slirp.c@ 15203

Last change on this file since 15203 was 15203, checked in by vboxsync, 16 years ago
slirp: comments, cosmetics
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	# ifdef VBOX_WITH_SLIRP_ICMP
16	# define ICMP_ENGAGE_EVENT(so, fdset1, fdset2) \
17	DO_ENGAGE_EVENT2((so), (fdset1), (fdset2), ICMP)
18	# else /* !VBOX_WITH_SLIRP_ICMP */
19	# define ICMP_ENGAGE_EVENT(so, fdset1, fdset2) \
20	/* no ICMP socket */
21	#endif /* !VBOX_WITH_SLIRP_ICMP */
22
23	# define DO_ENGAGE_EVENT1(so, fdset, label) \
24	do { \
25	FD_SET((so)->s, (fdset)); \
26	UPD_NFDS((so)->s); \
27	} while(0)
28
29
30	# define DO_ENGAGE_EVENT2(so, fdset1, fdset2, label) \
31	do { \
32	FD_SET((so)->s, (fdset1)); \
33	FD_SET((so)->s, (fdset2)); \
34	UPD_NFDS((so)->s); \
35	} while(0)
36
37	#else /* defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) && defined(RT_OS_WINDOWS) */
38
39	/*
40	* On Windows, we will be notified by IcmpSendEcho2() when the response arrives.
41	* So no call to WSAEventSelect necessary.
42	*/
43	# define ICMP_ENGAGE_EVENT(so, fdset1, fdset2) \
44	/* ignore */
45
46	# define DO_ENGAGE_EVENT1(so, fdset1, label) \
47	do { \
48	rc = WSAEventSelect((so)->s, VBOX_SOCKET_EVENT, FD_ALL_EVENTS); \
49	if (rc == SOCKET_ERROR) \
50	{ \
51	/* This should not happen */ \
52	error = WSAGetLastError(); \
53	LogRel(("WSAEventSelector (" #label ") error %d (so=%x, socket=%s, event=%x)\n", \
54	error, (so), (so)->s, VBOX_SOCKET_EVENT)); \
55	} \
56	} while(0)
57
58	# define DO_ENGAGE_EVENT2(so, fdset1, fdset2, label) \
59	O_ENGAGE_EVENT1((so), (fdset1), label)
60
61	#endif /* defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) && defined(RT_OS_WINDOWS) */
62
63	#define TCP_ENGAGE_EVENT1(so, fdset) \
64	DO_ENGAGE_EVENT1((so), (fdset), TCP)
65
66	#define TCP_ENGAGE_EVENT2(so, fdset1, fdset2) \
67	DO_ENGAGE_EVENT2((so), (fdset1), (fdset2), TCP)
68
69	#define UDP_ENGAGE_EVENT(so, fdset) \
70	DO_ENGAGE_EVENT1((so), (fdset), UDP)
71
72	static const uint8_t special_ethaddr[6] = {
73	0x52, 0x54, 0x00, 0x12, 0x35, 0x00
74	};
75
76	#ifdef RT_OS_WINDOWS
77
78	static int get_dns_addr_domain(PNATState pData, bool fVerbose,
79	struct in_addr *pdns_addr,
80	const char **ppszDomain)
81	{
82	int rc = 0;
83	FIXED_INFO *FixedInfo = NULL;
84	ULONG BufLen;
85	DWORD ret;
86	IP_ADDR_STRING *pIPAddr;
87	struct in_addr tmp_addr;
88
89	FixedInfo = (FIXED_INFO *)GlobalAlloc(GPTR, sizeof(FIXED_INFO));
90	BufLen = sizeof(FIXED_INFO);
91
92	/** @todo: this API returns all DNS servers, no matter whether the
93	* corresponding network adapter is disabled or not. Maybe replace
94	* this by GetAdapterAddresses(), which is XP/Vista only though. */
95	if (ERROR_BUFFER_OVERFLOW == GetNetworkParams(FixedInfo, &BufLen))
96	{
97	if (FixedInfo)
98	{
99	GlobalFree(FixedInfo);
100	FixedInfo = NULL;
101	}
102	FixedInfo = GlobalAlloc(GPTR, BufLen);
103	}
104
105	if ((ret = GetNetworkParams(FixedInfo, &BufLen)) != ERROR_SUCCESS)
106	{
107	Log(("GetNetworkParams failed. ret = %08x\n", (u_int)ret ));
108	if (FixedInfo)
109	{
110	GlobalFree(FixedInfo);
111	FixedInfo = NULL;
112	}
113	rc = -1;
114	goto get_dns_prefix;
115	}
116
117	pIPAddr = &(FixedInfo->DnsServerList);
118	inet_aton(pIPAddr->IpAddress.String, &tmp_addr);
119	Log(("nat: DNS Servers:\n"));
120	if (fVerbose \|\| pdns_addr->s_addr != tmp_addr.s_addr)
121	LogRel(("NAT: DNS address: %s\n", pIPAddr->IpAddress.String));
122	*pdns_addr = tmp_addr;
123
124	pIPAddr = FixedInfo -> DnsServerList.Next;
125	while (pIPAddr)
126	{
127	if (fVerbose)
128	LogRel(("NAT: ignored DNS address: %s\n", pIPAddr ->IpAddress.String));
129	pIPAddr = pIPAddr ->Next;
130	}
131	if (FixedInfo)
132	{
133	GlobalFree(FixedInfo);
134	FixedInfo = NULL;
135	}
136
137	get_dns_prefix:
138	if (ppszDomain)
139	{
140	OSVERSIONINFO ver;
141	char szDnsDomain[256];
142	DWORD dwSize = sizeof(szDnsDomain);
143
144	*ppszDomain = NULL;
145	GetVersionEx(&ver);
146	if (ver.dwMajorVersion >= 5)
147	{
148	/* GetComputerNameEx exists in Windows versions starting with 2000. */
149	if (GetComputerNameEx(ComputerNameDnsDomain, szDnsDomain, &dwSize))
150	{
151	if (szDnsDomain[0])
152	{
153	/* Just non-empty strings are valid. */
154	*ppszDomain = RTStrDup(szDnsDomain);
155	if (pData->fPassDomain)
156	{
157	if (fVerbose)
158	LogRel(("NAT: passing domain name %s\n", szDnsDomain));
159	}
160	else
161	Log(("nat: ignoring domain %s\n", szDnsDomain));
162	}
163	}
164	else
165	Log(("nat: GetComputerNameEx failed (%d)\n", GetLastError()));
166	}
167	}
168	return rc;
169	}
170
171	#else
172
173	static int get_dns_addr_domain(PNATState pData, bool fVerbose,
174	struct in_addr *pdns_addr,
175	const char **ppszDomain)
176	{
177	char buff[512];
178	char buff2[256];
179	FILE *f;
180	int found = 0;
181	struct in_addr tmp_addr;
182
183	#ifdef RT_OS_OS2
184	/* Try various locations. */
185	char *etc = getenv("ETC");
186	f = NULL;
187	if (etc)
188	{
189	snprintf(buff, sizeof(buff), "%s/RESOLV2", etc);
190	f = fopen(buff, "rt");
191	}
192	if (!f)
193	{
194	snprintf(buff, sizeof(buff), "%s/RESOLV2", _PATH_ETC);
195	f = fopen(buff, "rt");
196	}
197	if (!f)
198	{
199	snprintf(buff, sizeof(buff), "%s/resolv.conf", _PATH_ETC);
200	f = fopen(buff, "rt");
201	}
202	#else
203	f = fopen("/etc/resolv.conf", "r");
204	#endif
205	if (!f)
206	return -1;
207
208	if (ppszDomain)
209	*ppszDomain = NULL;
210	Log(("nat: DNS Servers:\n"));
211	while (fgets(buff, 512, f) != NULL)
212	{
213	if (sscanf(buff, "nameserver%*[ \t]%256s", buff2) == 1)
214	{
215	if (!inet_aton(buff2, &tmp_addr))
216	continue;
217	if (tmp_addr.s_addr == loopback_addr.s_addr)
218	tmp_addr = our_addr;
219	/* If it's the first one, set it to dns_addr */
220	if (!found)
221	{
222	if (fVerbose \|\| pdns_addr->s_addr != tmp_addr.s_addr)
223	LogRel(("NAT: DNS address: %s\n", buff2));
224	*pdns_addr = tmp_addr;
225	}
226	else
227	{
228	if (fVerbose)
229	LogRel(("NAT: ignored DNS address: %s\n", buff2));
230	}
231	found++;
232	}
233	if ( ppszDomain
234	&& (!strncmp(buff, "domain", 6) \|\| !strncmp(buff, "search", 6)))
235	{
236	/* Domain name/search list present. Pick first entry */
237	if (*ppszDomain == NULL)
238	{
239	char *tok;
240	char *saveptr;
241	tok = strtok_r(&buff[6], " \t\n", &saveptr);
242	if (tok)
243	{
244	*ppszDomain = RTStrDup(tok);
245	if (pData->fPassDomain)
246	{
247	if (fVerbose)
248	LogRel(("NAT: passing domain name %s\n", tok));
249	}
250	else
251	Log(("nat: ignoring domain %s\n", tok));
252	}
253	}
254	}
255	}
256	fclose(f);
257	if (!found)
258	return -1;
259	return 0;
260	}
261
262	#endif
263
264	int get_dns_addr(PNATState pData, struct in_addr *pdns_addr)
265	{
266	return get_dns_addr_domain(pData, false, pdns_addr, NULL);
267	}
268
269	int slirp_init(PNATState ppData, const char pszNetAddr, uint32_t u32Netmask,
270	bool fPassDomain, const char *pszTFTPPrefix,
271	const char pszBootFile, void pvUser)
272	{
273	int fNATfailed = 0;
274	PNATState pData = malloc(sizeof(NATState));
275	*ppData = pData;
276	if (!pData)
277	return VERR_NO_MEMORY;
278	if (u32Netmask & 0x1f)
279	/* CTL is x.x.x.15, bootp passes up to 16 IPs (15..31) */
280	return VERR_INVALID_PARAMETER;
281	memset(pData, '\0', sizeof(NATState));
282	pData->fPassDomain = fPassDomain;
283	pData->pvUser = pvUser;
284	#if ARCH_BITS == 64 && !defined(VBOX_WITH_BSD_REASS)
285	pData->cpvHashUsed = 1;
286	#endif
287	tftp_prefix = pszTFTPPrefix;
288	bootp_filename = pszBootFile;
289	pData->netmask = u32Netmask;
290
291	#ifdef RT_OS_WINDOWS
292	{
293	WSADATA Data;
294	WSAStartup(MAKEWORD(2,0), &Data);
295	}
296	# if defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC)
297	pData->phEvents[VBOX_SOCKET_EVENT_INDEX] = CreateEvent(NULL, FALSE, FALSE, NULL);
298	# endif
299	#endif
300
301	link_up = 1;
302
303	debug_init();
304	if_init(pData);
305	ip_init(pData);
306	#ifdef VBOX_WITH_SLIRP_ICMP
307	icmp_init(pData);
308	#endif /* VBOX_WITH_SLIRP_ICMP */
309
310	/* Initialise mbufs after setting the MTU */
311	m_init(pData);
312
313	/* set default addresses */
314	inet_aton("127.0.0.1", &loopback_addr);
315	inet_aton("127.0.0.1", &dns_addr);
316
317	if (get_dns_addr_domain(pData, true, &dns_addr, &pData->pszDomain) < 0)
318	fNATfailed = 1;
319
320	inet_aton(pszNetAddr, &special_addr);
321	alias_addr.s_addr = special_addr.s_addr \| htonl(CTL_ALIAS);
322	getouraddr(pData);
323	return fNATfailed ? VINF_NAT_DNS : VINF_SUCCESS;
324	}
325
326	/**
327	* Statistics counters.
328	*/
329	void slirp_register_timers(PNATState pData, PPDMDRVINS pDrvIns)
330	{
331	#ifndef VBOX_WITHOUT_RELEASE_STATISTICS
332	PDMDrvHlpSTAMRegisterF(pDrvIns, &pData->StatFill, STAMTYPE_PROFILE, STAMVISIBILITY_ALWAYS,
333	STAMUNIT_TICKS_PER_CALL, "Profiling slirp fills", "/Drivers/NAT%d/Fill", pDrvIns->iInstance);
334	PDMDrvHlpSTAMRegisterF(pDrvIns, &pData->StatPoll, STAMTYPE_PROFILE, STAMVISIBILITY_ALWAYS,
335	STAMUNIT_TICKS_PER_CALL, "Profiling slirp polls", "/Drivers/NAT%d/Poll", pDrvIns->iInstance);
336	PDMDrvHlpSTAMRegisterF(pDrvIns, &pData->StatFastTimer, STAMTYPE_PROFILE, STAMVISIBILITY_ALWAYS,
337	STAMUNIT_TICKS_PER_CALL, "Profiling slirp fast timer", "/Drivers/NAT%d/TimerFast", pDrvIns->iInstance);
338	PDMDrvHlpSTAMRegisterF(pDrvIns, &pData->StatSlowTimer, STAMTYPE_PROFILE, STAMVISIBILITY_ALWAYS,
339	STAMUNIT_TICKS_PER_CALL, "Profiling slirp slow timer", "/Drivers/NAT%d/TimerSlow", pDrvIns->iInstance);
340	PDMDrvHlpSTAMRegisterF(pDrvIns, &pData->StatTCP, STAMTYPE_COUNTER, STAMVISIBILITY_ALWAYS,
341	STAMUNIT_COUNT, "TCP sockets", "/Drivers/NAT%d/SockTCP", pDrvIns->iInstance);
342	PDMDrvHlpSTAMRegisterF(pDrvIns, &pData->StatTCPHot, STAMTYPE_COUNTER, STAMVISIBILITY_ALWAYS,
343	STAMUNIT_COUNT, "TCP sockets active", "/Drivers/NAT%d/SockTCPHot", pDrvIns->iInstance);
344	PDMDrvHlpSTAMRegisterF(pDrvIns, &pData->StatUDP, STAMTYPE_COUNTER, STAMVISIBILITY_ALWAYS,
345	STAMUNIT_COUNT, "UDP sockets", "/Drivers/NAT%d/SockUDP", pDrvIns->iInstance);
346	PDMDrvHlpSTAMRegisterF(pDrvIns, &pData->StatUDPHot, STAMTYPE_COUNTER, STAMVISIBILITY_ALWAYS,
347	STAMUNIT_COUNT, "UDP sockets active", "/Drivers/NAT%d/SockUDPHot", pDrvIns->iInstance);
348	#endif /* VBOX_WITHOUT_RELEASE_STATISTICS */
349	}
350
351	/**
352	* Marks the link as up, making it possible to establish new connections.
353	*/
354	void slirp_link_up(PNATState pData)
355	{
356	link_up = 1;
357	}
358
359	/**
360	* Marks the link as down and cleans up the current connections.
361	*/
362	void slirp_link_down(PNATState pData)
363	{
364	struct socket *so;
365
366	while ((so = tcb.so_next) != &tcb)
367	{
368	if (so->so_state & SS_NOFDREF \|\| so->s == -1)
369	sofree(pData, so);
370	else
371	tcp_drop(pData, sototcpcb(so), 0);
372	}
373
374	while ((so = udb.so_next) != &udb)
375	udp_detach(pData, so);
376
377	link_up = 0;
378	}
379
380	/**
381	* Terminates the slirp component.
382	*/
383	void slirp_term(PNATState pData)
384	{
385	if (pData->pszDomain)
386	RTStrFree((char )(void )pData->pszDomain);
387
388	#if ARCH_BITS == 64 && !defined(VBOX_WITH_BSD_REASS)
389	LogRel(("NAT: cpvHashUsed=%RU32 cpvHashCollisions=%RU32 cpvHashInserts=%RU64 cpvHashDone=%RU64\n",
390	pData->cpvHashUsed, pData->cpvHashCollisions, pData->cpvHashInserts, pData->cpvHashDone));
391	#endif
392
393	#ifdef VBOX_WITH_SLIRP_ICMP
394	# ifdef RT_OS_WINDOWS
395	pData->pfIcmpCloseHandle(pData->icmp_socket.sh);
396	# else
397	closesocket(pData->icmp_socket.s);
398	# endif
399	#endif
400
401	slirp_link_down(pData);
402	#ifdef RT_OS_WINDOWS
403	WSACleanup();
404	#endif
405	#ifdef LOG_ENABLED
406	Log(("\n"
407	"NAT statistics\n"
408	"--------------\n"
409	"\n"));
410	ipstats(pData);
411	tcpstats(pData);
412	udpstats(pData);
413	icmpstats(pData);
414	mbufstats(pData);
415	sockstats(pData);
416	Log(("\n"
417	"\n"
418	"\n"));
419	#endif
420	free(pData);
421	}
422
423
424	#define CONN_CANFSEND(so) (((so)->so_state & (SS_FCANTSENDMORE\|SS_ISFCONNECTED)) == SS_ISFCONNECTED)
425	#define CONN_CANFRCV(so) (((so)->so_state & (SS_FCANTRCVMORE\|SS_ISFCONNECTED)) == SS_ISFCONNECTED)
426	#define UPD_NFDS(x) if (nfds < (x)) nfds = (x)
427
428	/*
429	* curtime kept to an accuracy of 1ms
430	*/
431	#ifdef RT_OS_WINDOWS
432	static void updtime(PNATState pData)
433	{
434	struct _timeb tb;
435
436	_ftime(&tb);
437	curtime = (u_int)tb.time * (u_int)1000;
438	curtime += (u_int)tb.millitm;
439	}
440	#else
441	static void updtime(PNATState pData)
442	{
443	gettimeofday(&tt, 0);
444
445	curtime = (u_int)tt.tv_sec * (u_int)1000;
446	curtime += (u_int)tt.tv_usec / (u_int)1000;
447
448	if ((tt.tv_usec % 1000) >= 500)
449	curtime++;
450	}
451	#endif
452
453	void slirp_select_fill(PNATState pData, int *pnfds,
454	fd_set readfds, fd_set writefds, fd_set *xfds)
455	{
456	struct socket so, so_next;
457	int nfds;
458	#if defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) && defined(RT_OS_WINDOWS)
459	int rc;
460	int error;
461	#endif
462	#ifdef VBOX_WITH_BSD_REASS
463	int i;
464	#endif /* VBOX_WITH_BSD_REASS */
465
466	STAM_REL_PROFILE_START(&pData->StatFill, a);
467
468	nfds = *pnfds;
469
470	/*
471	* First, TCP sockets
472	*/
473	do_slowtimo = 0;
474	if (link_up)
475	{
476	/*
477	* *_slowtimo needs calling if there are IP fragments
478	* in the fragment queue, or there are TCP connections active
479	*/
480	#ifndef VBOX_WITH_BSD_REASS
481	do_slowtimo = ((tcb.so_next != &tcb)
482	\|\| ((struct ipasfrag )&ipq != u32_to_ptr(pData, ipq.next, struct ipasfrag )));
483	#else /* !VBOX_WITH_BSD_REASS */
484	/* XXX:
485	* triggering of fragment expiration should be the same but use new macroses
486	*/
487	do_slowtimo = (tcb.so_next != &tcb);
488	if (!do_slowtimo)
489	{
490	for (i = 0; i < IPREASS_NHASH; i++)
491	{
492	if (!TAILQ_EMPTY(&ipq[i]))
493	{
494	do_slowtimo = 1;
495	break;
496	}
497	}
498	}
499	#endif /* VBOX_WITH_BSD_REASS */
500
501	STAM_REL_COUNTER_RESET(&pData->StatTCP);
502	STAM_REL_COUNTER_RESET(&pData->StatTCPHot);
503
504	for (so = tcb.so_next; so != &tcb; so = so_next)
505	{
506	so_next = so->so_next;
507
508	STAM_REL_COUNTER_INC(&pData->StatTCP);
509
510	/*
511	* See if we need a tcp_fasttimo
512	*/
513	if (time_fasttimo == 0 && so->so_tcpcb->t_flags & TF_DELACK)
514	time_fasttimo = curtime; /* Flag when we want a fasttimo */
515
516	/*
517	* NOFDREF can include still connecting to local-host,
518	* newly socreated() sockets etc. Don't want to select these.
519	*/
520	if (so->so_state & SS_NOFDREF \|\| so->s == -1)
521	continue;
522
523	/*
524	* Set for reading sockets which are accepting
525	*/
526	if (so->so_state & SS_FACCEPTCONN)
527	{
528	STAM_REL_COUNTER_INC(&pData->StatTCPHot);
529	TCP_ENGAGE_EVENT1(so, readfds);
530	continue;
531	}
532
533	/*
534	* Set for writing sockets which are connecting
535	*/
536	if (so->so_state & SS_ISFCONNECTING)
537	{
538	STAM_REL_COUNTER_INC(&pData->StatTCPHot);
539	TCP_ENGAGE_EVENT1(so, writefds);
540	}
541
542	/*
543	* Set for writing if we are connected, can send more, and
544	* we have something to send
545	*/
546	if (CONN_CANFSEND(so) && so->so_rcv.sb_cc)
547	{
548	STAM_REL_COUNTER_INC(&pData->StatTCPHot);
549	TCP_ENGAGE_EVENT1(so, writefds);
550	}
551
552	/*
553	* Set for reading (and urgent data) if we are connected, can
554	* receive more, and we have room for it XXX /2 ?
555	*/
556	if (CONN_CANFRCV(so) && (so->so_snd.sb_cc < (so->so_snd.sb_datalen/2)))
557	{
558	STAM_REL_COUNTER_INC(&pData->StatTCPHot);
559	TCP_ENGAGE_EVENT2(so, readfds, xfds);
560	}
561	}
562
563	/*
564	* UDP sockets
565	*/
566	STAM_REL_COUNTER_RESET(&pData->StatUDP);
567	STAM_REL_COUNTER_RESET(&pData->StatUDPHot);
568
569	for (so = udb.so_next; so != &udb; so = so_next)
570	{
571	so_next = so->so_next;
572
573	STAM_REL_COUNTER_INC(&pData->StatUDP);
574
575	/*
576	* See if it's timed out
577	*/
578	if (so->so_expire)
579	{
580	if (so->so_expire <= curtime)
581	{
582	udp_detach(pData, so);
583	continue;
584	}
585	else
586	do_slowtimo = 1; /* Let socket expire */
587	}
588
589	/*
590	* When UDP packets are received from over the link, they're
591	* sendto()'d straight away, so no need for setting for writing
592	* Limit the number of packets queued by this session to 4.
593	* Note that even though we try and limit this to 4 packets,
594	* the session could have more queued if the packets needed
595	* to be fragmented.
596	*
597	* (XXX <= 4 ?)
598	*/
599	if ((so->so_state & SS_ISFCONNECTED) && so->so_queued <= 4)
600	{
601	STAM_REL_COUNTER_INC(&pData->StatUDPHot);
602	UDP_ENGAGE_EVENT(so, readfds);
603	}
604	}
605	ICMP_ENGAGE_EVENT(&pData->icmp_socket, readfds, writefds);
606	}
607
608	#if !defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) \|\| !defined(RT_OS_WINDOWS)
609	*pnfds = nfds;
610	#else
611	*pnfds = VBOX_EVENT_COUNT;
612	#endif
613
614	STAM_REL_PROFILE_STOP(&pData->StatFill, a);
615	}
616
617	void slirp_select_poll(PNATState pData, fd_set readfds, fd_set writefds, fd_set *xfds)
618	{
619	struct socket so, so_next;
620	int ret;
621	#if defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) && defined(RT_OS_WINDOWS)
622	WSANETWORKEVENTS NetworkEvents;
623	int rc;
624	int error;
625	#endif
626	static uint32_t stat_time;
627
628	STAM_REL_PROFILE_START(&pData->StatPoll, a);
629
630	/* Update time */
631	updtime(pData);
632
633	#ifdef LOG_ENABLED
634	if (curtime - stat_time > 10000)
635	{
636	stat_time = curtime;
637	sockstats(pData);
638	}
639	#endif
640
641	/*
642	* See if anything has timed out
643	*/
644	if (link_up)
645	{
646	if (time_fasttimo && ((curtime - time_fasttimo) >= 2))
647	{
648	STAM_REL_PROFILE_START(&pData->StatFastTimer, a);
649	tcp_fasttimo(pData);
650	time_fasttimo = 0;
651	STAM_REL_PROFILE_STOP(&pData->StatFastTimer, a);
652	}
653	if (do_slowtimo && ((curtime - last_slowtimo) >= 499))
654	{
655	STAM_REL_PROFILE_START(&pData->StatSlowTimer, a);
656	ip_slowtimo(pData);
657	tcp_slowtimo(pData);
658	last_slowtimo = curtime;
659	STAM_REL_PROFILE_STOP(&pData->StatSlowTimer, a);
660	}
661	}
662	#if defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) && defined(RT_OS_WINDOWS)
663	if (!readfds && !writefds && !xfds)
664	return; /* only timer update */
665	#endif
666
667	/*
668	* Check sockets
669	*/
670	if (link_up)
671	{
672	/*
673	* Check TCP sockets
674	*/
675	for (so = tcb.so_next; so != &tcb; so = so_next)
676	{
677	so_next = so->so_next;
678
679	/*
680	* FD_ISSET is meaningless on these sockets
681	* (and they can crash the program)
682	*/
683	if (so->so_state & SS_NOFDREF \|\| so->s == -1)
684	continue;
685
686	#if defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) && defined(RT_OS_WINDOWS)
687	rc = WSAEnumNetworkEvents(so->s, VBOX_SOCKET_EVENT, &NetworkEvents);
688	if (rc == SOCKET_ERROR)
689	{
690	error = WSAGetLastError();
691	LogRel(("WSAEnumNetworkEvents TCP error %d\n", error));
692	continue;
693	}
694	#endif
695
696	/*
697	* Check for URG data
698	* This will soread as well, so no need to
699	* test for readfds below if this succeeds
700	*/
701	#if !defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) \|\| !defined(RT_OS_WINDOWS)
702	if (FD_ISSET(so->s, xfds))
703	#else
704	/* out-of-band data */
705	if ((NetworkEvents.lNetworkEvents & FD_OOB) && NetworkEvents.iErrorCode[FD_OOB_BIT] == 0)
706	#endif
707	{
708	sorecvoob(pData, so);
709	}
710
711	/*
712	* Check sockets for reading
713	*/
714	#if !defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) \|\| !defined(RT_OS_WINDOWS)
715	else if (FD_ISSET(so->s, readfds))
716	#else
717	else if ((NetworkEvents.lNetworkEvents & FD_READ) && (NetworkEvents.iErrorCode[FD_READ_BIT] == 0))
718	#endif
719	{
720	/*
721	* Check for incoming connections
722	*/
723	if (so->so_state & SS_FACCEPTCONN)
724	{
725	tcp_connect(pData, so);
726	#if defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) && defined(RT_OS_WINDOWS)
727	if (!(NetworkEvents.lNetworkEvents & FD_CLOSE))
728	#endif
729	continue;
730	}
731
732	ret = soread(pData, so);
733	/* Output it if we read something */
734	if (ret > 0)
735	tcp_output(pData, sototcpcb(so));
736	}
737
738	#if defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) && defined(RT_OS_WINDOWS)
739	/*
740	* Check for FD_CLOSE events.
741	*/
742	if (NetworkEvents.lNetworkEvents & FD_CLOSE)
743	{
744	/*
745	* drain the socket
746	*/
747	for (;;)
748	{
749	ret = soread(pData, so);
750	if (ret > 0)
751	tcp_output(pData, sototcpcb(so));
752	else
753	break;
754	}
755	}
756	#endif
757
758	/*
759	* Check sockets for writing
760	*/
761	#if !defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) \|\| !defined(RT_OS_WINDOWS)
762	if (FD_ISSET(so->s, writefds))
763	#else
764	if ((NetworkEvents.lNetworkEvents & FD_WRITE) && (NetworkEvents.iErrorCode[FD_WRITE_BIT] == 0))
765	#endif
766	{
767	/*
768	* Check for non-blocking, still-connecting sockets
769	*/
770	if (so->so_state & SS_ISFCONNECTING)
771	{
772	/* Connected */
773	so->so_state &= ~SS_ISFCONNECTING;
774
775	/*
776	* This should be probably guarded by PROBE_CONN too. Anyway,
777	* we disable it on OS/2 because the below send call returns
778	* EFAULT which causes the opened TCP socket to close right
779	* after it has been opened and connected.
780	*/
781	#ifndef RT_OS_OS2
782	ret = send(so->s, (const char *)&ret, 0, 0);
783	if (ret < 0)
784	{
785	/* XXXXX Must fix, zero bytes is a NOP */
786	if ( errno == EAGAIN
787	\|\| errno == EWOULDBLOCK
788	\|\| errno == EINPROGRESS
789	\|\| errno == ENOTCONN)
790	continue;
791
792	/* else failed */
793	so->so_state = SS_NOFDREF;
794	}
795	/* else so->so_state &= ~SS_ISFCONNECTING; */
796	#endif
797
798	/*
799	* Continue tcp_input
800	*/
801	tcp_input(pData, (struct mbuf *)NULL, sizeof(struct ip), so);
802	/* continue; */
803	}
804	else
805	ret = sowrite(pData, so);
806	/*
807	* XXX If we wrote something (a lot), there could be the need
808	* for a window update. In the worst case, the remote will send
809	* a window probe to get things going again.
810	*/
811	}
812
813	/*
814	* Probe a still-connecting, non-blocking socket
815	* to check if it's still alive
816	*/
817	#ifdef PROBE_CONN
818	if (so->so_state & SS_ISFCONNECTING)
819	{
820	ret = recv(so->s, (char *)&ret, 0, 0);
821
822	if (ret < 0)
823	{
824	/* XXX */
825	if ( errno == EAGAIN
826	\|\| errno == EWOULDBLOCK
827	\|\| errno == EINPROGRESS
828	\|\| errno == ENOTCONN)
829	{
830	continue; /* Still connecting, continue */
831	}
832
833	/* else failed */
834	so->so_state = SS_NOFDREF;
835
836	/* tcp_input will take care of it */
837	}
838	else
839	{
840	ret = send(so->s, &ret, 0, 0);
841	if (ret < 0)
842	{
843	/* XXX */
844	if ( errno == EAGAIN
845	\|\| errno == EWOULDBLOCK
846	\|\| errno == EINPROGRESS
847	\|\| errno == ENOTCONN)
848	{
849	continue;
850	}
851	/* else failed */
852	so->so_state = SS_NOFDREF;
853	}
854	else
855	so->so_state &= ~SS_ISFCONNECTING;
856
857	}
858	tcp_input((struct mbuf *)NULL, sizeof(struct ip),so);
859	} /* SS_ISFCONNECTING */
860	#endif
861	}
862
863	/*
864	* Now UDP sockets.
865	* Incoming packets are sent straight away, they're not buffered.
866	* Incoming UDP data isn't buffered either.
867	*/
868	for (so = udb.so_next; so != &udb; so = so_next)
869	{
870	so_next = so->so_next;
871
872	#if defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) && defined(RT_OS_WINDOWS)
873	rc = WSAEnumNetworkEvents(so->s, VBOX_SOCKET_EVENT, &NetworkEvents);
874	if (rc == SOCKET_ERROR)
875	{
876	error = WSAGetLastError();
877	LogRel(("WSAEnumNetworkEvents TCP error %d\n", error));
878	continue;
879	}
880	#endif
881	#if !defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) \|\| !defined(RT_OS_WINDOWS)
882	if (so->s != -1 && FD_ISSET(so->s, readfds))
883	#else
884	if (((NetworkEvents.lNetworkEvents & FD_READ) && (NetworkEvents.iErrorCode[FD_READ_BIT] == 0)))
885	#endif
886	{
887	sorecvfrom(pData, so);
888	}
889	}
890
891	#if defined(VBOX_WITH_SLIRP_ICMP)
892	# if defined(RT_OS_WINDOWS)
893	sorecvfrom(pData, &pData->icmp_socket);
894	# else
895	if (so->s != -1 && FD_ISSET(so->s, readfds))
896	{
897	sorecvfrom(pData, &pData->icmp_socket);
898	}
899	# endif
900	#endif
901	}
902
903	/*
904	* See if we can start outputting
905	*/
906	if (if_queued && link_up)
907	if_start(pData);
908
909	STAM_REL_PROFILE_STOP(&pData->StatPoll, a);
910	}
911
912	#define ETH_ALEN 6
913	#define ETH_HLEN 14
914
915	#define ETH_P_IP 0x0800 /* Internet Protocol packet */
916	#define ETH_P_ARP 0x0806 /* Address Resolution packet */
917
918	#define ARPOP_REQUEST 1 /* ARP request */
919	#define ARPOP_REPLY 2 /* ARP reply */
920
921	struct ethhdr
922	{
923	unsigned char h_dest[ETH_ALEN]; /* destination eth addr */
924	unsigned char h_source[ETH_ALEN]; /* source ether addr */
925	unsigned short h_proto; /* packet type ID field */
926	};
927
928	struct arphdr
929	{
930	unsigned short ar_hrd; /* format of hardware address */
931	unsigned short ar_pro; /* format of protocol address */
932	unsigned char ar_hln; /* length of hardware address */
933	unsigned char ar_pln; /* length of protocol address */
934	unsigned short ar_op; /* ARP opcode (command) */
935
936	/*
937	* Ethernet looks like this : This bit is variable sized however...
938	*/
939	unsigned char ar_sha[ETH_ALEN]; /* sender hardware address */
940	unsigned char ar_sip[4]; /* sender IP address */
941	unsigned char ar_tha[ETH_ALEN]; /* target hardware address */
942	unsigned char ar_tip[4]; /* target IP address */
943	};
944
945	static
946	void arp_input(PNATState pData, const uint8_t *pkt, int pkt_len)
947	{
948	struct ethhdr eh = (struct ethhdr )pkt;
949	struct arphdr ah = (struct arphdr )(pkt + ETH_HLEN);
950	uint8_t arp_reply[ETH_HLEN + sizeof(struct arphdr)];
951	struct ethhdr reh = (struct ethhdr )arp_reply;
952	struct arphdr rah = (struct arphdr )(arp_reply + ETH_HLEN);
953	int ar_op;
954	struct ex_list *ex_ptr;
955	uint32_t htip = ntohl((uint32_t)ah->ar_tip);
956
957	ar_op = ntohs(ah->ar_op);
958	switch(ar_op)
959	{
960	case ARPOP_REQUEST:
961	if ((htip & pData->netmask) == ntohl(special_addr.s_addr))
962	{
963	if ( (htip & ~pData->netmask) == CTL_DNS
964	\|\| (htip & ~pData->netmask) == CTL_ALIAS)
965	goto arp_ok;
966	for (ex_ptr = exec_list; ex_ptr; ex_ptr = ex_ptr->ex_next)
967	{
968	if ((htip & ~pData->netmask) == ex_ptr->ex_addr)
969	goto arp_ok;
970	}
971	return;
972	arp_ok:
973	/* XXX: make an ARP request to have the client address */
974	memcpy(client_ethaddr, eh->h_source, ETH_ALEN);
975
976	/* ARP request for alias/dns mac address */
977	memcpy(reh->h_dest, pkt + ETH_ALEN, ETH_ALEN);
978	memcpy(reh->h_source, special_ethaddr, ETH_ALEN - 1);
979	reh->h_source[5] = ah->ar_tip[3];
980	reh->h_proto = htons(ETH_P_ARP);
981
982	rah->ar_hrd = htons(1);
983	rah->ar_pro = htons(ETH_P_IP);
984	rah->ar_hln = ETH_ALEN;
985	rah->ar_pln = 4;
986	rah->ar_op = htons(ARPOP_REPLY);
987	memcpy(rah->ar_sha, reh->h_source, ETH_ALEN);
988	memcpy(rah->ar_sip, ah->ar_tip, 4);
989	memcpy(rah->ar_tha, ah->ar_sha, ETH_ALEN);
990	memcpy(rah->ar_tip, ah->ar_sip, 4);
991	slirp_output(pData->pvUser, arp_reply, sizeof(arp_reply));
992	}
993	break;
994	default:
995	break;
996	}
997	}
998
999	void slirp_input(PNATState pData, const uint8_t *pkt, int pkt_len)
1000	{
1001	struct mbuf *m;
1002	int proto;
1003
1004	if (pkt_len < ETH_HLEN)
1005	return;
1006
1007	proto = ntohs((uint16_t )(pkt + 12));
1008	switch(proto)
1009	{
1010	case ETH_P_ARP:
1011	arp_input(pData, pkt, pkt_len);
1012	break;
1013	case ETH_P_IP:
1014	/* Update time. Important if the network is very quiet, as otherwise
1015	* the first outgoing connection gets an incorrect timestamp. */
1016	updtime(pData);
1017
1018	m = m_get(pData);
1019	if (!m)
1020	return;
1021	/* Note: we add to align the IP header */
1022	if (M_FREEROOM(m) < pkt_len + 2)
1023	{
1024	m_inc(m, pkt_len + 2);
1025	}
1026	m->m_len = pkt_len + 2;
1027	memcpy(m->m_data + 2, pkt, pkt_len);
1028
1029	m->m_data += 2 + ETH_HLEN;
1030	m->m_len -= 2 + ETH_HLEN;
1031
1032	ip_input(pData, m);
1033	break;
1034	default:
1035	break;
1036	}
1037	}
1038
1039	/* output the IP packet to the ethernet device */
1040	void if_encap(PNATState pData, const uint8_t *ip_data, int ip_data_len)
1041	{
1042	uint8_t buf[1600];
1043	struct ethhdr eh = (struct ethhdr )buf;
1044
1045	if (ip_data_len + ETH_HLEN > sizeof(buf))
1046	return;
1047
1048	memcpy(eh->h_dest, client_ethaddr, ETH_ALEN);
1049	memcpy(eh->h_source, special_ethaddr, ETH_ALEN - 1);
1050	/* XXX: not correct */
1051	eh->h_source[5] = CTL_ALIAS;
1052	eh->h_proto = htons(ETH_P_IP);
1053	memcpy(buf + sizeof(struct ethhdr), ip_data, ip_data_len);
1054	slirp_output(pData->pvUser, buf, ip_data_len + ETH_HLEN);
1055	}
1056
1057	int slirp_redir(PNATState pData, int is_udp, int host_port,
1058	struct in_addr guest_addr, int guest_port)
1059	{
1060	if (is_udp)
1061	{
1062	if (!udp_listen(pData, htons(host_port), guest_addr.s_addr,
1063	htons(guest_port), 0))
1064	return -1;
1065	}
1066	else
1067	{
1068	if (!solisten(pData, htons(host_port), guest_addr.s_addr,
1069	htons(guest_port), 0))
1070	return -1;
1071	}
1072	return 0;
1073	}
1074
1075	int slirp_add_exec(PNATState pData, int do_pty, const char *args, int addr_low_byte,
1076	int guest_port)
1077	{
1078	return add_exec(&exec_list, do_pty, (char *)args,
1079	addr_low_byte, htons(guest_port));
1080	}
1081
1082	void slirp_set_ethaddr(PNATState pData, const uint8_t *ethaddr)
1083	{
1084	memcpy(client_ethaddr, ethaddr, ETH_ALEN);
1085	}
1086
1087	#if defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) && defined(RT_OS_WINDOWS)
1088	HANDLE *slirp_get_events(PNATState pData)
1089	{
1090	return pData->phEvents;
1091	}
1092	void slirp_register_external_event(PNATState pData, HANDLE hEvent, int index)
1093	{
1094	pData->phEvents[index] = hEvent;
1095	}
1096	#endif
1097
1098	unsigned int slirp_get_timeout_ms(PNATState pData)
1099	{
1100	if (link_up)
1101	{
1102	if (time_fasttimo)
1103	return 2;
1104	if (do_slowtimo)
1105	return 500; /* see PR_SLOWHZ */
1106	}
1107	return 0;
1108	}

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

source: vbox/trunk/src/VBox/Devices/Network/slirp/slirp.c@ 15203

Download in other formats: