socket.c@ 33978

Last change on this file since 33978 was 33978, checked in by vboxsync, 14 years ago
NAT: log.
Property svn:eol-style set to `native` Property svn:keywords set to `Author Date Id Revision`
File size: 43.6 KB

Line
1	/* $Id: socket.c 33978 2010-11-11 12:00:13Z 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	#ifndef VBOX_WITH_SLIRP_BSD_SBUF
137	int
138	soread(PNATState pData, struct socket *so)
139	{
140	int n, nn, lss, total;
141	struct sbuf *sb = &so->so_snd;
142	size_t len = sb->sb_datalen - sb->sb_cc;
143	struct iovec iov[2];
144	int mss = so->so_tcpcb->t_maxseg;
145
146	STAM_PROFILE_START(&pData->StatIOread, a);
147	STAM_COUNTER_RESET(&pData->StatIORead_in_1);
148	STAM_COUNTER_RESET(&pData->StatIORead_in_2);
149
150	QSOCKET_LOCK(tcb);
151	SOCKET_LOCK(so);
152	QSOCKET_UNLOCK(tcb);
153
154	DEBUG_CALL("soread");
155	DEBUG_ARG("so = %lx", (long)so);
156
157	/*
158	* No need to check if there's enough room to read.
159	* soread wouldn't have been called if there weren't
160	*/
161
162	len = sb->sb_datalen - sb->sb_cc;
163
164	iov[0].iov_base = sb->sb_wptr;
165	iov[1].iov_base = 0;
166	iov[1].iov_len = 0;
167	if (sb->sb_wptr < sb->sb_rptr)
168	{
169	iov[0].iov_len = sb->sb_rptr - sb->sb_wptr;
170	/* Should never succeed, but... */
171	if (iov[0].iov_len > len)
172	iov[0].iov_len = len;
173	if (iov[0].iov_len > mss)
174	iov[0].iov_len -= iov[0].iov_len%mss;
175	n = 1;
176	}
177	else
178	{
179	iov[0].iov_len = (sb->sb_data + sb->sb_datalen) - sb->sb_wptr;
180	/* Should never succeed, but... */
181	if (iov[0].iov_len > len)
182	iov[0].iov_len = len;
183	len -= iov[0].iov_len;
184	if (len)
185	{
186	iov[1].iov_base = sb->sb_data;
187	iov[1].iov_len = sb->sb_rptr - sb->sb_data;
188	if (iov[1].iov_len > len)
189	iov[1].iov_len = len;
190	total = iov[0].iov_len + iov[1].iov_len;
191	if (total > mss)
192	{
193	lss = total % mss;
194	if (iov[1].iov_len > lss)
195	{
196	iov[1].iov_len -= lss;
197	n = 2;
198	}
199	else
200	{
201	lss -= iov[1].iov_len;
202	iov[0].iov_len -= lss;
203	n = 1;
204	}
205	}
206	else
207	n = 2;
208	}
209	else
210	{
211	if (iov[0].iov_len > mss)
212	iov[0].iov_len -= iov[0].iov_len%mss;
213	n = 1;
214	}
215	}
216
217	#ifdef HAVE_READV
218	nn = readv(so->s, (struct iovec *)iov, n);
219	DEBUG_MISC((dfd, " ... read nn = %d bytes\n", nn));
220	#else
221	nn = recv(so->s, iov[0].iov_base, iov[0].iov_len, (so->so_tcpcb->t_force? MSG_OOB:0));
222	#endif
223	if (nn <= 0)
224	{
225	/*
226	* Special case for WSAEnumNetworkEvents: If we receive 0 bytes that
227	* _could_ mean that the connection is closed. But we will receive an
228	* FD_CLOSE event later if the connection was _really_ closed. With
229	* www.youtube.com I see this very often. Closing the socket too early
230	* would be dangerous.
231	*/
232	int status;
233	unsigned long pending = 0;
234	status = ioctlsocket(so->s, FIONREAD, &pending);
235	if (status < 0)
236	Log(("NAT:error in WSAIoctl: %d\n", errno));
237	if (nn == 0 && (pending != 0))
238	{
239	SOCKET_UNLOCK(so);
240	STAM_PROFILE_STOP(&pData->StatIOread, a);
241	return 0;
242	}
243	if ( nn < 0
244	&& ( errno == EINTR
245	\|\| errno == EAGAIN
246	\|\| errno == EWOULDBLOCK))
247	{
248	SOCKET_UNLOCK(so);
249	STAM_PROFILE_STOP(&pData->StatIOread, a);
250	return 0;
251	}
252	else
253	{
254	/* nn == 0 means peer has performed an orderly shutdown */
255	DEBUG_MISC((dfd, " --- soread() disconnected, nn = %d, errno = %d-%s\n",
256	nn, errno, strerror(errno)));
257	sofcantrcvmore(so);
258	tcp_sockclosed(pData, sototcpcb(so));
259	SOCKET_UNLOCK(so);
260	STAM_PROFILE_STOP(&pData->StatIOread, a);
261	return -1;
262	}
263	}
264	STAM_STATS(
265	if (n == 1)
266	{
267	STAM_COUNTER_INC(&pData->StatIORead_in_1);
268	STAM_COUNTER_ADD(&pData->StatIORead_in_1_bytes, nn);
269	}
270	else
271	{
272	STAM_COUNTER_INC(&pData->StatIORead_in_2);
273	STAM_COUNTER_ADD(&pData->StatIORead_in_2_1st_bytes, nn);
274	}
275	);
276
277	#ifndef HAVE_READV
278	/*
279	* If there was no error, try and read the second time round
280	* We read again if n = 2 (ie, there's another part of the buffer)
281	* and we read as much as we could in the first read
282	* We don't test for <= 0 this time, because there legitimately
283	* might not be any more data (since the socket is non-blocking),
284	* a close will be detected on next iteration.
285	* A return of -1 wont (shouldn't) happen, since it didn't happen above
286	*/
287	if (n == 2 && nn == iov[0].iov_len)
288	{
289	int ret;
290	ret = recv(so->s, iov[1].iov_base, iov[1].iov_len, 0);
291	if (ret > 0)
292	nn += ret;
293	STAM_STATS(
294	if (ret > 0)
295	{
296	STAM_COUNTER_INC(&pData->StatIORead_in_2);
297	STAM_COUNTER_ADD(&pData->StatIORead_in_2_2nd_bytes, ret);
298	}
299	);
300	}
301
302	DEBUG_MISC((dfd, " ... read nn = %d bytes\n", nn));
303	#endif
304
305	/* Update fields */
306	sb->sb_cc += nn;
307	sb->sb_wptr += nn;
308	if (sb->sb_wptr >= (sb->sb_data + sb->sb_datalen))
309	sb->sb_wptr -= sb->sb_datalen;
310	STAM_PROFILE_STOP(&pData->StatIOread, a);
311	SOCKET_UNLOCK(so);
312	return nn;
313	}
314	#else /* VBOX_WITH_SLIRP_BSD_SBUF */
315	int
316	soread(PNATState pData, struct socket *so)
317	{
318	int n;
319	char *buf;
320	struct sbuf *sb = &so->so_snd;
321	size_t len = sbspace(sb);
322	int mss = so->so_tcpcb->t_maxseg;
323
324	STAM_PROFILE_START(&pData->StatIOread, a);
325	STAM_COUNTER_RESET(&pData->StatIORead_in_1);
326	STAM_COUNTER_RESET(&pData->StatIORead_in_2);
327
328	QSOCKET_LOCK(tcb);
329	SOCKET_LOCK(so);
330	QSOCKET_UNLOCK(tcb);
331
332	DEBUG_CALL("soread");
333	DEBUG_ARG("so = %lx", (long)so);
334
335	if (len > mss)
336	len -= len % mss;
337	buf = RTMemAlloc(len);
338	if (buf == NULL)
339	{
340	Log(("NAT: can't alloc enough memory\n"));
341	return -1;
342	}
343
344	n = recv(so->s, buf, len, (so->so_tcpcb->t_force? MSG_OOB:0));
345	if (n <= 0)
346	{
347	/*
348	* Special case for WSAEnumNetworkEvents: If we receive 0 bytes that
349	* _could_ mean that the connection is closed. But we will receive an
350	* FD_CLOSE event later if the connection was _really_ closed. With
351	* www.youtube.com I see this very often. Closing the socket too early
352	* would be dangerous.
353	*/
354	int status;
355	unsigned long pending = 0;
356	status = ioctlsocket(so->s, FIONREAD, &pending);
357	if (status < 0)
358	Log(("NAT:error in WSAIoctl: %d\n", errno));
359	if (n == 0 && (pending != 0))
360	{
361	SOCKET_UNLOCK(so);
362	STAM_PROFILE_STOP(&pData->StatIOread, a);
363	RTMemFree(buf);
364	return 0;
365	}
366	if ( n < 0
367	&& ( errno == EINTR
368	\|\| errno == EAGAIN
369	\|\| errno == EWOULDBLOCK))
370	{
371	SOCKET_UNLOCK(so);
372	STAM_PROFILE_STOP(&pData->StatIOread, a);
373	RTMemFree(buf);
374	return 0;
375	}
376	else
377	{
378	DEBUG_MISC((dfd, " --- soread() disconnected, n = %d, errno = %d-%s\n",
379	n, errno, strerror(errno)));
380	sofcantrcvmore(so);
381	tcp_sockclosed(pData, sototcpcb(so));
382	SOCKET_UNLOCK(so);
383	STAM_PROFILE_STOP(&pData->StatIOread, a);
384	RTMemFree(buf);
385	return -1;
386	}
387	}
388
389	sbuf_bcat(sb, buf, n);
390	RTMemFree(buf);
391	return n;
392	}
393	#endif
394
395	/*
396	* Get urgent data
397	*
398	* When the socket is created, we set it SO_OOBINLINE,
399	* so when OOB data arrives, we soread() it and everything
400	* in the send buffer is sent as urgent data
401	*/
402	void
403	sorecvoob(PNATState pData, struct socket *so)
404	{
405	struct tcpcb *tp = sototcpcb(so);
406	ssize_t ret;
407
408	DEBUG_CALL("sorecvoob");
409	DEBUG_ARG("so = %lx", (long)so);
410
411	/*
412	* We take a guess at how much urgent data has arrived.
413	* In most situations, when urgent data arrives, the next
414	* read() should get all the urgent data. This guess will
415	* be wrong however if more data arrives just after the
416	* urgent data, or the read() doesn't return all the
417	* urgent data.
418	*/
419	ret = soread(pData, so);
420	tp->snd_up = tp->snd_una + SBUF_LEN(&so->so_snd);
421	tp->t_force = 1;
422	tcp_output(pData, tp);
423	tp->t_force = 0;
424	}
425	#ifndef VBOX_WITH_SLIRP_BSD_SBUF
426	/*
427	* Send urgent data
428	* There's a lot duplicated code here, but...
429	*/
430	int
431	sosendoob(struct socket *so)
432	{
433	struct sbuf *sb = &so->so_rcv;
434	char buff[2048]; /* XXX Shouldn't be sending more oob data than this */
435
436	int n, len;
437
438	DEBUG_CALL("sosendoob");
439	DEBUG_ARG("so = %lx", (long)so);
440
441	if (so->so_urgc > sizeof(buff))
442	so->so_urgc = sizeof(buff); /* XXX */
443
444	if (sb->sb_rptr < sb->sb_wptr)
445	{
446	/* We can send it directly */
447	n = send(so->s, sb->sb_rptr, so->so_urgc, (MSG_OOB)); /* \|MSG_DONTWAIT)); */
448	so->so_urgc -= n;
449
450	DEBUG_MISC((dfd, " --- sent %d bytes urgent data, %d urgent bytes left\n",
451	n, so->so_urgc));
452	}
453	else
454	{
455	/*
456	* Since there's no sendv or sendtov like writev,
457	* we must copy all data to a linear buffer then
458	* send it all
459	*/
460	len = (sb->sb_data + sb->sb_datalen) - sb->sb_rptr;
461	if (len > so->so_urgc)
462	len = so->so_urgc;
463	memcpy(buff, sb->sb_rptr, len);
464	so->so_urgc -= len;
465	if (so->so_urgc)
466	{
467	n = sb->sb_wptr - sb->sb_data;
468	if (n > so->so_urgc)
469	n = so->so_urgc;
470	memcpy(buff + len, sb->sb_data, n);
471	so->so_urgc -= n;
472	len += n;
473	}
474	n = send(so->s, buff, len, (MSG_OOB)); /* \|MSG_DONTWAIT)); */
475	#ifdef DEBUG
476	if (n != len)
477	DEBUG_ERROR((dfd, "Didn't send all data urgently XXXXX\n"));
478	#endif
479	DEBUG_MISC((dfd, " ---2 sent %d bytes urgent data, %d urgent bytes left\n",
480	n, so->so_urgc));
481	}
482
483	sb->sb_cc -= n;
484	sb->sb_rptr += n;
485	if (sb->sb_rptr >= (sb->sb_data + sb->sb_datalen))
486	sb->sb_rptr -= sb->sb_datalen;
487
488	return n;
489	}
490
491	/*
492	* Write data from so_rcv to so's socket,
493	* updating all sbuf field as necessary
494	*/
495	int
496	sowrite(PNATState pData, struct socket *so)
497	{
498	int n, nn;
499	struct sbuf *sb = &so->so_rcv;
500	size_t len = sb->sb_cc;
501	struct iovec iov[2];
502
503	STAM_PROFILE_START(&pData->StatIOwrite, a);
504	STAM_COUNTER_RESET(&pData->StatIOWrite_in_1);
505	STAM_COUNTER_RESET(&pData->StatIOWrite_in_1_bytes);
506	STAM_COUNTER_RESET(&pData->StatIOWrite_in_2);
507	STAM_COUNTER_RESET(&pData->StatIOWrite_in_2_1st_bytes);
508	STAM_COUNTER_RESET(&pData->StatIOWrite_in_2_2nd_bytes);
509	STAM_COUNTER_RESET(&pData->StatIOWrite_no_w);
510	STAM_COUNTER_RESET(&pData->StatIOWrite_rest);
511	STAM_COUNTER_RESET(&pData->StatIOWrite_rest_bytes);
512	DEBUG_CALL("sowrite");
513	DEBUG_ARG("so = %lx", (long)so);
514	QSOCKET_LOCK(tcb);
515	SOCKET_LOCK(so);
516	QSOCKET_UNLOCK(tcb);
517	if (so->so_urgc)
518	{
519	sosendoob(so);
520	if (sb->sb_cc == 0)
521	{
522	SOCKET_UNLOCK(so);
523	STAM_PROFILE_STOP(&pData->StatIOwrite, a);
524	return 0;
525	}
526	}
527
528	/*
529	* No need to check if there's something to write,
530	* sowrite wouldn't have been called otherwise
531	*/
532
533	len = sb->sb_cc;
534
535	iov[0].iov_base = sb->sb_rptr;
536	iov[1].iov_base = 0;
537	iov[1].iov_len = 0;
538	if (sb->sb_rptr < sb->sb_wptr)
539	{
540	iov[0].iov_len = sb->sb_wptr - sb->sb_rptr;
541	/* Should never succeed, but... */
542	if (iov[0].iov_len > len)
543	iov[0].iov_len = len;
544	n = 1;
545	}
546	else
547	{
548	iov[0].iov_len = (sb->sb_data + sb->sb_datalen) - sb->sb_rptr;
549	if (iov[0].iov_len > len)
550	iov[0].iov_len = len;
551	len -= iov[0].iov_len;
552	if (len)
553	{
554	iov[1].iov_base = sb->sb_data;
555	iov[1].iov_len = sb->sb_wptr - sb->sb_data;
556	if (iov[1].iov_len > len)
557	iov[1].iov_len = len;
558	n = 2;
559	}
560	else
561	n = 1;
562	}
563	STAM_STATS({
564	if (n == 1)
565	{
566	STAM_COUNTER_INC(&pData->StatIOWrite_in_1);
567	STAM_COUNTER_ADD(&pData->StatIOWrite_in_1_bytes, iov[0].iov_len);
568	}
569	else
570	{
571	STAM_COUNTER_INC(&pData->StatIOWrite_in_2);
572	STAM_COUNTER_ADD(&pData->StatIOWrite_in_2_1st_bytes, iov[0].iov_len);
573	STAM_COUNTER_ADD(&pData->StatIOWrite_in_2_2nd_bytes, iov[1].iov_len);
574	}
575	});
576	/* Check if there's urgent data to send, and if so, send it */
577	#ifdef HAVE_READV
578	nn = writev(so->s, (const struct iovec *)iov, n);
579	DEBUG_MISC((dfd, " ... wrote nn = %d bytes\n", nn));
580	#else
581	nn = send(so->s, iov[0].iov_base, iov[0].iov_len, 0);
582	#endif
583	/* This should never happen, but people tell me it does shrug */
584	if ( nn < 0
585	&& ( errno == EAGAIN
586	\|\| errno == EINTR
587	\|\| errno == EWOULDBLOCK))
588	{
589	SOCKET_UNLOCK(so);
590	STAM_PROFILE_STOP(&pData->StatIOwrite, a);
591	return 0;
592	}
593
594	if (nn < 0 \|\| (nn == 0 && iov[0].iov_len > 0))
595	{
596	DEBUG_MISC((dfd, " --- sowrite disconnected, so->so_state = %x, errno = %d\n",
597	so->so_state, errno));
598	sofcantsendmore(so);
599	tcp_sockclosed(pData, sototcpcb(so));
600	SOCKET_UNLOCK(so);
601	STAM_PROFILE_STOP(&pData->StatIOwrite, a);
602	return -1;
603	}
604
605	#ifndef HAVE_READV
606	if (n == 2 && nn == iov[0].iov_len)
607	{
608	int ret;
609	ret = send(so->s, iov[1].iov_base, iov[1].iov_len, 0);
610	if (ret > 0)
611	nn += ret;
612	STAM_STATS({
613	if (ret > 0 && ret != iov[1].iov_len)
614	{
615	STAM_COUNTER_INC(&pData->StatIOWrite_rest);
616	STAM_COUNTER_ADD(&pData->StatIOWrite_rest_bytes, (iov[1].iov_len - ret));
617	}
618	});
619	}
620	DEBUG_MISC((dfd, " ... wrote nn = %d bytes\n", nn));
621	#endif
622
623	/* Update sbuf */
624	sb->sb_cc -= nn;
625	sb->sb_rptr += nn;
626	if (sb->sb_rptr >= (sb->sb_data + sb->sb_datalen))
627	sb->sb_rptr -= sb->sb_datalen;
628
629	/*
630	* If in DRAIN mode, and there's no more data, set
631	* it CANTSENDMORE
632	*/
633	if ((so->so_state & SS_FWDRAIN) && sb->sb_cc == 0)
634	sofcantsendmore(so);
635
636	SOCKET_UNLOCK(so);
637	STAM_PROFILE_STOP(&pData->StatIOwrite, a);
638	return nn;
639	}
640	#else /* VBOX_WITH_SLIRP_BSD_SBUF */
641	static int
642	do_sosend(struct socket *so, int fUrg)
643	{
644	struct sbuf *sb = &so->so_rcv;
645
646	int n, len;
647
648	DEBUG_CALL("sosendoob");
649	DEBUG_ARG("so = %lx", (long)so);
650
651	len = sbuf_len(sb);
652
653	n = send(so->s, sbuf_data(sb), len, (fUrg ? MSG_OOB : 0));
654	if (n < 0)
655	Log(("NAT: Can't sent sbuf via socket.\n"));
656	if (fUrg)
657	so->so_urgc -= n;
658	if (n > 0 && n < len)
659	{
660	char *ptr;
661	char *buff;
662	buff = RTMemAlloc(len);
663	if (buff == NULL)
664	{
665	Log(("NAT: No space to allocate temporal buffer\n"));
666	return -1;
667	}
668	ptr = sbuf_data(sb);
669	memcpy(buff, &ptr[n], len - n);
670	sbuf_bcpy(sb, buff, len - n);
671	RTMemFree(buff);
672	return n;
673	}
674	sbuf_clear(sb);
675	return n;
676	}
677	int
678	sosendoob(struct socket *so)
679	{
680	return do_sosend(so, 1);
681	}
682
683	/*
684	* Write data from so_rcv to so's socket,
685	* updating all sbuf field as necessary
686	*/
687	int
688	sowrite(PNATState pData, struct socket *so)
689	{
690	return do_sosend(so, 0);
691	}
692	#endif
693
694	/*
695	* recvfrom() a UDP socket
696	*/
697	void
698	sorecvfrom(PNATState pData, struct socket *so)
699	{
700	ssize_t ret = 0;
701	struct sockaddr_in addr;
702	socklen_t addrlen = sizeof(struct sockaddr_in);
703
704	DEBUG_CALL("sorecvfrom");
705	DEBUG_ARG("so = %lx", (long)so);
706
707	if (so->so_type == IPPROTO_ICMP)
708	{
709	/* This is a "ping" reply */
710	#ifdef RT_OS_WINDOWS
711	sorecvfrom_icmp_win(pData, so);
712	#else /* RT_OS_WINDOWS */
713	sorecvfrom_icmp_unix(pData, so);
714	#endif /* !RT_OS_WINDOWS */
715	udp_detach(pData, so);
716	}
717	else
718	{
719	/* A "normal" UDP packet */
720	struct mbuf *m;
721	ssize_t len;
722	u_long n = 0;
723	int size;
724	int rc = 0;
725	static int signalled = 0;
726
727	QSOCKET_LOCK(udb);
728	SOCKET_LOCK(so);
729	QSOCKET_UNLOCK(udb);
730
731	/How many data has been received ?/
732	/*
733	* 1. calculate how much we can read
734	* 2. read as much as possible
735	* 3. attach buffer to allocated header mbuf
736	*/
737	rc = ioctlsocket(so->s, FIONREAD, &n);
738	if (rc == -1)
739	{
740	if ( errno == EAGAIN
741	\|\| errno == EWOULDBLOCK
742	\|\| errno == EINPROGRESS
743	\|\| errno == ENOTCONN)
744	return;
745	else if (signalled == 0)
746	{
747	LogRel(("NAT: can't fetch amount of bytes on socket %R[natsock], so message will be truncated.\n", so));
748	signalled = 1;
749	}
750	return;
751	}
752
753	len = sizeof(struct udpiphdr);
754	if (n > (if_mtu - len))
755	{
756	n = if_mtu - len; /* can't read than we can put in the mbuf*/
757	}
758	len += n;
759
760	size = MCLBYTES;
761	if (len + if_maxlinkhdr < MSIZE)
762	size = MCLBYTES;
763	else if (len + if_maxlinkhdr < MCLBYTES)
764	size = MCLBYTES;
765	else if (len + if_maxlinkhdr < MJUM9BYTES)
766	size = MJUM9BYTES;
767	else if (len + if_maxlinkhdr < MJUM16BYTES)
768	size = MJUM16BYTES;
769	else
770	AssertMsgFailed(("Unsupported size"));
771
772	m = m_getjcl(pData, M_NOWAIT, MT_HEADER, M_PKTHDR, size);
773	if (m == NULL)
774	return;
775
776	m->m_data += ETH_HLEN;
777	m->m_pkthdr.header = mtod(m, void *);
778	m->m_data += sizeof(struct udpiphdr);
779	ret = recvfrom(so->s, mtod(m, char *), n, 0,
780	(struct sockaddr *)&addr, &addrlen);
781	/* @todo (vvl) check which flags and type should be passed */
782	m->m_len = ret;
783	if (ret < 0)
784	{
785	u_char code = ICMP_UNREACH_PORT;
786
787	if (errno == EHOSTUNREACH)
788	code = ICMP_UNREACH_HOST;
789	else if (errno == ENETUNREACH)
790	code = ICMP_UNREACH_NET;
791
792	m_freem(pData, m);
793	if ( errno == EAGAIN
794	\|\| errno == EWOULDBLOCK
795	\|\| errno == EINPROGRESS
796	\|\| errno == ENOTCONN)
797	{
798	return;
799	}
800
801	Log2((" rx error, tx icmp ICMP_UNREACH:%i\n", code));
802	icmp_error(pData, so->so_m, ICMP_UNREACH, code, 0, strerror(errno));
803	m_freem(pData, so->so_m);
804	so->so_m = NULL;
805	}
806	else
807	{
808	/*
809	* Hack: domain name lookup will be used the most for UDP,
810	* and since they'll only be used once there's no need
811	* for the 4 minute (or whatever) timeout... So we time them
812	* out much quicker (10 seconds for now...)
813	*/
814	if (so->so_expire)
815	{
816	if (so->so_fport != RT_H2N_U16_C(53))
817	so->so_expire = curtime + SO_EXPIRE;
818	}
819	/*
820	* last argument should be changed if Slirp will inject IP attributes
821	* Note: Here we can't check if dnsproxy's sent initial request
822	*/
823	if ( pData->fUseDnsProxy
824	&& so->so_fport == RT_H2N_U16_C(53))
825	dnsproxy_answer(pData, so, m);
826
827	#if 0
828	if (m->m_len == len)
829	{
830	m_inc(m, MINCSIZE);
831	m->m_len = 0;
832	}
833	#endif
834
835	/*
836	* If this packet was destined for CTL_ADDR,
837	* make it look like that's where it came from, done by udp_output
838	*/
839	udp_output(pData, so, m, &addr);
840	SOCKET_UNLOCK(so);
841	} /* rx error */
842	} /* if ping packet */
843	}
844
845	/*
846	* sendto() a socket
847	*/
848	int
849	sosendto(PNATState pData, struct socket so, struct mbuf m)
850	{
851	int ret;
852	struct sockaddr_in *paddr;
853	struct sockaddr addr;
854	#if 0
855	struct sockaddr_in host_addr;
856	#endif
857	caddr_t buf;
858	int mlen;
859
860	DEBUG_CALL("sosendto");
861	DEBUG_ARG("so = %lx", (long)so);
862	DEBUG_ARG("m = %lx", (long)m);
863
864	memset(&addr, 0, sizeof(struct sockaddr));
865	#ifdef RT_OS_DARWIN
866	addr.sa_len = sizeof(struct sockaddr_in);
867	#endif
868	paddr = (struct sockaddr_in *)&addr;
869	paddr->sin_family = AF_INET;
870	if ((so->so_faddr.s_addr & RT_H2N_U32(pData->netmask)) == pData->special_addr.s_addr)
871	{
872	/* It's an alias */
873	uint32_t last_byte = RT_N2H_U32(so->so_faddr.s_addr) & ~pData->netmask;
874	switch(last_byte)
875	{
876	#if 0
877	/* handle this case at 'default:' */
878	case CTL_BROADCAST:
879	addr.sin_addr.s_addr = INADDR_BROADCAST;
880	/* Send the packet to host to fully emulate broadcast */
881	/** @todo r=klaus: on Linux host this causes the host to receive
882	* the packet twice for some reason. And I cannot find any place
883	* in the man pages which states that sending a broadcast does not
884	* reach the host itself. */
885	host_addr.sin_family = AF_INET;
886	host_addr.sin_port = so->so_fport;
887	host_addr.sin_addr = our_addr;
888	sendto(so->s, m->m_data, m->m_len, 0,
889	(struct sockaddr *)&host_addr, sizeof (struct sockaddr));
890	break;
891	#endif
892	case CTL_DNS:
893	case CTL_ALIAS:
894	default:
895	if (last_byte == ~pData->netmask)
896	paddr->sin_addr.s_addr = INADDR_BROADCAST;
897	else
898	paddr->sin_addr = loopback_addr;
899	break;
900	}
901	}
902	else
903	paddr->sin_addr = so->so_faddr;
904	paddr->sin_port = so->so_fport;
905
906	DEBUG_MISC((dfd, " sendto()ing, addr.sin_port=%d, addr.sin_addr.s_addr=%.16s\n",
907	RT_N2H_U16(paddr->sin_port), inet_ntoa(paddr->sin_addr)));
908
909	/* Don't care what port we get */
910	mlen = m_length(m, NULL);
911	buf = RTMemAlloc(mlen);
912	if (buf == NULL)
913	{
914	return -1;
915	}
916	m_copydata(m, 0, mlen, buf);
917	ret = sendto(so->s, buf, mlen, 0,
918	(struct sockaddr *)&addr, sizeof (struct sockaddr));
919	RTMemFree(buf);
920	if (ret < 0)
921	{
922	Log2(("UDP: sendto fails (%s)\n", strerror(errno)));
923	return -1;
924	}
925
926	/*
927	* Kill the socket if there's no reply in 4 minutes,
928	* but only if it's an expirable socket
929	*/
930	if (so->so_expire)
931	so->so_expire = curtime + SO_EXPIRE;
932	so->so_state = SS_ISFCONNECTED; /* So that it gets select()ed */
933	return 0;
934	}
935
936	/*
937	* XXX This should really be tcp_listen
938	*/
939	struct socket *
940	solisten(PNATState pData, u_int32_t bind_addr, u_int port, u_int32_t laddr, u_int lport, int flags)
941	{
942	struct sockaddr_in addr;
943	struct socket *so;
944	socklen_t addrlen = sizeof(addr);
945	int s, opt = 1;
946	int status;
947
948	DEBUG_CALL("solisten");
949	DEBUG_ARG("port = %d", port);
950	DEBUG_ARG("laddr = %x", laddr);
951	DEBUG_ARG("lport = %d", lport);
952	DEBUG_ARG("flags = %x", flags);
953
954	if ((so = socreate()) == NULL)
955	{
956	/* RTMemFree(so); Not sofree() ??? free(NULL) == NOP */
957	return NULL;
958	}
959
960	/* Don't tcp_attach... we don't need so_snd nor so_rcv */
961	if ((so->so_tcpcb = tcp_newtcpcb(pData, so)) == NULL)
962	{
963	RTMemFree(so);
964	return NULL;
965	}
966
967	SOCKET_LOCK_CREATE(so);
968	SOCKET_LOCK(so);
969	QSOCKET_LOCK(tcb);
970	insque(pData, so,&tcb);
971	NSOCK_INC();
972	QSOCKET_UNLOCK(tcb);
973
974	/*
975	* SS_FACCEPTONCE sockets must time out.
976	*/
977	if (flags & SS_FACCEPTONCE)
978	so->so_tcpcb->t_timer[TCPT_KEEP] = TCPTV_KEEP_INIT*2;
979
980	so->so_state = (SS_FACCEPTCONN\|flags);
981	so->so_lport = lport; /* Kept in network format */
982	so->so_laddr.s_addr = laddr; /* Ditto */
983
984	memset(&addr, 0, sizeof(addr));
985	#ifdef RT_OS_DARWIN
986	addr.sin_len = sizeof(addr);
987	#endif
988	addr.sin_family = AF_INET;
989	addr.sin_addr.s_addr = bind_addr;
990	addr.sin_port = port;
991
992	if ( ((s = socket(AF_INET, SOCK_STREAM, 0)) < 0)
993	\|\| (setsockopt(s, SOL_SOCKET, SO_REUSEADDR,(char *)&opt, sizeof(int)) < 0)
994	\|\| (bind(s,(struct sockaddr *)&addr, sizeof(addr)) < 0)
995	\|\| (listen(s, 1) < 0))
996	{
997	#ifdef RT_OS_WINDOWS
998	int tmperrno = WSAGetLastError(); /* Don't clobber the real reason we failed */
999	closesocket(s);
1000	QSOCKET_LOCK(tcb);
1001	sofree(pData, so);
1002	QSOCKET_UNLOCK(tcb);
1003	/* Restore the real errno */
1004	WSASetLastError(tmperrno);
1005	#else
1006	int tmperrno = errno; /* Don't clobber the real reason we failed */
1007	close(s);
1008	QSOCKET_LOCK(tcb);
1009	sofree(pData, so);
1010	QSOCKET_UNLOCK(tcb);
1011	/* Restore the real errno */
1012	errno = tmperrno;
1013	#endif
1014	return NULL;
1015	}
1016	fd_nonblock(s);
1017	setsockopt(s, SOL_SOCKET, SO_OOBINLINE,(char *)&opt, sizeof(int));
1018
1019	getsockname(s,(struct sockaddr *)&addr,&addrlen);
1020	so->so_fport = addr.sin_port;
1021	/* set socket buffers */
1022	opt = pData->socket_rcv;
1023	status = setsockopt(s, SOL_SOCKET, SO_RCVBUF, (char *)&opt, sizeof(int));
1024	if (status < 0)
1025	{
1026	LogRel(("NAT: Error(%d) while setting RCV capacity to (%d)\n", errno, opt));
1027	goto no_sockopt;
1028	}
1029	opt = pData->socket_snd;
1030	status = setsockopt(s, SOL_SOCKET, SO_SNDBUF, (char *)&opt, sizeof(int));
1031	if (status < 0)
1032	{
1033	LogRel(("NAT: Error(%d) while setting SND capacity to (%d)\n", errno, opt));
1034	goto no_sockopt;
1035	}
1036	no_sockopt:
1037	if (addr.sin_addr.s_addr == 0 \|\| addr.sin_addr.s_addr == loopback_addr.s_addr)
1038	so->so_faddr = alias_addr;
1039	else
1040	so->so_faddr = addr.sin_addr;
1041
1042	so->s = s;
1043	SOCKET_UNLOCK(so);
1044	return so;
1045	}
1046
1047	/*
1048	* Data is available in so_rcv
1049	* Just write() the data to the socket
1050	* XXX not yet...
1051	*/
1052	void
1053	sorwakeup(struct socket *so)
1054	{
1055	#if 0
1056	sowrite(so);
1057	FD_CLR(so->s,&writefds);
1058	#endif
1059	}
1060
1061	/*
1062	* Data has been freed in so_snd
1063	* We have room for a read() if we want to
1064	* For now, don't read, it'll be done in the main loop
1065	*/
1066	void
1067	sowwakeup(struct socket *so)
1068	{
1069	}
1070
1071	/*
1072	* Various session state calls
1073	* XXX Should be #define's
1074	* The socket state stuff needs work, these often get call 2 or 3
1075	* times each when only 1 was needed
1076	*/
1077	void
1078	soisfconnecting(struct socket *so)
1079	{
1080	so->so_state &= ~(SS_NOFDREF\|SS_ISFCONNECTED\|SS_FCANTRCVMORE\|
1081	SS_FCANTSENDMORE\|SS_FWDRAIN);
1082	so->so_state \|= SS_ISFCONNECTING; /* Clobber other states */
1083	}
1084
1085	void
1086	soisfconnected(struct socket *so)
1087	{
1088	so->so_state &= ~(SS_ISFCONNECTING\|SS_FWDRAIN\|SS_NOFDREF);
1089	so->so_state \|= SS_ISFCONNECTED; /* Clobber other states */
1090	}
1091
1092	void
1093	sofcantrcvmore(struct socket *so)
1094	{
1095	if ((so->so_state & SS_NOFDREF) == 0)
1096	{
1097	shutdown(so->s, 0);
1098	}
1099	so->so_state &= ~(SS_ISFCONNECTING);
1100	if (so->so_state & SS_FCANTSENDMORE)
1101	so->so_state = SS_NOFDREF; /* Don't select it */
1102	/* XXX close() here as well? */
1103	else
1104	so->so_state \|= SS_FCANTRCVMORE;
1105	}
1106
1107	void
1108	sofcantsendmore(struct socket *so)
1109	{
1110	if ((so->so_state & SS_NOFDREF) == 0)
1111	shutdown(so->s, 1); /* send FIN to fhost */
1112
1113	so->so_state &= ~(SS_ISFCONNECTING);
1114	if (so->so_state & SS_FCANTRCVMORE)
1115	so->so_state = SS_NOFDREF; /* as above */
1116	else
1117	so->so_state \|= SS_FCANTSENDMORE;
1118	}
1119
1120	void
1121	soisfdisconnected(struct socket *so)
1122	{
1123	#if 0
1124	so->so_state &= ~(SS_ISFCONNECTING\|SS_ISFCONNECTED);
1125	close(so->s);
1126	so->so_state = SS_ISFDISCONNECTED;
1127	/*
1128	* XXX Do nothing ... ?
1129	*/
1130	#endif
1131	}
1132
1133	/*
1134	* Set write drain mode
1135	* Set CANTSENDMORE once all data has been write()n
1136	*/
1137	void
1138	sofwdrain(struct socket *so)
1139	{
1140	if (SBUF_LEN(&so->so_rcv))
1141	so->so_state \|= SS_FWDRAIN;
1142	else
1143	sofcantsendmore(so);
1144	}
1145
1146	static void
1147	send_icmp_to_guest(PNATState pData, char buff, size_t len, struct socket so, const struct sockaddr_in *addr)
1148	{
1149	struct ip *ip;
1150	uint32_t dst, src;
1151	char ip_copy[256];
1152	struct icmp *icp;
1153	int old_ip_len = 0;
1154	int hlen, original_hlen = 0;
1155	struct mbuf *m;
1156	struct icmp_msg *icm;
1157	uint8_t proto;
1158	int type = 0;
1159
1160	ip = (struct ip *)buff;
1161	/* Fix ip->ip_len to contain the total packet length including the header
1162	* in _host_ byte order for all OSes. On Darwin, that value already is in
1163	* host byte order. Solaris and Darwin report only the payload. */
1164	#ifndef RT_OS_DARWIN
1165	ip->ip_len = RT_N2H_U16(ip->ip_len);
1166	#endif
1167	hlen = (ip->ip_hl << 2);
1168	#if defined(RT_OS_SOLARIS) \|\| defined(RT_OS_DARWIN)
1169	ip->ip_len += hlen;
1170	#endif
1171	if (ip->ip_len < hlen + ICMP_MINLEN)
1172	{
1173	Log(("send_icmp_to_guest: ICMP header is too small to understand which type/subtype of the datagram\n"));
1174	return;
1175	}
1176	icp = (struct icmp )((char )ip + hlen);
1177
1178	Log(("ICMP:received msg(t:%d, c:%d)\n", icp->icmp_type, icp->icmp_code));
1179	if ( icp->icmp_type != ICMP_ECHOREPLY
1180	&& icp->icmp_type != ICMP_TIMXCEED
1181	&& icp->icmp_type != ICMP_UNREACH)
1182	{
1183	return;
1184	}
1185
1186	/*
1187	* ICMP_ECHOREPLY, ICMP_TIMXCEED, ICMP_UNREACH minimal header size is
1188	* ICMP_ECHOREPLY assuming data 0
1189	* icmp_{type(8), code(8), cksum(16),identifier(16),seqnum(16)}
1190	*/
1191	if (ip->ip_len < hlen + 8)
1192	{
1193	Log(("send_icmp_to_guest: NAT accept ICMP_{ECHOREPLY, TIMXCEED, UNREACH} the minimum size is 64 (see rfc792)\n"));
1194	return;
1195	}
1196
1197	type = icp->icmp_type;
1198	if ( type == ICMP_TIMXCEED
1199	\|\| type == ICMP_UNREACH)
1200	{
1201	/*
1202	* ICMP_TIMXCEED, ICMP_UNREACH minimal header size is
1203	* icmp_{type(8), code(8), cksum(16),unused(32)} + IP header + 64 bit of original datagram
1204	*/
1205	if (ip->ip_len < hlen + 2*8 + sizeof(struct ip))
1206	{
1207	Log(("send_icmp_to_guest: NAT accept ICMP_{TIMXCEED, UNREACH} the minimum size of ipheader + 64 bit of data (see rfc792)\n"));
1208	return;
1209	}
1210	ip = &icp->icmp_ip;
1211	}
1212
1213	icm = icmp_find_original_mbuf(pData, ip);
1214	if (icm == NULL)
1215	{
1216	Log(("NAT: Can't find the corresponding packet for the received ICMP\n"));
1217	return;
1218	}
1219
1220	m = icm->im_m;
1221	Assert(m != NULL);
1222
1223	src = addr->sin_addr.s_addr;
1224	if (type == ICMP_ECHOREPLY)
1225	{
1226	struct ip ip0 = mtod(m, struct ip );
1227	struct icmp icp0 = (struct icmp )((char *)ip0 + (ip0->ip_hl << 2));
1228	if (icp0->icmp_type != ICMP_ECHO)
1229	{
1230	Log(("NAT: we haven't found echo for this reply\n"));
1231	return;
1232	}
1233	/*
1234	* while combining buffer to send (see ip_icmp.c) we control ICMP header only,
1235	* IP header combined by OS network stack, our local copy of IP header contians values
1236	* in host byte order so no byte order conversion is required. IP headers fields are converting
1237	* in ip_output0 routine only.
1238	*/
1239	if ( (ip->ip_len - hlen)
1240	!= (ip0->ip_len - (ip0->ip_hl << 2)))
1241	{
1242	Log(("NAT: ECHO(%d) lenght doesn't match ECHOREPLY(%d)\n",
1243	(ip->ip_len - hlen), (ip0->ip_len - (ip0->ip_hl << 2))));
1244	return;
1245	}
1246	}
1247
1248	/* ip points on origianal ip header */
1249	ip = mtod(m, struct ip *);
1250	proto = ip->ip_p;
1251	/* Now ip is pointing on header we've sent from guest */
1252	if ( icp->icmp_type == ICMP_TIMXCEED
1253	\|\| icp->icmp_type == ICMP_UNREACH)
1254	{
1255	old_ip_len = (ip->ip_hl << 2) + 64;
1256	if (old_ip_len > sizeof(ip_copy))
1257	old_ip_len = sizeof(ip_copy);
1258	memcpy(ip_copy, ip, old_ip_len);
1259	}
1260
1261	/* source address from original IP packet*/
1262	dst = ip->ip_src.s_addr;
1263
1264	/* overide ther tail of old packet */
1265	ip = mtod(m, struct ip ); / ip is from mbuf we've overrided */
1266	original_hlen = ip->ip_hl << 2;
1267	/* saves original ip header and options */
1268	m_copyback(pData, m, original_hlen, len - hlen, buff + hlen);
1269	ip->ip_len = m_length(m, NULL);
1270	ip->ip_p = IPPROTO_ICMP; /* the original package could be whatever, but we're response via ICMP*/
1271
1272	icp = (struct icmp )((char )ip + (ip->ip_hl << 2));
1273	type = icp->icmp_type;
1274	if ( type == ICMP_TIMXCEED
1275	\|\| type == ICMP_UNREACH)
1276	{
1277	/* according RFC 793 error messages required copy of initial IP header + 64 bit */
1278	memcpy(&icp->icmp_ip, ip_copy, old_ip_len);
1279	ip->ip_tos = ((ip->ip_tos & 0x1E) \| 0xC0); /* high priority for errors */
1280	}
1281
1282	ip->ip_src.s_addr = src;
1283	ip->ip_dst.s_addr = dst;
1284	icmp_reflect(pData, m);
1285	LIST_REMOVE(icm, im_list);
1286	/* Don't call m_free here*/
1287
1288	if ( type == ICMP_TIMXCEED
1289	\|\| type == ICMP_UNREACH)
1290	{
1291	icm->im_so->so_m = NULL;
1292	switch (proto)
1293	{
1294	case IPPROTO_UDP:
1295	/XXX: so->so_m already freed so we shouldn't call sofree /
1296	udp_detach(pData, icm->im_so);
1297	break;
1298	case IPPROTO_TCP:
1299	/close tcp should be here /
1300	break;
1301	default:
1302	/* do nothing */
1303	break;
1304	}
1305	}
1306	RTMemFree(icm);
1307	}
1308
1309	#ifdef RT_OS_WINDOWS
1310	static void
1311	sorecvfrom_icmp_win(PNATState pData, struct socket *so)
1312	{
1313	int len;
1314	int i;
1315	struct ip *ip;
1316	struct mbuf *m;
1317	struct icmp *icp;
1318	struct icmp_msg *icm;
1319	struct ip ip_broken; / ICMP returns header + 64 bit of packet */
1320	uint32_t src;
1321	ICMP_ECHO_REPLY *icr;
1322	int hlen = 0;
1323	int data_len = 0;
1324	int nbytes = 0;
1325	u_char code = ~0;
1326	int out_len;
1327	int size;
1328
1329	len = pData->pfIcmpParseReplies(pData->pvIcmpBuffer, pData->szIcmpBuffer);
1330	if (len < 0)
1331	{
1332	LogRel(("NAT: Error (%d) occurred on ICMP receiving\n", GetLastError()));
1333	return;
1334	}
1335	if (len == 0)
1336	return; /* no error */
1337
1338	icr = (ICMP_ECHO_REPLY *)pData->pvIcmpBuffer;
1339	for (i = 0; i < len; ++i)
1340	{
1341	switch(icr[i].Status)
1342	{
1343	case IP_DEST_HOST_UNREACHABLE:
1344	code = (code != ~0 ? code : ICMP_UNREACH_HOST);
1345	case IP_DEST_NET_UNREACHABLE:
1346	code = (code != ~0 ? code : ICMP_UNREACH_NET);
1347	case IP_DEST_PROT_UNREACHABLE:
1348	code = (code != ~0 ? code : ICMP_UNREACH_PROTOCOL);
1349	/* UNREACH error inject here */
1350	case IP_DEST_PORT_UNREACHABLE:
1351	code = (code != ~0 ? code : ICMP_UNREACH_PORT);
1352	icmp_error(pData, so->so_m, ICMP_UNREACH, code, 0, "Error occurred!!!");
1353	m_freem(pData, so->so_m);
1354	so->so_m = NULL;
1355	break;
1356	case IP_SUCCESS: /* echo replied */
1357	out_len = ETH_HLEN + sizeof(struct ip) + 8;
1358	size;
1359	size = MCLBYTES;
1360	if (out_len < MSIZE)
1361	size = MCLBYTES;
1362	else if (out_len < MCLBYTES)
1363	size = MCLBYTES;
1364	else if (out_len < MJUM9BYTES)
1365	size = MJUM9BYTES;
1366	else if (out_len < MJUM16BYTES)
1367	size = MJUM16BYTES;
1368	else
1369	AssertMsgFailed(("Unsupported size"));
1370
1371	m = m_getjcl(pData, M_NOWAIT, MT_HEADER, M_PKTHDR, size);
1372	if (m == NULL)
1373	return;
1374	m->m_len = 0;
1375	m->m_data += if_maxlinkhdr;
1376	ip = mtod(m, struct ip *);
1377	ip->ip_src.s_addr = icr[i].Address;
1378	ip->ip_p = IPPROTO_ICMP;
1379	ip->ip_dst.s_addr = so->so_laddr.s_addr; /XXX: still the hack/
1380	data_len = sizeof(struct ip);
1381	ip->ip_hl = data_len >> 2; /* requiered for icmp_reflect, no IP options */
1382	ip->ip_ttl = icr[i].Options.Ttl;
1383
1384	icp = (struct icmp )&ip[1]; / no options */
1385	icp->icmp_type = ICMP_ECHOREPLY;
1386	icp->icmp_code = 0;
1387	icp->icmp_id = so->so_icmp_id;
1388	icp->icmp_seq = so->so_icmp_seq;
1389
1390	data_len += ICMP_MINLEN;
1391
1392	hlen = (ip->ip_hl << 2);
1393	m->m_pkthdr.header = mtod(m, void *);
1394	m->m_len = data_len;
1395
1396	m_copyback(pData, m, hlen + 8, icr[i].DataSize, icr[i].Data);
1397
1398	data_len += icr[i].DataSize;
1399
1400	ip->ip_len = data_len;
1401	m->m_len = ip->ip_len;
1402
1403	icmp_reflect(pData, m);
1404	break;
1405	case IP_TTL_EXPIRED_TRANSIT: /* TTL expired */
1406
1407	ip_broken = icr[i].Data;
1408	icm = icmp_find_original_mbuf(pData, ip_broken);
1409	if (icm == NULL) {
1410	Log(("ICMP: can't find original package (first double word %x)\n", (uint32_t )ip_broken));
1411	return;
1412	}
1413	m = icm->im_m;
1414	ip = mtod(m, struct ip *);
1415	ip->ip_ttl = icr[i].Options.Ttl;
1416	src = ip->ip_src.s_addr;
1417	ip->ip_dst.s_addr = src;
1418	ip->ip_dst.s_addr = icr[i].Address;
1419
1420	hlen = (ip->ip_hl << 2);
1421	icp = (struct icmp )((char )ip + hlen);
1422	ip_broken->ip_src.s_addr = src; /it packet sent from host not from guest/
1423	data_len = (ip_broken->ip_hl << 2) + 64;
1424
1425	m->m_len = data_len;
1426	m->m_pkthdr.header = mtod(m, void *);
1427	m_copyback(pData, m, ip->ip_hl >> 2, icr[i].DataSize, icr[i].Data);
1428	icmp_reflect(pData, m);
1429	break;
1430	default:
1431	Log(("ICMP(default): message with Status: %x was received from %x\n", icr[i].Status, icr[i].Address));
1432	break;
1433	}
1434	}
1435	}
1436	#else /* !RT_OS_WINDOWS */
1437	static void sorecvfrom_icmp_unix(PNATState pData, struct socket *so)
1438	{
1439	struct sockaddr_in addr;
1440	socklen_t addrlen = sizeof(struct sockaddr_in);
1441	struct ip ip;
1442	char *buff;
1443	int len = 0;
1444
1445	/* 1- step: read the ip header */
1446	len = recvfrom(so->s, &ip, sizeof(struct ip), MSG_PEEK,
1447	(struct sockaddr *)&addr, &addrlen);
1448	if ( len < 0
1449	&& ( errno == EAGAIN
1450	\|\| errno == EWOULDBLOCK
1451	\|\| errno == EINPROGRESS
1452	\|\| errno == ENOTCONN))
1453	{
1454	Log(("sorecvfrom_icmp_unix: 1 - step can't read IP datagramm (would block)\n"));
1455	return;
1456	}
1457
1458	if ( len < sizeof(struct ip)
1459	\|\| len < 0
1460	\|\| len == 0)
1461	{
1462	u_char code;
1463	code = ICMP_UNREACH_PORT;
1464
1465	if (errno == EHOSTUNREACH)
1466	code = ICMP_UNREACH_HOST;
1467	else if (errno == ENETUNREACH)
1468	code = ICMP_UNREACH_NET;
1469
1470	LogRel((" udp icmp rx errno = %d-%s\n",
1471	errno, strerror(errno)));
1472	icmp_error(pData, so->so_m, ICMP_UNREACH, code, 0, strerror(errno));
1473	m_freem(pData, so->so_m);
1474	so->so_m = NULL;
1475	Log(("sorecvfrom_icmp_unix: 1 - step can't read IP datagramm \n"));
1476	return;
1477	}
1478	/* basic check of IP header */
1479	if ( ip.ip_v != IPVERSION
1480	# ifndef RT_OS_DARWIN
1481	\|\| ip.ip_p != IPPROTO_ICMP
1482	# endif
1483	)
1484	{
1485	Log(("sorecvfrom_icmp_unix: 1 - step IP isn't IPv4 \n"));
1486	return;
1487	}
1488	# ifndef RT_OS_DARWIN
1489	/* Darwin reports the IP length already in host byte order. */
1490	ip.ip_len = RT_N2H_U16(ip.ip_len);
1491	# endif
1492	# if defined(RT_OS_SOLARIS) \|\| defined(RT_OS_DARWIN)
1493	/* Solaris and Darwin report the payload only */
1494	ip.ip_len += (ip.ip_hl << 2);
1495	# endif
1496	/* Note: ip->ip_len in host byte order (all OS) */
1497	len = ip.ip_len;
1498	buff = RTMemAlloc(len);
1499	if (buff == NULL)
1500	{
1501	Log(("sorecvfrom_icmp_unix: 1 - step can't allocate enought room for datagram\n"));
1502	return;
1503	}
1504	/* 2 - step: we're reading rest of the datagramm to the buffer */
1505	addrlen = sizeof(struct sockaddr_in);
1506	memset(&addr, 0, addrlen);
1507	len = recvfrom(so->s, buff, len, 0,
1508	(struct sockaddr *)&addr, &addrlen);
1509	if ( len < 0
1510	&& ( errno == EAGAIN
1511	\|\| errno == EWOULDBLOCK
1512	\|\| errno == EINPROGRESS
1513	\|\| errno == ENOTCONN))
1514	{
1515	Log(("sorecvfrom_icmp_unix: 2 - step can't read IP body (would block expected:%d)\n",
1516	ip.ip_len));
1517	RTMemFree(buff);
1518	return;
1519	}
1520	if ( len < 0
1521	\|\| len == 0)
1522	{
1523	Log(("sorecvfrom_icmp_unix: 2 - step read of the rest of datagramm is fallen (errno:%d, len:%d expected: %d)\n",
1524	errno, len, (ip.ip_len - sizeof(struct ip))));
1525	RTMemFree(buff);
1526	return;
1527	}
1528	/* len is modified in 2nd read, when the rest of the datagramm was read */
1529	send_icmp_to_guest(pData, buff, len, so, &addr);
1530	RTMemFree(buff);
1531	}
1532	#endif /* !RT_OS_WINDOWS */

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

source: vbox/trunk/src/VBox/Devices/Network/slirp/socket.c@ 33978

Download in other formats: