xref: /freebsd/sys/netinet/tcp_usrreq.c (revision afe61c15161c324a7af299a9b8457aba5afc92db)
1 /*
2  * Copyright (c) 1982, 1986, 1988, 1993
3  *	The Regents of the University of California.  All rights reserved.
4  *
5  * Redistribution and use in source and binary forms, with or without
6  * modification, are permitted provided that the following conditions
7  * are met:
8  * 1. Redistributions of source code must retain the above copyright
9  *    notice, this list of conditions and the following disclaimer.
10  * 2. Redistributions in binary form must reproduce the above copyright
11  *    notice, this list of conditions and the following disclaimer in the
12  *    documentation and/or other materials provided with the distribution.
13  * 3. All advertising materials mentioning features or use of this software
14  *    must display the following acknowledgement:
15  *	This product includes software developed by the University of
16  *	California, Berkeley and its contributors.
17  * 4. Neither the name of the University nor the names of its contributors
18  *    may be used to endorse or promote products derived from this software
19  *    without specific prior written permission.
20  *
21  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
22  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
23  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
24  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
25  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
26  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
27  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
28  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
29  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
30  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
31  * SUCH DAMAGE.
32  *
33  *	@(#)tcp_usrreq.c	8.2 (Berkeley) 1/3/94
34  */
35 
36 #include <sys/param.h>
37 #include <sys/systm.h>
38 #include <sys/malloc.h>
39 #include <sys/mbuf.h>
40 #include <sys/socket.h>
41 #include <sys/socketvar.h>
42 #include <sys/protosw.h>
43 #include <sys/errno.h>
44 #include <sys/stat.h>
45 
46 #include <net/if.h>
47 #include <net/route.h>
48 
49 #include <netinet/in.h>
50 #include <netinet/in_systm.h>
51 #include <netinet/ip.h>
52 #include <netinet/in_pcb.h>
53 #include <netinet/ip_var.h>
54 #include <netinet/tcp.h>
55 #include <netinet/tcp_fsm.h>
56 #include <netinet/tcp_seq.h>
57 #include <netinet/tcp_timer.h>
58 #include <netinet/tcp_var.h>
59 #include <netinet/tcpip.h>
60 #include <netinet/tcp_debug.h>
61 
62 /*
63  * TCP protocol interface to socket abstraction.
64  */
65 extern	char *tcpstates[];
66 
67 /*
68  * Process a TCP user request for TCP tb.  If this is a send request
69  * then m is the mbuf chain of send data.  If this is a timer expiration
70  * (called from the software clock routine), then timertype tells which timer.
71  */
72 /*ARGSUSED*/
73 int
74 tcp_usrreq(so, req, m, nam, control)
75 	struct socket *so;
76 	int req;
77 	struct mbuf *m, *nam, *control;
78 {
79 	register struct inpcb *inp;
80 	register struct tcpcb *tp = 0;
81 	int s;
82 	int error = 0;
83 	int ostate;
84 
85 	if (req == PRU_CONTROL)
86 		return (in_control(so, (int)m, (caddr_t)nam,
87 			(struct ifnet *)control));
88 	if (control && control->m_len) {
89 		m_freem(control);
90 		if (m)
91 			m_freem(m);
92 		return (EINVAL);
93 	}
94 
95 	s = splnet();
96 	inp = sotoinpcb(so);
97 	/*
98 	 * When a TCP is attached to a socket, then there will be
99 	 * a (struct inpcb) pointed at by the socket, and this
100 	 * structure will point at a subsidary (struct tcpcb).
101 	 */
102 	if (inp == 0 && req != PRU_ATTACH) {
103 		splx(s);
104 		return (EINVAL);		/* XXX */
105 	}
106 	if (inp) {
107 		tp = intotcpcb(inp);
108 		/* WHAT IF TP IS 0? */
109 #ifdef KPROF
110 		tcp_acounts[tp->t_state][req]++;
111 #endif
112 		ostate = tp->t_state;
113 	} else
114 		ostate = 0;
115 	switch (req) {
116 
117 	/*
118 	 * TCP attaches to socket via PRU_ATTACH, reserving space,
119 	 * and an internet control block.
120 	 */
121 	case PRU_ATTACH:
122 		if (inp) {
123 			error = EISCONN;
124 			break;
125 		}
126 		error = tcp_attach(so);
127 		if (error)
128 			break;
129 		if ((so->so_options & SO_LINGER) && so->so_linger == 0)
130 			so->so_linger = TCP_LINGERTIME;
131 		tp = sototcpcb(so);
132 		break;
133 
134 	/*
135 	 * PRU_DETACH detaches the TCP protocol from the socket.
136 	 * If the protocol state is non-embryonic, then can't
137 	 * do this directly: have to initiate a PRU_DISCONNECT,
138 	 * which may finish later; embryonic TCB's can just
139 	 * be discarded here.
140 	 */
141 	case PRU_DETACH:
142 		if (tp->t_state > TCPS_LISTEN)
143 			tp = tcp_disconnect(tp);
144 		else
145 			tp = tcp_close(tp);
146 		break;
147 
148 	/*
149 	 * Give the socket an address.
150 	 */
151 	case PRU_BIND:
152 		error = in_pcbbind(inp, nam);
153 		if (error)
154 			break;
155 		break;
156 
157 	/*
158 	 * Prepare to accept connections.
159 	 */
160 	case PRU_LISTEN:
161 		if (inp->inp_lport == 0)
162 			error = in_pcbbind(inp, (struct mbuf *)0);
163 		if (error == 0)
164 			tp->t_state = TCPS_LISTEN;
165 		break;
166 
167 	/*
168 	 * Initiate connection to peer.
169 	 * Create a template for use in transmissions on this connection.
170 	 * Enter SYN_SENT state, and mark socket as connecting.
171 	 * Start keep-alive timer, and seed output sequence space.
172 	 * Send initial segment on connection.
173 	 */
174 	case PRU_CONNECT:
175 		if (inp->inp_lport == 0) {
176 			error = in_pcbbind(inp, (struct mbuf *)0);
177 			if (error)
178 				break;
179 		}
180 		error = in_pcbconnect(inp, nam);
181 		if (error)
182 			break;
183 		tp->t_template = tcp_template(tp);
184 		if (tp->t_template == 0) {
185 			in_pcbdisconnect(inp);
186 			error = ENOBUFS;
187 			break;
188 		}
189 		/* Compute window scaling to request.  */
190 		while (tp->request_r_scale < TCP_MAX_WINSHIFT &&
191 		    (TCP_MAXWIN << tp->request_r_scale) < so->so_rcv.sb_hiwat)
192 			tp->request_r_scale++;
193 		soisconnecting(so);
194 		tcpstat.tcps_connattempt++;
195 		tp->t_state = TCPS_SYN_SENT;
196 		tp->t_timer[TCPT_KEEP] = TCPTV_KEEP_INIT;
197 		tp->iss = tcp_iss; tcp_iss += TCP_ISSINCR/2;
198 		tcp_sendseqinit(tp);
199 		error = tcp_output(tp);
200 		break;
201 
202 	/*
203 	 * Create a TCP connection between two sockets.
204 	 */
205 	case PRU_CONNECT2:
206 		error = EOPNOTSUPP;
207 		break;
208 
209 	/*
210 	 * Initiate disconnect from peer.
211 	 * If connection never passed embryonic stage, just drop;
212 	 * else if don't need to let data drain, then can just drop anyways,
213 	 * else have to begin TCP shutdown process: mark socket disconnecting,
214 	 * drain unread data, state switch to reflect user close, and
215 	 * send segment (e.g. FIN) to peer.  Socket will be really disconnected
216 	 * when peer sends FIN and acks ours.
217 	 *
218 	 * SHOULD IMPLEMENT LATER PRU_CONNECT VIA REALLOC TCPCB.
219 	 */
220 	case PRU_DISCONNECT:
221 		tp = tcp_disconnect(tp);
222 		break;
223 
224 	/*
225 	 * Accept a connection.  Essentially all the work is
226 	 * done at higher levels; just return the address
227 	 * of the peer, storing through addr.
228 	 */
229 	case PRU_ACCEPT:
230 		in_setpeeraddr(inp, nam);
231 		break;
232 
233 	/*
234 	 * Mark the connection as being incapable of further output.
235 	 */
236 	case PRU_SHUTDOWN:
237 		socantsendmore(so);
238 		tp = tcp_usrclosed(tp);
239 		if (tp)
240 			error = tcp_output(tp);
241 		break;
242 
243 	/*
244 	 * After a receive, possibly send window update to peer.
245 	 */
246 	case PRU_RCVD:
247 		(void) tcp_output(tp);
248 		break;
249 
250 	/*
251 	 * Do a send by putting data in output queue and updating urgent
252 	 * marker if URG set.  Possibly send more data.
253 	 */
254 	case PRU_SEND:
255 		sbappend(&so->so_snd, m);
256 		error = tcp_output(tp);
257 		break;
258 
259 	/*
260 	 * Abort the TCP.
261 	 */
262 	case PRU_ABORT:
263 		tp = tcp_drop(tp, ECONNABORTED);
264 		break;
265 
266 	case PRU_SENSE:
267 		((struct stat *) m)->st_blksize = so->so_snd.sb_hiwat;
268 		(void) splx(s);
269 		return (0);
270 
271 	case PRU_RCVOOB:
272 		if ((so->so_oobmark == 0 &&
273 		    (so->so_state & SS_RCVATMARK) == 0) ||
274 		    so->so_options & SO_OOBINLINE ||
275 		    tp->t_oobflags & TCPOOB_HADDATA) {
276 			error = EINVAL;
277 			break;
278 		}
279 		if ((tp->t_oobflags & TCPOOB_HAVEDATA) == 0) {
280 			error = EWOULDBLOCK;
281 			break;
282 		}
283 		m->m_len = 1;
284 		*mtod(m, caddr_t) = tp->t_iobc;
285 		if (((int)nam & MSG_PEEK) == 0)
286 			tp->t_oobflags ^= (TCPOOB_HAVEDATA | TCPOOB_HADDATA);
287 		break;
288 
289 	case PRU_SENDOOB:
290 		if (sbspace(&so->so_snd) < -512) {
291 			m_freem(m);
292 			error = ENOBUFS;
293 			break;
294 		}
295 		/*
296 		 * According to RFC961 (Assigned Protocols),
297 		 * the urgent pointer points to the last octet
298 		 * of urgent data.  We continue, however,
299 		 * to consider it to indicate the first octet
300 		 * of data past the urgent section.
301 		 * Otherwise, snd_up should be one lower.
302 		 */
303 		sbappend(&so->so_snd, m);
304 		tp->snd_up = tp->snd_una + so->so_snd.sb_cc;
305 		tp->t_force = 1;
306 		error = tcp_output(tp);
307 		tp->t_force = 0;
308 		break;
309 
310 	case PRU_SOCKADDR:
311 		in_setsockaddr(inp, nam);
312 		break;
313 
314 	case PRU_PEERADDR:
315 		in_setpeeraddr(inp, nam);
316 		break;
317 
318 	/*
319 	 * TCP slow timer went off; going through this
320 	 * routine for tracing's sake.
321 	 */
322 	case PRU_SLOWTIMO:
323 		tp = tcp_timers(tp, (int)nam);
324 		req |= (int)nam << 8;		/* for debug's sake */
325 		break;
326 
327 	default:
328 		panic("tcp_usrreq");
329 	}
330 	if (tp && (so->so_options & SO_DEBUG))
331 		tcp_trace(TA_USER, ostate, tp, (struct tcpiphdr *)0, req);
332 	splx(s);
333 	return (error);
334 }
335 
336 int
337 tcp_ctloutput(op, so, level, optname, mp)
338 	int op;
339 	struct socket *so;
340 	int level, optname;
341 	struct mbuf **mp;
342 {
343 	int error = 0, s;
344 	struct inpcb *inp;
345 	register struct tcpcb *tp;
346 	register struct mbuf *m;
347 	register int i;
348 
349 	s = splnet();
350 	inp = sotoinpcb(so);
351 	if (inp == NULL) {
352 		splx(s);
353 		if (op == PRCO_SETOPT && *mp)
354 			(void) m_free(*mp);
355 		return (ECONNRESET);
356 	}
357 	if (level != IPPROTO_TCP) {
358 		error = ip_ctloutput(op, so, level, optname, mp);
359 		splx(s);
360 		return (error);
361 	}
362 	tp = intotcpcb(inp);
363 
364 	switch (op) {
365 
366 	case PRCO_SETOPT:
367 		m = *mp;
368 		switch (optname) {
369 
370 		case TCP_NODELAY:
371 			if (m == NULL || m->m_len < sizeof (int))
372 				error = EINVAL;
373 			else if (*mtod(m, int *))
374 				tp->t_flags |= TF_NODELAY;
375 			else
376 				tp->t_flags &= ~TF_NODELAY;
377 			break;
378 
379 		case TCP_MAXSEG:
380 			if (m && (i = *mtod(m, int *)) > 0 && i <= tp->t_maxseg)
381 				tp->t_maxseg = i;
382 			else
383 				error = EINVAL;
384 			break;
385 
386 		default:
387 			error = ENOPROTOOPT;
388 			break;
389 		}
390 		if (m)
391 			(void) m_free(m);
392 		break;
393 
394 	case PRCO_GETOPT:
395 		*mp = m = m_get(M_WAIT, MT_SOOPTS);
396 		m->m_len = sizeof(int);
397 
398 		switch (optname) {
399 		case TCP_NODELAY:
400 			*mtod(m, int *) = tp->t_flags & TF_NODELAY;
401 			break;
402 		case TCP_MAXSEG:
403 			*mtod(m, int *) = tp->t_maxseg;
404 			break;
405 		default:
406 			error = ENOPROTOOPT;
407 			break;
408 		}
409 		break;
410 	}
411 	splx(s);
412 	return (error);
413 }
414 
415 /*
416  * tcp_sendspace and tcp_recvspace are the default send and receive window
417  * sizes, respectively.  These are obsolescent (this information should
418  * be set by the route).
419  */
420 #ifdef TCP_SMALLSPACE
421 u_long	tcp_sendspace = 1024*4;
422 u_long	tcp_recvspace = 1024*4;
423 #else
424 u_long	tcp_sendspace = 1024*16;
425 u_long	tcp_recvspace = 1024*16;
426 #endif
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 (tp->t_state < TCPS_ESTABLISHED)
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