slirp.c@ 53350

Last change on this file since 53350 was 53338, checked in by vboxsync, 10 years ago
NAT: if_encap: assert that mbuf has enough M_LEADINGSPACE to accomodate ethernet header.
Property svn:eol-style set to `native` Property svn:keywords set to `Author Date Id Revision`
File size: 67.7 KB

Line
1	/* $Id: slirp.c 53338 2014-11-16 01:56:38Z 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	* XXX: It shouldn't be non-Windows specific.
68	* resolv_conf_parser.h client's structure isn't OS specific, it's just need to be generalized a
69	* a bit to replace slirp_state.h DNS server (domain) lists with rcp_state like structure.
70	*/
71	# include "resolv_conf_parser.h"
72	#endif
73
74	#ifndef RT_OS_WINDOWS
75	# define DO_ENGAGE_EVENT1(so, fdset, label) \
76	do { \
77	if ( so->so_poll_index != -1 \
78	&& so->s == polls[so->so_poll_index].fd) \
79	{ \
80	polls[so->so_poll_index].events \|= N_(fdset ## _poll); \
81	break; \
82	} \
83	AssertRelease(poll_index < (nfds)); \
84	AssertRelease(poll_index >= 0 && poll_index < (nfds)); \
85	polls[poll_index].fd = (so)->s; \
86	(so)->so_poll_index = poll_index; \
87	polls[poll_index].events = N_(fdset ## _poll); \
88	polls[poll_index].revents = 0; \
89	poll_index++; \
90	} while (0)
91
92	# define DO_ENGAGE_EVENT2(so, fdset1, fdset2, label) \
93	do { \
94	if ( so->so_poll_index != -1 \
95	&& so->s == polls[so->so_poll_index].fd) \
96	{ \
97	polls[so->so_poll_index].events \|= \
98	N_(fdset1 ## _poll) \| N_(fdset2 ## _poll); \
99	break; \
100	} \
101	AssertRelease(poll_index < (nfds)); \
102	polls[poll_index].fd = (so)->s; \
103	(so)->so_poll_index = poll_index; \
104	polls[poll_index].events = \
105	N_(fdset1 ## _poll) \| N_(fdset2 ## _poll); \
106	poll_index++; \
107	} while (0)
108
109	# define DO_POLL_EVENTS(rc, error, so, events, label) do {} while (0)
110
111	/*
112	* DO_CHECK_FD_SET is used in dumping events on socket, including POLLNVAL.
113	* gcc warns about attempts to log POLLNVAL so construction in a last to lines
114	* used to catch POLLNVAL while logging and return false in case of error while
115	* normal usage.
116	*/
117	# define DO_CHECK_FD_SET(so, events, fdset) \
118	( ((so)->so_poll_index != -1) \
119	&& ((so)->so_poll_index <= ndfs) \
120	&& ((so)->s == polls[so->so_poll_index].fd) \
121	&& (polls[(so)->so_poll_index].revents & N_(fdset ## _poll)) \
122	&& ( N_(fdset ## _poll) == POLLNVAL \
123	\|\| !(polls[(so)->so_poll_index].revents & POLLNVAL)))
124
125	/* specific for Windows Winsock API */
126	# define DO_WIN_CHECK_FD_SET(so, events, fdset) 0
127
128	# ifndef RT_OS_LINUX
129	# define readfds_poll (POLLRDNORM)
130	# define writefds_poll (POLLWRNORM)
131	# else
132	# define readfds_poll (POLLIN)
133	# define writefds_poll (POLLOUT)
134	# endif
135	# define xfds_poll (POLLPRI)
136	# define closefds_poll (POLLHUP)
137	# define rderr_poll (POLLERR)
138	# if 0 /* unused yet */
139	# define rdhup_poll (POLLHUP)
140	# define nval_poll (POLLNVAL)
141	# endif
142
143	# define ICMP_ENGAGE_EVENT(so, fdset) \
144	do { \
145	if (pData->icmp_socket.s != -1) \
146	DO_ENGAGE_EVENT1((so), fdset, ICMP); \
147	} while (0)
148
149	#else /* RT_OS_WINDOWS */
150
151	/*
152	* On Windows, we will be notified by IcmpSendEcho2() when the response arrives.
153	* So no call to WSAEventSelect necessary.
154	*/
155	# define ICMP_ENGAGE_EVENT(so, fdset) do {} while (0)
156
157	/*
158	* On Windows we use FD_ALL_EVENTS to ensure that we don't miss any event.
159	*/
160	# define DO_ENGAGE_EVENT1(so, fdset1, label) \
161	do { \
162	rc = WSAEventSelect((so)->s, VBOX_SOCKET_EVENT, FD_ALL_EVENTS); \
163	if (rc == SOCKET_ERROR) \
164	{ \
165	/* This should not happen */ \
166	error = WSAGetLastError(); \
167	LogRel(("WSAEventSelect (" #label ") error %d (so=%x, socket=%s, event=%x)\n", \
168	error, (so), (so)->s, VBOX_SOCKET_EVENT)); \
169	} \
170	} while (0); \
171	CONTINUE(label)
172
173	# define DO_ENGAGE_EVENT2(so, fdset1, fdset2, label) \
174	DO_ENGAGE_EVENT1((so), (fdset1), label)
175
176	# define DO_POLL_EVENTS(rc, error, so, events, label) \
177	(rc) = WSAEnumNetworkEvents((so)->s, VBOX_SOCKET_EVENT, (events)); \
178	if ((rc) == SOCKET_ERROR) \
179	{ \
180	(error) = WSAGetLastError(); \
181	LogRel(("WSAEnumNetworkEvents %R[natsock] " #label " error %d\n", (so), (error))); \
182	LogFunc(("WSAEnumNetworkEvents %R[natsock] " #label " error %d\n", (so), (error))); \
183	CONTINUE(label); \
184	}
185
186	# define acceptds_win FD_ACCEPT
187	# define acceptds_win_bit FD_ACCEPT_BIT
188	# define readfds_win FD_READ
189	# define readfds_win_bit FD_READ_BIT
190	# define writefds_win FD_WRITE
191	# define writefds_win_bit FD_WRITE_BIT
192	# define xfds_win FD_OOB
193	# define xfds_win_bit FD_OOB_BIT
194	# define closefds_win FD_CLOSE
195	# define closefds_win_bit FD_CLOSE_BIT
196	# define connectfds_win FD_CONNECT
197	# define connectfds_win_bit FD_CONNECT_BIT
198
199	# define closefds_win FD_CLOSE
200	# define closefds_win_bit FD_CLOSE_BIT
201
202	# define DO_CHECK_FD_SET(so, events, fdset) \
203	(((events).lNetworkEvents & fdset ## _win) && ((events).iErrorCode[fdset ## _win_bit] == 0))
204
205	# define DO_WIN_CHECK_FD_SET(so, events, fdset) DO_CHECK_FD_SET((so), (events), fdset)
206	# define DO_UNIX_CHECK_FD_SET(so, events, fdset) 1 /specific for Unix API /
207
208	#endif /* RT_OS_WINDOWS */
209
210	#define TCP_ENGAGE_EVENT1(so, fdset) \
211	DO_ENGAGE_EVENT1((so), fdset, tcp)
212
213	#define TCP_ENGAGE_EVENT2(so, fdset1, fdset2) \
214	DO_ENGAGE_EVENT2((so), fdset1, fdset2, tcp)
215
216	#ifdef RT_OS_WINDOWS
217	# define WIN_TCP_ENGAGE_EVENT2(so, fdset, fdset2) TCP_ENGAGE_EVENT2(so, fdset1, fdset2)
218	#endif
219
220	#define UDP_ENGAGE_EVENT(so, fdset) \
221	DO_ENGAGE_EVENT1((so), fdset, udp)
222
223	#define POLL_TCP_EVENTS(rc, error, so, events) \
224	DO_POLL_EVENTS((rc), (error), (so), (events), tcp)
225
226	#define POLL_UDP_EVENTS(rc, error, so, events) \
227	DO_POLL_EVENTS((rc), (error), (so), (events), udp)
228
229	#define CHECK_FD_SET(so, events, set) \
230	(DO_CHECK_FD_SET((so), (events), set))
231
232	#define WIN_CHECK_FD_SET(so, events, set) \
233	(DO_WIN_CHECK_FD_SET((so), (events), set))
234
235	/*
236	* Loging macros
237	*/
238	#if VBOX_WITH_DEBUG_NAT_SOCKETS
239	# if defined(RT_OS_WINDOWS)
240	# define DO_LOG_NAT_SOCK(so, proto, winevent, r_fdset, w_fdset, x_fdset) \
241	do { \
242	LogRel((" " #proto " %R[natsock] %R[natwinnetevents]\n", (so), (winevent))); \
243	} while (0)
244	# else /* !RT_OS_WINDOWS */
245	# define DO_LOG_NAT_SOCK(so, proto, winevent, r_fdset, w_fdset, x_fdset) \
246	do { \
247	LogRel((" " #proto " %R[natsock] %s %s %s er: %s, %s, %s\n", (so), \
248	CHECK_FD_SET(so, ign ,r_fdset) ? "READ":"", \
249	CHECK_FD_SET(so, ign, w_fdset) ? "WRITE":"", \
250	CHECK_FD_SET(so, ign, x_fdset) ? "OOB":"", \
251	CHECK_FD_SET(so, ign, rderr) ? "RDERR":"", \
252	CHECK_FD_SET(so, ign, rdhup) ? "RDHUP":"", \
253	CHECK_FD_SET(so, ign, nval) ? "RDNVAL":"")); \
254	} while (0)
255	# endif /* !RT_OS_WINDOWS */
256	#else /* !VBOX_WITH_DEBUG_NAT_SOCKETS */
257	# define DO_LOG_NAT_SOCK(so, proto, winevent, r_fdset, w_fdset, x_fdset) do {} while (0)
258	#endif /* !VBOX_WITH_DEBUG_NAT_SOCKETS */
259
260	#define LOG_NAT_SOCK(so, proto, winevent, r_fdset, w_fdset, x_fdset) \
261	DO_LOG_NAT_SOCK((so), proto, (winevent), r_fdset, w_fdset, x_fdset)
262
263	static void activate_port_forwarding(PNATState, const uint8_t *pEther);
264
265	static const uint8_t special_ethaddr[6] =
266	{
267	0x52, 0x54, 0x00, 0x12, 0x35, 0x00
268	};
269
270	static const uint8_t broadcast_ethaddr[6] =
271	{
272	0xff, 0xff, 0xff, 0xff, 0xff, 0xff
273	};
274
275	const uint8_t zerro_ethaddr[6] =
276	{
277	0x0, 0x0, 0x0, 0x0, 0x0, 0x0
278	};
279
280	/**
281	* This helper routine do the checks in descriptions to
282	* ''fUnderPolling'' and ''fShouldBeRemoved'' flags
283	* @returns 1 if socket removed and 0 if no changes was made.
284	*/
285	static int slirpVerifyAndFreeSocket(PNATState pData, struct socket *pSocket)
286	{
287	AssertPtrReturn(pData, 0);
288	AssertPtrReturn(pSocket, 0);
289	AssertReturn(pSocket->fUnderPolling, 0);
290	if (pSocket->fShouldBeRemoved)
291	{
292	pSocket->fUnderPolling = 0;
293	sofree(pData, pSocket);
294	/* pSocket is PHANTOM, now */
295	return 1;
296	}
297	return 0;
298	}
299
300	int slirp_init(PNATState *ppData, uint32_t u32NetAddr, uint32_t u32Netmask,
301	bool fPassDomain, bool fUseHostResolver, int i32AliasMode,
302	int iIcmpCacheLimit, void *pvUser)
303	{
304	int rc;
305	PNATState pData;
306	if (u32Netmask & 0x1f)
307	/* CTL is x.x.x.15, bootp passes up to 16 IPs (15..31) */
308	return VERR_INVALID_PARAMETER;
309	pData = RTMemAllocZ(RT_ALIGN_Z(sizeof(NATState), sizeof(uint64_t)));
310	*ppData = pData;
311	if (!pData)
312	return VERR_NO_MEMORY;
313	pData->fPassDomain = !fUseHostResolver ? fPassDomain : false;
314	pData->fUseHostResolver = fUseHostResolver;
315	pData->fUseHostResolverPermanent = fUseHostResolver;
316	pData->pvUser = pvUser;
317	pData->netmask = u32Netmask;
318
319	rc = RTCritSectRwInit(&pData->CsRwHandlerChain);
320	if (RT_FAILURE(rc))
321	return rc;
322
323	/* sockets & TCP defaults */
324	pData->socket_rcv = 64 * _1K;
325	pData->socket_snd = 64 * _1K;
326	tcp_sndspace = 64 * _1K;
327	tcp_rcvspace = 64 * _1K;
328
329	/*
330	* Use the same default here as in DevNAT.cpp (SoMaxConnection CFGM value)
331	* to avoid release log noise.
332	*/
333	pData->soMaxConn = 10;
334
335	#ifdef RT_OS_WINDOWS
336	{
337	WSADATA Data;
338	RTLDRMOD hLdrMod;
339
340	WSAStartup(MAKEWORD(2, 0), &Data);
341
342	rc = RTLdrLoadSystem("Iphlpapi.dll", /* :fNoUnload */ true, &hLdrMod);
343	if (RT_SUCCESS(rc))
344	{
345	rc = RTLdrGetSymbol(hLdrMod, "GetAdaptersAddresses", (void **)&pData->pfGetAdaptersAddresses);
346	if (RT_FAILURE(rc))
347	LogRel(("NAT: Can't find GetAdapterAddresses in Iphlpapi.dll\n"));
348
349	RTLdrClose(hLdrMod);
350	}
351	}
352	pData->phEvents[VBOX_SOCKET_EVENT_INDEX] = CreateEvent(NULL, FALSE, FALSE, NULL);
353	#endif
354
355	rc = bootp_dhcp_init(pData);
356	if (RT_FAILURE(rc))
357	{
358	Log(("NAT: DHCP server initialization failed\n"));
359	RTMemFree(pData);
360	*ppData = NULL;
361	return rc;
362	}
363	debug_init(pData);
364	if_init(pData);
365	ip_init(pData);
366	icmp_init(pData, iIcmpCacheLimit);
367
368	/* Initialise mbufs after setting the MTU */
369	mbuf_init(pData);
370
371	pData->special_addr.s_addr = u32NetAddr;
372	pData->slirp_ethaddr = &special_ethaddr[0];
373	alias_addr.s_addr = pData->special_addr.s_addr \| RT_H2N_U32_C(CTL_ALIAS);
374	/* @todo: add ability to configure this staff */
375
376	/* set default addresses */
377	inet_aton("127.0.0.1", &loopback_addr);
378
379	rc = slirpTftpInit(pData);
380	AssertRCReturn(rc, VINF_NAT_DNS);
381
382	if (i32AliasMode & ~(PKT_ALIAS_LOG\|PKT_ALIAS_SAME_PORTS\|PKT_ALIAS_PROXY_ONLY))
383	{
384	Log(("NAT: alias mode %x is ignored\n", i32AliasMode));
385	i32AliasMode = 0;
386	}
387	pData->i32AliasMode = i32AliasMode;
388	getouraddr(pData);
389	{
390	int flags = 0;
391	struct in_addr proxy_addr;
392	pData->proxy_alias = LibAliasInit(pData, NULL);
393	if (pData->proxy_alias == NULL)
394	{
395	Log(("NAT: LibAlias default rule wasn't initialized\n"));
396	AssertMsgFailed(("NAT: LibAlias default rule wasn't initialized\n"));
397	}
398	flags = LibAliasSetMode(pData->proxy_alias, 0, 0);
399	#ifndef NO_FW_PUNCH
400	flags \|= PKT_ALIAS_PUNCH_FW;
401	#endif
402	flags \|= pData->i32AliasMode; /* do transparent proxying */
403	flags = LibAliasSetMode(pData->proxy_alias, flags, ~0);
404	proxy_addr.s_addr = RT_H2N_U32(RT_N2H_U32(pData->special_addr.s_addr) \| CTL_ALIAS);
405	LibAliasSetAddress(pData->proxy_alias, proxy_addr);
406	ftp_alias_load(pData);
407	nbt_alias_load(pData);
408	if (pData->fUseHostResolver)
409	dns_alias_load(pData);
410	}
411	#ifdef VBOX_WITH_NAT_SEND2HOME
412	/* @todo: we should know all interfaces available on host. */
413	pData->pInSockAddrHomeAddress = RTMemAllocZ(sizeof(struct sockaddr));
414	pData->cInHomeAddressSize = 1;
415	inet_aton("192.168.1.25", &pData->pInSockAddrHomeAddress[0].sin_addr);
416	pData->pInSockAddrHomeAddress[0].sin_family = AF_INET;
417	# ifdef RT_OS_DARWIN
418	pData->pInSockAddrHomeAddress[0].sin_len = sizeof(struct sockaddr_in);
419	# endif
420	#endif
421
422	slirp_link_up(pData);
423	return VINF_SUCCESS;
424	}
425
426	/**
427	* Register statistics.
428	*/
429	void slirp_register_statistics(PNATState pData, PPDMDRVINS pDrvIns)
430	{
431	#ifdef VBOX_WITH_STATISTICS
432	# define PROFILE_COUNTER(name, dsc) REGISTER_COUNTER(name, pData, STAMTYPE_PROFILE, STAMUNIT_TICKS_PER_CALL, dsc)
433	# define COUNTING_COUNTER(name, dsc) REGISTER_COUNTER(name, pData, STAMTYPE_COUNTER, STAMUNIT_COUNT, dsc)
434	# include "counters.h"
435	# undef COUNTER
436	/** @todo register statistics for the variables dumped by:
437	* ipstats(pData); tcpstats(pData); udpstats(pData); icmpstats(pData);
438	* mbufstats(pData); sockstats(pData); */
439	#else /* VBOX_WITH_STATISTICS */
440	NOREF(pData);
441	NOREF(pDrvIns);
442	#endif /* !VBOX_WITH_STATISTICS */
443	}
444
445	/**
446	* Deregister statistics.
447	*/
448	void slirp_deregister_statistics(PNATState pData, PPDMDRVINS pDrvIns)
449	{
450	if (pData == NULL)
451	return;
452	#ifdef VBOX_WITH_STATISTICS
453	# define PROFILE_COUNTER(name, dsc) DEREGISTER_COUNTER(name, pData)
454	# define COUNTING_COUNTER(name, dsc) DEREGISTER_COUNTER(name, pData)
455	# include "counters.h"
456	#else /* VBOX_WITH_STATISTICS */
457	NOREF(pData);
458	NOREF(pDrvIns);
459	#endif /* !VBOX_WITH_STATISTICS */
460	}
461
462	/**
463	* Marks the link as up, making it possible to establish new connections.
464	*/
465	void slirp_link_up(PNATState pData)
466	{
467	struct arp_cache_entry *ac;
468
469	if (link_up == 1)
470	return;
471
472	link_up = 1;
473
474	if (!pData->fUseHostResolverPermanent)
475	slirpInitializeDnsSettings(pData);
476
477	if (LIST_EMPTY(&pData->arp_cache))
478	return;
479
480	LIST_FOREACH(ac, &pData->arp_cache, list)
481	{
482	activate_port_forwarding(pData, ac->ether);
483	}
484	}
485
486	/**
487	* Marks the link as down and cleans up the current connections.
488	*/
489	void slirp_link_down(PNATState pData)
490	{
491	struct socket *so;
492	struct port_forward_rule *rule;
493
494	if (link_up == 0)
495	return;
496
497	slirpReleaseDnsSettings(pData);
498
499	while ((so = tcb.so_next) != &tcb)
500	{
501	/* Don't miss TCB releasing */
502	if ( !sototcpcb(so)
503	&& ( so->so_state & SS_NOFDREF
504	\|\| so->s == -1))
505	sofree(pData, so);
506	else
507	tcp_close(pData, sototcpcb(so));
508	}
509
510	while ((so = udb.so_next) != &udb)
511	udp_detach(pData, so);
512
513	/*
514	* Clear the active state of port-forwarding rules to force
515	* re-setup on restoration of communications.
516	*/
517	LIST_FOREACH(rule, &pData->port_forward_rule_head, list)
518	{
519	rule->activated = 0;
520	}
521	pData->cRedirectionsActive = 0;
522
523	link_up = 0;
524	}
525
526	/**
527	* Terminates the slirp component.
528	*/
529	void slirp_term(PNATState pData)
530	{
531	if (pData == NULL)
532	return;
533	icmp_finit(pData);
534
535	slirp_link_down(pData);
536	ftp_alias_unload(pData);
537	nbt_alias_unload(pData);
538	if (pData->fUseHostResolver)
539	{
540	dns_alias_unload(pData);
541	#ifdef VBOX_WITH_DNSMAPPING_IN_HOSTRESOLVER
542	while (!LIST_EMPTY(&pData->DNSMapHead))
543	{
544	PDNSMAPPINGENTRY pDnsEntry = LIST_FIRST(&pData->DNSMapHead);
545	LIST_REMOVE(pDnsEntry, MapList);
546	RTStrFree(pDnsEntry->pszCName);
547	RTMemFree(pDnsEntry);
548	}
549	#endif
550	}
551	while (!LIST_EMPTY(&instancehead))
552	{
553	struct libalias *la = LIST_FIRST(&instancehead);
554	/* libalias do all clean up */
555	LibAliasUninit(la);
556	}
557	while (!LIST_EMPTY(&pData->arp_cache))
558	{
559	struct arp_cache_entry *ac = LIST_FIRST(&pData->arp_cache);
560	LIST_REMOVE(ac, list);
561	RTMemFree(ac);
562	}
563	slirpTftpTerm(pData);
564	bootp_dhcp_fini(pData);
565	m_fini(pData);
566	#ifdef RT_OS_WINDOWS
567	WSACleanup();
568	#endif
569	#ifndef VBOX_WITH_SLIRP_BSD_SBUF
570	#ifdef LOG_ENABLED
571	Log(("\n"
572	"NAT statistics\n"
573	"--------------\n"
574	"\n"));
575	ipstats(pData);
576	tcpstats(pData);
577	udpstats(pData);
578	icmpstats(pData);
579	mbufstats(pData);
580	sockstats(pData);
581	Log(("\n"
582	"\n"
583	"\n"));
584	#endif
585	#endif
586	RTCritSectRwDelete(&pData->CsRwHandlerChain);
587	RTMemFree(pData);
588	}
589
590
591	#define CONN_CANFSEND(so) (((so)->so_state & (SS_FCANTSENDMORE\|SS_ISFCONNECTED)) == SS_ISFCONNECTED)
592	#define CONN_CANFRCV(so) (((so)->so_state & (SS_FCANTRCVMORE\|SS_ISFCONNECTED)) == SS_ISFCONNECTED)
593
594	/*
595	* curtime kept to an accuracy of 1ms
596	*/
597	static void updtime(PNATState pData)
598	{
599	#ifdef RT_OS_WINDOWS
600	struct _timeb tb;
601
602	_ftime(&tb);
603	curtime = (u_int)tb.time * (u_int)1000;
604	curtime += (u_int)tb.millitm;
605	#else
606	gettimeofday(&tt, 0);
607
608	curtime = (u_int)tt.tv_sec * (u_int)1000;
609	curtime += (u_int)tt.tv_usec / (u_int)1000;
610
611	if ((tt.tv_usec % 1000) >= 500)
612	curtime++;
613	#endif
614	}
615
616	#ifdef RT_OS_WINDOWS
617	void slirp_select_fill(PNATState pData, int *pnfds)
618	#else /* RT_OS_WINDOWS */
619	void slirp_select_fill(PNATState pData, int pnfds, struct pollfd polls)
620	#endif /* !RT_OS_WINDOWS */
621	{
622	struct socket so, so_next;
623	int nfds;
624	#if defined(RT_OS_WINDOWS)
625	int rc;
626	int error;
627	#else
628	int poll_index = 0;
629	#endif
630	int i;
631
632	STAM_PROFILE_START(&pData->StatFill, a);
633
634	nfds = *pnfds;
635
636	/*
637	* First, TCP sockets
638	*/
639	do_slowtimo = 0;
640	if (!link_up)
641	goto done;
642
643	/*
644	* *_slowtimo needs calling if there are IP fragments
645	* in the fragment queue, or there are TCP connections active
646	*/
647	/* XXX:
648	* triggering of fragment expiration should be the same but use new macroses
649	*/
650	do_slowtimo = (tcb.so_next != &tcb);
651	if (!do_slowtimo)
652	{
653	for (i = 0; i < IPREASS_NHASH; i++)
654	{
655	if (!TAILQ_EMPTY(&ipq[i]))
656	{
657	do_slowtimo = 1;
658	break;
659	}
660	}
661	}
662	/* always add the ICMP socket */
663	#ifndef RT_OS_WINDOWS
664	pData->icmp_socket.so_poll_index = -1;
665	#endif
666	ICMP_ENGAGE_EVENT(&pData->icmp_socket, readfds);
667
668	STAM_COUNTER_RESET(&pData->StatTCP);
669	STAM_COUNTER_RESET(&pData->StatTCPHot);
670
671	QSOCKET_FOREACH(so, so_next, tcp)
672	/* { */
673	Assert(so->so_type == IPPROTO_TCP);
674	#if !defined(RT_OS_WINDOWS)
675	so->so_poll_index = -1;
676	#endif
677	STAM_COUNTER_INC(&pData->StatTCP);
678	#ifdef VBOX_WITH_NAT_UDP_SOCKET_CLONE
679	/* TCP socket can't be cloned */
680	Assert((!so->so_cloneOf));
681	#endif
682	/*
683	* See if we need a tcp_fasttimo
684	*/
685	if ( time_fasttimo == 0
686	&& so->so_tcpcb != NULL
687	&& so->so_tcpcb->t_flags & TF_DELACK)
688	{
689	time_fasttimo = curtime; /* Flag when we want a fasttimo */
690	}
691
692	/*
693	* NOFDREF can include still connecting to local-host,
694	* newly socreated() sockets etc. Don't want to select these.
695	*/
696	if (so->so_state & SS_NOFDREF \|\| so->s == -1)
697	CONTINUE(tcp);
698
699	/*
700	* Set for reading sockets which are accepting
701	*/
702	if (so->so_state & SS_FACCEPTCONN)
703	{
704	STAM_COUNTER_INC(&pData->StatTCPHot);
705	TCP_ENGAGE_EVENT1(so, readfds);
706	CONTINUE(tcp);
707	}
708
709	/*
710	* Set for writing sockets which are connecting
711	*/
712	if (so->so_state & SS_ISFCONNECTING)
713	{
714	Log2(("connecting %R[natsock] engaged\n",so));
715	STAM_COUNTER_INC(&pData->StatTCPHot);
716	#ifdef RT_OS_WINDOWS
717	WIN_TCP_ENGAGE_EVENT2(so, writefds, connectfds);
718	#else
719	TCP_ENGAGE_EVENT1(so, writefds);
720	#endif
721	}
722
723	/*
724	* Set for writing if we are connected, can send more, and
725	* we have something to send
726	*/
727	if (CONN_CANFSEND(so) && SBUF_LEN(&so->so_rcv))
728	{
729	STAM_COUNTER_INC(&pData->StatTCPHot);
730	TCP_ENGAGE_EVENT1(so, writefds);
731	}
732
733	/*
734	* Set for reading (and urgent data) if we are connected, can
735	* receive more, and we have room for it XXX /2 ?
736	*/
737	/* @todo: vvl - check which predicat here will be more useful here in rerm of new sbufs. */
738	if ( CONN_CANFRCV(so)
739	&& (SBUF_LEN(&so->so_snd) < (SBUF_SIZE(&so->so_snd)/2))
740	#ifdef RT_OS_WINDOWS
741	&& !(so->so_state & SS_ISFCONNECTING)
742	#endif
743	)
744	{
745	STAM_COUNTER_INC(&pData->StatTCPHot);
746	TCP_ENGAGE_EVENT2(so, readfds, xfds);
747	}
748	LOOP_LABEL(tcp, so, so_next);
749	}
750
751	/*
752	* UDP sockets
753	*/
754	STAM_COUNTER_RESET(&pData->StatUDP);
755	STAM_COUNTER_RESET(&pData->StatUDPHot);
756
757	QSOCKET_FOREACH(so, so_next, udp)
758	/* { */
759
760	Assert(so->so_type == IPPROTO_UDP);
761	STAM_COUNTER_INC(&pData->StatUDP);
762	#if !defined(RT_OS_WINDOWS)
763	so->so_poll_index = -1;
764	#endif
765
766	/*
767	* See if it's timed out
768	*/
769	if (so->so_expire)
770	{
771	if (so->so_expire <= curtime)
772	{
773	Log2(("NAT: %R[natsock] expired\n", so));
774	if (so->so_timeout != NULL)
775	{
776	/* so_timeout - might change the so_expire value or
777	* drop so_timeout* from so.
778	*/
779	so->so_timeout(pData, so, so->so_timeout_arg);
780	/* on 4.2 so->
781	*/
782	if ( so_next->so_prev != so /* so_timeout freed the socket */
783	\|\| so->so_timeout) /* so_timeout just freed so_timeout */
784	CONTINUE_NO_UNLOCK(udp);
785	}
786	UDP_DETACH(pData, so, so_next);
787	CONTINUE_NO_UNLOCK(udp);
788	}
789	}
790	#ifdef VBOX_WITH_NAT_UDP_SOCKET_CLONE
791	if (so->so_cloneOf)
792	CONTINUE_NO_UNLOCK(udp);
793	#endif
794
795	/*
796	* When UDP packets are received from over the link, they're
797	* sendto()'d straight away, so no need for setting for writing
798	* Limit the number of packets queued by this session to 4.
799	* Note that even though we try and limit this to 4 packets,
800	* the session could have more queued if the packets needed
801	* to be fragmented.
802	*
803	* (XXX <= 4 ?)
804	*/
805	if ((so->so_state & SS_ISFCONNECTED) && so->so_queued <= 4)
806	{
807	STAM_COUNTER_INC(&pData->StatUDPHot);
808	UDP_ENGAGE_EVENT(so, readfds);
809	}
810	LOOP_LABEL(udp, so, so_next);
811	}
812	done:
813
814	#if defined(RT_OS_WINDOWS)
815	*pnfds = VBOX_EVENT_COUNT;
816	#else /* RT_OS_WINDOWS */
817	AssertRelease(poll_index <= *pnfds);
818	*pnfds = poll_index;
819	#endif /* !RT_OS_WINDOWS */
820
821	STAM_PROFILE_STOP(&pData->StatFill, a);
822	}
823
824
825	/**
826	* This function do Connection or sending tcp sequence to.
827	* @returns if true operation completed
828	* @note: functions call tcp_input that potentially could lead to tcp_drop
829	*/
830	static bool slirpConnectOrWrite(PNATState pData, struct socket *so, bool fConnectOnly)
831	{
832	int ret;
833	LogFlowFunc(("ENTER: so:%R[natsock], fConnectOnly:%RTbool\n", so, fConnectOnly));
834	/*
835	* Check for non-blocking, still-connecting sockets
836	*/
837	if (so->so_state & SS_ISFCONNECTING)
838	{
839	Log2(("connecting %R[natsock] catched\n", so));
840	/* Connected */
841	so->so_state &= ~SS_ISFCONNECTING;
842
843	/*
844	* This should be probably guarded by PROBE_CONN too. Anyway,
845	* we disable it on OS/2 because the below send call returns
846	* EFAULT which causes the opened TCP socket to close right
847	* after it has been opened and connected.
848	*/
849	#ifndef RT_OS_OS2
850	ret = send(so->s, (const char *)&ret, 0, 0);
851	if (ret < 0)
852	{
853	/* XXXXX Must fix, zero bytes is a NOP */
854	if ( soIgnorableErrorCode(errno)
855	\|\| errno == ENOTCONN)
856	{
857	LogFlowFunc(("LEAVE: false\n"));
858	return false;
859	}
860
861	/* else failed */
862	so->so_state = SS_NOFDREF;
863	}
864	/* else so->so_state &= ~SS_ISFCONNECTING; */
865	#endif
866
867	/*
868	* Continue tcp_input
869	*/
870	TCP_INPUT(pData, (struct mbuf *)NULL, sizeof(struct ip), so);
871	/* continue; */
872	}
873	else if (!fConnectOnly)
874	{
875	SOWRITE(ret, pData, so);
876	if (RT_LIKELY(ret > 0))
877	{
878	/*
879	* Make sure we will send window update to peer. This is
880	* a moral equivalent of calling tcp_output() for PRU_RCVD
881	* in tcp_usrreq() of the real stack.
882	*/
883	struct tcpcb *tp = sototcpcb(so);
884	if (RT_LIKELY(tp != NULL))
885	tp->t_flags \|= TF_DELACK;
886	}
887	}
888
889	LogFlowFunc(("LEAVE: true\n"));
890	return true;
891	}
892
893	#if defined(RT_OS_WINDOWS)
894	void slirp_select_poll(PNATState pData, int fTimeout, int fIcmp)
895	#else /* RT_OS_WINDOWS */
896	void slirp_select_poll(PNATState pData, struct pollfd *polls, int ndfs)
897	#endif /* !RT_OS_WINDOWS */
898	{
899	struct socket so, so_next;
900	int ret;
901	#if defined(RT_OS_WINDOWS)
902	WSANETWORKEVENTS NetworkEvents;
903	int rc;
904	int error;
905	#endif
906
907	STAM_PROFILE_START(&pData->StatPoll, a);
908
909	/* Update time */
910	updtime(pData);
911
912	/*
913	* See if anything has timed out
914	*/
915	if (link_up)
916	{
917	if (time_fasttimo && ((curtime - time_fasttimo) >= 2))
918	{
919	STAM_PROFILE_START(&pData->StatFastTimer, b);
920	tcp_fasttimo(pData);
921	time_fasttimo = 0;
922	STAM_PROFILE_STOP(&pData->StatFastTimer, b);
923	}
924	if (do_slowtimo && ((curtime - last_slowtimo) >= 499))
925	{
926	STAM_PROFILE_START(&pData->StatSlowTimer, c);
927	ip_slowtimo(pData);
928	tcp_slowtimo(pData);
929	last_slowtimo = curtime;
930	STAM_PROFILE_STOP(&pData->StatSlowTimer, c);
931	}
932	}
933	#if defined(RT_OS_WINDOWS)
934	if (fTimeout)
935	return; /* only timer update */
936	#endif
937
938	/*
939	* Check sockets
940	*/
941	if (!link_up)
942	goto done;
943	#if defined(RT_OS_WINDOWS)
944	/*XXX: before renaming please make see define
945	* fIcmp in slirp_state.h
946	*/
947	if (fIcmp)
948	sorecvfrom(pData, &pData->icmp_socket);
949	#else
950	if ( (pData->icmp_socket.s != -1)
951	&& CHECK_FD_SET(&pData->icmp_socket, ignored, readfds))
952	sorecvfrom(pData, &pData->icmp_socket);
953	#endif
954	/*
955	* Check TCP sockets
956	*/
957	QSOCKET_FOREACH(so, so_next, tcp)
958	/* { */
959	/* TCP socket can't be cloned */
960	#ifdef VBOX_WITH_NAT_UDP_SOCKET_CLONE
961	Assert((!so->so_cloneOf));
962	#endif
963	Assert(!so->fUnderPolling);
964	so->fUnderPolling = 1;
965	if (slirpVerifyAndFreeSocket(pData, so))
966	CONTINUE(tcp);
967	/*
968	* FD_ISSET is meaningless on these sockets
969	* (and they can crash the program)
970	*/
971	if (so->so_state & SS_NOFDREF \|\| so->s == -1)
972	{
973	so->fUnderPolling = 0;
974	CONTINUE(tcp);
975	}
976
977	POLL_TCP_EVENTS(rc, error, so, &NetworkEvents);
978
979	LOG_NAT_SOCK(so, TCP, &NetworkEvents, readfds, writefds, xfds);
980
981	if (so->so_state & SS_ISFCONNECTING)
982	{
983	int sockerr = 0;
984	#if !defined(RT_OS_WINDOWS)
985	{
986	int revents = 0;
987
988	/*
989	* Failed connect(2) is reported by poll(2) on
990	* different OSes with different combinations of
991	* POLLERR, POLLHUP, and POLLOUT.
992	*/
993	if ( CHECK_FD_SET(so, NetworkEvents, closefds) /* POLLHUP */
994	\|\| CHECK_FD_SET(so, NetworkEvents, rderr)) /* POLLERR */
995	{
996	revents = POLLHUP; /* squash to single "failed" flag */
997	}
998	#if defined(RT_OS_SOLARIS) \|\| defined(RT_OS_NETBSD)
999	/* Solaris and NetBSD report plain POLLOUT even on error */
1000	else if (CHECK_FD_SET(so, NetworkEvents, writefds)) /* POLLOUT */
1001	{
1002	revents = POLLOUT;
1003	}
1004	#endif
1005
1006	if (revents != 0)
1007	{
1008	socklen_t optlen = (socklen_t)sizeof(sockerr);
1009	ret = getsockopt(so->s, SOL_SOCKET, SO_ERROR, &sockerr, &optlen);
1010
1011	if ( RT_UNLIKELY(ret < 0)
1012	\|\| ( (revents & POLLHUP)
1013	&& RT_UNLIKELY(sockerr == 0)))
1014	sockerr = ETIMEDOUT;
1015	}
1016	}
1017	#else /* RT_OS_WINDOWS */
1018	{
1019	if (NetworkEvents.lNetworkEvents & FD_CONNECT)
1020	sockerr = NetworkEvents.iErrorCode[FD_CONNECT_BIT];
1021	}
1022	#endif
1023	if (sockerr != 0)
1024	{
1025	tcp_fconnect_failed(pData, so, sockerr);
1026	ret = slirpVerifyAndFreeSocket(pData, so);
1027	Assert(ret == 1); /* freed */
1028	CONTINUE(tcp);
1029	}
1030
1031	/*
1032	* XXX: For now just fall through to the old code to
1033	* handle successful connect(2).
1034	*/
1035	}
1036
1037	/*
1038	* Check for URG data
1039	* This will soread as well, so no need to
1040	* test for readfds below if this succeeds
1041	*/
1042
1043	/* out-of-band data */
1044	if ( CHECK_FD_SET(so, NetworkEvents, xfds)
1045	#ifdef RT_OS_DARWIN
1046	/* Darwin and probably BSD hosts generates POLLPRI\|POLLHUP event on receiving TCP.flags.{ACK\|URG\|FIN} this
1047	* combination on other Unixs hosts doesn't enter to this branch
1048	*/
1049	&& !CHECK_FD_SET(so, NetworkEvents, closefds)
1050	#endif
1051	#ifdef RT_OS_WINDOWS
1052	/**
1053	* In some cases FD_CLOSE comes with FD_OOB, that confuse tcp processing.
1054	*/
1055	&& !WIN_CHECK_FD_SET(so, NetworkEvents, closefds)
1056	#endif
1057	)
1058	{
1059	sorecvoob(pData, so);
1060	if (slirpVerifyAndFreeSocket(pData, so))
1061	CONTINUE(tcp);
1062	}
1063
1064	/*
1065	* Check sockets for reading
1066	*/
1067	else if ( CHECK_FD_SET(so, NetworkEvents, readfds)
1068	\|\| WIN_CHECK_FD_SET(so, NetworkEvents, acceptds))
1069	{
1070
1071	#ifdef RT_OS_WINDOWS
1072	if (WIN_CHECK_FD_SET(so, NetworkEvents, connectfds))
1073	{
1074	/* Finish connection first */
1075	/* should we ignore return value? */
1076	bool fRet = slirpConnectOrWrite(pData, so, true);
1077	LogFunc(("fRet:%RTbool\n", fRet));
1078	if (slirpVerifyAndFreeSocket(pData, so))
1079	CONTINUE(tcp);
1080	}
1081	#endif
1082	/*
1083	* Check for incoming connections
1084	*/
1085	if (so->so_state & SS_FACCEPTCONN)
1086	{
1087	TCP_CONNECT(pData, so);
1088	if (slirpVerifyAndFreeSocket(pData, so))
1089	CONTINUE(tcp);
1090	if (!CHECK_FD_SET(so, NetworkEvents, closefds))
1091	{
1092	so->fUnderPolling = 0;
1093	CONTINUE(tcp);
1094	}
1095	}
1096
1097	ret = soread(pData, so);
1098	if (slirpVerifyAndFreeSocket(pData, so))
1099	CONTINUE(tcp);
1100	/* Output it if we read something */
1101	if (RT_LIKELY(ret > 0))
1102	TCP_OUTPUT(pData, sototcpcb(so));
1103
1104	if (slirpVerifyAndFreeSocket(pData, so))
1105	CONTINUE(tcp);
1106	}
1107
1108	/*
1109	* Check for FD_CLOSE events.
1110	* in some cases once FD_CLOSE engaged on socket it could be flashed latter (for some reasons)
1111	*/
1112	if ( CHECK_FD_SET(so, NetworkEvents, closefds)
1113	\|\| (so->so_close == 1))
1114	{
1115	/*
1116	* drain the socket
1117	*/
1118	for (; so_next->so_prev == so
1119	&& !slirpVerifyAndFreeSocket(pData, so);)
1120	{
1121	ret = soread(pData, so);
1122	if (slirpVerifyAndFreeSocket(pData, so))
1123	break;
1124
1125	if (ret > 0)
1126	TCP_OUTPUT(pData, sototcpcb(so));
1127	else if (so_next->so_prev == so)
1128	{
1129	Log2(("%R[natsock] errno %d (%s)\n", so, errno, strerror(errno)));
1130	break;
1131	}
1132	}
1133
1134	/* if socket freed ''so'' is PHANTOM and next socket isn't points on it */
1135	if (so_next->so_prev == so)
1136	{
1137	/* mark the socket for termination _after_ it was drained */
1138	so->so_close = 1;
1139	/* No idea about Windows but on Posix, POLLHUP means that we can't send more.
1140	* Actually in the specific error scenario, POLLERR is set as well. */
1141	#ifndef RT_OS_WINDOWS
1142	if (CHECK_FD_SET(so, NetworkEvents, rderr))
1143	sofcantsendmore(so);
1144	#endif
1145	}
1146	if (so_next->so_prev == so)
1147	so->fUnderPolling = 0;
1148	CONTINUE(tcp);
1149	}
1150
1151	/*
1152	* Check sockets for writing
1153	*/
1154	if ( CHECK_FD_SET(so, NetworkEvents, writefds)
1155	#ifdef RT_OS_WINDOWS
1156	\|\| WIN_CHECK_FD_SET(so, NetworkEvents, connectfds)
1157	#endif
1158	)
1159	{
1160	int fConnectOrWriteSuccess = slirpConnectOrWrite(pData, so, false);
1161	/* slirpConnectOrWrite could return true even if tcp_input called tcp_drop,
1162	* so we should be ready to such situations.
1163	*/
1164	if (slirpVerifyAndFreeSocket(pData, so))
1165	CONTINUE(tcp);
1166	else if (!fConnectOrWriteSuccess)
1167	{
1168	so->fUnderPolling = 0;
1169	CONTINUE(tcp);
1170	}
1171	/* slirpConnectionOrWrite succeeded and socket wasn't dropped */
1172	}
1173
1174	/*
1175	* Probe a still-connecting, non-blocking socket
1176	* to check if it's still alive
1177	*/
1178	#ifdef PROBE_CONN
1179	if (so->so_state & SS_ISFCONNECTING)
1180	{
1181	ret = recv(so->s, (char *)&ret, 0, 0);
1182
1183	if (ret < 0)
1184	{
1185	/* XXX */
1186	if ( soIgnorableErrorCode(errno)
1187	\|\| errno == ENOTCONN)
1188	{
1189	CONTINUE(tcp); /* Still connecting, continue */
1190	}
1191
1192	/* else failed */
1193	so->so_state = SS_NOFDREF;
1194
1195	/* tcp_input will take care of it */
1196	}
1197	else
1198	{
1199	ret = send(so->s, &ret, 0, 0);
1200	if (ret < 0)
1201	{
1202	/* XXX */
1203	if ( soIgnorableErrorCode(errno)
1204	\|\| errno == ENOTCONN)
1205	{
1206	CONTINUE(tcp);
1207	}
1208	/* else failed */
1209	so->so_state = SS_NOFDREF;
1210	}
1211	else
1212	so->so_state &= ~SS_ISFCONNECTING;
1213
1214	}
1215	TCP_INPUT((struct mbuf *)NULL, sizeof(struct ip),so);
1216	} /* SS_ISFCONNECTING */
1217	#endif
1218	if (!slirpVerifyAndFreeSocket(pData, so))
1219	so->fUnderPolling = 0;
1220	LOOP_LABEL(tcp, so, so_next);
1221	}
1222
1223	/*
1224	* Now UDP sockets.
1225	* Incoming packets are sent straight away, they're not buffered.
1226	* Incoming UDP data isn't buffered either.
1227	*/
1228	QSOCKET_FOREACH(so, so_next, udp)
1229	/* { */
1230	#ifdef VBOX_WITH_NAT_UDP_SOCKET_CLONE
1231	if (so->so_cloneOf)
1232	CONTINUE_NO_UNLOCK(udp);
1233	#endif
1234	#if 0
1235	so->fUnderPolling = 1;
1236	if(slirpVerifyAndFreeSocket(pData, so));
1237	CONTINUE(udp);
1238	so->fUnderPolling = 0;
1239	#endif
1240
1241	POLL_UDP_EVENTS(rc, error, so, &NetworkEvents);
1242
1243	LOG_NAT_SOCK(so, UDP, &NetworkEvents, readfds, writefds, xfds);
1244
1245	if (so->s != -1 && CHECK_FD_SET(so, NetworkEvents, readfds))
1246	{
1247	SORECVFROM(pData, so);
1248	}
1249	LOOP_LABEL(udp, so, so_next);
1250	}
1251
1252	done:
1253
1254	STAM_PROFILE_STOP(&pData->StatPoll, a);
1255	}
1256
1257
1258	struct arphdr
1259	{
1260	unsigned short ar_hrd; /* format of hardware address */
1261	unsigned short ar_pro; /* format of protocol address */
1262	unsigned char ar_hln; /* length of hardware address */
1263	unsigned char ar_pln; /* length of protocol address */
1264	unsigned short ar_op; /* ARP opcode (command) */
1265
1266	/*
1267	* Ethernet looks like this : This bit is variable sized however...
1268	*/
1269	unsigned char ar_sha[ETH_ALEN]; /* sender hardware address */
1270	unsigned char ar_sip[4]; /* sender IP address */
1271	unsigned char ar_tha[ETH_ALEN]; /* target hardware address */
1272	unsigned char ar_tip[4]; /* target IP address */
1273	};
1274	AssertCompileSize(struct arphdr, 28);
1275
1276	static void arp_output(PNATState pData, const uint8_t pcu8EtherSource, const struct arphdr pcARPHeaderSource, uint32_t ip4TargetAddress)
1277	{
1278	struct ethhdr *pEtherHeaderResponse;
1279	struct arphdr *pARPHeaderResponse;
1280	uint32_t ip4TargetAddressInHostFormat;
1281	struct mbuf *pMbufResponse;
1282
1283	Assert((pcu8EtherSource));
1284	if (!pcu8EtherSource)
1285	return;
1286	ip4TargetAddressInHostFormat = RT_N2H_U32(ip4TargetAddress);
1287
1288	pMbufResponse = m_getcl(pData, M_NOWAIT, MT_HEADER, M_PKTHDR);
1289	if (!pMbufResponse)
1290	return;
1291	pEtherHeaderResponse = mtod(pMbufResponse, struct ethhdr *);
1292	/* @note: if_encap will swap src and dst*/
1293	memcpy(pEtherHeaderResponse->h_source, pcu8EtherSource, ETH_ALEN);
1294	pMbufResponse->m_data += ETH_HLEN;
1295	pARPHeaderResponse = mtod(pMbufResponse, struct arphdr *);
1296	pMbufResponse->m_len = sizeof(struct arphdr);
1297
1298	pARPHeaderResponse->ar_hrd = RT_H2N_U16_C(1);
1299	pARPHeaderResponse->ar_pro = RT_H2N_U16_C(ETH_P_IP);
1300	pARPHeaderResponse->ar_hln = ETH_ALEN;
1301	pARPHeaderResponse->ar_pln = 4;
1302	pARPHeaderResponse->ar_op = RT_H2N_U16_C(ARPOP_REPLY);
1303	memcpy(pARPHeaderResponse->ar_sha, special_ethaddr, ETH_ALEN);
1304
1305	if (!slirpMbufTagService(pData, pMbufResponse, (uint8_t)(ip4TargetAddressInHostFormat & ~pData->netmask)))
1306	{
1307	static bool fTagErrorReported;
1308	if (!fTagErrorReported)
1309	{
1310	LogRel(("NAT: couldn't add the tag(PACKET_SERVICE:%d)\n",
1311	(uint8_t)(ip4TargetAddressInHostFormat & ~pData->netmask)));
1312	fTagErrorReported = true;
1313	}
1314	}
1315	pARPHeaderResponse->ar_sha[5] = (uint8_t)(ip4TargetAddressInHostFormat & ~pData->netmask);
1316
1317	memcpy(pARPHeaderResponse->ar_sip, pcARPHeaderSource->ar_tip, 4);
1318	memcpy(pARPHeaderResponse->ar_tha, pcARPHeaderSource->ar_sha, ETH_ALEN);
1319	memcpy(pARPHeaderResponse->ar_tip, pcARPHeaderSource->ar_sip, 4);
1320	if_encap(pData, ETH_P_ARP, pMbufResponse, ETH_ENCAP_URG);
1321	}
1322	/**
1323	* @note This function will free m!
1324	*/
1325	static void arp_input(PNATState pData, struct mbuf *m)
1326	{
1327	struct ethhdr *pEtherHeader;
1328	struct arphdr *pARPHeader;
1329	uint32_t ip4TargetAddress;
1330
1331	int ar_op;
1332	pEtherHeader = mtod(m, struct ethhdr *);
1333	pARPHeader = (struct arphdr *)&pEtherHeader[1];
1334
1335	ar_op = RT_N2H_U16(pARPHeader->ar_op);
1336	ip4TargetAddress = (uint32_t)pARPHeader->ar_tip;
1337
1338	switch (ar_op)
1339	{
1340	case ARPOP_REQUEST:
1341	if ( CTL_CHECK(ip4TargetAddress, CTL_DNS)
1342	\|\| CTL_CHECK(ip4TargetAddress, CTL_ALIAS)
1343	\|\| CTL_CHECK(ip4TargetAddress, CTL_TFTP))
1344	arp_output(pData, pEtherHeader->h_source, pARPHeader, ip4TargetAddress);
1345
1346	/* Gratuitous ARP */
1347	if ( (uint32_t )pARPHeader->ar_sip == (uint32_t )pARPHeader->ar_tip
1348	&& memcmp(pARPHeader->ar_tha, broadcast_ethaddr, ETH_ALEN) == 0
1349	&& memcmp(pEtherHeader->h_dest, broadcast_ethaddr, ETH_ALEN) == 0)
1350	{
1351	/* We've received an announce about address assignment,
1352	* let's do an ARP cache update
1353	*/
1354	static bool fGratuitousArpReported;
1355	if (!fGratuitousArpReported)
1356	{
1357	LogRel(("NAT: Gratuitous ARP [IP:%RTnaipv4, ether:%RTmac]\n",
1358	pARPHeader->ar_sip, pARPHeader->ar_sha));
1359	fGratuitousArpReported = true;
1360	}
1361	slirp_arp_cache_update_or_add(pData, (uint32_t )pARPHeader->ar_sip, &pARPHeader->ar_sha[0]);
1362	}
1363	break;
1364
1365	case ARPOP_REPLY:
1366	slirp_arp_cache_update_or_add(pData, (uint32_t )pARPHeader->ar_sip, &pARPHeader->ar_sha[0]);
1367	break;
1368
1369	default:
1370	break;
1371	}
1372
1373	m_freem(pData, m);
1374	}
1375
1376	/**
1377	* Feed a packet into the slirp engine.
1378	*
1379	* @param m Data buffer, m_len is not valid.
1380	* @param cbBuf The length of the data in m.
1381	*/
1382	void slirp_input(PNATState pData, struct mbuf *m, size_t cbBuf)
1383	{
1384	int proto;
1385	static bool fWarnedIpv6;
1386	struct ethhdr *eh;
1387	uint8_t au8Ether[ETH_ALEN];
1388
1389	m->m_len = cbBuf;
1390	if (cbBuf < ETH_HLEN)
1391	{
1392	Log(("NAT: packet having size %d has been ignored\n", m->m_len));
1393	m_freem(pData, m);
1394	return;
1395	}
1396	eh = mtod(m, struct ethhdr *);
1397	proto = RT_N2H_U16(eh->h_proto);
1398
1399	memcpy(au8Ether, eh->h_source, ETH_ALEN);
1400
1401	switch(proto)
1402	{
1403	case ETH_P_ARP:
1404	arp_input(pData, m);
1405	break;
1406
1407	case ETH_P_IP:
1408	/* Update time. Important if the network is very quiet, as otherwise
1409	* the first outgoing connection gets an incorrect timestamp. */
1410	updtime(pData);
1411	m_adj(m, ETH_HLEN);
1412	M_ASSERTPKTHDR(m);
1413	m->m_pkthdr.header = mtod(m, void *);
1414	ip_input(pData, m);
1415	break;
1416
1417	case ETH_P_IPV6:
1418	m_freem(pData, m);
1419	if (!fWarnedIpv6)
1420	{
1421	LogRel(("NAT: IPv6 not supported\n"));
1422	fWarnedIpv6 = true;
1423	}
1424	break;
1425
1426	default:
1427	Log(("NAT: Unsupported protocol %x\n", proto));
1428	m_freem(pData, m);
1429	break;
1430	}
1431
1432	if (pData->cRedirectionsActive != pData->cRedirectionsStored)
1433	activate_port_forwarding(pData, au8Ether);
1434	}
1435
1436	/**
1437	* Output the IP packet to the ethernet device.
1438	*
1439	* @note This function will free m!
1440	*/
1441	void if_encap(PNATState pData, uint16_t eth_proto, struct mbuf *m, int flags)
1442	{
1443	struct ethhdr *eh;
1444	uint8_t *mbuf = NULL;
1445	size_t mlen = 0;
1446	STAM_PROFILE_START(&pData->StatIF_encap, a);
1447	LogFlowFunc(("ENTER: pData:%p, eth_proto:%RX16, m:%p, flags:%d\n",
1448	pData, eth_proto, m, flags));
1449
1450	M_ASSERTPKTHDR(m);
1451
1452	Assert(M_LEADINGSPACE(m) >= ETH_HLEN);
1453	m->m_data -= ETH_HLEN;
1454	m->m_len += ETH_HLEN;
1455	eh = mtod(m, struct ethhdr *);
1456	mlen = m->m_len;
1457
1458	if (memcmp(eh->h_source, special_ethaddr, ETH_ALEN) != 0)
1459	{
1460	struct m_tag *t = m_tag_first(m);
1461	uint8_t u8ServiceId = CTL_ALIAS;
1462	memcpy(eh->h_dest, eh->h_source, ETH_ALEN);
1463	memcpy(eh->h_source, special_ethaddr, ETH_ALEN);
1464	Assert(memcmp(eh->h_dest, special_ethaddr, ETH_ALEN) != 0);
1465	if (memcmp(eh->h_dest, zerro_ethaddr, ETH_ALEN) == 0)
1466	{
1467	/* don't do anything */
1468	m_freem(pData, m);
1469	goto done;
1470	}
1471	if ( t
1472	&& (t = m_tag_find(m, PACKET_SERVICE, NULL)))
1473	{
1474	Assert(t);
1475	u8ServiceId = (uint8_t )&t[1];
1476	}
1477	eh->h_source[5] = u8ServiceId;
1478	}
1479	/*
1480	* we're processing the chain, that isn't not expected.
1481	*/
1482	Assert((!m->m_next));
1483	if (m->m_next)
1484	{
1485	Log(("NAT: if_encap's recived the chain, dropping...\n"));
1486	m_freem(pData, m);
1487	goto done;
1488	}
1489	mbuf = mtod(m, uint8_t *);
1490	eh->h_proto = RT_H2N_U16(eth_proto);
1491	LogFunc(("eh(dst:%RTmac, src:%RTmac)\n", eh->h_dest, eh->h_source));
1492	if (flags & ETH_ENCAP_URG)
1493	slirp_urg_output(pData->pvUser, m, mbuf, mlen);
1494	else
1495	slirp_output(pData->pvUser, m, mbuf, mlen);
1496	done:
1497	STAM_PROFILE_STOP(&pData->StatIF_encap, a);
1498	LogFlowFuncLeave();
1499	}
1500
1501	/**
1502	* Still we're using dhcp server leasing to map ether to IP
1503	* @todo see rt_lookup_in_cache
1504	*/
1505	static uint32_t find_guest_ip(PNATState pData, const uint8_t *eth_addr)
1506	{
1507	uint32_t ip = INADDR_ANY;
1508	int rc;
1509
1510	if (eth_addr == NULL)
1511	return INADDR_ANY;
1512
1513	if ( memcmp(eth_addr, zerro_ethaddr, ETH_ALEN) == 0
1514	\|\| memcmp(eth_addr, broadcast_ethaddr, ETH_ALEN) == 0)
1515	return INADDR_ANY;
1516
1517	rc = slirp_arp_lookup_ip_by_ether(pData, eth_addr, &ip);
1518	if (RT_SUCCESS(rc))
1519	return ip;
1520
1521	bootp_cache_lookup_ip_by_ether(pData, eth_addr, &ip);
1522	/* ignore return code, ip will be set to INADDR_ANY on error */
1523	return ip;
1524	}
1525
1526	/**
1527	* We need check if we've activated port forwarding
1528	* for specific machine ... that of course relates to
1529	* service mode
1530	* @todo finish this for service case
1531	*/
1532	static void activate_port_forwarding(PNATState pData, const uint8_t *h_source)
1533	{
1534	struct port_forward_rule rule, tmp;
1535	const uint8_t *pu8EthSource = h_source;
1536
1537	/* check mac here */
1538	LIST_FOREACH_SAFE(rule, &pData->port_forward_rule_head, list, tmp)
1539	{
1540	struct socket *so;
1541	struct sockaddr sa;
1542	struct sockaddr_in *psin;
1543	socklen_t socketlen;
1544	int rc;
1545	uint32_t guest_addr; /* need to understand if we already give address to guest */
1546
1547	if (rule->activated)
1548	continue;
1549
1550	#ifdef VBOX_WITH_NAT_SERVICE
1551	/**
1552	* case when guest ip is INADDR_ANY shouldn't appear in NAT service
1553	*/
1554	Assert((rule->guest_addr.s_addr != INADDR_ANY));
1555	guest_addr = rule->guest_addr.s_addr;
1556	#else /* VBOX_WITH_NAT_SERVICE */
1557	guest_addr = find_guest_ip(pData, pu8EthSource);
1558	#endif /* !VBOX_WITH_NAT_SERVICE */
1559	if (guest_addr == INADDR_ANY)
1560	{
1561	/* the address wasn't granted */
1562	return;
1563	}
1564
1565	#if !defined(VBOX_WITH_NAT_SERVICE)
1566	if ( rule->guest_addr.s_addr != guest_addr
1567	&& rule->guest_addr.s_addr != INADDR_ANY)
1568	continue;
1569	if (rule->guest_addr.s_addr == INADDR_ANY)
1570	rule->guest_addr.s_addr = guest_addr;
1571	#endif
1572
1573	LogRel(("NAT: set redirect %s host %RTnaipv4:%d => guest %RTnaipv4:%d\n",
1574	rule->proto == IPPROTO_UDP ? "UDP" : "TCP",
1575	rule->bind_ip.s_addr, rule->host_port,
1576	guest_addr, rule->guest_port));
1577
1578	if (rule->proto == IPPROTO_UDP)
1579	so = udp_listen(pData, rule->bind_ip.s_addr, RT_H2N_U16(rule->host_port), guest_addr,
1580	RT_H2N_U16(rule->guest_port), 0);
1581	else
1582	so = solisten(pData, rule->bind_ip.s_addr, RT_H2N_U16(rule->host_port), guest_addr,
1583	RT_H2N_U16(rule->guest_port), 0);
1584
1585	if (so == NULL)
1586	goto remove_port_forwarding;
1587
1588	psin = (struct sockaddr_in *)&sa;
1589	psin->sin_family = AF_INET;
1590	psin->sin_port = 0;
1591	psin->sin_addr.s_addr = INADDR_ANY;
1592	socketlen = sizeof(struct sockaddr);
1593
1594	rc = getsockname(so->s, &sa, &socketlen);
1595	if (rc < 0 \|\| sa.sa_family != AF_INET)
1596	goto remove_port_forwarding;
1597
1598	rule->activated = 1;
1599	rule->so = so;
1600	pData->cRedirectionsActive++;
1601	continue;
1602
1603	remove_port_forwarding:
1604	LogRel(("NAT: failed to redirect %s %RTnaipv4:%d => %RTnaipv4:%d\n",
1605	(rule->proto == IPPROTO_UDP ? "UDP" : "TCP"),
1606	rule->bind_ip.s_addr, rule->host_port,
1607	guest_addr, rule->guest_port));
1608	LIST_REMOVE(rule, list);
1609	pData->cRedirectionsStored--;
1610	RTMemFree(rule);
1611	}
1612	}
1613
1614	/**
1615	* Changes in 3.1 instead of opening new socket do the following:
1616	* gain more information:
1617	* 1. bind IP
1618	* 2. host port
1619	* 3. guest port
1620	* 4. proto
1621	* 5. guest MAC address
1622	* the guest's MAC address is rather important for service, but we easily
1623	* could get it from VM configuration in DrvNAT or Service, the idea is activating
1624	* corresponding port-forwarding
1625	*/
1626	int slirp_add_redirect(PNATState pData, int is_udp, struct in_addr host_addr, int host_port,
1627	struct in_addr guest_addr, int guest_port, const uint8_t *ethaddr)
1628	{
1629	struct port_forward_rule *rule = NULL;
1630	LIST_FOREACH(rule, &pData->port_forward_rule_head, list)
1631	{
1632	if ( rule->proto == (is_udp ? IPPROTO_UDP : IPPROTO_TCP)
1633	&& rule->host_port == host_port
1634	&& rule->bind_ip.s_addr == host_addr.s_addr
1635	&& rule->guest_port == guest_port
1636	&& rule->guest_addr.s_addr == guest_addr.s_addr
1637	)
1638	return 0; /* rule has been already registered */
1639	}
1640
1641	rule = RTMemAllocZ(sizeof(struct port_forward_rule));
1642	if (rule == NULL)
1643	return 1;
1644
1645	rule->proto = (is_udp ? IPPROTO_UDP : IPPROTO_TCP);
1646	rule->host_port = host_port;
1647	rule->guest_port = guest_port;
1648	rule->guest_addr.s_addr = guest_addr.s_addr;
1649	rule->bind_ip.s_addr = host_addr.s_addr;
1650	if (ethaddr != NULL)
1651	memcpy(rule->mac_address, ethaddr, ETH_ALEN);
1652	/* @todo add mac address */
1653	LIST_INSERT_HEAD(&pData->port_forward_rule_head, rule, list);
1654	pData->cRedirectionsStored++;
1655	/* activate port-forwarding if guest has already got assigned IP */
1656	if ( ethaddr
1657	&& memcmp(ethaddr, zerro_ethaddr, ETH_ALEN))
1658	activate_port_forwarding(pData, ethaddr);
1659	return 0;
1660	}
1661
1662	int slirp_remove_redirect(PNATState pData, int is_udp, struct in_addr host_addr, int host_port,
1663	struct in_addr guest_addr, int guest_port)
1664	{
1665	struct port_forward_rule *rule = NULL;
1666	LIST_FOREACH(rule, &pData->port_forward_rule_head, list)
1667	{
1668	if ( rule->proto == (is_udp ? IPPROTO_UDP : IPPROTO_TCP)
1669	&& rule->host_port == host_port
1670	&& rule->guest_port == guest_port
1671	&& rule->bind_ip.s_addr == host_addr.s_addr
1672	&& rule->guest_addr.s_addr == guest_addr.s_addr
1673	&& rule->activated)
1674	{
1675	LogRel(("NAT: remove redirect %s host %RTnaipv4:%d => guest %RTnaipv4:%d\n",
1676	rule->proto == IPPROTO_UDP ? "UDP" : "TCP",
1677	rule->bind_ip.s_addr, rule->host_port,
1678	guest_addr.s_addr, rule->guest_port));
1679
1680	if (is_udp)
1681	udp_detach(pData, rule->so);
1682	else
1683	tcp_close(pData, sototcpcb(rule->so));
1684	LIST_REMOVE(rule, list);
1685	RTMemFree(rule);
1686	pData->cRedirectionsStored--;
1687	break;
1688	}
1689
1690	}
1691	return 0;
1692	}
1693
1694	void slirp_set_ethaddr_and_activate_port_forwarding(PNATState pData, const uint8_t *ethaddr, uint32_t GuestIP)
1695	{
1696	#ifndef VBOX_WITH_NAT_SERVICE
1697	memcpy(client_ethaddr, ethaddr, ETH_ALEN);
1698	#endif
1699	if (GuestIP != INADDR_ANY)
1700	{
1701	slirp_arp_cache_update_or_add(pData, GuestIP, ethaddr);
1702	activate_port_forwarding(pData, ethaddr);
1703	}
1704	}
1705
1706	#if defined(RT_OS_WINDOWS)
1707	HANDLE *slirp_get_events(PNATState pData)
1708	{
1709	return pData->phEvents;
1710	}
1711	void slirp_register_external_event(PNATState pData, HANDLE hEvent, int index)
1712	{
1713	pData->phEvents[index] = hEvent;
1714	}
1715	#endif
1716
1717	unsigned int slirp_get_timeout_ms(PNATState pData)
1718	{
1719	if (link_up)
1720	{
1721	if (time_fasttimo)
1722	return 2;
1723	if (do_slowtimo)
1724	return 500; /* see PR_SLOWHZ */
1725	}
1726	return 36001000; / one hour */
1727	}
1728
1729	#ifndef RT_OS_WINDOWS
1730	int slirp_get_nsock(PNATState pData)
1731	{
1732	return pData->nsock;
1733	}
1734	#endif
1735
1736	/*
1737	* this function called from NAT thread
1738	*/
1739	void slirp_post_sent(PNATState pData, void *pvArg)
1740	{
1741	struct mbuf m = (struct mbuf )pvArg;
1742	m_freem(pData, m);
1743	}
1744
1745	void slirp_set_dhcp_TFTP_prefix(PNATState pData, const char *tftpPrefix)
1746	{
1747	Log2(("tftp_prefix: %s\n", tftpPrefix));
1748	tftp_prefix = tftpPrefix;
1749	}
1750
1751	void slirp_set_dhcp_TFTP_bootfile(PNATState pData, const char *bootFile)
1752	{
1753	Log2(("bootFile: %s\n", bootFile));
1754	bootp_filename = bootFile;
1755	}
1756
1757	void slirp_set_dhcp_next_server(PNATState pData, const char *next_server)
1758	{
1759	Log2(("next_server: %s\n", next_server));
1760	if (next_server == NULL)
1761	pData->tftp_server.s_addr = RT_H2N_U32(RT_N2H_U32(pData->special_addr.s_addr) \| CTL_TFTP);
1762	else
1763	inet_aton(next_server, &pData->tftp_server);
1764	}
1765
1766	int slirp_set_binding_address(PNATState pData, char *addr)
1767	{
1768	if (addr == NULL \|\| (inet_aton(addr, &pData->bindIP) == 0))
1769	{
1770	pData->bindIP.s_addr = INADDR_ANY;
1771	return 1;
1772	}
1773	return 0;
1774	}
1775
1776	void slirp_set_dhcp_dns_proxy(PNATState pData, bool fDNSProxy)
1777	{
1778	if (!pData->fUseHostResolver)
1779	{
1780	Log2(("NAT: DNS proxy switched %s\n", (fDNSProxy ? "on" : "off")));
1781	pData->fUseDnsProxy = fDNSProxy;
1782	}
1783	else if (fDNSProxy)
1784	LogRel(("NAT: Host Resolver conflicts with DNS proxy, the last one was forcely ignored\n"));
1785	}
1786
1787	#define CHECK_ARG(name, val, lim_min, lim_max) \
1788	do { \
1789	if ((val) < (lim_min) \|\| (val) > (lim_max)) \
1790	{ \
1791	LogRel(("NAT: (" #name ":%d) has been ignored, " \
1792	"because out of range (%d, %d)\n", (val), (lim_min), (lim_max))); \
1793	return; \
1794	} \
1795	else \
1796	LogRel(("NAT: (" #name ":%d)\n", (val))); \
1797	} while (0)
1798
1799	void slirp_set_somaxconn(PNATState pData, int iSoMaxConn)
1800	{
1801	LogFlowFunc(("iSoMaxConn:d\n", iSoMaxConn));
1802	/* Conditions */
1803	if (iSoMaxConn > SOMAXCONN)
1804	{
1805	LogRel(("NAT: value of somaxconn(%d) bigger than SOMAXCONN(%d)\n", iSoMaxConn, SOMAXCONN));
1806	iSoMaxConn = SOMAXCONN;
1807	}
1808
1809	if (iSoMaxConn < 1)
1810	{
1811	LogRel(("NAT: proposed value(%d) of somaxconn is invalid, default value is used (%d)\n", iSoMaxConn, pData->soMaxConn));
1812	LogFlowFuncLeave();
1813	return;
1814	}
1815
1816	/* Asignment */
1817	if (pData->soMaxConn != iSoMaxConn)
1818	{
1819	LogRel(("NAT: value of somaxconn has been changed from %d to %d\n",
1820	pData->soMaxConn, iSoMaxConn));
1821	pData->soMaxConn = iSoMaxConn;
1822	}
1823	LogFlowFuncLeave();
1824	}
1825	/* don't allow user set less 8kB and more than 1M values */
1826	#define _8K_1M_CHECK_ARG(name, val) CHECK_ARG(name, (val), 8, 1024)
1827	void slirp_set_rcvbuf(PNATState pData, int kilobytes)
1828	{
1829	_8K_1M_CHECK_ARG("SOCKET_RCVBUF", kilobytes);
1830	pData->socket_rcv = kilobytes;
1831	}
1832	void slirp_set_sndbuf(PNATState pData, int kilobytes)
1833	{
1834	_8K_1M_CHECK_ARG("SOCKET_SNDBUF", kilobytes);
1835	pData->socket_snd = kilobytes * _1K;
1836	}
1837	void slirp_set_tcp_rcvspace(PNATState pData, int kilobytes)
1838	{
1839	_8K_1M_CHECK_ARG("TCP_RCVSPACE", kilobytes);
1840	tcp_rcvspace = kilobytes * _1K;
1841	}
1842	void slirp_set_tcp_sndspace(PNATState pData, int kilobytes)
1843	{
1844	_8K_1M_CHECK_ARG("TCP_SNDSPACE", kilobytes);
1845	tcp_sndspace = kilobytes * _1K;
1846	}
1847
1848	/*
1849	* Looking for Ether by ip in ARP-cache
1850	* Note: it´s responsible of caller to allocate buffer for result
1851	* @returns iprt status code
1852	*/
1853	int slirp_arp_lookup_ether_by_ip(PNATState pData, uint32_t ip, uint8_t *ether)
1854	{
1855	struct arp_cache_entry *ac;
1856
1857	if (ether == NULL)
1858	return VERR_INVALID_PARAMETER;
1859
1860	if (LIST_EMPTY(&pData->arp_cache))
1861	return VERR_NOT_FOUND;
1862
1863	LIST_FOREACH(ac, &pData->arp_cache, list)
1864	{
1865	if ( ac->ip == ip
1866	&& memcmp(ac->ether, broadcast_ethaddr, ETH_ALEN) != 0)
1867	{
1868	memcpy(ether, ac->ether, ETH_ALEN);
1869	return VINF_SUCCESS;
1870	}
1871	}
1872	return VERR_NOT_FOUND;
1873	}
1874
1875	/*
1876	* Looking for IP by Ether in ARP-cache
1877	* Note: it´s responsible of caller to allocate buffer for result
1878	* @returns 0 - if found, 1 - otherwise
1879	*/
1880	int slirp_arp_lookup_ip_by_ether(PNATState pData, const uint8_t ether, uint32_t ip)
1881	{
1882	struct arp_cache_entry *ac;
1883	*ip = INADDR_ANY;
1884
1885	if (LIST_EMPTY(&pData->arp_cache))
1886	return VERR_NOT_FOUND;
1887
1888	LIST_FOREACH(ac, &pData->arp_cache, list)
1889	{
1890	if (memcmp(ether, ac->ether, ETH_ALEN) == 0)
1891	{
1892	*ip = ac->ip;
1893	return VINF_SUCCESS;
1894	}
1895	}
1896	return VERR_NOT_FOUND;
1897	}
1898
1899	void slirp_arp_who_has(PNATState pData, uint32_t dst)
1900	{
1901	struct mbuf *m;
1902	struct ethhdr *ehdr;
1903	struct arphdr *ahdr;
1904	static bool fWarned = false;
1905	LogFlowFunc(("ENTER: %RTnaipv4\n", dst));
1906
1907	/* ARP request WHO HAS 0.0.0.0 is one of the signals
1908	* that something has been broken at Slirp. Investigating
1909	* pcap dumps it's easy to miss warning ARP requests being
1910	* focused on investigation of other protocols flow.
1911	*/
1912	#ifdef DEBUG_vvl
1913	Assert((dst != INADDR_ANY));
1914	NOREF(fWarned);
1915	#else
1916	if ( dst == INADDR_ANY
1917	&& !fWarned)
1918	{
1919	LogRel(("NAT:ARP: \"WHO HAS INADDR_ANY\" request has been detected\n"));
1920	fWarned = true;
1921	}
1922	#endif /* !DEBUG_vvl */
1923
1924	m = m_getcl(pData, M_NOWAIT, MT_HEADER, M_PKTHDR);
1925	if (m == NULL)
1926	{
1927	Log(("NAT: Can't alloc mbuf for ARP request\n"));
1928	LogFlowFuncLeave();
1929	return;
1930	}
1931	ehdr = mtod(m, struct ethhdr *);
1932	memset(ehdr->h_source, 0xff, ETH_ALEN);
1933	ahdr = (struct arphdr *)&ehdr[1];
1934	ahdr->ar_hrd = RT_H2N_U16_C(1);
1935	ahdr->ar_pro = RT_H2N_U16_C(ETH_P_IP);
1936	ahdr->ar_hln = ETH_ALEN;
1937	ahdr->ar_pln = 4;
1938	ahdr->ar_op = RT_H2N_U16_C(ARPOP_REQUEST);
1939	memcpy(ahdr->ar_sha, special_ethaddr, ETH_ALEN);
1940	/* we assume that this request come from gw, but not from DNS or TFTP */
1941	ahdr->ar_sha[5] = CTL_ALIAS;
1942	(uint32_t )ahdr->ar_sip = RT_H2N_U32(RT_N2H_U32(pData->special_addr.s_addr) \| CTL_ALIAS);
1943	memset(ahdr->ar_tha, 0xff, ETH_ALEN); /broadcast/
1944	(uint32_t )ahdr->ar_tip = dst;
1945	/* warn!!! should falls in mbuf minimal size */
1946	m->m_len = sizeof(struct arphdr) + ETH_HLEN;
1947	m->m_data += ETH_HLEN;
1948	m->m_len -= ETH_HLEN;
1949	if_encap(pData, ETH_P_ARP, m, ETH_ENCAP_URG);
1950	LogFlowFuncLeave();
1951	}
1952	#ifdef VBOX_WITH_DNSMAPPING_IN_HOSTRESOLVER
1953	void slirp_add_host_resolver_mapping(PNATState pData, const char pszHostName, const char pszHostNamePattern, uint32_t u32HostIP)
1954	{
1955	LogFlowFunc(("ENTER: pszHostName:%s, pszHostNamePattern:%s u32HostIP:%RTnaipv4\n",
1956	pszHostName ? pszHostName : "(null)",
1957	pszHostNamePattern ? pszHostNamePattern : "(null)",
1958	u32HostIP));
1959	if ( ( pszHostName
1960	\|\| pszHostNamePattern)
1961	&& u32HostIP != INADDR_ANY
1962	&& u32HostIP != INADDR_BROADCAST)
1963	{
1964	PDNSMAPPINGENTRY pDnsMapping = RTMemAllocZ(sizeof(DNSMAPPINGENTRY));
1965	if (!pDnsMapping)
1966	{
1967	LogFunc(("Can't allocate DNSMAPPINGENTRY\n"));
1968	LogFlowFuncLeave();
1969	return;
1970	}
1971	pDnsMapping->u32IpAddress = u32HostIP;
1972	if (pszHostName)
1973	pDnsMapping->pszCName = RTStrDup(pszHostName);
1974	else if (pszHostNamePattern)
1975	pDnsMapping->pszPattern = RTStrDup(pszHostNamePattern);
1976	if ( !pDnsMapping->pszCName
1977	&& !pDnsMapping->pszPattern)
1978	{
1979	LogFunc(("Can't allocate enough room for %s\n", pszHostName ? pszHostName : pszHostNamePattern));
1980	RTMemFree(pDnsMapping);
1981	LogFlowFuncLeave();
1982	return;
1983	}
1984	LIST_INSERT_HEAD(&pData->DNSMapHead, pDnsMapping, MapList);
1985	LogRel(("NAT: user-defined mapping %s: %RTnaipv4 is registered\n",
1986	pDnsMapping->pszCName ? pDnsMapping->pszCName : pDnsMapping->pszPattern,
1987	pDnsMapping->u32IpAddress));
1988	}
1989	LogFlowFuncLeave();
1990	}
1991	#endif
1992
1993	/* updates the arp cache
1994	* @note: this is helper function, slirp_arp_cache_update_or_add should be used.
1995	* @returns 0 - if has found and updated
1996	* 1 - if hasn't found.
1997	*/
1998	static inline int slirp_arp_cache_update(PNATState pData, uint32_t dst, const uint8_t *mac)
1999	{
2000	struct arp_cache_entry *ac;
2001	Assert(( memcmp(mac, broadcast_ethaddr, ETH_ALEN)
2002	&& memcmp(mac, zerro_ethaddr, ETH_ALEN)));
2003	LIST_FOREACH(ac, &pData->arp_cache, list)
2004	{
2005	if (ac->ip == dst)
2006	{
2007	memcpy(ac->ether, mac, ETH_ALEN);
2008	return 0;
2009	}
2010	}
2011	return 1;
2012	}
2013
2014	/**
2015	* add entry to the arp cache
2016	* @note: this is helper function, slirp_arp_cache_update_or_add should be used.
2017	*/
2018	static inline void slirp_arp_cache_add(PNATState pData, uint32_t ip, const uint8_t *ether)
2019	{
2020	struct arp_cache_entry *ac = NULL;
2021	Assert(( memcmp(ether, broadcast_ethaddr, ETH_ALEN)
2022	&& memcmp(ether, zerro_ethaddr, ETH_ALEN)));
2023	ac = RTMemAllocZ(sizeof(struct arp_cache_entry));
2024	if (ac == NULL)
2025	{
2026	Log(("NAT: Can't allocate arp cache entry\n"));
2027	return;
2028	}
2029	ac->ip = ip;
2030	memcpy(ac->ether, ether, ETH_ALEN);
2031	LIST_INSERT_HEAD(&pData->arp_cache, ac, list);
2032	}
2033
2034	/* updates or adds entry to the arp cache
2035	* @returns 0 - if has found and updated
2036	* 1 - if hasn't found.
2037	*/
2038	int slirp_arp_cache_update_or_add(PNATState pData, uint32_t dst, const uint8_t *mac)
2039	{
2040	if ( !memcmp(mac, broadcast_ethaddr, ETH_ALEN)
2041	\|\| !memcmp(mac, zerro_ethaddr, ETH_ALEN))
2042	{
2043	static bool fBroadcastEtherAddReported;
2044	if (!fBroadcastEtherAddReported)
2045	{
2046	LogRel(("NAT: Attempt to add pair [%RTmac:%RTnaipv4] in ARP cache was ignored\n",
2047	mac, dst));
2048	fBroadcastEtherAddReported = true;
2049	}
2050	return 1;
2051	}
2052	if (slirp_arp_cache_update(pData, dst, mac))
2053	slirp_arp_cache_add(pData, dst, mac);
2054
2055	return 0;
2056	}
2057
2058
2059	void slirp_set_mtu(PNATState pData, int mtu)
2060	{
2061	if (mtu < 20 \|\| mtu >= 16000)
2062	{
2063	LogRel(("NAT: mtu(%d) is out of range (20;16000] mtu forcely assigned to 1500\n", mtu));
2064	mtu = 1500;
2065	}
2066	/* MTU is maximum transition unit on */
2067	if_mtu =
2068	if_mru = mtu;
2069	}
2070
2071	/**
2072	* Info handler.
2073	*/
2074	void slirp_info(PNATState pData, const void pvArg, const char pszArgs)
2075	{
2076	struct socket so, so_next;
2077	struct arp_cache_entry *ac;
2078	struct port_forward_rule *rule;
2079	PCDBGFINFOHLP pHlp = (PCDBGFINFOHLP)pvArg;
2080	NOREF(pszArgs);
2081
2082	pHlp->pfnPrintf(pHlp, "NAT parameters: MTU=%d\n", if_mtu);
2083	pHlp->pfnPrintf(pHlp, "NAT TCP ports:\n");
2084	QSOCKET_FOREACH(so, so_next, tcp)
2085	/* { */
2086	pHlp->pfnPrintf(pHlp, " %R[natsock]\n", so);
2087	}
2088
2089	pHlp->pfnPrintf(pHlp, "NAT UDP ports:\n");
2090	QSOCKET_FOREACH(so, so_next, udp)
2091	/* { */
2092	pHlp->pfnPrintf(pHlp, " %R[natsock]\n", so);
2093	}
2094
2095	pHlp->pfnPrintf(pHlp, "NAT ARP cache:\n");
2096	LIST_FOREACH(ac, &pData->arp_cache, list)
2097	{
2098	pHlp->pfnPrintf(pHlp, " %RTnaipv4 %RTmac\n", ac->ip, &ac->ether);
2099	}
2100
2101	pHlp->pfnPrintf(pHlp, "NAT rules:\n");
2102	LIST_FOREACH(rule, &pData->port_forward_rule_head, list)
2103	{
2104	pHlp->pfnPrintf(pHlp, " %s %d => %RTnaipv4:%d %c\n",
2105	rule->proto == IPPROTO_UDP ? "UDP" : "TCP",
2106	rule->host_port, rule->guest_addr.s_addr, rule->guest_port,
2107	rule->activated ? ' ' : '*');
2108	}
2109	}
2110
2111	/**
2112	* @note: NATState::fUseHostResolver could be changed in bootp.c::dhcp_decode
2113	* @note: this function is executed on GUI/VirtualBox or main/VBoxHeadless thread.
2114	* @note: this function can potentially race with bootp.c::dhcp_decode (except Darwin)
2115	*/
2116	int slirp_host_network_configuration_change_strategy_selector(const PNATState pData)
2117	{
2118	if (pData->fUseHostResolverPermanent)
2119	return VBOX_NAT_DNS_HOSTRESOLVER;
2120
2121	if (pData->fUseDnsProxy) {
2122	#if HAVE_NOTIFICATION_FOR_DNS_UPDATE
2123	/* We dont conflict with bootp.c::dhcp_decode */
2124	struct rcp_state rcp_state;
2125	int rc;
2126
2127	rcp_state.rcps_flags \|= RCPSF_IGNORE_IPV6;
2128	rc = rcp_parse(&rcp_state, RESOLV_CONF_FILE);
2129	LogRelFunc(("NAT: rcp_parse:%Rrc old domain:%s new domain:%s\n",
2130	rc, LIST_FIRST(&pData->pDomainList)->dd_pszDomain,
2131	rcp_state.rcps_domain));
2132	if ( RT_FAILURE(rc)
2133	\|\| LIST_EMPTY(&pData->pDomainList))
2134	return VBOX_NAT_DNS_DNSPROXY;
2135
2136	if ( rcp_state.rcps_domain
2137	&& strcmp(rcp_state.rcps_domain, LIST_FIRST(&pData->pDomainList)->dd_pszDomain) == 0)
2138	return VBOX_NAT_DNS_DNSPROXY;
2139	else
2140	return VBOX_NAT_DNS_EXTERNAL;
2141	#else
2142	/* copy domain name */
2143	/* domain only compare with coy version */
2144	return VBOX_NAT_DNS_DNSPROXY;
2145	#endif
2146	}
2147	return VBOX_NAT_DNS_EXTERNAL;
2148	}

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

source: vbox/trunk/src/VBox/Devices/Network/slirp/slirp.c@ 53350

Download in other formats: