slirp.c@ 16226

最後變更在這個檔案從16226是 16226,由 vboxsync 提交於 16 年前

NAT:

DNS several DNS servers could be passed to guest now (with VBOX_WITH_MULTI_DNS)
on host server (if it registered) aliased with CTL_ALIAS, there are no reason use CTL_DNS for this puposes anymore, instead it could be used for built-in DNS if it will require.

屬性 svn:eol-style 設為 native

檔案大小: 37.7 KB

行
1	#include "slirp.h"
2	#ifdef RT_OS_OS2
3	# include <paths.h>
4	#endif
5
6	#include <VBox/err.h>
7	#include <VBox/pdmdrv.h>
8	#include <iprt/assert.h>
9
10	#if !defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) \|\| !defined(RT_OS_WINDOWS)
11
12	# define DO_ENGAGE_EVENT1(so, fdset, label) \
13	do { \
14	FD_SET((so)->s, (fdset)); \
15	UPD_NFDS((so)->s); \
16	} while(0)
17
18
19	# define DO_ENGAGE_EVENT2(so, fdset1, fdset2, label) \
20	do { \
21	FD_SET((so)->s, (fdset1)); \
22	FD_SET((so)->s, (fdset2)); \
23	UPD_NFDS((so)->s); \
24	} while(0)
25
26	# define DO_POLL_EVENTS(rc, error, so, events, label) do {} while (0)
27
28	# define DO_CHECK_FD_SET(so, events, fdset) (FD_ISSET((so)->s, (fdset)))
29
30	# define DO_WIN_CHECK_FD_SET(so, events, fdset ) 0 /* specific for Windows Winsock API */
31
32	# ifndef RT_OS_WINDOWS
33	# define ICMP_ENGAGE_EVENT(so, fdset) \
34	do { \
35	if (pData->icmp_socket.s != -1) \
36	DO_ENGAGE_EVENT1((so), (fdset), ICMP); \
37	} while (0)
38	# else /* !RT_OS_WINDOWS */
39	# define ICMP_ENGAGE_EVENT(so, fdset) do {} while(0)
40	#endif /* RT_OS_WINDOWS */
41
42	#else /* defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) && defined(RT_OS_WINDOWS) */
43
44	/*
45	* On Windows, we will be notified by IcmpSendEcho2() when the response arrives.
46	* So no call to WSAEventSelect necessary.
47	*/
48	# define ICMP_ENGAGE_EVENT(so, fdset) do {} while(0)
49
50	# define DO_ENGAGE_EVENT1(so, fdset1, label) \
51	do { \
52	rc = WSAEventSelect((so)->s, VBOX_SOCKET_EVENT, FD_ALL_EVENTS); \
53	if (rc == SOCKET_ERROR) \
54	{ \
55	/* This should not happen */ \
56	error = WSAGetLastError(); \
57	LogRel(("WSAEventSelector (" #label ") error %d (so=%x, socket=%s, event=%x)\n", \
58	error, (so), (so)->s, VBOX_SOCKET_EVENT)); \
59	} \
60	} while(0); \
61	continue
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 acceptds_win FD_ACCEPT
76	# define acceptds_win_bit FD_ACCEPT_BIT
77
78	# define readfds_win FD_READ
79	# define readfds_win_bit FD_READ_BIT
80
81	# define writefds_win FD_WRITE
82	# define writefds_win_bit FD_WRITE_BIT
83
84	# define xfds_win FD_OOB
85	# define xfds_win_bit FD_OOB_BIT
86
87	# define DO_CHECK_FD_SET(so, events, fdset) \
88	(((events).lNetworkEvents & fdset ## _win) && ((events).iErrorCode[fdset ## _win_bit] == 0))
89
90	# define DO_WIN_CHECK_FD_SET(so, events, fdset ) DO_CHECK_FD_SET((so), (events), fdset)
91
92	#endif /* defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) && defined(RT_OS_WINDOWS) */
93
94	#define TCP_ENGAGE_EVENT1(so, fdset) \
95	DO_ENGAGE_EVENT1((so), (fdset), TCP)
96
97	#define TCP_ENGAGE_EVENT2(so, fdset1, fdset2) \
98	DO_ENGAGE_EVENT2((so), (fdset1), (fdset2), TCP)
99
100	#define UDP_ENGAGE_EVENT(so, fdset) \
101	DO_ENGAGE_EVENT1((so), (fdset), UDP)
102
103	#define POLL_TCP_EVENTS(rc, error, so, events) \
104	DO_POLL_EVENTS((rc), (error), (so), (events), TCP)
105
106	#define POLL_UDP_EVENTS(rc, error, so, events) \
107	DO_POLL_EVENTS((rc), (error), (so), (events), UDP)
108
109	#define CHECK_FD_SET(so, events, set) \
110	(DO_CHECK_FD_SET((so), (events), set))
111
112	#define WIN_CHECK_FD_SET(so, events, set) \
113	(DO_WIN_CHECK_FD_SET((so), (events), set))
114
115	/*
116	* Loging macros
117	*/
118	#if VBOX_WITH_DEBUG_NAT_SOCKETS
119	# if defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) && defined(RT_OS_WINDOWS)
120	# define DO_LOG_NAT_SOCK(so, proto, winevent, r_fdset, w_fdset, x_fdset) \
121	do { \
122	LogRel((" " #proto "%R[natsock] %R[natwinnetevents]\n", (so), (winevent))); \
123	} while (0)
124	# else
125	# define DO_LOG_NAT_SOCK(so, proto, winevent, r_fdset, w_fdset, x_fdset) \
126	do { \
127	LogRel((" " #proto " %R[natsock] %s %s %s\n", (so), FD_ISSET((so)->s, (r_fdset))?"READ":"",\
128	FD_ISSET((so)->s, (w_fdset))?"WRITE":"", FD_ISSET((so)->s, (x_fdset))?"OOB":"")); \
129	} while (0)
130	# endif /* VBOX_WITH_DEBUG_NAT_SOCKETS */
131	#else
132	# define DO_LOG_NAT_SOCK(so, proto, winevent, r_fdset, w_fdset, x_fdset) do {} while (0)
133	#endif /* !VBOX_WITH_DEBUG_NAT_SOCKETS */
134
135	#define LOG_NAT_SOCK(so, proto, winevent, r_fdset, w_fdset, x_fdset) DO_LOG_NAT_SOCK((so), proto, (winevent), (r_fdset), (w_fdset), (x_fdset))
136
137	static const uint8_t special_ethaddr[6] =
138	{
139	0x52, 0x54, 0x00, 0x12, 0x35, 0x00
140	};
141
142	#ifdef RT_OS_WINDOWS
143
144	static int get_dns_addr_domain(PNATState pData, bool fVerbose,
145	struct in_addr *pdns_addr,
146	const char **ppszDomain)
147	{
148	int rc = 0;
149	FIXED_INFO *FixedInfo = NULL;
150	ULONG BufLen;
151	DWORD ret;
152	IP_ADDR_STRING *pIPAddr;
153	struct in_addr tmp_addr;
154	#ifdef VBOX_WITH_MULTI_DNS
155	struct in_addr local_mask;
156	struct in_addr local_network;
157	#endif
158
159	FixedInfo = (FIXED_INFO *)GlobalAlloc(GPTR, sizeof(FIXED_INFO));
160	BufLen = sizeof(FIXED_INFO);
161
162	/** @todo: this API returns all DNS servers, no matter whether the
163	* corresponding network adapter is disabled or not. Maybe replace
164	* this by GetAdapterAddresses(), which is XP/Vista only though. */
165	if (ERROR_BUFFER_OVERFLOW == GetNetworkParams(FixedInfo, &BufLen))
166	{
167	if (FixedInfo)
168	{
169	GlobalFree(FixedInfo);
170	FixedInfo = NULL;
171	}
172	FixedInfo = GlobalAlloc(GPTR, BufLen);
173	}
174
175	if ((ret = GetNetworkParams(FixedInfo, &BufLen)) != ERROR_SUCCESS)
176	{
177	Log(("GetNetworkParams failed. ret = %08x\n", (u_int)ret ));
178	if (FixedInfo)
179	{
180	GlobalFree(FixedInfo);
181	FixedInfo = NULL;
182	}
183	rc = -1;
184	goto get_dns_prefix;
185	}
186
187	#ifndef VBOX_WITH_MULTI_DNS
188	pIPAddr = &(FixedInfo->DnsServerList);
189	inet_aton(pIPAddr->IpAddress.String, &tmp_addr);
190	Log(("nat: DNS Servers:\n"));
191	if (fVerbose \|\| pdns_addr->s_addr != tmp_addr.s_addr)
192	LogRel(("NAT: DNS address: %s\n", pIPAddr->IpAddress.String));
193	*pdns_addr = tmp_addr;
194
195	pIPAddr = FixedInfo -> DnsServerList.Next;
196	while (pIPAddr)
197	{
198	if (fVerbose)
199	LogRel(("NAT: ignored DNS address: %s\n", pIPAddr ->IpAddress.String));
200	pIPAddr = pIPAddr ->Next;
201	}
202	#else
203	/localhost mask /
204	inet_aton("255.0.0.0", &local_mask);
205	inet_aton("127.0.0.0", &local_network);
206	for (pIPAddr = &FixedInfo->DnsServerList; pIPAddr != NULL; pIPAddr = pIPAddr->Next)
207	{
208	struct dns_entry *da = RTMemAllocZ(sizeof (struct dns_entry));
209	if (da == NULL)
210	{
211	LogRel(("can't alloc memory for DNS entry\n"));
212	return -1;
213	}
214	/check /
215	inet_aton(pIPAddr->IpAddress.String, &da->de_addr);
216	if ((ntohl(da->de_addr.s_addr) & ntohl(local_mask.s_addr)) == ntohl(local_network.s_addr)) {
217	da->de_addr.s_addr = htonl(ntohl(special_addr.s_addr) \| CTL_ALIAS);
218	}
219	LIST_INSERT_HEAD(&pData->dns_list_head, da, de_list);
220	}
221	#endif
222	if (FixedInfo)
223	{
224	GlobalFree(FixedInfo);
225	FixedInfo = NULL;
226	}
227
228	get_dns_prefix:
229	if (ppszDomain)
230	{
231	OSVERSIONINFO ver;
232	char szDnsDomain[256];
233	DWORD dwSize = sizeof(szDnsDomain);
234
235	*ppszDomain = NULL;
236	GetVersionEx(&ver);
237	if (ver.dwMajorVersion >= 5)
238	{
239	/* GetComputerNameEx exists in Windows versions starting with 2000. */
240	if (GetComputerNameEx(ComputerNameDnsDomain, szDnsDomain, &dwSize))
241	{
242	if (szDnsDomain[0])
243	{
244	/* Just non-empty strings are valid. */
245	*ppszDomain = RTStrDup(szDnsDomain);
246	if (pData->fPassDomain)
247	{
248	if (fVerbose)
249	LogRel(("NAT: passing domain name %s\n", szDnsDomain));
250	}
251	else
252	Log(("nat: ignoring domain %s\n", szDnsDomain));
253	}
254	}
255	else
256	Log(("nat: GetComputerNameEx failed (%d)\n", GetLastError()));
257	}
258	}
259	return rc;
260	}
261
262	#else
263
264	static int get_dns_addr_domain(PNATState pData, bool fVerbose,
265	struct in_addr *pdns_addr,
266	const char **ppszDomain)
267	{
268	char buff[512];
269	char buff2[256];
270	FILE *f;
271	int found = 0;
272	struct in_addr tmp_addr;
273
274	#ifdef RT_OS_OS2
275	/* Try various locations. */
276	char *etc = getenv("ETC");
277	f = NULL;
278	if (etc)
279	{
280	snprintf(buff, sizeof(buff), "%s/RESOLV2", etc);
281	f = fopen(buff, "rt");
282	}
283	if (!f)
284	{
285	snprintf(buff, sizeof(buff), "%s/RESOLV2", _PATH_ETC);
286	f = fopen(buff, "rt");
287	}
288	if (!f)
289	{
290	snprintf(buff, sizeof(buff), "%s/resolv.conf", _PATH_ETC);
291	f = fopen(buff, "rt");
292	}
293	#else
294	f = fopen("/etc/resolv.conf", "r");
295	#endif
296	if (!f)
297	return -1;
298
299	if (ppszDomain)
300	*ppszDomain = NULL;
301	Log(("nat: DNS Servers:\n"));
302	while (fgets(buff, 512, f) != NULL)
303	{
304	if (sscanf(buff, "nameserver%*[ \t]%256s", buff2) == 1)
305	{
306	if (!inet_aton(buff2, &tmp_addr))
307	continue;
308	/* If it's the first one, set it to dns_addr */
309	if (!found)
310	{
311	if (fVerbose \|\| pdns_addr->s_addr != tmp_addr.s_addr)
312	LogRel(("NAT: DNS address: %s\n", buff2));
313	*pdns_addr = tmp_addr;
314	}
315	else
316	{
317	if (fVerbose)
318	LogRel(("NAT: ignored DNS address: %s\n", buff2));
319	}
320	found++;
321	}
322	if ( ppszDomain
323	&& (!strncmp(buff, "domain", 6) \|\| !strncmp(buff, "search", 6)))
324	{
325	/* Domain name/search list present. Pick first entry */
326	if (*ppszDomain == NULL)
327	{
328	char *tok;
329	char *saveptr;
330	tok = strtok_r(&buff[6], " \t\n", &saveptr);
331	if (tok)
332	{
333	*ppszDomain = RTStrDup(tok);
334	if (pData->fPassDomain)
335	{
336	if (fVerbose)
337	LogRel(("NAT: passing domain name %s\n", tok));
338	}
339	else
340	Log(("nat: ignoring domain %s\n", tok));
341	}
342	}
343	}
344	}
345	fclose(f);
346	if (!found)
347	return -1;
348	return 0;
349	}
350
351	#endif
352
353	int get_dns_addr(PNATState pData, struct in_addr *pdns_addr)
354	{
355	return get_dns_addr_domain(pData, false, pdns_addr, NULL);
356	}
357
358	int slirp_init(PNATState ppData, const char pszNetAddr, uint32_t u32Netmask,
359	bool fPassDomain, const char *pszTFTPPrefix,
360	const char pszBootFile, void pvUser)
361	{
362	int fNATfailed = 0;
363	PNATState pData = RTMemAlloc(sizeof(NATState));
364	*ppData = pData;
365	if (!pData)
366	return VERR_NO_MEMORY;
367	if (u32Netmask & 0x1f)
368	/* CTL is x.x.x.15, bootp passes up to 16 IPs (15..31) */
369	return VERR_INVALID_PARAMETER;
370	memset(pData, '\0', sizeof(NATState));
371	pData->fPassDomain = fPassDomain;
372	pData->pvUser = pvUser;
373	tftp_prefix = pszTFTPPrefix;
374	bootp_filename = pszBootFile;
375	pData->netmask = u32Netmask;
376
377	#ifdef RT_OS_WINDOWS
378	{
379	WSADATA Data;
380	WSAStartup(MAKEWORD(2,0), &Data);
381	}
382	# if defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC)
383	pData->phEvents[VBOX_SOCKET_EVENT_INDEX] = CreateEvent(NULL, FALSE, FALSE, NULL);
384	# endif
385	#endif
386
387	link_up = 1;
388
389	debug_init();
390	if_init(pData);
391	ip_init(pData);
392	icmp_init(pData);
393
394	/* Initialise mbufs after setting the MTU */
395	m_init(pData);
396
397	inet_aton(pszNetAddr, &special_addr);
398	alias_addr.s_addr = special_addr.s_addr \| htonl(CTL_ALIAS);
399
400	/* set default addresses */
401	inet_aton("127.0.0.1", &loopback_addr);
402	#ifndef VBOX_WITH_MULTI_DNS
403	inet_aton("127.0.0.1", &dns_addr);
404
405	if (get_dns_addr_domain(pData, true, &dns_addr, &pData->pszDomain) < 0)
406	#else
407	if (get_dns_addr_domain(pData, true, NULL, &pData->pszDomain) < 0)
408	#endif
409	fNATfailed = 1;
410
411	getouraddr(pData);
412	return fNATfailed ? VINF_NAT_DNS : VINF_SUCCESS;
413	}
414
415	/**
416	* Statistics counters.
417	*/
418	void slirp_register_timers(PNATState pData, PPDMDRVINS pDrvIns)
419	{
420	#ifdef VBOX_WITH_STATISTICS
421	PDMDrvHlpSTAMRegisterF(pDrvIns, &pData->StatFill, STAMTYPE_PROFILE, STAMVISIBILITY_ALWAYS,
422	STAMUNIT_TICKS_PER_CALL, "Profiling slirp fills", "/Drivers/NAT%d/Fill", pDrvIns->iInstance);
423	PDMDrvHlpSTAMRegisterF(pDrvIns, &pData->StatPoll, STAMTYPE_PROFILE, STAMVISIBILITY_ALWAYS,
424	STAMUNIT_TICKS_PER_CALL, "Profiling slirp polls", "/Drivers/NAT%d/Poll", pDrvIns->iInstance);
425	PDMDrvHlpSTAMRegisterF(pDrvIns, &pData->StatFastTimer, STAMTYPE_PROFILE, STAMVISIBILITY_ALWAYS,
426	STAMUNIT_TICKS_PER_CALL, "Profiling slirp fast timer", "/Drivers/NAT%d/TimerFast", pDrvIns->iInstance);
427	PDMDrvHlpSTAMRegisterF(pDrvIns, &pData->StatSlowTimer, STAMTYPE_PROFILE, STAMVISIBILITY_ALWAYS,
428	STAMUNIT_TICKS_PER_CALL, "Profiling slirp slow timer", "/Drivers/NAT%d/TimerSlow", pDrvIns->iInstance);
429	PDMDrvHlpSTAMRegisterF(pDrvIns, &pData->StatTCP, STAMTYPE_COUNTER, STAMVISIBILITY_ALWAYS,
430	STAMUNIT_COUNT, "TCP sockets", "/Drivers/NAT%d/SockTCP", pDrvIns->iInstance);
431	PDMDrvHlpSTAMRegisterF(pDrvIns, &pData->StatTCPHot, STAMTYPE_COUNTER, STAMVISIBILITY_ALWAYS,
432	STAMUNIT_COUNT, "TCP sockets active", "/Drivers/NAT%d/SockTCPHot", pDrvIns->iInstance);
433	PDMDrvHlpSTAMRegisterF(pDrvIns, &pData->StatUDP, STAMTYPE_COUNTER, STAMVISIBILITY_ALWAYS,
434	STAMUNIT_COUNT, "UDP sockets", "/Drivers/NAT%d/SockUDP", pDrvIns->iInstance);
435	PDMDrvHlpSTAMRegisterF(pDrvIns, &pData->StatUDPHot, STAMTYPE_COUNTER, STAMVISIBILITY_ALWAYS,
436	STAMUNIT_COUNT, "UDP sockets active", "/Drivers/NAT%d/SockUDPHot", pDrvIns->iInstance);
437	#endif /* VBOX_WITH_STATISTICS */
438	}
439
440	/**
441	* Marks the link as up, making it possible to establish new connections.
442	*/
443	void slirp_link_up(PNATState pData)
444	{
445	link_up = 1;
446	}
447
448	/**
449	* Marks the link as down and cleans up the current connections.
450	*/
451	void slirp_link_down(PNATState pData)
452	{
453	struct socket *so;
454
455	while ((so = tcb.so_next) != &tcb)
456	{
457	if (so->so_state & SS_NOFDREF \|\| so->s == -1)
458	sofree(pData, so);
459	else
460	tcp_drop(pData, sototcpcb(so), 0);
461	}
462
463	while ((so = udb.so_next) != &udb)
464	udp_detach(pData, so);
465
466	link_up = 0;
467	}
468
469	/**
470	* Terminates the slirp component.
471	*/
472	void slirp_term(PNATState pData)
473	{
474	if (pData->pszDomain)
475	RTStrFree((char )(void )pData->pszDomain);
476
477	#ifdef RT_OS_WINDOWS
478	pData->pfIcmpCloseHandle(pData->icmp_socket.sh);
479	FreeLibrary(pData->hmIcmpLibrary);
480	RTMemFree(pData->pvIcmpBuffer);
481	# else
482	closesocket(pData->icmp_socket.s);
483	#endif
484
485	slirp_link_down(pData);
486	#ifdef RT_OS_WINDOWS
487	WSACleanup();
488	#endif
489	#ifdef LOG_ENABLED
490	Log(("\n"
491	"NAT statistics\n"
492	"--------------\n"
493	"\n"));
494	ipstats(pData);
495	tcpstats(pData);
496	udpstats(pData);
497	icmpstats(pData);
498	mbufstats(pData);
499	sockstats(pData);
500	Log(("\n"
501	"\n"
502	"\n"));
503	#endif
504	RTMemFree(pData);
505	}
506
507
508	#define CONN_CANFSEND(so) (((so)->so_state & (SS_FCANTSENDMORE\|SS_ISFCONNECTED)) == SS_ISFCONNECTED)
509	#define CONN_CANFRCV(so) (((so)->so_state & (SS_FCANTRCVMORE\|SS_ISFCONNECTED)) == SS_ISFCONNECTED)
510	#define UPD_NFDS(x) if (nfds < (x)) nfds = (x)
511
512	/*
513	* curtime kept to an accuracy of 1ms
514	*/
515	#ifdef RT_OS_WINDOWS
516	static void updtime(PNATState pData)
517	{
518	struct _timeb tb;
519
520	_ftime(&tb);
521	curtime = (u_int)tb.time * (u_int)1000;
522	curtime += (u_int)tb.millitm;
523	}
524	#else
525	static void updtime(PNATState pData)
526	{
527	gettimeofday(&tt, 0);
528
529	curtime = (u_int)tt.tv_sec * (u_int)1000;
530	curtime += (u_int)tt.tv_usec / (u_int)1000;
531
532	if ((tt.tv_usec % 1000) >= 500)
533	curtime++;
534	}
535	#endif
536
537	void slirp_select_fill(PNATState pData, int *pnfds,
538	fd_set readfds, fd_set writefds, fd_set *xfds)
539	{
540	struct socket so, so_next;
541	int nfds;
542	#if defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) && defined(RT_OS_WINDOWS)
543	int rc;
544	int error;
545	#endif
546	int i;
547
548	STAM_PROFILE_START(&pData->StatFill, a);
549
550	nfds = *pnfds;
551
552	/*
553	* First, TCP sockets
554	*/
555	do_slowtimo = 0;
556	if (link_up)
557	{
558	/*
559	* *_slowtimo needs calling if there are IP fragments
560	* in the fragment queue, or there are TCP connections active
561	*/
562	/* XXX:
563	* triggering of fragment expiration should be the same but use new macroses
564	*/
565	do_slowtimo = (tcb.so_next != &tcb);
566	if (!do_slowtimo)
567	{
568	for (i = 0; i < IPREASS_NHASH; i++)
569	{
570	if (!TAILQ_EMPTY(&ipq[i]))
571	{
572	do_slowtimo = 1;
573	break;
574	}
575	}
576	}
577	ICMP_ENGAGE_EVENT(&pData->icmp_socket, readfds);
578
579	STAM_COUNTER_RESET(&pData->StatTCP);
580	STAM_COUNTER_RESET(&pData->StatTCPHot);
581
582	for (so = tcb.so_next; so != &tcb; so = so_next)
583	{
584	so_next = so->so_next;
585
586	STAM_COUNTER_INC(&pData->StatTCP);
587
588	/*
589	* See if we need a tcp_fasttimo
590	*/
591	if (time_fasttimo == 0 && so->so_tcpcb->t_flags & TF_DELACK)
592	time_fasttimo = curtime; /* Flag when we want a fasttimo */
593
594	/*
595	* NOFDREF can include still connecting to local-host,
596	* newly socreated() sockets etc. Don't want to select these.
597	*/
598	if (so->so_state & SS_NOFDREF \|\| so->s == -1)
599	continue;
600
601	/*
602	* Set for reading sockets which are accepting
603	*/
604	if (so->so_state & SS_FACCEPTCONN)
605	{
606	STAM_COUNTER_INC(&pData->StatTCPHot);
607	TCP_ENGAGE_EVENT1(so, readfds);
608	continue;
609	}
610
611	/*
612	* Set for writing sockets which are connecting
613	*/
614	if (so->so_state & SS_ISFCONNECTING)
615	{
616	Log2(("connecting %R[natsock] engaged\n",so));
617	STAM_COUNTER_INC(&pData->StatTCPHot);
618	TCP_ENGAGE_EVENT1(so, writefds);
619	}
620
621	/*
622	* Set for writing if we are connected, can send more, and
623	* we have something to send
624	*/
625	if (CONN_CANFSEND(so) && so->so_rcv.sb_cc)
626	{
627	STAM_COUNTER_INC(&pData->StatTCPHot);
628	TCP_ENGAGE_EVENT1(so, writefds);
629	}
630
631	/*
632	* Set for reading (and urgent data) if we are connected, can
633	* receive more, and we have room for it XXX /2 ?
634	*/
635	if (CONN_CANFRCV(so) && (so->so_snd.sb_cc < (so->so_snd.sb_datalen/2)))
636	{
637	STAM_COUNTER_INC(&pData->StatTCPHot);
638	TCP_ENGAGE_EVENT2(so, readfds, xfds);
639	}
640	}
641
642	/*
643	* UDP sockets
644	*/
645	STAM_COUNTER_RESET(&pData->StatUDP);
646	STAM_COUNTER_RESET(&pData->StatUDPHot);
647
648	for (so = udb.so_next; so != &udb; so = so_next)
649	{
650	so_next = so->so_next;
651
652	STAM_COUNTER_INC(&pData->StatUDP);
653
654	/*
655	* See if it's timed out
656	*/
657	if (so->so_expire)
658	{
659	if (so->so_expire <= curtime)
660	{
661	udp_detach(pData, so);
662	continue;
663	}
664	else
665	do_slowtimo = 1; /* Let socket expire */
666	}
667
668	/*
669	* When UDP packets are received from over the link, they're
670	* sendto()'d straight away, so no need for setting for writing
671	* Limit the number of packets queued by this session to 4.
672	* Note that even though we try and limit this to 4 packets,
673	* the session could have more queued if the packets needed
674	* to be fragmented.
675	*
676	* (XXX <= 4 ?)
677	*/
678	if ((so->so_state & SS_ISFCONNECTED) && so->so_queued <= 4)
679	{
680	STAM_COUNTER_INC(&pData->StatUDPHot);
681	UDP_ENGAGE_EVENT(so, readfds);
682	}
683	}
684
685	}
686
687	#if !defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) \|\| !defined(RT_OS_WINDOWS)
688	*pnfds = nfds;
689	#else
690	*pnfds = VBOX_EVENT_COUNT;
691	#endif
692
693	STAM_PROFILE_STOP(&pData->StatFill, a);
694	}
695
696	#if defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) && defined(RT_OS_WINDOWS)
697	void slirp_select_poll(PNATState pData, int fTimeout, int fIcmp)
698	#else
699	void slirp_select_poll(PNATState pData, fd_set readfds, fd_set writefds, fd_set *xfds)
700	#endif
701	{
702	struct socket so, so_next;
703	int ret;
704	#if defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) && defined(RT_OS_WINDOWS)
705	WSANETWORKEVENTS NetworkEvents;
706	int rc;
707	int error;
708	#endif
709
710	STAM_PROFILE_START(&pData->StatPoll, a);
711
712	/* Update time */
713	updtime(pData);
714
715	/*
716	* See if anything has timed out
717	*/
718	if (link_up)
719	{
720	if (time_fasttimo && ((curtime - time_fasttimo) >= 2))
721	{
722	STAM_PROFILE_START(&pData->StatFastTimer, a);
723	tcp_fasttimo(pData);
724	time_fasttimo = 0;
725	STAM_PROFILE_STOP(&pData->StatFastTimer, a);
726	}
727	if (do_slowtimo && ((curtime - last_slowtimo) >= 499))
728	{
729	STAM_PROFILE_START(&pData->StatSlowTimer, a);
730	ip_slowtimo(pData);
731	tcp_slowtimo(pData);
732	last_slowtimo = curtime;
733	STAM_PROFILE_STOP(&pData->StatSlowTimer, a);
734	}
735	}
736	#if defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) && defined(RT_OS_WINDOWS)
737	if (fTimeout)
738	return; /* only timer update */
739	#endif
740
741	/*
742	* Check sockets
743	*/
744	if (link_up)
745	{
746	#if defined(RT_OS_WINDOWS)
747	/*XXX: before renaming please make see define
748	* fIcmp in slirp_state.h
749	*/
750	if (fIcmp)
751	sorecvfrom(pData, &pData->icmp_socket);
752	#else
753	if (pData->icmp_socket.s != -1 && FD_ISSET(pData->icmp_socket.s, readfds))
754	sorecvfrom(pData, &pData->icmp_socket);
755	#endif
756	/*
757	* Check TCP sockets
758	*/
759	for (so = tcb.so_next; so != &tcb; so = so_next)
760	{
761	so_next = so->so_next;
762
763	/*
764	* FD_ISSET is meaningless on these sockets
765	* (and they can crash the program)
766	*/
767	if (so->so_state & SS_NOFDREF \|\| so->s == -1)
768	continue;
769
770	POLL_TCP_EVENTS(rc, error, so, &NetworkEvents);
771
772	LOG_NAT_SOCK(so, TCP, &NetworkEvents, readfds, writefds, xfds);
773
774	/*
775	* Check for URG data
776	* This will soread as well, so no need to
777	* test for readfds below if this succeeds
778	*/
779
780	/* out-of-band data */
781	if (CHECK_FD_SET(so, NetworkEvents, xfds))
782	{
783	sorecvoob(pData, so);
784	}
785
786	/*
787	* Check sockets for reading
788	*/
789	else if ( CHECK_FD_SET(so, NetworkEvents, readfds)
790	\|\| WIN_CHECK_FD_SET(so, NetworkEvents, acceptds))
791	{
792	/*
793	* Check for incoming connections
794	*/
795	if (so->so_state & SS_FACCEPTCONN)
796	{
797	tcp_connect(pData, so);
798	#if defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) && defined(RT_OS_WINDOWS)
799	if (!(NetworkEvents.lNetworkEvents & FD_CLOSE))
800	#endif
801	continue;
802	}
803
804	ret = soread(pData, so, /fCloseIfNothingRead=/false);
805	/* Output it if we read something */
806	if (ret > 0)
807	tcp_output(pData, sototcpcb(so));
808	}
809
810	#if defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) && defined(RT_OS_WINDOWS)
811	/*
812	* Check for FD_CLOSE events.
813	*/
814	if (NetworkEvents.lNetworkEvents & FD_CLOSE)
815	{
816	/*
817	* drain the socket
818	*/
819	for (;;)
820	{
821	ret = soread(pData, so, /fCloseIfNothingRead=/true);
822	if (ret > 0)
823	tcp_output(pData, sototcpcb(so));
824	else
825	break;
826	}
827	}
828	#endif
829
830	/*
831	* Check sockets for writing
832	*/
833	if (CHECK_FD_SET(so, NetworkEvents, writefds))
834	{
835	/*
836	* Check for non-blocking, still-connecting sockets
837	*/
838	if (so->so_state & SS_ISFCONNECTING)
839	{
840	Log2(("connecting %R[natsock] catched\n", so));
841	/* Connected */
842	so->so_state &= ~SS_ISFCONNECTING;
843
844	/*
845	* This should be probably guarded by PROBE_CONN too. Anyway,
846	* we disable it on OS/2 because the below send call returns
847	* EFAULT which causes the opened TCP socket to close right
848	* after it has been opened and connected.
849	*/
850	#ifndef RT_OS_OS2
851	ret = send(so->s, (const char *)&ret, 0, 0);
852	if (ret < 0)
853	{
854	/* XXXXX Must fix, zero bytes is a NOP */
855	if ( errno == EAGAIN
856	\|\| errno == EWOULDBLOCK
857	\|\| errno == EINPROGRESS
858	\|\| errno == ENOTCONN)
859	continue;
860
861	/* else failed */
862	so->so_state = SS_NOFDREF;
863	}
864	/* else so->so_state &= ~SS_ISFCONNECTING; */
865	#endif
866
867	/*
868	* Continue tcp_input
869	*/
870	tcp_input(pData, (struct mbuf *)NULL, sizeof(struct ip), so);
871	/* continue; */
872	}
873	else
874	ret = sowrite(pData, so);
875	/*
876	* XXX If we wrote something (a lot), there could be the need
877	* for a window update. In the worst case, the remote will send
878	* a window probe to get things going again.
879	*/
880	}
881
882	/*
883	* Probe a still-connecting, non-blocking socket
884	* to check if it's still alive
885	*/
886	#ifdef PROBE_CONN
887	if (so->so_state & SS_ISFCONNECTING)
888	{
889	ret = recv(so->s, (char *)&ret, 0, 0);
890
891	if (ret < 0)
892	{
893	/* XXX */
894	if ( errno == EAGAIN
895	\|\| errno == EWOULDBLOCK
896	\|\| errno == EINPROGRESS
897	\|\| errno == ENOTCONN)
898	{
899	continue; /* Still connecting, continue */
900	}
901
902	/* else failed */
903	so->so_state = SS_NOFDREF;
904
905	/* tcp_input will take care of it */
906	}
907	else
908	{
909	ret = send(so->s, &ret, 0, 0);
910	if (ret < 0)
911	{
912	/* XXX */
913	if ( errno == EAGAIN
914	\|\| errno == EWOULDBLOCK
915	\|\| errno == EINPROGRESS
916	\|\| errno == ENOTCONN)
917	{
918	continue;
919	}
920	/* else failed */
921	so->so_state = SS_NOFDREF;
922	}
923	else
924	so->so_state &= ~SS_ISFCONNECTING;
925
926	}
927	tcp_input((struct mbuf *)NULL, sizeof(struct ip),so);
928	} /* SS_ISFCONNECTING */
929	#endif
930	}
931
932	/*
933	* Now UDP sockets.
934	* Incoming packets are sent straight away, they're not buffered.
935	* Incoming UDP data isn't buffered either.
936	*/
937	for (so = udb.so_next; so != &udb; so = so_next)
938	{
939	so_next = so->so_next;
940
941	POLL_UDP_EVENTS(rc, error, so, &NetworkEvents);
942
943	LOG_NAT_SOCK(so, UDP, &NetworkEvents, readfds, writefds, xfds);
944
945	if (so->s != -1 && CHECK_FD_SET(so, NetworkEvents, readfds))
946	{
947	sorecvfrom(pData, so);
948	}
949	}
950
951	}
952
953	/*
954	* See if we can start outputting
955	*/
956	if (if_queued && link_up)
957	if_start(pData);
958
959	STAM_PROFILE_STOP(&pData->StatPoll, a);
960	}
961
962	#define ETH_ALEN 6
963	#define ETH_HLEN 14
964
965	#define ARPOP_REQUEST 1 /* ARP request */
966	#define ARPOP_REPLY 2 /* ARP reply */
967
968	struct ethhdr
969	{
970	unsigned char h_dest[ETH_ALEN]; /* destination eth addr */
971	unsigned char h_source[ETH_ALEN]; /* source ether addr */
972	unsigned short h_proto; /* packet type ID field */
973	};
974
975	struct arphdr
976	{
977	unsigned short ar_hrd; /* format of hardware address */
978	unsigned short ar_pro; /* format of protocol address */
979	unsigned char ar_hln; /* length of hardware address */
980	unsigned char ar_pln; /* length of protocol address */
981	unsigned short ar_op; /* ARP opcode (command) */
982
983	/*
984	* Ethernet looks like this : This bit is variable sized however...
985	*/
986	unsigned char ar_sha[ETH_ALEN]; /* sender hardware address */
987	unsigned char ar_sip[4]; /* sender IP address */
988	unsigned char ar_tha[ETH_ALEN]; /* target hardware address */
989	unsigned char ar_tip[4]; /* target IP address */
990	};
991
992	static
993	#ifdef VBOX_WITH_SIMPLIFIED_SLIRP_SYNC
994	void arp_input(PNATState pData, struct mbuf *m)
995	#else
996	void arp_input(PNATState pData, const uint8_t *pkt, int pkt_len)
997	#endif
998	{
999	struct ethhdr *eh;
1000	struct ethhdr *reh;
1001	struct arphdr *ah;
1002	struct arphdr *rah;
1003	int ar_op;
1004	struct ex_list *ex_ptr;
1005	uint32_t htip;
1006	#ifndef VBOX_WITH_SIMPLIFIED_SLIRP_SYNC
1007	uint8_t arp_reply[sizeof(struct arphdr) + ETH_HLEN];
1008	eh = (struct ethhdr *)pkt;
1009	#else
1010	struct mbuf *mr;
1011	eh = mtod(m, struct ethhdr *);
1012	#endif
1013	ah = (struct arphdr *)&eh[1];
1014	htip = ntohl((uint32_t)ah->ar_tip);
1015
1016	#ifdef VBOX_WITH_SIMPLIFIED_SLIRP_SYNC
1017	mr = m_get(pData);
1018	mr->m_data += if_maxlinkhdr;
1019	mr->m_len = sizeof(struct arphdr);
1020	rah = mtod(mr, struct arphdr *);
1021	#else
1022	reh = (struct ethhdr *)arp_reply;
1023	rah = (struct arphdr *)&reh[1];
1024	#endif
1025
1026	ar_op = ntohs(ah->ar_op);
1027	switch(ar_op)
1028	{
1029	case ARPOP_REQUEST:
1030	if ((htip & pData->netmask) == ntohl(special_addr.s_addr))
1031	{
1032	if ( CTL_CHECK(htip,CTL_DNS)
1033	\|\| CTL_CHECK(htip, CTL_ALIAS)
1034	\|\| CTL_CHECK(htip, CTL_TFTP))
1035	goto arp_ok;
1036	for (ex_ptr = exec_list; ex_ptr; ex_ptr = ex_ptr->ex_next)
1037	{
1038	if ((htip & ~pData->netmask) == ex_ptr->ex_addr)
1039	goto arp_ok;
1040	}
1041	return;
1042	arp_ok:
1043
1044	#ifndef VBOX_WITH_SIMPLIFIED_SLIRP_SYNC
1045	memcpy(reh->h_dest, eh->h_source, ETH_ALEN);
1046	memcpy(reh->h_source, &special_addr, ETH_ALEN);
1047	reh->h_source[5] = ah->ar_tip[3];
1048	reh->h_proto = htons(ETH_P_ARP);
1049	#endif
1050	rah->ar_hrd = htons(1);
1051	rah->ar_pro = htons(ETH_P_IP);
1052	rah->ar_hln = ETH_ALEN;
1053	rah->ar_pln = 4;
1054	rah->ar_op = htons(ARPOP_REPLY);
1055	memcpy(rah->ar_sha, special_ethaddr, ETH_ALEN);
1056
1057	switch (htip & ~pData->netmask)
1058	{
1059	case CTL_DNS:
1060	case CTL_ALIAS:
1061	rah->ar_sha[5] = (uint8_t)(htip & ~pData->netmask);
1062	break;
1063	default:;
1064	}
1065
1066	memcpy(rah->ar_sip, ah->ar_tip, 4);
1067	memcpy(rah->ar_tha, ah->ar_sha, ETH_ALEN);
1068	memcpy(rah->ar_tip, ah->ar_sip, 4);
1069	#ifdef VBOX_WITH_SIMPLIFIED_SLIRP_SYNC
1070	if_encap(pData, ETH_P_ARP, mr);
1071	m_free(pData, m);
1072	#else
1073	slirp_output(pData->pvUser, arp_reply, sizeof(arp_reply));
1074	#endif
1075	}
1076	break;
1077	default:
1078	break;
1079	}
1080	}
1081
1082	void slirp_input(PNATState pData, const uint8_t *pkt, int pkt_len)
1083	{
1084	struct mbuf *m;
1085	int proto;
1086	static bool fWarnedIpv6;
1087
1088	if (pkt_len < ETH_HLEN)
1089	{
1090	LogRel(("NAT: packet having size %d has been ingnored\n", pkt_len));
1091	return;
1092	}
1093
1094	m = m_get(pData);
1095	if (!m)
1096	{
1097	LogRel(("can't allocate new mbuf\n"));
1098	return;
1099	}
1100
1101	/* Note: we add to align the IP header */
1102
1103	if (M_FREEROOM(m) < pkt_len)
1104	m_inc(m, pkt_len);
1105
1106	m->m_len = pkt_len ;
1107	memcpy(m->m_data, pkt, pkt_len);
1108
1109	proto = ntohs((uint16_t )(pkt + 12));
1110	switch(proto)
1111	{
1112	case ETH_P_ARP:
1113	#ifdef VBOX_WITH_SIMPLIFIED_SLIRP_SYNC
1114	arp_input(pData, m);
1115	#else
1116	arp_input(pData, pkt, pkt_len);
1117	m_free(pData, m);
1118	#endif
1119	break;
1120	case ETH_P_IP:
1121	/* Update time. Important if the network is very quiet, as otherwise
1122	* the first outgoing connection gets an incorrect timestamp. */
1123	updtime(pData);
1124	m->m_data += ETH_HLEN;
1125	m->m_len -= ETH_HLEN;
1126	ip_input(pData, m);
1127	break;
1128	case ETH_P_IPV6:
1129	m_free(pData, m);
1130	if (!fWarnedIpv6)
1131	{
1132	LogRel(("NAT: IPv6 not supported\n"));
1133	fWarnedIpv6 = true;
1134	}
1135	break;
1136	default:
1137	LogRel(("NAT: Unsupported protocol %x\n", proto));
1138	m_free(pData, m);
1139	break;
1140	}
1141	}
1142
1143	/* output the IP packet to the ethernet device */
1144	#ifdef VBOX_WITH_SIMPLIFIED_SLIRP_SYNC
1145	void if_encap(PNATState pData, uint16_t eth_proto, struct mbuf *m)
1146	#else
1147	void if_encap(PNATState pData, uint8_t *ip_data, int ip_data_len)
1148	#endif
1149	{
1150	#ifdef VBOX_WITH_SIMPLIFIED_SLIRP_SYNC
1151	struct ethhdr *eh;
1152	uint8_t *buf = RTMemAlloc(1600);
1153	m->m_data -= if_maxlinkhdr;
1154	m->m_len += ETH_HLEN;
1155	eh = mtod(m, struct ethhdr *);
1156	#else
1157	uint8_t buf[1600];
1158	struct ethhdr eh = (struct ethhdr )buf;
1159
1160	if (ip_data_len + ETH_HLEN > sizeof(buf))
1161	return;
1162
1163	memcpy(buf + sizeof(struct ethhdr), ip_data, ip_data_len);
1164	#endif
1165
1166
1167	memcpy(eh->h_dest, client_ethaddr, ETH_ALEN);
1168	memcpy(eh->h_source, special_ethaddr, ETH_ALEN - 1);
1169	/* XXX: not correct */
1170	eh->h_source[5] = CTL_ALIAS;
1171	#ifdef VBOX_WITH_SIMPLIFIED_SLIRP_SYNC
1172	eh->h_proto = htons(eth_proto);
1173	#if 0
1174	slirp_output(pData->pvUser, m, mtod(m, uint8_t *), m->m_len);
1175	#else
1176	memcpy(buf, mtod(m, uint8_t *), m->m_len);
1177	slirp_output(pData->pvUser, NULL, buf, m->m_len);
1178	m_free(pData, m);
1179	#endif
1180	#else
1181	eh->h_proto = htons(ETH_P_IP);
1182	slirp_output(pData->pvUser, buf, ip_data_len + ETH_HLEN);
1183	#endif
1184	}
1185
1186	int slirp_redir(PNATState pData, int is_udp, int host_port,
1187	struct in_addr guest_addr, int guest_port)
1188	{
1189	if (is_udp)
1190	{
1191	if (!udp_listen(pData, htons(host_port), guest_addr.s_addr,
1192	htons(guest_port), 0))
1193	return -1;
1194	}
1195	else
1196	{
1197	if (!solisten(pData, htons(host_port), guest_addr.s_addr,
1198	htons(guest_port), 0))
1199	return -1;
1200	}
1201	return 0;
1202	}
1203
1204	int slirp_add_exec(PNATState pData, int do_pty, const char *args, int addr_low_byte,
1205	int guest_port)
1206	{
1207	return add_exec(&exec_list, do_pty, (char *)args,
1208	addr_low_byte, htons(guest_port));
1209	}
1210
1211	void slirp_set_ethaddr(PNATState pData, const uint8_t *ethaddr)
1212	{
1213	memcpy(client_ethaddr, ethaddr, ETH_ALEN);
1214	}
1215
1216	#if defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) && defined(RT_OS_WINDOWS)
1217	HANDLE *slirp_get_events(PNATState pData)
1218	{
1219	return pData->phEvents;
1220	}
1221	void slirp_register_external_event(PNATState pData, HANDLE hEvent, int index)
1222	{
1223	pData->phEvents[index] = hEvent;
1224	}
1225	#endif
1226
1227	unsigned int slirp_get_timeout_ms(PNATState pData)
1228	{
1229	if (link_up)
1230	{
1231	if (time_fasttimo)
1232	return 2;
1233	if (do_slowtimo)
1234	return 500; /* see PR_SLOWHZ */
1235	}
1236	return 0;
1237	}
1238
1239	/*
1240	* this function called from NAT thread
1241	*/
1242	void slirp_post_sent(PNATState pData, void *pvArg)
1243	{
1244	struct socket *so = 0;
1245	struct tcpcb *tp = 0;
1246	struct mbuf m = (struct mbuf )pvArg;
1247	m_free(pData, m);
1248	}

注意: 瀏覽 TracBrowser 來幫助您使用儲存庫瀏覽器

source: vbox/trunk/src/VBox/Devices/Network/slirp/slirp.c@ 16226

以其他格式下載: