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; | |
5ffd5cdd DB |
173 | |
174 | /* We take the more conservative approach and | |
175 | * whitelist socket protocols that may pass. | |
176 | */ | |
177 | switch (sk->sk_protocol) { | |
178 | case NETLINK_ROUTE: | |
179 | case NETLINK_USERSOCK: | |
180 | case NETLINK_SOCK_DIAG: | |
181 | case NETLINK_NFLOG: | |
182 | case NETLINK_XFRM: | |
183 | case NETLINK_FIB_LOOKUP: | |
184 | case NETLINK_NETFILTER: | |
185 | case NETLINK_GENERIC: | |
498044bb | 186 | return true; |
5ffd5cdd DB |
187 | } |
188 | ||
498044bb | 189 | return false; |
5ffd5cdd DB |
190 | } |
191 | ||
bcbde0d4 DB |
192 | static int __netlink_deliver_tap_skb(struct sk_buff *skb, |
193 | struct net_device *dev) | |
194 | { | |
195 | struct sk_buff *nskb; | |
5ffd5cdd | 196 | struct sock *sk = skb->sk; |
bcbde0d4 DB |
197 | int ret = -ENOMEM; |
198 | ||
199 | dev_hold(dev); | |
200 | nskb = skb_clone(skb, GFP_ATOMIC); | |
201 | if (nskb) { | |
202 | nskb->dev = dev; | |
5ffd5cdd | 203 | nskb->protocol = htons((u16) sk->sk_protocol); |
604d13c9 DB |
204 | nskb->pkt_type = netlink_is_kernel(sk) ? |
205 | PACKET_KERNEL : PACKET_USER; | |
5ffd5cdd | 206 | |
bcbde0d4 DB |
207 | ret = dev_queue_xmit(nskb); |
208 | if (unlikely(ret > 0)) | |
209 | ret = net_xmit_errno(ret); | |
210 | } | |
211 | ||
212 | dev_put(dev); | |
213 | return ret; | |
214 | } | |
215 | ||
216 | static void __netlink_deliver_tap(struct sk_buff *skb) | |
217 | { | |
218 | int ret; | |
219 | struct netlink_tap *tmp; | |
220 | ||
5ffd5cdd DB |
221 | if (!netlink_filter_tap(skb)) |
222 | return; | |
223 | ||
bcbde0d4 DB |
224 | list_for_each_entry_rcu(tmp, &netlink_tap_all, list) { |
225 | ret = __netlink_deliver_tap_skb(skb, tmp->dev); | |
226 | if (unlikely(ret)) | |
227 | break; | |
228 | } | |
229 | } | |
230 | ||
231 | static void netlink_deliver_tap(struct sk_buff *skb) | |
232 | { | |
233 | rcu_read_lock(); | |
234 | ||
235 | if (unlikely(!list_empty(&netlink_tap_all))) | |
236 | __netlink_deliver_tap(skb); | |
237 | ||
238 | rcu_read_unlock(); | |
239 | } | |
240 | ||
73bfd370 DB |
241 | static void netlink_deliver_tap_kernel(struct sock *dst, struct sock *src, |
242 | struct sk_buff *skb) | |
243 | { | |
244 | if (!(netlink_is_kernel(dst) && netlink_is_kernel(src))) | |
245 | netlink_deliver_tap(skb); | |
246 | } | |
247 | ||
cd1df525 PM |
248 | static void netlink_overrun(struct sock *sk) |
249 | { | |
250 | struct netlink_sock *nlk = nlk_sk(sk); | |
251 | ||
252 | if (!(nlk->flags & NETLINK_RECV_NO_ENOBUFS)) { | |
253 | if (!test_and_set_bit(NETLINK_CONGESTED, &nlk_sk(sk)->state)) { | |
254 | sk->sk_err = ENOBUFS; | |
255 | sk->sk_error_report(sk); | |
256 | } | |
257 | } | |
258 | atomic_inc(&sk->sk_drops); | |
259 | } | |
260 | ||
261 | static void netlink_rcv_wake(struct sock *sk) | |
262 | { | |
263 | struct netlink_sock *nlk = nlk_sk(sk); | |
264 | ||
265 | if (skb_queue_empty(&sk->sk_receive_queue)) | |
266 | clear_bit(NETLINK_CONGESTED, &nlk->state); | |
267 | if (!test_bit(NETLINK_CONGESTED, &nlk->state)) | |
268 | wake_up_interruptible(&nlk->wait); | |
269 | } | |
270 | ||
ccdfcc39 | 271 | #ifdef CONFIG_NETLINK_MMAP |
9652e931 PM |
272 | static bool netlink_skb_is_mmaped(const struct sk_buff *skb) |
273 | { | |
274 | return NETLINK_CB(skb).flags & NETLINK_SKB_MMAPED; | |
275 | } | |
276 | ||
f9c22888 PM |
277 | static bool netlink_rx_is_mmaped(struct sock *sk) |
278 | { | |
279 | return nlk_sk(sk)->rx_ring.pg_vec != NULL; | |
280 | } | |
281 | ||
5fd96123 PM |
282 | static bool netlink_tx_is_mmaped(struct sock *sk) |
283 | { | |
284 | return nlk_sk(sk)->tx_ring.pg_vec != NULL; | |
285 | } | |
286 | ||
ccdfcc39 PM |
287 | static __pure struct page *pgvec_to_page(const void *addr) |
288 | { | |
289 | if (is_vmalloc_addr(addr)) | |
290 | return vmalloc_to_page(addr); | |
291 | else | |
292 | return virt_to_page(addr); | |
293 | } | |
294 | ||
295 | static void free_pg_vec(void **pg_vec, unsigned int order, unsigned int len) | |
296 | { | |
297 | unsigned int i; | |
298 | ||
299 | for (i = 0; i < len; i++) { | |
300 | if (pg_vec[i] != NULL) { | |
301 | if (is_vmalloc_addr(pg_vec[i])) | |
302 | vfree(pg_vec[i]); | |
303 | else | |
304 | free_pages((unsigned long)pg_vec[i], order); | |
305 | } | |
306 | } | |
307 | kfree(pg_vec); | |
308 | } | |
309 | ||
310 | static void *alloc_one_pg_vec_page(unsigned long order) | |
311 | { | |
312 | void *buffer; | |
313 | gfp_t gfp_flags = GFP_KERNEL | __GFP_COMP | __GFP_ZERO | | |
314 | __GFP_NOWARN | __GFP_NORETRY; | |
315 | ||
316 | buffer = (void *)__get_free_pages(gfp_flags, order); | |
317 | if (buffer != NULL) | |
318 | return buffer; | |
319 | ||
320 | buffer = vzalloc((1 << order) * PAGE_SIZE); | |
321 | if (buffer != NULL) | |
322 | return buffer; | |
323 | ||
324 | gfp_flags &= ~__GFP_NORETRY; | |
325 | return (void *)__get_free_pages(gfp_flags, order); | |
326 | } | |
327 | ||
328 | static void **alloc_pg_vec(struct netlink_sock *nlk, | |
329 | struct nl_mmap_req *req, unsigned int order) | |
330 | { | |
331 | unsigned int block_nr = req->nm_block_nr; | |
332 | unsigned int i; | |
8a849bb7 | 333 | void **pg_vec; |
ccdfcc39 PM |
334 | |
335 | pg_vec = kcalloc(block_nr, sizeof(void *), GFP_KERNEL); | |
336 | if (pg_vec == NULL) | |
337 | return NULL; | |
338 | ||
339 | for (i = 0; i < block_nr; i++) { | |
8a849bb7 | 340 | pg_vec[i] = alloc_one_pg_vec_page(order); |
ccdfcc39 PM |
341 | if (pg_vec[i] == NULL) |
342 | goto err1; | |
343 | } | |
344 | ||
345 | return pg_vec; | |
346 | err1: | |
347 | free_pg_vec(pg_vec, order, block_nr); | |
348 | return NULL; | |
349 | } | |
350 | ||
351 | static int netlink_set_ring(struct sock *sk, struct nl_mmap_req *req, | |
352 | bool closing, bool tx_ring) | |
353 | { | |
354 | struct netlink_sock *nlk = nlk_sk(sk); | |
355 | struct netlink_ring *ring; | |
356 | struct sk_buff_head *queue; | |
357 | void **pg_vec = NULL; | |
358 | unsigned int order = 0; | |
359 | int err; | |
360 | ||
361 | ring = tx_ring ? &nlk->tx_ring : &nlk->rx_ring; | |
362 | queue = tx_ring ? &sk->sk_write_queue : &sk->sk_receive_queue; | |
363 | ||
364 | if (!closing) { | |
365 | if (atomic_read(&nlk->mapped)) | |
366 | return -EBUSY; | |
367 | if (atomic_read(&ring->pending)) | |
368 | return -EBUSY; | |
369 | } | |
370 | ||
371 | if (req->nm_block_nr) { | |
372 | if (ring->pg_vec != NULL) | |
373 | return -EBUSY; | |
374 | ||
375 | if ((int)req->nm_block_size <= 0) | |
376 | return -EINVAL; | |
74e83b23 | 377 | if (!PAGE_ALIGNED(req->nm_block_size)) |
ccdfcc39 PM |
378 | return -EINVAL; |
379 | if (req->nm_frame_size < NL_MMAP_HDRLEN) | |
380 | return -EINVAL; | |
381 | if (!IS_ALIGNED(req->nm_frame_size, NL_MMAP_MSG_ALIGNMENT)) | |
382 | return -EINVAL; | |
383 | ||
384 | ring->frames_per_block = req->nm_block_size / | |
385 | req->nm_frame_size; | |
386 | if (ring->frames_per_block == 0) | |
387 | return -EINVAL; | |
388 | if (ring->frames_per_block * req->nm_block_nr != | |
389 | req->nm_frame_nr) | |
390 | return -EINVAL; | |
391 | ||
392 | order = get_order(req->nm_block_size); | |
393 | pg_vec = alloc_pg_vec(nlk, req, order); | |
394 | if (pg_vec == NULL) | |
395 | return -ENOMEM; | |
396 | } else { | |
397 | if (req->nm_frame_nr) | |
398 | return -EINVAL; | |
399 | } | |
400 | ||
401 | err = -EBUSY; | |
402 | mutex_lock(&nlk->pg_vec_lock); | |
403 | if (closing || atomic_read(&nlk->mapped) == 0) { | |
404 | err = 0; | |
405 | spin_lock_bh(&queue->lock); | |
406 | ||
407 | ring->frame_max = req->nm_frame_nr - 1; | |
408 | ring->head = 0; | |
409 | ring->frame_size = req->nm_frame_size; | |
410 | ring->pg_vec_pages = req->nm_block_size / PAGE_SIZE; | |
411 | ||
412 | swap(ring->pg_vec_len, req->nm_block_nr); | |
413 | swap(ring->pg_vec_order, order); | |
414 | swap(ring->pg_vec, pg_vec); | |
415 | ||
416 | __skb_queue_purge(queue); | |
417 | spin_unlock_bh(&queue->lock); | |
418 | ||
419 | WARN_ON(atomic_read(&nlk->mapped)); | |
420 | } | |
421 | mutex_unlock(&nlk->pg_vec_lock); | |
422 | ||
423 | if (pg_vec) | |
424 | free_pg_vec(pg_vec, order, req->nm_block_nr); | |
425 | return err; | |
426 | } | |
427 | ||
428 | static void netlink_mm_open(struct vm_area_struct *vma) | |
429 | { | |
430 | struct file *file = vma->vm_file; | |
431 | struct socket *sock = file->private_data; | |
432 | struct sock *sk = sock->sk; | |
433 | ||
434 | if (sk) | |
435 | atomic_inc(&nlk_sk(sk)->mapped); | |
436 | } | |
437 | ||
438 | static void netlink_mm_close(struct vm_area_struct *vma) | |
439 | { | |
440 | struct file *file = vma->vm_file; | |
441 | struct socket *sock = file->private_data; | |
442 | struct sock *sk = sock->sk; | |
443 | ||
444 | if (sk) | |
445 | atomic_dec(&nlk_sk(sk)->mapped); | |
446 | } | |
447 | ||
448 | static const struct vm_operations_struct netlink_mmap_ops = { | |
449 | .open = netlink_mm_open, | |
450 | .close = netlink_mm_close, | |
451 | }; | |
452 | ||
453 | static int netlink_mmap(struct file *file, struct socket *sock, | |
454 | struct vm_area_struct *vma) | |
455 | { | |
456 | struct sock *sk = sock->sk; | |
457 | struct netlink_sock *nlk = nlk_sk(sk); | |
458 | struct netlink_ring *ring; | |
459 | unsigned long start, size, expected; | |
460 | unsigned int i; | |
461 | int err = -EINVAL; | |
462 | ||
463 | if (vma->vm_pgoff) | |
464 | return -EINVAL; | |
465 | ||
466 | mutex_lock(&nlk->pg_vec_lock); | |
467 | ||
468 | expected = 0; | |
469 | for (ring = &nlk->rx_ring; ring <= &nlk->tx_ring; ring++) { | |
470 | if (ring->pg_vec == NULL) | |
471 | continue; | |
472 | expected += ring->pg_vec_len * ring->pg_vec_pages * PAGE_SIZE; | |
473 | } | |
474 | ||
475 | if (expected == 0) | |
476 | goto out; | |
477 | ||
478 | size = vma->vm_end - vma->vm_start; | |
479 | if (size != expected) | |
480 | goto out; | |
481 | ||
482 | start = vma->vm_start; | |
483 | for (ring = &nlk->rx_ring; ring <= &nlk->tx_ring; ring++) { | |
484 | if (ring->pg_vec == NULL) | |
485 | continue; | |
486 | ||
487 | for (i = 0; i < ring->pg_vec_len; i++) { | |
488 | struct page *page; | |
489 | void *kaddr = ring->pg_vec[i]; | |
490 | unsigned int pg_num; | |
491 | ||
492 | for (pg_num = 0; pg_num < ring->pg_vec_pages; pg_num++) { | |
493 | page = pgvec_to_page(kaddr); | |
494 | err = vm_insert_page(vma, start, page); | |
495 | if (err < 0) | |
496 | goto out; | |
497 | start += PAGE_SIZE; | |
498 | kaddr += PAGE_SIZE; | |
499 | } | |
500 | } | |
501 | } | |
502 | ||
503 | atomic_inc(&nlk->mapped); | |
504 | vma->vm_ops = &netlink_mmap_ops; | |
505 | err = 0; | |
506 | out: | |
507 | mutex_unlock(&nlk->pg_vec_lock); | |
7cdbac71 | 508 | return err; |
ccdfcc39 | 509 | } |
9652e931 PM |
510 | |
511 | static void netlink_frame_flush_dcache(const struct nl_mmap_hdr *hdr) | |
512 | { | |
513 | #if ARCH_IMPLEMENTS_FLUSH_DCACHE_PAGE == 1 | |
514 | struct page *p_start, *p_end; | |
515 | ||
516 | /* First page is flushed through netlink_{get,set}_status */ | |
517 | p_start = pgvec_to_page(hdr + PAGE_SIZE); | |
1d5085cb | 518 | p_end = pgvec_to_page((void *)hdr + NL_MMAP_HDRLEN + hdr->nm_len - 1); |
9652e931 PM |
519 | while (p_start <= p_end) { |
520 | flush_dcache_page(p_start); | |
521 | p_start++; | |
522 | } | |
523 | #endif | |
524 | } | |
525 | ||
526 | static enum nl_mmap_status netlink_get_status(const struct nl_mmap_hdr *hdr) | |
527 | { | |
528 | smp_rmb(); | |
529 | flush_dcache_page(pgvec_to_page(hdr)); | |
530 | return hdr->nm_status; | |
531 | } | |
532 | ||
533 | static void netlink_set_status(struct nl_mmap_hdr *hdr, | |
534 | enum nl_mmap_status status) | |
535 | { | |
536 | hdr->nm_status = status; | |
537 | flush_dcache_page(pgvec_to_page(hdr)); | |
538 | smp_wmb(); | |
539 | } | |
540 | ||
541 | static struct nl_mmap_hdr * | |
542 | __netlink_lookup_frame(const struct netlink_ring *ring, unsigned int pos) | |
543 | { | |
544 | unsigned int pg_vec_pos, frame_off; | |
545 | ||
546 | pg_vec_pos = pos / ring->frames_per_block; | |
547 | frame_off = pos % ring->frames_per_block; | |
548 | ||
549 | return ring->pg_vec[pg_vec_pos] + (frame_off * ring->frame_size); | |
550 | } | |
551 | ||
552 | static struct nl_mmap_hdr * | |
553 | netlink_lookup_frame(const struct netlink_ring *ring, unsigned int pos, | |
554 | enum nl_mmap_status status) | |
555 | { | |
556 | struct nl_mmap_hdr *hdr; | |
557 | ||
558 | hdr = __netlink_lookup_frame(ring, pos); | |
559 | if (netlink_get_status(hdr) != status) | |
560 | return NULL; | |
561 | ||
562 | return hdr; | |
563 | } | |
564 | ||
565 | static struct nl_mmap_hdr * | |
566 | netlink_current_frame(const struct netlink_ring *ring, | |
567 | enum nl_mmap_status status) | |
568 | { | |
569 | return netlink_lookup_frame(ring, ring->head, status); | |
570 | } | |
571 | ||
572 | static struct nl_mmap_hdr * | |
573 | netlink_previous_frame(const struct netlink_ring *ring, | |
574 | enum nl_mmap_status status) | |
575 | { | |
576 | unsigned int prev; | |
577 | ||
578 | prev = ring->head ? ring->head - 1 : ring->frame_max; | |
579 | return netlink_lookup_frame(ring, prev, status); | |
580 | } | |
581 | ||
582 | static void netlink_increment_head(struct netlink_ring *ring) | |
583 | { | |
584 | ring->head = ring->head != ring->frame_max ? ring->head + 1 : 0; | |
585 | } | |
586 | ||
587 | static void netlink_forward_ring(struct netlink_ring *ring) | |
588 | { | |
589 | unsigned int head = ring->head, pos = head; | |
590 | const struct nl_mmap_hdr *hdr; | |
591 | ||
592 | do { | |
593 | hdr = __netlink_lookup_frame(ring, pos); | |
594 | if (hdr->nm_status == NL_MMAP_STATUS_UNUSED) | |
595 | break; | |
596 | if (hdr->nm_status != NL_MMAP_STATUS_SKIP) | |
597 | break; | |
598 | netlink_increment_head(ring); | |
599 | } while (ring->head != head); | |
600 | } | |
601 | ||
cd1df525 PM |
602 | static bool netlink_dump_space(struct netlink_sock *nlk) |
603 | { | |
604 | struct netlink_ring *ring = &nlk->rx_ring; | |
605 | struct nl_mmap_hdr *hdr; | |
606 | unsigned int n; | |
607 | ||
608 | hdr = netlink_current_frame(ring, NL_MMAP_STATUS_UNUSED); | |
609 | if (hdr == NULL) | |
610 | return false; | |
611 | ||
612 | n = ring->head + ring->frame_max / 2; | |
613 | if (n > ring->frame_max) | |
614 | n -= ring->frame_max; | |
615 | ||
616 | hdr = __netlink_lookup_frame(ring, n); | |
617 | ||
618 | return hdr->nm_status == NL_MMAP_STATUS_UNUSED; | |
619 | } | |
620 | ||
9652e931 PM |
621 | static unsigned int netlink_poll(struct file *file, struct socket *sock, |
622 | poll_table *wait) | |
623 | { | |
624 | struct sock *sk = sock->sk; | |
625 | struct netlink_sock *nlk = nlk_sk(sk); | |
626 | unsigned int mask; | |
cd1df525 | 627 | int err; |
9652e931 | 628 | |
cd1df525 PM |
629 | if (nlk->rx_ring.pg_vec != NULL) { |
630 | /* Memory mapped sockets don't call recvmsg(), so flow control | |
631 | * for dumps is performed here. A dump is allowed to continue | |
632 | * if at least half the ring is unused. | |
633 | */ | |
16b304f3 | 634 | while (nlk->cb_running && netlink_dump_space(nlk)) { |
cd1df525 PM |
635 | err = netlink_dump(sk); |
636 | if (err < 0) { | |
ac30ef83 | 637 | sk->sk_err = -err; |
cd1df525 PM |
638 | sk->sk_error_report(sk); |
639 | break; | |
640 | } | |
641 | } | |
642 | netlink_rcv_wake(sk); | |
643 | } | |
5fd96123 | 644 | |
9652e931 PM |
645 | mask = datagram_poll(file, sock, wait); |
646 | ||
647 | spin_lock_bh(&sk->sk_receive_queue.lock); | |
648 | if (nlk->rx_ring.pg_vec) { | |
649 | netlink_forward_ring(&nlk->rx_ring); | |
650 | if (!netlink_previous_frame(&nlk->rx_ring, NL_MMAP_STATUS_UNUSED)) | |
651 | mask |= POLLIN | POLLRDNORM; | |
652 | } | |
653 | spin_unlock_bh(&sk->sk_receive_queue.lock); | |
654 | ||
655 | spin_lock_bh(&sk->sk_write_queue.lock); | |
656 | if (nlk->tx_ring.pg_vec) { | |
657 | if (netlink_current_frame(&nlk->tx_ring, NL_MMAP_STATUS_UNUSED)) | |
658 | mask |= POLLOUT | POLLWRNORM; | |
659 | } | |
660 | spin_unlock_bh(&sk->sk_write_queue.lock); | |
661 | ||
662 | return mask; | |
663 | } | |
664 | ||
665 | static struct nl_mmap_hdr *netlink_mmap_hdr(struct sk_buff *skb) | |
666 | { | |
667 | return (struct nl_mmap_hdr *)(skb->head - NL_MMAP_HDRLEN); | |
668 | } | |
669 | ||
670 | static void netlink_ring_setup_skb(struct sk_buff *skb, struct sock *sk, | |
671 | struct netlink_ring *ring, | |
672 | struct nl_mmap_hdr *hdr) | |
673 | { | |
674 | unsigned int size; | |
675 | void *data; | |
676 | ||
677 | size = ring->frame_size - NL_MMAP_HDRLEN; | |
678 | data = (void *)hdr + NL_MMAP_HDRLEN; | |
679 | ||
680 | skb->head = data; | |
681 | skb->data = data; | |
682 | skb_reset_tail_pointer(skb); | |
683 | skb->end = skb->tail + size; | |
684 | skb->len = 0; | |
685 | ||
686 | skb->destructor = netlink_skb_destructor; | |
687 | NETLINK_CB(skb).flags |= NETLINK_SKB_MMAPED; | |
688 | NETLINK_CB(skb).sk = sk; | |
689 | } | |
5fd96123 PM |
690 | |
691 | static int netlink_mmap_sendmsg(struct sock *sk, struct msghdr *msg, | |
692 | u32 dst_portid, u32 dst_group, | |
693 | struct sock_iocb *siocb) | |
694 | { | |
695 | struct netlink_sock *nlk = nlk_sk(sk); | |
696 | struct netlink_ring *ring; | |
697 | struct nl_mmap_hdr *hdr; | |
698 | struct sk_buff *skb; | |
699 | unsigned int maxlen; | |
700 | bool excl = true; | |
701 | int err = 0, len = 0; | |
702 | ||
703 | /* Netlink messages are validated by the receiver before processing. | |
704 | * In order to avoid userspace changing the contents of the message | |
705 | * after validation, the socket and the ring may only be used by a | |
706 | * single process, otherwise we fall back to copying. | |
707 | */ | |
708 | if (atomic_long_read(&sk->sk_socket->file->f_count) > 2 || | |
709 | atomic_read(&nlk->mapped) > 1) | |
710 | excl = false; | |
711 | ||
712 | mutex_lock(&nlk->pg_vec_lock); | |
713 | ||
714 | ring = &nlk->tx_ring; | |
715 | maxlen = ring->frame_size - NL_MMAP_HDRLEN; | |
716 | ||
717 | do { | |
718 | hdr = netlink_current_frame(ring, NL_MMAP_STATUS_VALID); | |
719 | if (hdr == NULL) { | |
720 | if (!(msg->msg_flags & MSG_DONTWAIT) && | |
721 | atomic_read(&nlk->tx_ring.pending)) | |
722 | schedule(); | |
723 | continue; | |
724 | } | |
725 | if (hdr->nm_len > maxlen) { | |
726 | err = -EINVAL; | |
727 | goto out; | |
728 | } | |
729 | ||
730 | netlink_frame_flush_dcache(hdr); | |
731 | ||
732 | if (likely(dst_portid == 0 && dst_group == 0 && excl)) { | |
733 | skb = alloc_skb_head(GFP_KERNEL); | |
734 | if (skb == NULL) { | |
735 | err = -ENOBUFS; | |
736 | goto out; | |
737 | } | |
738 | sock_hold(sk); | |
739 | netlink_ring_setup_skb(skb, sk, ring, hdr); | |
740 | NETLINK_CB(skb).flags |= NETLINK_SKB_TX; | |
741 | __skb_put(skb, hdr->nm_len); | |
742 | netlink_set_status(hdr, NL_MMAP_STATUS_RESERVED); | |
743 | atomic_inc(&ring->pending); | |
744 | } else { | |
745 | skb = alloc_skb(hdr->nm_len, GFP_KERNEL); | |
746 | if (skb == NULL) { | |
747 | err = -ENOBUFS; | |
748 | goto out; | |
749 | } | |
750 | __skb_put(skb, hdr->nm_len); | |
751 | memcpy(skb->data, (void *)hdr + NL_MMAP_HDRLEN, hdr->nm_len); | |
752 | netlink_set_status(hdr, NL_MMAP_STATUS_UNUSED); | |
753 | } | |
754 | ||
755 | netlink_increment_head(ring); | |
756 | ||
757 | NETLINK_CB(skb).portid = nlk->portid; | |
758 | NETLINK_CB(skb).dst_group = dst_group; | |
759 | NETLINK_CB(skb).creds = siocb->scm->creds; | |
760 | ||
761 | err = security_netlink_send(sk, skb); | |
762 | if (err) { | |
763 | kfree_skb(skb); | |
764 | goto out; | |
765 | } | |
766 | ||
767 | if (unlikely(dst_group)) { | |
768 | atomic_inc(&skb->users); | |
769 | netlink_broadcast(sk, skb, dst_portid, dst_group, | |
770 | GFP_KERNEL); | |
771 | } | |
772 | err = netlink_unicast(sk, skb, dst_portid, | |
773 | msg->msg_flags & MSG_DONTWAIT); | |
774 | if (err < 0) | |
775 | goto out; | |
776 | len += err; | |
777 | ||
778 | } while (hdr != NULL || | |
779 | (!(msg->msg_flags & MSG_DONTWAIT) && | |
780 | atomic_read(&nlk->tx_ring.pending))); | |
781 | ||
782 | if (len > 0) | |
783 | err = len; | |
784 | out: | |
785 | mutex_unlock(&nlk->pg_vec_lock); | |
786 | return err; | |
787 | } | |
f9c22888 PM |
788 | |
789 | static void netlink_queue_mmaped_skb(struct sock *sk, struct sk_buff *skb) | |
790 | { | |
791 | struct nl_mmap_hdr *hdr; | |
792 | ||
793 | hdr = netlink_mmap_hdr(skb); | |
794 | hdr->nm_len = skb->len; | |
795 | hdr->nm_group = NETLINK_CB(skb).dst_group; | |
796 | hdr->nm_pid = NETLINK_CB(skb).creds.pid; | |
1bf9310a ND |
797 | hdr->nm_uid = from_kuid(sk_user_ns(sk), NETLINK_CB(skb).creds.uid); |
798 | hdr->nm_gid = from_kgid(sk_user_ns(sk), NETLINK_CB(skb).creds.gid); | |
f9c22888 PM |
799 | netlink_frame_flush_dcache(hdr); |
800 | netlink_set_status(hdr, NL_MMAP_STATUS_VALID); | |
801 | ||
802 | NETLINK_CB(skb).flags |= NETLINK_SKB_DELIVERED; | |
803 | kfree_skb(skb); | |
804 | } | |
805 | ||
806 | static void netlink_ring_set_copied(struct sock *sk, struct sk_buff *skb) | |
807 | { | |
808 | struct netlink_sock *nlk = nlk_sk(sk); | |
809 | struct netlink_ring *ring = &nlk->rx_ring; | |
810 | struct nl_mmap_hdr *hdr; | |
811 | ||
812 | spin_lock_bh(&sk->sk_receive_queue.lock); | |
813 | hdr = netlink_current_frame(ring, NL_MMAP_STATUS_UNUSED); | |
814 | if (hdr == NULL) { | |
815 | spin_unlock_bh(&sk->sk_receive_queue.lock); | |
816 | kfree_skb(skb); | |
cd1df525 | 817 | netlink_overrun(sk); |
f9c22888 PM |
818 | return; |
819 | } | |
820 | netlink_increment_head(ring); | |
821 | __skb_queue_tail(&sk->sk_receive_queue, skb); | |
822 | spin_unlock_bh(&sk->sk_receive_queue.lock); | |
823 | ||
824 | hdr->nm_len = skb->len; | |
825 | hdr->nm_group = NETLINK_CB(skb).dst_group; | |
826 | hdr->nm_pid = NETLINK_CB(skb).creds.pid; | |
1bf9310a ND |
827 | hdr->nm_uid = from_kuid(sk_user_ns(sk), NETLINK_CB(skb).creds.uid); |
828 | hdr->nm_gid = from_kgid(sk_user_ns(sk), NETLINK_CB(skb).creds.gid); | |
f9c22888 PM |
829 | netlink_set_status(hdr, NL_MMAP_STATUS_COPY); |
830 | } | |
831 | ||
ccdfcc39 | 832 | #else /* CONFIG_NETLINK_MMAP */ |
9652e931 | 833 | #define netlink_skb_is_mmaped(skb) false |
f9c22888 | 834 | #define netlink_rx_is_mmaped(sk) false |
5fd96123 | 835 | #define netlink_tx_is_mmaped(sk) false |
ccdfcc39 | 836 | #define netlink_mmap sock_no_mmap |
9652e931 | 837 | #define netlink_poll datagram_poll |
5fd96123 | 838 | #define netlink_mmap_sendmsg(sk, msg, dst_portid, dst_group, siocb) 0 |
ccdfcc39 PM |
839 | #endif /* CONFIG_NETLINK_MMAP */ |
840 | ||
cf0a018a PM |
841 | static void netlink_skb_destructor(struct sk_buff *skb) |
842 | { | |
9652e931 PM |
843 | #ifdef CONFIG_NETLINK_MMAP |
844 | struct nl_mmap_hdr *hdr; | |
845 | struct netlink_ring *ring; | |
846 | struct sock *sk; | |
847 | ||
848 | /* If a packet from the kernel to userspace was freed because of an | |
849 | * error without being delivered to userspace, the kernel must reset | |
850 | * the status. In the direction userspace to kernel, the status is | |
851 | * always reset here after the packet was processed and freed. | |
852 | */ | |
853 | if (netlink_skb_is_mmaped(skb)) { | |
854 | hdr = netlink_mmap_hdr(skb); | |
855 | sk = NETLINK_CB(skb).sk; | |
856 | ||
5fd96123 PM |
857 | if (NETLINK_CB(skb).flags & NETLINK_SKB_TX) { |
858 | netlink_set_status(hdr, NL_MMAP_STATUS_UNUSED); | |
859 | ring = &nlk_sk(sk)->tx_ring; | |
860 | } else { | |
861 | if (!(NETLINK_CB(skb).flags & NETLINK_SKB_DELIVERED)) { | |
862 | hdr->nm_len = 0; | |
863 | netlink_set_status(hdr, NL_MMAP_STATUS_VALID); | |
864 | } | |
865 | ring = &nlk_sk(sk)->rx_ring; | |
9652e931 | 866 | } |
9652e931 PM |
867 | |
868 | WARN_ON(atomic_read(&ring->pending) == 0); | |
869 | atomic_dec(&ring->pending); | |
870 | sock_put(sk); | |
871 | ||
5e71d9d7 | 872 | skb->head = NULL; |
9652e931 PM |
873 | } |
874 | #endif | |
c05cdb1b | 875 | if (is_vmalloc_addr(skb->head)) { |
3a36515f PN |
876 | if (!skb->cloned || |
877 | !atomic_dec_return(&(skb_shinfo(skb)->dataref))) | |
878 | vfree(skb->head); | |
879 | ||
c05cdb1b PNA |
880 | skb->head = NULL; |
881 | } | |
9652e931 PM |
882 | if (skb->sk != NULL) |
883 | sock_rfree(skb); | |
cf0a018a PM |
884 | } |
885 | ||
886 | static void netlink_skb_set_owner_r(struct sk_buff *skb, struct sock *sk) | |
887 | { | |
888 | WARN_ON(skb->sk != NULL); | |
889 | skb->sk = sk; | |
890 | skb->destructor = netlink_skb_destructor; | |
891 | atomic_add(skb->truesize, &sk->sk_rmem_alloc); | |
892 | sk_mem_charge(sk, skb->truesize); | |
893 | } | |
894 | ||
1da177e4 LT |
895 | static void netlink_sock_destruct(struct sock *sk) |
896 | { | |
3f660d66 HX |
897 | struct netlink_sock *nlk = nlk_sk(sk); |
898 | ||
16b304f3 PS |
899 | if (nlk->cb_running) { |
900 | if (nlk->cb.done) | |
901 | nlk->cb.done(&nlk->cb); | |
6dc878a8 | 902 | |
16b304f3 PS |
903 | module_put(nlk->cb.module); |
904 | kfree_skb(nlk->cb.skb); | |
3f660d66 HX |
905 | } |
906 | ||
1da177e4 | 907 | skb_queue_purge(&sk->sk_receive_queue); |
ccdfcc39 PM |
908 | #ifdef CONFIG_NETLINK_MMAP |
909 | if (1) { | |
910 | struct nl_mmap_req req; | |
911 | ||
912 | memset(&req, 0, sizeof(req)); | |
913 | if (nlk->rx_ring.pg_vec) | |
914 | netlink_set_ring(sk, &req, true, false); | |
915 | memset(&req, 0, sizeof(req)); | |
916 | if (nlk->tx_ring.pg_vec) | |
917 | netlink_set_ring(sk, &req, true, true); | |
918 | } | |
919 | #endif /* CONFIG_NETLINK_MMAP */ | |
1da177e4 LT |
920 | |
921 | if (!sock_flag(sk, SOCK_DEAD)) { | |
6ac552fd | 922 | printk(KERN_ERR "Freeing alive netlink socket %p\n", sk); |
1da177e4 LT |
923 | return; |
924 | } | |
547b792c IJ |
925 | |
926 | WARN_ON(atomic_read(&sk->sk_rmem_alloc)); | |
927 | WARN_ON(atomic_read(&sk->sk_wmem_alloc)); | |
928 | WARN_ON(nlk_sk(sk)->groups); | |
1da177e4 LT |
929 | } |
930 | ||
6ac552fd PM |
931 | /* This lock without WQ_FLAG_EXCLUSIVE is good on UP and it is _very_ bad on |
932 | * SMP. Look, when several writers sleep and reader wakes them up, all but one | |
1da177e4 LT |
933 | * immediately hit write lock and grab all the cpus. Exclusive sleep solves |
934 | * this, _but_ remember, it adds useless work on UP machines. | |
935 | */ | |
936 | ||
d136f1bd | 937 | void netlink_table_grab(void) |
9a429c49 | 938 | __acquires(nl_table_lock) |
1da177e4 | 939 | { |
d136f1bd JB |
940 | might_sleep(); |
941 | ||
6abd219c | 942 | write_lock_irq(&nl_table_lock); |
1da177e4 LT |
943 | |
944 | if (atomic_read(&nl_table_users)) { | |
945 | DECLARE_WAITQUEUE(wait, current); | |
946 | ||
947 | add_wait_queue_exclusive(&nl_table_wait, &wait); | |
6ac552fd | 948 | for (;;) { |
1da177e4 LT |
949 | set_current_state(TASK_UNINTERRUPTIBLE); |
950 | if (atomic_read(&nl_table_users) == 0) | |
951 | break; | |
6abd219c | 952 | write_unlock_irq(&nl_table_lock); |
1da177e4 | 953 | schedule(); |
6abd219c | 954 | write_lock_irq(&nl_table_lock); |
1da177e4 LT |
955 | } |
956 | ||
957 | __set_current_state(TASK_RUNNING); | |
958 | remove_wait_queue(&nl_table_wait, &wait); | |
959 | } | |
960 | } | |
961 | ||
d136f1bd | 962 | void netlink_table_ungrab(void) |
9a429c49 | 963 | __releases(nl_table_lock) |
1da177e4 | 964 | { |
6abd219c | 965 | write_unlock_irq(&nl_table_lock); |
1da177e4 LT |
966 | wake_up(&nl_table_wait); |
967 | } | |
968 | ||
6ac552fd | 969 | static inline void |
1da177e4 LT |
970 | netlink_lock_table(void) |
971 | { | |
972 | /* read_lock() synchronizes us to netlink_table_grab */ | |
973 | ||
974 | read_lock(&nl_table_lock); | |
975 | atomic_inc(&nl_table_users); | |
976 | read_unlock(&nl_table_lock); | |
977 | } | |
978 | ||
6ac552fd | 979 | static inline void |
1da177e4 LT |
980 | netlink_unlock_table(void) |
981 | { | |
982 | if (atomic_dec_and_test(&nl_table_users)) | |
983 | wake_up(&nl_table_wait); | |
984 | } | |
985 | ||
da12c90e G |
986 | static bool netlink_compare(struct net *net, struct sock *sk) |
987 | { | |
988 | return net_eq(sock_net(sk), net); | |
989 | } | |
990 | ||
15e47304 | 991 | static struct sock *netlink_lookup(struct net *net, int protocol, u32 portid) |
1da177e4 | 992 | { |
da12c90e G |
993 | struct netlink_table *table = &nl_table[protocol]; |
994 | struct nl_portid_hash *hash = &table->hash; | |
1da177e4 LT |
995 | struct hlist_head *head; |
996 | struct sock *sk; | |
1da177e4 LT |
997 | |
998 | read_lock(&nl_table_lock); | |
15e47304 | 999 | head = nl_portid_hashfn(hash, portid); |
b67bfe0d | 1000 | sk_for_each(sk, head) { |
da12c90e G |
1001 | if (table->compare(net, sk) && |
1002 | (nlk_sk(sk)->portid == portid)) { | |
1da177e4 LT |
1003 | sock_hold(sk); |
1004 | goto found; | |
1005 | } | |
1006 | } | |
1007 | sk = NULL; | |
1008 | found: | |
1009 | read_unlock(&nl_table_lock); | |
1010 | return sk; | |
1011 | } | |
1012 | ||
15e47304 | 1013 | static struct hlist_head *nl_portid_hash_zalloc(size_t size) |
1da177e4 LT |
1014 | { |
1015 | if (size <= PAGE_SIZE) | |
ea72912c | 1016 | return kzalloc(size, GFP_ATOMIC); |
1da177e4 LT |
1017 | else |
1018 | return (struct hlist_head *) | |
ea72912c ED |
1019 | __get_free_pages(GFP_ATOMIC | __GFP_ZERO, |
1020 | get_order(size)); | |
1da177e4 LT |
1021 | } |
1022 | ||
15e47304 | 1023 | static void nl_portid_hash_free(struct hlist_head *table, size_t size) |
1da177e4 LT |
1024 | { |
1025 | if (size <= PAGE_SIZE) | |
1026 | kfree(table); | |
1027 | else | |
1028 | free_pages((unsigned long)table, get_order(size)); | |
1029 | } | |
1030 | ||
15e47304 | 1031 | static int nl_portid_hash_rehash(struct nl_portid_hash *hash, int grow) |
1da177e4 LT |
1032 | { |
1033 | unsigned int omask, mask, shift; | |
1034 | size_t osize, size; | |
1035 | struct hlist_head *otable, *table; | |
1036 | int i; | |
1037 | ||
1038 | omask = mask = hash->mask; | |
1039 | osize = size = (mask + 1) * sizeof(*table); | |
1040 | shift = hash->shift; | |
1041 | ||
1042 | if (grow) { | |
1043 | if (++shift > hash->max_shift) | |
1044 | return 0; | |
1045 | mask = mask * 2 + 1; | |
1046 | size *= 2; | |
1047 | } | |
1048 | ||
15e47304 | 1049 | table = nl_portid_hash_zalloc(size); |
1da177e4 LT |
1050 | if (!table) |
1051 | return 0; | |
1052 | ||
1da177e4 LT |
1053 | otable = hash->table; |
1054 | hash->table = table; | |
1055 | hash->mask = mask; | |
1056 | hash->shift = shift; | |
1057 | get_random_bytes(&hash->rnd, sizeof(hash->rnd)); | |
1058 | ||
1059 | for (i = 0; i <= omask; i++) { | |
1060 | struct sock *sk; | |
b67bfe0d | 1061 | struct hlist_node *tmp; |
1da177e4 | 1062 | |
b67bfe0d | 1063 | sk_for_each_safe(sk, tmp, &otable[i]) |
15e47304 | 1064 | __sk_add_node(sk, nl_portid_hashfn(hash, nlk_sk(sk)->portid)); |
1da177e4 LT |
1065 | } |
1066 | ||
15e47304 | 1067 | nl_portid_hash_free(otable, osize); |
1da177e4 LT |
1068 | hash->rehash_time = jiffies + 10 * 60 * HZ; |
1069 | return 1; | |
1070 | } | |
1071 | ||
15e47304 | 1072 | static inline int nl_portid_hash_dilute(struct nl_portid_hash *hash, int len) |
1da177e4 LT |
1073 | { |
1074 | int avg = hash->entries >> hash->shift; | |
1075 | ||
15e47304 | 1076 | if (unlikely(avg > 1) && nl_portid_hash_rehash(hash, 1)) |
1da177e4 LT |
1077 | return 1; |
1078 | ||
1079 | if (unlikely(len > avg) && time_after(jiffies, hash->rehash_time)) { | |
15e47304 | 1080 | nl_portid_hash_rehash(hash, 0); |
1da177e4 LT |
1081 | return 1; |
1082 | } | |
1083 | ||
1084 | return 0; | |
1085 | } | |
1086 | ||
90ddc4f0 | 1087 | static const struct proto_ops netlink_ops; |
1da177e4 | 1088 | |
4277a083 PM |
1089 | static void |
1090 | netlink_update_listeners(struct sock *sk) | |
1091 | { | |
1092 | struct netlink_table *tbl = &nl_table[sk->sk_protocol]; | |
4277a083 PM |
1093 | unsigned long mask; |
1094 | unsigned int i; | |
6d772ac5 ED |
1095 | struct listeners *listeners; |
1096 | ||
1097 | listeners = nl_deref_protected(tbl->listeners); | |
1098 | if (!listeners) | |
1099 | return; | |
4277a083 | 1100 | |
b4ff4f04 | 1101 | for (i = 0; i < NLGRPLONGS(tbl->groups); i++) { |
4277a083 | 1102 | mask = 0; |
b67bfe0d | 1103 | sk_for_each_bound(sk, &tbl->mc_list) { |
b4ff4f04 JB |
1104 | if (i < NLGRPLONGS(nlk_sk(sk)->ngroups)) |
1105 | mask |= nlk_sk(sk)->groups[i]; | |
1106 | } | |
6d772ac5 | 1107 | listeners->masks[i] = mask; |
4277a083 PM |
1108 | } |
1109 | /* this function is only called with the netlink table "grabbed", which | |
1110 | * makes sure updates are visible before bind or setsockopt return. */ | |
1111 | } | |
1112 | ||
15e47304 | 1113 | static int netlink_insert(struct sock *sk, struct net *net, u32 portid) |
1da177e4 | 1114 | { |
da12c90e G |
1115 | struct netlink_table *table = &nl_table[sk->sk_protocol]; |
1116 | struct nl_portid_hash *hash = &table->hash; | |
1da177e4 LT |
1117 | struct hlist_head *head; |
1118 | int err = -EADDRINUSE; | |
1119 | struct sock *osk; | |
1da177e4 LT |
1120 | int len; |
1121 | ||
1122 | netlink_table_grab(); | |
15e47304 | 1123 | head = nl_portid_hashfn(hash, portid); |
1da177e4 | 1124 | len = 0; |
b67bfe0d | 1125 | sk_for_each(osk, head) { |
da12c90e G |
1126 | if (table->compare(net, osk) && |
1127 | (nlk_sk(osk)->portid == portid)) | |
1da177e4 LT |
1128 | break; |
1129 | len++; | |
1130 | } | |
b67bfe0d | 1131 | if (osk) |
1da177e4 LT |
1132 | goto err; |
1133 | ||
1134 | err = -EBUSY; | |
15e47304 | 1135 | if (nlk_sk(sk)->portid) |
1da177e4 LT |
1136 | goto err; |
1137 | ||
1138 | err = -ENOMEM; | |
1139 | if (BITS_PER_LONG > 32 && unlikely(hash->entries >= UINT_MAX)) | |
1140 | goto err; | |
1141 | ||
15e47304 EB |
1142 | if (len && nl_portid_hash_dilute(hash, len)) |
1143 | head = nl_portid_hashfn(hash, portid); | |
1da177e4 | 1144 | hash->entries++; |
15e47304 | 1145 | nlk_sk(sk)->portid = portid; |
1da177e4 LT |
1146 | sk_add_node(sk, head); |
1147 | err = 0; | |
1148 | ||
1149 | err: | |
1150 | netlink_table_ungrab(); | |
1151 | return err; | |
1152 | } | |
1153 | ||
1154 | static void netlink_remove(struct sock *sk) | |
1155 | { | |
1156 | netlink_table_grab(); | |
d470e3b4 DM |
1157 | if (sk_del_node_init(sk)) |
1158 | nl_table[sk->sk_protocol].hash.entries--; | |
f7fa9b10 | 1159 | if (nlk_sk(sk)->subscriptions) |
1da177e4 LT |
1160 | __sk_del_bind_node(sk); |
1161 | netlink_table_ungrab(); | |
1162 | } | |
1163 | ||
1164 | static struct proto netlink_proto = { | |
1165 | .name = "NETLINK", | |
1166 | .owner = THIS_MODULE, | |
1167 | .obj_size = sizeof(struct netlink_sock), | |
1168 | }; | |
1169 | ||
1b8d7ae4 EB |
1170 | static int __netlink_create(struct net *net, struct socket *sock, |
1171 | struct mutex *cb_mutex, int protocol) | |
1da177e4 LT |
1172 | { |
1173 | struct sock *sk; | |
1174 | struct netlink_sock *nlk; | |
ab33a171 PM |
1175 | |
1176 | sock->ops = &netlink_ops; | |
1177 | ||
6257ff21 | 1178 | sk = sk_alloc(net, PF_NETLINK, GFP_KERNEL, &netlink_proto); |
ab33a171 PM |
1179 | if (!sk) |
1180 | return -ENOMEM; | |
1181 | ||
1182 | sock_init_data(sock, sk); | |
1183 | ||
1184 | nlk = nlk_sk(sk); | |
658cb354 | 1185 | if (cb_mutex) { |
ffa4d721 | 1186 | nlk->cb_mutex = cb_mutex; |
658cb354 | 1187 | } else { |
ffa4d721 PM |
1188 | nlk->cb_mutex = &nlk->cb_def_mutex; |
1189 | mutex_init(nlk->cb_mutex); | |
1190 | } | |
ab33a171 | 1191 | init_waitqueue_head(&nlk->wait); |
ccdfcc39 PM |
1192 | #ifdef CONFIG_NETLINK_MMAP |
1193 | mutex_init(&nlk->pg_vec_lock); | |
1194 | #endif | |
ab33a171 PM |
1195 | |
1196 | sk->sk_destruct = netlink_sock_destruct; | |
1197 | sk->sk_protocol = protocol; | |
1198 | return 0; | |
1199 | } | |
1200 | ||
3f378b68 EP |
1201 | static int netlink_create(struct net *net, struct socket *sock, int protocol, |
1202 | int kern) | |
ab33a171 PM |
1203 | { |
1204 | struct module *module = NULL; | |
af65bdfc | 1205 | struct mutex *cb_mutex; |
f7fa9b10 | 1206 | struct netlink_sock *nlk; |
4f520900 RGB |
1207 | int (*bind)(int group); |
1208 | void (*unbind)(int group); | |
ab33a171 | 1209 | int err = 0; |
1da177e4 LT |
1210 | |
1211 | sock->state = SS_UNCONNECTED; | |
1212 | ||
1213 | if (sock->type != SOCK_RAW && sock->type != SOCK_DGRAM) | |
1214 | return -ESOCKTNOSUPPORT; | |
1215 | ||
6ac552fd | 1216 | if (protocol < 0 || protocol >= MAX_LINKS) |
1da177e4 LT |
1217 | return -EPROTONOSUPPORT; |
1218 | ||
77247bbb | 1219 | netlink_lock_table(); |
95a5afca | 1220 | #ifdef CONFIG_MODULES |
ab33a171 | 1221 | if (!nl_table[protocol].registered) { |
77247bbb | 1222 | netlink_unlock_table(); |
4fdb3bb7 | 1223 | request_module("net-pf-%d-proto-%d", PF_NETLINK, protocol); |
77247bbb | 1224 | netlink_lock_table(); |
4fdb3bb7 | 1225 | } |
ab33a171 PM |
1226 | #endif |
1227 | if (nl_table[protocol].registered && | |
1228 | try_module_get(nl_table[protocol].module)) | |
1229 | module = nl_table[protocol].module; | |
974c37e9 AD |
1230 | else |
1231 | err = -EPROTONOSUPPORT; | |
af65bdfc | 1232 | cb_mutex = nl_table[protocol].cb_mutex; |
03292745 | 1233 | bind = nl_table[protocol].bind; |
4f520900 | 1234 | unbind = nl_table[protocol].unbind; |
77247bbb | 1235 | netlink_unlock_table(); |
4fdb3bb7 | 1236 | |
974c37e9 AD |
1237 | if (err < 0) |
1238 | goto out; | |
1239 | ||
6ac552fd PM |
1240 | err = __netlink_create(net, sock, cb_mutex, protocol); |
1241 | if (err < 0) | |
f7fa9b10 PM |
1242 | goto out_module; |
1243 | ||
6f756a8c | 1244 | local_bh_disable(); |
c1fd3b94 | 1245 | sock_prot_inuse_add(net, &netlink_proto, 1); |
6f756a8c DM |
1246 | local_bh_enable(); |
1247 | ||
f7fa9b10 | 1248 | nlk = nlk_sk(sock->sk); |
f7fa9b10 | 1249 | nlk->module = module; |
03292745 | 1250 | nlk->netlink_bind = bind; |
4f520900 | 1251 | nlk->netlink_unbind = unbind; |
ab33a171 PM |
1252 | out: |
1253 | return err; | |
1da177e4 | 1254 | |
ab33a171 PM |
1255 | out_module: |
1256 | module_put(module); | |
1257 | goto out; | |
1da177e4 LT |
1258 | } |
1259 | ||
1260 | static int netlink_release(struct socket *sock) | |
1261 | { | |
1262 | struct sock *sk = sock->sk; | |
1263 | struct netlink_sock *nlk; | |
1264 | ||
1265 | if (!sk) | |
1266 | return 0; | |
1267 | ||
1268 | netlink_remove(sk); | |
ac57b3a9 | 1269 | sock_orphan(sk); |
1da177e4 LT |
1270 | nlk = nlk_sk(sk); |
1271 | ||
3f660d66 HX |
1272 | /* |
1273 | * OK. Socket is unlinked, any packets that arrive now | |
1274 | * will be purged. | |
1275 | */ | |
1da177e4 | 1276 | |
1da177e4 LT |
1277 | sock->sk = NULL; |
1278 | wake_up_interruptible_all(&nlk->wait); | |
1279 | ||
1280 | skb_queue_purge(&sk->sk_write_queue); | |
1281 | ||
15e47304 | 1282 | if (nlk->portid) { |
1da177e4 | 1283 | struct netlink_notify n = { |
3b1e0a65 | 1284 | .net = sock_net(sk), |
1da177e4 | 1285 | .protocol = sk->sk_protocol, |
15e47304 | 1286 | .portid = nlk->portid, |
1da177e4 | 1287 | }; |
e041c683 AS |
1288 | atomic_notifier_call_chain(&netlink_chain, |
1289 | NETLINK_URELEASE, &n); | |
746fac4d | 1290 | } |
4fdb3bb7 | 1291 | |
5e7c001c | 1292 | module_put(nlk->module); |
4fdb3bb7 | 1293 | |
4277a083 | 1294 | netlink_table_grab(); |
aed81560 | 1295 | if (netlink_is_kernel(sk)) { |
869e58f8 DL |
1296 | BUG_ON(nl_table[sk->sk_protocol].registered == 0); |
1297 | if (--nl_table[sk->sk_protocol].registered == 0) { | |
6d772ac5 ED |
1298 | struct listeners *old; |
1299 | ||
1300 | old = nl_deref_protected(nl_table[sk->sk_protocol].listeners); | |
1301 | RCU_INIT_POINTER(nl_table[sk->sk_protocol].listeners, NULL); | |
1302 | kfree_rcu(old, rcu); | |
869e58f8 | 1303 | nl_table[sk->sk_protocol].module = NULL; |
9785e10a | 1304 | nl_table[sk->sk_protocol].bind = NULL; |
4f520900 | 1305 | nl_table[sk->sk_protocol].unbind = NULL; |
9785e10a | 1306 | nl_table[sk->sk_protocol].flags = 0; |
869e58f8 DL |
1307 | nl_table[sk->sk_protocol].registered = 0; |
1308 | } | |
658cb354 | 1309 | } else if (nlk->subscriptions) { |
4277a083 | 1310 | netlink_update_listeners(sk); |
658cb354 | 1311 | } |
4277a083 | 1312 | netlink_table_ungrab(); |
77247bbb | 1313 | |
f7fa9b10 PM |
1314 | kfree(nlk->groups); |
1315 | nlk->groups = NULL; | |
1316 | ||
3755810c | 1317 | local_bh_disable(); |
c1fd3b94 | 1318 | sock_prot_inuse_add(sock_net(sk), &netlink_proto, -1); |
3755810c | 1319 | local_bh_enable(); |
1da177e4 LT |
1320 | sock_put(sk); |
1321 | return 0; | |
1322 | } | |
1323 | ||
1324 | static int netlink_autobind(struct socket *sock) | |
1325 | { | |
1326 | struct sock *sk = sock->sk; | |
3b1e0a65 | 1327 | struct net *net = sock_net(sk); |
da12c90e G |
1328 | struct netlink_table *table = &nl_table[sk->sk_protocol]; |
1329 | struct nl_portid_hash *hash = &table->hash; | |
1da177e4 LT |
1330 | struct hlist_head *head; |
1331 | struct sock *osk; | |
15e47304 | 1332 | s32 portid = task_tgid_vnr(current); |
1da177e4 LT |
1333 | int err; |
1334 | static s32 rover = -4097; | |
1335 | ||
1336 | retry: | |
1337 | cond_resched(); | |
1338 | netlink_table_grab(); | |
15e47304 | 1339 | head = nl_portid_hashfn(hash, portid); |
b67bfe0d | 1340 | sk_for_each(osk, head) { |
da12c90e | 1341 | if (!table->compare(net, osk)) |
b4b51029 | 1342 | continue; |
15e47304 EB |
1343 | if (nlk_sk(osk)->portid == portid) { |
1344 | /* Bind collision, search negative portid values. */ | |
1345 | portid = rover--; | |
1da177e4 LT |
1346 | if (rover > -4097) |
1347 | rover = -4097; | |
1348 | netlink_table_ungrab(); | |
1349 | goto retry; | |
1350 | } | |
1351 | } | |
1352 | netlink_table_ungrab(); | |
1353 | ||
15e47304 | 1354 | err = netlink_insert(sk, net, portid); |
1da177e4 LT |
1355 | if (err == -EADDRINUSE) |
1356 | goto retry; | |
d470e3b4 DM |
1357 | |
1358 | /* If 2 threads race to autobind, that is fine. */ | |
1359 | if (err == -EBUSY) | |
1360 | err = 0; | |
1361 | ||
1362 | return err; | |
1da177e4 LT |
1363 | } |
1364 | ||
aa4cf945 EB |
1365 | /** |
1366 | * __netlink_ns_capable - General netlink message capability test | |
1367 | * @nsp: NETLINK_CB of the socket buffer holding a netlink command from userspace. | |
1368 | * @user_ns: The user namespace of the capability to use | |
1369 | * @cap: The capability to use | |
1370 | * | |
1371 | * Test to see if the opener of the socket we received the message | |
1372 | * from had when the netlink socket was created and the sender of the | |
1373 | * message has has the capability @cap in the user namespace @user_ns. | |
1374 | */ | |
1375 | bool __netlink_ns_capable(const struct netlink_skb_parms *nsp, | |
1376 | struct user_namespace *user_ns, int cap) | |
1377 | { | |
2d7a85f4 EB |
1378 | return ((nsp->flags & NETLINK_SKB_DST) || |
1379 | file_ns_capable(nsp->sk->sk_socket->file, user_ns, cap)) && | |
1380 | ns_capable(user_ns, cap); | |
aa4cf945 EB |
1381 | } |
1382 | EXPORT_SYMBOL(__netlink_ns_capable); | |
1383 | ||
1384 | /** | |
1385 | * netlink_ns_capable - General netlink message capability test | |
1386 | * @skb: socket buffer holding a netlink command from userspace | |
1387 | * @user_ns: The user namespace of the capability to use | |
1388 | * @cap: The capability to use | |
1389 | * | |
1390 | * Test to see if the opener of the socket we received the message | |
1391 | * from had when the netlink socket was created and the sender of the | |
1392 | * message has has the capability @cap in the user namespace @user_ns. | |
1393 | */ | |
1394 | bool netlink_ns_capable(const struct sk_buff *skb, | |
1395 | struct user_namespace *user_ns, int cap) | |
1396 | { | |
1397 | return __netlink_ns_capable(&NETLINK_CB(skb), user_ns, cap); | |
1398 | } | |
1399 | EXPORT_SYMBOL(netlink_ns_capable); | |
1400 | ||
1401 | /** | |
1402 | * netlink_capable - Netlink global message capability test | |
1403 | * @skb: socket buffer holding a netlink command from userspace | |
1404 | * @cap: The capability to use | |
1405 | * | |
1406 | * Test to see if the opener of the socket we received the message | |
1407 | * from had when the netlink socket was created and the sender of the | |
1408 | * message has has the capability @cap in all user namespaces. | |
1409 | */ | |
1410 | bool netlink_capable(const struct sk_buff *skb, int cap) | |
1411 | { | |
1412 | return netlink_ns_capable(skb, &init_user_ns, cap); | |
1413 | } | |
1414 | EXPORT_SYMBOL(netlink_capable); | |
1415 | ||
1416 | /** | |
1417 | * netlink_net_capable - Netlink network namespace message capability test | |
1418 | * @skb: socket buffer holding a netlink command from userspace | |
1419 | * @cap: The capability to use | |
1420 | * | |
1421 | * Test to see if the opener of the socket we received the message | |
1422 | * from had when the netlink socket was created and the sender of the | |
1423 | * message has has the capability @cap over the network namespace of | |
1424 | * the socket we received the message from. | |
1425 | */ | |
1426 | bool netlink_net_capable(const struct sk_buff *skb, int cap) | |
1427 | { | |
1428 | return netlink_ns_capable(skb, sock_net(skb->sk)->user_ns, cap); | |
1429 | } | |
1430 | EXPORT_SYMBOL(netlink_net_capable); | |
1431 | ||
5187cd05 | 1432 | static inline int netlink_allowed(const struct socket *sock, unsigned int flag) |
746fac4d | 1433 | { |
9785e10a | 1434 | return (nl_table[sock->sk->sk_protocol].flags & flag) || |
df008c91 | 1435 | ns_capable(sock_net(sock->sk)->user_ns, CAP_NET_ADMIN); |
746fac4d | 1436 | } |
1da177e4 | 1437 | |
f7fa9b10 PM |
1438 | static void |
1439 | netlink_update_subscriptions(struct sock *sk, unsigned int subscriptions) | |
1440 | { | |
1441 | struct netlink_sock *nlk = nlk_sk(sk); | |
1442 | ||
1443 | if (nlk->subscriptions && !subscriptions) | |
1444 | __sk_del_bind_node(sk); | |
1445 | else if (!nlk->subscriptions && subscriptions) | |
1446 | sk_add_bind_node(sk, &nl_table[sk->sk_protocol].mc_list); | |
1447 | nlk->subscriptions = subscriptions; | |
1448 | } | |
1449 | ||
b4ff4f04 | 1450 | static int netlink_realloc_groups(struct sock *sk) |
513c2500 PM |
1451 | { |
1452 | struct netlink_sock *nlk = nlk_sk(sk); | |
1453 | unsigned int groups; | |
b4ff4f04 | 1454 | unsigned long *new_groups; |
513c2500 PM |
1455 | int err = 0; |
1456 | ||
b4ff4f04 JB |
1457 | netlink_table_grab(); |
1458 | ||
513c2500 | 1459 | groups = nl_table[sk->sk_protocol].groups; |
b4ff4f04 | 1460 | if (!nl_table[sk->sk_protocol].registered) { |
513c2500 | 1461 | err = -ENOENT; |
b4ff4f04 JB |
1462 | goto out_unlock; |
1463 | } | |
513c2500 | 1464 | |
b4ff4f04 JB |
1465 | if (nlk->ngroups >= groups) |
1466 | goto out_unlock; | |
513c2500 | 1467 | |
b4ff4f04 JB |
1468 | new_groups = krealloc(nlk->groups, NLGRPSZ(groups), GFP_ATOMIC); |
1469 | if (new_groups == NULL) { | |
1470 | err = -ENOMEM; | |
1471 | goto out_unlock; | |
1472 | } | |
6ac552fd | 1473 | memset((char *)new_groups + NLGRPSZ(nlk->ngroups), 0, |
b4ff4f04 JB |
1474 | NLGRPSZ(groups) - NLGRPSZ(nlk->ngroups)); |
1475 | ||
1476 | nlk->groups = new_groups; | |
513c2500 | 1477 | nlk->ngroups = groups; |
b4ff4f04 JB |
1478 | out_unlock: |
1479 | netlink_table_ungrab(); | |
1480 | return err; | |
513c2500 PM |
1481 | } |
1482 | ||
4f520900 RGB |
1483 | static void netlink_unbind(int group, long unsigned int groups, |
1484 | struct netlink_sock *nlk) | |
1485 | { | |
1486 | int undo; | |
1487 | ||
1488 | if (!nlk->netlink_unbind) | |
1489 | return; | |
1490 | ||
1491 | for (undo = 0; undo < group; undo++) | |
1492 | if (test_bit(group, &groups)) | |
1493 | nlk->netlink_unbind(undo); | |
1494 | } | |
1495 | ||
6ac552fd PM |
1496 | static int netlink_bind(struct socket *sock, struct sockaddr *addr, |
1497 | int addr_len) | |
1da177e4 LT |
1498 | { |
1499 | struct sock *sk = sock->sk; | |
3b1e0a65 | 1500 | struct net *net = sock_net(sk); |
1da177e4 LT |
1501 | struct netlink_sock *nlk = nlk_sk(sk); |
1502 | struct sockaddr_nl *nladdr = (struct sockaddr_nl *)addr; | |
1503 | int err; | |
4f520900 | 1504 | long unsigned int groups = nladdr->nl_groups; |
746fac4d | 1505 | |
4e4b5376 HFS |
1506 | if (addr_len < sizeof(struct sockaddr_nl)) |
1507 | return -EINVAL; | |
1508 | ||
1da177e4 LT |
1509 | if (nladdr->nl_family != AF_NETLINK) |
1510 | return -EINVAL; | |
1511 | ||
1512 | /* Only superuser is allowed to listen multicasts */ | |
4f520900 | 1513 | if (groups) { |
5187cd05 | 1514 | if (!netlink_allowed(sock, NL_CFG_F_NONROOT_RECV)) |
513c2500 | 1515 | return -EPERM; |
b4ff4f04 JB |
1516 | err = netlink_realloc_groups(sk); |
1517 | if (err) | |
1518 | return err; | |
513c2500 | 1519 | } |
1da177e4 | 1520 | |
4f520900 | 1521 | if (nlk->portid) |
15e47304 | 1522 | if (nladdr->nl_pid != nlk->portid) |
1da177e4 | 1523 | return -EINVAL; |
4f520900 RGB |
1524 | |
1525 | if (nlk->netlink_bind && groups) { | |
1526 | int group; | |
1527 | ||
1528 | for (group = 0; group < nlk->ngroups; group++) { | |
1529 | if (!test_bit(group, &groups)) | |
1530 | continue; | |
1531 | err = nlk->netlink_bind(group); | |
1532 | if (!err) | |
1533 | continue; | |
1534 | netlink_unbind(group, groups, nlk); | |
1535 | return err; | |
1536 | } | |
1537 | } | |
1538 | ||
1539 | if (!nlk->portid) { | |
1da177e4 | 1540 | err = nladdr->nl_pid ? |
b4b51029 | 1541 | netlink_insert(sk, net, nladdr->nl_pid) : |
1da177e4 | 1542 | netlink_autobind(sock); |
4f520900 RGB |
1543 | if (err) { |
1544 | netlink_unbind(nlk->ngroups - 1, groups, nlk); | |
1da177e4 | 1545 | return err; |
4f520900 | 1546 | } |
1da177e4 LT |
1547 | } |
1548 | ||
4f520900 | 1549 | if (!groups && (nlk->groups == NULL || !(u32)nlk->groups[0])) |
1da177e4 LT |
1550 | return 0; |
1551 | ||
1552 | netlink_table_grab(); | |
f7fa9b10 | 1553 | netlink_update_subscriptions(sk, nlk->subscriptions + |
4f520900 | 1554 | hweight32(groups) - |
746fac4d | 1555 | hweight32(nlk->groups[0])); |
4f520900 | 1556 | nlk->groups[0] = (nlk->groups[0] & ~0xffffffffUL) | groups; |
4277a083 | 1557 | netlink_update_listeners(sk); |
1da177e4 LT |
1558 | netlink_table_ungrab(); |
1559 | ||
1560 | return 0; | |
1561 | } | |
1562 | ||
1563 | static int netlink_connect(struct socket *sock, struct sockaddr *addr, | |
1564 | int alen, int flags) | |
1565 | { | |
1566 | int err = 0; | |
1567 | struct sock *sk = sock->sk; | |
1568 | struct netlink_sock *nlk = nlk_sk(sk); | |
6ac552fd | 1569 | struct sockaddr_nl *nladdr = (struct sockaddr_nl *)addr; |
1da177e4 | 1570 | |
6503d961 CG |
1571 | if (alen < sizeof(addr->sa_family)) |
1572 | return -EINVAL; | |
1573 | ||
1da177e4 LT |
1574 | if (addr->sa_family == AF_UNSPEC) { |
1575 | sk->sk_state = NETLINK_UNCONNECTED; | |
15e47304 | 1576 | nlk->dst_portid = 0; |
d629b836 | 1577 | nlk->dst_group = 0; |
1da177e4 LT |
1578 | return 0; |
1579 | } | |
1580 | if (addr->sa_family != AF_NETLINK) | |
1581 | return -EINVAL; | |
1582 | ||
46833a86 | 1583 | if ((nladdr->nl_groups || nladdr->nl_pid) && |
5187cd05 | 1584 | !netlink_allowed(sock, NL_CFG_F_NONROOT_SEND)) |
1da177e4 LT |
1585 | return -EPERM; |
1586 | ||
15e47304 | 1587 | if (!nlk->portid) |
1da177e4 LT |
1588 | err = netlink_autobind(sock); |
1589 | ||
1590 | if (err == 0) { | |
1591 | sk->sk_state = NETLINK_CONNECTED; | |
15e47304 | 1592 | nlk->dst_portid = nladdr->nl_pid; |
d629b836 | 1593 | nlk->dst_group = ffs(nladdr->nl_groups); |
1da177e4 LT |
1594 | } |
1595 | ||
1596 | return err; | |
1597 | } | |
1598 | ||
6ac552fd PM |
1599 | static int netlink_getname(struct socket *sock, struct sockaddr *addr, |
1600 | int *addr_len, int peer) | |
1da177e4 LT |
1601 | { |
1602 | struct sock *sk = sock->sk; | |
1603 | struct netlink_sock *nlk = nlk_sk(sk); | |
13cfa97b | 1604 | DECLARE_SOCKADDR(struct sockaddr_nl *, nladdr, addr); |
746fac4d | 1605 | |
1da177e4 LT |
1606 | nladdr->nl_family = AF_NETLINK; |
1607 | nladdr->nl_pad = 0; | |
1608 | *addr_len = sizeof(*nladdr); | |
1609 | ||
1610 | if (peer) { | |
15e47304 | 1611 | nladdr->nl_pid = nlk->dst_portid; |
d629b836 | 1612 | nladdr->nl_groups = netlink_group_mask(nlk->dst_group); |
1da177e4 | 1613 | } else { |
15e47304 | 1614 | nladdr->nl_pid = nlk->portid; |
513c2500 | 1615 | nladdr->nl_groups = nlk->groups ? nlk->groups[0] : 0; |
1da177e4 LT |
1616 | } |
1617 | return 0; | |
1618 | } | |
1619 | ||
15e47304 | 1620 | static struct sock *netlink_getsockbyportid(struct sock *ssk, u32 portid) |
1da177e4 | 1621 | { |
1da177e4 LT |
1622 | struct sock *sock; |
1623 | struct netlink_sock *nlk; | |
1624 | ||
15e47304 | 1625 | sock = netlink_lookup(sock_net(ssk), ssk->sk_protocol, portid); |
1da177e4 LT |
1626 | if (!sock) |
1627 | return ERR_PTR(-ECONNREFUSED); | |
1628 | ||
1629 | /* Don't bother queuing skb if kernel socket has no input function */ | |
1630 | nlk = nlk_sk(sock); | |
cd40b7d3 | 1631 | if (sock->sk_state == NETLINK_CONNECTED && |
15e47304 | 1632 | nlk->dst_portid != nlk_sk(ssk)->portid) { |
1da177e4 LT |
1633 | sock_put(sock); |
1634 | return ERR_PTR(-ECONNREFUSED); | |
1635 | } | |
1636 | return sock; | |
1637 | } | |
1638 | ||
1639 | struct sock *netlink_getsockbyfilp(struct file *filp) | |
1640 | { | |
496ad9aa | 1641 | struct inode *inode = file_inode(filp); |
1da177e4 LT |
1642 | struct sock *sock; |
1643 | ||
1644 | if (!S_ISSOCK(inode->i_mode)) | |
1645 | return ERR_PTR(-ENOTSOCK); | |
1646 | ||
1647 | sock = SOCKET_I(inode)->sk; | |
1648 | if (sock->sk_family != AF_NETLINK) | |
1649 | return ERR_PTR(-EINVAL); | |
1650 | ||
1651 | sock_hold(sock); | |
1652 | return sock; | |
1653 | } | |
1654 | ||
3a36515f PN |
1655 | static struct sk_buff *netlink_alloc_large_skb(unsigned int size, |
1656 | int broadcast) | |
c05cdb1b PNA |
1657 | { |
1658 | struct sk_buff *skb; | |
1659 | void *data; | |
1660 | ||
3a36515f | 1661 | if (size <= NLMSG_GOODSIZE || broadcast) |
c05cdb1b PNA |
1662 | return alloc_skb(size, GFP_KERNEL); |
1663 | ||
3a36515f PN |
1664 | size = SKB_DATA_ALIGN(size) + |
1665 | SKB_DATA_ALIGN(sizeof(struct skb_shared_info)); | |
c05cdb1b PNA |
1666 | |
1667 | data = vmalloc(size); | |
1668 | if (data == NULL) | |
3a36515f | 1669 | return NULL; |
c05cdb1b | 1670 | |
3a36515f PN |
1671 | skb = build_skb(data, size); |
1672 | if (skb == NULL) | |
1673 | vfree(data); | |
1674 | else { | |
1675 | skb->head_frag = 0; | |
1676 | skb->destructor = netlink_skb_destructor; | |
1677 | } | |
c05cdb1b PNA |
1678 | |
1679 | return skb; | |
c05cdb1b PNA |
1680 | } |
1681 | ||
1da177e4 LT |
1682 | /* |
1683 | * Attach a skb to a netlink socket. | |
1684 | * The caller must hold a reference to the destination socket. On error, the | |
1685 | * reference is dropped. The skb is not send to the destination, just all | |
1686 | * all error checks are performed and memory in the queue is reserved. | |
1687 | * Return values: | |
1688 | * < 0: error. skb freed, reference to sock dropped. | |
1689 | * 0: continue | |
1690 | * 1: repeat lookup - reference dropped while waiting for socket memory. | |
1691 | */ | |
9457afee | 1692 | int netlink_attachskb(struct sock *sk, struct sk_buff *skb, |
c3d8d1e3 | 1693 | long *timeo, struct sock *ssk) |
1da177e4 LT |
1694 | { |
1695 | struct netlink_sock *nlk; | |
1696 | ||
1697 | nlk = nlk_sk(sk); | |
1698 | ||
5fd96123 PM |
1699 | if ((atomic_read(&sk->sk_rmem_alloc) > sk->sk_rcvbuf || |
1700 | test_bit(NETLINK_CONGESTED, &nlk->state)) && | |
1701 | !netlink_skb_is_mmaped(skb)) { | |
1da177e4 | 1702 | DECLARE_WAITQUEUE(wait, current); |
c3d8d1e3 | 1703 | if (!*timeo) { |
aed81560 | 1704 | if (!ssk || netlink_is_kernel(ssk)) |
1da177e4 LT |
1705 | netlink_overrun(sk); |
1706 | sock_put(sk); | |
1707 | kfree_skb(skb); | |
1708 | return -EAGAIN; | |
1709 | } | |
1710 | ||
1711 | __set_current_state(TASK_INTERRUPTIBLE); | |
1712 | add_wait_queue(&nlk->wait, &wait); | |
1713 | ||
1714 | if ((atomic_read(&sk->sk_rmem_alloc) > sk->sk_rcvbuf || | |
cd967e05 | 1715 | test_bit(NETLINK_CONGESTED, &nlk->state)) && |
1da177e4 | 1716 | !sock_flag(sk, SOCK_DEAD)) |
c3d8d1e3 | 1717 | *timeo = schedule_timeout(*timeo); |
1da177e4 LT |
1718 | |
1719 | __set_current_state(TASK_RUNNING); | |
1720 | remove_wait_queue(&nlk->wait, &wait); | |
1721 | sock_put(sk); | |
1722 | ||
1723 | if (signal_pending(current)) { | |
1724 | kfree_skb(skb); | |
c3d8d1e3 | 1725 | return sock_intr_errno(*timeo); |
1da177e4 LT |
1726 | } |
1727 | return 1; | |
1728 | } | |
cf0a018a | 1729 | netlink_skb_set_owner_r(skb, sk); |
1da177e4 LT |
1730 | return 0; |
1731 | } | |
1732 | ||
4a7e7c2a | 1733 | static int __netlink_sendskb(struct sock *sk, struct sk_buff *skb) |
1da177e4 | 1734 | { |
1da177e4 LT |
1735 | int len = skb->len; |
1736 | ||
bcbde0d4 DB |
1737 | netlink_deliver_tap(skb); |
1738 | ||
f9c22888 PM |
1739 | #ifdef CONFIG_NETLINK_MMAP |
1740 | if (netlink_skb_is_mmaped(skb)) | |
1741 | netlink_queue_mmaped_skb(sk, skb); | |
1742 | else if (netlink_rx_is_mmaped(sk)) | |
1743 | netlink_ring_set_copied(sk, skb); | |
1744 | else | |
1745 | #endif /* CONFIG_NETLINK_MMAP */ | |
1746 | skb_queue_tail(&sk->sk_receive_queue, skb); | |
676d2369 | 1747 | sk->sk_data_ready(sk); |
4a7e7c2a ED |
1748 | return len; |
1749 | } | |
1750 | ||
1751 | int netlink_sendskb(struct sock *sk, struct sk_buff *skb) | |
1752 | { | |
1753 | int len = __netlink_sendskb(sk, skb); | |
1754 | ||
1da177e4 LT |
1755 | sock_put(sk); |
1756 | return len; | |
1757 | } | |
1758 | ||
1759 | void netlink_detachskb(struct sock *sk, struct sk_buff *skb) | |
1760 | { | |
1761 | kfree_skb(skb); | |
1762 | sock_put(sk); | |
1763 | } | |
1764 | ||
b57ef81f | 1765 | static struct sk_buff *netlink_trim(struct sk_buff *skb, gfp_t allocation) |
1da177e4 LT |
1766 | { |
1767 | int delta; | |
1768 | ||
1298ca46 | 1769 | WARN_ON(skb->sk != NULL); |
5fd96123 PM |
1770 | if (netlink_skb_is_mmaped(skb)) |
1771 | return skb; | |
1da177e4 | 1772 | |
4305b541 | 1773 | delta = skb->end - skb->tail; |
c05cdb1b | 1774 | if (is_vmalloc_addr(skb->head) || delta * 2 < skb->truesize) |
1da177e4 LT |
1775 | return skb; |
1776 | ||
1777 | if (skb_shared(skb)) { | |
1778 | struct sk_buff *nskb = skb_clone(skb, allocation); | |
1779 | if (!nskb) | |
1780 | return skb; | |
8460c00f | 1781 | consume_skb(skb); |
1da177e4 LT |
1782 | skb = nskb; |
1783 | } | |
1784 | ||
1785 | if (!pskb_expand_head(skb, 0, -delta, allocation)) | |
1786 | skb->truesize -= delta; | |
1787 | ||
1788 | return skb; | |
1789 | } | |
1790 | ||
3fbc2905 EB |
1791 | static int netlink_unicast_kernel(struct sock *sk, struct sk_buff *skb, |
1792 | struct sock *ssk) | |
cd40b7d3 DL |
1793 | { |
1794 | int ret; | |
1795 | struct netlink_sock *nlk = nlk_sk(sk); | |
1796 | ||
1797 | ret = -ECONNREFUSED; | |
1798 | if (nlk->netlink_rcv != NULL) { | |
1799 | ret = skb->len; | |
cf0a018a | 1800 | netlink_skb_set_owner_r(skb, sk); |
e32123e5 | 1801 | NETLINK_CB(skb).sk = ssk; |
73bfd370 | 1802 | netlink_deliver_tap_kernel(sk, ssk, skb); |
cd40b7d3 | 1803 | nlk->netlink_rcv(skb); |
bfb253c9 ED |
1804 | consume_skb(skb); |
1805 | } else { | |
1806 | kfree_skb(skb); | |
cd40b7d3 | 1807 | } |
cd40b7d3 DL |
1808 | sock_put(sk); |
1809 | return ret; | |
1810 | } | |
1811 | ||
1812 | int netlink_unicast(struct sock *ssk, struct sk_buff *skb, | |
15e47304 | 1813 | u32 portid, int nonblock) |
1da177e4 LT |
1814 | { |
1815 | struct sock *sk; | |
1816 | int err; | |
1817 | long timeo; | |
1818 | ||
1819 | skb = netlink_trim(skb, gfp_any()); | |
1820 | ||
1821 | timeo = sock_sndtimeo(ssk, nonblock); | |
1822 | retry: | |
15e47304 | 1823 | sk = netlink_getsockbyportid(ssk, portid); |
1da177e4 LT |
1824 | if (IS_ERR(sk)) { |
1825 | kfree_skb(skb); | |
1826 | return PTR_ERR(sk); | |
1827 | } | |
cd40b7d3 | 1828 | if (netlink_is_kernel(sk)) |
3fbc2905 | 1829 | return netlink_unicast_kernel(sk, skb, ssk); |
cd40b7d3 | 1830 | |
b1153f29 | 1831 | if (sk_filter(sk, skb)) { |
84874607 | 1832 | err = skb->len; |
b1153f29 SH |
1833 | kfree_skb(skb); |
1834 | sock_put(sk); | |
1835 | return err; | |
1836 | } | |
1837 | ||
9457afee | 1838 | err = netlink_attachskb(sk, skb, &timeo, ssk); |
1da177e4 LT |
1839 | if (err == 1) |
1840 | goto retry; | |
1841 | if (err) | |
1842 | return err; | |
1843 | ||
7ee015e0 | 1844 | return netlink_sendskb(sk, skb); |
1da177e4 | 1845 | } |
6ac552fd | 1846 | EXPORT_SYMBOL(netlink_unicast); |
1da177e4 | 1847 | |
f9c22888 PM |
1848 | struct sk_buff *netlink_alloc_skb(struct sock *ssk, unsigned int size, |
1849 | u32 dst_portid, gfp_t gfp_mask) | |
1850 | { | |
1851 | #ifdef CONFIG_NETLINK_MMAP | |
1852 | struct sock *sk = NULL; | |
1853 | struct sk_buff *skb; | |
1854 | struct netlink_ring *ring; | |
1855 | struct nl_mmap_hdr *hdr; | |
1856 | unsigned int maxlen; | |
1857 | ||
1858 | sk = netlink_getsockbyportid(ssk, dst_portid); | |
1859 | if (IS_ERR(sk)) | |
1860 | goto out; | |
1861 | ||
1862 | ring = &nlk_sk(sk)->rx_ring; | |
1863 | /* fast-path without atomic ops for common case: non-mmaped receiver */ | |
1864 | if (ring->pg_vec == NULL) | |
1865 | goto out_put; | |
1866 | ||
aae9f0e2 TG |
1867 | if (ring->frame_size - NL_MMAP_HDRLEN < size) |
1868 | goto out_put; | |
1869 | ||
f9c22888 PM |
1870 | skb = alloc_skb_head(gfp_mask); |
1871 | if (skb == NULL) | |
1872 | goto err1; | |
1873 | ||
1874 | spin_lock_bh(&sk->sk_receive_queue.lock); | |
1875 | /* check again under lock */ | |
1876 | if (ring->pg_vec == NULL) | |
1877 | goto out_free; | |
1878 | ||
aae9f0e2 | 1879 | /* check again under lock */ |
f9c22888 PM |
1880 | maxlen = ring->frame_size - NL_MMAP_HDRLEN; |
1881 | if (maxlen < size) | |
1882 | goto out_free; | |
1883 | ||
1884 | netlink_forward_ring(ring); | |
1885 | hdr = netlink_current_frame(ring, NL_MMAP_STATUS_UNUSED); | |
1886 | if (hdr == NULL) | |
1887 | goto err2; | |
1888 | netlink_ring_setup_skb(skb, sk, ring, hdr); | |
1889 | netlink_set_status(hdr, NL_MMAP_STATUS_RESERVED); | |
1890 | atomic_inc(&ring->pending); | |
1891 | netlink_increment_head(ring); | |
1892 | ||
1893 | spin_unlock_bh(&sk->sk_receive_queue.lock); | |
1894 | return skb; | |
1895 | ||
1896 | err2: | |
1897 | kfree_skb(skb); | |
1898 | spin_unlock_bh(&sk->sk_receive_queue.lock); | |
cd1df525 | 1899 | netlink_overrun(sk); |
f9c22888 PM |
1900 | err1: |
1901 | sock_put(sk); | |
1902 | return NULL; | |
1903 | ||
1904 | out_free: | |
1905 | kfree_skb(skb); | |
1906 | spin_unlock_bh(&sk->sk_receive_queue.lock); | |
1907 | out_put: | |
1908 | sock_put(sk); | |
1909 | out: | |
1910 | #endif | |
1911 | return alloc_skb(size, gfp_mask); | |
1912 | } | |
1913 | EXPORT_SYMBOL_GPL(netlink_alloc_skb); | |
1914 | ||
4277a083 PM |
1915 | int netlink_has_listeners(struct sock *sk, unsigned int group) |
1916 | { | |
1917 | int res = 0; | |
5c398dc8 | 1918 | struct listeners *listeners; |
4277a083 | 1919 | |
aed81560 | 1920 | BUG_ON(!netlink_is_kernel(sk)); |
b4ff4f04 JB |
1921 | |
1922 | rcu_read_lock(); | |
1923 | listeners = rcu_dereference(nl_table[sk->sk_protocol].listeners); | |
1924 | ||
6d772ac5 | 1925 | if (listeners && group - 1 < nl_table[sk->sk_protocol].groups) |
5c398dc8 | 1926 | res = test_bit(group - 1, listeners->masks); |
b4ff4f04 JB |
1927 | |
1928 | rcu_read_unlock(); | |
1929 | ||
4277a083 PM |
1930 | return res; |
1931 | } | |
1932 | EXPORT_SYMBOL_GPL(netlink_has_listeners); | |
1933 | ||
b57ef81f | 1934 | static int netlink_broadcast_deliver(struct sock *sk, struct sk_buff *skb) |
1da177e4 LT |
1935 | { |
1936 | struct netlink_sock *nlk = nlk_sk(sk); | |
1937 | ||
1938 | if (atomic_read(&sk->sk_rmem_alloc) <= sk->sk_rcvbuf && | |
cd967e05 | 1939 | !test_bit(NETLINK_CONGESTED, &nlk->state)) { |
cf0a018a | 1940 | netlink_skb_set_owner_r(skb, sk); |
4a7e7c2a | 1941 | __netlink_sendskb(sk, skb); |
2c645800 | 1942 | return atomic_read(&sk->sk_rmem_alloc) > (sk->sk_rcvbuf >> 1); |
1da177e4 LT |
1943 | } |
1944 | return -1; | |
1945 | } | |
1946 | ||
1947 | struct netlink_broadcast_data { | |
1948 | struct sock *exclude_sk; | |
b4b51029 | 1949 | struct net *net; |
15e47304 | 1950 | u32 portid; |
1da177e4 LT |
1951 | u32 group; |
1952 | int failure; | |
ff491a73 | 1953 | int delivery_failure; |
1da177e4 LT |
1954 | int congested; |
1955 | int delivered; | |
7d877f3b | 1956 | gfp_t allocation; |
1da177e4 | 1957 | struct sk_buff *skb, *skb2; |
910a7e90 EB |
1958 | int (*tx_filter)(struct sock *dsk, struct sk_buff *skb, void *data); |
1959 | void *tx_data; | |
1da177e4 LT |
1960 | }; |
1961 | ||
46c9521f RR |
1962 | static void do_one_broadcast(struct sock *sk, |
1963 | struct netlink_broadcast_data *p) | |
1da177e4 LT |
1964 | { |
1965 | struct netlink_sock *nlk = nlk_sk(sk); | |
1966 | int val; | |
1967 | ||
1968 | if (p->exclude_sk == sk) | |
46c9521f | 1969 | return; |
1da177e4 | 1970 | |
15e47304 | 1971 | if (nlk->portid == p->portid || p->group - 1 >= nlk->ngroups || |
f7fa9b10 | 1972 | !test_bit(p->group - 1, nlk->groups)) |
46c9521f | 1973 | return; |
1da177e4 | 1974 | |
878628fb | 1975 | if (!net_eq(sock_net(sk), p->net)) |
46c9521f | 1976 | return; |
b4b51029 | 1977 | |
1da177e4 LT |
1978 | if (p->failure) { |
1979 | netlink_overrun(sk); | |
46c9521f | 1980 | return; |
1da177e4 LT |
1981 | } |
1982 | ||
1983 | sock_hold(sk); | |
1984 | if (p->skb2 == NULL) { | |
68acc024 | 1985 | if (skb_shared(p->skb)) { |
1da177e4 LT |
1986 | p->skb2 = skb_clone(p->skb, p->allocation); |
1987 | } else { | |
68acc024 TC |
1988 | p->skb2 = skb_get(p->skb); |
1989 | /* | |
1990 | * skb ownership may have been set when | |
1991 | * delivered to a previous socket. | |
1992 | */ | |
1993 | skb_orphan(p->skb2); | |
1da177e4 LT |
1994 | } |
1995 | } | |
1996 | if (p->skb2 == NULL) { | |
1997 | netlink_overrun(sk); | |
1998 | /* Clone failed. Notify ALL listeners. */ | |
1999 | p->failure = 1; | |
be0c22a4 PNA |
2000 | if (nlk->flags & NETLINK_BROADCAST_SEND_ERROR) |
2001 | p->delivery_failure = 1; | |
910a7e90 EB |
2002 | } else if (p->tx_filter && p->tx_filter(sk, p->skb2, p->tx_data)) { |
2003 | kfree_skb(p->skb2); | |
2004 | p->skb2 = NULL; | |
b1153f29 SH |
2005 | } else if (sk_filter(sk, p->skb2)) { |
2006 | kfree_skb(p->skb2); | |
2007 | p->skb2 = NULL; | |
1da177e4 LT |
2008 | } else if ((val = netlink_broadcast_deliver(sk, p->skb2)) < 0) { |
2009 | netlink_overrun(sk); | |
be0c22a4 PNA |
2010 | if (nlk->flags & NETLINK_BROADCAST_SEND_ERROR) |
2011 | p->delivery_failure = 1; | |
1da177e4 LT |
2012 | } else { |
2013 | p->congested |= val; | |
2014 | p->delivered = 1; | |
2015 | p->skb2 = NULL; | |
2016 | } | |
2017 | sock_put(sk); | |
1da177e4 LT |
2018 | } |
2019 | ||
15e47304 | 2020 | int netlink_broadcast_filtered(struct sock *ssk, struct sk_buff *skb, u32 portid, |
910a7e90 EB |
2021 | u32 group, gfp_t allocation, |
2022 | int (*filter)(struct sock *dsk, struct sk_buff *skb, void *data), | |
2023 | void *filter_data) | |
1da177e4 | 2024 | { |
3b1e0a65 | 2025 | struct net *net = sock_net(ssk); |
1da177e4 | 2026 | struct netlink_broadcast_data info; |
1da177e4 LT |
2027 | struct sock *sk; |
2028 | ||
2029 | skb = netlink_trim(skb, allocation); | |
2030 | ||
2031 | info.exclude_sk = ssk; | |
b4b51029 | 2032 | info.net = net; |
15e47304 | 2033 | info.portid = portid; |
1da177e4 LT |
2034 | info.group = group; |
2035 | info.failure = 0; | |
ff491a73 | 2036 | info.delivery_failure = 0; |
1da177e4 LT |
2037 | info.congested = 0; |
2038 | info.delivered = 0; | |
2039 | info.allocation = allocation; | |
2040 | info.skb = skb; | |
2041 | info.skb2 = NULL; | |
910a7e90 EB |
2042 | info.tx_filter = filter; |
2043 | info.tx_data = filter_data; | |
1da177e4 LT |
2044 | |
2045 | /* While we sleep in clone, do not allow to change socket list */ | |
2046 | ||
2047 | netlink_lock_table(); | |
2048 | ||
b67bfe0d | 2049 | sk_for_each_bound(sk, &nl_table[ssk->sk_protocol].mc_list) |
1da177e4 LT |
2050 | do_one_broadcast(sk, &info); |
2051 | ||
70d4bf6d | 2052 | consume_skb(skb); |
aa1c6a6f | 2053 | |
1da177e4 LT |
2054 | netlink_unlock_table(); |
2055 | ||
70d4bf6d NH |
2056 | if (info.delivery_failure) { |
2057 | kfree_skb(info.skb2); | |
ff491a73 | 2058 | return -ENOBUFS; |
658cb354 ED |
2059 | } |
2060 | consume_skb(info.skb2); | |
ff491a73 | 2061 | |
1da177e4 LT |
2062 | if (info.delivered) { |
2063 | if (info.congested && (allocation & __GFP_WAIT)) | |
2064 | yield(); | |
2065 | return 0; | |
2066 | } | |
1da177e4 LT |
2067 | return -ESRCH; |
2068 | } | |
910a7e90 EB |
2069 | EXPORT_SYMBOL(netlink_broadcast_filtered); |
2070 | ||
15e47304 | 2071 | int netlink_broadcast(struct sock *ssk, struct sk_buff *skb, u32 portid, |
910a7e90 EB |
2072 | u32 group, gfp_t allocation) |
2073 | { | |
15e47304 | 2074 | return netlink_broadcast_filtered(ssk, skb, portid, group, allocation, |
910a7e90 EB |
2075 | NULL, NULL); |
2076 | } | |
6ac552fd | 2077 | EXPORT_SYMBOL(netlink_broadcast); |
1da177e4 LT |
2078 | |
2079 | struct netlink_set_err_data { | |
2080 | struct sock *exclude_sk; | |
15e47304 | 2081 | u32 portid; |
1da177e4 LT |
2082 | u32 group; |
2083 | int code; | |
2084 | }; | |
2085 | ||
b57ef81f | 2086 | static int do_one_set_err(struct sock *sk, struct netlink_set_err_data *p) |
1da177e4 LT |
2087 | { |
2088 | struct netlink_sock *nlk = nlk_sk(sk); | |
1a50307b | 2089 | int ret = 0; |
1da177e4 LT |
2090 | |
2091 | if (sk == p->exclude_sk) | |
2092 | goto out; | |
2093 | ||
09ad9bc7 | 2094 | if (!net_eq(sock_net(sk), sock_net(p->exclude_sk))) |
b4b51029 EB |
2095 | goto out; |
2096 | ||
15e47304 | 2097 | if (nlk->portid == p->portid || p->group - 1 >= nlk->ngroups || |
f7fa9b10 | 2098 | !test_bit(p->group - 1, nlk->groups)) |
1da177e4 LT |
2099 | goto out; |
2100 | ||
1a50307b PNA |
2101 | if (p->code == ENOBUFS && nlk->flags & NETLINK_RECV_NO_ENOBUFS) { |
2102 | ret = 1; | |
2103 | goto out; | |
2104 | } | |
2105 | ||
1da177e4 LT |
2106 | sk->sk_err = p->code; |
2107 | sk->sk_error_report(sk); | |
2108 | out: | |
1a50307b | 2109 | return ret; |
1da177e4 LT |
2110 | } |
2111 | ||
4843b93c PNA |
2112 | /** |
2113 | * netlink_set_err - report error to broadcast listeners | |
2114 | * @ssk: the kernel netlink socket, as returned by netlink_kernel_create() | |
15e47304 | 2115 | * @portid: the PORTID of a process that we want to skip (if any) |
840e93f2 | 2116 | * @group: the broadcast group that will notice the error |
4843b93c | 2117 | * @code: error code, must be negative (as usual in kernelspace) |
1a50307b PNA |
2118 | * |
2119 | * This function returns the number of broadcast listeners that have set the | |
2120 | * NETLINK_RECV_NO_ENOBUFS socket option. | |
4843b93c | 2121 | */ |
15e47304 | 2122 | int netlink_set_err(struct sock *ssk, u32 portid, u32 group, int code) |
1da177e4 LT |
2123 | { |
2124 | struct netlink_set_err_data info; | |
1da177e4 | 2125 | struct sock *sk; |
1a50307b | 2126 | int ret = 0; |
1da177e4 LT |
2127 | |
2128 | info.exclude_sk = ssk; | |
15e47304 | 2129 | info.portid = portid; |
1da177e4 | 2130 | info.group = group; |
4843b93c PNA |
2131 | /* sk->sk_err wants a positive error value */ |
2132 | info.code = -code; | |
1da177e4 LT |
2133 | |
2134 | read_lock(&nl_table_lock); | |
2135 | ||
b67bfe0d | 2136 | sk_for_each_bound(sk, &nl_table[ssk->sk_protocol].mc_list) |
1a50307b | 2137 | ret += do_one_set_err(sk, &info); |
1da177e4 LT |
2138 | |
2139 | read_unlock(&nl_table_lock); | |
1a50307b | 2140 | return ret; |
1da177e4 | 2141 | } |
dd5b6ce6 | 2142 | EXPORT_SYMBOL(netlink_set_err); |
1da177e4 | 2143 | |
84659eb5 JB |
2144 | /* must be called with netlink table grabbed */ |
2145 | static void netlink_update_socket_mc(struct netlink_sock *nlk, | |
2146 | unsigned int group, | |
2147 | int is_new) | |
2148 | { | |
2149 | int old, new = !!is_new, subscriptions; | |
2150 | ||
2151 | old = test_bit(group - 1, nlk->groups); | |
2152 | subscriptions = nlk->subscriptions - old + new; | |
2153 | if (new) | |
2154 | __set_bit(group - 1, nlk->groups); | |
2155 | else | |
2156 | __clear_bit(group - 1, nlk->groups); | |
2157 | netlink_update_subscriptions(&nlk->sk, subscriptions); | |
2158 | netlink_update_listeners(&nlk->sk); | |
2159 | } | |
2160 | ||
9a4595bc | 2161 | static int netlink_setsockopt(struct socket *sock, int level, int optname, |
b7058842 | 2162 | char __user *optval, unsigned int optlen) |
9a4595bc PM |
2163 | { |
2164 | struct sock *sk = sock->sk; | |
2165 | struct netlink_sock *nlk = nlk_sk(sk); | |
eb496534 JB |
2166 | unsigned int val = 0; |
2167 | int err; | |
9a4595bc PM |
2168 | |
2169 | if (level != SOL_NETLINK) | |
2170 | return -ENOPROTOOPT; | |
2171 | ||
ccdfcc39 PM |
2172 | if (optname != NETLINK_RX_RING && optname != NETLINK_TX_RING && |
2173 | optlen >= sizeof(int) && | |
eb496534 | 2174 | get_user(val, (unsigned int __user *)optval)) |
9a4595bc PM |
2175 | return -EFAULT; |
2176 | ||
2177 | switch (optname) { | |
2178 | case NETLINK_PKTINFO: | |
2179 | if (val) | |
2180 | nlk->flags |= NETLINK_RECV_PKTINFO; | |
2181 | else | |
2182 | nlk->flags &= ~NETLINK_RECV_PKTINFO; | |
2183 | err = 0; | |
2184 | break; | |
2185 | case NETLINK_ADD_MEMBERSHIP: | |
2186 | case NETLINK_DROP_MEMBERSHIP: { | |
5187cd05 | 2187 | if (!netlink_allowed(sock, NL_CFG_F_NONROOT_RECV)) |
9a4595bc | 2188 | return -EPERM; |
b4ff4f04 JB |
2189 | err = netlink_realloc_groups(sk); |
2190 | if (err) | |
2191 | return err; | |
9a4595bc PM |
2192 | if (!val || val - 1 >= nlk->ngroups) |
2193 | return -EINVAL; | |
7774d5e0 | 2194 | if (optname == NETLINK_ADD_MEMBERSHIP && nlk->netlink_bind) { |
4f520900 RGB |
2195 | err = nlk->netlink_bind(val); |
2196 | if (err) | |
2197 | return err; | |
2198 | } | |
9a4595bc | 2199 | netlink_table_grab(); |
84659eb5 JB |
2200 | netlink_update_socket_mc(nlk, val, |
2201 | optname == NETLINK_ADD_MEMBERSHIP); | |
9a4595bc | 2202 | netlink_table_ungrab(); |
7774d5e0 RGB |
2203 | if (optname == NETLINK_DROP_MEMBERSHIP && nlk->netlink_unbind) |
2204 | nlk->netlink_unbind(val); | |
03292745 | 2205 | |
9a4595bc PM |
2206 | err = 0; |
2207 | break; | |
2208 | } | |
be0c22a4 PNA |
2209 | case NETLINK_BROADCAST_ERROR: |
2210 | if (val) | |
2211 | nlk->flags |= NETLINK_BROADCAST_SEND_ERROR; | |
2212 | else | |
2213 | nlk->flags &= ~NETLINK_BROADCAST_SEND_ERROR; | |
2214 | err = 0; | |
2215 | break; | |
38938bfe PNA |
2216 | case NETLINK_NO_ENOBUFS: |
2217 | if (val) { | |
2218 | nlk->flags |= NETLINK_RECV_NO_ENOBUFS; | |
cd967e05 | 2219 | clear_bit(NETLINK_CONGESTED, &nlk->state); |
38938bfe | 2220 | wake_up_interruptible(&nlk->wait); |
658cb354 | 2221 | } else { |
38938bfe | 2222 | nlk->flags &= ~NETLINK_RECV_NO_ENOBUFS; |
658cb354 | 2223 | } |
38938bfe PNA |
2224 | err = 0; |
2225 | break; | |
ccdfcc39 PM |
2226 | #ifdef CONFIG_NETLINK_MMAP |
2227 | case NETLINK_RX_RING: | |
2228 | case NETLINK_TX_RING: { | |
2229 | struct nl_mmap_req req; | |
2230 | ||
2231 | /* Rings might consume more memory than queue limits, require | |
2232 | * CAP_NET_ADMIN. | |
2233 | */ | |
2234 | if (!capable(CAP_NET_ADMIN)) | |
2235 | return -EPERM; | |
2236 | if (optlen < sizeof(req)) | |
2237 | return -EINVAL; | |
2238 | if (copy_from_user(&req, optval, sizeof(req))) | |
2239 | return -EFAULT; | |
2240 | err = netlink_set_ring(sk, &req, false, | |
2241 | optname == NETLINK_TX_RING); | |
2242 | break; | |
2243 | } | |
2244 | #endif /* CONFIG_NETLINK_MMAP */ | |
9a4595bc PM |
2245 | default: |
2246 | err = -ENOPROTOOPT; | |
2247 | } | |
2248 | return err; | |
2249 | } | |
2250 | ||
2251 | static int netlink_getsockopt(struct socket *sock, int level, int optname, | |
746fac4d | 2252 | char __user *optval, int __user *optlen) |
9a4595bc PM |
2253 | { |
2254 | struct sock *sk = sock->sk; | |
2255 | struct netlink_sock *nlk = nlk_sk(sk); | |
2256 | int len, val, err; | |
2257 | ||
2258 | if (level != SOL_NETLINK) | |
2259 | return -ENOPROTOOPT; | |
2260 | ||
2261 | if (get_user(len, optlen)) | |
2262 | return -EFAULT; | |
2263 | if (len < 0) | |
2264 | return -EINVAL; | |
2265 | ||
2266 | switch (optname) { | |
2267 | case NETLINK_PKTINFO: | |
2268 | if (len < sizeof(int)) | |
2269 | return -EINVAL; | |
2270 | len = sizeof(int); | |
2271 | val = nlk->flags & NETLINK_RECV_PKTINFO ? 1 : 0; | |
a27b58fe HC |
2272 | if (put_user(len, optlen) || |
2273 | put_user(val, optval)) | |
2274 | return -EFAULT; | |
9a4595bc PM |
2275 | err = 0; |
2276 | break; | |
be0c22a4 PNA |
2277 | case NETLINK_BROADCAST_ERROR: |
2278 | if (len < sizeof(int)) | |
2279 | return -EINVAL; | |
2280 | len = sizeof(int); | |
2281 | val = nlk->flags & NETLINK_BROADCAST_SEND_ERROR ? 1 : 0; | |
2282 | if (put_user(len, optlen) || | |
2283 | put_user(val, optval)) | |
2284 | return -EFAULT; | |
2285 | err = 0; | |
2286 | break; | |
38938bfe PNA |
2287 | case NETLINK_NO_ENOBUFS: |
2288 | if (len < sizeof(int)) | |
2289 | return -EINVAL; | |
2290 | len = sizeof(int); | |
2291 | val = nlk->flags & NETLINK_RECV_NO_ENOBUFS ? 1 : 0; | |
2292 | if (put_user(len, optlen) || | |
2293 | put_user(val, optval)) | |
2294 | return -EFAULT; | |
2295 | err = 0; | |
2296 | break; | |
9a4595bc PM |
2297 | default: |
2298 | err = -ENOPROTOOPT; | |
2299 | } | |
2300 | return err; | |
2301 | } | |
2302 | ||
2303 | static void netlink_cmsg_recv_pktinfo(struct msghdr *msg, struct sk_buff *skb) | |
2304 | { | |
2305 | struct nl_pktinfo info; | |
2306 | ||
2307 | info.group = NETLINK_CB(skb).dst_group; | |
2308 | put_cmsg(msg, SOL_NETLINK, NETLINK_PKTINFO, sizeof(info), &info); | |
2309 | } | |
2310 | ||
1da177e4 LT |
2311 | static int netlink_sendmsg(struct kiocb *kiocb, struct socket *sock, |
2312 | struct msghdr *msg, size_t len) | |
2313 | { | |
2314 | struct sock_iocb *siocb = kiocb_to_siocb(kiocb); | |
2315 | struct sock *sk = sock->sk; | |
2316 | struct netlink_sock *nlk = nlk_sk(sk); | |
342dfc30 | 2317 | DECLARE_SOCKADDR(struct sockaddr_nl *, addr, msg->msg_name); |
15e47304 | 2318 | u32 dst_portid; |
d629b836 | 2319 | u32 dst_group; |
1da177e4 LT |
2320 | struct sk_buff *skb; |
2321 | int err; | |
2322 | struct scm_cookie scm; | |
2d7a85f4 | 2323 | u32 netlink_skb_flags = 0; |
1da177e4 LT |
2324 | |
2325 | if (msg->msg_flags&MSG_OOB) | |
2326 | return -EOPNOTSUPP; | |
2327 | ||
16e57262 | 2328 | if (NULL == siocb->scm) |
1da177e4 | 2329 | siocb->scm = &scm; |
16e57262 | 2330 | |
e0e3cea4 | 2331 | err = scm_send(sock, msg, siocb->scm, true); |
1da177e4 LT |
2332 | if (err < 0) |
2333 | return err; | |
2334 | ||
2335 | if (msg->msg_namelen) { | |
b47030c7 | 2336 | err = -EINVAL; |
1da177e4 | 2337 | if (addr->nl_family != AF_NETLINK) |
b47030c7 | 2338 | goto out; |
15e47304 | 2339 | dst_portid = addr->nl_pid; |
d629b836 | 2340 | dst_group = ffs(addr->nl_groups); |
b47030c7 | 2341 | err = -EPERM; |
15e47304 | 2342 | if ((dst_group || dst_portid) && |
5187cd05 | 2343 | !netlink_allowed(sock, NL_CFG_F_NONROOT_SEND)) |
b47030c7 | 2344 | goto out; |
2d7a85f4 | 2345 | netlink_skb_flags |= NETLINK_SKB_DST; |
1da177e4 | 2346 | } else { |
15e47304 | 2347 | dst_portid = nlk->dst_portid; |
d629b836 | 2348 | dst_group = nlk->dst_group; |
1da177e4 LT |
2349 | } |
2350 | ||
15e47304 | 2351 | if (!nlk->portid) { |
1da177e4 LT |
2352 | err = netlink_autobind(sock); |
2353 | if (err) | |
2354 | goto out; | |
2355 | } | |
2356 | ||
5fd96123 PM |
2357 | if (netlink_tx_is_mmaped(sk) && |
2358 | msg->msg_iov->iov_base == NULL) { | |
2359 | err = netlink_mmap_sendmsg(sk, msg, dst_portid, dst_group, | |
2360 | siocb); | |
2361 | goto out; | |
2362 | } | |
2363 | ||
1da177e4 LT |
2364 | err = -EMSGSIZE; |
2365 | if (len > sk->sk_sndbuf - 32) | |
2366 | goto out; | |
2367 | err = -ENOBUFS; | |
3a36515f | 2368 | skb = netlink_alloc_large_skb(len, dst_group); |
6ac552fd | 2369 | if (skb == NULL) |
1da177e4 LT |
2370 | goto out; |
2371 | ||
15e47304 | 2372 | NETLINK_CB(skb).portid = nlk->portid; |
d629b836 | 2373 | NETLINK_CB(skb).dst_group = dst_group; |
dbe9a417 | 2374 | NETLINK_CB(skb).creds = siocb->scm->creds; |
2d7a85f4 | 2375 | NETLINK_CB(skb).flags = netlink_skb_flags; |
1da177e4 | 2376 | |
1da177e4 | 2377 | err = -EFAULT; |
6ac552fd | 2378 | if (memcpy_fromiovec(skb_put(skb, len), msg->msg_iov, len)) { |
1da177e4 LT |
2379 | kfree_skb(skb); |
2380 | goto out; | |
2381 | } | |
2382 | ||
2383 | err = security_netlink_send(sk, skb); | |
2384 | if (err) { | |
2385 | kfree_skb(skb); | |
2386 | goto out; | |
2387 | } | |
2388 | ||
d629b836 | 2389 | if (dst_group) { |
1da177e4 | 2390 | atomic_inc(&skb->users); |
15e47304 | 2391 | netlink_broadcast(sk, skb, dst_portid, dst_group, GFP_KERNEL); |
1da177e4 | 2392 | } |
15e47304 | 2393 | err = netlink_unicast(sk, skb, dst_portid, msg->msg_flags&MSG_DONTWAIT); |
1da177e4 LT |
2394 | |
2395 | out: | |
b47030c7 | 2396 | scm_destroy(siocb->scm); |
1da177e4 LT |
2397 | return err; |
2398 | } | |
2399 | ||
2400 | static int netlink_recvmsg(struct kiocb *kiocb, struct socket *sock, | |
2401 | struct msghdr *msg, size_t len, | |
2402 | int flags) | |
2403 | { | |
2404 | struct sock_iocb *siocb = kiocb_to_siocb(kiocb); | |
2405 | struct scm_cookie scm; | |
2406 | struct sock *sk = sock->sk; | |
2407 | struct netlink_sock *nlk = nlk_sk(sk); | |
2408 | int noblock = flags&MSG_DONTWAIT; | |
2409 | size_t copied; | |
68d6ac6d | 2410 | struct sk_buff *skb, *data_skb; |
b44d211e | 2411 | int err, ret; |
1da177e4 LT |
2412 | |
2413 | if (flags&MSG_OOB) | |
2414 | return -EOPNOTSUPP; | |
2415 | ||
2416 | copied = 0; | |
2417 | ||
6ac552fd PM |
2418 | skb = skb_recv_datagram(sk, flags, noblock, &err); |
2419 | if (skb == NULL) | |
1da177e4 LT |
2420 | goto out; |
2421 | ||
68d6ac6d JB |
2422 | data_skb = skb; |
2423 | ||
1dacc76d JB |
2424 | #ifdef CONFIG_COMPAT_NETLINK_MESSAGES |
2425 | if (unlikely(skb_shinfo(skb)->frag_list)) { | |
1dacc76d | 2426 | /* |
68d6ac6d JB |
2427 | * If this skb has a frag_list, then here that means that we |
2428 | * will have to use the frag_list skb's data for compat tasks | |
2429 | * and the regular skb's data for normal (non-compat) tasks. | |
1dacc76d | 2430 | * |
68d6ac6d JB |
2431 | * If we need to send the compat skb, assign it to the |
2432 | * 'data_skb' variable so that it will be used below for data | |
2433 | * copying. We keep 'skb' for everything else, including | |
2434 | * freeing both later. | |
1dacc76d | 2435 | */ |
68d6ac6d JB |
2436 | if (flags & MSG_CMSG_COMPAT) |
2437 | data_skb = skb_shinfo(skb)->frag_list; | |
1dacc76d JB |
2438 | } |
2439 | #endif | |
2440 | ||
9063e21f ED |
2441 | /* Record the max length of recvmsg() calls for future allocations */ |
2442 | nlk->max_recvmsg_len = max(nlk->max_recvmsg_len, len); | |
2443 | nlk->max_recvmsg_len = min_t(size_t, nlk->max_recvmsg_len, | |
2444 | 16384); | |
2445 | ||
68d6ac6d | 2446 | copied = data_skb->len; |
1da177e4 LT |
2447 | if (len < copied) { |
2448 | msg->msg_flags |= MSG_TRUNC; | |
2449 | copied = len; | |
2450 | } | |
2451 | ||
68d6ac6d JB |
2452 | skb_reset_transport_header(data_skb); |
2453 | err = skb_copy_datagram_iovec(data_skb, 0, msg->msg_iov, copied); | |
1da177e4 LT |
2454 | |
2455 | if (msg->msg_name) { | |
342dfc30 | 2456 | DECLARE_SOCKADDR(struct sockaddr_nl *, addr, msg->msg_name); |
1da177e4 LT |
2457 | addr->nl_family = AF_NETLINK; |
2458 | addr->nl_pad = 0; | |
15e47304 | 2459 | addr->nl_pid = NETLINK_CB(skb).portid; |
d629b836 | 2460 | addr->nl_groups = netlink_group_mask(NETLINK_CB(skb).dst_group); |
1da177e4 LT |
2461 | msg->msg_namelen = sizeof(*addr); |
2462 | } | |
2463 | ||
cc9a06cd PM |
2464 | if (nlk->flags & NETLINK_RECV_PKTINFO) |
2465 | netlink_cmsg_recv_pktinfo(msg, skb); | |
2466 | ||
1da177e4 LT |
2467 | if (NULL == siocb->scm) { |
2468 | memset(&scm, 0, sizeof(scm)); | |
2469 | siocb->scm = &scm; | |
2470 | } | |
2471 | siocb->scm->creds = *NETLINK_CREDS(skb); | |
188ccb55 | 2472 | if (flags & MSG_TRUNC) |
68d6ac6d | 2473 | copied = data_skb->len; |
daa3766e | 2474 | |
1da177e4 LT |
2475 | skb_free_datagram(sk, skb); |
2476 | ||
16b304f3 PS |
2477 | if (nlk->cb_running && |
2478 | atomic_read(&sk->sk_rmem_alloc) <= sk->sk_rcvbuf / 2) { | |
b44d211e AV |
2479 | ret = netlink_dump(sk); |
2480 | if (ret) { | |
ac30ef83 | 2481 | sk->sk_err = -ret; |
b44d211e AV |
2482 | sk->sk_error_report(sk); |
2483 | } | |
2484 | } | |
1da177e4 LT |
2485 | |
2486 | scm_recv(sock, msg, siocb->scm, flags); | |
1da177e4 LT |
2487 | out: |
2488 | netlink_rcv_wake(sk); | |
2489 | return err ? : copied; | |
2490 | } | |
2491 | ||
676d2369 | 2492 | static void netlink_data_ready(struct sock *sk) |
1da177e4 | 2493 | { |
cd40b7d3 | 2494 | BUG(); |
1da177e4 LT |
2495 | } |
2496 | ||
2497 | /* | |
746fac4d | 2498 | * We export these functions to other modules. They provide a |
1da177e4 LT |
2499 | * complete set of kernel non-blocking support for message |
2500 | * queueing. | |
2501 | */ | |
2502 | ||
2503 | struct sock * | |
9f00d977 PNA |
2504 | __netlink_kernel_create(struct net *net, int unit, struct module *module, |
2505 | struct netlink_kernel_cfg *cfg) | |
1da177e4 LT |
2506 | { |
2507 | struct socket *sock; | |
2508 | struct sock *sk; | |
77247bbb | 2509 | struct netlink_sock *nlk; |
5c398dc8 | 2510 | struct listeners *listeners = NULL; |
a31f2d17 PNA |
2511 | struct mutex *cb_mutex = cfg ? cfg->cb_mutex : NULL; |
2512 | unsigned int groups; | |
1da177e4 | 2513 | |
fab2caf6 | 2514 | BUG_ON(!nl_table); |
1da177e4 | 2515 | |
6ac552fd | 2516 | if (unit < 0 || unit >= MAX_LINKS) |
1da177e4 LT |
2517 | return NULL; |
2518 | ||
2519 | if (sock_create_lite(PF_NETLINK, SOCK_DGRAM, unit, &sock)) | |
2520 | return NULL; | |
2521 | ||
23fe1866 PE |
2522 | /* |
2523 | * We have to just have a reference on the net from sk, but don't | |
2524 | * get_net it. Besides, we cannot get and then put the net here. | |
2525 | * So we create one inside init_net and the move it to net. | |
2526 | */ | |
2527 | ||
2528 | if (__netlink_create(&init_net, sock, cb_mutex, unit) < 0) | |
2529 | goto out_sock_release_nosk; | |
2530 | ||
2531 | sk = sock->sk; | |
edf02087 | 2532 | sk_change_net(sk, net); |
4fdb3bb7 | 2533 | |
a31f2d17 | 2534 | if (!cfg || cfg->groups < 32) |
4277a083 | 2535 | groups = 32; |
a31f2d17 PNA |
2536 | else |
2537 | groups = cfg->groups; | |
4277a083 | 2538 | |
5c398dc8 | 2539 | listeners = kzalloc(sizeof(*listeners) + NLGRPSZ(groups), GFP_KERNEL); |
4277a083 PM |
2540 | if (!listeners) |
2541 | goto out_sock_release; | |
2542 | ||
1da177e4 | 2543 | sk->sk_data_ready = netlink_data_ready; |
a31f2d17 PNA |
2544 | if (cfg && cfg->input) |
2545 | nlk_sk(sk)->netlink_rcv = cfg->input; | |
1da177e4 | 2546 | |
b4b51029 | 2547 | if (netlink_insert(sk, net, 0)) |
77247bbb | 2548 | goto out_sock_release; |
4fdb3bb7 | 2549 | |
77247bbb PM |
2550 | nlk = nlk_sk(sk); |
2551 | nlk->flags |= NETLINK_KERNEL_SOCKET; | |
4fdb3bb7 | 2552 | |
4fdb3bb7 | 2553 | netlink_table_grab(); |
b4b51029 EB |
2554 | if (!nl_table[unit].registered) { |
2555 | nl_table[unit].groups = groups; | |
5c398dc8 | 2556 | rcu_assign_pointer(nl_table[unit].listeners, listeners); |
b4b51029 EB |
2557 | nl_table[unit].cb_mutex = cb_mutex; |
2558 | nl_table[unit].module = module; | |
9785e10a PNA |
2559 | if (cfg) { |
2560 | nl_table[unit].bind = cfg->bind; | |
2561 | nl_table[unit].flags = cfg->flags; | |
da12c90e G |
2562 | if (cfg->compare) |
2563 | nl_table[unit].compare = cfg->compare; | |
9785e10a | 2564 | } |
b4b51029 | 2565 | nl_table[unit].registered = 1; |
f937f1f4 JJ |
2566 | } else { |
2567 | kfree(listeners); | |
869e58f8 | 2568 | nl_table[unit].registered++; |
b4b51029 | 2569 | } |
4fdb3bb7 | 2570 | netlink_table_ungrab(); |
77247bbb PM |
2571 | return sk; |
2572 | ||
4fdb3bb7 | 2573 | out_sock_release: |
4277a083 | 2574 | kfree(listeners); |
9dfbec1f | 2575 | netlink_kernel_release(sk); |
23fe1866 PE |
2576 | return NULL; |
2577 | ||
2578 | out_sock_release_nosk: | |
4fdb3bb7 | 2579 | sock_release(sock); |
77247bbb | 2580 | return NULL; |
1da177e4 | 2581 | } |
9f00d977 | 2582 | EXPORT_SYMBOL(__netlink_kernel_create); |
b7c6ba6e DL |
2583 | |
2584 | void | |
2585 | netlink_kernel_release(struct sock *sk) | |
2586 | { | |
edf02087 | 2587 | sk_release_kernel(sk); |
b7c6ba6e DL |
2588 | } |
2589 | EXPORT_SYMBOL(netlink_kernel_release); | |
2590 | ||
d136f1bd | 2591 | int __netlink_change_ngroups(struct sock *sk, unsigned int groups) |
b4ff4f04 | 2592 | { |
5c398dc8 | 2593 | struct listeners *new, *old; |
b4ff4f04 | 2594 | struct netlink_table *tbl = &nl_table[sk->sk_protocol]; |
b4ff4f04 JB |
2595 | |
2596 | if (groups < 32) | |
2597 | groups = 32; | |
2598 | ||
b4ff4f04 | 2599 | if (NLGRPSZ(tbl->groups) < NLGRPSZ(groups)) { |
5c398dc8 ED |
2600 | new = kzalloc(sizeof(*new) + NLGRPSZ(groups), GFP_ATOMIC); |
2601 | if (!new) | |
d136f1bd | 2602 | return -ENOMEM; |
6d772ac5 | 2603 | old = nl_deref_protected(tbl->listeners); |
5c398dc8 ED |
2604 | memcpy(new->masks, old->masks, NLGRPSZ(tbl->groups)); |
2605 | rcu_assign_pointer(tbl->listeners, new); | |
2606 | ||
37b6b935 | 2607 | kfree_rcu(old, rcu); |
b4ff4f04 JB |
2608 | } |
2609 | tbl->groups = groups; | |
2610 | ||
d136f1bd JB |
2611 | return 0; |
2612 | } | |
2613 | ||
2614 | /** | |
2615 | * netlink_change_ngroups - change number of multicast groups | |
2616 | * | |
2617 | * This changes the number of multicast groups that are available | |
2618 | * on a certain netlink family. Note that it is not possible to | |
2619 | * change the number of groups to below 32. Also note that it does | |
2620 | * not implicitly call netlink_clear_multicast_users() when the | |
2621 | * number of groups is reduced. | |
2622 | * | |
2623 | * @sk: The kernel netlink socket, as returned by netlink_kernel_create(). | |
2624 | * @groups: The new number of groups. | |
2625 | */ | |
2626 | int netlink_change_ngroups(struct sock *sk, unsigned int groups) | |
2627 | { | |
2628 | int err; | |
2629 | ||
2630 | netlink_table_grab(); | |
2631 | err = __netlink_change_ngroups(sk, groups); | |
b4ff4f04 | 2632 | netlink_table_ungrab(); |
d136f1bd | 2633 | |
b4ff4f04 JB |
2634 | return err; |
2635 | } | |
b4ff4f04 | 2636 | |
b8273570 JB |
2637 | void __netlink_clear_multicast_users(struct sock *ksk, unsigned int group) |
2638 | { | |
2639 | struct sock *sk; | |
b8273570 JB |
2640 | struct netlink_table *tbl = &nl_table[ksk->sk_protocol]; |
2641 | ||
b67bfe0d | 2642 | sk_for_each_bound(sk, &tbl->mc_list) |
b8273570 JB |
2643 | netlink_update_socket_mc(nlk_sk(sk), group, 0); |
2644 | } | |
2645 | ||
a46621a3 | 2646 | struct nlmsghdr * |
15e47304 | 2647 | __nlmsg_put(struct sk_buff *skb, u32 portid, u32 seq, int type, int len, int flags) |
a46621a3 DV |
2648 | { |
2649 | struct nlmsghdr *nlh; | |
573ce260 | 2650 | int size = nlmsg_msg_size(len); |
a46621a3 | 2651 | |
23b45672 | 2652 | nlh = (struct nlmsghdr *)skb_put(skb, NLMSG_ALIGN(size)); |
a46621a3 DV |
2653 | nlh->nlmsg_type = type; |
2654 | nlh->nlmsg_len = size; | |
2655 | nlh->nlmsg_flags = flags; | |
15e47304 | 2656 | nlh->nlmsg_pid = portid; |
a46621a3 DV |
2657 | nlh->nlmsg_seq = seq; |
2658 | if (!__builtin_constant_p(size) || NLMSG_ALIGN(size) - size != 0) | |
573ce260 | 2659 | memset(nlmsg_data(nlh) + len, 0, NLMSG_ALIGN(size) - size); |
a46621a3 DV |
2660 | return nlh; |
2661 | } | |
2662 | EXPORT_SYMBOL(__nlmsg_put); | |
2663 | ||
1da177e4 LT |
2664 | /* |
2665 | * It looks a bit ugly. | |
2666 | * It would be better to create kernel thread. | |
2667 | */ | |
2668 | ||
2669 | static int netlink_dump(struct sock *sk) | |
2670 | { | |
2671 | struct netlink_sock *nlk = nlk_sk(sk); | |
2672 | struct netlink_callback *cb; | |
c7ac8679 | 2673 | struct sk_buff *skb = NULL; |
1da177e4 | 2674 | struct nlmsghdr *nlh; |
bf8b79e4 | 2675 | int len, err = -ENOBUFS; |
c7ac8679 | 2676 | int alloc_size; |
1da177e4 | 2677 | |
af65bdfc | 2678 | mutex_lock(nlk->cb_mutex); |
16b304f3 | 2679 | if (!nlk->cb_running) { |
bf8b79e4 TG |
2680 | err = -EINVAL; |
2681 | goto errout_skb; | |
1da177e4 LT |
2682 | } |
2683 | ||
16b304f3 | 2684 | cb = &nlk->cb; |
c7ac8679 GR |
2685 | alloc_size = max_t(int, cb->min_dump_alloc, NLMSG_GOODSIZE); |
2686 | ||
f9c22888 PM |
2687 | if (!netlink_rx_is_mmaped(sk) && |
2688 | atomic_read(&sk->sk_rmem_alloc) >= sk->sk_rcvbuf) | |
2689 | goto errout_skb; | |
9063e21f ED |
2690 | |
2691 | /* NLMSG_GOODSIZE is small to avoid high order allocations being | |
2692 | * required, but it makes sense to _attempt_ a 16K bytes allocation | |
2693 | * to reduce number of system calls on dump operations, if user | |
2694 | * ever provided a big enough buffer. | |
2695 | */ | |
2696 | if (alloc_size < nlk->max_recvmsg_len) { | |
2697 | skb = netlink_alloc_skb(sk, | |
2698 | nlk->max_recvmsg_len, | |
2699 | nlk->portid, | |
2700 | GFP_KERNEL | | |
2701 | __GFP_NOWARN | | |
2702 | __GFP_NORETRY); | |
2703 | /* available room should be exact amount to avoid MSG_TRUNC */ | |
2704 | if (skb) | |
2705 | skb_reserve(skb, skb_tailroom(skb) - | |
2706 | nlk->max_recvmsg_len); | |
2707 | } | |
2708 | if (!skb) | |
2709 | skb = netlink_alloc_skb(sk, alloc_size, nlk->portid, | |
2710 | GFP_KERNEL); | |
c7ac8679 | 2711 | if (!skb) |
c63d6ea3 | 2712 | goto errout_skb; |
f9c22888 | 2713 | netlink_skb_set_owner_r(skb, sk); |
c7ac8679 | 2714 | |
1da177e4 LT |
2715 | len = cb->dump(skb, cb); |
2716 | ||
2717 | if (len > 0) { | |
af65bdfc | 2718 | mutex_unlock(nlk->cb_mutex); |
b1153f29 SH |
2719 | |
2720 | if (sk_filter(sk, skb)) | |
2721 | kfree_skb(skb); | |
4a7e7c2a ED |
2722 | else |
2723 | __netlink_sendskb(sk, skb); | |
1da177e4 LT |
2724 | return 0; |
2725 | } | |
2726 | ||
bf8b79e4 TG |
2727 | nlh = nlmsg_put_answer(skb, cb, NLMSG_DONE, sizeof(len), NLM_F_MULTI); |
2728 | if (!nlh) | |
2729 | goto errout_skb; | |
2730 | ||
670dc283 JB |
2731 | nl_dump_check_consistent(cb, nlh); |
2732 | ||
bf8b79e4 TG |
2733 | memcpy(nlmsg_data(nlh), &len, sizeof(len)); |
2734 | ||
b1153f29 SH |
2735 | if (sk_filter(sk, skb)) |
2736 | kfree_skb(skb); | |
4a7e7c2a ED |
2737 | else |
2738 | __netlink_sendskb(sk, skb); | |
1da177e4 | 2739 | |
a8f74b22 TG |
2740 | if (cb->done) |
2741 | cb->done(cb); | |
1da177e4 | 2742 | |
16b304f3 PS |
2743 | nlk->cb_running = false; |
2744 | mutex_unlock(nlk->cb_mutex); | |
6dc878a8 | 2745 | module_put(cb->module); |
16b304f3 | 2746 | consume_skb(cb->skb); |
1da177e4 | 2747 | return 0; |
1797754e | 2748 | |
bf8b79e4 | 2749 | errout_skb: |
af65bdfc | 2750 | mutex_unlock(nlk->cb_mutex); |
bf8b79e4 | 2751 | kfree_skb(skb); |
bf8b79e4 | 2752 | return err; |
1da177e4 LT |
2753 | } |
2754 | ||
6dc878a8 G |
2755 | int __netlink_dump_start(struct sock *ssk, struct sk_buff *skb, |
2756 | const struct nlmsghdr *nlh, | |
2757 | struct netlink_dump_control *control) | |
1da177e4 LT |
2758 | { |
2759 | struct netlink_callback *cb; | |
2760 | struct sock *sk; | |
2761 | struct netlink_sock *nlk; | |
b44d211e | 2762 | int ret; |
1da177e4 | 2763 | |
f9c22888 PM |
2764 | /* Memory mapped dump requests need to be copied to avoid looping |
2765 | * on the pending state in netlink_mmap_sendmsg() while the CB hold | |
2766 | * a reference to the skb. | |
2767 | */ | |
2768 | if (netlink_skb_is_mmaped(skb)) { | |
2769 | skb = skb_copy(skb, GFP_KERNEL); | |
16b304f3 | 2770 | if (skb == NULL) |
f9c22888 | 2771 | return -ENOBUFS; |
f9c22888 PM |
2772 | } else |
2773 | atomic_inc(&skb->users); | |
2774 | ||
15e47304 | 2775 | sk = netlink_lookup(sock_net(ssk), ssk->sk_protocol, NETLINK_CB(skb).portid); |
1da177e4 | 2776 | if (sk == NULL) { |
16b304f3 PS |
2777 | ret = -ECONNREFUSED; |
2778 | goto error_free; | |
1da177e4 | 2779 | } |
6dc878a8 | 2780 | |
16b304f3 | 2781 | nlk = nlk_sk(sk); |
af65bdfc | 2782 | mutex_lock(nlk->cb_mutex); |
6dc878a8 | 2783 | /* A dump is in progress... */ |
16b304f3 | 2784 | if (nlk->cb_running) { |
6dc878a8 | 2785 | ret = -EBUSY; |
16b304f3 | 2786 | goto error_unlock; |
1da177e4 | 2787 | } |
6dc878a8 | 2788 | /* add reference of module which cb->dump belongs to */ |
16b304f3 | 2789 | if (!try_module_get(control->module)) { |
6dc878a8 | 2790 | ret = -EPROTONOSUPPORT; |
16b304f3 | 2791 | goto error_unlock; |
6dc878a8 G |
2792 | } |
2793 | ||
16b304f3 PS |
2794 | cb = &nlk->cb; |
2795 | memset(cb, 0, sizeof(*cb)); | |
2796 | cb->dump = control->dump; | |
2797 | cb->done = control->done; | |
2798 | cb->nlh = nlh; | |
2799 | cb->data = control->data; | |
2800 | cb->module = control->module; | |
2801 | cb->min_dump_alloc = control->min_dump_alloc; | |
2802 | cb->skb = skb; | |
2803 | ||
2804 | nlk->cb_running = true; | |
2805 | ||
af65bdfc | 2806 | mutex_unlock(nlk->cb_mutex); |
1da177e4 | 2807 | |
b44d211e | 2808 | ret = netlink_dump(sk); |
1da177e4 | 2809 | sock_put(sk); |
5c58298c | 2810 | |
b44d211e AV |
2811 | if (ret) |
2812 | return ret; | |
2813 | ||
5c58298c DL |
2814 | /* We successfully started a dump, by returning -EINTR we |
2815 | * signal not to send ACK even if it was requested. | |
2816 | */ | |
2817 | return -EINTR; | |
16b304f3 PS |
2818 | |
2819 | error_unlock: | |
2820 | sock_put(sk); | |
2821 | mutex_unlock(nlk->cb_mutex); | |
2822 | error_free: | |
2823 | kfree_skb(skb); | |
2824 | return ret; | |
1da177e4 | 2825 | } |
6dc878a8 | 2826 | EXPORT_SYMBOL(__netlink_dump_start); |
1da177e4 LT |
2827 | |
2828 | void netlink_ack(struct sk_buff *in_skb, struct nlmsghdr *nlh, int err) | |
2829 | { | |
2830 | struct sk_buff *skb; | |
2831 | struct nlmsghdr *rep; | |
2832 | struct nlmsgerr *errmsg; | |
339bf98f | 2833 | size_t payload = sizeof(*errmsg); |
1da177e4 | 2834 | |
339bf98f TG |
2835 | /* error messages get the original request appened */ |
2836 | if (err) | |
2837 | payload += nlmsg_len(nlh); | |
1da177e4 | 2838 | |
f9c22888 PM |
2839 | skb = netlink_alloc_skb(in_skb->sk, nlmsg_total_size(payload), |
2840 | NETLINK_CB(in_skb).portid, GFP_KERNEL); | |
1da177e4 LT |
2841 | if (!skb) { |
2842 | struct sock *sk; | |
2843 | ||
3b1e0a65 | 2844 | sk = netlink_lookup(sock_net(in_skb->sk), |
b4b51029 | 2845 | in_skb->sk->sk_protocol, |
15e47304 | 2846 | NETLINK_CB(in_skb).portid); |
1da177e4 LT |
2847 | if (sk) { |
2848 | sk->sk_err = ENOBUFS; | |
2849 | sk->sk_error_report(sk); | |
2850 | sock_put(sk); | |
2851 | } | |
2852 | return; | |
2853 | } | |
2854 | ||
15e47304 | 2855 | rep = __nlmsg_put(skb, NETLINK_CB(in_skb).portid, nlh->nlmsg_seq, |
5dba93ae | 2856 | NLMSG_ERROR, payload, 0); |
bf8b79e4 | 2857 | errmsg = nlmsg_data(rep); |
1da177e4 | 2858 | errmsg->error = err; |
bf8b79e4 | 2859 | memcpy(&errmsg->msg, nlh, err ? nlh->nlmsg_len : sizeof(*nlh)); |
15e47304 | 2860 | netlink_unicast(in_skb->sk, skb, NETLINK_CB(in_skb).portid, MSG_DONTWAIT); |
1da177e4 | 2861 | } |
6ac552fd | 2862 | EXPORT_SYMBOL(netlink_ack); |
1da177e4 | 2863 | |
cd40b7d3 | 2864 | int netlink_rcv_skb(struct sk_buff *skb, int (*cb)(struct sk_buff *, |
1d00a4eb | 2865 | struct nlmsghdr *)) |
82ace47a | 2866 | { |
82ace47a TG |
2867 | struct nlmsghdr *nlh; |
2868 | int err; | |
2869 | ||
2870 | while (skb->len >= nlmsg_total_size(0)) { | |
cd40b7d3 DL |
2871 | int msglen; |
2872 | ||
b529ccf2 | 2873 | nlh = nlmsg_hdr(skb); |
d35b6856 | 2874 | err = 0; |
82ace47a | 2875 | |
ad8e4b75 | 2876 | if (nlh->nlmsg_len < NLMSG_HDRLEN || skb->len < nlh->nlmsg_len) |
82ace47a TG |
2877 | return 0; |
2878 | ||
d35b6856 TG |
2879 | /* Only requests are handled by the kernel */ |
2880 | if (!(nlh->nlmsg_flags & NLM_F_REQUEST)) | |
5c58298c | 2881 | goto ack; |
45e7ae7f TG |
2882 | |
2883 | /* Skip control messages */ | |
2884 | if (nlh->nlmsg_type < NLMSG_MIN_TYPE) | |
5c58298c | 2885 | goto ack; |
d35b6856 | 2886 | |
1d00a4eb | 2887 | err = cb(skb, nlh); |
5c58298c DL |
2888 | if (err == -EINTR) |
2889 | goto skip; | |
2890 | ||
2891 | ack: | |
d35b6856 | 2892 | if (nlh->nlmsg_flags & NLM_F_ACK || err) |
82ace47a | 2893 | netlink_ack(skb, nlh, err); |
82ace47a | 2894 | |
5c58298c | 2895 | skip: |
6ac552fd | 2896 | msglen = NLMSG_ALIGN(nlh->nlmsg_len); |
cd40b7d3 DL |
2897 | if (msglen > skb->len) |
2898 | msglen = skb->len; | |
2899 | skb_pull(skb, msglen); | |
82ace47a TG |
2900 | } |
2901 | ||
2902 | return 0; | |
2903 | } | |
6ac552fd | 2904 | EXPORT_SYMBOL(netlink_rcv_skb); |
82ace47a | 2905 | |
d387f6ad TG |
2906 | /** |
2907 | * nlmsg_notify - send a notification netlink message | |
2908 | * @sk: netlink socket to use | |
2909 | * @skb: notification message | |
15e47304 | 2910 | * @portid: destination netlink portid for reports or 0 |
d387f6ad TG |
2911 | * @group: destination multicast group or 0 |
2912 | * @report: 1 to report back, 0 to disable | |
2913 | * @flags: allocation flags | |
2914 | */ | |
15e47304 | 2915 | int nlmsg_notify(struct sock *sk, struct sk_buff *skb, u32 portid, |
d387f6ad TG |
2916 | unsigned int group, int report, gfp_t flags) |
2917 | { | |
2918 | int err = 0; | |
2919 | ||
2920 | if (group) { | |
15e47304 | 2921 | int exclude_portid = 0; |
d387f6ad TG |
2922 | |
2923 | if (report) { | |
2924 | atomic_inc(&skb->users); | |
15e47304 | 2925 | exclude_portid = portid; |
d387f6ad TG |
2926 | } |
2927 | ||
1ce85fe4 PNA |
2928 | /* errors reported via destination sk->sk_err, but propagate |
2929 | * delivery errors if NETLINK_BROADCAST_ERROR flag is set */ | |
15e47304 | 2930 | err = nlmsg_multicast(sk, skb, exclude_portid, group, flags); |
d387f6ad TG |
2931 | } |
2932 | ||
1ce85fe4 PNA |
2933 | if (report) { |
2934 | int err2; | |
2935 | ||
15e47304 | 2936 | err2 = nlmsg_unicast(sk, skb, portid); |
1ce85fe4 PNA |
2937 | if (!err || err == -ESRCH) |
2938 | err = err2; | |
2939 | } | |
d387f6ad TG |
2940 | |
2941 | return err; | |
2942 | } | |
6ac552fd | 2943 | EXPORT_SYMBOL(nlmsg_notify); |
d387f6ad | 2944 | |
1da177e4 LT |
2945 | #ifdef CONFIG_PROC_FS |
2946 | struct nl_seq_iter { | |
e372c414 | 2947 | struct seq_net_private p; |
1da177e4 LT |
2948 | int link; |
2949 | int hash_idx; | |
2950 | }; | |
2951 | ||
2952 | static struct sock *netlink_seq_socket_idx(struct seq_file *seq, loff_t pos) | |
2953 | { | |
2954 | struct nl_seq_iter *iter = seq->private; | |
2955 | int i, j; | |
2956 | struct sock *s; | |
1da177e4 LT |
2957 | loff_t off = 0; |
2958 | ||
6ac552fd | 2959 | for (i = 0; i < MAX_LINKS; i++) { |
15e47304 | 2960 | struct nl_portid_hash *hash = &nl_table[i].hash; |
1da177e4 LT |
2961 | |
2962 | for (j = 0; j <= hash->mask; j++) { | |
b67bfe0d | 2963 | sk_for_each(s, &hash->table[j]) { |
1218854a | 2964 | if (sock_net(s) != seq_file_net(seq)) |
b4b51029 | 2965 | continue; |
1da177e4 LT |
2966 | if (off == pos) { |
2967 | iter->link = i; | |
2968 | iter->hash_idx = j; | |
2969 | return s; | |
2970 | } | |
2971 | ++off; | |
2972 | } | |
2973 | } | |
2974 | } | |
2975 | return NULL; | |
2976 | } | |
2977 | ||
2978 | static void *netlink_seq_start(struct seq_file *seq, loff_t *pos) | |
9a429c49 | 2979 | __acquires(nl_table_lock) |
1da177e4 LT |
2980 | { |
2981 | read_lock(&nl_table_lock); | |
2982 | return *pos ? netlink_seq_socket_idx(seq, *pos - 1) : SEQ_START_TOKEN; | |
2983 | } | |
2984 | ||
2985 | static void *netlink_seq_next(struct seq_file *seq, void *v, loff_t *pos) | |
2986 | { | |
2987 | struct sock *s; | |
2988 | struct nl_seq_iter *iter; | |
da12c90e | 2989 | struct net *net; |
1da177e4 LT |
2990 | int i, j; |
2991 | ||
2992 | ++*pos; | |
2993 | ||
2994 | if (v == SEQ_START_TOKEN) | |
2995 | return netlink_seq_socket_idx(seq, 0); | |
746fac4d | 2996 | |
da12c90e | 2997 | net = seq_file_net(seq); |
b4b51029 EB |
2998 | iter = seq->private; |
2999 | s = v; | |
3000 | do { | |
3001 | s = sk_next(s); | |
da12c90e | 3002 | } while (s && !nl_table[s->sk_protocol].compare(net, s)); |
1da177e4 LT |
3003 | if (s) |
3004 | return s; | |
3005 | ||
1da177e4 LT |
3006 | i = iter->link; |
3007 | j = iter->hash_idx + 1; | |
3008 | ||
3009 | do { | |
15e47304 | 3010 | struct nl_portid_hash *hash = &nl_table[i].hash; |
1da177e4 LT |
3011 | |
3012 | for (; j <= hash->mask; j++) { | |
3013 | s = sk_head(&hash->table[j]); | |
da12c90e G |
3014 | |
3015 | while (s && !nl_table[s->sk_protocol].compare(net, s)) | |
b4b51029 | 3016 | s = sk_next(s); |
1da177e4 LT |
3017 | if (s) { |
3018 | iter->link = i; | |
3019 | iter->hash_idx = j; | |
3020 | return s; | |
3021 | } | |
3022 | } | |
3023 | ||
3024 | j = 0; | |
3025 | } while (++i < MAX_LINKS); | |
3026 | ||
3027 | return NULL; | |
3028 | } | |
3029 | ||
3030 | static void netlink_seq_stop(struct seq_file *seq, void *v) | |
9a429c49 | 3031 | __releases(nl_table_lock) |
1da177e4 LT |
3032 | { |
3033 | read_unlock(&nl_table_lock); | |
3034 | } | |
3035 | ||
3036 | ||
3037 | static int netlink_seq_show(struct seq_file *seq, void *v) | |
3038 | { | |
658cb354 | 3039 | if (v == SEQ_START_TOKEN) { |
1da177e4 LT |
3040 | seq_puts(seq, |
3041 | "sk Eth Pid Groups " | |
cf0aa4e0 | 3042 | "Rmem Wmem Dump Locks Drops Inode\n"); |
658cb354 | 3043 | } else { |
1da177e4 LT |
3044 | struct sock *s = v; |
3045 | struct netlink_sock *nlk = nlk_sk(s); | |
3046 | ||
16b304f3 | 3047 | seq_printf(seq, "%pK %-3d %-6u %08x %-8d %-8d %d %-8d %-8d %-8lu\n", |
1da177e4 LT |
3048 | s, |
3049 | s->sk_protocol, | |
15e47304 | 3050 | nlk->portid, |
513c2500 | 3051 | nlk->groups ? (u32)nlk->groups[0] : 0, |
31e6d363 ED |
3052 | sk_rmem_alloc_get(s), |
3053 | sk_wmem_alloc_get(s), | |
16b304f3 | 3054 | nlk->cb_running, |
38938bfe | 3055 | atomic_read(&s->sk_refcnt), |
cf0aa4e0 MY |
3056 | atomic_read(&s->sk_drops), |
3057 | sock_i_ino(s) | |
1da177e4 LT |
3058 | ); |
3059 | ||
3060 | } | |
3061 | return 0; | |
3062 | } | |
3063 | ||
56b3d975 | 3064 | static const struct seq_operations netlink_seq_ops = { |
1da177e4 LT |
3065 | .start = netlink_seq_start, |
3066 | .next = netlink_seq_next, | |
3067 | .stop = netlink_seq_stop, | |
3068 | .show = netlink_seq_show, | |
3069 | }; | |
3070 | ||
3071 | ||
3072 | static int netlink_seq_open(struct inode *inode, struct file *file) | |
3073 | { | |
e372c414 DL |
3074 | return seq_open_net(inode, file, &netlink_seq_ops, |
3075 | sizeof(struct nl_seq_iter)); | |
b4b51029 EB |
3076 | } |
3077 | ||
da7071d7 | 3078 | static const struct file_operations netlink_seq_fops = { |
1da177e4 LT |
3079 | .owner = THIS_MODULE, |
3080 | .open = netlink_seq_open, | |
3081 | .read = seq_read, | |
3082 | .llseek = seq_lseek, | |
e372c414 | 3083 | .release = seq_release_net, |
1da177e4 LT |
3084 | }; |
3085 | ||
3086 | #endif | |
3087 | ||
3088 | int netlink_register_notifier(struct notifier_block *nb) | |
3089 | { | |
e041c683 | 3090 | return atomic_notifier_chain_register(&netlink_chain, nb); |
1da177e4 | 3091 | } |
6ac552fd | 3092 | EXPORT_SYMBOL(netlink_register_notifier); |
1da177e4 LT |
3093 | |
3094 | int netlink_unregister_notifier(struct notifier_block *nb) | |
3095 | { | |
e041c683 | 3096 | return atomic_notifier_chain_unregister(&netlink_chain, nb); |
1da177e4 | 3097 | } |
6ac552fd | 3098 | EXPORT_SYMBOL(netlink_unregister_notifier); |
746fac4d | 3099 | |
90ddc4f0 | 3100 | static const struct proto_ops netlink_ops = { |
1da177e4 LT |
3101 | .family = PF_NETLINK, |
3102 | .owner = THIS_MODULE, | |
3103 | .release = netlink_release, | |
3104 | .bind = netlink_bind, | |
3105 | .connect = netlink_connect, | |
3106 | .socketpair = sock_no_socketpair, | |
3107 | .accept = sock_no_accept, | |
3108 | .getname = netlink_getname, | |
9652e931 | 3109 | .poll = netlink_poll, |
1da177e4 LT |
3110 | .ioctl = sock_no_ioctl, |
3111 | .listen = sock_no_listen, | |
3112 | .shutdown = sock_no_shutdown, | |
9a4595bc PM |
3113 | .setsockopt = netlink_setsockopt, |
3114 | .getsockopt = netlink_getsockopt, | |
1da177e4 LT |
3115 | .sendmsg = netlink_sendmsg, |
3116 | .recvmsg = netlink_recvmsg, | |
ccdfcc39 | 3117 | .mmap = netlink_mmap, |
1da177e4 LT |
3118 | .sendpage = sock_no_sendpage, |
3119 | }; | |
3120 | ||
ec1b4cf7 | 3121 | static const struct net_proto_family netlink_family_ops = { |
1da177e4 LT |
3122 | .family = PF_NETLINK, |
3123 | .create = netlink_create, | |
3124 | .owner = THIS_MODULE, /* for consistency 8) */ | |
3125 | }; | |
3126 | ||
4665079c | 3127 | static int __net_init netlink_net_init(struct net *net) |
b4b51029 EB |
3128 | { |
3129 | #ifdef CONFIG_PROC_FS | |
d4beaa66 | 3130 | if (!proc_create("netlink", 0, net->proc_net, &netlink_seq_fops)) |
b4b51029 EB |
3131 | return -ENOMEM; |
3132 | #endif | |
3133 | return 0; | |
3134 | } | |
3135 | ||
4665079c | 3136 | static void __net_exit netlink_net_exit(struct net *net) |
b4b51029 EB |
3137 | { |
3138 | #ifdef CONFIG_PROC_FS | |
ece31ffd | 3139 | remove_proc_entry("netlink", net->proc_net); |
b4b51029 EB |
3140 | #endif |
3141 | } | |
3142 | ||
b963ea89 DM |
3143 | static void __init netlink_add_usersock_entry(void) |
3144 | { | |
5c398dc8 | 3145 | struct listeners *listeners; |
b963ea89 DM |
3146 | int groups = 32; |
3147 | ||
5c398dc8 | 3148 | listeners = kzalloc(sizeof(*listeners) + NLGRPSZ(groups), GFP_KERNEL); |
b963ea89 | 3149 | if (!listeners) |
5c398dc8 | 3150 | panic("netlink_add_usersock_entry: Cannot allocate listeners\n"); |
b963ea89 DM |
3151 | |
3152 | netlink_table_grab(); | |
3153 | ||
3154 | nl_table[NETLINK_USERSOCK].groups = groups; | |
5c398dc8 | 3155 | rcu_assign_pointer(nl_table[NETLINK_USERSOCK].listeners, listeners); |
b963ea89 DM |
3156 | nl_table[NETLINK_USERSOCK].module = THIS_MODULE; |
3157 | nl_table[NETLINK_USERSOCK].registered = 1; | |
9785e10a | 3158 | nl_table[NETLINK_USERSOCK].flags = NL_CFG_F_NONROOT_SEND; |
b963ea89 DM |
3159 | |
3160 | netlink_table_ungrab(); | |
3161 | } | |
3162 | ||
022cbae6 | 3163 | static struct pernet_operations __net_initdata netlink_net_ops = { |
b4b51029 EB |
3164 | .init = netlink_net_init, |
3165 | .exit = netlink_net_exit, | |
3166 | }; | |
3167 | ||
1da177e4 LT |
3168 | static int __init netlink_proto_init(void) |
3169 | { | |
1da177e4 | 3170 | int i; |
26ff5ddc | 3171 | unsigned long limit; |
1da177e4 LT |
3172 | unsigned int order; |
3173 | int err = proto_register(&netlink_proto, 0); | |
3174 | ||
3175 | if (err != 0) | |
3176 | goto out; | |
3177 | ||
fab25745 | 3178 | BUILD_BUG_ON(sizeof(struct netlink_skb_parms) > FIELD_SIZEOF(struct sk_buff, cb)); |
1da177e4 | 3179 | |
0da974f4 | 3180 | nl_table = kcalloc(MAX_LINKS, sizeof(*nl_table), GFP_KERNEL); |
fab2caf6 AM |
3181 | if (!nl_table) |
3182 | goto panic; | |
1da177e4 | 3183 | |
4481374c JB |
3184 | if (totalram_pages >= (128 * 1024)) |
3185 | limit = totalram_pages >> (21 - PAGE_SHIFT); | |
1da177e4 | 3186 | else |
4481374c | 3187 | limit = totalram_pages >> (23 - PAGE_SHIFT); |
1da177e4 | 3188 | |
26ff5ddc DC |
3189 | order = get_bitmask_order(limit) - 1 + PAGE_SHIFT; |
3190 | limit = (1UL << order) / sizeof(struct hlist_head); | |
3191 | order = get_bitmask_order(min(limit, (unsigned long)UINT_MAX)) - 1; | |
1da177e4 LT |
3192 | |
3193 | for (i = 0; i < MAX_LINKS; i++) { | |
15e47304 | 3194 | struct nl_portid_hash *hash = &nl_table[i].hash; |
1da177e4 | 3195 | |
15e47304 | 3196 | hash->table = nl_portid_hash_zalloc(1 * sizeof(*hash->table)); |
1da177e4 LT |
3197 | if (!hash->table) { |
3198 | while (i-- > 0) | |
15e47304 | 3199 | nl_portid_hash_free(nl_table[i].hash.table, |
1da177e4 LT |
3200 | 1 * sizeof(*hash->table)); |
3201 | kfree(nl_table); | |
fab2caf6 | 3202 | goto panic; |
1da177e4 | 3203 | } |
1da177e4 LT |
3204 | hash->max_shift = order; |
3205 | hash->shift = 0; | |
3206 | hash->mask = 0; | |
3207 | hash->rehash_time = jiffies; | |
ca15febf G |
3208 | |
3209 | nl_table[i].compare = netlink_compare; | |
1da177e4 LT |
3210 | } |
3211 | ||
bcbde0d4 DB |
3212 | INIT_LIST_HEAD(&netlink_tap_all); |
3213 | ||
b963ea89 DM |
3214 | netlink_add_usersock_entry(); |
3215 | ||
1da177e4 | 3216 | sock_register(&netlink_family_ops); |
b4b51029 | 3217 | register_pernet_subsys(&netlink_net_ops); |
746fac4d | 3218 | /* The netlink device handler may be needed early. */ |
1da177e4 LT |
3219 | rtnetlink_init(); |
3220 | out: | |
3221 | return err; | |
fab2caf6 AM |
3222 | panic: |
3223 | panic("netlink_init: Cannot allocate nl_table\n"); | |
1da177e4 LT |
3224 | } |
3225 | ||
1da177e4 | 3226 | core_initcall(netlink_proto_init); |