]>
Commit | Line | Data |
---|---|---|
1 | /* | |
2 | * Copyright (c) 2000 Apple Computer, Inc. All rights reserved. | |
3 | * | |
4 | * @APPLE_LICENSE_HEADER_START@ | |
5 | * | |
6 | * The contents of this file constitute Original Code as defined in and | |
7 | * are subject to the Apple Public Source License Version 1.1 (the | |
8 | * "License"). You may not use this file except in compliance with the | |
9 | * License. Please obtain a copy of the License at | |
10 | * http://www.apple.com/publicsource and read it before using this file. | |
11 | * | |
12 | * This Original Code and all software distributed under the License are | |
13 | * distributed on an "AS IS" basis, WITHOUT WARRANTY OF ANY KIND, EITHER | |
14 | * EXPRESS OR IMPLIED, AND APPLE HEREBY DISCLAIMS ALL SUCH WARRANTIES, | |
15 | * INCLUDING WITHOUT LIMITATION, ANY WARRANTIES OF MERCHANTABILITY, | |
16 | * FITNESS FOR A PARTICULAR PURPOSE OR NON-INFRINGEMENT. Please see the | |
17 | * License for the specific language governing rights and limitations | |
18 | * under the License. | |
19 | * | |
20 | * @APPLE_LICENSE_HEADER_END@ | |
21 | */ | |
22 | /* | |
23 | * Copyright (c) 1982, 1986, 1988, 1993 | |
24 | * The Regents of the University of California. All rights reserved. | |
25 | * | |
26 | * Redistribution and use in source and binary forms, with or without | |
27 | * modification, are permitted provided that the following conditions | |
28 | * are met: | |
29 | * 1. Redistributions of source code must retain the above copyright | |
30 | * notice, this list of conditions and the following disclaimer. | |
31 | * 2. Redistributions in binary form must reproduce the above copyright | |
32 | * notice, this list of conditions and the following disclaimer in the | |
33 | * documentation and/or other materials provided with the distribution. | |
34 | * 3. All advertising materials mentioning features or use of this software | |
35 | * must display the following acknowledgement: | |
36 | * This product includes software developed by the University of | |
37 | * California, Berkeley and its contributors. | |
38 | * 4. Neither the name of the University nor the names of its contributors | |
39 | * may be used to endorse or promote products derived from this software | |
40 | * without specific prior written permission. | |
41 | * | |
42 | * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND | |
43 | * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE | |
44 | * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE | |
45 | * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE | |
46 | * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL | |
47 | * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS | |
48 | * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) | |
49 | * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT | |
50 | * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY | |
51 | * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF | |
52 | * SUCH DAMAGE. | |
53 | * | |
54 | * From: @(#)tcp_usrreq.c 8.2 (Berkeley) 1/3/94 | |
55 | * $FreeBSD: src/sys/netinet/tcp_usrreq.c,v 1.51.2.9 2001/08/22 00:59:12 silby Exp $ | |
56 | */ | |
57 | ||
58 | ||
59 | #include <sys/param.h> | |
60 | #include <sys/systm.h> | |
61 | #include <sys/kernel.h> | |
62 | #include <sys/sysctl.h> | |
63 | #include <sys/mbuf.h> | |
64 | #if INET6 | |
65 | #include <sys/domain.h> | |
66 | #endif /* INET6 */ | |
67 | #include <sys/socket.h> | |
68 | #include <sys/socketvar.h> | |
69 | #include <sys/protosw.h> | |
70 | ||
71 | #include <net/if.h> | |
72 | #include <net/route.h> | |
73 | ||
74 | #include <netinet/in.h> | |
75 | #include <netinet/in_systm.h> | |
76 | #if INET6 | |
77 | #include <netinet/ip6.h> | |
78 | #endif | |
79 | #include <netinet/in_pcb.h> | |
80 | #if INET6 | |
81 | #include <netinet6/in6_pcb.h> | |
82 | #endif | |
83 | #include <netinet/in_var.h> | |
84 | #include <netinet/ip_var.h> | |
85 | #if INET6 | |
86 | #include <netinet6/ip6_var.h> | |
87 | #endif | |
88 | #include <netinet/tcp.h> | |
89 | #include <netinet/tcp_fsm.h> | |
90 | #include <netinet/tcp_seq.h> | |
91 | #include <netinet/tcp_timer.h> | |
92 | #include <netinet/tcp_var.h> | |
93 | #include <netinet/tcpip.h> | |
94 | #if TCPDEBUG | |
95 | #include <netinet/tcp_debug.h> | |
96 | #endif | |
97 | ||
98 | #if IPSEC | |
99 | #include <netinet6/ipsec.h> | |
100 | #endif /*IPSEC*/ | |
101 | ||
102 | /* | |
103 | * TCP protocol interface to socket abstraction. | |
104 | */ | |
105 | extern char *tcpstates[]; /* XXX ??? */ | |
106 | ||
107 | static int tcp_attach(struct socket *, struct proc *); | |
108 | static int tcp_connect(struct tcpcb *, struct sockaddr *, struct proc *); | |
109 | #if INET6 | |
110 | static int tcp6_connect(struct tcpcb *, struct sockaddr *, struct proc *); | |
111 | #endif /* INET6 */ | |
112 | static struct tcpcb * | |
113 | tcp_disconnect(struct tcpcb *); | |
114 | static struct tcpcb * | |
115 | tcp_usrclosed(struct tcpcb *); | |
116 | ||
117 | #if TCPDEBUG | |
118 | #define TCPDEBUG0 int ostate = 0 | |
119 | #define TCPDEBUG1() ostate = tp ? tp->t_state : 0 | |
120 | #define TCPDEBUG2(req) if (tp && (so->so_options & SO_DEBUG)) \ | |
121 | tcp_trace(TA_USER, ostate, tp, 0, 0, req) | |
122 | #else | |
123 | #define TCPDEBUG0 | |
124 | #define TCPDEBUG1() | |
125 | #define TCPDEBUG2(req) | |
126 | #endif | |
127 | ||
128 | /* | |
129 | * TCP attaches to socket via pru_attach(), reserving space, | |
130 | * and an internet control block. | |
131 | */ | |
132 | static int | |
133 | tcp_usr_attach(struct socket *so, int proto, struct proc *p) | |
134 | { | |
135 | int error; | |
136 | struct inpcb *inp = sotoinpcb(so); | |
137 | struct tcpcb *tp = 0; | |
138 | TCPDEBUG0; | |
139 | ||
140 | TCPDEBUG1(); | |
141 | if (inp) { | |
142 | error = EISCONN; | |
143 | goto out; | |
144 | } | |
145 | ||
146 | error = tcp_attach(so, p); | |
147 | if (error) | |
148 | goto out; | |
149 | ||
150 | if ((so->so_options & SO_LINGER) && so->so_linger == 0) | |
151 | so->so_linger = TCP_LINGERTIME * hz; | |
152 | tp = sototcpcb(so); | |
153 | out: | |
154 | TCPDEBUG2(PRU_ATTACH); | |
155 | return error; | |
156 | } | |
157 | ||
158 | /* | |
159 | * pru_detach() detaches the TCP protocol from the socket. | |
160 | * If the protocol state is non-embryonic, then can't | |
161 | * do this directly: have to initiate a pru_disconnect(), | |
162 | * which may finish later; embryonic TCB's can just | |
163 | * be discarded here. | |
164 | */ | |
165 | static int | |
166 | tcp_usr_detach(struct socket *so) | |
167 | { | |
168 | int error = 0; | |
169 | struct inpcb *inp = sotoinpcb(so); | |
170 | struct tcpcb *tp; | |
171 | TCPDEBUG0; | |
172 | ||
173 | if (inp == 0 || (inp->inp_state == INPCB_STATE_DEAD)) { | |
174 | return EINVAL; /* XXX */ | |
175 | } | |
176 | #if 1 | |
177 | lck_mtx_assert(((struct inpcb *)so->so_pcb)->inpcb_mtx, LCK_MTX_ASSERT_OWNED); | |
178 | #endif | |
179 | tp = intotcpcb(inp); | |
180 | /* In case we got disconnected from the peer */ | |
181 | if (tp == 0) | |
182 | goto out; | |
183 | TCPDEBUG1(); | |
184 | tp = tcp_disconnect(tp); | |
185 | out: | |
186 | TCPDEBUG2(PRU_DETACH); | |
187 | return error; | |
188 | } | |
189 | ||
190 | #define COMMON_START() TCPDEBUG0; \ | |
191 | do { \ | |
192 | if (inp == 0 || (inp->inp_state == INPCB_STATE_DEAD)) { \ | |
193 | return EINVAL; \ | |
194 | } \ | |
195 | tp = intotcpcb(inp); \ | |
196 | TCPDEBUG1(); \ | |
197 | } while(0) | |
198 | ||
199 | #define COMMON_END(req) out: TCPDEBUG2(req); return error; goto out | |
200 | ||
201 | ||
202 | /* | |
203 | * Give the socket an address. | |
204 | */ | |
205 | static int | |
206 | tcp_usr_bind(struct socket *so, struct sockaddr *nam, struct proc *p) | |
207 | { | |
208 | int error = 0; | |
209 | struct inpcb *inp = sotoinpcb(so); | |
210 | struct tcpcb *tp; | |
211 | struct sockaddr_in *sinp; | |
212 | ||
213 | COMMON_START(); | |
214 | ||
215 | /* | |
216 | * Must check for multicast addresses and disallow binding | |
217 | * to them. | |
218 | */ | |
219 | sinp = (struct sockaddr_in *)nam; | |
220 | if (sinp->sin_family == AF_INET && | |
221 | IN_MULTICAST(ntohl(sinp->sin_addr.s_addr))) { | |
222 | error = EAFNOSUPPORT; | |
223 | goto out; | |
224 | } | |
225 | error = in_pcbbind(inp, nam, p); | |
226 | if (error) | |
227 | goto out; | |
228 | COMMON_END(PRU_BIND); | |
229 | ||
230 | } | |
231 | ||
232 | #if INET6 | |
233 | static int | |
234 | tcp6_usr_bind(struct socket *so, struct sockaddr *nam, struct proc *p) | |
235 | { | |
236 | int error = 0; | |
237 | struct inpcb *inp = sotoinpcb(so); | |
238 | struct tcpcb *tp; | |
239 | struct sockaddr_in6 *sin6p; | |
240 | ||
241 | COMMON_START(); | |
242 | ||
243 | /* | |
244 | * Must check for multicast addresses and disallow binding | |
245 | * to them. | |
246 | */ | |
247 | sin6p = (struct sockaddr_in6 *)nam; | |
248 | if (sin6p->sin6_family == AF_INET6 && | |
249 | IN6_IS_ADDR_MULTICAST(&sin6p->sin6_addr)) { | |
250 | error = EAFNOSUPPORT; | |
251 | goto out; | |
252 | } | |
253 | inp->inp_vflag &= ~INP_IPV4; | |
254 | inp->inp_vflag |= INP_IPV6; | |
255 | if ((inp->inp_flags & IN6P_IPV6_V6ONLY) == 0) { | |
256 | if (IN6_IS_ADDR_UNSPECIFIED(&sin6p->sin6_addr)) | |
257 | inp->inp_vflag |= INP_IPV4; | |
258 | else if (IN6_IS_ADDR_V4MAPPED(&sin6p->sin6_addr)) { | |
259 | struct sockaddr_in sin; | |
260 | ||
261 | in6_sin6_2_sin(&sin, sin6p); | |
262 | inp->inp_vflag |= INP_IPV4; | |
263 | inp->inp_vflag &= ~INP_IPV6; | |
264 | error = in_pcbbind(inp, (struct sockaddr *)&sin, p); | |
265 | goto out; | |
266 | } | |
267 | } | |
268 | error = in6_pcbbind(inp, nam, p); | |
269 | if (error) | |
270 | goto out; | |
271 | COMMON_END(PRU_BIND); | |
272 | } | |
273 | #endif /* INET6 */ | |
274 | ||
275 | /* | |
276 | * Prepare to accept connections. | |
277 | */ | |
278 | static int | |
279 | tcp_usr_listen(struct socket *so, struct proc *p) | |
280 | { | |
281 | int error = 0; | |
282 | struct inpcb *inp = sotoinpcb(so); | |
283 | struct tcpcb *tp; | |
284 | ||
285 | COMMON_START(); | |
286 | if (inp->inp_lport == 0) | |
287 | error = in_pcbbind(inp, (struct sockaddr *)0, p); | |
288 | if (error == 0) | |
289 | tp->t_state = TCPS_LISTEN; | |
290 | COMMON_END(PRU_LISTEN); | |
291 | } | |
292 | ||
293 | #if INET6 | |
294 | static int | |
295 | tcp6_usr_listen(struct socket *so, struct proc *p) | |
296 | { | |
297 | int error = 0; | |
298 | struct inpcb *inp = sotoinpcb(so); | |
299 | struct tcpcb *tp; | |
300 | ||
301 | COMMON_START(); | |
302 | if (inp->inp_lport == 0) { | |
303 | inp->inp_vflag &= ~INP_IPV4; | |
304 | if ((inp->inp_flags & IN6P_IPV6_V6ONLY) == 0) | |
305 | inp->inp_vflag |= INP_IPV4; | |
306 | error = in6_pcbbind(inp, (struct sockaddr *)0, p); | |
307 | } | |
308 | if (error == 0) | |
309 | tp->t_state = TCPS_LISTEN; | |
310 | COMMON_END(PRU_LISTEN); | |
311 | } | |
312 | #endif /* INET6 */ | |
313 | ||
314 | /* | |
315 | * Initiate connection to peer. | |
316 | * Create a template for use in transmissions on this connection. | |
317 | * Enter SYN_SENT state, and mark socket as connecting. | |
318 | * Start keep-alive timer, and seed output sequence space. | |
319 | * Send initial segment on connection. | |
320 | */ | |
321 | static int | |
322 | tcp_usr_connect(struct socket *so, struct sockaddr *nam, struct proc *p) | |
323 | { | |
324 | int error = 0; | |
325 | struct inpcb *inp = sotoinpcb(so); | |
326 | struct tcpcb *tp; | |
327 | struct sockaddr_in *sinp; | |
328 | ||
329 | COMMON_START(); | |
330 | ||
331 | /* | |
332 | * Must disallow TCP ``connections'' to multicast addresses. | |
333 | */ | |
334 | sinp = (struct sockaddr_in *)nam; | |
335 | if (sinp->sin_family == AF_INET | |
336 | && IN_MULTICAST(ntohl(sinp->sin_addr.s_addr))) { | |
337 | error = EAFNOSUPPORT; | |
338 | goto out; | |
339 | } | |
340 | ||
341 | #ifndef __APPLE__ | |
342 | prison_remote_ip(p, 0, &sinp->sin_addr.s_addr); | |
343 | #endif | |
344 | ||
345 | if ((error = tcp_connect(tp, nam, p)) != 0) | |
346 | goto out; | |
347 | error = tcp_output(tp); | |
348 | COMMON_END(PRU_CONNECT); | |
349 | } | |
350 | ||
351 | #if INET6 | |
352 | static int | |
353 | tcp6_usr_connect(struct socket *so, struct sockaddr *nam, struct proc *p) | |
354 | { | |
355 | int error = 0; | |
356 | struct inpcb *inp = sotoinpcb(so); | |
357 | struct tcpcb *tp; | |
358 | struct sockaddr_in6 *sin6p; | |
359 | ||
360 | COMMON_START(); | |
361 | ||
362 | /* | |
363 | * Must disallow TCP ``connections'' to multicast addresses. | |
364 | */ | |
365 | sin6p = (struct sockaddr_in6 *)nam; | |
366 | if (sin6p->sin6_family == AF_INET6 | |
367 | && IN6_IS_ADDR_MULTICAST(&sin6p->sin6_addr)) { | |
368 | error = EAFNOSUPPORT; | |
369 | goto out; | |
370 | } | |
371 | ||
372 | if (IN6_IS_ADDR_V4MAPPED(&sin6p->sin6_addr)) { | |
373 | struct sockaddr_in sin; | |
374 | ||
375 | if ((inp->inp_flags & IN6P_IPV6_V6ONLY) != 0) | |
376 | return (EINVAL); | |
377 | ||
378 | in6_sin6_2_sin(&sin, sin6p); | |
379 | inp->inp_vflag |= INP_IPV4; | |
380 | inp->inp_vflag &= ~INP_IPV6; | |
381 | if ((error = tcp_connect(tp, (struct sockaddr *)&sin, p)) != 0) | |
382 | goto out; | |
383 | error = tcp_output(tp); | |
384 | goto out; | |
385 | } | |
386 | inp->inp_vflag &= ~INP_IPV4; | |
387 | inp->inp_vflag |= INP_IPV6; | |
388 | if ((error = tcp6_connect(tp, nam, p)) != 0) | |
389 | goto out; | |
390 | error = tcp_output(tp); | |
391 | if (error) | |
392 | goto out; | |
393 | COMMON_END(PRU_CONNECT); | |
394 | } | |
395 | #endif /* INET6 */ | |
396 | ||
397 | /* | |
398 | * Initiate disconnect from peer. | |
399 | * If connection never passed embryonic stage, just drop; | |
400 | * else if don't need to let data drain, then can just drop anyways, | |
401 | * else have to begin TCP shutdown process: mark socket disconnecting, | |
402 | * drain unread data, state switch to reflect user close, and | |
403 | * send segment (e.g. FIN) to peer. Socket will be really disconnected | |
404 | * when peer sends FIN and acks ours. | |
405 | * | |
406 | * SHOULD IMPLEMENT LATER PRU_CONNECT VIA REALLOC TCPCB. | |
407 | */ | |
408 | static int | |
409 | tcp_usr_disconnect(struct socket *so) | |
410 | { | |
411 | int error = 0; | |
412 | struct inpcb *inp = sotoinpcb(so); | |
413 | struct tcpcb *tp; | |
414 | ||
415 | #if 1 | |
416 | lck_mtx_assert(((struct inpcb *)so->so_pcb)->inpcb_mtx, LCK_MTX_ASSERT_OWNED); | |
417 | #endif | |
418 | COMMON_START(); | |
419 | /* In case we got disconnected from the peer */ | |
420 | if (tp == 0) | |
421 | goto out; | |
422 | tp = tcp_disconnect(tp); | |
423 | COMMON_END(PRU_DISCONNECT); | |
424 | } | |
425 | ||
426 | /* | |
427 | * Accept a connection. Essentially all the work is | |
428 | * done at higher levels; just return the address | |
429 | * of the peer, storing through addr. | |
430 | */ | |
431 | static int | |
432 | tcp_usr_accept(struct socket *so, struct sockaddr **nam) | |
433 | { | |
434 | int error = 0; | |
435 | struct inpcb *inp = sotoinpcb(so); | |
436 | struct tcpcb *tp = NULL; | |
437 | TCPDEBUG0; | |
438 | ||
439 | if (so->so_state & SS_ISDISCONNECTED) { | |
440 | error = ECONNABORTED; | |
441 | goto out; | |
442 | } | |
443 | if (inp == 0 || (inp->inp_state == INPCB_STATE_DEAD)) { | |
444 | return (EINVAL); | |
445 | } | |
446 | tp = intotcpcb(inp); | |
447 | TCPDEBUG1(); | |
448 | in_setpeeraddr(so, nam); | |
449 | COMMON_END(PRU_ACCEPT); | |
450 | } | |
451 | ||
452 | #if INET6 | |
453 | static int | |
454 | tcp6_usr_accept(struct socket *so, struct sockaddr **nam) | |
455 | { | |
456 | int error = 0; | |
457 | struct inpcb *inp = sotoinpcb(so); | |
458 | struct tcpcb *tp = NULL; | |
459 | TCPDEBUG0; | |
460 | ||
461 | if (so->so_state & SS_ISDISCONNECTED) { | |
462 | error = ECONNABORTED; | |
463 | goto out; | |
464 | } | |
465 | if (inp == 0 || (inp->inp_state == INPCB_STATE_DEAD)) { | |
466 | return (EINVAL); | |
467 | } | |
468 | tp = intotcpcb(inp); | |
469 | TCPDEBUG1(); | |
470 | in6_mapped_peeraddr(so, nam); | |
471 | COMMON_END(PRU_ACCEPT); | |
472 | } | |
473 | #endif /* INET6 */ | |
474 | /* | |
475 | * Mark the connection as being incapable of further output. | |
476 | */ | |
477 | static int | |
478 | tcp_usr_shutdown(struct socket *so) | |
479 | { | |
480 | int error = 0; | |
481 | struct inpcb *inp = sotoinpcb(so); | |
482 | struct tcpcb *tp; | |
483 | ||
484 | COMMON_START(); | |
485 | socantsendmore(so); | |
486 | /* In case we got disconnected from the peer */ | |
487 | if (tp == 0) | |
488 | goto out; | |
489 | tp = tcp_usrclosed(tp); | |
490 | if (tp) | |
491 | error = tcp_output(tp); | |
492 | COMMON_END(PRU_SHUTDOWN); | |
493 | } | |
494 | ||
495 | /* | |
496 | * After a receive, possibly send window update to peer. | |
497 | */ | |
498 | static int | |
499 | tcp_usr_rcvd(struct socket *so, int flags) | |
500 | { | |
501 | int error = 0; | |
502 | struct inpcb *inp = sotoinpcb(so); | |
503 | struct tcpcb *tp; | |
504 | ||
505 | COMMON_START(); | |
506 | /* In case we got disconnected from the peer */ | |
507 | if (tp == 0) | |
508 | goto out; | |
509 | tcp_output(tp); | |
510 | COMMON_END(PRU_RCVD); | |
511 | } | |
512 | ||
513 | /* | |
514 | * Do a send by putting data in output queue and updating urgent | |
515 | * marker if URG set. Possibly send more data. Unlike the other | |
516 | * pru_*() routines, the mbuf chains are our responsibility. We | |
517 | * must either enqueue them or free them. The other pru_* routines | |
518 | * generally are caller-frees. | |
519 | */ | |
520 | static int | |
521 | tcp_usr_send(struct socket *so, int flags, struct mbuf *m, | |
522 | struct sockaddr *nam, struct mbuf *control, struct proc *p) | |
523 | { | |
524 | int error = 0; | |
525 | struct inpcb *inp = sotoinpcb(so); | |
526 | struct tcpcb *tp; | |
527 | #if INET6 | |
528 | int isipv6; | |
529 | #endif | |
530 | TCPDEBUG0; | |
531 | ||
532 | if (inp == NULL || inp->inp_state == INPCB_STATE_DEAD) { | |
533 | /* | |
534 | * OOPS! we lost a race, the TCP session got reset after | |
535 | * we checked SS_CANTSENDMORE, eg: while doing uiomove or a | |
536 | * network interrupt in the non-splnet() section of sosend(). | |
537 | */ | |
538 | if (m) | |
539 | m_freem(m); | |
540 | if (control) | |
541 | m_freem(control); | |
542 | error = ECONNRESET; /* XXX EPIPE? */ | |
543 | tp = NULL; | |
544 | TCPDEBUG1(); | |
545 | goto out; | |
546 | } | |
547 | #if INET6 | |
548 | isipv6 = nam && nam->sa_family == AF_INET6; | |
549 | #endif /* INET6 */ | |
550 | tp = intotcpcb(inp); | |
551 | TCPDEBUG1(); | |
552 | if (control) { | |
553 | /* TCP doesn't do control messages (rights, creds, etc) */ | |
554 | if (control->m_len) { | |
555 | m_freem(control); | |
556 | if (m) | |
557 | m_freem(m); | |
558 | error = EINVAL; | |
559 | goto out; | |
560 | } | |
561 | m_freem(control); /* empty control, just free it */ | |
562 | } | |
563 | if(!(flags & PRUS_OOB)) { | |
564 | sbappend(&so->so_snd, m); | |
565 | if (nam && tp->t_state < TCPS_SYN_SENT) { | |
566 | /* | |
567 | * Do implied connect if not yet connected, | |
568 | * initialize window to default value, and | |
569 | * initialize maxseg/maxopd using peer's cached | |
570 | * MSS. | |
571 | */ | |
572 | #if INET6 | |
573 | if (isipv6) | |
574 | error = tcp6_connect(tp, nam, p); | |
575 | else | |
576 | #endif /* INET6 */ | |
577 | error = tcp_connect(tp, nam, p); | |
578 | if (error) | |
579 | goto out; | |
580 | tp->snd_wnd = TTCP_CLIENT_SND_WND; | |
581 | tcp_mss(tp, -1); | |
582 | } | |
583 | ||
584 | if (flags & PRUS_EOF) { | |
585 | /* | |
586 | * Close the send side of the connection after | |
587 | * the data is sent. | |
588 | */ | |
589 | socantsendmore(so); | |
590 | tp = tcp_usrclosed(tp); | |
591 | } | |
592 | if (tp != NULL) { | |
593 | if (flags & PRUS_MORETOCOME) | |
594 | tp->t_flags |= TF_MORETOCOME; | |
595 | error = tcp_output(tp); | |
596 | if (flags & PRUS_MORETOCOME) | |
597 | tp->t_flags &= ~TF_MORETOCOME; | |
598 | } | |
599 | } else { | |
600 | if (sbspace(&so->so_snd) < -512) { | |
601 | m_freem(m); | |
602 | error = ENOBUFS; | |
603 | goto out; | |
604 | } | |
605 | /* | |
606 | * According to RFC961 (Assigned Protocols), | |
607 | * the urgent pointer points to the last octet | |
608 | * of urgent data. We continue, however, | |
609 | * to consider it to indicate the first octet | |
610 | * of data past the urgent section. | |
611 | * Otherwise, snd_up should be one lower. | |
612 | */ | |
613 | sbappend(&so->so_snd, m); | |
614 | if (nam && tp->t_state < TCPS_SYN_SENT) { | |
615 | /* | |
616 | * Do implied connect if not yet connected, | |
617 | * initialize window to default value, and | |
618 | * initialize maxseg/maxopd using peer's cached | |
619 | * MSS. | |
620 | */ | |
621 | #if INET6 | |
622 | if (isipv6) | |
623 | error = tcp6_connect(tp, nam, p); | |
624 | else | |
625 | #endif /* INET6 */ | |
626 | error = tcp_connect(tp, nam, p); | |
627 | if (error) | |
628 | goto out; | |
629 | tp->snd_wnd = TTCP_CLIENT_SND_WND; | |
630 | tcp_mss(tp, -1); | |
631 | } | |
632 | tp->snd_up = tp->snd_una + so->so_snd.sb_cc; | |
633 | tp->t_force = 1; | |
634 | error = tcp_output(tp); | |
635 | tp->t_force = 0; | |
636 | } | |
637 | COMMON_END((flags & PRUS_OOB) ? PRU_SENDOOB : | |
638 | ((flags & PRUS_EOF) ? PRU_SEND_EOF : PRU_SEND)); | |
639 | } | |
640 | ||
641 | /* | |
642 | * Abort the TCP. | |
643 | */ | |
644 | static int | |
645 | tcp_usr_abort(struct socket *so) | |
646 | { | |
647 | int error = 0; | |
648 | struct inpcb *inp = sotoinpcb(so); | |
649 | struct tcpcb *tp; | |
650 | ||
651 | COMMON_START(); | |
652 | /* In case we got disconnected from the peer */ | |
653 | if (tp == 0) | |
654 | goto out; | |
655 | tp = tcp_drop(tp, ECONNABORTED); | |
656 | so->so_usecount--; | |
657 | COMMON_END(PRU_ABORT); | |
658 | } | |
659 | ||
660 | /* | |
661 | * Receive out-of-band data. | |
662 | */ | |
663 | static int | |
664 | tcp_usr_rcvoob(struct socket *so, struct mbuf *m, int flags) | |
665 | { | |
666 | int error = 0; | |
667 | struct inpcb *inp = sotoinpcb(so); | |
668 | struct tcpcb *tp; | |
669 | ||
670 | COMMON_START(); | |
671 | if ((so->so_oobmark == 0 && | |
672 | (so->so_state & SS_RCVATMARK) == 0) || | |
673 | so->so_options & SO_OOBINLINE || | |
674 | tp->t_oobflags & TCPOOB_HADDATA) { | |
675 | error = EINVAL; | |
676 | goto out; | |
677 | } | |
678 | if ((tp->t_oobflags & TCPOOB_HAVEDATA) == 0) { | |
679 | error = EWOULDBLOCK; | |
680 | goto out; | |
681 | } | |
682 | m->m_len = 1; | |
683 | *mtod(m, caddr_t) = tp->t_iobc; | |
684 | if ((flags & MSG_PEEK) == 0) | |
685 | tp->t_oobflags ^= (TCPOOB_HAVEDATA | TCPOOB_HADDATA); | |
686 | COMMON_END(PRU_RCVOOB); | |
687 | } | |
688 | ||
689 | /* xxx - should be const */ | |
690 | struct pr_usrreqs tcp_usrreqs = { | |
691 | tcp_usr_abort, tcp_usr_accept, tcp_usr_attach, tcp_usr_bind, | |
692 | tcp_usr_connect, pru_connect2_notsupp, in_control, tcp_usr_detach, | |
693 | tcp_usr_disconnect, tcp_usr_listen, in_setpeeraddr, tcp_usr_rcvd, | |
694 | tcp_usr_rcvoob, tcp_usr_send, pru_sense_null, tcp_usr_shutdown, | |
695 | in_setsockaddr, sosend, soreceive, pru_sopoll_notsupp | |
696 | }; | |
697 | ||
698 | #if INET6 | |
699 | struct pr_usrreqs tcp6_usrreqs = { | |
700 | tcp_usr_abort, tcp6_usr_accept, tcp_usr_attach, tcp6_usr_bind, | |
701 | tcp6_usr_connect, pru_connect2_notsupp, in6_control, tcp_usr_detach, | |
702 | tcp_usr_disconnect, tcp6_usr_listen, in6_mapped_peeraddr, tcp_usr_rcvd, | |
703 | tcp_usr_rcvoob, tcp_usr_send, pru_sense_null, tcp_usr_shutdown, | |
704 | in6_mapped_sockaddr, sosend, soreceive, pru_sopoll_notsupp | |
705 | }; | |
706 | #endif /* INET6 */ | |
707 | ||
708 | /* | |
709 | * Common subroutine to open a TCP connection to remote host specified | |
710 | * by struct sockaddr_in in mbuf *nam. Call in_pcbbind to assign a local | |
711 | * port number if needed. Call in_pcbladdr to do the routing and to choose | |
712 | * a local host address (interface). If there is an existing incarnation | |
713 | * of the same connection in TIME-WAIT state and if the remote host was | |
714 | * sending CC options and if the connection duration was < MSL, then | |
715 | * truncate the previous TIME-WAIT state and proceed. | |
716 | * Initialize connection parameters and enter SYN-SENT state. | |
717 | */ | |
718 | static int | |
719 | tcp_connect(tp, nam, p) | |
720 | register struct tcpcb *tp; | |
721 | struct sockaddr *nam; | |
722 | struct proc *p; | |
723 | { | |
724 | struct inpcb *inp = tp->t_inpcb, *oinp; | |
725 | struct socket *so = inp->inp_socket; | |
726 | struct tcpcb *otp; | |
727 | struct sockaddr_in *sin = (struct sockaddr_in *)nam; | |
728 | struct sockaddr_in *ifaddr; | |
729 | struct rmxp_tao *taop; | |
730 | struct rmxp_tao tao_noncached; | |
731 | int error; | |
732 | ||
733 | if (inp->inp_lport == 0) { | |
734 | error = in_pcbbind(inp, (struct sockaddr *)0, p); | |
735 | if (error) | |
736 | return error; | |
737 | } | |
738 | ||
739 | /* | |
740 | * Cannot simply call in_pcbconnect, because there might be an | |
741 | * earlier incarnation of this same connection still in | |
742 | * TIME_WAIT state, creating an ADDRINUSE error. | |
743 | */ | |
744 | error = in_pcbladdr(inp, nam, &ifaddr); | |
745 | if (error) | |
746 | return error; | |
747 | ||
748 | tcp_unlock(inp->inp_socket, 0, 0); | |
749 | oinp = in_pcblookup_hash(inp->inp_pcbinfo, | |
750 | sin->sin_addr, sin->sin_port, | |
751 | inp->inp_laddr.s_addr != INADDR_ANY ? inp->inp_laddr | |
752 | : ifaddr->sin_addr, | |
753 | inp->inp_lport, 0, NULL); | |
754 | ||
755 | tcp_lock(inp->inp_socket, 0, 0); | |
756 | if (oinp) { | |
757 | if (oinp != inp) /* 4143933: avoid deadlock if inp == oinp */ | |
758 | tcp_lock(oinp->inp_socket, 1, 0); | |
759 | if (in_pcb_checkstate(oinp, WNT_RELEASE, 1) == WNT_STOPUSING) { | |
760 | if (oinp != inp) | |
761 | tcp_unlock(oinp->inp_socket, 1, 0); | |
762 | goto skip_oinp; | |
763 | } | |
764 | ||
765 | if (oinp != inp && (otp = intotcpcb(oinp)) != NULL && | |
766 | otp->t_state == TCPS_TIME_WAIT && | |
767 | otp->t_starttime < tcp_msl && | |
768 | (otp->t_flags & TF_RCVD_CC)) | |
769 | otp = tcp_close(otp); | |
770 | else { | |
771 | printf("tcp_connect: inp=%x err=EADDRINUSE\n", inp); | |
772 | if (oinp != inp) | |
773 | tcp_unlock(oinp->inp_socket, 1, 0); | |
774 | return EADDRINUSE; | |
775 | } | |
776 | if (oinp != inp) | |
777 | tcp_unlock(oinp->inp_socket, 1, 0); | |
778 | } | |
779 | skip_oinp: | |
780 | if ((inp->inp_laddr.s_addr == INADDR_ANY ? ifaddr->sin_addr.s_addr : | |
781 | inp->inp_laddr.s_addr) == sin->sin_addr.s_addr && | |
782 | inp->inp_lport == sin->sin_port) | |
783 | return EINVAL; | |
784 | if (!lck_rw_try_lock_exclusive(inp->inp_pcbinfo->mtx)) { | |
785 | /*lock inversion issue, mostly with udp multicast packets */ | |
786 | socket_unlock(inp->inp_socket, 0); | |
787 | lck_rw_lock_exclusive(inp->inp_pcbinfo->mtx); | |
788 | socket_lock(inp->inp_socket, 0); | |
789 | } | |
790 | if (inp->inp_laddr.s_addr == INADDR_ANY) | |
791 | inp->inp_laddr = ifaddr->sin_addr; | |
792 | inp->inp_faddr = sin->sin_addr; | |
793 | inp->inp_fport = sin->sin_port; | |
794 | in_pcbrehash(inp); | |
795 | lck_rw_done(inp->inp_pcbinfo->mtx); | |
796 | ||
797 | /* Compute window scaling to request. */ | |
798 | while (tp->request_r_scale < TCP_MAX_WINSHIFT && | |
799 | (TCP_MAXWIN << tp->request_r_scale) < so->so_rcv.sb_hiwat) | |
800 | tp->request_r_scale++; | |
801 | ||
802 | soisconnecting(so); | |
803 | tcpstat.tcps_connattempt++; | |
804 | tp->t_state = TCPS_SYN_SENT; | |
805 | tp->t_timer[TCPT_KEEP] = tcp_keepinit; | |
806 | tp->iss = tcp_new_isn(tp); | |
807 | tcp_sendseqinit(tp); | |
808 | ||
809 | /* | |
810 | * Generate a CC value for this connection and | |
811 | * check whether CC or CCnew should be used. | |
812 | */ | |
813 | if ((taop = tcp_gettaocache(tp->t_inpcb)) == NULL) { | |
814 | taop = &tao_noncached; | |
815 | bzero(taop, sizeof(*taop)); | |
816 | } | |
817 | ||
818 | tp->cc_send = CC_INC(tcp_ccgen); | |
819 | if (taop->tao_ccsent != 0 && | |
820 | CC_GEQ(tp->cc_send, taop->tao_ccsent)) { | |
821 | taop->tao_ccsent = tp->cc_send; | |
822 | } else { | |
823 | taop->tao_ccsent = 0; | |
824 | tp->t_flags |= TF_SENDCCNEW; | |
825 | } | |
826 | ||
827 | return 0; | |
828 | } | |
829 | ||
830 | #if INET6 | |
831 | static int | |
832 | tcp6_connect(tp, nam, p) | |
833 | register struct tcpcb *tp; | |
834 | struct sockaddr *nam; | |
835 | struct proc *p; | |
836 | { | |
837 | struct inpcb *inp = tp->t_inpcb, *oinp; | |
838 | struct socket *so = inp->inp_socket; | |
839 | struct tcpcb *otp; | |
840 | struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *)nam; | |
841 | struct in6_addr addr6; | |
842 | struct rmxp_tao *taop; | |
843 | struct rmxp_tao tao_noncached; | |
844 | int error; | |
845 | ||
846 | if (inp->inp_lport == 0) { | |
847 | error = in6_pcbbind(inp, (struct sockaddr *)0, p); | |
848 | if (error) | |
849 | return error; | |
850 | } | |
851 | ||
852 | /* | |
853 | * Cannot simply call in_pcbconnect, because there might be an | |
854 | * earlier incarnation of this same connection still in | |
855 | * TIME_WAIT state, creating an ADDRINUSE error. | |
856 | */ | |
857 | error = in6_pcbladdr(inp, nam, &addr6); | |
858 | if (error) | |
859 | return error; | |
860 | tcp_unlock(inp->inp_socket, 0, 0); | |
861 | oinp = in6_pcblookup_hash(inp->inp_pcbinfo, | |
862 | &sin6->sin6_addr, sin6->sin6_port, | |
863 | IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr) | |
864 | ? &addr6 | |
865 | : &inp->in6p_laddr, | |
866 | inp->inp_lport, 0, NULL); | |
867 | tcp_lock(inp->inp_socket, 0, 0); | |
868 | if (oinp) { | |
869 | if (oinp != inp && (otp = intotcpcb(oinp)) != NULL && | |
870 | otp->t_state == TCPS_TIME_WAIT && | |
871 | otp->t_starttime < tcp_msl && | |
872 | (otp->t_flags & TF_RCVD_CC)) | |
873 | otp = tcp_close(otp); | |
874 | else | |
875 | return EADDRINUSE; | |
876 | } | |
877 | if (!lck_rw_try_lock_exclusive(inp->inp_pcbinfo->mtx)) { | |
878 | /*lock inversion issue, mostly with udp multicast packets */ | |
879 | socket_unlock(inp->inp_socket, 0); | |
880 | lck_rw_lock_exclusive(inp->inp_pcbinfo->mtx); | |
881 | socket_lock(inp->inp_socket, 0); | |
882 | } | |
883 | if (IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr)) | |
884 | inp->in6p_laddr = addr6; | |
885 | inp->in6p_faddr = sin6->sin6_addr; | |
886 | inp->inp_fport = sin6->sin6_port; | |
887 | if ((sin6->sin6_flowinfo & IPV6_FLOWINFO_MASK) != NULL) | |
888 | inp->in6p_flowinfo = sin6->sin6_flowinfo; | |
889 | in_pcbrehash(inp); | |
890 | lck_rw_done(inp->inp_pcbinfo->mtx); | |
891 | ||
892 | /* Compute window scaling to request. */ | |
893 | while (tp->request_r_scale < TCP_MAX_WINSHIFT && | |
894 | (TCP_MAXWIN << tp->request_r_scale) < so->so_rcv.sb_hiwat) | |
895 | tp->request_r_scale++; | |
896 | ||
897 | soisconnecting(so); | |
898 | tcpstat.tcps_connattempt++; | |
899 | tp->t_state = TCPS_SYN_SENT; | |
900 | tp->t_timer[TCPT_KEEP] = tcp_keepinit; | |
901 | tp->iss = tcp_new_isn(tp); | |
902 | tcp_sendseqinit(tp); | |
903 | ||
904 | /* | |
905 | * Generate a CC value for this connection and | |
906 | * check whether CC or CCnew should be used. | |
907 | */ | |
908 | if ((taop = tcp_gettaocache(tp->t_inpcb)) == NULL) { | |
909 | taop = &tao_noncached; | |
910 | bzero(taop, sizeof(*taop)); | |
911 | } | |
912 | ||
913 | tp->cc_send = CC_INC(tcp_ccgen); | |
914 | if (taop->tao_ccsent != 0 && | |
915 | CC_GEQ(tp->cc_send, taop->tao_ccsent)) { | |
916 | taop->tao_ccsent = tp->cc_send; | |
917 | } else { | |
918 | taop->tao_ccsent = 0; | |
919 | tp->t_flags |= TF_SENDCCNEW; | |
920 | } | |
921 | ||
922 | return 0; | |
923 | } | |
924 | #endif /* INET6 */ | |
925 | ||
926 | /* | |
927 | * The new sockopt interface makes it possible for us to block in the | |
928 | * copyin/out step (if we take a page fault). Taking a page fault at | |
929 | * splnet() is probably a Bad Thing. (Since sockets and pcbs both now | |
930 | * use TSM, there probably isn't any need for this function to run at | |
931 | * splnet() any more. This needs more examination.) | |
932 | */ | |
933 | int | |
934 | tcp_ctloutput(so, sopt) | |
935 | struct socket *so; | |
936 | struct sockopt *sopt; | |
937 | { | |
938 | int error, opt, optval; | |
939 | struct inpcb *inp; | |
940 | struct tcpcb *tp; | |
941 | ||
942 | error = 0; | |
943 | inp = sotoinpcb(so); | |
944 | if (inp == NULL) { | |
945 | return (ECONNRESET); | |
946 | } | |
947 | if (sopt->sopt_level != IPPROTO_TCP) { | |
948 | #if INET6 | |
949 | if (INP_CHECK_SOCKAF(so, AF_INET6)) | |
950 | error = ip6_ctloutput(so, sopt); | |
951 | else | |
952 | #endif /* INET6 */ | |
953 | error = ip_ctloutput(so, sopt); | |
954 | return (error); | |
955 | } | |
956 | tp = intotcpcb(inp); | |
957 | if (tp == NULL) { | |
958 | return (ECONNRESET); | |
959 | } | |
960 | ||
961 | switch (sopt->sopt_dir) { | |
962 | case SOPT_SET: | |
963 | switch (sopt->sopt_name) { | |
964 | case TCP_NODELAY: | |
965 | case TCP_NOOPT: | |
966 | case TCP_NOPUSH: | |
967 | error = sooptcopyin(sopt, &optval, sizeof optval, | |
968 | sizeof optval); | |
969 | if (error) | |
970 | break; | |
971 | ||
972 | switch (sopt->sopt_name) { | |
973 | case TCP_NODELAY: | |
974 | opt = TF_NODELAY; | |
975 | break; | |
976 | case TCP_NOOPT: | |
977 | opt = TF_NOOPT; | |
978 | break; | |
979 | case TCP_NOPUSH: | |
980 | opt = TF_NOPUSH; | |
981 | break; | |
982 | default: | |
983 | opt = 0; /* dead code to fool gcc */ | |
984 | break; | |
985 | } | |
986 | ||
987 | if (optval) | |
988 | tp->t_flags |= opt; | |
989 | else | |
990 | tp->t_flags &= ~opt; | |
991 | break; | |
992 | ||
993 | case TCP_MAXSEG: | |
994 | error = sooptcopyin(sopt, &optval, sizeof optval, | |
995 | sizeof optval); | |
996 | if (error) | |
997 | break; | |
998 | ||
999 | if (optval > 0 && optval <= tp->t_maxseg && | |
1000 | optval + 40 >= tcp_minmss) | |
1001 | tp->t_maxseg = optval; | |
1002 | else | |
1003 | error = EINVAL; | |
1004 | break; | |
1005 | ||
1006 | case TCP_KEEPALIVE: | |
1007 | error = sooptcopyin(sopt, &optval, sizeof optval, | |
1008 | sizeof optval); | |
1009 | if (error) | |
1010 | break; | |
1011 | if (optval < 0) | |
1012 | error = EINVAL; | |
1013 | else | |
1014 | tp->t_keepidle = optval * PR_SLOWHZ; | |
1015 | break; | |
1016 | ||
1017 | default: | |
1018 | error = ENOPROTOOPT; | |
1019 | break; | |
1020 | } | |
1021 | break; | |
1022 | ||
1023 | case SOPT_GET: | |
1024 | switch (sopt->sopt_name) { | |
1025 | case TCP_NODELAY: | |
1026 | optval = tp->t_flags & TF_NODELAY; | |
1027 | break; | |
1028 | case TCP_MAXSEG: | |
1029 | optval = tp->t_maxseg; | |
1030 | break; | |
1031 | case TCP_KEEPALIVE: | |
1032 | optval = tp->t_keepidle / PR_SLOWHZ; | |
1033 | break; | |
1034 | case TCP_NOOPT: | |
1035 | optval = tp->t_flags & TF_NOOPT; | |
1036 | break; | |
1037 | case TCP_NOPUSH: | |
1038 | optval = tp->t_flags & TF_NOPUSH; | |
1039 | break; | |
1040 | default: | |
1041 | error = ENOPROTOOPT; | |
1042 | break; | |
1043 | } | |
1044 | if (error == 0) | |
1045 | error = sooptcopyout(sopt, &optval, sizeof optval); | |
1046 | break; | |
1047 | } | |
1048 | return (error); | |
1049 | } | |
1050 | ||
1051 | /* | |
1052 | * tcp_sendspace and tcp_recvspace are the default send and receive window | |
1053 | * sizes, respectively. These are obsolescent (this information should | |
1054 | * be set by the route). | |
1055 | */ | |
1056 | u_long tcp_sendspace = 1024*16; | |
1057 | SYSCTL_INT(_net_inet_tcp, TCPCTL_SENDSPACE, sendspace, CTLFLAG_RW, | |
1058 | &tcp_sendspace , 0, "Maximum outgoing TCP datagram size"); | |
1059 | u_long tcp_recvspace = 1024*16; | |
1060 | SYSCTL_INT(_net_inet_tcp, TCPCTL_RECVSPACE, recvspace, CTLFLAG_RW, | |
1061 | &tcp_recvspace , 0, "Maximum incoming TCP datagram size"); | |
1062 | ||
1063 | __private_extern__ int tcp_sockthreshold = 256; | |
1064 | SYSCTL_INT(_net_inet_tcp, OID_AUTO, sockthreshold, CTLFLAG_RW, | |
1065 | &tcp_sockthreshold , 0, "TCP Socket size increased if less than threshold"); | |
1066 | ||
1067 | #define TCP_INCREASED_SPACE 65535 /* Automatically increase tcp send/rcv space to this value */ | |
1068 | /* | |
1069 | * Attach TCP protocol to socket, allocating | |
1070 | * internet protocol control block, tcp control block, | |
1071 | * bufer space, and entering LISTEN state if to accept connections. | |
1072 | */ | |
1073 | static int | |
1074 | tcp_attach(so, p) | |
1075 | struct socket *so; | |
1076 | struct proc *p; | |
1077 | { | |
1078 | register struct tcpcb *tp; | |
1079 | struct inpcb *inp; | |
1080 | int error; | |
1081 | #if INET6 | |
1082 | int isipv6 = INP_CHECK_SOCKAF(so, AF_INET6) != NULL; | |
1083 | #endif | |
1084 | ||
1085 | error = in_pcballoc(so, &tcbinfo, p); | |
1086 | if (error) | |
1087 | return (error); | |
1088 | ||
1089 | inp = sotoinpcb(so); | |
1090 | ||
1091 | if (so->so_snd.sb_hiwat == 0 || so->so_rcv.sb_hiwat == 0) { | |
1092 | /* | |
1093 | * The goal is to let clients have large send/rcv default windows (TCP_INCREASED_SPACE) | |
1094 | * while not hogging mbuf space for servers. This is done by watching a threshold | |
1095 | * of tcpcbs in use and bumping the default send and rcvspace only if under that threshold. | |
1096 | * The theory being that busy servers have a lot more active tcpcbs and don't want the potential | |
1097 | * memory penalty of having much larger sockbuffs. The sysctl allows to fine tune that threshold value. */ | |
1098 | ||
1099 | if (inp->inp_pcbinfo->ipi_count < tcp_sockthreshold) | |
1100 | error = soreserve(so, MAX(TCP_INCREASED_SPACE, tcp_sendspace), MAX(TCP_INCREASED_SPACE,tcp_recvspace)); | |
1101 | else | |
1102 | error = soreserve(so, tcp_sendspace, tcp_recvspace); | |
1103 | if (error) | |
1104 | return (error); | |
1105 | } | |
1106 | ||
1107 | #if INET6 | |
1108 | if (isipv6) { | |
1109 | inp->inp_vflag |= INP_IPV6; | |
1110 | inp->in6p_hops = -1; /* use kernel default */ | |
1111 | } | |
1112 | else | |
1113 | #endif /* INET6 */ | |
1114 | inp->inp_vflag |= INP_IPV4; | |
1115 | tp = tcp_newtcpcb(inp); | |
1116 | if (tp == 0) { | |
1117 | int nofd = so->so_state & SS_NOFDREF; /* XXX */ | |
1118 | ||
1119 | so->so_state &= ~SS_NOFDREF; /* don't free the socket yet */ | |
1120 | #if INET6 | |
1121 | if (isipv6) | |
1122 | in6_pcbdetach(inp); | |
1123 | else | |
1124 | #endif /* INET6 */ | |
1125 | in_pcbdetach(inp); | |
1126 | so->so_state |= nofd; | |
1127 | return (ENOBUFS); | |
1128 | } | |
1129 | tp->t_state = TCPS_CLOSED; | |
1130 | return (0); | |
1131 | } | |
1132 | ||
1133 | /* | |
1134 | * Initiate (or continue) disconnect. | |
1135 | * If embryonic state, just send reset (once). | |
1136 | * If in ``let data drain'' option and linger null, just drop. | |
1137 | * Otherwise (hard), mark socket disconnecting and drop | |
1138 | * current input data; switch states based on user close, and | |
1139 | * send segment to peer (with FIN). | |
1140 | */ | |
1141 | static struct tcpcb * | |
1142 | tcp_disconnect(tp) | |
1143 | register struct tcpcb *tp; | |
1144 | { | |
1145 | struct socket *so = tp->t_inpcb->inp_socket; | |
1146 | ||
1147 | if (tp->t_state < TCPS_ESTABLISHED) | |
1148 | tp = tcp_close(tp); | |
1149 | else if ((so->so_options & SO_LINGER) && so->so_linger == 0) | |
1150 | tp = tcp_drop(tp, 0); | |
1151 | else { | |
1152 | soisdisconnecting(so); | |
1153 | sbflush(&so->so_rcv); | |
1154 | tp = tcp_usrclosed(tp); | |
1155 | if (tp) | |
1156 | (void) tcp_output(tp); | |
1157 | } | |
1158 | return (tp); | |
1159 | } | |
1160 | ||
1161 | /* | |
1162 | * User issued close, and wish to trail through shutdown states: | |
1163 | * if never received SYN, just forget it. If got a SYN from peer, | |
1164 | * but haven't sent FIN, then go to FIN_WAIT_1 state to send peer a FIN. | |
1165 | * If already got a FIN from peer, then almost done; go to LAST_ACK | |
1166 | * state. In all other cases, have already sent FIN to peer (e.g. | |
1167 | * after PRU_SHUTDOWN), and just have to play tedious game waiting | |
1168 | * for peer to send FIN or not respond to keep-alives, etc. | |
1169 | * We can let the user exit from the close as soon as the FIN is acked. | |
1170 | */ | |
1171 | static struct tcpcb * | |
1172 | tcp_usrclosed(tp) | |
1173 | register struct tcpcb *tp; | |
1174 | { | |
1175 | ||
1176 | switch (tp->t_state) { | |
1177 | ||
1178 | case TCPS_CLOSED: | |
1179 | case TCPS_LISTEN: | |
1180 | tp->t_state = TCPS_CLOSED; | |
1181 | tp = tcp_close(tp); | |
1182 | break; | |
1183 | ||
1184 | case TCPS_SYN_SENT: | |
1185 | case TCPS_SYN_RECEIVED: | |
1186 | tp->t_flags |= TF_NEEDFIN; | |
1187 | break; | |
1188 | ||
1189 | case TCPS_ESTABLISHED: | |
1190 | tp->t_state = TCPS_FIN_WAIT_1; | |
1191 | break; | |
1192 | ||
1193 | case TCPS_CLOSE_WAIT: | |
1194 | tp->t_state = TCPS_LAST_ACK; | |
1195 | break; | |
1196 | } | |
1197 | if (tp && tp->t_state >= TCPS_FIN_WAIT_2) { | |
1198 | soisdisconnected(tp->t_inpcb->inp_socket); | |
1199 | /* To prevent the connection hanging in FIN_WAIT_2 forever. */ | |
1200 | if (tp->t_state == TCPS_FIN_WAIT_2) | |
1201 | tp->t_timer[TCPT_2MSL] = tcp_maxidle; | |
1202 | } | |
1203 | return (tp); | |
1204 | } | |
1205 |