slirp.c@ 15220

Last change on this file since 15220 was 15220, checked in by vboxsync, 16 years ago
slirp: readability
Property svn:eol-style set to `native`
File size: 34.1 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	# ifdef VBOX_WITH_SLIRP_ICMP
32	# define ICMP_ENGAGE_EVENT(so, fdset) \
33	do { \
34	if (pData->icmp_socket.s != -1) \
35	DO_ENGAGE_EVENT1((so), (fdset), ICMP); \
36	} while (0)
37	# else /* !VBOX_WITH_SLIRP_ICMP */
38	# define ICMP_ENGAGE_EVENT(so, fdset) do {} while (0)
39	# endif /* !VBOX_WITH_SLIRP_ICMP */
40
41	#else /* defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) && defined(RT_OS_WINDOWS) */
42
43	/*
44	* On Windows, we will be notified by IcmpSendEcho2() when the response arrives.
45	* So no call to WSAEventSelect necessary.
46	*/
47	# define ICMP_ENGAGE_EVENT(so, fdset) do {} while(0)
48
49	# define DO_ENGAGE_EVENT1(so, fdset1, label) \
50	do { \
51	rc = WSAEventSelect((so)->s, VBOX_SOCKET_EVENT, FD_ALL_EVENTS); \
52	if (rc == SOCKET_ERROR) \
53	{ \
54	/* This should not happen */ \
55	error = WSAGetLastError(); \
56	LogRel(("WSAEventSelector (" #label ") error %d (so=%x, socket=%s, event=%x)\n", \
57	error, (so), (so)->s, VBOX_SOCKET_EVENT)); \
58	} \
59	} while(0)
60
61	# define DO_ENGAGE_EVENT2(so, fdset1, fdset2, label) \
62	DO_ENGAGE_EVENT1((so), (fdset1), label)
63
64	# define DO_POLL_EVENTS(rc, error, so, events, label) \
65	(rc) = WSAEnumNetworkEvents((so)->s, VBOX_SOCKET_EVENT, (events)); \
66	if ((rc) == SOCKET_ERROR) \
67	{ \
68	(error) = WSAGetLastError(); \
69	LogRel(("WSAEnumNetworkEvents " #label " error %d\n", (error))); \
70	continue; \
71	}
72
73	#endif /* defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) && defined(RT_OS_WINDOWS) */
74
75	#define TCP_ENGAGE_EVENT1(so, fdset) \
76	DO_ENGAGE_EVENT1((so), (fdset), TCP)
77
78	#define TCP_ENGAGE_EVENT2(so, fdset1, fdset2) \
79	DO_ENGAGE_EVENT2((so), (fdset1), (fdset2), TCP)
80
81	#define UDP_ENGAGE_EVENT(so, fdset) \
82	DO_ENGAGE_EVENT1((so), (fdset), UDP)
83
84	#define POLL_TCP_EVENTS(rc, error, so, events) \
85	DO_POLL_EVENTS((rc), (error), (so), (events), TCP)
86
87	#define POLL_UDP_EVENTS(rc, error, so, events) \
88	DO_POLL_EVENTS((rc), (error), (so), (events), UDP)
89
90	static const uint8_t special_ethaddr[6] =
91	{
92	0x52, 0x54, 0x00, 0x12, 0x35, 0x00
93	};
94
95	#ifdef RT_OS_WINDOWS
96
97	static int get_dns_addr_domain(PNATState pData, bool fVerbose,
98	struct in_addr *pdns_addr,
99	const char **ppszDomain)
100	{
101	int rc = 0;
102	FIXED_INFO *FixedInfo = NULL;
103	ULONG BufLen;
104	DWORD ret;
105	IP_ADDR_STRING *pIPAddr;
106	struct in_addr tmp_addr;
107
108	FixedInfo = (FIXED_INFO *)GlobalAlloc(GPTR, sizeof(FIXED_INFO));
109	BufLen = sizeof(FIXED_INFO);
110
111	/** @todo: this API returns all DNS servers, no matter whether the
112	* corresponding network adapter is disabled or not. Maybe replace
113	* this by GetAdapterAddresses(), which is XP/Vista only though. */
114	if (ERROR_BUFFER_OVERFLOW == GetNetworkParams(FixedInfo, &BufLen))
115	{
116	if (FixedInfo)
117	{
118	GlobalFree(FixedInfo);
119	FixedInfo = NULL;
120	}
121	FixedInfo = GlobalAlloc(GPTR, BufLen);
122	}
123
124	if ((ret = GetNetworkParams(FixedInfo, &BufLen)) != ERROR_SUCCESS)
125	{
126	Log(("GetNetworkParams failed. ret = %08x\n", (u_int)ret ));
127	if (FixedInfo)
128	{
129	GlobalFree(FixedInfo);
130	FixedInfo = NULL;
131	}
132	rc = -1;
133	goto get_dns_prefix;
134	}
135
136	pIPAddr = &(FixedInfo->DnsServerList);
137	inet_aton(pIPAddr->IpAddress.String, &tmp_addr);
138	Log(("nat: DNS Servers:\n"));
139	if (fVerbose \|\| pdns_addr->s_addr != tmp_addr.s_addr)
140	LogRel(("NAT: DNS address: %s\n", pIPAddr->IpAddress.String));
141	*pdns_addr = tmp_addr;
142
143	pIPAddr = FixedInfo -> DnsServerList.Next;
144	while (pIPAddr)
145	{
146	if (fVerbose)
147	LogRel(("NAT: ignored DNS address: %s\n", pIPAddr ->IpAddress.String));
148	pIPAddr = pIPAddr ->Next;
149	}
150	if (FixedInfo)
151	{
152	GlobalFree(FixedInfo);
153	FixedInfo = NULL;
154	}
155
156	get_dns_prefix:
157	if (ppszDomain)
158	{
159	OSVERSIONINFO ver;
160	char szDnsDomain[256];
161	DWORD dwSize = sizeof(szDnsDomain);
162
163	*ppszDomain = NULL;
164	GetVersionEx(&ver);
165	if (ver.dwMajorVersion >= 5)
166	{
167	/* GetComputerNameEx exists in Windows versions starting with 2000. */
168	if (GetComputerNameEx(ComputerNameDnsDomain, szDnsDomain, &dwSize))
169	{
170	if (szDnsDomain[0])
171	{
172	/* Just non-empty strings are valid. */
173	*ppszDomain = RTStrDup(szDnsDomain);
174	if (pData->fPassDomain)
175	{
176	if (fVerbose)
177	LogRel(("NAT: passing domain name %s\n", szDnsDomain));
178	}
179	else
180	Log(("nat: ignoring domain %s\n", szDnsDomain));
181	}
182	}
183	else
184	Log(("nat: GetComputerNameEx failed (%d)\n", GetLastError()));
185	}
186	}
187	return rc;
188	}
189
190	#else
191
192	static int get_dns_addr_domain(PNATState pData, bool fVerbose,
193	struct in_addr *pdns_addr,
194	const char **ppszDomain)
195	{
196	char buff[512];
197	char buff2[256];
198	FILE *f;
199	int found = 0;
200	struct in_addr tmp_addr;
201
202	#ifdef RT_OS_OS2
203	/* Try various locations. */
204	char *etc = getenv("ETC");
205	f = NULL;
206	if (etc)
207	{
208	snprintf(buff, sizeof(buff), "%s/RESOLV2", etc);
209	f = fopen(buff, "rt");
210	}
211	if (!f)
212	{
213	snprintf(buff, sizeof(buff), "%s/RESOLV2", _PATH_ETC);
214	f = fopen(buff, "rt");
215	}
216	if (!f)
217	{
218	snprintf(buff, sizeof(buff), "%s/resolv.conf", _PATH_ETC);
219	f = fopen(buff, "rt");
220	}
221	#else
222	f = fopen("/etc/resolv.conf", "r");
223	#endif
224	if (!f)
225	return -1;
226
227	if (ppszDomain)
228	*ppszDomain = NULL;
229	Log(("nat: DNS Servers:\n"));
230	while (fgets(buff, 512, f) != NULL)
231	{
232	if (sscanf(buff, "nameserver%*[ \t]%256s", buff2) == 1)
233	{
234	if (!inet_aton(buff2, &tmp_addr))
235	continue;
236	if (tmp_addr.s_addr == loopback_addr.s_addr)
237	tmp_addr = our_addr;
238	/* If it's the first one, set it to dns_addr */
239	if (!found)
240	{
241	if (fVerbose \|\| pdns_addr->s_addr != tmp_addr.s_addr)
242	LogRel(("NAT: DNS address: %s\n", buff2));
243	*pdns_addr = tmp_addr;
244	}
245	else
246	{
247	if (fVerbose)
248	LogRel(("NAT: ignored DNS address: %s\n", buff2));
249	}
250	found++;
251	}
252	if ( ppszDomain
253	&& (!strncmp(buff, "domain", 6) \|\| !strncmp(buff, "search", 6)))
254	{
255	/* Domain name/search list present. Pick first entry */
256	if (*ppszDomain == NULL)
257	{
258	char *tok;
259	char *saveptr;
260	tok = strtok_r(&buff[6], " \t\n", &saveptr);
261	if (tok)
262	{
263	*ppszDomain = RTStrDup(tok);
264	if (pData->fPassDomain)
265	{
266	if (fVerbose)
267	LogRel(("NAT: passing domain name %s\n", tok));
268	}
269	else
270	Log(("nat: ignoring domain %s\n", tok));
271	}
272	}
273	}
274	}
275	fclose(f);
276	if (!found)
277	return -1;
278	return 0;
279	}
280
281	#endif
282
283	int get_dns_addr(PNATState pData, struct in_addr *pdns_addr)
284	{
285	return get_dns_addr_domain(pData, false, pdns_addr, NULL);
286	}
287
288	int slirp_init(PNATState ppData, const char pszNetAddr, uint32_t u32Netmask,
289	bool fPassDomain, const char *pszTFTPPrefix,
290	const char pszBootFile, void pvUser)
291	{
292	int fNATfailed = 0;
293	PNATState pData = malloc(sizeof(NATState));
294	*ppData = pData;
295	if (!pData)
296	return VERR_NO_MEMORY;
297	if (u32Netmask & 0x1f)
298	/* CTL is x.x.x.15, bootp passes up to 16 IPs (15..31) */
299	return VERR_INVALID_PARAMETER;
300	memset(pData, '\0', sizeof(NATState));
301	pData->fPassDomain = fPassDomain;
302	pData->pvUser = pvUser;
303	#if ARCH_BITS == 64 && !defined(VBOX_WITH_BSD_REASS)
304	pData->cpvHashUsed = 1;
305	#endif
306	tftp_prefix = pszTFTPPrefix;
307	bootp_filename = pszBootFile;
308	pData->netmask = u32Netmask;
309
310	#ifdef RT_OS_WINDOWS
311	{
312	WSADATA Data;
313	WSAStartup(MAKEWORD(2,0), &Data);
314	}
315	# if defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC)
316	pData->phEvents[VBOX_SOCKET_EVENT_INDEX] = CreateEvent(NULL, FALSE, FALSE, NULL);
317	# endif
318	#endif
319
320	link_up = 1;
321
322	debug_init();
323	if_init(pData);
324	ip_init(pData);
325	#ifdef VBOX_WITH_SLIRP_ICMP
326	icmp_init(pData);
327	#endif /* VBOX_WITH_SLIRP_ICMP */
328
329	/* Initialise mbufs after setting the MTU */
330	m_init(pData);
331
332	/* set default addresses */
333	inet_aton("127.0.0.1", &loopback_addr);
334	inet_aton("127.0.0.1", &dns_addr);
335
336	if (get_dns_addr_domain(pData, true, &dns_addr, &pData->pszDomain) < 0)
337	fNATfailed = 1;
338
339	inet_aton(pszNetAddr, &special_addr);
340	alias_addr.s_addr = special_addr.s_addr \| htonl(CTL_ALIAS);
341	getouraddr(pData);
342	return fNATfailed ? VINF_NAT_DNS : VINF_SUCCESS;
343	}
344
345	/**
346	* Statistics counters.
347	*/
348	void slirp_register_timers(PNATState pData, PPDMDRVINS pDrvIns)
349	{
350	#ifndef VBOX_WITHOUT_RELEASE_STATISTICS
351	PDMDrvHlpSTAMRegisterF(pDrvIns, &pData->StatFill, STAMTYPE_PROFILE, STAMVISIBILITY_ALWAYS,
352	STAMUNIT_TICKS_PER_CALL, "Profiling slirp fills", "/Drivers/NAT%d/Fill", pDrvIns->iInstance);
353	PDMDrvHlpSTAMRegisterF(pDrvIns, &pData->StatPoll, STAMTYPE_PROFILE, STAMVISIBILITY_ALWAYS,
354	STAMUNIT_TICKS_PER_CALL, "Profiling slirp polls", "/Drivers/NAT%d/Poll", pDrvIns->iInstance);
355	PDMDrvHlpSTAMRegisterF(pDrvIns, &pData->StatFastTimer, STAMTYPE_PROFILE, STAMVISIBILITY_ALWAYS,
356	STAMUNIT_TICKS_PER_CALL, "Profiling slirp fast timer", "/Drivers/NAT%d/TimerFast", pDrvIns->iInstance);
357	PDMDrvHlpSTAMRegisterF(pDrvIns, &pData->StatSlowTimer, STAMTYPE_PROFILE, STAMVISIBILITY_ALWAYS,
358	STAMUNIT_TICKS_PER_CALL, "Profiling slirp slow timer", "/Drivers/NAT%d/TimerSlow", pDrvIns->iInstance);
359	PDMDrvHlpSTAMRegisterF(pDrvIns, &pData->StatTCP, STAMTYPE_COUNTER, STAMVISIBILITY_ALWAYS,
360	STAMUNIT_COUNT, "TCP sockets", "/Drivers/NAT%d/SockTCP", pDrvIns->iInstance);
361	PDMDrvHlpSTAMRegisterF(pDrvIns, &pData->StatTCPHot, STAMTYPE_COUNTER, STAMVISIBILITY_ALWAYS,
362	STAMUNIT_COUNT, "TCP sockets active", "/Drivers/NAT%d/SockTCPHot", pDrvIns->iInstance);
363	PDMDrvHlpSTAMRegisterF(pDrvIns, &pData->StatUDP, STAMTYPE_COUNTER, STAMVISIBILITY_ALWAYS,
364	STAMUNIT_COUNT, "UDP sockets", "/Drivers/NAT%d/SockUDP", pDrvIns->iInstance);
365	PDMDrvHlpSTAMRegisterF(pDrvIns, &pData->StatUDPHot, STAMTYPE_COUNTER, STAMVISIBILITY_ALWAYS,
366	STAMUNIT_COUNT, "UDP sockets active", "/Drivers/NAT%d/SockUDPHot", pDrvIns->iInstance);
367	#endif /* VBOX_WITHOUT_RELEASE_STATISTICS */
368	}
369
370	/**
371	* Marks the link as up, making it possible to establish new connections.
372	*/
373	void slirp_link_up(PNATState pData)
374	{
375	link_up = 1;
376	}
377
378	/**
379	* Marks the link as down and cleans up the current connections.
380	*/
381	void slirp_link_down(PNATState pData)
382	{
383	struct socket *so;
384
385	while ((so = tcb.so_next) != &tcb)
386	{
387	if (so->so_state & SS_NOFDREF \|\| so->s == -1)
388	sofree(pData, so);
389	else
390	tcp_drop(pData, sototcpcb(so), 0);
391	}
392
393	while ((so = udb.so_next) != &udb)
394	udp_detach(pData, so);
395
396	link_up = 0;
397	}
398
399	/**
400	* Terminates the slirp component.
401	*/
402	void slirp_term(PNATState pData)
403	{
404	if (pData->pszDomain)
405	RTStrFree((char )(void )pData->pszDomain);
406
407	#if ARCH_BITS == 64 && !defined(VBOX_WITH_BSD_REASS)
408	LogRel(("NAT: cpvHashUsed=%RU32 cpvHashCollisions=%RU32 cpvHashInserts=%RU64 cpvHashDone=%RU64\n",
409	pData->cpvHashUsed, pData->cpvHashCollisions, pData->cpvHashInserts, pData->cpvHashDone));
410	#endif
411
412	#ifdef VBOX_WITH_SLIRP_ICMP
413	# ifdef RT_OS_WINDOWS
414	pData->pfIcmpCloseHandle(pData->icmp_socket.sh);
415	# else
416	closesocket(pData->icmp_socket.s);
417	# endif
418	#endif
419
420	slirp_link_down(pData);
421	#ifdef RT_OS_WINDOWS
422	WSACleanup();
423	#endif
424	#ifdef LOG_ENABLED
425	Log(("\n"
426	"NAT statistics\n"
427	"--------------\n"
428	"\n"));
429	ipstats(pData);
430	tcpstats(pData);
431	udpstats(pData);
432	icmpstats(pData);
433	mbufstats(pData);
434	sockstats(pData);
435	Log(("\n"
436	"\n"
437	"\n"));
438	#endif
439	free(pData);
440	}
441
442
443	#define CONN_CANFSEND(so) (((so)->so_state & (SS_FCANTSENDMORE\|SS_ISFCONNECTED)) == SS_ISFCONNECTED)
444	#define CONN_CANFRCV(so) (((so)->so_state & (SS_FCANTRCVMORE\|SS_ISFCONNECTED)) == SS_ISFCONNECTED)
445	#define UPD_NFDS(x) if (nfds < (x)) nfds = (x)
446
447	/*
448	* curtime kept to an accuracy of 1ms
449	*/
450	#ifdef RT_OS_WINDOWS
451	static void updtime(PNATState pData)
452	{
453	struct _timeb tb;
454
455	_ftime(&tb);
456	curtime = (u_int)tb.time * (u_int)1000;
457	curtime += (u_int)tb.millitm;
458	}
459	#else
460	static void updtime(PNATState pData)
461	{
462	gettimeofday(&tt, 0);
463
464	curtime = (u_int)tt.tv_sec * (u_int)1000;
465	curtime += (u_int)tt.tv_usec / (u_int)1000;
466
467	if ((tt.tv_usec % 1000) >= 500)
468	curtime++;
469	}
470	#endif
471
472	void slirp_select_fill(PNATState pData, int *pnfds,
473	fd_set readfds, fd_set writefds, fd_set *xfds)
474	{
475	struct socket so, so_next;
476	int nfds;
477	#if defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) && defined(RT_OS_WINDOWS)
478	int rc;
479	int error;
480	#endif
481	#ifdef VBOX_WITH_BSD_REASS
482	int i;
483	#endif /* VBOX_WITH_BSD_REASS */
484
485	STAM_REL_PROFILE_START(&pData->StatFill, a);
486
487	nfds = *pnfds;
488
489	/*
490	* First, TCP sockets
491	*/
492	do_slowtimo = 0;
493	if (link_up)
494	{
495	/*
496	* *_slowtimo needs calling if there are IP fragments
497	* in the fragment queue, or there are TCP connections active
498	*/
499	#ifndef VBOX_WITH_BSD_REASS
500	do_slowtimo = ((tcb.so_next != &tcb)
501	\|\| ((struct ipasfrag )&ipq != u32_to_ptr(pData, ipq.next, struct ipasfrag )));
502	#else /* !VBOX_WITH_BSD_REASS */
503	/* XXX:
504	* triggering of fragment expiration should be the same but use new macroses
505	*/
506	do_slowtimo = (tcb.so_next != &tcb);
507	if (!do_slowtimo)
508	{
509	for (i = 0; i < IPREASS_NHASH; i++)
510	{
511	if (!TAILQ_EMPTY(&ipq[i]))
512	{
513	do_slowtimo = 1;
514	break;
515	}
516	}
517	}
518	#endif /* VBOX_WITH_BSD_REASS */
519
520	STAM_REL_COUNTER_RESET(&pData->StatTCP);
521	STAM_REL_COUNTER_RESET(&pData->StatTCPHot);
522
523	for (so = tcb.so_next; so != &tcb; so = so_next)
524	{
525	so_next = so->so_next;
526
527	STAM_REL_COUNTER_INC(&pData->StatTCP);
528
529	/*
530	* See if we need a tcp_fasttimo
531	*/
532	if (time_fasttimo == 0 && so->so_tcpcb->t_flags & TF_DELACK)
533	time_fasttimo = curtime; /* Flag when we want a fasttimo */
534
535	/*
536	* NOFDREF can include still connecting to local-host,
537	* newly socreated() sockets etc. Don't want to select these.
538	*/
539	if (so->so_state & SS_NOFDREF \|\| so->s == -1)
540	continue;
541
542	/*
543	* Set for reading sockets which are accepting
544	*/
545	if (so->so_state & SS_FACCEPTCONN)
546	{
547	STAM_REL_COUNTER_INC(&pData->StatTCPHot);
548	TCP_ENGAGE_EVENT1(so, readfds);
549	continue;
550	}
551
552	/*
553	* Set for writing sockets which are connecting
554	*/
555	if (so->so_state & SS_ISFCONNECTING)
556	{
557	STAM_REL_COUNTER_INC(&pData->StatTCPHot);
558	TCP_ENGAGE_EVENT1(so, writefds);
559	}
560
561	/*
562	* Set for writing if we are connected, can send more, and
563	* we have something to send
564	*/
565	if (CONN_CANFSEND(so) && so->so_rcv.sb_cc)
566	{
567	STAM_REL_COUNTER_INC(&pData->StatTCPHot);
568	TCP_ENGAGE_EVENT1(so, writefds);
569	}
570
571	/*
572	* Set for reading (and urgent data) if we are connected, can
573	* receive more, and we have room for it XXX /2 ?
574	*/
575	if (CONN_CANFRCV(so) && (so->so_snd.sb_cc < (so->so_snd.sb_datalen/2)))
576	{
577	STAM_REL_COUNTER_INC(&pData->StatTCPHot);
578	TCP_ENGAGE_EVENT2(so, readfds, xfds);
579	}
580	}
581
582	/*
583	* UDP sockets
584	*/
585	STAM_REL_COUNTER_RESET(&pData->StatUDP);
586	STAM_REL_COUNTER_RESET(&pData->StatUDPHot);
587
588	for (so = udb.so_next; so != &udb; so = so_next)
589	{
590	so_next = so->so_next;
591
592	STAM_REL_COUNTER_INC(&pData->StatUDP);
593
594	/*
595	* See if it's timed out
596	*/
597	if (so->so_expire)
598	{
599	if (so->so_expire <= curtime)
600	{
601	udp_detach(pData, so);
602	continue;
603	}
604	else
605	do_slowtimo = 1; /* Let socket expire */
606	}
607
608	/*
609	* When UDP packets are received from over the link, they're
610	* sendto()'d straight away, so no need for setting for writing
611	* Limit the number of packets queued by this session to 4.
612	* Note that even though we try and limit this to 4 packets,
613	* the session could have more queued if the packets needed
614	* to be fragmented.
615	*
616	* (XXX <= 4 ?)
617	*/
618	if ((so->so_state & SS_ISFCONNECTED) && so->so_queued <= 4)
619	{
620	STAM_REL_COUNTER_INC(&pData->StatUDPHot);
621	UDP_ENGAGE_EVENT(so, readfds);
622	}
623	}
624
625	ICMP_ENGAGE_EVENT(&pData->icmp_socket, readfds);
626	}
627
628	#if !defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) \|\| !defined(RT_OS_WINDOWS)
629	*pnfds = nfds;
630	#else
631	*pnfds = VBOX_EVENT_COUNT;
632	#endif
633
634	STAM_REL_PROFILE_STOP(&pData->StatFill, a);
635	}
636
637	void slirp_select_poll(PNATState pData, fd_set readfds, fd_set writefds, fd_set *xfds)
638	{
639	struct socket so, so_next;
640	int ret;
641	#if defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) && defined(RT_OS_WINDOWS)
642	WSANETWORKEVENTS NetworkEvents;
643	int rc;
644	int error;
645	#endif
646	static uint32_t stat_time;
647
648	STAM_REL_PROFILE_START(&pData->StatPoll, a);
649
650	/* Update time */
651	updtime(pData);
652
653	#ifdef LOG_ENABLED
654	if (curtime - stat_time > 10000)
655	{
656	stat_time = curtime;
657	sockstats(pData);
658	}
659	#endif
660
661	/*
662	* See if anything has timed out
663	*/
664	if (link_up)
665	{
666	if (time_fasttimo && ((curtime - time_fasttimo) >= 2))
667	{
668	STAM_REL_PROFILE_START(&pData->StatFastTimer, a);
669	tcp_fasttimo(pData);
670	time_fasttimo = 0;
671	STAM_REL_PROFILE_STOP(&pData->StatFastTimer, a);
672	}
673	if (do_slowtimo && ((curtime - last_slowtimo) >= 499))
674	{
675	STAM_REL_PROFILE_START(&pData->StatSlowTimer, a);
676	ip_slowtimo(pData);
677	tcp_slowtimo(pData);
678	last_slowtimo = curtime;
679	STAM_REL_PROFILE_STOP(&pData->StatSlowTimer, a);
680	}
681	}
682	#if defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) && defined(RT_OS_WINDOWS)
683	if (!readfds && !writefds && !xfds)
684	return; /* only timer update */
685	#endif
686
687	/*
688	* Check sockets
689	*/
690	if (link_up)
691	{
692	/*
693	* Check TCP sockets
694	*/
695	for (so = tcb.so_next; so != &tcb; so = so_next)
696	{
697	so_next = so->so_next;
698
699	/*
700	* FD_ISSET is meaningless on these sockets
701	* (and they can crash the program)
702	*/
703	if (so->so_state & SS_NOFDREF \|\| so->s == -1)
704	continue;
705
706	POLL_TCP_EVENTS(rc, error, so, NetworkEvents);
707
708	/*
709	* Check for URG data
710	* This will soread as well, so no need to
711	* test for readfds below if this succeeds
712	*/
713	#if !defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) \|\| !defined(RT_OS_WINDOWS)
714	if (FD_ISSET(so->s, xfds))
715	#else
716	/* out-of-band data */
717	if ((NetworkEvents.lNetworkEvents & FD_OOB) && NetworkEvents.iErrorCode[FD_OOB_BIT] == 0)
718	#endif
719	{
720	sorecvoob(pData, so);
721	}
722
723	/*
724	* Check sockets for reading
725	*/
726	#if !defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) \|\| !defined(RT_OS_WINDOWS)
727	else if (FD_ISSET(so->s, readfds))
728	#else
729	else if ((NetworkEvents.lNetworkEvents & FD_READ) && (NetworkEvents.iErrorCode[FD_READ_BIT] == 0))
730	#endif
731	{
732	/*
733	* Check for incoming connections
734	*/
735	if (so->so_state & SS_FACCEPTCONN)
736	{
737	tcp_connect(pData, so);
738	#if defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) && defined(RT_OS_WINDOWS)
739	if (!(NetworkEvents.lNetworkEvents & FD_CLOSE))
740	#endif
741	continue;
742	}
743
744	ret = soread(pData, so);
745	/* Output it if we read something */
746	if (ret > 0)
747	tcp_output(pData, sototcpcb(so));
748	}
749
750	#if defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) && defined(RT_OS_WINDOWS)
751	/*
752	* Check for FD_CLOSE events.
753	*/
754	if (NetworkEvents.lNetworkEvents & FD_CLOSE)
755	{
756	/*
757	* drain the socket
758	*/
759	for (;;)
760	{
761	ret = soread(pData, so);
762	if (ret > 0)
763	tcp_output(pData, sototcpcb(so));
764	else
765	break;
766	}
767	}
768	#endif
769
770	/*
771	* Check sockets for writing
772	*/
773	#if !defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) \|\| !defined(RT_OS_WINDOWS)
774	if (FD_ISSET(so->s, writefds))
775	#else
776	if ((NetworkEvents.lNetworkEvents & FD_WRITE) && (NetworkEvents.iErrorCode[FD_WRITE_BIT] == 0))
777	#endif
778	{
779	/*
780	* Check for non-blocking, still-connecting sockets
781	*/
782	if (so->so_state & SS_ISFCONNECTING)
783	{
784	/* Connected */
785	so->so_state &= ~SS_ISFCONNECTING;
786
787	/*
788	* This should be probably guarded by PROBE_CONN too. Anyway,
789	* we disable it on OS/2 because the below send call returns
790	* EFAULT which causes the opened TCP socket to close right
791	* after it has been opened and connected.
792	*/
793	#ifndef RT_OS_OS2
794	ret = send(so->s, (const char *)&ret, 0, 0);
795	if (ret < 0)
796	{
797	/* XXXXX Must fix, zero bytes is a NOP */
798	if ( errno == EAGAIN
799	\|\| errno == EWOULDBLOCK
800	\|\| errno == EINPROGRESS
801	\|\| errno == ENOTCONN)
802	continue;
803
804	/* else failed */
805	so->so_state = SS_NOFDREF;
806	}
807	/* else so->so_state &= ~SS_ISFCONNECTING; */
808	#endif
809
810	/*
811	* Continue tcp_input
812	*/
813	tcp_input(pData, (struct mbuf *)NULL, sizeof(struct ip), so);
814	/* continue; */
815	}
816	else
817	ret = sowrite(pData, so);
818	/*
819	* XXX If we wrote something (a lot), there could be the need
820	* for a window update. In the worst case, the remote will send
821	* a window probe to get things going again.
822	*/
823	}
824
825	/*
826	* Probe a still-connecting, non-blocking socket
827	* to check if it's still alive
828	*/
829	#ifdef PROBE_CONN
830	if (so->so_state & SS_ISFCONNECTING)
831	{
832	ret = recv(so->s, (char *)&ret, 0, 0);
833
834	if (ret < 0)
835	{
836	/* XXX */
837	if ( errno == EAGAIN
838	\|\| errno == EWOULDBLOCK
839	\|\| errno == EINPROGRESS
840	\|\| errno == ENOTCONN)
841	{
842	continue; /* Still connecting, continue */
843	}
844
845	/* else failed */
846	so->so_state = SS_NOFDREF;
847
848	/* tcp_input will take care of it */
849	}
850	else
851	{
852	ret = send(so->s, &ret, 0, 0);
853	if (ret < 0)
854	{
855	/* XXX */
856	if ( errno == EAGAIN
857	\|\| errno == EWOULDBLOCK
858	\|\| errno == EINPROGRESS
859	\|\| errno == ENOTCONN)
860	{
861	continue;
862	}
863	/* else failed */
864	so->so_state = SS_NOFDREF;
865	}
866	else
867	so->so_state &= ~SS_ISFCONNECTING;
868
869	}
870	tcp_input((struct mbuf *)NULL, sizeof(struct ip),so);
871	} /* SS_ISFCONNECTING */
872	#endif
873	}
874
875	/*
876	* Now UDP sockets.
877	* Incoming packets are sent straight away, they're not buffered.
878	* Incoming UDP data isn't buffered either.
879	*/
880	for (so = udb.so_next; so != &udb; so = so_next)
881	{
882	so_next = so->so_next;
883
884	POLL_UDP_EVENTS(rc, error, so, NetworkEvents);
885
886	#if !defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) \|\| !defined(RT_OS_WINDOWS)
887	if (so->s != -1 && FD_ISSET(so->s, readfds))
888	#else
889	if (((NetworkEvents.lNetworkEvents & FD_READ) && (NetworkEvents.iErrorCode[FD_READ_BIT] == 0)))
890	#endif
891	{
892	sorecvfrom(pData, so);
893	}
894	}
895
896	#if defined(VBOX_WITH_SLIRP_ICMP)
897	# if defined(RT_OS_WINDOWS)
898	sorecvfrom(pData, &pData->icmp_socket);
899	# else
900	if (pData->icmp_socket.s != -1 && FD_ISSET(pData->icmp_socket.s, readfds))
901	sorecvfrom(pData, &pData->icmp_socket);
902	# endif
903	#endif
904	}
905
906	/*
907	* See if we can start outputting
908	*/
909	if (if_queued && link_up)
910	if_start(pData);
911
912	STAM_REL_PROFILE_STOP(&pData->StatPoll, a);
913	}
914
915	#define ETH_ALEN 6
916	#define ETH_HLEN 14
917
918	#define ETH_P_IP 0x0800 /* Internet Protocol packet */
919	#define ETH_P_ARP 0x0806 /* Address Resolution packet */
920
921	#define ARPOP_REQUEST 1 /* ARP request */
922	#define ARPOP_REPLY 2 /* ARP reply */
923
924	struct ethhdr
925	{
926	unsigned char h_dest[ETH_ALEN]; /* destination eth addr */
927	unsigned char h_source[ETH_ALEN]; /* source ether addr */
928	unsigned short h_proto; /* packet type ID field */
929	};
930
931	struct arphdr
932	{
933	unsigned short ar_hrd; /* format of hardware address */
934	unsigned short ar_pro; /* format of protocol address */
935	unsigned char ar_hln; /* length of hardware address */
936	unsigned char ar_pln; /* length of protocol address */
937	unsigned short ar_op; /* ARP opcode (command) */
938
939	/*
940	* Ethernet looks like this : This bit is variable sized however...
941	*/
942	unsigned char ar_sha[ETH_ALEN]; /* sender hardware address */
943	unsigned char ar_sip[4]; /* sender IP address */
944	unsigned char ar_tha[ETH_ALEN]; /* target hardware address */
945	unsigned char ar_tip[4]; /* target IP address */
946	};
947
948	static
949	void arp_input(PNATState pData, const uint8_t *pkt, int pkt_len)
950	{
951	struct ethhdr eh = (struct ethhdr )pkt;
952	struct arphdr ah = (struct arphdr )(pkt + ETH_HLEN);
953	uint8_t arp_reply[ETH_HLEN + sizeof(struct arphdr)];
954	struct ethhdr reh = (struct ethhdr )arp_reply;
955	struct arphdr rah = (struct arphdr )(arp_reply + ETH_HLEN);
956	int ar_op;
957	struct ex_list *ex_ptr;
958	uint32_t htip = ntohl((uint32_t)ah->ar_tip);
959
960	ar_op = ntohs(ah->ar_op);
961	switch(ar_op)
962	{
963	case ARPOP_REQUEST:
964	if ((htip & pData->netmask) == ntohl(special_addr.s_addr))
965	{
966	if ( (htip & ~pData->netmask) == CTL_DNS
967	\|\| (htip & ~pData->netmask) == CTL_ALIAS)
968	goto arp_ok;
969	for (ex_ptr = exec_list; ex_ptr; ex_ptr = ex_ptr->ex_next)
970	{
971	if ((htip & ~pData->netmask) == ex_ptr->ex_addr)
972	goto arp_ok;
973	}
974	return;
975	arp_ok:
976	/* XXX: make an ARP request to have the client address */
977	memcpy(client_ethaddr, eh->h_source, ETH_ALEN);
978
979	/* ARP request for alias/dns mac address */
980	memcpy(reh->h_dest, pkt + ETH_ALEN, ETH_ALEN);
981	memcpy(reh->h_source, special_ethaddr, ETH_ALEN - 1);
982	reh->h_source[5] = ah->ar_tip[3];
983	reh->h_proto = htons(ETH_P_ARP);
984
985	rah->ar_hrd = htons(1);
986	rah->ar_pro = htons(ETH_P_IP);
987	rah->ar_hln = ETH_ALEN;
988	rah->ar_pln = 4;
989	rah->ar_op = htons(ARPOP_REPLY);
990	memcpy(rah->ar_sha, reh->h_source, ETH_ALEN);
991	memcpy(rah->ar_sip, ah->ar_tip, 4);
992	memcpy(rah->ar_tha, ah->ar_sha, ETH_ALEN);
993	memcpy(rah->ar_tip, ah->ar_sip, 4);
994	slirp_output(pData->pvUser, arp_reply, sizeof(arp_reply));
995	}
996	break;
997	default:
998	break;
999	}
1000	}
1001
1002	void slirp_input(PNATState pData, const uint8_t *pkt, int pkt_len)
1003	{
1004	struct mbuf *m;
1005	int proto;
1006
1007	if (pkt_len < ETH_HLEN)
1008	return;
1009
1010	proto = ntohs((uint16_t )(pkt + 12));
1011	switch(proto)
1012	{
1013	case ETH_P_ARP:
1014	arp_input(pData, pkt, pkt_len);
1015	break;
1016	case ETH_P_IP:
1017	/* Update time. Important if the network is very quiet, as otherwise
1018	* the first outgoing connection gets an incorrect timestamp. */
1019	updtime(pData);
1020
1021	m = m_get(pData);
1022	if (!m)
1023	return;
1024	/* Note: we add to align the IP header */
1025	if (M_FREEROOM(m) < pkt_len + 2)
1026	{
1027	m_inc(m, pkt_len + 2);
1028	}
1029	m->m_len = pkt_len + 2;
1030	memcpy(m->m_data + 2, pkt, pkt_len);
1031
1032	m->m_data += 2 + ETH_HLEN;
1033	m->m_len -= 2 + ETH_HLEN;
1034
1035	ip_input(pData, m);
1036	break;
1037	default:
1038	break;
1039	}
1040	}
1041
1042	/* output the IP packet to the ethernet device */
1043	void if_encap(PNATState pData, const uint8_t *ip_data, int ip_data_len)
1044	{
1045	uint8_t buf[1600];
1046	struct ethhdr eh = (struct ethhdr )buf;
1047
1048	if (ip_data_len + ETH_HLEN > sizeof(buf))
1049	return;
1050
1051	memcpy(eh->h_dest, client_ethaddr, ETH_ALEN);
1052	memcpy(eh->h_source, special_ethaddr, ETH_ALEN - 1);
1053	/* XXX: not correct */
1054	eh->h_source[5] = CTL_ALIAS;
1055	eh->h_proto = htons(ETH_P_IP);
1056	memcpy(buf + sizeof(struct ethhdr), ip_data, ip_data_len);
1057	slirp_output(pData->pvUser, buf, ip_data_len + ETH_HLEN);
1058	}
1059
1060	int slirp_redir(PNATState pData, int is_udp, int host_port,
1061	struct in_addr guest_addr, int guest_port)
1062	{
1063	if (is_udp)
1064	{
1065	if (!udp_listen(pData, htons(host_port), guest_addr.s_addr,
1066	htons(guest_port), 0))
1067	return -1;
1068	}
1069	else
1070	{
1071	if (!solisten(pData, htons(host_port), guest_addr.s_addr,
1072	htons(guest_port), 0))
1073	return -1;
1074	}
1075	return 0;
1076	}
1077
1078	int slirp_add_exec(PNATState pData, int do_pty, const char *args, int addr_low_byte,
1079	int guest_port)
1080	{
1081	return add_exec(&exec_list, do_pty, (char *)args,
1082	addr_low_byte, htons(guest_port));
1083	}
1084
1085	void slirp_set_ethaddr(PNATState pData, const uint8_t *ethaddr)
1086	{
1087	memcpy(client_ethaddr, ethaddr, ETH_ALEN);
1088	}
1089
1090	#if defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) && defined(RT_OS_WINDOWS)
1091	HANDLE *slirp_get_events(PNATState pData)
1092	{
1093	return pData->phEvents;
1094	}
1095	void slirp_register_external_event(PNATState pData, HANDLE hEvent, int index)
1096	{
1097	pData->phEvents[index] = hEvent;
1098	}
1099	#endif
1100
1101	unsigned int slirp_get_timeout_ms(PNATState pData)
1102	{
1103	if (link_up)
1104	{
1105	if (time_fasttimo)
1106	return 2;
1107	if (do_slowtimo)
1108	return 500; /* see PR_SLOWHZ */
1109	}
1110	return 0;
1111	}

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

source: vbox/trunk/src/VBox/Devices/Network/slirp/slirp.c@ 15220

Download in other formats: