Commit | Line | Data |
---|---|---|
7c657876 ACM |
1 | /* |
2 | * net/dccp/proto.c | |
3 | * | |
4 | * An implementation of the DCCP protocol | |
5 | * Arnaldo Carvalho de Melo <acme@conectiva.com.br> | |
6 | * | |
7 | * This program is free software; you can redistribute it and/or modify it | |
8 | * under the terms of the GNU General Public License version 2 as | |
9 | * published by the Free Software Foundation. | |
10 | */ | |
11 | ||
12 | #include <linux/config.h> | |
13 | #include <linux/dccp.h> | |
14 | #include <linux/module.h> | |
15 | #include <linux/types.h> | |
16 | #include <linux/sched.h> | |
17 | #include <linux/kernel.h> | |
18 | #include <linux/skbuff.h> | |
19 | #include <linux/netdevice.h> | |
20 | #include <linux/in.h> | |
21 | #include <linux/if_arp.h> | |
22 | #include <linux/init.h> | |
23 | #include <linux/random.h> | |
24 | #include <net/checksum.h> | |
25 | ||
26 | #include <net/inet_common.h> | |
27 | #include <net/ip.h> | |
28 | #include <net/protocol.h> | |
29 | #include <net/sock.h> | |
30 | #include <net/xfrm.h> | |
31 | ||
32 | #include <asm/semaphore.h> | |
33 | #include <linux/spinlock.h> | |
34 | #include <linux/timer.h> | |
35 | #include <linux/delay.h> | |
36 | #include <linux/poll.h> | |
37 | #include <linux/dccp.h> | |
38 | ||
39 | #include "ccid.h" | |
40 | #include "dccp.h" | |
41 | ||
42 | DEFINE_SNMP_STAT(struct dccp_mib, dccp_statistics); | |
43 | ||
44 | atomic_t dccp_orphan_count = ATOMIC_INIT(0); | |
45 | ||
46 | static struct net_protocol dccp_protocol = { | |
47 | .handler = dccp_v4_rcv, | |
48 | .err_handler = dccp_v4_err, | |
49 | }; | |
50 | ||
51 | const char *dccp_packet_name(const int type) | |
52 | { | |
53 | static const char *dccp_packet_names[] = { | |
54 | [DCCP_PKT_REQUEST] = "REQUEST", | |
55 | [DCCP_PKT_RESPONSE] = "RESPONSE", | |
56 | [DCCP_PKT_DATA] = "DATA", | |
57 | [DCCP_PKT_ACK] = "ACK", | |
58 | [DCCP_PKT_DATAACK] = "DATAACK", | |
59 | [DCCP_PKT_CLOSEREQ] = "CLOSEREQ", | |
60 | [DCCP_PKT_CLOSE] = "CLOSE", | |
61 | [DCCP_PKT_RESET] = "RESET", | |
62 | [DCCP_PKT_SYNC] = "SYNC", | |
63 | [DCCP_PKT_SYNCACK] = "SYNCACK", | |
64 | }; | |
65 | ||
66 | if (type >= DCCP_NR_PKT_TYPES) | |
67 | return "INVALID"; | |
68 | else | |
69 | return dccp_packet_names[type]; | |
70 | } | |
71 | ||
72 | EXPORT_SYMBOL_GPL(dccp_packet_name); | |
73 | ||
74 | const char *dccp_state_name(const int state) | |
75 | { | |
76 | static char *dccp_state_names[] = { | |
77 | [DCCP_OPEN] = "OPEN", | |
78 | [DCCP_REQUESTING] = "REQUESTING", | |
79 | [DCCP_PARTOPEN] = "PARTOPEN", | |
80 | [DCCP_LISTEN] = "LISTEN", | |
81 | [DCCP_RESPOND] = "RESPOND", | |
82 | [DCCP_CLOSING] = "CLOSING", | |
83 | [DCCP_TIME_WAIT] = "TIME_WAIT", | |
84 | [DCCP_CLOSED] = "CLOSED", | |
85 | }; | |
86 | ||
87 | if (state >= DCCP_MAX_STATES) | |
88 | return "INVALID STATE!"; | |
89 | else | |
90 | return dccp_state_names[state]; | |
91 | } | |
92 | ||
93 | EXPORT_SYMBOL_GPL(dccp_state_name); | |
94 | ||
95 | static inline int dccp_listen_start(struct sock *sk) | |
96 | { | |
97 | dccp_sk(sk)->dccps_role = DCCP_ROLE_LISTEN; | |
98 | return inet_csk_listen_start(sk, TCP_SYNQ_HSIZE); | |
99 | } | |
100 | ||
101 | int dccp_disconnect(struct sock *sk, int flags) | |
102 | { | |
103 | struct inet_connection_sock *icsk = inet_csk(sk); | |
104 | struct inet_sock *inet = inet_sk(sk); | |
105 | int err = 0; | |
106 | const int old_state = sk->sk_state; | |
107 | ||
108 | if (old_state != DCCP_CLOSED) | |
109 | dccp_set_state(sk, DCCP_CLOSED); | |
110 | ||
111 | /* ABORT function of RFC793 */ | |
112 | if (old_state == DCCP_LISTEN) { | |
113 | inet_csk_listen_stop(sk); | |
114 | /* FIXME: do the active reset thing */ | |
115 | } else if (old_state == DCCP_REQUESTING) | |
116 | sk->sk_err = ECONNRESET; | |
117 | ||
118 | dccp_clear_xmit_timers(sk); | |
119 | __skb_queue_purge(&sk->sk_receive_queue); | |
120 | if (sk->sk_send_head != NULL) { | |
121 | __kfree_skb(sk->sk_send_head); | |
122 | sk->sk_send_head = NULL; | |
123 | } | |
124 | ||
125 | inet->dport = 0; | |
126 | ||
127 | if (!(sk->sk_userlocks & SOCK_BINDADDR_LOCK)) | |
128 | inet_reset_saddr(sk); | |
129 | ||
130 | sk->sk_shutdown = 0; | |
131 | sock_reset_flag(sk, SOCK_DONE); | |
132 | ||
133 | icsk->icsk_backoff = 0; | |
134 | inet_csk_delack_init(sk); | |
135 | __sk_dst_reset(sk); | |
136 | ||
137 | BUG_TRAP(!inet->num || icsk->icsk_bind_hash); | |
138 | ||
139 | sk->sk_error_report(sk); | |
140 | return err; | |
141 | } | |
142 | ||
143 | int dccp_ioctl(struct sock *sk, int cmd, unsigned long arg) | |
144 | { | |
145 | dccp_pr_debug("entry\n"); | |
146 | return -ENOIOCTLCMD; | |
147 | } | |
148 | ||
149 | int dccp_setsockopt(struct sock *sk, int level, int optname, | |
150 | char *optval, int optlen) | |
151 | { | |
152 | dccp_pr_debug("entry\n"); | |
153 | ||
154 | if (level != SOL_DCCP) | |
155 | return ip_setsockopt(sk, level, optname, optval, optlen); | |
156 | ||
157 | return -EOPNOTSUPP; | |
158 | } | |
159 | ||
160 | int dccp_getsockopt(struct sock *sk, int level, int optname, | |
161 | char *optval, int *optlen) | |
162 | { | |
163 | dccp_pr_debug("entry\n"); | |
164 | ||
165 | if (level != SOL_DCCP) | |
166 | return ip_getsockopt(sk, level, optname, optval, optlen); | |
167 | ||
168 | return -EOPNOTSUPP; | |
169 | } | |
170 | ||
171 | int dccp_sendmsg(struct kiocb *iocb, struct sock *sk, struct msghdr *msg, | |
172 | size_t len) | |
173 | { | |
174 | const struct dccp_sock *dp = dccp_sk(sk); | |
175 | const int flags = msg->msg_flags; | |
176 | const int noblock = flags & MSG_DONTWAIT; | |
177 | struct sk_buff *skb; | |
178 | int rc, size; | |
179 | long timeo; | |
180 | ||
181 | if (len > dp->dccps_mss_cache) | |
182 | return -EMSGSIZE; | |
183 | ||
184 | lock_sock(sk); | |
27258ee5 | 185 | timeo = sock_sndtimeo(sk, noblock); |
7c657876 ACM |
186 | |
187 | /* | |
188 | * We have to use sk_stream_wait_connect here to set sk_write_pending, | |
189 | * so that the trick in dccp_rcv_request_sent_state_process. | |
190 | */ | |
191 | /* Wait for a connection to finish. */ | |
192 | if ((1 << sk->sk_state) & ~(DCCPF_OPEN | DCCPF_PARTOPEN | DCCPF_CLOSING)) | |
193 | if ((rc = sk_stream_wait_connect(sk, &timeo)) != 0) | |
27258ee5 | 194 | goto out_release; |
7c657876 ACM |
195 | |
196 | size = sk->sk_prot->max_header + len; | |
197 | release_sock(sk); | |
198 | skb = sock_alloc_send_skb(sk, size, noblock, &rc); | |
199 | lock_sock(sk); | |
7c657876 ACM |
200 | if (skb == NULL) |
201 | goto out_release; | |
202 | ||
203 | skb_reserve(skb, sk->sk_prot->max_header); | |
204 | rc = memcpy_fromiovec(skb_put(skb, len), msg->msg_iov, len); | |
27258ee5 ACM |
205 | if (rc != 0) |
206 | goto out_discard; | |
207 | ||
208 | rc = dccp_write_xmit(sk, skb, len); | |
7c657876 ACM |
209 | out_release: |
210 | release_sock(sk); | |
211 | return rc ? : len; | |
27258ee5 ACM |
212 | out_discard: |
213 | kfree_skb(skb); | |
7c657876 | 214 | goto out_release; |
7c657876 ACM |
215 | } |
216 | ||
7c657876 ACM |
217 | int dccp_recvmsg(struct kiocb *iocb, struct sock *sk, struct msghdr *msg, |
218 | size_t len, int nonblock, int flags, int *addr_len) | |
219 | { | |
220 | const struct dccp_hdr *dh; | |
7c657876 ACM |
221 | long timeo; |
222 | ||
223 | lock_sock(sk); | |
224 | ||
531669a0 ACM |
225 | if (sk->sk_state == DCCP_LISTEN) { |
226 | len = -ENOTCONN; | |
7c657876 | 227 | goto out; |
7c657876 | 228 | } |
7c657876 | 229 | |
531669a0 | 230 | timeo = sock_rcvtimeo(sk, nonblock); |
7c657876 ACM |
231 | |
232 | do { | |
531669a0 | 233 | struct sk_buff *skb = skb_peek(&sk->sk_receive_queue); |
7c657876 | 234 | |
531669a0 ACM |
235 | if (skb == NULL) |
236 | goto verify_sock_status; | |
7c657876 | 237 | |
531669a0 | 238 | dh = dccp_hdr(skb); |
7c657876 | 239 | |
531669a0 ACM |
240 | if (dh->dccph_type == DCCP_PKT_DATA || |
241 | dh->dccph_type == DCCP_PKT_DATAACK) | |
242 | goto found_ok_skb; | |
7c657876 | 243 | |
531669a0 ACM |
244 | if (dh->dccph_type == DCCP_PKT_RESET || |
245 | dh->dccph_type == DCCP_PKT_CLOSE) { | |
246 | dccp_pr_debug("found fin ok!\n"); | |
247 | len = 0; | |
248 | goto found_fin_ok; | |
249 | } | |
250 | dccp_pr_debug("packet_type=%s\n", | |
251 | dccp_packet_name(dh->dccph_type)); | |
252 | sk_eat_skb(sk, skb); | |
253 | verify_sock_status: | |
254 | if (sock_flag(sk, SOCK_DONE)) { | |
255 | len = 0; | |
7c657876 | 256 | break; |
531669a0 | 257 | } |
7c657876 | 258 | |
531669a0 ACM |
259 | if (sk->sk_err) { |
260 | len = sock_error(sk); | |
261 | break; | |
262 | } | |
7c657876 | 263 | |
531669a0 ACM |
264 | if (sk->sk_shutdown & RCV_SHUTDOWN) { |
265 | len = 0; | |
266 | break; | |
267 | } | |
7c657876 | 268 | |
531669a0 ACM |
269 | if (sk->sk_state == DCCP_CLOSED) { |
270 | if (!sock_flag(sk, SOCK_DONE)) { | |
271 | /* This occurs when user tries to read | |
272 | * from never connected socket. | |
273 | */ | |
274 | len = -ENOTCONN; | |
7c657876 ACM |
275 | break; |
276 | } | |
531669a0 ACM |
277 | len = 0; |
278 | break; | |
7c657876 ACM |
279 | } |
280 | ||
531669a0 ACM |
281 | if (!timeo) { |
282 | len = -EAGAIN; | |
283 | break; | |
284 | } | |
7c657876 | 285 | |
531669a0 ACM |
286 | if (signal_pending(current)) { |
287 | len = sock_intr_errno(timeo); | |
288 | break; | |
289 | } | |
7c657876 | 290 | |
531669a0 | 291 | sk_wait_data(sk, &timeo); |
7c657876 | 292 | continue; |
7c657876 | 293 | found_ok_skb: |
531669a0 ACM |
294 | if (len > skb->len) |
295 | len = skb->len; | |
296 | else if (len < skb->len) | |
297 | msg->msg_flags |= MSG_TRUNC; | |
298 | ||
299 | if (skb_copy_datagram_iovec(skb, 0, msg->msg_iov, len)) { | |
300 | /* Exception. Bailout! */ | |
301 | len = -EFAULT; | |
302 | break; | |
7c657876 | 303 | } |
7c657876 ACM |
304 | found_fin_ok: |
305 | if (!(flags & MSG_PEEK)) | |
306 | sk_eat_skb(sk, skb); | |
307 | break; | |
531669a0 | 308 | } while (1); |
7c657876 ACM |
309 | out: |
310 | release_sock(sk); | |
531669a0 | 311 | return len; |
7c657876 ACM |
312 | } |
313 | ||
314 | static int inet_dccp_listen(struct socket *sock, int backlog) | |
315 | { | |
316 | struct sock *sk = sock->sk; | |
317 | unsigned char old_state; | |
318 | int err; | |
319 | ||
320 | lock_sock(sk); | |
321 | ||
322 | err = -EINVAL; | |
323 | if (sock->state != SS_UNCONNECTED || sock->type != SOCK_DCCP) | |
324 | goto out; | |
325 | ||
326 | old_state = sk->sk_state; | |
327 | if (!((1 << old_state) & (DCCPF_CLOSED | DCCPF_LISTEN))) | |
328 | goto out; | |
329 | ||
330 | /* Really, if the socket is already in listen state | |
331 | * we can only allow the backlog to be adjusted. | |
332 | */ | |
333 | if (old_state != DCCP_LISTEN) { | |
334 | /* | |
335 | * FIXME: here it probably should be sk->sk_prot->listen_start | |
336 | * see tcp_listen_start | |
337 | */ | |
338 | err = dccp_listen_start(sk); | |
339 | if (err) | |
340 | goto out; | |
341 | } | |
342 | sk->sk_max_ack_backlog = backlog; | |
343 | err = 0; | |
344 | ||
345 | out: | |
346 | release_sock(sk); | |
347 | return err; | |
348 | } | |
349 | ||
350 | static const unsigned char dccp_new_state[] = { | |
7690af3f ACM |
351 | /* current state: new state: action: */ |
352 | [0] = DCCP_CLOSED, | |
353 | [DCCP_OPEN] = DCCP_CLOSING | DCCP_ACTION_FIN, | |
354 | [DCCP_REQUESTING] = DCCP_CLOSED, | |
355 | [DCCP_PARTOPEN] = DCCP_CLOSING | DCCP_ACTION_FIN, | |
356 | [DCCP_LISTEN] = DCCP_CLOSED, | |
357 | [DCCP_RESPOND] = DCCP_CLOSED, | |
358 | [DCCP_CLOSING] = DCCP_CLOSED, | |
359 | [DCCP_TIME_WAIT] = DCCP_CLOSED, | |
360 | [DCCP_CLOSED] = DCCP_CLOSED, | |
7c657876 ACM |
361 | }; |
362 | ||
363 | static int dccp_close_state(struct sock *sk) | |
364 | { | |
365 | const int next = dccp_new_state[sk->sk_state]; | |
366 | const int ns = next & DCCP_STATE_MASK; | |
367 | ||
368 | if (ns != sk->sk_state) | |
369 | dccp_set_state(sk, ns); | |
370 | ||
371 | return next & DCCP_ACTION_FIN; | |
372 | } | |
373 | ||
374 | void dccp_close(struct sock *sk, long timeout) | |
375 | { | |
376 | struct sk_buff *skb; | |
377 | ||
378 | lock_sock(sk); | |
379 | ||
380 | sk->sk_shutdown = SHUTDOWN_MASK; | |
381 | ||
382 | if (sk->sk_state == DCCP_LISTEN) { | |
383 | dccp_set_state(sk, DCCP_CLOSED); | |
384 | ||
385 | /* Special case. */ | |
386 | inet_csk_listen_stop(sk); | |
387 | ||
388 | goto adjudge_to_death; | |
389 | } | |
390 | ||
391 | /* | |
392 | * We need to flush the recv. buffs. We do this only on the | |
393 | * descriptor close, not protocol-sourced closes, because the | |
394 | *reader process may not have drained the data yet! | |
395 | */ | |
396 | /* FIXME: check for unread data */ | |
397 | while ((skb = __skb_dequeue(&sk->sk_receive_queue)) != NULL) { | |
398 | __kfree_skb(skb); | |
399 | } | |
400 | ||
401 | if (sock_flag(sk, SOCK_LINGER) && !sk->sk_lingertime) { | |
402 | /* Check zero linger _after_ checking for unread data. */ | |
403 | sk->sk_prot->disconnect(sk, 0); | |
404 | } else if (dccp_close_state(sk)) { | |
405 | dccp_send_close(sk); | |
406 | } | |
407 | ||
408 | sk_stream_wait_close(sk, timeout); | |
409 | ||
410 | adjudge_to_death: | |
411 | release_sock(sk); | |
412 | /* | |
413 | * Now socket is owned by kernel and we acquire BH lock | |
414 | * to finish close. No need to check for user refs. | |
415 | */ | |
416 | local_bh_disable(); | |
417 | bh_lock_sock(sk); | |
418 | BUG_TRAP(!sock_owned_by_user(sk)); | |
419 | ||
420 | sock_hold(sk); | |
421 | sock_orphan(sk); | |
422 | ||
423 | if (sk->sk_state != DCCP_CLOSED) | |
424 | dccp_set_state(sk, DCCP_CLOSED); | |
425 | ||
426 | atomic_inc(&dccp_orphan_count); | |
427 | if (sk->sk_state == DCCP_CLOSED) | |
428 | inet_csk_destroy_sock(sk); | |
429 | ||
430 | /* Otherwise, socket is reprieved until protocol close. */ | |
431 | ||
432 | bh_unlock_sock(sk); | |
433 | local_bh_enable(); | |
434 | sock_put(sk); | |
435 | } | |
436 | ||
437 | void dccp_shutdown(struct sock *sk, int how) | |
438 | { | |
439 | dccp_pr_debug("entry\n"); | |
440 | } | |
441 | ||
442 | struct proto_ops inet_dccp_ops = { | |
443 | .family = PF_INET, | |
444 | .owner = THIS_MODULE, | |
445 | .release = inet_release, | |
446 | .bind = inet_bind, | |
447 | .connect = inet_stream_connect, | |
448 | .socketpair = sock_no_socketpair, | |
449 | .accept = inet_accept, | |
450 | .getname = inet_getname, | |
451 | .poll = sock_no_poll, | |
452 | .ioctl = inet_ioctl, | |
7690af3f ACM |
453 | /* FIXME: work on inet_listen to rename it to sock_common_listen */ |
454 | .listen = inet_dccp_listen, | |
7c657876 ACM |
455 | .shutdown = inet_shutdown, |
456 | .setsockopt = sock_common_setsockopt, | |
457 | .getsockopt = sock_common_getsockopt, | |
458 | .sendmsg = inet_sendmsg, | |
459 | .recvmsg = sock_common_recvmsg, | |
460 | .mmap = sock_no_mmap, | |
461 | .sendpage = sock_no_sendpage, | |
462 | }; | |
463 | ||
464 | extern struct net_proto_family inet_family_ops; | |
465 | ||
466 | static struct inet_protosw dccp_v4_protosw = { | |
467 | .type = SOCK_DCCP, | |
468 | .protocol = IPPROTO_DCCP, | |
469 | .prot = &dccp_v4_prot, | |
470 | .ops = &inet_dccp_ops, | |
471 | .capability = -1, | |
472 | .no_check = 0, | |
473 | .flags = 0, | |
474 | }; | |
475 | ||
476 | /* | |
477 | * This is the global socket data structure used for responding to | |
478 | * the Out-of-the-blue (OOTB) packets. A control sock will be created | |
479 | * for this socket at the initialization time. | |
480 | */ | |
481 | struct socket *dccp_ctl_socket; | |
482 | ||
483 | static char dccp_ctl_socket_err_msg[] __initdata = | |
484 | KERN_ERR "DCCP: Failed to create the control socket.\n"; | |
485 | ||
486 | static int __init dccp_ctl_sock_init(void) | |
487 | { | |
488 | int rc = sock_create_kern(PF_INET, SOCK_DCCP, IPPROTO_DCCP, | |
489 | &dccp_ctl_socket); | |
490 | if (rc < 0) | |
491 | printk(dccp_ctl_socket_err_msg); | |
492 | else { | |
493 | dccp_ctl_socket->sk->sk_allocation = GFP_ATOMIC; | |
494 | inet_sk(dccp_ctl_socket->sk)->uc_ttl = -1; | |
495 | ||
496 | /* Unhash it so that IP input processing does not even | |
497 | * see it, we do not wish this socket to see incoming | |
498 | * packets. | |
499 | */ | |
500 | dccp_ctl_socket->sk->sk_prot->unhash(dccp_ctl_socket->sk); | |
501 | } | |
502 | ||
503 | return rc; | |
504 | } | |
505 | ||
506 | static void __exit dccp_ctl_sock_exit(void) | |
507 | { | |
508 | if (dccp_ctl_socket != NULL) | |
509 | sock_release(dccp_ctl_socket); | |
510 | } | |
511 | ||
512 | static int __init init_dccp_v4_mibs(void) | |
513 | { | |
514 | int rc = -ENOMEM; | |
515 | ||
516 | dccp_statistics[0] = alloc_percpu(struct dccp_mib); | |
517 | if (dccp_statistics[0] == NULL) | |
518 | goto out; | |
519 | ||
520 | dccp_statistics[1] = alloc_percpu(struct dccp_mib); | |
521 | if (dccp_statistics[1] == NULL) | |
522 | goto out_free_one; | |
523 | ||
524 | rc = 0; | |
525 | out: | |
526 | return rc; | |
527 | out_free_one: | |
528 | free_percpu(dccp_statistics[0]); | |
529 | dccp_statistics[0] = NULL; | |
530 | goto out; | |
531 | ||
532 | } | |
533 | ||
534 | static int thash_entries; | |
535 | module_param(thash_entries, int, 0444); | |
536 | MODULE_PARM_DESC(thash_entries, "Number of ehash buckets"); | |
537 | ||
538 | int dccp_debug; | |
539 | module_param(dccp_debug, int, 0444); | |
540 | MODULE_PARM_DESC(dccp_debug, "Enable debug messages"); | |
541 | ||
542 | static int __init dccp_init(void) | |
543 | { | |
544 | unsigned long goal; | |
545 | int ehash_order, bhash_order, i; | |
546 | int rc = proto_register(&dccp_v4_prot, 1); | |
547 | ||
548 | if (rc) | |
549 | goto out; | |
550 | ||
7690af3f ACM |
551 | dccp_hashinfo.bind_bucket_cachep = |
552 | kmem_cache_create("dccp_bind_bucket", | |
553 | sizeof(struct inet_bind_bucket), 0, | |
554 | SLAB_HWCACHE_ALIGN, NULL, NULL); | |
7c657876 ACM |
555 | if (!dccp_hashinfo.bind_bucket_cachep) |
556 | goto out_proto_unregister; | |
557 | ||
558 | /* | |
559 | * Size and allocate the main established and bind bucket | |
560 | * hash tables. | |
561 | * | |
562 | * The methodology is similar to that of the buffer cache. | |
563 | */ | |
564 | if (num_physpages >= (128 * 1024)) | |
565 | goal = num_physpages >> (21 - PAGE_SHIFT); | |
566 | else | |
567 | goal = num_physpages >> (23 - PAGE_SHIFT); | |
568 | ||
569 | if (thash_entries) | |
7690af3f ACM |
570 | goal = (thash_entries * |
571 | sizeof(struct inet_ehash_bucket)) >> PAGE_SHIFT; | |
7c657876 ACM |
572 | for (ehash_order = 0; (1UL << ehash_order) < goal; ehash_order++) |
573 | ; | |
574 | do { | |
575 | dccp_hashinfo.ehash_size = (1UL << ehash_order) * PAGE_SIZE / | |
576 | sizeof(struct inet_ehash_bucket); | |
577 | dccp_hashinfo.ehash_size >>= 1; | |
7690af3f ACM |
578 | while (dccp_hashinfo.ehash_size & |
579 | (dccp_hashinfo.ehash_size - 1)) | |
7c657876 ACM |
580 | dccp_hashinfo.ehash_size--; |
581 | dccp_hashinfo.ehash = (struct inet_ehash_bucket *) | |
582 | __get_free_pages(GFP_ATOMIC, ehash_order); | |
583 | } while (!dccp_hashinfo.ehash && --ehash_order > 0); | |
584 | ||
585 | if (!dccp_hashinfo.ehash) { | |
586 | printk(KERN_CRIT "Failed to allocate DCCP " | |
587 | "established hash table\n"); | |
588 | goto out_free_bind_bucket_cachep; | |
589 | } | |
590 | ||
591 | for (i = 0; i < (dccp_hashinfo.ehash_size << 1); i++) { | |
592 | rwlock_init(&dccp_hashinfo.ehash[i].lock); | |
593 | INIT_HLIST_HEAD(&dccp_hashinfo.ehash[i].chain); | |
594 | } | |
595 | ||
596 | bhash_order = ehash_order; | |
597 | ||
598 | do { | |
599 | dccp_hashinfo.bhash_size = (1UL << bhash_order) * PAGE_SIZE / | |
600 | sizeof(struct inet_bind_hashbucket); | |
7690af3f ACM |
601 | if ((dccp_hashinfo.bhash_size > (64 * 1024)) && |
602 | bhash_order > 0) | |
7c657876 ACM |
603 | continue; |
604 | dccp_hashinfo.bhash = (struct inet_bind_hashbucket *) | |
605 | __get_free_pages(GFP_ATOMIC, bhash_order); | |
606 | } while (!dccp_hashinfo.bhash && --bhash_order >= 0); | |
607 | ||
608 | if (!dccp_hashinfo.bhash) { | |
609 | printk(KERN_CRIT "Failed to allocate DCCP bind hash table\n"); | |
610 | goto out_free_dccp_ehash; | |
611 | } | |
612 | ||
613 | for (i = 0; i < dccp_hashinfo.bhash_size; i++) { | |
614 | spin_lock_init(&dccp_hashinfo.bhash[i].lock); | |
615 | INIT_HLIST_HEAD(&dccp_hashinfo.bhash[i].chain); | |
616 | } | |
617 | ||
618 | if (init_dccp_v4_mibs()) | |
619 | goto out_free_dccp_bhash; | |
620 | ||
621 | rc = -EAGAIN; | |
622 | if (inet_add_protocol(&dccp_protocol, IPPROTO_DCCP)) | |
623 | goto out_free_dccp_v4_mibs; | |
624 | ||
625 | inet_register_protosw(&dccp_v4_protosw); | |
626 | ||
627 | rc = dccp_ctl_sock_init(); | |
628 | if (rc) | |
629 | goto out_unregister_protosw; | |
630 | out: | |
631 | return rc; | |
632 | out_unregister_protosw: | |
633 | inet_unregister_protosw(&dccp_v4_protosw); | |
634 | inet_del_protocol(&dccp_protocol, IPPROTO_DCCP); | |
635 | out_free_dccp_v4_mibs: | |
636 | free_percpu(dccp_statistics[0]); | |
637 | free_percpu(dccp_statistics[1]); | |
638 | dccp_statistics[0] = dccp_statistics[1] = NULL; | |
639 | out_free_dccp_bhash: | |
640 | free_pages((unsigned long)dccp_hashinfo.bhash, bhash_order); | |
641 | dccp_hashinfo.bhash = NULL; | |
642 | out_free_dccp_ehash: | |
643 | free_pages((unsigned long)dccp_hashinfo.ehash, ehash_order); | |
644 | dccp_hashinfo.ehash = NULL; | |
645 | out_free_bind_bucket_cachep: | |
646 | kmem_cache_destroy(dccp_hashinfo.bind_bucket_cachep); | |
647 | dccp_hashinfo.bind_bucket_cachep = NULL; | |
648 | out_proto_unregister: | |
649 | proto_unregister(&dccp_v4_prot); | |
650 | goto out; | |
651 | } | |
652 | ||
653 | static const char dccp_del_proto_err_msg[] __exitdata = | |
654 | KERN_ERR "can't remove dccp net_protocol\n"; | |
655 | ||
656 | static void __exit dccp_fini(void) | |
657 | { | |
658 | dccp_ctl_sock_exit(); | |
659 | ||
660 | inet_unregister_protosw(&dccp_v4_protosw); | |
661 | ||
662 | if (inet_del_protocol(&dccp_protocol, IPPROTO_DCCP) < 0) | |
663 | printk(dccp_del_proto_err_msg); | |
664 | ||
665 | /* Free the control endpoint. */ | |
666 | sock_release(dccp_ctl_socket); | |
667 | ||
668 | proto_unregister(&dccp_v4_prot); | |
669 | ||
670 | kmem_cache_destroy(dccp_hashinfo.bind_bucket_cachep); | |
671 | } | |
672 | ||
673 | module_init(dccp_init); | |
674 | module_exit(dccp_fini); | |
675 | ||
bb97d31f ACM |
676 | /* |
677 | * __stringify doesn't likes enums, so use SOCK_DCCP (6) and IPPROTO_DCCP (33) | |
678 | * values directly, Also cover the case where the protocol is not specified, | |
679 | * i.e. net-pf-PF_INET-proto-0-type-SOCK_DCCP | |
680 | */ | |
681 | MODULE_ALIAS("net-pf-" __stringify(PF_INET) "-proto-33-type-6"); | |
682 | MODULE_ALIAS("net-pf-" __stringify(PF_INET) "-proto-0-type-6"); | |
7c657876 ACM |
683 | MODULE_LICENSE("GPL"); |
684 | MODULE_AUTHOR("Arnaldo Carvalho de Melo <acme@conectiva.com.br>"); | |
685 | MODULE_DESCRIPTION("DCCP - Datagram Congestion Controlled Protocol"); |