slirp.c@ 52343

Last change on this file since 52343 was 52283, checked in by vboxsync, 10 years ago
NAT: Do pay attention to POLLERR and WSANETWORKEVENTS::iErrorCode and abort TCP connections on socket errors. That makes RST propagated to the guest properly (tickets 10525, 11696).
Property svn:eol-style set to `native` Property svn:keywords set to `Author Date Id Revision`
File size: 67.5 KB

Line
1	/* $Id: slirp.c 52283 2014-08-05 15:34:32Z 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	WSAStartup(MAKEWORD(2, 0), &Data);
339	}
340	pData->phEvents[VBOX_SOCKET_EVENT_INDEX] = CreateEvent(NULL, FALSE, FALSE, NULL);
341	#endif
342
343	rc = bootp_dhcp_init(pData);
344	if (RT_FAILURE(rc))
345	{
346	Log(("NAT: DHCP server initialization failed\n"));
347	RTMemFree(pData);
348	*ppData = NULL;
349	return rc;
350	}
351	debug_init(pData);
352	if_init(pData);
353	ip_init(pData);
354	icmp_init(pData, iIcmpCacheLimit);
355
356	/* Initialise mbufs after setting the MTU */
357	mbuf_init(pData);
358
359	pData->special_addr.s_addr = u32NetAddr;
360	pData->slirp_ethaddr = &special_ethaddr[0];
361	alias_addr.s_addr = pData->special_addr.s_addr \| RT_H2N_U32_C(CTL_ALIAS);
362	/* @todo: add ability to configure this staff */
363
364	/* set default addresses */
365	inet_aton("127.0.0.1", &loopback_addr);
366
367	rc = slirpTftpInit(pData);
368	AssertRCReturn(rc, VINF_NAT_DNS);
369
370	if (i32AliasMode & ~(PKT_ALIAS_LOG\|PKT_ALIAS_SAME_PORTS\|PKT_ALIAS_PROXY_ONLY))
371	{
372	Log(("NAT: alias mode %x is ignored\n", i32AliasMode));
373	i32AliasMode = 0;
374	}
375	pData->i32AliasMode = i32AliasMode;
376	getouraddr(pData);
377	{
378	int flags = 0;
379	struct in_addr proxy_addr;
380	pData->proxy_alias = LibAliasInit(pData, NULL);
381	if (pData->proxy_alias == NULL)
382	{
383	Log(("NAT: LibAlias default rule wasn't initialized\n"));
384	AssertMsgFailed(("NAT: LibAlias default rule wasn't initialized\n"));
385	}
386	flags = LibAliasSetMode(pData->proxy_alias, 0, 0);
387	#ifndef NO_FW_PUNCH
388	flags \|= PKT_ALIAS_PUNCH_FW;
389	#endif
390	flags \|= pData->i32AliasMode; /* do transparent proxying */
391	flags = LibAliasSetMode(pData->proxy_alias, flags, ~0);
392	proxy_addr.s_addr = RT_H2N_U32(RT_N2H_U32(pData->special_addr.s_addr) \| CTL_ALIAS);
393	LibAliasSetAddress(pData->proxy_alias, proxy_addr);
394	ftp_alias_load(pData);
395	nbt_alias_load(pData);
396	if (pData->fUseHostResolver)
397	dns_alias_load(pData);
398	}
399	#ifdef VBOX_WITH_NAT_SEND2HOME
400	/* @todo: we should know all interfaces available on host. */
401	pData->pInSockAddrHomeAddress = RTMemAllocZ(sizeof(struct sockaddr));
402	pData->cInHomeAddressSize = 1;
403	inet_aton("192.168.1.25", &pData->pInSockAddrHomeAddress[0].sin_addr);
404	pData->pInSockAddrHomeAddress[0].sin_family = AF_INET;
405	# ifdef RT_OS_DARWIN
406	pData->pInSockAddrHomeAddress[0].sin_len = sizeof(struct sockaddr_in);
407	# endif
408	#endif
409
410	slirp_link_up(pData);
411	return VINF_SUCCESS;
412	}
413
414	/**
415	* Register statistics.
416	*/
417	void slirp_register_statistics(PNATState pData, PPDMDRVINS pDrvIns)
418	{
419	#ifdef VBOX_WITH_STATISTICS
420	# define PROFILE_COUNTER(name, dsc) REGISTER_COUNTER(name, pData, STAMTYPE_PROFILE, STAMUNIT_TICKS_PER_CALL, dsc)
421	# define COUNTING_COUNTER(name, dsc) REGISTER_COUNTER(name, pData, STAMTYPE_COUNTER, STAMUNIT_COUNT, dsc)
422	# include "counters.h"
423	# undef COUNTER
424	/** @todo register statistics for the variables dumped by:
425	* ipstats(pData); tcpstats(pData); udpstats(pData); icmpstats(pData);
426	* mbufstats(pData); sockstats(pData); */
427	#else /* VBOX_WITH_STATISTICS */
428	NOREF(pData);
429	NOREF(pDrvIns);
430	#endif /* !VBOX_WITH_STATISTICS */
431	}
432
433	/**
434	* Deregister statistics.
435	*/
436	void slirp_deregister_statistics(PNATState pData, PPDMDRVINS pDrvIns)
437	{
438	if (pData == NULL)
439	return;
440	#ifdef VBOX_WITH_STATISTICS
441	# define PROFILE_COUNTER(name, dsc) DEREGISTER_COUNTER(name, pData)
442	# define COUNTING_COUNTER(name, dsc) DEREGISTER_COUNTER(name, pData)
443	# include "counters.h"
444	#else /* VBOX_WITH_STATISTICS */
445	NOREF(pData);
446	NOREF(pDrvIns);
447	#endif /* !VBOX_WITH_STATISTICS */
448	}
449
450	/**
451	* Marks the link as up, making it possible to establish new connections.
452	*/
453	void slirp_link_up(PNATState pData)
454	{
455	struct arp_cache_entry *ac;
456
457	if (link_up == 1)
458	return;
459
460	link_up = 1;
461
462	if (!pData->fUseHostResolverPermanent)
463	slirpInitializeDnsSettings(pData);
464
465	if (LIST_EMPTY(&pData->arp_cache))
466	return;
467
468	LIST_FOREACH(ac, &pData->arp_cache, list)
469	{
470	activate_port_forwarding(pData, ac->ether);
471	}
472	}
473
474	/**
475	* Marks the link as down and cleans up the current connections.
476	*/
477	void slirp_link_down(PNATState pData)
478	{
479	struct socket *so;
480	struct port_forward_rule *rule;
481
482	if (link_up == 0)
483	return;
484
485	slirpReleaseDnsSettings(pData);
486
487	while ((so = tcb.so_next) != &tcb)
488	{
489	/* Don't miss TCB releasing */
490	if ( !sototcpcb(so)
491	&& ( so->so_state & SS_NOFDREF
492	\|\| so->s == -1))
493	sofree(pData, so);
494	else
495	tcp_close(pData, sototcpcb(so));
496	}
497
498	while ((so = udb.so_next) != &udb)
499	udp_detach(pData, so);
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	if (pData == NULL)
520	return;
521	icmp_finit(pData);
522
523	slirp_link_down(pData);
524	ftp_alias_unload(pData);
525	nbt_alias_unload(pData);
526	if (pData->fUseHostResolver)
527	{
528	dns_alias_unload(pData);
529	#ifdef VBOX_WITH_DNSMAPPING_IN_HOSTRESOLVER
530	while (!LIST_EMPTY(&pData->DNSMapHead))
531	{
532	PDNSMAPPINGENTRY pDnsEntry = LIST_FIRST(&pData->DNSMapHead);
533	LIST_REMOVE(pDnsEntry, MapList);
534	RTStrFree(pDnsEntry->pszCName);
535	RTMemFree(pDnsEntry);
536	}
537	#endif
538	}
539	while (!LIST_EMPTY(&instancehead))
540	{
541	struct libalias *la = LIST_FIRST(&instancehead);
542	/* libalias do all clean up */
543	LibAliasUninit(la);
544	}
545	while (!LIST_EMPTY(&pData->arp_cache))
546	{
547	struct arp_cache_entry *ac = LIST_FIRST(&pData->arp_cache);
548	LIST_REMOVE(ac, list);
549	RTMemFree(ac);
550	}
551	slirpTftpTerm(pData);
552	bootp_dhcp_fini(pData);
553	m_fini(pData);
554	#ifdef RT_OS_WINDOWS
555	WSACleanup();
556	#endif
557	#ifndef VBOX_WITH_SLIRP_BSD_SBUF
558	#ifdef LOG_ENABLED
559	Log(("\n"
560	"NAT statistics\n"
561	"--------------\n"
562	"\n"));
563	ipstats(pData);
564	tcpstats(pData);
565	udpstats(pData);
566	icmpstats(pData);
567	mbufstats(pData);
568	sockstats(pData);
569	Log(("\n"
570	"\n"
571	"\n"));
572	#endif
573	#endif
574	RTCritSectRwDelete(&pData->CsRwHandlerChain);
575	RTMemFree(pData);
576	}
577
578
579	#define CONN_CANFSEND(so) (((so)->so_state & (SS_FCANTSENDMORE\|SS_ISFCONNECTED)) == SS_ISFCONNECTED)
580	#define CONN_CANFRCV(so) (((so)->so_state & (SS_FCANTRCVMORE\|SS_ISFCONNECTED)) == SS_ISFCONNECTED)
581
582	/*
583	* curtime kept to an accuracy of 1ms
584	*/
585	static void updtime(PNATState pData)
586	{
587	#ifdef RT_OS_WINDOWS
588	struct _timeb tb;
589
590	_ftime(&tb);
591	curtime = (u_int)tb.time * (u_int)1000;
592	curtime += (u_int)tb.millitm;
593	#else
594	gettimeofday(&tt, 0);
595
596	curtime = (u_int)tt.tv_sec * (u_int)1000;
597	curtime += (u_int)tt.tv_usec / (u_int)1000;
598
599	if ((tt.tv_usec % 1000) >= 500)
600	curtime++;
601	#endif
602	}
603
604	#ifdef RT_OS_WINDOWS
605	void slirp_select_fill(PNATState pData, int *pnfds)
606	#else /* RT_OS_WINDOWS */
607	void slirp_select_fill(PNATState pData, int pnfds, struct pollfd polls)
608	#endif /* !RT_OS_WINDOWS */
609	{
610	struct socket so, so_next;
611	int nfds;
612	#if defined(RT_OS_WINDOWS)
613	int rc;
614	int error;
615	#else
616	int poll_index = 0;
617	#endif
618	int i;
619
620	STAM_PROFILE_START(&pData->StatFill, a);
621
622	nfds = *pnfds;
623
624	/*
625	* First, TCP sockets
626	*/
627	do_slowtimo = 0;
628	if (!link_up)
629	goto done;
630
631	/*
632	* *_slowtimo needs calling if there are IP fragments
633	* in the fragment queue, or there are TCP connections active
634	*/
635	/* XXX:
636	* triggering of fragment expiration should be the same but use new macroses
637	*/
638	do_slowtimo = (tcb.so_next != &tcb);
639	if (!do_slowtimo)
640	{
641	for (i = 0; i < IPREASS_NHASH; i++)
642	{
643	if (!TAILQ_EMPTY(&ipq[i]))
644	{
645	do_slowtimo = 1;
646	break;
647	}
648	}
649	}
650	/* always add the ICMP socket */
651	#ifndef RT_OS_WINDOWS
652	pData->icmp_socket.so_poll_index = -1;
653	#endif
654	ICMP_ENGAGE_EVENT(&pData->icmp_socket, readfds);
655
656	STAM_COUNTER_RESET(&pData->StatTCP);
657	STAM_COUNTER_RESET(&pData->StatTCPHot);
658
659	QSOCKET_FOREACH(so, so_next, tcp)
660	/* { */
661	Assert(so->so_type == IPPROTO_TCP);
662	#if !defined(RT_OS_WINDOWS)
663	so->so_poll_index = -1;
664	#endif
665	STAM_COUNTER_INC(&pData->StatTCP);
666	#ifdef VBOX_WITH_NAT_UDP_SOCKET_CLONE
667	/* TCP socket can't be cloned */
668	Assert((!so->so_cloneOf));
669	#endif
670	/*
671	* See if we need a tcp_fasttimo
672	*/
673	if ( time_fasttimo == 0
674	&& so->so_tcpcb != NULL
675	&& so->so_tcpcb->t_flags & TF_DELACK)
676	{
677	time_fasttimo = curtime; /* Flag when we want a fasttimo */
678	}
679
680	/*
681	* NOFDREF can include still connecting to local-host,
682	* newly socreated() sockets etc. Don't want to select these.
683	*/
684	if (so->so_state & SS_NOFDREF \|\| so->s == -1)
685	CONTINUE(tcp);
686
687	/*
688	* Set for reading sockets which are accepting
689	*/
690	if (so->so_state & SS_FACCEPTCONN)
691	{
692	STAM_COUNTER_INC(&pData->StatTCPHot);
693	TCP_ENGAGE_EVENT1(so, readfds);
694	CONTINUE(tcp);
695	}
696
697	/*
698	* Set for writing sockets which are connecting
699	*/
700	if (so->so_state & SS_ISFCONNECTING)
701	{
702	Log2(("connecting %R[natsock] engaged\n",so));
703	STAM_COUNTER_INC(&pData->StatTCPHot);
704	#ifdef RT_OS_WINDOWS
705	WIN_TCP_ENGAGE_EVENT2(so, writefds, connectfds);
706	#else
707	TCP_ENGAGE_EVENT1(so, writefds);
708	#endif
709	}
710
711	/*
712	* Set for writing if we are connected, can send more, and
713	* we have something to send
714	*/
715	if (CONN_CANFSEND(so) && SBUF_LEN(&so->so_rcv))
716	{
717	STAM_COUNTER_INC(&pData->StatTCPHot);
718	TCP_ENGAGE_EVENT1(so, writefds);
719	}
720
721	/*
722	* Set for reading (and urgent data) if we are connected, can
723	* receive more, and we have room for it XXX /2 ?
724	*/
725	/* @todo: vvl - check which predicat here will be more useful here in rerm of new sbufs. */
726	if ( CONN_CANFRCV(so)
727	&& (SBUF_LEN(&so->so_snd) < (SBUF_SIZE(&so->so_snd)/2))
728	#ifdef RT_OS_WINDOWS
729	&& !(so->so_state & SS_ISFCONNECTING)
730	#endif
731	)
732	{
733	STAM_COUNTER_INC(&pData->StatTCPHot);
734	TCP_ENGAGE_EVENT2(so, readfds, xfds);
735	}
736	LOOP_LABEL(tcp, so, so_next);
737	}
738
739	/*
740	* UDP sockets
741	*/
742	STAM_COUNTER_RESET(&pData->StatUDP);
743	STAM_COUNTER_RESET(&pData->StatUDPHot);
744
745	QSOCKET_FOREACH(so, so_next, udp)
746	/* { */
747
748	Assert(so->so_type == IPPROTO_UDP);
749	STAM_COUNTER_INC(&pData->StatUDP);
750	#if !defined(RT_OS_WINDOWS)
751	so->so_poll_index = -1;
752	#endif
753
754	/*
755	* See if it's timed out
756	*/
757	if (so->so_expire)
758	{
759	if (so->so_expire <= curtime)
760	{
761	Log2(("NAT: %R[natsock] expired\n", so));
762	if (so->so_timeout != NULL)
763	{
764	/* so_timeout - might change the so_expire value or
765	* drop so_timeout* from so.
766	*/
767	so->so_timeout(pData, so, so->so_timeout_arg);
768	/* on 4.2 so->
769	*/
770	if ( so_next->so_prev != so /* so_timeout freed the socket */
771	\|\| so->so_timeout) /* so_timeout just freed so_timeout */
772	CONTINUE_NO_UNLOCK(udp);
773	}
774	UDP_DETACH(pData, so, so_next);
775	CONTINUE_NO_UNLOCK(udp);
776	}
777	}
778	#ifdef VBOX_WITH_NAT_UDP_SOCKET_CLONE
779	if (so->so_cloneOf)
780	CONTINUE_NO_UNLOCK(udp);
781	#endif
782
783	/*
784	* When UDP packets are received from over the link, they're
785	* sendto()'d straight away, so no need for setting for writing
786	* Limit the number of packets queued by this session to 4.
787	* Note that even though we try and limit this to 4 packets,
788	* the session could have more queued if the packets needed
789	* to be fragmented.
790	*
791	* (XXX <= 4 ?)
792	*/
793	if ((so->so_state & SS_ISFCONNECTED) && so->so_queued <= 4)
794	{
795	STAM_COUNTER_INC(&pData->StatUDPHot);
796	UDP_ENGAGE_EVENT(so, readfds);
797	}
798	LOOP_LABEL(udp, so, so_next);
799	}
800	done:
801
802	#if defined(RT_OS_WINDOWS)
803	*pnfds = VBOX_EVENT_COUNT;
804	#else /* RT_OS_WINDOWS */
805	AssertRelease(poll_index <= *pnfds);
806	*pnfds = poll_index;
807	#endif /* !RT_OS_WINDOWS */
808
809	STAM_PROFILE_STOP(&pData->StatFill, a);
810	}
811
812
813	/**
814	* This function do Connection or sending tcp sequence to.
815	* @returns if true operation completed
816	* @note: functions call tcp_input that potentially could lead to tcp_drop
817	*/
818	static bool slirpConnectOrWrite(PNATState pData, struct socket *so, bool fConnectOnly)
819	{
820	int ret;
821	LogFlowFunc(("ENTER: so:%R[natsock], fConnectOnly:%RTbool\n", so, fConnectOnly));
822	/*
823	* Check for non-blocking, still-connecting sockets
824	*/
825	if (so->so_state & SS_ISFCONNECTING)
826	{
827	Log2(("connecting %R[natsock] catched\n", so));
828	/* Connected */
829	so->so_state &= ~SS_ISFCONNECTING;
830
831	/*
832	* This should be probably guarded by PROBE_CONN too. Anyway,
833	* we disable it on OS/2 because the below send call returns
834	* EFAULT which causes the opened TCP socket to close right
835	* after it has been opened and connected.
836	*/
837	#ifndef RT_OS_OS2
838	ret = send(so->s, (const char *)&ret, 0, 0);
839	if (ret < 0)
840	{
841	/* XXXXX Must fix, zero bytes is a NOP */
842	if ( soIgnorableErrorCode(errno)
843	\|\| errno == ENOTCONN)
844	{
845	LogFlowFunc(("LEAVE: false\n"));
846	return false;
847	}
848
849	/* else failed */
850	so->so_state = SS_NOFDREF;
851	}
852	/* else so->so_state &= ~SS_ISFCONNECTING; */
853	#endif
854
855	/*
856	* Continue tcp_input
857	*/
858	TCP_INPUT(pData, (struct mbuf *)NULL, sizeof(struct ip), so);
859	/* continue; */
860	}
861	else if (!fConnectOnly)
862	{
863	SOWRITE(ret, pData, so);
864	if (RT_LIKELY(ret > 0))
865	{
866	/*
867	* Make sure we will send window update to peer. This is
868	* a moral equivalent of calling tcp_output() for PRU_RCVD
869	* in tcp_usrreq() of the real stack.
870	*/
871	struct tcpcb *tp = sototcpcb(so);
872	if (RT_LIKELY(tp != NULL))
873	tp->t_flags \|= TF_DELACK;
874	}
875	}
876
877	LogFlowFunc(("LEAVE: true\n"));
878	return true;
879	}
880
881	#if defined(RT_OS_WINDOWS)
882	void slirp_select_poll(PNATState pData, int fTimeout, int fIcmp)
883	#else /* RT_OS_WINDOWS */
884	void slirp_select_poll(PNATState pData, struct pollfd *polls, int ndfs)
885	#endif /* !RT_OS_WINDOWS */
886	{
887	struct socket so, so_next;
888	int ret;
889	#if defined(RT_OS_WINDOWS)
890	WSANETWORKEVENTS NetworkEvents;
891	int rc;
892	int error;
893	#endif
894
895	STAM_PROFILE_START(&pData->StatPoll, a);
896
897	/* Update time */
898	updtime(pData);
899
900	/*
901	* See if anything has timed out
902	*/
903	if (link_up)
904	{
905	if (time_fasttimo && ((curtime - time_fasttimo) >= 2))
906	{
907	STAM_PROFILE_START(&pData->StatFastTimer, b);
908	tcp_fasttimo(pData);
909	time_fasttimo = 0;
910	STAM_PROFILE_STOP(&pData->StatFastTimer, b);
911	}
912	if (do_slowtimo && ((curtime - last_slowtimo) >= 499))
913	{
914	STAM_PROFILE_START(&pData->StatSlowTimer, c);
915	ip_slowtimo(pData);
916	tcp_slowtimo(pData);
917	last_slowtimo = curtime;
918	STAM_PROFILE_STOP(&pData->StatSlowTimer, c);
919	}
920	}
921	#if defined(RT_OS_WINDOWS)
922	if (fTimeout)
923	return; /* only timer update */
924	#endif
925
926	/*
927	* Check sockets
928	*/
929	if (!link_up)
930	goto done;
931	#if defined(RT_OS_WINDOWS)
932	/*XXX: before renaming please make see define
933	* fIcmp in slirp_state.h
934	*/
935	if (fIcmp)
936	sorecvfrom(pData, &pData->icmp_socket);
937	#else
938	if ( (pData->icmp_socket.s != -1)
939	&& CHECK_FD_SET(&pData->icmp_socket, ignored, readfds))
940	sorecvfrom(pData, &pData->icmp_socket);
941	#endif
942	/*
943	* Check TCP sockets
944	*/
945	QSOCKET_FOREACH(so, so_next, tcp)
946	/* { */
947	/* TCP socket can't be cloned */
948	#ifdef VBOX_WITH_NAT_UDP_SOCKET_CLONE
949	Assert((!so->so_cloneOf));
950	#endif
951	Assert(!so->fUnderPolling);
952	so->fUnderPolling = 1;
953	if (slirpVerifyAndFreeSocket(pData, so))
954	CONTINUE(tcp);
955	/*
956	* FD_ISSET is meaningless on these sockets
957	* (and they can crash the program)
958	*/
959	if (so->so_state & SS_NOFDREF \|\| so->s == -1)
960	{
961	so->fUnderPolling = 0;
962	CONTINUE(tcp);
963	}
964
965	POLL_TCP_EVENTS(rc, error, so, &NetworkEvents);
966
967	LOG_NAT_SOCK(so, TCP, &NetworkEvents, readfds, writefds, xfds);
968
969	/*
970	* It's hard to hide error condition check behind existing
971	* macros without uglifying them further, so open-code it here
972	* instead.
973	*/
974	{
975	int pollerr = 0;
976	#if defined(RT_OS_WINDOWS)
977	{
978	int bit;
979	for (bit = 0; bit <= FD_CLOSE_BIT; ++bit)
980	{
981	if ( NetworkEvents.lNetworkEvents & (1 << bit)
982	&& NetworkEvents.iErrorCode[bit] != 0)
983	{
984	pollerr = NetworkEvents.iErrorCode[bit];
985	break;
986	}
987	}
988	}
989	#else
990	if (CHECK_FD_SET(so, NetworkEvents, rderr))
991	{
992	socklen_t optlen = (socklen_t)sizeof(int);
993	ret = getsockopt(so->s, SOL_SOCKET, SO_ERROR, &pollerr, &optlen);
994	if (ret < 0)
995	pollerr = ENETDOWN;
996	}
997	#endif
998	if (pollerr)
999	{
1000	if (so->so_state & SS_ISFCONNECTING)
1001	{
1002	/* "continue" tcp_input() to reject connection from guest */
1003	so->so_state = SS_NOFDREF;
1004	TCP_INPUT(pData, NULL, 0, so);
1005	}
1006	else
1007	{
1008	tcp_drop(pData, sototcpcb(so), pollerr);
1009	}
1010	ret = slirpVerifyAndFreeSocket(pData, so);
1011	Assert(ret == 1); /* freed */
1012	CONTINUE(tcp);
1013	}
1014	}
1015
1016	/*
1017	* Check for URG data
1018	* This will soread as well, so no need to
1019	* test for readfds below if this succeeds
1020	*/
1021
1022	/* out-of-band data */
1023	if ( CHECK_FD_SET(so, NetworkEvents, xfds)
1024	#ifdef RT_OS_DARWIN
1025	/* Darwin and probably BSD hosts generates POLLPRI\|POLLHUP event on receiving TCP.flags.{ACK\|URG\|FIN} this
1026	* combination on other Unixs hosts doesn't enter to this branch
1027	*/
1028	&& !CHECK_FD_SET(so, NetworkEvents, closefds)
1029	#endif
1030	#ifdef RT_OS_WINDOWS
1031	/**
1032	* In some cases FD_CLOSE comes with FD_OOB, that confuse tcp processing.
1033	*/
1034	&& !WIN_CHECK_FD_SET(so, NetworkEvents, closefds)
1035	#endif
1036	)
1037	{
1038	sorecvoob(pData, so);
1039	if (slirpVerifyAndFreeSocket(pData, so))
1040	CONTINUE(tcp);
1041	}
1042
1043	/*
1044	* Check sockets for reading
1045	*/
1046	else if ( CHECK_FD_SET(so, NetworkEvents, readfds)
1047	\|\| WIN_CHECK_FD_SET(so, NetworkEvents, acceptds))
1048	{
1049
1050	#ifdef RT_OS_WINDOWS
1051	if (WIN_CHECK_FD_SET(so, NetworkEvents, connectfds))
1052	{
1053	/* Finish connection first */
1054	/* should we ignore return value? */
1055	bool fRet = slirpConnectOrWrite(pData, so, true);
1056	LogFunc(("fRet:%RTbool\n", fRet));
1057	if (slirpVerifyAndFreeSocket(pData, so))
1058	CONTINUE(tcp);
1059	}
1060	#endif
1061	/*
1062	* Check for incoming connections
1063	*/
1064	if (so->so_state & SS_FACCEPTCONN)
1065	{
1066	TCP_CONNECT(pData, so);
1067	if (slirpVerifyAndFreeSocket(pData, so))
1068	CONTINUE(tcp);
1069	if (!CHECK_FD_SET(so, NetworkEvents, closefds))
1070	{
1071	so->fUnderPolling = 0;
1072	CONTINUE(tcp);
1073	}
1074	}
1075
1076	ret = soread(pData, so);
1077	if (slirpVerifyAndFreeSocket(pData, so))
1078	CONTINUE(tcp);
1079	/* Output it if we read something */
1080	if (RT_LIKELY(ret > 0))
1081	TCP_OUTPUT(pData, sototcpcb(so));
1082
1083	if (slirpVerifyAndFreeSocket(pData, so))
1084	CONTINUE(tcp);
1085	}
1086
1087	/*
1088	* Check for FD_CLOSE events.
1089	* in some cases once FD_CLOSE engaged on socket it could be flashed latter (for some reasons)
1090	*/
1091	if ( CHECK_FD_SET(so, NetworkEvents, closefds)
1092	\|\| (so->so_close == 1))
1093	{
1094	/*
1095	* drain the socket
1096	*/
1097	for (; so_next->so_prev == so
1098	&& !slirpVerifyAndFreeSocket(pData, so);)
1099	{
1100	ret = soread(pData, so);
1101	if (slirpVerifyAndFreeSocket(pData, so))
1102	break;
1103
1104	if (ret > 0)
1105	TCP_OUTPUT(pData, sototcpcb(so));
1106	else if (so_next->so_prev == so)
1107	{
1108	Log2(("%R[natsock] errno %d (%s)\n", so, errno, strerror(errno)));
1109	break;
1110	}
1111	}
1112
1113	/* if socket freed ''so'' is PHANTOM and next socket isn't points on it */
1114	if (so_next->so_prev == so)
1115	{
1116	/* mark the socket for termination _after_ it was drained */
1117	so->so_close = 1;
1118	}
1119	if (so_next->so_prev == so)
1120	so->fUnderPolling = 0;
1121	CONTINUE(tcp);
1122	}
1123
1124	/*
1125	* Check sockets for writing
1126	*/
1127	if ( CHECK_FD_SET(so, NetworkEvents, writefds)
1128	#ifdef RT_OS_WINDOWS
1129	\|\| WIN_CHECK_FD_SET(so, NetworkEvents, connectfds)
1130	#endif
1131	)
1132	{
1133	int fConnectOrWriteSuccess = slirpConnectOrWrite(pData, so, false);
1134	/* slirpConnectOrWrite could return true even if tcp_input called tcp_drop,
1135	* so we should be ready to such situations.
1136	*/
1137	if (slirpVerifyAndFreeSocket(pData, so))
1138	CONTINUE(tcp);
1139	else if (!fConnectOrWriteSuccess)
1140	{
1141	so->fUnderPolling = 0;
1142	CONTINUE(tcp);
1143	}
1144	/* slirpConnectionOrWrite succeeded and socket wasn't dropped */
1145	}
1146
1147	/*
1148	* Probe a still-connecting, non-blocking socket
1149	* to check if it's still alive
1150	*/
1151	#ifdef PROBE_CONN
1152	if (so->so_state & SS_ISFCONNECTING)
1153	{
1154	ret = recv(so->s, (char *)&ret, 0, 0);
1155
1156	if (ret < 0)
1157	{
1158	/* XXX */
1159	if ( soIgnorableErrorCode(errno)
1160	\|\| errno == ENOTCONN)
1161	{
1162	CONTINUE(tcp); /* Still connecting, continue */
1163	}
1164
1165	/* else failed */
1166	so->so_state = SS_NOFDREF;
1167
1168	/* tcp_input will take care of it */
1169	}
1170	else
1171	{
1172	ret = send(so->s, &ret, 0, 0);
1173	if (ret < 0)
1174	{
1175	/* XXX */
1176	if ( soIgnorableErrorCode(errno)
1177	\|\| errno == ENOTCONN)
1178	{
1179	CONTINUE(tcp);
1180	}
1181	/* else failed */
1182	so->so_state = SS_NOFDREF;
1183	}
1184	else
1185	so->so_state &= ~SS_ISFCONNECTING;
1186
1187	}
1188	TCP_INPUT((struct mbuf *)NULL, sizeof(struct ip),so);
1189	} /* SS_ISFCONNECTING */
1190	#endif
1191	if (!slirpVerifyAndFreeSocket(pData, so))
1192	so->fUnderPolling = 0;
1193	LOOP_LABEL(tcp, so, so_next);
1194	}
1195
1196	/*
1197	* Now UDP sockets.
1198	* Incoming packets are sent straight away, they're not buffered.
1199	* Incoming UDP data isn't buffered either.
1200	*/
1201	QSOCKET_FOREACH(so, so_next, udp)
1202	/* { */
1203	#ifdef VBOX_WITH_NAT_UDP_SOCKET_CLONE
1204	if (so->so_cloneOf)
1205	CONTINUE_NO_UNLOCK(udp);
1206	#endif
1207	#if 0
1208	so->fUnderPolling = 1;
1209	if(slirpVerifyAndFreeSocket(pData, so));
1210	CONTINUE(udp);
1211	so->fUnderPolling = 0;
1212	#endif
1213
1214	POLL_UDP_EVENTS(rc, error, so, &NetworkEvents);
1215
1216	LOG_NAT_SOCK(so, UDP, &NetworkEvents, readfds, writefds, xfds);
1217
1218	if (so->s != -1 && CHECK_FD_SET(so, NetworkEvents, readfds))
1219	{
1220	SORECVFROM(pData, so);
1221	}
1222	LOOP_LABEL(udp, so, so_next);
1223	}
1224
1225	done:
1226
1227	STAM_PROFILE_STOP(&pData->StatPoll, a);
1228	}
1229
1230
1231	struct arphdr
1232	{
1233	unsigned short ar_hrd; /* format of hardware address */
1234	unsigned short ar_pro; /* format of protocol address */
1235	unsigned char ar_hln; /* length of hardware address */
1236	unsigned char ar_pln; /* length of protocol address */
1237	unsigned short ar_op; /* ARP opcode (command) */
1238
1239	/*
1240	* Ethernet looks like this : This bit is variable sized however...
1241	*/
1242	unsigned char ar_sha[ETH_ALEN]; /* sender hardware address */
1243	unsigned char ar_sip[4]; /* sender IP address */
1244	unsigned char ar_tha[ETH_ALEN]; /* target hardware address */
1245	unsigned char ar_tip[4]; /* target IP address */
1246	};
1247	AssertCompileSize(struct arphdr, 28);
1248
1249	static void arp_output(PNATState pData, const uint8_t pcu8EtherSource, const struct arphdr pcARPHeaderSource, uint32_t ip4TargetAddress)
1250	{
1251	struct ethhdr *pEtherHeaderResponse;
1252	struct arphdr *pARPHeaderResponse;
1253	uint32_t ip4TargetAddressInHostFormat;
1254	struct mbuf *pMbufResponse;
1255
1256	Assert((pcu8EtherSource));
1257	if (!pcu8EtherSource)
1258	return;
1259	ip4TargetAddressInHostFormat = RT_N2H_U32(ip4TargetAddress);
1260
1261	pMbufResponse = m_getcl(pData, M_NOWAIT, MT_HEADER, M_PKTHDR);
1262	if (!pMbufResponse)
1263	return;
1264	pEtherHeaderResponse = mtod(pMbufResponse, struct ethhdr *);
1265	/* @note: if_encap will swap src and dst*/
1266	memcpy(pEtherHeaderResponse->h_source, pcu8EtherSource, ETH_ALEN);
1267	pMbufResponse->m_data += ETH_HLEN;
1268	pARPHeaderResponse = mtod(pMbufResponse, struct arphdr *);
1269	pMbufResponse->m_len = sizeof(struct arphdr);
1270
1271	pARPHeaderResponse->ar_hrd = RT_H2N_U16_C(1);
1272	pARPHeaderResponse->ar_pro = RT_H2N_U16_C(ETH_P_IP);
1273	pARPHeaderResponse->ar_hln = ETH_ALEN;
1274	pARPHeaderResponse->ar_pln = 4;
1275	pARPHeaderResponse->ar_op = RT_H2N_U16_C(ARPOP_REPLY);
1276	memcpy(pARPHeaderResponse->ar_sha, special_ethaddr, ETH_ALEN);
1277
1278	if (!slirpMbufTagService(pData, pMbufResponse, (uint8_t)(ip4TargetAddressInHostFormat & ~pData->netmask)))
1279	{
1280	static bool fTagErrorReported;
1281	if (!fTagErrorReported)
1282	{
1283	LogRel(("NAT: couldn't add the tag(PACKET_SERVICE:%d)\n",
1284	(uint8_t)(ip4TargetAddressInHostFormat & ~pData->netmask)));
1285	fTagErrorReported = true;
1286	}
1287	}
1288	pARPHeaderResponse->ar_sha[5] = (uint8_t)(ip4TargetAddressInHostFormat & ~pData->netmask);
1289
1290	memcpy(pARPHeaderResponse->ar_sip, pcARPHeaderSource->ar_tip, 4);
1291	memcpy(pARPHeaderResponse->ar_tha, pcARPHeaderSource->ar_sha, ETH_ALEN);
1292	memcpy(pARPHeaderResponse->ar_tip, pcARPHeaderSource->ar_sip, 4);
1293	if_encap(pData, ETH_P_ARP, pMbufResponse, ETH_ENCAP_URG);
1294	}
1295	/**
1296	* @note This function will free m!
1297	*/
1298	static void arp_input(PNATState pData, struct mbuf *m)
1299	{
1300	struct ethhdr *pEtherHeader;
1301	struct arphdr *pARPHeader;
1302	uint32_t ip4TargetAddress;
1303
1304	int ar_op;
1305	pEtherHeader = mtod(m, struct ethhdr *);
1306	pARPHeader = (struct arphdr *)&pEtherHeader[1];
1307
1308	ar_op = RT_N2H_U16(pARPHeader->ar_op);
1309	ip4TargetAddress = (uint32_t)pARPHeader->ar_tip;
1310
1311	switch (ar_op)
1312	{
1313	case ARPOP_REQUEST:
1314	if ( CTL_CHECK(ip4TargetAddress, CTL_DNS)
1315	\|\| CTL_CHECK(ip4TargetAddress, CTL_ALIAS)
1316	\|\| CTL_CHECK(ip4TargetAddress, CTL_TFTP))
1317	arp_output(pData, pEtherHeader->h_source, pARPHeader, ip4TargetAddress);
1318
1319	/* Gratuitous ARP */
1320	if ( (uint32_t )pARPHeader->ar_sip == (uint32_t )pARPHeader->ar_tip
1321	&& memcmp(pARPHeader->ar_tha, broadcast_ethaddr, ETH_ALEN) == 0
1322	&& memcmp(pEtherHeader->h_dest, broadcast_ethaddr, ETH_ALEN) == 0)
1323	{
1324	/* We've received an announce about address assignment,
1325	* let's do an ARP cache update
1326	*/
1327	static bool fGratuitousArpReported;
1328	if (!fGratuitousArpReported)
1329	{
1330	LogRel(("NAT: Gratuitous ARP [IP:%RTnaipv4, ether:%RTmac]\n",
1331	pARPHeader->ar_sip, pARPHeader->ar_sha));
1332	fGratuitousArpReported = true;
1333	}
1334	slirp_arp_cache_update_or_add(pData, (uint32_t )pARPHeader->ar_sip, &pARPHeader->ar_sha[0]);
1335	}
1336	break;
1337
1338	case ARPOP_REPLY:
1339	slirp_arp_cache_update_or_add(pData, (uint32_t )pARPHeader->ar_sip, &pARPHeader->ar_sha[0]);
1340	break;
1341
1342	default:
1343	break;
1344	}
1345
1346	m_freem(pData, m);
1347	}
1348
1349	/**
1350	* Feed a packet into the slirp engine.
1351	*
1352	* @param m Data buffer, m_len is not valid.
1353	* @param cbBuf The length of the data in m.
1354	*/
1355	void slirp_input(PNATState pData, struct mbuf *m, size_t cbBuf)
1356	{
1357	int proto;
1358	static bool fWarnedIpv6;
1359	struct ethhdr *eh;
1360	uint8_t au8Ether[ETH_ALEN];
1361
1362	m->m_len = cbBuf;
1363	if (cbBuf < ETH_HLEN)
1364	{
1365	Log(("NAT: packet having size %d has been ignored\n", m->m_len));
1366	m_freem(pData, m);
1367	return;
1368	}
1369	eh = mtod(m, struct ethhdr *);
1370	proto = RT_N2H_U16(eh->h_proto);
1371
1372	memcpy(au8Ether, eh->h_source, ETH_ALEN);
1373
1374	switch(proto)
1375	{
1376	case ETH_P_ARP:
1377	arp_input(pData, m);
1378	break;
1379
1380	case ETH_P_IP:
1381	/* Update time. Important if the network is very quiet, as otherwise
1382	* the first outgoing connection gets an incorrect timestamp. */
1383	updtime(pData);
1384	m_adj(m, ETH_HLEN);
1385	M_ASSERTPKTHDR(m);
1386	m->m_pkthdr.header = mtod(m, void *);
1387	ip_input(pData, m);
1388	break;
1389
1390	case ETH_P_IPV6:
1391	m_freem(pData, m);
1392	if (!fWarnedIpv6)
1393	{
1394	LogRel(("NAT: IPv6 not supported\n"));
1395	fWarnedIpv6 = true;
1396	}
1397	break;
1398
1399	default:
1400	Log(("NAT: Unsupported protocol %x\n", proto));
1401	m_freem(pData, m);
1402	break;
1403	}
1404
1405	if (pData->cRedirectionsActive != pData->cRedirectionsStored)
1406	activate_port_forwarding(pData, au8Ether);
1407	}
1408
1409	/**
1410	* Output the IP packet to the ethernet device.
1411	*
1412	* @note This function will free m!
1413	*/
1414	void if_encap(PNATState pData, uint16_t eth_proto, struct mbuf *m, int flags)
1415	{
1416	struct ethhdr *eh;
1417	uint8_t *mbuf = NULL;
1418	size_t mlen = 0;
1419	STAM_PROFILE_START(&pData->StatIF_encap, a);
1420	LogFlowFunc(("ENTER: pData:%p, eth_proto:%RX16, m:%p, flags:%d\n",
1421	pData, eth_proto, m, flags));
1422
1423	M_ASSERTPKTHDR(m);
1424	m->m_data -= ETH_HLEN;
1425	m->m_len += ETH_HLEN;
1426	eh = mtod(m, struct ethhdr *);
1427	mlen = m->m_len;
1428
1429	if (memcmp(eh->h_source, special_ethaddr, ETH_ALEN) != 0)
1430	{
1431	struct m_tag *t = m_tag_first(m);
1432	uint8_t u8ServiceId = CTL_ALIAS;
1433	memcpy(eh->h_dest, eh->h_source, ETH_ALEN);
1434	memcpy(eh->h_source, special_ethaddr, ETH_ALEN);
1435	Assert(memcmp(eh->h_dest, special_ethaddr, ETH_ALEN) != 0);
1436	if (memcmp(eh->h_dest, zerro_ethaddr, ETH_ALEN) == 0)
1437	{
1438	/* don't do anything */
1439	m_freem(pData, m);
1440	goto done;
1441	}
1442	if ( t
1443	&& (t = m_tag_find(m, PACKET_SERVICE, NULL)))
1444	{
1445	Assert(t);
1446	u8ServiceId = (uint8_t )&t[1];
1447	}
1448	eh->h_source[5] = u8ServiceId;
1449	}
1450	/*
1451	* we're processing the chain, that isn't not expected.
1452	*/
1453	Assert((!m->m_next));
1454	if (m->m_next)
1455	{
1456	Log(("NAT: if_encap's recived the chain, dropping...\n"));
1457	m_freem(pData, m);
1458	goto done;
1459	}
1460	mbuf = mtod(m, uint8_t *);
1461	eh->h_proto = RT_H2N_U16(eth_proto);
1462	LogFunc(("eh(dst:%RTmac, src:%RTmac)\n", eh->h_dest, eh->h_source));
1463	if (flags & ETH_ENCAP_URG)
1464	slirp_urg_output(pData->pvUser, m, mbuf, mlen);
1465	else
1466	slirp_output(pData->pvUser, m, mbuf, mlen);
1467	done:
1468	STAM_PROFILE_STOP(&pData->StatIF_encap, a);
1469	LogFlowFuncLeave();
1470	}
1471
1472	/**
1473	* Still we're using dhcp server leasing to map ether to IP
1474	* @todo see rt_lookup_in_cache
1475	*/
1476	static uint32_t find_guest_ip(PNATState pData, const uint8_t *eth_addr)
1477	{
1478	uint32_t ip = INADDR_ANY;
1479	int rc;
1480
1481	if (eth_addr == NULL)
1482	return INADDR_ANY;
1483
1484	if ( memcmp(eth_addr, zerro_ethaddr, ETH_ALEN) == 0
1485	\|\| memcmp(eth_addr, broadcast_ethaddr, ETH_ALEN) == 0)
1486	return INADDR_ANY;
1487
1488	rc = slirp_arp_lookup_ip_by_ether(pData, eth_addr, &ip);
1489	if (RT_SUCCESS(rc))
1490	return ip;
1491
1492	bootp_cache_lookup_ip_by_ether(pData, eth_addr, &ip);
1493	/* ignore return code, ip will be set to INADDR_ANY on error */
1494	return ip;
1495	}
1496
1497	/**
1498	* We need check if we've activated port forwarding
1499	* for specific machine ... that of course relates to
1500	* service mode
1501	* @todo finish this for service case
1502	*/
1503	static void activate_port_forwarding(PNATState pData, const uint8_t *h_source)
1504	{
1505	struct port_forward_rule rule, tmp;
1506	const uint8_t *pu8EthSource = h_source;
1507
1508	/* check mac here */
1509	LIST_FOREACH_SAFE(rule, &pData->port_forward_rule_head, list, tmp)
1510	{
1511	struct socket *so;
1512	struct alias_link *alias_link;
1513	struct libalias *lib;
1514	int flags;
1515	struct sockaddr sa;
1516	struct sockaddr_in *psin;
1517	socklen_t socketlen;
1518	struct in_addr alias;
1519	int rc;
1520	uint32_t guest_addr; /* need to understand if we already give address to guest */
1521
1522	if (rule->activated)
1523	continue;
1524
1525	#ifdef VBOX_WITH_NAT_SERVICE
1526	/**
1527	* case when guest ip is INADDR_ANY shouldn't appear in NAT service
1528	*/
1529	Assert((rule->guest_addr.s_addr != INADDR_ANY));
1530	guest_addr = rule->guest_addr.s_addr;
1531	#else /* VBOX_WITH_NAT_SERVICE */
1532	guest_addr = find_guest_ip(pData, pu8EthSource);
1533	#endif /* !VBOX_WITH_NAT_SERVICE */
1534	if (guest_addr == INADDR_ANY)
1535	{
1536	/* the address wasn't granted */
1537	return;
1538	}
1539
1540	#if !defined(VBOX_WITH_NAT_SERVICE)
1541	if ( rule->guest_addr.s_addr != guest_addr
1542	&& rule->guest_addr.s_addr != INADDR_ANY)
1543	continue;
1544	if (rule->guest_addr.s_addr == INADDR_ANY)
1545	rule->guest_addr.s_addr = guest_addr;
1546	#endif
1547
1548	LogRel(("NAT: set redirect %s host %RTnaipv4:%d => guest %RTnaipv4:%d\n",
1549	rule->proto == IPPROTO_UDP ? "UDP" : "TCP",
1550	rule->bind_ip.s_addr, rule->host_port,
1551	guest_addr, rule->guest_port));
1552
1553	if (rule->proto == IPPROTO_UDP)
1554	so = udp_listen(pData, rule->bind_ip.s_addr, RT_H2N_U16(rule->host_port), guest_addr,
1555	RT_H2N_U16(rule->guest_port), 0);
1556	else
1557	so = solisten(pData, rule->bind_ip.s_addr, RT_H2N_U16(rule->host_port), guest_addr,
1558	RT_H2N_U16(rule->guest_port), 0);
1559
1560	if (so == NULL)
1561	goto remove_port_forwarding;
1562
1563	psin = (struct sockaddr_in *)&sa;
1564	psin->sin_family = AF_INET;
1565	psin->sin_port = 0;
1566	psin->sin_addr.s_addr = INADDR_ANY;
1567	socketlen = sizeof(struct sockaddr);
1568
1569	rc = getsockname(so->s, &sa, &socketlen);
1570	if (rc < 0 \|\| sa.sa_family != AF_INET)
1571	goto remove_port_forwarding;
1572
1573	psin = (struct sockaddr_in *)&sa;
1574
1575	lib = LibAliasInit(pData, NULL);
1576	flags = LibAliasSetMode(lib, 0, 0);
1577	flags \|= pData->i32AliasMode;
1578	flags \|= PKT_ALIAS_REVERSE; /* set reverse */
1579	flags = LibAliasSetMode(lib, flags, ~0);
1580
1581	alias.s_addr = RT_H2N_U32(RT_N2H_U32(guest_addr) \| CTL_ALIAS);
1582	alias_link = LibAliasRedirectPort(lib, psin->sin_addr, RT_H2N_U16(rule->host_port),
1583	alias, RT_H2N_U16(rule->guest_port),
1584	pData->special_addr, -1, /* not very clear for now */
1585	rule->proto);
1586	if (!alias_link)
1587	goto remove_port_forwarding;
1588
1589	so->so_la = lib;
1590	rule->activated = 1;
1591	rule->so = so;
1592	pData->cRedirectionsActive++;
1593	continue;
1594
1595	remove_port_forwarding:
1596	LogRel(("NAT: failed to redirect %s %RTnaipv4:%d => %RTnaipv4:%d\n",
1597	(rule->proto == IPPROTO_UDP ? "UDP" : "TCP"),
1598	rule->bind_ip.s_addr, rule->host_port,
1599	guest_addr, rule->guest_port));
1600	LIST_REMOVE(rule, list);
1601	pData->cRedirectionsStored--;
1602	RTMemFree(rule);
1603	}
1604	}
1605
1606	/**
1607	* Changes in 3.1 instead of opening new socket do the following:
1608	* gain more information:
1609	* 1. bind IP
1610	* 2. host port
1611	* 3. guest port
1612	* 4. proto
1613	* 5. guest MAC address
1614	* the guest's MAC address is rather important for service, but we easily
1615	* could get it from VM configuration in DrvNAT or Service, the idea is activating
1616	* corresponding port-forwarding
1617	*/
1618	int slirp_add_redirect(PNATState pData, int is_udp, struct in_addr host_addr, int host_port,
1619	struct in_addr guest_addr, int guest_port, const uint8_t *ethaddr)
1620	{
1621	struct port_forward_rule *rule = NULL;
1622	LIST_FOREACH(rule, &pData->port_forward_rule_head, list)
1623	{
1624	if ( rule->proto == (is_udp ? IPPROTO_UDP : IPPROTO_TCP)
1625	&& rule->host_port == host_port
1626	&& rule->bind_ip.s_addr == host_addr.s_addr
1627	&& rule->guest_port == guest_port
1628	&& rule->guest_addr.s_addr == guest_addr.s_addr
1629	)
1630	return 0; /* rule has been already registered */
1631	}
1632
1633	rule = RTMemAllocZ(sizeof(struct port_forward_rule));
1634	if (rule == NULL)
1635	return 1;
1636
1637	rule->proto = (is_udp ? IPPROTO_UDP : IPPROTO_TCP);
1638	rule->host_port = host_port;
1639	rule->guest_port = guest_port;
1640	rule->guest_addr.s_addr = guest_addr.s_addr;
1641	rule->bind_ip.s_addr = host_addr.s_addr;
1642	if (ethaddr != NULL)
1643	memcpy(rule->mac_address, ethaddr, ETH_ALEN);
1644	/* @todo add mac address */
1645	LIST_INSERT_HEAD(&pData->port_forward_rule_head, rule, list);
1646	pData->cRedirectionsStored++;
1647	/* activate port-forwarding if guest has already got assigned IP */
1648	if ( ethaddr
1649	&& memcmp(ethaddr, zerro_ethaddr, ETH_ALEN))
1650	activate_port_forwarding(pData, ethaddr);
1651	return 0;
1652	}
1653
1654	int slirp_remove_redirect(PNATState pData, int is_udp, struct in_addr host_addr, int host_port,
1655	struct in_addr guest_addr, int guest_port)
1656	{
1657	struct port_forward_rule *rule = NULL;
1658	LIST_FOREACH(rule, &pData->port_forward_rule_head, list)
1659	{
1660	if ( rule->proto == (is_udp ? IPPROTO_UDP : IPPROTO_TCP)
1661	&& rule->host_port == host_port
1662	&& rule->guest_port == guest_port
1663	&& rule->bind_ip.s_addr == host_addr.s_addr
1664	&& rule->guest_addr.s_addr == guest_addr.s_addr
1665	&& rule->activated)
1666	{
1667	LogRel(("NAT: remove redirect %s host %RTnaipv4:%d => guest %RTnaipv4:%d\n",
1668	rule->proto == IPPROTO_UDP ? "UDP" : "TCP",
1669	rule->bind_ip.s_addr, rule->host_port,
1670	guest_addr.s_addr, rule->guest_port));
1671
1672	LibAliasUninit(rule->so->so_la);
1673	if (is_udp)
1674	udp_detach(pData, rule->so);
1675	else
1676	tcp_close(pData, sototcpcb(rule->so));
1677	LIST_REMOVE(rule, list);
1678	RTMemFree(rule);
1679	pData->cRedirectionsStored--;
1680	break;
1681	}
1682
1683	}
1684	return 0;
1685	}
1686
1687	void slirp_set_ethaddr_and_activate_port_forwarding(PNATState pData, const uint8_t *ethaddr, uint32_t GuestIP)
1688	{
1689	#ifndef VBOX_WITH_NAT_SERVICE
1690	memcpy(client_ethaddr, ethaddr, ETH_ALEN);
1691	#endif
1692	if (GuestIP != INADDR_ANY)
1693	{
1694	slirp_arp_cache_update_or_add(pData, GuestIP, ethaddr);
1695	activate_port_forwarding(pData, ethaddr);
1696	}
1697	}
1698
1699	#if defined(RT_OS_WINDOWS)
1700	HANDLE *slirp_get_events(PNATState pData)
1701	{
1702	return pData->phEvents;
1703	}
1704	void slirp_register_external_event(PNATState pData, HANDLE hEvent, int index)
1705	{
1706	pData->phEvents[index] = hEvent;
1707	}
1708	#endif
1709
1710	unsigned int slirp_get_timeout_ms(PNATState pData)
1711	{
1712	if (link_up)
1713	{
1714	if (time_fasttimo)
1715	return 2;
1716	if (do_slowtimo)
1717	return 500; /* see PR_SLOWHZ */
1718	}
1719	return 36001000; / one hour */
1720	}
1721
1722	#ifndef RT_OS_WINDOWS
1723	int slirp_get_nsock(PNATState pData)
1724	{
1725	return pData->nsock;
1726	}
1727	#endif
1728
1729	/*
1730	* this function called from NAT thread
1731	*/
1732	void slirp_post_sent(PNATState pData, void *pvArg)
1733	{
1734	struct mbuf m = (struct mbuf )pvArg;
1735	m_freem(pData, m);
1736	}
1737
1738	void slirp_set_dhcp_TFTP_prefix(PNATState pData, const char *tftpPrefix)
1739	{
1740	Log2(("tftp_prefix: %s\n", tftpPrefix));
1741	tftp_prefix = tftpPrefix;
1742	}
1743
1744	void slirp_set_dhcp_TFTP_bootfile(PNATState pData, const char *bootFile)
1745	{
1746	Log2(("bootFile: %s\n", bootFile));
1747	bootp_filename = bootFile;
1748	}
1749
1750	void slirp_set_dhcp_next_server(PNATState pData, const char *next_server)
1751	{
1752	Log2(("next_server: %s\n", next_server));
1753	if (next_server == NULL)
1754	pData->tftp_server.s_addr = RT_H2N_U32(RT_N2H_U32(pData->special_addr.s_addr) \| CTL_TFTP);
1755	else
1756	inet_aton(next_server, &pData->tftp_server);
1757	}
1758
1759	int slirp_set_binding_address(PNATState pData, char *addr)
1760	{
1761	if (addr == NULL \|\| (inet_aton(addr, &pData->bindIP) == 0))
1762	{
1763	pData->bindIP.s_addr = INADDR_ANY;
1764	return 1;
1765	}
1766	return 0;
1767	}
1768
1769	void slirp_set_dhcp_dns_proxy(PNATState pData, bool fDNSProxy)
1770	{
1771	if (!pData->fUseHostResolver)
1772	{
1773	Log2(("NAT: DNS proxy switched %s\n", (fDNSProxy ? "on" : "off")));
1774	pData->fUseDnsProxy = fDNSProxy;
1775	}
1776	else if (fDNSProxy)
1777	LogRel(("NAT: Host Resolver conflicts with DNS proxy, the last one was forcely ignored\n"));
1778	}
1779
1780	#define CHECK_ARG(name, val, lim_min, lim_max) \
1781	do { \
1782	if ((val) < (lim_min) \|\| (val) > (lim_max)) \
1783	{ \
1784	LogRel(("NAT: (" #name ":%d) has been ignored, " \
1785	"because out of range (%d, %d)\n", (val), (lim_min), (lim_max))); \
1786	return; \
1787	} \
1788	else \
1789	LogRel(("NAT: (" #name ":%d)\n", (val))); \
1790	} while (0)
1791
1792	void slirp_set_somaxconn(PNATState pData, int iSoMaxConn)
1793	{
1794	LogFlowFunc(("iSoMaxConn:d\n", iSoMaxConn));
1795	/* Conditions */
1796	if (iSoMaxConn > SOMAXCONN)
1797	{
1798	LogRel(("NAT: value of somaxconn(%d) bigger than SOMAXCONN(%d)\n", iSoMaxConn, SOMAXCONN));
1799	iSoMaxConn = SOMAXCONN;
1800	}
1801
1802	if (iSoMaxConn < 1)
1803	{
1804	LogRel(("NAT: proposed value(%d) of somaxconn is invalid, default value is used (%d)\n", iSoMaxConn, pData->soMaxConn));
1805	LogFlowFuncLeave();
1806	return;
1807	}
1808
1809	/* Asignment */
1810	if (pData->soMaxConn != iSoMaxConn)
1811	{
1812	LogRel(("NAT: value of somaxconn has been changed from %d to %d\n",
1813	pData->soMaxConn, iSoMaxConn));
1814	pData->soMaxConn = iSoMaxConn;
1815	}
1816	LogFlowFuncLeave();
1817	}
1818	/* don't allow user set less 8kB and more than 1M values */
1819	#define _8K_1M_CHECK_ARG(name, val) CHECK_ARG(name, (val), 8, 1024)
1820	void slirp_set_rcvbuf(PNATState pData, int kilobytes)
1821	{
1822	_8K_1M_CHECK_ARG("SOCKET_RCVBUF", kilobytes);
1823	pData->socket_rcv = kilobytes;
1824	}
1825	void slirp_set_sndbuf(PNATState pData, int kilobytes)
1826	{
1827	_8K_1M_CHECK_ARG("SOCKET_SNDBUF", kilobytes);
1828	pData->socket_snd = kilobytes * _1K;
1829	}
1830	void slirp_set_tcp_rcvspace(PNATState pData, int kilobytes)
1831	{
1832	_8K_1M_CHECK_ARG("TCP_RCVSPACE", kilobytes);
1833	tcp_rcvspace = kilobytes * _1K;
1834	}
1835	void slirp_set_tcp_sndspace(PNATState pData, int kilobytes)
1836	{
1837	_8K_1M_CHECK_ARG("TCP_SNDSPACE", kilobytes);
1838	tcp_sndspace = kilobytes * _1K;
1839	}
1840
1841	/*
1842	* Looking for Ether by ip in ARP-cache
1843	* Note: it´s responsible of caller to allocate buffer for result
1844	* @returns iprt status code
1845	*/
1846	int slirp_arp_lookup_ether_by_ip(PNATState pData, uint32_t ip, uint8_t *ether)
1847	{
1848	struct arp_cache_entry *ac;
1849
1850	if (ether == NULL)
1851	return VERR_INVALID_PARAMETER;
1852
1853	if (LIST_EMPTY(&pData->arp_cache))
1854	return VERR_NOT_FOUND;
1855
1856	LIST_FOREACH(ac, &pData->arp_cache, list)
1857	{
1858	if ( ac->ip == ip
1859	&& memcmp(ac->ether, broadcast_ethaddr, ETH_ALEN) != 0)
1860	{
1861	memcpy(ether, ac->ether, ETH_ALEN);
1862	return VINF_SUCCESS;
1863	}
1864	}
1865	return VERR_NOT_FOUND;
1866	}
1867
1868	/*
1869	* Looking for IP by Ether in ARP-cache
1870	* Note: it´s responsible of caller to allocate buffer for result
1871	* @returns 0 - if found, 1 - otherwise
1872	*/
1873	int slirp_arp_lookup_ip_by_ether(PNATState pData, const uint8_t ether, uint32_t ip)
1874	{
1875	struct arp_cache_entry *ac;
1876	*ip = INADDR_ANY;
1877
1878	if (LIST_EMPTY(&pData->arp_cache))
1879	return VERR_NOT_FOUND;
1880
1881	LIST_FOREACH(ac, &pData->arp_cache, list)
1882	{
1883	if (memcmp(ether, ac->ether, ETH_ALEN) == 0)
1884	{
1885	*ip = ac->ip;
1886	return VINF_SUCCESS;
1887	}
1888	}
1889	return VERR_NOT_FOUND;
1890	}
1891
1892	void slirp_arp_who_has(PNATState pData, uint32_t dst)
1893	{
1894	struct mbuf *m;
1895	struct ethhdr *ehdr;
1896	struct arphdr *ahdr;
1897	static bool fWarned = false;
1898	LogFlowFunc(("ENTER: %RTnaipv4\n", dst));
1899
1900	/* ARP request WHO HAS 0.0.0.0 is one of the signals
1901	* that something has been broken at Slirp. Investigating
1902	* pcap dumps it's easy to miss warning ARP requests being
1903	* focused on investigation of other protocols flow.
1904	*/
1905	#ifdef DEBUG_vvl
1906	Assert((dst != INADDR_ANY));
1907	NOREF(fWarned);
1908	#else
1909	if ( dst == INADDR_ANY
1910	&& !fWarned)
1911	{
1912	LogRel(("NAT:ARP: \"WHO HAS INADDR_ANY\" request has been detected\n"));
1913	fWarned = true;
1914	}
1915	#endif /* !DEBUG_vvl */
1916
1917	m = m_getcl(pData, M_NOWAIT, MT_HEADER, M_PKTHDR);
1918	if (m == NULL)
1919	{
1920	Log(("NAT: Can't alloc mbuf for ARP request\n"));
1921	LogFlowFuncLeave();
1922	return;
1923	}
1924	ehdr = mtod(m, struct ethhdr *);
1925	memset(ehdr->h_source, 0xff, ETH_ALEN);
1926	ahdr = (struct arphdr *)&ehdr[1];
1927	ahdr->ar_hrd = RT_H2N_U16_C(1);
1928	ahdr->ar_pro = RT_H2N_U16_C(ETH_P_IP);
1929	ahdr->ar_hln = ETH_ALEN;
1930	ahdr->ar_pln = 4;
1931	ahdr->ar_op = RT_H2N_U16_C(ARPOP_REQUEST);
1932	memcpy(ahdr->ar_sha, special_ethaddr, ETH_ALEN);
1933	/* we assume that this request come from gw, but not from DNS or TFTP */
1934	ahdr->ar_sha[5] = CTL_ALIAS;
1935	(uint32_t )ahdr->ar_sip = RT_H2N_U32(RT_N2H_U32(pData->special_addr.s_addr) \| CTL_ALIAS);
1936	memset(ahdr->ar_tha, 0xff, ETH_ALEN); /broadcast/
1937	(uint32_t )ahdr->ar_tip = dst;
1938	/* warn!!! should falls in mbuf minimal size */
1939	m->m_len = sizeof(struct arphdr) + ETH_HLEN;
1940	m->m_data += ETH_HLEN;
1941	m->m_len -= ETH_HLEN;
1942	if_encap(pData, ETH_P_ARP, m, ETH_ENCAP_URG);
1943	LogFlowFuncLeave();
1944	}
1945	#ifdef VBOX_WITH_DNSMAPPING_IN_HOSTRESOLVER
1946	void slirp_add_host_resolver_mapping(PNATState pData, const char pszHostName, const char pszHostNamePattern, uint32_t u32HostIP)
1947	{
1948	LogFlowFunc(("ENTER: pszHostName:%s, pszHostNamePattern:%s u32HostIP:%RTnaipv4\n",
1949	pszHostName ? pszHostName : "(null)",
1950	pszHostNamePattern ? pszHostNamePattern : "(null)",
1951	u32HostIP));
1952	if ( ( pszHostName
1953	\|\| pszHostNamePattern)
1954	&& u32HostIP != INADDR_ANY
1955	&& u32HostIP != INADDR_BROADCAST)
1956	{
1957	PDNSMAPPINGENTRY pDnsMapping = RTMemAllocZ(sizeof(DNSMAPPINGENTRY));
1958	if (!pDnsMapping)
1959	{
1960	LogFunc(("Can't allocate DNSMAPPINGENTRY\n"));
1961	LogFlowFuncLeave();
1962	return;
1963	}
1964	pDnsMapping->u32IpAddress = u32HostIP;
1965	if (pszHostName)
1966	pDnsMapping->pszCName = RTStrDup(pszHostName);
1967	else if (pszHostNamePattern)
1968	pDnsMapping->pszPattern = RTStrDup(pszHostNamePattern);
1969	if ( !pDnsMapping->pszCName
1970	&& !pDnsMapping->pszPattern)
1971	{
1972	LogFunc(("Can't allocate enough room for %s\n", pszHostName ? pszHostName : pszHostNamePattern));
1973	RTMemFree(pDnsMapping);
1974	LogFlowFuncLeave();
1975	return;
1976	}
1977	LIST_INSERT_HEAD(&pData->DNSMapHead, pDnsMapping, MapList);
1978	LogRel(("NAT: user-defined mapping %s: %RTnaipv4 is registered\n",
1979	pDnsMapping->pszCName ? pDnsMapping->pszCName : pDnsMapping->pszPattern,
1980	pDnsMapping->u32IpAddress));
1981	}
1982	LogFlowFuncLeave();
1983	}
1984	#endif
1985
1986	/* updates the arp cache
1987	* @note: this is helper function, slirp_arp_cache_update_or_add should be used.
1988	* @returns 0 - if has found and updated
1989	* 1 - if hasn't found.
1990	*/
1991	static inline int slirp_arp_cache_update(PNATState pData, uint32_t dst, const uint8_t *mac)
1992	{
1993	struct arp_cache_entry *ac;
1994	Assert(( memcmp(mac, broadcast_ethaddr, ETH_ALEN)
1995	&& memcmp(mac, zerro_ethaddr, ETH_ALEN)));
1996	LIST_FOREACH(ac, &pData->arp_cache, list)
1997	{
1998	if (ac->ip == dst)
1999	{
2000	memcpy(ac->ether, mac, ETH_ALEN);
2001	return 0;
2002	}
2003	}
2004	return 1;
2005	}
2006
2007	/**
2008	* add entry to the arp cache
2009	* @note: this is helper function, slirp_arp_cache_update_or_add should be used.
2010	*/
2011	static inline void slirp_arp_cache_add(PNATState pData, uint32_t ip, const uint8_t *ether)
2012	{
2013	struct arp_cache_entry *ac = NULL;
2014	Assert(( memcmp(ether, broadcast_ethaddr, ETH_ALEN)
2015	&& memcmp(ether, zerro_ethaddr, ETH_ALEN)));
2016	ac = RTMemAllocZ(sizeof(struct arp_cache_entry));
2017	if (ac == NULL)
2018	{
2019	Log(("NAT: Can't allocate arp cache entry\n"));
2020	return;
2021	}
2022	ac->ip = ip;
2023	memcpy(ac->ether, ether, ETH_ALEN);
2024	LIST_INSERT_HEAD(&pData->arp_cache, ac, list);
2025	}
2026
2027	/* updates or adds entry to the arp cache
2028	* @returns 0 - if has found and updated
2029	* 1 - if hasn't found.
2030	*/
2031	int slirp_arp_cache_update_or_add(PNATState pData, uint32_t dst, const uint8_t *mac)
2032	{
2033	if ( !memcmp(mac, broadcast_ethaddr, ETH_ALEN)
2034	\|\| !memcmp(mac, zerro_ethaddr, ETH_ALEN))
2035	{
2036	static bool fBroadcastEtherAddReported;
2037	if (!fBroadcastEtherAddReported)
2038	{
2039	LogRel(("NAT: Attempt to add pair [%RTmac:%RTnaipv4] in ARP cache was ignored\n",
2040	mac, dst));
2041	fBroadcastEtherAddReported = true;
2042	}
2043	return 1;
2044	}
2045	if (slirp_arp_cache_update(pData, dst, mac))
2046	slirp_arp_cache_add(pData, dst, mac);
2047
2048	return 0;
2049	}
2050
2051
2052	void slirp_set_mtu(PNATState pData, int mtu)
2053	{
2054	if (mtu < 20 \|\| mtu >= 16000)
2055	{
2056	LogRel(("NAT: mtu(%d) is out of range (20;16000] mtu forcely assigned to 1500\n", mtu));
2057	mtu = 1500;
2058	}
2059	/* MTU is maximum transition unit on */
2060	if_mtu =
2061	if_mru = mtu;
2062	}
2063
2064	/**
2065	* Info handler.
2066	*/
2067	void slirp_info(PNATState pData, const void pvArg, const char pszArgs)
2068	{
2069	struct socket so, so_next;
2070	struct arp_cache_entry *ac;
2071	struct port_forward_rule *rule;
2072	PCDBGFINFOHLP pHlp = (PCDBGFINFOHLP)pvArg;
2073	NOREF(pszArgs);
2074
2075	pHlp->pfnPrintf(pHlp, "NAT parameters: MTU=%d\n", if_mtu);
2076	pHlp->pfnPrintf(pHlp, "NAT TCP ports:\n");
2077	QSOCKET_FOREACH(so, so_next, tcp)
2078	/* { */
2079	pHlp->pfnPrintf(pHlp, " %R[natsock]\n", so);
2080	}
2081
2082	pHlp->pfnPrintf(pHlp, "NAT UDP ports:\n");
2083	QSOCKET_FOREACH(so, so_next, udp)
2084	/* { */
2085	pHlp->pfnPrintf(pHlp, " %R[natsock]\n", so);
2086	}
2087
2088	pHlp->pfnPrintf(pHlp, "NAT ARP cache:\n");
2089	LIST_FOREACH(ac, &pData->arp_cache, list)
2090	{
2091	pHlp->pfnPrintf(pHlp, " %RTnaipv4 %RTmac\n", ac->ip, &ac->ether);
2092	}
2093
2094	pHlp->pfnPrintf(pHlp, "NAT rules:\n");
2095	LIST_FOREACH(rule, &pData->port_forward_rule_head, list)
2096	{
2097	pHlp->pfnPrintf(pHlp, " %s %d => %RTnaipv4:%d %c\n",
2098	rule->proto == IPPROTO_UDP ? "UDP" : "TCP",
2099	rule->host_port, rule->guest_addr.s_addr, rule->guest_port,
2100	rule->activated ? ' ' : '*');
2101	}
2102	}
2103
2104	/**
2105	* @note: NATState::fUseHostResolver could be changed in bootp.c::dhcp_decode
2106	* @note: this function is executed on GUI/VirtualBox or main/VBoxHeadless thread.
2107	* @note: this function can potentially race with bootp.c::dhcp_decode (except Darwin)
2108	*/
2109	int slirp_host_network_configuration_change_strategy_selector(const PNATState pData)
2110	{
2111	if (pData->fUseHostResolverPermanent)
2112	return VBOX_NAT_DNS_HOSTRESOLVER;
2113
2114	if (pData->fUseDnsProxy) {
2115	#if HAVE_NOTIFICATION_FOR_DNS_UPDATE
2116	/* We dont conflict with bootp.c::dhcp_decode */
2117	struct rcp_state rcp_state;
2118	int rc;
2119
2120	rcp_state.rcps_flags \|= RCPSF_IGNORE_IPV6;
2121	rc = rcp_parse(&rcp_state, RESOLV_CONF_FILE);
2122	LogRelFunc(("NAT: rcp_parse:%Rrc old domain:%s new domain:%s\n",
2123	rc, LIST_FIRST(&pData->pDomainList)->dd_pszDomain,
2124	rcp_state.rcps_domain));
2125	if ( RT_FAILURE(rc)
2126	\|\| LIST_EMPTY(&pData->pDomainList))
2127	return VBOX_NAT_DNS_DNSPROXY;
2128
2129	if ( rcp_state.rcps_domain
2130	&& strcmp(rcp_state.rcps_domain, LIST_FIRST(&pData->pDomainList)->dd_pszDomain) == 0)
2131	return VBOX_NAT_DNS_DNSPROXY;
2132	else
2133	return VBOX_NAT_DNS_EXTERNAL;
2134	#else
2135	/* copy domain name */
2136	/* domain only compare with coy version */
2137	return VBOX_NAT_DNS_DNSPROXY;
2138	#endif
2139	}
2140	return VBOX_NAT_DNS_EXTERNAL;
2141	}

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

source: vbox/trunk/src/VBox/Devices/Network/slirp/slirp.c@ 52343

Download in other formats: