xref: /netbsd-src/sys/netinet/tcp_usrreq.c (revision 0b9f50897e9a9c6709320fafb4c3787fddcc0a45)
1 /*
2  * Copyright (c) 1982, 1986, 1988 Regents of the University of California.
3  * 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  *	from: @(#)tcp_usrreq.c	7.15 (Berkeley) 6/28/90
34  *	$Id: tcp_usrreq.c,v 1.4 1993/06/15 02:25:26 cgd Exp $
35  */
36 
37 #include "param.h"
38 #include "systm.h"
39 #include "malloc.h"
40 #include "mbuf.h"
41 #include "socket.h"
42 #include "socketvar.h"
43 #include "protosw.h"
44 #include "errno.h"
45 #include "stat.h"
46 
47 #include "../net/if.h"
48 #include "../net/route.h"
49 
50 #include "in.h"
51 #include "in_systm.h"
52 #include "ip.h"
53 #include "in_pcb.h"
54 #include "ip_var.h"
55 #include "tcp.h"
56 #include "tcp_fsm.h"
57 #include "tcp_seq.h"
58 #include "tcp_timer.h"
59 #include "tcp_var.h"
60 #include "tcpip.h"
61 #include "tcp_debug.h"
62 
63 /*
64  * TCP protocol interface to socket abstraction.
65  */
66 extern	char *tcpstates[];
67 struct	tcpcb *tcp_newtcpcb();
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 tcp_usrreq(so, req, m, nam, control)
76 	struct socket *so;
77 	int req;
78 	struct mbuf *m, *nam, *control;
79 {
80 	register struct inpcb *inp;
81 	register struct tcpcb *tp;
82 	int s;
83 	int error = 0;
84 	int ostate;
85 
86 	if (req == PRU_CONTROL)
87 		return (in_control(so, (int)m, (caddr_t)nam,
88 			(struct ifnet *)control));
89 	if (control && control->m_len) {
90 		m_freem(control);
91 		if (m)
92 			m_freem(m);
93 		return (EINVAL);
94 	}
95 
96 	s = splnet();
97 	inp = sotoinpcb(so);
98 	/*
99 	 * When a TCP is attached to a socket, then there will be
100 	 * a (struct inpcb) pointed at by the socket, and this
101 	 * structure will point at a subsidary (struct tcpcb).
102 	 */
103 	if (inp == 0 && req != PRU_ATTACH) {
104 		splx(s);
105 		return (EINVAL);		/* XXX */
106 	}
107 	if (inp) {
108 		tp = intotcpcb(inp);
109 		/* WHAT IF TP IS 0? */
110 #ifdef KPROF
111 		tcp_acounts[tp->t_state][req]++;
112 #endif
113 		ostate = tp->t_state;
114 	} else
115 		ostate = 0;
116 	switch (req) {
117 
118 	/*
119 	 * TCP attaches to socket via PRU_ATTACH, reserving space,
120 	 * and an internet control block.
121 	 */
122 	case PRU_ATTACH:
123 		if (inp) {
124 			error = EISCONN;
125 			break;
126 		}
127 		error = tcp_attach(so);
128 		if (error)
129 			break;
130 		if ((so->so_options & SO_LINGER) && so->so_linger == 0)
131 			so->so_linger = TCP_LINGERTIME;
132 		tp = sototcpcb(so);
133 		break;
134 
135 	/*
136 	 * PRU_DETACH detaches the TCP protocol from the socket.
137 	 * If the protocol state is non-embryonic, then can't
138 	 * do this directly: have to initiate a PRU_DISCONNECT,
139 	 * which may finish later; embryonic TCB's can just
140 	 * be discarded here.
141 	 */
142 	case PRU_DETACH:
143 		if (tp->t_state > TCPS_LISTEN)
144 			tp = tcp_disconnect(tp);
145 		else
146 			tp = tcp_close(tp);
147 		break;
148 
149 	/*
150 	 * Give the socket an address.
151 	 */
152 	case PRU_BIND:
153 		error = in_pcbbind(inp, nam);
154 		if (error)
155 			break;
156 		break;
157 
158 	/*
159 	 * Prepare to accept connections.
160 	 */
161 	case PRU_LISTEN:
162 		if (inp->inp_lport == 0)
163 			error = in_pcbbind(inp, (struct mbuf *)0);
164 		if (error == 0)
165 			tp->t_state = TCPS_LISTEN;
166 		break;
167 
168 	/*
169 	 * Initiate connection to peer.
170 	 * Create a template for use in transmissions on this connection.
171 	 * Enter SYN_SENT state, and mark socket as connecting.
172 	 * Start keep-alive timer, and seed output sequence space.
173 	 * Send initial segment on connection.
174 	 */
175 	case PRU_CONNECT:
176 		if (inp->inp_lport == 0) {
177 			error = in_pcbbind(inp, (struct mbuf *)0);
178 			if (error)
179 				break;
180 		}
181 		error = in_pcbconnect(inp, nam);
182 		if (error)
183 			break;
184 		tp->t_template = tcp_template(tp);
185 		if (tp->t_template == 0) {
186 			in_pcbdisconnect(inp);
187 			error = ENOBUFS;
188 			break;
189 		}
190 		soisconnecting(so);
191 		tcpstat.tcps_connattempt++;
192 		tp->t_state = TCPS_SYN_SENT;
193 		tp->t_timer[TCPT_KEEP] = TCPTV_KEEP_INIT;
194 		tp->iss = tcp_iss; tcp_iss += TCP_ISSINCR/2;
195 		tcp_sendseqinit(tp);
196 		error = tcp_output(tp);
197 		break;
198 
199 	/*
200 	 * Create a TCP connection between two sockets.
201 	 */
202 	case PRU_CONNECT2:
203 		error = EOPNOTSUPP;
204 		break;
205 
206 	/*
207 	 * Initiate disconnect from peer.
208 	 * If connection never passed embryonic stage, just drop;
209 	 * else if don't need to let data drain, then can just drop anyways,
210 	 * else have to begin TCP shutdown process: mark socket disconnecting,
211 	 * drain unread data, state switch to reflect user close, and
212 	 * send segment (e.g. FIN) to peer.  Socket will be really disconnected
213 	 * when peer sends FIN and acks ours.
214 	 *
215 	 * SHOULD IMPLEMENT LATER PRU_CONNECT VIA REALLOC TCPCB.
216 	 */
217 	case PRU_DISCONNECT:
218 		tp = tcp_disconnect(tp);
219 		break;
220 
221 	/*
222 	 * Accept a connection.  Essentially all the work is
223 	 * done at higher levels; just return the address
224 	 * of the peer, storing through addr.
225 	 */
226 	case PRU_ACCEPT: {
227 		struct sockaddr_in *sin = mtod(nam, struct sockaddr_in *);
228 
229 		nam->m_len = sizeof (struct sockaddr_in);
230 		sin->sin_family = AF_INET;
231 		sin->sin_len = sizeof(*sin);
232 		sin->sin_port = inp->inp_fport;
233 		sin->sin_addr = inp->inp_faddr;
234 		break;
235 		}
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 (((int)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, (int)nam);
328 		req |= (int)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 tcp_ctloutput(op, so, level, optname, mp)
341 	int op;
342 	struct socket *so;
343 	int level, optname;
344 	struct mbuf **mp;
345 {
346 	int error = 0;
347 	struct inpcb *inp = sotoinpcb(so);
348 	register struct tcpcb *tp = intotcpcb(inp);
349 	register struct mbuf *m;
350 
351 	if (level != IPPROTO_TCP)
352 		return (ip_ctloutput(op, so, level, optname, mp));
353 
354 	switch (op) {
355 
356 	case PRCO_SETOPT:
357 		m = *mp;
358 		switch (optname) {
359 
360 		case TCP_NODELAY:
361 			if (m == NULL || m->m_len < sizeof (int))
362 				error = EINVAL;
363 			else if (*mtod(m, int *))
364 				tp->t_flags |= TF_NODELAY;
365 			else
366 				tp->t_flags &= ~TF_NODELAY;
367 			break;
368 
369 		case TCP_MAXSEG:	/* not yet */
370 		default:
371 			error = EINVAL;
372 			break;
373 		}
374 		if (m)
375 			(void) m_free(m);
376 		break;
377 
378 	case PRCO_GETOPT:
379 		*mp = m = m_get(M_WAIT, MT_SOOPTS);
380 		m->m_len = sizeof(int);
381 
382 		switch (optname) {
383 		case TCP_NODELAY:
384 			*mtod(m, int *) = tp->t_flags & TF_NODELAY;
385 			break;
386 		case TCP_MAXSEG:
387 			*mtod(m, int *) = tp->t_maxseg;
388 			break;
389 		default:
390 			error = EINVAL;
391 			break;
392 		}
393 		break;
394 	}
395 	return (error);
396 }
397 
398 u_long	tcp_sendspace = 1024*8;
399 u_long	tcp_recvspace = 1024*8;
400 
401 /*
402  * Attach TCP protocol to socket, allocating
403  * internet protocol control block, tcp control block,
404  * bufer space, and entering LISTEN state if to accept connections.
405  */
406 tcp_attach(so)
407 	struct socket *so;
408 {
409 	register struct tcpcb *tp;
410 	struct inpcb *inp;
411 	int error;
412 
413 	if (so->so_snd.sb_hiwat == 0 || so->so_rcv.sb_hiwat == 0) {
414 		error = soreserve(so, tcp_sendspace, tcp_recvspace);
415 		if (error)
416 			return (error);
417 	}
418 	error = in_pcballoc(so, &tcb);
419 	if (error)
420 		return (error);
421 	inp = sotoinpcb(so);
422 	tp = tcp_newtcpcb(inp);
423 	if (tp == 0) {
424 		int nofd = so->so_state & SS_NOFDREF;	/* XXX */
425 
426 		so->so_state &= ~SS_NOFDREF;	/* don't free the socket yet */
427 		in_pcbdetach(inp);
428 		so->so_state |= nofd;
429 		return (ENOBUFS);
430 	}
431 	tp->t_state = TCPS_CLOSED;
432 	return (0);
433 }
434 
435 /*
436  * Initiate (or continue) disconnect.
437  * If embryonic state, just send reset (once).
438  * If in ``let data drain'' option and linger null, just drop.
439  * Otherwise (hard), mark socket disconnecting and drop
440  * current input data; switch states based on user close, and
441  * send segment to peer (with FIN).
442  */
443 struct tcpcb *
444 tcp_disconnect(tp)
445 	register struct tcpcb *tp;
446 {
447 	struct socket *so = tp->t_inpcb->inp_socket;
448 
449 	if (tp->t_state < TCPS_ESTABLISHED)
450 		tp = tcp_close(tp);
451 	else if ((so->so_options & SO_LINGER) && so->so_linger == 0)
452 		tp = tcp_drop(tp, 0);
453 	else {
454 		soisdisconnecting(so);
455 		sbflush(&so->so_rcv);
456 		tp = tcp_usrclosed(tp);
457 		if (tp)
458 			(void) tcp_output(tp);
459 	}
460 	return (tp);
461 }
462 
463 /*
464  * User issued close, and wish to trail through shutdown states:
465  * if never received SYN, just forget it.  If got a SYN from peer,
466  * but haven't sent FIN, then go to FIN_WAIT_1 state to send peer a FIN.
467  * If already got a FIN from peer, then almost done; go to LAST_ACK
468  * state.  In all other cases, have already sent FIN to peer (e.g.
469  * after PRU_SHUTDOWN), and just have to play tedious game waiting
470  * for peer to send FIN or not respond to keep-alives, etc.
471  * We can let the user exit from the close as soon as the FIN is acked.
472  */
473 struct tcpcb *
474 tcp_usrclosed(tp)
475 	register struct tcpcb *tp;
476 {
477 
478 	switch (tp->t_state) {
479 
480 	case TCPS_CLOSED:
481 	case TCPS_LISTEN:
482 	case TCPS_SYN_SENT:
483 		tp->t_state = TCPS_CLOSED;
484 		tp = tcp_close(tp);
485 		break;
486 
487 	case TCPS_SYN_RECEIVED:
488 	case TCPS_ESTABLISHED:
489 		tp->t_state = TCPS_FIN_WAIT_1;
490 		break;
491 
492 	case TCPS_CLOSE_WAIT:
493 		tp->t_state = TCPS_LAST_ACK;
494 		break;
495 	}
496 	if (tp && tp->t_state >= TCPS_FIN_WAIT_2)
497 		soisdisconnected(tp->t_inpcb->inp_socket);
498 	return (tp);
499 }
500