xref: /netbsd-src/sys/netinet/tcp_usrreq.c (revision ce0bb6e8d2e560ecacbe865a848624f94498063b)
1 /*	$NetBSD: tcp_usrreq.c,v 1.14 1995/04/13 06:36:53 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/kernel.h>
41 #include <sys/malloc.h>
42 #include <sys/mbuf.h>
43 #include <sys/socket.h>
44 #include <sys/socketvar.h>
45 #include <sys/protosw.h>
46 #include <sys/errno.h>
47 #include <sys/stat.h>
48 
49 #include <net/if.h>
50 #include <net/route.h>
51 
52 #include <netinet/in.h>
53 #include <netinet/in_systm.h>
54 #include <netinet/in_var.h>
55 #include <netinet/ip.h>
56 #include <netinet/in_pcb.h>
57 #include <netinet/ip_var.h>
58 #include <netinet/tcp.h>
59 #include <netinet/tcp_fsm.h>
60 #include <netinet/tcp_seq.h>
61 #include <netinet/tcp_timer.h>
62 #include <netinet/tcp_var.h>
63 #include <netinet/tcpip.h>
64 #include <netinet/tcp_debug.h>
65 
66 /*
67  * TCP protocol interface to socket abstraction.
68  */
69 extern	char *tcpstates[];
70 
71 /*
72  * Process a TCP user request for TCP tb.  If this is a send request
73  * then m is the mbuf chain of send data.  If this is a timer expiration
74  * (called from the software clock routine), then timertype tells which timer.
75  */
76 /*ARGSUSED*/
77 int
78 tcp_usrreq(so, req, m, nam, control)
79 	struct socket *so;
80 	int req;
81 	struct mbuf *m, *nam, *control;
82 {
83 	register struct inpcb *inp;
84 	register struct tcpcb *tp;
85 	int s;
86 	int error = 0;
87 	int ostate;
88 
89 	if (req == PRU_CONTROL)
90 		return (in_control(so, (long)m, (caddr_t)nam,
91 			(struct ifnet *)control));
92 	if (control && control->m_len) {
93 		m_freem(control);
94 		if (m)
95 			m_freem(m);
96 		return (EINVAL);
97 	}
98 
99 	s = splnet();
100 	inp = sotoinpcb(so);
101 	/*
102 	 * When a TCP is attached to a socket, then there will be
103 	 * a (struct inpcb) pointed at by the socket, and this
104 	 * structure will point at a subsidary (struct tcpcb).
105 	 */
106 	if (inp == 0 && req != PRU_ATTACH) {
107 		splx(s);
108 		return (EINVAL);		/* XXX */
109 	}
110 	if (inp) {
111 		tp = intotcpcb(inp);
112 		/* WHAT IF TP IS 0? */
113 #ifdef KPROF
114 		tcp_acounts[tp->t_state][req]++;
115 #endif
116 		ostate = tp->t_state;
117 	} else
118 		ostate = 0;
119 	switch (req) {
120 
121 	/*
122 	 * TCP attaches to socket via PRU_ATTACH, reserving space,
123 	 * and an internet control block.
124 	 */
125 	case PRU_ATTACH:
126 		if (inp) {
127 			error = EISCONN;
128 			break;
129 		}
130 		error = tcp_attach(so);
131 		if (error)
132 			break;
133 		if ((so->so_options & SO_LINGER) && so->so_linger == 0)
134 			so->so_linger = TCP_LINGERTIME * hz;
135 		tp = sototcpcb(so);
136 		break;
137 
138 	/*
139 	 * PRU_DETACH detaches the TCP protocol from the socket.
140 	 * If the protocol state is non-embryonic, then can't
141 	 * do this directly: have to initiate a PRU_DISCONNECT,
142 	 * which may finish later; embryonic TCB's can just
143 	 * be discarded here.
144 	 */
145 	case PRU_DETACH:
146 		if (tp->t_state > TCPS_LISTEN)
147 			tp = tcp_disconnect(tp);
148 		else
149 			tp = tcp_close(tp);
150 		break;
151 
152 	/*
153 	 * Give the socket an address.
154 	 */
155 	case PRU_BIND:
156 		error = in_pcbbind(inp, nam);
157 		if (error)
158 			break;
159 		break;
160 
161 	/*
162 	 * Prepare to accept connections.
163 	 */
164 	case PRU_LISTEN:
165 		if (inp->inp_lport == 0)
166 			error = in_pcbbind(inp, (struct mbuf *)0);
167 		if (error == 0)
168 			tp->t_state = TCPS_LISTEN;
169 		break;
170 
171 	/*
172 	 * Initiate connection to peer.
173 	 * Create a template for use in transmissions on this connection.
174 	 * Enter SYN_SENT state, and mark socket as connecting.
175 	 * Start keep-alive timer, and seed output sequence space.
176 	 * Send initial segment on connection.
177 	 */
178 	case PRU_CONNECT:
179 		if (inp->inp_lport == 0) {
180 			error = in_pcbbind(inp, (struct mbuf *)0);
181 			if (error)
182 				break;
183 		}
184 		error = in_pcbconnect(inp, nam);
185 		if (error)
186 			break;
187 		tp->t_template = tcp_template(tp);
188 		if (tp->t_template == 0) {
189 			in_pcbdisconnect(inp);
190 			error = ENOBUFS;
191 			break;
192 		}
193 		/* Compute window scaling to request.  */
194 		while (tp->request_r_scale < TCP_MAX_WINSHIFT &&
195 		    (TCP_MAXWIN << tp->request_r_scale) < so->so_rcv.sb_hiwat)
196 			tp->request_r_scale++;
197 		soisconnecting(so);
198 		tcpstat.tcps_connattempt++;
199 		tp->t_state = TCPS_SYN_SENT;
200 		tp->t_timer[TCPT_KEEP] = TCPTV_KEEP_INIT;
201 		tp->iss = tcp_iss; tcp_iss += TCP_ISSINCR/2;
202 		tcp_sendseqinit(tp);
203 		error = tcp_output(tp);
204 		break;
205 
206 	/*
207 	 * Create a TCP connection between two sockets.
208 	 */
209 	case PRU_CONNECT2:
210 		error = EOPNOTSUPP;
211 		break;
212 
213 	/*
214 	 * Initiate disconnect from peer.
215 	 * If connection never passed embryonic stage, just drop;
216 	 * else if don't need to let data drain, then can just drop anyways,
217 	 * else have to begin TCP shutdown process: mark socket disconnecting,
218 	 * drain unread data, state switch to reflect user close, and
219 	 * send segment (e.g. FIN) to peer.  Socket will be really disconnected
220 	 * when peer sends FIN and acks ours.
221 	 *
222 	 * SHOULD IMPLEMENT LATER PRU_CONNECT VIA REALLOC TCPCB.
223 	 */
224 	case PRU_DISCONNECT:
225 		tp = tcp_disconnect(tp);
226 		break;
227 
228 	/*
229 	 * Accept a connection.  Essentially all the work is
230 	 * done at higher levels; just return the address
231 	 * of the peer, storing through addr.
232 	 */
233 	case PRU_ACCEPT:
234 		in_setpeeraddr(inp, nam);
235 		break;
236 
237 	/*
238 	 * Mark the connection as being incapable of further output.
239 	 */
240 	case PRU_SHUTDOWN:
241 		socantsendmore(so);
242 		tp = tcp_usrclosed(tp);
243 		if (tp)
244 			error = tcp_output(tp);
245 		break;
246 
247 	/*
248 	 * After a receive, possibly send window update to peer.
249 	 */
250 	case PRU_RCVD:
251 		(void) tcp_output(tp);
252 		break;
253 
254 	/*
255 	 * Do a send by putting data in output queue and updating urgent
256 	 * marker if URG set.  Possibly send more data.
257 	 */
258 	case PRU_SEND:
259 		sbappend(&so->so_snd, m);
260 		error = tcp_output(tp);
261 		break;
262 
263 	/*
264 	 * Abort the TCP.
265 	 */
266 	case PRU_ABORT:
267 		tp = tcp_drop(tp, ECONNABORTED);
268 		break;
269 
270 	case PRU_SENSE:
271 		((struct stat *) m)->st_blksize = so->so_snd.sb_hiwat;
272 		(void) splx(s);
273 		return (0);
274 
275 	case PRU_RCVOOB:
276 		if ((so->so_oobmark == 0 &&
277 		    (so->so_state & SS_RCVATMARK) == 0) ||
278 		    so->so_options & SO_OOBINLINE ||
279 		    tp->t_oobflags & TCPOOB_HADDATA) {
280 			error = EINVAL;
281 			break;
282 		}
283 		if ((tp->t_oobflags & TCPOOB_HAVEDATA) == 0) {
284 			error = EWOULDBLOCK;
285 			break;
286 		}
287 		m->m_len = 1;
288 		*mtod(m, caddr_t) = tp->t_iobc;
289 		if (((long)nam & MSG_PEEK) == 0)
290 			tp->t_oobflags ^= (TCPOOB_HAVEDATA | TCPOOB_HADDATA);
291 		break;
292 
293 	case PRU_SENDOOB:
294 		if (sbspace(&so->so_snd) < -512) {
295 			m_freem(m);
296 			error = ENOBUFS;
297 			break;
298 		}
299 		/*
300 		 * According to RFC961 (Assigned Protocols),
301 		 * the urgent pointer points to the last octet
302 		 * of urgent data.  We continue, however,
303 		 * to consider it to indicate the first octet
304 		 * of data past the urgent section.
305 		 * Otherwise, snd_up should be one lower.
306 		 */
307 		sbappend(&so->so_snd, m);
308 		tp->snd_up = tp->snd_una + so->so_snd.sb_cc;
309 		tp->t_force = 1;
310 		error = tcp_output(tp);
311 		tp->t_force = 0;
312 		break;
313 
314 	case PRU_SOCKADDR:
315 		in_setsockaddr(inp, nam);
316 		break;
317 
318 	case PRU_PEERADDR:
319 		in_setpeeraddr(inp, nam);
320 		break;
321 
322 	/*
323 	 * TCP slow timer went off; going through this
324 	 * routine for tracing's sake.
325 	 */
326 	case PRU_SLOWTIMO:
327 		tp = tcp_timers(tp, (long)nam);
328 		req |= (long)nam << 8;		/* for debug's sake */
329 		break;
330 
331 	default:
332 		panic("tcp_usrreq");
333 	}
334 	if (tp && (so->so_options & SO_DEBUG))
335 		tcp_trace(TA_USER, ostate, tp, (struct tcpiphdr *)0, req);
336 	splx(s);
337 	return (error);
338 }
339 
340 int
341 tcp_ctloutput(op, so, level, optname, mp)
342 	int op;
343 	struct socket *so;
344 	int level, optname;
345 	struct mbuf **mp;
346 {
347 	int error = 0, s;
348 	struct inpcb *inp;
349 	register struct tcpcb *tp;
350 	register struct mbuf *m;
351 	register int i;
352 
353 	s = splnet();
354 	inp = sotoinpcb(so);
355 	if (inp == NULL) {
356 		splx(s);
357 		if (op == PRCO_SETOPT && *mp)
358 			(void) m_free(*mp);
359 		return (ECONNRESET);
360 	}
361 	if (level != IPPROTO_TCP) {
362 		error = ip_ctloutput(op, so, level, optname, mp);
363 		splx(s);
364 		return (error);
365 	}
366 	tp = intotcpcb(inp);
367 
368 	switch (op) {
369 
370 	case PRCO_SETOPT:
371 		m = *mp;
372 		switch (optname) {
373 
374 		case TCP_NODELAY:
375 			if (m == NULL || m->m_len < sizeof (int))
376 				error = EINVAL;
377 			else if (*mtod(m, int *))
378 				tp->t_flags |= TF_NODELAY;
379 			else
380 				tp->t_flags &= ~TF_NODELAY;
381 			break;
382 
383 		case TCP_MAXSEG:
384 			if (m && (i = *mtod(m, int *)) > 0 && i <= tp->t_maxseg)
385 				tp->t_maxseg = i;
386 			else
387 				error = EINVAL;
388 			break;
389 
390 		default:
391 			error = ENOPROTOOPT;
392 			break;
393 		}
394 		if (m)
395 			(void) m_free(m);
396 		break;
397 
398 	case PRCO_GETOPT:
399 		*mp = m = m_get(M_WAIT, MT_SOOPTS);
400 		m->m_len = sizeof(int);
401 
402 		switch (optname) {
403 		case TCP_NODELAY:
404 			*mtod(m, int *) = tp->t_flags & TF_NODELAY;
405 			break;
406 		case TCP_MAXSEG:
407 			*mtod(m, int *) = tp->t_maxseg;
408 			break;
409 		default:
410 			error = ENOPROTOOPT;
411 			break;
412 		}
413 		break;
414 	}
415 	splx(s);
416 	return (error);
417 }
418 
419 #ifndef TCP_SENDSPACE
420 #define	TCP_SENDSPACE	1024*16;
421 #endif
422 u_long	tcp_sendspace = TCP_SENDSPACE;
423 #ifndef TCP_RECVSPACE
424 #define	TCP_RECVSPACE	1024*16;
425 #endif
426 u_long	tcp_recvspace = TCP_RECVSPACE;
427 
428 /*
429  * Attach TCP protocol to socket, allocating
430  * internet protocol control block, tcp control block,
431  * bufer space, and entering LISTEN state if to accept connections.
432  */
433 int
434 tcp_attach(so)
435 	struct socket *so;
436 {
437 	register struct tcpcb *tp;
438 	struct inpcb *inp;
439 	int error;
440 
441 	if (so->so_snd.sb_hiwat == 0 || so->so_rcv.sb_hiwat == 0) {
442 		error = soreserve(so, tcp_sendspace, tcp_recvspace);
443 		if (error)
444 			return (error);
445 	}
446 	error = in_pcballoc(so, &tcb);
447 	if (error)
448 		return (error);
449 	inp = sotoinpcb(so);
450 	tp = tcp_newtcpcb(inp);
451 	if (tp == 0) {
452 		int nofd = so->so_state & SS_NOFDREF;	/* XXX */
453 
454 		so->so_state &= ~SS_NOFDREF;	/* don't free the socket yet */
455 		in_pcbdetach(inp);
456 		so->so_state |= nofd;
457 		return (ENOBUFS);
458 	}
459 	tp->t_state = TCPS_CLOSED;
460 	return (0);
461 }
462 
463 /*
464  * Initiate (or continue) disconnect.
465  * If embryonic state, just send reset (once).
466  * If in ``let data drain'' option and linger null, just drop.
467  * Otherwise (hard), mark socket disconnecting and drop
468  * current input data; switch states based on user close, and
469  * send segment to peer (with FIN).
470  */
471 struct tcpcb *
472 tcp_disconnect(tp)
473 	register struct tcpcb *tp;
474 {
475 	struct socket *so = tp->t_inpcb->inp_socket;
476 
477 	if (TCPS_HAVEESTABLISHED(tp->t_state) == 0)
478 		tp = tcp_close(tp);
479 	else if ((so->so_options & SO_LINGER) && so->so_linger == 0)
480 		tp = tcp_drop(tp, 0);
481 	else {
482 		soisdisconnecting(so);
483 		sbflush(&so->so_rcv);
484 		tp = tcp_usrclosed(tp);
485 		if (tp)
486 			(void) tcp_output(tp);
487 	}
488 	return (tp);
489 }
490 
491 /*
492  * User issued close, and wish to trail through shutdown states:
493  * if never received SYN, just forget it.  If got a SYN from peer,
494  * but haven't sent FIN, then go to FIN_WAIT_1 state to send peer a FIN.
495  * If already got a FIN from peer, then almost done; go to LAST_ACK
496  * state.  In all other cases, have already sent FIN to peer (e.g.
497  * after PRU_SHUTDOWN), and just have to play tedious game waiting
498  * for peer to send FIN or not respond to keep-alives, etc.
499  * We can let the user exit from the close as soon as the FIN is acked.
500  */
501 struct tcpcb *
502 tcp_usrclosed(tp)
503 	register struct tcpcb *tp;
504 {
505 
506 	switch (tp->t_state) {
507 
508 	case TCPS_CLOSED:
509 	case TCPS_LISTEN:
510 	case TCPS_SYN_SENT:
511 		tp->t_state = TCPS_CLOSED;
512 		tp = tcp_close(tp);
513 		break;
514 
515 	case TCPS_SYN_RECEIVED:
516 	case TCPS_ESTABLISHED:
517 		tp->t_state = TCPS_FIN_WAIT_1;
518 		break;
519 
520 	case TCPS_CLOSE_WAIT:
521 		tp->t_state = TCPS_LAST_ACK;
522 		break;
523 	}
524 	if (tp && tp->t_state >= TCPS_FIN_WAIT_2)
525 		soisdisconnected(tp->t_inpcb->inp_socket);
526 	return (tp);
527 }
528