slirp.c@ 14196

Last change on this file since 14196 was 14196, checked in by vboxsync, 16 years ago
NAT: Linux build fix; added some perfcounters
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/stam.h>
11	#include <VBox/pdmdrv.h>
12	#include <iprt/assert.h>
13
14	static const uint8_t special_ethaddr[6] = {
15	0x52, 0x54, 0x00, 0x12, 0x35, 0x00
16	};
17
18	#ifdef _WIN32
19
20	static int get_dns_addr_domain(PNATState pData, bool fVerbose,
21	struct in_addr *pdns_addr,
22	const char **ppszDomain)
23	{
24	int rc = 0;
25	FIXED_INFO *FixedInfo=NULL;
26	ULONG BufLen;
27	DWORD ret;
28	IP_ADDR_STRING *pIPAddr;
29	struct in_addr tmp_addr;
30
31	FixedInfo = (FIXED_INFO *)GlobalAlloc(GPTR, sizeof(FIXED_INFO));
32	BufLen = sizeof(FIXED_INFO);
33
34	/** @todo: this API returns all DNS servers, no matter whether the
35	* corresponding network adapter is disabled or not. Maybe replace
36	* this by GetAdapterAddresses(), which is XP/Vista only though. */
37	if (ERROR_BUFFER_OVERFLOW == GetNetworkParams(FixedInfo, &BufLen)) {
38	if (FixedInfo) {
39	GlobalFree(FixedInfo);
40	FixedInfo = NULL;
41	}
42	FixedInfo = GlobalAlloc(GPTR, BufLen);
43	}
44
45	if ((ret = GetNetworkParams(FixedInfo, &BufLen)) != ERROR_SUCCESS) {
46	Log(("GetNetworkParams failed. ret = %08x\n", (u_int)ret ));
47	if (FixedInfo) {
48	GlobalFree(FixedInfo);
49	FixedInfo = NULL;
50	}
51	rc = -1;
52	goto get_dns_prefix;
53	}
54
55	pIPAddr = &(FixedInfo->DnsServerList);
56	inet_aton(pIPAddr->IpAddress.String, &tmp_addr);
57	Log(("nat: DNS Servers:\n"));
58	if (fVerbose \|\| pdns_addr->s_addr != tmp_addr.s_addr)
59	LogRel(("NAT: DNS address: %s\n", pIPAddr->IpAddress.String));
60	*pdns_addr = tmp_addr;
61
62	pIPAddr = FixedInfo -> DnsServerList.Next;
63	while ( pIPAddr )
64	{
65	if (fVerbose)
66	LogRel(("NAT: ignored DNS address: %s\n", pIPAddr ->IpAddress.String));
67	pIPAddr = pIPAddr ->Next;
68	}
69	if (FixedInfo) {
70	GlobalFree(FixedInfo);
71	FixedInfo = NULL;
72	}
73
74	get_dns_prefix:
75	if (ppszDomain)
76	{
77	OSVERSIONINFO ver;
78	char szDnsDomain[256];
79	DWORD dwSize = sizeof(szDnsDomain);
80
81	*ppszDomain = NULL;
82	GetVersionEx(&ver);
83	if (ver.dwMajorVersion >= 5)
84	{
85	/* GetComputerNameEx exists in Windows versions starting with 2000. */
86	if (GetComputerNameEx(ComputerNameDnsDomain, szDnsDomain, &dwSize))
87	{
88	if (szDnsDomain[0])
89	{
90	/* Just non-empty strings are valid. */
91	*ppszDomain = RTStrDup(szDnsDomain);
92	if (pData->fPassDomain)
93	{
94	if (fVerbose)
95	LogRel(("NAT: passing domain name %s\n", szDnsDomain));
96	}
97	else
98	Log(("nat: ignoring domain %s\n", szDnsDomain));
99	}
100	}
101	else
102	Log(("nat: GetComputerNameEx failed (%d)\n", GetLastError()));
103	}
104	}
105	return rc;
106	}
107
108	#else
109
110	static int get_dns_addr_domain(PNATState pData, bool fVerbose,
111	struct in_addr *pdns_addr,
112	const char **ppszDomain)
113	{
114	char buff[512];
115	char buff2[256];
116	FILE *f;
117	int found = 0;
118	struct in_addr tmp_addr;
119
120	#ifdef RT_OS_OS2
121	/* Try various locations. */
122	char *etc = getenv("ETC");
123	f = NULL;
124	if (etc)
125	{
126	snprintf(buff, sizeof(buff), "%s/RESOLV2", etc);
127	f = fopen(buff, "rt");
128	}
129	if (!f) {
130	snprintf(buff, sizeof(buff), "%s/RESOLV2", _PATH_ETC);
131	f = fopen(buff, "rt");
132	}
133	if (!f) {
134	snprintf(buff, sizeof(buff), "%s/resolv.conf", _PATH_ETC);
135	f = fopen(buff, "rt");
136	}
137	#else
138	f = fopen("/etc/resolv.conf", "r");
139	#endif
140	if (!f)
141	return -1;
142
143	if (ppszDomain)
144	*ppszDomain = NULL;
145	Log(("nat: DNS Servers:\n"));
146	while (fgets(buff, 512, f) != NULL) {
147	if (sscanf(buff, "nameserver%*[ \t]%256s", buff2) == 1) {
148	if (!inet_aton(buff2, &tmp_addr))
149	continue;
150	if (tmp_addr.s_addr == loopback_addr.s_addr)
151	tmp_addr = our_addr;
152	/* If it's the first one, set it to dns_addr */
153	if (!found)
154	{
155	if (fVerbose \|\| pdns_addr->s_addr != tmp_addr.s_addr)
156	LogRel(("NAT: DNS address: %s\n", buff2));
157	*pdns_addr = tmp_addr;
158	}
159	else
160	{
161	if (fVerbose)
162	LogRel(("NAT: ignored DNS address: %s\n", buff2));
163	}
164	found++;
165	}
166	if ( ppszDomain
167	&& (!strncmp(buff, "domain", 6) \|\| !strncmp(buff, "search", 6)))
168	{
169	/* Domain name/search list present. Pick first entry */
170	if (*ppszDomain == NULL)
171	{
172	char *tok;
173	char *saveptr;
174	tok = strtok_r(&buff[6], " \t\n", &saveptr);
175	if (tok)
176	{
177	*ppszDomain = RTStrDup(tok);
178	if (pData->fPassDomain)
179	{
180	if (fVerbose)
181	LogRel(("NAT: passing domain name %s\n", tok));
182	}
183	else
184	Log(("nat: ignoring domain %s\n", tok));
185	}
186	}
187	}
188	}
189	fclose(f);
190	if (!found)
191	return -1;
192	return 0;
193	}
194
195	#endif
196
197	int get_dns_addr(PNATState pData, struct in_addr *pdns_addr)
198	{
199	return get_dns_addr_domain(pData, false, pdns_addr, NULL);
200	}
201
202	int slirp_init(PNATState ppData, const char pszNetAddr, uint32_t u32Netmask,
203	bool fPassDomain, const char *pszTFTPPrefix,
204	const char pszBootFile, void pvUser)
205	{
206	int fNATfailed = 0;
207	PNATState pData = malloc(sizeof(NATState));
208	*ppData = pData;
209	if (!pData)
210	return VERR_NO_MEMORY;
211	if (u32Netmask & 0x1f)
212	/* CTL is x.x.x.15, bootp passes up to 16 IPs (15..31) */
213	return VERR_INVALID_PARAMETER;
214	memset(pData, '\0', sizeof(NATState));
215	pData->fPassDomain = fPassDomain;
216	pData->pvUser = pvUser;
217	#if ARCH_BITS == 64
218	pData->cpvHashUsed = 1;
219	#endif
220	tftp_prefix = pszTFTPPrefix;
221	bootp_filename = pszBootFile;
222	pData->netmask = u32Netmask;
223
224	#ifdef _WIN32
225	{
226	WSADATA Data;
227	WSAStartup(MAKEWORD(2,0), &Data);
228	}
229	#if defined(VBOX_WITH_SIMPLEFIED_SLIRP_SYNC) && defined(RT_OS_WINDOWS)
230	pData->phEvents[VBOX_SOCKET_EVENT_INDEX] = CreateEvent(NULL, FALSE, FALSE, NULL);
231	#endif
232	#endif
233
234	Assert(sizeof(struct ip) == 20);
235	link_up = 1;
236
237	if_init(pData);
238	ip_init(pData);
239
240	/* Initialise mbufs after setting the MTU */
241	m_init(pData);
242
243	/* set default addresses */
244	inet_aton("127.0.0.1", &loopback_addr);
245	inet_aton("127.0.0.1", &dns_addr);
246
247	if (get_dns_addr_domain(pData, true, &dns_addr, &pData->pszDomain) < 0)
248	fNATfailed = 1;
249
250	inet_aton(pszNetAddr, &special_addr);
251	alias_addr.s_addr = special_addr.s_addr \| htonl(CTL_ALIAS);
252	getouraddr(pData);
253	return fNATfailed ? VINF_NAT_DNS : VINF_SUCCESS;
254	}
255
256	/**
257	* Statistics counters.
258	*/
259	void slirp_register_timers(PNATState pData, PPDMDRVINS pDrvIns)
260	{
261	#ifndef VBOX_WITHOUT_RELEASE_STATISTICS
262	PDMDrvHlpSTAMRegisterF(pDrvIns, &pData->StatFill, STAMTYPE_PROFILE, STAMVISIBILITY_ALWAYS,
263	STAMUNIT_TICKS_PER_CALL, "Profiling slirp fills", "/Drivers/NAT%d/Fill", pDrvIns->iInstance);
264	PDMDrvHlpSTAMRegisterF(pDrvIns, &pData->StatPoll, STAMTYPE_PROFILE, STAMVISIBILITY_ALWAYS,
265	STAMUNIT_TICKS_PER_CALL, "Profiling slirp polls", "/Drivers/NAT%d/Poll", pDrvIns->iInstance);
266	PDMDrvHlpSTAMRegisterF(pDrvIns, &pData->StatFastTimer, STAMTYPE_PROFILE, STAMVISIBILITY_ALWAYS,
267	STAMUNIT_TICKS_PER_CALL, "Profiling slirp fast timer", "/Drivers/NAT%d/TimerFast", pDrvIns->iInstance);
268	PDMDrvHlpSTAMRegisterF(pDrvIns, &pData->StatSlowTimer, STAMTYPE_PROFILE, STAMVISIBILITY_ALWAYS,
269	STAMUNIT_TICKS_PER_CALL, "Profiling slirp slow timer", "/Drivers/NAT%d/TimerSlow", pDrvIns->iInstance);
270	PDMDrvHlpSTAMRegisterF(pDrvIns, &pData->StatTCP, STAMTYPE_COUNTER, STAMVISIBILITY_ALWAYS,
271	STAMUNIT_COUNT, "TCP sockets", "/Drivers/NAT%d/SockTCP", pDrvIns->iInstance);
272	PDMDrvHlpSTAMRegisterF(pDrvIns, &pData->StatTCPHot, STAMTYPE_COUNTER, STAMVISIBILITY_ALWAYS,
273	STAMUNIT_COUNT, "TCP sockets active", "/Drivers/NAT%d/SockTCPHot", pDrvIns->iInstance);
274	PDMDrvHlpSTAMRegisterF(pDrvIns, &pData->StatUDP, STAMTYPE_COUNTER, STAMVISIBILITY_ALWAYS,
275	STAMUNIT_COUNT, "UDP sockets", "/Drivers/NAT%d/SockUDP", pDrvIns->iInstance);
276	PDMDrvHlpSTAMRegisterF(pDrvIns, &pData->StatTCPHot, STAMTYPE_COUNTER, STAMVISIBILITY_ALWAYS,
277	STAMUNIT_COUNT, "UDP sockets active", "/Drivers/NAT%d/SockUDPHot", pDrvIns->iInstance);
278	#endif /* VBOX_WITHOUT_RELEASE_STATISTICS */
279	}
280
281	/**
282	* Marks the link as up, making it possible to establish new connections.
283	*/
284	void slirp_link_up(PNATState pData)
285	{
286	link_up = 1;
287	}
288
289	/**
290	* Marks the link as down and cleans up the current connections.
291	*/
292	void slirp_link_down(PNATState pData)
293	{
294	struct socket *so;
295
296	while ((so = tcb.so_next) != &tcb)
297	{
298	if (so->so_state & SS_NOFDREF \|\| so->s == -1)
299	sofree(pData, so);
300	else
301	tcp_drop(pData, sototcpcb(so), 0);
302	}
303
304	while ((so = udb.so_next) != &udb)
305	udp_detach(pData, so);
306
307	link_up = 0;
308	}
309
310	/**
311	* Terminates the slirp component.
312	*/
313	void slirp_term(PNATState pData)
314	{
315	if (pData->pszDomain)
316	RTStrFree((char )(void )pData->pszDomain);
317
318	#if ARCH_BITS == 64
319	LogRel(("NAT: cpvHashUsed=%RU32 cpvHashCollisions=%RU32 cpvHashInserts=%RU64 cpvHashDone=%RU64\n",
320	pData->cpvHashUsed, pData->cpvHashCollisions, pData->cpvHashInserts, pData->cpvHashDone));
321	#endif
322
323	slirp_link_down(pData);
324	#ifdef WIN32
325	WSACleanup();
326	#endif
327	#ifdef LOG_ENABLED
328	Log(("\n"
329	"NAT statistics\n"
330	"--------------\n"
331	"\n"));
332	ipstats(pData);
333	tcpstats(pData);
334	udpstats(pData);
335	icmpstats(pData);
336	mbufstats(pData);
337	sockstats(pData);
338	Log(("\n"
339	"\n"
340	"\n"));
341	#endif
342	#if defined(VBOX_WITH_SIMPLEFIED_SLIRP_SYNC) && defined(RT_OS_WINDOWS)
343	free(pData->phEvents);
344	#endif
345	free(pData);
346	}
347
348
349	#define CONN_CANFSEND(so) (((so)->so_state & (SS_FCANTSENDMORE\|SS_ISFCONNECTED)) == SS_ISFCONNECTED)
350	#define CONN_CANFRCV(so) (((so)->so_state & (SS_FCANTRCVMORE\|SS_ISFCONNECTED)) == SS_ISFCONNECTED)
351	#define UPD_NFDS(x) if (nfds < (x)) nfds = (x)
352
353	/*
354	* curtime kept to an accuracy of 1ms
355	*/
356	#ifdef _WIN32
357	static void updtime(PNATState pData)
358	{
359	struct _timeb tb;
360
361	_ftime(&tb);
362	curtime = (u_int)tb.time * (u_int)1000;
363	curtime += (u_int)tb.millitm;
364	}
365	#else
366	static void updtime(PNATState pData)
367	{
368	gettimeofday(&tt, 0);
369
370	curtime = (u_int)tt.tv_sec * (u_int)1000;
371	curtime += (u_int)tt.tv_usec / (u_int)1000;
372
373	if ((tt.tv_usec % 1000) >= 500)
374	curtime++;
375	}
376	#endif
377
378	void slirp_select_fill(PNATState pData, int *pnfds,
379	fd_set readfds, fd_set writefds, fd_set *xfds)
380	{
381	struct socket so, so_next;
382	struct timeval timeout;
383	int nfds;
384	int tmp_time;
385	#if defined(VBOX_WITH_SIMPLEFIED_SLIRP_SYNC) && defined(RT_OS_WINDOWS)
386	int rc;
387	int error;
388	#endif
389
390	STAM_REL_PROFILE_START(&pData->StatFill, a);
391
392	nfds = *pnfds;
393	/*
394	* First, TCP sockets
395	*/
396	do_slowtimo = 0;
397	if (link_up) {
398	/*
399	* *_slowtimo needs calling if there are IP fragments
400	* in the fragment queue, or there are TCP connections active
401	*/
402	do_slowtimo = ((tcb.so_next != &tcb) \|\|
403	((struct ipasfrag )&ipq != u32_to_ptr(pData, ipq.next, struct ipasfrag )));
404
405	STAM_REL_COUNTER_RESET(&pData->StatTCP);
406	STAM_REL_COUNTER_RESET(&pData->StatTCPHot);
407
408	for (so = tcb.so_next; so != &tcb; so = so_next) {
409	so_next = so->so_next;
410
411	STAM_REL_COUNTER_INC(&pData->StatTCP);
412
413	/*
414	* See if we need a tcp_fasttimo
415	*/
416	if (time_fasttimo == 0 && so->so_tcpcb->t_flags & TF_DELACK)
417	time_fasttimo = curtime; /* Flag when we want a fasttimo */
418
419	/*
420	* NOFDREF can include still connecting to local-host,
421	* newly socreated() sockets etc. Don't want to select these.
422	*/
423	if (so->so_state & SS_NOFDREF \|\| so->s == -1)
424	continue;
425
426	/*
427	* Set for reading sockets which are accepting
428	*/
429	if (so->so_state & SS_FACCEPTCONN) {
430	STAM_REL_COUNTER_INC(&pData->StatTCPHot);
431	#if !defined(VBOX_WITH_SIMPLEFIED_SLIRP_SYNC) \|\| !defined(RT_OS_WINDOWS)
432	FD_SET(so->s, readfds);
433	UPD_NFDS(so->s);
434	#else
435	rc = WSAEventSelect(so->s, VBOX_SOCKET_EVENT, FD_READ\|FD_WRITE\|FD_ACCEPT\|FD_CONNECT\|FD_OOB);
436	if (rc == SOCKET_ERROR)
437	{
438	socket_error:
439	error = WSAGetLastError();
440	LogRel(("WSAEventSelector error %d (so=%x, socket=%s, event=%x)\n", error, so, so->s, VBOX_SOCKET_EVENT));
441	}
442	#endif
443	continue;
444	}
445
446	/*
447	* Set for writing sockets which are connecting
448	*/
449	if (so->so_state & SS_ISFCONNECTING) {
450	STAM_REL_COUNTER_INC(&pData->StatTCPHot);
451	#if !defined(VBOX_WITH_SIMPLEFIED_SLIRP_SYNC) \|\| !defined(RT_OS_WINDOWS)
452	FD_SET(so->s, writefds);
453	UPD_NFDS(so->s);
454	#else
455	rc = WSAEventSelect(so->s, VBOX_SOCKET_EVENT, FD_READ\|FD_WRITE\|FD_ACCEPT\|FD_CONNECT\|FD_OOB);
456	if (rc == SOCKET_ERROR)
457	goto socket_error;
458	#endif
459	continue;
460	}
461
462	/*
463	* Set for writing if we are connected, can send more, and
464	* we have something to send
465	*/
466	if (CONN_CANFSEND(so) && so->so_rcv.sb_cc) {
467	STAM_REL_COUNTER_INC(&pData->StatTCPHot);
468	#if !defined(VBOX_WITH_SIMPLEFIED_SLIRP_SYNC) \|\| !defined(RT_OS_WINDOWS)
469	FD_SET(so->s, writefds);
470	UPD_NFDS(so->s);
471	#else
472	rc = WSAEventSelect(so->s, VBOX_SOCKET_EVENT, FD_READ\|FD_WRITE\|FD_ACCEPT\|FD_CONNECT\|FD_OOB);
473	if (rc == SOCKET_ERROR)
474	goto socket_error;
475	continue; /XXX: we're using the widest mask for event/
476	#endif
477	}
478
479	/*
480	* Set for reading (and urgent data) if we are connected, can
481	* receive more, and we have room for it XXX /2 ?
482	*/
483	if (CONN_CANFRCV(so) && (so->so_snd.sb_cc < (so->so_snd.sb_datalen/2))) {
484	STAM_REL_COUNTER_INC(&pData->StatTCPHot);
485	#if !defined(VBOX_WITH_SIMPLEFIED_SLIRP_SYNC) \|\| !defined(RT_OS_WINDOWS)
486	FD_SET(so->s, readfds);
487	FD_SET(so->s, xfds);
488	UPD_NFDS(so->s);
489	#else
490	rc = WSAEventSelect(so->s, VBOX_SOCKET_EVENT, FD_OOB\|FD_READ\|FD_WRITE\|FD_ACCEPT\|FD_CONNECT);
491	if (rc == SOCKET_ERROR)
492	goto socket_error;
493	continue; /XXX: we're using the widest mask for event/
494	#endif
495	}
496	#if defined(VBOX_WITH_SIMPLEFIED_SLIRP_SYNC) && defined(RT_OS_WINDOWS)
497	rc = WSAEventSelect(so->s, NULL, 0);
498	if (rc == SOCKET_ERROR)
499	goto socket_error;
500	#endif
501	}
502
503	/*
504	* UDP sockets
505	*/
506	STAM_REL_COUNTER_RESET(&pData->StatUDP);
507	STAM_REL_COUNTER_RESET(&pData->StatUDPHot);
508
509	for (so = udb.so_next; so != &udb; so = so_next) {
510	so_next = so->so_next;
511
512	STAM_REL_COUNTER_INC(&pData->StatUDP);
513
514	/*
515	* See if it's timed out
516	*/
517	if (so->so_expire) {
518	if (so->so_expire <= curtime) {
519	udp_detach(pData, so);
520	continue;
521	} else
522	do_slowtimo = 1; /* Let socket expire */
523	}
524
525	/*
526	* When UDP packets are received from over the
527	* link, they're sendto()'d straight away, so
528	* no need for setting for writing
529	* Limit the number of packets queued by this session
530	* to 4. Note that even though we try and limit this
531	* to 4 packets, the session could have more queued
532	* if the packets needed to be fragmented
533	* (XXX <= 4 ?)
534	*/
535	if ((so->so_state & SS_ISFCONNECTED) && so->so_queued <= 4) {
536	STAM_REL_COUNTER_INC(&pData->StatUDPHot);
537	#if !defined(VBOX_WITH_SIMPLEFIED_SLIRP_SYNC) \|\| !defined(RT_OS_WINDOWS)
538	FD_SET(so->s, readfds);
539	UPD_NFDS(so->s);
540	#else
541	rc = WSAEventSelect(so->s, VBOX_SOCKET_EVENT, FD_READ\|FD_WRITE\|FD_OOB\|FD_ACCEPT);
542	if (rc == SOCKET_ERROR)
543	goto socket_error;
544	continue;
545	#endif
546	STAM_REL_COUNTER_INC(&pData->StatUDPHot);
547	}
548	#if defined(VBOX_WITH_SIMPLEFIED_SLIRP_SYNC) && defined(RT_OS_WINDOWS)
549	else
550	{
551	rc = WSAEventSelect(so->s, NULL, 0);
552	if (rc != SOCKET_ERROR)
553	goto socket_error;
554	}
555	#endif
556	}
557	}
558
559	/*
560	* Setup timeout to use minimum CPU usage, especially when idle
561	*/
562
563	/*
564	* First, see the timeout needed by *timo
565	*/
566	timeout.tv_sec = 0;
567	timeout.tv_usec = -1;
568	/*
569	* If a slowtimo is needed, set timeout to 500ms from the last
570	* slow timeout. If a fast timeout is needed, set timeout within
571	* 200ms of when it was requested.
572	*/
573	if (do_slowtimo) {
574	/* XXX + 10000 because some select()'s aren't that accurate */
575	timeout.tv_usec = ((500 - (curtime - last_slowtimo)) * 1000) + 10000;
576	if (timeout.tv_usec < 0)
577	timeout.tv_usec = 0;
578	else if (timeout.tv_usec > 510000)
579	timeout.tv_usec = 510000;
580
581	/* Can only fasttimo if we also slowtimo */
582	if (time_fasttimo) {
583	tmp_time = (200 - (curtime - time_fasttimo)) * 1000;
584	if (tmp_time < 0)
585	tmp_time = 0;
586
587	/* Choose the smallest of the 2 */
588	if (tmp_time < timeout.tv_usec)
589	timeout.tv_usec = (u_int)tmp_time;
590	}
591	}
592	#if !defined(VBOX_WITH_SIMPLEFIED_SLIRP_SYNC) \|\| !defined(RT_OS_WINDOWS)
593	*pnfds = nfds;
594	#else
595	*pnfds = VBOX_EVENT_COUNT;
596	#endif
597
598	STAM_REL_PROFILE_STOP(&pData->StatFill, a);
599	}
600
601	void slirp_select_poll(PNATState pData, fd_set readfds, fd_set writefds, fd_set *xfds)
602	{
603	struct socket so, so_next;
604	int ret;
605	#if defined(VBOX_WITH_SIMPLEFIED_SLIRP_SYNC) && defined(RT_OS_WINDOWS)
606	WSANETWORKEVENTS NetworkEvents;
607	int rc;
608	int error;
609	int timer_update = (readfds == NULL && writefds == NULL && xfds == NULL);
610	#endif
611	STAM_REL_PROFILE_START(&pData->StatPoll, a);
612
613	/* Update time */
614	updtime(pData);
615
616	/*
617	* See if anything has timed out
618	*/
619	if (link_up) {
620	if (time_fasttimo && ((curtime - time_fasttimo) >= 2)) {
621	STAM_REL_PROFILE_START(&pData->StatFastTimer, a);
622	tcp_fasttimo(pData);
623	time_fasttimo = 0;
624	STAM_REL_PROFILE_STOP(&pData->StatFastTimer, a);
625	}
626	if (do_slowtimo && ((curtime - last_slowtimo) >= 499)) {
627	STAM_REL_PROFILE_START(&pData->StatSlowTimer, a);
628	ip_slowtimo(pData);
629	tcp_slowtimo(pData);
630	last_slowtimo = curtime;
631	STAM_REL_PROFILE_STOP(&pData->StatSlowTimer, a);
632	}
633	}
634	#if defined(VBOX_WITH_SIMPLEFIED_SLIRP_SYNC) && defined(RT_OS_WINDOWS)
635	if (timer_update) return;
636	#endif
637
638	/*
639	* Check sockets
640	*/
641	if (link_up) {
642	/*
643	* Check TCP sockets
644	*/
645	for (so = tcb.so_next; so != &tcb; so = so_next) {
646	so_next = so->so_next;
647
648	/*
649	* FD_ISSET is meaningless on these sockets
650	* (and they can crash the program)
651	*/
652	if (so->so_state & SS_NOFDREF \|\| so->s == -1)
653	continue;
654	#if defined(VBOX_WITH_SIMPLEFIED_SLIRP_SYNC) && defined(RT_OS_WINDOWS)
655	rc = WSAEnumNetworkEvents(so->s, so->hNetworkEvent, &NetworkEvents);
656	if (rc == SOCKET_ERROR)
657	{
658	error = WSAGetLastError();
659	LogRel(("WSAEnumNetworkEvents TCP error %d\n", error));
660	continue;
661	}
662	#endif
663
664	/*
665	* Check for URG data
666	* This will soread as well, so no need to
667	* test for readfds below if this succeeds
668	*/
669	#if !defined(VBOX_WITH_SIMPLEFIED_SLIRP_SYNC) \|\| !defined(RT_OS_WINDOWS)
670	if (FD_ISSET(so->s, xfds))
671	#else
672	/* out-of-band data */
673	if ((NetworkEvents.lNetworkEvents & FD_OOB) && NetworkEvents.iErrorCode[FD_OOB_BIT] == 0)
674	#endif
675	sorecvoob(pData, so);
676	/*
677	* Check sockets for reading
678	*/
679	#if !defined(VBOX_WITH_SIMPLEFIED_SLIRP_SYNC) \|\| !defined(RT_OS_WINDOWS)
680	else if (FD_ISSET(so->s, readfds)) {
681	#else
682	else if ((NetworkEvents.lNetworkEvents & FD_READ) && (NetworkEvents.iErrorCode[FD_READ_BIT] == 0)) {
683	#endif
684	/*
685	* Check for incoming connections
686	*/
687	if (so->so_state & SS_FACCEPTCONN) {
688	tcp_connect(pData, so);
689	continue;
690	} /* else */
691	ret = soread(pData, so);
692
693	/* Output it if we read something */
694	if (ret > 0)
695	tcp_output(pData, sototcpcb(so));
696	}
697
698	/*
699	* Check sockets for writing
700	*/
701	#if !defined(VBOX_WITH_SIMPLEFIED_SLIRP_SYNC) \|\| !defined(RT_OS_WINDOWS)
702	if (FD_ISSET(so->s, writefds)) {
703	#else
704	if ((NetworkEvents.lNetworkEvents & FD_WRITE) && (NetworkEvents.iErrorCode[FD_WRITE_BIT] == 0)) {
705	#endif
706	/*
707	* Check for non-blocking, still-connecting sockets
708	*/
709	if (so->so_state & SS_ISFCONNECTING) {
710	/* Connected */
711	so->so_state &= ~SS_ISFCONNECTING;
712
713	/*
714	* This should be probably guarded by PROBE_CONN too. Anyway,
715	* we disable it on OS/2 because the below send call returns
716	* EFAULT which causes the opened TCP socket to close right
717	* after it has been opened and connected.
718	*/
719	#ifndef RT_OS_OS2
720	ret = send(so->s, (const char *)&ret, 0, 0);
721	if (ret < 0) {
722	/* XXXXX Must fix, zero bytes is a NOP */
723	if (errno == EAGAIN \|\| errno == EWOULDBLOCK \|\|
724	errno == EINPROGRESS \|\| errno == ENOTCONN) {
725	continue;
726	}
727
728	/* else failed */
729	so->so_state = SS_NOFDREF;
730	}
731	/* else so->so_state &= ~SS_ISFCONNECTING; */
732	#endif
733
734	/*
735	* Continue tcp_input
736	*/
737	tcp_input(pData, (struct mbuf *)NULL, sizeof(struct ip), so);
738	/* continue; */
739	} else
740	ret = sowrite(pData, so);
741	/*
742	* XXXXX If we wrote something (a lot), there
743	* could be a need for a window update.
744	* In the worst case, the remote will send
745	* a window probe to get things going again
746	*/
747	}
748
749	/*
750	* Probe a still-connecting, non-blocking socket
751	* to check if it's still alive
752	*/
753	#ifdef PROBE_CONN
754	if (so->so_state & SS_ISFCONNECTING) {
755	ret = recv(so->s, (char *)&ret, 0,0);
756
757	if (ret < 0) {
758	/* XXX */
759	if (errno == EAGAIN \|\| errno == EWOULDBLOCK \|\|
760	errno == EINPROGRESS \|\| errno == ENOTCONN) {
761	continue; /* Still connecting, continue */
762	}
763
764	/* else failed */
765	so->so_state = SS_NOFDREF;
766
767	/* tcp_input will take care of it */
768	} else {
769	ret = send(so->s, &ret, 0,0);
770	if (ret < 0) {
771	/* XXX */
772	if (errno == EAGAIN \|\| errno == EWOULDBLOCK \|\|
773	errno == EINPROGRESS \|\| errno == ENOTCONN) {
774	continue;
775	}
776	/* else failed */
777	so->so_state = SS_NOFDREF;
778	} else
779	so->so_state &= ~SS_ISFCONNECTING;
780
781	}
782	tcp_input((struct mbuf *)NULL, sizeof(struct ip),so);
783	} /* SS_ISFCONNECTING */
784	#endif
785	}
786
787	/*
788	* Now UDP sockets.
789	* Incoming packets are sent straight away, they're not buffered.
790	* Incoming UDP data isn't buffered either.
791	*/
792	for (so = udb.so_next; so != &udb; so = so_next) {
793	so_next = so->so_next;
794
795	#if defined(VBOX_WITH_SIMPLEFIED_SLIRP_SYNC) && defined(RT_OS_WINDOWS)
796	rc = WSAEnumNetworkEvents(so->s, VBOX_SOCKET_EVENT, &NetworkEvents);
797	if (rc == SOCKET_ERROR)
798	{
799	error = WSAGetLastError();
800	LogRel(("WSAEnumNetworkEvents TCP error %d\n", error));
801	continue;
802	}
803	#endif
804	#if !defined(VBOX_WITH_SIMPLEFIED_SLIRP_SYNC) \|\| !defined(RT_OS_WINDOWS)
805	if (so->s != -1 && FD_ISSET(so->s, readfds)) {
806	#else
807	if ((NetworkEvents.lNetworkEvents & FD_READ) && (NetworkEvents.iErrorCode[FD_READ_BIT] == 0)) {
808	#endif
809	sorecvfrom(pData, so);
810	}
811	}
812	}
813
814	/*
815	* See if we can start outputting
816	*/
817	if (if_queued && link_up)
818	if_start(pData);
819
820	STAM_REL_PROFILE_STOP(&pData->StatPoll, a);
821	}
822
823	#define ETH_ALEN 6
824	#define ETH_HLEN 14
825
826	#define ETH_P_IP 0x0800 /* Internet Protocol packet */
827	#define ETH_P_ARP 0x0806 /* Address Resolution packet */
828
829	#define ARPOP_REQUEST 1 /* ARP request */
830	#define ARPOP_REPLY 2 /* ARP reply */
831
832	struct ethhdr
833	{
834	unsigned char h_dest[ETH_ALEN]; /* destination eth addr */
835	unsigned char h_source[ETH_ALEN]; /* source ether addr */
836	unsigned short h_proto; /* packet type ID field */
837	};
838
839	struct arphdr
840	{
841	unsigned short ar_hrd; /* format of hardware address */
842	unsigned short ar_pro; /* format of protocol address */
843	unsigned char ar_hln; /* length of hardware address */
844	unsigned char ar_pln; /* length of protocol address */
845	unsigned short ar_op; /* ARP opcode (command) */
846
847	/*
848	* Ethernet looks like this : This bit is variable sized however...
849	*/
850	unsigned char ar_sha[ETH_ALEN]; /* sender hardware address */
851	unsigned char ar_sip[4]; /* sender IP address */
852	unsigned char ar_tha[ETH_ALEN]; /* target hardware address */
853	unsigned char ar_tip[4]; /* target IP address */
854	};
855
856	static
857	void arp_input(PNATState pData, const uint8_t *pkt, int pkt_len)
858	{
859	struct ethhdr eh = (struct ethhdr )pkt;
860	struct arphdr ah = (struct arphdr )(pkt + ETH_HLEN);
861	uint8_t arp_reply[ETH_HLEN + sizeof(struct arphdr)];
862	struct ethhdr reh = (struct ethhdr )arp_reply;
863	struct arphdr rah = (struct arphdr )(arp_reply + ETH_HLEN);
864	int ar_op;
865	struct ex_list *ex_ptr;
866	uint32_t htip = ntohl((uint32_t)ah->ar_tip);
867
868	ar_op = ntohs(ah->ar_op);
869	switch(ar_op) {
870	case ARPOP_REQUEST:
871	if ((htip & pData->netmask) == ntohl(special_addr.s_addr)) {
872	if ( (htip & ~pData->netmask) == CTL_DNS
873	\|\| (htip & ~pData->netmask) == CTL_ALIAS)
874	goto arp_ok;
875	for (ex_ptr = exec_list; ex_ptr; ex_ptr = ex_ptr->ex_next) {
876	if ((htip & ~pData->netmask) == ex_ptr->ex_addr)
877	goto arp_ok;
878	}
879	return;
880	arp_ok:
881	/* XXX: make an ARP request to have the client address */
882	memcpy(client_ethaddr, eh->h_source, ETH_ALEN);
883
884	/* ARP request for alias/dns mac address */
885	memcpy(reh->h_dest, pkt + ETH_ALEN, ETH_ALEN);
886	memcpy(reh->h_source, special_ethaddr, ETH_ALEN - 1);
887	reh->h_source[5] = ah->ar_tip[3];
888	reh->h_proto = htons(ETH_P_ARP);
889
890	rah->ar_hrd = htons(1);
891	rah->ar_pro = htons(ETH_P_IP);
892	rah->ar_hln = ETH_ALEN;
893	rah->ar_pln = 4;
894	rah->ar_op = htons(ARPOP_REPLY);
895	memcpy(rah->ar_sha, reh->h_source, ETH_ALEN);
896	memcpy(rah->ar_sip, ah->ar_tip, 4);
897	memcpy(rah->ar_tha, ah->ar_sha, ETH_ALEN);
898	memcpy(rah->ar_tip, ah->ar_sip, 4);
899	slirp_output(pData->pvUser, arp_reply, sizeof(arp_reply));
900	}
901	break;
902	default:
903	break;
904	}
905	}
906
907	void slirp_input(PNATState pData, const uint8_t *pkt, int pkt_len)
908	{
909	struct mbuf *m;
910	int proto;
911
912	if (pkt_len < ETH_HLEN)
913	return;
914
915	proto = ntohs((uint16_t )(pkt + 12));
916	switch(proto) {
917	case ETH_P_ARP:
918	arp_input(pData, pkt, pkt_len);
919	break;
920	case ETH_P_IP:
921	/* Update time. Important if the network is very quiet, as otherwise
922	* the first outgoing connection gets an incorrect timestamp. */
923	updtime(pData);
924
925	m = m_get(pData);
926	if (!m)
927	return;
928	/* Note: we add to align the IP header */
929	if (M_FREEROOM(m) < pkt_len + 2) {
930	m_inc(m, pkt_len + 2);
931	}
932	m->m_len = pkt_len + 2;
933	memcpy(m->m_data + 2, pkt, pkt_len);
934
935	m->m_data += 2 + ETH_HLEN;
936	m->m_len -= 2 + ETH_HLEN;
937
938	ip_input(pData, m);
939	break;
940	default:
941	break;
942	}
943	}
944
945	/* output the IP packet to the ethernet device */
946	void if_encap(PNATState pData, const uint8_t *ip_data, int ip_data_len)
947	{
948	uint8_t buf[1600];
949	struct ethhdr eh = (struct ethhdr )buf;
950
951	if (ip_data_len + ETH_HLEN > sizeof(buf))
952	return;
953
954	memcpy(eh->h_dest, client_ethaddr, ETH_ALEN);
955	memcpy(eh->h_source, special_ethaddr, ETH_ALEN - 1);
956	/* XXX: not correct */
957	eh->h_source[5] = CTL_ALIAS;
958	eh->h_proto = htons(ETH_P_IP);
959	memcpy(buf + sizeof(struct ethhdr), ip_data, ip_data_len);
960	slirp_output(pData->pvUser, buf, ip_data_len + ETH_HLEN);
961	}
962
963	int slirp_redir(PNATState pData, int is_udp, int host_port,
964	struct in_addr guest_addr, int guest_port)
965	{
966	if (is_udp) {
967	if (!udp_listen(pData, htons(host_port), guest_addr.s_addr,
968	htons(guest_port), 0))
969	return -1;
970	} else {
971	if (!solisten(pData, htons(host_port), guest_addr.s_addr,
972	htons(guest_port), 0))
973	return -1;
974	}
975	return 0;
976	}
977
978	int slirp_add_exec(PNATState pData, int do_pty, const char *args, int addr_low_byte,
979	int guest_port)
980	{
981	return add_exec(&exec_list, do_pty, (char *)args,
982	addr_low_byte, htons(guest_port));
983	}
984
985	void slirp_set_ethaddr(PNATState pData, const uint8_t *ethaddr)
986	{
987	memcpy(client_ethaddr, ethaddr, ETH_ALEN);
988	}
989
990	#if defined(VBOX_WITH_SIMPLEFIED_SLIRP_SYNC) && defined(RT_OS_WINDOWS)
991	HANDLE *slirp_get_events(PNATState pData)
992	{
993	return pData->phEvents;
994	}
995	void slirp_register_external_event(PNATState pData, HANDLE hEvent, int index)
996	{
997	pData->phEvents[index] = hEvent;
998	}
999	#endif

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

source: vbox/trunk/src/VBox/Devices/Network/slirp/slirp.c@ 14196

Download in other formats: