slirp.c@ 21004

Last change on this file since 21004 was 21004, checked in by vboxsync, 15 years ago
NAT: binding sockets to interface (default sockets and port-forwarding)
Property svn:eol-style set to `native`
File size: 54.3 KB

Line
1	#include "slirp.h"
2	#ifdef RT_OS_OS2
3	# include <paths.h>
4	#endif
5
6	#include <VBox/err.h>
7	#include <VBox/pdmdrv.h>
8	#include <iprt/assert.h>
9	#ifndef RT_OS_WINDOWS
10	# include <sys/ioctl.h>
11	# include <poll.h>
12	#else
13	# include <Winnls.h>
14	# define _WINSOCK2API_
15	# include <IPHlpApi.h>
16	#endif
17	#ifdef VBOX_WITH_SLIRP_ALIAS
18	# include <alias.h>
19	#endif
20
21	#if !defined(RT_OS_WINDOWS)
22
23	# define DO_ENGAGE_EVENT1(so, fdset, label) \
24	do { \
25	if( so->so_poll_index != -1 \
26	&& so->s == polls[so->so_poll_index].fd) { \
27	polls[so->so_poll_index].events \|= N_(fdset ## _poll); \
28	break; /* out of this loop */ \
29	} \
30	AssertRelease(poll_index < (nfds)); \
31	AssertRelease(poll_index >= 0 && poll_index < (nfds)); \
32	polls[poll_index].fd = (so)->s; \
33	(so)->so_poll_index = poll_index; \
34	polls[poll_index].events = N_(fdset ## _poll); \
35	polls[poll_index].revents = 0; \
36	poll_index++; \
37	} while(0)
38
39
40	# define DO_ENGAGE_EVENT2(so, fdset1, fdset2, label) \
41	do { \
42	if( so->so_poll_index != -1 \
43	&& so->s == polls[so->so_poll_index].fd) { \
44	polls[so->so_poll_index].events \|= \
45	N_(fdset1 ## _poll) \| N_(fdset1 ## _poll); \
46	break; /* out of this loop */ \
47	} \
48	AssertRelease(poll_index < (nfds)); \
49	polls[poll_index].fd = (so)->s; \
50	(so)->so_poll_index = poll_index; \
51	polls[poll_index].events = \
52	N_(fdset1 ## _poll) \| N_(fdset1 ## _poll); \
53	poll_index++; \
54	} while(0)
55
56	# define DO_POLL_EVENTS(rc, error, so, events, label) do {} while (0)
57
58	# define DO_CHECK_FD_SET(so, events, fdset) ( ((so)->so_poll_index != -1) \
59	&& ((so)->so_poll_index <= ndfs) \
60	&& ((so)->s == polls[so->so_poll_index].fd) \
61	&& (polls[(so)->so_poll_index].revents & N_(fdset ## _poll)))
62	# define DO_UNIX_CHECK_FD_SET(so, events, fdset ) DO_CHECK_FD_SET((so), (events), fdset) /specific for Unix API /
63	# define DO_WIN_CHECK_FD_SET(so, events, fdset ) 0 /* specific for Windows Winsock API */
64
65	# ifndef RT_OS_WINDOWS
66
67	# ifndef RT_OS_LINUX
68	# define readfds_poll (POLLRDNORM)
69	# define writefds_poll (POLLWRNORM)
70	# define xfds_poll (POLLRDBAND\|POLLWRBAND\|POLLPRI)
71	# else
72	# define readfds_poll (POLLIN)
73	# define writefds_poll (POLLOUT)
74	# define xfds_poll (POLLPRI)
75	# endif
76	# define rderr_poll (POLLERR)
77	# define rdhup_poll (POLLHUP)
78	# define nval_poll (POLLNVAL)
79
80	# define ICMP_ENGAGE_EVENT(so, fdset) \
81	do { \
82	if (pData->icmp_socket.s != -1) \
83	DO_ENGAGE_EVENT1((so), fdset, ICMP); \
84	} while (0)
85	# else /* !RT_OS_WINDOWS */
86	# define DO_WIN_CHECK_FD_SET(so, events, fdset ) DO_CHECK_FD_SET((so), (events), fdset)
87	# define ICMP_ENGAGE_EVENT(so, fdset) do {} while(0)
88	#endif /* RT_OS_WINDOWS */
89
90	#else /* defined(RT_OS_WINDOWS) */
91
92	/*
93	* On Windows, we will be notified by IcmpSendEcho2() when the response arrives.
94	* So no call to WSAEventSelect necessary.
95	*/
96	# define ICMP_ENGAGE_EVENT(so, fdset) do {} while(0)
97
98	# define DO_ENGAGE_EVENT1(so, fdset1, label) \
99	do { \
100	rc = WSAEventSelect((so)->s, VBOX_SOCKET_EVENT, FD_ALL_EVENTS); \
101	if (rc == SOCKET_ERROR) \
102	{ \
103	/* This should not happen */ \
104	error = WSAGetLastError(); \
105	LogRel(("WSAEventSelect (" #label ") error %d (so=%x, socket=%s, event=%x)\n", \
106	error, (so), (so)->s, VBOX_SOCKET_EVENT)); \
107	} \
108	} while(0); \
109	CONTINUE(label)
110
111	# define DO_ENGAGE_EVENT2(so, fdset1, fdset2, label) \
112	DO_ENGAGE_EVENT1((so), (fdset1), label)
113
114	# define DO_POLL_EVENTS(rc, error, so, events, label) \
115	(rc) = WSAEnumNetworkEvents((so)->s, VBOX_SOCKET_EVENT, (events)); \
116	if ((rc) == SOCKET_ERROR) \
117	{ \
118	(error) = WSAGetLastError(); \
119	LogRel(("WSAEnumNetworkEvents " #label " error %d\n", (error))); \
120	CONTINUE(label); \
121	}
122
123	# define acceptds_win FD_ACCEPT
124	# define acceptds_win_bit FD_ACCEPT_BIT
125
126	# define readfds_win FD_READ
127	# define readfds_win_bit FD_READ_BIT
128
129	# define writefds_win FD_WRITE
130	# define writefds_win_bit FD_WRITE_BIT
131
132	# define xfds_win FD_OOB
133	# define xfds_win_bit FD_OOB_BIT
134
135	# define DO_CHECK_FD_SET(so, events, fdset) \
136	(((events).lNetworkEvents & fdset ## _win) && ((events).iErrorCode[fdset ## _win_bit] == 0))
137
138	# define DO_WIN_CHECK_FD_SET(so, events, fdset ) DO_CHECK_FD_SET((so), (events), fdset)
139	# define DO_UNIX_CHECK_FD_SET(so, events, fdset ) 1 /specific for Unix API /
140
141	#endif /* defined(RT_OS_WINDOWS) */
142
143	#define TCP_ENGAGE_EVENT1(so, fdset) \
144	DO_ENGAGE_EVENT1((so), fdset, tcp)
145
146	#define TCP_ENGAGE_EVENT2(so, fdset1, fdset2) \
147	DO_ENGAGE_EVENT2((so), fdset1, fdset2, tcp)
148
149	#define UDP_ENGAGE_EVENT(so, fdset) \
150	DO_ENGAGE_EVENT1((so), fdset, udp)
151
152	#define POLL_TCP_EVENTS(rc, error, so, events) \
153	DO_POLL_EVENTS((rc), (error), (so), (events), tcp)
154
155	#define POLL_UDP_EVENTS(rc, error, so, events) \
156	DO_POLL_EVENTS((rc), (error), (so), (events), udp)
157
158	#define CHECK_FD_SET(so, events, set) \
159	(DO_CHECK_FD_SET((so), (events), set))
160
161	#define WIN_CHECK_FD_SET(so, events, set) \
162	(DO_WIN_CHECK_FD_SET((so), (events), set))
163	#define UNIX_CHECK_FD_SET(so, events, set) \
164	(DO_UNIX_CHECK_FD_SET(so, events, set))
165
166	/*
167	* Loging macros
168	*/
169	#if VBOX_WITH_DEBUG_NAT_SOCKETS
170	# if defined(RT_OS_WINDOWS)
171	# define DO_LOG_NAT_SOCK(so, proto, winevent, r_fdset, w_fdset, x_fdset) \
172	do { \
173	LogRel((" " #proto " %R[natsock] %R[natwinnetevents]\n", (so), (winevent))); \
174	} while (0)
175	# else /* RT_OS_WINDOWS */
176	# define DO_LOG_NAT_SOCK(so, proto, winevent, r_fdset, w_fdset, x_fdset) \
177	do { \
178	LogRel((" " #proto " %R[natsock] %s %s %s er: %s, %s, %s\n", (so), \
179	CHECK_FD_SET(so, ign ,r_fdset) ? "READ":"", \
180	CHECK_FD_SET(so, ign, w_fdset) ? "WRITE":"", \
181	CHECK_FD_SET(so, ign, x_fdset) ? "OOB":"", \
182	CHECK_FD_SET(so, ign, rderr) ? "RDERR":"", \
183	CHECK_FD_SET(so, ign, rdhup) ? "RDHUP":"", \
184	CHECK_FD_SET(so, ign, nval) ? "RDNVAL":"")); \
185	} while (0)
186	# endif /* !RT_OS_WINDOWS */
187	#else /* VBOX_WITH_DEBUG_NAT_SOCKETS */
188	# define DO_LOG_NAT_SOCK(so, proto, winevent, r_fdset, w_fdset, x_fdset) do {} while (0)
189	#endif /* !VBOX_WITH_DEBUG_NAT_SOCKETS */
190
191	#define LOG_NAT_SOCK(so, proto, winevent, r_fdset, w_fdset, x_fdset) DO_LOG_NAT_SOCK((so), proto, (winevent), r_fdset, w_fdset, x_fdset)
192
193	static const uint8_t special_ethaddr[6] =
194	{
195	0x52, 0x54, 0x00, 0x12, 0x35, 0x00
196	};
197
198	static const uint8_t broadcast_ethaddr[6] =
199	{
200	0xff, 0xff, 0xff, 0xff, 0xff, 0xff
201	};
202
203	const uint8_t zerro_ethaddr[6] =
204	{
205	0x0, 0x0, 0x0, 0x0, 0x0, 0x0
206	};
207
208	#ifdef RT_OS_WINDOWS
209	static int get_dns_addr_domain(PNATState pData, bool fVerbose,
210	struct in_addr *pdns_addr,
211	const char **ppszDomain)
212	{
213	/* Get amount of memory required for operation */
214	ULONG flags = GAA_FLAG_INCLUDE_PREFIX; /GAA_FLAG_INCLUDE_ALL_INTERFACES;/ /* all interfaces registered in NDIS */
215	PIP_ADAPTER_ADDRESSES addresses = NULL;
216	PIP_ADAPTER_ADDRESSES addr = NULL;
217	PIP_ADAPTER_DNS_SERVER_ADDRESS dns = NULL;
218	ULONG size = 0;
219	int wlen = 0;
220	char *suffix;
221	struct dns_entry *da = NULL;
222	struct dns_domain_entry *dd = NULL;
223	ULONG ret = ERROR_SUCCESS;
224
225	/* @todo add SKIPing flags to get only required information */
226
227	ret = pData->pfGetAdaptersAddresses(AF_INET, 0, NULL /* reserved */, addresses, &size);
228	if (ret != ERROR_BUFFER_OVERFLOW)
229	{
230	LogRel(("NAT: error %lu occurred on capacity detection operation\n", ret));
231	return -1;
232	}
233
234	if (size == 0)
235	{
236	LogRel(("NAT: Win socket API returns non capacity\n"));
237	return -1;
238	}
239
240	addresses = RTMemAllocZ(size);
241	if (addresses == NULL)
242	{
243	LogRel(("NAT: No memory available \n"));
244	return -1;
245	}
246
247	ret = pData->pfGetAdaptersAddresses(AF_INET, 0, NULL /* reserved */, addresses, &size);
248	if (ret != ERROR_SUCCESS)
249	{
250	LogRel(("NAT: error %lu occurred on fetching adapters info\n", ret));
251	RTMemFree(addresses);
252	return -1;
253	}
254	addr = addresses;
255	while(addr != NULL)
256	{
257	int found;
258	if (addr->OperStatus != IfOperStatusUp)
259	goto next;
260	dns = addr->FirstDnsServerAddress;
261	while (dns != NULL)
262	{
263	struct sockaddr *saddr = dns->Address.lpSockaddr;
264	if (saddr->sa_family != AF_INET)
265	goto next_dns;
266	/* add dns server to list */
267	da = RTMemAllocZ(sizeof(struct dns_entry));
268	if (da == NULL)
269	{
270	LogRel(("NAT: Can't allocate buffer for DNS entry\n"));
271	RTMemFree(addresses);
272	return VERR_NO_MEMORY;
273	}
274	LogRel(("NAT: adding %R[IP4] to DNS server list\n", &((struct sockaddr_in *)saddr)->sin_addr));
275	if ((((struct sockaddr_in *)saddr)->sin_addr.s_addr & htonl(IN_CLASSA_NET)) == ntohl(INADDR_LOOPBACK & IN_CLASSA_NET)) {
276	da->de_addr.s_addr = htonl(ntohl(special_addr.s_addr) \| CTL_ALIAS);
277	}
278	else
279	{
280	da->de_addr.s_addr = ((struct sockaddr_in *)saddr)->sin_addr.s_addr;
281	}
282	TAILQ_INSERT_HEAD(&pData->dns_list_head, da, de_list);
283
284	if (addr->DnsSuffix == NULL)
285	goto next_dns;
286
287	/uniq/
288	RTUtf16ToUtf8(addr->DnsSuffix, &suffix);
289
290	if (!suffix \|\| strlen(suffix) == 0) {
291	RTStrFree(suffix);
292	goto next_dns;
293	}
294
295	found = 0;
296	LIST_FOREACH(dd, &pData->dns_domain_list_head, dd_list)
297	{
298	if ( dd->dd_pszDomain != NULL
299	&& strcmp(dd->dd_pszDomain, suffix) == 0)
300	{
301	found = 1;
302	RTStrFree(suffix);
303	break;
304	}
305	}
306	if (found == 0)
307	{
308	dd = RTMemAllocZ(sizeof(struct dns_domain_entry));
309	if (dd == NULL)
310	{
311	LogRel(("NAT: not enough memory\n"));
312	RTStrFree(suffix);
313	RTMemFree(addresses);
314	return VERR_NO_MEMORY;
315	}
316	dd->dd_pszDomain = suffix;
317	LogRel(("NAT: adding domain name %s to search list\n", dd->dd_pszDomain));
318	LIST_INSERT_HEAD(&pData->dns_domain_list_head, dd, dd_list);
319	}
320	next_dns:
321	dns = dns->Next;
322	}
323	next:
324	addr = addr->Next;
325	}
326	RTMemFree(addresses);
327	return 0;
328	}
329
330	#else /* !RT_OS_WINDOWS */
331
332	static int get_dns_addr_domain(PNATState pData, bool fVerbose,
333	struct in_addr *pdns_addr,
334	const char **ppszDomain)
335	{
336	char buff[512];
337	char buff2[256];
338	FILE *f = NULL;
339	int found = 0;
340	struct in_addr tmp_addr;
341
342	#ifdef RT_OS_OS2
343	/* Try various locations. */
344	char *etc = getenv("ETC");
345	if (etc)
346	{
347	snprintf(buff, sizeof(buff), "%s/RESOLV2", etc);
348	f = fopen(buff, "rt");
349	}
350	if (!f)
351	{
352	snprintf(buff, sizeof(buff), "%s/RESOLV2", _PATH_ETC);
353	f = fopen(buff, "rt");
354	}
355	if (!f)
356	{
357	snprintf(buff, sizeof(buff), "%s/resolv.conf", _PATH_ETC);
358	f = fopen(buff, "rt");
359	}
360	#else
361	#ifndef DEBUG_vvl
362	f = fopen("/etc/resolv.conf", "r");
363	#else
364	char *home = getenv("HOME");
365	snprintf(buff, sizeof(buff), "%s/resolv.conf", home);
366	f = fopen(buff, "r");
367	if (f != NULL)
368	{
369	Log(("NAT: DNS we're using %s\n", buff));
370	}
371	else
372	{
373	f = fopen("/etc/resolv.conf", "r");
374	Log(("NAT: DNS we're using %s\n", buff));
375	}
376	#endif
377	#endif
378	if (!f)
379	return -1;
380
381	if (ppszDomain)
382	*ppszDomain = NULL;
383	Log(("nat: DNS Servers:\n"));
384	while (fgets(buff, 512, f) != NULL)
385	{
386	struct dns_entry *da = NULL;
387	if (sscanf(buff, "nameserver%*[ \t]%256s", buff2) == 1)
388	{
389	if (!inet_aton(buff2, &tmp_addr))
390	continue;
391	/localhost mask /
392	da = RTMemAllocZ(sizeof (struct dns_entry));
393	if (da == NULL)
394	{
395	LogRel(("can't alloc memory for DNS entry\n"));
396	return -1;
397	}
398	/check /
399	da->de_addr.s_addr = tmp_addr.s_addr;
400	if ((da->de_addr.s_addr & htonl(IN_CLASSA_NET)) == ntohl(INADDR_LOOPBACK & IN_CLASSA_NET)) {
401	da->de_addr.s_addr = htonl(ntohl(special_addr.s_addr) \| CTL_ALIAS);
402	}
403	TAILQ_INSERT_HEAD(&pData->dns_list_head, da, de_list);
404	found++;
405	}
406	if ((!strncmp(buff, "domain", 6) \|\| !strncmp(buff, "search", 6)))
407	{
408	char *tok;
409	char *saveptr;
410	struct dns_domain_entry *dd = NULL;
411	int found = 0;
412	tok = strtok_r(&buff[6], " \t\n", &saveptr);
413	LIST_FOREACH(dd, &pData->dns_domain_list_head, dd_list)
414	{
415	if( tok != NULL
416	&& strcmp(tok, dd->dd_pszDomain) == 0)
417	{
418	found = 1;
419	break;
420	}
421	}
422	if (tok != NULL && found == 0) {
423	dd = RTMemAllocZ(sizeof(struct dns_domain_entry));
424	if (dd == NULL)
425	{
426	LogRel(("NAT: not enought memory to add domain list\n"));
427	return VERR_NO_MEMORY;
428	}
429	dd->dd_pszDomain = RTStrDup(tok);
430	LogRel(("NAT: adding domain name %s to search list\n", dd->dd_pszDomain));
431	LIST_INSERT_HEAD(&pData->dns_domain_list_head, dd, dd_list);
432	}
433	}
434	}
435	fclose(f);
436	if (!found)
437	return -1;
438	return 0;
439	}
440
441	#endif
442
443	static int slirp_init_dns_list(PNATState pData)
444	{
445	TAILQ_INIT(&pData->dns_list_head);
446	LIST_INIT(&pData->dns_domain_list_head);
447	return get_dns_addr_domain(pData, true, NULL, NULL);
448	}
449
450	static void slirp_release_dns_list(PNATState pData)
451	{
452	struct dns_entry *de = NULL;
453	struct dns_domain_entry *dd = NULL;
454	while(!TAILQ_EMPTY(&pData->dns_list_head)) {
455	de = TAILQ_FIRST(&pData->dns_list_head);
456	TAILQ_REMOVE(&pData->dns_list_head, de, de_list);
457	RTMemFree(de);
458	}
459	while(!LIST_EMPTY(&pData->dns_domain_list_head)) {
460	dd = LIST_FIRST(&pData->dns_domain_list_head);
461	LIST_REMOVE(dd, dd_list);
462	if (dd->dd_pszDomain != NULL)
463	RTStrFree(dd->dd_pszDomain);
464	RTMemFree(dd);
465	}
466	}
467
468	int get_dns_addr(PNATState pData, struct in_addr *pdns_addr)
469	{
470	return get_dns_addr_domain(pData, false, pdns_addr, NULL);
471	}
472
473	#ifndef VBOX_WITH_NAT_SERVICE
474	int slirp_init(PNATState ppData, const char pszNetAddr, uint32_t u32Netmask,
475	bool fPassDomain, void *pvUser)
476	#else
477	int slirp_init(PNATState *ppData, uint32_t u32NetAddr, uint32_t u32Netmask,
478	bool fPassDomain, void *pvUser)
479	#endif
480	{
481	int fNATfailed = 0;
482	int rc;
483	PNATState pData = RTMemAllocZ(sizeof(NATState));
484	*ppData = pData;
485	if (!pData)
486	return VERR_NO_MEMORY;
487	if (u32Netmask & 0x1f)
488	/* CTL is x.x.x.15, bootp passes up to 16 IPs (15..31) */
489	return VERR_INVALID_PARAMETER;
490	pData->fPassDomain = fPassDomain;
491	pData->pvUser = pvUser;
492	pData->netmask = u32Netmask;
493
494	/* sockets & TCP defaults */
495	pData->socket_rcv = 64 * _1K;
496	pData->socket_snd = 64 * _1K;
497	tcp_sndspace = 64 * _1K;
498	tcp_rcvspace = 64 * _1K;
499
500	#ifdef RT_OS_WINDOWS
501	{
502	WSADATA Data;
503	WSAStartup(MAKEWORD(2, 0), &Data);
504	}
505	pData->phEvents[VBOX_SOCKET_EVENT_INDEX] = CreateEvent(NULL, FALSE, FALSE, NULL);
506	#endif
507	#ifdef VBOX_WITH_SLIRP_MT
508	QSOCKET_LOCK_CREATE(tcb);
509	QSOCKET_LOCK_CREATE(udb);
510	rc = RTReqCreateQueue(&pData->pReqQueue);
511	AssertReleaseRC(rc);
512	#endif
513
514	link_up = 1;
515
516	debug_init();
517	if_init(pData);
518	ip_init(pData);
519	icmp_init(pData);
520
521	/* Initialise mbufs after setting the MTU */
522	m_init(pData);
523
524	#ifndef VBOX_WITH_NAT_SERVICE
525	inet_aton(pszNetAddr, &special_addr);
526	#else
527	special_addr.s_addr = u32NetAddr;
528	#endif
529	#ifdef VBOX_WITHOUT_SLIRP_CLIENT_ETHER
530	pData->slirp_ethaddr = &special_ethaddr[0];
531	#endif
532	alias_addr.s_addr = special_addr.s_addr \| htonl(CTL_ALIAS);
533	/* @todo: add ability to configure this staff */
534
535	/* set default addresses */
536	inet_aton("127.0.0.1", &loopback_addr);
537	if (slirp_init_dns_list(pData) < 0)
538	fNATfailed = 1;
539
540	dnsproxy_init(pData);
541
542	getouraddr(pData);
543
544	#ifdef VBOX_WITH_SLIRP_ALIAS
545	{
546	int flags = 0;
547	struct in_addr proxy_addr;
548	pData->proxy_alias = LibAliasInit(pData, NULL);
549	if (pData->proxy_alias == NULL)
550	{
551	LogRel(("NAT: LibAlias default rule wasn't initialized\n"));
552	AssertMsgFailed(("NAT: LibAlias default rule wasn't initialized\n"));
553	}
554	flags = LibAliasSetMode(pData->proxy_alias, 0, 0);
555	flags \|= PKT_ALIAS_LOG; /* set logging */
556	flags = LibAliasSetMode(pData->proxy_alias, flags, ~0);
557	proxy_addr.s_addr = htonl(ntohl(special_addr.s_addr) \| CTL_ALIAS);
558	LibAliasSetAddress(pData->proxy_alias, proxy_addr);
559	ftp_alias_load();
560	nbt_alias_load();
561
562	}
563	#endif
564	return fNATfailed ? VINF_NAT_DNS : VINF_SUCCESS;
565	}
566
567	/**
568	* Register statistics.
569	*/
570	void slirp_register_statistics(PNATState pData, PPDMDRVINS pDrvIns)
571	{
572	#ifdef VBOX_WITH_STATISTICS
573	# define COUNTER(name, type, units, dsc) \
574	do { \
575	PDMDrvHlpSTAMRegisterF(pDrvIns, \
576	&pData->Stat ## name, \
577	type, \
578	STAMVISIBILITY_ALWAYS, \
579	units, \
580	dsc, \
581	"/Drivers/NAT%u/" #name, \
582	pDrvIns->iInstance); \
583	} while (0)
584
585	# define PROFILE_COUNTER(name, dsc) COUNTER(name, STAMTYPE_PROFILE, STAMUNIT_TICKS_PER_CALL, dsc)
586	# define COUNTING_COUNTER(name, dsc) COUNTER(name, STAMTYPE_COUNTER, STAMUNIT_COUNT, dsc)
587
588	# include "counters.h"
589
590	# undef COUNTER
591	# undef PROFILE_COUNTER
592	# undef COUNTING_COUNTER
593	/** @todo register statistics for the variables dumped by:
594	* ipstats(pData); tcpstats(pData); udpstats(pData); icmpstats(pData);
595	* mbufstats(pData); sockstats(pData); */
596	#endif /* VBOX_WITH_STATISTICS */
597	}
598
599	/**
600	* Deregister statistics.
601	*/
602	void slirp_deregister_statistics(PNATState pData, PPDMDRVINS pDrvIns)
603	{
604	#ifdef VBOX_WITH_STATISTICS
605	# define PROFILE_COUNTER(name, dsc) PDMDrvHlpSTAMDeregister(pDrvIns, &pData->Stat ## name)
606	# define COUNTING_COUNTER(name, dsc) PDMDrvHlpSTAMDeregister(pDrvIns, &pData->Stat ## name)
607
608	# include "counters.h"
609
610	# undef COUNTING_COUNTER
611	# undef PROFILE_COUNTER
612	#endif /* VBOX_WITH_STATISTICS */
613	}
614
615	/**
616	* Marks the link as up, making it possible to establish new connections.
617	*/
618	void slirp_link_up(PNATState pData)
619	{
620	link_up = 1;
621	}
622
623	/**
624	* Marks the link as down and cleans up the current connections.
625	*/
626	void slirp_link_down(PNATState pData)
627	{
628	struct socket *so;
629
630	while ((so = tcb.so_next) != &tcb)
631	{
632	if (so->so_state & SS_NOFDREF \|\| so->s == -1)
633	sofree(pData, so);
634	else
635	tcp_drop(pData, sototcpcb(so), 0);
636	}
637
638	while ((so = udb.so_next) != &udb)
639	udp_detach(pData, so);
640
641	link_up = 0;
642	}
643
644	/**
645	* Terminates the slirp component.
646	*/
647	void slirp_term(PNATState pData)
648	{
649	#ifdef RT_OS_WINDOWS
650	pData->pfIcmpCloseHandle(pData->icmp_socket.sh);
651	FreeLibrary(pData->hmIcmpLibrary);
652	RTMemFree(pData->pvIcmpBuffer);
653	# else
654	closesocket(pData->icmp_socket.s);
655	#endif
656
657	slirp_link_down(pData);
658	slirp_release_dns_list(pData);
659	#ifdef RT_OS_WINDOWS
660	WSACleanup();
661	#endif
662	#ifdef VBOX_WITH_SLIRP_ALIAS
663	while(!LIST_EMPTY(&instancehead)) {
664	struct libalias *la = LIST_FIRST(&instancehead);
665	/* libalias do all clean up */
666	LibAliasUninit(la);
667	}
668	#endif
669	#ifdef LOG_ENABLED
670	Log(("\n"
671	"NAT statistics\n"
672	"--------------\n"
673	"\n"));
674	ipstats(pData);
675	tcpstats(pData);
676	udpstats(pData);
677	icmpstats(pData);
678	mbufstats(pData);
679	sockstats(pData);
680	Log(("\n"
681	"\n"
682	"\n"));
683	#endif
684	RTMemFree(pData);
685	}
686
687
688	#define CONN_CANFSEND(so) (((so)->so_state & (SS_FCANTSENDMORE\|SS_ISFCONNECTED)) == SS_ISFCONNECTED)
689	#define CONN_CANFRCV(so) (((so)->so_state & (SS_FCANTRCVMORE\|SS_ISFCONNECTED)) == SS_ISFCONNECTED)
690	#define UPD_NFDS(x) do { if (nfds < (x)) nfds = (x); } while (0)
691
692	/*
693	* curtime kept to an accuracy of 1ms
694	*/
695	static void updtime(PNATState pData)
696	{
697	#ifdef RT_OS_WINDOWS
698	struct _timeb tb;
699
700	_ftime(&tb);
701	curtime = (u_int)tb.time * (u_int)1000;
702	curtime += (u_int)tb.millitm;
703	#else
704	gettimeofday(&tt, 0);
705
706	curtime = (u_int)tt.tv_sec * (u_int)1000;
707	curtime += (u_int)tt.tv_usec / (u_int)1000;
708
709	if ((tt.tv_usec % 1000) >= 500)
710	curtime++;
711	#endif
712	}
713
714	#ifdef RT_OS_WINDOWS
715	void slirp_select_fill(PNATState pData, int *pnfds)
716	#else /* RT_OS_WINDOWS */
717	void slirp_select_fill(PNATState pData, int pnfds, struct pollfd polls)
718	#endif /* !RT_OS_WINDOWS */
719	{
720	struct socket so, so_next;
721	int nfds;
722	#if defined(RT_OS_WINDOWS)
723	int rc;
724	int error;
725	#else
726	int poll_index = 0;
727	#endif
728	int i;
729
730	STAM_PROFILE_START(&pData->StatFill, a);
731
732	nfds = *pnfds;
733
734	/*
735	* First, TCP sockets
736	*/
737	do_slowtimo = 0;
738	if (!link_up)
739	goto done;
740	/*
741	* *_slowtimo needs calling if there are IP fragments
742	* in the fragment queue, or there are TCP connections active
743	*/
744	/* XXX:
745	* triggering of fragment expiration should be the same but use new macroses
746	*/
747	do_slowtimo = (tcb.so_next != &tcb);
748	if (!do_slowtimo)
749	{
750	for (i = 0; i < IPREASS_NHASH; i++)
751	{
752	if (!TAILQ_EMPTY(&ipq[i]))
753	{
754	do_slowtimo = 1;
755	break;
756	}
757	}
758	}
759	ICMP_ENGAGE_EVENT(&pData->icmp_socket, readfds);
760
761	STAM_COUNTER_RESET(&pData->StatTCP);
762	STAM_COUNTER_RESET(&pData->StatTCPHot);
763
764	QSOCKET_FOREACH(so, so_next, tcp)
765	/* { */
766	#if !defined(RT_OS_WINDOWS)
767	so->so_poll_index = -1;
768	#endif
769	STAM_COUNTER_INC(&pData->StatTCP);
770
771	/*
772	* See if we need a tcp_fasttimo
773	*/
774	if ( time_fasttimo == 0
775	&& so->so_tcpcb != NULL
776	&& so->so_tcpcb->t_flags & TF_DELACK)
777	time_fasttimo = curtime; /* Flag when we want a fasttimo */
778
779	/*
780	* NOFDREF can include still connecting to local-host,
781	* newly socreated() sockets etc. Don't want to select these.
782	*/
783	if (so->so_state & SS_NOFDREF \|\| so->s == -1)
784	CONTINUE(tcp);
785
786	/*
787	* Set for reading sockets which are accepting
788	*/
789	if (so->so_state & SS_FACCEPTCONN)
790	{
791	STAM_COUNTER_INC(&pData->StatTCPHot);
792	TCP_ENGAGE_EVENT1(so, readfds);
793	CONTINUE(tcp);
794	}
795
796	/*
797	* Set for writing sockets which are connecting
798	*/
799	if (so->so_state & SS_ISFCONNECTING)
800	{
801	Log2(("connecting %R[natsock] engaged\n",so));
802	STAM_COUNTER_INC(&pData->StatTCPHot);
803	TCP_ENGAGE_EVENT1(so, writefds);
804	}
805
806	/*
807	* Set for writing if we are connected, can send more, and
808	* we have something to send
809	*/
810	if (CONN_CANFSEND(so) && so->so_rcv.sb_cc)
811	{
812	STAM_COUNTER_INC(&pData->StatTCPHot);
813	TCP_ENGAGE_EVENT1(so, writefds);
814	}
815
816	/*
817	* Set for reading (and urgent data) if we are connected, can
818	* receive more, and we have room for it XXX /2 ?
819	*/
820	if (CONN_CANFRCV(so) && (so->so_snd.sb_cc < (so->so_snd.sb_datalen/2)))
821	{
822	STAM_COUNTER_INC(&pData->StatTCPHot);
823	TCP_ENGAGE_EVENT2(so, readfds, xfds);
824	}
825	LOOP_LABEL(tcp, so, so_next);
826	}
827
828	/*
829	* UDP sockets
830	*/
831	STAM_COUNTER_RESET(&pData->StatUDP);
832	STAM_COUNTER_RESET(&pData->StatUDPHot);
833
834	QSOCKET_FOREACH(so, so_next, udp)
835	/* { */
836
837	STAM_COUNTER_INC(&pData->StatUDP);
838	#if !defined(RT_OS_WINDOWS)
839	so->so_poll_index = -1;
840	#endif
841
842	/*
843	* See if it's timed out
844	*/
845	if (so->so_expire)
846	{
847	if (so->so_expire <= curtime)
848	{
849	Log2(("NAT: %R[natsock] expired\n", so));
850	if (so->so_timeout != NULL)
851	{
852	so->so_timeout(pData, so, so->so_timeout_arg);
853	}
854	#ifdef VBOX_WITH_SLIRP_MT
855	/* we need so_next for continue our cycle*/
856	so_next = so->so_next;
857	#endif
858	UDP_DETACH(pData, so, so_next);
859	CONTINUE_NO_UNLOCK(udp);
860	}
861	else
862	do_slowtimo = 1; /* Let socket expire */
863	}
864
865	/*
866	* When UDP packets are received from over the link, they're
867	* sendto()'d straight away, so no need for setting for writing
868	* Limit the number of packets queued by this session to 4.
869	* Note that even though we try and limit this to 4 packets,
870	* the session could have more queued if the packets needed
871	* to be fragmented.
872	*
873	* (XXX <= 4 ?)
874	*/
875	if ((so->so_state & SS_ISFCONNECTED) && so->so_queued <= 4)
876	{
877	STAM_COUNTER_INC(&pData->StatUDPHot);
878	UDP_ENGAGE_EVENT(so, readfds);
879	}
880	LOOP_LABEL(udp, so, so_next);
881	}
882	done:
883
884	#if defined(RT_OS_WINDOWS)
885	*pnfds = VBOX_EVENT_COUNT;
886	#else /* RT_OS_WINDOWS */
887	AssertRelease(poll_index <= *pnfds);
888	*pnfds = poll_index;
889	#endif /* !RT_OS_WINDOWS */
890
891	STAM_PROFILE_STOP(&pData->StatFill, a);
892	}
893
894	#if defined(RT_OS_WINDOWS)
895	void slirp_select_poll(PNATState pData, int fTimeout, int fIcmp)
896	#else /* RT_OS_WINDOWS */
897	void slirp_select_poll(PNATState pData, struct pollfd *polls, int ndfs)
898	#endif /* !RT_OS_WINDOWS */
899	{
900	struct socket so, so_next;
901	int ret;
902	#if defined(RT_OS_WINDOWS)
903	WSANETWORKEVENTS NetworkEvents;
904	int rc;
905	int error;
906	#else
907	int poll_index = 0;
908	#endif
909
910	STAM_PROFILE_START(&pData->StatPoll, a);
911
912	/* Update time */
913	updtime(pData);
914
915	/*
916	* See if anything has timed out
917	*/
918	if (link_up)
919	{
920	if (time_fasttimo && ((curtime - time_fasttimo) >= 2))
921	{
922	STAM_PROFILE_START(&pData->StatFastTimer, a);
923	tcp_fasttimo(pData);
924	time_fasttimo = 0;
925	STAM_PROFILE_STOP(&pData->StatFastTimer, a);
926	}
927	if (do_slowtimo && ((curtime - last_slowtimo) >= 499))
928	{
929	STAM_PROFILE_START(&pData->StatSlowTimer, a);
930	ip_slowtimo(pData);
931	tcp_slowtimo(pData);
932	last_slowtimo = curtime;
933	STAM_PROFILE_STOP(&pData->StatSlowTimer, a);
934	}
935	}
936	#if defined(RT_OS_WINDOWS)
937	if (fTimeout)
938	return; /* only timer update */
939	#endif
940
941	/*
942	* Check sockets
943	*/
944	if (!link_up)
945	goto done;
946	#if defined(RT_OS_WINDOWS)
947	/*XXX: before renaming please make see define
948	* fIcmp in slirp_state.h
949	*/
950	if (fIcmp)
951	sorecvfrom(pData, &pData->icmp_socket);
952	#else
953	if ( (pData->icmp_socket.s != -1)
954	&& CHECK_FD_SET(&pData->icmp_socket, ignored, readfds))
955	sorecvfrom(pData, &pData->icmp_socket);
956	#endif
957	/*
958	* Check TCP sockets
959	*/
960	QSOCKET_FOREACH(so, so_next, tcp)
961	/* { */
962
963	#ifdef VBOX_WITH_SLIRP_MT
964	if ( so->so_state & SS_NOFDREF
965	&& so->so_deleted == 1)
966	{
967	struct socket son, sop = NULL;
968	QSOCKET_LOCK(tcb);
969	if (so->so_next != NULL)
970	{
971	if (so->so_next != &tcb)
972	SOCKET_LOCK(so->so_next);
973	son = so->so_next;
974	}
975	if ( so->so_prev != &tcb
976	&& so->so_prev != NULL)
977	{
978	SOCKET_LOCK(so->so_prev);
979	sop = so->so_prev;
980	}
981	QSOCKET_UNLOCK(tcb);
982	remque(pData, so);
983	NSOCK_DEC();
984	SOCKET_UNLOCK(so);
985	SOCKET_LOCK_DESTROY(so);
986	RTMemFree(so);
987	so_next = son;
988	if (sop != NULL)
989	SOCKET_UNLOCK(sop);
990	CONTINUE_NO_UNLOCK(tcp);
991	}
992	#endif
993	/*
994	* FD_ISSET is meaningless on these sockets
995	* (and they can crash the program)
996	*/
997	if (so->so_state & SS_NOFDREF \|\| so->s == -1)
998	CONTINUE(tcp);
999
1000	POLL_TCP_EVENTS(rc, error, so, &NetworkEvents);
1001
1002	LOG_NAT_SOCK(so, TCP, &NetworkEvents, readfds, writefds, xfds);
1003
1004
1005	/*
1006	* Check for URG data
1007	* This will soread as well, so no need to
1008	* test for readfds below if this succeeds
1009	*/
1010
1011	/* out-of-band data */
1012	if (CHECK_FD_SET(so, NetworkEvents, xfds))
1013	{
1014	sorecvoob(pData, so);
1015	}
1016
1017	/*
1018	* Check sockets for reading
1019	*/
1020	else if ( CHECK_FD_SET(so, NetworkEvents, readfds)
1021	\|\| WIN_CHECK_FD_SET(so, NetworkEvents, acceptds))
1022	{
1023	/*
1024	* Check for incoming connections
1025	*/
1026	if (so->so_state & SS_FACCEPTCONN)
1027	{
1028	TCP_CONNECT(pData, so);
1029	#if defined(RT_OS_WINDOWS)
1030	if (!(NetworkEvents.lNetworkEvents & FD_CLOSE))
1031	#endif
1032	CONTINUE(tcp);
1033	}
1034
1035	ret = soread(pData, so);
1036	/* Output it if we read something */
1037	if (RT_LIKELY(ret > 0))
1038	TCP_OUTPUT(pData, sototcpcb(so));
1039	}
1040
1041	#if defined(RT_OS_WINDOWS)
1042	/*
1043	* Check for FD_CLOSE events.
1044	* in some cases once FD_CLOSE engaged on socket it could be flashed latter (for some reasons)
1045	*/
1046	if ( (NetworkEvents.lNetworkEvents & FD_CLOSE)
1047	\|\| (so->so_close == 1))
1048	{
1049	so->so_close = 1; /* mark it */
1050	/*
1051	* drain the socket
1052	*/
1053	for (;;)
1054	{
1055	ret = soread(pData, so);
1056	if (ret > 0)
1057	TCP_OUTPUT(pData, sototcpcb(so));
1058	else
1059	break;
1060	}
1061	CONTINUE(tcp);
1062	}
1063	#endif
1064
1065	/*
1066	* Check sockets for writing
1067	*/
1068	if (CHECK_FD_SET(so, NetworkEvents, writefds))
1069	{
1070	/*
1071	* Check for non-blocking, still-connecting sockets
1072	*/
1073	if (so->so_state & SS_ISFCONNECTING)
1074	{
1075	Log2(("connecting %R[natsock] catched\n", so));
1076	/* Connected */
1077	so->so_state &= ~SS_ISFCONNECTING;
1078
1079	/*
1080	* This should be probably guarded by PROBE_CONN too. Anyway,
1081	* we disable it on OS/2 because the below send call returns
1082	* EFAULT which causes the opened TCP socket to close right
1083	* after it has been opened and connected.
1084	*/
1085	#ifndef RT_OS_OS2
1086	ret = send(so->s, (const char *)&ret, 0, 0);
1087	if (ret < 0)
1088	{
1089	/* XXXXX Must fix, zero bytes is a NOP */
1090	if ( errno == EAGAIN
1091	\|\| errno == EWOULDBLOCK
1092	\|\| errno == EINPROGRESS
1093	\|\| errno == ENOTCONN)
1094	CONTINUE(tcp);
1095
1096	/* else failed */
1097	so->so_state = SS_NOFDREF;
1098	}
1099	/* else so->so_state &= ~SS_ISFCONNECTING; */
1100	#endif
1101
1102	/*
1103	* Continue tcp_input
1104	*/
1105	TCP_INPUT(pData, (struct mbuf *)NULL, sizeof(struct ip), so);
1106	/* continue; */
1107	}
1108	else
1109	SOWRITE(ret, pData, so);
1110	/*
1111	* XXX If we wrote something (a lot), there could be the need
1112	* for a window update. In the worst case, the remote will send
1113	* a window probe to get things going again.
1114	*/
1115	}
1116
1117	/*
1118	* Probe a still-connecting, non-blocking socket
1119	* to check if it's still alive
1120	*/
1121	#ifdef PROBE_CONN
1122	if (so->so_state & SS_ISFCONNECTING)
1123	{
1124	ret = recv(so->s, (char *)&ret, 0, 0);
1125
1126	if (ret < 0)
1127	{
1128	/* XXX */
1129	if ( errno == EAGAIN
1130	\|\| errno == EWOULDBLOCK
1131	\|\| errno == EINPROGRESS
1132	\|\| errno == ENOTCONN)
1133	{
1134	CONTINUE(tcp); /* Still connecting, continue */
1135	}
1136
1137	/* else failed */
1138	so->so_state = SS_NOFDREF;
1139
1140	/* tcp_input will take care of it */
1141	}
1142	else
1143	{
1144	ret = send(so->s, &ret, 0, 0);
1145	if (ret < 0)
1146	{
1147	/* XXX */
1148	if ( errno == EAGAIN
1149	\|\| errno == EWOULDBLOCK
1150	\|\| errno == EINPROGRESS
1151	\|\| errno == ENOTCONN)
1152	{
1153	CONTINUE(tcp);
1154	}
1155	/* else failed */
1156	so->so_state = SS_NOFDREF;
1157	}
1158	else
1159	so->so_state &= ~SS_ISFCONNECTING;
1160
1161	}
1162	TCP_INPUT((struct mbuf *)NULL, sizeof(struct ip),so);
1163	} /* SS_ISFCONNECTING */
1164	#endif
1165	#ifndef RT_OS_WINDOWS
1166	if ( UNIX_CHECK_FD_SET(so, NetworkEvents, rdhup)
1167	\|\| UNIX_CHECK_FD_SET(so, NetworkEvents, rderr))
1168	{
1169	int err;
1170	int inq, outq;
1171	int status;
1172	socklen_t optlen = sizeof(int);
1173	inq = outq = 0;
1174	status = getsockopt(so->s, SOL_SOCKET, SO_ERROR, &err, &optlen);
1175	if (status != 0)
1176	Log(("NAT: can't get error status from %R[natsock]\n", so));
1177	#ifndef RT_OS_SOLARIS
1178	status = ioctl(so->s, FIONREAD, &inq); /* tcp(7) recommends SIOCINQ which is Linux specific */
1179	if (status != 0 \|\| status != EINVAL)
1180	{
1181	/* EINVAL returned if socket in listen state tcp(7)*/
1182	Log(("NAT: can't get depth of IN queue status from %R[natsock]\n", so));
1183	}
1184	status = ioctl(so->s, TIOCOUTQ, &outq); /* SIOCOUTQ see previous comment */
1185	if (status != 0)
1186	Log(("NAT: can't get depth of OUT queue from %R[natsock]\n", so));
1187	#else
1188	/*
1189	* Solaris has bit different ioctl commands and its handlings
1190	* hint: streamio(7) I_NREAD
1191	*/
1192	#endif
1193	if ( so->so_state & SS_ISFCONNECTING
1194	\|\| UNIX_CHECK_FD_SET(so, NetworkEvents, readfds))
1195	{
1196	/**
1197	* Check if we need here take care about gracefull connection
1198	* @todo try with proxy server
1199	*/
1200	if (UNIX_CHECK_FD_SET(so, NetworkEvents, readfds))
1201	{
1202	/*
1203	* Never meet inq != 0 or outq != 0, anyway let it stay for a while
1204	* in case it happens we'll able to detect it.
1205	* Give TCP/IP stack wait or expire the socket.
1206	*/
1207	Log(("NAT: %R[natsock] err(%d:%s) s(in:%d,out:%d)happens on read I/O, "
1208	"other side close connection \n", so, err, strerror(err), inq, outq));
1209	CONTINUE(tcp);
1210	}
1211	goto tcp_input_close;
1212	}
1213	if ( !UNIX_CHECK_FD_SET(so, NetworkEvents, readfds)
1214	&& !UNIX_CHECK_FD_SET(so, NetworkEvents, writefds)
1215	&& !UNIX_CHECK_FD_SET(so, NetworkEvents, xfds))
1216	{
1217	Log(("NAT: system expires the socket %R[natsock] err(%d:%s) s(in:%d,out:%d) happens on non-I/O. ",
1218	so, err, strerror(err), inq, outq));
1219	goto tcp_input_close;
1220	}
1221	Log(("NAT: %R[natsock] we've met(%d:%s) s(in:%d, out:%d) unhandled combination hup (%d) "
1222	"rederr(%d) on (r:%d, w:%d, x:%d)\n",
1223	so, err, strerror(err),
1224	inq, outq,
1225	UNIX_CHECK_FD_SET(so, ign, rdhup),
1226	UNIX_CHECK_FD_SET(so, ign, rderr),
1227	UNIX_CHECK_FD_SET(so, ign, readfds),
1228	UNIX_CHECK_FD_SET(so, ign, writefds),
1229	UNIX_CHECK_FD_SET(so, ign, xfds)));
1230	/*
1231	* Give OS's TCP/IP stack a chance to resolve an issue or expire the socket.
1232	*/
1233	CONTINUE(tcp);
1234	tcp_input_close:
1235	so->so_state = SS_NOFDREF; /cause connection valid tcp connection termination and socket closing /
1236	TCP_INPUT(pData, (struct mbuf *)NULL, sizeof(struct ip), so);
1237	CONTINUE(tcp);
1238	}
1239	#endif
1240	LOOP_LABEL(tcp, so, so_next);
1241	}
1242
1243	/*
1244	* Now UDP sockets.
1245	* Incoming packets are sent straight away, they're not buffered.
1246	* Incoming UDP data isn't buffered either.
1247	*/
1248	QSOCKET_FOREACH(so, so_next, udp)
1249	/* { */
1250	#ifdef VBOX_WITH_SLIRP_MT
1251	if ( so->so_state & SS_NOFDREF
1252	&& so->so_deleted == 1)
1253	{
1254	struct socket son, sop = NULL;
1255	QSOCKET_LOCK(udb);
1256	if (so->so_next != NULL)
1257	{
1258	if (so->so_next != &udb)
1259	SOCKET_LOCK(so->so_next);
1260	son = so->so_next;
1261	}
1262	if ( so->so_prev != &udb
1263	&& so->so_prev != NULL)
1264	{
1265	SOCKET_LOCK(so->so_prev);
1266	sop = so->so_prev;
1267	}
1268	QSOCKET_UNLOCK(udb);
1269	remque(pData, so);
1270	NSOCK_DEC();
1271	SOCKET_UNLOCK(so);
1272	SOCKET_LOCK_DESTROY(so);
1273	RTMemFree(so);
1274	so_next = son;
1275	if (sop != NULL)
1276	SOCKET_UNLOCK(sop);
1277	CONTINUE_NO_UNLOCK(udp);
1278	}
1279	#endif
1280	POLL_UDP_EVENTS(rc, error, so, &NetworkEvents);
1281
1282	LOG_NAT_SOCK(so, UDP, &NetworkEvents, readfds, writefds, xfds);
1283
1284	if (so->s != -1 && CHECK_FD_SET(so, NetworkEvents, readfds))
1285	{
1286	SORECVFROM(pData, so);
1287	}
1288	LOOP_LABEL(udp, so, so_next);
1289	}
1290
1291	done:
1292	#ifndef VBOX_WITH_SLIRP_MT
1293	/*
1294	* See if we can start outputting
1295	*/
1296	if (if_queued && link_up)
1297	if_start(pData);
1298	#endif
1299
1300	STAM_PROFILE_STOP(&pData->StatPoll, a);
1301	}
1302
1303	#ifndef VBOX_WITHOUT_SLIRP_CLIENT_ETHER
1304	#define ETH_ALEN 6
1305	#define ETH_HLEN 14
1306
1307	#define ARPOP_REQUEST 1 /* ARP request */
1308	#define ARPOP_REPLY 2 /* ARP reply */
1309
1310	struct ethhdr
1311	{
1312	unsigned char h_dest[ETH_ALEN]; /* destination eth addr */
1313	unsigned char h_source[ETH_ALEN]; /* source ether addr */
1314	unsigned short h_proto; /* packet type ID field */
1315	};
1316	AssertCompileSize(struct ethhdr, 14);
1317	#endif
1318
1319	struct arphdr
1320	{
1321	unsigned short ar_hrd; /* format of hardware address */
1322	unsigned short ar_pro; /* format of protocol address */
1323	unsigned char ar_hln; /* length of hardware address */
1324	unsigned char ar_pln; /* length of protocol address */
1325	unsigned short ar_op; /* ARP opcode (command) */
1326
1327	/*
1328	* Ethernet looks like this : This bit is variable sized however...
1329	*/
1330	unsigned char ar_sha[ETH_ALEN]; /* sender hardware address */
1331	unsigned char ar_sip[4]; /* sender IP address */
1332	unsigned char ar_tha[ETH_ALEN]; /* target hardware address */
1333	unsigned char ar_tip[4]; /* target IP address */
1334	};
1335	AssertCompileSize(struct arphdr, 28);
1336
1337	static void arp_input(PNATState pData, struct mbuf *m)
1338	{
1339	struct ethhdr *eh;
1340	struct ethhdr *reh;
1341	struct arphdr *ah;
1342	struct arphdr *rah;
1343	int ar_op;
1344	struct ex_list *ex_ptr;
1345	uint32_t htip;
1346	uint32_t tip;
1347	struct mbuf *mr;
1348	eh = mtod(m, struct ethhdr *);
1349	ah = (struct arphdr *)&eh[1];
1350	htip = ntohl((uint32_t)ah->ar_tip);
1351	tip = (uint32_t)ah->ar_tip;
1352
1353	mr = m_get(pData);
1354	#ifdef VBOX_WITHOUT_SLIRP_CLIENT_ETHER
1355	reh = mtod(mr, struct ethhdr *);
1356	memcpy(reh->h_source, eh->h_source, ETH_ALEN); /* XXX: if_encap will swap src and dst*/
1357	Log4(("NAT: arp:%R[ether]->%R[ether]\n",
1358	reh->h_source, reh->h_dest));
1359	Log4(("NAT: arp: %R[IP4]\n", &tip));
1360	#endif
1361	mr->m_data += if_maxlinkhdr;
1362	mr->m_len = sizeof(struct arphdr);
1363	rah = mtod(mr, struct arphdr *);
1364
1365	ar_op = ntohs(ah->ar_op);
1366	switch(ar_op)
1367	{
1368	case ARPOP_REQUEST:
1369	#ifdef VBOX_WITH_NAT_SERVICE
1370	if (tip == special_addr.s_addr) goto arp_ok;
1371	#endif
1372	if ((htip & pData->netmask) == ntohl(special_addr.s_addr))
1373	{
1374	if ( CTL_CHECK(htip, CTL_DNS)
1375	\|\| CTL_CHECK(htip, CTL_ALIAS)
1376	\|\| CTL_CHECK(htip, CTL_TFTP))
1377	goto arp_ok;
1378	for (ex_ptr = exec_list; ex_ptr; ex_ptr = ex_ptr->ex_next)
1379	{
1380	if ((htip & ~pData->netmask) == ex_ptr->ex_addr)
1381	{
1382	goto arp_ok;
1383	}
1384	}
1385	return;
1386	arp_ok:
1387	rah->ar_hrd = htons(1);
1388	rah->ar_pro = htons(ETH_P_IP);
1389	rah->ar_hln = ETH_ALEN;
1390	rah->ar_pln = 4;
1391	rah->ar_op = htons(ARPOP_REPLY);
1392	memcpy(rah->ar_sha, special_ethaddr, ETH_ALEN);
1393
1394	switch (htip & ~pData->netmask)
1395	{
1396	case CTL_DNS:
1397	case CTL_ALIAS:
1398	rah->ar_sha[5] = (uint8_t)(htip & ~pData->netmask);
1399	break;
1400	default:;
1401	}
1402
1403	memcpy(rah->ar_sip, ah->ar_tip, 4);
1404	memcpy(rah->ar_tha, ah->ar_sha, ETH_ALEN);
1405	memcpy(rah->ar_tip, ah->ar_sip, 4);
1406	if_encap(pData, ETH_P_ARP, mr);
1407	m_free(pData, m);
1408	}
1409	break;
1410	default:
1411	break;
1412	}
1413	}
1414
1415	void slirp_input(PNATState pData, const uint8_t *pkt, int pkt_len)
1416	{
1417	struct mbuf *m;
1418	int proto;
1419	static bool fWarnedIpv6;
1420	struct ethhdr eh = (struct ethhdr)pkt;
1421
1422	Log2(("NAT: slirp_input %d\n", pkt_len));
1423	if (pkt_len < ETH_HLEN)
1424	{
1425	LogRel(("NAT: packet having size %d has been ingnored\n", pkt_len));
1426	return;
1427	}
1428	Log4(("NAT: in:%R[ether]->%R[ether]\n", &eh->h_source, &eh->h_dest));
1429	#ifdef VBOX_WITHOUT_SLIRP_CLIENT_ETHER
1430	if (memcmp(eh->h_source, special_ethaddr, ETH_ALEN) == 0)
1431	{
1432	/* @todo vasily: add ether logging routine in debug.c */
1433	Log(("NAT: packet was addressed to other MAC\n"));
1434	RTMemFree((void *)pkt);
1435	return;
1436	}
1437	#endif
1438
1439	m = m_get(pData);
1440	if (!m)
1441	{
1442	LogRel(("NAT: can't allocate new mbuf\n"));
1443	return;
1444	}
1445
1446	/* Note: we add to align the IP header */
1447
1448	if (M_FREEROOM(m) < pkt_len)
1449	m_inc(m, pkt_len);
1450
1451	m->m_len = pkt_len ;
1452	memcpy(m->m_data, pkt, pkt_len);
1453
1454	proto = ntohs((uint16_t )(pkt + 12));
1455	switch(proto)
1456	{
1457	case ETH_P_ARP:
1458	arp_input(pData, m);
1459	break;
1460	case ETH_P_IP:
1461	/* Update time. Important if the network is very quiet, as otherwise
1462	* the first outgoing connection gets an incorrect timestamp. */
1463	updtime(pData);
1464	m_adj(m, ETH_HLEN);
1465	ip_input(pData, m);
1466	break;
1467	case ETH_P_IPV6:
1468	m_free(pData, m);
1469	if (!fWarnedIpv6)
1470	{
1471	LogRel(("NAT: IPv6 not supported\n"));
1472	fWarnedIpv6 = true;
1473	}
1474	break;
1475	default:
1476	Log(("NAT: Unsupported protocol %x\n", proto));
1477	m_free(pData, m);
1478	break;
1479	}
1480	RTMemFree((void *)pkt);
1481	}
1482
1483	/* output the IP packet to the ethernet device */
1484	void if_encap(PNATState pData, uint16_t eth_proto, struct mbuf *m)
1485	{
1486	struct ethhdr *eh;
1487	uint8_t *buf = RTMemAlloc(1600);
1488
1489	m->m_data -= if_maxlinkhdr;
1490	m->m_len += ETH_HLEN;
1491	eh = mtod(m, struct ethhdr *);
1492
1493	if(MBUF_HEAD(m) != m->m_data)
1494	{
1495	LogRel(("NAT: ethernet detects corruption of the packet"));
1496	AssertMsgFailed(("!!Ethernet frame corrupted!!"));
1497	}
1498
1499	#ifndef VBOX_WITHOUT_SLIRP_CLIENT_ETHER
1500	memcpy(eh->h_dest, client_ethaddr, ETH_ALEN);
1501	memcpy(eh->h_source, special_ethaddr, ETH_ALEN - 1);
1502	/* XXX: not correct */
1503	eh->h_source[5] = CTL_ALIAS;
1504	#else
1505	if (memcmp(eh->h_source, special_ethaddr, ETH_ALEN) != 0)
1506	{
1507	memcpy(eh->h_dest, eh->h_source, ETH_ALEN);
1508	memcpy(eh->h_source, special_ethaddr, ETH_ALEN);
1509	Assert(memcmp(eh->h_dest, special_ethaddr, ETH_ALEN) != 0);
1510	if (memcmp(eh->h_dest, zerro_ethaddr, ETH_ALEN) == 0)
1511	{
1512	/* don't do anything */
1513	goto done;
1514	}
1515	}
1516	#endif
1517	eh->h_proto = htons(eth_proto);
1518	#if 0
1519	slirp_output(pData->pvUser, m, mtod(m, uint8_t *), m->m_len);
1520	#else
1521	memcpy(buf, mtod(m, uint8_t *), m->m_len);
1522	slirp_output(pData->pvUser, NULL, buf, m->m_len);
1523	done:
1524	m_free(pData, m);
1525	#endif
1526	}
1527
1528	int slirp_redir(PNATState pData, int is_udp, struct in_addr host_addr, int host_port,
1529	struct in_addr guest_addr, int guest_port)
1530	{
1531	struct socket *so;
1532	#ifdef VBOX_WITH_SLIRP_ALIAS
1533	struct alias_link *link;
1534	struct libalias *lib;
1535	int flags;
1536	struct sockaddr sa;
1537	struct sockaddr_in *psin;
1538	socklen_t socketlen;
1539	struct in_addr alias;
1540	int rc;
1541	#endif
1542	Log2(("NAT: set redirect %s hp:%d gp:%d\n", (is_udp?"UDP":"TCP"), host_port, guest_port));
1543	if (is_udp)
1544	{
1545	so = udp_listen(pData, host_addr.s_addr, htons(host_port), guest_addr.s_addr,
1546	htons(guest_port), 0);
1547	}
1548	else
1549	{
1550	so = solisten(pData, host_addr.s_addr, htons(host_port), guest_addr.s_addr,
1551	htons(guest_port), 0);
1552	}
1553	if (so == NULL)
1554	{
1555	return -1;
1556	}
1557	#ifndef VBOX_WITH_SLIRP_ALIAS
1558	Log2(("NAT: redirecting socket %R[natsock]\n", so));
1559	return (so != NULL ? 0 : -1);
1560	#else
1561
1562	psin = (struct sockaddr_in *)&sa;
1563	psin->sin_family = AF_INET;
1564	psin->sin_port = 0;
1565	psin->sin_addr.s_addr = INADDR_ANY;
1566	socketlen = sizeof(struct sockaddr);
1567
1568	rc = getsockname(so->s, &sa, &socketlen);
1569	if (rc < 0 \|\| sa.sa_family != AF_INET)
1570	{
1571	Log(("NAT: can't get socket's name\n"));
1572	return 1;
1573	}
1574
1575	psin = (struct sockaddr_in *)&sa;
1576
1577	#if 1
1578	lib = LibAliasInit(pData, NULL);
1579	flags = LibAliasSetMode(lib, 0, 0);
1580	flags \|= PKT_ALIAS_LOG; /* set logging */
1581	flags \|= PKT_ALIAS_REVERSE; /* set logging */
1582	flags = LibAliasSetMode(lib, flags, ~0);
1583	#else
1584	lib = LIST_FIRST(&instancehead);
1585	#endif
1586
1587	alias.s_addr = htonl(ntohl(guest_addr.s_addr) \| CTL_ALIAS);
1588	link = LibAliasRedirectPort(lib, psin->sin_addr, htons(host_port),
1589	alias, htons(guest_port),
1590	special_addr, -1, /* not very clear for now*/
1591	(is_udp ? IPPROTO_UDP : IPPROTO_TCP));
1592	if (link == NULL)
1593	{
1594	Log(("NAT: can't create redirect\n"));
1595	return 1;
1596	}
1597	so->so_la = lib;
1598
1599	return 0;
1600	#endif
1601	}
1602
1603	int slirp_add_exec(PNATState pData, int do_pty, const char *args, int addr_low_byte,
1604	int guest_port)
1605	{
1606	return add_exec(&exec_list, do_pty, (char *)args,
1607	addr_low_byte, htons(guest_port));
1608	}
1609
1610	void slirp_set_ethaddr(PNATState pData, const uint8_t *ethaddr)
1611	{
1612	#ifndef VBOX_WITHOUT_SLIRP_CLIENT_ETHER
1613	memcpy(client_ethaddr, ethaddr, ETH_ALEN);
1614	#endif
1615	}
1616
1617	#if defined(RT_OS_WINDOWS)
1618	HANDLE *slirp_get_events(PNATState pData)
1619	{
1620	return pData->phEvents;
1621	}
1622	void slirp_register_external_event(PNATState pData, HANDLE hEvent, int index)
1623	{
1624	pData->phEvents[index] = hEvent;
1625	}
1626	#endif
1627
1628	unsigned int slirp_get_timeout_ms(PNATState pData)
1629	{
1630	if (link_up)
1631	{
1632	if (time_fasttimo)
1633	return 2;
1634	if (do_slowtimo)
1635	return 500; /* see PR_SLOWHZ */
1636	}
1637	return 0;
1638	}
1639
1640	#ifndef RT_OS_WINDOWS
1641	int slirp_get_nsock(PNATState pData)
1642	{
1643	return pData->nsock;
1644	}
1645	#endif
1646
1647	/*
1648	* this function called from NAT thread
1649	*/
1650	void slirp_post_sent(PNATState pData, void *pvArg)
1651	{
1652	struct socket *so = 0;
1653	struct tcpcb *tp = 0;
1654	struct mbuf m = (struct mbuf )pvArg;
1655	m_free(pData, m);
1656	}
1657	#ifdef VBOX_WITH_SLIRP_MT
1658	void slirp_process_queue(PNATState pData)
1659	{
1660	RTReqProcess(pData->pReqQueue, RT_INDEFINITE_WAIT);
1661	}
1662	void *slirp_get_queue(PNATState pData)
1663	{
1664	return pData->pReqQueue;
1665	}
1666	#endif
1667
1668	uint16_t slirp_get_service(int proto, uint16_t dport, uint16_t sport)
1669	{
1670	uint16_t hdport, hsport, service;
1671	hdport = ntohs(dport);
1672	hsport = ntohs(sport);
1673	Log2(("proto: %d, dport: %d sport: %d\n", proto, hdport, hsport));
1674	service = 0;
1675	#if 0
1676	/* Always return 0 here */
1677	switch (hdport)
1678	{
1679	case 500:
1680	if (hsport != 500) /* vpnc by default try operate in src:500/dst:500 mode*/
1681	/* Not sure why this make Cisco VPN client's connection more stable,
1682	* at least on some servers
1683	*/
1684	service = sport;
1685	break;
1686	}
1687	#endif
1688	Log2(("service : %d\n", service));
1689	return htons(service);
1690	}
1691
1692	void slirp_set_dhcp_TFTP_prefix(PNATState pData, const char *tftpPrefix)
1693	{
1694	Log2(("tftp_prefix:%s\n", tftpPrefix));
1695	tftp_prefix = tftpPrefix;
1696	}
1697
1698	void slirp_set_dhcp_TFTP_bootfile(PNATState pData, const char *bootFile)
1699	{
1700	Log2(("bootFile:%s\n", bootFile));
1701	bootp_filename = bootFile;
1702	}
1703
1704	void slirp_set_dhcp_next_server(PNATState pData, const char *next_server)
1705	{
1706	Log2(("next_server:%s\n", next_server));
1707	if (next_server == NULL)
1708	pData->tftp_server.s_addr = htonl(ntohl(special_addr.s_addr) \| CTL_TFTP);
1709	else
1710	inet_aton(next_server, &pData->tftp_server);
1711	}
1712
1713	int slirp_set_binding_address(PNATState pData, char *addr)
1714	{
1715	if (addr == NULL \|\| (inet_aton(addr, &pData->bindIP) == 0))
1716	{
1717	pData->bindIP.s_addr = INADDR_ANY;
1718	return 1;
1719	}
1720	return 0;
1721	}
1722
1723	void slirp_set_dhcp_dns_proxy(PNATState pData, bool fDNSProxy)
1724	{
1725	Log2(("NAT: DNS proxy switched %s\n", (fDNSProxy ? "on" : "off")));
1726	pData->use_dns_proxy = fDNSProxy;
1727	}
1728
1729	#define CHECK_ARG(name, val, lim_min, lim_max) \
1730	do { \
1731	if ((val) < (lim_min) \|\| (val) > (lim_max)) \
1732	{ \
1733	LogRel(("NAT: (" #name ":%d) has been ignored, " \
1734	"because out of range (%d, %d)\n", (val), (lim_min), (lim_max))); \
1735	return; \
1736	} \
1737	else \
1738	{ \
1739	LogRel(("NAT: (" #name ":%d)\n", (val))); \
1740	} \
1741	} while (0)
1742
1743	/* don't allow user set less 8kB and more than 1M values */
1744	#define _8K_1M_CHECK_ARG(name, val) CHECK_ARG(name, (val), 8, 1024)
1745	void slirp_set_rcvbuf(PNATState pData, int kilobytes)
1746	{
1747	_8K_1M_CHECK_ARG("SOCKET_RCVBUF", kilobytes);
1748	pData->socket_rcv = kilobytes;
1749	}
1750	void slirp_set_sndbuf(PNATState pData, int kilobytes)
1751	{
1752	_8K_1M_CHECK_ARG("SOCKET_SNDBUF", kilobytes);
1753	pData->socket_snd = kilobytes * _1K;
1754	}
1755	void slirp_set_tcp_rcvspace(PNATState pData, int kilobytes)
1756	{
1757	_8K_1M_CHECK_ARG("TCP_RCVSPACE", kilobytes);
1758	tcp_rcvspace = kilobytes * _1K;
1759	}
1760	void slirp_set_tcp_sndspace(PNATState pData, int kilobytes)
1761	{
1762	_8K_1M_CHECK_ARG("TCP_SNDSPACE", kilobytes);
1763	tcp_sndspace = kilobytes * _1K;
1764	}
1765

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

source: vbox/trunk/src/VBox/Devices/Network/slirp/slirp.c@ 21004

Download in other formats: