slirp.c@ 16303

Last change on this file since 16303 was 16303, checked in by vboxsync, 16 years ago
NAT:DNS freeing DNS server list
Property svn:eol-style set to `native`
File size: 39.7 KB

Line
1	#include "slirp.h"
2	#ifdef RT_OS_OS2
3	# include <paths.h>
4	#endif
5
6	#include <VBox/err.h>
7	#include <VBox/pdmdrv.h>
8	#include <iprt/assert.h>
9
10	#ifdef VBOX_WITH_SLIRP_MT
11	# define CONTINUE(label) goto loop_end_ ## label ## _mt
12	/* @todo replace queue parameter with macrodinition */
13	# define LOOP_LABEL(label, so, sonext) loop_end_ ## label ## _mt: \
14	SOCKET_UNLOCK(so); \
15	QSOCKET_LOCK(_X(queue_ ## label ## _label)); \
16	(so) = (sonext)
17	#else
18	#define CONTINUE(label) continue;
19	# define LOOP_LABEL(label, so, sonext) /* empty*/
20	#endif
21	#if !defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) \|\| !defined(RT_OS_WINDOWS)
22
23	# define DO_ENGAGE_EVENT1(so, fdset, label) \
24	do { \
25	FD_SET((so)->s, (fdset)); \
26	UPD_NFDS((so)->s); \
27	} while(0)
28
29
30	# define DO_ENGAGE_EVENT2(so, fdset1, fdset2, label) \
31	do { \
32	FD_SET((so)->s, (fdset1)); \
33	FD_SET((so)->s, (fdset2)); \
34	UPD_NFDS((so)->s); \
35	} while(0)
36
37	# define DO_POLL_EVENTS(rc, error, so, events, label) do {} while (0)
38
39	# define DO_CHECK_FD_SET(so, events, fdset) (FD_ISSET((so)->s, (fdset)))
40
41	# define DO_WIN_CHECK_FD_SET(so, events, fdset ) 0 /* specific for Windows Winsock API */
42
43	# ifndef RT_OS_WINDOWS
44	# define ICMP_ENGAGE_EVENT(so, fdset) \
45	do { \
46	if (pData->icmp_socket.s != -1) \
47	DO_ENGAGE_EVENT1((so), (fdset), ICMP); \
48	} while (0)
49	# else /* !RT_OS_WINDOWS */
50	# define ICMP_ENGAGE_EVENT(so, fdset) do {} while(0)
51	#endif /* RT_OS_WINDOWS */
52
53	#else /* defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) && defined(RT_OS_WINDOWS) */
54
55	/*
56	* On Windows, we will be notified by IcmpSendEcho2() when the response arrives.
57	* So no call to WSAEventSelect necessary.
58	*/
59	# define ICMP_ENGAGE_EVENT(so, fdset) do {} while(0)
60
61	# define DO_ENGAGE_EVENT1(so, fdset1, label) \
62	do { \
63	rc = WSAEventSelect((so)->s, VBOX_SOCKET_EVENT, FD_ALL_EVENTS); \
64	if (rc == SOCKET_ERROR) \
65	{ \
66	/* This should not happen */ \
67	error = WSAGetLastError(); \
68	LogRel(("WSAEventSelector (" #label ") error %d (so=%x, socket=%s, event=%x)\n", \
69	error, (so), (so)->s, VBOX_SOCKET_EVENT)); \
70	} \
71	} while(0); \
72	CONTINUE(label)
73
74	# define DO_ENGAGE_EVENT2(so, fdset1, fdset2, label) \
75	DO_ENGAGE_EVENT1((so), (fdset1), label)
76
77	# define DO_POLL_EVENTS(rc, error, so, events, label) \
78	(rc) = WSAEnumNetworkEvents((so)->s, VBOX_SOCKET_EVENT, (events)); \
79	if ((rc) == SOCKET_ERROR) \
80	{ \
81	(error) = WSAGetLastError(); \
82	LogRel(("WSAEnumNetworkEvents " #label " error %d\n", (error))); \
83	CONTINUE(label); \
84	}
85
86	# define acceptds_win FD_ACCEPT
87	# define acceptds_win_bit FD_ACCEPT_BIT
88
89	# define readfds_win FD_READ
90	# define readfds_win_bit FD_READ_BIT
91
92	# define writefds_win FD_WRITE
93	# define writefds_win_bit FD_WRITE_BIT
94
95	# define xfds_win FD_OOB
96	# define xfds_win_bit FD_OOB_BIT
97
98	# define DO_CHECK_FD_SET(so, events, fdset) \
99	(((events).lNetworkEvents & fdset ## _win) && ((events).iErrorCode[fdset ## _win_bit] == 0))
100
101	# define DO_WIN_CHECK_FD_SET(so, events, fdset ) DO_CHECK_FD_SET((so), (events), fdset)
102
103	#endif /* defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) && defined(RT_OS_WINDOWS) */
104
105	#define TCP_ENGAGE_EVENT1(so, fdset) \
106	DO_ENGAGE_EVENT1((so), (fdset), tcp)
107
108	#define TCP_ENGAGE_EVENT2(so, fdset1, fdset2) \
109	DO_ENGAGE_EVENT2((so), (fdset1), (fdset2), tcp)
110
111	#define UDP_ENGAGE_EVENT(so, fdset) \
112	DO_ENGAGE_EVENT1((so), (fdset), udp)
113
114	#define POLL_TCP_EVENTS(rc, error, so, events) \
115	DO_POLL_EVENTS((rc), (error), (so), (events), tcp)
116
117	#define POLL_UDP_EVENTS(rc, error, so, events) \
118	DO_POLL_EVENTS((rc), (error), (so), (events), udp)
119
120	#define CHECK_FD_SET(so, events, set) \
121	(DO_CHECK_FD_SET((so), (events), set))
122
123	#define WIN_CHECK_FD_SET(so, events, set) \
124	(DO_WIN_CHECK_FD_SET((so), (events), set))
125
126	/*
127	* Loging macros
128	*/
129	#if VBOX_WITH_DEBUG_NAT_SOCKETS
130	# if defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) && defined(RT_OS_WINDOWS)
131	# define DO_LOG_NAT_SOCK(so, proto, winevent, r_fdset, w_fdset, x_fdset) \
132	do { \
133	LogRel((" " #proto "%R[natsock] %R[natwinnetevents]\n", (so), (winevent))); \
134	} while (0)
135	# else
136	# define DO_LOG_NAT_SOCK(so, proto, winevent, r_fdset, w_fdset, x_fdset) \
137	do { \
138	LogRel((" " #proto " %R[natsock] %s %s %s\n", (so), FD_ISSET((so)->s, (r_fdset))?"READ":"",\
139	FD_ISSET((so)->s, (w_fdset))?"WRITE":"", FD_ISSET((so)->s, (x_fdset))?"OOB":"")); \
140	} while (0)
141	# endif /* VBOX_WITH_DEBUG_NAT_SOCKETS */
142	#else
143	# define DO_LOG_NAT_SOCK(so, proto, winevent, r_fdset, w_fdset, x_fdset) do {} while (0)
144	#endif /* !VBOX_WITH_DEBUG_NAT_SOCKETS */
145
146	#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))
147
148	static const uint8_t special_ethaddr[6] =
149	{
150	0x52, 0x54, 0x00, 0x12, 0x35, 0x00
151	};
152
153	#ifdef RT_OS_WINDOWS
154
155	static int get_dns_addr_domain(PNATState pData, bool fVerbose,
156	struct in_addr *pdns_addr,
157	const char **ppszDomain)
158	{
159	int rc = 0;
160	FIXED_INFO *FixedInfo = NULL;
161	ULONG BufLen;
162	DWORD ret;
163	IP_ADDR_STRING *pIPAddr;
164	struct in_addr tmp_addr;
165
166	FixedInfo = (FIXED_INFO *)GlobalAlloc(GPTR, sizeof(FIXED_INFO));
167	BufLen = sizeof(FIXED_INFO);
168
169	/** @todo: this API returns all DNS servers, no matter whether the
170	* corresponding network adapter is disabled or not. Maybe replace
171	* this by GetAdapterAddresses(), which is XP/Vista only though. */
172	if (ERROR_BUFFER_OVERFLOW == GetNetworkParams(FixedInfo, &BufLen))
173	{
174	if (FixedInfo)
175	{
176	GlobalFree(FixedInfo);
177	FixedInfo = NULL;
178	}
179	FixedInfo = GlobalAlloc(GPTR, BufLen);
180	}
181
182	if ((ret = GetNetworkParams(FixedInfo, &BufLen)) != ERROR_SUCCESS)
183	{
184	Log(("GetNetworkParams failed. ret = %08x\n", (u_int)ret ));
185	if (FixedInfo)
186	{
187	GlobalFree(FixedInfo);
188	FixedInfo = NULL;
189	}
190	rc = -1;
191	goto get_dns_prefix;
192	}
193
194	#ifndef VBOX_WITH_MULTI_DNS
195	pIPAddr = &(FixedInfo->DnsServerList);
196	inet_aton(pIPAddr->IpAddress.String, &tmp_addr);
197	Log(("nat: DNS Servers:\n"));
198	if (fVerbose \|\| pdns_addr->s_addr != tmp_addr.s_addr)
199	LogRel(("NAT: DNS address: %s\n", pIPAddr->IpAddress.String));
200	*pdns_addr = tmp_addr;
201
202	pIPAddr = FixedInfo -> DnsServerList.Next;
203	while (pIPAddr)
204	{
205	if (fVerbose)
206	LogRel(("NAT: ignored DNS address: %s\n", pIPAddr ->IpAddress.String));
207	pIPAddr = pIPAddr ->Next;
208	}
209	#else
210	/localhost mask /
211	for (pIPAddr = &FixedInfo->DnsServerList; pIPAddr != NULL; pIPAddr = pIPAddr->Next)
212	{
213	struct dns_entry *da;
214	if(!inet_aton(pIPAddr->IpAddress.String, &tmp_addr))
215	continue;
216	da = RTMemAllocZ(sizeof (struct dns_entry));
217	if (da == NULL)
218	{
219	LogRel(("can't alloc memory for DNS entry\n"));
220	return -1;
221	}
222	/check /
223	if ((da->de_addr.s_addr & htonl(IN_CLASSA_NET)) == ntohl(INADDR_LOOPBACK & IN_CLASSA_NET)) {
224	da->de_addr.s_addr = htonl(ntohl(special_addr.s_addr) \| CTL_ALIAS);
225	}
226	LIST_INSERT_HEAD(&pData->dns_list_head, da, de_list);
227	}
228	#endif
229	if (FixedInfo)
230	{
231	GlobalFree(FixedInfo);
232	FixedInfo = NULL;
233	}
234
235	get_dns_prefix:
236	if (ppszDomain)
237	{
238	OSVERSIONINFO ver;
239	char szDnsDomain[256];
240	DWORD dwSize = sizeof(szDnsDomain);
241
242	*ppszDomain = NULL;
243	GetVersionEx(&ver);
244	if (ver.dwMajorVersion >= 5)
245	{
246	/* GetComputerNameEx exists in Windows versions starting with 2000. */
247	if (GetComputerNameEx(ComputerNameDnsDomain, szDnsDomain, &dwSize))
248	{
249	if (szDnsDomain[0])
250	{
251	/* Just non-empty strings are valid. */
252	*ppszDomain = RTStrDup(szDnsDomain);
253	if (pData->fPassDomain)
254	{
255	if (fVerbose)
256	LogRel(("NAT: passing domain name %s\n", szDnsDomain));
257	}
258	else
259	Log(("nat: ignoring domain %s\n", szDnsDomain));
260	}
261	}
262	else
263	Log(("nat: GetComputerNameEx failed (%d)\n", GetLastError()));
264	}
265	}
266	return rc;
267	}
268
269	#else
270
271	static int get_dns_addr_domain(PNATState pData, bool fVerbose,
272	struct in_addr *pdns_addr,
273	const char **ppszDomain)
274	{
275	char buff[512];
276	char buff2[256];
277	FILE *f;
278	int found = 0;
279	struct in_addr tmp_addr;
280
281	#ifdef RT_OS_OS2
282	/* Try various locations. */
283	char *etc = getenv("ETC");
284	f = NULL;
285	if (etc)
286	{
287	snprintf(buff, sizeof(buff), "%s/RESOLV2", etc);
288	f = fopen(buff, "rt");
289	}
290	if (!f)
291	{
292	snprintf(buff, sizeof(buff), "%s/RESOLV2", _PATH_ETC);
293	f = fopen(buff, "rt");
294	}
295	if (!f)
296	{
297	snprintf(buff, sizeof(buff), "%s/resolv.conf", _PATH_ETC);
298	f = fopen(buff, "rt");
299	}
300	#else
301	f = fopen("/etc/resolv.conf", "r");
302	#endif
303	if (!f)
304	return -1;
305
306	if (ppszDomain)
307	*ppszDomain = NULL;
308	Log(("nat: DNS Servers:\n"));
309	while (fgets(buff, 512, f) != NULL)
310	{
311	#ifdef VBOX_WITH_MULTI_DNS
312	struct dns_entry *da = NULL;
313	#endif
314	if (sscanf(buff, "nameserver%*[ \t]%256s", buff2) == 1)
315	{
316	if (!inet_aton(buff2, &tmp_addr))
317	continue;
318	#ifndef VBOX_WITH_MULTI_DNS
319	/* If it's the first one, set it to dns_addr */
320	if (!found)
321	{
322	if (fVerbose \|\| pdns_addr->s_addr != tmp_addr.s_addr)
323	LogRel(("NAT: DNS address: %s\n", buff2));
324	*pdns_addr = tmp_addr;
325	}
326	else
327	{
328	if (fVerbose)
329	LogRel(("NAT: ignored DNS address: %s\n", buff2));
330	}
331	#else
332	/localhost mask /
333	da = RTMemAllocZ(sizeof (struct dns_entry));
334	if (da == NULL)
335	{
336	LogRel(("can't alloc memory for DNS entry\n"));
337	return -1;
338	}
339	/check /
340	da->de_addr.s_addr = tmp_addr.s_addr;
341	if ((da->de_addr.s_addr & htonl(IN_CLASSA_NET)) == ntohl(INADDR_LOOPBACK & IN_CLASSA_NET)) {
342	da->de_addr.s_addr = htonl(ntohl(special_addr.s_addr) \| CTL_ALIAS);
343	}
344	LIST_INSERT_HEAD(&pData->dns_list_head, da, de_list);
345	#endif
346	found++;
347	}
348	if ( ppszDomain
349	&& (!strncmp(buff, "domain", 6) \|\| !strncmp(buff, "search", 6)))
350	{
351	/* Domain name/search list present. Pick first entry */
352	if (*ppszDomain == NULL)
353	{
354	char *tok;
355	char *saveptr;
356	tok = strtok_r(&buff[6], " \t\n", &saveptr);
357	if (tok)
358	{
359	*ppszDomain = RTStrDup(tok);
360	if (pData->fPassDomain)
361	{
362	if (fVerbose)
363	LogRel(("NAT: passing domain name %s\n", tok));
364	}
365	else
366	Log(("nat: ignoring domain %s\n", tok));
367	}
368	}
369	}
370	}
371	fclose(f);
372	if (!found)
373	return -1;
374	return 0;
375	}
376
377	#endif
378
379	int get_dns_addr(PNATState pData, struct in_addr *pdns_addr)
380	{
381	return get_dns_addr_domain(pData, false, pdns_addr, NULL);
382	}
383
384	int slirp_init(PNATState ppData, const char pszNetAddr, uint32_t u32Netmask,
385	bool fPassDomain, const char *pszTFTPPrefix,
386	const char pszBootFile, void pvUser)
387	{
388	int fNATfailed = 0;
389	PNATState pData = RTMemAlloc(sizeof(NATState));
390	*ppData = pData;
391	if (!pData)
392	return VERR_NO_MEMORY;
393	if (u32Netmask & 0x1f)
394	/* CTL is x.x.x.15, bootp passes up to 16 IPs (15..31) */
395	return VERR_INVALID_PARAMETER;
396	memset(pData, '\0', sizeof(NATState));
397	pData->fPassDomain = fPassDomain;
398	pData->pvUser = pvUser;
399	tftp_prefix = pszTFTPPrefix;
400	bootp_filename = pszBootFile;
401	pData->netmask = u32Netmask;
402
403	#ifdef RT_OS_WINDOWS
404	{
405	WSADATA Data;
406	WSAStartup(MAKEWORD(2,0), &Data);
407	}
408	# if defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC)
409	pData->phEvents[VBOX_SOCKET_EVENT_INDEX] = CreateEvent(NULL, FALSE, FALSE, NULL);
410	# endif
411	#endif
412
413	link_up = 1;
414
415	debug_init();
416	if_init(pData);
417	ip_init(pData);
418	icmp_init(pData);
419
420	/* Initialise mbufs after setting the MTU */
421	m_init(pData);
422
423	inet_aton(pszNetAddr, &special_addr);
424	alias_addr.s_addr = special_addr.s_addr \| htonl(CTL_ALIAS);
425	/* @todo: add ability to configure this staff */
426	pData->tftp_server.s_addr = htonl(ntohl(special_addr.s_addr) \| CTL_TFTP);
427
428	/* set default addresses */
429	inet_aton("127.0.0.1", &loopback_addr);
430	#ifndef VBOX_WITH_MULTI_DNS
431	inet_aton("127.0.0.1", &dns_addr);
432
433	if (get_dns_addr_domain(pData, true, &dns_addr, &pData->pszDomain) < 0)
434	#else
435	LIST_INIT(&pData->dns_list_head);
436	if (get_dns_addr_domain(pData, true, NULL, &pData->pszDomain) < 0)
437	#endif
438	fNATfailed = 1;
439
440	getouraddr(pData);
441	return fNATfailed ? VINF_NAT_DNS : VINF_SUCCESS;
442	}
443
444	/**
445	* Statistics counters.
446	*/
447	void slirp_register_timers(PNATState pData, PPDMDRVINS pDrvIns)
448	{
449	#ifdef VBOX_WITH_STATISTICS
450	PDMDrvHlpSTAMRegisterF(pDrvIns, &pData->StatFill, STAMTYPE_PROFILE, STAMVISIBILITY_ALWAYS,
451	STAMUNIT_TICKS_PER_CALL, "Profiling slirp fills", "/Drivers/NAT%d/Fill", pDrvIns->iInstance);
452	PDMDrvHlpSTAMRegisterF(pDrvIns, &pData->StatPoll, STAMTYPE_PROFILE, STAMVISIBILITY_ALWAYS,
453	STAMUNIT_TICKS_PER_CALL, "Profiling slirp polls", "/Drivers/NAT%d/Poll", pDrvIns->iInstance);
454	PDMDrvHlpSTAMRegisterF(pDrvIns, &pData->StatFastTimer, STAMTYPE_PROFILE, STAMVISIBILITY_ALWAYS,
455	STAMUNIT_TICKS_PER_CALL, "Profiling slirp fast timer", "/Drivers/NAT%d/TimerFast", pDrvIns->iInstance);
456	PDMDrvHlpSTAMRegisterF(pDrvIns, &pData->StatSlowTimer, STAMTYPE_PROFILE, STAMVISIBILITY_ALWAYS,
457	STAMUNIT_TICKS_PER_CALL, "Profiling slirp slow timer", "/Drivers/NAT%d/TimerSlow", pDrvIns->iInstance);
458	PDMDrvHlpSTAMRegisterF(pDrvIns, &pData->StatTCP, STAMTYPE_COUNTER, STAMVISIBILITY_ALWAYS,
459	STAMUNIT_COUNT, "TCP sockets", "/Drivers/NAT%d/SockTCP", pDrvIns->iInstance);
460	PDMDrvHlpSTAMRegisterF(pDrvIns, &pData->StatTCPHot, STAMTYPE_COUNTER, STAMVISIBILITY_ALWAYS,
461	STAMUNIT_COUNT, "TCP sockets active", "/Drivers/NAT%d/SockTCPHot", pDrvIns->iInstance);
462	PDMDrvHlpSTAMRegisterF(pDrvIns, &pData->StatUDP, STAMTYPE_COUNTER, STAMVISIBILITY_ALWAYS,
463	STAMUNIT_COUNT, "UDP sockets", "/Drivers/NAT%d/SockUDP", pDrvIns->iInstance);
464	PDMDrvHlpSTAMRegisterF(pDrvIns, &pData->StatUDPHot, STAMTYPE_COUNTER, STAMVISIBILITY_ALWAYS,
465	STAMUNIT_COUNT, "UDP sockets active", "/Drivers/NAT%d/SockUDPHot", pDrvIns->iInstance);
466	#endif /* VBOX_WITH_STATISTICS */
467	}
468
469	/**
470	* Marks the link as up, making it possible to establish new connections.
471	*/
472	void slirp_link_up(PNATState pData)
473	{
474	link_up = 1;
475	}
476
477	/**
478	* Marks the link as down and cleans up the current connections.
479	*/
480	void slirp_link_down(PNATState pData)
481	{
482	struct socket *so;
483
484	while ((so = tcb.so_next) != &tcb)
485	{
486	if (so->so_state & SS_NOFDREF \|\| so->s == -1)
487	sofree(pData, so);
488	else
489	tcp_drop(pData, sototcpcb(so), 0);
490	}
491
492	while ((so = udb.so_next) != &udb)
493	udp_detach(pData, so);
494
495	link_up = 0;
496	}
497
498	/**
499	* Terminates the slirp component.
500	*/
501	void slirp_term(PNATState pData)
502	{
503	#ifdef VBOX_WITH_MULTI_DNS
504	struct dns_entry *de = NULL;
505	#endif
506	if (pData->pszDomain)
507	RTStrFree((char )(void )pData->pszDomain);
508
509	#ifdef RT_OS_WINDOWS
510	pData->pfIcmpCloseHandle(pData->icmp_socket.sh);
511	FreeLibrary(pData->hmIcmpLibrary);
512	RTMemFree(pData->pvIcmpBuffer);
513	# else
514	closesocket(pData->icmp_socket.s);
515	#endif
516
517	slirp_link_down(pData);
518	#ifdef VBOX_WITH_MULTI_DNS
519	while(!LIST_EMPTY(&pData->dns_list_head)) {
520	de = LIST_FIRST(&pData->dns_list_head);
521	LIST_REMOVE(de, de_list);
522	RTMemFree(de);
523	}
524	#endif
525	#ifdef RT_OS_WINDOWS
526	WSACleanup();
527	#endif
528	#ifdef LOG_ENABLED
529	Log(("\n"
530	"NAT statistics\n"
531	"--------------\n"
532	"\n"));
533	ipstats(pData);
534	tcpstats(pData);
535	udpstats(pData);
536	icmpstats(pData);
537	mbufstats(pData);
538	sockstats(pData);
539	Log(("\n"
540	"\n"
541	"\n"));
542	#endif
543	RTMemFree(pData);
544	}
545
546
547	#define CONN_CANFSEND(so) (((so)->so_state & (SS_FCANTSENDMORE\|SS_ISFCONNECTED)) == SS_ISFCONNECTED)
548	#define CONN_CANFRCV(so) (((so)->so_state & (SS_FCANTRCVMORE\|SS_ISFCONNECTED)) == SS_ISFCONNECTED)
549	#define UPD_NFDS(x) if (nfds < (x)) nfds = (x)
550
551	/*
552	* curtime kept to an accuracy of 1ms
553	*/
554	#ifdef RT_OS_WINDOWS
555	static void updtime(PNATState pData)
556	{
557	struct _timeb tb;
558
559	_ftime(&tb);
560	curtime = (u_int)tb.time * (u_int)1000;
561	curtime += (u_int)tb.millitm;
562	}
563	#else
564	static void updtime(PNATState pData)
565	{
566	gettimeofday(&tt, 0);
567
568	curtime = (u_int)tt.tv_sec * (u_int)1000;
569	curtime += (u_int)tt.tv_usec / (u_int)1000;
570
571	if ((tt.tv_usec % 1000) >= 500)
572	curtime++;
573	}
574	#endif
575
576	void slirp_select_fill(PNATState pData, int *pnfds,
577	fd_set readfds, fd_set writefds, fd_set *xfds)
578	{
579	struct socket so, so_next;
580	int nfds;
581	#if defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) && defined(RT_OS_WINDOWS)
582	int rc;
583	int error;
584	#endif
585	int i;
586
587	STAM_PROFILE_START(&pData->StatFill, a);
588
589	nfds = *pnfds;
590
591	/*
592	* First, TCP sockets
593	*/
594	do_slowtimo = 0;
595	if (link_up)
596	{
597	/*
598	* *_slowtimo needs calling if there are IP fragments
599	* in the fragment queue, or there are TCP connections active
600	*/
601	/* XXX:
602	* triggering of fragment expiration should be the same but use new macroses
603	*/
604	do_slowtimo = (tcb.so_next != &tcb);
605	if (!do_slowtimo)
606	{
607	for (i = 0; i < IPREASS_NHASH; i++)
608	{
609	if (!TAILQ_EMPTY(&ipq[i]))
610	{
611	do_slowtimo = 1;
612	break;
613	}
614	}
615	}
616	ICMP_ENGAGE_EVENT(&pData->icmp_socket, readfds);
617
618	STAM_COUNTER_RESET(&pData->StatTCP);
619	STAM_COUNTER_RESET(&pData->StatTCPHot);
620
621	QSOCKET_FOREACH(so, so_next, tcp)
622	/* { */
623	STAM_COUNTER_INC(&pData->StatTCP);
624
625	/*
626	* See if we need a tcp_fasttimo
627	*/
628	if (time_fasttimo == 0 && so->so_tcpcb->t_flags & TF_DELACK)
629	time_fasttimo = curtime; /* Flag when we want a fasttimo */
630
631	/*
632	* NOFDREF can include still connecting to local-host,
633	* newly socreated() sockets etc. Don't want to select these.
634	*/
635	if (so->so_state & SS_NOFDREF \|\| so->s == -1)
636	CONTINUE(tcp);
637
638	/*
639	* Set for reading sockets which are accepting
640	*/
641	if (so->so_state & SS_FACCEPTCONN)
642	{
643	STAM_COUNTER_INC(&pData->StatTCPHot);
644	TCP_ENGAGE_EVENT1(so, readfds);
645	CONTINUE(tcp);
646	}
647
648	/*
649	* Set for writing sockets which are connecting
650	*/
651	if (so->so_state & SS_ISFCONNECTING)
652	{
653	Log2(("connecting %R[natsock] engaged\n",so));
654	STAM_COUNTER_INC(&pData->StatTCPHot);
655	TCP_ENGAGE_EVENT1(so, writefds);
656	}
657
658	/*
659	* Set for writing if we are connected, can send more, and
660	* we have something to send
661	*/
662	if (CONN_CANFSEND(so) && so->so_rcv.sb_cc)
663	{
664	STAM_COUNTER_INC(&pData->StatTCPHot);
665	TCP_ENGAGE_EVENT1(so, writefds);
666	}
667
668	/*
669	* Set for reading (and urgent data) if we are connected, can
670	* receive more, and we have room for it XXX /2 ?
671	*/
672	if (CONN_CANFRCV(so) && (so->so_snd.sb_cc < (so->so_snd.sb_datalen/2)))
673	{
674	STAM_COUNTER_INC(&pData->StatTCPHot);
675	TCP_ENGAGE_EVENT2(so, readfds, xfds);
676	}
677	LOOP_LABEL(tcp, so, so_next);
678	}
679
680	/*
681	* UDP sockets
682	*/
683	STAM_COUNTER_RESET(&pData->StatUDP);
684	STAM_COUNTER_RESET(&pData->StatUDPHot);
685
686	QSOCKET_FOREACH(so, so_next, udp)
687	/* { */
688	so_next = so->so_next;
689
690	STAM_COUNTER_INC(&pData->StatUDP);
691
692	/*
693	* See if it's timed out
694	*/
695	if (so->so_expire)
696	{
697	if (so->so_expire <= curtime)
698	{
699	udp_detach(pData, so);
700	CONTINUE(udp);
701	}
702	else
703	do_slowtimo = 1; /* Let socket expire */
704	}
705
706	/*
707	* When UDP packets are received from over the link, they're
708	* sendto()'d straight away, so no need for setting for writing
709	* Limit the number of packets queued by this session to 4.
710	* Note that even though we try and limit this to 4 packets,
711	* the session could have more queued if the packets needed
712	* to be fragmented.
713	*
714	* (XXX <= 4 ?)
715	*/
716	if ((so->so_state & SS_ISFCONNECTED) && so->so_queued <= 4)
717	{
718	STAM_COUNTER_INC(&pData->StatUDPHot);
719	UDP_ENGAGE_EVENT(so, readfds);
720	}
721	LOOP_LABEL(udp, so, so_next);
722	}
723
724	}
725
726	#if !defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) \|\| !defined(RT_OS_WINDOWS)
727	*pnfds = nfds;
728	#else
729	*pnfds = VBOX_EVENT_COUNT;
730	#endif
731
732	STAM_PROFILE_STOP(&pData->StatFill, a);
733	}
734
735	#if defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) && defined(RT_OS_WINDOWS)
736	void slirp_select_poll(PNATState pData, int fTimeout, int fIcmp)
737	#else
738	void slirp_select_poll(PNATState pData, fd_set readfds, fd_set writefds, fd_set *xfds)
739	#endif
740	{
741	struct socket so, so_next;
742	int ret;
743	#if defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) && defined(RT_OS_WINDOWS)
744	WSANETWORKEVENTS NetworkEvents;
745	int rc;
746	int error;
747	#endif
748
749	STAM_PROFILE_START(&pData->StatPoll, a);
750
751	/* Update time */
752	updtime(pData);
753
754	/*
755	* See if anything has timed out
756	*/
757	if (link_up)
758	{
759	if (time_fasttimo && ((curtime - time_fasttimo) >= 2))
760	{
761	STAM_PROFILE_START(&pData->StatFastTimer, a);
762	tcp_fasttimo(pData);
763	time_fasttimo = 0;
764	STAM_PROFILE_STOP(&pData->StatFastTimer, a);
765	}
766	if (do_slowtimo && ((curtime - last_slowtimo) >= 499))
767	{
768	STAM_PROFILE_START(&pData->StatSlowTimer, a);
769	ip_slowtimo(pData);
770	tcp_slowtimo(pData);
771	last_slowtimo = curtime;
772	STAM_PROFILE_STOP(&pData->StatSlowTimer, a);
773	}
774	}
775	#if defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) && defined(RT_OS_WINDOWS)
776	if (fTimeout)
777	return; /* only timer update */
778	#endif
779
780	/*
781	* Check sockets
782	*/
783	if (link_up)
784	{
785	#if defined(RT_OS_WINDOWS)
786	/*XXX: before renaming please make see define
787	* fIcmp in slirp_state.h
788	*/
789	if (fIcmp)
790	sorecvfrom(pData, &pData->icmp_socket);
791	#else
792	if (pData->icmp_socket.s != -1 && FD_ISSET(pData->icmp_socket.s, readfds))
793	sorecvfrom(pData, &pData->icmp_socket);
794	#endif
795	/*
796	* Check TCP sockets
797	*/
798	for (so = tcb.so_next; so != &tcb; so = so_next)
799	{
800	so_next = so->so_next;
801
802	/*
803	* FD_ISSET is meaningless on these sockets
804	* (and they can crash the program)
805	*/
806	if (so->so_state & SS_NOFDREF \|\| so->s == -1)
807	CONTINUE(tcp);
808
809	POLL_TCP_EVENTS(rc, error, so, &NetworkEvents);
810
811	LOG_NAT_SOCK(so, TCP, &NetworkEvents, readfds, writefds, xfds);
812
813	/*
814	* Check for URG data
815	* This will soread as well, so no need to
816	* test for readfds below if this succeeds
817	*/
818
819	/* out-of-band data */
820	if (CHECK_FD_SET(so, NetworkEvents, xfds))
821	{
822	sorecvoob(pData, so);
823	}
824
825	/*
826	* Check sockets for reading
827	*/
828	else if ( CHECK_FD_SET(so, NetworkEvents, readfds)
829	\|\| WIN_CHECK_FD_SET(so, NetworkEvents, acceptds))
830	{
831	/*
832	* Check for incoming connections
833	*/
834	if (so->so_state & SS_FACCEPTCONN)
835	{
836	tcp_connect(pData, so);
837	#if defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) && defined(RT_OS_WINDOWS)
838	if (!(NetworkEvents.lNetworkEvents & FD_CLOSE))
839	#endif
840	CONTINUE(tcp);
841	}
842
843	ret = soread(pData, so, /fCloseIfNothingRead=/false);
844	/* Output it if we read something */
845	if (ret > 0)
846	tcp_output(pData, sototcpcb(so));
847	}
848
849	#if defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) && defined(RT_OS_WINDOWS)
850	/*
851	* Check for FD_CLOSE events.
852	*/
853	if (NetworkEvents.lNetworkEvents & FD_CLOSE)
854	{
855	/*
856	* drain the socket
857	*/
858	for (;;)
859	{
860	ret = soread(pData, so, /fCloseIfNothingRead=/true);
861	if (ret > 0)
862	tcp_output(pData, sototcpcb(so));
863	else
864	break;
865	}
866	}
867	#endif
868
869	/*
870	* Check sockets for writing
871	*/
872	if (CHECK_FD_SET(so, NetworkEvents, writefds))
873	{
874	/*
875	* Check for non-blocking, still-connecting sockets
876	*/
877	if (so->so_state & SS_ISFCONNECTING)
878	{
879	Log2(("connecting %R[natsock] catched\n", so));
880	/* Connected */
881	so->so_state &= ~SS_ISFCONNECTING;
882
883	/*
884	* This should be probably guarded by PROBE_CONN too. Anyway,
885	* we disable it on OS/2 because the below send call returns
886	* EFAULT which causes the opened TCP socket to close right
887	* after it has been opened and connected.
888	*/
889	#ifndef RT_OS_OS2
890	ret = send(so->s, (const char *)&ret, 0, 0);
891	if (ret < 0)
892	{
893	/* XXXXX Must fix, zero bytes is a NOP */
894	if ( errno == EAGAIN
895	\|\| errno == EWOULDBLOCK
896	\|\| errno == EINPROGRESS
897	\|\| errno == ENOTCONN)
898	CONTINUE(tcp);
899
900	/* else failed */
901	so->so_state = SS_NOFDREF;
902	}
903	/* else so->so_state &= ~SS_ISFCONNECTING; */
904	#endif
905
906	/*
907	* Continue tcp_input
908	*/
909	tcp_input(pData, (struct mbuf *)NULL, sizeof(struct ip), so);
910	/* continue; */
911	}
912	else
913	ret = sowrite(pData, so);
914	/*
915	* XXX If we wrote something (a lot), there could be the need
916	* for a window update. In the worst case, the remote will send
917	* a window probe to get things going again.
918	*/
919	}
920
921	/*
922	* Probe a still-connecting, non-blocking socket
923	* to check if it's still alive
924	*/
925	#ifdef PROBE_CONN
926	if (so->so_state & SS_ISFCONNECTING)
927	{
928	ret = recv(so->s, (char *)&ret, 0, 0);
929
930	if (ret < 0)
931	{
932	/* XXX */
933	if ( errno == EAGAIN
934	\|\| errno == EWOULDBLOCK
935	\|\| errno == EINPROGRESS
936	\|\| errno == ENOTCONN)
937	{
938	CONTINUE(tcp); /* Still connecting, continue */
939	}
940
941	/* else failed */
942	so->so_state = SS_NOFDREF;
943
944	/* tcp_input will take care of it */
945	}
946	else
947	{
948	ret = send(so->s, &ret, 0, 0);
949	if (ret < 0)
950	{
951	/* XXX */
952	if ( errno == EAGAIN
953	\|\| errno == EWOULDBLOCK
954	\|\| errno == EINPROGRESS
955	\|\| errno == ENOTCONN)
956	{
957	CONTINUE(tcp);
958	}
959	/* else failed */
960	so->so_state = SS_NOFDREF;
961	}
962	else
963	so->so_state &= ~SS_ISFCONNECTING;
964
965	}
966	tcp_input((struct mbuf *)NULL, sizeof(struct ip),so);
967	} /* SS_ISFCONNECTING */
968	#endif
969	LOOP_LABEL(tcp, so, so_next);
970	}
971
972	/*
973	* Now UDP sockets.
974	* Incoming packets are sent straight away, they're not buffered.
975	* Incoming UDP data isn't buffered either.
976	*/
977	for (so = udb.so_next; so != &udb; so = so_next)
978	{
979	so_next = so->so_next;
980
981	POLL_UDP_EVENTS(rc, error, so, &NetworkEvents);
982
983	LOG_NAT_SOCK(so, UDP, &NetworkEvents, readfds, writefds, xfds);
984
985	if (so->s != -1 && CHECK_FD_SET(so, NetworkEvents, readfds))
986	{
987	sorecvfrom(pData, so);
988	}
989	LOOP_LABEL(udp, so, so_next);
990	}
991
992	}
993
994	#ifndef VBOX_WITH_SLIRP_MT
995	/*
996	* See if we can start outputting
997	*/
998	if (if_queued && link_up)
999	if_start(pData);
1000	#endif
1001
1002	STAM_PROFILE_STOP(&pData->StatPoll, a);
1003	}
1004
1005	#define ETH_ALEN 6
1006	#define ETH_HLEN 14
1007
1008	#define ARPOP_REQUEST 1 /* ARP request */
1009	#define ARPOP_REPLY 2 /* ARP reply */
1010
1011	struct ethhdr
1012	{
1013	unsigned char h_dest[ETH_ALEN]; /* destination eth addr */
1014	unsigned char h_source[ETH_ALEN]; /* source ether addr */
1015	unsigned short h_proto; /* packet type ID field */
1016	};
1017
1018	struct arphdr
1019	{
1020	unsigned short ar_hrd; /* format of hardware address */
1021	unsigned short ar_pro; /* format of protocol address */
1022	unsigned char ar_hln; /* length of hardware address */
1023	unsigned char ar_pln; /* length of protocol address */
1024	unsigned short ar_op; /* ARP opcode (command) */
1025
1026	/*
1027	* Ethernet looks like this : This bit is variable sized however...
1028	*/
1029	unsigned char ar_sha[ETH_ALEN]; /* sender hardware address */
1030	unsigned char ar_sip[4]; /* sender IP address */
1031	unsigned char ar_tha[ETH_ALEN]; /* target hardware address */
1032	unsigned char ar_tip[4]; /* target IP address */
1033	};
1034
1035	static
1036	#ifdef VBOX_WITH_SIMPLIFIED_SLIRP_SYNC
1037	void arp_input(PNATState pData, struct mbuf *m)
1038	#else
1039	void arp_input(PNATState pData, const uint8_t *pkt, int pkt_len)
1040	#endif
1041	{
1042	struct ethhdr *eh;
1043	struct ethhdr *reh;
1044	struct arphdr *ah;
1045	struct arphdr *rah;
1046	int ar_op;
1047	struct ex_list *ex_ptr;
1048	uint32_t htip;
1049	#ifndef VBOX_WITH_SIMPLIFIED_SLIRP_SYNC
1050	uint8_t arp_reply[sizeof(struct arphdr) + ETH_HLEN];
1051	eh = (struct ethhdr *)pkt;
1052	#else
1053	struct mbuf *mr;
1054	eh = mtod(m, struct ethhdr *);
1055	#endif
1056	ah = (struct arphdr *)&eh[1];
1057	htip = ntohl((uint32_t)ah->ar_tip);
1058
1059	#ifdef VBOX_WITH_SIMPLIFIED_SLIRP_SYNC
1060	mr = m_get(pData);
1061	mr->m_data += if_maxlinkhdr;
1062	mr->m_len = sizeof(struct arphdr);
1063	rah = mtod(mr, struct arphdr *);
1064	#else
1065	reh = (struct ethhdr *)arp_reply;
1066	rah = (struct arphdr *)&reh[1];
1067	#endif
1068
1069	ar_op = ntohs(ah->ar_op);
1070	switch(ar_op)
1071	{
1072	case ARPOP_REQUEST:
1073	if ((htip & pData->netmask) == ntohl(special_addr.s_addr))
1074	{
1075	if ( CTL_CHECK(htip,CTL_DNS)
1076	\|\| CTL_CHECK(htip, CTL_ALIAS)
1077	\|\| CTL_CHECK(htip, CTL_TFTP))
1078	goto arp_ok;
1079	for (ex_ptr = exec_list; ex_ptr; ex_ptr = ex_ptr->ex_next)
1080	{
1081	if ((htip & ~pData->netmask) == ex_ptr->ex_addr)
1082	goto arp_ok;
1083	}
1084	return;
1085	arp_ok:
1086
1087	#ifndef VBOX_WITH_SIMPLIFIED_SLIRP_SYNC
1088	memcpy(reh->h_dest, eh->h_source, ETH_ALEN);
1089	memcpy(reh->h_source, &special_addr, ETH_ALEN);
1090	reh->h_source[5] = ah->ar_tip[3];
1091	reh->h_proto = htons(ETH_P_ARP);
1092	#endif
1093	rah->ar_hrd = htons(1);
1094	rah->ar_pro = htons(ETH_P_IP);
1095	rah->ar_hln = ETH_ALEN;
1096	rah->ar_pln = 4;
1097	rah->ar_op = htons(ARPOP_REPLY);
1098	memcpy(rah->ar_sha, special_ethaddr, ETH_ALEN);
1099
1100	switch (htip & ~pData->netmask)
1101	{
1102	case CTL_DNS:
1103	case CTL_ALIAS:
1104	rah->ar_sha[5] = (uint8_t)(htip & ~pData->netmask);
1105	break;
1106	default:;
1107	}
1108
1109	memcpy(rah->ar_sip, ah->ar_tip, 4);
1110	memcpy(rah->ar_tha, ah->ar_sha, ETH_ALEN);
1111	memcpy(rah->ar_tip, ah->ar_sip, 4);
1112	#ifdef VBOX_WITH_SIMPLIFIED_SLIRP_SYNC
1113	if_encap(pData, ETH_P_ARP, mr);
1114	m_free(pData, m);
1115	#else
1116	slirp_output(pData->pvUser, arp_reply, sizeof(arp_reply));
1117	#endif
1118	}
1119	break;
1120	default:
1121	break;
1122	}
1123	}
1124
1125	void slirp_input(PNATState pData, const uint8_t *pkt, int pkt_len)
1126	{
1127	struct mbuf *m;
1128	int proto;
1129	static bool fWarnedIpv6;
1130
1131	if (pkt_len < ETH_HLEN)
1132	{
1133	LogRel(("NAT: packet having size %d has been ingnored\n", pkt_len));
1134	return;
1135	}
1136
1137	m = m_get(pData);
1138	if (!m)
1139	{
1140	LogRel(("can't allocate new mbuf\n"));
1141	return;
1142	}
1143
1144	/* Note: we add to align the IP header */
1145
1146	if (M_FREEROOM(m) < pkt_len)
1147	m_inc(m, pkt_len);
1148
1149	m->m_len = pkt_len ;
1150	memcpy(m->m_data, pkt, pkt_len);
1151
1152	proto = ntohs((uint16_t )(pkt + 12));
1153	switch(proto)
1154	{
1155	case ETH_P_ARP:
1156	#ifdef VBOX_WITH_SIMPLIFIED_SLIRP_SYNC
1157	arp_input(pData, m);
1158	#else
1159	arp_input(pData, pkt, pkt_len);
1160	m_free(pData, m);
1161	#endif
1162	break;
1163	case ETH_P_IP:
1164	/* Update time. Important if the network is very quiet, as otherwise
1165	* the first outgoing connection gets an incorrect timestamp. */
1166	updtime(pData);
1167	m->m_data += ETH_HLEN;
1168	m->m_len -= ETH_HLEN;
1169	ip_input(pData, m);
1170	break;
1171	case ETH_P_IPV6:
1172	m_free(pData, m);
1173	if (!fWarnedIpv6)
1174	{
1175	LogRel(("NAT: IPv6 not supported\n"));
1176	fWarnedIpv6 = true;
1177	}
1178	break;
1179	default:
1180	LogRel(("NAT: Unsupported protocol %x\n", proto));
1181	m_free(pData, m);
1182	break;
1183	}
1184	}
1185
1186	/* output the IP packet to the ethernet device */
1187	#ifdef VBOX_WITH_SIMPLIFIED_SLIRP_SYNC
1188	void if_encap(PNATState pData, uint16_t eth_proto, struct mbuf *m)
1189	#else
1190	void if_encap(PNATState pData, uint8_t *ip_data, int ip_data_len)
1191	#endif
1192	{
1193	#ifdef VBOX_WITH_SIMPLIFIED_SLIRP_SYNC
1194	struct ethhdr *eh;
1195	uint8_t *buf = RTMemAlloc(1600);
1196	m->m_data -= if_maxlinkhdr;
1197	m->m_len += ETH_HLEN;
1198	eh = mtod(m, struct ethhdr *);
1199	#else
1200	uint8_t buf[1600];
1201	struct ethhdr eh = (struct ethhdr )buf;
1202
1203	if (ip_data_len + ETH_HLEN > sizeof(buf))
1204	return;
1205
1206	memcpy(buf + sizeof(struct ethhdr), ip_data, ip_data_len);
1207	#endif
1208
1209
1210	memcpy(eh->h_dest, client_ethaddr, ETH_ALEN);
1211	memcpy(eh->h_source, special_ethaddr, ETH_ALEN - 1);
1212	/* XXX: not correct */
1213	eh->h_source[5] = CTL_ALIAS;
1214	#ifdef VBOX_WITH_SIMPLIFIED_SLIRP_SYNC
1215	eh->h_proto = htons(eth_proto);
1216	#if 0
1217	slirp_output(pData->pvUser, m, mtod(m, uint8_t *), m->m_len);
1218	#else
1219	memcpy(buf, mtod(m, uint8_t *), m->m_len);
1220	slirp_output(pData->pvUser, NULL, buf, m->m_len);
1221	m_free(pData, m);
1222	#endif
1223	#else
1224	eh->h_proto = htons(ETH_P_IP);
1225	slirp_output(pData->pvUser, buf, ip_data_len + ETH_HLEN);
1226	#endif
1227	}
1228
1229	int slirp_redir(PNATState pData, int is_udp, int host_port,
1230	struct in_addr guest_addr, int guest_port)
1231	{
1232	if (is_udp)
1233	{
1234	if (!udp_listen(pData, htons(host_port), guest_addr.s_addr,
1235	htons(guest_port), 0))
1236	return -1;
1237	}
1238	else
1239	{
1240	if (!solisten(pData, htons(host_port), guest_addr.s_addr,
1241	htons(guest_port), 0))
1242	return -1;
1243	}
1244	return 0;
1245	}
1246
1247	int slirp_add_exec(PNATState pData, int do_pty, const char *args, int addr_low_byte,
1248	int guest_port)
1249	{
1250	return add_exec(&exec_list, do_pty, (char *)args,
1251	addr_low_byte, htons(guest_port));
1252	}
1253
1254	void slirp_set_ethaddr(PNATState pData, const uint8_t *ethaddr)
1255	{
1256	memcpy(client_ethaddr, ethaddr, ETH_ALEN);
1257	}
1258
1259	#if defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) && defined(RT_OS_WINDOWS)
1260	HANDLE *slirp_get_events(PNATState pData)
1261	{
1262	return pData->phEvents;
1263	}
1264	void slirp_register_external_event(PNATState pData, HANDLE hEvent, int index)
1265	{
1266	pData->phEvents[index] = hEvent;
1267	}
1268	#endif
1269
1270	unsigned int slirp_get_timeout_ms(PNATState pData)
1271	{
1272	if (link_up)
1273	{
1274	if (time_fasttimo)
1275	return 2;
1276	if (do_slowtimo)
1277	return 500; /* see PR_SLOWHZ */
1278	}
1279	return 0;
1280	}
1281
1282	/*
1283	* this function called from NAT thread
1284	*/
1285	void slirp_post_sent(PNATState pData, void *pvArg)
1286	{
1287	struct socket *so = 0;
1288	struct tcpcb *tp = 0;
1289	struct mbuf m = (struct mbuf )pvArg;
1290	m_free(pData, m);
1291	}

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

source: vbox/trunk/src/VBox/Devices/Network/slirp/slirp.c@ 16303

Download in other formats: