slirp.c@ 14529

Last change on this file since 14529 was 14529, checked in by vboxsync, 16 years ago
removing freeing of inlined array
Property svn:eol-style set to `native`
File size: 34.7 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_SIMPLIFIED_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	free(pData);
342	}
343
344
345	#define CONN_CANFSEND(so) (((so)->so_state & (SS_FCANTSENDMORE\|SS_ISFCONNECTED)) == SS_ISFCONNECTED)
346	#define CONN_CANFRCV(so) (((so)->so_state & (SS_FCANTRCVMORE\|SS_ISFCONNECTED)) == SS_ISFCONNECTED)
347	#define UPD_NFDS(x) if (nfds < (x)) nfds = (x)
348
349	/*
350	* curtime kept to an accuracy of 1ms
351	*/
352	#ifdef _WIN32
353	static void updtime(PNATState pData)
354	{
355	struct _timeb tb;
356
357	_ftime(&tb);
358	curtime = (u_int)tb.time * (u_int)1000;
359	curtime += (u_int)tb.millitm;
360	}
361	#else
362	static void updtime(PNATState pData)
363	{
364	gettimeofday(&tt, 0);
365
366	curtime = (u_int)tt.tv_sec * (u_int)1000;
367	curtime += (u_int)tt.tv_usec / (u_int)1000;
368
369	if ((tt.tv_usec % 1000) >= 500)
370	curtime++;
371	}
372	#endif
373
374	void slirp_select_fill(PNATState pData, int *pnfds,
375	fd_set readfds, fd_set writefds, fd_set *xfds)
376	{
377	struct socket so, so_next;
378	int nfds;
379	#if defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) && defined(RT_OS_WINDOWS)
380	int rc;
381	int error;
382	#endif
383	#ifdef VBOX_WITH_BSD_REASS
384	int i;
385	#endif /* VBOX_WITH_BSD_REASS */
386
387	STAM_REL_PROFILE_START(&pData->StatFill, a);
388
389	nfds = *pnfds;
390	/*
391	* First, TCP sockets
392	*/
393	do_slowtimo = 0;
394	if (link_up) {
395	/*
396	* *_slowtimo needs calling if there are IP fragments
397	* in the fragment queue, or there are TCP connections active
398	*/
399	#ifndef VBOX_WITH_BSD_REASS
400	do_slowtimo = ((tcb.so_next != &tcb) \|\|
401	((struct ipasfrag )&ipq != u32_to_ptr(pData, ipq.next, struct ipasfrag )));
402	#else /* !VBOX_WITH_BSD_REASS */
403	/* XXX: triggering of fragment expiration should be the same but use
404	* new macroses
405	*/
406	for (i = 0; i < IPREASS_NHASH; i++) {
407	if (!TAILQ_EMPTY(&ipq[i])) {
408	do_slowtimo = 1;
409	break;
410	}
411	}
412	do_slowtimo \|= (tcb.so_next != &tcb);
413	#endif /* VBOX_WITH_BSD_REASS */
414
415	STAM_REL_COUNTER_RESET(&pData->StatTCP);
416	STAM_REL_COUNTER_RESET(&pData->StatTCPHot);
417
418	for (so = tcb.so_next; so != &tcb; so = so_next) {
419	so_next = so->so_next;
420
421	STAM_REL_COUNTER_INC(&pData->StatTCP);
422
423	/*
424	* See if we need a tcp_fasttimo
425	*/
426	if (time_fasttimo == 0 && so->so_tcpcb->t_flags & TF_DELACK)
427	time_fasttimo = curtime; /* Flag when we want a fasttimo */
428
429	/*
430	* NOFDREF can include still connecting to local-host,
431	* newly socreated() sockets etc. Don't want to select these.
432	*/
433	if (so->so_state & SS_NOFDREF \|\| so->s == -1)
434	continue;
435
436	/*
437	* Set for reading sockets which are accepting
438	*/
439	if (so->so_state & SS_FACCEPTCONN) {
440	STAM_REL_COUNTER_INC(&pData->StatTCPHot);
441	#if !defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) \|\| !defined(RT_OS_WINDOWS)
442	FD_SET(so->s, readfds);
443	UPD_NFDS(so->s);
444	#else
445	rc = WSAEventSelect(so->s, VBOX_SOCKET_EVENT, FD_READ\|FD_WRITE\|FD_ACCEPT\|FD_CONNECT\|FD_OOB);
446	if (rc == SOCKET_ERROR)
447	{
448	socket_error:
449	error = WSAGetLastError();
450	LogRel(("WSAEventSelector error %d (so=%x, socket=%s, event=%x)\n", error, so, so->s, VBOX_SOCKET_EVENT));
451	}
452	#endif
453	continue;
454	}
455
456	/*
457	* Set for writing sockets which are connecting
458	*/
459	if (so->so_state & SS_ISFCONNECTING) {
460	STAM_REL_COUNTER_INC(&pData->StatTCPHot);
461	#if !defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) \|\| !defined(RT_OS_WINDOWS)
462	FD_SET(so->s, writefds);
463	UPD_NFDS(so->s);
464	#else
465	rc = WSAEventSelect(so->s, VBOX_SOCKET_EVENT, FD_READ\|FD_WRITE\|FD_ACCEPT\|FD_CONNECT\|FD_OOB);
466	if (rc == SOCKET_ERROR)
467	goto socket_error;
468	#endif
469	continue;
470	}
471
472	/*
473	* Set for writing if we are connected, can send more, and
474	* we have something to send
475	*/
476	if (CONN_CANFSEND(so) && so->so_rcv.sb_cc) {
477	STAM_REL_COUNTER_INC(&pData->StatTCPHot);
478	#if !defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) \|\| !defined(RT_OS_WINDOWS)
479	FD_SET(so->s, writefds);
480	UPD_NFDS(so->s);
481	#else
482	rc = WSAEventSelect(so->s, VBOX_SOCKET_EVENT, FD_READ\|FD_WRITE\|FD_ACCEPT\|FD_CONNECT\|FD_OOB);
483	if (rc == SOCKET_ERROR)
484	goto socket_error;
485	continue; /XXX: we're using the widest mask for event/
486	#endif
487	}
488
489	/*
490	* Set for reading (and urgent data) if we are connected, can
491	* receive more, and we have room for it XXX /2 ?
492	*/
493	if (CONN_CANFRCV(so) && (so->so_snd.sb_cc < (so->so_snd.sb_datalen/2))) {
494	STAM_REL_COUNTER_INC(&pData->StatTCPHot);
495	#if !defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) \|\| !defined(RT_OS_WINDOWS)
496	FD_SET(so->s, readfds);
497	FD_SET(so->s, xfds);
498	UPD_NFDS(so->s);
499	#else
500	rc = WSAEventSelect(so->s, VBOX_SOCKET_EVENT, FD_OOB\|FD_READ\|FD_WRITE\|FD_ACCEPT\|FD_CONNECT);
501	if (rc == SOCKET_ERROR)
502	goto socket_error;
503	continue; /XXX: we're using the widest mask for event/
504	#endif
505	}
506	#if defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) && defined(RT_OS_WINDOWS)
507	rc = WSAEventSelect(so->s, NULL, 0);
508	if (rc == SOCKET_ERROR)
509	goto socket_error;
510	#endif
511	}
512
513	/*
514	* UDP sockets
515	*/
516	STAM_REL_COUNTER_RESET(&pData->StatUDP);
517	STAM_REL_COUNTER_RESET(&pData->StatUDPHot);
518
519	for (so = udb.so_next; so != &udb; so = so_next) {
520	so_next = so->so_next;
521
522	STAM_REL_COUNTER_INC(&pData->StatUDP);
523
524	/*
525	* See if it's timed out
526	*/
527	if (so->so_expire) {
528	if (so->so_expire <= curtime) {
529	udp_detach(pData, so);
530	continue;
531	} else
532	do_slowtimo = 1; /* Let socket expire */
533	}
534
535	/*
536	* When UDP packets are received from over the
537	* link, they're sendto()'d straight away, so
538	* no need for setting for writing
539	* Limit the number of packets queued by this session
540	* to 4. Note that even though we try and limit this
541	* to 4 packets, the session could have more queued
542	* if the packets needed to be fragmented
543	* (XXX <= 4 ?)
544	*/
545	if ((so->so_state & SS_ISFCONNECTED) && so->so_queued <= 4) {
546	STAM_REL_COUNTER_INC(&pData->StatUDPHot);
547	#if !defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) \|\| !defined(RT_OS_WINDOWS)
548	FD_SET(so->s, readfds);
549	UPD_NFDS(so->s);
550	#else
551	rc = WSAEventSelect(so->s, VBOX_SOCKET_EVENT, FD_READ\|FD_WRITE\|FD_OOB\|FD_ACCEPT);
552	if (rc == SOCKET_ERROR)
553	goto socket_error;
554	continue;
555	#endif
556	}
557	#if defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) && defined(RT_OS_WINDOWS)
558	else
559	{
560	rc = WSAEventSelect(so->s, NULL, 0);
561	if (rc != SOCKET_ERROR)
562	goto socket_error;
563	}
564	#endif
565	}
566	}
567
568	#if !defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) \|\| !defined(RT_OS_WINDOWS)
569	*pnfds = nfds;
570	#else
571	*pnfds = VBOX_EVENT_COUNT;
572	#endif
573
574	STAM_REL_PROFILE_STOP(&pData->StatFill, a);
575	}
576
577	void slirp_select_poll(PNATState pData, fd_set readfds, fd_set writefds, fd_set *xfds)
578	{
579	struct socket so, so_next;
580	int ret;
581	#if defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) && defined(RT_OS_WINDOWS)
582	WSANETWORKEVENTS NetworkEvents;
583	int rc;
584	int error;
585	int timer_update = (readfds == NULL && writefds == NULL && xfds == NULL);
586	#endif
587	STAM_REL_PROFILE_START(&pData->StatPoll, a);
588
589	/* Update time */
590	updtime(pData);
591
592	/*
593	* See if anything has timed out
594	*/
595	if (link_up) {
596	if (time_fasttimo && ((curtime - time_fasttimo) >= 2)) {
597	STAM_REL_PROFILE_START(&pData->StatFastTimer, a);
598	tcp_fasttimo(pData);
599	time_fasttimo = 0;
600	STAM_REL_PROFILE_STOP(&pData->StatFastTimer, a);
601	}
602	if (do_slowtimo && ((curtime - last_slowtimo) >= 499)) {
603	STAM_REL_PROFILE_START(&pData->StatSlowTimer, a);
604	ip_slowtimo(pData);
605	tcp_slowtimo(pData);
606	last_slowtimo = curtime;
607	STAM_REL_PROFILE_STOP(&pData->StatSlowTimer, a);
608	}
609	}
610	#if defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) && defined(RT_OS_WINDOWS)
611	if (timer_update) return;
612	#endif
613
614	/*
615	* Check sockets
616	*/
617	if (link_up) {
618	/*
619	* Check TCP sockets
620	*/
621	for (so = tcb.so_next; so != &tcb; so = so_next) {
622	so_next = so->so_next;
623
624	/*
625	* FD_ISSET is meaningless on these sockets
626	* (and they can crash the program)
627	*/
628	if (so->so_state & SS_NOFDREF \|\| so->s == -1)
629	continue;
630	#if defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) && defined(RT_OS_WINDOWS)
631	rc = WSAEnumNetworkEvents(so->s, VBOX_SOCKET_EVENT, &NetworkEvents);
632	if (rc == SOCKET_ERROR)
633	{
634	error = WSAGetLastError();
635	LogRel(("WSAEnumNetworkEvents TCP error %d\n", error));
636	continue;
637	}
638	#endif
639
640	/*
641	* Check for URG data
642	* This will soread as well, so no need to
643	* test for readfds below if this succeeds
644	*/
645	#if !defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) \|\| !defined(RT_OS_WINDOWS)
646	if (FD_ISSET(so->s, xfds))
647	#else
648	/* out-of-band data */
649	if ((NetworkEvents.lNetworkEvents & FD_OOB) && NetworkEvents.iErrorCode[FD_OOB_BIT] == 0)
650	#endif
651	sorecvoob(pData, so);
652	/*
653	* Check sockets for reading
654	*/
655	#if !defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) \|\| !defined(RT_OS_WINDOWS)
656	else if (FD_ISSET(so->s, readfds)) {
657	#else
658	else if ((NetworkEvents.lNetworkEvents & FD_READ) && (NetworkEvents.iErrorCode[FD_READ_BIT] == 0)) {
659	#endif
660	/*
661	* Check for incoming connections
662	*/
663	if (so->so_state & SS_FACCEPTCONN) {
664	tcp_connect(pData, so);
665	continue;
666	} /* else */
667	ret = soread(pData, so);
668
669	/* Output it if we read something */
670	if (ret > 0)
671	tcp_output(pData, sototcpcb(so));
672	}
673
674	/*
675	* Check sockets for writing
676	*/
677	#if !defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) \|\| !defined(RT_OS_WINDOWS)
678	if (FD_ISSET(so->s, writefds)) {
679	#else
680	if ((NetworkEvents.lNetworkEvents & FD_WRITE) && (NetworkEvents.iErrorCode[FD_WRITE_BIT] == 0)) {
681	#endif
682	/*
683	* Check for non-blocking, still-connecting sockets
684	*/
685	if (so->so_state & SS_ISFCONNECTING) {
686	/* Connected */
687	so->so_state &= ~SS_ISFCONNECTING;
688
689	/*
690	* This should be probably guarded by PROBE_CONN too. Anyway,
691	* we disable it on OS/2 because the below send call returns
692	* EFAULT which causes the opened TCP socket to close right
693	* after it has been opened and connected.
694	*/
695	#ifndef RT_OS_OS2
696	ret = send(so->s, (const char *)&ret, 0, 0);
697	if (ret < 0) {
698	/* XXXXX Must fix, zero bytes is a NOP */
699	if (errno == EAGAIN \|\| errno == EWOULDBLOCK \|\|
700	errno == EINPROGRESS \|\| errno == ENOTCONN) {
701	continue;
702	}
703
704	/* else failed */
705	so->so_state = SS_NOFDREF;
706	}
707	/* else so->so_state &= ~SS_ISFCONNECTING; */
708	#endif
709
710	/*
711	* Continue tcp_input
712	*/
713	tcp_input(pData, (struct mbuf *)NULL, sizeof(struct ip), so);
714	/* continue; */
715	} else
716	ret = sowrite(pData, so);
717	/*
718	* XXXXX If we wrote something (a lot), there
719	* could be a need for a window update.
720	* In the worst case, the remote will send
721	* a window probe to get things going again
722	*/
723	}
724
725	/*
726	* Probe a still-connecting, non-blocking socket
727	* to check if it's still alive
728	*/
729	#ifdef PROBE_CONN
730	if (so->so_state & SS_ISFCONNECTING) {
731	ret = recv(so->s, (char *)&ret, 0,0);
732
733	if (ret < 0) {
734	/* XXX */
735	if (errno == EAGAIN \|\| errno == EWOULDBLOCK \|\|
736	errno == EINPROGRESS \|\| errno == ENOTCONN) {
737	continue; /* Still connecting, continue */
738	}
739
740	/* else failed */
741	so->so_state = SS_NOFDREF;
742
743	/* tcp_input will take care of it */
744	} else {
745	ret = send(so->s, &ret, 0,0);
746	if (ret < 0) {
747	/* XXX */
748	if (errno == EAGAIN \|\| errno == EWOULDBLOCK \|\|
749	errno == EINPROGRESS \|\| errno == ENOTCONN) {
750	continue;
751	}
752	/* else failed */
753	so->so_state = SS_NOFDREF;
754	} else
755	so->so_state &= ~SS_ISFCONNECTING;
756
757	}
758	tcp_input((struct mbuf *)NULL, sizeof(struct ip),so);
759	} /* SS_ISFCONNECTING */
760	#endif
761	}
762
763	/*
764	* Now UDP sockets.
765	* Incoming packets are sent straight away, they're not buffered.
766	* Incoming UDP data isn't buffered either.
767	*/
768	for (so = udb.so_next; so != &udb; so = so_next) {
769	so_next = so->so_next;
770
771	#if defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) && defined(RT_OS_WINDOWS)
772	rc = WSAEnumNetworkEvents(so->s, VBOX_SOCKET_EVENT, &NetworkEvents);
773	if (rc == SOCKET_ERROR)
774	{
775	error = WSAGetLastError();
776	LogRel(("WSAEnumNetworkEvents TCP error %d\n", error));
777	continue;
778	}
779	#endif
780	#if !defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) \|\| !defined(RT_OS_WINDOWS)
781	if (so->s != -1 && FD_ISSET(so->s, readfds)) {
782	#else
783	if ((NetworkEvents.lNetworkEvents & FD_READ) && (NetworkEvents.iErrorCode[FD_READ_BIT] == 0)) {
784	#endif
785	sorecvfrom(pData, so);
786	}
787	}
788	}
789
790	/*
791	* See if we can start outputting
792	*/
793	if (if_queued && link_up)
794	if_start(pData);
795
796	STAM_REL_PROFILE_STOP(&pData->StatPoll, a);
797	}
798
799	#define ETH_ALEN 6
800	#define ETH_HLEN 14
801
802	#define ETH_P_IP 0x0800 /* Internet Protocol packet */
803	#define ETH_P_ARP 0x0806 /* Address Resolution packet */
804
805	#define ARPOP_REQUEST 1 /* ARP request */
806	#define ARPOP_REPLY 2 /* ARP reply */
807
808	struct ethhdr
809	{
810	unsigned char h_dest[ETH_ALEN]; /* destination eth addr */
811	unsigned char h_source[ETH_ALEN]; /* source ether addr */
812	unsigned short h_proto; /* packet type ID field */
813	};
814
815	struct arphdr
816	{
817	unsigned short ar_hrd; /* format of hardware address */
818	unsigned short ar_pro; /* format of protocol address */
819	unsigned char ar_hln; /* length of hardware address */
820	unsigned char ar_pln; /* length of protocol address */
821	unsigned short ar_op; /* ARP opcode (command) */
822
823	/*
824	* Ethernet looks like this : This bit is variable sized however...
825	*/
826	unsigned char ar_sha[ETH_ALEN]; /* sender hardware address */
827	unsigned char ar_sip[4]; /* sender IP address */
828	unsigned char ar_tha[ETH_ALEN]; /* target hardware address */
829	unsigned char ar_tip[4]; /* target IP address */
830	};
831
832	static
833	void arp_input(PNATState pData, const uint8_t *pkt, int pkt_len)
834	{
835	struct ethhdr eh = (struct ethhdr )pkt;
836	struct arphdr ah = (struct arphdr )(pkt + ETH_HLEN);
837	uint8_t arp_reply[ETH_HLEN + sizeof(struct arphdr)];
838	struct ethhdr reh = (struct ethhdr )arp_reply;
839	struct arphdr rah = (struct arphdr )(arp_reply + ETH_HLEN);
840	int ar_op;
841	struct ex_list *ex_ptr;
842	uint32_t htip = ntohl((uint32_t)ah->ar_tip);
843
844	ar_op = ntohs(ah->ar_op);
845	switch(ar_op) {
846	case ARPOP_REQUEST:
847	if ((htip & pData->netmask) == ntohl(special_addr.s_addr)) {
848	if ( (htip & ~pData->netmask) == CTL_DNS
849	\|\| (htip & ~pData->netmask) == CTL_ALIAS)
850	goto arp_ok;
851	for (ex_ptr = exec_list; ex_ptr; ex_ptr = ex_ptr->ex_next) {
852	if ((htip & ~pData->netmask) == ex_ptr->ex_addr)
853	goto arp_ok;
854	}
855	return;
856	arp_ok:
857	/* XXX: make an ARP request to have the client address */
858	memcpy(client_ethaddr, eh->h_source, ETH_ALEN);
859
860	/* ARP request for alias/dns mac address */
861	memcpy(reh->h_dest, pkt + ETH_ALEN, ETH_ALEN);
862	memcpy(reh->h_source, special_ethaddr, ETH_ALEN - 1);
863	reh->h_source[5] = ah->ar_tip[3];
864	reh->h_proto = htons(ETH_P_ARP);
865
866	rah->ar_hrd = htons(1);
867	rah->ar_pro = htons(ETH_P_IP);
868	rah->ar_hln = ETH_ALEN;
869	rah->ar_pln = 4;
870	rah->ar_op = htons(ARPOP_REPLY);
871	memcpy(rah->ar_sha, reh->h_source, ETH_ALEN);
872	memcpy(rah->ar_sip, ah->ar_tip, 4);
873	memcpy(rah->ar_tha, ah->ar_sha, ETH_ALEN);
874	memcpy(rah->ar_tip, ah->ar_sip, 4);
875	slirp_output(pData->pvUser, arp_reply, sizeof(arp_reply));
876	}
877	break;
878	default:
879	break;
880	}
881	}
882
883	void slirp_input(PNATState pData, const uint8_t *pkt, int pkt_len)
884	{
885	struct mbuf *m;
886	int proto;
887
888	if (pkt_len < ETH_HLEN)
889	return;
890
891	proto = ntohs((uint16_t )(pkt + 12));
892	switch(proto) {
893	case ETH_P_ARP:
894	arp_input(pData, pkt, pkt_len);
895	break;
896	case ETH_P_IP:
897	/* Update time. Important if the network is very quiet, as otherwise
898	* the first outgoing connection gets an incorrect timestamp. */
899	updtime(pData);
900
901	m = m_get(pData);
902	if (!m)
903	return;
904	/* Note: we add to align the IP header */
905	if (M_FREEROOM(m) < pkt_len + 2) {
906	m_inc(m, pkt_len + 2);
907	}
908	m->m_len = pkt_len + 2;
909	memcpy(m->m_data + 2, pkt, pkt_len);
910
911	m->m_data += 2 + ETH_HLEN;
912	m->m_len -= 2 + ETH_HLEN;
913
914	ip_input(pData, m);
915	break;
916	default:
917	break;
918	}
919	}
920
921	/* output the IP packet to the ethernet device */
922	void if_encap(PNATState pData, const uint8_t *ip_data, int ip_data_len)
923	{
924	uint8_t buf[1600];
925	struct ethhdr eh = (struct ethhdr )buf;
926
927	if (ip_data_len + ETH_HLEN > sizeof(buf))
928	return;
929
930	memcpy(eh->h_dest, client_ethaddr, ETH_ALEN);
931	memcpy(eh->h_source, special_ethaddr, ETH_ALEN - 1);
932	/* XXX: not correct */
933	eh->h_source[5] = CTL_ALIAS;
934	eh->h_proto = htons(ETH_P_IP);
935	memcpy(buf + sizeof(struct ethhdr), ip_data, ip_data_len);
936	slirp_output(pData->pvUser, buf, ip_data_len + ETH_HLEN);
937	}
938
939	int slirp_redir(PNATState pData, int is_udp, int host_port,
940	struct in_addr guest_addr, int guest_port)
941	{
942	if (is_udp) {
943	if (!udp_listen(pData, htons(host_port), guest_addr.s_addr,
944	htons(guest_port), 0))
945	return -1;
946	} else {
947	if (!solisten(pData, htons(host_port), guest_addr.s_addr,
948	htons(guest_port), 0))
949	return -1;
950	}
951	return 0;
952	}
953
954	int slirp_add_exec(PNATState pData, int do_pty, const char *args, int addr_low_byte,
955	int guest_port)
956	{
957	return add_exec(&exec_list, do_pty, (char *)args,
958	addr_low_byte, htons(guest_port));
959	}
960
961	void slirp_set_ethaddr(PNATState pData, const uint8_t *ethaddr)
962	{
963	memcpy(client_ethaddr, ethaddr, ETH_ALEN);
964	}
965
966	#if defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) && defined(RT_OS_WINDOWS)
967	HANDLE *slirp_get_events(PNATState pData)
968	{
969	return pData->phEvents;
970	}
971	void slirp_register_external_event(PNATState pData, HANDLE hEvent, int index)
972	{
973	pData->phEvents[index] = hEvent;
974	}
975	#endif
976
977	unsigned int slirp_get_timeout_ms(PNATState pData)
978	{
979	if (link_up)
980	{
981	if (time_fasttimo)
982	return 2;
983	if (do_slowtimo)
984	return 500;
985	}
986	return 0;
987	}

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

source: vbox/trunk/src/VBox/Devices/Network/slirp/slirp.c@ 14529

Download in other formats: