slirp.c@ 20716

Last change on this file since 20716 was 20713, checked in by vboxsync, 16 years ago
DrvNAT,slirp: Even simpler.
Property svn:eol-style set to `native`
File size: 58.4 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	* Register statistics.
715	*/
716	void slirp_register_statistics(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%u/" #name, \
728	pDrvIns->iInstance); \
729	} while (0)
730
731	# define PROFILE_COUNTER(name, dsc) COUNTER(name, STAMTYPE_PROFILE, STAMUNIT_TICKS_PER_CALL, dsc)
732	# define COUNTING_COUNTER(name, dsc) COUNTER(name, STAMTYPE_COUNTER, STAMUNIT_COUNT, dsc)
733
734	# include "counters.h"
735
736	# undef COUNTER
737	# undef PROFILE_COUNTER
738	# undef COUNTING_COUNTER
739	#endif /* VBOX_WITH_STATISTICS */
740	}
741
742	/**
743	* Deregister statistics.
744	*/
745	void slirp_deregister_statistics(PNATState pData, PPDMDRVINS pDrvIns)
746	{
747	#ifdef VBOX_WITH_STATISTICS
748	# define PROFILE_COUNTER(name, dsc) PDMDrvHlpSTAMDeregister(pDrvIns, &pData->Stat ## name)
749	# define COUNTING_COUNTER(name, dsc) PDMDrvHlpSTAMDeregister(pDrvIns, &pData->Stat ## name)
750
751	# include "counters.h"
752
753	# undef COUNTING_COUNTER
754	# undef PROFILE_COUNTER
755	#endif /* VBOX_WITH_STATISTICS */
756	}
757
758	/**
759	* Marks the link as up, making it possible to establish new connections.
760	*/
761	void slirp_link_up(PNATState pData)
762	{
763	link_up = 1;
764	}
765
766	/**
767	* Marks the link as down and cleans up the current connections.
768	*/
769	void slirp_link_down(PNATState pData)
770	{
771	struct socket *so;
772
773	while ((so = tcb.so_next) != &tcb)
774	{
775	if (so->so_state & SS_NOFDREF \|\| so->s == -1)
776	sofree(pData, so);
777	else
778	tcp_drop(pData, sototcpcb(so), 0);
779	}
780
781	while ((so = udb.so_next) != &udb)
782	udp_detach(pData, so);
783
784	link_up = 0;
785	}
786
787	/**
788	* Terminates the slirp component.
789	*/
790	void slirp_term(PNATState pData)
791	{
792	#ifndef VBOX_WITH_MULTI_DNS
793	if (pData->pszDomain)
794	RTStrFree((char )(void )pData->pszDomain);
795	#endif
796
797	#ifdef RT_OS_WINDOWS
798	pData->pfIcmpCloseHandle(pData->icmp_socket.sh);
799	FreeLibrary(pData->hmIcmpLibrary);
800	RTMemFree(pData->pvIcmpBuffer);
801	# else
802	closesocket(pData->icmp_socket.s);
803	#endif
804
805	slirp_link_down(pData);
806	#ifdef VBOX_WITH_MULTI_DNS
807	slirp_release_dns_list(pData);
808	#endif
809	#ifdef RT_OS_WINDOWS
810	WSACleanup();
811	#endif
812	#ifdef VBOX_WITH_SLIRP_ALIAS
813	while(!LIST_EMPTY(&instancehead)) {
814	struct libalias *la = LIST_FIRST(&instancehead);
815	/* libalias do all clean up */
816	LibAliasUninit(la);
817	}
818	#endif
819	#ifdef LOG_ENABLED
820	Log(("\n"
821	"NAT statistics\n"
822	"--------------\n"
823	"\n"));
824	ipstats(pData);
825	tcpstats(pData);
826	udpstats(pData);
827	icmpstats(pData);
828	mbufstats(pData);
829	sockstats(pData);
830	Log(("\n"
831	"\n"
832	"\n"));
833	#endif
834	RTMemFree(pData);
835	}
836
837
838	#define CONN_CANFSEND(so) (((so)->so_state & (SS_FCANTSENDMORE\|SS_ISFCONNECTED)) == SS_ISFCONNECTED)
839	#define CONN_CANFRCV(so) (((so)->so_state & (SS_FCANTRCVMORE\|SS_ISFCONNECTED)) == SS_ISFCONNECTED)
840	#define UPD_NFDS(x) do { if (nfds < (x)) nfds = (x); } while (0)
841
842	/*
843	* curtime kept to an accuracy of 1ms
844	*/
845	static void updtime(PNATState pData)
846	{
847	#ifdef RT_OS_WINDOWS
848	struct _timeb tb;
849
850	_ftime(&tb);
851	curtime = (u_int)tb.time * (u_int)1000;
852	curtime += (u_int)tb.millitm;
853	#else
854	gettimeofday(&tt, 0);
855
856	curtime = (u_int)tt.tv_sec * (u_int)1000;
857	curtime += (u_int)tt.tv_usec / (u_int)1000;
858
859	if ((tt.tv_usec % 1000) >= 500)
860	curtime++;
861	#endif
862	}
863
864	#ifdef RT_OS_WINDOWS
865	void slirp_select_fill(PNATState pData, int *pnfds)
866	#else /* RT_OS_WINDOWS */
867	void slirp_select_fill(PNATState pData, int pnfds, struct pollfd polls)
868	#endif /* !RT_OS_WINDOWS */
869	{
870	struct socket so, so_next;
871	int nfds;
872	#if defined(RT_OS_WINDOWS)
873	int rc;
874	int error;
875	#else
876	int poll_index = 0;
877	#endif
878	int i;
879
880	STAM_PROFILE_START(&pData->StatFill, a);
881
882	nfds = *pnfds;
883
884	/*
885	* First, TCP sockets
886	*/
887	do_slowtimo = 0;
888	if (!link_up)
889	goto done;
890	/*
891	* *_slowtimo needs calling if there are IP fragments
892	* in the fragment queue, or there are TCP connections active
893	*/
894	/* XXX:
895	* triggering of fragment expiration should be the same but use new macroses
896	*/
897	do_slowtimo = (tcb.so_next != &tcb);
898	if (!do_slowtimo)
899	{
900	for (i = 0; i < IPREASS_NHASH; i++)
901	{
902	if (!TAILQ_EMPTY(&ipq[i]))
903	{
904	do_slowtimo = 1;
905	break;
906	}
907	}
908	}
909	ICMP_ENGAGE_EVENT(&pData->icmp_socket, readfds);
910
911	STAM_COUNTER_RESET(&pData->StatTCP);
912	STAM_COUNTER_RESET(&pData->StatTCPHot);
913
914	QSOCKET_FOREACH(so, so_next, tcp)
915	/* { */
916	#if !defined(RT_OS_WINDOWS)
917	so->so_poll_index = -1;
918	#endif
919	STAM_COUNTER_INC(&pData->StatTCP);
920
921	/*
922	* See if we need a tcp_fasttimo
923	*/
924	if ( time_fasttimo == 0
925	&& so->so_tcpcb != NULL
926	&& so->so_tcpcb->t_flags & TF_DELACK)
927	time_fasttimo = curtime; /* Flag when we want a fasttimo */
928
929	/*
930	* NOFDREF can include still connecting to local-host,
931	* newly socreated() sockets etc. Don't want to select these.
932	*/
933	if (so->so_state & SS_NOFDREF \|\| so->s == -1)
934	CONTINUE(tcp);
935
936	/*
937	* Set for reading sockets which are accepting
938	*/
939	if (so->so_state & SS_FACCEPTCONN)
940	{
941	STAM_COUNTER_INC(&pData->StatTCPHot);
942	TCP_ENGAGE_EVENT1(so, readfds);
943	CONTINUE(tcp);
944	}
945
946	/*
947	* Set for writing sockets which are connecting
948	*/
949	if (so->so_state & SS_ISFCONNECTING)
950	{
951	Log2(("connecting %R[natsock] engaged\n",so));
952	STAM_COUNTER_INC(&pData->StatTCPHot);
953	TCP_ENGAGE_EVENT1(so, writefds);
954	}
955
956	/*
957	* Set for writing if we are connected, can send more, and
958	* we have something to send
959	*/
960	if (CONN_CANFSEND(so) && so->so_rcv.sb_cc)
961	{
962	STAM_COUNTER_INC(&pData->StatTCPHot);
963	TCP_ENGAGE_EVENT1(so, writefds);
964	}
965
966	/*
967	* Set for reading (and urgent data) if we are connected, can
968	* receive more, and we have room for it XXX /2 ?
969	*/
970	if (CONN_CANFRCV(so) && (so->so_snd.sb_cc < (so->so_snd.sb_datalen/2)))
971	{
972	STAM_COUNTER_INC(&pData->StatTCPHot);
973	TCP_ENGAGE_EVENT2(so, readfds, xfds);
974	}
975	LOOP_LABEL(tcp, so, so_next);
976	}
977
978	/*
979	* UDP sockets
980	*/
981	STAM_COUNTER_RESET(&pData->StatUDP);
982	STAM_COUNTER_RESET(&pData->StatUDPHot);
983
984	QSOCKET_FOREACH(so, so_next, udp)
985	/* { */
986
987	STAM_COUNTER_INC(&pData->StatUDP);
988	#if !defined(RT_OS_WINDOWS)
989	so->so_poll_index = -1;
990	#endif
991
992	/*
993	* See if it's timed out
994	*/
995	if (so->so_expire)
996	{
997	if (so->so_expire <= curtime)
998	{
999	#ifdef VBOX_WITH_SLIRP_DNS_PROXY
1000	Log2(("NAT: %R[natsock] expired\n", so));
1001	if (so->so_timeout != NULL)
1002	{
1003	so->so_timeout(pData, so, so->so_timeout_arg);
1004	}
1005	#endif
1006	#ifdef VBOX_WITH_SLIRP_MT
1007	/* we need so_next for continue our cycle*/
1008	so_next = so->so_next;
1009	#endif
1010	UDP_DETACH(pData, so, so_next);
1011	CONTINUE_NO_UNLOCK(udp);
1012	}
1013	else
1014	do_slowtimo = 1; /* Let socket expire */
1015	}
1016
1017	/*
1018	* When UDP packets are received from over the link, they're
1019	* sendto()'d straight away, so no need for setting for writing
1020	* Limit the number of packets queued by this session to 4.
1021	* Note that even though we try and limit this to 4 packets,
1022	* the session could have more queued if the packets needed
1023	* to be fragmented.
1024	*
1025	* (XXX <= 4 ?)
1026	*/
1027	if ((so->so_state & SS_ISFCONNECTED) && so->so_queued <= 4)
1028	{
1029	STAM_COUNTER_INC(&pData->StatUDPHot);
1030	UDP_ENGAGE_EVENT(so, readfds);
1031	}
1032	LOOP_LABEL(udp, so, so_next);
1033	}
1034	done:
1035
1036	#if defined(RT_OS_WINDOWS)
1037	*pnfds = VBOX_EVENT_COUNT;
1038	#else /* RT_OS_WINDOWS */
1039	AssertRelease(poll_index <= *pnfds);
1040	*pnfds = poll_index;
1041	#endif /* !RT_OS_WINDOWS */
1042
1043	STAM_PROFILE_STOP(&pData->StatFill, a);
1044	}
1045
1046	#if defined(RT_OS_WINDOWS)
1047	void slirp_select_poll(PNATState pData, int fTimeout, int fIcmp)
1048	#else /* RT_OS_WINDOWS */
1049	void slirp_select_poll(PNATState pData, struct pollfd *polls, int ndfs)
1050	#endif /* !RT_OS_WINDOWS */
1051	{
1052	struct socket so, so_next;
1053	int ret;
1054	#if defined(RT_OS_WINDOWS)
1055	WSANETWORKEVENTS NetworkEvents;
1056	int rc;
1057	int error;
1058	#else
1059	int poll_index = 0;
1060	#endif
1061
1062	STAM_PROFILE_START(&pData->StatPoll, a);
1063
1064	/* Update time */
1065	updtime(pData);
1066
1067	/*
1068	* See if anything has timed out
1069	*/
1070	if (link_up)
1071	{
1072	if (time_fasttimo && ((curtime - time_fasttimo) >= 2))
1073	{
1074	STAM_PROFILE_START(&pData->StatFastTimer, a);
1075	tcp_fasttimo(pData);
1076	time_fasttimo = 0;
1077	STAM_PROFILE_STOP(&pData->StatFastTimer, a);
1078	}
1079	if (do_slowtimo && ((curtime - last_slowtimo) >= 499))
1080	{
1081	STAM_PROFILE_START(&pData->StatSlowTimer, a);
1082	ip_slowtimo(pData);
1083	tcp_slowtimo(pData);
1084	last_slowtimo = curtime;
1085	STAM_PROFILE_STOP(&pData->StatSlowTimer, a);
1086	}
1087	}
1088	#if defined(RT_OS_WINDOWS)
1089	if (fTimeout)
1090	return; /* only timer update */
1091	#endif
1092
1093	/*
1094	* Check sockets
1095	*/
1096	if (!link_up)
1097	goto done;
1098	#if defined(RT_OS_WINDOWS)
1099	/*XXX: before renaming please make see define
1100	* fIcmp in slirp_state.h
1101	*/
1102	if (fIcmp)
1103	sorecvfrom(pData, &pData->icmp_socket);
1104	#else
1105	if ( (pData->icmp_socket.s != -1)
1106	&& CHECK_FD_SET(&pData->icmp_socket, ignored, readfds))
1107	sorecvfrom(pData, &pData->icmp_socket);
1108	#endif
1109	/*
1110	* Check TCP sockets
1111	*/
1112	QSOCKET_FOREACH(so, so_next, tcp)
1113	/* { */
1114
1115	#ifdef VBOX_WITH_SLIRP_MT
1116	if ( so->so_state & SS_NOFDREF
1117	&& so->so_deleted == 1)
1118	{
1119	struct socket son, sop = NULL;
1120	QSOCKET_LOCK(tcb);
1121	if (so->so_next != NULL)
1122	{
1123	if (so->so_next != &tcb)
1124	SOCKET_LOCK(so->so_next);
1125	son = so->so_next;
1126	}
1127	if ( so->so_prev != &tcb
1128	&& so->so_prev != NULL)
1129	{
1130	SOCKET_LOCK(so->so_prev);
1131	sop = so->so_prev;
1132	}
1133	QSOCKET_UNLOCK(tcb);
1134	remque(pData, so);
1135	NSOCK_DEC();
1136	SOCKET_UNLOCK(so);
1137	SOCKET_LOCK_DESTROY(so);
1138	RTMemFree(so);
1139	so_next = son;
1140	if (sop != NULL)
1141	SOCKET_UNLOCK(sop);
1142	CONTINUE_NO_UNLOCK(tcp);
1143	}
1144	#endif
1145	/*
1146	* FD_ISSET is meaningless on these sockets
1147	* (and they can crash the program)
1148	*/
1149	if (so->so_state & SS_NOFDREF \|\| so->s == -1)
1150	CONTINUE(tcp);
1151
1152	POLL_TCP_EVENTS(rc, error, so, &NetworkEvents);
1153
1154	LOG_NAT_SOCK(so, TCP, &NetworkEvents, readfds, writefds, xfds);
1155
1156
1157	/*
1158	* Check for URG data
1159	* This will soread as well, so no need to
1160	* test for readfds below if this succeeds
1161	*/
1162
1163	/* out-of-band data */
1164	if (CHECK_FD_SET(so, NetworkEvents, xfds))
1165	{
1166	sorecvoob(pData, so);
1167	}
1168
1169	/*
1170	* Check sockets for reading
1171	*/
1172	else if ( CHECK_FD_SET(so, NetworkEvents, readfds)
1173	\|\| WIN_CHECK_FD_SET(so, NetworkEvents, acceptds))
1174	{
1175	/*
1176	* Check for incoming connections
1177	*/
1178	if (so->so_state & SS_FACCEPTCONN)
1179	{
1180	TCP_CONNECT(pData, so);
1181	#if defined(RT_OS_WINDOWS)
1182	if (!(NetworkEvents.lNetworkEvents & FD_CLOSE))
1183	#endif
1184	CONTINUE(tcp);
1185	}
1186
1187	ret = soread(pData, so);
1188	/* Output it if we read something */
1189	if (RT_LIKELY(ret > 0))
1190	TCP_OUTPUT(pData, sototcpcb(so));
1191	}
1192
1193	#if defined(RT_OS_WINDOWS)
1194	/*
1195	* Check for FD_CLOSE events.
1196	* in some cases once FD_CLOSE engaged on socket it could be flashed latter (for some reasons)
1197	*/
1198	if ( (NetworkEvents.lNetworkEvents & FD_CLOSE)
1199	\|\| (so->so_close == 1))
1200	{
1201	so->so_close = 1; /* mark it */
1202	/*
1203	* drain the socket
1204	*/
1205	for (;;)
1206	{
1207	ret = soread(pData, so);
1208	if (ret > 0)
1209	TCP_OUTPUT(pData, sototcpcb(so));
1210	else
1211	break;
1212	}
1213	CONTINUE(tcp);
1214	}
1215	#endif
1216
1217	/*
1218	* Check sockets for writing
1219	*/
1220	if (CHECK_FD_SET(so, NetworkEvents, writefds))
1221	{
1222	/*
1223	* Check for non-blocking, still-connecting sockets
1224	*/
1225	if (so->so_state & SS_ISFCONNECTING)
1226	{
1227	Log2(("connecting %R[natsock] catched\n", so));
1228	/* Connected */
1229	so->so_state &= ~SS_ISFCONNECTING;
1230
1231	/*
1232	* This should be probably guarded by PROBE_CONN too. Anyway,
1233	* we disable it on OS/2 because the below send call returns
1234	* EFAULT which causes the opened TCP socket to close right
1235	* after it has been opened and connected.
1236	*/
1237	#ifndef RT_OS_OS2
1238	ret = send(so->s, (const char *)&ret, 0, 0);
1239	if (ret < 0)
1240	{
1241	/* XXXXX Must fix, zero bytes is a NOP */
1242	if ( errno == EAGAIN
1243	\|\| errno == EWOULDBLOCK
1244	\|\| errno == EINPROGRESS
1245	\|\| errno == ENOTCONN)
1246	CONTINUE(tcp);
1247
1248	/* else failed */
1249	so->so_state = SS_NOFDREF;
1250	}
1251	/* else so->so_state &= ~SS_ISFCONNECTING; */
1252	#endif
1253
1254	/*
1255	* Continue tcp_input
1256	*/
1257	TCP_INPUT(pData, (struct mbuf *)NULL, sizeof(struct ip), so);
1258	/* continue; */
1259	}
1260	else
1261	SOWRITE(ret, pData, so);
1262	/*
1263	* XXX If we wrote something (a lot), there could be the need
1264	* for a window update. In the worst case, the remote will send
1265	* a window probe to get things going again.
1266	*/
1267	}
1268
1269	/*
1270	* Probe a still-connecting, non-blocking socket
1271	* to check if it's still alive
1272	*/
1273	#ifdef PROBE_CONN
1274	if (so->so_state & SS_ISFCONNECTING)
1275	{
1276	ret = recv(so->s, (char *)&ret, 0, 0);
1277
1278	if (ret < 0)
1279	{
1280	/* XXX */
1281	if ( errno == EAGAIN
1282	\|\| errno == EWOULDBLOCK
1283	\|\| errno == EINPROGRESS
1284	\|\| errno == ENOTCONN)
1285	{
1286	CONTINUE(tcp); /* Still connecting, continue */
1287	}
1288
1289	/* else failed */
1290	so->so_state = SS_NOFDREF;
1291
1292	/* tcp_input will take care of it */
1293	}
1294	else
1295	{
1296	ret = send(so->s, &ret, 0, 0);
1297	if (ret < 0)
1298	{
1299	/* XXX */
1300	if ( errno == EAGAIN
1301	\|\| errno == EWOULDBLOCK
1302	\|\| errno == EINPROGRESS
1303	\|\| errno == ENOTCONN)
1304	{
1305	CONTINUE(tcp);
1306	}
1307	/* else failed */
1308	so->so_state = SS_NOFDREF;
1309	}
1310	else
1311	so->so_state &= ~SS_ISFCONNECTING;
1312
1313	}
1314	TCP_INPUT((struct mbuf *)NULL, sizeof(struct ip),so);
1315	} /* SS_ISFCONNECTING */
1316	#endif
1317	#ifndef RT_OS_WINDOWS
1318	if ( UNIX_CHECK_FD_SET(so, NetworkEvents, rdhup)
1319	\|\| UNIX_CHECK_FD_SET(so, NetworkEvents, rderr))
1320	{
1321	int err;
1322	int inq, outq;
1323	int status;
1324	socklen_t optlen = sizeof(int);
1325	inq = outq = 0;
1326	status = getsockopt(so->s, SOL_SOCKET, SO_ERROR, &err, &optlen);
1327	if (status != 0)
1328	Log(("NAT: can't get error status from %R[natsock]\n", so));
1329	#ifndef RT_OS_SOLARIS
1330	status = ioctl(so->s, FIONREAD, &inq); /* tcp(7) recommends SIOCINQ which is Linux specific */
1331	if (status != 0 \|\| status != EINVAL)
1332	{
1333	/* EINVAL returned if socket in listen state tcp(7)*/
1334	Log(("NAT: can't get depth of IN queue status from %R[natsock]\n", so));
1335	}
1336	status = ioctl(so->s, TIOCOUTQ, &outq); /* SIOCOUTQ see previous comment */
1337	if (status != 0)
1338	Log(("NAT: can't get depth of OUT queue from %R[natsock]\n", so));
1339	#else
1340	/*
1341	* Solaris has bit different ioctl commands and its handlings
1342	* hint: streamio(7) I_NREAD
1343	*/
1344	#endif
1345	if ( so->so_state & SS_ISFCONNECTING
1346	\|\| UNIX_CHECK_FD_SET(so, NetworkEvents, readfds))
1347	{
1348	/**
1349	* Check if we need here take care about gracefull connection
1350	* @todo try with proxy server
1351	*/
1352	if (UNIX_CHECK_FD_SET(so, NetworkEvents, readfds))
1353	{
1354	/*
1355	* Never meet inq != 0 or outq != 0, anyway let it stay for a while
1356	* in case it happens we'll able to detect it.
1357	* Give TCP/IP stack wait or expire the socket.
1358	*/
1359	Log(("NAT: %R[natsock] err(%d:%s) s(in:%d,out:%d)happens on read I/O, "
1360	"other side close connection \n", so, err, strerror(err), inq, outq));
1361	CONTINUE(tcp);
1362	}
1363	goto tcp_input_close;
1364	}
1365	if ( !UNIX_CHECK_FD_SET(so, NetworkEvents, readfds)
1366	&& !UNIX_CHECK_FD_SET(so, NetworkEvents, writefds)
1367	&& !UNIX_CHECK_FD_SET(so, NetworkEvents, xfds))
1368	{
1369	Log(("NAT: system expires the socket %R[natsock] err(%d:%s) s(in:%d,out:%d) happens on non-I/O. ",
1370	so, err, strerror(err), inq, outq));
1371	goto tcp_input_close;
1372	}
1373	Log(("NAT: %R[natsock] we've met(%d:%s) s(in:%d, out:%d) unhandled combination hup (%d) "
1374	"rederr(%d) on (r:%d, w:%d, x:%d)\n",
1375	so, err, strerror(err),
1376	inq, outq,
1377	UNIX_CHECK_FD_SET(so, ign, rdhup),
1378	UNIX_CHECK_FD_SET(so, ign, rderr),
1379	UNIX_CHECK_FD_SET(so, ign, readfds),
1380	UNIX_CHECK_FD_SET(so, ign, writefds),
1381	UNIX_CHECK_FD_SET(so, ign, xfds)));
1382	/*
1383	* Give OS's TCP/IP stack a chance to resolve an issue or expire the socket.
1384	*/
1385	CONTINUE(tcp);
1386	tcp_input_close:
1387	so->so_state = SS_NOFDREF; /cause connection valid tcp connection termination and socket closing /
1388	TCP_INPUT(pData, (struct mbuf *)NULL, sizeof(struct ip), so);
1389	CONTINUE(tcp);
1390	}
1391	#endif
1392	LOOP_LABEL(tcp, so, so_next);
1393	}
1394
1395	/*
1396	* Now UDP sockets.
1397	* Incoming packets are sent straight away, they're not buffered.
1398	* Incoming UDP data isn't buffered either.
1399	*/
1400	QSOCKET_FOREACH(so, so_next, udp)
1401	/* { */
1402	#ifdef VBOX_WITH_SLIRP_MT
1403	if ( so->so_state & SS_NOFDREF
1404	&& so->so_deleted == 1)
1405	{
1406	struct socket son, sop = NULL;
1407	QSOCKET_LOCK(udb);
1408	if (so->so_next != NULL)
1409	{
1410	if (so->so_next != &udb)
1411	SOCKET_LOCK(so->so_next);
1412	son = so->so_next;
1413	}
1414	if ( so->so_prev != &udb
1415	&& so->so_prev != NULL)
1416	{
1417	SOCKET_LOCK(so->so_prev);
1418	sop = so->so_prev;
1419	}
1420	QSOCKET_UNLOCK(udb);
1421	remque(pData, so);
1422	NSOCK_DEC();
1423	SOCKET_UNLOCK(so);
1424	SOCKET_LOCK_DESTROY(so);
1425	RTMemFree(so);
1426	so_next = son;
1427	if (sop != NULL)
1428	SOCKET_UNLOCK(sop);
1429	CONTINUE_NO_UNLOCK(udp);
1430	}
1431	#endif
1432	POLL_UDP_EVENTS(rc, error, so, &NetworkEvents);
1433
1434	LOG_NAT_SOCK(so, UDP, &NetworkEvents, readfds, writefds, xfds);
1435
1436	if (so->s != -1 && CHECK_FD_SET(so, NetworkEvents, readfds))
1437	{
1438	SORECVFROM(pData, so);
1439	}
1440	LOOP_LABEL(udp, so, so_next);
1441	}
1442
1443	done:
1444	#ifndef VBOX_WITH_SLIRP_MT
1445	/*
1446	* See if we can start outputting
1447	*/
1448	if (if_queued && link_up)
1449	if_start(pData);
1450	#endif
1451
1452	STAM_PROFILE_STOP(&pData->StatPoll, a);
1453	}
1454
1455	#ifndef VBOX_WITHOUT_SLIRP_CLIENT_ETHER
1456	#define ETH_ALEN 6
1457	#define ETH_HLEN 14
1458
1459	#define ARPOP_REQUEST 1 /* ARP request */
1460	#define ARPOP_REPLY 2 /* ARP reply */
1461
1462	struct ethhdr
1463	{
1464	unsigned char h_dest[ETH_ALEN]; /* destination eth addr */
1465	unsigned char h_source[ETH_ALEN]; /* source ether addr */
1466	unsigned short h_proto; /* packet type ID field */
1467	};
1468	AssertCompileSize(struct ethhdr, 14);
1469	#endif
1470
1471	struct arphdr
1472	{
1473	unsigned short ar_hrd; /* format of hardware address */
1474	unsigned short ar_pro; /* format of protocol address */
1475	unsigned char ar_hln; /* length of hardware address */
1476	unsigned char ar_pln; /* length of protocol address */
1477	unsigned short ar_op; /* ARP opcode (command) */
1478
1479	/*
1480	* Ethernet looks like this : This bit is variable sized however...
1481	*/
1482	unsigned char ar_sha[ETH_ALEN]; /* sender hardware address */
1483	unsigned char ar_sip[4]; /* sender IP address */
1484	unsigned char ar_tha[ETH_ALEN]; /* target hardware address */
1485	unsigned char ar_tip[4]; /* target IP address */
1486	};
1487	AssertCompileSize(struct arphdr, 28);
1488
1489	static void arp_input(PNATState pData, struct mbuf *m)
1490	{
1491	struct ethhdr *eh;
1492	struct ethhdr *reh;
1493	struct arphdr *ah;
1494	struct arphdr *rah;
1495	int ar_op;
1496	struct ex_list *ex_ptr;
1497	uint32_t htip;
1498	uint32_t tip;
1499	struct mbuf *mr;
1500	eh = mtod(m, struct ethhdr *);
1501	ah = (struct arphdr *)&eh[1];
1502	htip = ntohl((uint32_t)ah->ar_tip);
1503	tip = (uint32_t)ah->ar_tip;
1504
1505	mr = m_get(pData);
1506	#ifdef VBOX_WITHOUT_SLIRP_CLIENT_ETHER
1507	reh = mtod(mr, struct ethhdr *);
1508	memcpy(reh->h_source, eh->h_source, ETH_ALEN); /* XXX: if_encap will swap src and dst*/
1509	Log4(("NAT: arp:%R[ether]->%R[ether]\n",
1510	reh->h_source, reh->h_dest));
1511	Log4(("NAT: arp: %R[IP4]\n", &tip));
1512	#endif
1513	mr->m_data += if_maxlinkhdr;
1514	mr->m_len = sizeof(struct arphdr);
1515	rah = mtod(mr, struct arphdr *);
1516
1517	ar_op = ntohs(ah->ar_op);
1518	switch(ar_op)
1519	{
1520	case ARPOP_REQUEST:
1521	#ifdef VBOX_WITH_NAT_SERVICE
1522	if (tip == special_addr.s_addr) goto arp_ok;
1523	#endif
1524	if ((htip & pData->netmask) == ntohl(special_addr.s_addr))
1525	{
1526	if ( CTL_CHECK(htip, CTL_DNS)
1527	\|\| CTL_CHECK(htip, CTL_ALIAS)
1528	\|\| CTL_CHECK(htip, CTL_TFTP))
1529	goto arp_ok;
1530	for (ex_ptr = exec_list; ex_ptr; ex_ptr = ex_ptr->ex_next)
1531	{
1532	if ((htip & ~pData->netmask) == ex_ptr->ex_addr)
1533	{
1534	goto arp_ok;
1535	}
1536	}
1537	return;
1538	arp_ok:
1539	rah->ar_hrd = htons(1);
1540	rah->ar_pro = htons(ETH_P_IP);
1541	rah->ar_hln = ETH_ALEN;
1542	rah->ar_pln = 4;
1543	rah->ar_op = htons(ARPOP_REPLY);
1544	memcpy(rah->ar_sha, special_ethaddr, ETH_ALEN);
1545
1546	switch (htip & ~pData->netmask)
1547	{
1548	case CTL_DNS:
1549	case CTL_ALIAS:
1550	rah->ar_sha[5] = (uint8_t)(htip & ~pData->netmask);
1551	break;
1552	default:;
1553	}
1554
1555	memcpy(rah->ar_sip, ah->ar_tip, 4);
1556	memcpy(rah->ar_tha, ah->ar_sha, ETH_ALEN);
1557	memcpy(rah->ar_tip, ah->ar_sip, 4);
1558	if_encap(pData, ETH_P_ARP, mr);
1559	m_free(pData, m);
1560	}
1561	break;
1562	default:
1563	break;
1564	}
1565	}
1566
1567	void slirp_input(PNATState pData, const uint8_t *pkt, int pkt_len)
1568	{
1569	struct mbuf *m;
1570	int proto;
1571	static bool fWarnedIpv6;
1572	struct ethhdr eh = (struct ethhdr)pkt;
1573
1574	Log2(("NAT: slirp_input %d\n", pkt_len));
1575	if (pkt_len < ETH_HLEN)
1576	{
1577	LogRel(("NAT: packet having size %d has been ingnored\n", pkt_len));
1578	return;
1579	}
1580	Log4(("NAT: in:%R[ether]->%R[ether]\n", &eh->h_source, &eh->h_dest));
1581	#ifdef VBOX_WITHOUT_SLIRP_CLIENT_ETHER
1582	if (memcmp(eh->h_source, special_ethaddr, ETH_ALEN) == 0)
1583	{
1584	/* @todo vasily: add ether logging routine in debug.c */
1585	Log(("NAT: packet was addressed to other MAC\n"));
1586	RTMemFree((void *)pkt);
1587	return;
1588	}
1589	#endif
1590
1591	m = m_get(pData);
1592	if (!m)
1593	{
1594	LogRel(("NAT: can't allocate new mbuf\n"));
1595	return;
1596	}
1597
1598	/* Note: we add to align the IP header */
1599
1600	if (M_FREEROOM(m) < pkt_len)
1601	m_inc(m, pkt_len);
1602
1603	m->m_len = pkt_len ;
1604	memcpy(m->m_data, pkt, pkt_len);
1605
1606	proto = ntohs((uint16_t )(pkt + 12));
1607	switch(proto)
1608	{
1609	case ETH_P_ARP:
1610	arp_input(pData, m);
1611	break;
1612	case ETH_P_IP:
1613	/* Update time. Important if the network is very quiet, as otherwise
1614	* the first outgoing connection gets an incorrect timestamp. */
1615	updtime(pData);
1616	m_adj(m, ETH_HLEN);
1617	ip_input(pData, m);
1618	break;
1619	case ETH_P_IPV6:
1620	m_free(pData, m);
1621	if (!fWarnedIpv6)
1622	{
1623	LogRel(("NAT: IPv6 not supported\n"));
1624	fWarnedIpv6 = true;
1625	}
1626	break;
1627	default:
1628	Log(("NAT: Unsupported protocol %x\n", proto));
1629	m_free(pData, m);
1630	break;
1631	}
1632	RTMemFree((void *)pkt);
1633	}
1634
1635	/* output the IP packet to the ethernet device */
1636	void if_encap(PNATState pData, uint16_t eth_proto, struct mbuf *m)
1637	{
1638	struct ethhdr *eh;
1639	uint8_t *buf = RTMemAlloc(1600);
1640
1641	m->m_data -= if_maxlinkhdr;
1642	m->m_len += ETH_HLEN;
1643	eh = mtod(m, struct ethhdr *);
1644
1645	if(MBUF_HEAD(m) != m->m_data)
1646	{
1647	LogRel(("NAT: ethernet detects corruption of the packet"));
1648	AssertMsgFailed(("!!Ethernet frame corrupted!!"));
1649	}
1650
1651	#ifndef VBOX_WITHOUT_SLIRP_CLIENT_ETHER
1652	memcpy(eh->h_dest, client_ethaddr, ETH_ALEN);
1653	memcpy(eh->h_source, special_ethaddr, ETH_ALEN - 1);
1654	/* XXX: not correct */
1655	eh->h_source[5] = CTL_ALIAS;
1656	#else
1657	if (memcmp(eh->h_source, special_ethaddr, ETH_ALEN) != 0)
1658	{
1659	memcpy(eh->h_dest, eh->h_source, ETH_ALEN);
1660	memcpy(eh->h_source, special_ethaddr, ETH_ALEN);
1661	Assert(memcmp(eh->h_dest, special_ethaddr, ETH_ALEN) != 0);
1662	if (memcmp(eh->h_dest, zerro_ethaddr, ETH_ALEN) == 0)
1663	{
1664	/* don't do anything */
1665	goto done;
1666	}
1667	}
1668	#endif
1669	eh->h_proto = htons(eth_proto);
1670	#if 0
1671	slirp_output(pData->pvUser, m, mtod(m, uint8_t *), m->m_len);
1672	#else
1673	memcpy(buf, mtod(m, uint8_t *), m->m_len);
1674	slirp_output(pData->pvUser, NULL, buf, m->m_len);
1675	done:
1676	m_free(pData, m);
1677	#endif
1678	}
1679
1680	int slirp_redir(PNATState pData, int is_udp, int host_port,
1681	struct in_addr guest_addr, int guest_port)
1682	{
1683	struct socket *so;
1684	#ifdef VBOX_WITH_SLIRP_ALIAS
1685	struct alias_link *link;
1686	struct libalias *lib;
1687	int flags;
1688	struct sockaddr sa;
1689	struct sockaddr_in *psin;
1690	socklen_t socketlen;
1691	struct in_addr alias;
1692	int rc;
1693	#endif
1694	Log2(("NAT: set redirect %s hp:%d gp:%d\n", (is_udp?"UDP":"TCP"), host_port, guest_port));
1695	if (is_udp)
1696	{
1697	so = udp_listen(pData, htons(host_port), guest_addr.s_addr,
1698	htons(guest_port), 0);
1699	}
1700	else
1701	{
1702	so = solisten(pData, htons(host_port), guest_addr.s_addr,
1703	htons(guest_port), 0);
1704	}
1705	#ifndef VBOX_WITH_SLIRP_ALIAS
1706	Log2(("NAT: redirecting socket %R[natsock]\n", so));
1707	return (so != NULL ? 0 : -1);
1708	#else
1709
1710	psin = (struct sockaddr_in *)&sa;
1711	psin->sin_family = AF_INET;
1712	psin->sin_port = 0;
1713	psin->sin_addr.s_addr = INADDR_ANY;
1714	socketlen = sizeof(struct sockaddr);
1715
1716	rc = getsockname(so->s, &sa, &socketlen);
1717	if (rc < 0 \|\| sa.sa_family != AF_INET)
1718	{
1719	Log(("NAT: can't get socket's name\n"));
1720	return 1;
1721	}
1722
1723	psin = (struct sockaddr_in *)&sa;
1724
1725	#if 1
1726	lib = LibAliasInit(pData, NULL);
1727	flags = LibAliasSetMode(lib, 0, 0);
1728	flags \|= PKT_ALIAS_LOG; /* set logging */
1729	flags \|= PKT_ALIAS_REVERSE; /* set logging */
1730	flags = LibAliasSetMode(lib, flags, ~0);
1731	#else
1732	lib = LIST_FIRST(&instancehead);
1733	#endif
1734
1735	alias.s_addr = htonl(ntohl(guest_addr.s_addr) \| CTL_ALIAS);
1736	link = LibAliasRedirectPort(lib, psin->sin_addr, htons(host_port),
1737	alias, htons(guest_port),
1738	special_addr, -1, /* not very clear for now*/
1739	(is_udp ? IPPROTO_UDP : IPPROTO_TCP));
1740	if (link == NULL)
1741	{
1742	Log(("NAT: can't create redirect\n"));
1743	return 1;
1744	}
1745	so->so_la = lib;
1746
1747	return 0;
1748	#endif
1749	}
1750
1751	int slirp_add_exec(PNATState pData, int do_pty, const char *args, int addr_low_byte,
1752	int guest_port)
1753	{
1754	return add_exec(&exec_list, do_pty, (char *)args,
1755	addr_low_byte, htons(guest_port));
1756	}
1757
1758	void slirp_set_ethaddr(PNATState pData, const uint8_t *ethaddr)
1759	{
1760	#ifndef VBOX_WITHOUT_SLIRP_CLIENT_ETHER
1761	memcpy(client_ethaddr, ethaddr, ETH_ALEN);
1762	#endif
1763	}
1764
1765	#if defined(RT_OS_WINDOWS)
1766	HANDLE *slirp_get_events(PNATState pData)
1767	{
1768	return pData->phEvents;
1769	}
1770	void slirp_register_external_event(PNATState pData, HANDLE hEvent, int index)
1771	{
1772	pData->phEvents[index] = hEvent;
1773	}
1774	#endif
1775
1776	unsigned int slirp_get_timeout_ms(PNATState pData)
1777	{
1778	if (link_up)
1779	{
1780	if (time_fasttimo)
1781	return 2;
1782	if (do_slowtimo)
1783	return 500; /* see PR_SLOWHZ */
1784	}
1785	return 0;
1786	}
1787
1788	#ifndef RT_OS_WINDOWS
1789	int slirp_get_nsock(PNATState pData)
1790	{
1791	return pData->nsock;
1792	}
1793	#endif
1794
1795	/*
1796	* this function called from NAT thread
1797	*/
1798	void slirp_post_sent(PNATState pData, void *pvArg)
1799	{
1800	struct socket *so = 0;
1801	struct tcpcb *tp = 0;
1802	struct mbuf m = (struct mbuf )pvArg;
1803	m_free(pData, m);
1804	}
1805	#ifdef VBOX_WITH_SLIRP_MT
1806	void slirp_process_queue(PNATState pData)
1807	{
1808	RTReqProcess(pData->pReqQueue, RT_INDEFINITE_WAIT);
1809	}
1810	void *slirp_get_queue(PNATState pData)
1811	{
1812	return pData->pReqQueue;
1813	}
1814	#endif
1815
1816	uint16_t slirp_get_service(int proto, uint16_t dport, uint16_t sport)
1817	{
1818	uint16_t hdport, hsport, service;
1819	hdport = ntohs(dport);
1820	hsport = ntohs(sport);
1821	Log2(("proto: %d, dport: %d sport: %d\n", proto, hdport, hsport));
1822	service = 0;
1823	#if 0
1824	/* Always return 0 here */
1825	switch (hdport)
1826	{
1827	case 500:
1828	if (hsport != 500) /* vpnc by default try operate in src:500/dst:500 mode*/
1829	/* Not sure why this make Cisco VPN client's connection more stable,
1830	* at least on some servers
1831	*/
1832	service = sport;
1833	break;
1834	}
1835	#endif
1836	Log2(("service : %d\n", service));
1837	return htons(service);
1838	}
1839
1840	void slirp_set_dhcp_TFTP_prefix(PNATState pData, const char *tftpPrefix)
1841	{
1842	Log2(("tftp_prefix:%s\n", tftpPrefix));
1843	tftp_prefix = tftpPrefix;
1844	}
1845
1846	void slirp_set_dhcp_TFTP_bootfile(PNATState pData, const char *bootFile)
1847	{
1848	Log2(("bootFile:%s\n", bootFile));
1849	bootp_filename = bootFile;
1850	}
1851
1852	void slirp_set_dhcp_next_server(PNATState pData, const char *next_server)
1853	{
1854	Log2(("next_server:%s\n", next_server));
1855	if (next_server == NULL)
1856	pData->tftp_server.s_addr = htonl(ntohl(special_addr.s_addr) \| CTL_TFTP);
1857	else
1858	inet_aton(next_server, &pData->tftp_server);
1859	}
1860	#ifdef VBOX_WITH_SLIRP_DNS_PROXY
1861	void slirp_set_dhcp_dns_proxy(PNATState pData, bool fDNSProxy)
1862	{
1863	Log2(("NAT: DNS proxy switched %s\n", (fDNSProxy ? "on" : "off")));
1864	pData->use_dns_proxy = fDNSProxy;
1865	}
1866	#endif
1867
1868	#define CHECK_ARG(name, val, lim_min, lim_max) \
1869	do { \
1870	if ((val) < (lim_min) \|\| (val) > (lim_max)) \
1871	{ \
1872	LogRel(("NAT: (" #name ":%d) has been ignored, " \
1873	"because out of range (%d, %d)\n", (val), (lim_min), (lim_max))); \
1874	return; \
1875	} \
1876	else \
1877	{ \
1878	LogRel(("NAT: (" #name ":%d)\n", (val))); \
1879	} \
1880	} while (0)
1881
1882	/* don't allow user set less 8kB and more than 1M values */
1883	#define _8K_1M_CHECK_ARG(name, val) CHECK_ARG(name, (val), 8, 1024)
1884	void slirp_set_rcvbuf(PNATState pData, int kilobytes)
1885	{
1886	_8K_1M_CHECK_ARG("SOCKET_RCVBUF", kilobytes);
1887	pData->socket_rcv = kilobytes;
1888	}
1889	void slirp_set_sndbuf(PNATState pData, int kilobytes)
1890	{
1891	_8K_1M_CHECK_ARG("SOCKET_SNDBUF", kilobytes);
1892	pData->socket_snd = kilobytes * _1K;
1893	}
1894	void slirp_set_tcp_rcvspace(PNATState pData, int kilobytes)
1895	{
1896	_8K_1M_CHECK_ARG("TCP_RCVSPACE", kilobytes);
1897	tcp_rcvspace = kilobytes * _1K;
1898	}
1899	void slirp_set_tcp_sndspace(PNATState pData, int kilobytes)
1900	{
1901	_8K_1M_CHECK_ARG("TCP_SNDSPACE", kilobytes);
1902	tcp_sndspace = kilobytes * _1K;
1903	}
1904

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

source: vbox/trunk/src/VBox/Devices/Network/slirp/slirp.c@ 20716

Download in other formats: