socket.c@ 15287

Last change on this file since 15287 was 15261, checked in by vboxsync, 16 years ago
slirp: less noise
Property svn:eol-style set to `native`
File size: 27.1 KB

Line
1	/*
2	* Copyright (c) 1995 Danny Gasparovski.
3	*
4	* Please read the file COPYRIGHT for the
5	* terms and conditions of the copyright.
6	*/
7
8	#define WANT_SYS_IOCTL_H
9	#include <slirp.h>
10	#include "ip_icmp.h"
11	#include "main.h"
12	#ifdef __sun__
13	#include <sys/filio.h>
14	#endif
15	#if defined(VBOX_WITH_SLIRP_ICMP) && defined (RT_OS_WINDOWS)
16	#include <iphlpapi.h>
17	#include <icmpapi.h>
18	#endif
19
20	#ifdef VBOX_WITH_SLIRP_ICMP
21	static void send_icmp_to_guest(PNATState, char , size_t, struct socket , const struct sockaddr_in *);
22	static void sorecvfrom_icmp_win(PNATState, struct socket *);
23	#endif
24	static void sorecvfrom_icmp_unix(PNATState, struct socket *);
25
26	void
27	so_init()
28	{
29	}
30
31
32	struct socket *
33	solookup(struct socket *head, struct in_addr laddr,
34	u_int lport, struct in_addr faddr, u_int fport)
35	{
36	struct socket *so;
37
38	for (so = head->so_next; so != head; so = so->so_next)
39	{
40	if ( so->so_lport == lport
41	&& so->so_laddr.s_addr == laddr.s_addr
42	&& so->so_faddr.s_addr == faddr.s_addr
43	&& so->so_fport == fport)
44	return so;
45	}
46
47	return (struct socket *)NULL;
48	}
49
50	/*
51	* Create a new socket, initialise the fields
52	* It is the responsibility of the caller to
53	* insque() it into the correct linked-list
54	*/
55	struct socket *
56	socreate()
57	{
58	struct socket *so;
59
60	so = (struct socket *)malloc(sizeof(struct socket));
61	if(so)
62	{
63	memset(so, 0, sizeof(struct socket));
64	so->so_state = SS_NOFDREF;
65	so->s = -1;
66	}
67	return so;
68	}
69
70	/*
71	* remque and free a socket, clobber cache
72	*/
73	void
74	sofree(PNATState pData, struct socket *so)
75	{
76	if (so == tcp_last_so)
77	tcp_last_so = &tcb;
78	else if (so == udp_last_so)
79	udp_last_so = &udb;
80
81	m_free(pData, so->so_m);
82
83	if(so->so_next && so->so_prev)
84	remque(pData, so); /* crashes if so is not in a queue */
85
86	free(so);
87	}
88
89	/*
90	* Read from so's socket into sb_snd, updating all relevant sbuf fields
91	* NOTE: This will only be called if it is select()ed for reading, so
92	* a read() of 0 (or less) means it's disconnected
93	*/
94	int
95	soread(PNATState pData, struct socket *so, int fCloseIfNothingRead)
96	{
97	int n, nn, lss, total;
98	struct sbuf *sb = &so->so_snd;
99	size_t len = sb->sb_datalen - sb->sb_cc;
100	struct iovec iov[2];
101	int mss = so->so_tcpcb->t_maxseg;
102
103	DEBUG_CALL("soread");
104	DEBUG_ARG("so = %lx", (long )so);
105
106	/*
107	* No need to check if there's enough room to read.
108	* soread wouldn't have been called if there weren't
109	*/
110
111	len = sb->sb_datalen - sb->sb_cc;
112
113	iov[0].iov_base = sb->sb_wptr;
114	iov[1].iov_base = 0;
115	iov[1].iov_len = 0;
116	if (sb->sb_wptr < sb->sb_rptr)
117	{
118	iov[0].iov_len = sb->sb_rptr - sb->sb_wptr;
119	/* Should never succeed, but... */
120	if (iov[0].iov_len > len)
121	iov[0].iov_len = len;
122	if (iov[0].iov_len > mss)
123	iov[0].iov_len -= iov[0].iov_len%mss;
124	n = 1;
125	}
126	else
127	{
128	iov[0].iov_len = (sb->sb_data + sb->sb_datalen) - sb->sb_wptr;
129	/* Should never succeed, but... */
130	if (iov[0].iov_len > len)
131	iov[0].iov_len = len;
132	len -= iov[0].iov_len;
133	if (len)
134	{
135	iov[1].iov_base = sb->sb_data;
136	iov[1].iov_len = sb->sb_rptr - sb->sb_data;
137	if(iov[1].iov_len > len)
138	iov[1].iov_len = len;
139	total = iov[0].iov_len + iov[1].iov_len;
140	if (total > mss)
141	{
142	lss = total % mss;
143	if (iov[1].iov_len > lss)
144	{
145	iov[1].iov_len -= lss;
146	n = 2;
147	}
148	else
149	{
150	lss -= iov[1].iov_len;
151	iov[0].iov_len -= lss;
152	n = 1;
153	}
154	}
155	else
156	n = 2;
157	}
158	else
159	{
160	if (iov[0].iov_len > mss)
161	iov[0].iov_len -= iov[0].iov_len%mss;
162	n = 1;
163	}
164	}
165
166	#ifdef HAVE_READV
167	nn = readv(so->s, (struct iovec *)iov, n);
168	DEBUG_MISC((dfd, " ... read nn = %d bytes\n", nn));
169	#else
170	nn = recv(so->s, iov[0].iov_base, iov[0].iov_len,0);
171	#endif
172	if (nn <= 0)
173	{
174	#if defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) && defined(RT_OS_WINDOWS)
175	/*
176	* Special case for WSAEnumNetworkEvents: If we receive 0 bytes that
177	* _could_ mean that the connection is closed. But we will receive an
178	* FD_CLOSE event later if the connection was _really_ closed. With
179	* www.youtube.com I see this very often. Closing the socket too early
180	* would be dangerous.
181	*/
182	if (nn == 0 && !fCloseIfNothingRead)
183	return 0;
184	#endif
185	if (nn < 0 && (errno == EINTR \|\| errno == EAGAIN \|\| errno == EWOULDBLOCK))
186	return 0;
187	else
188	{
189	/* nn == 0 means peer has performed an orderly shutdown */
190	DEBUG_MISC((dfd, " --- soread() disconnected, nn = %d, errno = %d-%s\n",
191	nn, errno,strerror(errno)));
192	sofcantrcvmore(so);
193	tcp_sockclosed(pData, sototcpcb(so));
194	return -1;
195	}
196	}
197
198	#ifndef HAVE_READV
199	/*
200	* If there was no error, try and read the second time round
201	* We read again if n = 2 (ie, there's another part of the buffer)
202	* and we read as much as we could in the first read
203	* We don't test for <= 0 this time, because there legitimately
204	* might not be any more data (since the socket is non-blocking),
205	* a close will be detected on next iteration.
206	* A return of -1 wont (shouldn't) happen, since it didn't happen above
207	*/
208	if (n == 2 && nn == iov[0].iov_len)
209	{
210	int ret;
211	ret = recv(so->s, iov[1].iov_base, iov[1].iov_len,0);
212	if (ret > 0)
213	nn += ret;
214	}
215
216	DEBUG_MISC((dfd, " ... read nn = %d bytes\n", nn));
217	#endif
218
219	/* Update fields */
220	sb->sb_cc += nn;
221	sb->sb_wptr += nn;
222	if (sb->sb_wptr >= (sb->sb_data + sb->sb_datalen))
223	sb->sb_wptr -= sb->sb_datalen;
224	return nn;
225	}
226
227	/*
228	* Get urgent data
229	*
230	* When the socket is created, we set it SO_OOBINLINE,
231	* so when OOB data arrives, we soread() it and everything
232	* in the send buffer is sent as urgent data
233	*/
234	void
235	sorecvoob(PNATState pData, struct socket *so)
236	{
237	struct tcpcb *tp = sototcpcb(so);
238
239	DEBUG_CALL("sorecvoob");
240	DEBUG_ARG("so = %lx", (long)so);
241
242	/*
243	* We take a guess at how much urgent data has arrived.
244	* In most situations, when urgent data arrives, the next
245	* read() should get all the urgent data. This guess will
246	* be wrong however if more data arrives just after the
247	* urgent data, or the read() doesn't return all the
248	* urgent data.
249	*/
250	soread(pData, so, /fCloseIfNothingRead=/false);
251	tp->snd_up = tp->snd_una + so->so_snd.sb_cc;
252	tp->t_force = 1;
253	tcp_output(pData, tp);
254	tp->t_force = 0;
255	}
256
257	/*
258	* Send urgent data
259	* There's a lot duplicated code here, but...
260	*/
261	int
262	sosendoob(struct socket *so)
263	{
264	struct sbuf *sb = &so->so_rcv;
265	char buff[2048]; /* XXX Shouldn't be sending more oob data than this */
266
267	int n, len;
268
269	DEBUG_CALL("sosendoob");
270	DEBUG_ARG("so = %lx", (long)so);
271	DEBUG_ARG("sb->sb_cc = %d", sb->sb_cc);
272
273	if (so->so_urgc > sizeof(buff))
274	so->so_urgc = sizeof(buff); /* XXX */
275
276	if (sb->sb_rptr < sb->sb_wptr)
277	{
278	/* We can send it directly */
279	n = send(so->s, sb->sb_rptr, so->so_urgc, (MSG_OOB)); /* \|MSG_DONTWAIT)); */
280	so->so_urgc -= n;
281
282	DEBUG_MISC((dfd, " --- sent %d bytes urgent data, %d urgent bytes left\n",
283	n, so->so_urgc));
284	}
285	else
286	{
287	/*
288	* Since there's no sendv or sendtov like writev,
289	* we must copy all data to a linear buffer then
290	* send it all
291	*/
292	len = (sb->sb_data + sb->sb_datalen) - sb->sb_rptr;
293	if (len > so->so_urgc) len = so->so_urgc;
294	memcpy(buff, sb->sb_rptr, len);
295	so->so_urgc -= len;
296	if (so->so_urgc)
297	{
298	n = sb->sb_wptr - sb->sb_data;
299	if (n > so->so_urgc)
300	n = so->so_urgc;
301	memcpy(buff + len, sb->sb_data, n);
302	so->so_urgc -= n;
303	len += n;
304	}
305	n = send(so->s, buff, len, (MSG_OOB)); /* \|MSG_DONTWAIT)); */
306	#ifdef DEBUG
307	if (n != len)
308	DEBUG_ERROR((dfd, "Didn't send all data urgently XXXXX\n"));
309	#endif
310	DEBUG_MISC((dfd, " ---2 sent %d bytes urgent data, %d urgent bytes left\n",
311	n, so->so_urgc));
312	}
313
314	sb->sb_cc -= n;
315	sb->sb_rptr += n;
316	if (sb->sb_rptr >= (sb->sb_data + sb->sb_datalen))
317	sb->sb_rptr -= sb->sb_datalen;
318
319	return n;
320	}
321
322	/*
323	* Write data from so_rcv to so's socket,
324	* updating all sbuf field as necessary
325	*/
326	int
327	sowrite(PNATState pData, struct socket *so)
328	{
329	int n,nn;
330	struct sbuf *sb = &so->so_rcv;
331	size_t len = sb->sb_cc;
332	struct iovec iov[2];
333
334	DEBUG_CALL("sowrite");
335	DEBUG_ARG("so = %lx", (long)so);
336
337	if (so->so_urgc)
338	{
339	sosendoob(so);
340	if (sb->sb_cc == 0)
341	return 0;
342	}
343
344	/*
345	* No need to check if there's something to write,
346	* sowrite wouldn't have been called otherwise
347	*/
348
349	len = sb->sb_cc;
350
351	iov[0].iov_base = sb->sb_rptr;
352	iov[1].iov_base = 0;
353	iov[1].iov_len = 0;
354	if (sb->sb_rptr < sb->sb_wptr)
355	{
356	iov[0].iov_len = sb->sb_wptr - sb->sb_rptr;
357	/* Should never succeed, but... */
358	if (iov[0].iov_len > len)
359	iov[0].iov_len = len;
360	n = 1;
361	}
362	else
363	{
364	iov[0].iov_len = (sb->sb_data + sb->sb_datalen) - sb->sb_rptr;
365	if (iov[0].iov_len > len)
366	iov[0].iov_len = len;
367	len -= iov[0].iov_len;
368	if (len)
369	{
370	iov[1].iov_base = sb->sb_data;
371	iov[1].iov_len = sb->sb_wptr - sb->sb_data;
372	if (iov[1].iov_len > len)
373	iov[1].iov_len = len;
374	n = 2;
375	}
376	else
377	n = 1;
378	}
379	/* Check if there's urgent data to send, and if so, send it */
380	#ifdef HAVE_READV
381	nn = writev(so->s, (const struct iovec *)iov, n);
382	DEBUG_MISC((dfd, " ... wrote nn = %d bytes\n", nn));
383	#else
384	nn = send(so->s, iov[0].iov_base, iov[0].iov_len, 0);
385	#endif
386	/* This should never happen, but people tell me it does shrug */
387	if (nn < 0 && (errno == EAGAIN \|\| errno == EINTR \|\| errno == EWOULDBLOCK))
388	return 0;
389
390	if (nn < 0 \|\| (nn == 0 && iov[0].iov_len > 0))
391	{
392	DEBUG_MISC((dfd, " --- sowrite disconnected, so->so_state = %x, errno = %d\n",
393	so->so_state, errno));
394	sofcantsendmore(so);
395	tcp_sockclosed(pData, sototcpcb(so));
396	return -1;
397	}
398
399	#ifndef HAVE_READV
400	if (n == 2 && nn == iov[0].iov_len)
401	{
402	int ret;
403	ret = send(so->s, iov[1].iov_base, iov[1].iov_len,0);
404	if (ret > 0)
405	nn += ret;
406	}
407	DEBUG_MISC((dfd, " ... wrote nn = %d bytes\n", nn));
408	#endif
409
410	/* Update sbuf */
411	sb->sb_cc -= nn;
412	sb->sb_rptr += nn;
413	if (sb->sb_rptr >= (sb->sb_data + sb->sb_datalen))
414	sb->sb_rptr -= sb->sb_datalen;
415
416	/*
417	* If in DRAIN mode, and there's no more data, set
418	* it CANTSENDMORE
419	*/
420	if ((so->so_state & SS_FWDRAIN) && sb->sb_cc == 0)
421	sofcantsendmore(so);
422
423	return nn;
424	}
425
426	/*
427	* recvfrom() a UDP socket
428	*/
429	void
430	sorecvfrom(PNATState pData, struct socket *so)
431	{
432	struct sockaddr_in addr;
433	socklen_t addrlen = sizeof(struct sockaddr_in);
434
435	DEBUG_CALL("sorecvfrom");
436	DEBUG_ARG("so = %lx", (long)so);
437
438	if (so->so_type == IPPROTO_ICMP)
439	{
440	/* This is a "ping" reply */
441	#if !defined(VBOX_WITH_SLIRP_ICMP) \|\| (defined(VBOX_WITH_SLIRP_ICMP) && !defined(RT_OS_WINDOWS))
442	sorecvfrom_icmp_unix(pData, so);
443	#endif
444	#if defined(VBOX_WITH_SLIRP_ICMP) && defined(RT_OS_WINDOWS)
445	sorecvfrom_icmp_win(pData, so);
446	#endif
447	udp_detach(pData, so);
448	}
449	else
450	{
451	/* A "normal" UDP packet */
452	struct mbuf *m;
453	size_t len;
454	u_long n;
455
456	if (!(m = m_get(pData)))
457	return;
458	m->m_data += if_maxlinkhdr;
459
460	/*
461	* XXX Shouldn't FIONREAD packets destined for port 53,
462	* but I don't know the max packet size for DNS lookups
463	*/
464	len = M_FREEROOM(m);
465	/* if (so->so_fport != htons(53)) */
466	{
467	ioctlsocket(so->s, FIONREAD, &n);
468
469	if (n > len)
470	{
471	n = (m->m_data - m->m_dat) + m->m_len + n + 1;
472	m_inc(m, n);
473	len = M_FREEROOM(m);
474	}
475	}
476
477	m->m_len = recvfrom(so->s, m->m_data, len, 0,
478	(struct sockaddr *)&addr, &addrlen);
479	DEBUG_MISC((dfd, " did recvfrom %d, errno = %d-%s\n",
480	m->m_len, errno,strerror(errno)));
481	if(m->m_len < 0)
482	{
483	u_char code=ICMP_UNREACH_PORT;
484
485	if (errno == EHOSTUNREACH)
486	code=ICMP_UNREACH_HOST;
487	else if(errno == ENETUNREACH)
488	code=ICMP_UNREACH_NET;
489
490	DEBUG_MISC((dfd," rx error, tx icmp ICMP_UNREACH:%i\n", code));
491	icmp_error(pData, so->so_m, ICMP_UNREACH,code, 0,strerror(errno));
492	m_free(pData, m);
493	}
494	else
495	{
496	/*
497	* Hack: domain name lookup will be used the most for UDP,
498	* and since they'll only be used once there's no need
499	* for the 4 minute (or whatever) timeout... So we time them
500	* out much quicker (10 seconds for now...)
501	*/
502	if (so->so_expire)
503	{
504	if (so->so_fport == htons(53))
505	so->so_expire = curtime + SO_EXPIREFAST;
506	else
507	so->so_expire = curtime + SO_EXPIRE;
508	}
509
510	#if 0
511	if (m->m_len == len)
512	{
513	m_inc(m, MINCSIZE);
514	m->m_len = 0;
515	}
516	#endif
517
518	/*
519	* If this packet was destined for CTL_ADDR,
520	* make it look like that's where it came from, done by udp_output
521	*/
522	udp_output(pData, so, m, &addr);
523	} /* rx error */
524	} /* if ping packet */
525	}
526
527	/*
528	* sendto() a socket
529	*/
530	int
531	sosendto(PNATState pData, struct socket so, struct mbuf m)
532	{
533	int ret;
534	struct sockaddr_in addr;
535	#if 0
536	struct sockaddr_in host_addr;
537	#endif
538
539	DEBUG_CALL("sosendto");
540	DEBUG_ARG("so = %lx", (long)so);
541	DEBUG_ARG("m = %lx", (long)m);
542
543	addr.sin_family = AF_INET;
544	if ((so->so_faddr.s_addr & htonl(pData->netmask)) == special_addr.s_addr)
545	{
546	/* It's an alias */
547	uint32_t last_byte = ntohl(so->so_faddr.s_addr) & ~pData->netmask;
548	switch(last_byte)
549	{
550	#if 0
551	/* handle this case at 'default:' */
552	case CTL_BROADCAST:
553	addr.sin_addr.s_addr = INADDR_BROADCAST;
554	/* Send the packet to host to fully emulate broadcast */
555	/** @todo r=klaus: on Linux host this causes the host to receive
556	* the packet twice for some reason. And I cannot find any place
557	* in the man pages which states that sending a broadcast does not
558	* reach the host itself. */
559	host_addr.sin_family = AF_INET;
560	host_addr.sin_port = so->so_fport;
561	host_addr.sin_addr = our_addr;
562	sendto(so->s, m->m_data, m->m_len, 0,
563	(struct sockaddr *)&host_addr, sizeof (struct sockaddr));
564	break;
565	#endif
566	case CTL_DNS:
567	if (!get_dns_addr(pData, &dns_addr))
568	addr.sin_addr = dns_addr;
569	else
570	addr.sin_addr = loopback_addr;
571	break;
572	case CTL_ALIAS:
573	default:
574	if (last_byte == ~pData->netmask)
575	addr.sin_addr.s_addr = INADDR_BROADCAST;
576	else
577	addr.sin_addr = loopback_addr;
578	break;
579	}
580	}
581	else
582	addr.sin_addr = so->so_faddr;
583	addr.sin_port = so->so_fport;
584
585	DEBUG_MISC((dfd, " sendto()ing, addr.sin_port=%d, addr.sin_addr.s_addr=%.16s\n",
586	ntohs(addr.sin_port), inet_ntoa(addr.sin_addr)));
587
588	/* Don't care what port we get */
589	ret = sendto(so->s, m->m_data, m->m_len, 0,
590	(struct sockaddr *)&addr, sizeof (struct sockaddr));
591	if (ret < 0)
592	return -1;
593
594	/*
595	* Kill the socket if there's no reply in 4 minutes,
596	* but only if it's an expirable socket
597	*/
598	if (so->so_expire)
599	so->so_expire = curtime + SO_EXPIRE;
600	so->so_state = SS_ISFCONNECTED; /* So that it gets select()ed */
601	return 0;
602	}
603
604	/*
605	* XXX This should really be tcp_listen
606	*/
607	struct socket *
608	solisten(PNATState pData, u_int port, u_int32_t laddr, u_int lport, int flags)
609	{
610	struct sockaddr_in addr;
611	struct socket *so;
612	socklen_t addrlen = sizeof(addr);
613	int s, opt = 1;
614
615	DEBUG_CALL("solisten");
616	DEBUG_ARG("port = %d", port);
617	DEBUG_ARG("laddr = %x", laddr);
618	DEBUG_ARG("lport = %d", lport);
619	DEBUG_ARG("flags = %x", flags);
620
621	if ((so = socreate()) == NULL)
622	{
623	/* free(so); Not sofree() ??? free(NULL) == NOP */
624	return NULL;
625	}
626
627	/* Don't tcp_attach... we don't need so_snd nor so_rcv */
628	if ((so->so_tcpcb = tcp_newtcpcb(pData, so)) == NULL)
629	{
630	free(so);
631	return NULL;
632	}
633	insque(pData, so,&tcb);
634
635	/*
636	* SS_FACCEPTONCE sockets must time out.
637	*/
638	if (flags & SS_FACCEPTONCE)
639	so->so_tcpcb->t_timer[TCPT_KEEP] = TCPTV_KEEP_INIT*2;
640
641	so->so_state = (SS_FACCEPTCONN\|flags);
642	so->so_lport = lport; /* Kept in network format */
643	so->so_laddr.s_addr = laddr; /* Ditto */
644
645	addr.sin_family = AF_INET;
646	addr.sin_addr.s_addr = INADDR_ANY;
647	addr.sin_port = port;
648
649	if ( ((s = socket(AF_INET,SOCK_STREAM,0)) < 0)
650	\|\| (setsockopt(s,SOL_SOCKET,SO_REUSEADDR,(char *)&opt,sizeof(int)) < 0)
651	\|\| (bind(s,(struct sockaddr *)&addr, sizeof(addr)) < 0)
652	\|\| (listen(s,1) < 0))
653	{
654	#ifdef RT_OS_WINDOWS
655	int tmperrno = WSAGetLastError(); /* Don't clobber the real reason we failed */
656	closesocket(s);
657	sofree(pData, so);
658	/* Restore the real errno */
659	WSASetLastError(tmperrno);
660	#else
661	int tmperrno = errno; /* Don't clobber the real reason we failed */
662	close(s);
663	sofree(pData, so);
664	/* Restore the real errno */
665	errno = tmperrno;
666	#endif
667	return NULL;
668	}
669	setsockopt(s,SOL_SOCKET,SO_OOBINLINE,(char *)&opt,sizeof(int));
670
671	getsockname(s,(struct sockaddr *)&addr,&addrlen);
672	so->so_fport = addr.sin_port;
673	if (addr.sin_addr.s_addr == 0 \|\| addr.sin_addr.s_addr == loopback_addr.s_addr)
674	so->so_faddr = alias_addr;
675	else
676	so->so_faddr = addr.sin_addr;
677
678	so->s = s;
679	return so;
680	}
681
682	/*
683	* Data is available in so_rcv
684	* Just write() the data to the socket
685	* XXX not yet...
686	*/
687	void
688	sorwakeup(struct socket *so)
689	{
690	#if 0
691	sowrite(so);
692	FD_CLR(so->s,&writefds);
693	#endif
694	}
695
696	/*
697	* Data has been freed in so_snd
698	* We have room for a read() if we want to
699	* For now, don't read, it'll be done in the main loop
700	*/
701	void
702	sowwakeup(struct socket *so)
703	{
704	}
705
706	/*
707	* Various session state calls
708	* XXX Should be #define's
709	* The socket state stuff needs work, these often get call 2 or 3
710	* times each when only 1 was needed
711	*/
712	void
713	soisfconnecting(struct socket *so)
714	{
715	so->so_state &= ~(SS_NOFDREF\|SS_ISFCONNECTED\|SS_FCANTRCVMORE\|
716	SS_FCANTSENDMORE\|SS_FWDRAIN);
717	so->so_state \|= SS_ISFCONNECTING; /* Clobber other states */
718	}
719
720	void
721	soisfconnected(struct socket *so)
722	{
723	so->so_state &= ~(SS_ISFCONNECTING\|SS_FWDRAIN\|SS_NOFDREF);
724	so->so_state \|= SS_ISFCONNECTED; /* Clobber other states */
725	}
726
727	void
728	sofcantrcvmore(struct socket *so)
729	{
730	if ((so->so_state & SS_NOFDREF) == 0)
731	{
732	shutdown(so->s,0);
733	}
734	so->so_state &= ~(SS_ISFCONNECTING);
735	if (so->so_state & SS_FCANTSENDMORE)
736	so->so_state = SS_NOFDREF; /* Don't select it */
737	/* XXX close() here as well? */
738	else
739	so->so_state \|= SS_FCANTRCVMORE;
740	}
741
742	void
743	sofcantsendmore(struct socket *so)
744	{
745	if ((so->so_state & SS_NOFDREF) == 0)
746	shutdown(so->s, 1); /* send FIN to fhost */
747
748	so->so_state &= ~(SS_ISFCONNECTING);
749	if (so->so_state & SS_FCANTRCVMORE)
750	so->so_state = SS_NOFDREF; /* as above */
751	else
752	so->so_state \|= SS_FCANTSENDMORE;
753	}
754
755	void
756	soisfdisconnected(struct socket *so)
757	{
758	#if 0
759	so->so_state &= ~(SS_ISFCONNECTING\|SS_ISFCONNECTED);
760	close(so->s);
761	so->so_state = SS_ISFDISCONNECTED;
762	/*
763	* XXX Do nothing ... ?
764	*/
765	#endif
766	}
767
768	/*
769	* Set write drain mode
770	* Set CANTSENDMORE once all data has been write()n
771	*/
772	void
773	sofwdrain(struct socket *so)
774	{
775	if (so->so_rcv.sb_cc)
776	so->so_state \|= SS_FWDRAIN;
777	else
778	sofcantsendmore(so);
779	}
780
781	#ifdef VBOX_WITH_SLIRP_ICMP
782	static void
783	send_icmp_to_guest(PNATState pData, char buff, size_t len, struct socket so, const struct sockaddr_in *addr)
784	{
785	struct ip *ip;
786	uint32_t dst,src;
787	char ip_copy[256];
788	struct icmp *icp;
789	int old_ip_len;
790	struct mbuf *m;
791	struct icmp_msg *icm;
792
793	ip = (struct ip *)buff;
794	icp = (struct icmp )((char )ip + (ip->ip_hl << 2));
795
796	if (icp->icmp_type != ICMP_ECHOREPLY && icp->icmp_type != ICMP_TIMXCEED)
797	{
798	LogRel(("received ICMP(t:%d, c:%d)\n", icp->icmp_type, icp->icmp_code));
799	return;
800	}
801
802	if (icp->icmp_type == ICMP_TIMXCEED)
803	ip = &icp->icmp_ip;
804
805	icm = icmp_find_original_mbuf(pData, ip);
806
807	if (icm == NULL)
808	{
809	LogRel(("NAT: Can't find the corresponding packet for the received ICMP\n"));
810	return;
811	}
812
813	m = icm->im_m;
814	Assert(m != NULL);
815
816	src = addr->sin_addr.s_addr;
817
818	ip = mtod(m, struct ip *);
819	/* Now ip is pointing on header we've sent from guest */
820	if (icp->icmp_type == ICMP_TIMXCEED)
821	{
822	old_ip_len = (ip->ip_hl << 2) + 64;
823	memcpy(ip_copy, ip, old_ip_len);
824	}
825
826	/* source address from original IP packet*/
827	dst = ip->ip_src.s_addr;
828
829	/* overide ther tail of old packet */
830	memcpy(m->m_data, buff, len);
831	m->m_len = len;
832	ip = mtod(m, struct ip ); / ip is from mbuf we've overrided */
833
834	icp = (struct icmp )((char )ip + (ip->ip_hl << 2));
835	if (icp->icmp_type == ICMP_TIMXCEED)
836	{
837	/* according RFC 793 error messages required copy of initial IP header + 64 bit */
838	memcpy(&icp->icmp_ip, ip_copy, old_ip_len);
839	ip->ip_tos=((ip->ip_tos & 0x1E) \| 0xC0); /* high priority for errors */
840	}
841
842	/* the low level expects fields to be in host format so let's convert them*/
843	NTOHS(ip->ip_len);
844	NTOHS(ip->ip_off);
845	NTOHS(ip->ip_id);
846	ip->ip_src.s_addr = src;
847	ip->ip_dst.s_addr = dst;
848	icmp_reflect(pData, m);
849	LIST_REMOVE(icm, im_list);
850	/* Don't call m_free here*/
851	free(icm);
852	}
853
854	# ifdef RT_OS_WINDOWS
855	static void
856	sorecvfrom_icmp_win(PNATState pData, struct socket *so)
857	{
858	int len;
859	int i;
860	struct ip *ip;
861	struct mbuf *m;
862	struct icmp *icp;
863	struct icmp_msg *icm;
864	struct ip ip_broken; / ICMP returns header + 64 bit of packet */
865	uint32_t src;
866	ICMP_ECHO_REPLY *icr;
867	u_char code = ~0;
868	len = pData->pfIcmpParseReplies(pData->pvIcmpBuffer, pData->szIcmpBuffer);
869	if (len < 0)
870	{
871	LogRel(("NAT: Error (%d) occurred on ICMP receiving\n", GetLastError()));
872	return;
873	}
874	if (len == 0)
875	return; /* no error */
876	LogRel(("IcmpParseReplies returns %ld\n", len));
877	icr = (ICMP_ECHO_REPLY *)pData->pvIcmpBuffer;
878	for (i = 0; i < len; ++i)
879	{
880	switch(icr[i].Status)
881	{
882	case IP_DEST_HOST_UNREACHABLE:
883	code = (code != ~0 ? code : ICMP_UNREACH_HOST);
884	case IP_DEST_NET_UNREACHABLE:
885	code = (code != ~0 ? code : ICMP_UNREACH_NET);
886	case IP_DEST_PROT_UNREACHABLE:
887	code = (code != ~0 ? code : ICMP_UNREACH_PROTOCOL);
888	/* UNREACH error inject here */
889	case IP_DEST_PORT_UNREACHABLE:
890	code = (code != ~0 ? code : ICMP_UNREACH_PORT);
891	icmp_error(pData, so->so_m, ICMP_UNREACH, code, 0, "Error occurred!!!");
892	break;
893	case IP_SUCCESS: /* echo replied */
894	m = m_get(pData);
895	ip = mtod(m, struct ip *);
896	ip->ip_src.s_addr = icr[i].Address;
897	ip->ip_p = IPPROTO_ICMP;
898	ip->ip_dst.s_addr = so->so_laddr.s_addr; /XXX: still the hack/
899	ip->ip_hl = sizeof(struct ip) >> 2; /* requiered for icmp_reflect, no IP options */
900	ip->ip_ttl = icr[i].Options.Ttl;
901
902	icp = (struct icmp )&ip[1]; / no options */
903	icp->icmp_type = ICMP_ECHOREPLY;
904	icp->icmp_code = 0;
905	icp->icmp_id = so->so_icmp_id;
906	icp->icmp_seq = so->so_icmp_seq;
907	memcpy(icp->icmp_data, icr[i].Data, icr[i].DataSize);
908
909	ip->ip_len = sizeof(struct ip) + ICMP_MINLEN + icr[i].DataSize;
910	m->m_len = ip->ip_len;
911
912	icmp_reflect(pData, m);
913	case IP_TTL_EXPIRED_TRANSIT: /* TTL expired */
914
915	ip_broken = icr[i].Data;
916	icm = icmp_find_original_mbuf(pData, ip_broken);
917	if (icm == NULL) {
918	LogRel(("ICMP: can't find original package (first double word %x)\n", (uint32_t )ip_broken));
919	return;
920	}
921	m = icm->im_m;
922	ip = mtod(m, struct ip *);
923	ip->ip_ttl = icr[i].Options.Ttl;
924	src = ip->ip_src.s_addr;
925	ip->ip_dst.s_addr = src;
926	ip->ip_dst.s_addr = icr[i].Address;
927	icp = (struct icmp )((char )ip + (ip->ip_hl << 2));
928	ip_broken->ip_src.s_addr = src; /it packet sent from host not from guest/
929	memcpy(icp->icmp_data, ip_broken, (ip_broken->ip_hl << 2) + 64);
930	icmp_reflect(pData, m);
931	break;
932	default:
933	LogRel(("ICMP(default): message with Status: %x was received from %x\n", icr[i].Status, icr[i].Address));
934	break;
935	}
936	}
937	}
938	# endif /* RT_OS_WINDOWS */
939	#endif /* VBOX_WITH_SLIRP_ICMP */
940
941	static void sorecvfrom_icmp_unix(PNATState pData, struct socket *so)
942	{
943	struct sockaddr_in addr;
944	socklen_t addrlen = sizeof(struct sockaddr_in);
945	char buff[1500];
946	int len;
947	len = recvfrom(so->s, buff, 1500, 0,
948	(struct sockaddr *)&addr, &addrlen);
949	/* XXX Check if reply is "correct"? */
950
951	if (len == -1 \|\| len == 0)
952	{
953	u_char code = ICMP_UNREACH_PORT;
954
955	if (errno == EHOSTUNREACH)
956	code=ICMP_UNREACH_HOST;
957	else if(errno == ENETUNREACH)
958	code=ICMP_UNREACH_NET;
959
960	DEBUG_MISC((dfd," udp icmp rx errno = %d-%s\n",
961	errno,strerror(errno)));
962	icmp_error(pData, so->so_m, ICMP_UNREACH,code, 0,strerror(errno));
963	}
964	else
965	{
966	#ifdef VBOX_WITH_SLIRP_ICMP
967	send_icmp_to_guest(pData, buff, len, so, &addr);
968	#else
969	icmp_reflect(pData, so->so_m);
970	so->so_m = 0; /* Don't m_free() it again! */
971	#endif
972	}
973	}
974

Note: See TracBrowser for help on using the repository browser.

source: vbox/trunk/src/VBox/Devices/Network/slirp/socket.c@ 15287

Download in other formats: