slirp.c@ 15089

Last change on this file since 15089 was 15089, checked in by vboxsync, 16 years ago
slirp:icmp: Win ICMP api can't be handled in the WSA socket way so it should be handled separately.
Property svn:eol-style set to `native`
File size: 33.4 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	#if defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC)
14	# if !defined(RT_OS_WINDOWS)
15	# define DO_ENGAGE_EVENT1(so, fdset, label) \
16	do { \
17	FD_SET((so)->s, (fdset)); \
18	UPD_NFDS((so)->s); \
19	} while(0)
20
21
22	# define DO_ENGAGE_EVENT2(so, fdset0, fdset1, label) \
23	do { \
24	FD_SET((so)->s, (fdset0)); \
25	FD_SET((so)->s, (fdset1)); \
26	UPD_NFDS((so)->s); \
27	} while(0)
28	# else /* !RT_OS_WINDOWS */
29	# define DO_ENGAGE_EVENT1(so, fdset0, label) \
30	do { \
31	rc = WSAEventSelect((so)->s, VBOX_SOCKET_EVENT, FD_ALL_EVENTS); \
32	if (rc == SOCKET_ERROR) \
33	{ \
34	/* This should not happen */ \
35	error = WSAGetLastError(); \
36	LogRel(("WSAEventSelector (" #label ") error %d (so=%x, socket=%s, event=%x)\n", \
37	error, (so), (so)->s, VBOX_SOCKET_EVENT)); \
38	} \
39	} while(0); \
40
41	# define DO_ENGAGE_EVENT2(so, fdset0, fdset1, label) DO_ENGAGE_EVENT1((so), (fdset0), label)
42	# endif /* RT_OS_WINDOWS */
43
44	# define TCP_ENGAGE_EVENT1(so, fdset0) \
45	DO_ENGAGE_EVENT1((so), (fdset0), TCP)
46
47	# define TCP_ENGAGE_EVENT2(so, fdset0, fdset1) \
48	DO_ENGAGE_EVENT2((so), (fdset0), (fdset1), TCP)
49
50	#define UDP_ENGAGE_EVENT(so, fdset) DO_ENGAGE_EVENT1((so), (fdset), UDP)
51	#endif /* VBOX_WITH_SIMPLIFIED_SLIRP_SYNC */
52
53	static const uint8_t special_ethaddr[6] = {
54	0x52, 0x54, 0x00, 0x12, 0x35, 0x00
55	};
56
57	#ifdef RT_OS_WINDOWS
58
59	static int get_dns_addr_domain(PNATState pData, bool fVerbose,
60	struct in_addr *pdns_addr,
61	const char **ppszDomain)
62	{
63	int rc = 0;
64	FIXED_INFO *FixedInfo = NULL;
65	ULONG BufLen;
66	DWORD ret;
67	IP_ADDR_STRING *pIPAddr;
68	struct in_addr tmp_addr;
69
70	FixedInfo = (FIXED_INFO *)GlobalAlloc(GPTR, sizeof(FIXED_INFO));
71	BufLen = sizeof(FIXED_INFO);
72
73	/** @todo: this API returns all DNS servers, no matter whether the
74	* corresponding network adapter is disabled or not. Maybe replace
75	* this by GetAdapterAddresses(), which is XP/Vista only though. */
76	if (ERROR_BUFFER_OVERFLOW == GetNetworkParams(FixedInfo, &BufLen))
77	{
78	if (FixedInfo)
79	{
80	GlobalFree(FixedInfo);
81	FixedInfo = NULL;
82	}
83	FixedInfo = GlobalAlloc(GPTR, BufLen);
84	}
85
86	if ((ret = GetNetworkParams(FixedInfo, &BufLen)) != ERROR_SUCCESS)
87	{
88	Log(("GetNetworkParams failed. ret = %08x\n", (u_int)ret ));
89	if (FixedInfo)
90	{
91	GlobalFree(FixedInfo);
92	FixedInfo = NULL;
93	}
94	rc = -1;
95	goto get_dns_prefix;
96	}
97
98	pIPAddr = &(FixedInfo->DnsServerList);
99	inet_aton(pIPAddr->IpAddress.String, &tmp_addr);
100	Log(("nat: DNS Servers:\n"));
101	if (fVerbose \|\| pdns_addr->s_addr != tmp_addr.s_addr)
102	LogRel(("NAT: DNS address: %s\n", pIPAddr->IpAddress.String));
103	*pdns_addr = tmp_addr;
104
105	pIPAddr = FixedInfo -> DnsServerList.Next;
106	while (pIPAddr)
107	{
108	if (fVerbose)
109	LogRel(("NAT: ignored DNS address: %s\n", pIPAddr ->IpAddress.String));
110	pIPAddr = pIPAddr ->Next;
111	}
112	if (FixedInfo)
113	{
114	GlobalFree(FixedInfo);
115	FixedInfo = NULL;
116	}
117
118	get_dns_prefix:
119	if (ppszDomain)
120	{
121	OSVERSIONINFO ver;
122	char szDnsDomain[256];
123	DWORD dwSize = sizeof(szDnsDomain);
124
125	*ppszDomain = NULL;
126	GetVersionEx(&ver);
127	if (ver.dwMajorVersion >= 5)
128	{
129	/* GetComputerNameEx exists in Windows versions starting with 2000. */
130	if (GetComputerNameEx(ComputerNameDnsDomain, szDnsDomain, &dwSize))
131	{
132	if (szDnsDomain[0])
133	{
134	/* Just non-empty strings are valid. */
135	*ppszDomain = RTStrDup(szDnsDomain);
136	if (pData->fPassDomain)
137	{
138	if (fVerbose)
139	LogRel(("NAT: passing domain name %s\n", szDnsDomain));
140	}
141	else
142	Log(("nat: ignoring domain %s\n", szDnsDomain));
143	}
144	}
145	else
146	Log(("nat: GetComputerNameEx failed (%d)\n", GetLastError()));
147	}
148	}
149	return rc;
150	}
151
152	#else
153
154	static int get_dns_addr_domain(PNATState pData, bool fVerbose,
155	struct in_addr *pdns_addr,
156	const char **ppszDomain)
157	{
158	char buff[512];
159	char buff2[256];
160	FILE *f;
161	int found = 0;
162	struct in_addr tmp_addr;
163
164	#ifdef RT_OS_OS2
165	/* Try various locations. */
166	char *etc = getenv("ETC");
167	f = NULL;
168	if (etc)
169	{
170	snprintf(buff, sizeof(buff), "%s/RESOLV2", etc);
171	f = fopen(buff, "rt");
172	}
173	if (!f)
174	{
175	snprintf(buff, sizeof(buff), "%s/RESOLV2", _PATH_ETC);
176	f = fopen(buff, "rt");
177	}
178	if (!f)
179	{
180	snprintf(buff, sizeof(buff), "%s/resolv.conf", _PATH_ETC);
181	f = fopen(buff, "rt");
182	}
183	#else
184	f = fopen("/etc/resolv.conf", "r");
185	#endif
186	if (!f)
187	return -1;
188
189	if (ppszDomain)
190	*ppszDomain = NULL;
191	Log(("nat: DNS Servers:\n"));
192	while (fgets(buff, 512, f) != NULL)
193	{
194	if (sscanf(buff, "nameserver%*[ \t]%256s", buff2) == 1)
195	{
196	if (!inet_aton(buff2, &tmp_addr))
197	continue;
198	if (tmp_addr.s_addr == loopback_addr.s_addr)
199	tmp_addr = our_addr;
200	/* If it's the first one, set it to dns_addr */
201	if (!found)
202	{
203	if (fVerbose \|\| pdns_addr->s_addr != tmp_addr.s_addr)
204	LogRel(("NAT: DNS address: %s\n", buff2));
205	*pdns_addr = tmp_addr;
206	}
207	else
208	{
209	if (fVerbose)
210	LogRel(("NAT: ignored DNS address: %s\n", buff2));
211	}
212	found++;
213	}
214	if ( ppszDomain
215	&& (!strncmp(buff, "domain", 6) \|\| !strncmp(buff, "search", 6)))
216	{
217	/* Domain name/search list present. Pick first entry */
218	if (*ppszDomain == NULL)
219	{
220	char *tok;
221	char *saveptr;
222	tok = strtok_r(&buff[6], " \t\n", &saveptr);
223	if (tok)
224	{
225	*ppszDomain = RTStrDup(tok);
226	if (pData->fPassDomain)
227	{
228	if (fVerbose)
229	LogRel(("NAT: passing domain name %s\n", tok));
230	}
231	else
232	Log(("nat: ignoring domain %s\n", tok));
233	}
234	}
235	}
236	}
237	fclose(f);
238	if (!found)
239	return -1;
240	return 0;
241	}
242
243	#endif
244
245	int get_dns_addr(PNATState pData, struct in_addr *pdns_addr)
246	{
247	return get_dns_addr_domain(pData, false, pdns_addr, NULL);
248	}
249
250	int slirp_init(PNATState ppData, const char pszNetAddr, uint32_t u32Netmask,
251	bool fPassDomain, const char *pszTFTPPrefix,
252	const char pszBootFile, void pvUser)
253	{
254	int fNATfailed = 0;
255	PNATState pData = malloc(sizeof(NATState));
256	*ppData = pData;
257	if (!pData)
258	return VERR_NO_MEMORY;
259	if (u32Netmask & 0x1f)
260	/* CTL is x.x.x.15, bootp passes up to 16 IPs (15..31) */
261	return VERR_INVALID_PARAMETER;
262	memset(pData, '\0', sizeof(NATState));
263	pData->fPassDomain = fPassDomain;
264	pData->pvUser = pvUser;
265	#if ARCH_BITS == 64 && !defined(VBOX_WITH_BSD_REASS)
266	pData->cpvHashUsed = 1;
267	#endif
268	tftp_prefix = pszTFTPPrefix;
269	bootp_filename = pszBootFile;
270	pData->netmask = u32Netmask;
271
272	#ifdef RT_OS_WINDOWS
273	{
274	WSADATA Data;
275	WSAStartup(MAKEWORD(2,0), &Data);
276	}
277	# if defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC)
278	pData->phEvents[VBOX_SOCKET_EVENT_INDEX] = CreateEvent(NULL, FALSE, FALSE, NULL);
279	# endif
280	#endif
281
282	Assert(sizeof(struct ip) == 20);
283	link_up = 1;
284
285	debug_init();
286	if_init(pData);
287	ip_init(pData);
288	#ifdef VBOX_WITH_SLIRP_ICMP
289	icmp_init(pData);
290	#endif /* VBOX_WITH_SLIRP_ICMP */
291
292	/* Initialise mbufs after setting the MTU */
293	m_init(pData);
294
295	/* set default addresses */
296	inet_aton("127.0.0.1", &loopback_addr);
297	inet_aton("127.0.0.1", &dns_addr);
298
299	if (get_dns_addr_domain(pData, true, &dns_addr, &pData->pszDomain) < 0)
300	fNATfailed = 1;
301
302	inet_aton(pszNetAddr, &special_addr);
303	alias_addr.s_addr = special_addr.s_addr \| htonl(CTL_ALIAS);
304	getouraddr(pData);
305	return fNATfailed ? VINF_NAT_DNS : VINF_SUCCESS;
306	}
307
308	/**
309	* Statistics counters.
310	*/
311	void slirp_register_timers(PNATState pData, PPDMDRVINS pDrvIns)
312	{
313	#ifndef VBOX_WITHOUT_RELEASE_STATISTICS
314	PDMDrvHlpSTAMRegisterF(pDrvIns, &pData->StatFill, STAMTYPE_PROFILE, STAMVISIBILITY_ALWAYS,
315	STAMUNIT_TICKS_PER_CALL, "Profiling slirp fills", "/Drivers/NAT%d/Fill", pDrvIns->iInstance);
316	PDMDrvHlpSTAMRegisterF(pDrvIns, &pData->StatPoll, STAMTYPE_PROFILE, STAMVISIBILITY_ALWAYS,
317	STAMUNIT_TICKS_PER_CALL, "Profiling slirp polls", "/Drivers/NAT%d/Poll", pDrvIns->iInstance);
318	PDMDrvHlpSTAMRegisterF(pDrvIns, &pData->StatFastTimer, STAMTYPE_PROFILE, STAMVISIBILITY_ALWAYS,
319	STAMUNIT_TICKS_PER_CALL, "Profiling slirp fast timer", "/Drivers/NAT%d/TimerFast", pDrvIns->iInstance);
320	PDMDrvHlpSTAMRegisterF(pDrvIns, &pData->StatSlowTimer, STAMTYPE_PROFILE, STAMVISIBILITY_ALWAYS,
321	STAMUNIT_TICKS_PER_CALL, "Profiling slirp slow timer", "/Drivers/NAT%d/TimerSlow", pDrvIns->iInstance);
322	PDMDrvHlpSTAMRegisterF(pDrvIns, &pData->StatTCP, STAMTYPE_COUNTER, STAMVISIBILITY_ALWAYS,
323	STAMUNIT_COUNT, "TCP sockets", "/Drivers/NAT%d/SockTCP", pDrvIns->iInstance);
324	PDMDrvHlpSTAMRegisterF(pDrvIns, &pData->StatTCPHot, STAMTYPE_COUNTER, STAMVISIBILITY_ALWAYS,
325	STAMUNIT_COUNT, "TCP sockets active", "/Drivers/NAT%d/SockTCPHot", pDrvIns->iInstance);
326	PDMDrvHlpSTAMRegisterF(pDrvIns, &pData->StatUDP, STAMTYPE_COUNTER, STAMVISIBILITY_ALWAYS,
327	STAMUNIT_COUNT, "UDP sockets", "/Drivers/NAT%d/SockUDP", pDrvIns->iInstance);
328	PDMDrvHlpSTAMRegisterF(pDrvIns, &pData->StatUDPHot, STAMTYPE_COUNTER, STAMVISIBILITY_ALWAYS,
329	STAMUNIT_COUNT, "UDP sockets active", "/Drivers/NAT%d/SockUDPHot", pDrvIns->iInstance);
330	#endif /* VBOX_WITHOUT_RELEASE_STATISTICS */
331	}
332
333	/**
334	* Marks the link as up, making it possible to establish new connections.
335	*/
336	void slirp_link_up(PNATState pData)
337	{
338	link_up = 1;
339	}
340
341	/**
342	* Marks the link as down and cleans up the current connections.
343	*/
344	void slirp_link_down(PNATState pData)
345	{
346	struct socket *so;
347
348	while ((so = tcb.so_next) != &tcb)
349	{
350	if (so->so_state & SS_NOFDREF \|\| so->s == -1)
351	sofree(pData, so);
352	else
353	tcp_drop(pData, sototcpcb(so), 0);
354	}
355
356	while ((so = udb.so_next) != &udb)
357	udp_detach(pData, so);
358
359	link_up = 0;
360	}
361
362	/**
363	* Terminates the slirp component.
364	*/
365	void slirp_term(PNATState pData)
366	{
367	if (pData->pszDomain)
368	RTStrFree((char )(void )pData->pszDomain);
369
370	#if ARCH_BITS == 64 && !defined(VBOX_WITH_BSD_REASS)
371	LogRel(("NAT: cpvHashUsed=%RU32 cpvHashCollisions=%RU32 cpvHashInserts=%RU64 cpvHashDone=%RU64\n",
372	pData->cpvHashUsed, pData->cpvHashCollisions, pData->cpvHashInserts, pData->cpvHashDone));
373	#endif
374
375	slirp_link_down(pData);
376	#ifdef RT_OS_WINDOWS
377	WSACleanup();
378	#endif
379	#ifdef LOG_ENABLED
380	Log(("\n"
381	"NAT statistics\n"
382	"--------------\n"
383	"\n"));
384	ipstats(pData);
385	tcpstats(pData);
386	udpstats(pData);
387	icmpstats(pData);
388	mbufstats(pData);
389	sockstats(pData);
390	Log(("\n"
391	"\n"
392	"\n"));
393	#endif
394	free(pData);
395	}
396
397
398	#define CONN_CANFSEND(so) (((so)->so_state & (SS_FCANTSENDMORE\|SS_ISFCONNECTED)) == SS_ISFCONNECTED)
399	#define CONN_CANFRCV(so) (((so)->so_state & (SS_FCANTRCVMORE\|SS_ISFCONNECTED)) == SS_ISFCONNECTED)
400	#define UPD_NFDS(x) if (nfds < (x)) nfds = (x)
401
402	/*
403	* curtime kept to an accuracy of 1ms
404	*/
405	#ifdef RT_OS_WINDOWS
406	static void updtime(PNATState pData)
407	{
408	struct _timeb tb;
409
410	_ftime(&tb);
411	curtime = (u_int)tb.time * (u_int)1000;
412	curtime += (u_int)tb.millitm;
413	}
414	#else
415	static void updtime(PNATState pData)
416	{
417	gettimeofday(&tt, 0);
418
419	curtime = (u_int)tt.tv_sec * (u_int)1000;
420	curtime += (u_int)tt.tv_usec / (u_int)1000;
421
422	if ((tt.tv_usec % 1000) >= 500)
423	curtime++;
424	}
425	#endif
426
427	void slirp_select_fill(PNATState pData, int *pnfds,
428	fd_set readfds, fd_set writefds, fd_set *xfds)
429	{
430	struct socket so, so_next;
431	int nfds;
432	#if defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) && defined(RT_OS_WINDOWS)
433	int rc;
434	int error;
435	#endif
436	#ifdef VBOX_WITH_BSD_REASS
437	int i;
438	#endif /* VBOX_WITH_BSD_REASS */
439
440	STAM_REL_PROFILE_START(&pData->StatFill, a);
441
442	nfds = *pnfds;
443
444	/*
445	* First, TCP sockets
446	*/
447	do_slowtimo = 0;
448	if (link_up)
449	{
450	/*
451	* *_slowtimo needs calling if there are IP fragments
452	* in the fragment queue, or there are TCP connections active
453	*/
454	#ifndef VBOX_WITH_BSD_REASS
455	do_slowtimo = ((tcb.so_next != &tcb)
456	\|\| ((struct ipasfrag )&ipq != u32_to_ptr(pData, ipq.next, struct ipasfrag )));
457	#else /* !VBOX_WITH_BSD_REASS */
458	/* XXX:
459	* triggering of fragment expiration should be the same but use new macroses
460	*/
461	do_slowtimo = (tcb.so_next != &tcb);
462	if (!do_slowtimo)
463	{
464	for (i = 0; i < IPREASS_NHASH; i++)
465	{
466	if (!TAILQ_EMPTY(&ipq[i]))
467	{
468	do_slowtimo = 1;
469	break;
470	}
471	}
472	}
473	#endif /* VBOX_WITH_BSD_REASS */
474
475	STAM_REL_COUNTER_RESET(&pData->StatTCP);
476	STAM_REL_COUNTER_RESET(&pData->StatTCPHot);
477
478	for (so = tcb.so_next; so != &tcb; so = so_next)
479	{
480	so_next = so->so_next;
481
482	STAM_REL_COUNTER_INC(&pData->StatTCP);
483
484	/*
485	* See if we need a tcp_fasttimo
486	*/
487	if (time_fasttimo == 0 && so->so_tcpcb->t_flags & TF_DELACK)
488	time_fasttimo = curtime; /* Flag when we want a fasttimo */
489
490	/*
491	* NOFDREF can include still connecting to local-host,
492	* newly socreated() sockets etc. Don't want to select these.
493	*/
494	if (so->so_state & SS_NOFDREF \|\| so->s == -1)
495	continue;
496
497	/*
498	* Set for reading sockets which are accepting
499	*/
500	if (so->so_state & SS_FACCEPTCONN)
501	{
502	STAM_REL_COUNTER_INC(&pData->StatTCPHot);
503	TCP_ENGAGE_EVENT1(so, readfds);
504	continue;
505	}
506
507	/*
508	* Set for writing sockets which are connecting
509	*/
510	if (so->so_state & SS_ISFCONNECTING)
511	{
512	STAM_REL_COUNTER_INC(&pData->StatTCPHot);
513	TCP_ENGAGE_EVENT1(so, writefds);
514	}
515
516	/*
517	* Set for writing if we are connected, can send more, and
518	* we have something to send
519	*/
520	if (CONN_CANFSEND(so) && so->so_rcv.sb_cc)
521	{
522	STAM_REL_COUNTER_INC(&pData->StatTCPHot);
523	TCP_ENGAGE_EVENT1(so, writefds);
524	}
525
526	/*
527	* Set for reading (and urgent data) if we are connected, can
528	* receive more, and we have room for it XXX /2 ?
529	*/
530	if (CONN_CANFRCV(so) && (so->so_snd.sb_cc < (so->so_snd.sb_datalen/2)))
531	{
532	STAM_REL_COUNTER_INC(&pData->StatTCPHot);
533	TCP_ENGAGE_EVENT2(so, readfds, xfds);
534	}
535	}
536
537	/*
538	* UDP sockets
539	*/
540	STAM_REL_COUNTER_RESET(&pData->StatUDP);
541	STAM_REL_COUNTER_RESET(&pData->StatUDPHot);
542
543	for (so = udb.so_next; so != &udb; so = so_next)
544	{
545	so_next = so->so_next;
546
547	STAM_REL_COUNTER_INC(&pData->StatUDP);
548
549	/*
550	* See if it's timed out
551	*/
552	if (so->so_expire)
553	{
554	if (so->so_expire <= curtime)
555	{
556	udp_detach(pData, so);
557	continue;
558	}
559	else
560	do_slowtimo = 1; /* Let socket expire */
561	}
562
563	/*
564	* When UDP packets are received from over the link, they're
565	* sendto()'d straight away, so no need for setting for writing
566	* Limit the number of packets queued by this session to 4.
567	* Note that even though we try and limit this to 4 packets,
568	* the session could have more queued if the packets needed
569	* to be fragmented.
570	*
571	* (XXX <= 4 ?)
572	*/
573	if ((so->so_state & SS_ISFCONNECTED) && so->so_queued <= 4)
574	{
575	STAM_REL_COUNTER_INC(&pData->StatUDPHot);
576	UDP_ENGAGE_EVENT(so, readfds);
577	}
578	}
579	}
580
581	#if !defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) \|\| !defined(RT_OS_WINDOWS)
582	*pnfds = nfds;
583	#else
584	*pnfds = VBOX_EVENT_COUNT;
585	#endif
586
587	STAM_REL_PROFILE_STOP(&pData->StatFill, a);
588	}
589
590	void slirp_select_poll(PNATState pData, fd_set readfds, fd_set writefds, fd_set *xfds)
591	{
592	struct socket so, so_next;
593	int ret;
594	#if defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) && defined(RT_OS_WINDOWS)
595	WSANETWORKEVENTS NetworkEvents;
596	int rc;
597	int error;
598	#endif
599	static uint32_t stat_time;
600
601	STAM_REL_PROFILE_START(&pData->StatPoll, a);
602
603	/* Update time */
604	updtime(pData);
605
606	#ifdef LOG_ENABLED
607	if (curtime - stat_time > 10000)
608	{
609	stat_time = curtime;
610	sockstats(pData);
611	}
612	#endif
613
614	/*
615	* See if anything has timed out
616	*/
617	if (link_up)
618	{
619	if (time_fasttimo && ((curtime - time_fasttimo) >= 2))
620	{
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	{
628	STAM_REL_PROFILE_START(&pData->StatSlowTimer, a);
629	ip_slowtimo(pData);
630	tcp_slowtimo(pData);
631	last_slowtimo = curtime;
632	STAM_REL_PROFILE_STOP(&pData->StatSlowTimer, a);
633	}
634	}
635	#if defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) && defined(RT_OS_WINDOWS)
636	if (!readfds && !writefds && !xfds)
637	return; /* only timer update */
638	#endif
639
640	/*
641	* Check sockets
642	*/
643	if (link_up)
644	{
645	/*
646	* Check TCP sockets
647	*/
648	for (so = tcb.so_next; so != &tcb; so = so_next)
649	{
650	so_next = so->so_next;
651
652	/*
653	* FD_ISSET is meaningless on these sockets
654	* (and they can crash the program)
655	*/
656	if (so->so_state & SS_NOFDREF \|\| so->s == -1)
657	continue;
658
659	#if defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) && defined(RT_OS_WINDOWS)
660	rc = WSAEnumNetworkEvents(so->s, VBOX_SOCKET_EVENT, &NetworkEvents);
661	if (rc == SOCKET_ERROR)
662	{
663	error = WSAGetLastError();
664	LogRel(("WSAEnumNetworkEvents TCP error %d\n", error));
665	continue;
666	}
667	#endif
668
669	/*
670	* Check for URG data
671	* This will soread as well, so no need to
672	* test for readfds below if this succeeds
673	*/
674	#if !defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) \|\| !defined(RT_OS_WINDOWS)
675	if (FD_ISSET(so->s, xfds))
676	#else
677	/* out-of-band data */
678	if ((NetworkEvents.lNetworkEvents & FD_OOB) && NetworkEvents.iErrorCode[FD_OOB_BIT] == 0)
679	#endif
680	{
681	sorecvoob(pData, so);
682	}
683
684	/*
685	* Check sockets for reading
686	*/
687	#if !defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) \|\| !defined(RT_OS_WINDOWS)
688	else if (FD_ISSET(so->s, readfds))
689	#else
690	else if ((NetworkEvents.lNetworkEvents & FD_READ) && (NetworkEvents.iErrorCode[FD_READ_BIT] == 0))
691	#endif
692	{
693	/*
694	* Check for incoming connections
695	*/
696	if (so->so_state & SS_FACCEPTCONN)
697	{
698	tcp_connect(pData, so);
699	#if defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) && defined(RT_OS_WINDOWS)
700	if (!NetworkEvents.lNetworkEvents & FD_CLOSE)
701	#endif
702	continue;
703	}
704
705	ret = soread(pData, so);
706	/* Output it if we read something */
707	if (ret > 0)
708	tcp_output(pData, sototcpcb(so));
709	}
710
711	#if defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) && defined(RT_OS_WINDOWS)
712	/*
713	* Check for FD_CLOSE events.
714	*/
715	if (NetworkEvents.lNetworkEvents & FD_CLOSE)
716	{
717	/*
718	* drain the socket
719	*/
720	for (;;)
721	{
722	ret = soread(pData, so);
723	if (ret > 0)
724	tcp_output(pData, sototcpcb(so));
725	else
726	break;
727	}
728	}
729	#endif
730
731	/*
732	* Check sockets for writing
733	*/
734	#if !defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) \|\| !defined(RT_OS_WINDOWS)
735	if (FD_ISSET(so->s, writefds))
736	#else
737	if ((NetworkEvents.lNetworkEvents & FD_WRITE) && (NetworkEvents.iErrorCode[FD_WRITE_BIT] == 0))
738	#endif
739	{
740	/*
741	* Check for non-blocking, still-connecting sockets
742	*/
743	if (so->so_state & SS_ISFCONNECTING)
744	{
745	/* Connected */
746	so->so_state &= ~SS_ISFCONNECTING;
747
748	/*
749	* This should be probably guarded by PROBE_CONN too. Anyway,
750	* we disable it on OS/2 because the below send call returns
751	* EFAULT which causes the opened TCP socket to close right
752	* after it has been opened and connected.
753	*/
754	#ifndef RT_OS_OS2
755	ret = send(so->s, (const char *)&ret, 0, 0);
756	if (ret < 0)
757	{
758	/* XXXXX Must fix, zero bytes is a NOP */
759	if ( errno == EAGAIN
760	\|\| errno == EWOULDBLOCK
761	\|\| errno == EINPROGRESS
762	\|\| errno == ENOTCONN)
763	continue;
764
765	/* else failed */
766	so->so_state = SS_NOFDREF;
767	}
768	/* else so->so_state &= ~SS_ISFCONNECTING; */
769	#endif
770
771	/*
772	* Continue tcp_input
773	*/
774	tcp_input(pData, (struct mbuf *)NULL, sizeof(struct ip), so);
775	/* continue; */
776	}
777	else
778	ret = sowrite(pData, so);
779	/*
780	* XXX If we wrote something (a lot), there could be the need
781	* for a window update. In the worst case, the remote will send
782	* a window probe to get things going again.
783	*/
784	}
785
786	/*
787	* Probe a still-connecting, non-blocking socket
788	* to check if it's still alive
789	*/
790	#ifdef PROBE_CONN
791	if (so->so_state & SS_ISFCONNECTING)
792	{
793	ret = recv(so->s, (char *)&ret, 0, 0);
794
795	if (ret < 0)
796	{
797	/* XXX */
798	if ( errno == EAGAIN
799	\|\| errno == EWOULDBLOCK
800	\|\| errno == EINPROGRESS
801	\|\| errno == ENOTCONN)
802	{
803	continue; /* Still connecting, continue */
804	}
805
806	/* else failed */
807	so->so_state = SS_NOFDREF;
808
809	/* tcp_input will take care of it */
810	}
811	else
812	{
813	ret = send(so->s, &ret, 0, 0);
814	if (ret < 0)
815	{
816	/* XXX */
817	if ( errno == EAGAIN
818	\|\| errno == EWOULDBLOCK
819	\|\| errno == EINPROGRESS
820	\|\| errno == ENOTCONN)
821	{
822	continue;
823	}
824	/* else failed */
825	so->so_state = SS_NOFDREF;
826	}
827	else
828	so->so_state &= ~SS_ISFCONNECTING;
829
830	}
831	tcp_input((struct mbuf *)NULL, sizeof(struct ip),so);
832	} /* SS_ISFCONNECTING */
833	#endif
834	}
835
836	/*
837	* Now UDP sockets.
838	* Incoming packets are sent straight away, they're not buffered.
839	* Incoming UDP data isn't buffered either.
840	*/
841	for (so = udb.so_next; so != &udb; so = so_next)
842	{
843	so_next = so->so_next;
844
845	#if defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) && defined(RT_OS_WINDOWS)
846	rc = WSAEnumNetworkEvents(so->s, VBOX_SOCKET_EVENT, &NetworkEvents);
847	if (rc == SOCKET_ERROR)
848	{
849	error = WSAGetLastError();
850	LogRel(("WSAEnumNetworkEvents TCP error %d\n", error));
851	continue;
852	}
853	#endif
854	#if !defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) \|\| !defined(RT_OS_WINDOWS)
855	if (so->s != -1 && FD_ISSET(so->s, readfds))
856	#else
857	if (((NetworkEvents.lNetworkEvents & FD_READ) && (NetworkEvents.iErrorCode[FD_READ_BIT] == 0)))
858	#endif
859	{
860	sorecvfrom(pData, so);
861	}
862	}
863	#if defined(VBOX_WITH_SLIRP_ICMP) && defined(RT_OS_WINDOWS)
864	sorecvfrom(pData, &pData->icmp_socket);
865	#endif
866	}
867
868	/*
869	* See if we can start outputting
870	*/
871	if (if_queued && link_up)
872	if_start(pData);
873
874	STAM_REL_PROFILE_STOP(&pData->StatPoll, a);
875	}
876
877	#define ETH_ALEN 6
878	#define ETH_HLEN 14
879
880	#define ETH_P_IP 0x0800 /* Internet Protocol packet */
881	#define ETH_P_ARP 0x0806 /* Address Resolution packet */
882
883	#define ARPOP_REQUEST 1 /* ARP request */
884	#define ARPOP_REPLY 2 /* ARP reply */
885
886	struct ethhdr
887	{
888	unsigned char h_dest[ETH_ALEN]; /* destination eth addr */
889	unsigned char h_source[ETH_ALEN]; /* source ether addr */
890	unsigned short h_proto; /* packet type ID field */
891	};
892
893	struct arphdr
894	{
895	unsigned short ar_hrd; /* format of hardware address */
896	unsigned short ar_pro; /* format of protocol address */
897	unsigned char ar_hln; /* length of hardware address */
898	unsigned char ar_pln; /* length of protocol address */
899	unsigned short ar_op; /* ARP opcode (command) */
900
901	/*
902	* Ethernet looks like this : This bit is variable sized however...
903	*/
904	unsigned char ar_sha[ETH_ALEN]; /* sender hardware address */
905	unsigned char ar_sip[4]; /* sender IP address */
906	unsigned char ar_tha[ETH_ALEN]; /* target hardware address */
907	unsigned char ar_tip[4]; /* target IP address */
908	};
909
910	static
911	void arp_input(PNATState pData, const uint8_t *pkt, int pkt_len)
912	{
913	struct ethhdr eh = (struct ethhdr )pkt;
914	struct arphdr ah = (struct arphdr )(pkt + ETH_HLEN);
915	uint8_t arp_reply[ETH_HLEN + sizeof(struct arphdr)];
916	struct ethhdr reh = (struct ethhdr )arp_reply;
917	struct arphdr rah = (struct arphdr )(arp_reply + ETH_HLEN);
918	int ar_op;
919	struct ex_list *ex_ptr;
920	uint32_t htip = ntohl((uint32_t)ah->ar_tip);
921
922	ar_op = ntohs(ah->ar_op);
923	switch(ar_op)
924	{
925	case ARPOP_REQUEST:
926	if ((htip & pData->netmask) == ntohl(special_addr.s_addr))
927	{
928	if ( (htip & ~pData->netmask) == CTL_DNS
929	\|\| (htip & ~pData->netmask) == CTL_ALIAS)
930	goto arp_ok;
931	for (ex_ptr = exec_list; ex_ptr; ex_ptr = ex_ptr->ex_next)
932	{
933	if ((htip & ~pData->netmask) == ex_ptr->ex_addr)
934	goto arp_ok;
935	}
936	return;
937	arp_ok:
938	/* XXX: make an ARP request to have the client address */
939	memcpy(client_ethaddr, eh->h_source, ETH_ALEN);
940
941	/* ARP request for alias/dns mac address */
942	memcpy(reh->h_dest, pkt + ETH_ALEN, ETH_ALEN);
943	memcpy(reh->h_source, special_ethaddr, ETH_ALEN - 1);
944	reh->h_source[5] = ah->ar_tip[3];
945	reh->h_proto = htons(ETH_P_ARP);
946
947	rah->ar_hrd = htons(1);
948	rah->ar_pro = htons(ETH_P_IP);
949	rah->ar_hln = ETH_ALEN;
950	rah->ar_pln = 4;
951	rah->ar_op = htons(ARPOP_REPLY);
952	memcpy(rah->ar_sha, reh->h_source, ETH_ALEN);
953	memcpy(rah->ar_sip, ah->ar_tip, 4);
954	memcpy(rah->ar_tha, ah->ar_sha, ETH_ALEN);
955	memcpy(rah->ar_tip, ah->ar_sip, 4);
956	slirp_output(pData->pvUser, arp_reply, sizeof(arp_reply));
957	}
958	break;
959	default:
960	break;
961	}
962	}
963
964	void slirp_input(PNATState pData, const uint8_t *pkt, int pkt_len)
965	{
966	struct mbuf *m;
967	int proto;
968
969	if (pkt_len < ETH_HLEN)
970	return;
971
972	proto = ntohs((uint16_t )(pkt + 12));
973	switch(proto)
974	{
975	case ETH_P_ARP:
976	arp_input(pData, pkt, pkt_len);
977	break;
978	case ETH_P_IP:
979	/* Update time. Important if the network is very quiet, as otherwise
980	* the first outgoing connection gets an incorrect timestamp. */
981	updtime(pData);
982
983	m = m_get(pData);
984	if (!m)
985	return;
986	/* Note: we add to align the IP header */
987	if (M_FREEROOM(m) < pkt_len + 2)
988	{
989	m_inc(m, pkt_len + 2);
990	}
991	m->m_len = pkt_len + 2;
992	memcpy(m->m_data + 2, pkt, pkt_len);
993
994	m->m_data += 2 + ETH_HLEN;
995	m->m_len -= 2 + ETH_HLEN;
996
997	ip_input(pData, m);
998	break;
999	default:
1000	break;
1001	}
1002	}
1003
1004	/* output the IP packet to the ethernet device */
1005	void if_encap(PNATState pData, const uint8_t *ip_data, int ip_data_len)
1006	{
1007	uint8_t buf[1600];
1008	struct ethhdr eh = (struct ethhdr )buf;
1009
1010	if (ip_data_len + ETH_HLEN > sizeof(buf))
1011	return;
1012
1013	memcpy(eh->h_dest, client_ethaddr, ETH_ALEN);
1014	memcpy(eh->h_source, special_ethaddr, ETH_ALEN - 1);
1015	/* XXX: not correct */
1016	eh->h_source[5] = CTL_ALIAS;
1017	eh->h_proto = htons(ETH_P_IP);
1018	memcpy(buf + sizeof(struct ethhdr), ip_data, ip_data_len);
1019	slirp_output(pData->pvUser, buf, ip_data_len + ETH_HLEN);
1020	}
1021
1022	int slirp_redir(PNATState pData, int is_udp, int host_port,
1023	struct in_addr guest_addr, int guest_port)
1024	{
1025	if (is_udp)
1026	{
1027	if (!udp_listen(pData, htons(host_port), guest_addr.s_addr,
1028	htons(guest_port), 0))
1029	return -1;
1030	}
1031	else
1032	{
1033	if (!solisten(pData, htons(host_port), guest_addr.s_addr,
1034	htons(guest_port), 0))
1035	return -1;
1036	}
1037	return 0;
1038	}
1039
1040	int slirp_add_exec(PNATState pData, int do_pty, const char *args, int addr_low_byte,
1041	int guest_port)
1042	{
1043	return add_exec(&exec_list, do_pty, (char *)args,
1044	addr_low_byte, htons(guest_port));
1045	}
1046
1047	void slirp_set_ethaddr(PNATState pData, const uint8_t *ethaddr)
1048	{
1049	memcpy(client_ethaddr, ethaddr, ETH_ALEN);
1050	}
1051
1052	#if defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) && defined(RT_OS_WINDOWS)
1053	HANDLE *slirp_get_events(PNATState pData)
1054	{
1055	return pData->phEvents;
1056	}
1057	void slirp_register_external_event(PNATState pData, HANDLE hEvent, int index)
1058	{
1059	pData->phEvents[index] = hEvent;
1060	}
1061	#endif
1062
1063	unsigned int slirp_get_timeout_ms(PNATState pData)
1064	{
1065	if (link_up)
1066	{
1067	if (time_fasttimo)
1068	return 2;
1069	if (do_slowtimo)
1070	return 500; /* see PR_SLOWHZ */
1071	}
1072	return 0;
1073	}

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

source: vbox/trunk/src/VBox/Devices/Network/slirp/slirp.c@ 15089

Download in other formats: