slirp.c@ 20376

Last change on this file since 20376 was 20376, checked in by vboxsync, 15 years ago
NAT: slirp performance counters
Property svn:eol-style set to `native`
File size: 56.1 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	#ifdef RT_OS_WINDOWS
636	{
637	WSADATA Data;
638	WSAStartup(MAKEWORD(2, 0), &Data);
639	}
640	pData->phEvents[VBOX_SOCKET_EVENT_INDEX] = CreateEvent(NULL, FALSE, FALSE, NULL);
641	#endif
642	#ifdef VBOX_WITH_SLIRP_MT
643	QSOCKET_LOCK_CREATE(tcb);
644	QSOCKET_LOCK_CREATE(udb);
645	rc = RTReqCreateQueue(&pData->pReqQueue);
646	AssertReleaseRC(rc);
647	#endif
648
649	link_up = 1;
650
651	debug_init();
652	if_init(pData);
653	ip_init(pData);
654	icmp_init(pData);
655
656	/* Initialise mbufs after setting the MTU */
657	m_init(pData);
658
659	#ifndef VBOX_WITH_NAT_SERVICE
660	inet_aton(pszNetAddr, &special_addr);
661	#else
662	special_addr.s_addr = u32NetAddr;
663	#endif
664	#ifdef VBOX_WITHOUT_SLIRP_CLIENT_ETHER
665	pData->slirp_ethaddr = &special_ethaddr[0];
666	#endif
667	alias_addr.s_addr = special_addr.s_addr \| htonl(CTL_ALIAS);
668	/* @todo: add ability to configure this staff */
669
670	/* set default addresses */
671	inet_aton("127.0.0.1", &loopback_addr);
672	#ifndef VBOX_WITH_MULTI_DNS
673	inet_aton("127.0.0.1", &dns_addr);
674
675	if (get_dns_addr_domain(pData, true, &dns_addr, &pData->pszDomain) < 0)
676	#else
677	if (slirp_init_dns_list(pData) < 0)
678	#endif
679	fNATfailed = 1;
680	#ifdef VBOX_WITH_SLIRP_DNS_PROXY
681	dnsproxy_init(pData);
682	#endif
683
684	getouraddr(pData);
685
686	#ifdef VBOX_WITH_SLIRP_ALIAS
687	{
688	int flags = 0;
689	pData->proxy_alias = LibAliasInit(pData, NULL);
690	if (pData->proxy_alias == NULL)
691	{
692	LogRel(("NAT: LibAlias default rule wasn't initialized\n"));
693	AssertMsgFailed(("NAT: LibAlias default rule wasn't initialized\n"));
694	}
695	flags = LibAliasSetMode(pData->proxy_alias, 0, 0);
696	flags \|= PKT_ALIAS_LOG; /* set logging */
697	flags = LibAliasSetMode(pData->proxy_alias, flags, ~0);
698	LibAliasSetAddress(pData->proxy_alias, special_addr);
699	ftp_alias_load();
700	nbt_alias_load();
701
702	}
703	#endif
704	return fNATfailed ? VINF_NAT_DNS : VINF_SUCCESS;
705	}
706
707	/**
708	* Statistics counters.
709	*/
710	void slirp_register_timers(PNATState pData, PPDMDRVINS pDrvIns)
711	{
712	#ifdef VBOX_WITH_STATISTICS
713	# define COUNTER(name, type, units, dsc) \
714	do{ \
715	PDMDrvHlpSTAMRegisterF(pDrvIns, \
716	&pData->Stat ## name, \
717	type, \
718	STAMVISIBILITY_ALWAYS, \
719	units, \
720	dsc, \
721	"/Drivers/NAT%d/Stat" #name, pDrvIns->iInstance); \
722	}while(0)
723
724	# define PROFILE_COUNTER(name, dsc) \
725	COUNTER(name, STAMTYPE_PROFILE, STAMUNIT_TICKS_PER_CALL, dsc)
726	# define COUNTING_COUTER(name, dsc) \
727	COUNTER(name, STAMTYPE_COUNTER, STAMUNIT_COUNT, dsc)
728
729	#include "counters.h"
730
731	#undef COUNTER
732	#undef PROFILE_COUNTER
733	#undef COUNTING_COUTER
734
735	#endif /* VBOX_WITH_STATISTICS */
736	}
737
738	/**
739	* Marks the link as up, making it possible to establish new connections.
740	*/
741	void slirp_link_up(PNATState pData)
742	{
743	link_up = 1;
744	}
745
746	/**
747	* Marks the link as down and cleans up the current connections.
748	*/
749	void slirp_link_down(PNATState pData)
750	{
751	struct socket *so;
752
753	while ((so = tcb.so_next) != &tcb)
754	{
755	if (so->so_state & SS_NOFDREF \|\| so->s == -1)
756	sofree(pData, so);
757	else
758	tcp_drop(pData, sototcpcb(so), 0);
759	}
760
761	while ((so = udb.so_next) != &udb)
762	udp_detach(pData, so);
763
764	link_up = 0;
765	}
766
767	/**
768	* Terminates the slirp component.
769	*/
770	void slirp_term(PNATState pData)
771	{
772	#ifndef VBOX_WITH_MULTI_DNS
773	if (pData->pszDomain)
774	RTStrFree((char )(void )pData->pszDomain);
775	#endif
776
777	#ifdef RT_OS_WINDOWS
778	pData->pfIcmpCloseHandle(pData->icmp_socket.sh);
779	FreeLibrary(pData->hmIcmpLibrary);
780	RTMemFree(pData->pvIcmpBuffer);
781	# else
782	closesocket(pData->icmp_socket.s);
783	#endif
784
785	slirp_link_down(pData);
786	#ifdef VBOX_WITH_MULTI_DNS
787	slirp_release_dns_list(pData);
788	#endif
789	#ifdef RT_OS_WINDOWS
790	WSACleanup();
791	#endif
792	#ifdef VBOX_WITH_SLIRP_ALIAS
793	while(!LIST_EMPTY(&instancehead)) {
794	struct libalias *la = LIST_FIRST(&instancehead);
795	/* libalias do all clean up */
796	LibAliasUninit(la);
797	}
798	#endif
799	#ifdef LOG_ENABLED
800	Log(("\n"
801	"NAT statistics\n"
802	"--------------\n"
803	"\n"));
804	ipstats(pData);
805	tcpstats(pData);
806	udpstats(pData);
807	icmpstats(pData);
808	mbufstats(pData);
809	sockstats(pData);
810	Log(("\n"
811	"\n"
812	"\n"));
813	#endif
814	RTMemFree(pData);
815	}
816
817
818	#define CONN_CANFSEND(so) (((so)->so_state & (SS_FCANTSENDMORE\|SS_ISFCONNECTED)) == SS_ISFCONNECTED)
819	#define CONN_CANFRCV(so) (((so)->so_state & (SS_FCANTRCVMORE\|SS_ISFCONNECTED)) == SS_ISFCONNECTED)
820	#define UPD_NFDS(x) do { if (nfds < (x)) nfds = (x); } while (0)
821
822	/*
823	* curtime kept to an accuracy of 1ms
824	*/
825	static void updtime(PNATState pData)
826	{
827	#ifdef RT_OS_WINDOWS
828	struct _timeb tb;
829
830	_ftime(&tb);
831	curtime = (u_int)tb.time * (u_int)1000;
832	curtime += (u_int)tb.millitm;
833	#else
834	gettimeofday(&tt, 0);
835
836	curtime = (u_int)tt.tv_sec * (u_int)1000;
837	curtime += (u_int)tt.tv_usec / (u_int)1000;
838
839	if ((tt.tv_usec % 1000) >= 500)
840	curtime++;
841	#endif
842	}
843
844	#ifdef RT_OS_WINDOWS
845	void slirp_select_fill(PNATState pData, int *pnfds)
846	#else /* RT_OS_WINDOWS */
847	void slirp_select_fill(PNATState pData, int pnfds, struct pollfd polls)
848	#endif /* !RT_OS_WINDOWS */
849	{
850	struct socket so, so_next;
851	int nfds;
852	#if defined(RT_OS_WINDOWS)
853	int rc;
854	int error;
855	#else
856	int poll_index = 0;
857	#endif
858	int i;
859
860	SLIRP_PROFILE_START(Fill, a);
861
862	nfds = *pnfds;
863
864	/*
865	* First, TCP sockets
866	*/
867	do_slowtimo = 0;
868	if (!link_up)
869	goto done;
870	/*
871	* *_slowtimo needs calling if there are IP fragments
872	* in the fragment queue, or there are TCP connections active
873	*/
874	/* XXX:
875	* triggering of fragment expiration should be the same but use new macroses
876	*/
877	do_slowtimo = (tcb.so_next != &tcb);
878	if (!do_slowtimo)
879	{
880	for (i = 0; i < IPREASS_NHASH; i++)
881	{
882	if (!TAILQ_EMPTY(&ipq[i]))
883	{
884	do_slowtimo = 1;
885	break;
886	}
887	}
888	}
889	ICMP_ENGAGE_EVENT(&pData->icmp_socket, readfds);
890
891	SLIRP_COUNTER_RESET(TCP);
892	SLIRP_COUNTER_RESET(TCPHot);
893
894	QSOCKET_FOREACH(so, so_next, tcp)
895	/* { */
896	#if !defined(RT_OS_WINDOWS)
897	so->so_poll_index = -1;
898	#endif
899	SLIRP_COUNTER_INC(TCP);
900
901	/*
902	* See if we need a tcp_fasttimo
903	*/
904	if ( time_fasttimo == 0
905	&& so->so_tcpcb != NULL
906	&& so->so_tcpcb->t_flags & TF_DELACK)
907	time_fasttimo = curtime; /* Flag when we want a fasttimo */
908
909	/*
910	* NOFDREF can include still connecting to local-host,
911	* newly socreated() sockets etc. Don't want to select these.
912	*/
913	if (so->so_state & SS_NOFDREF \|\| so->s == -1)
914	CONTINUE(tcp);
915
916	/*
917	* Set for reading sockets which are accepting
918	*/
919	if (so->so_state & SS_FACCEPTCONN)
920	{
921	SLIRP_COUNTER_INC(TCPHot);
922	TCP_ENGAGE_EVENT1(so, readfds);
923	CONTINUE(tcp);
924	}
925
926	/*
927	* Set for writing sockets which are connecting
928	*/
929	if (so->so_state & SS_ISFCONNECTING)
930	{
931	Log2(("connecting %R[natsock] engaged\n",so));
932	SLIRP_COUNTER_INC(TCPHot);
933	TCP_ENGAGE_EVENT1(so, writefds);
934	}
935
936	/*
937	* Set for writing if we are connected, can send more, and
938	* we have something to send
939	*/
940	if (CONN_CANFSEND(so) && so->so_rcv.sb_cc)
941	{
942	SLIRP_COUNTER_INC(TCPHot);
943	TCP_ENGAGE_EVENT1(so, writefds);
944	}
945
946	/*
947	* Set for reading (and urgent data) if we are connected, can
948	* receive more, and we have room for it XXX /2 ?
949	*/
950	if (CONN_CANFRCV(so) && (so->so_snd.sb_cc < (so->so_snd.sb_datalen/2)))
951	{
952	SLIRP_COUNTER_INC(TCPHot);
953	TCP_ENGAGE_EVENT2(so, readfds, xfds);
954	}
955	LOOP_LABEL(tcp, so, so_next);
956	}
957
958	/*
959	* UDP sockets
960	*/
961	SLIRP_COUNTER_RESET(UDP);
962	SLIRP_COUNTER_RESET(UDPHot);
963
964	QSOCKET_FOREACH(so, so_next, udp)
965	/* { */
966
967	SLIRP_COUNTER_INC(UDP);
968	#if !defined(RT_OS_WINDOWS)
969	so->so_poll_index = -1;
970	#endif
971
972	/*
973	* See if it's timed out
974	*/
975	if (so->so_expire)
976	{
977	if (so->so_expire <= curtime)
978	{
979	#ifdef VBOX_WITH_SLIRP_DNS_PROXY
980	Log2(("NAT: %R[natsock] expired\n", so));
981	if (so->so_timeout != NULL)
982	{
983	so->so_timeout(pData, so, so->so_timeout_arg);
984	}
985	#endif
986	#ifdef VBOX_WITH_SLIRP_MT
987	/* we need so_next for continue our cycle*/
988	so_next = so->so_next;
989	#endif
990	UDP_DETACH(pData, so, so_next);
991	CONTINUE_NO_UNLOCK(udp);
992	}
993	else
994	do_slowtimo = 1; /* Let socket expire */
995	}
996
997	/*
998	* When UDP packets are received from over the link, they're
999	* sendto()'d straight away, so no need for setting for writing
1000	* Limit the number of packets queued by this session to 4.
1001	* Note that even though we try and limit this to 4 packets,
1002	* the session could have more queued if the packets needed
1003	* to be fragmented.
1004	*
1005	* (XXX <= 4 ?)
1006	*/
1007	if ((so->so_state & SS_ISFCONNECTED) && so->so_queued <= 4)
1008	{
1009	SLIRP_COUNTER_INC(UDPHot);
1010	UDP_ENGAGE_EVENT(so, readfds);
1011	}
1012	LOOP_LABEL(udp, so, so_next);
1013	}
1014	done:
1015
1016	#if defined(RT_OS_WINDOWS)
1017	*pnfds = VBOX_EVENT_COUNT;
1018	#else /* RT_OS_WINDOWS */
1019	AssertRelease(poll_index <= *pnfds);
1020	*pnfds = poll_index;
1021	#endif /* !RT_OS_WINDOWS */
1022
1023	SLIRP_PROFILE_STOP(Fill, a);
1024	}
1025
1026	#if defined(RT_OS_WINDOWS)
1027	void slirp_select_poll(PNATState pData, int fTimeout, int fIcmp)
1028	#else /* RT_OS_WINDOWS */
1029	void slirp_select_poll(PNATState pData, struct pollfd *polls, int ndfs)
1030	#endif /* !RT_OS_WINDOWS */
1031	{
1032	struct socket so, so_next;
1033	int ret;
1034	#if defined(RT_OS_WINDOWS)
1035	WSANETWORKEVENTS NetworkEvents;
1036	int rc;
1037	int error;
1038	#else
1039	int poll_index = 0;
1040	#endif
1041
1042	SLIRP_PROFILE_START(Poll, a);
1043
1044	/* Update time */
1045	updtime(pData);
1046
1047	/*
1048	* See if anything has timed out
1049	*/
1050	if (link_up)
1051	{
1052	if (time_fasttimo && ((curtime - time_fasttimo) >= 2))
1053	{
1054	SLIRP_PROFILE_START(FastTimer, a);
1055	tcp_fasttimo(pData);
1056	time_fasttimo = 0;
1057	SLIRP_PROFILE_STOP(FastTimer, a);
1058	}
1059	if (do_slowtimo && ((curtime - last_slowtimo) >= 499))
1060	{
1061	SLIRP_PROFILE_START(SlowTimer, a);
1062	ip_slowtimo(pData);
1063	tcp_slowtimo(pData);
1064	last_slowtimo = curtime;
1065	SLIRP_PROFILE_STOP(SlowTimer, a);
1066	}
1067	}
1068	#if defined(RT_OS_WINDOWS)
1069	if (fTimeout)
1070	return; /* only timer update */
1071	#endif
1072
1073	/*
1074	* Check sockets
1075	*/
1076	if (!link_up)
1077	goto done;
1078	#if defined(RT_OS_WINDOWS)
1079	/*XXX: before renaming please make see define
1080	* fIcmp in slirp_state.h
1081	*/
1082	if (fIcmp)
1083	sorecvfrom(pData, &pData->icmp_socket);
1084	#else
1085	if ( (pData->icmp_socket.s != -1)
1086	&& CHECK_FD_SET(&pData->icmp_socket, ignored, readfds))
1087	sorecvfrom(pData, &pData->icmp_socket);
1088	#endif
1089	/*
1090	* Check TCP sockets
1091	*/
1092	QSOCKET_FOREACH(so, so_next, tcp)
1093	/* { */
1094
1095	#ifdef VBOX_WITH_SLIRP_MT
1096	if ( so->so_state & SS_NOFDREF
1097	&& so->so_deleted == 1)
1098	{
1099	struct socket son, sop = NULL;
1100	QSOCKET_LOCK(tcb);
1101	if (so->so_next != NULL)
1102	{
1103	if (so->so_next != &tcb)
1104	SOCKET_LOCK(so->so_next);
1105	son = so->so_next;
1106	}
1107	if ( so->so_prev != &tcb
1108	&& so->so_prev != NULL)
1109	{
1110	SOCKET_LOCK(so->so_prev);
1111	sop = so->so_prev;
1112	}
1113	QSOCKET_UNLOCK(tcb);
1114	remque(pData, so);
1115	NSOCK_DEC();
1116	SOCKET_UNLOCK(so);
1117	SOCKET_LOCK_DESTROY(so);
1118	RTMemFree(so);
1119	so_next = son;
1120	if (sop != NULL)
1121	SOCKET_UNLOCK(sop);
1122	CONTINUE_NO_UNLOCK(tcp);
1123	}
1124	#endif
1125	/*
1126	* FD_ISSET is meaningless on these sockets
1127	* (and they can crash the program)
1128	*/
1129	if (so->so_state & SS_NOFDREF \|\| so->s == -1)
1130	CONTINUE(tcp);
1131
1132	POLL_TCP_EVENTS(rc, error, so, &NetworkEvents);
1133
1134	LOG_NAT_SOCK(so, TCP, &NetworkEvents, readfds, writefds, xfds);
1135
1136
1137	/*
1138	* Check for URG data
1139	* This will soread as well, so no need to
1140	* test for readfds below if this succeeds
1141	*/
1142
1143	/* out-of-band data */
1144	if (CHECK_FD_SET(so, NetworkEvents, xfds))
1145	{
1146	sorecvoob(pData, so);
1147	}
1148
1149	/*
1150	* Check sockets for reading
1151	*/
1152	else if ( CHECK_FD_SET(so, NetworkEvents, readfds)
1153	\|\| WIN_CHECK_FD_SET(so, NetworkEvents, acceptds))
1154	{
1155	/*
1156	* Check for incoming connections
1157	*/
1158	if (so->so_state & SS_FACCEPTCONN)
1159	{
1160	TCP_CONNECT(pData, so);
1161	#if defined(RT_OS_WINDOWS)
1162	if (!(NetworkEvents.lNetworkEvents & FD_CLOSE))
1163	#endif
1164	CONTINUE(tcp);
1165	}
1166
1167	ret = soread(pData, so);
1168	/* Output it if we read something */
1169	if (RT_LIKELY(ret > 0))
1170	TCP_OUTPUT(pData, sototcpcb(so));
1171	}
1172
1173	#if defined(RT_OS_WINDOWS)
1174	/*
1175	* Check for FD_CLOSE events.
1176	* in some cases once FD_CLOSE engaged on socket it could be flashed latter (for some reasons)
1177	*/
1178	if ( (NetworkEvents.lNetworkEvents & FD_CLOSE)
1179	\|\| (so->so_close == 1))
1180	{
1181	so->so_close = 1; /* mark it */
1182	/*
1183	* drain the socket
1184	*/
1185	for (;;)
1186	{
1187	ret = soread(pData, so);
1188	if (ret > 0)
1189	TCP_OUTPUT(pData, sototcpcb(so));
1190	else
1191	break;
1192	}
1193	CONTINUE(tcp);
1194	}
1195	#endif
1196
1197	/*
1198	* Check sockets for writing
1199	*/
1200	if (CHECK_FD_SET(so, NetworkEvents, writefds))
1201	{
1202	/*
1203	* Check for non-blocking, still-connecting sockets
1204	*/
1205	if (so->so_state & SS_ISFCONNECTING)
1206	{
1207	Log2(("connecting %R[natsock] catched\n", so));
1208	/* Connected */
1209	so->so_state &= ~SS_ISFCONNECTING;
1210
1211	/*
1212	* This should be probably guarded by PROBE_CONN too. Anyway,
1213	* we disable it on OS/2 because the below send call returns
1214	* EFAULT which causes the opened TCP socket to close right
1215	* after it has been opened and connected.
1216	*/
1217	#ifndef RT_OS_OS2
1218	ret = send(so->s, (const char *)&ret, 0, 0);
1219	if (ret < 0)
1220	{
1221	/* XXXXX Must fix, zero bytes is a NOP */
1222	if ( errno == EAGAIN
1223	\|\| errno == EWOULDBLOCK
1224	\|\| errno == EINPROGRESS
1225	\|\| errno == ENOTCONN)
1226	CONTINUE(tcp);
1227
1228	/* else failed */
1229	so->so_state = SS_NOFDREF;
1230	}
1231	/* else so->so_state &= ~SS_ISFCONNECTING; */
1232	#endif
1233
1234	/*
1235	* Continue tcp_input
1236	*/
1237	TCP_INPUT(pData, (struct mbuf *)NULL, sizeof(struct ip), so);
1238	/* continue; */
1239	}
1240	else
1241	SOWRITE(ret, pData, so);
1242	/*
1243	* XXX If we wrote something (a lot), there could be the need
1244	* for a window update. In the worst case, the remote will send
1245	* a window probe to get things going again.
1246	*/
1247	}
1248
1249	/*
1250	* Probe a still-connecting, non-blocking socket
1251	* to check if it's still alive
1252	*/
1253	#ifdef PROBE_CONN
1254	if (so->so_state & SS_ISFCONNECTING)
1255	{
1256	ret = recv(so->s, (char *)&ret, 0, 0);
1257
1258	if (ret < 0)
1259	{
1260	/* XXX */
1261	if ( errno == EAGAIN
1262	\|\| errno == EWOULDBLOCK
1263	\|\| errno == EINPROGRESS
1264	\|\| errno == ENOTCONN)
1265	{
1266	CONTINUE(tcp); /* Still connecting, continue */
1267	}
1268
1269	/* else failed */
1270	so->so_state = SS_NOFDREF;
1271
1272	/* tcp_input will take care of it */
1273	}
1274	else
1275	{
1276	ret = send(so->s, &ret, 0, 0);
1277	if (ret < 0)
1278	{
1279	/* XXX */
1280	if ( errno == EAGAIN
1281	\|\| errno == EWOULDBLOCK
1282	\|\| errno == EINPROGRESS
1283	\|\| errno == ENOTCONN)
1284	{
1285	CONTINUE(tcp);
1286	}
1287	/* else failed */
1288	so->so_state = SS_NOFDREF;
1289	}
1290	else
1291	so->so_state &= ~SS_ISFCONNECTING;
1292
1293	}
1294	TCP_INPUT((struct mbuf *)NULL, sizeof(struct ip),so);
1295	} /* SS_ISFCONNECTING */
1296	#endif
1297	#ifndef RT_OS_WINDOWS
1298	if ( UNIX_CHECK_FD_SET(so, NetworkEvents, rdhup)
1299	\|\| UNIX_CHECK_FD_SET(so, NetworkEvents, rderr))
1300	{
1301	int err;
1302	int inq, outq;
1303	int status;
1304	socklen_t optlen = sizeof(int);
1305	inq = outq = 0;
1306	status = getsockopt(so->s, SOL_SOCKET, SO_ERROR, &err, &optlen);
1307	if (status != 0)
1308	Log(("NAT: can't get error status from %R[natsock]\n", so));
1309	#ifndef RT_OS_SOLARIS
1310	status = ioctl(so->s, FIONREAD, &inq); /* tcp(7) recommends SIOCINQ which is Linux specific */
1311	if (status != 0 \|\| status != EINVAL)
1312	{
1313	/* EINVAL returned if socket in listen state tcp(7)*/
1314	Log(("NAT: can't get depth of IN queue status from %R[natsock]\n", so));
1315	}
1316	status = ioctl(so->s, TIOCOUTQ, &outq); /* SIOCOUTQ see previous comment */
1317	if (status != 0)
1318	Log(("NAT: can't get depth of OUT queue from %R[natsock]\n", so));
1319	#else
1320	/*
1321	* Solaris has bit different ioctl commands and its handlings
1322	* hint: streamio(7) I_NREAD
1323	*/
1324	#endif
1325	if ( so->so_state & SS_ISFCONNECTING
1326	\|\| UNIX_CHECK_FD_SET(so, NetworkEvents, readfds))
1327	{
1328	/**
1329	* Check if we need here take care about gracefull connection
1330	* @todo try with proxy server
1331	*/
1332	if (UNIX_CHECK_FD_SET(so, NetworkEvents, readfds))
1333	{
1334	/*
1335	* Never meet inq != 0 or outq != 0, anyway let it stay for a while
1336	* in case it happens we'll able to detect it.
1337	* Give TCP/IP stack wait or expire the socket.
1338	*/
1339	Log(("NAT: %R[natsock] err(%d:%s) s(in:%d,out:%d)happens on read I/O, "
1340	"other side close connection \n", so, err, strerror(err), inq, outq));
1341	CONTINUE(tcp);
1342	}
1343	goto tcp_input_close;
1344	}
1345	if ( !UNIX_CHECK_FD_SET(so, NetworkEvents, readfds)
1346	&& !UNIX_CHECK_FD_SET(so, NetworkEvents, writefds)
1347	&& !UNIX_CHECK_FD_SET(so, NetworkEvents, xfds))
1348	{
1349	Log(("NAT: system expires the socket %R[natsock] err(%d:%s) s(in:%d,out:%d) happens on non-I/O. ",
1350	so, err, strerror(err), inq, outq));
1351	goto tcp_input_close;
1352	}
1353	Log(("NAT: %R[natsock] we've met(%d:%s) s(in:%d, out:%d) unhandled combination hup (%d) "
1354	"rederr(%d) on (r:%d, w:%d, x:%d)\n",
1355	so, err, strerror(err),
1356	inq, outq,
1357	UNIX_CHECK_FD_SET(so, ign, rdhup),
1358	UNIX_CHECK_FD_SET(so, ign, rderr),
1359	UNIX_CHECK_FD_SET(so, ign, readfds),
1360	UNIX_CHECK_FD_SET(so, ign, writefds),
1361	UNIX_CHECK_FD_SET(so, ign, xfds)));
1362	/*
1363	* Give OS's TCP/IP stack a chance to resolve an issue or expire the socket.
1364	*/
1365	CONTINUE(tcp);
1366	tcp_input_close:
1367	so->so_state = SS_NOFDREF; /cause connection valid tcp connection termination and socket closing /
1368	TCP_INPUT(pData, (struct mbuf *)NULL, sizeof(struct ip), so);
1369	CONTINUE(tcp);
1370	}
1371	#endif
1372	LOOP_LABEL(tcp, so, so_next);
1373	}
1374
1375	/*
1376	* Now UDP sockets.
1377	* Incoming packets are sent straight away, they're not buffered.
1378	* Incoming UDP data isn't buffered either.
1379	*/
1380	QSOCKET_FOREACH(so, so_next, udp)
1381	/* { */
1382	#ifdef VBOX_WITH_SLIRP_MT
1383	if ( so->so_state & SS_NOFDREF
1384	&& so->so_deleted == 1)
1385	{
1386	struct socket son, sop = NULL;
1387	QSOCKET_LOCK(udb);
1388	if (so->so_next != NULL)
1389	{
1390	if (so->so_next != &udb)
1391	SOCKET_LOCK(so->so_next);
1392	son = so->so_next;
1393	}
1394	if ( so->so_prev != &udb
1395	&& so->so_prev != NULL)
1396	{
1397	SOCKET_LOCK(so->so_prev);
1398	sop = so->so_prev;
1399	}
1400	QSOCKET_UNLOCK(udb);
1401	remque(pData, so);
1402	NSOCK_DEC();
1403	SOCKET_UNLOCK(so);
1404	SOCKET_LOCK_DESTROY(so);
1405	RTMemFree(so);
1406	so_next = son;
1407	if (sop != NULL)
1408	SOCKET_UNLOCK(sop);
1409	CONTINUE_NO_UNLOCK(udp);
1410	}
1411	#endif
1412	POLL_UDP_EVENTS(rc, error, so, &NetworkEvents);
1413
1414	LOG_NAT_SOCK(so, UDP, &NetworkEvents, readfds, writefds, xfds);
1415
1416	if (so->s != -1 && CHECK_FD_SET(so, NetworkEvents, readfds))
1417	{
1418	SORECVFROM(pData, so);
1419	}
1420	LOOP_LABEL(udp, so, so_next);
1421	}
1422
1423	done:
1424	#ifndef VBOX_WITH_SLIRP_MT
1425	/*
1426	* See if we can start outputting
1427	*/
1428	if (if_queued && link_up)
1429	if_start(pData);
1430	#endif
1431
1432	SLIRP_PROFILE_STOP(Poll, a);
1433	}
1434
1435	#ifndef VBOX_WITHOUT_SLIRP_CLIENT_ETHER
1436	#define ETH_ALEN 6
1437	#define ETH_HLEN 14
1438
1439	#define ARPOP_REQUEST 1 /* ARP request */
1440	#define ARPOP_REPLY 2 /* ARP reply */
1441
1442	struct ethhdr
1443	{
1444	unsigned char h_dest[ETH_ALEN]; /* destination eth addr */
1445	unsigned char h_source[ETH_ALEN]; /* source ether addr */
1446	unsigned short h_proto; /* packet type ID field */
1447	};
1448	AssertCompileSize(struct ethhdr, 14);
1449	#endif
1450
1451	struct arphdr
1452	{
1453	unsigned short ar_hrd; /* format of hardware address */
1454	unsigned short ar_pro; /* format of protocol address */
1455	unsigned char ar_hln; /* length of hardware address */
1456	unsigned char ar_pln; /* length of protocol address */
1457	unsigned short ar_op; /* ARP opcode (command) */
1458
1459	/*
1460	* Ethernet looks like this : This bit is variable sized however...
1461	*/
1462	unsigned char ar_sha[ETH_ALEN]; /* sender hardware address */
1463	unsigned char ar_sip[4]; /* sender IP address */
1464	unsigned char ar_tha[ETH_ALEN]; /* target hardware address */
1465	unsigned char ar_tip[4]; /* target IP address */
1466	};
1467	AssertCompileSize(struct arphdr, 28);
1468
1469	static void arp_input(PNATState pData, struct mbuf *m)
1470	{
1471	struct ethhdr *eh;
1472	struct ethhdr *reh;
1473	struct arphdr *ah;
1474	struct arphdr *rah;
1475	int ar_op;
1476	struct ex_list *ex_ptr;
1477	uint32_t htip;
1478	uint32_t tip;
1479	struct mbuf *mr;
1480	eh = mtod(m, struct ethhdr *);
1481	ah = (struct arphdr *)&eh[1];
1482	htip = ntohl((uint32_t)ah->ar_tip);
1483	tip = (uint32_t)ah->ar_tip;
1484
1485	mr = m_get(pData);
1486	#ifdef VBOX_WITHOUT_SLIRP_CLIENT_ETHER
1487	reh = mtod(mr, struct ethhdr *);
1488	memcpy(reh->h_source, eh->h_source, ETH_ALEN); /* XXX: if_encap will swap src and dst*/
1489	Log4(("NAT: arp:%R[ether]->%R[ether]\n",
1490	reh->h_source, reh->h_dest));
1491	Log4(("NAT: arp: %R[IP4]\n", &tip));
1492	#endif
1493	mr->m_data += if_maxlinkhdr;
1494	mr->m_len = sizeof(struct arphdr);
1495	rah = mtod(mr, struct arphdr *);
1496
1497	ar_op = ntohs(ah->ar_op);
1498	switch(ar_op)
1499	{
1500	case ARPOP_REQUEST:
1501	#ifdef VBOX_WITH_NAT_SERVICE
1502	if (tip == special_addr.s_addr) goto arp_ok;
1503	#endif
1504	if ((htip & pData->netmask) == ntohl(special_addr.s_addr))
1505	{
1506	if ( CTL_CHECK(htip, CTL_DNS)
1507	\|\| CTL_CHECK(htip, CTL_ALIAS)
1508	\|\| CTL_CHECK(htip, CTL_TFTP))
1509	goto arp_ok;
1510	for (ex_ptr = exec_list; ex_ptr; ex_ptr = ex_ptr->ex_next)
1511	{
1512	if ((htip & ~pData->netmask) == ex_ptr->ex_addr)
1513	{
1514	goto arp_ok;
1515	}
1516	}
1517	return;
1518	arp_ok:
1519	rah->ar_hrd = htons(1);
1520	rah->ar_pro = htons(ETH_P_IP);
1521	rah->ar_hln = ETH_ALEN;
1522	rah->ar_pln = 4;
1523	rah->ar_op = htons(ARPOP_REPLY);
1524	memcpy(rah->ar_sha, special_ethaddr, ETH_ALEN);
1525
1526	switch (htip & ~pData->netmask)
1527	{
1528	case CTL_DNS:
1529	case CTL_ALIAS:
1530	rah->ar_sha[5] = (uint8_t)(htip & ~pData->netmask);
1531	break;
1532	default:;
1533	}
1534
1535	memcpy(rah->ar_sip, ah->ar_tip, 4);
1536	memcpy(rah->ar_tha, ah->ar_sha, ETH_ALEN);
1537	memcpy(rah->ar_tip, ah->ar_sip, 4);
1538	if_encap(pData, ETH_P_ARP, mr);
1539	m_free(pData, m);
1540	}
1541	break;
1542	default:
1543	break;
1544	}
1545	}
1546
1547	void slirp_input(PNATState pData, const uint8_t *pkt, int pkt_len)
1548	{
1549	struct mbuf *m;
1550	int proto;
1551	static bool fWarnedIpv6;
1552	struct ethhdr eh = (struct ethhdr)pkt;
1553
1554	Log2(("NAT: slirp_input %d\n", pkt_len));
1555	if (pkt_len < ETH_HLEN)
1556	{
1557	LogRel(("NAT: packet having size %d has been ingnored\n", pkt_len));
1558	return;
1559	}
1560	Log4(("NAT: in:%R[ether]->%R[ether]\n", &eh->h_source, &eh->h_dest));
1561	#ifdef VBOX_WITHOUT_SLIRP_CLIENT_ETHER
1562	if (memcmp(eh->h_source, special_ethaddr, ETH_ALEN) == 0)
1563	{
1564	/* @todo vasily: add ether logging routine in debug.c */
1565	Log(("NAT: packet was addressed to other MAC\n"));
1566	RTMemFree((void *)pkt);
1567	return;
1568	}
1569	#endif
1570
1571	m = m_get(pData);
1572	if (!m)
1573	{
1574	LogRel(("NAT: can't allocate new mbuf\n"));
1575	return;
1576	}
1577
1578	/* Note: we add to align the IP header */
1579
1580	if (M_FREEROOM(m) < pkt_len)
1581	m_inc(m, pkt_len);
1582
1583	m->m_len = pkt_len ;
1584	memcpy(m->m_data, pkt, pkt_len);
1585
1586	proto = ntohs((uint16_t )(pkt + 12));
1587	switch(proto)
1588	{
1589	case ETH_P_ARP:
1590	arp_input(pData, m);
1591	break;
1592	case ETH_P_IP:
1593	/* Update time. Important if the network is very quiet, as otherwise
1594	* the first outgoing connection gets an incorrect timestamp. */
1595	updtime(pData);
1596	m_adj(m, ETH_HLEN);
1597	ip_input(pData, m);
1598	break;
1599	case ETH_P_IPV6:
1600	m_free(pData, m);
1601	if (!fWarnedIpv6)
1602	{
1603	LogRel(("NAT: IPv6 not supported\n"));
1604	fWarnedIpv6 = true;
1605	}
1606	break;
1607	default:
1608	Log(("NAT: Unsupported protocol %x\n", proto));
1609	m_free(pData, m);
1610	break;
1611	}
1612	RTMemFree((void *)pkt);
1613	}
1614
1615	/* output the IP packet to the ethernet device */
1616	void if_encap(PNATState pData, uint16_t eth_proto, struct mbuf *m)
1617	{
1618	struct ethhdr *eh;
1619	uint8_t *buf = RTMemAlloc(1600);
1620
1621	m->m_data -= if_maxlinkhdr;
1622	m->m_len += ETH_HLEN;
1623	eh = mtod(m, struct ethhdr *);
1624
1625	if(MBUF_HEAD(m) != m->m_data)
1626	{
1627	LogRel(("NAT: ethernet detects corruption of the packet"));
1628	AssertMsgFailed(("!!Ethernet frame corrupted!!"));
1629	}
1630
1631	#ifndef VBOX_WITHOUT_SLIRP_CLIENT_ETHER
1632	memcpy(eh->h_dest, client_ethaddr, ETH_ALEN);
1633	memcpy(eh->h_source, special_ethaddr, ETH_ALEN - 1);
1634	/* XXX: not correct */
1635	eh->h_source[5] = CTL_ALIAS;
1636	#else
1637	if (memcmp(eh->h_source, special_ethaddr, ETH_ALEN) != 0)
1638	{
1639	memcpy(eh->h_dest, eh->h_source, ETH_ALEN);
1640	memcpy(eh->h_source, special_ethaddr, ETH_ALEN);
1641	Assert(memcmp(eh->h_dest, special_ethaddr, ETH_ALEN) != 0);
1642	if (memcmp(eh->h_dest, zerro_ethaddr, ETH_ALEN) == 0)
1643	{
1644	/* don't do anything */
1645	goto done;
1646	}
1647	}
1648	#endif
1649	eh->h_proto = htons(eth_proto);
1650	#if 0
1651	slirp_output(pData->pvUser, m, mtod(m, uint8_t *), m->m_len);
1652	#else
1653	memcpy(buf, mtod(m, uint8_t *), m->m_len);
1654	slirp_output(pData->pvUser, NULL, buf, m->m_len);
1655	done:
1656	m_free(pData, m);
1657	#endif
1658	}
1659
1660	int slirp_redir(PNATState pData, int is_udp, int host_port,
1661	struct in_addr guest_addr, int guest_port)
1662	{
1663	struct socket *so;
1664	#ifdef VBOX_WITH_SLIRP_ALIAS
1665	struct alias_link *link;
1666	struct libalias *lib;
1667	int flags;
1668	struct sockaddr sa;
1669	struct sockaddr_in *psin;
1670	socklen_t socketlen;
1671	struct in_addr alias;
1672	int rc;
1673	#endif
1674	Log2(("NAT: set redirect %s hp:%d gp:%d\n", (is_udp?"UDP":"TCP"), host_port, guest_port));
1675	if (is_udp)
1676	{
1677	so = udp_listen(pData, htons(host_port), guest_addr.s_addr,
1678	htons(guest_port), 0);
1679	}
1680	else
1681	{
1682	so = solisten(pData, htons(host_port), guest_addr.s_addr,
1683	htons(guest_port), 0);
1684	}
1685	#ifndef VBOX_WITH_SLIRP_ALIAS
1686	Log2(("NAT: redirecting socket %R[natsock]\n", so));
1687	return (so != NULL ? 0 : -1);
1688	#else
1689
1690	psin = (struct sockaddr_in *)&sa;
1691	psin->sin_family = AF_INET;
1692	psin->sin_port = 0;
1693	psin->sin_addr.s_addr = INADDR_ANY;
1694	socketlen = sizeof(struct sockaddr);
1695
1696	rc = getsockname(so->s, &sa, &socketlen);
1697	if (rc < 0 \|\| sa.sa_family != AF_INET)
1698	{
1699	Log(("NAT: can't get socket's name\n"));
1700	return 1;
1701	}
1702
1703	psin = (struct sockaddr_in *)&sa;
1704
1705	#if 1
1706	lib = LibAliasInit(pData, NULL);
1707	flags = LibAliasSetMode(lib, 0, 0);
1708	flags \|= PKT_ALIAS_LOG; /* set logging */
1709	flags \|= PKT_ALIAS_REVERSE; /* set logging */
1710	flags = LibAliasSetMode(lib, flags, ~0);
1711	#else
1712	lib = LIST_FIRST(&instancehead);
1713	#endif
1714
1715	alias.s_addr = htonl(ntohl(guest_addr.s_addr) \| CTL_ALIAS);
1716	link = LibAliasRedirectPort(lib, psin->sin_addr, htons(host_port),
1717	alias, htons(guest_port),
1718	special_addr, -1, /* not very clear for now*/
1719	(is_udp ? IPPROTO_UDP : IPPROTO_TCP));
1720	if (link == NULL)
1721	{
1722	Log(("NAT: can't create redirect\n"));
1723	return 1;
1724	}
1725	so->so_la = lib;
1726
1727	return 0;
1728	#endif
1729	}
1730
1731	int slirp_add_exec(PNATState pData, int do_pty, const char *args, int addr_low_byte,
1732	int guest_port)
1733	{
1734	return add_exec(&exec_list, do_pty, (char *)args,
1735	addr_low_byte, htons(guest_port));
1736	}
1737
1738	void slirp_set_ethaddr(PNATState pData, const uint8_t *ethaddr)
1739	{
1740	#ifndef VBOX_WITHOUT_SLIRP_CLIENT_ETHER
1741	memcpy(client_ethaddr, ethaddr, ETH_ALEN);
1742	#endif
1743	}
1744
1745	#if defined(RT_OS_WINDOWS)
1746	HANDLE *slirp_get_events(PNATState pData)
1747	{
1748	return pData->phEvents;
1749	}
1750	void slirp_register_external_event(PNATState pData, HANDLE hEvent, int index)
1751	{
1752	pData->phEvents[index] = hEvent;
1753	}
1754	#endif
1755
1756	unsigned int slirp_get_timeout_ms(PNATState pData)
1757	{
1758	if (link_up)
1759	{
1760	if (time_fasttimo)
1761	return 2;
1762	if (do_slowtimo)
1763	return 500; /* see PR_SLOWHZ */
1764	}
1765	return 0;
1766	}
1767
1768	#ifndef RT_OS_WINDOWS
1769	int slirp_get_nsock(PNATState pData)
1770	{
1771	return pData->nsock;
1772	}
1773	#endif
1774
1775	/*
1776	* this function called from NAT thread
1777	*/
1778	void slirp_post_sent(PNATState pData, void *pvArg)
1779	{
1780	struct socket *so = 0;
1781	struct tcpcb *tp = 0;
1782	struct mbuf m = (struct mbuf )pvArg;
1783	m_free(pData, m);
1784	}
1785	#ifdef VBOX_WITH_SLIRP_MT
1786	void slirp_process_queue(PNATState pData)
1787	{
1788	RTReqProcess(pData->pReqQueue, RT_INDEFINITE_WAIT);
1789	}
1790	void *slirp_get_queue(PNATState pData)
1791	{
1792	return pData->pReqQueue;
1793	}
1794	#endif
1795
1796	uint16_t slirp_get_service(int proto, uint16_t dport, uint16_t sport)
1797	{
1798	uint16_t hdport, hsport, service;
1799	hdport = ntohs(dport);
1800	hsport = ntohs(sport);
1801	Log2(("proto: %d, dport: %d sport: %d\n", proto, hdport, hsport));
1802	service = 0;
1803	#if 0
1804	/* Always return 0 here */
1805	switch (hdport)
1806	{
1807	case 500:
1808	if (hsport != 500) /* vpnc by default try operate in src:500/dst:500 mode*/
1809	/* Not sure why this make Cisco VPN client's connection more stable,
1810	* at least on some servers
1811	*/
1812	service = sport;
1813	break;
1814	}
1815	#endif
1816	Log2(("service : %d\n", service));
1817	return htons(service);
1818	}
1819
1820	void slirp_set_dhcp_TFTP_prefix(PNATState pData, const char *tftpPrefix)
1821	{
1822	Log2(("tftp_prefix:%s\n", tftpPrefix));
1823	tftp_prefix = tftpPrefix;
1824	}
1825
1826	void slirp_set_dhcp_TFTP_bootfile(PNATState pData, const char *bootFile)
1827	{
1828	Log2(("bootFile:%s\n", bootFile));
1829	bootp_filename = bootFile;
1830	}
1831
1832	void slirp_set_dhcp_next_server(PNATState pData, const char *next_server)
1833	{
1834	Log2(("next_server:%s\n", next_server));
1835	if (next_server == NULL)
1836	pData->tftp_server.s_addr = htonl(ntohl(special_addr.s_addr) \| CTL_TFTP);
1837	else
1838	inet_aton(next_server, &pData->tftp_server);
1839	}
1840	#ifdef VBOX_WITH_SLIRP_DNS_PROXY
1841	void slirp_set_dhcp_dns_proxy(PNATState pData, bool fDNSProxy)
1842	{
1843	Log2(("NAT: DNS proxy switched %s\n", (fDNSProxy ? "on" : "off")));
1844	pData->use_dns_proxy = fDNSProxy;
1845	}
1846	#endif

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

source: vbox/trunk/src/VBox/Devices/Network/slirp/slirp.c@ 20376

Download in other formats: