slirp.c@ 16213

Last change on this file since 16213 was 16213, checked in by vboxsync, 16 years ago

NAT:

introduces TFTP special_address | CTL_TFTP(4) for built-in address
can connect to other then built-in tftp

Property svn:eol-style set to native

File size: 36.7 KB

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

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

source: vbox/trunk/src/VBox/Devices/Network/slirp/slirp.c@ 16213

Download in other formats: