slirp.c@ 15194

Last change on this file since 15194 was 15194, checked in by vboxsync, 16 years ago
Slirp:ICMP: Proper shutdowning of ICMP raw socket on Unix and ICMP file handle on Windows
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) \|\| !defined(RT_OS_WINDOWS)
14	# define DO_ENGAGE_EVENT1(so, fdset, label) \
15	do { \
16	FD_SET((so)->s, (fdset)); \
17	UPD_NFDS((so)->s); \
18	} while(0)
19
20
21	# define DO_ENGAGE_EVENT2(so, fdset0, fdset1, label) \
22	do { \
23	FD_SET((so)->s, (fdset0)); \
24	FD_SET((so)->s, (fdset1)); \
25	UPD_NFDS((so)->s); \
26	} while(0)
27	#else /* defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) && defined(RT_OS_WINDOWS) */
28	# define DO_ENGAGE_EVENT1(so, fdset0, label) \
29	do { \
30	rc = WSAEventSelect((so)->s, VBOX_SOCKET_EVENT, FD_ALL_EVENTS); \
31	if (rc == SOCKET_ERROR) \
32	{ \
33	/* This should not happen */ \
34	error = WSAGetLastError(); \
35	LogRel(("WSAEventSelector (" #label ") error %d (so=%x, socket=%s, event=%x)\n", \
36	error, (so), (so)->s, VBOX_SOCKET_EVENT)); \
37	} \
38	} while(0)
39
40	# define DO_ENGAGE_EVENT2(so, fdset0, fdset1, label) DO_ENGAGE_EVENT1((so), (fdset0), label)
41	#endif /* defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) && defined(RT_OS_WINDOWS) */
42
43	# define TCP_ENGAGE_EVENT1(so, fdset0) \
44	DO_ENGAGE_EVENT1((so), (fdset0), TCP)
45
46	# define TCP_ENGAGE_EVENT2(so, fdset0, fdset1) \
47	DO_ENGAGE_EVENT2((so), (fdset0), (fdset1), TCP)
48
49	#define UDP_ENGAGE_EVENT(so, fdset) \
50	DO_ENGAGE_EVENT1((so), (fdset), UDP)
51
52	static const uint8_t special_ethaddr[6] = {
53	0x52, 0x54, 0x00, 0x12, 0x35, 0x00
54	};
55
56	#ifdef RT_OS_WINDOWS
57
58	static int get_dns_addr_domain(PNATState pData, bool fVerbose,
59	struct in_addr *pdns_addr,
60	const char **ppszDomain)
61	{
62	int rc = 0;
63	FIXED_INFO *FixedInfo = NULL;
64	ULONG BufLen;
65	DWORD ret;
66	IP_ADDR_STRING *pIPAddr;
67	struct in_addr tmp_addr;
68
69	FixedInfo = (FIXED_INFO *)GlobalAlloc(GPTR, sizeof(FIXED_INFO));
70	BufLen = sizeof(FIXED_INFO);
71
72	/** @todo: this API returns all DNS servers, no matter whether the
73	* corresponding network adapter is disabled or not. Maybe replace
74	* this by GetAdapterAddresses(), which is XP/Vista only though. */
75	if (ERROR_BUFFER_OVERFLOW == GetNetworkParams(FixedInfo, &BufLen))
76	{
77	if (FixedInfo)
78	{
79	GlobalFree(FixedInfo);
80	FixedInfo = NULL;
81	}
82	FixedInfo = GlobalAlloc(GPTR, BufLen);
83	}
84
85	if ((ret = GetNetworkParams(FixedInfo, &BufLen)) != ERROR_SUCCESS)
86	{
87	Log(("GetNetworkParams failed. ret = %08x\n", (u_int)ret ));
88	if (FixedInfo)
89	{
90	GlobalFree(FixedInfo);
91	FixedInfo = NULL;
92	}
93	rc = -1;
94	goto get_dns_prefix;
95	}
96
97	pIPAddr = &(FixedInfo->DnsServerList);
98	inet_aton(pIPAddr->IpAddress.String, &tmp_addr);
99	Log(("nat: DNS Servers:\n"));
100	if (fVerbose \|\| pdns_addr->s_addr != tmp_addr.s_addr)
101	LogRel(("NAT: DNS address: %s\n", pIPAddr->IpAddress.String));
102	*pdns_addr = tmp_addr;
103
104	pIPAddr = FixedInfo -> DnsServerList.Next;
105	while (pIPAddr)
106	{
107	if (fVerbose)
108	LogRel(("NAT: ignored DNS address: %s\n", pIPAddr ->IpAddress.String));
109	pIPAddr = pIPAddr ->Next;
110	}
111	if (FixedInfo)
112	{
113	GlobalFree(FixedInfo);
114	FixedInfo = NULL;
115	}
116
117	get_dns_prefix:
118	if (ppszDomain)
119	{
120	OSVERSIONINFO ver;
121	char szDnsDomain[256];
122	DWORD dwSize = sizeof(szDnsDomain);
123
124	*ppszDomain = NULL;
125	GetVersionEx(&ver);
126	if (ver.dwMajorVersion >= 5)
127	{
128	/* GetComputerNameEx exists in Windows versions starting with 2000. */
129	if (GetComputerNameEx(ComputerNameDnsDomain, szDnsDomain, &dwSize))
130	{
131	if (szDnsDomain[0])
132	{
133	/* Just non-empty strings are valid. */
134	*ppszDomain = RTStrDup(szDnsDomain);
135	if (pData->fPassDomain)
136	{
137	if (fVerbose)
138	LogRel(("NAT: passing domain name %s\n", szDnsDomain));
139	}
140	else
141	Log(("nat: ignoring domain %s\n", szDnsDomain));
142	}
143	}
144	else
145	Log(("nat: GetComputerNameEx failed (%d)\n", GetLastError()));
146	}
147	}
148	return rc;
149	}
150
151	#else
152
153	static int get_dns_addr_domain(PNATState pData, bool fVerbose,
154	struct in_addr *pdns_addr,
155	const char **ppszDomain)
156	{
157	char buff[512];
158	char buff2[256];
159	FILE *f;
160	int found = 0;
161	struct in_addr tmp_addr;
162
163	#ifdef RT_OS_OS2
164	/* Try various locations. */
165	char *etc = getenv("ETC");
166	f = NULL;
167	if (etc)
168	{
169	snprintf(buff, sizeof(buff), "%s/RESOLV2", etc);
170	f = fopen(buff, "rt");
171	}
172	if (!f)
173	{
174	snprintf(buff, sizeof(buff), "%s/RESOLV2", _PATH_ETC);
175	f = fopen(buff, "rt");
176	}
177	if (!f)
178	{
179	snprintf(buff, sizeof(buff), "%s/resolv.conf", _PATH_ETC);
180	f = fopen(buff, "rt");
181	}
182	#else
183	f = fopen("/etc/resolv.conf", "r");
184	#endif
185	if (!f)
186	return -1;
187
188	if (ppszDomain)
189	*ppszDomain = NULL;
190	Log(("nat: DNS Servers:\n"));
191	while (fgets(buff, 512, f) != NULL)
192	{
193	if (sscanf(buff, "nameserver%*[ \t]%256s", buff2) == 1)
194	{
195	if (!inet_aton(buff2, &tmp_addr))
196	continue;
197	if (tmp_addr.s_addr == loopback_addr.s_addr)
198	tmp_addr = our_addr;
199	/* If it's the first one, set it to dns_addr */
200	if (!found)
201	{
202	if (fVerbose \|\| pdns_addr->s_addr != tmp_addr.s_addr)
203	LogRel(("NAT: DNS address: %s\n", buff2));
204	*pdns_addr = tmp_addr;
205	}
206	else
207	{
208	if (fVerbose)
209	LogRel(("NAT: ignored DNS address: %s\n", buff2));
210	}
211	found++;
212	}
213	if ( ppszDomain
214	&& (!strncmp(buff, "domain", 6) \|\| !strncmp(buff, "search", 6)))
215	{
216	/* Domain name/search list present. Pick first entry */
217	if (*ppszDomain == NULL)
218	{
219	char *tok;
220	char *saveptr;
221	tok = strtok_r(&buff[6], " \t\n", &saveptr);
222	if (tok)
223	{
224	*ppszDomain = RTStrDup(tok);
225	if (pData->fPassDomain)
226	{
227	if (fVerbose)
228	LogRel(("NAT: passing domain name %s\n", tok));
229	}
230	else
231	Log(("nat: ignoring domain %s\n", tok));
232	}
233	}
234	}
235	}
236	fclose(f);
237	if (!found)
238	return -1;
239	return 0;
240	}
241
242	#endif
243
244	int get_dns_addr(PNATState pData, struct in_addr *pdns_addr)
245	{
246	return get_dns_addr_domain(pData, false, pdns_addr, NULL);
247	}
248
249	int slirp_init(PNATState ppData, const char pszNetAddr, uint32_t u32Netmask,
250	bool fPassDomain, const char *pszTFTPPrefix,
251	const char pszBootFile, void pvUser)
252	{
253	int fNATfailed = 0;
254	PNATState pData = malloc(sizeof(NATState));
255	*ppData = pData;
256	if (!pData)
257	return VERR_NO_MEMORY;
258	if (u32Netmask & 0x1f)
259	/* CTL is x.x.x.15, bootp passes up to 16 IPs (15..31) */
260	return VERR_INVALID_PARAMETER;
261	memset(pData, '\0', sizeof(NATState));
262	pData->fPassDomain = fPassDomain;
263	pData->pvUser = pvUser;
264	#if ARCH_BITS == 64 && !defined(VBOX_WITH_BSD_REASS)
265	pData->cpvHashUsed = 1;
266	#endif
267	tftp_prefix = pszTFTPPrefix;
268	bootp_filename = pszBootFile;
269	pData->netmask = u32Netmask;
270
271	#ifdef RT_OS_WINDOWS
272	{
273	WSADATA Data;
274	WSAStartup(MAKEWORD(2,0), &Data);
275	}
276	# if defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC)
277	pData->phEvents[VBOX_SOCKET_EVENT_INDEX] = CreateEvent(NULL, FALSE, FALSE, NULL);
278	# endif
279	#endif
280
281	link_up = 1;
282
283	debug_init();
284	if_init(pData);
285	ip_init(pData);
286	#ifdef VBOX_WITH_SLIRP_ICMP
287	icmp_init(pData);
288	#endif /* VBOX_WITH_SLIRP_ICMP */
289
290	/* Initialise mbufs after setting the MTU */
291	m_init(pData);
292
293	/* set default addresses */
294	inet_aton("127.0.0.1", &loopback_addr);
295	inet_aton("127.0.0.1", &dns_addr);
296
297	if (get_dns_addr_domain(pData, true, &dns_addr, &pData->pszDomain) < 0)
298	fNATfailed = 1;
299
300	inet_aton(pszNetAddr, &special_addr);
301	alias_addr.s_addr = special_addr.s_addr \| htonl(CTL_ALIAS);
302	getouraddr(pData);
303	return fNATfailed ? VINF_NAT_DNS : VINF_SUCCESS;
304	}
305
306	/**
307	* Statistics counters.
308	*/
309	void slirp_register_timers(PNATState pData, PPDMDRVINS pDrvIns)
310	{
311	#ifndef VBOX_WITHOUT_RELEASE_STATISTICS
312	PDMDrvHlpSTAMRegisterF(pDrvIns, &pData->StatFill, STAMTYPE_PROFILE, STAMVISIBILITY_ALWAYS,
313	STAMUNIT_TICKS_PER_CALL, "Profiling slirp fills", "/Drivers/NAT%d/Fill", pDrvIns->iInstance);
314	PDMDrvHlpSTAMRegisterF(pDrvIns, &pData->StatPoll, STAMTYPE_PROFILE, STAMVISIBILITY_ALWAYS,
315	STAMUNIT_TICKS_PER_CALL, "Profiling slirp polls", "/Drivers/NAT%d/Poll", pDrvIns->iInstance);
316	PDMDrvHlpSTAMRegisterF(pDrvIns, &pData->StatFastTimer, STAMTYPE_PROFILE, STAMVISIBILITY_ALWAYS,
317	STAMUNIT_TICKS_PER_CALL, "Profiling slirp fast timer", "/Drivers/NAT%d/TimerFast", pDrvIns->iInstance);
318	PDMDrvHlpSTAMRegisterF(pDrvIns, &pData->StatSlowTimer, STAMTYPE_PROFILE, STAMVISIBILITY_ALWAYS,
319	STAMUNIT_TICKS_PER_CALL, "Profiling slirp slow timer", "/Drivers/NAT%d/TimerSlow", pDrvIns->iInstance);
320	PDMDrvHlpSTAMRegisterF(pDrvIns, &pData->StatTCP, STAMTYPE_COUNTER, STAMVISIBILITY_ALWAYS,
321	STAMUNIT_COUNT, "TCP sockets", "/Drivers/NAT%d/SockTCP", pDrvIns->iInstance);
322	PDMDrvHlpSTAMRegisterF(pDrvIns, &pData->StatTCPHot, STAMTYPE_COUNTER, STAMVISIBILITY_ALWAYS,
323	STAMUNIT_COUNT, "TCP sockets active", "/Drivers/NAT%d/SockTCPHot", pDrvIns->iInstance);
324	PDMDrvHlpSTAMRegisterF(pDrvIns, &pData->StatUDP, STAMTYPE_COUNTER, STAMVISIBILITY_ALWAYS,
325	STAMUNIT_COUNT, "UDP sockets", "/Drivers/NAT%d/SockUDP", pDrvIns->iInstance);
326	PDMDrvHlpSTAMRegisterF(pDrvIns, &pData->StatUDPHot, STAMTYPE_COUNTER, STAMVISIBILITY_ALWAYS,
327	STAMUNIT_COUNT, "UDP sockets active", "/Drivers/NAT%d/SockUDPHot", pDrvIns->iInstance);
328	#endif /* VBOX_WITHOUT_RELEASE_STATISTICS */
329	}
330
331	/**
332	* Marks the link as up, making it possible to establish new connections.
333	*/
334	void slirp_link_up(PNATState pData)
335	{
336	link_up = 1;
337	}
338
339	/**
340	* Marks the link as down and cleans up the current connections.
341	*/
342	void slirp_link_down(PNATState pData)
343	{
344	struct socket *so;
345
346	while ((so = tcb.so_next) != &tcb)
347	{
348	if (so->so_state & SS_NOFDREF \|\| so->s == -1)
349	sofree(pData, so);
350	else
351	tcp_drop(pData, sototcpcb(so), 0);
352	}
353
354	while ((so = udb.so_next) != &udb)
355	udp_detach(pData, so);
356
357	link_up = 0;
358	}
359
360	/**
361	* Terminates the slirp component.
362	*/
363	void slirp_term(PNATState pData)
364	{
365	if (pData->pszDomain)
366	RTStrFree((char )(void )pData->pszDomain);
367
368	#if ARCH_BITS == 64 && !defined(VBOX_WITH_BSD_REASS)
369	LogRel(("NAT: cpvHashUsed=%RU32 cpvHashCollisions=%RU32 cpvHashInserts=%RU64 cpvHashDone=%RU64\n",
370	pData->cpvHashUsed, pData->cpvHashCollisions, pData->cpvHashInserts, pData->cpvHashDone));
371	#endif
372
373	#ifdef VBOX_WITH_SLIRP_ICMP
374	# ifdef RT_OS_WINDOWS
375	pData->pfIcmpCloseHandle(pData->icmp_socket.sh);
376	# else
377	closesocket(pData->icmp_socket.s);
378	# endif
379	#endif
380
381	slirp_link_down(pData);
382	#ifdef RT_OS_WINDOWS
383	WSACleanup();
384	#endif
385	#ifdef LOG_ENABLED
386	Log(("\n"
387	"NAT statistics\n"
388	"--------------\n"
389	"\n"));
390	ipstats(pData);
391	tcpstats(pData);
392	udpstats(pData);
393	icmpstats(pData);
394	mbufstats(pData);
395	sockstats(pData);
396	Log(("\n"
397	"\n"
398	"\n"));
399	#endif
400	free(pData);
401	}
402
403
404	#define CONN_CANFSEND(so) (((so)->so_state & (SS_FCANTSENDMORE\|SS_ISFCONNECTED)) == SS_ISFCONNECTED)
405	#define CONN_CANFRCV(so) (((so)->so_state & (SS_FCANTRCVMORE\|SS_ISFCONNECTED)) == SS_ISFCONNECTED)
406	#define UPD_NFDS(x) if (nfds < (x)) nfds = (x)
407
408	/*
409	* curtime kept to an accuracy of 1ms
410	*/
411	#ifdef RT_OS_WINDOWS
412	static void updtime(PNATState pData)
413	{
414	struct _timeb tb;
415
416	_ftime(&tb);
417	curtime = (u_int)tb.time * (u_int)1000;
418	curtime += (u_int)tb.millitm;
419	}
420	#else
421	static void updtime(PNATState pData)
422	{
423	gettimeofday(&tt, 0);
424
425	curtime = (u_int)tt.tv_sec * (u_int)1000;
426	curtime += (u_int)tt.tv_usec / (u_int)1000;
427
428	if ((tt.tv_usec % 1000) >= 500)
429	curtime++;
430	}
431	#endif
432
433	void slirp_select_fill(PNATState pData, int *pnfds,
434	fd_set readfds, fd_set writefds, fd_set *xfds)
435	{
436	struct socket so, so_next;
437	int nfds;
438	#if defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) && defined(RT_OS_WINDOWS)
439	int rc;
440	int error;
441	#endif
442	#ifdef VBOX_WITH_BSD_REASS
443	int i;
444	#endif /* VBOX_WITH_BSD_REASS */
445
446	STAM_REL_PROFILE_START(&pData->StatFill, a);
447
448	nfds = *pnfds;
449
450	/*
451	* First, TCP sockets
452	*/
453	do_slowtimo = 0;
454	if (link_up)
455	{
456	/*
457	* *_slowtimo needs calling if there are IP fragments
458	* in the fragment queue, or there are TCP connections active
459	*/
460	#ifndef VBOX_WITH_BSD_REASS
461	do_slowtimo = ((tcb.so_next != &tcb)
462	\|\| ((struct ipasfrag )&ipq != u32_to_ptr(pData, ipq.next, struct ipasfrag )));
463	#else /* !VBOX_WITH_BSD_REASS */
464	/* XXX:
465	* triggering of fragment expiration should be the same but use new macroses
466	*/
467	do_slowtimo = (tcb.so_next != &tcb);
468	if (!do_slowtimo)
469	{
470	for (i = 0; i < IPREASS_NHASH; i++)
471	{
472	if (!TAILQ_EMPTY(&ipq[i]))
473	{
474	do_slowtimo = 1;
475	break;
476	}
477	}
478	}
479	#endif /* VBOX_WITH_BSD_REASS */
480
481	STAM_REL_COUNTER_RESET(&pData->StatTCP);
482	STAM_REL_COUNTER_RESET(&pData->StatTCPHot);
483
484	for (so = tcb.so_next; so != &tcb; so = so_next)
485	{
486	so_next = so->so_next;
487
488	STAM_REL_COUNTER_INC(&pData->StatTCP);
489
490	/*
491	* See if we need a tcp_fasttimo
492	*/
493	if (time_fasttimo == 0 && so->so_tcpcb->t_flags & TF_DELACK)
494	time_fasttimo = curtime; /* Flag when we want a fasttimo */
495
496	/*
497	* NOFDREF can include still connecting to local-host,
498	* newly socreated() sockets etc. Don't want to select these.
499	*/
500	if (so->so_state & SS_NOFDREF \|\| so->s == -1)
501	continue;
502
503	/*
504	* Set for reading sockets which are accepting
505	*/
506	if (so->so_state & SS_FACCEPTCONN)
507	{
508	STAM_REL_COUNTER_INC(&pData->StatTCPHot);
509	TCP_ENGAGE_EVENT1(so, readfds);
510	continue;
511	}
512
513	/*
514	* Set for writing sockets which are connecting
515	*/
516	if (so->so_state & SS_ISFCONNECTING)
517	{
518	STAM_REL_COUNTER_INC(&pData->StatTCPHot);
519	TCP_ENGAGE_EVENT1(so, writefds);
520	}
521
522	/*
523	* Set for writing if we are connected, can send more, and
524	* we have something to send
525	*/
526	if (CONN_CANFSEND(so) && so->so_rcv.sb_cc)
527	{
528	STAM_REL_COUNTER_INC(&pData->StatTCPHot);
529	TCP_ENGAGE_EVENT1(so, writefds);
530	}
531
532	/*
533	* Set for reading (and urgent data) if we are connected, can
534	* receive more, and we have room for it XXX /2 ?
535	*/
536	if (CONN_CANFRCV(so) && (so->so_snd.sb_cc < (so->so_snd.sb_datalen/2)))
537	{
538	STAM_REL_COUNTER_INC(&pData->StatTCPHot);
539	TCP_ENGAGE_EVENT2(so, readfds, xfds);
540	}
541	}
542
543	/*
544	* UDP sockets
545	*/
546	STAM_REL_COUNTER_RESET(&pData->StatUDP);
547	STAM_REL_COUNTER_RESET(&pData->StatUDPHot);
548
549	for (so = udb.so_next; so != &udb; so = so_next)
550	{
551	so_next = so->so_next;
552
553	STAM_REL_COUNTER_INC(&pData->StatUDP);
554
555	/*
556	* See if it's timed out
557	*/
558	if (so->so_expire)
559	{
560	if (so->so_expire <= curtime)
561	{
562	udp_detach(pData, so);
563	continue;
564	}
565	else
566	do_slowtimo = 1; /* Let socket expire */
567	}
568
569	/*
570	* When UDP packets are received from over the link, they're
571	* sendto()'d straight away, so no need for setting for writing
572	* Limit the number of packets queued by this session to 4.
573	* Note that even though we try and limit this to 4 packets,
574	* the session could have more queued if the packets needed
575	* to be fragmented.
576	*
577	* (XXX <= 4 ?)
578	*/
579	if ((so->so_state & SS_ISFCONNECTED) && so->so_queued <= 4)
580	{
581	STAM_REL_COUNTER_INC(&pData->StatUDPHot);
582	UDP_ENGAGE_EVENT(so, readfds);
583	}
584	}
585	}
586
587	#if !defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) \|\| !defined(RT_OS_WINDOWS)
588	*pnfds = nfds;
589	#else
590	*pnfds = VBOX_EVENT_COUNT;
591	#endif
592
593	STAM_REL_PROFILE_STOP(&pData->StatFill, a);
594	}
595
596	void slirp_select_poll(PNATState pData, fd_set readfds, fd_set writefds, fd_set *xfds)
597	{
598	struct socket so, so_next;
599	int ret;
600	#if defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) && defined(RT_OS_WINDOWS)
601	WSANETWORKEVENTS NetworkEvents;
602	int rc;
603	int error;
604	#endif
605	static uint32_t stat_time;
606
607	STAM_REL_PROFILE_START(&pData->StatPoll, a);
608
609	/* Update time */
610	updtime(pData);
611
612	#ifdef LOG_ENABLED
613	if (curtime - stat_time > 10000)
614	{
615	stat_time = curtime;
616	sockstats(pData);
617	}
618	#endif
619
620	/*
621	* See if anything has timed out
622	*/
623	if (link_up)
624	{
625	if (time_fasttimo && ((curtime - time_fasttimo) >= 2))
626	{
627	STAM_REL_PROFILE_START(&pData->StatFastTimer, a);
628	tcp_fasttimo(pData);
629	time_fasttimo = 0;
630	STAM_REL_PROFILE_STOP(&pData->StatFastTimer, a);
631	}
632	if (do_slowtimo && ((curtime - last_slowtimo) >= 499))
633	{
634	STAM_REL_PROFILE_START(&pData->StatSlowTimer, a);
635	ip_slowtimo(pData);
636	tcp_slowtimo(pData);
637	last_slowtimo = curtime;
638	STAM_REL_PROFILE_STOP(&pData->StatSlowTimer, a);
639	}
640	}
641	#if defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) && defined(RT_OS_WINDOWS)
642	if (!readfds && !writefds && !xfds)
643	return; /* only timer update */
644	#endif
645
646	/*
647	* Check sockets
648	*/
649	if (link_up)
650	{
651	/*
652	* Check TCP sockets
653	*/
654	for (so = tcb.so_next; so != &tcb; so = so_next)
655	{
656	so_next = so->so_next;
657
658	/*
659	* FD_ISSET is meaningless on these sockets
660	* (and they can crash the program)
661	*/
662	if (so->so_state & SS_NOFDREF \|\| so->s == -1)
663	continue;
664
665	#if defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) && defined(RT_OS_WINDOWS)
666	rc = WSAEnumNetworkEvents(so->s, VBOX_SOCKET_EVENT, &NetworkEvents);
667	if (rc == SOCKET_ERROR)
668	{
669	error = WSAGetLastError();
670	LogRel(("WSAEnumNetworkEvents TCP error %d\n", error));
671	continue;
672	}
673	#endif
674
675	/*
676	* Check for URG data
677	* This will soread as well, so no need to
678	* test for readfds below if this succeeds
679	*/
680	#if !defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) \|\| !defined(RT_OS_WINDOWS)
681	if (FD_ISSET(so->s, xfds))
682	#else
683	/* out-of-band data */
684	if ((NetworkEvents.lNetworkEvents & FD_OOB) && NetworkEvents.iErrorCode[FD_OOB_BIT] == 0)
685	#endif
686	{
687	sorecvoob(pData, so);
688	}
689
690	/*
691	* Check sockets for reading
692	*/
693	#if !defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) \|\| !defined(RT_OS_WINDOWS)
694	else if (FD_ISSET(so->s, readfds))
695	#else
696	else if ((NetworkEvents.lNetworkEvents & FD_READ) && (NetworkEvents.iErrorCode[FD_READ_BIT] == 0))
697	#endif
698	{
699	/*
700	* Check for incoming connections
701	*/
702	if (so->so_state & SS_FACCEPTCONN)
703	{
704	tcp_connect(pData, so);
705	#if defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) && defined(RT_OS_WINDOWS)
706	if (!NetworkEvents.lNetworkEvents & FD_CLOSE)
707	#endif
708	continue;
709	}
710
711	ret = soread(pData, so);
712	/* Output it if we read something */
713	if (ret > 0)
714	tcp_output(pData, sototcpcb(so));
715	}
716
717	#if defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) && defined(RT_OS_WINDOWS)
718	/*
719	* Check for FD_CLOSE events.
720	*/
721	if (NetworkEvents.lNetworkEvents & FD_CLOSE)
722	{
723	/*
724	* drain the socket
725	*/
726	for (;;)
727	{
728	ret = soread(pData, so);
729	if (ret > 0)
730	tcp_output(pData, sototcpcb(so));
731	else
732	break;
733	}
734	}
735	#endif
736
737	/*
738	* Check sockets for writing
739	*/
740	#if !defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) \|\| !defined(RT_OS_WINDOWS)
741	if (FD_ISSET(so->s, writefds))
742	#else
743	if ((NetworkEvents.lNetworkEvents & FD_WRITE) && (NetworkEvents.iErrorCode[FD_WRITE_BIT] == 0))
744	#endif
745	{
746	/*
747	* Check for non-blocking, still-connecting sockets
748	*/
749	if (so->so_state & SS_ISFCONNECTING)
750	{
751	/* Connected */
752	so->so_state &= ~SS_ISFCONNECTING;
753
754	/*
755	* This should be probably guarded by PROBE_CONN too. Anyway,
756	* we disable it on OS/2 because the below send call returns
757	* EFAULT which causes the opened TCP socket to close right
758	* after it has been opened and connected.
759	*/
760	#ifndef RT_OS_OS2
761	ret = send(so->s, (const char *)&ret, 0, 0);
762	if (ret < 0)
763	{
764	/* XXXXX Must fix, zero bytes is a NOP */
765	if ( errno == EAGAIN
766	\|\| errno == EWOULDBLOCK
767	\|\| errno == EINPROGRESS
768	\|\| errno == ENOTCONN)
769	continue;
770
771	/* else failed */
772	so->so_state = SS_NOFDREF;
773	}
774	/* else so->so_state &= ~SS_ISFCONNECTING; */
775	#endif
776
777	/*
778	* Continue tcp_input
779	*/
780	tcp_input(pData, (struct mbuf *)NULL, sizeof(struct ip), so);
781	/* continue; */
782	}
783	else
784	ret = sowrite(pData, so);
785	/*
786	* XXX If we wrote something (a lot), there could be the need
787	* for a window update. In the worst case, the remote will send
788	* a window probe to get things going again.
789	*/
790	}
791
792	/*
793	* Probe a still-connecting, non-blocking socket
794	* to check if it's still alive
795	*/
796	#ifdef PROBE_CONN
797	if (so->so_state & SS_ISFCONNECTING)
798	{
799	ret = recv(so->s, (char *)&ret, 0, 0);
800
801	if (ret < 0)
802	{
803	/* XXX */
804	if ( errno == EAGAIN
805	\|\| errno == EWOULDBLOCK
806	\|\| errno == EINPROGRESS
807	\|\| errno == ENOTCONN)
808	{
809	continue; /* Still connecting, continue */
810	}
811
812	/* else failed */
813	so->so_state = SS_NOFDREF;
814
815	/* tcp_input will take care of it */
816	}
817	else
818	{
819	ret = send(so->s, &ret, 0, 0);
820	if (ret < 0)
821	{
822	/* XXX */
823	if ( errno == EAGAIN
824	\|\| errno == EWOULDBLOCK
825	\|\| errno == EINPROGRESS
826	\|\| errno == ENOTCONN)
827	{
828	continue;
829	}
830	/* else failed */
831	so->so_state = SS_NOFDREF;
832	}
833	else
834	so->so_state &= ~SS_ISFCONNECTING;
835
836	}
837	tcp_input((struct mbuf *)NULL, sizeof(struct ip),so);
838	} /* SS_ISFCONNECTING */
839	#endif
840	}
841
842	/*
843	* Now UDP sockets.
844	* Incoming packets are sent straight away, they're not buffered.
845	* Incoming UDP data isn't buffered either.
846	*/
847	for (so = udb.so_next; so != &udb; so = so_next)
848	{
849	so_next = so->so_next;
850
851	#if defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) && defined(RT_OS_WINDOWS)
852	rc = WSAEnumNetworkEvents(so->s, VBOX_SOCKET_EVENT, &NetworkEvents);
853	if (rc == SOCKET_ERROR)
854	{
855	error = WSAGetLastError();
856	LogRel(("WSAEnumNetworkEvents TCP error %d\n", error));
857	continue;
858	}
859	#endif
860	#if !defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) \|\| !defined(RT_OS_WINDOWS)
861	if (so->s != -1 && FD_ISSET(so->s, readfds))
862	#else
863	if (((NetworkEvents.lNetworkEvents & FD_READ) && (NetworkEvents.iErrorCode[FD_READ_BIT] == 0)))
864	#endif
865	{
866	sorecvfrom(pData, so);
867	}
868	}
869	#if defined(VBOX_WITH_SLIRP_ICMP) && defined(RT_OS_WINDOWS)
870	sorecvfrom(pData, &pData->icmp_socket);
871	#endif
872	}
873
874	/*
875	* See if we can start outputting
876	*/
877	if (if_queued && link_up)
878	if_start(pData);
879
880	STAM_REL_PROFILE_STOP(&pData->StatPoll, a);
881	}
882
883	#define ETH_ALEN 6
884	#define ETH_HLEN 14
885
886	#define ETH_P_IP 0x0800 /* Internet Protocol packet */
887	#define ETH_P_ARP 0x0806 /* Address Resolution packet */
888
889	#define ARPOP_REQUEST 1 /* ARP request */
890	#define ARPOP_REPLY 2 /* ARP reply */
891
892	struct ethhdr
893	{
894	unsigned char h_dest[ETH_ALEN]; /* destination eth addr */
895	unsigned char h_source[ETH_ALEN]; /* source ether addr */
896	unsigned short h_proto; /* packet type ID field */
897	};
898
899	struct arphdr
900	{
901	unsigned short ar_hrd; /* format of hardware address */
902	unsigned short ar_pro; /* format of protocol address */
903	unsigned char ar_hln; /* length of hardware address */
904	unsigned char ar_pln; /* length of protocol address */
905	unsigned short ar_op; /* ARP opcode (command) */
906
907	/*
908	* Ethernet looks like this : This bit is variable sized however...
909	*/
910	unsigned char ar_sha[ETH_ALEN]; /* sender hardware address */
911	unsigned char ar_sip[4]; /* sender IP address */
912	unsigned char ar_tha[ETH_ALEN]; /* target hardware address */
913	unsigned char ar_tip[4]; /* target IP address */
914	};
915
916	static
917	void arp_input(PNATState pData, const uint8_t *pkt, int pkt_len)
918	{
919	struct ethhdr eh = (struct ethhdr )pkt;
920	struct arphdr ah = (struct arphdr )(pkt + ETH_HLEN);
921	uint8_t arp_reply[ETH_HLEN + sizeof(struct arphdr)];
922	struct ethhdr reh = (struct ethhdr )arp_reply;
923	struct arphdr rah = (struct arphdr )(arp_reply + ETH_HLEN);
924	int ar_op;
925	struct ex_list *ex_ptr;
926	uint32_t htip = ntohl((uint32_t)ah->ar_tip);
927
928	ar_op = ntohs(ah->ar_op);
929	switch(ar_op)
930	{
931	case ARPOP_REQUEST:
932	if ((htip & pData->netmask) == ntohl(special_addr.s_addr))
933	{
934	if ( (htip & ~pData->netmask) == CTL_DNS
935	\|\| (htip & ~pData->netmask) == CTL_ALIAS)
936	goto arp_ok;
937	for (ex_ptr = exec_list; ex_ptr; ex_ptr = ex_ptr->ex_next)
938	{
939	if ((htip & ~pData->netmask) == ex_ptr->ex_addr)
940	goto arp_ok;
941	}
942	return;
943	arp_ok:
944	/* XXX: make an ARP request to have the client address */
945	memcpy(client_ethaddr, eh->h_source, ETH_ALEN);
946
947	/* ARP request for alias/dns mac address */
948	memcpy(reh->h_dest, pkt + ETH_ALEN, ETH_ALEN);
949	memcpy(reh->h_source, special_ethaddr, ETH_ALEN - 1);
950	reh->h_source[5] = ah->ar_tip[3];
951	reh->h_proto = htons(ETH_P_ARP);
952
953	rah->ar_hrd = htons(1);
954	rah->ar_pro = htons(ETH_P_IP);
955	rah->ar_hln = ETH_ALEN;
956	rah->ar_pln = 4;
957	rah->ar_op = htons(ARPOP_REPLY);
958	memcpy(rah->ar_sha, reh->h_source, ETH_ALEN);
959	memcpy(rah->ar_sip, ah->ar_tip, 4);
960	memcpy(rah->ar_tha, ah->ar_sha, ETH_ALEN);
961	memcpy(rah->ar_tip, ah->ar_sip, 4);
962	slirp_output(pData->pvUser, arp_reply, sizeof(arp_reply));
963	}
964	break;
965	default:
966	break;
967	}
968	}
969
970	void slirp_input(PNATState pData, const uint8_t *pkt, int pkt_len)
971	{
972	struct mbuf *m;
973	int proto;
974
975	if (pkt_len < ETH_HLEN)
976	return;
977
978	proto = ntohs((uint16_t )(pkt + 12));
979	switch(proto)
980	{
981	case ETH_P_ARP:
982	arp_input(pData, pkt, pkt_len);
983	break;
984	case ETH_P_IP:
985	/* Update time. Important if the network is very quiet, as otherwise
986	* the first outgoing connection gets an incorrect timestamp. */
987	updtime(pData);
988
989	m = m_get(pData);
990	if (!m)
991	return;
992	/* Note: we add to align the IP header */
993	if (M_FREEROOM(m) < pkt_len + 2)
994	{
995	m_inc(m, pkt_len + 2);
996	}
997	m->m_len = pkt_len + 2;
998	memcpy(m->m_data + 2, pkt, pkt_len);
999
1000	m->m_data += 2 + ETH_HLEN;
1001	m->m_len -= 2 + ETH_HLEN;
1002
1003	ip_input(pData, m);
1004	break;
1005	default:
1006	break;
1007	}
1008	}
1009
1010	/* output the IP packet to the ethernet device */
1011	void if_encap(PNATState pData, const uint8_t *ip_data, int ip_data_len)
1012	{
1013	uint8_t buf[1600];
1014	struct ethhdr eh = (struct ethhdr )buf;
1015
1016	if (ip_data_len + ETH_HLEN > sizeof(buf))
1017	return;
1018
1019	memcpy(eh->h_dest, client_ethaddr, ETH_ALEN);
1020	memcpy(eh->h_source, special_ethaddr, ETH_ALEN - 1);
1021	/* XXX: not correct */
1022	eh->h_source[5] = CTL_ALIAS;
1023	eh->h_proto = htons(ETH_P_IP);
1024	memcpy(buf + sizeof(struct ethhdr), ip_data, ip_data_len);
1025	slirp_output(pData->pvUser, buf, ip_data_len + ETH_HLEN);
1026	}
1027
1028	int slirp_redir(PNATState pData, int is_udp, int host_port,
1029	struct in_addr guest_addr, int guest_port)
1030	{
1031	if (is_udp)
1032	{
1033	if (!udp_listen(pData, htons(host_port), guest_addr.s_addr,
1034	htons(guest_port), 0))
1035	return -1;
1036	}
1037	else
1038	{
1039	if (!solisten(pData, htons(host_port), guest_addr.s_addr,
1040	htons(guest_port), 0))
1041	return -1;
1042	}
1043	return 0;
1044	}
1045
1046	int slirp_add_exec(PNATState pData, int do_pty, const char *args, int addr_low_byte,
1047	int guest_port)
1048	{
1049	return add_exec(&exec_list, do_pty, (char *)args,
1050	addr_low_byte, htons(guest_port));
1051	}
1052
1053	void slirp_set_ethaddr(PNATState pData, const uint8_t *ethaddr)
1054	{
1055	memcpy(client_ethaddr, ethaddr, ETH_ALEN);
1056	}
1057
1058	#if defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) && defined(RT_OS_WINDOWS)
1059	HANDLE *slirp_get_events(PNATState pData)
1060	{
1061	return pData->phEvents;
1062	}
1063	void slirp_register_external_event(PNATState pData, HANDLE hEvent, int index)
1064	{
1065	pData->phEvents[index] = hEvent;
1066	}
1067	#endif
1068
1069	unsigned int slirp_get_timeout_ms(PNATState pData)
1070	{
1071	if (link_up)
1072	{
1073	if (time_fasttimo)
1074	return 2;
1075	if (do_slowtimo)
1076	return 500; /* see PR_SLOWHZ */
1077	}
1078	return 0;
1079	}

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

source: vbox/trunk/src/VBox/Devices/Network/slirp/slirp.c@ 15194

Download in other formats: