slirp.c@ 15230

Last change on this file since 15230 was 15223, checked in by vboxsync, 16 years ago
slirp: fix win burns
Property svn:eol-style set to `native`
File size: 33.9 KB

Line
1	#include "slirp.h"
2	#ifdef RT_OS_OS2
3	# include <paths.h>
4	#endif
5
6	/* disable these counters for the final release */
7	/* #define VBOX_WITHOUT_RELEASE_STATISTICS */
8
9	#include <VBox/err.h>
10	#include <VBox/pdmdrv.h>
11	#include <iprt/assert.h>
12
13	#if !defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) \|\| !defined(RT_OS_WINDOWS)
14
15	# define DO_ENGAGE_EVENT1(so, fdset, label) \
16	do { \
17	FD_SET((so)->s, (fdset)); \
18	UPD_NFDS((so)->s); \
19	} while(0)
20
21
22	# define DO_ENGAGE_EVENT2(so, fdset1, fdset2, label) \
23	do { \
24	FD_SET((so)->s, (fdset1)); \
25	FD_SET((so)->s, (fdset2)); \
26	UPD_NFDS((so)->s); \
27	} while(0)
28
29	# define DO_POLL_EVENTS(rc, error, so, events, label) do {} while (0)
30
31	#define DO_CHECK_FD_SET(so, events, fdset) (FD_ISSET((so)->s, (fdset)))
32
33	# ifdef VBOX_WITH_SLIRP_ICMP
34	# define ICMP_ENGAGE_EVENT(so, fdset) \
35	do { \
36	if (pData->icmp_socket.s != -1) \
37	DO_ENGAGE_EVENT1((so), (fdset), ICMP); \
38	} while (0)
39	# else /* !VBOX_WITH_SLIRP_ICMP */
40	# define ICMP_ENGAGE_EVENT(so, fdset) do {} while (0)
41	# endif /* !VBOX_WITH_SLIRP_ICMP */
42
43	#else /* defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) && defined(RT_OS_WINDOWS) */
44
45	/*
46	* On Windows, we will be notified by IcmpSendEcho2() when the response arrives.
47	* So no call to WSAEventSelect necessary.
48	*/
49	# define ICMP_ENGAGE_EVENT(so, fdset) do {} while(0)
50
51	# define DO_ENGAGE_EVENT1(so, fdset1, label) \
52	do { \
53	rc = WSAEventSelect((so)->s, VBOX_SOCKET_EVENT, FD_ALL_EVENTS); \
54	if (rc == SOCKET_ERROR) \
55	{ \
56	/* This should not happen */ \
57	error = WSAGetLastError(); \
58	LogRel(("WSAEventSelector (" #label ") error %d (so=%x, socket=%s, event=%x)\n", \
59	error, (so), (so)->s, VBOX_SOCKET_EVENT)); \
60	} \
61	} while(0)
62
63	# define DO_ENGAGE_EVENT2(so, fdset1, fdset2, label) \
64	DO_ENGAGE_EVENT1((so), (fdset1), label)
65
66	# define DO_POLL_EVENTS(rc, error, so, events, label) \
67	(rc) = WSAEnumNetworkEvents((so)->s, VBOX_SOCKET_EVENT, (events)); \
68	if ((rc) == SOCKET_ERROR) \
69	{ \
70	(error) = WSAGetLastError(); \
71	LogRel(("WSAEnumNetworkEvents " #label " error %d\n", (error))); \
72	continue; \
73	}
74
75	# define readfds_win FD_READ
76	# define readfds_win_bit FD_READ_BIT
77
78	# define writefds_win FD_WRITE
79	# define writefds_win_bit FD_WRITE_BIT
80
81	# define xfds_win FD_OOB
82	# define xfds_win_bit FD_OOB_BIT
83
84	# define DO_CHECK_FD_SET(so, events, fdset) \
85	(((events).lNetworkEvents & fdset ## _win) && ((events).iErrorCode[fdset ## _win_bit] == 0))
86
87
88	#endif /* defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) && defined(RT_OS_WINDOWS) */
89
90	#define TCP_ENGAGE_EVENT1(so, fdset) \
91	DO_ENGAGE_EVENT1((so), (fdset), TCP)
92
93	#define TCP_ENGAGE_EVENT2(so, fdset1, fdset2) \
94	DO_ENGAGE_EVENT2((so), (fdset1), (fdset2), TCP)
95
96	#define UDP_ENGAGE_EVENT(so, fdset) \
97	DO_ENGAGE_EVENT1((so), (fdset), UDP)
98
99	#define POLL_TCP_EVENTS(rc, error, so, events) \
100	DO_POLL_EVENTS((rc), (error), (so), (events), TCP)
101
102	#define POLL_UDP_EVENTS(rc, error, so, events) \
103	DO_POLL_EVENTS((rc), (error), (so), (events), UDP)
104
105	#define CHECK_FD_SET(so, events, set) \
106	(DO_CHECK_FD_SET((so), (events), set))
107
108	static const uint8_t special_ethaddr[6] =
109	{
110	0x52, 0x54, 0x00, 0x12, 0x35, 0x00
111	};
112
113	#ifdef RT_OS_WINDOWS
114
115	static int get_dns_addr_domain(PNATState pData, bool fVerbose,
116	struct in_addr *pdns_addr,
117	const char **ppszDomain)
118	{
119	int rc = 0;
120	FIXED_INFO *FixedInfo = NULL;
121	ULONG BufLen;
122	DWORD ret;
123	IP_ADDR_STRING *pIPAddr;
124	struct in_addr tmp_addr;
125
126	FixedInfo = (FIXED_INFO *)GlobalAlloc(GPTR, sizeof(FIXED_INFO));
127	BufLen = sizeof(FIXED_INFO);
128
129	/** @todo: this API returns all DNS servers, no matter whether the
130	* corresponding network adapter is disabled or not. Maybe replace
131	* this by GetAdapterAddresses(), which is XP/Vista only though. */
132	if (ERROR_BUFFER_OVERFLOW == GetNetworkParams(FixedInfo, &BufLen))
133	{
134	if (FixedInfo)
135	{
136	GlobalFree(FixedInfo);
137	FixedInfo = NULL;
138	}
139	FixedInfo = GlobalAlloc(GPTR, BufLen);
140	}
141
142	if ((ret = GetNetworkParams(FixedInfo, &BufLen)) != ERROR_SUCCESS)
143	{
144	Log(("GetNetworkParams failed. ret = %08x\n", (u_int)ret ));
145	if (FixedInfo)
146	{
147	GlobalFree(FixedInfo);
148	FixedInfo = NULL;
149	}
150	rc = -1;
151	goto get_dns_prefix;
152	}
153
154	pIPAddr = &(FixedInfo->DnsServerList);
155	inet_aton(pIPAddr->IpAddress.String, &tmp_addr);
156	Log(("nat: DNS Servers:\n"));
157	if (fVerbose \|\| pdns_addr->s_addr != tmp_addr.s_addr)
158	LogRel(("NAT: DNS address: %s\n", pIPAddr->IpAddress.String));
159	*pdns_addr = tmp_addr;
160
161	pIPAddr = FixedInfo -> DnsServerList.Next;
162	while (pIPAddr)
163	{
164	if (fVerbose)
165	LogRel(("NAT: ignored DNS address: %s\n", pIPAddr ->IpAddress.String));
166	pIPAddr = pIPAddr ->Next;
167	}
168	if (FixedInfo)
169	{
170	GlobalFree(FixedInfo);
171	FixedInfo = NULL;
172	}
173
174	get_dns_prefix:
175	if (ppszDomain)
176	{
177	OSVERSIONINFO ver;
178	char szDnsDomain[256];
179	DWORD dwSize = sizeof(szDnsDomain);
180
181	*ppszDomain = NULL;
182	GetVersionEx(&ver);
183	if (ver.dwMajorVersion >= 5)
184	{
185	/* GetComputerNameEx exists in Windows versions starting with 2000. */
186	if (GetComputerNameEx(ComputerNameDnsDomain, szDnsDomain, &dwSize))
187	{
188	if (szDnsDomain[0])
189	{
190	/* Just non-empty strings are valid. */
191	*ppszDomain = RTStrDup(szDnsDomain);
192	if (pData->fPassDomain)
193	{
194	if (fVerbose)
195	LogRel(("NAT: passing domain name %s\n", szDnsDomain));
196	}
197	else
198	Log(("nat: ignoring domain %s\n", szDnsDomain));
199	}
200	}
201	else
202	Log(("nat: GetComputerNameEx failed (%d)\n", GetLastError()));
203	}
204	}
205	return rc;
206	}
207
208	#else
209
210	static int get_dns_addr_domain(PNATState pData, bool fVerbose,
211	struct in_addr *pdns_addr,
212	const char **ppszDomain)
213	{
214	char buff[512];
215	char buff2[256];
216	FILE *f;
217	int found = 0;
218	struct in_addr tmp_addr;
219
220	#ifdef RT_OS_OS2
221	/* Try various locations. */
222	char *etc = getenv("ETC");
223	f = NULL;
224	if (etc)
225	{
226	snprintf(buff, sizeof(buff), "%s/RESOLV2", etc);
227	f = fopen(buff, "rt");
228	}
229	if (!f)
230	{
231	snprintf(buff, sizeof(buff), "%s/RESOLV2", _PATH_ETC);
232	f = fopen(buff, "rt");
233	}
234	if (!f)
235	{
236	snprintf(buff, sizeof(buff), "%s/resolv.conf", _PATH_ETC);
237	f = fopen(buff, "rt");
238	}
239	#else
240	f = fopen("/etc/resolv.conf", "r");
241	#endif
242	if (!f)
243	return -1;
244
245	if (ppszDomain)
246	*ppszDomain = NULL;
247	Log(("nat: DNS Servers:\n"));
248	while (fgets(buff, 512, f) != NULL)
249	{
250	if (sscanf(buff, "nameserver%*[ \t]%256s", buff2) == 1)
251	{
252	if (!inet_aton(buff2, &tmp_addr))
253	continue;
254	if (tmp_addr.s_addr == loopback_addr.s_addr)
255	tmp_addr = our_addr;
256	/* If it's the first one, set it to dns_addr */
257	if (!found)
258	{
259	if (fVerbose \|\| pdns_addr->s_addr != tmp_addr.s_addr)
260	LogRel(("NAT: DNS address: %s\n", buff2));
261	*pdns_addr = tmp_addr;
262	}
263	else
264	{
265	if (fVerbose)
266	LogRel(("NAT: ignored DNS address: %s\n", buff2));
267	}
268	found++;
269	}
270	if ( ppszDomain
271	&& (!strncmp(buff, "domain", 6) \|\| !strncmp(buff, "search", 6)))
272	{
273	/* Domain name/search list present. Pick first entry */
274	if (*ppszDomain == NULL)
275	{
276	char *tok;
277	char *saveptr;
278	tok = strtok_r(&buff[6], " \t\n", &saveptr);
279	if (tok)
280	{
281	*ppszDomain = RTStrDup(tok);
282	if (pData->fPassDomain)
283	{
284	if (fVerbose)
285	LogRel(("NAT: passing domain name %s\n", tok));
286	}
287	else
288	Log(("nat: ignoring domain %s\n", tok));
289	}
290	}
291	}
292	}
293	fclose(f);
294	if (!found)
295	return -1;
296	return 0;
297	}
298
299	#endif
300
301	int get_dns_addr(PNATState pData, struct in_addr *pdns_addr)
302	{
303	return get_dns_addr_domain(pData, false, pdns_addr, NULL);
304	}
305
306	int slirp_init(PNATState ppData, const char pszNetAddr, uint32_t u32Netmask,
307	bool fPassDomain, const char *pszTFTPPrefix,
308	const char pszBootFile, void pvUser)
309	{
310	int fNATfailed = 0;
311	PNATState pData = malloc(sizeof(NATState));
312	*ppData = pData;
313	if (!pData)
314	return VERR_NO_MEMORY;
315	if (u32Netmask & 0x1f)
316	/* CTL is x.x.x.15, bootp passes up to 16 IPs (15..31) */
317	return VERR_INVALID_PARAMETER;
318	memset(pData, '\0', sizeof(NATState));
319	pData->fPassDomain = fPassDomain;
320	pData->pvUser = pvUser;
321	#if ARCH_BITS == 64 && !defined(VBOX_WITH_BSD_REASS)
322	pData->cpvHashUsed = 1;
323	#endif
324	tftp_prefix = pszTFTPPrefix;
325	bootp_filename = pszBootFile;
326	pData->netmask = u32Netmask;
327
328	#ifdef RT_OS_WINDOWS
329	{
330	WSADATA Data;
331	WSAStartup(MAKEWORD(2,0), &Data);
332	}
333	# if defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC)
334	pData->phEvents[VBOX_SOCKET_EVENT_INDEX] = CreateEvent(NULL, FALSE, FALSE, NULL);
335	# endif
336	#endif
337
338	link_up = 1;
339
340	debug_init();
341	if_init(pData);
342	ip_init(pData);
343	#ifdef VBOX_WITH_SLIRP_ICMP
344	icmp_init(pData);
345	#endif /* VBOX_WITH_SLIRP_ICMP */
346
347	/* Initialise mbufs after setting the MTU */
348	m_init(pData);
349
350	/* set default addresses */
351	inet_aton("127.0.0.1", &loopback_addr);
352	inet_aton("127.0.0.1", &dns_addr);
353
354	if (get_dns_addr_domain(pData, true, &dns_addr, &pData->pszDomain) < 0)
355	fNATfailed = 1;
356
357	inet_aton(pszNetAddr, &special_addr);
358	alias_addr.s_addr = special_addr.s_addr \| htonl(CTL_ALIAS);
359	getouraddr(pData);
360	return fNATfailed ? VINF_NAT_DNS : VINF_SUCCESS;
361	}
362
363	/**
364	* Statistics counters.
365	*/
366	void slirp_register_timers(PNATState pData, PPDMDRVINS pDrvIns)
367	{
368	#ifndef VBOX_WITHOUT_RELEASE_STATISTICS
369	PDMDrvHlpSTAMRegisterF(pDrvIns, &pData->StatFill, STAMTYPE_PROFILE, STAMVISIBILITY_ALWAYS,
370	STAMUNIT_TICKS_PER_CALL, "Profiling slirp fills", "/Drivers/NAT%d/Fill", pDrvIns->iInstance);
371	PDMDrvHlpSTAMRegisterF(pDrvIns, &pData->StatPoll, STAMTYPE_PROFILE, STAMVISIBILITY_ALWAYS,
372	STAMUNIT_TICKS_PER_CALL, "Profiling slirp polls", "/Drivers/NAT%d/Poll", pDrvIns->iInstance);
373	PDMDrvHlpSTAMRegisterF(pDrvIns, &pData->StatFastTimer, STAMTYPE_PROFILE, STAMVISIBILITY_ALWAYS,
374	STAMUNIT_TICKS_PER_CALL, "Profiling slirp fast timer", "/Drivers/NAT%d/TimerFast", pDrvIns->iInstance);
375	PDMDrvHlpSTAMRegisterF(pDrvIns, &pData->StatSlowTimer, STAMTYPE_PROFILE, STAMVISIBILITY_ALWAYS,
376	STAMUNIT_TICKS_PER_CALL, "Profiling slirp slow timer", "/Drivers/NAT%d/TimerSlow", pDrvIns->iInstance);
377	PDMDrvHlpSTAMRegisterF(pDrvIns, &pData->StatTCP, STAMTYPE_COUNTER, STAMVISIBILITY_ALWAYS,
378	STAMUNIT_COUNT, "TCP sockets", "/Drivers/NAT%d/SockTCP", pDrvIns->iInstance);
379	PDMDrvHlpSTAMRegisterF(pDrvIns, &pData->StatTCPHot, STAMTYPE_COUNTER, STAMVISIBILITY_ALWAYS,
380	STAMUNIT_COUNT, "TCP sockets active", "/Drivers/NAT%d/SockTCPHot", pDrvIns->iInstance);
381	PDMDrvHlpSTAMRegisterF(pDrvIns, &pData->StatUDP, STAMTYPE_COUNTER, STAMVISIBILITY_ALWAYS,
382	STAMUNIT_COUNT, "UDP sockets", "/Drivers/NAT%d/SockUDP", pDrvIns->iInstance);
383	PDMDrvHlpSTAMRegisterF(pDrvIns, &pData->StatUDPHot, STAMTYPE_COUNTER, STAMVISIBILITY_ALWAYS,
384	STAMUNIT_COUNT, "UDP sockets active", "/Drivers/NAT%d/SockUDPHot", pDrvIns->iInstance);
385	#endif /* VBOX_WITHOUT_RELEASE_STATISTICS */
386	}
387
388	/**
389	* Marks the link as up, making it possible to establish new connections.
390	*/
391	void slirp_link_up(PNATState pData)
392	{
393	link_up = 1;
394	}
395
396	/**
397	* Marks the link as down and cleans up the current connections.
398	*/
399	void slirp_link_down(PNATState pData)
400	{
401	struct socket *so;
402
403	while ((so = tcb.so_next) != &tcb)
404	{
405	if (so->so_state & SS_NOFDREF \|\| so->s == -1)
406	sofree(pData, so);
407	else
408	tcp_drop(pData, sototcpcb(so), 0);
409	}
410
411	while ((so = udb.so_next) != &udb)
412	udp_detach(pData, so);
413
414	link_up = 0;
415	}
416
417	/**
418	* Terminates the slirp component.
419	*/
420	void slirp_term(PNATState pData)
421	{
422	if (pData->pszDomain)
423	RTStrFree((char )(void )pData->pszDomain);
424
425	#if ARCH_BITS == 64 && !defined(VBOX_WITH_BSD_REASS)
426	LogRel(("NAT: cpvHashUsed=%RU32 cpvHashCollisions=%RU32 cpvHashInserts=%RU64 cpvHashDone=%RU64\n",
427	pData->cpvHashUsed, pData->cpvHashCollisions, pData->cpvHashInserts, pData->cpvHashDone));
428	#endif
429
430	#ifdef VBOX_WITH_SLIRP_ICMP
431	# ifdef RT_OS_WINDOWS
432	pData->pfIcmpCloseHandle(pData->icmp_socket.sh);
433	# else
434	closesocket(pData->icmp_socket.s);
435	# endif
436	#endif
437
438	slirp_link_down(pData);
439	#ifdef RT_OS_WINDOWS
440	WSACleanup();
441	#endif
442	#ifdef LOG_ENABLED
443	Log(("\n"
444	"NAT statistics\n"
445	"--------------\n"
446	"\n"));
447	ipstats(pData);
448	tcpstats(pData);
449	udpstats(pData);
450	icmpstats(pData);
451	mbufstats(pData);
452	sockstats(pData);
453	Log(("\n"
454	"\n"
455	"\n"));
456	#endif
457	free(pData);
458	}
459
460
461	#define CONN_CANFSEND(so) (((so)->so_state & (SS_FCANTSENDMORE\|SS_ISFCONNECTED)) == SS_ISFCONNECTED)
462	#define CONN_CANFRCV(so) (((so)->so_state & (SS_FCANTRCVMORE\|SS_ISFCONNECTED)) == SS_ISFCONNECTED)
463	#define UPD_NFDS(x) if (nfds < (x)) nfds = (x)
464
465	/*
466	* curtime kept to an accuracy of 1ms
467	*/
468	#ifdef RT_OS_WINDOWS
469	static void updtime(PNATState pData)
470	{
471	struct _timeb tb;
472
473	_ftime(&tb);
474	curtime = (u_int)tb.time * (u_int)1000;
475	curtime += (u_int)tb.millitm;
476	}
477	#else
478	static void updtime(PNATState pData)
479	{
480	gettimeofday(&tt, 0);
481
482	curtime = (u_int)tt.tv_sec * (u_int)1000;
483	curtime += (u_int)tt.tv_usec / (u_int)1000;
484
485	if ((tt.tv_usec % 1000) >= 500)
486	curtime++;
487	}
488	#endif
489
490	void slirp_select_fill(PNATState pData, int *pnfds,
491	fd_set readfds, fd_set writefds, fd_set *xfds)
492	{
493	struct socket so, so_next;
494	int nfds;
495	#if defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) && defined(RT_OS_WINDOWS)
496	int rc;
497	int error;
498	#endif
499	#ifdef VBOX_WITH_BSD_REASS
500	int i;
501	#endif /* VBOX_WITH_BSD_REASS */
502
503	STAM_REL_PROFILE_START(&pData->StatFill, a);
504
505	nfds = *pnfds;
506
507	/*
508	* First, TCP sockets
509	*/
510	do_slowtimo = 0;
511	if (link_up)
512	{
513	/*
514	* *_slowtimo needs calling if there are IP fragments
515	* in the fragment queue, or there are TCP connections active
516	*/
517	#ifndef VBOX_WITH_BSD_REASS
518	do_slowtimo = ((tcb.so_next != &tcb)
519	\|\| ((struct ipasfrag )&ipq != u32_to_ptr(pData, ipq.next, struct ipasfrag )));
520	#else /* !VBOX_WITH_BSD_REASS */
521	/* XXX:
522	* triggering of fragment expiration should be the same but use new macroses
523	*/
524	do_slowtimo = (tcb.so_next != &tcb);
525	if (!do_slowtimo)
526	{
527	for (i = 0; i < IPREASS_NHASH; i++)
528	{
529	if (!TAILQ_EMPTY(&ipq[i]))
530	{
531	do_slowtimo = 1;
532	break;
533	}
534	}
535	}
536	#endif /* VBOX_WITH_BSD_REASS */
537
538	STAM_REL_COUNTER_RESET(&pData->StatTCP);
539	STAM_REL_COUNTER_RESET(&pData->StatTCPHot);
540
541	for (so = tcb.so_next; so != &tcb; so = so_next)
542	{
543	so_next = so->so_next;
544
545	STAM_REL_COUNTER_INC(&pData->StatTCP);
546
547	/*
548	* See if we need a tcp_fasttimo
549	*/
550	if (time_fasttimo == 0 && so->so_tcpcb->t_flags & TF_DELACK)
551	time_fasttimo = curtime; /* Flag when we want a fasttimo */
552
553	/*
554	* NOFDREF can include still connecting to local-host,
555	* newly socreated() sockets etc. Don't want to select these.
556	*/
557	if (so->so_state & SS_NOFDREF \|\| so->s == -1)
558	continue;
559
560	/*
561	* Set for reading sockets which are accepting
562	*/
563	if (so->so_state & SS_FACCEPTCONN)
564	{
565	STAM_REL_COUNTER_INC(&pData->StatTCPHot);
566	TCP_ENGAGE_EVENT1(so, readfds);
567	continue;
568	}
569
570	/*
571	* Set for writing sockets which are connecting
572	*/
573	if (so->so_state & SS_ISFCONNECTING)
574	{
575	STAM_REL_COUNTER_INC(&pData->StatTCPHot);
576	TCP_ENGAGE_EVENT1(so, writefds);
577	}
578
579	/*
580	* Set for writing if we are connected, can send more, and
581	* we have something to send
582	*/
583	if (CONN_CANFSEND(so) && so->so_rcv.sb_cc)
584	{
585	STAM_REL_COUNTER_INC(&pData->StatTCPHot);
586	TCP_ENGAGE_EVENT1(so, writefds);
587	}
588
589	/*
590	* Set for reading (and urgent data) if we are connected, can
591	* receive more, and we have room for it XXX /2 ?
592	*/
593	if (CONN_CANFRCV(so) && (so->so_snd.sb_cc < (so->so_snd.sb_datalen/2)))
594	{
595	STAM_REL_COUNTER_INC(&pData->StatTCPHot);
596	TCP_ENGAGE_EVENT2(so, readfds, xfds);
597	}
598	}
599
600	/*
601	* UDP sockets
602	*/
603	STAM_REL_COUNTER_RESET(&pData->StatUDP);
604	STAM_REL_COUNTER_RESET(&pData->StatUDPHot);
605
606	for (so = udb.so_next; so != &udb; so = so_next)
607	{
608	so_next = so->so_next;
609
610	STAM_REL_COUNTER_INC(&pData->StatUDP);
611
612	/*
613	* See if it's timed out
614	*/
615	if (so->so_expire)
616	{
617	if (so->so_expire <= curtime)
618	{
619	udp_detach(pData, so);
620	continue;
621	}
622	else
623	do_slowtimo = 1; /* Let socket expire */
624	}
625
626	/*
627	* When UDP packets are received from over the link, they're
628	* sendto()'d straight away, so no need for setting for writing
629	* Limit the number of packets queued by this session to 4.
630	* Note that even though we try and limit this to 4 packets,
631	* the session could have more queued if the packets needed
632	* to be fragmented.
633	*
634	* (XXX <= 4 ?)
635	*/
636	if ((so->so_state & SS_ISFCONNECTED) && so->so_queued <= 4)
637	{
638	STAM_REL_COUNTER_INC(&pData->StatUDPHot);
639	UDP_ENGAGE_EVENT(so, readfds);
640	}
641	}
642
643	ICMP_ENGAGE_EVENT(&pData->icmp_socket, readfds);
644	}
645
646	#if !defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) \|\| !defined(RT_OS_WINDOWS)
647	*pnfds = nfds;
648	#else
649	*pnfds = VBOX_EVENT_COUNT;
650	#endif
651
652	STAM_REL_PROFILE_STOP(&pData->StatFill, a);
653	}
654
655	void slirp_select_poll(PNATState pData, fd_set readfds, fd_set writefds, fd_set *xfds)
656	{
657	struct socket so, so_next;
658	int ret;
659	#if defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) && defined(RT_OS_WINDOWS)
660	WSANETWORKEVENTS NetworkEvents;
661	int rc;
662	int error;
663	#endif
664	static uint32_t stat_time;
665
666	STAM_REL_PROFILE_START(&pData->StatPoll, a);
667
668	/* Update time */
669	updtime(pData);
670
671	#ifdef LOG_ENABLED
672	if (curtime - stat_time > 10000)
673	{
674	stat_time = curtime;
675	sockstats(pData);
676	}
677	#endif
678
679	/*
680	* See if anything has timed out
681	*/
682	if (link_up)
683	{
684	if (time_fasttimo && ((curtime - time_fasttimo) >= 2))
685	{
686	STAM_REL_PROFILE_START(&pData->StatFastTimer, a);
687	tcp_fasttimo(pData);
688	time_fasttimo = 0;
689	STAM_REL_PROFILE_STOP(&pData->StatFastTimer, a);
690	}
691	if (do_slowtimo && ((curtime - last_slowtimo) >= 499))
692	{
693	STAM_REL_PROFILE_START(&pData->StatSlowTimer, a);
694	ip_slowtimo(pData);
695	tcp_slowtimo(pData);
696	last_slowtimo = curtime;
697	STAM_REL_PROFILE_STOP(&pData->StatSlowTimer, a);
698	}
699	}
700	#if defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) && defined(RT_OS_WINDOWS)
701	if (!readfds && !writefds && !xfds)
702	return; /* only timer update */
703	#endif
704
705	/*
706	* Check sockets
707	*/
708	if (link_up)
709	{
710	/*
711	* Check TCP sockets
712	*/
713	for (so = tcb.so_next; so != &tcb; so = so_next)
714	{
715	so_next = so->so_next;
716
717	/*
718	* FD_ISSET is meaningless on these sockets
719	* (and they can crash the program)
720	*/
721	if (so->so_state & SS_NOFDREF \|\| so->s == -1)
722	continue;
723
724	POLL_TCP_EVENTS(rc, error, so, &NetworkEvents);
725
726	/*
727	* Check for URG data
728	* This will soread as well, so no need to
729	* test for readfds below if this succeeds
730	*/
731
732	/* out-of-band data */
733	if (CHECK_FD_SET(so, NetworkEvents, xfds))
734	{
735	sorecvoob(pData, so);
736	}
737
738	/*
739	* Check sockets for reading
740	*/
741	else if (CHECK_FD_SET(so, NetworkEvents, readfds))
742	{
743	/*
744	* Check for incoming connections
745	*/
746	if (so->so_state & SS_FACCEPTCONN)
747	{
748	tcp_connect(pData, so);
749	#if defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) && defined(RT_OS_WINDOWS)
750	if (!(NetworkEvents.lNetworkEvents & FD_CLOSE))
751	#endif
752	continue;
753	}
754
755	ret = soread(pData, so);
756	/* Output it if we read something */
757	if (ret > 0)
758	tcp_output(pData, sototcpcb(so));
759	}
760
761	#if defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) && defined(RT_OS_WINDOWS)
762	/*
763	* Check for FD_CLOSE events.
764	*/
765	if (NetworkEvents.lNetworkEvents & FD_CLOSE)
766	{
767	/*
768	* drain the socket
769	*/
770	for (;;)
771	{
772	ret = soread(pData, so);
773	if (ret > 0)
774	tcp_output(pData, sototcpcb(so));
775	else
776	break;
777	}
778	}
779	#endif
780
781	/*
782	* Check sockets for writing
783	*/
784	if (CHECK_FD_SET(so, NetworkEvents, writefds))
785	{
786	/*
787	* Check for non-blocking, still-connecting sockets
788	*/
789	if (so->so_state & SS_ISFCONNECTING)
790	{
791	/* Connected */
792	so->so_state &= ~SS_ISFCONNECTING;
793
794	/*
795	* This should be probably guarded by PROBE_CONN too. Anyway,
796	* we disable it on OS/2 because the below send call returns
797	* EFAULT which causes the opened TCP socket to close right
798	* after it has been opened and connected.
799	*/
800	#ifndef RT_OS_OS2
801	ret = send(so->s, (const char *)&ret, 0, 0);
802	if (ret < 0)
803	{
804	/* XXXXX Must fix, zero bytes is a NOP */
805	if ( errno == EAGAIN
806	\|\| errno == EWOULDBLOCK
807	\|\| errno == EINPROGRESS
808	\|\| errno == ENOTCONN)
809	continue;
810
811	/* else failed */
812	so->so_state = SS_NOFDREF;
813	}
814	/* else so->so_state &= ~SS_ISFCONNECTING; */
815	#endif
816
817	/*
818	* Continue tcp_input
819	*/
820	tcp_input(pData, (struct mbuf *)NULL, sizeof(struct ip), so);
821	/* continue; */
822	}
823	else
824	ret = sowrite(pData, so);
825	/*
826	* XXX If we wrote something (a lot), there could be the need
827	* for a window update. In the worst case, the remote will send
828	* a window probe to get things going again.
829	*/
830	}
831
832	/*
833	* Probe a still-connecting, non-blocking socket
834	* to check if it's still alive
835	*/
836	#ifdef PROBE_CONN
837	if (so->so_state & SS_ISFCONNECTING)
838	{
839	ret = recv(so->s, (char *)&ret, 0, 0);
840
841	if (ret < 0)
842	{
843	/* XXX */
844	if ( errno == EAGAIN
845	\|\| errno == EWOULDBLOCK
846	\|\| errno == EINPROGRESS
847	\|\| errno == ENOTCONN)
848	{
849	continue; /* Still connecting, continue */
850	}
851
852	/* else failed */
853	so->so_state = SS_NOFDREF;
854
855	/* tcp_input will take care of it */
856	}
857	else
858	{
859	ret = send(so->s, &ret, 0, 0);
860	if (ret < 0)
861	{
862	/* XXX */
863	if ( errno == EAGAIN
864	\|\| errno == EWOULDBLOCK
865	\|\| errno == EINPROGRESS
866	\|\| errno == ENOTCONN)
867	{
868	continue;
869	}
870	/* else failed */
871	so->so_state = SS_NOFDREF;
872	}
873	else
874	so->so_state &= ~SS_ISFCONNECTING;
875
876	}
877	tcp_input((struct mbuf *)NULL, sizeof(struct ip),so);
878	} /* SS_ISFCONNECTING */
879	#endif
880	}
881
882	/*
883	* Now UDP sockets.
884	* Incoming packets are sent straight away, they're not buffered.
885	* Incoming UDP data isn't buffered either.
886	*/
887	for (so = udb.so_next; so != &udb; so = so_next)
888	{
889	so_next = so->so_next;
890
891	POLL_UDP_EVENTS(rc, error, so, &NetworkEvents);
892
893	if (so->s != -1 && CHECK_FD_SET(so, NetworkEvents, readfds))
894	{
895	sorecvfrom(pData, so);
896	}
897	}
898
899	#if defined(VBOX_WITH_SLIRP_ICMP)
900	# if defined(RT_OS_WINDOWS)
901	sorecvfrom(pData, &pData->icmp_socket);
902	# else
903	if (pData->icmp_socket.s != -1 && FD_ISSET(pData->icmp_socket.s, readfds))
904	sorecvfrom(pData, &pData->icmp_socket);
905	# endif
906	#endif
907	}
908
909	/*
910	* See if we can start outputting
911	*/
912	if (if_queued && link_up)
913	if_start(pData);
914
915	STAM_REL_PROFILE_STOP(&pData->StatPoll, a);
916	}
917
918	#define ETH_ALEN 6
919	#define ETH_HLEN 14
920
921	#define ETH_P_IP 0x0800 /* Internet Protocol packet */
922	#define ETH_P_ARP 0x0806 /* Address Resolution packet */
923
924	#define ARPOP_REQUEST 1 /* ARP request */
925	#define ARPOP_REPLY 2 /* ARP reply */
926
927	struct ethhdr
928	{
929	unsigned char h_dest[ETH_ALEN]; /* destination eth addr */
930	unsigned char h_source[ETH_ALEN]; /* source ether addr */
931	unsigned short h_proto; /* packet type ID field */
932	};
933
934	struct arphdr
935	{
936	unsigned short ar_hrd; /* format of hardware address */
937	unsigned short ar_pro; /* format of protocol address */
938	unsigned char ar_hln; /* length of hardware address */
939	unsigned char ar_pln; /* length of protocol address */
940	unsigned short ar_op; /* ARP opcode (command) */
941
942	/*
943	* Ethernet looks like this : This bit is variable sized however...
944	*/
945	unsigned char ar_sha[ETH_ALEN]; /* sender hardware address */
946	unsigned char ar_sip[4]; /* sender IP address */
947	unsigned char ar_tha[ETH_ALEN]; /* target hardware address */
948	unsigned char ar_tip[4]; /* target IP address */
949	};
950
951	static
952	void arp_input(PNATState pData, const uint8_t *pkt, int pkt_len)
953	{
954	struct ethhdr eh = (struct ethhdr )pkt;
955	struct arphdr ah = (struct arphdr )(pkt + ETH_HLEN);
956	uint8_t arp_reply[ETH_HLEN + sizeof(struct arphdr)];
957	struct ethhdr reh = (struct ethhdr )arp_reply;
958	struct arphdr rah = (struct arphdr )(arp_reply + ETH_HLEN);
959	int ar_op;
960	struct ex_list *ex_ptr;
961	uint32_t htip = ntohl((uint32_t)ah->ar_tip);
962
963	ar_op = ntohs(ah->ar_op);
964	switch(ar_op)
965	{
966	case ARPOP_REQUEST:
967	if ((htip & pData->netmask) == ntohl(special_addr.s_addr))
968	{
969	if ( (htip & ~pData->netmask) == CTL_DNS
970	\|\| (htip & ~pData->netmask) == CTL_ALIAS)
971	goto arp_ok;
972	for (ex_ptr = exec_list; ex_ptr; ex_ptr = ex_ptr->ex_next)
973	{
974	if ((htip & ~pData->netmask) == ex_ptr->ex_addr)
975	goto arp_ok;
976	}
977	return;
978	arp_ok:
979	/* XXX: make an ARP request to have the client address */
980	memcpy(client_ethaddr, eh->h_source, ETH_ALEN);
981
982	/* ARP request for alias/dns mac address */
983	memcpy(reh->h_dest, pkt + ETH_ALEN, ETH_ALEN);
984	memcpy(reh->h_source, special_ethaddr, ETH_ALEN - 1);
985	reh->h_source[5] = ah->ar_tip[3];
986	reh->h_proto = htons(ETH_P_ARP);
987
988	rah->ar_hrd = htons(1);
989	rah->ar_pro = htons(ETH_P_IP);
990	rah->ar_hln = ETH_ALEN;
991	rah->ar_pln = 4;
992	rah->ar_op = htons(ARPOP_REPLY);
993	memcpy(rah->ar_sha, reh->h_source, ETH_ALEN);
994	memcpy(rah->ar_sip, ah->ar_tip, 4);
995	memcpy(rah->ar_tha, ah->ar_sha, ETH_ALEN);
996	memcpy(rah->ar_tip, ah->ar_sip, 4);
997	slirp_output(pData->pvUser, arp_reply, sizeof(arp_reply));
998	}
999	break;
1000	default:
1001	break;
1002	}
1003	}
1004
1005	void slirp_input(PNATState pData, const uint8_t *pkt, int pkt_len)
1006	{
1007	struct mbuf *m;
1008	int proto;
1009
1010	if (pkt_len < ETH_HLEN)
1011	return;
1012
1013	proto = ntohs((uint16_t )(pkt + 12));
1014	switch(proto)
1015	{
1016	case ETH_P_ARP:
1017	arp_input(pData, pkt, pkt_len);
1018	break;
1019	case ETH_P_IP:
1020	/* Update time. Important if the network is very quiet, as otherwise
1021	* the first outgoing connection gets an incorrect timestamp. */
1022	updtime(pData);
1023
1024	m = m_get(pData);
1025	if (!m)
1026	return;
1027	/* Note: we add to align the IP header */
1028	if (M_FREEROOM(m) < pkt_len + 2)
1029	{
1030	m_inc(m, pkt_len + 2);
1031	}
1032	m->m_len = pkt_len + 2;
1033	memcpy(m->m_data + 2, pkt, pkt_len);
1034
1035	m->m_data += 2 + ETH_HLEN;
1036	m->m_len -= 2 + ETH_HLEN;
1037
1038	ip_input(pData, m);
1039	break;
1040	default:
1041	break;
1042	}
1043	}
1044
1045	/* output the IP packet to the ethernet device */
1046	void if_encap(PNATState pData, const uint8_t *ip_data, int ip_data_len)
1047	{
1048	uint8_t buf[1600];
1049	struct ethhdr eh = (struct ethhdr )buf;
1050
1051	if (ip_data_len + ETH_HLEN > sizeof(buf))
1052	return;
1053
1054	memcpy(eh->h_dest, client_ethaddr, ETH_ALEN);
1055	memcpy(eh->h_source, special_ethaddr, ETH_ALEN - 1);
1056	/* XXX: not correct */
1057	eh->h_source[5] = CTL_ALIAS;
1058	eh->h_proto = htons(ETH_P_IP);
1059	memcpy(buf + sizeof(struct ethhdr), ip_data, ip_data_len);
1060	slirp_output(pData->pvUser, buf, ip_data_len + ETH_HLEN);
1061	}
1062
1063	int slirp_redir(PNATState pData, int is_udp, int host_port,
1064	struct in_addr guest_addr, int guest_port)
1065	{
1066	if (is_udp)
1067	{
1068	if (!udp_listen(pData, htons(host_port), guest_addr.s_addr,
1069	htons(guest_port), 0))
1070	return -1;
1071	}
1072	else
1073	{
1074	if (!solisten(pData, htons(host_port), guest_addr.s_addr,
1075	htons(guest_port), 0))
1076	return -1;
1077	}
1078	return 0;
1079	}
1080
1081	int slirp_add_exec(PNATState pData, int do_pty, const char *args, int addr_low_byte,
1082	int guest_port)
1083	{
1084	return add_exec(&exec_list, do_pty, (char *)args,
1085	addr_low_byte, htons(guest_port));
1086	}
1087
1088	void slirp_set_ethaddr(PNATState pData, const uint8_t *ethaddr)
1089	{
1090	memcpy(client_ethaddr, ethaddr, ETH_ALEN);
1091	}
1092
1093	#if defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) && defined(RT_OS_WINDOWS)
1094	HANDLE *slirp_get_events(PNATState pData)
1095	{
1096	return pData->phEvents;
1097	}
1098	void slirp_register_external_event(PNATState pData, HANDLE hEvent, int index)
1099	{
1100	pData->phEvents[index] = hEvent;
1101	}
1102	#endif
1103
1104	unsigned int slirp_get_timeout_ms(PNATState pData)
1105	{
1106	if (link_up)
1107	{
1108	if (time_fasttimo)
1109	return 2;
1110	if (do_slowtimo)
1111	return 500; /* see PR_SLOWHZ */
1112	}
1113	return 0;
1114	}

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

source: vbox/trunk/src/VBox/Devices/Network/slirp/slirp.c@ 15230

Download in other formats: