slirp.c@ 57550

Last change on this file since 57550 was 57550, checked in by vboxsync, 9 years ago
NAT: improved error message
Property svn:eol-style set to `native` Property svn:keywords set to `Author Date Id Revision`
File size: 67.4 KB

Line
1	/* $Id: slirp.c 57550 2015-08-26 14:19:11Z vboxsync $ */
2	/** @file
3	* NAT - slirp glue.
4	*/
5
6	/*
7	* Copyright (C) 2006-2015 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)
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	{
308	/* CTL is x.x.x.15, bootp passes up to 16 IPs (15..31) */
309	LogRel(("The last 5 bits of the netmask (%RTnaipv4) need to be unset\n", RT_BE2H_U32(u32Netmask)));
310	return VERR_INVALID_PARAMETER;
311	}
312	pData = RTMemAllocZ(RT_ALIGN_Z(sizeof(NATState), sizeof(uint64_t)));
313	*ppData = pData;
314	if (!pData)
315	return VERR_NO_MEMORY;
316	pData->fPassDomain = !fUseHostResolver ? fPassDomain : false;
317	pData->fUseHostResolver = fUseHostResolver;
318	pData->fUseHostResolverPermanent = fUseHostResolver;
319	pData->pvUser = pvUser;
320	pData->netmask = u32Netmask;
321
322	rc = RTCritSectRwInit(&pData->CsRwHandlerChain);
323	if (RT_FAILURE(rc))
324	return rc;
325
326	/* sockets & TCP defaults */
327	pData->socket_rcv = 64 * _1K;
328	pData->socket_snd = 64 * _1K;
329	tcp_sndspace = 64 * _1K;
330	tcp_rcvspace = 64 * _1K;
331
332	/*
333	* Use the same default here as in DevNAT.cpp (SoMaxConnection CFGM value)
334	* to avoid release log noise.
335	*/
336	pData->soMaxConn = 10;
337
338	#ifdef RT_OS_WINDOWS
339	{
340	WSADATA Data;
341	RTLDRMOD hLdrMod;
342
343	WSAStartup(MAKEWORD(2, 0), &Data);
344
345	rc = RTLdrLoadSystem("Iphlpapi.dll", /* :fNoUnload */ true, &hLdrMod);
346	if (RT_SUCCESS(rc))
347	{
348	rc = RTLdrGetSymbol(hLdrMod, "GetAdaptersAddresses", (void **)&pData->pfGetAdaptersAddresses);
349	if (RT_FAILURE(rc))
350	LogRel(("NAT: Can't find GetAdapterAddresses in Iphlpapi.dll\n"));
351
352	RTLdrClose(hLdrMod);
353	}
354	}
355	pData->phEvents[VBOX_SOCKET_EVENT_INDEX] = CreateEvent(NULL, FALSE, FALSE, NULL);
356	#endif
357
358	rc = bootp_dhcp_init(pData);
359	if (RT_FAILURE(rc))
360	{
361	Log(("NAT: DHCP server initialization failed\n"));
362	RTMemFree(pData);
363	*ppData = NULL;
364	return rc;
365	}
366	debug_init(pData);
367	if_init(pData);
368	ip_init(pData);
369	icmp_init(pData, iIcmpCacheLimit);
370
371	/* Initialise mbufs after setting the MTU */
372	mbuf_init(pData);
373
374	pData->special_addr.s_addr = u32NetAddr;
375	pData->slirp_ethaddr = &special_ethaddr[0];
376	alias_addr.s_addr = pData->special_addr.s_addr \| RT_H2N_U32_C(CTL_ALIAS);
377	/* @todo: add ability to configure this staff */
378
379	/* set default addresses */
380	inet_aton("127.0.0.1", &loopback_addr);
381
382	rc = slirpTftpInit(pData);
383	AssertRCReturn(rc, VINF_NAT_DNS);
384
385	if (i32AliasMode & ~(PKT_ALIAS_LOG\|PKT_ALIAS_SAME_PORTS\|PKT_ALIAS_PROXY_ONLY))
386	{
387	Log(("NAT: alias mode %x is ignored\n", i32AliasMode));
388	i32AliasMode = 0;
389	}
390	pData->i32AliasMode = i32AliasMode;
391	getouraddr(pData);
392	{
393	int flags = 0;
394	struct in_addr proxy_addr;
395	pData->proxy_alias = LibAliasInit(pData, NULL);
396	if (pData->proxy_alias == NULL)
397	{
398	Log(("NAT: LibAlias default rule wasn't initialized\n"));
399	AssertMsgFailed(("NAT: LibAlias default rule wasn't initialized\n"));
400	}
401	flags = LibAliasSetMode(pData->proxy_alias, 0, 0);
402	#ifndef NO_FW_PUNCH
403	flags \|= PKT_ALIAS_PUNCH_FW;
404	#endif
405	flags \|= pData->i32AliasMode; /* do transparent proxying */
406	flags = LibAliasSetMode(pData->proxy_alias, flags, ~0);
407	proxy_addr.s_addr = RT_H2N_U32(RT_N2H_U32(pData->special_addr.s_addr) \| CTL_ALIAS);
408	LibAliasSetAddress(pData->proxy_alias, proxy_addr);
409	ftp_alias_load(pData);
410	nbt_alias_load(pData);
411	if (pData->fUseHostResolver)
412	dns_alias_load(pData);
413	}
414	#ifdef VBOX_WITH_NAT_SEND2HOME
415	/* @todo: we should know all interfaces available on host. */
416	pData->pInSockAddrHomeAddress = RTMemAllocZ(sizeof(struct sockaddr));
417	pData->cInHomeAddressSize = 1;
418	inet_aton("192.168.1.25", &pData->pInSockAddrHomeAddress[0].sin_addr);
419	pData->pInSockAddrHomeAddress[0].sin_family = AF_INET;
420	# ifdef RT_OS_DARWIN
421	pData->pInSockAddrHomeAddress[0].sin_len = sizeof(struct sockaddr_in);
422	# endif
423	#endif
424
425	slirp_link_up(pData);
426	return VINF_SUCCESS;
427	}
428
429	/**
430	* Register statistics.
431	*/
432	void slirp_register_statistics(PNATState pData, PPDMDRVINS pDrvIns)
433	{
434	#ifdef VBOX_WITH_STATISTICS
435	# define PROFILE_COUNTER(name, dsc) REGISTER_COUNTER(name, pData, STAMTYPE_PROFILE, STAMUNIT_TICKS_PER_CALL, dsc)
436	# define COUNTING_COUNTER(name, dsc) REGISTER_COUNTER(name, pData, STAMTYPE_COUNTER, STAMUNIT_COUNT, dsc)
437	# include "counters.h"
438	# undef COUNTER
439	/** @todo register statistics for the variables dumped by:
440	* ipstats(pData); tcpstats(pData); udpstats(pData); icmpstats(pData);
441	* mbufstats(pData); sockstats(pData); */
442	#else /* VBOX_WITH_STATISTICS */
443	NOREF(pData);
444	NOREF(pDrvIns);
445	#endif /* !VBOX_WITH_STATISTICS */
446	}
447
448	/**
449	* Deregister statistics.
450	*/
451	void slirp_deregister_statistics(PNATState pData, PPDMDRVINS pDrvIns)
452	{
453	if (pData == NULL)
454	return;
455	#ifdef VBOX_WITH_STATISTICS
456	# define PROFILE_COUNTER(name, dsc) DEREGISTER_COUNTER(name, pData)
457	# define COUNTING_COUNTER(name, dsc) DEREGISTER_COUNTER(name, pData)
458	# include "counters.h"
459	#else /* VBOX_WITH_STATISTICS */
460	NOREF(pData);
461	NOREF(pDrvIns);
462	#endif /* !VBOX_WITH_STATISTICS */
463	}
464
465	/**
466	* Marks the link as up, making it possible to establish new connections.
467	*/
468	void slirp_link_up(PNATState pData)
469	{
470	struct arp_cache_entry *ac;
471
472	if (link_up == 1)
473	return;
474
475	link_up = 1;
476
477	if (!pData->fUseHostResolverPermanent)
478	slirpInitializeDnsSettings(pData);
479
480	if (LIST_EMPTY(&pData->arp_cache))
481	return;
482
483	LIST_FOREACH(ac, &pData->arp_cache, list)
484	{
485	activate_port_forwarding(pData, ac->ether);
486	}
487	}
488
489	/**
490	* Marks the link as down and cleans up the current connections.
491	*/
492	void slirp_link_down(PNATState pData)
493	{
494	struct port_forward_rule *rule;
495
496	if (link_up == 0)
497	return;
498
499	slirpReleaseDnsSettings(pData);
500
501	/*
502	* Clear the active state of port-forwarding rules to force
503	* re-setup on restoration of communications.
504	*/
505	LIST_FOREACH(rule, &pData->port_forward_rule_head, list)
506	{
507	rule->activated = 0;
508	}
509	pData->cRedirectionsActive = 0;
510
511	link_up = 0;
512	}
513
514	/**
515	* Terminates the slirp component.
516	*/
517	void slirp_term(PNATState pData)
518	{
519	struct socket *so;
520
521	if (pData == NULL)
522	return;
523
524	icmp_finit(pData);
525
526	while ((so = tcb.so_next) != &tcb)
527	{
528	/* Don't miss TCB releasing */
529	if ( !sototcpcb(so)
530	&& ( so->so_state & SS_NOFDREF
531	\|\| so->s == -1))
532	sofree(pData, so);
533	else
534	tcp_close(pData, sototcpcb(so));
535	}
536
537	while ((so = udb.so_next) != &udb)
538	udp_detach(pData, so);
539
540	slirp_link_down(pData);
541	ftp_alias_unload(pData);
542	nbt_alias_unload(pData);
543	if (pData->fUseHostResolver)
544	{
545	dns_alias_unload(pData);
546	#ifdef VBOX_WITH_DNSMAPPING_IN_HOSTRESOLVER
547	while (!LIST_EMPTY(&pData->DNSMapHead))
548	{
549	PDNSMAPPINGENTRY pDnsEntry = LIST_FIRST(&pData->DNSMapHead);
550	LIST_REMOVE(pDnsEntry, MapList);
551	RTStrFree(pDnsEntry->pszCName);
552	RTMemFree(pDnsEntry);
553	}
554	#endif
555	}
556	while (!LIST_EMPTY(&instancehead))
557	{
558	struct libalias *la = LIST_FIRST(&instancehead);
559	/* libalias do all clean up */
560	LibAliasUninit(la);
561	}
562	while (!LIST_EMPTY(&pData->arp_cache))
563	{
564	struct arp_cache_entry *ac = LIST_FIRST(&pData->arp_cache);
565	LIST_REMOVE(ac, list);
566	RTMemFree(ac);
567	}
568	slirpTftpTerm(pData);
569	bootp_dhcp_fini(pData);
570	m_fini(pData);
571	#ifdef RT_OS_WINDOWS
572	WSACleanup();
573	#endif
574	#ifdef LOG_ENABLED
575	Log(("\n"
576	"NAT statistics\n"
577	"--------------\n"
578	"\n"));
579	ipstats(pData);
580	tcpstats(pData);
581	udpstats(pData);
582	icmpstats(pData);
583	mbufstats(pData);
584	sockstats(pData);
585	Log(("\n"
586	"\n"
587	"\n"));
588	#endif
589	RTCritSectRwDelete(&pData->CsRwHandlerChain);
590	RTMemFree(pData);
591	}
592
593
594	#define CONN_CANFSEND(so) (((so)->so_state & (SS_FCANTSENDMORE\|SS_ISFCONNECTED)) == SS_ISFCONNECTED)
595	#define CONN_CANFRCV(so) (((so)->so_state & (SS_FCANTRCVMORE\|SS_ISFCONNECTED)) == SS_ISFCONNECTED)
596
597	/*
598	* curtime kept to an accuracy of 1ms
599	*/
600	static void updtime(PNATState pData)
601	{
602	#ifdef RT_OS_WINDOWS
603	struct _timeb tb;
604
605	_ftime(&tb);
606	curtime = (u_int)tb.time * (u_int)1000;
607	curtime += (u_int)tb.millitm;
608	#else
609	gettimeofday(&tt, 0);
610
611	curtime = (u_int)tt.tv_sec * (u_int)1000;
612	curtime += (u_int)tt.tv_usec / (u_int)1000;
613
614	if ((tt.tv_usec % 1000) >= 500)
615	curtime++;
616	#endif
617	}
618
619	#ifdef RT_OS_WINDOWS
620	void slirp_select_fill(PNATState pData, int *pnfds)
621	#else /* RT_OS_WINDOWS */
622	void slirp_select_fill(PNATState pData, int pnfds, struct pollfd polls)
623	#endif /* !RT_OS_WINDOWS */
624	{
625	struct socket so, so_next;
626	int nfds;
627	#if defined(RT_OS_WINDOWS)
628	int rc;
629	int error;
630	#else
631	int poll_index = 0;
632	#endif
633	int i;
634
635	STAM_PROFILE_START(&pData->StatFill, a);
636
637	nfds = *pnfds;
638
639	/*
640	* First, TCP sockets
641	*/
642	do_slowtimo = 0;
643	if (!link_up)
644	goto done;
645
646	/*
647	* *_slowtimo needs calling if there are IP fragments
648	* in the fragment queue, or there are TCP connections active
649	*/
650	/* XXX:
651	* triggering of fragment expiration should be the same but use new macroses
652	*/
653	do_slowtimo = (tcb.so_next != &tcb);
654	if (!do_slowtimo)
655	{
656	for (i = 0; i < IPREASS_NHASH; i++)
657	{
658	if (!TAILQ_EMPTY(&ipq[i]))
659	{
660	do_slowtimo = 1;
661	break;
662	}
663	}
664	}
665	/* always add the ICMP socket */
666	#ifndef RT_OS_WINDOWS
667	pData->icmp_socket.so_poll_index = -1;
668	#endif
669	ICMP_ENGAGE_EVENT(&pData->icmp_socket, readfds);
670
671	STAM_COUNTER_RESET(&pData->StatTCP);
672	STAM_COUNTER_RESET(&pData->StatTCPHot);
673
674	QSOCKET_FOREACH(so, so_next, tcp)
675	/* { */
676	Assert(so->so_type == IPPROTO_TCP);
677	#if !defined(RT_OS_WINDOWS)
678	so->so_poll_index = -1;
679	#endif
680	STAM_COUNTER_INC(&pData->StatTCP);
681	#ifdef VBOX_WITH_NAT_UDP_SOCKET_CLONE
682	/* TCP socket can't be cloned */
683	Assert((!so->so_cloneOf));
684	#endif
685	/*
686	* See if we need a tcp_fasttimo
687	*/
688	if ( time_fasttimo == 0
689	&& so->so_tcpcb != NULL
690	&& so->so_tcpcb->t_flags & TF_DELACK)
691	{
692	time_fasttimo = curtime; /* Flag when we want a fasttimo */
693	}
694
695	/*
696	* NOFDREF can include still connecting to local-host,
697	* newly socreated() sockets etc. Don't want to select these.
698	*/
699	if (so->so_state & SS_NOFDREF \|\| so->s == -1)
700	CONTINUE(tcp);
701
702	/*
703	* Set for reading sockets which are accepting
704	*/
705	if (so->so_state & SS_FACCEPTCONN)
706	{
707	STAM_COUNTER_INC(&pData->StatTCPHot);
708	TCP_ENGAGE_EVENT1(so, readfds);
709	CONTINUE(tcp);
710	}
711
712	/*
713	* Set for writing sockets which are connecting
714	*/
715	if (so->so_state & SS_ISFCONNECTING)
716	{
717	Log2(("connecting %R[natsock] engaged\n",so));
718	STAM_COUNTER_INC(&pData->StatTCPHot);
719	#ifdef RT_OS_WINDOWS
720	WIN_TCP_ENGAGE_EVENT2(so, writefds, connectfds);
721	#else
722	TCP_ENGAGE_EVENT1(so, writefds);
723	#endif
724	}
725
726	/*
727	* Set for writing if we are connected, can send more, and
728	* we have something to send
729	*/
730	if (CONN_CANFSEND(so) && SBUF_LEN(&so->so_rcv))
731	{
732	STAM_COUNTER_INC(&pData->StatTCPHot);
733	TCP_ENGAGE_EVENT1(so, writefds);
734	}
735
736	/*
737	* Set for reading (and urgent data) if we are connected, can
738	* receive more, and we have room for it XXX /2 ?
739	*/
740	/* @todo: vvl - check which predicat here will be more useful here in rerm of new sbufs. */
741	if ( CONN_CANFRCV(so)
742	&& (SBUF_LEN(&so->so_snd) < (SBUF_SIZE(&so->so_snd)/2))
743	#ifdef RT_OS_WINDOWS
744	&& !(so->so_state & SS_ISFCONNECTING)
745	#endif
746	)
747	{
748	STAM_COUNTER_INC(&pData->StatTCPHot);
749	TCP_ENGAGE_EVENT2(so, readfds, xfds);
750	}
751	LOOP_LABEL(tcp, so, so_next);
752	}
753
754	/*
755	* UDP sockets
756	*/
757	STAM_COUNTER_RESET(&pData->StatUDP);
758	STAM_COUNTER_RESET(&pData->StatUDPHot);
759
760	QSOCKET_FOREACH(so, so_next, udp)
761	/* { */
762
763	Assert(so->so_type == IPPROTO_UDP);
764	STAM_COUNTER_INC(&pData->StatUDP);
765	#if !defined(RT_OS_WINDOWS)
766	so->so_poll_index = -1;
767	#endif
768
769	/*
770	* See if it's timed out
771	*/
772	if (so->so_expire)
773	{
774	if (so->so_expire <= curtime)
775	{
776	Log2(("NAT: %R[natsock] expired\n", so));
777	if (so->so_timeout != NULL)
778	{
779	/* so_timeout - might change the so_expire value or
780	* drop so_timeout* from so.
781	*/
782	so->so_timeout(pData, so, so->so_timeout_arg);
783	/* on 4.2 so->
784	*/
785	if ( so_next->so_prev != so /* so_timeout freed the socket */
786	\|\| so->so_timeout) /* so_timeout just freed so_timeout */
787	CONTINUE_NO_UNLOCK(udp);
788	}
789	UDP_DETACH(pData, so, so_next);
790	CONTINUE_NO_UNLOCK(udp);
791	}
792	}
793	#ifdef VBOX_WITH_NAT_UDP_SOCKET_CLONE
794	if (so->so_cloneOf)
795	CONTINUE_NO_UNLOCK(udp);
796	#endif
797
798	/*
799	* When UDP packets are received from over the link, they're
800	* sendto()'d straight away, so no need for setting for writing
801	* Limit the number of packets queued by this session to 4.
802	* Note that even though we try and limit this to 4 packets,
803	* the session could have more queued if the packets needed
804	* to be fragmented.
805	*
806	* (XXX <= 4 ?)
807	*/
808	if ((so->so_state & SS_ISFCONNECTED) && so->so_queued <= 4)
809	{
810	STAM_COUNTER_INC(&pData->StatUDPHot);
811	UDP_ENGAGE_EVENT(so, readfds);
812	}
813	LOOP_LABEL(udp, so, so_next);
814	}
815	done:
816
817	#if defined(RT_OS_WINDOWS)
818	*pnfds = VBOX_EVENT_COUNT;
819	#else /* RT_OS_WINDOWS */
820	AssertRelease(poll_index <= *pnfds);
821	*pnfds = poll_index;
822	#endif /* !RT_OS_WINDOWS */
823
824	STAM_PROFILE_STOP(&pData->StatFill, a);
825	}
826
827
828	/**
829	* This function do Connection or sending tcp sequence to.
830	* @returns if true operation completed
831	* @note: functions call tcp_input that potentially could lead to tcp_drop
832	*/
833	static bool slirpConnectOrWrite(PNATState pData, struct socket *so, bool fConnectOnly)
834	{
835	int ret;
836	LogFlowFunc(("ENTER: so:%R[natsock], fConnectOnly:%RTbool\n", so, fConnectOnly));
837	/*
838	* Check for non-blocking, still-connecting sockets
839	*/
840	if (so->so_state & SS_ISFCONNECTING)
841	{
842	Log2(("connecting %R[natsock] catched\n", so));
843	/* Connected */
844	so->so_state &= ~SS_ISFCONNECTING;
845
846	/*
847	* This should be probably guarded by PROBE_CONN too. Anyway,
848	* we disable it on OS/2 because the below send call returns
849	* EFAULT which causes the opened TCP socket to close right
850	* after it has been opened and connected.
851	*/
852	#ifndef RT_OS_OS2
853	ret = send(so->s, (const char *)&ret, 0, 0);
854	if (ret < 0)
855	{
856	/* XXXXX Must fix, zero bytes is a NOP */
857	if ( soIgnorableErrorCode(errno)
858	\|\| errno == ENOTCONN)
859	{
860	LogFlowFunc(("LEAVE: false\n"));
861	return false;
862	}
863
864	/* else failed */
865	so->so_state = SS_NOFDREF;
866	}
867	/* else so->so_state &= ~SS_ISFCONNECTING; */
868	#endif
869
870	/*
871	* Continue tcp_input
872	*/
873	TCP_INPUT(pData, (struct mbuf *)NULL, sizeof(struct ip), so);
874	/* continue; */
875	}
876	else if (!fConnectOnly)
877	{
878	SOWRITE(ret, pData, so);
879	if (RT_LIKELY(ret > 0))
880	{
881	/*
882	* Make sure we will send window update to peer. This is
883	* a moral equivalent of calling tcp_output() for PRU_RCVD
884	* in tcp_usrreq() of the real stack.
885	*/
886	struct tcpcb *tp = sototcpcb(so);
887	if (RT_LIKELY(tp != NULL))
888	tp->t_flags \|= TF_DELACK;
889	}
890	}
891
892	LogFlowFunc(("LEAVE: true\n"));
893	return true;
894	}
895
896	#if defined(RT_OS_WINDOWS)
897	void slirp_select_poll(PNATState pData, int fTimeout)
898	#else /* RT_OS_WINDOWS */
899	void slirp_select_poll(PNATState pData, struct pollfd *polls, int ndfs)
900	#endif /* !RT_OS_WINDOWS */
901	{
902	struct socket so, so_next;
903	int ret;
904	#if defined(RT_OS_WINDOWS)
905	WSANETWORKEVENTS NetworkEvents;
906	int rc;
907	int error;
908	#endif
909
910	STAM_PROFILE_START(&pData->StatPoll, a);
911
912	/* Update time */
913	updtime(pData);
914
915	/*
916	* See if anything has timed out
917	*/
918	if (link_up)
919	{
920	if (time_fasttimo && ((curtime - time_fasttimo) >= 2))
921	{
922	STAM_PROFILE_START(&pData->StatFastTimer, b);
923	tcp_fasttimo(pData);
924	time_fasttimo = 0;
925	STAM_PROFILE_STOP(&pData->StatFastTimer, b);
926	}
927	if (do_slowtimo && ((curtime - last_slowtimo) >= 499))
928	{
929	STAM_PROFILE_START(&pData->StatSlowTimer, c);
930	ip_slowtimo(pData);
931	tcp_slowtimo(pData);
932	last_slowtimo = curtime;
933	STAM_PROFILE_STOP(&pData->StatSlowTimer, c);
934	}
935	}
936	#if defined(RT_OS_WINDOWS)
937	if (fTimeout)
938	return; /* only timer update */
939	#endif
940
941	/*
942	* Check sockets
943	*/
944	if (!link_up)
945	goto done;
946	#if defined(RT_OS_WINDOWS)
947	icmpwin_process(pData);
948	#else
949	if ( (pData->icmp_socket.s != -1)
950	&& CHECK_FD_SET(&pData->icmp_socket, ignored, readfds))
951	sorecvfrom(pData, &pData->icmp_socket);
952	#endif
953	/*
954	* Check TCP sockets
955	*/
956	QSOCKET_FOREACH(so, so_next, tcp)
957	/* { */
958	/* TCP socket can't be cloned */
959	#ifdef VBOX_WITH_NAT_UDP_SOCKET_CLONE
960	Assert((!so->so_cloneOf));
961	#endif
962	Assert(!so->fUnderPolling);
963	so->fUnderPolling = 1;
964	if (slirpVerifyAndFreeSocket(pData, so))
965	CONTINUE(tcp);
966	/*
967	* FD_ISSET is meaningless on these sockets
968	* (and they can crash the program)
969	*/
970	if (so->so_state & SS_NOFDREF \|\| so->s == -1)
971	{
972	so->fUnderPolling = 0;
973	CONTINUE(tcp);
974	}
975
976	POLL_TCP_EVENTS(rc, error, so, &NetworkEvents);
977
978	LOG_NAT_SOCK(so, TCP, &NetworkEvents, readfds, writefds, xfds);
979
980	if (so->so_state & SS_ISFCONNECTING)
981	{
982	int sockerr = 0;
983	#if !defined(RT_OS_WINDOWS)
984	{
985	int revents = 0;
986
987	/*
988	* Failed connect(2) is reported by poll(2) on
989	* different OSes with different combinations of
990	* POLLERR, POLLHUP, and POLLOUT.
991	*/
992	if ( CHECK_FD_SET(so, NetworkEvents, closefds) /* POLLHUP */
993	\|\| CHECK_FD_SET(so, NetworkEvents, rderr)) /* POLLERR */
994	{
995	revents = POLLHUP; /* squash to single "failed" flag */
996	}
997	#if defined(RT_OS_SOLARIS) \|\| defined(RT_OS_NETBSD)
998	/* Solaris and NetBSD report plain POLLOUT even on error */
999	else if (CHECK_FD_SET(so, NetworkEvents, writefds)) /* POLLOUT */
1000	{
1001	revents = POLLOUT;
1002	}
1003	#endif
1004
1005	if (revents != 0)
1006	{
1007	socklen_t optlen = (socklen_t)sizeof(sockerr);
1008	ret = getsockopt(so->s, SOL_SOCKET, SO_ERROR, &sockerr, &optlen);
1009
1010	if ( RT_UNLIKELY(ret < 0)
1011	\|\| ( (revents & POLLHUP)
1012	&& RT_UNLIKELY(sockerr == 0)))
1013	sockerr = ETIMEDOUT;
1014	}
1015	}
1016	#else /* RT_OS_WINDOWS */
1017	{
1018	if (NetworkEvents.lNetworkEvents & FD_CONNECT)
1019	sockerr = NetworkEvents.iErrorCode[FD_CONNECT_BIT];
1020	}
1021	#endif
1022	if (sockerr != 0)
1023	{
1024	tcp_fconnect_failed(pData, so, sockerr);
1025	ret = slirpVerifyAndFreeSocket(pData, so);
1026	Assert(ret == 1); /* freed */
1027	CONTINUE(tcp);
1028	}
1029
1030	/*
1031	* XXX: For now just fall through to the old code to
1032	* handle successful connect(2).
1033	*/
1034	}
1035
1036	/*
1037	* Check for URG data
1038	* This will soread as well, so no need to
1039	* test for readfds below if this succeeds
1040	*/
1041
1042	/* out-of-band data */
1043	if ( CHECK_FD_SET(so, NetworkEvents, xfds)
1044	#ifdef RT_OS_DARWIN
1045	/* Darwin and probably BSD hosts generates POLLPRI\|POLLHUP event on receiving TCP.flags.{ACK\|URG\|FIN} this
1046	* combination on other Unixs hosts doesn't enter to this branch
1047	*/
1048	&& !CHECK_FD_SET(so, NetworkEvents, closefds)
1049	#endif
1050	#ifdef RT_OS_WINDOWS
1051	/**
1052	* In some cases FD_CLOSE comes with FD_OOB, that confuse tcp processing.
1053	*/
1054	&& !WIN_CHECK_FD_SET(so, NetworkEvents, closefds)
1055	#endif
1056	)
1057	{
1058	sorecvoob(pData, so);
1059	if (slirpVerifyAndFreeSocket(pData, so))
1060	CONTINUE(tcp);
1061	}
1062
1063	/*
1064	* Check sockets for reading
1065	*/
1066	else if ( CHECK_FD_SET(so, NetworkEvents, readfds)
1067	\|\| WIN_CHECK_FD_SET(so, NetworkEvents, acceptds))
1068	{
1069
1070	#ifdef RT_OS_WINDOWS
1071	if (WIN_CHECK_FD_SET(so, NetworkEvents, connectfds))
1072	{
1073	/* Finish connection first */
1074	/* should we ignore return value? */
1075	bool fRet = slirpConnectOrWrite(pData, so, true);
1076	LogFunc(("fRet:%RTbool\n", fRet));
1077	if (slirpVerifyAndFreeSocket(pData, so))
1078	CONTINUE(tcp);
1079	}
1080	#endif
1081	/*
1082	* Check for incoming connections
1083	*/
1084	if (so->so_state & SS_FACCEPTCONN)
1085	{
1086	TCP_CONNECT(pData, so);
1087	if (slirpVerifyAndFreeSocket(pData, so))
1088	CONTINUE(tcp);
1089	if (!CHECK_FD_SET(so, NetworkEvents, closefds))
1090	{
1091	so->fUnderPolling = 0;
1092	CONTINUE(tcp);
1093	}
1094	}
1095
1096	ret = soread(pData, so);
1097	if (slirpVerifyAndFreeSocket(pData, so))
1098	CONTINUE(tcp);
1099	/* Output it if we read something */
1100	if (RT_LIKELY(ret > 0))
1101	TCP_OUTPUT(pData, sototcpcb(so));
1102
1103	if (slirpVerifyAndFreeSocket(pData, so))
1104	CONTINUE(tcp);
1105	}
1106
1107	/*
1108	* Check for FD_CLOSE events.
1109	* in some cases once FD_CLOSE engaged on socket it could be flashed latter (for some reasons)
1110	*/
1111	if ( CHECK_FD_SET(so, NetworkEvents, closefds)
1112	\|\| (so->so_close == 1))
1113	{
1114	/*
1115	* drain the socket
1116	*/
1117	for (; so_next->so_prev == so
1118	&& !slirpVerifyAndFreeSocket(pData, so);)
1119	{
1120	ret = soread(pData, so);
1121	if (slirpVerifyAndFreeSocket(pData, so))
1122	break;
1123
1124	if (ret > 0)
1125	TCP_OUTPUT(pData, sototcpcb(so));
1126	else if (so_next->so_prev == so)
1127	{
1128	Log2(("%R[natsock] errno %d (%s)\n", so, errno, strerror(errno)));
1129	break;
1130	}
1131	}
1132
1133	/* if socket freed ''so'' is PHANTOM and next socket isn't points on it */
1134	if (so_next->so_prev == so)
1135	{
1136	/* mark the socket for termination _after_ it was drained */
1137	so->so_close = 1;
1138	/* No idea about Windows but on Posix, POLLHUP means that we can't send more.
1139	* Actually in the specific error scenario, POLLERR is set as well. */
1140	#ifndef RT_OS_WINDOWS
1141	if (CHECK_FD_SET(so, NetworkEvents, rderr))
1142	sofcantsendmore(so);
1143	#endif
1144	}
1145	if (so_next->so_prev == so)
1146	so->fUnderPolling = 0;
1147	CONTINUE(tcp);
1148	}
1149
1150	/*
1151	* Check sockets for writing
1152	*/
1153	if ( CHECK_FD_SET(so, NetworkEvents, writefds)
1154	#ifdef RT_OS_WINDOWS
1155	\|\| WIN_CHECK_FD_SET(so, NetworkEvents, connectfds)
1156	#endif
1157	)
1158	{
1159	int fConnectOrWriteSuccess = slirpConnectOrWrite(pData, so, false);
1160	/* slirpConnectOrWrite could return true even if tcp_input called tcp_drop,
1161	* so we should be ready to such situations.
1162	*/
1163	if (slirpVerifyAndFreeSocket(pData, so))
1164	CONTINUE(tcp);
1165	else if (!fConnectOrWriteSuccess)
1166	{
1167	so->fUnderPolling = 0;
1168	CONTINUE(tcp);
1169	}
1170	/* slirpConnectionOrWrite succeeded and socket wasn't dropped */
1171	}
1172
1173	/*
1174	* Probe a still-connecting, non-blocking socket
1175	* to check if it's still alive
1176	*/
1177	#ifdef PROBE_CONN
1178	if (so->so_state & SS_ISFCONNECTING)
1179	{
1180	ret = recv(so->s, (char *)&ret, 0, 0);
1181
1182	if (ret < 0)
1183	{
1184	/* XXX */
1185	if ( soIgnorableErrorCode(errno)
1186	\|\| errno == ENOTCONN)
1187	{
1188	CONTINUE(tcp); /* Still connecting, continue */
1189	}
1190
1191	/* else failed */
1192	so->so_state = SS_NOFDREF;
1193
1194	/* tcp_input will take care of it */
1195	}
1196	else
1197	{
1198	ret = send(so->s, &ret, 0, 0);
1199	if (ret < 0)
1200	{
1201	/* XXX */
1202	if ( soIgnorableErrorCode(errno)
1203	\|\| errno == ENOTCONN)
1204	{
1205	CONTINUE(tcp);
1206	}
1207	/* else failed */
1208	so->so_state = SS_NOFDREF;
1209	}
1210	else
1211	so->so_state &= ~SS_ISFCONNECTING;
1212
1213	}
1214	TCP_INPUT((struct mbuf *)NULL, sizeof(struct ip),so);
1215	} /* SS_ISFCONNECTING */
1216	#endif
1217	if (!slirpVerifyAndFreeSocket(pData, so))
1218	so->fUnderPolling = 0;
1219	LOOP_LABEL(tcp, so, so_next);
1220	}
1221
1222	/*
1223	* Now UDP sockets.
1224	* Incoming packets are sent straight away, they're not buffered.
1225	* Incoming UDP data isn't buffered either.
1226	*/
1227	QSOCKET_FOREACH(so, so_next, udp)
1228	/* { */
1229	#ifdef VBOX_WITH_NAT_UDP_SOCKET_CLONE
1230	if (so->so_cloneOf)
1231	CONTINUE_NO_UNLOCK(udp);
1232	#endif
1233	#if 0
1234	so->fUnderPolling = 1;
1235	if(slirpVerifyAndFreeSocket(pData, so));
1236	CONTINUE(udp);
1237	so->fUnderPolling = 0;
1238	#endif
1239
1240	POLL_UDP_EVENTS(rc, error, so, &NetworkEvents);
1241
1242	LOG_NAT_SOCK(so, UDP, &NetworkEvents, readfds, writefds, xfds);
1243
1244	if (so->s != -1 && CHECK_FD_SET(so, NetworkEvents, readfds))
1245	{
1246	SORECVFROM(pData, so);
1247	}
1248	LOOP_LABEL(udp, so, so_next);
1249	}
1250
1251	done:
1252
1253	STAM_PROFILE_STOP(&pData->StatPoll, a);
1254	}
1255
1256
1257	struct arphdr
1258	{
1259	unsigned short ar_hrd; /* format of hardware address */
1260	unsigned short ar_pro; /* format of protocol address */
1261	unsigned char ar_hln; /* length of hardware address */
1262	unsigned char ar_pln; /* length of protocol address */
1263	unsigned short ar_op; /* ARP opcode (command) */
1264
1265	/*
1266	* Ethernet looks like this : This bit is variable sized however...
1267	*/
1268	unsigned char ar_sha[ETH_ALEN]; /* sender hardware address */
1269	unsigned char ar_sip[4]; /* sender IP address */
1270	unsigned char ar_tha[ETH_ALEN]; /* target hardware address */
1271	unsigned char ar_tip[4]; /* target IP address */
1272	};
1273	AssertCompileSize(struct arphdr, 28);
1274
1275	static void arp_output(PNATState pData, const uint8_t pcu8EtherSource, const struct arphdr pcARPHeaderSource, uint32_t ip4TargetAddress)
1276	{
1277	struct ethhdr *pEtherHeaderResponse;
1278	struct arphdr *pARPHeaderResponse;
1279	uint32_t ip4TargetAddressInHostFormat;
1280	struct mbuf *pMbufResponse;
1281
1282	Assert((pcu8EtherSource));
1283	if (!pcu8EtherSource)
1284	return;
1285	ip4TargetAddressInHostFormat = RT_N2H_U32(ip4TargetAddress);
1286
1287	pMbufResponse = m_getcl(pData, M_NOWAIT, MT_HEADER, M_PKTHDR);
1288	if (!pMbufResponse)
1289	return;
1290	pEtherHeaderResponse = mtod(pMbufResponse, struct ethhdr *);
1291	/* @note: if_encap will swap src and dst*/
1292	memcpy(pEtherHeaderResponse->h_source, pcu8EtherSource, ETH_ALEN);
1293	pMbufResponse->m_data += ETH_HLEN;
1294	pARPHeaderResponse = mtod(pMbufResponse, struct arphdr *);
1295	pMbufResponse->m_len = sizeof(struct arphdr);
1296
1297	pARPHeaderResponse->ar_hrd = RT_H2N_U16_C(1);
1298	pARPHeaderResponse->ar_pro = RT_H2N_U16_C(ETH_P_IP);
1299	pARPHeaderResponse->ar_hln = ETH_ALEN;
1300	pARPHeaderResponse->ar_pln = 4;
1301	pARPHeaderResponse->ar_op = RT_H2N_U16_C(ARPOP_REPLY);
1302	memcpy(pARPHeaderResponse->ar_sha, special_ethaddr, ETH_ALEN);
1303
1304	if (!slirpMbufTagService(pData, pMbufResponse, (uint8_t)(ip4TargetAddressInHostFormat & ~pData->netmask)))
1305	{
1306	static bool fTagErrorReported;
1307	if (!fTagErrorReported)
1308	{
1309	LogRel(("NAT: couldn't add the tag(PACKET_SERVICE:%d)\n",
1310	(uint8_t)(ip4TargetAddressInHostFormat & ~pData->netmask)));
1311	fTagErrorReported = true;
1312	}
1313	}
1314	pARPHeaderResponse->ar_sha[5] = (uint8_t)(ip4TargetAddressInHostFormat & ~pData->netmask);
1315
1316	memcpy(pARPHeaderResponse->ar_sip, pcARPHeaderSource->ar_tip, 4);
1317	memcpy(pARPHeaderResponse->ar_tha, pcARPHeaderSource->ar_sha, ETH_ALEN);
1318	memcpy(pARPHeaderResponse->ar_tip, pcARPHeaderSource->ar_sip, 4);
1319	if_encap(pData, ETH_P_ARP, pMbufResponse, ETH_ENCAP_URG);
1320	}
1321	/**
1322	* @note This function will free m!
1323	*/
1324	static void arp_input(PNATState pData, struct mbuf *m)
1325	{
1326	struct ethhdr *pEtherHeader;
1327	struct arphdr *pARPHeader;
1328	uint32_t ip4TargetAddress;
1329
1330	int ar_op;
1331	pEtherHeader = mtod(m, struct ethhdr *);
1332	pARPHeader = (struct arphdr *)&pEtherHeader[1];
1333
1334	ar_op = RT_N2H_U16(pARPHeader->ar_op);
1335	ip4TargetAddress = (uint32_t)pARPHeader->ar_tip;
1336
1337	switch (ar_op)
1338	{
1339	case ARPOP_REQUEST:
1340	if ( CTL_CHECK(ip4TargetAddress, CTL_DNS)
1341	\|\| CTL_CHECK(ip4TargetAddress, CTL_ALIAS)
1342	\|\| CTL_CHECK(ip4TargetAddress, CTL_TFTP))
1343	arp_output(pData, pEtherHeader->h_source, pARPHeader, ip4TargetAddress);
1344
1345	/* Gratuitous ARP */
1346	if ( (uint32_t )pARPHeader->ar_sip == (uint32_t )pARPHeader->ar_tip
1347	&& memcmp(pARPHeader->ar_tha, broadcast_ethaddr, ETH_ALEN) == 0
1348	&& memcmp(pEtherHeader->h_dest, broadcast_ethaddr, ETH_ALEN) == 0)
1349	{
1350	/* We've received an announce about address assignment,
1351	* let's do an ARP cache update
1352	*/
1353	static bool fGratuitousArpReported;
1354	if (!fGratuitousArpReported)
1355	{
1356	LogRel(("NAT: Gratuitous ARP [IP:%RTnaipv4, ether:%RTmac]\n",
1357	(uint32_t )pARPHeader->ar_sip, pARPHeader->ar_sha));
1358	fGratuitousArpReported = true;
1359	}
1360	slirp_arp_cache_update_or_add(pData, (uint32_t )pARPHeader->ar_sip, &pARPHeader->ar_sha[0]);
1361	}
1362	break;
1363
1364	case ARPOP_REPLY:
1365	slirp_arp_cache_update_or_add(pData, (uint32_t )pARPHeader->ar_sip, &pARPHeader->ar_sha[0]);
1366	break;
1367
1368	default:
1369	break;
1370	}
1371
1372	m_freem(pData, m);
1373	}
1374
1375	/**
1376	* Feed a packet into the slirp engine.
1377	*
1378	* @param m Data buffer, m_len is not valid.
1379	* @param cbBuf The length of the data in m.
1380	*/
1381	void slirp_input(PNATState pData, struct mbuf *m, size_t cbBuf)
1382	{
1383	int proto;
1384	static bool fWarnedIpv6;
1385	struct ethhdr *eh;
1386	uint8_t au8Ether[ETH_ALEN];
1387
1388	m->m_len = cbBuf;
1389	if (cbBuf < ETH_HLEN)
1390	{
1391	Log(("NAT: packet having size %d has been ignored\n", m->m_len));
1392	m_freem(pData, m);
1393	return;
1394	}
1395	eh = mtod(m, struct ethhdr *);
1396	proto = RT_N2H_U16(eh->h_proto);
1397
1398	memcpy(au8Ether, eh->h_source, ETH_ALEN);
1399
1400	switch(proto)
1401	{
1402	case ETH_P_ARP:
1403	arp_input(pData, m);
1404	break;
1405
1406	case ETH_P_IP:
1407	/* Update time. Important if the network is very quiet, as otherwise
1408	* the first outgoing connection gets an incorrect timestamp. */
1409	updtime(pData);
1410	m_adj(m, ETH_HLEN);
1411	M_ASSERTPKTHDR(m);
1412	m->m_pkthdr.header = mtod(m, void *);
1413	ip_input(pData, m);
1414	break;
1415
1416	case ETH_P_IPV6:
1417	m_freem(pData, m);
1418	if (!fWarnedIpv6)
1419	{
1420	LogRel(("NAT: IPv6 not supported\n"));
1421	fWarnedIpv6 = true;
1422	}
1423	break;
1424
1425	default:
1426	Log(("NAT: Unsupported protocol %x\n", proto));
1427	m_freem(pData, m);
1428	break;
1429	}
1430
1431	if (pData->cRedirectionsActive != pData->cRedirectionsStored)
1432	activate_port_forwarding(pData, au8Ether);
1433	}
1434
1435	/**
1436	* Output the IP packet to the ethernet device.
1437	*
1438	* @note This function will free m!
1439	*/
1440	void if_encap(PNATState pData, uint16_t eth_proto, struct mbuf *m, int flags)
1441	{
1442	struct ethhdr *eh;
1443	uint8_t *mbuf = NULL;
1444	size_t mlen = 0;
1445	STAM_PROFILE_START(&pData->StatIF_encap, a);
1446	LogFlowFunc(("ENTER: pData:%p, eth_proto:%RX16, m:%p, flags:%d\n",
1447	pData, eth_proto, m, flags));
1448
1449	M_ASSERTPKTHDR(m);
1450
1451	Assert(M_LEADINGSPACE(m) >= ETH_HLEN);
1452	m->m_data -= ETH_HLEN;
1453	m->m_len += ETH_HLEN;
1454	eh = mtod(m, struct ethhdr *);
1455	mlen = m->m_len;
1456
1457	if (memcmp(eh->h_source, special_ethaddr, ETH_ALEN) != 0)
1458	{
1459	struct m_tag *t = m_tag_first(m);
1460	uint8_t u8ServiceId = CTL_ALIAS;
1461	memcpy(eh->h_dest, eh->h_source, ETH_ALEN);
1462	memcpy(eh->h_source, special_ethaddr, ETH_ALEN);
1463	Assert(memcmp(eh->h_dest, special_ethaddr, ETH_ALEN) != 0);
1464	if (memcmp(eh->h_dest, zerro_ethaddr, ETH_ALEN) == 0)
1465	{
1466	/* don't do anything */
1467	m_freem(pData, m);
1468	goto done;
1469	}
1470	if ( t
1471	&& (t = m_tag_find(m, PACKET_SERVICE, NULL)))
1472	{
1473	Assert(t);
1474	u8ServiceId = (uint8_t )&t[1];
1475	}
1476	eh->h_source[5] = u8ServiceId;
1477	}
1478	/*
1479	* we're processing the chain, that isn't not expected.
1480	*/
1481	Assert((!m->m_next));
1482	if (m->m_next)
1483	{
1484	Log(("NAT: if_encap's recived the chain, dropping...\n"));
1485	m_freem(pData, m);
1486	goto done;
1487	}
1488	mbuf = mtod(m, uint8_t *);
1489	eh->h_proto = RT_H2N_U16(eth_proto);
1490	LogFunc(("eh(dst:%RTmac, src:%RTmac)\n", eh->h_dest, eh->h_source));
1491	if (flags & ETH_ENCAP_URG)
1492	slirp_urg_output(pData->pvUser, m, mbuf, mlen);
1493	else
1494	slirp_output(pData->pvUser, m, mbuf, mlen);
1495	done:
1496	STAM_PROFILE_STOP(&pData->StatIF_encap, a);
1497	LogFlowFuncLeave();
1498	}
1499
1500	/**
1501	* Still we're using dhcp server leasing to map ether to IP
1502	* @todo see rt_lookup_in_cache
1503	*/
1504	static uint32_t find_guest_ip(PNATState pData, const uint8_t *eth_addr)
1505	{
1506	uint32_t ip = INADDR_ANY;
1507	int rc;
1508
1509	if (eth_addr == NULL)
1510	return INADDR_ANY;
1511
1512	if ( memcmp(eth_addr, zerro_ethaddr, ETH_ALEN) == 0
1513	\|\| memcmp(eth_addr, broadcast_ethaddr, ETH_ALEN) == 0)
1514	return INADDR_ANY;
1515
1516	rc = slirp_arp_lookup_ip_by_ether(pData, eth_addr, &ip);
1517	if (RT_SUCCESS(rc))
1518	return ip;
1519
1520	bootp_cache_lookup_ip_by_ether(pData, eth_addr, &ip);
1521	/* ignore return code, ip will be set to INADDR_ANY on error */
1522	return ip;
1523	}
1524
1525	/**
1526	* We need check if we've activated port forwarding
1527	* for specific machine ... that of course relates to
1528	* service mode
1529	* @todo finish this for service case
1530	*/
1531	static void activate_port_forwarding(PNATState pData, const uint8_t *h_source)
1532	{
1533	struct port_forward_rule rule, tmp;
1534	const uint8_t *pu8EthSource = h_source;
1535
1536	/* check mac here */
1537	LIST_FOREACH_SAFE(rule, &pData->port_forward_rule_head, list, tmp)
1538	{
1539	struct socket *so;
1540	struct sockaddr sa;
1541	struct sockaddr_in *psin;
1542	socklen_t socketlen;
1543	int rc;
1544	uint32_t guest_addr; /* need to understand if we already give address to guest */
1545
1546	if (rule->activated)
1547	continue;
1548
1549	guest_addr = find_guest_ip(pData, pu8EthSource);
1550	if (guest_addr == INADDR_ANY)
1551	{
1552	/* the address wasn't granted */
1553	return;
1554	}
1555
1556	if ( rule->guest_addr.s_addr != guest_addr
1557	&& rule->guest_addr.s_addr != INADDR_ANY)
1558	continue;
1559	if (rule->guest_addr.s_addr == INADDR_ANY)
1560	rule->guest_addr.s_addr = guest_addr;
1561
1562	LogRel(("NAT: set redirect %s host %RTnaipv4:%d => guest %RTnaipv4:%d\n",
1563	rule->proto == IPPROTO_UDP ? "UDP" : "TCP",
1564	rule->bind_ip.s_addr, rule->host_port,
1565	guest_addr, rule->guest_port));
1566
1567	if (rule->proto == IPPROTO_UDP)
1568	so = udp_listen(pData, rule->bind_ip.s_addr, RT_H2N_U16(rule->host_port), guest_addr,
1569	RT_H2N_U16(rule->guest_port), 0);
1570	else
1571	so = solisten(pData, rule->bind_ip.s_addr, RT_H2N_U16(rule->host_port), guest_addr,
1572	RT_H2N_U16(rule->guest_port), 0);
1573
1574	if (so == NULL)
1575	goto remove_port_forwarding;
1576
1577	psin = (struct sockaddr_in *)&sa;
1578	psin->sin_family = AF_INET;
1579	psin->sin_port = 0;
1580	psin->sin_addr.s_addr = INADDR_ANY;
1581	socketlen = sizeof(struct sockaddr);
1582
1583	rc = getsockname(so->s, &sa, &socketlen);
1584	if (rc < 0 \|\| sa.sa_family != AF_INET)
1585	goto remove_port_forwarding;
1586
1587	rule->activated = 1;
1588	rule->so = so;
1589	pData->cRedirectionsActive++;
1590	continue;
1591
1592	remove_port_forwarding:
1593	LogRel(("NAT: failed to redirect %s %RTnaipv4:%d => %RTnaipv4:%d\n",
1594	(rule->proto == IPPROTO_UDP ? "UDP" : "TCP"),
1595	rule->bind_ip.s_addr, rule->host_port,
1596	guest_addr, rule->guest_port));
1597	LIST_REMOVE(rule, list);
1598	pData->cRedirectionsStored--;
1599	RTMemFree(rule);
1600	}
1601	}
1602
1603	/**
1604	* Changes in 3.1 instead of opening new socket do the following:
1605	* gain more information:
1606	* 1. bind IP
1607	* 2. host port
1608	* 3. guest port
1609	* 4. proto
1610	* 5. guest MAC address
1611	* the guest's MAC address is rather important for service, but we easily
1612	* could get it from VM configuration in DrvNAT or Service, the idea is activating
1613	* corresponding port-forwarding
1614	*/
1615	int slirp_add_redirect(PNATState pData, int is_udp, struct in_addr host_addr, int host_port,
1616	struct in_addr guest_addr, int guest_port, const uint8_t *ethaddr)
1617	{
1618	struct port_forward_rule *rule = NULL;
1619	LIST_FOREACH(rule, &pData->port_forward_rule_head, list)
1620	{
1621	if ( rule->proto == (is_udp ? IPPROTO_UDP : IPPROTO_TCP)
1622	&& rule->host_port == host_port
1623	&& rule->bind_ip.s_addr == host_addr.s_addr
1624	&& rule->guest_port == guest_port
1625	&& rule->guest_addr.s_addr == guest_addr.s_addr
1626	)
1627	return 0; /* rule has been already registered */
1628	}
1629
1630	rule = RTMemAllocZ(sizeof(struct port_forward_rule));
1631	if (rule == NULL)
1632	return 1;
1633
1634	rule->proto = (is_udp ? IPPROTO_UDP : IPPROTO_TCP);
1635	rule->host_port = host_port;
1636	rule->guest_port = guest_port;
1637	rule->guest_addr.s_addr = guest_addr.s_addr;
1638	rule->bind_ip.s_addr = host_addr.s_addr;
1639	if (ethaddr != NULL)
1640	memcpy(rule->mac_address, ethaddr, ETH_ALEN);
1641	/* @todo add mac address */
1642	LIST_INSERT_HEAD(&pData->port_forward_rule_head, rule, list);
1643	pData->cRedirectionsStored++;
1644	/* activate port-forwarding if guest has already got assigned IP */
1645	if ( ethaddr
1646	&& memcmp(ethaddr, zerro_ethaddr, ETH_ALEN))
1647	activate_port_forwarding(pData, ethaddr);
1648	return 0;
1649	}
1650
1651	int slirp_remove_redirect(PNATState pData, int is_udp, struct in_addr host_addr, int host_port,
1652	struct in_addr guest_addr, int guest_port)
1653	{
1654	struct port_forward_rule *rule = NULL;
1655	LIST_FOREACH(rule, &pData->port_forward_rule_head, list)
1656	{
1657	if ( rule->proto == (is_udp ? IPPROTO_UDP : IPPROTO_TCP)
1658	&& rule->host_port == host_port
1659	&& rule->guest_port == guest_port
1660	&& rule->bind_ip.s_addr == host_addr.s_addr
1661	&& rule->guest_addr.s_addr == guest_addr.s_addr
1662	&& rule->activated)
1663	{
1664	LogRel(("NAT: remove redirect %s host %RTnaipv4:%d => guest %RTnaipv4:%d\n",
1665	rule->proto == IPPROTO_UDP ? "UDP" : "TCP",
1666	rule->bind_ip.s_addr, rule->host_port,
1667	guest_addr.s_addr, rule->guest_port));
1668
1669	if (is_udp)
1670	udp_detach(pData, rule->so);
1671	else
1672	tcp_close(pData, sototcpcb(rule->so));
1673	LIST_REMOVE(rule, list);
1674	RTMemFree(rule);
1675	pData->cRedirectionsStored--;
1676	break;
1677	}
1678
1679	}
1680	return 0;
1681	}
1682
1683	void slirp_set_ethaddr_and_activate_port_forwarding(PNATState pData, const uint8_t *ethaddr, uint32_t GuestIP)
1684	{
1685	memcpy(client_ethaddr, ethaddr, ETH_ALEN);
1686	if (GuestIP != INADDR_ANY)
1687	{
1688	slirp_arp_cache_update_or_add(pData, GuestIP, ethaddr);
1689	activate_port_forwarding(pData, ethaddr);
1690	}
1691	}
1692
1693	#if defined(RT_OS_WINDOWS)
1694	HANDLE *slirp_get_events(PNATState pData)
1695	{
1696	return pData->phEvents;
1697	}
1698	void slirp_register_external_event(PNATState pData, HANDLE hEvent, int index)
1699	{
1700	pData->phEvents[index] = hEvent;
1701	}
1702	#endif
1703
1704	unsigned int slirp_get_timeout_ms(PNATState pData)
1705	{
1706	if (link_up)
1707	{
1708	if (time_fasttimo)
1709	return 2;
1710	if (do_slowtimo)
1711	return 500; /* see PR_SLOWHZ */
1712	}
1713	return 36001000; / one hour */
1714	}
1715
1716	#ifndef RT_OS_WINDOWS
1717	int slirp_get_nsock(PNATState pData)
1718	{
1719	return pData->nsock;
1720	}
1721	#endif
1722
1723	/*
1724	* this function called from NAT thread
1725	*/
1726	void slirp_post_sent(PNATState pData, void *pvArg)
1727	{
1728	struct mbuf m = (struct mbuf )pvArg;
1729	m_freem(pData, m);
1730	}
1731
1732	void slirp_set_dhcp_TFTP_prefix(PNATState pData, const char *tftpPrefix)
1733	{
1734	Log2(("tftp_prefix: %s\n", tftpPrefix));
1735	tftp_prefix = tftpPrefix;
1736	}
1737
1738	void slirp_set_dhcp_TFTP_bootfile(PNATState pData, const char *bootFile)
1739	{
1740	Log2(("bootFile: %s\n", bootFile));
1741	bootp_filename = bootFile;
1742	}
1743
1744	void slirp_set_dhcp_next_server(PNATState pData, const char *next_server)
1745	{
1746	Log2(("next_server: %s\n", next_server));
1747	if (next_server == NULL)
1748	pData->tftp_server.s_addr = RT_H2N_U32(RT_N2H_U32(pData->special_addr.s_addr) \| CTL_TFTP);
1749	else
1750	inet_aton(next_server, &pData->tftp_server);
1751	}
1752
1753	int slirp_set_binding_address(PNATState pData, char *addr)
1754	{
1755	if (addr == NULL \|\| (inet_aton(addr, &pData->bindIP) == 0))
1756	{
1757	pData->bindIP.s_addr = INADDR_ANY;
1758	return 1;
1759	}
1760	return 0;
1761	}
1762
1763	void slirp_set_dhcp_dns_proxy(PNATState pData, bool fDNSProxy)
1764	{
1765	if (!pData->fUseHostResolver)
1766	{
1767	Log2(("NAT: DNS proxy switched %s\n", (fDNSProxy ? "on" : "off")));
1768	pData->fUseDnsProxy = fDNSProxy;
1769	}
1770	else if (fDNSProxy)
1771	LogRel(("NAT: Host Resolver conflicts with DNS proxy, the last one was forcely ignored\n"));
1772	}
1773
1774	#define CHECK_ARG(name, val, lim_min, lim_max) \
1775	do { \
1776	if ((val) < (lim_min) \|\| (val) > (lim_max)) \
1777	{ \
1778	LogRel(("NAT: (" #name ":%d) has been ignored, " \
1779	"because out of range (%d, %d)\n", (val), (lim_min), (lim_max))); \
1780	return; \
1781	} \
1782	else \
1783	LogRel(("NAT: (" #name ":%d)\n", (val))); \
1784	} while (0)
1785
1786	void slirp_set_somaxconn(PNATState pData, int iSoMaxConn)
1787	{
1788	LogFlowFunc(("iSoMaxConn:%d\n", iSoMaxConn));
1789	/* Conditions */
1790	if (iSoMaxConn > SOMAXCONN)
1791	{
1792	LogRel(("NAT: value of somaxconn(%d) bigger than SOMAXCONN(%d)\n", iSoMaxConn, SOMAXCONN));
1793	iSoMaxConn = SOMAXCONN;
1794	}
1795
1796	if (iSoMaxConn < 1)
1797	{
1798	LogRel(("NAT: proposed value(%d) of somaxconn is invalid, default value is used (%d)\n", iSoMaxConn, pData->soMaxConn));
1799	LogFlowFuncLeave();
1800	return;
1801	}
1802
1803	/* Asignment */
1804	if (pData->soMaxConn != iSoMaxConn)
1805	{
1806	LogRel(("NAT: value of somaxconn has been changed from %d to %d\n",
1807	pData->soMaxConn, iSoMaxConn));
1808	pData->soMaxConn = iSoMaxConn;
1809	}
1810	LogFlowFuncLeave();
1811	}
1812	/* don't allow user set less 8kB and more than 1M values */
1813	#define _8K_1M_CHECK_ARG(name, val) CHECK_ARG(name, (val), 8, 1024)
1814	void slirp_set_rcvbuf(PNATState pData, int kilobytes)
1815	{
1816	_8K_1M_CHECK_ARG("SOCKET_RCVBUF", kilobytes);
1817	pData->socket_rcv = kilobytes;
1818	}
1819	void slirp_set_sndbuf(PNATState pData, int kilobytes)
1820	{
1821	_8K_1M_CHECK_ARG("SOCKET_SNDBUF", kilobytes);
1822	pData->socket_snd = kilobytes * _1K;
1823	}
1824	void slirp_set_tcp_rcvspace(PNATState pData, int kilobytes)
1825	{
1826	_8K_1M_CHECK_ARG("TCP_RCVSPACE", kilobytes);
1827	tcp_rcvspace = kilobytes * _1K;
1828	}
1829	void slirp_set_tcp_sndspace(PNATState pData, int kilobytes)
1830	{
1831	_8K_1M_CHECK_ARG("TCP_SNDSPACE", kilobytes);
1832	tcp_sndspace = kilobytes * _1K;
1833	}
1834
1835	/*
1836	* Looking for Ether by ip in ARP-cache
1837	* Note: it´s responsible of caller to allocate buffer for result
1838	* @returns iprt status code
1839	*/
1840	int slirp_arp_lookup_ether_by_ip(PNATState pData, uint32_t ip, uint8_t *ether)
1841	{
1842	struct arp_cache_entry *ac;
1843
1844	if (ether == NULL)
1845	return VERR_INVALID_PARAMETER;
1846
1847	if (LIST_EMPTY(&pData->arp_cache))
1848	return VERR_NOT_FOUND;
1849
1850	LIST_FOREACH(ac, &pData->arp_cache, list)
1851	{
1852	if ( ac->ip == ip
1853	&& memcmp(ac->ether, broadcast_ethaddr, ETH_ALEN) != 0)
1854	{
1855	memcpy(ether, ac->ether, ETH_ALEN);
1856	return VINF_SUCCESS;
1857	}
1858	}
1859	return VERR_NOT_FOUND;
1860	}
1861
1862	/*
1863	* Looking for IP by Ether in ARP-cache
1864	* Note: it´s responsible of caller to allocate buffer for result
1865	* @returns 0 - if found, 1 - otherwise
1866	*/
1867	int slirp_arp_lookup_ip_by_ether(PNATState pData, const uint8_t ether, uint32_t ip)
1868	{
1869	struct arp_cache_entry *ac;
1870	*ip = INADDR_ANY;
1871
1872	if (LIST_EMPTY(&pData->arp_cache))
1873	return VERR_NOT_FOUND;
1874
1875	LIST_FOREACH(ac, &pData->arp_cache, list)
1876	{
1877	if (memcmp(ether, ac->ether, ETH_ALEN) == 0)
1878	{
1879	*ip = ac->ip;
1880	return VINF_SUCCESS;
1881	}
1882	}
1883	return VERR_NOT_FOUND;
1884	}
1885
1886	void slirp_arp_who_has(PNATState pData, uint32_t dst)
1887	{
1888	struct mbuf *m;
1889	struct ethhdr *ehdr;
1890	struct arphdr *ahdr;
1891	static bool fWarned = false;
1892	LogFlowFunc(("ENTER: %RTnaipv4\n", dst));
1893
1894	/* ARP request WHO HAS 0.0.0.0 is one of the signals
1895	* that something has been broken at Slirp. Investigating
1896	* pcap dumps it's easy to miss warning ARP requests being
1897	* focused on investigation of other protocols flow.
1898	*/
1899	#ifdef DEBUG_vvl
1900	Assert((dst != INADDR_ANY));
1901	NOREF(fWarned);
1902	#else
1903	if ( dst == INADDR_ANY
1904	&& !fWarned)
1905	{
1906	LogRel(("NAT:ARP: \"WHO HAS INADDR_ANY\" request has been detected\n"));
1907	fWarned = true;
1908	}
1909	#endif /* !DEBUG_vvl */
1910
1911	m = m_getcl(pData, M_NOWAIT, MT_HEADER, M_PKTHDR);
1912	if (m == NULL)
1913	{
1914	Log(("NAT: Can't alloc mbuf for ARP request\n"));
1915	LogFlowFuncLeave();
1916	return;
1917	}
1918	ehdr = mtod(m, struct ethhdr *);
1919	memset(ehdr->h_source, 0xff, ETH_ALEN);
1920	ahdr = (struct arphdr *)&ehdr[1];
1921	ahdr->ar_hrd = RT_H2N_U16_C(1);
1922	ahdr->ar_pro = RT_H2N_U16_C(ETH_P_IP);
1923	ahdr->ar_hln = ETH_ALEN;
1924	ahdr->ar_pln = 4;
1925	ahdr->ar_op = RT_H2N_U16_C(ARPOP_REQUEST);
1926	memcpy(ahdr->ar_sha, special_ethaddr, ETH_ALEN);
1927	/* we assume that this request come from gw, but not from DNS or TFTP */
1928	ahdr->ar_sha[5] = CTL_ALIAS;
1929	(uint32_t )ahdr->ar_sip = RT_H2N_U32(RT_N2H_U32(pData->special_addr.s_addr) \| CTL_ALIAS);
1930	memset(ahdr->ar_tha, 0xff, ETH_ALEN); /broadcast/
1931	(uint32_t )ahdr->ar_tip = dst;
1932	/* warn!!! should falls in mbuf minimal size */
1933	m->m_len = sizeof(struct arphdr) + ETH_HLEN;
1934	m->m_data += ETH_HLEN;
1935	m->m_len -= ETH_HLEN;
1936	if_encap(pData, ETH_P_ARP, m, ETH_ENCAP_URG);
1937	LogFlowFuncLeave();
1938	}
1939	#ifdef VBOX_WITH_DNSMAPPING_IN_HOSTRESOLVER
1940	void slirp_add_host_resolver_mapping(PNATState pData, const char pszHostName, const char pszHostNamePattern, uint32_t u32HostIP)
1941	{
1942	LogFlowFunc(("ENTER: pszHostName:%s, pszHostNamePattern:%s u32HostIP:%RTnaipv4\n",
1943	pszHostName ? pszHostName : "(null)",
1944	pszHostNamePattern ? pszHostNamePattern : "(null)",
1945	u32HostIP));
1946	if ( ( pszHostName
1947	\|\| pszHostNamePattern)
1948	&& u32HostIP != INADDR_ANY
1949	&& u32HostIP != INADDR_BROADCAST)
1950	{
1951	PDNSMAPPINGENTRY pDnsMapping = RTMemAllocZ(sizeof(DNSMAPPINGENTRY));
1952	if (!pDnsMapping)
1953	{
1954	LogFunc(("Can't allocate DNSMAPPINGENTRY\n"));
1955	LogFlowFuncLeave();
1956	return;
1957	}
1958	pDnsMapping->u32IpAddress = u32HostIP;
1959	if (pszHostName)
1960	pDnsMapping->pszCName = RTStrDup(pszHostName);
1961	else if (pszHostNamePattern)
1962	pDnsMapping->pszPattern = RTStrDup(pszHostNamePattern);
1963	if ( !pDnsMapping->pszCName
1964	&& !pDnsMapping->pszPattern)
1965	{
1966	LogFunc(("Can't allocate enough room for %s\n", pszHostName ? pszHostName : pszHostNamePattern));
1967	RTMemFree(pDnsMapping);
1968	LogFlowFuncLeave();
1969	return;
1970	}
1971	LIST_INSERT_HEAD(&pData->DNSMapHead, pDnsMapping, MapList);
1972	LogRel(("NAT: user-defined mapping %s: %RTnaipv4 is registered\n",
1973	pDnsMapping->pszCName ? pDnsMapping->pszCName : pDnsMapping->pszPattern,
1974	pDnsMapping->u32IpAddress));
1975	}
1976	LogFlowFuncLeave();
1977	}
1978	#endif
1979
1980	/* updates the arp cache
1981	* @note: this is helper function, slirp_arp_cache_update_or_add should be used.
1982	* @returns 0 - if has found and updated
1983	* 1 - if hasn't found.
1984	*/
1985	static inline int slirp_arp_cache_update(PNATState pData, uint32_t dst, const uint8_t *mac)
1986	{
1987	struct arp_cache_entry *ac;
1988	Assert(( memcmp(mac, broadcast_ethaddr, ETH_ALEN)
1989	&& memcmp(mac, zerro_ethaddr, ETH_ALEN)));
1990	LIST_FOREACH(ac, &pData->arp_cache, list)
1991	{
1992	if (ac->ip == dst)
1993	{
1994	memcpy(ac->ether, mac, ETH_ALEN);
1995	return 0;
1996	}
1997	}
1998	return 1;
1999	}
2000
2001	/**
2002	* add entry to the arp cache
2003	* @note: this is helper function, slirp_arp_cache_update_or_add should be used.
2004	*/
2005	static inline void slirp_arp_cache_add(PNATState pData, uint32_t ip, const uint8_t *ether)
2006	{
2007	struct arp_cache_entry *ac = NULL;
2008	Assert(( memcmp(ether, broadcast_ethaddr, ETH_ALEN)
2009	&& memcmp(ether, zerro_ethaddr, ETH_ALEN)));
2010	ac = RTMemAllocZ(sizeof(struct arp_cache_entry));
2011	if (ac == NULL)
2012	{
2013	Log(("NAT: Can't allocate arp cache entry\n"));
2014	return;
2015	}
2016	ac->ip = ip;
2017	memcpy(ac->ether, ether, ETH_ALEN);
2018	LIST_INSERT_HEAD(&pData->arp_cache, ac, list);
2019	}
2020
2021	/* updates or adds entry to the arp cache
2022	* @returns 0 - if has found and updated
2023	* 1 - if hasn't found.
2024	*/
2025	int slirp_arp_cache_update_or_add(PNATState pData, uint32_t dst, const uint8_t *mac)
2026	{
2027	if ( !memcmp(mac, broadcast_ethaddr, ETH_ALEN)
2028	\|\| !memcmp(mac, zerro_ethaddr, ETH_ALEN))
2029	{
2030	static bool fBroadcastEtherAddReported;
2031	if (!fBroadcastEtherAddReported)
2032	{
2033	LogRel(("NAT: Attempt to add pair [%RTmac:%RTnaipv4] in ARP cache was ignored\n",
2034	mac, dst));
2035	fBroadcastEtherAddReported = true;
2036	}
2037	return 1;
2038	}
2039	if (slirp_arp_cache_update(pData, dst, mac))
2040	slirp_arp_cache_add(pData, dst, mac);
2041
2042	return 0;
2043	}
2044
2045
2046	void slirp_set_mtu(PNATState pData, int mtu)
2047	{
2048	if (mtu < 20 \|\| mtu >= 16000)
2049	{
2050	LogRel(("NAT: mtu(%d) is out of range (20;16000] mtu forcely assigned to 1500\n", mtu));
2051	mtu = 1500;
2052	}
2053	/* MTU is maximum transition unit on */
2054	if_mtu =
2055	if_mru = mtu;
2056	}
2057
2058	/**
2059	* Info handler.
2060	*/
2061	void slirp_info(PNATState pData, const void pvArg, const char pszArgs)
2062	{
2063	struct socket so, so_next;
2064	struct arp_cache_entry *ac;
2065	struct port_forward_rule *rule;
2066	PCDBGFINFOHLP pHlp = (PCDBGFINFOHLP)pvArg;
2067	NOREF(pszArgs);
2068
2069	pHlp->pfnPrintf(pHlp, "NAT parameters: MTU=%d\n", if_mtu);
2070	pHlp->pfnPrintf(pHlp, "NAT TCP ports:\n");
2071	QSOCKET_FOREACH(so, so_next, tcp)
2072	/* { */
2073	pHlp->pfnPrintf(pHlp, " %R[natsock]\n", so);
2074	}
2075
2076	pHlp->pfnPrintf(pHlp, "NAT UDP ports:\n");
2077	QSOCKET_FOREACH(so, so_next, udp)
2078	/* { */
2079	pHlp->pfnPrintf(pHlp, " %R[natsock]\n", so);
2080	}
2081
2082	pHlp->pfnPrintf(pHlp, "NAT ARP cache:\n");
2083	LIST_FOREACH(ac, &pData->arp_cache, list)
2084	{
2085	pHlp->pfnPrintf(pHlp, " %RTnaipv4 %RTmac\n", ac->ip, &ac->ether);
2086	}
2087
2088	pHlp->pfnPrintf(pHlp, "NAT rules:\n");
2089	LIST_FOREACH(rule, &pData->port_forward_rule_head, list)
2090	{
2091	pHlp->pfnPrintf(pHlp, " %s %d => %RTnaipv4:%d %c\n",
2092	rule->proto == IPPROTO_UDP ? "UDP" : "TCP",
2093	rule->host_port, rule->guest_addr.s_addr, rule->guest_port,
2094	rule->activated ? ' ' : '*');
2095	}
2096	}
2097
2098	/**
2099	* @note: NATState::fUseHostResolver could be changed in bootp.c::dhcp_decode
2100	* @note: this function is executed on GUI/VirtualBox or main/VBoxHeadless thread.
2101	* @note: this function can potentially race with bootp.c::dhcp_decode (except Darwin)
2102	*/
2103	int slirp_host_network_configuration_change_strategy_selector(const PNATState pData)
2104	{
2105	if (pData->fUseHostResolverPermanent)
2106	return VBOX_NAT_DNS_HOSTRESOLVER;
2107
2108	if (pData->fUseDnsProxy) {
2109	#if HAVE_NOTIFICATION_FOR_DNS_UPDATE /* XXX */ && !defined(RT_OS_WINDOWS)
2110	/* We dont conflict with bootp.c::dhcp_decode */
2111	struct rcp_state rcp_state;
2112	int rc;
2113
2114	rcp_state.rcps_flags \|= RCPSF_IGNORE_IPV6;
2115	rc = rcp_parse(&rcp_state, RESOLV_CONF_FILE);
2116	LogRelFunc(("NAT: rcp_parse:%Rrc old domain:%s new domain:%s\n",
2117	rc, LIST_EMPTY(&pData->pDomainList)
2118	? "(null)"
2119	: LIST_FIRST(&pData->pDomainList)->dd_pszDomain,
2120	rcp_state.rcps_domain));
2121	if ( RT_FAILURE(rc)
2122	\|\| LIST_EMPTY(&pData->pDomainList))
2123	return VBOX_NAT_DNS_DNSPROXY;
2124
2125	if ( rcp_state.rcps_domain
2126	&& strcmp(rcp_state.rcps_domain, LIST_FIRST(&pData->pDomainList)->dd_pszDomain) == 0)
2127	return VBOX_NAT_DNS_DNSPROXY;
2128	else
2129	return VBOX_NAT_DNS_EXTERNAL;
2130	#else
2131	/* copy domain name */
2132	/* domain only compare with coy version */
2133	return VBOX_NAT_DNS_DNSPROXY;
2134	#endif
2135	}
2136	return VBOX_NAT_DNS_EXTERNAL;
2137	}

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

source: vbox/trunk/src/VBox/Devices/Network/slirp/slirp.c@ 57550

Download in other formats: