slirp.c@ 14232

Last change on this file since 14232 was 14232, checked in by vboxsync, 16 years ago
Adapted versions of ip_reass, ip_freef and ip_slowtimo from FreeBSD (compilable on Linux, need some debugging)
Property svn:eol-style set to `native`
File size: 30.3 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->StatUDPHot, 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	#ifdef VBOX_WITH_BSD_REASS
389	int i;
390	#endif /* VBOX_WITH_BSD_REASS */
391
392	STAM_REL_PROFILE_START(&pData->StatFill, a);
393
394	nfds = *pnfds;
395	/*
396	* First, TCP sockets
397	*/
398	do_slowtimo = 0;
399	if (link_up) {
400	/*
401	* *_slowtimo needs calling if there are IP fragments
402	* in the fragment queue, or there are TCP connections active
403	*/
404	#ifndef VBOX_WITH_BSD_REASS
405	do_slowtimo = ((tcb.so_next != &tcb) \|\|
406	((struct ipasfrag )&ipq != u32_to_ptr(pData, ipq.next, struct ipasfrag )));
407	#else /* !VBOX_WITH_BSD_REASS */
408	/* XXX: triggering of fragment expiration should be the same but use
409	* new macroses
410	*/
411	for (i = 0; i < IPREASS_NHASH; i++) {
412	if (!TAILQ_EMPTY(&ipq[i])) {
413	do_slowtimo = 1;
414	break;
415	}
416	}
417	do_slowtimo \|= (tcb.so_next != &tcb);
418	#endif /* VBOX_WITH_BSD_REASS */
419
420	STAM_REL_COUNTER_RESET(&pData->StatTCP);
421	STAM_REL_COUNTER_RESET(&pData->StatTCPHot);
422
423	for (so = tcb.so_next; so != &tcb; so = so_next) {
424	so_next = so->so_next;
425
426	STAM_REL_COUNTER_INC(&pData->StatTCP);
427
428	/*
429	* See if we need a tcp_fasttimo
430	*/
431	if (time_fasttimo == 0 && so->so_tcpcb->t_flags & TF_DELACK)
432	time_fasttimo = curtime; /* Flag when we want a fasttimo */
433
434	/*
435	* NOFDREF can include still connecting to local-host,
436	* newly socreated() sockets etc. Don't want to select these.
437	*/
438	if (so->so_state & SS_NOFDREF \|\| so->s == -1)
439	continue;
440
441	/*
442	* Set for reading sockets which are accepting
443	*/
444	if (so->so_state & SS_FACCEPTCONN) {
445	STAM_REL_COUNTER_INC(&pData->StatTCPHot);
446	#if !defined(VBOX_WITH_SIMPLEFIED_SLIRP_SYNC) \|\| !defined(RT_OS_WINDOWS)
447	FD_SET(so->s, readfds);
448	UPD_NFDS(so->s);
449	#else
450	rc = WSAEventSelect(so->s, VBOX_SOCKET_EVENT, FD_READ\|FD_WRITE\|FD_ACCEPT\|FD_CONNECT\|FD_OOB);
451	if (rc == SOCKET_ERROR)
452	{
453	socket_error:
454	error = WSAGetLastError();
455	LogRel(("WSAEventSelector error %d (so=%x, socket=%s, event=%x)\n", error, so, so->s, VBOX_SOCKET_EVENT));
456	}
457	#endif
458	continue;
459	}
460
461	/*
462	* Set for writing sockets which are connecting
463	*/
464	if (so->so_state & SS_ISFCONNECTING) {
465	STAM_REL_COUNTER_INC(&pData->StatTCPHot);
466	#if !defined(VBOX_WITH_SIMPLEFIED_SLIRP_SYNC) \|\| !defined(RT_OS_WINDOWS)
467	FD_SET(so->s, writefds);
468	UPD_NFDS(so->s);
469	#else
470	rc = WSAEventSelect(so->s, VBOX_SOCKET_EVENT, FD_READ\|FD_WRITE\|FD_ACCEPT\|FD_CONNECT\|FD_OOB);
471	if (rc == SOCKET_ERROR)
472	goto socket_error;
473	#endif
474	continue;
475	}
476
477	/*
478	* Set for writing if we are connected, can send more, and
479	* we have something to send
480	*/
481	if (CONN_CANFSEND(so) && so->so_rcv.sb_cc) {
482	STAM_REL_COUNTER_INC(&pData->StatTCPHot);
483	#if !defined(VBOX_WITH_SIMPLEFIED_SLIRP_SYNC) \|\| !defined(RT_OS_WINDOWS)
484	FD_SET(so->s, writefds);
485	UPD_NFDS(so->s);
486	#else
487	rc = WSAEventSelect(so->s, VBOX_SOCKET_EVENT, FD_READ\|FD_WRITE\|FD_ACCEPT\|FD_CONNECT\|FD_OOB);
488	if (rc == SOCKET_ERROR)
489	goto socket_error;
490	continue; /XXX: we're using the widest mask for event/
491	#endif
492	}
493
494	/*
495	* Set for reading (and urgent data) if we are connected, can
496	* receive more, and we have room for it XXX /2 ?
497	*/
498	if (CONN_CANFRCV(so) && (so->so_snd.sb_cc < (so->so_snd.sb_datalen/2))) {
499	STAM_REL_COUNTER_INC(&pData->StatTCPHot);
500	#if !defined(VBOX_WITH_SIMPLEFIED_SLIRP_SYNC) \|\| !defined(RT_OS_WINDOWS)
501	FD_SET(so->s, readfds);
502	FD_SET(so->s, xfds);
503	UPD_NFDS(so->s);
504	#else
505	rc = WSAEventSelect(so->s, VBOX_SOCKET_EVENT, FD_OOB\|FD_READ\|FD_WRITE\|FD_ACCEPT\|FD_CONNECT);
506	if (rc == SOCKET_ERROR)
507	goto socket_error;
508	continue; /XXX: we're using the widest mask for event/
509	#endif
510	}
511	#if defined(VBOX_WITH_SIMPLEFIED_SLIRP_SYNC) && defined(RT_OS_WINDOWS)
512	rc = WSAEventSelect(so->s, NULL, 0);
513	if (rc == SOCKET_ERROR)
514	goto socket_error;
515	#endif
516	}
517
518	/*
519	* UDP sockets
520	*/
521	STAM_REL_COUNTER_RESET(&pData->StatUDP);
522	STAM_REL_COUNTER_RESET(&pData->StatUDPHot);
523
524	for (so = udb.so_next; so != &udb; so = so_next) {
525	so_next = so->so_next;
526
527	STAM_REL_COUNTER_INC(&pData->StatUDP);
528
529	/*
530	* See if it's timed out
531	*/
532	if (so->so_expire) {
533	if (so->so_expire <= curtime) {
534	udp_detach(pData, so);
535	continue;
536	} else
537	do_slowtimo = 1; /* Let socket expire */
538	}
539
540	/*
541	* When UDP packets are received from over the
542	* link, they're sendto()'d straight away, so
543	* no need for setting for writing
544	* Limit the number of packets queued by this session
545	* to 4. Note that even though we try and limit this
546	* to 4 packets, the session could have more queued
547	* if the packets needed to be fragmented
548	* (XXX <= 4 ?)
549	*/
550	if ((so->so_state & SS_ISFCONNECTED) && so->so_queued <= 4) {
551	STAM_REL_COUNTER_INC(&pData->StatUDPHot);
552	#if !defined(VBOX_WITH_SIMPLEFIED_SLIRP_SYNC) \|\| !defined(RT_OS_WINDOWS)
553	FD_SET(so->s, readfds);
554	UPD_NFDS(so->s);
555	#else
556	rc = WSAEventSelect(so->s, VBOX_SOCKET_EVENT, FD_READ\|FD_WRITE\|FD_OOB\|FD_ACCEPT);
557	if (rc == SOCKET_ERROR)
558	goto socket_error;
559	continue;
560	#endif
561	}
562	#if defined(VBOX_WITH_SIMPLEFIED_SLIRP_SYNC) && defined(RT_OS_WINDOWS)
563	else
564	{
565	rc = WSAEventSelect(so->s, NULL, 0);
566	if (rc != SOCKET_ERROR)
567	goto socket_error;
568	}
569	#endif
570	}
571	}
572
573	/*
574	* Setup timeout to use minimum CPU usage, especially when idle
575	*/
576
577	/*
578	* First, see the timeout needed by *timo
579	*/
580	timeout.tv_sec = 0;
581	timeout.tv_usec = -1;
582	/*
583	* If a slowtimo is needed, set timeout to 500ms from the last
584	* slow timeout. If a fast timeout is needed, set timeout within
585	* 200ms of when it was requested.
586	*/
587	if (do_slowtimo) {
588	/* XXX + 10000 because some select()'s aren't that accurate */
589	timeout.tv_usec = ((500 - (curtime - last_slowtimo)) * 1000) + 10000;
590	if (timeout.tv_usec < 0)
591	timeout.tv_usec = 0;
592	else if (timeout.tv_usec > 510000)
593	timeout.tv_usec = 510000;
594
595	/* Can only fasttimo if we also slowtimo */
596	if (time_fasttimo) {
597	tmp_time = (200 - (curtime - time_fasttimo)) * 1000;
598	if (tmp_time < 0)
599	tmp_time = 0;
600
601	/* Choose the smallest of the 2 */
602	if (tmp_time < timeout.tv_usec)
603	timeout.tv_usec = (u_int)tmp_time;
604	}
605	}
606	#if !defined(VBOX_WITH_SIMPLEFIED_SLIRP_SYNC) \|\| !defined(RT_OS_WINDOWS)
607	*pnfds = nfds;
608	#else
609	*pnfds = VBOX_EVENT_COUNT;
610	#endif
611
612	STAM_REL_PROFILE_STOP(&pData->StatFill, a);
613	}
614
615	void slirp_select_poll(PNATState pData, fd_set readfds, fd_set writefds, fd_set *xfds)
616	{
617	struct socket so, so_next;
618	int ret;
619	#if defined(VBOX_WITH_SIMPLEFIED_SLIRP_SYNC) && defined(RT_OS_WINDOWS)
620	WSANETWORKEVENTS NetworkEvents;
621	int rc;
622	int error;
623	int timer_update = (readfds == NULL && writefds == NULL && xfds == NULL);
624	#endif
625	STAM_REL_PROFILE_START(&pData->StatPoll, a);
626
627	/* Update time */
628	updtime(pData);
629
630	/*
631	* See if anything has timed out
632	*/
633	if (link_up) {
634	if (time_fasttimo && ((curtime - time_fasttimo) >= 2)) {
635	STAM_REL_PROFILE_START(&pData->StatFastTimer, a);
636	tcp_fasttimo(pData);
637	time_fasttimo = 0;
638	STAM_REL_PROFILE_STOP(&pData->StatFastTimer, a);
639	}
640	if (do_slowtimo && ((curtime - last_slowtimo) >= 499)) {
641	STAM_REL_PROFILE_START(&pData->StatSlowTimer, a);
642	ip_slowtimo(pData);
643	tcp_slowtimo(pData);
644	last_slowtimo = curtime;
645	STAM_REL_PROFILE_STOP(&pData->StatSlowTimer, a);
646	}
647	}
648	#if defined(VBOX_WITH_SIMPLEFIED_SLIRP_SYNC) && defined(RT_OS_WINDOWS)
649	if (timer_update) return;
650	#endif
651
652	/*
653	* Check sockets
654	*/
655	if (link_up) {
656	/*
657	* Check TCP sockets
658	*/
659	for (so = tcb.so_next; so != &tcb; so = so_next) {
660	so_next = so->so_next;
661
662	/*
663	* FD_ISSET is meaningless on these sockets
664	* (and they can crash the program)
665	*/
666	if (so->so_state & SS_NOFDREF \|\| so->s == -1)
667	continue;
668	#if defined(VBOX_WITH_SIMPLEFIED_SLIRP_SYNC) && defined(RT_OS_WINDOWS)
669	rc = WSAEnumNetworkEvents(so->s, VBOX_SOCKET_EVENT, &NetworkEvents);
670	if (rc == SOCKET_ERROR)
671	{
672	error = WSAGetLastError();
673	LogRel(("WSAEnumNetworkEvents TCP error %d\n", error));
674	continue;
675	}
676	#endif
677
678	/*
679	* Check for URG data
680	* This will soread as well, so no need to
681	* test for readfds below if this succeeds
682	*/
683	#if !defined(VBOX_WITH_SIMPLEFIED_SLIRP_SYNC) \|\| !defined(RT_OS_WINDOWS)
684	if (FD_ISSET(so->s, xfds))
685	#else
686	/* out-of-band data */
687	if ((NetworkEvents.lNetworkEvents & FD_OOB) && NetworkEvents.iErrorCode[FD_OOB_BIT] == 0)
688	#endif
689	sorecvoob(pData, so);
690	/*
691	* Check sockets for reading
692	*/
693	#if !defined(VBOX_WITH_SIMPLEFIED_SLIRP_SYNC) \|\| !defined(RT_OS_WINDOWS)
694	else if (FD_ISSET(so->s, readfds)) {
695	#else
696	else if ((NetworkEvents.lNetworkEvents & FD_READ) && (NetworkEvents.iErrorCode[FD_READ_BIT] == 0)) {
697	#endif
698	/*
699	* Check for incoming connections
700	*/
701	if (so->so_state & SS_FACCEPTCONN) {
702	tcp_connect(pData, so);
703	continue;
704	} /* else */
705	ret = soread(pData, so);
706
707	/* Output it if we read something */
708	if (ret > 0)
709	tcp_output(pData, sototcpcb(so));
710	}
711
712	/*
713	* Check sockets for writing
714	*/
715	#if !defined(VBOX_WITH_SIMPLEFIED_SLIRP_SYNC) \|\| !defined(RT_OS_WINDOWS)
716	if (FD_ISSET(so->s, writefds)) {
717	#else
718	if ((NetworkEvents.lNetworkEvents & FD_WRITE) && (NetworkEvents.iErrorCode[FD_WRITE_BIT] == 0)) {
719	#endif
720	/*
721	* Check for non-blocking, still-connecting sockets
722	*/
723	if (so->so_state & SS_ISFCONNECTING) {
724	/* Connected */
725	so->so_state &= ~SS_ISFCONNECTING;
726
727	/*
728	* This should be probably guarded by PROBE_CONN too. Anyway,
729	* we disable it on OS/2 because the below send call returns
730	* EFAULT which causes the opened TCP socket to close right
731	* after it has been opened and connected.
732	*/
733	#ifndef RT_OS_OS2
734	ret = send(so->s, (const char *)&ret, 0, 0);
735	if (ret < 0) {
736	/* XXXXX Must fix, zero bytes is a NOP */
737	if (errno == EAGAIN \|\| errno == EWOULDBLOCK \|\|
738	errno == EINPROGRESS \|\| errno == ENOTCONN) {
739	continue;
740	}
741
742	/* else failed */
743	so->so_state = SS_NOFDREF;
744	}
745	/* else so->so_state &= ~SS_ISFCONNECTING; */
746	#endif
747
748	/*
749	* Continue tcp_input
750	*/
751	tcp_input(pData, (struct mbuf *)NULL, sizeof(struct ip), so);
752	/* continue; */
753	} else
754	ret = sowrite(pData, so);
755	/*
756	* XXXXX If we wrote something (a lot), there
757	* could be a need for a window update.
758	* In the worst case, the remote will send
759	* a window probe to get things going again
760	*/
761	}
762
763	/*
764	* Probe a still-connecting, non-blocking socket
765	* to check if it's still alive
766	*/
767	#ifdef PROBE_CONN
768	if (so->so_state & SS_ISFCONNECTING) {
769	ret = recv(so->s, (char *)&ret, 0,0);
770
771	if (ret < 0) {
772	/* XXX */
773	if (errno == EAGAIN \|\| errno == EWOULDBLOCK \|\|
774	errno == EINPROGRESS \|\| errno == ENOTCONN) {
775	continue; /* Still connecting, continue */
776	}
777
778	/* else failed */
779	so->so_state = SS_NOFDREF;
780
781	/* tcp_input will take care of it */
782	} else {
783	ret = send(so->s, &ret, 0,0);
784	if (ret < 0) {
785	/* XXX */
786	if (errno == EAGAIN \|\| errno == EWOULDBLOCK \|\|
787	errno == EINPROGRESS \|\| errno == ENOTCONN) {
788	continue;
789	}
790	/* else failed */
791	so->so_state = SS_NOFDREF;
792	} else
793	so->so_state &= ~SS_ISFCONNECTING;
794
795	}
796	tcp_input((struct mbuf *)NULL, sizeof(struct ip),so);
797	} /* SS_ISFCONNECTING */
798	#endif
799	}
800
801	/*
802	* Now UDP sockets.
803	* Incoming packets are sent straight away, they're not buffered.
804	* Incoming UDP data isn't buffered either.
805	*/
806	for (so = udb.so_next; so != &udb; so = so_next) {
807	so_next = so->so_next;
808
809	#if defined(VBOX_WITH_SIMPLEFIED_SLIRP_SYNC) && defined(RT_OS_WINDOWS)
810	rc = WSAEnumNetworkEvents(so->s, VBOX_SOCKET_EVENT, &NetworkEvents);
811	if (rc == SOCKET_ERROR)
812	{
813	error = WSAGetLastError();
814	LogRel(("WSAEnumNetworkEvents TCP error %d\n", error));
815	continue;
816	}
817	#endif
818	#if !defined(VBOX_WITH_SIMPLEFIED_SLIRP_SYNC) \|\| !defined(RT_OS_WINDOWS)
819	if (so->s != -1 && FD_ISSET(so->s, readfds)) {
820	#else
821	if ((NetworkEvents.lNetworkEvents & FD_READ) && (NetworkEvents.iErrorCode[FD_READ_BIT] == 0)) {
822	#endif
823	sorecvfrom(pData, so);
824	}
825	}
826	}
827
828	/*
829	* See if we can start outputting
830	*/
831	if (if_queued && link_up)
832	if_start(pData);
833
834	STAM_REL_PROFILE_STOP(&pData->StatPoll, a);
835	}
836
837	#define ETH_ALEN 6
838	#define ETH_HLEN 14
839
840	#define ETH_P_IP 0x0800 /* Internet Protocol packet */
841	#define ETH_P_ARP 0x0806 /* Address Resolution packet */
842
843	#define ARPOP_REQUEST 1 /* ARP request */
844	#define ARPOP_REPLY 2 /* ARP reply */
845
846	struct ethhdr
847	{
848	unsigned char h_dest[ETH_ALEN]; /* destination eth addr */
849	unsigned char h_source[ETH_ALEN]; /* source ether addr */
850	unsigned short h_proto; /* packet type ID field */
851	};
852
853	struct arphdr
854	{
855	unsigned short ar_hrd; /* format of hardware address */
856	unsigned short ar_pro; /* format of protocol address */
857	unsigned char ar_hln; /* length of hardware address */
858	unsigned char ar_pln; /* length of protocol address */
859	unsigned short ar_op; /* ARP opcode (command) */
860
861	/*
862	* Ethernet looks like this : This bit is variable sized however...
863	*/
864	unsigned char ar_sha[ETH_ALEN]; /* sender hardware address */
865	unsigned char ar_sip[4]; /* sender IP address */
866	unsigned char ar_tha[ETH_ALEN]; /* target hardware address */
867	unsigned char ar_tip[4]; /* target IP address */
868	};
869
870	static
871	void arp_input(PNATState pData, const uint8_t *pkt, int pkt_len)
872	{
873	struct ethhdr eh = (struct ethhdr )pkt;
874	struct arphdr ah = (struct arphdr )(pkt + ETH_HLEN);
875	uint8_t arp_reply[ETH_HLEN + sizeof(struct arphdr)];
876	struct ethhdr reh = (struct ethhdr )arp_reply;
877	struct arphdr rah = (struct arphdr )(arp_reply + ETH_HLEN);
878	int ar_op;
879	struct ex_list *ex_ptr;
880	uint32_t htip = ntohl((uint32_t)ah->ar_tip);
881
882	ar_op = ntohs(ah->ar_op);
883	switch(ar_op) {
884	case ARPOP_REQUEST:
885	if ((htip & pData->netmask) == ntohl(special_addr.s_addr)) {
886	if ( (htip & ~pData->netmask) == CTL_DNS
887	\|\| (htip & ~pData->netmask) == CTL_ALIAS)
888	goto arp_ok;
889	for (ex_ptr = exec_list; ex_ptr; ex_ptr = ex_ptr->ex_next) {
890	if ((htip & ~pData->netmask) == ex_ptr->ex_addr)
891	goto arp_ok;
892	}
893	return;
894	arp_ok:
895	/* XXX: make an ARP request to have the client address */
896	memcpy(client_ethaddr, eh->h_source, ETH_ALEN);
897
898	/* ARP request for alias/dns mac address */
899	memcpy(reh->h_dest, pkt + ETH_ALEN, ETH_ALEN);
900	memcpy(reh->h_source, special_ethaddr, ETH_ALEN - 1);
901	reh->h_source[5] = ah->ar_tip[3];
902	reh->h_proto = htons(ETH_P_ARP);
903
904	rah->ar_hrd = htons(1);
905	rah->ar_pro = htons(ETH_P_IP);
906	rah->ar_hln = ETH_ALEN;
907	rah->ar_pln = 4;
908	rah->ar_op = htons(ARPOP_REPLY);
909	memcpy(rah->ar_sha, reh->h_source, ETH_ALEN);
910	memcpy(rah->ar_sip, ah->ar_tip, 4);
911	memcpy(rah->ar_tha, ah->ar_sha, ETH_ALEN);
912	memcpy(rah->ar_tip, ah->ar_sip, 4);
913	slirp_output(pData->pvUser, arp_reply, sizeof(arp_reply));
914	}
915	break;
916	default:
917	break;
918	}
919	}
920
921	void slirp_input(PNATState pData, const uint8_t *pkt, int pkt_len)
922	{
923	struct mbuf *m;
924	int proto;
925
926	if (pkt_len < ETH_HLEN)
927	return;
928
929	proto = ntohs((uint16_t )(pkt + 12));
930	switch(proto) {
931	case ETH_P_ARP:
932	arp_input(pData, pkt, pkt_len);
933	break;
934	case ETH_P_IP:
935	/* Update time. Important if the network is very quiet, as otherwise
936	* the first outgoing connection gets an incorrect timestamp. */
937	updtime(pData);
938
939	m = m_get(pData);
940	if (!m)
941	return;
942	/* Note: we add to align the IP header */
943	if (M_FREEROOM(m) < pkt_len + 2) {
944	m_inc(m, pkt_len + 2);
945	}
946	m->m_len = pkt_len + 2;
947	memcpy(m->m_data + 2, pkt, pkt_len);
948
949	m->m_data += 2 + ETH_HLEN;
950	m->m_len -= 2 + ETH_HLEN;
951
952	ip_input(pData, m);
953	break;
954	default:
955	break;
956	}
957	}
958
959	/* output the IP packet to the ethernet device */
960	void if_encap(PNATState pData, const uint8_t *ip_data, int ip_data_len)
961	{
962	uint8_t buf[1600];
963	struct ethhdr eh = (struct ethhdr )buf;
964
965	if (ip_data_len + ETH_HLEN > sizeof(buf))
966	return;
967
968	memcpy(eh->h_dest, client_ethaddr, ETH_ALEN);
969	memcpy(eh->h_source, special_ethaddr, ETH_ALEN - 1);
970	/* XXX: not correct */
971	eh->h_source[5] = CTL_ALIAS;
972	eh->h_proto = htons(ETH_P_IP);
973	memcpy(buf + sizeof(struct ethhdr), ip_data, ip_data_len);
974	slirp_output(pData->pvUser, buf, ip_data_len + ETH_HLEN);
975	}
976
977	int slirp_redir(PNATState pData, int is_udp, int host_port,
978	struct in_addr guest_addr, int guest_port)
979	{
980	if (is_udp) {
981	if (!udp_listen(pData, htons(host_port), guest_addr.s_addr,
982	htons(guest_port), 0))
983	return -1;
984	} else {
985	if (!solisten(pData, htons(host_port), guest_addr.s_addr,
986	htons(guest_port), 0))
987	return -1;
988	}
989	return 0;
990	}
991
992	int slirp_add_exec(PNATState pData, int do_pty, const char *args, int addr_low_byte,
993	int guest_port)
994	{
995	return add_exec(&exec_list, do_pty, (char *)args,
996	addr_low_byte, htons(guest_port));
997	}
998
999	void slirp_set_ethaddr(PNATState pData, const uint8_t *ethaddr)
1000	{
1001	memcpy(client_ethaddr, ethaddr, ETH_ALEN);
1002	}
1003
1004	#if defined(VBOX_WITH_SIMPLEFIED_SLIRP_SYNC) && defined(RT_OS_WINDOWS)
1005	HANDLE *slirp_get_events(PNATState pData)
1006	{
1007	return pData->phEvents;
1008	}
1009	void slirp_register_external_event(PNATState pData, HANDLE hEvent, int index)
1010	{
1011	pData->phEvents[index] = hEvent;
1012	}
1013	#endif

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

source: vbox/trunk/src/VBox/Devices/Network/slirp/slirp.c@ 14232

Download in other formats: