socket.c@ 28915

Last change on this file since 28915 was 28800, checked in by vboxsync, 14 years ago
Automated rebranding to Oracle copyright/license strings via filemuncher
Property svn:eol-style set to `native` Property svn:keywords set to `Author Date Id Revision`
File size: 42.9 KB

Line
1	/* $Id: socket.c 28800 2010-04-27 08:22:32Z vboxsync $ */
2	/** @file
3	* NAT - socket handling.
4	*/
5
6	/*
7	* Copyright (C) 2006-2010 Oracle Corporation
8	*
9	* This file is part of VirtualBox Open Source Edition (OSE), as
10	* available from http://www.virtualbox.org. This file is free software;
11	* you can redistribute it and/or modify it under the terms of the GNU
12	* General Public License (GPL) as published by the Free Software
13	* Foundation, in version 2 as it comes in the "COPYING" file of the
14	* VirtualBox OSE distribution. VirtualBox OSE is distributed in the
15	* hope that it will be useful, but WITHOUT ANY WARRANTY of any kind.
16	*/
17
18	/*
19	* This code is based on:
20	*
21	* Copyright (c) 1995 Danny Gasparovski.
22	*
23	* Please read the file COPYRIGHT for the
24	* terms and conditions of the copyright.
25	*/
26
27	#define WANT_SYS_IOCTL_H
28	#include <slirp.h>
29	#include "ip_icmp.h"
30	#include "main.h"
31	#ifdef __sun__
32	#include <sys/filio.h>
33	#endif
34	#include <VBox/pdmdrv.h>
35	#if defined (RT_OS_WINDOWS)
36	#include <iphlpapi.h>
37	#include <icmpapi.h>
38	#endif
39
40
41	static void send_icmp_to_guest(PNATState, char , size_t, struct socket , const struct sockaddr_in *);
42	#ifdef RT_OS_WINDOWS
43	static void sorecvfrom_icmp_win(PNATState, struct socket *);
44	#else /* RT_OS_WINDOWS */
45	static void sorecvfrom_icmp_unix(PNATState, struct socket *);
46	#endif /* !RT_OS_WINDOWS */
47
48	void
49	so_init()
50	{
51	}
52
53	struct socket *
54	solookup(struct socket *head, struct in_addr laddr,
55	u_int lport, struct in_addr faddr, u_int fport)
56	{
57	struct socket *so;
58
59	for (so = head->so_next; so != head; so = so->so_next)
60	{
61	if ( so->so_lport == lport
62	&& so->so_laddr.s_addr == laddr.s_addr
63	&& so->so_faddr.s_addr == faddr.s_addr
64	&& so->so_fport == fport)
65	return so;
66	}
67
68	return (struct socket *)NULL;
69	}
70
71	/*
72	* Create a new socket, initialise the fields
73	* It is the responsibility of the caller to
74	* insque() it into the correct linked-list
75	*/
76	struct socket *
77	socreate()
78	{
79	struct socket *so;
80
81	so = (struct socket *)RTMemAllocZ(sizeof(struct socket));
82	if (so)
83	{
84	so->so_state = SS_NOFDREF;
85	so->s = -1;
86	#if !defined(RT_OS_WINDOWS)
87	so->so_poll_index = -1;
88	#endif
89	}
90	return so;
91	}
92
93	/*
94	* remque and free a socket, clobber cache
95	* VBOX_WITH_SLIRP_MT: before sofree queue should be locked, because
96	* in sofree we don't know from which queue item beeing removed.
97	*/
98	void
99	sofree(PNATState pData, struct socket *so)
100	{
101	struct socket *so_prev = NULL;
102	if (so == tcp_last_so)
103	tcp_last_so = &tcb;
104	else if (so == udp_last_so)
105	udp_last_so = &udb;
106
107	/* check if mbuf haven't been already freed */
108	if (so->so_m != NULL)
109	m_freem(pData, so->so_m);
110	#ifndef VBOX_WITH_SLIRP_MT
111	if (so->so_next && so->so_prev)
112	{
113	remque(pData, so); /* crashes if so is not in a queue */
114	NSOCK_DEC();
115	}
116
117	RTMemFree(so);
118	#else
119	so->so_deleted = 1;
120	#endif
121	}
122
123	#ifdef VBOX_WITH_SLIRP_MT
124	void
125	soread_queue(PNATState pData, struct socket so, int ret)
126	{
127	*ret = soread(pData, so);
128	}
129	#endif
130
131	/*
132	* Read from so's socket into sb_snd, updating all relevant sbuf fields
133	* NOTE: This will only be called if it is select()ed for reading, so
134	* a read() of 0 (or less) means it's disconnected
135	*/
136	int
137	soread(PNATState pData, struct socket *so)
138	{
139	int n, nn, lss, total;
140	struct sbuf *sb = &so->so_snd;
141	size_t len = sb->sb_datalen - sb->sb_cc;
142	struct iovec iov[2];
143	int mss = so->so_tcpcb->t_maxseg;
144
145	STAM_PROFILE_START(&pData->StatIOread, a);
146	STAM_COUNTER_RESET(&pData->StatIORead_in_1);
147	STAM_COUNTER_RESET(&pData->StatIORead_in_2);
148
149	QSOCKET_LOCK(tcb);
150	SOCKET_LOCK(so);
151	QSOCKET_UNLOCK(tcb);
152
153	DEBUG_CALL("soread");
154	DEBUG_ARG("so = %lx", (long)so);
155
156	/*
157	* No need to check if there's enough room to read.
158	* soread wouldn't have been called if there weren't
159	*/
160
161	len = sb->sb_datalen - sb->sb_cc;
162
163	iov[0].iov_base = sb->sb_wptr;
164	iov[1].iov_base = 0;
165	iov[1].iov_len = 0;
166	if (sb->sb_wptr < sb->sb_rptr)
167	{
168	iov[0].iov_len = sb->sb_rptr - sb->sb_wptr;
169	/* Should never succeed, but... */
170	if (iov[0].iov_len > len)
171	iov[0].iov_len = len;
172	if (iov[0].iov_len > mss)
173	iov[0].iov_len -= iov[0].iov_len%mss;
174	n = 1;
175	}
176	else
177	{
178	iov[0].iov_len = (sb->sb_data + sb->sb_datalen) - sb->sb_wptr;
179	/* Should never succeed, but... */
180	if (iov[0].iov_len > len)
181	iov[0].iov_len = len;
182	len -= iov[0].iov_len;
183	if (len)
184	{
185	iov[1].iov_base = sb->sb_data;
186	iov[1].iov_len = sb->sb_rptr - sb->sb_data;
187	if (iov[1].iov_len > len)
188	iov[1].iov_len = len;
189	total = iov[0].iov_len + iov[1].iov_len;
190	if (total > mss)
191	{
192	lss = total % mss;
193	if (iov[1].iov_len > lss)
194	{
195	iov[1].iov_len -= lss;
196	n = 2;
197	}
198	else
199	{
200	lss -= iov[1].iov_len;
201	iov[0].iov_len -= lss;
202	n = 1;
203	}
204	}
205	else
206	n = 2;
207	}
208	else
209	{
210	if (iov[0].iov_len > mss)
211	iov[0].iov_len -= iov[0].iov_len%mss;
212	n = 1;
213	}
214	}
215
216	#ifdef HAVE_READV
217	nn = readv(so->s, (struct iovec *)iov, n);
218	DEBUG_MISC((dfd, " ... read nn = %d bytes\n", nn));
219	#else
220	nn = recv(so->s, iov[0].iov_base, iov[0].iov_len, (so->so_tcpcb->t_force? MSG_OOB:0));
221	#endif
222	if (nn <= 0)
223	{
224	/*
225	* Special case for WSAEnumNetworkEvents: If we receive 0 bytes that
226	* _could_ mean that the connection is closed. But we will receive an
227	* FD_CLOSE event later if the connection was _really_ closed. With
228	* www.youtube.com I see this very often. Closing the socket too early
229	* would be dangerous.
230	*/
231	int status;
232	unsigned long pending = 0;
233	status = ioctlsocket(so->s, FIONREAD, &pending);
234	if (status < 0)
235	LogRel(("NAT:error in WSAIoctl: %d\n", errno));
236	if (nn == 0 && (pending != 0))
237	{
238	SOCKET_UNLOCK(so);
239	STAM_PROFILE_STOP(&pData->StatIOread, a);
240	return 0;
241	}
242	if ( nn < 0
243	&& ( errno == EINTR
244	\|\| errno == EAGAIN
245	\|\| errno == EWOULDBLOCK))
246	{
247	SOCKET_UNLOCK(so);
248	STAM_PROFILE_STOP(&pData->StatIOread, a);
249	return 0;
250	}
251	else
252	{
253	/* nn == 0 means peer has performed an orderly shutdown */
254	DEBUG_MISC((dfd, " --- soread() disconnected, nn = %d, errno = %d-%s\n",
255	nn, errno, strerror(errno)));
256	sofcantrcvmore(so);
257	tcp_sockclosed(pData, sototcpcb(so));
258	SOCKET_UNLOCK(so);
259	STAM_PROFILE_STOP(&pData->StatIOread, a);
260	return -1;
261	}
262	}
263	STAM_STATS(
264	if (n == 1)
265	{
266	STAM_COUNTER_INC(&pData->StatIORead_in_1);
267	STAM_COUNTER_ADD(&pData->StatIORead_in_1_bytes, nn);
268	}
269	else
270	{
271	STAM_COUNTER_INC(&pData->StatIORead_in_2);
272	STAM_COUNTER_ADD(&pData->StatIORead_in_2_1st_bytes, nn);
273	}
274	);
275
276	#ifndef HAVE_READV
277	/*
278	* If there was no error, try and read the second time round
279	* We read again if n = 2 (ie, there's another part of the buffer)
280	* and we read as much as we could in the first read
281	* We don't test for <= 0 this time, because there legitimately
282	* might not be any more data (since the socket is non-blocking),
283	* a close will be detected on next iteration.
284	* A return of -1 wont (shouldn't) happen, since it didn't happen above
285	*/
286	if (n == 2 && nn == iov[0].iov_len)
287	{
288	int ret;
289	ret = recv(so->s, iov[1].iov_base, iov[1].iov_len, 0);
290	if (ret > 0)
291	nn += ret;
292	STAM_STATS(
293	if (ret > 0)
294	{
295	STAM_COUNTER_INC(&pData->StatIORead_in_2);
296	STAM_COUNTER_ADD(&pData->StatIORead_in_2_2nd_bytes, ret);
297	}
298	);
299	}
300
301	DEBUG_MISC((dfd, " ... read nn = %d bytes\n", nn));
302	#endif
303
304	/* Update fields */
305	sb->sb_cc += nn;
306	sb->sb_wptr += nn;
307	if (sb->sb_wptr >= (sb->sb_data + sb->sb_datalen))
308	sb->sb_wptr -= sb->sb_datalen;
309	STAM_PROFILE_STOP(&pData->StatIOread, a);
310	SOCKET_UNLOCK(so);
311	return nn;
312	}
313
314	/*
315	* Get urgent data
316	*
317	* When the socket is created, we set it SO_OOBINLINE,
318	* so when OOB data arrives, we soread() it and everything
319	* in the send buffer is sent as urgent data
320	*/
321	void
322	sorecvoob(PNATState pData, struct socket *so)
323	{
324	struct tcpcb *tp = sototcpcb(so);
325	ssize_t ret;
326
327	DEBUG_CALL("sorecvoob");
328	DEBUG_ARG("so = %lx", (long)so);
329
330	/*
331	* We take a guess at how much urgent data has arrived.
332	* In most situations, when urgent data arrives, the next
333	* read() should get all the urgent data. This guess will
334	* be wrong however if more data arrives just after the
335	* urgent data, or the read() doesn't return all the
336	* urgent data.
337	*/
338	ret = soread(pData, so);
339	tp->snd_up = tp->snd_una + so->so_snd.sb_cc;
340	tp->t_force = 1;
341	tcp_output(pData, tp);
342	tp->t_force = 0;
343	}
344
345	/*
346	* Send urgent data
347	* There's a lot duplicated code here, but...
348	*/
349	int
350	sosendoob(struct socket *so)
351	{
352	struct sbuf *sb = &so->so_rcv;
353	char buff[2048]; /* XXX Shouldn't be sending more oob data than this */
354
355	int n, len;
356
357	DEBUG_CALL("sosendoob");
358	DEBUG_ARG("so = %lx", (long)so);
359	DEBUG_ARG("sb->sb_cc = %d", sb->sb_cc);
360
361	if (so->so_urgc > sizeof(buff))
362	so->so_urgc = sizeof(buff); /* XXX */
363
364	if (sb->sb_rptr < sb->sb_wptr)
365	{
366	/* We can send it directly */
367	n = send(so->s, sb->sb_rptr, so->so_urgc, (MSG_OOB)); /* \|MSG_DONTWAIT)); */
368	so->so_urgc -= n;
369
370	DEBUG_MISC((dfd, " --- sent %d bytes urgent data, %d urgent bytes left\n",
371	n, so->so_urgc));
372	}
373	else
374	{
375	/*
376	* Since there's no sendv or sendtov like writev,
377	* we must copy all data to a linear buffer then
378	* send it all
379	*/
380	len = (sb->sb_data + sb->sb_datalen) - sb->sb_rptr;
381	if (len > so->so_urgc)
382	len = so->so_urgc;
383	memcpy(buff, sb->sb_rptr, len);
384	so->so_urgc -= len;
385	if (so->so_urgc)
386	{
387	n = sb->sb_wptr - sb->sb_data;
388	if (n > so->so_urgc)
389	n = so->so_urgc;
390	memcpy(buff + len, sb->sb_data, n);
391	so->so_urgc -= n;
392	len += n;
393	}
394	n = send(so->s, buff, len, (MSG_OOB)); /* \|MSG_DONTWAIT)); */
395	#ifdef DEBUG
396	if (n != len)
397	DEBUG_ERROR((dfd, "Didn't send all data urgently XXXXX\n"));
398	#endif
399	DEBUG_MISC((dfd, " ---2 sent %d bytes urgent data, %d urgent bytes left\n",
400	n, so->so_urgc));
401	}
402
403	sb->sb_cc -= n;
404	sb->sb_rptr += n;
405	if (sb->sb_rptr >= (sb->sb_data + sb->sb_datalen))
406	sb->sb_rptr -= sb->sb_datalen;
407
408	return n;
409	}
410
411	/*
412	* Write data from so_rcv to so's socket,
413	* updating all sbuf field as necessary
414	*/
415	int
416	sowrite(PNATState pData, struct socket *so)
417	{
418	int n, nn;
419	struct sbuf *sb = &so->so_rcv;
420	size_t len = sb->sb_cc;
421	struct iovec iov[2];
422
423	STAM_PROFILE_START(&pData->StatIOwrite, a);
424	STAM_COUNTER_RESET(&pData->StatIOWrite_in_1);
425	STAM_COUNTER_RESET(&pData->StatIOWrite_in_1_bytes);
426	STAM_COUNTER_RESET(&pData->StatIOWrite_in_2);
427	STAM_COUNTER_RESET(&pData->StatIOWrite_in_2_1st_bytes);
428	STAM_COUNTER_RESET(&pData->StatIOWrite_in_2_2nd_bytes);
429	STAM_COUNTER_RESET(&pData->StatIOWrite_no_w);
430	STAM_COUNTER_RESET(&pData->StatIOWrite_rest);
431	STAM_COUNTER_RESET(&pData->StatIOWrite_rest_bytes);
432	DEBUG_CALL("sowrite");
433	DEBUG_ARG("so = %lx", (long)so);
434	QSOCKET_LOCK(tcb);
435	SOCKET_LOCK(so);
436	QSOCKET_UNLOCK(tcb);
437	if (so->so_urgc)
438	{
439	sosendoob(so);
440	if (sb->sb_cc == 0)
441	{
442	SOCKET_UNLOCK(so);
443	STAM_PROFILE_STOP(&pData->StatIOwrite, a);
444	return 0;
445	}
446	}
447
448	/*
449	* No need to check if there's something to write,
450	* sowrite wouldn't have been called otherwise
451	*/
452
453	len = sb->sb_cc;
454
455	iov[0].iov_base = sb->sb_rptr;
456	iov[1].iov_base = 0;
457	iov[1].iov_len = 0;
458	if (sb->sb_rptr < sb->sb_wptr)
459	{
460	iov[0].iov_len = sb->sb_wptr - sb->sb_rptr;
461	/* Should never succeed, but... */
462	if (iov[0].iov_len > len)
463	iov[0].iov_len = len;
464	n = 1;
465	}
466	else
467	{
468	iov[0].iov_len = (sb->sb_data + sb->sb_datalen) - sb->sb_rptr;
469	if (iov[0].iov_len > len)
470	iov[0].iov_len = len;
471	len -= iov[0].iov_len;
472	if (len)
473	{
474	iov[1].iov_base = sb->sb_data;
475	iov[1].iov_len = sb->sb_wptr - sb->sb_data;
476	if (iov[1].iov_len > len)
477	iov[1].iov_len = len;
478	n = 2;
479	}
480	else
481	n = 1;
482	}
483	STAM_STATS({
484	if (n == 1)
485	{
486	STAM_COUNTER_INC(&pData->StatIOWrite_in_1);
487	STAM_COUNTER_ADD(&pData->StatIOWrite_in_1_bytes, iov[0].iov_len);
488	}
489	else
490	{
491	STAM_COUNTER_INC(&pData->StatIOWrite_in_2);
492	STAM_COUNTER_ADD(&pData->StatIOWrite_in_2_1st_bytes, iov[0].iov_len);
493	STAM_COUNTER_ADD(&pData->StatIOWrite_in_2_2nd_bytes, iov[1].iov_len);
494	}
495	});
496	/* Check if there's urgent data to send, and if so, send it */
497	#ifdef HAVE_READV
498	nn = writev(so->s, (const struct iovec *)iov, n);
499	DEBUG_MISC((dfd, " ... wrote nn = %d bytes\n", nn));
500	#else
501	nn = send(so->s, iov[0].iov_base, iov[0].iov_len, 0);
502	#endif
503	/* This should never happen, but people tell me it does shrug */
504	if ( nn < 0
505	&& ( errno == EAGAIN
506	\|\| errno == EINTR
507	\|\| errno == EWOULDBLOCK))
508	{
509	SOCKET_UNLOCK(so);
510	STAM_PROFILE_STOP(&pData->StatIOwrite, a);
511	return 0;
512	}
513
514	if (nn < 0 \|\| (nn == 0 && iov[0].iov_len > 0))
515	{
516	DEBUG_MISC((dfd, " --- sowrite disconnected, so->so_state = %x, errno = %d\n",
517	so->so_state, errno));
518	sofcantsendmore(so);
519	tcp_sockclosed(pData, sototcpcb(so));
520	SOCKET_UNLOCK(so);
521	STAM_PROFILE_STOP(&pData->StatIOwrite, a);
522	return -1;
523	}
524
525	#ifndef HAVE_READV
526	if (n == 2 && nn == iov[0].iov_len)
527	{
528	int ret;
529	ret = send(so->s, iov[1].iov_base, iov[1].iov_len, 0);
530	if (ret > 0)
531	nn += ret;
532	STAM_STATS({
533	if (ret > 0 && ret != iov[1].iov_len)
534	{
535	STAM_COUNTER_INC(&pData->StatIOWrite_rest);
536	STAM_COUNTER_ADD(&pData->StatIOWrite_rest_bytes, (ret - iov[1].iov_len));
537	}
538	});
539	}
540	DEBUG_MISC((dfd, " ... wrote nn = %d bytes\n", nn));
541	#endif
542
543	/* Update sbuf */
544	sb->sb_cc -= nn;
545	sb->sb_rptr += nn;
546	if (sb->sb_rptr >= (sb->sb_data + sb->sb_datalen))
547	sb->sb_rptr -= sb->sb_datalen;
548
549	/*
550	* If in DRAIN mode, and there's no more data, set
551	* it CANTSENDMORE
552	*/
553	if ((so->so_state & SS_FWDRAIN) && sb->sb_cc == 0)
554	sofcantsendmore(so);
555
556	SOCKET_UNLOCK(so);
557	STAM_PROFILE_STOP(&pData->StatIOwrite, a);
558	return nn;
559	}
560
561	/*
562	* recvfrom() a UDP socket
563	*/
564	void
565	sorecvfrom(PNATState pData, struct socket *so)
566	{
567	ssize_t ret = 0;
568	struct sockaddr_in addr;
569	socklen_t addrlen = sizeof(struct sockaddr_in);
570
571	DEBUG_CALL("sorecvfrom");
572	DEBUG_ARG("so = %lx", (long)so);
573
574	if (so->so_type == IPPROTO_ICMP)
575	{
576	/* This is a "ping" reply */
577	#ifdef RT_OS_WINDOWS
578	sorecvfrom_icmp_win(pData, so);
579	#else /* RT_OS_WINDOWS */
580	sorecvfrom_icmp_unix(pData, so);
581	#endif /* !RT_OS_WINDOWS */
582	udp_detach(pData, so);
583	}
584	else
585	{
586	/* A "normal" UDP packet */
587	struct mbuf *m;
588	ssize_t len;
589	u_long n = 0;
590	#ifdef VBOX_WITH_SLIRP_BSD_MBUF
591	int size;
592	#endif
593	int rc = 0;
594	static int signalled = 0;
595
596	QSOCKET_LOCK(udb);
597	SOCKET_LOCK(so);
598	QSOCKET_UNLOCK(udb);
599
600	#ifndef VBOX_WITH_SLIRP_BSD_MBUF
601	if (!(m = m_get(pData)))
602	{
603	SOCKET_UNLOCK(so);
604	return;
605	}
606	/* adjust both parameters to maks M_FREEROOM calculate correct */
607	m->m_data += if_maxlinkhdr + sizeof(struct udphdr) + sizeof(struct ip);
608
609	/*
610	* XXX Shouldn't FIONREAD packets destined for port 53,
611	* but I don't know the max packet size for DNS lookups
612	*/
613	len = M_FREEROOM(m);
614	/* if (so->so_fport != RT_H2N_U16_C(53)) */
615	rc = ioctlsocket(so->s, FIONREAD, &n);
616	if ( rc == -1
617	&& ( errno == EAGAIN
618	\|\| errno == EWOULDBLOCK
619	\|\| errno == EINPROGRESS
620	\|\| errno == ENOTCONN))
621	{
622	m_freem(pData, m);
623	return;
624	}
625
626	Log2(("NAT: %R[natsock] ioctlsocket before read "
627	"(rc:%d errno:%d, n:%d)\n", so, rc, errno, n));
628
629	if (rc == -1 && signalled == 0)
630	{
631	LogRel(("NAT: can't fetch amount of bytes on socket %R[natsock], so message will be truncated.\n", so));
632	signalled = 1;
633	m_freem(pData, m);
634	return;
635	}
636
637	if (rc != -1 && n > len)
638	{
639	n = (m->m_data - m->m_dat) + m->m_len + n + 1;
640	m_inc(m, n);
641	len = M_FREEROOM(m);
642	}
643	ret = recvfrom(so->s, m->m_data, len, 0,
644	(struct sockaddr *)&addr, &addrlen);
645	Log2(("NAT: %R[natsock] ioctlsocket after read "
646	"(rc:%d errno:%d, n:%d) ret:%d, len:%d\n", so,
647	rc, errno, n, ret, len));
648	#else
649	/How many data has been received ?/
650	/*
651	* 1. calculate how much we can read
652	* 2. read as much as possible
653	* 3. attach buffer to allocated header mbuf
654	*/
655	rc = ioctlsocket(so->s, FIONREAD, &n);
656	if (rc == -1 && signalled == 0)
657	{
658	LogRel(("NAT: can't fetch amount of bytes on socket %R[natsock], so message will be truncated.\n", so));
659	signalled = 1;
660	}
661
662	len = sizeof(struct udpiphdr) + ETH_HLEN;
663	if (n > (if_mtu - len))
664	{
665	n = if_mtu - len; /* can't read than we can put in the mbuf*/
666	}
667	len += n;
668
669	size = MCLBYTES;
670	if (len < MSIZE)
671	size = MCLBYTES;
672	else if (len < MCLBYTES)
673	size = MCLBYTES;
674	else if (len < MJUM9BYTES)
675	size = MJUM9BYTES;
676	else if (len < MJUM16BYTES)
677	size = MJUM16BYTES;
678	else
679	AssertMsgFailed(("Unsupported size"));
680
681	m = m_getjcl(pData, M_NOWAIT, MT_HEADER, M_PKTHDR, size);
682	if (m == NULL)
683	return;
684	m->m_data += ETH_HLEN;
685	m->m_pkthdr.header = mtod(m, void *);
686	m->m_data += sizeof(struct udpiphdr);
687	ret = recvfrom(so->s, mtod(m, char *), n, 0,
688	(struct sockaddr *)&addr, &addrlen);
689	/* @todo (vvl) check which flags and type should be passed */
690	#endif
691	m->m_len = ret;
692	if (ret < 0)
693	{
694	u_char code = ICMP_UNREACH_PORT;
695
696	if (errno == EHOSTUNREACH)
697	code = ICMP_UNREACH_HOST;
698	else if (errno == ENETUNREACH)
699	code = ICMP_UNREACH_NET;
700
701	m_freem(pData, m);
702	if ( errno == EAGAIN
703	\|\| errno == EWOULDBLOCK
704	\|\| errno == EINPROGRESS
705	\|\| errno == ENOTCONN)
706	{
707	return;
708	}
709
710	Log2((" rx error, tx icmp ICMP_UNREACH:%i\n", code));
711	icmp_error(pData, so->so_m, ICMP_UNREACH, code, 0, strerror(errno));
712	so->so_m = NULL;
713	}
714	else
715	{
716	/*
717	* Hack: domain name lookup will be used the most for UDP,
718	* and since they'll only be used once there's no need
719	* for the 4 minute (or whatever) timeout... So we time them
720	* out much quicker (10 seconds for now...)
721	*/
722	if (so->so_expire)
723	{
724	if (so->so_fport != RT_H2N_U16_C(53))
725	so->so_expire = curtime + SO_EXPIRE;
726	}
727	/*
728	* last argument should be changed if Slirp will inject IP attributes
729	* Note: Here we can't check if dnsproxy's sent initial request
730	*/
731	#ifndef VBOX_WITH_SLIRP_BSD_MBUF
732	if (so->so_fport == RT_H2N_U16_C(53))
733	dnsproxy_answer(pData, so, m);
734	#endif
735
736	#if 0
737	if (m->m_len == len)
738	{
739	m_inc(m, MINCSIZE);
740	m->m_len = 0;
741	}
742	#endif
743
744	/*
745	* If this packet was destined for CTL_ADDR,
746	* make it look like that's where it came from, done by udp_output
747	*/
748	udp_output(pData, so, m, &addr);
749	SOCKET_UNLOCK(so);
750	} /* rx error */
751	} /* if ping packet */
752	}
753
754	/*
755	* sendto() a socket
756	*/
757	int
758	sosendto(PNATState pData, struct socket so, struct mbuf m)
759	{
760	int ret;
761	struct sockaddr_in *paddr;
762	struct sockaddr addr;
763	#if 0
764	struct sockaddr_in host_addr;
765	#endif
766	#ifdef VBOX_WITH_SLIRP_BSD_MBUF
767	caddr_t buf;
768	int mlen;
769	#endif
770
771	DEBUG_CALL("sosendto");
772	DEBUG_ARG("so = %lx", (long)so);
773	DEBUG_ARG("m = %lx", (long)m);
774
775	memset(&addr, 0, sizeof(struct sockaddr));
776	#ifdef RT_OS_DARWIN
777	addr.sa_len = sizeof(struct sockaddr_in);
778	#endif
779	paddr = (struct sockaddr_in *)&addr;
780	paddr->sin_family = AF_INET;
781	if ((so->so_faddr.s_addr & RT_H2N_U32(pData->netmask)) == pData->special_addr.s_addr)
782	{
783	/* It's an alias */
784	uint32_t last_byte = RT_N2H_U32(so->so_faddr.s_addr) & ~pData->netmask;
785	switch(last_byte)
786	{
787	#if 0
788	/* handle this case at 'default:' */
789	case CTL_BROADCAST:
790	addr.sin_addr.s_addr = INADDR_BROADCAST;
791	/* Send the packet to host to fully emulate broadcast */
792	/** @todo r=klaus: on Linux host this causes the host to receive
793	* the packet twice for some reason. And I cannot find any place
794	* in the man pages which states that sending a broadcast does not
795	* reach the host itself. */
796	host_addr.sin_family = AF_INET;
797	host_addr.sin_port = so->so_fport;
798	host_addr.sin_addr = our_addr;
799	sendto(so->s, m->m_data, m->m_len, 0,
800	(struct sockaddr *)&host_addr, sizeof (struct sockaddr));
801	break;
802	#endif
803	case CTL_DNS:
804	case CTL_ALIAS:
805	default:
806	if (last_byte == ~pData->netmask)
807	paddr->sin_addr.s_addr = INADDR_BROADCAST;
808	else
809	paddr->sin_addr = loopback_addr;
810	break;
811	}
812	}
813	else
814	paddr->sin_addr = so->so_faddr;
815	paddr->sin_port = so->so_fport;
816
817	DEBUG_MISC((dfd, " sendto()ing, addr.sin_port=%d, addr.sin_addr.s_addr=%.16s\n",
818	RT_N2H_U16(paddr->sin_port), inet_ntoa(paddr->sin_addr)));
819
820	/* Don't care what port we get */
821	#ifndef VBOX_WITH_SLIRP_BSD_MBUF
822	ret = sendto(so->s, m->m_data, m->m_len, 0, &addr, sizeof (struct sockaddr_in));
823	#else
824	mlen = m_length(m, NULL);
825	buf = RTMemAlloc(mlen);
826	if (buf == NULL)
827	{
828	return -1;
829	}
830	m_copydata(m, 0, mlen, buf);
831	ret = sendto(so->s, buf, mlen, 0,
832	(struct sockaddr *)&addr, sizeof (struct sockaddr));
833	#endif
834	if (ret < 0)
835	{
836	Log2(("UDP: sendto fails (%s)\n", strerror(errno)));
837	return -1;
838	}
839
840	/*
841	* Kill the socket if there's no reply in 4 minutes,
842	* but only if it's an expirable socket
843	*/
844	if (so->so_expire)
845	so->so_expire = curtime + SO_EXPIRE;
846	so->so_state = SS_ISFCONNECTED; /* So that it gets select()ed */
847	return 0;
848	}
849
850	/*
851	* XXX This should really be tcp_listen
852	*/
853	struct socket *
854	solisten(PNATState pData, u_int32_t bind_addr, u_int port, u_int32_t laddr, u_int lport, int flags)
855	{
856	struct sockaddr_in addr;
857	struct socket *so;
858	socklen_t addrlen = sizeof(addr);
859	int s, opt = 1;
860	int status;
861
862	DEBUG_CALL("solisten");
863	DEBUG_ARG("port = %d", port);
864	DEBUG_ARG("laddr = %x", laddr);
865	DEBUG_ARG("lport = %d", lport);
866	DEBUG_ARG("flags = %x", flags);
867
868	if ((so = socreate()) == NULL)
869	{
870	/* RTMemFree(so); Not sofree() ??? free(NULL) == NOP */
871	return NULL;
872	}
873
874	/* Don't tcp_attach... we don't need so_snd nor so_rcv */
875	if ((so->so_tcpcb = tcp_newtcpcb(pData, so)) == NULL)
876	{
877	RTMemFree(so);
878	return NULL;
879	}
880
881	SOCKET_LOCK_CREATE(so);
882	SOCKET_LOCK(so);
883	QSOCKET_LOCK(tcb);
884	insque(pData, so,&tcb);
885	NSOCK_INC();
886	QSOCKET_UNLOCK(tcb);
887
888	/*
889	* SS_FACCEPTONCE sockets must time out.
890	*/
891	if (flags & SS_FACCEPTONCE)
892	so->so_tcpcb->t_timer[TCPT_KEEP] = TCPTV_KEEP_INIT*2;
893
894	so->so_state = (SS_FACCEPTCONN\|flags);
895	so->so_lport = lport; /* Kept in network format */
896	so->so_laddr.s_addr = laddr; /* Ditto */
897
898	memset(&addr, 0, sizeof(addr));
899	#ifdef RT_OS_DARWIN
900	addr.sin_len = sizeof(addr);
901	#endif
902	addr.sin_family = AF_INET;
903	addr.sin_addr.s_addr = bind_addr;
904	addr.sin_port = port;
905
906	if ( ((s = socket(AF_INET, SOCK_STREAM, 0)) < 0)
907	\|\| (setsockopt(s, SOL_SOCKET, SO_REUSEADDR,(char *)&opt, sizeof(int)) < 0)
908	\|\| (bind(s,(struct sockaddr *)&addr, sizeof(addr)) < 0)
909	\|\| (listen(s, 1) < 0))
910	{
911	#ifdef RT_OS_WINDOWS
912	int tmperrno = WSAGetLastError(); /* Don't clobber the real reason we failed */
913	closesocket(s);
914	QSOCKET_LOCK(tcb);
915	sofree(pData, so);
916	QSOCKET_UNLOCK(tcb);
917	/* Restore the real errno */
918	WSASetLastError(tmperrno);
919	#else
920	int tmperrno = errno; /* Don't clobber the real reason we failed */
921	close(s);
922	QSOCKET_LOCK(tcb);
923	sofree(pData, so);
924	QSOCKET_UNLOCK(tcb);
925	/* Restore the real errno */
926	errno = tmperrno;
927	#endif
928	return NULL;
929	}
930	fd_nonblock(s);
931	setsockopt(s, SOL_SOCKET, SO_OOBINLINE,(char *)&opt, sizeof(int));
932
933	getsockname(s,(struct sockaddr *)&addr,&addrlen);
934	so->so_fport = addr.sin_port;
935	/* set socket buffers */
936	opt = pData->socket_rcv;
937	status = setsockopt(s, SOL_SOCKET, SO_RCVBUF, (char *)&opt, sizeof(int));
938	if (status < 0)
939	{
940	LogRel(("NAT: Error(%d) while setting RCV capacity to (%d)\n", errno, opt));
941	goto no_sockopt;
942	}
943	opt = pData->socket_snd;
944	status = setsockopt(s, SOL_SOCKET, SO_SNDBUF, (char *)&opt, sizeof(int));
945	if (status < 0)
946	{
947	LogRel(("NAT: Error(%d) while setting SND capacity to (%d)\n", errno, opt));
948	goto no_sockopt;
949	}
950	no_sockopt:
951	if (addr.sin_addr.s_addr == 0 \|\| addr.sin_addr.s_addr == loopback_addr.s_addr)
952	so->so_faddr = alias_addr;
953	else
954	so->so_faddr = addr.sin_addr;
955
956	so->s = s;
957	SOCKET_UNLOCK(so);
958	return so;
959	}
960
961	/*
962	* Data is available in so_rcv
963	* Just write() the data to the socket
964	* XXX not yet...
965	*/
966	void
967	sorwakeup(struct socket *so)
968	{
969	#if 0
970	sowrite(so);
971	FD_CLR(so->s,&writefds);
972	#endif
973	}
974
975	/*
976	* Data has been freed in so_snd
977	* We have room for a read() if we want to
978	* For now, don't read, it'll be done in the main loop
979	*/
980	void
981	sowwakeup(struct socket *so)
982	{
983	}
984
985	/*
986	* Various session state calls
987	* XXX Should be #define's
988	* The socket state stuff needs work, these often get call 2 or 3
989	* times each when only 1 was needed
990	*/
991	void
992	soisfconnecting(struct socket *so)
993	{
994	so->so_state &= ~(SS_NOFDREF\|SS_ISFCONNECTED\|SS_FCANTRCVMORE\|
995	SS_FCANTSENDMORE\|SS_FWDRAIN);
996	so->so_state \|= SS_ISFCONNECTING; /* Clobber other states */
997	}
998
999	void
1000	soisfconnected(struct socket *so)
1001	{
1002	so->so_state &= ~(SS_ISFCONNECTING\|SS_FWDRAIN\|SS_NOFDREF);
1003	so->so_state \|= SS_ISFCONNECTED; /* Clobber other states */
1004	}
1005
1006	void
1007	sofcantrcvmore(struct socket *so)
1008	{
1009	if ((so->so_state & SS_NOFDREF) == 0)
1010	{
1011	shutdown(so->s, 0);
1012	}
1013	so->so_state &= ~(SS_ISFCONNECTING);
1014	if (so->so_state & SS_FCANTSENDMORE)
1015	so->so_state = SS_NOFDREF; /* Don't select it */
1016	/* XXX close() here as well? */
1017	else
1018	so->so_state \|= SS_FCANTRCVMORE;
1019	}
1020
1021	void
1022	sofcantsendmore(struct socket *so)
1023	{
1024	if ((so->so_state & SS_NOFDREF) == 0)
1025	shutdown(so->s, 1); /* send FIN to fhost */
1026
1027	so->so_state &= ~(SS_ISFCONNECTING);
1028	if (so->so_state & SS_FCANTRCVMORE)
1029	so->so_state = SS_NOFDREF; /* as above */
1030	else
1031	so->so_state \|= SS_FCANTSENDMORE;
1032	}
1033
1034	void
1035	soisfdisconnected(struct socket *so)
1036	{
1037	#if 0
1038	so->so_state &= ~(SS_ISFCONNECTING\|SS_ISFCONNECTED);
1039	close(so->s);
1040	so->so_state = SS_ISFDISCONNECTED;
1041	/*
1042	* XXX Do nothing ... ?
1043	*/
1044	#endif
1045	}
1046
1047	/*
1048	* Set write drain mode
1049	* Set CANTSENDMORE once all data has been write()n
1050	*/
1051	void
1052	sofwdrain(struct socket *so)
1053	{
1054	if (so->so_rcv.sb_cc)
1055	so->so_state \|= SS_FWDRAIN;
1056	else
1057	sofcantsendmore(so);
1058	}
1059
1060	static void
1061	send_icmp_to_guest(PNATState pData, char buff, size_t len, struct socket so, const struct sockaddr_in *addr)
1062	{
1063	struct ip *ip;
1064	uint32_t dst, src;
1065	char ip_copy[256];
1066	struct icmp *icp;
1067	int old_ip_len = 0;
1068	int hlen, original_hlen = 0;
1069	struct mbuf *m;
1070	struct icmp_msg *icm;
1071	uint8_t proto;
1072	int type = 0;
1073	#ifndef VBOX_WITH_SLIRP_BSD_MBUF
1074	int m_room;
1075	#endif
1076
1077	ip = (struct ip *)buff;
1078	/* Fix ip->ip_len to contain the total packet length including the header
1079	* in _host_ byte order for all OSes. On Darwin, that value already is in
1080	* host byte order. Solaris and Darwin report only the payload. */
1081	#ifndef RT_OS_DARWIN
1082	ip->ip_len = RT_N2H_U16(ip->ip_len);
1083	#endif
1084	hlen = (ip->ip_hl << 2);
1085	#if defined(RT_OS_SOLARIS) \|\| defined(RT_OS_DARWIN)
1086	ip->ip_len += hlen;
1087	#endif
1088	if (ip->ip_len < hlen + ICMP_MINLEN)
1089	{
1090	Log(("send_icmp_to_guest: ICMP header is too small to understand which type/subtype of the datagram\n"));
1091	return;
1092	}
1093	icp = (struct icmp )((char )ip + hlen);
1094
1095	Log(("ICMP:received msg(t:%d, c:%d)\n", icp->icmp_type, icp->icmp_code));
1096	if ( icp->icmp_type != ICMP_ECHOREPLY
1097	&& icp->icmp_type != ICMP_TIMXCEED
1098	&& icp->icmp_type != ICMP_UNREACH)
1099	{
1100	return;
1101	}
1102
1103	/*
1104	* ICMP_ECHOREPLY, ICMP_TIMXCEED, ICMP_UNREACH minimal header size is
1105	* ICMP_ECHOREPLY assuming data 0
1106	* icmp_{type(8), code(8), cksum(16),identifier(16),seqnum(16)}
1107	*/
1108	if (ip->ip_len < hlen + 8)
1109	{
1110	Log(("send_icmp_to_guest: NAT accept ICMP_{ECHOREPLY, TIMXCEED, UNREACH} the minimum size is 64 (see rfc792)\n"));
1111	return;
1112	}
1113
1114	type = icp->icmp_type;
1115	if ( type == ICMP_TIMXCEED
1116	\|\| type == ICMP_UNREACH)
1117	{
1118	/*
1119	* ICMP_TIMXCEED, ICMP_UNREACH minimal header size is
1120	* icmp_{type(8), code(8), cksum(16),unused(32)} + IP header + 64 bit of original datagram
1121	*/
1122	if (ip->ip_len < hlen + 2*8 + sizeof(struct ip))
1123	{
1124	Log(("send_icmp_to_guest: NAT accept ICMP_{TIMXCEED, UNREACH} the minimum size of ipheader + 64 bit of data (see rfc792)\n"));
1125	return;
1126	}
1127	ip = &icp->icmp_ip;
1128	}
1129
1130	icm = icmp_find_original_mbuf(pData, ip);
1131	if (icm == NULL)
1132	{
1133	Log(("NAT: Can't find the corresponding packet for the received ICMP\n"));
1134	return;
1135	}
1136
1137	m = icm->im_m;
1138	Assert(m != NULL);
1139
1140	src = addr->sin_addr.s_addr;
1141	if (type == ICMP_ECHOREPLY)
1142	{
1143	struct ip ip0 = mtod(m, struct ip );
1144	struct icmp icp0 = (struct icmp )((char *)ip0 + (ip0->ip_hl << 2));
1145	if (icp0->icmp_type != ICMP_ECHO)
1146	{
1147	Log(("NAT: we haven't found echo for this reply\n"));
1148	return;
1149	}
1150	/*
1151	* while combining buffer to send (see ip_icmp.c) we control ICMP header only,
1152	* IP header combined by OS network stack, our local copy of IP header contians values
1153	* in host byte order so no byte order conversion is required. IP headers fields are converting
1154	* in ip_output0 routine only.
1155	*/
1156	if ( (ip->ip_len - hlen)
1157	!= (ip0->ip_len - (ip0->ip_hl << 2)))
1158	{
1159	Log(("NAT: ECHO(%d) lenght doesn't match ECHOREPLY(%d)\n",
1160	(ip->ip_len - hlen), (ip0->ip_len - (ip0->ip_hl << 2))));
1161	return;
1162	}
1163	}
1164
1165	/* ip points on origianal ip header */
1166	ip = mtod(m, struct ip *);
1167	proto = ip->ip_p;
1168	/* Now ip is pointing on header we've sent from guest */
1169	if ( icp->icmp_type == ICMP_TIMXCEED
1170	\|\| icp->icmp_type == ICMP_UNREACH)
1171	{
1172	old_ip_len = (ip->ip_hl << 2) + 64;
1173	if (old_ip_len > sizeof(ip_copy))
1174	old_ip_len = sizeof(ip_copy);
1175	memcpy(ip_copy, ip, old_ip_len);
1176	}
1177
1178	/* source address from original IP packet*/
1179	dst = ip->ip_src.s_addr;
1180
1181	/* overide ther tail of old packet */
1182	ip = mtod(m, struct ip ); / ip is from mbuf we've overrided */
1183	original_hlen = ip->ip_hl << 2;
1184	/* saves original ip header and options */
1185	#ifdef VBOX_WITH_SLIRP_BSD_MBUF
1186	m_copyback(pData, m, original_hlen, len - hlen, buff + hlen);
1187	ip->ip_len = m_length(m, NULL);
1188	#else
1189	/* m_room space in the saved m buffer */
1190	m_room = M_ROOM(m);
1191	if (m_room < len - hlen + original_hlen)
1192	{
1193	/* we need involve ether header length into new buffer buffer calculation */
1194	m_inc(m, if_maxlinkhdr + len - hlen + original_hlen);
1195	if (m->m_size < if_maxlinkhdr + len - hlen + original_hlen)
1196	{
1197	Log(("send_icmp_to_guest: extending buffer was failed (packet is dropped)\n"));
1198	return;
1199	}
1200	}
1201	memcpy(m->m_data + original_hlen, buff + hlen, len - hlen);
1202	m->m_len = len - hlen + original_hlen;
1203	ip->ip_len = m->m_len;
1204	#endif
1205	ip->ip_p = IPPROTO_ICMP; /* the original package could be whatever, but we're response via ICMP*/
1206
1207	icp = (struct icmp )((char )ip + (ip->ip_hl << 2));
1208	type = icp->icmp_type;
1209	if ( type == ICMP_TIMXCEED
1210	\|\| type == ICMP_UNREACH)
1211	{
1212	/* according RFC 793 error messages required copy of initial IP header + 64 bit */
1213	memcpy(&icp->icmp_ip, ip_copy, old_ip_len);
1214	ip->ip_tos = ((ip->ip_tos & 0x1E) \| 0xC0); /* high priority for errors */
1215	}
1216
1217	ip->ip_src.s_addr = src;
1218	ip->ip_dst.s_addr = dst;
1219	icmp_reflect(pData, m);
1220	LIST_REMOVE(icm, im_list);
1221	/* Don't call m_free here*/
1222
1223	if ( type == ICMP_TIMXCEED
1224	\|\| type == ICMP_UNREACH)
1225	{
1226	icm->im_so->so_m = NULL;
1227	switch (proto)
1228	{
1229	case IPPROTO_UDP:
1230	/XXX: so->so_m already freed so we shouldn't call sofree /
1231	udp_detach(pData, icm->im_so);
1232	break;
1233	case IPPROTO_TCP:
1234	/close tcp should be here /
1235	break;
1236	default:
1237	/* do nothing */
1238	break;
1239	}
1240	}
1241	RTMemFree(icm);
1242	}
1243
1244	#ifdef RT_OS_WINDOWS
1245	static void
1246	sorecvfrom_icmp_win(PNATState pData, struct socket *so)
1247	{
1248	int len;
1249	int i;
1250	struct ip *ip;
1251	struct mbuf *m;
1252	struct icmp *icp;
1253	struct icmp_msg *icm;
1254	struct ip ip_broken; / ICMP returns header + 64 bit of packet */
1255	uint32_t src;
1256	ICMP_ECHO_REPLY *icr;
1257	int hlen = 0;
1258	int data_len = 0;
1259	int nbytes = 0;
1260	u_char code = ~0;
1261	int out_len;
1262	int size;
1263
1264	len = pData->pfIcmpParseReplies(pData->pvIcmpBuffer, pData->szIcmpBuffer);
1265	if (len < 0)
1266	{
1267	LogRel(("NAT: Error (%d) occurred on ICMP receiving\n", GetLastError()));
1268	return;
1269	}
1270	if (len == 0)
1271	return; /* no error */
1272
1273	icr = (ICMP_ECHO_REPLY *)pData->pvIcmpBuffer;
1274	for (i = 0; i < len; ++i)
1275	{
1276	switch(icr[i].Status)
1277	{
1278	case IP_DEST_HOST_UNREACHABLE:
1279	code = (code != ~0 ? code : ICMP_UNREACH_HOST);
1280	case IP_DEST_NET_UNREACHABLE:
1281	code = (code != ~0 ? code : ICMP_UNREACH_NET);
1282	case IP_DEST_PROT_UNREACHABLE:
1283	code = (code != ~0 ? code : ICMP_UNREACH_PROTOCOL);
1284	/* UNREACH error inject here */
1285	case IP_DEST_PORT_UNREACHABLE:
1286	code = (code != ~0 ? code : ICMP_UNREACH_PORT);
1287	icmp_error(pData, so->so_m, ICMP_UNREACH, code, 0, "Error occurred!!!");
1288	so->so_m = NULL;
1289	break;
1290	case IP_SUCCESS: /* echo replied */
1291	# ifndef VBOX_WITH_SLIRP_BSD_MBUF
1292	m = m_get(pData);
1293	# else
1294	out_len = ETH_HLEN + sizeof(struct ip) + 8;
1295	size;
1296	size = MCLBYTES;
1297	if (out_len < MSIZE)
1298	size = MCLBYTES;
1299	else if (out_len < MCLBYTES)
1300	size = MCLBYTES;
1301	else if (out_len < MJUM9BYTES)
1302	size = MJUM9BYTES;
1303	else if (out_len < MJUM16BYTES)
1304	size = MJUM16BYTES;
1305	else
1306	AssertMsgFailed(("Unsupported size"));
1307
1308	m = m_getjcl(pData, M_NOWAIT, MT_HEADER, M_PKTHDR, size);
1309	if (m == NULL)
1310	return;
1311	# endif
1312	m->m_len = 0;
1313	m->m_data += if_maxlinkhdr;
1314	ip = mtod(m, struct ip *);
1315	ip->ip_src.s_addr = icr[i].Address;
1316	ip->ip_p = IPPROTO_ICMP;
1317	ip->ip_dst.s_addr = so->so_laddr.s_addr; /XXX: still the hack/
1318	data_len = sizeof(struct ip);
1319	ip->ip_hl = data_len >> 2; /* requiered for icmp_reflect, no IP options */
1320	ip->ip_ttl = icr[i].Options.Ttl;
1321
1322	icp = (struct icmp )&ip[1]; / no options */
1323	icp->icmp_type = ICMP_ECHOREPLY;
1324	icp->icmp_code = 0;
1325	icp->icmp_id = so->so_icmp_id;
1326	icp->icmp_seq = so->so_icmp_seq;
1327
1328	data_len += ICMP_MINLEN;
1329
1330	# ifndef VBOX_WITH_SLIRP_BSD_MBUF
1331	nbytes = (data_len + icr[i].DataSize > m->m_size? m->m_size - data_len: icr[i].DataSize);
1332	memcpy(icp->icmp_data, icr[i].Data, nbytes);
1333	# else
1334	hlen = (ip->ip_hl << 2);
1335	m->m_pkthdr.header = mtod(m, void *);
1336	m->m_len = data_len;
1337
1338	m_copyback(pData, m, hlen + 8, icr[i].DataSize, icr[i].Data);
1339	# endif
1340
1341	data_len += icr[i].DataSize;
1342
1343	ip->ip_len = data_len;
1344	m->m_len = ip->ip_len;
1345
1346	icmp_reflect(pData, m);
1347	break;
1348	case IP_TTL_EXPIRED_TRANSIT: /* TTL expired */
1349
1350	ip_broken = icr[i].Data;
1351	icm = icmp_find_original_mbuf(pData, ip_broken);
1352	if (icm == NULL) {
1353	Log(("ICMP: can't find original package (first double word %x)\n", (uint32_t )ip_broken));
1354	return;
1355	}
1356	m = icm->im_m;
1357	ip = mtod(m, struct ip *);
1358	ip->ip_ttl = icr[i].Options.Ttl;
1359	src = ip->ip_src.s_addr;
1360	ip->ip_dst.s_addr = src;
1361	ip->ip_dst.s_addr = icr[i].Address;
1362
1363	hlen = (ip->ip_hl << 2);
1364	icp = (struct icmp )((char )ip + hlen);
1365	ip_broken->ip_src.s_addr = src; /it packet sent from host not from guest/
1366	data_len = (ip_broken->ip_hl << 2) + 64;
1367
1368	#ifndef VBOX_WITH_SLIRP_BSD_MBUF
1369	nbytes =(hlen + ICMP_MINLEN + data_len > m->m_size? m->m_size - (hlen + ICMP_MINLEN): data_len);
1370	memcpy(icp->icmp_data, ip_broken, nbytes);
1371	#else
1372	m->m_len = data_len;
1373	m->m_pkthdr.header = mtod(m, void *);
1374	m_copyback(pData, m, ip->ip_hl >> 2, icr[i].DataSize, icr[i].Data);
1375	#endif
1376	icmp_reflect(pData, m);
1377	break;
1378	default:
1379	Log(("ICMP(default): message with Status: %x was received from %x\n", icr[i].Status, icr[i].Address));
1380	break;
1381	}
1382	}
1383	}
1384	#else /* !RT_OS_WINDOWS */
1385	static void sorecvfrom_icmp_unix(PNATState pData, struct socket *so)
1386	{
1387	struct sockaddr_in addr;
1388	socklen_t addrlen = sizeof(struct sockaddr_in);
1389	struct ip ip;
1390	char *buff;
1391	int len = 0;
1392
1393	/* 1- step: read the ip header */
1394	len = recvfrom(so->s, &ip, sizeof(struct ip), MSG_PEEK,
1395	(struct sockaddr *)&addr, &addrlen);
1396	if ( len < 0
1397	&& ( errno == EAGAIN
1398	\|\| errno == EWOULDBLOCK
1399	\|\| errno == EINPROGRESS
1400	\|\| errno == ENOTCONN))
1401	{
1402	Log(("sorecvfrom_icmp_unix: 1 - step can't read IP datagramm (would block)\n"));
1403	return;
1404	}
1405
1406	if ( len < sizeof(struct ip)
1407	\|\| len < 0
1408	\|\| len == 0)
1409	{
1410	u_char code;
1411	code = ICMP_UNREACH_PORT;
1412
1413	if (errno == EHOSTUNREACH)
1414	code = ICMP_UNREACH_HOST;
1415	else if (errno == ENETUNREACH)
1416	code = ICMP_UNREACH_NET;
1417
1418	LogRel((" udp icmp rx errno = %d-%s\n",
1419	errno, strerror(errno)));
1420	icmp_error(pData, so->so_m, ICMP_UNREACH, code, 0, strerror(errno));
1421	so->so_m = NULL;
1422	Log(("sorecvfrom_icmp_unix: 1 - step can't read IP datagramm \n"));
1423	return;
1424	}
1425	/* basic check of IP header */
1426	if ( ip.ip_v != IPVERSION
1427	# ifndef RT_OS_DARWIN
1428	\|\| ip.ip_p != IPPROTO_ICMP
1429	# endif
1430	)
1431	{
1432	Log(("sorecvfrom_icmp_unix: 1 - step IP isn't IPv4 \n"));
1433	return;
1434	}
1435	# ifndef RT_OS_DARWIN
1436	/* Darwin reports the IP length already in host byte order. */
1437	ip.ip_len = RT_N2H_U16(ip.ip_len);
1438	# endif
1439	# if defined(RT_OS_SOLARIS) \|\| defined(RT_OS_DARWIN)
1440	/* Solaris and Darwin report the payload only */
1441	ip.ip_len += (ip.ip_hl << 2);
1442	# endif
1443	/* Note: ip->ip_len in host byte order (all OS) */
1444	len = ip.ip_len;
1445	buff = RTMemAlloc(len);
1446	if (buff == NULL)
1447	{
1448	Log(("sorecvfrom_icmp_unix: 1 - step can't allocate enought room for datagram\n"));
1449	return;
1450	}
1451	/* 2 - step: we're reading rest of the datagramm to the buffer */
1452	addrlen = sizeof(struct sockaddr_in);
1453	memset(&addr, 0, addrlen);
1454	len = recvfrom(so->s, buff, len, 0,
1455	(struct sockaddr *)&addr, &addrlen);
1456	if ( len < 0
1457	&& ( errno == EAGAIN
1458	\|\| errno == EWOULDBLOCK
1459	\|\| errno == EINPROGRESS
1460	\|\| errno == ENOTCONN))
1461	{
1462	Log(("sorecvfrom_icmp_unix: 2 - step can't read IP body (would block expected:%d)\n",
1463	ip.ip_len));
1464	RTMemFree(buff);
1465	return;
1466	}
1467	if ( len < 0
1468	\|\| len == 0)
1469	{
1470	Log(("sorecvfrom_icmp_unix: 2 - step read of the rest of datagramm is fallen (errno:%d, len:%d expected: %d)\n",
1471	errno, len, (ip.ip_len - sizeof(struct ip))));
1472	RTMemFree(buff);
1473	return;
1474	}
1475	/* len is modified in 2nd read, when the rest of the datagramm was read */
1476	send_icmp_to_guest(pData, buff, len, so, &addr);
1477	RTMemFree(buff);
1478	}
1479	#endif /* !RT_OS_WINDOWS */

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

source: vbox/trunk/src/VBox/Devices/Network/slirp/socket.c@ 28915

Download in other formats: