slirp.c@ 49435

Last change on this file since 49435 was 49347, checked in by vboxsync, 11 years ago
NAT: Use RW critsect to protect the handler chain list against concurrent access
Property svn:eol-style set to `native` Property svn:keywords set to `Author Date Id Revision`
File size: 64.1 KB

Line
1	/* $Id: slirp.c 49347 2013-10-31 14:08:05Z vboxsync $ */
2	/** @file
3	* NAT - slirp glue.
4	*/
5
6	/*
7	* Copyright (C) 2006-2012 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	* libslirp glue
22	*
23	* Copyright (c) 2004-2008 Fabrice Bellard
24	*
25	* Permission is hereby granted, free of charge, to any person obtaining a copy
26	* of this software and associated documentation files (the "Software"), to deal
27	* in the Software without restriction, including without limitation the rights
28	* to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
29	* copies of the Software, and to permit persons to whom the Software is
30	* furnished to do so, subject to the following conditions:
31	*
32	* The above copyright notice and this permission notice shall be included in
33	* all copies or substantial portions of the Software.
34	*
35	* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
36	* IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
37	* FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
38	* THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
39	* LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
40	* OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
41	* THE SOFTWARE.
42	*/
43
44	#include "slirp.h"
45	#ifdef RT_OS_OS2
46	# include <paths.h>
47	#endif
48
49	#include <VBox/err.h>
50	#include <VBox/vmm/dbgf.h>
51	#include <VBox/vmm/pdmdrv.h>
52	#include <iprt/assert.h>
53	#include <iprt/file.h>
54	#ifndef RT_OS_WINDOWS
55	# include <sys/ioctl.h>
56	# include <poll.h>
57	# include <netinet/in.h>
58	#else
59	# include <Winnls.h>
60	# define _WINSOCK2API_
61	# include <IPHlpApi.h>
62	#endif
63	#include <alias.h>
64
65	#ifndef RT_OS_WINDOWS
66
67	# define DO_ENGAGE_EVENT1(so, fdset, label) \
68	do { \
69	if ( so->so_poll_index != -1 \
70	&& so->s == polls[so->so_poll_index].fd) \
71	{ \
72	polls[so->so_poll_index].events \|= N_(fdset ## _poll); \
73	break; \
74	} \
75	AssertRelease(poll_index < (nfds)); \
76	AssertRelease(poll_index >= 0 && poll_index < (nfds)); \
77	polls[poll_index].fd = (so)->s; \
78	(so)->so_poll_index = poll_index; \
79	polls[poll_index].events = N_(fdset ## _poll); \
80	polls[poll_index].revents = 0; \
81	poll_index++; \
82	} while (0)
83
84	# define DO_ENGAGE_EVENT2(so, fdset1, fdset2, label) \
85	do { \
86	if ( so->so_poll_index != -1 \
87	&& so->s == polls[so->so_poll_index].fd) \
88	{ \
89	polls[so->so_poll_index].events \|= \
90	N_(fdset1 ## _poll) \| N_(fdset2 ## _poll); \
91	break; \
92	} \
93	AssertRelease(poll_index < (nfds)); \
94	polls[poll_index].fd = (so)->s; \
95	(so)->so_poll_index = poll_index; \
96	polls[poll_index].events = \
97	N_(fdset1 ## _poll) \| N_(fdset2 ## _poll); \
98	poll_index++; \
99	} while (0)
100
101	# define DO_POLL_EVENTS(rc, error, so, events, label) do {} while (0)
102
103	/*
104	* DO_CHECK_FD_SET is used in dumping events on socket, including POLLNVAL.
105	* gcc warns about attempts to log POLLNVAL so construction in a last to lines
106	* used to catch POLLNVAL while logging and return false in case of error while
107	* normal usage.
108	*/
109	# define DO_CHECK_FD_SET(so, events, fdset) \
110	( ((so)->so_poll_index != -1) \
111	&& ((so)->so_poll_index <= ndfs) \
112	&& ((so)->s == polls[so->so_poll_index].fd) \
113	&& (polls[(so)->so_poll_index].revents & N_(fdset ## _poll)) \
114	&& ( N_(fdset ## _poll) == POLLNVAL \
115	\|\| !(polls[(so)->so_poll_index].revents & POLLNVAL)))
116
117	/* specific for Windows Winsock API */
118	# define DO_WIN_CHECK_FD_SET(so, events, fdset) 0
119
120	# ifndef RT_OS_LINUX
121	# define readfds_poll (POLLRDNORM)
122	# define writefds_poll (POLLWRNORM)
123	# else
124	# define readfds_poll (POLLIN)
125	# define writefds_poll (POLLOUT)
126	# endif
127	# define xfds_poll (POLLPRI)
128	# define closefds_poll (POLLHUP)
129	# define rderr_poll (POLLERR)
130	# if 0 /* unused yet */
131	# define rdhup_poll (POLLHUP)
132	# define nval_poll (POLLNVAL)
133	# endif
134
135	# define ICMP_ENGAGE_EVENT(so, fdset) \
136	do { \
137	if (pData->icmp_socket.s != -1) \
138	DO_ENGAGE_EVENT1((so), fdset, ICMP); \
139	} while (0)
140
141	#else /* RT_OS_WINDOWS */
142
143	/*
144	* On Windows, we will be notified by IcmpSendEcho2() when the response arrives.
145	* So no call to WSAEventSelect necessary.
146	*/
147	# define ICMP_ENGAGE_EVENT(so, fdset) do {} while (0)
148
149	/*
150	* On Windows we use FD_ALL_EVENTS to ensure that we don't miss any event.
151	*/
152	# define DO_ENGAGE_EVENT1(so, fdset1, label) \
153	do { \
154	rc = WSAEventSelect((so)->s, VBOX_SOCKET_EVENT, FD_ALL_EVENTS); \
155	if (rc == SOCKET_ERROR) \
156	{ \
157	/* This should not happen */ \
158	error = WSAGetLastError(); \
159	LogRel(("WSAEventSelect (" #label ") error %d (so=%x, socket=%s, event=%x)\n", \
160	error, (so), (so)->s, VBOX_SOCKET_EVENT)); \
161	} \
162	} while (0); \
163	CONTINUE(label)
164
165	# define DO_ENGAGE_EVENT2(so, fdset1, fdset2, label) \
166	DO_ENGAGE_EVENT1((so), (fdset1), label)
167
168	# define DO_POLL_EVENTS(rc, error, so, events, label) \
169	(rc) = WSAEnumNetworkEvents((so)->s, VBOX_SOCKET_EVENT, (events)); \
170	if ((rc) == SOCKET_ERROR) \
171	{ \
172	(error) = WSAGetLastError(); \
173	LogRel(("WSAEnumNetworkEvents %R[natsock] " #label " error %d\n", (so), (error))); \
174	LogFunc(("WSAEnumNetworkEvents %R[natsock] " #label " error %d\n", (so), (error))); \
175	CONTINUE(label); \
176	}
177
178	# define acceptds_win FD_ACCEPT
179	# define acceptds_win_bit FD_ACCEPT_BIT
180	# define readfds_win FD_READ
181	# define readfds_win_bit FD_READ_BIT
182	# define writefds_win FD_WRITE
183	# define writefds_win_bit FD_WRITE_BIT
184	# define xfds_win FD_OOB
185	# define xfds_win_bit FD_OOB_BIT
186	# define closefds_win FD_CLOSE
187	# define closefds_win_bit FD_CLOSE_BIT
188	# define connectfds_win FD_CONNECT
189	# define connectfds_win_bit FD_CONNECT_BIT
190
191	# define closefds_win FD_CLOSE
192	# define closefds_win_bit FD_CLOSE_BIT
193
194	# define DO_CHECK_FD_SET(so, events, fdset) \
195	(((events).lNetworkEvents & fdset ## _win) && ((events).iErrorCode[fdset ## _win_bit] == 0))
196
197	# define DO_WIN_CHECK_FD_SET(so, events, fdset) DO_CHECK_FD_SET((so), (events), fdset)
198	# define DO_UNIX_CHECK_FD_SET(so, events, fdset) 1 /specific for Unix API /
199
200	#endif /* RT_OS_WINDOWS */
201
202	#define TCP_ENGAGE_EVENT1(so, fdset) \
203	DO_ENGAGE_EVENT1((so), fdset, tcp)
204
205	#define TCP_ENGAGE_EVENT2(so, fdset1, fdset2) \
206	DO_ENGAGE_EVENT2((so), fdset1, fdset2, tcp)
207
208	#ifdef RT_OS_WINDOWS
209	# define WIN_TCP_ENGAGE_EVENT2(so, fdset, fdset2) TCP_ENGAGE_EVENT2(so, fdset1, fdset2)
210	#endif
211
212	#define UDP_ENGAGE_EVENT(so, fdset) \
213	DO_ENGAGE_EVENT1((so), fdset, udp)
214
215	#define POLL_TCP_EVENTS(rc, error, so, events) \
216	DO_POLL_EVENTS((rc), (error), (so), (events), tcp)
217
218	#define POLL_UDP_EVENTS(rc, error, so, events) \
219	DO_POLL_EVENTS((rc), (error), (so), (events), udp)
220
221	#define CHECK_FD_SET(so, events, set) \
222	(DO_CHECK_FD_SET((so), (events), set))
223
224	#define WIN_CHECK_FD_SET(so, events, set) \
225	(DO_WIN_CHECK_FD_SET((so), (events), set))
226
227	/*
228	* Loging macros
229	*/
230	#if VBOX_WITH_DEBUG_NAT_SOCKETS
231	# if defined(RT_OS_WINDOWS)
232	# define DO_LOG_NAT_SOCK(so, proto, winevent, r_fdset, w_fdset, x_fdset) \
233	do { \
234	LogRel((" " #proto " %R[natsock] %R[natwinnetevents]\n", (so), (winevent))); \
235	} while (0)
236	# else /* !RT_OS_WINDOWS */
237	# define DO_LOG_NAT_SOCK(so, proto, winevent, r_fdset, w_fdset, x_fdset) \
238	do { \
239	LogRel((" " #proto " %R[natsock] %s %s %s er: %s, %s, %s\n", (so), \
240	CHECK_FD_SET(so, ign ,r_fdset) ? "READ":"", \
241	CHECK_FD_SET(so, ign, w_fdset) ? "WRITE":"", \
242	CHECK_FD_SET(so, ign, x_fdset) ? "OOB":"", \
243	CHECK_FD_SET(so, ign, rderr) ? "RDERR":"", \
244	CHECK_FD_SET(so, ign, rdhup) ? "RDHUP":"", \
245	CHECK_FD_SET(so, ign, nval) ? "RDNVAL":"")); \
246	} while (0)
247	# endif /* !RT_OS_WINDOWS */
248	#else /* !VBOX_WITH_DEBUG_NAT_SOCKETS */
249	# define DO_LOG_NAT_SOCK(so, proto, winevent, r_fdset, w_fdset, x_fdset) do {} while (0)
250	#endif /* !VBOX_WITH_DEBUG_NAT_SOCKETS */
251
252	#define LOG_NAT_SOCK(so, proto, winevent, r_fdset, w_fdset, x_fdset) \
253	DO_LOG_NAT_SOCK((so), proto, (winevent), r_fdset, w_fdset, x_fdset)
254
255	static void activate_port_forwarding(PNATState, const uint8_t *pEther);
256
257	static const uint8_t special_ethaddr[6] =
258	{
259	0x52, 0x54, 0x00, 0x12, 0x35, 0x00
260	};
261
262	static const uint8_t broadcast_ethaddr[6] =
263	{
264	0xff, 0xff, 0xff, 0xff, 0xff, 0xff
265	};
266
267	const uint8_t zerro_ethaddr[6] =
268	{
269	0x0, 0x0, 0x0, 0x0, 0x0, 0x0
270	};
271
272	/**
273	* This helper routine do the checks in descriptions to
274	* ''fUnderPolling'' and ''fShouldBeRemoved'' flags
275	* @returns 1 if socket removed and 0 if no changes was made.
276	*/
277	static int slirpVerifyAndFreeSocket(PNATState pData, struct socket *pSocket)
278	{
279	AssertPtrReturn(pData, 0);
280	AssertPtrReturn(pSocket, 0);
281	AssertReturn(pSocket->fUnderPolling, 0);
282	if (pSocket->fShouldBeRemoved)
283	{
284	pSocket->fUnderPolling = 0;
285	sofree(pData, pSocket);
286	/* pSocket is PHANTOM, now */
287	return 1;
288	}
289	return 0;
290	}
291
292	int slirp_init(PNATState *ppData, uint32_t u32NetAddr, uint32_t u32Netmask,
293	bool fPassDomain, bool fUseHostResolver, int i32AliasMode,
294	int iIcmpCacheLimit, void *pvUser)
295	{
296	int rc;
297	PNATState pData;
298	if (u32Netmask & 0x1f)
299	/* CTL is x.x.x.15, bootp passes up to 16 IPs (15..31) */
300	return VERR_INVALID_PARAMETER;
301	pData = RTMemAllocZ(RT_ALIGN_Z(sizeof(NATState), sizeof(uint64_t)));
302	*ppData = pData;
303	if (!pData)
304	return VERR_NO_MEMORY;
305	pData->fPassDomain = !fUseHostResolver ? fPassDomain : false;
306	pData->fUseHostResolver = fUseHostResolver;
307	pData->fUseHostResolverPermanent = fUseHostResolver;
308	pData->pvUser = pvUser;
309	pData->netmask = u32Netmask;
310
311	rc = RTCritSectRwInit(&pData->CsRwHandlerChain);
312	if (RT_FAILURE(rc))
313	return rc;
314
315	/* sockets & TCP defaults */
316	pData->socket_rcv = 64 * _1K;
317	pData->socket_snd = 64 * _1K;
318	tcp_sndspace = 64 * _1K;
319	tcp_rcvspace = 64 * _1K;
320
321	/*
322	* Use the same default here as in DevNAT.cpp (SoMaxConnection CFGM value)
323	* to avoid release log noise.
324	*/
325	pData->soMaxConn = 10;
326
327	#ifdef RT_OS_WINDOWS
328	{
329	WSADATA Data;
330	WSAStartup(MAKEWORD(2, 0), &Data);
331	}
332	pData->phEvents[VBOX_SOCKET_EVENT_INDEX] = CreateEvent(NULL, FALSE, FALSE, NULL);
333	#endif
334
335	link_up = 1;
336
337	rc = bootp_dhcp_init(pData);
338	if (RT_FAILURE(rc))
339	{
340	Log(("NAT: DHCP server initialization failed\n"));
341	RTMemFree(pData);
342	*ppData = NULL;
343	return rc;
344	}
345	debug_init(pData);
346	if_init(pData);
347	ip_init(pData);
348	icmp_init(pData, iIcmpCacheLimit);
349
350	/* Initialise mbufs after setting the MTU */
351	mbuf_init(pData);
352
353	pData->special_addr.s_addr = u32NetAddr;
354	pData->slirp_ethaddr = &special_ethaddr[0];
355	alias_addr.s_addr = pData->special_addr.s_addr \| RT_H2N_U32_C(CTL_ALIAS);
356	/* @todo: add ability to configure this staff */
357
358	/* set default addresses */
359	inet_aton("127.0.0.1", &loopback_addr);
360
361	if (!fUseHostResolver)
362	{
363	rc = slirpInitializeDnsSettings(pData);
364	AssertRCReturn(rc, VINF_NAT_DNS);
365	}
366
367	rc = slirpTftpInit(pData);
368	AssertRCReturn(rc, VINF_NAT_DNS);
369
370	if (i32AliasMode & ~(PKT_ALIAS_LOG\|PKT_ALIAS_SAME_PORTS\|PKT_ALIAS_PROXY_ONLY))
371	{
372	Log(("NAT: alias mode %x is ignored\n", i32AliasMode));
373	i32AliasMode = 0;
374	}
375	pData->i32AliasMode = i32AliasMode;
376	getouraddr(pData);
377	{
378	int flags = 0;
379	struct in_addr proxy_addr;
380	pData->proxy_alias = LibAliasInit(pData, NULL);
381	if (pData->proxy_alias == NULL)
382	{
383	Log(("NAT: LibAlias default rule wasn't initialized\n"));
384	AssertMsgFailed(("NAT: LibAlias default rule wasn't initialized\n"));
385	}
386	flags = LibAliasSetMode(pData->proxy_alias, 0, 0);
387	#ifndef NO_FW_PUNCH
388	flags \|= PKT_ALIAS_PUNCH_FW;
389	#endif
390	flags \|= pData->i32AliasMode; /* do transparent proxying */
391	flags = LibAliasSetMode(pData->proxy_alias, flags, ~0);
392	proxy_addr.s_addr = RT_H2N_U32(RT_N2H_U32(pData->special_addr.s_addr) \| CTL_ALIAS);
393	LibAliasSetAddress(pData->proxy_alias, proxy_addr);
394	ftp_alias_load(pData);
395	nbt_alias_load(pData);
396	if (pData->fUseHostResolver)
397	dns_alias_load(pData);
398	}
399	#ifdef VBOX_WITH_NAT_SEND2HOME
400	/* @todo: we should know all interfaces available on host. */
401	pData->pInSockAddrHomeAddress = RTMemAllocZ(sizeof(struct sockaddr));
402	pData->cInHomeAddressSize = 1;
403	inet_aton("192.168.1.25", &pData->pInSockAddrHomeAddress[0].sin_addr);
404	pData->pInSockAddrHomeAddress[0].sin_family = AF_INET;
405	# ifdef RT_OS_DARWIN
406	pData->pInSockAddrHomeAddress[0].sin_len = sizeof(struct sockaddr_in);
407	# endif
408	#endif
409	return VINF_SUCCESS;
410	}
411
412	/**
413	* Register statistics.
414	*/
415	void slirp_register_statistics(PNATState pData, PPDMDRVINS pDrvIns)
416	{
417	#ifdef VBOX_WITH_STATISTICS
418	# define PROFILE_COUNTER(name, dsc) REGISTER_COUNTER(name, pData, STAMTYPE_PROFILE, STAMUNIT_TICKS_PER_CALL, dsc)
419	# define COUNTING_COUNTER(name, dsc) REGISTER_COUNTER(name, pData, STAMTYPE_COUNTER, STAMUNIT_COUNT, dsc)
420	# include "counters.h"
421	# undef COUNTER
422	/** @todo register statistics for the variables dumped by:
423	* ipstats(pData); tcpstats(pData); udpstats(pData); icmpstats(pData);
424	* mbufstats(pData); sockstats(pData); */
425	#else /* VBOX_WITH_STATISTICS */
426	NOREF(pData);
427	NOREF(pDrvIns);
428	#endif /* !VBOX_WITH_STATISTICS */
429	}
430
431	/**
432	* Deregister statistics.
433	*/
434	void slirp_deregister_statistics(PNATState pData, PPDMDRVINS pDrvIns)
435	{
436	if (pData == NULL)
437	return;
438	#ifdef VBOX_WITH_STATISTICS
439	# define PROFILE_COUNTER(name, dsc) DEREGISTER_COUNTER(name, pData)
440	# define COUNTING_COUNTER(name, dsc) DEREGISTER_COUNTER(name, pData)
441	# include "counters.h"
442	#else /* VBOX_WITH_STATISTICS */
443	NOREF(pData);
444	NOREF(pDrvIns);
445	#endif /* !VBOX_WITH_STATISTICS */
446	}
447
448	/**
449	* Marks the link as up, making it possible to establish new connections.
450	*/
451	void slirp_link_up(PNATState pData)
452	{
453	struct arp_cache_entry *ac;
454	link_up = 1;
455
456	if (LIST_EMPTY(&pData->arp_cache))
457	return;
458
459	LIST_FOREACH(ac, &pData->arp_cache, list)
460	{
461	activate_port_forwarding(pData, ac->ether);
462	}
463	}
464
465	/**
466	* Marks the link as down and cleans up the current connections.
467	*/
468	void slirp_link_down(PNATState pData)
469	{
470	struct socket *so;
471	struct port_forward_rule *rule;
472
473	while ((so = tcb.so_next) != &tcb)
474	{
475	/* Don't miss TCB releasing */
476	if ( !sototcpcb(so)
477	&& ( so->so_state & SS_NOFDREF
478	\|\| so->s == -1))
479	sofree(pData, so);
480	else
481	tcp_close(pData, sototcpcb(so));
482	}
483
484	while ((so = udb.so_next) != &udb)
485	udp_detach(pData, so);
486
487	/*
488	* Clear the active state of port-forwarding rules to force
489	* re-setup on restoration of communications.
490	*/
491	LIST_FOREACH(rule, &pData->port_forward_rule_head, list)
492	{
493	rule->activated = 0;
494	}
495	pData->cRedirectionsActive = 0;
496
497	link_up = 0;
498	}
499
500	/**
501	* Terminates the slirp component.
502	*/
503	void slirp_term(PNATState pData)
504	{
505	if (pData == NULL)
506	return;
507	icmp_finit(pData);
508
509	slirp_link_down(pData);
510	slirpReleaseDnsSettings(pData);
511	ftp_alias_unload(pData);
512	nbt_alias_unload(pData);
513	if (pData->fUseHostResolver)
514	{
515	dns_alias_unload(pData);
516	#ifdef VBOX_WITH_DNSMAPPING_IN_HOSTRESOLVER
517	while (!LIST_EMPTY(&pData->DNSMapHead))
518	{
519	PDNSMAPPINGENTRY pDnsEntry = LIST_FIRST(&pData->DNSMapHead);
520	LIST_REMOVE(pDnsEntry, MapList);
521	RTStrFree(pDnsEntry->pszCName);
522	RTMemFree(pDnsEntry);
523	}
524	#endif
525	}
526	while (!LIST_EMPTY(&instancehead))
527	{
528	struct libalias *la = LIST_FIRST(&instancehead);
529	/* libalias do all clean up */
530	LibAliasUninit(la);
531	}
532	while (!LIST_EMPTY(&pData->arp_cache))
533	{
534	struct arp_cache_entry *ac = LIST_FIRST(&pData->arp_cache);
535	LIST_REMOVE(ac, list);
536	RTMemFree(ac);
537	}
538	slirpTftpTerm(pData);
539	bootp_dhcp_fini(pData);
540	m_fini(pData);
541	#ifdef RT_OS_WINDOWS
542	WSACleanup();
543	#endif
544	#ifndef VBOX_WITH_SLIRP_BSD_SBUF
545	#ifdef LOG_ENABLED
546	Log(("\n"
547	"NAT statistics\n"
548	"--------------\n"
549	"\n"));
550	ipstats(pData);
551	tcpstats(pData);
552	udpstats(pData);
553	icmpstats(pData);
554	mbufstats(pData);
555	sockstats(pData);
556	Log(("\n"
557	"\n"
558	"\n"));
559	#endif
560	#endif
561	RTCritSectRwDelete(&pData->CsRwHandlerChain);
562	RTMemFree(pData);
563	}
564
565
566	#define CONN_CANFSEND(so) (((so)->so_state & (SS_FCANTSENDMORE\|SS_ISFCONNECTED)) == SS_ISFCONNECTED)
567	#define CONN_CANFRCV(so) (((so)->so_state & (SS_FCANTRCVMORE\|SS_ISFCONNECTED)) == SS_ISFCONNECTED)
568
569	/*
570	* curtime kept to an accuracy of 1ms
571	*/
572	static void updtime(PNATState pData)
573	{
574	#ifdef RT_OS_WINDOWS
575	struct _timeb tb;
576
577	_ftime(&tb);
578	curtime = (u_int)tb.time * (u_int)1000;
579	curtime += (u_int)tb.millitm;
580	#else
581	gettimeofday(&tt, 0);
582
583	curtime = (u_int)tt.tv_sec * (u_int)1000;
584	curtime += (u_int)tt.tv_usec / (u_int)1000;
585
586	if ((tt.tv_usec % 1000) >= 500)
587	curtime++;
588	#endif
589	}
590
591	#ifdef RT_OS_WINDOWS
592	void slirp_select_fill(PNATState pData, int *pnfds)
593	#else /* RT_OS_WINDOWS */
594	void slirp_select_fill(PNATState pData, int pnfds, struct pollfd polls)
595	#endif /* !RT_OS_WINDOWS */
596	{
597	struct socket so, so_next;
598	int nfds;
599	#if defined(RT_OS_WINDOWS)
600	int rc;
601	int error;
602	#else
603	int poll_index = 0;
604	#endif
605	int i;
606
607	STAM_PROFILE_START(&pData->StatFill, a);
608
609	nfds = *pnfds;
610
611	/*
612	* First, TCP sockets
613	*/
614	do_slowtimo = 0;
615	if (!link_up)
616	goto done;
617
618	/*
619	* *_slowtimo needs calling if there are IP fragments
620	* in the fragment queue, or there are TCP connections active
621	*/
622	/* XXX:
623	* triggering of fragment expiration should be the same but use new macroses
624	*/
625	do_slowtimo = (tcb.so_next != &tcb);
626	if (!do_slowtimo)
627	{
628	for (i = 0; i < IPREASS_NHASH; i++)
629	{
630	if (!TAILQ_EMPTY(&ipq[i]))
631	{
632	do_slowtimo = 1;
633	break;
634	}
635	}
636	}
637	/* always add the ICMP socket */
638	#ifndef RT_OS_WINDOWS
639	pData->icmp_socket.so_poll_index = -1;
640	#endif
641	ICMP_ENGAGE_EVENT(&pData->icmp_socket, readfds);
642
643	STAM_COUNTER_RESET(&pData->StatTCP);
644	STAM_COUNTER_RESET(&pData->StatTCPHot);
645
646	QSOCKET_FOREACH(so, so_next, tcp)
647	/* { */
648	Assert(so->so_type == IPPROTO_TCP);
649	#if !defined(RT_OS_WINDOWS)
650	so->so_poll_index = -1;
651	#endif
652	STAM_COUNTER_INC(&pData->StatTCP);
653	#ifdef VBOX_WITH_NAT_UDP_SOCKET_CLONE
654	/* TCP socket can't be cloned */
655	Assert((!so->so_cloneOf));
656	#endif
657	/*
658	* See if we need a tcp_fasttimo
659	*/
660	if ( time_fasttimo == 0
661	&& so->so_tcpcb != NULL
662	&& so->so_tcpcb->t_flags & TF_DELACK)
663	{
664	time_fasttimo = curtime; /* Flag when we want a fasttimo */
665	}
666
667	/*
668	* NOFDREF can include still connecting to local-host,
669	* newly socreated() sockets etc. Don't want to select these.
670	*/
671	if (so->so_state & SS_NOFDREF \|\| so->s == -1)
672	CONTINUE(tcp);
673
674	/*
675	* Set for reading sockets which are accepting
676	*/
677	if (so->so_state & SS_FACCEPTCONN)
678	{
679	STAM_COUNTER_INC(&pData->StatTCPHot);
680	TCP_ENGAGE_EVENT1(so, readfds);
681	CONTINUE(tcp);
682	}
683
684	/*
685	* Set for writing sockets which are connecting
686	*/
687	if (so->so_state & SS_ISFCONNECTING)
688	{
689	Log2(("connecting %R[natsock] engaged\n",so));
690	STAM_COUNTER_INC(&pData->StatTCPHot);
691	#ifdef RT_OS_WINDOWS
692	WIN_TCP_ENGAGE_EVENT2(so, writefds, connectfds);
693	#else
694	TCP_ENGAGE_EVENT1(so, writefds);
695	#endif
696	}
697
698	/*
699	* Set for writing if we are connected, can send more, and
700	* we have something to send
701	*/
702	if (CONN_CANFSEND(so) && SBUF_LEN(&so->so_rcv))
703	{
704	STAM_COUNTER_INC(&pData->StatTCPHot);
705	TCP_ENGAGE_EVENT1(so, writefds);
706	}
707
708	/*
709	* Set for reading (and urgent data) if we are connected, can
710	* receive more, and we have room for it XXX /2 ?
711	*/
712	/* @todo: vvl - check which predicat here will be more useful here in rerm of new sbufs. */
713	if ( CONN_CANFRCV(so)
714	&& (SBUF_LEN(&so->so_snd) < (SBUF_SIZE(&so->so_snd)/2))
715	#ifdef RT_OS_WINDOWS
716	&& !(so->so_state & SS_ISFCONNECTING)
717	#endif
718	)
719	{
720	STAM_COUNTER_INC(&pData->StatTCPHot);
721	TCP_ENGAGE_EVENT2(so, readfds, xfds);
722	}
723	LOOP_LABEL(tcp, so, so_next);
724	}
725
726	/*
727	* UDP sockets
728	*/
729	STAM_COUNTER_RESET(&pData->StatUDP);
730	STAM_COUNTER_RESET(&pData->StatUDPHot);
731
732	QSOCKET_FOREACH(so, so_next, udp)
733	/* { */
734
735	Assert(so->so_type == IPPROTO_UDP);
736	STAM_COUNTER_INC(&pData->StatUDP);
737	#if !defined(RT_OS_WINDOWS)
738	so->so_poll_index = -1;
739	#endif
740
741	/*
742	* See if it's timed out
743	*/
744	if (so->so_expire)
745	{
746	if (so->so_expire <= curtime)
747	{
748	Log2(("NAT: %R[natsock] expired\n", so));
749	if (so->so_timeout != NULL)
750	{
751	/* so_timeout - might change the so_expire value or
752	* drop so_timeout* from so.
753	*/
754	so->so_timeout(pData, so, so->so_timeout_arg);
755	/* on 4.2 so->
756	*/
757	if ( so_next->so_prev != so /* so_timeout freed the socket */
758	\|\| so->so_timeout) /* so_timeout just freed so_timeout */
759	CONTINUE_NO_UNLOCK(udp);
760	}
761	UDP_DETACH(pData, so, so_next);
762	CONTINUE_NO_UNLOCK(udp);
763	}
764	}
765	#ifdef VBOX_WITH_NAT_UDP_SOCKET_CLONE
766	if (so->so_cloneOf)
767	CONTINUE_NO_UNLOCK(udp);
768	#endif
769
770	/*
771	* When UDP packets are received from over the link, they're
772	* sendto()'d straight away, so no need for setting for writing
773	* Limit the number of packets queued by this session to 4.
774	* Note that even though we try and limit this to 4 packets,
775	* the session could have more queued if the packets needed
776	* to be fragmented.
777	*
778	* (XXX <= 4 ?)
779	*/
780	if ((so->so_state & SS_ISFCONNECTED) && so->so_queued <= 4)
781	{
782	STAM_COUNTER_INC(&pData->StatUDPHot);
783	UDP_ENGAGE_EVENT(so, readfds);
784	}
785	LOOP_LABEL(udp, so, so_next);
786	}
787	done:
788
789	#if defined(RT_OS_WINDOWS)
790	*pnfds = VBOX_EVENT_COUNT;
791	#else /* RT_OS_WINDOWS */
792	AssertRelease(poll_index <= *pnfds);
793	*pnfds = poll_index;
794	#endif /* !RT_OS_WINDOWS */
795
796	STAM_PROFILE_STOP(&pData->StatFill, a);
797	}
798
799
800	/**
801	* This function do Connection or sending tcp sequence to.
802	* @returns if true operation completed
803	* @note: functions call tcp_input that potentially could lead to tcp_drop
804	*/
805	static bool slirpConnectOrWrite(PNATState pData, struct socket *so, bool fConnectOnly)
806	{
807	int ret;
808	LogFlowFunc(("ENTER: so:%R[natsock], fConnectOnly:%RTbool\n", so, fConnectOnly));
809	/*
810	* Check for non-blocking, still-connecting sockets
811	*/
812	if (so->so_state & SS_ISFCONNECTING)
813	{
814	Log2(("connecting %R[natsock] catched\n", so));
815	/* Connected */
816	so->so_state &= ~SS_ISFCONNECTING;
817
818	/*
819	* This should be probably guarded by PROBE_CONN too. Anyway,
820	* we disable it on OS/2 because the below send call returns
821	* EFAULT which causes the opened TCP socket to close right
822	* after it has been opened and connected.
823	*/
824	#ifndef RT_OS_OS2
825	ret = send(so->s, (const char *)&ret, 0, 0);
826	if (ret < 0)
827	{
828	/* XXXXX Must fix, zero bytes is a NOP */
829	if ( soIgnorableErrorCode(errno)
830	\|\| errno == ENOTCONN)
831	{
832	LogFlowFunc(("LEAVE: false\n"));
833	return false;
834	}
835
836	/* else failed */
837	so->so_state = SS_NOFDREF;
838	}
839	/* else so->so_state &= ~SS_ISFCONNECTING; */
840	#endif
841
842	/*
843	* Continue tcp_input
844	*/
845	TCP_INPUT(pData, (struct mbuf *)NULL, sizeof(struct ip), so);
846	/* continue; */
847	}
848	else if (!fConnectOnly)
849	SOWRITE(ret, pData, so);
850	/*
851	* XXX If we wrote something (a lot), there could be the need
852	* for a window update. In the worst case, the remote will send
853	* a window probe to get things going again.
854	*/
855	LogFlowFunc(("LEAVE: true\n"));
856	return true;
857	}
858
859	#if defined(RT_OS_WINDOWS)
860	void slirp_select_poll(PNATState pData, int fTimeout, int fIcmp)
861	#else /* RT_OS_WINDOWS */
862	void slirp_select_poll(PNATState pData, struct pollfd *polls, int ndfs)
863	#endif /* !RT_OS_WINDOWS */
864	{
865	struct socket so, so_next;
866	int ret;
867	#if defined(RT_OS_WINDOWS)
868	WSANETWORKEVENTS NetworkEvents;
869	int rc;
870	int error;
871	#endif
872
873	STAM_PROFILE_START(&pData->StatPoll, a);
874
875	/* Update time */
876	updtime(pData);
877
878	/*
879	* See if anything has timed out
880	*/
881	if (link_up)
882	{
883	if (time_fasttimo && ((curtime - time_fasttimo) >= 2))
884	{
885	STAM_PROFILE_START(&pData->StatFastTimer, b);
886	tcp_fasttimo(pData);
887	time_fasttimo = 0;
888	STAM_PROFILE_STOP(&pData->StatFastTimer, b);
889	}
890	if (do_slowtimo && ((curtime - last_slowtimo) >= 499))
891	{
892	STAM_PROFILE_START(&pData->StatSlowTimer, c);
893	ip_slowtimo(pData);
894	tcp_slowtimo(pData);
895	last_slowtimo = curtime;
896	STAM_PROFILE_STOP(&pData->StatSlowTimer, c);
897	}
898	}
899	#if defined(RT_OS_WINDOWS)
900	if (fTimeout)
901	return; /* only timer update */
902	#endif
903
904	/*
905	* Check sockets
906	*/
907	if (!link_up)
908	goto done;
909	#if defined(RT_OS_WINDOWS)
910	/*XXX: before renaming please make see define
911	* fIcmp in slirp_state.h
912	*/
913	if (fIcmp)
914	sorecvfrom(pData, &pData->icmp_socket);
915	#else
916	if ( (pData->icmp_socket.s != -1)
917	&& CHECK_FD_SET(&pData->icmp_socket, ignored, readfds))
918	sorecvfrom(pData, &pData->icmp_socket);
919	#endif
920	/*
921	* Check TCP sockets
922	*/
923	QSOCKET_FOREACH(so, so_next, tcp)
924	/* { */
925	/* TCP socket can't be cloned */
926	#ifdef VBOX_WITH_NAT_UDP_SOCKET_CLONE
927	Assert((!so->so_cloneOf));
928	#endif
929	Assert(!so->fUnderPolling);
930	so->fUnderPolling = 1;
931	if (slirpVerifyAndFreeSocket(pData, so))
932	CONTINUE(tcp);
933	/*
934	* FD_ISSET is meaningless on these sockets
935	* (and they can crash the program)
936	*/
937	if (so->so_state & SS_NOFDREF \|\| so->s == -1)
938	{
939	so->fUnderPolling = 0;
940	CONTINUE(tcp);
941	}
942
943	POLL_TCP_EVENTS(rc, error, so, &NetworkEvents);
944
945	LOG_NAT_SOCK(so, TCP, &NetworkEvents, readfds, writefds, xfds);
946
947
948	/*
949	* Check for URG data
950	* This will soread as well, so no need to
951	* test for readfds below if this succeeds
952	*/
953
954	/* out-of-band data */
955	if ( CHECK_FD_SET(so, NetworkEvents, xfds)
956	#ifdef RT_OS_DARWIN
957	/* Darwin and probably BSD hosts generates POLLPRI\|POLLHUP event on receiving TCP.flags.{ACK\|URG\|FIN} this
958	* combination on other Unixs hosts doesn't enter to this branch
959	*/
960	&& !CHECK_FD_SET(so, NetworkEvents, closefds)
961	#endif
962	#ifdef RT_OS_WINDOWS
963	/**
964	* In some cases FD_CLOSE comes with FD_OOB, that confuse tcp processing.
965	*/
966	&& !WIN_CHECK_FD_SET(so, NetworkEvents, closefds)
967	#endif
968	)
969	{
970	sorecvoob(pData, so);
971	if (slirpVerifyAndFreeSocket(pData, so))
972	CONTINUE(tcp);
973	}
974
975	/*
976	* Check sockets for reading
977	*/
978	else if ( CHECK_FD_SET(so, NetworkEvents, readfds)
979	\|\| WIN_CHECK_FD_SET(so, NetworkEvents, acceptds))
980	{
981
982	#ifdef RT_OS_WINDOWS
983	if (WIN_CHECK_FD_SET(so, NetworkEvents, connectfds))
984	{
985	/* Finish connection first */
986	/* should we ignore return value? */
987	bool fRet = slirpConnectOrWrite(pData, so, true);
988	LogFunc(("fRet:%RTbool\n", fRet));
989	if (slirpVerifyAndFreeSocket(pData, so))
990	CONTINUE(tcp);
991	}
992	#endif
993	/*
994	* Check for incoming connections
995	*/
996	if (so->so_state & SS_FACCEPTCONN)
997	{
998	TCP_CONNECT(pData, so);
999	if (slirpVerifyAndFreeSocket(pData, so))
1000	CONTINUE(tcp);
1001	if (!CHECK_FD_SET(so, NetworkEvents, closefds))
1002	{
1003	so->fUnderPolling = 0;
1004	CONTINUE(tcp);
1005	}
1006	}
1007
1008	ret = soread(pData, so);
1009	if (slirpVerifyAndFreeSocket(pData, so))
1010	CONTINUE(tcp);
1011	/* Output it if we read something */
1012	if (RT_LIKELY(ret > 0))
1013	TCP_OUTPUT(pData, sototcpcb(so));
1014
1015	if (slirpVerifyAndFreeSocket(pData, so))
1016	CONTINUE(tcp);
1017	}
1018
1019	/*
1020	* Check for FD_CLOSE events.
1021	* in some cases once FD_CLOSE engaged on socket it could be flashed latter (for some reasons)
1022	*/
1023	if ( CHECK_FD_SET(so, NetworkEvents, closefds)
1024	\|\| (so->so_close == 1))
1025	{
1026	/*
1027	* drain the socket
1028	*/
1029	for (; so_next->so_prev == so
1030	&& !slirpVerifyAndFreeSocket(pData, so);)
1031	{
1032	ret = soread(pData, so);
1033	if (slirpVerifyAndFreeSocket(pData, so))
1034	break;
1035
1036	if (ret > 0)
1037	TCP_OUTPUT(pData, sototcpcb(so));
1038	else if (so_next->so_prev == so)
1039	{
1040	Log2(("%R[natsock] errno %d (%s)\n", so, errno, strerror(errno)));
1041	break;
1042	}
1043	}
1044
1045	/* if socket freed ''so'' is PHANTOM and next socket isn't points on it */
1046	if (so_next->so_prev == so)
1047	{
1048	/* mark the socket for termination _after_ it was drained */
1049	so->so_close = 1;
1050	/* No idea about Windows but on Posix, POLLHUP means that we can't send more.
1051	* Actually in the specific error scenario, POLLERR is set as well. */
1052	#ifndef RT_OS_WINDOWS
1053	if (CHECK_FD_SET(so, NetworkEvents, rderr))
1054	sofcantsendmore(so);
1055	#endif
1056	}
1057	if (so_next->so_prev == so)
1058	so->fUnderPolling = 0;
1059	CONTINUE(tcp);
1060	}
1061
1062	/*
1063	* Check sockets for writing
1064	*/
1065	if ( CHECK_FD_SET(so, NetworkEvents, writefds)
1066	#ifdef RT_OS_WINDOWS
1067	\|\| WIN_CHECK_FD_SET(so, NetworkEvents, connectfds)
1068	#endif
1069	)
1070	{
1071	int fConnectOrWriteSuccess = slirpConnectOrWrite(pData, so, false);
1072	/* slirpConnectOrWrite could return true even if tcp_input called tcp_drop,
1073	* so we should be ready to such situations.
1074	*/
1075	if (slirpVerifyAndFreeSocket(pData, so))
1076	CONTINUE(tcp);
1077	else if (!fConnectOrWriteSuccess)
1078	{
1079	so->fUnderPolling = 0;
1080	CONTINUE(tcp);
1081	}
1082	/* slirpConnectionOrWrite succeeded and socket wasn't dropped */
1083	}
1084
1085	/*
1086	* Probe a still-connecting, non-blocking socket
1087	* to check if it's still alive
1088	*/
1089	#ifdef PROBE_CONN
1090	if (so->so_state & SS_ISFCONNECTING)
1091	{
1092	ret = recv(so->s, (char *)&ret, 0, 0);
1093
1094	if (ret < 0)
1095	{
1096	/* XXX */
1097	if ( soIgnorableErrorCode(errno)
1098	\|\| errno == ENOTCONN)
1099	{
1100	CONTINUE(tcp); /* Still connecting, continue */
1101	}
1102
1103	/* else failed */
1104	so->so_state = SS_NOFDREF;
1105
1106	/* tcp_input will take care of it */
1107	}
1108	else
1109	{
1110	ret = send(so->s, &ret, 0, 0);
1111	if (ret < 0)
1112	{
1113	/* XXX */
1114	if ( soIgnorableErrorCode(errno)
1115	\|\| errno == ENOTCONN)
1116	{
1117	CONTINUE(tcp);
1118	}
1119	/* else failed */
1120	so->so_state = SS_NOFDREF;
1121	}
1122	else
1123	so->so_state &= ~SS_ISFCONNECTING;
1124
1125	}
1126	TCP_INPUT((struct mbuf *)NULL, sizeof(struct ip),so);
1127	} /* SS_ISFCONNECTING */
1128	#endif
1129	if (!slirpVerifyAndFreeSocket(pData, so))
1130	so->fUnderPolling = 0;
1131	LOOP_LABEL(tcp, so, so_next);
1132	}
1133
1134	/*
1135	* Now UDP sockets.
1136	* Incoming packets are sent straight away, they're not buffered.
1137	* Incoming UDP data isn't buffered either.
1138	*/
1139	QSOCKET_FOREACH(so, so_next, udp)
1140	/* { */
1141	#ifdef VBOX_WITH_NAT_UDP_SOCKET_CLONE
1142	if (so->so_cloneOf)
1143	CONTINUE_NO_UNLOCK(udp);
1144	#endif
1145	#if 0
1146	so->fUnderPolling = 1;
1147	if(slirpVerifyAndFreeSocket(pData, so));
1148	CONTINUE(udp);
1149	so->fUnderPolling = 0;
1150	#endif
1151
1152	POLL_UDP_EVENTS(rc, error, so, &NetworkEvents);
1153
1154	LOG_NAT_SOCK(so, UDP, &NetworkEvents, readfds, writefds, xfds);
1155
1156	if (so->s != -1 && CHECK_FD_SET(so, NetworkEvents, readfds))
1157	{
1158	SORECVFROM(pData, so);
1159	}
1160	LOOP_LABEL(udp, so, so_next);
1161	}
1162
1163	done:
1164
1165	STAM_PROFILE_STOP(&pData->StatPoll, a);
1166	}
1167
1168
1169	struct arphdr
1170	{
1171	unsigned short ar_hrd; /* format of hardware address */
1172	unsigned short ar_pro; /* format of protocol address */
1173	unsigned char ar_hln; /* length of hardware address */
1174	unsigned char ar_pln; /* length of protocol address */
1175	unsigned short ar_op; /* ARP opcode (command) */
1176
1177	/*
1178	* Ethernet looks like this : This bit is variable sized however...
1179	*/
1180	unsigned char ar_sha[ETH_ALEN]; /* sender hardware address */
1181	unsigned char ar_sip[4]; /* sender IP address */
1182	unsigned char ar_tha[ETH_ALEN]; /* target hardware address */
1183	unsigned char ar_tip[4]; /* target IP address */
1184	};
1185	AssertCompileSize(struct arphdr, 28);
1186
1187	static void arp_output(PNATState pData, const uint8_t pcu8EtherSource, const struct arphdr pcARPHeaderSource, uint32_t ip4TargetAddress)
1188	{
1189	struct ethhdr *pEtherHeaderResponse;
1190	struct arphdr *pARPHeaderResponse;
1191	uint32_t ip4TargetAddressInHostFormat;
1192	struct mbuf *pMbufResponse;
1193
1194	Assert((pcu8EtherSource));
1195	if (!pcu8EtherSource)
1196	return;
1197	ip4TargetAddressInHostFormat = RT_N2H_U32(ip4TargetAddress);
1198
1199	pMbufResponse = m_getcl(pData, M_NOWAIT, MT_HEADER, M_PKTHDR);
1200	if (!pMbufResponse)
1201	return;
1202	pEtherHeaderResponse = mtod(pMbufResponse, struct ethhdr *);
1203	/* @note: if_encap will swap src and dst*/
1204	memcpy(pEtherHeaderResponse->h_source, pcu8EtherSource, ETH_ALEN);
1205	pMbufResponse->m_data += ETH_HLEN;
1206	pARPHeaderResponse = mtod(pMbufResponse, struct arphdr *);
1207	pMbufResponse->m_len = sizeof(struct arphdr);
1208
1209	pARPHeaderResponse->ar_hrd = RT_H2N_U16_C(1);
1210	pARPHeaderResponse->ar_pro = RT_H2N_U16_C(ETH_P_IP);
1211	pARPHeaderResponse->ar_hln = ETH_ALEN;
1212	pARPHeaderResponse->ar_pln = 4;
1213	pARPHeaderResponse->ar_op = RT_H2N_U16_C(ARPOP_REPLY);
1214	memcpy(pARPHeaderResponse->ar_sha, special_ethaddr, ETH_ALEN);
1215
1216	if (!slirpMbufTagService(pData, pMbufResponse, (uint8_t)(ip4TargetAddressInHostFormat & ~pData->netmask)))
1217	{
1218	static bool fTagErrorReported;
1219	if (!fTagErrorReported)
1220	{
1221	LogRel(("NAT: couldn't add the tag(PACKET_SERVICE:%d)\n",
1222	(uint8_t)(ip4TargetAddressInHostFormat & ~pData->netmask)));
1223	fTagErrorReported = true;
1224	}
1225	}
1226	pARPHeaderResponse->ar_sha[5] = (uint8_t)(ip4TargetAddressInHostFormat & ~pData->netmask);
1227
1228	memcpy(pARPHeaderResponse->ar_sip, pcARPHeaderSource->ar_tip, 4);
1229	memcpy(pARPHeaderResponse->ar_tha, pcARPHeaderSource->ar_sha, ETH_ALEN);
1230	memcpy(pARPHeaderResponse->ar_tip, pcARPHeaderSource->ar_sip, 4);
1231	if_encap(pData, ETH_P_ARP, pMbufResponse, ETH_ENCAP_URG);
1232	}
1233	/**
1234	* @note This function will free m!
1235	*/
1236	static void arp_input(PNATState pData, struct mbuf *m)
1237	{
1238	struct ethhdr *pEtherHeader;
1239	struct arphdr *pARPHeader;
1240	uint32_t ip4TargetAddress;
1241
1242	int ar_op;
1243	pEtherHeader = mtod(m, struct ethhdr *);
1244	pARPHeader = (struct arphdr *)&pEtherHeader[1];
1245
1246	ar_op = RT_N2H_U16(pARPHeader->ar_op);
1247	ip4TargetAddress = (uint32_t)pARPHeader->ar_tip;
1248
1249	switch (ar_op)
1250	{
1251	case ARPOP_REQUEST:
1252	if ( CTL_CHECK(ip4TargetAddress, CTL_DNS)
1253	\|\| CTL_CHECK(ip4TargetAddress, CTL_ALIAS)
1254	\|\| CTL_CHECK(ip4TargetAddress, CTL_TFTP))
1255	arp_output(pData, pEtherHeader->h_source, pARPHeader, ip4TargetAddress);
1256
1257	/* Gratuitous ARP */
1258	if ( (uint32_t )pARPHeader->ar_sip == (uint32_t )pARPHeader->ar_tip
1259	&& memcmp(pARPHeader->ar_tha, broadcast_ethaddr, ETH_ALEN) == 0
1260	&& memcmp(pEtherHeader->h_dest, broadcast_ethaddr, ETH_ALEN) == 0)
1261	{
1262	/* We've received an announce about address assignment,
1263	* let's do an ARP cache update
1264	*/
1265	static bool fGratuitousArpReported;
1266	if (!fGratuitousArpReported)
1267	{
1268	LogRel(("NAT: Gratuitous ARP [IP:%RTnaipv4, ether:%RTmac]\n",
1269	pARPHeader->ar_sip, pARPHeader->ar_sha));
1270	fGratuitousArpReported = true;
1271	}
1272	slirp_arp_cache_update_or_add(pData, (uint32_t )pARPHeader->ar_sip, &pARPHeader->ar_sha[0]);
1273	}
1274	break;
1275
1276	case ARPOP_REPLY:
1277	slirp_arp_cache_update_or_add(pData, (uint32_t )pARPHeader->ar_sip, &pARPHeader->ar_sha[0]);
1278	break;
1279
1280	default:
1281	break;
1282	}
1283
1284	m_freem(pData, m);
1285	}
1286
1287	/**
1288	* Feed a packet into the slirp engine.
1289	*
1290	* @param m Data buffer, m_len is not valid.
1291	* @param cbBuf The length of the data in m.
1292	*/
1293	void slirp_input(PNATState pData, struct mbuf *m, size_t cbBuf)
1294	{
1295	int proto;
1296	static bool fWarnedIpv6;
1297	struct ethhdr *eh;
1298	uint8_t au8Ether[ETH_ALEN];
1299
1300	m->m_len = cbBuf;
1301	if (cbBuf < ETH_HLEN)
1302	{
1303	Log(("NAT: packet having size %d has been ignored\n", m->m_len));
1304	m_freem(pData, m);
1305	return;
1306	}
1307	eh = mtod(m, struct ethhdr *);
1308	proto = RT_N2H_U16(eh->h_proto);
1309
1310	memcpy(au8Ether, eh->h_source, ETH_ALEN);
1311
1312	switch(proto)
1313	{
1314	case ETH_P_ARP:
1315	arp_input(pData, m);
1316	break;
1317
1318	case ETH_P_IP:
1319	/* Update time. Important if the network is very quiet, as otherwise
1320	* the first outgoing connection gets an incorrect timestamp. */
1321	updtime(pData);
1322	m_adj(m, ETH_HLEN);
1323	M_ASSERTPKTHDR(m);
1324	m->m_pkthdr.header = mtod(m, void *);
1325	ip_input(pData, m);
1326	break;
1327
1328	case ETH_P_IPV6:
1329	m_freem(pData, m);
1330	if (!fWarnedIpv6)
1331	{
1332	LogRel(("NAT: IPv6 not supported\n"));
1333	fWarnedIpv6 = true;
1334	}
1335	break;
1336
1337	default:
1338	Log(("NAT: Unsupported protocol %x\n", proto));
1339	m_freem(pData, m);
1340	break;
1341	}
1342
1343	if (pData->cRedirectionsActive != pData->cRedirectionsStored)
1344	activate_port_forwarding(pData, au8Ether);
1345	}
1346
1347	/**
1348	* Output the IP packet to the ethernet device.
1349	*
1350	* @note This function will free m!
1351	*/
1352	void if_encap(PNATState pData, uint16_t eth_proto, struct mbuf *m, int flags)
1353	{
1354	struct ethhdr *eh;
1355	uint8_t *mbuf = NULL;
1356	size_t mlen = 0;
1357	STAM_PROFILE_START(&pData->StatIF_encap, a);
1358	LogFlowFunc(("ENTER: pData:%p, eth_proto:%RX16, m:%p, flags:%d\n",
1359	pData, eth_proto, m, flags));
1360
1361	M_ASSERTPKTHDR(m);
1362	m->m_data -= ETH_HLEN;
1363	m->m_len += ETH_HLEN;
1364	eh = mtod(m, struct ethhdr *);
1365	mlen = m->m_len;
1366
1367	if (memcmp(eh->h_source, special_ethaddr, ETH_ALEN) != 0)
1368	{
1369	struct m_tag *t = m_tag_first(m);
1370	uint8_t u8ServiceId = CTL_ALIAS;
1371	memcpy(eh->h_dest, eh->h_source, ETH_ALEN);
1372	memcpy(eh->h_source, special_ethaddr, ETH_ALEN);
1373	Assert(memcmp(eh->h_dest, special_ethaddr, ETH_ALEN) != 0);
1374	if (memcmp(eh->h_dest, zerro_ethaddr, ETH_ALEN) == 0)
1375	{
1376	/* don't do anything */
1377	m_freem(pData, m);
1378	goto done;
1379	}
1380	if ( t
1381	&& (t = m_tag_find(m, PACKET_SERVICE, NULL)))
1382	{
1383	Assert(t);
1384	u8ServiceId = (uint8_t )&t[1];
1385	}
1386	eh->h_source[5] = u8ServiceId;
1387	}
1388	/*
1389	* we're processing the chain, that isn't not expected.
1390	*/
1391	Assert((!m->m_next));
1392	if (m->m_next)
1393	{
1394	Log(("NAT: if_encap's recived the chain, dropping...\n"));
1395	m_freem(pData, m);
1396	goto done;
1397	}
1398	mbuf = mtod(m, uint8_t *);
1399	eh->h_proto = RT_H2N_U16(eth_proto);
1400	LogFunc(("eh(dst:%RTmac, src:%RTmac)\n", eh->h_dest, eh->h_source));
1401	if (flags & ETH_ENCAP_URG)
1402	slirp_urg_output(pData->pvUser, m, mbuf, mlen);
1403	else
1404	slirp_output(pData->pvUser, m, mbuf, mlen);
1405	done:
1406	STAM_PROFILE_STOP(&pData->StatIF_encap, a);
1407	LogFlowFuncLeave();
1408	}
1409
1410	/**
1411	* Still we're using dhcp server leasing to map ether to IP
1412	* @todo see rt_lookup_in_cache
1413	*/
1414	static uint32_t find_guest_ip(PNATState pData, const uint8_t *eth_addr)
1415	{
1416	uint32_t ip = INADDR_ANY;
1417	int rc;
1418
1419	if (eth_addr == NULL)
1420	return INADDR_ANY;
1421
1422	if ( memcmp(eth_addr, zerro_ethaddr, ETH_ALEN) == 0
1423	\|\| memcmp(eth_addr, broadcast_ethaddr, ETH_ALEN) == 0)
1424	return INADDR_ANY;
1425
1426	rc = slirp_arp_lookup_ip_by_ether(pData, eth_addr, &ip);
1427	if (RT_SUCCESS(rc))
1428	return ip;
1429
1430	bootp_cache_lookup_ip_by_ether(pData, eth_addr, &ip);
1431	/* ignore return code, ip will be set to INADDR_ANY on error */
1432	return ip;
1433	}
1434
1435	/**
1436	* We need check if we've activated port forwarding
1437	* for specific machine ... that of course relates to
1438	* service mode
1439	* @todo finish this for service case
1440	*/
1441	static void activate_port_forwarding(PNATState pData, const uint8_t *h_source)
1442	{
1443	struct port_forward_rule rule, tmp;
1444	const uint8_t *pu8EthSource = h_source;
1445
1446	/* check mac here */
1447	LIST_FOREACH_SAFE(rule, &pData->port_forward_rule_head, list, tmp)
1448	{
1449	struct socket *so;
1450	struct alias_link *alias_link;
1451	struct libalias *lib;
1452	int flags;
1453	struct sockaddr sa;
1454	struct sockaddr_in *psin;
1455	socklen_t socketlen;
1456	struct in_addr alias;
1457	int rc;
1458	uint32_t guest_addr; /* need to understand if we already give address to guest */
1459
1460	if (rule->activated)
1461	continue;
1462
1463	#ifdef VBOX_WITH_NAT_SERVICE
1464	/**
1465	* case when guest ip is INADDR_ANY shouldn't appear in NAT service
1466	*/
1467	Assert((rule->guest_addr.s_addr != INADDR_ANY));
1468	guest_addr = rule->guest_addr.s_addr;
1469	#else /* VBOX_WITH_NAT_SERVICE */
1470	guest_addr = find_guest_ip(pData, pu8EthSource);
1471	#endif /* !VBOX_WITH_NAT_SERVICE */
1472	if (guest_addr == INADDR_ANY)
1473	{
1474	/* the address wasn't granted */
1475	return;
1476	}
1477
1478	#if !defined(VBOX_WITH_NAT_SERVICE)
1479	if ( rule->guest_addr.s_addr != guest_addr
1480	&& rule->guest_addr.s_addr != INADDR_ANY)
1481	continue;
1482	if (rule->guest_addr.s_addr == INADDR_ANY)
1483	rule->guest_addr.s_addr = guest_addr;
1484	#endif
1485
1486	LogRel(("NAT: set redirect %s host port %d => guest port %d @ %RTnaipv4\n",
1487	rule->proto == IPPROTO_UDP ? "UDP" : "TCP", rule->host_port, rule->guest_port, guest_addr));
1488
1489	if (rule->proto == IPPROTO_UDP)
1490	so = udp_listen(pData, rule->bind_ip.s_addr, RT_H2N_U16(rule->host_port), guest_addr,
1491	RT_H2N_U16(rule->guest_port), 0);
1492	else
1493	so = solisten(pData, rule->bind_ip.s_addr, RT_H2N_U16(rule->host_port), guest_addr,
1494	RT_H2N_U16(rule->guest_port), 0);
1495
1496	if (so == NULL)
1497	goto remove_port_forwarding;
1498
1499	psin = (struct sockaddr_in *)&sa;
1500	psin->sin_family = AF_INET;
1501	psin->sin_port = 0;
1502	psin->sin_addr.s_addr = INADDR_ANY;
1503	socketlen = sizeof(struct sockaddr);
1504
1505	rc = getsockname(so->s, &sa, &socketlen);
1506	if (rc < 0 \|\| sa.sa_family != AF_INET)
1507	goto remove_port_forwarding;
1508
1509	psin = (struct sockaddr_in *)&sa;
1510
1511	lib = LibAliasInit(pData, NULL);
1512	flags = LibAliasSetMode(lib, 0, 0);
1513	flags \|= pData->i32AliasMode;
1514	flags \|= PKT_ALIAS_REVERSE; /* set reverse */
1515	flags = LibAliasSetMode(lib, flags, ~0);
1516
1517	alias.s_addr = RT_H2N_U32(RT_N2H_U32(guest_addr) \| CTL_ALIAS);
1518	alias_link = LibAliasRedirectPort(lib, psin->sin_addr, RT_H2N_U16(rule->host_port),
1519	alias, RT_H2N_U16(rule->guest_port),
1520	pData->special_addr, -1, /* not very clear for now */
1521	rule->proto);
1522	if (!alias_link)
1523	goto remove_port_forwarding;
1524
1525	so->so_la = lib;
1526	rule->activated = 1;
1527	rule->so = so;
1528	pData->cRedirectionsActive++;
1529	continue;
1530
1531	remove_port_forwarding:
1532	LogRel(("NAT: failed to redirect %s %d => %d\n",
1533	(rule->proto == IPPROTO_UDP?"UDP":"TCP"), rule->host_port, rule->guest_port));
1534	LIST_REMOVE(rule, list);
1535	pData->cRedirectionsStored--;
1536	RTMemFree(rule);
1537	}
1538	}
1539
1540	/**
1541	* Changes in 3.1 instead of opening new socket do the following:
1542	* gain more information:
1543	* 1. bind IP
1544	* 2. host port
1545	* 3. guest port
1546	* 4. proto
1547	* 5. guest MAC address
1548	* the guest's MAC address is rather important for service, but we easily
1549	* could get it from VM configuration in DrvNAT or Service, the idea is activating
1550	* corresponding port-forwarding
1551	*/
1552	int slirp_add_redirect(PNATState pData, int is_udp, struct in_addr host_addr, int host_port,
1553	struct in_addr guest_addr, int guest_port, const uint8_t *ethaddr)
1554	{
1555	struct port_forward_rule *rule = NULL;
1556	LIST_FOREACH(rule, &pData->port_forward_rule_head, list)
1557	{
1558	if ( rule->proto == (is_udp ? IPPROTO_UDP : IPPROTO_TCP)
1559	&& rule->host_port == host_port
1560	&& rule->bind_ip.s_addr == host_addr.s_addr
1561	&& rule->guest_port == guest_port
1562	&& rule->guest_addr.s_addr == guest_addr.s_addr
1563	)
1564	return 0; /* rule has been already registered */
1565	}
1566
1567	rule = RTMemAllocZ(sizeof(struct port_forward_rule));
1568	if (rule == NULL)
1569	return 1;
1570
1571	rule->proto = (is_udp ? IPPROTO_UDP : IPPROTO_TCP);
1572	rule->host_port = host_port;
1573	rule->guest_port = guest_port;
1574	rule->guest_addr.s_addr = guest_addr.s_addr;
1575	rule->bind_ip.s_addr = host_addr.s_addr;
1576	if (ethaddr != NULL)
1577	memcpy(rule->mac_address, ethaddr, ETH_ALEN);
1578	/* @todo add mac address */
1579	LIST_INSERT_HEAD(&pData->port_forward_rule_head, rule, list);
1580	pData->cRedirectionsStored++;
1581	/* activate port-forwarding if guest has already got assigned IP */
1582	if ( ethaddr
1583	&& memcmp(ethaddr, zerro_ethaddr, ETH_ALEN))
1584	activate_port_forwarding(pData, ethaddr);
1585	return 0;
1586	}
1587
1588	int slirp_remove_redirect(PNATState pData, int is_udp, struct in_addr host_addr, int host_port,
1589	struct in_addr guest_addr, int guest_port)
1590	{
1591	struct port_forward_rule *rule = NULL;
1592	LIST_FOREACH(rule, &pData->port_forward_rule_head, list)
1593	{
1594	if ( rule->proto == (is_udp ? IPPROTO_UDP : IPPROTO_TCP)
1595	&& rule->host_port == host_port
1596	&& rule->guest_port == guest_port
1597	&& rule->bind_ip.s_addr == host_addr.s_addr
1598	&& rule->guest_addr.s_addr == guest_addr.s_addr
1599	&& rule->activated)
1600	{
1601	LogRel(("NAT: remove redirect %s host port %d => guest port %d @ %RTnaipv4\n",
1602	rule->proto == IPPROTO_UDP ? "UDP" : "TCP", rule->host_port, rule->guest_port, guest_addr));
1603
1604	LibAliasUninit(rule->so->so_la);
1605	if (is_udp)
1606	udp_detach(pData, rule->so);
1607	else
1608	tcp_close(pData, sototcpcb(rule->so));
1609	LIST_REMOVE(rule, list);
1610	RTMemFree(rule);
1611	pData->cRedirectionsStored--;
1612	break;
1613	}
1614
1615	}
1616	return 0;
1617	}
1618
1619	void slirp_set_ethaddr_and_activate_port_forwarding(PNATState pData, const uint8_t *ethaddr, uint32_t GuestIP)
1620	{
1621	#ifndef VBOX_WITH_NAT_SERVICE
1622	memcpy(client_ethaddr, ethaddr, ETH_ALEN);
1623	#endif
1624	if (GuestIP != INADDR_ANY)
1625	{
1626	slirp_arp_cache_update_or_add(pData, GuestIP, ethaddr);
1627	activate_port_forwarding(pData, ethaddr);
1628	}
1629	}
1630
1631	#if defined(RT_OS_WINDOWS)
1632	HANDLE *slirp_get_events(PNATState pData)
1633	{
1634	return pData->phEvents;
1635	}
1636	void slirp_register_external_event(PNATState pData, HANDLE hEvent, int index)
1637	{
1638	pData->phEvents[index] = hEvent;
1639	}
1640	#endif
1641
1642	unsigned int slirp_get_timeout_ms(PNATState pData)
1643	{
1644	if (link_up)
1645	{
1646	if (time_fasttimo)
1647	return 2;
1648	if (do_slowtimo)
1649	return 500; /* see PR_SLOWHZ */
1650	}
1651	return 36001000; / one hour */
1652	}
1653
1654	#ifndef RT_OS_WINDOWS
1655	int slirp_get_nsock(PNATState pData)
1656	{
1657	return pData->nsock;
1658	}
1659	#endif
1660
1661	/*
1662	* this function called from NAT thread
1663	*/
1664	void slirp_post_sent(PNATState pData, void *pvArg)
1665	{
1666	struct mbuf m = (struct mbuf )pvArg;
1667	m_freem(pData, m);
1668	}
1669
1670	void slirp_set_dhcp_TFTP_prefix(PNATState pData, const char *tftpPrefix)
1671	{
1672	Log2(("tftp_prefix: %s\n", tftpPrefix));
1673	tftp_prefix = tftpPrefix;
1674	}
1675
1676	void slirp_set_dhcp_TFTP_bootfile(PNATState pData, const char *bootFile)
1677	{
1678	Log2(("bootFile: %s\n", bootFile));
1679	bootp_filename = bootFile;
1680	}
1681
1682	void slirp_set_dhcp_next_server(PNATState pData, const char *next_server)
1683	{
1684	Log2(("next_server: %s\n", next_server));
1685	if (next_server == NULL)
1686	pData->tftp_server.s_addr = RT_H2N_U32(RT_N2H_U32(pData->special_addr.s_addr) \| CTL_TFTP);
1687	else
1688	inet_aton(next_server, &pData->tftp_server);
1689	}
1690
1691	int slirp_set_binding_address(PNATState pData, char *addr)
1692	{
1693	if (addr == NULL \|\| (inet_aton(addr, &pData->bindIP) == 0))
1694	{
1695	pData->bindIP.s_addr = INADDR_ANY;
1696	return 1;
1697	}
1698	return 0;
1699	}
1700
1701	void slirp_set_dhcp_dns_proxy(PNATState pData, bool fDNSProxy)
1702	{
1703	if (!pData->fUseHostResolver)
1704	{
1705	Log2(("NAT: DNS proxy switched %s\n", (fDNSProxy ? "on" : "off")));
1706	pData->fUseDnsProxy = fDNSProxy;
1707	}
1708	else if (fDNSProxy)
1709	LogRel(("NAT: Host Resolver conflicts with DNS proxy, the last one was forcely ignored\n"));
1710	}
1711
1712	#define CHECK_ARG(name, val, lim_min, lim_max) \
1713	do { \
1714	if ((val) < (lim_min) \|\| (val) > (lim_max)) \
1715	{ \
1716	LogRel(("NAT: (" #name ":%d) has been ignored, " \
1717	"because out of range (%d, %d)\n", (val), (lim_min), (lim_max))); \
1718	return; \
1719	} \
1720	else \
1721	LogRel(("NAT: (" #name ":%d)\n", (val))); \
1722	} while (0)
1723
1724	void slirp_set_somaxconn(PNATState pData, int iSoMaxConn)
1725	{
1726	LogFlowFunc(("iSoMaxConn:d\n", iSoMaxConn));
1727	/* Conditions */
1728	if (iSoMaxConn > SOMAXCONN)
1729	{
1730	LogRel(("NAT: value of somaxconn(%d) bigger than SOMAXCONN(%d)\n", iSoMaxConn, SOMAXCONN));
1731	iSoMaxConn = SOMAXCONN;
1732	}
1733
1734	if (iSoMaxConn < 1)
1735	{
1736	LogRel(("NAT: proposed value(%d) of somaxconn is invalid, default value is used (%d)\n", iSoMaxConn, pData->soMaxConn));
1737	LogFlowFuncLeave();
1738	return;
1739	}
1740
1741	/* Asignment */
1742	if (pData->soMaxConn != iSoMaxConn)
1743	{
1744	LogRel(("NAT: value of somaxconn has been changed from %d to %d\n",
1745	pData->soMaxConn, iSoMaxConn));
1746	pData->soMaxConn = iSoMaxConn;
1747	}
1748	LogFlowFuncLeave();
1749	}
1750	/* don't allow user set less 8kB and more than 1M values */
1751	#define _8K_1M_CHECK_ARG(name, val) CHECK_ARG(name, (val), 8, 1024)
1752	void slirp_set_rcvbuf(PNATState pData, int kilobytes)
1753	{
1754	_8K_1M_CHECK_ARG("SOCKET_RCVBUF", kilobytes);
1755	pData->socket_rcv = kilobytes;
1756	}
1757	void slirp_set_sndbuf(PNATState pData, int kilobytes)
1758	{
1759	_8K_1M_CHECK_ARG("SOCKET_SNDBUF", kilobytes);
1760	pData->socket_snd = kilobytes * _1K;
1761	}
1762	void slirp_set_tcp_rcvspace(PNATState pData, int kilobytes)
1763	{
1764	_8K_1M_CHECK_ARG("TCP_RCVSPACE", kilobytes);
1765	tcp_rcvspace = kilobytes * _1K;
1766	}
1767	void slirp_set_tcp_sndspace(PNATState pData, int kilobytes)
1768	{
1769	_8K_1M_CHECK_ARG("TCP_SNDSPACE", kilobytes);
1770	tcp_sndspace = kilobytes * _1K;
1771	}
1772
1773	/*
1774	* Looking for Ether by ip in ARP-cache
1775	* Note: it´s responsible of caller to allocate buffer for result
1776	* @returns iprt status code
1777	*/
1778	int slirp_arp_lookup_ether_by_ip(PNATState pData, uint32_t ip, uint8_t *ether)
1779	{
1780	struct arp_cache_entry *ac;
1781
1782	if (ether == NULL)
1783	return VERR_INVALID_PARAMETER;
1784
1785	if (LIST_EMPTY(&pData->arp_cache))
1786	return VERR_NOT_FOUND;
1787
1788	LIST_FOREACH(ac, &pData->arp_cache, list)
1789	{
1790	if ( ac->ip == ip
1791	&& memcmp(ac->ether, broadcast_ethaddr, ETH_ALEN) != 0)
1792	{
1793	memcpy(ether, ac->ether, ETH_ALEN);
1794	return VINF_SUCCESS;
1795	}
1796	}
1797	return VERR_NOT_FOUND;
1798	}
1799
1800	/*
1801	* Looking for IP by Ether in ARP-cache
1802	* Note: it´s responsible of caller to allocate buffer for result
1803	* @returns 0 - if found, 1 - otherwise
1804	*/
1805	int slirp_arp_lookup_ip_by_ether(PNATState pData, const uint8_t ether, uint32_t ip)
1806	{
1807	struct arp_cache_entry *ac;
1808	*ip = INADDR_ANY;
1809
1810	if (LIST_EMPTY(&pData->arp_cache))
1811	return VERR_NOT_FOUND;
1812
1813	LIST_FOREACH(ac, &pData->arp_cache, list)
1814	{
1815	if (memcmp(ether, ac->ether, ETH_ALEN) == 0)
1816	{
1817	*ip = ac->ip;
1818	return VINF_SUCCESS;
1819	}
1820	}
1821	return VERR_NOT_FOUND;
1822	}
1823
1824	void slirp_arp_who_has(PNATState pData, uint32_t dst)
1825	{
1826	struct mbuf *m;
1827	struct ethhdr *ehdr;
1828	struct arphdr *ahdr;
1829	static bool fWarned = false;
1830	LogFlowFunc(("ENTER: %RTnaipv4\n", dst));
1831
1832	/* ARP request WHO HAS 0.0.0.0 is one of the signals
1833	* that something has been broken at Slirp. Investigating
1834	* pcap dumps it's easy to miss warning ARP requests being
1835	* focused on investigation of other protocols flow.
1836	*/
1837	#ifdef DEBUG_vvl
1838	Assert((dst != INADDR_ANY));
1839	NOREF(fWarned);
1840	#else
1841	if ( dst == INADDR_ANY
1842	&& !fWarned)
1843	{
1844	LogRel(("NAT:ARP: \"WHO HAS INADDR_ANY\" request has been detected\n"));
1845	fWarned = true;
1846	}
1847	#endif /* !DEBUG_vvl */
1848
1849	m = m_getcl(pData, M_NOWAIT, MT_HEADER, M_PKTHDR);
1850	if (m == NULL)
1851	{
1852	Log(("NAT: Can't alloc mbuf for ARP request\n"));
1853	LogFlowFuncLeave();
1854	return;
1855	}
1856	ehdr = mtod(m, struct ethhdr *);
1857	memset(ehdr->h_source, 0xff, ETH_ALEN);
1858	ahdr = (struct arphdr *)&ehdr[1];
1859	ahdr->ar_hrd = RT_H2N_U16_C(1);
1860	ahdr->ar_pro = RT_H2N_U16_C(ETH_P_IP);
1861	ahdr->ar_hln = ETH_ALEN;
1862	ahdr->ar_pln = 4;
1863	ahdr->ar_op = RT_H2N_U16_C(ARPOP_REQUEST);
1864	memcpy(ahdr->ar_sha, special_ethaddr, ETH_ALEN);
1865	/* we assume that this request come from gw, but not from DNS or TFTP */
1866	ahdr->ar_sha[5] = CTL_ALIAS;
1867	(uint32_t )ahdr->ar_sip = RT_H2N_U32(RT_N2H_U32(pData->special_addr.s_addr) \| CTL_ALIAS);
1868	memset(ahdr->ar_tha, 0xff, ETH_ALEN); /broadcast/
1869	(uint32_t )ahdr->ar_tip = dst;
1870	/* warn!!! should falls in mbuf minimal size */
1871	m->m_len = sizeof(struct arphdr) + ETH_HLEN;
1872	m->m_data += ETH_HLEN;
1873	m->m_len -= ETH_HLEN;
1874	if_encap(pData, ETH_P_ARP, m, ETH_ENCAP_URG);
1875	LogFlowFuncLeave();
1876	}
1877	#ifdef VBOX_WITH_DNSMAPPING_IN_HOSTRESOLVER
1878	void slirp_add_host_resolver_mapping(PNATState pData, const char pszHostName, const char pszHostNamePattern, uint32_t u32HostIP)
1879	{
1880	LogFlowFunc(("ENTER: pszHostName:%s, pszHostNamePattern:%s u32HostIP:%RTnaipv4\n",
1881	pszHostName ? pszHostName : "(null)",
1882	pszHostNamePattern ? pszHostNamePattern : "(null)",
1883	u32HostIP));
1884	if ( ( pszHostName
1885	\|\| pszHostNamePattern)
1886	&& u32HostIP != INADDR_ANY
1887	&& u32HostIP != INADDR_BROADCAST)
1888	{
1889	PDNSMAPPINGENTRY pDnsMapping = RTMemAllocZ(sizeof(DNSMAPPINGENTRY));
1890	if (!pDnsMapping)
1891	{
1892	LogFunc(("Can't allocate DNSMAPPINGENTRY\n"));
1893	LogFlowFuncLeave();
1894	return;
1895	}
1896	pDnsMapping->u32IpAddress = u32HostIP;
1897	if (pszHostName)
1898	pDnsMapping->pszCName = RTStrDup(pszHostName);
1899	else if (pszHostNamePattern)
1900	pDnsMapping->pszPattern = RTStrDup(pszHostNamePattern);
1901	if ( !pDnsMapping->pszCName
1902	&& !pDnsMapping->pszPattern)
1903	{
1904	LogFunc(("Can't allocate enough room for %s\n", pszHostName ? pszHostName : pszHostNamePattern));
1905	RTMemFree(pDnsMapping);
1906	LogFlowFuncLeave();
1907	return;
1908	}
1909	LIST_INSERT_HEAD(&pData->DNSMapHead, pDnsMapping, MapList);
1910	LogRel(("NAT: user-defined mapping %s: %RTnaipv4 is registered\n",
1911	pDnsMapping->pszCName ? pDnsMapping->pszCName : pDnsMapping->pszPattern,
1912	pDnsMapping->u32IpAddress));
1913	}
1914	LogFlowFuncLeave();
1915	}
1916	#endif
1917
1918	/* updates the arp cache
1919	* @note: this is helper function, slirp_arp_cache_update_or_add should be used.
1920	* @returns 0 - if has found and updated
1921	* 1 - if hasn't found.
1922	*/
1923	static inline int slirp_arp_cache_update(PNATState pData, uint32_t dst, const uint8_t *mac)
1924	{
1925	struct arp_cache_entry *ac;
1926	Assert(( memcmp(mac, broadcast_ethaddr, ETH_ALEN)
1927	&& memcmp(mac, zerro_ethaddr, ETH_ALEN)));
1928	LIST_FOREACH(ac, &pData->arp_cache, list)
1929	{
1930	if (!memcmp(ac->ether, mac, ETH_ALEN))
1931	{
1932	ac->ip = dst;
1933	return 0;
1934	}
1935	}
1936	return 1;
1937	}
1938
1939	/**
1940	* add entry to the arp cache
1941	* @note: this is helper function, slirp_arp_cache_update_or_add should be used.
1942	*/
1943	static inline void slirp_arp_cache_add(PNATState pData, uint32_t ip, const uint8_t *ether)
1944	{
1945	struct arp_cache_entry *ac = NULL;
1946	Assert(( memcmp(ether, broadcast_ethaddr, ETH_ALEN)
1947	&& memcmp(ether, zerro_ethaddr, ETH_ALEN)));
1948	ac = RTMemAllocZ(sizeof(struct arp_cache_entry));
1949	if (ac == NULL)
1950	{
1951	Log(("NAT: Can't allocate arp cache entry\n"));
1952	return;
1953	}
1954	ac->ip = ip;
1955	memcpy(ac->ether, ether, ETH_ALEN);
1956	LIST_INSERT_HEAD(&pData->arp_cache, ac, list);
1957	}
1958
1959	/* updates or adds entry to the arp cache
1960	* @returns 0 - if has found and updated
1961	* 1 - if hasn't found.
1962	*/
1963	int slirp_arp_cache_update_or_add(PNATState pData, uint32_t dst, const uint8_t *mac)
1964	{
1965	if ( !memcmp(mac, broadcast_ethaddr, ETH_ALEN)
1966	\|\| !memcmp(mac, zerro_ethaddr, ETH_ALEN))
1967	{
1968	static bool fBroadcastEtherAddReported;
1969	if (!fBroadcastEtherAddReported)
1970	{
1971	LogRel(("NAT: Attempt to add pair [%RTmac:%RTnaipv4] in ARP cache was ignored\n",
1972	mac, dst));
1973	fBroadcastEtherAddReported = true;
1974	}
1975	return 1;
1976	}
1977	if (slirp_arp_cache_update(pData, dst, mac))
1978	slirp_arp_cache_add(pData, dst, mac);
1979
1980	return 0;
1981	}
1982
1983
1984	void slirp_set_mtu(PNATState pData, int mtu)
1985	{
1986	if (mtu < 20 \|\| mtu >= 16000)
1987	{
1988	LogRel(("NAT: mtu(%d) is out of range (20;16000] mtu forcely assigned to 1500\n", mtu));
1989	mtu = 1500;
1990	}
1991	/* MTU is maximum transition unit on */
1992	if_mtu =
1993	if_mru = mtu;
1994	}
1995
1996	/**
1997	* Info handler.
1998	*/
1999	void slirp_info(PNATState pData, const void pvArg, const char pszArgs)
2000	{
2001	struct socket so, so_next;
2002	struct arp_cache_entry *ac;
2003	struct port_forward_rule *rule;
2004	PCDBGFINFOHLP pHlp = (PCDBGFINFOHLP)pvArg;
2005	NOREF(pszArgs);
2006
2007	pHlp->pfnPrintf(pHlp, "NAT parameters: MTU=%d\n", if_mtu);
2008	pHlp->pfnPrintf(pHlp, "NAT TCP ports:\n");
2009	QSOCKET_FOREACH(so, so_next, tcp)
2010	/* { */
2011	pHlp->pfnPrintf(pHlp, " %R[natsock]\n", so);
2012	}
2013
2014	pHlp->pfnPrintf(pHlp, "NAT UDP ports:\n");
2015	QSOCKET_FOREACH(so, so_next, udp)
2016	/* { */
2017	pHlp->pfnPrintf(pHlp, " %R[natsock]\n", so);
2018	}
2019
2020	pHlp->pfnPrintf(pHlp, "NAT ARP cache:\n");
2021	LIST_FOREACH(ac, &pData->arp_cache, list)
2022	{
2023	pHlp->pfnPrintf(pHlp, " %RTnaipv4 %RTmac\n", ac->ip, &ac->ether);
2024	}
2025
2026	pHlp->pfnPrintf(pHlp, "NAT rules:\n");
2027	LIST_FOREACH(rule, &pData->port_forward_rule_head, list)
2028	{
2029	pHlp->pfnPrintf(pHlp, " %s %d => %RTnaipv4:%d %c\n",
2030	rule->proto == IPPROTO_UDP ? "UDP" : "TCP",
2031	rule->host_port, rule->guest_addr.s_addr, rule->guest_port,
2032	rule->activated ? ' ' : '*');
2033	}
2034	}

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

source: vbox/trunk/src/VBox/Devices/Network/slirp/slirp.c@ 49435

Download in other formats: