slirp.c@ 15091

Last change on this file since 15091 was 15091, checked in by vboxsync, 16 years ago
this is AssertCompile in the header
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	link_up = 1;
283
284	debug_init();
285	if_init(pData);
286	ip_init(pData);
287	#ifdef VBOX_WITH_SLIRP_ICMP
288	icmp_init(pData);
289	#endif /* VBOX_WITH_SLIRP_ICMP */
290
291	/* Initialise mbufs after setting the MTU */
292	m_init(pData);
293
294	/* set default addresses */
295	inet_aton("127.0.0.1", &loopback_addr);
296	inet_aton("127.0.0.1", &dns_addr);
297
298	if (get_dns_addr_domain(pData, true, &dns_addr, &pData->pszDomain) < 0)
299	fNATfailed = 1;
300
301	inet_aton(pszNetAddr, &special_addr);
302	alias_addr.s_addr = special_addr.s_addr \| htonl(CTL_ALIAS);
303	getouraddr(pData);
304	return fNATfailed ? VINF_NAT_DNS : VINF_SUCCESS;
305	}
306
307	/**
308	* Statistics counters.
309	*/
310	void slirp_register_timers(PNATState pData, PPDMDRVINS pDrvIns)
311	{
312	#ifndef VBOX_WITHOUT_RELEASE_STATISTICS
313	PDMDrvHlpSTAMRegisterF(pDrvIns, &pData->StatFill, STAMTYPE_PROFILE, STAMVISIBILITY_ALWAYS,
314	STAMUNIT_TICKS_PER_CALL, "Profiling slirp fills", "/Drivers/NAT%d/Fill", pDrvIns->iInstance);
315	PDMDrvHlpSTAMRegisterF(pDrvIns, &pData->StatPoll, STAMTYPE_PROFILE, STAMVISIBILITY_ALWAYS,
316	STAMUNIT_TICKS_PER_CALL, "Profiling slirp polls", "/Drivers/NAT%d/Poll", pDrvIns->iInstance);
317	PDMDrvHlpSTAMRegisterF(pDrvIns, &pData->StatFastTimer, STAMTYPE_PROFILE, STAMVISIBILITY_ALWAYS,
318	STAMUNIT_TICKS_PER_CALL, "Profiling slirp fast timer", "/Drivers/NAT%d/TimerFast", pDrvIns->iInstance);
319	PDMDrvHlpSTAMRegisterF(pDrvIns, &pData->StatSlowTimer, STAMTYPE_PROFILE, STAMVISIBILITY_ALWAYS,
320	STAMUNIT_TICKS_PER_CALL, "Profiling slirp slow timer", "/Drivers/NAT%d/TimerSlow", pDrvIns->iInstance);
321	PDMDrvHlpSTAMRegisterF(pDrvIns, &pData->StatTCP, STAMTYPE_COUNTER, STAMVISIBILITY_ALWAYS,
322	STAMUNIT_COUNT, "TCP sockets", "/Drivers/NAT%d/SockTCP", pDrvIns->iInstance);
323	PDMDrvHlpSTAMRegisterF(pDrvIns, &pData->StatTCPHot, STAMTYPE_COUNTER, STAMVISIBILITY_ALWAYS,
324	STAMUNIT_COUNT, "TCP sockets active", "/Drivers/NAT%d/SockTCPHot", pDrvIns->iInstance);
325	PDMDrvHlpSTAMRegisterF(pDrvIns, &pData->StatUDP, STAMTYPE_COUNTER, STAMVISIBILITY_ALWAYS,
326	STAMUNIT_COUNT, "UDP sockets", "/Drivers/NAT%d/SockUDP", pDrvIns->iInstance);
327	PDMDrvHlpSTAMRegisterF(pDrvIns, &pData->StatUDPHot, STAMTYPE_COUNTER, STAMVISIBILITY_ALWAYS,
328	STAMUNIT_COUNT, "UDP sockets active", "/Drivers/NAT%d/SockUDPHot", pDrvIns->iInstance);
329	#endif /* VBOX_WITHOUT_RELEASE_STATISTICS */
330	}
331
332	/**
333	* Marks the link as up, making it possible to establish new connections.
334	*/
335	void slirp_link_up(PNATState pData)
336	{
337	link_up = 1;
338	}
339
340	/**
341	* Marks the link as down and cleans up the current connections.
342	*/
343	void slirp_link_down(PNATState pData)
344	{
345	struct socket *so;
346
347	while ((so = tcb.so_next) != &tcb)
348	{
349	if (so->so_state & SS_NOFDREF \|\| so->s == -1)
350	sofree(pData, so);
351	else
352	tcp_drop(pData, sototcpcb(so), 0);
353	}
354
355	while ((so = udb.so_next) != &udb)
356	udp_detach(pData, so);
357
358	link_up = 0;
359	}
360
361	/**
362	* Terminates the slirp component.
363	*/
364	void slirp_term(PNATState pData)
365	{
366	if (pData->pszDomain)
367	RTStrFree((char )(void )pData->pszDomain);
368
369	#if ARCH_BITS == 64 && !defined(VBOX_WITH_BSD_REASS)
370	LogRel(("NAT: cpvHashUsed=%RU32 cpvHashCollisions=%RU32 cpvHashInserts=%RU64 cpvHashDone=%RU64\n",
371	pData->cpvHashUsed, pData->cpvHashCollisions, pData->cpvHashInserts, pData->cpvHashDone));
372	#endif
373
374	slirp_link_down(pData);
375	#ifdef RT_OS_WINDOWS
376	WSACleanup();
377	#endif
378	#ifdef LOG_ENABLED
379	Log(("\n"
380	"NAT statistics\n"
381	"--------------\n"
382	"\n"));
383	ipstats(pData);
384	tcpstats(pData);
385	udpstats(pData);
386	icmpstats(pData);
387	mbufstats(pData);
388	sockstats(pData);
389	Log(("\n"
390	"\n"
391	"\n"));
392	#endif
393	free(pData);
394	}
395
396
397	#define CONN_CANFSEND(so) (((so)->so_state & (SS_FCANTSENDMORE\|SS_ISFCONNECTED)) == SS_ISFCONNECTED)
398	#define CONN_CANFRCV(so) (((so)->so_state & (SS_FCANTRCVMORE\|SS_ISFCONNECTED)) == SS_ISFCONNECTED)
399	#define UPD_NFDS(x) if (nfds < (x)) nfds = (x)
400
401	/*
402	* curtime kept to an accuracy of 1ms
403	*/
404	#ifdef RT_OS_WINDOWS
405	static void updtime(PNATState pData)
406	{
407	struct _timeb tb;
408
409	_ftime(&tb);
410	curtime = (u_int)tb.time * (u_int)1000;
411	curtime += (u_int)tb.millitm;
412	}
413	#else
414	static void updtime(PNATState pData)
415	{
416	gettimeofday(&tt, 0);
417
418	curtime = (u_int)tt.tv_sec * (u_int)1000;
419	curtime += (u_int)tt.tv_usec / (u_int)1000;
420
421	if ((tt.tv_usec % 1000) >= 500)
422	curtime++;
423	}
424	#endif
425
426	void slirp_select_fill(PNATState pData, int *pnfds,
427	fd_set readfds, fd_set writefds, fd_set *xfds)
428	{
429	struct socket so, so_next;
430	int nfds;
431	#if defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) && defined(RT_OS_WINDOWS)
432	int rc;
433	int error;
434	#endif
435	#ifdef VBOX_WITH_BSD_REASS
436	int i;
437	#endif /* VBOX_WITH_BSD_REASS */
438
439	STAM_REL_PROFILE_START(&pData->StatFill, a);
440
441	nfds = *pnfds;
442
443	/*
444	* First, TCP sockets
445	*/
446	do_slowtimo = 0;
447	if (link_up)
448	{
449	/*
450	* *_slowtimo needs calling if there are IP fragments
451	* in the fragment queue, or there are TCP connections active
452	*/
453	#ifndef VBOX_WITH_BSD_REASS
454	do_slowtimo = ((tcb.so_next != &tcb)
455	\|\| ((struct ipasfrag )&ipq != u32_to_ptr(pData, ipq.next, struct ipasfrag )));
456	#else /* !VBOX_WITH_BSD_REASS */
457	/* XXX:
458	* triggering of fragment expiration should be the same but use new macroses
459	*/
460	do_slowtimo = (tcb.so_next != &tcb);
461	if (!do_slowtimo)
462	{
463	for (i = 0; i < IPREASS_NHASH; i++)
464	{
465	if (!TAILQ_EMPTY(&ipq[i]))
466	{
467	do_slowtimo = 1;
468	break;
469	}
470	}
471	}
472	#endif /* VBOX_WITH_BSD_REASS */
473
474	STAM_REL_COUNTER_RESET(&pData->StatTCP);
475	STAM_REL_COUNTER_RESET(&pData->StatTCPHot);
476
477	for (so = tcb.so_next; so != &tcb; so = so_next)
478	{
479	so_next = so->so_next;
480
481	STAM_REL_COUNTER_INC(&pData->StatTCP);
482
483	/*
484	* See if we need a tcp_fasttimo
485	*/
486	if (time_fasttimo == 0 && so->so_tcpcb->t_flags & TF_DELACK)
487	time_fasttimo = curtime; /* Flag when we want a fasttimo */
488
489	/*
490	* NOFDREF can include still connecting to local-host,
491	* newly socreated() sockets etc. Don't want to select these.
492	*/
493	if (so->so_state & SS_NOFDREF \|\| so->s == -1)
494	continue;
495
496	/*
497	* Set for reading sockets which are accepting
498	*/
499	if (so->so_state & SS_FACCEPTCONN)
500	{
501	STAM_REL_COUNTER_INC(&pData->StatTCPHot);
502	TCP_ENGAGE_EVENT1(so, readfds);
503	continue;
504	}
505
506	/*
507	* Set for writing sockets which are connecting
508	*/
509	if (so->so_state & SS_ISFCONNECTING)
510	{
511	STAM_REL_COUNTER_INC(&pData->StatTCPHot);
512	TCP_ENGAGE_EVENT1(so, writefds);
513	}
514
515	/*
516	* Set for writing if we are connected, can send more, and
517	* we have something to send
518	*/
519	if (CONN_CANFSEND(so) && so->so_rcv.sb_cc)
520	{
521	STAM_REL_COUNTER_INC(&pData->StatTCPHot);
522	TCP_ENGAGE_EVENT1(so, writefds);
523	}
524
525	/*
526	* Set for reading (and urgent data) if we are connected, can
527	* receive more, and we have room for it XXX /2 ?
528	*/
529	if (CONN_CANFRCV(so) && (so->so_snd.sb_cc < (so->so_snd.sb_datalen/2)))
530	{
531	STAM_REL_COUNTER_INC(&pData->StatTCPHot);
532	TCP_ENGAGE_EVENT2(so, readfds, xfds);
533	}
534	}
535
536	/*
537	* UDP sockets
538	*/
539	STAM_REL_COUNTER_RESET(&pData->StatUDP);
540	STAM_REL_COUNTER_RESET(&pData->StatUDPHot);
541
542	for (so = udb.so_next; so != &udb; so = so_next)
543	{
544	so_next = so->so_next;
545
546	STAM_REL_COUNTER_INC(&pData->StatUDP);
547
548	/*
549	* See if it's timed out
550	*/
551	if (so->so_expire)
552	{
553	if (so->so_expire <= curtime)
554	{
555	udp_detach(pData, so);
556	continue;
557	}
558	else
559	do_slowtimo = 1; /* Let socket expire */
560	}
561
562	/*
563	* When UDP packets are received from over the link, they're
564	* sendto()'d straight away, so no need for setting for writing
565	* Limit the number of packets queued by this session to 4.
566	* Note that even though we try and limit this to 4 packets,
567	* the session could have more queued if the packets needed
568	* to be fragmented.
569	*
570	* (XXX <= 4 ?)
571	*/
572	if ((so->so_state & SS_ISFCONNECTED) && so->so_queued <= 4)
573	{
574	STAM_REL_COUNTER_INC(&pData->StatUDPHot);
575	UDP_ENGAGE_EVENT(so, readfds);
576	}
577	}
578	}
579
580	#if !defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) \|\| !defined(RT_OS_WINDOWS)
581	*pnfds = nfds;
582	#else
583	*pnfds = VBOX_EVENT_COUNT;
584	#endif
585
586	STAM_REL_PROFILE_STOP(&pData->StatFill, a);
587	}
588
589	void slirp_select_poll(PNATState pData, fd_set readfds, fd_set writefds, fd_set *xfds)
590	{
591	struct socket so, so_next;
592	int ret;
593	#if defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) && defined(RT_OS_WINDOWS)
594	WSANETWORKEVENTS NetworkEvents;
595	int rc;
596	int error;
597	#endif
598	static uint32_t stat_time;
599
600	STAM_REL_PROFILE_START(&pData->StatPoll, a);
601
602	/* Update time */
603	updtime(pData);
604
605	#ifdef LOG_ENABLED
606	if (curtime - stat_time > 10000)
607	{
608	stat_time = curtime;
609	sockstats(pData);
610	}
611	#endif
612
613	/*
614	* See if anything has timed out
615	*/
616	if (link_up)
617	{
618	if (time_fasttimo && ((curtime - time_fasttimo) >= 2))
619	{
620	STAM_REL_PROFILE_START(&pData->StatFastTimer, a);
621	tcp_fasttimo(pData);
622	time_fasttimo = 0;
623	STAM_REL_PROFILE_STOP(&pData->StatFastTimer, a);
624	}
625	if (do_slowtimo && ((curtime - last_slowtimo) >= 499))
626	{
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_SIMPLIFIED_SLIRP_SYNC) && defined(RT_OS_WINDOWS)
635	if (!readfds && !writefds && !xfds)
636	return; /* only timer update */
637	#endif
638
639	/*
640	* Check sockets
641	*/
642	if (link_up)
643	{
644	/*
645	* Check TCP sockets
646	*/
647	for (so = tcb.so_next; so != &tcb; so = so_next)
648	{
649	so_next = so->so_next;
650
651	/*
652	* FD_ISSET is meaningless on these sockets
653	* (and they can crash the program)
654	*/
655	if (so->so_state & SS_NOFDREF \|\| so->s == -1)
656	continue;
657
658	#if defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) && defined(RT_OS_WINDOWS)
659	rc = WSAEnumNetworkEvents(so->s, VBOX_SOCKET_EVENT, &NetworkEvents);
660	if (rc == SOCKET_ERROR)
661	{
662	error = WSAGetLastError();
663	LogRel(("WSAEnumNetworkEvents TCP error %d\n", error));
664	continue;
665	}
666	#endif
667
668	/*
669	* Check for URG data
670	* This will soread as well, so no need to
671	* test for readfds below if this succeeds
672	*/
673	#if !defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) \|\| !defined(RT_OS_WINDOWS)
674	if (FD_ISSET(so->s, xfds))
675	#else
676	/* out-of-band data */
677	if ((NetworkEvents.lNetworkEvents & FD_OOB) && NetworkEvents.iErrorCode[FD_OOB_BIT] == 0)
678	#endif
679	{
680	sorecvoob(pData, so);
681	}
682
683	/*
684	* Check sockets for reading
685	*/
686	#if !defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) \|\| !defined(RT_OS_WINDOWS)
687	else if (FD_ISSET(so->s, readfds))
688	#else
689	else if ((NetworkEvents.lNetworkEvents & FD_READ) && (NetworkEvents.iErrorCode[FD_READ_BIT] == 0))
690	#endif
691	{
692	/*
693	* Check for incoming connections
694	*/
695	if (so->so_state & SS_FACCEPTCONN)
696	{
697	tcp_connect(pData, so);
698	#if defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) && defined(RT_OS_WINDOWS)
699	if (!NetworkEvents.lNetworkEvents & FD_CLOSE)
700	#endif
701	continue;
702	}
703
704	ret = soread(pData, so);
705	/* Output it if we read something */
706	if (ret > 0)
707	tcp_output(pData, sototcpcb(so));
708	}
709
710	#if defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) && defined(RT_OS_WINDOWS)
711	/*
712	* Check for FD_CLOSE events.
713	*/
714	if (NetworkEvents.lNetworkEvents & FD_CLOSE)
715	{
716	/*
717	* drain the socket
718	*/
719	for (;;)
720	{
721	ret = soread(pData, so);
722	if (ret > 0)
723	tcp_output(pData, sototcpcb(so));
724	else
725	break;
726	}
727	}
728	#endif
729
730	/*
731	* Check sockets for writing
732	*/
733	#if !defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) \|\| !defined(RT_OS_WINDOWS)
734	if (FD_ISSET(so->s, writefds))
735	#else
736	if ((NetworkEvents.lNetworkEvents & FD_WRITE) && (NetworkEvents.iErrorCode[FD_WRITE_BIT] == 0))
737	#endif
738	{
739	/*
740	* Check for non-blocking, still-connecting sockets
741	*/
742	if (so->so_state & SS_ISFCONNECTING)
743	{
744	/* Connected */
745	so->so_state &= ~SS_ISFCONNECTING;
746
747	/*
748	* This should be probably guarded by PROBE_CONN too. Anyway,
749	* we disable it on OS/2 because the below send call returns
750	* EFAULT which causes the opened TCP socket to close right
751	* after it has been opened and connected.
752	*/
753	#ifndef RT_OS_OS2
754	ret = send(so->s, (const char *)&ret, 0, 0);
755	if (ret < 0)
756	{
757	/* XXXXX Must fix, zero bytes is a NOP */
758	if ( errno == EAGAIN
759	\|\| errno == EWOULDBLOCK
760	\|\| errno == EINPROGRESS
761	\|\| errno == ENOTCONN)
762	continue;
763
764	/* else failed */
765	so->so_state = SS_NOFDREF;
766	}
767	/* else so->so_state &= ~SS_ISFCONNECTING; */
768	#endif
769
770	/*
771	* Continue tcp_input
772	*/
773	tcp_input(pData, (struct mbuf *)NULL, sizeof(struct ip), so);
774	/* continue; */
775	}
776	else
777	ret = sowrite(pData, so);
778	/*
779	* XXX If we wrote something (a lot), there could be the need
780	* for a window update. In the worst case, the remote will send
781	* a window probe to get things going again.
782	*/
783	}
784
785	/*
786	* Probe a still-connecting, non-blocking socket
787	* to check if it's still alive
788	*/
789	#ifdef PROBE_CONN
790	if (so->so_state & SS_ISFCONNECTING)
791	{
792	ret = recv(so->s, (char *)&ret, 0, 0);
793
794	if (ret < 0)
795	{
796	/* XXX */
797	if ( errno == EAGAIN
798	\|\| errno == EWOULDBLOCK
799	\|\| errno == EINPROGRESS
800	\|\| errno == ENOTCONN)
801	{
802	continue; /* Still connecting, continue */
803	}
804
805	/* else failed */
806	so->so_state = SS_NOFDREF;
807
808	/* tcp_input will take care of it */
809	}
810	else
811	{
812	ret = send(so->s, &ret, 0, 0);
813	if (ret < 0)
814	{
815	/* XXX */
816	if ( errno == EAGAIN
817	\|\| errno == EWOULDBLOCK
818	\|\| errno == EINPROGRESS
819	\|\| errno == ENOTCONN)
820	{
821	continue;
822	}
823	/* else failed */
824	so->so_state = SS_NOFDREF;
825	}
826	else
827	so->so_state &= ~SS_ISFCONNECTING;
828
829	}
830	tcp_input((struct mbuf *)NULL, sizeof(struct ip),so);
831	} /* SS_ISFCONNECTING */
832	#endif
833	}
834
835	/*
836	* Now UDP sockets.
837	* Incoming packets are sent straight away, they're not buffered.
838	* Incoming UDP data isn't buffered either.
839	*/
840	for (so = udb.so_next; so != &udb; so = so_next)
841	{
842	so_next = so->so_next;
843
844	#if defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) && defined(RT_OS_WINDOWS)
845	rc = WSAEnumNetworkEvents(so->s, VBOX_SOCKET_EVENT, &NetworkEvents);
846	if (rc == SOCKET_ERROR)
847	{
848	error = WSAGetLastError();
849	LogRel(("WSAEnumNetworkEvents TCP error %d\n", error));
850	continue;
851	}
852	#endif
853	#if !defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) \|\| !defined(RT_OS_WINDOWS)
854	if (so->s != -1 && FD_ISSET(so->s, readfds))
855	#else
856	if (((NetworkEvents.lNetworkEvents & FD_READ) && (NetworkEvents.iErrorCode[FD_READ_BIT] == 0)))
857	#endif
858	{
859	sorecvfrom(pData, so);
860	}
861	}
862	#if defined(VBOX_WITH_SLIRP_ICMP) && defined(RT_OS_WINDOWS)
863	sorecvfrom(pData, &pData->icmp_socket);
864	#endif
865	}
866
867	/*
868	* See if we can start outputting
869	*/
870	if (if_queued && link_up)
871	if_start(pData);
872
873	STAM_REL_PROFILE_STOP(&pData->StatPoll, a);
874	}
875
876	#define ETH_ALEN 6
877	#define ETH_HLEN 14
878
879	#define ETH_P_IP 0x0800 /* Internet Protocol packet */
880	#define ETH_P_ARP 0x0806 /* Address Resolution packet */
881
882	#define ARPOP_REQUEST 1 /* ARP request */
883	#define ARPOP_REPLY 2 /* ARP reply */
884
885	struct ethhdr
886	{
887	unsigned char h_dest[ETH_ALEN]; /* destination eth addr */
888	unsigned char h_source[ETH_ALEN]; /* source ether addr */
889	unsigned short h_proto; /* packet type ID field */
890	};
891
892	struct arphdr
893	{
894	unsigned short ar_hrd; /* format of hardware address */
895	unsigned short ar_pro; /* format of protocol address */
896	unsigned char ar_hln; /* length of hardware address */
897	unsigned char ar_pln; /* length of protocol address */
898	unsigned short ar_op; /* ARP opcode (command) */
899
900	/*
901	* Ethernet looks like this : This bit is variable sized however...
902	*/
903	unsigned char ar_sha[ETH_ALEN]; /* sender hardware address */
904	unsigned char ar_sip[4]; /* sender IP address */
905	unsigned char ar_tha[ETH_ALEN]; /* target hardware address */
906	unsigned char ar_tip[4]; /* target IP address */
907	};
908
909	static
910	void arp_input(PNATState pData, const uint8_t *pkt, int pkt_len)
911	{
912	struct ethhdr eh = (struct ethhdr )pkt;
913	struct arphdr ah = (struct arphdr )(pkt + ETH_HLEN);
914	uint8_t arp_reply[ETH_HLEN + sizeof(struct arphdr)];
915	struct ethhdr reh = (struct ethhdr )arp_reply;
916	struct arphdr rah = (struct arphdr )(arp_reply + ETH_HLEN);
917	int ar_op;
918	struct ex_list *ex_ptr;
919	uint32_t htip = ntohl((uint32_t)ah->ar_tip);
920
921	ar_op = ntohs(ah->ar_op);
922	switch(ar_op)
923	{
924	case ARPOP_REQUEST:
925	if ((htip & pData->netmask) == ntohl(special_addr.s_addr))
926	{
927	if ( (htip & ~pData->netmask) == CTL_DNS
928	\|\| (htip & ~pData->netmask) == CTL_ALIAS)
929	goto arp_ok;
930	for (ex_ptr = exec_list; ex_ptr; ex_ptr = ex_ptr->ex_next)
931	{
932	if ((htip & ~pData->netmask) == ex_ptr->ex_addr)
933	goto arp_ok;
934	}
935	return;
936	arp_ok:
937	/* XXX: make an ARP request to have the client address */
938	memcpy(client_ethaddr, eh->h_source, ETH_ALEN);
939
940	/* ARP request for alias/dns mac address */
941	memcpy(reh->h_dest, pkt + ETH_ALEN, ETH_ALEN);
942	memcpy(reh->h_source, special_ethaddr, ETH_ALEN - 1);
943	reh->h_source[5] = ah->ar_tip[3];
944	reh->h_proto = htons(ETH_P_ARP);
945
946	rah->ar_hrd = htons(1);
947	rah->ar_pro = htons(ETH_P_IP);
948	rah->ar_hln = ETH_ALEN;
949	rah->ar_pln = 4;
950	rah->ar_op = htons(ARPOP_REPLY);
951	memcpy(rah->ar_sha, reh->h_source, ETH_ALEN);
952	memcpy(rah->ar_sip, ah->ar_tip, 4);
953	memcpy(rah->ar_tha, ah->ar_sha, ETH_ALEN);
954	memcpy(rah->ar_tip, ah->ar_sip, 4);
955	slirp_output(pData->pvUser, arp_reply, sizeof(arp_reply));
956	}
957	break;
958	default:
959	break;
960	}
961	}
962
963	void slirp_input(PNATState pData, const uint8_t *pkt, int pkt_len)
964	{
965	struct mbuf *m;
966	int proto;
967
968	if (pkt_len < ETH_HLEN)
969	return;
970
971	proto = ntohs((uint16_t )(pkt + 12));
972	switch(proto)
973	{
974	case ETH_P_ARP:
975	arp_input(pData, pkt, pkt_len);
976	break;
977	case ETH_P_IP:
978	/* Update time. Important if the network is very quiet, as otherwise
979	* the first outgoing connection gets an incorrect timestamp. */
980	updtime(pData);
981
982	m = m_get(pData);
983	if (!m)
984	return;
985	/* Note: we add to align the IP header */
986	if (M_FREEROOM(m) < pkt_len + 2)
987	{
988	m_inc(m, pkt_len + 2);
989	}
990	m->m_len = pkt_len + 2;
991	memcpy(m->m_data + 2, pkt, pkt_len);
992
993	m->m_data += 2 + ETH_HLEN;
994	m->m_len -= 2 + ETH_HLEN;
995
996	ip_input(pData, m);
997	break;
998	default:
999	break;
1000	}
1001	}
1002
1003	/* output the IP packet to the ethernet device */
1004	void if_encap(PNATState pData, const uint8_t *ip_data, int ip_data_len)
1005	{
1006	uint8_t buf[1600];
1007	struct ethhdr eh = (struct ethhdr )buf;
1008
1009	if (ip_data_len + ETH_HLEN > sizeof(buf))
1010	return;
1011
1012	memcpy(eh->h_dest, client_ethaddr, ETH_ALEN);
1013	memcpy(eh->h_source, special_ethaddr, ETH_ALEN - 1);
1014	/* XXX: not correct */
1015	eh->h_source[5] = CTL_ALIAS;
1016	eh->h_proto = htons(ETH_P_IP);
1017	memcpy(buf + sizeof(struct ethhdr), ip_data, ip_data_len);
1018	slirp_output(pData->pvUser, buf, ip_data_len + ETH_HLEN);
1019	}
1020
1021	int slirp_redir(PNATState pData, int is_udp, int host_port,
1022	struct in_addr guest_addr, int guest_port)
1023	{
1024	if (is_udp)
1025	{
1026	if (!udp_listen(pData, htons(host_port), guest_addr.s_addr,
1027	htons(guest_port), 0))
1028	return -1;
1029	}
1030	else
1031	{
1032	if (!solisten(pData, htons(host_port), guest_addr.s_addr,
1033	htons(guest_port), 0))
1034	return -1;
1035	}
1036	return 0;
1037	}
1038
1039	int slirp_add_exec(PNATState pData, int do_pty, const char *args, int addr_low_byte,
1040	int guest_port)
1041	{
1042	return add_exec(&exec_list, do_pty, (char *)args,
1043	addr_low_byte, htons(guest_port));
1044	}
1045
1046	void slirp_set_ethaddr(PNATState pData, const uint8_t *ethaddr)
1047	{
1048	memcpy(client_ethaddr, ethaddr, ETH_ALEN);
1049	}
1050
1051	#if defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) && defined(RT_OS_WINDOWS)
1052	HANDLE *slirp_get_events(PNATState pData)
1053	{
1054	return pData->phEvents;
1055	}
1056	void slirp_register_external_event(PNATState pData, HANDLE hEvent, int index)
1057	{
1058	pData->phEvents[index] = hEvent;
1059	}
1060	#endif
1061
1062	unsigned int slirp_get_timeout_ms(PNATState pData)
1063	{
1064	if (link_up)
1065	{
1066	if (time_fasttimo)
1067	return 2;
1068	if (do_slowtimo)
1069	return 500; /* see PR_SLOWHZ */
1070	}
1071	return 0;
1072	}

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

source: vbox/trunk/src/VBox/Devices/Network/slirp/slirp.c@ 15091

Download in other formats: