xref: /netbsd-src/sys/netinet/tcp_usrreq.c (revision 61f282557f0bc41c0b762c629a2f4c14be8b7591)
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  *	@(#)tcp_usrreq.c	7.15 (Berkeley) 6/28/90
34  */
35 
36 #include "param.h"
37 #include "systm.h"
38 #include "malloc.h"
39 #include "mbuf.h"
40 #include "socket.h"
41 #include "socketvar.h"
42 #include "protosw.h"
43 #include "errno.h"
44 #include "stat.h"
45 
46 #include "../net/if.h"
47 #include "../net/route.h"
48 
49 #include "in.h"
50 #include "in_systm.h"
51 #include "ip.h"
52 #include "in_pcb.h"
53 #include "ip_var.h"
54 #include "tcp.h"
55 #include "tcp_fsm.h"
56 #include "tcp_seq.h"
57 #include "tcp_timer.h"
58 #include "tcp_var.h"
59 #include "tcpip.h"
60 #include "tcp_debug.h"
61 
62 /*
63  * TCP protocol interface to socket abstraction.
64  */
65 extern	char *tcpstates[];
66 struct	tcpcb *tcp_newtcpcb();
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 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;
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 		soisconnecting(so);
190 		tcpstat.tcps_connattempt++;
191 		tp->t_state = TCPS_SYN_SENT;
192 		tp->t_timer[TCPT_KEEP] = TCPTV_KEEP_INIT;
193 		tp->iss = tcp_iss; tcp_iss += TCP_ISSINCR/2;
194 		tcp_sendseqinit(tp);
195 		error = tcp_output(tp);
196 		break;
197 
198 	/*
199 	 * Create a TCP connection between two sockets.
200 	 */
201 	case PRU_CONNECT2:
202 		error = EOPNOTSUPP;
203 		break;
204 
205 	/*
206 	 * Initiate disconnect from peer.
207 	 * If connection never passed embryonic stage, just drop;
208 	 * else if don't need to let data drain, then can just drop anyways,
209 	 * else have to begin TCP shutdown process: mark socket disconnecting,
210 	 * drain unread data, state switch to reflect user close, and
211 	 * send segment (e.g. FIN) to peer.  Socket will be really disconnected
212 	 * when peer sends FIN and acks ours.
213 	 *
214 	 * SHOULD IMPLEMENT LATER PRU_CONNECT VIA REALLOC TCPCB.
215 	 */
216 	case PRU_DISCONNECT:
217 		tp = tcp_disconnect(tp);
218 		break;
219 
220 	/*
221 	 * Accept a connection.  Essentially all the work is
222 	 * done at higher levels; just return the address
223 	 * of the peer, storing through addr.
224 	 */
225 	case PRU_ACCEPT: {
226 		struct sockaddr_in *sin = mtod(nam, struct sockaddr_in *);
227 
228 		nam->m_len = sizeof (struct sockaddr_in);
229 		sin->sin_family = AF_INET;
230 		sin->sin_len = sizeof(*sin);
231 		sin->sin_port = inp->inp_fport;
232 		sin->sin_addr = inp->inp_faddr;
233 		break;
234 		}
235 
236 	/*
237 	 * Mark the connection as being incapable of further output.
238 	 */
239 	case PRU_SHUTDOWN:
240 		socantsendmore(so);
241 		tp = tcp_usrclosed(tp);
242 		if (tp)
243 			error = tcp_output(tp);
244 		break;
245 
246 	/*
247 	 * After a receive, possibly send window update to peer.
248 	 */
249 	case PRU_RCVD:
250 		(void) tcp_output(tp);
251 		break;
252 
253 	/*
254 	 * Do a send by putting data in output queue and updating urgent
255 	 * marker if URG set.  Possibly send more data.
256 	 */
257 	case PRU_SEND:
258 		sbappend(&so->so_snd, m);
259 		error = tcp_output(tp);
260 		break;
261 
262 	/*
263 	 * Abort the TCP.
264 	 */
265 	case PRU_ABORT:
266 		tp = tcp_drop(tp, ECONNABORTED);
267 		break;
268 
269 	case PRU_SENSE:
270 		((struct stat *) m)->st_blksize = so->so_snd.sb_hiwat;
271 		(void) splx(s);
272 		return (0);
273 
274 	case PRU_RCVOOB:
275 		if ((so->so_oobmark == 0 &&
276 		    (so->so_state & SS_RCVATMARK) == 0) ||
277 		    so->so_options & SO_OOBINLINE ||
278 		    tp->t_oobflags & TCPOOB_HADDATA) {
279 			error = EINVAL;
280 			break;
281 		}
282 		if ((tp->t_oobflags & TCPOOB_HAVEDATA) == 0) {
283 			error = EWOULDBLOCK;
284 			break;
285 		}
286 		m->m_len = 1;
287 		*mtod(m, caddr_t) = tp->t_iobc;
288 		if (((int)nam & MSG_PEEK) == 0)
289 			tp->t_oobflags ^= (TCPOOB_HAVEDATA | TCPOOB_HADDATA);
290 		break;
291 
292 	case PRU_SENDOOB:
293 		if (sbspace(&so->so_snd) < -512) {
294 			m_freem(m);
295 			error = ENOBUFS;
296 			break;
297 		}
298 		/*
299 		 * According to RFC961 (Assigned Protocols),
300 		 * the urgent pointer points to the last octet
301 		 * of urgent data.  We continue, however,
302 		 * to consider it to indicate the first octet
303 		 * of data past the urgent section.
304 		 * Otherwise, snd_up should be one lower.
305 		 */
306 		sbappend(&so->so_snd, m);
307 		tp->snd_up = tp->snd_una + so->so_snd.sb_cc;
308 		tp->t_force = 1;
309 		error = tcp_output(tp);
310 		tp->t_force = 0;
311 		break;
312 
313 	case PRU_SOCKADDR:
314 		in_setsockaddr(inp, nam);
315 		break;
316 
317 	case PRU_PEERADDR:
318 		in_setpeeraddr(inp, nam);
319 		break;
320 
321 	/*
322 	 * TCP slow timer went off; going through this
323 	 * routine for tracing's sake.
324 	 */
325 	case PRU_SLOWTIMO:
326 		tp = tcp_timers(tp, (int)nam);
327 		req |= (int)nam << 8;		/* for debug's sake */
328 		break;
329 
330 	default:
331 		panic("tcp_usrreq");
332 	}
333 	if (tp && (so->so_options & SO_DEBUG))
334 		tcp_trace(TA_USER, ostate, tp, (struct tcpiphdr *)0, req);
335 	splx(s);
336 	return (error);
337 }
338 
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;
346 	struct inpcb *inp = sotoinpcb(so);
347 	register struct tcpcb *tp = intotcpcb(inp);
348 	register struct mbuf *m;
349 
350 	if (level != IPPROTO_TCP)
351 		return (ip_ctloutput(op, so, level, optname, mp));
352 
353 	switch (op) {
354 
355 	case PRCO_SETOPT:
356 		m = *mp;
357 		switch (optname) {
358 
359 		case TCP_NODELAY:
360 			if (m == NULL || m->m_len < sizeof (int))
361 				error = EINVAL;
362 			else if (*mtod(m, int *))
363 				tp->t_flags |= TF_NODELAY;
364 			else
365 				tp->t_flags &= ~TF_NODELAY;
366 			break;
367 
368 		case TCP_MAXSEG:	/* not yet */
369 		default:
370 			error = EINVAL;
371 			break;
372 		}
373 		if (m)
374 			(void) m_free(m);
375 		break;
376 
377 	case PRCO_GETOPT:
378 		*mp = m = m_get(M_WAIT, MT_SOOPTS);
379 		m->m_len = sizeof(int);
380 
381 		switch (optname) {
382 		case TCP_NODELAY:
383 			*mtod(m, int *) = tp->t_flags & TF_NODELAY;
384 			break;
385 		case TCP_MAXSEG:
386 			*mtod(m, int *) = tp->t_maxseg;
387 			break;
388 		default:
389 			error = EINVAL;
390 			break;
391 		}
392 		break;
393 	}
394 	return (error);
395 }
396 
397 u_long	tcp_sendspace = 1024*4;
398 u_long	tcp_recvspace = 1024*4;
399 
400 /*
401  * Attach TCP protocol to socket, allocating
402  * internet protocol control block, tcp control block,
403  * bufer space, and entering LISTEN state if to accept connections.
404  */
405 tcp_attach(so)
406 	struct socket *so;
407 {
408 	register struct tcpcb *tp;
409 	struct inpcb *inp;
410 	int error;
411 
412 	if (so->so_snd.sb_hiwat == 0 || so->so_rcv.sb_hiwat == 0) {
413 		error = soreserve(so, tcp_sendspace, tcp_recvspace);
414 		if (error)
415 			return (error);
416 	}
417 	error = in_pcballoc(so, &tcb);
418 	if (error)
419 		return (error);
420 	inp = sotoinpcb(so);
421 	tp = tcp_newtcpcb(inp);
422 	if (tp == 0) {
423 		int nofd = so->so_state & SS_NOFDREF;	/* XXX */
424 
425 		so->so_state &= ~SS_NOFDREF;	/* don't free the socket yet */
426 		in_pcbdetach(inp);
427 		so->so_state |= nofd;
428 		return (ENOBUFS);
429 	}
430 	tp->t_state = TCPS_CLOSED;
431 	return (0);
432 }
433 
434 /*
435  * Initiate (or continue) disconnect.
436  * If embryonic state, just send reset (once).
437  * If in ``let data drain'' option and linger null, just drop.
438  * Otherwise (hard), mark socket disconnecting and drop
439  * current input data; switch states based on user close, and
440  * send segment to peer (with FIN).
441  */
442 struct tcpcb *
443 tcp_disconnect(tp)
444 	register struct tcpcb *tp;
445 {
446 	struct socket *so = tp->t_inpcb->inp_socket;
447 
448 	if (tp->t_state < TCPS_ESTABLISHED)
449 		tp = tcp_close(tp);
450 	else if ((so->so_options & SO_LINGER) && so->so_linger == 0)
451 		tp = tcp_drop(tp, 0);
452 	else {
453 		soisdisconnecting(so);
454 		sbflush(&so->so_rcv);
455 		tp = tcp_usrclosed(tp);
456 		if (tp)
457 			(void) tcp_output(tp);
458 	}
459 	return (tp);
460 }
461 
462 /*
463  * User issued close, and wish to trail through shutdown states:
464  * if never received SYN, just forget it.  If got a SYN from peer,
465  * but haven't sent FIN, then go to FIN_WAIT_1 state to send peer a FIN.
466  * If already got a FIN from peer, then almost done; go to LAST_ACK
467  * state.  In all other cases, have already sent FIN to peer (e.g.
468  * after PRU_SHUTDOWN), and just have to play tedious game waiting
469  * for peer to send FIN or not respond to keep-alives, etc.
470  * We can let the user exit from the close as soon as the FIN is acked.
471  */
472 struct tcpcb *
473 tcp_usrclosed(tp)
474 	register struct tcpcb *tp;
475 {
476 
477 	switch (tp->t_state) {
478 
479 	case TCPS_CLOSED:
480 	case TCPS_LISTEN:
481 	case TCPS_SYN_SENT:
482 		tp->t_state = TCPS_CLOSED;
483 		tp = tcp_close(tp);
484 		break;
485 
486 	case TCPS_SYN_RECEIVED:
487 	case TCPS_ESTABLISHED:
488 		tp->t_state = TCPS_FIN_WAIT_1;
489 		break;
490 
491 	case TCPS_CLOSE_WAIT:
492 		tp->t_state = TCPS_LAST_ACK;
493 		break;
494 	}
495 	if (tp && tp->t_state >= TCPS_FIN_WAIT_2)
496 		soisdisconnected(tp->t_inpcb->inp_socket);
497 	return (tp);
498 }
499