virtio-net: Add new RX filter controls
[qemu-kvm/markmc.git] / slirp / socket.c
blob098132a1bb9f6958cd8f96575fbdb0493bffce0f
1 /*
2 * Copyright (c) 1995 Danny Gasparovski.
4 * Please read the file COPYRIGHT for the
5 * terms and conditions of the copyright.
6 */
8 #include "qemu-common.h"
9 #include <slirp.h>
10 #include "ip_icmp.h"
11 #ifdef __sun__
12 #include <sys/filio.h>
13 #endif
15 static void sofcantrcvmore(struct socket *so);
16 static void sofcantsendmore(struct socket *so);
18 #if 0
19 static void
20 so_init()
22 /* Nothing yet */
24 #endif
26 struct socket *
27 solookup(struct socket *head, struct in_addr laddr, u_int lport,
28 struct in_addr faddr, u_int fport)
30 struct socket *so;
32 for (so = head->so_next; so != head; so = so->so_next) {
33 if (so->so_lport == lport &&
34 so->so_laddr.s_addr == laddr.s_addr &&
35 so->so_faddr.s_addr == faddr.s_addr &&
36 so->so_fport == fport)
37 break;
40 if (so == head)
41 return (struct socket *)NULL;
42 return so;
47 * Create a new socket, initialise the fields
48 * It is the responsibility of the caller to
49 * insque() it into the correct linked-list
51 struct socket *
52 socreate(void)
54 struct socket *so;
56 so = (struct socket *)malloc(sizeof(struct socket));
57 if(so) {
58 memset(so, 0, sizeof(struct socket));
59 so->so_state = SS_NOFDREF;
60 so->s = -1;
62 return(so);
66 * remque and free a socket, clobber cache
68 void
69 sofree(struct socket *so)
71 if (so->so_emu==EMU_RSH && so->extra) {
72 sofree(so->extra);
73 so->extra=NULL;
75 if (so == tcp_last_so)
76 tcp_last_so = &tcb;
77 else if (so == udp_last_so)
78 udp_last_so = &udb;
80 m_free(so->so_m);
82 if(so->so_next && so->so_prev)
83 remque(so); /* crashes if so is not in a queue */
85 free(so);
88 size_t sopreprbuf(struct socket *so, struct iovec *iov, int *np)
90 int n, lss, total;
91 struct sbuf *sb = &so->so_snd;
92 int len = sb->sb_datalen - sb->sb_cc;
93 int mss = so->so_tcpcb->t_maxseg;
95 DEBUG_CALL("sopreprbuf");
96 DEBUG_ARG("so = %lx", (long )so);
98 len = sb->sb_datalen - sb->sb_cc;
100 if (len <= 0)
101 return 0;
103 iov[0].iov_base = sb->sb_wptr;
104 iov[1].iov_base = NULL;
105 iov[1].iov_len = 0;
106 if (sb->sb_wptr < sb->sb_rptr) {
107 iov[0].iov_len = sb->sb_rptr - sb->sb_wptr;
108 /* Should never succeed, but... */
109 if (iov[0].iov_len > len)
110 iov[0].iov_len = len;
111 if (iov[0].iov_len > mss)
112 iov[0].iov_len -= iov[0].iov_len%mss;
113 n = 1;
114 } else {
115 iov[0].iov_len = (sb->sb_data + sb->sb_datalen) - sb->sb_wptr;
116 /* Should never succeed, but... */
117 if (iov[0].iov_len > len) iov[0].iov_len = len;
118 len -= iov[0].iov_len;
119 if (len) {
120 iov[1].iov_base = sb->sb_data;
121 iov[1].iov_len = sb->sb_rptr - sb->sb_data;
122 if(iov[1].iov_len > len)
123 iov[1].iov_len = len;
124 total = iov[0].iov_len + iov[1].iov_len;
125 if (total > mss) {
126 lss = total%mss;
127 if (iov[1].iov_len > lss) {
128 iov[1].iov_len -= lss;
129 n = 2;
130 } else {
131 lss -= iov[1].iov_len;
132 iov[0].iov_len -= lss;
133 n = 1;
135 } else
136 n = 2;
137 } else {
138 if (iov[0].iov_len > mss)
139 iov[0].iov_len -= iov[0].iov_len%mss;
140 n = 1;
143 if (np)
144 *np = n;
146 return iov[0].iov_len + (n - 1) * iov[1].iov_len;
150 * Read from so's socket into sb_snd, updating all relevant sbuf fields
151 * NOTE: This will only be called if it is select()ed for reading, so
152 * a read() of 0 (or less) means it's disconnected
155 soread(struct socket *so)
157 int n, nn;
158 struct sbuf *sb = &so->so_snd;
159 struct iovec iov[2];
161 DEBUG_CALL("soread");
162 DEBUG_ARG("so = %lx", (long )so);
165 * No need to check if there's enough room to read.
166 * soread wouldn't have been called if there weren't
168 sopreprbuf(so, iov, &n);
170 #ifdef HAVE_READV
171 nn = readv(so->s, (struct iovec *)iov, n);
172 DEBUG_MISC((dfd, " ... read nn = %d bytes\n", nn));
173 #else
174 nn = recv(so->s, iov[0].iov_base, iov[0].iov_len,0);
175 #endif
176 if (nn <= 0) {
177 if (nn < 0 && (errno == EINTR || errno == EAGAIN))
178 return 0;
179 else {
180 DEBUG_MISC((dfd, " --- soread() disconnected, nn = %d, errno = %d-%s\n", nn, errno,strerror(errno)));
181 sofcantrcvmore(so);
182 tcp_sockclosed(sototcpcb(so));
183 return -1;
187 #ifndef HAVE_READV
189 * If there was no error, try and read the second time round
190 * We read again if n = 2 (ie, there's another part of the buffer)
191 * and we read as much as we could in the first read
192 * We don't test for <= 0 this time, because there legitimately
193 * might not be any more data (since the socket is non-blocking),
194 * a close will be detected on next iteration.
195 * A return of -1 wont (shouldn't) happen, since it didn't happen above
197 if (n == 2 && nn == iov[0].iov_len) {
198 int ret;
199 ret = recv(so->s, iov[1].iov_base, iov[1].iov_len,0);
200 if (ret > 0)
201 nn += ret;
204 DEBUG_MISC((dfd, " ... read nn = %d bytes\n", nn));
205 #endif
207 /* Update fields */
208 sb->sb_cc += nn;
209 sb->sb_wptr += nn;
210 if (sb->sb_wptr >= (sb->sb_data + sb->sb_datalen))
211 sb->sb_wptr -= sb->sb_datalen;
212 return nn;
215 int soreadbuf(struct socket *so, const char *buf, int size)
217 int n, nn, copy = size;
218 struct sbuf *sb = &so->so_snd;
219 struct iovec iov[2];
221 DEBUG_CALL("soreadbuf");
222 DEBUG_ARG("so = %lx", (long )so);
225 * No need to check if there's enough room to read.
226 * soread wouldn't have been called if there weren't
228 if (sopreprbuf(so, iov, &n) < size)
229 goto err;
231 nn = MIN(iov[0].iov_len, copy);
232 memcpy(iov[0].iov_base, buf, nn);
234 copy -= nn;
235 buf += nn;
237 if (copy == 0)
238 goto done;
240 memcpy(iov[1].iov_base, buf, copy);
242 done:
243 /* Update fields */
244 sb->sb_cc += size;
245 sb->sb_wptr += size;
246 if (sb->sb_wptr >= (sb->sb_data + sb->sb_datalen))
247 sb->sb_wptr -= sb->sb_datalen;
248 return size;
249 err:
251 sofcantrcvmore(so);
252 tcp_sockclosed(sototcpcb(so));
253 fprintf(stderr, "soreadbuf buffer to small");
254 return -1;
258 * Get urgent data
260 * When the socket is created, we set it SO_OOBINLINE,
261 * so when OOB data arrives, we soread() it and everything
262 * in the send buffer is sent as urgent data
264 void
265 sorecvoob(struct socket *so)
267 struct tcpcb *tp = sototcpcb(so);
269 DEBUG_CALL("sorecvoob");
270 DEBUG_ARG("so = %lx", (long)so);
273 * We take a guess at how much urgent data has arrived.
274 * In most situations, when urgent data arrives, the next
275 * read() should get all the urgent data. This guess will
276 * be wrong however if more data arrives just after the
277 * urgent data, or the read() doesn't return all the
278 * urgent data.
280 soread(so);
281 tp->snd_up = tp->snd_una + so->so_snd.sb_cc;
282 tp->t_force = 1;
283 tcp_output(tp);
284 tp->t_force = 0;
288 * Send urgent data
289 * There's a lot duplicated code here, but...
292 sosendoob(struct socket *so)
294 struct sbuf *sb = &so->so_rcv;
295 char buff[2048]; /* XXX Shouldn't be sending more oob data than this */
297 int n, len;
299 DEBUG_CALL("sosendoob");
300 DEBUG_ARG("so = %lx", (long)so);
301 DEBUG_ARG("sb->sb_cc = %d", sb->sb_cc);
303 if (so->so_urgc > 2048)
304 so->so_urgc = 2048; /* XXXX */
306 if (sb->sb_rptr < sb->sb_wptr) {
307 /* We can send it directly */
308 n = slirp_send(so, sb->sb_rptr, so->so_urgc, (MSG_OOB)); /* |MSG_DONTWAIT)); */
309 so->so_urgc -= n;
311 DEBUG_MISC((dfd, " --- sent %d bytes urgent data, %d urgent bytes left\n", n, so->so_urgc));
312 } else {
314 * Since there's no sendv or sendtov like writev,
315 * we must copy all data to a linear buffer then
316 * send it all
318 len = (sb->sb_data + sb->sb_datalen) - sb->sb_rptr;
319 if (len > so->so_urgc) len = so->so_urgc;
320 memcpy(buff, sb->sb_rptr, len);
321 so->so_urgc -= len;
322 if (so->so_urgc) {
323 n = sb->sb_wptr - sb->sb_data;
324 if (n > so->so_urgc) n = so->so_urgc;
325 memcpy((buff + len), sb->sb_data, n);
326 so->so_urgc -= n;
327 len += n;
329 n = slirp_send(so, buff, len, (MSG_OOB)); /* |MSG_DONTWAIT)); */
330 #ifdef DEBUG
331 if (n != len)
332 DEBUG_ERROR((dfd, "Didn't send all data urgently XXXXX\n"));
333 #endif
334 DEBUG_MISC((dfd, " ---2 sent %d bytes urgent data, %d urgent bytes left\n", n, so->so_urgc));
337 sb->sb_cc -= n;
338 sb->sb_rptr += n;
339 if (sb->sb_rptr >= (sb->sb_data + sb->sb_datalen))
340 sb->sb_rptr -= sb->sb_datalen;
342 return n;
346 * Write data from so_rcv to so's socket,
347 * updating all sbuf field as necessary
350 sowrite(struct socket *so)
352 int n,nn;
353 struct sbuf *sb = &so->so_rcv;
354 int len = sb->sb_cc;
355 struct iovec iov[2];
357 DEBUG_CALL("sowrite");
358 DEBUG_ARG("so = %lx", (long)so);
360 if (so->so_urgc) {
361 sosendoob(so);
362 if (sb->sb_cc == 0)
363 return 0;
367 * No need to check if there's something to write,
368 * sowrite wouldn't have been called otherwise
371 len = sb->sb_cc;
373 iov[0].iov_base = sb->sb_rptr;
374 iov[1].iov_base = NULL;
375 iov[1].iov_len = 0;
376 if (sb->sb_rptr < sb->sb_wptr) {
377 iov[0].iov_len = sb->sb_wptr - sb->sb_rptr;
378 /* Should never succeed, but... */
379 if (iov[0].iov_len > len) iov[0].iov_len = len;
380 n = 1;
381 } else {
382 iov[0].iov_len = (sb->sb_data + sb->sb_datalen) - sb->sb_rptr;
383 if (iov[0].iov_len > len) iov[0].iov_len = len;
384 len -= iov[0].iov_len;
385 if (len) {
386 iov[1].iov_base = sb->sb_data;
387 iov[1].iov_len = sb->sb_wptr - sb->sb_data;
388 if (iov[1].iov_len > len) iov[1].iov_len = len;
389 n = 2;
390 } else
391 n = 1;
393 /* Check if there's urgent data to send, and if so, send it */
395 #ifdef HAVE_READV
396 nn = writev(so->s, (const struct iovec *)iov, n);
398 DEBUG_MISC((dfd, " ... wrote nn = %d bytes\n", nn));
399 #else
400 nn = slirp_send(so, iov[0].iov_base, iov[0].iov_len,0);
401 #endif
402 /* This should never happen, but people tell me it does *shrug* */
403 if (nn < 0 && (errno == EAGAIN || errno == EINTR))
404 return 0;
406 if (nn <= 0) {
407 DEBUG_MISC((dfd, " --- sowrite disconnected, so->so_state = %x, errno = %d\n",
408 so->so_state, errno));
409 sofcantsendmore(so);
410 tcp_sockclosed(sototcpcb(so));
411 return -1;
414 #ifndef HAVE_READV
415 if (n == 2 && nn == iov[0].iov_len) {
416 int ret;
417 ret = slirp_send(so, iov[1].iov_base, iov[1].iov_len,0);
418 if (ret > 0)
419 nn += ret;
421 DEBUG_MISC((dfd, " ... wrote nn = %d bytes\n", nn));
422 #endif
424 /* Update sbuf */
425 sb->sb_cc -= nn;
426 sb->sb_rptr += nn;
427 if (sb->sb_rptr >= (sb->sb_data + sb->sb_datalen))
428 sb->sb_rptr -= sb->sb_datalen;
431 * If in DRAIN mode, and there's no more data, set
432 * it CANTSENDMORE
434 if ((so->so_state & SS_FWDRAIN) && sb->sb_cc == 0)
435 sofcantsendmore(so);
437 return nn;
441 * recvfrom() a UDP socket
443 void
444 sorecvfrom(struct socket *so)
446 struct sockaddr_in addr;
447 socklen_t addrlen = sizeof(struct sockaddr_in);
449 DEBUG_CALL("sorecvfrom");
450 DEBUG_ARG("so = %lx", (long)so);
452 if (so->so_type == IPPROTO_ICMP) { /* This is a "ping" reply */
453 char buff[256];
454 int len;
456 len = recvfrom(so->s, buff, 256, 0,
457 (struct sockaddr *)&addr, &addrlen);
458 /* XXX Check if reply is "correct"? */
460 if(len == -1 || len == 0) {
461 u_char code=ICMP_UNREACH_PORT;
463 if(errno == EHOSTUNREACH) code=ICMP_UNREACH_HOST;
464 else if(errno == ENETUNREACH) code=ICMP_UNREACH_NET;
466 DEBUG_MISC((dfd," udp icmp rx errno = %d-%s\n",
467 errno,strerror(errno)));
468 icmp_error(so->so_m, ICMP_UNREACH,code, 0,strerror(errno));
469 } else {
470 icmp_reflect(so->so_m);
471 so->so_m = NULL; /* Don't m_free() it again! */
473 /* No need for this socket anymore, udp_detach it */
474 udp_detach(so);
475 } else { /* A "normal" UDP packet */
476 struct mbuf *m;
477 int len, n;
479 if (!(m = m_get())) return;
480 m->m_data += IF_MAXLINKHDR;
483 * XXX Shouldn't FIONREAD packets destined for port 53,
484 * but I don't know the max packet size for DNS lookups
486 len = M_FREEROOM(m);
487 /* if (so->so_fport != htons(53)) { */
488 ioctlsocket(so->s, FIONREAD, &n);
490 if (n > len) {
491 n = (m->m_data - m->m_dat) + m->m_len + n + 1;
492 m_inc(m, n);
493 len = M_FREEROOM(m);
495 /* } */
497 m->m_len = recvfrom(so->s, m->m_data, len, 0,
498 (struct sockaddr *)&addr, &addrlen);
499 DEBUG_MISC((dfd, " did recvfrom %d, errno = %d-%s\n",
500 m->m_len, errno,strerror(errno)));
501 if(m->m_len<0) {
502 u_char code=ICMP_UNREACH_PORT;
504 if(errno == EHOSTUNREACH) code=ICMP_UNREACH_HOST;
505 else if(errno == ENETUNREACH) code=ICMP_UNREACH_NET;
507 DEBUG_MISC((dfd," rx error, tx icmp ICMP_UNREACH:%i\n", code));
508 icmp_error(so->so_m, ICMP_UNREACH,code, 0,strerror(errno));
509 m_free(m);
510 } else {
512 * Hack: domain name lookup will be used the most for UDP,
513 * and since they'll only be used once there's no need
514 * for the 4 minute (or whatever) timeout... So we time them
515 * out much quicker (10 seconds for now...)
517 if (so->so_expire) {
518 if (so->so_fport == htons(53))
519 so->so_expire = curtime + SO_EXPIREFAST;
520 else
521 so->so_expire = curtime + SO_EXPIRE;
524 /* if (m->m_len == len) {
525 * m_inc(m, MINCSIZE);
526 * m->m_len = 0;
531 * If this packet was destined for CTL_ADDR,
532 * make it look like that's where it came from, done by udp_output
534 udp_output(so, m, &addr);
535 } /* rx error */
536 } /* if ping packet */
540 * sendto() a socket
543 sosendto(struct socket *so, struct mbuf *m)
545 int ret;
546 struct sockaddr_in addr;
548 DEBUG_CALL("sosendto");
549 DEBUG_ARG("so = %lx", (long)so);
550 DEBUG_ARG("m = %lx", (long)m);
552 addr.sin_family = AF_INET;
553 if ((so->so_faddr.s_addr & htonl(0xffffff00)) == special_addr.s_addr) {
554 /* It's an alias */
555 switch(ntohl(so->so_faddr.s_addr) & 0xff) {
556 case CTL_DNS:
557 addr.sin_addr = dns_addr;
558 break;
559 case CTL_ALIAS:
560 default:
561 addr.sin_addr = loopback_addr;
562 break;
564 } else
565 addr.sin_addr = so->so_faddr;
566 addr.sin_port = so->so_fport;
568 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)));
570 /* Don't care what port we get */
571 ret = sendto(so->s, m->m_data, m->m_len, 0,
572 (struct sockaddr *)&addr, sizeof (struct sockaddr));
573 if (ret < 0)
574 return -1;
577 * Kill the socket if there's no reply in 4 minutes,
578 * but only if it's an expirable socket
580 if (so->so_expire)
581 so->so_expire = curtime + SO_EXPIRE;
582 so->so_state = SS_ISFCONNECTED; /* So that it gets select()ed */
583 return 0;
587 * XXX This should really be tcp_listen
589 struct socket *
590 solisten(u_int port, u_int32_t laddr, u_int lport, int flags)
592 struct sockaddr_in addr;
593 struct socket *so;
594 int s, opt = 1;
595 socklen_t addrlen = sizeof(addr);
597 DEBUG_CALL("solisten");
598 DEBUG_ARG("port = %d", port);
599 DEBUG_ARG("laddr = %x", laddr);
600 DEBUG_ARG("lport = %d", lport);
601 DEBUG_ARG("flags = %x", flags);
603 if ((so = socreate()) == NULL) {
604 /* free(so); Not sofree() ??? free(NULL) == NOP */
605 return NULL;
608 /* Don't tcp_attach... we don't need so_snd nor so_rcv */
609 if ((so->so_tcpcb = tcp_newtcpcb(so)) == NULL) {
610 free(so);
611 return NULL;
613 insque(so,&tcb);
616 * SS_FACCEPTONCE sockets must time out.
618 if (flags & SS_FACCEPTONCE)
619 so->so_tcpcb->t_timer[TCPT_KEEP] = TCPTV_KEEP_INIT*2;
621 so->so_state = (SS_FACCEPTCONN|flags);
622 so->so_lport = lport; /* Kept in network format */
623 so->so_laddr.s_addr = laddr; /* Ditto */
625 addr.sin_family = AF_INET;
626 addr.sin_addr.s_addr = INADDR_ANY;
627 addr.sin_port = port;
629 if (((s = socket(AF_INET,SOCK_STREAM,0)) < 0) ||
630 (setsockopt(s,SOL_SOCKET,SO_REUSEADDR,(char *)&opt,sizeof(int)) < 0) ||
631 (bind(s,(struct sockaddr *)&addr, sizeof(addr)) < 0) ||
632 (listen(s,1) < 0)) {
633 int tmperrno = errno; /* Don't clobber the real reason we failed */
635 close(s);
636 sofree(so);
637 /* Restore the real errno */
638 #ifdef _WIN32
639 WSASetLastError(tmperrno);
640 #else
641 errno = tmperrno;
642 #endif
643 return NULL;
645 setsockopt(s,SOL_SOCKET,SO_OOBINLINE,(char *)&opt,sizeof(int));
647 getsockname(s,(struct sockaddr *)&addr,&addrlen);
648 so->so_fport = addr.sin_port;
649 if (addr.sin_addr.s_addr == 0 || addr.sin_addr.s_addr == loopback_addr.s_addr)
650 so->so_faddr = alias_addr;
651 else
652 so->so_faddr = addr.sin_addr;
654 so->s = s;
655 return so;
658 #if 0
660 * Data is available in so_rcv
661 * Just write() the data to the socket
662 * XXX not yet...
664 static void
665 sorwakeup(so)
666 struct socket *so;
668 /* sowrite(so); */
669 /* FD_CLR(so->s,&writefds); */
673 * Data has been freed in so_snd
674 * We have room for a read() if we want to
675 * For now, don't read, it'll be done in the main loop
677 static void
678 sowwakeup(so)
679 struct socket *so;
681 /* Nothing, yet */
683 #endif
686 * Various session state calls
687 * XXX Should be #define's
688 * The socket state stuff needs work, these often get call 2 or 3
689 * times each when only 1 was needed
691 void
692 soisfconnecting(struct socket *so)
694 so->so_state &= ~(SS_NOFDREF|SS_ISFCONNECTED|SS_FCANTRCVMORE|
695 SS_FCANTSENDMORE|SS_FWDRAIN);
696 so->so_state |= SS_ISFCONNECTING; /* Clobber other states */
699 void
700 soisfconnected(struct socket *so)
702 so->so_state &= ~(SS_ISFCONNECTING|SS_FWDRAIN|SS_NOFDREF);
703 so->so_state |= SS_ISFCONNECTED; /* Clobber other states */
706 static void
707 sofcantrcvmore(struct socket *so)
709 if ((so->so_state & SS_NOFDREF) == 0) {
710 shutdown(so->s,0);
711 if(global_writefds) {
712 FD_CLR(so->s,global_writefds);
715 so->so_state &= ~(SS_ISFCONNECTING);
716 if (so->so_state & SS_FCANTSENDMORE)
717 so->so_state = SS_NOFDREF; /* Don't select it */ /* XXX close() here as well? */
718 else
719 so->so_state |= SS_FCANTRCVMORE;
722 static void
723 sofcantsendmore(struct socket *so)
725 if ((so->so_state & SS_NOFDREF) == 0) {
726 shutdown(so->s,1); /* send FIN to fhost */
727 if (global_readfds) {
728 FD_CLR(so->s,global_readfds);
730 if (global_xfds) {
731 FD_CLR(so->s,global_xfds);
734 so->so_state &= ~(SS_ISFCONNECTING);
735 if (so->so_state & SS_FCANTRCVMORE)
736 so->so_state = SS_NOFDREF; /* as above */
737 else
738 so->so_state |= SS_FCANTSENDMORE;
741 void
742 soisfdisconnected(struct socket *so)
744 /* so->so_state &= ~(SS_ISFCONNECTING|SS_ISFCONNECTED); */
745 /* close(so->s); */
746 /* so->so_state = SS_ISFDISCONNECTED; */
748 * XXX Do nothing ... ?
753 * Set write drain mode
754 * Set CANTSENDMORE once all data has been write()n
756 void
757 sofwdrain(struct socket *so)
759 if (so->so_rcv.sb_cc)
760 so->so_state |= SS_FWDRAIN;
761 else
762 sofcantsendmore(so);