slirp.c@ 16291

Last change on this file since 16291 was 16291, checked in by vboxsync, 16 years ago
NAT: multi threading. Introduces set of macroces to use locking/unlocking/creating/destroing mutexes Also every socket enqueueing is accomponement with lock creation (it doen't work yet, compilable on Windows don't know about Linux)
Property svn:eol-style set to `native`
File size: 38.9 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	#ifdef VBOX_WITH_SLIRP_MT
11	# define CONTINUE(label) goto loop_end_ ## label ## _mt
12	/* @todo replace queue parameter with macrodinition */
13	# define LOOP_LABEL(label, so, sonext) loop_end_ ## label ## _mt: \
14	SOCKET_UNLOCK(so); \
15	QSOCKET_LOCK(_X(queue_ ## label ## _label)); \
16	(so) = (sonext)
17	#else
18	#define CONTINUE(label) continue;
19	# define LOOP_LABEL(label, so, sonext) /* empty*/
20	#endif
21	#if !defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) \|\| !defined(RT_OS_WINDOWS)
22
23	# define DO_ENGAGE_EVENT1(so, fdset, label) \
24	do { \
25	FD_SET((so)->s, (fdset)); \
26	UPD_NFDS((so)->s); \
27	} while(0)
28
29
30	# define DO_ENGAGE_EVENT2(so, fdset1, fdset2, label) \
31	do { \
32	FD_SET((so)->s, (fdset1)); \
33	FD_SET((so)->s, (fdset2)); \
34	UPD_NFDS((so)->s); \
35	} while(0)
36
37	# define DO_POLL_EVENTS(rc, error, so, events, label) do {} while (0)
38
39	# define DO_CHECK_FD_SET(so, events, fdset) (FD_ISSET((so)->s, (fdset)))
40
41	# define DO_WIN_CHECK_FD_SET(so, events, fdset ) 0 /* specific for Windows Winsock API */
42
43	# ifndef RT_OS_WINDOWS
44	# define ICMP_ENGAGE_EVENT(so, fdset) \
45	do { \
46	if (pData->icmp_socket.s != -1) \
47	DO_ENGAGE_EVENT1((so), (fdset), ICMP); \
48	} while (0)
49	# else /* !RT_OS_WINDOWS */
50	# define ICMP_ENGAGE_EVENT(so, fdset) do {} while(0)
51	#endif /* RT_OS_WINDOWS */
52
53	#else /* defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) && defined(RT_OS_WINDOWS) */
54
55	/*
56	* On Windows, we will be notified by IcmpSendEcho2() when the response arrives.
57	* So no call to WSAEventSelect necessary.
58	*/
59	# define ICMP_ENGAGE_EVENT(so, fdset) do {} while(0)
60
61	# define DO_ENGAGE_EVENT1(so, fdset1, label) \
62	do { \
63	rc = WSAEventSelect((so)->s, VBOX_SOCKET_EVENT, FD_ALL_EVENTS); \
64	if (rc == SOCKET_ERROR) \
65	{ \
66	/* This should not happen */ \
67	error = WSAGetLastError(); \
68	LogRel(("WSAEventSelector (" #label ") error %d (so=%x, socket=%s, event=%x)\n", \
69	error, (so), (so)->s, VBOX_SOCKET_EVENT)); \
70	} \
71	} while(0); \
72	CONTINUE(label)
73
74	# define DO_ENGAGE_EVENT2(so, fdset1, fdset2, label) \
75	DO_ENGAGE_EVENT1((so), (fdset1), label)
76
77	# define DO_POLL_EVENTS(rc, error, so, events, label) \
78	(rc) = WSAEnumNetworkEvents((so)->s, VBOX_SOCKET_EVENT, (events)); \
79	if ((rc) == SOCKET_ERROR) \
80	{ \
81	(error) = WSAGetLastError(); \
82	LogRel(("WSAEnumNetworkEvents " #label " error %d\n", (error))); \
83	CONTINUE(label); \
84	}
85
86	# define acceptds_win FD_ACCEPT
87	# define acceptds_win_bit FD_ACCEPT_BIT
88
89	# define readfds_win FD_READ
90	# define readfds_win_bit FD_READ_BIT
91
92	# define writefds_win FD_WRITE
93	# define writefds_win_bit FD_WRITE_BIT
94
95	# define xfds_win FD_OOB
96	# define xfds_win_bit FD_OOB_BIT
97
98	# define DO_CHECK_FD_SET(so, events, fdset) \
99	(((events).lNetworkEvents & fdset ## _win) && ((events).iErrorCode[fdset ## _win_bit] == 0))
100
101	# define DO_WIN_CHECK_FD_SET(so, events, fdset ) DO_CHECK_FD_SET((so), (events), fdset)
102
103	#endif /* defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) && defined(RT_OS_WINDOWS) */
104
105	#define TCP_ENGAGE_EVENT1(so, fdset) \
106	DO_ENGAGE_EVENT1((so), (fdset), tcp)
107
108	#define TCP_ENGAGE_EVENT2(so, fdset1, fdset2) \
109	DO_ENGAGE_EVENT2((so), (fdset1), (fdset2), tcp)
110
111	#define UDP_ENGAGE_EVENT(so, fdset) \
112	DO_ENGAGE_EVENT1((so), (fdset), udp)
113
114	#define POLL_TCP_EVENTS(rc, error, so, events) \
115	DO_POLL_EVENTS((rc), (error), (so), (events), tcp)
116
117	#define POLL_UDP_EVENTS(rc, error, so, events) \
118	DO_POLL_EVENTS((rc), (error), (so), (events), udp)
119
120	#define CHECK_FD_SET(so, events, set) \
121	(DO_CHECK_FD_SET((so), (events), set))
122
123	#define WIN_CHECK_FD_SET(so, events, set) \
124	(DO_WIN_CHECK_FD_SET((so), (events), set))
125
126	/*
127	* Loging macros
128	*/
129	#if VBOX_WITH_DEBUG_NAT_SOCKETS
130	# if defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) && defined(RT_OS_WINDOWS)
131	# define DO_LOG_NAT_SOCK(so, proto, winevent, r_fdset, w_fdset, x_fdset) \
132	do { \
133	LogRel((" " #proto "%R[natsock] %R[natwinnetevents]\n", (so), (winevent))); \
134	} while (0)
135	# else
136	# define DO_LOG_NAT_SOCK(so, proto, winevent, r_fdset, w_fdset, x_fdset) \
137	do { \
138	LogRel((" " #proto " %R[natsock] %s %s %s\n", (so), FD_ISSET((so)->s, (r_fdset))?"READ":"",\
139	FD_ISSET((so)->s, (w_fdset))?"WRITE":"", FD_ISSET((so)->s, (x_fdset))?"OOB":"")); \
140	} while (0)
141	# endif /* VBOX_WITH_DEBUG_NAT_SOCKETS */
142	#else
143	# define DO_LOG_NAT_SOCK(so, proto, winevent, r_fdset, w_fdset, x_fdset) do {} while (0)
144	#endif /* !VBOX_WITH_DEBUG_NAT_SOCKETS */
145
146	#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))
147
148	static const uint8_t special_ethaddr[6] =
149	{
150	0x52, 0x54, 0x00, 0x12, 0x35, 0x00
151	};
152
153	#ifdef RT_OS_WINDOWS
154
155	static int get_dns_addr_domain(PNATState pData, bool fVerbose,
156	struct in_addr *pdns_addr,
157	const char **ppszDomain)
158	{
159	int rc = 0;
160	FIXED_INFO *FixedInfo = NULL;
161	ULONG BufLen;
162	DWORD ret;
163	IP_ADDR_STRING *pIPAddr;
164	struct in_addr tmp_addr;
165	#ifdef VBOX_WITH_MULTI_DNS
166	struct in_addr local_mask;
167	struct in_addr local_network;
168	#endif
169
170	FixedInfo = (FIXED_INFO *)GlobalAlloc(GPTR, sizeof(FIXED_INFO));
171	BufLen = sizeof(FIXED_INFO);
172
173	/** @todo: this API returns all DNS servers, no matter whether the
174	* corresponding network adapter is disabled or not. Maybe replace
175	* this by GetAdapterAddresses(), which is XP/Vista only though. */
176	if (ERROR_BUFFER_OVERFLOW == GetNetworkParams(FixedInfo, &BufLen))
177	{
178	if (FixedInfo)
179	{
180	GlobalFree(FixedInfo);
181	FixedInfo = NULL;
182	}
183	FixedInfo = GlobalAlloc(GPTR, BufLen);
184	}
185
186	if ((ret = GetNetworkParams(FixedInfo, &BufLen)) != ERROR_SUCCESS)
187	{
188	Log(("GetNetworkParams failed. ret = %08x\n", (u_int)ret ));
189	if (FixedInfo)
190	{
191	GlobalFree(FixedInfo);
192	FixedInfo = NULL;
193	}
194	rc = -1;
195	goto get_dns_prefix;
196	}
197
198	#ifndef VBOX_WITH_MULTI_DNS
199	pIPAddr = &(FixedInfo->DnsServerList);
200	inet_aton(pIPAddr->IpAddress.String, &tmp_addr);
201	Log(("nat: DNS Servers:\n"));
202	if (fVerbose \|\| pdns_addr->s_addr != tmp_addr.s_addr)
203	LogRel(("NAT: DNS address: %s\n", pIPAddr->IpAddress.String));
204	*pdns_addr = tmp_addr;
205
206	pIPAddr = FixedInfo -> DnsServerList.Next;
207	while (pIPAddr)
208	{
209	if (fVerbose)
210	LogRel(("NAT: ignored DNS address: %s\n", pIPAddr ->IpAddress.String));
211	pIPAddr = pIPAddr ->Next;
212	}
213	#else
214	/localhost mask /
215	inet_aton("255.0.0.0", &local_mask);
216	inet_aton("127.0.0.0", &local_network);
217	for (pIPAddr = &FixedInfo->DnsServerList; pIPAddr != NULL; pIPAddr = pIPAddr->Next)
218	{
219	struct dns_entry *da = RTMemAllocZ(sizeof (struct dns_entry));
220	if (da == NULL)
221	{
222	LogRel(("can't alloc memory for DNS entry\n"));
223	return -1;
224	}
225	/check /
226	inet_aton(pIPAddr->IpAddress.String, &da->de_addr);
227	if ((ntohl(da->de_addr.s_addr) & ntohl(local_mask.s_addr)) == ntohl(local_network.s_addr)) {
228	da->de_addr.s_addr = htonl(ntohl(special_addr.s_addr) \| CTL_ALIAS);
229	}
230	LIST_INSERT_HEAD(&pData->dns_list_head, da, de_list);
231	}
232	#endif
233	if (FixedInfo)
234	{
235	GlobalFree(FixedInfo);
236	FixedInfo = NULL;
237	}
238
239	get_dns_prefix:
240	if (ppszDomain)
241	{
242	OSVERSIONINFO ver;
243	char szDnsDomain[256];
244	DWORD dwSize = sizeof(szDnsDomain);
245
246	*ppszDomain = NULL;
247	GetVersionEx(&ver);
248	if (ver.dwMajorVersion >= 5)
249	{
250	/* GetComputerNameEx exists in Windows versions starting with 2000. */
251	if (GetComputerNameEx(ComputerNameDnsDomain, szDnsDomain, &dwSize))
252	{
253	if (szDnsDomain[0])
254	{
255	/* Just non-empty strings are valid. */
256	*ppszDomain = RTStrDup(szDnsDomain);
257	if (pData->fPassDomain)
258	{
259	if (fVerbose)
260	LogRel(("NAT: passing domain name %s\n", szDnsDomain));
261	}
262	else
263	Log(("nat: ignoring domain %s\n", szDnsDomain));
264	}
265	}
266	else
267	Log(("nat: GetComputerNameEx failed (%d)\n", GetLastError()));
268	}
269	}
270	return rc;
271	}
272
273	#else
274
275	static int get_dns_addr_domain(PNATState pData, bool fVerbose,
276	struct in_addr *pdns_addr,
277	const char **ppszDomain)
278	{
279	char buff[512];
280	char buff2[256];
281	FILE *f;
282	int found = 0;
283	struct in_addr tmp_addr;
284
285	#ifdef RT_OS_OS2
286	/* Try various locations. */
287	char *etc = getenv("ETC");
288	f = NULL;
289	if (etc)
290	{
291	snprintf(buff, sizeof(buff), "%s/RESOLV2", etc);
292	f = fopen(buff, "rt");
293	}
294	if (!f)
295	{
296	snprintf(buff, sizeof(buff), "%s/RESOLV2", _PATH_ETC);
297	f = fopen(buff, "rt");
298	}
299	if (!f)
300	{
301	snprintf(buff, sizeof(buff), "%s/resolv.conf", _PATH_ETC);
302	f = fopen(buff, "rt");
303	}
304	#else
305	f = fopen("/etc/resolv.conf", "r");
306	#endif
307	if (!f)
308	return -1;
309
310	if (ppszDomain)
311	*ppszDomain = NULL;
312	Log(("nat: DNS Servers:\n"));
313	while (fgets(buff, 512, f) != NULL)
314	{
315	if (sscanf(buff, "nameserver%*[ \t]%256s", buff2) == 1)
316	{
317	if (!inet_aton(buff2, &tmp_addr))
318	continue;
319	/* If it's the first one, set it to dns_addr */
320	if (!found)
321	{
322	if (fVerbose \|\| pdns_addr->s_addr != tmp_addr.s_addr)
323	LogRel(("NAT: DNS address: %s\n", buff2));
324	*pdns_addr = tmp_addr;
325	}
326	else
327	{
328	if (fVerbose)
329	LogRel(("NAT: ignored DNS address: %s\n", buff2));
330	}
331	found++;
332	}
333	if ( ppszDomain
334	&& (!strncmp(buff, "domain", 6) \|\| !strncmp(buff, "search", 6)))
335	{
336	/* Domain name/search list present. Pick first entry */
337	if (*ppszDomain == NULL)
338	{
339	char *tok;
340	char *saveptr;
341	tok = strtok_r(&buff[6], " \t\n", &saveptr);
342	if (tok)
343	{
344	*ppszDomain = RTStrDup(tok);
345	if (pData->fPassDomain)
346	{
347	if (fVerbose)
348	LogRel(("NAT: passing domain name %s\n", tok));
349	}
350	else
351	Log(("nat: ignoring domain %s\n", tok));
352	}
353	}
354	}
355	}
356	fclose(f);
357	if (!found)
358	return -1;
359	return 0;
360	}
361
362	#endif
363
364	int get_dns_addr(PNATState pData, struct in_addr *pdns_addr)
365	{
366	return get_dns_addr_domain(pData, false, pdns_addr, NULL);
367	}
368
369	int slirp_init(PNATState ppData, const char pszNetAddr, uint32_t u32Netmask,
370	bool fPassDomain, const char *pszTFTPPrefix,
371	const char pszBootFile, void pvUser)
372	{
373	int fNATfailed = 0;
374	PNATState pData = RTMemAlloc(sizeof(NATState));
375	*ppData = pData;
376	if (!pData)
377	return VERR_NO_MEMORY;
378	if (u32Netmask & 0x1f)
379	/* CTL is x.x.x.15, bootp passes up to 16 IPs (15..31) */
380	return VERR_INVALID_PARAMETER;
381	memset(pData, '\0', sizeof(NATState));
382	pData->fPassDomain = fPassDomain;
383	pData->pvUser = pvUser;
384	tftp_prefix = pszTFTPPrefix;
385	bootp_filename = pszBootFile;
386	pData->netmask = u32Netmask;
387
388	#ifdef RT_OS_WINDOWS
389	{
390	WSADATA Data;
391	WSAStartup(MAKEWORD(2,0), &Data);
392	}
393	# if defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC)
394	pData->phEvents[VBOX_SOCKET_EVENT_INDEX] = CreateEvent(NULL, FALSE, FALSE, NULL);
395	# endif
396	#endif
397
398	link_up = 1;
399
400	debug_init();
401	if_init(pData);
402	ip_init(pData);
403	icmp_init(pData);
404
405	/* Initialise mbufs after setting the MTU */
406	m_init(pData);
407
408	inet_aton(pszNetAddr, &special_addr);
409	alias_addr.s_addr = special_addr.s_addr \| htonl(CTL_ALIAS);
410	/* @todo: add ability to configure this staff */
411	pData->tftp_server.s_addr = htonl(ntohl(special_addr.s_addr) \| CTL_TFTP);
412
413	/* set default addresses */
414	inet_aton("127.0.0.1", &loopback_addr);
415	#ifndef VBOX_WITH_MULTI_DNS
416	inet_aton("127.0.0.1", &dns_addr);
417
418	if (get_dns_addr_domain(pData, true, &dns_addr, &pData->pszDomain) < 0)
419	#else
420	if (get_dns_addr_domain(pData, true, NULL, &pData->pszDomain) < 0)
421	#endif
422	fNATfailed = 1;
423
424	getouraddr(pData);
425	return fNATfailed ? VINF_NAT_DNS : VINF_SUCCESS;
426	}
427
428	/**
429	* Statistics counters.
430	*/
431	void slirp_register_timers(PNATState pData, PPDMDRVINS pDrvIns)
432	{
433	#ifdef VBOX_WITH_STATISTICS
434	PDMDrvHlpSTAMRegisterF(pDrvIns, &pData->StatFill, STAMTYPE_PROFILE, STAMVISIBILITY_ALWAYS,
435	STAMUNIT_TICKS_PER_CALL, "Profiling slirp fills", "/Drivers/NAT%d/Fill", pDrvIns->iInstance);
436	PDMDrvHlpSTAMRegisterF(pDrvIns, &pData->StatPoll, STAMTYPE_PROFILE, STAMVISIBILITY_ALWAYS,
437	STAMUNIT_TICKS_PER_CALL, "Profiling slirp polls", "/Drivers/NAT%d/Poll", pDrvIns->iInstance);
438	PDMDrvHlpSTAMRegisterF(pDrvIns, &pData->StatFastTimer, STAMTYPE_PROFILE, STAMVISIBILITY_ALWAYS,
439	STAMUNIT_TICKS_PER_CALL, "Profiling slirp fast timer", "/Drivers/NAT%d/TimerFast", pDrvIns->iInstance);
440	PDMDrvHlpSTAMRegisterF(pDrvIns, &pData->StatSlowTimer, STAMTYPE_PROFILE, STAMVISIBILITY_ALWAYS,
441	STAMUNIT_TICKS_PER_CALL, "Profiling slirp slow timer", "/Drivers/NAT%d/TimerSlow", pDrvIns->iInstance);
442	PDMDrvHlpSTAMRegisterF(pDrvIns, &pData->StatTCP, STAMTYPE_COUNTER, STAMVISIBILITY_ALWAYS,
443	STAMUNIT_COUNT, "TCP sockets", "/Drivers/NAT%d/SockTCP", pDrvIns->iInstance);
444	PDMDrvHlpSTAMRegisterF(pDrvIns, &pData->StatTCPHot, STAMTYPE_COUNTER, STAMVISIBILITY_ALWAYS,
445	STAMUNIT_COUNT, "TCP sockets active", "/Drivers/NAT%d/SockTCPHot", pDrvIns->iInstance);
446	PDMDrvHlpSTAMRegisterF(pDrvIns, &pData->StatUDP, STAMTYPE_COUNTER, STAMVISIBILITY_ALWAYS,
447	STAMUNIT_COUNT, "UDP sockets", "/Drivers/NAT%d/SockUDP", pDrvIns->iInstance);
448	PDMDrvHlpSTAMRegisterF(pDrvIns, &pData->StatUDPHot, STAMTYPE_COUNTER, STAMVISIBILITY_ALWAYS,
449	STAMUNIT_COUNT, "UDP sockets active", "/Drivers/NAT%d/SockUDPHot", pDrvIns->iInstance);
450	#endif /* VBOX_WITH_STATISTICS */
451	}
452
453	/**
454	* Marks the link as up, making it possible to establish new connections.
455	*/
456	void slirp_link_up(PNATState pData)
457	{
458	link_up = 1;
459	}
460
461	/**
462	* Marks the link as down and cleans up the current connections.
463	*/
464	void slirp_link_down(PNATState pData)
465	{
466	struct socket *so;
467
468	while ((so = tcb.so_next) != &tcb)
469	{
470	if (so->so_state & SS_NOFDREF \|\| so->s == -1)
471	sofree(pData, so);
472	else
473	tcp_drop(pData, sototcpcb(so), 0);
474	}
475
476	while ((so = udb.so_next) != &udb)
477	udp_detach(pData, so);
478
479	link_up = 0;
480	}
481
482	/**
483	* Terminates the slirp component.
484	*/
485	void slirp_term(PNATState pData)
486	{
487	if (pData->pszDomain)
488	RTStrFree((char )(void )pData->pszDomain);
489
490	#ifdef RT_OS_WINDOWS
491	pData->pfIcmpCloseHandle(pData->icmp_socket.sh);
492	FreeLibrary(pData->hmIcmpLibrary);
493	RTMemFree(pData->pvIcmpBuffer);
494	# else
495	closesocket(pData->icmp_socket.s);
496	#endif
497
498	slirp_link_down(pData);
499	#ifdef RT_OS_WINDOWS
500	WSACleanup();
501	#endif
502	#ifdef LOG_ENABLED
503	Log(("\n"
504	"NAT statistics\n"
505	"--------------\n"
506	"\n"));
507	ipstats(pData);
508	tcpstats(pData);
509	udpstats(pData);
510	icmpstats(pData);
511	mbufstats(pData);
512	sockstats(pData);
513	Log(("\n"
514	"\n"
515	"\n"));
516	#endif
517	RTMemFree(pData);
518	}
519
520
521	#define CONN_CANFSEND(so) (((so)->so_state & (SS_FCANTSENDMORE\|SS_ISFCONNECTED)) == SS_ISFCONNECTED)
522	#define CONN_CANFRCV(so) (((so)->so_state & (SS_FCANTRCVMORE\|SS_ISFCONNECTED)) == SS_ISFCONNECTED)
523	#define UPD_NFDS(x) if (nfds < (x)) nfds = (x)
524
525	/*
526	* curtime kept to an accuracy of 1ms
527	*/
528	#ifdef RT_OS_WINDOWS
529	static void updtime(PNATState pData)
530	{
531	struct _timeb tb;
532
533	_ftime(&tb);
534	curtime = (u_int)tb.time * (u_int)1000;
535	curtime += (u_int)tb.millitm;
536	}
537	#else
538	static void updtime(PNATState pData)
539	{
540	gettimeofday(&tt, 0);
541
542	curtime = (u_int)tt.tv_sec * (u_int)1000;
543	curtime += (u_int)tt.tv_usec / (u_int)1000;
544
545	if ((tt.tv_usec % 1000) >= 500)
546	curtime++;
547	}
548	#endif
549
550	void slirp_select_fill(PNATState pData, int *pnfds,
551	fd_set readfds, fd_set writefds, fd_set *xfds)
552	{
553	struct socket so, so_next;
554	int nfds;
555	#if defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) && defined(RT_OS_WINDOWS)
556	int rc;
557	int error;
558	#endif
559	int i;
560
561	STAM_PROFILE_START(&pData->StatFill, a);
562
563	nfds = *pnfds;
564
565	/*
566	* First, TCP sockets
567	*/
568	do_slowtimo = 0;
569	if (link_up)
570	{
571	/*
572	* *_slowtimo needs calling if there are IP fragments
573	* in the fragment queue, or there are TCP connections active
574	*/
575	/* XXX:
576	* triggering of fragment expiration should be the same but use new macroses
577	*/
578	do_slowtimo = (tcb.so_next != &tcb);
579	if (!do_slowtimo)
580	{
581	for (i = 0; i < IPREASS_NHASH; i++)
582	{
583	if (!TAILQ_EMPTY(&ipq[i]))
584	{
585	do_slowtimo = 1;
586	break;
587	}
588	}
589	}
590	ICMP_ENGAGE_EVENT(&pData->icmp_socket, readfds);
591
592	STAM_COUNTER_RESET(&pData->StatTCP);
593	STAM_COUNTER_RESET(&pData->StatTCPHot);
594
595	QSOCKET_FOREACH(so, so_next, tcp)
596	/* { */
597	STAM_COUNTER_INC(&pData->StatTCP);
598
599	/*
600	* See if we need a tcp_fasttimo
601	*/
602	if (time_fasttimo == 0 && so->so_tcpcb->t_flags & TF_DELACK)
603	time_fasttimo = curtime; /* Flag when we want a fasttimo */
604
605	/*
606	* NOFDREF can include still connecting to local-host,
607	* newly socreated() sockets etc. Don't want to select these.
608	*/
609	if (so->so_state & SS_NOFDREF \|\| so->s == -1)
610	CONTINUE(tcp);
611
612	/*
613	* Set for reading sockets which are accepting
614	*/
615	if (so->so_state & SS_FACCEPTCONN)
616	{
617	STAM_COUNTER_INC(&pData->StatTCPHot);
618	TCP_ENGAGE_EVENT1(so, readfds);
619	CONTINUE(tcp);
620	}
621
622	/*
623	* Set for writing sockets which are connecting
624	*/
625	if (so->so_state & SS_ISFCONNECTING)
626	{
627	Log2(("connecting %R[natsock] engaged\n",so));
628	STAM_COUNTER_INC(&pData->StatTCPHot);
629	TCP_ENGAGE_EVENT1(so, writefds);
630	}
631
632	/*
633	* Set for writing if we are connected, can send more, and
634	* we have something to send
635	*/
636	if (CONN_CANFSEND(so) && so->so_rcv.sb_cc)
637	{
638	STAM_COUNTER_INC(&pData->StatTCPHot);
639	TCP_ENGAGE_EVENT1(so, writefds);
640	}
641
642	/*
643	* Set for reading (and urgent data) if we are connected, can
644	* receive more, and we have room for it XXX /2 ?
645	*/
646	if (CONN_CANFRCV(so) && (so->so_snd.sb_cc < (so->so_snd.sb_datalen/2)))
647	{
648	STAM_COUNTER_INC(&pData->StatTCPHot);
649	TCP_ENGAGE_EVENT2(so, readfds, xfds);
650	}
651	LOOP_LABEL(tcp, so, so_next);
652	}
653
654	/*
655	* UDP sockets
656	*/
657	STAM_COUNTER_RESET(&pData->StatUDP);
658	STAM_COUNTER_RESET(&pData->StatUDPHot);
659
660	QSOCKET_FOREACH(so, so_next, udp)
661	/* { */
662	so_next = so->so_next;
663
664	STAM_COUNTER_INC(&pData->StatUDP);
665
666	/*
667	* See if it's timed out
668	*/
669	if (so->so_expire)
670	{
671	if (so->so_expire <= curtime)
672	{
673	udp_detach(pData, so);
674	CONTINUE(udp);
675	}
676	else
677	do_slowtimo = 1; /* Let socket expire */
678	}
679
680	/*
681	* When UDP packets are received from over the link, they're
682	* sendto()'d straight away, so no need for setting for writing
683	* Limit the number of packets queued by this session to 4.
684	* Note that even though we try and limit this to 4 packets,
685	* the session could have more queued if the packets needed
686	* to be fragmented.
687	*
688	* (XXX <= 4 ?)
689	*/
690	if ((so->so_state & SS_ISFCONNECTED) && so->so_queued <= 4)
691	{
692	STAM_COUNTER_INC(&pData->StatUDPHot);
693	UDP_ENGAGE_EVENT(so, readfds);
694	}
695	LOOP_LABEL(udp, so, so_next);
696	}
697
698	}
699
700	#if !defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) \|\| !defined(RT_OS_WINDOWS)
701	*pnfds = nfds;
702	#else
703	*pnfds = VBOX_EVENT_COUNT;
704	#endif
705
706	STAM_PROFILE_STOP(&pData->StatFill, a);
707	}
708
709	#if defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) && defined(RT_OS_WINDOWS)
710	void slirp_select_poll(PNATState pData, int fTimeout, int fIcmp)
711	#else
712	void slirp_select_poll(PNATState pData, fd_set readfds, fd_set writefds, fd_set *xfds)
713	#endif
714	{
715	struct socket so, so_next;
716	int ret;
717	#if defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) && defined(RT_OS_WINDOWS)
718	WSANETWORKEVENTS NetworkEvents;
719	int rc;
720	int error;
721	#endif
722
723	STAM_PROFILE_START(&pData->StatPoll, a);
724
725	/* Update time */
726	updtime(pData);
727
728	/*
729	* See if anything has timed out
730	*/
731	if (link_up)
732	{
733	if (time_fasttimo && ((curtime - time_fasttimo) >= 2))
734	{
735	STAM_PROFILE_START(&pData->StatFastTimer, a);
736	tcp_fasttimo(pData);
737	time_fasttimo = 0;
738	STAM_PROFILE_STOP(&pData->StatFastTimer, a);
739	}
740	if (do_slowtimo && ((curtime - last_slowtimo) >= 499))
741	{
742	STAM_PROFILE_START(&pData->StatSlowTimer, a);
743	ip_slowtimo(pData);
744	tcp_slowtimo(pData);
745	last_slowtimo = curtime;
746	STAM_PROFILE_STOP(&pData->StatSlowTimer, a);
747	}
748	}
749	#if defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) && defined(RT_OS_WINDOWS)
750	if (fTimeout)
751	return; /* only timer update */
752	#endif
753
754	/*
755	* Check sockets
756	*/
757	if (link_up)
758	{
759	#if defined(RT_OS_WINDOWS)
760	/*XXX: before renaming please make see define
761	* fIcmp in slirp_state.h
762	*/
763	if (fIcmp)
764	sorecvfrom(pData, &pData->icmp_socket);
765	#else
766	if (pData->icmp_socket.s != -1 && FD_ISSET(pData->icmp_socket.s, readfds))
767	sorecvfrom(pData, &pData->icmp_socket);
768	#endif
769	/*
770	* Check TCP sockets
771	*/
772	for (so = tcb.so_next; so != &tcb; so = so_next)
773	{
774	so_next = so->so_next;
775
776	/*
777	* FD_ISSET is meaningless on these sockets
778	* (and they can crash the program)
779	*/
780	if (so->so_state & SS_NOFDREF \|\| so->s == -1)
781	CONTINUE(tcp);
782
783	POLL_TCP_EVENTS(rc, error, so, &NetworkEvents);
784
785	LOG_NAT_SOCK(so, TCP, &NetworkEvents, readfds, writefds, xfds);
786
787	/*
788	* Check for URG data
789	* This will soread as well, so no need to
790	* test for readfds below if this succeeds
791	*/
792
793	/* out-of-band data */
794	if (CHECK_FD_SET(so, NetworkEvents, xfds))
795	{
796	sorecvoob(pData, so);
797	}
798
799	/*
800	* Check sockets for reading
801	*/
802	else if ( CHECK_FD_SET(so, NetworkEvents, readfds)
803	\|\| WIN_CHECK_FD_SET(so, NetworkEvents, acceptds))
804	{
805	/*
806	* Check for incoming connections
807	*/
808	if (so->so_state & SS_FACCEPTCONN)
809	{
810	tcp_connect(pData, so);
811	#if defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) && defined(RT_OS_WINDOWS)
812	if (!(NetworkEvents.lNetworkEvents & FD_CLOSE))
813	#endif
814	CONTINUE(tcp);
815	}
816
817	ret = soread(pData, so, /fCloseIfNothingRead=/false);
818	/* Output it if we read something */
819	if (ret > 0)
820	tcp_output(pData, sototcpcb(so));
821	}
822
823	#if defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) && defined(RT_OS_WINDOWS)
824	/*
825	* Check for FD_CLOSE events.
826	*/
827	if (NetworkEvents.lNetworkEvents & FD_CLOSE)
828	{
829	/*
830	* drain the socket
831	*/
832	for (;;)
833	{
834	ret = soread(pData, so, /fCloseIfNothingRead=/true);
835	if (ret > 0)
836	tcp_output(pData, sototcpcb(so));
837	else
838	break;
839	}
840	}
841	#endif
842
843	/*
844	* Check sockets for writing
845	*/
846	if (CHECK_FD_SET(so, NetworkEvents, writefds))
847	{
848	/*
849	* Check for non-blocking, still-connecting sockets
850	*/
851	if (so->so_state & SS_ISFCONNECTING)
852	{
853	Log2(("connecting %R[natsock] catched\n", so));
854	/* Connected */
855	so->so_state &= ~SS_ISFCONNECTING;
856
857	/*
858	* This should be probably guarded by PROBE_CONN too. Anyway,
859	* we disable it on OS/2 because the below send call returns
860	* EFAULT which causes the opened TCP socket to close right
861	* after it has been opened and connected.
862	*/
863	#ifndef RT_OS_OS2
864	ret = send(so->s, (const char *)&ret, 0, 0);
865	if (ret < 0)
866	{
867	/* XXXXX Must fix, zero bytes is a NOP */
868	if ( errno == EAGAIN
869	\|\| errno == EWOULDBLOCK
870	\|\| errno == EINPROGRESS
871	\|\| errno == ENOTCONN)
872	CONTINUE(tcp);
873
874	/* else failed */
875	so->so_state = SS_NOFDREF;
876	}
877	/* else so->so_state &= ~SS_ISFCONNECTING; */
878	#endif
879
880	/*
881	* Continue tcp_input
882	*/
883	tcp_input(pData, (struct mbuf *)NULL, sizeof(struct ip), so);
884	/* continue; */
885	}
886	else
887	ret = sowrite(pData, so);
888	/*
889	* XXX If we wrote something (a lot), there could be the need
890	* for a window update. In the worst case, the remote will send
891	* a window probe to get things going again.
892	*/
893	}
894
895	/*
896	* Probe a still-connecting, non-blocking socket
897	* to check if it's still alive
898	*/
899	#ifdef PROBE_CONN
900	if (so->so_state & SS_ISFCONNECTING)
901	{
902	ret = recv(so->s, (char *)&ret, 0, 0);
903
904	if (ret < 0)
905	{
906	/* XXX */
907	if ( errno == EAGAIN
908	\|\| errno == EWOULDBLOCK
909	\|\| errno == EINPROGRESS
910	\|\| errno == ENOTCONN)
911	{
912	CONTINUE(tcp); /* Still connecting, continue */
913	}
914
915	/* else failed */
916	so->so_state = SS_NOFDREF;
917
918	/* tcp_input will take care of it */
919	}
920	else
921	{
922	ret = send(so->s, &ret, 0, 0);
923	if (ret < 0)
924	{
925	/* XXX */
926	if ( errno == EAGAIN
927	\|\| errno == EWOULDBLOCK
928	\|\| errno == EINPROGRESS
929	\|\| errno == ENOTCONN)
930	{
931	CONTINUE(tcp);
932	}
933	/* else failed */
934	so->so_state = SS_NOFDREF;
935	}
936	else
937	so->so_state &= ~SS_ISFCONNECTING;
938
939	}
940	tcp_input((struct mbuf *)NULL, sizeof(struct ip),so);
941	} /* SS_ISFCONNECTING */
942	#endif
943	LOOP_LABEL(tcp, so, so_next);
944	}
945
946	/*
947	* Now UDP sockets.
948	* Incoming packets are sent straight away, they're not buffered.
949	* Incoming UDP data isn't buffered either.
950	*/
951	for (so = udb.so_next; so != &udb; so = so_next)
952	{
953	so_next = so->so_next;
954
955	POLL_UDP_EVENTS(rc, error, so, &NetworkEvents);
956
957	LOG_NAT_SOCK(so, UDP, &NetworkEvents, readfds, writefds, xfds);
958
959	if (so->s != -1 && CHECK_FD_SET(so, NetworkEvents, readfds))
960	{
961	sorecvfrom(pData, so);
962	}
963	LOOP_LABEL(udp, so, so_next);
964	}
965
966	}
967
968	#ifndef VBOX_WITH_SLIRP_MT
969	/*
970	* See if we can start outputting
971	*/
972	if (if_queued && link_up)
973	if_start(pData);
974	#endif
975
976	STAM_PROFILE_STOP(&pData->StatPoll, a);
977	}
978
979	#define ETH_ALEN 6
980	#define ETH_HLEN 14
981
982	#define ARPOP_REQUEST 1 /* ARP request */
983	#define ARPOP_REPLY 2 /* ARP reply */
984
985	struct ethhdr
986	{
987	unsigned char h_dest[ETH_ALEN]; /* destination eth addr */
988	unsigned char h_source[ETH_ALEN]; /* source ether addr */
989	unsigned short h_proto; /* packet type ID field */
990	};
991
992	struct arphdr
993	{
994	unsigned short ar_hrd; /* format of hardware address */
995	unsigned short ar_pro; /* format of protocol address */
996	unsigned char ar_hln; /* length of hardware address */
997	unsigned char ar_pln; /* length of protocol address */
998	unsigned short ar_op; /* ARP opcode (command) */
999
1000	/*
1001	* Ethernet looks like this : This bit is variable sized however...
1002	*/
1003	unsigned char ar_sha[ETH_ALEN]; /* sender hardware address */
1004	unsigned char ar_sip[4]; /* sender IP address */
1005	unsigned char ar_tha[ETH_ALEN]; /* target hardware address */
1006	unsigned char ar_tip[4]; /* target IP address */
1007	};
1008
1009	static
1010	#ifdef VBOX_WITH_SIMPLIFIED_SLIRP_SYNC
1011	void arp_input(PNATState pData, struct mbuf *m)
1012	#else
1013	void arp_input(PNATState pData, const uint8_t *pkt, int pkt_len)
1014	#endif
1015	{
1016	struct ethhdr *eh;
1017	struct ethhdr *reh;
1018	struct arphdr *ah;
1019	struct arphdr *rah;
1020	int ar_op;
1021	struct ex_list *ex_ptr;
1022	uint32_t htip;
1023	#ifndef VBOX_WITH_SIMPLIFIED_SLIRP_SYNC
1024	uint8_t arp_reply[sizeof(struct arphdr) + ETH_HLEN];
1025	eh = (struct ethhdr *)pkt;
1026	#else
1027	struct mbuf *mr;
1028	eh = mtod(m, struct ethhdr *);
1029	#endif
1030	ah = (struct arphdr *)&eh[1];
1031	htip = ntohl((uint32_t)ah->ar_tip);
1032
1033	#ifdef VBOX_WITH_SIMPLIFIED_SLIRP_SYNC
1034	mr = m_get(pData);
1035	mr->m_data += if_maxlinkhdr;
1036	mr->m_len = sizeof(struct arphdr);
1037	rah = mtod(mr, struct arphdr *);
1038	#else
1039	reh = (struct ethhdr *)arp_reply;
1040	rah = (struct arphdr *)&reh[1];
1041	#endif
1042
1043	ar_op = ntohs(ah->ar_op);
1044	switch(ar_op)
1045	{
1046	case ARPOP_REQUEST:
1047	if ((htip & pData->netmask) == ntohl(special_addr.s_addr))
1048	{
1049	if ( CTL_CHECK(htip,CTL_DNS)
1050	\|\| CTL_CHECK(htip, CTL_ALIAS)
1051	\|\| CTL_CHECK(htip, CTL_TFTP))
1052	goto arp_ok;
1053	for (ex_ptr = exec_list; ex_ptr; ex_ptr = ex_ptr->ex_next)
1054	{
1055	if ((htip & ~pData->netmask) == ex_ptr->ex_addr)
1056	goto arp_ok;
1057	}
1058	return;
1059	arp_ok:
1060
1061	#ifndef VBOX_WITH_SIMPLIFIED_SLIRP_SYNC
1062	memcpy(reh->h_dest, eh->h_source, ETH_ALEN);
1063	memcpy(reh->h_source, &special_addr, ETH_ALEN);
1064	reh->h_source[5] = ah->ar_tip[3];
1065	reh->h_proto = htons(ETH_P_ARP);
1066	#endif
1067	rah->ar_hrd = htons(1);
1068	rah->ar_pro = htons(ETH_P_IP);
1069	rah->ar_hln = ETH_ALEN;
1070	rah->ar_pln = 4;
1071	rah->ar_op = htons(ARPOP_REPLY);
1072	memcpy(rah->ar_sha, special_ethaddr, ETH_ALEN);
1073
1074	switch (htip & ~pData->netmask)
1075	{
1076	case CTL_DNS:
1077	case CTL_ALIAS:
1078	rah->ar_sha[5] = (uint8_t)(htip & ~pData->netmask);
1079	break;
1080	default:;
1081	}
1082
1083	memcpy(rah->ar_sip, ah->ar_tip, 4);
1084	memcpy(rah->ar_tha, ah->ar_sha, ETH_ALEN);
1085	memcpy(rah->ar_tip, ah->ar_sip, 4);
1086	#ifdef VBOX_WITH_SIMPLIFIED_SLIRP_SYNC
1087	if_encap(pData, ETH_P_ARP, mr);
1088	m_free(pData, m);
1089	#else
1090	slirp_output(pData->pvUser, arp_reply, sizeof(arp_reply));
1091	#endif
1092	}
1093	break;
1094	default:
1095	break;
1096	}
1097	}
1098
1099	void slirp_input(PNATState pData, const uint8_t *pkt, int pkt_len)
1100	{
1101	struct mbuf *m;
1102	int proto;
1103	static bool fWarnedIpv6;
1104
1105	if (pkt_len < ETH_HLEN)
1106	{
1107	LogRel(("NAT: packet having size %d has been ingnored\n", pkt_len));
1108	return;
1109	}
1110
1111	m = m_get(pData);
1112	if (!m)
1113	{
1114	LogRel(("can't allocate new mbuf\n"));
1115	return;
1116	}
1117
1118	/* Note: we add to align the IP header */
1119
1120	if (M_FREEROOM(m) < pkt_len)
1121	m_inc(m, pkt_len);
1122
1123	m->m_len = pkt_len ;
1124	memcpy(m->m_data, pkt, pkt_len);
1125
1126	proto = ntohs((uint16_t )(pkt + 12));
1127	switch(proto)
1128	{
1129	case ETH_P_ARP:
1130	#ifdef VBOX_WITH_SIMPLIFIED_SLIRP_SYNC
1131	arp_input(pData, m);
1132	#else
1133	arp_input(pData, pkt, pkt_len);
1134	m_free(pData, m);
1135	#endif
1136	break;
1137	case ETH_P_IP:
1138	/* Update time. Important if the network is very quiet, as otherwise
1139	* the first outgoing connection gets an incorrect timestamp. */
1140	updtime(pData);
1141	m->m_data += ETH_HLEN;
1142	m->m_len -= ETH_HLEN;
1143	ip_input(pData, m);
1144	break;
1145	case ETH_P_IPV6:
1146	m_free(pData, m);
1147	if (!fWarnedIpv6)
1148	{
1149	LogRel(("NAT: IPv6 not supported\n"));
1150	fWarnedIpv6 = true;
1151	}
1152	break;
1153	default:
1154	LogRel(("NAT: Unsupported protocol %x\n", proto));
1155	m_free(pData, m);
1156	break;
1157	}
1158	}
1159
1160	/* output the IP packet to the ethernet device */
1161	#ifdef VBOX_WITH_SIMPLIFIED_SLIRP_SYNC
1162	void if_encap(PNATState pData, uint16_t eth_proto, struct mbuf *m)
1163	#else
1164	void if_encap(PNATState pData, uint8_t *ip_data, int ip_data_len)
1165	#endif
1166	{
1167	#ifdef VBOX_WITH_SIMPLIFIED_SLIRP_SYNC
1168	struct ethhdr *eh;
1169	uint8_t *buf = RTMemAlloc(1600);
1170	m->m_data -= if_maxlinkhdr;
1171	m->m_len += ETH_HLEN;
1172	eh = mtod(m, struct ethhdr *);
1173	#else
1174	uint8_t buf[1600];
1175	struct ethhdr eh = (struct ethhdr )buf;
1176
1177	if (ip_data_len + ETH_HLEN > sizeof(buf))
1178	return;
1179
1180	memcpy(buf + sizeof(struct ethhdr), ip_data, ip_data_len);
1181	#endif
1182
1183
1184	memcpy(eh->h_dest, client_ethaddr, ETH_ALEN);
1185	memcpy(eh->h_source, special_ethaddr, ETH_ALEN - 1);
1186	/* XXX: not correct */
1187	eh->h_source[5] = CTL_ALIAS;
1188	#ifdef VBOX_WITH_SIMPLIFIED_SLIRP_SYNC
1189	eh->h_proto = htons(eth_proto);
1190	#if 0
1191	slirp_output(pData->pvUser, m, mtod(m, uint8_t *), m->m_len);
1192	#else
1193	memcpy(buf, mtod(m, uint8_t *), m->m_len);
1194	slirp_output(pData->pvUser, NULL, buf, m->m_len);
1195	m_free(pData, m);
1196	#endif
1197	#else
1198	eh->h_proto = htons(ETH_P_IP);
1199	slirp_output(pData->pvUser, buf, ip_data_len + ETH_HLEN);
1200	#endif
1201	}
1202
1203	int slirp_redir(PNATState pData, int is_udp, int host_port,
1204	struct in_addr guest_addr, int guest_port)
1205	{
1206	if (is_udp)
1207	{
1208	if (!udp_listen(pData, htons(host_port), guest_addr.s_addr,
1209	htons(guest_port), 0))
1210	return -1;
1211	}
1212	else
1213	{
1214	if (!solisten(pData, htons(host_port), guest_addr.s_addr,
1215	htons(guest_port), 0))
1216	return -1;
1217	}
1218	return 0;
1219	}
1220
1221	int slirp_add_exec(PNATState pData, int do_pty, const char *args, int addr_low_byte,
1222	int guest_port)
1223	{
1224	return add_exec(&exec_list, do_pty, (char *)args,
1225	addr_low_byte, htons(guest_port));
1226	}
1227
1228	void slirp_set_ethaddr(PNATState pData, const uint8_t *ethaddr)
1229	{
1230	memcpy(client_ethaddr, ethaddr, ETH_ALEN);
1231	}
1232
1233	#if defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) && defined(RT_OS_WINDOWS)
1234	HANDLE *slirp_get_events(PNATState pData)
1235	{
1236	return pData->phEvents;
1237	}
1238	void slirp_register_external_event(PNATState pData, HANDLE hEvent, int index)
1239	{
1240	pData->phEvents[index] = hEvent;
1241	}
1242	#endif
1243
1244	unsigned int slirp_get_timeout_ms(PNATState pData)
1245	{
1246	if (link_up)
1247	{
1248	if (time_fasttimo)
1249	return 2;
1250	if (do_slowtimo)
1251	return 500; /* see PR_SLOWHZ */
1252	}
1253	return 0;
1254	}
1255
1256	/*
1257	* this function called from NAT thread
1258	*/
1259	void slirp_post_sent(PNATState pData, void *pvArg)
1260	{
1261	struct socket *so = 0;
1262	struct tcpcb *tp = 0;
1263	struct mbuf m = (struct mbuf )pvArg;
1264	m_free(pData, m);
1265	}

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

source: vbox/trunk/src/VBox/Devices/Network/slirp/slirp.c@ 16291

Download in other formats: