slirp.c@ 14200

Last change on this file since 14200 was 14200, checked in by vboxsync, 16 years ago
try to fix the slirp build (PVM define on Mac)
Property svn:eol-style set to `native`
File size: 29.9 KB

Line
1	#include "slirp.h"
2	#ifdef RT_OS_OS2
3	# include <paths.h>
4	#endif
5
6	/* disable these counters for the final release */
7	/* #define VBOX_WITHOUT_RELEASE_STATISTICS */
8
9	#include <VBox/err.h>
10	#include <VBox/pdmdrv.h>
11	#include <iprt/assert.h>
12
13	static const uint8_t special_ethaddr[6] = {
14	0x52, 0x54, 0x00, 0x12, 0x35, 0x00
15	};
16
17	#ifdef _WIN32
18
19	static int get_dns_addr_domain(PNATState pData, bool fVerbose,
20	struct in_addr *pdns_addr,
21	const char **ppszDomain)
22	{
23	int rc = 0;
24	FIXED_INFO *FixedInfo=NULL;
25	ULONG BufLen;
26	DWORD ret;
27	IP_ADDR_STRING *pIPAddr;
28	struct in_addr tmp_addr;
29
30	FixedInfo = (FIXED_INFO *)GlobalAlloc(GPTR, sizeof(FIXED_INFO));
31	BufLen = sizeof(FIXED_INFO);
32
33	/** @todo: this API returns all DNS servers, no matter whether the
34	* corresponding network adapter is disabled or not. Maybe replace
35	* this by GetAdapterAddresses(), which is XP/Vista only though. */
36	if (ERROR_BUFFER_OVERFLOW == GetNetworkParams(FixedInfo, &BufLen)) {
37	if (FixedInfo) {
38	GlobalFree(FixedInfo);
39	FixedInfo = NULL;
40	}
41	FixedInfo = GlobalAlloc(GPTR, BufLen);
42	}
43
44	if ((ret = GetNetworkParams(FixedInfo, &BufLen)) != ERROR_SUCCESS) {
45	Log(("GetNetworkParams failed. ret = %08x\n", (u_int)ret ));
46	if (FixedInfo) {
47	GlobalFree(FixedInfo);
48	FixedInfo = NULL;
49	}
50	rc = -1;
51	goto get_dns_prefix;
52	}
53
54	pIPAddr = &(FixedInfo->DnsServerList);
55	inet_aton(pIPAddr->IpAddress.String, &tmp_addr);
56	Log(("nat: DNS Servers:\n"));
57	if (fVerbose \|\| pdns_addr->s_addr != tmp_addr.s_addr)
58	LogRel(("NAT: DNS address: %s\n", pIPAddr->IpAddress.String));
59	*pdns_addr = tmp_addr;
60
61	pIPAddr = FixedInfo -> DnsServerList.Next;
62	while ( pIPAddr )
63	{
64	if (fVerbose)
65	LogRel(("NAT: ignored DNS address: %s\n", pIPAddr ->IpAddress.String));
66	pIPAddr = pIPAddr ->Next;
67	}
68	if (FixedInfo) {
69	GlobalFree(FixedInfo);
70	FixedInfo = NULL;
71	}
72
73	get_dns_prefix:
74	if (ppszDomain)
75	{
76	OSVERSIONINFO ver;
77	char szDnsDomain[256];
78	DWORD dwSize = sizeof(szDnsDomain);
79
80	*ppszDomain = NULL;
81	GetVersionEx(&ver);
82	if (ver.dwMajorVersion >= 5)
83	{
84	/* GetComputerNameEx exists in Windows versions starting with 2000. */
85	if (GetComputerNameEx(ComputerNameDnsDomain, szDnsDomain, &dwSize))
86	{
87	if (szDnsDomain[0])
88	{
89	/* Just non-empty strings are valid. */
90	*ppszDomain = RTStrDup(szDnsDomain);
91	if (pData->fPassDomain)
92	{
93	if (fVerbose)
94	LogRel(("NAT: passing domain name %s\n", szDnsDomain));
95	}
96	else
97	Log(("nat: ignoring domain %s\n", szDnsDomain));
98	}
99	}
100	else
101	Log(("nat: GetComputerNameEx failed (%d)\n", GetLastError()));
102	}
103	}
104	return rc;
105	}
106
107	#else
108
109	static int get_dns_addr_domain(PNATState pData, bool fVerbose,
110	struct in_addr *pdns_addr,
111	const char **ppszDomain)
112	{
113	char buff[512];
114	char buff2[256];
115	FILE *f;
116	int found = 0;
117	struct in_addr tmp_addr;
118
119	#ifdef RT_OS_OS2
120	/* Try various locations. */
121	char *etc = getenv("ETC");
122	f = NULL;
123	if (etc)
124	{
125	snprintf(buff, sizeof(buff), "%s/RESOLV2", etc);
126	f = fopen(buff, "rt");
127	}
128	if (!f) {
129	snprintf(buff, sizeof(buff), "%s/RESOLV2", _PATH_ETC);
130	f = fopen(buff, "rt");
131	}
132	if (!f) {
133	snprintf(buff, sizeof(buff), "%s/resolv.conf", _PATH_ETC);
134	f = fopen(buff, "rt");
135	}
136	#else
137	f = fopen("/etc/resolv.conf", "r");
138	#endif
139	if (!f)
140	return -1;
141
142	if (ppszDomain)
143	*ppszDomain = NULL;
144	Log(("nat: DNS Servers:\n"));
145	while (fgets(buff, 512, f) != NULL) {
146	if (sscanf(buff, "nameserver%*[ \t]%256s", buff2) == 1) {
147	if (!inet_aton(buff2, &tmp_addr))
148	continue;
149	if (tmp_addr.s_addr == loopback_addr.s_addr)
150	tmp_addr = our_addr;
151	/* If it's the first one, set it to dns_addr */
152	if (!found)
153	{
154	if (fVerbose \|\| pdns_addr->s_addr != tmp_addr.s_addr)
155	LogRel(("NAT: DNS address: %s\n", buff2));
156	*pdns_addr = tmp_addr;
157	}
158	else
159	{
160	if (fVerbose)
161	LogRel(("NAT: ignored DNS address: %s\n", buff2));
162	}
163	found++;
164	}
165	if ( ppszDomain
166	&& (!strncmp(buff, "domain", 6) \|\| !strncmp(buff, "search", 6)))
167	{
168	/* Domain name/search list present. Pick first entry */
169	if (*ppszDomain == NULL)
170	{
171	char *tok;
172	char *saveptr;
173	tok = strtok_r(&buff[6], " \t\n", &saveptr);
174	if (tok)
175	{
176	*ppszDomain = RTStrDup(tok);
177	if (pData->fPassDomain)
178	{
179	if (fVerbose)
180	LogRel(("NAT: passing domain name %s\n", tok));
181	}
182	else
183	Log(("nat: ignoring domain %s\n", tok));
184	}
185	}
186	}
187	}
188	fclose(f);
189	if (!found)
190	return -1;
191	return 0;
192	}
193
194	#endif
195
196	int get_dns_addr(PNATState pData, struct in_addr *pdns_addr)
197	{
198	return get_dns_addr_domain(pData, false, pdns_addr, NULL);
199	}
200
201	int slirp_init(PNATState ppData, const char pszNetAddr, uint32_t u32Netmask,
202	bool fPassDomain, const char *pszTFTPPrefix,
203	const char pszBootFile, void pvUser)
204	{
205	int fNATfailed = 0;
206	PNATState pData = malloc(sizeof(NATState));
207	*ppData = pData;
208	if (!pData)
209	return VERR_NO_MEMORY;
210	if (u32Netmask & 0x1f)
211	/* CTL is x.x.x.15, bootp passes up to 16 IPs (15..31) */
212	return VERR_INVALID_PARAMETER;
213	memset(pData, '\0', sizeof(NATState));
214	pData->fPassDomain = fPassDomain;
215	pData->pvUser = pvUser;
216	#if ARCH_BITS == 64
217	pData->cpvHashUsed = 1;
218	#endif
219	tftp_prefix = pszTFTPPrefix;
220	bootp_filename = pszBootFile;
221	pData->netmask = u32Netmask;
222
223	#ifdef _WIN32
224	{
225	WSADATA Data;
226	WSAStartup(MAKEWORD(2,0), &Data);
227	}
228	#if defined(VBOX_WITH_SIMPLEFIED_SLIRP_SYNC) && defined(RT_OS_WINDOWS)
229	pData->phEvents[VBOX_SOCKET_EVENT_INDEX] = CreateEvent(NULL, FALSE, FALSE, NULL);
230	#endif
231	#endif
232
233	Assert(sizeof(struct ip) == 20);
234	link_up = 1;
235
236	if_init(pData);
237	ip_init(pData);
238
239	/* Initialise mbufs after setting the MTU */
240	m_init(pData);
241
242	/* set default addresses */
243	inet_aton("127.0.0.1", &loopback_addr);
244	inet_aton("127.0.0.1", &dns_addr);
245
246	if (get_dns_addr_domain(pData, true, &dns_addr, &pData->pszDomain) < 0)
247	fNATfailed = 1;
248
249	inet_aton(pszNetAddr, &special_addr);
250	alias_addr.s_addr = special_addr.s_addr \| htonl(CTL_ALIAS);
251	getouraddr(pData);
252	return fNATfailed ? VINF_NAT_DNS : VINF_SUCCESS;
253	}
254
255	/**
256	* Statistics counters.
257	*/
258	void slirp_register_timers(PNATState pData, PPDMDRVINS pDrvIns)
259	{
260	#ifndef VBOX_WITHOUT_RELEASE_STATISTICS
261	PDMDrvHlpSTAMRegisterF(pDrvIns, &pData->StatFill, STAMTYPE_PROFILE, STAMVISIBILITY_ALWAYS,
262	STAMUNIT_TICKS_PER_CALL, "Profiling slirp fills", "/Drivers/NAT%d/Fill", pDrvIns->iInstance);
263	PDMDrvHlpSTAMRegisterF(pDrvIns, &pData->StatPoll, STAMTYPE_PROFILE, STAMVISIBILITY_ALWAYS,
264	STAMUNIT_TICKS_PER_CALL, "Profiling slirp polls", "/Drivers/NAT%d/Poll", pDrvIns->iInstance);
265	PDMDrvHlpSTAMRegisterF(pDrvIns, &pData->StatFastTimer, STAMTYPE_PROFILE, STAMVISIBILITY_ALWAYS,
266	STAMUNIT_TICKS_PER_CALL, "Profiling slirp fast timer", "/Drivers/NAT%d/TimerFast", pDrvIns->iInstance);
267	PDMDrvHlpSTAMRegisterF(pDrvIns, &pData->StatSlowTimer, STAMTYPE_PROFILE, STAMVISIBILITY_ALWAYS,
268	STAMUNIT_TICKS_PER_CALL, "Profiling slirp slow timer", "/Drivers/NAT%d/TimerSlow", pDrvIns->iInstance);
269	PDMDrvHlpSTAMRegisterF(pDrvIns, &pData->StatTCP, STAMTYPE_COUNTER, STAMVISIBILITY_ALWAYS,
270	STAMUNIT_COUNT, "TCP sockets", "/Drivers/NAT%d/SockTCP", pDrvIns->iInstance);
271	PDMDrvHlpSTAMRegisterF(pDrvIns, &pData->StatTCPHot, STAMTYPE_COUNTER, STAMVISIBILITY_ALWAYS,
272	STAMUNIT_COUNT, "TCP sockets active", "/Drivers/NAT%d/SockTCPHot", pDrvIns->iInstance);
273	PDMDrvHlpSTAMRegisterF(pDrvIns, &pData->StatUDP, STAMTYPE_COUNTER, STAMVISIBILITY_ALWAYS,
274	STAMUNIT_COUNT, "UDP sockets", "/Drivers/NAT%d/SockUDP", pDrvIns->iInstance);
275	PDMDrvHlpSTAMRegisterF(pDrvIns, &pData->StatTCPHot, STAMTYPE_COUNTER, STAMVISIBILITY_ALWAYS,
276	STAMUNIT_COUNT, "UDP sockets active", "/Drivers/NAT%d/SockUDPHot", pDrvIns->iInstance);
277	#endif /* VBOX_WITHOUT_RELEASE_STATISTICS */
278	}
279
280	/**
281	* Marks the link as up, making it possible to establish new connections.
282	*/
283	void slirp_link_up(PNATState pData)
284	{
285	link_up = 1;
286	}
287
288	/**
289	* Marks the link as down and cleans up the current connections.
290	*/
291	void slirp_link_down(PNATState pData)
292	{
293	struct socket *so;
294
295	while ((so = tcb.so_next) != &tcb)
296	{
297	if (so->so_state & SS_NOFDREF \|\| so->s == -1)
298	sofree(pData, so);
299	else
300	tcp_drop(pData, sototcpcb(so), 0);
301	}
302
303	while ((so = udb.so_next) != &udb)
304	udp_detach(pData, so);
305
306	link_up = 0;
307	}
308
309	/**
310	* Terminates the slirp component.
311	*/
312	void slirp_term(PNATState pData)
313	{
314	if (pData->pszDomain)
315	RTStrFree((char )(void )pData->pszDomain);
316
317	#if ARCH_BITS == 64
318	LogRel(("NAT: cpvHashUsed=%RU32 cpvHashCollisions=%RU32 cpvHashInserts=%RU64 cpvHashDone=%RU64\n",
319	pData->cpvHashUsed, pData->cpvHashCollisions, pData->cpvHashInserts, pData->cpvHashDone));
320	#endif
321
322	slirp_link_down(pData);
323	#ifdef WIN32
324	WSACleanup();
325	#endif
326	#ifdef LOG_ENABLED
327	Log(("\n"
328	"NAT statistics\n"
329	"--------------\n"
330	"\n"));
331	ipstats(pData);
332	tcpstats(pData);
333	udpstats(pData);
334	icmpstats(pData);
335	mbufstats(pData);
336	sockstats(pData);
337	Log(("\n"
338	"\n"
339	"\n"));
340	#endif
341	#if defined(VBOX_WITH_SIMPLEFIED_SLIRP_SYNC) && defined(RT_OS_WINDOWS)
342	free(pData->phEvents);
343	#endif
344	free(pData);
345	}
346
347
348	#define CONN_CANFSEND(so) (((so)->so_state & (SS_FCANTSENDMORE\|SS_ISFCONNECTED)) == SS_ISFCONNECTED)
349	#define CONN_CANFRCV(so) (((so)->so_state & (SS_FCANTRCVMORE\|SS_ISFCONNECTED)) == SS_ISFCONNECTED)
350	#define UPD_NFDS(x) if (nfds < (x)) nfds = (x)
351
352	/*
353	* curtime kept to an accuracy of 1ms
354	*/
355	#ifdef _WIN32
356	static void updtime(PNATState pData)
357	{
358	struct _timeb tb;
359
360	_ftime(&tb);
361	curtime = (u_int)tb.time * (u_int)1000;
362	curtime += (u_int)tb.millitm;
363	}
364	#else
365	static void updtime(PNATState pData)
366	{
367	gettimeofday(&tt, 0);
368
369	curtime = (u_int)tt.tv_sec * (u_int)1000;
370	curtime += (u_int)tt.tv_usec / (u_int)1000;
371
372	if ((tt.tv_usec % 1000) >= 500)
373	curtime++;
374	}
375	#endif
376
377	void slirp_select_fill(PNATState pData, int *pnfds,
378	fd_set readfds, fd_set writefds, fd_set *xfds)
379	{
380	struct socket so, so_next;
381	struct timeval timeout;
382	int nfds;
383	int tmp_time;
384	#if defined(VBOX_WITH_SIMPLEFIED_SLIRP_SYNC) && defined(RT_OS_WINDOWS)
385	int rc;
386	int error;
387	#endif
388
389	STAM_REL_PROFILE_START(&pData->StatFill, a);
390
391	nfds = *pnfds;
392	/*
393	* First, TCP sockets
394	*/
395	do_slowtimo = 0;
396	if (link_up) {
397	/*
398	* *_slowtimo needs calling if there are IP fragments
399	* in the fragment queue, or there are TCP connections active
400	*/
401	do_slowtimo = ((tcb.so_next != &tcb) \|\|
402	((struct ipasfrag )&ipq != u32_to_ptr(pData, ipq.next, struct ipasfrag )));
403
404	STAM_REL_COUNTER_RESET(&pData->StatTCP);
405	STAM_REL_COUNTER_RESET(&pData->StatTCPHot);
406
407	for (so = tcb.so_next; so != &tcb; so = so_next) {
408	so_next = so->so_next;
409
410	STAM_REL_COUNTER_INC(&pData->StatTCP);
411
412	/*
413	* See if we need a tcp_fasttimo
414	*/
415	if (time_fasttimo == 0 && so->so_tcpcb->t_flags & TF_DELACK)
416	time_fasttimo = curtime; /* Flag when we want a fasttimo */
417
418	/*
419	* NOFDREF can include still connecting to local-host,
420	* newly socreated() sockets etc. Don't want to select these.
421	*/
422	if (so->so_state & SS_NOFDREF \|\| so->s == -1)
423	continue;
424
425	/*
426	* Set for reading sockets which are accepting
427	*/
428	if (so->so_state & SS_FACCEPTCONN) {
429	STAM_REL_COUNTER_INC(&pData->StatTCPHot);
430	#if !defined(VBOX_WITH_SIMPLEFIED_SLIRP_SYNC) \|\| !defined(RT_OS_WINDOWS)
431	FD_SET(so->s, readfds);
432	UPD_NFDS(so->s);
433	#else
434	rc = WSAEventSelect(so->s, VBOX_SOCKET_EVENT, FD_READ\|FD_WRITE\|FD_ACCEPT\|FD_CONNECT\|FD_OOB);
435	if (rc == SOCKET_ERROR)
436	{
437	socket_error:
438	error = WSAGetLastError();
439	LogRel(("WSAEventSelector error %d (so=%x, socket=%s, event=%x)\n", error, so, so->s, VBOX_SOCKET_EVENT));
440	}
441	#endif
442	continue;
443	}
444
445	/*
446	* Set for writing sockets which are connecting
447	*/
448	if (so->so_state & SS_ISFCONNECTING) {
449	STAM_REL_COUNTER_INC(&pData->StatTCPHot);
450	#if !defined(VBOX_WITH_SIMPLEFIED_SLIRP_SYNC) \|\| !defined(RT_OS_WINDOWS)
451	FD_SET(so->s, writefds);
452	UPD_NFDS(so->s);
453	#else
454	rc = WSAEventSelect(so->s, VBOX_SOCKET_EVENT, FD_READ\|FD_WRITE\|FD_ACCEPT\|FD_CONNECT\|FD_OOB);
455	if (rc == SOCKET_ERROR)
456	goto socket_error;
457	#endif
458	continue;
459	}
460
461	/*
462	* Set for writing if we are connected, can send more, and
463	* we have something to send
464	*/
465	if (CONN_CANFSEND(so) && so->so_rcv.sb_cc) {
466	STAM_REL_COUNTER_INC(&pData->StatTCPHot);
467	#if !defined(VBOX_WITH_SIMPLEFIED_SLIRP_SYNC) \|\| !defined(RT_OS_WINDOWS)
468	FD_SET(so->s, writefds);
469	UPD_NFDS(so->s);
470	#else
471	rc = WSAEventSelect(so->s, VBOX_SOCKET_EVENT, FD_READ\|FD_WRITE\|FD_ACCEPT\|FD_CONNECT\|FD_OOB);
472	if (rc == SOCKET_ERROR)
473	goto socket_error;
474	continue; /XXX: we're using the widest mask for event/
475	#endif
476	}
477
478	/*
479	* Set for reading (and urgent data) if we are connected, can
480	* receive more, and we have room for it XXX /2 ?
481	*/
482	if (CONN_CANFRCV(so) && (so->so_snd.sb_cc < (so->so_snd.sb_datalen/2))) {
483	STAM_REL_COUNTER_INC(&pData->StatTCPHot);
484	#if !defined(VBOX_WITH_SIMPLEFIED_SLIRP_SYNC) \|\| !defined(RT_OS_WINDOWS)
485	FD_SET(so->s, readfds);
486	FD_SET(so->s, xfds);
487	UPD_NFDS(so->s);
488	#else
489	rc = WSAEventSelect(so->s, VBOX_SOCKET_EVENT, FD_OOB\|FD_READ\|FD_WRITE\|FD_ACCEPT\|FD_CONNECT);
490	if (rc == SOCKET_ERROR)
491	goto socket_error;
492	continue; /XXX: we're using the widest mask for event/
493	#endif
494	}
495	#if defined(VBOX_WITH_SIMPLEFIED_SLIRP_SYNC) && defined(RT_OS_WINDOWS)
496	rc = WSAEventSelect(so->s, NULL, 0);
497	if (rc == SOCKET_ERROR)
498	goto socket_error;
499	#endif
500	}
501
502	/*
503	* UDP sockets
504	*/
505	STAM_REL_COUNTER_RESET(&pData->StatUDP);
506	STAM_REL_COUNTER_RESET(&pData->StatUDPHot);
507
508	for (so = udb.so_next; so != &udb; so = so_next) {
509	so_next = so->so_next;
510
511	STAM_REL_COUNTER_INC(&pData->StatUDP);
512
513	/*
514	* See if it's timed out
515	*/
516	if (so->so_expire) {
517	if (so->so_expire <= curtime) {
518	udp_detach(pData, so);
519	continue;
520	} else
521	do_slowtimo = 1; /* Let socket expire */
522	}
523
524	/*
525	* When UDP packets are received from over the
526	* link, they're sendto()'d straight away, so
527	* no need for setting for writing
528	* Limit the number of packets queued by this session
529	* to 4. Note that even though we try and limit this
530	* to 4 packets, the session could have more queued
531	* if the packets needed to be fragmented
532	* (XXX <= 4 ?)
533	*/
534	if ((so->so_state & SS_ISFCONNECTED) && so->so_queued <= 4) {
535	STAM_REL_COUNTER_INC(&pData->StatUDPHot);
536	#if !defined(VBOX_WITH_SIMPLEFIED_SLIRP_SYNC) \|\| !defined(RT_OS_WINDOWS)
537	FD_SET(so->s, readfds);
538	UPD_NFDS(so->s);
539	#else
540	rc = WSAEventSelect(so->s, VBOX_SOCKET_EVENT, FD_READ\|FD_WRITE\|FD_OOB\|FD_ACCEPT);
541	if (rc == SOCKET_ERROR)
542	goto socket_error;
543	continue;
544	#endif
545	STAM_REL_COUNTER_INC(&pData->StatUDPHot);
546	}
547	#if defined(VBOX_WITH_SIMPLEFIED_SLIRP_SYNC) && defined(RT_OS_WINDOWS)
548	else
549	{
550	rc = WSAEventSelect(so->s, NULL, 0);
551	if (rc != SOCKET_ERROR)
552	goto socket_error;
553	}
554	#endif
555	}
556	}
557
558	/*
559	* Setup timeout to use minimum CPU usage, especially when idle
560	*/
561
562	/*
563	* First, see the timeout needed by *timo
564	*/
565	timeout.tv_sec = 0;
566	timeout.tv_usec = -1;
567	/*
568	* If a slowtimo is needed, set timeout to 500ms from the last
569	* slow timeout. If a fast timeout is needed, set timeout within
570	* 200ms of when it was requested.
571	*/
572	if (do_slowtimo) {
573	/* XXX + 10000 because some select()'s aren't that accurate */
574	timeout.tv_usec = ((500 - (curtime - last_slowtimo)) * 1000) + 10000;
575	if (timeout.tv_usec < 0)
576	timeout.tv_usec = 0;
577	else if (timeout.tv_usec > 510000)
578	timeout.tv_usec = 510000;
579
580	/* Can only fasttimo if we also slowtimo */
581	if (time_fasttimo) {
582	tmp_time = (200 - (curtime - time_fasttimo)) * 1000;
583	if (tmp_time < 0)
584	tmp_time = 0;
585
586	/* Choose the smallest of the 2 */
587	if (tmp_time < timeout.tv_usec)
588	timeout.tv_usec = (u_int)tmp_time;
589	}
590	}
591	#if !defined(VBOX_WITH_SIMPLEFIED_SLIRP_SYNC) \|\| !defined(RT_OS_WINDOWS)
592	*pnfds = nfds;
593	#else
594	*pnfds = VBOX_EVENT_COUNT;
595	#endif
596
597	STAM_REL_PROFILE_STOP(&pData->StatFill, a);
598	}
599
600	void slirp_select_poll(PNATState pData, fd_set readfds, fd_set writefds, fd_set *xfds)
601	{
602	struct socket so, so_next;
603	int ret;
604	#if defined(VBOX_WITH_SIMPLEFIED_SLIRP_SYNC) && defined(RT_OS_WINDOWS)
605	WSANETWORKEVENTS NetworkEvents;
606	int rc;
607	int error;
608	int timer_update = (readfds == NULL && writefds == NULL && xfds == NULL);
609	#endif
610	STAM_REL_PROFILE_START(&pData->StatPoll, a);
611
612	/* Update time */
613	updtime(pData);
614
615	/*
616	* See if anything has timed out
617	*/
618	if (link_up) {
619	if (time_fasttimo && ((curtime - time_fasttimo) >= 2)) {
620	STAM_REL_PROFILE_START(&pData->StatFastTimer, a);
621	tcp_fasttimo(pData);
622	time_fasttimo = 0;
623	STAM_REL_PROFILE_STOP(&pData->StatFastTimer, a);
624	}
625	if (do_slowtimo && ((curtime - last_slowtimo) >= 499)) {
626	STAM_REL_PROFILE_START(&pData->StatSlowTimer, a);
627	ip_slowtimo(pData);
628	tcp_slowtimo(pData);
629	last_slowtimo = curtime;
630	STAM_REL_PROFILE_STOP(&pData->StatSlowTimer, a);
631	}
632	}
633	#if defined(VBOX_WITH_SIMPLEFIED_SLIRP_SYNC) && defined(RT_OS_WINDOWS)
634	if (timer_update) return;
635	#endif
636
637	/*
638	* Check sockets
639	*/
640	if (link_up) {
641	/*
642	* Check TCP sockets
643	*/
644	for (so = tcb.so_next; so != &tcb; so = so_next) {
645	so_next = so->so_next;
646
647	/*
648	* FD_ISSET is meaningless on these sockets
649	* (and they can crash the program)
650	*/
651	if (so->so_state & SS_NOFDREF \|\| so->s == -1)
652	continue;
653	#if defined(VBOX_WITH_SIMPLEFIED_SLIRP_SYNC) && defined(RT_OS_WINDOWS)
654	rc = WSAEnumNetworkEvents(so->s, VBOX_SOCKET_EVENT, &NetworkEvents);
655	if (rc == SOCKET_ERROR)
656	{
657	error = WSAGetLastError();
658	LogRel(("WSAEnumNetworkEvents TCP error %d\n", error));
659	continue;
660	}
661	#endif
662
663	/*
664	* Check for URG data
665	* This will soread as well, so no need to
666	* test for readfds below if this succeeds
667	*/
668	#if !defined(VBOX_WITH_SIMPLEFIED_SLIRP_SYNC) \|\| !defined(RT_OS_WINDOWS)
669	if (FD_ISSET(so->s, xfds))
670	#else
671	/* out-of-band data */
672	if ((NetworkEvents.lNetworkEvents & FD_OOB) && NetworkEvents.iErrorCode[FD_OOB_BIT] == 0)
673	#endif
674	sorecvoob(pData, so);
675	/*
676	* Check sockets for reading
677	*/
678	#if !defined(VBOX_WITH_SIMPLEFIED_SLIRP_SYNC) \|\| !defined(RT_OS_WINDOWS)
679	else if (FD_ISSET(so->s, readfds)) {
680	#else
681	else if ((NetworkEvents.lNetworkEvents & FD_READ) && (NetworkEvents.iErrorCode[FD_READ_BIT] == 0)) {
682	#endif
683	/*
684	* Check for incoming connections
685	*/
686	if (so->so_state & SS_FACCEPTCONN) {
687	tcp_connect(pData, so);
688	continue;
689	} /* else */
690	ret = soread(pData, so);
691
692	/* Output it if we read something */
693	if (ret > 0)
694	tcp_output(pData, sototcpcb(so));
695	}
696
697	/*
698	* Check sockets for writing
699	*/
700	#if !defined(VBOX_WITH_SIMPLEFIED_SLIRP_SYNC) \|\| !defined(RT_OS_WINDOWS)
701	if (FD_ISSET(so->s, writefds)) {
702	#else
703	if ((NetworkEvents.lNetworkEvents & FD_WRITE) && (NetworkEvents.iErrorCode[FD_WRITE_BIT] == 0)) {
704	#endif
705	/*
706	* Check for non-blocking, still-connecting sockets
707	*/
708	if (so->so_state & SS_ISFCONNECTING) {
709	/* Connected */
710	so->so_state &= ~SS_ISFCONNECTING;
711
712	/*
713	* This should be probably guarded by PROBE_CONN too. Anyway,
714	* we disable it on OS/2 because the below send call returns
715	* EFAULT which causes the opened TCP socket to close right
716	* after it has been opened and connected.
717	*/
718	#ifndef RT_OS_OS2
719	ret = send(so->s, (const char *)&ret, 0, 0);
720	if (ret < 0) {
721	/* XXXXX Must fix, zero bytes is a NOP */
722	if (errno == EAGAIN \|\| errno == EWOULDBLOCK \|\|
723	errno == EINPROGRESS \|\| errno == ENOTCONN) {
724	continue;
725	}
726
727	/* else failed */
728	so->so_state = SS_NOFDREF;
729	}
730	/* else so->so_state &= ~SS_ISFCONNECTING; */
731	#endif
732
733	/*
734	* Continue tcp_input
735	*/
736	tcp_input(pData, (struct mbuf *)NULL, sizeof(struct ip), so);
737	/* continue; */
738	} else
739	ret = sowrite(pData, so);
740	/*
741	* XXXXX If we wrote something (a lot), there
742	* could be a need for a window update.
743	* In the worst case, the remote will send
744	* a window probe to get things going again
745	*/
746	}
747
748	/*
749	* Probe a still-connecting, non-blocking socket
750	* to check if it's still alive
751	*/
752	#ifdef PROBE_CONN
753	if (so->so_state & SS_ISFCONNECTING) {
754	ret = recv(so->s, (char *)&ret, 0,0);
755
756	if (ret < 0) {
757	/* XXX */
758	if (errno == EAGAIN \|\| errno == EWOULDBLOCK \|\|
759	errno == EINPROGRESS \|\| errno == ENOTCONN) {
760	continue; /* Still connecting, continue */
761	}
762
763	/* else failed */
764	so->so_state = SS_NOFDREF;
765
766	/* tcp_input will take care of it */
767	} else {
768	ret = send(so->s, &ret, 0,0);
769	if (ret < 0) {
770	/* XXX */
771	if (errno == EAGAIN \|\| errno == EWOULDBLOCK \|\|
772	errno == EINPROGRESS \|\| errno == ENOTCONN) {
773	continue;
774	}
775	/* else failed */
776	so->so_state = SS_NOFDREF;
777	} else
778	so->so_state &= ~SS_ISFCONNECTING;
779
780	}
781	tcp_input((struct mbuf *)NULL, sizeof(struct ip),so);
782	} /* SS_ISFCONNECTING */
783	#endif
784	}
785
786	/*
787	* Now UDP sockets.
788	* Incoming packets are sent straight away, they're not buffered.
789	* Incoming UDP data isn't buffered either.
790	*/
791	for (so = udb.so_next; so != &udb; so = so_next) {
792	so_next = so->so_next;
793
794	#if defined(VBOX_WITH_SIMPLEFIED_SLIRP_SYNC) && defined(RT_OS_WINDOWS)
795	rc = WSAEnumNetworkEvents(so->s, VBOX_SOCKET_EVENT, &NetworkEvents);
796	if (rc == SOCKET_ERROR)
797	{
798	error = WSAGetLastError();
799	LogRel(("WSAEnumNetworkEvents TCP error %d\n", error));
800	continue;
801	}
802	#endif
803	#if !defined(VBOX_WITH_SIMPLEFIED_SLIRP_SYNC) \|\| !defined(RT_OS_WINDOWS)
804	if (so->s != -1 && FD_ISSET(so->s, readfds)) {
805	#else
806	if ((NetworkEvents.lNetworkEvents & FD_READ) && (NetworkEvents.iErrorCode[FD_READ_BIT] == 0)) {
807	#endif
808	sorecvfrom(pData, so);
809	}
810	}
811	}
812
813	/*
814	* See if we can start outputting
815	*/
816	if (if_queued && link_up)
817	if_start(pData);
818
819	STAM_REL_PROFILE_STOP(&pData->StatPoll, a);
820	}
821
822	#define ETH_ALEN 6
823	#define ETH_HLEN 14
824
825	#define ETH_P_IP 0x0800 /* Internet Protocol packet */
826	#define ETH_P_ARP 0x0806 /* Address Resolution packet */
827
828	#define ARPOP_REQUEST 1 /* ARP request */
829	#define ARPOP_REPLY 2 /* ARP reply */
830
831	struct ethhdr
832	{
833	unsigned char h_dest[ETH_ALEN]; /* destination eth addr */
834	unsigned char h_source[ETH_ALEN]; /* source ether addr */
835	unsigned short h_proto; /* packet type ID field */
836	};
837
838	struct arphdr
839	{
840	unsigned short ar_hrd; /* format of hardware address */
841	unsigned short ar_pro; /* format of protocol address */
842	unsigned char ar_hln; /* length of hardware address */
843	unsigned char ar_pln; /* length of protocol address */
844	unsigned short ar_op; /* ARP opcode (command) */
845
846	/*
847	* Ethernet looks like this : This bit is variable sized however...
848	*/
849	unsigned char ar_sha[ETH_ALEN]; /* sender hardware address */
850	unsigned char ar_sip[4]; /* sender IP address */
851	unsigned char ar_tha[ETH_ALEN]; /* target hardware address */
852	unsigned char ar_tip[4]; /* target IP address */
853	};
854
855	static
856	void arp_input(PNATState pData, const uint8_t *pkt, int pkt_len)
857	{
858	struct ethhdr eh = (struct ethhdr )pkt;
859	struct arphdr ah = (struct arphdr )(pkt + ETH_HLEN);
860	uint8_t arp_reply[ETH_HLEN + sizeof(struct arphdr)];
861	struct ethhdr reh = (struct ethhdr )arp_reply;
862	struct arphdr rah = (struct arphdr )(arp_reply + ETH_HLEN);
863	int ar_op;
864	struct ex_list *ex_ptr;
865	uint32_t htip = ntohl((uint32_t)ah->ar_tip);
866
867	ar_op = ntohs(ah->ar_op);
868	switch(ar_op) {
869	case ARPOP_REQUEST:
870	if ((htip & pData->netmask) == ntohl(special_addr.s_addr)) {
871	if ( (htip & ~pData->netmask) == CTL_DNS
872	\|\| (htip & ~pData->netmask) == CTL_ALIAS)
873	goto arp_ok;
874	for (ex_ptr = exec_list; ex_ptr; ex_ptr = ex_ptr->ex_next) {
875	if ((htip & ~pData->netmask) == ex_ptr->ex_addr)
876	goto arp_ok;
877	}
878	return;
879	arp_ok:
880	/* XXX: make an ARP request to have the client address */
881	memcpy(client_ethaddr, eh->h_source, ETH_ALEN);
882
883	/* ARP request for alias/dns mac address */
884	memcpy(reh->h_dest, pkt + ETH_ALEN, ETH_ALEN);
885	memcpy(reh->h_source, special_ethaddr, ETH_ALEN - 1);
886	reh->h_source[5] = ah->ar_tip[3];
887	reh->h_proto = htons(ETH_P_ARP);
888
889	rah->ar_hrd = htons(1);
890	rah->ar_pro = htons(ETH_P_IP);
891	rah->ar_hln = ETH_ALEN;
892	rah->ar_pln = 4;
893	rah->ar_op = htons(ARPOP_REPLY);
894	memcpy(rah->ar_sha, reh->h_source, ETH_ALEN);
895	memcpy(rah->ar_sip, ah->ar_tip, 4);
896	memcpy(rah->ar_tha, ah->ar_sha, ETH_ALEN);
897	memcpy(rah->ar_tip, ah->ar_sip, 4);
898	slirp_output(pData->pvUser, arp_reply, sizeof(arp_reply));
899	}
900	break;
901	default:
902	break;
903	}
904	}
905
906	void slirp_input(PNATState pData, const uint8_t *pkt, int pkt_len)
907	{
908	struct mbuf *m;
909	int proto;
910
911	if (pkt_len < ETH_HLEN)
912	return;
913
914	proto = ntohs((uint16_t )(pkt + 12));
915	switch(proto) {
916	case ETH_P_ARP:
917	arp_input(pData, pkt, pkt_len);
918	break;
919	case ETH_P_IP:
920	/* Update time. Important if the network is very quiet, as otherwise
921	* the first outgoing connection gets an incorrect timestamp. */
922	updtime(pData);
923
924	m = m_get(pData);
925	if (!m)
926	return;
927	/* Note: we add to align the IP header */
928	if (M_FREEROOM(m) < pkt_len + 2) {
929	m_inc(m, pkt_len + 2);
930	}
931	m->m_len = pkt_len + 2;
932	memcpy(m->m_data + 2, pkt, pkt_len);
933
934	m->m_data += 2 + ETH_HLEN;
935	m->m_len -= 2 + ETH_HLEN;
936
937	ip_input(pData, m);
938	break;
939	default:
940	break;
941	}
942	}
943
944	/* output the IP packet to the ethernet device */
945	void if_encap(PNATState pData, const uint8_t *ip_data, int ip_data_len)
946	{
947	uint8_t buf[1600];
948	struct ethhdr eh = (struct ethhdr )buf;
949
950	if (ip_data_len + ETH_HLEN > sizeof(buf))
951	return;
952
953	memcpy(eh->h_dest, client_ethaddr, ETH_ALEN);
954	memcpy(eh->h_source, special_ethaddr, ETH_ALEN - 1);
955	/* XXX: not correct */
956	eh->h_source[5] = CTL_ALIAS;
957	eh->h_proto = htons(ETH_P_IP);
958	memcpy(buf + sizeof(struct ethhdr), ip_data, ip_data_len);
959	slirp_output(pData->pvUser, buf, ip_data_len + ETH_HLEN);
960	}
961
962	int slirp_redir(PNATState pData, int is_udp, int host_port,
963	struct in_addr guest_addr, int guest_port)
964	{
965	if (is_udp) {
966	if (!udp_listen(pData, htons(host_port), guest_addr.s_addr,
967	htons(guest_port), 0))
968	return -1;
969	} else {
970	if (!solisten(pData, htons(host_port), guest_addr.s_addr,
971	htons(guest_port), 0))
972	return -1;
973	}
974	return 0;
975	}
976
977	int slirp_add_exec(PNATState pData, int do_pty, const char *args, int addr_low_byte,
978	int guest_port)
979	{
980	return add_exec(&exec_list, do_pty, (char *)args,
981	addr_low_byte, htons(guest_port));
982	}
983
984	void slirp_set_ethaddr(PNATState pData, const uint8_t *ethaddr)
985	{
986	memcpy(client_ethaddr, ethaddr, ETH_ALEN);
987	}
988
989	#if defined(VBOX_WITH_SIMPLEFIED_SLIRP_SYNC) && defined(RT_OS_WINDOWS)
990	HANDLE *slirp_get_events(PNATState pData)
991	{
992	return pData->phEvents;
993	}
994	void slirp_register_external_event(PNATState pData, HANDLE hEvent, int index)
995	{
996	pData->phEvents[index] = hEvent;
997	}
998	#endif

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

source: vbox/trunk/src/VBox/Devices/Network/slirp/slirp.c@ 14200

Download in other formats: