Commit | Line | Data |
---|---|---|
1da177e4 LT |
1 | /* |
2 | * NETLINK Kernel-user communication protocol. | |
3 | * | |
113aa838 | 4 | * Authors: Alan Cox <alan@lxorguk.ukuu.org.uk> |
1da177e4 | 5 | * Alexey Kuznetsov <kuznet@ms2.inr.ac.ru> |
cd1df525 | 6 | * Patrick McHardy <kaber@trash.net> |
1da177e4 LT |
7 | * |
8 | * This program is free software; you can redistribute it and/or | |
9 | * modify it under the terms of the GNU General Public License | |
10 | * as published by the Free Software Foundation; either version | |
11 | * 2 of the License, or (at your option) any later version. | |
746fac4d | 12 | * |
1da177e4 LT |
13 | * Tue Jun 26 14:36:48 MEST 2001 Herbert "herp" Rosmanith |
14 | * added netlink_proto_exit | |
15 | * Tue Jan 22 18:32:44 BRST 2002 Arnaldo C. de Melo <acme@conectiva.com.br> | |
16 | * use nlk_sk, as sk->protinfo is on a diet 8) | |
4fdb3bb7 HW |
17 | * Fri Jul 22 19:51:12 MEST 2005 Harald Welte <laforge@gnumonks.org> |
18 | * - inc module use count of module that owns | |
19 | * the kernel socket in case userspace opens | |
20 | * socket of same protocol | |
21 | * - remove all module support, since netlink is | |
22 | * mandatory if CONFIG_NET=y these days | |
1da177e4 LT |
23 | */ |
24 | ||
1da177e4 LT |
25 | #include <linux/module.h> |
26 | ||
4fc268d2 | 27 | #include <linux/capability.h> |
1da177e4 LT |
28 | #include <linux/kernel.h> |
29 | #include <linux/init.h> | |
1da177e4 LT |
30 | #include <linux/signal.h> |
31 | #include <linux/sched.h> | |
32 | #include <linux/errno.h> | |
33 | #include <linux/string.h> | |
34 | #include <linux/stat.h> | |
35 | #include <linux/socket.h> | |
36 | #include <linux/un.h> | |
37 | #include <linux/fcntl.h> | |
38 | #include <linux/termios.h> | |
39 | #include <linux/sockios.h> | |
40 | #include <linux/net.h> | |
41 | #include <linux/fs.h> | |
42 | #include <linux/slab.h> | |
43 | #include <asm/uaccess.h> | |
44 | #include <linux/skbuff.h> | |
45 | #include <linux/netdevice.h> | |
46 | #include <linux/rtnetlink.h> | |
47 | #include <linux/proc_fs.h> | |
48 | #include <linux/seq_file.h> | |
1da177e4 LT |
49 | #include <linux/notifier.h> |
50 | #include <linux/security.h> | |
51 | #include <linux/jhash.h> | |
52 | #include <linux/jiffies.h> | |
53 | #include <linux/random.h> | |
54 | #include <linux/bitops.h> | |
55 | #include <linux/mm.h> | |
56 | #include <linux/types.h> | |
54e0f520 | 57 | #include <linux/audit.h> |
af65bdfc | 58 | #include <linux/mutex.h> |
ccdfcc39 | 59 | #include <linux/vmalloc.h> |
bcbde0d4 | 60 | #include <linux/if_arp.h> |
e341694e | 61 | #include <linux/rhashtable.h> |
9652e931 | 62 | #include <asm/cacheflush.h> |
e341694e | 63 | #include <linux/hash.h> |
ee1c2442 | 64 | #include <linux/genetlink.h> |
54e0f520 | 65 | |
457c4cbc | 66 | #include <net/net_namespace.h> |
1da177e4 LT |
67 | #include <net/sock.h> |
68 | #include <net/scm.h> | |
82ace47a | 69 | #include <net/netlink.h> |
1da177e4 | 70 | |
0f29c768 | 71 | #include "af_netlink.h" |
1da177e4 | 72 | |
5c398dc8 ED |
73 | struct listeners { |
74 | struct rcu_head rcu; | |
75 | unsigned long masks[0]; | |
6c04bb18 JB |
76 | }; |
77 | ||
cd967e05 | 78 | /* state bits */ |
cc3a572f | 79 | #define NETLINK_S_CONGESTED 0x0 |
cd967e05 PM |
80 | |
81 | /* flags */ | |
cc3a572f ND |
82 | #define NETLINK_F_KERNEL_SOCKET 0x1 |
83 | #define NETLINK_F_RECV_PKTINFO 0x2 | |
84 | #define NETLINK_F_BROADCAST_SEND_ERROR 0x4 | |
85 | #define NETLINK_F_RECV_NO_ENOBUFS 0x8 | |
59324cf3 | 86 | #define NETLINK_F_LISTEN_ALL_NSID 0x10 |
77247bbb | 87 | |
035c4c16 | 88 | static inline int netlink_is_kernel(struct sock *sk) |
aed81560 | 89 | { |
cc3a572f | 90 | return nlk_sk(sk)->flags & NETLINK_F_KERNEL_SOCKET; |
aed81560 DL |
91 | } |
92 | ||
0f29c768 AV |
93 | struct netlink_table *nl_table; |
94 | EXPORT_SYMBOL_GPL(nl_table); | |
1da177e4 LT |
95 | |
96 | static DECLARE_WAIT_QUEUE_HEAD(nl_table_wait); | |
97 | ||
98 | static int netlink_dump(struct sock *sk); | |
9652e931 | 99 | static void netlink_skb_destructor(struct sk_buff *skb); |
1da177e4 | 100 | |
78fd1d0a | 101 | /* nl_table locking explained: |
21e4902a | 102 | * Lookup and traversal are protected with an RCU read-side lock. Insertion |
c5adde94 | 103 | * and removal are protected with per bucket lock while using RCU list |
21e4902a TG |
104 | * modification primitives and may run in parallel to RCU protected lookups. |
105 | * Destruction of the Netlink socket may only occur *after* nl_table_lock has | |
106 | * been acquired * either during or after the socket has been removed from | |
107 | * the list and after an RCU grace period. | |
78fd1d0a | 108 | */ |
0f29c768 AV |
109 | DEFINE_RWLOCK(nl_table_lock); |
110 | EXPORT_SYMBOL_GPL(nl_table_lock); | |
1da177e4 LT |
111 | static atomic_t nl_table_users = ATOMIC_INIT(0); |
112 | ||
6d772ac5 ED |
113 | #define nl_deref_protected(X) rcu_dereference_protected(X, lockdep_is_held(&nl_table_lock)); |
114 | ||
e041c683 | 115 | static ATOMIC_NOTIFIER_HEAD(netlink_chain); |
1da177e4 | 116 | |
bcbde0d4 DB |
117 | static DEFINE_SPINLOCK(netlink_tap_lock); |
118 | static struct list_head netlink_tap_all __read_mostly; | |
119 | ||
c428ecd1 HX |
120 | static const struct rhashtable_params netlink_rhashtable_params; |
121 | ||
b57ef81f | 122 | static inline u32 netlink_group_mask(u32 group) |
d629b836 PM |
123 | { |
124 | return group ? 1 << (group - 1) : 0; | |
125 | } | |
126 | ||
bcbde0d4 DB |
127 | int netlink_add_tap(struct netlink_tap *nt) |
128 | { | |
129 | if (unlikely(nt->dev->type != ARPHRD_NETLINK)) | |
130 | return -EINVAL; | |
131 | ||
132 | spin_lock(&netlink_tap_lock); | |
133 | list_add_rcu(&nt->list, &netlink_tap_all); | |
134 | spin_unlock(&netlink_tap_lock); | |
135 | ||
fcd4d35e | 136 | __module_get(nt->module); |
bcbde0d4 DB |
137 | |
138 | return 0; | |
139 | } | |
140 | EXPORT_SYMBOL_GPL(netlink_add_tap); | |
141 | ||
2173f8d9 | 142 | static int __netlink_remove_tap(struct netlink_tap *nt) |
bcbde0d4 DB |
143 | { |
144 | bool found = false; | |
145 | struct netlink_tap *tmp; | |
146 | ||
147 | spin_lock(&netlink_tap_lock); | |
148 | ||
149 | list_for_each_entry(tmp, &netlink_tap_all, list) { | |
150 | if (nt == tmp) { | |
151 | list_del_rcu(&nt->list); | |
152 | found = true; | |
153 | goto out; | |
154 | } | |
155 | } | |
156 | ||
157 | pr_warn("__netlink_remove_tap: %p not found\n", nt); | |
158 | out: | |
159 | spin_unlock(&netlink_tap_lock); | |
160 | ||
161 | if (found && nt->module) | |
162 | module_put(nt->module); | |
163 | ||
164 | return found ? 0 : -ENODEV; | |
165 | } | |
bcbde0d4 DB |
166 | |
167 | int netlink_remove_tap(struct netlink_tap *nt) | |
168 | { | |
169 | int ret; | |
170 | ||
171 | ret = __netlink_remove_tap(nt); | |
172 | synchronize_net(); | |
173 | ||
174 | return ret; | |
175 | } | |
176 | EXPORT_SYMBOL_GPL(netlink_remove_tap); | |
177 | ||
5ffd5cdd DB |
178 | static bool netlink_filter_tap(const struct sk_buff *skb) |
179 | { | |
180 | struct sock *sk = skb->sk; | |
5ffd5cdd DB |
181 | |
182 | /* We take the more conservative approach and | |
183 | * whitelist socket protocols that may pass. | |
184 | */ | |
185 | switch (sk->sk_protocol) { | |
186 | case NETLINK_ROUTE: | |
187 | case NETLINK_USERSOCK: | |
188 | case NETLINK_SOCK_DIAG: | |
189 | case NETLINK_NFLOG: | |
190 | case NETLINK_XFRM: | |
191 | case NETLINK_FIB_LOOKUP: | |
192 | case NETLINK_NETFILTER: | |
193 | case NETLINK_GENERIC: | |
498044bb | 194 | return true; |
5ffd5cdd DB |
195 | } |
196 | ||
498044bb | 197 | return false; |
5ffd5cdd DB |
198 | } |
199 | ||
bcbde0d4 DB |
200 | static int __netlink_deliver_tap_skb(struct sk_buff *skb, |
201 | struct net_device *dev) | |
202 | { | |
203 | struct sk_buff *nskb; | |
5ffd5cdd | 204 | struct sock *sk = skb->sk; |
bcbde0d4 DB |
205 | int ret = -ENOMEM; |
206 | ||
207 | dev_hold(dev); | |
208 | nskb = skb_clone(skb, GFP_ATOMIC); | |
209 | if (nskb) { | |
210 | nskb->dev = dev; | |
5ffd5cdd | 211 | nskb->protocol = htons((u16) sk->sk_protocol); |
604d13c9 DB |
212 | nskb->pkt_type = netlink_is_kernel(sk) ? |
213 | PACKET_KERNEL : PACKET_USER; | |
4e48ed88 | 214 | skb_reset_network_header(nskb); |
bcbde0d4 DB |
215 | ret = dev_queue_xmit(nskb); |
216 | if (unlikely(ret > 0)) | |
217 | ret = net_xmit_errno(ret); | |
218 | } | |
219 | ||
220 | dev_put(dev); | |
221 | return ret; | |
222 | } | |
223 | ||
224 | static void __netlink_deliver_tap(struct sk_buff *skb) | |
225 | { | |
226 | int ret; | |
227 | struct netlink_tap *tmp; | |
228 | ||
5ffd5cdd DB |
229 | if (!netlink_filter_tap(skb)) |
230 | return; | |
231 | ||
bcbde0d4 DB |
232 | list_for_each_entry_rcu(tmp, &netlink_tap_all, list) { |
233 | ret = __netlink_deliver_tap_skb(skb, tmp->dev); | |
234 | if (unlikely(ret)) | |
235 | break; | |
236 | } | |
237 | } | |
238 | ||
239 | static void netlink_deliver_tap(struct sk_buff *skb) | |
240 | { | |
241 | rcu_read_lock(); | |
242 | ||
243 | if (unlikely(!list_empty(&netlink_tap_all))) | |
244 | __netlink_deliver_tap(skb); | |
245 | ||
246 | rcu_read_unlock(); | |
247 | } | |
248 | ||
73bfd370 DB |
249 | static void netlink_deliver_tap_kernel(struct sock *dst, struct sock *src, |
250 | struct sk_buff *skb) | |
251 | { | |
252 | if (!(netlink_is_kernel(dst) && netlink_is_kernel(src))) | |
253 | netlink_deliver_tap(skb); | |
254 | } | |
255 | ||
cd1df525 PM |
256 | static void netlink_overrun(struct sock *sk) |
257 | { | |
258 | struct netlink_sock *nlk = nlk_sk(sk); | |
259 | ||
cc3a572f ND |
260 | if (!(nlk->flags & NETLINK_F_RECV_NO_ENOBUFS)) { |
261 | if (!test_and_set_bit(NETLINK_S_CONGESTED, | |
262 | &nlk_sk(sk)->state)) { | |
cd1df525 PM |
263 | sk->sk_err = ENOBUFS; |
264 | sk->sk_error_report(sk); | |
265 | } | |
266 | } | |
267 | atomic_inc(&sk->sk_drops); | |
268 | } | |
269 | ||
270 | static void netlink_rcv_wake(struct sock *sk) | |
271 | { | |
272 | struct netlink_sock *nlk = nlk_sk(sk); | |
273 | ||
274 | if (skb_queue_empty(&sk->sk_receive_queue)) | |
cc3a572f ND |
275 | clear_bit(NETLINK_S_CONGESTED, &nlk->state); |
276 | if (!test_bit(NETLINK_S_CONGESTED, &nlk->state)) | |
cd1df525 PM |
277 | wake_up_interruptible(&nlk->wait); |
278 | } | |
279 | ||
ccdfcc39 | 280 | #ifdef CONFIG_NETLINK_MMAP |
9652e931 PM |
281 | static bool netlink_skb_is_mmaped(const struct sk_buff *skb) |
282 | { | |
283 | return NETLINK_CB(skb).flags & NETLINK_SKB_MMAPED; | |
284 | } | |
285 | ||
f9c22888 PM |
286 | static bool netlink_rx_is_mmaped(struct sock *sk) |
287 | { | |
288 | return nlk_sk(sk)->rx_ring.pg_vec != NULL; | |
289 | } | |
290 | ||
5fd96123 PM |
291 | static bool netlink_tx_is_mmaped(struct sock *sk) |
292 | { | |
293 | return nlk_sk(sk)->tx_ring.pg_vec != NULL; | |
294 | } | |
295 | ||
ccdfcc39 PM |
296 | static __pure struct page *pgvec_to_page(const void *addr) |
297 | { | |
298 | if (is_vmalloc_addr(addr)) | |
299 | return vmalloc_to_page(addr); | |
300 | else | |
301 | return virt_to_page(addr); | |
302 | } | |
303 | ||
304 | static void free_pg_vec(void **pg_vec, unsigned int order, unsigned int len) | |
305 | { | |
306 | unsigned int i; | |
307 | ||
308 | for (i = 0; i < len; i++) { | |
309 | if (pg_vec[i] != NULL) { | |
310 | if (is_vmalloc_addr(pg_vec[i])) | |
311 | vfree(pg_vec[i]); | |
312 | else | |
313 | free_pages((unsigned long)pg_vec[i], order); | |
314 | } | |
315 | } | |
316 | kfree(pg_vec); | |
317 | } | |
318 | ||
319 | static void *alloc_one_pg_vec_page(unsigned long order) | |
320 | { | |
321 | void *buffer; | |
322 | gfp_t gfp_flags = GFP_KERNEL | __GFP_COMP | __GFP_ZERO | | |
323 | __GFP_NOWARN | __GFP_NORETRY; | |
324 | ||
325 | buffer = (void *)__get_free_pages(gfp_flags, order); | |
326 | if (buffer != NULL) | |
327 | return buffer; | |
328 | ||
329 | buffer = vzalloc((1 << order) * PAGE_SIZE); | |
330 | if (buffer != NULL) | |
331 | return buffer; | |
332 | ||
333 | gfp_flags &= ~__GFP_NORETRY; | |
334 | return (void *)__get_free_pages(gfp_flags, order); | |
335 | } | |
336 | ||
337 | static void **alloc_pg_vec(struct netlink_sock *nlk, | |
338 | struct nl_mmap_req *req, unsigned int order) | |
339 | { | |
340 | unsigned int block_nr = req->nm_block_nr; | |
341 | unsigned int i; | |
8a849bb7 | 342 | void **pg_vec; |
ccdfcc39 PM |
343 | |
344 | pg_vec = kcalloc(block_nr, sizeof(void *), GFP_KERNEL); | |
345 | if (pg_vec == NULL) | |
346 | return NULL; | |
347 | ||
348 | for (i = 0; i < block_nr; i++) { | |
8a849bb7 | 349 | pg_vec[i] = alloc_one_pg_vec_page(order); |
ccdfcc39 PM |
350 | if (pg_vec[i] == NULL) |
351 | goto err1; | |
352 | } | |
353 | ||
354 | return pg_vec; | |
355 | err1: | |
356 | free_pg_vec(pg_vec, order, block_nr); | |
357 | return NULL; | |
358 | } | |
359 | ||
360 | static int netlink_set_ring(struct sock *sk, struct nl_mmap_req *req, | |
361 | bool closing, bool tx_ring) | |
362 | { | |
363 | struct netlink_sock *nlk = nlk_sk(sk); | |
364 | struct netlink_ring *ring; | |
365 | struct sk_buff_head *queue; | |
366 | void **pg_vec = NULL; | |
367 | unsigned int order = 0; | |
368 | int err; | |
369 | ||
370 | ring = tx_ring ? &nlk->tx_ring : &nlk->rx_ring; | |
371 | queue = tx_ring ? &sk->sk_write_queue : &sk->sk_receive_queue; | |
372 | ||
373 | if (!closing) { | |
374 | if (atomic_read(&nlk->mapped)) | |
375 | return -EBUSY; | |
376 | if (atomic_read(&ring->pending)) | |
377 | return -EBUSY; | |
378 | } | |
379 | ||
380 | if (req->nm_block_nr) { | |
381 | if (ring->pg_vec != NULL) | |
382 | return -EBUSY; | |
383 | ||
384 | if ((int)req->nm_block_size <= 0) | |
385 | return -EINVAL; | |
74e83b23 | 386 | if (!PAGE_ALIGNED(req->nm_block_size)) |
ccdfcc39 PM |
387 | return -EINVAL; |
388 | if (req->nm_frame_size < NL_MMAP_HDRLEN) | |
389 | return -EINVAL; | |
390 | if (!IS_ALIGNED(req->nm_frame_size, NL_MMAP_MSG_ALIGNMENT)) | |
391 | return -EINVAL; | |
392 | ||
393 | ring->frames_per_block = req->nm_block_size / | |
394 | req->nm_frame_size; | |
395 | if (ring->frames_per_block == 0) | |
396 | return -EINVAL; | |
397 | if (ring->frames_per_block * req->nm_block_nr != | |
398 | req->nm_frame_nr) | |
399 | return -EINVAL; | |
400 | ||
401 | order = get_order(req->nm_block_size); | |
402 | pg_vec = alloc_pg_vec(nlk, req, order); | |
403 | if (pg_vec == NULL) | |
404 | return -ENOMEM; | |
405 | } else { | |
406 | if (req->nm_frame_nr) | |
407 | return -EINVAL; | |
408 | } | |
409 | ||
410 | err = -EBUSY; | |
411 | mutex_lock(&nlk->pg_vec_lock); | |
412 | if (closing || atomic_read(&nlk->mapped) == 0) { | |
413 | err = 0; | |
414 | spin_lock_bh(&queue->lock); | |
415 | ||
416 | ring->frame_max = req->nm_frame_nr - 1; | |
417 | ring->head = 0; | |
418 | ring->frame_size = req->nm_frame_size; | |
419 | ring->pg_vec_pages = req->nm_block_size / PAGE_SIZE; | |
420 | ||
421 | swap(ring->pg_vec_len, req->nm_block_nr); | |
422 | swap(ring->pg_vec_order, order); | |
423 | swap(ring->pg_vec, pg_vec); | |
424 | ||
425 | __skb_queue_purge(queue); | |
426 | spin_unlock_bh(&queue->lock); | |
427 | ||
428 | WARN_ON(atomic_read(&nlk->mapped)); | |
429 | } | |
430 | mutex_unlock(&nlk->pg_vec_lock); | |
431 | ||
432 | if (pg_vec) | |
433 | free_pg_vec(pg_vec, order, req->nm_block_nr); | |
434 | return err; | |
435 | } | |
436 | ||
437 | static void netlink_mm_open(struct vm_area_struct *vma) | |
438 | { | |
439 | struct file *file = vma->vm_file; | |
440 | struct socket *sock = file->private_data; | |
441 | struct sock *sk = sock->sk; | |
442 | ||
443 | if (sk) | |
444 | atomic_inc(&nlk_sk(sk)->mapped); | |
445 | } | |
446 | ||
447 | static void netlink_mm_close(struct vm_area_struct *vma) | |
448 | { | |
449 | struct file *file = vma->vm_file; | |
450 | struct socket *sock = file->private_data; | |
451 | struct sock *sk = sock->sk; | |
452 | ||
453 | if (sk) | |
454 | atomic_dec(&nlk_sk(sk)->mapped); | |
455 | } | |
456 | ||
457 | static const struct vm_operations_struct netlink_mmap_ops = { | |
458 | .open = netlink_mm_open, | |
459 | .close = netlink_mm_close, | |
460 | }; | |
461 | ||
462 | static int netlink_mmap(struct file *file, struct socket *sock, | |
463 | struct vm_area_struct *vma) | |
464 | { | |
465 | struct sock *sk = sock->sk; | |
466 | struct netlink_sock *nlk = nlk_sk(sk); | |
467 | struct netlink_ring *ring; | |
468 | unsigned long start, size, expected; | |
469 | unsigned int i; | |
470 | int err = -EINVAL; | |
471 | ||
472 | if (vma->vm_pgoff) | |
473 | return -EINVAL; | |
474 | ||
475 | mutex_lock(&nlk->pg_vec_lock); | |
476 | ||
477 | expected = 0; | |
478 | for (ring = &nlk->rx_ring; ring <= &nlk->tx_ring; ring++) { | |
479 | if (ring->pg_vec == NULL) | |
480 | continue; | |
481 | expected += ring->pg_vec_len * ring->pg_vec_pages * PAGE_SIZE; | |
482 | } | |
483 | ||
484 | if (expected == 0) | |
485 | goto out; | |
486 | ||
487 | size = vma->vm_end - vma->vm_start; | |
488 | if (size != expected) | |
489 | goto out; | |
490 | ||
491 | start = vma->vm_start; | |
492 | for (ring = &nlk->rx_ring; ring <= &nlk->tx_ring; ring++) { | |
493 | if (ring->pg_vec == NULL) | |
494 | continue; | |
495 | ||
496 | for (i = 0; i < ring->pg_vec_len; i++) { | |
497 | struct page *page; | |
498 | void *kaddr = ring->pg_vec[i]; | |
499 | unsigned int pg_num; | |
500 | ||
501 | for (pg_num = 0; pg_num < ring->pg_vec_pages; pg_num++) { | |
502 | page = pgvec_to_page(kaddr); | |
503 | err = vm_insert_page(vma, start, page); | |
504 | if (err < 0) | |
505 | goto out; | |
506 | start += PAGE_SIZE; | |
507 | kaddr += PAGE_SIZE; | |
508 | } | |
509 | } | |
510 | } | |
511 | ||
512 | atomic_inc(&nlk->mapped); | |
513 | vma->vm_ops = &netlink_mmap_ops; | |
514 | err = 0; | |
515 | out: | |
516 | mutex_unlock(&nlk->pg_vec_lock); | |
7cdbac71 | 517 | return err; |
ccdfcc39 | 518 | } |
9652e931 | 519 | |
4682a035 | 520 | static void netlink_frame_flush_dcache(const struct nl_mmap_hdr *hdr, unsigned int nm_len) |
9652e931 PM |
521 | { |
522 | #if ARCH_IMPLEMENTS_FLUSH_DCACHE_PAGE == 1 | |
523 | struct page *p_start, *p_end; | |
524 | ||
525 | /* First page is flushed through netlink_{get,set}_status */ | |
526 | p_start = pgvec_to_page(hdr + PAGE_SIZE); | |
4682a035 | 527 | p_end = pgvec_to_page((void *)hdr + NL_MMAP_HDRLEN + nm_len - 1); |
9652e931 PM |
528 | while (p_start <= p_end) { |
529 | flush_dcache_page(p_start); | |
530 | p_start++; | |
531 | } | |
532 | #endif | |
533 | } | |
534 | ||
535 | static enum nl_mmap_status netlink_get_status(const struct nl_mmap_hdr *hdr) | |
536 | { | |
537 | smp_rmb(); | |
538 | flush_dcache_page(pgvec_to_page(hdr)); | |
539 | return hdr->nm_status; | |
540 | } | |
541 | ||
542 | static void netlink_set_status(struct nl_mmap_hdr *hdr, | |
543 | enum nl_mmap_status status) | |
544 | { | |
a18e6a18 | 545 | smp_mb(); |
9652e931 PM |
546 | hdr->nm_status = status; |
547 | flush_dcache_page(pgvec_to_page(hdr)); | |
9652e931 PM |
548 | } |
549 | ||
550 | static struct nl_mmap_hdr * | |
551 | __netlink_lookup_frame(const struct netlink_ring *ring, unsigned int pos) | |
552 | { | |
553 | unsigned int pg_vec_pos, frame_off; | |
554 | ||
555 | pg_vec_pos = pos / ring->frames_per_block; | |
556 | frame_off = pos % ring->frames_per_block; | |
557 | ||
558 | return ring->pg_vec[pg_vec_pos] + (frame_off * ring->frame_size); | |
559 | } | |
560 | ||
561 | static struct nl_mmap_hdr * | |
562 | netlink_lookup_frame(const struct netlink_ring *ring, unsigned int pos, | |
563 | enum nl_mmap_status status) | |
564 | { | |
565 | struct nl_mmap_hdr *hdr; | |
566 | ||
567 | hdr = __netlink_lookup_frame(ring, pos); | |
568 | if (netlink_get_status(hdr) != status) | |
569 | return NULL; | |
570 | ||
571 | return hdr; | |
572 | } | |
573 | ||
574 | static struct nl_mmap_hdr * | |
575 | netlink_current_frame(const struct netlink_ring *ring, | |
576 | enum nl_mmap_status status) | |
577 | { | |
578 | return netlink_lookup_frame(ring, ring->head, status); | |
579 | } | |
580 | ||
581 | static struct nl_mmap_hdr * | |
582 | netlink_previous_frame(const struct netlink_ring *ring, | |
583 | enum nl_mmap_status status) | |
584 | { | |
585 | unsigned int prev; | |
586 | ||
587 | prev = ring->head ? ring->head - 1 : ring->frame_max; | |
588 | return netlink_lookup_frame(ring, prev, status); | |
589 | } | |
590 | ||
591 | static void netlink_increment_head(struct netlink_ring *ring) | |
592 | { | |
593 | ring->head = ring->head != ring->frame_max ? ring->head + 1 : 0; | |
594 | } | |
595 | ||
596 | static void netlink_forward_ring(struct netlink_ring *ring) | |
597 | { | |
598 | unsigned int head = ring->head, pos = head; | |
599 | const struct nl_mmap_hdr *hdr; | |
600 | ||
601 | do { | |
602 | hdr = __netlink_lookup_frame(ring, pos); | |
603 | if (hdr->nm_status == NL_MMAP_STATUS_UNUSED) | |
604 | break; | |
605 | if (hdr->nm_status != NL_MMAP_STATUS_SKIP) | |
606 | break; | |
607 | netlink_increment_head(ring); | |
608 | } while (ring->head != head); | |
609 | } | |
610 | ||
cd1df525 PM |
611 | static bool netlink_dump_space(struct netlink_sock *nlk) |
612 | { | |
613 | struct netlink_ring *ring = &nlk->rx_ring; | |
614 | struct nl_mmap_hdr *hdr; | |
615 | unsigned int n; | |
616 | ||
617 | hdr = netlink_current_frame(ring, NL_MMAP_STATUS_UNUSED); | |
618 | if (hdr == NULL) | |
619 | return false; | |
620 | ||
621 | n = ring->head + ring->frame_max / 2; | |
622 | if (n > ring->frame_max) | |
623 | n -= ring->frame_max; | |
624 | ||
625 | hdr = __netlink_lookup_frame(ring, n); | |
626 | ||
627 | return hdr->nm_status == NL_MMAP_STATUS_UNUSED; | |
628 | } | |
629 | ||
9652e931 PM |
630 | static unsigned int netlink_poll(struct file *file, struct socket *sock, |
631 | poll_table *wait) | |
632 | { | |
633 | struct sock *sk = sock->sk; | |
634 | struct netlink_sock *nlk = nlk_sk(sk); | |
635 | unsigned int mask; | |
cd1df525 | 636 | int err; |
9652e931 | 637 | |
cd1df525 PM |
638 | if (nlk->rx_ring.pg_vec != NULL) { |
639 | /* Memory mapped sockets don't call recvmsg(), so flow control | |
640 | * for dumps is performed here. A dump is allowed to continue | |
641 | * if at least half the ring is unused. | |
642 | */ | |
16b304f3 | 643 | while (nlk->cb_running && netlink_dump_space(nlk)) { |
cd1df525 PM |
644 | err = netlink_dump(sk); |
645 | if (err < 0) { | |
ac30ef83 | 646 | sk->sk_err = -err; |
cd1df525 PM |
647 | sk->sk_error_report(sk); |
648 | break; | |
649 | } | |
650 | } | |
651 | netlink_rcv_wake(sk); | |
652 | } | |
5fd96123 | 653 | |
9652e931 PM |
654 | mask = datagram_poll(file, sock, wait); |
655 | ||
656 | spin_lock_bh(&sk->sk_receive_queue.lock); | |
657 | if (nlk->rx_ring.pg_vec) { | |
658 | netlink_forward_ring(&nlk->rx_ring); | |
659 | if (!netlink_previous_frame(&nlk->rx_ring, NL_MMAP_STATUS_UNUSED)) | |
660 | mask |= POLLIN | POLLRDNORM; | |
661 | } | |
662 | spin_unlock_bh(&sk->sk_receive_queue.lock); | |
663 | ||
664 | spin_lock_bh(&sk->sk_write_queue.lock); | |
665 | if (nlk->tx_ring.pg_vec) { | |
666 | if (netlink_current_frame(&nlk->tx_ring, NL_MMAP_STATUS_UNUSED)) | |
667 | mask |= POLLOUT | POLLWRNORM; | |
668 | } | |
669 | spin_unlock_bh(&sk->sk_write_queue.lock); | |
670 | ||
671 | return mask; | |
672 | } | |
673 | ||
674 | static struct nl_mmap_hdr *netlink_mmap_hdr(struct sk_buff *skb) | |
675 | { | |
676 | return (struct nl_mmap_hdr *)(skb->head - NL_MMAP_HDRLEN); | |
677 | } | |
678 | ||
679 | static void netlink_ring_setup_skb(struct sk_buff *skb, struct sock *sk, | |
680 | struct netlink_ring *ring, | |
681 | struct nl_mmap_hdr *hdr) | |
682 | { | |
683 | unsigned int size; | |
684 | void *data; | |
685 | ||
686 | size = ring->frame_size - NL_MMAP_HDRLEN; | |
687 | data = (void *)hdr + NL_MMAP_HDRLEN; | |
688 | ||
689 | skb->head = data; | |
690 | skb->data = data; | |
691 | skb_reset_tail_pointer(skb); | |
692 | skb->end = skb->tail + size; | |
693 | skb->len = 0; | |
694 | ||
695 | skb->destructor = netlink_skb_destructor; | |
696 | NETLINK_CB(skb).flags |= NETLINK_SKB_MMAPED; | |
697 | NETLINK_CB(skb).sk = sk; | |
698 | } | |
5fd96123 PM |
699 | |
700 | static int netlink_mmap_sendmsg(struct sock *sk, struct msghdr *msg, | |
701 | u32 dst_portid, u32 dst_group, | |
7cc05662 | 702 | struct scm_cookie *scm) |
5fd96123 PM |
703 | { |
704 | struct netlink_sock *nlk = nlk_sk(sk); | |
705 | struct netlink_ring *ring; | |
706 | struct nl_mmap_hdr *hdr; | |
707 | struct sk_buff *skb; | |
708 | unsigned int maxlen; | |
5fd96123 PM |
709 | int err = 0, len = 0; |
710 | ||
5fd96123 PM |
711 | mutex_lock(&nlk->pg_vec_lock); |
712 | ||
713 | ring = &nlk->tx_ring; | |
714 | maxlen = ring->frame_size - NL_MMAP_HDRLEN; | |
715 | ||
716 | do { | |
4682a035 DM |
717 | unsigned int nm_len; |
718 | ||
5fd96123 PM |
719 | hdr = netlink_current_frame(ring, NL_MMAP_STATUS_VALID); |
720 | if (hdr == NULL) { | |
721 | if (!(msg->msg_flags & MSG_DONTWAIT) && | |
722 | atomic_read(&nlk->tx_ring.pending)) | |
723 | schedule(); | |
724 | continue; | |
725 | } | |
4682a035 DM |
726 | |
727 | nm_len = ACCESS_ONCE(hdr->nm_len); | |
728 | if (nm_len > maxlen) { | |
5fd96123 PM |
729 | err = -EINVAL; |
730 | goto out; | |
731 | } | |
732 | ||
4682a035 | 733 | netlink_frame_flush_dcache(hdr, nm_len); |
5fd96123 | 734 | |
4682a035 DM |
735 | skb = alloc_skb(nm_len, GFP_KERNEL); |
736 | if (skb == NULL) { | |
737 | err = -ENOBUFS; | |
738 | goto out; | |
5fd96123 | 739 | } |
4682a035 DM |
740 | __skb_put(skb, nm_len); |
741 | memcpy(skb->data, (void *)hdr + NL_MMAP_HDRLEN, nm_len); | |
742 | netlink_set_status(hdr, NL_MMAP_STATUS_UNUSED); | |
5fd96123 PM |
743 | |
744 | netlink_increment_head(ring); | |
745 | ||
746 | NETLINK_CB(skb).portid = nlk->portid; | |
747 | NETLINK_CB(skb).dst_group = dst_group; | |
7cc05662 | 748 | NETLINK_CB(skb).creds = scm->creds; |
5fd96123 PM |
749 | |
750 | err = security_netlink_send(sk, skb); | |
751 | if (err) { | |
752 | kfree_skb(skb); | |
753 | goto out; | |
754 | } | |
755 | ||
756 | if (unlikely(dst_group)) { | |
757 | atomic_inc(&skb->users); | |
758 | netlink_broadcast(sk, skb, dst_portid, dst_group, | |
759 | GFP_KERNEL); | |
760 | } | |
761 | err = netlink_unicast(sk, skb, dst_portid, | |
762 | msg->msg_flags & MSG_DONTWAIT); | |
763 | if (err < 0) | |
764 | goto out; | |
765 | len += err; | |
766 | ||
767 | } while (hdr != NULL || | |
768 | (!(msg->msg_flags & MSG_DONTWAIT) && | |
769 | atomic_read(&nlk->tx_ring.pending))); | |
770 | ||
771 | if (len > 0) | |
772 | err = len; | |
773 | out: | |
774 | mutex_unlock(&nlk->pg_vec_lock); | |
775 | return err; | |
776 | } | |
f9c22888 PM |
777 | |
778 | static void netlink_queue_mmaped_skb(struct sock *sk, struct sk_buff *skb) | |
779 | { | |
780 | struct nl_mmap_hdr *hdr; | |
781 | ||
782 | hdr = netlink_mmap_hdr(skb); | |
783 | hdr->nm_len = skb->len; | |
784 | hdr->nm_group = NETLINK_CB(skb).dst_group; | |
785 | hdr->nm_pid = NETLINK_CB(skb).creds.pid; | |
1bf9310a ND |
786 | hdr->nm_uid = from_kuid(sk_user_ns(sk), NETLINK_CB(skb).creds.uid); |
787 | hdr->nm_gid = from_kgid(sk_user_ns(sk), NETLINK_CB(skb).creds.gid); | |
4682a035 | 788 | netlink_frame_flush_dcache(hdr, hdr->nm_len); |
f9c22888 PM |
789 | netlink_set_status(hdr, NL_MMAP_STATUS_VALID); |
790 | ||
791 | NETLINK_CB(skb).flags |= NETLINK_SKB_DELIVERED; | |
792 | kfree_skb(skb); | |
793 | } | |
794 | ||
795 | static void netlink_ring_set_copied(struct sock *sk, struct sk_buff *skb) | |
796 | { | |
797 | struct netlink_sock *nlk = nlk_sk(sk); | |
798 | struct netlink_ring *ring = &nlk->rx_ring; | |
799 | struct nl_mmap_hdr *hdr; | |
800 | ||
801 | spin_lock_bh(&sk->sk_receive_queue.lock); | |
802 | hdr = netlink_current_frame(ring, NL_MMAP_STATUS_UNUSED); | |
803 | if (hdr == NULL) { | |
804 | spin_unlock_bh(&sk->sk_receive_queue.lock); | |
805 | kfree_skb(skb); | |
cd1df525 | 806 | netlink_overrun(sk); |
f9c22888 PM |
807 | return; |
808 | } | |
809 | netlink_increment_head(ring); | |
810 | __skb_queue_tail(&sk->sk_receive_queue, skb); | |
811 | spin_unlock_bh(&sk->sk_receive_queue.lock); | |
812 | ||
813 | hdr->nm_len = skb->len; | |
814 | hdr->nm_group = NETLINK_CB(skb).dst_group; | |
815 | hdr->nm_pid = NETLINK_CB(skb).creds.pid; | |
1bf9310a ND |
816 | hdr->nm_uid = from_kuid(sk_user_ns(sk), NETLINK_CB(skb).creds.uid); |
817 | hdr->nm_gid = from_kgid(sk_user_ns(sk), NETLINK_CB(skb).creds.gid); | |
f9c22888 PM |
818 | netlink_set_status(hdr, NL_MMAP_STATUS_COPY); |
819 | } | |
820 | ||
ccdfcc39 | 821 | #else /* CONFIG_NETLINK_MMAP */ |
9652e931 | 822 | #define netlink_skb_is_mmaped(skb) false |
f9c22888 | 823 | #define netlink_rx_is_mmaped(sk) false |
5fd96123 | 824 | #define netlink_tx_is_mmaped(sk) false |
ccdfcc39 | 825 | #define netlink_mmap sock_no_mmap |
9652e931 | 826 | #define netlink_poll datagram_poll |
7cc05662 | 827 | #define netlink_mmap_sendmsg(sk, msg, dst_portid, dst_group, scm) 0 |
ccdfcc39 PM |
828 | #endif /* CONFIG_NETLINK_MMAP */ |
829 | ||
cf0a018a PM |
830 | static void netlink_skb_destructor(struct sk_buff *skb) |
831 | { | |
9652e931 PM |
832 | #ifdef CONFIG_NETLINK_MMAP |
833 | struct nl_mmap_hdr *hdr; | |
834 | struct netlink_ring *ring; | |
835 | struct sock *sk; | |
836 | ||
837 | /* If a packet from the kernel to userspace was freed because of an | |
838 | * error without being delivered to userspace, the kernel must reset | |
839 | * the status. In the direction userspace to kernel, the status is | |
840 | * always reset here after the packet was processed and freed. | |
841 | */ | |
842 | if (netlink_skb_is_mmaped(skb)) { | |
843 | hdr = netlink_mmap_hdr(skb); | |
844 | sk = NETLINK_CB(skb).sk; | |
845 | ||
5fd96123 PM |
846 | if (NETLINK_CB(skb).flags & NETLINK_SKB_TX) { |
847 | netlink_set_status(hdr, NL_MMAP_STATUS_UNUSED); | |
848 | ring = &nlk_sk(sk)->tx_ring; | |
849 | } else { | |
850 | if (!(NETLINK_CB(skb).flags & NETLINK_SKB_DELIVERED)) { | |
851 | hdr->nm_len = 0; | |
852 | netlink_set_status(hdr, NL_MMAP_STATUS_VALID); | |
853 | } | |
854 | ring = &nlk_sk(sk)->rx_ring; | |
9652e931 | 855 | } |
9652e931 PM |
856 | |
857 | WARN_ON(atomic_read(&ring->pending) == 0); | |
858 | atomic_dec(&ring->pending); | |
859 | sock_put(sk); | |
860 | ||
5e71d9d7 | 861 | skb->head = NULL; |
9652e931 PM |
862 | } |
863 | #endif | |
c05cdb1b | 864 | if (is_vmalloc_addr(skb->head)) { |
3a36515f PN |
865 | if (!skb->cloned || |
866 | !atomic_dec_return(&(skb_shinfo(skb)->dataref))) | |
867 | vfree(skb->head); | |
868 | ||
c05cdb1b PNA |
869 | skb->head = NULL; |
870 | } | |
9652e931 PM |
871 | if (skb->sk != NULL) |
872 | sock_rfree(skb); | |
cf0a018a PM |
873 | } |
874 | ||
875 | static void netlink_skb_set_owner_r(struct sk_buff *skb, struct sock *sk) | |
876 | { | |
877 | WARN_ON(skb->sk != NULL); | |
878 | skb->sk = sk; | |
879 | skb->destructor = netlink_skb_destructor; | |
880 | atomic_add(skb->truesize, &sk->sk_rmem_alloc); | |
881 | sk_mem_charge(sk, skb->truesize); | |
882 | } | |
883 | ||
1da177e4 LT |
884 | static void netlink_sock_destruct(struct sock *sk) |
885 | { | |
3f660d66 HX |
886 | struct netlink_sock *nlk = nlk_sk(sk); |
887 | ||
16b304f3 PS |
888 | if (nlk->cb_running) { |
889 | if (nlk->cb.done) | |
890 | nlk->cb.done(&nlk->cb); | |
6dc878a8 | 891 | |
16b304f3 PS |
892 | module_put(nlk->cb.module); |
893 | kfree_skb(nlk->cb.skb); | |
3f660d66 HX |
894 | } |
895 | ||
1da177e4 | 896 | skb_queue_purge(&sk->sk_receive_queue); |
ccdfcc39 PM |
897 | #ifdef CONFIG_NETLINK_MMAP |
898 | if (1) { | |
899 | struct nl_mmap_req req; | |
900 | ||
901 | memset(&req, 0, sizeof(req)); | |
902 | if (nlk->rx_ring.pg_vec) | |
903 | netlink_set_ring(sk, &req, true, false); | |
904 | memset(&req, 0, sizeof(req)); | |
905 | if (nlk->tx_ring.pg_vec) | |
906 | netlink_set_ring(sk, &req, true, true); | |
907 | } | |
908 | #endif /* CONFIG_NETLINK_MMAP */ | |
1da177e4 LT |
909 | |
910 | if (!sock_flag(sk, SOCK_DEAD)) { | |
6ac552fd | 911 | printk(KERN_ERR "Freeing alive netlink socket %p\n", sk); |
1da177e4 LT |
912 | return; |
913 | } | |
547b792c IJ |
914 | |
915 | WARN_ON(atomic_read(&sk->sk_rmem_alloc)); | |
916 | WARN_ON(atomic_read(&sk->sk_wmem_alloc)); | |
917 | WARN_ON(nlk_sk(sk)->groups); | |
1da177e4 LT |
918 | } |
919 | ||
6ac552fd PM |
920 | /* This lock without WQ_FLAG_EXCLUSIVE is good on UP and it is _very_ bad on |
921 | * SMP. Look, when several writers sleep and reader wakes them up, all but one | |
1da177e4 LT |
922 | * immediately hit write lock and grab all the cpus. Exclusive sleep solves |
923 | * this, _but_ remember, it adds useless work on UP machines. | |
924 | */ | |
925 | ||
d136f1bd | 926 | void netlink_table_grab(void) |
9a429c49 | 927 | __acquires(nl_table_lock) |
1da177e4 | 928 | { |
d136f1bd JB |
929 | might_sleep(); |
930 | ||
6abd219c | 931 | write_lock_irq(&nl_table_lock); |
1da177e4 LT |
932 | |
933 | if (atomic_read(&nl_table_users)) { | |
934 | DECLARE_WAITQUEUE(wait, current); | |
935 | ||
936 | add_wait_queue_exclusive(&nl_table_wait, &wait); | |
6ac552fd | 937 | for (;;) { |
1da177e4 LT |
938 | set_current_state(TASK_UNINTERRUPTIBLE); |
939 | if (atomic_read(&nl_table_users) == 0) | |
940 | break; | |
6abd219c | 941 | write_unlock_irq(&nl_table_lock); |
1da177e4 | 942 | schedule(); |
6abd219c | 943 | write_lock_irq(&nl_table_lock); |
1da177e4 LT |
944 | } |
945 | ||
946 | __set_current_state(TASK_RUNNING); | |
947 | remove_wait_queue(&nl_table_wait, &wait); | |
948 | } | |
949 | } | |
950 | ||
d136f1bd | 951 | void netlink_table_ungrab(void) |
9a429c49 | 952 | __releases(nl_table_lock) |
1da177e4 | 953 | { |
6abd219c | 954 | write_unlock_irq(&nl_table_lock); |
1da177e4 LT |
955 | wake_up(&nl_table_wait); |
956 | } | |
957 | ||
6ac552fd | 958 | static inline void |
1da177e4 LT |
959 | netlink_lock_table(void) |
960 | { | |
961 | /* read_lock() synchronizes us to netlink_table_grab */ | |
962 | ||
963 | read_lock(&nl_table_lock); | |
964 | atomic_inc(&nl_table_users); | |
965 | read_unlock(&nl_table_lock); | |
966 | } | |
967 | ||
6ac552fd | 968 | static inline void |
1da177e4 LT |
969 | netlink_unlock_table(void) |
970 | { | |
971 | if (atomic_dec_and_test(&nl_table_users)) | |
972 | wake_up(&nl_table_wait); | |
973 | } | |
974 | ||
e341694e | 975 | struct netlink_compare_arg |
1da177e4 | 976 | { |
c428ecd1 | 977 | possible_net_t pnet; |
e341694e TG |
978 | u32 portid; |
979 | }; | |
1da177e4 | 980 | |
8f2ddaac HX |
981 | /* Doing sizeof directly may yield 4 extra bytes on 64-bit. */ |
982 | #define netlink_compare_arg_len \ | |
983 | (offsetof(struct netlink_compare_arg, portid) + sizeof(u32)) | |
c428ecd1 HX |
984 | |
985 | static inline int netlink_compare(struct rhashtable_compare_arg *arg, | |
986 | const void *ptr) | |
1da177e4 | 987 | { |
c428ecd1 HX |
988 | const struct netlink_compare_arg *x = arg->key; |
989 | const struct netlink_sock *nlk = ptr; | |
1da177e4 | 990 | |
c428ecd1 HX |
991 | return nlk->portid != x->portid || |
992 | !net_eq(sock_net(&nlk->sk), read_pnet(&x->pnet)); | |
993 | } | |
994 | ||
995 | static void netlink_compare_arg_init(struct netlink_compare_arg *arg, | |
996 | struct net *net, u32 portid) | |
997 | { | |
998 | memset(arg, 0, sizeof(*arg)); | |
999 | write_pnet(&arg->pnet, net); | |
1000 | arg->portid = portid; | |
1da177e4 LT |
1001 | } |
1002 | ||
e341694e TG |
1003 | static struct sock *__netlink_lookup(struct netlink_table *table, u32 portid, |
1004 | struct net *net) | |
1da177e4 | 1005 | { |
c428ecd1 | 1006 | struct netlink_compare_arg arg; |
1da177e4 | 1007 | |
c428ecd1 HX |
1008 | netlink_compare_arg_init(&arg, net, portid); |
1009 | return rhashtable_lookup_fast(&table->hash, &arg, | |
1010 | netlink_rhashtable_params); | |
1da177e4 LT |
1011 | } |
1012 | ||
c428ecd1 | 1013 | static int __netlink_insert(struct netlink_table *table, struct sock *sk) |
c5adde94 | 1014 | { |
c428ecd1 | 1015 | struct netlink_compare_arg arg; |
c5adde94 | 1016 | |
c428ecd1 HX |
1017 | netlink_compare_arg_init(&arg, sock_net(sk), nlk_sk(sk)->portid); |
1018 | return rhashtable_lookup_insert_key(&table->hash, &arg, | |
1019 | &nlk_sk(sk)->node, | |
1020 | netlink_rhashtable_params); | |
c5adde94 YX |
1021 | } |
1022 | ||
e341694e | 1023 | static struct sock *netlink_lookup(struct net *net, int protocol, u32 portid) |
1da177e4 | 1024 | { |
e341694e TG |
1025 | struct netlink_table *table = &nl_table[protocol]; |
1026 | struct sock *sk; | |
1da177e4 | 1027 | |
e341694e TG |
1028 | rcu_read_lock(); |
1029 | sk = __netlink_lookup(table, portid, net); | |
1030 | if (sk) | |
1031 | sock_hold(sk); | |
1032 | rcu_read_unlock(); | |
1da177e4 | 1033 | |
e341694e | 1034 | return sk; |
1da177e4 LT |
1035 | } |
1036 | ||
90ddc4f0 | 1037 | static const struct proto_ops netlink_ops; |
1da177e4 | 1038 | |
4277a083 PM |
1039 | static void |
1040 | netlink_update_listeners(struct sock *sk) | |
1041 | { | |
1042 | struct netlink_table *tbl = &nl_table[sk->sk_protocol]; | |
4277a083 PM |
1043 | unsigned long mask; |
1044 | unsigned int i; | |
6d772ac5 ED |
1045 | struct listeners *listeners; |
1046 | ||
1047 | listeners = nl_deref_protected(tbl->listeners); | |
1048 | if (!listeners) | |
1049 | return; | |
4277a083 | 1050 | |
b4ff4f04 | 1051 | for (i = 0; i < NLGRPLONGS(tbl->groups); i++) { |
4277a083 | 1052 | mask = 0; |
b67bfe0d | 1053 | sk_for_each_bound(sk, &tbl->mc_list) { |
b4ff4f04 JB |
1054 | if (i < NLGRPLONGS(nlk_sk(sk)->ngroups)) |
1055 | mask |= nlk_sk(sk)->groups[i]; | |
1056 | } | |
6d772ac5 | 1057 | listeners->masks[i] = mask; |
4277a083 PM |
1058 | } |
1059 | /* this function is only called with the netlink table "grabbed", which | |
1060 | * makes sure updates are visible before bind or setsockopt return. */ | |
1061 | } | |
1062 | ||
8ea65f4a | 1063 | static int netlink_insert(struct sock *sk, u32 portid) |
1da177e4 | 1064 | { |
da12c90e | 1065 | struct netlink_table *table = &nl_table[sk->sk_protocol]; |
919d9db9 | 1066 | int err; |
1da177e4 | 1067 | |
c5adde94 | 1068 | lock_sock(sk); |
1da177e4 LT |
1069 | |
1070 | err = -EBUSY; | |
15e47304 | 1071 | if (nlk_sk(sk)->portid) |
1da177e4 LT |
1072 | goto err; |
1073 | ||
1074 | err = -ENOMEM; | |
97defe1e TG |
1075 | if (BITS_PER_LONG > 32 && |
1076 | unlikely(atomic_read(&table->hash.nelems) >= UINT_MAX)) | |
1da177e4 LT |
1077 | goto err; |
1078 | ||
15e47304 | 1079 | nlk_sk(sk)->portid = portid; |
e341694e | 1080 | sock_hold(sk); |
919d9db9 | 1081 | |
c428ecd1 HX |
1082 | err = __netlink_insert(table, sk); |
1083 | if (err) { | |
1084 | if (err == -EEXIST) | |
1085 | err = -EADDRINUSE; | |
c5adde94 | 1086 | sock_put(sk); |
919d9db9 HX |
1087 | } |
1088 | ||
1da177e4 | 1089 | err: |
c5adde94 | 1090 | release_sock(sk); |
1da177e4 LT |
1091 | return err; |
1092 | } | |
1093 | ||
1094 | static void netlink_remove(struct sock *sk) | |
1095 | { | |
e341694e TG |
1096 | struct netlink_table *table; |
1097 | ||
e341694e | 1098 | table = &nl_table[sk->sk_protocol]; |
c428ecd1 HX |
1099 | if (!rhashtable_remove_fast(&table->hash, &nlk_sk(sk)->node, |
1100 | netlink_rhashtable_params)) { | |
e341694e TG |
1101 | WARN_ON(atomic_read(&sk->sk_refcnt) == 1); |
1102 | __sock_put(sk); | |
1103 | } | |
e341694e | 1104 | |
1da177e4 | 1105 | netlink_table_grab(); |
b10dcb3b | 1106 | if (nlk_sk(sk)->subscriptions) { |
1da177e4 | 1107 | __sk_del_bind_node(sk); |
b10dcb3b JB |
1108 | netlink_update_listeners(sk); |
1109 | } | |
ee1c2442 JB |
1110 | if (sk->sk_protocol == NETLINK_GENERIC) |
1111 | atomic_inc(&genl_sk_destructing_cnt); | |
1da177e4 LT |
1112 | netlink_table_ungrab(); |
1113 | } | |
1114 | ||
1115 | static struct proto netlink_proto = { | |
1116 | .name = "NETLINK", | |
1117 | .owner = THIS_MODULE, | |
1118 | .obj_size = sizeof(struct netlink_sock), | |
1119 | }; | |
1120 | ||
1b8d7ae4 | 1121 | static int __netlink_create(struct net *net, struct socket *sock, |
11aa9c28 EB |
1122 | struct mutex *cb_mutex, int protocol, |
1123 | int kern) | |
1da177e4 LT |
1124 | { |
1125 | struct sock *sk; | |
1126 | struct netlink_sock *nlk; | |
ab33a171 PM |
1127 | |
1128 | sock->ops = &netlink_ops; | |
1129 | ||
11aa9c28 | 1130 | sk = sk_alloc(net, PF_NETLINK, GFP_KERNEL, &netlink_proto, kern); |
ab33a171 PM |
1131 | if (!sk) |
1132 | return -ENOMEM; | |
1133 | ||
1134 | sock_init_data(sock, sk); | |
1135 | ||
1136 | nlk = nlk_sk(sk); | |
658cb354 | 1137 | if (cb_mutex) { |
ffa4d721 | 1138 | nlk->cb_mutex = cb_mutex; |
658cb354 | 1139 | } else { |
ffa4d721 PM |
1140 | nlk->cb_mutex = &nlk->cb_def_mutex; |
1141 | mutex_init(nlk->cb_mutex); | |
1142 | } | |
ab33a171 | 1143 | init_waitqueue_head(&nlk->wait); |
ccdfcc39 PM |
1144 | #ifdef CONFIG_NETLINK_MMAP |
1145 | mutex_init(&nlk->pg_vec_lock); | |
1146 | #endif | |
ab33a171 PM |
1147 | |
1148 | sk->sk_destruct = netlink_sock_destruct; | |
1149 | sk->sk_protocol = protocol; | |
1150 | return 0; | |
1151 | } | |
1152 | ||
3f378b68 EP |
1153 | static int netlink_create(struct net *net, struct socket *sock, int protocol, |
1154 | int kern) | |
ab33a171 PM |
1155 | { |
1156 | struct module *module = NULL; | |
af65bdfc | 1157 | struct mutex *cb_mutex; |
f7fa9b10 | 1158 | struct netlink_sock *nlk; |
023e2cfa JB |
1159 | int (*bind)(struct net *net, int group); |
1160 | void (*unbind)(struct net *net, int group); | |
ab33a171 | 1161 | int err = 0; |
1da177e4 LT |
1162 | |
1163 | sock->state = SS_UNCONNECTED; | |
1164 | ||
1165 | if (sock->type != SOCK_RAW && sock->type != SOCK_DGRAM) | |
1166 | return -ESOCKTNOSUPPORT; | |
1167 | ||
6ac552fd | 1168 | if (protocol < 0 || protocol >= MAX_LINKS) |
1da177e4 LT |
1169 | return -EPROTONOSUPPORT; |
1170 | ||
77247bbb | 1171 | netlink_lock_table(); |
95a5afca | 1172 | #ifdef CONFIG_MODULES |
ab33a171 | 1173 | if (!nl_table[protocol].registered) { |
77247bbb | 1174 | netlink_unlock_table(); |
4fdb3bb7 | 1175 | request_module("net-pf-%d-proto-%d", PF_NETLINK, protocol); |
77247bbb | 1176 | netlink_lock_table(); |
4fdb3bb7 | 1177 | } |
ab33a171 PM |
1178 | #endif |
1179 | if (nl_table[protocol].registered && | |
1180 | try_module_get(nl_table[protocol].module)) | |
1181 | module = nl_table[protocol].module; | |
974c37e9 AD |
1182 | else |
1183 | err = -EPROTONOSUPPORT; | |
af65bdfc | 1184 | cb_mutex = nl_table[protocol].cb_mutex; |
03292745 | 1185 | bind = nl_table[protocol].bind; |
4f520900 | 1186 | unbind = nl_table[protocol].unbind; |
77247bbb | 1187 | netlink_unlock_table(); |
4fdb3bb7 | 1188 | |
974c37e9 AD |
1189 | if (err < 0) |
1190 | goto out; | |
1191 | ||
11aa9c28 | 1192 | err = __netlink_create(net, sock, cb_mutex, protocol, kern); |
6ac552fd | 1193 | if (err < 0) |
f7fa9b10 PM |
1194 | goto out_module; |
1195 | ||
6f756a8c | 1196 | local_bh_disable(); |
c1fd3b94 | 1197 | sock_prot_inuse_add(net, &netlink_proto, 1); |
6f756a8c DM |
1198 | local_bh_enable(); |
1199 | ||
f7fa9b10 | 1200 | nlk = nlk_sk(sock->sk); |
f7fa9b10 | 1201 | nlk->module = module; |
03292745 | 1202 | nlk->netlink_bind = bind; |
4f520900 | 1203 | nlk->netlink_unbind = unbind; |
ab33a171 PM |
1204 | out: |
1205 | return err; | |
1da177e4 | 1206 | |
ab33a171 PM |
1207 | out_module: |
1208 | module_put(module); | |
1209 | goto out; | |
1da177e4 LT |
1210 | } |
1211 | ||
21e4902a TG |
1212 | static void deferred_put_nlk_sk(struct rcu_head *head) |
1213 | { | |
1214 | struct netlink_sock *nlk = container_of(head, struct netlink_sock, rcu); | |
1215 | ||
1216 | sock_put(&nlk->sk); | |
1217 | } | |
1218 | ||
1da177e4 LT |
1219 | static int netlink_release(struct socket *sock) |
1220 | { | |
1221 | struct sock *sk = sock->sk; | |
1222 | struct netlink_sock *nlk; | |
1223 | ||
1224 | if (!sk) | |
1225 | return 0; | |
1226 | ||
1227 | netlink_remove(sk); | |
ac57b3a9 | 1228 | sock_orphan(sk); |
1da177e4 LT |
1229 | nlk = nlk_sk(sk); |
1230 | ||
3f660d66 HX |
1231 | /* |
1232 | * OK. Socket is unlinked, any packets that arrive now | |
1233 | * will be purged. | |
1234 | */ | |
1da177e4 | 1235 | |
ee1c2442 JB |
1236 | /* must not acquire netlink_table_lock in any way again before unbind |
1237 | * and notifying genetlink is done as otherwise it might deadlock | |
1238 | */ | |
1239 | if (nlk->netlink_unbind) { | |
1240 | int i; | |
1241 | ||
1242 | for (i = 0; i < nlk->ngroups; i++) | |
1243 | if (test_bit(i, nlk->groups)) | |
1244 | nlk->netlink_unbind(sock_net(sk), i + 1); | |
1245 | } | |
1246 | if (sk->sk_protocol == NETLINK_GENERIC && | |
1247 | atomic_dec_return(&genl_sk_destructing_cnt) == 0) | |
1248 | wake_up(&genl_sk_destructing_waitq); | |
1249 | ||
1da177e4 LT |
1250 | sock->sk = NULL; |
1251 | wake_up_interruptible_all(&nlk->wait); | |
1252 | ||
1253 | skb_queue_purge(&sk->sk_write_queue); | |
1254 | ||
15e47304 | 1255 | if (nlk->portid) { |
1da177e4 | 1256 | struct netlink_notify n = { |
3b1e0a65 | 1257 | .net = sock_net(sk), |
1da177e4 | 1258 | .protocol = sk->sk_protocol, |
15e47304 | 1259 | .portid = nlk->portid, |
1da177e4 | 1260 | }; |
e041c683 AS |
1261 | atomic_notifier_call_chain(&netlink_chain, |
1262 | NETLINK_URELEASE, &n); | |
746fac4d | 1263 | } |
4fdb3bb7 | 1264 | |
5e7c001c | 1265 | module_put(nlk->module); |
4fdb3bb7 | 1266 | |
aed81560 | 1267 | if (netlink_is_kernel(sk)) { |
b10dcb3b | 1268 | netlink_table_grab(); |
869e58f8 DL |
1269 | BUG_ON(nl_table[sk->sk_protocol].registered == 0); |
1270 | if (--nl_table[sk->sk_protocol].registered == 0) { | |
6d772ac5 ED |
1271 | struct listeners *old; |
1272 | ||
1273 | old = nl_deref_protected(nl_table[sk->sk_protocol].listeners); | |
1274 | RCU_INIT_POINTER(nl_table[sk->sk_protocol].listeners, NULL); | |
1275 | kfree_rcu(old, rcu); | |
869e58f8 | 1276 | nl_table[sk->sk_protocol].module = NULL; |
9785e10a | 1277 | nl_table[sk->sk_protocol].bind = NULL; |
4f520900 | 1278 | nl_table[sk->sk_protocol].unbind = NULL; |
9785e10a | 1279 | nl_table[sk->sk_protocol].flags = 0; |
869e58f8 DL |
1280 | nl_table[sk->sk_protocol].registered = 0; |
1281 | } | |
b10dcb3b | 1282 | netlink_table_ungrab(); |
658cb354 | 1283 | } |
77247bbb | 1284 | |
f7fa9b10 PM |
1285 | kfree(nlk->groups); |
1286 | nlk->groups = NULL; | |
1287 | ||
3755810c | 1288 | local_bh_disable(); |
c1fd3b94 | 1289 | sock_prot_inuse_add(sock_net(sk), &netlink_proto, -1); |
3755810c | 1290 | local_bh_enable(); |
21e4902a | 1291 | call_rcu(&nlk->rcu, deferred_put_nlk_sk); |
1da177e4 LT |
1292 | return 0; |
1293 | } | |
1294 | ||
1295 | static int netlink_autobind(struct socket *sock) | |
1296 | { | |
1297 | struct sock *sk = sock->sk; | |
3b1e0a65 | 1298 | struct net *net = sock_net(sk); |
da12c90e | 1299 | struct netlink_table *table = &nl_table[sk->sk_protocol]; |
15e47304 | 1300 | s32 portid = task_tgid_vnr(current); |
1da177e4 | 1301 | int err; |
b9fbe709 HX |
1302 | s32 rover = -4096; |
1303 | bool ok; | |
1da177e4 LT |
1304 | |
1305 | retry: | |
1306 | cond_resched(); | |
e341694e | 1307 | rcu_read_lock(); |
b9fbe709 HX |
1308 | ok = !__netlink_lookup(table, portid, net); |
1309 | rcu_read_unlock(); | |
1310 | if (!ok) { | |
e341694e | 1311 | /* Bind collision, search negative portid values. */ |
b9fbe709 HX |
1312 | if (rover == -4096) |
1313 | /* rover will be in range [S32_MIN, -4097] */ | |
1314 | rover = S32_MIN + prandom_u32_max(-4096 - S32_MIN); | |
1315 | else if (rover >= -4096) | |
e341694e | 1316 | rover = -4097; |
b9fbe709 | 1317 | portid = rover--; |
e341694e | 1318 | goto retry; |
1da177e4 | 1319 | } |
1da177e4 | 1320 | |
8ea65f4a | 1321 | err = netlink_insert(sk, portid); |
1da177e4 LT |
1322 | if (err == -EADDRINUSE) |
1323 | goto retry; | |
d470e3b4 DM |
1324 | |
1325 | /* If 2 threads race to autobind, that is fine. */ | |
1326 | if (err == -EBUSY) | |
1327 | err = 0; | |
1328 | ||
1329 | return err; | |
1da177e4 LT |
1330 | } |
1331 | ||
aa4cf945 EB |
1332 | /** |
1333 | * __netlink_ns_capable - General netlink message capability test | |
1334 | * @nsp: NETLINK_CB of the socket buffer holding a netlink command from userspace. | |
1335 | * @user_ns: The user namespace of the capability to use | |
1336 | * @cap: The capability to use | |
1337 | * | |
1338 | * Test to see if the opener of the socket we received the message | |
1339 | * from had when the netlink socket was created and the sender of the | |
1340 | * message has has the capability @cap in the user namespace @user_ns. | |
1341 | */ | |
1342 | bool __netlink_ns_capable(const struct netlink_skb_parms *nsp, | |
1343 | struct user_namespace *user_ns, int cap) | |
1344 | { | |
2d7a85f4 EB |
1345 | return ((nsp->flags & NETLINK_SKB_DST) || |
1346 | file_ns_capable(nsp->sk->sk_socket->file, user_ns, cap)) && | |
1347 | ns_capable(user_ns, cap); | |
aa4cf945 EB |
1348 | } |
1349 | EXPORT_SYMBOL(__netlink_ns_capable); | |
1350 | ||
1351 | /** | |
1352 | * netlink_ns_capable - General netlink message capability test | |
1353 | * @skb: socket buffer holding a netlink command from userspace | |
1354 | * @user_ns: The user namespace of the capability to use | |
1355 | * @cap: The capability to use | |
1356 | * | |
1357 | * Test to see if the opener of the socket we received the message | |
1358 | * from had when the netlink socket was created and the sender of the | |
1359 | * message has has the capability @cap in the user namespace @user_ns. | |
1360 | */ | |
1361 | bool netlink_ns_capable(const struct sk_buff *skb, | |
1362 | struct user_namespace *user_ns, int cap) | |
1363 | { | |
1364 | return __netlink_ns_capable(&NETLINK_CB(skb), user_ns, cap); | |
1365 | } | |
1366 | EXPORT_SYMBOL(netlink_ns_capable); | |
1367 | ||
1368 | /** | |
1369 | * netlink_capable - Netlink global message capability test | |
1370 | * @skb: socket buffer holding a netlink command from userspace | |
1371 | * @cap: The capability to use | |
1372 | * | |
1373 | * Test to see if the opener of the socket we received the message | |
1374 | * from had when the netlink socket was created and the sender of the | |
1375 | * message has has the capability @cap in all user namespaces. | |
1376 | */ | |
1377 | bool netlink_capable(const struct sk_buff *skb, int cap) | |
1378 | { | |
1379 | return netlink_ns_capable(skb, &init_user_ns, cap); | |
1380 | } | |
1381 | EXPORT_SYMBOL(netlink_capable); | |
1382 | ||
1383 | /** | |
1384 | * netlink_net_capable - Netlink network namespace message capability test | |
1385 | * @skb: socket buffer holding a netlink command from userspace | |
1386 | * @cap: The capability to use | |
1387 | * | |
1388 | * Test to see if the opener of the socket we received the message | |
1389 | * from had when the netlink socket was created and the sender of the | |
1390 | * message has has the capability @cap over the network namespace of | |
1391 | * the socket we received the message from. | |
1392 | */ | |
1393 | bool netlink_net_capable(const struct sk_buff *skb, int cap) | |
1394 | { | |
1395 | return netlink_ns_capable(skb, sock_net(skb->sk)->user_ns, cap); | |
1396 | } | |
1397 | EXPORT_SYMBOL(netlink_net_capable); | |
1398 | ||
5187cd05 | 1399 | static inline int netlink_allowed(const struct socket *sock, unsigned int flag) |
746fac4d | 1400 | { |
9785e10a | 1401 | return (nl_table[sock->sk->sk_protocol].flags & flag) || |
df008c91 | 1402 | ns_capable(sock_net(sock->sk)->user_ns, CAP_NET_ADMIN); |
746fac4d | 1403 | } |
1da177e4 | 1404 | |
f7fa9b10 PM |
1405 | static void |
1406 | netlink_update_subscriptions(struct sock *sk, unsigned int subscriptions) | |
1407 | { | |
1408 | struct netlink_sock *nlk = nlk_sk(sk); | |
1409 | ||
1410 | if (nlk->subscriptions && !subscriptions) | |
1411 | __sk_del_bind_node(sk); | |
1412 | else if (!nlk->subscriptions && subscriptions) | |
1413 | sk_add_bind_node(sk, &nl_table[sk->sk_protocol].mc_list); | |
1414 | nlk->subscriptions = subscriptions; | |
1415 | } | |
1416 | ||
b4ff4f04 | 1417 | static int netlink_realloc_groups(struct sock *sk) |
513c2500 PM |
1418 | { |
1419 | struct netlink_sock *nlk = nlk_sk(sk); | |
1420 | unsigned int groups; | |
b4ff4f04 | 1421 | unsigned long *new_groups; |
513c2500 PM |
1422 | int err = 0; |
1423 | ||
b4ff4f04 JB |
1424 | netlink_table_grab(); |
1425 | ||
513c2500 | 1426 | groups = nl_table[sk->sk_protocol].groups; |
b4ff4f04 | 1427 | if (!nl_table[sk->sk_protocol].registered) { |
513c2500 | 1428 | err = -ENOENT; |
b4ff4f04 JB |
1429 | goto out_unlock; |
1430 | } | |
513c2500 | 1431 | |
b4ff4f04 JB |
1432 | if (nlk->ngroups >= groups) |
1433 | goto out_unlock; | |
513c2500 | 1434 | |
b4ff4f04 JB |
1435 | new_groups = krealloc(nlk->groups, NLGRPSZ(groups), GFP_ATOMIC); |
1436 | if (new_groups == NULL) { | |
1437 | err = -ENOMEM; | |
1438 | goto out_unlock; | |
1439 | } | |
6ac552fd | 1440 | memset((char *)new_groups + NLGRPSZ(nlk->ngroups), 0, |
b4ff4f04 JB |
1441 | NLGRPSZ(groups) - NLGRPSZ(nlk->ngroups)); |
1442 | ||
1443 | nlk->groups = new_groups; | |
513c2500 | 1444 | nlk->ngroups = groups; |
b4ff4f04 JB |
1445 | out_unlock: |
1446 | netlink_table_ungrab(); | |
1447 | return err; | |
513c2500 PM |
1448 | } |
1449 | ||
02c81ab9 | 1450 | static void netlink_undo_bind(int group, long unsigned int groups, |
023e2cfa | 1451 | struct sock *sk) |
4f520900 | 1452 | { |
023e2cfa | 1453 | struct netlink_sock *nlk = nlk_sk(sk); |
4f520900 RGB |
1454 | int undo; |
1455 | ||
1456 | if (!nlk->netlink_unbind) | |
1457 | return; | |
1458 | ||
1459 | for (undo = 0; undo < group; undo++) | |
6251edd9 | 1460 | if (test_bit(undo, &groups)) |
8b7c36d8 | 1461 | nlk->netlink_unbind(sock_net(sk), undo + 1); |
4f520900 RGB |
1462 | } |
1463 | ||
6ac552fd PM |
1464 | static int netlink_bind(struct socket *sock, struct sockaddr *addr, |
1465 | int addr_len) | |
1da177e4 LT |
1466 | { |
1467 | struct sock *sk = sock->sk; | |
3b1e0a65 | 1468 | struct net *net = sock_net(sk); |
1da177e4 LT |
1469 | struct netlink_sock *nlk = nlk_sk(sk); |
1470 | struct sockaddr_nl *nladdr = (struct sockaddr_nl *)addr; | |
1471 | int err; | |
4f520900 | 1472 | long unsigned int groups = nladdr->nl_groups; |
746fac4d | 1473 | |
4e4b5376 HFS |
1474 | if (addr_len < sizeof(struct sockaddr_nl)) |
1475 | return -EINVAL; | |
1476 | ||
1da177e4 LT |
1477 | if (nladdr->nl_family != AF_NETLINK) |
1478 | return -EINVAL; | |
1479 | ||
1480 | /* Only superuser is allowed to listen multicasts */ | |
4f520900 | 1481 | if (groups) { |
5187cd05 | 1482 | if (!netlink_allowed(sock, NL_CFG_F_NONROOT_RECV)) |
513c2500 | 1483 | return -EPERM; |
b4ff4f04 JB |
1484 | err = netlink_realloc_groups(sk); |
1485 | if (err) | |
1486 | return err; | |
513c2500 | 1487 | } |
1da177e4 | 1488 | |
4f520900 | 1489 | if (nlk->portid) |
15e47304 | 1490 | if (nladdr->nl_pid != nlk->portid) |
1da177e4 | 1491 | return -EINVAL; |
4f520900 RGB |
1492 | |
1493 | if (nlk->netlink_bind && groups) { | |
1494 | int group; | |
1495 | ||
1496 | for (group = 0; group < nlk->ngroups; group++) { | |
1497 | if (!test_bit(group, &groups)) | |
1498 | continue; | |
8b7c36d8 | 1499 | err = nlk->netlink_bind(net, group + 1); |
4f520900 RGB |
1500 | if (!err) |
1501 | continue; | |
023e2cfa | 1502 | netlink_undo_bind(group, groups, sk); |
4f520900 RGB |
1503 | return err; |
1504 | } | |
1505 | } | |
1506 | ||
1507 | if (!nlk->portid) { | |
1da177e4 | 1508 | err = nladdr->nl_pid ? |
8ea65f4a | 1509 | netlink_insert(sk, nladdr->nl_pid) : |
1da177e4 | 1510 | netlink_autobind(sock); |
4f520900 | 1511 | if (err) { |
023e2cfa | 1512 | netlink_undo_bind(nlk->ngroups, groups, sk); |
1da177e4 | 1513 | return err; |
4f520900 | 1514 | } |
1da177e4 LT |
1515 | } |
1516 | ||
4f520900 | 1517 | if (!groups && (nlk->groups == NULL || !(u32)nlk->groups[0])) |
1da177e4 LT |
1518 | return 0; |
1519 | ||
1520 | netlink_table_grab(); | |
f7fa9b10 | 1521 | netlink_update_subscriptions(sk, nlk->subscriptions + |
4f520900 | 1522 | hweight32(groups) - |
746fac4d | 1523 | hweight32(nlk->groups[0])); |
4f520900 | 1524 | nlk->groups[0] = (nlk->groups[0] & ~0xffffffffUL) | groups; |
4277a083 | 1525 | netlink_update_listeners(sk); |
1da177e4 LT |
1526 | netlink_table_ungrab(); |
1527 | ||
1528 | return 0; | |
1529 | } | |
1530 | ||
1531 | static int netlink_connect(struct socket *sock, struct sockaddr *addr, | |
1532 | int alen, int flags) | |
1533 | { | |
1534 | int err = 0; | |
1535 | struct sock *sk = sock->sk; | |
1536 | struct netlink_sock *nlk = nlk_sk(sk); | |
6ac552fd | 1537 | struct sockaddr_nl *nladdr = (struct sockaddr_nl *)addr; |
1da177e4 | 1538 | |
6503d961 CG |
1539 | if (alen < sizeof(addr->sa_family)) |
1540 | return -EINVAL; | |
1541 | ||
1da177e4 LT |
1542 | if (addr->sa_family == AF_UNSPEC) { |
1543 | sk->sk_state = NETLINK_UNCONNECTED; | |
15e47304 | 1544 | nlk->dst_portid = 0; |
d629b836 | 1545 | nlk->dst_group = 0; |
1da177e4 LT |
1546 | return 0; |
1547 | } | |
1548 | if (addr->sa_family != AF_NETLINK) | |
1549 | return -EINVAL; | |
1550 | ||
46833a86 | 1551 | if ((nladdr->nl_groups || nladdr->nl_pid) && |
5187cd05 | 1552 | !netlink_allowed(sock, NL_CFG_F_NONROOT_SEND)) |
1da177e4 LT |
1553 | return -EPERM; |
1554 | ||
15e47304 | 1555 | if (!nlk->portid) |
1da177e4 LT |
1556 | err = netlink_autobind(sock); |
1557 | ||
1558 | if (err == 0) { | |
1559 | sk->sk_state = NETLINK_CONNECTED; | |
15e47304 | 1560 | nlk->dst_portid = nladdr->nl_pid; |
d629b836 | 1561 | nlk->dst_group = ffs(nladdr->nl_groups); |
1da177e4 LT |
1562 | } |
1563 | ||
1564 | return err; | |
1565 | } | |
1566 | ||
6ac552fd PM |
1567 | static int netlink_getname(struct socket *sock, struct sockaddr *addr, |
1568 | int *addr_len, int peer) | |
1da177e4 LT |
1569 | { |
1570 | struct sock *sk = sock->sk; | |
1571 | struct netlink_sock *nlk = nlk_sk(sk); | |
13cfa97b | 1572 | DECLARE_SOCKADDR(struct sockaddr_nl *, nladdr, addr); |
746fac4d | 1573 | |
1da177e4 LT |
1574 | nladdr->nl_family = AF_NETLINK; |
1575 | nladdr->nl_pad = 0; | |
1576 | *addr_len = sizeof(*nladdr); | |
1577 | ||
1578 | if (peer) { | |
15e47304 | 1579 | nladdr->nl_pid = nlk->dst_portid; |
d629b836 | 1580 | nladdr->nl_groups = netlink_group_mask(nlk->dst_group); |
1da177e4 | 1581 | } else { |
15e47304 | 1582 | nladdr->nl_pid = nlk->portid; |
513c2500 | 1583 | nladdr->nl_groups = nlk->groups ? nlk->groups[0] : 0; |
1da177e4 LT |
1584 | } |
1585 | return 0; | |
1586 | } | |
1587 | ||
15e47304 | 1588 | static struct sock *netlink_getsockbyportid(struct sock *ssk, u32 portid) |
1da177e4 | 1589 | { |
1da177e4 LT |
1590 | struct sock *sock; |
1591 | struct netlink_sock *nlk; | |
1592 | ||
15e47304 | 1593 | sock = netlink_lookup(sock_net(ssk), ssk->sk_protocol, portid); |
1da177e4 LT |
1594 | if (!sock) |
1595 | return ERR_PTR(-ECONNREFUSED); | |
1596 | ||
1597 | /* Don't bother queuing skb if kernel socket has no input function */ | |
1598 | nlk = nlk_sk(sock); | |
cd40b7d3 | 1599 | if (sock->sk_state == NETLINK_CONNECTED && |
15e47304 | 1600 | nlk->dst_portid != nlk_sk(ssk)->portid) { |
1da177e4 LT |
1601 | sock_put(sock); |
1602 | return ERR_PTR(-ECONNREFUSED); | |
1603 | } | |
1604 | return sock; | |
1605 | } | |
1606 | ||
1607 | struct sock *netlink_getsockbyfilp(struct file *filp) | |
1608 | { | |
496ad9aa | 1609 | struct inode *inode = file_inode(filp); |
1da177e4 LT |
1610 | struct sock *sock; |
1611 | ||
1612 | if (!S_ISSOCK(inode->i_mode)) | |
1613 | return ERR_PTR(-ENOTSOCK); | |
1614 | ||
1615 | sock = SOCKET_I(inode)->sk; | |
1616 | if (sock->sk_family != AF_NETLINK) | |
1617 | return ERR_PTR(-EINVAL); | |
1618 | ||
1619 | sock_hold(sock); | |
1620 | return sock; | |
1621 | } | |
1622 | ||
3a36515f PN |
1623 | static struct sk_buff *netlink_alloc_large_skb(unsigned int size, |
1624 | int broadcast) | |
c05cdb1b PNA |
1625 | { |
1626 | struct sk_buff *skb; | |
1627 | void *data; | |
1628 | ||
3a36515f | 1629 | if (size <= NLMSG_GOODSIZE || broadcast) |
c05cdb1b PNA |
1630 | return alloc_skb(size, GFP_KERNEL); |
1631 | ||
3a36515f PN |
1632 | size = SKB_DATA_ALIGN(size) + |
1633 | SKB_DATA_ALIGN(sizeof(struct skb_shared_info)); | |
c05cdb1b PNA |
1634 | |
1635 | data = vmalloc(size); | |
1636 | if (data == NULL) | |
3a36515f | 1637 | return NULL; |
c05cdb1b | 1638 | |
2ea2f62c | 1639 | skb = __build_skb(data, size); |
3a36515f PN |
1640 | if (skb == NULL) |
1641 | vfree(data); | |
2ea2f62c | 1642 | else |
3a36515f | 1643 | skb->destructor = netlink_skb_destructor; |
c05cdb1b PNA |
1644 | |
1645 | return skb; | |
c05cdb1b PNA |
1646 | } |
1647 | ||
1da177e4 LT |
1648 | /* |
1649 | * Attach a skb to a netlink socket. | |
1650 | * The caller must hold a reference to the destination socket. On error, the | |
1651 | * reference is dropped. The skb is not send to the destination, just all | |
1652 | * all error checks are performed and memory in the queue is reserved. | |
1653 | * Return values: | |
1654 | * < 0: error. skb freed, reference to sock dropped. | |
1655 | * 0: continue | |
1656 | * 1: repeat lookup - reference dropped while waiting for socket memory. | |
1657 | */ | |
9457afee | 1658 | int netlink_attachskb(struct sock *sk, struct sk_buff *skb, |
c3d8d1e3 | 1659 | long *timeo, struct sock *ssk) |
1da177e4 LT |
1660 | { |
1661 | struct netlink_sock *nlk; | |
1662 | ||
1663 | nlk = nlk_sk(sk); | |
1664 | ||
5fd96123 | 1665 | if ((atomic_read(&sk->sk_rmem_alloc) > sk->sk_rcvbuf || |
cc3a572f | 1666 | test_bit(NETLINK_S_CONGESTED, &nlk->state)) && |
5fd96123 | 1667 | !netlink_skb_is_mmaped(skb)) { |
1da177e4 | 1668 | DECLARE_WAITQUEUE(wait, current); |
c3d8d1e3 | 1669 | if (!*timeo) { |
aed81560 | 1670 | if (!ssk || netlink_is_kernel(ssk)) |
1da177e4 LT |
1671 | netlink_overrun(sk); |
1672 | sock_put(sk); | |
1673 | kfree_skb(skb); | |
1674 | return -EAGAIN; | |
1675 | } | |
1676 | ||
1677 | __set_current_state(TASK_INTERRUPTIBLE); | |
1678 | add_wait_queue(&nlk->wait, &wait); | |
1679 | ||
1680 | if ((atomic_read(&sk->sk_rmem_alloc) > sk->sk_rcvbuf || | |
cc3a572f | 1681 | test_bit(NETLINK_S_CONGESTED, &nlk->state)) && |
1da177e4 | 1682 | !sock_flag(sk, SOCK_DEAD)) |
c3d8d1e3 | 1683 | *timeo = schedule_timeout(*timeo); |
1da177e4 LT |
1684 | |
1685 | __set_current_state(TASK_RUNNING); | |
1686 | remove_wait_queue(&nlk->wait, &wait); | |
1687 | sock_put(sk); | |
1688 | ||
1689 | if (signal_pending(current)) { | |
1690 | kfree_skb(skb); | |
c3d8d1e3 | 1691 | return sock_intr_errno(*timeo); |
1da177e4 LT |
1692 | } |
1693 | return 1; | |
1694 | } | |
cf0a018a | 1695 | netlink_skb_set_owner_r(skb, sk); |
1da177e4 LT |
1696 | return 0; |
1697 | } | |
1698 | ||
4a7e7c2a | 1699 | static int __netlink_sendskb(struct sock *sk, struct sk_buff *skb) |
1da177e4 | 1700 | { |
1da177e4 LT |
1701 | int len = skb->len; |
1702 | ||
bcbde0d4 DB |
1703 | netlink_deliver_tap(skb); |
1704 | ||
f9c22888 PM |
1705 | #ifdef CONFIG_NETLINK_MMAP |
1706 | if (netlink_skb_is_mmaped(skb)) | |
1707 | netlink_queue_mmaped_skb(sk, skb); | |
1708 | else if (netlink_rx_is_mmaped(sk)) | |
1709 | netlink_ring_set_copied(sk, skb); | |
1710 | else | |
1711 | #endif /* CONFIG_NETLINK_MMAP */ | |
1712 | skb_queue_tail(&sk->sk_receive_queue, skb); | |
676d2369 | 1713 | sk->sk_data_ready(sk); |
4a7e7c2a ED |
1714 | return len; |
1715 | } | |
1716 | ||
1717 | int netlink_sendskb(struct sock *sk, struct sk_buff *skb) | |
1718 | { | |
1719 | int len = __netlink_sendskb(sk, skb); | |
1720 | ||
1da177e4 LT |
1721 | sock_put(sk); |
1722 | return len; | |
1723 | } | |
1724 | ||
1725 | void netlink_detachskb(struct sock *sk, struct sk_buff *skb) | |
1726 | { | |
1727 | kfree_skb(skb); | |
1728 | sock_put(sk); | |
1729 | } | |
1730 | ||
b57ef81f | 1731 | static struct sk_buff *netlink_trim(struct sk_buff *skb, gfp_t allocation) |
1da177e4 LT |
1732 | { |
1733 | int delta; | |
1734 | ||
1298ca46 | 1735 | WARN_ON(skb->sk != NULL); |
5fd96123 PM |
1736 | if (netlink_skb_is_mmaped(skb)) |
1737 | return skb; | |
1da177e4 | 1738 | |
4305b541 | 1739 | delta = skb->end - skb->tail; |
c05cdb1b | 1740 | if (is_vmalloc_addr(skb->head) || delta * 2 < skb->truesize) |
1da177e4 LT |
1741 | return skb; |
1742 | ||
1743 | if (skb_shared(skb)) { | |
1744 | struct sk_buff *nskb = skb_clone(skb, allocation); | |
1745 | if (!nskb) | |
1746 | return skb; | |
8460c00f | 1747 | consume_skb(skb); |
1da177e4 LT |
1748 | skb = nskb; |
1749 | } | |
1750 | ||
1751 | if (!pskb_expand_head(skb, 0, -delta, allocation)) | |
1752 | skb->truesize -= delta; | |
1753 | ||
1754 | return skb; | |
1755 | } | |
1756 | ||
3fbc2905 EB |
1757 | static int netlink_unicast_kernel(struct sock *sk, struct sk_buff *skb, |
1758 | struct sock *ssk) | |
cd40b7d3 DL |
1759 | { |
1760 | int ret; | |
1761 | struct netlink_sock *nlk = nlk_sk(sk); | |
1762 | ||
1763 | ret = -ECONNREFUSED; | |
1764 | if (nlk->netlink_rcv != NULL) { | |
1765 | ret = skb->len; | |
cf0a018a | 1766 | netlink_skb_set_owner_r(skb, sk); |
e32123e5 | 1767 | NETLINK_CB(skb).sk = ssk; |
73bfd370 | 1768 | netlink_deliver_tap_kernel(sk, ssk, skb); |
cd40b7d3 | 1769 | nlk->netlink_rcv(skb); |
bfb253c9 ED |
1770 | consume_skb(skb); |
1771 | } else { | |
1772 | kfree_skb(skb); | |
cd40b7d3 | 1773 | } |
cd40b7d3 DL |
1774 | sock_put(sk); |
1775 | return ret; | |
1776 | } | |
1777 | ||
1778 | int netlink_unicast(struct sock *ssk, struct sk_buff *skb, | |
15e47304 | 1779 | u32 portid, int nonblock) |
1da177e4 LT |
1780 | { |
1781 | struct sock *sk; | |
1782 | int err; | |
1783 | long timeo; | |
1784 | ||
1785 | skb = netlink_trim(skb, gfp_any()); | |
1786 | ||
1787 | timeo = sock_sndtimeo(ssk, nonblock); | |
1788 | retry: | |
15e47304 | 1789 | sk = netlink_getsockbyportid(ssk, portid); |
1da177e4 LT |
1790 | if (IS_ERR(sk)) { |
1791 | kfree_skb(skb); | |
1792 | return PTR_ERR(sk); | |
1793 | } | |
cd40b7d3 | 1794 | if (netlink_is_kernel(sk)) |
3fbc2905 | 1795 | return netlink_unicast_kernel(sk, skb, ssk); |
cd40b7d3 | 1796 | |
b1153f29 | 1797 | if (sk_filter(sk, skb)) { |
84874607 | 1798 | err = skb->len; |
b1153f29 SH |
1799 | kfree_skb(skb); |
1800 | sock_put(sk); | |
1801 | return err; | |
1802 | } | |
1803 | ||
9457afee | 1804 | err = netlink_attachskb(sk, skb, &timeo, ssk); |
1da177e4 LT |
1805 | if (err == 1) |
1806 | goto retry; | |
1807 | if (err) | |
1808 | return err; | |
1809 | ||
7ee015e0 | 1810 | return netlink_sendskb(sk, skb); |
1da177e4 | 1811 | } |
6ac552fd | 1812 | EXPORT_SYMBOL(netlink_unicast); |
1da177e4 | 1813 | |
f9c22888 PM |
1814 | struct sk_buff *netlink_alloc_skb(struct sock *ssk, unsigned int size, |
1815 | u32 dst_portid, gfp_t gfp_mask) | |
1816 | { | |
1817 | #ifdef CONFIG_NETLINK_MMAP | |
1818 | struct sock *sk = NULL; | |
1819 | struct sk_buff *skb; | |
1820 | struct netlink_ring *ring; | |
1821 | struct nl_mmap_hdr *hdr; | |
1822 | unsigned int maxlen; | |
1823 | ||
1824 | sk = netlink_getsockbyportid(ssk, dst_portid); | |
1825 | if (IS_ERR(sk)) | |
1826 | goto out; | |
1827 | ||
1828 | ring = &nlk_sk(sk)->rx_ring; | |
1829 | /* fast-path without atomic ops for common case: non-mmaped receiver */ | |
1830 | if (ring->pg_vec == NULL) | |
1831 | goto out_put; | |
1832 | ||
aae9f0e2 TG |
1833 | if (ring->frame_size - NL_MMAP_HDRLEN < size) |
1834 | goto out_put; | |
1835 | ||
f9c22888 PM |
1836 | skb = alloc_skb_head(gfp_mask); |
1837 | if (skb == NULL) | |
1838 | goto err1; | |
1839 | ||
1840 | spin_lock_bh(&sk->sk_receive_queue.lock); | |
1841 | /* check again under lock */ | |
1842 | if (ring->pg_vec == NULL) | |
1843 | goto out_free; | |
1844 | ||
aae9f0e2 | 1845 | /* check again under lock */ |
f9c22888 PM |
1846 | maxlen = ring->frame_size - NL_MMAP_HDRLEN; |
1847 | if (maxlen < size) | |
1848 | goto out_free; | |
1849 | ||
1850 | netlink_forward_ring(ring); | |
1851 | hdr = netlink_current_frame(ring, NL_MMAP_STATUS_UNUSED); | |
1852 | if (hdr == NULL) | |
1853 | goto err2; | |
1854 | netlink_ring_setup_skb(skb, sk, ring, hdr); | |
1855 | netlink_set_status(hdr, NL_MMAP_STATUS_RESERVED); | |
1856 | atomic_inc(&ring->pending); | |
1857 | netlink_increment_head(ring); | |
1858 | ||
1859 | spin_unlock_bh(&sk->sk_receive_queue.lock); | |
1860 | return skb; | |
1861 | ||
1862 | err2: | |
1863 | kfree_skb(skb); | |
1864 | spin_unlock_bh(&sk->sk_receive_queue.lock); | |
cd1df525 | 1865 | netlink_overrun(sk); |
f9c22888 PM |
1866 | err1: |
1867 | sock_put(sk); | |
1868 | return NULL; | |
1869 | ||
1870 | out_free: | |
1871 | kfree_skb(skb); | |
1872 | spin_unlock_bh(&sk->sk_receive_queue.lock); | |
1873 | out_put: | |
1874 | sock_put(sk); | |
1875 | out: | |
1876 | #endif | |
1877 | return alloc_skb(size, gfp_mask); | |
1878 | } | |
1879 | EXPORT_SYMBOL_GPL(netlink_alloc_skb); | |
1880 | ||
4277a083 PM |
1881 | int netlink_has_listeners(struct sock *sk, unsigned int group) |
1882 | { | |
1883 | int res = 0; | |
5c398dc8 | 1884 | struct listeners *listeners; |
4277a083 | 1885 | |
aed81560 | 1886 | BUG_ON(!netlink_is_kernel(sk)); |
b4ff4f04 JB |
1887 | |
1888 | rcu_read_lock(); | |
1889 | listeners = rcu_dereference(nl_table[sk->sk_protocol].listeners); | |
1890 | ||
6d772ac5 | 1891 | if (listeners && group - 1 < nl_table[sk->sk_protocol].groups) |
5c398dc8 | 1892 | res = test_bit(group - 1, listeners->masks); |
b4ff4f04 JB |
1893 | |
1894 | rcu_read_unlock(); | |
1895 | ||
4277a083 PM |
1896 | return res; |
1897 | } | |
1898 | EXPORT_SYMBOL_GPL(netlink_has_listeners); | |
1899 | ||
b57ef81f | 1900 | static int netlink_broadcast_deliver(struct sock *sk, struct sk_buff *skb) |
1da177e4 LT |
1901 | { |
1902 | struct netlink_sock *nlk = nlk_sk(sk); | |
1903 | ||
1904 | if (atomic_read(&sk->sk_rmem_alloc) <= sk->sk_rcvbuf && | |
cc3a572f | 1905 | !test_bit(NETLINK_S_CONGESTED, &nlk->state)) { |
cf0a018a | 1906 | netlink_skb_set_owner_r(skb, sk); |
4a7e7c2a | 1907 | __netlink_sendskb(sk, skb); |
2c645800 | 1908 | return atomic_read(&sk->sk_rmem_alloc) > (sk->sk_rcvbuf >> 1); |
1da177e4 LT |
1909 | } |
1910 | return -1; | |
1911 | } | |
1912 | ||
1913 | struct netlink_broadcast_data { | |
1914 | struct sock *exclude_sk; | |
b4b51029 | 1915 | struct net *net; |
15e47304 | 1916 | u32 portid; |
1da177e4 LT |
1917 | u32 group; |
1918 | int failure; | |
ff491a73 | 1919 | int delivery_failure; |
1da177e4 LT |
1920 | int congested; |
1921 | int delivered; | |
7d877f3b | 1922 | gfp_t allocation; |
1da177e4 | 1923 | struct sk_buff *skb, *skb2; |
910a7e90 EB |
1924 | int (*tx_filter)(struct sock *dsk, struct sk_buff *skb, void *data); |
1925 | void *tx_data; | |
1da177e4 LT |
1926 | }; |
1927 | ||
46c9521f RR |
1928 | static void do_one_broadcast(struct sock *sk, |
1929 | struct netlink_broadcast_data *p) | |
1da177e4 LT |
1930 | { |
1931 | struct netlink_sock *nlk = nlk_sk(sk); | |
1932 | int val; | |
1933 | ||
1934 | if (p->exclude_sk == sk) | |
46c9521f | 1935 | return; |
1da177e4 | 1936 | |
15e47304 | 1937 | if (nlk->portid == p->portid || p->group - 1 >= nlk->ngroups || |
f7fa9b10 | 1938 | !test_bit(p->group - 1, nlk->groups)) |
46c9521f | 1939 | return; |
1da177e4 | 1940 | |
59324cf3 ND |
1941 | if (!net_eq(sock_net(sk), p->net)) { |
1942 | if (!(nlk->flags & NETLINK_F_LISTEN_ALL_NSID)) | |
1943 | return; | |
1944 | ||
1945 | if (!peernet_has_id(sock_net(sk), p->net)) | |
1946 | return; | |
1947 | ||
1948 | if (!file_ns_capable(sk->sk_socket->file, p->net->user_ns, | |
1949 | CAP_NET_BROADCAST)) | |
1950 | return; | |
1951 | } | |
b4b51029 | 1952 | |
1da177e4 LT |
1953 | if (p->failure) { |
1954 | netlink_overrun(sk); | |
46c9521f | 1955 | return; |
1da177e4 LT |
1956 | } |
1957 | ||
1958 | sock_hold(sk); | |
1959 | if (p->skb2 == NULL) { | |
68acc024 | 1960 | if (skb_shared(p->skb)) { |
1da177e4 LT |
1961 | p->skb2 = skb_clone(p->skb, p->allocation); |
1962 | } else { | |
68acc024 TC |
1963 | p->skb2 = skb_get(p->skb); |
1964 | /* | |
1965 | * skb ownership may have been set when | |
1966 | * delivered to a previous socket. | |
1967 | */ | |
1968 | skb_orphan(p->skb2); | |
1da177e4 LT |
1969 | } |
1970 | } | |
1971 | if (p->skb2 == NULL) { | |
1972 | netlink_overrun(sk); | |
1973 | /* Clone failed. Notify ALL listeners. */ | |
1974 | p->failure = 1; | |
cc3a572f | 1975 | if (nlk->flags & NETLINK_F_BROADCAST_SEND_ERROR) |
be0c22a4 | 1976 | p->delivery_failure = 1; |
59324cf3 ND |
1977 | goto out; |
1978 | } | |
1979 | if (p->tx_filter && p->tx_filter(sk, p->skb2, p->tx_data)) { | |
910a7e90 EB |
1980 | kfree_skb(p->skb2); |
1981 | p->skb2 = NULL; | |
59324cf3 ND |
1982 | goto out; |
1983 | } | |
1984 | if (sk_filter(sk, p->skb2)) { | |
b1153f29 SH |
1985 | kfree_skb(p->skb2); |
1986 | p->skb2 = NULL; | |
59324cf3 ND |
1987 | goto out; |
1988 | } | |
1989 | NETLINK_CB(p->skb2).nsid = peernet2id(sock_net(sk), p->net); | |
1990 | NETLINK_CB(p->skb2).nsid_is_set = true; | |
1991 | val = netlink_broadcast_deliver(sk, p->skb2); | |
1992 | if (val < 0) { | |
1da177e4 | 1993 | netlink_overrun(sk); |
cc3a572f | 1994 | if (nlk->flags & NETLINK_F_BROADCAST_SEND_ERROR) |
be0c22a4 | 1995 | p->delivery_failure = 1; |
1da177e4 LT |
1996 | } else { |
1997 | p->congested |= val; | |
1998 | p->delivered = 1; | |
1999 | p->skb2 = NULL; | |
2000 | } | |
59324cf3 | 2001 | out: |
1da177e4 | 2002 | sock_put(sk); |
1da177e4 LT |
2003 | } |
2004 | ||
15e47304 | 2005 | int netlink_broadcast_filtered(struct sock *ssk, struct sk_buff *skb, u32 portid, |
910a7e90 EB |
2006 | u32 group, gfp_t allocation, |
2007 | int (*filter)(struct sock *dsk, struct sk_buff *skb, void *data), | |
2008 | void *filter_data) | |
1da177e4 | 2009 | { |
3b1e0a65 | 2010 | struct net *net = sock_net(ssk); |
1da177e4 | 2011 | struct netlink_broadcast_data info; |
1da177e4 LT |
2012 | struct sock *sk; |
2013 | ||
2014 | skb = netlink_trim(skb, allocation); | |
2015 | ||
2016 | info.exclude_sk = ssk; | |
b4b51029 | 2017 | info.net = net; |
15e47304 | 2018 | info.portid = portid; |
1da177e4 LT |
2019 | info.group = group; |
2020 | info.failure = 0; | |
ff491a73 | 2021 | info.delivery_failure = 0; |
1da177e4 LT |
2022 | info.congested = 0; |
2023 | info.delivered = 0; | |
2024 | info.allocation = allocation; | |
2025 | info.skb = skb; | |
2026 | info.skb2 = NULL; | |
910a7e90 EB |
2027 | info.tx_filter = filter; |
2028 | info.tx_data = filter_data; | |
1da177e4 LT |
2029 | |
2030 | /* While we sleep in clone, do not allow to change socket list */ | |
2031 | ||
2032 | netlink_lock_table(); | |
2033 | ||
b67bfe0d | 2034 | sk_for_each_bound(sk, &nl_table[ssk->sk_protocol].mc_list) |
1da177e4 LT |
2035 | do_one_broadcast(sk, &info); |
2036 | ||
70d4bf6d | 2037 | consume_skb(skb); |
aa1c6a6f | 2038 | |
1da177e4 LT |
2039 | netlink_unlock_table(); |
2040 | ||
70d4bf6d NH |
2041 | if (info.delivery_failure) { |
2042 | kfree_skb(info.skb2); | |
ff491a73 | 2043 | return -ENOBUFS; |
658cb354 ED |
2044 | } |
2045 | consume_skb(info.skb2); | |
ff491a73 | 2046 | |
1da177e4 LT |
2047 | if (info.delivered) { |
2048 | if (info.congested && (allocation & __GFP_WAIT)) | |
2049 | yield(); | |
2050 | return 0; | |
2051 | } | |
1da177e4 LT |
2052 | return -ESRCH; |
2053 | } | |
910a7e90 EB |
2054 | EXPORT_SYMBOL(netlink_broadcast_filtered); |
2055 | ||
15e47304 | 2056 | int netlink_broadcast(struct sock *ssk, struct sk_buff *skb, u32 portid, |
910a7e90 EB |
2057 | u32 group, gfp_t allocation) |
2058 | { | |
15e47304 | 2059 | return netlink_broadcast_filtered(ssk, skb, portid, group, allocation, |
910a7e90 EB |
2060 | NULL, NULL); |
2061 | } | |
6ac552fd | 2062 | EXPORT_SYMBOL(netlink_broadcast); |
1da177e4 LT |
2063 | |
2064 | struct netlink_set_err_data { | |
2065 | struct sock *exclude_sk; | |
15e47304 | 2066 | u32 portid; |
1da177e4 LT |
2067 | u32 group; |
2068 | int code; | |
2069 | }; | |
2070 | ||
b57ef81f | 2071 | static int do_one_set_err(struct sock *sk, struct netlink_set_err_data *p) |
1da177e4 LT |
2072 | { |
2073 | struct netlink_sock *nlk = nlk_sk(sk); | |
1a50307b | 2074 | int ret = 0; |
1da177e4 LT |
2075 | |
2076 | if (sk == p->exclude_sk) | |
2077 | goto out; | |
2078 | ||
09ad9bc7 | 2079 | if (!net_eq(sock_net(sk), sock_net(p->exclude_sk))) |
b4b51029 EB |
2080 | goto out; |
2081 | ||
15e47304 | 2082 | if (nlk->portid == p->portid || p->group - 1 >= nlk->ngroups || |
f7fa9b10 | 2083 | !test_bit(p->group - 1, nlk->groups)) |
1da177e4 LT |
2084 | goto out; |
2085 | ||
cc3a572f | 2086 | if (p->code == ENOBUFS && nlk->flags & NETLINK_F_RECV_NO_ENOBUFS) { |
1a50307b PNA |
2087 | ret = 1; |
2088 | goto out; | |
2089 | } | |
2090 | ||
1da177e4 LT |
2091 | sk->sk_err = p->code; |
2092 | sk->sk_error_report(sk); | |
2093 | out: | |
1a50307b | 2094 | return ret; |
1da177e4 LT |
2095 | } |
2096 | ||
4843b93c PNA |
2097 | /** |
2098 | * netlink_set_err - report error to broadcast listeners | |
2099 | * @ssk: the kernel netlink socket, as returned by netlink_kernel_create() | |
15e47304 | 2100 | * @portid: the PORTID of a process that we want to skip (if any) |
840e93f2 | 2101 | * @group: the broadcast group that will notice the error |
4843b93c | 2102 | * @code: error code, must be negative (as usual in kernelspace) |
1a50307b PNA |
2103 | * |
2104 | * This function returns the number of broadcast listeners that have set the | |
cc3a572f | 2105 | * NETLINK_NO_ENOBUFS socket option. |
4843b93c | 2106 | */ |
15e47304 | 2107 | int netlink_set_err(struct sock *ssk, u32 portid, u32 group, int code) |
1da177e4 LT |
2108 | { |
2109 | struct netlink_set_err_data info; | |
1da177e4 | 2110 | struct sock *sk; |
1a50307b | 2111 | int ret = 0; |
1da177e4 LT |
2112 | |
2113 | info.exclude_sk = ssk; | |
15e47304 | 2114 | info.portid = portid; |
1da177e4 | 2115 | info.group = group; |
4843b93c PNA |
2116 | /* sk->sk_err wants a positive error value */ |
2117 | info.code = -code; | |
1da177e4 LT |
2118 | |
2119 | read_lock(&nl_table_lock); | |
2120 | ||
b67bfe0d | 2121 | sk_for_each_bound(sk, &nl_table[ssk->sk_protocol].mc_list) |
1a50307b | 2122 | ret += do_one_set_err(sk, &info); |
1da177e4 LT |
2123 | |
2124 | read_unlock(&nl_table_lock); | |
1a50307b | 2125 | return ret; |
1da177e4 | 2126 | } |
dd5b6ce6 | 2127 | EXPORT_SYMBOL(netlink_set_err); |
1da177e4 | 2128 | |
84659eb5 JB |
2129 | /* must be called with netlink table grabbed */ |
2130 | static void netlink_update_socket_mc(struct netlink_sock *nlk, | |
2131 | unsigned int group, | |
2132 | int is_new) | |
2133 | { | |
2134 | int old, new = !!is_new, subscriptions; | |
2135 | ||
2136 | old = test_bit(group - 1, nlk->groups); | |
2137 | subscriptions = nlk->subscriptions - old + new; | |
2138 | if (new) | |
2139 | __set_bit(group - 1, nlk->groups); | |
2140 | else | |
2141 | __clear_bit(group - 1, nlk->groups); | |
2142 | netlink_update_subscriptions(&nlk->sk, subscriptions); | |
2143 | netlink_update_listeners(&nlk->sk); | |
2144 | } | |
2145 | ||
9a4595bc | 2146 | static int netlink_setsockopt(struct socket *sock, int level, int optname, |
b7058842 | 2147 | char __user *optval, unsigned int optlen) |
9a4595bc PM |
2148 | { |
2149 | struct sock *sk = sock->sk; | |
2150 | struct netlink_sock *nlk = nlk_sk(sk); | |
eb496534 JB |
2151 | unsigned int val = 0; |
2152 | int err; | |
9a4595bc PM |
2153 | |
2154 | if (level != SOL_NETLINK) | |
2155 | return -ENOPROTOOPT; | |
2156 | ||
ccdfcc39 PM |
2157 | if (optname != NETLINK_RX_RING && optname != NETLINK_TX_RING && |
2158 | optlen >= sizeof(int) && | |
eb496534 | 2159 | get_user(val, (unsigned int __user *)optval)) |
9a4595bc PM |
2160 | return -EFAULT; |
2161 | ||
2162 | switch (optname) { | |
2163 | case NETLINK_PKTINFO: | |
2164 | if (val) | |
cc3a572f | 2165 | nlk->flags |= NETLINK_F_RECV_PKTINFO; |
9a4595bc | 2166 | else |
cc3a572f | 2167 | nlk->flags &= ~NETLINK_F_RECV_PKTINFO; |
9a4595bc PM |
2168 | err = 0; |
2169 | break; | |
2170 | case NETLINK_ADD_MEMBERSHIP: | |
2171 | case NETLINK_DROP_MEMBERSHIP: { | |
5187cd05 | 2172 | if (!netlink_allowed(sock, NL_CFG_F_NONROOT_RECV)) |
9a4595bc | 2173 | return -EPERM; |
b4ff4f04 JB |
2174 | err = netlink_realloc_groups(sk); |
2175 | if (err) | |
2176 | return err; | |
9a4595bc PM |
2177 | if (!val || val - 1 >= nlk->ngroups) |
2178 | return -EINVAL; | |
7774d5e0 | 2179 | if (optname == NETLINK_ADD_MEMBERSHIP && nlk->netlink_bind) { |
023e2cfa | 2180 | err = nlk->netlink_bind(sock_net(sk), val); |
4f520900 RGB |
2181 | if (err) |
2182 | return err; | |
2183 | } | |
9a4595bc | 2184 | netlink_table_grab(); |
84659eb5 JB |
2185 | netlink_update_socket_mc(nlk, val, |
2186 | optname == NETLINK_ADD_MEMBERSHIP); | |
9a4595bc | 2187 | netlink_table_ungrab(); |
7774d5e0 | 2188 | if (optname == NETLINK_DROP_MEMBERSHIP && nlk->netlink_unbind) |
023e2cfa | 2189 | nlk->netlink_unbind(sock_net(sk), val); |
03292745 | 2190 | |
9a4595bc PM |
2191 | err = 0; |
2192 | break; | |
2193 | } | |
be0c22a4 PNA |
2194 | case NETLINK_BROADCAST_ERROR: |
2195 | if (val) | |
cc3a572f | 2196 | nlk->flags |= NETLINK_F_BROADCAST_SEND_ERROR; |
be0c22a4 | 2197 | else |
cc3a572f | 2198 | nlk->flags &= ~NETLINK_F_BROADCAST_SEND_ERROR; |
be0c22a4 PNA |
2199 | err = 0; |
2200 | break; | |
38938bfe PNA |
2201 | case NETLINK_NO_ENOBUFS: |
2202 | if (val) { | |
cc3a572f ND |
2203 | nlk->flags |= NETLINK_F_RECV_NO_ENOBUFS; |
2204 | clear_bit(NETLINK_S_CONGESTED, &nlk->state); | |
38938bfe | 2205 | wake_up_interruptible(&nlk->wait); |
658cb354 | 2206 | } else { |
cc3a572f | 2207 | nlk->flags &= ~NETLINK_F_RECV_NO_ENOBUFS; |
658cb354 | 2208 | } |
38938bfe PNA |
2209 | err = 0; |
2210 | break; | |
ccdfcc39 PM |
2211 | #ifdef CONFIG_NETLINK_MMAP |
2212 | case NETLINK_RX_RING: | |
2213 | case NETLINK_TX_RING: { | |
2214 | struct nl_mmap_req req; | |
2215 | ||
2216 | /* Rings might consume more memory than queue limits, require | |
2217 | * CAP_NET_ADMIN. | |
2218 | */ | |
2219 | if (!capable(CAP_NET_ADMIN)) | |
2220 | return -EPERM; | |
2221 | if (optlen < sizeof(req)) | |
2222 | return -EINVAL; | |
2223 | if (copy_from_user(&req, optval, sizeof(req))) | |
2224 | return -EFAULT; | |
2225 | err = netlink_set_ring(sk, &req, false, | |
2226 | optname == NETLINK_TX_RING); | |
2227 | break; | |
2228 | } | |
2229 | #endif /* CONFIG_NETLINK_MMAP */ | |
59324cf3 ND |
2230 | case NETLINK_LISTEN_ALL_NSID: |
2231 | if (!ns_capable(sock_net(sk)->user_ns, CAP_NET_BROADCAST)) | |
2232 | return -EPERM; | |
2233 | ||
2234 | if (val) | |
2235 | nlk->flags |= NETLINK_F_LISTEN_ALL_NSID; | |
2236 | else | |
2237 | nlk->flags &= ~NETLINK_F_LISTEN_ALL_NSID; | |
2238 | err = 0; | |
2239 | break; | |
9a4595bc PM |
2240 | default: |
2241 | err = -ENOPROTOOPT; | |
2242 | } | |
2243 | return err; | |
2244 | } | |
2245 | ||
2246 | static int netlink_getsockopt(struct socket *sock, int level, int optname, | |
746fac4d | 2247 | char __user *optval, int __user *optlen) |
9a4595bc PM |
2248 | { |
2249 | struct sock *sk = sock->sk; | |
2250 | struct netlink_sock *nlk = nlk_sk(sk); | |
2251 | int len, val, err; | |
2252 | ||
2253 | if (level != SOL_NETLINK) | |
2254 | return -ENOPROTOOPT; | |
2255 | ||
2256 | if (get_user(len, optlen)) | |
2257 | return -EFAULT; | |
2258 | if (len < 0) | |
2259 | return -EINVAL; | |
2260 | ||
2261 | switch (optname) { | |
2262 | case NETLINK_PKTINFO: | |
2263 | if (len < sizeof(int)) | |
2264 | return -EINVAL; | |
2265 | len = sizeof(int); | |
cc3a572f | 2266 | val = nlk->flags & NETLINK_F_RECV_PKTINFO ? 1 : 0; |
a27b58fe HC |
2267 | if (put_user(len, optlen) || |
2268 | put_user(val, optval)) | |
2269 | return -EFAULT; | |
9a4595bc PM |
2270 | err = 0; |
2271 | break; | |
be0c22a4 PNA |
2272 | case NETLINK_BROADCAST_ERROR: |
2273 | if (len < sizeof(int)) | |
2274 | return -EINVAL; | |
2275 | len = sizeof(int); | |
cc3a572f | 2276 | val = nlk->flags & NETLINK_F_BROADCAST_SEND_ERROR ? 1 : 0; |
be0c22a4 PNA |
2277 | if (put_user(len, optlen) || |
2278 | put_user(val, optval)) | |
2279 | return -EFAULT; | |
2280 | err = 0; | |
2281 | break; | |
38938bfe PNA |
2282 | case NETLINK_NO_ENOBUFS: |
2283 | if (len < sizeof(int)) | |
2284 | return -EINVAL; | |
2285 | len = sizeof(int); | |
cc3a572f | 2286 | val = nlk->flags & NETLINK_F_RECV_NO_ENOBUFS ? 1 : 0; |
38938bfe PNA |
2287 | if (put_user(len, optlen) || |
2288 | put_user(val, optval)) | |
2289 | return -EFAULT; | |
2290 | err = 0; | |
2291 | break; | |
9a4595bc PM |
2292 | default: |
2293 | err = -ENOPROTOOPT; | |
2294 | } | |
2295 | return err; | |
2296 | } | |
2297 | ||
2298 | static void netlink_cmsg_recv_pktinfo(struct msghdr *msg, struct sk_buff *skb) | |
2299 | { | |
2300 | struct nl_pktinfo info; | |
2301 | ||
2302 | info.group = NETLINK_CB(skb).dst_group; | |
2303 | put_cmsg(msg, SOL_NETLINK, NETLINK_PKTINFO, sizeof(info), &info); | |
2304 | } | |
2305 | ||
59324cf3 ND |
2306 | static void netlink_cmsg_listen_all_nsid(struct sock *sk, struct msghdr *msg, |
2307 | struct sk_buff *skb) | |
2308 | { | |
2309 | if (!NETLINK_CB(skb).nsid_is_set) | |
2310 | return; | |
2311 | ||
2312 | put_cmsg(msg, SOL_NETLINK, NETLINK_LISTEN_ALL_NSID, sizeof(int), | |
2313 | &NETLINK_CB(skb).nsid); | |
2314 | } | |
2315 | ||
1b784140 | 2316 | static int netlink_sendmsg(struct socket *sock, struct msghdr *msg, size_t len) |
1da177e4 | 2317 | { |
1da177e4 LT |
2318 | struct sock *sk = sock->sk; |
2319 | struct netlink_sock *nlk = nlk_sk(sk); | |
342dfc30 | 2320 | DECLARE_SOCKADDR(struct sockaddr_nl *, addr, msg->msg_name); |
15e47304 | 2321 | u32 dst_portid; |
d629b836 | 2322 | u32 dst_group; |
1da177e4 LT |
2323 | struct sk_buff *skb; |
2324 | int err; | |
2325 | struct scm_cookie scm; | |
2d7a85f4 | 2326 | u32 netlink_skb_flags = 0; |
1da177e4 LT |
2327 | |
2328 | if (msg->msg_flags&MSG_OOB) | |
2329 | return -EOPNOTSUPP; | |
2330 | ||
7cc05662 | 2331 | err = scm_send(sock, msg, &scm, true); |
1da177e4 LT |
2332 | if (err < 0) |
2333 | return err; | |
2334 | ||
2335 | if (msg->msg_namelen) { | |
b47030c7 | 2336 | err = -EINVAL; |
1da177e4 | 2337 | if (addr->nl_family != AF_NETLINK) |
b47030c7 | 2338 | goto out; |
15e47304 | 2339 | dst_portid = addr->nl_pid; |
d629b836 | 2340 | dst_group = ffs(addr->nl_groups); |
b47030c7 | 2341 | err = -EPERM; |
15e47304 | 2342 | if ((dst_group || dst_portid) && |
5187cd05 | 2343 | !netlink_allowed(sock, NL_CFG_F_NONROOT_SEND)) |
b47030c7 | 2344 | goto out; |
2d7a85f4 | 2345 | netlink_skb_flags |= NETLINK_SKB_DST; |
1da177e4 | 2346 | } else { |
15e47304 | 2347 | dst_portid = nlk->dst_portid; |
d629b836 | 2348 | dst_group = nlk->dst_group; |
1da177e4 LT |
2349 | } |
2350 | ||
15e47304 | 2351 | if (!nlk->portid) { |
1da177e4 LT |
2352 | err = netlink_autobind(sock); |
2353 | if (err) | |
2354 | goto out; | |
2355 | } | |
2356 | ||
a8866ff6 AV |
2357 | /* It's a really convoluted way for userland to ask for mmaped |
2358 | * sendmsg(), but that's what we've got... | |
2359 | */ | |
5fd96123 | 2360 | if (netlink_tx_is_mmaped(sk) && |
a8866ff6 AV |
2361 | msg->msg_iter.type == ITER_IOVEC && |
2362 | msg->msg_iter.nr_segs == 1 && | |
c0371da6 | 2363 | msg->msg_iter.iov->iov_base == NULL) { |
5fd96123 | 2364 | err = netlink_mmap_sendmsg(sk, msg, dst_portid, dst_group, |
7cc05662 | 2365 | &scm); |
5fd96123 PM |
2366 | goto out; |
2367 | } | |
2368 | ||
1da177e4 LT |
2369 | err = -EMSGSIZE; |
2370 | if (len > sk->sk_sndbuf - 32) | |
2371 | goto out; | |
2372 | err = -ENOBUFS; | |
3a36515f | 2373 | skb = netlink_alloc_large_skb(len, dst_group); |
6ac552fd | 2374 | if (skb == NULL) |
1da177e4 LT |
2375 | goto out; |
2376 | ||
15e47304 | 2377 | NETLINK_CB(skb).portid = nlk->portid; |
d629b836 | 2378 | NETLINK_CB(skb).dst_group = dst_group; |
7cc05662 | 2379 | NETLINK_CB(skb).creds = scm.creds; |
2d7a85f4 | 2380 | NETLINK_CB(skb).flags = netlink_skb_flags; |
1da177e4 | 2381 | |
1da177e4 | 2382 | err = -EFAULT; |
6ce8e9ce | 2383 | if (memcpy_from_msg(skb_put(skb, len), msg, len)) { |
1da177e4 LT |
2384 | kfree_skb(skb); |
2385 | goto out; | |
2386 | } | |
2387 | ||
2388 | err = security_netlink_send(sk, skb); | |
2389 | if (err) { | |
2390 | kfree_skb(skb); | |
2391 | goto out; | |
2392 | } | |
2393 | ||
d629b836 | 2394 | if (dst_group) { |
1da177e4 | 2395 | atomic_inc(&skb->users); |
15e47304 | 2396 | netlink_broadcast(sk, skb, dst_portid, dst_group, GFP_KERNEL); |
1da177e4 | 2397 | } |
15e47304 | 2398 | err = netlink_unicast(sk, skb, dst_portid, msg->msg_flags&MSG_DONTWAIT); |
1da177e4 LT |
2399 | |
2400 | out: | |
7cc05662 | 2401 | scm_destroy(&scm); |
1da177e4 LT |
2402 | return err; |
2403 | } | |
2404 | ||
1b784140 | 2405 | static int netlink_recvmsg(struct socket *sock, struct msghdr *msg, size_t len, |
1da177e4 LT |
2406 | int flags) |
2407 | { | |
1da177e4 LT |
2408 | struct scm_cookie scm; |
2409 | struct sock *sk = sock->sk; | |
2410 | struct netlink_sock *nlk = nlk_sk(sk); | |
2411 | int noblock = flags&MSG_DONTWAIT; | |
2412 | size_t copied; | |
68d6ac6d | 2413 | struct sk_buff *skb, *data_skb; |
b44d211e | 2414 | int err, ret; |
1da177e4 LT |
2415 | |
2416 | if (flags&MSG_OOB) | |
2417 | return -EOPNOTSUPP; | |
2418 | ||
2419 | copied = 0; | |
2420 | ||
6ac552fd PM |
2421 | skb = skb_recv_datagram(sk, flags, noblock, &err); |
2422 | if (skb == NULL) | |
1da177e4 LT |
2423 | goto out; |
2424 | ||
68d6ac6d JB |
2425 | data_skb = skb; |
2426 | ||
1dacc76d JB |
2427 | #ifdef CONFIG_COMPAT_NETLINK_MESSAGES |
2428 | if (unlikely(skb_shinfo(skb)->frag_list)) { | |
1dacc76d | 2429 | /* |
68d6ac6d JB |
2430 | * If this skb has a frag_list, then here that means that we |
2431 | * will have to use the frag_list skb's data for compat tasks | |
2432 | * and the regular skb's data for normal (non-compat) tasks. | |
1dacc76d | 2433 | * |
68d6ac6d JB |
2434 | * If we need to send the compat skb, assign it to the |
2435 | * 'data_skb' variable so that it will be used below for data | |
2436 | * copying. We keep 'skb' for everything else, including | |
2437 | * freeing both later. | |
1dacc76d | 2438 | */ |
68d6ac6d JB |
2439 | if (flags & MSG_CMSG_COMPAT) |
2440 | data_skb = skb_shinfo(skb)->frag_list; | |
1dacc76d JB |
2441 | } |
2442 | #endif | |
2443 | ||
9063e21f ED |
2444 | /* Record the max length of recvmsg() calls for future allocations */ |
2445 | nlk->max_recvmsg_len = max(nlk->max_recvmsg_len, len); | |
2446 | nlk->max_recvmsg_len = min_t(size_t, nlk->max_recvmsg_len, | |
2447 | 16384); | |
2448 | ||
68d6ac6d | 2449 | copied = data_skb->len; |
1da177e4 LT |
2450 | if (len < copied) { |
2451 | msg->msg_flags |= MSG_TRUNC; | |
2452 | copied = len; | |
2453 | } | |
2454 | ||
68d6ac6d | 2455 | skb_reset_transport_header(data_skb); |
51f3d02b | 2456 | err = skb_copy_datagram_msg(data_skb, 0, msg, copied); |
1da177e4 LT |
2457 | |
2458 | if (msg->msg_name) { | |
342dfc30 | 2459 | DECLARE_SOCKADDR(struct sockaddr_nl *, addr, msg->msg_name); |
1da177e4 LT |
2460 | addr->nl_family = AF_NETLINK; |
2461 | addr->nl_pad = 0; | |
15e47304 | 2462 | addr->nl_pid = NETLINK_CB(skb).portid; |
d629b836 | 2463 | addr->nl_groups = netlink_group_mask(NETLINK_CB(skb).dst_group); |
1da177e4 LT |
2464 | msg->msg_namelen = sizeof(*addr); |
2465 | } | |
2466 | ||
cc3a572f | 2467 | if (nlk->flags & NETLINK_F_RECV_PKTINFO) |
cc9a06cd | 2468 | netlink_cmsg_recv_pktinfo(msg, skb); |
59324cf3 ND |
2469 | if (nlk->flags & NETLINK_F_LISTEN_ALL_NSID) |
2470 | netlink_cmsg_listen_all_nsid(sk, msg, skb); | |
cc9a06cd | 2471 | |
7cc05662 CH |
2472 | memset(&scm, 0, sizeof(scm)); |
2473 | scm.creds = *NETLINK_CREDS(skb); | |
188ccb55 | 2474 | if (flags & MSG_TRUNC) |
68d6ac6d | 2475 | copied = data_skb->len; |
daa3766e | 2476 | |
1da177e4 LT |
2477 | skb_free_datagram(sk, skb); |
2478 | ||
16b304f3 PS |
2479 | if (nlk->cb_running && |
2480 | atomic_read(&sk->sk_rmem_alloc) <= sk->sk_rcvbuf / 2) { | |
b44d211e AV |
2481 | ret = netlink_dump(sk); |
2482 | if (ret) { | |
ac30ef83 | 2483 | sk->sk_err = -ret; |
b44d211e AV |
2484 | sk->sk_error_report(sk); |
2485 | } | |
2486 | } | |
1da177e4 | 2487 | |
7cc05662 | 2488 | scm_recv(sock, msg, &scm, flags); |
1da177e4 LT |
2489 | out: |
2490 | netlink_rcv_wake(sk); | |
2491 | return err ? : copied; | |
2492 | } | |
2493 | ||
676d2369 | 2494 | static void netlink_data_ready(struct sock *sk) |
1da177e4 | 2495 | { |
cd40b7d3 | 2496 | BUG(); |
1da177e4 LT |
2497 | } |
2498 | ||
2499 | /* | |
746fac4d | 2500 | * We export these functions to other modules. They provide a |
1da177e4 LT |
2501 | * complete set of kernel non-blocking support for message |
2502 | * queueing. | |
2503 | */ | |
2504 | ||
2505 | struct sock * | |
9f00d977 PNA |
2506 | __netlink_kernel_create(struct net *net, int unit, struct module *module, |
2507 | struct netlink_kernel_cfg *cfg) | |
1da177e4 LT |
2508 | { |
2509 | struct socket *sock; | |
2510 | struct sock *sk; | |
77247bbb | 2511 | struct netlink_sock *nlk; |
5c398dc8 | 2512 | struct listeners *listeners = NULL; |
a31f2d17 PNA |
2513 | struct mutex *cb_mutex = cfg ? cfg->cb_mutex : NULL; |
2514 | unsigned int groups; | |
1da177e4 | 2515 | |
fab2caf6 | 2516 | BUG_ON(!nl_table); |
1da177e4 | 2517 | |
6ac552fd | 2518 | if (unit < 0 || unit >= MAX_LINKS) |
1da177e4 LT |
2519 | return NULL; |
2520 | ||
2521 | if (sock_create_lite(PF_NETLINK, SOCK_DGRAM, unit, &sock)) | |
2522 | return NULL; | |
13d3078e EB |
2523 | |
2524 | if (__netlink_create(net, sock, cb_mutex, unit, 1) < 0) | |
23fe1866 PE |
2525 | goto out_sock_release_nosk; |
2526 | ||
2527 | sk = sock->sk; | |
4fdb3bb7 | 2528 | |
a31f2d17 | 2529 | if (!cfg || cfg->groups < 32) |
4277a083 | 2530 | groups = 32; |
a31f2d17 PNA |
2531 | else |
2532 | groups = cfg->groups; | |
4277a083 | 2533 | |
5c398dc8 | 2534 | listeners = kzalloc(sizeof(*listeners) + NLGRPSZ(groups), GFP_KERNEL); |
4277a083 PM |
2535 | if (!listeners) |
2536 | goto out_sock_release; | |
2537 | ||
1da177e4 | 2538 | sk->sk_data_ready = netlink_data_ready; |
a31f2d17 PNA |
2539 | if (cfg && cfg->input) |
2540 | nlk_sk(sk)->netlink_rcv = cfg->input; | |
1da177e4 | 2541 | |
8ea65f4a | 2542 | if (netlink_insert(sk, 0)) |
77247bbb | 2543 | goto out_sock_release; |
4fdb3bb7 | 2544 | |
77247bbb | 2545 | nlk = nlk_sk(sk); |
cc3a572f | 2546 | nlk->flags |= NETLINK_F_KERNEL_SOCKET; |
4fdb3bb7 | 2547 | |
4fdb3bb7 | 2548 | netlink_table_grab(); |
b4b51029 EB |
2549 | if (!nl_table[unit].registered) { |
2550 | nl_table[unit].groups = groups; | |
5c398dc8 | 2551 | rcu_assign_pointer(nl_table[unit].listeners, listeners); |
b4b51029 EB |
2552 | nl_table[unit].cb_mutex = cb_mutex; |
2553 | nl_table[unit].module = module; | |
9785e10a PNA |
2554 | if (cfg) { |
2555 | nl_table[unit].bind = cfg->bind; | |
6251edd9 | 2556 | nl_table[unit].unbind = cfg->unbind; |
9785e10a | 2557 | nl_table[unit].flags = cfg->flags; |
da12c90e G |
2558 | if (cfg->compare) |
2559 | nl_table[unit].compare = cfg->compare; | |
9785e10a | 2560 | } |
b4b51029 | 2561 | nl_table[unit].registered = 1; |
f937f1f4 JJ |
2562 | } else { |
2563 | kfree(listeners); | |
869e58f8 | 2564 | nl_table[unit].registered++; |
b4b51029 | 2565 | } |
4fdb3bb7 | 2566 | netlink_table_ungrab(); |
77247bbb PM |
2567 | return sk; |
2568 | ||
4fdb3bb7 | 2569 | out_sock_release: |
4277a083 | 2570 | kfree(listeners); |
9dfbec1f | 2571 | netlink_kernel_release(sk); |
23fe1866 PE |
2572 | return NULL; |
2573 | ||
2574 | out_sock_release_nosk: | |
4fdb3bb7 | 2575 | sock_release(sock); |
77247bbb | 2576 | return NULL; |
1da177e4 | 2577 | } |
9f00d977 | 2578 | EXPORT_SYMBOL(__netlink_kernel_create); |
b7c6ba6e DL |
2579 | |
2580 | void | |
2581 | netlink_kernel_release(struct sock *sk) | |
2582 | { | |
13d3078e EB |
2583 | if (sk == NULL || sk->sk_socket == NULL) |
2584 | return; | |
2585 | ||
2586 | sock_release(sk->sk_socket); | |
b7c6ba6e DL |
2587 | } |
2588 | EXPORT_SYMBOL(netlink_kernel_release); | |
2589 | ||
d136f1bd | 2590 | int __netlink_change_ngroups(struct sock *sk, unsigned int groups) |
b4ff4f04 | 2591 | { |
5c398dc8 | 2592 | struct listeners *new, *old; |
b4ff4f04 | 2593 | struct netlink_table *tbl = &nl_table[sk->sk_protocol]; |
b4ff4f04 JB |
2594 | |
2595 | if (groups < 32) | |
2596 | groups = 32; | |
2597 | ||
b4ff4f04 | 2598 | if (NLGRPSZ(tbl->groups) < NLGRPSZ(groups)) { |
5c398dc8 ED |
2599 | new = kzalloc(sizeof(*new) + NLGRPSZ(groups), GFP_ATOMIC); |
2600 | if (!new) | |
d136f1bd | 2601 | return -ENOMEM; |
6d772ac5 | 2602 | old = nl_deref_protected(tbl->listeners); |
5c398dc8 ED |
2603 | memcpy(new->masks, old->masks, NLGRPSZ(tbl->groups)); |
2604 | rcu_assign_pointer(tbl->listeners, new); | |
2605 | ||
37b6b935 | 2606 | kfree_rcu(old, rcu); |
b4ff4f04 JB |
2607 | } |
2608 | tbl->groups = groups; | |
2609 | ||
d136f1bd JB |
2610 | return 0; |
2611 | } | |
2612 | ||
2613 | /** | |
2614 | * netlink_change_ngroups - change number of multicast groups | |
2615 | * | |
2616 | * This changes the number of multicast groups that are available | |
2617 | * on a certain netlink family. Note that it is not possible to | |
2618 | * change the number of groups to below 32. Also note that it does | |
2619 | * not implicitly call netlink_clear_multicast_users() when the | |
2620 | * number of groups is reduced. | |
2621 | * | |
2622 | * @sk: The kernel netlink socket, as returned by netlink_kernel_create(). | |
2623 | * @groups: The new number of groups. | |
2624 | */ | |
2625 | int netlink_change_ngroups(struct sock *sk, unsigned int groups) | |
2626 | { | |
2627 | int err; | |
2628 | ||
2629 | netlink_table_grab(); | |
2630 | err = __netlink_change_ngroups(sk, groups); | |
b4ff4f04 | 2631 | netlink_table_ungrab(); |
d136f1bd | 2632 | |
b4ff4f04 JB |
2633 | return err; |
2634 | } | |
b4ff4f04 | 2635 | |
b8273570 JB |
2636 | void __netlink_clear_multicast_users(struct sock *ksk, unsigned int group) |
2637 | { | |
2638 | struct sock *sk; | |
b8273570 JB |
2639 | struct netlink_table *tbl = &nl_table[ksk->sk_protocol]; |
2640 | ||
b67bfe0d | 2641 | sk_for_each_bound(sk, &tbl->mc_list) |
b8273570 JB |
2642 | netlink_update_socket_mc(nlk_sk(sk), group, 0); |
2643 | } | |
2644 | ||
a46621a3 | 2645 | struct nlmsghdr * |
15e47304 | 2646 | __nlmsg_put(struct sk_buff *skb, u32 portid, u32 seq, int type, int len, int flags) |
a46621a3 DV |
2647 | { |
2648 | struct nlmsghdr *nlh; | |
573ce260 | 2649 | int size = nlmsg_msg_size(len); |
a46621a3 | 2650 | |
23b45672 | 2651 | nlh = (struct nlmsghdr *)skb_put(skb, NLMSG_ALIGN(size)); |
a46621a3 DV |
2652 | nlh->nlmsg_type = type; |
2653 | nlh->nlmsg_len = size; | |
2654 | nlh->nlmsg_flags = flags; | |
15e47304 | 2655 | nlh->nlmsg_pid = portid; |
a46621a3 DV |
2656 | nlh->nlmsg_seq = seq; |
2657 | if (!__builtin_constant_p(size) || NLMSG_ALIGN(size) - size != 0) | |
573ce260 | 2658 | memset(nlmsg_data(nlh) + len, 0, NLMSG_ALIGN(size) - size); |
a46621a3 DV |
2659 | return nlh; |
2660 | } | |
2661 | EXPORT_SYMBOL(__nlmsg_put); | |
2662 | ||
1da177e4 LT |
2663 | /* |
2664 | * It looks a bit ugly. | |
2665 | * It would be better to create kernel thread. | |
2666 | */ | |
2667 | ||
2668 | static int netlink_dump(struct sock *sk) | |
2669 | { | |
2670 | struct netlink_sock *nlk = nlk_sk(sk); | |
2671 | struct netlink_callback *cb; | |
c7ac8679 | 2672 | struct sk_buff *skb = NULL; |
1da177e4 | 2673 | struct nlmsghdr *nlh; |
bf8b79e4 | 2674 | int len, err = -ENOBUFS; |
c7ac8679 | 2675 | int alloc_size; |
1da177e4 | 2676 | |
af65bdfc | 2677 | mutex_lock(nlk->cb_mutex); |
16b304f3 | 2678 | if (!nlk->cb_running) { |
bf8b79e4 TG |
2679 | err = -EINVAL; |
2680 | goto errout_skb; | |
1da177e4 LT |
2681 | } |
2682 | ||
16b304f3 | 2683 | cb = &nlk->cb; |
c7ac8679 GR |
2684 | alloc_size = max_t(int, cb->min_dump_alloc, NLMSG_GOODSIZE); |
2685 | ||
f9c22888 PM |
2686 | if (!netlink_rx_is_mmaped(sk) && |
2687 | atomic_read(&sk->sk_rmem_alloc) >= sk->sk_rcvbuf) | |
2688 | goto errout_skb; | |
9063e21f ED |
2689 | |
2690 | /* NLMSG_GOODSIZE is small to avoid high order allocations being | |
2691 | * required, but it makes sense to _attempt_ a 16K bytes allocation | |
2692 | * to reduce number of system calls on dump operations, if user | |
2693 | * ever provided a big enough buffer. | |
2694 | */ | |
2695 | if (alloc_size < nlk->max_recvmsg_len) { | |
2696 | skb = netlink_alloc_skb(sk, | |
2697 | nlk->max_recvmsg_len, | |
2698 | nlk->portid, | |
2699 | GFP_KERNEL | | |
2700 | __GFP_NOWARN | | |
2701 | __GFP_NORETRY); | |
2702 | /* available room should be exact amount to avoid MSG_TRUNC */ | |
2703 | if (skb) | |
2704 | skb_reserve(skb, skb_tailroom(skb) - | |
2705 | nlk->max_recvmsg_len); | |
2706 | } | |
2707 | if (!skb) | |
2708 | skb = netlink_alloc_skb(sk, alloc_size, nlk->portid, | |
2709 | GFP_KERNEL); | |
c7ac8679 | 2710 | if (!skb) |
c63d6ea3 | 2711 | goto errout_skb; |
f9c22888 | 2712 | netlink_skb_set_owner_r(skb, sk); |
c7ac8679 | 2713 | |
1da177e4 LT |
2714 | len = cb->dump(skb, cb); |
2715 | ||
2716 | if (len > 0) { | |
af65bdfc | 2717 | mutex_unlock(nlk->cb_mutex); |
b1153f29 SH |
2718 | |
2719 | if (sk_filter(sk, skb)) | |
2720 | kfree_skb(skb); | |
4a7e7c2a ED |
2721 | else |
2722 | __netlink_sendskb(sk, skb); | |
1da177e4 LT |
2723 | return 0; |
2724 | } | |
2725 | ||
bf8b79e4 TG |
2726 | nlh = nlmsg_put_answer(skb, cb, NLMSG_DONE, sizeof(len), NLM_F_MULTI); |
2727 | if (!nlh) | |
2728 | goto errout_skb; | |
2729 | ||
670dc283 JB |
2730 | nl_dump_check_consistent(cb, nlh); |
2731 | ||
bf8b79e4 TG |
2732 | memcpy(nlmsg_data(nlh), &len, sizeof(len)); |
2733 | ||
b1153f29 SH |
2734 | if (sk_filter(sk, skb)) |
2735 | kfree_skb(skb); | |
4a7e7c2a ED |
2736 | else |
2737 | __netlink_sendskb(sk, skb); | |
1da177e4 | 2738 | |
a8f74b22 TG |
2739 | if (cb->done) |
2740 | cb->done(cb); | |
1da177e4 | 2741 | |
16b304f3 PS |
2742 | nlk->cb_running = false; |
2743 | mutex_unlock(nlk->cb_mutex); | |
6dc878a8 | 2744 | module_put(cb->module); |
16b304f3 | 2745 | consume_skb(cb->skb); |
1da177e4 | 2746 | return 0; |
1797754e | 2747 | |
bf8b79e4 | 2748 | errout_skb: |
af65bdfc | 2749 | mutex_unlock(nlk->cb_mutex); |
bf8b79e4 | 2750 | kfree_skb(skb); |
bf8b79e4 | 2751 | return err; |
1da177e4 LT |
2752 | } |
2753 | ||
6dc878a8 G |
2754 | int __netlink_dump_start(struct sock *ssk, struct sk_buff *skb, |
2755 | const struct nlmsghdr *nlh, | |
2756 | struct netlink_dump_control *control) | |
1da177e4 LT |
2757 | { |
2758 | struct netlink_callback *cb; | |
2759 | struct sock *sk; | |
2760 | struct netlink_sock *nlk; | |
b44d211e | 2761 | int ret; |
1da177e4 | 2762 | |
f9c22888 PM |
2763 | /* Memory mapped dump requests need to be copied to avoid looping |
2764 | * on the pending state in netlink_mmap_sendmsg() while the CB hold | |
2765 | * a reference to the skb. | |
2766 | */ | |
2767 | if (netlink_skb_is_mmaped(skb)) { | |
2768 | skb = skb_copy(skb, GFP_KERNEL); | |
16b304f3 | 2769 | if (skb == NULL) |
f9c22888 | 2770 | return -ENOBUFS; |
f9c22888 PM |
2771 | } else |
2772 | atomic_inc(&skb->users); | |
2773 | ||
15e47304 | 2774 | sk = netlink_lookup(sock_net(ssk), ssk->sk_protocol, NETLINK_CB(skb).portid); |
1da177e4 | 2775 | if (sk == NULL) { |
16b304f3 PS |
2776 | ret = -ECONNREFUSED; |
2777 | goto error_free; | |
1da177e4 | 2778 | } |
6dc878a8 | 2779 | |
16b304f3 | 2780 | nlk = nlk_sk(sk); |
af65bdfc | 2781 | mutex_lock(nlk->cb_mutex); |
6dc878a8 | 2782 | /* A dump is in progress... */ |
16b304f3 | 2783 | if (nlk->cb_running) { |
6dc878a8 | 2784 | ret = -EBUSY; |
16b304f3 | 2785 | goto error_unlock; |
1da177e4 | 2786 | } |
6dc878a8 | 2787 | /* add reference of module which cb->dump belongs to */ |
16b304f3 | 2788 | if (!try_module_get(control->module)) { |
6dc878a8 | 2789 | ret = -EPROTONOSUPPORT; |
16b304f3 | 2790 | goto error_unlock; |
6dc878a8 G |
2791 | } |
2792 | ||
16b304f3 PS |
2793 | cb = &nlk->cb; |
2794 | memset(cb, 0, sizeof(*cb)); | |
2795 | cb->dump = control->dump; | |
2796 | cb->done = control->done; | |
2797 | cb->nlh = nlh; | |
2798 | cb->data = control->data; | |
2799 | cb->module = control->module; | |
2800 | cb->min_dump_alloc = control->min_dump_alloc; | |
2801 | cb->skb = skb; | |
2802 | ||
2803 | nlk->cb_running = true; | |
2804 | ||
af65bdfc | 2805 | mutex_unlock(nlk->cb_mutex); |
1da177e4 | 2806 | |
b44d211e | 2807 | ret = netlink_dump(sk); |
1da177e4 | 2808 | sock_put(sk); |
5c58298c | 2809 | |
b44d211e AV |
2810 | if (ret) |
2811 | return ret; | |
2812 | ||
5c58298c DL |
2813 | /* We successfully started a dump, by returning -EINTR we |
2814 | * signal not to send ACK even if it was requested. | |
2815 | */ | |
2816 | return -EINTR; | |
16b304f3 PS |
2817 | |
2818 | error_unlock: | |
2819 | sock_put(sk); | |
2820 | mutex_unlock(nlk->cb_mutex); | |
2821 | error_free: | |
2822 | kfree_skb(skb); | |
2823 | return ret; | |
1da177e4 | 2824 | } |
6dc878a8 | 2825 | EXPORT_SYMBOL(__netlink_dump_start); |
1da177e4 LT |
2826 | |
2827 | void netlink_ack(struct sk_buff *in_skb, struct nlmsghdr *nlh, int err) | |
2828 | { | |
2829 | struct sk_buff *skb; | |
2830 | struct nlmsghdr *rep; | |
2831 | struct nlmsgerr *errmsg; | |
339bf98f | 2832 | size_t payload = sizeof(*errmsg); |
1da177e4 | 2833 | |
339bf98f TG |
2834 | /* error messages get the original request appened */ |
2835 | if (err) | |
2836 | payload += nlmsg_len(nlh); | |
1da177e4 | 2837 | |
f9c22888 PM |
2838 | skb = netlink_alloc_skb(in_skb->sk, nlmsg_total_size(payload), |
2839 | NETLINK_CB(in_skb).portid, GFP_KERNEL); | |
1da177e4 LT |
2840 | if (!skb) { |
2841 | struct sock *sk; | |
2842 | ||
3b1e0a65 | 2843 | sk = netlink_lookup(sock_net(in_skb->sk), |
b4b51029 | 2844 | in_skb->sk->sk_protocol, |
15e47304 | 2845 | NETLINK_CB(in_skb).portid); |
1da177e4 LT |
2846 | if (sk) { |
2847 | sk->sk_err = ENOBUFS; | |
2848 | sk->sk_error_report(sk); | |
2849 | sock_put(sk); | |
2850 | } | |
2851 | return; | |
2852 | } | |
2853 | ||
15e47304 | 2854 | rep = __nlmsg_put(skb, NETLINK_CB(in_skb).portid, nlh->nlmsg_seq, |
5dba93ae | 2855 | NLMSG_ERROR, payload, 0); |
bf8b79e4 | 2856 | errmsg = nlmsg_data(rep); |
1da177e4 | 2857 | errmsg->error = err; |
bf8b79e4 | 2858 | memcpy(&errmsg->msg, nlh, err ? nlh->nlmsg_len : sizeof(*nlh)); |
15e47304 | 2859 | netlink_unicast(in_skb->sk, skb, NETLINK_CB(in_skb).portid, MSG_DONTWAIT); |
1da177e4 | 2860 | } |
6ac552fd | 2861 | EXPORT_SYMBOL(netlink_ack); |
1da177e4 | 2862 | |
cd40b7d3 | 2863 | int netlink_rcv_skb(struct sk_buff *skb, int (*cb)(struct sk_buff *, |
1d00a4eb | 2864 | struct nlmsghdr *)) |
82ace47a | 2865 | { |
82ace47a TG |
2866 | struct nlmsghdr *nlh; |
2867 | int err; | |
2868 | ||
2869 | while (skb->len >= nlmsg_total_size(0)) { | |
cd40b7d3 DL |
2870 | int msglen; |
2871 | ||
b529ccf2 | 2872 | nlh = nlmsg_hdr(skb); |
d35b6856 | 2873 | err = 0; |
82ace47a | 2874 | |
ad8e4b75 | 2875 | if (nlh->nlmsg_len < NLMSG_HDRLEN || skb->len < nlh->nlmsg_len) |
82ace47a TG |
2876 | return 0; |
2877 | ||
d35b6856 TG |
2878 | /* Only requests are handled by the kernel */ |
2879 | if (!(nlh->nlmsg_flags & NLM_F_REQUEST)) | |
5c58298c | 2880 | goto ack; |
45e7ae7f TG |
2881 | |
2882 | /* Skip control messages */ | |
2883 | if (nlh->nlmsg_type < NLMSG_MIN_TYPE) | |
5c58298c | 2884 | goto ack; |
d35b6856 | 2885 | |
1d00a4eb | 2886 | err = cb(skb, nlh); |
5c58298c DL |
2887 | if (err == -EINTR) |
2888 | goto skip; | |
2889 | ||
2890 | ack: | |
d35b6856 | 2891 | if (nlh->nlmsg_flags & NLM_F_ACK || err) |
82ace47a | 2892 | netlink_ack(skb, nlh, err); |
82ace47a | 2893 | |
5c58298c | 2894 | skip: |
6ac552fd | 2895 | msglen = NLMSG_ALIGN(nlh->nlmsg_len); |
cd40b7d3 DL |
2896 | if (msglen > skb->len) |
2897 | msglen = skb->len; | |
2898 | skb_pull(skb, msglen); | |
82ace47a TG |
2899 | } |
2900 | ||
2901 | return 0; | |
2902 | } | |
6ac552fd | 2903 | EXPORT_SYMBOL(netlink_rcv_skb); |
82ace47a | 2904 | |
d387f6ad TG |
2905 | /** |
2906 | * nlmsg_notify - send a notification netlink message | |
2907 | * @sk: netlink socket to use | |
2908 | * @skb: notification message | |
15e47304 | 2909 | * @portid: destination netlink portid for reports or 0 |
d387f6ad TG |
2910 | * @group: destination multicast group or 0 |
2911 | * @report: 1 to report back, 0 to disable | |
2912 | * @flags: allocation flags | |
2913 | */ | |
15e47304 | 2914 | int nlmsg_notify(struct sock *sk, struct sk_buff *skb, u32 portid, |
d387f6ad TG |
2915 | unsigned int group, int report, gfp_t flags) |
2916 | { | |
2917 | int err = 0; | |
2918 | ||
2919 | if (group) { | |
15e47304 | 2920 | int exclude_portid = 0; |
d387f6ad TG |
2921 | |
2922 | if (report) { | |
2923 | atomic_inc(&skb->users); | |
15e47304 | 2924 | exclude_portid = portid; |
d387f6ad TG |
2925 | } |
2926 | ||
1ce85fe4 PNA |
2927 | /* errors reported via destination sk->sk_err, but propagate |
2928 | * delivery errors if NETLINK_BROADCAST_ERROR flag is set */ | |
15e47304 | 2929 | err = nlmsg_multicast(sk, skb, exclude_portid, group, flags); |
d387f6ad TG |
2930 | } |
2931 | ||
1ce85fe4 PNA |
2932 | if (report) { |
2933 | int err2; | |
2934 | ||
15e47304 | 2935 | err2 = nlmsg_unicast(sk, skb, portid); |
1ce85fe4 PNA |
2936 | if (!err || err == -ESRCH) |
2937 | err = err2; | |
2938 | } | |
d387f6ad TG |
2939 | |
2940 | return err; | |
2941 | } | |
6ac552fd | 2942 | EXPORT_SYMBOL(nlmsg_notify); |
d387f6ad | 2943 | |
1da177e4 LT |
2944 | #ifdef CONFIG_PROC_FS |
2945 | struct nl_seq_iter { | |
e372c414 | 2946 | struct seq_net_private p; |
56d28b1e | 2947 | struct rhashtable_iter hti; |
1da177e4 | 2948 | int link; |
1da177e4 LT |
2949 | }; |
2950 | ||
56d28b1e | 2951 | static int netlink_walk_start(struct nl_seq_iter *iter) |
1da177e4 | 2952 | { |
56d28b1e | 2953 | int err; |
1da177e4 | 2954 | |
56d28b1e HX |
2955 | err = rhashtable_walk_init(&nl_table[iter->link].hash, &iter->hti); |
2956 | if (err) { | |
2957 | iter->link = MAX_LINKS; | |
2958 | return err; | |
1da177e4 | 2959 | } |
56d28b1e HX |
2960 | |
2961 | err = rhashtable_walk_start(&iter->hti); | |
2962 | return err == -EAGAIN ? 0 : err; | |
1da177e4 LT |
2963 | } |
2964 | ||
56d28b1e | 2965 | static void netlink_walk_stop(struct nl_seq_iter *iter) |
1da177e4 | 2966 | { |
56d28b1e HX |
2967 | rhashtable_walk_stop(&iter->hti); |
2968 | rhashtable_walk_exit(&iter->hti); | |
1da177e4 LT |
2969 | } |
2970 | ||
56d28b1e | 2971 | static void *__netlink_seq_next(struct seq_file *seq) |
1da177e4 | 2972 | { |
56d28b1e | 2973 | struct nl_seq_iter *iter = seq->private; |
e341694e | 2974 | struct netlink_sock *nlk; |
1da177e4 | 2975 | |
56d28b1e HX |
2976 | do { |
2977 | for (;;) { | |
2978 | int err; | |
1da177e4 | 2979 | |
56d28b1e | 2980 | nlk = rhashtable_walk_next(&iter->hti); |
746fac4d | 2981 | |
56d28b1e HX |
2982 | if (IS_ERR(nlk)) { |
2983 | if (PTR_ERR(nlk) == -EAGAIN) | |
2984 | continue; | |
e341694e | 2985 | |
56d28b1e HX |
2986 | return nlk; |
2987 | } | |
1da177e4 | 2988 | |
56d28b1e HX |
2989 | if (nlk) |
2990 | break; | |
1da177e4 | 2991 | |
56d28b1e HX |
2992 | netlink_walk_stop(iter); |
2993 | if (++iter->link >= MAX_LINKS) | |
2994 | return NULL; | |
da12c90e | 2995 | |
56d28b1e HX |
2996 | err = netlink_walk_start(iter); |
2997 | if (err) | |
2998 | return ERR_PTR(err); | |
1da177e4 | 2999 | } |
56d28b1e | 3000 | } while (sock_net(&nlk->sk) != seq_file_net(seq)); |
1da177e4 | 3001 | |
56d28b1e HX |
3002 | return nlk; |
3003 | } | |
1da177e4 | 3004 | |
56d28b1e HX |
3005 | static void *netlink_seq_start(struct seq_file *seq, loff_t *posp) |
3006 | { | |
3007 | struct nl_seq_iter *iter = seq->private; | |
3008 | void *obj = SEQ_START_TOKEN; | |
3009 | loff_t pos; | |
3010 | int err; | |
3011 | ||
3012 | iter->link = 0; | |
3013 | ||
3014 | err = netlink_walk_start(iter); | |
3015 | if (err) | |
3016 | return ERR_PTR(err); | |
3017 | ||
3018 | for (pos = *posp; pos && obj && !IS_ERR(obj); pos--) | |
3019 | obj = __netlink_seq_next(seq); | |
3020 | ||
3021 | return obj; | |
3022 | } | |
3023 | ||
3024 | static void *netlink_seq_next(struct seq_file *seq, void *v, loff_t *pos) | |
3025 | { | |
3026 | ++*pos; | |
3027 | return __netlink_seq_next(seq); | |
1da177e4 LT |
3028 | } |
3029 | ||
3030 | static void netlink_seq_stop(struct seq_file *seq, void *v) | |
3031 | { | |
56d28b1e HX |
3032 | struct nl_seq_iter *iter = seq->private; |
3033 | ||
3034 | if (iter->link >= MAX_LINKS) | |
3035 | return; | |
3036 | ||
3037 | netlink_walk_stop(iter); | |
1da177e4 LT |
3038 | } |
3039 | ||
3040 | ||
3041 | static int netlink_seq_show(struct seq_file *seq, void *v) | |
3042 | { | |
658cb354 | 3043 | if (v == SEQ_START_TOKEN) { |
1da177e4 LT |
3044 | seq_puts(seq, |
3045 | "sk Eth Pid Groups " | |
cf0aa4e0 | 3046 | "Rmem Wmem Dump Locks Drops Inode\n"); |
658cb354 | 3047 | } else { |
1da177e4 LT |
3048 | struct sock *s = v; |
3049 | struct netlink_sock *nlk = nlk_sk(s); | |
3050 | ||
16b304f3 | 3051 | seq_printf(seq, "%pK %-3d %-6u %08x %-8d %-8d %d %-8d %-8d %-8lu\n", |
1da177e4 LT |
3052 | s, |
3053 | s->sk_protocol, | |
15e47304 | 3054 | nlk->portid, |
513c2500 | 3055 | nlk->groups ? (u32)nlk->groups[0] : 0, |
31e6d363 ED |
3056 | sk_rmem_alloc_get(s), |
3057 | sk_wmem_alloc_get(s), | |
16b304f3 | 3058 | nlk->cb_running, |
38938bfe | 3059 | atomic_read(&s->sk_refcnt), |
cf0aa4e0 MY |
3060 | atomic_read(&s->sk_drops), |
3061 | sock_i_ino(s) | |
1da177e4 LT |
3062 | ); |
3063 | ||
3064 | } | |
3065 | return 0; | |
3066 | } | |
3067 | ||
56b3d975 | 3068 | static const struct seq_operations netlink_seq_ops = { |
1da177e4 LT |
3069 | .start = netlink_seq_start, |
3070 | .next = netlink_seq_next, | |
3071 | .stop = netlink_seq_stop, | |
3072 | .show = netlink_seq_show, | |
3073 | }; | |
3074 | ||
3075 | ||
3076 | static int netlink_seq_open(struct inode *inode, struct file *file) | |
3077 | { | |
e372c414 DL |
3078 | return seq_open_net(inode, file, &netlink_seq_ops, |
3079 | sizeof(struct nl_seq_iter)); | |
b4b51029 EB |
3080 | } |
3081 | ||
da7071d7 | 3082 | static const struct file_operations netlink_seq_fops = { |
1da177e4 LT |
3083 | .owner = THIS_MODULE, |
3084 | .open = netlink_seq_open, | |
3085 | .read = seq_read, | |
3086 | .llseek = seq_lseek, | |
e372c414 | 3087 | .release = seq_release_net, |
1da177e4 LT |
3088 | }; |
3089 | ||
3090 | #endif | |
3091 | ||
3092 | int netlink_register_notifier(struct notifier_block *nb) | |
3093 | { | |
e041c683 | 3094 | return atomic_notifier_chain_register(&netlink_chain, nb); |
1da177e4 | 3095 | } |
6ac552fd | 3096 | EXPORT_SYMBOL(netlink_register_notifier); |
1da177e4 LT |
3097 | |
3098 | int netlink_unregister_notifier(struct notifier_block *nb) | |
3099 | { | |
e041c683 | 3100 | return atomic_notifier_chain_unregister(&netlink_chain, nb); |
1da177e4 | 3101 | } |
6ac552fd | 3102 | EXPORT_SYMBOL(netlink_unregister_notifier); |
746fac4d | 3103 | |
90ddc4f0 | 3104 | static const struct proto_ops netlink_ops = { |
1da177e4 LT |
3105 | .family = PF_NETLINK, |
3106 | .owner = THIS_MODULE, | |
3107 | .release = netlink_release, | |
3108 | .bind = netlink_bind, | |
3109 | .connect = netlink_connect, | |
3110 | .socketpair = sock_no_socketpair, | |
3111 | .accept = sock_no_accept, | |
3112 | .getname = netlink_getname, | |
9652e931 | 3113 | .poll = netlink_poll, |
1da177e4 LT |
3114 | .ioctl = sock_no_ioctl, |
3115 | .listen = sock_no_listen, | |
3116 | .shutdown = sock_no_shutdown, | |
9a4595bc PM |
3117 | .setsockopt = netlink_setsockopt, |
3118 | .getsockopt = netlink_getsockopt, | |
1da177e4 LT |
3119 | .sendmsg = netlink_sendmsg, |
3120 | .recvmsg = netlink_recvmsg, | |
ccdfcc39 | 3121 | .mmap = netlink_mmap, |
1da177e4 LT |
3122 | .sendpage = sock_no_sendpage, |
3123 | }; | |
3124 | ||
ec1b4cf7 | 3125 | static const struct net_proto_family netlink_family_ops = { |
1da177e4 LT |
3126 | .family = PF_NETLINK, |
3127 | .create = netlink_create, | |
3128 | .owner = THIS_MODULE, /* for consistency 8) */ | |
3129 | }; | |
3130 | ||
4665079c | 3131 | static int __net_init netlink_net_init(struct net *net) |
b4b51029 EB |
3132 | { |
3133 | #ifdef CONFIG_PROC_FS | |
d4beaa66 | 3134 | if (!proc_create("netlink", 0, net->proc_net, &netlink_seq_fops)) |
b4b51029 EB |
3135 | return -ENOMEM; |
3136 | #endif | |
3137 | return 0; | |
3138 | } | |
3139 | ||
4665079c | 3140 | static void __net_exit netlink_net_exit(struct net *net) |
b4b51029 EB |
3141 | { |
3142 | #ifdef CONFIG_PROC_FS | |
ece31ffd | 3143 | remove_proc_entry("netlink", net->proc_net); |
b4b51029 EB |
3144 | #endif |
3145 | } | |
3146 | ||
b963ea89 DM |
3147 | static void __init netlink_add_usersock_entry(void) |
3148 | { | |
5c398dc8 | 3149 | struct listeners *listeners; |
b963ea89 DM |
3150 | int groups = 32; |
3151 | ||
5c398dc8 | 3152 | listeners = kzalloc(sizeof(*listeners) + NLGRPSZ(groups), GFP_KERNEL); |
b963ea89 | 3153 | if (!listeners) |
5c398dc8 | 3154 | panic("netlink_add_usersock_entry: Cannot allocate listeners\n"); |
b963ea89 DM |
3155 | |
3156 | netlink_table_grab(); | |
3157 | ||
3158 | nl_table[NETLINK_USERSOCK].groups = groups; | |
5c398dc8 | 3159 | rcu_assign_pointer(nl_table[NETLINK_USERSOCK].listeners, listeners); |
b963ea89 DM |
3160 | nl_table[NETLINK_USERSOCK].module = THIS_MODULE; |
3161 | nl_table[NETLINK_USERSOCK].registered = 1; | |
9785e10a | 3162 | nl_table[NETLINK_USERSOCK].flags = NL_CFG_F_NONROOT_SEND; |
b963ea89 DM |
3163 | |
3164 | netlink_table_ungrab(); | |
3165 | } | |
3166 | ||
022cbae6 | 3167 | static struct pernet_operations __net_initdata netlink_net_ops = { |
b4b51029 EB |
3168 | .init = netlink_net_init, |
3169 | .exit = netlink_net_exit, | |
3170 | }; | |
3171 | ||
49f7b33e | 3172 | static inline u32 netlink_hash(const void *data, u32 len, u32 seed) |
c428ecd1 HX |
3173 | { |
3174 | const struct netlink_sock *nlk = data; | |
3175 | struct netlink_compare_arg arg; | |
3176 | ||
3177 | netlink_compare_arg_init(&arg, sock_net(&nlk->sk), nlk->portid); | |
11b58ba1 | 3178 | return jhash2((u32 *)&arg, netlink_compare_arg_len / sizeof(u32), seed); |
c428ecd1 HX |
3179 | } |
3180 | ||
3181 | static const struct rhashtable_params netlink_rhashtable_params = { | |
3182 | .head_offset = offsetof(struct netlink_sock, node), | |
3183 | .key_len = netlink_compare_arg_len, | |
c428ecd1 HX |
3184 | .obj_hashfn = netlink_hash, |
3185 | .obj_cmpfn = netlink_compare, | |
b5e2c150 | 3186 | .automatic_shrinking = true, |
c428ecd1 HX |
3187 | }; |
3188 | ||
1da177e4 LT |
3189 | static int __init netlink_proto_init(void) |
3190 | { | |
1da177e4 | 3191 | int i; |
1da177e4 LT |
3192 | int err = proto_register(&netlink_proto, 0); |
3193 | ||
3194 | if (err != 0) | |
3195 | goto out; | |
3196 | ||
fab25745 | 3197 | BUILD_BUG_ON(sizeof(struct netlink_skb_parms) > FIELD_SIZEOF(struct sk_buff, cb)); |
1da177e4 | 3198 | |
0da974f4 | 3199 | nl_table = kcalloc(MAX_LINKS, sizeof(*nl_table), GFP_KERNEL); |
fab2caf6 AM |
3200 | if (!nl_table) |
3201 | goto panic; | |
1da177e4 | 3202 | |
1da177e4 | 3203 | for (i = 0; i < MAX_LINKS; i++) { |
c428ecd1 HX |
3204 | if (rhashtable_init(&nl_table[i].hash, |
3205 | &netlink_rhashtable_params) < 0) { | |
e341694e TG |
3206 | while (--i > 0) |
3207 | rhashtable_destroy(&nl_table[i].hash); | |
1da177e4 | 3208 | kfree(nl_table); |
fab2caf6 | 3209 | goto panic; |
1da177e4 | 3210 | } |
1da177e4 LT |
3211 | } |
3212 | ||
bcbde0d4 DB |
3213 | INIT_LIST_HEAD(&netlink_tap_all); |
3214 | ||
b963ea89 DM |
3215 | netlink_add_usersock_entry(); |
3216 | ||
1da177e4 | 3217 | sock_register(&netlink_family_ops); |
b4b51029 | 3218 | register_pernet_subsys(&netlink_net_ops); |
746fac4d | 3219 | /* The netlink device handler may be needed early. */ |
1da177e4 LT |
3220 | rtnetlink_init(); |
3221 | out: | |
3222 | return err; | |
fab2caf6 AM |
3223 | panic: |
3224 | panic("netlink_init: Cannot allocate nl_table\n"); | |
1da177e4 LT |
3225 | } |
3226 | ||
1da177e4 | 3227 | core_initcall(netlink_proto_init); |