slirp.c@ 15243

Last change on this file since 15243 was 15243, checked in by vboxsync, 16 years ago
slirp: another fix for windows hosts
Property svn:eol-style set to `native`
File size: 34.0 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) \|\| !defined(RT_OS_WINDOWS)
14
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, fdset1, fdset2, label) \
23	do { \
24	FD_SET((so)->s, (fdset1)); \
25	FD_SET((so)->s, (fdset2)); \
26	UPD_NFDS((so)->s); \
27	} while(0)
28
29	# define DO_POLL_EVENTS(rc, error, so, events, label) do {} while (0)
30
31	#define DO_CHECK_FD_SET(so, events, fdset) (FD_ISSET((so)->s, (fdset)))
32
33	# ifdef VBOX_WITH_SLIRP_ICMP
34	# define ICMP_ENGAGE_EVENT(so, fdset) \
35	do { \
36	if (pData->icmp_socket.s != -1) \
37	DO_ENGAGE_EVENT1((so), (fdset), ICMP); \
38	} while (0)
39	# else /* !VBOX_WITH_SLIRP_ICMP */
40	# define ICMP_ENGAGE_EVENT(so, fdset) do {} while (0)
41	# endif /* !VBOX_WITH_SLIRP_ICMP */
42
43	#else /* defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) && defined(RT_OS_WINDOWS) */
44
45	/*
46	* On Windows, we will be notified by IcmpSendEcho2() when the response arrives.
47	* So no call to WSAEventSelect necessary.
48	*/
49	# define ICMP_ENGAGE_EVENT(so, fdset) do {} while(0)
50
51	# define DO_ENGAGE_EVENT1(so, fdset1, label) \
52	do { \
53	rc = WSAEventSelect((so)->s, VBOX_SOCKET_EVENT, FD_ALL_EVENTS); \
54	if (rc == SOCKET_ERROR) \
55	{ \
56	/* This should not happen */ \
57	error = WSAGetLastError(); \
58	LogRel(("WSAEventSelector (" #label ") error %d (so=%x, socket=%s, event=%x)\n", \
59	error, (so), (so)->s, VBOX_SOCKET_EVENT)); \
60	} \
61	} while(0); \
62	continue
63
64	# define DO_ENGAGE_EVENT2(so, fdset1, fdset2, label) \
65	DO_ENGAGE_EVENT1((so), (fdset1), label)
66
67	# define DO_POLL_EVENTS(rc, error, so, events, label) \
68	(rc) = WSAEnumNetworkEvents((so)->s, VBOX_SOCKET_EVENT, (events)); \
69	if ((rc) == SOCKET_ERROR) \
70	{ \
71	(error) = WSAGetLastError(); \
72	LogRel(("WSAEnumNetworkEvents " #label " error %d\n", (error))); \
73	continue; \
74	}
75
76	# define readfds_win FD_READ
77	# define readfds_win_bit FD_READ_BIT
78
79	# define writefds_win FD_WRITE
80	# define writefds_win_bit FD_WRITE_BIT
81
82	# define xfds_win FD_OOB
83	# define xfds_win_bit FD_OOB_BIT
84
85	# define DO_CHECK_FD_SET(so, events, fdset) \
86	(((events).lNetworkEvents & fdset ## _win) && ((events).iErrorCode[fdset ## _win_bit] == 0))
87
88
89	#endif /* defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) && defined(RT_OS_WINDOWS) */
90
91	#define TCP_ENGAGE_EVENT1(so, fdset) \
92	DO_ENGAGE_EVENT1((so), (fdset), TCP)
93
94	#define TCP_ENGAGE_EVENT2(so, fdset1, fdset2) \
95	DO_ENGAGE_EVENT2((so), (fdset1), (fdset2), TCP)
96
97	#define UDP_ENGAGE_EVENT(so, fdset) \
98	DO_ENGAGE_EVENT1((so), (fdset), UDP)
99
100	#define POLL_TCP_EVENTS(rc, error, so, events) \
101	DO_POLL_EVENTS((rc), (error), (so), (events), TCP)
102
103	#define POLL_UDP_EVENTS(rc, error, so, events) \
104	DO_POLL_EVENTS((rc), (error), (so), (events), UDP)
105
106	#define CHECK_FD_SET(so, events, set) \
107	(DO_CHECK_FD_SET((so), (events), set))
108
109	static const uint8_t special_ethaddr[6] =
110	{
111	0x52, 0x54, 0x00, 0x12, 0x35, 0x00
112	};
113
114	#ifdef RT_OS_WINDOWS
115
116	static int get_dns_addr_domain(PNATState pData, bool fVerbose,
117	struct in_addr *pdns_addr,
118	const char **ppszDomain)
119	{
120	int rc = 0;
121	FIXED_INFO *FixedInfo = NULL;
122	ULONG BufLen;
123	DWORD ret;
124	IP_ADDR_STRING *pIPAddr;
125	struct in_addr tmp_addr;
126
127	FixedInfo = (FIXED_INFO *)GlobalAlloc(GPTR, sizeof(FIXED_INFO));
128	BufLen = sizeof(FIXED_INFO);
129
130	/** @todo: this API returns all DNS servers, no matter whether the
131	* corresponding network adapter is disabled or not. Maybe replace
132	* this by GetAdapterAddresses(), which is XP/Vista only though. */
133	if (ERROR_BUFFER_OVERFLOW == GetNetworkParams(FixedInfo, &BufLen))
134	{
135	if (FixedInfo)
136	{
137	GlobalFree(FixedInfo);
138	FixedInfo = NULL;
139	}
140	FixedInfo = GlobalAlloc(GPTR, BufLen);
141	}
142
143	if ((ret = GetNetworkParams(FixedInfo, &BufLen)) != ERROR_SUCCESS)
144	{
145	Log(("GetNetworkParams failed. ret = %08x\n", (u_int)ret ));
146	if (FixedInfo)
147	{
148	GlobalFree(FixedInfo);
149	FixedInfo = NULL;
150	}
151	rc = -1;
152	goto get_dns_prefix;
153	}
154
155	pIPAddr = &(FixedInfo->DnsServerList);
156	inet_aton(pIPAddr->IpAddress.String, &tmp_addr);
157	Log(("nat: DNS Servers:\n"));
158	if (fVerbose \|\| pdns_addr->s_addr != tmp_addr.s_addr)
159	LogRel(("NAT: DNS address: %s\n", pIPAddr->IpAddress.String));
160	*pdns_addr = tmp_addr;
161
162	pIPAddr = FixedInfo -> DnsServerList.Next;
163	while (pIPAddr)
164	{
165	if (fVerbose)
166	LogRel(("NAT: ignored DNS address: %s\n", pIPAddr ->IpAddress.String));
167	pIPAddr = pIPAddr ->Next;
168	}
169	if (FixedInfo)
170	{
171	GlobalFree(FixedInfo);
172	FixedInfo = NULL;
173	}
174
175	get_dns_prefix:
176	if (ppszDomain)
177	{
178	OSVERSIONINFO ver;
179	char szDnsDomain[256];
180	DWORD dwSize = sizeof(szDnsDomain);
181
182	*ppszDomain = NULL;
183	GetVersionEx(&ver);
184	if (ver.dwMajorVersion >= 5)
185	{
186	/* GetComputerNameEx exists in Windows versions starting with 2000. */
187	if (GetComputerNameEx(ComputerNameDnsDomain, szDnsDomain, &dwSize))
188	{
189	if (szDnsDomain[0])
190	{
191	/* Just non-empty strings are valid. */
192	*ppszDomain = RTStrDup(szDnsDomain);
193	if (pData->fPassDomain)
194	{
195	if (fVerbose)
196	LogRel(("NAT: passing domain name %s\n", szDnsDomain));
197	}
198	else
199	Log(("nat: ignoring domain %s\n", szDnsDomain));
200	}
201	}
202	else
203	Log(("nat: GetComputerNameEx failed (%d)\n", GetLastError()));
204	}
205	}
206	return rc;
207	}
208
209	#else
210
211	static int get_dns_addr_domain(PNATState pData, bool fVerbose,
212	struct in_addr *pdns_addr,
213	const char **ppszDomain)
214	{
215	char buff[512];
216	char buff2[256];
217	FILE *f;
218	int found = 0;
219	struct in_addr tmp_addr;
220
221	#ifdef RT_OS_OS2
222	/* Try various locations. */
223	char *etc = getenv("ETC");
224	f = NULL;
225	if (etc)
226	{
227	snprintf(buff, sizeof(buff), "%s/RESOLV2", etc);
228	f = fopen(buff, "rt");
229	}
230	if (!f)
231	{
232	snprintf(buff, sizeof(buff), "%s/RESOLV2", _PATH_ETC);
233	f = fopen(buff, "rt");
234	}
235	if (!f)
236	{
237	snprintf(buff, sizeof(buff), "%s/resolv.conf", _PATH_ETC);
238	f = fopen(buff, "rt");
239	}
240	#else
241	f = fopen("/etc/resolv.conf", "r");
242	#endif
243	if (!f)
244	return -1;
245
246	if (ppszDomain)
247	*ppszDomain = NULL;
248	Log(("nat: DNS Servers:\n"));
249	while (fgets(buff, 512, f) != NULL)
250	{
251	if (sscanf(buff, "nameserver%*[ \t]%256s", buff2) == 1)
252	{
253	if (!inet_aton(buff2, &tmp_addr))
254	continue;
255	if (tmp_addr.s_addr == loopback_addr.s_addr)
256	tmp_addr = our_addr;
257	/* If it's the first one, set it to dns_addr */
258	if (!found)
259	{
260	if (fVerbose \|\| pdns_addr->s_addr != tmp_addr.s_addr)
261	LogRel(("NAT: DNS address: %s\n", buff2));
262	*pdns_addr = tmp_addr;
263	}
264	else
265	{
266	if (fVerbose)
267	LogRel(("NAT: ignored DNS address: %s\n", buff2));
268	}
269	found++;
270	}
271	if ( ppszDomain
272	&& (!strncmp(buff, "domain", 6) \|\| !strncmp(buff, "search", 6)))
273	{
274	/* Domain name/search list present. Pick first entry */
275	if (*ppszDomain == NULL)
276	{
277	char *tok;
278	char *saveptr;
279	tok = strtok_r(&buff[6], " \t\n", &saveptr);
280	if (tok)
281	{
282	*ppszDomain = RTStrDup(tok);
283	if (pData->fPassDomain)
284	{
285	if (fVerbose)
286	LogRel(("NAT: passing domain name %s\n", tok));
287	}
288	else
289	Log(("nat: ignoring domain %s\n", tok));
290	}
291	}
292	}
293	}
294	fclose(f);
295	if (!found)
296	return -1;
297	return 0;
298	}
299
300	#endif
301
302	int get_dns_addr(PNATState pData, struct in_addr *pdns_addr)
303	{
304	return get_dns_addr_domain(pData, false, pdns_addr, NULL);
305	}
306
307	int slirp_init(PNATState ppData, const char pszNetAddr, uint32_t u32Netmask,
308	bool fPassDomain, const char *pszTFTPPrefix,
309	const char pszBootFile, void pvUser)
310	{
311	int fNATfailed = 0;
312	PNATState pData = malloc(sizeof(NATState));
313	*ppData = pData;
314	if (!pData)
315	return VERR_NO_MEMORY;
316	if (u32Netmask & 0x1f)
317	/* CTL is x.x.x.15, bootp passes up to 16 IPs (15..31) */
318	return VERR_INVALID_PARAMETER;
319	memset(pData, '\0', sizeof(NATState));
320	pData->fPassDomain = fPassDomain;
321	pData->pvUser = pvUser;
322	#if ARCH_BITS == 64 && !defined(VBOX_WITH_BSD_REASS)
323	pData->cpvHashUsed = 1;
324	#endif
325	tftp_prefix = pszTFTPPrefix;
326	bootp_filename = pszBootFile;
327	pData->netmask = u32Netmask;
328
329	#ifdef RT_OS_WINDOWS
330	{
331	WSADATA Data;
332	WSAStartup(MAKEWORD(2,0), &Data);
333	}
334	# if defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC)
335	pData->phEvents[VBOX_SOCKET_EVENT_INDEX] = CreateEvent(NULL, FALSE, FALSE, NULL);
336	# endif
337	#endif
338
339	link_up = 1;
340
341	debug_init();
342	if_init(pData);
343	ip_init(pData);
344	#ifdef VBOX_WITH_SLIRP_ICMP
345	icmp_init(pData);
346	#endif /* VBOX_WITH_SLIRP_ICMP */
347
348	/* Initialise mbufs after setting the MTU */
349	m_init(pData);
350
351	/* set default addresses */
352	inet_aton("127.0.0.1", &loopback_addr);
353	inet_aton("127.0.0.1", &dns_addr);
354
355	if (get_dns_addr_domain(pData, true, &dns_addr, &pData->pszDomain) < 0)
356	fNATfailed = 1;
357
358	inet_aton(pszNetAddr, &special_addr);
359	alias_addr.s_addr = special_addr.s_addr \| htonl(CTL_ALIAS);
360	getouraddr(pData);
361	return fNATfailed ? VINF_NAT_DNS : VINF_SUCCESS;
362	}
363
364	/**
365	* Statistics counters.
366	*/
367	void slirp_register_timers(PNATState pData, PPDMDRVINS pDrvIns)
368	{
369	#ifndef VBOX_WITHOUT_RELEASE_STATISTICS
370	PDMDrvHlpSTAMRegisterF(pDrvIns, &pData->StatFill, STAMTYPE_PROFILE, STAMVISIBILITY_ALWAYS,
371	STAMUNIT_TICKS_PER_CALL, "Profiling slirp fills", "/Drivers/NAT%d/Fill", pDrvIns->iInstance);
372	PDMDrvHlpSTAMRegisterF(pDrvIns, &pData->StatPoll, STAMTYPE_PROFILE, STAMVISIBILITY_ALWAYS,
373	STAMUNIT_TICKS_PER_CALL, "Profiling slirp polls", "/Drivers/NAT%d/Poll", pDrvIns->iInstance);
374	PDMDrvHlpSTAMRegisterF(pDrvIns, &pData->StatFastTimer, STAMTYPE_PROFILE, STAMVISIBILITY_ALWAYS,
375	STAMUNIT_TICKS_PER_CALL, "Profiling slirp fast timer", "/Drivers/NAT%d/TimerFast", pDrvIns->iInstance);
376	PDMDrvHlpSTAMRegisterF(pDrvIns, &pData->StatSlowTimer, STAMTYPE_PROFILE, STAMVISIBILITY_ALWAYS,
377	STAMUNIT_TICKS_PER_CALL, "Profiling slirp slow timer", "/Drivers/NAT%d/TimerSlow", pDrvIns->iInstance);
378	PDMDrvHlpSTAMRegisterF(pDrvIns, &pData->StatTCP, STAMTYPE_COUNTER, STAMVISIBILITY_ALWAYS,
379	STAMUNIT_COUNT, "TCP sockets", "/Drivers/NAT%d/SockTCP", pDrvIns->iInstance);
380	PDMDrvHlpSTAMRegisterF(pDrvIns, &pData->StatTCPHot, STAMTYPE_COUNTER, STAMVISIBILITY_ALWAYS,
381	STAMUNIT_COUNT, "TCP sockets active", "/Drivers/NAT%d/SockTCPHot", pDrvIns->iInstance);
382	PDMDrvHlpSTAMRegisterF(pDrvIns, &pData->StatUDP, STAMTYPE_COUNTER, STAMVISIBILITY_ALWAYS,
383	STAMUNIT_COUNT, "UDP sockets", "/Drivers/NAT%d/SockUDP", pDrvIns->iInstance);
384	PDMDrvHlpSTAMRegisterF(pDrvIns, &pData->StatUDPHot, STAMTYPE_COUNTER, STAMVISIBILITY_ALWAYS,
385	STAMUNIT_COUNT, "UDP sockets active", "/Drivers/NAT%d/SockUDPHot", pDrvIns->iInstance);
386	#endif /* VBOX_WITHOUT_RELEASE_STATISTICS */
387	}
388
389	/**
390	* Marks the link as up, making it possible to establish new connections.
391	*/
392	void slirp_link_up(PNATState pData)
393	{
394	link_up = 1;
395	}
396
397	/**
398	* Marks the link as down and cleans up the current connections.
399	*/
400	void slirp_link_down(PNATState pData)
401	{
402	struct socket *so;
403
404	while ((so = tcb.so_next) != &tcb)
405	{
406	if (so->so_state & SS_NOFDREF \|\| so->s == -1)
407	sofree(pData, so);
408	else
409	tcp_drop(pData, sototcpcb(so), 0);
410	}
411
412	while ((so = udb.so_next) != &udb)
413	udp_detach(pData, so);
414
415	link_up = 0;
416	}
417
418	/**
419	* Terminates the slirp component.
420	*/
421	void slirp_term(PNATState pData)
422	{
423	if (pData->pszDomain)
424	RTStrFree((char )(void )pData->pszDomain);
425
426	#if ARCH_BITS == 64 && !defined(VBOX_WITH_BSD_REASS)
427	LogRel(("NAT: cpvHashUsed=%RU32 cpvHashCollisions=%RU32 cpvHashInserts=%RU64 cpvHashDone=%RU64\n",
428	pData->cpvHashUsed, pData->cpvHashCollisions, pData->cpvHashInserts, pData->cpvHashDone));
429	#endif
430
431	#ifdef VBOX_WITH_SLIRP_ICMP
432	# ifdef RT_OS_WINDOWS
433	pData->pfIcmpCloseHandle(pData->icmp_socket.sh);
434	# else
435	closesocket(pData->icmp_socket.s);
436	# endif
437	#endif
438
439	slirp_link_down(pData);
440	#ifdef RT_OS_WINDOWS
441	WSACleanup();
442	#endif
443	#ifdef LOG_ENABLED
444	Log(("\n"
445	"NAT statistics\n"
446	"--------------\n"
447	"\n"));
448	ipstats(pData);
449	tcpstats(pData);
450	udpstats(pData);
451	icmpstats(pData);
452	mbufstats(pData);
453	sockstats(pData);
454	Log(("\n"
455	"\n"
456	"\n"));
457	#endif
458	free(pData);
459	}
460
461
462	#define CONN_CANFSEND(so) (((so)->so_state & (SS_FCANTSENDMORE\|SS_ISFCONNECTED)) == SS_ISFCONNECTED)
463	#define CONN_CANFRCV(so) (((so)->so_state & (SS_FCANTRCVMORE\|SS_ISFCONNECTED)) == SS_ISFCONNECTED)
464	#define UPD_NFDS(x) if (nfds < (x)) nfds = (x)
465
466	/*
467	* curtime kept to an accuracy of 1ms
468	*/
469	#ifdef RT_OS_WINDOWS
470	static void updtime(PNATState pData)
471	{
472	struct _timeb tb;
473
474	_ftime(&tb);
475	curtime = (u_int)tb.time * (u_int)1000;
476	curtime += (u_int)tb.millitm;
477	}
478	#else
479	static void updtime(PNATState pData)
480	{
481	gettimeofday(&tt, 0);
482
483	curtime = (u_int)tt.tv_sec * (u_int)1000;
484	curtime += (u_int)tt.tv_usec / (u_int)1000;
485
486	if ((tt.tv_usec % 1000) >= 500)
487	curtime++;
488	}
489	#endif
490
491	void slirp_select_fill(PNATState pData, int *pnfds,
492	fd_set readfds, fd_set writefds, fd_set *xfds)
493	{
494	struct socket so, so_next;
495	int nfds;
496	#if defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) && defined(RT_OS_WINDOWS)
497	int rc;
498	int error;
499	#endif
500	#ifdef VBOX_WITH_BSD_REASS
501	int i;
502	#endif /* VBOX_WITH_BSD_REASS */
503
504	STAM_REL_PROFILE_START(&pData->StatFill, a);
505
506	nfds = *pnfds;
507
508	/*
509	* First, TCP sockets
510	*/
511	do_slowtimo = 0;
512	if (link_up)
513	{
514	/*
515	* *_slowtimo needs calling if there are IP fragments
516	* in the fragment queue, or there are TCP connections active
517	*/
518	#ifndef VBOX_WITH_BSD_REASS
519	do_slowtimo = ((tcb.so_next != &tcb)
520	\|\| ((struct ipasfrag )&ipq != u32_to_ptr(pData, ipq.next, struct ipasfrag )));
521	#else /* !VBOX_WITH_BSD_REASS */
522	/* XXX:
523	* triggering of fragment expiration should be the same but use new macroses
524	*/
525	do_slowtimo = (tcb.so_next != &tcb);
526	if (!do_slowtimo)
527	{
528	for (i = 0; i < IPREASS_NHASH; i++)
529	{
530	if (!TAILQ_EMPTY(&ipq[i]))
531	{
532	do_slowtimo = 1;
533	break;
534	}
535	}
536	}
537	#endif /* VBOX_WITH_BSD_REASS */
538
539	STAM_REL_COUNTER_RESET(&pData->StatTCP);
540	STAM_REL_COUNTER_RESET(&pData->StatTCPHot);
541
542	for (so = tcb.so_next; so != &tcb; so = so_next)
543	{
544	so_next = so->so_next;
545
546	STAM_REL_COUNTER_INC(&pData->StatTCP);
547
548	/*
549	* See if we need a tcp_fasttimo
550	*/
551	if (time_fasttimo == 0 && so->so_tcpcb->t_flags & TF_DELACK)
552	time_fasttimo = curtime; /* Flag when we want a fasttimo */
553
554	/*
555	* NOFDREF can include still connecting to local-host,
556	* newly socreated() sockets etc. Don't want to select these.
557	*/
558	if (so->so_state & SS_NOFDREF \|\| so->s == -1)
559	continue;
560
561	/*
562	* Set for reading sockets which are accepting
563	*/
564	if (so->so_state & SS_FACCEPTCONN)
565	{
566	STAM_REL_COUNTER_INC(&pData->StatTCPHot);
567	TCP_ENGAGE_EVENT1(so, readfds);
568	continue;
569	}
570
571	/*
572	* Set for writing sockets which are connecting
573	*/
574	if (so->so_state & SS_ISFCONNECTING)
575	{
576	STAM_REL_COUNTER_INC(&pData->StatTCPHot);
577	TCP_ENGAGE_EVENT1(so, writefds);
578	}
579
580	/*
581	* Set for writing if we are connected, can send more, and
582	* we have something to send
583	*/
584	if (CONN_CANFSEND(so) && so->so_rcv.sb_cc)
585	{
586	STAM_REL_COUNTER_INC(&pData->StatTCPHot);
587	TCP_ENGAGE_EVENT1(so, writefds);
588	}
589
590	/*
591	* Set for reading (and urgent data) if we are connected, can
592	* receive more, and we have room for it XXX /2 ?
593	*/
594	if (CONN_CANFRCV(so) && (so->so_snd.sb_cc < (so->so_snd.sb_datalen/2)))
595	{
596	STAM_REL_COUNTER_INC(&pData->StatTCPHot);
597	TCP_ENGAGE_EVENT2(so, readfds, xfds);
598	}
599	}
600
601	/*
602	* UDP sockets
603	*/
604	STAM_REL_COUNTER_RESET(&pData->StatUDP);
605	STAM_REL_COUNTER_RESET(&pData->StatUDPHot);
606
607	for (so = udb.so_next; so != &udb; so = so_next)
608	{
609	so_next = so->so_next;
610
611	STAM_REL_COUNTER_INC(&pData->StatUDP);
612
613	/*
614	* See if it's timed out
615	*/
616	if (so->so_expire)
617	{
618	if (so->so_expire <= curtime)
619	{
620	udp_detach(pData, so);
621	continue;
622	}
623	else
624	do_slowtimo = 1; /* Let socket expire */
625	}
626
627	/*
628	* When UDP packets are received from over the link, they're
629	* sendto()'d straight away, so no need for setting for writing
630	* Limit the number of packets queued by this session to 4.
631	* Note that even though we try and limit this to 4 packets,
632	* the session could have more queued if the packets needed
633	* to be fragmented.
634	*
635	* (XXX <= 4 ?)
636	*/
637	if ((so->so_state & SS_ISFCONNECTED) && so->so_queued <= 4)
638	{
639	STAM_REL_COUNTER_INC(&pData->StatUDPHot);
640	UDP_ENGAGE_EVENT(so, readfds);
641	}
642	}
643
644	ICMP_ENGAGE_EVENT(&pData->icmp_socket, readfds);
645	}
646
647	#if !defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) \|\| !defined(RT_OS_WINDOWS)
648	*pnfds = nfds;
649	#else
650	*pnfds = VBOX_EVENT_COUNT;
651	#endif
652
653	STAM_REL_PROFILE_STOP(&pData->StatFill, a);
654	}
655
656	void slirp_select_poll(PNATState pData, fd_set readfds, fd_set writefds, fd_set *xfds)
657	{
658	struct socket so, so_next;
659	int ret;
660	#if defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) && defined(RT_OS_WINDOWS)
661	WSANETWORKEVENTS NetworkEvents;
662	int rc;
663	int error;
664	#endif
665	static uint32_t stat_time;
666
667	STAM_REL_PROFILE_START(&pData->StatPoll, a);
668
669	/* Update time */
670	updtime(pData);
671
672	#ifdef LOG_ENABLED
673	if (curtime - stat_time > 10000)
674	{
675	stat_time = curtime;
676	sockstats(pData);
677	}
678	#endif
679
680	/*
681	* See if anything has timed out
682	*/
683	if (link_up)
684	{
685	if (time_fasttimo && ((curtime - time_fasttimo) >= 2))
686	{
687	STAM_REL_PROFILE_START(&pData->StatFastTimer, a);
688	tcp_fasttimo(pData);
689	time_fasttimo = 0;
690	STAM_REL_PROFILE_STOP(&pData->StatFastTimer, a);
691	}
692	if (do_slowtimo && ((curtime - last_slowtimo) >= 499))
693	{
694	STAM_REL_PROFILE_START(&pData->StatSlowTimer, a);
695	ip_slowtimo(pData);
696	tcp_slowtimo(pData);
697	last_slowtimo = curtime;
698	STAM_REL_PROFILE_STOP(&pData->StatSlowTimer, a);
699	}
700	}
701	#if defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) && defined(RT_OS_WINDOWS)
702	if (!readfds && !writefds && !xfds)
703	return; /* only timer update */
704	#endif
705
706	/*
707	* Check sockets
708	*/
709	if (link_up)
710	{
711	/*
712	* Check TCP sockets
713	*/
714	for (so = tcb.so_next; so != &tcb; so = so_next)
715	{
716	so_next = so->so_next;
717
718	/*
719	* FD_ISSET is meaningless on these sockets
720	* (and they can crash the program)
721	*/
722	if (so->so_state & SS_NOFDREF \|\| so->s == -1)
723	continue;
724
725	POLL_TCP_EVENTS(rc, error, so, &NetworkEvents);
726
727	/*
728	* Check for URG data
729	* This will soread as well, so no need to
730	* test for readfds below if this succeeds
731	*/
732
733	/* out-of-band data */
734	if (CHECK_FD_SET(so, NetworkEvents, xfds))
735	{
736	sorecvoob(pData, so);
737	}
738
739	/*
740	* Check sockets for reading
741	*/
742	else if (CHECK_FD_SET(so, NetworkEvents, readfds))
743	{
744	/*
745	* Check for incoming connections
746	*/
747	if (so->so_state & SS_FACCEPTCONN)
748	{
749	tcp_connect(pData, so);
750	#if defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) && defined(RT_OS_WINDOWS)
751	if (!(NetworkEvents.lNetworkEvents & FD_CLOSE))
752	#endif
753	continue;
754	}
755
756	ret = soread(pData, so, /fCloseIfNothingRead=/false);
757	/* Output it if we read something */
758	if (ret > 0)
759	tcp_output(pData, sototcpcb(so));
760	}
761
762	#if defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) && defined(RT_OS_WINDOWS)
763	/*
764	* Check for FD_CLOSE events.
765	*/
766	if (NetworkEvents.lNetworkEvents & FD_CLOSE)
767	{
768	/*
769	* drain the socket
770	*/
771	for (;;)
772	{
773	ret = soread(pData, so, /fCloseIfNothingRead=/true);
774	if (ret > 0)
775	tcp_output(pData, sototcpcb(so));
776	else
777	break;
778	}
779	}
780	#endif
781
782	/*
783	* Check sockets for writing
784	*/
785	if (CHECK_FD_SET(so, NetworkEvents, writefds))
786	{
787	/*
788	* Check for non-blocking, still-connecting sockets
789	*/
790	if (so->so_state & SS_ISFCONNECTING)
791	{
792	/* Connected */
793	so->so_state &= ~SS_ISFCONNECTING;
794
795	/*
796	* This should be probably guarded by PROBE_CONN too. Anyway,
797	* we disable it on OS/2 because the below send call returns
798	* EFAULT which causes the opened TCP socket to close right
799	* after it has been opened and connected.
800	*/
801	#ifndef RT_OS_OS2
802	ret = send(so->s, (const char *)&ret, 0, 0);
803	if (ret < 0)
804	{
805	/* XXXXX Must fix, zero bytes is a NOP */
806	if ( errno == EAGAIN
807	\|\| errno == EWOULDBLOCK
808	\|\| errno == EINPROGRESS
809	\|\| errno == ENOTCONN)
810	continue;
811
812	/* else failed */
813	so->so_state = SS_NOFDREF;
814	}
815	/* else so->so_state &= ~SS_ISFCONNECTING; */
816	#endif
817
818	/*
819	* Continue tcp_input
820	*/
821	tcp_input(pData, (struct mbuf *)NULL, sizeof(struct ip), so);
822	/* continue; */
823	}
824	else
825	ret = sowrite(pData, so);
826	/*
827	* XXX If we wrote something (a lot), there could be the need
828	* for a window update. In the worst case, the remote will send
829	* a window probe to get things going again.
830	*/
831	}
832
833	/*
834	* Probe a still-connecting, non-blocking socket
835	* to check if it's still alive
836	*/
837	#ifdef PROBE_CONN
838	if (so->so_state & SS_ISFCONNECTING)
839	{
840	ret = recv(so->s, (char *)&ret, 0, 0);
841
842	if (ret < 0)
843	{
844	/* XXX */
845	if ( errno == EAGAIN
846	\|\| errno == EWOULDBLOCK
847	\|\| errno == EINPROGRESS
848	\|\| errno == ENOTCONN)
849	{
850	continue; /* Still connecting, continue */
851	}
852
853	/* else failed */
854	so->so_state = SS_NOFDREF;
855
856	/* tcp_input will take care of it */
857	}
858	else
859	{
860	ret = send(so->s, &ret, 0, 0);
861	if (ret < 0)
862	{
863	/* XXX */
864	if ( errno == EAGAIN
865	\|\| errno == EWOULDBLOCK
866	\|\| errno == EINPROGRESS
867	\|\| errno == ENOTCONN)
868	{
869	continue;
870	}
871	/* else failed */
872	so->so_state = SS_NOFDREF;
873	}
874	else
875	so->so_state &= ~SS_ISFCONNECTING;
876
877	}
878	tcp_input((struct mbuf *)NULL, sizeof(struct ip),so);
879	} /* SS_ISFCONNECTING */
880	#endif
881	}
882
883	/*
884	* Now UDP sockets.
885	* Incoming packets are sent straight away, they're not buffered.
886	* Incoming UDP data isn't buffered either.
887	*/
888	for (so = udb.so_next; so != &udb; so = so_next)
889	{
890	so_next = so->so_next;
891
892	POLL_UDP_EVENTS(rc, error, so, &NetworkEvents);
893
894	if (so->s != -1 && CHECK_FD_SET(so, NetworkEvents, readfds))
895	{
896	sorecvfrom(pData, so);
897	}
898	}
899
900	#if defined(VBOX_WITH_SLIRP_ICMP)
901	# if defined(RT_OS_WINDOWS)
902	sorecvfrom(pData, &pData->icmp_socket);
903	# else
904	if (pData->icmp_socket.s != -1 && FD_ISSET(pData->icmp_socket.s, readfds))
905	sorecvfrom(pData, &pData->icmp_socket);
906	# endif
907	#endif
908	}
909
910	/*
911	* See if we can start outputting
912	*/
913	if (if_queued && link_up)
914	if_start(pData);
915
916	STAM_REL_PROFILE_STOP(&pData->StatPoll, a);
917	}
918
919	#define ETH_ALEN 6
920	#define ETH_HLEN 14
921
922	#define ETH_P_IP 0x0800 /* Internet Protocol packet */
923	#define ETH_P_ARP 0x0806 /* Address Resolution packet */
924
925	#define ARPOP_REQUEST 1 /* ARP request */
926	#define ARPOP_REPLY 2 /* ARP reply */
927
928	struct ethhdr
929	{
930	unsigned char h_dest[ETH_ALEN]; /* destination eth addr */
931	unsigned char h_source[ETH_ALEN]; /* source ether addr */
932	unsigned short h_proto; /* packet type ID field */
933	};
934
935	struct arphdr
936	{
937	unsigned short ar_hrd; /* format of hardware address */
938	unsigned short ar_pro; /* format of protocol address */
939	unsigned char ar_hln; /* length of hardware address */
940	unsigned char ar_pln; /* length of protocol address */
941	unsigned short ar_op; /* ARP opcode (command) */
942
943	/*
944	* Ethernet looks like this : This bit is variable sized however...
945	*/
946	unsigned char ar_sha[ETH_ALEN]; /* sender hardware address */
947	unsigned char ar_sip[4]; /* sender IP address */
948	unsigned char ar_tha[ETH_ALEN]; /* target hardware address */
949	unsigned char ar_tip[4]; /* target IP address */
950	};
951
952	static
953	void arp_input(PNATState pData, const uint8_t *pkt, int pkt_len)
954	{
955	struct ethhdr eh = (struct ethhdr )pkt;
956	struct arphdr ah = (struct arphdr )(pkt + ETH_HLEN);
957	uint8_t arp_reply[ETH_HLEN + sizeof(struct arphdr)];
958	struct ethhdr reh = (struct ethhdr )arp_reply;
959	struct arphdr rah = (struct arphdr )(arp_reply + ETH_HLEN);
960	int ar_op;
961	struct ex_list *ex_ptr;
962	uint32_t htip = ntohl((uint32_t)ah->ar_tip);
963
964	ar_op = ntohs(ah->ar_op);
965	switch(ar_op)
966	{
967	case ARPOP_REQUEST:
968	if ((htip & pData->netmask) == ntohl(special_addr.s_addr))
969	{
970	if ( (htip & ~pData->netmask) == CTL_DNS
971	\|\| (htip & ~pData->netmask) == CTL_ALIAS)
972	goto arp_ok;
973	for (ex_ptr = exec_list; ex_ptr; ex_ptr = ex_ptr->ex_next)
974	{
975	if ((htip & ~pData->netmask) == ex_ptr->ex_addr)
976	goto arp_ok;
977	}
978	return;
979	arp_ok:
980	/* XXX: make an ARP request to have the client address */
981	memcpy(client_ethaddr, eh->h_source, ETH_ALEN);
982
983	/* ARP request for alias/dns mac address */
984	memcpy(reh->h_dest, pkt + ETH_ALEN, ETH_ALEN);
985	memcpy(reh->h_source, special_ethaddr, ETH_ALEN - 1);
986	reh->h_source[5] = ah->ar_tip[3];
987	reh->h_proto = htons(ETH_P_ARP);
988
989	rah->ar_hrd = htons(1);
990	rah->ar_pro = htons(ETH_P_IP);
991	rah->ar_hln = ETH_ALEN;
992	rah->ar_pln = 4;
993	rah->ar_op = htons(ARPOP_REPLY);
994	memcpy(rah->ar_sha, reh->h_source, ETH_ALEN);
995	memcpy(rah->ar_sip, ah->ar_tip, 4);
996	memcpy(rah->ar_tha, ah->ar_sha, ETH_ALEN);
997	memcpy(rah->ar_tip, ah->ar_sip, 4);
998	slirp_output(pData->pvUser, arp_reply, sizeof(arp_reply));
999	}
1000	break;
1001	default:
1002	break;
1003	}
1004	}
1005
1006	void slirp_input(PNATState pData, const uint8_t *pkt, int pkt_len)
1007	{
1008	struct mbuf *m;
1009	int proto;
1010
1011	if (pkt_len < ETH_HLEN)
1012	return;
1013
1014	proto = ntohs((uint16_t )(pkt + 12));
1015	switch(proto)
1016	{
1017	case ETH_P_ARP:
1018	arp_input(pData, pkt, pkt_len);
1019	break;
1020	case ETH_P_IP:
1021	/* Update time. Important if the network is very quiet, as otherwise
1022	* the first outgoing connection gets an incorrect timestamp. */
1023	updtime(pData);
1024
1025	m = m_get(pData);
1026	if (!m)
1027	return;
1028	/* Note: we add to align the IP header */
1029	if (M_FREEROOM(m) < pkt_len + 2)
1030	{
1031	m_inc(m, pkt_len + 2);
1032	}
1033	m->m_len = pkt_len + 2;
1034	memcpy(m->m_data + 2, pkt, pkt_len);
1035
1036	m->m_data += 2 + ETH_HLEN;
1037	m->m_len -= 2 + ETH_HLEN;
1038
1039	ip_input(pData, m);
1040	break;
1041	default:
1042	break;
1043	}
1044	}
1045
1046	/* output the IP packet to the ethernet device */
1047	void if_encap(PNATState pData, const uint8_t *ip_data, int ip_data_len)
1048	{
1049	uint8_t buf[1600];
1050	struct ethhdr eh = (struct ethhdr )buf;
1051
1052	if (ip_data_len + ETH_HLEN > sizeof(buf))
1053	return;
1054
1055	memcpy(eh->h_dest, client_ethaddr, ETH_ALEN);
1056	memcpy(eh->h_source, special_ethaddr, ETH_ALEN - 1);
1057	/* XXX: not correct */
1058	eh->h_source[5] = CTL_ALIAS;
1059	eh->h_proto = htons(ETH_P_IP);
1060	memcpy(buf + sizeof(struct ethhdr), ip_data, ip_data_len);
1061	slirp_output(pData->pvUser, buf, ip_data_len + ETH_HLEN);
1062	}
1063
1064	int slirp_redir(PNATState pData, int is_udp, int host_port,
1065	struct in_addr guest_addr, int guest_port)
1066	{
1067	if (is_udp)
1068	{
1069	if (!udp_listen(pData, htons(host_port), guest_addr.s_addr,
1070	htons(guest_port), 0))
1071	return -1;
1072	}
1073	else
1074	{
1075	if (!solisten(pData, htons(host_port), guest_addr.s_addr,
1076	htons(guest_port), 0))
1077	return -1;
1078	}
1079	return 0;
1080	}
1081
1082	int slirp_add_exec(PNATState pData, int do_pty, const char *args, int addr_low_byte,
1083	int guest_port)
1084	{
1085	return add_exec(&exec_list, do_pty, (char *)args,
1086	addr_low_byte, htons(guest_port));
1087	}
1088
1089	void slirp_set_ethaddr(PNATState pData, const uint8_t *ethaddr)
1090	{
1091	memcpy(client_ethaddr, ethaddr, ETH_ALEN);
1092	}
1093
1094	#if defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) && defined(RT_OS_WINDOWS)
1095	HANDLE *slirp_get_events(PNATState pData)
1096	{
1097	return pData->phEvents;
1098	}
1099	void slirp_register_external_event(PNATState pData, HANDLE hEvent, int index)
1100	{
1101	pData->phEvents[index] = hEvent;
1102	}
1103	#endif
1104
1105	unsigned int slirp_get_timeout_ms(PNATState pData)
1106	{
1107	if (link_up)
1108	{
1109	if (time_fasttimo)
1110	return 2;
1111	if (do_slowtimo)
1112	return 500; /* see PR_SLOWHZ */
1113	}
1114	return 0;
1115	}

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

source: vbox/trunk/src/VBox/Devices/Network/slirp/slirp.c@ 15243

Download in other formats: