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 |
0a6a3a23 | 87 | #define NETLINK_F_CAP_ACK 0x20 |
77247bbb | 88 | |
035c4c16 | 89 | static inline int netlink_is_kernel(struct sock *sk) |
aed81560 | 90 | { |
cc3a572f | 91 | return nlk_sk(sk)->flags & NETLINK_F_KERNEL_SOCKET; |
aed81560 DL |
92 | } |
93 | ||
91dd93f9 | 94 | struct netlink_table *nl_table __read_mostly; |
0f29c768 | 95 | EXPORT_SYMBOL_GPL(nl_table); |
1da177e4 LT |
96 | |
97 | static DECLARE_WAIT_QUEUE_HEAD(nl_table_wait); | |
98 | ||
99 | static int netlink_dump(struct sock *sk); | |
9652e931 | 100 | static void netlink_skb_destructor(struct sk_buff *skb); |
1da177e4 | 101 | |
78fd1d0a | 102 | /* nl_table locking explained: |
21e4902a | 103 | * Lookup and traversal are protected with an RCU read-side lock. Insertion |
c5adde94 | 104 | * and removal are protected with per bucket lock while using RCU list |
21e4902a TG |
105 | * modification primitives and may run in parallel to RCU protected lookups. |
106 | * Destruction of the Netlink socket may only occur *after* nl_table_lock has | |
107 | * been acquired * either during or after the socket has been removed from | |
108 | * the list and after an RCU grace period. | |
78fd1d0a | 109 | */ |
0f29c768 AV |
110 | DEFINE_RWLOCK(nl_table_lock); |
111 | EXPORT_SYMBOL_GPL(nl_table_lock); | |
1da177e4 LT |
112 | static atomic_t nl_table_users = ATOMIC_INIT(0); |
113 | ||
6d772ac5 ED |
114 | #define nl_deref_protected(X) rcu_dereference_protected(X, lockdep_is_held(&nl_table_lock)); |
115 | ||
e041c683 | 116 | static ATOMIC_NOTIFIER_HEAD(netlink_chain); |
1da177e4 | 117 | |
bcbde0d4 DB |
118 | static DEFINE_SPINLOCK(netlink_tap_lock); |
119 | static struct list_head netlink_tap_all __read_mostly; | |
120 | ||
c428ecd1 HX |
121 | static const struct rhashtable_params netlink_rhashtable_params; |
122 | ||
b57ef81f | 123 | static inline u32 netlink_group_mask(u32 group) |
d629b836 PM |
124 | { |
125 | return group ? 1 << (group - 1) : 0; | |
126 | } | |
127 | ||
1853c949 DB |
128 | static struct sk_buff *netlink_to_full_skb(const struct sk_buff *skb, |
129 | gfp_t gfp_mask) | |
130 | { | |
131 | unsigned int len = skb_end_offset(skb); | |
132 | struct sk_buff *new; | |
133 | ||
134 | new = alloc_skb(len, gfp_mask); | |
135 | if (new == NULL) | |
136 | return NULL; | |
137 | ||
138 | NETLINK_CB(new).portid = NETLINK_CB(skb).portid; | |
139 | NETLINK_CB(new).dst_group = NETLINK_CB(skb).dst_group; | |
140 | NETLINK_CB(new).creds = NETLINK_CB(skb).creds; | |
141 | ||
142 | memcpy(skb_put(new, len), skb->data, len); | |
143 | return new; | |
144 | } | |
145 | ||
bcbde0d4 DB |
146 | int netlink_add_tap(struct netlink_tap *nt) |
147 | { | |
148 | if (unlikely(nt->dev->type != ARPHRD_NETLINK)) | |
149 | return -EINVAL; | |
150 | ||
151 | spin_lock(&netlink_tap_lock); | |
152 | list_add_rcu(&nt->list, &netlink_tap_all); | |
153 | spin_unlock(&netlink_tap_lock); | |
154 | ||
fcd4d35e | 155 | __module_get(nt->module); |
bcbde0d4 DB |
156 | |
157 | return 0; | |
158 | } | |
159 | EXPORT_SYMBOL_GPL(netlink_add_tap); | |
160 | ||
2173f8d9 | 161 | static int __netlink_remove_tap(struct netlink_tap *nt) |
bcbde0d4 DB |
162 | { |
163 | bool found = false; | |
164 | struct netlink_tap *tmp; | |
165 | ||
166 | spin_lock(&netlink_tap_lock); | |
167 | ||
168 | list_for_each_entry(tmp, &netlink_tap_all, list) { | |
169 | if (nt == tmp) { | |
170 | list_del_rcu(&nt->list); | |
171 | found = true; | |
172 | goto out; | |
173 | } | |
174 | } | |
175 | ||
176 | pr_warn("__netlink_remove_tap: %p not found\n", nt); | |
177 | out: | |
178 | spin_unlock(&netlink_tap_lock); | |
179 | ||
92b80eb3 | 180 | if (found) |
bcbde0d4 DB |
181 | module_put(nt->module); |
182 | ||
183 | return found ? 0 : -ENODEV; | |
184 | } | |
bcbde0d4 DB |
185 | |
186 | int netlink_remove_tap(struct netlink_tap *nt) | |
187 | { | |
188 | int ret; | |
189 | ||
190 | ret = __netlink_remove_tap(nt); | |
191 | synchronize_net(); | |
192 | ||
193 | return ret; | |
194 | } | |
195 | EXPORT_SYMBOL_GPL(netlink_remove_tap); | |
196 | ||
5ffd5cdd DB |
197 | static bool netlink_filter_tap(const struct sk_buff *skb) |
198 | { | |
199 | struct sock *sk = skb->sk; | |
5ffd5cdd DB |
200 | |
201 | /* We take the more conservative approach and | |
202 | * whitelist socket protocols that may pass. | |
203 | */ | |
204 | switch (sk->sk_protocol) { | |
205 | case NETLINK_ROUTE: | |
206 | case NETLINK_USERSOCK: | |
207 | case NETLINK_SOCK_DIAG: | |
208 | case NETLINK_NFLOG: | |
209 | case NETLINK_XFRM: | |
210 | case NETLINK_FIB_LOOKUP: | |
211 | case NETLINK_NETFILTER: | |
212 | case NETLINK_GENERIC: | |
498044bb | 213 | return true; |
5ffd5cdd DB |
214 | } |
215 | ||
498044bb | 216 | return false; |
5ffd5cdd DB |
217 | } |
218 | ||
bcbde0d4 DB |
219 | static int __netlink_deliver_tap_skb(struct sk_buff *skb, |
220 | struct net_device *dev) | |
221 | { | |
222 | struct sk_buff *nskb; | |
5ffd5cdd | 223 | struct sock *sk = skb->sk; |
bcbde0d4 DB |
224 | int ret = -ENOMEM; |
225 | ||
226 | dev_hold(dev); | |
1853c949 DB |
227 | |
228 | if (netlink_skb_is_mmaped(skb) || is_vmalloc_addr(skb->head)) | |
229 | nskb = netlink_to_full_skb(skb, GFP_ATOMIC); | |
230 | else | |
231 | nskb = skb_clone(skb, GFP_ATOMIC); | |
bcbde0d4 DB |
232 | if (nskb) { |
233 | nskb->dev = dev; | |
5ffd5cdd | 234 | nskb->protocol = htons((u16) sk->sk_protocol); |
604d13c9 DB |
235 | nskb->pkt_type = netlink_is_kernel(sk) ? |
236 | PACKET_KERNEL : PACKET_USER; | |
4e48ed88 | 237 | skb_reset_network_header(nskb); |
bcbde0d4 DB |
238 | ret = dev_queue_xmit(nskb); |
239 | if (unlikely(ret > 0)) | |
240 | ret = net_xmit_errno(ret); | |
241 | } | |
242 | ||
243 | dev_put(dev); | |
244 | return ret; | |
245 | } | |
246 | ||
247 | static void __netlink_deliver_tap(struct sk_buff *skb) | |
248 | { | |
249 | int ret; | |
250 | struct netlink_tap *tmp; | |
251 | ||
5ffd5cdd DB |
252 | if (!netlink_filter_tap(skb)) |
253 | return; | |
254 | ||
bcbde0d4 DB |
255 | list_for_each_entry_rcu(tmp, &netlink_tap_all, list) { |
256 | ret = __netlink_deliver_tap_skb(skb, tmp->dev); | |
257 | if (unlikely(ret)) | |
258 | break; | |
259 | } | |
260 | } | |
261 | ||
262 | static void netlink_deliver_tap(struct sk_buff *skb) | |
263 | { | |
264 | rcu_read_lock(); | |
265 | ||
266 | if (unlikely(!list_empty(&netlink_tap_all))) | |
267 | __netlink_deliver_tap(skb); | |
268 | ||
269 | rcu_read_unlock(); | |
270 | } | |
271 | ||
73bfd370 DB |
272 | static void netlink_deliver_tap_kernel(struct sock *dst, struct sock *src, |
273 | struct sk_buff *skb) | |
274 | { | |
275 | if (!(netlink_is_kernel(dst) && netlink_is_kernel(src))) | |
276 | netlink_deliver_tap(skb); | |
277 | } | |
278 | ||
cd1df525 PM |
279 | static void netlink_overrun(struct sock *sk) |
280 | { | |
281 | struct netlink_sock *nlk = nlk_sk(sk); | |
282 | ||
cc3a572f ND |
283 | if (!(nlk->flags & NETLINK_F_RECV_NO_ENOBUFS)) { |
284 | if (!test_and_set_bit(NETLINK_S_CONGESTED, | |
285 | &nlk_sk(sk)->state)) { | |
cd1df525 PM |
286 | sk->sk_err = ENOBUFS; |
287 | sk->sk_error_report(sk); | |
288 | } | |
289 | } | |
290 | atomic_inc(&sk->sk_drops); | |
291 | } | |
292 | ||
293 | static void netlink_rcv_wake(struct sock *sk) | |
294 | { | |
295 | struct netlink_sock *nlk = nlk_sk(sk); | |
296 | ||
297 | if (skb_queue_empty(&sk->sk_receive_queue)) | |
cc3a572f ND |
298 | clear_bit(NETLINK_S_CONGESTED, &nlk->state); |
299 | if (!test_bit(NETLINK_S_CONGESTED, &nlk->state)) | |
cd1df525 PM |
300 | wake_up_interruptible(&nlk->wait); |
301 | } | |
302 | ||
ccdfcc39 | 303 | #ifdef CONFIG_NETLINK_MMAP |
f9c22888 PM |
304 | static bool netlink_rx_is_mmaped(struct sock *sk) |
305 | { | |
306 | return nlk_sk(sk)->rx_ring.pg_vec != NULL; | |
307 | } | |
308 | ||
5fd96123 PM |
309 | static bool netlink_tx_is_mmaped(struct sock *sk) |
310 | { | |
311 | return nlk_sk(sk)->tx_ring.pg_vec != NULL; | |
312 | } | |
313 | ||
ccdfcc39 PM |
314 | static __pure struct page *pgvec_to_page(const void *addr) |
315 | { | |
316 | if (is_vmalloc_addr(addr)) | |
317 | return vmalloc_to_page(addr); | |
318 | else | |
319 | return virt_to_page(addr); | |
320 | } | |
321 | ||
322 | static void free_pg_vec(void **pg_vec, unsigned int order, unsigned int len) | |
323 | { | |
324 | unsigned int i; | |
325 | ||
326 | for (i = 0; i < len; i++) { | |
327 | if (pg_vec[i] != NULL) { | |
328 | if (is_vmalloc_addr(pg_vec[i])) | |
329 | vfree(pg_vec[i]); | |
330 | else | |
331 | free_pages((unsigned long)pg_vec[i], order); | |
332 | } | |
333 | } | |
334 | kfree(pg_vec); | |
335 | } | |
336 | ||
337 | static void *alloc_one_pg_vec_page(unsigned long order) | |
338 | { | |
339 | void *buffer; | |
340 | gfp_t gfp_flags = GFP_KERNEL | __GFP_COMP | __GFP_ZERO | | |
341 | __GFP_NOWARN | __GFP_NORETRY; | |
342 | ||
343 | buffer = (void *)__get_free_pages(gfp_flags, order); | |
344 | if (buffer != NULL) | |
345 | return buffer; | |
346 | ||
347 | buffer = vzalloc((1 << order) * PAGE_SIZE); | |
348 | if (buffer != NULL) | |
349 | return buffer; | |
350 | ||
351 | gfp_flags &= ~__GFP_NORETRY; | |
352 | return (void *)__get_free_pages(gfp_flags, order); | |
353 | } | |
354 | ||
355 | static void **alloc_pg_vec(struct netlink_sock *nlk, | |
356 | struct nl_mmap_req *req, unsigned int order) | |
357 | { | |
358 | unsigned int block_nr = req->nm_block_nr; | |
359 | unsigned int i; | |
8a849bb7 | 360 | void **pg_vec; |
ccdfcc39 PM |
361 | |
362 | pg_vec = kcalloc(block_nr, sizeof(void *), GFP_KERNEL); | |
363 | if (pg_vec == NULL) | |
364 | return NULL; | |
365 | ||
366 | for (i = 0; i < block_nr; i++) { | |
8a849bb7 | 367 | pg_vec[i] = alloc_one_pg_vec_page(order); |
ccdfcc39 PM |
368 | if (pg_vec[i] == NULL) |
369 | goto err1; | |
370 | } | |
371 | ||
372 | return pg_vec; | |
373 | err1: | |
374 | free_pg_vec(pg_vec, order, block_nr); | |
375 | return NULL; | |
376 | } | |
377 | ||
0470eb99 FW |
378 | |
379 | static void | |
380 | __netlink_set_ring(struct sock *sk, struct nl_mmap_req *req, bool tx_ring, void **pg_vec, | |
381 | unsigned int order) | |
382 | { | |
383 | struct netlink_sock *nlk = nlk_sk(sk); | |
384 | struct sk_buff_head *queue; | |
385 | struct netlink_ring *ring; | |
386 | ||
387 | queue = tx_ring ? &sk->sk_write_queue : &sk->sk_receive_queue; | |
388 | ring = tx_ring ? &nlk->tx_ring : &nlk->rx_ring; | |
389 | ||
390 | spin_lock_bh(&queue->lock); | |
391 | ||
392 | ring->frame_max = req->nm_frame_nr - 1; | |
393 | ring->head = 0; | |
394 | ring->frame_size = req->nm_frame_size; | |
395 | ring->pg_vec_pages = req->nm_block_size / PAGE_SIZE; | |
396 | ||
397 | swap(ring->pg_vec_len, req->nm_block_nr); | |
398 | swap(ring->pg_vec_order, order); | |
399 | swap(ring->pg_vec, pg_vec); | |
400 | ||
401 | __skb_queue_purge(queue); | |
402 | spin_unlock_bh(&queue->lock); | |
403 | ||
404 | WARN_ON(atomic_read(&nlk->mapped)); | |
405 | ||
406 | if (pg_vec) | |
407 | free_pg_vec(pg_vec, order, req->nm_block_nr); | |
408 | } | |
409 | ||
ccdfcc39 | 410 | static int netlink_set_ring(struct sock *sk, struct nl_mmap_req *req, |
0470eb99 | 411 | bool tx_ring) |
ccdfcc39 PM |
412 | { |
413 | struct netlink_sock *nlk = nlk_sk(sk); | |
414 | struct netlink_ring *ring; | |
ccdfcc39 PM |
415 | void **pg_vec = NULL; |
416 | unsigned int order = 0; | |
ccdfcc39 PM |
417 | |
418 | ring = tx_ring ? &nlk->tx_ring : &nlk->rx_ring; | |
ccdfcc39 | 419 | |
0470eb99 FW |
420 | if (atomic_read(&nlk->mapped)) |
421 | return -EBUSY; | |
422 | if (atomic_read(&ring->pending)) | |
423 | return -EBUSY; | |
ccdfcc39 PM |
424 | |
425 | if (req->nm_block_nr) { | |
426 | if (ring->pg_vec != NULL) | |
427 | return -EBUSY; | |
428 | ||
429 | if ((int)req->nm_block_size <= 0) | |
430 | return -EINVAL; | |
74e83b23 | 431 | if (!PAGE_ALIGNED(req->nm_block_size)) |
ccdfcc39 PM |
432 | return -EINVAL; |
433 | if (req->nm_frame_size < NL_MMAP_HDRLEN) | |
434 | return -EINVAL; | |
435 | if (!IS_ALIGNED(req->nm_frame_size, NL_MMAP_MSG_ALIGNMENT)) | |
436 | return -EINVAL; | |
437 | ||
438 | ring->frames_per_block = req->nm_block_size / | |
439 | req->nm_frame_size; | |
440 | if (ring->frames_per_block == 0) | |
441 | return -EINVAL; | |
442 | if (ring->frames_per_block * req->nm_block_nr != | |
443 | req->nm_frame_nr) | |
444 | return -EINVAL; | |
445 | ||
446 | order = get_order(req->nm_block_size); | |
447 | pg_vec = alloc_pg_vec(nlk, req, order); | |
448 | if (pg_vec == NULL) | |
449 | return -ENOMEM; | |
450 | } else { | |
451 | if (req->nm_frame_nr) | |
452 | return -EINVAL; | |
453 | } | |
454 | ||
ccdfcc39 | 455 | mutex_lock(&nlk->pg_vec_lock); |
0470eb99 FW |
456 | if (atomic_read(&nlk->mapped) == 0) { |
457 | __netlink_set_ring(sk, req, tx_ring, pg_vec, order); | |
458 | mutex_unlock(&nlk->pg_vec_lock); | |
459 | return 0; | |
ccdfcc39 | 460 | } |
0470eb99 | 461 | |
ccdfcc39 PM |
462 | mutex_unlock(&nlk->pg_vec_lock); |
463 | ||
464 | if (pg_vec) | |
465 | free_pg_vec(pg_vec, order, req->nm_block_nr); | |
0470eb99 FW |
466 | |
467 | return -EBUSY; | |
ccdfcc39 PM |
468 | } |
469 | ||
470 | static void netlink_mm_open(struct vm_area_struct *vma) | |
471 | { | |
472 | struct file *file = vma->vm_file; | |
473 | struct socket *sock = file->private_data; | |
474 | struct sock *sk = sock->sk; | |
475 | ||
476 | if (sk) | |
477 | atomic_inc(&nlk_sk(sk)->mapped); | |
478 | } | |
479 | ||
480 | static void netlink_mm_close(struct vm_area_struct *vma) | |
481 | { | |
482 | struct file *file = vma->vm_file; | |
483 | struct socket *sock = file->private_data; | |
484 | struct sock *sk = sock->sk; | |
485 | ||
486 | if (sk) | |
487 | atomic_dec(&nlk_sk(sk)->mapped); | |
488 | } | |
489 | ||
490 | static const struct vm_operations_struct netlink_mmap_ops = { | |
491 | .open = netlink_mm_open, | |
492 | .close = netlink_mm_close, | |
493 | }; | |
494 | ||
495 | static int netlink_mmap(struct file *file, struct socket *sock, | |
496 | struct vm_area_struct *vma) | |
497 | { | |
498 | struct sock *sk = sock->sk; | |
499 | struct netlink_sock *nlk = nlk_sk(sk); | |
500 | struct netlink_ring *ring; | |
501 | unsigned long start, size, expected; | |
502 | unsigned int i; | |
503 | int err = -EINVAL; | |
504 | ||
505 | if (vma->vm_pgoff) | |
506 | return -EINVAL; | |
507 | ||
508 | mutex_lock(&nlk->pg_vec_lock); | |
509 | ||
510 | expected = 0; | |
511 | for (ring = &nlk->rx_ring; ring <= &nlk->tx_ring; ring++) { | |
512 | if (ring->pg_vec == NULL) | |
513 | continue; | |
514 | expected += ring->pg_vec_len * ring->pg_vec_pages * PAGE_SIZE; | |
515 | } | |
516 | ||
517 | if (expected == 0) | |
518 | goto out; | |
519 | ||
520 | size = vma->vm_end - vma->vm_start; | |
521 | if (size != expected) | |
522 | goto out; | |
523 | ||
524 | start = vma->vm_start; | |
525 | for (ring = &nlk->rx_ring; ring <= &nlk->tx_ring; ring++) { | |
526 | if (ring->pg_vec == NULL) | |
527 | continue; | |
528 | ||
529 | for (i = 0; i < ring->pg_vec_len; i++) { | |
530 | struct page *page; | |
531 | void *kaddr = ring->pg_vec[i]; | |
532 | unsigned int pg_num; | |
533 | ||
534 | for (pg_num = 0; pg_num < ring->pg_vec_pages; pg_num++) { | |
535 | page = pgvec_to_page(kaddr); | |
536 | err = vm_insert_page(vma, start, page); | |
537 | if (err < 0) | |
538 | goto out; | |
539 | start += PAGE_SIZE; | |
540 | kaddr += PAGE_SIZE; | |
541 | } | |
542 | } | |
543 | } | |
544 | ||
545 | atomic_inc(&nlk->mapped); | |
546 | vma->vm_ops = &netlink_mmap_ops; | |
547 | err = 0; | |
548 | out: | |
549 | mutex_unlock(&nlk->pg_vec_lock); | |
7cdbac71 | 550 | return err; |
ccdfcc39 | 551 | } |
9652e931 | 552 | |
4682a035 | 553 | static void netlink_frame_flush_dcache(const struct nl_mmap_hdr *hdr, unsigned int nm_len) |
9652e931 PM |
554 | { |
555 | #if ARCH_IMPLEMENTS_FLUSH_DCACHE_PAGE == 1 | |
556 | struct page *p_start, *p_end; | |
557 | ||
558 | /* First page is flushed through netlink_{get,set}_status */ | |
559 | p_start = pgvec_to_page(hdr + PAGE_SIZE); | |
4682a035 | 560 | p_end = pgvec_to_page((void *)hdr + NL_MMAP_HDRLEN + nm_len - 1); |
9652e931 PM |
561 | while (p_start <= p_end) { |
562 | flush_dcache_page(p_start); | |
563 | p_start++; | |
564 | } | |
565 | #endif | |
566 | } | |
567 | ||
568 | static enum nl_mmap_status netlink_get_status(const struct nl_mmap_hdr *hdr) | |
569 | { | |
570 | smp_rmb(); | |
571 | flush_dcache_page(pgvec_to_page(hdr)); | |
572 | return hdr->nm_status; | |
573 | } | |
574 | ||
575 | static void netlink_set_status(struct nl_mmap_hdr *hdr, | |
576 | enum nl_mmap_status status) | |
577 | { | |
a18e6a18 | 578 | smp_mb(); |
9652e931 PM |
579 | hdr->nm_status = status; |
580 | flush_dcache_page(pgvec_to_page(hdr)); | |
9652e931 PM |
581 | } |
582 | ||
583 | static struct nl_mmap_hdr * | |
584 | __netlink_lookup_frame(const struct netlink_ring *ring, unsigned int pos) | |
585 | { | |
586 | unsigned int pg_vec_pos, frame_off; | |
587 | ||
588 | pg_vec_pos = pos / ring->frames_per_block; | |
589 | frame_off = pos % ring->frames_per_block; | |
590 | ||
591 | return ring->pg_vec[pg_vec_pos] + (frame_off * ring->frame_size); | |
592 | } | |
593 | ||
594 | static struct nl_mmap_hdr * | |
595 | netlink_lookup_frame(const struct netlink_ring *ring, unsigned int pos, | |
596 | enum nl_mmap_status status) | |
597 | { | |
598 | struct nl_mmap_hdr *hdr; | |
599 | ||
600 | hdr = __netlink_lookup_frame(ring, pos); | |
601 | if (netlink_get_status(hdr) != status) | |
602 | return NULL; | |
603 | ||
604 | return hdr; | |
605 | } | |
606 | ||
607 | static struct nl_mmap_hdr * | |
608 | netlink_current_frame(const struct netlink_ring *ring, | |
609 | enum nl_mmap_status status) | |
610 | { | |
611 | return netlink_lookup_frame(ring, ring->head, status); | |
612 | } | |
613 | ||
9652e931 PM |
614 | static void netlink_increment_head(struct netlink_ring *ring) |
615 | { | |
616 | ring->head = ring->head != ring->frame_max ? ring->head + 1 : 0; | |
617 | } | |
618 | ||
619 | static void netlink_forward_ring(struct netlink_ring *ring) | |
620 | { | |
7084a315 | 621 | unsigned int head = ring->head; |
9652e931 PM |
622 | const struct nl_mmap_hdr *hdr; |
623 | ||
624 | do { | |
7084a315 | 625 | hdr = __netlink_lookup_frame(ring, ring->head); |
9652e931 PM |
626 | if (hdr->nm_status == NL_MMAP_STATUS_UNUSED) |
627 | break; | |
628 | if (hdr->nm_status != NL_MMAP_STATUS_SKIP) | |
629 | break; | |
630 | netlink_increment_head(ring); | |
631 | } while (ring->head != head); | |
632 | } | |
633 | ||
0ef70770 KM |
634 | static bool netlink_has_valid_frame(struct netlink_ring *ring) |
635 | { | |
636 | unsigned int head = ring->head, pos = head; | |
637 | const struct nl_mmap_hdr *hdr; | |
638 | ||
639 | do { | |
640 | hdr = __netlink_lookup_frame(ring, pos); | |
641 | if (hdr->nm_status == NL_MMAP_STATUS_VALID) | |
642 | return true; | |
643 | pos = pos != 0 ? pos - 1 : ring->frame_max; | |
644 | } while (pos != head); | |
645 | ||
646 | return false; | |
647 | } | |
648 | ||
cd1df525 PM |
649 | static bool netlink_dump_space(struct netlink_sock *nlk) |
650 | { | |
651 | struct netlink_ring *ring = &nlk->rx_ring; | |
652 | struct nl_mmap_hdr *hdr; | |
653 | unsigned int n; | |
654 | ||
655 | hdr = netlink_current_frame(ring, NL_MMAP_STATUS_UNUSED); | |
656 | if (hdr == NULL) | |
657 | return false; | |
658 | ||
659 | n = ring->head + ring->frame_max / 2; | |
660 | if (n > ring->frame_max) | |
661 | n -= ring->frame_max; | |
662 | ||
663 | hdr = __netlink_lookup_frame(ring, n); | |
664 | ||
665 | return hdr->nm_status == NL_MMAP_STATUS_UNUSED; | |
666 | } | |
667 | ||
9652e931 PM |
668 | static unsigned int netlink_poll(struct file *file, struct socket *sock, |
669 | poll_table *wait) | |
670 | { | |
671 | struct sock *sk = sock->sk; | |
672 | struct netlink_sock *nlk = nlk_sk(sk); | |
673 | unsigned int mask; | |
cd1df525 | 674 | int err; |
9652e931 | 675 | |
cd1df525 PM |
676 | if (nlk->rx_ring.pg_vec != NULL) { |
677 | /* Memory mapped sockets don't call recvmsg(), so flow control | |
678 | * for dumps is performed here. A dump is allowed to continue | |
679 | * if at least half the ring is unused. | |
680 | */ | |
16b304f3 | 681 | while (nlk->cb_running && netlink_dump_space(nlk)) { |
cd1df525 PM |
682 | err = netlink_dump(sk); |
683 | if (err < 0) { | |
ac30ef83 | 684 | sk->sk_err = -err; |
cd1df525 PM |
685 | sk->sk_error_report(sk); |
686 | break; | |
687 | } | |
688 | } | |
689 | netlink_rcv_wake(sk); | |
690 | } | |
5fd96123 | 691 | |
9652e931 PM |
692 | mask = datagram_poll(file, sock, wait); |
693 | ||
a66e3656 DB |
694 | /* We could already have received frames in the normal receive |
695 | * queue, that will show up as NL_MMAP_STATUS_COPY in the ring, | |
696 | * so if mask contains pollin/etc already, there's no point | |
697 | * walking the ring. | |
698 | */ | |
699 | if ((mask & (POLLIN | POLLRDNORM)) != (POLLIN | POLLRDNORM)) { | |
700 | spin_lock_bh(&sk->sk_receive_queue.lock); | |
701 | if (nlk->rx_ring.pg_vec) { | |
702 | if (netlink_has_valid_frame(&nlk->rx_ring)) | |
703 | mask |= POLLIN | POLLRDNORM; | |
704 | } | |
705 | spin_unlock_bh(&sk->sk_receive_queue.lock); | |
9652e931 | 706 | } |
9652e931 PM |
707 | |
708 | spin_lock_bh(&sk->sk_write_queue.lock); | |
709 | if (nlk->tx_ring.pg_vec) { | |
710 | if (netlink_current_frame(&nlk->tx_ring, NL_MMAP_STATUS_UNUSED)) | |
711 | mask |= POLLOUT | POLLWRNORM; | |
712 | } | |
713 | spin_unlock_bh(&sk->sk_write_queue.lock); | |
714 | ||
715 | return mask; | |
716 | } | |
717 | ||
718 | static struct nl_mmap_hdr *netlink_mmap_hdr(struct sk_buff *skb) | |
719 | { | |
720 | return (struct nl_mmap_hdr *)(skb->head - NL_MMAP_HDRLEN); | |
721 | } | |
722 | ||
723 | static void netlink_ring_setup_skb(struct sk_buff *skb, struct sock *sk, | |
724 | struct netlink_ring *ring, | |
725 | struct nl_mmap_hdr *hdr) | |
726 | { | |
727 | unsigned int size; | |
728 | void *data; | |
729 | ||
730 | size = ring->frame_size - NL_MMAP_HDRLEN; | |
731 | data = (void *)hdr + NL_MMAP_HDRLEN; | |
732 | ||
733 | skb->head = data; | |
734 | skb->data = data; | |
735 | skb_reset_tail_pointer(skb); | |
736 | skb->end = skb->tail + size; | |
737 | skb->len = 0; | |
738 | ||
739 | skb->destructor = netlink_skb_destructor; | |
740 | NETLINK_CB(skb).flags |= NETLINK_SKB_MMAPED; | |
741 | NETLINK_CB(skb).sk = sk; | |
742 | } | |
5fd96123 PM |
743 | |
744 | static int netlink_mmap_sendmsg(struct sock *sk, struct msghdr *msg, | |
745 | u32 dst_portid, u32 dst_group, | |
7cc05662 | 746 | struct scm_cookie *scm) |
5fd96123 PM |
747 | { |
748 | struct netlink_sock *nlk = nlk_sk(sk); | |
749 | struct netlink_ring *ring; | |
750 | struct nl_mmap_hdr *hdr; | |
751 | struct sk_buff *skb; | |
752 | unsigned int maxlen; | |
5fd96123 PM |
753 | int err = 0, len = 0; |
754 | ||
5fd96123 PM |
755 | mutex_lock(&nlk->pg_vec_lock); |
756 | ||
757 | ring = &nlk->tx_ring; | |
758 | maxlen = ring->frame_size - NL_MMAP_HDRLEN; | |
759 | ||
760 | do { | |
4682a035 DM |
761 | unsigned int nm_len; |
762 | ||
5fd96123 PM |
763 | hdr = netlink_current_frame(ring, NL_MMAP_STATUS_VALID); |
764 | if (hdr == NULL) { | |
765 | if (!(msg->msg_flags & MSG_DONTWAIT) && | |
766 | atomic_read(&nlk->tx_ring.pending)) | |
767 | schedule(); | |
768 | continue; | |
769 | } | |
4682a035 DM |
770 | |
771 | nm_len = ACCESS_ONCE(hdr->nm_len); | |
772 | if (nm_len > maxlen) { | |
5fd96123 PM |
773 | err = -EINVAL; |
774 | goto out; | |
775 | } | |
776 | ||
4682a035 | 777 | netlink_frame_flush_dcache(hdr, nm_len); |
5fd96123 | 778 | |
4682a035 DM |
779 | skb = alloc_skb(nm_len, GFP_KERNEL); |
780 | if (skb == NULL) { | |
781 | err = -ENOBUFS; | |
782 | goto out; | |
5fd96123 | 783 | } |
4682a035 DM |
784 | __skb_put(skb, nm_len); |
785 | memcpy(skb->data, (void *)hdr + NL_MMAP_HDRLEN, nm_len); | |
786 | netlink_set_status(hdr, NL_MMAP_STATUS_UNUSED); | |
5fd96123 PM |
787 | |
788 | netlink_increment_head(ring); | |
789 | ||
790 | NETLINK_CB(skb).portid = nlk->portid; | |
791 | NETLINK_CB(skb).dst_group = dst_group; | |
7cc05662 | 792 | NETLINK_CB(skb).creds = scm->creds; |
5fd96123 PM |
793 | |
794 | err = security_netlink_send(sk, skb); | |
795 | if (err) { | |
796 | kfree_skb(skb); | |
797 | goto out; | |
798 | } | |
799 | ||
800 | if (unlikely(dst_group)) { | |
801 | atomic_inc(&skb->users); | |
802 | netlink_broadcast(sk, skb, dst_portid, dst_group, | |
803 | GFP_KERNEL); | |
804 | } | |
805 | err = netlink_unicast(sk, skb, dst_portid, | |
806 | msg->msg_flags & MSG_DONTWAIT); | |
807 | if (err < 0) | |
808 | goto out; | |
809 | len += err; | |
810 | ||
811 | } while (hdr != NULL || | |
812 | (!(msg->msg_flags & MSG_DONTWAIT) && | |
813 | atomic_read(&nlk->tx_ring.pending))); | |
814 | ||
815 | if (len > 0) | |
816 | err = len; | |
817 | out: | |
818 | mutex_unlock(&nlk->pg_vec_lock); | |
819 | return err; | |
820 | } | |
f9c22888 PM |
821 | |
822 | static void netlink_queue_mmaped_skb(struct sock *sk, struct sk_buff *skb) | |
823 | { | |
824 | struct nl_mmap_hdr *hdr; | |
825 | ||
826 | hdr = netlink_mmap_hdr(skb); | |
827 | hdr->nm_len = skb->len; | |
828 | hdr->nm_group = NETLINK_CB(skb).dst_group; | |
829 | hdr->nm_pid = NETLINK_CB(skb).creds.pid; | |
1bf9310a ND |
830 | hdr->nm_uid = from_kuid(sk_user_ns(sk), NETLINK_CB(skb).creds.uid); |
831 | hdr->nm_gid = from_kgid(sk_user_ns(sk), NETLINK_CB(skb).creds.gid); | |
4682a035 | 832 | netlink_frame_flush_dcache(hdr, hdr->nm_len); |
f9c22888 PM |
833 | netlink_set_status(hdr, NL_MMAP_STATUS_VALID); |
834 | ||
835 | NETLINK_CB(skb).flags |= NETLINK_SKB_DELIVERED; | |
836 | kfree_skb(skb); | |
837 | } | |
838 | ||
839 | static void netlink_ring_set_copied(struct sock *sk, struct sk_buff *skb) | |
840 | { | |
841 | struct netlink_sock *nlk = nlk_sk(sk); | |
842 | struct netlink_ring *ring = &nlk->rx_ring; | |
843 | struct nl_mmap_hdr *hdr; | |
844 | ||
845 | spin_lock_bh(&sk->sk_receive_queue.lock); | |
846 | hdr = netlink_current_frame(ring, NL_MMAP_STATUS_UNUSED); | |
847 | if (hdr == NULL) { | |
848 | spin_unlock_bh(&sk->sk_receive_queue.lock); | |
849 | kfree_skb(skb); | |
cd1df525 | 850 | netlink_overrun(sk); |
f9c22888 PM |
851 | return; |
852 | } | |
853 | netlink_increment_head(ring); | |
854 | __skb_queue_tail(&sk->sk_receive_queue, skb); | |
855 | spin_unlock_bh(&sk->sk_receive_queue.lock); | |
856 | ||
857 | hdr->nm_len = skb->len; | |
858 | hdr->nm_group = NETLINK_CB(skb).dst_group; | |
859 | hdr->nm_pid = NETLINK_CB(skb).creds.pid; | |
1bf9310a ND |
860 | hdr->nm_uid = from_kuid(sk_user_ns(sk), NETLINK_CB(skb).creds.uid); |
861 | hdr->nm_gid = from_kgid(sk_user_ns(sk), NETLINK_CB(skb).creds.gid); | |
f9c22888 PM |
862 | netlink_set_status(hdr, NL_MMAP_STATUS_COPY); |
863 | } | |
864 | ||
ccdfcc39 | 865 | #else /* CONFIG_NETLINK_MMAP */ |
f9c22888 | 866 | #define netlink_rx_is_mmaped(sk) false |
5fd96123 | 867 | #define netlink_tx_is_mmaped(sk) false |
ccdfcc39 | 868 | #define netlink_mmap sock_no_mmap |
9652e931 | 869 | #define netlink_poll datagram_poll |
7cc05662 | 870 | #define netlink_mmap_sendmsg(sk, msg, dst_portid, dst_group, scm) 0 |
ccdfcc39 PM |
871 | #endif /* CONFIG_NETLINK_MMAP */ |
872 | ||
cf0a018a PM |
873 | static void netlink_skb_destructor(struct sk_buff *skb) |
874 | { | |
9652e931 PM |
875 | #ifdef CONFIG_NETLINK_MMAP |
876 | struct nl_mmap_hdr *hdr; | |
877 | struct netlink_ring *ring; | |
878 | struct sock *sk; | |
879 | ||
880 | /* If a packet from the kernel to userspace was freed because of an | |
881 | * error without being delivered to userspace, the kernel must reset | |
882 | * the status. In the direction userspace to kernel, the status is | |
883 | * always reset here after the packet was processed and freed. | |
884 | */ | |
885 | if (netlink_skb_is_mmaped(skb)) { | |
886 | hdr = netlink_mmap_hdr(skb); | |
887 | sk = NETLINK_CB(skb).sk; | |
888 | ||
5fd96123 PM |
889 | if (NETLINK_CB(skb).flags & NETLINK_SKB_TX) { |
890 | netlink_set_status(hdr, NL_MMAP_STATUS_UNUSED); | |
891 | ring = &nlk_sk(sk)->tx_ring; | |
892 | } else { | |
893 | if (!(NETLINK_CB(skb).flags & NETLINK_SKB_DELIVERED)) { | |
894 | hdr->nm_len = 0; | |
895 | netlink_set_status(hdr, NL_MMAP_STATUS_VALID); | |
896 | } | |
897 | ring = &nlk_sk(sk)->rx_ring; | |
9652e931 | 898 | } |
9652e931 PM |
899 | |
900 | WARN_ON(atomic_read(&ring->pending) == 0); | |
901 | atomic_dec(&ring->pending); | |
902 | sock_put(sk); | |
903 | ||
5e71d9d7 | 904 | skb->head = NULL; |
9652e931 PM |
905 | } |
906 | #endif | |
c05cdb1b | 907 | if (is_vmalloc_addr(skb->head)) { |
3a36515f PN |
908 | if (!skb->cloned || |
909 | !atomic_dec_return(&(skb_shinfo(skb)->dataref))) | |
910 | vfree(skb->head); | |
911 | ||
c05cdb1b PNA |
912 | skb->head = NULL; |
913 | } | |
9652e931 PM |
914 | if (skb->sk != NULL) |
915 | sock_rfree(skb); | |
cf0a018a PM |
916 | } |
917 | ||
918 | static void netlink_skb_set_owner_r(struct sk_buff *skb, struct sock *sk) | |
919 | { | |
920 | WARN_ON(skb->sk != NULL); | |
921 | skb->sk = sk; | |
922 | skb->destructor = netlink_skb_destructor; | |
923 | atomic_add(skb->truesize, &sk->sk_rmem_alloc); | |
924 | sk_mem_charge(sk, skb->truesize); | |
925 | } | |
926 | ||
1da177e4 LT |
927 | static void netlink_sock_destruct(struct sock *sk) |
928 | { | |
3f660d66 HX |
929 | struct netlink_sock *nlk = nlk_sk(sk); |
930 | ||
16b304f3 PS |
931 | if (nlk->cb_running) { |
932 | if (nlk->cb.done) | |
933 | nlk->cb.done(&nlk->cb); | |
6dc878a8 | 934 | |
16b304f3 PS |
935 | module_put(nlk->cb.module); |
936 | kfree_skb(nlk->cb.skb); | |
3f660d66 HX |
937 | } |
938 | ||
1da177e4 | 939 | skb_queue_purge(&sk->sk_receive_queue); |
ccdfcc39 PM |
940 | #ifdef CONFIG_NETLINK_MMAP |
941 | if (1) { | |
942 | struct nl_mmap_req req; | |
943 | ||
944 | memset(&req, 0, sizeof(req)); | |
945 | if (nlk->rx_ring.pg_vec) | |
0470eb99 | 946 | __netlink_set_ring(sk, &req, false, NULL, 0); |
ccdfcc39 PM |
947 | memset(&req, 0, sizeof(req)); |
948 | if (nlk->tx_ring.pg_vec) | |
0470eb99 | 949 | __netlink_set_ring(sk, &req, true, NULL, 0); |
ccdfcc39 PM |
950 | } |
951 | #endif /* CONFIG_NETLINK_MMAP */ | |
1da177e4 LT |
952 | |
953 | if (!sock_flag(sk, SOCK_DEAD)) { | |
6ac552fd | 954 | printk(KERN_ERR "Freeing alive netlink socket %p\n", sk); |
1da177e4 LT |
955 | return; |
956 | } | |
547b792c IJ |
957 | |
958 | WARN_ON(atomic_read(&sk->sk_rmem_alloc)); | |
959 | WARN_ON(atomic_read(&sk->sk_wmem_alloc)); | |
960 | WARN_ON(nlk_sk(sk)->groups); | |
1da177e4 LT |
961 | } |
962 | ||
6ac552fd PM |
963 | /* This lock without WQ_FLAG_EXCLUSIVE is good on UP and it is _very_ bad on |
964 | * SMP. Look, when several writers sleep and reader wakes them up, all but one | |
1da177e4 LT |
965 | * immediately hit write lock and grab all the cpus. Exclusive sleep solves |
966 | * this, _but_ remember, it adds useless work on UP machines. | |
967 | */ | |
968 | ||
d136f1bd | 969 | void netlink_table_grab(void) |
9a429c49 | 970 | __acquires(nl_table_lock) |
1da177e4 | 971 | { |
d136f1bd JB |
972 | might_sleep(); |
973 | ||
6abd219c | 974 | write_lock_irq(&nl_table_lock); |
1da177e4 LT |
975 | |
976 | if (atomic_read(&nl_table_users)) { | |
977 | DECLARE_WAITQUEUE(wait, current); | |
978 | ||
979 | add_wait_queue_exclusive(&nl_table_wait, &wait); | |
6ac552fd | 980 | for (;;) { |
1da177e4 LT |
981 | set_current_state(TASK_UNINTERRUPTIBLE); |
982 | if (atomic_read(&nl_table_users) == 0) | |
983 | break; | |
6abd219c | 984 | write_unlock_irq(&nl_table_lock); |
1da177e4 | 985 | schedule(); |
6abd219c | 986 | write_lock_irq(&nl_table_lock); |
1da177e4 LT |
987 | } |
988 | ||
989 | __set_current_state(TASK_RUNNING); | |
990 | remove_wait_queue(&nl_table_wait, &wait); | |
991 | } | |
992 | } | |
993 | ||
d136f1bd | 994 | void netlink_table_ungrab(void) |
9a429c49 | 995 | __releases(nl_table_lock) |
1da177e4 | 996 | { |
6abd219c | 997 | write_unlock_irq(&nl_table_lock); |
1da177e4 LT |
998 | wake_up(&nl_table_wait); |
999 | } | |
1000 | ||
6ac552fd | 1001 | static inline void |
1da177e4 LT |
1002 | netlink_lock_table(void) |
1003 | { | |
1004 | /* read_lock() synchronizes us to netlink_table_grab */ | |
1005 | ||
1006 | read_lock(&nl_table_lock); | |
1007 | atomic_inc(&nl_table_users); | |
1008 | read_unlock(&nl_table_lock); | |
1009 | } | |
1010 | ||
6ac552fd | 1011 | static inline void |
1da177e4 LT |
1012 | netlink_unlock_table(void) |
1013 | { | |
1014 | if (atomic_dec_and_test(&nl_table_users)) | |
1015 | wake_up(&nl_table_wait); | |
1016 | } | |
1017 | ||
e341694e | 1018 | struct netlink_compare_arg |
1da177e4 | 1019 | { |
c428ecd1 | 1020 | possible_net_t pnet; |
e341694e TG |
1021 | u32 portid; |
1022 | }; | |
1da177e4 | 1023 | |
8f2ddaac HX |
1024 | /* Doing sizeof directly may yield 4 extra bytes on 64-bit. */ |
1025 | #define netlink_compare_arg_len \ | |
1026 | (offsetof(struct netlink_compare_arg, portid) + sizeof(u32)) | |
c428ecd1 HX |
1027 | |
1028 | static inline int netlink_compare(struct rhashtable_compare_arg *arg, | |
1029 | const void *ptr) | |
1da177e4 | 1030 | { |
c428ecd1 HX |
1031 | const struct netlink_compare_arg *x = arg->key; |
1032 | const struct netlink_sock *nlk = ptr; | |
1da177e4 | 1033 | |
da314c99 | 1034 | return nlk->portid != x->portid || |
c428ecd1 HX |
1035 | !net_eq(sock_net(&nlk->sk), read_pnet(&x->pnet)); |
1036 | } | |
1037 | ||
1038 | static void netlink_compare_arg_init(struct netlink_compare_arg *arg, | |
1039 | struct net *net, u32 portid) | |
1040 | { | |
1041 | memset(arg, 0, sizeof(*arg)); | |
1042 | write_pnet(&arg->pnet, net); | |
1043 | arg->portid = portid; | |
1da177e4 LT |
1044 | } |
1045 | ||
e341694e TG |
1046 | static struct sock *__netlink_lookup(struct netlink_table *table, u32 portid, |
1047 | struct net *net) | |
1da177e4 | 1048 | { |
c428ecd1 | 1049 | struct netlink_compare_arg arg; |
1da177e4 | 1050 | |
c428ecd1 HX |
1051 | netlink_compare_arg_init(&arg, net, portid); |
1052 | return rhashtable_lookup_fast(&table->hash, &arg, | |
1053 | netlink_rhashtable_params); | |
1da177e4 LT |
1054 | } |
1055 | ||
c428ecd1 | 1056 | static int __netlink_insert(struct netlink_table *table, struct sock *sk) |
c5adde94 | 1057 | { |
c428ecd1 | 1058 | struct netlink_compare_arg arg; |
c5adde94 | 1059 | |
da314c99 | 1060 | netlink_compare_arg_init(&arg, sock_net(sk), nlk_sk(sk)->portid); |
c428ecd1 HX |
1061 | return rhashtable_lookup_insert_key(&table->hash, &arg, |
1062 | &nlk_sk(sk)->node, | |
1063 | netlink_rhashtable_params); | |
c5adde94 YX |
1064 | } |
1065 | ||
e341694e | 1066 | static struct sock *netlink_lookup(struct net *net, int protocol, u32 portid) |
1da177e4 | 1067 | { |
e341694e TG |
1068 | struct netlink_table *table = &nl_table[protocol]; |
1069 | struct sock *sk; | |
1da177e4 | 1070 | |
e341694e TG |
1071 | rcu_read_lock(); |
1072 | sk = __netlink_lookup(table, portid, net); | |
1073 | if (sk) | |
1074 | sock_hold(sk); | |
1075 | rcu_read_unlock(); | |
1da177e4 | 1076 | |
e341694e | 1077 | return sk; |
1da177e4 LT |
1078 | } |
1079 | ||
90ddc4f0 | 1080 | static const struct proto_ops netlink_ops; |
1da177e4 | 1081 | |
4277a083 PM |
1082 | static void |
1083 | netlink_update_listeners(struct sock *sk) | |
1084 | { | |
1085 | struct netlink_table *tbl = &nl_table[sk->sk_protocol]; | |
4277a083 PM |
1086 | unsigned long mask; |
1087 | unsigned int i; | |
6d772ac5 ED |
1088 | struct listeners *listeners; |
1089 | ||
1090 | listeners = nl_deref_protected(tbl->listeners); | |
1091 | if (!listeners) | |
1092 | return; | |
4277a083 | 1093 | |
b4ff4f04 | 1094 | for (i = 0; i < NLGRPLONGS(tbl->groups); i++) { |
4277a083 | 1095 | mask = 0; |
b67bfe0d | 1096 | sk_for_each_bound(sk, &tbl->mc_list) { |
b4ff4f04 JB |
1097 | if (i < NLGRPLONGS(nlk_sk(sk)->ngroups)) |
1098 | mask |= nlk_sk(sk)->groups[i]; | |
1099 | } | |
6d772ac5 | 1100 | listeners->masks[i] = mask; |
4277a083 PM |
1101 | } |
1102 | /* this function is only called with the netlink table "grabbed", which | |
1103 | * makes sure updates are visible before bind or setsockopt return. */ | |
1104 | } | |
1105 | ||
8ea65f4a | 1106 | static int netlink_insert(struct sock *sk, u32 portid) |
1da177e4 | 1107 | { |
da12c90e | 1108 | struct netlink_table *table = &nl_table[sk->sk_protocol]; |
919d9db9 | 1109 | int err; |
1da177e4 | 1110 | |
c5adde94 | 1111 | lock_sock(sk); |
1da177e4 | 1112 | |
da314c99 HX |
1113 | err = nlk_sk(sk)->portid == portid ? 0 : -EBUSY; |
1114 | if (nlk_sk(sk)->bound) | |
1da177e4 LT |
1115 | goto err; |
1116 | ||
1117 | err = -ENOMEM; | |
97defe1e TG |
1118 | if (BITS_PER_LONG > 32 && |
1119 | unlikely(atomic_read(&table->hash.nelems) >= UINT_MAX)) | |
1da177e4 LT |
1120 | goto err; |
1121 | ||
da314c99 | 1122 | nlk_sk(sk)->portid = portid; |
e341694e | 1123 | sock_hold(sk); |
919d9db9 | 1124 | |
c428ecd1 HX |
1125 | err = __netlink_insert(table, sk); |
1126 | if (err) { | |
4e7c1330 DB |
1127 | /* In case the hashtable backend returns with -EBUSY |
1128 | * from here, it must not escape to the caller. | |
1129 | */ | |
1130 | if (unlikely(err == -EBUSY)) | |
1131 | err = -EOVERFLOW; | |
c428ecd1 HX |
1132 | if (err == -EEXIST) |
1133 | err = -EADDRINUSE; | |
c5adde94 | 1134 | sock_put(sk); |
1f770c0a | 1135 | goto err; |
919d9db9 HX |
1136 | } |
1137 | ||
da314c99 HX |
1138 | /* We need to ensure that the socket is hashed and visible. */ |
1139 | smp_wmb(); | |
1140 | nlk_sk(sk)->bound = portid; | |
1f770c0a | 1141 | |
1da177e4 | 1142 | err: |
c5adde94 | 1143 | release_sock(sk); |
1da177e4 LT |
1144 | return err; |
1145 | } | |
1146 | ||
1147 | static void netlink_remove(struct sock *sk) | |
1148 | { | |
e341694e TG |
1149 | struct netlink_table *table; |
1150 | ||
e341694e | 1151 | table = &nl_table[sk->sk_protocol]; |
c428ecd1 HX |
1152 | if (!rhashtable_remove_fast(&table->hash, &nlk_sk(sk)->node, |
1153 | netlink_rhashtable_params)) { | |
e341694e TG |
1154 | WARN_ON(atomic_read(&sk->sk_refcnt) == 1); |
1155 | __sock_put(sk); | |
1156 | } | |
e341694e | 1157 | |
1da177e4 | 1158 | netlink_table_grab(); |
b10dcb3b | 1159 | if (nlk_sk(sk)->subscriptions) { |
1da177e4 | 1160 | __sk_del_bind_node(sk); |
b10dcb3b JB |
1161 | netlink_update_listeners(sk); |
1162 | } | |
ee1c2442 JB |
1163 | if (sk->sk_protocol == NETLINK_GENERIC) |
1164 | atomic_inc(&genl_sk_destructing_cnt); | |
1da177e4 LT |
1165 | netlink_table_ungrab(); |
1166 | } | |
1167 | ||
1168 | static struct proto netlink_proto = { | |
1169 | .name = "NETLINK", | |
1170 | .owner = THIS_MODULE, | |
1171 | .obj_size = sizeof(struct netlink_sock), | |
1172 | }; | |
1173 | ||
1b8d7ae4 | 1174 | static int __netlink_create(struct net *net, struct socket *sock, |
11aa9c28 EB |
1175 | struct mutex *cb_mutex, int protocol, |
1176 | int kern) | |
1da177e4 LT |
1177 | { |
1178 | struct sock *sk; | |
1179 | struct netlink_sock *nlk; | |
ab33a171 PM |
1180 | |
1181 | sock->ops = &netlink_ops; | |
1182 | ||
11aa9c28 | 1183 | sk = sk_alloc(net, PF_NETLINK, GFP_KERNEL, &netlink_proto, kern); |
ab33a171 PM |
1184 | if (!sk) |
1185 | return -ENOMEM; | |
1186 | ||
1187 | sock_init_data(sock, sk); | |
1188 | ||
1189 | nlk = nlk_sk(sk); | |
658cb354 | 1190 | if (cb_mutex) { |
ffa4d721 | 1191 | nlk->cb_mutex = cb_mutex; |
658cb354 | 1192 | } else { |
ffa4d721 PM |
1193 | nlk->cb_mutex = &nlk->cb_def_mutex; |
1194 | mutex_init(nlk->cb_mutex); | |
1195 | } | |
ab33a171 | 1196 | init_waitqueue_head(&nlk->wait); |
ccdfcc39 PM |
1197 | #ifdef CONFIG_NETLINK_MMAP |
1198 | mutex_init(&nlk->pg_vec_lock); | |
1199 | #endif | |
ab33a171 PM |
1200 | |
1201 | sk->sk_destruct = netlink_sock_destruct; | |
1202 | sk->sk_protocol = protocol; | |
1203 | return 0; | |
1204 | } | |
1205 | ||
3f378b68 EP |
1206 | static int netlink_create(struct net *net, struct socket *sock, int protocol, |
1207 | int kern) | |
ab33a171 PM |
1208 | { |
1209 | struct module *module = NULL; | |
af65bdfc | 1210 | struct mutex *cb_mutex; |
f7fa9b10 | 1211 | struct netlink_sock *nlk; |
023e2cfa JB |
1212 | int (*bind)(struct net *net, int group); |
1213 | void (*unbind)(struct net *net, int group); | |
ab33a171 | 1214 | int err = 0; |
1da177e4 LT |
1215 | |
1216 | sock->state = SS_UNCONNECTED; | |
1217 | ||
1218 | if (sock->type != SOCK_RAW && sock->type != SOCK_DGRAM) | |
1219 | return -ESOCKTNOSUPPORT; | |
1220 | ||
6ac552fd | 1221 | if (protocol < 0 || protocol >= MAX_LINKS) |
1da177e4 LT |
1222 | return -EPROTONOSUPPORT; |
1223 | ||
77247bbb | 1224 | netlink_lock_table(); |
95a5afca | 1225 | #ifdef CONFIG_MODULES |
ab33a171 | 1226 | if (!nl_table[protocol].registered) { |
77247bbb | 1227 | netlink_unlock_table(); |
4fdb3bb7 | 1228 | request_module("net-pf-%d-proto-%d", PF_NETLINK, protocol); |
77247bbb | 1229 | netlink_lock_table(); |
4fdb3bb7 | 1230 | } |
ab33a171 PM |
1231 | #endif |
1232 | if (nl_table[protocol].registered && | |
1233 | try_module_get(nl_table[protocol].module)) | |
1234 | module = nl_table[protocol].module; | |
974c37e9 AD |
1235 | else |
1236 | err = -EPROTONOSUPPORT; | |
af65bdfc | 1237 | cb_mutex = nl_table[protocol].cb_mutex; |
03292745 | 1238 | bind = nl_table[protocol].bind; |
4f520900 | 1239 | unbind = nl_table[protocol].unbind; |
77247bbb | 1240 | netlink_unlock_table(); |
4fdb3bb7 | 1241 | |
974c37e9 AD |
1242 | if (err < 0) |
1243 | goto out; | |
1244 | ||
11aa9c28 | 1245 | err = __netlink_create(net, sock, cb_mutex, protocol, kern); |
6ac552fd | 1246 | if (err < 0) |
f7fa9b10 PM |
1247 | goto out_module; |
1248 | ||
6f756a8c | 1249 | local_bh_disable(); |
c1fd3b94 | 1250 | sock_prot_inuse_add(net, &netlink_proto, 1); |
6f756a8c DM |
1251 | local_bh_enable(); |
1252 | ||
f7fa9b10 | 1253 | nlk = nlk_sk(sock->sk); |
f7fa9b10 | 1254 | nlk->module = module; |
03292745 | 1255 | nlk->netlink_bind = bind; |
4f520900 | 1256 | nlk->netlink_unbind = unbind; |
ab33a171 PM |
1257 | out: |
1258 | return err; | |
1da177e4 | 1259 | |
ab33a171 PM |
1260 | out_module: |
1261 | module_put(module); | |
1262 | goto out; | |
1da177e4 LT |
1263 | } |
1264 | ||
21e4902a TG |
1265 | static void deferred_put_nlk_sk(struct rcu_head *head) |
1266 | { | |
1267 | struct netlink_sock *nlk = container_of(head, struct netlink_sock, rcu); | |
1268 | ||
1269 | sock_put(&nlk->sk); | |
1270 | } | |
1271 | ||
1da177e4 LT |
1272 | static int netlink_release(struct socket *sock) |
1273 | { | |
1274 | struct sock *sk = sock->sk; | |
1275 | struct netlink_sock *nlk; | |
1276 | ||
1277 | if (!sk) | |
1278 | return 0; | |
1279 | ||
1280 | netlink_remove(sk); | |
ac57b3a9 | 1281 | sock_orphan(sk); |
1da177e4 LT |
1282 | nlk = nlk_sk(sk); |
1283 | ||
3f660d66 HX |
1284 | /* |
1285 | * OK. Socket is unlinked, any packets that arrive now | |
1286 | * will be purged. | |
1287 | */ | |
1da177e4 | 1288 | |
ee1c2442 JB |
1289 | /* must not acquire netlink_table_lock in any way again before unbind |
1290 | * and notifying genetlink is done as otherwise it might deadlock | |
1291 | */ | |
1292 | if (nlk->netlink_unbind) { | |
1293 | int i; | |
1294 | ||
1295 | for (i = 0; i < nlk->ngroups; i++) | |
1296 | if (test_bit(i, nlk->groups)) | |
1297 | nlk->netlink_unbind(sock_net(sk), i + 1); | |
1298 | } | |
1299 | if (sk->sk_protocol == NETLINK_GENERIC && | |
1300 | atomic_dec_return(&genl_sk_destructing_cnt) == 0) | |
1301 | wake_up(&genl_sk_destructing_waitq); | |
1302 | ||
1da177e4 LT |
1303 | sock->sk = NULL; |
1304 | wake_up_interruptible_all(&nlk->wait); | |
1305 | ||
1306 | skb_queue_purge(&sk->sk_write_queue); | |
1307 | ||
15e47304 | 1308 | if (nlk->portid) { |
1da177e4 | 1309 | struct netlink_notify n = { |
3b1e0a65 | 1310 | .net = sock_net(sk), |
1da177e4 | 1311 | .protocol = sk->sk_protocol, |
15e47304 | 1312 | .portid = nlk->portid, |
1da177e4 | 1313 | }; |
e041c683 AS |
1314 | atomic_notifier_call_chain(&netlink_chain, |
1315 | NETLINK_URELEASE, &n); | |
746fac4d | 1316 | } |
4fdb3bb7 | 1317 | |
5e7c001c | 1318 | module_put(nlk->module); |
4fdb3bb7 | 1319 | |
aed81560 | 1320 | if (netlink_is_kernel(sk)) { |
b10dcb3b | 1321 | netlink_table_grab(); |
869e58f8 DL |
1322 | BUG_ON(nl_table[sk->sk_protocol].registered == 0); |
1323 | if (--nl_table[sk->sk_protocol].registered == 0) { | |
6d772ac5 ED |
1324 | struct listeners *old; |
1325 | ||
1326 | old = nl_deref_protected(nl_table[sk->sk_protocol].listeners); | |
1327 | RCU_INIT_POINTER(nl_table[sk->sk_protocol].listeners, NULL); | |
1328 | kfree_rcu(old, rcu); | |
869e58f8 | 1329 | nl_table[sk->sk_protocol].module = NULL; |
9785e10a | 1330 | nl_table[sk->sk_protocol].bind = NULL; |
4f520900 | 1331 | nl_table[sk->sk_protocol].unbind = NULL; |
9785e10a | 1332 | nl_table[sk->sk_protocol].flags = 0; |
869e58f8 DL |
1333 | nl_table[sk->sk_protocol].registered = 0; |
1334 | } | |
b10dcb3b | 1335 | netlink_table_ungrab(); |
658cb354 | 1336 | } |
77247bbb | 1337 | |
f7fa9b10 PM |
1338 | kfree(nlk->groups); |
1339 | nlk->groups = NULL; | |
1340 | ||
3755810c | 1341 | local_bh_disable(); |
c1fd3b94 | 1342 | sock_prot_inuse_add(sock_net(sk), &netlink_proto, -1); |
3755810c | 1343 | local_bh_enable(); |
21e4902a | 1344 | call_rcu(&nlk->rcu, deferred_put_nlk_sk); |
1da177e4 LT |
1345 | return 0; |
1346 | } | |
1347 | ||
1348 | static int netlink_autobind(struct socket *sock) | |
1349 | { | |
1350 | struct sock *sk = sock->sk; | |
3b1e0a65 | 1351 | struct net *net = sock_net(sk); |
da12c90e | 1352 | struct netlink_table *table = &nl_table[sk->sk_protocol]; |
15e47304 | 1353 | s32 portid = task_tgid_vnr(current); |
1da177e4 | 1354 | int err; |
b9fbe709 HX |
1355 | s32 rover = -4096; |
1356 | bool ok; | |
1da177e4 LT |
1357 | |
1358 | retry: | |
1359 | cond_resched(); | |
e341694e | 1360 | rcu_read_lock(); |
b9fbe709 HX |
1361 | ok = !__netlink_lookup(table, portid, net); |
1362 | rcu_read_unlock(); | |
1363 | if (!ok) { | |
e341694e | 1364 | /* Bind collision, search negative portid values. */ |
b9fbe709 HX |
1365 | if (rover == -4096) |
1366 | /* rover will be in range [S32_MIN, -4097] */ | |
1367 | rover = S32_MIN + prandom_u32_max(-4096 - S32_MIN); | |
1368 | else if (rover >= -4096) | |
e341694e | 1369 | rover = -4097; |
b9fbe709 | 1370 | portid = rover--; |
e341694e | 1371 | goto retry; |
1da177e4 | 1372 | } |
1da177e4 | 1373 | |
8ea65f4a | 1374 | err = netlink_insert(sk, portid); |
1da177e4 LT |
1375 | if (err == -EADDRINUSE) |
1376 | goto retry; | |
d470e3b4 DM |
1377 | |
1378 | /* If 2 threads race to autobind, that is fine. */ | |
1379 | if (err == -EBUSY) | |
1380 | err = 0; | |
1381 | ||
1382 | return err; | |
1da177e4 LT |
1383 | } |
1384 | ||
aa4cf945 EB |
1385 | /** |
1386 | * __netlink_ns_capable - General netlink message capability test | |
1387 | * @nsp: NETLINK_CB of the socket buffer holding a netlink command from userspace. | |
1388 | * @user_ns: The user namespace of the capability to use | |
1389 | * @cap: The capability to use | |
1390 | * | |
1391 | * Test to see if the opener of the socket we received the message | |
1392 | * from had when the netlink socket was created and the sender of the | |
1393 | * message has has the capability @cap in the user namespace @user_ns. | |
1394 | */ | |
1395 | bool __netlink_ns_capable(const struct netlink_skb_parms *nsp, | |
1396 | struct user_namespace *user_ns, int cap) | |
1397 | { | |
2d7a85f4 EB |
1398 | return ((nsp->flags & NETLINK_SKB_DST) || |
1399 | file_ns_capable(nsp->sk->sk_socket->file, user_ns, cap)) && | |
1400 | ns_capable(user_ns, cap); | |
aa4cf945 EB |
1401 | } |
1402 | EXPORT_SYMBOL(__netlink_ns_capable); | |
1403 | ||
1404 | /** | |
1405 | * netlink_ns_capable - General netlink message capability test | |
1406 | * @skb: socket buffer holding a netlink command from userspace | |
1407 | * @user_ns: The user namespace of the capability to use | |
1408 | * @cap: The capability to use | |
1409 | * | |
1410 | * Test to see if the opener of the socket we received the message | |
1411 | * from had when the netlink socket was created and the sender of the | |
1412 | * message has has the capability @cap in the user namespace @user_ns. | |
1413 | */ | |
1414 | bool netlink_ns_capable(const struct sk_buff *skb, | |
1415 | struct user_namespace *user_ns, int cap) | |
1416 | { | |
1417 | return __netlink_ns_capable(&NETLINK_CB(skb), user_ns, cap); | |
1418 | } | |
1419 | EXPORT_SYMBOL(netlink_ns_capable); | |
1420 | ||
1421 | /** | |
1422 | * netlink_capable - Netlink global message capability test | |
1423 | * @skb: socket buffer holding a netlink command from userspace | |
1424 | * @cap: The capability to use | |
1425 | * | |
1426 | * Test to see if the opener of the socket we received the message | |
1427 | * from had when the netlink socket was created and the sender of the | |
1428 | * message has has the capability @cap in all user namespaces. | |
1429 | */ | |
1430 | bool netlink_capable(const struct sk_buff *skb, int cap) | |
1431 | { | |
1432 | return netlink_ns_capable(skb, &init_user_ns, cap); | |
1433 | } | |
1434 | EXPORT_SYMBOL(netlink_capable); | |
1435 | ||
1436 | /** | |
1437 | * netlink_net_capable - Netlink network namespace message capability test | |
1438 | * @skb: socket buffer holding a netlink command from userspace | |
1439 | * @cap: The capability to use | |
1440 | * | |
1441 | * Test to see if the opener of the socket we received the message | |
1442 | * from had when the netlink socket was created and the sender of the | |
1443 | * message has has the capability @cap over the network namespace of | |
1444 | * the socket we received the message from. | |
1445 | */ | |
1446 | bool netlink_net_capable(const struct sk_buff *skb, int cap) | |
1447 | { | |
1448 | return netlink_ns_capable(skb, sock_net(skb->sk)->user_ns, cap); | |
1449 | } | |
1450 | EXPORT_SYMBOL(netlink_net_capable); | |
1451 | ||
5187cd05 | 1452 | static inline int netlink_allowed(const struct socket *sock, unsigned int flag) |
746fac4d | 1453 | { |
9785e10a | 1454 | return (nl_table[sock->sk->sk_protocol].flags & flag) || |
df008c91 | 1455 | ns_capable(sock_net(sock->sk)->user_ns, CAP_NET_ADMIN); |
746fac4d | 1456 | } |
1da177e4 | 1457 | |
f7fa9b10 PM |
1458 | static void |
1459 | netlink_update_subscriptions(struct sock *sk, unsigned int subscriptions) | |
1460 | { | |
1461 | struct netlink_sock *nlk = nlk_sk(sk); | |
1462 | ||
1463 | if (nlk->subscriptions && !subscriptions) | |
1464 | __sk_del_bind_node(sk); | |
1465 | else if (!nlk->subscriptions && subscriptions) | |
1466 | sk_add_bind_node(sk, &nl_table[sk->sk_protocol].mc_list); | |
1467 | nlk->subscriptions = subscriptions; | |
1468 | } | |
1469 | ||
b4ff4f04 | 1470 | static int netlink_realloc_groups(struct sock *sk) |
513c2500 PM |
1471 | { |
1472 | struct netlink_sock *nlk = nlk_sk(sk); | |
1473 | unsigned int groups; | |
b4ff4f04 | 1474 | unsigned long *new_groups; |
513c2500 PM |
1475 | int err = 0; |
1476 | ||
b4ff4f04 JB |
1477 | netlink_table_grab(); |
1478 | ||
513c2500 | 1479 | groups = nl_table[sk->sk_protocol].groups; |
b4ff4f04 | 1480 | if (!nl_table[sk->sk_protocol].registered) { |
513c2500 | 1481 | err = -ENOENT; |
b4ff4f04 JB |
1482 | goto out_unlock; |
1483 | } | |
513c2500 | 1484 | |
b4ff4f04 JB |
1485 | if (nlk->ngroups >= groups) |
1486 | goto out_unlock; | |
513c2500 | 1487 | |
b4ff4f04 JB |
1488 | new_groups = krealloc(nlk->groups, NLGRPSZ(groups), GFP_ATOMIC); |
1489 | if (new_groups == NULL) { | |
1490 | err = -ENOMEM; | |
1491 | goto out_unlock; | |
1492 | } | |
6ac552fd | 1493 | memset((char *)new_groups + NLGRPSZ(nlk->ngroups), 0, |
b4ff4f04 JB |
1494 | NLGRPSZ(groups) - NLGRPSZ(nlk->ngroups)); |
1495 | ||
1496 | nlk->groups = new_groups; | |
513c2500 | 1497 | nlk->ngroups = groups; |
b4ff4f04 JB |
1498 | out_unlock: |
1499 | netlink_table_ungrab(); | |
1500 | return err; | |
513c2500 PM |
1501 | } |
1502 | ||
02c81ab9 | 1503 | static void netlink_undo_bind(int group, long unsigned int groups, |
023e2cfa | 1504 | struct sock *sk) |
4f520900 | 1505 | { |
023e2cfa | 1506 | struct netlink_sock *nlk = nlk_sk(sk); |
4f520900 RGB |
1507 | int undo; |
1508 | ||
1509 | if (!nlk->netlink_unbind) | |
1510 | return; | |
1511 | ||
1512 | for (undo = 0; undo < group; undo++) | |
6251edd9 | 1513 | if (test_bit(undo, &groups)) |
8b7c36d8 | 1514 | nlk->netlink_unbind(sock_net(sk), undo + 1); |
4f520900 RGB |
1515 | } |
1516 | ||
6ac552fd PM |
1517 | static int netlink_bind(struct socket *sock, struct sockaddr *addr, |
1518 | int addr_len) | |
1da177e4 LT |
1519 | { |
1520 | struct sock *sk = sock->sk; | |
3b1e0a65 | 1521 | struct net *net = sock_net(sk); |
1da177e4 LT |
1522 | struct netlink_sock *nlk = nlk_sk(sk); |
1523 | struct sockaddr_nl *nladdr = (struct sockaddr_nl *)addr; | |
1524 | int err; | |
4f520900 | 1525 | long unsigned int groups = nladdr->nl_groups; |
da314c99 | 1526 | bool bound; |
746fac4d | 1527 | |
4e4b5376 HFS |
1528 | if (addr_len < sizeof(struct sockaddr_nl)) |
1529 | return -EINVAL; | |
1530 | ||
1da177e4 LT |
1531 | if (nladdr->nl_family != AF_NETLINK) |
1532 | return -EINVAL; | |
1533 | ||
1534 | /* Only superuser is allowed to listen multicasts */ | |
4f520900 | 1535 | if (groups) { |
5187cd05 | 1536 | if (!netlink_allowed(sock, NL_CFG_F_NONROOT_RECV)) |
513c2500 | 1537 | return -EPERM; |
b4ff4f04 JB |
1538 | err = netlink_realloc_groups(sk); |
1539 | if (err) | |
1540 | return err; | |
513c2500 | 1541 | } |
1da177e4 | 1542 | |
da314c99 HX |
1543 | bound = nlk->bound; |
1544 | if (bound) { | |
1545 | /* Ensure nlk->portid is up-to-date. */ | |
1546 | smp_rmb(); | |
1547 | ||
15e47304 | 1548 | if (nladdr->nl_pid != nlk->portid) |
1da177e4 | 1549 | return -EINVAL; |
da314c99 | 1550 | } |
4f520900 RGB |
1551 | |
1552 | if (nlk->netlink_bind && groups) { | |
1553 | int group; | |
1554 | ||
1555 | for (group = 0; group < nlk->ngroups; group++) { | |
1556 | if (!test_bit(group, &groups)) | |
1557 | continue; | |
8b7c36d8 | 1558 | err = nlk->netlink_bind(net, group + 1); |
4f520900 RGB |
1559 | if (!err) |
1560 | continue; | |
023e2cfa | 1561 | netlink_undo_bind(group, groups, sk); |
4f520900 RGB |
1562 | return err; |
1563 | } | |
1564 | } | |
1565 | ||
da314c99 HX |
1566 | /* No need for barriers here as we return to user-space without |
1567 | * using any of the bound attributes. | |
1568 | */ | |
1569 | if (!bound) { | |
1da177e4 | 1570 | err = nladdr->nl_pid ? |
8ea65f4a | 1571 | netlink_insert(sk, nladdr->nl_pid) : |
1da177e4 | 1572 | netlink_autobind(sock); |
4f520900 | 1573 | if (err) { |
023e2cfa | 1574 | netlink_undo_bind(nlk->ngroups, groups, sk); |
1da177e4 | 1575 | return err; |
4f520900 | 1576 | } |
1da177e4 LT |
1577 | } |
1578 | ||
4f520900 | 1579 | if (!groups && (nlk->groups == NULL || !(u32)nlk->groups[0])) |
1da177e4 LT |
1580 | return 0; |
1581 | ||
1582 | netlink_table_grab(); | |
f7fa9b10 | 1583 | netlink_update_subscriptions(sk, nlk->subscriptions + |
4f520900 | 1584 | hweight32(groups) - |
746fac4d | 1585 | hweight32(nlk->groups[0])); |
4f520900 | 1586 | nlk->groups[0] = (nlk->groups[0] & ~0xffffffffUL) | groups; |
4277a083 | 1587 | netlink_update_listeners(sk); |
1da177e4 LT |
1588 | netlink_table_ungrab(); |
1589 | ||
1590 | return 0; | |
1591 | } | |
1592 | ||
1593 | static int netlink_connect(struct socket *sock, struct sockaddr *addr, | |
1594 | int alen, int flags) | |
1595 | { | |
1596 | int err = 0; | |
1597 | struct sock *sk = sock->sk; | |
1598 | struct netlink_sock *nlk = nlk_sk(sk); | |
6ac552fd | 1599 | struct sockaddr_nl *nladdr = (struct sockaddr_nl *)addr; |
1da177e4 | 1600 | |
6503d961 CG |
1601 | if (alen < sizeof(addr->sa_family)) |
1602 | return -EINVAL; | |
1603 | ||
1da177e4 LT |
1604 | if (addr->sa_family == AF_UNSPEC) { |
1605 | sk->sk_state = NETLINK_UNCONNECTED; | |
15e47304 | 1606 | nlk->dst_portid = 0; |
d629b836 | 1607 | nlk->dst_group = 0; |
1da177e4 LT |
1608 | return 0; |
1609 | } | |
1610 | if (addr->sa_family != AF_NETLINK) | |
1611 | return -EINVAL; | |
1612 | ||
46833a86 | 1613 | if ((nladdr->nl_groups || nladdr->nl_pid) && |
5187cd05 | 1614 | !netlink_allowed(sock, NL_CFG_F_NONROOT_SEND)) |
1da177e4 LT |
1615 | return -EPERM; |
1616 | ||
da314c99 HX |
1617 | /* No need for barriers here as we return to user-space without |
1618 | * using any of the bound attributes. | |
1619 | */ | |
1620 | if (!nlk->bound) | |
1da177e4 LT |
1621 | err = netlink_autobind(sock); |
1622 | ||
1623 | if (err == 0) { | |
1624 | sk->sk_state = NETLINK_CONNECTED; | |
15e47304 | 1625 | nlk->dst_portid = nladdr->nl_pid; |
d629b836 | 1626 | nlk->dst_group = ffs(nladdr->nl_groups); |
1da177e4 LT |
1627 | } |
1628 | ||
1629 | return err; | |
1630 | } | |
1631 | ||
6ac552fd PM |
1632 | static int netlink_getname(struct socket *sock, struct sockaddr *addr, |
1633 | int *addr_len, int peer) | |
1da177e4 LT |
1634 | { |
1635 | struct sock *sk = sock->sk; | |
1636 | struct netlink_sock *nlk = nlk_sk(sk); | |
13cfa97b | 1637 | DECLARE_SOCKADDR(struct sockaddr_nl *, nladdr, addr); |
746fac4d | 1638 | |
1da177e4 LT |
1639 | nladdr->nl_family = AF_NETLINK; |
1640 | nladdr->nl_pad = 0; | |
1641 | *addr_len = sizeof(*nladdr); | |
1642 | ||
1643 | if (peer) { | |
15e47304 | 1644 | nladdr->nl_pid = nlk->dst_portid; |
d629b836 | 1645 | nladdr->nl_groups = netlink_group_mask(nlk->dst_group); |
1da177e4 | 1646 | } else { |
15e47304 | 1647 | nladdr->nl_pid = nlk->portid; |
513c2500 | 1648 | nladdr->nl_groups = nlk->groups ? nlk->groups[0] : 0; |
1da177e4 LT |
1649 | } |
1650 | return 0; | |
1651 | } | |
1652 | ||
15e47304 | 1653 | static struct sock *netlink_getsockbyportid(struct sock *ssk, u32 portid) |
1da177e4 | 1654 | { |
1da177e4 LT |
1655 | struct sock *sock; |
1656 | struct netlink_sock *nlk; | |
1657 | ||
15e47304 | 1658 | sock = netlink_lookup(sock_net(ssk), ssk->sk_protocol, portid); |
1da177e4 LT |
1659 | if (!sock) |
1660 | return ERR_PTR(-ECONNREFUSED); | |
1661 | ||
1662 | /* Don't bother queuing skb if kernel socket has no input function */ | |
1663 | nlk = nlk_sk(sock); | |
cd40b7d3 | 1664 | if (sock->sk_state == NETLINK_CONNECTED && |
15e47304 | 1665 | nlk->dst_portid != nlk_sk(ssk)->portid) { |
1da177e4 LT |
1666 | sock_put(sock); |
1667 | return ERR_PTR(-ECONNREFUSED); | |
1668 | } | |
1669 | return sock; | |
1670 | } | |
1671 | ||
1672 | struct sock *netlink_getsockbyfilp(struct file *filp) | |
1673 | { | |
496ad9aa | 1674 | struct inode *inode = file_inode(filp); |
1da177e4 LT |
1675 | struct sock *sock; |
1676 | ||
1677 | if (!S_ISSOCK(inode->i_mode)) | |
1678 | return ERR_PTR(-ENOTSOCK); | |
1679 | ||
1680 | sock = SOCKET_I(inode)->sk; | |
1681 | if (sock->sk_family != AF_NETLINK) | |
1682 | return ERR_PTR(-EINVAL); | |
1683 | ||
1684 | sock_hold(sock); | |
1685 | return sock; | |
1686 | } | |
1687 | ||
3a36515f PN |
1688 | static struct sk_buff *netlink_alloc_large_skb(unsigned int size, |
1689 | int broadcast) | |
c05cdb1b PNA |
1690 | { |
1691 | struct sk_buff *skb; | |
1692 | void *data; | |
1693 | ||
3a36515f | 1694 | if (size <= NLMSG_GOODSIZE || broadcast) |
c05cdb1b PNA |
1695 | return alloc_skb(size, GFP_KERNEL); |
1696 | ||
3a36515f PN |
1697 | size = SKB_DATA_ALIGN(size) + |
1698 | SKB_DATA_ALIGN(sizeof(struct skb_shared_info)); | |
c05cdb1b PNA |
1699 | |
1700 | data = vmalloc(size); | |
1701 | if (data == NULL) | |
3a36515f | 1702 | return NULL; |
c05cdb1b | 1703 | |
2ea2f62c | 1704 | skb = __build_skb(data, size); |
3a36515f PN |
1705 | if (skb == NULL) |
1706 | vfree(data); | |
2ea2f62c | 1707 | else |
3a36515f | 1708 | skb->destructor = netlink_skb_destructor; |
c05cdb1b PNA |
1709 | |
1710 | return skb; | |
c05cdb1b PNA |
1711 | } |
1712 | ||
1da177e4 LT |
1713 | /* |
1714 | * Attach a skb to a netlink socket. | |
1715 | * The caller must hold a reference to the destination socket. On error, the | |
1716 | * reference is dropped. The skb is not send to the destination, just all | |
1717 | * all error checks are performed and memory in the queue is reserved. | |
1718 | * Return values: | |
1719 | * < 0: error. skb freed, reference to sock dropped. | |
1720 | * 0: continue | |
1721 | * 1: repeat lookup - reference dropped while waiting for socket memory. | |
1722 | */ | |
9457afee | 1723 | int netlink_attachskb(struct sock *sk, struct sk_buff *skb, |
c3d8d1e3 | 1724 | long *timeo, struct sock *ssk) |
1da177e4 LT |
1725 | { |
1726 | struct netlink_sock *nlk; | |
1727 | ||
1728 | nlk = nlk_sk(sk); | |
1729 | ||
5fd96123 | 1730 | if ((atomic_read(&sk->sk_rmem_alloc) > sk->sk_rcvbuf || |
cc3a572f | 1731 | test_bit(NETLINK_S_CONGESTED, &nlk->state)) && |
5fd96123 | 1732 | !netlink_skb_is_mmaped(skb)) { |
1da177e4 | 1733 | DECLARE_WAITQUEUE(wait, current); |
c3d8d1e3 | 1734 | if (!*timeo) { |
aed81560 | 1735 | if (!ssk || netlink_is_kernel(ssk)) |
1da177e4 LT |
1736 | netlink_overrun(sk); |
1737 | sock_put(sk); | |
1738 | kfree_skb(skb); | |
1739 | return -EAGAIN; | |
1740 | } | |
1741 | ||
1742 | __set_current_state(TASK_INTERRUPTIBLE); | |
1743 | add_wait_queue(&nlk->wait, &wait); | |
1744 | ||
1745 | if ((atomic_read(&sk->sk_rmem_alloc) > sk->sk_rcvbuf || | |
cc3a572f | 1746 | test_bit(NETLINK_S_CONGESTED, &nlk->state)) && |
1da177e4 | 1747 | !sock_flag(sk, SOCK_DEAD)) |
c3d8d1e3 | 1748 | *timeo = schedule_timeout(*timeo); |
1da177e4 LT |
1749 | |
1750 | __set_current_state(TASK_RUNNING); | |
1751 | remove_wait_queue(&nlk->wait, &wait); | |
1752 | sock_put(sk); | |
1753 | ||
1754 | if (signal_pending(current)) { | |
1755 | kfree_skb(skb); | |
c3d8d1e3 | 1756 | return sock_intr_errno(*timeo); |
1da177e4 LT |
1757 | } |
1758 | return 1; | |
1759 | } | |
cf0a018a | 1760 | netlink_skb_set_owner_r(skb, sk); |
1da177e4 LT |
1761 | return 0; |
1762 | } | |
1763 | ||
4a7e7c2a | 1764 | static int __netlink_sendskb(struct sock *sk, struct sk_buff *skb) |
1da177e4 | 1765 | { |
1da177e4 LT |
1766 | int len = skb->len; |
1767 | ||
bcbde0d4 DB |
1768 | netlink_deliver_tap(skb); |
1769 | ||
f9c22888 PM |
1770 | #ifdef CONFIG_NETLINK_MMAP |
1771 | if (netlink_skb_is_mmaped(skb)) | |
1772 | netlink_queue_mmaped_skb(sk, skb); | |
1773 | else if (netlink_rx_is_mmaped(sk)) | |
1774 | netlink_ring_set_copied(sk, skb); | |
1775 | else | |
1776 | #endif /* CONFIG_NETLINK_MMAP */ | |
1777 | skb_queue_tail(&sk->sk_receive_queue, skb); | |
676d2369 | 1778 | sk->sk_data_ready(sk); |
4a7e7c2a ED |
1779 | return len; |
1780 | } | |
1781 | ||
1782 | int netlink_sendskb(struct sock *sk, struct sk_buff *skb) | |
1783 | { | |
1784 | int len = __netlink_sendskb(sk, skb); | |
1785 | ||
1da177e4 LT |
1786 | sock_put(sk); |
1787 | return len; | |
1788 | } | |
1789 | ||
1790 | void netlink_detachskb(struct sock *sk, struct sk_buff *skb) | |
1791 | { | |
1792 | kfree_skb(skb); | |
1793 | sock_put(sk); | |
1794 | } | |
1795 | ||
b57ef81f | 1796 | static struct sk_buff *netlink_trim(struct sk_buff *skb, gfp_t allocation) |
1da177e4 LT |
1797 | { |
1798 | int delta; | |
1799 | ||
1298ca46 | 1800 | WARN_ON(skb->sk != NULL); |
5fd96123 PM |
1801 | if (netlink_skb_is_mmaped(skb)) |
1802 | return skb; | |
1da177e4 | 1803 | |
4305b541 | 1804 | delta = skb->end - skb->tail; |
c05cdb1b | 1805 | if (is_vmalloc_addr(skb->head) || delta * 2 < skb->truesize) |
1da177e4 LT |
1806 | return skb; |
1807 | ||
1808 | if (skb_shared(skb)) { | |
1809 | struct sk_buff *nskb = skb_clone(skb, allocation); | |
1810 | if (!nskb) | |
1811 | return skb; | |
8460c00f | 1812 | consume_skb(skb); |
1da177e4 LT |
1813 | skb = nskb; |
1814 | } | |
1815 | ||
1816 | if (!pskb_expand_head(skb, 0, -delta, allocation)) | |
1817 | skb->truesize -= delta; | |
1818 | ||
1819 | return skb; | |
1820 | } | |
1821 | ||
3fbc2905 EB |
1822 | static int netlink_unicast_kernel(struct sock *sk, struct sk_buff *skb, |
1823 | struct sock *ssk) | |
cd40b7d3 DL |
1824 | { |
1825 | int ret; | |
1826 | struct netlink_sock *nlk = nlk_sk(sk); | |
1827 | ||
1828 | ret = -ECONNREFUSED; | |
1829 | if (nlk->netlink_rcv != NULL) { | |
1830 | ret = skb->len; | |
cf0a018a | 1831 | netlink_skb_set_owner_r(skb, sk); |
e32123e5 | 1832 | NETLINK_CB(skb).sk = ssk; |
73bfd370 | 1833 | netlink_deliver_tap_kernel(sk, ssk, skb); |
cd40b7d3 | 1834 | nlk->netlink_rcv(skb); |
bfb253c9 ED |
1835 | consume_skb(skb); |
1836 | } else { | |
1837 | kfree_skb(skb); | |
cd40b7d3 | 1838 | } |
cd40b7d3 DL |
1839 | sock_put(sk); |
1840 | return ret; | |
1841 | } | |
1842 | ||
1843 | int netlink_unicast(struct sock *ssk, struct sk_buff *skb, | |
15e47304 | 1844 | u32 portid, int nonblock) |
1da177e4 LT |
1845 | { |
1846 | struct sock *sk; | |
1847 | int err; | |
1848 | long timeo; | |
1849 | ||
1850 | skb = netlink_trim(skb, gfp_any()); | |
1851 | ||
1852 | timeo = sock_sndtimeo(ssk, nonblock); | |
1853 | retry: | |
15e47304 | 1854 | sk = netlink_getsockbyportid(ssk, portid); |
1da177e4 LT |
1855 | if (IS_ERR(sk)) { |
1856 | kfree_skb(skb); | |
1857 | return PTR_ERR(sk); | |
1858 | } | |
cd40b7d3 | 1859 | if (netlink_is_kernel(sk)) |
3fbc2905 | 1860 | return netlink_unicast_kernel(sk, skb, ssk); |
cd40b7d3 | 1861 | |
b1153f29 | 1862 | if (sk_filter(sk, skb)) { |
84874607 | 1863 | err = skb->len; |
b1153f29 SH |
1864 | kfree_skb(skb); |
1865 | sock_put(sk); | |
1866 | return err; | |
1867 | } | |
1868 | ||
9457afee | 1869 | err = netlink_attachskb(sk, skb, &timeo, ssk); |
1da177e4 LT |
1870 | if (err == 1) |
1871 | goto retry; | |
1872 | if (err) | |
1873 | return err; | |
1874 | ||
7ee015e0 | 1875 | return netlink_sendskb(sk, skb); |
1da177e4 | 1876 | } |
6ac552fd | 1877 | EXPORT_SYMBOL(netlink_unicast); |
1da177e4 | 1878 | |
6bb0fef4 DB |
1879 | struct sk_buff *__netlink_alloc_skb(struct sock *ssk, unsigned int size, |
1880 | unsigned int ldiff, u32 dst_portid, | |
1881 | gfp_t gfp_mask) | |
f9c22888 PM |
1882 | { |
1883 | #ifdef CONFIG_NETLINK_MMAP | |
6bb0fef4 | 1884 | unsigned int maxlen, linear_size; |
f9c22888 PM |
1885 | struct sock *sk = NULL; |
1886 | struct sk_buff *skb; | |
1887 | struct netlink_ring *ring; | |
1888 | struct nl_mmap_hdr *hdr; | |
f9c22888 PM |
1889 | |
1890 | sk = netlink_getsockbyportid(ssk, dst_portid); | |
1891 | if (IS_ERR(sk)) | |
1892 | goto out; | |
1893 | ||
1894 | ring = &nlk_sk(sk)->rx_ring; | |
1895 | /* fast-path without atomic ops for common case: non-mmaped receiver */ | |
1896 | if (ring->pg_vec == NULL) | |
1897 | goto out_put; | |
1898 | ||
6bb0fef4 DB |
1899 | /* We need to account the full linear size needed as a ring |
1900 | * slot cannot have non-linear parts. | |
1901 | */ | |
1902 | linear_size = size + ldiff; | |
1903 | if (ring->frame_size - NL_MMAP_HDRLEN < linear_size) | |
aae9f0e2 TG |
1904 | goto out_put; |
1905 | ||
f9c22888 PM |
1906 | skb = alloc_skb_head(gfp_mask); |
1907 | if (skb == NULL) | |
1908 | goto err1; | |
1909 | ||
1910 | spin_lock_bh(&sk->sk_receive_queue.lock); | |
1911 | /* check again under lock */ | |
1912 | if (ring->pg_vec == NULL) | |
1913 | goto out_free; | |
1914 | ||
aae9f0e2 | 1915 | /* check again under lock */ |
f9c22888 | 1916 | maxlen = ring->frame_size - NL_MMAP_HDRLEN; |
6bb0fef4 | 1917 | if (maxlen < linear_size) |
f9c22888 PM |
1918 | goto out_free; |
1919 | ||
1920 | netlink_forward_ring(ring); | |
1921 | hdr = netlink_current_frame(ring, NL_MMAP_STATUS_UNUSED); | |
1922 | if (hdr == NULL) | |
1923 | goto err2; | |
6bb0fef4 | 1924 | |
f9c22888 PM |
1925 | netlink_ring_setup_skb(skb, sk, ring, hdr); |
1926 | netlink_set_status(hdr, NL_MMAP_STATUS_RESERVED); | |
1927 | atomic_inc(&ring->pending); | |
1928 | netlink_increment_head(ring); | |
1929 | ||
1930 | spin_unlock_bh(&sk->sk_receive_queue.lock); | |
1931 | return skb; | |
1932 | ||
1933 | err2: | |
1934 | kfree_skb(skb); | |
1935 | spin_unlock_bh(&sk->sk_receive_queue.lock); | |
cd1df525 | 1936 | netlink_overrun(sk); |
f9c22888 PM |
1937 | err1: |
1938 | sock_put(sk); | |
1939 | return NULL; | |
1940 | ||
1941 | out_free: | |
1942 | kfree_skb(skb); | |
1943 | spin_unlock_bh(&sk->sk_receive_queue.lock); | |
1944 | out_put: | |
1945 | sock_put(sk); | |
1946 | out: | |
1947 | #endif | |
1948 | return alloc_skb(size, gfp_mask); | |
1949 | } | |
6bb0fef4 | 1950 | EXPORT_SYMBOL_GPL(__netlink_alloc_skb); |
f9c22888 | 1951 | |
4277a083 PM |
1952 | int netlink_has_listeners(struct sock *sk, unsigned int group) |
1953 | { | |
1954 | int res = 0; | |
5c398dc8 | 1955 | struct listeners *listeners; |
4277a083 | 1956 | |
aed81560 | 1957 | BUG_ON(!netlink_is_kernel(sk)); |
b4ff4f04 JB |
1958 | |
1959 | rcu_read_lock(); | |
1960 | listeners = rcu_dereference(nl_table[sk->sk_protocol].listeners); | |
1961 | ||
6d772ac5 | 1962 | if (listeners && group - 1 < nl_table[sk->sk_protocol].groups) |
5c398dc8 | 1963 | res = test_bit(group - 1, listeners->masks); |
b4ff4f04 JB |
1964 | |
1965 | rcu_read_unlock(); | |
1966 | ||
4277a083 PM |
1967 | return res; |
1968 | } | |
1969 | EXPORT_SYMBOL_GPL(netlink_has_listeners); | |
1970 | ||
b57ef81f | 1971 | static int netlink_broadcast_deliver(struct sock *sk, struct sk_buff *skb) |
1da177e4 LT |
1972 | { |
1973 | struct netlink_sock *nlk = nlk_sk(sk); | |
1974 | ||
1975 | if (atomic_read(&sk->sk_rmem_alloc) <= sk->sk_rcvbuf && | |
cc3a572f | 1976 | !test_bit(NETLINK_S_CONGESTED, &nlk->state)) { |
cf0a018a | 1977 | netlink_skb_set_owner_r(skb, sk); |
4a7e7c2a | 1978 | __netlink_sendskb(sk, skb); |
2c645800 | 1979 | return atomic_read(&sk->sk_rmem_alloc) > (sk->sk_rcvbuf >> 1); |
1da177e4 LT |
1980 | } |
1981 | return -1; | |
1982 | } | |
1983 | ||
1984 | struct netlink_broadcast_data { | |
1985 | struct sock *exclude_sk; | |
b4b51029 | 1986 | struct net *net; |
15e47304 | 1987 | u32 portid; |
1da177e4 LT |
1988 | u32 group; |
1989 | int failure; | |
ff491a73 | 1990 | int delivery_failure; |
1da177e4 LT |
1991 | int congested; |
1992 | int delivered; | |
7d877f3b | 1993 | gfp_t allocation; |
1da177e4 | 1994 | struct sk_buff *skb, *skb2; |
910a7e90 EB |
1995 | int (*tx_filter)(struct sock *dsk, struct sk_buff *skb, void *data); |
1996 | void *tx_data; | |
1da177e4 LT |
1997 | }; |
1998 | ||
46c9521f RR |
1999 | static void do_one_broadcast(struct sock *sk, |
2000 | struct netlink_broadcast_data *p) | |
1da177e4 LT |
2001 | { |
2002 | struct netlink_sock *nlk = nlk_sk(sk); | |
2003 | int val; | |
2004 | ||
2005 | if (p->exclude_sk == sk) | |
46c9521f | 2006 | return; |
1da177e4 | 2007 | |
15e47304 | 2008 | if (nlk->portid == p->portid || p->group - 1 >= nlk->ngroups || |
f7fa9b10 | 2009 | !test_bit(p->group - 1, nlk->groups)) |
46c9521f | 2010 | return; |
1da177e4 | 2011 | |
59324cf3 ND |
2012 | if (!net_eq(sock_net(sk), p->net)) { |
2013 | if (!(nlk->flags & NETLINK_F_LISTEN_ALL_NSID)) | |
2014 | return; | |
2015 | ||
2016 | if (!peernet_has_id(sock_net(sk), p->net)) | |
2017 | return; | |
2018 | ||
2019 | if (!file_ns_capable(sk->sk_socket->file, p->net->user_ns, | |
2020 | CAP_NET_BROADCAST)) | |
2021 | return; | |
2022 | } | |
b4b51029 | 2023 | |
1da177e4 LT |
2024 | if (p->failure) { |
2025 | netlink_overrun(sk); | |
46c9521f | 2026 | return; |
1da177e4 LT |
2027 | } |
2028 | ||
2029 | sock_hold(sk); | |
2030 | if (p->skb2 == NULL) { | |
68acc024 | 2031 | if (skb_shared(p->skb)) { |
1da177e4 LT |
2032 | p->skb2 = skb_clone(p->skb, p->allocation); |
2033 | } else { | |
68acc024 TC |
2034 | p->skb2 = skb_get(p->skb); |
2035 | /* | |
2036 | * skb ownership may have been set when | |
2037 | * delivered to a previous socket. | |
2038 | */ | |
2039 | skb_orphan(p->skb2); | |
1da177e4 LT |
2040 | } |
2041 | } | |
2042 | if (p->skb2 == NULL) { | |
2043 | netlink_overrun(sk); | |
2044 | /* Clone failed. Notify ALL listeners. */ | |
2045 | p->failure = 1; | |
cc3a572f | 2046 | if (nlk->flags & NETLINK_F_BROADCAST_SEND_ERROR) |
be0c22a4 | 2047 | p->delivery_failure = 1; |
59324cf3 ND |
2048 | goto out; |
2049 | } | |
2050 | if (p->tx_filter && p->tx_filter(sk, p->skb2, p->tx_data)) { | |
910a7e90 EB |
2051 | kfree_skb(p->skb2); |
2052 | p->skb2 = NULL; | |
59324cf3 ND |
2053 | goto out; |
2054 | } | |
2055 | if (sk_filter(sk, p->skb2)) { | |
b1153f29 SH |
2056 | kfree_skb(p->skb2); |
2057 | p->skb2 = NULL; | |
59324cf3 ND |
2058 | goto out; |
2059 | } | |
2060 | NETLINK_CB(p->skb2).nsid = peernet2id(sock_net(sk), p->net); | |
2061 | NETLINK_CB(p->skb2).nsid_is_set = true; | |
2062 | val = netlink_broadcast_deliver(sk, p->skb2); | |
2063 | if (val < 0) { | |
1da177e4 | 2064 | netlink_overrun(sk); |
cc3a572f | 2065 | if (nlk->flags & NETLINK_F_BROADCAST_SEND_ERROR) |
be0c22a4 | 2066 | p->delivery_failure = 1; |
1da177e4 LT |
2067 | } else { |
2068 | p->congested |= val; | |
2069 | p->delivered = 1; | |
2070 | p->skb2 = NULL; | |
2071 | } | |
59324cf3 | 2072 | out: |
1da177e4 | 2073 | sock_put(sk); |
1da177e4 LT |
2074 | } |
2075 | ||
15e47304 | 2076 | int netlink_broadcast_filtered(struct sock *ssk, struct sk_buff *skb, u32 portid, |
910a7e90 EB |
2077 | u32 group, gfp_t allocation, |
2078 | int (*filter)(struct sock *dsk, struct sk_buff *skb, void *data), | |
2079 | void *filter_data) | |
1da177e4 | 2080 | { |
3b1e0a65 | 2081 | struct net *net = sock_net(ssk); |
1da177e4 | 2082 | struct netlink_broadcast_data info; |
1da177e4 LT |
2083 | struct sock *sk; |
2084 | ||
2085 | skb = netlink_trim(skb, allocation); | |
2086 | ||
2087 | info.exclude_sk = ssk; | |
b4b51029 | 2088 | info.net = net; |
15e47304 | 2089 | info.portid = portid; |
1da177e4 LT |
2090 | info.group = group; |
2091 | info.failure = 0; | |
ff491a73 | 2092 | info.delivery_failure = 0; |
1da177e4 LT |
2093 | info.congested = 0; |
2094 | info.delivered = 0; | |
2095 | info.allocation = allocation; | |
2096 | info.skb = skb; | |
2097 | info.skb2 = NULL; | |
910a7e90 EB |
2098 | info.tx_filter = filter; |
2099 | info.tx_data = filter_data; | |
1da177e4 LT |
2100 | |
2101 | /* While we sleep in clone, do not allow to change socket list */ | |
2102 | ||
2103 | netlink_lock_table(); | |
2104 | ||
b67bfe0d | 2105 | sk_for_each_bound(sk, &nl_table[ssk->sk_protocol].mc_list) |
1da177e4 LT |
2106 | do_one_broadcast(sk, &info); |
2107 | ||
70d4bf6d | 2108 | consume_skb(skb); |
aa1c6a6f | 2109 | |
1da177e4 LT |
2110 | netlink_unlock_table(); |
2111 | ||
70d4bf6d NH |
2112 | if (info.delivery_failure) { |
2113 | kfree_skb(info.skb2); | |
ff491a73 | 2114 | return -ENOBUFS; |
658cb354 ED |
2115 | } |
2116 | consume_skb(info.skb2); | |
ff491a73 | 2117 | |
1da177e4 | 2118 | if (info.delivered) { |
d0164adc | 2119 | if (info.congested && gfpflags_allow_blocking(allocation)) |
1da177e4 LT |
2120 | yield(); |
2121 | return 0; | |
2122 | } | |
1da177e4 LT |
2123 | return -ESRCH; |
2124 | } | |
910a7e90 EB |
2125 | EXPORT_SYMBOL(netlink_broadcast_filtered); |
2126 | ||
15e47304 | 2127 | int netlink_broadcast(struct sock *ssk, struct sk_buff *skb, u32 portid, |
910a7e90 EB |
2128 | u32 group, gfp_t allocation) |
2129 | { | |
15e47304 | 2130 | return netlink_broadcast_filtered(ssk, skb, portid, group, allocation, |
910a7e90 EB |
2131 | NULL, NULL); |
2132 | } | |
6ac552fd | 2133 | EXPORT_SYMBOL(netlink_broadcast); |
1da177e4 LT |
2134 | |
2135 | struct netlink_set_err_data { | |
2136 | struct sock *exclude_sk; | |
15e47304 | 2137 | u32 portid; |
1da177e4 LT |
2138 | u32 group; |
2139 | int code; | |
2140 | }; | |
2141 | ||
b57ef81f | 2142 | static int do_one_set_err(struct sock *sk, struct netlink_set_err_data *p) |
1da177e4 LT |
2143 | { |
2144 | struct netlink_sock *nlk = nlk_sk(sk); | |
1a50307b | 2145 | int ret = 0; |
1da177e4 LT |
2146 | |
2147 | if (sk == p->exclude_sk) | |
2148 | goto out; | |
2149 | ||
09ad9bc7 | 2150 | if (!net_eq(sock_net(sk), sock_net(p->exclude_sk))) |
b4b51029 EB |
2151 | goto out; |
2152 | ||
15e47304 | 2153 | if (nlk->portid == p->portid || p->group - 1 >= nlk->ngroups || |
f7fa9b10 | 2154 | !test_bit(p->group - 1, nlk->groups)) |
1da177e4 LT |
2155 | goto out; |
2156 | ||
cc3a572f | 2157 | if (p->code == ENOBUFS && nlk->flags & NETLINK_F_RECV_NO_ENOBUFS) { |
1a50307b PNA |
2158 | ret = 1; |
2159 | goto out; | |
2160 | } | |
2161 | ||
1da177e4 LT |
2162 | sk->sk_err = p->code; |
2163 | sk->sk_error_report(sk); | |
2164 | out: | |
1a50307b | 2165 | return ret; |
1da177e4 LT |
2166 | } |
2167 | ||
4843b93c PNA |
2168 | /** |
2169 | * netlink_set_err - report error to broadcast listeners | |
2170 | * @ssk: the kernel netlink socket, as returned by netlink_kernel_create() | |
15e47304 | 2171 | * @portid: the PORTID of a process that we want to skip (if any) |
840e93f2 | 2172 | * @group: the broadcast group that will notice the error |
4843b93c | 2173 | * @code: error code, must be negative (as usual in kernelspace) |
1a50307b PNA |
2174 | * |
2175 | * This function returns the number of broadcast listeners that have set the | |
cc3a572f | 2176 | * NETLINK_NO_ENOBUFS socket option. |
4843b93c | 2177 | */ |
15e47304 | 2178 | int netlink_set_err(struct sock *ssk, u32 portid, u32 group, int code) |
1da177e4 LT |
2179 | { |
2180 | struct netlink_set_err_data info; | |
1da177e4 | 2181 | struct sock *sk; |
1a50307b | 2182 | int ret = 0; |
1da177e4 LT |
2183 | |
2184 | info.exclude_sk = ssk; | |
15e47304 | 2185 | info.portid = portid; |
1da177e4 | 2186 | info.group = group; |
4843b93c PNA |
2187 | /* sk->sk_err wants a positive error value */ |
2188 | info.code = -code; | |
1da177e4 LT |
2189 | |
2190 | read_lock(&nl_table_lock); | |
2191 | ||
b67bfe0d | 2192 | sk_for_each_bound(sk, &nl_table[ssk->sk_protocol].mc_list) |
1a50307b | 2193 | ret += do_one_set_err(sk, &info); |
1da177e4 LT |
2194 | |
2195 | read_unlock(&nl_table_lock); | |
1a50307b | 2196 | return ret; |
1da177e4 | 2197 | } |
dd5b6ce6 | 2198 | EXPORT_SYMBOL(netlink_set_err); |
1da177e4 | 2199 | |
84659eb5 JB |
2200 | /* must be called with netlink table grabbed */ |
2201 | static void netlink_update_socket_mc(struct netlink_sock *nlk, | |
2202 | unsigned int group, | |
2203 | int is_new) | |
2204 | { | |
2205 | int old, new = !!is_new, subscriptions; | |
2206 | ||
2207 | old = test_bit(group - 1, nlk->groups); | |
2208 | subscriptions = nlk->subscriptions - old + new; | |
2209 | if (new) | |
2210 | __set_bit(group - 1, nlk->groups); | |
2211 | else | |
2212 | __clear_bit(group - 1, nlk->groups); | |
2213 | netlink_update_subscriptions(&nlk->sk, subscriptions); | |
2214 | netlink_update_listeners(&nlk->sk); | |
2215 | } | |
2216 | ||
9a4595bc | 2217 | static int netlink_setsockopt(struct socket *sock, int level, int optname, |
b7058842 | 2218 | char __user *optval, unsigned int optlen) |
9a4595bc PM |
2219 | { |
2220 | struct sock *sk = sock->sk; | |
2221 | struct netlink_sock *nlk = nlk_sk(sk); | |
eb496534 JB |
2222 | unsigned int val = 0; |
2223 | int err; | |
9a4595bc PM |
2224 | |
2225 | if (level != SOL_NETLINK) | |
2226 | return -ENOPROTOOPT; | |
2227 | ||
ccdfcc39 PM |
2228 | if (optname != NETLINK_RX_RING && optname != NETLINK_TX_RING && |
2229 | optlen >= sizeof(int) && | |
eb496534 | 2230 | get_user(val, (unsigned int __user *)optval)) |
9a4595bc PM |
2231 | return -EFAULT; |
2232 | ||
2233 | switch (optname) { | |
2234 | case NETLINK_PKTINFO: | |
2235 | if (val) | |
cc3a572f | 2236 | nlk->flags |= NETLINK_F_RECV_PKTINFO; |
9a4595bc | 2237 | else |
cc3a572f | 2238 | nlk->flags &= ~NETLINK_F_RECV_PKTINFO; |
9a4595bc PM |
2239 | err = 0; |
2240 | break; | |
2241 | case NETLINK_ADD_MEMBERSHIP: | |
2242 | case NETLINK_DROP_MEMBERSHIP: { | |
5187cd05 | 2243 | if (!netlink_allowed(sock, NL_CFG_F_NONROOT_RECV)) |
9a4595bc | 2244 | return -EPERM; |
b4ff4f04 JB |
2245 | err = netlink_realloc_groups(sk); |
2246 | if (err) | |
2247 | return err; | |
9a4595bc PM |
2248 | if (!val || val - 1 >= nlk->ngroups) |
2249 | return -EINVAL; | |
7774d5e0 | 2250 | if (optname == NETLINK_ADD_MEMBERSHIP && nlk->netlink_bind) { |
023e2cfa | 2251 | err = nlk->netlink_bind(sock_net(sk), val); |
4f520900 RGB |
2252 | if (err) |
2253 | return err; | |
2254 | } | |
9a4595bc | 2255 | netlink_table_grab(); |
84659eb5 JB |
2256 | netlink_update_socket_mc(nlk, val, |
2257 | optname == NETLINK_ADD_MEMBERSHIP); | |
9a4595bc | 2258 | netlink_table_ungrab(); |
7774d5e0 | 2259 | if (optname == NETLINK_DROP_MEMBERSHIP && nlk->netlink_unbind) |
023e2cfa | 2260 | nlk->netlink_unbind(sock_net(sk), val); |
03292745 | 2261 | |
9a4595bc PM |
2262 | err = 0; |
2263 | break; | |
2264 | } | |
be0c22a4 PNA |
2265 | case NETLINK_BROADCAST_ERROR: |
2266 | if (val) | |
cc3a572f | 2267 | nlk->flags |= NETLINK_F_BROADCAST_SEND_ERROR; |
be0c22a4 | 2268 | else |
cc3a572f | 2269 | nlk->flags &= ~NETLINK_F_BROADCAST_SEND_ERROR; |
be0c22a4 PNA |
2270 | err = 0; |
2271 | break; | |
38938bfe PNA |
2272 | case NETLINK_NO_ENOBUFS: |
2273 | if (val) { | |
cc3a572f ND |
2274 | nlk->flags |= NETLINK_F_RECV_NO_ENOBUFS; |
2275 | clear_bit(NETLINK_S_CONGESTED, &nlk->state); | |
38938bfe | 2276 | wake_up_interruptible(&nlk->wait); |
658cb354 | 2277 | } else { |
cc3a572f | 2278 | nlk->flags &= ~NETLINK_F_RECV_NO_ENOBUFS; |
658cb354 | 2279 | } |
38938bfe PNA |
2280 | err = 0; |
2281 | break; | |
ccdfcc39 PM |
2282 | #ifdef CONFIG_NETLINK_MMAP |
2283 | case NETLINK_RX_RING: | |
2284 | case NETLINK_TX_RING: { | |
2285 | struct nl_mmap_req req; | |
2286 | ||
2287 | /* Rings might consume more memory than queue limits, require | |
2288 | * CAP_NET_ADMIN. | |
2289 | */ | |
2290 | if (!capable(CAP_NET_ADMIN)) | |
2291 | return -EPERM; | |
2292 | if (optlen < sizeof(req)) | |
2293 | return -EINVAL; | |
2294 | if (copy_from_user(&req, optval, sizeof(req))) | |
2295 | return -EFAULT; | |
0470eb99 | 2296 | err = netlink_set_ring(sk, &req, |
ccdfcc39 PM |
2297 | optname == NETLINK_TX_RING); |
2298 | break; | |
2299 | } | |
2300 | #endif /* CONFIG_NETLINK_MMAP */ | |
59324cf3 ND |
2301 | case NETLINK_LISTEN_ALL_NSID: |
2302 | if (!ns_capable(sock_net(sk)->user_ns, CAP_NET_BROADCAST)) | |
2303 | return -EPERM; | |
2304 | ||
2305 | if (val) | |
2306 | nlk->flags |= NETLINK_F_LISTEN_ALL_NSID; | |
2307 | else | |
2308 | nlk->flags &= ~NETLINK_F_LISTEN_ALL_NSID; | |
2309 | err = 0; | |
2310 | break; | |
0a6a3a23 CR |
2311 | case NETLINK_CAP_ACK: |
2312 | if (val) | |
2313 | nlk->flags |= NETLINK_F_CAP_ACK; | |
2314 | else | |
2315 | nlk->flags &= ~NETLINK_F_CAP_ACK; | |
2316 | err = 0; | |
2317 | break; | |
9a4595bc PM |
2318 | default: |
2319 | err = -ENOPROTOOPT; | |
2320 | } | |
2321 | return err; | |
2322 | } | |
2323 | ||
2324 | static int netlink_getsockopt(struct socket *sock, int level, int optname, | |
746fac4d | 2325 | char __user *optval, int __user *optlen) |
9a4595bc PM |
2326 | { |
2327 | struct sock *sk = sock->sk; | |
2328 | struct netlink_sock *nlk = nlk_sk(sk); | |
2329 | int len, val, err; | |
2330 | ||
2331 | if (level != SOL_NETLINK) | |
2332 | return -ENOPROTOOPT; | |
2333 | ||
2334 | if (get_user(len, optlen)) | |
2335 | return -EFAULT; | |
2336 | if (len < 0) | |
2337 | return -EINVAL; | |
2338 | ||
2339 | switch (optname) { | |
2340 | case NETLINK_PKTINFO: | |
2341 | if (len < sizeof(int)) | |
2342 | return -EINVAL; | |
2343 | len = sizeof(int); | |
cc3a572f | 2344 | val = nlk->flags & NETLINK_F_RECV_PKTINFO ? 1 : 0; |
a27b58fe HC |
2345 | if (put_user(len, optlen) || |
2346 | put_user(val, optval)) | |
2347 | return -EFAULT; | |
9a4595bc PM |
2348 | err = 0; |
2349 | break; | |
be0c22a4 PNA |
2350 | case NETLINK_BROADCAST_ERROR: |
2351 | if (len < sizeof(int)) | |
2352 | return -EINVAL; | |
2353 | len = sizeof(int); | |
cc3a572f | 2354 | val = nlk->flags & NETLINK_F_BROADCAST_SEND_ERROR ? 1 : 0; |
be0c22a4 PNA |
2355 | if (put_user(len, optlen) || |
2356 | put_user(val, optval)) | |
2357 | return -EFAULT; | |
2358 | err = 0; | |
2359 | break; | |
38938bfe PNA |
2360 | case NETLINK_NO_ENOBUFS: |
2361 | if (len < sizeof(int)) | |
2362 | return -EINVAL; | |
2363 | len = sizeof(int); | |
cc3a572f | 2364 | val = nlk->flags & NETLINK_F_RECV_NO_ENOBUFS ? 1 : 0; |
38938bfe PNA |
2365 | if (put_user(len, optlen) || |
2366 | put_user(val, optval)) | |
2367 | return -EFAULT; | |
2368 | err = 0; | |
2369 | break; | |
b42be38b DH |
2370 | case NETLINK_LIST_MEMBERSHIPS: { |
2371 | int pos, idx, shift; | |
2372 | ||
2373 | err = 0; | |
47191d65 | 2374 | netlink_lock_table(); |
b42be38b DH |
2375 | for (pos = 0; pos * 8 < nlk->ngroups; pos += sizeof(u32)) { |
2376 | if (len - pos < sizeof(u32)) | |
2377 | break; | |
2378 | ||
2379 | idx = pos / sizeof(unsigned long); | |
2380 | shift = (pos % sizeof(unsigned long)) * 8; | |
2381 | if (put_user((u32)(nlk->groups[idx] >> shift), | |
2382 | (u32 __user *)(optval + pos))) { | |
2383 | err = -EFAULT; | |
2384 | break; | |
2385 | } | |
2386 | } | |
2387 | if (put_user(ALIGN(nlk->ngroups / 8, sizeof(u32)), optlen)) | |
2388 | err = -EFAULT; | |
47191d65 | 2389 | netlink_unlock_table(); |
b42be38b DH |
2390 | break; |
2391 | } | |
0a6a3a23 CR |
2392 | case NETLINK_CAP_ACK: |
2393 | if (len < sizeof(int)) | |
2394 | return -EINVAL; | |
2395 | len = sizeof(int); | |
2396 | val = nlk->flags & NETLINK_F_CAP_ACK ? 1 : 0; | |
2397 | if (put_user(len, optlen) || | |
2398 | put_user(val, optval)) | |
2399 | return -EFAULT; | |
2400 | err = 0; | |
2401 | break; | |
9a4595bc PM |
2402 | default: |
2403 | err = -ENOPROTOOPT; | |
2404 | } | |
2405 | return err; | |
2406 | } | |
2407 | ||
2408 | static void netlink_cmsg_recv_pktinfo(struct msghdr *msg, struct sk_buff *skb) | |
2409 | { | |
2410 | struct nl_pktinfo info; | |
2411 | ||
2412 | info.group = NETLINK_CB(skb).dst_group; | |
2413 | put_cmsg(msg, SOL_NETLINK, NETLINK_PKTINFO, sizeof(info), &info); | |
2414 | } | |
2415 | ||
59324cf3 ND |
2416 | static void netlink_cmsg_listen_all_nsid(struct sock *sk, struct msghdr *msg, |
2417 | struct sk_buff *skb) | |
2418 | { | |
2419 | if (!NETLINK_CB(skb).nsid_is_set) | |
2420 | return; | |
2421 | ||
2422 | put_cmsg(msg, SOL_NETLINK, NETLINK_LISTEN_ALL_NSID, sizeof(int), | |
2423 | &NETLINK_CB(skb).nsid); | |
2424 | } | |
2425 | ||
1b784140 | 2426 | static int netlink_sendmsg(struct socket *sock, struct msghdr *msg, size_t len) |
1da177e4 | 2427 | { |
1da177e4 LT |
2428 | struct sock *sk = sock->sk; |
2429 | struct netlink_sock *nlk = nlk_sk(sk); | |
342dfc30 | 2430 | DECLARE_SOCKADDR(struct sockaddr_nl *, addr, msg->msg_name); |
15e47304 | 2431 | u32 dst_portid; |
d629b836 | 2432 | u32 dst_group; |
1da177e4 LT |
2433 | struct sk_buff *skb; |
2434 | int err; | |
2435 | struct scm_cookie scm; | |
2d7a85f4 | 2436 | u32 netlink_skb_flags = 0; |
1da177e4 LT |
2437 | |
2438 | if (msg->msg_flags&MSG_OOB) | |
2439 | return -EOPNOTSUPP; | |
2440 | ||
7cc05662 | 2441 | err = scm_send(sock, msg, &scm, true); |
1da177e4 LT |
2442 | if (err < 0) |
2443 | return err; | |
2444 | ||
2445 | if (msg->msg_namelen) { | |
b47030c7 | 2446 | err = -EINVAL; |
1da177e4 | 2447 | if (addr->nl_family != AF_NETLINK) |
b47030c7 | 2448 | goto out; |
15e47304 | 2449 | dst_portid = addr->nl_pid; |
d629b836 | 2450 | dst_group = ffs(addr->nl_groups); |
b47030c7 | 2451 | err = -EPERM; |
15e47304 | 2452 | if ((dst_group || dst_portid) && |
5187cd05 | 2453 | !netlink_allowed(sock, NL_CFG_F_NONROOT_SEND)) |
b47030c7 | 2454 | goto out; |
2d7a85f4 | 2455 | netlink_skb_flags |= NETLINK_SKB_DST; |
1da177e4 | 2456 | } else { |
15e47304 | 2457 | dst_portid = nlk->dst_portid; |
d629b836 | 2458 | dst_group = nlk->dst_group; |
1da177e4 LT |
2459 | } |
2460 | ||
da314c99 | 2461 | if (!nlk->bound) { |
1da177e4 LT |
2462 | err = netlink_autobind(sock); |
2463 | if (err) | |
2464 | goto out; | |
da314c99 HX |
2465 | } else { |
2466 | /* Ensure nlk is hashed and visible. */ | |
2467 | smp_rmb(); | |
1da177e4 LT |
2468 | } |
2469 | ||
a8866ff6 AV |
2470 | /* It's a really convoluted way for userland to ask for mmaped |
2471 | * sendmsg(), but that's what we've got... | |
2472 | */ | |
5fd96123 | 2473 | if (netlink_tx_is_mmaped(sk) && |
c953e239 | 2474 | iter_is_iovec(&msg->msg_iter) && |
a8866ff6 | 2475 | msg->msg_iter.nr_segs == 1 && |
c0371da6 | 2476 | msg->msg_iter.iov->iov_base == NULL) { |
5fd96123 | 2477 | err = netlink_mmap_sendmsg(sk, msg, dst_portid, dst_group, |
7cc05662 | 2478 | &scm); |
5fd96123 PM |
2479 | goto out; |
2480 | } | |
2481 | ||
1da177e4 LT |
2482 | err = -EMSGSIZE; |
2483 | if (len > sk->sk_sndbuf - 32) | |
2484 | goto out; | |
2485 | err = -ENOBUFS; | |
3a36515f | 2486 | skb = netlink_alloc_large_skb(len, dst_group); |
6ac552fd | 2487 | if (skb == NULL) |
1da177e4 LT |
2488 | goto out; |
2489 | ||
15e47304 | 2490 | NETLINK_CB(skb).portid = nlk->portid; |
d629b836 | 2491 | NETLINK_CB(skb).dst_group = dst_group; |
7cc05662 | 2492 | NETLINK_CB(skb).creds = scm.creds; |
2d7a85f4 | 2493 | NETLINK_CB(skb).flags = netlink_skb_flags; |
1da177e4 | 2494 | |
1da177e4 | 2495 | err = -EFAULT; |
6ce8e9ce | 2496 | if (memcpy_from_msg(skb_put(skb, len), msg, len)) { |
1da177e4 LT |
2497 | kfree_skb(skb); |
2498 | goto out; | |
2499 | } | |
2500 | ||
2501 | err = security_netlink_send(sk, skb); | |
2502 | if (err) { | |
2503 | kfree_skb(skb); | |
2504 | goto out; | |
2505 | } | |
2506 | ||
d629b836 | 2507 | if (dst_group) { |
1da177e4 | 2508 | atomic_inc(&skb->users); |
15e47304 | 2509 | netlink_broadcast(sk, skb, dst_portid, dst_group, GFP_KERNEL); |
1da177e4 | 2510 | } |
15e47304 | 2511 | err = netlink_unicast(sk, skb, dst_portid, msg->msg_flags&MSG_DONTWAIT); |
1da177e4 LT |
2512 | |
2513 | out: | |
7cc05662 | 2514 | scm_destroy(&scm); |
1da177e4 LT |
2515 | return err; |
2516 | } | |
2517 | ||
1b784140 | 2518 | static int netlink_recvmsg(struct socket *sock, struct msghdr *msg, size_t len, |
1da177e4 LT |
2519 | int flags) |
2520 | { | |
1da177e4 LT |
2521 | struct scm_cookie scm; |
2522 | struct sock *sk = sock->sk; | |
2523 | struct netlink_sock *nlk = nlk_sk(sk); | |
2524 | int noblock = flags&MSG_DONTWAIT; | |
2525 | size_t copied; | |
68d6ac6d | 2526 | struct sk_buff *skb, *data_skb; |
b44d211e | 2527 | int err, ret; |
1da177e4 LT |
2528 | |
2529 | if (flags&MSG_OOB) | |
2530 | return -EOPNOTSUPP; | |
2531 | ||
2532 | copied = 0; | |
2533 | ||
6ac552fd PM |
2534 | skb = skb_recv_datagram(sk, flags, noblock, &err); |
2535 | if (skb == NULL) | |
1da177e4 LT |
2536 | goto out; |
2537 | ||
68d6ac6d JB |
2538 | data_skb = skb; |
2539 | ||
1dacc76d JB |
2540 | #ifdef CONFIG_COMPAT_NETLINK_MESSAGES |
2541 | if (unlikely(skb_shinfo(skb)->frag_list)) { | |
1dacc76d | 2542 | /* |
68d6ac6d JB |
2543 | * If this skb has a frag_list, then here that means that we |
2544 | * will have to use the frag_list skb's data for compat tasks | |
2545 | * and the regular skb's data for normal (non-compat) tasks. | |
1dacc76d | 2546 | * |
68d6ac6d JB |
2547 | * If we need to send the compat skb, assign it to the |
2548 | * 'data_skb' variable so that it will be used below for data | |
2549 | * copying. We keep 'skb' for everything else, including | |
2550 | * freeing both later. | |
1dacc76d | 2551 | */ |
68d6ac6d JB |
2552 | if (flags & MSG_CMSG_COMPAT) |
2553 | data_skb = skb_shinfo(skb)->frag_list; | |
1dacc76d JB |
2554 | } |
2555 | #endif | |
2556 | ||
9063e21f ED |
2557 | /* Record the max length of recvmsg() calls for future allocations */ |
2558 | nlk->max_recvmsg_len = max(nlk->max_recvmsg_len, len); | |
2559 | nlk->max_recvmsg_len = min_t(size_t, nlk->max_recvmsg_len, | |
2560 | 16384); | |
2561 | ||
68d6ac6d | 2562 | copied = data_skb->len; |
1da177e4 LT |
2563 | if (len < copied) { |
2564 | msg->msg_flags |= MSG_TRUNC; | |
2565 | copied = len; | |
2566 | } | |
2567 | ||
68d6ac6d | 2568 | skb_reset_transport_header(data_skb); |
51f3d02b | 2569 | err = skb_copy_datagram_msg(data_skb, 0, msg, copied); |
1da177e4 LT |
2570 | |
2571 | if (msg->msg_name) { | |
342dfc30 | 2572 | DECLARE_SOCKADDR(struct sockaddr_nl *, addr, msg->msg_name); |
1da177e4 LT |
2573 | addr->nl_family = AF_NETLINK; |
2574 | addr->nl_pad = 0; | |
15e47304 | 2575 | addr->nl_pid = NETLINK_CB(skb).portid; |
d629b836 | 2576 | addr->nl_groups = netlink_group_mask(NETLINK_CB(skb).dst_group); |
1da177e4 LT |
2577 | msg->msg_namelen = sizeof(*addr); |
2578 | } | |
2579 | ||
cc3a572f | 2580 | if (nlk->flags & NETLINK_F_RECV_PKTINFO) |
cc9a06cd | 2581 | netlink_cmsg_recv_pktinfo(msg, skb); |
59324cf3 ND |
2582 | if (nlk->flags & NETLINK_F_LISTEN_ALL_NSID) |
2583 | netlink_cmsg_listen_all_nsid(sk, msg, skb); | |
cc9a06cd | 2584 | |
7cc05662 CH |
2585 | memset(&scm, 0, sizeof(scm)); |
2586 | scm.creds = *NETLINK_CREDS(skb); | |
188ccb55 | 2587 | if (flags & MSG_TRUNC) |
68d6ac6d | 2588 | copied = data_skb->len; |
daa3766e | 2589 | |
1da177e4 LT |
2590 | skb_free_datagram(sk, skb); |
2591 | ||
16b304f3 PS |
2592 | if (nlk->cb_running && |
2593 | atomic_read(&sk->sk_rmem_alloc) <= sk->sk_rcvbuf / 2) { | |
b44d211e AV |
2594 | ret = netlink_dump(sk); |
2595 | if (ret) { | |
ac30ef83 | 2596 | sk->sk_err = -ret; |
b44d211e AV |
2597 | sk->sk_error_report(sk); |
2598 | } | |
2599 | } | |
1da177e4 | 2600 | |
7cc05662 | 2601 | scm_recv(sock, msg, &scm, flags); |
1da177e4 LT |
2602 | out: |
2603 | netlink_rcv_wake(sk); | |
2604 | return err ? : copied; | |
2605 | } | |
2606 | ||
676d2369 | 2607 | static void netlink_data_ready(struct sock *sk) |
1da177e4 | 2608 | { |
cd40b7d3 | 2609 | BUG(); |
1da177e4 LT |
2610 | } |
2611 | ||
2612 | /* | |
746fac4d | 2613 | * We export these functions to other modules. They provide a |
1da177e4 LT |
2614 | * complete set of kernel non-blocking support for message |
2615 | * queueing. | |
2616 | */ | |
2617 | ||
2618 | struct sock * | |
9f00d977 PNA |
2619 | __netlink_kernel_create(struct net *net, int unit, struct module *module, |
2620 | struct netlink_kernel_cfg *cfg) | |
1da177e4 LT |
2621 | { |
2622 | struct socket *sock; | |
2623 | struct sock *sk; | |
77247bbb | 2624 | struct netlink_sock *nlk; |
5c398dc8 | 2625 | struct listeners *listeners = NULL; |
a31f2d17 PNA |
2626 | struct mutex *cb_mutex = cfg ? cfg->cb_mutex : NULL; |
2627 | unsigned int groups; | |
1da177e4 | 2628 | |
fab2caf6 | 2629 | BUG_ON(!nl_table); |
1da177e4 | 2630 | |
6ac552fd | 2631 | if (unit < 0 || unit >= MAX_LINKS) |
1da177e4 LT |
2632 | return NULL; |
2633 | ||
2634 | if (sock_create_lite(PF_NETLINK, SOCK_DGRAM, unit, &sock)) | |
2635 | return NULL; | |
13d3078e EB |
2636 | |
2637 | if (__netlink_create(net, sock, cb_mutex, unit, 1) < 0) | |
23fe1866 PE |
2638 | goto out_sock_release_nosk; |
2639 | ||
2640 | sk = sock->sk; | |
4fdb3bb7 | 2641 | |
a31f2d17 | 2642 | if (!cfg || cfg->groups < 32) |
4277a083 | 2643 | groups = 32; |
a31f2d17 PNA |
2644 | else |
2645 | groups = cfg->groups; | |
4277a083 | 2646 | |
5c398dc8 | 2647 | listeners = kzalloc(sizeof(*listeners) + NLGRPSZ(groups), GFP_KERNEL); |
4277a083 PM |
2648 | if (!listeners) |
2649 | goto out_sock_release; | |
2650 | ||
1da177e4 | 2651 | sk->sk_data_ready = netlink_data_ready; |
a31f2d17 PNA |
2652 | if (cfg && cfg->input) |
2653 | nlk_sk(sk)->netlink_rcv = cfg->input; | |
1da177e4 | 2654 | |
8ea65f4a | 2655 | if (netlink_insert(sk, 0)) |
77247bbb | 2656 | goto out_sock_release; |
4fdb3bb7 | 2657 | |
77247bbb | 2658 | nlk = nlk_sk(sk); |
cc3a572f | 2659 | nlk->flags |= NETLINK_F_KERNEL_SOCKET; |
4fdb3bb7 | 2660 | |
4fdb3bb7 | 2661 | netlink_table_grab(); |
b4b51029 EB |
2662 | if (!nl_table[unit].registered) { |
2663 | nl_table[unit].groups = groups; | |
5c398dc8 | 2664 | rcu_assign_pointer(nl_table[unit].listeners, listeners); |
b4b51029 EB |
2665 | nl_table[unit].cb_mutex = cb_mutex; |
2666 | nl_table[unit].module = module; | |
9785e10a PNA |
2667 | if (cfg) { |
2668 | nl_table[unit].bind = cfg->bind; | |
6251edd9 | 2669 | nl_table[unit].unbind = cfg->unbind; |
9785e10a | 2670 | nl_table[unit].flags = cfg->flags; |
da12c90e G |
2671 | if (cfg->compare) |
2672 | nl_table[unit].compare = cfg->compare; | |
9785e10a | 2673 | } |
b4b51029 | 2674 | nl_table[unit].registered = 1; |
f937f1f4 JJ |
2675 | } else { |
2676 | kfree(listeners); | |
869e58f8 | 2677 | nl_table[unit].registered++; |
b4b51029 | 2678 | } |
4fdb3bb7 | 2679 | netlink_table_ungrab(); |
77247bbb PM |
2680 | return sk; |
2681 | ||
4fdb3bb7 | 2682 | out_sock_release: |
4277a083 | 2683 | kfree(listeners); |
9dfbec1f | 2684 | netlink_kernel_release(sk); |
23fe1866 PE |
2685 | return NULL; |
2686 | ||
2687 | out_sock_release_nosk: | |
4fdb3bb7 | 2688 | sock_release(sock); |
77247bbb | 2689 | return NULL; |
1da177e4 | 2690 | } |
9f00d977 | 2691 | EXPORT_SYMBOL(__netlink_kernel_create); |
b7c6ba6e DL |
2692 | |
2693 | void | |
2694 | netlink_kernel_release(struct sock *sk) | |
2695 | { | |
13d3078e EB |
2696 | if (sk == NULL || sk->sk_socket == NULL) |
2697 | return; | |
2698 | ||
2699 | sock_release(sk->sk_socket); | |
b7c6ba6e DL |
2700 | } |
2701 | EXPORT_SYMBOL(netlink_kernel_release); | |
2702 | ||
d136f1bd | 2703 | int __netlink_change_ngroups(struct sock *sk, unsigned int groups) |
b4ff4f04 | 2704 | { |
5c398dc8 | 2705 | struct listeners *new, *old; |
b4ff4f04 | 2706 | struct netlink_table *tbl = &nl_table[sk->sk_protocol]; |
b4ff4f04 JB |
2707 | |
2708 | if (groups < 32) | |
2709 | groups = 32; | |
2710 | ||
b4ff4f04 | 2711 | if (NLGRPSZ(tbl->groups) < NLGRPSZ(groups)) { |
5c398dc8 ED |
2712 | new = kzalloc(sizeof(*new) + NLGRPSZ(groups), GFP_ATOMIC); |
2713 | if (!new) | |
d136f1bd | 2714 | return -ENOMEM; |
6d772ac5 | 2715 | old = nl_deref_protected(tbl->listeners); |
5c398dc8 ED |
2716 | memcpy(new->masks, old->masks, NLGRPSZ(tbl->groups)); |
2717 | rcu_assign_pointer(tbl->listeners, new); | |
2718 | ||
37b6b935 | 2719 | kfree_rcu(old, rcu); |
b4ff4f04 JB |
2720 | } |
2721 | tbl->groups = groups; | |
2722 | ||
d136f1bd JB |
2723 | return 0; |
2724 | } | |
2725 | ||
2726 | /** | |
2727 | * netlink_change_ngroups - change number of multicast groups | |
2728 | * | |
2729 | * This changes the number of multicast groups that are available | |
2730 | * on a certain netlink family. Note that it is not possible to | |
2731 | * change the number of groups to below 32. Also note that it does | |
2732 | * not implicitly call netlink_clear_multicast_users() when the | |
2733 | * number of groups is reduced. | |
2734 | * | |
2735 | * @sk: The kernel netlink socket, as returned by netlink_kernel_create(). | |
2736 | * @groups: The new number of groups. | |
2737 | */ | |
2738 | int netlink_change_ngroups(struct sock *sk, unsigned int groups) | |
2739 | { | |
2740 | int err; | |
2741 | ||
2742 | netlink_table_grab(); | |
2743 | err = __netlink_change_ngroups(sk, groups); | |
b4ff4f04 | 2744 | netlink_table_ungrab(); |
d136f1bd | 2745 | |
b4ff4f04 JB |
2746 | return err; |
2747 | } | |
b4ff4f04 | 2748 | |
b8273570 JB |
2749 | void __netlink_clear_multicast_users(struct sock *ksk, unsigned int group) |
2750 | { | |
2751 | struct sock *sk; | |
b8273570 JB |
2752 | struct netlink_table *tbl = &nl_table[ksk->sk_protocol]; |
2753 | ||
b67bfe0d | 2754 | sk_for_each_bound(sk, &tbl->mc_list) |
b8273570 JB |
2755 | netlink_update_socket_mc(nlk_sk(sk), group, 0); |
2756 | } | |
2757 | ||
a46621a3 | 2758 | struct nlmsghdr * |
15e47304 | 2759 | __nlmsg_put(struct sk_buff *skb, u32 portid, u32 seq, int type, int len, int flags) |
a46621a3 DV |
2760 | { |
2761 | struct nlmsghdr *nlh; | |
573ce260 | 2762 | int size = nlmsg_msg_size(len); |
a46621a3 | 2763 | |
23b45672 | 2764 | nlh = (struct nlmsghdr *)skb_put(skb, NLMSG_ALIGN(size)); |
a46621a3 DV |
2765 | nlh->nlmsg_type = type; |
2766 | nlh->nlmsg_len = size; | |
2767 | nlh->nlmsg_flags = flags; | |
15e47304 | 2768 | nlh->nlmsg_pid = portid; |
a46621a3 DV |
2769 | nlh->nlmsg_seq = seq; |
2770 | if (!__builtin_constant_p(size) || NLMSG_ALIGN(size) - size != 0) | |
573ce260 | 2771 | memset(nlmsg_data(nlh) + len, 0, NLMSG_ALIGN(size) - size); |
a46621a3 DV |
2772 | return nlh; |
2773 | } | |
2774 | EXPORT_SYMBOL(__nlmsg_put); | |
2775 | ||
1da177e4 LT |
2776 | /* |
2777 | * It looks a bit ugly. | |
2778 | * It would be better to create kernel thread. | |
2779 | */ | |
2780 | ||
2781 | static int netlink_dump(struct sock *sk) | |
2782 | { | |
2783 | struct netlink_sock *nlk = nlk_sk(sk); | |
2784 | struct netlink_callback *cb; | |
c7ac8679 | 2785 | struct sk_buff *skb = NULL; |
1da177e4 | 2786 | struct nlmsghdr *nlh; |
bf8b79e4 | 2787 | int len, err = -ENOBUFS; |
db65a3aa | 2788 | int alloc_min_size; |
c7ac8679 | 2789 | int alloc_size; |
1da177e4 | 2790 | |
af65bdfc | 2791 | mutex_lock(nlk->cb_mutex); |
16b304f3 | 2792 | if (!nlk->cb_running) { |
bf8b79e4 TG |
2793 | err = -EINVAL; |
2794 | goto errout_skb; | |
1da177e4 LT |
2795 | } |
2796 | ||
f9c22888 PM |
2797 | if (!netlink_rx_is_mmaped(sk) && |
2798 | atomic_read(&sk->sk_rmem_alloc) >= sk->sk_rcvbuf) | |
2799 | goto errout_skb; | |
9063e21f ED |
2800 | |
2801 | /* NLMSG_GOODSIZE is small to avoid high order allocations being | |
2802 | * required, but it makes sense to _attempt_ a 16K bytes allocation | |
2803 | * to reduce number of system calls on dump operations, if user | |
2804 | * ever provided a big enough buffer. | |
2805 | */ | |
db65a3aa AR |
2806 | cb = &nlk->cb; |
2807 | alloc_min_size = max_t(int, cb->min_dump_alloc, NLMSG_GOODSIZE); | |
2808 | ||
2809 | if (alloc_min_size < nlk->max_recvmsg_len) { | |
2810 | alloc_size = nlk->max_recvmsg_len; | |
2811 | skb = netlink_alloc_skb(sk, alloc_size, nlk->portid, | |
9063e21f ED |
2812 | GFP_KERNEL | |
2813 | __GFP_NOWARN | | |
2814 | __GFP_NORETRY); | |
9063e21f | 2815 | } |
db65a3aa AR |
2816 | if (!skb) { |
2817 | alloc_size = alloc_min_size; | |
9063e21f ED |
2818 | skb = netlink_alloc_skb(sk, alloc_size, nlk->portid, |
2819 | GFP_KERNEL); | |
db65a3aa | 2820 | } |
c7ac8679 | 2821 | if (!skb) |
c63d6ea3 | 2822 | goto errout_skb; |
db65a3aa AR |
2823 | |
2824 | /* Trim skb to allocated size. User is expected to provide buffer as | |
2825 | * large as max(min_dump_alloc, 16KiB (mac_recvmsg_len capped at | |
2826 | * netlink_recvmsg())). dump will pack as many smaller messages as | |
2827 | * could fit within the allocated skb. skb is typically allocated | |
2828 | * with larger space than required (could be as much as near 2x the | |
2829 | * requested size with align to next power of 2 approach). Allowing | |
2830 | * dump to use the excess space makes it difficult for a user to have a | |
2831 | * reasonable static buffer based on the expected largest dump of a | |
2832 | * single netdev. The outcome is MSG_TRUNC error. | |
2833 | */ | |
2834 | skb_reserve(skb, skb_tailroom(skb) - alloc_size); | |
f9c22888 | 2835 | netlink_skb_set_owner_r(skb, sk); |
c7ac8679 | 2836 | |
1da177e4 LT |
2837 | len = cb->dump(skb, cb); |
2838 | ||
2839 | if (len > 0) { | |
af65bdfc | 2840 | mutex_unlock(nlk->cb_mutex); |
b1153f29 SH |
2841 | |
2842 | if (sk_filter(sk, skb)) | |
2843 | kfree_skb(skb); | |
4a7e7c2a ED |
2844 | else |
2845 | __netlink_sendskb(sk, skb); | |
1da177e4 LT |
2846 | return 0; |
2847 | } | |
2848 | ||
bf8b79e4 TG |
2849 | nlh = nlmsg_put_answer(skb, cb, NLMSG_DONE, sizeof(len), NLM_F_MULTI); |
2850 | if (!nlh) | |
2851 | goto errout_skb; | |
2852 | ||
670dc283 JB |
2853 | nl_dump_check_consistent(cb, nlh); |
2854 | ||
bf8b79e4 TG |
2855 | memcpy(nlmsg_data(nlh), &len, sizeof(len)); |
2856 | ||
b1153f29 SH |
2857 | if (sk_filter(sk, skb)) |
2858 | kfree_skb(skb); | |
4a7e7c2a ED |
2859 | else |
2860 | __netlink_sendskb(sk, skb); | |
1da177e4 | 2861 | |
a8f74b22 TG |
2862 | if (cb->done) |
2863 | cb->done(cb); | |
1da177e4 | 2864 | |
16b304f3 PS |
2865 | nlk->cb_running = false; |
2866 | mutex_unlock(nlk->cb_mutex); | |
6dc878a8 | 2867 | module_put(cb->module); |
16b304f3 | 2868 | consume_skb(cb->skb); |
1da177e4 | 2869 | return 0; |
1797754e | 2870 | |
bf8b79e4 | 2871 | errout_skb: |
af65bdfc | 2872 | mutex_unlock(nlk->cb_mutex); |
bf8b79e4 | 2873 | kfree_skb(skb); |
bf8b79e4 | 2874 | return err; |
1da177e4 LT |
2875 | } |
2876 | ||
6dc878a8 G |
2877 | int __netlink_dump_start(struct sock *ssk, struct sk_buff *skb, |
2878 | const struct nlmsghdr *nlh, | |
2879 | struct netlink_dump_control *control) | |
1da177e4 LT |
2880 | { |
2881 | struct netlink_callback *cb; | |
2882 | struct sock *sk; | |
2883 | struct netlink_sock *nlk; | |
b44d211e | 2884 | int ret; |
1da177e4 | 2885 | |
f9c22888 PM |
2886 | /* Memory mapped dump requests need to be copied to avoid looping |
2887 | * on the pending state in netlink_mmap_sendmsg() while the CB hold | |
2888 | * a reference to the skb. | |
2889 | */ | |
2890 | if (netlink_skb_is_mmaped(skb)) { | |
2891 | skb = skb_copy(skb, GFP_KERNEL); | |
16b304f3 | 2892 | if (skb == NULL) |
f9c22888 | 2893 | return -ENOBUFS; |
f9c22888 PM |
2894 | } else |
2895 | atomic_inc(&skb->users); | |
2896 | ||
15e47304 | 2897 | sk = netlink_lookup(sock_net(ssk), ssk->sk_protocol, NETLINK_CB(skb).portid); |
1da177e4 | 2898 | if (sk == NULL) { |
16b304f3 PS |
2899 | ret = -ECONNREFUSED; |
2900 | goto error_free; | |
1da177e4 | 2901 | } |
6dc878a8 | 2902 | |
16b304f3 | 2903 | nlk = nlk_sk(sk); |
af65bdfc | 2904 | mutex_lock(nlk->cb_mutex); |
6dc878a8 | 2905 | /* A dump is in progress... */ |
16b304f3 | 2906 | if (nlk->cb_running) { |
6dc878a8 | 2907 | ret = -EBUSY; |
16b304f3 | 2908 | goto error_unlock; |
1da177e4 | 2909 | } |
6dc878a8 | 2910 | /* add reference of module which cb->dump belongs to */ |
16b304f3 | 2911 | if (!try_module_get(control->module)) { |
6dc878a8 | 2912 | ret = -EPROTONOSUPPORT; |
16b304f3 | 2913 | goto error_unlock; |
6dc878a8 G |
2914 | } |
2915 | ||
16b304f3 PS |
2916 | cb = &nlk->cb; |
2917 | memset(cb, 0, sizeof(*cb)); | |
2918 | cb->dump = control->dump; | |
2919 | cb->done = control->done; | |
2920 | cb->nlh = nlh; | |
2921 | cb->data = control->data; | |
2922 | cb->module = control->module; | |
2923 | cb->min_dump_alloc = control->min_dump_alloc; | |
2924 | cb->skb = skb; | |
2925 | ||
2926 | nlk->cb_running = true; | |
2927 | ||
af65bdfc | 2928 | mutex_unlock(nlk->cb_mutex); |
1da177e4 | 2929 | |
b44d211e | 2930 | ret = netlink_dump(sk); |
1da177e4 | 2931 | sock_put(sk); |
5c58298c | 2932 | |
b44d211e AV |
2933 | if (ret) |
2934 | return ret; | |
2935 | ||
5c58298c DL |
2936 | /* We successfully started a dump, by returning -EINTR we |
2937 | * signal not to send ACK even if it was requested. | |
2938 | */ | |
2939 | return -EINTR; | |
16b304f3 PS |
2940 | |
2941 | error_unlock: | |
2942 | sock_put(sk); | |
2943 | mutex_unlock(nlk->cb_mutex); | |
2944 | error_free: | |
2945 | kfree_skb(skb); | |
2946 | return ret; | |
1da177e4 | 2947 | } |
6dc878a8 | 2948 | EXPORT_SYMBOL(__netlink_dump_start); |
1da177e4 LT |
2949 | |
2950 | void netlink_ack(struct sk_buff *in_skb, struct nlmsghdr *nlh, int err) | |
2951 | { | |
2952 | struct sk_buff *skb; | |
2953 | struct nlmsghdr *rep; | |
2954 | struct nlmsgerr *errmsg; | |
339bf98f | 2955 | size_t payload = sizeof(*errmsg); |
0a6a3a23 | 2956 | struct netlink_sock *nlk = nlk_sk(NETLINK_CB(in_skb).sk); |
1da177e4 | 2957 | |
0a6a3a23 CR |
2958 | /* Error messages get the original request appened, unless the user |
2959 | * requests to cap the error message. | |
2960 | */ | |
2961 | if (!(nlk->flags & NETLINK_F_CAP_ACK) && err) | |
339bf98f | 2962 | payload += nlmsg_len(nlh); |
1da177e4 | 2963 | |
f9c22888 PM |
2964 | skb = netlink_alloc_skb(in_skb->sk, nlmsg_total_size(payload), |
2965 | NETLINK_CB(in_skb).portid, GFP_KERNEL); | |
1da177e4 LT |
2966 | if (!skb) { |
2967 | struct sock *sk; | |
2968 | ||
3b1e0a65 | 2969 | sk = netlink_lookup(sock_net(in_skb->sk), |
b4b51029 | 2970 | in_skb->sk->sk_protocol, |
15e47304 | 2971 | NETLINK_CB(in_skb).portid); |
1da177e4 LT |
2972 | if (sk) { |
2973 | sk->sk_err = ENOBUFS; | |
2974 | sk->sk_error_report(sk); | |
2975 | sock_put(sk); | |
2976 | } | |
2977 | return; | |
2978 | } | |
2979 | ||
15e47304 | 2980 | rep = __nlmsg_put(skb, NETLINK_CB(in_skb).portid, nlh->nlmsg_seq, |
5dba93ae | 2981 | NLMSG_ERROR, payload, 0); |
bf8b79e4 | 2982 | errmsg = nlmsg_data(rep); |
1da177e4 | 2983 | errmsg->error = err; |
0a6a3a23 | 2984 | memcpy(&errmsg->msg, nlh, payload > sizeof(*errmsg) ? nlh->nlmsg_len : sizeof(*nlh)); |
15e47304 | 2985 | netlink_unicast(in_skb->sk, skb, NETLINK_CB(in_skb).portid, MSG_DONTWAIT); |
1da177e4 | 2986 | } |
6ac552fd | 2987 | EXPORT_SYMBOL(netlink_ack); |
1da177e4 | 2988 | |
cd40b7d3 | 2989 | int netlink_rcv_skb(struct sk_buff *skb, int (*cb)(struct sk_buff *, |
1d00a4eb | 2990 | struct nlmsghdr *)) |
82ace47a | 2991 | { |
82ace47a TG |
2992 | struct nlmsghdr *nlh; |
2993 | int err; | |
2994 | ||
2995 | while (skb->len >= nlmsg_total_size(0)) { | |
cd40b7d3 DL |
2996 | int msglen; |
2997 | ||
b529ccf2 | 2998 | nlh = nlmsg_hdr(skb); |
d35b6856 | 2999 | err = 0; |
82ace47a | 3000 | |
ad8e4b75 | 3001 | if (nlh->nlmsg_len < NLMSG_HDRLEN || skb->len < nlh->nlmsg_len) |
82ace47a TG |
3002 | return 0; |
3003 | ||
d35b6856 TG |
3004 | /* Only requests are handled by the kernel */ |
3005 | if (!(nlh->nlmsg_flags & NLM_F_REQUEST)) | |
5c58298c | 3006 | goto ack; |
45e7ae7f TG |
3007 | |
3008 | /* Skip control messages */ | |
3009 | if (nlh->nlmsg_type < NLMSG_MIN_TYPE) | |
5c58298c | 3010 | goto ack; |
d35b6856 | 3011 | |
1d00a4eb | 3012 | err = cb(skb, nlh); |
5c58298c DL |
3013 | if (err == -EINTR) |
3014 | goto skip; | |
3015 | ||
3016 | ack: | |
d35b6856 | 3017 | if (nlh->nlmsg_flags & NLM_F_ACK || err) |
82ace47a | 3018 | netlink_ack(skb, nlh, err); |
82ace47a | 3019 | |
5c58298c | 3020 | skip: |
6ac552fd | 3021 | msglen = NLMSG_ALIGN(nlh->nlmsg_len); |
cd40b7d3 DL |
3022 | if (msglen > skb->len) |
3023 | msglen = skb->len; | |
3024 | skb_pull(skb, msglen); | |
82ace47a TG |
3025 | } |
3026 | ||
3027 | return 0; | |
3028 | } | |
6ac552fd | 3029 | EXPORT_SYMBOL(netlink_rcv_skb); |
82ace47a | 3030 | |
d387f6ad TG |
3031 | /** |
3032 | * nlmsg_notify - send a notification netlink message | |
3033 | * @sk: netlink socket to use | |
3034 | * @skb: notification message | |
15e47304 | 3035 | * @portid: destination netlink portid for reports or 0 |
d387f6ad TG |
3036 | * @group: destination multicast group or 0 |
3037 | * @report: 1 to report back, 0 to disable | |
3038 | * @flags: allocation flags | |
3039 | */ | |
15e47304 | 3040 | int nlmsg_notify(struct sock *sk, struct sk_buff *skb, u32 portid, |
d387f6ad TG |
3041 | unsigned int group, int report, gfp_t flags) |
3042 | { | |
3043 | int err = 0; | |
3044 | ||
3045 | if (group) { | |
15e47304 | 3046 | int exclude_portid = 0; |
d387f6ad TG |
3047 | |
3048 | if (report) { | |
3049 | atomic_inc(&skb->users); | |
15e47304 | 3050 | exclude_portid = portid; |
d387f6ad TG |
3051 | } |
3052 | ||
1ce85fe4 PNA |
3053 | /* errors reported via destination sk->sk_err, but propagate |
3054 | * delivery errors if NETLINK_BROADCAST_ERROR flag is set */ | |
15e47304 | 3055 | err = nlmsg_multicast(sk, skb, exclude_portid, group, flags); |
d387f6ad TG |
3056 | } |
3057 | ||
1ce85fe4 PNA |
3058 | if (report) { |
3059 | int err2; | |
3060 | ||
15e47304 | 3061 | err2 = nlmsg_unicast(sk, skb, portid); |
1ce85fe4 PNA |
3062 | if (!err || err == -ESRCH) |
3063 | err = err2; | |
3064 | } | |
d387f6ad TG |
3065 | |
3066 | return err; | |
3067 | } | |
6ac552fd | 3068 | EXPORT_SYMBOL(nlmsg_notify); |
d387f6ad | 3069 | |
1da177e4 LT |
3070 | #ifdef CONFIG_PROC_FS |
3071 | struct nl_seq_iter { | |
e372c414 | 3072 | struct seq_net_private p; |
56d28b1e | 3073 | struct rhashtable_iter hti; |
1da177e4 | 3074 | int link; |
1da177e4 LT |
3075 | }; |
3076 | ||
56d28b1e | 3077 | static int netlink_walk_start(struct nl_seq_iter *iter) |
1da177e4 | 3078 | { |
56d28b1e | 3079 | int err; |
1da177e4 | 3080 | |
56d28b1e HX |
3081 | err = rhashtable_walk_init(&nl_table[iter->link].hash, &iter->hti); |
3082 | if (err) { | |
3083 | iter->link = MAX_LINKS; | |
3084 | return err; | |
1da177e4 | 3085 | } |
56d28b1e HX |
3086 | |
3087 | err = rhashtable_walk_start(&iter->hti); | |
3088 | return err == -EAGAIN ? 0 : err; | |
1da177e4 LT |
3089 | } |
3090 | ||
56d28b1e | 3091 | static void netlink_walk_stop(struct nl_seq_iter *iter) |
1da177e4 | 3092 | { |
56d28b1e HX |
3093 | rhashtable_walk_stop(&iter->hti); |
3094 | rhashtable_walk_exit(&iter->hti); | |
1da177e4 LT |
3095 | } |
3096 | ||
56d28b1e | 3097 | static void *__netlink_seq_next(struct seq_file *seq) |
1da177e4 | 3098 | { |
56d28b1e | 3099 | struct nl_seq_iter *iter = seq->private; |
e341694e | 3100 | struct netlink_sock *nlk; |
1da177e4 | 3101 | |
56d28b1e HX |
3102 | do { |
3103 | for (;;) { | |
3104 | int err; | |
1da177e4 | 3105 | |
56d28b1e | 3106 | nlk = rhashtable_walk_next(&iter->hti); |
746fac4d | 3107 | |
56d28b1e HX |
3108 | if (IS_ERR(nlk)) { |
3109 | if (PTR_ERR(nlk) == -EAGAIN) | |
3110 | continue; | |
e341694e | 3111 | |
56d28b1e HX |
3112 | return nlk; |
3113 | } | |
1da177e4 | 3114 | |
56d28b1e HX |
3115 | if (nlk) |
3116 | break; | |
1da177e4 | 3117 | |
56d28b1e HX |
3118 | netlink_walk_stop(iter); |
3119 | if (++iter->link >= MAX_LINKS) | |
3120 | return NULL; | |
da12c90e | 3121 | |
56d28b1e HX |
3122 | err = netlink_walk_start(iter); |
3123 | if (err) | |
3124 | return ERR_PTR(err); | |
1da177e4 | 3125 | } |
56d28b1e | 3126 | } while (sock_net(&nlk->sk) != seq_file_net(seq)); |
1da177e4 | 3127 | |
56d28b1e HX |
3128 | return nlk; |
3129 | } | |
1da177e4 | 3130 | |
56d28b1e HX |
3131 | static void *netlink_seq_start(struct seq_file *seq, loff_t *posp) |
3132 | { | |
3133 | struct nl_seq_iter *iter = seq->private; | |
3134 | void *obj = SEQ_START_TOKEN; | |
3135 | loff_t pos; | |
3136 | int err; | |
3137 | ||
3138 | iter->link = 0; | |
3139 | ||
3140 | err = netlink_walk_start(iter); | |
3141 | if (err) | |
3142 | return ERR_PTR(err); | |
3143 | ||
3144 | for (pos = *posp; pos && obj && !IS_ERR(obj); pos--) | |
3145 | obj = __netlink_seq_next(seq); | |
3146 | ||
3147 | return obj; | |
3148 | } | |
3149 | ||
3150 | static void *netlink_seq_next(struct seq_file *seq, void *v, loff_t *pos) | |
3151 | { | |
3152 | ++*pos; | |
3153 | return __netlink_seq_next(seq); | |
1da177e4 LT |
3154 | } |
3155 | ||
3156 | static void netlink_seq_stop(struct seq_file *seq, void *v) | |
3157 | { | |
56d28b1e HX |
3158 | struct nl_seq_iter *iter = seq->private; |
3159 | ||
3160 | if (iter->link >= MAX_LINKS) | |
3161 | return; | |
3162 | ||
3163 | netlink_walk_stop(iter); | |
1da177e4 LT |
3164 | } |
3165 | ||
3166 | ||
3167 | static int netlink_seq_show(struct seq_file *seq, void *v) | |
3168 | { | |
658cb354 | 3169 | if (v == SEQ_START_TOKEN) { |
1da177e4 LT |
3170 | seq_puts(seq, |
3171 | "sk Eth Pid Groups " | |
cf0aa4e0 | 3172 | "Rmem Wmem Dump Locks Drops Inode\n"); |
658cb354 | 3173 | } else { |
1da177e4 LT |
3174 | struct sock *s = v; |
3175 | struct netlink_sock *nlk = nlk_sk(s); | |
3176 | ||
16b304f3 | 3177 | seq_printf(seq, "%pK %-3d %-6u %08x %-8d %-8d %d %-8d %-8d %-8lu\n", |
1da177e4 LT |
3178 | s, |
3179 | s->sk_protocol, | |
15e47304 | 3180 | nlk->portid, |
513c2500 | 3181 | nlk->groups ? (u32)nlk->groups[0] : 0, |
31e6d363 ED |
3182 | sk_rmem_alloc_get(s), |
3183 | sk_wmem_alloc_get(s), | |
16b304f3 | 3184 | nlk->cb_running, |
38938bfe | 3185 | atomic_read(&s->sk_refcnt), |
cf0aa4e0 MY |
3186 | atomic_read(&s->sk_drops), |
3187 | sock_i_ino(s) | |
1da177e4 LT |
3188 | ); |
3189 | ||
3190 | } | |
3191 | return 0; | |
3192 | } | |
3193 | ||
56b3d975 | 3194 | static const struct seq_operations netlink_seq_ops = { |
1da177e4 LT |
3195 | .start = netlink_seq_start, |
3196 | .next = netlink_seq_next, | |
3197 | .stop = netlink_seq_stop, | |
3198 | .show = netlink_seq_show, | |
3199 | }; | |
3200 | ||
3201 | ||
3202 | static int netlink_seq_open(struct inode *inode, struct file *file) | |
3203 | { | |
e372c414 DL |
3204 | return seq_open_net(inode, file, &netlink_seq_ops, |
3205 | sizeof(struct nl_seq_iter)); | |
b4b51029 EB |
3206 | } |
3207 | ||
da7071d7 | 3208 | static const struct file_operations netlink_seq_fops = { |
1da177e4 LT |
3209 | .owner = THIS_MODULE, |
3210 | .open = netlink_seq_open, | |
3211 | .read = seq_read, | |
3212 | .llseek = seq_lseek, | |
e372c414 | 3213 | .release = seq_release_net, |
1da177e4 LT |
3214 | }; |
3215 | ||
3216 | #endif | |
3217 | ||
3218 | int netlink_register_notifier(struct notifier_block *nb) | |
3219 | { | |
e041c683 | 3220 | return atomic_notifier_chain_register(&netlink_chain, nb); |
1da177e4 | 3221 | } |
6ac552fd | 3222 | EXPORT_SYMBOL(netlink_register_notifier); |
1da177e4 LT |
3223 | |
3224 | int netlink_unregister_notifier(struct notifier_block *nb) | |
3225 | { | |
e041c683 | 3226 | return atomic_notifier_chain_unregister(&netlink_chain, nb); |
1da177e4 | 3227 | } |
6ac552fd | 3228 | EXPORT_SYMBOL(netlink_unregister_notifier); |
746fac4d | 3229 | |
90ddc4f0 | 3230 | static const struct proto_ops netlink_ops = { |
1da177e4 LT |
3231 | .family = PF_NETLINK, |
3232 | .owner = THIS_MODULE, | |
3233 | .release = netlink_release, | |
3234 | .bind = netlink_bind, | |
3235 | .connect = netlink_connect, | |
3236 | .socketpair = sock_no_socketpair, | |
3237 | .accept = sock_no_accept, | |
3238 | .getname = netlink_getname, | |
9652e931 | 3239 | .poll = netlink_poll, |
1da177e4 LT |
3240 | .ioctl = sock_no_ioctl, |
3241 | .listen = sock_no_listen, | |
3242 | .shutdown = sock_no_shutdown, | |
9a4595bc PM |
3243 | .setsockopt = netlink_setsockopt, |
3244 | .getsockopt = netlink_getsockopt, | |
1da177e4 LT |
3245 | .sendmsg = netlink_sendmsg, |
3246 | .recvmsg = netlink_recvmsg, | |
ccdfcc39 | 3247 | .mmap = netlink_mmap, |
1da177e4 LT |
3248 | .sendpage = sock_no_sendpage, |
3249 | }; | |
3250 | ||
ec1b4cf7 | 3251 | static const struct net_proto_family netlink_family_ops = { |
1da177e4 LT |
3252 | .family = PF_NETLINK, |
3253 | .create = netlink_create, | |
3254 | .owner = THIS_MODULE, /* for consistency 8) */ | |
3255 | }; | |
3256 | ||
4665079c | 3257 | static int __net_init netlink_net_init(struct net *net) |
b4b51029 EB |
3258 | { |
3259 | #ifdef CONFIG_PROC_FS | |
d4beaa66 | 3260 | if (!proc_create("netlink", 0, net->proc_net, &netlink_seq_fops)) |
b4b51029 EB |
3261 | return -ENOMEM; |
3262 | #endif | |
3263 | return 0; | |
3264 | } | |
3265 | ||
4665079c | 3266 | static void __net_exit netlink_net_exit(struct net *net) |
b4b51029 EB |
3267 | { |
3268 | #ifdef CONFIG_PROC_FS | |
ece31ffd | 3269 | remove_proc_entry("netlink", net->proc_net); |
b4b51029 EB |
3270 | #endif |
3271 | } | |
3272 | ||
b963ea89 DM |
3273 | static void __init netlink_add_usersock_entry(void) |
3274 | { | |
5c398dc8 | 3275 | struct listeners *listeners; |
b963ea89 DM |
3276 | int groups = 32; |
3277 | ||
5c398dc8 | 3278 | listeners = kzalloc(sizeof(*listeners) + NLGRPSZ(groups), GFP_KERNEL); |
b963ea89 | 3279 | if (!listeners) |
5c398dc8 | 3280 | panic("netlink_add_usersock_entry: Cannot allocate listeners\n"); |
b963ea89 DM |
3281 | |
3282 | netlink_table_grab(); | |
3283 | ||
3284 | nl_table[NETLINK_USERSOCK].groups = groups; | |
5c398dc8 | 3285 | rcu_assign_pointer(nl_table[NETLINK_USERSOCK].listeners, listeners); |
b963ea89 DM |
3286 | nl_table[NETLINK_USERSOCK].module = THIS_MODULE; |
3287 | nl_table[NETLINK_USERSOCK].registered = 1; | |
9785e10a | 3288 | nl_table[NETLINK_USERSOCK].flags = NL_CFG_F_NONROOT_SEND; |
b963ea89 DM |
3289 | |
3290 | netlink_table_ungrab(); | |
3291 | } | |
3292 | ||
022cbae6 | 3293 | static struct pernet_operations __net_initdata netlink_net_ops = { |
b4b51029 EB |
3294 | .init = netlink_net_init, |
3295 | .exit = netlink_net_exit, | |
3296 | }; | |
3297 | ||
49f7b33e | 3298 | static inline u32 netlink_hash(const void *data, u32 len, u32 seed) |
c428ecd1 HX |
3299 | { |
3300 | const struct netlink_sock *nlk = data; | |
3301 | struct netlink_compare_arg arg; | |
3302 | ||
da314c99 | 3303 | netlink_compare_arg_init(&arg, sock_net(&nlk->sk), nlk->portid); |
11b58ba1 | 3304 | return jhash2((u32 *)&arg, netlink_compare_arg_len / sizeof(u32), seed); |
c428ecd1 HX |
3305 | } |
3306 | ||
3307 | static const struct rhashtable_params netlink_rhashtable_params = { | |
3308 | .head_offset = offsetof(struct netlink_sock, node), | |
3309 | .key_len = netlink_compare_arg_len, | |
c428ecd1 HX |
3310 | .obj_hashfn = netlink_hash, |
3311 | .obj_cmpfn = netlink_compare, | |
b5e2c150 | 3312 | .automatic_shrinking = true, |
c428ecd1 HX |
3313 | }; |
3314 | ||
1da177e4 LT |
3315 | static int __init netlink_proto_init(void) |
3316 | { | |
1da177e4 | 3317 | int i; |
1da177e4 LT |
3318 | int err = proto_register(&netlink_proto, 0); |
3319 | ||
3320 | if (err != 0) | |
3321 | goto out; | |
3322 | ||
fab25745 | 3323 | BUILD_BUG_ON(sizeof(struct netlink_skb_parms) > FIELD_SIZEOF(struct sk_buff, cb)); |
1da177e4 | 3324 | |
0da974f4 | 3325 | nl_table = kcalloc(MAX_LINKS, sizeof(*nl_table), GFP_KERNEL); |
fab2caf6 AM |
3326 | if (!nl_table) |
3327 | goto panic; | |
1da177e4 | 3328 | |
1da177e4 | 3329 | for (i = 0; i < MAX_LINKS; i++) { |
c428ecd1 HX |
3330 | if (rhashtable_init(&nl_table[i].hash, |
3331 | &netlink_rhashtable_params) < 0) { | |
e341694e TG |
3332 | while (--i > 0) |
3333 | rhashtable_destroy(&nl_table[i].hash); | |
1da177e4 | 3334 | kfree(nl_table); |
fab2caf6 | 3335 | goto panic; |
1da177e4 | 3336 | } |
1da177e4 LT |
3337 | } |
3338 | ||
bcbde0d4 DB |
3339 | INIT_LIST_HEAD(&netlink_tap_all); |
3340 | ||
b963ea89 DM |
3341 | netlink_add_usersock_entry(); |
3342 | ||
1da177e4 | 3343 | sock_register(&netlink_family_ops); |
b4b51029 | 3344 | register_pernet_subsys(&netlink_net_ops); |
746fac4d | 3345 | /* The netlink device handler may be needed early. */ |
1da177e4 LT |
3346 | rtnetlink_init(); |
3347 | out: | |
3348 | return err; | |
fab2caf6 AM |
3349 | panic: |
3350 | panic("netlink_init: Cannot allocate nl_table\n"); | |
1da177e4 LT |
3351 | } |
3352 | ||
1da177e4 | 3353 | core_initcall(netlink_proto_init); |