slirp.c@ 65455

Last change on this file since 65455 was 64679, checked in by vboxsync, 8 years ago

NAT: slirp_select_poll - don't stop after draining half-closed socket,
we still need to check it for POLLOUT.

This whole block could use some refactoring to clarify repeated
conditions and spooky action at a distance, but this fix is intended
to have minimal diff.

Property svn:eol-style set to native
Property svn:keywords set to Author Date Id Revision

File size: 63.5 KB

Line
1	/* $Id: slirp.c 64679 2016-11-15 23:51:15Z vboxsync $ */
2	/** @file
3	* NAT - slirp glue.
4	*/
5
6	/*
7	* Copyright (C) 2006-2016 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 <iprt/win/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	#ifdef 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 const uint8_t special_ethaddr[6] =
264	{
265	0x52, 0x54, 0x00, 0x12, 0x35, 0x00
266	};
267
268	static const uint8_t broadcast_ethaddr[6] =
269	{
270	0xff, 0xff, 0xff, 0xff, 0xff, 0xff
271	};
272
273	const uint8_t zerro_ethaddr[6] =
274	{
275	0x0, 0x0, 0x0, 0x0, 0x0, 0x0
276	};
277
278	/**
279	* This helper routine do the checks in descriptions to
280	* ''fUnderPolling'' and ''fShouldBeRemoved'' flags
281	* @returns 1 if socket removed and 0 if no changes was made.
282	*/
283	static int slirpVerifyAndFreeSocket(PNATState pData, struct socket *pSocket)
284	{
285	AssertPtrReturn(pData, 0);
286	AssertPtrReturn(pSocket, 0);
287	AssertReturn(pSocket->fUnderPolling, 0);
288	if (pSocket->fShouldBeRemoved)
289	{
290	pSocket->fUnderPolling = 0;
291	sofree(pData, pSocket);
292	/* pSocket is PHANTOM, now */
293	return 1;
294	}
295	return 0;
296	}
297
298	int slirp_init(PNATState *ppData, uint32_t u32NetAddr, uint32_t u32Netmask,
299	bool fPassDomain, bool fUseHostResolver, int i32AliasMode,
300	int iIcmpCacheLimit, void *pvUser)
301	{
302	int rc;
303	PNATState pData;
304	if (u32Netmask & 0x1f)
305	{
306	/* CTL is x.x.x.15, bootp passes up to 16 IPs (15..31) */
307	LogRel(("NAT: The last 5 bits of the netmask (%RTnaipv4) need to be unset\n", RT_BE2H_U32(u32Netmask)));
308	return VERR_INVALID_PARAMETER;
309	}
310	pData = RTMemAllocZ(RT_ALIGN_Z(sizeof(NATState), sizeof(uint64_t)));
311	*ppData = pData;
312	if (!pData)
313	return VERR_NO_MEMORY;
314	pData->fPassDomain = !fUseHostResolver ? fPassDomain : false;
315	pData->fUseHostResolver = fUseHostResolver;
316	pData->fUseHostResolverPermanent = fUseHostResolver;
317	pData->pvUser = pvUser;
318	pData->netmask = u32Netmask;
319
320	rc = RTCritSectRwInit(&pData->CsRwHandlerChain);
321	if (RT_FAILURE(rc))
322	return rc;
323
324	/* sockets & TCP defaults */
325	pData->socket_rcv = 64 * _1K;
326	pData->socket_snd = 64 * _1K;
327	tcp_sndspace = 64 * _1K;
328	tcp_rcvspace = 64 * _1K;
329
330	/*
331	* Use the same default here as in DevNAT.cpp (SoMaxConnection CFGM value)
332	* to avoid release log noise.
333	*/
334	pData->soMaxConn = 10;
335
336	#ifdef RT_OS_WINDOWS
337	{
338	WSADATA Data;
339	RTLDRMOD hLdrMod;
340
341	WSAStartup(MAKEWORD(2, 0), &Data);
342
343	rc = RTLdrLoadSystem("Iphlpapi.dll", true /fNoUnload/, &hLdrMod);
344	if (RT_SUCCESS(rc))
345	{
346	rc = RTLdrGetSymbol(hLdrMod, "GetAdaptersAddresses", (void **)&pData->pfnGetAdaptersAddresses);
347	if (RT_FAILURE(rc))
348	LogRel(("NAT: Can't find GetAdapterAddresses in Iphlpapi.dll\n"));
349
350	RTLdrClose(hLdrMod);
351	}
352	}
353	pData->phEvents[VBOX_SOCKET_EVENT_INDEX] = CreateEvent(NULL, FALSE, FALSE, NULL);
354	#endif
355
356	rc = bootp_dhcp_init(pData);
357	if (RT_FAILURE(rc))
358	{
359	Log(("NAT: DHCP server initialization failed\n"));
360	RTMemFree(pData);
361	*ppData = NULL;
362	return rc;
363	}
364	debug_init(pData);
365	if_init(pData);
366	ip_init(pData);
367	icmp_init(pData, iIcmpCacheLimit);
368
369	/* Initialise mbufs after setting the MTU */
370	mbuf_init(pData);
371
372	pData->special_addr.s_addr = u32NetAddr;
373	pData->slirp_ethaddr = &special_ethaddr[0];
374	alias_addr.s_addr = pData->special_addr.s_addr \| RT_H2N_U32_C(CTL_ALIAS);
375	/** @todo add ability to configure this staff */
376
377	/*
378	* Some guests won't reacquire DHCP lease on link flap when VM is
379	* restored. Instead of forcing users to explicitly set CTL_GUEST
380	* in port-forwarding rules, provide it as initial guess here.
381	*/
382	slirp_update_guest_addr_guess(pData,
383	pData->special_addr.s_addr \| RT_H2N_U32_C(CTL_GUEST),
384	"initialization");
385
386	/* set default addresses */
387	inet_aton("127.0.0.1", &loopback_addr);
388
389	rc = slirpTftpInit(pData);
390	AssertRCReturn(rc, VINF_NAT_DNS);
391
392	if (i32AliasMode & ~(PKT_ALIAS_LOG\|PKT_ALIAS_SAME_PORTS\|PKT_ALIAS_PROXY_ONLY))
393	{
394	Log(("NAT: alias mode %x is ignored\n", i32AliasMode));
395	i32AliasMode = 0;
396	}
397	pData->i32AliasMode = i32AliasMode;
398	getouraddr(pData);
399	{
400	int flags = 0;
401	struct in_addr proxy_addr;
402	pData->proxy_alias = LibAliasInit(pData, NULL);
403	if (pData->proxy_alias == NULL)
404	{
405	Log(("NAT: LibAlias default rule wasn't initialized\n"));
406	AssertMsgFailed(("NAT: LibAlias default rule wasn't initialized\n"));
407	}
408	flags = LibAliasSetMode(pData->proxy_alias, 0, 0);
409	#ifndef NO_FW_PUNCH
410	flags \|= PKT_ALIAS_PUNCH_FW;
411	#endif
412	flags \|= pData->i32AliasMode; /* do transparent proxying */
413	flags = LibAliasSetMode(pData->proxy_alias, flags, ~0U);
414	proxy_addr.s_addr = RT_H2N_U32(RT_N2H_U32(pData->special_addr.s_addr) \| CTL_ALIAS);
415	LibAliasSetAddress(pData->proxy_alias, proxy_addr);
416	ftp_alias_load(pData);
417	nbt_alias_load(pData);
418	}
419	#ifdef VBOX_WITH_NAT_SEND2HOME
420	/** @todo we should know all interfaces available on host. */
421	pData->pInSockAddrHomeAddress = RTMemAllocZ(sizeof(struct sockaddr));
422	pData->cInHomeAddressSize = 1;
423	inet_aton("192.168.1.25", &pData->pInSockAddrHomeAddress[0].sin_addr);
424	pData->pInSockAddrHomeAddress[0].sin_family = AF_INET;
425	# ifdef RT_OS_DARWIN
426	pData->pInSockAddrHomeAddress[0].sin_len = sizeof(struct sockaddr_in);
427	# endif
428	#endif
429
430	#ifdef VBOX_WITH_DNSMAPPING_IN_HOSTRESOLVER
431	STAILQ_INIT(&pData->DNSMapNames);
432	STAILQ_INIT(&pData->DNSMapPatterns);
433	#endif
434
435	slirp_link_up(pData);
436	return VINF_SUCCESS;
437	}
438
439	/**
440	* Register statistics.
441	*/
442	void slirp_register_statistics(PNATState pData, PPDMDRVINS pDrvIns)
443	{
444	#ifdef VBOX_WITH_STATISTICS
445	# define PROFILE_COUNTER(name, dsc) REGISTER_COUNTER(name, pData, STAMTYPE_PROFILE, STAMUNIT_TICKS_PER_CALL, dsc)
446	# define COUNTING_COUNTER(name, dsc) REGISTER_COUNTER(name, pData, STAMTYPE_COUNTER, STAMUNIT_COUNT, dsc)
447	# include "counters.h"
448	# undef COUNTER
449	/** @todo register statistics for the variables dumped by:
450	* ipstats(pData); tcpstats(pData); udpstats(pData); icmpstats(pData);
451	* mbufstats(pData); sockstats(pData); */
452	#else /* VBOX_WITH_STATISTICS */
453	NOREF(pData);
454	NOREF(pDrvIns);
455	#endif /* !VBOX_WITH_STATISTICS */
456	}
457
458	/**
459	* Deregister statistics.
460	*/
461	void slirp_deregister_statistics(PNATState pData, PPDMDRVINS pDrvIns)
462	{
463	if (pData == NULL)
464	return;
465	#ifdef VBOX_WITH_STATISTICS
466	# define PROFILE_COUNTER(name, dsc) DEREGISTER_COUNTER(name, pData)
467	# define COUNTING_COUNTER(name, dsc) DEREGISTER_COUNTER(name, pData)
468	# include "counters.h"
469	#else /* VBOX_WITH_STATISTICS */
470	NOREF(pData);
471	NOREF(pDrvIns);
472	#endif /* !VBOX_WITH_STATISTICS */
473	}
474
475	/**
476	* Marks the link as up, making it possible to establish new connections.
477	*/
478	void slirp_link_up(PNATState pData)
479	{
480	if (link_up == 1)
481	return;
482
483	link_up = 1;
484
485	if (!pData->fUseHostResolverPermanent)
486	slirpInitializeDnsSettings(pData);
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	if (link_up == 0)
495	return;
496
497	slirpReleaseDnsSettings(pData);
498
499	link_up = 0;
500	}
501
502	/**
503	* Terminates the slirp component.
504	*/
505	void slirp_term(PNATState pData)
506	{
507	struct socket *so;
508
509	if (pData == NULL)
510	return;
511
512	icmp_finit(pData);
513
514	while ((so = tcb.so_next) != &tcb)
515	{
516	/* Don't miss TCB releasing */
517	if ( !sototcpcb(so)
518	&& ( so->so_state & SS_NOFDREF
519	\|\| so->s == -1))
520	sofree(pData, so);
521	else
522	tcp_close(pData, sototcpcb(so));
523	}
524
525	while ((so = udb.so_next) != &udb)
526	udp_detach(pData, so);
527
528	slirp_link_down(pData);
529	ftp_alias_unload(pData);
530	nbt_alias_unload(pData);
531
532	#ifdef VBOX_WITH_DNSMAPPING_IN_HOSTRESOLVER
533	{
534	DNSMAPPINGHEAD *heads[2];
535	int i;
536
537	heads[0] = &pData->DNSMapNames;
538	heads[1] = &pData->DNSMapPatterns;
539	for (i = 0; i < RT_ELEMENTS(heads); ++i)
540	{
541	while (!STAILQ_EMPTY(heads[i]))
542	{
543	PDNSMAPPINGENTRY pDnsEntry = STAILQ_FIRST(heads[i]);
544	STAILQ_REMOVE_HEAD(heads[i], MapList);
545	RTStrFree(pDnsEntry->pszName);
546	RTMemFree(pDnsEntry);
547	}
548	}
549	}
550	#endif
551
552	while (!LIST_EMPTY(&instancehead))
553	{
554	struct libalias *la = LIST_FIRST(&instancehead);
555	/* libalias do all clean up */
556	LibAliasUninit(la);
557	}
558	while (!LIST_EMPTY(&pData->arp_cache))
559	{
560	struct arp_cache_entry *ac = LIST_FIRST(&pData->arp_cache);
561	LIST_REMOVE(ac, list);
562	RTMemFree(ac);
563	}
564	slirpTftpTerm(pData);
565	bootp_dhcp_fini(pData);
566	m_fini(pData);
567	#ifdef RT_OS_WINDOWS
568	WSACleanup();
569	#endif
570	#ifdef LOG_ENABLED
571	Log(("\n"
572	"NAT statistics\n"
573	"--------------\n"
574	"\n"));
575	ipstats(pData);
576	tcpstats(pData);
577	udpstats(pData);
578	icmpstats(pData);
579	mbufstats(pData);
580	sockstats(pData);
581	Log(("\n"
582	"\n"
583	"\n"));
584	#endif
585	RTCritSectRwDelete(&pData->CsRwHandlerChain);
586	RTMemFree(pData);
587	}
588
589
590	#define CONN_CANFSEND(so) (((so)->so_state & (SS_FCANTSENDMORE\|SS_ISFCONNECTED)) == SS_ISFCONNECTED)
591	#define CONN_CANFRCV(so) (((so)->so_state & (SS_FCANTRCVMORE\|SS_ISFCONNECTED)) == SS_ISFCONNECTED)
592
593	/*
594	* curtime kept to an accuracy of 1ms
595	*/
596	static void updtime(PNATState pData)
597	{
598	#ifdef RT_OS_WINDOWS
599	struct _timeb tb;
600
601	_ftime(&tb);
602	curtime = (u_int)tb.time * (u_int)1000;
603	curtime += (u_int)tb.millitm;
604	#else
605	gettimeofday(&tt, 0);
606
607	curtime = (u_int)tt.tv_sec * (u_int)1000;
608	curtime += (u_int)tt.tv_usec / (u_int)1000;
609
610	if ((tt.tv_usec % 1000) >= 500)
611	curtime++;
612	#endif
613	}
614
615	#ifdef RT_OS_WINDOWS
616	void slirp_select_fill(PNATState pData, int *pnfds)
617	#else /* RT_OS_WINDOWS */
618	void slirp_select_fill(PNATState pData, int pnfds, struct pollfd polls)
619	#endif /* !RT_OS_WINDOWS */
620	{
621	struct socket so, so_next;
622	int nfds;
623	#if defined(RT_OS_WINDOWS)
624	int rc;
625	int error;
626	#else
627	int poll_index = 0;
628	#endif
629	int i;
630
631	STAM_PROFILE_START(&pData->StatFill, a);
632
633	nfds = *pnfds;
634
635	/*
636	* First, TCP sockets
637	*/
638	do_slowtimo = 0;
639	if (!link_up)
640	goto done;
641
642	/*
643	* *_slowtimo needs calling if there are IP fragments
644	* in the fragment queue, or there are TCP connections active
645	*/
646	/* XXX:
647	* triggering of fragment expiration should be the same but use new macroses
648	*/
649	do_slowtimo = (tcb.so_next != &tcb);
650	if (!do_slowtimo)
651	{
652	for (i = 0; i < IPREASS_NHASH; i++)
653	{
654	if (!TAILQ_EMPTY(&ipq[i]))
655	{
656	do_slowtimo = 1;
657	break;
658	}
659	}
660	}
661	/* always add the ICMP socket */
662	#ifndef RT_OS_WINDOWS
663	pData->icmp_socket.so_poll_index = -1;
664	#endif
665	ICMP_ENGAGE_EVENT(&pData->icmp_socket, readfds);
666
667	STAM_COUNTER_RESET(&pData->StatTCP);
668	STAM_COUNTER_RESET(&pData->StatTCPHot);
669
670	QSOCKET_FOREACH(so, so_next, tcp)
671	/* { */
672	Assert(so->so_type == IPPROTO_TCP);
673	#if !defined(RT_OS_WINDOWS)
674	so->so_poll_index = -1;
675	#endif
676	STAM_COUNTER_INC(&pData->StatTCP);
677	#ifdef VBOX_WITH_NAT_UDP_SOCKET_CLONE
678	/* TCP socket can't be cloned */
679	Assert((!so->so_cloneOf));
680	#endif
681	/*
682	* See if we need a tcp_fasttimo
683	*/
684	if ( time_fasttimo == 0
685	&& so->so_tcpcb != NULL
686	&& so->so_tcpcb->t_flags & TF_DELACK)
687	{
688	time_fasttimo = curtime; /* Flag when we want a fasttimo */
689	}
690
691	/*
692	* NOFDREF can include still connecting to local-host,
693	* newly socreated() sockets etc. Don't want to select these.
694	*/
695	if (so->so_state & SS_NOFDREF \|\| so->s == -1)
696	CONTINUE(tcp);
697
698	/*
699	* Set for reading sockets which are accepting
700	*/
701	if (so->so_state & SS_FACCEPTCONN)
702	{
703	STAM_COUNTER_INC(&pData->StatTCPHot);
704	TCP_ENGAGE_EVENT1(so, readfds);
705	CONTINUE(tcp);
706	}
707
708	/*
709	* Set for writing sockets which are connecting
710	*/
711	if (so->so_state & SS_ISFCONNECTING)
712	{
713	Log2(("connecting %R[natsock] engaged\n",so));
714	STAM_COUNTER_INC(&pData->StatTCPHot);
715	#ifdef RT_OS_WINDOWS
716	WIN_TCP_ENGAGE_EVENT2(so, writefds, connectfds);
717	#else
718	TCP_ENGAGE_EVENT1(so, writefds);
719	#endif
720	}
721
722	/*
723	* Set for writing if we are connected, can send more, and
724	* we have something to send
725	*/
726	if (CONN_CANFSEND(so) && SBUF_LEN(&so->so_rcv))
727	{
728	STAM_COUNTER_INC(&pData->StatTCPHot);
729	TCP_ENGAGE_EVENT1(so, writefds);
730	}
731
732	/*
733	* Set for reading (and urgent data) if we are connected, can
734	* receive more, and we have room for it XXX /2 ?
735	*/
736	/** @todo vvl - check which predicat here will be more useful here in rerm of new sbufs. */
737	if ( CONN_CANFRCV(so)
738	&& (SBUF_LEN(&so->so_snd) < (SBUF_SIZE(&so->so_snd)/2))
739	#ifdef RT_OS_WINDOWS
740	&& !(so->so_state & SS_ISFCONNECTING)
741	#endif
742	)
743	{
744	STAM_COUNTER_INC(&pData->StatTCPHot);
745	TCP_ENGAGE_EVENT2(so, readfds, xfds);
746	}
747	LOOP_LABEL(tcp, so, so_next);
748	}
749
750	/*
751	* UDP sockets
752	*/
753	STAM_COUNTER_RESET(&pData->StatUDP);
754	STAM_COUNTER_RESET(&pData->StatUDPHot);
755
756	QSOCKET_FOREACH(so, so_next, udp)
757	/* { */
758
759	Assert(so->so_type == IPPROTO_UDP);
760	STAM_COUNTER_INC(&pData->StatUDP);
761	#if !defined(RT_OS_WINDOWS)
762	so->so_poll_index = -1;
763	#endif
764
765	/*
766	* See if it's timed out
767	*/
768	if (so->so_expire)
769	{
770	if (so->so_expire <= curtime)
771	{
772	Log2(("NAT: %R[natsock] expired\n", so));
773	if (so->so_timeout != NULL)
774	{
775	/* so_timeout - might change the so_expire value or
776	* drop so_timeout* from so.
777	*/
778	so->so_timeout(pData, so, so->so_timeout_arg);
779	/* on 4.2 so->
780	*/
781	if ( so_next->so_prev != so /* so_timeout freed the socket */
782	\|\| so->so_timeout) /* so_timeout just freed so_timeout */
783	CONTINUE_NO_UNLOCK(udp);
784	}
785	UDP_DETACH(pData, so, so_next);
786	CONTINUE_NO_UNLOCK(udp);
787	}
788	}
789	#ifdef VBOX_WITH_NAT_UDP_SOCKET_CLONE
790	if (so->so_cloneOf)
791	CONTINUE_NO_UNLOCK(udp);
792	#endif
793
794	/*
795	* When UDP packets are received from over the link, they're
796	* sendto()'d straight away, so no need for setting for writing
797	* Limit the number of packets queued by this session to 4.
798	* Note that even though we try and limit this to 4 packets,
799	* the session could have more queued if the packets needed
800	* to be fragmented.
801	*
802	* (XXX <= 4 ?)
803	*/
804	if ((so->so_state & SS_ISFCONNECTED) && so->so_queued <= 4)
805	{
806	STAM_COUNTER_INC(&pData->StatUDPHot);
807	UDP_ENGAGE_EVENT(so, readfds);
808	}
809	LOOP_LABEL(udp, so, so_next);
810	}
811	done:
812
813	#if defined(RT_OS_WINDOWS)
814	*pnfds = VBOX_EVENT_COUNT;
815	#else /* RT_OS_WINDOWS */
816	AssertRelease(poll_index <= *pnfds);
817	*pnfds = poll_index;
818	#endif /* !RT_OS_WINDOWS */
819
820	STAM_PROFILE_STOP(&pData->StatFill, a);
821	}
822
823
824	/**
825	* This function do Connection or sending tcp sequence to.
826	* @returns if true operation completed
827	* @note: functions call tcp_input that potentially could lead to tcp_drop
828	*/
829	static bool slirpConnectOrWrite(PNATState pData, struct socket *so, bool fConnectOnly)
830	{
831	int ret;
832	LogFlowFunc(("ENTER: so:%R[natsock], fConnectOnly:%RTbool\n", so, fConnectOnly));
833	/*
834	* Check for non-blocking, still-connecting sockets
835	*/
836	if (so->so_state & SS_ISFCONNECTING)
837	{
838	Log2(("connecting %R[natsock] catched\n", so));
839	/* Connected */
840	so->so_state &= ~SS_ISFCONNECTING;
841
842	/*
843	* This should be probably guarded by PROBE_CONN too. Anyway,
844	* we disable it on OS/2 because the below send call returns
845	* EFAULT which causes the opened TCP socket to close right
846	* after it has been opened and connected.
847	*/
848	#ifndef RT_OS_OS2
849	ret = send(so->s, (const char *)&ret, 0, 0);
850	if (ret < 0)
851	{
852	/* XXXXX Must fix, zero bytes is a NOP */
853	if ( soIgnorableErrorCode(errno)
854	\|\| errno == ENOTCONN)
855	{
856	LogFlowFunc(("LEAVE: false\n"));
857	return false;
858	}
859
860	/* else failed */
861	so->so_state = SS_NOFDREF;
862	}
863	/* else so->so_state &= ~SS_ISFCONNECTING; */
864	#endif
865
866	/*
867	* Continue tcp_input
868	*/
869	TCP_INPUT(pData, (struct mbuf *)NULL, sizeof(struct ip), so);
870	/* continue; */
871	}
872	else if (!fConnectOnly)
873	{
874	SOWRITE(ret, pData, so);
875	if (RT_LIKELY(ret > 0))
876	{
877	/*
878	* Make sure we will send window update to peer. This is
879	* a moral equivalent of calling tcp_output() for PRU_RCVD
880	* in tcp_usrreq() of the real stack.
881	*/
882	struct tcpcb *tp = sototcpcb(so);
883	if (RT_LIKELY(tp != NULL))
884	tp->t_flags \|= TF_DELACK;
885	}
886	}
887
888	LogFlowFunc(("LEAVE: true\n"));
889	return true;
890	}
891
892	#if defined(RT_OS_WINDOWS)
893	void slirp_select_poll(PNATState pData, int fTimeout)
894	#else /* RT_OS_WINDOWS */
895	void slirp_select_poll(PNATState pData, struct pollfd *polls, int ndfs)
896	#endif /* !RT_OS_WINDOWS */
897	{
898	struct socket so, so_next;
899	int ret;
900	#if defined(RT_OS_WINDOWS)
901	WSANETWORKEVENTS NetworkEvents;
902	int rc;
903	int error;
904	#endif
905
906	STAM_PROFILE_START(&pData->StatPoll, a);
907
908	/* Update time */
909	updtime(pData);
910
911	/*
912	* See if anything has timed out
913	*/
914	if (link_up)
915	{
916	if (time_fasttimo && ((curtime - time_fasttimo) >= 2))
917	{
918	STAM_PROFILE_START(&pData->StatFastTimer, b);
919	tcp_fasttimo(pData);
920	time_fasttimo = 0;
921	STAM_PROFILE_STOP(&pData->StatFastTimer, b);
922	}
923	if (do_slowtimo && ((curtime - last_slowtimo) >= 499))
924	{
925	STAM_PROFILE_START(&pData->StatSlowTimer, c);
926	ip_slowtimo(pData);
927	tcp_slowtimo(pData);
928	last_slowtimo = curtime;
929	STAM_PROFILE_STOP(&pData->StatSlowTimer, c);
930	}
931	}
932	#if defined(RT_OS_WINDOWS)
933	if (fTimeout)
934	return; /* only timer update */
935	#endif
936
937	/*
938	* Check sockets
939	*/
940	if (!link_up)
941	goto done;
942	#if defined(RT_OS_WINDOWS)
943	icmpwin_process(pData);
944	#else
945	if ( (pData->icmp_socket.s != -1)
946	&& CHECK_FD_SET(&pData->icmp_socket, ignored, readfds))
947	sorecvfrom(pData, &pData->icmp_socket);
948	#endif
949	/*
950	* Check TCP sockets
951	*/
952	QSOCKET_FOREACH(so, so_next, tcp)
953	/* { */
954	/* TCP socket can't be cloned */
955	#ifdef VBOX_WITH_NAT_UDP_SOCKET_CLONE
956	Assert((!so->so_cloneOf));
957	#endif
958	Assert(!so->fUnderPolling);
959	so->fUnderPolling = 1;
960	if (slirpVerifyAndFreeSocket(pData, so))
961	CONTINUE(tcp);
962	/*
963	* FD_ISSET is meaningless on these sockets
964	* (and they can crash the program)
965	*/
966	if (so->so_state & SS_NOFDREF \|\| so->s == -1)
967	{
968	so->fUnderPolling = 0;
969	CONTINUE(tcp);
970	}
971
972	POLL_TCP_EVENTS(rc, error, so, &NetworkEvents);
973
974	LOG_NAT_SOCK(so, TCP, &NetworkEvents, readfds, writefds, xfds);
975
976	if (so->so_state & SS_ISFCONNECTING)
977	{
978	int sockerr = 0;
979	#if !defined(RT_OS_WINDOWS)
980	{
981	int revents = 0;
982
983	/*
984	* Failed connect(2) is reported by poll(2) on
985	* different OSes with different combinations of
986	* POLLERR, POLLHUP, and POLLOUT.
987	*/
988	if ( CHECK_FD_SET(so, NetworkEvents, closefds) /* POLLHUP */
989	\|\| CHECK_FD_SET(so, NetworkEvents, rderr)) /* POLLERR */
990	{
991	revents = POLLHUP; /* squash to single "failed" flag */
992	}
993	#if defined(RT_OS_SOLARIS) \|\| defined(RT_OS_NETBSD)
994	/* Solaris and NetBSD report plain POLLOUT even on error */
995	else if (CHECK_FD_SET(so, NetworkEvents, writefds)) /* POLLOUT */
996	{
997	revents = POLLOUT;
998	}
999	#endif
1000
1001	if (revents != 0)
1002	{
1003	socklen_t optlen = (socklen_t)sizeof(sockerr);
1004	ret = getsockopt(so->s, SOL_SOCKET, SO_ERROR, &sockerr, &optlen);
1005
1006	if ( RT_UNLIKELY(ret < 0)
1007	\|\| ( (revents & POLLHUP)
1008	&& RT_UNLIKELY(sockerr == 0)))
1009	sockerr = ETIMEDOUT;
1010	}
1011	}
1012	#else /* RT_OS_WINDOWS */
1013	{
1014	if (NetworkEvents.lNetworkEvents & FD_CONNECT)
1015	sockerr = NetworkEvents.iErrorCode[FD_CONNECT_BIT];
1016	}
1017	#endif
1018	if (sockerr != 0)
1019	{
1020	tcp_fconnect_failed(pData, so, sockerr);
1021	ret = slirpVerifyAndFreeSocket(pData, so);
1022	Assert(ret == 1); /* freed */
1023	CONTINUE(tcp);
1024	}
1025
1026	/*
1027	* XXX: For now just fall through to the old code to
1028	* handle successful connect(2).
1029	*/
1030	}
1031
1032	/*
1033	* Check for URG data
1034	* This will soread as well, so no need to
1035	* test for readfds below if this succeeds
1036	*/
1037
1038	/* out-of-band data */
1039	if ( CHECK_FD_SET(so, NetworkEvents, xfds)
1040	#ifdef RT_OS_DARWIN
1041	/* Darwin and probably BSD hosts generates POLLPRI\|POLLHUP event on receiving TCP.flags.{ACK\|URG\|FIN} this
1042	* combination on other Unixs hosts doesn't enter to this branch
1043	*/
1044	&& !CHECK_FD_SET(so, NetworkEvents, closefds)
1045	#endif
1046	#ifdef RT_OS_WINDOWS
1047	/**
1048	* In some cases FD_CLOSE comes with FD_OOB, that confuse tcp processing.
1049	*/
1050	&& !WIN_CHECK_FD_SET(so, NetworkEvents, closefds)
1051	#endif
1052	)
1053	{
1054	sorecvoob(pData, so);
1055	if (slirpVerifyAndFreeSocket(pData, so))
1056	CONTINUE(tcp);
1057	}
1058
1059	/*
1060	* Check sockets for reading
1061	*/
1062	else if ( CHECK_FD_SET(so, NetworkEvents, readfds)
1063	\|\| WIN_CHECK_FD_SET(so, NetworkEvents, acceptds))
1064	{
1065
1066	#ifdef RT_OS_WINDOWS
1067	if (WIN_CHECK_FD_SET(so, NetworkEvents, connectfds))
1068	{
1069	/* Finish connection first */
1070	/* should we ignore return value? */
1071	bool fRet = slirpConnectOrWrite(pData, so, true);
1072	LogFunc(("fRet:%RTbool\n", fRet)); NOREF(fRet);
1073	if (slirpVerifyAndFreeSocket(pData, so))
1074	CONTINUE(tcp);
1075	}
1076	#endif
1077	/*
1078	* Check for incoming connections
1079	*/
1080	if (so->so_state & SS_FACCEPTCONN)
1081	{
1082	TCP_CONNECT(pData, so);
1083	if (slirpVerifyAndFreeSocket(pData, so))
1084	CONTINUE(tcp);
1085	if (!CHECK_FD_SET(so, NetworkEvents, closefds))
1086	{
1087	so->fUnderPolling = 0;
1088	CONTINUE(tcp);
1089	}
1090	}
1091
1092	ret = soread(pData, so);
1093	if (slirpVerifyAndFreeSocket(pData, so))
1094	CONTINUE(tcp);
1095	/* Output it if we read something */
1096	if (RT_LIKELY(ret > 0))
1097	TCP_OUTPUT(pData, sototcpcb(so));
1098
1099	if (slirpVerifyAndFreeSocket(pData, so))
1100	CONTINUE(tcp);
1101	}
1102
1103	/*
1104	* Check for FD_CLOSE events.
1105	* in some cases once FD_CLOSE engaged on socket it could be flashed latter (for some reasons)
1106	*/
1107	if ( CHECK_FD_SET(so, NetworkEvents, closefds)
1108	\|\| (so->so_close == 1))
1109	{
1110	/*
1111	* drain the socket
1112	*/
1113	for (; so_next->so_prev == so
1114	&& !slirpVerifyAndFreeSocket(pData, so);)
1115	{
1116	ret = soread(pData, so);
1117	if (slirpVerifyAndFreeSocket(pData, so))
1118	break;
1119
1120	if (ret > 0)
1121	TCP_OUTPUT(pData, sototcpcb(so));
1122	else if (so_next->so_prev == so)
1123	{
1124	Log2(("%R[natsock] errno %d (%s)\n", so, errno, strerror(errno)));
1125	break;
1126	}
1127	}
1128
1129	/* if socket freed ''so'' is PHANTOM and next socket isn't points on it */
1130	if (so_next->so_prev != so)
1131	{
1132	CONTINUE(tcp);
1133	}
1134	else
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	}
1146
1147	/*
1148	* Check sockets for writing
1149	*/
1150	if ( CHECK_FD_SET(so, NetworkEvents, writefds)
1151	#ifdef RT_OS_WINDOWS
1152	\|\| WIN_CHECK_FD_SET(so, NetworkEvents, connectfds)
1153	#endif
1154	)
1155	{
1156	int fConnectOrWriteSuccess = slirpConnectOrWrite(pData, so, false);
1157	/* slirpConnectOrWrite could return true even if tcp_input called tcp_drop,
1158	* so we should be ready to such situations.
1159	*/
1160	if (slirpVerifyAndFreeSocket(pData, so))
1161	CONTINUE(tcp);
1162	else if (!fConnectOrWriteSuccess)
1163	{
1164	so->fUnderPolling = 0;
1165	CONTINUE(tcp);
1166	}
1167	/* slirpConnectionOrWrite succeeded and socket wasn't dropped */
1168	}
1169
1170	/*
1171	* Probe a still-connecting, non-blocking socket
1172	* to check if it's still alive
1173	*/
1174	#ifdef PROBE_CONN
1175	if (so->so_state & SS_ISFCONNECTING)
1176	{
1177	ret = recv(so->s, (char *)&ret, 0, 0);
1178
1179	if (ret < 0)
1180	{
1181	/* XXX */
1182	if ( soIgnorableErrorCode(errno)
1183	\|\| errno == ENOTCONN)
1184	{
1185	CONTINUE(tcp); /* Still connecting, continue */
1186	}
1187
1188	/* else failed */
1189	so->so_state = SS_NOFDREF;
1190
1191	/* tcp_input will take care of it */
1192	}
1193	else
1194	{
1195	ret = send(so->s, &ret, 0, 0);
1196	if (ret < 0)
1197	{
1198	/* XXX */
1199	if ( soIgnorableErrorCode(errno)
1200	\|\| errno == ENOTCONN)
1201	{
1202	CONTINUE(tcp);
1203	}
1204	/* else failed */
1205	so->so_state = SS_NOFDREF;
1206	}
1207	else
1208	so->so_state &= ~SS_ISFCONNECTING;
1209
1210	}
1211	TCP_INPUT((struct mbuf *)NULL, sizeof(struct ip),so);
1212	} /* SS_ISFCONNECTING */
1213	#endif
1214	if (!slirpVerifyAndFreeSocket(pData, so))
1215	so->fUnderPolling = 0;
1216	LOOP_LABEL(tcp, so, so_next);
1217	}
1218
1219	/*
1220	* Now UDP sockets.
1221	* Incoming packets are sent straight away, they're not buffered.
1222	* Incoming UDP data isn't buffered either.
1223	*/
1224	QSOCKET_FOREACH(so, so_next, udp)
1225	/* { */
1226	#ifdef VBOX_WITH_NAT_UDP_SOCKET_CLONE
1227	if (so->so_cloneOf)
1228	CONTINUE_NO_UNLOCK(udp);
1229	#endif
1230	#if 0
1231	so->fUnderPolling = 1;
1232	if(slirpVerifyAndFreeSocket(pData, so));
1233	CONTINUE(udp);
1234	so->fUnderPolling = 0;
1235	#endif
1236
1237	POLL_UDP_EVENTS(rc, error, so, &NetworkEvents);
1238
1239	LOG_NAT_SOCK(so, UDP, &NetworkEvents, readfds, writefds, xfds);
1240
1241	if (so->s != -1 && CHECK_FD_SET(so, NetworkEvents, readfds))
1242	{
1243	SORECVFROM(pData, so);
1244	}
1245	LOOP_LABEL(udp, so, so_next);
1246	}
1247
1248	done:
1249
1250	STAM_PROFILE_STOP(&pData->StatPoll, a);
1251	}
1252
1253
1254	struct arphdr
1255	{
1256	unsigned short ar_hrd; /* format of hardware address */
1257	unsigned short ar_pro; /* format of protocol address */
1258	unsigned char ar_hln; /* length of hardware address */
1259	unsigned char ar_pln; /* length of protocol address */
1260	unsigned short ar_op; /* ARP opcode (command) */
1261
1262	/*
1263	* Ethernet looks like this : This bit is variable sized however...
1264	*/
1265	unsigned char ar_sha[ETH_ALEN]; /* sender hardware address */
1266	unsigned char ar_sip[4]; /* sender IP address */
1267	unsigned char ar_tha[ETH_ALEN]; /* target hardware address */
1268	unsigned char ar_tip[4]; /* target IP address */
1269	};
1270	AssertCompileSize(struct arphdr, 28);
1271
1272	static void arp_output(PNATState pData, const uint8_t pcu8EtherSource, const struct arphdr pcARPHeaderSource, uint32_t ip4TargetAddress)
1273	{
1274	struct ethhdr *pEtherHeaderResponse;
1275	struct arphdr *pARPHeaderResponse;
1276	uint32_t ip4TargetAddressInHostFormat;
1277	struct mbuf *pMbufResponse;
1278
1279	Assert((pcu8EtherSource));
1280	if (!pcu8EtherSource)
1281	return;
1282	ip4TargetAddressInHostFormat = RT_N2H_U32(ip4TargetAddress);
1283
1284	pMbufResponse = m_getcl(pData, M_NOWAIT, MT_HEADER, M_PKTHDR);
1285	if (!pMbufResponse)
1286	return;
1287	pEtherHeaderResponse = mtod(pMbufResponse, struct ethhdr *);
1288	/* @note: if_encap will swap src and dst*/
1289	memcpy(pEtherHeaderResponse->h_source, pcu8EtherSource, ETH_ALEN);
1290	pMbufResponse->m_data += ETH_HLEN;
1291	pARPHeaderResponse = mtod(pMbufResponse, struct arphdr *);
1292	pMbufResponse->m_len = sizeof(struct arphdr);
1293
1294	pARPHeaderResponse->ar_hrd = RT_H2N_U16_C(1);
1295	pARPHeaderResponse->ar_pro = RT_H2N_U16_C(ETH_P_IP);
1296	pARPHeaderResponse->ar_hln = ETH_ALEN;
1297	pARPHeaderResponse->ar_pln = 4;
1298	pARPHeaderResponse->ar_op = RT_H2N_U16_C(ARPOP_REPLY);
1299	memcpy(pARPHeaderResponse->ar_sha, special_ethaddr, ETH_ALEN);
1300
1301	if (!slirpMbufTagService(pData, pMbufResponse, (uint8_t)(ip4TargetAddressInHostFormat & ~pData->netmask)))
1302	{
1303	static bool fTagErrorReported;
1304	if (!fTagErrorReported)
1305	{
1306	LogRel(("NAT: Couldn't add the tag(PACKET_SERVICE:%d)\n",
1307	(uint8_t)(ip4TargetAddressInHostFormat & ~pData->netmask)));
1308	fTagErrorReported = true;
1309	}
1310	}
1311	pARPHeaderResponse->ar_sha[5] = (uint8_t)(ip4TargetAddressInHostFormat & ~pData->netmask);
1312
1313	memcpy(pARPHeaderResponse->ar_sip, pcARPHeaderSource->ar_tip, 4);
1314	memcpy(pARPHeaderResponse->ar_tha, pcARPHeaderSource->ar_sha, ETH_ALEN);
1315	memcpy(pARPHeaderResponse->ar_tip, pcARPHeaderSource->ar_sip, 4);
1316	if_encap(pData, ETH_P_ARP, pMbufResponse, ETH_ENCAP_URG);
1317	}
1318
1319	/**
1320	* @note This function will free m!
1321	*/
1322	static void arp_input(PNATState pData, struct mbuf *m)
1323	{
1324	struct ethhdr *pEtherHeader;
1325	struct arphdr *pARPHeader;
1326	uint32_t ip4TargetAddress;
1327
1328	int ar_op;
1329	pEtherHeader = mtod(m, struct ethhdr *);
1330	pARPHeader = (struct arphdr *)&pEtherHeader[1];
1331
1332	ar_op = RT_N2H_U16(pARPHeader->ar_op);
1333	ip4TargetAddress = (uint32_t)pARPHeader->ar_tip;
1334
1335	switch (ar_op)
1336	{
1337	case ARPOP_REQUEST:
1338	if ( CTL_CHECK(ip4TargetAddress, CTL_DNS)
1339	\|\| CTL_CHECK(ip4TargetAddress, CTL_ALIAS)
1340	\|\| CTL_CHECK(ip4TargetAddress, CTL_TFTP))
1341	{
1342	slirp_update_guest_addr_guess(pData, (uint32_t )pARPHeader->ar_sip, "arp request");
1343	arp_output(pData, pEtherHeader->h_source, pARPHeader, ip4TargetAddress);
1344	break;
1345	}
1346
1347	/* Gratuitous ARP */
1348	if ( (uint32_t )pARPHeader->ar_sip == (uint32_t )pARPHeader->ar_tip
1349	&& ( memcmp(pARPHeader->ar_tha, zerro_ethaddr, ETH_ALEN) == 0
1350	\|\| memcmp(pARPHeader->ar_tha, broadcast_ethaddr, ETH_ALEN) == 0)
1351	&& memcmp(pEtherHeader->h_dest, broadcast_ethaddr, ETH_ALEN) == 0)
1352	{
1353	LogRel2(("NAT: Gratuitous ARP from %RTnaipv4 at %RTmac\n",
1354	(uint32_t )pARPHeader->ar_sip, pARPHeader->ar_sha));
1355	slirp_update_guest_addr_guess(pData, (uint32_t )pARPHeader->ar_sip, "gratuitous arp");
1356	slirp_arp_cache_update_or_add(pData, (uint32_t )pARPHeader->ar_sip, &pARPHeader->ar_sha[0]);
1357	}
1358	break;
1359
1360	case ARPOP_REPLY:
1361	slirp_arp_cache_update_or_add(pData, (uint32_t )pARPHeader->ar_sip, &pARPHeader->ar_sha[0]);
1362	break;
1363
1364	default:
1365	break;
1366	}
1367
1368	m_freem(pData, m);
1369	}
1370
1371	/**
1372	* Feed a packet into the slirp engine.
1373	*
1374	* @param m Data buffer, m_len is not valid.
1375	* @param cbBuf The length of the data in m.
1376	*/
1377	void slirp_input(PNATState pData, struct mbuf *m, size_t cbBuf)
1378	{
1379	int proto;
1380	static bool fWarnedIpv6;
1381	struct ethhdr *eh;
1382
1383	m->m_len = (int)cbBuf; Assert((size_t)m->m_len == cbBuf);
1384	if (cbBuf < ETH_HLEN)
1385	{
1386	Log(("NAT: packet having size %d has been ignored\n", m->m_len));
1387	m_freem(pData, m);
1388	return;
1389	}
1390
1391	eh = mtod(m, struct ethhdr *);
1392	proto = RT_N2H_U16(eh->h_proto);
1393	switch(proto)
1394	{
1395	case ETH_P_ARP:
1396	arp_input(pData, m);
1397	break;
1398
1399	case ETH_P_IP:
1400	/* Update time. Important if the network is very quiet, as otherwise
1401	* the first outgoing connection gets an incorrect timestamp. */
1402	updtime(pData);
1403	m_adj(m, ETH_HLEN);
1404	M_ASSERTPKTHDR(m);
1405	m->m_pkthdr.header = mtod(m, void *);
1406	ip_input(pData, m);
1407	break;
1408
1409	case ETH_P_IPV6:
1410	m_freem(pData, m);
1411	if (!fWarnedIpv6)
1412	{
1413	LogRel(("NAT: IPv6 not supported\n"));
1414	fWarnedIpv6 = true;
1415	}
1416	break;
1417
1418	default:
1419	Log(("NAT: Unsupported protocol %x\n", proto));
1420	m_freem(pData, m);
1421	break;
1422	}
1423	}
1424
1425	/**
1426	* Output the IP packet to the ethernet device.
1427	*
1428	* @note This function will free m!
1429	*/
1430	void if_encap(PNATState pData, uint16_t eth_proto, struct mbuf *m, int flags)
1431	{
1432	struct ethhdr *eh;
1433	uint8_t *mbuf = NULL;
1434	int mlen;
1435	STAM_PROFILE_START(&pData->StatIF_encap, a);
1436	LogFlowFunc(("ENTER: pData:%p, eth_proto:%RX16, m:%p, flags:%d\n",
1437	pData, eth_proto, m, flags));
1438
1439	M_ASSERTPKTHDR(m);
1440
1441	Assert(M_LEADINGSPACE(m) >= ETH_HLEN);
1442	m->m_data -= ETH_HLEN;
1443	m->m_len += ETH_HLEN;
1444	eh = mtod(m, struct ethhdr *);
1445	mlen = m->m_len;
1446
1447	if (memcmp(eh->h_source, special_ethaddr, ETH_ALEN) != 0)
1448	{
1449	struct m_tag *t = m_tag_first(m);
1450	uint8_t u8ServiceId = CTL_ALIAS;
1451	memcpy(eh->h_dest, eh->h_source, ETH_ALEN);
1452	memcpy(eh->h_source, special_ethaddr, ETH_ALEN);
1453	Assert(memcmp(eh->h_dest, special_ethaddr, ETH_ALEN) != 0);
1454	if (memcmp(eh->h_dest, zerro_ethaddr, ETH_ALEN) == 0)
1455	{
1456	/* don't do anything */
1457	m_freem(pData, m);
1458	goto done;
1459	}
1460	if ( t
1461	&& (t = m_tag_find(m, PACKET_SERVICE, NULL)))
1462	{
1463	Assert(t);
1464	u8ServiceId = (uint8_t )&t[1];
1465	}
1466	eh->h_source[5] = u8ServiceId;
1467	}
1468	/*
1469	* we're processing the chain, that isn't not expected.
1470	*/
1471	Assert((!m->m_next));
1472	if (m->m_next)
1473	{
1474	Log(("NAT: if_encap's recived the chain, dropping...\n"));
1475	m_freem(pData, m);
1476	goto done;
1477	}
1478	mbuf = mtod(m, uint8_t *);
1479	eh->h_proto = RT_H2N_U16(eth_proto);
1480	LogFunc(("eh(dst:%RTmac, src:%RTmac)\n", eh->h_dest, eh->h_source));
1481	if (flags & ETH_ENCAP_URG)
1482	slirp_urg_output(pData->pvUser, m, mbuf, mlen);
1483	else
1484	slirp_output(pData->pvUser, m, mbuf, mlen);
1485	done:
1486	STAM_PROFILE_STOP(&pData->StatIF_encap, a);
1487	LogFlowFuncLeave();
1488	}
1489
1490
1491	void
1492	slirp_update_guest_addr_guess(PNATState pData, uint32_t guess, const char *msg)
1493	{
1494	Assert(msg != NULL);
1495
1496	if (pData->guest_addr_guess.s_addr == guess)
1497	{
1498	LogRel2(("NAT: Guest address guess %RTnaipv4 re-confirmed by %s\n",
1499	pData->guest_addr_guess.s_addr, msg));
1500	return;
1501	}
1502
1503	if (pData->guest_addr_guess.s_addr == INADDR_ANY)
1504	{
1505	pData->guest_addr_guess.s_addr = guess;
1506	LogRel(("NAT: Guest address guess set to %RTnaipv4 by %s\n",
1507	pData->guest_addr_guess.s_addr, msg));
1508	return;
1509	}
1510	else
1511	{
1512	LogRel(("NAT: Guest address guess changed from %RTnaipv4 to %RTnaipv4 by %s\n",
1513	pData->guest_addr_guess.s_addr, guess, msg));
1514	pData->guest_addr_guess.s_addr = guess;
1515	return;
1516	}
1517	}
1518
1519
1520	static struct port_forward_rule *
1521	slirp_find_redirect(PNATState pData,
1522	int is_udp,
1523	struct in_addr host_addr, int host_port,
1524	struct in_addr guest_addr, int guest_port)
1525	{
1526	struct port_forward_rule *rule;
1527	uint16_t proto = (is_udp ? IPPROTO_UDP : IPPROTO_TCP);
1528
1529	LIST_FOREACH(rule, &pData->port_forward_rule_head, list)
1530	{
1531	if ( rule->proto == proto
1532	&& rule->host_port == host_port
1533	&& rule->bind_ip.s_addr == host_addr.s_addr
1534	&& rule->guest_port == guest_port
1535	&& rule->guest_addr.s_addr == guest_addr.s_addr)
1536	{
1537	return rule;
1538	}
1539	}
1540
1541	return NULL;
1542	}
1543
1544
1545	int slirp_add_redirect(PNATState pData, int is_udp, struct in_addr host_addr, int host_port,
1546	struct in_addr guest_addr, int guest_port)
1547	{
1548	struct port_forward_rule *rule;
1549
1550	rule = slirp_find_redirect(pData, is_udp, host_addr, host_port, guest_addr, guest_port);
1551	if (rule != NULL) /* rule has been already registered */
1552	{
1553	/* XXX: this shouldn't happen */
1554	return 0;
1555	}
1556
1557	rule = RTMemAllocZ(sizeof(struct port_forward_rule));
1558	if (rule == NULL)
1559	return 1;
1560
1561	rule->proto = (is_udp ? IPPROTO_UDP : IPPROTO_TCP);
1562	rule->bind_ip.s_addr = host_addr.s_addr;
1563	rule->host_port = host_port;
1564	rule->guest_addr.s_addr = guest_addr.s_addr;
1565	rule->guest_port = guest_port;
1566
1567	if (rule->proto == IPPROTO_UDP)
1568	rule->so = udp_listen(pData, rule->bind_ip.s_addr, RT_H2N_U16(rule->host_port),
1569	rule->guest_addr.s_addr, RT_H2N_U16(rule->guest_port), 0);
1570	else
1571	rule->so = solisten(pData, rule->bind_ip.s_addr, RT_H2N_U16(rule->host_port),
1572	rule->guest_addr.s_addr, RT_H2N_U16(rule->guest_port), 0);
1573
1574	if (rule->so == NULL)
1575	{
1576	LogRel(("NAT: Failed to redirect %s %RTnaipv4:%d -> %RTnaipv4:%d (%s)\n",
1577	rule->proto == IPPROTO_UDP ? "UDP" : "TCP",
1578	rule->bind_ip.s_addr, rule->host_port,
1579	guest_addr, rule->guest_port, strerror(errno)));
1580	RTMemFree(rule);
1581	return 1;
1582	}
1583
1584	LogRel(("NAT: Set redirect %s %RTnaipv4:%d -> %RTnaipv4:%d\n",
1585	rule->proto == IPPROTO_UDP ? "UDP" : "TCP",
1586	rule->bind_ip.s_addr, rule->host_port,
1587	guest_addr, rule->guest_port));
1588
1589	LIST_INSERT_HEAD(&pData->port_forward_rule_head, rule, list);
1590	return 0;
1591	}
1592
1593
1594	int slirp_remove_redirect(PNATState pData, int is_udp, struct in_addr host_addr, int host_port,
1595	struct in_addr guest_addr, int guest_port)
1596	{
1597	struct port_forward_rule *rule;
1598
1599	rule = slirp_find_redirect(pData, is_udp, host_addr, host_port, guest_addr, guest_port);
1600	if (rule == NULL)
1601	{
1602	LogRel(("NAT: Unable to find redirect %s %RTnaipv4:%d -> %RTnaipv4:%d\n",
1603	is_udp ? "UDP" : "TCP",
1604	host_addr.s_addr, host_port,
1605	guest_addr.s_addr, guest_port));
1606	return 0;
1607	}
1608
1609	LogRel(("NAT: Remove redirect %s %RTnaipv4:%d -> %RTnaipv4:%d\n",
1610	rule->proto == IPPROTO_UDP ? "UDP" : "TCP",
1611	rule->bind_ip.s_addr, rule->host_port,
1612	guest_addr.s_addr, rule->guest_port));
1613
1614	if (rule->so != NULL)
1615	{
1616	if (is_udp)
1617	udp_detach(pData, rule->so);
1618	else
1619	tcp_close(pData, sototcpcb(rule->so));
1620	}
1621
1622	LIST_REMOVE(rule, list);
1623	RTMemFree(rule);
1624	return 0;
1625	}
1626
1627
1628	#if defined(RT_OS_WINDOWS)
1629	HANDLE *slirp_get_events(PNATState pData)
1630	{
1631	return pData->phEvents;
1632	}
1633	void slirp_register_external_event(PNATState pData, HANDLE hEvent, int index)
1634	{
1635	pData->phEvents[index] = hEvent;
1636	}
1637	#endif
1638
1639	unsigned int slirp_get_timeout_ms(PNATState pData)
1640	{
1641	if (link_up)
1642	{
1643	if (time_fasttimo)
1644	return 2;
1645	if (do_slowtimo)
1646	return 500; /* see PR_SLOWHZ */
1647	}
1648	return 36001000; / one hour */
1649	}
1650
1651	#ifndef RT_OS_WINDOWS
1652	int slirp_get_nsock(PNATState pData)
1653	{
1654	return pData->nsock;
1655	}
1656	#endif
1657
1658	/*
1659	* this function called from NAT thread
1660	*/
1661	void slirp_post_sent(PNATState pData, void *pvArg)
1662	{
1663	struct mbuf m = (struct mbuf )pvArg;
1664	m_freem(pData, m);
1665	}
1666
1667	void slirp_set_dhcp_TFTP_prefix(PNATState pData, const char *tftpPrefix)
1668	{
1669	Log2(("tftp_prefix: %s\n", tftpPrefix));
1670	tftp_prefix = tftpPrefix;
1671	}
1672
1673	void slirp_set_dhcp_TFTP_bootfile(PNATState pData, const char *bootFile)
1674	{
1675	Log2(("bootFile: %s\n", bootFile));
1676	bootp_filename = bootFile;
1677	}
1678
1679	void slirp_set_dhcp_next_server(PNATState pData, const char *next_server)
1680	{
1681	Log2(("next_server: %s\n", next_server));
1682	if (next_server == NULL)
1683	pData->tftp_server.s_addr = RT_H2N_U32(RT_N2H_U32(pData->special_addr.s_addr) \| CTL_TFTP);
1684	else
1685	inet_aton(next_server, &pData->tftp_server);
1686	}
1687
1688	int slirp_set_binding_address(PNATState pData, char *addr)
1689	{
1690	if (addr == NULL \|\| (inet_aton(addr, &pData->bindIP) == 0))
1691	{
1692	pData->bindIP.s_addr = INADDR_ANY;
1693	return 1;
1694	}
1695	return 0;
1696	}
1697
1698	void slirp_set_dhcp_dns_proxy(PNATState pData, bool fDNSProxy)
1699	{
1700	if (!pData->fUseHostResolver)
1701	{
1702	Log2(("NAT: DNS proxy switched %s\n", (fDNSProxy ? "on" : "off")));
1703	pData->fUseDnsProxy = fDNSProxy;
1704	}
1705	else if (fDNSProxy)
1706	LogRel(("NAT: Host Resolver conflicts with DNS proxy, the last one was forcely ignored\n"));
1707	}
1708
1709	#define CHECK_ARG(name, val, lim_min, lim_max) \
1710	do { \
1711	if ((val) < (lim_min) \|\| (val) > (lim_max)) \
1712	{ \
1713	LogRel(("NAT: (" #name ":%d) has been ignored, " \
1714	"because out of range (%d, %d)\n", (val), (lim_min), (lim_max))); \
1715	return; \
1716	} \
1717	else \
1718	LogRel(("NAT: (" #name ":%d)\n", (val))); \
1719	} while (0)
1720
1721	void slirp_set_somaxconn(PNATState pData, int iSoMaxConn)
1722	{
1723	LogFlowFunc(("iSoMaxConn:%d\n", iSoMaxConn));
1724	/* Conditions */
1725	if (iSoMaxConn > SOMAXCONN)
1726	{
1727	LogRel(("NAT: value of somaxconn(%d) bigger than SOMAXCONN(%d)\n", iSoMaxConn, SOMAXCONN));
1728	iSoMaxConn = SOMAXCONN;
1729	}
1730
1731	if (iSoMaxConn < 1)
1732	{
1733	LogRel(("NAT: proposed value(%d) of somaxconn is invalid, default value is used (%d)\n", iSoMaxConn, pData->soMaxConn));
1734	LogFlowFuncLeave();
1735	return;
1736	}
1737
1738	/* Asignment */
1739	if (pData->soMaxConn != iSoMaxConn)
1740	{
1741	LogRel(("NAT: value of somaxconn has been changed from %d to %d\n",
1742	pData->soMaxConn, iSoMaxConn));
1743	pData->soMaxConn = iSoMaxConn;
1744	}
1745	LogFlowFuncLeave();
1746	}
1747	/* don't allow user set less 8kB and more than 1M values */
1748	#define _8K_1M_CHECK_ARG(name, val) CHECK_ARG(name, (val), 8, 1024)
1749	void slirp_set_rcvbuf(PNATState pData, int kilobytes)
1750	{
1751	_8K_1M_CHECK_ARG("SOCKET_RCVBUF", kilobytes);
1752	pData->socket_rcv = kilobytes;
1753	}
1754	void slirp_set_sndbuf(PNATState pData, int kilobytes)
1755	{
1756	_8K_1M_CHECK_ARG("SOCKET_SNDBUF", kilobytes);
1757	pData->socket_snd = kilobytes * _1K;
1758	}
1759	void slirp_set_tcp_rcvspace(PNATState pData, int kilobytes)
1760	{
1761	_8K_1M_CHECK_ARG("TCP_RCVSPACE", kilobytes);
1762	tcp_rcvspace = kilobytes * _1K;
1763	}
1764	void slirp_set_tcp_sndspace(PNATState pData, int kilobytes)
1765	{
1766	_8K_1M_CHECK_ARG("TCP_SNDSPACE", kilobytes);
1767	tcp_sndspace = kilobytes * _1K;
1768	}
1769
1770	/*
1771	* Looking for Ether by ip in ARP-cache
1772	* Note: it´s responsible of caller to allocate buffer for result
1773	* @returns iprt status code
1774	*/
1775	int slirp_arp_lookup_ether_by_ip(PNATState pData, uint32_t ip, uint8_t *ether)
1776	{
1777	struct arp_cache_entry *ac;
1778
1779	if (ether == NULL)
1780	return VERR_INVALID_PARAMETER;
1781
1782	if (LIST_EMPTY(&pData->arp_cache))
1783	return VERR_NOT_FOUND;
1784
1785	LIST_FOREACH(ac, &pData->arp_cache, list)
1786	{
1787	if ( ac->ip == ip
1788	&& memcmp(ac->ether, broadcast_ethaddr, ETH_ALEN) != 0)
1789	{
1790	memcpy(ether, ac->ether, ETH_ALEN);
1791	return VINF_SUCCESS;
1792	}
1793	}
1794	return VERR_NOT_FOUND;
1795	}
1796
1797	/*
1798	* Looking for IP by Ether in ARP-cache
1799	* Note: it´s responsible of caller to allocate buffer for result
1800	* @returns 0 - if found, 1 - otherwise
1801	*/
1802	int slirp_arp_lookup_ip_by_ether(PNATState pData, const uint8_t ether, uint32_t ip)
1803	{
1804	struct arp_cache_entry *ac;
1805	*ip = INADDR_ANY;
1806
1807	if (LIST_EMPTY(&pData->arp_cache))
1808	return VERR_NOT_FOUND;
1809
1810	LIST_FOREACH(ac, &pData->arp_cache, list)
1811	{
1812	if (memcmp(ether, ac->ether, ETH_ALEN) == 0)
1813	{
1814	*ip = ac->ip;
1815	return VINF_SUCCESS;
1816	}
1817	}
1818	return VERR_NOT_FOUND;
1819	}
1820
1821	void slirp_arp_who_has(PNATState pData, uint32_t dst)
1822	{
1823	struct mbuf *m;
1824	struct ethhdr *ehdr;
1825	struct arphdr *ahdr;
1826	static bool fWarned = false;
1827	LogFlowFunc(("ENTER: %RTnaipv4\n", dst));
1828
1829	/* ARP request WHO HAS 0.0.0.0 is one of the signals
1830	* that something has been broken at Slirp. Investigating
1831	* pcap dumps it's easy to miss warning ARP requests being
1832	* focused on investigation of other protocols flow.
1833	*/
1834	#ifdef DEBUG_vvl
1835	Assert((dst != INADDR_ANY));
1836	NOREF(fWarned);
1837	#else
1838	if ( dst == INADDR_ANY
1839	&& !fWarned)
1840	{
1841	LogRel(("NAT: ARP: \"WHO HAS INADDR_ANY\" request has been detected\n"));
1842	fWarned = true;
1843	}
1844	#endif /* !DEBUG_vvl */
1845
1846	m = m_getcl(pData, M_NOWAIT, MT_HEADER, M_PKTHDR);
1847	if (m == NULL)
1848	{
1849	Log(("NAT: Can't alloc mbuf for ARP request\n"));
1850	LogFlowFuncLeave();
1851	return;
1852	}
1853	ehdr = mtod(m, struct ethhdr *);
1854	memset(ehdr->h_source, 0xff, ETH_ALEN);
1855	ahdr = (struct arphdr *)&ehdr[1];
1856	ahdr->ar_hrd = RT_H2N_U16_C(1);
1857	ahdr->ar_pro = RT_H2N_U16_C(ETH_P_IP);
1858	ahdr->ar_hln = ETH_ALEN;
1859	ahdr->ar_pln = 4;
1860	ahdr->ar_op = RT_H2N_U16_C(ARPOP_REQUEST);
1861	memcpy(ahdr->ar_sha, special_ethaddr, ETH_ALEN);
1862	/* we assume that this request come from gw, but not from DNS or TFTP */
1863	ahdr->ar_sha[5] = CTL_ALIAS;
1864	(uint32_t )ahdr->ar_sip = RT_H2N_U32(RT_N2H_U32(pData->special_addr.s_addr) \| CTL_ALIAS);
1865	memset(ahdr->ar_tha, 0xff, ETH_ALEN); /broadcast/
1866	(uint32_t )ahdr->ar_tip = dst;
1867	/* warn!!! should falls in mbuf minimal size */
1868	m->m_len = sizeof(struct arphdr) + ETH_HLEN;
1869	m->m_data += ETH_HLEN;
1870	m->m_len -= ETH_HLEN;
1871	if_encap(pData, ETH_P_ARP, m, ETH_ENCAP_URG);
1872	LogFlowFuncLeave();
1873	}
1874
1875
1876	/* updates the arp cache
1877	* @note: this is helper function, slirp_arp_cache_update_or_add should be used.
1878	* @returns 0 - if has found and updated
1879	* 1 - if hasn't found.
1880	*/
1881	static inline int slirp_arp_cache_update(PNATState pData, uint32_t dst, const uint8_t *mac)
1882	{
1883	struct arp_cache_entry *ac;
1884	Assert(( memcmp(mac, broadcast_ethaddr, ETH_ALEN)
1885	&& memcmp(mac, zerro_ethaddr, ETH_ALEN)));
1886	LIST_FOREACH(ac, &pData->arp_cache, list)
1887	{
1888	if (ac->ip == dst)
1889	{
1890	memcpy(ac->ether, mac, ETH_ALEN);
1891	return 0;
1892	}
1893	}
1894	return 1;
1895	}
1896
1897	/**
1898	* add entry to the arp cache
1899	* @note: this is helper function, slirp_arp_cache_update_or_add should be used.
1900	*/
1901	static inline void slirp_arp_cache_add(PNATState pData, uint32_t ip, const uint8_t *ether)
1902	{
1903	struct arp_cache_entry *ac = NULL;
1904	Assert(( memcmp(ether, broadcast_ethaddr, ETH_ALEN)
1905	&& memcmp(ether, zerro_ethaddr, ETH_ALEN)));
1906	ac = RTMemAllocZ(sizeof(struct arp_cache_entry));
1907	if (ac == NULL)
1908	{
1909	Log(("NAT: Can't allocate arp cache entry\n"));
1910	return;
1911	}
1912	ac->ip = ip;
1913	memcpy(ac->ether, ether, ETH_ALEN);
1914	LIST_INSERT_HEAD(&pData->arp_cache, ac, list);
1915	}
1916
1917	/* updates or adds entry to the arp cache
1918	* @returns 0 - if has found and updated
1919	* 1 - if hasn't found.
1920	*/
1921	int slirp_arp_cache_update_or_add(PNATState pData, uint32_t dst, const uint8_t *mac)
1922	{
1923	if ( !memcmp(mac, broadcast_ethaddr, ETH_ALEN)
1924	\|\| !memcmp(mac, zerro_ethaddr, ETH_ALEN))
1925	{
1926	static bool fBroadcastEtherAddReported;
1927	if (!fBroadcastEtherAddReported)
1928	{
1929	LogRel(("NAT: Attempt to add pair [%RTmac:%RTnaipv4] in ARP cache was ignored\n",
1930	mac, dst));
1931	fBroadcastEtherAddReported = true;
1932	}
1933	return 1;
1934	}
1935	if (slirp_arp_cache_update(pData, dst, mac))
1936	slirp_arp_cache_add(pData, dst, mac);
1937
1938	return 0;
1939	}
1940
1941
1942	void slirp_set_mtu(PNATState pData, int mtu)
1943	{
1944	if (mtu < 20 \|\| mtu >= 16000)
1945	{
1946	LogRel(("NAT: MTU(%d) is out of range (20;16000] mtu forcely assigned to 1500\n", mtu));
1947	mtu = 1500;
1948	}
1949	/* MTU is maximum transition unit on */
1950	if_mtu =
1951	if_mru = mtu;
1952	}
1953
1954	/**
1955	* Info handler.
1956	*/
1957	void slirp_info(PNATState pData, const void pvArg, const char pszArgs)
1958	{
1959	struct socket so, so_next;
1960	struct arp_cache_entry *ac;
1961	struct port_forward_rule *rule;
1962	PCDBGFINFOHLP pHlp = (PCDBGFINFOHLP)pvArg;
1963	NOREF(pszArgs);
1964
1965	pHlp->pfnPrintf(pHlp, "NAT parameters: MTU=%d\n", if_mtu);
1966	pHlp->pfnPrintf(pHlp, "NAT TCP ports:\n");
1967	QSOCKET_FOREACH(so, so_next, tcp)
1968	/* { */
1969	pHlp->pfnPrintf(pHlp, " %R[natsock]\n", so);
1970	}
1971
1972	pHlp->pfnPrintf(pHlp, "NAT UDP ports:\n");
1973	QSOCKET_FOREACH(so, so_next, udp)
1974	/* { */
1975	pHlp->pfnPrintf(pHlp, " %R[natsock]\n", so);
1976	}
1977
1978	pHlp->pfnPrintf(pHlp, "NAT ARP cache:\n");
1979	LIST_FOREACH(ac, &pData->arp_cache, list)
1980	{
1981	pHlp->pfnPrintf(pHlp, " %RTnaipv4 %RTmac\n", ac->ip, &ac->ether);
1982	}
1983
1984	pHlp->pfnPrintf(pHlp, "NAT rules:\n");
1985	LIST_FOREACH(rule, &pData->port_forward_rule_head, list)
1986	{
1987	pHlp->pfnPrintf(pHlp, " %s %d => %RTnaipv4:%d %c\n",
1988	rule->proto == IPPROTO_UDP ? "UDP" : "TCP",
1989	rule->host_port, rule->guest_addr.s_addr, rule->guest_port,
1990	rule->activated ? ' ' : '*');
1991	}
1992	}
1993
1994	/**
1995	* @note: NATState::fUseHostResolver could be changed in bootp.c::dhcp_decode
1996	* @note: this function is executed on GUI/VirtualBox or main/VBoxHeadless thread.
1997	* @note: this function can potentially race with bootp.c::dhcp_decode (except Darwin)
1998	*/
1999	int slirp_host_network_configuration_change_strategy_selector(const PNATState pData)
2000	{
2001	if (pData->fUseHostResolverPermanent)
2002	return VBOX_NAT_DNS_HOSTRESOLVER;
2003
2004	if (pData->fUseDnsProxy) {
2005	#if HAVE_NOTIFICATION_FOR_DNS_UPDATE /* XXX */ && !defined(RT_OS_WINDOWS)
2006	/* We dont conflict with bootp.c::dhcp_decode */
2007	struct rcp_state rcp_state;
2008	int rc;
2009
2010	rcp_state.rcps_flags \|= RCPSF_IGNORE_IPV6;
2011	rc = rcp_parse(&rcp_state, RESOLV_CONF_FILE);
2012	LogRelFunc(("NAT: rcp_parse:%Rrc old domain:%s new domain:%s\n",
2013	rc, LIST_EMPTY(&pData->pDomainList)
2014	? "(null)"
2015	: LIST_FIRST(&pData->pDomainList)->dd_pszDomain,
2016	rcp_state.rcps_domain));
2017	if ( RT_FAILURE(rc)
2018	\|\| LIST_EMPTY(&pData->pDomainList))
2019	return VBOX_NAT_DNS_DNSPROXY;
2020
2021	if ( rcp_state.rcps_domain
2022	&& strcmp(rcp_state.rcps_domain, LIST_FIRST(&pData->pDomainList)->dd_pszDomain) == 0)
2023	return VBOX_NAT_DNS_DNSPROXY;
2024	else
2025	return VBOX_NAT_DNS_EXTERNAL;
2026	#else
2027	/* copy domain name */
2028	/* domain only compare with coy version */
2029	return VBOX_NAT_DNS_DNSPROXY;
2030	#endif
2031	}
2032	return VBOX_NAT_DNS_EXTERNAL;
2033	}

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

source: vbox/trunk/src/VBox/Devices/Network/slirp/slirp.c@ 65455

Download in other formats: