Commit | Line | Data |
---|---|---|
1da177e4 LT |
1 | /* |
2 | * NETLINK Kernel-user communication protocol. | |
3 | * | |
113aa838 | 4 | * Authors: Alan Cox <alan@lxorguk.ukuu.org.uk> |
1da177e4 | 5 | * Alexey Kuznetsov <kuznet@ms2.inr.ac.ru> |
cd1df525 | 6 | * Patrick McHardy <kaber@trash.net> |
1da177e4 LT |
7 | * |
8 | * This program is free software; you can redistribute it and/or | |
9 | * modify it under the terms of the GNU General Public License | |
10 | * as published by the Free Software Foundation; either version | |
11 | * 2 of the License, or (at your option) any later version. | |
746fac4d | 12 | * |
1da177e4 LT |
13 | * Tue Jun 26 14:36:48 MEST 2001 Herbert "herp" Rosmanith |
14 | * added netlink_proto_exit | |
15 | * Tue Jan 22 18:32:44 BRST 2002 Arnaldo C. de Melo <acme@conectiva.com.br> | |
16 | * use nlk_sk, as sk->protinfo is on a diet 8) | |
4fdb3bb7 HW |
17 | * Fri Jul 22 19:51:12 MEST 2005 Harald Welte <laforge@gnumonks.org> |
18 | * - inc module use count of module that owns | |
19 | * the kernel socket in case userspace opens | |
20 | * socket of same protocol | |
21 | * - remove all module support, since netlink is | |
22 | * mandatory if CONFIG_NET=y these days | |
1da177e4 LT |
23 | */ |
24 | ||
1da177e4 LT |
25 | #include <linux/module.h> |
26 | ||
4fc268d2 | 27 | #include <linux/capability.h> |
1da177e4 LT |
28 | #include <linux/kernel.h> |
29 | #include <linux/init.h> | |
1da177e4 LT |
30 | #include <linux/signal.h> |
31 | #include <linux/sched.h> | |
32 | #include <linux/errno.h> | |
33 | #include <linux/string.h> | |
34 | #include <linux/stat.h> | |
35 | #include <linux/socket.h> | |
36 | #include <linux/un.h> | |
37 | #include <linux/fcntl.h> | |
38 | #include <linux/termios.h> | |
39 | #include <linux/sockios.h> | |
40 | #include <linux/net.h> | |
41 | #include <linux/fs.h> | |
42 | #include <linux/slab.h> | |
43 | #include <asm/uaccess.h> | |
44 | #include <linux/skbuff.h> | |
45 | #include <linux/netdevice.h> | |
46 | #include <linux/rtnetlink.h> | |
47 | #include <linux/proc_fs.h> | |
48 | #include <linux/seq_file.h> | |
1da177e4 LT |
49 | #include <linux/notifier.h> |
50 | #include <linux/security.h> | |
51 | #include <linux/jhash.h> | |
52 | #include <linux/jiffies.h> | |
53 | #include <linux/random.h> | |
54 | #include <linux/bitops.h> | |
55 | #include <linux/mm.h> | |
56 | #include <linux/types.h> | |
54e0f520 | 57 | #include <linux/audit.h> |
af65bdfc | 58 | #include <linux/mutex.h> |
ccdfcc39 | 59 | #include <linux/vmalloc.h> |
bcbde0d4 | 60 | #include <linux/if_arp.h> |
9652e931 | 61 | #include <asm/cacheflush.h> |
54e0f520 | 62 | |
457c4cbc | 63 | #include <net/net_namespace.h> |
1da177e4 LT |
64 | #include <net/sock.h> |
65 | #include <net/scm.h> | |
82ace47a | 66 | #include <net/netlink.h> |
1da177e4 | 67 | |
0f29c768 | 68 | #include "af_netlink.h" |
1da177e4 | 69 | |
5c398dc8 ED |
70 | struct listeners { |
71 | struct rcu_head rcu; | |
72 | unsigned long masks[0]; | |
6c04bb18 JB |
73 | }; |
74 | ||
cd967e05 PM |
75 | /* state bits */ |
76 | #define NETLINK_CONGESTED 0x0 | |
77 | ||
78 | /* flags */ | |
77247bbb | 79 | #define NETLINK_KERNEL_SOCKET 0x1 |
9a4595bc | 80 | #define NETLINK_RECV_PKTINFO 0x2 |
be0c22a4 | 81 | #define NETLINK_BROADCAST_SEND_ERROR 0x4 |
38938bfe | 82 | #define NETLINK_RECV_NO_ENOBUFS 0x8 |
77247bbb | 83 | |
035c4c16 | 84 | static inline int netlink_is_kernel(struct sock *sk) |
aed81560 DL |
85 | { |
86 | return nlk_sk(sk)->flags & NETLINK_KERNEL_SOCKET; | |
87 | } | |
88 | ||
0f29c768 AV |
89 | struct netlink_table *nl_table; |
90 | EXPORT_SYMBOL_GPL(nl_table); | |
1da177e4 LT |
91 | |
92 | static DECLARE_WAIT_QUEUE_HEAD(nl_table_wait); | |
93 | ||
94 | static int netlink_dump(struct sock *sk); | |
9652e931 | 95 | static void netlink_skb_destructor(struct sk_buff *skb); |
1da177e4 | 96 | |
0f29c768 AV |
97 | DEFINE_RWLOCK(nl_table_lock); |
98 | EXPORT_SYMBOL_GPL(nl_table_lock); | |
1da177e4 LT |
99 | static atomic_t nl_table_users = ATOMIC_INIT(0); |
100 | ||
6d772ac5 ED |
101 | #define nl_deref_protected(X) rcu_dereference_protected(X, lockdep_is_held(&nl_table_lock)); |
102 | ||
e041c683 | 103 | static ATOMIC_NOTIFIER_HEAD(netlink_chain); |
1da177e4 | 104 | |
bcbde0d4 DB |
105 | static DEFINE_SPINLOCK(netlink_tap_lock); |
106 | static struct list_head netlink_tap_all __read_mostly; | |
107 | ||
b57ef81f | 108 | static inline u32 netlink_group_mask(u32 group) |
d629b836 PM |
109 | { |
110 | return group ? 1 << (group - 1) : 0; | |
111 | } | |
112 | ||
15e47304 | 113 | static inline struct hlist_head *nl_portid_hashfn(struct nl_portid_hash *hash, u32 portid) |
1da177e4 | 114 | { |
15e47304 | 115 | return &hash->table[jhash_1word(portid, hash->rnd) & hash->mask]; |
1da177e4 LT |
116 | } |
117 | ||
bcbde0d4 DB |
118 | int netlink_add_tap(struct netlink_tap *nt) |
119 | { | |
120 | if (unlikely(nt->dev->type != ARPHRD_NETLINK)) | |
121 | return -EINVAL; | |
122 | ||
123 | spin_lock(&netlink_tap_lock); | |
124 | list_add_rcu(&nt->list, &netlink_tap_all); | |
125 | spin_unlock(&netlink_tap_lock); | |
126 | ||
127 | if (nt->module) | |
128 | __module_get(nt->module); | |
129 | ||
130 | return 0; | |
131 | } | |
132 | EXPORT_SYMBOL_GPL(netlink_add_tap); | |
133 | ||
134 | int __netlink_remove_tap(struct netlink_tap *nt) | |
135 | { | |
136 | bool found = false; | |
137 | struct netlink_tap *tmp; | |
138 | ||
139 | spin_lock(&netlink_tap_lock); | |
140 | ||
141 | list_for_each_entry(tmp, &netlink_tap_all, list) { | |
142 | if (nt == tmp) { | |
143 | list_del_rcu(&nt->list); | |
144 | found = true; | |
145 | goto out; | |
146 | } | |
147 | } | |
148 | ||
149 | pr_warn("__netlink_remove_tap: %p not found\n", nt); | |
150 | out: | |
151 | spin_unlock(&netlink_tap_lock); | |
152 | ||
153 | if (found && nt->module) | |
154 | module_put(nt->module); | |
155 | ||
156 | return found ? 0 : -ENODEV; | |
157 | } | |
158 | EXPORT_SYMBOL_GPL(__netlink_remove_tap); | |
159 | ||
160 | int netlink_remove_tap(struct netlink_tap *nt) | |
161 | { | |
162 | int ret; | |
163 | ||
164 | ret = __netlink_remove_tap(nt); | |
165 | synchronize_net(); | |
166 | ||
167 | return ret; | |
168 | } | |
169 | EXPORT_SYMBOL_GPL(netlink_remove_tap); | |
170 | ||
5ffd5cdd DB |
171 | static bool netlink_filter_tap(const struct sk_buff *skb) |
172 | { | |
173 | struct sock *sk = skb->sk; | |
174 | bool pass = false; | |
175 | ||
176 | /* We take the more conservative approach and | |
177 | * whitelist socket protocols that may pass. | |
178 | */ | |
179 | switch (sk->sk_protocol) { | |
180 | case NETLINK_ROUTE: | |
181 | case NETLINK_USERSOCK: | |
182 | case NETLINK_SOCK_DIAG: | |
183 | case NETLINK_NFLOG: | |
184 | case NETLINK_XFRM: | |
185 | case NETLINK_FIB_LOOKUP: | |
186 | case NETLINK_NETFILTER: | |
187 | case NETLINK_GENERIC: | |
188 | pass = true; | |
189 | break; | |
190 | } | |
191 | ||
192 | return pass; | |
193 | } | |
194 | ||
bcbde0d4 DB |
195 | static int __netlink_deliver_tap_skb(struct sk_buff *skb, |
196 | struct net_device *dev) | |
197 | { | |
198 | struct sk_buff *nskb; | |
5ffd5cdd | 199 | struct sock *sk = skb->sk; |
bcbde0d4 DB |
200 | int ret = -ENOMEM; |
201 | ||
202 | dev_hold(dev); | |
203 | nskb = skb_clone(skb, GFP_ATOMIC); | |
204 | if (nskb) { | |
205 | nskb->dev = dev; | |
5ffd5cdd | 206 | nskb->protocol = htons((u16) sk->sk_protocol); |
604d13c9 DB |
207 | nskb->pkt_type = netlink_is_kernel(sk) ? |
208 | PACKET_KERNEL : PACKET_USER; | |
5ffd5cdd | 209 | |
bcbde0d4 DB |
210 | ret = dev_queue_xmit(nskb); |
211 | if (unlikely(ret > 0)) | |
212 | ret = net_xmit_errno(ret); | |
213 | } | |
214 | ||
215 | dev_put(dev); | |
216 | return ret; | |
217 | } | |
218 | ||
219 | static void __netlink_deliver_tap(struct sk_buff *skb) | |
220 | { | |
221 | int ret; | |
222 | struct netlink_tap *tmp; | |
223 | ||
5ffd5cdd DB |
224 | if (!netlink_filter_tap(skb)) |
225 | return; | |
226 | ||
bcbde0d4 DB |
227 | list_for_each_entry_rcu(tmp, &netlink_tap_all, list) { |
228 | ret = __netlink_deliver_tap_skb(skb, tmp->dev); | |
229 | if (unlikely(ret)) | |
230 | break; | |
231 | } | |
232 | } | |
233 | ||
234 | static void netlink_deliver_tap(struct sk_buff *skb) | |
235 | { | |
236 | rcu_read_lock(); | |
237 | ||
238 | if (unlikely(!list_empty(&netlink_tap_all))) | |
239 | __netlink_deliver_tap(skb); | |
240 | ||
241 | rcu_read_unlock(); | |
242 | } | |
243 | ||
73bfd370 DB |
244 | static void netlink_deliver_tap_kernel(struct sock *dst, struct sock *src, |
245 | struct sk_buff *skb) | |
246 | { | |
247 | if (!(netlink_is_kernel(dst) && netlink_is_kernel(src))) | |
248 | netlink_deliver_tap(skb); | |
249 | } | |
250 | ||
cd1df525 PM |
251 | static void netlink_overrun(struct sock *sk) |
252 | { | |
253 | struct netlink_sock *nlk = nlk_sk(sk); | |
254 | ||
255 | if (!(nlk->flags & NETLINK_RECV_NO_ENOBUFS)) { | |
256 | if (!test_and_set_bit(NETLINK_CONGESTED, &nlk_sk(sk)->state)) { | |
257 | sk->sk_err = ENOBUFS; | |
258 | sk->sk_error_report(sk); | |
259 | } | |
260 | } | |
261 | atomic_inc(&sk->sk_drops); | |
262 | } | |
263 | ||
264 | static void netlink_rcv_wake(struct sock *sk) | |
265 | { | |
266 | struct netlink_sock *nlk = nlk_sk(sk); | |
267 | ||
268 | if (skb_queue_empty(&sk->sk_receive_queue)) | |
269 | clear_bit(NETLINK_CONGESTED, &nlk->state); | |
270 | if (!test_bit(NETLINK_CONGESTED, &nlk->state)) | |
271 | wake_up_interruptible(&nlk->wait); | |
272 | } | |
273 | ||
ccdfcc39 | 274 | #ifdef CONFIG_NETLINK_MMAP |
9652e931 PM |
275 | static bool netlink_skb_is_mmaped(const struct sk_buff *skb) |
276 | { | |
277 | return NETLINK_CB(skb).flags & NETLINK_SKB_MMAPED; | |
278 | } | |
279 | ||
f9c22888 PM |
280 | static bool netlink_rx_is_mmaped(struct sock *sk) |
281 | { | |
282 | return nlk_sk(sk)->rx_ring.pg_vec != NULL; | |
283 | } | |
284 | ||
5fd96123 PM |
285 | static bool netlink_tx_is_mmaped(struct sock *sk) |
286 | { | |
287 | return nlk_sk(sk)->tx_ring.pg_vec != NULL; | |
288 | } | |
289 | ||
ccdfcc39 PM |
290 | static __pure struct page *pgvec_to_page(const void *addr) |
291 | { | |
292 | if (is_vmalloc_addr(addr)) | |
293 | return vmalloc_to_page(addr); | |
294 | else | |
295 | return virt_to_page(addr); | |
296 | } | |
297 | ||
298 | static void free_pg_vec(void **pg_vec, unsigned int order, unsigned int len) | |
299 | { | |
300 | unsigned int i; | |
301 | ||
302 | for (i = 0; i < len; i++) { | |
303 | if (pg_vec[i] != NULL) { | |
304 | if (is_vmalloc_addr(pg_vec[i])) | |
305 | vfree(pg_vec[i]); | |
306 | else | |
307 | free_pages((unsigned long)pg_vec[i], order); | |
308 | } | |
309 | } | |
310 | kfree(pg_vec); | |
311 | } | |
312 | ||
313 | static void *alloc_one_pg_vec_page(unsigned long order) | |
314 | { | |
315 | void *buffer; | |
316 | gfp_t gfp_flags = GFP_KERNEL | __GFP_COMP | __GFP_ZERO | | |
317 | __GFP_NOWARN | __GFP_NORETRY; | |
318 | ||
319 | buffer = (void *)__get_free_pages(gfp_flags, order); | |
320 | if (buffer != NULL) | |
321 | return buffer; | |
322 | ||
323 | buffer = vzalloc((1 << order) * PAGE_SIZE); | |
324 | if (buffer != NULL) | |
325 | return buffer; | |
326 | ||
327 | gfp_flags &= ~__GFP_NORETRY; | |
328 | return (void *)__get_free_pages(gfp_flags, order); | |
329 | } | |
330 | ||
331 | static void **alloc_pg_vec(struct netlink_sock *nlk, | |
332 | struct nl_mmap_req *req, unsigned int order) | |
333 | { | |
334 | unsigned int block_nr = req->nm_block_nr; | |
335 | unsigned int i; | |
8a849bb7 | 336 | void **pg_vec; |
ccdfcc39 PM |
337 | |
338 | pg_vec = kcalloc(block_nr, sizeof(void *), GFP_KERNEL); | |
339 | if (pg_vec == NULL) | |
340 | return NULL; | |
341 | ||
342 | for (i = 0; i < block_nr; i++) { | |
8a849bb7 | 343 | pg_vec[i] = alloc_one_pg_vec_page(order); |
ccdfcc39 PM |
344 | if (pg_vec[i] == NULL) |
345 | goto err1; | |
346 | } | |
347 | ||
348 | return pg_vec; | |
349 | err1: | |
350 | free_pg_vec(pg_vec, order, block_nr); | |
351 | return NULL; | |
352 | } | |
353 | ||
354 | static int netlink_set_ring(struct sock *sk, struct nl_mmap_req *req, | |
355 | bool closing, bool tx_ring) | |
356 | { | |
357 | struct netlink_sock *nlk = nlk_sk(sk); | |
358 | struct netlink_ring *ring; | |
359 | struct sk_buff_head *queue; | |
360 | void **pg_vec = NULL; | |
361 | unsigned int order = 0; | |
362 | int err; | |
363 | ||
364 | ring = tx_ring ? &nlk->tx_ring : &nlk->rx_ring; | |
365 | queue = tx_ring ? &sk->sk_write_queue : &sk->sk_receive_queue; | |
366 | ||
367 | if (!closing) { | |
368 | if (atomic_read(&nlk->mapped)) | |
369 | return -EBUSY; | |
370 | if (atomic_read(&ring->pending)) | |
371 | return -EBUSY; | |
372 | } | |
373 | ||
374 | if (req->nm_block_nr) { | |
375 | if (ring->pg_vec != NULL) | |
376 | return -EBUSY; | |
377 | ||
378 | if ((int)req->nm_block_size <= 0) | |
379 | return -EINVAL; | |
380 | if (!IS_ALIGNED(req->nm_block_size, PAGE_SIZE)) | |
381 | return -EINVAL; | |
382 | if (req->nm_frame_size < NL_MMAP_HDRLEN) | |
383 | return -EINVAL; | |
384 | if (!IS_ALIGNED(req->nm_frame_size, NL_MMAP_MSG_ALIGNMENT)) | |
385 | return -EINVAL; | |
386 | ||
387 | ring->frames_per_block = req->nm_block_size / | |
388 | req->nm_frame_size; | |
389 | if (ring->frames_per_block == 0) | |
390 | return -EINVAL; | |
391 | if (ring->frames_per_block * req->nm_block_nr != | |
392 | req->nm_frame_nr) | |
393 | return -EINVAL; | |
394 | ||
395 | order = get_order(req->nm_block_size); | |
396 | pg_vec = alloc_pg_vec(nlk, req, order); | |
397 | if (pg_vec == NULL) | |
398 | return -ENOMEM; | |
399 | } else { | |
400 | if (req->nm_frame_nr) | |
401 | return -EINVAL; | |
402 | } | |
403 | ||
404 | err = -EBUSY; | |
405 | mutex_lock(&nlk->pg_vec_lock); | |
406 | if (closing || atomic_read(&nlk->mapped) == 0) { | |
407 | err = 0; | |
408 | spin_lock_bh(&queue->lock); | |
409 | ||
410 | ring->frame_max = req->nm_frame_nr - 1; | |
411 | ring->head = 0; | |
412 | ring->frame_size = req->nm_frame_size; | |
413 | ring->pg_vec_pages = req->nm_block_size / PAGE_SIZE; | |
414 | ||
415 | swap(ring->pg_vec_len, req->nm_block_nr); | |
416 | swap(ring->pg_vec_order, order); | |
417 | swap(ring->pg_vec, pg_vec); | |
418 | ||
419 | __skb_queue_purge(queue); | |
420 | spin_unlock_bh(&queue->lock); | |
421 | ||
422 | WARN_ON(atomic_read(&nlk->mapped)); | |
423 | } | |
424 | mutex_unlock(&nlk->pg_vec_lock); | |
425 | ||
426 | if (pg_vec) | |
427 | free_pg_vec(pg_vec, order, req->nm_block_nr); | |
428 | return err; | |
429 | } | |
430 | ||
431 | static void netlink_mm_open(struct vm_area_struct *vma) | |
432 | { | |
433 | struct file *file = vma->vm_file; | |
434 | struct socket *sock = file->private_data; | |
435 | struct sock *sk = sock->sk; | |
436 | ||
437 | if (sk) | |
438 | atomic_inc(&nlk_sk(sk)->mapped); | |
439 | } | |
440 | ||
441 | static void netlink_mm_close(struct vm_area_struct *vma) | |
442 | { | |
443 | struct file *file = vma->vm_file; | |
444 | struct socket *sock = file->private_data; | |
445 | struct sock *sk = sock->sk; | |
446 | ||
447 | if (sk) | |
448 | atomic_dec(&nlk_sk(sk)->mapped); | |
449 | } | |
450 | ||
451 | static const struct vm_operations_struct netlink_mmap_ops = { | |
452 | .open = netlink_mm_open, | |
453 | .close = netlink_mm_close, | |
454 | }; | |
455 | ||
456 | static int netlink_mmap(struct file *file, struct socket *sock, | |
457 | struct vm_area_struct *vma) | |
458 | { | |
459 | struct sock *sk = sock->sk; | |
460 | struct netlink_sock *nlk = nlk_sk(sk); | |
461 | struct netlink_ring *ring; | |
462 | unsigned long start, size, expected; | |
463 | unsigned int i; | |
464 | int err = -EINVAL; | |
465 | ||
466 | if (vma->vm_pgoff) | |
467 | return -EINVAL; | |
468 | ||
469 | mutex_lock(&nlk->pg_vec_lock); | |
470 | ||
471 | expected = 0; | |
472 | for (ring = &nlk->rx_ring; ring <= &nlk->tx_ring; ring++) { | |
473 | if (ring->pg_vec == NULL) | |
474 | continue; | |
475 | expected += ring->pg_vec_len * ring->pg_vec_pages * PAGE_SIZE; | |
476 | } | |
477 | ||
478 | if (expected == 0) | |
479 | goto out; | |
480 | ||
481 | size = vma->vm_end - vma->vm_start; | |
482 | if (size != expected) | |
483 | goto out; | |
484 | ||
485 | start = vma->vm_start; | |
486 | for (ring = &nlk->rx_ring; ring <= &nlk->tx_ring; ring++) { | |
487 | if (ring->pg_vec == NULL) | |
488 | continue; | |
489 | ||
490 | for (i = 0; i < ring->pg_vec_len; i++) { | |
491 | struct page *page; | |
492 | void *kaddr = ring->pg_vec[i]; | |
493 | unsigned int pg_num; | |
494 | ||
495 | for (pg_num = 0; pg_num < ring->pg_vec_pages; pg_num++) { | |
496 | page = pgvec_to_page(kaddr); | |
497 | err = vm_insert_page(vma, start, page); | |
498 | if (err < 0) | |
499 | goto out; | |
500 | start += PAGE_SIZE; | |
501 | kaddr += PAGE_SIZE; | |
502 | } | |
503 | } | |
504 | } | |
505 | ||
506 | atomic_inc(&nlk->mapped); | |
507 | vma->vm_ops = &netlink_mmap_ops; | |
508 | err = 0; | |
509 | out: | |
510 | mutex_unlock(&nlk->pg_vec_lock); | |
7cdbac71 | 511 | return err; |
ccdfcc39 | 512 | } |
9652e931 PM |
513 | |
514 | static void netlink_frame_flush_dcache(const struct nl_mmap_hdr *hdr) | |
515 | { | |
516 | #if ARCH_IMPLEMENTS_FLUSH_DCACHE_PAGE == 1 | |
517 | struct page *p_start, *p_end; | |
518 | ||
519 | /* First page is flushed through netlink_{get,set}_status */ | |
520 | p_start = pgvec_to_page(hdr + PAGE_SIZE); | |
1d5085cb | 521 | p_end = pgvec_to_page((void *)hdr + NL_MMAP_HDRLEN + hdr->nm_len - 1); |
9652e931 PM |
522 | while (p_start <= p_end) { |
523 | flush_dcache_page(p_start); | |
524 | p_start++; | |
525 | } | |
526 | #endif | |
527 | } | |
528 | ||
529 | static enum nl_mmap_status netlink_get_status(const struct nl_mmap_hdr *hdr) | |
530 | { | |
531 | smp_rmb(); | |
532 | flush_dcache_page(pgvec_to_page(hdr)); | |
533 | return hdr->nm_status; | |
534 | } | |
535 | ||
536 | static void netlink_set_status(struct nl_mmap_hdr *hdr, | |
537 | enum nl_mmap_status status) | |
538 | { | |
539 | hdr->nm_status = status; | |
540 | flush_dcache_page(pgvec_to_page(hdr)); | |
541 | smp_wmb(); | |
542 | } | |
543 | ||
544 | static struct nl_mmap_hdr * | |
545 | __netlink_lookup_frame(const struct netlink_ring *ring, unsigned int pos) | |
546 | { | |
547 | unsigned int pg_vec_pos, frame_off; | |
548 | ||
549 | pg_vec_pos = pos / ring->frames_per_block; | |
550 | frame_off = pos % ring->frames_per_block; | |
551 | ||
552 | return ring->pg_vec[pg_vec_pos] + (frame_off * ring->frame_size); | |
553 | } | |
554 | ||
555 | static struct nl_mmap_hdr * | |
556 | netlink_lookup_frame(const struct netlink_ring *ring, unsigned int pos, | |
557 | enum nl_mmap_status status) | |
558 | { | |
559 | struct nl_mmap_hdr *hdr; | |
560 | ||
561 | hdr = __netlink_lookup_frame(ring, pos); | |
562 | if (netlink_get_status(hdr) != status) | |
563 | return NULL; | |
564 | ||
565 | return hdr; | |
566 | } | |
567 | ||
568 | static struct nl_mmap_hdr * | |
569 | netlink_current_frame(const struct netlink_ring *ring, | |
570 | enum nl_mmap_status status) | |
571 | { | |
572 | return netlink_lookup_frame(ring, ring->head, status); | |
573 | } | |
574 | ||
575 | static struct nl_mmap_hdr * | |
576 | netlink_previous_frame(const struct netlink_ring *ring, | |
577 | enum nl_mmap_status status) | |
578 | { | |
579 | unsigned int prev; | |
580 | ||
581 | prev = ring->head ? ring->head - 1 : ring->frame_max; | |
582 | return netlink_lookup_frame(ring, prev, status); | |
583 | } | |
584 | ||
585 | static void netlink_increment_head(struct netlink_ring *ring) | |
586 | { | |
587 | ring->head = ring->head != ring->frame_max ? ring->head + 1 : 0; | |
588 | } | |
589 | ||
590 | static void netlink_forward_ring(struct netlink_ring *ring) | |
591 | { | |
592 | unsigned int head = ring->head, pos = head; | |
593 | const struct nl_mmap_hdr *hdr; | |
594 | ||
595 | do { | |
596 | hdr = __netlink_lookup_frame(ring, pos); | |
597 | if (hdr->nm_status == NL_MMAP_STATUS_UNUSED) | |
598 | break; | |
599 | if (hdr->nm_status != NL_MMAP_STATUS_SKIP) | |
600 | break; | |
601 | netlink_increment_head(ring); | |
602 | } while (ring->head != head); | |
603 | } | |
604 | ||
cd1df525 PM |
605 | static bool netlink_dump_space(struct netlink_sock *nlk) |
606 | { | |
607 | struct netlink_ring *ring = &nlk->rx_ring; | |
608 | struct nl_mmap_hdr *hdr; | |
609 | unsigned int n; | |
610 | ||
611 | hdr = netlink_current_frame(ring, NL_MMAP_STATUS_UNUSED); | |
612 | if (hdr == NULL) | |
613 | return false; | |
614 | ||
615 | n = ring->head + ring->frame_max / 2; | |
616 | if (n > ring->frame_max) | |
617 | n -= ring->frame_max; | |
618 | ||
619 | hdr = __netlink_lookup_frame(ring, n); | |
620 | ||
621 | return hdr->nm_status == NL_MMAP_STATUS_UNUSED; | |
622 | } | |
623 | ||
9652e931 PM |
624 | static unsigned int netlink_poll(struct file *file, struct socket *sock, |
625 | poll_table *wait) | |
626 | { | |
627 | struct sock *sk = sock->sk; | |
628 | struct netlink_sock *nlk = nlk_sk(sk); | |
629 | unsigned int mask; | |
cd1df525 | 630 | int err; |
9652e931 | 631 | |
cd1df525 PM |
632 | if (nlk->rx_ring.pg_vec != NULL) { |
633 | /* Memory mapped sockets don't call recvmsg(), so flow control | |
634 | * for dumps is performed here. A dump is allowed to continue | |
635 | * if at least half the ring is unused. | |
636 | */ | |
16b304f3 | 637 | while (nlk->cb_running && netlink_dump_space(nlk)) { |
cd1df525 PM |
638 | err = netlink_dump(sk); |
639 | if (err < 0) { | |
640 | sk->sk_err = err; | |
641 | sk->sk_error_report(sk); | |
642 | break; | |
643 | } | |
644 | } | |
645 | netlink_rcv_wake(sk); | |
646 | } | |
5fd96123 | 647 | |
9652e931 PM |
648 | mask = datagram_poll(file, sock, wait); |
649 | ||
650 | spin_lock_bh(&sk->sk_receive_queue.lock); | |
651 | if (nlk->rx_ring.pg_vec) { | |
652 | netlink_forward_ring(&nlk->rx_ring); | |
653 | if (!netlink_previous_frame(&nlk->rx_ring, NL_MMAP_STATUS_UNUSED)) | |
654 | mask |= POLLIN | POLLRDNORM; | |
655 | } | |
656 | spin_unlock_bh(&sk->sk_receive_queue.lock); | |
657 | ||
658 | spin_lock_bh(&sk->sk_write_queue.lock); | |
659 | if (nlk->tx_ring.pg_vec) { | |
660 | if (netlink_current_frame(&nlk->tx_ring, NL_MMAP_STATUS_UNUSED)) | |
661 | mask |= POLLOUT | POLLWRNORM; | |
662 | } | |
663 | spin_unlock_bh(&sk->sk_write_queue.lock); | |
664 | ||
665 | return mask; | |
666 | } | |
667 | ||
668 | static struct nl_mmap_hdr *netlink_mmap_hdr(struct sk_buff *skb) | |
669 | { | |
670 | return (struct nl_mmap_hdr *)(skb->head - NL_MMAP_HDRLEN); | |
671 | } | |
672 | ||
673 | static void netlink_ring_setup_skb(struct sk_buff *skb, struct sock *sk, | |
674 | struct netlink_ring *ring, | |
675 | struct nl_mmap_hdr *hdr) | |
676 | { | |
677 | unsigned int size; | |
678 | void *data; | |
679 | ||
680 | size = ring->frame_size - NL_MMAP_HDRLEN; | |
681 | data = (void *)hdr + NL_MMAP_HDRLEN; | |
682 | ||
683 | skb->head = data; | |
684 | skb->data = data; | |
685 | skb_reset_tail_pointer(skb); | |
686 | skb->end = skb->tail + size; | |
687 | skb->len = 0; | |
688 | ||
689 | skb->destructor = netlink_skb_destructor; | |
690 | NETLINK_CB(skb).flags |= NETLINK_SKB_MMAPED; | |
691 | NETLINK_CB(skb).sk = sk; | |
692 | } | |
5fd96123 PM |
693 | |
694 | static int netlink_mmap_sendmsg(struct sock *sk, struct msghdr *msg, | |
695 | u32 dst_portid, u32 dst_group, | |
696 | struct sock_iocb *siocb) | |
697 | { | |
698 | struct netlink_sock *nlk = nlk_sk(sk); | |
699 | struct netlink_ring *ring; | |
700 | struct nl_mmap_hdr *hdr; | |
701 | struct sk_buff *skb; | |
702 | unsigned int maxlen; | |
703 | bool excl = true; | |
704 | int err = 0, len = 0; | |
705 | ||
706 | /* Netlink messages are validated by the receiver before processing. | |
707 | * In order to avoid userspace changing the contents of the message | |
708 | * after validation, the socket and the ring may only be used by a | |
709 | * single process, otherwise we fall back to copying. | |
710 | */ | |
711 | if (atomic_long_read(&sk->sk_socket->file->f_count) > 2 || | |
712 | atomic_read(&nlk->mapped) > 1) | |
713 | excl = false; | |
714 | ||
715 | mutex_lock(&nlk->pg_vec_lock); | |
716 | ||
717 | ring = &nlk->tx_ring; | |
718 | maxlen = ring->frame_size - NL_MMAP_HDRLEN; | |
719 | ||
720 | do { | |
721 | hdr = netlink_current_frame(ring, NL_MMAP_STATUS_VALID); | |
722 | if (hdr == NULL) { | |
723 | if (!(msg->msg_flags & MSG_DONTWAIT) && | |
724 | atomic_read(&nlk->tx_ring.pending)) | |
725 | schedule(); | |
726 | continue; | |
727 | } | |
728 | if (hdr->nm_len > maxlen) { | |
729 | err = -EINVAL; | |
730 | goto out; | |
731 | } | |
732 | ||
733 | netlink_frame_flush_dcache(hdr); | |
734 | ||
735 | if (likely(dst_portid == 0 && dst_group == 0 && excl)) { | |
736 | skb = alloc_skb_head(GFP_KERNEL); | |
737 | if (skb == NULL) { | |
738 | err = -ENOBUFS; | |
739 | goto out; | |
740 | } | |
741 | sock_hold(sk); | |
742 | netlink_ring_setup_skb(skb, sk, ring, hdr); | |
743 | NETLINK_CB(skb).flags |= NETLINK_SKB_TX; | |
744 | __skb_put(skb, hdr->nm_len); | |
745 | netlink_set_status(hdr, NL_MMAP_STATUS_RESERVED); | |
746 | atomic_inc(&ring->pending); | |
747 | } else { | |
748 | skb = alloc_skb(hdr->nm_len, GFP_KERNEL); | |
749 | if (skb == NULL) { | |
750 | err = -ENOBUFS; | |
751 | goto out; | |
752 | } | |
753 | __skb_put(skb, hdr->nm_len); | |
754 | memcpy(skb->data, (void *)hdr + NL_MMAP_HDRLEN, hdr->nm_len); | |
755 | netlink_set_status(hdr, NL_MMAP_STATUS_UNUSED); | |
756 | } | |
757 | ||
758 | netlink_increment_head(ring); | |
759 | ||
760 | NETLINK_CB(skb).portid = nlk->portid; | |
761 | NETLINK_CB(skb).dst_group = dst_group; | |
762 | NETLINK_CB(skb).creds = siocb->scm->creds; | |
763 | ||
764 | err = security_netlink_send(sk, skb); | |
765 | if (err) { | |
766 | kfree_skb(skb); | |
767 | goto out; | |
768 | } | |
769 | ||
770 | if (unlikely(dst_group)) { | |
771 | atomic_inc(&skb->users); | |
772 | netlink_broadcast(sk, skb, dst_portid, dst_group, | |
773 | GFP_KERNEL); | |
774 | } | |
775 | err = netlink_unicast(sk, skb, dst_portid, | |
776 | msg->msg_flags & MSG_DONTWAIT); | |
777 | if (err < 0) | |
778 | goto out; | |
779 | len += err; | |
780 | ||
781 | } while (hdr != NULL || | |
782 | (!(msg->msg_flags & MSG_DONTWAIT) && | |
783 | atomic_read(&nlk->tx_ring.pending))); | |
784 | ||
785 | if (len > 0) | |
786 | err = len; | |
787 | out: | |
788 | mutex_unlock(&nlk->pg_vec_lock); | |
789 | return err; | |
790 | } | |
f9c22888 PM |
791 | |
792 | static void netlink_queue_mmaped_skb(struct sock *sk, struct sk_buff *skb) | |
793 | { | |
794 | struct nl_mmap_hdr *hdr; | |
795 | ||
796 | hdr = netlink_mmap_hdr(skb); | |
797 | hdr->nm_len = skb->len; | |
798 | hdr->nm_group = NETLINK_CB(skb).dst_group; | |
799 | hdr->nm_pid = NETLINK_CB(skb).creds.pid; | |
1bf9310a ND |
800 | hdr->nm_uid = from_kuid(sk_user_ns(sk), NETLINK_CB(skb).creds.uid); |
801 | hdr->nm_gid = from_kgid(sk_user_ns(sk), NETLINK_CB(skb).creds.gid); | |
f9c22888 PM |
802 | netlink_frame_flush_dcache(hdr); |
803 | netlink_set_status(hdr, NL_MMAP_STATUS_VALID); | |
804 | ||
805 | NETLINK_CB(skb).flags |= NETLINK_SKB_DELIVERED; | |
806 | kfree_skb(skb); | |
807 | } | |
808 | ||
809 | static void netlink_ring_set_copied(struct sock *sk, struct sk_buff *skb) | |
810 | { | |
811 | struct netlink_sock *nlk = nlk_sk(sk); | |
812 | struct netlink_ring *ring = &nlk->rx_ring; | |
813 | struct nl_mmap_hdr *hdr; | |
814 | ||
815 | spin_lock_bh(&sk->sk_receive_queue.lock); | |
816 | hdr = netlink_current_frame(ring, NL_MMAP_STATUS_UNUSED); | |
817 | if (hdr == NULL) { | |
818 | spin_unlock_bh(&sk->sk_receive_queue.lock); | |
819 | kfree_skb(skb); | |
cd1df525 | 820 | netlink_overrun(sk); |
f9c22888 PM |
821 | return; |
822 | } | |
823 | netlink_increment_head(ring); | |
824 | __skb_queue_tail(&sk->sk_receive_queue, skb); | |
825 | spin_unlock_bh(&sk->sk_receive_queue.lock); | |
826 | ||
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); | |
f9c22888 PM |
832 | netlink_set_status(hdr, NL_MMAP_STATUS_COPY); |
833 | } | |
834 | ||
ccdfcc39 | 835 | #else /* CONFIG_NETLINK_MMAP */ |
9652e931 | 836 | #define netlink_skb_is_mmaped(skb) false |
f9c22888 | 837 | #define netlink_rx_is_mmaped(sk) false |
5fd96123 | 838 | #define netlink_tx_is_mmaped(sk) false |
ccdfcc39 | 839 | #define netlink_mmap sock_no_mmap |
9652e931 | 840 | #define netlink_poll datagram_poll |
5fd96123 | 841 | #define netlink_mmap_sendmsg(sk, msg, dst_portid, dst_group, siocb) 0 |
ccdfcc39 PM |
842 | #endif /* CONFIG_NETLINK_MMAP */ |
843 | ||
cf0a018a PM |
844 | static void netlink_skb_destructor(struct sk_buff *skb) |
845 | { | |
9652e931 PM |
846 | #ifdef CONFIG_NETLINK_MMAP |
847 | struct nl_mmap_hdr *hdr; | |
848 | struct netlink_ring *ring; | |
849 | struct sock *sk; | |
850 | ||
851 | /* If a packet from the kernel to userspace was freed because of an | |
852 | * error without being delivered to userspace, the kernel must reset | |
853 | * the status. In the direction userspace to kernel, the status is | |
854 | * always reset here after the packet was processed and freed. | |
855 | */ | |
856 | if (netlink_skb_is_mmaped(skb)) { | |
857 | hdr = netlink_mmap_hdr(skb); | |
858 | sk = NETLINK_CB(skb).sk; | |
859 | ||
5fd96123 PM |
860 | if (NETLINK_CB(skb).flags & NETLINK_SKB_TX) { |
861 | netlink_set_status(hdr, NL_MMAP_STATUS_UNUSED); | |
862 | ring = &nlk_sk(sk)->tx_ring; | |
863 | } else { | |
864 | if (!(NETLINK_CB(skb).flags & NETLINK_SKB_DELIVERED)) { | |
865 | hdr->nm_len = 0; | |
866 | netlink_set_status(hdr, NL_MMAP_STATUS_VALID); | |
867 | } | |
868 | ring = &nlk_sk(sk)->rx_ring; | |
9652e931 | 869 | } |
9652e931 PM |
870 | |
871 | WARN_ON(atomic_read(&ring->pending) == 0); | |
872 | atomic_dec(&ring->pending); | |
873 | sock_put(sk); | |
874 | ||
5e71d9d7 | 875 | skb->head = NULL; |
9652e931 PM |
876 | } |
877 | #endif | |
c05cdb1b | 878 | if (is_vmalloc_addr(skb->head)) { |
3a36515f PN |
879 | if (!skb->cloned || |
880 | !atomic_dec_return(&(skb_shinfo(skb)->dataref))) | |
881 | vfree(skb->head); | |
882 | ||
c05cdb1b PNA |
883 | skb->head = NULL; |
884 | } | |
9652e931 PM |
885 | if (skb->sk != NULL) |
886 | sock_rfree(skb); | |
cf0a018a PM |
887 | } |
888 | ||
889 | static void netlink_skb_set_owner_r(struct sk_buff *skb, struct sock *sk) | |
890 | { | |
891 | WARN_ON(skb->sk != NULL); | |
892 | skb->sk = sk; | |
893 | skb->destructor = netlink_skb_destructor; | |
894 | atomic_add(skb->truesize, &sk->sk_rmem_alloc); | |
895 | sk_mem_charge(sk, skb->truesize); | |
896 | } | |
897 | ||
1da177e4 LT |
898 | static void netlink_sock_destruct(struct sock *sk) |
899 | { | |
3f660d66 HX |
900 | struct netlink_sock *nlk = nlk_sk(sk); |
901 | ||
16b304f3 PS |
902 | if (nlk->cb_running) { |
903 | if (nlk->cb.done) | |
904 | nlk->cb.done(&nlk->cb); | |
6dc878a8 | 905 | |
16b304f3 PS |
906 | module_put(nlk->cb.module); |
907 | kfree_skb(nlk->cb.skb); | |
3f660d66 HX |
908 | } |
909 | ||
1da177e4 | 910 | skb_queue_purge(&sk->sk_receive_queue); |
ccdfcc39 PM |
911 | #ifdef CONFIG_NETLINK_MMAP |
912 | if (1) { | |
913 | struct nl_mmap_req req; | |
914 | ||
915 | memset(&req, 0, sizeof(req)); | |
916 | if (nlk->rx_ring.pg_vec) | |
917 | netlink_set_ring(sk, &req, true, false); | |
918 | memset(&req, 0, sizeof(req)); | |
919 | if (nlk->tx_ring.pg_vec) | |
920 | netlink_set_ring(sk, &req, true, true); | |
921 | } | |
922 | #endif /* CONFIG_NETLINK_MMAP */ | |
1da177e4 LT |
923 | |
924 | if (!sock_flag(sk, SOCK_DEAD)) { | |
6ac552fd | 925 | printk(KERN_ERR "Freeing alive netlink socket %p\n", sk); |
1da177e4 LT |
926 | return; |
927 | } | |
547b792c IJ |
928 | |
929 | WARN_ON(atomic_read(&sk->sk_rmem_alloc)); | |
930 | WARN_ON(atomic_read(&sk->sk_wmem_alloc)); | |
931 | WARN_ON(nlk_sk(sk)->groups); | |
1da177e4 LT |
932 | } |
933 | ||
6ac552fd PM |
934 | /* This lock without WQ_FLAG_EXCLUSIVE is good on UP and it is _very_ bad on |
935 | * SMP. Look, when several writers sleep and reader wakes them up, all but one | |
1da177e4 LT |
936 | * immediately hit write lock and grab all the cpus. Exclusive sleep solves |
937 | * this, _but_ remember, it adds useless work on UP machines. | |
938 | */ | |
939 | ||
d136f1bd | 940 | void netlink_table_grab(void) |
9a429c49 | 941 | __acquires(nl_table_lock) |
1da177e4 | 942 | { |
d136f1bd JB |
943 | might_sleep(); |
944 | ||
6abd219c | 945 | write_lock_irq(&nl_table_lock); |
1da177e4 LT |
946 | |
947 | if (atomic_read(&nl_table_users)) { | |
948 | DECLARE_WAITQUEUE(wait, current); | |
949 | ||
950 | add_wait_queue_exclusive(&nl_table_wait, &wait); | |
6ac552fd | 951 | for (;;) { |
1da177e4 LT |
952 | set_current_state(TASK_UNINTERRUPTIBLE); |
953 | if (atomic_read(&nl_table_users) == 0) | |
954 | break; | |
6abd219c | 955 | write_unlock_irq(&nl_table_lock); |
1da177e4 | 956 | schedule(); |
6abd219c | 957 | write_lock_irq(&nl_table_lock); |
1da177e4 LT |
958 | } |
959 | ||
960 | __set_current_state(TASK_RUNNING); | |
961 | remove_wait_queue(&nl_table_wait, &wait); | |
962 | } | |
963 | } | |
964 | ||
d136f1bd | 965 | void netlink_table_ungrab(void) |
9a429c49 | 966 | __releases(nl_table_lock) |
1da177e4 | 967 | { |
6abd219c | 968 | write_unlock_irq(&nl_table_lock); |
1da177e4 LT |
969 | wake_up(&nl_table_wait); |
970 | } | |
971 | ||
6ac552fd | 972 | static inline void |
1da177e4 LT |
973 | netlink_lock_table(void) |
974 | { | |
975 | /* read_lock() synchronizes us to netlink_table_grab */ | |
976 | ||
977 | read_lock(&nl_table_lock); | |
978 | atomic_inc(&nl_table_users); | |
979 | read_unlock(&nl_table_lock); | |
980 | } | |
981 | ||
6ac552fd | 982 | static inline void |
1da177e4 LT |
983 | netlink_unlock_table(void) |
984 | { | |
985 | if (atomic_dec_and_test(&nl_table_users)) | |
986 | wake_up(&nl_table_wait); | |
987 | } | |
988 | ||
da12c90e G |
989 | static bool netlink_compare(struct net *net, struct sock *sk) |
990 | { | |
991 | return net_eq(sock_net(sk), net); | |
992 | } | |
993 | ||
15e47304 | 994 | static struct sock *netlink_lookup(struct net *net, int protocol, u32 portid) |
1da177e4 | 995 | { |
da12c90e G |
996 | struct netlink_table *table = &nl_table[protocol]; |
997 | struct nl_portid_hash *hash = &table->hash; | |
1da177e4 LT |
998 | struct hlist_head *head; |
999 | struct sock *sk; | |
1da177e4 LT |
1000 | |
1001 | read_lock(&nl_table_lock); | |
15e47304 | 1002 | head = nl_portid_hashfn(hash, portid); |
b67bfe0d | 1003 | sk_for_each(sk, head) { |
da12c90e G |
1004 | if (table->compare(net, sk) && |
1005 | (nlk_sk(sk)->portid == portid)) { | |
1da177e4 LT |
1006 | sock_hold(sk); |
1007 | goto found; | |
1008 | } | |
1009 | } | |
1010 | sk = NULL; | |
1011 | found: | |
1012 | read_unlock(&nl_table_lock); | |
1013 | return sk; | |
1014 | } | |
1015 | ||
15e47304 | 1016 | static struct hlist_head *nl_portid_hash_zalloc(size_t size) |
1da177e4 LT |
1017 | { |
1018 | if (size <= PAGE_SIZE) | |
ea72912c | 1019 | return kzalloc(size, GFP_ATOMIC); |
1da177e4 LT |
1020 | else |
1021 | return (struct hlist_head *) | |
ea72912c ED |
1022 | __get_free_pages(GFP_ATOMIC | __GFP_ZERO, |
1023 | get_order(size)); | |
1da177e4 LT |
1024 | } |
1025 | ||
15e47304 | 1026 | static void nl_portid_hash_free(struct hlist_head *table, size_t size) |
1da177e4 LT |
1027 | { |
1028 | if (size <= PAGE_SIZE) | |
1029 | kfree(table); | |
1030 | else | |
1031 | free_pages((unsigned long)table, get_order(size)); | |
1032 | } | |
1033 | ||
15e47304 | 1034 | static int nl_portid_hash_rehash(struct nl_portid_hash *hash, int grow) |
1da177e4 LT |
1035 | { |
1036 | unsigned int omask, mask, shift; | |
1037 | size_t osize, size; | |
1038 | struct hlist_head *otable, *table; | |
1039 | int i; | |
1040 | ||
1041 | omask = mask = hash->mask; | |
1042 | osize = size = (mask + 1) * sizeof(*table); | |
1043 | shift = hash->shift; | |
1044 | ||
1045 | if (grow) { | |
1046 | if (++shift > hash->max_shift) | |
1047 | return 0; | |
1048 | mask = mask * 2 + 1; | |
1049 | size *= 2; | |
1050 | } | |
1051 | ||
15e47304 | 1052 | table = nl_portid_hash_zalloc(size); |
1da177e4 LT |
1053 | if (!table) |
1054 | return 0; | |
1055 | ||
1da177e4 LT |
1056 | otable = hash->table; |
1057 | hash->table = table; | |
1058 | hash->mask = mask; | |
1059 | hash->shift = shift; | |
1060 | get_random_bytes(&hash->rnd, sizeof(hash->rnd)); | |
1061 | ||
1062 | for (i = 0; i <= omask; i++) { | |
1063 | struct sock *sk; | |
b67bfe0d | 1064 | struct hlist_node *tmp; |
1da177e4 | 1065 | |
b67bfe0d | 1066 | sk_for_each_safe(sk, tmp, &otable[i]) |
15e47304 | 1067 | __sk_add_node(sk, nl_portid_hashfn(hash, nlk_sk(sk)->portid)); |
1da177e4 LT |
1068 | } |
1069 | ||
15e47304 | 1070 | nl_portid_hash_free(otable, osize); |
1da177e4 LT |
1071 | hash->rehash_time = jiffies + 10 * 60 * HZ; |
1072 | return 1; | |
1073 | } | |
1074 | ||
15e47304 | 1075 | static inline int nl_portid_hash_dilute(struct nl_portid_hash *hash, int len) |
1da177e4 LT |
1076 | { |
1077 | int avg = hash->entries >> hash->shift; | |
1078 | ||
15e47304 | 1079 | if (unlikely(avg > 1) && nl_portid_hash_rehash(hash, 1)) |
1da177e4 LT |
1080 | return 1; |
1081 | ||
1082 | if (unlikely(len > avg) && time_after(jiffies, hash->rehash_time)) { | |
15e47304 | 1083 | nl_portid_hash_rehash(hash, 0); |
1da177e4 LT |
1084 | return 1; |
1085 | } | |
1086 | ||
1087 | return 0; | |
1088 | } | |
1089 | ||
90ddc4f0 | 1090 | static const struct proto_ops netlink_ops; |
1da177e4 | 1091 | |
4277a083 PM |
1092 | static void |
1093 | netlink_update_listeners(struct sock *sk) | |
1094 | { | |
1095 | struct netlink_table *tbl = &nl_table[sk->sk_protocol]; | |
4277a083 PM |
1096 | unsigned long mask; |
1097 | unsigned int i; | |
6d772ac5 ED |
1098 | struct listeners *listeners; |
1099 | ||
1100 | listeners = nl_deref_protected(tbl->listeners); | |
1101 | if (!listeners) | |
1102 | return; | |
4277a083 | 1103 | |
b4ff4f04 | 1104 | for (i = 0; i < NLGRPLONGS(tbl->groups); i++) { |
4277a083 | 1105 | mask = 0; |
b67bfe0d | 1106 | sk_for_each_bound(sk, &tbl->mc_list) { |
b4ff4f04 JB |
1107 | if (i < NLGRPLONGS(nlk_sk(sk)->ngroups)) |
1108 | mask |= nlk_sk(sk)->groups[i]; | |
1109 | } | |
6d772ac5 | 1110 | listeners->masks[i] = mask; |
4277a083 PM |
1111 | } |
1112 | /* this function is only called with the netlink table "grabbed", which | |
1113 | * makes sure updates are visible before bind or setsockopt return. */ | |
1114 | } | |
1115 | ||
15e47304 | 1116 | static int netlink_insert(struct sock *sk, struct net *net, u32 portid) |
1da177e4 | 1117 | { |
da12c90e G |
1118 | struct netlink_table *table = &nl_table[sk->sk_protocol]; |
1119 | struct nl_portid_hash *hash = &table->hash; | |
1da177e4 LT |
1120 | struct hlist_head *head; |
1121 | int err = -EADDRINUSE; | |
1122 | struct sock *osk; | |
1da177e4 LT |
1123 | int len; |
1124 | ||
1125 | netlink_table_grab(); | |
15e47304 | 1126 | head = nl_portid_hashfn(hash, portid); |
1da177e4 | 1127 | len = 0; |
b67bfe0d | 1128 | sk_for_each(osk, head) { |
da12c90e G |
1129 | if (table->compare(net, osk) && |
1130 | (nlk_sk(osk)->portid == portid)) | |
1da177e4 LT |
1131 | break; |
1132 | len++; | |
1133 | } | |
b67bfe0d | 1134 | if (osk) |
1da177e4 LT |
1135 | goto err; |
1136 | ||
1137 | err = -EBUSY; | |
15e47304 | 1138 | if (nlk_sk(sk)->portid) |
1da177e4 LT |
1139 | goto err; |
1140 | ||
1141 | err = -ENOMEM; | |
1142 | if (BITS_PER_LONG > 32 && unlikely(hash->entries >= UINT_MAX)) | |
1143 | goto err; | |
1144 | ||
15e47304 EB |
1145 | if (len && nl_portid_hash_dilute(hash, len)) |
1146 | head = nl_portid_hashfn(hash, portid); | |
1da177e4 | 1147 | hash->entries++; |
15e47304 | 1148 | nlk_sk(sk)->portid = portid; |
1da177e4 LT |
1149 | sk_add_node(sk, head); |
1150 | err = 0; | |
1151 | ||
1152 | err: | |
1153 | netlink_table_ungrab(); | |
1154 | return err; | |
1155 | } | |
1156 | ||
1157 | static void netlink_remove(struct sock *sk) | |
1158 | { | |
1159 | netlink_table_grab(); | |
d470e3b4 DM |
1160 | if (sk_del_node_init(sk)) |
1161 | nl_table[sk->sk_protocol].hash.entries--; | |
f7fa9b10 | 1162 | if (nlk_sk(sk)->subscriptions) |
1da177e4 LT |
1163 | __sk_del_bind_node(sk); |
1164 | netlink_table_ungrab(); | |
1165 | } | |
1166 | ||
1167 | static struct proto netlink_proto = { | |
1168 | .name = "NETLINK", | |
1169 | .owner = THIS_MODULE, | |
1170 | .obj_size = sizeof(struct netlink_sock), | |
1171 | }; | |
1172 | ||
1b8d7ae4 EB |
1173 | static int __netlink_create(struct net *net, struct socket *sock, |
1174 | struct mutex *cb_mutex, int protocol) | |
1da177e4 LT |
1175 | { |
1176 | struct sock *sk; | |
1177 | struct netlink_sock *nlk; | |
ab33a171 PM |
1178 | |
1179 | sock->ops = &netlink_ops; | |
1180 | ||
6257ff21 | 1181 | sk = sk_alloc(net, PF_NETLINK, GFP_KERNEL, &netlink_proto); |
ab33a171 PM |
1182 | if (!sk) |
1183 | return -ENOMEM; | |
1184 | ||
1185 | sock_init_data(sock, sk); | |
1186 | ||
1187 | nlk = nlk_sk(sk); | |
658cb354 | 1188 | if (cb_mutex) { |
ffa4d721 | 1189 | nlk->cb_mutex = cb_mutex; |
658cb354 | 1190 | } else { |
ffa4d721 PM |
1191 | nlk->cb_mutex = &nlk->cb_def_mutex; |
1192 | mutex_init(nlk->cb_mutex); | |
1193 | } | |
ab33a171 | 1194 | init_waitqueue_head(&nlk->wait); |
ccdfcc39 PM |
1195 | #ifdef CONFIG_NETLINK_MMAP |
1196 | mutex_init(&nlk->pg_vec_lock); | |
1197 | #endif | |
ab33a171 PM |
1198 | |
1199 | sk->sk_destruct = netlink_sock_destruct; | |
1200 | sk->sk_protocol = protocol; | |
1201 | return 0; | |
1202 | } | |
1203 | ||
3f378b68 EP |
1204 | static int netlink_create(struct net *net, struct socket *sock, int protocol, |
1205 | int kern) | |
ab33a171 PM |
1206 | { |
1207 | struct module *module = NULL; | |
af65bdfc | 1208 | struct mutex *cb_mutex; |
f7fa9b10 | 1209 | struct netlink_sock *nlk; |
03292745 | 1210 | void (*bind)(int group); |
ab33a171 | 1211 | int err = 0; |
1da177e4 LT |
1212 | |
1213 | sock->state = SS_UNCONNECTED; | |
1214 | ||
1215 | if (sock->type != SOCK_RAW && sock->type != SOCK_DGRAM) | |
1216 | return -ESOCKTNOSUPPORT; | |
1217 | ||
6ac552fd | 1218 | if (protocol < 0 || protocol >= MAX_LINKS) |
1da177e4 LT |
1219 | return -EPROTONOSUPPORT; |
1220 | ||
77247bbb | 1221 | netlink_lock_table(); |
95a5afca | 1222 | #ifdef CONFIG_MODULES |
ab33a171 | 1223 | if (!nl_table[protocol].registered) { |
77247bbb | 1224 | netlink_unlock_table(); |
4fdb3bb7 | 1225 | request_module("net-pf-%d-proto-%d", PF_NETLINK, protocol); |
77247bbb | 1226 | netlink_lock_table(); |
4fdb3bb7 | 1227 | } |
ab33a171 PM |
1228 | #endif |
1229 | if (nl_table[protocol].registered && | |
1230 | try_module_get(nl_table[protocol].module)) | |
1231 | module = nl_table[protocol].module; | |
974c37e9 AD |
1232 | else |
1233 | err = -EPROTONOSUPPORT; | |
af65bdfc | 1234 | cb_mutex = nl_table[protocol].cb_mutex; |
03292745 | 1235 | bind = nl_table[protocol].bind; |
77247bbb | 1236 | netlink_unlock_table(); |
4fdb3bb7 | 1237 | |
974c37e9 AD |
1238 | if (err < 0) |
1239 | goto out; | |
1240 | ||
6ac552fd PM |
1241 | err = __netlink_create(net, sock, cb_mutex, protocol); |
1242 | if (err < 0) | |
f7fa9b10 PM |
1243 | goto out_module; |
1244 | ||
6f756a8c | 1245 | local_bh_disable(); |
c1fd3b94 | 1246 | sock_prot_inuse_add(net, &netlink_proto, 1); |
6f756a8c DM |
1247 | local_bh_enable(); |
1248 | ||
f7fa9b10 | 1249 | nlk = nlk_sk(sock->sk); |
f7fa9b10 | 1250 | nlk->module = module; |
03292745 | 1251 | nlk->netlink_bind = bind; |
ab33a171 PM |
1252 | out: |
1253 | return err; | |
1da177e4 | 1254 | |
ab33a171 PM |
1255 | out_module: |
1256 | module_put(module); | |
1257 | goto out; | |
1da177e4 LT |
1258 | } |
1259 | ||
1260 | static int netlink_release(struct socket *sock) | |
1261 | { | |
1262 | struct sock *sk = sock->sk; | |
1263 | struct netlink_sock *nlk; | |
1264 | ||
1265 | if (!sk) | |
1266 | return 0; | |
1267 | ||
1268 | netlink_remove(sk); | |
ac57b3a9 | 1269 | sock_orphan(sk); |
1da177e4 LT |
1270 | nlk = nlk_sk(sk); |
1271 | ||
3f660d66 HX |
1272 | /* |
1273 | * OK. Socket is unlinked, any packets that arrive now | |
1274 | * will be purged. | |
1275 | */ | |
1da177e4 | 1276 | |
1da177e4 LT |
1277 | sock->sk = NULL; |
1278 | wake_up_interruptible_all(&nlk->wait); | |
1279 | ||
1280 | skb_queue_purge(&sk->sk_write_queue); | |
1281 | ||
15e47304 | 1282 | if (nlk->portid) { |
1da177e4 | 1283 | struct netlink_notify n = { |
3b1e0a65 | 1284 | .net = sock_net(sk), |
1da177e4 | 1285 | .protocol = sk->sk_protocol, |
15e47304 | 1286 | .portid = nlk->portid, |
1da177e4 | 1287 | }; |
e041c683 AS |
1288 | atomic_notifier_call_chain(&netlink_chain, |
1289 | NETLINK_URELEASE, &n); | |
746fac4d | 1290 | } |
4fdb3bb7 | 1291 | |
5e7c001c | 1292 | module_put(nlk->module); |
4fdb3bb7 | 1293 | |
4277a083 | 1294 | netlink_table_grab(); |
aed81560 | 1295 | if (netlink_is_kernel(sk)) { |
869e58f8 DL |
1296 | BUG_ON(nl_table[sk->sk_protocol].registered == 0); |
1297 | if (--nl_table[sk->sk_protocol].registered == 0) { | |
6d772ac5 ED |
1298 | struct listeners *old; |
1299 | ||
1300 | old = nl_deref_protected(nl_table[sk->sk_protocol].listeners); | |
1301 | RCU_INIT_POINTER(nl_table[sk->sk_protocol].listeners, NULL); | |
1302 | kfree_rcu(old, rcu); | |
869e58f8 | 1303 | nl_table[sk->sk_protocol].module = NULL; |
9785e10a PNA |
1304 | nl_table[sk->sk_protocol].bind = NULL; |
1305 | nl_table[sk->sk_protocol].flags = 0; | |
869e58f8 DL |
1306 | nl_table[sk->sk_protocol].registered = 0; |
1307 | } | |
658cb354 | 1308 | } else if (nlk->subscriptions) { |
4277a083 | 1309 | netlink_update_listeners(sk); |
658cb354 | 1310 | } |
4277a083 | 1311 | netlink_table_ungrab(); |
77247bbb | 1312 | |
f7fa9b10 PM |
1313 | kfree(nlk->groups); |
1314 | nlk->groups = NULL; | |
1315 | ||
3755810c | 1316 | local_bh_disable(); |
c1fd3b94 | 1317 | sock_prot_inuse_add(sock_net(sk), &netlink_proto, -1); |
3755810c | 1318 | local_bh_enable(); |
1da177e4 LT |
1319 | sock_put(sk); |
1320 | return 0; | |
1321 | } | |
1322 | ||
1323 | static int netlink_autobind(struct socket *sock) | |
1324 | { | |
1325 | struct sock *sk = sock->sk; | |
3b1e0a65 | 1326 | struct net *net = sock_net(sk); |
da12c90e G |
1327 | struct netlink_table *table = &nl_table[sk->sk_protocol]; |
1328 | struct nl_portid_hash *hash = &table->hash; | |
1da177e4 LT |
1329 | struct hlist_head *head; |
1330 | struct sock *osk; | |
15e47304 | 1331 | s32 portid = task_tgid_vnr(current); |
1da177e4 LT |
1332 | int err; |
1333 | static s32 rover = -4097; | |
1334 | ||
1335 | retry: | |
1336 | cond_resched(); | |
1337 | netlink_table_grab(); | |
15e47304 | 1338 | head = nl_portid_hashfn(hash, portid); |
b67bfe0d | 1339 | sk_for_each(osk, head) { |
da12c90e | 1340 | if (!table->compare(net, osk)) |
b4b51029 | 1341 | continue; |
15e47304 EB |
1342 | if (nlk_sk(osk)->portid == portid) { |
1343 | /* Bind collision, search negative portid values. */ | |
1344 | portid = rover--; | |
1da177e4 LT |
1345 | if (rover > -4097) |
1346 | rover = -4097; | |
1347 | netlink_table_ungrab(); | |
1348 | goto retry; | |
1349 | } | |
1350 | } | |
1351 | netlink_table_ungrab(); | |
1352 | ||
15e47304 | 1353 | err = netlink_insert(sk, net, portid); |
1da177e4 LT |
1354 | if (err == -EADDRINUSE) |
1355 | goto retry; | |
d470e3b4 DM |
1356 | |
1357 | /* If 2 threads race to autobind, that is fine. */ | |
1358 | if (err == -EBUSY) | |
1359 | err = 0; | |
1360 | ||
1361 | return err; | |
1da177e4 LT |
1362 | } |
1363 | ||
b57ef81f | 1364 | static inline int netlink_capable(const struct socket *sock, unsigned int flag) |
746fac4d | 1365 | { |
9785e10a | 1366 | return (nl_table[sock->sk->sk_protocol].flags & flag) || |
df008c91 | 1367 | ns_capable(sock_net(sock->sk)->user_ns, CAP_NET_ADMIN); |
746fac4d | 1368 | } |
1da177e4 | 1369 | |
f7fa9b10 PM |
1370 | static void |
1371 | netlink_update_subscriptions(struct sock *sk, unsigned int subscriptions) | |
1372 | { | |
1373 | struct netlink_sock *nlk = nlk_sk(sk); | |
1374 | ||
1375 | if (nlk->subscriptions && !subscriptions) | |
1376 | __sk_del_bind_node(sk); | |
1377 | else if (!nlk->subscriptions && subscriptions) | |
1378 | sk_add_bind_node(sk, &nl_table[sk->sk_protocol].mc_list); | |
1379 | nlk->subscriptions = subscriptions; | |
1380 | } | |
1381 | ||
b4ff4f04 | 1382 | static int netlink_realloc_groups(struct sock *sk) |
513c2500 PM |
1383 | { |
1384 | struct netlink_sock *nlk = nlk_sk(sk); | |
1385 | unsigned int groups; | |
b4ff4f04 | 1386 | unsigned long *new_groups; |
513c2500 PM |
1387 | int err = 0; |
1388 | ||
b4ff4f04 JB |
1389 | netlink_table_grab(); |
1390 | ||
513c2500 | 1391 | groups = nl_table[sk->sk_protocol].groups; |
b4ff4f04 | 1392 | if (!nl_table[sk->sk_protocol].registered) { |
513c2500 | 1393 | err = -ENOENT; |
b4ff4f04 JB |
1394 | goto out_unlock; |
1395 | } | |
513c2500 | 1396 | |
b4ff4f04 JB |
1397 | if (nlk->ngroups >= groups) |
1398 | goto out_unlock; | |
513c2500 | 1399 | |
b4ff4f04 JB |
1400 | new_groups = krealloc(nlk->groups, NLGRPSZ(groups), GFP_ATOMIC); |
1401 | if (new_groups == NULL) { | |
1402 | err = -ENOMEM; | |
1403 | goto out_unlock; | |
1404 | } | |
6ac552fd | 1405 | memset((char *)new_groups + NLGRPSZ(nlk->ngroups), 0, |
b4ff4f04 JB |
1406 | NLGRPSZ(groups) - NLGRPSZ(nlk->ngroups)); |
1407 | ||
1408 | nlk->groups = new_groups; | |
513c2500 | 1409 | nlk->ngroups = groups; |
b4ff4f04 JB |
1410 | out_unlock: |
1411 | netlink_table_ungrab(); | |
1412 | return err; | |
513c2500 PM |
1413 | } |
1414 | ||
6ac552fd PM |
1415 | static int netlink_bind(struct socket *sock, struct sockaddr *addr, |
1416 | int addr_len) | |
1da177e4 LT |
1417 | { |
1418 | struct sock *sk = sock->sk; | |
3b1e0a65 | 1419 | struct net *net = sock_net(sk); |
1da177e4 LT |
1420 | struct netlink_sock *nlk = nlk_sk(sk); |
1421 | struct sockaddr_nl *nladdr = (struct sockaddr_nl *)addr; | |
1422 | int err; | |
746fac4d | 1423 | |
4e4b5376 HFS |
1424 | if (addr_len < sizeof(struct sockaddr_nl)) |
1425 | return -EINVAL; | |
1426 | ||
1da177e4 LT |
1427 | if (nladdr->nl_family != AF_NETLINK) |
1428 | return -EINVAL; | |
1429 | ||
1430 | /* Only superuser is allowed to listen multicasts */ | |
513c2500 | 1431 | if (nladdr->nl_groups) { |
9785e10a | 1432 | if (!netlink_capable(sock, NL_CFG_F_NONROOT_RECV)) |
513c2500 | 1433 | return -EPERM; |
b4ff4f04 JB |
1434 | err = netlink_realloc_groups(sk); |
1435 | if (err) | |
1436 | return err; | |
513c2500 | 1437 | } |
1da177e4 | 1438 | |
15e47304 EB |
1439 | if (nlk->portid) { |
1440 | if (nladdr->nl_pid != nlk->portid) | |
1da177e4 LT |
1441 | return -EINVAL; |
1442 | } else { | |
1443 | err = nladdr->nl_pid ? | |
b4b51029 | 1444 | netlink_insert(sk, net, nladdr->nl_pid) : |
1da177e4 LT |
1445 | netlink_autobind(sock); |
1446 | if (err) | |
1447 | return err; | |
1448 | } | |
1449 | ||
513c2500 | 1450 | if (!nladdr->nl_groups && (nlk->groups == NULL || !(u32)nlk->groups[0])) |
1da177e4 LT |
1451 | return 0; |
1452 | ||
1453 | netlink_table_grab(); | |
f7fa9b10 | 1454 | netlink_update_subscriptions(sk, nlk->subscriptions + |
746fac4d YH |
1455 | hweight32(nladdr->nl_groups) - |
1456 | hweight32(nlk->groups[0])); | |
1457 | nlk->groups[0] = (nlk->groups[0] & ~0xffffffffUL) | nladdr->nl_groups; | |
4277a083 | 1458 | netlink_update_listeners(sk); |
1da177e4 LT |
1459 | netlink_table_ungrab(); |
1460 | ||
03292745 PNA |
1461 | if (nlk->netlink_bind && nlk->groups[0]) { |
1462 | int i; | |
1463 | ||
1464 | for (i=0; i<nlk->ngroups; i++) { | |
1465 | if (test_bit(i, nlk->groups)) | |
1466 | nlk->netlink_bind(i); | |
1467 | } | |
1468 | } | |
1469 | ||
1da177e4 LT |
1470 | return 0; |
1471 | } | |
1472 | ||
1473 | static int netlink_connect(struct socket *sock, struct sockaddr *addr, | |
1474 | int alen, int flags) | |
1475 | { | |
1476 | int err = 0; | |
1477 | struct sock *sk = sock->sk; | |
1478 | struct netlink_sock *nlk = nlk_sk(sk); | |
6ac552fd | 1479 | struct sockaddr_nl *nladdr = (struct sockaddr_nl *)addr; |
1da177e4 | 1480 | |
6503d961 CG |
1481 | if (alen < sizeof(addr->sa_family)) |
1482 | return -EINVAL; | |
1483 | ||
1da177e4 LT |
1484 | if (addr->sa_family == AF_UNSPEC) { |
1485 | sk->sk_state = NETLINK_UNCONNECTED; | |
15e47304 | 1486 | nlk->dst_portid = 0; |
d629b836 | 1487 | nlk->dst_group = 0; |
1da177e4 LT |
1488 | return 0; |
1489 | } | |
1490 | if (addr->sa_family != AF_NETLINK) | |
1491 | return -EINVAL; | |
1492 | ||
1493 | /* Only superuser is allowed to send multicasts */ | |
9785e10a | 1494 | if (nladdr->nl_groups && !netlink_capable(sock, NL_CFG_F_NONROOT_SEND)) |
1da177e4 LT |
1495 | return -EPERM; |
1496 | ||
15e47304 | 1497 | if (!nlk->portid) |
1da177e4 LT |
1498 | err = netlink_autobind(sock); |
1499 | ||
1500 | if (err == 0) { | |
1501 | sk->sk_state = NETLINK_CONNECTED; | |
15e47304 | 1502 | nlk->dst_portid = nladdr->nl_pid; |
d629b836 | 1503 | nlk->dst_group = ffs(nladdr->nl_groups); |
1da177e4 LT |
1504 | } |
1505 | ||
1506 | return err; | |
1507 | } | |
1508 | ||
6ac552fd PM |
1509 | static int netlink_getname(struct socket *sock, struct sockaddr *addr, |
1510 | int *addr_len, int peer) | |
1da177e4 LT |
1511 | { |
1512 | struct sock *sk = sock->sk; | |
1513 | struct netlink_sock *nlk = nlk_sk(sk); | |
13cfa97b | 1514 | DECLARE_SOCKADDR(struct sockaddr_nl *, nladdr, addr); |
746fac4d | 1515 | |
1da177e4 LT |
1516 | nladdr->nl_family = AF_NETLINK; |
1517 | nladdr->nl_pad = 0; | |
1518 | *addr_len = sizeof(*nladdr); | |
1519 | ||
1520 | if (peer) { | |
15e47304 | 1521 | nladdr->nl_pid = nlk->dst_portid; |
d629b836 | 1522 | nladdr->nl_groups = netlink_group_mask(nlk->dst_group); |
1da177e4 | 1523 | } else { |
15e47304 | 1524 | nladdr->nl_pid = nlk->portid; |
513c2500 | 1525 | nladdr->nl_groups = nlk->groups ? nlk->groups[0] : 0; |
1da177e4 LT |
1526 | } |
1527 | return 0; | |
1528 | } | |
1529 | ||
15e47304 | 1530 | static struct sock *netlink_getsockbyportid(struct sock *ssk, u32 portid) |
1da177e4 | 1531 | { |
1da177e4 LT |
1532 | struct sock *sock; |
1533 | struct netlink_sock *nlk; | |
1534 | ||
15e47304 | 1535 | sock = netlink_lookup(sock_net(ssk), ssk->sk_protocol, portid); |
1da177e4 LT |
1536 | if (!sock) |
1537 | return ERR_PTR(-ECONNREFUSED); | |
1538 | ||
1539 | /* Don't bother queuing skb if kernel socket has no input function */ | |
1540 | nlk = nlk_sk(sock); | |
cd40b7d3 | 1541 | if (sock->sk_state == NETLINK_CONNECTED && |
15e47304 | 1542 | nlk->dst_portid != nlk_sk(ssk)->portid) { |
1da177e4 LT |
1543 | sock_put(sock); |
1544 | return ERR_PTR(-ECONNREFUSED); | |
1545 | } | |
1546 | return sock; | |
1547 | } | |
1548 | ||
1549 | struct sock *netlink_getsockbyfilp(struct file *filp) | |
1550 | { | |
496ad9aa | 1551 | struct inode *inode = file_inode(filp); |
1da177e4 LT |
1552 | struct sock *sock; |
1553 | ||
1554 | if (!S_ISSOCK(inode->i_mode)) | |
1555 | return ERR_PTR(-ENOTSOCK); | |
1556 | ||
1557 | sock = SOCKET_I(inode)->sk; | |
1558 | if (sock->sk_family != AF_NETLINK) | |
1559 | return ERR_PTR(-EINVAL); | |
1560 | ||
1561 | sock_hold(sock); | |
1562 | return sock; | |
1563 | } | |
1564 | ||
3a36515f PN |
1565 | static struct sk_buff *netlink_alloc_large_skb(unsigned int size, |
1566 | int broadcast) | |
c05cdb1b PNA |
1567 | { |
1568 | struct sk_buff *skb; | |
1569 | void *data; | |
1570 | ||
3a36515f | 1571 | if (size <= NLMSG_GOODSIZE || broadcast) |
c05cdb1b PNA |
1572 | return alloc_skb(size, GFP_KERNEL); |
1573 | ||
3a36515f PN |
1574 | size = SKB_DATA_ALIGN(size) + |
1575 | SKB_DATA_ALIGN(sizeof(struct skb_shared_info)); | |
c05cdb1b PNA |
1576 | |
1577 | data = vmalloc(size); | |
1578 | if (data == NULL) | |
3a36515f | 1579 | return NULL; |
c05cdb1b | 1580 | |
3a36515f PN |
1581 | skb = build_skb(data, size); |
1582 | if (skb == NULL) | |
1583 | vfree(data); | |
1584 | else { | |
1585 | skb->head_frag = 0; | |
1586 | skb->destructor = netlink_skb_destructor; | |
1587 | } | |
c05cdb1b PNA |
1588 | |
1589 | return skb; | |
c05cdb1b PNA |
1590 | } |
1591 | ||
1da177e4 LT |
1592 | /* |
1593 | * Attach a skb to a netlink socket. | |
1594 | * The caller must hold a reference to the destination socket. On error, the | |
1595 | * reference is dropped. The skb is not send to the destination, just all | |
1596 | * all error checks are performed and memory in the queue is reserved. | |
1597 | * Return values: | |
1598 | * < 0: error. skb freed, reference to sock dropped. | |
1599 | * 0: continue | |
1600 | * 1: repeat lookup - reference dropped while waiting for socket memory. | |
1601 | */ | |
9457afee | 1602 | int netlink_attachskb(struct sock *sk, struct sk_buff *skb, |
c3d8d1e3 | 1603 | long *timeo, struct sock *ssk) |
1da177e4 LT |
1604 | { |
1605 | struct netlink_sock *nlk; | |
1606 | ||
1607 | nlk = nlk_sk(sk); | |
1608 | ||
5fd96123 PM |
1609 | if ((atomic_read(&sk->sk_rmem_alloc) > sk->sk_rcvbuf || |
1610 | test_bit(NETLINK_CONGESTED, &nlk->state)) && | |
1611 | !netlink_skb_is_mmaped(skb)) { | |
1da177e4 | 1612 | DECLARE_WAITQUEUE(wait, current); |
c3d8d1e3 | 1613 | if (!*timeo) { |
aed81560 | 1614 | if (!ssk || netlink_is_kernel(ssk)) |
1da177e4 LT |
1615 | netlink_overrun(sk); |
1616 | sock_put(sk); | |
1617 | kfree_skb(skb); | |
1618 | return -EAGAIN; | |
1619 | } | |
1620 | ||
1621 | __set_current_state(TASK_INTERRUPTIBLE); | |
1622 | add_wait_queue(&nlk->wait, &wait); | |
1623 | ||
1624 | if ((atomic_read(&sk->sk_rmem_alloc) > sk->sk_rcvbuf || | |
cd967e05 | 1625 | test_bit(NETLINK_CONGESTED, &nlk->state)) && |
1da177e4 | 1626 | !sock_flag(sk, SOCK_DEAD)) |
c3d8d1e3 | 1627 | *timeo = schedule_timeout(*timeo); |
1da177e4 LT |
1628 | |
1629 | __set_current_state(TASK_RUNNING); | |
1630 | remove_wait_queue(&nlk->wait, &wait); | |
1631 | sock_put(sk); | |
1632 | ||
1633 | if (signal_pending(current)) { | |
1634 | kfree_skb(skb); | |
c3d8d1e3 | 1635 | return sock_intr_errno(*timeo); |
1da177e4 LT |
1636 | } |
1637 | return 1; | |
1638 | } | |
cf0a018a | 1639 | netlink_skb_set_owner_r(skb, sk); |
1da177e4 LT |
1640 | return 0; |
1641 | } | |
1642 | ||
4a7e7c2a | 1643 | static int __netlink_sendskb(struct sock *sk, struct sk_buff *skb) |
1da177e4 | 1644 | { |
1da177e4 LT |
1645 | int len = skb->len; |
1646 | ||
bcbde0d4 DB |
1647 | netlink_deliver_tap(skb); |
1648 | ||
f9c22888 PM |
1649 | #ifdef CONFIG_NETLINK_MMAP |
1650 | if (netlink_skb_is_mmaped(skb)) | |
1651 | netlink_queue_mmaped_skb(sk, skb); | |
1652 | else if (netlink_rx_is_mmaped(sk)) | |
1653 | netlink_ring_set_copied(sk, skb); | |
1654 | else | |
1655 | #endif /* CONFIG_NETLINK_MMAP */ | |
1656 | skb_queue_tail(&sk->sk_receive_queue, skb); | |
1da177e4 | 1657 | sk->sk_data_ready(sk, len); |
4a7e7c2a ED |
1658 | return len; |
1659 | } | |
1660 | ||
1661 | int netlink_sendskb(struct sock *sk, struct sk_buff *skb) | |
1662 | { | |
1663 | int len = __netlink_sendskb(sk, skb); | |
1664 | ||
1da177e4 LT |
1665 | sock_put(sk); |
1666 | return len; | |
1667 | } | |
1668 | ||
1669 | void netlink_detachskb(struct sock *sk, struct sk_buff *skb) | |
1670 | { | |
1671 | kfree_skb(skb); | |
1672 | sock_put(sk); | |
1673 | } | |
1674 | ||
b57ef81f | 1675 | static struct sk_buff *netlink_trim(struct sk_buff *skb, gfp_t allocation) |
1da177e4 LT |
1676 | { |
1677 | int delta; | |
1678 | ||
1298ca46 | 1679 | WARN_ON(skb->sk != NULL); |
5fd96123 PM |
1680 | if (netlink_skb_is_mmaped(skb)) |
1681 | return skb; | |
1da177e4 | 1682 | |
4305b541 | 1683 | delta = skb->end - skb->tail; |
c05cdb1b | 1684 | if (is_vmalloc_addr(skb->head) || delta * 2 < skb->truesize) |
1da177e4 LT |
1685 | return skb; |
1686 | ||
1687 | if (skb_shared(skb)) { | |
1688 | struct sk_buff *nskb = skb_clone(skb, allocation); | |
1689 | if (!nskb) | |
1690 | return skb; | |
8460c00f | 1691 | consume_skb(skb); |
1da177e4 LT |
1692 | skb = nskb; |
1693 | } | |
1694 | ||
1695 | if (!pskb_expand_head(skb, 0, -delta, allocation)) | |
1696 | skb->truesize -= delta; | |
1697 | ||
1698 | return skb; | |
1699 | } | |
1700 | ||
3fbc2905 EB |
1701 | static int netlink_unicast_kernel(struct sock *sk, struct sk_buff *skb, |
1702 | struct sock *ssk) | |
cd40b7d3 DL |
1703 | { |
1704 | int ret; | |
1705 | struct netlink_sock *nlk = nlk_sk(sk); | |
1706 | ||
1707 | ret = -ECONNREFUSED; | |
1708 | if (nlk->netlink_rcv != NULL) { | |
1709 | ret = skb->len; | |
cf0a018a | 1710 | netlink_skb_set_owner_r(skb, sk); |
e32123e5 | 1711 | NETLINK_CB(skb).sk = ssk; |
73bfd370 | 1712 | netlink_deliver_tap_kernel(sk, ssk, skb); |
cd40b7d3 | 1713 | nlk->netlink_rcv(skb); |
bfb253c9 ED |
1714 | consume_skb(skb); |
1715 | } else { | |
1716 | kfree_skb(skb); | |
cd40b7d3 | 1717 | } |
cd40b7d3 DL |
1718 | sock_put(sk); |
1719 | return ret; | |
1720 | } | |
1721 | ||
1722 | int netlink_unicast(struct sock *ssk, struct sk_buff *skb, | |
15e47304 | 1723 | u32 portid, int nonblock) |
1da177e4 LT |
1724 | { |
1725 | struct sock *sk; | |
1726 | int err; | |
1727 | long timeo; | |
1728 | ||
1729 | skb = netlink_trim(skb, gfp_any()); | |
1730 | ||
1731 | timeo = sock_sndtimeo(ssk, nonblock); | |
1732 | retry: | |
15e47304 | 1733 | sk = netlink_getsockbyportid(ssk, portid); |
1da177e4 LT |
1734 | if (IS_ERR(sk)) { |
1735 | kfree_skb(skb); | |
1736 | return PTR_ERR(sk); | |
1737 | } | |
cd40b7d3 | 1738 | if (netlink_is_kernel(sk)) |
3fbc2905 | 1739 | return netlink_unicast_kernel(sk, skb, ssk); |
cd40b7d3 | 1740 | |
b1153f29 | 1741 | if (sk_filter(sk, skb)) { |
84874607 | 1742 | err = skb->len; |
b1153f29 SH |
1743 | kfree_skb(skb); |
1744 | sock_put(sk); | |
1745 | return err; | |
1746 | } | |
1747 | ||
9457afee | 1748 | err = netlink_attachskb(sk, skb, &timeo, ssk); |
1da177e4 LT |
1749 | if (err == 1) |
1750 | goto retry; | |
1751 | if (err) | |
1752 | return err; | |
1753 | ||
7ee015e0 | 1754 | return netlink_sendskb(sk, skb); |
1da177e4 | 1755 | } |
6ac552fd | 1756 | EXPORT_SYMBOL(netlink_unicast); |
1da177e4 | 1757 | |
f9c22888 PM |
1758 | struct sk_buff *netlink_alloc_skb(struct sock *ssk, unsigned int size, |
1759 | u32 dst_portid, gfp_t gfp_mask) | |
1760 | { | |
1761 | #ifdef CONFIG_NETLINK_MMAP | |
1762 | struct sock *sk = NULL; | |
1763 | struct sk_buff *skb; | |
1764 | struct netlink_ring *ring; | |
1765 | struct nl_mmap_hdr *hdr; | |
1766 | unsigned int maxlen; | |
1767 | ||
1768 | sk = netlink_getsockbyportid(ssk, dst_portid); | |
1769 | if (IS_ERR(sk)) | |
1770 | goto out; | |
1771 | ||
1772 | ring = &nlk_sk(sk)->rx_ring; | |
1773 | /* fast-path without atomic ops for common case: non-mmaped receiver */ | |
1774 | if (ring->pg_vec == NULL) | |
1775 | goto out_put; | |
1776 | ||
1777 | skb = alloc_skb_head(gfp_mask); | |
1778 | if (skb == NULL) | |
1779 | goto err1; | |
1780 | ||
1781 | spin_lock_bh(&sk->sk_receive_queue.lock); | |
1782 | /* check again under lock */ | |
1783 | if (ring->pg_vec == NULL) | |
1784 | goto out_free; | |
1785 | ||
1786 | maxlen = ring->frame_size - NL_MMAP_HDRLEN; | |
1787 | if (maxlen < size) | |
1788 | goto out_free; | |
1789 | ||
1790 | netlink_forward_ring(ring); | |
1791 | hdr = netlink_current_frame(ring, NL_MMAP_STATUS_UNUSED); | |
1792 | if (hdr == NULL) | |
1793 | goto err2; | |
1794 | netlink_ring_setup_skb(skb, sk, ring, hdr); | |
1795 | netlink_set_status(hdr, NL_MMAP_STATUS_RESERVED); | |
1796 | atomic_inc(&ring->pending); | |
1797 | netlink_increment_head(ring); | |
1798 | ||
1799 | spin_unlock_bh(&sk->sk_receive_queue.lock); | |
1800 | return skb; | |
1801 | ||
1802 | err2: | |
1803 | kfree_skb(skb); | |
1804 | spin_unlock_bh(&sk->sk_receive_queue.lock); | |
cd1df525 | 1805 | netlink_overrun(sk); |
f9c22888 PM |
1806 | err1: |
1807 | sock_put(sk); | |
1808 | return NULL; | |
1809 | ||
1810 | out_free: | |
1811 | kfree_skb(skb); | |
1812 | spin_unlock_bh(&sk->sk_receive_queue.lock); | |
1813 | out_put: | |
1814 | sock_put(sk); | |
1815 | out: | |
1816 | #endif | |
1817 | return alloc_skb(size, gfp_mask); | |
1818 | } | |
1819 | EXPORT_SYMBOL_GPL(netlink_alloc_skb); | |
1820 | ||
4277a083 PM |
1821 | int netlink_has_listeners(struct sock *sk, unsigned int group) |
1822 | { | |
1823 | int res = 0; | |
5c398dc8 | 1824 | struct listeners *listeners; |
4277a083 | 1825 | |
aed81560 | 1826 | BUG_ON(!netlink_is_kernel(sk)); |
b4ff4f04 JB |
1827 | |
1828 | rcu_read_lock(); | |
1829 | listeners = rcu_dereference(nl_table[sk->sk_protocol].listeners); | |
1830 | ||
6d772ac5 | 1831 | if (listeners && group - 1 < nl_table[sk->sk_protocol].groups) |
5c398dc8 | 1832 | res = test_bit(group - 1, listeners->masks); |
b4ff4f04 JB |
1833 | |
1834 | rcu_read_unlock(); | |
1835 | ||
4277a083 PM |
1836 | return res; |
1837 | } | |
1838 | EXPORT_SYMBOL_GPL(netlink_has_listeners); | |
1839 | ||
b57ef81f | 1840 | static int netlink_broadcast_deliver(struct sock *sk, struct sk_buff *skb) |
1da177e4 LT |
1841 | { |
1842 | struct netlink_sock *nlk = nlk_sk(sk); | |
1843 | ||
1844 | if (atomic_read(&sk->sk_rmem_alloc) <= sk->sk_rcvbuf && | |
cd967e05 | 1845 | !test_bit(NETLINK_CONGESTED, &nlk->state)) { |
cf0a018a | 1846 | netlink_skb_set_owner_r(skb, sk); |
4a7e7c2a | 1847 | __netlink_sendskb(sk, skb); |
2c645800 | 1848 | return atomic_read(&sk->sk_rmem_alloc) > (sk->sk_rcvbuf >> 1); |
1da177e4 LT |
1849 | } |
1850 | return -1; | |
1851 | } | |
1852 | ||
1853 | struct netlink_broadcast_data { | |
1854 | struct sock *exclude_sk; | |
b4b51029 | 1855 | struct net *net; |
15e47304 | 1856 | u32 portid; |
1da177e4 LT |
1857 | u32 group; |
1858 | int failure; | |
ff491a73 | 1859 | int delivery_failure; |
1da177e4 LT |
1860 | int congested; |
1861 | int delivered; | |
7d877f3b | 1862 | gfp_t allocation; |
1da177e4 | 1863 | struct sk_buff *skb, *skb2; |
910a7e90 EB |
1864 | int (*tx_filter)(struct sock *dsk, struct sk_buff *skb, void *data); |
1865 | void *tx_data; | |
1da177e4 LT |
1866 | }; |
1867 | ||
b57ef81f | 1868 | static int do_one_broadcast(struct sock *sk, |
1da177e4 LT |
1869 | struct netlink_broadcast_data *p) |
1870 | { | |
1871 | struct netlink_sock *nlk = nlk_sk(sk); | |
1872 | int val; | |
1873 | ||
1874 | if (p->exclude_sk == sk) | |
1875 | goto out; | |
1876 | ||
15e47304 | 1877 | if (nlk->portid == p->portid || p->group - 1 >= nlk->ngroups || |
f7fa9b10 | 1878 | !test_bit(p->group - 1, nlk->groups)) |
1da177e4 LT |
1879 | goto out; |
1880 | ||
878628fb | 1881 | if (!net_eq(sock_net(sk), p->net)) |
b4b51029 EB |
1882 | goto out; |
1883 | ||
1da177e4 LT |
1884 | if (p->failure) { |
1885 | netlink_overrun(sk); | |
1886 | goto out; | |
1887 | } | |
1888 | ||
1889 | sock_hold(sk); | |
1890 | if (p->skb2 == NULL) { | |
68acc024 | 1891 | if (skb_shared(p->skb)) { |
1da177e4 LT |
1892 | p->skb2 = skb_clone(p->skb, p->allocation); |
1893 | } else { | |
68acc024 TC |
1894 | p->skb2 = skb_get(p->skb); |
1895 | /* | |
1896 | * skb ownership may have been set when | |
1897 | * delivered to a previous socket. | |
1898 | */ | |
1899 | skb_orphan(p->skb2); | |
1da177e4 LT |
1900 | } |
1901 | } | |
1902 | if (p->skb2 == NULL) { | |
1903 | netlink_overrun(sk); | |
1904 | /* Clone failed. Notify ALL listeners. */ | |
1905 | p->failure = 1; | |
be0c22a4 PNA |
1906 | if (nlk->flags & NETLINK_BROADCAST_SEND_ERROR) |
1907 | p->delivery_failure = 1; | |
910a7e90 EB |
1908 | } else if (p->tx_filter && p->tx_filter(sk, p->skb2, p->tx_data)) { |
1909 | kfree_skb(p->skb2); | |
1910 | p->skb2 = NULL; | |
b1153f29 SH |
1911 | } else if (sk_filter(sk, p->skb2)) { |
1912 | kfree_skb(p->skb2); | |
1913 | p->skb2 = NULL; | |
1da177e4 LT |
1914 | } else if ((val = netlink_broadcast_deliver(sk, p->skb2)) < 0) { |
1915 | netlink_overrun(sk); | |
be0c22a4 PNA |
1916 | if (nlk->flags & NETLINK_BROADCAST_SEND_ERROR) |
1917 | p->delivery_failure = 1; | |
1da177e4 LT |
1918 | } else { |
1919 | p->congested |= val; | |
1920 | p->delivered = 1; | |
1921 | p->skb2 = NULL; | |
1922 | } | |
1923 | sock_put(sk); | |
1924 | ||
1925 | out: | |
1926 | return 0; | |
1927 | } | |
1928 | ||
15e47304 | 1929 | int netlink_broadcast_filtered(struct sock *ssk, struct sk_buff *skb, u32 portid, |
910a7e90 EB |
1930 | u32 group, gfp_t allocation, |
1931 | int (*filter)(struct sock *dsk, struct sk_buff *skb, void *data), | |
1932 | void *filter_data) | |
1da177e4 | 1933 | { |
3b1e0a65 | 1934 | struct net *net = sock_net(ssk); |
1da177e4 | 1935 | struct netlink_broadcast_data info; |
1da177e4 LT |
1936 | struct sock *sk; |
1937 | ||
1938 | skb = netlink_trim(skb, allocation); | |
1939 | ||
1940 | info.exclude_sk = ssk; | |
b4b51029 | 1941 | info.net = net; |
15e47304 | 1942 | info.portid = portid; |
1da177e4 LT |
1943 | info.group = group; |
1944 | info.failure = 0; | |
ff491a73 | 1945 | info.delivery_failure = 0; |
1da177e4 LT |
1946 | info.congested = 0; |
1947 | info.delivered = 0; | |
1948 | info.allocation = allocation; | |
1949 | info.skb = skb; | |
1950 | info.skb2 = NULL; | |
910a7e90 EB |
1951 | info.tx_filter = filter; |
1952 | info.tx_data = filter_data; | |
1da177e4 LT |
1953 | |
1954 | /* While we sleep in clone, do not allow to change socket list */ | |
1955 | ||
1956 | netlink_lock_table(); | |
1957 | ||
b67bfe0d | 1958 | sk_for_each_bound(sk, &nl_table[ssk->sk_protocol].mc_list) |
1da177e4 LT |
1959 | do_one_broadcast(sk, &info); |
1960 | ||
70d4bf6d | 1961 | consume_skb(skb); |
aa1c6a6f | 1962 | |
1da177e4 LT |
1963 | netlink_unlock_table(); |
1964 | ||
70d4bf6d NH |
1965 | if (info.delivery_failure) { |
1966 | kfree_skb(info.skb2); | |
ff491a73 | 1967 | return -ENOBUFS; |
658cb354 ED |
1968 | } |
1969 | consume_skb(info.skb2); | |
ff491a73 | 1970 | |
1da177e4 LT |
1971 | if (info.delivered) { |
1972 | if (info.congested && (allocation & __GFP_WAIT)) | |
1973 | yield(); | |
1974 | return 0; | |
1975 | } | |
1da177e4 LT |
1976 | return -ESRCH; |
1977 | } | |
910a7e90 EB |
1978 | EXPORT_SYMBOL(netlink_broadcast_filtered); |
1979 | ||
15e47304 | 1980 | int netlink_broadcast(struct sock *ssk, struct sk_buff *skb, u32 portid, |
910a7e90 EB |
1981 | u32 group, gfp_t allocation) |
1982 | { | |
15e47304 | 1983 | return netlink_broadcast_filtered(ssk, skb, portid, group, allocation, |
910a7e90 EB |
1984 | NULL, NULL); |
1985 | } | |
6ac552fd | 1986 | EXPORT_SYMBOL(netlink_broadcast); |
1da177e4 LT |
1987 | |
1988 | struct netlink_set_err_data { | |
1989 | struct sock *exclude_sk; | |
15e47304 | 1990 | u32 portid; |
1da177e4 LT |
1991 | u32 group; |
1992 | int code; | |
1993 | }; | |
1994 | ||
b57ef81f | 1995 | static int do_one_set_err(struct sock *sk, struct netlink_set_err_data *p) |
1da177e4 LT |
1996 | { |
1997 | struct netlink_sock *nlk = nlk_sk(sk); | |
1a50307b | 1998 | int ret = 0; |
1da177e4 LT |
1999 | |
2000 | if (sk == p->exclude_sk) | |
2001 | goto out; | |
2002 | ||
09ad9bc7 | 2003 | if (!net_eq(sock_net(sk), sock_net(p->exclude_sk))) |
b4b51029 EB |
2004 | goto out; |
2005 | ||
15e47304 | 2006 | if (nlk->portid == p->portid || p->group - 1 >= nlk->ngroups || |
f7fa9b10 | 2007 | !test_bit(p->group - 1, nlk->groups)) |
1da177e4 LT |
2008 | goto out; |
2009 | ||
1a50307b PNA |
2010 | if (p->code == ENOBUFS && nlk->flags & NETLINK_RECV_NO_ENOBUFS) { |
2011 | ret = 1; | |
2012 | goto out; | |
2013 | } | |
2014 | ||
1da177e4 LT |
2015 | sk->sk_err = p->code; |
2016 | sk->sk_error_report(sk); | |
2017 | out: | |
1a50307b | 2018 | return ret; |
1da177e4 LT |
2019 | } |
2020 | ||
4843b93c PNA |
2021 | /** |
2022 | * netlink_set_err - report error to broadcast listeners | |
2023 | * @ssk: the kernel netlink socket, as returned by netlink_kernel_create() | |
15e47304 | 2024 | * @portid: the PORTID of a process that we want to skip (if any) |
840e93f2 | 2025 | * @group: the broadcast group that will notice the error |
4843b93c | 2026 | * @code: error code, must be negative (as usual in kernelspace) |
1a50307b PNA |
2027 | * |
2028 | * This function returns the number of broadcast listeners that have set the | |
2029 | * NETLINK_RECV_NO_ENOBUFS socket option. | |
4843b93c | 2030 | */ |
15e47304 | 2031 | int netlink_set_err(struct sock *ssk, u32 portid, u32 group, int code) |
1da177e4 LT |
2032 | { |
2033 | struct netlink_set_err_data info; | |
1da177e4 | 2034 | struct sock *sk; |
1a50307b | 2035 | int ret = 0; |
1da177e4 LT |
2036 | |
2037 | info.exclude_sk = ssk; | |
15e47304 | 2038 | info.portid = portid; |
1da177e4 | 2039 | info.group = group; |
4843b93c PNA |
2040 | /* sk->sk_err wants a positive error value */ |
2041 | info.code = -code; | |
1da177e4 LT |
2042 | |
2043 | read_lock(&nl_table_lock); | |
2044 | ||
b67bfe0d | 2045 | sk_for_each_bound(sk, &nl_table[ssk->sk_protocol].mc_list) |
1a50307b | 2046 | ret += do_one_set_err(sk, &info); |
1da177e4 LT |
2047 | |
2048 | read_unlock(&nl_table_lock); | |
1a50307b | 2049 | return ret; |
1da177e4 | 2050 | } |
dd5b6ce6 | 2051 | EXPORT_SYMBOL(netlink_set_err); |
1da177e4 | 2052 | |
84659eb5 JB |
2053 | /* must be called with netlink table grabbed */ |
2054 | static void netlink_update_socket_mc(struct netlink_sock *nlk, | |
2055 | unsigned int group, | |
2056 | int is_new) | |
2057 | { | |
2058 | int old, new = !!is_new, subscriptions; | |
2059 | ||
2060 | old = test_bit(group - 1, nlk->groups); | |
2061 | subscriptions = nlk->subscriptions - old + new; | |
2062 | if (new) | |
2063 | __set_bit(group - 1, nlk->groups); | |
2064 | else | |
2065 | __clear_bit(group - 1, nlk->groups); | |
2066 | netlink_update_subscriptions(&nlk->sk, subscriptions); | |
2067 | netlink_update_listeners(&nlk->sk); | |
2068 | } | |
2069 | ||
9a4595bc | 2070 | static int netlink_setsockopt(struct socket *sock, int level, int optname, |
b7058842 | 2071 | char __user *optval, unsigned int optlen) |
9a4595bc PM |
2072 | { |
2073 | struct sock *sk = sock->sk; | |
2074 | struct netlink_sock *nlk = nlk_sk(sk); | |
eb496534 JB |
2075 | unsigned int val = 0; |
2076 | int err; | |
9a4595bc PM |
2077 | |
2078 | if (level != SOL_NETLINK) | |
2079 | return -ENOPROTOOPT; | |
2080 | ||
ccdfcc39 PM |
2081 | if (optname != NETLINK_RX_RING && optname != NETLINK_TX_RING && |
2082 | optlen >= sizeof(int) && | |
eb496534 | 2083 | get_user(val, (unsigned int __user *)optval)) |
9a4595bc PM |
2084 | return -EFAULT; |
2085 | ||
2086 | switch (optname) { | |
2087 | case NETLINK_PKTINFO: | |
2088 | if (val) | |
2089 | nlk->flags |= NETLINK_RECV_PKTINFO; | |
2090 | else | |
2091 | nlk->flags &= ~NETLINK_RECV_PKTINFO; | |
2092 | err = 0; | |
2093 | break; | |
2094 | case NETLINK_ADD_MEMBERSHIP: | |
2095 | case NETLINK_DROP_MEMBERSHIP: { | |
9785e10a | 2096 | if (!netlink_capable(sock, NL_CFG_F_NONROOT_RECV)) |
9a4595bc | 2097 | return -EPERM; |
b4ff4f04 JB |
2098 | err = netlink_realloc_groups(sk); |
2099 | if (err) | |
2100 | return err; | |
9a4595bc PM |
2101 | if (!val || val - 1 >= nlk->ngroups) |
2102 | return -EINVAL; | |
2103 | netlink_table_grab(); | |
84659eb5 JB |
2104 | netlink_update_socket_mc(nlk, val, |
2105 | optname == NETLINK_ADD_MEMBERSHIP); | |
9a4595bc | 2106 | netlink_table_ungrab(); |
03292745 PNA |
2107 | |
2108 | if (nlk->netlink_bind) | |
2109 | nlk->netlink_bind(val); | |
2110 | ||
9a4595bc PM |
2111 | err = 0; |
2112 | break; | |
2113 | } | |
be0c22a4 PNA |
2114 | case NETLINK_BROADCAST_ERROR: |
2115 | if (val) | |
2116 | nlk->flags |= NETLINK_BROADCAST_SEND_ERROR; | |
2117 | else | |
2118 | nlk->flags &= ~NETLINK_BROADCAST_SEND_ERROR; | |
2119 | err = 0; | |
2120 | break; | |
38938bfe PNA |
2121 | case NETLINK_NO_ENOBUFS: |
2122 | if (val) { | |
2123 | nlk->flags |= NETLINK_RECV_NO_ENOBUFS; | |
cd967e05 | 2124 | clear_bit(NETLINK_CONGESTED, &nlk->state); |
38938bfe | 2125 | wake_up_interruptible(&nlk->wait); |
658cb354 | 2126 | } else { |
38938bfe | 2127 | nlk->flags &= ~NETLINK_RECV_NO_ENOBUFS; |
658cb354 | 2128 | } |
38938bfe PNA |
2129 | err = 0; |
2130 | break; | |
ccdfcc39 PM |
2131 | #ifdef CONFIG_NETLINK_MMAP |
2132 | case NETLINK_RX_RING: | |
2133 | case NETLINK_TX_RING: { | |
2134 | struct nl_mmap_req req; | |
2135 | ||
2136 | /* Rings might consume more memory than queue limits, require | |
2137 | * CAP_NET_ADMIN. | |
2138 | */ | |
2139 | if (!capable(CAP_NET_ADMIN)) | |
2140 | return -EPERM; | |
2141 | if (optlen < sizeof(req)) | |
2142 | return -EINVAL; | |
2143 | if (copy_from_user(&req, optval, sizeof(req))) | |
2144 | return -EFAULT; | |
2145 | err = netlink_set_ring(sk, &req, false, | |
2146 | optname == NETLINK_TX_RING); | |
2147 | break; | |
2148 | } | |
2149 | #endif /* CONFIG_NETLINK_MMAP */ | |
9a4595bc PM |
2150 | default: |
2151 | err = -ENOPROTOOPT; | |
2152 | } | |
2153 | return err; | |
2154 | } | |
2155 | ||
2156 | static int netlink_getsockopt(struct socket *sock, int level, int optname, | |
746fac4d | 2157 | char __user *optval, int __user *optlen) |
9a4595bc PM |
2158 | { |
2159 | struct sock *sk = sock->sk; | |
2160 | struct netlink_sock *nlk = nlk_sk(sk); | |
2161 | int len, val, err; | |
2162 | ||
2163 | if (level != SOL_NETLINK) | |
2164 | return -ENOPROTOOPT; | |
2165 | ||
2166 | if (get_user(len, optlen)) | |
2167 | return -EFAULT; | |
2168 | if (len < 0) | |
2169 | return -EINVAL; | |
2170 | ||
2171 | switch (optname) { | |
2172 | case NETLINK_PKTINFO: | |
2173 | if (len < sizeof(int)) | |
2174 | return -EINVAL; | |
2175 | len = sizeof(int); | |
2176 | val = nlk->flags & NETLINK_RECV_PKTINFO ? 1 : 0; | |
a27b58fe HC |
2177 | if (put_user(len, optlen) || |
2178 | put_user(val, optval)) | |
2179 | return -EFAULT; | |
9a4595bc PM |
2180 | err = 0; |
2181 | break; | |
be0c22a4 PNA |
2182 | case NETLINK_BROADCAST_ERROR: |
2183 | if (len < sizeof(int)) | |
2184 | return -EINVAL; | |
2185 | len = sizeof(int); | |
2186 | val = nlk->flags & NETLINK_BROADCAST_SEND_ERROR ? 1 : 0; | |
2187 | if (put_user(len, optlen) || | |
2188 | put_user(val, optval)) | |
2189 | return -EFAULT; | |
2190 | err = 0; | |
2191 | break; | |
38938bfe PNA |
2192 | case NETLINK_NO_ENOBUFS: |
2193 | if (len < sizeof(int)) | |
2194 | return -EINVAL; | |
2195 | len = sizeof(int); | |
2196 | val = nlk->flags & NETLINK_RECV_NO_ENOBUFS ? 1 : 0; | |
2197 | if (put_user(len, optlen) || | |
2198 | put_user(val, optval)) | |
2199 | return -EFAULT; | |
2200 | err = 0; | |
2201 | break; | |
9a4595bc PM |
2202 | default: |
2203 | err = -ENOPROTOOPT; | |
2204 | } | |
2205 | return err; | |
2206 | } | |
2207 | ||
2208 | static void netlink_cmsg_recv_pktinfo(struct msghdr *msg, struct sk_buff *skb) | |
2209 | { | |
2210 | struct nl_pktinfo info; | |
2211 | ||
2212 | info.group = NETLINK_CB(skb).dst_group; | |
2213 | put_cmsg(msg, SOL_NETLINK, NETLINK_PKTINFO, sizeof(info), &info); | |
2214 | } | |
2215 | ||
1da177e4 LT |
2216 | static int netlink_sendmsg(struct kiocb *kiocb, struct socket *sock, |
2217 | struct msghdr *msg, size_t len) | |
2218 | { | |
2219 | struct sock_iocb *siocb = kiocb_to_siocb(kiocb); | |
2220 | struct sock *sk = sock->sk; | |
2221 | struct netlink_sock *nlk = nlk_sk(sk); | |
6ac552fd | 2222 | struct sockaddr_nl *addr = msg->msg_name; |
15e47304 | 2223 | u32 dst_portid; |
d629b836 | 2224 | u32 dst_group; |
1da177e4 LT |
2225 | struct sk_buff *skb; |
2226 | int err; | |
2227 | struct scm_cookie scm; | |
2228 | ||
2229 | if (msg->msg_flags&MSG_OOB) | |
2230 | return -EOPNOTSUPP; | |
2231 | ||
16e57262 | 2232 | if (NULL == siocb->scm) |
1da177e4 | 2233 | siocb->scm = &scm; |
16e57262 | 2234 | |
e0e3cea4 | 2235 | err = scm_send(sock, msg, siocb->scm, true); |
1da177e4 LT |
2236 | if (err < 0) |
2237 | return err; | |
2238 | ||
2239 | if (msg->msg_namelen) { | |
b47030c7 | 2240 | err = -EINVAL; |
1da177e4 | 2241 | if (addr->nl_family != AF_NETLINK) |
b47030c7 | 2242 | goto out; |
15e47304 | 2243 | dst_portid = addr->nl_pid; |
d629b836 | 2244 | dst_group = ffs(addr->nl_groups); |
b47030c7 | 2245 | err = -EPERM; |
15e47304 | 2246 | if ((dst_group || dst_portid) && |
9785e10a | 2247 | !netlink_capable(sock, NL_CFG_F_NONROOT_SEND)) |
b47030c7 | 2248 | goto out; |
1da177e4 | 2249 | } else { |
15e47304 | 2250 | dst_portid = nlk->dst_portid; |
d629b836 | 2251 | dst_group = nlk->dst_group; |
1da177e4 LT |
2252 | } |
2253 | ||
15e47304 | 2254 | if (!nlk->portid) { |
1da177e4 LT |
2255 | err = netlink_autobind(sock); |
2256 | if (err) | |
2257 | goto out; | |
2258 | } | |
2259 | ||
5fd96123 PM |
2260 | if (netlink_tx_is_mmaped(sk) && |
2261 | msg->msg_iov->iov_base == NULL) { | |
2262 | err = netlink_mmap_sendmsg(sk, msg, dst_portid, dst_group, | |
2263 | siocb); | |
2264 | goto out; | |
2265 | } | |
2266 | ||
1da177e4 LT |
2267 | err = -EMSGSIZE; |
2268 | if (len > sk->sk_sndbuf - 32) | |
2269 | goto out; | |
2270 | err = -ENOBUFS; | |
3a36515f | 2271 | skb = netlink_alloc_large_skb(len, dst_group); |
6ac552fd | 2272 | if (skb == NULL) |
1da177e4 LT |
2273 | goto out; |
2274 | ||
15e47304 | 2275 | NETLINK_CB(skb).portid = nlk->portid; |
d629b836 | 2276 | NETLINK_CB(skb).dst_group = dst_group; |
dbe9a417 | 2277 | NETLINK_CB(skb).creds = siocb->scm->creds; |
1da177e4 | 2278 | |
1da177e4 | 2279 | err = -EFAULT; |
6ac552fd | 2280 | if (memcpy_fromiovec(skb_put(skb, len), msg->msg_iov, len)) { |
1da177e4 LT |
2281 | kfree_skb(skb); |
2282 | goto out; | |
2283 | } | |
2284 | ||
2285 | err = security_netlink_send(sk, skb); | |
2286 | if (err) { | |
2287 | kfree_skb(skb); | |
2288 | goto out; | |
2289 | } | |
2290 | ||
d629b836 | 2291 | if (dst_group) { |
1da177e4 | 2292 | atomic_inc(&skb->users); |
15e47304 | 2293 | netlink_broadcast(sk, skb, dst_portid, dst_group, GFP_KERNEL); |
1da177e4 | 2294 | } |
15e47304 | 2295 | err = netlink_unicast(sk, skb, dst_portid, msg->msg_flags&MSG_DONTWAIT); |
1da177e4 LT |
2296 | |
2297 | out: | |
b47030c7 | 2298 | scm_destroy(siocb->scm); |
1da177e4 LT |
2299 | return err; |
2300 | } | |
2301 | ||
2302 | static int netlink_recvmsg(struct kiocb *kiocb, struct socket *sock, | |
2303 | struct msghdr *msg, size_t len, | |
2304 | int flags) | |
2305 | { | |
2306 | struct sock_iocb *siocb = kiocb_to_siocb(kiocb); | |
2307 | struct scm_cookie scm; | |
2308 | struct sock *sk = sock->sk; | |
2309 | struct netlink_sock *nlk = nlk_sk(sk); | |
2310 | int noblock = flags&MSG_DONTWAIT; | |
2311 | size_t copied; | |
68d6ac6d | 2312 | struct sk_buff *skb, *data_skb; |
b44d211e | 2313 | int err, ret; |
1da177e4 LT |
2314 | |
2315 | if (flags&MSG_OOB) | |
2316 | return -EOPNOTSUPP; | |
2317 | ||
2318 | copied = 0; | |
2319 | ||
6ac552fd PM |
2320 | skb = skb_recv_datagram(sk, flags, noblock, &err); |
2321 | if (skb == NULL) | |
1da177e4 LT |
2322 | goto out; |
2323 | ||
68d6ac6d JB |
2324 | data_skb = skb; |
2325 | ||
1dacc76d JB |
2326 | #ifdef CONFIG_COMPAT_NETLINK_MESSAGES |
2327 | if (unlikely(skb_shinfo(skb)->frag_list)) { | |
1dacc76d | 2328 | /* |
68d6ac6d JB |
2329 | * If this skb has a frag_list, then here that means that we |
2330 | * will have to use the frag_list skb's data for compat tasks | |
2331 | * and the regular skb's data for normal (non-compat) tasks. | |
1dacc76d | 2332 | * |
68d6ac6d JB |
2333 | * If we need to send the compat skb, assign it to the |
2334 | * 'data_skb' variable so that it will be used below for data | |
2335 | * copying. We keep 'skb' for everything else, including | |
2336 | * freeing both later. | |
1dacc76d | 2337 | */ |
68d6ac6d JB |
2338 | if (flags & MSG_CMSG_COMPAT) |
2339 | data_skb = skb_shinfo(skb)->frag_list; | |
1dacc76d JB |
2340 | } |
2341 | #endif | |
2342 | ||
68d6ac6d | 2343 | copied = data_skb->len; |
1da177e4 LT |
2344 | if (len < copied) { |
2345 | msg->msg_flags |= MSG_TRUNC; | |
2346 | copied = len; | |
2347 | } | |
2348 | ||
68d6ac6d JB |
2349 | skb_reset_transport_header(data_skb); |
2350 | err = skb_copy_datagram_iovec(data_skb, 0, msg->msg_iov, copied); | |
1da177e4 LT |
2351 | |
2352 | if (msg->msg_name) { | |
6ac552fd | 2353 | struct sockaddr_nl *addr = (struct sockaddr_nl *)msg->msg_name; |
1da177e4 LT |
2354 | addr->nl_family = AF_NETLINK; |
2355 | addr->nl_pad = 0; | |
15e47304 | 2356 | addr->nl_pid = NETLINK_CB(skb).portid; |
d629b836 | 2357 | addr->nl_groups = netlink_group_mask(NETLINK_CB(skb).dst_group); |
1da177e4 LT |
2358 | msg->msg_namelen = sizeof(*addr); |
2359 | } | |
2360 | ||
cc9a06cd PM |
2361 | if (nlk->flags & NETLINK_RECV_PKTINFO) |
2362 | netlink_cmsg_recv_pktinfo(msg, skb); | |
2363 | ||
1da177e4 LT |
2364 | if (NULL == siocb->scm) { |
2365 | memset(&scm, 0, sizeof(scm)); | |
2366 | siocb->scm = &scm; | |
2367 | } | |
2368 | siocb->scm->creds = *NETLINK_CREDS(skb); | |
188ccb55 | 2369 | if (flags & MSG_TRUNC) |
68d6ac6d | 2370 | copied = data_skb->len; |
daa3766e | 2371 | |
1da177e4 LT |
2372 | skb_free_datagram(sk, skb); |
2373 | ||
16b304f3 PS |
2374 | if (nlk->cb_running && |
2375 | atomic_read(&sk->sk_rmem_alloc) <= sk->sk_rcvbuf / 2) { | |
b44d211e AV |
2376 | ret = netlink_dump(sk); |
2377 | if (ret) { | |
2378 | sk->sk_err = ret; | |
2379 | sk->sk_error_report(sk); | |
2380 | } | |
2381 | } | |
1da177e4 LT |
2382 | |
2383 | scm_recv(sock, msg, siocb->scm, flags); | |
1da177e4 LT |
2384 | out: |
2385 | netlink_rcv_wake(sk); | |
2386 | return err ? : copied; | |
2387 | } | |
2388 | ||
2389 | static void netlink_data_ready(struct sock *sk, int len) | |
2390 | { | |
cd40b7d3 | 2391 | BUG(); |
1da177e4 LT |
2392 | } |
2393 | ||
2394 | /* | |
746fac4d | 2395 | * We export these functions to other modules. They provide a |
1da177e4 LT |
2396 | * complete set of kernel non-blocking support for message |
2397 | * queueing. | |
2398 | */ | |
2399 | ||
2400 | struct sock * | |
9f00d977 PNA |
2401 | __netlink_kernel_create(struct net *net, int unit, struct module *module, |
2402 | struct netlink_kernel_cfg *cfg) | |
1da177e4 LT |
2403 | { |
2404 | struct socket *sock; | |
2405 | struct sock *sk; | |
77247bbb | 2406 | struct netlink_sock *nlk; |
5c398dc8 | 2407 | struct listeners *listeners = NULL; |
a31f2d17 PNA |
2408 | struct mutex *cb_mutex = cfg ? cfg->cb_mutex : NULL; |
2409 | unsigned int groups; | |
1da177e4 | 2410 | |
fab2caf6 | 2411 | BUG_ON(!nl_table); |
1da177e4 | 2412 | |
6ac552fd | 2413 | if (unit < 0 || unit >= MAX_LINKS) |
1da177e4 LT |
2414 | return NULL; |
2415 | ||
2416 | if (sock_create_lite(PF_NETLINK, SOCK_DGRAM, unit, &sock)) | |
2417 | return NULL; | |
2418 | ||
23fe1866 PE |
2419 | /* |
2420 | * We have to just have a reference on the net from sk, but don't | |
2421 | * get_net it. Besides, we cannot get and then put the net here. | |
2422 | * So we create one inside init_net and the move it to net. | |
2423 | */ | |
2424 | ||
2425 | if (__netlink_create(&init_net, sock, cb_mutex, unit) < 0) | |
2426 | goto out_sock_release_nosk; | |
2427 | ||
2428 | sk = sock->sk; | |
edf02087 | 2429 | sk_change_net(sk, net); |
4fdb3bb7 | 2430 | |
a31f2d17 | 2431 | if (!cfg || cfg->groups < 32) |
4277a083 | 2432 | groups = 32; |
a31f2d17 PNA |
2433 | else |
2434 | groups = cfg->groups; | |
4277a083 | 2435 | |
5c398dc8 | 2436 | listeners = kzalloc(sizeof(*listeners) + NLGRPSZ(groups), GFP_KERNEL); |
4277a083 PM |
2437 | if (!listeners) |
2438 | goto out_sock_release; | |
2439 | ||
1da177e4 | 2440 | sk->sk_data_ready = netlink_data_ready; |
a31f2d17 PNA |
2441 | if (cfg && cfg->input) |
2442 | nlk_sk(sk)->netlink_rcv = cfg->input; | |
1da177e4 | 2443 | |
b4b51029 | 2444 | if (netlink_insert(sk, net, 0)) |
77247bbb | 2445 | goto out_sock_release; |
4fdb3bb7 | 2446 | |
77247bbb PM |
2447 | nlk = nlk_sk(sk); |
2448 | nlk->flags |= NETLINK_KERNEL_SOCKET; | |
4fdb3bb7 | 2449 | |
4fdb3bb7 | 2450 | netlink_table_grab(); |
b4b51029 EB |
2451 | if (!nl_table[unit].registered) { |
2452 | nl_table[unit].groups = groups; | |
5c398dc8 | 2453 | rcu_assign_pointer(nl_table[unit].listeners, listeners); |
b4b51029 EB |
2454 | nl_table[unit].cb_mutex = cb_mutex; |
2455 | nl_table[unit].module = module; | |
9785e10a PNA |
2456 | if (cfg) { |
2457 | nl_table[unit].bind = cfg->bind; | |
2458 | nl_table[unit].flags = cfg->flags; | |
da12c90e G |
2459 | if (cfg->compare) |
2460 | nl_table[unit].compare = cfg->compare; | |
9785e10a | 2461 | } |
b4b51029 | 2462 | nl_table[unit].registered = 1; |
f937f1f4 JJ |
2463 | } else { |
2464 | kfree(listeners); | |
869e58f8 | 2465 | nl_table[unit].registered++; |
b4b51029 | 2466 | } |
4fdb3bb7 | 2467 | netlink_table_ungrab(); |
77247bbb PM |
2468 | return sk; |
2469 | ||
4fdb3bb7 | 2470 | out_sock_release: |
4277a083 | 2471 | kfree(listeners); |
9dfbec1f | 2472 | netlink_kernel_release(sk); |
23fe1866 PE |
2473 | return NULL; |
2474 | ||
2475 | out_sock_release_nosk: | |
4fdb3bb7 | 2476 | sock_release(sock); |
77247bbb | 2477 | return NULL; |
1da177e4 | 2478 | } |
9f00d977 | 2479 | EXPORT_SYMBOL(__netlink_kernel_create); |
b7c6ba6e DL |
2480 | |
2481 | void | |
2482 | netlink_kernel_release(struct sock *sk) | |
2483 | { | |
edf02087 | 2484 | sk_release_kernel(sk); |
b7c6ba6e DL |
2485 | } |
2486 | EXPORT_SYMBOL(netlink_kernel_release); | |
2487 | ||
d136f1bd | 2488 | int __netlink_change_ngroups(struct sock *sk, unsigned int groups) |
b4ff4f04 | 2489 | { |
5c398dc8 | 2490 | struct listeners *new, *old; |
b4ff4f04 | 2491 | struct netlink_table *tbl = &nl_table[sk->sk_protocol]; |
b4ff4f04 JB |
2492 | |
2493 | if (groups < 32) | |
2494 | groups = 32; | |
2495 | ||
b4ff4f04 | 2496 | if (NLGRPSZ(tbl->groups) < NLGRPSZ(groups)) { |
5c398dc8 ED |
2497 | new = kzalloc(sizeof(*new) + NLGRPSZ(groups), GFP_ATOMIC); |
2498 | if (!new) | |
d136f1bd | 2499 | return -ENOMEM; |
6d772ac5 | 2500 | old = nl_deref_protected(tbl->listeners); |
5c398dc8 ED |
2501 | memcpy(new->masks, old->masks, NLGRPSZ(tbl->groups)); |
2502 | rcu_assign_pointer(tbl->listeners, new); | |
2503 | ||
37b6b935 | 2504 | kfree_rcu(old, rcu); |
b4ff4f04 JB |
2505 | } |
2506 | tbl->groups = groups; | |
2507 | ||
d136f1bd JB |
2508 | return 0; |
2509 | } | |
2510 | ||
2511 | /** | |
2512 | * netlink_change_ngroups - change number of multicast groups | |
2513 | * | |
2514 | * This changes the number of multicast groups that are available | |
2515 | * on a certain netlink family. Note that it is not possible to | |
2516 | * change the number of groups to below 32. Also note that it does | |
2517 | * not implicitly call netlink_clear_multicast_users() when the | |
2518 | * number of groups is reduced. | |
2519 | * | |
2520 | * @sk: The kernel netlink socket, as returned by netlink_kernel_create(). | |
2521 | * @groups: The new number of groups. | |
2522 | */ | |
2523 | int netlink_change_ngroups(struct sock *sk, unsigned int groups) | |
2524 | { | |
2525 | int err; | |
2526 | ||
2527 | netlink_table_grab(); | |
2528 | err = __netlink_change_ngroups(sk, groups); | |
b4ff4f04 | 2529 | netlink_table_ungrab(); |
d136f1bd | 2530 | |
b4ff4f04 JB |
2531 | return err; |
2532 | } | |
b4ff4f04 | 2533 | |
b8273570 JB |
2534 | void __netlink_clear_multicast_users(struct sock *ksk, unsigned int group) |
2535 | { | |
2536 | struct sock *sk; | |
b8273570 JB |
2537 | struct netlink_table *tbl = &nl_table[ksk->sk_protocol]; |
2538 | ||
b67bfe0d | 2539 | sk_for_each_bound(sk, &tbl->mc_list) |
b8273570 JB |
2540 | netlink_update_socket_mc(nlk_sk(sk), group, 0); |
2541 | } | |
2542 | ||
84659eb5 JB |
2543 | /** |
2544 | * netlink_clear_multicast_users - kick off multicast listeners | |
2545 | * | |
2546 | * This function removes all listeners from the given group. | |
2547 | * @ksk: The kernel netlink socket, as returned by | |
2548 | * netlink_kernel_create(). | |
2549 | * @group: The multicast group to clear. | |
2550 | */ | |
2551 | void netlink_clear_multicast_users(struct sock *ksk, unsigned int group) | |
2552 | { | |
84659eb5 | 2553 | netlink_table_grab(); |
b8273570 | 2554 | __netlink_clear_multicast_users(ksk, group); |
84659eb5 JB |
2555 | netlink_table_ungrab(); |
2556 | } | |
84659eb5 | 2557 | |
a46621a3 | 2558 | struct nlmsghdr * |
15e47304 | 2559 | __nlmsg_put(struct sk_buff *skb, u32 portid, u32 seq, int type, int len, int flags) |
a46621a3 DV |
2560 | { |
2561 | struct nlmsghdr *nlh; | |
573ce260 | 2562 | int size = nlmsg_msg_size(len); |
a46621a3 DV |
2563 | |
2564 | nlh = (struct nlmsghdr*)skb_put(skb, NLMSG_ALIGN(size)); | |
2565 | nlh->nlmsg_type = type; | |
2566 | nlh->nlmsg_len = size; | |
2567 | nlh->nlmsg_flags = flags; | |
15e47304 | 2568 | nlh->nlmsg_pid = portid; |
a46621a3 DV |
2569 | nlh->nlmsg_seq = seq; |
2570 | if (!__builtin_constant_p(size) || NLMSG_ALIGN(size) - size != 0) | |
573ce260 | 2571 | memset(nlmsg_data(nlh) + len, 0, NLMSG_ALIGN(size) - size); |
a46621a3 DV |
2572 | return nlh; |
2573 | } | |
2574 | EXPORT_SYMBOL(__nlmsg_put); | |
2575 | ||
1da177e4 LT |
2576 | /* |
2577 | * It looks a bit ugly. | |
2578 | * It would be better to create kernel thread. | |
2579 | */ | |
2580 | ||
2581 | static int netlink_dump(struct sock *sk) | |
2582 | { | |
2583 | struct netlink_sock *nlk = nlk_sk(sk); | |
2584 | struct netlink_callback *cb; | |
c7ac8679 | 2585 | struct sk_buff *skb = NULL; |
1da177e4 | 2586 | struct nlmsghdr *nlh; |
bf8b79e4 | 2587 | int len, err = -ENOBUFS; |
c7ac8679 | 2588 | int alloc_size; |
1da177e4 | 2589 | |
af65bdfc | 2590 | mutex_lock(nlk->cb_mutex); |
16b304f3 | 2591 | if (!nlk->cb_running) { |
bf8b79e4 TG |
2592 | err = -EINVAL; |
2593 | goto errout_skb; | |
1da177e4 LT |
2594 | } |
2595 | ||
16b304f3 | 2596 | cb = &nlk->cb; |
c7ac8679 GR |
2597 | alloc_size = max_t(int, cb->min_dump_alloc, NLMSG_GOODSIZE); |
2598 | ||
f9c22888 PM |
2599 | if (!netlink_rx_is_mmaped(sk) && |
2600 | atomic_read(&sk->sk_rmem_alloc) >= sk->sk_rcvbuf) | |
2601 | goto errout_skb; | |
2602 | skb = netlink_alloc_skb(sk, alloc_size, nlk->portid, GFP_KERNEL); | |
c7ac8679 | 2603 | if (!skb) |
c63d6ea3 | 2604 | goto errout_skb; |
f9c22888 | 2605 | netlink_skb_set_owner_r(skb, sk); |
c7ac8679 | 2606 | |
1da177e4 LT |
2607 | len = cb->dump(skb, cb); |
2608 | ||
2609 | if (len > 0) { | |
af65bdfc | 2610 | mutex_unlock(nlk->cb_mutex); |
b1153f29 SH |
2611 | |
2612 | if (sk_filter(sk, skb)) | |
2613 | kfree_skb(skb); | |
4a7e7c2a ED |
2614 | else |
2615 | __netlink_sendskb(sk, skb); | |
1da177e4 LT |
2616 | return 0; |
2617 | } | |
2618 | ||
bf8b79e4 TG |
2619 | nlh = nlmsg_put_answer(skb, cb, NLMSG_DONE, sizeof(len), NLM_F_MULTI); |
2620 | if (!nlh) | |
2621 | goto errout_skb; | |
2622 | ||
670dc283 JB |
2623 | nl_dump_check_consistent(cb, nlh); |
2624 | ||
bf8b79e4 TG |
2625 | memcpy(nlmsg_data(nlh), &len, sizeof(len)); |
2626 | ||
b1153f29 SH |
2627 | if (sk_filter(sk, skb)) |
2628 | kfree_skb(skb); | |
4a7e7c2a ED |
2629 | else |
2630 | __netlink_sendskb(sk, skb); | |
1da177e4 | 2631 | |
a8f74b22 TG |
2632 | if (cb->done) |
2633 | cb->done(cb); | |
1da177e4 | 2634 | |
16b304f3 PS |
2635 | nlk->cb_running = false; |
2636 | mutex_unlock(nlk->cb_mutex); | |
6dc878a8 | 2637 | module_put(cb->module); |
16b304f3 | 2638 | consume_skb(cb->skb); |
1da177e4 | 2639 | return 0; |
1797754e | 2640 | |
bf8b79e4 | 2641 | errout_skb: |
af65bdfc | 2642 | mutex_unlock(nlk->cb_mutex); |
bf8b79e4 | 2643 | kfree_skb(skb); |
bf8b79e4 | 2644 | return err; |
1da177e4 LT |
2645 | } |
2646 | ||
6dc878a8 G |
2647 | int __netlink_dump_start(struct sock *ssk, struct sk_buff *skb, |
2648 | const struct nlmsghdr *nlh, | |
2649 | struct netlink_dump_control *control) | |
1da177e4 LT |
2650 | { |
2651 | struct netlink_callback *cb; | |
2652 | struct sock *sk; | |
2653 | struct netlink_sock *nlk; | |
b44d211e | 2654 | int ret; |
1da177e4 | 2655 | |
f9c22888 PM |
2656 | /* Memory mapped dump requests need to be copied to avoid looping |
2657 | * on the pending state in netlink_mmap_sendmsg() while the CB hold | |
2658 | * a reference to the skb. | |
2659 | */ | |
2660 | if (netlink_skb_is_mmaped(skb)) { | |
2661 | skb = skb_copy(skb, GFP_KERNEL); | |
16b304f3 | 2662 | if (skb == NULL) |
f9c22888 | 2663 | return -ENOBUFS; |
f9c22888 PM |
2664 | } else |
2665 | atomic_inc(&skb->users); | |
2666 | ||
15e47304 | 2667 | sk = netlink_lookup(sock_net(ssk), ssk->sk_protocol, NETLINK_CB(skb).portid); |
1da177e4 | 2668 | if (sk == NULL) { |
16b304f3 PS |
2669 | ret = -ECONNREFUSED; |
2670 | goto error_free; | |
1da177e4 | 2671 | } |
6dc878a8 | 2672 | |
16b304f3 | 2673 | nlk = nlk_sk(sk); |
af65bdfc | 2674 | mutex_lock(nlk->cb_mutex); |
6dc878a8 | 2675 | /* A dump is in progress... */ |
16b304f3 | 2676 | if (nlk->cb_running) { |
6dc878a8 | 2677 | ret = -EBUSY; |
16b304f3 | 2678 | goto error_unlock; |
1da177e4 | 2679 | } |
6dc878a8 | 2680 | /* add reference of module which cb->dump belongs to */ |
16b304f3 | 2681 | if (!try_module_get(control->module)) { |
6dc878a8 | 2682 | ret = -EPROTONOSUPPORT; |
16b304f3 | 2683 | goto error_unlock; |
6dc878a8 G |
2684 | } |
2685 | ||
16b304f3 PS |
2686 | cb = &nlk->cb; |
2687 | memset(cb, 0, sizeof(*cb)); | |
2688 | cb->dump = control->dump; | |
2689 | cb->done = control->done; | |
2690 | cb->nlh = nlh; | |
2691 | cb->data = control->data; | |
2692 | cb->module = control->module; | |
2693 | cb->min_dump_alloc = control->min_dump_alloc; | |
2694 | cb->skb = skb; | |
2695 | ||
2696 | nlk->cb_running = true; | |
2697 | ||
af65bdfc | 2698 | mutex_unlock(nlk->cb_mutex); |
1da177e4 | 2699 | |
b44d211e | 2700 | ret = netlink_dump(sk); |
1da177e4 | 2701 | sock_put(sk); |
5c58298c | 2702 | |
b44d211e AV |
2703 | if (ret) |
2704 | return ret; | |
2705 | ||
5c58298c DL |
2706 | /* We successfully started a dump, by returning -EINTR we |
2707 | * signal not to send ACK even if it was requested. | |
2708 | */ | |
2709 | return -EINTR; | |
16b304f3 PS |
2710 | |
2711 | error_unlock: | |
2712 | sock_put(sk); | |
2713 | mutex_unlock(nlk->cb_mutex); | |
2714 | error_free: | |
2715 | kfree_skb(skb); | |
2716 | return ret; | |
1da177e4 | 2717 | } |
6dc878a8 | 2718 | EXPORT_SYMBOL(__netlink_dump_start); |
1da177e4 LT |
2719 | |
2720 | void netlink_ack(struct sk_buff *in_skb, struct nlmsghdr *nlh, int err) | |
2721 | { | |
2722 | struct sk_buff *skb; | |
2723 | struct nlmsghdr *rep; | |
2724 | struct nlmsgerr *errmsg; | |
339bf98f | 2725 | size_t payload = sizeof(*errmsg); |
1da177e4 | 2726 | |
339bf98f TG |
2727 | /* error messages get the original request appened */ |
2728 | if (err) | |
2729 | payload += nlmsg_len(nlh); | |
1da177e4 | 2730 | |
f9c22888 PM |
2731 | skb = netlink_alloc_skb(in_skb->sk, nlmsg_total_size(payload), |
2732 | NETLINK_CB(in_skb).portid, GFP_KERNEL); | |
1da177e4 LT |
2733 | if (!skb) { |
2734 | struct sock *sk; | |
2735 | ||
3b1e0a65 | 2736 | sk = netlink_lookup(sock_net(in_skb->sk), |
b4b51029 | 2737 | in_skb->sk->sk_protocol, |
15e47304 | 2738 | NETLINK_CB(in_skb).portid); |
1da177e4 LT |
2739 | if (sk) { |
2740 | sk->sk_err = ENOBUFS; | |
2741 | sk->sk_error_report(sk); | |
2742 | sock_put(sk); | |
2743 | } | |
2744 | return; | |
2745 | } | |
2746 | ||
15e47304 | 2747 | rep = __nlmsg_put(skb, NETLINK_CB(in_skb).portid, nlh->nlmsg_seq, |
5dba93ae | 2748 | NLMSG_ERROR, payload, 0); |
bf8b79e4 | 2749 | errmsg = nlmsg_data(rep); |
1da177e4 | 2750 | errmsg->error = err; |
bf8b79e4 | 2751 | memcpy(&errmsg->msg, nlh, err ? nlh->nlmsg_len : sizeof(*nlh)); |
15e47304 | 2752 | netlink_unicast(in_skb->sk, skb, NETLINK_CB(in_skb).portid, MSG_DONTWAIT); |
1da177e4 | 2753 | } |
6ac552fd | 2754 | EXPORT_SYMBOL(netlink_ack); |
1da177e4 | 2755 | |
cd40b7d3 | 2756 | int netlink_rcv_skb(struct sk_buff *skb, int (*cb)(struct sk_buff *, |
1d00a4eb | 2757 | struct nlmsghdr *)) |
82ace47a | 2758 | { |
82ace47a TG |
2759 | struct nlmsghdr *nlh; |
2760 | int err; | |
2761 | ||
2762 | while (skb->len >= nlmsg_total_size(0)) { | |
cd40b7d3 DL |
2763 | int msglen; |
2764 | ||
b529ccf2 | 2765 | nlh = nlmsg_hdr(skb); |
d35b6856 | 2766 | err = 0; |
82ace47a | 2767 | |
ad8e4b75 | 2768 | if (nlh->nlmsg_len < NLMSG_HDRLEN || skb->len < nlh->nlmsg_len) |
82ace47a TG |
2769 | return 0; |
2770 | ||
d35b6856 TG |
2771 | /* Only requests are handled by the kernel */ |
2772 | if (!(nlh->nlmsg_flags & NLM_F_REQUEST)) | |
5c58298c | 2773 | goto ack; |
45e7ae7f TG |
2774 | |
2775 | /* Skip control messages */ | |
2776 | if (nlh->nlmsg_type < NLMSG_MIN_TYPE) | |
5c58298c | 2777 | goto ack; |
d35b6856 | 2778 | |
1d00a4eb | 2779 | err = cb(skb, nlh); |
5c58298c DL |
2780 | if (err == -EINTR) |
2781 | goto skip; | |
2782 | ||
2783 | ack: | |
d35b6856 | 2784 | if (nlh->nlmsg_flags & NLM_F_ACK || err) |
82ace47a | 2785 | netlink_ack(skb, nlh, err); |
82ace47a | 2786 | |
5c58298c | 2787 | skip: |
6ac552fd | 2788 | msglen = NLMSG_ALIGN(nlh->nlmsg_len); |
cd40b7d3 DL |
2789 | if (msglen > skb->len) |
2790 | msglen = skb->len; | |
2791 | skb_pull(skb, msglen); | |
82ace47a TG |
2792 | } |
2793 | ||
2794 | return 0; | |
2795 | } | |
6ac552fd | 2796 | EXPORT_SYMBOL(netlink_rcv_skb); |
82ace47a | 2797 | |
d387f6ad TG |
2798 | /** |
2799 | * nlmsg_notify - send a notification netlink message | |
2800 | * @sk: netlink socket to use | |
2801 | * @skb: notification message | |
15e47304 | 2802 | * @portid: destination netlink portid for reports or 0 |
d387f6ad TG |
2803 | * @group: destination multicast group or 0 |
2804 | * @report: 1 to report back, 0 to disable | |
2805 | * @flags: allocation flags | |
2806 | */ | |
15e47304 | 2807 | int nlmsg_notify(struct sock *sk, struct sk_buff *skb, u32 portid, |
d387f6ad TG |
2808 | unsigned int group, int report, gfp_t flags) |
2809 | { | |
2810 | int err = 0; | |
2811 | ||
2812 | if (group) { | |
15e47304 | 2813 | int exclude_portid = 0; |
d387f6ad TG |
2814 | |
2815 | if (report) { | |
2816 | atomic_inc(&skb->users); | |
15e47304 | 2817 | exclude_portid = portid; |
d387f6ad TG |
2818 | } |
2819 | ||
1ce85fe4 PNA |
2820 | /* errors reported via destination sk->sk_err, but propagate |
2821 | * delivery errors if NETLINK_BROADCAST_ERROR flag is set */ | |
15e47304 | 2822 | err = nlmsg_multicast(sk, skb, exclude_portid, group, flags); |
d387f6ad TG |
2823 | } |
2824 | ||
1ce85fe4 PNA |
2825 | if (report) { |
2826 | int err2; | |
2827 | ||
15e47304 | 2828 | err2 = nlmsg_unicast(sk, skb, portid); |
1ce85fe4 PNA |
2829 | if (!err || err == -ESRCH) |
2830 | err = err2; | |
2831 | } | |
d387f6ad TG |
2832 | |
2833 | return err; | |
2834 | } | |
6ac552fd | 2835 | EXPORT_SYMBOL(nlmsg_notify); |
d387f6ad | 2836 | |
1da177e4 LT |
2837 | #ifdef CONFIG_PROC_FS |
2838 | struct nl_seq_iter { | |
e372c414 | 2839 | struct seq_net_private p; |
1da177e4 LT |
2840 | int link; |
2841 | int hash_idx; | |
2842 | }; | |
2843 | ||
2844 | static struct sock *netlink_seq_socket_idx(struct seq_file *seq, loff_t pos) | |
2845 | { | |
2846 | struct nl_seq_iter *iter = seq->private; | |
2847 | int i, j; | |
2848 | struct sock *s; | |
1da177e4 LT |
2849 | loff_t off = 0; |
2850 | ||
6ac552fd | 2851 | for (i = 0; i < MAX_LINKS; i++) { |
15e47304 | 2852 | struct nl_portid_hash *hash = &nl_table[i].hash; |
1da177e4 LT |
2853 | |
2854 | for (j = 0; j <= hash->mask; j++) { | |
b67bfe0d | 2855 | sk_for_each(s, &hash->table[j]) { |
1218854a | 2856 | if (sock_net(s) != seq_file_net(seq)) |
b4b51029 | 2857 | continue; |
1da177e4 LT |
2858 | if (off == pos) { |
2859 | iter->link = i; | |
2860 | iter->hash_idx = j; | |
2861 | return s; | |
2862 | } | |
2863 | ++off; | |
2864 | } | |
2865 | } | |
2866 | } | |
2867 | return NULL; | |
2868 | } | |
2869 | ||
2870 | static void *netlink_seq_start(struct seq_file *seq, loff_t *pos) | |
9a429c49 | 2871 | __acquires(nl_table_lock) |
1da177e4 LT |
2872 | { |
2873 | read_lock(&nl_table_lock); | |
2874 | return *pos ? netlink_seq_socket_idx(seq, *pos - 1) : SEQ_START_TOKEN; | |
2875 | } | |
2876 | ||
2877 | static void *netlink_seq_next(struct seq_file *seq, void *v, loff_t *pos) | |
2878 | { | |
2879 | struct sock *s; | |
2880 | struct nl_seq_iter *iter; | |
da12c90e | 2881 | struct net *net; |
1da177e4 LT |
2882 | int i, j; |
2883 | ||
2884 | ++*pos; | |
2885 | ||
2886 | if (v == SEQ_START_TOKEN) | |
2887 | return netlink_seq_socket_idx(seq, 0); | |
746fac4d | 2888 | |
da12c90e | 2889 | net = seq_file_net(seq); |
b4b51029 EB |
2890 | iter = seq->private; |
2891 | s = v; | |
2892 | do { | |
2893 | s = sk_next(s); | |
da12c90e | 2894 | } while (s && !nl_table[s->sk_protocol].compare(net, s)); |
1da177e4 LT |
2895 | if (s) |
2896 | return s; | |
2897 | ||
1da177e4 LT |
2898 | i = iter->link; |
2899 | j = iter->hash_idx + 1; | |
2900 | ||
2901 | do { | |
15e47304 | 2902 | struct nl_portid_hash *hash = &nl_table[i].hash; |
1da177e4 LT |
2903 | |
2904 | for (; j <= hash->mask; j++) { | |
2905 | s = sk_head(&hash->table[j]); | |
da12c90e G |
2906 | |
2907 | while (s && !nl_table[s->sk_protocol].compare(net, s)) | |
b4b51029 | 2908 | s = sk_next(s); |
1da177e4 LT |
2909 | if (s) { |
2910 | iter->link = i; | |
2911 | iter->hash_idx = j; | |
2912 | return s; | |
2913 | } | |
2914 | } | |
2915 | ||
2916 | j = 0; | |
2917 | } while (++i < MAX_LINKS); | |
2918 | ||
2919 | return NULL; | |
2920 | } | |
2921 | ||
2922 | static void netlink_seq_stop(struct seq_file *seq, void *v) | |
9a429c49 | 2923 | __releases(nl_table_lock) |
1da177e4 LT |
2924 | { |
2925 | read_unlock(&nl_table_lock); | |
2926 | } | |
2927 | ||
2928 | ||
2929 | static int netlink_seq_show(struct seq_file *seq, void *v) | |
2930 | { | |
658cb354 | 2931 | if (v == SEQ_START_TOKEN) { |
1da177e4 LT |
2932 | seq_puts(seq, |
2933 | "sk Eth Pid Groups " | |
cf0aa4e0 | 2934 | "Rmem Wmem Dump Locks Drops Inode\n"); |
658cb354 | 2935 | } else { |
1da177e4 LT |
2936 | struct sock *s = v; |
2937 | struct netlink_sock *nlk = nlk_sk(s); | |
2938 | ||
16b304f3 | 2939 | seq_printf(seq, "%pK %-3d %-6u %08x %-8d %-8d %d %-8d %-8d %-8lu\n", |
1da177e4 LT |
2940 | s, |
2941 | s->sk_protocol, | |
15e47304 | 2942 | nlk->portid, |
513c2500 | 2943 | nlk->groups ? (u32)nlk->groups[0] : 0, |
31e6d363 ED |
2944 | sk_rmem_alloc_get(s), |
2945 | sk_wmem_alloc_get(s), | |
16b304f3 | 2946 | nlk->cb_running, |
38938bfe | 2947 | atomic_read(&s->sk_refcnt), |
cf0aa4e0 MY |
2948 | atomic_read(&s->sk_drops), |
2949 | sock_i_ino(s) | |
1da177e4 LT |
2950 | ); |
2951 | ||
2952 | } | |
2953 | return 0; | |
2954 | } | |
2955 | ||
56b3d975 | 2956 | static const struct seq_operations netlink_seq_ops = { |
1da177e4 LT |
2957 | .start = netlink_seq_start, |
2958 | .next = netlink_seq_next, | |
2959 | .stop = netlink_seq_stop, | |
2960 | .show = netlink_seq_show, | |
2961 | }; | |
2962 | ||
2963 | ||
2964 | static int netlink_seq_open(struct inode *inode, struct file *file) | |
2965 | { | |
e372c414 DL |
2966 | return seq_open_net(inode, file, &netlink_seq_ops, |
2967 | sizeof(struct nl_seq_iter)); | |
b4b51029 EB |
2968 | } |
2969 | ||
da7071d7 | 2970 | static const struct file_operations netlink_seq_fops = { |
1da177e4 LT |
2971 | .owner = THIS_MODULE, |
2972 | .open = netlink_seq_open, | |
2973 | .read = seq_read, | |
2974 | .llseek = seq_lseek, | |
e372c414 | 2975 | .release = seq_release_net, |
1da177e4 LT |
2976 | }; |
2977 | ||
2978 | #endif | |
2979 | ||
2980 | int netlink_register_notifier(struct notifier_block *nb) | |
2981 | { | |
e041c683 | 2982 | return atomic_notifier_chain_register(&netlink_chain, nb); |
1da177e4 | 2983 | } |
6ac552fd | 2984 | EXPORT_SYMBOL(netlink_register_notifier); |
1da177e4 LT |
2985 | |
2986 | int netlink_unregister_notifier(struct notifier_block *nb) | |
2987 | { | |
e041c683 | 2988 | return atomic_notifier_chain_unregister(&netlink_chain, nb); |
1da177e4 | 2989 | } |
6ac552fd | 2990 | EXPORT_SYMBOL(netlink_unregister_notifier); |
746fac4d | 2991 | |
90ddc4f0 | 2992 | static const struct proto_ops netlink_ops = { |
1da177e4 LT |
2993 | .family = PF_NETLINK, |
2994 | .owner = THIS_MODULE, | |
2995 | .release = netlink_release, | |
2996 | .bind = netlink_bind, | |
2997 | .connect = netlink_connect, | |
2998 | .socketpair = sock_no_socketpair, | |
2999 | .accept = sock_no_accept, | |
3000 | .getname = netlink_getname, | |
9652e931 | 3001 | .poll = netlink_poll, |
1da177e4 LT |
3002 | .ioctl = sock_no_ioctl, |
3003 | .listen = sock_no_listen, | |
3004 | .shutdown = sock_no_shutdown, | |
9a4595bc PM |
3005 | .setsockopt = netlink_setsockopt, |
3006 | .getsockopt = netlink_getsockopt, | |
1da177e4 LT |
3007 | .sendmsg = netlink_sendmsg, |
3008 | .recvmsg = netlink_recvmsg, | |
ccdfcc39 | 3009 | .mmap = netlink_mmap, |
1da177e4 LT |
3010 | .sendpage = sock_no_sendpage, |
3011 | }; | |
3012 | ||
ec1b4cf7 | 3013 | static const struct net_proto_family netlink_family_ops = { |
1da177e4 LT |
3014 | .family = PF_NETLINK, |
3015 | .create = netlink_create, | |
3016 | .owner = THIS_MODULE, /* for consistency 8) */ | |
3017 | }; | |
3018 | ||
4665079c | 3019 | static int __net_init netlink_net_init(struct net *net) |
b4b51029 EB |
3020 | { |
3021 | #ifdef CONFIG_PROC_FS | |
d4beaa66 | 3022 | if (!proc_create("netlink", 0, net->proc_net, &netlink_seq_fops)) |
b4b51029 EB |
3023 | return -ENOMEM; |
3024 | #endif | |
3025 | return 0; | |
3026 | } | |
3027 | ||
4665079c | 3028 | static void __net_exit netlink_net_exit(struct net *net) |
b4b51029 EB |
3029 | { |
3030 | #ifdef CONFIG_PROC_FS | |
ece31ffd | 3031 | remove_proc_entry("netlink", net->proc_net); |
b4b51029 EB |
3032 | #endif |
3033 | } | |
3034 | ||
b963ea89 DM |
3035 | static void __init netlink_add_usersock_entry(void) |
3036 | { | |
5c398dc8 | 3037 | struct listeners *listeners; |
b963ea89 DM |
3038 | int groups = 32; |
3039 | ||
5c398dc8 | 3040 | listeners = kzalloc(sizeof(*listeners) + NLGRPSZ(groups), GFP_KERNEL); |
b963ea89 | 3041 | if (!listeners) |
5c398dc8 | 3042 | panic("netlink_add_usersock_entry: Cannot allocate listeners\n"); |
b963ea89 DM |
3043 | |
3044 | netlink_table_grab(); | |
3045 | ||
3046 | nl_table[NETLINK_USERSOCK].groups = groups; | |
5c398dc8 | 3047 | rcu_assign_pointer(nl_table[NETLINK_USERSOCK].listeners, listeners); |
b963ea89 DM |
3048 | nl_table[NETLINK_USERSOCK].module = THIS_MODULE; |
3049 | nl_table[NETLINK_USERSOCK].registered = 1; | |
9785e10a | 3050 | nl_table[NETLINK_USERSOCK].flags = NL_CFG_F_NONROOT_SEND; |
b963ea89 DM |
3051 | |
3052 | netlink_table_ungrab(); | |
3053 | } | |
3054 | ||
022cbae6 | 3055 | static struct pernet_operations __net_initdata netlink_net_ops = { |
b4b51029 EB |
3056 | .init = netlink_net_init, |
3057 | .exit = netlink_net_exit, | |
3058 | }; | |
3059 | ||
1da177e4 LT |
3060 | static int __init netlink_proto_init(void) |
3061 | { | |
1da177e4 | 3062 | int i; |
26ff5ddc | 3063 | unsigned long limit; |
1da177e4 LT |
3064 | unsigned int order; |
3065 | int err = proto_register(&netlink_proto, 0); | |
3066 | ||
3067 | if (err != 0) | |
3068 | goto out; | |
3069 | ||
fab25745 | 3070 | BUILD_BUG_ON(sizeof(struct netlink_skb_parms) > FIELD_SIZEOF(struct sk_buff, cb)); |
1da177e4 | 3071 | |
0da974f4 | 3072 | nl_table = kcalloc(MAX_LINKS, sizeof(*nl_table), GFP_KERNEL); |
fab2caf6 AM |
3073 | if (!nl_table) |
3074 | goto panic; | |
1da177e4 | 3075 | |
4481374c JB |
3076 | if (totalram_pages >= (128 * 1024)) |
3077 | limit = totalram_pages >> (21 - PAGE_SHIFT); | |
1da177e4 | 3078 | else |
4481374c | 3079 | limit = totalram_pages >> (23 - PAGE_SHIFT); |
1da177e4 | 3080 | |
26ff5ddc DC |
3081 | order = get_bitmask_order(limit) - 1 + PAGE_SHIFT; |
3082 | limit = (1UL << order) / sizeof(struct hlist_head); | |
3083 | order = get_bitmask_order(min(limit, (unsigned long)UINT_MAX)) - 1; | |
1da177e4 LT |
3084 | |
3085 | for (i = 0; i < MAX_LINKS; i++) { | |
15e47304 | 3086 | struct nl_portid_hash *hash = &nl_table[i].hash; |
1da177e4 | 3087 | |
15e47304 | 3088 | hash->table = nl_portid_hash_zalloc(1 * sizeof(*hash->table)); |
1da177e4 LT |
3089 | if (!hash->table) { |
3090 | while (i-- > 0) | |
15e47304 | 3091 | nl_portid_hash_free(nl_table[i].hash.table, |
1da177e4 LT |
3092 | 1 * sizeof(*hash->table)); |
3093 | kfree(nl_table); | |
fab2caf6 | 3094 | goto panic; |
1da177e4 | 3095 | } |
1da177e4 LT |
3096 | hash->max_shift = order; |
3097 | hash->shift = 0; | |
3098 | hash->mask = 0; | |
3099 | hash->rehash_time = jiffies; | |
ca15febf G |
3100 | |
3101 | nl_table[i].compare = netlink_compare; | |
1da177e4 LT |
3102 | } |
3103 | ||
bcbde0d4 DB |
3104 | INIT_LIST_HEAD(&netlink_tap_all); |
3105 | ||
b963ea89 DM |
3106 | netlink_add_usersock_entry(); |
3107 | ||
1da177e4 | 3108 | sock_register(&netlink_family_ops); |
b4b51029 | 3109 | register_pernet_subsys(&netlink_net_ops); |
746fac4d | 3110 | /* The netlink device handler may be needed early. */ |
1da177e4 LT |
3111 | rtnetlink_init(); |
3112 | out: | |
3113 | return err; | |
fab2caf6 AM |
3114 | panic: |
3115 | panic("netlink_init: Cannot allocate nl_table\n"); | |
1da177e4 LT |
3116 | } |
3117 | ||
1da177e4 | 3118 | core_initcall(netlink_proto_init); |