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> |
54e0f520 | 64 | |
457c4cbc | 65 | #include <net/net_namespace.h> |
1da177e4 LT |
66 | #include <net/sock.h> |
67 | #include <net/scm.h> | |
82ace47a | 68 | #include <net/netlink.h> |
1da177e4 | 69 | |
0f29c768 | 70 | #include "af_netlink.h" |
1da177e4 | 71 | |
5c398dc8 ED |
72 | struct listeners { |
73 | struct rcu_head rcu; | |
74 | unsigned long masks[0]; | |
6c04bb18 JB |
75 | }; |
76 | ||
cd967e05 PM |
77 | /* state bits */ |
78 | #define NETLINK_CONGESTED 0x0 | |
79 | ||
80 | /* flags */ | |
77247bbb | 81 | #define NETLINK_KERNEL_SOCKET 0x1 |
9a4595bc | 82 | #define NETLINK_RECV_PKTINFO 0x2 |
be0c22a4 | 83 | #define NETLINK_BROADCAST_SEND_ERROR 0x4 |
38938bfe | 84 | #define NETLINK_RECV_NO_ENOBUFS 0x8 |
77247bbb | 85 | |
035c4c16 | 86 | static inline int netlink_is_kernel(struct sock *sk) |
aed81560 DL |
87 | { |
88 | return nlk_sk(sk)->flags & NETLINK_KERNEL_SOCKET; | |
89 | } | |
90 | ||
0f29c768 AV |
91 | struct netlink_table *nl_table; |
92 | EXPORT_SYMBOL_GPL(nl_table); | |
1da177e4 LT |
93 | |
94 | static DECLARE_WAIT_QUEUE_HEAD(nl_table_wait); | |
95 | ||
96 | static int netlink_dump(struct sock *sk); | |
9652e931 | 97 | static void netlink_skb_destructor(struct sk_buff *skb); |
1da177e4 | 98 | |
78fd1d0a | 99 | /* nl_table locking explained: |
21e4902a | 100 | * Lookup and traversal are protected with an RCU read-side lock. Insertion |
c5adde94 | 101 | * and removal are protected with per bucket lock while using RCU list |
21e4902a TG |
102 | * modification primitives and may run in parallel to RCU protected lookups. |
103 | * Destruction of the Netlink socket may only occur *after* nl_table_lock has | |
104 | * been acquired * either during or after the socket has been removed from | |
105 | * the list and after an RCU grace period. | |
78fd1d0a | 106 | */ |
0f29c768 AV |
107 | DEFINE_RWLOCK(nl_table_lock); |
108 | EXPORT_SYMBOL_GPL(nl_table_lock); | |
1da177e4 LT |
109 | static atomic_t nl_table_users = ATOMIC_INIT(0); |
110 | ||
6d772ac5 ED |
111 | #define nl_deref_protected(X) rcu_dereference_protected(X, lockdep_is_held(&nl_table_lock)); |
112 | ||
e041c683 | 113 | static ATOMIC_NOTIFIER_HEAD(netlink_chain); |
1da177e4 | 114 | |
bcbde0d4 DB |
115 | static DEFINE_SPINLOCK(netlink_tap_lock); |
116 | static struct list_head netlink_tap_all __read_mostly; | |
117 | ||
b57ef81f | 118 | static inline u32 netlink_group_mask(u32 group) |
d629b836 PM |
119 | { |
120 | return group ? 1 << (group - 1) : 0; | |
121 | } | |
122 | ||
bcbde0d4 DB |
123 | int netlink_add_tap(struct netlink_tap *nt) |
124 | { | |
125 | if (unlikely(nt->dev->type != ARPHRD_NETLINK)) | |
126 | return -EINVAL; | |
127 | ||
128 | spin_lock(&netlink_tap_lock); | |
129 | list_add_rcu(&nt->list, &netlink_tap_all); | |
130 | spin_unlock(&netlink_tap_lock); | |
131 | ||
fcd4d35e | 132 | __module_get(nt->module); |
bcbde0d4 DB |
133 | |
134 | return 0; | |
135 | } | |
136 | EXPORT_SYMBOL_GPL(netlink_add_tap); | |
137 | ||
2173f8d9 | 138 | static int __netlink_remove_tap(struct netlink_tap *nt) |
bcbde0d4 DB |
139 | { |
140 | bool found = false; | |
141 | struct netlink_tap *tmp; | |
142 | ||
143 | spin_lock(&netlink_tap_lock); | |
144 | ||
145 | list_for_each_entry(tmp, &netlink_tap_all, list) { | |
146 | if (nt == tmp) { | |
147 | list_del_rcu(&nt->list); | |
148 | found = true; | |
149 | goto out; | |
150 | } | |
151 | } | |
152 | ||
153 | pr_warn("__netlink_remove_tap: %p not found\n", nt); | |
154 | out: | |
155 | spin_unlock(&netlink_tap_lock); | |
156 | ||
157 | if (found && nt->module) | |
158 | module_put(nt->module); | |
159 | ||
160 | return found ? 0 : -ENODEV; | |
161 | } | |
bcbde0d4 DB |
162 | |
163 | int netlink_remove_tap(struct netlink_tap *nt) | |
164 | { | |
165 | int ret; | |
166 | ||
167 | ret = __netlink_remove_tap(nt); | |
168 | synchronize_net(); | |
169 | ||
170 | return ret; | |
171 | } | |
172 | EXPORT_SYMBOL_GPL(netlink_remove_tap); | |
173 | ||
5ffd5cdd DB |
174 | static bool netlink_filter_tap(const struct sk_buff *skb) |
175 | { | |
176 | struct sock *sk = skb->sk; | |
5ffd5cdd DB |
177 | |
178 | /* We take the more conservative approach and | |
179 | * whitelist socket protocols that may pass. | |
180 | */ | |
181 | switch (sk->sk_protocol) { | |
182 | case NETLINK_ROUTE: | |
183 | case NETLINK_USERSOCK: | |
184 | case NETLINK_SOCK_DIAG: | |
185 | case NETLINK_NFLOG: | |
186 | case NETLINK_XFRM: | |
187 | case NETLINK_FIB_LOOKUP: | |
188 | case NETLINK_NETFILTER: | |
189 | case NETLINK_GENERIC: | |
498044bb | 190 | return true; |
5ffd5cdd DB |
191 | } |
192 | ||
498044bb | 193 | return false; |
5ffd5cdd DB |
194 | } |
195 | ||
bcbde0d4 DB |
196 | static int __netlink_deliver_tap_skb(struct sk_buff *skb, |
197 | struct net_device *dev) | |
198 | { | |
199 | struct sk_buff *nskb; | |
5ffd5cdd | 200 | struct sock *sk = skb->sk; |
bcbde0d4 DB |
201 | int ret = -ENOMEM; |
202 | ||
203 | dev_hold(dev); | |
204 | nskb = skb_clone(skb, GFP_ATOMIC); | |
205 | if (nskb) { | |
206 | nskb->dev = dev; | |
5ffd5cdd | 207 | nskb->protocol = htons((u16) sk->sk_protocol); |
604d13c9 DB |
208 | nskb->pkt_type = netlink_is_kernel(sk) ? |
209 | PACKET_KERNEL : PACKET_USER; | |
4e48ed88 | 210 | skb_reset_network_header(nskb); |
bcbde0d4 DB |
211 | ret = dev_queue_xmit(nskb); |
212 | if (unlikely(ret > 0)) | |
213 | ret = net_xmit_errno(ret); | |
214 | } | |
215 | ||
216 | dev_put(dev); | |
217 | return ret; | |
218 | } | |
219 | ||
220 | static void __netlink_deliver_tap(struct sk_buff *skb) | |
221 | { | |
222 | int ret; | |
223 | struct netlink_tap *tmp; | |
224 | ||
5ffd5cdd DB |
225 | if (!netlink_filter_tap(skb)) |
226 | return; | |
227 | ||
bcbde0d4 DB |
228 | list_for_each_entry_rcu(tmp, &netlink_tap_all, list) { |
229 | ret = __netlink_deliver_tap_skb(skb, tmp->dev); | |
230 | if (unlikely(ret)) | |
231 | break; | |
232 | } | |
233 | } | |
234 | ||
235 | static void netlink_deliver_tap(struct sk_buff *skb) | |
236 | { | |
237 | rcu_read_lock(); | |
238 | ||
239 | if (unlikely(!list_empty(&netlink_tap_all))) | |
240 | __netlink_deliver_tap(skb); | |
241 | ||
242 | rcu_read_unlock(); | |
243 | } | |
244 | ||
73bfd370 DB |
245 | static void netlink_deliver_tap_kernel(struct sock *dst, struct sock *src, |
246 | struct sk_buff *skb) | |
247 | { | |
248 | if (!(netlink_is_kernel(dst) && netlink_is_kernel(src))) | |
249 | netlink_deliver_tap(skb); | |
250 | } | |
251 | ||
cd1df525 PM |
252 | static void netlink_overrun(struct sock *sk) |
253 | { | |
254 | struct netlink_sock *nlk = nlk_sk(sk); | |
255 | ||
256 | if (!(nlk->flags & NETLINK_RECV_NO_ENOBUFS)) { | |
257 | if (!test_and_set_bit(NETLINK_CONGESTED, &nlk_sk(sk)->state)) { | |
258 | sk->sk_err = ENOBUFS; | |
259 | sk->sk_error_report(sk); | |
260 | } | |
261 | } | |
262 | atomic_inc(&sk->sk_drops); | |
263 | } | |
264 | ||
265 | static void netlink_rcv_wake(struct sock *sk) | |
266 | { | |
267 | struct netlink_sock *nlk = nlk_sk(sk); | |
268 | ||
269 | if (skb_queue_empty(&sk->sk_receive_queue)) | |
270 | clear_bit(NETLINK_CONGESTED, &nlk->state); | |
271 | if (!test_bit(NETLINK_CONGESTED, &nlk->state)) | |
272 | wake_up_interruptible(&nlk->wait); | |
273 | } | |
274 | ||
ccdfcc39 | 275 | #ifdef CONFIG_NETLINK_MMAP |
9652e931 PM |
276 | static bool netlink_skb_is_mmaped(const struct sk_buff *skb) |
277 | { | |
278 | return NETLINK_CB(skb).flags & NETLINK_SKB_MMAPED; | |
279 | } | |
280 | ||
f9c22888 PM |
281 | static bool netlink_rx_is_mmaped(struct sock *sk) |
282 | { | |
283 | return nlk_sk(sk)->rx_ring.pg_vec != NULL; | |
284 | } | |
285 | ||
5fd96123 PM |
286 | static bool netlink_tx_is_mmaped(struct sock *sk) |
287 | { | |
288 | return nlk_sk(sk)->tx_ring.pg_vec != NULL; | |
289 | } | |
290 | ||
ccdfcc39 PM |
291 | static __pure struct page *pgvec_to_page(const void *addr) |
292 | { | |
293 | if (is_vmalloc_addr(addr)) | |
294 | return vmalloc_to_page(addr); | |
295 | else | |
296 | return virt_to_page(addr); | |
297 | } | |
298 | ||
299 | static void free_pg_vec(void **pg_vec, unsigned int order, unsigned int len) | |
300 | { | |
301 | unsigned int i; | |
302 | ||
303 | for (i = 0; i < len; i++) { | |
304 | if (pg_vec[i] != NULL) { | |
305 | if (is_vmalloc_addr(pg_vec[i])) | |
306 | vfree(pg_vec[i]); | |
307 | else | |
308 | free_pages((unsigned long)pg_vec[i], order); | |
309 | } | |
310 | } | |
311 | kfree(pg_vec); | |
312 | } | |
313 | ||
314 | static void *alloc_one_pg_vec_page(unsigned long order) | |
315 | { | |
316 | void *buffer; | |
317 | gfp_t gfp_flags = GFP_KERNEL | __GFP_COMP | __GFP_ZERO | | |
318 | __GFP_NOWARN | __GFP_NORETRY; | |
319 | ||
320 | buffer = (void *)__get_free_pages(gfp_flags, order); | |
321 | if (buffer != NULL) | |
322 | return buffer; | |
323 | ||
324 | buffer = vzalloc((1 << order) * PAGE_SIZE); | |
325 | if (buffer != NULL) | |
326 | return buffer; | |
327 | ||
328 | gfp_flags &= ~__GFP_NORETRY; | |
329 | return (void *)__get_free_pages(gfp_flags, order); | |
330 | } | |
331 | ||
332 | static void **alloc_pg_vec(struct netlink_sock *nlk, | |
333 | struct nl_mmap_req *req, unsigned int order) | |
334 | { | |
335 | unsigned int block_nr = req->nm_block_nr; | |
336 | unsigned int i; | |
8a849bb7 | 337 | void **pg_vec; |
ccdfcc39 PM |
338 | |
339 | pg_vec = kcalloc(block_nr, sizeof(void *), GFP_KERNEL); | |
340 | if (pg_vec == NULL) | |
341 | return NULL; | |
342 | ||
343 | for (i = 0; i < block_nr; i++) { | |
8a849bb7 | 344 | pg_vec[i] = alloc_one_pg_vec_page(order); |
ccdfcc39 PM |
345 | if (pg_vec[i] == NULL) |
346 | goto err1; | |
347 | } | |
348 | ||
349 | return pg_vec; | |
350 | err1: | |
351 | free_pg_vec(pg_vec, order, block_nr); | |
352 | return NULL; | |
353 | } | |
354 | ||
355 | static int netlink_set_ring(struct sock *sk, struct nl_mmap_req *req, | |
356 | bool closing, bool tx_ring) | |
357 | { | |
358 | struct netlink_sock *nlk = nlk_sk(sk); | |
359 | struct netlink_ring *ring; | |
360 | struct sk_buff_head *queue; | |
361 | void **pg_vec = NULL; | |
362 | unsigned int order = 0; | |
363 | int err; | |
364 | ||
365 | ring = tx_ring ? &nlk->tx_ring : &nlk->rx_ring; | |
366 | queue = tx_ring ? &sk->sk_write_queue : &sk->sk_receive_queue; | |
367 | ||
368 | if (!closing) { | |
369 | if (atomic_read(&nlk->mapped)) | |
370 | return -EBUSY; | |
371 | if (atomic_read(&ring->pending)) | |
372 | return -EBUSY; | |
373 | } | |
374 | ||
375 | if (req->nm_block_nr) { | |
376 | if (ring->pg_vec != NULL) | |
377 | return -EBUSY; | |
378 | ||
379 | if ((int)req->nm_block_size <= 0) | |
380 | return -EINVAL; | |
74e83b23 | 381 | if (!PAGE_ALIGNED(req->nm_block_size)) |
ccdfcc39 PM |
382 | return -EINVAL; |
383 | if (req->nm_frame_size < NL_MMAP_HDRLEN) | |
384 | return -EINVAL; | |
385 | if (!IS_ALIGNED(req->nm_frame_size, NL_MMAP_MSG_ALIGNMENT)) | |
386 | return -EINVAL; | |
387 | ||
388 | ring->frames_per_block = req->nm_block_size / | |
389 | req->nm_frame_size; | |
390 | if (ring->frames_per_block == 0) | |
391 | return -EINVAL; | |
392 | if (ring->frames_per_block * req->nm_block_nr != | |
393 | req->nm_frame_nr) | |
394 | return -EINVAL; | |
395 | ||
396 | order = get_order(req->nm_block_size); | |
397 | pg_vec = alloc_pg_vec(nlk, req, order); | |
398 | if (pg_vec == NULL) | |
399 | return -ENOMEM; | |
400 | } else { | |
401 | if (req->nm_frame_nr) | |
402 | return -EINVAL; | |
403 | } | |
404 | ||
405 | err = -EBUSY; | |
406 | mutex_lock(&nlk->pg_vec_lock); | |
407 | if (closing || atomic_read(&nlk->mapped) == 0) { | |
408 | err = 0; | |
409 | spin_lock_bh(&queue->lock); | |
410 | ||
411 | ring->frame_max = req->nm_frame_nr - 1; | |
412 | ring->head = 0; | |
413 | ring->frame_size = req->nm_frame_size; | |
414 | ring->pg_vec_pages = req->nm_block_size / PAGE_SIZE; | |
415 | ||
416 | swap(ring->pg_vec_len, req->nm_block_nr); | |
417 | swap(ring->pg_vec_order, order); | |
418 | swap(ring->pg_vec, pg_vec); | |
419 | ||
420 | __skb_queue_purge(queue); | |
421 | spin_unlock_bh(&queue->lock); | |
422 | ||
423 | WARN_ON(atomic_read(&nlk->mapped)); | |
424 | } | |
425 | mutex_unlock(&nlk->pg_vec_lock); | |
426 | ||
427 | if (pg_vec) | |
428 | free_pg_vec(pg_vec, order, req->nm_block_nr); | |
429 | return err; | |
430 | } | |
431 | ||
432 | static void netlink_mm_open(struct vm_area_struct *vma) | |
433 | { | |
434 | struct file *file = vma->vm_file; | |
435 | struct socket *sock = file->private_data; | |
436 | struct sock *sk = sock->sk; | |
437 | ||
438 | if (sk) | |
439 | atomic_inc(&nlk_sk(sk)->mapped); | |
440 | } | |
441 | ||
442 | static void netlink_mm_close(struct vm_area_struct *vma) | |
443 | { | |
444 | struct file *file = vma->vm_file; | |
445 | struct socket *sock = file->private_data; | |
446 | struct sock *sk = sock->sk; | |
447 | ||
448 | if (sk) | |
449 | atomic_dec(&nlk_sk(sk)->mapped); | |
450 | } | |
451 | ||
452 | static const struct vm_operations_struct netlink_mmap_ops = { | |
453 | .open = netlink_mm_open, | |
454 | .close = netlink_mm_close, | |
455 | }; | |
456 | ||
457 | static int netlink_mmap(struct file *file, struct socket *sock, | |
458 | struct vm_area_struct *vma) | |
459 | { | |
460 | struct sock *sk = sock->sk; | |
461 | struct netlink_sock *nlk = nlk_sk(sk); | |
462 | struct netlink_ring *ring; | |
463 | unsigned long start, size, expected; | |
464 | unsigned int i; | |
465 | int err = -EINVAL; | |
466 | ||
467 | if (vma->vm_pgoff) | |
468 | return -EINVAL; | |
469 | ||
470 | mutex_lock(&nlk->pg_vec_lock); | |
471 | ||
472 | expected = 0; | |
473 | for (ring = &nlk->rx_ring; ring <= &nlk->tx_ring; ring++) { | |
474 | if (ring->pg_vec == NULL) | |
475 | continue; | |
476 | expected += ring->pg_vec_len * ring->pg_vec_pages * PAGE_SIZE; | |
477 | } | |
478 | ||
479 | if (expected == 0) | |
480 | goto out; | |
481 | ||
482 | size = vma->vm_end - vma->vm_start; | |
483 | if (size != expected) | |
484 | goto out; | |
485 | ||
486 | start = vma->vm_start; | |
487 | for (ring = &nlk->rx_ring; ring <= &nlk->tx_ring; ring++) { | |
488 | if (ring->pg_vec == NULL) | |
489 | continue; | |
490 | ||
491 | for (i = 0; i < ring->pg_vec_len; i++) { | |
492 | struct page *page; | |
493 | void *kaddr = ring->pg_vec[i]; | |
494 | unsigned int pg_num; | |
495 | ||
496 | for (pg_num = 0; pg_num < ring->pg_vec_pages; pg_num++) { | |
497 | page = pgvec_to_page(kaddr); | |
498 | err = vm_insert_page(vma, start, page); | |
499 | if (err < 0) | |
500 | goto out; | |
501 | start += PAGE_SIZE; | |
502 | kaddr += PAGE_SIZE; | |
503 | } | |
504 | } | |
505 | } | |
506 | ||
507 | atomic_inc(&nlk->mapped); | |
508 | vma->vm_ops = &netlink_mmap_ops; | |
509 | err = 0; | |
510 | out: | |
511 | mutex_unlock(&nlk->pg_vec_lock); | |
7cdbac71 | 512 | return err; |
ccdfcc39 | 513 | } |
9652e931 | 514 | |
4682a035 | 515 | static void netlink_frame_flush_dcache(const struct nl_mmap_hdr *hdr, unsigned int nm_len) |
9652e931 PM |
516 | { |
517 | #if ARCH_IMPLEMENTS_FLUSH_DCACHE_PAGE == 1 | |
518 | struct page *p_start, *p_end; | |
519 | ||
520 | /* First page is flushed through netlink_{get,set}_status */ | |
521 | p_start = pgvec_to_page(hdr + PAGE_SIZE); | |
4682a035 | 522 | p_end = pgvec_to_page((void *)hdr + NL_MMAP_HDRLEN + nm_len - 1); |
9652e931 PM |
523 | while (p_start <= p_end) { |
524 | flush_dcache_page(p_start); | |
525 | p_start++; | |
526 | } | |
527 | #endif | |
528 | } | |
529 | ||
530 | static enum nl_mmap_status netlink_get_status(const struct nl_mmap_hdr *hdr) | |
531 | { | |
532 | smp_rmb(); | |
533 | flush_dcache_page(pgvec_to_page(hdr)); | |
534 | return hdr->nm_status; | |
535 | } | |
536 | ||
537 | static void netlink_set_status(struct nl_mmap_hdr *hdr, | |
538 | enum nl_mmap_status status) | |
539 | { | |
a18e6a18 | 540 | smp_mb(); |
9652e931 PM |
541 | hdr->nm_status = status; |
542 | flush_dcache_page(pgvec_to_page(hdr)); | |
9652e931 PM |
543 | } |
544 | ||
545 | static struct nl_mmap_hdr * | |
546 | __netlink_lookup_frame(const struct netlink_ring *ring, unsigned int pos) | |
547 | { | |
548 | unsigned int pg_vec_pos, frame_off; | |
549 | ||
550 | pg_vec_pos = pos / ring->frames_per_block; | |
551 | frame_off = pos % ring->frames_per_block; | |
552 | ||
553 | return ring->pg_vec[pg_vec_pos] + (frame_off * ring->frame_size); | |
554 | } | |
555 | ||
556 | static struct nl_mmap_hdr * | |
557 | netlink_lookup_frame(const struct netlink_ring *ring, unsigned int pos, | |
558 | enum nl_mmap_status status) | |
559 | { | |
560 | struct nl_mmap_hdr *hdr; | |
561 | ||
562 | hdr = __netlink_lookup_frame(ring, pos); | |
563 | if (netlink_get_status(hdr) != status) | |
564 | return NULL; | |
565 | ||
566 | return hdr; | |
567 | } | |
568 | ||
569 | static struct nl_mmap_hdr * | |
570 | netlink_current_frame(const struct netlink_ring *ring, | |
571 | enum nl_mmap_status status) | |
572 | { | |
573 | return netlink_lookup_frame(ring, ring->head, status); | |
574 | } | |
575 | ||
576 | static struct nl_mmap_hdr * | |
577 | netlink_previous_frame(const struct netlink_ring *ring, | |
578 | enum nl_mmap_status status) | |
579 | { | |
580 | unsigned int prev; | |
581 | ||
582 | prev = ring->head ? ring->head - 1 : ring->frame_max; | |
583 | return netlink_lookup_frame(ring, prev, status); | |
584 | } | |
585 | ||
586 | static void netlink_increment_head(struct netlink_ring *ring) | |
587 | { | |
588 | ring->head = ring->head != ring->frame_max ? ring->head + 1 : 0; | |
589 | } | |
590 | ||
591 | static void netlink_forward_ring(struct netlink_ring *ring) | |
592 | { | |
593 | unsigned int head = ring->head, pos = head; | |
594 | const struct nl_mmap_hdr *hdr; | |
595 | ||
596 | do { | |
597 | hdr = __netlink_lookup_frame(ring, pos); | |
598 | if (hdr->nm_status == NL_MMAP_STATUS_UNUSED) | |
599 | break; | |
600 | if (hdr->nm_status != NL_MMAP_STATUS_SKIP) | |
601 | break; | |
602 | netlink_increment_head(ring); | |
603 | } while (ring->head != head); | |
604 | } | |
605 | ||
cd1df525 PM |
606 | static bool netlink_dump_space(struct netlink_sock *nlk) |
607 | { | |
608 | struct netlink_ring *ring = &nlk->rx_ring; | |
609 | struct nl_mmap_hdr *hdr; | |
610 | unsigned int n; | |
611 | ||
612 | hdr = netlink_current_frame(ring, NL_MMAP_STATUS_UNUSED); | |
613 | if (hdr == NULL) | |
614 | return false; | |
615 | ||
616 | n = ring->head + ring->frame_max / 2; | |
617 | if (n > ring->frame_max) | |
618 | n -= ring->frame_max; | |
619 | ||
620 | hdr = __netlink_lookup_frame(ring, n); | |
621 | ||
622 | return hdr->nm_status == NL_MMAP_STATUS_UNUSED; | |
623 | } | |
624 | ||
9652e931 PM |
625 | static unsigned int netlink_poll(struct file *file, struct socket *sock, |
626 | poll_table *wait) | |
627 | { | |
628 | struct sock *sk = sock->sk; | |
629 | struct netlink_sock *nlk = nlk_sk(sk); | |
630 | unsigned int mask; | |
cd1df525 | 631 | int err; |
9652e931 | 632 | |
cd1df525 PM |
633 | if (nlk->rx_ring.pg_vec != NULL) { |
634 | /* Memory mapped sockets don't call recvmsg(), so flow control | |
635 | * for dumps is performed here. A dump is allowed to continue | |
636 | * if at least half the ring is unused. | |
637 | */ | |
16b304f3 | 638 | while (nlk->cb_running && netlink_dump_space(nlk)) { |
cd1df525 PM |
639 | err = netlink_dump(sk); |
640 | if (err < 0) { | |
ac30ef83 | 641 | sk->sk_err = -err; |
cd1df525 PM |
642 | sk->sk_error_report(sk); |
643 | break; | |
644 | } | |
645 | } | |
646 | netlink_rcv_wake(sk); | |
647 | } | |
5fd96123 | 648 | |
9652e931 PM |
649 | mask = datagram_poll(file, sock, wait); |
650 | ||
651 | spin_lock_bh(&sk->sk_receive_queue.lock); | |
652 | if (nlk->rx_ring.pg_vec) { | |
653 | netlink_forward_ring(&nlk->rx_ring); | |
654 | if (!netlink_previous_frame(&nlk->rx_ring, NL_MMAP_STATUS_UNUSED)) | |
655 | mask |= POLLIN | POLLRDNORM; | |
656 | } | |
657 | spin_unlock_bh(&sk->sk_receive_queue.lock); | |
658 | ||
659 | spin_lock_bh(&sk->sk_write_queue.lock); | |
660 | if (nlk->tx_ring.pg_vec) { | |
661 | if (netlink_current_frame(&nlk->tx_ring, NL_MMAP_STATUS_UNUSED)) | |
662 | mask |= POLLOUT | POLLWRNORM; | |
663 | } | |
664 | spin_unlock_bh(&sk->sk_write_queue.lock); | |
665 | ||
666 | return mask; | |
667 | } | |
668 | ||
669 | static struct nl_mmap_hdr *netlink_mmap_hdr(struct sk_buff *skb) | |
670 | { | |
671 | return (struct nl_mmap_hdr *)(skb->head - NL_MMAP_HDRLEN); | |
672 | } | |
673 | ||
674 | static void netlink_ring_setup_skb(struct sk_buff *skb, struct sock *sk, | |
675 | struct netlink_ring *ring, | |
676 | struct nl_mmap_hdr *hdr) | |
677 | { | |
678 | unsigned int size; | |
679 | void *data; | |
680 | ||
681 | size = ring->frame_size - NL_MMAP_HDRLEN; | |
682 | data = (void *)hdr + NL_MMAP_HDRLEN; | |
683 | ||
684 | skb->head = data; | |
685 | skb->data = data; | |
686 | skb_reset_tail_pointer(skb); | |
687 | skb->end = skb->tail + size; | |
688 | skb->len = 0; | |
689 | ||
690 | skb->destructor = netlink_skb_destructor; | |
691 | NETLINK_CB(skb).flags |= NETLINK_SKB_MMAPED; | |
692 | NETLINK_CB(skb).sk = sk; | |
693 | } | |
5fd96123 PM |
694 | |
695 | static int netlink_mmap_sendmsg(struct sock *sk, struct msghdr *msg, | |
696 | u32 dst_portid, u32 dst_group, | |
697 | struct sock_iocb *siocb) | |
698 | { | |
699 | struct netlink_sock *nlk = nlk_sk(sk); | |
700 | struct netlink_ring *ring; | |
701 | struct nl_mmap_hdr *hdr; | |
702 | struct sk_buff *skb; | |
703 | unsigned int maxlen; | |
5fd96123 PM |
704 | int err = 0, len = 0; |
705 | ||
5fd96123 PM |
706 | mutex_lock(&nlk->pg_vec_lock); |
707 | ||
708 | ring = &nlk->tx_ring; | |
709 | maxlen = ring->frame_size - NL_MMAP_HDRLEN; | |
710 | ||
711 | do { | |
4682a035 DM |
712 | unsigned int nm_len; |
713 | ||
5fd96123 PM |
714 | hdr = netlink_current_frame(ring, NL_MMAP_STATUS_VALID); |
715 | if (hdr == NULL) { | |
716 | if (!(msg->msg_flags & MSG_DONTWAIT) && | |
717 | atomic_read(&nlk->tx_ring.pending)) | |
718 | schedule(); | |
719 | continue; | |
720 | } | |
4682a035 DM |
721 | |
722 | nm_len = ACCESS_ONCE(hdr->nm_len); | |
723 | if (nm_len > maxlen) { | |
5fd96123 PM |
724 | err = -EINVAL; |
725 | goto out; | |
726 | } | |
727 | ||
4682a035 | 728 | netlink_frame_flush_dcache(hdr, nm_len); |
5fd96123 | 729 | |
4682a035 DM |
730 | skb = alloc_skb(nm_len, GFP_KERNEL); |
731 | if (skb == NULL) { | |
732 | err = -ENOBUFS; | |
733 | goto out; | |
5fd96123 | 734 | } |
4682a035 DM |
735 | __skb_put(skb, nm_len); |
736 | memcpy(skb->data, (void *)hdr + NL_MMAP_HDRLEN, nm_len); | |
737 | netlink_set_status(hdr, NL_MMAP_STATUS_UNUSED); | |
5fd96123 PM |
738 | |
739 | netlink_increment_head(ring); | |
740 | ||
741 | NETLINK_CB(skb).portid = nlk->portid; | |
742 | NETLINK_CB(skb).dst_group = dst_group; | |
743 | NETLINK_CB(skb).creds = siocb->scm->creds; | |
744 | ||
745 | err = security_netlink_send(sk, skb); | |
746 | if (err) { | |
747 | kfree_skb(skb); | |
748 | goto out; | |
749 | } | |
750 | ||
751 | if (unlikely(dst_group)) { | |
752 | atomic_inc(&skb->users); | |
753 | netlink_broadcast(sk, skb, dst_portid, dst_group, | |
754 | GFP_KERNEL); | |
755 | } | |
756 | err = netlink_unicast(sk, skb, dst_portid, | |
757 | msg->msg_flags & MSG_DONTWAIT); | |
758 | if (err < 0) | |
759 | goto out; | |
760 | len += err; | |
761 | ||
762 | } while (hdr != NULL || | |
763 | (!(msg->msg_flags & MSG_DONTWAIT) && | |
764 | atomic_read(&nlk->tx_ring.pending))); | |
765 | ||
766 | if (len > 0) | |
767 | err = len; | |
768 | out: | |
769 | mutex_unlock(&nlk->pg_vec_lock); | |
770 | return err; | |
771 | } | |
f9c22888 PM |
772 | |
773 | static void netlink_queue_mmaped_skb(struct sock *sk, struct sk_buff *skb) | |
774 | { | |
775 | struct nl_mmap_hdr *hdr; | |
776 | ||
777 | hdr = netlink_mmap_hdr(skb); | |
778 | hdr->nm_len = skb->len; | |
779 | hdr->nm_group = NETLINK_CB(skb).dst_group; | |
780 | hdr->nm_pid = NETLINK_CB(skb).creds.pid; | |
1bf9310a ND |
781 | hdr->nm_uid = from_kuid(sk_user_ns(sk), NETLINK_CB(skb).creds.uid); |
782 | hdr->nm_gid = from_kgid(sk_user_ns(sk), NETLINK_CB(skb).creds.gid); | |
4682a035 | 783 | netlink_frame_flush_dcache(hdr, hdr->nm_len); |
f9c22888 PM |
784 | netlink_set_status(hdr, NL_MMAP_STATUS_VALID); |
785 | ||
786 | NETLINK_CB(skb).flags |= NETLINK_SKB_DELIVERED; | |
787 | kfree_skb(skb); | |
788 | } | |
789 | ||
790 | static void netlink_ring_set_copied(struct sock *sk, struct sk_buff *skb) | |
791 | { | |
792 | struct netlink_sock *nlk = nlk_sk(sk); | |
793 | struct netlink_ring *ring = &nlk->rx_ring; | |
794 | struct nl_mmap_hdr *hdr; | |
795 | ||
796 | spin_lock_bh(&sk->sk_receive_queue.lock); | |
797 | hdr = netlink_current_frame(ring, NL_MMAP_STATUS_UNUSED); | |
798 | if (hdr == NULL) { | |
799 | spin_unlock_bh(&sk->sk_receive_queue.lock); | |
800 | kfree_skb(skb); | |
cd1df525 | 801 | netlink_overrun(sk); |
f9c22888 PM |
802 | return; |
803 | } | |
804 | netlink_increment_head(ring); | |
805 | __skb_queue_tail(&sk->sk_receive_queue, skb); | |
806 | spin_unlock_bh(&sk->sk_receive_queue.lock); | |
807 | ||
808 | hdr->nm_len = skb->len; | |
809 | hdr->nm_group = NETLINK_CB(skb).dst_group; | |
810 | hdr->nm_pid = NETLINK_CB(skb).creds.pid; | |
1bf9310a ND |
811 | hdr->nm_uid = from_kuid(sk_user_ns(sk), NETLINK_CB(skb).creds.uid); |
812 | hdr->nm_gid = from_kgid(sk_user_ns(sk), NETLINK_CB(skb).creds.gid); | |
f9c22888 PM |
813 | netlink_set_status(hdr, NL_MMAP_STATUS_COPY); |
814 | } | |
815 | ||
ccdfcc39 | 816 | #else /* CONFIG_NETLINK_MMAP */ |
9652e931 | 817 | #define netlink_skb_is_mmaped(skb) false |
f9c22888 | 818 | #define netlink_rx_is_mmaped(sk) false |
5fd96123 | 819 | #define netlink_tx_is_mmaped(sk) false |
ccdfcc39 | 820 | #define netlink_mmap sock_no_mmap |
9652e931 | 821 | #define netlink_poll datagram_poll |
5fd96123 | 822 | #define netlink_mmap_sendmsg(sk, msg, dst_portid, dst_group, siocb) 0 |
ccdfcc39 PM |
823 | #endif /* CONFIG_NETLINK_MMAP */ |
824 | ||
cf0a018a PM |
825 | static void netlink_skb_destructor(struct sk_buff *skb) |
826 | { | |
9652e931 PM |
827 | #ifdef CONFIG_NETLINK_MMAP |
828 | struct nl_mmap_hdr *hdr; | |
829 | struct netlink_ring *ring; | |
830 | struct sock *sk; | |
831 | ||
832 | /* If a packet from the kernel to userspace was freed because of an | |
833 | * error without being delivered to userspace, the kernel must reset | |
834 | * the status. In the direction userspace to kernel, the status is | |
835 | * always reset here after the packet was processed and freed. | |
836 | */ | |
837 | if (netlink_skb_is_mmaped(skb)) { | |
838 | hdr = netlink_mmap_hdr(skb); | |
839 | sk = NETLINK_CB(skb).sk; | |
840 | ||
5fd96123 PM |
841 | if (NETLINK_CB(skb).flags & NETLINK_SKB_TX) { |
842 | netlink_set_status(hdr, NL_MMAP_STATUS_UNUSED); | |
843 | ring = &nlk_sk(sk)->tx_ring; | |
844 | } else { | |
845 | if (!(NETLINK_CB(skb).flags & NETLINK_SKB_DELIVERED)) { | |
846 | hdr->nm_len = 0; | |
847 | netlink_set_status(hdr, NL_MMAP_STATUS_VALID); | |
848 | } | |
849 | ring = &nlk_sk(sk)->rx_ring; | |
9652e931 | 850 | } |
9652e931 PM |
851 | |
852 | WARN_ON(atomic_read(&ring->pending) == 0); | |
853 | atomic_dec(&ring->pending); | |
854 | sock_put(sk); | |
855 | ||
5e71d9d7 | 856 | skb->head = NULL; |
9652e931 PM |
857 | } |
858 | #endif | |
c05cdb1b | 859 | if (is_vmalloc_addr(skb->head)) { |
3a36515f PN |
860 | if (!skb->cloned || |
861 | !atomic_dec_return(&(skb_shinfo(skb)->dataref))) | |
862 | vfree(skb->head); | |
863 | ||
c05cdb1b PNA |
864 | skb->head = NULL; |
865 | } | |
9652e931 PM |
866 | if (skb->sk != NULL) |
867 | sock_rfree(skb); | |
cf0a018a PM |
868 | } |
869 | ||
870 | static void netlink_skb_set_owner_r(struct sk_buff *skb, struct sock *sk) | |
871 | { | |
872 | WARN_ON(skb->sk != NULL); | |
873 | skb->sk = sk; | |
874 | skb->destructor = netlink_skb_destructor; | |
875 | atomic_add(skb->truesize, &sk->sk_rmem_alloc); | |
876 | sk_mem_charge(sk, skb->truesize); | |
877 | } | |
878 | ||
1da177e4 LT |
879 | static void netlink_sock_destruct(struct sock *sk) |
880 | { | |
3f660d66 HX |
881 | struct netlink_sock *nlk = nlk_sk(sk); |
882 | ||
16b304f3 PS |
883 | if (nlk->cb_running) { |
884 | if (nlk->cb.done) | |
885 | nlk->cb.done(&nlk->cb); | |
6dc878a8 | 886 | |
16b304f3 PS |
887 | module_put(nlk->cb.module); |
888 | kfree_skb(nlk->cb.skb); | |
3f660d66 HX |
889 | } |
890 | ||
1da177e4 | 891 | skb_queue_purge(&sk->sk_receive_queue); |
ccdfcc39 PM |
892 | #ifdef CONFIG_NETLINK_MMAP |
893 | if (1) { | |
894 | struct nl_mmap_req req; | |
895 | ||
896 | memset(&req, 0, sizeof(req)); | |
897 | if (nlk->rx_ring.pg_vec) | |
898 | netlink_set_ring(sk, &req, true, false); | |
899 | memset(&req, 0, sizeof(req)); | |
900 | if (nlk->tx_ring.pg_vec) | |
901 | netlink_set_ring(sk, &req, true, true); | |
902 | } | |
903 | #endif /* CONFIG_NETLINK_MMAP */ | |
1da177e4 LT |
904 | |
905 | if (!sock_flag(sk, SOCK_DEAD)) { | |
6ac552fd | 906 | printk(KERN_ERR "Freeing alive netlink socket %p\n", sk); |
1da177e4 LT |
907 | return; |
908 | } | |
547b792c IJ |
909 | |
910 | WARN_ON(atomic_read(&sk->sk_rmem_alloc)); | |
911 | WARN_ON(atomic_read(&sk->sk_wmem_alloc)); | |
912 | WARN_ON(nlk_sk(sk)->groups); | |
1da177e4 LT |
913 | } |
914 | ||
6ac552fd PM |
915 | /* This lock without WQ_FLAG_EXCLUSIVE is good on UP and it is _very_ bad on |
916 | * SMP. Look, when several writers sleep and reader wakes them up, all but one | |
1da177e4 LT |
917 | * immediately hit write lock and grab all the cpus. Exclusive sleep solves |
918 | * this, _but_ remember, it adds useless work on UP machines. | |
919 | */ | |
920 | ||
d136f1bd | 921 | void netlink_table_grab(void) |
9a429c49 | 922 | __acquires(nl_table_lock) |
1da177e4 | 923 | { |
d136f1bd JB |
924 | might_sleep(); |
925 | ||
6abd219c | 926 | write_lock_irq(&nl_table_lock); |
1da177e4 LT |
927 | |
928 | if (atomic_read(&nl_table_users)) { | |
929 | DECLARE_WAITQUEUE(wait, current); | |
930 | ||
931 | add_wait_queue_exclusive(&nl_table_wait, &wait); | |
6ac552fd | 932 | for (;;) { |
1da177e4 LT |
933 | set_current_state(TASK_UNINTERRUPTIBLE); |
934 | if (atomic_read(&nl_table_users) == 0) | |
935 | break; | |
6abd219c | 936 | write_unlock_irq(&nl_table_lock); |
1da177e4 | 937 | schedule(); |
6abd219c | 938 | write_lock_irq(&nl_table_lock); |
1da177e4 LT |
939 | } |
940 | ||
941 | __set_current_state(TASK_RUNNING); | |
942 | remove_wait_queue(&nl_table_wait, &wait); | |
943 | } | |
944 | } | |
945 | ||
d136f1bd | 946 | void netlink_table_ungrab(void) |
9a429c49 | 947 | __releases(nl_table_lock) |
1da177e4 | 948 | { |
6abd219c | 949 | write_unlock_irq(&nl_table_lock); |
1da177e4 LT |
950 | wake_up(&nl_table_wait); |
951 | } | |
952 | ||
6ac552fd | 953 | static inline void |
1da177e4 LT |
954 | netlink_lock_table(void) |
955 | { | |
956 | /* read_lock() synchronizes us to netlink_table_grab */ | |
957 | ||
958 | read_lock(&nl_table_lock); | |
959 | atomic_inc(&nl_table_users); | |
960 | read_unlock(&nl_table_lock); | |
961 | } | |
962 | ||
6ac552fd | 963 | static inline void |
1da177e4 LT |
964 | netlink_unlock_table(void) |
965 | { | |
966 | if (atomic_dec_and_test(&nl_table_users)) | |
967 | wake_up(&nl_table_wait); | |
968 | } | |
969 | ||
e341694e | 970 | struct netlink_compare_arg |
1da177e4 | 971 | { |
e341694e TG |
972 | struct net *net; |
973 | u32 portid; | |
974 | }; | |
1da177e4 | 975 | |
e341694e | 976 | static bool netlink_compare(void *ptr, void *arg) |
1da177e4 | 977 | { |
e341694e TG |
978 | struct netlink_compare_arg *x = arg; |
979 | struct sock *sk = ptr; | |
1da177e4 | 980 | |
e341694e TG |
981 | return nlk_sk(sk)->portid == x->portid && |
982 | net_eq(sock_net(sk), x->net); | |
1da177e4 LT |
983 | } |
984 | ||
e341694e TG |
985 | static struct sock *__netlink_lookup(struct netlink_table *table, u32 portid, |
986 | struct net *net) | |
1da177e4 | 987 | { |
e341694e TG |
988 | struct netlink_compare_arg arg = { |
989 | .net = net, | |
990 | .portid = portid, | |
991 | }; | |
1da177e4 | 992 | |
8d24c0b4 | 993 | return rhashtable_lookup_compare(&table->hash, &portid, |
e341694e | 994 | &netlink_compare, &arg); |
1da177e4 LT |
995 | } |
996 | ||
c5adde94 YX |
997 | static bool __netlink_insert(struct netlink_table *table, struct sock *sk, |
998 | struct net *net) | |
999 | { | |
1000 | struct netlink_compare_arg arg = { | |
1001 | .net = net, | |
1002 | .portid = nlk_sk(sk)->portid, | |
1003 | }; | |
1004 | ||
1005 | return rhashtable_lookup_compare_insert(&table->hash, | |
1006 | &nlk_sk(sk)->node, | |
1007 | &netlink_compare, &arg); | |
1008 | } | |
1009 | ||
e341694e | 1010 | static struct sock *netlink_lookup(struct net *net, int protocol, u32 portid) |
1da177e4 | 1011 | { |
e341694e TG |
1012 | struct netlink_table *table = &nl_table[protocol]; |
1013 | struct sock *sk; | |
1da177e4 | 1014 | |
e341694e TG |
1015 | rcu_read_lock(); |
1016 | sk = __netlink_lookup(table, portid, net); | |
1017 | if (sk) | |
1018 | sock_hold(sk); | |
1019 | rcu_read_unlock(); | |
1da177e4 | 1020 | |
e341694e | 1021 | return sk; |
1da177e4 LT |
1022 | } |
1023 | ||
90ddc4f0 | 1024 | static const struct proto_ops netlink_ops; |
1da177e4 | 1025 | |
4277a083 PM |
1026 | static void |
1027 | netlink_update_listeners(struct sock *sk) | |
1028 | { | |
1029 | struct netlink_table *tbl = &nl_table[sk->sk_protocol]; | |
4277a083 PM |
1030 | unsigned long mask; |
1031 | unsigned int i; | |
6d772ac5 ED |
1032 | struct listeners *listeners; |
1033 | ||
1034 | listeners = nl_deref_protected(tbl->listeners); | |
1035 | if (!listeners) | |
1036 | return; | |
4277a083 | 1037 | |
b4ff4f04 | 1038 | for (i = 0; i < NLGRPLONGS(tbl->groups); i++) { |
4277a083 | 1039 | mask = 0; |
b67bfe0d | 1040 | sk_for_each_bound(sk, &tbl->mc_list) { |
b4ff4f04 JB |
1041 | if (i < NLGRPLONGS(nlk_sk(sk)->ngroups)) |
1042 | mask |= nlk_sk(sk)->groups[i]; | |
1043 | } | |
6d772ac5 | 1044 | listeners->masks[i] = mask; |
4277a083 PM |
1045 | } |
1046 | /* this function is only called with the netlink table "grabbed", which | |
1047 | * makes sure updates are visible before bind or setsockopt return. */ | |
1048 | } | |
1049 | ||
15e47304 | 1050 | static int netlink_insert(struct sock *sk, struct net *net, u32 portid) |
1da177e4 | 1051 | { |
da12c90e | 1052 | struct netlink_table *table = &nl_table[sk->sk_protocol]; |
1da177e4 | 1053 | int err = -EADDRINUSE; |
1da177e4 | 1054 | |
c5adde94 | 1055 | lock_sock(sk); |
1da177e4 LT |
1056 | |
1057 | err = -EBUSY; | |
15e47304 | 1058 | if (nlk_sk(sk)->portid) |
1da177e4 LT |
1059 | goto err; |
1060 | ||
1061 | err = -ENOMEM; | |
97defe1e TG |
1062 | if (BITS_PER_LONG > 32 && |
1063 | unlikely(atomic_read(&table->hash.nelems) >= UINT_MAX)) | |
1da177e4 LT |
1064 | goto err; |
1065 | ||
15e47304 | 1066 | nlk_sk(sk)->portid = portid; |
e341694e | 1067 | sock_hold(sk); |
c5adde94 YX |
1068 | if (__netlink_insert(table, sk, net)) |
1069 | err = 0; | |
1070 | else | |
1071 | sock_put(sk); | |
1da177e4 | 1072 | err: |
c5adde94 | 1073 | release_sock(sk); |
1da177e4 LT |
1074 | return err; |
1075 | } | |
1076 | ||
1077 | static void netlink_remove(struct sock *sk) | |
1078 | { | |
e341694e TG |
1079 | struct netlink_table *table; |
1080 | ||
e341694e | 1081 | table = &nl_table[sk->sk_protocol]; |
6eba8224 | 1082 | if (rhashtable_remove(&table->hash, &nlk_sk(sk)->node)) { |
e341694e TG |
1083 | WARN_ON(atomic_read(&sk->sk_refcnt) == 1); |
1084 | __sock_put(sk); | |
1085 | } | |
e341694e | 1086 | |
1da177e4 | 1087 | netlink_table_grab(); |
b10dcb3b | 1088 | if (nlk_sk(sk)->subscriptions) { |
1da177e4 | 1089 | __sk_del_bind_node(sk); |
b10dcb3b JB |
1090 | netlink_update_listeners(sk); |
1091 | } | |
1da177e4 LT |
1092 | netlink_table_ungrab(); |
1093 | } | |
1094 | ||
1095 | static struct proto netlink_proto = { | |
1096 | .name = "NETLINK", | |
1097 | .owner = THIS_MODULE, | |
1098 | .obj_size = sizeof(struct netlink_sock), | |
1099 | }; | |
1100 | ||
1b8d7ae4 EB |
1101 | static int __netlink_create(struct net *net, struct socket *sock, |
1102 | struct mutex *cb_mutex, int protocol) | |
1da177e4 LT |
1103 | { |
1104 | struct sock *sk; | |
1105 | struct netlink_sock *nlk; | |
ab33a171 PM |
1106 | |
1107 | sock->ops = &netlink_ops; | |
1108 | ||
6257ff21 | 1109 | sk = sk_alloc(net, PF_NETLINK, GFP_KERNEL, &netlink_proto); |
ab33a171 PM |
1110 | if (!sk) |
1111 | return -ENOMEM; | |
1112 | ||
1113 | sock_init_data(sock, sk); | |
1114 | ||
1115 | nlk = nlk_sk(sk); | |
658cb354 | 1116 | if (cb_mutex) { |
ffa4d721 | 1117 | nlk->cb_mutex = cb_mutex; |
658cb354 | 1118 | } else { |
ffa4d721 PM |
1119 | nlk->cb_mutex = &nlk->cb_def_mutex; |
1120 | mutex_init(nlk->cb_mutex); | |
1121 | } | |
ab33a171 | 1122 | init_waitqueue_head(&nlk->wait); |
ccdfcc39 PM |
1123 | #ifdef CONFIG_NETLINK_MMAP |
1124 | mutex_init(&nlk->pg_vec_lock); | |
1125 | #endif | |
ab33a171 PM |
1126 | |
1127 | sk->sk_destruct = netlink_sock_destruct; | |
1128 | sk->sk_protocol = protocol; | |
1129 | return 0; | |
1130 | } | |
1131 | ||
3f378b68 EP |
1132 | static int netlink_create(struct net *net, struct socket *sock, int protocol, |
1133 | int kern) | |
ab33a171 PM |
1134 | { |
1135 | struct module *module = NULL; | |
af65bdfc | 1136 | struct mutex *cb_mutex; |
f7fa9b10 | 1137 | struct netlink_sock *nlk; |
023e2cfa JB |
1138 | int (*bind)(struct net *net, int group); |
1139 | void (*unbind)(struct net *net, int group); | |
ab33a171 | 1140 | int err = 0; |
1da177e4 LT |
1141 | |
1142 | sock->state = SS_UNCONNECTED; | |
1143 | ||
1144 | if (sock->type != SOCK_RAW && sock->type != SOCK_DGRAM) | |
1145 | return -ESOCKTNOSUPPORT; | |
1146 | ||
6ac552fd | 1147 | if (protocol < 0 || protocol >= MAX_LINKS) |
1da177e4 LT |
1148 | return -EPROTONOSUPPORT; |
1149 | ||
77247bbb | 1150 | netlink_lock_table(); |
95a5afca | 1151 | #ifdef CONFIG_MODULES |
ab33a171 | 1152 | if (!nl_table[protocol].registered) { |
77247bbb | 1153 | netlink_unlock_table(); |
4fdb3bb7 | 1154 | request_module("net-pf-%d-proto-%d", PF_NETLINK, protocol); |
77247bbb | 1155 | netlink_lock_table(); |
4fdb3bb7 | 1156 | } |
ab33a171 PM |
1157 | #endif |
1158 | if (nl_table[protocol].registered && | |
1159 | try_module_get(nl_table[protocol].module)) | |
1160 | module = nl_table[protocol].module; | |
974c37e9 AD |
1161 | else |
1162 | err = -EPROTONOSUPPORT; | |
af65bdfc | 1163 | cb_mutex = nl_table[protocol].cb_mutex; |
03292745 | 1164 | bind = nl_table[protocol].bind; |
4f520900 | 1165 | unbind = nl_table[protocol].unbind; |
77247bbb | 1166 | netlink_unlock_table(); |
4fdb3bb7 | 1167 | |
974c37e9 AD |
1168 | if (err < 0) |
1169 | goto out; | |
1170 | ||
6ac552fd PM |
1171 | err = __netlink_create(net, sock, cb_mutex, protocol); |
1172 | if (err < 0) | |
f7fa9b10 PM |
1173 | goto out_module; |
1174 | ||
6f756a8c | 1175 | local_bh_disable(); |
c1fd3b94 | 1176 | sock_prot_inuse_add(net, &netlink_proto, 1); |
6f756a8c DM |
1177 | local_bh_enable(); |
1178 | ||
f7fa9b10 | 1179 | nlk = nlk_sk(sock->sk); |
f7fa9b10 | 1180 | nlk->module = module; |
03292745 | 1181 | nlk->netlink_bind = bind; |
4f520900 | 1182 | nlk->netlink_unbind = unbind; |
ab33a171 PM |
1183 | out: |
1184 | return err; | |
1da177e4 | 1185 | |
ab33a171 PM |
1186 | out_module: |
1187 | module_put(module); | |
1188 | goto out; | |
1da177e4 LT |
1189 | } |
1190 | ||
21e4902a TG |
1191 | static void deferred_put_nlk_sk(struct rcu_head *head) |
1192 | { | |
1193 | struct netlink_sock *nlk = container_of(head, struct netlink_sock, rcu); | |
1194 | ||
1195 | sock_put(&nlk->sk); | |
1196 | } | |
1197 | ||
1da177e4 LT |
1198 | static int netlink_release(struct socket *sock) |
1199 | { | |
1200 | struct sock *sk = sock->sk; | |
1201 | struct netlink_sock *nlk; | |
1202 | ||
1203 | if (!sk) | |
1204 | return 0; | |
1205 | ||
1206 | netlink_remove(sk); | |
ac57b3a9 | 1207 | sock_orphan(sk); |
1da177e4 LT |
1208 | nlk = nlk_sk(sk); |
1209 | ||
3f660d66 HX |
1210 | /* |
1211 | * OK. Socket is unlinked, any packets that arrive now | |
1212 | * will be purged. | |
1213 | */ | |
1da177e4 | 1214 | |
1da177e4 LT |
1215 | sock->sk = NULL; |
1216 | wake_up_interruptible_all(&nlk->wait); | |
1217 | ||
1218 | skb_queue_purge(&sk->sk_write_queue); | |
1219 | ||
15e47304 | 1220 | if (nlk->portid) { |
1da177e4 | 1221 | struct netlink_notify n = { |
3b1e0a65 | 1222 | .net = sock_net(sk), |
1da177e4 | 1223 | .protocol = sk->sk_protocol, |
15e47304 | 1224 | .portid = nlk->portid, |
1da177e4 | 1225 | }; |
e041c683 AS |
1226 | atomic_notifier_call_chain(&netlink_chain, |
1227 | NETLINK_URELEASE, &n); | |
746fac4d | 1228 | } |
4fdb3bb7 | 1229 | |
5e7c001c | 1230 | module_put(nlk->module); |
4fdb3bb7 | 1231 | |
aed81560 | 1232 | if (netlink_is_kernel(sk)) { |
b10dcb3b | 1233 | netlink_table_grab(); |
869e58f8 DL |
1234 | BUG_ON(nl_table[sk->sk_protocol].registered == 0); |
1235 | if (--nl_table[sk->sk_protocol].registered == 0) { | |
6d772ac5 ED |
1236 | struct listeners *old; |
1237 | ||
1238 | old = nl_deref_protected(nl_table[sk->sk_protocol].listeners); | |
1239 | RCU_INIT_POINTER(nl_table[sk->sk_protocol].listeners, NULL); | |
1240 | kfree_rcu(old, rcu); | |
869e58f8 | 1241 | nl_table[sk->sk_protocol].module = NULL; |
9785e10a | 1242 | nl_table[sk->sk_protocol].bind = NULL; |
4f520900 | 1243 | nl_table[sk->sk_protocol].unbind = NULL; |
9785e10a | 1244 | nl_table[sk->sk_protocol].flags = 0; |
869e58f8 DL |
1245 | nl_table[sk->sk_protocol].registered = 0; |
1246 | } | |
b10dcb3b | 1247 | netlink_table_ungrab(); |
658cb354 | 1248 | } |
77247bbb | 1249 | |
7d68536b JB |
1250 | if (nlk->netlink_unbind) { |
1251 | int i; | |
1252 | ||
1253 | for (i = 0; i < nlk->ngroups; i++) | |
1254 | if (test_bit(i, nlk->groups)) | |
023e2cfa | 1255 | nlk->netlink_unbind(sock_net(sk), i + 1); |
7d68536b | 1256 | } |
f7fa9b10 PM |
1257 | kfree(nlk->groups); |
1258 | nlk->groups = NULL; | |
1259 | ||
3755810c | 1260 | local_bh_disable(); |
c1fd3b94 | 1261 | sock_prot_inuse_add(sock_net(sk), &netlink_proto, -1); |
3755810c | 1262 | local_bh_enable(); |
21e4902a | 1263 | call_rcu(&nlk->rcu, deferred_put_nlk_sk); |
1da177e4 LT |
1264 | return 0; |
1265 | } | |
1266 | ||
1267 | static int netlink_autobind(struct socket *sock) | |
1268 | { | |
1269 | struct sock *sk = sock->sk; | |
3b1e0a65 | 1270 | struct net *net = sock_net(sk); |
da12c90e | 1271 | struct netlink_table *table = &nl_table[sk->sk_protocol]; |
15e47304 | 1272 | s32 portid = task_tgid_vnr(current); |
1da177e4 LT |
1273 | int err; |
1274 | static s32 rover = -4097; | |
1275 | ||
1276 | retry: | |
1277 | cond_resched(); | |
e341694e TG |
1278 | rcu_read_lock(); |
1279 | if (__netlink_lookup(table, portid, net)) { | |
1280 | /* Bind collision, search negative portid values. */ | |
1281 | portid = rover--; | |
1282 | if (rover > -4097) | |
1283 | rover = -4097; | |
1284 | rcu_read_unlock(); | |
1285 | goto retry; | |
1da177e4 | 1286 | } |
e341694e | 1287 | rcu_read_unlock(); |
1da177e4 | 1288 | |
15e47304 | 1289 | err = netlink_insert(sk, net, portid); |
1da177e4 LT |
1290 | if (err == -EADDRINUSE) |
1291 | goto retry; | |
d470e3b4 DM |
1292 | |
1293 | /* If 2 threads race to autobind, that is fine. */ | |
1294 | if (err == -EBUSY) | |
1295 | err = 0; | |
1296 | ||
1297 | return err; | |
1da177e4 LT |
1298 | } |
1299 | ||
aa4cf945 EB |
1300 | /** |
1301 | * __netlink_ns_capable - General netlink message capability test | |
1302 | * @nsp: NETLINK_CB of the socket buffer holding a netlink command from userspace. | |
1303 | * @user_ns: The user namespace of the capability to use | |
1304 | * @cap: The capability to use | |
1305 | * | |
1306 | * Test to see if the opener of the socket we received the message | |
1307 | * from had when the netlink socket was created and the sender of the | |
1308 | * message has has the capability @cap in the user namespace @user_ns. | |
1309 | */ | |
1310 | bool __netlink_ns_capable(const struct netlink_skb_parms *nsp, | |
1311 | struct user_namespace *user_ns, int cap) | |
1312 | { | |
2d7a85f4 EB |
1313 | return ((nsp->flags & NETLINK_SKB_DST) || |
1314 | file_ns_capable(nsp->sk->sk_socket->file, user_ns, cap)) && | |
1315 | ns_capable(user_ns, cap); | |
aa4cf945 EB |
1316 | } |
1317 | EXPORT_SYMBOL(__netlink_ns_capable); | |
1318 | ||
1319 | /** | |
1320 | * netlink_ns_capable - General netlink message capability test | |
1321 | * @skb: socket buffer holding a netlink command from userspace | |
1322 | * @user_ns: The user namespace of the capability to use | |
1323 | * @cap: The capability to use | |
1324 | * | |
1325 | * Test to see if the opener of the socket we received the message | |
1326 | * from had when the netlink socket was created and the sender of the | |
1327 | * message has has the capability @cap in the user namespace @user_ns. | |
1328 | */ | |
1329 | bool netlink_ns_capable(const struct sk_buff *skb, | |
1330 | struct user_namespace *user_ns, int cap) | |
1331 | { | |
1332 | return __netlink_ns_capable(&NETLINK_CB(skb), user_ns, cap); | |
1333 | } | |
1334 | EXPORT_SYMBOL(netlink_ns_capable); | |
1335 | ||
1336 | /** | |
1337 | * netlink_capable - Netlink global message capability test | |
1338 | * @skb: socket buffer holding a netlink command from userspace | |
1339 | * @cap: The capability to use | |
1340 | * | |
1341 | * Test to see if the opener of the socket we received the message | |
1342 | * from had when the netlink socket was created and the sender of the | |
1343 | * message has has the capability @cap in all user namespaces. | |
1344 | */ | |
1345 | bool netlink_capable(const struct sk_buff *skb, int cap) | |
1346 | { | |
1347 | return netlink_ns_capable(skb, &init_user_ns, cap); | |
1348 | } | |
1349 | EXPORT_SYMBOL(netlink_capable); | |
1350 | ||
1351 | /** | |
1352 | * netlink_net_capable - Netlink network namespace message capability test | |
1353 | * @skb: socket buffer holding a netlink command from userspace | |
1354 | * @cap: The capability to use | |
1355 | * | |
1356 | * Test to see if the opener of the socket we received the message | |
1357 | * from had when the netlink socket was created and the sender of the | |
1358 | * message has has the capability @cap over the network namespace of | |
1359 | * the socket we received the message from. | |
1360 | */ | |
1361 | bool netlink_net_capable(const struct sk_buff *skb, int cap) | |
1362 | { | |
1363 | return netlink_ns_capable(skb, sock_net(skb->sk)->user_ns, cap); | |
1364 | } | |
1365 | EXPORT_SYMBOL(netlink_net_capable); | |
1366 | ||
5187cd05 | 1367 | static inline int netlink_allowed(const struct socket *sock, unsigned int flag) |
746fac4d | 1368 | { |
9785e10a | 1369 | return (nl_table[sock->sk->sk_protocol].flags & flag) || |
df008c91 | 1370 | ns_capable(sock_net(sock->sk)->user_ns, CAP_NET_ADMIN); |
746fac4d | 1371 | } |
1da177e4 | 1372 | |
f7fa9b10 PM |
1373 | static void |
1374 | netlink_update_subscriptions(struct sock *sk, unsigned int subscriptions) | |
1375 | { | |
1376 | struct netlink_sock *nlk = nlk_sk(sk); | |
1377 | ||
1378 | if (nlk->subscriptions && !subscriptions) | |
1379 | __sk_del_bind_node(sk); | |
1380 | else if (!nlk->subscriptions && subscriptions) | |
1381 | sk_add_bind_node(sk, &nl_table[sk->sk_protocol].mc_list); | |
1382 | nlk->subscriptions = subscriptions; | |
1383 | } | |
1384 | ||
b4ff4f04 | 1385 | static int netlink_realloc_groups(struct sock *sk) |
513c2500 PM |
1386 | { |
1387 | struct netlink_sock *nlk = nlk_sk(sk); | |
1388 | unsigned int groups; | |
b4ff4f04 | 1389 | unsigned long *new_groups; |
513c2500 PM |
1390 | int err = 0; |
1391 | ||
b4ff4f04 JB |
1392 | netlink_table_grab(); |
1393 | ||
513c2500 | 1394 | groups = nl_table[sk->sk_protocol].groups; |
b4ff4f04 | 1395 | if (!nl_table[sk->sk_protocol].registered) { |
513c2500 | 1396 | err = -ENOENT; |
b4ff4f04 JB |
1397 | goto out_unlock; |
1398 | } | |
513c2500 | 1399 | |
b4ff4f04 JB |
1400 | if (nlk->ngroups >= groups) |
1401 | goto out_unlock; | |
513c2500 | 1402 | |
b4ff4f04 JB |
1403 | new_groups = krealloc(nlk->groups, NLGRPSZ(groups), GFP_ATOMIC); |
1404 | if (new_groups == NULL) { | |
1405 | err = -ENOMEM; | |
1406 | goto out_unlock; | |
1407 | } | |
6ac552fd | 1408 | memset((char *)new_groups + NLGRPSZ(nlk->ngroups), 0, |
b4ff4f04 JB |
1409 | NLGRPSZ(groups) - NLGRPSZ(nlk->ngroups)); |
1410 | ||
1411 | nlk->groups = new_groups; | |
513c2500 | 1412 | nlk->ngroups = groups; |
b4ff4f04 JB |
1413 | out_unlock: |
1414 | netlink_table_ungrab(); | |
1415 | return err; | |
513c2500 PM |
1416 | } |
1417 | ||
02c81ab9 | 1418 | static void netlink_undo_bind(int group, long unsigned int groups, |
023e2cfa | 1419 | struct sock *sk) |
4f520900 | 1420 | { |
023e2cfa | 1421 | struct netlink_sock *nlk = nlk_sk(sk); |
4f520900 RGB |
1422 | int undo; |
1423 | ||
1424 | if (!nlk->netlink_unbind) | |
1425 | return; | |
1426 | ||
1427 | for (undo = 0; undo < group; undo++) | |
6251edd9 | 1428 | if (test_bit(undo, &groups)) |
023e2cfa | 1429 | nlk->netlink_unbind(sock_net(sk), undo); |
4f520900 RGB |
1430 | } |
1431 | ||
6ac552fd PM |
1432 | static int netlink_bind(struct socket *sock, struct sockaddr *addr, |
1433 | int addr_len) | |
1da177e4 LT |
1434 | { |
1435 | struct sock *sk = sock->sk; | |
3b1e0a65 | 1436 | struct net *net = sock_net(sk); |
1da177e4 LT |
1437 | struct netlink_sock *nlk = nlk_sk(sk); |
1438 | struct sockaddr_nl *nladdr = (struct sockaddr_nl *)addr; | |
1439 | int err; | |
4f520900 | 1440 | long unsigned int groups = nladdr->nl_groups; |
746fac4d | 1441 | |
4e4b5376 HFS |
1442 | if (addr_len < sizeof(struct sockaddr_nl)) |
1443 | return -EINVAL; | |
1444 | ||
1da177e4 LT |
1445 | if (nladdr->nl_family != AF_NETLINK) |
1446 | return -EINVAL; | |
1447 | ||
1448 | /* Only superuser is allowed to listen multicasts */ | |
4f520900 | 1449 | if (groups) { |
5187cd05 | 1450 | if (!netlink_allowed(sock, NL_CFG_F_NONROOT_RECV)) |
513c2500 | 1451 | return -EPERM; |
b4ff4f04 JB |
1452 | err = netlink_realloc_groups(sk); |
1453 | if (err) | |
1454 | return err; | |
513c2500 | 1455 | } |
1da177e4 | 1456 | |
4f520900 | 1457 | if (nlk->portid) |
15e47304 | 1458 | if (nladdr->nl_pid != nlk->portid) |
1da177e4 | 1459 | return -EINVAL; |
4f520900 RGB |
1460 | |
1461 | if (nlk->netlink_bind && groups) { | |
1462 | int group; | |
1463 | ||
1464 | for (group = 0; group < nlk->ngroups; group++) { | |
1465 | if (!test_bit(group, &groups)) | |
1466 | continue; | |
023e2cfa | 1467 | err = nlk->netlink_bind(net, group); |
4f520900 RGB |
1468 | if (!err) |
1469 | continue; | |
023e2cfa | 1470 | netlink_undo_bind(group, groups, sk); |
4f520900 RGB |
1471 | return err; |
1472 | } | |
1473 | } | |
1474 | ||
1475 | if (!nlk->portid) { | |
1da177e4 | 1476 | err = nladdr->nl_pid ? |
b4b51029 | 1477 | netlink_insert(sk, net, nladdr->nl_pid) : |
1da177e4 | 1478 | netlink_autobind(sock); |
4f520900 | 1479 | if (err) { |
023e2cfa | 1480 | netlink_undo_bind(nlk->ngroups, groups, sk); |
1da177e4 | 1481 | return err; |
4f520900 | 1482 | } |
1da177e4 LT |
1483 | } |
1484 | ||
4f520900 | 1485 | if (!groups && (nlk->groups == NULL || !(u32)nlk->groups[0])) |
1da177e4 LT |
1486 | return 0; |
1487 | ||
1488 | netlink_table_grab(); | |
f7fa9b10 | 1489 | netlink_update_subscriptions(sk, nlk->subscriptions + |
4f520900 | 1490 | hweight32(groups) - |
746fac4d | 1491 | hweight32(nlk->groups[0])); |
4f520900 | 1492 | nlk->groups[0] = (nlk->groups[0] & ~0xffffffffUL) | groups; |
4277a083 | 1493 | netlink_update_listeners(sk); |
1da177e4 LT |
1494 | netlink_table_ungrab(); |
1495 | ||
1496 | return 0; | |
1497 | } | |
1498 | ||
1499 | static int netlink_connect(struct socket *sock, struct sockaddr *addr, | |
1500 | int alen, int flags) | |
1501 | { | |
1502 | int err = 0; | |
1503 | struct sock *sk = sock->sk; | |
1504 | struct netlink_sock *nlk = nlk_sk(sk); | |
6ac552fd | 1505 | struct sockaddr_nl *nladdr = (struct sockaddr_nl *)addr; |
1da177e4 | 1506 | |
6503d961 CG |
1507 | if (alen < sizeof(addr->sa_family)) |
1508 | return -EINVAL; | |
1509 | ||
1da177e4 LT |
1510 | if (addr->sa_family == AF_UNSPEC) { |
1511 | sk->sk_state = NETLINK_UNCONNECTED; | |
15e47304 | 1512 | nlk->dst_portid = 0; |
d629b836 | 1513 | nlk->dst_group = 0; |
1da177e4 LT |
1514 | return 0; |
1515 | } | |
1516 | if (addr->sa_family != AF_NETLINK) | |
1517 | return -EINVAL; | |
1518 | ||
46833a86 | 1519 | if ((nladdr->nl_groups || nladdr->nl_pid) && |
5187cd05 | 1520 | !netlink_allowed(sock, NL_CFG_F_NONROOT_SEND)) |
1da177e4 LT |
1521 | return -EPERM; |
1522 | ||
15e47304 | 1523 | if (!nlk->portid) |
1da177e4 LT |
1524 | err = netlink_autobind(sock); |
1525 | ||
1526 | if (err == 0) { | |
1527 | sk->sk_state = NETLINK_CONNECTED; | |
15e47304 | 1528 | nlk->dst_portid = nladdr->nl_pid; |
d629b836 | 1529 | nlk->dst_group = ffs(nladdr->nl_groups); |
1da177e4 LT |
1530 | } |
1531 | ||
1532 | return err; | |
1533 | } | |
1534 | ||
6ac552fd PM |
1535 | static int netlink_getname(struct socket *sock, struct sockaddr *addr, |
1536 | int *addr_len, int peer) | |
1da177e4 LT |
1537 | { |
1538 | struct sock *sk = sock->sk; | |
1539 | struct netlink_sock *nlk = nlk_sk(sk); | |
13cfa97b | 1540 | DECLARE_SOCKADDR(struct sockaddr_nl *, nladdr, addr); |
746fac4d | 1541 | |
1da177e4 LT |
1542 | nladdr->nl_family = AF_NETLINK; |
1543 | nladdr->nl_pad = 0; | |
1544 | *addr_len = sizeof(*nladdr); | |
1545 | ||
1546 | if (peer) { | |
15e47304 | 1547 | nladdr->nl_pid = nlk->dst_portid; |
d629b836 | 1548 | nladdr->nl_groups = netlink_group_mask(nlk->dst_group); |
1da177e4 | 1549 | } else { |
15e47304 | 1550 | nladdr->nl_pid = nlk->portid; |
513c2500 | 1551 | nladdr->nl_groups = nlk->groups ? nlk->groups[0] : 0; |
1da177e4 LT |
1552 | } |
1553 | return 0; | |
1554 | } | |
1555 | ||
15e47304 | 1556 | static struct sock *netlink_getsockbyportid(struct sock *ssk, u32 portid) |
1da177e4 | 1557 | { |
1da177e4 LT |
1558 | struct sock *sock; |
1559 | struct netlink_sock *nlk; | |
1560 | ||
15e47304 | 1561 | sock = netlink_lookup(sock_net(ssk), ssk->sk_protocol, portid); |
1da177e4 LT |
1562 | if (!sock) |
1563 | return ERR_PTR(-ECONNREFUSED); | |
1564 | ||
1565 | /* Don't bother queuing skb if kernel socket has no input function */ | |
1566 | nlk = nlk_sk(sock); | |
cd40b7d3 | 1567 | if (sock->sk_state == NETLINK_CONNECTED && |
15e47304 | 1568 | nlk->dst_portid != nlk_sk(ssk)->portid) { |
1da177e4 LT |
1569 | sock_put(sock); |
1570 | return ERR_PTR(-ECONNREFUSED); | |
1571 | } | |
1572 | return sock; | |
1573 | } | |
1574 | ||
1575 | struct sock *netlink_getsockbyfilp(struct file *filp) | |
1576 | { | |
496ad9aa | 1577 | struct inode *inode = file_inode(filp); |
1da177e4 LT |
1578 | struct sock *sock; |
1579 | ||
1580 | if (!S_ISSOCK(inode->i_mode)) | |
1581 | return ERR_PTR(-ENOTSOCK); | |
1582 | ||
1583 | sock = SOCKET_I(inode)->sk; | |
1584 | if (sock->sk_family != AF_NETLINK) | |
1585 | return ERR_PTR(-EINVAL); | |
1586 | ||
1587 | sock_hold(sock); | |
1588 | return sock; | |
1589 | } | |
1590 | ||
3a36515f PN |
1591 | static struct sk_buff *netlink_alloc_large_skb(unsigned int size, |
1592 | int broadcast) | |
c05cdb1b PNA |
1593 | { |
1594 | struct sk_buff *skb; | |
1595 | void *data; | |
1596 | ||
3a36515f | 1597 | if (size <= NLMSG_GOODSIZE || broadcast) |
c05cdb1b PNA |
1598 | return alloc_skb(size, GFP_KERNEL); |
1599 | ||
3a36515f PN |
1600 | size = SKB_DATA_ALIGN(size) + |
1601 | SKB_DATA_ALIGN(sizeof(struct skb_shared_info)); | |
c05cdb1b PNA |
1602 | |
1603 | data = vmalloc(size); | |
1604 | if (data == NULL) | |
3a36515f | 1605 | return NULL; |
c05cdb1b | 1606 | |
3a36515f PN |
1607 | skb = build_skb(data, size); |
1608 | if (skb == NULL) | |
1609 | vfree(data); | |
1610 | else { | |
1611 | skb->head_frag = 0; | |
1612 | skb->destructor = netlink_skb_destructor; | |
1613 | } | |
c05cdb1b PNA |
1614 | |
1615 | return skb; | |
c05cdb1b PNA |
1616 | } |
1617 | ||
1da177e4 LT |
1618 | /* |
1619 | * Attach a skb to a netlink socket. | |
1620 | * The caller must hold a reference to the destination socket. On error, the | |
1621 | * reference is dropped. The skb is not send to the destination, just all | |
1622 | * all error checks are performed and memory in the queue is reserved. | |
1623 | * Return values: | |
1624 | * < 0: error. skb freed, reference to sock dropped. | |
1625 | * 0: continue | |
1626 | * 1: repeat lookup - reference dropped while waiting for socket memory. | |
1627 | */ | |
9457afee | 1628 | int netlink_attachskb(struct sock *sk, struct sk_buff *skb, |
c3d8d1e3 | 1629 | long *timeo, struct sock *ssk) |
1da177e4 LT |
1630 | { |
1631 | struct netlink_sock *nlk; | |
1632 | ||
1633 | nlk = nlk_sk(sk); | |
1634 | ||
5fd96123 PM |
1635 | if ((atomic_read(&sk->sk_rmem_alloc) > sk->sk_rcvbuf || |
1636 | test_bit(NETLINK_CONGESTED, &nlk->state)) && | |
1637 | !netlink_skb_is_mmaped(skb)) { | |
1da177e4 | 1638 | DECLARE_WAITQUEUE(wait, current); |
c3d8d1e3 | 1639 | if (!*timeo) { |
aed81560 | 1640 | if (!ssk || netlink_is_kernel(ssk)) |
1da177e4 LT |
1641 | netlink_overrun(sk); |
1642 | sock_put(sk); | |
1643 | kfree_skb(skb); | |
1644 | return -EAGAIN; | |
1645 | } | |
1646 | ||
1647 | __set_current_state(TASK_INTERRUPTIBLE); | |
1648 | add_wait_queue(&nlk->wait, &wait); | |
1649 | ||
1650 | if ((atomic_read(&sk->sk_rmem_alloc) > sk->sk_rcvbuf || | |
cd967e05 | 1651 | test_bit(NETLINK_CONGESTED, &nlk->state)) && |
1da177e4 | 1652 | !sock_flag(sk, SOCK_DEAD)) |
c3d8d1e3 | 1653 | *timeo = schedule_timeout(*timeo); |
1da177e4 LT |
1654 | |
1655 | __set_current_state(TASK_RUNNING); | |
1656 | remove_wait_queue(&nlk->wait, &wait); | |
1657 | sock_put(sk); | |
1658 | ||
1659 | if (signal_pending(current)) { | |
1660 | kfree_skb(skb); | |
c3d8d1e3 | 1661 | return sock_intr_errno(*timeo); |
1da177e4 LT |
1662 | } |
1663 | return 1; | |
1664 | } | |
cf0a018a | 1665 | netlink_skb_set_owner_r(skb, sk); |
1da177e4 LT |
1666 | return 0; |
1667 | } | |
1668 | ||
4a7e7c2a | 1669 | static int __netlink_sendskb(struct sock *sk, struct sk_buff *skb) |
1da177e4 | 1670 | { |
1da177e4 LT |
1671 | int len = skb->len; |
1672 | ||
bcbde0d4 DB |
1673 | netlink_deliver_tap(skb); |
1674 | ||
f9c22888 PM |
1675 | #ifdef CONFIG_NETLINK_MMAP |
1676 | if (netlink_skb_is_mmaped(skb)) | |
1677 | netlink_queue_mmaped_skb(sk, skb); | |
1678 | else if (netlink_rx_is_mmaped(sk)) | |
1679 | netlink_ring_set_copied(sk, skb); | |
1680 | else | |
1681 | #endif /* CONFIG_NETLINK_MMAP */ | |
1682 | skb_queue_tail(&sk->sk_receive_queue, skb); | |
676d2369 | 1683 | sk->sk_data_ready(sk); |
4a7e7c2a ED |
1684 | return len; |
1685 | } | |
1686 | ||
1687 | int netlink_sendskb(struct sock *sk, struct sk_buff *skb) | |
1688 | { | |
1689 | int len = __netlink_sendskb(sk, skb); | |
1690 | ||
1da177e4 LT |
1691 | sock_put(sk); |
1692 | return len; | |
1693 | } | |
1694 | ||
1695 | void netlink_detachskb(struct sock *sk, struct sk_buff *skb) | |
1696 | { | |
1697 | kfree_skb(skb); | |
1698 | sock_put(sk); | |
1699 | } | |
1700 | ||
b57ef81f | 1701 | static struct sk_buff *netlink_trim(struct sk_buff *skb, gfp_t allocation) |
1da177e4 LT |
1702 | { |
1703 | int delta; | |
1704 | ||
1298ca46 | 1705 | WARN_ON(skb->sk != NULL); |
5fd96123 PM |
1706 | if (netlink_skb_is_mmaped(skb)) |
1707 | return skb; | |
1da177e4 | 1708 | |
4305b541 | 1709 | delta = skb->end - skb->tail; |
c05cdb1b | 1710 | if (is_vmalloc_addr(skb->head) || delta * 2 < skb->truesize) |
1da177e4 LT |
1711 | return skb; |
1712 | ||
1713 | if (skb_shared(skb)) { | |
1714 | struct sk_buff *nskb = skb_clone(skb, allocation); | |
1715 | if (!nskb) | |
1716 | return skb; | |
8460c00f | 1717 | consume_skb(skb); |
1da177e4 LT |
1718 | skb = nskb; |
1719 | } | |
1720 | ||
1721 | if (!pskb_expand_head(skb, 0, -delta, allocation)) | |
1722 | skb->truesize -= delta; | |
1723 | ||
1724 | return skb; | |
1725 | } | |
1726 | ||
3fbc2905 EB |
1727 | static int netlink_unicast_kernel(struct sock *sk, struct sk_buff *skb, |
1728 | struct sock *ssk) | |
cd40b7d3 DL |
1729 | { |
1730 | int ret; | |
1731 | struct netlink_sock *nlk = nlk_sk(sk); | |
1732 | ||
1733 | ret = -ECONNREFUSED; | |
1734 | if (nlk->netlink_rcv != NULL) { | |
1735 | ret = skb->len; | |
cf0a018a | 1736 | netlink_skb_set_owner_r(skb, sk); |
e32123e5 | 1737 | NETLINK_CB(skb).sk = ssk; |
73bfd370 | 1738 | netlink_deliver_tap_kernel(sk, ssk, skb); |
cd40b7d3 | 1739 | nlk->netlink_rcv(skb); |
bfb253c9 ED |
1740 | consume_skb(skb); |
1741 | } else { | |
1742 | kfree_skb(skb); | |
cd40b7d3 | 1743 | } |
cd40b7d3 DL |
1744 | sock_put(sk); |
1745 | return ret; | |
1746 | } | |
1747 | ||
1748 | int netlink_unicast(struct sock *ssk, struct sk_buff *skb, | |
15e47304 | 1749 | u32 portid, int nonblock) |
1da177e4 LT |
1750 | { |
1751 | struct sock *sk; | |
1752 | int err; | |
1753 | long timeo; | |
1754 | ||
1755 | skb = netlink_trim(skb, gfp_any()); | |
1756 | ||
1757 | timeo = sock_sndtimeo(ssk, nonblock); | |
1758 | retry: | |
15e47304 | 1759 | sk = netlink_getsockbyportid(ssk, portid); |
1da177e4 LT |
1760 | if (IS_ERR(sk)) { |
1761 | kfree_skb(skb); | |
1762 | return PTR_ERR(sk); | |
1763 | } | |
cd40b7d3 | 1764 | if (netlink_is_kernel(sk)) |
3fbc2905 | 1765 | return netlink_unicast_kernel(sk, skb, ssk); |
cd40b7d3 | 1766 | |
b1153f29 | 1767 | if (sk_filter(sk, skb)) { |
84874607 | 1768 | err = skb->len; |
b1153f29 SH |
1769 | kfree_skb(skb); |
1770 | sock_put(sk); | |
1771 | return err; | |
1772 | } | |
1773 | ||
9457afee | 1774 | err = netlink_attachskb(sk, skb, &timeo, ssk); |
1da177e4 LT |
1775 | if (err == 1) |
1776 | goto retry; | |
1777 | if (err) | |
1778 | return err; | |
1779 | ||
7ee015e0 | 1780 | return netlink_sendskb(sk, skb); |
1da177e4 | 1781 | } |
6ac552fd | 1782 | EXPORT_SYMBOL(netlink_unicast); |
1da177e4 | 1783 | |
f9c22888 PM |
1784 | struct sk_buff *netlink_alloc_skb(struct sock *ssk, unsigned int size, |
1785 | u32 dst_portid, gfp_t gfp_mask) | |
1786 | { | |
1787 | #ifdef CONFIG_NETLINK_MMAP | |
1788 | struct sock *sk = NULL; | |
1789 | struct sk_buff *skb; | |
1790 | struct netlink_ring *ring; | |
1791 | struct nl_mmap_hdr *hdr; | |
1792 | unsigned int maxlen; | |
1793 | ||
1794 | sk = netlink_getsockbyportid(ssk, dst_portid); | |
1795 | if (IS_ERR(sk)) | |
1796 | goto out; | |
1797 | ||
1798 | ring = &nlk_sk(sk)->rx_ring; | |
1799 | /* fast-path without atomic ops for common case: non-mmaped receiver */ | |
1800 | if (ring->pg_vec == NULL) | |
1801 | goto out_put; | |
1802 | ||
aae9f0e2 TG |
1803 | if (ring->frame_size - NL_MMAP_HDRLEN < size) |
1804 | goto out_put; | |
1805 | ||
f9c22888 PM |
1806 | skb = alloc_skb_head(gfp_mask); |
1807 | if (skb == NULL) | |
1808 | goto err1; | |
1809 | ||
1810 | spin_lock_bh(&sk->sk_receive_queue.lock); | |
1811 | /* check again under lock */ | |
1812 | if (ring->pg_vec == NULL) | |
1813 | goto out_free; | |
1814 | ||
aae9f0e2 | 1815 | /* check again under lock */ |
f9c22888 PM |
1816 | maxlen = ring->frame_size - NL_MMAP_HDRLEN; |
1817 | if (maxlen < size) | |
1818 | goto out_free; | |
1819 | ||
1820 | netlink_forward_ring(ring); | |
1821 | hdr = netlink_current_frame(ring, NL_MMAP_STATUS_UNUSED); | |
1822 | if (hdr == NULL) | |
1823 | goto err2; | |
1824 | netlink_ring_setup_skb(skb, sk, ring, hdr); | |
1825 | netlink_set_status(hdr, NL_MMAP_STATUS_RESERVED); | |
1826 | atomic_inc(&ring->pending); | |
1827 | netlink_increment_head(ring); | |
1828 | ||
1829 | spin_unlock_bh(&sk->sk_receive_queue.lock); | |
1830 | return skb; | |
1831 | ||
1832 | err2: | |
1833 | kfree_skb(skb); | |
1834 | spin_unlock_bh(&sk->sk_receive_queue.lock); | |
cd1df525 | 1835 | netlink_overrun(sk); |
f9c22888 PM |
1836 | err1: |
1837 | sock_put(sk); | |
1838 | return NULL; | |
1839 | ||
1840 | out_free: | |
1841 | kfree_skb(skb); | |
1842 | spin_unlock_bh(&sk->sk_receive_queue.lock); | |
1843 | out_put: | |
1844 | sock_put(sk); | |
1845 | out: | |
1846 | #endif | |
1847 | return alloc_skb(size, gfp_mask); | |
1848 | } | |
1849 | EXPORT_SYMBOL_GPL(netlink_alloc_skb); | |
1850 | ||
4277a083 PM |
1851 | int netlink_has_listeners(struct sock *sk, unsigned int group) |
1852 | { | |
1853 | int res = 0; | |
5c398dc8 | 1854 | struct listeners *listeners; |
4277a083 | 1855 | |
aed81560 | 1856 | BUG_ON(!netlink_is_kernel(sk)); |
b4ff4f04 JB |
1857 | |
1858 | rcu_read_lock(); | |
1859 | listeners = rcu_dereference(nl_table[sk->sk_protocol].listeners); | |
1860 | ||
6d772ac5 | 1861 | if (listeners && group - 1 < nl_table[sk->sk_protocol].groups) |
5c398dc8 | 1862 | res = test_bit(group - 1, listeners->masks); |
b4ff4f04 JB |
1863 | |
1864 | rcu_read_unlock(); | |
1865 | ||
4277a083 PM |
1866 | return res; |
1867 | } | |
1868 | EXPORT_SYMBOL_GPL(netlink_has_listeners); | |
1869 | ||
b57ef81f | 1870 | static int netlink_broadcast_deliver(struct sock *sk, struct sk_buff *skb) |
1da177e4 LT |
1871 | { |
1872 | struct netlink_sock *nlk = nlk_sk(sk); | |
1873 | ||
1874 | if (atomic_read(&sk->sk_rmem_alloc) <= sk->sk_rcvbuf && | |
cd967e05 | 1875 | !test_bit(NETLINK_CONGESTED, &nlk->state)) { |
cf0a018a | 1876 | netlink_skb_set_owner_r(skb, sk); |
4a7e7c2a | 1877 | __netlink_sendskb(sk, skb); |
2c645800 | 1878 | return atomic_read(&sk->sk_rmem_alloc) > (sk->sk_rcvbuf >> 1); |
1da177e4 LT |
1879 | } |
1880 | return -1; | |
1881 | } | |
1882 | ||
1883 | struct netlink_broadcast_data { | |
1884 | struct sock *exclude_sk; | |
b4b51029 | 1885 | struct net *net; |
15e47304 | 1886 | u32 portid; |
1da177e4 LT |
1887 | u32 group; |
1888 | int failure; | |
ff491a73 | 1889 | int delivery_failure; |
1da177e4 LT |
1890 | int congested; |
1891 | int delivered; | |
7d877f3b | 1892 | gfp_t allocation; |
1da177e4 | 1893 | struct sk_buff *skb, *skb2; |
910a7e90 EB |
1894 | int (*tx_filter)(struct sock *dsk, struct sk_buff *skb, void *data); |
1895 | void *tx_data; | |
1da177e4 LT |
1896 | }; |
1897 | ||
46c9521f RR |
1898 | static void do_one_broadcast(struct sock *sk, |
1899 | struct netlink_broadcast_data *p) | |
1da177e4 LT |
1900 | { |
1901 | struct netlink_sock *nlk = nlk_sk(sk); | |
1902 | int val; | |
1903 | ||
1904 | if (p->exclude_sk == sk) | |
46c9521f | 1905 | return; |
1da177e4 | 1906 | |
15e47304 | 1907 | if (nlk->portid == p->portid || p->group - 1 >= nlk->ngroups || |
f7fa9b10 | 1908 | !test_bit(p->group - 1, nlk->groups)) |
46c9521f | 1909 | return; |
1da177e4 | 1910 | |
878628fb | 1911 | if (!net_eq(sock_net(sk), p->net)) |
46c9521f | 1912 | return; |
b4b51029 | 1913 | |
1da177e4 LT |
1914 | if (p->failure) { |
1915 | netlink_overrun(sk); | |
46c9521f | 1916 | return; |
1da177e4 LT |
1917 | } |
1918 | ||
1919 | sock_hold(sk); | |
1920 | if (p->skb2 == NULL) { | |
68acc024 | 1921 | if (skb_shared(p->skb)) { |
1da177e4 LT |
1922 | p->skb2 = skb_clone(p->skb, p->allocation); |
1923 | } else { | |
68acc024 TC |
1924 | p->skb2 = skb_get(p->skb); |
1925 | /* | |
1926 | * skb ownership may have been set when | |
1927 | * delivered to a previous socket. | |
1928 | */ | |
1929 | skb_orphan(p->skb2); | |
1da177e4 LT |
1930 | } |
1931 | } | |
1932 | if (p->skb2 == NULL) { | |
1933 | netlink_overrun(sk); | |
1934 | /* Clone failed. Notify ALL listeners. */ | |
1935 | p->failure = 1; | |
be0c22a4 PNA |
1936 | if (nlk->flags & NETLINK_BROADCAST_SEND_ERROR) |
1937 | p->delivery_failure = 1; | |
910a7e90 EB |
1938 | } else if (p->tx_filter && p->tx_filter(sk, p->skb2, p->tx_data)) { |
1939 | kfree_skb(p->skb2); | |
1940 | p->skb2 = NULL; | |
b1153f29 SH |
1941 | } else if (sk_filter(sk, p->skb2)) { |
1942 | kfree_skb(p->skb2); | |
1943 | p->skb2 = NULL; | |
1da177e4 LT |
1944 | } else if ((val = netlink_broadcast_deliver(sk, p->skb2)) < 0) { |
1945 | netlink_overrun(sk); | |
be0c22a4 PNA |
1946 | if (nlk->flags & NETLINK_BROADCAST_SEND_ERROR) |
1947 | p->delivery_failure = 1; | |
1da177e4 LT |
1948 | } else { |
1949 | p->congested |= val; | |
1950 | p->delivered = 1; | |
1951 | p->skb2 = NULL; | |
1952 | } | |
1953 | sock_put(sk); | |
1da177e4 LT |
1954 | } |
1955 | ||
15e47304 | 1956 | int netlink_broadcast_filtered(struct sock *ssk, struct sk_buff *skb, u32 portid, |
910a7e90 EB |
1957 | u32 group, gfp_t allocation, |
1958 | int (*filter)(struct sock *dsk, struct sk_buff *skb, void *data), | |
1959 | void *filter_data) | |
1da177e4 | 1960 | { |
3b1e0a65 | 1961 | struct net *net = sock_net(ssk); |
1da177e4 | 1962 | struct netlink_broadcast_data info; |
1da177e4 LT |
1963 | struct sock *sk; |
1964 | ||
1965 | skb = netlink_trim(skb, allocation); | |
1966 | ||
1967 | info.exclude_sk = ssk; | |
b4b51029 | 1968 | info.net = net; |
15e47304 | 1969 | info.portid = portid; |
1da177e4 LT |
1970 | info.group = group; |
1971 | info.failure = 0; | |
ff491a73 | 1972 | info.delivery_failure = 0; |
1da177e4 LT |
1973 | info.congested = 0; |
1974 | info.delivered = 0; | |
1975 | info.allocation = allocation; | |
1976 | info.skb = skb; | |
1977 | info.skb2 = NULL; | |
910a7e90 EB |
1978 | info.tx_filter = filter; |
1979 | info.tx_data = filter_data; | |
1da177e4 LT |
1980 | |
1981 | /* While we sleep in clone, do not allow to change socket list */ | |
1982 | ||
1983 | netlink_lock_table(); | |
1984 | ||
b67bfe0d | 1985 | sk_for_each_bound(sk, &nl_table[ssk->sk_protocol].mc_list) |
1da177e4 LT |
1986 | do_one_broadcast(sk, &info); |
1987 | ||
70d4bf6d | 1988 | consume_skb(skb); |
aa1c6a6f | 1989 | |
1da177e4 LT |
1990 | netlink_unlock_table(); |
1991 | ||
70d4bf6d NH |
1992 | if (info.delivery_failure) { |
1993 | kfree_skb(info.skb2); | |
ff491a73 | 1994 | return -ENOBUFS; |
658cb354 ED |
1995 | } |
1996 | consume_skb(info.skb2); | |
ff491a73 | 1997 | |
1da177e4 LT |
1998 | if (info.delivered) { |
1999 | if (info.congested && (allocation & __GFP_WAIT)) | |
2000 | yield(); | |
2001 | return 0; | |
2002 | } | |
1da177e4 LT |
2003 | return -ESRCH; |
2004 | } | |
910a7e90 EB |
2005 | EXPORT_SYMBOL(netlink_broadcast_filtered); |
2006 | ||
15e47304 | 2007 | int netlink_broadcast(struct sock *ssk, struct sk_buff *skb, u32 portid, |
910a7e90 EB |
2008 | u32 group, gfp_t allocation) |
2009 | { | |
15e47304 | 2010 | return netlink_broadcast_filtered(ssk, skb, portid, group, allocation, |
910a7e90 EB |
2011 | NULL, NULL); |
2012 | } | |
6ac552fd | 2013 | EXPORT_SYMBOL(netlink_broadcast); |
1da177e4 LT |
2014 | |
2015 | struct netlink_set_err_data { | |
2016 | struct sock *exclude_sk; | |
15e47304 | 2017 | u32 portid; |
1da177e4 LT |
2018 | u32 group; |
2019 | int code; | |
2020 | }; | |
2021 | ||
b57ef81f | 2022 | static int do_one_set_err(struct sock *sk, struct netlink_set_err_data *p) |
1da177e4 LT |
2023 | { |
2024 | struct netlink_sock *nlk = nlk_sk(sk); | |
1a50307b | 2025 | int ret = 0; |
1da177e4 LT |
2026 | |
2027 | if (sk == p->exclude_sk) | |
2028 | goto out; | |
2029 | ||
09ad9bc7 | 2030 | if (!net_eq(sock_net(sk), sock_net(p->exclude_sk))) |
b4b51029 EB |
2031 | goto out; |
2032 | ||
15e47304 | 2033 | if (nlk->portid == p->portid || p->group - 1 >= nlk->ngroups || |
f7fa9b10 | 2034 | !test_bit(p->group - 1, nlk->groups)) |
1da177e4 LT |
2035 | goto out; |
2036 | ||
1a50307b PNA |
2037 | if (p->code == ENOBUFS && nlk->flags & NETLINK_RECV_NO_ENOBUFS) { |
2038 | ret = 1; | |
2039 | goto out; | |
2040 | } | |
2041 | ||
1da177e4 LT |
2042 | sk->sk_err = p->code; |
2043 | sk->sk_error_report(sk); | |
2044 | out: | |
1a50307b | 2045 | return ret; |
1da177e4 LT |
2046 | } |
2047 | ||
4843b93c PNA |
2048 | /** |
2049 | * netlink_set_err - report error to broadcast listeners | |
2050 | * @ssk: the kernel netlink socket, as returned by netlink_kernel_create() | |
15e47304 | 2051 | * @portid: the PORTID of a process that we want to skip (if any) |
840e93f2 | 2052 | * @group: the broadcast group that will notice the error |
4843b93c | 2053 | * @code: error code, must be negative (as usual in kernelspace) |
1a50307b PNA |
2054 | * |
2055 | * This function returns the number of broadcast listeners that have set the | |
2056 | * NETLINK_RECV_NO_ENOBUFS socket option. | |
4843b93c | 2057 | */ |
15e47304 | 2058 | int netlink_set_err(struct sock *ssk, u32 portid, u32 group, int code) |
1da177e4 LT |
2059 | { |
2060 | struct netlink_set_err_data info; | |
1da177e4 | 2061 | struct sock *sk; |
1a50307b | 2062 | int ret = 0; |
1da177e4 LT |
2063 | |
2064 | info.exclude_sk = ssk; | |
15e47304 | 2065 | info.portid = portid; |
1da177e4 | 2066 | info.group = group; |
4843b93c PNA |
2067 | /* sk->sk_err wants a positive error value */ |
2068 | info.code = -code; | |
1da177e4 LT |
2069 | |
2070 | read_lock(&nl_table_lock); | |
2071 | ||
b67bfe0d | 2072 | sk_for_each_bound(sk, &nl_table[ssk->sk_protocol].mc_list) |
1a50307b | 2073 | ret += do_one_set_err(sk, &info); |
1da177e4 LT |
2074 | |
2075 | read_unlock(&nl_table_lock); | |
1a50307b | 2076 | return ret; |
1da177e4 | 2077 | } |
dd5b6ce6 | 2078 | EXPORT_SYMBOL(netlink_set_err); |
1da177e4 | 2079 | |
84659eb5 JB |
2080 | /* must be called with netlink table grabbed */ |
2081 | static void netlink_update_socket_mc(struct netlink_sock *nlk, | |
2082 | unsigned int group, | |
2083 | int is_new) | |
2084 | { | |
2085 | int old, new = !!is_new, subscriptions; | |
2086 | ||
2087 | old = test_bit(group - 1, nlk->groups); | |
2088 | subscriptions = nlk->subscriptions - old + new; | |
2089 | if (new) | |
2090 | __set_bit(group - 1, nlk->groups); | |
2091 | else | |
2092 | __clear_bit(group - 1, nlk->groups); | |
2093 | netlink_update_subscriptions(&nlk->sk, subscriptions); | |
2094 | netlink_update_listeners(&nlk->sk); | |
2095 | } | |
2096 | ||
9a4595bc | 2097 | static int netlink_setsockopt(struct socket *sock, int level, int optname, |
b7058842 | 2098 | char __user *optval, unsigned int optlen) |
9a4595bc PM |
2099 | { |
2100 | struct sock *sk = sock->sk; | |
2101 | struct netlink_sock *nlk = nlk_sk(sk); | |
eb496534 JB |
2102 | unsigned int val = 0; |
2103 | int err; | |
9a4595bc PM |
2104 | |
2105 | if (level != SOL_NETLINK) | |
2106 | return -ENOPROTOOPT; | |
2107 | ||
ccdfcc39 PM |
2108 | if (optname != NETLINK_RX_RING && optname != NETLINK_TX_RING && |
2109 | optlen >= sizeof(int) && | |
eb496534 | 2110 | get_user(val, (unsigned int __user *)optval)) |
9a4595bc PM |
2111 | return -EFAULT; |
2112 | ||
2113 | switch (optname) { | |
2114 | case NETLINK_PKTINFO: | |
2115 | if (val) | |
2116 | nlk->flags |= NETLINK_RECV_PKTINFO; | |
2117 | else | |
2118 | nlk->flags &= ~NETLINK_RECV_PKTINFO; | |
2119 | err = 0; | |
2120 | break; | |
2121 | case NETLINK_ADD_MEMBERSHIP: | |
2122 | case NETLINK_DROP_MEMBERSHIP: { | |
5187cd05 | 2123 | if (!netlink_allowed(sock, NL_CFG_F_NONROOT_RECV)) |
9a4595bc | 2124 | return -EPERM; |
b4ff4f04 JB |
2125 | err = netlink_realloc_groups(sk); |
2126 | if (err) | |
2127 | return err; | |
9a4595bc PM |
2128 | if (!val || val - 1 >= nlk->ngroups) |
2129 | return -EINVAL; | |
7774d5e0 | 2130 | if (optname == NETLINK_ADD_MEMBERSHIP && nlk->netlink_bind) { |
023e2cfa | 2131 | err = nlk->netlink_bind(sock_net(sk), val); |
4f520900 RGB |
2132 | if (err) |
2133 | return err; | |
2134 | } | |
9a4595bc | 2135 | netlink_table_grab(); |
84659eb5 JB |
2136 | netlink_update_socket_mc(nlk, val, |
2137 | optname == NETLINK_ADD_MEMBERSHIP); | |
9a4595bc | 2138 | netlink_table_ungrab(); |
7774d5e0 | 2139 | if (optname == NETLINK_DROP_MEMBERSHIP && nlk->netlink_unbind) |
023e2cfa | 2140 | nlk->netlink_unbind(sock_net(sk), val); |
03292745 | 2141 | |
9a4595bc PM |
2142 | err = 0; |
2143 | break; | |
2144 | } | |
be0c22a4 PNA |
2145 | case NETLINK_BROADCAST_ERROR: |
2146 | if (val) | |
2147 | nlk->flags |= NETLINK_BROADCAST_SEND_ERROR; | |
2148 | else | |
2149 | nlk->flags &= ~NETLINK_BROADCAST_SEND_ERROR; | |
2150 | err = 0; | |
2151 | break; | |
38938bfe PNA |
2152 | case NETLINK_NO_ENOBUFS: |
2153 | if (val) { | |
2154 | nlk->flags |= NETLINK_RECV_NO_ENOBUFS; | |
cd967e05 | 2155 | clear_bit(NETLINK_CONGESTED, &nlk->state); |
38938bfe | 2156 | wake_up_interruptible(&nlk->wait); |
658cb354 | 2157 | } else { |
38938bfe | 2158 | nlk->flags &= ~NETLINK_RECV_NO_ENOBUFS; |
658cb354 | 2159 | } |
38938bfe PNA |
2160 | err = 0; |
2161 | break; | |
ccdfcc39 PM |
2162 | #ifdef CONFIG_NETLINK_MMAP |
2163 | case NETLINK_RX_RING: | |
2164 | case NETLINK_TX_RING: { | |
2165 | struct nl_mmap_req req; | |
2166 | ||
2167 | /* Rings might consume more memory than queue limits, require | |
2168 | * CAP_NET_ADMIN. | |
2169 | */ | |
2170 | if (!capable(CAP_NET_ADMIN)) | |
2171 | return -EPERM; | |
2172 | if (optlen < sizeof(req)) | |
2173 | return -EINVAL; | |
2174 | if (copy_from_user(&req, optval, sizeof(req))) | |
2175 | return -EFAULT; | |
2176 | err = netlink_set_ring(sk, &req, false, | |
2177 | optname == NETLINK_TX_RING); | |
2178 | break; | |
2179 | } | |
2180 | #endif /* CONFIG_NETLINK_MMAP */ | |
9a4595bc PM |
2181 | default: |
2182 | err = -ENOPROTOOPT; | |
2183 | } | |
2184 | return err; | |
2185 | } | |
2186 | ||
2187 | static int netlink_getsockopt(struct socket *sock, int level, int optname, | |
746fac4d | 2188 | char __user *optval, int __user *optlen) |
9a4595bc PM |
2189 | { |
2190 | struct sock *sk = sock->sk; | |
2191 | struct netlink_sock *nlk = nlk_sk(sk); | |
2192 | int len, val, err; | |
2193 | ||
2194 | if (level != SOL_NETLINK) | |
2195 | return -ENOPROTOOPT; | |
2196 | ||
2197 | if (get_user(len, optlen)) | |
2198 | return -EFAULT; | |
2199 | if (len < 0) | |
2200 | return -EINVAL; | |
2201 | ||
2202 | switch (optname) { | |
2203 | case NETLINK_PKTINFO: | |
2204 | if (len < sizeof(int)) | |
2205 | return -EINVAL; | |
2206 | len = sizeof(int); | |
2207 | val = nlk->flags & NETLINK_RECV_PKTINFO ? 1 : 0; | |
a27b58fe HC |
2208 | if (put_user(len, optlen) || |
2209 | put_user(val, optval)) | |
2210 | return -EFAULT; | |
9a4595bc PM |
2211 | err = 0; |
2212 | break; | |
be0c22a4 PNA |
2213 | case NETLINK_BROADCAST_ERROR: |
2214 | if (len < sizeof(int)) | |
2215 | return -EINVAL; | |
2216 | len = sizeof(int); | |
2217 | val = nlk->flags & NETLINK_BROADCAST_SEND_ERROR ? 1 : 0; | |
2218 | if (put_user(len, optlen) || | |
2219 | put_user(val, optval)) | |
2220 | return -EFAULT; | |
2221 | err = 0; | |
2222 | break; | |
38938bfe PNA |
2223 | case NETLINK_NO_ENOBUFS: |
2224 | if (len < sizeof(int)) | |
2225 | return -EINVAL; | |
2226 | len = sizeof(int); | |
2227 | val = nlk->flags & NETLINK_RECV_NO_ENOBUFS ? 1 : 0; | |
2228 | if (put_user(len, optlen) || | |
2229 | put_user(val, optval)) | |
2230 | return -EFAULT; | |
2231 | err = 0; | |
2232 | break; | |
9a4595bc PM |
2233 | default: |
2234 | err = -ENOPROTOOPT; | |
2235 | } | |
2236 | return err; | |
2237 | } | |
2238 | ||
2239 | static void netlink_cmsg_recv_pktinfo(struct msghdr *msg, struct sk_buff *skb) | |
2240 | { | |
2241 | struct nl_pktinfo info; | |
2242 | ||
2243 | info.group = NETLINK_CB(skb).dst_group; | |
2244 | put_cmsg(msg, SOL_NETLINK, NETLINK_PKTINFO, sizeof(info), &info); | |
2245 | } | |
2246 | ||
1da177e4 LT |
2247 | static int netlink_sendmsg(struct kiocb *kiocb, struct socket *sock, |
2248 | struct msghdr *msg, size_t len) | |
2249 | { | |
2250 | struct sock_iocb *siocb = kiocb_to_siocb(kiocb); | |
2251 | struct sock *sk = sock->sk; | |
2252 | struct netlink_sock *nlk = nlk_sk(sk); | |
342dfc30 | 2253 | DECLARE_SOCKADDR(struct sockaddr_nl *, addr, msg->msg_name); |
15e47304 | 2254 | u32 dst_portid; |
d629b836 | 2255 | u32 dst_group; |
1da177e4 LT |
2256 | struct sk_buff *skb; |
2257 | int err; | |
2258 | struct scm_cookie scm; | |
2d7a85f4 | 2259 | u32 netlink_skb_flags = 0; |
1da177e4 LT |
2260 | |
2261 | if (msg->msg_flags&MSG_OOB) | |
2262 | return -EOPNOTSUPP; | |
2263 | ||
16e57262 | 2264 | if (NULL == siocb->scm) |
1da177e4 | 2265 | siocb->scm = &scm; |
16e57262 | 2266 | |
e0e3cea4 | 2267 | err = scm_send(sock, msg, siocb->scm, true); |
1da177e4 LT |
2268 | if (err < 0) |
2269 | return err; | |
2270 | ||
2271 | if (msg->msg_namelen) { | |
b47030c7 | 2272 | err = -EINVAL; |
1da177e4 | 2273 | if (addr->nl_family != AF_NETLINK) |
b47030c7 | 2274 | goto out; |
15e47304 | 2275 | dst_portid = addr->nl_pid; |
d629b836 | 2276 | dst_group = ffs(addr->nl_groups); |
b47030c7 | 2277 | err = -EPERM; |
15e47304 | 2278 | if ((dst_group || dst_portid) && |
5187cd05 | 2279 | !netlink_allowed(sock, NL_CFG_F_NONROOT_SEND)) |
b47030c7 | 2280 | goto out; |
2d7a85f4 | 2281 | netlink_skb_flags |= NETLINK_SKB_DST; |
1da177e4 | 2282 | } else { |
15e47304 | 2283 | dst_portid = nlk->dst_portid; |
d629b836 | 2284 | dst_group = nlk->dst_group; |
1da177e4 LT |
2285 | } |
2286 | ||
15e47304 | 2287 | if (!nlk->portid) { |
1da177e4 LT |
2288 | err = netlink_autobind(sock); |
2289 | if (err) | |
2290 | goto out; | |
2291 | } | |
2292 | ||
5fd96123 | 2293 | if (netlink_tx_is_mmaped(sk) && |
c0371da6 | 2294 | msg->msg_iter.iov->iov_base == NULL) { |
5fd96123 PM |
2295 | err = netlink_mmap_sendmsg(sk, msg, dst_portid, dst_group, |
2296 | siocb); | |
2297 | goto out; | |
2298 | } | |
2299 | ||
1da177e4 LT |
2300 | err = -EMSGSIZE; |
2301 | if (len > sk->sk_sndbuf - 32) | |
2302 | goto out; | |
2303 | err = -ENOBUFS; | |
3a36515f | 2304 | skb = netlink_alloc_large_skb(len, dst_group); |
6ac552fd | 2305 | if (skb == NULL) |
1da177e4 LT |
2306 | goto out; |
2307 | ||
15e47304 | 2308 | NETLINK_CB(skb).portid = nlk->portid; |
d629b836 | 2309 | NETLINK_CB(skb).dst_group = dst_group; |
dbe9a417 | 2310 | NETLINK_CB(skb).creds = siocb->scm->creds; |
2d7a85f4 | 2311 | NETLINK_CB(skb).flags = netlink_skb_flags; |
1da177e4 | 2312 | |
1da177e4 | 2313 | err = -EFAULT; |
6ce8e9ce | 2314 | if (memcpy_from_msg(skb_put(skb, len), msg, len)) { |
1da177e4 LT |
2315 | kfree_skb(skb); |
2316 | goto out; | |
2317 | } | |
2318 | ||
2319 | err = security_netlink_send(sk, skb); | |
2320 | if (err) { | |
2321 | kfree_skb(skb); | |
2322 | goto out; | |
2323 | } | |
2324 | ||
d629b836 | 2325 | if (dst_group) { |
1da177e4 | 2326 | atomic_inc(&skb->users); |
15e47304 | 2327 | netlink_broadcast(sk, skb, dst_portid, dst_group, GFP_KERNEL); |
1da177e4 | 2328 | } |
15e47304 | 2329 | err = netlink_unicast(sk, skb, dst_portid, msg->msg_flags&MSG_DONTWAIT); |
1da177e4 LT |
2330 | |
2331 | out: | |
b47030c7 | 2332 | scm_destroy(siocb->scm); |
1da177e4 LT |
2333 | return err; |
2334 | } | |
2335 | ||
2336 | static int netlink_recvmsg(struct kiocb *kiocb, struct socket *sock, | |
2337 | struct msghdr *msg, size_t len, | |
2338 | int flags) | |
2339 | { | |
2340 | struct sock_iocb *siocb = kiocb_to_siocb(kiocb); | |
2341 | struct scm_cookie scm; | |
2342 | struct sock *sk = sock->sk; | |
2343 | struct netlink_sock *nlk = nlk_sk(sk); | |
2344 | int noblock = flags&MSG_DONTWAIT; | |
2345 | size_t copied; | |
68d6ac6d | 2346 | struct sk_buff *skb, *data_skb; |
b44d211e | 2347 | int err, ret; |
1da177e4 LT |
2348 | |
2349 | if (flags&MSG_OOB) | |
2350 | return -EOPNOTSUPP; | |
2351 | ||
2352 | copied = 0; | |
2353 | ||
6ac552fd PM |
2354 | skb = skb_recv_datagram(sk, flags, noblock, &err); |
2355 | if (skb == NULL) | |
1da177e4 LT |
2356 | goto out; |
2357 | ||
68d6ac6d JB |
2358 | data_skb = skb; |
2359 | ||
1dacc76d JB |
2360 | #ifdef CONFIG_COMPAT_NETLINK_MESSAGES |
2361 | if (unlikely(skb_shinfo(skb)->frag_list)) { | |
1dacc76d | 2362 | /* |
68d6ac6d JB |
2363 | * If this skb has a frag_list, then here that means that we |
2364 | * will have to use the frag_list skb's data for compat tasks | |
2365 | * and the regular skb's data for normal (non-compat) tasks. | |
1dacc76d | 2366 | * |
68d6ac6d JB |
2367 | * If we need to send the compat skb, assign it to the |
2368 | * 'data_skb' variable so that it will be used below for data | |
2369 | * copying. We keep 'skb' for everything else, including | |
2370 | * freeing both later. | |
1dacc76d | 2371 | */ |
68d6ac6d JB |
2372 | if (flags & MSG_CMSG_COMPAT) |
2373 | data_skb = skb_shinfo(skb)->frag_list; | |
1dacc76d JB |
2374 | } |
2375 | #endif | |
2376 | ||
9063e21f ED |
2377 | /* Record the max length of recvmsg() calls for future allocations */ |
2378 | nlk->max_recvmsg_len = max(nlk->max_recvmsg_len, len); | |
2379 | nlk->max_recvmsg_len = min_t(size_t, nlk->max_recvmsg_len, | |
2380 | 16384); | |
2381 | ||
68d6ac6d | 2382 | copied = data_skb->len; |
1da177e4 LT |
2383 | if (len < copied) { |
2384 | msg->msg_flags |= MSG_TRUNC; | |
2385 | copied = len; | |
2386 | } | |
2387 | ||
68d6ac6d | 2388 | skb_reset_transport_header(data_skb); |
51f3d02b | 2389 | err = skb_copy_datagram_msg(data_skb, 0, msg, copied); |
1da177e4 LT |
2390 | |
2391 | if (msg->msg_name) { | |
342dfc30 | 2392 | DECLARE_SOCKADDR(struct sockaddr_nl *, addr, msg->msg_name); |
1da177e4 LT |
2393 | addr->nl_family = AF_NETLINK; |
2394 | addr->nl_pad = 0; | |
15e47304 | 2395 | addr->nl_pid = NETLINK_CB(skb).portid; |
d629b836 | 2396 | addr->nl_groups = netlink_group_mask(NETLINK_CB(skb).dst_group); |
1da177e4 LT |
2397 | msg->msg_namelen = sizeof(*addr); |
2398 | } | |
2399 | ||
cc9a06cd PM |
2400 | if (nlk->flags & NETLINK_RECV_PKTINFO) |
2401 | netlink_cmsg_recv_pktinfo(msg, skb); | |
2402 | ||
1da177e4 LT |
2403 | if (NULL == siocb->scm) { |
2404 | memset(&scm, 0, sizeof(scm)); | |
2405 | siocb->scm = &scm; | |
2406 | } | |
2407 | siocb->scm->creds = *NETLINK_CREDS(skb); | |
188ccb55 | 2408 | if (flags & MSG_TRUNC) |
68d6ac6d | 2409 | copied = data_skb->len; |
daa3766e | 2410 | |
1da177e4 LT |
2411 | skb_free_datagram(sk, skb); |
2412 | ||
16b304f3 PS |
2413 | if (nlk->cb_running && |
2414 | atomic_read(&sk->sk_rmem_alloc) <= sk->sk_rcvbuf / 2) { | |
b44d211e AV |
2415 | ret = netlink_dump(sk); |
2416 | if (ret) { | |
ac30ef83 | 2417 | sk->sk_err = -ret; |
b44d211e AV |
2418 | sk->sk_error_report(sk); |
2419 | } | |
2420 | } | |
1da177e4 LT |
2421 | |
2422 | scm_recv(sock, msg, siocb->scm, flags); | |
1da177e4 LT |
2423 | out: |
2424 | netlink_rcv_wake(sk); | |
2425 | return err ? : copied; | |
2426 | } | |
2427 | ||
676d2369 | 2428 | static void netlink_data_ready(struct sock *sk) |
1da177e4 | 2429 | { |
cd40b7d3 | 2430 | BUG(); |
1da177e4 LT |
2431 | } |
2432 | ||
2433 | /* | |
746fac4d | 2434 | * We export these functions to other modules. They provide a |
1da177e4 LT |
2435 | * complete set of kernel non-blocking support for message |
2436 | * queueing. | |
2437 | */ | |
2438 | ||
2439 | struct sock * | |
9f00d977 PNA |
2440 | __netlink_kernel_create(struct net *net, int unit, struct module *module, |
2441 | struct netlink_kernel_cfg *cfg) | |
1da177e4 LT |
2442 | { |
2443 | struct socket *sock; | |
2444 | struct sock *sk; | |
77247bbb | 2445 | struct netlink_sock *nlk; |
5c398dc8 | 2446 | struct listeners *listeners = NULL; |
a31f2d17 PNA |
2447 | struct mutex *cb_mutex = cfg ? cfg->cb_mutex : NULL; |
2448 | unsigned int groups; | |
1da177e4 | 2449 | |
fab2caf6 | 2450 | BUG_ON(!nl_table); |
1da177e4 | 2451 | |
6ac552fd | 2452 | if (unit < 0 || unit >= MAX_LINKS) |
1da177e4 LT |
2453 | return NULL; |
2454 | ||
2455 | if (sock_create_lite(PF_NETLINK, SOCK_DGRAM, unit, &sock)) | |
2456 | return NULL; | |
2457 | ||
23fe1866 PE |
2458 | /* |
2459 | * We have to just have a reference on the net from sk, but don't | |
2460 | * get_net it. Besides, we cannot get and then put the net here. | |
2461 | * So we create one inside init_net and the move it to net. | |
2462 | */ | |
2463 | ||
2464 | if (__netlink_create(&init_net, sock, cb_mutex, unit) < 0) | |
2465 | goto out_sock_release_nosk; | |
2466 | ||
2467 | sk = sock->sk; | |
edf02087 | 2468 | sk_change_net(sk, net); |
4fdb3bb7 | 2469 | |
a31f2d17 | 2470 | if (!cfg || cfg->groups < 32) |
4277a083 | 2471 | groups = 32; |
a31f2d17 PNA |
2472 | else |
2473 | groups = cfg->groups; | |
4277a083 | 2474 | |
5c398dc8 | 2475 | listeners = kzalloc(sizeof(*listeners) + NLGRPSZ(groups), GFP_KERNEL); |
4277a083 PM |
2476 | if (!listeners) |
2477 | goto out_sock_release; | |
2478 | ||
1da177e4 | 2479 | sk->sk_data_ready = netlink_data_ready; |
a31f2d17 PNA |
2480 | if (cfg && cfg->input) |
2481 | nlk_sk(sk)->netlink_rcv = cfg->input; | |
1da177e4 | 2482 | |
b4b51029 | 2483 | if (netlink_insert(sk, net, 0)) |
77247bbb | 2484 | goto out_sock_release; |
4fdb3bb7 | 2485 | |
77247bbb PM |
2486 | nlk = nlk_sk(sk); |
2487 | nlk->flags |= NETLINK_KERNEL_SOCKET; | |
4fdb3bb7 | 2488 | |
4fdb3bb7 | 2489 | netlink_table_grab(); |
b4b51029 EB |
2490 | if (!nl_table[unit].registered) { |
2491 | nl_table[unit].groups = groups; | |
5c398dc8 | 2492 | rcu_assign_pointer(nl_table[unit].listeners, listeners); |
b4b51029 EB |
2493 | nl_table[unit].cb_mutex = cb_mutex; |
2494 | nl_table[unit].module = module; | |
9785e10a PNA |
2495 | if (cfg) { |
2496 | nl_table[unit].bind = cfg->bind; | |
6251edd9 | 2497 | nl_table[unit].unbind = cfg->unbind; |
9785e10a | 2498 | nl_table[unit].flags = cfg->flags; |
da12c90e G |
2499 | if (cfg->compare) |
2500 | nl_table[unit].compare = cfg->compare; | |
9785e10a | 2501 | } |
b4b51029 | 2502 | nl_table[unit].registered = 1; |
f937f1f4 JJ |
2503 | } else { |
2504 | kfree(listeners); | |
869e58f8 | 2505 | nl_table[unit].registered++; |
b4b51029 | 2506 | } |
4fdb3bb7 | 2507 | netlink_table_ungrab(); |
77247bbb PM |
2508 | return sk; |
2509 | ||
4fdb3bb7 | 2510 | out_sock_release: |
4277a083 | 2511 | kfree(listeners); |
9dfbec1f | 2512 | netlink_kernel_release(sk); |
23fe1866 PE |
2513 | return NULL; |
2514 | ||
2515 | out_sock_release_nosk: | |
4fdb3bb7 | 2516 | sock_release(sock); |
77247bbb | 2517 | return NULL; |
1da177e4 | 2518 | } |
9f00d977 | 2519 | EXPORT_SYMBOL(__netlink_kernel_create); |
b7c6ba6e DL |
2520 | |
2521 | void | |
2522 | netlink_kernel_release(struct sock *sk) | |
2523 | { | |
edf02087 | 2524 | sk_release_kernel(sk); |
b7c6ba6e DL |
2525 | } |
2526 | EXPORT_SYMBOL(netlink_kernel_release); | |
2527 | ||
d136f1bd | 2528 | int __netlink_change_ngroups(struct sock *sk, unsigned int groups) |
b4ff4f04 | 2529 | { |
5c398dc8 | 2530 | struct listeners *new, *old; |
b4ff4f04 | 2531 | struct netlink_table *tbl = &nl_table[sk->sk_protocol]; |
b4ff4f04 JB |
2532 | |
2533 | if (groups < 32) | |
2534 | groups = 32; | |
2535 | ||
b4ff4f04 | 2536 | if (NLGRPSZ(tbl->groups) < NLGRPSZ(groups)) { |
5c398dc8 ED |
2537 | new = kzalloc(sizeof(*new) + NLGRPSZ(groups), GFP_ATOMIC); |
2538 | if (!new) | |
d136f1bd | 2539 | return -ENOMEM; |
6d772ac5 | 2540 | old = nl_deref_protected(tbl->listeners); |
5c398dc8 ED |
2541 | memcpy(new->masks, old->masks, NLGRPSZ(tbl->groups)); |
2542 | rcu_assign_pointer(tbl->listeners, new); | |
2543 | ||
37b6b935 | 2544 | kfree_rcu(old, rcu); |
b4ff4f04 JB |
2545 | } |
2546 | tbl->groups = groups; | |
2547 | ||
d136f1bd JB |
2548 | return 0; |
2549 | } | |
2550 | ||
2551 | /** | |
2552 | * netlink_change_ngroups - change number of multicast groups | |
2553 | * | |
2554 | * This changes the number of multicast groups that are available | |
2555 | * on a certain netlink family. Note that it is not possible to | |
2556 | * change the number of groups to below 32. Also note that it does | |
2557 | * not implicitly call netlink_clear_multicast_users() when the | |
2558 | * number of groups is reduced. | |
2559 | * | |
2560 | * @sk: The kernel netlink socket, as returned by netlink_kernel_create(). | |
2561 | * @groups: The new number of groups. | |
2562 | */ | |
2563 | int netlink_change_ngroups(struct sock *sk, unsigned int groups) | |
2564 | { | |
2565 | int err; | |
2566 | ||
2567 | netlink_table_grab(); | |
2568 | err = __netlink_change_ngroups(sk, groups); | |
b4ff4f04 | 2569 | netlink_table_ungrab(); |
d136f1bd | 2570 | |
b4ff4f04 JB |
2571 | return err; |
2572 | } | |
b4ff4f04 | 2573 | |
b8273570 JB |
2574 | void __netlink_clear_multicast_users(struct sock *ksk, unsigned int group) |
2575 | { | |
2576 | struct sock *sk; | |
b8273570 JB |
2577 | struct netlink_table *tbl = &nl_table[ksk->sk_protocol]; |
2578 | ||
b67bfe0d | 2579 | sk_for_each_bound(sk, &tbl->mc_list) |
b8273570 JB |
2580 | netlink_update_socket_mc(nlk_sk(sk), group, 0); |
2581 | } | |
2582 | ||
a46621a3 | 2583 | struct nlmsghdr * |
15e47304 | 2584 | __nlmsg_put(struct sk_buff *skb, u32 portid, u32 seq, int type, int len, int flags) |
a46621a3 DV |
2585 | { |
2586 | struct nlmsghdr *nlh; | |
573ce260 | 2587 | int size = nlmsg_msg_size(len); |
a46621a3 | 2588 | |
23b45672 | 2589 | nlh = (struct nlmsghdr *)skb_put(skb, NLMSG_ALIGN(size)); |
a46621a3 DV |
2590 | nlh->nlmsg_type = type; |
2591 | nlh->nlmsg_len = size; | |
2592 | nlh->nlmsg_flags = flags; | |
15e47304 | 2593 | nlh->nlmsg_pid = portid; |
a46621a3 DV |
2594 | nlh->nlmsg_seq = seq; |
2595 | if (!__builtin_constant_p(size) || NLMSG_ALIGN(size) - size != 0) | |
573ce260 | 2596 | memset(nlmsg_data(nlh) + len, 0, NLMSG_ALIGN(size) - size); |
a46621a3 DV |
2597 | return nlh; |
2598 | } | |
2599 | EXPORT_SYMBOL(__nlmsg_put); | |
2600 | ||
1da177e4 LT |
2601 | /* |
2602 | * It looks a bit ugly. | |
2603 | * It would be better to create kernel thread. | |
2604 | */ | |
2605 | ||
2606 | static int netlink_dump(struct sock *sk) | |
2607 | { | |
2608 | struct netlink_sock *nlk = nlk_sk(sk); | |
2609 | struct netlink_callback *cb; | |
c7ac8679 | 2610 | struct sk_buff *skb = NULL; |
1da177e4 | 2611 | struct nlmsghdr *nlh; |
bf8b79e4 | 2612 | int len, err = -ENOBUFS; |
c7ac8679 | 2613 | int alloc_size; |
1da177e4 | 2614 | |
af65bdfc | 2615 | mutex_lock(nlk->cb_mutex); |
16b304f3 | 2616 | if (!nlk->cb_running) { |
bf8b79e4 TG |
2617 | err = -EINVAL; |
2618 | goto errout_skb; | |
1da177e4 LT |
2619 | } |
2620 | ||
16b304f3 | 2621 | cb = &nlk->cb; |
c7ac8679 GR |
2622 | alloc_size = max_t(int, cb->min_dump_alloc, NLMSG_GOODSIZE); |
2623 | ||
f9c22888 PM |
2624 | if (!netlink_rx_is_mmaped(sk) && |
2625 | atomic_read(&sk->sk_rmem_alloc) >= sk->sk_rcvbuf) | |
2626 | goto errout_skb; | |
9063e21f ED |
2627 | |
2628 | /* NLMSG_GOODSIZE is small to avoid high order allocations being | |
2629 | * required, but it makes sense to _attempt_ a 16K bytes allocation | |
2630 | * to reduce number of system calls on dump operations, if user | |
2631 | * ever provided a big enough buffer. | |
2632 | */ | |
2633 | if (alloc_size < nlk->max_recvmsg_len) { | |
2634 | skb = netlink_alloc_skb(sk, | |
2635 | nlk->max_recvmsg_len, | |
2636 | nlk->portid, | |
2637 | GFP_KERNEL | | |
2638 | __GFP_NOWARN | | |
2639 | __GFP_NORETRY); | |
2640 | /* available room should be exact amount to avoid MSG_TRUNC */ | |
2641 | if (skb) | |
2642 | skb_reserve(skb, skb_tailroom(skb) - | |
2643 | nlk->max_recvmsg_len); | |
2644 | } | |
2645 | if (!skb) | |
2646 | skb = netlink_alloc_skb(sk, alloc_size, nlk->portid, | |
2647 | GFP_KERNEL); | |
c7ac8679 | 2648 | if (!skb) |
c63d6ea3 | 2649 | goto errout_skb; |
f9c22888 | 2650 | netlink_skb_set_owner_r(skb, sk); |
c7ac8679 | 2651 | |
1da177e4 LT |
2652 | len = cb->dump(skb, cb); |
2653 | ||
2654 | if (len > 0) { | |
af65bdfc | 2655 | mutex_unlock(nlk->cb_mutex); |
b1153f29 SH |
2656 | |
2657 | if (sk_filter(sk, skb)) | |
2658 | kfree_skb(skb); | |
4a7e7c2a ED |
2659 | else |
2660 | __netlink_sendskb(sk, skb); | |
1da177e4 LT |
2661 | return 0; |
2662 | } | |
2663 | ||
bf8b79e4 TG |
2664 | nlh = nlmsg_put_answer(skb, cb, NLMSG_DONE, sizeof(len), NLM_F_MULTI); |
2665 | if (!nlh) | |
2666 | goto errout_skb; | |
2667 | ||
670dc283 JB |
2668 | nl_dump_check_consistent(cb, nlh); |
2669 | ||
bf8b79e4 TG |
2670 | memcpy(nlmsg_data(nlh), &len, sizeof(len)); |
2671 | ||
b1153f29 SH |
2672 | if (sk_filter(sk, skb)) |
2673 | kfree_skb(skb); | |
4a7e7c2a ED |
2674 | else |
2675 | __netlink_sendskb(sk, skb); | |
1da177e4 | 2676 | |
a8f74b22 TG |
2677 | if (cb->done) |
2678 | cb->done(cb); | |
1da177e4 | 2679 | |
16b304f3 PS |
2680 | nlk->cb_running = false; |
2681 | mutex_unlock(nlk->cb_mutex); | |
6dc878a8 | 2682 | module_put(cb->module); |
16b304f3 | 2683 | consume_skb(cb->skb); |
1da177e4 | 2684 | return 0; |
1797754e | 2685 | |
bf8b79e4 | 2686 | errout_skb: |
af65bdfc | 2687 | mutex_unlock(nlk->cb_mutex); |
bf8b79e4 | 2688 | kfree_skb(skb); |
bf8b79e4 | 2689 | return err; |
1da177e4 LT |
2690 | } |
2691 | ||
6dc878a8 G |
2692 | int __netlink_dump_start(struct sock *ssk, struct sk_buff *skb, |
2693 | const struct nlmsghdr *nlh, | |
2694 | struct netlink_dump_control *control) | |
1da177e4 LT |
2695 | { |
2696 | struct netlink_callback *cb; | |
2697 | struct sock *sk; | |
2698 | struct netlink_sock *nlk; | |
b44d211e | 2699 | int ret; |
1da177e4 | 2700 | |
f9c22888 PM |
2701 | /* Memory mapped dump requests need to be copied to avoid looping |
2702 | * on the pending state in netlink_mmap_sendmsg() while the CB hold | |
2703 | * a reference to the skb. | |
2704 | */ | |
2705 | if (netlink_skb_is_mmaped(skb)) { | |
2706 | skb = skb_copy(skb, GFP_KERNEL); | |
16b304f3 | 2707 | if (skb == NULL) |
f9c22888 | 2708 | return -ENOBUFS; |
f9c22888 PM |
2709 | } else |
2710 | atomic_inc(&skb->users); | |
2711 | ||
15e47304 | 2712 | sk = netlink_lookup(sock_net(ssk), ssk->sk_protocol, NETLINK_CB(skb).portid); |
1da177e4 | 2713 | if (sk == NULL) { |
16b304f3 PS |
2714 | ret = -ECONNREFUSED; |
2715 | goto error_free; | |
1da177e4 | 2716 | } |
6dc878a8 | 2717 | |
16b304f3 | 2718 | nlk = nlk_sk(sk); |
af65bdfc | 2719 | mutex_lock(nlk->cb_mutex); |
6dc878a8 | 2720 | /* A dump is in progress... */ |
16b304f3 | 2721 | if (nlk->cb_running) { |
6dc878a8 | 2722 | ret = -EBUSY; |
16b304f3 | 2723 | goto error_unlock; |
1da177e4 | 2724 | } |
6dc878a8 | 2725 | /* add reference of module which cb->dump belongs to */ |
16b304f3 | 2726 | if (!try_module_get(control->module)) { |
6dc878a8 | 2727 | ret = -EPROTONOSUPPORT; |
16b304f3 | 2728 | goto error_unlock; |
6dc878a8 G |
2729 | } |
2730 | ||
16b304f3 PS |
2731 | cb = &nlk->cb; |
2732 | memset(cb, 0, sizeof(*cb)); | |
2733 | cb->dump = control->dump; | |
2734 | cb->done = control->done; | |
2735 | cb->nlh = nlh; | |
2736 | cb->data = control->data; | |
2737 | cb->module = control->module; | |
2738 | cb->min_dump_alloc = control->min_dump_alloc; | |
2739 | cb->skb = skb; | |
2740 | ||
2741 | nlk->cb_running = true; | |
2742 | ||
af65bdfc | 2743 | mutex_unlock(nlk->cb_mutex); |
1da177e4 | 2744 | |
b44d211e | 2745 | ret = netlink_dump(sk); |
1da177e4 | 2746 | sock_put(sk); |
5c58298c | 2747 | |
b44d211e AV |
2748 | if (ret) |
2749 | return ret; | |
2750 | ||
5c58298c DL |
2751 | /* We successfully started a dump, by returning -EINTR we |
2752 | * signal not to send ACK even if it was requested. | |
2753 | */ | |
2754 | return -EINTR; | |
16b304f3 PS |
2755 | |
2756 | error_unlock: | |
2757 | sock_put(sk); | |
2758 | mutex_unlock(nlk->cb_mutex); | |
2759 | error_free: | |
2760 | kfree_skb(skb); | |
2761 | return ret; | |
1da177e4 | 2762 | } |
6dc878a8 | 2763 | EXPORT_SYMBOL(__netlink_dump_start); |
1da177e4 LT |
2764 | |
2765 | void netlink_ack(struct sk_buff *in_skb, struct nlmsghdr *nlh, int err) | |
2766 | { | |
2767 | struct sk_buff *skb; | |
2768 | struct nlmsghdr *rep; | |
2769 | struct nlmsgerr *errmsg; | |
339bf98f | 2770 | size_t payload = sizeof(*errmsg); |
1da177e4 | 2771 | |
339bf98f TG |
2772 | /* error messages get the original request appened */ |
2773 | if (err) | |
2774 | payload += nlmsg_len(nlh); | |
1da177e4 | 2775 | |
f9c22888 PM |
2776 | skb = netlink_alloc_skb(in_skb->sk, nlmsg_total_size(payload), |
2777 | NETLINK_CB(in_skb).portid, GFP_KERNEL); | |
1da177e4 LT |
2778 | if (!skb) { |
2779 | struct sock *sk; | |
2780 | ||
3b1e0a65 | 2781 | sk = netlink_lookup(sock_net(in_skb->sk), |
b4b51029 | 2782 | in_skb->sk->sk_protocol, |
15e47304 | 2783 | NETLINK_CB(in_skb).portid); |
1da177e4 LT |
2784 | if (sk) { |
2785 | sk->sk_err = ENOBUFS; | |
2786 | sk->sk_error_report(sk); | |
2787 | sock_put(sk); | |
2788 | } | |
2789 | return; | |
2790 | } | |
2791 | ||
15e47304 | 2792 | rep = __nlmsg_put(skb, NETLINK_CB(in_skb).portid, nlh->nlmsg_seq, |
5dba93ae | 2793 | NLMSG_ERROR, payload, 0); |
bf8b79e4 | 2794 | errmsg = nlmsg_data(rep); |
1da177e4 | 2795 | errmsg->error = err; |
bf8b79e4 | 2796 | memcpy(&errmsg->msg, nlh, err ? nlh->nlmsg_len : sizeof(*nlh)); |
15e47304 | 2797 | netlink_unicast(in_skb->sk, skb, NETLINK_CB(in_skb).portid, MSG_DONTWAIT); |
1da177e4 | 2798 | } |
6ac552fd | 2799 | EXPORT_SYMBOL(netlink_ack); |
1da177e4 | 2800 | |
cd40b7d3 | 2801 | int netlink_rcv_skb(struct sk_buff *skb, int (*cb)(struct sk_buff *, |
1d00a4eb | 2802 | struct nlmsghdr *)) |
82ace47a | 2803 | { |
82ace47a TG |
2804 | struct nlmsghdr *nlh; |
2805 | int err; | |
2806 | ||
2807 | while (skb->len >= nlmsg_total_size(0)) { | |
cd40b7d3 DL |
2808 | int msglen; |
2809 | ||
b529ccf2 | 2810 | nlh = nlmsg_hdr(skb); |
d35b6856 | 2811 | err = 0; |
82ace47a | 2812 | |
ad8e4b75 | 2813 | if (nlh->nlmsg_len < NLMSG_HDRLEN || skb->len < nlh->nlmsg_len) |
82ace47a TG |
2814 | return 0; |
2815 | ||
d35b6856 TG |
2816 | /* Only requests are handled by the kernel */ |
2817 | if (!(nlh->nlmsg_flags & NLM_F_REQUEST)) | |
5c58298c | 2818 | goto ack; |
45e7ae7f TG |
2819 | |
2820 | /* Skip control messages */ | |
2821 | if (nlh->nlmsg_type < NLMSG_MIN_TYPE) | |
5c58298c | 2822 | goto ack; |
d35b6856 | 2823 | |
1d00a4eb | 2824 | err = cb(skb, nlh); |
5c58298c DL |
2825 | if (err == -EINTR) |
2826 | goto skip; | |
2827 | ||
2828 | ack: | |
d35b6856 | 2829 | if (nlh->nlmsg_flags & NLM_F_ACK || err) |
82ace47a | 2830 | netlink_ack(skb, nlh, err); |
82ace47a | 2831 | |
5c58298c | 2832 | skip: |
6ac552fd | 2833 | msglen = NLMSG_ALIGN(nlh->nlmsg_len); |
cd40b7d3 DL |
2834 | if (msglen > skb->len) |
2835 | msglen = skb->len; | |
2836 | skb_pull(skb, msglen); | |
82ace47a TG |
2837 | } |
2838 | ||
2839 | return 0; | |
2840 | } | |
6ac552fd | 2841 | EXPORT_SYMBOL(netlink_rcv_skb); |
82ace47a | 2842 | |
d387f6ad TG |
2843 | /** |
2844 | * nlmsg_notify - send a notification netlink message | |
2845 | * @sk: netlink socket to use | |
2846 | * @skb: notification message | |
15e47304 | 2847 | * @portid: destination netlink portid for reports or 0 |
d387f6ad TG |
2848 | * @group: destination multicast group or 0 |
2849 | * @report: 1 to report back, 0 to disable | |
2850 | * @flags: allocation flags | |
2851 | */ | |
15e47304 | 2852 | int nlmsg_notify(struct sock *sk, struct sk_buff *skb, u32 portid, |
d387f6ad TG |
2853 | unsigned int group, int report, gfp_t flags) |
2854 | { | |
2855 | int err = 0; | |
2856 | ||
2857 | if (group) { | |
15e47304 | 2858 | int exclude_portid = 0; |
d387f6ad TG |
2859 | |
2860 | if (report) { | |
2861 | atomic_inc(&skb->users); | |
15e47304 | 2862 | exclude_portid = portid; |
d387f6ad TG |
2863 | } |
2864 | ||
1ce85fe4 PNA |
2865 | /* errors reported via destination sk->sk_err, but propagate |
2866 | * delivery errors if NETLINK_BROADCAST_ERROR flag is set */ | |
15e47304 | 2867 | err = nlmsg_multicast(sk, skb, exclude_portid, group, flags); |
d387f6ad TG |
2868 | } |
2869 | ||
1ce85fe4 PNA |
2870 | if (report) { |
2871 | int err2; | |
2872 | ||
15e47304 | 2873 | err2 = nlmsg_unicast(sk, skb, portid); |
1ce85fe4 PNA |
2874 | if (!err || err == -ESRCH) |
2875 | err = err2; | |
2876 | } | |
d387f6ad TG |
2877 | |
2878 | return err; | |
2879 | } | |
6ac552fd | 2880 | EXPORT_SYMBOL(nlmsg_notify); |
d387f6ad | 2881 | |
1da177e4 LT |
2882 | #ifdef CONFIG_PROC_FS |
2883 | struct nl_seq_iter { | |
e372c414 | 2884 | struct seq_net_private p; |
1da177e4 LT |
2885 | int link; |
2886 | int hash_idx; | |
2887 | }; | |
2888 | ||
2889 | static struct sock *netlink_seq_socket_idx(struct seq_file *seq, loff_t pos) | |
2890 | { | |
2891 | struct nl_seq_iter *iter = seq->private; | |
2892 | int i, j; | |
e341694e | 2893 | struct netlink_sock *nlk; |
1da177e4 | 2894 | struct sock *s; |
1da177e4 LT |
2895 | loff_t off = 0; |
2896 | ||
6ac552fd | 2897 | for (i = 0; i < MAX_LINKS; i++) { |
e341694e | 2898 | struct rhashtable *ht = &nl_table[i].hash; |
67a24ac1 | 2899 | const struct bucket_table *tbl = rht_dereference_rcu(ht->tbl, ht); |
e341694e TG |
2900 | |
2901 | for (j = 0; j < tbl->size; j++) { | |
88d6ed15 TG |
2902 | struct rhash_head *node; |
2903 | ||
2904 | rht_for_each_entry_rcu(nlk, node, tbl, j, node) { | |
e341694e | 2905 | s = (struct sock *)nlk; |
1da177e4 | 2906 | |
1218854a | 2907 | if (sock_net(s) != seq_file_net(seq)) |
b4b51029 | 2908 | continue; |
1da177e4 LT |
2909 | if (off == pos) { |
2910 | iter->link = i; | |
2911 | iter->hash_idx = j; | |
2912 | return s; | |
2913 | } | |
2914 | ++off; | |
2915 | } | |
2916 | } | |
2917 | } | |
2918 | return NULL; | |
2919 | } | |
2920 | ||
2921 | static void *netlink_seq_start(struct seq_file *seq, loff_t *pos) | |
21e4902a | 2922 | __acquires(RCU) |
1da177e4 | 2923 | { |
e341694e | 2924 | rcu_read_lock(); |
1da177e4 LT |
2925 | return *pos ? netlink_seq_socket_idx(seq, *pos - 1) : SEQ_START_TOKEN; |
2926 | } | |
2927 | ||
2928 | static void *netlink_seq_next(struct seq_file *seq, void *v, loff_t *pos) | |
2929 | { | |
78fd1d0a | 2930 | struct rhashtable *ht; |
88d6ed15 TG |
2931 | const struct bucket_table *tbl; |
2932 | struct rhash_head *node; | |
e341694e | 2933 | struct netlink_sock *nlk; |
1da177e4 | 2934 | struct nl_seq_iter *iter; |
da12c90e | 2935 | struct net *net; |
1da177e4 LT |
2936 | int i, j; |
2937 | ||
2938 | ++*pos; | |
2939 | ||
2940 | if (v == SEQ_START_TOKEN) | |
2941 | return netlink_seq_socket_idx(seq, 0); | |
746fac4d | 2942 | |
da12c90e | 2943 | net = seq_file_net(seq); |
b4b51029 | 2944 | iter = seq->private; |
e341694e TG |
2945 | nlk = v; |
2946 | ||
78fd1d0a TG |
2947 | i = iter->link; |
2948 | ht = &nl_table[i].hash; | |
88d6ed15 TG |
2949 | tbl = rht_dereference_rcu(ht->tbl, ht); |
2950 | rht_for_each_entry_rcu_continue(nlk, node, nlk->node.next, tbl, iter->hash_idx, node) | |
e341694e TG |
2951 | if (net_eq(sock_net((struct sock *)nlk), net)) |
2952 | return nlk; | |
1da177e4 | 2953 | |
1da177e4 LT |
2954 | j = iter->hash_idx + 1; |
2955 | ||
2956 | do { | |
da12c90e | 2957 | |
e341694e | 2958 | for (; j < tbl->size; j++) { |
88d6ed15 | 2959 | rht_for_each_entry_rcu(nlk, node, tbl, j, node) { |
e341694e TG |
2960 | if (net_eq(sock_net((struct sock *)nlk), net)) { |
2961 | iter->link = i; | |
2962 | iter->hash_idx = j; | |
2963 | return nlk; | |
2964 | } | |
1da177e4 LT |
2965 | } |
2966 | } | |
2967 | ||
2968 | j = 0; | |
2969 | } while (++i < MAX_LINKS); | |
2970 | ||
2971 | return NULL; | |
2972 | } | |
2973 | ||
2974 | static void netlink_seq_stop(struct seq_file *seq, void *v) | |
21e4902a | 2975 | __releases(RCU) |
1da177e4 | 2976 | { |
e341694e | 2977 | rcu_read_unlock(); |
1da177e4 LT |
2978 | } |
2979 | ||
2980 | ||
2981 | static int netlink_seq_show(struct seq_file *seq, void *v) | |
2982 | { | |
658cb354 | 2983 | if (v == SEQ_START_TOKEN) { |
1da177e4 LT |
2984 | seq_puts(seq, |
2985 | "sk Eth Pid Groups " | |
cf0aa4e0 | 2986 | "Rmem Wmem Dump Locks Drops Inode\n"); |
658cb354 | 2987 | } else { |
1da177e4 LT |
2988 | struct sock *s = v; |
2989 | struct netlink_sock *nlk = nlk_sk(s); | |
2990 | ||
16b304f3 | 2991 | seq_printf(seq, "%pK %-3d %-6u %08x %-8d %-8d %d %-8d %-8d %-8lu\n", |
1da177e4 LT |
2992 | s, |
2993 | s->sk_protocol, | |
15e47304 | 2994 | nlk->portid, |
513c2500 | 2995 | nlk->groups ? (u32)nlk->groups[0] : 0, |
31e6d363 ED |
2996 | sk_rmem_alloc_get(s), |
2997 | sk_wmem_alloc_get(s), | |
16b304f3 | 2998 | nlk->cb_running, |
38938bfe | 2999 | atomic_read(&s->sk_refcnt), |
cf0aa4e0 MY |
3000 | atomic_read(&s->sk_drops), |
3001 | sock_i_ino(s) | |
1da177e4 LT |
3002 | ); |
3003 | ||
3004 | } | |
3005 | return 0; | |
3006 | } | |
3007 | ||
56b3d975 | 3008 | static const struct seq_operations netlink_seq_ops = { |
1da177e4 LT |
3009 | .start = netlink_seq_start, |
3010 | .next = netlink_seq_next, | |
3011 | .stop = netlink_seq_stop, | |
3012 | .show = netlink_seq_show, | |
3013 | }; | |
3014 | ||
3015 | ||
3016 | static int netlink_seq_open(struct inode *inode, struct file *file) | |
3017 | { | |
e372c414 DL |
3018 | return seq_open_net(inode, file, &netlink_seq_ops, |
3019 | sizeof(struct nl_seq_iter)); | |
b4b51029 EB |
3020 | } |
3021 | ||
da7071d7 | 3022 | static const struct file_operations netlink_seq_fops = { |
1da177e4 LT |
3023 | .owner = THIS_MODULE, |
3024 | .open = netlink_seq_open, | |
3025 | .read = seq_read, | |
3026 | .llseek = seq_lseek, | |
e372c414 | 3027 | .release = seq_release_net, |
1da177e4 LT |
3028 | }; |
3029 | ||
3030 | #endif | |
3031 | ||
3032 | int netlink_register_notifier(struct notifier_block *nb) | |
3033 | { | |
e041c683 | 3034 | return atomic_notifier_chain_register(&netlink_chain, nb); |
1da177e4 | 3035 | } |
6ac552fd | 3036 | EXPORT_SYMBOL(netlink_register_notifier); |
1da177e4 LT |
3037 | |
3038 | int netlink_unregister_notifier(struct notifier_block *nb) | |
3039 | { | |
e041c683 | 3040 | return atomic_notifier_chain_unregister(&netlink_chain, nb); |
1da177e4 | 3041 | } |
6ac552fd | 3042 | EXPORT_SYMBOL(netlink_unregister_notifier); |
746fac4d | 3043 | |
90ddc4f0 | 3044 | static const struct proto_ops netlink_ops = { |
1da177e4 LT |
3045 | .family = PF_NETLINK, |
3046 | .owner = THIS_MODULE, | |
3047 | .release = netlink_release, | |
3048 | .bind = netlink_bind, | |
3049 | .connect = netlink_connect, | |
3050 | .socketpair = sock_no_socketpair, | |
3051 | .accept = sock_no_accept, | |
3052 | .getname = netlink_getname, | |
9652e931 | 3053 | .poll = netlink_poll, |
1da177e4 LT |
3054 | .ioctl = sock_no_ioctl, |
3055 | .listen = sock_no_listen, | |
3056 | .shutdown = sock_no_shutdown, | |
9a4595bc PM |
3057 | .setsockopt = netlink_setsockopt, |
3058 | .getsockopt = netlink_getsockopt, | |
1da177e4 LT |
3059 | .sendmsg = netlink_sendmsg, |
3060 | .recvmsg = netlink_recvmsg, | |
ccdfcc39 | 3061 | .mmap = netlink_mmap, |
1da177e4 LT |
3062 | .sendpage = sock_no_sendpage, |
3063 | }; | |
3064 | ||
ec1b4cf7 | 3065 | static const struct net_proto_family netlink_family_ops = { |
1da177e4 LT |
3066 | .family = PF_NETLINK, |
3067 | .create = netlink_create, | |
3068 | .owner = THIS_MODULE, /* for consistency 8) */ | |
3069 | }; | |
3070 | ||
4665079c | 3071 | static int __net_init netlink_net_init(struct net *net) |
b4b51029 EB |
3072 | { |
3073 | #ifdef CONFIG_PROC_FS | |
d4beaa66 | 3074 | if (!proc_create("netlink", 0, net->proc_net, &netlink_seq_fops)) |
b4b51029 EB |
3075 | return -ENOMEM; |
3076 | #endif | |
3077 | return 0; | |
3078 | } | |
3079 | ||
4665079c | 3080 | static void __net_exit netlink_net_exit(struct net *net) |
b4b51029 EB |
3081 | { |
3082 | #ifdef CONFIG_PROC_FS | |
ece31ffd | 3083 | remove_proc_entry("netlink", net->proc_net); |
b4b51029 EB |
3084 | #endif |
3085 | } | |
3086 | ||
b963ea89 DM |
3087 | static void __init netlink_add_usersock_entry(void) |
3088 | { | |
5c398dc8 | 3089 | struct listeners *listeners; |
b963ea89 DM |
3090 | int groups = 32; |
3091 | ||
5c398dc8 | 3092 | listeners = kzalloc(sizeof(*listeners) + NLGRPSZ(groups), GFP_KERNEL); |
b963ea89 | 3093 | if (!listeners) |
5c398dc8 | 3094 | panic("netlink_add_usersock_entry: Cannot allocate listeners\n"); |
b963ea89 DM |
3095 | |
3096 | netlink_table_grab(); | |
3097 | ||
3098 | nl_table[NETLINK_USERSOCK].groups = groups; | |
5c398dc8 | 3099 | rcu_assign_pointer(nl_table[NETLINK_USERSOCK].listeners, listeners); |
b963ea89 DM |
3100 | nl_table[NETLINK_USERSOCK].module = THIS_MODULE; |
3101 | nl_table[NETLINK_USERSOCK].registered = 1; | |
9785e10a | 3102 | nl_table[NETLINK_USERSOCK].flags = NL_CFG_F_NONROOT_SEND; |
b963ea89 DM |
3103 | |
3104 | netlink_table_ungrab(); | |
3105 | } | |
3106 | ||
022cbae6 | 3107 | static struct pernet_operations __net_initdata netlink_net_ops = { |
b4b51029 EB |
3108 | .init = netlink_net_init, |
3109 | .exit = netlink_net_exit, | |
3110 | }; | |
3111 | ||
1da177e4 LT |
3112 | static int __init netlink_proto_init(void) |
3113 | { | |
1da177e4 | 3114 | int i; |
1da177e4 | 3115 | int err = proto_register(&netlink_proto, 0); |
e341694e TG |
3116 | struct rhashtable_params ht_params = { |
3117 | .head_offset = offsetof(struct netlink_sock, node), | |
3118 | .key_offset = offsetof(struct netlink_sock, portid), | |
3119 | .key_len = sizeof(u32), /* portid */ | |
7f19fc5e | 3120 | .hashfn = jhash, |
e341694e TG |
3121 | .max_shift = 16, /* 64K */ |
3122 | .grow_decision = rht_grow_above_75, | |
3123 | .shrink_decision = rht_shrink_below_30, | |
e341694e | 3124 | }; |
1da177e4 LT |
3125 | |
3126 | if (err != 0) | |
3127 | goto out; | |
3128 | ||
fab25745 | 3129 | BUILD_BUG_ON(sizeof(struct netlink_skb_parms) > FIELD_SIZEOF(struct sk_buff, cb)); |
1da177e4 | 3130 | |
0da974f4 | 3131 | nl_table = kcalloc(MAX_LINKS, sizeof(*nl_table), GFP_KERNEL); |
fab2caf6 AM |
3132 | if (!nl_table) |
3133 | goto panic; | |
1da177e4 | 3134 | |
1da177e4 | 3135 | for (i = 0; i < MAX_LINKS; i++) { |
e341694e TG |
3136 | if (rhashtable_init(&nl_table[i].hash, &ht_params) < 0) { |
3137 | while (--i > 0) | |
3138 | rhashtable_destroy(&nl_table[i].hash); | |
1da177e4 | 3139 | kfree(nl_table); |
fab2caf6 | 3140 | goto panic; |
1da177e4 | 3141 | } |
1da177e4 LT |
3142 | } |
3143 | ||
bcbde0d4 DB |
3144 | INIT_LIST_HEAD(&netlink_tap_all); |
3145 | ||
b963ea89 DM |
3146 | netlink_add_usersock_entry(); |
3147 | ||
1da177e4 | 3148 | sock_register(&netlink_family_ops); |
b4b51029 | 3149 | register_pernet_subsys(&netlink_net_ops); |
746fac4d | 3150 | /* The netlink device handler may be needed early. */ |
1da177e4 LT |
3151 | rtnetlink_init(); |
3152 | out: | |
3153 | return err; | |
fab2caf6 AM |
3154 | panic: |
3155 | panic("netlink_init: Cannot allocate nl_table\n"); | |
1da177e4 LT |
3156 | } |
3157 | ||
1da177e4 | 3158 | core_initcall(netlink_proto_init); |