xref: /netbsd-src/sys/netinet/tcp_usrreq.c (revision 6ea46cb5e46c49111a6ecf3bcbe3c7e2730fe9f6)
1 /*	$NetBSD: tcp_usrreq.c,v 1.10 1994/06/29 06:38:52 cgd Exp $	*/
2 
3 /*
4  * Copyright (c) 1982, 1986, 1988, 1993
5  *	The Regents of the University of California.  All rights reserved.
6  *
7  * Redistribution and use in source and binary forms, with or without
8  * modification, are permitted provided that the following conditions
9  * are met:
10  * 1. Redistributions of source code must retain the above copyright
11  *    notice, this list of conditions and the following disclaimer.
12  * 2. Redistributions in binary form must reproduce the above copyright
13  *    notice, this list of conditions and the following disclaimer in the
14  *    documentation and/or other materials provided with the distribution.
15  * 3. All advertising materials mentioning features or use of this software
16  *    must display the following acknowledgement:
17  *	This product includes software developed by the University of
18  *	California, Berkeley and its contributors.
19  * 4. Neither the name of the University nor the names of its contributors
20  *    may be used to endorse or promote products derived from this software
21  *    without specific prior written permission.
22  *
23  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
24  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
25  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
26  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
27  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
28  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
29  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
30  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
31  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
32  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
33  * SUCH DAMAGE.
34  *
35  *	@(#)tcp_usrreq.c	8.2 (Berkeley) 1/3/94
36  */
37 
38 #include <sys/param.h>
39 #include <sys/systm.h>
40 #include <sys/malloc.h>
41 #include <sys/mbuf.h>
42 #include <sys/socket.h>
43 #include <sys/socketvar.h>
44 #include <sys/protosw.h>
45 #include <sys/errno.h>
46 #include <sys/stat.h>
47 
48 #include <net/if.h>
49 #include <net/route.h>
50 
51 #include <netinet/in.h>
52 #include <netinet/in_systm.h>
53 #include <netinet/ip.h>
54 #include <netinet/in_pcb.h>
55 #include <netinet/ip_var.h>
56 #include <netinet/tcp.h>
57 #include <netinet/tcp_fsm.h>
58 #include <netinet/tcp_seq.h>
59 #include <netinet/tcp_timer.h>
60 #include <netinet/tcp_var.h>
61 #include <netinet/tcpip.h>
62 #include <netinet/tcp_debug.h>
63 
64 /*
65  * TCP protocol interface to socket abstraction.
66  */
67 extern	char *tcpstates[];
68 
69 /*
70  * Process a TCP user request for TCP tb.  If this is a send request
71  * then m is the mbuf chain of send data.  If this is a timer expiration
72  * (called from the software clock routine), then timertype tells which timer.
73  */
74 /*ARGSUSED*/
75 int
76 tcp_usrreq(so, req, m, nam, control)
77 	struct socket *so;
78 	int req;
79 	struct mbuf *m, *nam, *control;
80 {
81 	register struct inpcb *inp;
82 	register struct tcpcb *tp;
83 	int s;
84 	int error = 0;
85 	int ostate;
86 
87 	if (req == PRU_CONTROL)
88 		return (in_control(so, (int)m, (caddr_t)nam,
89 			(struct ifnet *)control));
90 	if (control && control->m_len) {
91 		m_freem(control);
92 		if (m)
93 			m_freem(m);
94 		return (EINVAL);
95 	}
96 
97 	s = splnet();
98 	inp = sotoinpcb(so);
99 	/*
100 	 * When a TCP is attached to a socket, then there will be
101 	 * a (struct inpcb) pointed at by the socket, and this
102 	 * structure will point at a subsidary (struct tcpcb).
103 	 */
104 	if (inp == 0 && req != PRU_ATTACH) {
105 		splx(s);
106 		return (EINVAL);		/* XXX */
107 	}
108 	if (inp) {
109 		tp = intotcpcb(inp);
110 		/* WHAT IF TP IS 0? */
111 #ifdef KPROF
112 		tcp_acounts[tp->t_state][req]++;
113 #endif
114 		ostate = tp->t_state;
115 	} else
116 		ostate = 0;
117 	switch (req) {
118 
119 	/*
120 	 * TCP attaches to socket via PRU_ATTACH, reserving space,
121 	 * and an internet control block.
122 	 */
123 	case PRU_ATTACH:
124 		if (inp) {
125 			error = EISCONN;
126 			break;
127 		}
128 		error = tcp_attach(so);
129 		if (error)
130 			break;
131 		if ((so->so_options & SO_LINGER) && so->so_linger == 0)
132 			so->so_linger = TCP_LINGERTIME;
133 		tp = sototcpcb(so);
134 		break;
135 
136 	/*
137 	 * PRU_DETACH detaches the TCP protocol from the socket.
138 	 * If the protocol state is non-embryonic, then can't
139 	 * do this directly: have to initiate a PRU_DISCONNECT,
140 	 * which may finish later; embryonic TCB's can just
141 	 * be discarded here.
142 	 */
143 	case PRU_DETACH:
144 		if (tp->t_state > TCPS_LISTEN)
145 			tp = tcp_disconnect(tp);
146 		else
147 			tp = tcp_close(tp);
148 		break;
149 
150 	/*
151 	 * Give the socket an address.
152 	 */
153 	case PRU_BIND:
154 		error = in_pcbbind(inp, nam);
155 		if (error)
156 			break;
157 		break;
158 
159 	/*
160 	 * Prepare to accept connections.
161 	 */
162 	case PRU_LISTEN:
163 		if (inp->inp_lport == 0)
164 			error = in_pcbbind(inp, (struct mbuf *)0);
165 		if (error == 0)
166 			tp->t_state = TCPS_LISTEN;
167 		break;
168 
169 	/*
170 	 * Initiate connection to peer.
171 	 * Create a template for use in transmissions on this connection.
172 	 * Enter SYN_SENT state, and mark socket as connecting.
173 	 * Start keep-alive timer, and seed output sequence space.
174 	 * Send initial segment on connection.
175 	 */
176 	case PRU_CONNECT:
177 		if (inp->inp_lport == 0) {
178 			error = in_pcbbind(inp, (struct mbuf *)0);
179 			if (error)
180 				break;
181 		}
182 		error = in_pcbconnect(inp, nam);
183 		if (error)
184 			break;
185 		tp->t_template = tcp_template(tp);
186 		if (tp->t_template == 0) {
187 			in_pcbdisconnect(inp);
188 			error = ENOBUFS;
189 			break;
190 		}
191 		/* Compute window scaling to request.  */
192 		while (tp->request_r_scale < TCP_MAX_WINSHIFT &&
193 		    (TCP_MAXWIN << tp->request_r_scale) < so->so_rcv.sb_hiwat)
194 			tp->request_r_scale++;
195 		soisconnecting(so);
196 		tcpstat.tcps_connattempt++;
197 		tp->t_state = TCPS_SYN_SENT;
198 		tp->t_timer[TCPT_KEEP] = TCPTV_KEEP_INIT;
199 		tp->iss = tcp_iss; tcp_iss += TCP_ISSINCR/2;
200 		tcp_sendseqinit(tp);
201 		error = tcp_output(tp);
202 		break;
203 
204 	/*
205 	 * Create a TCP connection between two sockets.
206 	 */
207 	case PRU_CONNECT2:
208 		error = EOPNOTSUPP;
209 		break;
210 
211 	/*
212 	 * Initiate disconnect from peer.
213 	 * If connection never passed embryonic stage, just drop;
214 	 * else if don't need to let data drain, then can just drop anyways,
215 	 * else have to begin TCP shutdown process: mark socket disconnecting,
216 	 * drain unread data, state switch to reflect user close, and
217 	 * send segment (e.g. FIN) to peer.  Socket will be really disconnected
218 	 * when peer sends FIN and acks ours.
219 	 *
220 	 * SHOULD IMPLEMENT LATER PRU_CONNECT VIA REALLOC TCPCB.
221 	 */
222 	case PRU_DISCONNECT:
223 		tp = tcp_disconnect(tp);
224 		break;
225 
226 	/*
227 	 * Accept a connection.  Essentially all the work is
228 	 * done at higher levels; just return the address
229 	 * of the peer, storing through addr.
230 	 */
231 	case PRU_ACCEPT:
232 		in_setpeeraddr(inp, nam);
233 		break;
234 
235 	/*
236 	 * Mark the connection as being incapable of further output.
237 	 */
238 	case PRU_SHUTDOWN:
239 		socantsendmore(so);
240 		tp = tcp_usrclosed(tp);
241 		if (tp)
242 			error = tcp_output(tp);
243 		break;
244 
245 	/*
246 	 * After a receive, possibly send window update to peer.
247 	 */
248 	case PRU_RCVD:
249 		(void) tcp_output(tp);
250 		break;
251 
252 	/*
253 	 * Do a send by putting data in output queue and updating urgent
254 	 * marker if URG set.  Possibly send more data.
255 	 */
256 	case PRU_SEND:
257 		sbappend(&so->so_snd, m);
258 		error = tcp_output(tp);
259 		break;
260 
261 	/*
262 	 * Abort the TCP.
263 	 */
264 	case PRU_ABORT:
265 		tp = tcp_drop(tp, ECONNABORTED);
266 		break;
267 
268 	case PRU_SENSE:
269 		((struct stat *) m)->st_blksize = so->so_snd.sb_hiwat;
270 		(void) splx(s);
271 		return (0);
272 
273 	case PRU_RCVOOB:
274 		if ((so->so_oobmark == 0 &&
275 		    (so->so_state & SS_RCVATMARK) == 0) ||
276 		    so->so_options & SO_OOBINLINE ||
277 		    tp->t_oobflags & TCPOOB_HADDATA) {
278 			error = EINVAL;
279 			break;
280 		}
281 		if ((tp->t_oobflags & TCPOOB_HAVEDATA) == 0) {
282 			error = EWOULDBLOCK;
283 			break;
284 		}
285 		m->m_len = 1;
286 		*mtod(m, caddr_t) = tp->t_iobc;
287 		if (((int)nam & MSG_PEEK) == 0)
288 			tp->t_oobflags ^= (TCPOOB_HAVEDATA | TCPOOB_HADDATA);
289 		break;
290 
291 	case PRU_SENDOOB:
292 		if (sbspace(&so->so_snd) < -512) {
293 			m_freem(m);
294 			error = ENOBUFS;
295 			break;
296 		}
297 		/*
298 		 * According to RFC961 (Assigned Protocols),
299 		 * the urgent pointer points to the last octet
300 		 * of urgent data.  We continue, however,
301 		 * to consider it to indicate the first octet
302 		 * of data past the urgent section.
303 		 * Otherwise, snd_up should be one lower.
304 		 */
305 		sbappend(&so->so_snd, m);
306 		tp->snd_up = tp->snd_una + so->so_snd.sb_cc;
307 		tp->t_force = 1;
308 		error = tcp_output(tp);
309 		tp->t_force = 0;
310 		break;
311 
312 	case PRU_SOCKADDR:
313 		in_setsockaddr(inp, nam);
314 		break;
315 
316 	case PRU_PEERADDR:
317 		in_setpeeraddr(inp, nam);
318 		break;
319 
320 	/*
321 	 * TCP slow timer went off; going through this
322 	 * routine for tracing's sake.
323 	 */
324 	case PRU_SLOWTIMO:
325 		tp = tcp_timers(tp, (int)nam);
326 		req |= (int)nam << 8;		/* for debug's sake */
327 		break;
328 
329 	default:
330 		panic("tcp_usrreq");
331 	}
332 	if (tp && (so->so_options & SO_DEBUG))
333 		tcp_trace(TA_USER, ostate, tp, (struct tcpiphdr *)0, req);
334 	splx(s);
335 	return (error);
336 }
337 
338 int
339 tcp_ctloutput(op, so, level, optname, mp)
340 	int op;
341 	struct socket *so;
342 	int level, optname;
343 	struct mbuf **mp;
344 {
345 	int error = 0, s;
346 	struct inpcb *inp;
347 	register struct tcpcb *tp;
348 	register struct mbuf *m;
349 	register int i;
350 
351 	s = splnet();
352 	inp = sotoinpcb(so);
353 	if (inp == NULL) {
354 		splx(s);
355 		if (op == PRCO_SETOPT && *mp)
356 			(void) m_free(*mp);
357 		return (ECONNRESET);
358 	}
359 	if (level != IPPROTO_TCP) {
360 		error = ip_ctloutput(op, so, level, optname, mp);
361 		splx(s);
362 		return (error);
363 	}
364 	tp = intotcpcb(inp);
365 
366 	switch (op) {
367 
368 	case PRCO_SETOPT:
369 		m = *mp;
370 		switch (optname) {
371 
372 		case TCP_NODELAY:
373 			if (m == NULL || m->m_len < sizeof (int))
374 				error = EINVAL;
375 			else if (*mtod(m, int *))
376 				tp->t_flags |= TF_NODELAY;
377 			else
378 				tp->t_flags &= ~TF_NODELAY;
379 			break;
380 
381 		case TCP_MAXSEG:
382 			if (m && (i = *mtod(m, int *)) > 0 && i <= tp->t_maxseg)
383 				tp->t_maxseg = i;
384 			else
385 				error = EINVAL;
386 			break;
387 
388 		default:
389 			error = ENOPROTOOPT;
390 			break;
391 		}
392 		if (m)
393 			(void) m_free(m);
394 		break;
395 
396 	case PRCO_GETOPT:
397 		*mp = m = m_get(M_WAIT, MT_SOOPTS);
398 		m->m_len = sizeof(int);
399 
400 		switch (optname) {
401 		case TCP_NODELAY:
402 			*mtod(m, int *) = tp->t_flags & TF_NODELAY;
403 			break;
404 		case TCP_MAXSEG:
405 			*mtod(m, int *) = tp->t_maxseg;
406 			break;
407 		default:
408 			error = ENOPROTOOPT;
409 			break;
410 		}
411 		break;
412 	}
413 	splx(s);
414 	return (error);
415 }
416 
417 u_long	tcp_sendspace = 1024*8;
418 u_long	tcp_recvspace = 1024*8;
419 
420 /*
421  * Attach TCP protocol to socket, allocating
422  * internet protocol control block, tcp control block,
423  * bufer space, and entering LISTEN state if to accept connections.
424  */
425 int
426 tcp_attach(so)
427 	struct socket *so;
428 {
429 	register struct tcpcb *tp;
430 	struct inpcb *inp;
431 	int error;
432 
433 	if (so->so_snd.sb_hiwat == 0 || so->so_rcv.sb_hiwat == 0) {
434 		error = soreserve(so, tcp_sendspace, tcp_recvspace);
435 		if (error)
436 			return (error);
437 	}
438 	error = in_pcballoc(so, &tcb);
439 	if (error)
440 		return (error);
441 	inp = sotoinpcb(so);
442 	tp = tcp_newtcpcb(inp);
443 	if (tp == 0) {
444 		int nofd = so->so_state & SS_NOFDREF;	/* XXX */
445 
446 		so->so_state &= ~SS_NOFDREF;	/* don't free the socket yet */
447 		in_pcbdetach(inp);
448 		so->so_state |= nofd;
449 		return (ENOBUFS);
450 	}
451 	tp->t_state = TCPS_CLOSED;
452 	return (0);
453 }
454 
455 /*
456  * Initiate (or continue) disconnect.
457  * If embryonic state, just send reset (once).
458  * If in ``let data drain'' option and linger null, just drop.
459  * Otherwise (hard), mark socket disconnecting and drop
460  * current input data; switch states based on user close, and
461  * send segment to peer (with FIN).
462  */
463 struct tcpcb *
464 tcp_disconnect(tp)
465 	register struct tcpcb *tp;
466 {
467 	struct socket *so = tp->t_inpcb->inp_socket;
468 
469 	if (tp->t_state < TCPS_ESTABLISHED)
470 		tp = tcp_close(tp);
471 	else if ((so->so_options & SO_LINGER) && so->so_linger == 0)
472 		tp = tcp_drop(tp, 0);
473 	else {
474 		soisdisconnecting(so);
475 		sbflush(&so->so_rcv);
476 		tp = tcp_usrclosed(tp);
477 		if (tp)
478 			(void) tcp_output(tp);
479 	}
480 	return (tp);
481 }
482 
483 /*
484  * User issued close, and wish to trail through shutdown states:
485  * if never received SYN, just forget it.  If got a SYN from peer,
486  * but haven't sent FIN, then go to FIN_WAIT_1 state to send peer a FIN.
487  * If already got a FIN from peer, then almost done; go to LAST_ACK
488  * state.  In all other cases, have already sent FIN to peer (e.g.
489  * after PRU_SHUTDOWN), and just have to play tedious game waiting
490  * for peer to send FIN or not respond to keep-alives, etc.
491  * We can let the user exit from the close as soon as the FIN is acked.
492  */
493 struct tcpcb *
494 tcp_usrclosed(tp)
495 	register struct tcpcb *tp;
496 {
497 
498 	switch (tp->t_state) {
499 
500 	case TCPS_CLOSED:
501 	case TCPS_LISTEN:
502 	case TCPS_SYN_SENT:
503 		tp->t_state = TCPS_CLOSED;
504 		tp = tcp_close(tp);
505 		break;
506 
507 	case TCPS_SYN_RECEIVED:
508 	case TCPS_ESTABLISHED:
509 		tp->t_state = TCPS_FIN_WAIT_1;
510 		break;
511 
512 	case TCPS_CLOSE_WAIT:
513 		tp->t_state = TCPS_LAST_ACK;
514 		break;
515 	}
516 	if (tp && tp->t_state >= TCPS_FIN_WAIT_2)
517 		soisdisconnected(tp->t_inpcb->inp_socket);
518 	return (tp);
519 }
520