1/*
2 * Copyright (c) 1982, 1986, 1988, 1990, 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. Neither the name of the University nor the names of its contributors
14 *    may be used to endorse or promote products derived from this software
15 *    without specific prior written permission.
16 *
17 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
18 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
19 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
20 * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
21 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
22 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
23 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
24 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
25 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
26 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
27 * SUCH DAMAGE.
28 *
29 *	@(#)tcp_subr.c	8.1 (Berkeley) 6/10/93
30 * tcp_subr.c,v 1.5 1994/10/08 22:39:58 phk Exp
31 */
32
33/*
34 * Changes and additions relating to SLiRP
35 * Copyright (c) 1995 Danny Gasparovski.
36 *
37 * Please read the file COPYRIGHT for the
38 * terms and conditions of the copyright.
39 */
40
41#define WANT_SYS_IOCTL_H
42#include <slirp.h>
43#include "proxy_common.h"
44
45/* patchable/settable parameters for tcp */
46/* Don't do rfc1323 performance enhancements */
47#define TCP_DO_RFC1323 0
48
49/*
50 * Tcp initialization
51 */
52void
53tcp_init(void)
54{
55	tcp_iss = 1;		/* wrong */
56	tcb.so_next = tcb.so_prev = &tcb;
57}
58
59/*
60 * Create template to be used to send tcp packets on a connection.
61 * Call after host entry created, fills
62 * in a skeletal tcp/ip header, minimizing the amount of work
63 * necessary when the connection is used.
64 */
65/* struct tcpiphdr * */
66void
67tcp_template(struct tcpcb *tp)
68{
69	struct socket *so = tp->t_socket;
70	register struct tcpiphdr *n = &tp->t_template;
71
72	n->ti_mbuf = NULL;
73	n->ti_x1 = 0;
74	n->ti_pr = IPPROTO_TCP;
75	n->ti_len = htons(sizeof (struct tcpiphdr) - sizeof (struct ip));
76	n->ti_src = ip_seth(so->so_faddr_ip);
77	n->ti_dst = ip_seth(so->so_laddr_ip);
78	n->ti_sport = port_seth(so->so_faddr_port);
79	n->ti_dport = port_seth(so->so_laddr_port);
80
81	n->ti_seq = 0;
82	n->ti_ack = 0;
83	n->ti_x2 = 0;
84	n->ti_off = 5;
85	n->ti_flags = 0;
86	n->ti_win = 0;
87	n->ti_sum = 0;
88	n->ti_urp = 0;
89}
90
91/*
92 * Send a single message to the TCP at address specified by
93 * the given TCP/IP header.  If m == 0, then we make a copy
94 * of the tcpiphdr at ti and send directly to the addressed host.
95 * This is used to force keep alive messages out using the TCP
96 * template for a connection tp->t_template.  If flags are given
97 * then we send a message back to the TCP which originated the
98 * segment ti, and discard the mbuf containing it and any other
99 * attached mbufs.
100 *
101 * In any case the ack and sequence number of the transmitted
102 * segment are as specified by the parameters.
103 */
104void
105tcp_respond(struct tcpcb *tp, struct tcpiphdr *ti, struct mbuf *m,
106            tcp_seq ack, tcp_seq seq, int flags)
107{
108	register int tlen;
109	int win = 0;
110
111	DEBUG_CALL("tcp_respond");
112	DEBUG_ARG("tp = %lx", (long)tp);
113	DEBUG_ARG("ti = %lx", (long)ti);
114	DEBUG_ARG("m = %lx", (long)m);
115	DEBUG_ARG("ack = %u", ack);
116	DEBUG_ARG("seq = %u", seq);
117	DEBUG_ARG("flags = %x", flags);
118
119	if (tp)
120		win = sbspace(&tp->t_socket->so_rcv);
121        if (m == NULL) {
122		if ((m = m_get()) == NULL)
123			return;
124#ifdef TCP_COMPAT_42
125		tlen = 1;
126#else
127		tlen = 0;
128#endif
129		m->m_data += IF_MAXLINKHDR;
130		*mtod(m, struct tcpiphdr *) = *ti;
131		ti = mtod(m, struct tcpiphdr *);
132		flags = TH_ACK;
133	} else {
134		/*
135		 * ti points into m so the next line is just making
136		 * the mbuf point to ti
137		 */
138		m->m_data = (caddr_t)ti;
139
140		m->m_len = sizeof (struct tcpiphdr);
141		tlen = 0;
142#define xchg(a,b,type) { type t; t=a; a=b; b=t; }
143		xchg(ti->ti_dst, ti->ti_src, ipaddr_t);
144		xchg(ti->ti_dport, ti->ti_sport, port_t);
145#undef xchg
146	}
147	ti->ti_len = htons((u_short)(sizeof (struct tcphdr) + tlen));
148	tlen += sizeof (struct tcpiphdr);
149	m->m_len = tlen;
150
151        ti->ti_mbuf = NULL;
152	ti->ti_x1 = 0;
153	ti->ti_seq = htonl(seq);
154	ti->ti_ack = htonl(ack);
155	ti->ti_x2 = 0;
156	ti->ti_off = sizeof (struct tcphdr) >> 2;
157	ti->ti_flags = flags;
158	if (tp)
159		ti->ti_win = htons((u_int16_t) (win >> tp->rcv_scale));
160	else
161		ti->ti_win = htons((u_int16_t)win);
162	ti->ti_urp = 0;
163	ti->ti_sum = 0;
164	ti->ti_sum = cksum(m, tlen);
165	((struct ip *)ti)->ip_len = tlen;
166
167	if(flags & TH_RST)
168	  ((struct ip *)ti)->ip_ttl = MAXTTL;
169	else
170	  ((struct ip *)ti)->ip_ttl = IPDEFTTL;
171
172	(void) ip_output((struct socket *)0, m);
173}
174
175/*
176 * Create a new TCP control block, making an
177 * empty reassembly queue and hooking it to the argument
178 * protocol control block.
179 */
180struct tcpcb *
181tcp_newtcpcb(struct socket *so)
182{
183	register struct tcpcb *tp;
184
185	tp = (struct tcpcb *)malloc(sizeof(*tp));
186	if (tp == NULL)
187		return ((struct tcpcb *)0);
188
189	memset((char *) tp, 0, sizeof(struct tcpcb));
190	tp->seg_next = tp->seg_prev = (struct tcpiphdr*)tp;
191	tp->t_maxseg = TCP_MSS;
192
193	tp->t_flags = TCP_DO_RFC1323 ? (TF_REQ_SCALE|TF_REQ_TSTMP) : 0;
194	tp->t_socket = so;
195
196	/*
197	 * Init srtt to TCPTV_SRTTBASE (0), so we can tell that we have no
198	 * rtt estimate.  Set rttvar so that srtt + 2 * rttvar gives
199	 * reasonable initial retransmit time.
200	 */
201	tp->t_srtt = TCPTV_SRTTBASE;
202	tp->t_rttvar = TCPTV_SRTTDFLT << 2;
203	tp->t_rttmin = TCPTV_MIN;
204
205	TCPT_RANGESET(tp->t_rxtcur,
206	    ((TCPTV_SRTTBASE >> 2) + (TCPTV_SRTTDFLT << 2)) >> 1,
207	    TCPTV_MIN, TCPTV_REXMTMAX);
208
209	tp->snd_cwnd = TCP_MAXWIN << TCP_MAX_WINSHIFT;
210	tp->snd_ssthresh = TCP_MAXWIN << TCP_MAX_WINSHIFT;
211	tp->t_state = TCPS_CLOSED;
212
213	so->so_tcpcb = tp;
214
215	return (tp);
216}
217
218/*
219 * Drop a TCP connection, reporting
220 * the specified error.  If connection is synchronized,
221 * then send a RST to peer.
222 */
223struct tcpcb *tcp_drop(struct tcpcb *tp, int err)
224{
225/* tcp_drop(tp, errno)
226	register struct tcpcb *tp;
227	int errno;
228{
229*/
230
231	DEBUG_CALL("tcp_drop");
232	DEBUG_ARG("tp = %lx", (long)tp);
233	DEBUG_ARG("errno = %d", errno);
234
235	if (TCPS_HAVERCVDSYN(tp->t_state)) {
236		tp->t_state = TCPS_CLOSED;
237		(void) tcp_output(tp);
238		STAT(tcpstat.tcps_drops++);
239	} else
240		STAT(tcpstat.tcps_conndrops++);
241/*	if (errno == ETIMEDOUT && tp->t_softerror)
242 *		errno = tp->t_softerror;
243 */
244/*	so->so_error = errno; */
245	return (tcp_close(tp));
246}
247
248/*
249 * Close a TCP control block:
250 *	discard all space held by the tcp
251 *	discard internet protocol block
252 *	wake up any sleepers
253 */
254struct tcpcb *
255tcp_close(struct tcpcb *tp)
256{
257	register struct tcpiphdr *t;
258	struct socket *so = tp->t_socket;
259	register struct mbuf *m;
260
261	DEBUG_CALL("tcp_close");
262	DEBUG_ARG("tp = %lx", (long )tp);
263
264	/* free the reassembly queue, if any */
265	t = tcpfrag_list_first(tp);
266	while (!tcpfrag_list_end(t, tp)) {
267		t = tcpiphdr_next(t);
268		m = tcpiphdr_prev(t)->ti_mbuf;
269		remque(tcpiphdr2qlink(tcpiphdr_prev(t)));
270		m_freem(m);
271	}
272	/* It's static */
273/*	if (tp->t_template)
274 *		(void) m_free(dtom(tp->t_template));
275 */
276/*	free(tp, M_PCB);  */
277	free(tp);
278        so->so_tcpcb = NULL;
279	soisfdisconnected(so);
280	/* clobber input socket cache if we're closing the cached connection */
281	if (so == tcp_last_so)
282		tcp_last_so = &tcb;
283	socket_close(so->s);
284	sbfree(&so->so_rcv);
285	sbfree(&so->so_snd);
286	sofree(so);
287	STAT(tcpstat.tcps_closed++);
288	return ((struct tcpcb *)0);
289}
290
291#ifdef notdef
292void
293tcp_drain()
294{
295	/* XXX */
296}
297
298/*
299 * When a source quench is received, close congestion window
300 * to one segment.  We will gradually open it again as we proceed.
301 */
302void
303tcp_quench(i, errno)
304
305	int errno;
306{
307	struct tcpcb *tp = intotcpcb(inp);
308
309	if (tp)
310		tp->snd_cwnd = tp->t_maxseg;
311}
312
313#endif /* notdef */
314
315/*
316 * TCP protocol interface to socket abstraction.
317 */
318
319/*
320 * User issued close, and wish to trail through shutdown states:
321 * if never received SYN, just forget it.  If got a SYN from peer,
322 * but haven't sent FIN, then go to FIN_WAIT_1 state to send peer a FIN.
323 * If already got a FIN from peer, then almost done; go to LAST_ACK
324 * state.  In all other cases, have already sent FIN to peer (e.g.
325 * after PRU_SHUTDOWN), and just have to play tedious game waiting
326 * for peer to send FIN or not respond to keep-alives, etc.
327 * We can let the user exit from the close as soon as the FIN is acked.
328 */
329void
330tcp_sockclosed(struct tcpcb *tp)
331{
332
333	DEBUG_CALL("tcp_sockclosed");
334	DEBUG_ARG("tp = %lx", (long)tp);
335
336	switch (tp->t_state) {
337
338	case TCPS_CLOSED:
339	case TCPS_LISTEN:
340	case TCPS_SYN_SENT:
341		tp->t_state = TCPS_CLOSED;
342		tp = tcp_close(tp);
343		break;
344
345	case TCPS_SYN_RECEIVED:
346	case TCPS_ESTABLISHED:
347		tp->t_state = TCPS_FIN_WAIT_1;
348		break;
349
350	case TCPS_CLOSE_WAIT:
351		tp->t_state = TCPS_LAST_ACK;
352		break;
353	}
354/*	soisfdisconnecting(tp->t_socket); */
355	if (tp && tp->t_state >= TCPS_FIN_WAIT_2)
356		soisfdisconnected(tp->t_socket);
357	if (tp)
358		tcp_output(tp);
359}
360
361static void
362tcp_proxy_event( struct socket*  so,
363                 int             s,
364                 ProxyEvent      event )
365{
366    so->so_state &= ~SS_PROXIFIED;
367
368    if (event == PROXY_EVENT_CONNECTED) {
369        so->s         = s;
370        so->so_state &= ~(SS_ISFCONNECTING);
371    }
372    else {
373        so->so_state = SS_NOFDREF;
374    }
375
376    /* continue the connect */
377    tcp_input(NULL, sizeof(struct ip), so);
378}
379
380
381/* Tests if an IP address corresponds to a special Qemu service (eg, the DNS
382 * server or the gateway - see ctl.h) and if so, rewrites it with the real
383 * address of the service.
384 */
385int is_qemu_special_address(unsigned long dst_addr, unsigned long *redir_addr)
386{
387  if ((dst_addr & 0xffffff00) == special_addr_ip) {
388    /* It's an alias */
389
390    int  last_byte = dst_addr & 0xff;
391
392    if (CTL_IS_DNS(last_byte))
393      *redir_addr = dns_addr[last_byte - CTL_DNS];
394    else
395      *redir_addr = loopback_addr_ip;
396
397    return 1;
398  }
399  return 0;
400}
401
402
403/*
404 * Connect to a host on the Internet
405 * Called by tcp_input
406 * Only do a connect, the tcp fields will be set in tcp_input
407 * return 0 if there's a result of the connect,
408 * else return -1 means we're still connecting
409 * The return value is almost always -1 since the socket is
410 * nonblocking.  Connect returns after the SYN is sent, and does
411 * not wait for ACK+SYN.
412 */
413int tcp_fconnect(struct socket *so)
414{
415  int ret=0;
416    int try_proxy = 0;
417    SockAddress    sockaddr;
418    unsigned long       sock_ip;
419    int                 sock_port;
420    int redirect_happened = 0; /* for logging new src ip/port after connect */
421
422    DEBUG_CALL("tcp_fconnect");
423    DEBUG_ARG("so = %lx", (long )so);
424
425    /* when true, a connection that otherwise would be dropped will instead be
426     * redirected to the sink ('-net-forward-tcp2sink') */
427    int forward_dropped_to_sink = 0;
428    time_t timestamp = time(NULL);
429
430    /*-------------------------------------------------------------*/
431    /* User mode network stack restrictions */
432    if (slirp_should_drop(so->so_faddr_ip, so->so_faddr_port, IPPROTO_TCP)) {
433
434      /* If forwarding to sink is enabled, don't actually drop it */
435      if (slirp_should_forward_dropped_tcp2sink()) {
436        slirp_drop_log(
437            "About to be dropped TCP allowed to sink: "
438            "0x%08lx:0x%04x -> 0x%08lx:0x%04x %ld\n",
439              so->so_laddr_ip,
440              so->so_laddr_port,
441              so->so_faddr_ip,
442              so->so_faddr_port,
443              timestamp
444        );
445        forward_dropped_to_sink = 1;
446      }
447      else {
448        slirp_drop_log(
449            "Dropped TCP: 0x%08lx:0x%04x -> 0x%08lx:0x%04x %ld\n",
450            so->so_laddr_ip,
451            so->so_laddr_port,
452            so->so_faddr_ip,
453            so->so_faddr_port,
454            timestamp
455        );
456        //errno = ENETUNREACH;
457        errno = ECONNREFUSED;
458        return -1;
459      }
460    } else {
461      slirp_drop_log(
462          "Allowed TCP: 0x%08lx:0x%04x -> 0x%08lx:0x%04x %ld\n",
463          so->so_laddr_ip,
464          so->so_laddr_port,
465          so->so_faddr_ip,
466          so->so_faddr_port,
467          timestamp
468      );
469    }
470    /*-------------------------------------------------------------*/
471
472    if ((ret=so->s=socket_create_inet(SOCKET_STREAM)) >= 0)
473    {
474        int  s = so->s;
475
476        socket_set_nonblock(s);
477        socket_set_xreuseaddr(s);
478        socket_set_oobinline(s);
479
480
481        if (forward_dropped_to_sink) {
482
483          /* This connection would normally be dropped, but since forwarding of
484           * dropped connections is enabled, redirect it to the sink */
485          sock_ip = slirp_get_tcp_sink_ip();
486          sock_port= slirp_get_tcp_sink_port();
487          redirect_happened = 1;
488        }
489        else {   /* An allowed connection */
490
491          unsigned long faddr;
492          int fport;
493
494          /* Determine if the connection should be redirected
495           * due to a -net-forward rule */
496          /* faddr and fport are modified only on success */
497          if (slirp_should_net_forward(so->so_faddr_ip, so->so_faddr_port,
498                                       &faddr, &fport)) {
499            redirect_happened = 1;
500            sock_ip = faddr;              /* forced dst addr */
501            sock_port= fport;                 /* forced dst port */
502          }
503          /* Determine if this is a connection to a special qemu service,
504           * and change the destination address accordingly.
505           * 'faddr' is modified only on success  */
506          else if (is_qemu_special_address(so->so_faddr_ip, &faddr)) {
507
508          /* We keep the original destination port. If a special service
509           * listens on a different port than the standard, then appropriate
510           * forwarding should be set up using -net-forward, e.g., as it is
511           * the case with Mawler's DNS traffic, which is redirected to the
512           * special DNS port:
513           * -net-forward 0.0.0.0:0.0.0.0:53:127.0.0.1:21737 */
514
515            sock_ip = faddr;         /* real DNS/gateway addr */
516            sock_port= so->so_faddr_port;/* original dst port */
517
518          }
519          /* A normal connection - keep the original destination addr/port */
520          else {
521
522            try_proxy = 1;
523
524            sock_ip = so->so_faddr_ip;  /* original dst addr */
525            sock_port= so->so_faddr_port;   /* original dst port */
526          }
527        }
528
529        DEBUG_MISC((dfd, " connect()ing, addr=%s, proxy=%d\n",
530                    sock_address_to_string(&sockaddr), try_proxy));
531
532        sock_address_init_inet( &sockaddr, sock_ip, sock_port );
533
534        if (try_proxy) {
535            if (!proxy_manager_add(&sockaddr, SOCKET_STREAM,
536                                   (ProxyEventFunc) tcp_proxy_event, so)) {
537                soisfconnecting(so);
538                so->s         = -1;
539                so->so_state |= SS_PROXIFIED;
540                return 0;
541            }
542        }
543
544        /* We don't care what port we get */
545        socket_connect(s, &sockaddr);
546
547        if (redirect_happened) {
548          SockAddress local_addr;
549          if (socket_get_address(s, &local_addr)) {
550            fprintf (stderr,
551                     "Warning: tcp_fconnect: could not get socket name\n");
552          }
553          slirp_drop_log(
554              "Redirected TCP: orig 0x%08lx:0x%04x -> 0x%08lx:0x%04x "
555              "new 0x%08lx:0x%04x -> 0x%08lx:0x%04x %ld\n",
556              so->so_laddr_ip, so->so_laddr_port,
557              so->so_faddr_ip, so->so_laddr_port,
558              sock_address_get_ip(&local_addr),
559              sock_address_get_port(&local_addr),
560              sock_ip, sock_port, timestamp
561          );
562        }
563
564       /*
565        * If it's not in progress, it failed, so we just return 0,
566        * without clearing SS_NOFDREF
567        */
568        soisfconnecting(so);
569  }
570
571  return(ret);
572}
573
574/*
575 * Accept the socket and connect to the local-host
576 *
577 * We have a problem. The correct thing to do would be
578 * to first connect to the local-host, and only if the
579 * connection is accepted, then do an accept() here.
580 * But, a) we need to know who's trying to connect
581 * to the socket to be able to SYN the local-host, and
582 * b) we are already connected to the foreign host by
583 * the time it gets to accept(), so... We simply accept
584 * here and SYN the local-host.
585 */
586void
587tcp_connect(struct socket *inso)
588{
589	struct socket *so;
590	SockAddress   addr;
591	uint32_t      addr_ip;
592	struct tcpcb *tp;
593	int s;
594
595	DEBUG_CALL("tcp_connect");
596	DEBUG_ARG("inso = %lx", (long)inso);
597
598	/*
599	 * If it's an SS_ACCEPTONCE socket, no need to socreate()
600	 * another socket, just use the accept() socket.
601	 */
602	if (inso->so_state & SS_FACCEPTONCE) {
603		/* FACCEPTONCE already have a tcpcb */
604		so = inso;
605	} else {
606		if ((so = socreate()) == NULL) {
607			/* If it failed, get rid of the pending connection */
608			socket_close(socket_accept(inso->s, NULL));
609			return;
610		}
611		if (tcp_attach(so) < 0) {
612			free(so); /* NOT sofree */
613			return;
614		}
615		so->so_laddr_ip   = inso->so_laddr_ip;
616		so->so_laddr_port = inso->so_laddr_port;
617	}
618
619	(void) tcp_mss(sototcpcb(so), 0);
620
621	if ((s = socket_accept(inso->s, &addr)) < 0) {
622		tcp_close(sototcpcb(so)); /* This will sofree() as well */
623		return;
624	}
625	socket_set_nonblock(s);
626	socket_set_xreuseaddr(s);
627	socket_set_oobinline(s);
628	socket_set_nodelay(s);
629
630	so->so_faddr_port = sock_address_get_port(&addr);
631
632	addr_ip = sock_address_get_ip(&addr);
633
634	so->so_faddr_ip = addr_ip;
635	/* Translate connections from localhost to the real hostname */
636	if (addr_ip == 0 || addr_ip == loopback_addr_ip)
637	   so->so_faddr_ip = alias_addr_ip;
638
639	/* Close the accept() socket, set right state */
640	if (inso->so_state & SS_FACCEPTONCE) {
641		socket_close(so->s); /* If we only accept once, close the accept() socket */
642		so->so_state = SS_NOFDREF; /* Don't select it yet, even though we have an FD */
643					   /* if it's not FACCEPTONCE, it's already NOFDREF */
644	}
645	so->s = s;
646
647	so->so_iptos = tcp_tos(so);
648	tp = sototcpcb(so);
649
650	tcp_template(tp);
651
652	/* Compute window scaling to request.  */
653/*	while (tp->request_r_scale < TCP_MAX_WINSHIFT &&
654 *		(TCP_MAXWIN << tp->request_r_scale) < so->so_rcv.sb_hiwat)
655 *		tp->request_r_scale++;
656 */
657
658/*	soisconnecting(so); */ /* NOFDREF used instead */
659	STAT(tcpstat.tcps_connattempt++);
660
661	tp->t_state = TCPS_SYN_SENT;
662	tp->t_timer[TCPT_KEEP] = TCPTV_KEEP_INIT;
663	tp->iss = tcp_iss;
664	tcp_iss += TCP_ISSINCR/2;
665	tcp_sendseqinit(tp);
666	tcp_output(tp);
667}
668
669/*
670 * Attach a TCPCB to a socket.
671 */
672int
673tcp_attach(struct socket *so)
674{
675	if ((so->so_tcpcb = tcp_newtcpcb(so)) == NULL)
676	   return -1;
677
678	insque(so, &tcb);
679
680	return 0;
681}
682
683/*
684 * Set the socket's type of service field
685 */
686static const struct tos_t tcptos[] = {
687	  {0, 20, IPTOS_THROUGHPUT, 0},	/* ftp data */
688	  {21, 21, IPTOS_LOWDELAY,  EMU_FTP},	/* ftp control */
689	  {0, 23, IPTOS_LOWDELAY, 0},	/* telnet */
690	  {0, 80, IPTOS_THROUGHPUT, 0},	/* WWW */
691	  {0, 513, IPTOS_LOWDELAY, EMU_RLOGIN|EMU_NOCONNECT},	/* rlogin */
692	  {0, 514, IPTOS_LOWDELAY, EMU_RSH|EMU_NOCONNECT},	/* shell */
693	  {0, 544, IPTOS_LOWDELAY, EMU_KSH},		/* kshell */
694	  {0, 543, IPTOS_LOWDELAY, 0},	/* klogin */
695	  {0, 6667, IPTOS_THROUGHPUT, EMU_IRC},	/* IRC */
696	  {0, 6668, IPTOS_THROUGHPUT, EMU_IRC},	/* IRC undernet */
697	  {0, 7070, IPTOS_LOWDELAY, EMU_REALAUDIO }, /* RealAudio control */
698	  {0, 113, IPTOS_LOWDELAY, EMU_IDENT }, /* identd protocol */
699	  {0, 0, 0, 0}
700};
701
702#ifdef CONFIG_QEMU
703static
704#endif
705struct emu_t *tcpemu = NULL;
706
707/*
708 * Return TOS according to the above table
709 */
710u_int8_t
711tcp_tos(struct socket *so)
712{
713	int i = 0;
714	struct emu_t *emup;
715
716	while(tcptos[i].tos) {
717		if ((tcptos[i].fport && so->so_faddr_port == tcptos[i].fport) ||
718		    (tcptos[i].lport && so->so_laddr_port == tcptos[i].lport)) {
719			so->so_emu = tcptos[i].emu;
720			return tcptos[i].tos;
721		}
722		i++;
723	}
724
725	/* Nope, lets see if there's a user-added one */
726	for (emup = tcpemu; emup; emup = emup->next) {
727		if ((emup->fport && (so->so_faddr_port == emup->fport)) ||
728		    (emup->lport && (so->so_laddr_port == emup->lport))) {
729			so->so_emu = emup->emu;
730			return emup->tos;
731		}
732	}
733	return 0;
734}
735
736#if 0
737int do_echo = -1;
738#endif
739
740/*
741 * Emulate programs that try and connect to us
742 * This includes ftp (the data connection is
743 * initiated by the server) and IRC (DCC CHAT and
744 * DCC SEND) for now
745 *
746 * NOTE: It's possible to crash SLiRP by sending it
747 * unstandard strings to emulate... if this is a problem,
748 * more checks are needed here
749 *
750 * XXX Assumes the whole command came in one packet
751 *
752 * XXX Some ftp clients will have their TOS set to
753 * LOWDELAY and so Nagel will kick in.  Because of this,
754 * we'll get the first letter, followed by the rest, so
755 * we simply scan for ORT instead of PORT...
756 * DCC doesn't have this problem because there's other stuff
757 * in the packet before the DCC command.
758 *
759 * Return 1 if the mbuf m is still valid and should be
760 * sbappend()ed
761 *
762 * NOTE: if you return 0 you MUST m_free() the mbuf!
763 */
764int
765tcp_emu(struct socket *so, struct mbuf *m)
766{
767	u_int n1, n2, n3, n4, n5, n6;
768        char buff[257];
769	u_int32_t laddr;
770	u_int lport;
771	char *bptr;
772
773	DEBUG_CALL("tcp_emu");
774	DEBUG_ARG("so = %lx", (long)so);
775	DEBUG_ARG("m = %lx", (long)m);
776
777	switch(so->so_emu) {
778		int x, i;
779
780	 case EMU_IDENT:
781		/*
782		 * Identification protocol as per rfc-1413
783		 */
784
785		{
786			struct socket *tmpso;
787			SockAddress    addr;
788			struct sbuf *so_rcv = &so->so_rcv;
789
790			memcpy(so_rcv->sb_wptr, m->m_data, m->m_len);
791			so_rcv->sb_wptr += m->m_len;
792			so_rcv->sb_rptr += m->m_len;
793			m->m_data[m->m_len] = 0; /* NULL terminate */
794			if (strchr(m->m_data, '\r') || strchr(m->m_data, '\n')) {
795				if (sscanf(so_rcv->sb_data, "%u%*[ ,]%u", &n1, &n2) == 2) {
796					/* n2 is the one on our host */
797					for (tmpso = tcb.so_next; tmpso != &tcb; tmpso = tmpso->so_next) {
798						if (tmpso->so_laddr_ip == so->so_laddr_ip &&
799						    tmpso->so_laddr_port == n2 &&
800						    tmpso->so_faddr_ip == so->so_faddr_ip &&
801						    tmpso->so_faddr_port == n1) {
802							if (socket_get_address(tmpso->s, &addr) == 0)
803							   n2 = sock_address_get_port(&addr);
804							break;
805						}
806					}
807				}
808                                so_rcv->sb_cc = snprintf(so_rcv->sb_data,
809                                                         so_rcv->sb_datalen,
810                                                         "%d,%d\r\n", n1, n2);
811				so_rcv->sb_rptr = so_rcv->sb_data;
812				so_rcv->sb_wptr = so_rcv->sb_data + so_rcv->sb_cc;
813			}
814			m_free(m);
815			return 0;
816		}
817
818        case EMU_FTP: /* ftp */
819                *(m->m_data+m->m_len) = 0; /* NUL terminate for strstr */
820		if ((bptr = (char *)strstr(m->m_data, "ORT")) != NULL) {
821			/*
822			 * Need to emulate the PORT command
823			 */
824			x = sscanf(bptr, "ORT %u,%u,%u,%u,%u,%u\r\n%256[^\177]",
825				   &n1, &n2, &n3, &n4, &n5, &n6, buff);
826			if (x < 6)
827			   return 1;
828
829			laddr = (n1 << 24) | (n2 << 16) | (n3 << 8) | (n4);
830			lport = (n5 << 8) | (n6);
831
832			if ((so = solisten(0, laddr, lport, SS_FACCEPTONCE)) == NULL)
833			   return 1;
834
835			n6 = so->so_faddr_port;
836
837			n5 = (n6 >> 8) & 0xff;
838			n6 &= 0xff;
839
840			laddr = so->so_faddr_ip;
841
842			n1 = ((laddr >> 24) & 0xff);
843			n2 = ((laddr >> 16) & 0xff);
844			n3 = ((laddr >> 8)  & 0xff);
845			n4 =  (laddr & 0xff);
846
847			m->m_len = bptr - m->m_data; /* Adjust length */
848                        m->m_len += snprintf(bptr, m->m_hdr.mh_size - m->m_len,
849                                             "ORT %d,%d,%d,%d,%d,%d\r\n%s",
850                                             n1, n2, n3, n4, n5, n6, x==7?buff:"");
851			return 1;
852		} else if ((bptr = (char *)strstr(m->m_data, "27 Entering")) != NULL) {
853			/*
854			 * Need to emulate the PASV response
855			 */
856			x = sscanf(bptr, "27 Entering Passive Mode (%u,%u,%u,%u,%u,%u)\r\n%256[^\177]",
857				   &n1, &n2, &n3, &n4, &n5, &n6, buff);
858			if (x < 6)
859			   return 1;
860
861			laddr = (n1 << 24) | (n2 << 16) | (n3 << 8) | (n4);
862			lport = (n5 << 8) | (n6);
863
864			if ((so = solisten(0, laddr, lport, SS_FACCEPTONCE)) == NULL)
865			   return 1;
866
867			n6 = so->so_faddr_port;
868
869			n5 = (n6 >> 8) & 0xff;
870			n6 &= 0xff;
871
872			laddr = so->so_faddr_ip;
873
874			n1 = ((laddr >> 24) & 0xff);
875			n2 = ((laddr >> 16) & 0xff);
876			n3 = ((laddr >> 8)  & 0xff);
877			n4 =  (laddr & 0xff);
878
879			m->m_len = bptr - m->m_data; /* Adjust length */
880			m->m_len += snprintf(bptr, m->m_hdr.mh_size - m->m_len,
881                                             "27 Entering Passive Mode (%d,%d,%d,%d,%d,%d)\r\n%s",
882                                             n1, n2, n3, n4, n5, n6, x==7?buff:"");
883
884			return 1;
885		}
886
887		return 1;
888
889	 case EMU_KSH:
890		/*
891		 * The kshell (Kerberos rsh) and shell services both pass
892		 * a local port port number to carry signals to the server
893		 * and stderr to the client.  It is passed at the beginning
894		 * of the connection as a NUL-terminated decimal ASCII string.
895		 */
896		so->so_emu = 0;
897		for (lport = 0, i = 0; i < m->m_len-1; ++i) {
898			if (m->m_data[i] < '0' || m->m_data[i] > '9')
899				return 1;       /* invalid number */
900			lport *= 10;
901			lport += m->m_data[i] - '0';
902		}
903		if (m->m_data[m->m_len-1] == '\0' && lport != 0 &&
904		    (so = solisten(0, so->so_laddr_ip, lport, SS_FACCEPTONCE)) != NULL)
905			m->m_len = snprintf(m->m_data, m->m_hdr.mh_size, "%d",
906                                so->so_faddr_port) + 1;
907		return 1;
908
909	 case EMU_IRC:
910		/*
911		 * Need to emulate DCC CHAT, DCC SEND and DCC MOVE
912		 */
913		*(m->m_data+m->m_len) = 0; /* NULL terminate the string for strstr */
914		if ((bptr = (char *)strstr(m->m_data, "DCC")) == NULL)
915			 return 1;
916
917		/* The %256s is for the broken mIRC */
918		if (sscanf(bptr, "DCC CHAT %256s %u %u", buff, &laddr, &lport) == 3) {
919			if ((so = solisten(0, laddr, lport, SS_FACCEPTONCE)) == NULL)
920				return 1;
921
922			m->m_len = bptr - m->m_data; /* Adjust length */
923                        m->m_len += snprintf(bptr, m->m_hdr.mh_size,
924                                             "DCC CHAT chat %lu %u%c\n",
925			     (unsigned long) so->so_faddr_ip,
926			     so->so_faddr_port, 1);
927		} else if (sscanf(bptr, "DCC SEND %256s %u %u %u", buff, &laddr, &lport, &n1) == 4) {
928			if ((so = solisten(0, laddr, lport, SS_FACCEPTONCE)) == NULL)
929				return 1;
930
931			m->m_len = bptr - m->m_data; /* Adjust length */
932                        m->m_len += snprintf(bptr, m->m_hdr.mh_size,
933                                             "DCC SEND %s %u %u %u%c\n", buff,
934			      so->so_faddr_ip, so->so_faddr_port, n1, 1);
935		} else if (sscanf(bptr, "DCC MOVE %256s %u %u %u", buff, &laddr, &lport, &n1) == 4) {
936			if ((so = solisten(0, laddr, lport, SS_FACCEPTONCE)) == NULL)
937				return 1;
938
939			m->m_len = bptr - m->m_data; /* Adjust length */
940                        m->m_len += snprintf(bptr, m->m_hdr.mh_size,
941                                             "DCC MOVE %s %lu %u %u%c\n", buff,
942			      (unsigned long)so->so_faddr_ip,
943			      so->so_faddr_port, n1, 1);
944		}
945		return 1;
946
947	 case EMU_REALAUDIO:
948                /*
949		 * RealAudio emulation - JP. We must try to parse the incoming
950		 * data and try to find the two characters that contain the
951		 * port number. Then we redirect an udp port and replace the
952		 * number with the real port we got.
953		 *
954		 * The 1.0 beta versions of the player are not supported
955		 * any more.
956		 *
957		 * A typical packet for player version 1.0 (release version):
958		 *
959		 * 0000:50 4E 41 00 05
960		 * 0000:00 01 00 02 1B D7 00 00 67 E6 6C DC 63 00 12 50 .....×..gælÜc..P
961		 * 0010:4E 43 4C 49 45 4E 54 20 31 30 31 20 41 4C 50 48 NCLIENT 101 ALPH
962		 * 0020:41 6C 00 00 52 00 17 72 61 66 69 6C 65 73 2F 76 Al..R..rafiles/v
963		 * 0030:6F 61 2F 65 6E 67 6C 69 73 68 5F 2E 72 61 79 42 oa/english_.rayB
964		 *
965		 * Now the port number 0x1BD7 is found at offset 0x04 of the
966		 * Now the port number 0x1BD7 is found at offset 0x04 of the
967		 * second packet. This time we received five bytes first and
968		 * then the rest. You never know how many bytes you get.
969		 *
970		 * A typical packet for player version 2.0 (beta):
971		 *
972		 * 0000:50 4E 41 00 06 00 02 00 00 00 01 00 02 1B C1 00 PNA...........Á.
973		 * 0010:00 67 75 78 F5 63 00 0A 57 69 6E 32 2E 30 2E 30 .guxõc..Win2.0.0
974		 * 0020:2E 35 6C 00 00 52 00 1C 72 61 66 69 6C 65 73 2F .5l..R..rafiles/
975		 * 0030:77 65 62 73 69 74 65 2F 32 30 72 65 6C 65 61 73 website/20releas
976		 * 0040:65 2E 72 61 79 53 00 00 06 36 42                e.rayS...6B
977		 *
978		 * Port number 0x1BC1 is found at offset 0x0d.
979		 *
980		 * This is just a horrible switch statement. Variable ra tells
981		 * us where we're going.
982		 */
983
984		bptr = m->m_data;
985		while (bptr < m->m_data + m->m_len) {
986			u_short p;
987			static int ra = 0;
988			char ra_tbl[4];
989
990			ra_tbl[0] = 0x50;
991			ra_tbl[1] = 0x4e;
992			ra_tbl[2] = 0x41;
993			ra_tbl[3] = 0;
994
995			switch (ra) {
996			 case 0:
997			 case 2:
998			 case 3:
999				if (*bptr++ != ra_tbl[ra]) {
1000					ra = 0;
1001					continue;
1002				}
1003				break;
1004
1005			 case 1:
1006				/*
1007				 * We may get 0x50 several times, ignore them
1008				 */
1009				if (*bptr == 0x50) {
1010					ra = 1;
1011					bptr++;
1012					continue;
1013				} else if (*bptr++ != ra_tbl[ra]) {
1014					ra = 0;
1015					continue;
1016				}
1017				break;
1018
1019			 case 4:
1020				/*
1021				 * skip version number
1022				 */
1023				bptr++;
1024				break;
1025
1026			 case 5:
1027				/*
1028				 * The difference between versions 1.0 and
1029				 * 2.0 is here. For future versions of
1030				 * the player this may need to be modified.
1031				 */
1032				if (*(bptr + 1) == 0x02)
1033				   bptr += 8;
1034				else
1035				   bptr += 4;
1036				break;
1037
1038			 case 6:
1039				/* This is the field containing the port
1040				 * number that RA-player is listening to.
1041				 */
1042				lport = (((u_char*)bptr)[0] << 8)
1043				+ ((u_char *)bptr)[1];
1044				if (lport < 6970)
1045				   lport += 256;   /* don't know why */
1046				if (lport < 6970 || lport > 7170)
1047				   return 1;       /* failed */
1048
1049				/* try to get udp port between 6970 - 7170 */
1050				for (p = 6970; p < 7071; p++) {
1051					if (udp_listen( p,
1052						       so->so_laddr_ip,
1053						       lport,
1054						       SS_FACCEPTONCE)) {
1055						break;
1056					}
1057				}
1058				if (p == 7071)
1059				   p = 0;
1060				*(u_char *)bptr++ = (p >> 8) & 0xff;
1061				*(u_char *)bptr++ = p & 0xff;
1062				ra = 0;
1063				return 1;   /* port redirected, we're done */
1064				break;
1065
1066			 default:
1067				ra = 0;
1068			}
1069			ra++;
1070		}
1071		return 1;
1072
1073	 default:
1074		/* Ooops, not emulated, won't call tcp_emu again */
1075		so->so_emu = 0;
1076		return 1;
1077	}
1078}
1079
1080/*
1081 * Do misc. config of SLiRP while its running.
1082 * Return 0 if this connections is to be closed, 1 otherwise,
1083 * return 2 if this is a command-line connection
1084 */
1085int
1086tcp_ctl(struct socket *so)
1087{
1088	struct sbuf *sb = &so->so_snd;
1089	int command;
1090 	struct ex_list *ex_ptr;
1091	int do_pty;
1092        //	struct socket *tmpso;
1093
1094	DEBUG_CALL("tcp_ctl");
1095	DEBUG_ARG("so = %lx", (long )so);
1096
1097#if 0
1098	/*
1099	 * Check if they're authorised
1100	 */
1101	if (ctl_addr_ip && (ctl_addr_ip == -1 || (so->so_laddr_ip != ctl_addr_ip))) {
1102		sb->sb_cc = sprintf(sb->sb_wptr,"Error: Permission denied.\r\n");
1103		sb->sb_wptr += sb->sb_cc;
1104		return 0;
1105	}
1106#endif
1107	command = (so->so_faddr_ip & 0xff);
1108
1109	switch(command) {
1110	default: /* Check for exec's */
1111
1112		/*
1113		 * Check if it's pty_exec
1114		 */
1115		for (ex_ptr = exec_list; ex_ptr; ex_ptr = ex_ptr->ex_next) {
1116			if (ex_ptr->ex_fport == so->so_faddr_port &&
1117			    command == ex_ptr->ex_addr) {
1118				if (ex_ptr->ex_pty == 3) {
1119					so->s = -1;
1120					so->extra = (void *)ex_ptr->ex_exec;
1121					return 1;
1122				}
1123				do_pty = ex_ptr->ex_pty;
1124				goto do_exec;
1125			}
1126		}
1127
1128		/*
1129		 * Nothing bound..
1130		 */
1131		/* tcp_fconnect(so); */
1132
1133		/* FALLTHROUGH */
1134	case CTL_ALIAS:
1135          sb->sb_cc = snprintf(sb->sb_wptr, sb->sb_datalen - (sb->sb_wptr - sb->sb_data),
1136                               "Error: No application configured.\r\n");
1137	  sb->sb_wptr += sb->sb_cc;
1138	  return(0);
1139
1140	do_exec:
1141		DEBUG_MISC((dfd, " executing %s \n",ex_ptr->ex_exec));
1142		return(fork_exec(so, ex_ptr->ex_exec, do_pty));
1143
1144#if 0
1145	case CTL_CMD:
1146	   for (tmpso = tcb.so_next; tmpso != &tcb; tmpso = tmpso->so_next) {
1147	     if (tmpso->so_emu == EMU_CTL &&
1148		 !(tmpso->so_tcpcb?
1149		   (tmpso->so_tcpcb->t_state & (TCPS_TIME_WAIT|TCPS_LAST_ACK))
1150		   :0)) {
1151	       /* Ooops, control connection already active */
1152	       sb->sb_cc = sprintf(sb->sb_wptr,"Sorry, already connected.\r\n");
1153	       sb->sb_wptr += sb->sb_cc;
1154	       return 0;
1155	     }
1156	   }
1157	   so->so_emu = EMU_CTL;
1158	   ctl_password_ok = 0;
1159	   sb->sb_cc = sprintf(sb->sb_wptr, "Slirp command-line ready (type \"help\" for help).\r\nSlirp> ");
1160	   sb->sb_wptr += sb->sb_cc;
1161	   do_echo=-1;
1162	   return(2);
1163#endif
1164	}
1165}
1166