slirp.c@ 15219

Last change on this file since 15219 was 15210, checked in by vboxsync, 16 years ago
build fix
Property svn:eol-style set to `native`
File size: 33.8 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	# ifdef VBOX_WITH_SLIRP_ICMP
30	# define ICMP_ENGAGE_EVENT(so, fdset) \
31	do { \
32	if (pData->icmp_socket.s != -1) \
33	DO_ENGAGE_EVENT1((so), (fdset), ICMP); \
34	} while (0)
35	# else /* !VBOX_WITH_SLIRP_ICMP */
36	# define ICMP_ENGAGE_EVENT(so, fdset) do {} while (0)
37	# endif /* !VBOX_WITH_SLIRP_ICMP */
38
39	#else /* defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) && defined(RT_OS_WINDOWS) */
40
41	/*
42	* On Windows, we will be notified by IcmpSendEcho2() when the response arrives.
43	* So no call to WSAEventSelect necessary.
44	*/
45	# define ICMP_ENGAGE_EVENT(so, fdset) do {} while(0)
46
47	# define DO_ENGAGE_EVENT1(so, fdset1, label) \
48	do { \
49	rc = WSAEventSelect((so)->s, VBOX_SOCKET_EVENT, FD_ALL_EVENTS); \
50	if (rc == SOCKET_ERROR) \
51	{ \
52	/* This should not happen */ \
53	error = WSAGetLastError(); \
54	LogRel(("WSAEventSelector (" #label ") error %d (so=%x, socket=%s, event=%x)\n", \
55	error, (so), (so)->s, VBOX_SOCKET_EVENT)); \
56	} \
57	} while(0)
58
59	# define DO_ENGAGE_EVENT2(so, fdset1, fdset2, label) \
60	DO_ENGAGE_EVENT1((so), (fdset1), label)
61
62	#endif /* defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) && defined(RT_OS_WINDOWS) */
63
64	#define TCP_ENGAGE_EVENT1(so, fdset) \
65	DO_ENGAGE_EVENT1((so), (fdset), TCP)
66
67	#define TCP_ENGAGE_EVENT2(so, fdset1, fdset2) \
68	DO_ENGAGE_EVENT2((so), (fdset1), (fdset2), TCP)
69
70	#define UDP_ENGAGE_EVENT(so, fdset) \
71	DO_ENGAGE_EVENT1((so), (fdset), UDP)
72
73	static const uint8_t special_ethaddr[6] =
74	{
75	0x52, 0x54, 0x00, 0x12, 0x35, 0x00
76	};
77
78	#ifdef RT_OS_WINDOWS
79
80	static int get_dns_addr_domain(PNATState pData, bool fVerbose,
81	struct in_addr *pdns_addr,
82	const char **ppszDomain)
83	{
84	int rc = 0;
85	FIXED_INFO *FixedInfo = NULL;
86	ULONG BufLen;
87	DWORD ret;
88	IP_ADDR_STRING *pIPAddr;
89	struct in_addr tmp_addr;
90
91	FixedInfo = (FIXED_INFO *)GlobalAlloc(GPTR, sizeof(FIXED_INFO));
92	BufLen = sizeof(FIXED_INFO);
93
94	/** @todo: this API returns all DNS servers, no matter whether the
95	* corresponding network adapter is disabled or not. Maybe replace
96	* this by GetAdapterAddresses(), which is XP/Vista only though. */
97	if (ERROR_BUFFER_OVERFLOW == GetNetworkParams(FixedInfo, &BufLen))
98	{
99	if (FixedInfo)
100	{
101	GlobalFree(FixedInfo);
102	FixedInfo = NULL;
103	}
104	FixedInfo = GlobalAlloc(GPTR, BufLen);
105	}
106
107	if ((ret = GetNetworkParams(FixedInfo, &BufLen)) != ERROR_SUCCESS)
108	{
109	Log(("GetNetworkParams failed. ret = %08x\n", (u_int)ret ));
110	if (FixedInfo)
111	{
112	GlobalFree(FixedInfo);
113	FixedInfo = NULL;
114	}
115	rc = -1;
116	goto get_dns_prefix;
117	}
118
119	pIPAddr = &(FixedInfo->DnsServerList);
120	inet_aton(pIPAddr->IpAddress.String, &tmp_addr);
121	Log(("nat: DNS Servers:\n"));
122	if (fVerbose \|\| pdns_addr->s_addr != tmp_addr.s_addr)
123	LogRel(("NAT: DNS address: %s\n", pIPAddr->IpAddress.String));
124	*pdns_addr = tmp_addr;
125
126	pIPAddr = FixedInfo -> DnsServerList.Next;
127	while (pIPAddr)
128	{
129	if (fVerbose)
130	LogRel(("NAT: ignored DNS address: %s\n", pIPAddr ->IpAddress.String));
131	pIPAddr = pIPAddr ->Next;
132	}
133	if (FixedInfo)
134	{
135	GlobalFree(FixedInfo);
136	FixedInfo = NULL;
137	}
138
139	get_dns_prefix:
140	if (ppszDomain)
141	{
142	OSVERSIONINFO ver;
143	char szDnsDomain[256];
144	DWORD dwSize = sizeof(szDnsDomain);
145
146	*ppszDomain = NULL;
147	GetVersionEx(&ver);
148	if (ver.dwMajorVersion >= 5)
149	{
150	/* GetComputerNameEx exists in Windows versions starting with 2000. */
151	if (GetComputerNameEx(ComputerNameDnsDomain, szDnsDomain, &dwSize))
152	{
153	if (szDnsDomain[0])
154	{
155	/* Just non-empty strings are valid. */
156	*ppszDomain = RTStrDup(szDnsDomain);
157	if (pData->fPassDomain)
158	{
159	if (fVerbose)
160	LogRel(("NAT: passing domain name %s\n", szDnsDomain));
161	}
162	else
163	Log(("nat: ignoring domain %s\n", szDnsDomain));
164	}
165	}
166	else
167	Log(("nat: GetComputerNameEx failed (%d)\n", GetLastError()));
168	}
169	}
170	return rc;
171	}
172
173	#else
174
175	static int get_dns_addr_domain(PNATState pData, bool fVerbose,
176	struct in_addr *pdns_addr,
177	const char **ppszDomain)
178	{
179	char buff[512];
180	char buff2[256];
181	FILE *f;
182	int found = 0;
183	struct in_addr tmp_addr;
184
185	#ifdef RT_OS_OS2
186	/* Try various locations. */
187	char *etc = getenv("ETC");
188	f = NULL;
189	if (etc)
190	{
191	snprintf(buff, sizeof(buff), "%s/RESOLV2", etc);
192	f = fopen(buff, "rt");
193	}
194	if (!f)
195	{
196	snprintf(buff, sizeof(buff), "%s/RESOLV2", _PATH_ETC);
197	f = fopen(buff, "rt");
198	}
199	if (!f)
200	{
201	snprintf(buff, sizeof(buff), "%s/resolv.conf", _PATH_ETC);
202	f = fopen(buff, "rt");
203	}
204	#else
205	f = fopen("/etc/resolv.conf", "r");
206	#endif
207	if (!f)
208	return -1;
209
210	if (ppszDomain)
211	*ppszDomain = NULL;
212	Log(("nat: DNS Servers:\n"));
213	while (fgets(buff, 512, f) != NULL)
214	{
215	if (sscanf(buff, "nameserver%*[ \t]%256s", buff2) == 1)
216	{
217	if (!inet_aton(buff2, &tmp_addr))
218	continue;
219	if (tmp_addr.s_addr == loopback_addr.s_addr)
220	tmp_addr = our_addr;
221	/* If it's the first one, set it to dns_addr */
222	if (!found)
223	{
224	if (fVerbose \|\| pdns_addr->s_addr != tmp_addr.s_addr)
225	LogRel(("NAT: DNS address: %s\n", buff2));
226	*pdns_addr = tmp_addr;
227	}
228	else
229	{
230	if (fVerbose)
231	LogRel(("NAT: ignored DNS address: %s\n", buff2));
232	}
233	found++;
234	}
235	if ( ppszDomain
236	&& (!strncmp(buff, "domain", 6) \|\| !strncmp(buff, "search", 6)))
237	{
238	/* Domain name/search list present. Pick first entry */
239	if (*ppszDomain == NULL)
240	{
241	char *tok;
242	char *saveptr;
243	tok = strtok_r(&buff[6], " \t\n", &saveptr);
244	if (tok)
245	{
246	*ppszDomain = RTStrDup(tok);
247	if (pData->fPassDomain)
248	{
249	if (fVerbose)
250	LogRel(("NAT: passing domain name %s\n", tok));
251	}
252	else
253	Log(("nat: ignoring domain %s\n", tok));
254	}
255	}
256	}
257	}
258	fclose(f);
259	if (!found)
260	return -1;
261	return 0;
262	}
263
264	#endif
265
266	int get_dns_addr(PNATState pData, struct in_addr *pdns_addr)
267	{
268	return get_dns_addr_domain(pData, false, pdns_addr, NULL);
269	}
270
271	int slirp_init(PNATState ppData, const char pszNetAddr, uint32_t u32Netmask,
272	bool fPassDomain, const char *pszTFTPPrefix,
273	const char pszBootFile, void pvUser)
274	{
275	int fNATfailed = 0;
276	PNATState pData = malloc(sizeof(NATState));
277	*ppData = pData;
278	if (!pData)
279	return VERR_NO_MEMORY;
280	if (u32Netmask & 0x1f)
281	/* CTL is x.x.x.15, bootp passes up to 16 IPs (15..31) */
282	return VERR_INVALID_PARAMETER;
283	memset(pData, '\0', sizeof(NATState));
284	pData->fPassDomain = fPassDomain;
285	pData->pvUser = pvUser;
286	#if ARCH_BITS == 64 && !defined(VBOX_WITH_BSD_REASS)
287	pData->cpvHashUsed = 1;
288	#endif
289	tftp_prefix = pszTFTPPrefix;
290	bootp_filename = pszBootFile;
291	pData->netmask = u32Netmask;
292
293	#ifdef RT_OS_WINDOWS
294	{
295	WSADATA Data;
296	WSAStartup(MAKEWORD(2,0), &Data);
297	}
298	# if defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC)
299	pData->phEvents[VBOX_SOCKET_EVENT_INDEX] = CreateEvent(NULL, FALSE, FALSE, NULL);
300	# endif
301	#endif
302
303	link_up = 1;
304
305	debug_init();
306	if_init(pData);
307	ip_init(pData);
308	#ifdef VBOX_WITH_SLIRP_ICMP
309	icmp_init(pData);
310	#endif /* VBOX_WITH_SLIRP_ICMP */
311
312	/* Initialise mbufs after setting the MTU */
313	m_init(pData);
314
315	/* set default addresses */
316	inet_aton("127.0.0.1", &loopback_addr);
317	inet_aton("127.0.0.1", &dns_addr);
318
319	if (get_dns_addr_domain(pData, true, &dns_addr, &pData->pszDomain) < 0)
320	fNATfailed = 1;
321
322	inet_aton(pszNetAddr, &special_addr);
323	alias_addr.s_addr = special_addr.s_addr \| htonl(CTL_ALIAS);
324	getouraddr(pData);
325	return fNATfailed ? VINF_NAT_DNS : VINF_SUCCESS;
326	}
327
328	/**
329	* Statistics counters.
330	*/
331	void slirp_register_timers(PNATState pData, PPDMDRVINS pDrvIns)
332	{
333	#ifndef VBOX_WITHOUT_RELEASE_STATISTICS
334	PDMDrvHlpSTAMRegisterF(pDrvIns, &pData->StatFill, STAMTYPE_PROFILE, STAMVISIBILITY_ALWAYS,
335	STAMUNIT_TICKS_PER_CALL, "Profiling slirp fills", "/Drivers/NAT%d/Fill", pDrvIns->iInstance);
336	PDMDrvHlpSTAMRegisterF(pDrvIns, &pData->StatPoll, STAMTYPE_PROFILE, STAMVISIBILITY_ALWAYS,
337	STAMUNIT_TICKS_PER_CALL, "Profiling slirp polls", "/Drivers/NAT%d/Poll", pDrvIns->iInstance);
338	PDMDrvHlpSTAMRegisterF(pDrvIns, &pData->StatFastTimer, STAMTYPE_PROFILE, STAMVISIBILITY_ALWAYS,
339	STAMUNIT_TICKS_PER_CALL, "Profiling slirp fast timer", "/Drivers/NAT%d/TimerFast", pDrvIns->iInstance);
340	PDMDrvHlpSTAMRegisterF(pDrvIns, &pData->StatSlowTimer, STAMTYPE_PROFILE, STAMVISIBILITY_ALWAYS,
341	STAMUNIT_TICKS_PER_CALL, "Profiling slirp slow timer", "/Drivers/NAT%d/TimerSlow", pDrvIns->iInstance);
342	PDMDrvHlpSTAMRegisterF(pDrvIns, &pData->StatTCP, STAMTYPE_COUNTER, STAMVISIBILITY_ALWAYS,
343	STAMUNIT_COUNT, "TCP sockets", "/Drivers/NAT%d/SockTCP", pDrvIns->iInstance);
344	PDMDrvHlpSTAMRegisterF(pDrvIns, &pData->StatTCPHot, STAMTYPE_COUNTER, STAMVISIBILITY_ALWAYS,
345	STAMUNIT_COUNT, "TCP sockets active", "/Drivers/NAT%d/SockTCPHot", pDrvIns->iInstance);
346	PDMDrvHlpSTAMRegisterF(pDrvIns, &pData->StatUDP, STAMTYPE_COUNTER, STAMVISIBILITY_ALWAYS,
347	STAMUNIT_COUNT, "UDP sockets", "/Drivers/NAT%d/SockUDP", pDrvIns->iInstance);
348	PDMDrvHlpSTAMRegisterF(pDrvIns, &pData->StatUDPHot, STAMTYPE_COUNTER, STAMVISIBILITY_ALWAYS,
349	STAMUNIT_COUNT, "UDP sockets active", "/Drivers/NAT%d/SockUDPHot", pDrvIns->iInstance);
350	#endif /* VBOX_WITHOUT_RELEASE_STATISTICS */
351	}
352
353	/**
354	* Marks the link as up, making it possible to establish new connections.
355	*/
356	void slirp_link_up(PNATState pData)
357	{
358	link_up = 1;
359	}
360
361	/**
362	* Marks the link as down and cleans up the current connections.
363	*/
364	void slirp_link_down(PNATState pData)
365	{
366	struct socket *so;
367
368	while ((so = tcb.so_next) != &tcb)
369	{
370	if (so->so_state & SS_NOFDREF \|\| so->s == -1)
371	sofree(pData, so);
372	else
373	tcp_drop(pData, sototcpcb(so), 0);
374	}
375
376	while ((so = udb.so_next) != &udb)
377	udp_detach(pData, so);
378
379	link_up = 0;
380	}
381
382	/**
383	* Terminates the slirp component.
384	*/
385	void slirp_term(PNATState pData)
386	{
387	if (pData->pszDomain)
388	RTStrFree((char )(void )pData->pszDomain);
389
390	#if ARCH_BITS == 64 && !defined(VBOX_WITH_BSD_REASS)
391	LogRel(("NAT: cpvHashUsed=%RU32 cpvHashCollisions=%RU32 cpvHashInserts=%RU64 cpvHashDone=%RU64\n",
392	pData->cpvHashUsed, pData->cpvHashCollisions, pData->cpvHashInserts, pData->cpvHashDone));
393	#endif
394
395	#ifdef VBOX_WITH_SLIRP_ICMP
396	# ifdef RT_OS_WINDOWS
397	pData->pfIcmpCloseHandle(pData->icmp_socket.sh);
398	# else
399	closesocket(pData->icmp_socket.s);
400	# endif
401	#endif
402
403	slirp_link_down(pData);
404	#ifdef RT_OS_WINDOWS
405	WSACleanup();
406	#endif
407	#ifdef LOG_ENABLED
408	Log(("\n"
409	"NAT statistics\n"
410	"--------------\n"
411	"\n"));
412	ipstats(pData);
413	tcpstats(pData);
414	udpstats(pData);
415	icmpstats(pData);
416	mbufstats(pData);
417	sockstats(pData);
418	Log(("\n"
419	"\n"
420	"\n"));
421	#endif
422	free(pData);
423	}
424
425
426	#define CONN_CANFSEND(so) (((so)->so_state & (SS_FCANTSENDMORE\|SS_ISFCONNECTED)) == SS_ISFCONNECTED)
427	#define CONN_CANFRCV(so) (((so)->so_state & (SS_FCANTRCVMORE\|SS_ISFCONNECTED)) == SS_ISFCONNECTED)
428	#define UPD_NFDS(x) if (nfds < (x)) nfds = (x)
429
430	/*
431	* curtime kept to an accuracy of 1ms
432	*/
433	#ifdef RT_OS_WINDOWS
434	static void updtime(PNATState pData)
435	{
436	struct _timeb tb;
437
438	_ftime(&tb);
439	curtime = (u_int)tb.time * (u_int)1000;
440	curtime += (u_int)tb.millitm;
441	}
442	#else
443	static void updtime(PNATState pData)
444	{
445	gettimeofday(&tt, 0);
446
447	curtime = (u_int)tt.tv_sec * (u_int)1000;
448	curtime += (u_int)tt.tv_usec / (u_int)1000;
449
450	if ((tt.tv_usec % 1000) >= 500)
451	curtime++;
452	}
453	#endif
454
455	void slirp_select_fill(PNATState pData, int *pnfds,
456	fd_set readfds, fd_set writefds, fd_set *xfds)
457	{
458	struct socket so, so_next;
459	int nfds;
460	#if defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) && defined(RT_OS_WINDOWS)
461	int rc;
462	int error;
463	#endif
464	#ifdef VBOX_WITH_BSD_REASS
465	int i;
466	#endif /* VBOX_WITH_BSD_REASS */
467
468	STAM_REL_PROFILE_START(&pData->StatFill, a);
469
470	nfds = *pnfds;
471
472	/*
473	* First, TCP sockets
474	*/
475	do_slowtimo = 0;
476	if (link_up)
477	{
478	/*
479	* *_slowtimo needs calling if there are IP fragments
480	* in the fragment queue, or there are TCP connections active
481	*/
482	#ifndef VBOX_WITH_BSD_REASS
483	do_slowtimo = ((tcb.so_next != &tcb)
484	\|\| ((struct ipasfrag )&ipq != u32_to_ptr(pData, ipq.next, struct ipasfrag )));
485	#else /* !VBOX_WITH_BSD_REASS */
486	/* XXX:
487	* triggering of fragment expiration should be the same but use new macroses
488	*/
489	do_slowtimo = (tcb.so_next != &tcb);
490	if (!do_slowtimo)
491	{
492	for (i = 0; i < IPREASS_NHASH; i++)
493	{
494	if (!TAILQ_EMPTY(&ipq[i]))
495	{
496	do_slowtimo = 1;
497	break;
498	}
499	}
500	}
501	#endif /* VBOX_WITH_BSD_REASS */
502
503	STAM_REL_COUNTER_RESET(&pData->StatTCP);
504	STAM_REL_COUNTER_RESET(&pData->StatTCPHot);
505
506	for (so = tcb.so_next; so != &tcb; so = so_next)
507	{
508	so_next = so->so_next;
509
510	STAM_REL_COUNTER_INC(&pData->StatTCP);
511
512	/*
513	* See if we need a tcp_fasttimo
514	*/
515	if (time_fasttimo == 0 && so->so_tcpcb->t_flags & TF_DELACK)
516	time_fasttimo = curtime; /* Flag when we want a fasttimo */
517
518	/*
519	* NOFDREF can include still connecting to local-host,
520	* newly socreated() sockets etc. Don't want to select these.
521	*/
522	if (so->so_state & SS_NOFDREF \|\| so->s == -1)
523	continue;
524
525	/*
526	* Set for reading sockets which are accepting
527	*/
528	if (so->so_state & SS_FACCEPTCONN)
529	{
530	STAM_REL_COUNTER_INC(&pData->StatTCPHot);
531	TCP_ENGAGE_EVENT1(so, readfds);
532	continue;
533	}
534
535	/*
536	* Set for writing sockets which are connecting
537	*/
538	if (so->so_state & SS_ISFCONNECTING)
539	{
540	STAM_REL_COUNTER_INC(&pData->StatTCPHot);
541	TCP_ENGAGE_EVENT1(so, writefds);
542	}
543
544	/*
545	* Set for writing if we are connected, can send more, and
546	* we have something to send
547	*/
548	if (CONN_CANFSEND(so) && so->so_rcv.sb_cc)
549	{
550	STAM_REL_COUNTER_INC(&pData->StatTCPHot);
551	TCP_ENGAGE_EVENT1(so, writefds);
552	}
553
554	/*
555	* Set for reading (and urgent data) if we are connected, can
556	* receive more, and we have room for it XXX /2 ?
557	*/
558	if (CONN_CANFRCV(so) && (so->so_snd.sb_cc < (so->so_snd.sb_datalen/2)))
559	{
560	STAM_REL_COUNTER_INC(&pData->StatTCPHot);
561	TCP_ENGAGE_EVENT2(so, readfds, xfds);
562	}
563	}
564
565	/*
566	* UDP sockets
567	*/
568	STAM_REL_COUNTER_RESET(&pData->StatUDP);
569	STAM_REL_COUNTER_RESET(&pData->StatUDPHot);
570
571	for (so = udb.so_next; so != &udb; so = so_next)
572	{
573	so_next = so->so_next;
574
575	STAM_REL_COUNTER_INC(&pData->StatUDP);
576
577	/*
578	* See if it's timed out
579	*/
580	if (so->so_expire)
581	{
582	if (so->so_expire <= curtime)
583	{
584	udp_detach(pData, so);
585	continue;
586	}
587	else
588	do_slowtimo = 1; /* Let socket expire */
589	}
590
591	/*
592	* When UDP packets are received from over the link, they're
593	* sendto()'d straight away, so no need for setting for writing
594	* Limit the number of packets queued by this session to 4.
595	* Note that even though we try and limit this to 4 packets,
596	* the session could have more queued if the packets needed
597	* to be fragmented.
598	*
599	* (XXX <= 4 ?)
600	*/
601	if ((so->so_state & SS_ISFCONNECTED) && so->so_queued <= 4)
602	{
603	STAM_REL_COUNTER_INC(&pData->StatUDPHot);
604	UDP_ENGAGE_EVENT(so, readfds);
605	}
606	}
607
608	ICMP_ENGAGE_EVENT(&pData->icmp_socket, readfds);
609	}
610
611	#if !defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) \|\| !defined(RT_OS_WINDOWS)
612	*pnfds = nfds;
613	#else
614	*pnfds = VBOX_EVENT_COUNT;
615	#endif
616
617	STAM_REL_PROFILE_STOP(&pData->StatFill, a);
618	}
619
620	void slirp_select_poll(PNATState pData, fd_set readfds, fd_set writefds, fd_set *xfds)
621	{
622	struct socket so, so_next;
623	int ret;
624	#if defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) && defined(RT_OS_WINDOWS)
625	WSANETWORKEVENTS NetworkEvents;
626	int rc;
627	int error;
628	#endif
629	static uint32_t stat_time;
630
631	STAM_REL_PROFILE_START(&pData->StatPoll, a);
632
633	/* Update time */
634	updtime(pData);
635
636	#ifdef LOG_ENABLED
637	if (curtime - stat_time > 10000)
638	{
639	stat_time = curtime;
640	sockstats(pData);
641	}
642	#endif
643
644	/*
645	* See if anything has timed out
646	*/
647	if (link_up)
648	{
649	if (time_fasttimo && ((curtime - time_fasttimo) >= 2))
650	{
651	STAM_REL_PROFILE_START(&pData->StatFastTimer, a);
652	tcp_fasttimo(pData);
653	time_fasttimo = 0;
654	STAM_REL_PROFILE_STOP(&pData->StatFastTimer, a);
655	}
656	if (do_slowtimo && ((curtime - last_slowtimo) >= 499))
657	{
658	STAM_REL_PROFILE_START(&pData->StatSlowTimer, a);
659	ip_slowtimo(pData);
660	tcp_slowtimo(pData);
661	last_slowtimo = curtime;
662	STAM_REL_PROFILE_STOP(&pData->StatSlowTimer, a);
663	}
664	}
665	#if defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) && defined(RT_OS_WINDOWS)
666	if (!readfds && !writefds && !xfds)
667	return; /* only timer update */
668	#endif
669
670	/*
671	* Check sockets
672	*/
673	if (link_up)
674	{
675	/*
676	* Check TCP sockets
677	*/
678	for (so = tcb.so_next; so != &tcb; so = so_next)
679	{
680	so_next = so->so_next;
681
682	/*
683	* FD_ISSET is meaningless on these sockets
684	* (and they can crash the program)
685	*/
686	if (so->so_state & SS_NOFDREF \|\| so->s == -1)
687	continue;
688
689	#if defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) && defined(RT_OS_WINDOWS)
690	rc = WSAEnumNetworkEvents(so->s, VBOX_SOCKET_EVENT, &NetworkEvents);
691	if (rc == SOCKET_ERROR)
692	{
693	error = WSAGetLastError();
694	LogRel(("WSAEnumNetworkEvents TCP error %d\n", error));
695	continue;
696	}
697	#endif
698
699	/*
700	* Check for URG data
701	* This will soread as well, so no need to
702	* test for readfds below if this succeeds
703	*/
704	#if !defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) \|\| !defined(RT_OS_WINDOWS)
705	if (FD_ISSET(so->s, xfds))
706	#else
707	/* out-of-band data */
708	if ((NetworkEvents.lNetworkEvents & FD_OOB) && NetworkEvents.iErrorCode[FD_OOB_BIT] == 0)
709	#endif
710	{
711	sorecvoob(pData, so);
712	}
713
714	/*
715	* Check sockets for reading
716	*/
717	#if !defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) \|\| !defined(RT_OS_WINDOWS)
718	else if (FD_ISSET(so->s, readfds))
719	#else
720	else if ((NetworkEvents.lNetworkEvents & FD_READ) && (NetworkEvents.iErrorCode[FD_READ_BIT] == 0))
721	#endif
722	{
723	/*
724	* Check for incoming connections
725	*/
726	if (so->so_state & SS_FACCEPTCONN)
727	{
728	tcp_connect(pData, so);
729	#if defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) && defined(RT_OS_WINDOWS)
730	if (!(NetworkEvents.lNetworkEvents & FD_CLOSE))
731	#endif
732	continue;
733	}
734
735	ret = soread(pData, so);
736	/* Output it if we read something */
737	if (ret > 0)
738	tcp_output(pData, sototcpcb(so));
739	}
740
741	#if defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) && defined(RT_OS_WINDOWS)
742	/*
743	* Check for FD_CLOSE events.
744	*/
745	if (NetworkEvents.lNetworkEvents & FD_CLOSE)
746	{
747	/*
748	* drain the socket
749	*/
750	for (;;)
751	{
752	ret = soread(pData, so);
753	if (ret > 0)
754	tcp_output(pData, sototcpcb(so));
755	else
756	break;
757	}
758	}
759	#endif
760
761	/*
762	* Check sockets for writing
763	*/
764	#if !defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) \|\| !defined(RT_OS_WINDOWS)
765	if (FD_ISSET(so->s, writefds))
766	#else
767	if ((NetworkEvents.lNetworkEvents & FD_WRITE) && (NetworkEvents.iErrorCode[FD_WRITE_BIT] == 0))
768	#endif
769	{
770	/*
771	* Check for non-blocking, still-connecting sockets
772	*/
773	if (so->so_state & SS_ISFCONNECTING)
774	{
775	/* Connected */
776	so->so_state &= ~SS_ISFCONNECTING;
777
778	/*
779	* This should be probably guarded by PROBE_CONN too. Anyway,
780	* we disable it on OS/2 because the below send call returns
781	* EFAULT which causes the opened TCP socket to close right
782	* after it has been opened and connected.
783	*/
784	#ifndef RT_OS_OS2
785	ret = send(so->s, (const char *)&ret, 0, 0);
786	if (ret < 0)
787	{
788	/* XXXXX Must fix, zero bytes is a NOP */
789	if ( errno == EAGAIN
790	\|\| errno == EWOULDBLOCK
791	\|\| errno == EINPROGRESS
792	\|\| errno == ENOTCONN)
793	continue;
794
795	/* else failed */
796	so->so_state = SS_NOFDREF;
797	}
798	/* else so->so_state &= ~SS_ISFCONNECTING; */
799	#endif
800
801	/*
802	* Continue tcp_input
803	*/
804	tcp_input(pData, (struct mbuf *)NULL, sizeof(struct ip), so);
805	/* continue; */
806	}
807	else
808	ret = sowrite(pData, so);
809	/*
810	* XXX If we wrote something (a lot), there could be the need
811	* for a window update. In the worst case, the remote will send
812	* a window probe to get things going again.
813	*/
814	}
815
816	/*
817	* Probe a still-connecting, non-blocking socket
818	* to check if it's still alive
819	*/
820	#ifdef PROBE_CONN
821	if (so->so_state & SS_ISFCONNECTING)
822	{
823	ret = recv(so->s, (char *)&ret, 0, 0);
824
825	if (ret < 0)
826	{
827	/* XXX */
828	if ( errno == EAGAIN
829	\|\| errno == EWOULDBLOCK
830	\|\| errno == EINPROGRESS
831	\|\| errno == ENOTCONN)
832	{
833	continue; /* Still connecting, continue */
834	}
835
836	/* else failed */
837	so->so_state = SS_NOFDREF;
838
839	/* tcp_input will take care of it */
840	}
841	else
842	{
843	ret = send(so->s, &ret, 0, 0);
844	if (ret < 0)
845	{
846	/* XXX */
847	if ( errno == EAGAIN
848	\|\| errno == EWOULDBLOCK
849	\|\| errno == EINPROGRESS
850	\|\| errno == ENOTCONN)
851	{
852	continue;
853	}
854	/* else failed */
855	so->so_state = SS_NOFDREF;
856	}
857	else
858	so->so_state &= ~SS_ISFCONNECTING;
859
860	}
861	tcp_input((struct mbuf *)NULL, sizeof(struct ip),so);
862	} /* SS_ISFCONNECTING */
863	#endif
864	}
865
866	/*
867	* Now UDP sockets.
868	* Incoming packets are sent straight away, they're not buffered.
869	* Incoming UDP data isn't buffered either.
870	*/
871	for (so = udb.so_next; so != &udb; so = so_next)
872	{
873	so_next = so->so_next;
874
875	#if defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) && defined(RT_OS_WINDOWS)
876	rc = WSAEnumNetworkEvents(so->s, VBOX_SOCKET_EVENT, &NetworkEvents);
877	if (rc == SOCKET_ERROR)
878	{
879	error = WSAGetLastError();
880	LogRel(("WSAEnumNetworkEvents TCP error %d\n", error));
881	continue;
882	}
883	#endif
884	#if !defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) \|\| !defined(RT_OS_WINDOWS)
885	if (so->s != -1 && FD_ISSET(so->s, readfds))
886	#else
887	if (((NetworkEvents.lNetworkEvents & FD_READ) && (NetworkEvents.iErrorCode[FD_READ_BIT] == 0)))
888	#endif
889	{
890	sorecvfrom(pData, so);
891	}
892	}
893
894	#if defined(VBOX_WITH_SLIRP_ICMP)
895	# if defined(RT_OS_WINDOWS)
896	sorecvfrom(pData, &pData->icmp_socket);
897	# else
898	if (pData->icmp_socket.s != -1 && FD_ISSET(pData->icmp_socket.s, readfds))
899	sorecvfrom(pData, &pData->icmp_socket);
900	# endif
901	#endif
902	}
903
904	/*
905	* See if we can start outputting
906	*/
907	if (if_queued && link_up)
908	if_start(pData);
909
910	STAM_REL_PROFILE_STOP(&pData->StatPoll, a);
911	}
912
913	#define ETH_ALEN 6
914	#define ETH_HLEN 14
915
916	#define ETH_P_IP 0x0800 /* Internet Protocol packet */
917	#define ETH_P_ARP 0x0806 /* Address Resolution packet */
918
919	#define ARPOP_REQUEST 1 /* ARP request */
920	#define ARPOP_REPLY 2 /* ARP reply */
921
922	struct ethhdr
923	{
924	unsigned char h_dest[ETH_ALEN]; /* destination eth addr */
925	unsigned char h_source[ETH_ALEN]; /* source ether addr */
926	unsigned short h_proto; /* packet type ID field */
927	};
928
929	struct arphdr
930	{
931	unsigned short ar_hrd; /* format of hardware address */
932	unsigned short ar_pro; /* format of protocol address */
933	unsigned char ar_hln; /* length of hardware address */
934	unsigned char ar_pln; /* length of protocol address */
935	unsigned short ar_op; /* ARP opcode (command) */
936
937	/*
938	* Ethernet looks like this : This bit is variable sized however...
939	*/
940	unsigned char ar_sha[ETH_ALEN]; /* sender hardware address */
941	unsigned char ar_sip[4]; /* sender IP address */
942	unsigned char ar_tha[ETH_ALEN]; /* target hardware address */
943	unsigned char ar_tip[4]; /* target IP address */
944	};
945
946	static
947	void arp_input(PNATState pData, const uint8_t *pkt, int pkt_len)
948	{
949	struct ethhdr eh = (struct ethhdr )pkt;
950	struct arphdr ah = (struct arphdr )(pkt + ETH_HLEN);
951	uint8_t arp_reply[ETH_HLEN + sizeof(struct arphdr)];
952	struct ethhdr reh = (struct ethhdr )arp_reply;
953	struct arphdr rah = (struct arphdr )(arp_reply + ETH_HLEN);
954	int ar_op;
955	struct ex_list *ex_ptr;
956	uint32_t htip = ntohl((uint32_t)ah->ar_tip);
957
958	ar_op = ntohs(ah->ar_op);
959	switch(ar_op)
960	{
961	case ARPOP_REQUEST:
962	if ((htip & pData->netmask) == ntohl(special_addr.s_addr))
963	{
964	if ( (htip & ~pData->netmask) == CTL_DNS
965	\|\| (htip & ~pData->netmask) == CTL_ALIAS)
966	goto arp_ok;
967	for (ex_ptr = exec_list; ex_ptr; ex_ptr = ex_ptr->ex_next)
968	{
969	if ((htip & ~pData->netmask) == ex_ptr->ex_addr)
970	goto arp_ok;
971	}
972	return;
973	arp_ok:
974	/* XXX: make an ARP request to have the client address */
975	memcpy(client_ethaddr, eh->h_source, ETH_ALEN);
976
977	/* ARP request for alias/dns mac address */
978	memcpy(reh->h_dest, pkt + ETH_ALEN, ETH_ALEN);
979	memcpy(reh->h_source, special_ethaddr, ETH_ALEN - 1);
980	reh->h_source[5] = ah->ar_tip[3];
981	reh->h_proto = htons(ETH_P_ARP);
982
983	rah->ar_hrd = htons(1);
984	rah->ar_pro = htons(ETH_P_IP);
985	rah->ar_hln = ETH_ALEN;
986	rah->ar_pln = 4;
987	rah->ar_op = htons(ARPOP_REPLY);
988	memcpy(rah->ar_sha, reh->h_source, ETH_ALEN);
989	memcpy(rah->ar_sip, ah->ar_tip, 4);
990	memcpy(rah->ar_tha, ah->ar_sha, ETH_ALEN);
991	memcpy(rah->ar_tip, ah->ar_sip, 4);
992	slirp_output(pData->pvUser, arp_reply, sizeof(arp_reply));
993	}
994	break;
995	default:
996	break;
997	}
998	}
999
1000	void slirp_input(PNATState pData, const uint8_t *pkt, int pkt_len)
1001	{
1002	struct mbuf *m;
1003	int proto;
1004
1005	if (pkt_len < ETH_HLEN)
1006	return;
1007
1008	proto = ntohs((uint16_t )(pkt + 12));
1009	switch(proto)
1010	{
1011	case ETH_P_ARP:
1012	arp_input(pData, pkt, pkt_len);
1013	break;
1014	case ETH_P_IP:
1015	/* Update time. Important if the network is very quiet, as otherwise
1016	* the first outgoing connection gets an incorrect timestamp. */
1017	updtime(pData);
1018
1019	m = m_get(pData);
1020	if (!m)
1021	return;
1022	/* Note: we add to align the IP header */
1023	if (M_FREEROOM(m) < pkt_len + 2)
1024	{
1025	m_inc(m, pkt_len + 2);
1026	}
1027	m->m_len = pkt_len + 2;
1028	memcpy(m->m_data + 2, pkt, pkt_len);
1029
1030	m->m_data += 2 + ETH_HLEN;
1031	m->m_len -= 2 + ETH_HLEN;
1032
1033	ip_input(pData, m);
1034	break;
1035	default:
1036	break;
1037	}
1038	}
1039
1040	/* output the IP packet to the ethernet device */
1041	void if_encap(PNATState pData, const uint8_t *ip_data, int ip_data_len)
1042	{
1043	uint8_t buf[1600];
1044	struct ethhdr eh = (struct ethhdr )buf;
1045
1046	if (ip_data_len + ETH_HLEN > sizeof(buf))
1047	return;
1048
1049	memcpy(eh->h_dest, client_ethaddr, ETH_ALEN);
1050	memcpy(eh->h_source, special_ethaddr, ETH_ALEN - 1);
1051	/* XXX: not correct */
1052	eh->h_source[5] = CTL_ALIAS;
1053	eh->h_proto = htons(ETH_P_IP);
1054	memcpy(buf + sizeof(struct ethhdr), ip_data, ip_data_len);
1055	slirp_output(pData->pvUser, buf, ip_data_len + ETH_HLEN);
1056	}
1057
1058	int slirp_redir(PNATState pData, int is_udp, int host_port,
1059	struct in_addr guest_addr, int guest_port)
1060	{
1061	if (is_udp)
1062	{
1063	if (!udp_listen(pData, htons(host_port), guest_addr.s_addr,
1064	htons(guest_port), 0))
1065	return -1;
1066	}
1067	else
1068	{
1069	if (!solisten(pData, htons(host_port), guest_addr.s_addr,
1070	htons(guest_port), 0))
1071	return -1;
1072	}
1073	return 0;
1074	}
1075
1076	int slirp_add_exec(PNATState pData, int do_pty, const char *args, int addr_low_byte,
1077	int guest_port)
1078	{
1079	return add_exec(&exec_list, do_pty, (char *)args,
1080	addr_low_byte, htons(guest_port));
1081	}
1082
1083	void slirp_set_ethaddr(PNATState pData, const uint8_t *ethaddr)
1084	{
1085	memcpy(client_ethaddr, ethaddr, ETH_ALEN);
1086	}
1087
1088	#if defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) && defined(RT_OS_WINDOWS)
1089	HANDLE *slirp_get_events(PNATState pData)
1090	{
1091	return pData->phEvents;
1092	}
1093	void slirp_register_external_event(PNATState pData, HANDLE hEvent, int index)
1094	{
1095	pData->phEvents[index] = hEvent;
1096	}
1097	#endif
1098
1099	unsigned int slirp_get_timeout_ms(PNATState pData)
1100	{
1101	if (link_up)
1102	{
1103	if (time_fasttimo)
1104	return 2;
1105	if (do_slowtimo)
1106	return 500; /* see PR_SLOWHZ */
1107	}
1108	return 0;
1109	}

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

source: vbox/trunk/src/VBox/Devices/Network/slirp/slirp.c@ 15219

Download in other formats: