slirp.c@ 15106

Last change on this file since 15106 was 15095, checked in by vboxsync, 16 years ago
slirp: tabs
Property svn:eol-style set to `native`
File size: 33.3 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	slirp_link_down(pData);
374	#ifdef RT_OS_WINDOWS
375	WSACleanup();
376	#endif
377	#ifdef LOG_ENABLED
378	Log(("\n"
379	"NAT statistics\n"
380	"--------------\n"
381	"\n"));
382	ipstats(pData);
383	tcpstats(pData);
384	udpstats(pData);
385	icmpstats(pData);
386	mbufstats(pData);
387	sockstats(pData);
388	Log(("\n"
389	"\n"
390	"\n"));
391	#endif
392	free(pData);
393	}
394
395
396	#define CONN_CANFSEND(so) (((so)->so_state & (SS_FCANTSENDMORE\|SS_ISFCONNECTED)) == SS_ISFCONNECTED)
397	#define CONN_CANFRCV(so) (((so)->so_state & (SS_FCANTRCVMORE\|SS_ISFCONNECTED)) == SS_ISFCONNECTED)
398	#define UPD_NFDS(x) if (nfds < (x)) nfds = (x)
399
400	/*
401	* curtime kept to an accuracy of 1ms
402	*/
403	#ifdef RT_OS_WINDOWS
404	static void updtime(PNATState pData)
405	{
406	struct _timeb tb;
407
408	_ftime(&tb);
409	curtime = (u_int)tb.time * (u_int)1000;
410	curtime += (u_int)tb.millitm;
411	}
412	#else
413	static void updtime(PNATState pData)
414	{
415	gettimeofday(&tt, 0);
416
417	curtime = (u_int)tt.tv_sec * (u_int)1000;
418	curtime += (u_int)tt.tv_usec / (u_int)1000;
419
420	if ((tt.tv_usec % 1000) >= 500)
421	curtime++;
422	}
423	#endif
424
425	void slirp_select_fill(PNATState pData, int *pnfds,
426	fd_set readfds, fd_set writefds, fd_set *xfds)
427	{
428	struct socket so, so_next;
429	int nfds;
430	#if defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) && defined(RT_OS_WINDOWS)
431	int rc;
432	int error;
433	#endif
434	#ifdef VBOX_WITH_BSD_REASS
435	int i;
436	#endif /* VBOX_WITH_BSD_REASS */
437
438	STAM_REL_PROFILE_START(&pData->StatFill, a);
439
440	nfds = *pnfds;
441
442	/*
443	* First, TCP sockets
444	*/
445	do_slowtimo = 0;
446	if (link_up)
447	{
448	/*
449	* *_slowtimo needs calling if there are IP fragments
450	* in the fragment queue, or there are TCP connections active
451	*/
452	#ifndef VBOX_WITH_BSD_REASS
453	do_slowtimo = ((tcb.so_next != &tcb)
454	\|\| ((struct ipasfrag )&ipq != u32_to_ptr(pData, ipq.next, struct ipasfrag )));
455	#else /* !VBOX_WITH_BSD_REASS */
456	/* XXX:
457	* triggering of fragment expiration should be the same but use new macroses
458	*/
459	do_slowtimo = (tcb.so_next != &tcb);
460	if (!do_slowtimo)
461	{
462	for (i = 0; i < IPREASS_NHASH; i++)
463	{
464	if (!TAILQ_EMPTY(&ipq[i]))
465	{
466	do_slowtimo = 1;
467	break;
468	}
469	}
470	}
471	#endif /* VBOX_WITH_BSD_REASS */
472
473	STAM_REL_COUNTER_RESET(&pData->StatTCP);
474	STAM_REL_COUNTER_RESET(&pData->StatTCPHot);
475
476	for (so = tcb.so_next; so != &tcb; so = so_next)
477	{
478	so_next = so->so_next;
479
480	STAM_REL_COUNTER_INC(&pData->StatTCP);
481
482	/*
483	* See if we need a tcp_fasttimo
484	*/
485	if (time_fasttimo == 0 && so->so_tcpcb->t_flags & TF_DELACK)
486	time_fasttimo = curtime; /* Flag when we want a fasttimo */
487
488	/*
489	* NOFDREF can include still connecting to local-host,
490	* newly socreated() sockets etc. Don't want to select these.
491	*/
492	if (so->so_state & SS_NOFDREF \|\| so->s == -1)
493	continue;
494
495	/*
496	* Set for reading sockets which are accepting
497	*/
498	if (so->so_state & SS_FACCEPTCONN)
499	{
500	STAM_REL_COUNTER_INC(&pData->StatTCPHot);
501	TCP_ENGAGE_EVENT1(so, readfds);
502	continue;
503	}
504
505	/*
506	* Set for writing sockets which are connecting
507	*/
508	if (so->so_state & SS_ISFCONNECTING)
509	{
510	STAM_REL_COUNTER_INC(&pData->StatTCPHot);
511	TCP_ENGAGE_EVENT1(so, writefds);
512	}
513
514	/*
515	* Set for writing if we are connected, can send more, and
516	* we have something to send
517	*/
518	if (CONN_CANFSEND(so) && so->so_rcv.sb_cc)
519	{
520	STAM_REL_COUNTER_INC(&pData->StatTCPHot);
521	TCP_ENGAGE_EVENT1(so, writefds);
522	}
523
524	/*
525	* Set for reading (and urgent data) if we are connected, can
526	* receive more, and we have room for it XXX /2 ?
527	*/
528	if (CONN_CANFRCV(so) && (so->so_snd.sb_cc < (so->so_snd.sb_datalen/2)))
529	{
530	STAM_REL_COUNTER_INC(&pData->StatTCPHot);
531	TCP_ENGAGE_EVENT2(so, readfds, xfds);
532	}
533	}
534
535	/*
536	* UDP sockets
537	*/
538	STAM_REL_COUNTER_RESET(&pData->StatUDP);
539	STAM_REL_COUNTER_RESET(&pData->StatUDPHot);
540
541	for (so = udb.so_next; so != &udb; so = so_next)
542	{
543	so_next = so->so_next;
544
545	STAM_REL_COUNTER_INC(&pData->StatUDP);
546
547	/*
548	* See if it's timed out
549	*/
550	if (so->so_expire)
551	{
552	if (so->so_expire <= curtime)
553	{
554	udp_detach(pData, so);
555	continue;
556	}
557	else
558	do_slowtimo = 1; /* Let socket expire */
559	}
560
561	/*
562	* When UDP packets are received from over the link, they're
563	* sendto()'d straight away, so no need for setting for writing
564	* Limit the number of packets queued by this session to 4.
565	* Note that even though we try and limit this to 4 packets,
566	* the session could have more queued if the packets needed
567	* to be fragmented.
568	*
569	* (XXX <= 4 ?)
570	*/
571	if ((so->so_state & SS_ISFCONNECTED) && so->so_queued <= 4)
572	{
573	STAM_REL_COUNTER_INC(&pData->StatUDPHot);
574	UDP_ENGAGE_EVENT(so, readfds);
575	}
576	}
577	}
578
579	#if !defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) \|\| !defined(RT_OS_WINDOWS)
580	*pnfds = nfds;
581	#else
582	*pnfds = VBOX_EVENT_COUNT;
583	#endif
584
585	STAM_REL_PROFILE_STOP(&pData->StatFill, a);
586	}
587
588	void slirp_select_poll(PNATState pData, fd_set readfds, fd_set writefds, fd_set *xfds)
589	{
590	struct socket so, so_next;
591	int ret;
592	#if defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) && defined(RT_OS_WINDOWS)
593	WSANETWORKEVENTS NetworkEvents;
594	int rc;
595	int error;
596	#endif
597	static uint32_t stat_time;
598
599	STAM_REL_PROFILE_START(&pData->StatPoll, a);
600
601	/* Update time */
602	updtime(pData);
603
604	#ifdef LOG_ENABLED
605	if (curtime - stat_time > 10000)
606	{
607	stat_time = curtime;
608	sockstats(pData);
609	}
610	#endif
611
612	/*
613	* See if anything has timed out
614	*/
615	if (link_up)
616	{
617	if (time_fasttimo && ((curtime - time_fasttimo) >= 2))
618	{
619	STAM_REL_PROFILE_START(&pData->StatFastTimer, a);
620	tcp_fasttimo(pData);
621	time_fasttimo = 0;
622	STAM_REL_PROFILE_STOP(&pData->StatFastTimer, a);
623	}
624	if (do_slowtimo && ((curtime - last_slowtimo) >= 499))
625	{
626	STAM_REL_PROFILE_START(&pData->StatSlowTimer, a);
627	ip_slowtimo(pData);
628	tcp_slowtimo(pData);
629	last_slowtimo = curtime;
630	STAM_REL_PROFILE_STOP(&pData->StatSlowTimer, a);
631	}
632	}
633	#if defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) && defined(RT_OS_WINDOWS)
634	if (!readfds && !writefds && !xfds)
635	return; /* only timer update */
636	#endif
637
638	/*
639	* Check sockets
640	*/
641	if (link_up)
642	{
643	/*
644	* Check TCP sockets
645	*/
646	for (so = tcb.so_next; so != &tcb; so = so_next)
647	{
648	so_next = so->so_next;
649
650	/*
651	* FD_ISSET is meaningless on these sockets
652	* (and they can crash the program)
653	*/
654	if (so->so_state & SS_NOFDREF \|\| so->s == -1)
655	continue;
656
657	#if defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) && defined(RT_OS_WINDOWS)
658	rc = WSAEnumNetworkEvents(so->s, VBOX_SOCKET_EVENT, &NetworkEvents);
659	if (rc == SOCKET_ERROR)
660	{
661	error = WSAGetLastError();
662	LogRel(("WSAEnumNetworkEvents TCP error %d\n", error));
663	continue;
664	}
665	#endif
666
667	/*
668	* Check for URG data
669	* This will soread as well, so no need to
670	* test for readfds below if this succeeds
671	*/
672	#if !defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) \|\| !defined(RT_OS_WINDOWS)
673	if (FD_ISSET(so->s, xfds))
674	#else
675	/* out-of-band data */
676	if ((NetworkEvents.lNetworkEvents & FD_OOB) && NetworkEvents.iErrorCode[FD_OOB_BIT] == 0)
677	#endif
678	{
679	sorecvoob(pData, so);
680	}
681
682	/*
683	* Check sockets for reading
684	*/
685	#if !defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) \|\| !defined(RT_OS_WINDOWS)
686	else if (FD_ISSET(so->s, readfds))
687	#else
688	else if ((NetworkEvents.lNetworkEvents & FD_READ) && (NetworkEvents.iErrorCode[FD_READ_BIT] == 0))
689	#endif
690	{
691	/*
692	* Check for incoming connections
693	*/
694	if (so->so_state & SS_FACCEPTCONN)
695	{
696	tcp_connect(pData, so);
697	#if defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) && defined(RT_OS_WINDOWS)
698	if (!NetworkEvents.lNetworkEvents & FD_CLOSE)
699	#endif
700	continue;
701	}
702
703	ret = soread(pData, so);
704	/* Output it if we read something */
705	if (ret > 0)
706	tcp_output(pData, sototcpcb(so));
707	}
708
709	#if defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) && defined(RT_OS_WINDOWS)
710	/*
711	* Check for FD_CLOSE events.
712	*/
713	if (NetworkEvents.lNetworkEvents & FD_CLOSE)
714	{
715	/*
716	* drain the socket
717	*/
718	for (;;)
719	{
720	ret = soread(pData, so);
721	if (ret > 0)
722	tcp_output(pData, sototcpcb(so));
723	else
724	break;
725	}
726	}
727	#endif
728
729	/*
730	* Check sockets for writing
731	*/
732	#if !defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) \|\| !defined(RT_OS_WINDOWS)
733	if (FD_ISSET(so->s, writefds))
734	#else
735	if ((NetworkEvents.lNetworkEvents & FD_WRITE) && (NetworkEvents.iErrorCode[FD_WRITE_BIT] == 0))
736	#endif
737	{
738	/*
739	* Check for non-blocking, still-connecting sockets
740	*/
741	if (so->so_state & SS_ISFCONNECTING)
742	{
743	/* Connected */
744	so->so_state &= ~SS_ISFCONNECTING;
745
746	/*
747	* This should be probably guarded by PROBE_CONN too. Anyway,
748	* we disable it on OS/2 because the below send call returns
749	* EFAULT which causes the opened TCP socket to close right
750	* after it has been opened and connected.
751	*/
752	#ifndef RT_OS_OS2
753	ret = send(so->s, (const char *)&ret, 0, 0);
754	if (ret < 0)
755	{
756	/* XXXXX Must fix, zero bytes is a NOP */
757	if ( errno == EAGAIN
758	\|\| errno == EWOULDBLOCK
759	\|\| errno == EINPROGRESS
760	\|\| errno == ENOTCONN)
761	continue;
762
763	/* else failed */
764	so->so_state = SS_NOFDREF;
765	}
766	/* else so->so_state &= ~SS_ISFCONNECTING; */
767	#endif
768
769	/*
770	* Continue tcp_input
771	*/
772	tcp_input(pData, (struct mbuf *)NULL, sizeof(struct ip), so);
773	/* continue; */
774	}
775	else
776	ret = sowrite(pData, so);
777	/*
778	* XXX If we wrote something (a lot), there could be the need
779	* for a window update. In the worst case, the remote will send
780	* a window probe to get things going again.
781	*/
782	}
783
784	/*
785	* Probe a still-connecting, non-blocking socket
786	* to check if it's still alive
787	*/
788	#ifdef PROBE_CONN
789	if (so->so_state & SS_ISFCONNECTING)
790	{
791	ret = recv(so->s, (char *)&ret, 0, 0);
792
793	if (ret < 0)
794	{
795	/* XXX */
796	if ( errno == EAGAIN
797	\|\| errno == EWOULDBLOCK
798	\|\| errno == EINPROGRESS
799	\|\| errno == ENOTCONN)
800	{
801	continue; /* Still connecting, continue */
802	}
803
804	/* else failed */
805	so->so_state = SS_NOFDREF;
806
807	/* tcp_input will take care of it */
808	}
809	else
810	{
811	ret = send(so->s, &ret, 0, 0);
812	if (ret < 0)
813	{
814	/* XXX */
815	if ( errno == EAGAIN
816	\|\| errno == EWOULDBLOCK
817	\|\| errno == EINPROGRESS
818	\|\| errno == ENOTCONN)
819	{
820	continue;
821	}
822	/* else failed */
823	so->so_state = SS_NOFDREF;
824	}
825	else
826	so->so_state &= ~SS_ISFCONNECTING;
827
828	}
829	tcp_input((struct mbuf *)NULL, sizeof(struct ip),so);
830	} /* SS_ISFCONNECTING */
831	#endif
832	}
833
834	/*
835	* Now UDP sockets.
836	* Incoming packets are sent straight away, they're not buffered.
837	* Incoming UDP data isn't buffered either.
838	*/
839	for (so = udb.so_next; so != &udb; so = so_next)
840	{
841	so_next = so->so_next;
842
843	#if defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) && defined(RT_OS_WINDOWS)
844	rc = WSAEnumNetworkEvents(so->s, VBOX_SOCKET_EVENT, &NetworkEvents);
845	if (rc == SOCKET_ERROR)
846	{
847	error = WSAGetLastError();
848	LogRel(("WSAEnumNetworkEvents TCP error %d\n", error));
849	continue;
850	}
851	#endif
852	#if !defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) \|\| !defined(RT_OS_WINDOWS)
853	if (so->s != -1 && FD_ISSET(so->s, readfds))
854	#else
855	if (((NetworkEvents.lNetworkEvents & FD_READ) && (NetworkEvents.iErrorCode[FD_READ_BIT] == 0)))
856	#endif
857	{
858	sorecvfrom(pData, so);
859	}
860	}
861	#if defined(VBOX_WITH_SLIRP_ICMP) && defined(RT_OS_WINDOWS)
862	sorecvfrom(pData, &pData->icmp_socket);
863	#endif
864	}
865
866	/*
867	* See if we can start outputting
868	*/
869	if (if_queued && link_up)
870	if_start(pData);
871
872	STAM_REL_PROFILE_STOP(&pData->StatPoll, a);
873	}
874
875	#define ETH_ALEN 6
876	#define ETH_HLEN 14
877
878	#define ETH_P_IP 0x0800 /* Internet Protocol packet */
879	#define ETH_P_ARP 0x0806 /* Address Resolution packet */
880
881	#define ARPOP_REQUEST 1 /* ARP request */
882	#define ARPOP_REPLY 2 /* ARP reply */
883
884	struct ethhdr
885	{
886	unsigned char h_dest[ETH_ALEN]; /* destination eth addr */
887	unsigned char h_source[ETH_ALEN]; /* source ether addr */
888	unsigned short h_proto; /* packet type ID field */
889	};
890
891	struct arphdr
892	{
893	unsigned short ar_hrd; /* format of hardware address */
894	unsigned short ar_pro; /* format of protocol address */
895	unsigned char ar_hln; /* length of hardware address */
896	unsigned char ar_pln; /* length of protocol address */
897	unsigned short ar_op; /* ARP opcode (command) */
898
899	/*
900	* Ethernet looks like this : This bit is variable sized however...
901	*/
902	unsigned char ar_sha[ETH_ALEN]; /* sender hardware address */
903	unsigned char ar_sip[4]; /* sender IP address */
904	unsigned char ar_tha[ETH_ALEN]; /* target hardware address */
905	unsigned char ar_tip[4]; /* target IP address */
906	};
907
908	static
909	void arp_input(PNATState pData, const uint8_t *pkt, int pkt_len)
910	{
911	struct ethhdr eh = (struct ethhdr )pkt;
912	struct arphdr ah = (struct arphdr )(pkt + ETH_HLEN);
913	uint8_t arp_reply[ETH_HLEN + sizeof(struct arphdr)];
914	struct ethhdr reh = (struct ethhdr )arp_reply;
915	struct arphdr rah = (struct arphdr )(arp_reply + ETH_HLEN);
916	int ar_op;
917	struct ex_list *ex_ptr;
918	uint32_t htip = ntohl((uint32_t)ah->ar_tip);
919
920	ar_op = ntohs(ah->ar_op);
921	switch(ar_op)
922	{
923	case ARPOP_REQUEST:
924	if ((htip & pData->netmask) == ntohl(special_addr.s_addr))
925	{
926	if ( (htip & ~pData->netmask) == CTL_DNS
927	\|\| (htip & ~pData->netmask) == CTL_ALIAS)
928	goto arp_ok;
929	for (ex_ptr = exec_list; ex_ptr; ex_ptr = ex_ptr->ex_next)
930	{
931	if ((htip & ~pData->netmask) == ex_ptr->ex_addr)
932	goto arp_ok;
933	}
934	return;
935	arp_ok:
936	/* XXX: make an ARP request to have the client address */
937	memcpy(client_ethaddr, eh->h_source, ETH_ALEN);
938
939	/* ARP request for alias/dns mac address */
940	memcpy(reh->h_dest, pkt + ETH_ALEN, ETH_ALEN);
941	memcpy(reh->h_source, special_ethaddr, ETH_ALEN - 1);
942	reh->h_source[5] = ah->ar_tip[3];
943	reh->h_proto = htons(ETH_P_ARP);
944
945	rah->ar_hrd = htons(1);
946	rah->ar_pro = htons(ETH_P_IP);
947	rah->ar_hln = ETH_ALEN;
948	rah->ar_pln = 4;
949	rah->ar_op = htons(ARPOP_REPLY);
950	memcpy(rah->ar_sha, reh->h_source, ETH_ALEN);
951	memcpy(rah->ar_sip, ah->ar_tip, 4);
952	memcpy(rah->ar_tha, ah->ar_sha, ETH_ALEN);
953	memcpy(rah->ar_tip, ah->ar_sip, 4);
954	slirp_output(pData->pvUser, arp_reply, sizeof(arp_reply));
955	}
956	break;
957	default:
958	break;
959	}
960	}
961
962	void slirp_input(PNATState pData, const uint8_t *pkt, int pkt_len)
963	{
964	struct mbuf *m;
965	int proto;
966
967	if (pkt_len < ETH_HLEN)
968	return;
969
970	proto = ntohs((uint16_t )(pkt + 12));
971	switch(proto)
972	{
973	case ETH_P_ARP:
974	arp_input(pData, pkt, pkt_len);
975	break;
976	case ETH_P_IP:
977	/* Update time. Important if the network is very quiet, as otherwise
978	* the first outgoing connection gets an incorrect timestamp. */
979	updtime(pData);
980
981	m = m_get(pData);
982	if (!m)
983	return;
984	/* Note: we add to align the IP header */
985	if (M_FREEROOM(m) < pkt_len + 2)
986	{
987	m_inc(m, pkt_len + 2);
988	}
989	m->m_len = pkt_len + 2;
990	memcpy(m->m_data + 2, pkt, pkt_len);
991
992	m->m_data += 2 + ETH_HLEN;
993	m->m_len -= 2 + ETH_HLEN;
994
995	ip_input(pData, m);
996	break;
997	default:
998	break;
999	}
1000	}
1001
1002	/* output the IP packet to the ethernet device */
1003	void if_encap(PNATState pData, const uint8_t *ip_data, int ip_data_len)
1004	{
1005	uint8_t buf[1600];
1006	struct ethhdr eh = (struct ethhdr )buf;
1007
1008	if (ip_data_len + ETH_HLEN > sizeof(buf))
1009	return;
1010
1011	memcpy(eh->h_dest, client_ethaddr, ETH_ALEN);
1012	memcpy(eh->h_source, special_ethaddr, ETH_ALEN - 1);
1013	/* XXX: not correct */
1014	eh->h_source[5] = CTL_ALIAS;
1015	eh->h_proto = htons(ETH_P_IP);
1016	memcpy(buf + sizeof(struct ethhdr), ip_data, ip_data_len);
1017	slirp_output(pData->pvUser, buf, ip_data_len + ETH_HLEN);
1018	}
1019
1020	int slirp_redir(PNATState pData, int is_udp, int host_port,
1021	struct in_addr guest_addr, int guest_port)
1022	{
1023	if (is_udp)
1024	{
1025	if (!udp_listen(pData, htons(host_port), guest_addr.s_addr,
1026	htons(guest_port), 0))
1027	return -1;
1028	}
1029	else
1030	{
1031	if (!solisten(pData, htons(host_port), guest_addr.s_addr,
1032	htons(guest_port), 0))
1033	return -1;
1034	}
1035	return 0;
1036	}
1037
1038	int slirp_add_exec(PNATState pData, int do_pty, const char *args, int addr_low_byte,
1039	int guest_port)
1040	{
1041	return add_exec(&exec_list, do_pty, (char *)args,
1042	addr_low_byte, htons(guest_port));
1043	}
1044
1045	void slirp_set_ethaddr(PNATState pData, const uint8_t *ethaddr)
1046	{
1047	memcpy(client_ethaddr, ethaddr, ETH_ALEN);
1048	}
1049
1050	#if defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) && defined(RT_OS_WINDOWS)
1051	HANDLE *slirp_get_events(PNATState pData)
1052	{
1053	return pData->phEvents;
1054	}
1055	void slirp_register_external_event(PNATState pData, HANDLE hEvent, int index)
1056	{
1057	pData->phEvents[index] = hEvent;
1058	}
1059	#endif
1060
1061	unsigned int slirp_get_timeout_ms(PNATState pData)
1062	{
1063	if (link_up)
1064	{
1065	if (time_fasttimo)
1066	return 2;
1067	if (do_slowtimo)
1068	return 500; /* see PR_SLOWHZ */
1069	}
1070	return 0;
1071	}

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

source: vbox/trunk/src/VBox/Devices/Network/slirp/slirp.c@ 15106

Download in other formats: