2 * Copyright (c) 1995 Danny Gasparovski.
4 * Please read the file COPYRIGHT for the
5 * terms and conditions of the copyright.
8 #define WANT_SYS_IOCTL_H
21 solookup(head, laddr, lport, faddr, fport)
30 for (so = head->so_next; so != head; so = so->so_next) {
31 if (so->so_lport == lport &&
32 so->so_laddr.s_addr == laddr.s_addr &&
33 so->so_faddr.s_addr == faddr.s_addr &&
34 so->so_fport == fport)
39 return (struct socket *)NULL;
45 * Create a new socket, initialise the fields
46 * It is the responsibility of the caller to
47 * insque() it into the correct linked-list
54 so = (struct socket *)malloc(sizeof(struct socket));
56 memset(so, 0, sizeof(struct socket));
57 so->so_state = SS_NOFDREF;
64 * remque and free a socket, clobber cache
70 if (so->so_emu==EMU_RSH && so->extra) {
74 if (so == tcp_last_so)
76 else if (so == udp_last_so)
81 if(so->so_next && so->so_prev)
82 remque(so); /* crashes if so is not in a queue */
88 * Read from so's socket into sb_snd, updating all relevant sbuf fields
89 * NOTE: This will only be called if it is select()ed for reading, so
90 * a read() of 0 (or less) means it's disconnected
96 int n, nn, lss, total;
97 struct sbuf *sb = &so->so_snd;
98 int len = sb->sb_datalen - sb->sb_cc;
100 int mss = so->so_tcpcb->t_maxseg;
102 DEBUG_CALL("soread");
103 DEBUG_ARG("so = %lx", (long )so);
106 * No need to check if there's enough room to read.
107 * soread wouldn't have been called if there weren't
110 len = sb->sb_datalen - sb->sb_cc;
112 iov[0].iov_base = sb->sb_wptr;
113 if (sb->sb_wptr < sb->sb_rptr) {
114 iov[0].iov_len = sb->sb_rptr - sb->sb_wptr;
115 /* Should never succeed, but... */
116 if (iov[0].iov_len > len)
117 iov[0].iov_len = len;
118 if (iov[0].iov_len > mss)
119 iov[0].iov_len -= iov[0].iov_len%mss;
122 iov[0].iov_len = (sb->sb_data + sb->sb_datalen) - sb->sb_wptr;
123 /* Should never succeed, but... */
124 if (iov[0].iov_len > len) iov[0].iov_len = len;
125 len -= iov[0].iov_len;
127 iov[1].iov_base = sb->sb_data;
128 iov[1].iov_len = sb->sb_rptr - sb->sb_data;
129 if(iov[1].iov_len > len)
130 iov[1].iov_len = len;
131 total = iov[0].iov_len + iov[1].iov_len;
134 if (iov[1].iov_len > lss) {
135 iov[1].iov_len -= lss;
138 lss -= iov[1].iov_len;
139 iov[0].iov_len -= lss;
145 if (iov[0].iov_len > mss)
146 iov[0].iov_len -= iov[0].iov_len%mss;
152 nn = readv(so->s, (struct iovec *)iov, n);
153 DEBUG_MISC((dfd, " ... read nn = %d bytes\n", nn));
155 nn = recv(so->s, iov[0].iov_base, iov[0].iov_len,0);
158 if (nn < 0 && (errno == EINTR || errno == EAGAIN))
161 DEBUG_MISC((dfd, " --- soread() disconnected, nn = %d, errno = %d-%s\n", nn, errno,strerror(errno)));
163 tcp_sockclosed(sototcpcb(so));
170 * If there was no error, try and read the second time round
171 * We read again if n = 2 (ie, there's another part of the buffer)
172 * and we read as much as we could in the first read
173 * We don't test for <= 0 this time, because there legitimately
174 * might not be any more data (since the socket is non-blocking),
175 * a close will be detected on next iteration.
176 * A return of -1 wont (shouldn't) happen, since it didn't happen above
178 if (n == 2 && nn == iov[0].iov_len)
179 nn += recv(so->s, iov[1].iov_base, iov[1].iov_len,0);
181 DEBUG_MISC((dfd, " ... read nn = %d bytes\n", nn));
187 if (sb->sb_wptr >= (sb->sb_data + sb->sb_datalen))
188 sb->sb_wptr -= sb->sb_datalen;
195 * When the socket is created, we set it SO_OOBINLINE,
196 * so when OOB data arrives, we soread() it and everything
197 * in the send buffer is sent as urgent data
203 struct tcpcb *tp = sototcpcb(so);
205 DEBUG_CALL("sorecvoob");
206 DEBUG_ARG("so = %lx", (long)so);
209 * We take a guess at how much urgent data has arrived.
210 * In most situations, when urgent data arrives, the next
211 * read() should get all the urgent data. This guess will
212 * be wrong however if more data arrives just after the
213 * urgent data, or the read() doesn't return all the
217 tp->snd_up = tp->snd_una + so->so_snd.sb_cc;
225 * There's a lot duplicated code here, but...
231 struct sbuf *sb = &so->so_rcv;
232 char buff[2048]; /* XXX Shouldn't be sending more oob data than this */
236 DEBUG_CALL("sosendoob");
237 DEBUG_ARG("so = %lx", (long)so);
238 DEBUG_ARG("sb->sb_cc = %d", sb->sb_cc);
240 if (so->so_urgc > 2048)
241 so->so_urgc = 2048; /* XXXX */
243 if (sb->sb_rptr < sb->sb_wptr) {
244 /* We can send it directly */
245 n = send(so->s, sb->sb_rptr, so->so_urgc, (MSG_OOB)); /* |MSG_DONTWAIT)); */
248 DEBUG_MISC((dfd, " --- sent %d bytes urgent data, %d urgent bytes left\n", n, so->so_urgc));
251 * Since there's no sendv or sendtov like writev,
252 * we must copy all data to a linear buffer then
255 len = (sb->sb_data + sb->sb_datalen) - sb->sb_rptr;
256 if (len > so->so_urgc) len = so->so_urgc;
257 memcpy(buff, sb->sb_rptr, len);
260 n = sb->sb_wptr - sb->sb_data;
261 if (n > so->so_urgc) n = so->so_urgc;
262 memcpy((buff + len), sb->sb_data, n);
266 n = send(so->s, buff, len, (MSG_OOB)); /* |MSG_DONTWAIT)); */
269 DEBUG_ERROR((dfd, "Didn't send all data urgently XXXXX\n"));
271 DEBUG_MISC((dfd, " ---2 sent %d bytes urgent data, %d urgent bytes left\n", n, so->so_urgc));
276 if (sb->sb_rptr >= (sb->sb_data + sb->sb_datalen))
277 sb->sb_rptr -= sb->sb_datalen;
283 * Write data from so_rcv to so's socket,
284 * updating all sbuf field as necessary
291 struct sbuf *sb = &so->so_rcv;
295 DEBUG_CALL("sowrite");
296 DEBUG_ARG("so = %lx", (long)so);
305 * No need to check if there's something to write,
306 * sowrite wouldn't have been called otherwise
311 iov[0].iov_base = sb->sb_rptr;
312 if (sb->sb_rptr < sb->sb_wptr) {
313 iov[0].iov_len = sb->sb_wptr - sb->sb_rptr;
314 /* Should never succeed, but... */
315 if (iov[0].iov_len > len) iov[0].iov_len = len;
318 iov[0].iov_len = (sb->sb_data + sb->sb_datalen) - sb->sb_rptr;
319 if (iov[0].iov_len > len) iov[0].iov_len = len;
320 len -= iov[0].iov_len;
322 iov[1].iov_base = sb->sb_data;
323 iov[1].iov_len = sb->sb_wptr - sb->sb_data;
324 if (iov[1].iov_len > len) iov[1].iov_len = len;
329 /* Check if there's urgent data to send, and if so, send it */
332 nn = writev(so->s, (const struct iovec *)iov, n);
334 DEBUG_MISC((dfd, " ... wrote nn = %d bytes\n", nn));
336 nn = send(so->s, iov[0].iov_base, iov[0].iov_len,0);
338 /* This should never happen, but people tell me it does *shrug* */
339 if (nn < 0 && (errno == EAGAIN || errno == EINTR))
343 DEBUG_MISC((dfd, " --- sowrite disconnected, so->so_state = %x, errno = %d\n",
344 so->so_state, errno));
346 tcp_sockclosed(sototcpcb(so));
351 if (n == 2 && nn == iov[0].iov_len)
352 nn += send(so->s, iov[1].iov_base, iov[1].iov_len,0);
353 DEBUG_MISC((dfd, " ... wrote nn = %d bytes\n", nn));
359 if (sb->sb_rptr >= (sb->sb_data + sb->sb_datalen))
360 sb->sb_rptr -= sb->sb_datalen;
363 * If in DRAIN mode, and there's no more data, set
366 if ((so->so_state & SS_FWDRAIN) && sb->sb_cc == 0)
373 * recvfrom() a UDP socket
379 struct sockaddr_in addr;
380 int addrlen = sizeof(struct sockaddr_in);
382 DEBUG_CALL("sorecvfrom");
383 DEBUG_ARG("so = %lx", (long)so);
385 if (so->so_type == IPPROTO_ICMP) { /* This is a "ping" reply */
389 len = recvfrom(so->s, buff, 256, 0,
390 (struct sockaddr *)&addr, &addrlen);
391 /* XXX Check if reply is "correct"? */
393 if(len == -1 || len == 0) {
394 u_char code=ICMP_UNREACH_PORT;
396 if(errno == EHOSTUNREACH) code=ICMP_UNREACH_HOST;
397 else if(errno == ENETUNREACH) code=ICMP_UNREACH_NET;
399 DEBUG_MISC((dfd," udp icmp rx errno = %d-%s\n",
400 errno,strerror(errno)));
401 icmp_error(so->so_m, ICMP_UNREACH,code, 0,strerror(errno));
403 icmp_reflect(so->so_m);
404 so->so_m = 0; /* Don't m_free() it again! */
406 /* No need for this socket anymore, udp_detach it */
408 } else { /* A "normal" UDP packet */
412 if (!(m = m_get())) return;
413 m->m_data += if_maxlinkhdr;
416 * XXX Shouldn't FIONREAD packets destined for port 53,
417 * but I don't know the max packet size for DNS lookups
420 /* if (so->so_fport != htons(53)) { */
421 ioctlsocket(so->s, FIONREAD, &n);
424 n = (m->m_data - m->m_dat) + m->m_len + n + 1;
430 m->m_len = recvfrom(so->s, m->m_data, len, 0,
431 (struct sockaddr *)&addr, &addrlen);
432 DEBUG_MISC((dfd, " did recvfrom %d, errno = %d-%s\n",
433 m->m_len, errno,strerror(errno)));
435 u_char code=ICMP_UNREACH_PORT;
437 if(errno == EHOSTUNREACH) code=ICMP_UNREACH_HOST;
438 else if(errno == ENETUNREACH) code=ICMP_UNREACH_NET;
440 DEBUG_MISC((dfd," rx error, tx icmp ICMP_UNREACH:%i\n", code));
441 icmp_error(so->so_m, ICMP_UNREACH,code, 0,strerror(errno));
445 * Hack: domain name lookup will be used the most for UDP,
446 * and since they'll only be used once there's no need
447 * for the 4 minute (or whatever) timeout... So we time them
448 * out much quicker (10 seconds for now...)
451 if (so->so_fport == htons(53))
452 so->so_expire = curtime + SO_EXPIREFAST;
454 so->so_expire = curtime + SO_EXPIRE;
457 /* if (m->m_len == len) {
458 * m_inc(m, MINCSIZE);
464 * If this packet was destined for CTL_ADDR,
465 * make it look like that's where it came from, done by udp_output
467 udp_output(so, m, &addr);
469 } /* if ping packet */
481 struct sockaddr_in addr;
483 DEBUG_CALL("sosendto");
484 DEBUG_ARG("so = %lx", (long)so);
485 DEBUG_ARG("m = %lx", (long)m);
487 addr.sin_family = AF_INET;
488 if ((so->so_faddr.s_addr & htonl(0xffffff00)) == special_addr.s_addr) {
490 switch(ntohl(so->so_faddr.s_addr) & 0xff) {
492 addr.sin_addr = dns_addr;
496 addr.sin_addr = loopback_addr;
500 addr.sin_addr = so->so_faddr;
501 addr.sin_port = so->so_fport;
503 DEBUG_MISC((dfd, " sendto()ing, addr.sin_port=%d, addr.sin_addr.s_addr=%.16s\n", ntohs(addr.sin_port), inet_ntoa(addr.sin_addr)));
505 /* Don't care what port we get */
506 ret = sendto(so->s, m->m_data, m->m_len, 0,
507 (struct sockaddr *)&addr, sizeof (struct sockaddr));
512 * Kill the socket if there's no reply in 4 minutes,
513 * but only if it's an expirable socket
516 so->so_expire = curtime + SO_EXPIRE;
517 so->so_state = SS_ISFCONNECTED; /* So that it gets select()ed */
522 * XXX This should really be tcp_listen
525 solisten(port, laddr, lport, flags)
531 struct sockaddr_in addr;
533 int s, addrlen = sizeof(addr), opt = 1;
535 DEBUG_CALL("solisten");
536 DEBUG_ARG("port = %d", port);
537 DEBUG_ARG("laddr = %x", laddr);
538 DEBUG_ARG("lport = %d", lport);
539 DEBUG_ARG("flags = %x", flags);
541 if ((so = socreate()) == NULL) {
542 /* free(so); Not sofree() ??? free(NULL) == NOP */
546 /* Don't tcp_attach... we don't need so_snd nor so_rcv */
547 if ((so->so_tcpcb = tcp_newtcpcb(so)) == NULL) {
554 * SS_FACCEPTONCE sockets must time out.
556 if (flags & SS_FACCEPTONCE)
557 so->so_tcpcb->t_timer[TCPT_KEEP] = TCPTV_KEEP_INIT*2;
559 so->so_state = (SS_FACCEPTCONN|flags);
560 so->so_lport = lport; /* Kept in network format */
561 so->so_laddr.s_addr = laddr; /* Ditto */
563 addr.sin_family = AF_INET;
564 addr.sin_addr.s_addr = INADDR_ANY;
565 addr.sin_port = port;
567 if (((s = socket(AF_INET,SOCK_STREAM,0)) < 0) ||
568 (bind(s,(struct sockaddr *)&addr, sizeof(addr)) < 0) ||
570 int tmperrno = errno; /* Don't clobber the real reason we failed */
574 /* Restore the real errno */
576 WSASetLastError(tmperrno);
582 setsockopt(s,SOL_SOCKET,SO_REUSEADDR,(char *)&opt,sizeof(int));
583 setsockopt(s,SOL_SOCKET,SO_OOBINLINE,(char *)&opt,sizeof(int));
585 getsockname(s,(struct sockaddr *)&addr,&addrlen);
586 so->so_fport = addr.sin_port;
587 if (addr.sin_addr.s_addr == 0 || addr.sin_addr.s_addr == loopback_addr.s_addr)
588 so->so_faddr = our_addr;
590 so->so_faddr = addr.sin_addr;
597 * Data is available in so_rcv
598 * Just write() the data to the socket
606 /* FD_CLR(so->s,&writefds); */
610 * Data has been freed in so_snd
611 * We have room for a read() if we want to
612 * For now, don't read, it'll be done in the main loop
622 * Various session state calls
623 * XXX Should be #define's
624 * The socket state stuff needs work, these often get call 2 or 3
625 * times each when only 1 was needed
629 register struct socket *so;
631 so->so_state &= ~(SS_NOFDREF|SS_ISFCONNECTED|SS_FCANTRCVMORE|
632 SS_FCANTSENDMORE|SS_FWDRAIN);
633 so->so_state |= SS_ISFCONNECTING; /* Clobber other states */
638 register struct socket *so;
640 so->so_state &= ~(SS_ISFCONNECTING|SS_FWDRAIN|SS_NOFDREF);
641 so->so_state |= SS_ISFCONNECTED; /* Clobber other states */
648 if ((so->so_state & SS_NOFDREF) == 0) {
650 if(global_writefds) {
651 FD_CLR(so->s,global_writefds);
654 so->so_state &= ~(SS_ISFCONNECTING);
655 if (so->so_state & SS_FCANTSENDMORE)
656 so->so_state = SS_NOFDREF; /* Don't select it */ /* XXX close() here as well? */
658 so->so_state |= SS_FCANTRCVMORE;
665 if ((so->so_state & SS_NOFDREF) == 0) {
666 shutdown(so->s,1); /* send FIN to fhost */
667 if (global_readfds) {
668 FD_CLR(so->s,global_readfds);
671 FD_CLR(so->s,global_xfds);
674 so->so_state &= ~(SS_ISFCONNECTING);
675 if (so->so_state & SS_FCANTRCVMORE)
676 so->so_state = SS_NOFDREF; /* as above */
678 so->so_state |= SS_FCANTSENDMORE;
682 soisfdisconnected(so)
685 /* so->so_state &= ~(SS_ISFCONNECTING|SS_ISFCONNECTED); */
687 /* so->so_state = SS_ISFDISCONNECTED; */
689 * XXX Do nothing ... ?
694 * Set write drain mode
695 * Set CANTSENDMORE once all data has been write()n
701 if (so->so_rcv.sb_cc)
702 so->so_state |= SS_FWDRAIN;