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) {
180
ret = recv(so->s, iov[1].iov_base, iov[1].iov_len,0);
185
DEBUG_MISC((dfd, " ... read nn = %d bytes\n", nn));
191
if (sb->sb_wptr >= (sb->sb_data + sb->sb_datalen))
192
sb->sb_wptr -= sb->sb_datalen;
199
* When the socket is created, we set it SO_OOBINLINE,
200
* so when OOB data arrives, we soread() it and everything
201
* in the send buffer is sent as urgent data
207
struct tcpcb *tp = sototcpcb(so);
209
DEBUG_CALL("sorecvoob");
210
DEBUG_ARG("so = %lx", (long)so);
213
* We take a guess at how much urgent data has arrived.
214
* In most situations, when urgent data arrives, the next
215
* read() should get all the urgent data. This guess will
216
* be wrong however if more data arrives just after the
217
* urgent data, or the read() doesn't return all the
221
tp->snd_up = tp->snd_una + so->so_snd.sb_cc;
229
* There's a lot duplicated code here, but...
235
struct sbuf *sb = &so->so_rcv;
236
char buff[2048]; /* XXX Shouldn't be sending more oob data than this */
240
DEBUG_CALL("sosendoob");
241
DEBUG_ARG("so = %lx", (long)so);
242
DEBUG_ARG("sb->sb_cc = %d", sb->sb_cc);
244
if (so->so_urgc > 2048)
245
so->so_urgc = 2048; /* XXXX */
247
if (sb->sb_rptr < sb->sb_wptr) {
248
/* We can send it directly */
249
n = send(so->s, sb->sb_rptr, so->so_urgc, (MSG_OOB)); /* |MSG_DONTWAIT)); */
252
DEBUG_MISC((dfd, " --- sent %d bytes urgent data, %d urgent bytes left\n", n, so->so_urgc));
255
* Since there's no sendv or sendtov like writev,
256
* we must copy all data to a linear buffer then
259
len = (sb->sb_data + sb->sb_datalen) - sb->sb_rptr;
260
if (len > so->so_urgc) len = so->so_urgc;
261
memcpy(buff, sb->sb_rptr, len);
264
n = sb->sb_wptr - sb->sb_data;
265
if (n > so->so_urgc) n = so->so_urgc;
266
memcpy((buff + len), sb->sb_data, n);
270
n = send(so->s, buff, len, (MSG_OOB)); /* |MSG_DONTWAIT)); */
273
DEBUG_ERROR((dfd, "Didn't send all data urgently XXXXX\n"));
275
DEBUG_MISC((dfd, " ---2 sent %d bytes urgent data, %d urgent bytes left\n", n, so->so_urgc));
280
if (sb->sb_rptr >= (sb->sb_data + sb->sb_datalen))
281
sb->sb_rptr -= sb->sb_datalen;
287
* Write data from so_rcv to so's socket,
288
* updating all sbuf field as necessary
295
struct sbuf *sb = &so->so_rcv;
299
DEBUG_CALL("sowrite");
300
DEBUG_ARG("so = %lx", (long)so);
309
* No need to check if there's something to write,
310
* sowrite wouldn't have been called otherwise
315
iov[0].iov_base = sb->sb_rptr;
316
if (sb->sb_rptr < sb->sb_wptr) {
317
iov[0].iov_len = sb->sb_wptr - sb->sb_rptr;
318
/* Should never succeed, but... */
319
if (iov[0].iov_len > len) iov[0].iov_len = len;
322
iov[0].iov_len = (sb->sb_data + sb->sb_datalen) - sb->sb_rptr;
323
if (iov[0].iov_len > len) iov[0].iov_len = len;
324
len -= iov[0].iov_len;
326
iov[1].iov_base = sb->sb_data;
327
iov[1].iov_len = sb->sb_wptr - sb->sb_data;
328
if (iov[1].iov_len > len) iov[1].iov_len = len;
333
/* Check if there's urgent data to send, and if so, send it */
336
nn = writev(so->s, (const struct iovec *)iov, n);
338
DEBUG_MISC((dfd, " ... wrote nn = %d bytes\n", nn));
340
nn = send(so->s, iov[0].iov_base, iov[0].iov_len,0);
342
/* This should never happen, but people tell me it does *shrug* */
343
if (nn < 0 && (errno == EAGAIN || errno == EINTR))
347
DEBUG_MISC((dfd, " --- sowrite disconnected, so->so_state = %x, errno = %d\n",
348
so->so_state, errno));
350
tcp_sockclosed(sototcpcb(so));
355
if (n == 2 && nn == iov[0].iov_len) {
357
ret = send(so->s, iov[1].iov_base, iov[1].iov_len,0);
361
DEBUG_MISC((dfd, " ... wrote nn = %d bytes\n", nn));
367
if (sb->sb_rptr >= (sb->sb_data + sb->sb_datalen))
368
sb->sb_rptr -= sb->sb_datalen;
371
* If in DRAIN mode, and there's no more data, set
374
if ((so->so_state & SS_FWDRAIN) && sb->sb_cc == 0)
381
* recvfrom() a UDP socket
387
struct sockaddr_in addr;
388
socklen_t addrlen = sizeof(struct sockaddr_in);
390
DEBUG_CALL("sorecvfrom");
391
DEBUG_ARG("so = %lx", (long)so);
393
if (so->so_type == IPPROTO_ICMP) { /* This is a "ping" reply */
397
len = recvfrom(so->s, buff, 256, 0,
398
(struct sockaddr *)&addr, &addrlen);
399
/* XXX Check if reply is "correct"? */
401
if(len == -1 || len == 0) {
402
u_char code=ICMP_UNREACH_PORT;
404
if(errno == EHOSTUNREACH) code=ICMP_UNREACH_HOST;
405
else if(errno == ENETUNREACH) code=ICMP_UNREACH_NET;
407
DEBUG_MISC((dfd," udp icmp rx errno = %d-%s\n",
408
errno,strerror(errno)));
409
icmp_error(so->so_m, ICMP_UNREACH,code, 0,strerror(errno));
411
icmp_reflect(so->so_m);
412
so->so_m = 0; /* Don't m_free() it again! */
414
/* No need for this socket anymore, udp_detach it */
416
} else { /* A "normal" UDP packet */
420
if (!(m = m_get())) return;
421
m->m_data += if_maxlinkhdr;
424
* XXX Shouldn't FIONREAD packets destined for port 53,
425
* but I don't know the max packet size for DNS lookups
428
/* if (so->so_fport != htons(53)) { */
429
ioctlsocket(so->s, FIONREAD, &n);
432
n = (m->m_data - m->m_dat) + m->m_len + n + 1;
438
m->m_len = recvfrom(so->s, m->m_data, len, 0,
439
(struct sockaddr *)&addr, &addrlen);
440
DEBUG_MISC((dfd, " did recvfrom %d, errno = %d-%s\n",
441
m->m_len, errno,strerror(errno)));
443
u_char code=ICMP_UNREACH_PORT;
445
if(errno == EHOSTUNREACH) code=ICMP_UNREACH_HOST;
446
else if(errno == ENETUNREACH) code=ICMP_UNREACH_NET;
448
DEBUG_MISC((dfd," rx error, tx icmp ICMP_UNREACH:%i\n", code));
449
icmp_error(so->so_m, ICMP_UNREACH,code, 0,strerror(errno));
453
* Hack: domain name lookup will be used the most for UDP,
454
* and since they'll only be used once there's no need
455
* for the 4 minute (or whatever) timeout... So we time them
456
* out much quicker (10 seconds for now...)
459
if (so->so_fport == htons(53))
460
so->so_expire = curtime + SO_EXPIREFAST;
462
so->so_expire = curtime + SO_EXPIRE;
465
/* if (m->m_len == len) {
466
* m_inc(m, MINCSIZE);
472
* If this packet was destined for CTL_ADDR,
473
* make it look like that's where it came from, done by udp_output
475
udp_output(so, m, &addr);
477
} /* if ping packet */
489
struct sockaddr_in addr;
491
DEBUG_CALL("sosendto");
492
DEBUG_ARG("so = %lx", (long)so);
493
DEBUG_ARG("m = %lx", (long)m);
495
addr.sin_family = AF_INET;
496
if ((so->so_faddr.s_addr & htonl(0xffffff00)) == special_addr.s_addr) {
498
switch(ntohl(so->so_faddr.s_addr) & 0xff) {
500
addr.sin_addr = dns_addr;
504
addr.sin_addr = loopback_addr;
508
addr.sin_addr = so->so_faddr;
509
addr.sin_port = so->so_fport;
511
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)));
513
/* Don't care what port we get */
514
ret = sendto(so->s, m->m_data, m->m_len, 0,
515
(struct sockaddr *)&addr, sizeof (struct sockaddr));
520
* Kill the socket if there's no reply in 4 minutes,
521
* but only if it's an expirable socket
524
so->so_expire = curtime + SO_EXPIRE;
525
so->so_state = SS_ISFCONNECTED; /* So that it gets select()ed */
530
* XXX This should really be tcp_listen
533
solisten(port, laddr, lport, flags)
539
struct sockaddr_in addr;
542
socklen_t addrlen = sizeof(addr);
545
DEBUG_CALL("solisten");
546
DEBUG_ARG("port = %d", port);
547
DEBUG_ARG("laddr = %x", laddr);
548
DEBUG_ARG("lport = %d", lport);
549
DEBUG_ARG("flags = %x", flags);
551
if ((so = socreate()) == NULL) {
552
/* free(so); Not sofree() ??? free(NULL) == NOP */
556
/* Don't tcp_attach... we don't need so_snd nor so_rcv */
557
if ((so->so_tcpcb = tcp_newtcpcb(so)) == NULL) {
564
* SS_FACCEPTONCE sockets must time out.
566
if (flags & SS_FACCEPTONCE)
567
so->so_tcpcb->t_timer[TCPT_KEEP] = TCPTV_KEEP_INIT*2;
569
so->so_state = (SS_FACCEPTCONN|flags);
570
so->so_lport = lport; /* Kept in network format */
571
so->so_laddr.s_addr = laddr; /* Ditto */
573
addr.sin_family = AF_INET;
574
addr.sin_addr.s_addr = INADDR_ANY;
575
addr.sin_port = port;
577
if (((s = socket(AF_INET,SOCK_STREAM,0)) < 0) ||
578
(bind(s,(struct sockaddr *)&addr, sizeof(addr)) < 0) ||
580
int tmperrno = errno; /* Don't clobber the real reason we failed */
584
/* Restore the real errno */
586
WSASetLastError(tmperrno);
592
setsockopt(s,SOL_SOCKET,SO_REUSEADDR,(char *)&opt,sizeof(int));
593
setsockopt(s,SOL_SOCKET,SO_OOBINLINE,(char *)&opt,sizeof(int));
595
getsockname(s,(struct sockaddr *)&addr,&addrlen);
596
so->so_fport = addr.sin_port;
597
if (addr.sin_addr.s_addr == 0 || addr.sin_addr.s_addr == loopback_addr.s_addr)
598
so->so_faddr = our_addr;
600
so->so_faddr = addr.sin_addr;
607
* Data is available in so_rcv
608
* Just write() the data to the socket
616
/* FD_CLR(so->s,&writefds); */
620
* Data has been freed in so_snd
621
* We have room for a read() if we want to
622
* For now, don't read, it'll be done in the main loop
632
* Various session state calls
633
* XXX Should be #define's
634
* The socket state stuff needs work, these often get call 2 or 3
635
* times each when only 1 was needed
639
register struct socket *so;
641
so->so_state &= ~(SS_NOFDREF|SS_ISFCONNECTED|SS_FCANTRCVMORE|
642
SS_FCANTSENDMORE|SS_FWDRAIN);
643
so->so_state |= SS_ISFCONNECTING; /* Clobber other states */
648
register struct socket *so;
650
so->so_state &= ~(SS_ISFCONNECTING|SS_FWDRAIN|SS_NOFDREF);
651
so->so_state |= SS_ISFCONNECTED; /* Clobber other states */
658
if ((so->so_state & SS_NOFDREF) == 0) {
660
if(global_writefds) {
661
FD_CLR(so->s,global_writefds);
664
so->so_state &= ~(SS_ISFCONNECTING);
665
if (so->so_state & SS_FCANTSENDMORE)
666
so->so_state = SS_NOFDREF; /* Don't select it */ /* XXX close() here as well? */
668
so->so_state |= SS_FCANTRCVMORE;
675
if ((so->so_state & SS_NOFDREF) == 0) {
676
shutdown(so->s,1); /* send FIN to fhost */
677
if (global_readfds) {
678
FD_CLR(so->s,global_readfds);
681
FD_CLR(so->s,global_xfds);
684
so->so_state &= ~(SS_ISFCONNECTING);
685
if (so->so_state & SS_FCANTRCVMORE)
686
so->so_state = SS_NOFDREF; /* as above */
688
so->so_state |= SS_FCANTSENDMORE;
692
soisfdisconnected(so)
695
/* so->so_state &= ~(SS_ISFCONNECTING|SS_ISFCONNECTED); */
697
/* so->so_state = SS_ISFDISCONNECTED; */
699
* XXX Do nothing ... ?
704
* Set write drain mode
705
* Set CANTSENDMORE once all data has been write()n
711
if (so->so_rcv.sb_cc)
712
so->so_state |= SS_FWDRAIN;