slirp.c@ 13951

Last change on this file since 13951 was 13951, checked in by vboxsync, 16 years ago
fixed win build for 38966
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	#include <VBox/err.h>
7	#include <iprt/assert.h>
8	#ifdef VBOX_WITH_SYNC_SLIRP
9	#include <iprt/semaphore.h>
10	#endif
11
12	static const uint8_t special_ethaddr[6] = {
13	0x52, 0x54, 0x00, 0x12, 0x35, 0x00
14	};
15
16	#ifdef _WIN32
17
18	static int get_dns_addr_domain(PNATState pData, bool fVerbose,
19	struct in_addr *pdns_addr,
20	const char **ppszDomain)
21	{
22	int rc = 0;
23	FIXED_INFO *FixedInfo=NULL;
24	ULONG BufLen;
25	DWORD ret;
26	IP_ADDR_STRING *pIPAddr;
27	struct in_addr tmp_addr;
28
29	FixedInfo = (FIXED_INFO *)GlobalAlloc(GPTR, sizeof(FIXED_INFO));
30	BufLen = sizeof(FIXED_INFO);
31
32	/** @todo: this API returns all DNS servers, no matter whether the
33	* corresponding network adapter is disabled or not. Maybe replace
34	* this by GetAdapterAddresses(), which is XP/Vista only though. */
35	if (ERROR_BUFFER_OVERFLOW == GetNetworkParams(FixedInfo, &BufLen)) {
36	if (FixedInfo) {
37	GlobalFree(FixedInfo);
38	FixedInfo = NULL;
39	}
40	FixedInfo = GlobalAlloc(GPTR, BufLen);
41	}
42
43	if ((ret = GetNetworkParams(FixedInfo, &BufLen)) != ERROR_SUCCESS) {
44	Log(("GetNetworkParams failed. ret = %08x\n", (u_int)ret ));
45	if (FixedInfo) {
46	GlobalFree(FixedInfo);
47	FixedInfo = NULL;
48	}
49	rc = -1;
50	goto get_dns_prefix;
51	}
52
53	pIPAddr = &(FixedInfo->DnsServerList);
54	inet_aton(pIPAddr->IpAddress.String, &tmp_addr);
55	Log(("nat: DNS Servers:\n"));
56	if (fVerbose \|\| pdns_addr->s_addr != tmp_addr.s_addr)
57	LogRel(("NAT: DNS address: %s\n", pIPAddr->IpAddress.String));
58	*pdns_addr = tmp_addr;
59
60	pIPAddr = FixedInfo -> DnsServerList.Next;
61	while ( pIPAddr )
62	{
63	if (fVerbose)
64	LogRel(("NAT: ignored DNS address: %s\n", pIPAddr ->IpAddress.String));
65	pIPAddr = pIPAddr ->Next;
66	}
67	if (FixedInfo) {
68	GlobalFree(FixedInfo);
69	FixedInfo = NULL;
70	}
71
72	get_dns_prefix:
73	if (ppszDomain)
74	{
75	OSVERSIONINFO ver;
76	char szDnsDomain[256];
77	DWORD dwSize = sizeof(szDnsDomain);
78
79	*ppszDomain = NULL;
80	GetVersionEx(&ver);
81	if (ver.dwMajorVersion >= 5)
82	{
83	/* GetComputerNameEx exists in Windows versions starting with 2000. */
84	if (GetComputerNameEx(ComputerNameDnsDomain, szDnsDomain, &dwSize))
85	{
86	if (szDnsDomain[0])
87	{
88	/* Just non-empty strings are valid. */
89	*ppszDomain = RTStrDup(szDnsDomain);
90	if (pData->fPassDomain)
91	{
92	if (fVerbose)
93	LogRel(("NAT: passing domain name %s\n", szDnsDomain));
94	}
95	else
96	Log(("nat: ignoring domain %s\n", szDnsDomain));
97	}
98	}
99	else
100	Log(("nat: GetComputerNameEx failed (%d)\n", GetLastError()));
101	}
102	}
103	return rc;
104	}
105
106	#else
107
108	static int get_dns_addr_domain(PNATState pData, bool fVerbose,
109	struct in_addr *pdns_addr,
110	const char **ppszDomain)
111	{
112	char buff[512];
113	char buff2[256];
114	FILE *f;
115	int found = 0;
116	struct in_addr tmp_addr;
117
118	#ifdef RT_OS_OS2
119	/* Try various locations. */
120	char *etc = getenv("ETC");
121	f = NULL;
122	if (etc)
123	{
124	snprintf(buff, sizeof(buff), "%s/RESOLV2", etc);
125	f = fopen(buff, "rt");
126	}
127	if (!f) {
128	snprintf(buff, sizeof(buff), "%s/RESOLV2", _PATH_ETC);
129	f = fopen(buff, "rt");
130	}
131	if (!f) {
132	snprintf(buff, sizeof(buff), "%s/resolv.conf", _PATH_ETC);
133	f = fopen(buff, "rt");
134	}
135	#else
136	f = fopen("/etc/resolv.conf", "r");
137	#endif
138	if (!f)
139	return -1;
140
141	if (ppszDomain)
142	*ppszDomain = NULL;
143	Log(("nat: DNS Servers:\n"));
144	while (fgets(buff, 512, f) != NULL) {
145	if (sscanf(buff, "nameserver%*[ \t]%256s", buff2) == 1) {
146	if (!inet_aton(buff2, &tmp_addr))
147	continue;
148	if (tmp_addr.s_addr == loopback_addr.s_addr)
149	tmp_addr = our_addr;
150	/* If it's the first one, set it to dns_addr */
151	if (!found)
152	{
153	if (fVerbose \|\| pdns_addr->s_addr != tmp_addr.s_addr)
154	LogRel(("NAT: DNS address: %s\n", buff2));
155	*pdns_addr = tmp_addr;
156	}
157	else
158	{
159	if (fVerbose)
160	LogRel(("NAT: ignored DNS address: %s\n", buff2));
161	}
162	found++;
163	}
164	if ( ppszDomain
165	&& (!strncmp(buff, "domain", 6) \|\| !strncmp(buff, "search", 6)))
166	{
167	/* Domain name/search list present. Pick first entry */
168	if (*ppszDomain == NULL)
169	{
170	char *tok;
171	char *saveptr;
172	tok = strtok_r(&buff[6], " \t\n", &saveptr);
173	if (tok)
174	{
175	*ppszDomain = RTStrDup(tok);
176	if (pData->fPassDomain)
177	{
178	if (fVerbose)
179	LogRel(("NAT: passing domain name %s\n", tok));
180	}
181	else
182	Log(("nat: ignoring domain %s\n", tok));
183	}
184	}
185	}
186	}
187	fclose(f);
188	if (!found)
189	return -1;
190	return 0;
191	}
192
193	#endif
194
195	int get_dns_addr(PNATState pData, struct in_addr *pdns_addr)
196	{
197	return get_dns_addr_domain(pData, false, pdns_addr, NULL);
198	}
199
200	int slirp_init(PNATState ppData, const char pszNetAddr, uint32_t u32Netmask,
201	bool fPassDomain, const char *pszTFTPPrefix,
202	const char pszBootFile, void pvUser)
203	{
204	int fNATfailed = 0;
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	#endif
228
229	VBOX_SLIRP_LOCK_CREATE(&pData->tcb_mutex);
230	VBOX_SLIRP_LOCK_CREATE(&pData->tcp_last_so_mutex);
231	VBOX_SLIRP_LOCK_CREATE(&pData->udb_mutex);
232	VBOX_SLIRP_LOCK_CREATE(&pData->udp_last_so_mutex);
233	VBOX_SLIRP_LOCK_CREATE(&pData->if_queued_mutex);
234	VBOX_SLIRP_LOCK_CREATE(&pData->next_m_mutex);
235
236	Assert(sizeof(struct ip) == 20);
237	link_up = 1;
238
239	if_init(pData);
240	ip_init(pData);
241
242	/* Initialise mbufs after setting the MTU */
243	m_init(pData);
244
245	/* set default addresses */
246	inet_aton("127.0.0.1", &loopback_addr);
247	inet_aton("127.0.0.1", &dns_addr);
248
249	if (get_dns_addr_domain(pData, true, &dns_addr, &pData->pszDomain) < 0)
250	fNATfailed = 1;
251
252	inet_aton(pszNetAddr, &special_addr);
253	alias_addr.s_addr = special_addr.s_addr \| htonl(CTL_ALIAS);
254	getouraddr(pData);
255	return fNATfailed ? VINF_NAT_DNS : VINF_SUCCESS;
256	}
257
258	/**
259	* Marks the link as up, making it possible to establish new connections.
260	*/
261	void slirp_link_up(PNATState pData)
262	{
263	link_up = 1;
264	}
265
266	/**
267	* Marks the link as down and cleans up the current connections.
268	*/
269	void slirp_link_down(PNATState pData)
270	{
271	struct socket *so;
272
273	while ((so = tcb.so_next) != &tcb)
274	{
275	if (so->so_state & SS_NOFDREF \|\| so->s == -1)
276	sofree(pData, so);
277	else
278	tcp_drop(pData, sototcpcb(so), 0);
279	}
280
281	while ((so = udb.so_next) != &udb)
282	udp_detach(pData, so);
283
284	link_up = 0;
285	}
286
287	/**
288	* Terminates the slirp component.
289	*/
290	void slirp_term(PNATState pData)
291	{
292	if (pData->pszDomain)
293	RTStrFree((char )(void )pData->pszDomain);
294
295	#if ARCH_BITS == 64
296	LogRel(("NAT: cpvHashUsed=%RU32 cpvHashCollisions=%RU32 cpvHashInserts=%RU64 cpvHashDone=%RU64\n",
297	pData->cpvHashUsed, pData->cpvHashCollisions, pData->cpvHashInserts, pData->cpvHashDone));
298	#endif
299
300	slirp_link_down(pData);
301	#ifdef WIN32
302	WSACleanup();
303	#endif
304	#ifdef LOG_ENABLED
305	Log(("\n"
306	"NAT statistics\n"
307	"--------------\n"
308	"\n"));
309	ipstats(pData);
310	tcpstats(pData);
311	udpstats(pData);
312	icmpstats(pData);
313	mbufstats(pData);
314	sockstats(pData);
315	Log(("\n"
316	"\n"
317	"\n"));
318	#endif
319	free(pData);
320	}
321
322
323	#define CONN_CANFSEND(so) (((so)->so_state & (SS_FCANTSENDMORE\|SS_ISFCONNECTED)) == SS_ISFCONNECTED)
324	#define CONN_CANFRCV(so) (((so)->so_state & (SS_FCANTRCVMORE\|SS_ISFCONNECTED)) == SS_ISFCONNECTED)
325	#define UPD_NFDS(x) if (nfds < (x)) nfds = (x)
326
327	/*
328	* curtime kept to an accuracy of 1ms
329	*/
330	#ifdef _WIN32
331	static void updtime(PNATState pData)
332	{
333	struct _timeb tb;
334
335	_ftime(&tb);
336	curtime = (u_int)tb.time * (u_int)1000;
337	curtime += (u_int)tb.millitm;
338	}
339	#else
340	static void updtime(PNATState pData)
341	{
342	gettimeofday(&tt, 0);
343
344	curtime = (u_int)tt.tv_sec * (u_int)1000;
345	curtime += (u_int)tt.tv_usec / (u_int)1000;
346
347	if ((tt.tv_usec % 1000) >= 500)
348	curtime++;
349	}
350	#endif
351
352	void slirp_select_fill(PNATState pData, int *pnfds,
353	fd_set readfds, fd_set writefds, fd_set *xfds)
354	{
355	struct socket so, so_next;
356	struct timeval timeout;
357	int nfds;
358	int tmp_time;
359
360	nfds = *pnfds;
361	/*
362	* First, TCP sockets
363	*/
364	#ifndef VBOX_WITH_SYNC_SLIRP
365	do_slowtimo = 0;
366	#endif
367	if (link_up) {
368	/*
369	* *_slowtimo needs calling if there are IP fragments
370	* in the fragment queue, or there are TCP connections active
371	*/
372	VBOX_SLIRP_LOCK(pData->tcb_mutex);
373	#ifndef VBOX_WITH_SYNC_SLIRP
374	do_slowtimo = ((tcb.so_next != &tcb) \|\|
375	((struct ipasfrag )&ipq != u32_to_ptr(pData, ipq.next, struct ipasfrag )));
376	#endif
377
378	so = tcb.so_next;
379	#ifndef VBOX_WITH_SYNC_SLIRP
380	for (so = tcb.so_next; so != &tcb; so = so_next) {
381	#else
382	while (1) {
383	tcp_loop_begin:
384	if (so == &tcb) {
385	VBOX_SLIRP_UNLOCK(pData->tcb_mutex);
386	break;
387	}
388	#endif
389	so_next = so->so_next;
390	VBOX_SLIRP_UNLOCK(pData->tcb_mutex);
391
392	/*
393	* See if we need a tcp_fasttimo
394	*/
395	#ifndef VBOX_SLIRP_UNLOCK
396	if (time_fasttimo == 0
397	&& so->so_tcpcb
398	&& so->so_tcpcb->t_flags & TF_DELACK)
399	time_fasttimo = curtime; /* Flag when we want a fasttimo */
400	#endif
401
402	/*
403	* NOFDREF can include still connecting to local-host,
404	* newly socreated() sockets etc. Don't want to select these.
405	*/
406	if (so->so_state & SS_NOFDREF \|\| so->s == -1)
407	goto before_loop_ends;
408
409	/*
410	* Set for reading sockets which are accepting
411	*/
412	if (so->so_state & SS_FACCEPTCONN) {
413	FD_SET(so->s, readfds);
414	UPD_NFDS(so->s);
415	goto before_loop_ends;
416	}
417	#ifndef VBOX_WITH_SYNC_SLIRP
418	/*
419	* Set for writing sockets which are connecting
420	*/
421	if (so->so_state & SS_ISFCONNECTING) {
422	FD_SET(so->s, writefds);
423	UPD_NFDS(so->s);
424	goto before_loop_ends;
425	}
426
427	/*
428	* Set for writing if we are connected, can send more, and
429	* we have something to send
430	*/
431	if (CONN_CANFSEND(so) && so->so_rcv.sb_cc) {
432	FD_SET(so->s, writefds);
433	UPD_NFDS(so->s);
434	}
435	#endif
436
437	/*
438	* Set for reading (and urgent data) if we are connected, can
439	* receive more, and we have room for it XXX /2 ?
440	*/
441	if (CONN_CANFRCV(so) && (so->so_snd.sb_cc < (so->so_snd.sb_datalen/2))) {
442	FD_SET(so->s, readfds);
443	FD_SET(so->s, xfds);
444	UPD_NFDS(so->s);
445	}
446	before_loop_ends:
447	/Release of global tcb mutex happens in the head of loop/
448	VBOX_SLIRP_LOCK(pData->tcb_mutex);
449	#ifdef VBOX_WITH_SYNC_SLIRP
450	so = so_next;
451	#endif
452	}
453
454	/*
455	* UDP sockets
456	*/
457	VBOX_SLIRP_LOCK(pData->udb_mutex);
458	so = udb.so_next;
459	#ifndef VBOX_WITH_SYNC_SLIRP
460	for (so = udb.so_next; so != &udb; so = so_next) {
461	#else
462	while(1) {
463	if (so == &udb) {
464	VBOX_SLIRP_UNLOCK(pData->udb_mutex);
465	break;
466	}
467	#endif
468	so_next = so->so_next;
469	VBOX_SLIRP_UNLOCK(pData->udb_mutex);
470
471	/*
472	* See if it's timed out
473	*/
474	#ifndef VBOX_WITH_SYNC_SLIRP
475	if (so->so_expire) {
476	if (so->so_expire <= curtime) {
477	udp_detach(pData, so);
478	goto before_udp_loop_end;
479	} else
480	do_slowtimo = 1; /* Let socket expire */
481	}
482	#endif
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	FD_SET(so->s, readfds);
496	UPD_NFDS(so->s);
497	}
498	before_udp_loop_end:
499	VBOX_SLIRP_LOCK(pData->udb_mutex);
500	#ifdef VBOX_WITH_SYNC_SLIRP
501	so = so_next;
502	#endif
503	}
504	}
505
506	/*
507	* Setup timeout to use minimum CPU usage, especially when idle
508	*/
509
510	#ifndef VBOX_WITH_SYNC_SLIRP
511	/*
512	* First, see the timeout needed by *timo
513	*/
514	timeout.tv_sec = 0;
515	timeout.tv_usec = -1;
516	/*
517	* If a slowtimo is needed, set timeout to 500ms from the last
518	* slow timeout. If a fast timeout is needed, set timeout within
519	* 200ms of when it was requested.
520	*/
521	if (do_slowtimo) {
522	/* XXX + 10000 because some select()'s aren't that accurate */
523	timeout.tv_usec = ((500 - (curtime - last_slowtimo)) * 1000) + 10000;
524	if (timeout.tv_usec < 0)
525	timeout.tv_usec = 0;
526	else if (timeout.tv_usec > 510000)
527	timeout.tv_usec = 510000;
528
529	/* Can only fasttimo if we also slowtimo */
530	if (time_fasttimo) {
531	tmp_time = (200 - (curtime - time_fasttimo)) * 1000;
532	if (tmp_time < 0)
533	tmp_time = 0;
534
535	/* Choose the smallest of the 2 */
536	if (tmp_time < timeout.tv_usec)
537	timeout.tv_usec = (u_int)tmp_time;
538	}
539	}
540	#endif
541	*pnfds = nfds;
542	}
543
544	void slirp_select_poll(PNATState pData, fd_set readfds, fd_set writefds, fd_set *xfds)
545	{
546	struct socket so, so_next;
547	int ret;
548
549	/* Update time */
550	updtime(pData);
551
552	/*
553	* See if anything has timed out
554	*/
555	#ifndef VBOX_WITH_SYNC_SLIRP
556	if (link_up) {
557	if (time_fasttimo && ((curtime - time_fasttimo) >= 2)) {
558	tcp_fasttimo(pData);
559	time_fasttimo = 0;
560	}
561	if (do_slowtimo && ((curtime - last_slowtimo) >= 499)) {
562	ip_slowtimo(pData);
563	tcp_slowtimo(pData);
564	last_slowtimo = curtime;
565	}
566	}
567	#endif
568
569	/*
570	* Check sockets
571	*/
572	if (link_up) {
573	/*
574	* Check TCP sockets
575	*/
576	VBOX_SLIRP_LOCK(pData->tcb_mutex);
577	so = tcb.so_next;
578	#ifndef VBOX_WITH_SYNC_SLIRP
579	for (so = tcb.so_next; so != &tcb; so = so_next) {
580	#else
581	while (1) {
582	loop_begin:
583	if (so == &tcb) {
584	VBOX_SLIRP_UNLOCK(pData->tcb_mutex);
585	break;
586	}
587	#endif
588	so_next = so->so_next;
589
590
591	VBOX_SLIRP_UNLOCK(pData->tcb_mutex);
592
593	/*
594	* FD_ISSET is meaningless on these sockets
595	* (and they can crash the program)
596	*/
597	if (so->so_state & SS_NOFDREF \|\| so->s == -1)
598	goto before_loop_ends;
599
600	/*
601	* Check for URG data
602	* This will soread as well, so no need to
603	* test for readfds below if this succeeds
604	*/
605	if (FD_ISSET(so->s, xfds))
606	sorecvoob(pData, so);
607	/*
608	* Check sockets for reading
609	*/
610	else if (FD_ISSET(so->s, readfds)) {
611	/*
612	* Check for incoming connections
613	*/
614	if (so->so_state & SS_FACCEPTCONN) {
615	tcp_connect(pData, so);
616	goto before_loop_ends;
617	} /* else */
618	ret = soread(pData, so);
619
620	/* Output it if we read something */
621	if (ret > 0)
622	tcp_output(pData, sototcpcb(so));
623	}
624
625	#ifndef VBOX_WITH_SYNC_SLIRP
626	/*
627	* Check sockets for writing
628	*/
629	if (FD_ISSET(so->s, writefds)) {
630	/*
631	* Check for non-blocking, still-connecting sockets
632	*/
633	if (so->so_state & SS_ISFCONNECTING) {
634	/* Connected */
635	so->so_state &= ~SS_ISFCONNECTING;
636
637	/*
638	* This should be probably guarded by PROBE_CONN too. Anyway,
639	* we disable it on OS/2 because the below send call returns
640	* EFAULT which causes the opened TCP socket to close right
641	* after it has been opened and connected.
642	*/
643	#ifndef RT_OS_OS2
644	ret = send(so->s, (const char *)&ret, 0, 0);
645	if (ret < 0) {
646	/* XXXXX Must fix, zero bytes is a NOP */
647	if (errno == EAGAIN \|\| errno == EWOULDBLOCK \|\|
648	errno == EINPROGRESS \|\| errno == ENOTCONN)
649	goto before_loop_ends;
650
651	/* else failed */
652	so->so_state = SS_NOFDREF;
653	}
654	/* else so->so_state &= ~SS_ISFCONNECTING; */
655	#endif
656
657	/*
658	* Continue tcp_input
659	*/
660	tcp_input(pData, (struct mbuf *)NULL, sizeof(struct ip), so);
661	/* continue; */
662	} else
663	ret = sowrite(pData, so);
664	/*
665	* XXXXX If we wrote something (a lot), there
666	* could be a need for a window update.
667	* In the worst case, the remote will send
668	* a window probe to get things going again
669	*/
670	}
671	#endif
672
673	/*
674	* Probe a still-connecting, non-blocking socket
675	* to check if it's still alive
676	*/
677	#ifdef PROBE_CONN
678	if (so->so_state & SS_ISFCONNECTING) {
679	ret = recv(so->s, (char *)&ret, 0,0);
680
681	if (ret < 0) {
682	/* XXX */
683	if (errno == EAGAIN \|\| errno == EWOULDBLOCK \|\|
684	errno == EINPROGRESS \|\| errno == ENOTCONN)
685	goto before_loop_ends;/* Still connecting, continue */
686
687	/* else failed */
688	so->so_state = SS_NOFDREF;
689
690	/* tcp_input will take care of it */
691	} else {
692	ret = send(so->s, &ret, 0,0);
693	if (ret < 0) {
694	/* XXX */
695	if (errno == EAGAIN \|\| errno == EWOULDBLOCK \|\|
696	errno == EINPROGRESS \|\| errno == ENOTCONN)
697	goto before_loop_ends;
698	/* else failed */
699	so->so_state = SS_NOFDREF;
700	} else
701	so->so_state &= ~SS_ISFCONNECTING;
702
703	}
704	tcp_input((struct mbuf *)NULL, sizeof(struct ip),so);
705	} /* SS_ISFCONNECTING */
706	#endif
707	before_loop_ends:
708	VBOX_SLIRP_LOCK(pData->tcb_mutex);
709	#ifdef VBOX_WITH_SYNC_SLIRP
710	so = so_next;
711	#endif
712	}
713
714	/*
715	* Now UDP sockets.
716	* Incoming packets are sent straight away, they're not buffered.
717	* Incoming UDP data isn't buffered either.
718	*/
719	VBOX_SLIRP_LOCK(pData->udb_mutex);
720	so = udb.so_next;
721	#ifndef VBOX_WITH_SYNC_SLIRP
722	for (so = udb.so_next; so != &udb; so = so_next) {
723	#else
724	while(1) {
725	if (so == &udb) {
726	VBOX_SLIRP_UNLOCK(pData->udb_mutex);
727	break;
728	}
729	#endif
730	so_next = so->so_next;
731	VBOX_SLIRP_UNLOCK(pData->udb_mutex);
732
733	if (so->s != -1 && FD_ISSET(so->s, readfds)) {
734	sorecvfrom(pData, so);
735	}
736	VBOX_SLIRP_LOCK(pData->udb_mutex);
737	#ifdef VBOX_WITH_SYNC_SLIRP
738	so = so_next;
739	#endif
740	}
741	}
742
743	/*
744	* See if we can start outputting
745	*/
746	#ifndef VBOX_WITH_SYNC_SLIRP
747	if (if_queued && link_up)
748	if_start(pData);
749	#else
750	#if 0
751	if (link_up) {
752	VBOX_SLIRP_LOCK(pData->if_queued_mutex);
753	if (if_queued > 0){
754	VBOX_SLIRP_UNLOCK(pData->if_queued_mutex);
755	if_start(pData);
756	}
757	else {
758	VBOX_SLIRP_UNLOCK(pData->if_queued_mutex);
759	}
760	}
761	#endif
762	#endif
763	}
764
765	#define ETH_ALEN 6
766	#define ETH_HLEN 14
767
768	#define ETH_P_IP 0x0800 /* Internet Protocol packet */
769	#define ETH_P_ARP 0x0806 /* Address Resolution packet */
770
771	#define ARPOP_REQUEST 1 /* ARP request */
772	#define ARPOP_REPLY 2 /* ARP reply */
773
774	struct ethhdr
775	{
776	unsigned char h_dest[ETH_ALEN]; /* destination eth addr */
777	unsigned char h_source[ETH_ALEN]; /* source ether addr */
778	unsigned short h_proto; /* packet type ID field */
779	};
780
781	struct arphdr
782	{
783	unsigned short ar_hrd; /* format of hardware address */
784	unsigned short ar_pro; /* format of protocol address */
785	unsigned char ar_hln; /* length of hardware address */
786	unsigned char ar_pln; /* length of protocol address */
787	unsigned short ar_op; /* ARP opcode (command) */
788
789	/*
790	* Ethernet looks like this : This bit is variable sized however...
791	*/
792	unsigned char ar_sha[ETH_ALEN]; /* sender hardware address */
793	unsigned char ar_sip[4]; /* sender IP address */
794	unsigned char ar_tha[ETH_ALEN]; /* target hardware address */
795	unsigned char ar_tip[4]; /* target IP address */
796	};
797
798	static
799	void arp_input(PNATState pData, const uint8_t *pkt, int pkt_len)
800	{
801	struct ethhdr eh = (struct ethhdr )pkt;
802	struct arphdr ah = (struct arphdr )(pkt + ETH_HLEN);
803	uint8_t arp_reply[ETH_HLEN + sizeof(struct arphdr)];
804	struct ethhdr reh = (struct ethhdr )arp_reply;
805	struct arphdr rah = (struct arphdr )(arp_reply + ETH_HLEN);
806	int ar_op;
807	struct ex_list *ex_ptr;
808	uint32_t htip = ntohl((uint32_t)ah->ar_tip);
809
810	ar_op = ntohs(ah->ar_op);
811	switch(ar_op) {
812	case ARPOP_REQUEST:
813	if ((htip & pData->netmask) == ntohl(special_addr.s_addr)) {
814	if ( (htip & ~pData->netmask) == CTL_DNS
815	\|\| (htip & ~pData->netmask) == CTL_ALIAS)
816	goto arp_ok;
817	for (ex_ptr = exec_list; ex_ptr; ex_ptr = ex_ptr->ex_next) {
818	if ((htip & ~pData->netmask) == ex_ptr->ex_addr)
819	goto arp_ok;
820	}
821	return;
822	arp_ok:
823	/* XXX: make an ARP request to have the client address */
824	memcpy(client_ethaddr, eh->h_source, ETH_ALEN);
825
826	/* ARP request for alias/dns mac address */
827	memcpy(reh->h_dest, pkt + ETH_ALEN, ETH_ALEN);
828	memcpy(reh->h_source, special_ethaddr, ETH_ALEN - 1);
829	reh->h_source[5] = ah->ar_tip[3];
830	reh->h_proto = htons(ETH_P_ARP);
831
832	rah->ar_hrd = htons(1);
833	rah->ar_pro = htons(ETH_P_IP);
834	rah->ar_hln = ETH_ALEN;
835	rah->ar_pln = 4;
836	rah->ar_op = htons(ARPOP_REPLY);
837	memcpy(rah->ar_sha, reh->h_source, ETH_ALEN);
838	memcpy(rah->ar_sip, ah->ar_tip, 4);
839	memcpy(rah->ar_tha, ah->ar_sha, ETH_ALEN);
840	memcpy(rah->ar_tip, ah->ar_sip, 4);
841	slirp_output(pData->pvUser, arp_reply, sizeof(arp_reply));
842	}
843	break;
844	default:
845	break;
846	}
847	}
848
849	void slirp_input(PNATState pData, const uint8_t *pkt, int pkt_len)
850	{
851	struct mbuf *m;
852	int proto;
853
854	if (pkt_len < ETH_HLEN)
855	return;
856
857	proto = ntohs((uint16_t )(pkt + 12));
858	switch(proto) {
859	case ETH_P_ARP:
860	arp_input(pData, pkt, pkt_len);
861	break;
862	case ETH_P_IP:
863	/* Update time. Important if the network is very quiet, as otherwise
864	* the first outgoing connection gets an incorrect timestamp. */
865	updtime(pData);
866
867	m = m_get(pData);
868	if (!m)
869	return;
870	/* Note: we add to align the IP header */
871	if (M_FREEROOM(m) < pkt_len + 2) {
872	m_inc(m, pkt_len + 2);
873	}
874	m->m_len = pkt_len + 2;
875	memcpy(m->m_data + 2, pkt, pkt_len);
876
877	m->m_data += 2 + ETH_HLEN;
878	m->m_len -= 2 + ETH_HLEN;
879
880	ip_input(pData, m);
881	break;
882	default:
883	break;
884	}
885	}
886
887	/* output the IP packet to the ethernet device */
888	void if_encap(PNATState pData, const uint8_t *ip_data, int ip_data_len)
889	{
890	uint8_t buf[1600];
891	struct ethhdr eh = (struct ethhdr )buf;
892
893	if (ip_data_len + ETH_HLEN > sizeof(buf))
894	return;
895
896	memcpy(eh->h_dest, client_ethaddr, ETH_ALEN);
897	memcpy(eh->h_source, special_ethaddr, ETH_ALEN - 1);
898	/* XXX: not correct */
899	eh->h_source[5] = CTL_ALIAS;
900	eh->h_proto = htons(ETH_P_IP);
901	memcpy(buf + sizeof(struct ethhdr), ip_data, ip_data_len);
902	slirp_output(pData->pvUser, buf, ip_data_len + ETH_HLEN);
903	}
904
905	int slirp_redir(PNATState pData, int is_udp, int host_port,
906	struct in_addr guest_addr, int guest_port)
907	{
908	if (is_udp) {
909	if (!udp_listen(pData, htons(host_port), guest_addr.s_addr,
910	htons(guest_port), 0))
911	return -1;
912	} else {
913	if (!solisten(pData, htons(host_port), guest_addr.s_addr,
914	htons(guest_port), 0))
915	return -1;
916	}
917	return 0;
918	}
919
920	int slirp_add_exec(PNATState pData, int do_pty, const char *args, int addr_low_byte,
921	int guest_port)
922	{
923	return add_exec(&exec_list, do_pty, (char *)args,
924	addr_low_byte, htons(guest_port));
925	}
926
927	void slirp_set_ethaddr(PNATState pData, const uint8_t *ethaddr)
928	{
929	memcpy(client_ethaddr, ethaddr, ETH_ALEN);
930	}
931
932	#ifdef VBOX_WITH_SYNC_SLIRP
933	void slirp_fasttmr(PNATState pData)
934	{
935	struct socket so, so_next;
936	updtime(pData);
937	time_fasttimo = 0;
938	#if 1
939	VBOX_SLIRP_LOCK(pData->tcb_mutex);
940	so = tcb.so_next;
941	while(1) {
942	if (so == &tcb) {
943	VBOX_SLIRP_UNLOCK(pData->tcb_mutex);
944	break;
945	}
946	so_next = so->so_next;
947	VBOX_SLIRP_UNLOCK(pData->tcb_mutex);
948	if (time_fasttimo == 0
949	&& so->so_tcpcb
950	&& so->so_tcpcb->t_flags & TF_DELACK)
951	time_fasttimo = curtime; /* Flag when we want a fasttimo */
952	VBOX_SLIRP_LOCK(pData->tcb_mutex);
953	so = so_next;
954	}
955	#endif
956	if (time_fasttimo) {
957	tcp_fasttimo(pData);
958	time_fasttimo = 0;
959	}
960	}
961
962	void slirp_slowtmr(PNATState pData)
963	{
964	struct socket so, so_next;
965	updtime(pData);
966	do_slowtimo = 0;
967	#if 1
968	VBOX_SLIRP_LOCK(pData->tcb_mutex);
969	do_slowtimo = ((tcb.so_next != &tcb) \|\|
970	((struct ipasfrag )&ipq != u32_to_ptr(pData, ipq.next, struct ipasfrag )));
971
972	VBOX_SLIRP_UNLOCK(pData->tcb_mutex);
973
974	VBOX_SLIRP_LOCK(pData->udb_mutex);
975	so = udb.so_next;
976	while(1) {
977	if (so == &udb) {
978	VBOX_SLIRP_UNLOCK(pData->udb_mutex);
979	break;
980	}
981	so_next = so->so_next;
982	VBOX_SLIRP_UNLOCK(pData->udb_mutex);
983
984	if (so->so_expire) {
985	if (so->so_expire <= curtime) {
986	udp_detach(pData, so);
987	goto before_loop_ends;
988	}
989	do_slowtimo = 1;
990	}
991	before_loop_ends:
992	VBOX_SLIRP_LOCK(pData->udb_mutex);
993	so = so_next;
994	}
995	#endif
996	if (do_slowtimo) {
997	tcp_slowtimo(pData);
998	ip_slowtimo(pData);
999	last_slowtimo = curtime;
1000	}
1001	}
1002
1003	/selects open and ready sockets for write/
1004	void slirp_send_fill(PNATState pData, int pnfds, fd_set writefds)
1005	{
1006	struct socket so, so_next;
1007	int nfds = *pnfds;
1008
1009	if (link_up == 0) return;
1010
1011	VBOX_SLIRP_LOCK(pData->tcb_mutex);
1012	so = tcb.so_next;
1013	while (1) {
1014	if (so == &tcb) {
1015	VBOX_SLIRP_UNLOCK(pData->tcb_mutex);
1016	break;
1017	}
1018	so_next = so->so_next;
1019	VBOX_SLIRP_UNLOCK(pData->tcb_mutex);
1020
1021	if (so->so_state & SS_NOFDREF \|\| so->s == -1)
1022	goto before_loop_ends;
1023
1024	if (so->so_state & SS_ISFCONNECTING) {
1025	FD_SET(so->s, writefds);
1026	UPD_NFDS(so->s);
1027	goto before_loop_ends;
1028	}
1029
1030	if (CONN_CANFSEND(so) && so->so_rcv.sb_cc) {
1031	FD_SET(so->s, writefds);
1032	UPD_NFDS(so->s);
1033	}
1034
1035	before_loop_ends:
1036	VBOX_SLIRP_LOCK(pData->tcb_mutex);
1037	so = so_next;
1038	}
1039	*pnfds = nfds;
1040	}
1041	/triggers socket output /
1042	void slirp_send_trigger(PNATState pData, int pnfds, fd_set writefds)
1043	{
1044	struct socket so, so_next;
1045	int nfds = *pnfds;
1046	int ret;
1047
1048	if (link_up == 0) return;
1049
1050	VBOX_SLIRP_LOCK(pData->tcb_mutex);
1051	so = tcb.so_next;
1052	while (1) {
1053	if (so == &tcb) {
1054	VBOX_SLIRP_UNLOCK(pData->tcb_mutex);
1055	break;
1056	}
1057	so_next = so->so_next;
1058	VBOX_SLIRP_UNLOCK(pData->tcb_mutex);
1059	/*
1060	* Check sockets for writing
1061	*/
1062	if (FD_ISSET(so->s, writefds)) {
1063	/*
1064	* Check for non-blocking, still-connecting sockets
1065	*/
1066	if (so->so_state & SS_ISFCONNECTING) {
1067	/* Connected */
1068	so->so_state &= ~SS_ISFCONNECTING;
1069
1070	/*
1071	* This should be probably guarded by PROBE_CONN too. Anyway,
1072	* we disable it on OS/2 because the below send call returns
1073	* EFAULT which causes the opened TCP socket to close right
1074	* after it has been opened and connected.
1075	*/
1076	#ifndef RT_OS_OS2
1077	ret = send(so->s, (const char *)&ret, 0, 0);
1078	if (ret < 0) {
1079	/* XXXXX Must fix, zero bytes is a NOP */
1080	if (errno == EAGAIN \|\| errno == EWOULDBLOCK \|\|
1081	errno == EINPROGRESS \|\| errno == ENOTCONN)
1082	goto before_loop_ends;
1083
1084	/* else failed */
1085	so->so_state = SS_NOFDREF;
1086	}
1087	/* else so->so_state &= ~SS_ISFCONNECTING; */
1088	#endif
1089
1090	/*
1091	* Continue tcp_input
1092	*/
1093	tcp_input(pData, (struct mbuf *)NULL, sizeof(struct ip), so);
1094	/* continue; */
1095	} else
1096	ret = sowrite(pData, so);
1097	/*
1098	* XXXXX If we wrote something (a lot), there
1099	* could be a need for a window update.
1100	* In the worst case, the remote will send
1101	* a window probe to get things going again
1102	*/
1103	}
1104
1105	before_loop_ends:
1106	VBOX_SLIRP_LOCK(pData->tcb_mutex);
1107	so = so_next;
1108	}
1109	}
1110	#endif

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

source: vbox/trunk/src/VBox/Devices/Network/slirp/slirp.c@ 13951

Download in other formats: