xref: /netbsd-src/sys/netinet/tcp_usrreq.c (revision 05a9db8e4f5a54955cacec0e01c45d54b958765c)
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 1.8 1994/01/10 23:27:45 mycroft Exp $
35  */
36 
37 #include <sys/param.h>
38 #include <sys/systm.h>
39 #include <sys/malloc.h>
40 #include <sys/mbuf.h>
41 #include <sys/socket.h>
42 #include <sys/socketvar.h>
43 #include <sys/protosw.h>
44 #include <sys/errno.h>
45 #include <sys/stat.h>
46 
47 #include <net/if.h>
48 #include <net/route.h>
49 
50 #include <netinet/in.h>
51 #include <netinet/in_systm.h>
52 #include <netinet/ip.h>
53 #include <netinet/in_pcb.h>
54 #include <netinet/ip_var.h>
55 #include <netinet/tcp.h>
56 #include <netinet/tcp_fsm.h>
57 #include <netinet/tcp_seq.h>
58 #include <netinet/tcp_timer.h>
59 #include <netinet/tcp_var.h>
60 #include <netinet/tcpip.h>
61 #include <netinet/tcp_debug.h>
62 
63 /*
64  * TCP protocol interface to socket abstraction.
65  */
66 extern	char *tcpstates[];
67 
68 /*
69  * Process a TCP user request for TCP tb.  If this is a send request
70  * then m is the mbuf chain of send data.  If this is a timer expiration
71  * (called from the software clock routine), then timertype tells which timer.
72  */
73 /*ARGSUSED*/
74 int
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 		in_setpeeraddr(inp, nam);
228 		break;
229 
230 	/*
231 	 * Mark the connection as being incapable of further output.
232 	 */
233 	case PRU_SHUTDOWN:
234 		socantsendmore(so);
235 		tp = tcp_usrclosed(tp);
236 		if (tp)
237 			error = tcp_output(tp);
238 		break;
239 
240 	/*
241 	 * After a receive, possibly send window update to peer.
242 	 */
243 	case PRU_RCVD:
244 		(void) tcp_output(tp);
245 		break;
246 
247 	/*
248 	 * Do a send by putting data in output queue and updating urgent
249 	 * marker if URG set.  Possibly send more data.
250 	 */
251 	case PRU_SEND:
252 		sbappend(&so->so_snd, m);
253 		error = tcp_output(tp);
254 		break;
255 
256 	/*
257 	 * Abort the TCP.
258 	 */
259 	case PRU_ABORT:
260 		tp = tcp_drop(tp, ECONNABORTED);
261 		break;
262 
263 	case PRU_SENSE:
264 		((struct stat *) m)->st_blksize = so->so_snd.sb_hiwat;
265 		(void) splx(s);
266 		return (0);
267 
268 	case PRU_RCVOOB:
269 		if ((so->so_oobmark == 0 &&
270 		    (so->so_state & SS_RCVATMARK) == 0) ||
271 		    so->so_options & SO_OOBINLINE ||
272 		    tp->t_oobflags & TCPOOB_HADDATA) {
273 			error = EINVAL;
274 			break;
275 		}
276 		if ((tp->t_oobflags & TCPOOB_HAVEDATA) == 0) {
277 			error = EWOULDBLOCK;
278 			break;
279 		}
280 		m->m_len = 1;
281 		*mtod(m, caddr_t) = tp->t_iobc;
282 		if (((int)nam & MSG_PEEK) == 0)
283 			tp->t_oobflags ^= (TCPOOB_HAVEDATA | TCPOOB_HADDATA);
284 		break;
285 
286 	case PRU_SENDOOB:
287 		if (sbspace(&so->so_snd) < -512) {
288 			m_freem(m);
289 			error = ENOBUFS;
290 			break;
291 		}
292 		/*
293 		 * According to RFC961 (Assigned Protocols),
294 		 * the urgent pointer points to the last octet
295 		 * of urgent data.  We continue, however,
296 		 * to consider it to indicate the first octet
297 		 * of data past the urgent section.
298 		 * Otherwise, snd_up should be one lower.
299 		 */
300 		sbappend(&so->so_snd, m);
301 		tp->snd_up = tp->snd_una + so->so_snd.sb_cc;
302 		tp->t_force = 1;
303 		error = tcp_output(tp);
304 		tp->t_force = 0;
305 		break;
306 
307 	case PRU_SOCKADDR:
308 		in_setsockaddr(inp, nam);
309 		break;
310 
311 	case PRU_PEERADDR:
312 		in_setpeeraddr(inp, nam);
313 		break;
314 
315 	/*
316 	 * TCP slow timer went off; going through this
317 	 * routine for tracing's sake.
318 	 */
319 	case PRU_SLOWTIMO:
320 		tp = tcp_timers(tp, (int)nam);
321 		req |= (int)nam << 8;		/* for debug's sake */
322 		break;
323 
324 	default:
325 		panic("tcp_usrreq");
326 	}
327 	if (tp && (so->so_options & SO_DEBUG))
328 		tcp_trace(TA_USER, ostate, tp, (struct tcpiphdr *)0, req);
329 	splx(s);
330 	return (error);
331 }
332 
333 int
334 tcp_ctloutput(op, so, level, optname, mp)
335 	int op;
336 	struct socket *so;
337 	int level, optname;
338 	struct mbuf **mp;
339 {
340 	int error = 0;
341 	struct inpcb *inp = sotoinpcb(so);
342 	register struct tcpcb *tp = intotcpcb(inp);
343 	register struct mbuf *m;
344 
345 	if (level != IPPROTO_TCP)
346 		return (ip_ctloutput(op, so, level, optname, mp));
347 
348 	switch (op) {
349 
350 	case PRCO_SETOPT:
351 		m = *mp;
352 		switch (optname) {
353 
354 		case TCP_NODELAY:
355 			if (m == NULL || m->m_len < sizeof (int))
356 				error = EINVAL;
357 			else if (*mtod(m, int *))
358 				tp->t_flags |= TF_NODELAY;
359 			else
360 				tp->t_flags &= ~TF_NODELAY;
361 			break;
362 
363 		case TCP_MAXSEG:	/* not yet */
364 		default:
365 			error = EINVAL;
366 			break;
367 		}
368 		if (m)
369 			(void) m_free(m);
370 		break;
371 
372 	case PRCO_GETOPT:
373 		*mp = m = m_get(M_WAIT, MT_SOOPTS);
374 		m->m_len = sizeof(int);
375 
376 		switch (optname) {
377 		case TCP_NODELAY:
378 			*mtod(m, int *) = tp->t_flags & TF_NODELAY;
379 			break;
380 		case TCP_MAXSEG:
381 			*mtod(m, int *) = tp->t_maxseg;
382 			break;
383 		default:
384 			error = EINVAL;
385 			break;
386 		}
387 		break;
388 	}
389 	return (error);
390 }
391 
392 u_long	tcp_sendspace = 1024*8;
393 u_long	tcp_recvspace = 1024*8;
394 
395 /*
396  * Attach TCP protocol to socket, allocating
397  * internet protocol control block, tcp control block,
398  * bufer space, and entering LISTEN state if to accept connections.
399  */
400 int
401 tcp_attach(so)
402 	struct socket *so;
403 {
404 	register struct tcpcb *tp;
405 	struct inpcb *inp;
406 	int error;
407 
408 	if (so->so_snd.sb_hiwat == 0 || so->so_rcv.sb_hiwat == 0) {
409 		error = soreserve(so, tcp_sendspace, tcp_recvspace);
410 		if (error)
411 			return (error);
412 	}
413 	error = in_pcballoc(so, &tcb);
414 	if (error)
415 		return (error);
416 	inp = sotoinpcb(so);
417 	tp = tcp_newtcpcb(inp);
418 	if (tp == 0) {
419 		int nofd = so->so_state & SS_NOFDREF;	/* XXX */
420 
421 		so->so_state &= ~SS_NOFDREF;	/* don't free the socket yet */
422 		in_pcbdetach(inp);
423 		so->so_state |= nofd;
424 		return (ENOBUFS);
425 	}
426 	tp->t_state = TCPS_CLOSED;
427 	return (0);
428 }
429 
430 /*
431  * Initiate (or continue) disconnect.
432  * If embryonic state, just send reset (once).
433  * If in ``let data drain'' option and linger null, just drop.
434  * Otherwise (hard), mark socket disconnecting and drop
435  * current input data; switch states based on user close, and
436  * send segment to peer (with FIN).
437  */
438 struct tcpcb *
439 tcp_disconnect(tp)
440 	register struct tcpcb *tp;
441 {
442 	struct socket *so = tp->t_inpcb->inp_socket;
443 
444 	if (tp->t_state < TCPS_ESTABLISHED)
445 		tp = tcp_close(tp);
446 	else if ((so->so_options & SO_LINGER) && so->so_linger == 0)
447 		tp = tcp_drop(tp, 0);
448 	else {
449 		soisdisconnecting(so);
450 		sbflush(&so->so_rcv);
451 		tp = tcp_usrclosed(tp);
452 		if (tp)
453 			(void) tcp_output(tp);
454 	}
455 	return (tp);
456 }
457 
458 /*
459  * User issued close, and wish to trail through shutdown states:
460  * if never received SYN, just forget it.  If got a SYN from peer,
461  * but haven't sent FIN, then go to FIN_WAIT_1 state to send peer a FIN.
462  * If already got a FIN from peer, then almost done; go to LAST_ACK
463  * state.  In all other cases, have already sent FIN to peer (e.g.
464  * after PRU_SHUTDOWN), and just have to play tedious game waiting
465  * for peer to send FIN or not respond to keep-alives, etc.
466  * We can let the user exit from the close as soon as the FIN is acked.
467  */
468 struct tcpcb *
469 tcp_usrclosed(tp)
470 	register struct tcpcb *tp;
471 {
472 
473 	switch (tp->t_state) {
474 
475 	case TCPS_CLOSED:
476 	case TCPS_LISTEN:
477 	case TCPS_SYN_SENT:
478 		tp->t_state = TCPS_CLOSED;
479 		tp = tcp_close(tp);
480 		break;
481 
482 	case TCPS_SYN_RECEIVED:
483 	case TCPS_ESTABLISHED:
484 		tp->t_state = TCPS_FIN_WAIT_1;
485 		break;
486 
487 	case TCPS_CLOSE_WAIT:
488 		tp->t_state = TCPS_LAST_ACK;
489 		break;
490 	}
491 	if (tp && tp->t_state >= TCPS_FIN_WAIT_2)
492 		soisdisconnected(tp->t_inpcb->inp_socket);
493 	return (tp);
494 }
495