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 | ||
2173f8d9 | 134 | static int __netlink_remove_tap(struct netlink_tap *nt) |
bcbde0d4 DB |
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 | } | |
bcbde0d4 DB |
158 | |
159 | int netlink_remove_tap(struct netlink_tap *nt) | |
160 | { | |
161 | int ret; | |
162 | ||
163 | ret = __netlink_remove_tap(nt); | |
164 | synchronize_net(); | |
165 | ||
166 | return ret; | |
167 | } | |
168 | EXPORT_SYMBOL_GPL(netlink_remove_tap); | |
169 | ||
5ffd5cdd DB |
170 | static bool netlink_filter_tap(const struct sk_buff *skb) |
171 | { | |
172 | struct sock *sk = skb->sk; | |
173 | bool pass = false; | |
174 | ||
175 | /* We take the more conservative approach and | |
176 | * whitelist socket protocols that may pass. | |
177 | */ | |
178 | switch (sk->sk_protocol) { | |
179 | case NETLINK_ROUTE: | |
180 | case NETLINK_USERSOCK: | |
181 | case NETLINK_SOCK_DIAG: | |
182 | case NETLINK_NFLOG: | |
183 | case NETLINK_XFRM: | |
184 | case NETLINK_FIB_LOOKUP: | |
185 | case NETLINK_NETFILTER: | |
186 | case NETLINK_GENERIC: | |
187 | pass = true; | |
188 | break; | |
189 | } | |
190 | ||
191 | return pass; | |
192 | } | |
193 | ||
bcbde0d4 DB |
194 | static int __netlink_deliver_tap_skb(struct sk_buff *skb, |
195 | struct net_device *dev) | |
196 | { | |
197 | struct sk_buff *nskb; | |
5ffd5cdd | 198 | struct sock *sk = skb->sk; |
bcbde0d4 DB |
199 | int ret = -ENOMEM; |
200 | ||
201 | dev_hold(dev); | |
202 | nskb = skb_clone(skb, GFP_ATOMIC); | |
203 | if (nskb) { | |
204 | nskb->dev = dev; | |
5ffd5cdd | 205 | nskb->protocol = htons((u16) sk->sk_protocol); |
604d13c9 DB |
206 | nskb->pkt_type = netlink_is_kernel(sk) ? |
207 | PACKET_KERNEL : PACKET_USER; | |
5ffd5cdd | 208 | |
bcbde0d4 DB |
209 | ret = dev_queue_xmit(nskb); |
210 | if (unlikely(ret > 0)) | |
211 | ret = net_xmit_errno(ret); | |
212 | } | |
213 | ||
214 | dev_put(dev); | |
215 | return ret; | |
216 | } | |
217 | ||
218 | static void __netlink_deliver_tap(struct sk_buff *skb) | |
219 | { | |
220 | int ret; | |
221 | struct netlink_tap *tmp; | |
222 | ||
5ffd5cdd DB |
223 | if (!netlink_filter_tap(skb)) |
224 | return; | |
225 | ||
bcbde0d4 DB |
226 | list_for_each_entry_rcu(tmp, &netlink_tap_all, list) { |
227 | ret = __netlink_deliver_tap_skb(skb, tmp->dev); | |
228 | if (unlikely(ret)) | |
229 | break; | |
230 | } | |
231 | } | |
232 | ||
233 | static void netlink_deliver_tap(struct sk_buff *skb) | |
234 | { | |
235 | rcu_read_lock(); | |
236 | ||
237 | if (unlikely(!list_empty(&netlink_tap_all))) | |
238 | __netlink_deliver_tap(skb); | |
239 | ||
240 | rcu_read_unlock(); | |
241 | } | |
242 | ||
73bfd370 DB |
243 | static void netlink_deliver_tap_kernel(struct sock *dst, struct sock *src, |
244 | struct sk_buff *skb) | |
245 | { | |
246 | if (!(netlink_is_kernel(dst) && netlink_is_kernel(src))) | |
247 | netlink_deliver_tap(skb); | |
248 | } | |
249 | ||
cd1df525 PM |
250 | static void netlink_overrun(struct sock *sk) |
251 | { | |
252 | struct netlink_sock *nlk = nlk_sk(sk); | |
253 | ||
254 | if (!(nlk->flags & NETLINK_RECV_NO_ENOBUFS)) { | |
255 | if (!test_and_set_bit(NETLINK_CONGESTED, &nlk_sk(sk)->state)) { | |
256 | sk->sk_err = ENOBUFS; | |
257 | sk->sk_error_report(sk); | |
258 | } | |
259 | } | |
260 | atomic_inc(&sk->sk_drops); | |
261 | } | |
262 | ||
263 | static void netlink_rcv_wake(struct sock *sk) | |
264 | { | |
265 | struct netlink_sock *nlk = nlk_sk(sk); | |
266 | ||
267 | if (skb_queue_empty(&sk->sk_receive_queue)) | |
268 | clear_bit(NETLINK_CONGESTED, &nlk->state); | |
269 | if (!test_bit(NETLINK_CONGESTED, &nlk->state)) | |
270 | wake_up_interruptible(&nlk->wait); | |
271 | } | |
272 | ||
ccdfcc39 | 273 | #ifdef CONFIG_NETLINK_MMAP |
9652e931 PM |
274 | static bool netlink_skb_is_mmaped(const struct sk_buff *skb) |
275 | { | |
276 | return NETLINK_CB(skb).flags & NETLINK_SKB_MMAPED; | |
277 | } | |
278 | ||
f9c22888 PM |
279 | static bool netlink_rx_is_mmaped(struct sock *sk) |
280 | { | |
281 | return nlk_sk(sk)->rx_ring.pg_vec != NULL; | |
282 | } | |
283 | ||
5fd96123 PM |
284 | static bool netlink_tx_is_mmaped(struct sock *sk) |
285 | { | |
286 | return nlk_sk(sk)->tx_ring.pg_vec != NULL; | |
287 | } | |
288 | ||
ccdfcc39 PM |
289 | static __pure struct page *pgvec_to_page(const void *addr) |
290 | { | |
291 | if (is_vmalloc_addr(addr)) | |
292 | return vmalloc_to_page(addr); | |
293 | else | |
294 | return virt_to_page(addr); | |
295 | } | |
296 | ||
297 | static void free_pg_vec(void **pg_vec, unsigned int order, unsigned int len) | |
298 | { | |
299 | unsigned int i; | |
300 | ||
301 | for (i = 0; i < len; i++) { | |
302 | if (pg_vec[i] != NULL) { | |
303 | if (is_vmalloc_addr(pg_vec[i])) | |
304 | vfree(pg_vec[i]); | |
305 | else | |
306 | free_pages((unsigned long)pg_vec[i], order); | |
307 | } | |
308 | } | |
309 | kfree(pg_vec); | |
310 | } | |
311 | ||
312 | static void *alloc_one_pg_vec_page(unsigned long order) | |
313 | { | |
314 | void *buffer; | |
315 | gfp_t gfp_flags = GFP_KERNEL | __GFP_COMP | __GFP_ZERO | | |
316 | __GFP_NOWARN | __GFP_NORETRY; | |
317 | ||
318 | buffer = (void *)__get_free_pages(gfp_flags, order); | |
319 | if (buffer != NULL) | |
320 | return buffer; | |
321 | ||
322 | buffer = vzalloc((1 << order) * PAGE_SIZE); | |
323 | if (buffer != NULL) | |
324 | return buffer; | |
325 | ||
326 | gfp_flags &= ~__GFP_NORETRY; | |
327 | return (void *)__get_free_pages(gfp_flags, order); | |
328 | } | |
329 | ||
330 | static void **alloc_pg_vec(struct netlink_sock *nlk, | |
331 | struct nl_mmap_req *req, unsigned int order) | |
332 | { | |
333 | unsigned int block_nr = req->nm_block_nr; | |
334 | unsigned int i; | |
8a849bb7 | 335 | void **pg_vec; |
ccdfcc39 PM |
336 | |
337 | pg_vec = kcalloc(block_nr, sizeof(void *), GFP_KERNEL); | |
338 | if (pg_vec == NULL) | |
339 | return NULL; | |
340 | ||
341 | for (i = 0; i < block_nr; i++) { | |
8a849bb7 | 342 | pg_vec[i] = alloc_one_pg_vec_page(order); |
ccdfcc39 PM |
343 | if (pg_vec[i] == NULL) |
344 | goto err1; | |
345 | } | |
346 | ||
347 | return pg_vec; | |
348 | err1: | |
349 | free_pg_vec(pg_vec, order, block_nr); | |
350 | return NULL; | |
351 | } | |
352 | ||
353 | static int netlink_set_ring(struct sock *sk, struct nl_mmap_req *req, | |
354 | bool closing, bool tx_ring) | |
355 | { | |
356 | struct netlink_sock *nlk = nlk_sk(sk); | |
357 | struct netlink_ring *ring; | |
358 | struct sk_buff_head *queue; | |
359 | void **pg_vec = NULL; | |
360 | unsigned int order = 0; | |
361 | int err; | |
362 | ||
363 | ring = tx_ring ? &nlk->tx_ring : &nlk->rx_ring; | |
364 | queue = tx_ring ? &sk->sk_write_queue : &sk->sk_receive_queue; | |
365 | ||
366 | if (!closing) { | |
367 | if (atomic_read(&nlk->mapped)) | |
368 | return -EBUSY; | |
369 | if (atomic_read(&ring->pending)) | |
370 | return -EBUSY; | |
371 | } | |
372 | ||
373 | if (req->nm_block_nr) { | |
374 | if (ring->pg_vec != NULL) | |
375 | return -EBUSY; | |
376 | ||
377 | if ((int)req->nm_block_size <= 0) | |
378 | return -EINVAL; | |
379 | if (!IS_ALIGNED(req->nm_block_size, PAGE_SIZE)) | |
380 | return -EINVAL; | |
381 | if (req->nm_frame_size < NL_MMAP_HDRLEN) | |
382 | return -EINVAL; | |
383 | if (!IS_ALIGNED(req->nm_frame_size, NL_MMAP_MSG_ALIGNMENT)) | |
384 | return -EINVAL; | |
385 | ||
386 | ring->frames_per_block = req->nm_block_size / | |
387 | req->nm_frame_size; | |
388 | if (ring->frames_per_block == 0) | |
389 | return -EINVAL; | |
390 | if (ring->frames_per_block * req->nm_block_nr != | |
391 | req->nm_frame_nr) | |
392 | return -EINVAL; | |
393 | ||
394 | order = get_order(req->nm_block_size); | |
395 | pg_vec = alloc_pg_vec(nlk, req, order); | |
396 | if (pg_vec == NULL) | |
397 | return -ENOMEM; | |
398 | } else { | |
399 | if (req->nm_frame_nr) | |
400 | return -EINVAL; | |
401 | } | |
402 | ||
403 | err = -EBUSY; | |
404 | mutex_lock(&nlk->pg_vec_lock); | |
405 | if (closing || atomic_read(&nlk->mapped) == 0) { | |
406 | err = 0; | |
407 | spin_lock_bh(&queue->lock); | |
408 | ||
409 | ring->frame_max = req->nm_frame_nr - 1; | |
410 | ring->head = 0; | |
411 | ring->frame_size = req->nm_frame_size; | |
412 | ring->pg_vec_pages = req->nm_block_size / PAGE_SIZE; | |
413 | ||
414 | swap(ring->pg_vec_len, req->nm_block_nr); | |
415 | swap(ring->pg_vec_order, order); | |
416 | swap(ring->pg_vec, pg_vec); | |
417 | ||
418 | __skb_queue_purge(queue); | |
419 | spin_unlock_bh(&queue->lock); | |
420 | ||
421 | WARN_ON(atomic_read(&nlk->mapped)); | |
422 | } | |
423 | mutex_unlock(&nlk->pg_vec_lock); | |
424 | ||
425 | if (pg_vec) | |
426 | free_pg_vec(pg_vec, order, req->nm_block_nr); | |
427 | return err; | |
428 | } | |
429 | ||
430 | static void netlink_mm_open(struct vm_area_struct *vma) | |
431 | { | |
432 | struct file *file = vma->vm_file; | |
433 | struct socket *sock = file->private_data; | |
434 | struct sock *sk = sock->sk; | |
435 | ||
436 | if (sk) | |
437 | atomic_inc(&nlk_sk(sk)->mapped); | |
438 | } | |
439 | ||
440 | static void netlink_mm_close(struct vm_area_struct *vma) | |
441 | { | |
442 | struct file *file = vma->vm_file; | |
443 | struct socket *sock = file->private_data; | |
444 | struct sock *sk = sock->sk; | |
445 | ||
446 | if (sk) | |
447 | atomic_dec(&nlk_sk(sk)->mapped); | |
448 | } | |
449 | ||
450 | static const struct vm_operations_struct netlink_mmap_ops = { | |
451 | .open = netlink_mm_open, | |
452 | .close = netlink_mm_close, | |
453 | }; | |
454 | ||
455 | static int netlink_mmap(struct file *file, struct socket *sock, | |
456 | struct vm_area_struct *vma) | |
457 | { | |
458 | struct sock *sk = sock->sk; | |
459 | struct netlink_sock *nlk = nlk_sk(sk); | |
460 | struct netlink_ring *ring; | |
461 | unsigned long start, size, expected; | |
462 | unsigned int i; | |
463 | int err = -EINVAL; | |
464 | ||
465 | if (vma->vm_pgoff) | |
466 | return -EINVAL; | |
467 | ||
468 | mutex_lock(&nlk->pg_vec_lock); | |
469 | ||
470 | expected = 0; | |
471 | for (ring = &nlk->rx_ring; ring <= &nlk->tx_ring; ring++) { | |
472 | if (ring->pg_vec == NULL) | |
473 | continue; | |
474 | expected += ring->pg_vec_len * ring->pg_vec_pages * PAGE_SIZE; | |
475 | } | |
476 | ||
477 | if (expected == 0) | |
478 | goto out; | |
479 | ||
480 | size = vma->vm_end - vma->vm_start; | |
481 | if (size != expected) | |
482 | goto out; | |
483 | ||
484 | start = vma->vm_start; | |
485 | for (ring = &nlk->rx_ring; ring <= &nlk->tx_ring; ring++) { | |
486 | if (ring->pg_vec == NULL) | |
487 | continue; | |
488 | ||
489 | for (i = 0; i < ring->pg_vec_len; i++) { | |
490 | struct page *page; | |
491 | void *kaddr = ring->pg_vec[i]; | |
492 | unsigned int pg_num; | |
493 | ||
494 | for (pg_num = 0; pg_num < ring->pg_vec_pages; pg_num++) { | |
495 | page = pgvec_to_page(kaddr); | |
496 | err = vm_insert_page(vma, start, page); | |
497 | if (err < 0) | |
498 | goto out; | |
499 | start += PAGE_SIZE; | |
500 | kaddr += PAGE_SIZE; | |
501 | } | |
502 | } | |
503 | } | |
504 | ||
505 | atomic_inc(&nlk->mapped); | |
506 | vma->vm_ops = &netlink_mmap_ops; | |
507 | err = 0; | |
508 | out: | |
509 | mutex_unlock(&nlk->pg_vec_lock); | |
7cdbac71 | 510 | return err; |
ccdfcc39 | 511 | } |
9652e931 PM |
512 | |
513 | static void netlink_frame_flush_dcache(const struct nl_mmap_hdr *hdr) | |
514 | { | |
515 | #if ARCH_IMPLEMENTS_FLUSH_DCACHE_PAGE == 1 | |
516 | struct page *p_start, *p_end; | |
517 | ||
518 | /* First page is flushed through netlink_{get,set}_status */ | |
519 | p_start = pgvec_to_page(hdr + PAGE_SIZE); | |
1d5085cb | 520 | p_end = pgvec_to_page((void *)hdr + NL_MMAP_HDRLEN + hdr->nm_len - 1); |
9652e931 PM |
521 | while (p_start <= p_end) { |
522 | flush_dcache_page(p_start); | |
523 | p_start++; | |
524 | } | |
525 | #endif | |
526 | } | |
527 | ||
528 | static enum nl_mmap_status netlink_get_status(const struct nl_mmap_hdr *hdr) | |
529 | { | |
530 | smp_rmb(); | |
531 | flush_dcache_page(pgvec_to_page(hdr)); | |
532 | return hdr->nm_status; | |
533 | } | |
534 | ||
535 | static void netlink_set_status(struct nl_mmap_hdr *hdr, | |
536 | enum nl_mmap_status status) | |
537 | { | |
538 | hdr->nm_status = status; | |
539 | flush_dcache_page(pgvec_to_page(hdr)); | |
540 | smp_wmb(); | |
541 | } | |
542 | ||
543 | static struct nl_mmap_hdr * | |
544 | __netlink_lookup_frame(const struct netlink_ring *ring, unsigned int pos) | |
545 | { | |
546 | unsigned int pg_vec_pos, frame_off; | |
547 | ||
548 | pg_vec_pos = pos / ring->frames_per_block; | |
549 | frame_off = pos % ring->frames_per_block; | |
550 | ||
551 | return ring->pg_vec[pg_vec_pos] + (frame_off * ring->frame_size); | |
552 | } | |
553 | ||
554 | static struct nl_mmap_hdr * | |
555 | netlink_lookup_frame(const struct netlink_ring *ring, unsigned int pos, | |
556 | enum nl_mmap_status status) | |
557 | { | |
558 | struct nl_mmap_hdr *hdr; | |
559 | ||
560 | hdr = __netlink_lookup_frame(ring, pos); | |
561 | if (netlink_get_status(hdr) != status) | |
562 | return NULL; | |
563 | ||
564 | return hdr; | |
565 | } | |
566 | ||
567 | static struct nl_mmap_hdr * | |
568 | netlink_current_frame(const struct netlink_ring *ring, | |
569 | enum nl_mmap_status status) | |
570 | { | |
571 | return netlink_lookup_frame(ring, ring->head, status); | |
572 | } | |
573 | ||
574 | static struct nl_mmap_hdr * | |
575 | netlink_previous_frame(const struct netlink_ring *ring, | |
576 | enum nl_mmap_status status) | |
577 | { | |
578 | unsigned int prev; | |
579 | ||
580 | prev = ring->head ? ring->head - 1 : ring->frame_max; | |
581 | return netlink_lookup_frame(ring, prev, status); | |
582 | } | |
583 | ||
584 | static void netlink_increment_head(struct netlink_ring *ring) | |
585 | { | |
586 | ring->head = ring->head != ring->frame_max ? ring->head + 1 : 0; | |
587 | } | |
588 | ||
589 | static void netlink_forward_ring(struct netlink_ring *ring) | |
590 | { | |
591 | unsigned int head = ring->head, pos = head; | |
592 | const struct nl_mmap_hdr *hdr; | |
593 | ||
594 | do { | |
595 | hdr = __netlink_lookup_frame(ring, pos); | |
596 | if (hdr->nm_status == NL_MMAP_STATUS_UNUSED) | |
597 | break; | |
598 | if (hdr->nm_status != NL_MMAP_STATUS_SKIP) | |
599 | break; | |
600 | netlink_increment_head(ring); | |
601 | } while (ring->head != head); | |
602 | } | |
603 | ||
cd1df525 PM |
604 | static bool netlink_dump_space(struct netlink_sock *nlk) |
605 | { | |
606 | struct netlink_ring *ring = &nlk->rx_ring; | |
607 | struct nl_mmap_hdr *hdr; | |
608 | unsigned int n; | |
609 | ||
610 | hdr = netlink_current_frame(ring, NL_MMAP_STATUS_UNUSED); | |
611 | if (hdr == NULL) | |
612 | return false; | |
613 | ||
614 | n = ring->head + ring->frame_max / 2; | |
615 | if (n > ring->frame_max) | |
616 | n -= ring->frame_max; | |
617 | ||
618 | hdr = __netlink_lookup_frame(ring, n); | |
619 | ||
620 | return hdr->nm_status == NL_MMAP_STATUS_UNUSED; | |
621 | } | |
622 | ||
9652e931 PM |
623 | static unsigned int netlink_poll(struct file *file, struct socket *sock, |
624 | poll_table *wait) | |
625 | { | |
626 | struct sock *sk = sock->sk; | |
627 | struct netlink_sock *nlk = nlk_sk(sk); | |
628 | unsigned int mask; | |
cd1df525 | 629 | int err; |
9652e931 | 630 | |
cd1df525 PM |
631 | if (nlk->rx_ring.pg_vec != NULL) { |
632 | /* Memory mapped sockets don't call recvmsg(), so flow control | |
633 | * for dumps is performed here. A dump is allowed to continue | |
634 | * if at least half the ring is unused. | |
635 | */ | |
16b304f3 | 636 | while (nlk->cb_running && netlink_dump_space(nlk)) { |
cd1df525 PM |
637 | err = netlink_dump(sk); |
638 | if (err < 0) { | |
639 | sk->sk_err = err; | |
640 | sk->sk_error_report(sk); | |
641 | break; | |
642 | } | |
643 | } | |
644 | netlink_rcv_wake(sk); | |
645 | } | |
5fd96123 | 646 | |
9652e931 PM |
647 | mask = datagram_poll(file, sock, wait); |
648 | ||
649 | spin_lock_bh(&sk->sk_receive_queue.lock); | |
650 | if (nlk->rx_ring.pg_vec) { | |
651 | netlink_forward_ring(&nlk->rx_ring); | |
652 | if (!netlink_previous_frame(&nlk->rx_ring, NL_MMAP_STATUS_UNUSED)) | |
653 | mask |= POLLIN | POLLRDNORM; | |
654 | } | |
655 | spin_unlock_bh(&sk->sk_receive_queue.lock); | |
656 | ||
657 | spin_lock_bh(&sk->sk_write_queue.lock); | |
658 | if (nlk->tx_ring.pg_vec) { | |
659 | if (netlink_current_frame(&nlk->tx_ring, NL_MMAP_STATUS_UNUSED)) | |
660 | mask |= POLLOUT | POLLWRNORM; | |
661 | } | |
662 | spin_unlock_bh(&sk->sk_write_queue.lock); | |
663 | ||
664 | return mask; | |
665 | } | |
666 | ||
667 | static struct nl_mmap_hdr *netlink_mmap_hdr(struct sk_buff *skb) | |
668 | { | |
669 | return (struct nl_mmap_hdr *)(skb->head - NL_MMAP_HDRLEN); | |
670 | } | |
671 | ||
672 | static void netlink_ring_setup_skb(struct sk_buff *skb, struct sock *sk, | |
673 | struct netlink_ring *ring, | |
674 | struct nl_mmap_hdr *hdr) | |
675 | { | |
676 | unsigned int size; | |
677 | void *data; | |
678 | ||
679 | size = ring->frame_size - NL_MMAP_HDRLEN; | |
680 | data = (void *)hdr + NL_MMAP_HDRLEN; | |
681 | ||
682 | skb->head = data; | |
683 | skb->data = data; | |
684 | skb_reset_tail_pointer(skb); | |
685 | skb->end = skb->tail + size; | |
686 | skb->len = 0; | |
687 | ||
688 | skb->destructor = netlink_skb_destructor; | |
689 | NETLINK_CB(skb).flags |= NETLINK_SKB_MMAPED; | |
690 | NETLINK_CB(skb).sk = sk; | |
691 | } | |
5fd96123 PM |
692 | |
693 | static int netlink_mmap_sendmsg(struct sock *sk, struct msghdr *msg, | |
694 | u32 dst_portid, u32 dst_group, | |
695 | struct sock_iocb *siocb) | |
696 | { | |
697 | struct netlink_sock *nlk = nlk_sk(sk); | |
698 | struct netlink_ring *ring; | |
699 | struct nl_mmap_hdr *hdr; | |
700 | struct sk_buff *skb; | |
701 | unsigned int maxlen; | |
702 | bool excl = true; | |
703 | int err = 0, len = 0; | |
704 | ||
705 | /* Netlink messages are validated by the receiver before processing. | |
706 | * In order to avoid userspace changing the contents of the message | |
707 | * after validation, the socket and the ring may only be used by a | |
708 | * single process, otherwise we fall back to copying. | |
709 | */ | |
710 | if (atomic_long_read(&sk->sk_socket->file->f_count) > 2 || | |
711 | atomic_read(&nlk->mapped) > 1) | |
712 | excl = false; | |
713 | ||
714 | mutex_lock(&nlk->pg_vec_lock); | |
715 | ||
716 | ring = &nlk->tx_ring; | |
717 | maxlen = ring->frame_size - NL_MMAP_HDRLEN; | |
718 | ||
719 | do { | |
720 | hdr = netlink_current_frame(ring, NL_MMAP_STATUS_VALID); | |
721 | if (hdr == NULL) { | |
722 | if (!(msg->msg_flags & MSG_DONTWAIT) && | |
723 | atomic_read(&nlk->tx_ring.pending)) | |
724 | schedule(); | |
725 | continue; | |
726 | } | |
727 | if (hdr->nm_len > maxlen) { | |
728 | err = -EINVAL; | |
729 | goto out; | |
730 | } | |
731 | ||
732 | netlink_frame_flush_dcache(hdr); | |
733 | ||
734 | if (likely(dst_portid == 0 && dst_group == 0 && excl)) { | |
735 | skb = alloc_skb_head(GFP_KERNEL); | |
736 | if (skb == NULL) { | |
737 | err = -ENOBUFS; | |
738 | goto out; | |
739 | } | |
740 | sock_hold(sk); | |
741 | netlink_ring_setup_skb(skb, sk, ring, hdr); | |
742 | NETLINK_CB(skb).flags |= NETLINK_SKB_TX; | |
743 | __skb_put(skb, hdr->nm_len); | |
744 | netlink_set_status(hdr, NL_MMAP_STATUS_RESERVED); | |
745 | atomic_inc(&ring->pending); | |
746 | } else { | |
747 | skb = alloc_skb(hdr->nm_len, GFP_KERNEL); | |
748 | if (skb == NULL) { | |
749 | err = -ENOBUFS; | |
750 | goto out; | |
751 | } | |
752 | __skb_put(skb, hdr->nm_len); | |
753 | memcpy(skb->data, (void *)hdr + NL_MMAP_HDRLEN, hdr->nm_len); | |
754 | netlink_set_status(hdr, NL_MMAP_STATUS_UNUSED); | |
755 | } | |
756 | ||
757 | netlink_increment_head(ring); | |
758 | ||
759 | NETLINK_CB(skb).portid = nlk->portid; | |
760 | NETLINK_CB(skb).dst_group = dst_group; | |
761 | NETLINK_CB(skb).creds = siocb->scm->creds; | |
762 | ||
763 | err = security_netlink_send(sk, skb); | |
764 | if (err) { | |
765 | kfree_skb(skb); | |
766 | goto out; | |
767 | } | |
768 | ||
769 | if (unlikely(dst_group)) { | |
770 | atomic_inc(&skb->users); | |
771 | netlink_broadcast(sk, skb, dst_portid, dst_group, | |
772 | GFP_KERNEL); | |
773 | } | |
774 | err = netlink_unicast(sk, skb, dst_portid, | |
775 | msg->msg_flags & MSG_DONTWAIT); | |
776 | if (err < 0) | |
777 | goto out; | |
778 | len += err; | |
779 | ||
780 | } while (hdr != NULL || | |
781 | (!(msg->msg_flags & MSG_DONTWAIT) && | |
782 | atomic_read(&nlk->tx_ring.pending))); | |
783 | ||
784 | if (len > 0) | |
785 | err = len; | |
786 | out: | |
787 | mutex_unlock(&nlk->pg_vec_lock); | |
788 | return err; | |
789 | } | |
f9c22888 PM |
790 | |
791 | static void netlink_queue_mmaped_skb(struct sock *sk, struct sk_buff *skb) | |
792 | { | |
793 | struct nl_mmap_hdr *hdr; | |
794 | ||
795 | hdr = netlink_mmap_hdr(skb); | |
796 | hdr->nm_len = skb->len; | |
797 | hdr->nm_group = NETLINK_CB(skb).dst_group; | |
798 | hdr->nm_pid = NETLINK_CB(skb).creds.pid; | |
1bf9310a ND |
799 | hdr->nm_uid = from_kuid(sk_user_ns(sk), NETLINK_CB(skb).creds.uid); |
800 | hdr->nm_gid = from_kgid(sk_user_ns(sk), NETLINK_CB(skb).creds.gid); | |
f9c22888 PM |
801 | netlink_frame_flush_dcache(hdr); |
802 | netlink_set_status(hdr, NL_MMAP_STATUS_VALID); | |
803 | ||
804 | NETLINK_CB(skb).flags |= NETLINK_SKB_DELIVERED; | |
805 | kfree_skb(skb); | |
806 | } | |
807 | ||
808 | static void netlink_ring_set_copied(struct sock *sk, struct sk_buff *skb) | |
809 | { | |
810 | struct netlink_sock *nlk = nlk_sk(sk); | |
811 | struct netlink_ring *ring = &nlk->rx_ring; | |
812 | struct nl_mmap_hdr *hdr; | |
813 | ||
814 | spin_lock_bh(&sk->sk_receive_queue.lock); | |
815 | hdr = netlink_current_frame(ring, NL_MMAP_STATUS_UNUSED); | |
816 | if (hdr == NULL) { | |
817 | spin_unlock_bh(&sk->sk_receive_queue.lock); | |
818 | kfree_skb(skb); | |
cd1df525 | 819 | netlink_overrun(sk); |
f9c22888 PM |
820 | return; |
821 | } | |
822 | netlink_increment_head(ring); | |
823 | __skb_queue_tail(&sk->sk_receive_queue, skb); | |
824 | spin_unlock_bh(&sk->sk_receive_queue.lock); | |
825 | ||
826 | hdr->nm_len = skb->len; | |
827 | hdr->nm_group = NETLINK_CB(skb).dst_group; | |
828 | hdr->nm_pid = NETLINK_CB(skb).creds.pid; | |
1bf9310a ND |
829 | hdr->nm_uid = from_kuid(sk_user_ns(sk), NETLINK_CB(skb).creds.uid); |
830 | hdr->nm_gid = from_kgid(sk_user_ns(sk), NETLINK_CB(skb).creds.gid); | |
f9c22888 PM |
831 | netlink_set_status(hdr, NL_MMAP_STATUS_COPY); |
832 | } | |
833 | ||
ccdfcc39 | 834 | #else /* CONFIG_NETLINK_MMAP */ |
9652e931 | 835 | #define netlink_skb_is_mmaped(skb) false |
f9c22888 | 836 | #define netlink_rx_is_mmaped(sk) false |
5fd96123 | 837 | #define netlink_tx_is_mmaped(sk) false |
ccdfcc39 | 838 | #define netlink_mmap sock_no_mmap |
9652e931 | 839 | #define netlink_poll datagram_poll |
5fd96123 | 840 | #define netlink_mmap_sendmsg(sk, msg, dst_portid, dst_group, siocb) 0 |
ccdfcc39 PM |
841 | #endif /* CONFIG_NETLINK_MMAP */ |
842 | ||
cf0a018a PM |
843 | static void netlink_skb_destructor(struct sk_buff *skb) |
844 | { | |
9652e931 PM |
845 | #ifdef CONFIG_NETLINK_MMAP |
846 | struct nl_mmap_hdr *hdr; | |
847 | struct netlink_ring *ring; | |
848 | struct sock *sk; | |
849 | ||
850 | /* If a packet from the kernel to userspace was freed because of an | |
851 | * error without being delivered to userspace, the kernel must reset | |
852 | * the status. In the direction userspace to kernel, the status is | |
853 | * always reset here after the packet was processed and freed. | |
854 | */ | |
855 | if (netlink_skb_is_mmaped(skb)) { | |
856 | hdr = netlink_mmap_hdr(skb); | |
857 | sk = NETLINK_CB(skb).sk; | |
858 | ||
5fd96123 PM |
859 | if (NETLINK_CB(skb).flags & NETLINK_SKB_TX) { |
860 | netlink_set_status(hdr, NL_MMAP_STATUS_UNUSED); | |
861 | ring = &nlk_sk(sk)->tx_ring; | |
862 | } else { | |
863 | if (!(NETLINK_CB(skb).flags & NETLINK_SKB_DELIVERED)) { | |
864 | hdr->nm_len = 0; | |
865 | netlink_set_status(hdr, NL_MMAP_STATUS_VALID); | |
866 | } | |
867 | ring = &nlk_sk(sk)->rx_ring; | |
9652e931 | 868 | } |
9652e931 PM |
869 | |
870 | WARN_ON(atomic_read(&ring->pending) == 0); | |
871 | atomic_dec(&ring->pending); | |
872 | sock_put(sk); | |
873 | ||
5e71d9d7 | 874 | skb->head = NULL; |
9652e931 PM |
875 | } |
876 | #endif | |
c05cdb1b | 877 | if (is_vmalloc_addr(skb->head)) { |
3a36515f PN |
878 | if (!skb->cloned || |
879 | !atomic_dec_return(&(skb_shinfo(skb)->dataref))) | |
880 | vfree(skb->head); | |
881 | ||
c05cdb1b PNA |
882 | skb->head = NULL; |
883 | } | |
9652e931 PM |
884 | if (skb->sk != NULL) |
885 | sock_rfree(skb); | |
cf0a018a PM |
886 | } |
887 | ||
888 | static void netlink_skb_set_owner_r(struct sk_buff *skb, struct sock *sk) | |
889 | { | |
890 | WARN_ON(skb->sk != NULL); | |
891 | skb->sk = sk; | |
892 | skb->destructor = netlink_skb_destructor; | |
893 | atomic_add(skb->truesize, &sk->sk_rmem_alloc); | |
894 | sk_mem_charge(sk, skb->truesize); | |
895 | } | |
896 | ||
1da177e4 LT |
897 | static void netlink_sock_destruct(struct sock *sk) |
898 | { | |
3f660d66 HX |
899 | struct netlink_sock *nlk = nlk_sk(sk); |
900 | ||
16b304f3 PS |
901 | if (nlk->cb_running) { |
902 | if (nlk->cb.done) | |
903 | nlk->cb.done(&nlk->cb); | |
6dc878a8 | 904 | |
16b304f3 PS |
905 | module_put(nlk->cb.module); |
906 | kfree_skb(nlk->cb.skb); | |
3f660d66 HX |
907 | } |
908 | ||
1da177e4 | 909 | skb_queue_purge(&sk->sk_receive_queue); |
ccdfcc39 PM |
910 | #ifdef CONFIG_NETLINK_MMAP |
911 | if (1) { | |
912 | struct nl_mmap_req req; | |
913 | ||
914 | memset(&req, 0, sizeof(req)); | |
915 | if (nlk->rx_ring.pg_vec) | |
916 | netlink_set_ring(sk, &req, true, false); | |
917 | memset(&req, 0, sizeof(req)); | |
918 | if (nlk->tx_ring.pg_vec) | |
919 | netlink_set_ring(sk, &req, true, true); | |
920 | } | |
921 | #endif /* CONFIG_NETLINK_MMAP */ | |
1da177e4 LT |
922 | |
923 | if (!sock_flag(sk, SOCK_DEAD)) { | |
6ac552fd | 924 | printk(KERN_ERR "Freeing alive netlink socket %p\n", sk); |
1da177e4 LT |
925 | return; |
926 | } | |
547b792c IJ |
927 | |
928 | WARN_ON(atomic_read(&sk->sk_rmem_alloc)); | |
929 | WARN_ON(atomic_read(&sk->sk_wmem_alloc)); | |
930 | WARN_ON(nlk_sk(sk)->groups); | |
1da177e4 LT |
931 | } |
932 | ||
6ac552fd PM |
933 | /* This lock without WQ_FLAG_EXCLUSIVE is good on UP and it is _very_ bad on |
934 | * SMP. Look, when several writers sleep and reader wakes them up, all but one | |
1da177e4 LT |
935 | * immediately hit write lock and grab all the cpus. Exclusive sleep solves |
936 | * this, _but_ remember, it adds useless work on UP machines. | |
937 | */ | |
938 | ||
d136f1bd | 939 | void netlink_table_grab(void) |
9a429c49 | 940 | __acquires(nl_table_lock) |
1da177e4 | 941 | { |
d136f1bd JB |
942 | might_sleep(); |
943 | ||
6abd219c | 944 | write_lock_irq(&nl_table_lock); |
1da177e4 LT |
945 | |
946 | if (atomic_read(&nl_table_users)) { | |
947 | DECLARE_WAITQUEUE(wait, current); | |
948 | ||
949 | add_wait_queue_exclusive(&nl_table_wait, &wait); | |
6ac552fd | 950 | for (;;) { |
1da177e4 LT |
951 | set_current_state(TASK_UNINTERRUPTIBLE); |
952 | if (atomic_read(&nl_table_users) == 0) | |
953 | break; | |
6abd219c | 954 | write_unlock_irq(&nl_table_lock); |
1da177e4 | 955 | schedule(); |
6abd219c | 956 | write_lock_irq(&nl_table_lock); |
1da177e4 LT |
957 | } |
958 | ||
959 | __set_current_state(TASK_RUNNING); | |
960 | remove_wait_queue(&nl_table_wait, &wait); | |
961 | } | |
962 | } | |
963 | ||
d136f1bd | 964 | void netlink_table_ungrab(void) |
9a429c49 | 965 | __releases(nl_table_lock) |
1da177e4 | 966 | { |
6abd219c | 967 | write_unlock_irq(&nl_table_lock); |
1da177e4 LT |
968 | wake_up(&nl_table_wait); |
969 | } | |
970 | ||
6ac552fd | 971 | static inline void |
1da177e4 LT |
972 | netlink_lock_table(void) |
973 | { | |
974 | /* read_lock() synchronizes us to netlink_table_grab */ | |
975 | ||
976 | read_lock(&nl_table_lock); | |
977 | atomic_inc(&nl_table_users); | |
978 | read_unlock(&nl_table_lock); | |
979 | } | |
980 | ||
6ac552fd | 981 | static inline void |
1da177e4 LT |
982 | netlink_unlock_table(void) |
983 | { | |
984 | if (atomic_dec_and_test(&nl_table_users)) | |
985 | wake_up(&nl_table_wait); | |
986 | } | |
987 | ||
da12c90e G |
988 | static bool netlink_compare(struct net *net, struct sock *sk) |
989 | { | |
990 | return net_eq(sock_net(sk), net); | |
991 | } | |
992 | ||
15e47304 | 993 | static struct sock *netlink_lookup(struct net *net, int protocol, u32 portid) |
1da177e4 | 994 | { |
da12c90e G |
995 | struct netlink_table *table = &nl_table[protocol]; |
996 | struct nl_portid_hash *hash = &table->hash; | |
1da177e4 LT |
997 | struct hlist_head *head; |
998 | struct sock *sk; | |
1da177e4 LT |
999 | |
1000 | read_lock(&nl_table_lock); | |
15e47304 | 1001 | head = nl_portid_hashfn(hash, portid); |
b67bfe0d | 1002 | sk_for_each(sk, head) { |
da12c90e G |
1003 | if (table->compare(net, sk) && |
1004 | (nlk_sk(sk)->portid == portid)) { | |
1da177e4 LT |
1005 | sock_hold(sk); |
1006 | goto found; | |
1007 | } | |
1008 | } | |
1009 | sk = NULL; | |
1010 | found: | |
1011 | read_unlock(&nl_table_lock); | |
1012 | return sk; | |
1013 | } | |
1014 | ||
15e47304 | 1015 | static struct hlist_head *nl_portid_hash_zalloc(size_t size) |
1da177e4 LT |
1016 | { |
1017 | if (size <= PAGE_SIZE) | |
ea72912c | 1018 | return kzalloc(size, GFP_ATOMIC); |
1da177e4 LT |
1019 | else |
1020 | return (struct hlist_head *) | |
ea72912c ED |
1021 | __get_free_pages(GFP_ATOMIC | __GFP_ZERO, |
1022 | get_order(size)); | |
1da177e4 LT |
1023 | } |
1024 | ||
15e47304 | 1025 | static void nl_portid_hash_free(struct hlist_head *table, size_t size) |
1da177e4 LT |
1026 | { |
1027 | if (size <= PAGE_SIZE) | |
1028 | kfree(table); | |
1029 | else | |
1030 | free_pages((unsigned long)table, get_order(size)); | |
1031 | } | |
1032 | ||
15e47304 | 1033 | static int nl_portid_hash_rehash(struct nl_portid_hash *hash, int grow) |
1da177e4 LT |
1034 | { |
1035 | unsigned int omask, mask, shift; | |
1036 | size_t osize, size; | |
1037 | struct hlist_head *otable, *table; | |
1038 | int i; | |
1039 | ||
1040 | omask = mask = hash->mask; | |
1041 | osize = size = (mask + 1) * sizeof(*table); | |
1042 | shift = hash->shift; | |
1043 | ||
1044 | if (grow) { | |
1045 | if (++shift > hash->max_shift) | |
1046 | return 0; | |
1047 | mask = mask * 2 + 1; | |
1048 | size *= 2; | |
1049 | } | |
1050 | ||
15e47304 | 1051 | table = nl_portid_hash_zalloc(size); |
1da177e4 LT |
1052 | if (!table) |
1053 | return 0; | |
1054 | ||
1da177e4 LT |
1055 | otable = hash->table; |
1056 | hash->table = table; | |
1057 | hash->mask = mask; | |
1058 | hash->shift = shift; | |
1059 | get_random_bytes(&hash->rnd, sizeof(hash->rnd)); | |
1060 | ||
1061 | for (i = 0; i <= omask; i++) { | |
1062 | struct sock *sk; | |
b67bfe0d | 1063 | struct hlist_node *tmp; |
1da177e4 | 1064 | |
b67bfe0d | 1065 | sk_for_each_safe(sk, tmp, &otable[i]) |
15e47304 | 1066 | __sk_add_node(sk, nl_portid_hashfn(hash, nlk_sk(sk)->portid)); |
1da177e4 LT |
1067 | } |
1068 | ||
15e47304 | 1069 | nl_portid_hash_free(otable, osize); |
1da177e4 LT |
1070 | hash->rehash_time = jiffies + 10 * 60 * HZ; |
1071 | return 1; | |
1072 | } | |
1073 | ||
15e47304 | 1074 | static inline int nl_portid_hash_dilute(struct nl_portid_hash *hash, int len) |
1da177e4 LT |
1075 | { |
1076 | int avg = hash->entries >> hash->shift; | |
1077 | ||
15e47304 | 1078 | if (unlikely(avg > 1) && nl_portid_hash_rehash(hash, 1)) |
1da177e4 LT |
1079 | return 1; |
1080 | ||
1081 | if (unlikely(len > avg) && time_after(jiffies, hash->rehash_time)) { | |
15e47304 | 1082 | nl_portid_hash_rehash(hash, 0); |
1da177e4 LT |
1083 | return 1; |
1084 | } | |
1085 | ||
1086 | return 0; | |
1087 | } | |
1088 | ||
90ddc4f0 | 1089 | static const struct proto_ops netlink_ops; |
1da177e4 | 1090 | |
4277a083 PM |
1091 | static void |
1092 | netlink_update_listeners(struct sock *sk) | |
1093 | { | |
1094 | struct netlink_table *tbl = &nl_table[sk->sk_protocol]; | |
4277a083 PM |
1095 | unsigned long mask; |
1096 | unsigned int i; | |
6d772ac5 ED |
1097 | struct listeners *listeners; |
1098 | ||
1099 | listeners = nl_deref_protected(tbl->listeners); | |
1100 | if (!listeners) | |
1101 | return; | |
4277a083 | 1102 | |
b4ff4f04 | 1103 | for (i = 0; i < NLGRPLONGS(tbl->groups); i++) { |
4277a083 | 1104 | mask = 0; |
b67bfe0d | 1105 | sk_for_each_bound(sk, &tbl->mc_list) { |
b4ff4f04 JB |
1106 | if (i < NLGRPLONGS(nlk_sk(sk)->ngroups)) |
1107 | mask |= nlk_sk(sk)->groups[i]; | |
1108 | } | |
6d772ac5 | 1109 | listeners->masks[i] = mask; |
4277a083 PM |
1110 | } |
1111 | /* this function is only called with the netlink table "grabbed", which | |
1112 | * makes sure updates are visible before bind or setsockopt return. */ | |
1113 | } | |
1114 | ||
15e47304 | 1115 | static int netlink_insert(struct sock *sk, struct net *net, u32 portid) |
1da177e4 | 1116 | { |
da12c90e G |
1117 | struct netlink_table *table = &nl_table[sk->sk_protocol]; |
1118 | struct nl_portid_hash *hash = &table->hash; | |
1da177e4 LT |
1119 | struct hlist_head *head; |
1120 | int err = -EADDRINUSE; | |
1121 | struct sock *osk; | |
1da177e4 LT |
1122 | int len; |
1123 | ||
1124 | netlink_table_grab(); | |
15e47304 | 1125 | head = nl_portid_hashfn(hash, portid); |
1da177e4 | 1126 | len = 0; |
b67bfe0d | 1127 | sk_for_each(osk, head) { |
da12c90e G |
1128 | if (table->compare(net, osk) && |
1129 | (nlk_sk(osk)->portid == portid)) | |
1da177e4 LT |
1130 | break; |
1131 | len++; | |
1132 | } | |
b67bfe0d | 1133 | if (osk) |
1da177e4 LT |
1134 | goto err; |
1135 | ||
1136 | err = -EBUSY; | |
15e47304 | 1137 | if (nlk_sk(sk)->portid) |
1da177e4 LT |
1138 | goto err; |
1139 | ||
1140 | err = -ENOMEM; | |
1141 | if (BITS_PER_LONG > 32 && unlikely(hash->entries >= UINT_MAX)) | |
1142 | goto err; | |
1143 | ||
15e47304 EB |
1144 | if (len && nl_portid_hash_dilute(hash, len)) |
1145 | head = nl_portid_hashfn(hash, portid); | |
1da177e4 | 1146 | hash->entries++; |
15e47304 | 1147 | nlk_sk(sk)->portid = portid; |
1da177e4 LT |
1148 | sk_add_node(sk, head); |
1149 | err = 0; | |
1150 | ||
1151 | err: | |
1152 | netlink_table_ungrab(); | |
1153 | return err; | |
1154 | } | |
1155 | ||
1156 | static void netlink_remove(struct sock *sk) | |
1157 | { | |
1158 | netlink_table_grab(); | |
d470e3b4 DM |
1159 | if (sk_del_node_init(sk)) |
1160 | nl_table[sk->sk_protocol].hash.entries--; | |
f7fa9b10 | 1161 | if (nlk_sk(sk)->subscriptions) |
1da177e4 LT |
1162 | __sk_del_bind_node(sk); |
1163 | netlink_table_ungrab(); | |
1164 | } | |
1165 | ||
1166 | static struct proto netlink_proto = { | |
1167 | .name = "NETLINK", | |
1168 | .owner = THIS_MODULE, | |
1169 | .obj_size = sizeof(struct netlink_sock), | |
1170 | }; | |
1171 | ||
1b8d7ae4 EB |
1172 | static int __netlink_create(struct net *net, struct socket *sock, |
1173 | struct mutex *cb_mutex, int protocol) | |
1da177e4 LT |
1174 | { |
1175 | struct sock *sk; | |
1176 | struct netlink_sock *nlk; | |
ab33a171 PM |
1177 | |
1178 | sock->ops = &netlink_ops; | |
1179 | ||
6257ff21 | 1180 | sk = sk_alloc(net, PF_NETLINK, GFP_KERNEL, &netlink_proto); |
ab33a171 PM |
1181 | if (!sk) |
1182 | return -ENOMEM; | |
1183 | ||
1184 | sock_init_data(sock, sk); | |
1185 | ||
1186 | nlk = nlk_sk(sk); | |
658cb354 | 1187 | if (cb_mutex) { |
ffa4d721 | 1188 | nlk->cb_mutex = cb_mutex; |
658cb354 | 1189 | } else { |
ffa4d721 PM |
1190 | nlk->cb_mutex = &nlk->cb_def_mutex; |
1191 | mutex_init(nlk->cb_mutex); | |
1192 | } | |
ab33a171 | 1193 | init_waitqueue_head(&nlk->wait); |
ccdfcc39 PM |
1194 | #ifdef CONFIG_NETLINK_MMAP |
1195 | mutex_init(&nlk->pg_vec_lock); | |
1196 | #endif | |
ab33a171 PM |
1197 | |
1198 | sk->sk_destruct = netlink_sock_destruct; | |
1199 | sk->sk_protocol = protocol; | |
1200 | return 0; | |
1201 | } | |
1202 | ||
3f378b68 EP |
1203 | static int netlink_create(struct net *net, struct socket *sock, int protocol, |
1204 | int kern) | |
ab33a171 PM |
1205 | { |
1206 | struct module *module = NULL; | |
af65bdfc | 1207 | struct mutex *cb_mutex; |
f7fa9b10 | 1208 | struct netlink_sock *nlk; |
4f520900 RGB |
1209 | int (*bind)(int group); |
1210 | void (*unbind)(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; |
4f520900 | 1236 | unbind = nl_table[protocol].unbind; |
77247bbb | 1237 | netlink_unlock_table(); |
4fdb3bb7 | 1238 | |
974c37e9 AD |
1239 | if (err < 0) |
1240 | goto out; | |
1241 | ||
6ac552fd PM |
1242 | err = __netlink_create(net, sock, cb_mutex, protocol); |
1243 | if (err < 0) | |
f7fa9b10 PM |
1244 | goto out_module; |
1245 | ||
6f756a8c | 1246 | local_bh_disable(); |
c1fd3b94 | 1247 | sock_prot_inuse_add(net, &netlink_proto, 1); |
6f756a8c DM |
1248 | local_bh_enable(); |
1249 | ||
f7fa9b10 | 1250 | nlk = nlk_sk(sock->sk); |
f7fa9b10 | 1251 | nlk->module = module; |
03292745 | 1252 | nlk->netlink_bind = bind; |
4f520900 | 1253 | nlk->netlink_unbind = unbind; |
ab33a171 PM |
1254 | out: |
1255 | return err; | |
1da177e4 | 1256 | |
ab33a171 PM |
1257 | out_module: |
1258 | module_put(module); | |
1259 | goto out; | |
1da177e4 LT |
1260 | } |
1261 | ||
1262 | static int netlink_release(struct socket *sock) | |
1263 | { | |
1264 | struct sock *sk = sock->sk; | |
1265 | struct netlink_sock *nlk; | |
1266 | ||
1267 | if (!sk) | |
1268 | return 0; | |
1269 | ||
1270 | netlink_remove(sk); | |
ac57b3a9 | 1271 | sock_orphan(sk); |
1da177e4 LT |
1272 | nlk = nlk_sk(sk); |
1273 | ||
3f660d66 HX |
1274 | /* |
1275 | * OK. Socket is unlinked, any packets that arrive now | |
1276 | * will be purged. | |
1277 | */ | |
1da177e4 | 1278 | |
1da177e4 LT |
1279 | sock->sk = NULL; |
1280 | wake_up_interruptible_all(&nlk->wait); | |
1281 | ||
1282 | skb_queue_purge(&sk->sk_write_queue); | |
1283 | ||
15e47304 | 1284 | if (nlk->portid) { |
1da177e4 | 1285 | struct netlink_notify n = { |
3b1e0a65 | 1286 | .net = sock_net(sk), |
1da177e4 | 1287 | .protocol = sk->sk_protocol, |
15e47304 | 1288 | .portid = nlk->portid, |
1da177e4 | 1289 | }; |
e041c683 AS |
1290 | atomic_notifier_call_chain(&netlink_chain, |
1291 | NETLINK_URELEASE, &n); | |
746fac4d | 1292 | } |
4fdb3bb7 | 1293 | |
5e7c001c | 1294 | module_put(nlk->module); |
4fdb3bb7 | 1295 | |
4277a083 | 1296 | netlink_table_grab(); |
aed81560 | 1297 | if (netlink_is_kernel(sk)) { |
869e58f8 DL |
1298 | BUG_ON(nl_table[sk->sk_protocol].registered == 0); |
1299 | if (--nl_table[sk->sk_protocol].registered == 0) { | |
6d772ac5 ED |
1300 | struct listeners *old; |
1301 | ||
1302 | old = nl_deref_protected(nl_table[sk->sk_protocol].listeners); | |
1303 | RCU_INIT_POINTER(nl_table[sk->sk_protocol].listeners, NULL); | |
1304 | kfree_rcu(old, rcu); | |
869e58f8 | 1305 | nl_table[sk->sk_protocol].module = NULL; |
9785e10a | 1306 | nl_table[sk->sk_protocol].bind = NULL; |
4f520900 | 1307 | nl_table[sk->sk_protocol].unbind = NULL; |
9785e10a | 1308 | nl_table[sk->sk_protocol].flags = 0; |
869e58f8 DL |
1309 | nl_table[sk->sk_protocol].registered = 0; |
1310 | } | |
658cb354 | 1311 | } else if (nlk->subscriptions) { |
4277a083 | 1312 | netlink_update_listeners(sk); |
658cb354 | 1313 | } |
4277a083 | 1314 | netlink_table_ungrab(); |
77247bbb | 1315 | |
f7fa9b10 PM |
1316 | kfree(nlk->groups); |
1317 | nlk->groups = NULL; | |
1318 | ||
3755810c | 1319 | local_bh_disable(); |
c1fd3b94 | 1320 | sock_prot_inuse_add(sock_net(sk), &netlink_proto, -1); |
3755810c | 1321 | local_bh_enable(); |
1da177e4 LT |
1322 | sock_put(sk); |
1323 | return 0; | |
1324 | } | |
1325 | ||
1326 | static int netlink_autobind(struct socket *sock) | |
1327 | { | |
1328 | struct sock *sk = sock->sk; | |
3b1e0a65 | 1329 | struct net *net = sock_net(sk); |
da12c90e G |
1330 | struct netlink_table *table = &nl_table[sk->sk_protocol]; |
1331 | struct nl_portid_hash *hash = &table->hash; | |
1da177e4 LT |
1332 | struct hlist_head *head; |
1333 | struct sock *osk; | |
15e47304 | 1334 | s32 portid = task_tgid_vnr(current); |
1da177e4 LT |
1335 | int err; |
1336 | static s32 rover = -4097; | |
1337 | ||
1338 | retry: | |
1339 | cond_resched(); | |
1340 | netlink_table_grab(); | |
15e47304 | 1341 | head = nl_portid_hashfn(hash, portid); |
b67bfe0d | 1342 | sk_for_each(osk, head) { |
da12c90e | 1343 | if (!table->compare(net, osk)) |
b4b51029 | 1344 | continue; |
15e47304 EB |
1345 | if (nlk_sk(osk)->portid == portid) { |
1346 | /* Bind collision, search negative portid values. */ | |
1347 | portid = rover--; | |
1da177e4 LT |
1348 | if (rover > -4097) |
1349 | rover = -4097; | |
1350 | netlink_table_ungrab(); | |
1351 | goto retry; | |
1352 | } | |
1353 | } | |
1354 | netlink_table_ungrab(); | |
1355 | ||
15e47304 | 1356 | err = netlink_insert(sk, net, portid); |
1da177e4 LT |
1357 | if (err == -EADDRINUSE) |
1358 | goto retry; | |
d470e3b4 DM |
1359 | |
1360 | /* If 2 threads race to autobind, that is fine. */ | |
1361 | if (err == -EBUSY) | |
1362 | err = 0; | |
1363 | ||
1364 | return err; | |
1da177e4 LT |
1365 | } |
1366 | ||
b57ef81f | 1367 | static inline int netlink_capable(const struct socket *sock, unsigned int flag) |
746fac4d | 1368 | { |
9785e10a | 1369 | return (nl_table[sock->sk->sk_protocol].flags & flag) || |
df008c91 | 1370 | ns_capable(sock_net(sock->sk)->user_ns, CAP_NET_ADMIN); |
746fac4d | 1371 | } |
1da177e4 | 1372 | |
f7fa9b10 PM |
1373 | static void |
1374 | netlink_update_subscriptions(struct sock *sk, unsigned int subscriptions) | |
1375 | { | |
1376 | struct netlink_sock *nlk = nlk_sk(sk); | |
1377 | ||
1378 | if (nlk->subscriptions && !subscriptions) | |
1379 | __sk_del_bind_node(sk); | |
1380 | else if (!nlk->subscriptions && subscriptions) | |
1381 | sk_add_bind_node(sk, &nl_table[sk->sk_protocol].mc_list); | |
1382 | nlk->subscriptions = subscriptions; | |
1383 | } | |
1384 | ||
b4ff4f04 | 1385 | static int netlink_realloc_groups(struct sock *sk) |
513c2500 PM |
1386 | { |
1387 | struct netlink_sock *nlk = nlk_sk(sk); | |
1388 | unsigned int groups; | |
b4ff4f04 | 1389 | unsigned long *new_groups; |
513c2500 PM |
1390 | int err = 0; |
1391 | ||
b4ff4f04 JB |
1392 | netlink_table_grab(); |
1393 | ||
513c2500 | 1394 | groups = nl_table[sk->sk_protocol].groups; |
b4ff4f04 | 1395 | if (!nl_table[sk->sk_protocol].registered) { |
513c2500 | 1396 | err = -ENOENT; |
b4ff4f04 JB |
1397 | goto out_unlock; |
1398 | } | |
513c2500 | 1399 | |
b4ff4f04 JB |
1400 | if (nlk->ngroups >= groups) |
1401 | goto out_unlock; | |
513c2500 | 1402 | |
b4ff4f04 JB |
1403 | new_groups = krealloc(nlk->groups, NLGRPSZ(groups), GFP_ATOMIC); |
1404 | if (new_groups == NULL) { | |
1405 | err = -ENOMEM; | |
1406 | goto out_unlock; | |
1407 | } | |
6ac552fd | 1408 | memset((char *)new_groups + NLGRPSZ(nlk->ngroups), 0, |
b4ff4f04 JB |
1409 | NLGRPSZ(groups) - NLGRPSZ(nlk->ngroups)); |
1410 | ||
1411 | nlk->groups = new_groups; | |
513c2500 | 1412 | nlk->ngroups = groups; |
b4ff4f04 JB |
1413 | out_unlock: |
1414 | netlink_table_ungrab(); | |
1415 | return err; | |
513c2500 PM |
1416 | } |
1417 | ||
4f520900 RGB |
1418 | static void netlink_unbind(int group, long unsigned int groups, |
1419 | struct netlink_sock *nlk) | |
1420 | { | |
1421 | int undo; | |
1422 | ||
1423 | if (!nlk->netlink_unbind) | |
1424 | return; | |
1425 | ||
1426 | for (undo = 0; undo < group; undo++) | |
1427 | if (test_bit(group, &groups)) | |
1428 | nlk->netlink_unbind(undo); | |
1429 | } | |
1430 | ||
6ac552fd PM |
1431 | static int netlink_bind(struct socket *sock, struct sockaddr *addr, |
1432 | int addr_len) | |
1da177e4 LT |
1433 | { |
1434 | struct sock *sk = sock->sk; | |
3b1e0a65 | 1435 | struct net *net = sock_net(sk); |
1da177e4 LT |
1436 | struct netlink_sock *nlk = nlk_sk(sk); |
1437 | struct sockaddr_nl *nladdr = (struct sockaddr_nl *)addr; | |
1438 | int err; | |
4f520900 | 1439 | long unsigned int groups = nladdr->nl_groups; |
746fac4d | 1440 | |
4e4b5376 HFS |
1441 | if (addr_len < sizeof(struct sockaddr_nl)) |
1442 | return -EINVAL; | |
1443 | ||
1da177e4 LT |
1444 | if (nladdr->nl_family != AF_NETLINK) |
1445 | return -EINVAL; | |
1446 | ||
1447 | /* Only superuser is allowed to listen multicasts */ | |
4f520900 | 1448 | if (groups) { |
9785e10a | 1449 | if (!netlink_capable(sock, NL_CFG_F_NONROOT_RECV)) |
513c2500 | 1450 | return -EPERM; |
b4ff4f04 JB |
1451 | err = netlink_realloc_groups(sk); |
1452 | if (err) | |
1453 | return err; | |
513c2500 | 1454 | } |
1da177e4 | 1455 | |
4f520900 | 1456 | if (nlk->portid) |
15e47304 | 1457 | if (nladdr->nl_pid != nlk->portid) |
1da177e4 | 1458 | return -EINVAL; |
4f520900 RGB |
1459 | |
1460 | if (nlk->netlink_bind && groups) { | |
1461 | int group; | |
1462 | ||
1463 | for (group = 0; group < nlk->ngroups; group++) { | |
1464 | if (!test_bit(group, &groups)) | |
1465 | continue; | |
1466 | err = nlk->netlink_bind(group); | |
1467 | if (!err) | |
1468 | continue; | |
1469 | netlink_unbind(group, groups, nlk); | |
1470 | return err; | |
1471 | } | |
1472 | } | |
1473 | ||
1474 | if (!nlk->portid) { | |
1da177e4 | 1475 | err = nladdr->nl_pid ? |
b4b51029 | 1476 | netlink_insert(sk, net, nladdr->nl_pid) : |
1da177e4 | 1477 | netlink_autobind(sock); |
4f520900 RGB |
1478 | if (err) { |
1479 | netlink_unbind(nlk->ngroups - 1, groups, nlk); | |
1da177e4 | 1480 | return err; |
4f520900 | 1481 | } |
1da177e4 LT |
1482 | } |
1483 | ||
4f520900 | 1484 | if (!groups && (nlk->groups == NULL || !(u32)nlk->groups[0])) |
1da177e4 LT |
1485 | return 0; |
1486 | ||
1487 | netlink_table_grab(); | |
f7fa9b10 | 1488 | netlink_update_subscriptions(sk, nlk->subscriptions + |
4f520900 | 1489 | hweight32(groups) - |
746fac4d | 1490 | hweight32(nlk->groups[0])); |
4f520900 | 1491 | nlk->groups[0] = (nlk->groups[0] & ~0xffffffffUL) | groups; |
4277a083 | 1492 | netlink_update_listeners(sk); |
1da177e4 LT |
1493 | netlink_table_ungrab(); |
1494 | ||
1495 | return 0; | |
1496 | } | |
1497 | ||
1498 | static int netlink_connect(struct socket *sock, struct sockaddr *addr, | |
1499 | int alen, int flags) | |
1500 | { | |
1501 | int err = 0; | |
1502 | struct sock *sk = sock->sk; | |
1503 | struct netlink_sock *nlk = nlk_sk(sk); | |
6ac552fd | 1504 | struct sockaddr_nl *nladdr = (struct sockaddr_nl *)addr; |
1da177e4 | 1505 | |
6503d961 CG |
1506 | if (alen < sizeof(addr->sa_family)) |
1507 | return -EINVAL; | |
1508 | ||
1da177e4 LT |
1509 | if (addr->sa_family == AF_UNSPEC) { |
1510 | sk->sk_state = NETLINK_UNCONNECTED; | |
15e47304 | 1511 | nlk->dst_portid = 0; |
d629b836 | 1512 | nlk->dst_group = 0; |
1da177e4 LT |
1513 | return 0; |
1514 | } | |
1515 | if (addr->sa_family != AF_NETLINK) | |
1516 | return -EINVAL; | |
1517 | ||
46833a86 MP |
1518 | if ((nladdr->nl_groups || nladdr->nl_pid) && |
1519 | !netlink_capable(sock, NL_CFG_F_NONROOT_SEND)) | |
1da177e4 LT |
1520 | return -EPERM; |
1521 | ||
15e47304 | 1522 | if (!nlk->portid) |
1da177e4 LT |
1523 | err = netlink_autobind(sock); |
1524 | ||
1525 | if (err == 0) { | |
1526 | sk->sk_state = NETLINK_CONNECTED; | |
15e47304 | 1527 | nlk->dst_portid = nladdr->nl_pid; |
d629b836 | 1528 | nlk->dst_group = ffs(nladdr->nl_groups); |
1da177e4 LT |
1529 | } |
1530 | ||
1531 | return err; | |
1532 | } | |
1533 | ||
6ac552fd PM |
1534 | static int netlink_getname(struct socket *sock, struct sockaddr *addr, |
1535 | int *addr_len, int peer) | |
1da177e4 LT |
1536 | { |
1537 | struct sock *sk = sock->sk; | |
1538 | struct netlink_sock *nlk = nlk_sk(sk); | |
13cfa97b | 1539 | DECLARE_SOCKADDR(struct sockaddr_nl *, nladdr, addr); |
746fac4d | 1540 | |
1da177e4 LT |
1541 | nladdr->nl_family = AF_NETLINK; |
1542 | nladdr->nl_pad = 0; | |
1543 | *addr_len = sizeof(*nladdr); | |
1544 | ||
1545 | if (peer) { | |
15e47304 | 1546 | nladdr->nl_pid = nlk->dst_portid; |
d629b836 | 1547 | nladdr->nl_groups = netlink_group_mask(nlk->dst_group); |
1da177e4 | 1548 | } else { |
15e47304 | 1549 | nladdr->nl_pid = nlk->portid; |
513c2500 | 1550 | nladdr->nl_groups = nlk->groups ? nlk->groups[0] : 0; |
1da177e4 LT |
1551 | } |
1552 | return 0; | |
1553 | } | |
1554 | ||
15e47304 | 1555 | static struct sock *netlink_getsockbyportid(struct sock *ssk, u32 portid) |
1da177e4 | 1556 | { |
1da177e4 LT |
1557 | struct sock *sock; |
1558 | struct netlink_sock *nlk; | |
1559 | ||
15e47304 | 1560 | sock = netlink_lookup(sock_net(ssk), ssk->sk_protocol, portid); |
1da177e4 LT |
1561 | if (!sock) |
1562 | return ERR_PTR(-ECONNREFUSED); | |
1563 | ||
1564 | /* Don't bother queuing skb if kernel socket has no input function */ | |
1565 | nlk = nlk_sk(sock); | |
cd40b7d3 | 1566 | if (sock->sk_state == NETLINK_CONNECTED && |
15e47304 | 1567 | nlk->dst_portid != nlk_sk(ssk)->portid) { |
1da177e4 LT |
1568 | sock_put(sock); |
1569 | return ERR_PTR(-ECONNREFUSED); | |
1570 | } | |
1571 | return sock; | |
1572 | } | |
1573 | ||
1574 | struct sock *netlink_getsockbyfilp(struct file *filp) | |
1575 | { | |
496ad9aa | 1576 | struct inode *inode = file_inode(filp); |
1da177e4 LT |
1577 | struct sock *sock; |
1578 | ||
1579 | if (!S_ISSOCK(inode->i_mode)) | |
1580 | return ERR_PTR(-ENOTSOCK); | |
1581 | ||
1582 | sock = SOCKET_I(inode)->sk; | |
1583 | if (sock->sk_family != AF_NETLINK) | |
1584 | return ERR_PTR(-EINVAL); | |
1585 | ||
1586 | sock_hold(sock); | |
1587 | return sock; | |
1588 | } | |
1589 | ||
3a36515f PN |
1590 | static struct sk_buff *netlink_alloc_large_skb(unsigned int size, |
1591 | int broadcast) | |
c05cdb1b PNA |
1592 | { |
1593 | struct sk_buff *skb; | |
1594 | void *data; | |
1595 | ||
3a36515f | 1596 | if (size <= NLMSG_GOODSIZE || broadcast) |
c05cdb1b PNA |
1597 | return alloc_skb(size, GFP_KERNEL); |
1598 | ||
3a36515f PN |
1599 | size = SKB_DATA_ALIGN(size) + |
1600 | SKB_DATA_ALIGN(sizeof(struct skb_shared_info)); | |
c05cdb1b PNA |
1601 | |
1602 | data = vmalloc(size); | |
1603 | if (data == NULL) | |
3a36515f | 1604 | return NULL; |
c05cdb1b | 1605 | |
3a36515f PN |
1606 | skb = build_skb(data, size); |
1607 | if (skb == NULL) | |
1608 | vfree(data); | |
1609 | else { | |
1610 | skb->head_frag = 0; | |
1611 | skb->destructor = netlink_skb_destructor; | |
1612 | } | |
c05cdb1b PNA |
1613 | |
1614 | return skb; | |
c05cdb1b PNA |
1615 | } |
1616 | ||
1da177e4 LT |
1617 | /* |
1618 | * Attach a skb to a netlink socket. | |
1619 | * The caller must hold a reference to the destination socket. On error, the | |
1620 | * reference is dropped. The skb is not send to the destination, just all | |
1621 | * all error checks are performed and memory in the queue is reserved. | |
1622 | * Return values: | |
1623 | * < 0: error. skb freed, reference to sock dropped. | |
1624 | * 0: continue | |
1625 | * 1: repeat lookup - reference dropped while waiting for socket memory. | |
1626 | */ | |
9457afee | 1627 | int netlink_attachskb(struct sock *sk, struct sk_buff *skb, |
c3d8d1e3 | 1628 | long *timeo, struct sock *ssk) |
1da177e4 LT |
1629 | { |
1630 | struct netlink_sock *nlk; | |
1631 | ||
1632 | nlk = nlk_sk(sk); | |
1633 | ||
5fd96123 PM |
1634 | if ((atomic_read(&sk->sk_rmem_alloc) > sk->sk_rcvbuf || |
1635 | test_bit(NETLINK_CONGESTED, &nlk->state)) && | |
1636 | !netlink_skb_is_mmaped(skb)) { | |
1da177e4 | 1637 | DECLARE_WAITQUEUE(wait, current); |
c3d8d1e3 | 1638 | if (!*timeo) { |
aed81560 | 1639 | if (!ssk || netlink_is_kernel(ssk)) |
1da177e4 LT |
1640 | netlink_overrun(sk); |
1641 | sock_put(sk); | |
1642 | kfree_skb(skb); | |
1643 | return -EAGAIN; | |
1644 | } | |
1645 | ||
1646 | __set_current_state(TASK_INTERRUPTIBLE); | |
1647 | add_wait_queue(&nlk->wait, &wait); | |
1648 | ||
1649 | if ((atomic_read(&sk->sk_rmem_alloc) > sk->sk_rcvbuf || | |
cd967e05 | 1650 | test_bit(NETLINK_CONGESTED, &nlk->state)) && |
1da177e4 | 1651 | !sock_flag(sk, SOCK_DEAD)) |
c3d8d1e3 | 1652 | *timeo = schedule_timeout(*timeo); |
1da177e4 LT |
1653 | |
1654 | __set_current_state(TASK_RUNNING); | |
1655 | remove_wait_queue(&nlk->wait, &wait); | |
1656 | sock_put(sk); | |
1657 | ||
1658 | if (signal_pending(current)) { | |
1659 | kfree_skb(skb); | |
c3d8d1e3 | 1660 | return sock_intr_errno(*timeo); |
1da177e4 LT |
1661 | } |
1662 | return 1; | |
1663 | } | |
cf0a018a | 1664 | netlink_skb_set_owner_r(skb, sk); |
1da177e4 LT |
1665 | return 0; |
1666 | } | |
1667 | ||
4a7e7c2a | 1668 | static int __netlink_sendskb(struct sock *sk, struct sk_buff *skb) |
1da177e4 | 1669 | { |
1da177e4 LT |
1670 | int len = skb->len; |
1671 | ||
bcbde0d4 DB |
1672 | netlink_deliver_tap(skb); |
1673 | ||
f9c22888 PM |
1674 | #ifdef CONFIG_NETLINK_MMAP |
1675 | if (netlink_skb_is_mmaped(skb)) | |
1676 | netlink_queue_mmaped_skb(sk, skb); | |
1677 | else if (netlink_rx_is_mmaped(sk)) | |
1678 | netlink_ring_set_copied(sk, skb); | |
1679 | else | |
1680 | #endif /* CONFIG_NETLINK_MMAP */ | |
1681 | skb_queue_tail(&sk->sk_receive_queue, skb); | |
676d2369 | 1682 | sk->sk_data_ready(sk); |
4a7e7c2a ED |
1683 | return len; |
1684 | } | |
1685 | ||
1686 | int netlink_sendskb(struct sock *sk, struct sk_buff *skb) | |
1687 | { | |
1688 | int len = __netlink_sendskb(sk, skb); | |
1689 | ||
1da177e4 LT |
1690 | sock_put(sk); |
1691 | return len; | |
1692 | } | |
1693 | ||
1694 | void netlink_detachskb(struct sock *sk, struct sk_buff *skb) | |
1695 | { | |
1696 | kfree_skb(skb); | |
1697 | sock_put(sk); | |
1698 | } | |
1699 | ||
b57ef81f | 1700 | static struct sk_buff *netlink_trim(struct sk_buff *skb, gfp_t allocation) |
1da177e4 LT |
1701 | { |
1702 | int delta; | |
1703 | ||
1298ca46 | 1704 | WARN_ON(skb->sk != NULL); |
5fd96123 PM |
1705 | if (netlink_skb_is_mmaped(skb)) |
1706 | return skb; | |
1da177e4 | 1707 | |
4305b541 | 1708 | delta = skb->end - skb->tail; |
c05cdb1b | 1709 | if (is_vmalloc_addr(skb->head) || delta * 2 < skb->truesize) |
1da177e4 LT |
1710 | return skb; |
1711 | ||
1712 | if (skb_shared(skb)) { | |
1713 | struct sk_buff *nskb = skb_clone(skb, allocation); | |
1714 | if (!nskb) | |
1715 | return skb; | |
8460c00f | 1716 | consume_skb(skb); |
1da177e4 LT |
1717 | skb = nskb; |
1718 | } | |
1719 | ||
1720 | if (!pskb_expand_head(skb, 0, -delta, allocation)) | |
1721 | skb->truesize -= delta; | |
1722 | ||
1723 | return skb; | |
1724 | } | |
1725 | ||
3fbc2905 EB |
1726 | static int netlink_unicast_kernel(struct sock *sk, struct sk_buff *skb, |
1727 | struct sock *ssk) | |
cd40b7d3 DL |
1728 | { |
1729 | int ret; | |
1730 | struct netlink_sock *nlk = nlk_sk(sk); | |
1731 | ||
1732 | ret = -ECONNREFUSED; | |
1733 | if (nlk->netlink_rcv != NULL) { | |
1734 | ret = skb->len; | |
cf0a018a | 1735 | netlink_skb_set_owner_r(skb, sk); |
e32123e5 | 1736 | NETLINK_CB(skb).sk = ssk; |
73bfd370 | 1737 | netlink_deliver_tap_kernel(sk, ssk, skb); |
cd40b7d3 | 1738 | nlk->netlink_rcv(skb); |
bfb253c9 ED |
1739 | consume_skb(skb); |
1740 | } else { | |
1741 | kfree_skb(skb); | |
cd40b7d3 | 1742 | } |
cd40b7d3 DL |
1743 | sock_put(sk); |
1744 | return ret; | |
1745 | } | |
1746 | ||
1747 | int netlink_unicast(struct sock *ssk, struct sk_buff *skb, | |
15e47304 | 1748 | u32 portid, int nonblock) |
1da177e4 LT |
1749 | { |
1750 | struct sock *sk; | |
1751 | int err; | |
1752 | long timeo; | |
1753 | ||
1754 | skb = netlink_trim(skb, gfp_any()); | |
1755 | ||
1756 | timeo = sock_sndtimeo(ssk, nonblock); | |
1757 | retry: | |
15e47304 | 1758 | sk = netlink_getsockbyportid(ssk, portid); |
1da177e4 LT |
1759 | if (IS_ERR(sk)) { |
1760 | kfree_skb(skb); | |
1761 | return PTR_ERR(sk); | |
1762 | } | |
cd40b7d3 | 1763 | if (netlink_is_kernel(sk)) |
3fbc2905 | 1764 | return netlink_unicast_kernel(sk, skb, ssk); |
cd40b7d3 | 1765 | |
b1153f29 | 1766 | if (sk_filter(sk, skb)) { |
84874607 | 1767 | err = skb->len; |
b1153f29 SH |
1768 | kfree_skb(skb); |
1769 | sock_put(sk); | |
1770 | return err; | |
1771 | } | |
1772 | ||
9457afee | 1773 | err = netlink_attachskb(sk, skb, &timeo, ssk); |
1da177e4 LT |
1774 | if (err == 1) |
1775 | goto retry; | |
1776 | if (err) | |
1777 | return err; | |
1778 | ||
7ee015e0 | 1779 | return netlink_sendskb(sk, skb); |
1da177e4 | 1780 | } |
6ac552fd | 1781 | EXPORT_SYMBOL(netlink_unicast); |
1da177e4 | 1782 | |
f9c22888 PM |
1783 | struct sk_buff *netlink_alloc_skb(struct sock *ssk, unsigned int size, |
1784 | u32 dst_portid, gfp_t gfp_mask) | |
1785 | { | |
1786 | #ifdef CONFIG_NETLINK_MMAP | |
1787 | struct sock *sk = NULL; | |
1788 | struct sk_buff *skb; | |
1789 | struct netlink_ring *ring; | |
1790 | struct nl_mmap_hdr *hdr; | |
1791 | unsigned int maxlen; | |
1792 | ||
1793 | sk = netlink_getsockbyportid(ssk, dst_portid); | |
1794 | if (IS_ERR(sk)) | |
1795 | goto out; | |
1796 | ||
1797 | ring = &nlk_sk(sk)->rx_ring; | |
1798 | /* fast-path without atomic ops for common case: non-mmaped receiver */ | |
1799 | if (ring->pg_vec == NULL) | |
1800 | goto out_put; | |
1801 | ||
aae9f0e2 TG |
1802 | if (ring->frame_size - NL_MMAP_HDRLEN < size) |
1803 | goto out_put; | |
1804 | ||
f9c22888 PM |
1805 | skb = alloc_skb_head(gfp_mask); |
1806 | if (skb == NULL) | |
1807 | goto err1; | |
1808 | ||
1809 | spin_lock_bh(&sk->sk_receive_queue.lock); | |
1810 | /* check again under lock */ | |
1811 | if (ring->pg_vec == NULL) | |
1812 | goto out_free; | |
1813 | ||
aae9f0e2 | 1814 | /* check again under lock */ |
f9c22888 PM |
1815 | maxlen = ring->frame_size - NL_MMAP_HDRLEN; |
1816 | if (maxlen < size) | |
1817 | goto out_free; | |
1818 | ||
1819 | netlink_forward_ring(ring); | |
1820 | hdr = netlink_current_frame(ring, NL_MMAP_STATUS_UNUSED); | |
1821 | if (hdr == NULL) | |
1822 | goto err2; | |
1823 | netlink_ring_setup_skb(skb, sk, ring, hdr); | |
1824 | netlink_set_status(hdr, NL_MMAP_STATUS_RESERVED); | |
1825 | atomic_inc(&ring->pending); | |
1826 | netlink_increment_head(ring); | |
1827 | ||
1828 | spin_unlock_bh(&sk->sk_receive_queue.lock); | |
1829 | return skb; | |
1830 | ||
1831 | err2: | |
1832 | kfree_skb(skb); | |
1833 | spin_unlock_bh(&sk->sk_receive_queue.lock); | |
cd1df525 | 1834 | netlink_overrun(sk); |
f9c22888 PM |
1835 | err1: |
1836 | sock_put(sk); | |
1837 | return NULL; | |
1838 | ||
1839 | out_free: | |
1840 | kfree_skb(skb); | |
1841 | spin_unlock_bh(&sk->sk_receive_queue.lock); | |
1842 | out_put: | |
1843 | sock_put(sk); | |
1844 | out: | |
1845 | #endif | |
1846 | return alloc_skb(size, gfp_mask); | |
1847 | } | |
1848 | EXPORT_SYMBOL_GPL(netlink_alloc_skb); | |
1849 | ||
4277a083 PM |
1850 | int netlink_has_listeners(struct sock *sk, unsigned int group) |
1851 | { | |
1852 | int res = 0; | |
5c398dc8 | 1853 | struct listeners *listeners; |
4277a083 | 1854 | |
aed81560 | 1855 | BUG_ON(!netlink_is_kernel(sk)); |
b4ff4f04 JB |
1856 | |
1857 | rcu_read_lock(); | |
1858 | listeners = rcu_dereference(nl_table[sk->sk_protocol].listeners); | |
1859 | ||
6d772ac5 | 1860 | if (listeners && group - 1 < nl_table[sk->sk_protocol].groups) |
5c398dc8 | 1861 | res = test_bit(group - 1, listeners->masks); |
b4ff4f04 JB |
1862 | |
1863 | rcu_read_unlock(); | |
1864 | ||
4277a083 PM |
1865 | return res; |
1866 | } | |
1867 | EXPORT_SYMBOL_GPL(netlink_has_listeners); | |
1868 | ||
b57ef81f | 1869 | static int netlink_broadcast_deliver(struct sock *sk, struct sk_buff *skb) |
1da177e4 LT |
1870 | { |
1871 | struct netlink_sock *nlk = nlk_sk(sk); | |
1872 | ||
1873 | if (atomic_read(&sk->sk_rmem_alloc) <= sk->sk_rcvbuf && | |
cd967e05 | 1874 | !test_bit(NETLINK_CONGESTED, &nlk->state)) { |
cf0a018a | 1875 | netlink_skb_set_owner_r(skb, sk); |
4a7e7c2a | 1876 | __netlink_sendskb(sk, skb); |
2c645800 | 1877 | return atomic_read(&sk->sk_rmem_alloc) > (sk->sk_rcvbuf >> 1); |
1da177e4 LT |
1878 | } |
1879 | return -1; | |
1880 | } | |
1881 | ||
1882 | struct netlink_broadcast_data { | |
1883 | struct sock *exclude_sk; | |
b4b51029 | 1884 | struct net *net; |
15e47304 | 1885 | u32 portid; |
1da177e4 LT |
1886 | u32 group; |
1887 | int failure; | |
ff491a73 | 1888 | int delivery_failure; |
1da177e4 LT |
1889 | int congested; |
1890 | int delivered; | |
7d877f3b | 1891 | gfp_t allocation; |
1da177e4 | 1892 | struct sk_buff *skb, *skb2; |
910a7e90 EB |
1893 | int (*tx_filter)(struct sock *dsk, struct sk_buff *skb, void *data); |
1894 | void *tx_data; | |
1da177e4 LT |
1895 | }; |
1896 | ||
b57ef81f | 1897 | static int do_one_broadcast(struct sock *sk, |
1da177e4 LT |
1898 | struct netlink_broadcast_data *p) |
1899 | { | |
1900 | struct netlink_sock *nlk = nlk_sk(sk); | |
1901 | int val; | |
1902 | ||
1903 | if (p->exclude_sk == sk) | |
1904 | goto out; | |
1905 | ||
15e47304 | 1906 | if (nlk->portid == p->portid || p->group - 1 >= nlk->ngroups || |
f7fa9b10 | 1907 | !test_bit(p->group - 1, nlk->groups)) |
1da177e4 LT |
1908 | goto out; |
1909 | ||
878628fb | 1910 | if (!net_eq(sock_net(sk), p->net)) |
b4b51029 EB |
1911 | goto out; |
1912 | ||
1da177e4 LT |
1913 | if (p->failure) { |
1914 | netlink_overrun(sk); | |
1915 | goto out; | |
1916 | } | |
1917 | ||
1918 | sock_hold(sk); | |
1919 | if (p->skb2 == NULL) { | |
68acc024 | 1920 | if (skb_shared(p->skb)) { |
1da177e4 LT |
1921 | p->skb2 = skb_clone(p->skb, p->allocation); |
1922 | } else { | |
68acc024 TC |
1923 | p->skb2 = skb_get(p->skb); |
1924 | /* | |
1925 | * skb ownership may have been set when | |
1926 | * delivered to a previous socket. | |
1927 | */ | |
1928 | skb_orphan(p->skb2); | |
1da177e4 LT |
1929 | } |
1930 | } | |
1931 | if (p->skb2 == NULL) { | |
1932 | netlink_overrun(sk); | |
1933 | /* Clone failed. Notify ALL listeners. */ | |
1934 | p->failure = 1; | |
be0c22a4 PNA |
1935 | if (nlk->flags & NETLINK_BROADCAST_SEND_ERROR) |
1936 | p->delivery_failure = 1; | |
910a7e90 EB |
1937 | } else if (p->tx_filter && p->tx_filter(sk, p->skb2, p->tx_data)) { |
1938 | kfree_skb(p->skb2); | |
1939 | p->skb2 = NULL; | |
b1153f29 SH |
1940 | } else if (sk_filter(sk, p->skb2)) { |
1941 | kfree_skb(p->skb2); | |
1942 | p->skb2 = NULL; | |
1da177e4 LT |
1943 | } else if ((val = netlink_broadcast_deliver(sk, p->skb2)) < 0) { |
1944 | netlink_overrun(sk); | |
be0c22a4 PNA |
1945 | if (nlk->flags & NETLINK_BROADCAST_SEND_ERROR) |
1946 | p->delivery_failure = 1; | |
1da177e4 LT |
1947 | } else { |
1948 | p->congested |= val; | |
1949 | p->delivered = 1; | |
1950 | p->skb2 = NULL; | |
1951 | } | |
1952 | sock_put(sk); | |
1953 | ||
1954 | out: | |
1955 | return 0; | |
1956 | } | |
1957 | ||
15e47304 | 1958 | int netlink_broadcast_filtered(struct sock *ssk, struct sk_buff *skb, u32 portid, |
910a7e90 EB |
1959 | u32 group, gfp_t allocation, |
1960 | int (*filter)(struct sock *dsk, struct sk_buff *skb, void *data), | |
1961 | void *filter_data) | |
1da177e4 | 1962 | { |
3b1e0a65 | 1963 | struct net *net = sock_net(ssk); |
1da177e4 | 1964 | struct netlink_broadcast_data info; |
1da177e4 LT |
1965 | struct sock *sk; |
1966 | ||
1967 | skb = netlink_trim(skb, allocation); | |
1968 | ||
1969 | info.exclude_sk = ssk; | |
b4b51029 | 1970 | info.net = net; |
15e47304 | 1971 | info.portid = portid; |
1da177e4 LT |
1972 | info.group = group; |
1973 | info.failure = 0; | |
ff491a73 | 1974 | info.delivery_failure = 0; |
1da177e4 LT |
1975 | info.congested = 0; |
1976 | info.delivered = 0; | |
1977 | info.allocation = allocation; | |
1978 | info.skb = skb; | |
1979 | info.skb2 = NULL; | |
910a7e90 EB |
1980 | info.tx_filter = filter; |
1981 | info.tx_data = filter_data; | |
1da177e4 LT |
1982 | |
1983 | /* While we sleep in clone, do not allow to change socket list */ | |
1984 | ||
1985 | netlink_lock_table(); | |
1986 | ||
b67bfe0d | 1987 | sk_for_each_bound(sk, &nl_table[ssk->sk_protocol].mc_list) |
1da177e4 LT |
1988 | do_one_broadcast(sk, &info); |
1989 | ||
70d4bf6d | 1990 | consume_skb(skb); |
aa1c6a6f | 1991 | |
1da177e4 LT |
1992 | netlink_unlock_table(); |
1993 | ||
70d4bf6d NH |
1994 | if (info.delivery_failure) { |
1995 | kfree_skb(info.skb2); | |
ff491a73 | 1996 | return -ENOBUFS; |
658cb354 ED |
1997 | } |
1998 | consume_skb(info.skb2); | |
ff491a73 | 1999 | |
1da177e4 LT |
2000 | if (info.delivered) { |
2001 | if (info.congested && (allocation & __GFP_WAIT)) | |
2002 | yield(); | |
2003 | return 0; | |
2004 | } | |
1da177e4 LT |
2005 | return -ESRCH; |
2006 | } | |
910a7e90 EB |
2007 | EXPORT_SYMBOL(netlink_broadcast_filtered); |
2008 | ||
15e47304 | 2009 | int netlink_broadcast(struct sock *ssk, struct sk_buff *skb, u32 portid, |
910a7e90 EB |
2010 | u32 group, gfp_t allocation) |
2011 | { | |
15e47304 | 2012 | return netlink_broadcast_filtered(ssk, skb, portid, group, allocation, |
910a7e90 EB |
2013 | NULL, NULL); |
2014 | } | |
6ac552fd | 2015 | EXPORT_SYMBOL(netlink_broadcast); |
1da177e4 LT |
2016 | |
2017 | struct netlink_set_err_data { | |
2018 | struct sock *exclude_sk; | |
15e47304 | 2019 | u32 portid; |
1da177e4 LT |
2020 | u32 group; |
2021 | int code; | |
2022 | }; | |
2023 | ||
b57ef81f | 2024 | static int do_one_set_err(struct sock *sk, struct netlink_set_err_data *p) |
1da177e4 LT |
2025 | { |
2026 | struct netlink_sock *nlk = nlk_sk(sk); | |
1a50307b | 2027 | int ret = 0; |
1da177e4 LT |
2028 | |
2029 | if (sk == p->exclude_sk) | |
2030 | goto out; | |
2031 | ||
09ad9bc7 | 2032 | if (!net_eq(sock_net(sk), sock_net(p->exclude_sk))) |
b4b51029 EB |
2033 | goto out; |
2034 | ||
15e47304 | 2035 | if (nlk->portid == p->portid || p->group - 1 >= nlk->ngroups || |
f7fa9b10 | 2036 | !test_bit(p->group - 1, nlk->groups)) |
1da177e4 LT |
2037 | goto out; |
2038 | ||
1a50307b PNA |
2039 | if (p->code == ENOBUFS && nlk->flags & NETLINK_RECV_NO_ENOBUFS) { |
2040 | ret = 1; | |
2041 | goto out; | |
2042 | } | |
2043 | ||
1da177e4 LT |
2044 | sk->sk_err = p->code; |
2045 | sk->sk_error_report(sk); | |
2046 | out: | |
1a50307b | 2047 | return ret; |
1da177e4 LT |
2048 | } |
2049 | ||
4843b93c PNA |
2050 | /** |
2051 | * netlink_set_err - report error to broadcast listeners | |
2052 | * @ssk: the kernel netlink socket, as returned by netlink_kernel_create() | |
15e47304 | 2053 | * @portid: the PORTID of a process that we want to skip (if any) |
840e93f2 | 2054 | * @group: the broadcast group that will notice the error |
4843b93c | 2055 | * @code: error code, must be negative (as usual in kernelspace) |
1a50307b PNA |
2056 | * |
2057 | * This function returns the number of broadcast listeners that have set the | |
2058 | * NETLINK_RECV_NO_ENOBUFS socket option. | |
4843b93c | 2059 | */ |
15e47304 | 2060 | int netlink_set_err(struct sock *ssk, u32 portid, u32 group, int code) |
1da177e4 LT |
2061 | { |
2062 | struct netlink_set_err_data info; | |
1da177e4 | 2063 | struct sock *sk; |
1a50307b | 2064 | int ret = 0; |
1da177e4 LT |
2065 | |
2066 | info.exclude_sk = ssk; | |
15e47304 | 2067 | info.portid = portid; |
1da177e4 | 2068 | info.group = group; |
4843b93c PNA |
2069 | /* sk->sk_err wants a positive error value */ |
2070 | info.code = -code; | |
1da177e4 LT |
2071 | |
2072 | read_lock(&nl_table_lock); | |
2073 | ||
b67bfe0d | 2074 | sk_for_each_bound(sk, &nl_table[ssk->sk_protocol].mc_list) |
1a50307b | 2075 | ret += do_one_set_err(sk, &info); |
1da177e4 LT |
2076 | |
2077 | read_unlock(&nl_table_lock); | |
1a50307b | 2078 | return ret; |
1da177e4 | 2079 | } |
dd5b6ce6 | 2080 | EXPORT_SYMBOL(netlink_set_err); |
1da177e4 | 2081 | |
84659eb5 JB |
2082 | /* must be called with netlink table grabbed */ |
2083 | static void netlink_update_socket_mc(struct netlink_sock *nlk, | |
2084 | unsigned int group, | |
2085 | int is_new) | |
2086 | { | |
2087 | int old, new = !!is_new, subscriptions; | |
2088 | ||
2089 | old = test_bit(group - 1, nlk->groups); | |
2090 | subscriptions = nlk->subscriptions - old + new; | |
2091 | if (new) | |
2092 | __set_bit(group - 1, nlk->groups); | |
2093 | else | |
2094 | __clear_bit(group - 1, nlk->groups); | |
2095 | netlink_update_subscriptions(&nlk->sk, subscriptions); | |
2096 | netlink_update_listeners(&nlk->sk); | |
2097 | } | |
2098 | ||
9a4595bc | 2099 | static int netlink_setsockopt(struct socket *sock, int level, int optname, |
b7058842 | 2100 | char __user *optval, unsigned int optlen) |
9a4595bc PM |
2101 | { |
2102 | struct sock *sk = sock->sk; | |
2103 | struct netlink_sock *nlk = nlk_sk(sk); | |
eb496534 JB |
2104 | unsigned int val = 0; |
2105 | int err; | |
9a4595bc PM |
2106 | |
2107 | if (level != SOL_NETLINK) | |
2108 | return -ENOPROTOOPT; | |
2109 | ||
ccdfcc39 PM |
2110 | if (optname != NETLINK_RX_RING && optname != NETLINK_TX_RING && |
2111 | optlen >= sizeof(int) && | |
eb496534 | 2112 | get_user(val, (unsigned int __user *)optval)) |
9a4595bc PM |
2113 | return -EFAULT; |
2114 | ||
2115 | switch (optname) { | |
2116 | case NETLINK_PKTINFO: | |
2117 | if (val) | |
2118 | nlk->flags |= NETLINK_RECV_PKTINFO; | |
2119 | else | |
2120 | nlk->flags &= ~NETLINK_RECV_PKTINFO; | |
2121 | err = 0; | |
2122 | break; | |
2123 | case NETLINK_ADD_MEMBERSHIP: | |
2124 | case NETLINK_DROP_MEMBERSHIP: { | |
9785e10a | 2125 | if (!netlink_capable(sock, NL_CFG_F_NONROOT_RECV)) |
9a4595bc | 2126 | return -EPERM; |
b4ff4f04 JB |
2127 | err = netlink_realloc_groups(sk); |
2128 | if (err) | |
2129 | return err; | |
9a4595bc PM |
2130 | if (!val || val - 1 >= nlk->ngroups) |
2131 | return -EINVAL; | |
4f520900 RGB |
2132 | if (nlk->netlink_bind) { |
2133 | err = nlk->netlink_bind(val); | |
2134 | if (err) | |
2135 | return err; | |
2136 | } | |
9a4595bc | 2137 | netlink_table_grab(); |
84659eb5 JB |
2138 | netlink_update_socket_mc(nlk, val, |
2139 | optname == NETLINK_ADD_MEMBERSHIP); | |
9a4595bc | 2140 | netlink_table_ungrab(); |
03292745 | 2141 | |
9a4595bc PM |
2142 | err = 0; |
2143 | break; | |
2144 | } | |
be0c22a4 PNA |
2145 | case NETLINK_BROADCAST_ERROR: |
2146 | if (val) | |
2147 | nlk->flags |= NETLINK_BROADCAST_SEND_ERROR; | |
2148 | else | |
2149 | nlk->flags &= ~NETLINK_BROADCAST_SEND_ERROR; | |
2150 | err = 0; | |
2151 | break; | |
38938bfe PNA |
2152 | case NETLINK_NO_ENOBUFS: |
2153 | if (val) { | |
2154 | nlk->flags |= NETLINK_RECV_NO_ENOBUFS; | |
cd967e05 | 2155 | clear_bit(NETLINK_CONGESTED, &nlk->state); |
38938bfe | 2156 | wake_up_interruptible(&nlk->wait); |
658cb354 | 2157 | } else { |
38938bfe | 2158 | nlk->flags &= ~NETLINK_RECV_NO_ENOBUFS; |
658cb354 | 2159 | } |
38938bfe PNA |
2160 | err = 0; |
2161 | break; | |
ccdfcc39 PM |
2162 | #ifdef CONFIG_NETLINK_MMAP |
2163 | case NETLINK_RX_RING: | |
2164 | case NETLINK_TX_RING: { | |
2165 | struct nl_mmap_req req; | |
2166 | ||
2167 | /* Rings might consume more memory than queue limits, require | |
2168 | * CAP_NET_ADMIN. | |
2169 | */ | |
2170 | if (!capable(CAP_NET_ADMIN)) | |
2171 | return -EPERM; | |
2172 | if (optlen < sizeof(req)) | |
2173 | return -EINVAL; | |
2174 | if (copy_from_user(&req, optval, sizeof(req))) | |
2175 | return -EFAULT; | |
2176 | err = netlink_set_ring(sk, &req, false, | |
2177 | optname == NETLINK_TX_RING); | |
2178 | break; | |
2179 | } | |
2180 | #endif /* CONFIG_NETLINK_MMAP */ | |
9a4595bc PM |
2181 | default: |
2182 | err = -ENOPROTOOPT; | |
2183 | } | |
2184 | return err; | |
2185 | } | |
2186 | ||
2187 | static int netlink_getsockopt(struct socket *sock, int level, int optname, | |
746fac4d | 2188 | char __user *optval, int __user *optlen) |
9a4595bc PM |
2189 | { |
2190 | struct sock *sk = sock->sk; | |
2191 | struct netlink_sock *nlk = nlk_sk(sk); | |
2192 | int len, val, err; | |
2193 | ||
2194 | if (level != SOL_NETLINK) | |
2195 | return -ENOPROTOOPT; | |
2196 | ||
2197 | if (get_user(len, optlen)) | |
2198 | return -EFAULT; | |
2199 | if (len < 0) | |
2200 | return -EINVAL; | |
2201 | ||
2202 | switch (optname) { | |
2203 | case NETLINK_PKTINFO: | |
2204 | if (len < sizeof(int)) | |
2205 | return -EINVAL; | |
2206 | len = sizeof(int); | |
2207 | val = nlk->flags & NETLINK_RECV_PKTINFO ? 1 : 0; | |
a27b58fe HC |
2208 | if (put_user(len, optlen) || |
2209 | put_user(val, optval)) | |
2210 | return -EFAULT; | |
9a4595bc PM |
2211 | err = 0; |
2212 | break; | |
be0c22a4 PNA |
2213 | case NETLINK_BROADCAST_ERROR: |
2214 | if (len < sizeof(int)) | |
2215 | return -EINVAL; | |
2216 | len = sizeof(int); | |
2217 | val = nlk->flags & NETLINK_BROADCAST_SEND_ERROR ? 1 : 0; | |
2218 | if (put_user(len, optlen) || | |
2219 | put_user(val, optval)) | |
2220 | return -EFAULT; | |
2221 | err = 0; | |
2222 | break; | |
38938bfe PNA |
2223 | case NETLINK_NO_ENOBUFS: |
2224 | if (len < sizeof(int)) | |
2225 | return -EINVAL; | |
2226 | len = sizeof(int); | |
2227 | val = nlk->flags & NETLINK_RECV_NO_ENOBUFS ? 1 : 0; | |
2228 | if (put_user(len, optlen) || | |
2229 | put_user(val, optval)) | |
2230 | return -EFAULT; | |
2231 | err = 0; | |
2232 | break; | |
9a4595bc PM |
2233 | default: |
2234 | err = -ENOPROTOOPT; | |
2235 | } | |
2236 | return err; | |
2237 | } | |
2238 | ||
2239 | static void netlink_cmsg_recv_pktinfo(struct msghdr *msg, struct sk_buff *skb) | |
2240 | { | |
2241 | struct nl_pktinfo info; | |
2242 | ||
2243 | info.group = NETLINK_CB(skb).dst_group; | |
2244 | put_cmsg(msg, SOL_NETLINK, NETLINK_PKTINFO, sizeof(info), &info); | |
2245 | } | |
2246 | ||
1da177e4 LT |
2247 | static int netlink_sendmsg(struct kiocb *kiocb, struct socket *sock, |
2248 | struct msghdr *msg, size_t len) | |
2249 | { | |
2250 | struct sock_iocb *siocb = kiocb_to_siocb(kiocb); | |
2251 | struct sock *sk = sock->sk; | |
2252 | struct netlink_sock *nlk = nlk_sk(sk); | |
342dfc30 | 2253 | DECLARE_SOCKADDR(struct sockaddr_nl *, addr, msg->msg_name); |
15e47304 | 2254 | u32 dst_portid; |
d629b836 | 2255 | u32 dst_group; |
1da177e4 LT |
2256 | struct sk_buff *skb; |
2257 | int err; | |
2258 | struct scm_cookie scm; | |
2259 | ||
2260 | if (msg->msg_flags&MSG_OOB) | |
2261 | return -EOPNOTSUPP; | |
2262 | ||
16e57262 | 2263 | if (NULL == siocb->scm) |
1da177e4 | 2264 | siocb->scm = &scm; |
16e57262 | 2265 | |
e0e3cea4 | 2266 | err = scm_send(sock, msg, siocb->scm, true); |
1da177e4 LT |
2267 | if (err < 0) |
2268 | return err; | |
2269 | ||
2270 | if (msg->msg_namelen) { | |
b47030c7 | 2271 | err = -EINVAL; |
1da177e4 | 2272 | if (addr->nl_family != AF_NETLINK) |
b47030c7 | 2273 | goto out; |
15e47304 | 2274 | dst_portid = addr->nl_pid; |
d629b836 | 2275 | dst_group = ffs(addr->nl_groups); |
b47030c7 | 2276 | err = -EPERM; |
15e47304 | 2277 | if ((dst_group || dst_portid) && |
9785e10a | 2278 | !netlink_capable(sock, NL_CFG_F_NONROOT_SEND)) |
b47030c7 | 2279 | goto out; |
1da177e4 | 2280 | } else { |
15e47304 | 2281 | dst_portid = nlk->dst_portid; |
d629b836 | 2282 | dst_group = nlk->dst_group; |
1da177e4 LT |
2283 | } |
2284 | ||
15e47304 | 2285 | if (!nlk->portid) { |
1da177e4 LT |
2286 | err = netlink_autobind(sock); |
2287 | if (err) | |
2288 | goto out; | |
2289 | } | |
2290 | ||
5fd96123 PM |
2291 | if (netlink_tx_is_mmaped(sk) && |
2292 | msg->msg_iov->iov_base == NULL) { | |
2293 | err = netlink_mmap_sendmsg(sk, msg, dst_portid, dst_group, | |
2294 | siocb); | |
2295 | goto out; | |
2296 | } | |
2297 | ||
1da177e4 LT |
2298 | err = -EMSGSIZE; |
2299 | if (len > sk->sk_sndbuf - 32) | |
2300 | goto out; | |
2301 | err = -ENOBUFS; | |
3a36515f | 2302 | skb = netlink_alloc_large_skb(len, dst_group); |
6ac552fd | 2303 | if (skb == NULL) |
1da177e4 LT |
2304 | goto out; |
2305 | ||
15e47304 | 2306 | NETLINK_CB(skb).portid = nlk->portid; |
d629b836 | 2307 | NETLINK_CB(skb).dst_group = dst_group; |
dbe9a417 | 2308 | NETLINK_CB(skb).creds = siocb->scm->creds; |
1da177e4 | 2309 | |
1da177e4 | 2310 | err = -EFAULT; |
6ac552fd | 2311 | if (memcpy_fromiovec(skb_put(skb, len), msg->msg_iov, len)) { |
1da177e4 LT |
2312 | kfree_skb(skb); |
2313 | goto out; | |
2314 | } | |
2315 | ||
2316 | err = security_netlink_send(sk, skb); | |
2317 | if (err) { | |
2318 | kfree_skb(skb); | |
2319 | goto out; | |
2320 | } | |
2321 | ||
d629b836 | 2322 | if (dst_group) { |
1da177e4 | 2323 | atomic_inc(&skb->users); |
15e47304 | 2324 | netlink_broadcast(sk, skb, dst_portid, dst_group, GFP_KERNEL); |
1da177e4 | 2325 | } |
15e47304 | 2326 | err = netlink_unicast(sk, skb, dst_portid, msg->msg_flags&MSG_DONTWAIT); |
1da177e4 LT |
2327 | |
2328 | out: | |
b47030c7 | 2329 | scm_destroy(siocb->scm); |
1da177e4 LT |
2330 | return err; |
2331 | } | |
2332 | ||
2333 | static int netlink_recvmsg(struct kiocb *kiocb, struct socket *sock, | |
2334 | struct msghdr *msg, size_t len, | |
2335 | int flags) | |
2336 | { | |
2337 | struct sock_iocb *siocb = kiocb_to_siocb(kiocb); | |
2338 | struct scm_cookie scm; | |
2339 | struct sock *sk = sock->sk; | |
2340 | struct netlink_sock *nlk = nlk_sk(sk); | |
2341 | int noblock = flags&MSG_DONTWAIT; | |
2342 | size_t copied; | |
68d6ac6d | 2343 | struct sk_buff *skb, *data_skb; |
b44d211e | 2344 | int err, ret; |
1da177e4 LT |
2345 | |
2346 | if (flags&MSG_OOB) | |
2347 | return -EOPNOTSUPP; | |
2348 | ||
2349 | copied = 0; | |
2350 | ||
6ac552fd PM |
2351 | skb = skb_recv_datagram(sk, flags, noblock, &err); |
2352 | if (skb == NULL) | |
1da177e4 LT |
2353 | goto out; |
2354 | ||
68d6ac6d JB |
2355 | data_skb = skb; |
2356 | ||
1dacc76d JB |
2357 | #ifdef CONFIG_COMPAT_NETLINK_MESSAGES |
2358 | if (unlikely(skb_shinfo(skb)->frag_list)) { | |
1dacc76d | 2359 | /* |
68d6ac6d JB |
2360 | * If this skb has a frag_list, then here that means that we |
2361 | * will have to use the frag_list skb's data for compat tasks | |
2362 | * and the regular skb's data for normal (non-compat) tasks. | |
1dacc76d | 2363 | * |
68d6ac6d JB |
2364 | * If we need to send the compat skb, assign it to the |
2365 | * 'data_skb' variable so that it will be used below for data | |
2366 | * copying. We keep 'skb' for everything else, including | |
2367 | * freeing both later. | |
1dacc76d | 2368 | */ |
68d6ac6d JB |
2369 | if (flags & MSG_CMSG_COMPAT) |
2370 | data_skb = skb_shinfo(skb)->frag_list; | |
1dacc76d JB |
2371 | } |
2372 | #endif | |
2373 | ||
9063e21f ED |
2374 | /* Record the max length of recvmsg() calls for future allocations */ |
2375 | nlk->max_recvmsg_len = max(nlk->max_recvmsg_len, len); | |
2376 | nlk->max_recvmsg_len = min_t(size_t, nlk->max_recvmsg_len, | |
2377 | 16384); | |
2378 | ||
68d6ac6d | 2379 | copied = data_skb->len; |
1da177e4 LT |
2380 | if (len < copied) { |
2381 | msg->msg_flags |= MSG_TRUNC; | |
2382 | copied = len; | |
2383 | } | |
2384 | ||
68d6ac6d JB |
2385 | skb_reset_transport_header(data_skb); |
2386 | err = skb_copy_datagram_iovec(data_skb, 0, msg->msg_iov, copied); | |
1da177e4 LT |
2387 | |
2388 | if (msg->msg_name) { | |
342dfc30 | 2389 | DECLARE_SOCKADDR(struct sockaddr_nl *, addr, msg->msg_name); |
1da177e4 LT |
2390 | addr->nl_family = AF_NETLINK; |
2391 | addr->nl_pad = 0; | |
15e47304 | 2392 | addr->nl_pid = NETLINK_CB(skb).portid; |
d629b836 | 2393 | addr->nl_groups = netlink_group_mask(NETLINK_CB(skb).dst_group); |
1da177e4 LT |
2394 | msg->msg_namelen = sizeof(*addr); |
2395 | } | |
2396 | ||
cc9a06cd PM |
2397 | if (nlk->flags & NETLINK_RECV_PKTINFO) |
2398 | netlink_cmsg_recv_pktinfo(msg, skb); | |
2399 | ||
1da177e4 LT |
2400 | if (NULL == siocb->scm) { |
2401 | memset(&scm, 0, sizeof(scm)); | |
2402 | siocb->scm = &scm; | |
2403 | } | |
2404 | siocb->scm->creds = *NETLINK_CREDS(skb); | |
188ccb55 | 2405 | if (flags & MSG_TRUNC) |
68d6ac6d | 2406 | copied = data_skb->len; |
daa3766e | 2407 | |
1da177e4 LT |
2408 | skb_free_datagram(sk, skb); |
2409 | ||
16b304f3 PS |
2410 | if (nlk->cb_running && |
2411 | atomic_read(&sk->sk_rmem_alloc) <= sk->sk_rcvbuf / 2) { | |
b44d211e AV |
2412 | ret = netlink_dump(sk); |
2413 | if (ret) { | |
2414 | sk->sk_err = ret; | |
2415 | sk->sk_error_report(sk); | |
2416 | } | |
2417 | } | |
1da177e4 LT |
2418 | |
2419 | scm_recv(sock, msg, siocb->scm, flags); | |
1da177e4 LT |
2420 | out: |
2421 | netlink_rcv_wake(sk); | |
2422 | return err ? : copied; | |
2423 | } | |
2424 | ||
676d2369 | 2425 | static void netlink_data_ready(struct sock *sk) |
1da177e4 | 2426 | { |
cd40b7d3 | 2427 | BUG(); |
1da177e4 LT |
2428 | } |
2429 | ||
2430 | /* | |
746fac4d | 2431 | * We export these functions to other modules. They provide a |
1da177e4 LT |
2432 | * complete set of kernel non-blocking support for message |
2433 | * queueing. | |
2434 | */ | |
2435 | ||
2436 | struct sock * | |
9f00d977 PNA |
2437 | __netlink_kernel_create(struct net *net, int unit, struct module *module, |
2438 | struct netlink_kernel_cfg *cfg) | |
1da177e4 LT |
2439 | { |
2440 | struct socket *sock; | |
2441 | struct sock *sk; | |
77247bbb | 2442 | struct netlink_sock *nlk; |
5c398dc8 | 2443 | struct listeners *listeners = NULL; |
a31f2d17 PNA |
2444 | struct mutex *cb_mutex = cfg ? cfg->cb_mutex : NULL; |
2445 | unsigned int groups; | |
1da177e4 | 2446 | |
fab2caf6 | 2447 | BUG_ON(!nl_table); |
1da177e4 | 2448 | |
6ac552fd | 2449 | if (unit < 0 || unit >= MAX_LINKS) |
1da177e4 LT |
2450 | return NULL; |
2451 | ||
2452 | if (sock_create_lite(PF_NETLINK, SOCK_DGRAM, unit, &sock)) | |
2453 | return NULL; | |
2454 | ||
23fe1866 PE |
2455 | /* |
2456 | * We have to just have a reference on the net from sk, but don't | |
2457 | * get_net it. Besides, we cannot get and then put the net here. | |
2458 | * So we create one inside init_net and the move it to net. | |
2459 | */ | |
2460 | ||
2461 | if (__netlink_create(&init_net, sock, cb_mutex, unit) < 0) | |
2462 | goto out_sock_release_nosk; | |
2463 | ||
2464 | sk = sock->sk; | |
edf02087 | 2465 | sk_change_net(sk, net); |
4fdb3bb7 | 2466 | |
a31f2d17 | 2467 | if (!cfg || cfg->groups < 32) |
4277a083 | 2468 | groups = 32; |
a31f2d17 PNA |
2469 | else |
2470 | groups = cfg->groups; | |
4277a083 | 2471 | |
5c398dc8 | 2472 | listeners = kzalloc(sizeof(*listeners) + NLGRPSZ(groups), GFP_KERNEL); |
4277a083 PM |
2473 | if (!listeners) |
2474 | goto out_sock_release; | |
2475 | ||
1da177e4 | 2476 | sk->sk_data_ready = netlink_data_ready; |
a31f2d17 PNA |
2477 | if (cfg && cfg->input) |
2478 | nlk_sk(sk)->netlink_rcv = cfg->input; | |
1da177e4 | 2479 | |
b4b51029 | 2480 | if (netlink_insert(sk, net, 0)) |
77247bbb | 2481 | goto out_sock_release; |
4fdb3bb7 | 2482 | |
77247bbb PM |
2483 | nlk = nlk_sk(sk); |
2484 | nlk->flags |= NETLINK_KERNEL_SOCKET; | |
4fdb3bb7 | 2485 | |
4fdb3bb7 | 2486 | netlink_table_grab(); |
b4b51029 EB |
2487 | if (!nl_table[unit].registered) { |
2488 | nl_table[unit].groups = groups; | |
5c398dc8 | 2489 | rcu_assign_pointer(nl_table[unit].listeners, listeners); |
b4b51029 EB |
2490 | nl_table[unit].cb_mutex = cb_mutex; |
2491 | nl_table[unit].module = module; | |
9785e10a PNA |
2492 | if (cfg) { |
2493 | nl_table[unit].bind = cfg->bind; | |
2494 | nl_table[unit].flags = cfg->flags; | |
da12c90e G |
2495 | if (cfg->compare) |
2496 | nl_table[unit].compare = cfg->compare; | |
9785e10a | 2497 | } |
b4b51029 | 2498 | nl_table[unit].registered = 1; |
f937f1f4 JJ |
2499 | } else { |
2500 | kfree(listeners); | |
869e58f8 | 2501 | nl_table[unit].registered++; |
b4b51029 | 2502 | } |
4fdb3bb7 | 2503 | netlink_table_ungrab(); |
77247bbb PM |
2504 | return sk; |
2505 | ||
4fdb3bb7 | 2506 | out_sock_release: |
4277a083 | 2507 | kfree(listeners); |
9dfbec1f | 2508 | netlink_kernel_release(sk); |
23fe1866 PE |
2509 | return NULL; |
2510 | ||
2511 | out_sock_release_nosk: | |
4fdb3bb7 | 2512 | sock_release(sock); |
77247bbb | 2513 | return NULL; |
1da177e4 | 2514 | } |
9f00d977 | 2515 | EXPORT_SYMBOL(__netlink_kernel_create); |
b7c6ba6e DL |
2516 | |
2517 | void | |
2518 | netlink_kernel_release(struct sock *sk) | |
2519 | { | |
edf02087 | 2520 | sk_release_kernel(sk); |
b7c6ba6e DL |
2521 | } |
2522 | EXPORT_SYMBOL(netlink_kernel_release); | |
2523 | ||
d136f1bd | 2524 | int __netlink_change_ngroups(struct sock *sk, unsigned int groups) |
b4ff4f04 | 2525 | { |
5c398dc8 | 2526 | struct listeners *new, *old; |
b4ff4f04 | 2527 | struct netlink_table *tbl = &nl_table[sk->sk_protocol]; |
b4ff4f04 JB |
2528 | |
2529 | if (groups < 32) | |
2530 | groups = 32; | |
2531 | ||
b4ff4f04 | 2532 | if (NLGRPSZ(tbl->groups) < NLGRPSZ(groups)) { |
5c398dc8 ED |
2533 | new = kzalloc(sizeof(*new) + NLGRPSZ(groups), GFP_ATOMIC); |
2534 | if (!new) | |
d136f1bd | 2535 | return -ENOMEM; |
6d772ac5 | 2536 | old = nl_deref_protected(tbl->listeners); |
5c398dc8 ED |
2537 | memcpy(new->masks, old->masks, NLGRPSZ(tbl->groups)); |
2538 | rcu_assign_pointer(tbl->listeners, new); | |
2539 | ||
37b6b935 | 2540 | kfree_rcu(old, rcu); |
b4ff4f04 JB |
2541 | } |
2542 | tbl->groups = groups; | |
2543 | ||
d136f1bd JB |
2544 | return 0; |
2545 | } | |
2546 | ||
2547 | /** | |
2548 | * netlink_change_ngroups - change number of multicast groups | |
2549 | * | |
2550 | * This changes the number of multicast groups that are available | |
2551 | * on a certain netlink family. Note that it is not possible to | |
2552 | * change the number of groups to below 32. Also note that it does | |
2553 | * not implicitly call netlink_clear_multicast_users() when the | |
2554 | * number of groups is reduced. | |
2555 | * | |
2556 | * @sk: The kernel netlink socket, as returned by netlink_kernel_create(). | |
2557 | * @groups: The new number of groups. | |
2558 | */ | |
2559 | int netlink_change_ngroups(struct sock *sk, unsigned int groups) | |
2560 | { | |
2561 | int err; | |
2562 | ||
2563 | netlink_table_grab(); | |
2564 | err = __netlink_change_ngroups(sk, groups); | |
b4ff4f04 | 2565 | netlink_table_ungrab(); |
d136f1bd | 2566 | |
b4ff4f04 JB |
2567 | return err; |
2568 | } | |
b4ff4f04 | 2569 | |
b8273570 JB |
2570 | void __netlink_clear_multicast_users(struct sock *ksk, unsigned int group) |
2571 | { | |
2572 | struct sock *sk; | |
b8273570 JB |
2573 | struct netlink_table *tbl = &nl_table[ksk->sk_protocol]; |
2574 | ||
b67bfe0d | 2575 | sk_for_each_bound(sk, &tbl->mc_list) |
b8273570 JB |
2576 | netlink_update_socket_mc(nlk_sk(sk), group, 0); |
2577 | } | |
2578 | ||
a46621a3 | 2579 | struct nlmsghdr * |
15e47304 | 2580 | __nlmsg_put(struct sk_buff *skb, u32 portid, u32 seq, int type, int len, int flags) |
a46621a3 DV |
2581 | { |
2582 | struct nlmsghdr *nlh; | |
573ce260 | 2583 | int size = nlmsg_msg_size(len); |
a46621a3 | 2584 | |
23b45672 | 2585 | nlh = (struct nlmsghdr *)skb_put(skb, NLMSG_ALIGN(size)); |
a46621a3 DV |
2586 | nlh->nlmsg_type = type; |
2587 | nlh->nlmsg_len = size; | |
2588 | nlh->nlmsg_flags = flags; | |
15e47304 | 2589 | nlh->nlmsg_pid = portid; |
a46621a3 DV |
2590 | nlh->nlmsg_seq = seq; |
2591 | if (!__builtin_constant_p(size) || NLMSG_ALIGN(size) - size != 0) | |
573ce260 | 2592 | memset(nlmsg_data(nlh) + len, 0, NLMSG_ALIGN(size) - size); |
a46621a3 DV |
2593 | return nlh; |
2594 | } | |
2595 | EXPORT_SYMBOL(__nlmsg_put); | |
2596 | ||
1da177e4 LT |
2597 | /* |
2598 | * It looks a bit ugly. | |
2599 | * It would be better to create kernel thread. | |
2600 | */ | |
2601 | ||
2602 | static int netlink_dump(struct sock *sk) | |
2603 | { | |
2604 | struct netlink_sock *nlk = nlk_sk(sk); | |
2605 | struct netlink_callback *cb; | |
c7ac8679 | 2606 | struct sk_buff *skb = NULL; |
1da177e4 | 2607 | struct nlmsghdr *nlh; |
bf8b79e4 | 2608 | int len, err = -ENOBUFS; |
c7ac8679 | 2609 | int alloc_size; |
1da177e4 | 2610 | |
af65bdfc | 2611 | mutex_lock(nlk->cb_mutex); |
16b304f3 | 2612 | if (!nlk->cb_running) { |
bf8b79e4 TG |
2613 | err = -EINVAL; |
2614 | goto errout_skb; | |
1da177e4 LT |
2615 | } |
2616 | ||
16b304f3 | 2617 | cb = &nlk->cb; |
c7ac8679 GR |
2618 | alloc_size = max_t(int, cb->min_dump_alloc, NLMSG_GOODSIZE); |
2619 | ||
f9c22888 PM |
2620 | if (!netlink_rx_is_mmaped(sk) && |
2621 | atomic_read(&sk->sk_rmem_alloc) >= sk->sk_rcvbuf) | |
2622 | goto errout_skb; | |
9063e21f ED |
2623 | |
2624 | /* NLMSG_GOODSIZE is small to avoid high order allocations being | |
2625 | * required, but it makes sense to _attempt_ a 16K bytes allocation | |
2626 | * to reduce number of system calls on dump operations, if user | |
2627 | * ever provided a big enough buffer. | |
2628 | */ | |
2629 | if (alloc_size < nlk->max_recvmsg_len) { | |
2630 | skb = netlink_alloc_skb(sk, | |
2631 | nlk->max_recvmsg_len, | |
2632 | nlk->portid, | |
2633 | GFP_KERNEL | | |
2634 | __GFP_NOWARN | | |
2635 | __GFP_NORETRY); | |
2636 | /* available room should be exact amount to avoid MSG_TRUNC */ | |
2637 | if (skb) | |
2638 | skb_reserve(skb, skb_tailroom(skb) - | |
2639 | nlk->max_recvmsg_len); | |
2640 | } | |
2641 | if (!skb) | |
2642 | skb = netlink_alloc_skb(sk, alloc_size, nlk->portid, | |
2643 | GFP_KERNEL); | |
c7ac8679 | 2644 | if (!skb) |
c63d6ea3 | 2645 | goto errout_skb; |
f9c22888 | 2646 | netlink_skb_set_owner_r(skb, sk); |
c7ac8679 | 2647 | |
1da177e4 LT |
2648 | len = cb->dump(skb, cb); |
2649 | ||
2650 | if (len > 0) { | |
af65bdfc | 2651 | mutex_unlock(nlk->cb_mutex); |
b1153f29 SH |
2652 | |
2653 | if (sk_filter(sk, skb)) | |
2654 | kfree_skb(skb); | |
4a7e7c2a ED |
2655 | else |
2656 | __netlink_sendskb(sk, skb); | |
1da177e4 LT |
2657 | return 0; |
2658 | } | |
2659 | ||
bf8b79e4 TG |
2660 | nlh = nlmsg_put_answer(skb, cb, NLMSG_DONE, sizeof(len), NLM_F_MULTI); |
2661 | if (!nlh) | |
2662 | goto errout_skb; | |
2663 | ||
670dc283 JB |
2664 | nl_dump_check_consistent(cb, nlh); |
2665 | ||
bf8b79e4 TG |
2666 | memcpy(nlmsg_data(nlh), &len, sizeof(len)); |
2667 | ||
b1153f29 SH |
2668 | if (sk_filter(sk, skb)) |
2669 | kfree_skb(skb); | |
4a7e7c2a ED |
2670 | else |
2671 | __netlink_sendskb(sk, skb); | |
1da177e4 | 2672 | |
a8f74b22 TG |
2673 | if (cb->done) |
2674 | cb->done(cb); | |
1da177e4 | 2675 | |
16b304f3 PS |
2676 | nlk->cb_running = false; |
2677 | mutex_unlock(nlk->cb_mutex); | |
6dc878a8 | 2678 | module_put(cb->module); |
16b304f3 | 2679 | consume_skb(cb->skb); |
1da177e4 | 2680 | return 0; |
1797754e | 2681 | |
bf8b79e4 | 2682 | errout_skb: |
af65bdfc | 2683 | mutex_unlock(nlk->cb_mutex); |
bf8b79e4 | 2684 | kfree_skb(skb); |
bf8b79e4 | 2685 | return err; |
1da177e4 LT |
2686 | } |
2687 | ||
6dc878a8 G |
2688 | int __netlink_dump_start(struct sock *ssk, struct sk_buff *skb, |
2689 | const struct nlmsghdr *nlh, | |
2690 | struct netlink_dump_control *control) | |
1da177e4 LT |
2691 | { |
2692 | struct netlink_callback *cb; | |
2693 | struct sock *sk; | |
2694 | struct netlink_sock *nlk; | |
b44d211e | 2695 | int ret; |
1da177e4 | 2696 | |
f9c22888 PM |
2697 | /* Memory mapped dump requests need to be copied to avoid looping |
2698 | * on the pending state in netlink_mmap_sendmsg() while the CB hold | |
2699 | * a reference to the skb. | |
2700 | */ | |
2701 | if (netlink_skb_is_mmaped(skb)) { | |
2702 | skb = skb_copy(skb, GFP_KERNEL); | |
16b304f3 | 2703 | if (skb == NULL) |
f9c22888 | 2704 | return -ENOBUFS; |
f9c22888 PM |
2705 | } else |
2706 | atomic_inc(&skb->users); | |
2707 | ||
15e47304 | 2708 | sk = netlink_lookup(sock_net(ssk), ssk->sk_protocol, NETLINK_CB(skb).portid); |
1da177e4 | 2709 | if (sk == NULL) { |
16b304f3 PS |
2710 | ret = -ECONNREFUSED; |
2711 | goto error_free; | |
1da177e4 | 2712 | } |
6dc878a8 | 2713 | |
16b304f3 | 2714 | nlk = nlk_sk(sk); |
af65bdfc | 2715 | mutex_lock(nlk->cb_mutex); |
6dc878a8 | 2716 | /* A dump is in progress... */ |
16b304f3 | 2717 | if (nlk->cb_running) { |
6dc878a8 | 2718 | ret = -EBUSY; |
16b304f3 | 2719 | goto error_unlock; |
1da177e4 | 2720 | } |
6dc878a8 | 2721 | /* add reference of module which cb->dump belongs to */ |
16b304f3 | 2722 | if (!try_module_get(control->module)) { |
6dc878a8 | 2723 | ret = -EPROTONOSUPPORT; |
16b304f3 | 2724 | goto error_unlock; |
6dc878a8 G |
2725 | } |
2726 | ||
16b304f3 PS |
2727 | cb = &nlk->cb; |
2728 | memset(cb, 0, sizeof(*cb)); | |
2729 | cb->dump = control->dump; | |
2730 | cb->done = control->done; | |
2731 | cb->nlh = nlh; | |
2732 | cb->data = control->data; | |
2733 | cb->module = control->module; | |
2734 | cb->min_dump_alloc = control->min_dump_alloc; | |
2735 | cb->skb = skb; | |
2736 | ||
2737 | nlk->cb_running = true; | |
2738 | ||
af65bdfc | 2739 | mutex_unlock(nlk->cb_mutex); |
1da177e4 | 2740 | |
b44d211e | 2741 | ret = netlink_dump(sk); |
1da177e4 | 2742 | sock_put(sk); |
5c58298c | 2743 | |
b44d211e AV |
2744 | if (ret) |
2745 | return ret; | |
2746 | ||
5c58298c DL |
2747 | /* We successfully started a dump, by returning -EINTR we |
2748 | * signal not to send ACK even if it was requested. | |
2749 | */ | |
2750 | return -EINTR; | |
16b304f3 PS |
2751 | |
2752 | error_unlock: | |
2753 | sock_put(sk); | |
2754 | mutex_unlock(nlk->cb_mutex); | |
2755 | error_free: | |
2756 | kfree_skb(skb); | |
2757 | return ret; | |
1da177e4 | 2758 | } |
6dc878a8 | 2759 | EXPORT_SYMBOL(__netlink_dump_start); |
1da177e4 LT |
2760 | |
2761 | void netlink_ack(struct sk_buff *in_skb, struct nlmsghdr *nlh, int err) | |
2762 | { | |
2763 | struct sk_buff *skb; | |
2764 | struct nlmsghdr *rep; | |
2765 | struct nlmsgerr *errmsg; | |
339bf98f | 2766 | size_t payload = sizeof(*errmsg); |
1da177e4 | 2767 | |
339bf98f TG |
2768 | /* error messages get the original request appened */ |
2769 | if (err) | |
2770 | payload += nlmsg_len(nlh); | |
1da177e4 | 2771 | |
f9c22888 PM |
2772 | skb = netlink_alloc_skb(in_skb->sk, nlmsg_total_size(payload), |
2773 | NETLINK_CB(in_skb).portid, GFP_KERNEL); | |
1da177e4 LT |
2774 | if (!skb) { |
2775 | struct sock *sk; | |
2776 | ||
3b1e0a65 | 2777 | sk = netlink_lookup(sock_net(in_skb->sk), |
b4b51029 | 2778 | in_skb->sk->sk_protocol, |
15e47304 | 2779 | NETLINK_CB(in_skb).portid); |
1da177e4 LT |
2780 | if (sk) { |
2781 | sk->sk_err = ENOBUFS; | |
2782 | sk->sk_error_report(sk); | |
2783 | sock_put(sk); | |
2784 | } | |
2785 | return; | |
2786 | } | |
2787 | ||
15e47304 | 2788 | rep = __nlmsg_put(skb, NETLINK_CB(in_skb).portid, nlh->nlmsg_seq, |
5dba93ae | 2789 | NLMSG_ERROR, payload, 0); |
bf8b79e4 | 2790 | errmsg = nlmsg_data(rep); |
1da177e4 | 2791 | errmsg->error = err; |
bf8b79e4 | 2792 | memcpy(&errmsg->msg, nlh, err ? nlh->nlmsg_len : sizeof(*nlh)); |
15e47304 | 2793 | netlink_unicast(in_skb->sk, skb, NETLINK_CB(in_skb).portid, MSG_DONTWAIT); |
1da177e4 | 2794 | } |
6ac552fd | 2795 | EXPORT_SYMBOL(netlink_ack); |
1da177e4 | 2796 | |
cd40b7d3 | 2797 | int netlink_rcv_skb(struct sk_buff *skb, int (*cb)(struct sk_buff *, |
1d00a4eb | 2798 | struct nlmsghdr *)) |
82ace47a | 2799 | { |
82ace47a TG |
2800 | struct nlmsghdr *nlh; |
2801 | int err; | |
2802 | ||
2803 | while (skb->len >= nlmsg_total_size(0)) { | |
cd40b7d3 DL |
2804 | int msglen; |
2805 | ||
b529ccf2 | 2806 | nlh = nlmsg_hdr(skb); |
d35b6856 | 2807 | err = 0; |
82ace47a | 2808 | |
ad8e4b75 | 2809 | if (nlh->nlmsg_len < NLMSG_HDRLEN || skb->len < nlh->nlmsg_len) |
82ace47a TG |
2810 | return 0; |
2811 | ||
d35b6856 TG |
2812 | /* Only requests are handled by the kernel */ |
2813 | if (!(nlh->nlmsg_flags & NLM_F_REQUEST)) | |
5c58298c | 2814 | goto ack; |
45e7ae7f TG |
2815 | |
2816 | /* Skip control messages */ | |
2817 | if (nlh->nlmsg_type < NLMSG_MIN_TYPE) | |
5c58298c | 2818 | goto ack; |
d35b6856 | 2819 | |
1d00a4eb | 2820 | err = cb(skb, nlh); |
5c58298c DL |
2821 | if (err == -EINTR) |
2822 | goto skip; | |
2823 | ||
2824 | ack: | |
d35b6856 | 2825 | if (nlh->nlmsg_flags & NLM_F_ACK || err) |
82ace47a | 2826 | netlink_ack(skb, nlh, err); |
82ace47a | 2827 | |
5c58298c | 2828 | skip: |
6ac552fd | 2829 | msglen = NLMSG_ALIGN(nlh->nlmsg_len); |
cd40b7d3 DL |
2830 | if (msglen > skb->len) |
2831 | msglen = skb->len; | |
2832 | skb_pull(skb, msglen); | |
82ace47a TG |
2833 | } |
2834 | ||
2835 | return 0; | |
2836 | } | |
6ac552fd | 2837 | EXPORT_SYMBOL(netlink_rcv_skb); |
82ace47a | 2838 | |
d387f6ad TG |
2839 | /** |
2840 | * nlmsg_notify - send a notification netlink message | |
2841 | * @sk: netlink socket to use | |
2842 | * @skb: notification message | |
15e47304 | 2843 | * @portid: destination netlink portid for reports or 0 |
d387f6ad TG |
2844 | * @group: destination multicast group or 0 |
2845 | * @report: 1 to report back, 0 to disable | |
2846 | * @flags: allocation flags | |
2847 | */ | |
15e47304 | 2848 | int nlmsg_notify(struct sock *sk, struct sk_buff *skb, u32 portid, |
d387f6ad TG |
2849 | unsigned int group, int report, gfp_t flags) |
2850 | { | |
2851 | int err = 0; | |
2852 | ||
2853 | if (group) { | |
15e47304 | 2854 | int exclude_portid = 0; |
d387f6ad TG |
2855 | |
2856 | if (report) { | |
2857 | atomic_inc(&skb->users); | |
15e47304 | 2858 | exclude_portid = portid; |
d387f6ad TG |
2859 | } |
2860 | ||
1ce85fe4 PNA |
2861 | /* errors reported via destination sk->sk_err, but propagate |
2862 | * delivery errors if NETLINK_BROADCAST_ERROR flag is set */ | |
15e47304 | 2863 | err = nlmsg_multicast(sk, skb, exclude_portid, group, flags); |
d387f6ad TG |
2864 | } |
2865 | ||
1ce85fe4 PNA |
2866 | if (report) { |
2867 | int err2; | |
2868 | ||
15e47304 | 2869 | err2 = nlmsg_unicast(sk, skb, portid); |
1ce85fe4 PNA |
2870 | if (!err || err == -ESRCH) |
2871 | err = err2; | |
2872 | } | |
d387f6ad TG |
2873 | |
2874 | return err; | |
2875 | } | |
6ac552fd | 2876 | EXPORT_SYMBOL(nlmsg_notify); |
d387f6ad | 2877 | |
1da177e4 LT |
2878 | #ifdef CONFIG_PROC_FS |
2879 | struct nl_seq_iter { | |
e372c414 | 2880 | struct seq_net_private p; |
1da177e4 LT |
2881 | int link; |
2882 | int hash_idx; | |
2883 | }; | |
2884 | ||
2885 | static struct sock *netlink_seq_socket_idx(struct seq_file *seq, loff_t pos) | |
2886 | { | |
2887 | struct nl_seq_iter *iter = seq->private; | |
2888 | int i, j; | |
2889 | struct sock *s; | |
1da177e4 LT |
2890 | loff_t off = 0; |
2891 | ||
6ac552fd | 2892 | for (i = 0; i < MAX_LINKS; i++) { |
15e47304 | 2893 | struct nl_portid_hash *hash = &nl_table[i].hash; |
1da177e4 LT |
2894 | |
2895 | for (j = 0; j <= hash->mask; j++) { | |
b67bfe0d | 2896 | sk_for_each(s, &hash->table[j]) { |
1218854a | 2897 | if (sock_net(s) != seq_file_net(seq)) |
b4b51029 | 2898 | continue; |
1da177e4 LT |
2899 | if (off == pos) { |
2900 | iter->link = i; | |
2901 | iter->hash_idx = j; | |
2902 | return s; | |
2903 | } | |
2904 | ++off; | |
2905 | } | |
2906 | } | |
2907 | } | |
2908 | return NULL; | |
2909 | } | |
2910 | ||
2911 | static void *netlink_seq_start(struct seq_file *seq, loff_t *pos) | |
9a429c49 | 2912 | __acquires(nl_table_lock) |
1da177e4 LT |
2913 | { |
2914 | read_lock(&nl_table_lock); | |
2915 | return *pos ? netlink_seq_socket_idx(seq, *pos - 1) : SEQ_START_TOKEN; | |
2916 | } | |
2917 | ||
2918 | static void *netlink_seq_next(struct seq_file *seq, void *v, loff_t *pos) | |
2919 | { | |
2920 | struct sock *s; | |
2921 | struct nl_seq_iter *iter; | |
da12c90e | 2922 | struct net *net; |
1da177e4 LT |
2923 | int i, j; |
2924 | ||
2925 | ++*pos; | |
2926 | ||
2927 | if (v == SEQ_START_TOKEN) | |
2928 | return netlink_seq_socket_idx(seq, 0); | |
746fac4d | 2929 | |
da12c90e | 2930 | net = seq_file_net(seq); |
b4b51029 EB |
2931 | iter = seq->private; |
2932 | s = v; | |
2933 | do { | |
2934 | s = sk_next(s); | |
da12c90e | 2935 | } while (s && !nl_table[s->sk_protocol].compare(net, s)); |
1da177e4 LT |
2936 | if (s) |
2937 | return s; | |
2938 | ||
1da177e4 LT |
2939 | i = iter->link; |
2940 | j = iter->hash_idx + 1; | |
2941 | ||
2942 | do { | |
15e47304 | 2943 | struct nl_portid_hash *hash = &nl_table[i].hash; |
1da177e4 LT |
2944 | |
2945 | for (; j <= hash->mask; j++) { | |
2946 | s = sk_head(&hash->table[j]); | |
da12c90e G |
2947 | |
2948 | while (s && !nl_table[s->sk_protocol].compare(net, s)) | |
b4b51029 | 2949 | s = sk_next(s); |
1da177e4 LT |
2950 | if (s) { |
2951 | iter->link = i; | |
2952 | iter->hash_idx = j; | |
2953 | return s; | |
2954 | } | |
2955 | } | |
2956 | ||
2957 | j = 0; | |
2958 | } while (++i < MAX_LINKS); | |
2959 | ||
2960 | return NULL; | |
2961 | } | |
2962 | ||
2963 | static void netlink_seq_stop(struct seq_file *seq, void *v) | |
9a429c49 | 2964 | __releases(nl_table_lock) |
1da177e4 LT |
2965 | { |
2966 | read_unlock(&nl_table_lock); | |
2967 | } | |
2968 | ||
2969 | ||
2970 | static int netlink_seq_show(struct seq_file *seq, void *v) | |
2971 | { | |
658cb354 | 2972 | if (v == SEQ_START_TOKEN) { |
1da177e4 LT |
2973 | seq_puts(seq, |
2974 | "sk Eth Pid Groups " | |
cf0aa4e0 | 2975 | "Rmem Wmem Dump Locks Drops Inode\n"); |
658cb354 | 2976 | } else { |
1da177e4 LT |
2977 | struct sock *s = v; |
2978 | struct netlink_sock *nlk = nlk_sk(s); | |
2979 | ||
16b304f3 | 2980 | seq_printf(seq, "%pK %-3d %-6u %08x %-8d %-8d %d %-8d %-8d %-8lu\n", |
1da177e4 LT |
2981 | s, |
2982 | s->sk_protocol, | |
15e47304 | 2983 | nlk->portid, |
513c2500 | 2984 | nlk->groups ? (u32)nlk->groups[0] : 0, |
31e6d363 ED |
2985 | sk_rmem_alloc_get(s), |
2986 | sk_wmem_alloc_get(s), | |
16b304f3 | 2987 | nlk->cb_running, |
38938bfe | 2988 | atomic_read(&s->sk_refcnt), |
cf0aa4e0 MY |
2989 | atomic_read(&s->sk_drops), |
2990 | sock_i_ino(s) | |
1da177e4 LT |
2991 | ); |
2992 | ||
2993 | } | |
2994 | return 0; | |
2995 | } | |
2996 | ||
56b3d975 | 2997 | static const struct seq_operations netlink_seq_ops = { |
1da177e4 LT |
2998 | .start = netlink_seq_start, |
2999 | .next = netlink_seq_next, | |
3000 | .stop = netlink_seq_stop, | |
3001 | .show = netlink_seq_show, | |
3002 | }; | |
3003 | ||
3004 | ||
3005 | static int netlink_seq_open(struct inode *inode, struct file *file) | |
3006 | { | |
e372c414 DL |
3007 | return seq_open_net(inode, file, &netlink_seq_ops, |
3008 | sizeof(struct nl_seq_iter)); | |
b4b51029 EB |
3009 | } |
3010 | ||
da7071d7 | 3011 | static const struct file_operations netlink_seq_fops = { |
1da177e4 LT |
3012 | .owner = THIS_MODULE, |
3013 | .open = netlink_seq_open, | |
3014 | .read = seq_read, | |
3015 | .llseek = seq_lseek, | |
e372c414 | 3016 | .release = seq_release_net, |
1da177e4 LT |
3017 | }; |
3018 | ||
3019 | #endif | |
3020 | ||
3021 | int netlink_register_notifier(struct notifier_block *nb) | |
3022 | { | |
e041c683 | 3023 | return atomic_notifier_chain_register(&netlink_chain, nb); |
1da177e4 | 3024 | } |
6ac552fd | 3025 | EXPORT_SYMBOL(netlink_register_notifier); |
1da177e4 LT |
3026 | |
3027 | int netlink_unregister_notifier(struct notifier_block *nb) | |
3028 | { | |
e041c683 | 3029 | return atomic_notifier_chain_unregister(&netlink_chain, nb); |
1da177e4 | 3030 | } |
6ac552fd | 3031 | EXPORT_SYMBOL(netlink_unregister_notifier); |
746fac4d | 3032 | |
90ddc4f0 | 3033 | static const struct proto_ops netlink_ops = { |
1da177e4 LT |
3034 | .family = PF_NETLINK, |
3035 | .owner = THIS_MODULE, | |
3036 | .release = netlink_release, | |
3037 | .bind = netlink_bind, | |
3038 | .connect = netlink_connect, | |
3039 | .socketpair = sock_no_socketpair, | |
3040 | .accept = sock_no_accept, | |
3041 | .getname = netlink_getname, | |
9652e931 | 3042 | .poll = netlink_poll, |
1da177e4 LT |
3043 | .ioctl = sock_no_ioctl, |
3044 | .listen = sock_no_listen, | |
3045 | .shutdown = sock_no_shutdown, | |
9a4595bc PM |
3046 | .setsockopt = netlink_setsockopt, |
3047 | .getsockopt = netlink_getsockopt, | |
1da177e4 LT |
3048 | .sendmsg = netlink_sendmsg, |
3049 | .recvmsg = netlink_recvmsg, | |
ccdfcc39 | 3050 | .mmap = netlink_mmap, |
1da177e4 LT |
3051 | .sendpage = sock_no_sendpage, |
3052 | }; | |
3053 | ||
ec1b4cf7 | 3054 | static const struct net_proto_family netlink_family_ops = { |
1da177e4 LT |
3055 | .family = PF_NETLINK, |
3056 | .create = netlink_create, | |
3057 | .owner = THIS_MODULE, /* for consistency 8) */ | |
3058 | }; | |
3059 | ||
4665079c | 3060 | static int __net_init netlink_net_init(struct net *net) |
b4b51029 EB |
3061 | { |
3062 | #ifdef CONFIG_PROC_FS | |
d4beaa66 | 3063 | if (!proc_create("netlink", 0, net->proc_net, &netlink_seq_fops)) |
b4b51029 EB |
3064 | return -ENOMEM; |
3065 | #endif | |
3066 | return 0; | |
3067 | } | |
3068 | ||
4665079c | 3069 | static void __net_exit netlink_net_exit(struct net *net) |
b4b51029 EB |
3070 | { |
3071 | #ifdef CONFIG_PROC_FS | |
ece31ffd | 3072 | remove_proc_entry("netlink", net->proc_net); |
b4b51029 EB |
3073 | #endif |
3074 | } | |
3075 | ||
b963ea89 DM |
3076 | static void __init netlink_add_usersock_entry(void) |
3077 | { | |
5c398dc8 | 3078 | struct listeners *listeners; |
b963ea89 DM |
3079 | int groups = 32; |
3080 | ||
5c398dc8 | 3081 | listeners = kzalloc(sizeof(*listeners) + NLGRPSZ(groups), GFP_KERNEL); |
b963ea89 | 3082 | if (!listeners) |
5c398dc8 | 3083 | panic("netlink_add_usersock_entry: Cannot allocate listeners\n"); |
b963ea89 DM |
3084 | |
3085 | netlink_table_grab(); | |
3086 | ||
3087 | nl_table[NETLINK_USERSOCK].groups = groups; | |
5c398dc8 | 3088 | rcu_assign_pointer(nl_table[NETLINK_USERSOCK].listeners, listeners); |
b963ea89 DM |
3089 | nl_table[NETLINK_USERSOCK].module = THIS_MODULE; |
3090 | nl_table[NETLINK_USERSOCK].registered = 1; | |
9785e10a | 3091 | nl_table[NETLINK_USERSOCK].flags = NL_CFG_F_NONROOT_SEND; |
b963ea89 DM |
3092 | |
3093 | netlink_table_ungrab(); | |
3094 | } | |
3095 | ||
022cbae6 | 3096 | static struct pernet_operations __net_initdata netlink_net_ops = { |
b4b51029 EB |
3097 | .init = netlink_net_init, |
3098 | .exit = netlink_net_exit, | |
3099 | }; | |
3100 | ||
1da177e4 LT |
3101 | static int __init netlink_proto_init(void) |
3102 | { | |
1da177e4 | 3103 | int i; |
26ff5ddc | 3104 | unsigned long limit; |
1da177e4 LT |
3105 | unsigned int order; |
3106 | int err = proto_register(&netlink_proto, 0); | |
3107 | ||
3108 | if (err != 0) | |
3109 | goto out; | |
3110 | ||
fab25745 | 3111 | BUILD_BUG_ON(sizeof(struct netlink_skb_parms) > FIELD_SIZEOF(struct sk_buff, cb)); |
1da177e4 | 3112 | |
0da974f4 | 3113 | nl_table = kcalloc(MAX_LINKS, sizeof(*nl_table), GFP_KERNEL); |
fab2caf6 AM |
3114 | if (!nl_table) |
3115 | goto panic; | |
1da177e4 | 3116 | |
4481374c JB |
3117 | if (totalram_pages >= (128 * 1024)) |
3118 | limit = totalram_pages >> (21 - PAGE_SHIFT); | |
1da177e4 | 3119 | else |
4481374c | 3120 | limit = totalram_pages >> (23 - PAGE_SHIFT); |
1da177e4 | 3121 | |
26ff5ddc DC |
3122 | order = get_bitmask_order(limit) - 1 + PAGE_SHIFT; |
3123 | limit = (1UL << order) / sizeof(struct hlist_head); | |
3124 | order = get_bitmask_order(min(limit, (unsigned long)UINT_MAX)) - 1; | |
1da177e4 LT |
3125 | |
3126 | for (i = 0; i < MAX_LINKS; i++) { | |
15e47304 | 3127 | struct nl_portid_hash *hash = &nl_table[i].hash; |
1da177e4 | 3128 | |
15e47304 | 3129 | hash->table = nl_portid_hash_zalloc(1 * sizeof(*hash->table)); |
1da177e4 LT |
3130 | if (!hash->table) { |
3131 | while (i-- > 0) | |
15e47304 | 3132 | nl_portid_hash_free(nl_table[i].hash.table, |
1da177e4 LT |
3133 | 1 * sizeof(*hash->table)); |
3134 | kfree(nl_table); | |
fab2caf6 | 3135 | goto panic; |
1da177e4 | 3136 | } |
1da177e4 LT |
3137 | hash->max_shift = order; |
3138 | hash->shift = 0; | |
3139 | hash->mask = 0; | |
3140 | hash->rehash_time = jiffies; | |
ca15febf G |
3141 | |
3142 | nl_table[i].compare = netlink_compare; | |
1da177e4 LT |
3143 | } |
3144 | ||
bcbde0d4 DB |
3145 | INIT_LIST_HEAD(&netlink_tap_all); |
3146 | ||
b963ea89 DM |
3147 | netlink_add_usersock_entry(); |
3148 | ||
1da177e4 | 3149 | sock_register(&netlink_family_ops); |
b4b51029 | 3150 | register_pernet_subsys(&netlink_net_ops); |
746fac4d | 3151 | /* The netlink device handler may be needed early. */ |
1da177e4 LT |
3152 | rtnetlink_init(); |
3153 | out: | |
3154 | return err; | |
fab2caf6 AM |
3155 | panic: |
3156 | panic("netlink_init: Cannot allocate nl_table\n"); | |
1da177e4 LT |
3157 | } |
3158 | ||
1da177e4 | 3159 | core_initcall(netlink_proto_init); |