socket.c@ 27976

Last change on this file since 27976 was 27976, checked in by vboxsync, 15 years ago
*: scm cleans up whitespace and adds a new line at the end of ApplianceimplPrivate.h.
Property svn:eol-style set to `native`
File size: 41.3 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	#include <VBox/pdmdrv.h>
16	#if defined (RT_OS_WINDOWS)
17	#include <iphlpapi.h>
18	#include <icmpapi.h>
19	#endif
20
21
22	static void send_icmp_to_guest(PNATState, char , size_t, struct socket , const struct sockaddr_in *);
23	#ifdef RT_OS_WINDOWS
24	static void sorecvfrom_icmp_win(PNATState, struct socket *);
25	#else /* RT_OS_WINDOWS */
26	static void sorecvfrom_icmp_unix(PNATState, struct socket *);
27	#endif /* !RT_OS_WINDOWS */
28
29	void
30	so_init()
31	{
32	}
33
34	struct socket *
35	solookup(struct socket *head, struct in_addr laddr,
36	u_int lport, struct in_addr faddr, u_int fport)
37	{
38	struct socket *so;
39
40	for (so = head->so_next; so != head; so = so->so_next)
41	{
42	if ( so->so_lport == lport
43	&& so->so_laddr.s_addr == laddr.s_addr
44	&& so->so_faddr.s_addr == faddr.s_addr
45	&& so->so_fport == fport)
46	return so;
47	}
48
49	return (struct socket *)NULL;
50	}
51
52	/*
53	* Create a new socket, initialise the fields
54	* It is the responsibility of the caller to
55	* insque() it into the correct linked-list
56	*/
57	struct socket *
58	socreate()
59	{
60	struct socket *so;
61
62	so = (struct socket *)RTMemAllocZ(sizeof(struct socket));
63	if (so)
64	{
65	so->so_state = SS_NOFDREF;
66	so->s = -1;
67	#if !defined(RT_OS_WINDOWS)
68	so->so_poll_index = -1;
69	#endif
70	}
71	return so;
72	}
73
74	/*
75	* remque and free a socket, clobber cache
76	* VBOX_WITH_SLIRP_MT: before sofree queue should be locked, because
77	* in sofree we don't know from which queue item beeing removed.
78	*/
79	void
80	sofree(PNATState pData, struct socket *so)
81	{
82	struct socket *so_prev = NULL;
83	if (so == tcp_last_so)
84	tcp_last_so = &tcb;
85	else if (so == udp_last_so)
86	udp_last_so = &udb;
87
88	/* check if mbuf haven't been already freed */
89	if (so->so_m != NULL)
90	m_free(pData, so->so_m);
91	#ifndef VBOX_WITH_SLIRP_MT
92	if (so->so_next && so->so_prev)
93	{
94	remque(pData, so); /* crashes if so is not in a queue */
95	NSOCK_DEC();
96	}
97
98	RTMemFree(so);
99	#else
100	so->so_deleted = 1;
101	#endif
102	}
103
104	#ifdef VBOX_WITH_SLIRP_MT
105	void
106	soread_queue(PNATState pData, struct socket so, int ret)
107	{
108	*ret = soread(pData, so);
109	}
110	#endif
111
112	/*
113	* Read from so's socket into sb_snd, updating all relevant sbuf fields
114	* NOTE: This will only be called if it is select()ed for reading, so
115	* a read() of 0 (or less) means it's disconnected
116	*/
117	int
118	soread(PNATState pData, struct socket *so)
119	{
120	int n, nn, lss, total;
121	struct sbuf *sb = &so->so_snd;
122	size_t len = sb->sb_datalen - sb->sb_cc;
123	struct iovec iov[2];
124	int mss = so->so_tcpcb->t_maxseg;
125
126	STAM_PROFILE_START(&pData->StatIOread, a);
127	STAM_COUNTER_RESET(&pData->StatIORead_in_1);
128	STAM_COUNTER_RESET(&pData->StatIORead_in_2);
129
130	QSOCKET_LOCK(tcb);
131	SOCKET_LOCK(so);
132	QSOCKET_UNLOCK(tcb);
133
134	DEBUG_CALL("soread");
135	DEBUG_ARG("so = %lx", (long)so);
136
137	/*
138	* No need to check if there's enough room to read.
139	* soread wouldn't have been called if there weren't
140	*/
141
142	len = sb->sb_datalen - sb->sb_cc;
143
144	iov[0].iov_base = sb->sb_wptr;
145	iov[1].iov_base = 0;
146	iov[1].iov_len = 0;
147	if (sb->sb_wptr < sb->sb_rptr)
148	{
149	iov[0].iov_len = sb->sb_rptr - sb->sb_wptr;
150	/* Should never succeed, but... */
151	if (iov[0].iov_len > len)
152	iov[0].iov_len = len;
153	if (iov[0].iov_len > mss)
154	iov[0].iov_len -= iov[0].iov_len%mss;
155	n = 1;
156	}
157	else
158	{
159	iov[0].iov_len = (sb->sb_data + sb->sb_datalen) - sb->sb_wptr;
160	/* Should never succeed, but... */
161	if (iov[0].iov_len > len)
162	iov[0].iov_len = len;
163	len -= iov[0].iov_len;
164	if (len)
165	{
166	iov[1].iov_base = sb->sb_data;
167	iov[1].iov_len = sb->sb_rptr - sb->sb_data;
168	if (iov[1].iov_len > len)
169	iov[1].iov_len = len;
170	total = iov[0].iov_len + iov[1].iov_len;
171	if (total > mss)
172	{
173	lss = total % mss;
174	if (iov[1].iov_len > lss)
175	{
176	iov[1].iov_len -= lss;
177	n = 2;
178	}
179	else
180	{
181	lss -= iov[1].iov_len;
182	iov[0].iov_len -= lss;
183	n = 1;
184	}
185	}
186	else
187	n = 2;
188	}
189	else
190	{
191	if (iov[0].iov_len > mss)
192	iov[0].iov_len -= iov[0].iov_len%mss;
193	n = 1;
194	}
195	}
196
197	#ifdef HAVE_READV
198	nn = readv(so->s, (struct iovec *)iov, n);
199	DEBUG_MISC((dfd, " ... read nn = %d bytes\n", nn));
200	#else
201	nn = recv(so->s, iov[0].iov_base, iov[0].iov_len, 0);
202	#endif
203	if (nn <= 0)
204	{
205	#ifdef RT_OS_WINDOWS
206	/*
207	* Special case for WSAEnumNetworkEvents: If we receive 0 bytes that
208	* _could_ mean that the connection is closed. But we will receive an
209	* FD_CLOSE event later if the connection was _really_ closed. With
210	* www.youtube.com I see this very often. Closing the socket too early
211	* would be dangerous.
212	*/
213	int status, ignored;
214	unsigned long pending = 0;
215	status = WSAIoctl(so->s, FIONREAD, NULL, 0, &pending, sizeof(unsigned long), &ignored, NULL, NULL);
216	if (status < 0)
217	LogRel(("NAT:error in WSAIoctl: %d\n", WSAGetLastError()));
218	if (nn == 0 && (pending != 0))
219	{
220	SOCKET_UNLOCK(so);
221	STAM_PROFILE_STOP(&pData->StatIOread, a);
222	return 0;
223	}
224	#endif
225	if ( nn < 0
226	&& ( errno == EINTR
227	\|\| errno == EAGAIN
228	\|\| errno == EWOULDBLOCK))
229	{
230	SOCKET_UNLOCK(so);
231	STAM_PROFILE_STOP(&pData->StatIOread, a);
232	return 0;
233	}
234	else
235	{
236	/* nn == 0 means peer has performed an orderly shutdown */
237	DEBUG_MISC((dfd, " --- soread() disconnected, nn = %d, errno = %d-%s\n",
238	nn, errno, strerror(errno)));
239	sofcantrcvmore(so);
240	tcp_sockclosed(pData, sototcpcb(so));
241	SOCKET_UNLOCK(so);
242	STAM_PROFILE_STOP(&pData->StatIOread, a);
243	return -1;
244	}
245	}
246	STAM_STATS(
247	if (n == 1)
248	{
249	STAM_COUNTER_INC(&pData->StatIORead_in_1);
250	STAM_COUNTER_ADD(&pData->StatIORead_in_1_bytes, nn);
251	}
252	else
253	{
254	STAM_COUNTER_INC(&pData->StatIORead_in_2);
255	STAM_COUNTER_ADD(&pData->StatIORead_in_2_1st_bytes, nn);
256	}
257	);
258
259	#ifndef HAVE_READV
260	/*
261	* If there was no error, try and read the second time round
262	* We read again if n = 2 (ie, there's another part of the buffer)
263	* and we read as much as we could in the first read
264	* We don't test for <= 0 this time, because there legitimately
265	* might not be any more data (since the socket is non-blocking),
266	* a close will be detected on next iteration.
267	* A return of -1 wont (shouldn't) happen, since it didn't happen above
268	*/
269	if (n == 2 && nn == iov[0].iov_len)
270	{
271	int ret;
272	ret = recv(so->s, iov[1].iov_base, iov[1].iov_len, 0);
273	if (ret > 0)
274	nn += ret;
275	STAM_STATS(
276	if (ret > 0)
277	{
278	STAM_COUNTER_INC(&pData->StatIORead_in_2);
279	STAM_COUNTER_ADD(&pData->StatIORead_in_2_2nd_bytes, ret);
280	}
281	);
282	}
283
284	DEBUG_MISC((dfd, " ... read nn = %d bytes\n", nn));
285	#endif
286
287	/* Update fields */
288	sb->sb_cc += nn;
289	sb->sb_wptr += nn;
290	if (sb->sb_wptr >= (sb->sb_data + sb->sb_datalen))
291	sb->sb_wptr -= sb->sb_datalen;
292	STAM_PROFILE_STOP(&pData->StatIOread, a);
293	SOCKET_UNLOCK(so);
294	return nn;
295	}
296
297	/*
298	* Get urgent data
299	*
300	* When the socket is created, we set it SO_OOBINLINE,
301	* so when OOB data arrives, we soread() it and everything
302	* in the send buffer is sent as urgent data
303	*/
304	void
305	sorecvoob(PNATState pData, struct socket *so)
306	{
307	struct tcpcb *tp = sototcpcb(so);
308
309	DEBUG_CALL("sorecvoob");
310	DEBUG_ARG("so = %lx", (long)so);
311
312	/*
313	* We take a guess at how much urgent data has arrived.
314	* In most situations, when urgent data arrives, the next
315	* read() should get all the urgent data. This guess will
316	* be wrong however if more data arrives just after the
317	* urgent data, or the read() doesn't return all the
318	* urgent data.
319	*/
320	soread(pData, so);
321	tp->snd_up = tp->snd_una + so->so_snd.sb_cc;
322	tp->t_force = 1;
323	tcp_output(pData, tp);
324	tp->t_force = 0;
325	}
326
327	/*
328	* Send urgent data
329	* There's a lot duplicated code here, but...
330	*/
331	int
332	sosendoob(struct socket *so)
333	{
334	struct sbuf *sb = &so->so_rcv;
335	char buff[2048]; /* XXX Shouldn't be sending more oob data than this */
336
337	int n, len;
338
339	DEBUG_CALL("sosendoob");
340	DEBUG_ARG("so = %lx", (long)so);
341	DEBUG_ARG("sb->sb_cc = %d", sb->sb_cc);
342
343	if (so->so_urgc > sizeof(buff))
344	so->so_urgc = sizeof(buff); /* XXX */
345
346	if (sb->sb_rptr < sb->sb_wptr)
347	{
348	/* We can send it directly */
349	n = send(so->s, sb->sb_rptr, so->so_urgc, (MSG_OOB)); /* \|MSG_DONTWAIT)); */
350	so->so_urgc -= n;
351
352	DEBUG_MISC((dfd, " --- sent %d bytes urgent data, %d urgent bytes left\n",
353	n, so->so_urgc));
354	}
355	else
356	{
357	/*
358	* Since there's no sendv or sendtov like writev,
359	* we must copy all data to a linear buffer then
360	* send it all
361	*/
362	len = (sb->sb_data + sb->sb_datalen) - sb->sb_rptr;
363	if (len > so->so_urgc)
364	len = so->so_urgc;
365	memcpy(buff, sb->sb_rptr, len);
366	so->so_urgc -= len;
367	if (so->so_urgc)
368	{
369	n = sb->sb_wptr - sb->sb_data;
370	if (n > so->so_urgc)
371	n = so->so_urgc;
372	memcpy(buff + len, sb->sb_data, n);
373	so->so_urgc -= n;
374	len += n;
375	}
376	n = send(so->s, buff, len, (MSG_OOB)); /* \|MSG_DONTWAIT)); */
377	#ifdef DEBUG
378	if (n != len)
379	DEBUG_ERROR((dfd, "Didn't send all data urgently XXXXX\n"));
380	#endif
381	DEBUG_MISC((dfd, " ---2 sent %d bytes urgent data, %d urgent bytes left\n",
382	n, so->so_urgc));
383	}
384
385	sb->sb_cc -= n;
386	sb->sb_rptr += n;
387	if (sb->sb_rptr >= (sb->sb_data + sb->sb_datalen))
388	sb->sb_rptr -= sb->sb_datalen;
389
390	return n;
391	}
392
393	/*
394	* Write data from so_rcv to so's socket,
395	* updating all sbuf field as necessary
396	*/
397	int
398	sowrite(PNATState pData, struct socket *so)
399	{
400	int n, nn;
401	struct sbuf *sb = &so->so_rcv;
402	size_t len = sb->sb_cc;
403	struct iovec iov[2];
404
405	STAM_PROFILE_START(&pData->StatIOwrite, a);
406	STAM_COUNTER_RESET(&pData->StatIOWrite_in_1);
407	STAM_COUNTER_RESET(&pData->StatIOWrite_in_1_bytes);
408	STAM_COUNTER_RESET(&pData->StatIOWrite_in_2);
409	STAM_COUNTER_RESET(&pData->StatIOWrite_in_2_1st_bytes);
410	STAM_COUNTER_RESET(&pData->StatIOWrite_in_2_2nd_bytes);
411	STAM_COUNTER_RESET(&pData->StatIOWrite_no_w);
412	STAM_COUNTER_RESET(&pData->StatIOWrite_rest);
413	STAM_COUNTER_RESET(&pData->StatIOWrite_rest_bytes);
414	DEBUG_CALL("sowrite");
415	DEBUG_ARG("so = %lx", (long)so);
416	QSOCKET_LOCK(tcb);
417	SOCKET_LOCK(so);
418	QSOCKET_UNLOCK(tcb);
419	if (so->so_urgc)
420	{
421	sosendoob(so);
422	if (sb->sb_cc == 0)
423	{
424	SOCKET_UNLOCK(so);
425	STAM_PROFILE_STOP(&pData->StatIOwrite, a);
426	return 0;
427	}
428	}
429
430	/*
431	* No need to check if there's something to write,
432	* sowrite wouldn't have been called otherwise
433	*/
434
435	len = sb->sb_cc;
436
437	iov[0].iov_base = sb->sb_rptr;
438	iov[1].iov_base = 0;
439	iov[1].iov_len = 0;
440	if (sb->sb_rptr < sb->sb_wptr)
441	{
442	iov[0].iov_len = sb->sb_wptr - sb->sb_rptr;
443	/* Should never succeed, but... */
444	if (iov[0].iov_len > len)
445	iov[0].iov_len = len;
446	n = 1;
447	}
448	else
449	{
450	iov[0].iov_len = (sb->sb_data + sb->sb_datalen) - sb->sb_rptr;
451	if (iov[0].iov_len > len)
452	iov[0].iov_len = len;
453	len -= iov[0].iov_len;
454	if (len)
455	{
456	iov[1].iov_base = sb->sb_data;
457	iov[1].iov_len = sb->sb_wptr - sb->sb_data;
458	if (iov[1].iov_len > len)
459	iov[1].iov_len = len;
460	n = 2;
461	}
462	else
463	n = 1;
464	}
465	STAM_STATS({
466	if (n == 1)
467	{
468	STAM_COUNTER_INC(&pData->StatIOWrite_in_1);
469	STAM_COUNTER_ADD(&pData->StatIOWrite_in_1_bytes, iov[0].iov_len);
470	}
471	else
472	{
473	STAM_COUNTER_INC(&pData->StatIOWrite_in_2);
474	STAM_COUNTER_ADD(&pData->StatIOWrite_in_2_1st_bytes, iov[0].iov_len);
475	STAM_COUNTER_ADD(&pData->StatIOWrite_in_2_2nd_bytes, iov[1].iov_len);
476	}
477	});
478	/* Check if there's urgent data to send, and if so, send it */
479	#ifdef HAVE_READV
480	nn = writev(so->s, (const struct iovec *)iov, n);
481	DEBUG_MISC((dfd, " ... wrote nn = %d bytes\n", nn));
482	#else
483	nn = send(so->s, iov[0].iov_base, iov[0].iov_len, 0);
484	#endif
485	/* This should never happen, but people tell me it does shrug */
486	if ( nn < 0
487	&& ( errno == EAGAIN
488	\|\| errno == EINTR
489	\|\| errno == EWOULDBLOCK))
490	{
491	SOCKET_UNLOCK(so);
492	STAM_PROFILE_STOP(&pData->StatIOwrite, a);
493	return 0;
494	}
495
496	if (nn < 0 \|\| (nn == 0 && iov[0].iov_len > 0))
497	{
498	DEBUG_MISC((dfd, " --- sowrite disconnected, so->so_state = %x, errno = %d\n",
499	so->so_state, errno));
500	sofcantsendmore(so);
501	tcp_sockclosed(pData, sototcpcb(so));
502	SOCKET_UNLOCK(so);
503	STAM_PROFILE_STOP(&pData->StatIOwrite, a);
504	return -1;
505	}
506
507	#ifndef HAVE_READV
508	if (n == 2 && nn == iov[0].iov_len)
509	{
510	int ret;
511	ret = send(so->s, iov[1].iov_base, iov[1].iov_len, 0);
512	if (ret > 0)
513	nn += ret;
514	STAM_STATS({
515	if (ret > 0 && ret != iov[1].iov_len)
516	{
517	STAM_COUNTER_INC(&pData->StatIOWrite_rest);
518	STAM_COUNTER_ADD(&pData->StatIOWrite_rest_bytes, (ret - iov[1].iov_len));
519	}
520	});
521	}
522	DEBUG_MISC((dfd, " ... wrote nn = %d bytes\n", nn));
523	#endif
524
525	/* Update sbuf */
526	sb->sb_cc -= nn;
527	sb->sb_rptr += nn;
528	if (sb->sb_rptr >= (sb->sb_data + sb->sb_datalen))
529	sb->sb_rptr -= sb->sb_datalen;
530
531	/*
532	* If in DRAIN mode, and there's no more data, set
533	* it CANTSENDMORE
534	*/
535	if ((so->so_state & SS_FWDRAIN) && sb->sb_cc == 0)
536	sofcantsendmore(so);
537
538	SOCKET_UNLOCK(so);
539	STAM_PROFILE_STOP(&pData->StatIOwrite, a);
540	return nn;
541	}
542
543	/*
544	* recvfrom() a UDP socket
545	*/
546	void
547	sorecvfrom(PNATState pData, struct socket *so)
548	{
549	ssize_t ret = 0;
550	struct sockaddr_in addr;
551	socklen_t addrlen = sizeof(struct sockaddr_in);
552
553	DEBUG_CALL("sorecvfrom");
554	DEBUG_ARG("so = %lx", (long)so);
555
556	if (so->so_type == IPPROTO_ICMP)
557	{
558	/* This is a "ping" reply */
559	#ifdef RT_OS_WINDOWS
560	sorecvfrom_icmp_win(pData, so);
561	#else /* RT_OS_WINDOWS */
562	sorecvfrom_icmp_unix(pData, so);
563	#endif /* !RT_OS_WINDOWS */
564	udp_detach(pData, so);
565	}
566	else
567	{
568	/* A "normal" UDP packet */
569	struct mbuf *m;
570	ssize_t len;
571	u_long n = 0;
572	#ifdef VBOX_WITH_SLIRP_BSD_MBUF
573	int size;
574	#endif
575	int rc = 0;
576	static int signalled = 0;
577
578	QSOCKET_LOCK(udb);
579	SOCKET_LOCK(so);
580	QSOCKET_UNLOCK(udb);
581
582	#ifndef VBOX_WITH_SLIRP_BSD_MBUF
583	if (!(m = m_get(pData)))
584	{
585	SOCKET_UNLOCK(so);
586	return;
587	}
588	/* adjust both parameters to maks M_FREEROOM calculate correct */
589	m->m_data += if_maxlinkhdr + sizeof(struct udphdr) + sizeof(struct ip);
590
591	/*
592	* XXX Shouldn't FIONREAD packets destined for port 53,
593	* but I don't know the max packet size for DNS lookups
594	*/
595	len = M_FREEROOM(m);
596	/* if (so->so_fport != RT_H2N_U16_C(53)) */
597	rc = ioctlsocket(so->s, FIONREAD, &n);
598	if ( rc == -1
599	&& ( errno == EAGAIN
600	\|\| errno == EWOULDBLOCK
601	\|\| errno == EINPROGRESS
602	\|\| errno == ENOTCONN))
603	{
604	m_free(pData, m);
605	return;
606	}
607
608	Log2(("NAT: %R[natsock] ioctlsocket before read "
609	"(rc:%d errno:%d, n:%d)\n", so, rc, errno, n));
610
611	if (rc == -1 && signalled == 0)
612	{
613	LogRel(("NAT: can't fetch amount of bytes on socket %R[natsock], so message will be truncated.\n", so));
614	signalled = 1;
615	m_free(pData, m);
616	return;
617	}
618
619	if (rc != -1 && n > len)
620	{
621	n = (m->m_data - m->m_dat) + m->m_len + n + 1;
622	m_inc(m, n);
623	len = M_FREEROOM(m);
624	}
625	ret = recvfrom(so->s, m->m_data, len, 0,
626	(struct sockaddr *)&addr, &addrlen);
627	Log2(("NAT: %R[natsock] ioctlsocket after read "
628	"(rc:%d errno:%d, n:%d) ret:%d, len:%d\n", so,
629	rc, errno, n, ret, len));
630	#else
631	/How many data has been received ?/
632	/*
633	* 1. calculate how much we can read
634	* 2. read as much as possible
635	* 3. attach buffer to allocated header mbuf
636	*/
637	rc = ioctlsocket(so->s, FIONREAD, &n);
638	if (rc == -1 && signalled == 0)
639	{
640	LogRel(("NAT: can't fetch amount of bytes on socket %R[natsock], so message will be truncated.\n", so));
641	signalled = 1;
642	}
643
644	len = sizeof(struct udpiphdr) + ETH_HLEN;
645	if (n > (if_mtu - len))
646	{
647	n = if_mtu - len; /* can't read than we can put in the mbuf*/
648	}
649	len += n;
650
651	size = MCLBYTES;
652	if (len < MSIZE)
653	size = MCLBYTES;
654	else if (len < MCLBYTES)
655	size = MCLBYTES;
656	else if (len < MJUM9BYTES)
657	size = MJUM9BYTES;
658	else if (len < MJUM16BYTES)
659	size = MJUM16BYTES;
660	else
661	AssertMsgFailed(("Unsupported size"));
662
663	m = m_getjcl(pData, M_NOWAIT, MT_HEADER, M_PKTHDR, size);
664	m->m_data += ETH_HLEN;
665	m->m_pkthdr.header = mtod(m, void *);
666	m->m_data += sizeof(struct udpiphdr);
667	ret = recvfrom(so->s, mtod(m, char *), n, 0,
668	(struct sockaddr *)&addr, &addrlen);
669	/* @todo (vvl) check which flags and type should be passed */
670	#endif
671	m->m_len = ret;
672	if (ret < 0)
673	{
674	u_char code = ICMP_UNREACH_PORT;
675
676	if (errno == EHOSTUNREACH)
677	code = ICMP_UNREACH_HOST;
678	else if (errno == ENETUNREACH)
679	code = ICMP_UNREACH_NET;
680
681	m_free(pData, m);
682	if ( errno == EAGAIN
683	\|\| errno == EWOULDBLOCK
684	\|\| errno == EINPROGRESS
685	\|\| errno == ENOTCONN)
686	{
687	return;
688	}
689
690	Log2((" rx error, tx icmp ICMP_UNREACH:%i\n", code));
691	icmp_error(pData, so->so_m, ICMP_UNREACH, code, 0, strerror(errno));
692	so->so_m = NULL;
693	}
694	else
695	{
696	/*
697	* Hack: domain name lookup will be used the most for UDP,
698	* and since they'll only be used once there's no need
699	* for the 4 minute (or whatever) timeout... So we time them
700	* out much quicker (10 seconds for now...)
701	*/
702	if (so->so_expire)
703	{
704	if (so->so_fport != RT_H2N_U16_C(53))
705	so->so_expire = curtime + SO_EXPIRE;
706	}
707	/*
708	* last argument should be changed if Slirp will inject IP attributes
709	* Note: Here we can't check if dnsproxy's sent initial request
710	*/
711	#ifndef VBOX_WITH_SLIRP_BSD_MBUF
712	if (so->so_fport == RT_H2N_U16_C(53))
713	dnsproxy_answer(pData, so, m);
714	#endif
715
716	#if 0
717	if (m->m_len == len)
718	{
719	m_inc(m, MINCSIZE);
720	m->m_len = 0;
721	}
722	#endif
723
724	/*
725	* If this packet was destined for CTL_ADDR,
726	* make it look like that's where it came from, done by udp_output
727	*/
728	udp_output(pData, so, m, &addr);
729	SOCKET_UNLOCK(so);
730	} /* rx error */
731	} /* if ping packet */
732	}
733
734	/*
735	* sendto() a socket
736	*/
737	int
738	sosendto(PNATState pData, struct socket so, struct mbuf m)
739	{
740	int ret;
741	struct sockaddr_in *paddr;
742	struct sockaddr addr;
743	#if 0
744	struct sockaddr_in host_addr;
745	#endif
746	#ifdef VBOX_WITH_SLIRP_BSD_MBUF
747	caddr_t buf;
748	int mlen;
749	#endif
750
751	DEBUG_CALL("sosendto");
752	DEBUG_ARG("so = %lx", (long)so);
753	DEBUG_ARG("m = %lx", (long)m);
754
755	memset(&addr, 0, sizeof(struct sockaddr));
756	#ifdef RT_OS_DARWIN
757	addr.sa_len = sizeof(struct sockaddr_in);
758	#endif
759	paddr = (struct sockaddr_in *)&addr;
760	paddr->sin_family = AF_INET;
761	if ((so->so_faddr.s_addr & RT_H2N_U32(pData->netmask)) == pData->special_addr.s_addr)
762	{
763	/* It's an alias */
764	uint32_t last_byte = RT_N2H_U32(so->so_faddr.s_addr) & ~pData->netmask;
765	switch(last_byte)
766	{
767	#if 0
768	/* handle this case at 'default:' */
769	case CTL_BROADCAST:
770	addr.sin_addr.s_addr = INADDR_BROADCAST;
771	/* Send the packet to host to fully emulate broadcast */
772	/** @todo r=klaus: on Linux host this causes the host to receive
773	* the packet twice for some reason. And I cannot find any place
774	* in the man pages which states that sending a broadcast does not
775	* reach the host itself. */
776	host_addr.sin_family = AF_INET;
777	host_addr.sin_port = so->so_fport;
778	host_addr.sin_addr = our_addr;
779	sendto(so->s, m->m_data, m->m_len, 0,
780	(struct sockaddr *)&host_addr, sizeof (struct sockaddr));
781	break;
782	#endif
783	case CTL_DNS:
784	case CTL_ALIAS:
785	default:
786	if (last_byte == ~pData->netmask)
787	paddr->sin_addr.s_addr = INADDR_BROADCAST;
788	else
789	paddr->sin_addr = loopback_addr;
790	break;
791	}
792	}
793	else
794	paddr->sin_addr = so->so_faddr;
795	paddr->sin_port = so->so_fport;
796
797	DEBUG_MISC((dfd, " sendto()ing, addr.sin_port=%d, addr.sin_addr.s_addr=%.16s\n",
798	RT_N2H_U16(paddr->sin_port), inet_ntoa(paddr->sin_addr)));
799
800	/* Don't care what port we get */
801	#ifndef VBOX_WITH_SLIRP_BSD_MBUF
802	ret = sendto(so->s, m->m_data, m->m_len, 0, &addr, sizeof (struct sockaddr_in));
803	#else
804	mlen = m_length(m, NULL);
805	buf = RTMemAlloc(mlen);
806	if (buf == NULL)
807	{
808	return -1;
809	}
810	m_copydata(m, 0, mlen, buf);
811	ret = sendto(so->s, buf, mlen, 0,
812	(struct sockaddr *)&addr, sizeof (struct sockaddr));
813	#endif
814	if (ret < 0)
815	{
816	Log2(("UDP: sendto fails (%s)\n", strerror(errno)));
817	return -1;
818	}
819
820	/*
821	* Kill the socket if there's no reply in 4 minutes,
822	* but only if it's an expirable socket
823	*/
824	if (so->so_expire)
825	so->so_expire = curtime + SO_EXPIRE;
826	so->so_state = SS_ISFCONNECTED; /* So that it gets select()ed */
827	return 0;
828	}
829
830	/*
831	* XXX This should really be tcp_listen
832	*/
833	struct socket *
834	solisten(PNATState pData, u_int32_t bind_addr, u_int port, u_int32_t laddr, u_int lport, int flags)
835	{
836	struct sockaddr_in addr;
837	struct socket *so;
838	socklen_t addrlen = sizeof(addr);
839	int s, opt = 1;
840	int status;
841
842	DEBUG_CALL("solisten");
843	DEBUG_ARG("port = %d", port);
844	DEBUG_ARG("laddr = %x", laddr);
845	DEBUG_ARG("lport = %d", lport);
846	DEBUG_ARG("flags = %x", flags);
847
848	if ((so = socreate()) == NULL)
849	{
850	/* RTMemFree(so); Not sofree() ??? free(NULL) == NOP */
851	return NULL;
852	}
853
854	/* Don't tcp_attach... we don't need so_snd nor so_rcv */
855	if ((so->so_tcpcb = tcp_newtcpcb(pData, so)) == NULL)
856	{
857	RTMemFree(so);
858	return NULL;
859	}
860
861	SOCKET_LOCK_CREATE(so);
862	SOCKET_LOCK(so);
863	QSOCKET_LOCK(tcb);
864	insque(pData, so,&tcb);
865	NSOCK_INC();
866	QSOCKET_UNLOCK(tcb);
867
868	/*
869	* SS_FACCEPTONCE sockets must time out.
870	*/
871	if (flags & SS_FACCEPTONCE)
872	so->so_tcpcb->t_timer[TCPT_KEEP] = TCPTV_KEEP_INIT*2;
873
874	so->so_state = (SS_FACCEPTCONN\|flags);
875	so->so_lport = lport; /* Kept in network format */
876	so->so_laddr.s_addr = laddr; /* Ditto */
877
878	memset(&addr, 0, sizeof(addr));
879	#ifdef RT_OS_DARWIN
880	addr.sin_len = sizeof(addr);
881	#endif
882	addr.sin_family = AF_INET;
883	addr.sin_addr.s_addr = bind_addr;
884	addr.sin_port = port;
885
886	if ( ((s = socket(AF_INET, SOCK_STREAM, 0)) < 0)
887	\|\| (setsockopt(s, SOL_SOCKET, SO_REUSEADDR,(char *)&opt, sizeof(int)) < 0)
888	\|\| (bind(s,(struct sockaddr *)&addr, sizeof(addr)) < 0)
889	\|\| (listen(s, 1) < 0))
890	{
891	#ifdef RT_OS_WINDOWS
892	int tmperrno = WSAGetLastError(); /* Don't clobber the real reason we failed */
893	closesocket(s);
894	QSOCKET_LOCK(tcb);
895	sofree(pData, so);
896	QSOCKET_UNLOCK(tcb);
897	/* Restore the real errno */
898	WSASetLastError(tmperrno);
899	#else
900	int tmperrno = errno; /* Don't clobber the real reason we failed */
901	close(s);
902	QSOCKET_LOCK(tcb);
903	sofree(pData, so);
904	QSOCKET_UNLOCK(tcb);
905	/* Restore the real errno */
906	errno = tmperrno;
907	#endif
908	return NULL;
909	}
910	fd_nonblock(s);
911	setsockopt(s, SOL_SOCKET, SO_OOBINLINE,(char *)&opt, sizeof(int));
912
913	getsockname(s,(struct sockaddr *)&addr,&addrlen);
914	so->so_fport = addr.sin_port;
915	/* set socket buffers */
916	opt = pData->socket_rcv;
917	status = setsockopt(s, SOL_SOCKET, SO_RCVBUF, (char *)&opt, sizeof(int));
918	if (status < 0)
919	{
920	LogRel(("NAT: Error(%d) while setting RCV capacity to (%d)\n", errno, opt));
921	goto no_sockopt;
922	}
923	opt = pData->socket_snd;
924	status = setsockopt(s, SOL_SOCKET, SO_SNDBUF, (char *)&opt, sizeof(int));
925	if (status < 0)
926	{
927	LogRel(("NAT: Error(%d) while setting SND capacity to (%d)\n", errno, opt));
928	goto no_sockopt;
929	}
930	no_sockopt:
931	if (addr.sin_addr.s_addr == 0 \|\| addr.sin_addr.s_addr == loopback_addr.s_addr)
932	so->so_faddr = alias_addr;
933	else
934	so->so_faddr = addr.sin_addr;
935
936	so->s = s;
937	SOCKET_UNLOCK(so);
938	return so;
939	}
940
941	/*
942	* Data is available in so_rcv
943	* Just write() the data to the socket
944	* XXX not yet...
945	*/
946	void
947	sorwakeup(struct socket *so)
948	{
949	#if 0
950	sowrite(so);
951	FD_CLR(so->s,&writefds);
952	#endif
953	}
954
955	/*
956	* Data has been freed in so_snd
957	* We have room for a read() if we want to
958	* For now, don't read, it'll be done in the main loop
959	*/
960	void
961	sowwakeup(struct socket *so)
962	{
963	}
964
965	/*
966	* Various session state calls
967	* XXX Should be #define's
968	* The socket state stuff needs work, these often get call 2 or 3
969	* times each when only 1 was needed
970	*/
971	void
972	soisfconnecting(struct socket *so)
973	{
974	so->so_state &= ~(SS_NOFDREF\|SS_ISFCONNECTED\|SS_FCANTRCVMORE\|
975	SS_FCANTSENDMORE\|SS_FWDRAIN);
976	so->so_state \|= SS_ISFCONNECTING; /* Clobber other states */
977	}
978
979	void
980	soisfconnected(struct socket *so)
981	{
982	so->so_state &= ~(SS_ISFCONNECTING\|SS_FWDRAIN\|SS_NOFDREF);
983	so->so_state \|= SS_ISFCONNECTED; /* Clobber other states */
984	}
985
986	void
987	sofcantrcvmore(struct socket *so)
988	{
989	if ((so->so_state & SS_NOFDREF) == 0)
990	{
991	shutdown(so->s, 0);
992	}
993	so->so_state &= ~(SS_ISFCONNECTING);
994	if (so->so_state & SS_FCANTSENDMORE)
995	so->so_state = SS_NOFDREF; /* Don't select it */
996	/* XXX close() here as well? */
997	else
998	so->so_state \|= SS_FCANTRCVMORE;
999	}
1000
1001	void
1002	sofcantsendmore(struct socket *so)
1003	{
1004	if ((so->so_state & SS_NOFDREF) == 0)
1005	shutdown(so->s, 1); /* send FIN to fhost */
1006
1007	so->so_state &= ~(SS_ISFCONNECTING);
1008	if (so->so_state & SS_FCANTRCVMORE)
1009	so->so_state = SS_NOFDREF; /* as above */
1010	else
1011	so->so_state \|= SS_FCANTSENDMORE;
1012	}
1013
1014	void
1015	soisfdisconnected(struct socket *so)
1016	{
1017	#if 0
1018	so->so_state &= ~(SS_ISFCONNECTING\|SS_ISFCONNECTED);
1019	close(so->s);
1020	so->so_state = SS_ISFDISCONNECTED;
1021	/*
1022	* XXX Do nothing ... ?
1023	*/
1024	#endif
1025	}
1026
1027	/*
1028	* Set write drain mode
1029	* Set CANTSENDMORE once all data has been write()n
1030	*/
1031	void
1032	sofwdrain(struct socket *so)
1033	{
1034	if (so->so_rcv.sb_cc)
1035	so->so_state \|= SS_FWDRAIN;
1036	else
1037	sofcantsendmore(so);
1038	}
1039
1040	static void
1041	send_icmp_to_guest(PNATState pData, char buff, size_t len, struct socket so, const struct sockaddr_in *addr)
1042	{
1043	struct ip *ip;
1044	uint32_t dst, src;
1045	char ip_copy[256];
1046	struct icmp *icp;
1047	int old_ip_len = 0;
1048	int hlen, original_hlen = 0;
1049	struct mbuf *m;
1050	struct icmp_msg *icm;
1051	uint8_t proto;
1052	int type = 0;
1053	int m_room;
1054
1055	ip = (struct ip *)buff;
1056	/* Fix ip->ip_len to contain the total packet length including the header
1057	* in _host_ byte order for all OSes. On Darwin, that value already is in
1058	* host byte order. Solaris and Darwin report only the payload. */
1059	#ifndef RT_OS_DARWIN
1060	ip->ip_len = RT_N2H_U16(ip->ip_len);
1061	#endif
1062	hlen = (ip->ip_hl << 2);
1063	#if defined(RT_OS_SOLARIS) \|\| defined(RT_OS_DARWIN)
1064	ip->ip_len += hlen;
1065	#endif
1066	if (ip->ip_len < hlen + ICMP_MINLEN)
1067	{
1068	Log(("send_icmp_to_guest: ICMP header is too small to understand which type/subtype of the datagram\n"));
1069	return;
1070	}
1071	icp = (struct icmp )((char )ip + hlen);
1072
1073	Log(("ICMP:received msg(t:%d, c:%d)\n", icp->icmp_type, icp->icmp_code));
1074	if ( icp->icmp_type != ICMP_ECHOREPLY
1075	&& icp->icmp_type != ICMP_TIMXCEED
1076	&& icp->icmp_type != ICMP_UNREACH)
1077	{
1078	return;
1079	}
1080
1081	/*
1082	* ICMP_ECHOREPLY, ICMP_TIMXCEED, ICMP_UNREACH minimal header size is
1083	* ICMP_ECHOREPLY assuming data 0
1084	* icmp_{type(8), code(8), cksum(16),identifier(16),seqnum(16)}
1085	*/
1086	if (ip->ip_len < hlen + 8)
1087	{
1088	Log(("send_icmp_to_guest: NAT accept ICMP_{ECHOREPLY, TIMXCEED, UNREACH} the minimum size is 64 (see rfc792)\n"));
1089	return;
1090	}
1091
1092	type = icp->icmp_type;
1093	if ( type == ICMP_TIMXCEED
1094	\|\| type == ICMP_UNREACH)
1095	{
1096	/*
1097	* ICMP_TIMXCEED, ICMP_UNREACH minimal header size is
1098	* icmp_{type(8), code(8), cksum(16),unused(32)} + IP header + 64 bit of original datagram
1099	*/
1100	if (ip->ip_len < hlen + 2*8 + sizeof(struct ip))
1101	{
1102	Log(("send_icmp_to_guest: NAT accept ICMP_{TIMXCEED, UNREACH} the minimum size of ipheader + 64 bit of data (see rfc792)\n"));
1103	return;
1104	}
1105	ip = &icp->icmp_ip;
1106	}
1107
1108	icm = icmp_find_original_mbuf(pData, ip);
1109	if (icm == NULL)
1110	{
1111	Log(("NAT: Can't find the corresponding packet for the received ICMP\n"));
1112	return;
1113	}
1114
1115	m = icm->im_m;
1116	Assert(m != NULL);
1117
1118	src = addr->sin_addr.s_addr;
1119	if (type == ICMP_ECHOREPLY)
1120	{
1121	struct ip ip0 = mtod(m, struct ip );
1122	struct icmp icp0 = (struct icmp )((char *)ip0 + (ip0->ip_hl << 2));
1123	if (icp0->icmp_type != ICMP_ECHO)
1124	{
1125	Log(("NAT: we haven't found echo for this reply\n"));
1126	return;
1127	}
1128	/*
1129	* while combining buffer to send (see ip_icmp.c) we control ICMP header only,
1130	* IP header combined by OS network stack, our local copy of IP header contians values
1131	* in host byte order so no byte order conversion is required. IP headers fields are converting
1132	* in ip_output0 routine only.
1133	*/
1134	if ( (ip->ip_len - hlen)
1135	!= (ip0->ip_len - (ip0->ip_hl << 2)))
1136	{
1137	Log(("NAT: ECHO(%d) lenght doesn't match ECHOREPLY(%d)\n",
1138	(ip->ip_len - hlen), (ip0->ip_len - (ip0->ip_hl << 2))));
1139	return;
1140	}
1141	}
1142
1143	/* ip points on origianal ip header */
1144	ip = mtod(m, struct ip *);
1145	proto = ip->ip_p;
1146	/* Now ip is pointing on header we've sent from guest */
1147	if ( icp->icmp_type == ICMP_TIMXCEED
1148	\|\| icp->icmp_type == ICMP_UNREACH)
1149	{
1150	old_ip_len = (ip->ip_hl << 2) + 64;
1151	if (old_ip_len > sizeof(ip_copy))
1152	old_ip_len = sizeof(ip_copy);
1153	memcpy(ip_copy, ip, old_ip_len);
1154	}
1155
1156	/* source address from original IP packet*/
1157	dst = ip->ip_src.s_addr;
1158
1159	/* overide ther tail of old packet */
1160	ip = mtod(m, struct ip ); / ip is from mbuf we've overrided */
1161	original_hlen = ip->ip_hl << 2;
1162	/* saves original ip header and options */
1163	#ifdef VBOX_WITH_SLIRP_BSD_MBUF
1164	m_copyback(pData, m, original_hlen, len - hlen, buff + hlen);
1165	ip->ip_len = m_length(m, NULL);
1166	#else
1167	/* m_room space in the saved m buffer */
1168	m_room = M_ROOM(m);
1169	if (m_room < len - hlen + original_hlen)
1170	{
1171	/* we need involve ether header length into new buffer buffer calculation */
1172	m_inc(m, if_maxlinkhdr + len - hlen + original_hlen);
1173	if (m->m_size < if_maxlinkhdr + len - hlen + original_hlen)
1174	{
1175	Log(("send_icmp_to_guest: extending buffer was failed (packet is dropped)\n"));
1176	return;
1177	}
1178	}
1179	memcpy(m->m_data + original_hlen, buff + hlen, len - hlen);
1180	m->m_len = len - hlen + original_hlen;
1181	ip->ip_len = m->m_len;
1182	#endif
1183	ip->ip_p = IPPROTO_ICMP; /* the original package could be whatever, but we're response via ICMP*/
1184
1185	icp = (struct icmp )((char )ip + (ip->ip_hl << 2));
1186	type = icp->icmp_type;
1187	if ( type == ICMP_TIMXCEED
1188	\|\| type == ICMP_UNREACH)
1189	{
1190	/* according RFC 793 error messages required copy of initial IP header + 64 bit */
1191	memcpy(&icp->icmp_ip, ip_copy, old_ip_len);
1192	ip->ip_tos = ((ip->ip_tos & 0x1E) \| 0xC0); /* high priority for errors */
1193	}
1194
1195	ip->ip_src.s_addr = src;
1196	ip->ip_dst.s_addr = dst;
1197	icmp_reflect(pData, m);
1198	LIST_REMOVE(icm, im_list);
1199	/* Don't call m_free here*/
1200
1201	if ( type == ICMP_TIMXCEED
1202	\|\| type == ICMP_UNREACH)
1203	{
1204	icm->im_so->so_m = NULL;
1205	switch (proto)
1206	{
1207	case IPPROTO_UDP:
1208	/XXX: so->so_m already freed so we shouldn't call sofree /
1209	udp_detach(pData, icm->im_so);
1210	break;
1211	case IPPROTO_TCP:
1212	/close tcp should be here /
1213	break;
1214	default:
1215	/* do nothing */
1216	break;
1217	}
1218	}
1219	RTMemFree(icm);
1220	}
1221
1222	#ifdef RT_OS_WINDOWS
1223	static void
1224	sorecvfrom_icmp_win(PNATState pData, struct socket *so)
1225	{
1226	int len;
1227	int i;
1228	struct ip *ip;
1229	struct mbuf *m;
1230	struct icmp *icp;
1231	struct icmp_msg *icm;
1232	struct ip ip_broken; / ICMP returns header + 64 bit of packet */
1233	uint32_t src;
1234	ICMP_ECHO_REPLY *icr;
1235	int hlen = 0;
1236	int data_len = 0;
1237	int nbytes = 0;
1238	u_char code = ~0;
1239
1240	len = pData->pfIcmpParseReplies(pData->pvIcmpBuffer, pData->szIcmpBuffer);
1241	if (len < 0)
1242	{
1243	LogRel(("NAT: Error (%d) occurred on ICMP receiving\n", GetLastError()));
1244	return;
1245	}
1246	if (len == 0)
1247	return; /* no error */
1248
1249	icr = (ICMP_ECHO_REPLY *)pData->pvIcmpBuffer;
1250	for (i = 0; i < len; ++i)
1251	{
1252	switch(icr[i].Status)
1253	{
1254	case IP_DEST_HOST_UNREACHABLE:
1255	code = (code != ~0 ? code : ICMP_UNREACH_HOST);
1256	case IP_DEST_NET_UNREACHABLE:
1257	code = (code != ~0 ? code : ICMP_UNREACH_NET);
1258	case IP_DEST_PROT_UNREACHABLE:
1259	code = (code != ~0 ? code : ICMP_UNREACH_PROTOCOL);
1260	/* UNREACH error inject here */
1261	case IP_DEST_PORT_UNREACHABLE:
1262	code = (code != ~0 ? code : ICMP_UNREACH_PORT);
1263	icmp_error(pData, so->so_m, ICMP_UNREACH, code, 0, "Error occurred!!!");
1264	so->so_m = NULL;
1265	break;
1266	case IP_SUCCESS: /* echo replied */
1267	# ifndef VBOX_WITH_SLIRP_BSD_MBUF
1268	m = m_get(pData);
1269	# else
1270	m = m_gethdr(pData, M_NOWAIT, MT_HEADER);
1271	# endif
1272	m->m_data += if_maxlinkhdr;
1273	ip = mtod(m, struct ip *);
1274	ip->ip_src.s_addr = icr[i].Address;
1275	ip->ip_p = IPPROTO_ICMP;
1276	ip->ip_dst.s_addr = so->so_laddr.s_addr; /XXX: still the hack/
1277	data_len = sizeof(struct ip);
1278	ip->ip_hl = data_len >> 2; /* requiered for icmp_reflect, no IP options */
1279	ip->ip_ttl = icr[i].Options.Ttl;
1280
1281	icp = (struct icmp )&ip[1]; / no options */
1282	icp->icmp_type = ICMP_ECHOREPLY;
1283	icp->icmp_code = 0;
1284	icp->icmp_id = so->so_icmp_id;
1285	icp->icmp_seq = so->so_icmp_seq;
1286
1287	data_len += ICMP_MINLEN;
1288
1289	# ifndef VBOX_WITH_SLIRP_BSD_MBUF
1290	nbytes = (data_len + icr[i].DataSize > m->m_size? m->m_size - data_len: icr[i].DataSize);
1291	memcpy(icp->icmp_data, icr[i].Data, nbytes);
1292	# else
1293	AssertMsgFailed(("ICMP"));
1294	# endif
1295
1296	data_len += icr[i].DataSize;
1297
1298	ip->ip_len = data_len;
1299	m->m_len = ip->ip_len;
1300
1301	icmp_reflect(pData, m);
1302	break;
1303	case IP_TTL_EXPIRED_TRANSIT: /* TTL expired */
1304
1305	ip_broken = icr[i].Data;
1306	icm = icmp_find_original_mbuf(pData, ip_broken);
1307	if (icm == NULL) {
1308	Log(("ICMP: can't find original package (first double word %x)\n", (uint32_t )ip_broken));
1309	return;
1310	}
1311	m = icm->im_m;
1312	ip = mtod(m, struct ip *);
1313	ip->ip_ttl = icr[i].Options.Ttl;
1314	src = ip->ip_src.s_addr;
1315	ip->ip_dst.s_addr = src;
1316	ip->ip_dst.s_addr = icr[i].Address;
1317
1318	hlen = (ip->ip_hl << 2);
1319	icp = (struct icmp )((char )ip + hlen);
1320	ip_broken->ip_src.s_addr = src; /it packet sent from host not from guest/
1321	data_len = (ip_broken->ip_hl << 2) + 64;
1322
1323	#ifndef VBOX_WITH_SLIRP_BSD_MBUF
1324	nbytes =(hlen + ICMP_MINLEN + data_len > m->m_size? m->m_size - (hlen + ICMP_MINLEN): data_len);
1325	memcpy(icp->icmp_data, ip_broken, nbytes);
1326	#else
1327	AssertMsgFailed(("ICMP"));
1328	#endif
1329	icmp_reflect(pData, m);
1330	break;
1331	default:
1332	Log(("ICMP(default): message with Status: %x was received from %x\n", icr[i].Status, icr[i].Address));
1333	break;
1334	}
1335	}
1336	}
1337	#else /* !RT_OS_WINDOWS */
1338	static void sorecvfrom_icmp_unix(PNATState pData, struct socket *so)
1339	{
1340	struct sockaddr_in addr;
1341	socklen_t addrlen = sizeof(struct sockaddr_in);
1342	struct ip ip;
1343	char *buff;
1344	int len = 0;
1345
1346	/* 1- step: read the ip header */
1347	len = recvfrom(so->s, &ip, sizeof(struct ip), MSG_PEEK,
1348	(struct sockaddr *)&addr, &addrlen);
1349	if ( len < 0
1350	&& ( errno == EAGAIN
1351	\|\| errno == EWOULDBLOCK
1352	\|\| errno == EINPROGRESS
1353	\|\| errno == ENOTCONN))
1354	{
1355	Log(("sorecvfrom_icmp_unix: 1 - step can't read IP datagramm (would block)\n"));
1356	return;
1357	}
1358
1359	if ( len < sizeof(struct ip)
1360	\|\| len < 0
1361	\|\| len == 0)
1362	{
1363	u_char code;
1364	code = ICMP_UNREACH_PORT;
1365
1366	if (errno == EHOSTUNREACH)
1367	code = ICMP_UNREACH_HOST;
1368	else if (errno == ENETUNREACH)
1369	code = ICMP_UNREACH_NET;
1370
1371	LogRel((" udp icmp rx errno = %d-%s\n",
1372	errno, strerror(errno)));
1373	icmp_error(pData, so->so_m, ICMP_UNREACH, code, 0, strerror(errno));
1374	so->so_m = NULL;
1375	Log(("sorecvfrom_icmp_unix: 1 - step can't read IP datagramm \n"));
1376	return;
1377	}
1378	/* basic check of IP header */
1379	if ( ip.ip_v != IPVERSION
1380	# ifndef RT_OS_DARWIN
1381	\|\| ip.ip_p != IPPROTO_ICMP
1382	# endif
1383	)
1384	{
1385	Log(("sorecvfrom_icmp_unix: 1 - step IP isn't IPv4 \n"));
1386	return;
1387	}
1388	# ifndef RT_OS_DARWIN
1389	/* Darwin reports the IP length already in host byte order. */
1390	ip.ip_len = RT_N2H_U16(ip.ip_len);
1391	# endif
1392	# if defined(RT_OS_SOLARIS) \|\| defined(RT_OS_DARWIN)
1393	/* Solaris and Darwin report the payload only */
1394	ip.ip_len += (ip.ip_hl << 2);
1395	# endif
1396	/* Note: ip->ip_len in host byte order (all OS) */
1397	len = ip.ip_len;
1398	buff = RTMemAlloc(len);
1399	if (buff == NULL)
1400	{
1401	Log(("sorecvfrom_icmp_unix: 1 - step can't allocate enought room for datagram\n"));
1402	return;
1403	}
1404	/* 2 - step: we're reading rest of the datagramm to the buffer */
1405	addrlen = sizeof(struct sockaddr_in);
1406	memset(&addr, 0, addrlen);
1407	len = recvfrom(so->s, buff, len, 0,
1408	(struct sockaddr *)&addr, &addrlen);
1409	if ( len < 0
1410	&& ( errno == EAGAIN
1411	\|\| errno == EWOULDBLOCK
1412	\|\| errno == EINPROGRESS
1413	\|\| errno == ENOTCONN))
1414	{
1415	Log(("sorecvfrom_icmp_unix: 2 - step can't read IP body (would block expected:%d)\n",
1416	ip.ip_len));
1417	RTMemFree(buff);
1418	return;
1419	}
1420	if ( len < 0
1421	\|\| len == 0)
1422	{
1423	Log(("sorecvfrom_icmp_unix: 2 - step read of the rest of datagramm is fallen (errno:%d, len:%d expected: %d)\n",
1424	errno, len, (ip.ip_len - sizeof(struct ip))));
1425	RTMemFree(buff);
1426	return;
1427	}
1428	/* len is modified in 2nd read, when the rest of the datagramm was read */
1429	send_icmp_to_guest(pData, buff, len, so, &addr);
1430	RTMemFree(buff);
1431	}
1432	#endif /* !RT_OS_WINDOWS */

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

source: vbox/trunk/src/VBox/Devices/Network/slirp/socket.c@ 27976

Download in other formats: