slirp.c@ 14161

Last change on this file since 14161 was 14161, checked in by vboxsync, 16 years ago
removing all references to cEvent from polling routine
Property svn:eol-style set to `native`
File size: 25.5 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 <iprt/assert.h>
8
9	static const uint8_t special_ethaddr[6] = {
10	0x52, 0x54, 0x00, 0x12, 0x35, 0x00
11	};
12
13	#ifdef _WIN32
14
15	static int get_dns_addr_domain(PNATState pData, bool fVerbose,
16	struct in_addr *pdns_addr,
17	const char **ppszDomain)
18	{
19	int rc = 0;
20	FIXED_INFO *FixedInfo=NULL;
21	ULONG BufLen;
22	DWORD ret;
23	IP_ADDR_STRING *pIPAddr;
24	struct in_addr tmp_addr;
25
26	FixedInfo = (FIXED_INFO *)GlobalAlloc(GPTR, sizeof(FIXED_INFO));
27	BufLen = sizeof(FIXED_INFO);
28
29	/** @todo: this API returns all DNS servers, no matter whether the
30	* corresponding network adapter is disabled or not. Maybe replace
31	* this by GetAdapterAddresses(), which is XP/Vista only though. */
32	if (ERROR_BUFFER_OVERFLOW == GetNetworkParams(FixedInfo, &BufLen)) {
33	if (FixedInfo) {
34	GlobalFree(FixedInfo);
35	FixedInfo = NULL;
36	}
37	FixedInfo = GlobalAlloc(GPTR, BufLen);
38	}
39
40	if ((ret = GetNetworkParams(FixedInfo, &BufLen)) != ERROR_SUCCESS) {
41	Log(("GetNetworkParams failed. ret = %08x\n", (u_int)ret ));
42	if (FixedInfo) {
43	GlobalFree(FixedInfo);
44	FixedInfo = NULL;
45	}
46	rc = -1;
47	goto get_dns_prefix;
48	}
49
50	pIPAddr = &(FixedInfo->DnsServerList);
51	inet_aton(pIPAddr->IpAddress.String, &tmp_addr);
52	Log(("nat: DNS Servers:\n"));
53	if (fVerbose \|\| pdns_addr->s_addr != tmp_addr.s_addr)
54	LogRel(("NAT: DNS address: %s\n", pIPAddr->IpAddress.String));
55	*pdns_addr = tmp_addr;
56
57	pIPAddr = FixedInfo -> DnsServerList.Next;
58	while ( pIPAddr )
59	{
60	if (fVerbose)
61	LogRel(("NAT: ignored DNS address: %s\n", pIPAddr ->IpAddress.String));
62	pIPAddr = pIPAddr ->Next;
63	}
64	if (FixedInfo) {
65	GlobalFree(FixedInfo);
66	FixedInfo = NULL;
67	}
68
69	get_dns_prefix:
70	if (ppszDomain)
71	{
72	OSVERSIONINFO ver;
73	char szDnsDomain[256];
74	DWORD dwSize = sizeof(szDnsDomain);
75
76	*ppszDomain = NULL;
77	GetVersionEx(&ver);
78	if (ver.dwMajorVersion >= 5)
79	{
80	/* GetComputerNameEx exists in Windows versions starting with 2000. */
81	if (GetComputerNameEx(ComputerNameDnsDomain, szDnsDomain, &dwSize))
82	{
83	if (szDnsDomain[0])
84	{
85	/* Just non-empty strings are valid. */
86	*ppszDomain = RTStrDup(szDnsDomain);
87	if (pData->fPassDomain)
88	{
89	if (fVerbose)
90	LogRel(("NAT: passing domain name %s\n", szDnsDomain));
91	}
92	else
93	Log(("nat: ignoring domain %s\n", szDnsDomain));
94	}
95	}
96	else
97	Log(("nat: GetComputerNameEx failed (%d)\n", GetLastError()));
98	}
99	}
100	return rc;
101	}
102
103	#else
104
105	static int get_dns_addr_domain(PNATState pData, bool fVerbose,
106	struct in_addr *pdns_addr,
107	const char **ppszDomain)
108	{
109	char buff[512];
110	char buff2[256];
111	FILE *f;
112	int found = 0;
113	struct in_addr tmp_addr;
114
115	#ifdef RT_OS_OS2
116	/* Try various locations. */
117	char *etc = getenv("ETC");
118	f = NULL;
119	if (etc)
120	{
121	snprintf(buff, sizeof(buff), "%s/RESOLV2", etc);
122	f = fopen(buff, "rt");
123	}
124	if (!f) {
125	snprintf(buff, sizeof(buff), "%s/RESOLV2", _PATH_ETC);
126	f = fopen(buff, "rt");
127	}
128	if (!f) {
129	snprintf(buff, sizeof(buff), "%s/resolv.conf", _PATH_ETC);
130	f = fopen(buff, "rt");
131	}
132	#else
133	f = fopen("/etc/resolv.conf", "r");
134	#endif
135	if (!f)
136	return -1;
137
138	if (ppszDomain)
139	*ppszDomain = NULL;
140	Log(("nat: DNS Servers:\n"));
141	while (fgets(buff, 512, f) != NULL) {
142	if (sscanf(buff, "nameserver%*[ \t]%256s", buff2) == 1) {
143	if (!inet_aton(buff2, &tmp_addr))
144	continue;
145	if (tmp_addr.s_addr == loopback_addr.s_addr)
146	tmp_addr = our_addr;
147	/* If it's the first one, set it to dns_addr */
148	if (!found)
149	{
150	if (fVerbose \|\| pdns_addr->s_addr != tmp_addr.s_addr)
151	LogRel(("NAT: DNS address: %s\n", buff2));
152	*pdns_addr = tmp_addr;
153	}
154	else
155	{
156	if (fVerbose)
157	LogRel(("NAT: ignored DNS address: %s\n", buff2));
158	}
159	found++;
160	}
161	if ( ppszDomain
162	&& (!strncmp(buff, "domain", 6) \|\| !strncmp(buff, "search", 6)))
163	{
164	/* Domain name/search list present. Pick first entry */
165	if (*ppszDomain == NULL)
166	{
167	char *tok;
168	char *saveptr;
169	tok = strtok_r(&buff[6], " \t\n", &saveptr);
170	if (tok)
171	{
172	*ppszDomain = RTStrDup(tok);
173	if (pData->fPassDomain)
174	{
175	if (fVerbose)
176	LogRel(("NAT: passing domain name %s\n", tok));
177	}
178	else
179	Log(("nat: ignoring domain %s\n", tok));
180	}
181	}
182	}
183	}
184	fclose(f);
185	if (!found)
186	return -1;
187	return 0;
188	}
189
190	#endif
191
192	int get_dns_addr(PNATState pData, struct in_addr *pdns_addr)
193	{
194	return get_dns_addr_domain(pData, false, pdns_addr, NULL);
195	}
196
197	int slirp_init(PNATState ppData, const char pszNetAddr, uint32_t u32Netmask,
198	bool fPassDomain, const char *pszTFTPPrefix,
199	const char pszBootFile, void pvUser)
200	{
201	int fNATfailed = 0;
202	#if defined(VBOX_WITH_SIMPLEFIED_SLIRP_SYNC) && defined(RT_OS_WINDOWS)
203	int i;
204	#endif
205	PNATState pData = malloc(sizeof(NATState));
206	*ppData = pData;
207	if (!pData)
208	return VERR_NO_MEMORY;
209	if (u32Netmask & 0x1f)
210	/* CTL is x.x.x.15, bootp passes up to 16 IPs (15..31) */
211	return VERR_INVALID_PARAMETER;
212	memset(pData, '\0', sizeof(NATState));
213	pData->fPassDomain = fPassDomain;
214	pData->pvUser = pvUser;
215	#if ARCH_BITS == 64
216	pData->cpvHashUsed = 1;
217	#endif
218	tftp_prefix = pszTFTPPrefix;
219	bootp_filename = pszBootFile;
220	pData->netmask = u32Netmask;
221
222	#ifdef _WIN32
223	{
224	WSADATA Data;
225	WSAStartup(MAKEWORD(2,0), &Data);
226	}
227	#ifdef VBOX_WITH_SIMPLEFIED_SLIRP_SYNC
228	/XXX:probably should be configurable/
229	pData->cMaxEvent = WSA_MAXIMUM_WAIT_EVENTS;
230	pData->phEvents = malloc(sizeof(HANDLE) * pData->cMaxEvent);
231	for (i = 1; i < WSA_MAXIMUM_WAIT_EVENTS; ++i) {
232	pData->phEvents[i] = CreateEvent(NULL, FALSE, FALSE, NULL);
233	}
234	#endif
235	#endif
236
237	Assert(sizeof(struct ip) == 20);
238	link_up = 1;
239
240	if_init(pData);
241	ip_init(pData);
242
243	/* Initialise mbufs after setting the MTU */
244	m_init(pData);
245
246	/* set default addresses */
247	inet_aton("127.0.0.1", &loopback_addr);
248	inet_aton("127.0.0.1", &dns_addr);
249
250	if (get_dns_addr_domain(pData, true, &dns_addr, &pData->pszDomain) < 0)
251	fNATfailed = 1;
252
253	inet_aton(pszNetAddr, &special_addr);
254	alias_addr.s_addr = special_addr.s_addr \| htonl(CTL_ALIAS);
255	getouraddr(pData);
256	return fNATfailed ? VINF_NAT_DNS : VINF_SUCCESS;
257	}
258
259	/**
260	* Marks the link as up, making it possible to establish new connections.
261	*/
262	void slirp_link_up(PNATState pData)
263	{
264	link_up = 1;
265	}
266
267	/**
268	* Marks the link as down and cleans up the current connections.
269	*/
270	void slirp_link_down(PNATState pData)
271	{
272	struct socket *so;
273
274	while ((so = tcb.so_next) != &tcb)
275	{
276	if (so->so_state & SS_NOFDREF \|\| so->s == -1)
277	sofree(pData, so);
278	else
279	tcp_drop(pData, sototcpcb(so), 0);
280	}
281
282	while ((so = udb.so_next) != &udb)
283	udp_detach(pData, so);
284
285	link_up = 0;
286	}
287
288	/**
289	* Terminates the slirp component.
290	*/
291	void slirp_term(PNATState pData)
292	{
293	if (pData->pszDomain)
294	RTStrFree((char )(void )pData->pszDomain);
295
296	#if ARCH_BITS == 64
297	LogRel(("NAT: cpvHashUsed=%RU32 cpvHashCollisions=%RU32 cpvHashInserts=%RU64 cpvHashDone=%RU64\n",
298	pData->cpvHashUsed, pData->cpvHashCollisions, pData->cpvHashInserts, pData->cpvHashDone));
299	#endif
300
301	slirp_link_down(pData);
302	#ifdef WIN32
303	WSACleanup();
304	#endif
305	#ifdef LOG_ENABLED
306	Log(("\n"
307	"NAT statistics\n"
308	"--------------\n"
309	"\n"));
310	ipstats(pData);
311	tcpstats(pData);
312	udpstats(pData);
313	icmpstats(pData);
314	mbufstats(pData);
315	sockstats(pData);
316	Log(("\n"
317	"\n"
318	"\n"));
319	#endif
320	#if defined(VBOX_WITH_SIMPLEFIED_SLIRP_SYNC) && defined(RT_OS_WINDOWS)
321	free(pData->phEvents);
322	#endif
323	free(pData);
324	}
325
326
327	#define CONN_CANFSEND(so) (((so)->so_state & (SS_FCANTSENDMORE\|SS_ISFCONNECTED)) == SS_ISFCONNECTED)
328	#define CONN_CANFRCV(so) (((so)->so_state & (SS_FCANTRCVMORE\|SS_ISFCONNECTED)) == SS_ISFCONNECTED)
329	#define UPD_NFDS(x) if (nfds < (x)) nfds = (x)
330
331	/*
332	* curtime kept to an accuracy of 1ms
333	*/
334	#ifdef _WIN32
335	static void updtime(PNATState pData)
336	{
337	struct _timeb tb;
338
339	_ftime(&tb);
340	curtime = (u_int)tb.time * (u_int)1000;
341	curtime += (u_int)tb.millitm;
342	}
343	#else
344	static void updtime(PNATState pData)
345	{
346	gettimeofday(&tt, 0);
347
348	curtime = (u_int)tt.tv_sec * (u_int)1000;
349	curtime += (u_int)tt.tv_usec / (u_int)1000;
350
351	if ((tt.tv_usec % 1000) >= 500)
352	curtime++;
353	}
354	#endif
355
356	void slirp_select_fill(PNATState pData, int *pnfds,
357	fd_set readfds, fd_set writefds, fd_set *xfds)
358	{
359	struct socket so, so_next;
360	struct timeval timeout;
361	int nfds;
362	int tmp_time;
363	#if defined(VBOX_WITH_SIMPLEFIED_SLIRP_SYNC) && defined(RT_OS_WINDOWS)
364	int rc;
365	/* Number of valid entries.
366	* 1st event for drvNATSend() */
367	int cElements;
368	#endif
369
370	nfds = *pnfds;
371	/*
372	* First, TCP sockets
373	*/
374	do_slowtimo = 0;
375	if (link_up) {
376	/*
377	* *_slowtimo needs calling if there are IP fragments
378	* in the fragment queue, or there are TCP connections active
379	*/
380	do_slowtimo = ((tcb.so_next != &tcb) \|\|
381	((struct ipasfrag )&ipq != u32_to_ptr(pData, ipq.next, struct ipasfrag )));
382
383	#if defined(VBOX_WITH_SIMPLEFIED_SLIRP_SYNC) && defined(RT_OS_WINDOWS)
384	/*
385	* Make this array static with a fixed maximum
386	* 1st event for drvNATSend()
387	*/
388	cElements = 1;
389	#endif
390
391	for (so = tcb.so_next; so != &tcb; so = so_next) {
392	so_next = so->so_next;
393
394	/*
395	* See if we need a tcp_fasttimo
396	*/
397	if (time_fasttimo == 0 && so->so_tcpcb->t_flags & TF_DELACK)
398	time_fasttimo = curtime; /* Flag when we want a fasttimo */
399
400	/*
401	* NOFDREF can include still connecting to local-host,
402	* newly socreated() sockets etc. Don't want to select these.
403	*/
404	if (so->so_state & SS_NOFDREF \|\| so->s == -1)
405	continue;
406
407	/*
408	* Set for reading sockets which are accepting
409	*/
410	if (so->so_state & SS_FACCEPTCONN) {
411	#if !defined(VBOX_WITH_SIMPLEFIED_SLIRP_SYNC) \|\| !defined(RT_OS_WINDOWS)
412	FD_SET(so->s, readfds);
413	UPD_NFDS(so->s);
414	#else
415	rc = WSAEventSelect(so->s, so->hNetworkEvent, FD_READ\|FD_WRITE\|FD_ACCEPT\|FD_CONNECT\|FD_OOB);
416	AssertRelease(rc != SOCKET_ERROR);
417	#endif
418	continue;
419	}
420
421	/*
422	* Set for writing sockets which are connecting
423	*/
424	if (so->so_state & SS_ISFCONNECTING) {
425	#if !defined(VBOX_WITH_SIMPLEFIED_SLIRP_SYNC) \|\| !defined(RT_OS_WINDOWS)
426	FD_SET(so->s, writefds);
427	UPD_NFDS(so->s);
428	#else
429	rc = WSAEventSelect(so->s, so->hNetworkEvent, FD_READ\|FD_WRITE\|FD_ACCEPT\|FD_CONNECT\|FD_OOB);
430	AssertRelease(rc != SOCKET_ERROR);
431	#endif
432	continue;
433	}
434
435	/*
436	* Set for writing if we are connected, can send more, and
437	* we have something to send
438	*/
439	if (CONN_CANFSEND(so) && so->so_rcv.sb_cc) {
440	#if !defined(VBOX_WITH_SIMPLEFIED_SLIRP_SYNC) \|\| !defined(RT_OS_WINDOWS)
441	FD_SET(so->s, writefds);
442	UPD_NFDS(so->s);
443	#else
444	rc = WSAEventSelect(so->s, so->hNetworkEvent, FD_READ\|FD_WRITE\|FD_ACCEPT\|FD_CONNECT\|FD_OOB);
445	AssertRelease(rc != SOCKET_ERROR);
446	continue; /XXX: we're using the widest mask for event/
447	#endif
448	}
449
450	/*
451	* Set for reading (and urgent data) if we are connected, can
452	* receive more, and we have room for it XXX /2 ?
453	*/
454	if (CONN_CANFRCV(so) && (so->so_snd.sb_cc < (so->so_snd.sb_datalen/2))) {
455	#if !defined(VBOX_WITH_SIMPLEFIED_SLIRP_SYNC) \|\| !defined(RT_OS_WINDOWS)
456	FD_SET(so->s, readfds);
457	FD_SET(so->s, xfds);
458	UPD_NFDS(so->s);
459	#else
460	rc = WSAEventSelect(so->s, so->hNetworkEvent, FD_OOB\|FD_READ\|FD_WRITE\|FD_ACCEPT\|FD_CONNECT);
461	AssertRelease(rc != SOCKET_ERROR);
462	continue; /XXX: we're using the widest mask for event/
463	#endif
464	}
465	}
466
467	/*
468	* UDP sockets
469	*/
470	for (so = udb.so_next; so != &udb; so = so_next) {
471	so_next = so->so_next;
472
473	/*
474	* See if it's timed out
475	*/
476	if (so->so_expire) {
477	if (so->so_expire <= curtime) {
478	udp_detach(pData, so);
479	continue;
480	} else
481	do_slowtimo = 1; /* Let socket expire */
482	}
483
484	/*
485	* When UDP packets are received from over the
486	* link, they're sendto()'d straight away, so
487	* no need for setting for writing
488	* Limit the number of packets queued by this session
489	* to 4. Note that even though we try and limit this
490	* to 4 packets, the session could have more queued
491	* if the packets needed to be fragmented
492	* (XXX <= 4 ?)
493	*/
494	if ((so->so_state & SS_ISFCONNECTED) && so->so_queued <= 4) {
495	#if !defined(VBOX_WITH_SIMPLEFIED_SLIRP_SYNC) \|\| !defined(RT_OS_WINDOWS)
496	FD_SET(so->s, readfds);
497	UPD_NFDS(so->s);
498	#else
499	rc = WSAEventSelect(so->s, so->hNetworkEvent, FD_READ\|FD_WRITE\|FD_OOB\|FD_ACCEPT);
500	AssertRelease(rc != SOCKET_ERROR);
501	#endif
502	}
503	}
504	}
505
506	/*
507	* Setup timeout to use minimum CPU usage, especially when idle
508	*/
509
510	/*
511	* First, see the timeout needed by *timo
512	*/
513	timeout.tv_sec = 0;
514	timeout.tv_usec = -1;
515	/*
516	* If a slowtimo is needed, set timeout to 500ms from the last
517	* slow timeout. If a fast timeout is needed, set timeout within
518	* 200ms of when it was requested.
519	*/
520	if (do_slowtimo) {
521	/* XXX + 10000 because some select()'s aren't that accurate */
522	timeout.tv_usec = ((500 - (curtime - last_slowtimo)) * 1000) + 10000;
523	if (timeout.tv_usec < 0)
524	timeout.tv_usec = 0;
525	else if (timeout.tv_usec > 510000)
526	timeout.tv_usec = 510000;
527
528	/* Can only fasttimo if we also slowtimo */
529	if (time_fasttimo) {
530	tmp_time = (200 - (curtime - time_fasttimo)) * 1000;
531	if (tmp_time < 0)
532	tmp_time = 0;
533
534	/* Choose the smallest of the 2 */
535	if (tmp_time < timeout.tv_usec)
536	timeout.tv_usec = (u_int)tmp_time;
537	}
538	}
539	#if !defined(VBOX_WITH_SIMPLEFIED_SLIRP_SYNC) \|\| !defined(RT_OS_WINDOWS)
540	*pnfds = nfds;
541	#else
542	*pnfds = WSA_MAXIMUM_WAIT_EVENTS;
543	#endif
544	}
545
546	void slirp_select_poll(PNATState pData, fd_set readfds, fd_set writefds, fd_set *xfds)
547	{
548	struct socket so, so_next;
549	int ret;
550	#if defined(VBOX_WITH_SIMPLEFIED_SLIRP_SYNC) && defined(RT_OS_WINDOWS)
551	WSANETWORKEVENTS NetworkEvents;
552	int rc;
553	int timer_update = (readfds == NULL && writefds == NULL && xfds == NULL);
554	#endif
555
556	/* Update time */
557	updtime(pData);
558
559	/*
560	* See if anything has timed out
561	*/
562	if (link_up) {
563	if (time_fasttimo && ((curtime - time_fasttimo) >= 2)) {
564	tcp_fasttimo(pData);
565	time_fasttimo = 0;
566	}
567	if (do_slowtimo && ((curtime - last_slowtimo) >= 499)) {
568	ip_slowtimo(pData);
569	tcp_slowtimo(pData);
570	last_slowtimo = curtime;
571	}
572	}
573	#if defined(VBOX_WITH_SIMPLEFIED_SLIRP_SYNC) && defined(RT_OS_WINDOWS)
574	if (timer_update) return;
575	#endif
576
577	/*
578	* Check sockets
579	*/
580	if (link_up) {
581	/*
582	* Check TCP sockets
583	*/
584	for (so = tcb.so_next; so != &tcb; so = so_next) {
585	so_next = so->so_next;
586
587	/*
588	* FD_ISSET is meaningless on these sockets
589	* (and they can crash the program)
590	*/
591	if (so->so_state & SS_NOFDREF \|\| so->s == -1)
592	continue;
593	#if defined(VBOX_WITH_SIMPLEFIED_SLIRP_SYNC) && defined(RT_OS_WINDOWS)
594	rc = WSAEnumNetworkEvents(so->s, so->hNetworkEvent, &NetworkEvents);
595	AssertRelease(rc != SOCKET_ERROR);
596	#endif
597
598	/*
599	* Check for URG data
600	* This will soread as well, so no need to
601	* test for readfds below if this succeeds
602	*/
603	#if !defined(VBOX_WITH_SIMPLEFIED_SLIRP_SYNC) \|\| !defined(RT_OS_WINDOWS)
604	if (FD_ISSET(so->s, xfds))
605	#else
606	if ((NetworkEvents.lNetworkEvents & FD_OOB) && NetworkEvents.iErrorCode[FD_OOB_BIT] == 0)
607	#endif
608	sorecvoob(pData, so);
609	/*
610	* Check sockets for reading
611	*/
612	#if !defined(VBOX_WITH_SIMPLEFIED_SLIRP_SYNC) \|\| !defined(RT_OS_WINDOWS)
613	else if (FD_ISSET(so->s, readfds)) {
614	#else
615	else if ((NetworkEvents.lNetworkEvents & FD_READ) && (NetworkEvents.iErrorCode[FD_READ_BIT] == 0)) {
616	#endif
617	/*
618	* Check for incoming connections
619	*/
620	if (so->so_state & SS_FACCEPTCONN) {
621	tcp_connect(pData, so);
622	continue;
623	} /* else */
624	ret = soread(pData, so);
625
626	/* Output it if we read something */
627	if (ret > 0)
628	tcp_output(pData, sototcpcb(so));
629	}
630
631	/*
632	* Check sockets for writing
633	*/
634	#if !defined(VBOX_WITH_SIMPLEFIED_SLIRP_SYNC) \|\| !defined(RT_OS_WINDOWS)
635	if (FD_ISSET(so->s, writefds)) {
636	#else
637	if ((NetworkEvents.lNetworkEvents & FD_WRITE) && (NetworkEvents.iErrorCode[FD_WRITE_BIT] == 0)) {
638	#endif
639	/*
640	* Check for non-blocking, still-connecting sockets
641	*/
642	if (so->so_state & SS_ISFCONNECTING) {
643	/* Connected */
644	so->so_state &= ~SS_ISFCONNECTING;
645
646	/*
647	* This should be probably guarded by PROBE_CONN too. Anyway,
648	* we disable it on OS/2 because the below send call returns
649	* EFAULT which causes the opened TCP socket to close right
650	* after it has been opened and connected.
651	*/
652	#ifndef RT_OS_OS2
653	ret = send(so->s, (const char *)&ret, 0, 0);
654	if (ret < 0) {
655	/* XXXXX Must fix, zero bytes is a NOP */
656	if (errno == EAGAIN \|\| errno == EWOULDBLOCK \|\|
657	errno == EINPROGRESS \|\| errno == ENOTCONN) {
658	continue;
659	}
660
661	/* else failed */
662	so->so_state = SS_NOFDREF;
663	}
664	/* else so->so_state &= ~SS_ISFCONNECTING; */
665	#endif
666
667	/*
668	* Continue tcp_input
669	*/
670	tcp_input(pData, (struct mbuf *)NULL, sizeof(struct ip), so);
671	/* continue; */
672	} else
673	ret = sowrite(pData, so);
674	/*
675	* XXXXX If we wrote something (a lot), there
676	* could be a need for a window update.
677	* In the worst case, the remote will send
678	* a window probe to get things going again
679	*/
680	}
681
682	/*
683	* Probe a still-connecting, non-blocking socket
684	* to check if it's still alive
685	*/
686	#ifdef PROBE_CONN
687	if (so->so_state & SS_ISFCONNECTING) {
688	ret = recv(so->s, (char *)&ret, 0,0);
689
690	if (ret < 0) {
691	/* XXX */
692	if (errno == EAGAIN \|\| errno == EWOULDBLOCK \|\|
693	errno == EINPROGRESS \|\| errno == ENOTCONN) {
694	continue; /* Still connecting, continue */
695	}
696
697	/* else failed */
698	so->so_state = SS_NOFDREF;
699
700	/* tcp_input will take care of it */
701	} else {
702	ret = send(so->s, &ret, 0,0);
703	if (ret < 0) {
704	/* XXX */
705	if (errno == EAGAIN \|\| errno == EWOULDBLOCK \|\|
706	errno == EINPROGRESS \|\| errno == ENOTCONN) {
707	continue;
708	}
709	/* else failed */
710	so->so_state = SS_NOFDREF;
711	} else
712	so->so_state &= ~SS_ISFCONNECTING;
713
714	}
715	tcp_input((struct mbuf *)NULL, sizeof(struct ip),so);
716	} /* SS_ISFCONNECTING */
717	#endif
718	}
719
720	/*
721	* Now UDP sockets.
722	* Incoming packets are sent straight away, they're not buffered.
723	* Incoming UDP data isn't buffered either.
724	*/
725	for (so = udb.so_next; so != &udb; so = so_next) {
726	so_next = so->so_next;
727
728	#if defined(VBOX_WITH_SIMPLEFIED_SLIRP_SYNC) && defined(RT_OS_WINDOWS)
729	rc = WSAEnumNetworkEvents(so->s, so->hNetworkEvent, &NetworkEvents);
730	AssertRelease(rc != SOCKET_ERROR);
731	#endif
732	#if !defined(VBOX_WITH_SIMPLEFIED_SLIRP_SYNC) \|\| !defined(RT_OS_WINDOWS)
733	if (so->s != -1 && FD_ISSET(so->s, readfds)) {
734	#else
735	if ((NetworkEvents.lNetworkEvents & FD_READ) && (NetworkEvents.iErrorCode[FD_READ_BIT] == 0)) {
736	#endif
737	sorecvfrom(pData, so);
738	}
739	}
740	}
741
742	/*
743	* See if we can start outputting
744	*/
745	if (if_queued && link_up)
746	if_start(pData);
747	}
748
749	#define ETH_ALEN 6
750	#define ETH_HLEN 14
751
752	#define ETH_P_IP 0x0800 /* Internet Protocol packet */
753	#define ETH_P_ARP 0x0806 /* Address Resolution packet */
754
755	#define ARPOP_REQUEST 1 /* ARP request */
756	#define ARPOP_REPLY 2 /* ARP reply */
757
758	struct ethhdr
759	{
760	unsigned char h_dest[ETH_ALEN]; /* destination eth addr */
761	unsigned char h_source[ETH_ALEN]; /* source ether addr */
762	unsigned short h_proto; /* packet type ID field */
763	};
764
765	struct arphdr
766	{
767	unsigned short ar_hrd; /* format of hardware address */
768	unsigned short ar_pro; /* format of protocol address */
769	unsigned char ar_hln; /* length of hardware address */
770	unsigned char ar_pln; /* length of protocol address */
771	unsigned short ar_op; /* ARP opcode (command) */
772
773	/*
774	* Ethernet looks like this : This bit is variable sized however...
775	*/
776	unsigned char ar_sha[ETH_ALEN]; /* sender hardware address */
777	unsigned char ar_sip[4]; /* sender IP address */
778	unsigned char ar_tha[ETH_ALEN]; /* target hardware address */
779	unsigned char ar_tip[4]; /* target IP address */
780	};
781
782	static
783	void arp_input(PNATState pData, const uint8_t *pkt, int pkt_len)
784	{
785	struct ethhdr eh = (struct ethhdr )pkt;
786	struct arphdr ah = (struct arphdr )(pkt + ETH_HLEN);
787	uint8_t arp_reply[ETH_HLEN + sizeof(struct arphdr)];
788	struct ethhdr reh = (struct ethhdr )arp_reply;
789	struct arphdr rah = (struct arphdr )(arp_reply + ETH_HLEN);
790	int ar_op;
791	struct ex_list *ex_ptr;
792	uint32_t htip = ntohl((uint32_t)ah->ar_tip);
793
794	ar_op = ntohs(ah->ar_op);
795	switch(ar_op) {
796	case ARPOP_REQUEST:
797	if ((htip & pData->netmask) == ntohl(special_addr.s_addr)) {
798	if ( (htip & ~pData->netmask) == CTL_DNS
799	\|\| (htip & ~pData->netmask) == CTL_ALIAS)
800	goto arp_ok;
801	for (ex_ptr = exec_list; ex_ptr; ex_ptr = ex_ptr->ex_next) {
802	if ((htip & ~pData->netmask) == ex_ptr->ex_addr)
803	goto arp_ok;
804	}
805	return;
806	arp_ok:
807	/* XXX: make an ARP request to have the client address */
808	memcpy(client_ethaddr, eh->h_source, ETH_ALEN);
809
810	/* ARP request for alias/dns mac address */
811	memcpy(reh->h_dest, pkt + ETH_ALEN, ETH_ALEN);
812	memcpy(reh->h_source, special_ethaddr, ETH_ALEN - 1);
813	reh->h_source[5] = ah->ar_tip[3];
814	reh->h_proto = htons(ETH_P_ARP);
815
816	rah->ar_hrd = htons(1);
817	rah->ar_pro = htons(ETH_P_IP);
818	rah->ar_hln = ETH_ALEN;
819	rah->ar_pln = 4;
820	rah->ar_op = htons(ARPOP_REPLY);
821	memcpy(rah->ar_sha, reh->h_source, ETH_ALEN);
822	memcpy(rah->ar_sip, ah->ar_tip, 4);
823	memcpy(rah->ar_tha, ah->ar_sha, ETH_ALEN);
824	memcpy(rah->ar_tip, ah->ar_sip, 4);
825	slirp_output(pData->pvUser, arp_reply, sizeof(arp_reply));
826	}
827	break;
828	default:
829	break;
830	}
831	}
832
833	void slirp_input(PNATState pData, const uint8_t *pkt, int pkt_len)
834	{
835	struct mbuf *m;
836	int proto;
837
838	if (pkt_len < ETH_HLEN)
839	return;
840
841	proto = ntohs((uint16_t )(pkt + 12));
842	switch(proto) {
843	case ETH_P_ARP:
844	arp_input(pData, pkt, pkt_len);
845	break;
846	case ETH_P_IP:
847	/* Update time. Important if the network is very quiet, as otherwise
848	* the first outgoing connection gets an incorrect timestamp. */
849	updtime(pData);
850
851	m = m_get(pData);
852	if (!m)
853	return;
854	/* Note: we add to align the IP header */
855	if (M_FREEROOM(m) < pkt_len + 2) {
856	m_inc(m, pkt_len + 2);
857	}
858	m->m_len = pkt_len + 2;
859	memcpy(m->m_data + 2, pkt, pkt_len);
860
861	m->m_data += 2 + ETH_HLEN;
862	m->m_len -= 2 + ETH_HLEN;
863
864	ip_input(pData, m);
865	break;
866	default:
867	break;
868	}
869	}
870
871	/* output the IP packet to the ethernet device */
872	void if_encap(PNATState pData, const uint8_t *ip_data, int ip_data_len)
873	{
874	uint8_t buf[1600];
875	struct ethhdr eh = (struct ethhdr )buf;
876
877	if (ip_data_len + ETH_HLEN > sizeof(buf))
878	return;
879
880	memcpy(eh->h_dest, client_ethaddr, ETH_ALEN);
881	memcpy(eh->h_source, special_ethaddr, ETH_ALEN - 1);
882	/* XXX: not correct */
883	eh->h_source[5] = CTL_ALIAS;
884	eh->h_proto = htons(ETH_P_IP);
885	memcpy(buf + sizeof(struct ethhdr), ip_data, ip_data_len);
886	slirp_output(pData->pvUser, buf, ip_data_len + ETH_HLEN);
887	}
888
889	int slirp_redir(PNATState pData, int is_udp, int host_port,
890	struct in_addr guest_addr, int guest_port)
891	{
892	if (is_udp) {
893	if (!udp_listen(pData, htons(host_port), guest_addr.s_addr,
894	htons(guest_port), 0))
895	return -1;
896	} else {
897	if (!solisten(pData, htons(host_port), guest_addr.s_addr,
898	htons(guest_port), 0))
899	return -1;
900	}
901	return 0;
902	}
903
904	int slirp_add_exec(PNATState pData, int do_pty, const char *args, int addr_low_byte,
905	int guest_port)
906	{
907	return add_exec(&exec_list, do_pty, (char *)args,
908	addr_low_byte, htons(guest_port));
909	}
910
911	void slirp_set_ethaddr(PNATState pData, const uint8_t *ethaddr)
912	{
913	memcpy(client_ethaddr, ethaddr, ETH_ALEN);
914	}
915
916	#if defined(VBOX_WITH_SIMPLEFIED_SLIRP_SYNC) && defined(RT_OS_WINDOWS)
917	HANDLE *slirp_get_events(PNATState pData)
918	{
919	return (pData->phEvents);
920	}
921	void slirp_register_external_event(PNATState pData, HANDLE hEvent)
922	{
923	pData->phEvents[0] = hEvent;
924	}
925	#endif

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

source: vbox/trunk/src/VBox/Devices/Network/slirp/slirp.c@ 14161

Download in other formats: