slirp.c@ 55268

Last change on this file since 55268 was 54827, checked in by vboxsync, 10 years ago
NAT: G/c ancient abandoned VBOX_WITH_SLIRP_BSD_SBUF experiment.
Property svn:eol-style set to `native` Property svn:keywords set to `Author Date Id Revision`
File size: 67.2 KB

Line
1	/* $Id: slirp.c 54827 2015-03-18 04:59:50Z 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)
204
205	# define DO_WIN_CHECK_FD_SET(so, events, fdset) DO_CHECK_FD_SET((so), (events), fdset)
206	# define DO_UNIX_CHECK_FD_SET(so, events, fdset) 1 /specific for Unix API /
207
208	#endif /* RT_OS_WINDOWS */
209
210	#define TCP_ENGAGE_EVENT1(so, fdset) \
211	DO_ENGAGE_EVENT1((so), fdset, tcp)
212
213	#define TCP_ENGAGE_EVENT2(so, fdset1, fdset2) \
214	DO_ENGAGE_EVENT2((so), fdset1, fdset2, tcp)
215
216	#ifdef RT_OS_WINDOWS
217	# define WIN_TCP_ENGAGE_EVENT2(so, fdset, fdset2) TCP_ENGAGE_EVENT2(so, fdset1, fdset2)
218	#endif
219
220	#define UDP_ENGAGE_EVENT(so, fdset) \
221	DO_ENGAGE_EVENT1((so), fdset, udp)
222
223	#define POLL_TCP_EVENTS(rc, error, so, events) \
224	DO_POLL_EVENTS((rc), (error), (so), (events), tcp)
225
226	#define POLL_UDP_EVENTS(rc, error, so, events) \
227	DO_POLL_EVENTS((rc), (error), (so), (events), udp)
228
229	#define CHECK_FD_SET(so, events, set) \
230	(DO_CHECK_FD_SET((so), (events), set))
231
232	#define WIN_CHECK_FD_SET(so, events, set) \
233	(DO_WIN_CHECK_FD_SET((so), (events), set))
234
235	/*
236	* Loging macros
237	*/
238	#if VBOX_WITH_DEBUG_NAT_SOCKETS
239	# if defined(RT_OS_WINDOWS)
240	# define DO_LOG_NAT_SOCK(so, proto, winevent, r_fdset, w_fdset, x_fdset) \
241	do { \
242	LogRel((" " #proto " %R[natsock] %R[natwinnetevents]\n", (so), (winevent))); \
243	} while (0)
244	# else /* !RT_OS_WINDOWS */
245	# define DO_LOG_NAT_SOCK(so, proto, winevent, r_fdset, w_fdset, x_fdset) \
246	do { \
247	LogRel((" " #proto " %R[natsock] %s %s %s er: %s, %s, %s\n", (so), \
248	CHECK_FD_SET(so, ign ,r_fdset) ? "READ":"", \
249	CHECK_FD_SET(so, ign, w_fdset) ? "WRITE":"", \
250	CHECK_FD_SET(so, ign, x_fdset) ? "OOB":"", \
251	CHECK_FD_SET(so, ign, rderr) ? "RDERR":"", \
252	CHECK_FD_SET(so, ign, rdhup) ? "RDHUP":"", \
253	CHECK_FD_SET(so, ign, nval) ? "RDNVAL":"")); \
254	} while (0)
255	# endif /* !RT_OS_WINDOWS */
256	#else /* !VBOX_WITH_DEBUG_NAT_SOCKETS */
257	# define DO_LOG_NAT_SOCK(so, proto, winevent, r_fdset, w_fdset, x_fdset) do {} while (0)
258	#endif /* !VBOX_WITH_DEBUG_NAT_SOCKETS */
259
260	#define LOG_NAT_SOCK(so, proto, winevent, r_fdset, w_fdset, x_fdset) \
261	DO_LOG_NAT_SOCK((so), proto, (winevent), r_fdset, w_fdset, x_fdset)
262
263	static void activate_port_forwarding(PNATState, const uint8_t *pEther);
264
265	static const uint8_t special_ethaddr[6] =
266	{
267	0x52, 0x54, 0x00, 0x12, 0x35, 0x00
268	};
269
270	static const uint8_t broadcast_ethaddr[6] =
271	{
272	0xff, 0xff, 0xff, 0xff, 0xff, 0xff
273	};
274
275	const uint8_t zerro_ethaddr[6] =
276	{
277	0x0, 0x0, 0x0, 0x0, 0x0, 0x0
278	};
279
280	/**
281	* This helper routine do the checks in descriptions to
282	* ''fUnderPolling'' and ''fShouldBeRemoved'' flags
283	* @returns 1 if socket removed and 0 if no changes was made.
284	*/
285	static int slirpVerifyAndFreeSocket(PNATState pData, struct socket *pSocket)
286	{
287	AssertPtrReturn(pData, 0);
288	AssertPtrReturn(pSocket, 0);
289	AssertReturn(pSocket->fUnderPolling, 0);
290	if (pSocket->fShouldBeRemoved)
291	{
292	pSocket->fUnderPolling = 0;
293	sofree(pData, pSocket);
294	/* pSocket is PHANTOM, now */
295	return 1;
296	}
297	return 0;
298	}
299
300	int slirp_init(PNATState *ppData, uint32_t u32NetAddr, uint32_t u32Netmask,
301	bool fPassDomain, bool fUseHostResolver, int i32AliasMode,
302	int iIcmpCacheLimit, void *pvUser)
303	{
304	int rc;
305	PNATState pData;
306	if (u32Netmask & 0x1f)
307	/* CTL is x.x.x.15, bootp passes up to 16 IPs (15..31) */
308	return VERR_INVALID_PARAMETER;
309	pData = RTMemAllocZ(RT_ALIGN_Z(sizeof(NATState), sizeof(uint64_t)));
310	*ppData = pData;
311	if (!pData)
312	return VERR_NO_MEMORY;
313	pData->fPassDomain = !fUseHostResolver ? fPassDomain : false;
314	pData->fUseHostResolver = fUseHostResolver;
315	pData->fUseHostResolverPermanent = fUseHostResolver;
316	pData->pvUser = pvUser;
317	pData->netmask = u32Netmask;
318
319	rc = RTCritSectRwInit(&pData->CsRwHandlerChain);
320	if (RT_FAILURE(rc))
321	return rc;
322
323	/* sockets & TCP defaults */
324	pData->socket_rcv = 64 * _1K;
325	pData->socket_snd = 64 * _1K;
326	tcp_sndspace = 64 * _1K;
327	tcp_rcvspace = 64 * _1K;
328
329	/*
330	* Use the same default here as in DevNAT.cpp (SoMaxConnection CFGM value)
331	* to avoid release log noise.
332	*/
333	pData->soMaxConn = 10;
334
335	#ifdef RT_OS_WINDOWS
336	{
337	WSADATA Data;
338	RTLDRMOD hLdrMod;
339
340	WSAStartup(MAKEWORD(2, 0), &Data);
341
342	rc = RTLdrLoadSystem("Iphlpapi.dll", /* :fNoUnload */ true, &hLdrMod);
343	if (RT_SUCCESS(rc))
344	{
345	rc = RTLdrGetSymbol(hLdrMod, "GetAdaptersAddresses", (void **)&pData->pfGetAdaptersAddresses);
346	if (RT_FAILURE(rc))
347	LogRel(("NAT: Can't find GetAdapterAddresses in Iphlpapi.dll\n"));
348
349	RTLdrClose(hLdrMod);
350	}
351	}
352	pData->phEvents[VBOX_SOCKET_EVENT_INDEX] = CreateEvent(NULL, FALSE, FALSE, NULL);
353	#endif
354
355	rc = bootp_dhcp_init(pData);
356	if (RT_FAILURE(rc))
357	{
358	Log(("NAT: DHCP server initialization failed\n"));
359	RTMemFree(pData);
360	*ppData = NULL;
361	return rc;
362	}
363	debug_init(pData);
364	if_init(pData);
365	ip_init(pData);
366	icmp_init(pData, iIcmpCacheLimit);
367
368	/* Initialise mbufs after setting the MTU */
369	mbuf_init(pData);
370
371	pData->special_addr.s_addr = u32NetAddr;
372	pData->slirp_ethaddr = &special_ethaddr[0];
373	alias_addr.s_addr = pData->special_addr.s_addr \| RT_H2N_U32_C(CTL_ALIAS);
374	/* @todo: add ability to configure this staff */
375
376	/* set default addresses */
377	inet_aton("127.0.0.1", &loopback_addr);
378
379	rc = slirpTftpInit(pData);
380	AssertRCReturn(rc, VINF_NAT_DNS);
381
382	if (i32AliasMode & ~(PKT_ALIAS_LOG\|PKT_ALIAS_SAME_PORTS\|PKT_ALIAS_PROXY_ONLY))
383	{
384	Log(("NAT: alias mode %x is ignored\n", i32AliasMode));
385	i32AliasMode = 0;
386	}
387	pData->i32AliasMode = i32AliasMode;
388	getouraddr(pData);
389	{
390	int flags = 0;
391	struct in_addr proxy_addr;
392	pData->proxy_alias = LibAliasInit(pData, NULL);
393	if (pData->proxy_alias == NULL)
394	{
395	Log(("NAT: LibAlias default rule wasn't initialized\n"));
396	AssertMsgFailed(("NAT: LibAlias default rule wasn't initialized\n"));
397	}
398	flags = LibAliasSetMode(pData->proxy_alias, 0, 0);
399	#ifndef NO_FW_PUNCH
400	flags \|= PKT_ALIAS_PUNCH_FW;
401	#endif
402	flags \|= pData->i32AliasMode; /* do transparent proxying */
403	flags = LibAliasSetMode(pData->proxy_alias, flags, ~0);
404	proxy_addr.s_addr = RT_H2N_U32(RT_N2H_U32(pData->special_addr.s_addr) \| CTL_ALIAS);
405	LibAliasSetAddress(pData->proxy_alias, proxy_addr);
406	ftp_alias_load(pData);
407	nbt_alias_load(pData);
408	if (pData->fUseHostResolver)
409	dns_alias_load(pData);
410	}
411	#ifdef VBOX_WITH_NAT_SEND2HOME
412	/* @todo: we should know all interfaces available on host. */
413	pData->pInSockAddrHomeAddress = RTMemAllocZ(sizeof(struct sockaddr));
414	pData->cInHomeAddressSize = 1;
415	inet_aton("192.168.1.25", &pData->pInSockAddrHomeAddress[0].sin_addr);
416	pData->pInSockAddrHomeAddress[0].sin_family = AF_INET;
417	# ifdef RT_OS_DARWIN
418	pData->pInSockAddrHomeAddress[0].sin_len = sizeof(struct sockaddr_in);
419	# endif
420	#endif
421
422	slirp_link_up(pData);
423	return VINF_SUCCESS;
424	}
425
426	/**
427	* Register statistics.
428	*/
429	void slirp_register_statistics(PNATState pData, PPDMDRVINS pDrvIns)
430	{
431	#ifdef VBOX_WITH_STATISTICS
432	# define PROFILE_COUNTER(name, dsc) REGISTER_COUNTER(name, pData, STAMTYPE_PROFILE, STAMUNIT_TICKS_PER_CALL, dsc)
433	# define COUNTING_COUNTER(name, dsc) REGISTER_COUNTER(name, pData, STAMTYPE_COUNTER, STAMUNIT_COUNT, dsc)
434	# include "counters.h"
435	# undef COUNTER
436	/** @todo register statistics for the variables dumped by:
437	* ipstats(pData); tcpstats(pData); udpstats(pData); icmpstats(pData);
438	* mbufstats(pData); sockstats(pData); */
439	#else /* VBOX_WITH_STATISTICS */
440	NOREF(pData);
441	NOREF(pDrvIns);
442	#endif /* !VBOX_WITH_STATISTICS */
443	}
444
445	/**
446	* Deregister statistics.
447	*/
448	void slirp_deregister_statistics(PNATState pData, PPDMDRVINS pDrvIns)
449	{
450	if (pData == NULL)
451	return;
452	#ifdef VBOX_WITH_STATISTICS
453	# define PROFILE_COUNTER(name, dsc) DEREGISTER_COUNTER(name, pData)
454	# define COUNTING_COUNTER(name, dsc) DEREGISTER_COUNTER(name, pData)
455	# include "counters.h"
456	#else /* VBOX_WITH_STATISTICS */
457	NOREF(pData);
458	NOREF(pDrvIns);
459	#endif /* !VBOX_WITH_STATISTICS */
460	}
461
462	/**
463	* Marks the link as up, making it possible to establish new connections.
464	*/
465	void slirp_link_up(PNATState pData)
466	{
467	struct arp_cache_entry *ac;
468
469	if (link_up == 1)
470	return;
471
472	link_up = 1;
473
474	if (!pData->fUseHostResolverPermanent)
475	slirpInitializeDnsSettings(pData);
476
477	if (LIST_EMPTY(&pData->arp_cache))
478	return;
479
480	LIST_FOREACH(ac, &pData->arp_cache, list)
481	{
482	activate_port_forwarding(pData, ac->ether);
483	}
484	}
485
486	/**
487	* Marks the link as down and cleans up the current connections.
488	*/
489	void slirp_link_down(PNATState pData)
490	{
491	struct socket *so;
492	struct port_forward_rule *rule;
493
494	if (link_up == 0)
495	return;
496
497	slirpReleaseDnsSettings(pData);
498
499	while ((so = tcb.so_next) != &tcb)
500	{
501	/* Don't miss TCB releasing */
502	if ( !sototcpcb(so)
503	&& ( so->so_state & SS_NOFDREF
504	\|\| so->s == -1))
505	sofree(pData, so);
506	else
507	tcp_close(pData, sototcpcb(so));
508	}
509
510	while ((so = udb.so_next) != &udb)
511	udp_detach(pData, so);
512
513	/*
514	* Clear the active state of port-forwarding rules to force
515	* re-setup on restoration of communications.
516	*/
517	LIST_FOREACH(rule, &pData->port_forward_rule_head, list)
518	{
519	rule->activated = 0;
520	}
521	pData->cRedirectionsActive = 0;
522
523	link_up = 0;
524	}
525
526	/**
527	* Terminates the slirp component.
528	*/
529	void slirp_term(PNATState pData)
530	{
531	if (pData == NULL)
532	return;
533	icmp_finit(pData);
534
535	slirp_link_down(pData);
536	ftp_alias_unload(pData);
537	nbt_alias_unload(pData);
538	if (pData->fUseHostResolver)
539	{
540	dns_alias_unload(pData);
541	#ifdef VBOX_WITH_DNSMAPPING_IN_HOSTRESOLVER
542	while (!LIST_EMPTY(&pData->DNSMapHead))
543	{
544	PDNSMAPPINGENTRY pDnsEntry = LIST_FIRST(&pData->DNSMapHead);
545	LIST_REMOVE(pDnsEntry, MapList);
546	RTStrFree(pDnsEntry->pszCName);
547	RTMemFree(pDnsEntry);
548	}
549	#endif
550	}
551	while (!LIST_EMPTY(&instancehead))
552	{
553	struct libalias *la = LIST_FIRST(&instancehead);
554	/* libalias do all clean up */
555	LibAliasUninit(la);
556	}
557	while (!LIST_EMPTY(&pData->arp_cache))
558	{
559	struct arp_cache_entry *ac = LIST_FIRST(&pData->arp_cache);
560	LIST_REMOVE(ac, list);
561	RTMemFree(ac);
562	}
563	slirpTftpTerm(pData);
564	bootp_dhcp_fini(pData);
565	m_fini(pData);
566	#ifdef RT_OS_WINDOWS
567	WSACleanup();
568	#endif
569	#ifdef LOG_ENABLED
570	Log(("\n"
571	"NAT statistics\n"
572	"--------------\n"
573	"\n"));
574	ipstats(pData);
575	tcpstats(pData);
576	udpstats(pData);
577	icmpstats(pData);
578	mbufstats(pData);
579	sockstats(pData);
580	Log(("\n"
581	"\n"
582	"\n"));
583	#endif
584	RTCritSectRwDelete(&pData->CsRwHandlerChain);
585	RTMemFree(pData);
586	}
587
588
589	#define CONN_CANFSEND(so) (((so)->so_state & (SS_FCANTSENDMORE\|SS_ISFCONNECTED)) == SS_ISFCONNECTED)
590	#define CONN_CANFRCV(so) (((so)->so_state & (SS_FCANTRCVMORE\|SS_ISFCONNECTED)) == SS_ISFCONNECTED)
591
592	/*
593	* curtime kept to an accuracy of 1ms
594	*/
595	static void updtime(PNATState pData)
596	{
597	#ifdef RT_OS_WINDOWS
598	struct _timeb tb;
599
600	_ftime(&tb);
601	curtime = (u_int)tb.time * (u_int)1000;
602	curtime += (u_int)tb.millitm;
603	#else
604	gettimeofday(&tt, 0);
605
606	curtime = (u_int)tt.tv_sec * (u_int)1000;
607	curtime += (u_int)tt.tv_usec / (u_int)1000;
608
609	if ((tt.tv_usec % 1000) >= 500)
610	curtime++;
611	#endif
612	}
613
614	#ifdef RT_OS_WINDOWS
615	void slirp_select_fill(PNATState pData, int *pnfds)
616	#else /* RT_OS_WINDOWS */
617	void slirp_select_fill(PNATState pData, int pnfds, struct pollfd polls)
618	#endif /* !RT_OS_WINDOWS */
619	{
620	struct socket so, so_next;
621	int nfds;
622	#if defined(RT_OS_WINDOWS)
623	int rc;
624	int error;
625	#else
626	int poll_index = 0;
627	#endif
628	int i;
629
630	STAM_PROFILE_START(&pData->StatFill, a);
631
632	nfds = *pnfds;
633
634	/*
635	* First, TCP sockets
636	*/
637	do_slowtimo = 0;
638	if (!link_up)
639	goto done;
640
641	/*
642	* *_slowtimo needs calling if there are IP fragments
643	* in the fragment queue, or there are TCP connections active
644	*/
645	/* XXX:
646	* triggering of fragment expiration should be the same but use new macroses
647	*/
648	do_slowtimo = (tcb.so_next != &tcb);
649	if (!do_slowtimo)
650	{
651	for (i = 0; i < IPREASS_NHASH; i++)
652	{
653	if (!TAILQ_EMPTY(&ipq[i]))
654	{
655	do_slowtimo = 1;
656	break;
657	}
658	}
659	}
660	/* always add the ICMP socket */
661	#ifndef RT_OS_WINDOWS
662	pData->icmp_socket.so_poll_index = -1;
663	#endif
664	ICMP_ENGAGE_EVENT(&pData->icmp_socket, readfds);
665
666	STAM_COUNTER_RESET(&pData->StatTCP);
667	STAM_COUNTER_RESET(&pData->StatTCPHot);
668
669	QSOCKET_FOREACH(so, so_next, tcp)
670	/* { */
671	Assert(so->so_type == IPPROTO_TCP);
672	#if !defined(RT_OS_WINDOWS)
673	so->so_poll_index = -1;
674	#endif
675	STAM_COUNTER_INC(&pData->StatTCP);
676	#ifdef VBOX_WITH_NAT_UDP_SOCKET_CLONE
677	/* TCP socket can't be cloned */
678	Assert((!so->so_cloneOf));
679	#endif
680	/*
681	* See if we need a tcp_fasttimo
682	*/
683	if ( time_fasttimo == 0
684	&& so->so_tcpcb != NULL
685	&& so->so_tcpcb->t_flags & TF_DELACK)
686	{
687	time_fasttimo = curtime; /* Flag when we want a fasttimo */
688	}
689
690	/*
691	* NOFDREF can include still connecting to local-host,
692	* newly socreated() sockets etc. Don't want to select these.
693	*/
694	if (so->so_state & SS_NOFDREF \|\| so->s == -1)
695	CONTINUE(tcp);
696
697	/*
698	* Set for reading sockets which are accepting
699	*/
700	if (so->so_state & SS_FACCEPTCONN)
701	{
702	STAM_COUNTER_INC(&pData->StatTCPHot);
703	TCP_ENGAGE_EVENT1(so, readfds);
704	CONTINUE(tcp);
705	}
706
707	/*
708	* Set for writing sockets which are connecting
709	*/
710	if (so->so_state & SS_ISFCONNECTING)
711	{
712	Log2(("connecting %R[natsock] engaged\n",so));
713	STAM_COUNTER_INC(&pData->StatTCPHot);
714	#ifdef RT_OS_WINDOWS
715	WIN_TCP_ENGAGE_EVENT2(so, writefds, connectfds);
716	#else
717	TCP_ENGAGE_EVENT1(so, writefds);
718	#endif
719	}
720
721	/*
722	* Set for writing if we are connected, can send more, and
723	* we have something to send
724	*/
725	if (CONN_CANFSEND(so) && SBUF_LEN(&so->so_rcv))
726	{
727	STAM_COUNTER_INC(&pData->StatTCPHot);
728	TCP_ENGAGE_EVENT1(so, writefds);
729	}
730
731	/*
732	* Set for reading (and urgent data) if we are connected, can
733	* receive more, and we have room for it XXX /2 ?
734	*/
735	/* @todo: vvl - check which predicat here will be more useful here in rerm of new sbufs. */
736	if ( CONN_CANFRCV(so)
737	&& (SBUF_LEN(&so->so_snd) < (SBUF_SIZE(&so->so_snd)/2))
738	#ifdef RT_OS_WINDOWS
739	&& !(so->so_state & SS_ISFCONNECTING)
740	#endif
741	)
742	{
743	STAM_COUNTER_INC(&pData->StatTCPHot);
744	TCP_ENGAGE_EVENT2(so, readfds, xfds);
745	}
746	LOOP_LABEL(tcp, so, so_next);
747	}
748
749	/*
750	* UDP sockets
751	*/
752	STAM_COUNTER_RESET(&pData->StatUDP);
753	STAM_COUNTER_RESET(&pData->StatUDPHot);
754
755	QSOCKET_FOREACH(so, so_next, udp)
756	/* { */
757
758	Assert(so->so_type == IPPROTO_UDP);
759	STAM_COUNTER_INC(&pData->StatUDP);
760	#if !defined(RT_OS_WINDOWS)
761	so->so_poll_index = -1;
762	#endif
763
764	/*
765	* See if it's timed out
766	*/
767	if (so->so_expire)
768	{
769	if (so->so_expire <= curtime)
770	{
771	Log2(("NAT: %R[natsock] expired\n", so));
772	if (so->so_timeout != NULL)
773	{
774	/* so_timeout - might change the so_expire value or
775	* drop so_timeout* from so.
776	*/
777	so->so_timeout(pData, so, so->so_timeout_arg);
778	/* on 4.2 so->
779	*/
780	if ( so_next->so_prev != so /* so_timeout freed the socket */
781	\|\| so->so_timeout) /* so_timeout just freed so_timeout */
782	CONTINUE_NO_UNLOCK(udp);
783	}
784	UDP_DETACH(pData, so, so_next);
785	CONTINUE_NO_UNLOCK(udp);
786	}
787	}
788	#ifdef VBOX_WITH_NAT_UDP_SOCKET_CLONE
789	if (so->so_cloneOf)
790	CONTINUE_NO_UNLOCK(udp);
791	#endif
792
793	/*
794	* When UDP packets are received from over the link, they're
795	* sendto()'d straight away, so no need for setting for writing
796	* Limit the number of packets queued by this session to 4.
797	* Note that even though we try and limit this to 4 packets,
798	* the session could have more queued if the packets needed
799	* to be fragmented.
800	*
801	* (XXX <= 4 ?)
802	*/
803	if ((so->so_state & SS_ISFCONNECTED) && so->so_queued <= 4)
804	{
805	STAM_COUNTER_INC(&pData->StatUDPHot);
806	UDP_ENGAGE_EVENT(so, readfds);
807	}
808	LOOP_LABEL(udp, so, so_next);
809	}
810	done:
811
812	#if defined(RT_OS_WINDOWS)
813	*pnfds = VBOX_EVENT_COUNT;
814	#else /* RT_OS_WINDOWS */
815	AssertRelease(poll_index <= *pnfds);
816	*pnfds = poll_index;
817	#endif /* !RT_OS_WINDOWS */
818
819	STAM_PROFILE_STOP(&pData->StatFill, a);
820	}
821
822
823	/**
824	* This function do Connection or sending tcp sequence to.
825	* @returns if true operation completed
826	* @note: functions call tcp_input that potentially could lead to tcp_drop
827	*/
828	static bool slirpConnectOrWrite(PNATState pData, struct socket *so, bool fConnectOnly)
829	{
830	int ret;
831	LogFlowFunc(("ENTER: so:%R[natsock], fConnectOnly:%RTbool\n", so, fConnectOnly));
832	/*
833	* Check for non-blocking, still-connecting sockets
834	*/
835	if (so->so_state & SS_ISFCONNECTING)
836	{
837	Log2(("connecting %R[natsock] catched\n", so));
838	/* Connected */
839	so->so_state &= ~SS_ISFCONNECTING;
840
841	/*
842	* This should be probably guarded by PROBE_CONN too. Anyway,
843	* we disable it on OS/2 because the below send call returns
844	* EFAULT which causes the opened TCP socket to close right
845	* after it has been opened and connected.
846	*/
847	#ifndef RT_OS_OS2
848	ret = send(so->s, (const char *)&ret, 0, 0);
849	if (ret < 0)
850	{
851	/* XXXXX Must fix, zero bytes is a NOP */
852	if ( soIgnorableErrorCode(errno)
853	\|\| errno == ENOTCONN)
854	{
855	LogFlowFunc(("LEAVE: false\n"));
856	return false;
857	}
858
859	/* else failed */
860	so->so_state = SS_NOFDREF;
861	}
862	/* else so->so_state &= ~SS_ISFCONNECTING; */
863	#endif
864
865	/*
866	* Continue tcp_input
867	*/
868	TCP_INPUT(pData, (struct mbuf *)NULL, sizeof(struct ip), so);
869	/* continue; */
870	}
871	else if (!fConnectOnly)
872	{
873	SOWRITE(ret, pData, so);
874	if (RT_LIKELY(ret > 0))
875	{
876	/*
877	* Make sure we will send window update to peer. This is
878	* a moral equivalent of calling tcp_output() for PRU_RCVD
879	* in tcp_usrreq() of the real stack.
880	*/
881	struct tcpcb *tp = sototcpcb(so);
882	if (RT_LIKELY(tp != NULL))
883	tp->t_flags \|= TF_DELACK;
884	}
885	}
886
887	LogFlowFunc(("LEAVE: true\n"));
888	return true;
889	}
890
891	#if defined(RT_OS_WINDOWS)
892	void slirp_select_poll(PNATState pData, int fTimeout)
893	#else /* RT_OS_WINDOWS */
894	void slirp_select_poll(PNATState pData, struct pollfd *polls, int ndfs)
895	#endif /* !RT_OS_WINDOWS */
896	{
897	struct socket so, so_next;
898	int ret;
899	#if defined(RT_OS_WINDOWS)
900	WSANETWORKEVENTS NetworkEvents;
901	int rc;
902	int error;
903	#endif
904
905	STAM_PROFILE_START(&pData->StatPoll, a);
906
907	/* Update time */
908	updtime(pData);
909
910	/*
911	* See if anything has timed out
912	*/
913	if (link_up)
914	{
915	if (time_fasttimo && ((curtime - time_fasttimo) >= 2))
916	{
917	STAM_PROFILE_START(&pData->StatFastTimer, b);
918	tcp_fasttimo(pData);
919	time_fasttimo = 0;
920	STAM_PROFILE_STOP(&pData->StatFastTimer, b);
921	}
922	if (do_slowtimo && ((curtime - last_slowtimo) >= 499))
923	{
924	STAM_PROFILE_START(&pData->StatSlowTimer, c);
925	ip_slowtimo(pData);
926	tcp_slowtimo(pData);
927	last_slowtimo = curtime;
928	STAM_PROFILE_STOP(&pData->StatSlowTimer, c);
929	}
930	}
931	#if defined(RT_OS_WINDOWS)
932	if (fTimeout)
933	return; /* only timer update */
934	#endif
935
936	/*
937	* Check sockets
938	*/
939	if (!link_up)
940	goto done;
941	#if defined(RT_OS_WINDOWS)
942	icmpwin_process(pData);
943	#else
944	if ( (pData->icmp_socket.s != -1)
945	&& CHECK_FD_SET(&pData->icmp_socket, ignored, readfds))
946	sorecvfrom(pData, &pData->icmp_socket);
947	#endif
948	/*
949	* Check TCP sockets
950	*/
951	QSOCKET_FOREACH(so, so_next, tcp)
952	/* { */
953	/* TCP socket can't be cloned */
954	#ifdef VBOX_WITH_NAT_UDP_SOCKET_CLONE
955	Assert((!so->so_cloneOf));
956	#endif
957	Assert(!so->fUnderPolling);
958	so->fUnderPolling = 1;
959	if (slirpVerifyAndFreeSocket(pData, so))
960	CONTINUE(tcp);
961	/*
962	* FD_ISSET is meaningless on these sockets
963	* (and they can crash the program)
964	*/
965	if (so->so_state & SS_NOFDREF \|\| so->s == -1)
966	{
967	so->fUnderPolling = 0;
968	CONTINUE(tcp);
969	}
970
971	POLL_TCP_EVENTS(rc, error, so, &NetworkEvents);
972
973	LOG_NAT_SOCK(so, TCP, &NetworkEvents, readfds, writefds, xfds);
974
975	if (so->so_state & SS_ISFCONNECTING)
976	{
977	int sockerr = 0;
978	#if !defined(RT_OS_WINDOWS)
979	{
980	int revents = 0;
981
982	/*
983	* Failed connect(2) is reported by poll(2) on
984	* different OSes with different combinations of
985	* POLLERR, POLLHUP, and POLLOUT.
986	*/
987	if ( CHECK_FD_SET(so, NetworkEvents, closefds) /* POLLHUP */
988	\|\| CHECK_FD_SET(so, NetworkEvents, rderr)) /* POLLERR */
989	{
990	revents = POLLHUP; /* squash to single "failed" flag */
991	}
992	#if defined(RT_OS_SOLARIS) \|\| defined(RT_OS_NETBSD)
993	/* Solaris and NetBSD report plain POLLOUT even on error */
994	else if (CHECK_FD_SET(so, NetworkEvents, writefds)) /* POLLOUT */
995	{
996	revents = POLLOUT;
997	}
998	#endif
999
1000	if (revents != 0)
1001	{
1002	socklen_t optlen = (socklen_t)sizeof(sockerr);
1003	ret = getsockopt(so->s, SOL_SOCKET, SO_ERROR, &sockerr, &optlen);
1004
1005	if ( RT_UNLIKELY(ret < 0)
1006	\|\| ( (revents & POLLHUP)
1007	&& RT_UNLIKELY(sockerr == 0)))
1008	sockerr = ETIMEDOUT;
1009	}
1010	}
1011	#else /* RT_OS_WINDOWS */
1012	{
1013	if (NetworkEvents.lNetworkEvents & FD_CONNECT)
1014	sockerr = NetworkEvents.iErrorCode[FD_CONNECT_BIT];
1015	}
1016	#endif
1017	if (sockerr != 0)
1018	{
1019	tcp_fconnect_failed(pData, so, sockerr);
1020	ret = slirpVerifyAndFreeSocket(pData, so);
1021	Assert(ret == 1); /* freed */
1022	CONTINUE(tcp);
1023	}
1024
1025	/*
1026	* XXX: For now just fall through to the old code to
1027	* handle successful connect(2).
1028	*/
1029	}
1030
1031	/*
1032	* Check for URG data
1033	* This will soread as well, so no need to
1034	* test for readfds below if this succeeds
1035	*/
1036
1037	/* out-of-band data */
1038	if ( CHECK_FD_SET(so, NetworkEvents, xfds)
1039	#ifdef RT_OS_DARWIN
1040	/* Darwin and probably BSD hosts generates POLLPRI\|POLLHUP event on receiving TCP.flags.{ACK\|URG\|FIN} this
1041	* combination on other Unixs hosts doesn't enter to this branch
1042	*/
1043	&& !CHECK_FD_SET(so, NetworkEvents, closefds)
1044	#endif
1045	#ifdef RT_OS_WINDOWS
1046	/**
1047	* In some cases FD_CLOSE comes with FD_OOB, that confuse tcp processing.
1048	*/
1049	&& !WIN_CHECK_FD_SET(so, NetworkEvents, closefds)
1050	#endif
1051	)
1052	{
1053	sorecvoob(pData, so);
1054	if (slirpVerifyAndFreeSocket(pData, so))
1055	CONTINUE(tcp);
1056	}
1057
1058	/*
1059	* Check sockets for reading
1060	*/
1061	else if ( CHECK_FD_SET(so, NetworkEvents, readfds)
1062	\|\| WIN_CHECK_FD_SET(so, NetworkEvents, acceptds))
1063	{
1064
1065	#ifdef RT_OS_WINDOWS
1066	if (WIN_CHECK_FD_SET(so, NetworkEvents, connectfds))
1067	{
1068	/* Finish connection first */
1069	/* should we ignore return value? */
1070	bool fRet = slirpConnectOrWrite(pData, so, true);
1071	LogFunc(("fRet:%RTbool\n", fRet));
1072	if (slirpVerifyAndFreeSocket(pData, so))
1073	CONTINUE(tcp);
1074	}
1075	#endif
1076	/*
1077	* Check for incoming connections
1078	*/
1079	if (so->so_state & SS_FACCEPTCONN)
1080	{
1081	TCP_CONNECT(pData, so);
1082	if (slirpVerifyAndFreeSocket(pData, so))
1083	CONTINUE(tcp);
1084	if (!CHECK_FD_SET(so, NetworkEvents, closefds))
1085	{
1086	so->fUnderPolling = 0;
1087	CONTINUE(tcp);
1088	}
1089	}
1090
1091	ret = soread(pData, so);
1092	if (slirpVerifyAndFreeSocket(pData, so))
1093	CONTINUE(tcp);
1094	/* Output it if we read something */
1095	if (RT_LIKELY(ret > 0))
1096	TCP_OUTPUT(pData, sototcpcb(so));
1097
1098	if (slirpVerifyAndFreeSocket(pData, so))
1099	CONTINUE(tcp);
1100	}
1101
1102	/*
1103	* Check for FD_CLOSE events.
1104	* in some cases once FD_CLOSE engaged on socket it could be flashed latter (for some reasons)
1105	*/
1106	if ( CHECK_FD_SET(so, NetworkEvents, closefds)
1107	\|\| (so->so_close == 1))
1108	{
1109	/*
1110	* drain the socket
1111	*/
1112	for (; so_next->so_prev == so
1113	&& !slirpVerifyAndFreeSocket(pData, so);)
1114	{
1115	ret = soread(pData, so);
1116	if (slirpVerifyAndFreeSocket(pData, so))
1117	break;
1118
1119	if (ret > 0)
1120	TCP_OUTPUT(pData, sototcpcb(so));
1121	else if (so_next->so_prev == so)
1122	{
1123	Log2(("%R[natsock] errno %d (%s)\n", so, errno, strerror(errno)));
1124	break;
1125	}
1126	}
1127
1128	/* if socket freed ''so'' is PHANTOM and next socket isn't points on it */
1129	if (so_next->so_prev == so)
1130	{
1131	/* mark the socket for termination _after_ it was drained */
1132	so->so_close = 1;
1133	/* No idea about Windows but on Posix, POLLHUP means that we can't send more.
1134	* Actually in the specific error scenario, POLLERR is set as well. */
1135	#ifndef RT_OS_WINDOWS
1136	if (CHECK_FD_SET(so, NetworkEvents, rderr))
1137	sofcantsendmore(so);
1138	#endif
1139	}
1140	if (so_next->so_prev == so)
1141	so->fUnderPolling = 0;
1142	CONTINUE(tcp);
1143	}
1144
1145	/*
1146	* Check sockets for writing
1147	*/
1148	if ( CHECK_FD_SET(so, NetworkEvents, writefds)
1149	#ifdef RT_OS_WINDOWS
1150	\|\| WIN_CHECK_FD_SET(so, NetworkEvents, connectfds)
1151	#endif
1152	)
1153	{
1154	int fConnectOrWriteSuccess = slirpConnectOrWrite(pData, so, false);
1155	/* slirpConnectOrWrite could return true even if tcp_input called tcp_drop,
1156	* so we should be ready to such situations.
1157	*/
1158	if (slirpVerifyAndFreeSocket(pData, so))
1159	CONTINUE(tcp);
1160	else if (!fConnectOrWriteSuccess)
1161	{
1162	so->fUnderPolling = 0;
1163	CONTINUE(tcp);
1164	}
1165	/* slirpConnectionOrWrite succeeded and socket wasn't dropped */
1166	}
1167
1168	/*
1169	* Probe a still-connecting, non-blocking socket
1170	* to check if it's still alive
1171	*/
1172	#ifdef PROBE_CONN
1173	if (so->so_state & SS_ISFCONNECTING)
1174	{
1175	ret = recv(so->s, (char *)&ret, 0, 0);
1176
1177	if (ret < 0)
1178	{
1179	/* XXX */
1180	if ( soIgnorableErrorCode(errno)
1181	\|\| errno == ENOTCONN)
1182	{
1183	CONTINUE(tcp); /* Still connecting, continue */
1184	}
1185
1186	/* else failed */
1187	so->so_state = SS_NOFDREF;
1188
1189	/* tcp_input will take care of it */
1190	}
1191	else
1192	{
1193	ret = send(so->s, &ret, 0, 0);
1194	if (ret < 0)
1195	{
1196	/* XXX */
1197	if ( soIgnorableErrorCode(errno)
1198	\|\| errno == ENOTCONN)
1199	{
1200	CONTINUE(tcp);
1201	}
1202	/* else failed */
1203	so->so_state = SS_NOFDREF;
1204	}
1205	else
1206	so->so_state &= ~SS_ISFCONNECTING;
1207
1208	}
1209	TCP_INPUT((struct mbuf *)NULL, sizeof(struct ip),so);
1210	} /* SS_ISFCONNECTING */
1211	#endif
1212	if (!slirpVerifyAndFreeSocket(pData, so))
1213	so->fUnderPolling = 0;
1214	LOOP_LABEL(tcp, so, so_next);
1215	}
1216
1217	/*
1218	* Now UDP sockets.
1219	* Incoming packets are sent straight away, they're not buffered.
1220	* Incoming UDP data isn't buffered either.
1221	*/
1222	QSOCKET_FOREACH(so, so_next, udp)
1223	/* { */
1224	#ifdef VBOX_WITH_NAT_UDP_SOCKET_CLONE
1225	if (so->so_cloneOf)
1226	CONTINUE_NO_UNLOCK(udp);
1227	#endif
1228	#if 0
1229	so->fUnderPolling = 1;
1230	if(slirpVerifyAndFreeSocket(pData, so));
1231	CONTINUE(udp);
1232	so->fUnderPolling = 0;
1233	#endif
1234
1235	POLL_UDP_EVENTS(rc, error, so, &NetworkEvents);
1236
1237	LOG_NAT_SOCK(so, UDP, &NetworkEvents, readfds, writefds, xfds);
1238
1239	if (so->s != -1 && CHECK_FD_SET(so, NetworkEvents, readfds))
1240	{
1241	SORECVFROM(pData, so);
1242	}
1243	LOOP_LABEL(udp, so, so_next);
1244	}
1245
1246	done:
1247
1248	STAM_PROFILE_STOP(&pData->StatPoll, a);
1249	}
1250
1251
1252	struct arphdr
1253	{
1254	unsigned short ar_hrd; /* format of hardware address */
1255	unsigned short ar_pro; /* format of protocol address */
1256	unsigned char ar_hln; /* length of hardware address */
1257	unsigned char ar_pln; /* length of protocol address */
1258	unsigned short ar_op; /* ARP opcode (command) */
1259
1260	/*
1261	* Ethernet looks like this : This bit is variable sized however...
1262	*/
1263	unsigned char ar_sha[ETH_ALEN]; /* sender hardware address */
1264	unsigned char ar_sip[4]; /* sender IP address */
1265	unsigned char ar_tha[ETH_ALEN]; /* target hardware address */
1266	unsigned char ar_tip[4]; /* target IP address */
1267	};
1268	AssertCompileSize(struct arphdr, 28);
1269
1270	static void arp_output(PNATState pData, const uint8_t pcu8EtherSource, const struct arphdr pcARPHeaderSource, uint32_t ip4TargetAddress)
1271	{
1272	struct ethhdr *pEtherHeaderResponse;
1273	struct arphdr *pARPHeaderResponse;
1274	uint32_t ip4TargetAddressInHostFormat;
1275	struct mbuf *pMbufResponse;
1276
1277	Assert((pcu8EtherSource));
1278	if (!pcu8EtherSource)
1279	return;
1280	ip4TargetAddressInHostFormat = RT_N2H_U32(ip4TargetAddress);
1281
1282	pMbufResponse = m_getcl(pData, M_NOWAIT, MT_HEADER, M_PKTHDR);
1283	if (!pMbufResponse)
1284	return;
1285	pEtherHeaderResponse = mtod(pMbufResponse, struct ethhdr *);
1286	/* @note: if_encap will swap src and dst*/
1287	memcpy(pEtherHeaderResponse->h_source, pcu8EtherSource, ETH_ALEN);
1288	pMbufResponse->m_data += ETH_HLEN;
1289	pARPHeaderResponse = mtod(pMbufResponse, struct arphdr *);
1290	pMbufResponse->m_len = sizeof(struct arphdr);
1291
1292	pARPHeaderResponse->ar_hrd = RT_H2N_U16_C(1);
1293	pARPHeaderResponse->ar_pro = RT_H2N_U16_C(ETH_P_IP);
1294	pARPHeaderResponse->ar_hln = ETH_ALEN;
1295	pARPHeaderResponse->ar_pln = 4;
1296	pARPHeaderResponse->ar_op = RT_H2N_U16_C(ARPOP_REPLY);
1297	memcpy(pARPHeaderResponse->ar_sha, special_ethaddr, ETH_ALEN);
1298
1299	if (!slirpMbufTagService(pData, pMbufResponse, (uint8_t)(ip4TargetAddressInHostFormat & ~pData->netmask)))
1300	{
1301	static bool fTagErrorReported;
1302	if (!fTagErrorReported)
1303	{
1304	LogRel(("NAT: couldn't add the tag(PACKET_SERVICE:%d)\n",
1305	(uint8_t)(ip4TargetAddressInHostFormat & ~pData->netmask)));
1306	fTagErrorReported = true;
1307	}
1308	}
1309	pARPHeaderResponse->ar_sha[5] = (uint8_t)(ip4TargetAddressInHostFormat & ~pData->netmask);
1310
1311	memcpy(pARPHeaderResponse->ar_sip, pcARPHeaderSource->ar_tip, 4);
1312	memcpy(pARPHeaderResponse->ar_tha, pcARPHeaderSource->ar_sha, ETH_ALEN);
1313	memcpy(pARPHeaderResponse->ar_tip, pcARPHeaderSource->ar_sip, 4);
1314	if_encap(pData, ETH_P_ARP, pMbufResponse, ETH_ENCAP_URG);
1315	}
1316	/**
1317	* @note This function will free m!
1318	*/
1319	static void arp_input(PNATState pData, struct mbuf *m)
1320	{
1321	struct ethhdr *pEtherHeader;
1322	struct arphdr *pARPHeader;
1323	uint32_t ip4TargetAddress;
1324
1325	int ar_op;
1326	pEtherHeader = mtod(m, struct ethhdr *);
1327	pARPHeader = (struct arphdr *)&pEtherHeader[1];
1328
1329	ar_op = RT_N2H_U16(pARPHeader->ar_op);
1330	ip4TargetAddress = (uint32_t)pARPHeader->ar_tip;
1331
1332	switch (ar_op)
1333	{
1334	case ARPOP_REQUEST:
1335	if ( CTL_CHECK(ip4TargetAddress, CTL_DNS)
1336	\|\| CTL_CHECK(ip4TargetAddress, CTL_ALIAS)
1337	\|\| CTL_CHECK(ip4TargetAddress, CTL_TFTP))
1338	arp_output(pData, pEtherHeader->h_source, pARPHeader, ip4TargetAddress);
1339
1340	/* Gratuitous ARP */
1341	if ( (uint32_t )pARPHeader->ar_sip == (uint32_t )pARPHeader->ar_tip
1342	&& memcmp(pARPHeader->ar_tha, broadcast_ethaddr, ETH_ALEN) == 0
1343	&& memcmp(pEtherHeader->h_dest, broadcast_ethaddr, ETH_ALEN) == 0)
1344	{
1345	/* We've received an announce about address assignment,
1346	* let's do an ARP cache update
1347	*/
1348	static bool fGratuitousArpReported;
1349	if (!fGratuitousArpReported)
1350	{
1351	LogRel(("NAT: Gratuitous ARP [IP:%RTnaipv4, ether:%RTmac]\n",
1352	pARPHeader->ar_sip, pARPHeader->ar_sha));
1353	fGratuitousArpReported = true;
1354	}
1355	slirp_arp_cache_update_or_add(pData, (uint32_t )pARPHeader->ar_sip, &pARPHeader->ar_sha[0]);
1356	}
1357	break;
1358
1359	case ARPOP_REPLY:
1360	slirp_arp_cache_update_or_add(pData, (uint32_t )pARPHeader->ar_sip, &pARPHeader->ar_sha[0]);
1361	break;
1362
1363	default:
1364	break;
1365	}
1366
1367	m_freem(pData, m);
1368	}
1369
1370	/**
1371	* Feed a packet into the slirp engine.
1372	*
1373	* @param m Data buffer, m_len is not valid.
1374	* @param cbBuf The length of the data in m.
1375	*/
1376	void slirp_input(PNATState pData, struct mbuf *m, size_t cbBuf)
1377	{
1378	int proto;
1379	static bool fWarnedIpv6;
1380	struct ethhdr *eh;
1381	uint8_t au8Ether[ETH_ALEN];
1382
1383	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	eh = mtod(m, struct ethhdr *);
1391	proto = RT_N2H_U16(eh->h_proto);
1392
1393	memcpy(au8Ether, eh->h_source, ETH_ALEN);
1394
1395	switch(proto)
1396	{
1397	case ETH_P_ARP:
1398	arp_input(pData, m);
1399	break;
1400
1401	case ETH_P_IP:
1402	/* Update time. Important if the network is very quiet, as otherwise
1403	* the first outgoing connection gets an incorrect timestamp. */
1404	updtime(pData);
1405	m_adj(m, ETH_HLEN);
1406	M_ASSERTPKTHDR(m);
1407	m->m_pkthdr.header = mtod(m, void *);
1408	ip_input(pData, m);
1409	break;
1410
1411	case ETH_P_IPV6:
1412	m_freem(pData, m);
1413	if (!fWarnedIpv6)
1414	{
1415	LogRel(("NAT: IPv6 not supported\n"));
1416	fWarnedIpv6 = true;
1417	}
1418	break;
1419
1420	default:
1421	Log(("NAT: Unsupported protocol %x\n", proto));
1422	m_freem(pData, m);
1423	break;
1424	}
1425
1426	if (pData->cRedirectionsActive != pData->cRedirectionsStored)
1427	activate_port_forwarding(pData, au8Ether);
1428	}
1429
1430	/**
1431	* Output the IP packet to the ethernet device.
1432	*
1433	* @note This function will free m!
1434	*/
1435	void if_encap(PNATState pData, uint16_t eth_proto, struct mbuf *m, int flags)
1436	{
1437	struct ethhdr *eh;
1438	uint8_t *mbuf = NULL;
1439	size_t mlen = 0;
1440	STAM_PROFILE_START(&pData->StatIF_encap, a);
1441	LogFlowFunc(("ENTER: pData:%p, eth_proto:%RX16, m:%p, flags:%d\n",
1442	pData, eth_proto, m, flags));
1443
1444	M_ASSERTPKTHDR(m);
1445
1446	Assert(M_LEADINGSPACE(m) >= ETH_HLEN);
1447	m->m_data -= ETH_HLEN;
1448	m->m_len += ETH_HLEN;
1449	eh = mtod(m, struct ethhdr *);
1450	mlen = m->m_len;
1451
1452	if (memcmp(eh->h_source, special_ethaddr, ETH_ALEN) != 0)
1453	{
1454	struct m_tag *t = m_tag_first(m);
1455	uint8_t u8ServiceId = CTL_ALIAS;
1456	memcpy(eh->h_dest, eh->h_source, ETH_ALEN);
1457	memcpy(eh->h_source, special_ethaddr, ETH_ALEN);
1458	Assert(memcmp(eh->h_dest, special_ethaddr, ETH_ALEN) != 0);
1459	if (memcmp(eh->h_dest, zerro_ethaddr, ETH_ALEN) == 0)
1460	{
1461	/* don't do anything */
1462	m_freem(pData, m);
1463	goto done;
1464	}
1465	if ( t
1466	&& (t = m_tag_find(m, PACKET_SERVICE, NULL)))
1467	{
1468	Assert(t);
1469	u8ServiceId = (uint8_t )&t[1];
1470	}
1471	eh->h_source[5] = u8ServiceId;
1472	}
1473	/*
1474	* we're processing the chain, that isn't not expected.
1475	*/
1476	Assert((!m->m_next));
1477	if (m->m_next)
1478	{
1479	Log(("NAT: if_encap's recived the chain, dropping...\n"));
1480	m_freem(pData, m);
1481	goto done;
1482	}
1483	mbuf = mtod(m, uint8_t *);
1484	eh->h_proto = RT_H2N_U16(eth_proto);
1485	LogFunc(("eh(dst:%RTmac, src:%RTmac)\n", eh->h_dest, eh->h_source));
1486	if (flags & ETH_ENCAP_URG)
1487	slirp_urg_output(pData->pvUser, m, mbuf, mlen);
1488	else
1489	slirp_output(pData->pvUser, m, mbuf, mlen);
1490	done:
1491	STAM_PROFILE_STOP(&pData->StatIF_encap, a);
1492	LogFlowFuncLeave();
1493	}
1494
1495	/**
1496	* Still we're using dhcp server leasing to map ether to IP
1497	* @todo see rt_lookup_in_cache
1498	*/
1499	static uint32_t find_guest_ip(PNATState pData, const uint8_t *eth_addr)
1500	{
1501	uint32_t ip = INADDR_ANY;
1502	int rc;
1503
1504	if (eth_addr == NULL)
1505	return INADDR_ANY;
1506
1507	if ( memcmp(eth_addr, zerro_ethaddr, ETH_ALEN) == 0
1508	\|\| memcmp(eth_addr, broadcast_ethaddr, ETH_ALEN) == 0)
1509	return INADDR_ANY;
1510
1511	rc = slirp_arp_lookup_ip_by_ether(pData, eth_addr, &ip);
1512	if (RT_SUCCESS(rc))
1513	return ip;
1514
1515	bootp_cache_lookup_ip_by_ether(pData, eth_addr, &ip);
1516	/* ignore return code, ip will be set to INADDR_ANY on error */
1517	return ip;
1518	}
1519
1520	/**
1521	* We need check if we've activated port forwarding
1522	* for specific machine ... that of course relates to
1523	* service mode
1524	* @todo finish this for service case
1525	*/
1526	static void activate_port_forwarding(PNATState pData, const uint8_t *h_source)
1527	{
1528	struct port_forward_rule rule, tmp;
1529	const uint8_t *pu8EthSource = h_source;
1530
1531	/* check mac here */
1532	LIST_FOREACH_SAFE(rule, &pData->port_forward_rule_head, list, tmp)
1533	{
1534	struct socket *so;
1535	struct sockaddr sa;
1536	struct sockaddr_in *psin;
1537	socklen_t socketlen;
1538	int rc;
1539	uint32_t guest_addr; /* need to understand if we already give address to guest */
1540
1541	if (rule->activated)
1542	continue;
1543
1544	guest_addr = find_guest_ip(pData, pu8EthSource);
1545	if (guest_addr == INADDR_ANY)
1546	{
1547	/* the address wasn't granted */
1548	return;
1549	}
1550
1551	if ( rule->guest_addr.s_addr != guest_addr
1552	&& rule->guest_addr.s_addr != INADDR_ANY)
1553	continue;
1554	if (rule->guest_addr.s_addr == INADDR_ANY)
1555	rule->guest_addr.s_addr = guest_addr;
1556
1557	LogRel(("NAT: set redirect %s host %RTnaipv4:%d => guest %RTnaipv4:%d\n",
1558	rule->proto == IPPROTO_UDP ? "UDP" : "TCP",
1559	rule->bind_ip.s_addr, rule->host_port,
1560	guest_addr, rule->guest_port));
1561
1562	if (rule->proto == IPPROTO_UDP)
1563	so = udp_listen(pData, rule->bind_ip.s_addr, RT_H2N_U16(rule->host_port), guest_addr,
1564	RT_H2N_U16(rule->guest_port), 0);
1565	else
1566	so = solisten(pData, rule->bind_ip.s_addr, RT_H2N_U16(rule->host_port), guest_addr,
1567	RT_H2N_U16(rule->guest_port), 0);
1568
1569	if (so == NULL)
1570	goto remove_port_forwarding;
1571
1572	psin = (struct sockaddr_in *)&sa;
1573	psin->sin_family = AF_INET;
1574	psin->sin_port = 0;
1575	psin->sin_addr.s_addr = INADDR_ANY;
1576	socketlen = sizeof(struct sockaddr);
1577
1578	rc = getsockname(so->s, &sa, &socketlen);
1579	if (rc < 0 \|\| sa.sa_family != AF_INET)
1580	goto remove_port_forwarding;
1581
1582	rule->activated = 1;
1583	rule->so = so;
1584	pData->cRedirectionsActive++;
1585	continue;
1586
1587	remove_port_forwarding:
1588	LogRel(("NAT: failed to redirect %s %RTnaipv4:%d => %RTnaipv4:%d\n",
1589	(rule->proto == IPPROTO_UDP ? "UDP" : "TCP"),
1590	rule->bind_ip.s_addr, rule->host_port,
1591	guest_addr, rule->guest_port));
1592	LIST_REMOVE(rule, list);
1593	pData->cRedirectionsStored--;
1594	RTMemFree(rule);
1595	}
1596	}
1597
1598	/**
1599	* Changes in 3.1 instead of opening new socket do the following:
1600	* gain more information:
1601	* 1. bind IP
1602	* 2. host port
1603	* 3. guest port
1604	* 4. proto
1605	* 5. guest MAC address
1606	* the guest's MAC address is rather important for service, but we easily
1607	* could get it from VM configuration in DrvNAT or Service, the idea is activating
1608	* corresponding port-forwarding
1609	*/
1610	int slirp_add_redirect(PNATState pData, int is_udp, struct in_addr host_addr, int host_port,
1611	struct in_addr guest_addr, int guest_port, const uint8_t *ethaddr)
1612	{
1613	struct port_forward_rule *rule = NULL;
1614	LIST_FOREACH(rule, &pData->port_forward_rule_head, list)
1615	{
1616	if ( rule->proto == (is_udp ? IPPROTO_UDP : IPPROTO_TCP)
1617	&& rule->host_port == host_port
1618	&& rule->bind_ip.s_addr == host_addr.s_addr
1619	&& rule->guest_port == guest_port
1620	&& rule->guest_addr.s_addr == guest_addr.s_addr
1621	)
1622	return 0; /* rule has been already registered */
1623	}
1624
1625	rule = RTMemAllocZ(sizeof(struct port_forward_rule));
1626	if (rule == NULL)
1627	return 1;
1628
1629	rule->proto = (is_udp ? IPPROTO_UDP : IPPROTO_TCP);
1630	rule->host_port = host_port;
1631	rule->guest_port = guest_port;
1632	rule->guest_addr.s_addr = guest_addr.s_addr;
1633	rule->bind_ip.s_addr = host_addr.s_addr;
1634	if (ethaddr != NULL)
1635	memcpy(rule->mac_address, ethaddr, ETH_ALEN);
1636	/* @todo add mac address */
1637	LIST_INSERT_HEAD(&pData->port_forward_rule_head, rule, list);
1638	pData->cRedirectionsStored++;
1639	/* activate port-forwarding if guest has already got assigned IP */
1640	if ( ethaddr
1641	&& memcmp(ethaddr, zerro_ethaddr, ETH_ALEN))
1642	activate_port_forwarding(pData, ethaddr);
1643	return 0;
1644	}
1645
1646	int slirp_remove_redirect(PNATState pData, int is_udp, struct in_addr host_addr, int host_port,
1647	struct in_addr guest_addr, int guest_port)
1648	{
1649	struct port_forward_rule *rule = NULL;
1650	LIST_FOREACH(rule, &pData->port_forward_rule_head, list)
1651	{
1652	if ( rule->proto == (is_udp ? IPPROTO_UDP : IPPROTO_TCP)
1653	&& rule->host_port == host_port
1654	&& rule->guest_port == guest_port
1655	&& rule->bind_ip.s_addr == host_addr.s_addr
1656	&& rule->guest_addr.s_addr == guest_addr.s_addr
1657	&& rule->activated)
1658	{
1659	LogRel(("NAT: remove redirect %s host %RTnaipv4:%d => guest %RTnaipv4:%d\n",
1660	rule->proto == IPPROTO_UDP ? "UDP" : "TCP",
1661	rule->bind_ip.s_addr, rule->host_port,
1662	guest_addr.s_addr, rule->guest_port));
1663
1664	if (is_udp)
1665	udp_detach(pData, rule->so);
1666	else
1667	tcp_close(pData, sototcpcb(rule->so));
1668	LIST_REMOVE(rule, list);
1669	RTMemFree(rule);
1670	pData->cRedirectionsStored--;
1671	break;
1672	}
1673
1674	}
1675	return 0;
1676	}
1677
1678	void slirp_set_ethaddr_and_activate_port_forwarding(PNATState pData, const uint8_t *ethaddr, uint32_t GuestIP)
1679	{
1680	memcpy(client_ethaddr, ethaddr, ETH_ALEN);
1681	if (GuestIP != INADDR_ANY)
1682	{
1683	slirp_arp_cache_update_or_add(pData, GuestIP, ethaddr);
1684	activate_port_forwarding(pData, ethaddr);
1685	}
1686	}
1687
1688	#if defined(RT_OS_WINDOWS)
1689	HANDLE *slirp_get_events(PNATState pData)
1690	{
1691	return pData->phEvents;
1692	}
1693	void slirp_register_external_event(PNATState pData, HANDLE hEvent, int index)
1694	{
1695	pData->phEvents[index] = hEvent;
1696	}
1697	#endif
1698
1699	unsigned int slirp_get_timeout_ms(PNATState pData)
1700	{
1701	if (link_up)
1702	{
1703	if (time_fasttimo)
1704	return 2;
1705	if (do_slowtimo)
1706	return 500; /* see PR_SLOWHZ */
1707	}
1708	return 36001000; / one hour */
1709	}
1710
1711	#ifndef RT_OS_WINDOWS
1712	int slirp_get_nsock(PNATState pData)
1713	{
1714	return pData->nsock;
1715	}
1716	#endif
1717
1718	/*
1719	* this function called from NAT thread
1720	*/
1721	void slirp_post_sent(PNATState pData, void *pvArg)
1722	{
1723	struct mbuf m = (struct mbuf )pvArg;
1724	m_freem(pData, m);
1725	}
1726
1727	void slirp_set_dhcp_TFTP_prefix(PNATState pData, const char *tftpPrefix)
1728	{
1729	Log2(("tftp_prefix: %s\n", tftpPrefix));
1730	tftp_prefix = tftpPrefix;
1731	}
1732
1733	void slirp_set_dhcp_TFTP_bootfile(PNATState pData, const char *bootFile)
1734	{
1735	Log2(("bootFile: %s\n", bootFile));
1736	bootp_filename = bootFile;
1737	}
1738
1739	void slirp_set_dhcp_next_server(PNATState pData, const char *next_server)
1740	{
1741	Log2(("next_server: %s\n", next_server));
1742	if (next_server == NULL)
1743	pData->tftp_server.s_addr = RT_H2N_U32(RT_N2H_U32(pData->special_addr.s_addr) \| CTL_TFTP);
1744	else
1745	inet_aton(next_server, &pData->tftp_server);
1746	}
1747
1748	int slirp_set_binding_address(PNATState pData, char *addr)
1749	{
1750	if (addr == NULL \|\| (inet_aton(addr, &pData->bindIP) == 0))
1751	{
1752	pData->bindIP.s_addr = INADDR_ANY;
1753	return 1;
1754	}
1755	return 0;
1756	}
1757
1758	void slirp_set_dhcp_dns_proxy(PNATState pData, bool fDNSProxy)
1759	{
1760	if (!pData->fUseHostResolver)
1761	{
1762	Log2(("NAT: DNS proxy switched %s\n", (fDNSProxy ? "on" : "off")));
1763	pData->fUseDnsProxy = fDNSProxy;
1764	}
1765	else if (fDNSProxy)
1766	LogRel(("NAT: Host Resolver conflicts with DNS proxy, the last one was forcely ignored\n"));
1767	}
1768
1769	#define CHECK_ARG(name, val, lim_min, lim_max) \
1770	do { \
1771	if ((val) < (lim_min) \|\| (val) > (lim_max)) \
1772	{ \
1773	LogRel(("NAT: (" #name ":%d) has been ignored, " \
1774	"because out of range (%d, %d)\n", (val), (lim_min), (lim_max))); \
1775	return; \
1776	} \
1777	else \
1778	LogRel(("NAT: (" #name ":%d)\n", (val))); \
1779	} while (0)
1780
1781	void slirp_set_somaxconn(PNATState pData, int iSoMaxConn)
1782	{
1783	LogFlowFunc(("iSoMaxConn:d\n", iSoMaxConn));
1784	/* Conditions */
1785	if (iSoMaxConn > SOMAXCONN)
1786	{
1787	LogRel(("NAT: value of somaxconn(%d) bigger than SOMAXCONN(%d)\n", iSoMaxConn, SOMAXCONN));
1788	iSoMaxConn = SOMAXCONN;
1789	}
1790
1791	if (iSoMaxConn < 1)
1792	{
1793	LogRel(("NAT: proposed value(%d) of somaxconn is invalid, default value is used (%d)\n", iSoMaxConn, pData->soMaxConn));
1794	LogFlowFuncLeave();
1795	return;
1796	}
1797
1798	/* Asignment */
1799	if (pData->soMaxConn != iSoMaxConn)
1800	{
1801	LogRel(("NAT: value of somaxconn has been changed from %d to %d\n",
1802	pData->soMaxConn, iSoMaxConn));
1803	pData->soMaxConn = iSoMaxConn;
1804	}
1805	LogFlowFuncLeave();
1806	}
1807	/* don't allow user set less 8kB and more than 1M values */
1808	#define _8K_1M_CHECK_ARG(name, val) CHECK_ARG(name, (val), 8, 1024)
1809	void slirp_set_rcvbuf(PNATState pData, int kilobytes)
1810	{
1811	_8K_1M_CHECK_ARG("SOCKET_RCVBUF", kilobytes);
1812	pData->socket_rcv = kilobytes;
1813	}
1814	void slirp_set_sndbuf(PNATState pData, int kilobytes)
1815	{
1816	_8K_1M_CHECK_ARG("SOCKET_SNDBUF", kilobytes);
1817	pData->socket_snd = kilobytes * _1K;
1818	}
1819	void slirp_set_tcp_rcvspace(PNATState pData, int kilobytes)
1820	{
1821	_8K_1M_CHECK_ARG("TCP_RCVSPACE", kilobytes);
1822	tcp_rcvspace = kilobytes * _1K;
1823	}
1824	void slirp_set_tcp_sndspace(PNATState pData, int kilobytes)
1825	{
1826	_8K_1M_CHECK_ARG("TCP_SNDSPACE", kilobytes);
1827	tcp_sndspace = kilobytes * _1K;
1828	}
1829
1830	/*
1831	* Looking for Ether by ip in ARP-cache
1832	* Note: it´s responsible of caller to allocate buffer for result
1833	* @returns iprt status code
1834	*/
1835	int slirp_arp_lookup_ether_by_ip(PNATState pData, uint32_t ip, uint8_t *ether)
1836	{
1837	struct arp_cache_entry *ac;
1838
1839	if (ether == NULL)
1840	return VERR_INVALID_PARAMETER;
1841
1842	if (LIST_EMPTY(&pData->arp_cache))
1843	return VERR_NOT_FOUND;
1844
1845	LIST_FOREACH(ac, &pData->arp_cache, list)
1846	{
1847	if ( ac->ip == ip
1848	&& memcmp(ac->ether, broadcast_ethaddr, ETH_ALEN) != 0)
1849	{
1850	memcpy(ether, ac->ether, ETH_ALEN);
1851	return VINF_SUCCESS;
1852	}
1853	}
1854	return VERR_NOT_FOUND;
1855	}
1856
1857	/*
1858	* Looking for IP by Ether in ARP-cache
1859	* Note: it´s responsible of caller to allocate buffer for result
1860	* @returns 0 - if found, 1 - otherwise
1861	*/
1862	int slirp_arp_lookup_ip_by_ether(PNATState pData, const uint8_t ether, uint32_t ip)
1863	{
1864	struct arp_cache_entry *ac;
1865	*ip = INADDR_ANY;
1866
1867	if (LIST_EMPTY(&pData->arp_cache))
1868	return VERR_NOT_FOUND;
1869
1870	LIST_FOREACH(ac, &pData->arp_cache, list)
1871	{
1872	if (memcmp(ether, ac->ether, ETH_ALEN) == 0)
1873	{
1874	*ip = ac->ip;
1875	return VINF_SUCCESS;
1876	}
1877	}
1878	return VERR_NOT_FOUND;
1879	}
1880
1881	void slirp_arp_who_has(PNATState pData, uint32_t dst)
1882	{
1883	struct mbuf *m;
1884	struct ethhdr *ehdr;
1885	struct arphdr *ahdr;
1886	static bool fWarned = false;
1887	LogFlowFunc(("ENTER: %RTnaipv4\n", dst));
1888
1889	/* ARP request WHO HAS 0.0.0.0 is one of the signals
1890	* that something has been broken at Slirp. Investigating
1891	* pcap dumps it's easy to miss warning ARP requests being
1892	* focused on investigation of other protocols flow.
1893	*/
1894	#ifdef DEBUG_vvl
1895	Assert((dst != INADDR_ANY));
1896	NOREF(fWarned);
1897	#else
1898	if ( dst == INADDR_ANY
1899	&& !fWarned)
1900	{
1901	LogRel(("NAT:ARP: \"WHO HAS INADDR_ANY\" request has been detected\n"));
1902	fWarned = true;
1903	}
1904	#endif /* !DEBUG_vvl */
1905
1906	m = m_getcl(pData, M_NOWAIT, MT_HEADER, M_PKTHDR);
1907	if (m == NULL)
1908	{
1909	Log(("NAT: Can't alloc mbuf for ARP request\n"));
1910	LogFlowFuncLeave();
1911	return;
1912	}
1913	ehdr = mtod(m, struct ethhdr *);
1914	memset(ehdr->h_source, 0xff, ETH_ALEN);
1915	ahdr = (struct arphdr *)&ehdr[1];
1916	ahdr->ar_hrd = RT_H2N_U16_C(1);
1917	ahdr->ar_pro = RT_H2N_U16_C(ETH_P_IP);
1918	ahdr->ar_hln = ETH_ALEN;
1919	ahdr->ar_pln = 4;
1920	ahdr->ar_op = RT_H2N_U16_C(ARPOP_REQUEST);
1921	memcpy(ahdr->ar_sha, special_ethaddr, ETH_ALEN);
1922	/* we assume that this request come from gw, but not from DNS or TFTP */
1923	ahdr->ar_sha[5] = CTL_ALIAS;
1924	(uint32_t )ahdr->ar_sip = RT_H2N_U32(RT_N2H_U32(pData->special_addr.s_addr) \| CTL_ALIAS);
1925	memset(ahdr->ar_tha, 0xff, ETH_ALEN); /broadcast/
1926	(uint32_t )ahdr->ar_tip = dst;
1927	/* warn!!! should falls in mbuf minimal size */
1928	m->m_len = sizeof(struct arphdr) + ETH_HLEN;
1929	m->m_data += ETH_HLEN;
1930	m->m_len -= ETH_HLEN;
1931	if_encap(pData, ETH_P_ARP, m, ETH_ENCAP_URG);
1932	LogFlowFuncLeave();
1933	}
1934	#ifdef VBOX_WITH_DNSMAPPING_IN_HOSTRESOLVER
1935	void slirp_add_host_resolver_mapping(PNATState pData, const char pszHostName, const char pszHostNamePattern, uint32_t u32HostIP)
1936	{
1937	LogFlowFunc(("ENTER: pszHostName:%s, pszHostNamePattern:%s u32HostIP:%RTnaipv4\n",
1938	pszHostName ? pszHostName : "(null)",
1939	pszHostNamePattern ? pszHostNamePattern : "(null)",
1940	u32HostIP));
1941	if ( ( pszHostName
1942	\|\| pszHostNamePattern)
1943	&& u32HostIP != INADDR_ANY
1944	&& u32HostIP != INADDR_BROADCAST)
1945	{
1946	PDNSMAPPINGENTRY pDnsMapping = RTMemAllocZ(sizeof(DNSMAPPINGENTRY));
1947	if (!pDnsMapping)
1948	{
1949	LogFunc(("Can't allocate DNSMAPPINGENTRY\n"));
1950	LogFlowFuncLeave();
1951	return;
1952	}
1953	pDnsMapping->u32IpAddress = u32HostIP;
1954	if (pszHostName)
1955	pDnsMapping->pszCName = RTStrDup(pszHostName);
1956	else if (pszHostNamePattern)
1957	pDnsMapping->pszPattern = RTStrDup(pszHostNamePattern);
1958	if ( !pDnsMapping->pszCName
1959	&& !pDnsMapping->pszPattern)
1960	{
1961	LogFunc(("Can't allocate enough room for %s\n", pszHostName ? pszHostName : pszHostNamePattern));
1962	RTMemFree(pDnsMapping);
1963	LogFlowFuncLeave();
1964	return;
1965	}
1966	LIST_INSERT_HEAD(&pData->DNSMapHead, pDnsMapping, MapList);
1967	LogRel(("NAT: user-defined mapping %s: %RTnaipv4 is registered\n",
1968	pDnsMapping->pszCName ? pDnsMapping->pszCName : pDnsMapping->pszPattern,
1969	pDnsMapping->u32IpAddress));
1970	}
1971	LogFlowFuncLeave();
1972	}
1973	#endif
1974
1975	/* updates the arp cache
1976	* @note: this is helper function, slirp_arp_cache_update_or_add should be used.
1977	* @returns 0 - if has found and updated
1978	* 1 - if hasn't found.
1979	*/
1980	static inline int slirp_arp_cache_update(PNATState pData, uint32_t dst, const uint8_t *mac)
1981	{
1982	struct arp_cache_entry *ac;
1983	Assert(( memcmp(mac, broadcast_ethaddr, ETH_ALEN)
1984	&& memcmp(mac, zerro_ethaddr, ETH_ALEN)));
1985	LIST_FOREACH(ac, &pData->arp_cache, list)
1986	{
1987	if (ac->ip == dst)
1988	{
1989	memcpy(ac->ether, mac, ETH_ALEN);
1990	return 0;
1991	}
1992	}
1993	return 1;
1994	}
1995
1996	/**
1997	* add entry to the arp cache
1998	* @note: this is helper function, slirp_arp_cache_update_or_add should be used.
1999	*/
2000	static inline void slirp_arp_cache_add(PNATState pData, uint32_t ip, const uint8_t *ether)
2001	{
2002	struct arp_cache_entry *ac = NULL;
2003	Assert(( memcmp(ether, broadcast_ethaddr, ETH_ALEN)
2004	&& memcmp(ether, zerro_ethaddr, ETH_ALEN)));
2005	ac = RTMemAllocZ(sizeof(struct arp_cache_entry));
2006	if (ac == NULL)
2007	{
2008	Log(("NAT: Can't allocate arp cache entry\n"));
2009	return;
2010	}
2011	ac->ip = ip;
2012	memcpy(ac->ether, ether, ETH_ALEN);
2013	LIST_INSERT_HEAD(&pData->arp_cache, ac, list);
2014	}
2015
2016	/* updates or adds entry to the arp cache
2017	* @returns 0 - if has found and updated
2018	* 1 - if hasn't found.
2019	*/
2020	int slirp_arp_cache_update_or_add(PNATState pData, uint32_t dst, const uint8_t *mac)
2021	{
2022	if ( !memcmp(mac, broadcast_ethaddr, ETH_ALEN)
2023	\|\| !memcmp(mac, zerro_ethaddr, ETH_ALEN))
2024	{
2025	static bool fBroadcastEtherAddReported;
2026	if (!fBroadcastEtherAddReported)
2027	{
2028	LogRel(("NAT: Attempt to add pair [%RTmac:%RTnaipv4] in ARP cache was ignored\n",
2029	mac, dst));
2030	fBroadcastEtherAddReported = true;
2031	}
2032	return 1;
2033	}
2034	if (slirp_arp_cache_update(pData, dst, mac))
2035	slirp_arp_cache_add(pData, dst, mac);
2036
2037	return 0;
2038	}
2039
2040
2041	void slirp_set_mtu(PNATState pData, int mtu)
2042	{
2043	if (mtu < 20 \|\| mtu >= 16000)
2044	{
2045	LogRel(("NAT: mtu(%d) is out of range (20;16000] mtu forcely assigned to 1500\n", mtu));
2046	mtu = 1500;
2047	}
2048	/* MTU is maximum transition unit on */
2049	if_mtu =
2050	if_mru = mtu;
2051	}
2052
2053	/**
2054	* Info handler.
2055	*/
2056	void slirp_info(PNATState pData, const void pvArg, const char pszArgs)
2057	{
2058	struct socket so, so_next;
2059	struct arp_cache_entry *ac;
2060	struct port_forward_rule *rule;
2061	PCDBGFINFOHLP pHlp = (PCDBGFINFOHLP)pvArg;
2062	NOREF(pszArgs);
2063
2064	pHlp->pfnPrintf(pHlp, "NAT parameters: MTU=%d\n", if_mtu);
2065	pHlp->pfnPrintf(pHlp, "NAT TCP ports:\n");
2066	QSOCKET_FOREACH(so, so_next, tcp)
2067	/* { */
2068	pHlp->pfnPrintf(pHlp, " %R[natsock]\n", so);
2069	}
2070
2071	pHlp->pfnPrintf(pHlp, "NAT UDP ports:\n");
2072	QSOCKET_FOREACH(so, so_next, udp)
2073	/* { */
2074	pHlp->pfnPrintf(pHlp, " %R[natsock]\n", so);
2075	}
2076
2077	pHlp->pfnPrintf(pHlp, "NAT ARP cache:\n");
2078	LIST_FOREACH(ac, &pData->arp_cache, list)
2079	{
2080	pHlp->pfnPrintf(pHlp, " %RTnaipv4 %RTmac\n", ac->ip, &ac->ether);
2081	}
2082
2083	pHlp->pfnPrintf(pHlp, "NAT rules:\n");
2084	LIST_FOREACH(rule, &pData->port_forward_rule_head, list)
2085	{
2086	pHlp->pfnPrintf(pHlp, " %s %d => %RTnaipv4:%d %c\n",
2087	rule->proto == IPPROTO_UDP ? "UDP" : "TCP",
2088	rule->host_port, rule->guest_addr.s_addr, rule->guest_port,
2089	rule->activated ? ' ' : '*');
2090	}
2091	}
2092
2093	/**
2094	* @note: NATState::fUseHostResolver could be changed in bootp.c::dhcp_decode
2095	* @note: this function is executed on GUI/VirtualBox or main/VBoxHeadless thread.
2096	* @note: this function can potentially race with bootp.c::dhcp_decode (except Darwin)
2097	*/
2098	int slirp_host_network_configuration_change_strategy_selector(const PNATState pData)
2099	{
2100	if (pData->fUseHostResolverPermanent)
2101	return VBOX_NAT_DNS_HOSTRESOLVER;
2102
2103	if (pData->fUseDnsProxy) {
2104	#if HAVE_NOTIFICATION_FOR_DNS_UPDATE /* XXX */ && !defined(RT_OS_WINDOWS)
2105	/* We dont conflict with bootp.c::dhcp_decode */
2106	struct rcp_state rcp_state;
2107	int rc;
2108
2109	rcp_state.rcps_flags \|= RCPSF_IGNORE_IPV6;
2110	rc = rcp_parse(&rcp_state, RESOLV_CONF_FILE);
2111	LogRelFunc(("NAT: rcp_parse:%Rrc old domain:%s new domain:%s\n",
2112	rc, LIST_EMPTY(&pData->pDomainList)
2113	? "(null)"
2114	: LIST_FIRST(&pData->pDomainList)->dd_pszDomain,
2115	rcp_state.rcps_domain));
2116	if ( RT_FAILURE(rc)
2117	\|\| LIST_EMPTY(&pData->pDomainList))
2118	return VBOX_NAT_DNS_DNSPROXY;
2119
2120	if ( rcp_state.rcps_domain
2121	&& strcmp(rcp_state.rcps_domain, LIST_FIRST(&pData->pDomainList)->dd_pszDomain) == 0)
2122	return VBOX_NAT_DNS_DNSPROXY;
2123	else
2124	return VBOX_NAT_DNS_EXTERNAL;
2125	#else
2126	/* copy domain name */
2127	/* domain only compare with coy version */
2128	return VBOX_NAT_DNS_DNSPROXY;
2129	#endif
2130	}
2131	return VBOX_NAT_DNS_EXTERNAL;
2132	}

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

source: vbox/trunk/src/VBox/Devices/Network/slirp/slirp.c@ 55268

Download in other formats: