slirp.c@ 16254

Last change on this file since 16254 was 16254, checked in by vboxsync, 16 years ago
NAT: clean up dhcp a bit, and fox the TFTP server address calculation
Property svn:eol-style set to `native`
File size: 37.8 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	#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	/* @todo: add ability to configure this staff */
400	pData->tftp_server.s_addr = htonl(ntohl(special_addr.s_addr) \| CTL_TFTP);
401
402	/* set default addresses */
403	inet_aton("127.0.0.1", &loopback_addr);
404	#ifndef VBOX_WITH_MULTI_DNS
405	inet_aton("127.0.0.1", &dns_addr);
406
407	if (get_dns_addr_domain(pData, true, &dns_addr, &pData->pszDomain) < 0)
408	#else
409	if (get_dns_addr_domain(pData, true, NULL, &pData->pszDomain) < 0)
410	#endif
411	fNATfailed = 1;
412
413	getouraddr(pData);
414	return fNATfailed ? VINF_NAT_DNS : VINF_SUCCESS;
415	}
416
417	/**
418	* Statistics counters.
419	*/
420	void slirp_register_timers(PNATState pData, PPDMDRVINS pDrvIns)
421	{
422	#ifdef VBOX_WITH_STATISTICS
423	PDMDrvHlpSTAMRegisterF(pDrvIns, &pData->StatFill, STAMTYPE_PROFILE, STAMVISIBILITY_ALWAYS,
424	STAMUNIT_TICKS_PER_CALL, "Profiling slirp fills", "/Drivers/NAT%d/Fill", pDrvIns->iInstance);
425	PDMDrvHlpSTAMRegisterF(pDrvIns, &pData->StatPoll, STAMTYPE_PROFILE, STAMVISIBILITY_ALWAYS,
426	STAMUNIT_TICKS_PER_CALL, "Profiling slirp polls", "/Drivers/NAT%d/Poll", pDrvIns->iInstance);
427	PDMDrvHlpSTAMRegisterF(pDrvIns, &pData->StatFastTimer, STAMTYPE_PROFILE, STAMVISIBILITY_ALWAYS,
428	STAMUNIT_TICKS_PER_CALL, "Profiling slirp fast timer", "/Drivers/NAT%d/TimerFast", pDrvIns->iInstance);
429	PDMDrvHlpSTAMRegisterF(pDrvIns, &pData->StatSlowTimer, STAMTYPE_PROFILE, STAMVISIBILITY_ALWAYS,
430	STAMUNIT_TICKS_PER_CALL, "Profiling slirp slow timer", "/Drivers/NAT%d/TimerSlow", pDrvIns->iInstance);
431	PDMDrvHlpSTAMRegisterF(pDrvIns, &pData->StatTCP, STAMTYPE_COUNTER, STAMVISIBILITY_ALWAYS,
432	STAMUNIT_COUNT, "TCP sockets", "/Drivers/NAT%d/SockTCP", pDrvIns->iInstance);
433	PDMDrvHlpSTAMRegisterF(pDrvIns, &pData->StatTCPHot, STAMTYPE_COUNTER, STAMVISIBILITY_ALWAYS,
434	STAMUNIT_COUNT, "TCP sockets active", "/Drivers/NAT%d/SockTCPHot", pDrvIns->iInstance);
435	PDMDrvHlpSTAMRegisterF(pDrvIns, &pData->StatUDP, STAMTYPE_COUNTER, STAMVISIBILITY_ALWAYS,
436	STAMUNIT_COUNT, "UDP sockets", "/Drivers/NAT%d/SockUDP", pDrvIns->iInstance);
437	PDMDrvHlpSTAMRegisterF(pDrvIns, &pData->StatUDPHot, STAMTYPE_COUNTER, STAMVISIBILITY_ALWAYS,
438	STAMUNIT_COUNT, "UDP sockets active", "/Drivers/NAT%d/SockUDPHot", pDrvIns->iInstance);
439	#endif /* VBOX_WITH_STATISTICS */
440	}
441
442	/**
443	* Marks the link as up, making it possible to establish new connections.
444	*/
445	void slirp_link_up(PNATState pData)
446	{
447	link_up = 1;
448	}
449
450	/**
451	* Marks the link as down and cleans up the current connections.
452	*/
453	void slirp_link_down(PNATState pData)
454	{
455	struct socket *so;
456
457	while ((so = tcb.so_next) != &tcb)
458	{
459	if (so->so_state & SS_NOFDREF \|\| so->s == -1)
460	sofree(pData, so);
461	else
462	tcp_drop(pData, sototcpcb(so), 0);
463	}
464
465	while ((so = udb.so_next) != &udb)
466	udp_detach(pData, so);
467
468	link_up = 0;
469	}
470
471	/**
472	* Terminates the slirp component.
473	*/
474	void slirp_term(PNATState pData)
475	{
476	if (pData->pszDomain)
477	RTStrFree((char )(void )pData->pszDomain);
478
479	#ifdef RT_OS_WINDOWS
480	pData->pfIcmpCloseHandle(pData->icmp_socket.sh);
481	FreeLibrary(pData->hmIcmpLibrary);
482	RTMemFree(pData->pvIcmpBuffer);
483	# else
484	closesocket(pData->icmp_socket.s);
485	#endif
486
487	slirp_link_down(pData);
488	#ifdef RT_OS_WINDOWS
489	WSACleanup();
490	#endif
491	#ifdef LOG_ENABLED
492	Log(("\n"
493	"NAT statistics\n"
494	"--------------\n"
495	"\n"));
496	ipstats(pData);
497	tcpstats(pData);
498	udpstats(pData);
499	icmpstats(pData);
500	mbufstats(pData);
501	sockstats(pData);
502	Log(("\n"
503	"\n"
504	"\n"));
505	#endif
506	RTMemFree(pData);
507	}
508
509
510	#define CONN_CANFSEND(so) (((so)->so_state & (SS_FCANTSENDMORE\|SS_ISFCONNECTED)) == SS_ISFCONNECTED)
511	#define CONN_CANFRCV(so) (((so)->so_state & (SS_FCANTRCVMORE\|SS_ISFCONNECTED)) == SS_ISFCONNECTED)
512	#define UPD_NFDS(x) if (nfds < (x)) nfds = (x)
513
514	/*
515	* curtime kept to an accuracy of 1ms
516	*/
517	#ifdef RT_OS_WINDOWS
518	static void updtime(PNATState pData)
519	{
520	struct _timeb tb;
521
522	_ftime(&tb);
523	curtime = (u_int)tb.time * (u_int)1000;
524	curtime += (u_int)tb.millitm;
525	}
526	#else
527	static void updtime(PNATState pData)
528	{
529	gettimeofday(&tt, 0);
530
531	curtime = (u_int)tt.tv_sec * (u_int)1000;
532	curtime += (u_int)tt.tv_usec / (u_int)1000;
533
534	if ((tt.tv_usec % 1000) >= 500)
535	curtime++;
536	}
537	#endif
538
539	void slirp_select_fill(PNATState pData, int *pnfds,
540	fd_set readfds, fd_set writefds, fd_set *xfds)
541	{
542	struct socket so, so_next;
543	int nfds;
544	#if defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) && defined(RT_OS_WINDOWS)
545	int rc;
546	int error;
547	#endif
548	int i;
549
550	STAM_PROFILE_START(&pData->StatFill, a);
551
552	nfds = *pnfds;
553
554	/*
555	* First, TCP sockets
556	*/
557	do_slowtimo = 0;
558	if (link_up)
559	{
560	/*
561	* *_slowtimo needs calling if there are IP fragments
562	* in the fragment queue, or there are TCP connections active
563	*/
564	/* XXX:
565	* triggering of fragment expiration should be the same but use new macroses
566	*/
567	do_slowtimo = (tcb.so_next != &tcb);
568	if (!do_slowtimo)
569	{
570	for (i = 0; i < IPREASS_NHASH; i++)
571	{
572	if (!TAILQ_EMPTY(&ipq[i]))
573	{
574	do_slowtimo = 1;
575	break;
576	}
577	}
578	}
579	ICMP_ENGAGE_EVENT(&pData->icmp_socket, readfds);
580
581	STAM_COUNTER_RESET(&pData->StatTCP);
582	STAM_COUNTER_RESET(&pData->StatTCPHot);
583
584	for (so = tcb.so_next; so != &tcb; so = so_next)
585	{
586	so_next = so->so_next;
587
588	STAM_COUNTER_INC(&pData->StatTCP);
589
590	/*
591	* See if we need a tcp_fasttimo
592	*/
593	if (time_fasttimo == 0 && so->so_tcpcb->t_flags & TF_DELACK)
594	time_fasttimo = curtime; /* Flag when we want a fasttimo */
595
596	/*
597	* NOFDREF can include still connecting to local-host,
598	* newly socreated() sockets etc. Don't want to select these.
599	*/
600	if (so->so_state & SS_NOFDREF \|\| so->s == -1)
601	continue;
602
603	/*
604	* Set for reading sockets which are accepting
605	*/
606	if (so->so_state & SS_FACCEPTCONN)
607	{
608	STAM_COUNTER_INC(&pData->StatTCPHot);
609	TCP_ENGAGE_EVENT1(so, readfds);
610	continue;
611	}
612
613	/*
614	* Set for writing sockets which are connecting
615	*/
616	if (so->so_state & SS_ISFCONNECTING)
617	{
618	Log2(("connecting %R[natsock] engaged\n",so));
619	STAM_COUNTER_INC(&pData->StatTCPHot);
620	TCP_ENGAGE_EVENT1(so, writefds);
621	}
622
623	/*
624	* Set for writing if we are connected, can send more, and
625	* we have something to send
626	*/
627	if (CONN_CANFSEND(so) && so->so_rcv.sb_cc)
628	{
629	STAM_COUNTER_INC(&pData->StatTCPHot);
630	TCP_ENGAGE_EVENT1(so, writefds);
631	}
632
633	/*
634	* Set for reading (and urgent data) if we are connected, can
635	* receive more, and we have room for it XXX /2 ?
636	*/
637	if (CONN_CANFRCV(so) && (so->so_snd.sb_cc < (so->so_snd.sb_datalen/2)))
638	{
639	STAM_COUNTER_INC(&pData->StatTCPHot);
640	TCP_ENGAGE_EVENT2(so, readfds, xfds);
641	}
642	}
643
644	/*
645	* UDP sockets
646	*/
647	STAM_COUNTER_RESET(&pData->StatUDP);
648	STAM_COUNTER_RESET(&pData->StatUDPHot);
649
650	for (so = udb.so_next; so != &udb; so = so_next)
651	{
652	so_next = so->so_next;
653
654	STAM_COUNTER_INC(&pData->StatUDP);
655
656	/*
657	* See if it's timed out
658	*/
659	if (so->so_expire)
660	{
661	if (so->so_expire <= curtime)
662	{
663	udp_detach(pData, so);
664	continue;
665	}
666	else
667	do_slowtimo = 1; /* Let socket expire */
668	}
669
670	/*
671	* When UDP packets are received from over the link, they're
672	* sendto()'d straight away, so no need for setting for writing
673	* Limit the number of packets queued by this session to 4.
674	* Note that even though we try and limit this to 4 packets,
675	* the session could have more queued if the packets needed
676	* to be fragmented.
677	*
678	* (XXX <= 4 ?)
679	*/
680	if ((so->so_state & SS_ISFCONNECTED) && so->so_queued <= 4)
681	{
682	STAM_COUNTER_INC(&pData->StatUDPHot);
683	UDP_ENGAGE_EVENT(so, readfds);
684	}
685	}
686
687	}
688
689	#if !defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) \|\| !defined(RT_OS_WINDOWS)
690	*pnfds = nfds;
691	#else
692	*pnfds = VBOX_EVENT_COUNT;
693	#endif
694
695	STAM_PROFILE_STOP(&pData->StatFill, a);
696	}
697
698	#if defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) && defined(RT_OS_WINDOWS)
699	void slirp_select_poll(PNATState pData, int fTimeout, int fIcmp)
700	#else
701	void slirp_select_poll(PNATState pData, fd_set readfds, fd_set writefds, fd_set *xfds)
702	#endif
703	{
704	struct socket so, so_next;
705	int ret;
706	#if defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) && defined(RT_OS_WINDOWS)
707	WSANETWORKEVENTS NetworkEvents;
708	int rc;
709	int error;
710	#endif
711
712	STAM_PROFILE_START(&pData->StatPoll, a);
713
714	/* Update time */
715	updtime(pData);
716
717	/*
718	* See if anything has timed out
719	*/
720	if (link_up)
721	{
722	if (time_fasttimo && ((curtime - time_fasttimo) >= 2))
723	{
724	STAM_PROFILE_START(&pData->StatFastTimer, a);
725	tcp_fasttimo(pData);
726	time_fasttimo = 0;
727	STAM_PROFILE_STOP(&pData->StatFastTimer, a);
728	}
729	if (do_slowtimo && ((curtime - last_slowtimo) >= 499))
730	{
731	STAM_PROFILE_START(&pData->StatSlowTimer, a);
732	ip_slowtimo(pData);
733	tcp_slowtimo(pData);
734	last_slowtimo = curtime;
735	STAM_PROFILE_STOP(&pData->StatSlowTimer, a);
736	}
737	}
738	#if defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) && defined(RT_OS_WINDOWS)
739	if (fTimeout)
740	return; /* only timer update */
741	#endif
742
743	/*
744	* Check sockets
745	*/
746	if (link_up)
747	{
748	#if defined(RT_OS_WINDOWS)
749	/*XXX: before renaming please make see define
750	* fIcmp in slirp_state.h
751	*/
752	if (fIcmp)
753	sorecvfrom(pData, &pData->icmp_socket);
754	#else
755	if (pData->icmp_socket.s != -1 && FD_ISSET(pData->icmp_socket.s, readfds))
756	sorecvfrom(pData, &pData->icmp_socket);
757	#endif
758	/*
759	* Check TCP sockets
760	*/
761	for (so = tcb.so_next; so != &tcb; so = so_next)
762	{
763	so_next = so->so_next;
764
765	/*
766	* FD_ISSET is meaningless on these sockets
767	* (and they can crash the program)
768	*/
769	if (so->so_state & SS_NOFDREF \|\| so->s == -1)
770	continue;
771
772	POLL_TCP_EVENTS(rc, error, so, &NetworkEvents);
773
774	LOG_NAT_SOCK(so, TCP, &NetworkEvents, readfds, writefds, xfds);
775
776	/*
777	* Check for URG data
778	* This will soread as well, so no need to
779	* test for readfds below if this succeeds
780	*/
781
782	/* out-of-band data */
783	if (CHECK_FD_SET(so, NetworkEvents, xfds))
784	{
785	sorecvoob(pData, so);
786	}
787
788	/*
789	* Check sockets for reading
790	*/
791	else if ( CHECK_FD_SET(so, NetworkEvents, readfds)
792	\|\| WIN_CHECK_FD_SET(so, NetworkEvents, acceptds))
793	{
794	/*
795	* Check for incoming connections
796	*/
797	if (so->so_state & SS_FACCEPTCONN)
798	{
799	tcp_connect(pData, so);
800	#if defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) && defined(RT_OS_WINDOWS)
801	if (!(NetworkEvents.lNetworkEvents & FD_CLOSE))
802	#endif
803	continue;
804	}
805
806	ret = soread(pData, so, /fCloseIfNothingRead=/false);
807	/* Output it if we read something */
808	if (ret > 0)
809	tcp_output(pData, sototcpcb(so));
810	}
811
812	#if defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) && defined(RT_OS_WINDOWS)
813	/*
814	* Check for FD_CLOSE events.
815	*/
816	if (NetworkEvents.lNetworkEvents & FD_CLOSE)
817	{
818	/*
819	* drain the socket
820	*/
821	for (;;)
822	{
823	ret = soread(pData, so, /fCloseIfNothingRead=/true);
824	if (ret > 0)
825	tcp_output(pData, sototcpcb(so));
826	else
827	break;
828	}
829	}
830	#endif
831
832	/*
833	* Check sockets for writing
834	*/
835	if (CHECK_FD_SET(so, NetworkEvents, writefds))
836	{
837	/*
838	* Check for non-blocking, still-connecting sockets
839	*/
840	if (so->so_state & SS_ISFCONNECTING)
841	{
842	Log2(("connecting %R[natsock] catched\n", so));
843	/* Connected */
844	so->so_state &= ~SS_ISFCONNECTING;
845
846	/*
847	* This should be probably guarded by PROBE_CONN too. Anyway,
848	* we disable it on OS/2 because the below send call returns
849	* EFAULT which causes the opened TCP socket to close right
850	* after it has been opened and connected.
851	*/
852	#ifndef RT_OS_OS2
853	ret = send(so->s, (const char *)&ret, 0, 0);
854	if (ret < 0)
855	{
856	/* XXXXX Must fix, zero bytes is a NOP */
857	if ( errno == EAGAIN
858	\|\| errno == EWOULDBLOCK
859	\|\| errno == EINPROGRESS
860	\|\| errno == ENOTCONN)
861	continue;
862
863	/* else failed */
864	so->so_state = SS_NOFDREF;
865	}
866	/* else so->so_state &= ~SS_ISFCONNECTING; */
867	#endif
868
869	/*
870	* Continue tcp_input
871	*/
872	tcp_input(pData, (struct mbuf *)NULL, sizeof(struct ip), so);
873	/* continue; */
874	}
875	else
876	ret = sowrite(pData, so);
877	/*
878	* XXX If we wrote something (a lot), there could be the need
879	* for a window update. In the worst case, the remote will send
880	* a window probe to get things going again.
881	*/
882	}
883
884	/*
885	* Probe a still-connecting, non-blocking socket
886	* to check if it's still alive
887	*/
888	#ifdef PROBE_CONN
889	if (so->so_state & SS_ISFCONNECTING)
890	{
891	ret = recv(so->s, (char *)&ret, 0, 0);
892
893	if (ret < 0)
894	{
895	/* XXX */
896	if ( errno == EAGAIN
897	\|\| errno == EWOULDBLOCK
898	\|\| errno == EINPROGRESS
899	\|\| errno == ENOTCONN)
900	{
901	continue; /* Still connecting, continue */
902	}
903
904	/* else failed */
905	so->so_state = SS_NOFDREF;
906
907	/* tcp_input will take care of it */
908	}
909	else
910	{
911	ret = send(so->s, &ret, 0, 0);
912	if (ret < 0)
913	{
914	/* XXX */
915	if ( errno == EAGAIN
916	\|\| errno == EWOULDBLOCK
917	\|\| errno == EINPROGRESS
918	\|\| errno == ENOTCONN)
919	{
920	continue;
921	}
922	/* else failed */
923	so->so_state = SS_NOFDREF;
924	}
925	else
926	so->so_state &= ~SS_ISFCONNECTING;
927
928	}
929	tcp_input((struct mbuf *)NULL, sizeof(struct ip),so);
930	} /* SS_ISFCONNECTING */
931	#endif
932	}
933
934	/*
935	* Now UDP sockets.
936	* Incoming packets are sent straight away, they're not buffered.
937	* Incoming UDP data isn't buffered either.
938	*/
939	for (so = udb.so_next; so != &udb; so = so_next)
940	{
941	so_next = so->so_next;
942
943	POLL_UDP_EVENTS(rc, error, so, &NetworkEvents);
944
945	LOG_NAT_SOCK(so, UDP, &NetworkEvents, readfds, writefds, xfds);
946
947	if (so->s != -1 && CHECK_FD_SET(so, NetworkEvents, readfds))
948	{
949	sorecvfrom(pData, so);
950	}
951	}
952
953	}
954
955	/*
956	* See if we can start outputting
957	*/
958	if (if_queued && link_up)
959	if_start(pData);
960
961	STAM_PROFILE_STOP(&pData->StatPoll, a);
962	}
963
964	#define ETH_ALEN 6
965	#define ETH_HLEN 14
966
967	#define ARPOP_REQUEST 1 /* ARP request */
968	#define ARPOP_REPLY 2 /* ARP reply */
969
970	struct ethhdr
971	{
972	unsigned char h_dest[ETH_ALEN]; /* destination eth addr */
973	unsigned char h_source[ETH_ALEN]; /* source ether addr */
974	unsigned short h_proto; /* packet type ID field */
975	};
976
977	struct arphdr
978	{
979	unsigned short ar_hrd; /* format of hardware address */
980	unsigned short ar_pro; /* format of protocol address */
981	unsigned char ar_hln; /* length of hardware address */
982	unsigned char ar_pln; /* length of protocol address */
983	unsigned short ar_op; /* ARP opcode (command) */
984
985	/*
986	* Ethernet looks like this : This bit is variable sized however...
987	*/
988	unsigned char ar_sha[ETH_ALEN]; /* sender hardware address */
989	unsigned char ar_sip[4]; /* sender IP address */
990	unsigned char ar_tha[ETH_ALEN]; /* target hardware address */
991	unsigned char ar_tip[4]; /* target IP address */
992	};
993
994	static
995	#ifdef VBOX_WITH_SIMPLIFIED_SLIRP_SYNC
996	void arp_input(PNATState pData, struct mbuf *m)
997	#else
998	void arp_input(PNATState pData, const uint8_t *pkt, int pkt_len)
999	#endif
1000	{
1001	struct ethhdr *eh;
1002	struct ethhdr *reh;
1003	struct arphdr *ah;
1004	struct arphdr *rah;
1005	int ar_op;
1006	struct ex_list *ex_ptr;
1007	uint32_t htip;
1008	#ifndef VBOX_WITH_SIMPLIFIED_SLIRP_SYNC
1009	uint8_t arp_reply[sizeof(struct arphdr) + ETH_HLEN];
1010	eh = (struct ethhdr *)pkt;
1011	#else
1012	struct mbuf *mr;
1013	eh = mtod(m, struct ethhdr *);
1014	#endif
1015	ah = (struct arphdr *)&eh[1];
1016	htip = ntohl((uint32_t)ah->ar_tip);
1017
1018	#ifdef VBOX_WITH_SIMPLIFIED_SLIRP_SYNC
1019	mr = m_get(pData);
1020	mr->m_data += if_maxlinkhdr;
1021	mr->m_len = sizeof(struct arphdr);
1022	rah = mtod(mr, struct arphdr *);
1023	#else
1024	reh = (struct ethhdr *)arp_reply;
1025	rah = (struct arphdr *)&reh[1];
1026	#endif
1027
1028	ar_op = ntohs(ah->ar_op);
1029	switch(ar_op)
1030	{
1031	case ARPOP_REQUEST:
1032	if ((htip & pData->netmask) == ntohl(special_addr.s_addr))
1033	{
1034	if ( CTL_CHECK(htip,CTL_DNS)
1035	\|\| CTL_CHECK(htip, CTL_ALIAS)
1036	\|\| CTL_CHECK(htip, CTL_TFTP))
1037	goto arp_ok;
1038	for (ex_ptr = exec_list; ex_ptr; ex_ptr = ex_ptr->ex_next)
1039	{
1040	if ((htip & ~pData->netmask) == ex_ptr->ex_addr)
1041	goto arp_ok;
1042	}
1043	return;
1044	arp_ok:
1045
1046	#ifndef VBOX_WITH_SIMPLIFIED_SLIRP_SYNC
1047	memcpy(reh->h_dest, eh->h_source, ETH_ALEN);
1048	memcpy(reh->h_source, &special_addr, ETH_ALEN);
1049	reh->h_source[5] = ah->ar_tip[3];
1050	reh->h_proto = htons(ETH_P_ARP);
1051	#endif
1052	rah->ar_hrd = htons(1);
1053	rah->ar_pro = htons(ETH_P_IP);
1054	rah->ar_hln = ETH_ALEN;
1055	rah->ar_pln = 4;
1056	rah->ar_op = htons(ARPOP_REPLY);
1057	memcpy(rah->ar_sha, special_ethaddr, ETH_ALEN);
1058
1059	switch (htip & ~pData->netmask)
1060	{
1061	case CTL_DNS:
1062	case CTL_ALIAS:
1063	rah->ar_sha[5] = (uint8_t)(htip & ~pData->netmask);
1064	break;
1065	default:;
1066	}
1067
1068	memcpy(rah->ar_sip, ah->ar_tip, 4);
1069	memcpy(rah->ar_tha, ah->ar_sha, ETH_ALEN);
1070	memcpy(rah->ar_tip, ah->ar_sip, 4);
1071	#ifdef VBOX_WITH_SIMPLIFIED_SLIRP_SYNC
1072	if_encap(pData, ETH_P_ARP, mr);
1073	m_free(pData, m);
1074	#else
1075	slirp_output(pData->pvUser, arp_reply, sizeof(arp_reply));
1076	#endif
1077	}
1078	break;
1079	default:
1080	break;
1081	}
1082	}
1083
1084	void slirp_input(PNATState pData, const uint8_t *pkt, int pkt_len)
1085	{
1086	struct mbuf *m;
1087	int proto;
1088	static bool fWarnedIpv6;
1089
1090	if (pkt_len < ETH_HLEN)
1091	{
1092	LogRel(("NAT: packet having size %d has been ingnored\n", pkt_len));
1093	return;
1094	}
1095
1096	m = m_get(pData);
1097	if (!m)
1098	{
1099	LogRel(("can't allocate new mbuf\n"));
1100	return;
1101	}
1102
1103	/* Note: we add to align the IP header */
1104
1105	if (M_FREEROOM(m) < pkt_len)
1106	m_inc(m, pkt_len);
1107
1108	m->m_len = pkt_len ;
1109	memcpy(m->m_data, pkt, pkt_len);
1110
1111	proto = ntohs((uint16_t )(pkt + 12));
1112	switch(proto)
1113	{
1114	case ETH_P_ARP:
1115	#ifdef VBOX_WITH_SIMPLIFIED_SLIRP_SYNC
1116	arp_input(pData, m);
1117	#else
1118	arp_input(pData, pkt, pkt_len);
1119	m_free(pData, m);
1120	#endif
1121	break;
1122	case ETH_P_IP:
1123	/* Update time. Important if the network is very quiet, as otherwise
1124	* the first outgoing connection gets an incorrect timestamp. */
1125	updtime(pData);
1126	m->m_data += ETH_HLEN;
1127	m->m_len -= ETH_HLEN;
1128	ip_input(pData, m);
1129	break;
1130	case ETH_P_IPV6:
1131	m_free(pData, m);
1132	if (!fWarnedIpv6)
1133	{
1134	LogRel(("NAT: IPv6 not supported\n"));
1135	fWarnedIpv6 = true;
1136	}
1137	break;
1138	default:
1139	LogRel(("NAT: Unsupported protocol %x\n", proto));
1140	m_free(pData, m);
1141	break;
1142	}
1143	}
1144
1145	/* output the IP packet to the ethernet device */
1146	#ifdef VBOX_WITH_SIMPLIFIED_SLIRP_SYNC
1147	void if_encap(PNATState pData, uint16_t eth_proto, struct mbuf *m)
1148	#else
1149	void if_encap(PNATState pData, uint8_t *ip_data, int ip_data_len)
1150	#endif
1151	{
1152	#ifdef VBOX_WITH_SIMPLIFIED_SLIRP_SYNC
1153	struct ethhdr *eh;
1154	uint8_t *buf = RTMemAlloc(1600);
1155	m->m_data -= if_maxlinkhdr;
1156	m->m_len += ETH_HLEN;
1157	eh = mtod(m, struct ethhdr *);
1158	#else
1159	uint8_t buf[1600];
1160	struct ethhdr eh = (struct ethhdr )buf;
1161
1162	if (ip_data_len + ETH_HLEN > sizeof(buf))
1163	return;
1164
1165	memcpy(buf + sizeof(struct ethhdr), ip_data, ip_data_len);
1166	#endif
1167
1168
1169	memcpy(eh->h_dest, client_ethaddr, ETH_ALEN);
1170	memcpy(eh->h_source, special_ethaddr, ETH_ALEN - 1);
1171	/* XXX: not correct */
1172	eh->h_source[5] = CTL_ALIAS;
1173	#ifdef VBOX_WITH_SIMPLIFIED_SLIRP_SYNC
1174	eh->h_proto = htons(eth_proto);
1175	#if 0
1176	slirp_output(pData->pvUser, m, mtod(m, uint8_t *), m->m_len);
1177	#else
1178	memcpy(buf, mtod(m, uint8_t *), m->m_len);
1179	slirp_output(pData->pvUser, NULL, buf, m->m_len);
1180	m_free(pData, m);
1181	#endif
1182	#else
1183	eh->h_proto = htons(ETH_P_IP);
1184	slirp_output(pData->pvUser, buf, ip_data_len + ETH_HLEN);
1185	#endif
1186	}
1187
1188	int slirp_redir(PNATState pData, int is_udp, int host_port,
1189	struct in_addr guest_addr, int guest_port)
1190	{
1191	if (is_udp)
1192	{
1193	if (!udp_listen(pData, htons(host_port), guest_addr.s_addr,
1194	htons(guest_port), 0))
1195	return -1;
1196	}
1197	else
1198	{
1199	if (!solisten(pData, htons(host_port), guest_addr.s_addr,
1200	htons(guest_port), 0))
1201	return -1;
1202	}
1203	return 0;
1204	}
1205
1206	int slirp_add_exec(PNATState pData, int do_pty, const char *args, int addr_low_byte,
1207	int guest_port)
1208	{
1209	return add_exec(&exec_list, do_pty, (char *)args,
1210	addr_low_byte, htons(guest_port));
1211	}
1212
1213	void slirp_set_ethaddr(PNATState pData, const uint8_t *ethaddr)
1214	{
1215	memcpy(client_ethaddr, ethaddr, ETH_ALEN);
1216	}
1217
1218	#if defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) && defined(RT_OS_WINDOWS)
1219	HANDLE *slirp_get_events(PNATState pData)
1220	{
1221	return pData->phEvents;
1222	}
1223	void slirp_register_external_event(PNATState pData, HANDLE hEvent, int index)
1224	{
1225	pData->phEvents[index] = hEvent;
1226	}
1227	#endif
1228
1229	unsigned int slirp_get_timeout_ms(PNATState pData)
1230	{
1231	if (link_up)
1232	{
1233	if (time_fasttimo)
1234	return 2;
1235	if (do_slowtimo)
1236	return 500; /* see PR_SLOWHZ */
1237	}
1238	return 0;
1239	}
1240
1241	/*
1242	* this function called from NAT thread
1243	*/
1244	void slirp_post_sent(PNATState pData, void *pvArg)
1245	{
1246	struct socket *so = 0;
1247	struct tcpcb *tp = 0;
1248	struct mbuf m = (struct mbuf )pvArg;
1249	m_free(pData, m);
1250	}

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

source: vbox/trunk/src/VBox/Devices/Network/slirp/slirp.c@ 16254

Download in other formats: