bd | 949 | void netlink_table_ungrab(void) |
9a429c49 | 950 | __releases(nl_table_lock) |
1da177e4 | 951 | { |
6abd219c | 952 | write_unlock_irq(&nl_table_lock); |
1da177e4 LT |
953 | wake_up(&nl_table_wait); |
954 | } | |
955 | ||
6ac552fd | 956 | static inline void |
1da177e4 LT |
957 | netlink_lock_table(void) |
958 | { | |
959 | /* read_lock() synchronizes us to netlink_table_grab */ | |
960 | ||
961 | read_lock(&nl_table_lock); | |
962 | atomic_inc(&nl_table_users); | |
963 | read_unlock(&nl_table_lock); | |
964 | } | |
965 | ||
6ac552fd | 966 | static inline void |
1da177e4 LT |
967 | netlink_unlock_table(void) |
968 | { | |
969 | if (atomic_dec_and_test(&nl_table_users)) | |
970 | wake_up(&nl_table_wait); | |
971 | } | |
972 | ||
e341694e | 973 | struct netlink_compare_arg |
1da177e4 | 974 | { |
c428ecd1 | 975 | possible_net_t pnet; |
e341694e | 976 | u32 portid; |
c428ecd1 | 977 | char trailer[]; |
e341694e | 978 | }; |
1da177e4 | 979 | |
c428ecd1 HX |
980 | #define netlink_compare_arg_len offsetof(struct netlink_compare_arg, trailer) |
981 | ||
982 | static inline int netlink_compare(struct rhashtable_compare_arg *arg, | |
983 | const void *ptr) | |
1da177e4 | 984 | { |
c428ecd1 HX |
985 | const struct netlink_compare_arg *x = arg->key; |
986 | const struct netlink_sock *nlk = ptr; | |
1da177e4 | 987 | |
c428ecd1 HX |
988 | return nlk->portid != x->portid || |
989 | !net_eq(sock_net(&nlk->sk), read_pnet(&x->pnet)); | |
990 | } | |
991 | ||
992 | static void netlink_compare_arg_init(struct netlink_compare_arg *arg, | |
993 | struct net *net, u32 portid) | |
994 | { | |
995 | memset(arg, 0, sizeof(*arg)); | |
996 | write_pnet(&arg->pnet, net); | |
997 | arg->portid = portid; | |
1da177e4 LT |
998 | } |
999 | ||
e341694e TG |
1000 | static struct sock *__netlink_lookup(struct netlink_table *table, u32 portid, |
1001 | struct net *net) | |
1da177e4 | 1002 | { |
c428ecd1 | 1003 | struct netlink_compare_arg arg; |
1da177e4 | 1004 | |
c428ecd1 HX |
1005 | netlink_compare_arg_init(&arg, net, portid); |
1006 | return rhashtable_lookup_fast(&table->hash, &arg, | |
1007 | netlink_rhashtable_params); | |
1da177e4 LT |
1008 | } |
1009 | ||
c428ecd1 | 1010 | static int __netlink_insert(struct netlink_table *table, struct sock *sk) |
c5adde94 | 1011 | { |
c428ecd1 | 1012 | struct netlink_compare_arg arg; |
c5adde94 | 1013 | |
c428ecd1 HX |
1014 | netlink_compare_arg_init(&arg, sock_net(sk), nlk_sk(sk)->portid); |
1015 | return rhashtable_lookup_insert_key(&table->hash, &arg, | |
1016 | &nlk_sk(sk)->node, | |
1017 | netlink_rhashtable_params); | |
c5adde94 YX |
1018 | } |
1019 | ||
e341694e | 1020 | static struct sock *netlink_lookup(struct net *net, int protocol, u32 portid) |
1da177e4 | 1021 | { |
e341694e TG |
1022 | struct netlink_table *table = &nl_table[protocol]; |
1023 | struct sock *sk; | |
1da177e4 | 1024 | |
e341694e TG |
1025 | rcu_read_lock(); |
1026 | sk = __netlink_lookup(table, portid, net); | |
1027 | if (sk) | |
1028 | sock_hold(sk); | |
1029 | rcu_read_unlock(); | |
1da177e4 | 1030 | |
e341694e | 1031 | return sk; |
1da177e4 LT |
1032 | } |
1033 | ||
90ddc4f0 | 1034 | static const struct proto_ops netlink_ops; |
1da177e4 | 1035 | |
4277a083 PM |
1036 | static void |
1037 | netlink_update_listeners(struct sock *sk) | |
1038 | { | |
1039 | struct netlink_table *tbl = &nl_table[sk->sk_protocol]; | |
4277a083 PM |
1040 | unsigned long mask; |
1041 | unsigned int i; | |
6d772ac5 ED |
1042 | struct listeners *listeners; |
1043 | ||
1044 | listeners = nl_deref_protected(tbl->listeners); | |
1045 | if (!listeners) | |
1046 | return; | |
4277a083 | 1047 | |
b4ff4f04 | 1048 | for (i = 0; i < NLGRPLONGS(tbl->groups); i++) { |
4277a083 | 1049 | mask = 0; |
b67bfe0d | 1050 | sk_for_each_bound(sk, &tbl->mc_list) { |
b4ff4f04 JB |
1051 | if (i < NLGRPLONGS(nlk_sk(sk)->ngroups)) |
1052 | mask |= nlk_sk(sk)->groups[i]; | |
1053 | } | |
6d772ac5 | 1054 | listeners->masks[i] = mask; |
4277a083 PM |
1055 | } |
1056 | /* this function is only called with the netlink table "grabbed", which | |
1057 | * makes sure updates are visible before bind or setsockopt return. */ | |
1058 | } | |
1059 | ||
8ea65f4a | 1060 | static int netlink_insert(struct sock *sk, u32 portid) |
1da177e4 | 1061 | { |
da12c90e | 1062 | struct netlink_table *table = &nl_table[sk->sk_protocol]; |
919d9db9 | 1063 | int err; |
1da177e4 | 1064 | |
c5adde94 | 1065 | lock_sock(sk); |
1da177e4 LT |
1066 | |
1067 | err = -EBUSY; | |
15e47304 | 1068 | if (nlk_sk(sk)->portid) |
1da177e4 LT |
1069 | goto err; |
1070 | ||
1071 | err = -ENOMEM; | |
97defe1e TG |
1072 | if (BITS_PER_LONG > 32 && |
1073 | unlikely(atomic_read(&table->hash.nelems) >= UINT_MAX)) | |
1da177e4 LT |
1074 | goto err; |
1075 | ||
15e47304 | 1076 | nlk_sk(sk)->portid = portid; |
e341694e | 1077 | sock_hold(sk); |
919d9db9 | 1078 | |
c428ecd1 HX |
1079 | err = __netlink_insert(table, sk); |
1080 | if (err) { | |
1081 | if (err == -EEXIST) | |
1082 | err = -EADDRINUSE; | |
c5adde94 | 1083 | sock_put(sk); |
919d9db9 HX |
1084 | } |
1085 | ||
1da177e4 | 1086 | err: |
c5adde94 | 1087 | release_sock(sk); |
1da177e4 LT |
1088 | return err; |
1089 | } | |
1090 | ||
1091 | static void netlink_remove(struct sock *sk) | |
1092 | { | |
e341694e TG |
1093 | struct netlink_table *table; |
1094 | ||
e341694e | 1095 | table = &nl_table[sk->sk_protocol]; |
c428ecd1 HX |
1096 | if (!rhashtable_remove_fast(&table->hash, &nlk_sk(sk)->node, |
1097 | netlink_rhashtable_params)) { | |
e341694e TG |
1098 | WARN_ON(atomic_read(&sk->sk_refcnt) == 1); |
1099 | __sock_put(sk); | |
1100 | } | |
e341694e | 1101 | |
1da177e4 | 1102 | netlink_table_grab(); |
b10dcb3b | 1103 | if (nlk_sk(sk)->subscriptions) { |
1da177e4 | 1104 | __sk_del_bind_node(sk); |
b10dcb3b JB |
1105 | netlink_update_listeners(sk); |
1106 | } | |
ee1c2442 JB |
1107 | if (sk->sk_protocol == NETLINK_GENERIC) |
1108 | atomic_inc(&genl_sk_destructing_cnt); | |
1da177e4 LT |
1109 | netlink_table_ungrab(); |
1110 | } | |
1111 | ||
1112 | static struct proto netlink_proto = { | |
1113 | .name = "NETLINK", | |
1114 | .owner = THIS_MODULE, | |
1115 | .obj_size = sizeof(struct netlink_sock), | |
1116 | }; | |
1117 | ||
1b8d7ae4 EB |
1118 | static int __netlink_create(struct net *net, struct socket *sock, |
1119 | struct mutex *cb_mutex, int protocol) | |
1da177e4 LT |
1120 | { |
1121 | struct sock *sk; | |
1122 | struct netlink_sock *nlk; | |
ab33a171 PM |
1123 | |
1124 | sock->ops = &netlink_ops; | |
1125 | ||
6257ff21 | 1126 | sk = sk_alloc(net, PF_NETLINK, GFP_KERNEL, &netlink_proto); |
ab33a171 PM |
1127 | if (!sk) |
1128 | return -ENOMEM; | |
1129 | ||
1130 | sock_init_data(sock, sk); | |
1131 | ||
1132 | nlk = nlk_sk(sk); | |
658cb354 | 1133 | if (cb_mutex) { |
ffa4d721 | 1134 | nlk->cb_mutex = cb_mutex; |
658cb354 | 1135 | } else { |
ffa4d721 PM |
1136 | nlk->cb_mutex = &nlk->cb_def_mutex; |
1137 | mutex_init(nlk->cb_mutex); | |
1138 | } | |
ab33a171 | 1139 | init_waitqueue_head(&nlk->wait); |
ccdfcc39 PM |
1140 | #ifdef CONFIG_NETLINK_MMAP |
1141 | mutex_init(&nlk->pg_vec_lock); | |
1142 | #endif | |
ab33a171 PM |
1143 | |
1144 | sk->sk_destruct = netlink_sock_destruct; | |
1145 | sk->sk_protocol = protocol; | |
1146 | return 0; | |
1147 | } | |
1148 | ||
3f378b68 EP |
1149 | static int netlink_create(struct net *net, struct socket *sock, int protocol, |
1150 | int kern) | |
ab33a171 PM |
1151 | { |
1152 | struct module *module = NULL; | |
af65bdfc | 1153 | struct mutex *cb_mutex; |
f7fa9b10 | 1154 | struct netlink_sock *nlk; |
023e2cfa JB |
1155 | int (*bind)(struct net *net, int group); |
1156 | void (*unbind)(struct net *net, int group); | |
ab33a171 | 1157 | int err = 0; |
1da177e4 LT |
1158 | |
1159 | sock->state = SS_UNCONNECTED; | |
1160 | ||
1161 | if (sock->type != SOCK_RAW && sock->type != SOCK_DGRAM) | |
1162 | return -ESOCKTNOSUPPORT; | |
1163 | ||
6ac552fd | 1164 | if (protocol < 0 || protocol >= MAX_LINKS) |
1da177e4 LT |
1165 | return -EPROTONOSUPPORT; |
1166 | ||
77247bbb | 1167 | netlink_lock_table(); |
95a5afca | 1168 | #ifdef CONFIG_MODULES |
ab33a171 | 1169 | if (!nl_table[protocol].registered) { |
77247bbb | 1170 | netlink_unlock_table(); |
4fdb3bb7 | 1171 | request_module("net-pf-%d-proto-%d", PF_NETLINK, protocol); |
77247bbb | 1172 | netlink_lock_table(); |
4fdb3bb7 | 1173 | } |
ab33a171 PM |
1174 | #endif |
1175 | if (nl_table[protocol].registered && | |
1176 | try_module_get(nl_table[protocol].module)) | |
1177 | module = nl_table[protocol].module; | |
974c37e9 AD |
1178 | else |
1179 | err = -EPROTONOSUPPORT; | |
af65bdfc | 1180 | cb_mutex = nl_table[protocol].cb_mutex; |
03292745 | 1181 | bind = nl_table[protocol].bind; |
4f520900 | 1182 | unbind = nl_table[protocol].unbind; |
77247bbb | 1183 | netlink_unlock_table(); |
4fdb3bb7 | 1184 | |
974c37e9 AD |
1185 | if (err < 0) |
1186 | goto out; | |
1187 | ||
6ac552fd PM |
1188 | err = __netlink_create(net, sock, cb_mutex, protocol); |
1189 | if (err < 0) | |
f7fa9b10 PM |
1190 | goto out_module; |
1191 | ||
6f756a8c | 1192 | local_bh_disable(); |
c1fd3b94 | 1193 | sock_prot_inuse_add(net, &netlink_proto, 1); |
6f756a8c DM |
1194 | local_bh_enable(); |
1195 | ||
f7fa9b10 | 1196 | nlk = nlk_sk(sock->sk); |
f7fa9b10 | 1197 | nlk->module = module; |
03292745 | 1198 | nlk->netlink_bind = bind; |
4f520900 | 1199 | nlk->netlink_unbind = unbind; |
ab33a171 PM |
1200 | out: |
1201 | return err; | |
1da177e4 | 1202 | |
ab33a171 PM |
1203 | out_module: |
1204 | module_put(module); | |
1205 | goto out; | |
1da177e4 LT |
1206 | } |
1207 | ||
21e4902a TG |
1208 | static void deferred_put_nlk_sk(struct rcu_head *head) |
1209 | { | |
1210 | struct netlink_sock *nlk = container_of(head, struct netlink_sock, rcu); | |
1211 | ||
1212 | sock_put(&nlk->sk); | |
1213 | } | |
1214 | ||
1da177e4 LT |
1215 | static int netlink_release(struct socket *sock) |
1216 | { | |
1217 | struct sock *sk = sock->sk; | |
1218 | struct netlink_sock *nlk; | |
1219 | ||
1220 | if (!sk) | |
1221 | return 0; | |
1222 | ||
1223 | netlink_remove(sk); | |
ac57b3a9 | 1224 | sock_orphan(sk); |
1da177e4 LT |
1225 | nlk = nlk_sk(sk); |
1226 | ||
3f660d66 HX |
1227 | /* |
1228 | * OK. Socket is unlinked, any packets that arrive now | |
1229 | * will be purged. | |
1230 | */ | |
1da177e4 | 1231 | |
ee1c2442 JB |
1232 | /* must not acquire netlink_table_lock in any way again before unbind |
1233 | * and notifying genetlink is done as otherwise it might deadlock | |
1234 | */ | |
1235 | if (nlk->netlink_unbind) { | |
1236 | int i; | |
1237 | ||
1238 | for (i = 0; i < nlk->ngroups; i++) | |
1239 | if (test_bit(i, nlk->groups)) | |
1240 | nlk->netlink_unbind(sock_net(sk), i + 1); | |
1241 | } | |
1242 | if (sk->sk_protocol == NETLINK_GENERIC && | |
1243 | atomic_dec_return(&genl_sk_destructing_cnt) == 0) | |
1244 | wake_up(&genl_sk_destructing_waitq); | |
1245 | ||
1da177e4 LT |
1246 | sock->sk = NULL; |
1247 | wake_up_interruptible_all(&nlk->wait); | |
1248 | ||
1249 | skb_queue_purge(&sk->sk_write_queue); | |
1250 | ||
15e47304 | 1251 | if (nlk->portid) { |
1da177e4 | 1252 | struct netlink_notify n = { |
3b1e0a65 | 1253 | .net = sock_net(sk), |
1da177e4 | 1254 | .protocol = sk->sk_protocol, |
15e47304 | 1255 | .portid = nlk->portid, |
1da177e4 | 1256 | }; |
e041c683 AS |
1257 | atomic_notifier_call_chain(&netlink_chain, |
1258 | NETLINK_URELEASE, &n); | |
746fac4d | 1259 | } |
4fdb3bb7 | 1260 | |
5e7c001c | 1261 | module_put(nlk->module); |
4fdb3bb7 | 1262 | |
aed81560 | 1263 | if (netlink_is_kernel(sk)) { |
b10dcb3b | 1264 | netlink_table_grab(); |
869e58f8 DL |
1265 | BUG_ON(nl_table[sk->sk_protocol].registered == 0); |
1266 | if (--nl_table[sk->sk_protocol].registered == 0) { | |
6d772ac5 ED |
1267 | struct listeners *old; |
1268 | ||
1269 | old = nl_deref_protected(nl_table[sk->sk_protocol].listeners); | |
1270 | RCU_INIT_POINTER(nl_table[sk->sk_protocol].listeners, NULL); | |
1271 | kfree_rcu(old, rcu); | |
869e58f8 | 1272 | nl_table[sk->sk_protocol].module = NULL; |
9785e10a | 1273 | nl_table[sk->sk_protocol].bind = NULL; |
4f520900 | 1274 | nl_table[sk->sk_protocol].unbind = NULL; |
9785e10a | 1275 | nl_table[sk->sk_protocol].flags = 0; |
869e58f8 DL |
1276 | nl_table[sk->sk_protocol].registered = 0; |
1277 | } | |
b10dcb3b | 1278 | netlink_table_ungrab(); |
658cb354 | 1279 | } |
77247bbb | 1280 | |
f7fa9b10 PM |
1281 | kfree(nlk->groups); |
1282 | nlk->groups = NULL; | |
1283 | ||
3755810c | 1284 | local_bh_disable(); |
c1fd3b94 | 1285 | sock_prot_inuse_add(sock_net(sk), &netlink_proto, -1); |
3755810c | 1286 | local_bh_enable(); |
21e4902a | 1287 | call_rcu(&nlk->rcu, deferred_put_nlk_sk); |
1da177e4 LT |
1288 | return 0; |
1289 | } | |
1290 | ||
1291 | static int netlink_autobind(struct socket *sock) | |
1292 | { | |
1293 | struct sock *sk = sock->sk; | |
3b1e0a65 | 1294 | struct net *net = sock_net(sk); |
da12c90e | 1295 | struct netlink_table *table = &nl_table[sk->sk_protocol]; |
15e47304 | 1296 | s32 portid = task_tgid_vnr(current); |
1da177e4 LT |
1297 | int err; |
1298 | static s32 rover = -4097; | |
1299 | ||
1300 | retry: | |
1301 | cond_resched(); | |
e341694e TG |
1302 | rcu_read_lock(); |
1303 | if (__netlink_lookup(table, portid, net)) { | |
1304 | /* Bind collision, search negative portid values. */ | |
1305 | portid = rover--; | |
1306 | if (rover > -4097) | |
1307 | rover = -4097; | |
1308 | rcu_read_unlock(); | |
1309 | goto retry; | |
1da177e4 | 1310 | } |
e341694e | 1311 | rcu_read_unlock(); |
1da177e4 | 1312 | |
8ea65f4a | 1313 | err = netlink_insert(sk, portid); |
1da177e4 LT |
1314 | if (err == -EADDRINUSE) |
1315 | goto retry; | |
d470e3b4 DM |
1316 | |
1317 | /* If 2 threads race to autobind, that is fine. */ | |
1318 | if (err == -EBUSY) | |
1319 | err = 0; | |
1320 | ||
1321 | return err; | |
1da177e4 LT |
1322 | } |
1323 | ||
aa4cf945 EB |
1324 | /** |
1325 | * __netlink_ns_capable - General netlink message capability test | |
1326 | * @nsp: NETLINK_CB of the socket buffer holding a netlink command from userspace. | |
1327 | * @user_ns: The user namespace of the capability to use | |
1328 | * @cap: The capability to use | |
1329 | * | |
1330 | * Test to see if the opener of the socket we received the message | |
1331 | * from had when the netlink socket was created and the sender of the | |
1332 | * message has has the capability @cap in the user namespace @user_ns. | |
1333 | */ | |
1334 | bool __netlink_ns_capable(const struct netlink_skb_parms *nsp, | |
1335 | struct user_namespace *user_ns, int cap) | |
1336 | { | |
2d7a85f4 EB |
1337 | return ((nsp->flags & NETLINK_SKB_DST) || |
1338 | file_ns_capable(nsp->sk->sk_socket->file, user_ns, cap)) && | |
1339 | ns_capable(user_ns, cap); | |
aa4cf945 EB |
1340 | } |
1341 | EXPORT_SYMBOL(__netlink_ns_capable); | |
1342 | ||
1343 | /** | |
1344 | * netlink_ns_capable - General netlink message capability test | |
1345 | * @skb: socket buffer holding a netlink command from userspace | |
1346 | * @user_ns: The user namespace of the capability to use | |
1347 | * @cap: The capability to use | |
1348 | * | |
1349 | * Test to see if the opener of the socket we received the message | |
1350 | * from had when the netlink socket was created and the sender of the | |
1351 | * message has has the capability @cap in the user namespace @user_ns. | |
1352 | */ | |
1353 | bool netlink_ns_capable(const struct sk_buff *skb, | |
1354 | struct user_namespace *user_ns, int cap) | |
1355 | { | |
1356 | return __netlink_ns_capable(&NETLINK_CB(skb), user_ns, cap); | |
1357 | } | |
1358 | EXPORT_SYMBOL(netlink_ns_capable); | |
1359 | ||
1360 | /** | |
1361 | * netlink_capable - Netlink global message capability test | |
1362 | * @skb: socket buffer holding a netlink command from userspace | |
1363 | * @cap: The capability to use | |
1364 | * | |
1365 | * Test to see if the opener of the socket we received the message | |
1366 | * from had when the netlink socket was created and the sender of the | |
1367 | * message has has the capability @cap in all user namespaces. | |
1368 | */ | |
1369 | bool netlink_capable(const struct sk_buff *skb, int cap) | |
1370 | { | |
1371 | return netlink_ns_capable(skb, &init_user_ns, cap); | |
1372 | } | |
1373 | EXPORT_SYMBOL(netlink_capable); | |
1374 | ||
1375 | /** | |
1376 | * netlink_net_capable - Netlink network namespace message capability test | |
1377 | * @skb: socket buffer holding a netlink command from userspace | |
1378 | * @cap: The capability to use | |
1379 | * | |
1380 | * Test to see if the opener of the socket we received the message | |
1381 | * from had when the netlink socket was created and the sender of the | |
1382 | * message has has the capability @cap over the network namespace of | |
1383 | * the socket we received the message from. | |
1384 | */ | |
1385 | bool netlink_net_capable(const struct sk_buff *skb, int cap) | |
1386 | { | |
1387 | return netlink_ns_capable(skb, sock_net(skb->sk)->user_ns, cap); | |
1388 | } | |
1389 | EXPORT_SYMBOL(netlink_net_capable); | |
1390 | ||
5187cd05 | 1391 | static inline int netlink_allowed(const struct socket *sock, unsigned int flag) |
746fac4d | 1392 | { |
9785e10a | 1393 | return (nl_table[sock->sk->sk_protocol].flags & flag) || |
df008c91 | 1394 | ns_capable(sock_net(sock->sk)->user_ns, CAP_NET_ADMIN); |
746fac4d | 1395 | } |
1da177e4 | 1396 | |
f7fa9b10 PM |
1397 | static void |
1398 | netlink_update_subscriptions(struct sock *sk, unsigned int subscriptions) | |
1399 | { | |
1400 | struct netlink_sock *nlk = nlk_sk(sk); | |
1401 | ||
1402 | if (nlk->subscriptions && !subscriptions) | |
1403 | __sk_del_bind_node(sk); | |
1404 | else if (!nlk->subscriptions && subscriptions) | |
1405 | sk_add_bind_node(sk, &nl_table[sk->sk_protocol].mc_list); | |
1406 | nlk->subscriptions = subscriptions; | |
1407 | } | |
1408 | ||
b4ff4f04 | 1409 | static int netlink_realloc_groups(struct sock *sk) |
513c2500 PM |
1410 | { |
1411 | struct netlink_sock *nlk = nlk_sk(sk); | |
1412 | unsigned int groups; | |
b4ff4f04 | 1413 | unsigned long *new_groups; |
513c2500 PM |
1414 | int err = 0; |
1415 | ||
b4ff4f04 JB |
1416 | netlink_table_grab(); |
1417 | ||
513c2500 | 1418 | groups = nl_table[sk->sk_protocol].groups; |
b4ff4f04 | 1419 | if (!nl_table[sk->sk_protocol].registered) { |
513c2500 | 1420 | err = -ENOENT; |
b4ff4f04 JB |
1421 | goto out_unlock; |
1422 | } | |
513c2500 | 1423 | |
b4ff4f04 JB |
1424 | if (nlk->ngroups >= groups) |
1425 | goto out_unlock; | |
513c2500 | 1426 | |
b4ff4f04 JB |
1427 | new_groups = krealloc(nlk->groups, NLGRPSZ(groups), GFP_ATOMIC); |
1428 | if (new_groups == NULL) { | |
1429 | err = -ENOMEM; | |
1430 | goto out_unlock; | |
1431 | } | |
6ac552fd | 1432 | memset((char *)new_groups + NLGRPSZ(nlk->ngroups), 0, |
b4ff4f04 JB |
1433 | NLGRPSZ(groups) - NLGRPSZ(nlk->ngroups)); |
1434 | ||
1435 | nlk->groups = new_groups; | |
513c2500 | 1436 | nlk->ngroups = groups; |
b4ff4f04 JB |
1437 | out_unlock: |
1438 | netlink_table_ungrab(); | |
1439 | return err; | |
513c2500 PM |
1440 | } |
1441 | ||
02c81ab9 | 1442 | static void netlink_undo_bind(int group, long unsigned int groups, |
023e2cfa | 1443 | struct sock *sk) |
4f520900 | 1444 | { |
023e2cfa | 1445 | struct netlink_sock *nlk = nlk_sk(sk); |
4f520900 RGB |
1446 | int undo; |
1447 | ||
1448 | if (!nlk->netlink_unbind) | |
1449 | return; | |
1450 | ||
1451 | for (undo = 0; undo < group; undo++) | |
6251edd9 | 1452 | if (test_bit(undo, &groups)) |
8b7c36d8 | 1453 | nlk->netlink_unbind(sock_net(sk), undo + 1); |
4f520900 RGB |
1454 | } |
1455 | ||
6ac552fd PM |
1456 | static int netlink_bind(struct socket *sock, struct sockaddr *addr, |
1457 | int addr_len) | |
1da177e4 LT |
1458 | { |
1459 | struct sock *sk = sock->sk; | |
3b1e0a65 | 1460 | struct net *net = sock_net(sk); |
1da177e4 LT |
1461 | struct netlink_sock *nlk = nlk_sk(sk); |
1462 | struct sockaddr_nl *nladdr = (struct sockaddr_nl *)addr; | |
1463 | int err; | |
4f520900 | 1464 | long unsigned int groups = nladdr->nl_groups; |
746fac4d | 1465 | |
4e4b5376 HFS |
1466 | if (addr_len < sizeof(struct sockaddr_nl)) |
1467 | return -EINVAL; | |
1468 | ||
1da177e4 LT |
1469 | if (nladdr->nl_family != AF_NETLINK) |
1470 | return -EINVAL; | |
1471 | ||
1472 | /* Only superuser is allowed to listen multicasts */ | |
4f520900 | 1473 | if (groups) { |
5187cd05 | 1474 | if (!netlink_allowed(sock, NL_CFG_F_NONROOT_RECV)) |
513c2500 | 1475 | return -EPERM; |
b4ff4f04 JB |
1476 | err = netlink_realloc_groups(sk); |
1477 | if (err) | |
1478 | return err; | |
513c2500 | 1479 | } |
1da177e4 | 1480 | |
4f520900 | 1481 | if (nlk->portid) |
15e47304 | 1482 | if (nladdr->nl_pid != nlk->portid) |
1da177e4 | 1483 | return -EINVAL; |
4f520900 RGB |
1484 | |
1485 | if (nlk->netlink_bind && groups) { | |
1486 | int group; | |
1487 | ||
1488 | for (group = 0; group < nlk->ngroups; group++) { | |
1489 | if (!test_bit(group, &groups)) | |
1490 | continue; | |
8b7c36d8 | 1491 | err = nlk->netlink_bind(net, group + 1); |
4f520900 RGB |
1492 | if (!err) |
1493 | continue; | |
023e2cfa | 1494 | netlink_undo_bind(group, groups, sk); |
4f520900 RGB |
1495 | return err; |
1496 | } | |
1497 | } | |
1498 | ||
1499 | if (!nlk->portid) { | |
1da177e4 | 1500 | err = nladdr->nl_pid ? |
8ea65f4a | 1501 | netlink_insert(sk, nladdr->nl_pid) : |
1da177e4 | 1502 | netlink_autobind(sock); |
4f520900 | 1503 | if (err) { |
023e2cfa | 1504 | netlink_undo_bind(nlk->ngroups, groups, sk); |
1da177e4 | 1505 | return err; |
4f520900 | 1506 | } |
1da177e4 LT |
1507 | } |
1508 | ||
4f520900 | 1509 | if (!groups && (nlk->groups == NULL || !(u32)nlk->groups[0])) |
1da177e4 LT |
1510 | return 0; |
1511 | ||
1512 | netlink_table_grab(); | |
f7fa9b10 | 1513 | netlink_update_subscriptions(sk, nlk->subscriptions + |
4f520900 | 1514 | hweight32(groups) - |
746fac4d | 1515 | hweight32(nlk->groups[0])); |
4f520900 | 1516 | nlk->groups[0] = (nlk->groups[0] & ~0xffffffffUL) | groups; |
4277a083 | 1517 | netlink_update_listeners(sk); |
1da177e4 LT |
1518 | netlink_table_ungrab(); |
1519 | ||
1520 | return 0; | |
1521 | } | |
1522 | ||
1523 | static int netlink_connect(struct socket *sock, struct sockaddr *addr, | |
1524 | int alen, int flags) | |
1525 | { | |
1526 | int err = 0; | |
1527 | struct sock *sk = sock->sk; | |
1528 | struct netlink_sock *nlk = nlk_sk(sk); | |
6ac552fd | 1529 | struct sockaddr_nl *nladdr = (struct sockaddr_nl *)addr; |
1da177e4 | 1530 | |
6503d961 CG |
1531 | if (alen < sizeof(addr->sa_family)) |
1532 | return -EINVAL; | |
1533 | ||
1da177e4 LT |
1534 | if (addr->sa_family == AF_UNSPEC) { |
1535 | sk->sk_state = NETLINK_UNCONNECTED; | |
15e47304 | 1536 | nlk->dst_portid = 0; |
d629b836 | 1537 | nlk->dst_group = 0; |
1da177e4 LT |
1538 | return 0; |
1539 | } | |
1540 | if (addr->sa_family != AF_NETLINK) | |
1541 | return -EINVAL; | |
1542 | ||
46833a86 | 1543 | if ((nladdr->nl_groups || nladdr->nl_pid) && |
5187cd05 | 1544 | !netlink_allowed(sock, NL_CFG_F_NONROOT_SEND)) |
1da177e4 LT |
1545 | return -EPERM; |
1546 | ||
15e47304 | 1547 | if (!nlk->portid) |
1da177e4 LT |
1548 | err = netlink_autobind(sock); |
1549 | ||
1550 | if (err == 0) { | |
1551 | sk->sk_state = NETLINK_CONNECTED; | |
15e47304 | 1552 | nlk->dst_portid = nladdr->nl_pid; |
d629b836 | 1553 | nlk->dst_group = ffs(nladdr->nl_groups); |
1da177e4 LT |
1554 | } |
1555 | ||
1556 | return err; | |
1557 | } | |
1558 | ||
6ac552fd PM |
1559 | static int netlink_getname(struct socket *sock, struct sockaddr *addr, |
1560 | int *addr_len, int peer) | |
1da177e4 LT |
1561 | { |
1562 | struct sock *sk = sock->sk; | |
1563 | struct netlink_sock *nlk = nlk_sk(sk); | |
13cfa97b | 1564 | DECLARE_SOCKADDR(struct sockaddr_nl *, nladdr, addr); |
746fac4d | 1565 | |
1da177e4 LT |
1566 | nladdr->nl_family = AF_NETLINK; |
1567 | nladdr->nl_pad = 0; | |
1568 | *addr_len = sizeof(*nladdr); | |
1569 | ||
1570 | if (peer) { | |
15e47304 | 1571 | nladdr->nl_pid = nlk->dst_portid; |
d629b836 | 1572 | nladdr->nl_groups = netlink_group_mask(nlk->dst_group); |
1da177e4 | 1573 | } else { |
15e47304 | 1574 | nladdr->nl_pid = nlk->portid; |
513c2500 | 1575 | nladdr->nl_groups = nlk->groups ? nlk->groups[0] : 0; |
1da177e4 LT |
1576 | } |
1577 | return 0; | |
1578 | } | |
1579 | ||
15e47304 | 1580 | static struct sock *netlink_getsockbyportid(struct sock *ssk, u32 portid) |
1da177e4 | 1581 | { |
1da177e4 LT |
1582 | struct sock *sock; |
1583 | struct netlink_sock *nlk; | |
1584 | ||
15e47304 | 1585 | sock = netlink_lookup(sock_net(ssk), ssk->sk_protocol, portid); |
1da177e4 LT |
1586 | if (!sock) |
1587 | return ERR_PTR(-ECONNREFUSED); | |
1588 | ||
1589 | /* Don't bother queuing skb if kernel socket has no input function */ | |
1590 | nlk = nlk_sk(sock); | |
cd40b7d3 | 1591 | if (sock->sk_state == NETLINK_CONNECTED && |
15e47304 | 1592 | nlk->dst_portid != nlk_sk(ssk)->portid) { |
1da177e4 LT |
1593 | sock_put(sock); |
1594 | return ERR_PTR(-ECONNREFUSED); | |
1595 | } | |
1596 | return sock; | |
1597 | } | |
1598 | ||
1599 | struct sock *netlink_getsockbyfilp(struct file *filp) | |
1600 | { | |
496ad9aa | 1601 | struct inode *inode = file_inode(filp); |
1da177e4 LT |
1602 | struct sock *sock; |
1603 | ||
1604 | if (!S_ISSOCK(inode->i_mode)) | |
1605 | return ERR_PTR(-ENOTSOCK); | |
1606 | ||
1607 | sock = SOCKET_I(inode)->sk; | |
1608 | if (sock->sk_family != AF_NETLINK) | |
1609 | return ERR_PTR(-EINVAL); | |
1610 | ||
1611 | sock_hold(sock); | |
1612 | return sock; | |
1613 | } | |
1614 | ||
3a36515f PN |
1615 | static struct sk_buff *netlink_alloc_large_skb(unsigned int size, |
1616 | int broadcast) | |
c05cdb1b PNA |
1617 | { |
1618 | struct sk_buff *skb; | |
1619 | void *data; | |
1620 | ||
3a36515f | 1621 | if (size <= NLMSG_GOODSIZE || broadcast) |
c05cdb1b PNA |
1622 | return alloc_skb(size, GFP_KERNEL); |
1623 | ||
3a36515f PN |
1624 | size = SKB_DATA_ALIGN(size) + |
1625 | SKB_DATA_ALIGN(sizeof(struct skb_shared_info)); | |
c05cdb1b PNA |
1626 | |
1627 | data = vmalloc(size); | |
1628 | if (data == NULL) | |
3a36515f | 1629 | return NULL; |
c05cdb1b | 1630 | |
3a36515f PN |
1631 | skb = build_skb(data, size); |
1632 | if (skb == NULL) | |
1633 | vfree(data); | |
1634 | else { | |
1635 | skb->head_frag = 0; | |
1636 | skb->destructor = netlink_skb_destructor; | |
1637 | } | |
c05cdb1b PNA |
1638 | |
1639 | return skb; | |
c05cdb1b PNA |
1640 | } |
1641 | ||
1da177e4 LT |
1642 | /* |
1643 | * Attach a skb to a netlink socket. | |
1644 | * The caller must hold a reference to the destination socket. On error, the | |
1645 | * reference is dropped. The skb is not send to the destination, just all | |
1646 | * all error checks are performed and memory in the queue is reserved. | |
1647 | * Return values: | |
1648 | * < 0: error. skb freed, reference to sock dropped. | |
1649 | * 0: continue | |
1650 | * 1: repeat lookup - reference dropped while waiting for socket memory. | |
1651 | */ | |
9457afee | 1652 | int netlink_attachskb(struct sock *sk, struct sk_buff *skb, |
c3d8d1e3 | 1653 | long *timeo, struct sock *ssk) |
1da177e4 LT |
1654 | { |
1655 | struct netlink_sock *nlk; | |
1656 | ||
1657 | nlk = nlk_sk(sk); | |
1658 | ||
5fd96123 PM |
1659 | if ((atomic_read(&sk->sk_rmem_alloc) > sk->sk_rcvbuf || |
1660 | test_bit(NETLINK_CONGESTED, &nlk->state)) && | |
1661 | !netlink_skb_is_mmaped(skb)) { | |
1da177e4 | 1662 | DECLARE_WAITQUEUE(wait, current); |
c3d8d1e3 | 1663 | if (!*timeo) { |
aed81560 | 1664 | if (!ssk || netlink_is_kernel(ssk)) |
1da177e4 LT |
1665 | netlink_overrun(sk); |
1666 | sock_put(sk); | |
1667 | kfree_skb(skb); | |
1668 | return -EAGAIN; | |
1669 | } | |
1670 | ||
1671 | __set_current_state(TASK_INTERRUPTIBLE); | |
1672 | add_wait_queue(&nlk->wait, &wait); | |
1673 | ||
1674 | if ((atomic_read(&sk->sk_rmem_alloc) > sk->sk_rcvbuf || | |
cd967e05 | 1675 | test_bit(NETLINK_CONGESTED, &nlk->state)) && |
1da177e4 | 1676 | !sock_flag(sk, SOCK_DEAD)) |
c3d8d1e3 | 1677 | *timeo = schedule_timeout(*timeo); |
1da177e4 LT |
1678 | |
1679 | __set_current_state(TASK_RUNNING); | |
1680 | remove_wait_queue(&nlk->wait, &wait); | |
1681 | sock_put(sk); | |
1682 | ||
1683 | if (signal_pending(current)) { | |
1684 | kfree_skb(skb); | |
c3d8d1e3 | 1685 | return sock_intr_errno(*timeo); |
1da177e4 LT |
1686 | } |
1687 | return 1; | |
1688 | } | |
cf0a018a | 1689 | netlink_skb_set_owner_r(skb, sk); |
1da177e4 LT |
1690 | return 0; |
1691 | } | |
1692 | ||
4a7e7c2a | 1693 | static int __netlink_sendskb(struct sock *sk, struct sk_buff *skb) |
1da177e4 | 1694 | { |
1da177e4 LT |
1695 | int len = skb->len; |
1696 | ||
bcbde0d4 DB |
1697 | netlink_deliver_tap(skb); |
1698 | ||
f9c22888 PM |
1699 | #ifdef CONFIG_NETLINK_MMAP |
1700 | if (netlink_skb_is_mmaped(skb)) | |
1701 | netlink_queue_mmaped_skb(sk, skb); | |
1702 | else if (netlink_rx_is_mmaped(sk)) | |
1703 | netlink_ring_set_copied(sk, skb); | |
1704 | else | |
1705 | #endif /* CONFIG_NETLINK_MMAP */ | |
1706 | skb_queue_tail(&sk->sk_receive_queue, skb); | |
676d2369 | 1707 | sk->sk_data_ready(sk); |
4a7e7c2a ED |
1708 | return len; |
1709 | } | |
1710 | ||
1711 | int netlink_sendskb(struct sock *sk, struct sk_buff *skb) | |
1712 | { | |
1713 | int len = __netlink_sendskb(sk, skb); | |
1714 | ||
1da177e4 LT |
1715 | sock_put(sk); |
1716 | return len; | |
1717 | } | |
1718 | ||
1719 | void netlink_detachskb(struct sock *sk, struct sk_buff *skb) | |
1720 | { | |
1721 | kfree_skb(skb); | |
1722 | sock_put(sk); | |
1723 | } | |
1724 | ||
b57ef81f | 1725 | static struct sk_buff *netlink_trim(struct sk_buff *skb, gfp_t allocation) |
1da177e4 LT |
1726 | { |
1727 | int delta; | |
1728 | ||
1298ca46 | 1729 | WARN_ON(skb->sk != NULL); |
5fd96123 PM |
1730 | if (netlink_skb_is_mmaped(skb)) |
1731 | return skb; | |
1da177e4 | 1732 | |
4305b541 | 1733 | delta = skb->end - skb->tail; |
c05cdb1b | 1734 | if (is_vmalloc_addr(skb->head) || delta * 2 < skb->truesize) |
1da177e4 LT |
1735 | return skb; |
1736 | ||
1737 | if (skb_shared(skb)) { | |
1738 | struct sk_buff *nskb = skb_clone(skb, allocation); | |
1739 | if (!nskb) | |
1740 | return skb; | |
8460c00f | 1741 | consume_skb(skb); |
1da177e4 LT |
1742 | skb = nskb; |
1743 | } | |
1744 | ||
1745 | if (!pskb_expand_head(skb, 0, -delta, allocation)) | |
1746 | skb->truesize -= delta; | |
1747 | ||
1748 | return skb; | |
1749 | } | |
1750 | ||
3fbc2905 EB |
1751 | static int netlink_unicast_kernel(struct sock *sk, struct sk_buff *skb, |
1752 | struct sock *ssk) | |
cd40b7d3 DL |
1753 | { |
1754 | int ret; | |
1755 | struct netlink_sock *nlk = nlk_sk(sk); | |
1756 | ||
1757 | ret = -ECONNREFUSED; | |
1758 | if (nlk->netlink_rcv != NULL) { | |
1759 | ret = skb->len; | |
cf0a018a | 1760 | netlink_skb_set_owner_r(skb, sk); |
e32123e5 | 1761 | NETLINK_CB(skb).sk = ssk; |
73bfd370 | 1762 | netlink_deliver_tap_kernel(sk, ssk, skb); |
cd40b7d3 | 1763 | nlk->netlink_rcv(skb); |
bfb253c9 ED |
1764 | consume_skb(skb); |
1765 | } else { | |
1766 | kfree_skb(skb); | |
cd40b7d3 | 1767 | } |
cd40b7d3 DL |
1768 | sock_put(sk); |
1769 | return ret; | |
1770 | } | |
1771 | ||
1772 | int netlink_unicast(struct sock *ssk, struct sk_buff *skb, | |
15e47304 | 1773 | u32 portid, int nonblock) |
1da177e4 LT |
1774 | { |
1775 | struct sock *sk; | |
1776 | int err; | |
1777 | long timeo; | |
1778 | ||
1779 | skb = netlink_trim(skb, gfp_any()); | |
1780 | ||
1781 | timeo = sock_sndtimeo(ssk, nonblock); | |
1782 | retry: | |
15e47304 | 1783 | sk = netlink_getsockbyportid(ssk, portid); |
1da177e4 LT |
1784 | if (IS_ERR(sk)) { |
1785 | kfree_skb(skb); | |
1786 | return PTR_ERR(sk); | |
1787 | } | |
cd40b7d3 | 1788 | if (netlink_is_kernel(sk)) |
3fbc2905 | 1789 | return netlink_unicast_kernel(sk, skb, ssk); |
cd40b7d3 | 1790 | |
b1153f29 | 1791 | if (sk_filter(sk, skb)) { |
84874607 | 1792 | err = skb->len; |
b1153f29 SH |
1793 | kfree_skb(skb); |
1794 | sock_put(sk); | |
1795 | return err; | |
1796 | } | |
1797 | ||
9457afee | 1798 | err = netlink_attachskb(sk, skb, &timeo, ssk); |
1da177e4 LT |
1799 | if (err == 1) |
1800 | goto retry; | |
1801 | if (err) | |
1802 | return err; | |
1803 | ||
7ee015e0 | 1804 | return netlink_sendskb(sk, skb); |
1da177e4 | 1805 | } |
6ac552fd | 1806 | EXPORT_SYMBOL(netlink_unicast); |
1da177e4 | 1807 | |
f9c22888 PM |
1808 | struct sk_buff *netlink_alloc_skb(struct sock *ssk, unsigned int size, |
1809 | u32 dst_portid, gfp_t gfp_mask) | |
1810 | { | |
1811 | #ifdef CONFIG_NETLINK_MMAP | |
1812 | struct sock *sk = NULL; | |
1813 | struct sk_buff *skb; | |
1814 | struct netlink_ring *ring; | |
1815 | struct nl_mmap_hdr *hdr; | |
1816 | unsigned int maxlen; | |
1817 | ||
1818 | sk = netlink_getsockbyportid(ssk, dst_portid); | |
1819 | if (IS_ERR(sk)) | |
1820 | goto out; | |
1821 | ||
1822 | ring = &nlk_sk(sk)->rx_ring; | |
1823 | /* fast-path without atomic ops for common case: non-mmaped receiver */ | |
1824 | if (ring->pg_vec == NULL) | |
1825 | goto out_put; | |
1826 | ||
aae9f0e2 TG |
1827 | if (ring->frame_size - NL_MMAP_HDRLEN < size) |
1828 | goto out_put; | |
1829 | ||
f9c22888 PM |
1830 | skb = alloc_skb_head(gfp_mask); |
1831 | if (skb == NULL) | |
1832 | goto err1; | |
1833 | ||
1834 | spin_lock_bh(&sk->sk_receive_queue.lock); | |
1835 | /* check again under lock */ | |
1836 | if (ring->pg_vec == NULL) | |
1837 | goto out_free; | |
1838 | ||
aae9f0e2 | 1839 | /* check again under lock */ |
f9c22888 PM |
1840 | maxlen = ring->frame_size - NL_MMAP_HDRLEN; |
1841 | if (maxlen < size) | |
1842 | goto out_free; | |
1843 | ||
1844 | netlink_forward_ring(ring); | |
1845 | hdr = netlink_current_frame(ring, NL_MMAP_STATUS_UNUSED); | |
1846 | if (hdr == NULL) | |
1847 | goto err2; | |
1848 | netlink_ring_setup_skb(skb, sk, ring, hdr); | |
1849 | netlink_set_status(hdr, NL_MMAP_STATUS_RESERVED); | |
1850 | atomic_inc(&ring->pending); | |
1851 | netlink_increment_head(ring); | |
1852 | ||
1853 | spin_unlock_bh(&sk->sk_receive_queue.lock); | |
1854 | return skb; | |
1855 | ||
1856 | err2: | |
1857 | kfree_skb(skb); | |
1858 | spin_unlock_bh(&sk->sk_receive_queue.lock); | |
cd1df525 | 1859 | netlink_overrun(sk); |
f9c22888 PM |
1860 | err1: |
1861 | sock_put(sk); | |
1862 | return NULL; | |
1863 | ||
1864 | out_free: | |
1865 | kfree_skb(skb); | |
1866 | spin_unlock_bh(&sk->sk_receive_queue.lock); | |
1867 | out_put: | |
1868 | sock_put(sk); | |
1869 | out: | |
1870 | #endif | |
1871 | return alloc_skb(size, gfp_mask); | |
1872 | } | |
1873 | EXPORT_SYMBOL_GPL(netlink_alloc_skb); | |
1874 | ||
4277a083 PM |
1875 | int netlink_has_listeners(struct sock *sk, unsigned int group) |
1876 | { | |
1877 | int res = 0; | |
5c398dc8 | 1878 | struct listeners *listeners; |
4277a083 | 1879 | |
aed81560 | 1880 | BUG_ON(!netlink_is_kernel(sk)); |
b4ff4f04 JB |
1881 | |
1882 | rcu_read_lock(); | |
1883 | listeners = rcu_dereference(nl_table[sk->sk_protocol].listeners); | |
1884 | ||
6d772ac5 | 1885 | if (listeners && group - 1 < nl_table[sk->sk_protocol].groups) |
5c398dc8 | 1886 | res = test_bit(group - 1, listeners->masks); |
b4ff4f04 JB |
1887 | |
1888 | rcu_read_unlock(); | |
1889 | ||
4277a083 PM |
1890 | return res; |
1891 | } | |
1892 | EXPORT_SYMBOL_GPL(netlink_has_listeners); | |
1893 | ||
b57ef81f | 1894 | static int netlink_broadcast_deliver(struct sock *sk, struct sk_buff *skb) |
1da177e4 LT |
1895 | { |
1896 | struct netlink_sock *nlk = nlk_sk(sk); | |
1897 | ||
1898 | if (atomic_read(&sk->sk_rmem_alloc) <= sk->sk_rcvbuf && | |
cd967e05 | 1899 | !test_bit(NETLINK_CONGESTED, &nlk->state)) { |
cf0a018a | 1900 | netlink_skb_set_owner_r(skb, sk); |
4a7e7c2a | 1901 | __netlink_sendskb(sk, skb); |
2c645800 | 1902 | return atomic_read(&sk->sk_rmem_alloc) > (sk->sk_rcvbuf >> 1); |
1da177e4 LT |
1903 | } |
1904 | return -1; | |
1905 | } | |
1906 | ||
1907 | struct netlink_broadcast_data { | |
1908 | struct sock *exclude_sk; | |
b4b51029 | 1909 | struct net *net; |
15e47304 | 1910 | u32 portid; |
1da177e4 LT |
1911 | u32 group; |
1912 | int failure; | |
ff491a73 | 1913 | int delivery_failure; |
1da177e4 LT |
1914 | int congested; |
1915 | int delivered; | |
7d877f3b | 1916 | gfp_t allocation; |
1da177e4 | 1917 | struct sk_buff *skb, *skb2; |
910a7e90 EB |
1918 | int (*tx_filter)(struct sock *dsk, struct sk_buff *skb, void *data); |
1919 | void *tx_data; | |
1da177e4 LT |
1920 | }; |
1921 | ||
46c9521f RR |
1922 | static void do_one_broadcast(struct sock *sk, |
1923 | struct netlink_broadcast_data *p) | |
1da177e4 LT |
1924 | { |
1925 | struct netlink_sock *nlk = nlk_sk(sk); | |
1926 | int val; | |
1927 | ||
1928 | if (p->exclude_sk == sk) | |
46c9521f | 1929 | return; |
1da177e4 | 1930 | |
15e47304 | 1931 | if (nlk->portid == p->portid || p->group - 1 >= nlk->ngroups || |
f7fa9b10 | 1932 | !test_bit(p->group - 1, nlk->groups)) |
46c9521f | 1933 | return; |
1da177e4 | 1934 | |
878628fb | 1935 | if (!net_eq(sock_net(sk), p->net)) |
46c9521f | 1936 | return; |
b4b51029 | 1937 | |
1da177e4 LT |
1938 | if (p->failure) { |
1939 | netlink_overrun(sk); | |
46c9521f | 1940 | return; |
1da177e4 LT |
1941 | } |
1942 | ||
1943 | sock_hold(sk); | |
1944 | if (p->skb2 == NULL) { | |
68acc024 | 1945 | if (skb_shared(p->skb)) { |
1da177e4 LT |
1946 | p->skb2 = skb_clone(p->skb, p->allocation); |
1947 | } else { | |
68acc024 TC |
1948 | p->skb2 = skb_get(p->skb); |
1949 | /* | |
1950 | * skb ownership may have been set when | |
1951 | * delivered to a previous socket. | |
1952 | */ | |
1953 | skb_orphan(p->skb2); | |
1da177e4 LT |
1954 | } |
1955 | } | |
1956 | if (p->skb2 == NULL) { | |
1957 | netlink_overrun(sk); | |
1958 | /* Clone failed. Notify ALL listeners. */ | |
1959 | p->failure = 1; | |
be0c22a4 PNA |
1960 | if (nlk->flags & NETLINK_BROADCAST_SEND_ERROR) |
1961 | p->delivery_failure = 1; | |
910a7e90 EB |
1962 | } else if (p->tx_filter && p->tx_filter(sk, p->skb2, p->tx_data)) { |
1963 | kfree_skb(p->skb2); | |
1964 | p->skb2 = NULL; | |
b1153f29 SH |
1965 | } else if (sk_filter(sk, p->skb2)) { |
1966 | kfree_skb(p->skb2); | |
1967 | p->skb2 = NULL; | |
1da177e4 LT |
1968 | } else if ((val = netlink_broadcast_deliver(sk, p->skb2)) < 0) { |
1969 | netlink_overrun(sk); | |
be0c22a4 PNA |
1970 | if (nlk->flags & NETLINK_BROADCAST_SEND_ERROR) |
1971 | p->delivery_failure = 1; | |
1da177e4 LT |
1972 | } else { |
1973 | p->congested |= val; | |
1974 | p->delivered = 1; | |
1975 | p->skb2 = NULL; | |
1976 | } | |
1977 | sock_put(sk); | |
1da177e4 LT |
1978 | } |
1979 | ||
15e47304 | 1980 | int netlink_broadcast_filtered(struct sock *ssk, struct sk_buff *skb, u32 portid, |
910a7e90 EB |
1981 | u32 group, gfp_t allocation, |
1982 | int (*filter)(struct sock *dsk, struct sk_buff *skb, void *data), | |
1983 | void *filter_data) | |
1da177e4 | 1984 | { |
3b1e0a65 | 1985 | struct net *net = sock_net(ssk); |
1da177e4 | 1986 | struct netlink_broadcast_data info; |
1da177e4 LT |
1987 | struct sock *sk; |
1988 | ||
1989 | skb = netlink_trim(skb, allocation); | |
1990 | ||
1991 | info.exclude_sk = ssk; | |
b4b51029 | 1992 | info.net = net; |
15e47304 | 1993 | info.portid = portid; |
1da177e4 LT |
1994 | info.group = group; |
1995 | info.failure = 0; | |
ff491a73 | 1996 | info.delivery_failure = 0; |
1da177e4 LT |
1997 | info.congested = 0; |
1998 | info.delivered = 0; | |
1999 | info.allocation = allocation; | |
2000 | info.skb = skb; | |
2001 | info.skb2 = NULL; | |
910a7e90 EB |
2002 | info.tx_filter = filter; |
2003 | info.tx_data = filter_data; | |
1da177e4 LT |
2004 | |
2005 | /* While we sleep in clone, do not allow to change socket list */ | |
2006 | ||
2007 | netlink_lock_table(); | |
2008 | ||
b67bfe0d | 2009 | sk_for_each_bound(sk, &nl_table[ssk->sk_protocol].mc_list) |
1da177e4 LT |
2010 | do_one_broadcast(sk, &info); |
2011 | ||
70d4bf6d | 2012 | consume_skb(skb); |
aa1c6a6f | 2013 | |
1da177e4 LT |
2014 | netlink_unlock_table(); |
2015 | ||
70d4bf6d NH |
2016 | if (info.delivery_failure) { |
2017 | kfree_skb(info.skb2); | |
ff491a73 | 2018 | return -ENOBUFS; |
658cb354 ED |
2019 | } |
2020 | consume_skb(info.skb2); | |
ff491a73 | 2021 | |
1da177e4 LT |
2022 | if (info.delivered) { |
2023 | if (info.congested && (allocation & __GFP_WAIT)) | |
2024 | yield(); | |
2025 | return 0; | |
2026 | } | |
1da177e4 LT |
2027 | return -ESRCH; |
2028 | } | |
910a7e90 EB |
2029 | EXPORT_SYMBOL(netlink_broadcast_filtered); |
2030 | ||
15e47304 | 2031 | int netlink_broadcast(struct sock *ssk, struct sk_buff *skb, u32 portid, |
910a7e90 EB |
2032 | u32 group, gfp_t allocation) |
2033 | { | |
15e47304 | 2034 | return netlink_broadcast_filtered(ssk, skb, portid, group, allocation, |
910a7e90 EB |
2035 | NULL, NULL); |
2036 | } | |
6ac552fd | 2037 | EXPORT_SYMBOL(netlink_broadcast); |
1da177e4 LT |
2038 | |
2039 | struct netlink_set_err_data { | |
2040 | struct sock *exclude_sk; | |
15e47304 | 2041 | u32 portid; |
1da177e4 LT |
2042 | u32 group; |
2043 | int code; | |
2044 | }; | |
2045 | ||
b57ef81f | 2046 | static int do_one_set_err(struct sock *sk, struct netlink_set_err_data *p) |
1da177e4 LT |
2047 | { |
2048 | struct netlink_sock *nlk = nlk_sk(sk); | |
1a50307b | 2049 | int ret = 0; |
1da177e4 LT |
2050 | |
2051 | if (sk == p->exclude_sk) | |
2052 | goto out; | |
2053 | ||
09ad9bc7 | 2054 | if (!net_eq(sock_net(sk), sock_net(p->exclude_sk))) |
b4b51029 EB |
2055 | goto out; |
2056 | ||
15e47304 | 2057 | if (nlk->portid == p->portid || p->group - 1 >= nlk->ngroups || |
f7fa9b10 | 2058 | !test_bit(p->group - 1, nlk->groups)) |
1da177e4 LT |
2059 | goto out; |
2060 | ||
1a50307b PNA |
2061 | if (p->code == ENOBUFS && nlk->flags & NETLINK_RECV_NO_ENOBUFS) { |
2062 | ret = 1; | |
2063 | goto out; | |
2064 | } | |
2065 | ||
1da177e4 LT |
2066 | sk->sk_err = p->code; |
2067 | sk->sk_error_report(sk); | |
2068 | out: | |
1a50307b | 2069 | return ret; |
1da177e4 LT |
2070 | } |
2071 | ||
4843b93c PNA |
2072 | /** |
2073 | * netlink_set_err - report error to broadcast listeners | |
2074 | * @ssk: the kernel netlink socket, as returned by netlink_kernel_create() | |
15e47304 | 2075 | * @portid: the PORTID of a process that we want to skip (if any) |
840e93f2 | 2076 | * @group: the broadcast group that will notice the error |
4843b93c | 2077 | * @code: error code, must be negative (as usual in kernelspace) |
1a50307b PNA |
2078 | * |
2079 | * This function returns the number of broadcast listeners that have set the | |
2080 | * NETLINK_RECV_NO_ENOBUFS socket option. | |
4843b93c | 2081 | */ |
15e47304 | 2082 | int netlink_set_err(struct sock *ssk, u32 portid, u32 group, int code) |
1da177e4 LT |
2083 | { |
2084 | struct netlink_set_err_data info; | |
1da177e4 | 2085 | struct sock *sk; |
1a50307b | 2086 | int ret = 0; |
1da177e4 LT |
2087 | |
2088 | info.exclude_sk = ssk; | |
15e47304 | 2089 | info.portid = portid; |
1da177e4 | 2090 | info.group = group; |
4843b93c PNA |
2091 | /* sk->sk_err wants a positive error value */ |
2092 | info.code = -code; | |
1da177e4 LT |
2093 | |
2094 | read_lock(&nl_table_lock); | |
2095 | ||
b67bfe0d | 2096 | sk_for_each_bound(sk, &nl_table[ssk->sk_protocol].mc_list) |
1a50307b | 2097 | ret += do_one_set_err(sk, &info); |
1da177e4 LT |
2098 | |
2099 | read_unlock(&nl_table_lock); | |
1a50307b | 2100 | return ret; |
1da177e4 | 2101 | } |
dd5b6ce6 | 2102 | EXPORT_SYMBOL(netlink_set_err); |
1da177e4 | 2103 | |
84659eb5 JB |
2104 | /* must be called with netlink table grabbed */ |
2105 | static void netlink_update_socket_mc(struct netlink_sock *nlk, | |
2106 | unsigned int group, | |
2107 | int is_new) | |
2108 | { | |
2109 | int old, new = !!is_new, subscriptions; | |
2110 | ||
2111 | old = test_bit(group - 1, nlk->groups); | |
2112 | subscriptions = nlk->subscriptions - old + new; | |
2113 | if (new) | |
2114 | __set_bit(group - 1, nlk->groups); | |
2115 | else | |
2116 | __clear_bit(group - 1, nlk->groups); | |
2117 | netlink_update_subscriptions(&nlk->sk, subscriptions); | |
2118 | netlink_update_listeners(&nlk->sk); | |
2119 | } | |
2120 | ||
9a4595bc | 2121 | static int netlink_setsockopt(struct socket *sock, int level, int optname, |
b7058842 | 2122 | char __user *optval, unsigned int optlen) |
9a4595bc PM |
2123 | { |
2124 | struct sock *sk = sock->sk; | |
2125 | struct netlink_sock *nlk = nlk_sk(sk); | |
eb496534 JB |
2126 | unsigned int val = 0; |
2127 | int err; | |
9a4595bc PM |
2128 | |
2129 | if (level != SOL_NETLINK) | |
2130 | return -ENOPROTOOPT; | |
2131 | ||
ccdfcc39 PM |
2132 | if (optname != NETLINK_RX_RING && optname != NETLINK_TX_RING && |
2133 | optlen >= sizeof(int) && | |
eb496534 | 2134 | get_user(val, (unsigned int __user *)optval)) |
9a4595bc PM |
2135 | return -EFAULT; |
2136 | ||
2137 | switch (optname) { | |
2138 | case NETLINK_PKTINFO: | |
2139 | if (val) | |
2140 | nlk->flags |= NETLINK_RECV_PKTINFO; | |
2141 | else | |
2142 | nlk->flags &= ~NETLINK_RECV_PKTINFO; | |
2143 | err = 0; | |
2144 | break; | |
2145 | case NETLINK_ADD_MEMBERSHIP: | |
2146 | case NETLINK_DROP_MEMBERSHIP: { | |
5187cd05 | 2147 | if (!netlink_allowed(sock, NL_CFG_F_NONROOT_RECV)) |
9a4595bc | 2148 | return -EPERM; |
b4ff4f04 JB |
2149 | err = netlink_realloc_groups(sk); |
2150 | if (err) | |
2151 | return err; | |
9a4595bc PM |
2152 | if (!val || val - 1 >= nlk->ngroups) |
2153 | return -EINVAL; | |
7774d5e0 | 2154 | if (optname == NETLINK_ADD_MEMBERSHIP && nlk->netlink_bind) { |
023e2cfa | 2155 | err = nlk->netlink_bind(sock_net(sk), val); |
4f520900 RGB |
2156 | if (err) |
2157 | return err; | |
2158 | } | |
9a4595bc | 2159 | netlink_table_grab(); |
84659eb5 JB |
2160 | netlink_update_socket_mc(nlk, val, |
2161 | optname == NETLINK_ADD_MEMBERSHIP); | |
9a4595bc | 2162 | netlink_table_ungrab(); |
7774d5e0 | 2163 | if (optname == NETLINK_DROP_MEMBERSHIP && nlk->netlink_unbind) |
023e2cfa | 2164 | nlk->netlink_unbind(sock_net(sk), val); |
03292745 | 2165 | |
9a4595bc PM |
2166 | err = 0; |
2167 | break; | |
2168 | } | |
be0c22a4 PNA |
2169 | case NETLINK_BROADCAST_ERROR: |
2170 | if (val) | |
2171 | nlk->flags |= NETLINK_BROADCAST_SEND_ERROR; | |
2172 | else | |
2173 | nlk->flags &= ~NETLINK_BROADCAST_SEND_ERROR; | |
2174 | err = 0; | |
2175 | break; | |
38938bfe PNA |
2176 | case NETLINK_NO_ENOBUFS: |
2177 | if (val) { | |
2178 | nlk->flags |= NETLINK_RECV_NO_ENOBUFS; | |
cd967e05 | 2179 | clear_bit(NETLINK_CONGESTED, &nlk->state); |
38938bfe | 2180 | wake_up_interruptible(&nlk->wait); |
658cb354 | 2181 | } else { |
38938bfe | 2182 | nlk->flags &= ~NETLINK_RECV_NO_ENOBUFS; |
658cb354 | 2183 | } |
38938bfe PNA |
2184 | err = 0; |
2185 | break; | |
ccdfcc39 PM |
2186 | #ifdef CONFIG_NETLINK_MMAP |
2187 | case NETLINK_RX_RING: | |
2188 | case NETLINK_TX_RING: { | |
2189 | struct nl_mmap_req req; | |
2190 | ||
2191 | /* Rings might consume more memory than queue limits, require | |
2192 | * CAP_NET_ADMIN. | |
2193 | */ | |
2194 | if (!capable(CAP_NET_ADMIN)) | |
2195 | return -EPERM; | |
2196 | if (optlen < sizeof(req)) | |
2197 | return -EINVAL; | |
2198 | if (copy_from_user(&req, optval, sizeof(req))) | |
2199 | return -EFAULT; | |
2200 | err = netlink_set_ring(sk, &req, false, | |
2201 | optname == NETLINK_TX_RING); | |
2202 | break; | |
2203 | } | |
2204 | #endif /* CONFIG_NETLINK_MMAP */ | |
9a4595bc PM |
2205 | default: |
2206 | err = -ENOPROTOOPT; | |
2207 | } | |
2208 | return err; | |
2209 | } | |
2210 | ||
2211 | static int netlink_getsockopt(struct socket *sock, int level, int optname, | |
746fac4d | 2212 | char __user *optval, int __user *optlen) |
9a4595bc PM |
2213 | { |
2214 | struct sock *sk = sock->sk; | |
2215 | struct netlink_sock *nlk = nlk_sk(sk); | |
2216 | int len, val, err; | |
2217 | ||
2218 | if (level != SOL_NETLINK) | |
2219 | return -ENOPROTOOPT; | |
2220 | ||
2221 | if (get_user(len, optlen)) | |
2222 | return -EFAULT; | |
2223 | if (len < 0) | |
2224 | return -EINVAL; | |
2225 | ||
2226 | switch (optname) { | |
2227 | case NETLINK_PKTINFO: | |
2228 | if (len < sizeof(int)) | |
2229 | return -EINVAL; | |
2230 | len = sizeof(int); | |
2231 | val = nlk->flags & NETLINK_RECV_PKTINFO ? 1 : 0; | |
a27b58fe HC |
2232 | if (put_user(len, optlen) || |
2233 | put_user(val, optval)) | |
2234 | return -EFAULT; | |
9a4595bc PM |
2235 | err = 0; |
2236 | break; | |
be0c22a4 PNA |
2237 | case NETLINK_BROADCAST_ERROR: |
2238 | if (len < sizeof(int)) | |
2239 | return -EINVAL; | |
2240 | len = sizeof(int); | |
2241 | val = nlk->flags & NETLINK_BROADCAST_SEND_ERROR ? 1 : 0; | |
2242 | if (put_user(len, optlen) || | |
2243 | put_user(val, optval)) | |
2244 | return -EFAULT; | |
2245 | err = 0; | |
2246 | break; | |
38938bfe PNA |
2247 | case NETLINK_NO_ENOBUFS: |
2248 | if (len < sizeof(int)) | |
2249 | return -EINVAL; | |
2250 | len = sizeof(int); | |
2251 | val = nlk->flags & NETLINK_RECV_NO_ENOBUFS ? 1 : 0; | |
2252 | if (put_user(len, optlen) || | |
2253 | put_user(val, optval)) | |
2254 | return -EFAULT; | |
2255 | err = 0; | |
2256 | break; | |
9a4595bc PM |
2257 | default: |
2258 | err = -ENOPROTOOPT; | |
2259 | } | |
2260 | return err; | |
2261 | } | |
2262 | ||
2263 | static void netlink_cmsg_recv_pktinfo(struct msghdr *msg, struct sk_buff *skb) | |
2264 | { | |
2265 | struct nl_pktinfo info; | |
2266 | ||
2267 | info.group = NETLINK_CB(skb).dst_group; | |
2268 | put_cmsg(msg, SOL_NETLINK, NETLINK_PKTINFO, sizeof(info), &info); | |
2269 | } | |
2270 | ||
1b784140 | 2271 | static int netlink_sendmsg(struct socket *sock, struct msghdr *msg, size_t len) |
1da177e4 | 2272 | { |
1da177e4 LT |
2273 | struct sock *sk = sock->sk; |
2274 | struct netlink_sock *nlk = nlk_sk(sk); | |
342dfc30 | 2275 | DECLARE_SOCKADDR(struct sockaddr_nl *, addr, msg->msg_name); |
15e47304 | 2276 | u32 dst_portid; |
d629b836 | 2277 | u32 dst_group; |
1da177e4 LT |
2278 | struct sk_buff *skb; |
2279 | int err; | |
2280 | struct scm_cookie scm; | |
2d7a85f4 | 2281 | u32 netlink_skb_flags = 0; |
1da177e4 LT |
2282 | |
2283 | if (msg->msg_flags&MSG_OOB) | |
2284 | return -EOPNOTSUPP; | |
2285 | ||
7cc05662 | 2286 | err = scm_send(sock, msg, &scm, true); |
1da177e4 LT |
2287 | if (err < 0) |
2288 | return err; | |
2289 | ||
2290 | if (msg->msg_namelen) { | |
b47030c7 | 2291 | err = -EINVAL; |
1da177e4 | 2292 | if (addr->nl_family != AF_NETLINK) |
b47030c7 | 2293 | goto out; |
15e47304 | 2294 | dst_portid = addr->nl_pid; |
d629b836 | 2295 | dst_group = ffs(addr->nl_groups); |
b47030c7 | 2296 | err = -EPERM; |
15e47304 | 2297 | if ((dst_group || dst_portid) && |
5187cd05 | 2298 | !netlink_allowed(sock, NL_CFG_F_NONROOT_SEND)) |
b47030c7 | 2299 | goto out; |
2d7a85f4 | 2300 | netlink_skb_flags |= NETLINK_SKB_DST; |
1da177e4 | 2301 | } else { |
15e47304 | 2302 | dst_portid = nlk->dst_portid; |
d629b836 | 2303 | dst_group = nlk->dst_group; |
1da177e4 LT |
2304 | } |
2305 | ||
15e47304 | 2306 | if (!nlk->portid) { |
1da177e4 LT |
2307 | err = netlink_autobind(sock); |
2308 | if (err) | |
2309 | goto out; | |
2310 | } | |
2311 | ||
a8866ff6 AV |
2312 | /* It's a really convoluted way for userland to ask for mmaped |
2313 | * sendmsg(), but that's what we've got... | |
2314 | */ | |
5fd96123 | 2315 | if (netlink_tx_is_mmaped(sk) && |
a8866ff6 AV |
2316 | msg->msg_iter.type == ITER_IOVEC && |
2317 | msg->msg_iter.nr_segs == 1 && | |
c0371da6 | 2318 | msg->msg_iter.iov->iov_base == NULL) { |
5fd96123 | 2319 | err = netlink_mmap_sendmsg(sk, msg, dst_portid, dst_group, |
7cc05662 | 2320 | &scm); |
5fd96123 PM |
2321 | goto out; |
2322 | } | |
2323 | ||
1da177e4 LT |
2324 | err = -EMSGSIZE; |
2325 | if (len > sk->sk_sndbuf - 32) | |
2326 | goto out; | |
2327 | err = -ENOBUFS; | |
3a36515f | 2328 | skb = netlink_alloc_large_skb(len, dst_group); |
6ac552fd | 2329 | if (skb == NULL) |
1da177e4 LT |
2330 | goto out; |
2331 | ||
15e47304 | 2332 | NETLINK_CB(skb).portid = nlk->portid; |
d629b836 | 2333 | NETLINK_CB(skb).dst_group = dst_group; |
7cc05662 | 2334 | NETLINK_CB(skb).creds = scm.creds; |
2d7a85f4 | 2335 | NETLINK_CB(skb).flags = netlink_skb_flags; |
1da177e4 | 2336 | |
1da177e4 | 2337 | err = -EFAULT; |
6ce8e9ce | 2338 | if (memcpy_from_msg(skb_put(skb, len), msg, len)) { |
1da177e4 LT |
2339 | kfree_skb(skb); |
2340 | goto out; | |
2341 | } | |
2342 | ||
2343 | err = security_netlink_send(sk, skb); | |
2344 | if (err) { | |
2345 | kfree_skb(skb); | |
2346 | goto out; | |
2347 | } | |
2348 | ||
d629b836 | 2349 | if (dst_group) { |
1da177e4 | 2350 | atomic_inc(&skb->users); |
15e47304 | 2351 | netlink_broadcast(sk, skb, dst_portid, dst_group, GFP_KERNEL); |
1da177e4 | 2352 | } |
15e47304 | 2353 | err = netlink_unicast(sk, skb, dst_portid, msg->msg_flags&MSG_DONTWAIT); |
1da177e4 LT |
2354 | |
2355 | out: | |
7cc05662 | 2356 | scm_destroy(&scm); |
1da177e4 LT |
2357 | return err; |
2358 | } | |
2359 | ||
1b784140 | 2360 | static int netlink_recvmsg(struct socket *sock, struct msghdr *msg, size_t len, |
1da177e4 LT |
2361 | int flags) |
2362 | { | |
1da177e4 LT |
2363 | struct scm_cookie scm; |
2364 | struct sock *sk = sock->sk; | |
2365 | struct netlink_sock *nlk = nlk_sk(sk); | |
2366 | int noblock = flags&MSG_DONTWAIT; | |
2367 | size_t copied; | |
68d6ac6d | 2368 | struct sk_buff *skb, *data_skb; |
b44d211e | 2369 | int err, ret; |
1da177e4 LT |
2370 | |
2371 | if (flags&MSG_OOB) | |
2372 | return -EOPNOTSUPP; | |
2373 | ||
2374 | copied = 0; | |
2375 | ||
6ac552fd PM |
2376 | skb = skb_recv_datagram(sk, flags, noblock, &err); |
2377 | if (skb == NULL) | |
1da177e4 LT |
2378 | goto out; |
2379 | ||
68d6ac6d JB |
2380 | data_skb = skb; |
2381 | ||
1dacc76d JB |
2382 | #ifdef CONFIG_COMPAT_NETLINK_MESSAGES |
2383 | if (unlikely(skb_shinfo(skb)->frag_list)) { | |
1dacc76d | 2384 | /* |
68d6ac6d JB |
2385 | * If this skb has a frag_list, then here that means that we |
2386 | * will have to use the frag_list skb's data for compat tasks | |
2387 | * and the regular skb's data for normal (non-compat) tasks. | |
1dacc76d | 2388 | * |
68d6ac6d JB |
2389 | * If we need to send the compat skb, assign it to the |
2390 | * 'data_skb' variable so that it will be used below for data | |
2391 | * copying. We keep 'skb' for everything else, including | |
2392 | * freeing both later. | |
1dacc76d | 2393 | */ |
68d6ac6d JB |
2394 | if (flags & MSG_CMSG_COMPAT) |
2395 | data_skb = skb_shinfo(skb)->frag_list; | |
1dacc76d JB |
2396 | } |
2397 | #endif | |
2398 | ||
9063e21f ED |
2399 | /* Record the max length of recvmsg() calls for future allocations */ |
2400 | nlk->max_recvmsg_len = max(nlk->max_recvmsg_len, len); | |
2401 | nlk->max_recvmsg_len = min_t(size_t, nlk->max_recvmsg_len, | |
2402 | 16384); | |
2403 | ||
68d6ac6d | 2404 | copied = data_skb->len; |
1da177e4 LT |
2405 | if (len < copied) { |
2406 | msg->msg_flags |= MSG_TRUNC; | |
2407 | copied = len; | |
2408 | } | |
2409 | ||
68d6ac6d | 2410 | skb_reset_transport_header(data_skb); |
51f3d02b | 2411 | err = skb_copy_datagram_msg(data_skb, 0, msg, copied); |
1da177e4 LT |
2412 | |
2413 | if (msg->msg_name) { | |
342dfc30 | 2414 | DECLARE_SOCKADDR(struct sockaddr_nl *, addr, msg->msg_name); |
1da177e4 LT |
2415 | addr->nl_family = AF_NETLINK; |
2416 | addr->nl_pad = 0; | |
15e47304 | 2417 | addr->nl_pid = NETLINK_CB(skb).portid; |
d629b836 | 2418 | addr->nl_groups = netlink_group_mask(NETLINK_CB(skb).dst_group); |
1da177e4 LT |
2419 | msg->msg_namelen = sizeof(*addr); |
2420 | } | |
2421 | ||
cc9a06cd PM |
2422 | if (nlk->flags & NETLINK_RECV_PKTINFO) |
2423 | netlink_cmsg_recv_pktinfo(msg, skb); | |
2424 | ||
7cc05662 CH |
2425 | memset(&scm, 0, sizeof(scm)); |
2426 | scm.creds = *NETLINK_CREDS(skb); | |
188ccb55 | 2427 | if (flags & MSG_TRUNC) |
68d6ac6d | 2428 | copied = data_skb->len; |
daa3766e | 2429 | |
1da177e4 LT |
2430 | skb_free_datagram(sk, skb); |
2431 | ||
16b304f3 PS |
2432 | if (nlk->cb_running && |
2433 | atomic_read(&sk->sk_rmem_alloc) <= sk->sk_rcvbuf / 2) { | |
b44d211e AV |
2434 | ret = netlink_dump(sk); |
2435 | if (ret) { | |
ac30ef83 | 2436 | sk->sk_err = -ret; |
b44d211e AV |
2437 | sk->sk_error_report(sk); |
2438 | } | |
2439 | } | |
1da177e4 | 2440 | |
7cc05662 | 2441 | scm_recv(sock, msg, &scm, flags); |
1da177e4 LT |
2442 | out: |
2443 | netlink_rcv_wake(sk); | |
2444 | return err ? : copied; | |
2445 | } | |
2446 | ||
676d2369 | 2447 | static void netlink_data_ready(struct sock *sk) |
1da177e4 | 2448 | { |
cd40b7d3 | 2449 | BUG(); |
1da177e4 LT |
2450 | } |
2451 | ||
2452 | /* | |
746fac4d | 2453 | * We export these functions to other modules. They provide a |
1da177e4 LT |
2454 | * complete set of kernel non-blocking support for message |
2455 | * queueing. | |
2456 | */ | |
2457 | ||
2458 | struct sock * | |
9f00d977 PNA |
2459 | __netlink_kernel_create(struct net *net, int unit, struct module *module, |
2460 | struct netlink_kernel_cfg *cfg) | |
1da177e4 LT |
2461 | { |
2462 | struct socket *sock; | |
2463 | struct sock *sk; | |
77247bbb | 2464 | struct netlink_sock *nlk; |
5c398dc8 | 2465 | struct listeners *listeners = NULL; |
a31f2d17 PNA |
2466 | struct mutex *cb_mutex = cfg ? cfg->cb_mutex : NULL; |
2467 | unsigned int groups; | |
1da177e4 | 2468 | |
fab2caf6 | 2469 | BUG_ON(!nl_table); |
1da177e4 | 2470 | |
6ac552fd | 2471 | if (unit < 0 || unit >= MAX_LINKS) |
1da177e4 LT |
2472 | return NULL; |
2473 | ||
2474 | if (sock_create_lite(PF_NETLINK, SOCK_DGRAM, unit, &sock)) | |
2475 | return NULL; | |
2476 | ||
23fe1866 PE |
2477 | /* |
2478 | * We have to just have a reference on the net from sk, but don't | |
2479 | * get_net it. Besides, we cannot get and then put the net here. | |
2480 | * So we create one inside init_net and the move it to net. | |
2481 | */ | |
2482 | ||
2483 | if (__netlink_create(&init_net, sock, cb_mutex, unit) < 0) | |
2484 | goto out_sock_release_nosk; | |
2485 | ||
2486 | sk = sock->sk; | |
edf02087 | 2487 | sk_change_net(sk, net); |
4fdb3bb7 | 2488 | |
a31f2d17 | 2489 | if (!cfg || cfg->groups < 32) |
4277a083 | 2490 | groups = 32; |
a31f2d17 PNA |
2491 | else |
2492 | groups = cfg->groups; | |
4277a083 | 2493 | |
5c398dc8 | 2494 | listeners = kzalloc(sizeof(*listeners) + NLGRPSZ(groups), GFP_KERNEL); |
4277a083 PM |
2495 | if (!listeners) |
2496 | goto out_sock_release; | |
2497 | ||
1da177e4 | 2498 | sk->sk_data_ready = netlink_data_ready; |
a31f2d17 PNA |
2499 | if (cfg && cfg->input) |
2500 | nlk_sk(sk)->netlink_rcv = cfg->input; | |
1da177e4 | 2501 | |
8ea65f4a | 2502 | if (netlink_insert(sk, 0)) |
77247bbb | 2503 | goto out_sock_release; |
4fdb3bb7 | 2504 | |
77247bbb PM |
2505 | nlk = nlk_sk(sk); |
2506 | nlk->flags |= NETLINK_KERNEL_SOCKET; | |
4fdb3bb7 | 2507 | |
4fdb3bb7 | 2508 | netlink_table_grab(); |
b4b51029 EB |
2509 | if (!nl_table[unit].registered) { |
2510 | nl_table[unit].groups = groups; | |
5c398dc8 | 2511 | rcu_assign_pointer(nl_table[unit].listeners, listeners); |
b4b51029 EB |
2512 | nl_table[unit].cb_mutex = cb_mutex; |
2513 | nl_table[unit].module = module; | |
9785e10a PNA |
2514 | if (cfg) { |
2515 | nl_table[unit].bind = cfg->bind; | |
6251edd9 | 2516 | nl_table[unit].unbind = cfg->unbind; |
9785e10a | 2517 | nl_table[unit].flags = cfg->flags; |
da12c90e G |
2518 | if (cfg->compare) |
2519 | nl_table[unit].compare = cfg->compare; | |
9785e10a | 2520 | } |
b4b51029 | 2521 | nl_table[unit].registered = 1; |
f937f1f4 JJ |
2522 | } else { |
2523 | kfree(listeners); | |
869e58f8 | 2524 | nl_table[unit].registered++; |
b4b51029 | 2525 | } |
4fdb3bb7 | 2526 | netlink_table_ungrab(); |
77247bbb PM |
2527 | return sk; |
2528 | ||
4fdb3bb7 | 2529 | out_sock_release: |
4277a083 | 2530 | kfree(listeners); |
9dfbec1f | 2531 | netlink_kernel_release(sk); |
23fe1866 PE |
2532 | return NULL; |
2533 | ||
2534 | out_sock_release_nosk: | |
4fdb3bb7 | 2535 | sock_release(sock); |
77247bbb | 2536 | return NULL; |
1da177e4 | 2537 | } |
9f00d977 | 2538 | EXPORT_SYMBOL(__netlink_kernel_create); |
b7c6ba6e DL |
2539 | |
2540 | void | |
2541 | netlink_kernel_release(struct sock *sk) | |
2542 | { | |
edf02087 | 2543 | sk_release_kernel(sk); |
b7c6ba6e DL |
2544 | } |
2545 | EXPORT_SYMBOL(netlink_kernel_release); | |
2546 | ||
d136f1bd | 2547 | int __netlink_change_ngroups(struct sock *sk, unsigned int groups) |
b4ff4f04 | 2548 | { |
5c398dc8 | 2549 | struct listeners *new, *old; |
b4ff4f04 | 2550 | struct netlink_table *tbl = &nl_table[sk->sk_protocol]; |
b4ff4f04 JB |
2551 | |
2552 | if (groups < 32) | |
2553 | groups = 32; | |
2554 | ||
b4ff4f04 | 2555 | if (NLGRPSZ(tbl->groups) < NLGRPSZ(groups)) { |
5c398dc8 ED |
2556 | new = kzalloc(sizeof(*new) + NLGRPSZ(groups), GFP_ATOMIC); |
2557 | if (!new) | |
d136f1bd | 2558 | return -ENOMEM; |
6d772ac5 | 2559 | old = nl_deref_protected(tbl->listeners); |
5c398dc8 ED |
2560 | memcpy(new->masks, old->masks, NLGRPSZ(tbl->groups)); |
2561 | rcu_assign_pointer(tbl->listeners, new); | |
2562 | ||
37b6b935 | 2563 | kfree_rcu(old, rcu); |
b4ff4f04 JB |
2564 | } |
2565 | tbl->groups = groups; | |
2566 | ||
d136f1bd JB |
2567 | return 0; |
2568 | } | |
2569 | ||
2570 | /** | |
2571 | * netlink_change_ngroups - change number of multicast groups | |
2572 | * | |
2573 | * This changes the number of multicast groups that are available | |
2574 | * on a certain netlink family. Note that it is not possible to | |
2575 | * change the number of groups to below 32. Also note that it does | |
2576 | * not implicitly call netlink_clear_multicast_users() when the | |
2577 | * number of groups is reduced. | |
2578 | * | |
2579 | * @sk: The kernel netlink socket, as returned by netlink_kernel_create(). | |
2580 | * @groups: The new number of groups. | |
2581 | */ | |
2582 | int netlink_change_ngroups(struct sock *sk, unsigned int groups) | |
2583 | { | |
2584 | int err; | |
2585 | ||
2586 | netlink_table_grab(); | |
2587 | err = __netlink_change_ngroups(sk, groups); | |
b4ff4f04 | 2588 | netlink_table_ungrab(); |
d136f1bd | 2589 | |
b4ff4f04 JB |
2590 | return err; |
2591 | } | |
b4ff4f04 | 2592 | |
b8273570 JB |
2593 | void __netlink_clear_multicast_users(struct sock *ksk, unsigned int group) |
2594 | { | |
2595 | struct sock *sk; | |
b8273570 JB |
2596 | struct netlink_table *tbl = &nl_table[ksk->sk_protocol]; |
2597 | ||
b67bfe0d | 2598 | sk_for_each_bound(sk, &tbl->mc_list) |
b8273570 JB |
2599 | netlink_update_socket_mc(nlk_sk(sk), group, 0); |
2600 | } | |
2601 | ||
a46621a3 | 2602 | struct nlmsghdr * |
15e47304 | 2603 | __nlmsg_put(struct sk_buff *skb, u32 portid, u32 seq, int type, int len, int flags) |
a46621a3 DV |
2604 | { |
2605 | struct nlmsghdr *nlh; | |
573ce260 | 2606 | int size = nlmsg_msg_size(len); |
a46621a3 | 2607 | |
23b45672 | 2608 | nlh = (struct nlmsghdr *)skb_put(skb, NLMSG_ALIGN(size)); |
a46621a3 DV |
2609 | nlh->nlmsg_type = type; |
2610 | nlh->nlmsg_len = size; | |
2611 | nlh->nlmsg_flags = flags; | |
15e47304 | 2612 | nlh->nlmsg_pid = portid; |
a46621a3 DV |
2613 | nlh->nlmsg_seq = seq; |
2614 | if (!__builtin_constant_p(size) || NLMSG_ALIGN(size) - size != 0) | |
573ce260 | 2615 | memset(nlmsg_data(nlh) + len, 0, NLMSG_ALIGN(size) - size); |
a46621a3 DV |
2616 | return nlh; |
2617 | } | |
2618 | EXPORT_SYMBOL(__nlmsg_put); | |
2619 | ||
1da177e4 LT |
2620 | /* |
2621 | * It looks a bit ugly. | |
2622 | * It would be better to create kernel thread. | |
2623 | */ | |
2624 | ||
2625 | static int netlink_dump(struct sock *sk) | |
2626 | { | |
2627 | struct netlink_sock *nlk = nlk_sk(sk); | |
2628 | struct netlink_callback *cb; | |
c7ac8679 | 2629 | struct sk_buff *skb = NULL; |
1da177e4 | 2630 | struct nlmsghdr *nlh; |
bf8b79e4 | 2631 | int len, err = -ENOBUFS; |
c7ac8679 | 2632 | int alloc_size; |
1da177e4 | 2633 | |
af65bdfc | 2634 | mutex_lock(nlk->cb_mutex); |
16b304f3 | 2635 | if (!nlk->cb_running) { |
bf8b79e4 TG |
2636 | err = -EINVAL; |
2637 | goto errout_skb; | |
1da177e4 LT |
2638 | } |
2639 | ||
16b304f3 | 2640 | cb = &nlk->cb; |
c7ac8679 GR |
2641 | alloc_size = max_t(int, cb->min_dump_alloc, NLMSG_GOODSIZE); |
2642 | ||
f9c22888 PM |
2643 | if (!netlink_rx_is_mmaped(sk) && |
2644 | atomic_read(&sk->sk_rmem_alloc) >= sk->sk_rcvbuf) | |
2645 | goto errout_skb; | |
9063e21f ED |
2646 | |
2647 | /* NLMSG_GOODSIZE is small to avoid high order allocations being | |
2648 | * required, but it makes sense to _attempt_ a 16K bytes allocation | |
2649 | * to reduce number of system calls on dump operations, if user | |
2650 | * ever provided a big enough buffer. | |
2651 | */ | |
2652 | if (alloc_size < nlk->max_recvmsg_len) { | |
2653 | skb = netlink_alloc_skb(sk, | |
2654 | nlk->max_recvmsg_len, | |
2655 | nlk->portid, | |
2656 | GFP_KERNEL | | |
2657 | __GFP_NOWARN | | |
2658 | __GFP_NORETRY); | |
2659 | /* available room should be exact amount to avoid MSG_TRUNC */ | |
2660 | if (skb) | |
2661 | skb_reserve(skb, skb_tailroom(skb) - | |
2662 | nlk->max_recvmsg_len); | |
2663 | } | |
2664 | if (!skb) | |
2665 | skb = netlink_alloc_skb(sk, alloc_size, nlk->portid, | |
2666 | GFP_KERNEL); | |
c7ac8679 | 2667 | if (!skb) |
c63d6ea3 | 2668 | goto errout_skb; |
f9c22888 | 2669 | netlink_skb_set_owner_r(skb, sk); |
c7ac8679 | 2670 | |
1da177e4 LT |
2671 | len = cb->dump(skb, cb); |
2672 | ||
2673 | if (len > 0) { | |
af65bdfc | 2674 | mutex_unlock(nlk->cb_mutex); |
b1153f29 SH |
2675 | |
2676 | if (sk_filter(sk, skb)) | |
2677 | kfree_skb(skb); | |
4a7e7c2a ED |
2678 | else |
2679 | __netlink_sendskb(sk, skb); | |
1da177e4 LT |
2680 | return 0; |
2681 | } | |
2682 | ||
bf8b79e4 TG |
2683 | nlh = nlmsg_put_answer(skb, cb, NLMSG_DONE, sizeof(len), NLM_F_MULTI); |
2684 | if (!nlh) | |
2685 | goto errout_skb; | |
2686 | ||
670dc283 JB |
2687 | nl_dump_check_consistent(cb, nlh); |
2688 | ||
bf8b79e4 TG |
2689 | memcpy(nlmsg_data(nlh), &len, sizeof(len)); |
2690 | ||
b1153f29 SH |
2691 | if (sk_filter(sk, skb)) |
2692 | kfree_skb(skb); | |
4a7e7c2a ED |
2693 | else |
2694 | __netlink_sendskb(sk, skb); | |
1da177e4 | 2695 | |
a8f74b22 TG |
2696 | if (cb->done) |
2697 | cb->done(cb); | |
1da177e4 | 2698 | |
16b304f3 PS |
2699 | nlk->cb_running = false; |
2700 | mutex_unlock(nlk->cb_mutex); | |
6dc878a8 | 2701 | module_put(cb->module); |
16b304f3 | 2702 | consume_skb(cb->skb); |
1da177e4 | 2703 | return 0; |
1797754e | 2704 | |
bf8b79e4 | 2705 | errout_skb: |
af65bdfc | 2706 | mutex_unlock(nlk->cb_mutex); |
bf8b79e4 | 2707 | kfree_skb(skb); |
bf8b79e4 | 2708 | return err; |
1da177e4 LT |
2709 | } |
2710 | ||
6dc878a8 G |
2711 | int __netlink_dump_start(struct sock *ssk, struct sk_buff *skb, |
2712 | const struct nlmsghdr *nlh, | |
2713 | struct netlink_dump_control *control) | |
1da177e4 LT |
2714 | { |
2715 | struct netlink_callback *cb; | |
2716 | struct sock *sk; | |
2717 | struct netlink_sock *nlk; | |
b44d211e | 2718 | int ret; |
1da177e4 | 2719 | |
f9c22888 PM |
2720 | /* Memory mapped dump requests need to be copied to avoid looping |
2721 | * on the pending state in netlink_mmap_sendmsg() while the CB hold | |
2722 | * a reference to the skb. | |
2723 | */ | |
2724 | if (netlink_skb_is_mmaped(skb)) { | |
2725 | skb = skb_copy(skb, GFP_KERNEL); | |
16b304f3 | 2726 | if (skb == NULL) |
f9c22888 | 2727 | return -ENOBUFS; |
f9c22888 PM |
2728 | } else |
2729 | atomic_inc(&skb->users); | |
2730 | ||
15e47304 | 2731 | sk = netlink_lookup(sock_net(ssk), ssk->sk_protocol, NETLINK_CB(skb).portid); |
1da177e4 | 2732 | if (sk == NULL) { |
16b304f3 PS |
2733 | ret = -ECONNREFUSED; |
2734 | goto error_free; | |
1da177e4 | 2735 | } |
6dc878a8 | 2736 | |
16b304f3 | 2737 | nlk = nlk_sk(sk); |
af65bdfc | 2738 | mutex_lock(nlk->cb_mutex); |
6dc878a8 | 2739 | /* A dump is in progress... */ |
16b304f3 | 2740 | if (nlk->cb_running) { |
6dc878a8 | 2741 | ret = -EBUSY; |
16b304f3 | 2742 | goto error_unlock; |
1da177e4 | 2743 | } |
6dc878a8 | 2744 | /* add reference of module which cb->dump belongs to */ |
16b304f3 | 2745 | if (!try_module_get(control->module)) { |
6dc878a8 | 2746 | ret = -EPROTONOSUPPORT; |
16b304f3 | 2747 | goto error_unlock; |
6dc878a8 G |
2748 | } |
2749 | ||
16b304f3 PS |
2750 | cb = &nlk->cb; |
2751 | memset(cb, 0, sizeof(*cb)); | |
2752 | cb->dump = control->dump; | |
2753 | cb->done = control->done; | |
2754 | cb->nlh = nlh; | |
2755 | cb->data = control->data; | |
2756 | cb->module = control->module; | |
2757 | cb->min_dump_alloc = control->min_dump_alloc; | |
2758 | cb->skb = skb; | |
2759 | ||
2760 | nlk->cb_running = true; | |
2761 | ||
af65bdfc | 2762 | mutex_unlock(nlk->cb_mutex); |
1da177e4 | 2763 | |
b44d211e | 2764 | ret = netlink_dump(sk); |
1da177e4 | 2765 | sock_put(sk); |
5c58298c | 2766 | |
b44d211e AV |
2767 | if (ret) |
2768 | return ret; | |
2769 | ||
5c58298c DL |
2770 | /* We successfully started a dump, by returning -EINTR we |
2771 | * signal not to send ACK even if it was requested. | |
2772 | */ | |
2773 | return -EINTR; | |
16b304f3 PS |
2774 | |
2775 | error_unlock: | |
2776 | sock_put(sk); | |
2777 | mutex_unlock(nlk->cb_mutex); | |
2778 | error_free: | |
2779 | kfree_skb(skb); | |
2780 | return ret; | |
1da177e4 | 2781 | } |
6dc878a8 | 2782 | EXPORT_SYMBOL(__netlink_dump_start); |
1da177e4 LT |
2783 | |
2784 | void netlink_ack(struct sk_buff *in_skb, struct nlmsghdr *nlh, int err) | |
2785 | { | |
2786 | struct sk_buff *skb; | |
2787 | struct nlmsghdr *rep; | |
2788 | struct nlmsgerr *errmsg; | |
339bf98f | 2789 | size_t payload = sizeof(*errmsg); |
1da177e4 | 2790 | |
339bf98f TG |
2791 | /* error messages get the original request appened */ |
2792 | if (err) | |
2793 | payload += nlmsg_len(nlh); | |
1da177e4 | 2794 | |
f9c22888 PM |
2795 | skb = netlink_alloc_skb(in_skb->sk, nlmsg_total_size(payload), |
2796 | NETLINK_CB(in_skb).portid, GFP_KERNEL); | |
1da177e4 LT |
2797 | if (!skb) { |
2798 | struct sock *sk; | |
2799 | ||
3b1e0a65 | 2800 | sk = netlink_lookup(sock_net(in_skb->sk), |
b4b51029 | 2801 | in_skb->sk->sk_protocol, |
15e47304 | 2802 | NETLINK_CB(in_skb).portid); |
1da177e4 LT |
2803 | if (sk) { |
2804 | sk->sk_err = ENOBUFS; | |
2805 | sk->sk_error_report(sk); | |
2806 | sock_put(sk); | |
2807 | } | |
2808 | return; | |
2809 | } | |
2810 | ||
15e47304 | 2811 | rep = __nlmsg_put(skb, NETLINK_CB(in_skb).portid, nlh->nlmsg_seq, |
5dba93ae | 2812 | NLMSG_ERROR, payload, 0); |
bf8b79e4 | 2813 | errmsg = nlmsg_data(rep); |
1da177e4 | 2814 | errmsg->error = err; |
bf8b79e4 | 2815 | memcpy(&errmsg->msg, nlh, err ? nlh->nlmsg_len : sizeof(*nlh)); |
15e47304 | 2816 | netlink_unicast(in_skb->sk, skb, NETLINK_CB(in_skb).portid, MSG_DONTWAIT); |
1da177e4 | 2817 | } |
6ac552fd | 2818 | EXPORT_SYMBOL(netlink_ack); |
1da177e4 | 2819 | |
cd40b7d3 | 2820 | int netlink_rcv_skb(struct sk_buff *skb, int (*cb)(struct sk_buff *, |
1d00a4eb | 2821 | struct nlmsghdr *)) |
82ace47a | 2822 | { |
82ace47a TG |
2823 | struct nlmsghdr *nlh; |
2824 | int err; | |
2825 | ||
2826 | while (skb->len >= nlmsg_total_size(0)) { | |
cd40b7d3 DL |
2827 | int msglen; |
2828 | ||
b529ccf2 | 2829 | nlh = nlmsg_hdr(skb); |
d35b6856 | 2830 | err = 0; |
82ace47a | 2831 | |
ad8e4b75 | 2832 | if (nlh->nlmsg_len < NLMSG_HDRLEN || skb->len < nlh->nlmsg_len) |
82ace47a TG |
2833 | return 0; |
2834 | ||
d35b6856 TG |
2835 | /* Only requests are handled by the kernel */ |
2836 | if (!(nlh->nlmsg_flags & NLM_F_REQUEST)) | |
5c58298c | 2837 | goto ack; |
45e7ae7f TG |
2838 | |
2839 | /* Skip control messages */ | |
2840 | if (nlh->nlmsg_type < NLMSG_MIN_TYPE) | |
5c58298c | 2841 | goto ack; |
d35b6856 | 2842 | |
1d00a4eb | 2843 | err = cb(skb, nlh); |
5c58298c DL |
2844 | if (err == -EINTR) |
2845 | goto skip; | |
2846 | ||
2847 | ack: | |
d35b6856 | 2848 | if (nlh->nlmsg_flags & NLM_F_ACK || err) |
82ace47a | 2849 | netlink_ack(skb, nlh, err); |
82ace47a | 2850 | |
5c58298c | 2851 | skip: |
6ac552fd | 2852 | msglen = NLMSG_ALIGN(nlh->nlmsg_len); |
cd40b7d3 DL |
2853 | if (msglen > skb->len) |
2854 | msglen = skb->len; | |
2855 | skb_pull(skb, msglen); | |
82ace47a TG |
2856 | } |
2857 | ||
2858 | return 0; | |
2859 | } | |
6ac552fd | 2860 | EXPORT_SYMBOL(netlink_rcv_skb); |
82ace47a | 2861 | |
d387f6ad TG |
2862 | /** |
2863 | * nlmsg_notify - send a notification netlink message | |
2864 | * @sk: netlink socket to use | |
2865 | * @skb: notification message | |
15e47304 | 2866 | * @portid: destination netlink portid for reports or 0 |
d387f6ad TG |
2867 | * @group: destination multicast group or 0 |
2868 | * @report: 1 to report back, 0 to disable | |
2869 | * @flags: allocation flags | |
2870 | */ | |
15e47304 | 2871 | int nlmsg_notify(struct sock *sk, struct sk_buff *skb, u32 portid, |
d387f6ad TG |
2872 | unsigned int group, int report, gfp_t flags) |
2873 | { | |
2874 | int err = 0; | |
2875 | ||
2876 | if (group) { | |
15e47304 | 2877 | int exclude_portid = 0; |
d387f6ad TG |
2878 | |
2879 | if (report) { | |
2880 | atomic_inc(&skb->users); | |
15e47304 | 2881 | exclude_portid = portid; |
d387f6ad TG |
2882 | } |
2883 | ||
1ce85fe4 PNA |
2884 | /* errors reported via destination sk->sk_err, but propagate |
2885 | * delivery errors if NETLINK_BROADCAST_ERROR flag is set */ | |
15e47304 | 2886 | err = nlmsg_multicast(sk, skb, exclude_portid, group, flags); |
d387f6ad TG |
2887 | } |
2888 | ||
1ce85fe4 PNA |
2889 | if (report) { |
2890 | int err2; | |
2891 | ||
15e47304 | 2892 | err2 = nlmsg_unicast(sk, skb, portid); |
1ce85fe4 PNA |
2893 | if (!err || err == -ESRCH) |
2894 | err = err2; | |
2895 | } | |
d387f6ad TG |
2896 | |
2897 | return err; | |
2898 | } | |
6ac552fd | 2899 | EXPORT_SYMBOL(nlmsg_notify); |
d387f6ad | 2900 | |
1da177e4 LT |
2901 | #ifdef CONFIG_PROC_FS |
2902 | struct nl_seq_iter { | |
e372c414 | 2903 | struct seq_net_private p; |
56d28b1e | 2904 | struct rhashtable_iter hti; |
1da177e4 | 2905 | int link; |
1da177e4 LT |
2906 | }; |
2907 | ||
56d28b1e | 2908 | static int netlink_walk_start(struct nl_seq_iter *iter) |
1da177e4 | 2909 | { |
56d28b1e | 2910 | int err; |
1da177e4 | 2911 | |
56d28b1e HX |
2912 | err = rhashtable_walk_init(&nl_table[iter->link].hash, &iter->hti); |
2913 | if (err) { | |
2914 | iter->link = MAX_LINKS; | |
2915 | return err; | |
1da177e4 | 2916 | } |
56d28b1e HX |
2917 | |
2918 | err = rhashtable_walk_start(&iter->hti); | |
2919 | return err == -EAGAIN ? 0 : err; | |
1da177e4 LT |
2920 | } |
2921 | ||
56d28b1e | 2922 | static void netlink_walk_stop(struct nl_seq_iter *iter) |
1da177e4 | 2923 | { |
56d28b1e HX |
2924 | rhashtable_walk_stop(&iter->hti); |
2925 | rhashtable_walk_exit(&iter->hti); | |
1da177e4 LT |
2926 | } |
2927 | ||
56d28b1e | 2928 | static void *__netlink_seq_next(struct seq_file *seq) |
1da177e4 | 2929 | { |
56d28b1e | 2930 | struct nl_seq_iter *iter = seq->private; |
e341694e | 2931 | struct netlink_sock *nlk; |
1da177e4 | 2932 | |
56d28b1e HX |
2933 | do { |
2934 | for (;;) { | |
2935 | int err; | |
1da177e4 | 2936 | |
56d28b1e | 2937 | nlk = rhashtable_walk_next(&iter->hti); |
746fac4d | 2938 | |
56d28b1e HX |
2939 | if (IS_ERR(nlk)) { |
2940 | if (PTR_ERR(nlk) == -EAGAIN) | |
2941 | continue; | |
e341694e | 2942 | |
56d28b1e HX |
2943 | return nlk; |
2944 | } | |
1da177e4 | 2945 | |
56d28b1e HX |
2946 | if (nlk) |
2947 | break; | |
1da177e4 | 2948 | |
56d28b1e HX |
2949 | netlink_walk_stop(iter); |
2950 | if (++iter->link >= MAX_LINKS) | |
2951 | return NULL; | |
da12c90e | 2952 | |
56d28b1e HX |
2953 | err = netlink_walk_start(iter); |
2954 | if (err) | |
2955 | return ERR_PTR(err); | |
1da177e4 | 2956 | } |
56d28b1e | 2957 | } while (sock_net(&nlk->sk) != seq_file_net(seq)); |
1da177e4 | 2958 | |
56d28b1e HX |
2959 | return nlk; |
2960 | } | |
1da177e4 | 2961 | |
56d28b1e HX |
2962 | static void *netlink_seq_start(struct seq_file *seq, loff_t *posp) |
2963 | { | |
2964 | struct nl_seq_iter *iter = seq->private; | |
2965 | void *obj = SEQ_START_TOKEN; | |
2966 | loff_t pos; | |
2967 | int err; | |
2968 | ||
2969 | iter->link = 0; | |
2970 | ||
2971 | err = netlink_walk_start(iter); | |
2972 | if (err) | |
2973 | return ERR_PTR(err); | |
2974 | ||
2975 | for (pos = *posp; pos && obj && !IS_ERR(obj); pos--) | |
2976 | obj = __netlink_seq_next(seq); | |
2977 | ||
2978 | return obj; | |
2979 | } | |
2980 | ||
2981 | static void *netlink_seq_next(struct seq_file *seq, void *v, loff_t *pos) | |
2982 | { | |
2983 | ++*pos; | |
2984 | return __netlink_seq_next(seq); | |
1da177e4 LT |
2985 | } |
2986 | ||
2987 | static void netlink_seq_stop(struct seq_file *seq, void *v) | |
2988 | { | |
56d28b1e HX |
2989 | struct nl_seq_iter *iter = seq->private; |
2990 | ||
2991 | if (iter->link >= MAX_LINKS) | |
2992 | return; | |
2993 | ||
2994 | netlink_walk_stop(iter); | |
1da177e4 LT |
2995 | } |
2996 | ||
2997 | ||
2998 | static int netlink_seq_show(struct seq_file *seq, void *v) | |
2999 | { | |
658cb354 | 3000 | if (v == SEQ_START_TOKEN) { |
1da177e4 LT |
3001 | seq_puts(seq, |
3002 | "sk Eth Pid Groups " | |
cf0aa4e0 | 3003 | "Rmem Wmem Dump Locks Drops Inode\n"); |
658cb354 | 3004 | } else { |
1da177e4 LT |
3005 | struct sock *s = v; |
3006 | struct netlink_sock *nlk = nlk_sk(s); | |
3007 | ||
16b304f3 | 3008 | seq_printf(seq, "%pK %-3d %-6u %08x %-8d %-8d %d %-8d %-8d %-8lu\n", |
1da177e4 LT |
3009 | s, |
3010 | s->sk_protocol, | |
15e47304 | 3011 | nlk->portid, |
513c2500 | 3012 | nlk->groups ? (u32)nlk->groups[0] : 0, |
31e6d363 ED |
3013 | sk_rmem_alloc_get(s), |
3014 | sk_wmem_alloc_get(s), | |
16b304f3 | 3015 | nlk->cb_running, |
38938bfe | 3016 | atomic_read(&s->sk_refcnt), |
cf0aa4e0 MY |
3017 | atomic_read(&s->sk_drops), |
3018 | sock_i_ino(s) | |
1da177e4 LT |
3019 | ); |
3020 | ||
3021 | } | |
3022 | return 0; | |
3023 | } | |
3024 | ||
56b3d975 | 3025 | static const struct seq_operations netlink_seq_ops = { |
1da177e4 LT |
3026 | .start = netlink_seq_start, |
3027 | .next = netlink_seq_next, | |
3028 | .stop = netlink_seq_stop, | |
3029 | .show = netlink_seq_show, | |
3030 | }; | |
3031 | ||
3032 | ||
3033 | static int netlink_seq_open(struct inode *inode, struct file *file) | |
3034 | { | |
e372c414 DL |
3035 | return seq_open_net(inode, file, &netlink_seq_ops, |
3036 | sizeof(struct nl_seq_iter)); | |
b4b51029 EB |
3037 | } |
3038 | ||
da7071d7 | 3039 | static const struct file_operations netlink_seq_fops = { |
1da177e4 LT |
3040 | .owner = THIS_MODULE, |
3041 | .open = netlink_seq_open, | |
3042 | .read = seq_read, | |
3043 | .llseek = seq_lseek, | |
e372c414 | 3044 | .release = seq_release_net, |
1da177e4 LT |
3045 | }; |
3046 | ||
3047 | #endif | |
3048 | ||
3049 | int netlink_register_notifier(struct notifier_block *nb) | |
3050 | { | |
e041c683 | 3051 | return atomic_notifier_chain_register(&netlink_chain, nb); |
1da177e4 | 3052 | } |
6ac552fd | 3053 | EXPORT_SYMBOL(netlink_register_notifier); |
1da177e4 LT |
3054 | |
3055 | int netlink_unregister_notifier(struct notifier_block *nb) | |
3056 | { | |
e041c683 | 3057 | return atomic_notifier_chain_unregister(&netlink_chain, nb); |
1da177e4 | 3058 | } |
6ac552fd | 3059 | EXPORT_SYMBOL(netlink_unregister_notifier); |
746fac4d | 3060 | |
90ddc4f0 | 3061 | static const struct proto_ops netlink_ops = { |
1da177e4 LT |
3062 | .family = PF_NETLINK, |
3063 | .owner = THIS_MODULE, | |
3064 | .release = netlink_release, | |
3065 | .bind = netlink_bind, | |
3066 | .connect = netlink_connect, | |
3067 | .socketpair = sock_no_socketpair, | |
3068 | .accept = sock_no_accept, | |
3069 | .getname = netlink_getname, | |
9652e931 | 3070 | .poll = netlink_poll, |
1da177e4 LT |
3071 | .ioctl = sock_no_ioctl, |
3072 | .listen = sock_no_listen, | |
3073 | .shutdown = sock_no_shutdown, | |
9a4595bc PM |
3074 | .setsockopt = netlink_setsockopt, |
3075 | .getsockopt = netlink_getsockopt, | |
1da177e4 LT |
3076 | .sendmsg = netlink_sendmsg, |
3077 | .recvmsg = netlink_recvmsg, | |
ccdfcc39 | 3078 | .mmap = netlink_mmap, |
1da177e4 LT |
3079 | .sendpage = sock_no_sendpage, |
3080 | }; | |
3081 | ||
ec1b4cf7 | 3082 | static const struct net_proto_family netlink_family_ops = { |
1da177e4 LT |
3083 | .family = PF_NETLINK, |
3084 | .create = netlink_create, | |
3085 | .owner = THIS_MODULE, /* for consistency 8) */ | |
3086 | }; | |
3087 | ||
4665079c | 3088 | static int __net_init netlink_net_init(struct net *net) |
b4b51029 EB |
3089 | { |
3090 | #ifdef CONFIG_PROC_FS | |
d4beaa66 | 3091 | if (!proc_create("netlink", 0, net->proc_net, &netlink_seq_fops)) |
b4b51029 EB |
3092 | return -ENOMEM; |
3093 | #endif | |
3094 | return 0; | |
3095 | } | |
3096 | ||
4665079c | 3097 | static void __net_exit netlink_net_exit(struct net *net) |
b4b51029 EB |
3098 | { |
3099 | #ifdef CONFIG_PROC_FS | |
ece31ffd | 3100 | remove_proc_entry("netlink", net->proc_net); |
b4b51029 EB |
3101 | #endif |
3102 | } | |
3103 | ||
b963ea89 DM |
3104 | static void __init netlink_add_usersock_entry(void) |
3105 | { | |
5c398dc8 | 3106 | struct listeners *listeners; |
b963ea89 DM |
3107 | int groups = 32; |
3108 | ||
5c398dc8 | 3109 | listeners = kzalloc(sizeof(*listeners) + NLGRPSZ(groups), GFP_KERNEL); |
b963ea89 | 3110 | if (!listeners) |
5c398dc8 | 3111 | panic("netlink_add_usersock_entry: Cannot allocate listeners\n"); |
b963ea89 DM |
3112 | |
3113 | netlink_table_grab(); | |
3114 | ||
3115 | nl_table[NETLINK_USERSOCK].groups = groups; | |
5c398dc8 | 3116 | rcu_assign_pointer(nl_table[NETLINK_USERSOCK].listeners, listeners); |
b963ea89 DM |
3117 | nl_table[NETLINK_USERSOCK].module = THIS_MODULE; |
3118 | nl_table[NETLINK_USERSOCK].registered = 1; | |
9785e10a | 3119 | nl_table[NETLINK_USERSOCK].flags = NL_CFG_F_NONROOT_SEND; |
b963ea89 DM |
3120 | |
3121 | netlink_table_ungrab(); | |
3122 | } | |
3123 | ||
022cbae6 | 3124 | static struct pernet_operations __net_initdata netlink_net_ops = { |
b4b51029 EB |
3125 | .init = netlink_net_init, |
3126 | .exit = netlink_net_exit, | |
3127 | }; | |
3128 | ||
c428ecd1 HX |
3129 | static inline u32 netlink_hash(const void *data, u32 seed) |
3130 | { | |
3131 | const struct netlink_sock *nlk = data; | |
3132 | struct netlink_compare_arg arg; | |
3133 | ||
3134 | netlink_compare_arg_init(&arg, sock_net(&nlk->sk), nlk->portid); | |
3135 | return jhash(&arg, netlink_compare_arg_len, seed); | |
3136 | } | |
3137 | ||
3138 | static const struct rhashtable_params netlink_rhashtable_params = { | |
3139 | .head_offset = offsetof(struct netlink_sock, node), | |
3140 | .key_len = netlink_compare_arg_len, | |
3141 | .hashfn = jhash, | |
3142 | .obj_hashfn = netlink_hash, | |
3143 | .obj_cmpfn = netlink_compare, | |
3144 | .max_size = 65536, | |
3145 | }; | |
3146 | ||
1da177e4 LT |
3147 | static int __init netlink_proto_init(void) |
3148 | { | |
1da177e4 | 3149 | int i; |
1da177e4 LT |
3150 | int err = proto_register(&netlink_proto, 0); |
3151 | ||
3152 | if (err != 0) | |
3153 | goto out; | |
3154 | ||
fab25745 | 3155 | BUILD_BUG_ON(sizeof(struct netlink_skb_parms) > FIELD_SIZEOF(struct sk_buff, cb)); |
1da177e4 | 3156 | |
0da974f4 | 3157 | nl_table = kcalloc(MAX_LINKS, sizeof(*nl_table), GFP_KERNEL); |
fab2caf6 AM |
3158 | if (!nl_table) |
3159 | goto panic; | |
1da177e4 | 3160 | |
1da177e4 | 3161 | for (i = 0; i < MAX_LINKS; i++) { |
c428ecd1 HX |
3162 | if (rhashtable_init(&nl_table[i].hash, |
3163 | &netlink_rhashtable_params) < 0) { | |
e341694e TG |
3164 | while (--i > 0) |
3165 | rhashtable_destroy(&nl_table[i].hash); | |
1da177e4 | 3166 | kfree(nl_table); |
fab2caf6 | 3167 | goto panic; |
1da177e4 | 3168 | } |
1da177e4 LT |
3169 | } |
3170 | ||
bcbde0d4 DB |
3171 | INIT_LIST_HEAD(&netlink_tap_all); |
3172 | ||
b963ea89 DM |
3173 | netlink_add_usersock_entry(); |
3174 | ||
1da177e4 | 3175 | sock_register(&netlink_family_ops); |
b4b51029 | 3176 | register_pernet_subsys(&netlink_net_ops); |
746fac4d | 3177 | /* The netlink device handler may be needed early. */ |
1da177e4 LT |
3178 | rtnetlink_init(); |
3179 | out: | |
3180 | return err; | |
fab2caf6 AM |
3181 | panic: |
3182 | panic("netlink_init: Cannot allocate nl_table\n"); | |
1da177e4 LT |
3183 | } |
3184 | ||
1da177e4 | 3185 | core_initcall(netlink_proto_init); |