slirp.c@ 20703

Last change on this file since 20703 was 20574, checked in by vboxsync, 16 years ago
NAT: typo (missed 'N')
Property svn:eol-style set to `native`
File size: 59.2 KB

Line
1	#include "slirp.h"
2	#ifdef RT_OS_OS2
3	# include <paths.h>
4	#endif
5	#ifdef VBOX_WITH_SLIRP_DNS_PROXY
6	#include "dnsproxy/dnsproxy.h"
7	#endif
8
9	#include <VBox/err.h>
10	#include <VBox/pdmdrv.h>
11	#include <iprt/assert.h>
12	#ifndef RT_OS_WINDOWS
13	# include <sys/ioctl.h>
14	# include <poll.h>
15	#else
16	# include <Winnls.h>
17	# define _WINSOCK2API_
18	# include <IPHlpApi.h>
19	#endif
20	#ifdef VBOX_WITH_SLIRP_ALIAS
21	# include<alias.h>
22	#endif
23
24	#if !defined(RT_OS_WINDOWS)
25
26	# define DO_ENGAGE_EVENT1(so, fdset, label) \
27	do { \
28	if( so->so_poll_index != -1 \
29	&& so->s == polls[so->so_poll_index].fd) { \
30	polls[so->so_poll_index].events \|= N_(fdset ## _poll); \
31	break; /* out of this loop */ \
32	} \
33	AssertRelease(poll_index < (nfds)); \
34	AssertRelease(poll_index >= 0 && poll_index < (nfds)); \
35	polls[poll_index].fd = (so)->s; \
36	(so)->so_poll_index = poll_index; \
37	polls[poll_index].events = N_(fdset ## _poll); \
38	polls[poll_index].revents = 0; \
39	poll_index++; \
40	} while(0)
41
42
43	# define DO_ENGAGE_EVENT2(so, fdset1, fdset2, label) \
44	do { \
45	if( so->so_poll_index != -1 \
46	&& so->s == polls[so->so_poll_index].fd) { \
47	polls[so->so_poll_index].events \|= \
48	N_(fdset1 ## _poll) \| N_(fdset1 ## _poll); \
49	break; /* out of this loop */ \
50	} \
51	AssertRelease(poll_index < (nfds)); \
52	polls[poll_index].fd = (so)->s; \
53	(so)->so_poll_index = poll_index; \
54	polls[poll_index].events = \
55	N_(fdset1 ## _poll) \| N_(fdset1 ## _poll); \
56	poll_index++; \
57	} while(0)
58
59	# define DO_POLL_EVENTS(rc, error, so, events, label) do {} while (0)
60
61	# define DO_CHECK_FD_SET(so, events, fdset) ( ((so)->so_poll_index != -1) \
62	&& ((so)->so_poll_index <= ndfs) \
63	&& ((so)->s == polls[so->so_poll_index].fd) \
64	&& (polls[(so)->so_poll_index].revents & N_(fdset ## _poll)))
65	# define DO_UNIX_CHECK_FD_SET(so, events, fdset ) DO_CHECK_FD_SET((so), (events), fdset) /specific for Unix API /
66	# define DO_WIN_CHECK_FD_SET(so, events, fdset ) 0 /* specific for Windows Winsock API */
67
68	# ifndef RT_OS_WINDOWS
69
70	# ifndef RT_OS_LINUX
71	# define readfds_poll (POLLRDNORM)
72	# define writefds_poll (POLLWRNORM)
73	# define xfds_poll (POLLRDBAND\|POLLWRBAND\|POLLPRI)
74	# else
75	# define readfds_poll (POLLIN)
76	# define writefds_poll (POLLOUT)
77	# define xfds_poll (POLLPRI)
78	# endif
79	# define rderr_poll (POLLERR)
80	# define rdhup_poll (POLLHUP)
81	# define nval_poll (POLLNVAL)
82
83	# define ICMP_ENGAGE_EVENT(so, fdset) \
84	do { \
85	if (pData->icmp_socket.s != -1) \
86	DO_ENGAGE_EVENT1((so), fdset, ICMP); \
87	} while (0)
88	# else /* !RT_OS_WINDOWS */
89	# define DO_WIN_CHECK_FD_SET(so, events, fdset ) DO_CHECK_FD_SET((so), (events), fdset)
90	# define ICMP_ENGAGE_EVENT(so, fdset) do {} while(0)
91	#endif /* RT_OS_WINDOWS */
92
93	#else /* defined(RT_OS_WINDOWS) */
94
95	/*
96	* On Windows, we will be notified by IcmpSendEcho2() when the response arrives.
97	* So no call to WSAEventSelect necessary.
98	*/
99	# define ICMP_ENGAGE_EVENT(so, fdset) do {} while(0)
100
101	# define DO_ENGAGE_EVENT1(so, fdset1, label) \
102	do { \
103	rc = WSAEventSelect((so)->s, VBOX_SOCKET_EVENT, FD_ALL_EVENTS); \
104	if (rc == SOCKET_ERROR) \
105	{ \
106	/* This should not happen */ \
107	error = WSAGetLastError(); \
108	LogRel(("WSAEventSelect (" #label ") error %d (so=%x, socket=%s, event=%x)\n", \
109	error, (so), (so)->s, VBOX_SOCKET_EVENT)); \
110	} \
111	} while(0); \
112	CONTINUE(label)
113
114	# define DO_ENGAGE_EVENT2(so, fdset1, fdset2, label) \
115	DO_ENGAGE_EVENT1((so), (fdset1), label)
116
117	# define DO_POLL_EVENTS(rc, error, so, events, label) \
118	(rc) = WSAEnumNetworkEvents((so)->s, VBOX_SOCKET_EVENT, (events)); \
119	if ((rc) == SOCKET_ERROR) \
120	{ \
121	(error) = WSAGetLastError(); \
122	LogRel(("WSAEnumNetworkEvents " #label " error %d\n", (error))); \
123	CONTINUE(label); \
124	}
125
126	# define acceptds_win FD_ACCEPT
127	# define acceptds_win_bit FD_ACCEPT_BIT
128
129	# define readfds_win FD_READ
130	# define readfds_win_bit FD_READ_BIT
131
132	# define writefds_win FD_WRITE
133	# define writefds_win_bit FD_WRITE_BIT
134
135	# define xfds_win FD_OOB
136	# define xfds_win_bit FD_OOB_BIT
137
138	# define DO_CHECK_FD_SET(so, events, fdset) \
139	(((events).lNetworkEvents & fdset ## _win) && ((events).iErrorCode[fdset ## _win_bit] == 0))
140
141	# define DO_WIN_CHECK_FD_SET(so, events, fdset ) DO_CHECK_FD_SET((so), (events), fdset)
142	# define DO_UNIX_CHECK_FD_SET(so, events, fdset ) 1 /specific for Unix API /
143
144	#endif /* defined(RT_OS_WINDOWS) */
145
146	#define TCP_ENGAGE_EVENT1(so, fdset) \
147	DO_ENGAGE_EVENT1((so), fdset, tcp)
148
149	#define TCP_ENGAGE_EVENT2(so, fdset1, fdset2) \
150	DO_ENGAGE_EVENT2((so), fdset1, fdset2, tcp)
151
152	#define UDP_ENGAGE_EVENT(so, fdset) \
153	DO_ENGAGE_EVENT1((so), fdset, udp)
154
155	#define POLL_TCP_EVENTS(rc, error, so, events) \
156	DO_POLL_EVENTS((rc), (error), (so), (events), tcp)
157
158	#define POLL_UDP_EVENTS(rc, error, so, events) \
159	DO_POLL_EVENTS((rc), (error), (so), (events), udp)
160
161	#define CHECK_FD_SET(so, events, set) \
162	(DO_CHECK_FD_SET((so), (events), set))
163
164	#define WIN_CHECK_FD_SET(so, events, set) \
165	(DO_WIN_CHECK_FD_SET((so), (events), set))
166	#define UNIX_CHECK_FD_SET(so, events, set) \
167	(DO_UNIX_CHECK_FD_SET(so, events, set))
168
169	/*
170	* Loging macros
171	*/
172	#if VBOX_WITH_DEBUG_NAT_SOCKETS
173	# if defined(RT_OS_WINDOWS)
174	# define DO_LOG_NAT_SOCK(so, proto, winevent, r_fdset, w_fdset, x_fdset) \
175	do { \
176	LogRel((" " #proto " %R[natsock] %R[natwinnetevents]\n", (so), (winevent))); \
177	} while (0)
178	# else /* RT_OS_WINDOWS */
179	# define DO_LOG_NAT_SOCK(so, proto, winevent, r_fdset, w_fdset, x_fdset) \
180	do { \
181	LogRel((" " #proto " %R[natsock] %s %s %s er: %s, %s, %s\n", (so), \
182	CHECK_FD_SET(so, ign ,r_fdset) ? "READ":"", \
183	CHECK_FD_SET(so, ign, w_fdset) ? "WRITE":"", \
184	CHECK_FD_SET(so, ign, x_fdset) ? "OOB":"", \
185	CHECK_FD_SET(so, ign, rderr) ? "RDERR":"", \
186	CHECK_FD_SET(so, ign, rdhup) ? "RDHUP":"", \
187	CHECK_FD_SET(so, ign, nval) ? "RDNVAL":"")); \
188	} while (0)
189	# endif /* !RT_OS_WINDOWS */
190	#else /* VBOX_WITH_DEBUG_NAT_SOCKETS */
191	# define DO_LOG_NAT_SOCK(so, proto, winevent, r_fdset, w_fdset, x_fdset) do {} while (0)
192	#endif /* !VBOX_WITH_DEBUG_NAT_SOCKETS */
193
194	#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)
195
196	static const uint8_t special_ethaddr[6] =
197	{
198	0x52, 0x54, 0x00, 0x12, 0x35, 0x00
199	};
200
201	static const uint8_t broadcast_ethaddr[6] =
202	{
203	0xff, 0xff, 0xff, 0xff, 0xff, 0xff
204	};
205
206	const uint8_t zerro_ethaddr[6] =
207	{
208	0x0, 0x0, 0x0, 0x0, 0x0, 0x0
209	};
210
211	#ifdef RT_OS_WINDOWS
212	# ifndef VBOX_WITH_MULTI_DNS
213	static int get_dns_addr_domain(PNATState pData, bool fVerbose,
214	struct in_addr *pdns_addr,
215	const char **ppszDomain)
216	{
217	int rc = 0;
218	FIXED_INFO *FixedInfo = NULL;
219	ULONG BufLen;
220	DWORD ret;
221	IP_ADDR_STRING *pIPAddr;
222	struct in_addr tmp_addr;
223
224	FixedInfo = (FIXED_INFO *)GlobalAlloc(GPTR, sizeof(FIXED_INFO));
225	BufLen = sizeof(FIXED_INFO);
226
227	/** @todo: this API returns all DNS servers, no matter whether the
228	* corresponding network adapter is disabled or not. Maybe replace
229	* this by GetAdapterAddresses(), which is XP/Vista only though. */
230	if (ERROR_BUFFER_OVERFLOW == GetNetworkParams(FixedInfo, &BufLen))
231	{
232	if (FixedInfo)
233	{
234	GlobalFree(FixedInfo);
235	FixedInfo = NULL;
236	}
237	FixedInfo = GlobalAlloc(GPTR, BufLen);
238	}
239
240	if ((ret = GetNetworkParams(FixedInfo, &BufLen)) != ERROR_SUCCESS)
241	{
242	Log(("GetNetworkParams failed. ret = %08x\n", (u_int)ret ));
243	if (FixedInfo)
244	{
245	GlobalFree(FixedInfo);
246	FixedInfo = NULL;
247	}
248	rc = -1;
249	goto get_dns_prefix;
250	}
251
252	pIPAddr = &(FixedInfo->DnsServerList);
253	inet_aton(pIPAddr->IpAddress.String, &tmp_addr);
254	Log(("nat: DNS Servers:\n"));
255	if (fVerbose \|\| pdns_addr->s_addr != tmp_addr.s_addr)
256	LogRel(("NAT: DNS address: %s\n", pIPAddr->IpAddress.String));
257	*pdns_addr = tmp_addr;
258
259	pIPAddr = FixedInfo -> DnsServerList.Next;
260	while (pIPAddr)
261	{
262	if (fVerbose)
263	LogRel(("NAT: ignored DNS address: %s\n", pIPAddr ->IpAddress.String));
264	pIPAddr = pIPAddr ->Next;
265	}
266	if (FixedInfo)
267	{
268	GlobalFree(FixedInfo);
269	FixedInfo = NULL;
270	}
271
272	get_dns_prefix:
273	if (ppszDomain)
274	{
275	OSVERSIONINFO ver;
276	char szDnsDomain[256];
277	DWORD dwSize = sizeof(szDnsDomain);
278
279	*ppszDomain = NULL;
280	GetVersionEx(&ver);
281	if (ver.dwMajorVersion >= 5)
282	{
283	/* GetComputerNameEx exists in Windows versions starting with 2000. */
284	if (GetComputerNameEx(ComputerNameDnsDomain, szDnsDomain, &dwSize))
285	{
286	if (szDnsDomain[0])
287	{
288	/* Just non-empty strings are valid. */
289	*ppszDomain = RTStrDup(szDnsDomain);
290	if (pData->fPassDomain)
291	{
292	if (fVerbose)
293	LogRel(("NAT: passing domain name %s\n", szDnsDomain));
294	}
295	else
296	Log(("nat: ignoring domain %s\n", szDnsDomain));
297	}
298	}
299	else
300	Log(("nat: GetComputerNameEx failed (%d)\n", GetLastError()));
301	}
302	}
303	return rc;
304	}
305	# else /* !VBOX_WITH_MULTI_DNS */
306	static int get_dns_addr_domain(PNATState pData, bool fVerbose,
307	struct in_addr *pdns_addr,
308	const char **ppszDomain)
309	{
310	/* Get amount of memory required for operation */
311	ULONG flags = GAA_FLAG_INCLUDE_PREFIX; /GAA_FLAG_INCLUDE_ALL_INTERFACES;/ /* all interfaces registered in NDIS */
312	PIP_ADAPTER_ADDRESSES addresses = NULL;
313	PIP_ADAPTER_ADDRESSES addr = NULL;
314	PIP_ADAPTER_DNS_SERVER_ADDRESS dns = NULL;
315	ULONG size = 0;
316	int wlen = 0;
317	char *suffix;
318	struct dns_entry *da = NULL;
319	struct dns_domain_entry *dd = NULL;
320	ULONG ret = ERROR_SUCCESS;
321
322	/* @todo add SKIPing flags to get only required information */
323
324	ret = pData->pfGetAdaptersAddresses(AF_INET, 0, NULL /* reserved */, addresses, &size);
325	if (ret != ERROR_BUFFER_OVERFLOW)
326	{
327	LogRel(("NAT: error %lu occured on capacity detection operation\n", ret));
328	return -1;
329	}
330
331	if (size == 0)
332	{
333	LogRel(("NAT: Win socket API returns non capacity\n"));
334	return -1;
335	}
336
337	addresses = RTMemAllocZ(size);
338	if (addresses == NULL)
339	{
340	LogRel(("NAT: No memory available \n"));
341	return -1;
342	}
343
344	ret = pData->pfGetAdaptersAddresses(AF_INET, 0, NULL /* reserved */, addresses, &size);
345	if (ret != ERROR_SUCCESS)
346	{
347	LogRel(("NAT: error %lu occured on fetching adapters info\n", ret));
348	RTMemFree(addresses);
349	return -1;
350	}
351	addr = addresses;
352	while(addr != NULL)
353	{
354	int found;
355	if (addr->OperStatus != IfOperStatusUp)
356	goto next;
357	dns = addr->FirstDnsServerAddress;
358	while (dns != NULL)
359	{
360	struct sockaddr *saddr = dns->Address.lpSockaddr;
361	if (saddr->sa_family != AF_INET)
362	goto next_dns;
363	/* add dns server to list */
364	da = RTMemAllocZ(sizeof(struct dns_entry));
365	if (da == NULL)
366	{
367	LogRel(("NAT: Can't allocate buffer for DNS entry\n"));
368	RTMemFree(addresses);
369	return VERR_NO_MEMORY;
370	}
371	LogRel(("NAT: adding %R[IP4] to DNS server list\n", &((struct sockaddr_in *)saddr)->sin_addr));
372	if ((((struct sockaddr_in *)saddr)->sin_addr.s_addr & htonl(IN_CLASSA_NET)) == ntohl(INADDR_LOOPBACK & IN_CLASSA_NET)) {
373	da->de_addr.s_addr = htonl(ntohl(special_addr.s_addr) \| CTL_ALIAS);
374	}
375	else
376	{
377	da->de_addr.s_addr = ((struct sockaddr_in *)saddr)->sin_addr.s_addr;
378	}
379	TAILQ_INSERT_HEAD(&pData->dns_list_head, da, de_list);
380
381	if (addr->DnsSuffix == NULL)
382	goto next_dns;
383
384	/uniq/
385	RTUtf16ToUtf8(addr->DnsSuffix, &suffix);
386
387	if (!suffix \|\| strlen(suffix) == 0) {
388	RTStrFree(suffix);
389	goto next_dns;
390	}
391
392	found = 0;
393	LIST_FOREACH(dd, &pData->dns_domain_list_head, dd_list)
394	{
395	if ( dd->dd_pszDomain != NULL
396	&& strcmp(dd->dd_pszDomain, suffix) == 0)
397	{
398	found = 1;
399	RTStrFree(suffix);
400	break;
401	}
402	}
403	if (found == 0)
404	{
405	dd = RTMemAllocZ(sizeof(struct dns_domain_entry));
406	if (dd == NULL)
407	{
408	LogRel(("NAT: not enough memory\n"));
409	RTStrFree(suffix);
410	RTMemFree(addresses);
411	return VERR_NO_MEMORY;
412	}
413	dd->dd_pszDomain = suffix;
414	LogRel(("NAT: adding domain name %s to search list\n", dd->dd_pszDomain));
415	LIST_INSERT_HEAD(&pData->dns_domain_list_head, dd, dd_list);
416	}
417	next_dns:
418	dns = dns->Next;
419	}
420	next:
421	addr = addr->Next;
422	}
423	RTMemFree(addresses);
424	return 0;
425	}
426	# endif /* VBOX_WITH_MULTI_DNS */
427
428	#else /* !RT_OS_WINDOWS */
429
430	static int get_dns_addr_domain(PNATState pData, bool fVerbose,
431	struct in_addr *pdns_addr,
432	const char **ppszDomain)
433	{
434	char buff[512];
435	char buff2[256];
436	FILE *f = NULL;
437	int found = 0;
438	struct in_addr tmp_addr;
439
440	#ifdef RT_OS_OS2
441	/* Try various locations. */
442	char *etc = getenv("ETC");
443	if (etc)
444	{
445	snprintf(buff, sizeof(buff), "%s/RESOLV2", etc);
446	f = fopen(buff, "rt");
447	}
448	if (!f)
449	{
450	snprintf(buff, sizeof(buff), "%s/RESOLV2", _PATH_ETC);
451	f = fopen(buff, "rt");
452	}
453	if (!f)
454	{
455	snprintf(buff, sizeof(buff), "%s/resolv.conf", _PATH_ETC);
456	f = fopen(buff, "rt");
457	}
458	#else
459	#ifndef DEBUG_vvl
460	f = fopen("/etc/resolv.conf", "r");
461	#else
462	char *home = getenv("HOME");
463	snprintf(buff, sizeof(buff), "%s/resolv.conf", home);
464	f = fopen(buff, "r");
465	if (f != NULL)
466	{
467	Log(("NAT: DNS we're using %s\n", buff));
468	}
469	else
470	{
471	f = fopen("/etc/resolv.conf", "r");
472	Log(("NAT: DNS we're using %s\n", buff));
473	}
474	#endif
475	#endif
476	if (!f)
477	return -1;
478
479	if (ppszDomain)
480	*ppszDomain = NULL;
481	Log(("nat: DNS Servers:\n"));
482	while (fgets(buff, 512, f) != NULL)
483	{
484	#ifdef VBOX_WITH_MULTI_DNS
485	struct dns_entry *da = NULL;
486	#endif
487	if (sscanf(buff, "nameserver%*[ \t]%256s", buff2) == 1)
488	{
489	if (!inet_aton(buff2, &tmp_addr))
490	continue;
491	#ifndef VBOX_WITH_MULTI_DNS
492	/* If it's the first one, set it to dns_addr */
493	if (!found)
494	{
495	if (fVerbose \|\| pdns_addr->s_addr != tmp_addr.s_addr)
496	LogRel(("NAT: DNS address: %s\n", buff2));
497	*pdns_addr = tmp_addr;
498	}
499	else
500	{
501	if (fVerbose)
502	LogRel(("NAT: ignored DNS address: %s\n", buff2));
503	}
504	#else
505	/localhost mask /
506	da = RTMemAllocZ(sizeof (struct dns_entry));
507	if (da == NULL)
508	{
509	LogRel(("can't alloc memory for DNS entry\n"));
510	return -1;
511	}
512	/check /
513	da->de_addr.s_addr = tmp_addr.s_addr;
514	if ((da->de_addr.s_addr & htonl(IN_CLASSA_NET)) == ntohl(INADDR_LOOPBACK & IN_CLASSA_NET)) {
515	da->de_addr.s_addr = htonl(ntohl(special_addr.s_addr) \| CTL_ALIAS);
516	}
517	TAILQ_INSERT_HEAD(&pData->dns_list_head, da, de_list);
518	#endif
519	found++;
520	}
521	#ifndef VBOX_WITH_MULTI_DNS
522	if ( ppszDomain
523	&& (!strncmp(buff, "domain", 6) \|\| !strncmp(buff, "search", 6)))
524	{
525	/* Domain name/search list present. Pick first entry */
526	if (*ppszDomain == NULL)
527	{
528	char *tok;
529	char *saveptr;
530	tok = strtok_r(&buff[6], " \t\n", &saveptr);
531	if (tok)
532	{
533	*ppszDomain = RTStrDup(tok);
534	if (pData->fPassDomain)
535	{
536	if (fVerbose)
537	LogRel(("NAT: passing domain name %s\n", tok));
538	}
539	else
540	Log(("nat: ignoring domain %s\n", tok));
541	}
542	}
543	}
544	#else
545	if ((!strncmp(buff, "domain", 6) \|\| !strncmp(buff, "search", 6)))
546	{
547	char *tok;
548	char *saveptr;
549	struct dns_domain_entry *dd = NULL;
550	int found = 0;
551	tok = strtok_r(&buff[6], " \t\n", &saveptr);
552	LIST_FOREACH(dd, &pData->dns_domain_list_head, dd_list)
553	{
554	if( tok != NULL
555	&& strcmp(tok, dd->dd_pszDomain) == 0)
556	{
557	found = 1;
558	break;
559	}
560	}
561	if (tok != NULL && found == 0) {
562	dd = RTMemAllocZ(sizeof(struct dns_domain_entry));
563	if (dd == NULL)
564	{
565	LogRel(("NAT: not enought memory to add domain list\n"));
566	return VERR_NO_MEMORY;
567	}
568	dd->dd_pszDomain = RTStrDup(tok);
569	LogRel(("NAT: adding domain name %s to search list\n", dd->dd_pszDomain));
570	LIST_INSERT_HEAD(&pData->dns_domain_list_head, dd, dd_list);
571	}
572	}
573	#endif
574	}
575	fclose(f);
576	if (!found)
577	return -1;
578	return 0;
579	}
580
581	#endif
582	#ifdef VBOX_WITH_MULTI_DNS
583	static int slirp_init_dns_list(PNATState pData)
584	{
585	TAILQ_INIT(&pData->dns_list_head);
586	LIST_INIT(&pData->dns_domain_list_head);
587	return get_dns_addr_domain(pData, true, NULL, NULL);
588	}
589
590	static void slirp_release_dns_list(PNATState pData)
591	{
592	struct dns_entry *de = NULL;
593	struct dns_domain_entry *dd = NULL;
594	while(!TAILQ_EMPTY(&pData->dns_list_head)) {
595	de = TAILQ_FIRST(&pData->dns_list_head);
596	TAILQ_REMOVE(&pData->dns_list_head, de, de_list);
597	RTMemFree(de);
598	}
599	while(!LIST_EMPTY(&pData->dns_domain_list_head)) {
600	dd = LIST_FIRST(&pData->dns_domain_list_head);
601	LIST_REMOVE(dd, dd_list);
602	if (dd->dd_pszDomain != NULL)
603	RTStrFree(dd->dd_pszDomain);
604	RTMemFree(dd);
605	}
606	}
607	#endif
608
609	int get_dns_addr(PNATState pData, struct in_addr *pdns_addr)
610	{
611	return get_dns_addr_domain(pData, false, pdns_addr, NULL);
612	}
613
614	#ifndef VBOX_WITH_NAT_SERVICE
615	int slirp_init(PNATState ppData, const char pszNetAddr, uint32_t u32Netmask,
616	bool fPassDomain, void *pvUser)
617	#else
618	int slirp_init(PNATState *ppData, uint32_t u32NetAddr, uint32_t u32Netmask,
619	bool fPassDomain, void *pvUser)
620	#endif
621	{
622	int fNATfailed = 0;
623	int rc;
624	PNATState pData = RTMemAllocZ(sizeof(NATState));
625	*ppData = pData;
626	if (!pData)
627	return VERR_NO_MEMORY;
628	if (u32Netmask & 0x1f)
629	/* CTL is x.x.x.15, bootp passes up to 16 IPs (15..31) */
630	return VERR_INVALID_PARAMETER;
631	pData->fPassDomain = fPassDomain;
632	pData->pvUser = pvUser;
633	pData->netmask = u32Netmask;
634
635	/* sockets & TCP defaults */
636	pData->socket_rcv = 64 * _1K;
637	pData->socket_snd = 64 * _1K;
638	tcp_sndspace = 64 * _1K;
639	tcp_rcvspace = 64 * _1K;
640
641	#ifdef RT_OS_WINDOWS
642	{
643	WSADATA Data;
644	WSAStartup(MAKEWORD(2, 0), &Data);
645	}
646	pData->phEvents[VBOX_SOCKET_EVENT_INDEX] = CreateEvent(NULL, FALSE, FALSE, NULL);
647	#endif
648	#ifdef VBOX_WITH_SLIRP_MT
649	QSOCKET_LOCK_CREATE(tcb);
650	QSOCKET_LOCK_CREATE(udb);
651	rc = RTReqCreateQueue(&pData->pReqQueue);
652	AssertReleaseRC(rc);
653	#endif
654
655	link_up = 1;
656
657	debug_init();
658	if_init(pData);
659	ip_init(pData);
660	icmp_init(pData);
661
662	/* Initialise mbufs after setting the MTU */
663	m_init(pData);
664
665	#ifndef VBOX_WITH_NAT_SERVICE
666	inet_aton(pszNetAddr, &special_addr);
667	#else
668	special_addr.s_addr = u32NetAddr;
669	#endif
670	#ifdef VBOX_WITHOUT_SLIRP_CLIENT_ETHER
671	pData->slirp_ethaddr = &special_ethaddr[0];
672	#endif
673	alias_addr.s_addr = special_addr.s_addr \| htonl(CTL_ALIAS);
674	/* @todo: add ability to configure this staff */
675
676	/* set default addresses */
677	inet_aton("127.0.0.1", &loopback_addr);
678	#ifndef VBOX_WITH_MULTI_DNS
679	inet_aton("127.0.0.1", &dns_addr);
680
681	if (get_dns_addr_domain(pData, true, &dns_addr, &pData->pszDomain) < 0)
682	#else
683	if (slirp_init_dns_list(pData) < 0)
684	#endif
685	fNATfailed = 1;
686	#ifdef VBOX_WITH_SLIRP_DNS_PROXY
687	dnsproxy_init(pData);
688	#endif
689
690	getouraddr(pData);
691
692	#ifdef VBOX_WITH_SLIRP_ALIAS
693	{
694	int flags = 0;
695	pData->proxy_alias = LibAliasInit(pData, NULL);
696	if (pData->proxy_alias == NULL)
697	{
698	LogRel(("NAT: LibAlias default rule wasn't initialized\n"));
699	AssertMsgFailed(("NAT: LibAlias default rule wasn't initialized\n"));
700	}
701	flags = LibAliasSetMode(pData->proxy_alias, 0, 0);
702	flags \|= PKT_ALIAS_LOG; /* set logging */
703	flags = LibAliasSetMode(pData->proxy_alias, flags, ~0);
704	LibAliasSetAddress(pData->proxy_alias, special_addr);
705	ftp_alias_load();
706	nbt_alias_load();
707
708	}
709	#endif
710	return fNATfailed ? VINF_NAT_DNS : VINF_SUCCESS;
711	}
712
713	/**
714	* Statistics counters.
715	*/
716	void slirp_register_timers(PNATState pData, PPDMDRVINS pDrvIns)
717	{
718	#ifdef VBOX_WITH_STATISTICS
719	# define COUNTER(name, type, units, dsc) \
720	do{ \
721	PDMDrvHlpSTAMRegisterF(pDrvIns, \
722	&pData->Stat ## name, \
723	type, \
724	STAMVISIBILITY_ALWAYS, \
725	units, \
726	dsc, \
727	"/Drivers/NAT%d/Stat" #name, pDrvIns->iInstance); \
728	}while(0)
729
730	# define PROFILE_COUNTER(name, dsc) \
731	COUNTER(name, STAMTYPE_PROFILE, STAMUNIT_TICKS_PER_CALL, dsc);
732	# define COUNTING_COUNTER(name, dsc) \
733	COUNTER(name, STAMTYPE_COUNTER, STAMUNIT_COUNT, dsc);
734
735	#include "counters.h"
736
737	#undef COUNTER
738	#undef PROFILE_COUNTER
739	#undef COUNTING_COUNTER
740
741	#endif /* VBOX_WITH_STATISTICS */
742	}
743
744	/**
745	* Marks the link as up, making it possible to establish new connections.
746	*/
747	void slirp_link_up(PNATState pData)
748	{
749	link_up = 1;
750	}
751
752	/**
753	* Marks the link as down and cleans up the current connections.
754	*/
755	void slirp_link_down(PNATState pData)
756	{
757	struct socket *so;
758
759	while ((so = tcb.so_next) != &tcb)
760	{
761	if (so->so_state & SS_NOFDREF \|\| so->s == -1)
762	sofree(pData, so);
763	else
764	tcp_drop(pData, sototcpcb(so), 0);
765	}
766
767	while ((so = udb.so_next) != &udb)
768	udp_detach(pData, so);
769
770	link_up = 0;
771	}
772
773	/**
774	* Terminates the slirp component.
775	*/
776	void slirp_term(PNATState pData)
777	{
778	#ifndef VBOX_WITH_MULTI_DNS
779	if (pData->pszDomain)
780	RTStrFree((char )(void )pData->pszDomain);
781	#endif
782
783	#ifdef RT_OS_WINDOWS
784	pData->pfIcmpCloseHandle(pData->icmp_socket.sh);
785	FreeLibrary(pData->hmIcmpLibrary);
786	RTMemFree(pData->pvIcmpBuffer);
787	# else
788	closesocket(pData->icmp_socket.s);
789	#endif
790
791	slirp_link_down(pData);
792	#ifdef VBOX_WITH_MULTI_DNS
793	slirp_release_dns_list(pData);
794	#endif
795	#ifdef RT_OS_WINDOWS
796	WSACleanup();
797	#endif
798	#ifdef VBOX_WITH_SLIRP_ALIAS
799	while(!LIST_EMPTY(&instancehead)) {
800	struct libalias *la = LIST_FIRST(&instancehead);
801	/* libalias do all clean up */
802	LibAliasUninit(la);
803	}
804	#endif
805	#ifdef LOG_ENABLED
806	Log(("\n"
807	"NAT statistics\n"
808	"--------------\n"
809	"\n"));
810	ipstats(pData);
811	tcpstats(pData);
812	udpstats(pData);
813	icmpstats(pData);
814	mbufstats(pData);
815	sockstats(pData);
816	Log(("\n"
817	"\n"
818	"\n"));
819	#endif
820	RTMemFree(pData);
821	}
822
823
824	#define CONN_CANFSEND(so) (((so)->so_state & (SS_FCANTSENDMORE\|SS_ISFCONNECTED)) == SS_ISFCONNECTED)
825	#define CONN_CANFRCV(so) (((so)->so_state & (SS_FCANTRCVMORE\|SS_ISFCONNECTED)) == SS_ISFCONNECTED)
826	#define UPD_NFDS(x) do { if (nfds < (x)) nfds = (x); } while (0)
827
828	/*
829	* curtime kept to an accuracy of 1ms
830	*/
831	static void updtime(PNATState pData)
832	{
833	#ifdef RT_OS_WINDOWS
834	struct _timeb tb;
835
836	_ftime(&tb);
837	curtime = (u_int)tb.time * (u_int)1000;
838	curtime += (u_int)tb.millitm;
839	#else
840	gettimeofday(&tt, 0);
841
842	curtime = (u_int)tt.tv_sec * (u_int)1000;
843	curtime += (u_int)tt.tv_usec / (u_int)1000;
844
845	if ((tt.tv_usec % 1000) >= 500)
846	curtime++;
847	#endif
848	}
849
850	#ifdef RT_OS_WINDOWS
851	void slirp_select_fill(PNATState pData, int *pnfds)
852	#else /* RT_OS_WINDOWS */
853	void slirp_select_fill(PNATState pData, int pnfds, struct pollfd polls)
854	#endif /* !RT_OS_WINDOWS */
855	{
856	struct socket so, so_next;
857	int nfds;
858	#if defined(RT_OS_WINDOWS)
859	int rc;
860	int error;
861	#else
862	int poll_index = 0;
863	#endif
864	int i;
865
866	SLIRP_PROFILE_START(Fill, a);
867
868	nfds = *pnfds;
869
870	/*
871	* First, TCP sockets
872	*/
873	do_slowtimo = 0;
874	if (!link_up)
875	goto done;
876	/*
877	* *_slowtimo needs calling if there are IP fragments
878	* in the fragment queue, or there are TCP connections active
879	*/
880	/* XXX:
881	* triggering of fragment expiration should be the same but use new macroses
882	*/
883	do_slowtimo = (tcb.so_next != &tcb);
884	if (!do_slowtimo)
885	{
886	for (i = 0; i < IPREASS_NHASH; i++)
887	{
888	if (!TAILQ_EMPTY(&ipq[i]))
889	{
890	do_slowtimo = 1;
891	break;
892	}
893	}
894	}
895	ICMP_ENGAGE_EVENT(&pData->icmp_socket, readfds);
896
897	SLIRP_COUNTER_RESET(TCP);
898	SLIRP_COUNTER_RESET(TCPHot);
899
900	QSOCKET_FOREACH(so, so_next, tcp)
901	/* { */
902	#if !defined(RT_OS_WINDOWS)
903	so->so_poll_index = -1;
904	#endif
905	SLIRP_COUNTER_INC(TCP);
906
907	/*
908	* See if we need a tcp_fasttimo
909	*/
910	if ( time_fasttimo == 0
911	&& so->so_tcpcb != NULL
912	&& so->so_tcpcb->t_flags & TF_DELACK)
913	time_fasttimo = curtime; /* Flag when we want a fasttimo */
914
915	/*
916	* NOFDREF can include still connecting to local-host,
917	* newly socreated() sockets etc. Don't want to select these.
918	*/
919	if (so->so_state & SS_NOFDREF \|\| so->s == -1)
920	CONTINUE(tcp);
921
922	/*
923	* Set for reading sockets which are accepting
924	*/
925	if (so->so_state & SS_FACCEPTCONN)
926	{
927	SLIRP_COUNTER_INC(TCPHot);
928	TCP_ENGAGE_EVENT1(so, readfds);
929	CONTINUE(tcp);
930	}
931
932	/*
933	* Set for writing sockets which are connecting
934	*/
935	if (so->so_state & SS_ISFCONNECTING)
936	{
937	Log2(("connecting %R[natsock] engaged\n",so));
938	SLIRP_COUNTER_INC(TCPHot);
939	TCP_ENGAGE_EVENT1(so, writefds);
940	}
941
942	/*
943	* Set for writing if we are connected, can send more, and
944	* we have something to send
945	*/
946	if (CONN_CANFSEND(so) && so->so_rcv.sb_cc)
947	{
948	SLIRP_COUNTER_INC(TCPHot);
949	TCP_ENGAGE_EVENT1(so, writefds);
950	}
951
952	/*
953	* Set for reading (and urgent data) if we are connected, can
954	* receive more, and we have room for it XXX /2 ?
955	*/
956	if (CONN_CANFRCV(so) && (so->so_snd.sb_cc < (so->so_snd.sb_datalen/2)))
957	{
958	SLIRP_COUNTER_INC(TCPHot);
959	TCP_ENGAGE_EVENT2(so, readfds, xfds);
960	}
961	LOOP_LABEL(tcp, so, so_next);
962	}
963
964	/*
965	* UDP sockets
966	*/
967	SLIRP_COUNTER_RESET(UDP);
968	SLIRP_COUNTER_RESET(UDPHot);
969
970	QSOCKET_FOREACH(so, so_next, udp)
971	/* { */
972
973	SLIRP_COUNTER_INC(UDP);
974	#if !defined(RT_OS_WINDOWS)
975	so->so_poll_index = -1;
976	#endif
977
978	/*
979	* See if it's timed out
980	*/
981	if (so->so_expire)
982	{
983	if (so->so_expire <= curtime)
984	{
985	#ifdef VBOX_WITH_SLIRP_DNS_PROXY
986	Log2(("NAT: %R[natsock] expired\n", so));
987	if (so->so_timeout != NULL)
988	{
989	so->so_timeout(pData, so, so->so_timeout_arg);
990	}
991	#endif
992	#ifdef VBOX_WITH_SLIRP_MT
993	/* we need so_next for continue our cycle*/
994	so_next = so->so_next;
995	#endif
996	UDP_DETACH(pData, so, so_next);
997	CONTINUE_NO_UNLOCK(udp);
998	}
999	else
1000	do_slowtimo = 1; /* Let socket expire */
1001	}
1002
1003	/*
1004	* When UDP packets are received from over the link, they're
1005	* sendto()'d straight away, so no need for setting for writing
1006	* Limit the number of packets queued by this session to 4.
1007	* Note that even though we try and limit this to 4 packets,
1008	* the session could have more queued if the packets needed
1009	* to be fragmented.
1010	*
1011	* (XXX <= 4 ?)
1012	*/
1013	if ((so->so_state & SS_ISFCONNECTED) && so->so_queued <= 4)
1014	{
1015	SLIRP_COUNTER_INC(UDPHot);
1016	UDP_ENGAGE_EVENT(so, readfds);
1017	}
1018	LOOP_LABEL(udp, so, so_next);
1019	}
1020	done:
1021
1022	#if defined(RT_OS_WINDOWS)
1023	*pnfds = VBOX_EVENT_COUNT;
1024	#else /* RT_OS_WINDOWS */
1025	AssertRelease(poll_index <= *pnfds);
1026	*pnfds = poll_index;
1027	#endif /* !RT_OS_WINDOWS */
1028
1029	SLIRP_PROFILE_STOP(Fill, a);
1030	}
1031
1032	#if defined(RT_OS_WINDOWS)
1033	void slirp_select_poll(PNATState pData, int fTimeout, int fIcmp)
1034	#else /* RT_OS_WINDOWS */
1035	void slirp_select_poll(PNATState pData, struct pollfd *polls, int ndfs)
1036	#endif /* !RT_OS_WINDOWS */
1037	{
1038	struct socket so, so_next;
1039	int ret;
1040	#if defined(RT_OS_WINDOWS)
1041	WSANETWORKEVENTS NetworkEvents;
1042	int rc;
1043	int error;
1044	#else
1045	int poll_index = 0;
1046	#endif
1047
1048	SLIRP_PROFILE_START(Poll, a);
1049
1050	/* Update time */
1051	updtime(pData);
1052
1053	/*
1054	* See if anything has timed out
1055	*/
1056	if (link_up)
1057	{
1058	if (time_fasttimo && ((curtime - time_fasttimo) >= 2))
1059	{
1060	SLIRP_PROFILE_START(FastTimer, a);
1061	tcp_fasttimo(pData);
1062	time_fasttimo = 0;
1063	SLIRP_PROFILE_STOP(FastTimer, a);
1064	}
1065	if (do_slowtimo && ((curtime - last_slowtimo) >= 499))
1066	{
1067	SLIRP_PROFILE_START(SlowTimer, a);
1068	ip_slowtimo(pData);
1069	tcp_slowtimo(pData);
1070	last_slowtimo = curtime;
1071	SLIRP_PROFILE_STOP(SlowTimer, a);
1072	}
1073	}
1074	#if defined(RT_OS_WINDOWS)
1075	if (fTimeout)
1076	return; /* only timer update */
1077	#endif
1078
1079	/*
1080	* Check sockets
1081	*/
1082	if (!link_up)
1083	goto done;
1084	#if defined(RT_OS_WINDOWS)
1085	/*XXX: before renaming please make see define
1086	* fIcmp in slirp_state.h
1087	*/
1088	if (fIcmp)
1089	sorecvfrom(pData, &pData->icmp_socket);
1090	#else
1091	if ( (pData->icmp_socket.s != -1)
1092	&& CHECK_FD_SET(&pData->icmp_socket, ignored, readfds))
1093	sorecvfrom(pData, &pData->icmp_socket);
1094	#endif
1095	/*
1096	* Check TCP sockets
1097	*/
1098	QSOCKET_FOREACH(so, so_next, tcp)
1099	/* { */
1100
1101	#ifdef VBOX_WITH_SLIRP_MT
1102	if ( so->so_state & SS_NOFDREF
1103	&& so->so_deleted == 1)
1104	{
1105	struct socket son, sop = NULL;
1106	QSOCKET_LOCK(tcb);
1107	if (so->so_next != NULL)
1108	{
1109	if (so->so_next != &tcb)
1110	SOCKET_LOCK(so->so_next);
1111	son = so->so_next;
1112	}
1113	if ( so->so_prev != &tcb
1114	&& so->so_prev != NULL)
1115	{
1116	SOCKET_LOCK(so->so_prev);
1117	sop = so->so_prev;
1118	}
1119	QSOCKET_UNLOCK(tcb);
1120	remque(pData, so);
1121	NSOCK_DEC();
1122	SOCKET_UNLOCK(so);
1123	SOCKET_LOCK_DESTROY(so);
1124	RTMemFree(so);
1125	so_next = son;
1126	if (sop != NULL)
1127	SOCKET_UNLOCK(sop);
1128	CONTINUE_NO_UNLOCK(tcp);
1129	}
1130	#endif
1131	/*
1132	* FD_ISSET is meaningless on these sockets
1133	* (and they can crash the program)
1134	*/
1135	if (so->so_state & SS_NOFDREF \|\| so->s == -1)
1136	CONTINUE(tcp);
1137
1138	POLL_TCP_EVENTS(rc, error, so, &NetworkEvents);
1139
1140	LOG_NAT_SOCK(so, TCP, &NetworkEvents, readfds, writefds, xfds);
1141
1142
1143	/*
1144	* Check for URG data
1145	* This will soread as well, so no need to
1146	* test for readfds below if this succeeds
1147	*/
1148
1149	/* out-of-band data */
1150	if (CHECK_FD_SET(so, NetworkEvents, xfds))
1151	{
1152	sorecvoob(pData, so);
1153	}
1154
1155	/*
1156	* Check sockets for reading
1157	*/
1158	else if ( CHECK_FD_SET(so, NetworkEvents, readfds)
1159	\|\| WIN_CHECK_FD_SET(so, NetworkEvents, acceptds))
1160	{
1161	/*
1162	* Check for incoming connections
1163	*/
1164	if (so->so_state & SS_FACCEPTCONN)
1165	{
1166	TCP_CONNECT(pData, so);
1167	#if defined(RT_OS_WINDOWS)
1168	if (!(NetworkEvents.lNetworkEvents & FD_CLOSE))
1169	#endif
1170	CONTINUE(tcp);
1171	}
1172
1173	ret = soread(pData, so);
1174	/* Output it if we read something */
1175	if (RT_LIKELY(ret > 0))
1176	TCP_OUTPUT(pData, sototcpcb(so));
1177	}
1178
1179	#if defined(RT_OS_WINDOWS)
1180	/*
1181	* Check for FD_CLOSE events.
1182	* in some cases once FD_CLOSE engaged on socket it could be flashed latter (for some reasons)
1183	*/
1184	if ( (NetworkEvents.lNetworkEvents & FD_CLOSE)
1185	\|\| (so->so_close == 1))
1186	{
1187	so->so_close = 1; /* mark it */
1188	/*
1189	* drain the socket
1190	*/
1191	for (;;)
1192	{
1193	ret = soread(pData, so);
1194	if (ret > 0)
1195	TCP_OUTPUT(pData, sototcpcb(so));
1196	else
1197	break;
1198	}
1199	CONTINUE(tcp);
1200	}
1201	#endif
1202
1203	/*
1204	* Check sockets for writing
1205	*/
1206	if (CHECK_FD_SET(so, NetworkEvents, writefds))
1207	{
1208	/*
1209	* Check for non-blocking, still-connecting sockets
1210	*/
1211	if (so->so_state & SS_ISFCONNECTING)
1212	{
1213	Log2(("connecting %R[natsock] catched\n", so));
1214	/* Connected */
1215	so->so_state &= ~SS_ISFCONNECTING;
1216
1217	/*
1218	* This should be probably guarded by PROBE_CONN too. Anyway,
1219	* we disable it on OS/2 because the below send call returns
1220	* EFAULT which causes the opened TCP socket to close right
1221	* after it has been opened and connected.
1222	*/
1223	#ifndef RT_OS_OS2
1224	ret = send(so->s, (const char *)&ret, 0, 0);
1225	if (ret < 0)
1226	{
1227	/* XXXXX Must fix, zero bytes is a NOP */
1228	if ( errno == EAGAIN
1229	\|\| errno == EWOULDBLOCK
1230	\|\| errno == EINPROGRESS
1231	\|\| errno == ENOTCONN)
1232	CONTINUE(tcp);
1233
1234	/* else failed */
1235	so->so_state = SS_NOFDREF;
1236	}
1237	/* else so->so_state &= ~SS_ISFCONNECTING; */
1238	#endif
1239
1240	/*
1241	* Continue tcp_input
1242	*/
1243	TCP_INPUT(pData, (struct mbuf *)NULL, sizeof(struct ip), so);
1244	/* continue; */
1245	}
1246	else
1247	SOWRITE(ret, pData, so);
1248	/*
1249	* XXX If we wrote something (a lot), there could be the need
1250	* for a window update. In the worst case, the remote will send
1251	* a window probe to get things going again.
1252	*/
1253	}
1254
1255	/*
1256	* Probe a still-connecting, non-blocking socket
1257	* to check if it's still alive
1258	*/
1259	#ifdef PROBE_CONN
1260	if (so->so_state & SS_ISFCONNECTING)
1261	{
1262	ret = recv(so->s, (char *)&ret, 0, 0);
1263
1264	if (ret < 0)
1265	{
1266	/* XXX */
1267	if ( errno == EAGAIN
1268	\|\| errno == EWOULDBLOCK
1269	\|\| errno == EINPROGRESS
1270	\|\| errno == ENOTCONN)
1271	{
1272	CONTINUE(tcp); /* Still connecting, continue */
1273	}
1274
1275	/* else failed */
1276	so->so_state = SS_NOFDREF;
1277
1278	/* tcp_input will take care of it */
1279	}
1280	else
1281	{
1282	ret = send(so->s, &ret, 0, 0);
1283	if (ret < 0)
1284	{
1285	/* XXX */
1286	if ( errno == EAGAIN
1287	\|\| errno == EWOULDBLOCK
1288	\|\| errno == EINPROGRESS
1289	\|\| errno == ENOTCONN)
1290	{
1291	CONTINUE(tcp);
1292	}
1293	/* else failed */
1294	so->so_state = SS_NOFDREF;
1295	}
1296	else
1297	so->so_state &= ~SS_ISFCONNECTING;
1298
1299	}
1300	TCP_INPUT((struct mbuf *)NULL, sizeof(struct ip),so);
1301	} /* SS_ISFCONNECTING */
1302	#endif
1303	#ifndef RT_OS_WINDOWS
1304	if ( UNIX_CHECK_FD_SET(so, NetworkEvents, rdhup)
1305	\|\| UNIX_CHECK_FD_SET(so, NetworkEvents, rderr))
1306	{
1307	int err;
1308	int inq, outq;
1309	int status;
1310	socklen_t optlen = sizeof(int);
1311	inq = outq = 0;
1312	status = getsockopt(so->s, SOL_SOCKET, SO_ERROR, &err, &optlen);
1313	if (status != 0)
1314	Log(("NAT: can't get error status from %R[natsock]\n", so));
1315	#ifndef RT_OS_SOLARIS
1316	status = ioctl(so->s, FIONREAD, &inq); /* tcp(7) recommends SIOCINQ which is Linux specific */
1317	if (status != 0 \|\| status != EINVAL)
1318	{
1319	/* EINVAL returned if socket in listen state tcp(7)*/
1320	Log(("NAT: can't get depth of IN queue status from %R[natsock]\n", so));
1321	}
1322	status = ioctl(so->s, TIOCOUTQ, &outq); /* SIOCOUTQ see previous comment */
1323	if (status != 0)
1324	Log(("NAT: can't get depth of OUT queue from %R[natsock]\n", so));
1325	#else
1326	/*
1327	* Solaris has bit different ioctl commands and its handlings
1328	* hint: streamio(7) I_NREAD
1329	*/
1330	#endif
1331	if ( so->so_state & SS_ISFCONNECTING
1332	\|\| UNIX_CHECK_FD_SET(so, NetworkEvents, readfds))
1333	{
1334	/**
1335	* Check if we need here take care about gracefull connection
1336	* @todo try with proxy server
1337	*/
1338	if (UNIX_CHECK_FD_SET(so, NetworkEvents, readfds))
1339	{
1340	/*
1341	* Never meet inq != 0 or outq != 0, anyway let it stay for a while
1342	* in case it happens we'll able to detect it.
1343	* Give TCP/IP stack wait or expire the socket.
1344	*/
1345	Log(("NAT: %R[natsock] err(%d:%s) s(in:%d,out:%d)happens on read I/O, "
1346	"other side close connection \n", so, err, strerror(err), inq, outq));
1347	CONTINUE(tcp);
1348	}
1349	goto tcp_input_close;
1350	}
1351	if ( !UNIX_CHECK_FD_SET(so, NetworkEvents, readfds)
1352	&& !UNIX_CHECK_FD_SET(so, NetworkEvents, writefds)
1353	&& !UNIX_CHECK_FD_SET(so, NetworkEvents, xfds))
1354	{
1355	Log(("NAT: system expires the socket %R[natsock] err(%d:%s) s(in:%d,out:%d) happens on non-I/O. ",
1356	so, err, strerror(err), inq, outq));
1357	goto tcp_input_close;
1358	}
1359	Log(("NAT: %R[natsock] we've met(%d:%s) s(in:%d, out:%d) unhandled combination hup (%d) "
1360	"rederr(%d) on (r:%d, w:%d, x:%d)\n",
1361	so, err, strerror(err),
1362	inq, outq,
1363	UNIX_CHECK_FD_SET(so, ign, rdhup),
1364	UNIX_CHECK_FD_SET(so, ign, rderr),
1365	UNIX_CHECK_FD_SET(so, ign, readfds),
1366	UNIX_CHECK_FD_SET(so, ign, writefds),
1367	UNIX_CHECK_FD_SET(so, ign, xfds)));
1368	/*
1369	* Give OS's TCP/IP stack a chance to resolve an issue or expire the socket.
1370	*/
1371	CONTINUE(tcp);
1372	tcp_input_close:
1373	so->so_state = SS_NOFDREF; /cause connection valid tcp connection termination and socket closing /
1374	TCP_INPUT(pData, (struct mbuf *)NULL, sizeof(struct ip), so);
1375	CONTINUE(tcp);
1376	}
1377	#endif
1378	LOOP_LABEL(tcp, so, so_next);
1379	}
1380
1381	/*
1382	* Now UDP sockets.
1383	* Incoming packets are sent straight away, they're not buffered.
1384	* Incoming UDP data isn't buffered either.
1385	*/
1386	QSOCKET_FOREACH(so, so_next, udp)
1387	/* { */
1388	#ifdef VBOX_WITH_SLIRP_MT
1389	if ( so->so_state & SS_NOFDREF
1390	&& so->so_deleted == 1)
1391	{
1392	struct socket son, sop = NULL;
1393	QSOCKET_LOCK(udb);
1394	if (so->so_next != NULL)
1395	{
1396	if (so->so_next != &udb)
1397	SOCKET_LOCK(so->so_next);
1398	son = so->so_next;
1399	}
1400	if ( so->so_prev != &udb
1401	&& so->so_prev != NULL)
1402	{
1403	SOCKET_LOCK(so->so_prev);
1404	sop = so->so_prev;
1405	}
1406	QSOCKET_UNLOCK(udb);
1407	remque(pData, so);
1408	NSOCK_DEC();
1409	SOCKET_UNLOCK(so);
1410	SOCKET_LOCK_DESTROY(so);
1411	RTMemFree(so);
1412	so_next = son;
1413	if (sop != NULL)
1414	SOCKET_UNLOCK(sop);
1415	CONTINUE_NO_UNLOCK(udp);
1416	}
1417	#endif
1418	POLL_UDP_EVENTS(rc, error, so, &NetworkEvents);
1419
1420	LOG_NAT_SOCK(so, UDP, &NetworkEvents, readfds, writefds, xfds);
1421
1422	if (so->s != -1 && CHECK_FD_SET(so, NetworkEvents, readfds))
1423	{
1424	SORECVFROM(pData, so);
1425	}
1426	LOOP_LABEL(udp, so, so_next);
1427	}
1428
1429	done:
1430	#ifndef VBOX_WITH_SLIRP_MT
1431	/*
1432	* See if we can start outputting
1433	*/
1434	if (if_queued && link_up)
1435	if_start(pData);
1436	#endif
1437
1438	SLIRP_PROFILE_STOP(Poll, a);
1439	}
1440
1441	#ifndef VBOX_WITHOUT_SLIRP_CLIENT_ETHER
1442	#define ETH_ALEN 6
1443	#define ETH_HLEN 14
1444
1445	#define ARPOP_REQUEST 1 /* ARP request */
1446	#define ARPOP_REPLY 2 /* ARP reply */
1447
1448	struct ethhdr
1449	{
1450	unsigned char h_dest[ETH_ALEN]; /* destination eth addr */
1451	unsigned char h_source[ETH_ALEN]; /* source ether addr */
1452	unsigned short h_proto; /* packet type ID field */
1453	};
1454	AssertCompileSize(struct ethhdr, 14);
1455	#endif
1456
1457	struct arphdr
1458	{
1459	unsigned short ar_hrd; /* format of hardware address */
1460	unsigned short ar_pro; /* format of protocol address */
1461	unsigned char ar_hln; /* length of hardware address */
1462	unsigned char ar_pln; /* length of protocol address */
1463	unsigned short ar_op; /* ARP opcode (command) */
1464
1465	/*
1466	* Ethernet looks like this : This bit is variable sized however...
1467	*/
1468	unsigned char ar_sha[ETH_ALEN]; /* sender hardware address */
1469	unsigned char ar_sip[4]; /* sender IP address */
1470	unsigned char ar_tha[ETH_ALEN]; /* target hardware address */
1471	unsigned char ar_tip[4]; /* target IP address */
1472	};
1473	AssertCompileSize(struct arphdr, 28);
1474
1475	static void arp_input(PNATState pData, struct mbuf *m)
1476	{
1477	struct ethhdr *eh;
1478	struct ethhdr *reh;
1479	struct arphdr *ah;
1480	struct arphdr *rah;
1481	int ar_op;
1482	struct ex_list *ex_ptr;
1483	uint32_t htip;
1484	uint32_t tip;
1485	struct mbuf *mr;
1486	eh = mtod(m, struct ethhdr *);
1487	ah = (struct arphdr *)&eh[1];
1488	htip = ntohl((uint32_t)ah->ar_tip);
1489	tip = (uint32_t)ah->ar_tip;
1490
1491	mr = m_get(pData);
1492	#ifdef VBOX_WITHOUT_SLIRP_CLIENT_ETHER
1493	reh = mtod(mr, struct ethhdr *);
1494	memcpy(reh->h_source, eh->h_source, ETH_ALEN); /* XXX: if_encap will swap src and dst*/
1495	Log4(("NAT: arp:%R[ether]->%R[ether]\n",
1496	reh->h_source, reh->h_dest));
1497	Log4(("NAT: arp: %R[IP4]\n", &tip));
1498	#endif
1499	mr->m_data += if_maxlinkhdr;
1500	mr->m_len = sizeof(struct arphdr);
1501	rah = mtod(mr, struct arphdr *);
1502
1503	ar_op = ntohs(ah->ar_op);
1504	switch(ar_op)
1505	{
1506	case ARPOP_REQUEST:
1507	#ifdef VBOX_WITH_NAT_SERVICE
1508	if (tip == special_addr.s_addr) goto arp_ok;
1509	#endif
1510	if ((htip & pData->netmask) == ntohl(special_addr.s_addr))
1511	{
1512	if ( CTL_CHECK(htip, CTL_DNS)
1513	\|\| CTL_CHECK(htip, CTL_ALIAS)
1514	\|\| CTL_CHECK(htip, CTL_TFTP))
1515	goto arp_ok;
1516	for (ex_ptr = exec_list; ex_ptr; ex_ptr = ex_ptr->ex_next)
1517	{
1518	if ((htip & ~pData->netmask) == ex_ptr->ex_addr)
1519	{
1520	goto arp_ok;
1521	}
1522	}
1523	return;
1524	arp_ok:
1525	rah->ar_hrd = htons(1);
1526	rah->ar_pro = htons(ETH_P_IP);
1527	rah->ar_hln = ETH_ALEN;
1528	rah->ar_pln = 4;
1529	rah->ar_op = htons(ARPOP_REPLY);
1530	memcpy(rah->ar_sha, special_ethaddr, ETH_ALEN);
1531
1532	switch (htip & ~pData->netmask)
1533	{
1534	case CTL_DNS:
1535	case CTL_ALIAS:
1536	rah->ar_sha[5] = (uint8_t)(htip & ~pData->netmask);
1537	break;
1538	default:;
1539	}
1540
1541	memcpy(rah->ar_sip, ah->ar_tip, 4);
1542	memcpy(rah->ar_tha, ah->ar_sha, ETH_ALEN);
1543	memcpy(rah->ar_tip, ah->ar_sip, 4);
1544	if_encap(pData, ETH_P_ARP, mr);
1545	m_free(pData, m);
1546	}
1547	break;
1548	default:
1549	break;
1550	}
1551	}
1552
1553	void slirp_input(PNATState pData, const uint8_t *pkt, int pkt_len)
1554	{
1555	struct mbuf *m;
1556	int proto;
1557	static bool fWarnedIpv6;
1558	struct ethhdr eh = (struct ethhdr)pkt;
1559
1560	Log2(("NAT: slirp_input %d\n", pkt_len));
1561	if (pkt_len < ETH_HLEN)
1562	{
1563	LogRel(("NAT: packet having size %d has been ingnored\n", pkt_len));
1564	return;
1565	}
1566	Log4(("NAT: in:%R[ether]->%R[ether]\n", &eh->h_source, &eh->h_dest));
1567	#ifdef VBOX_WITHOUT_SLIRP_CLIENT_ETHER
1568	if (memcmp(eh->h_source, special_ethaddr, ETH_ALEN) == 0)
1569	{
1570	/* @todo vasily: add ether logging routine in debug.c */
1571	Log(("NAT: packet was addressed to other MAC\n"));
1572	RTMemFree((void *)pkt);
1573	return;
1574	}
1575	#endif
1576
1577	m = m_get(pData);
1578	if (!m)
1579	{
1580	LogRel(("NAT: can't allocate new mbuf\n"));
1581	return;
1582	}
1583
1584	/* Note: we add to align the IP header */
1585
1586	if (M_FREEROOM(m) < pkt_len)
1587	m_inc(m, pkt_len);
1588
1589	m->m_len = pkt_len ;
1590	memcpy(m->m_data, pkt, pkt_len);
1591
1592	proto = ntohs((uint16_t )(pkt + 12));
1593	switch(proto)
1594	{
1595	case ETH_P_ARP:
1596	arp_input(pData, m);
1597	break;
1598	case ETH_P_IP:
1599	/* Update time. Important if the network is very quiet, as otherwise
1600	* the first outgoing connection gets an incorrect timestamp. */
1601	updtime(pData);
1602	m_adj(m, ETH_HLEN);
1603	ip_input(pData, m);
1604	break;
1605	case ETH_P_IPV6:
1606	m_free(pData, m);
1607	if (!fWarnedIpv6)
1608	{
1609	LogRel(("NAT: IPv6 not supported\n"));
1610	fWarnedIpv6 = true;
1611	}
1612	break;
1613	default:
1614	Log(("NAT: Unsupported protocol %x\n", proto));
1615	m_free(pData, m);
1616	break;
1617	}
1618	RTMemFree((void *)pkt);
1619	}
1620
1621	/* output the IP packet to the ethernet device */
1622	void if_encap(PNATState pData, uint16_t eth_proto, struct mbuf *m)
1623	{
1624	struct ethhdr *eh;
1625	uint8_t *buf = RTMemAlloc(1600);
1626
1627	m->m_data -= if_maxlinkhdr;
1628	m->m_len += ETH_HLEN;
1629	eh = mtod(m, struct ethhdr *);
1630
1631	if(MBUF_HEAD(m) != m->m_data)
1632	{
1633	LogRel(("NAT: ethernet detects corruption of the packet"));
1634	AssertMsgFailed(("!!Ethernet frame corrupted!!"));
1635	}
1636
1637	#ifndef VBOX_WITHOUT_SLIRP_CLIENT_ETHER
1638	memcpy(eh->h_dest, client_ethaddr, ETH_ALEN);
1639	memcpy(eh->h_source, special_ethaddr, ETH_ALEN - 1);
1640	/* XXX: not correct */
1641	eh->h_source[5] = CTL_ALIAS;
1642	#else
1643	if (memcmp(eh->h_source, special_ethaddr, ETH_ALEN) != 0)
1644	{
1645	memcpy(eh->h_dest, eh->h_source, ETH_ALEN);
1646	memcpy(eh->h_source, special_ethaddr, ETH_ALEN);
1647	Assert(memcmp(eh->h_dest, special_ethaddr, ETH_ALEN) != 0);
1648	if (memcmp(eh->h_dest, zerro_ethaddr, ETH_ALEN) == 0)
1649	{
1650	/* don't do anything */
1651	goto done;
1652	}
1653	}
1654	#endif
1655	eh->h_proto = htons(eth_proto);
1656	#if 0
1657	slirp_output(pData->pvUser, m, mtod(m, uint8_t *), m->m_len);
1658	#else
1659	memcpy(buf, mtod(m, uint8_t *), m->m_len);
1660	slirp_output(pData->pvUser, NULL, buf, m->m_len);
1661	done:
1662	m_free(pData, m);
1663	#endif
1664	}
1665
1666	int slirp_redir(PNATState pData, int is_udp, int host_port,
1667	struct in_addr guest_addr, int guest_port)
1668	{
1669	struct socket *so;
1670	#ifdef VBOX_WITH_SLIRP_ALIAS
1671	struct alias_link *link;
1672	struct libalias *lib;
1673	int flags;
1674	struct sockaddr sa;
1675	struct sockaddr_in *psin;
1676	socklen_t socketlen;
1677	struct in_addr alias;
1678	int rc;
1679	#endif
1680	Log2(("NAT: set redirect %s hp:%d gp:%d\n", (is_udp?"UDP":"TCP"), host_port, guest_port));
1681	if (is_udp)
1682	{
1683	so = udp_listen(pData, htons(host_port), guest_addr.s_addr,
1684	htons(guest_port), 0);
1685	}
1686	else
1687	{
1688	so = solisten(pData, htons(host_port), guest_addr.s_addr,
1689	htons(guest_port), 0);
1690	}
1691	#ifndef VBOX_WITH_SLIRP_ALIAS
1692	Log2(("NAT: redirecting socket %R[natsock]\n", so));
1693	return (so != NULL ? 0 : -1);
1694	#else
1695
1696	psin = (struct sockaddr_in *)&sa;
1697	psin->sin_family = AF_INET;
1698	psin->sin_port = 0;
1699	psin->sin_addr.s_addr = INADDR_ANY;
1700	socketlen = sizeof(struct sockaddr);
1701
1702	rc = getsockname(so->s, &sa, &socketlen);
1703	if (rc < 0 \|\| sa.sa_family != AF_INET)
1704	{
1705	Log(("NAT: can't get socket's name\n"));
1706	return 1;
1707	}
1708
1709	psin = (struct sockaddr_in *)&sa;
1710
1711	#if 1
1712	lib = LibAliasInit(pData, NULL);
1713	flags = LibAliasSetMode(lib, 0, 0);
1714	flags \|= PKT_ALIAS_LOG; /* set logging */
1715	flags \|= PKT_ALIAS_REVERSE; /* set logging */
1716	flags = LibAliasSetMode(lib, flags, ~0);
1717	#else
1718	lib = LIST_FIRST(&instancehead);
1719	#endif
1720
1721	alias.s_addr = htonl(ntohl(guest_addr.s_addr) \| CTL_ALIAS);
1722	link = LibAliasRedirectPort(lib, psin->sin_addr, htons(host_port),
1723	alias, htons(guest_port),
1724	special_addr, -1, /* not very clear for now*/
1725	(is_udp ? IPPROTO_UDP : IPPROTO_TCP));
1726	if (link == NULL)
1727	{
1728	Log(("NAT: can't create redirect\n"));
1729	return 1;
1730	}
1731	so->so_la = lib;
1732
1733	return 0;
1734	#endif
1735	}
1736
1737	int slirp_add_exec(PNATState pData, int do_pty, const char *args, int addr_low_byte,
1738	int guest_port)
1739	{
1740	return add_exec(&exec_list, do_pty, (char *)args,
1741	addr_low_byte, htons(guest_port));
1742	}
1743
1744	void slirp_set_ethaddr(PNATState pData, const uint8_t *ethaddr)
1745	{
1746	#ifndef VBOX_WITHOUT_SLIRP_CLIENT_ETHER
1747	memcpy(client_ethaddr, ethaddr, ETH_ALEN);
1748	#endif
1749	}
1750
1751	#if defined(RT_OS_WINDOWS)
1752	HANDLE *slirp_get_events(PNATState pData)
1753	{
1754	return pData->phEvents;
1755	}
1756	void slirp_register_external_event(PNATState pData, HANDLE hEvent, int index)
1757	{
1758	pData->phEvents[index] = hEvent;
1759	}
1760	#endif
1761
1762	unsigned int slirp_get_timeout_ms(PNATState pData)
1763	{
1764	if (link_up)
1765	{
1766	if (time_fasttimo)
1767	return 2;
1768	if (do_slowtimo)
1769	return 500; /* see PR_SLOWHZ */
1770	}
1771	return 0;
1772	}
1773
1774	#ifndef RT_OS_WINDOWS
1775	int slirp_get_nsock(PNATState pData)
1776	{
1777	return pData->nsock;
1778	}
1779	#endif
1780
1781	/*
1782	* this function called from NAT thread
1783	*/
1784	void slirp_post_sent(PNATState pData, void *pvArg)
1785	{
1786	struct socket *so = 0;
1787	struct tcpcb *tp = 0;
1788	struct mbuf m = (struct mbuf )pvArg;
1789	m_free(pData, m);
1790	}
1791	#ifdef VBOX_WITH_SLIRP_MT
1792	void slirp_process_queue(PNATState pData)
1793	{
1794	RTReqProcess(pData->pReqQueue, RT_INDEFINITE_WAIT);
1795	}
1796	void *slirp_get_queue(PNATState pData)
1797	{
1798	return pData->pReqQueue;
1799	}
1800	#endif
1801
1802	uint16_t slirp_get_service(int proto, uint16_t dport, uint16_t sport)
1803	{
1804	uint16_t hdport, hsport, service;
1805	hdport = ntohs(dport);
1806	hsport = ntohs(sport);
1807	Log2(("proto: %d, dport: %d sport: %d\n", proto, hdport, hsport));
1808	service = 0;
1809	#if 0
1810	/* Always return 0 here */
1811	switch (hdport)
1812	{
1813	case 500:
1814	if (hsport != 500) /* vpnc by default try operate in src:500/dst:500 mode*/
1815	/* Not sure why this make Cisco VPN client's connection more stable,
1816	* at least on some servers
1817	*/
1818	service = sport;
1819	break;
1820	}
1821	#endif
1822	Log2(("service : %d\n", service));
1823	return htons(service);
1824	}
1825
1826	void slirp_set_dhcp_TFTP_prefix(PNATState pData, const char *tftpPrefix)
1827	{
1828	Log2(("tftp_prefix:%s\n", tftpPrefix));
1829	tftp_prefix = tftpPrefix;
1830	}
1831
1832	void slirp_set_dhcp_TFTP_bootfile(PNATState pData, const char *bootFile)
1833	{
1834	Log2(("bootFile:%s\n", bootFile));
1835	bootp_filename = bootFile;
1836	}
1837
1838	void slirp_set_dhcp_next_server(PNATState pData, const char *next_server)
1839	{
1840	Log2(("next_server:%s\n", next_server));
1841	if (next_server == NULL)
1842	pData->tftp_server.s_addr = htonl(ntohl(special_addr.s_addr) \| CTL_TFTP);
1843	else
1844	inet_aton(next_server, &pData->tftp_server);
1845	}
1846	#ifdef VBOX_WITH_SLIRP_DNS_PROXY
1847	void slirp_set_dhcp_dns_proxy(PNATState pData, bool fDNSProxy)
1848	{
1849	Log2(("NAT: DNS proxy switched %s\n", (fDNSProxy ? "on" : "off")));
1850	pData->use_dns_proxy = fDNSProxy;
1851	}
1852	#endif
1853
1854	#define CHECK_ARG(name, val, lim_min, lim_max) \
1855	do { \
1856	if ((val) < (lim_min) \|\| (val) > (lim_max)) \
1857	{ \
1858	LogRel(("NAT: (" #name ":%d) has been ignored, " \
1859	"because out of range (%d, %d)\n", (val), (lim_min), (lim_max))); \
1860	return; \
1861	} \
1862	else \
1863	{ \
1864	LogRel(("NAT: (" #name ":%d)\n", (val))); \
1865	} \
1866	} while (0)
1867
1868	/* don't allow user set less 8kB and more than 1M values */
1869	#define _8K_1M_CHECK_ARG(name, val) CHECK_ARG(name, (val), 8, 1024)
1870	void slirp_set_rcvbuf(PNATState pData, int kilobytes)
1871	{
1872	_8K_1M_CHECK_ARG("SOCKET_RCVBUF", kilobytes);
1873	pData->socket_rcv = kilobytes;
1874	}
1875	void slirp_set_sndbuf(PNATState pData, int kilobytes)
1876	{
1877	_8K_1M_CHECK_ARG("SOCKET_SNDBUF", kilobytes);
1878	pData->socket_snd = kilobytes * _1K;
1879	}
1880	void slirp_set_tcp_rcvspace(PNATState pData, int kilobytes)
1881	{
1882	_8K_1M_CHECK_ARG("TCP_RCVSPACE", kilobytes);
1883	tcp_rcvspace = kilobytes * _1K;
1884	}
1885	void slirp_set_tcp_sndspace(PNATState pData, int kilobytes)
1886	{
1887	_8K_1M_CHECK_ARG("TCP_SNDSPACE", kilobytes);
1888	tcp_sndspace = kilobytes * _1K;
1889	}
1890
1891	#ifdef VBOX_WITH_STATISTICS
1892	/definitions /
1893	#define COUNTING_COUNTER(name, dsc) \
1894	void slirp_counting_counter_##name##_reset(PNATState); \
1895	void slirp_counting_counter_##name##_inc(PNATState); \
1896	void slirp_counting_counter_##name##_add(PNATState,int ); \
1897	/* @todo think abaout it */
1898	#define PROFILE_COUNTER(name, dsc)
1899	#include "counters.h"
1900	#undef COUNTING_COUNTER
1901	#undef PROFILE_COUNTER
1902
1903	/declarations/
1904	#define COUNTING_COUNTER(name, dsc) \
1905	void slirp_counting_counter_##name##_reset(PNATState pData) \
1906	{ \
1907	SLIRP_COUNTER_RESET(name); \
1908	} \
1909	void slirp_counting_counter_##name##_inc(PNATState pData) \
1910	{ \
1911	SLIRP_COUNTER_INC(name); \
1912	} \
1913	void slirp_counting_counter_##name##_add(PNATState pData,int val) \
1914	{ \
1915	SLIRP_COUNTER_ADD(name, val); \
1916	}
1917	/* @todo think abaout it */
1918	#define PROFILE_COUNTER(name, dsc)
1919	#include "counters.h"
1920	#undef COUNTING_COUNTER
1921	#undef PROFILE_COUNTER
1922	#endif

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

source: vbox/trunk/src/VBox/Devices/Network/slirp/slirp.c@ 20703

Download in other formats: