slirp.c@ 15388

Last change on this file since 15388 was 15384, checked in by vboxsync, 16 years ago
slirp: removed debugging code
Property svn:eol-style set to `native`
File size: 35.6 KB

Line
1	#include "slirp.h"
2	#ifdef RT_OS_OS2
3	# include <paths.h>
4	#endif
5
6	/* disable these counters for the final release */
7	/* #define VBOX_WITHOUT_RELEASE_STATISTICS */
8
9	#include <VBox/err.h>
10	#include <VBox/pdmdrv.h>
11	#include <iprt/assert.h>
12
13	#if !defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) \|\| !defined(RT_OS_WINDOWS)
14
15	# define DO_ENGAGE_EVENT1(so, fdset, label) \
16	do { \
17	FD_SET((so)->s, (fdset)); \
18	UPD_NFDS((so)->s); \
19	} while(0)
20
21
22	# define DO_ENGAGE_EVENT2(so, fdset1, fdset2, label) \
23	do { \
24	FD_SET((so)->s, (fdset1)); \
25	FD_SET((so)->s, (fdset2)); \
26	UPD_NFDS((so)->s); \
27	} while(0)
28
29	# define DO_POLL_EVENTS(rc, error, so, events, label) do {} while (0)
30
31	#define DO_CHECK_FD_SET(so, events, fdset) (FD_ISSET((so)->s, (fdset)))
32
33	# ifdef VBOX_WITH_SLIRP_ICMP
34	# define ICMP_ENGAGE_EVENT(so, fdset) \
35	do { \
36	if (pData->icmp_socket.s != -1) \
37	DO_ENGAGE_EVENT1((so), (fdset), ICMP); \
38	} while (0)
39	# else /* !VBOX_WITH_SLIRP_ICMP */
40	# define ICMP_ENGAGE_EVENT(so, fdset) do {} while (0)
41	# endif /* !VBOX_WITH_SLIRP_ICMP */
42
43	#else /* defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) && defined(RT_OS_WINDOWS) */
44
45	/*
46	* On Windows, we will be notified by IcmpSendEcho2() when the response arrives.
47	* So no call to WSAEventSelect necessary.
48	*/
49	# define ICMP_ENGAGE_EVENT(so, fdset) do {} while(0)
50
51	# define DO_ENGAGE_EVENT1(so, fdset1, label) \
52	do { \
53	rc = WSAEventSelect((so)->s, VBOX_SOCKET_EVENT, FD_ALL_EVENTS); \
54	if (rc == SOCKET_ERROR) \
55	{ \
56	/* This should not happen */ \
57	error = WSAGetLastError(); \
58	LogRel(("WSAEventSelector (" #label ") error %d (so=%x, socket=%s, event=%x)\n", \
59	error, (so), (so)->s, VBOX_SOCKET_EVENT)); \
60	} \
61	} while(0); \
62	continue
63
64	# define DO_ENGAGE_EVENT2(so, fdset1, fdset2, label) \
65	DO_ENGAGE_EVENT1((so), (fdset1), label)
66
67	# define DO_POLL_EVENTS(rc, error, so, events, label) \
68	(rc) = WSAEnumNetworkEvents((so)->s, VBOX_SOCKET_EVENT, (events)); \
69	if ((rc) == SOCKET_ERROR) \
70	{ \
71	(error) = WSAGetLastError(); \
72	LogRel(("WSAEnumNetworkEvents " #label " error %d\n", (error))); \
73	continue; \
74	}
75
76	# define readfds_win FD_READ
77	# define readfds_win_bit FD_READ_BIT
78
79	# define writefds_win FD_WRITE
80	# define writefds_win_bit FD_WRITE_BIT
81
82	# define xfds_win FD_OOB
83	# define xfds_win_bit FD_OOB_BIT
84
85	# define DO_CHECK_FD_SET(so, events, fdset) \
86	(((events).lNetworkEvents & fdset ## _win) && ((events).iErrorCode[fdset ## _win_bit] == 0))
87
88
89	#endif /* defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) && defined(RT_OS_WINDOWS) */
90
91	#define TCP_ENGAGE_EVENT1(so, fdset) \
92	DO_ENGAGE_EVENT1((so), (fdset), TCP)
93
94	#define TCP_ENGAGE_EVENT2(so, fdset1, fdset2) \
95	DO_ENGAGE_EVENT2((so), (fdset1), (fdset2), TCP)
96
97	#define UDP_ENGAGE_EVENT(so, fdset) \
98	DO_ENGAGE_EVENT1((so), (fdset), UDP)
99
100	#define POLL_TCP_EVENTS(rc, error, so, events) \
101	DO_POLL_EVENTS((rc), (error), (so), (events), TCP)
102
103	#define POLL_UDP_EVENTS(rc, error, so, events) \
104	DO_POLL_EVENTS((rc), (error), (so), (events), UDP)
105
106	#define CHECK_FD_SET(so, events, set) \
107	(DO_CHECK_FD_SET((so), (events), set))
108
109	/*
110	* Loging macros
111	*/
112	#if VBOX_WITH_DEBUG_NAT_SOCKETS
113	# if defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) && defined(RT_OS_WINDOWS)
114	# define DO_LOG_NAT_SOCK(so, proto, winevent, r_fdset, w_fdset, x_fdset) \
115	do { \
116	LogRel((" " #proto "%R[natsock] %R[natwinnetevents]\n", (so), (winevent))); \
117	} while (0)
118	# else
119	# define DO_LOG_NAT_SOCK(so, proto, winevent, r_fdset, w_fdset, x_fdset) \
120	do { \
121	LogRel((" " #proto " %R[natsock] %s %s %s\n", (so), FD_ISSET((so)->s, (r_fdset))?"READ":"",\
122	FD_ISSET((so)->s, (w_fdset))?"WRITE":"", FD_ISSET((so)->s, (x_fdset))?"OOB":"")); \
123	} while (0)
124	# endif /* VBOX_WITH_DEBUG_NAT_SOCKETS */
125	#else
126	# define DO_LOG_NAT_SOCK(so, proto, winevent, r_fdset, w_fdset, x_fdset) do {} while (0)
127	#endif /* !VBOX_WITH_DEBUG_NAT_SOCKETS */
128
129	#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))
130
131	static const uint8_t special_ethaddr[6] =
132	{
133	0x52, 0x54, 0x00, 0x12, 0x35, 0x00
134	};
135
136	#ifdef RT_OS_WINDOWS
137
138	static int get_dns_addr_domain(PNATState pData, bool fVerbose,
139	struct in_addr *pdns_addr,
140	const char **ppszDomain)
141	{
142	int rc = 0;
143	FIXED_INFO *FixedInfo = NULL;
144	ULONG BufLen;
145	DWORD ret;
146	IP_ADDR_STRING *pIPAddr;
147	struct in_addr tmp_addr;
148
149	FixedInfo = (FIXED_INFO *)GlobalAlloc(GPTR, sizeof(FIXED_INFO));
150	BufLen = sizeof(FIXED_INFO);
151
152	/** @todo: this API returns all DNS servers, no matter whether the
153	* corresponding network adapter is disabled or not. Maybe replace
154	* this by GetAdapterAddresses(), which is XP/Vista only though. */
155	if (ERROR_BUFFER_OVERFLOW == GetNetworkParams(FixedInfo, &BufLen))
156	{
157	if (FixedInfo)
158	{
159	GlobalFree(FixedInfo);
160	FixedInfo = NULL;
161	}
162	FixedInfo = GlobalAlloc(GPTR, BufLen);
163	}
164
165	if ((ret = GetNetworkParams(FixedInfo, &BufLen)) != ERROR_SUCCESS)
166	{
167	Log(("GetNetworkParams failed. ret = %08x\n", (u_int)ret ));
168	if (FixedInfo)
169	{
170	GlobalFree(FixedInfo);
171	FixedInfo = NULL;
172	}
173	rc = -1;
174	goto get_dns_prefix;
175	}
176
177	pIPAddr = &(FixedInfo->DnsServerList);
178	inet_aton(pIPAddr->IpAddress.String, &tmp_addr);
179	Log(("nat: DNS Servers:\n"));
180	if (fVerbose \|\| pdns_addr->s_addr != tmp_addr.s_addr)
181	LogRel(("NAT: DNS address: %s\n", pIPAddr->IpAddress.String));
182	*pdns_addr = tmp_addr;
183
184	pIPAddr = FixedInfo -> DnsServerList.Next;
185	while (pIPAddr)
186	{
187	if (fVerbose)
188	LogRel(("NAT: ignored DNS address: %s\n", pIPAddr ->IpAddress.String));
189	pIPAddr = pIPAddr ->Next;
190	}
191	if (FixedInfo)
192	{
193	GlobalFree(FixedInfo);
194	FixedInfo = NULL;
195	}
196
197	get_dns_prefix:
198	if (ppszDomain)
199	{
200	OSVERSIONINFO ver;
201	char szDnsDomain[256];
202	DWORD dwSize = sizeof(szDnsDomain);
203
204	*ppszDomain = NULL;
205	GetVersionEx(&ver);
206	if (ver.dwMajorVersion >= 5)
207	{
208	/* GetComputerNameEx exists in Windows versions starting with 2000. */
209	if (GetComputerNameEx(ComputerNameDnsDomain, szDnsDomain, &dwSize))
210	{
211	if (szDnsDomain[0])
212	{
213	/* Just non-empty strings are valid. */
214	*ppszDomain = RTStrDup(szDnsDomain);
215	if (pData->fPassDomain)
216	{
217	if (fVerbose)
218	LogRel(("NAT: passing domain name %s\n", szDnsDomain));
219	}
220	else
221	Log(("nat: ignoring domain %s\n", szDnsDomain));
222	}
223	}
224	else
225	Log(("nat: GetComputerNameEx failed (%d)\n", GetLastError()));
226	}
227	}
228	return rc;
229	}
230
231	#else
232
233	static int get_dns_addr_domain(PNATState pData, bool fVerbose,
234	struct in_addr *pdns_addr,
235	const char **ppszDomain)
236	{
237	char buff[512];
238	char buff2[256];
239	FILE *f;
240	int found = 0;
241	struct in_addr tmp_addr;
242
243	#ifdef RT_OS_OS2
244	/* Try various locations. */
245	char *etc = getenv("ETC");
246	f = NULL;
247	if (etc)
248	{
249	snprintf(buff, sizeof(buff), "%s/RESOLV2", etc);
250	f = fopen(buff, "rt");
251	}
252	if (!f)
253	{
254	snprintf(buff, sizeof(buff), "%s/RESOLV2", _PATH_ETC);
255	f = fopen(buff, "rt");
256	}
257	if (!f)
258	{
259	snprintf(buff, sizeof(buff), "%s/resolv.conf", _PATH_ETC);
260	f = fopen(buff, "rt");
261	}
262	#else
263	f = fopen("/etc/resolv.conf", "r");
264	#endif
265	if (!f)
266	return -1;
267
268	if (ppszDomain)
269	*ppszDomain = NULL;
270	Log(("nat: DNS Servers:\n"));
271	while (fgets(buff, 512, f) != NULL)
272	{
273	if (sscanf(buff, "nameserver%*[ \t]%256s", buff2) == 1)
274	{
275	if (!inet_aton(buff2, &tmp_addr))
276	continue;
277	if (tmp_addr.s_addr == loopback_addr.s_addr)
278	tmp_addr = our_addr;
279	/* If it's the first one, set it to dns_addr */
280	if (!found)
281	{
282	if (fVerbose \|\| pdns_addr->s_addr != tmp_addr.s_addr)
283	LogRel(("NAT: DNS address: %s\n", buff2));
284	*pdns_addr = tmp_addr;
285	}
286	else
287	{
288	if (fVerbose)
289	LogRel(("NAT: ignored DNS address: %s\n", buff2));
290	}
291	found++;
292	}
293	if ( ppszDomain
294	&& (!strncmp(buff, "domain", 6) \|\| !strncmp(buff, "search", 6)))
295	{
296	/* Domain name/search list present. Pick first entry */
297	if (*ppszDomain == NULL)
298	{
299	char *tok;
300	char *saveptr;
301	tok = strtok_r(&buff[6], " \t\n", &saveptr);
302	if (tok)
303	{
304	*ppszDomain = RTStrDup(tok);
305	if (pData->fPassDomain)
306	{
307	if (fVerbose)
308	LogRel(("NAT: passing domain name %s\n", tok));
309	}
310	else
311	Log(("nat: ignoring domain %s\n", tok));
312	}
313	}
314	}
315	}
316	fclose(f);
317	if (!found)
318	return -1;
319	return 0;
320	}
321
322	#endif
323
324	int get_dns_addr(PNATState pData, struct in_addr *pdns_addr)
325	{
326	return get_dns_addr_domain(pData, false, pdns_addr, NULL);
327	}
328
329	int slirp_init(PNATState ppData, const char pszNetAddr, uint32_t u32Netmask,
330	bool fPassDomain, const char *pszTFTPPrefix,
331	const char pszBootFile, void pvUser)
332	{
333	int fNATfailed = 0;
334	PNATState pData = malloc(sizeof(NATState));
335	*ppData = pData;
336	if (!pData)
337	return VERR_NO_MEMORY;
338	if (u32Netmask & 0x1f)
339	/* CTL is x.x.x.15, bootp passes up to 16 IPs (15..31) */
340	return VERR_INVALID_PARAMETER;
341	memset(pData, '\0', sizeof(NATState));
342	pData->fPassDomain = fPassDomain;
343	pData->pvUser = pvUser;
344	#if ARCH_BITS == 64 && !defined(VBOX_WITH_BSD_REASS)
345	pData->cpvHashUsed = 1;
346	#endif
347	tftp_prefix = pszTFTPPrefix;
348	bootp_filename = pszBootFile;
349	pData->netmask = u32Netmask;
350
351	#ifdef RT_OS_WINDOWS
352	{
353	WSADATA Data;
354	WSAStartup(MAKEWORD(2,0), &Data);
355	}
356	# if defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC)
357	pData->phEvents[VBOX_SOCKET_EVENT_INDEX] = CreateEvent(NULL, FALSE, FALSE, NULL);
358	# endif
359	#endif
360
361	link_up = 1;
362
363	debug_init();
364	if_init(pData);
365	ip_init(pData);
366	#ifdef VBOX_WITH_SLIRP_ICMP
367	icmp_init(pData);
368	#endif /* VBOX_WITH_SLIRP_ICMP */
369
370	/* Initialise mbufs after setting the MTU */
371	m_init(pData);
372
373	/* set default addresses */
374	inet_aton("127.0.0.1", &loopback_addr);
375	inet_aton("127.0.0.1", &dns_addr);
376
377	if (get_dns_addr_domain(pData, true, &dns_addr, &pData->pszDomain) < 0)
378	fNATfailed = 1;
379
380	inet_aton(pszNetAddr, &special_addr);
381	alias_addr.s_addr = special_addr.s_addr \| htonl(CTL_ALIAS);
382	getouraddr(pData);
383	return fNATfailed ? VINF_NAT_DNS : VINF_SUCCESS;
384	}
385
386	/**
387	* Statistics counters.
388	*/
389	void slirp_register_timers(PNATState pData, PPDMDRVINS pDrvIns)
390	{
391	#ifndef VBOX_WITHOUT_RELEASE_STATISTICS
392	PDMDrvHlpSTAMRegisterF(pDrvIns, &pData->StatFill, STAMTYPE_PROFILE, STAMVISIBILITY_ALWAYS,
393	STAMUNIT_TICKS_PER_CALL, "Profiling slirp fills", "/Drivers/NAT%d/Fill", pDrvIns->iInstance);
394	PDMDrvHlpSTAMRegisterF(pDrvIns, &pData->StatPoll, STAMTYPE_PROFILE, STAMVISIBILITY_ALWAYS,
395	STAMUNIT_TICKS_PER_CALL, "Profiling slirp polls", "/Drivers/NAT%d/Poll", pDrvIns->iInstance);
396	PDMDrvHlpSTAMRegisterF(pDrvIns, &pData->StatFastTimer, STAMTYPE_PROFILE, STAMVISIBILITY_ALWAYS,
397	STAMUNIT_TICKS_PER_CALL, "Profiling slirp fast timer", "/Drivers/NAT%d/TimerFast", pDrvIns->iInstance);
398	PDMDrvHlpSTAMRegisterF(pDrvIns, &pData->StatSlowTimer, STAMTYPE_PROFILE, STAMVISIBILITY_ALWAYS,
399	STAMUNIT_TICKS_PER_CALL, "Profiling slirp slow timer", "/Drivers/NAT%d/TimerSlow", pDrvIns->iInstance);
400	PDMDrvHlpSTAMRegisterF(pDrvIns, &pData->StatTCP, STAMTYPE_COUNTER, STAMVISIBILITY_ALWAYS,
401	STAMUNIT_COUNT, "TCP sockets", "/Drivers/NAT%d/SockTCP", pDrvIns->iInstance);
402	PDMDrvHlpSTAMRegisterF(pDrvIns, &pData->StatTCPHot, STAMTYPE_COUNTER, STAMVISIBILITY_ALWAYS,
403	STAMUNIT_COUNT, "TCP sockets active", "/Drivers/NAT%d/SockTCPHot", pDrvIns->iInstance);
404	PDMDrvHlpSTAMRegisterF(pDrvIns, &pData->StatUDP, STAMTYPE_COUNTER, STAMVISIBILITY_ALWAYS,
405	STAMUNIT_COUNT, "UDP sockets", "/Drivers/NAT%d/SockUDP", pDrvIns->iInstance);
406	PDMDrvHlpSTAMRegisterF(pDrvIns, &pData->StatUDPHot, STAMTYPE_COUNTER, STAMVISIBILITY_ALWAYS,
407	STAMUNIT_COUNT, "UDP sockets active", "/Drivers/NAT%d/SockUDPHot", pDrvIns->iInstance);
408	#endif /* VBOX_WITHOUT_RELEASE_STATISTICS */
409	}
410
411	/**
412	* Marks the link as up, making it possible to establish new connections.
413	*/
414	void slirp_link_up(PNATState pData)
415	{
416	link_up = 1;
417	}
418
419	/**
420	* Marks the link as down and cleans up the current connections.
421	*/
422	void slirp_link_down(PNATState pData)
423	{
424	struct socket *so;
425
426	while ((so = tcb.so_next) != &tcb)
427	{
428	if (so->so_state & SS_NOFDREF \|\| so->s == -1)
429	sofree(pData, so);
430	else
431	tcp_drop(pData, sototcpcb(so), 0);
432	}
433
434	while ((so = udb.so_next) != &udb)
435	udp_detach(pData, so);
436
437	link_up = 0;
438	}
439
440	/**
441	* Terminates the slirp component.
442	*/
443	void slirp_term(PNATState pData)
444	{
445	if (pData->pszDomain)
446	RTStrFree((char )(void )pData->pszDomain);
447
448	#if ARCH_BITS == 64 && !defined(VBOX_WITH_BSD_REASS)
449	LogRel(("NAT: cpvHashUsed=%RU32 cpvHashCollisions=%RU32 cpvHashInserts=%RU64 cpvHashDone=%RU64\n",
450	pData->cpvHashUsed, pData->cpvHashCollisions, pData->cpvHashInserts, pData->cpvHashDone));
451	#endif
452
453	#ifdef VBOX_WITH_SLIRP_ICMP
454	# ifdef RT_OS_WINDOWS
455	pData->pfIcmpCloseHandle(pData->icmp_socket.sh);
456	FreeLibrary(pData->hmIcmpLibrary);
457	free(pData->pvIcmpBuffer);
458	# else
459	closesocket(pData->icmp_socket.s);
460	# endif
461	#endif
462
463	slirp_link_down(pData);
464	#ifdef RT_OS_WINDOWS
465	WSACleanup();
466	#endif
467	#ifdef LOG_ENABLED
468	Log(("\n"
469	"NAT statistics\n"
470	"--------------\n"
471	"\n"));
472	ipstats(pData);
473	tcpstats(pData);
474	udpstats(pData);
475	icmpstats(pData);
476	mbufstats(pData);
477	sockstats(pData);
478	Log(("\n"
479	"\n"
480	"\n"));
481	#endif
482	free(pData);
483	}
484
485
486	#define CONN_CANFSEND(so) (((so)->so_state & (SS_FCANTSENDMORE\|SS_ISFCONNECTED)) == SS_ISFCONNECTED)
487	#define CONN_CANFRCV(so) (((so)->so_state & (SS_FCANTRCVMORE\|SS_ISFCONNECTED)) == SS_ISFCONNECTED)
488	#define UPD_NFDS(x) if (nfds < (x)) nfds = (x)
489
490	/*
491	* curtime kept to an accuracy of 1ms
492	*/
493	#ifdef RT_OS_WINDOWS
494	static void updtime(PNATState pData)
495	{
496	struct _timeb tb;
497
498	_ftime(&tb);
499	curtime = (u_int)tb.time * (u_int)1000;
500	curtime += (u_int)tb.millitm;
501	}
502	#else
503	static void updtime(PNATState pData)
504	{
505	gettimeofday(&tt, 0);
506
507	curtime = (u_int)tt.tv_sec * (u_int)1000;
508	curtime += (u_int)tt.tv_usec / (u_int)1000;
509
510	if ((tt.tv_usec % 1000) >= 500)
511	curtime++;
512	}
513	#endif
514
515	void slirp_select_fill(PNATState pData, int *pnfds,
516	fd_set readfds, fd_set writefds, fd_set *xfds)
517	{
518	struct socket so, so_next;
519	int nfds;
520	#if defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) && defined(RT_OS_WINDOWS)
521	int rc;
522	int error;
523	#endif
524	#ifdef VBOX_WITH_BSD_REASS
525	int i;
526	#endif /* VBOX_WITH_BSD_REASS */
527
528	STAM_REL_PROFILE_START(&pData->StatFill, a);
529
530	nfds = *pnfds;
531
532	/*
533	* First, TCP sockets
534	*/
535	do_slowtimo = 0;
536	if (link_up)
537	{
538	/*
539	* *_slowtimo needs calling if there are IP fragments
540	* in the fragment queue, or there are TCP connections active
541	*/
542	#ifndef VBOX_WITH_BSD_REASS
543	do_slowtimo = ((tcb.so_next != &tcb)
544	\|\| ((struct ipasfrag )&ipq != u32_to_ptr(pData, ipq.next, struct ipasfrag )));
545	#else /* !VBOX_WITH_BSD_REASS */
546	/* XXX:
547	* triggering of fragment expiration should be the same but use new macroses
548	*/
549	do_slowtimo = (tcb.so_next != &tcb);
550	if (!do_slowtimo)
551	{
552	for (i = 0; i < IPREASS_NHASH; i++)
553	{
554	if (!TAILQ_EMPTY(&ipq[i]))
555	{
556	do_slowtimo = 1;
557	break;
558	}
559	}
560	}
561	#endif /* VBOX_WITH_BSD_REASS */
562	ICMP_ENGAGE_EVENT(&pData->icmp_socket, readfds);
563
564	STAM_REL_COUNTER_RESET(&pData->StatTCP);
565	STAM_REL_COUNTER_RESET(&pData->StatTCPHot);
566
567	for (so = tcb.so_next; so != &tcb; so = so_next)
568	{
569	so_next = so->so_next;
570
571	STAM_REL_COUNTER_INC(&pData->StatTCP);
572
573	/*
574	* See if we need a tcp_fasttimo
575	*/
576	if (time_fasttimo == 0 && so->so_tcpcb->t_flags & TF_DELACK)
577	time_fasttimo = curtime; /* Flag when we want a fasttimo */
578
579	/*
580	* NOFDREF can include still connecting to local-host,
581	* newly socreated() sockets etc. Don't want to select these.
582	*/
583	if (so->so_state & SS_NOFDREF \|\| so->s == -1)
584	continue;
585
586	/*
587	* Set for reading sockets which are accepting
588	*/
589	if (so->so_state & SS_FACCEPTCONN)
590	{
591	STAM_REL_COUNTER_INC(&pData->StatTCPHot);
592	TCP_ENGAGE_EVENT1(so, readfds);
593	continue;
594	}
595
596	/*
597	* Set for writing sockets which are connecting
598	*/
599	if (so->so_state & SS_ISFCONNECTING)
600	{
601	STAM_REL_COUNTER_INC(&pData->StatTCPHot);
602	TCP_ENGAGE_EVENT1(so, writefds);
603	}
604
605	/*
606	* Set for writing if we are connected, can send more, and
607	* we have something to send
608	*/
609	if (CONN_CANFSEND(so) && so->so_rcv.sb_cc)
610	{
611	STAM_REL_COUNTER_INC(&pData->StatTCPHot);
612	TCP_ENGAGE_EVENT1(so, writefds);
613	}
614
615	/*
616	* Set for reading (and urgent data) if we are connected, can
617	* receive more, and we have room for it XXX /2 ?
618	*/
619	if (CONN_CANFRCV(so) && (so->so_snd.sb_cc < (so->so_snd.sb_datalen/2)))
620	{
621	STAM_REL_COUNTER_INC(&pData->StatTCPHot);
622	TCP_ENGAGE_EVENT2(so, readfds, xfds);
623	}
624	}
625
626	/*
627	* UDP sockets
628	*/
629	STAM_REL_COUNTER_RESET(&pData->StatUDP);
630	STAM_REL_COUNTER_RESET(&pData->StatUDPHot);
631
632	for (so = udb.so_next; so != &udb; so = so_next)
633	{
634	so_next = so->so_next;
635
636	STAM_REL_COUNTER_INC(&pData->StatUDP);
637
638	/*
639	* See if it's timed out
640	*/
641	if (so->so_expire)
642	{
643	if (so->so_expire <= curtime)
644	{
645	udp_detach(pData, so);
646	continue;
647	}
648	else
649	do_slowtimo = 1; /* Let socket expire */
650	}
651
652	/*
653	* When UDP packets are received from over the link, they're
654	* sendto()'d straight away, so no need for setting for writing
655	* Limit the number of packets queued by this session to 4.
656	* Note that even though we try and limit this to 4 packets,
657	* the session could have more queued if the packets needed
658	* to be fragmented.
659	*
660	* (XXX <= 4 ?)
661	*/
662	if ((so->so_state & SS_ISFCONNECTED) && so->so_queued <= 4)
663	{
664	STAM_REL_COUNTER_INC(&pData->StatUDPHot);
665	UDP_ENGAGE_EVENT(so, readfds);
666	}
667	}
668
669	}
670
671	#if !defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) \|\| !defined(RT_OS_WINDOWS)
672	*pnfds = nfds;
673	#else
674	*pnfds = VBOX_EVENT_COUNT;
675	#endif
676
677	STAM_REL_PROFILE_STOP(&pData->StatFill, a);
678	}
679
680	#if defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) && defined(RT_OS_WINDOWS)
681	void slirp_select_poll(PNATState pData, int fTimeout, int fIcmp)
682	#else
683	void slirp_select_poll(PNATState pData, fd_set readfds, fd_set writefds, fd_set *xfds)
684	#endif
685	{
686	struct socket so, so_next;
687	int ret;
688	#if defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) && defined(RT_OS_WINDOWS)
689	WSANETWORKEVENTS NetworkEvents;
690	int rc;
691	int error;
692	#endif
693
694	STAM_REL_PROFILE_START(&pData->StatPoll, a);
695
696	/* Update time */
697	updtime(pData);
698
699	/*
700	* See if anything has timed out
701	*/
702	if (link_up)
703	{
704	if (time_fasttimo && ((curtime - time_fasttimo) >= 2))
705	{
706	STAM_REL_PROFILE_START(&pData->StatFastTimer, a);
707	tcp_fasttimo(pData);
708	time_fasttimo = 0;
709	STAM_REL_PROFILE_STOP(&pData->StatFastTimer, a);
710	}
711	if (do_slowtimo && ((curtime - last_slowtimo) >= 499))
712	{
713	STAM_REL_PROFILE_START(&pData->StatSlowTimer, a);
714	ip_slowtimo(pData);
715	tcp_slowtimo(pData);
716	last_slowtimo = curtime;
717	STAM_REL_PROFILE_STOP(&pData->StatSlowTimer, a);
718	}
719	}
720	#if defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) && defined(RT_OS_WINDOWS)
721	if (fTimeout)
722	return; /* only timer update */
723	#endif
724
725	/*
726	* Check sockets
727	*/
728	if (link_up)
729	{
730	#if defined(VBOX_WITH_SLIRP_ICMP)
731	# if defined(RT_OS_WINDOWS)
732	if (fIcmp)
733	sorecvfrom(pData, &pData->icmp_socket);
734	# else
735	if (pData->icmp_socket.s != -1 && FD_ISSET(pData->icmp_socket.s, readfds))
736	sorecvfrom(pData, &pData->icmp_socket);
737	# endif
738	#endif
739	/*
740	* Check TCP sockets
741	*/
742	for (so = tcb.so_next; so != &tcb; so = so_next)
743	{
744	so_next = so->so_next;
745
746	/*
747	* FD_ISSET is meaningless on these sockets
748	* (and they can crash the program)
749	*/
750	if (so->so_state & SS_NOFDREF \|\| so->s == -1)
751	continue;
752
753	POLL_TCP_EVENTS(rc, error, so, &NetworkEvents);
754
755	LOG_NAT_SOCK(so, TCP, &NetworkEvents, readfds, writefds, xfds);
756
757	/*
758	* Check for URG data
759	* This will soread as well, so no need to
760	* test for readfds below if this succeeds
761	*/
762
763	/* out-of-band data */
764	if (CHECK_FD_SET(so, NetworkEvents, xfds))
765	{
766	sorecvoob(pData, so);
767	}
768
769	/*
770	* Check sockets for reading
771	*/
772	else if (CHECK_FD_SET(so, NetworkEvents, readfds))
773	{
774	/*
775	* Check for incoming connections
776	*/
777	if (so->so_state & SS_FACCEPTCONN)
778	{
779	tcp_connect(pData, so);
780	#if defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) && defined(RT_OS_WINDOWS)
781	if (!(NetworkEvents.lNetworkEvents & FD_CLOSE))
782	#endif
783	continue;
784	}
785
786	ret = soread(pData, so, /fCloseIfNothingRead=/false);
787	/* Output it if we read something */
788	if (ret > 0)
789	tcp_output(pData, sototcpcb(so));
790	}
791
792	#if defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) && defined(RT_OS_WINDOWS)
793	/*
794	* Check for FD_CLOSE events.
795	*/
796	if (NetworkEvents.lNetworkEvents & FD_CLOSE)
797	{
798	/*
799	* drain the socket
800	*/
801	for (;;)
802	{
803	ret = soread(pData, so, /fCloseIfNothingRead=/true);
804	if (ret > 0)
805	tcp_output(pData, sototcpcb(so));
806	else
807	break;
808	}
809	}
810	#endif
811
812	/*
813	* Check sockets for writing
814	*/
815	if (CHECK_FD_SET(so, NetworkEvents, writefds))
816	{
817	/*
818	* Check for non-blocking, still-connecting sockets
819	*/
820	if (so->so_state & SS_ISFCONNECTING)
821	{
822	/* Connected */
823	so->so_state &= ~SS_ISFCONNECTING;
824
825	/*
826	* This should be probably guarded by PROBE_CONN too. Anyway,
827	* we disable it on OS/2 because the below send call returns
828	* EFAULT which causes the opened TCP socket to close right
829	* after it has been opened and connected.
830	*/
831	#ifndef RT_OS_OS2
832	ret = send(so->s, (const char *)&ret, 0, 0);
833	if (ret < 0)
834	{
835	/* XXXXX Must fix, zero bytes is a NOP */
836	if ( errno == EAGAIN
837	\|\| errno == EWOULDBLOCK
838	\|\| errno == EINPROGRESS
839	\|\| errno == ENOTCONN)
840	continue;
841
842	/* else failed */
843	so->so_state = SS_NOFDREF;
844	}
845	/* else so->so_state &= ~SS_ISFCONNECTING; */
846	#endif
847
848	/*
849	* Continue tcp_input
850	*/
851	tcp_input(pData, (struct mbuf *)NULL, sizeof(struct ip), so);
852	/* continue; */
853	}
854	else
855	ret = sowrite(pData, so);
856	/*
857	* XXX If we wrote something (a lot), there could be the need
858	* for a window update. In the worst case, the remote will send
859	* a window probe to get things going again.
860	*/
861	}
862
863	/*
864	* Probe a still-connecting, non-blocking socket
865	* to check if it's still alive
866	*/
867	#ifdef PROBE_CONN
868	if (so->so_state & SS_ISFCONNECTING)
869	{
870	ret = recv(so->s, (char *)&ret, 0, 0);
871
872	if (ret < 0)
873	{
874	/* XXX */
875	if ( errno == EAGAIN
876	\|\| errno == EWOULDBLOCK
877	\|\| errno == EINPROGRESS
878	\|\| errno == ENOTCONN)
879	{
880	continue; /* Still connecting, continue */
881	}
882
883	/* else failed */
884	so->so_state = SS_NOFDREF;
885
886	/* tcp_input will take care of it */
887	}
888	else
889	{
890	ret = send(so->s, &ret, 0, 0);
891	if (ret < 0)
892	{
893	/* XXX */
894	if ( errno == EAGAIN
895	\|\| errno == EWOULDBLOCK
896	\|\| errno == EINPROGRESS
897	\|\| errno == ENOTCONN)
898	{
899	continue;
900	}
901	/* else failed */
902	so->so_state = SS_NOFDREF;
903	}
904	else
905	so->so_state &= ~SS_ISFCONNECTING;
906
907	}
908	tcp_input((struct mbuf *)NULL, sizeof(struct ip),so);
909	} /* SS_ISFCONNECTING */
910	#endif
911	}
912
913	/*
914	* Now UDP sockets.
915	* Incoming packets are sent straight away, they're not buffered.
916	* Incoming UDP data isn't buffered either.
917	*/
918	for (so = udb.so_next; so != &udb; so = so_next)
919	{
920	so_next = so->so_next;
921
922	POLL_UDP_EVENTS(rc, error, so, &NetworkEvents);
923
924	LOG_NAT_SOCK(so, UDP, &NetworkEvents, readfds, writefds, xfds);
925
926	if (so->s != -1 && CHECK_FD_SET(so, NetworkEvents, readfds))
927	{
928	sorecvfrom(pData, so);
929	}
930	}
931
932	#if 0
933	#if defined(VBOX_WITH_SLIRP_ICMP)
934	# if defined(RT_OS_WINDOWS)
935	if (fIcmp)
936	sorecvfrom(pData, &pData->icmp_socket);
937	# else
938	if (pData->icmp_socket.s != -1 && FD_ISSET(pData->icmp_socket.s, readfds))
939	sorecvfrom(pData, &pData->icmp_socket);
940	# endif
941	#endif
942	#endif
943	}
944
945	/*
946	* See if we can start outputting
947	*/
948	if (if_queued && link_up)
949	if_start(pData);
950
951	STAM_REL_PROFILE_STOP(&pData->StatPoll, a);
952	}
953
954	#define ETH_ALEN 6
955	#define ETH_HLEN 14
956
957	#define ETH_P_IP 0x0800 /* Internet Protocol packet */
958	#define ETH_P_ARP 0x0806 /* Address Resolution packet */
959
960	#define ARPOP_REQUEST 1 /* ARP request */
961	#define ARPOP_REPLY 2 /* ARP reply */
962
963	struct ethhdr
964	{
965	unsigned char h_dest[ETH_ALEN]; /* destination eth addr */
966	unsigned char h_source[ETH_ALEN]; /* source ether addr */
967	unsigned short h_proto; /* packet type ID field */
968	};
969
970	struct arphdr
971	{
972	unsigned short ar_hrd; /* format of hardware address */
973	unsigned short ar_pro; /* format of protocol address */
974	unsigned char ar_hln; /* length of hardware address */
975	unsigned char ar_pln; /* length of protocol address */
976	unsigned short ar_op; /* ARP opcode (command) */
977
978	/*
979	* Ethernet looks like this : This bit is variable sized however...
980	*/
981	unsigned char ar_sha[ETH_ALEN]; /* sender hardware address */
982	unsigned char ar_sip[4]; /* sender IP address */
983	unsigned char ar_tha[ETH_ALEN]; /* target hardware address */
984	unsigned char ar_tip[4]; /* target IP address */
985	};
986
987	static
988	void arp_input(PNATState pData, const uint8_t *pkt, int pkt_len)
989	{
990	struct ethhdr eh = (struct ethhdr )pkt;
991	struct arphdr ah = (struct arphdr )(pkt + ETH_HLEN);
992	uint8_t arp_reply[ETH_HLEN + sizeof(struct arphdr)];
993	struct ethhdr reh = (struct ethhdr )arp_reply;
994	struct arphdr rah = (struct arphdr )(arp_reply + ETH_HLEN);
995	int ar_op;
996	struct ex_list *ex_ptr;
997	uint32_t htip = ntohl((uint32_t)ah->ar_tip);
998
999	ar_op = ntohs(ah->ar_op);
1000	switch(ar_op)
1001	{
1002	case ARPOP_REQUEST:
1003	if ((htip & pData->netmask) == ntohl(special_addr.s_addr))
1004	{
1005	if ( (htip & ~pData->netmask) == CTL_DNS
1006	\|\| (htip & ~pData->netmask) == CTL_ALIAS)
1007	goto arp_ok;
1008	for (ex_ptr = exec_list; ex_ptr; ex_ptr = ex_ptr->ex_next)
1009	{
1010	if ((htip & ~pData->netmask) == ex_ptr->ex_addr)
1011	goto arp_ok;
1012	}
1013	return;
1014	arp_ok:
1015	/* XXX: make an ARP request to have the client address */
1016	memcpy(client_ethaddr, eh->h_source, ETH_ALEN);
1017
1018	/* ARP request for alias/dns mac address */
1019	memcpy(reh->h_dest, pkt + ETH_ALEN, ETH_ALEN);
1020	memcpy(reh->h_source, special_ethaddr, ETH_ALEN - 1);
1021	reh->h_source[5] = ah->ar_tip[3];
1022	reh->h_proto = htons(ETH_P_ARP);
1023
1024	rah->ar_hrd = htons(1);
1025	rah->ar_pro = htons(ETH_P_IP);
1026	rah->ar_hln = ETH_ALEN;
1027	rah->ar_pln = 4;
1028	rah->ar_op = htons(ARPOP_REPLY);
1029	memcpy(rah->ar_sha, reh->h_source, ETH_ALEN);
1030	memcpy(rah->ar_sip, ah->ar_tip, 4);
1031	memcpy(rah->ar_tha, ah->ar_sha, ETH_ALEN);
1032	memcpy(rah->ar_tip, ah->ar_sip, 4);
1033	slirp_output(pData->pvUser, arp_reply, sizeof(arp_reply));
1034	}
1035	break;
1036	default:
1037	break;
1038	}
1039	}
1040
1041	void slirp_input(PNATState pData, const uint8_t *pkt, int pkt_len)
1042	{
1043	struct mbuf *m;
1044	int proto;
1045
1046	if (pkt_len < ETH_HLEN)
1047	return;
1048
1049	proto = ntohs((uint16_t )(pkt + 12));
1050	switch(proto)
1051	{
1052	case ETH_P_ARP:
1053	arp_input(pData, pkt, pkt_len);
1054	break;
1055	case ETH_P_IP:
1056	/* Update time. Important if the network is very quiet, as otherwise
1057	* the first outgoing connection gets an incorrect timestamp. */
1058	updtime(pData);
1059
1060	m = m_get(pData);
1061	if (!m)
1062	return;
1063	/* Note: we add to align the IP header */
1064	if (M_FREEROOM(m) < pkt_len + 2)
1065	{
1066	m_inc(m, pkt_len + 2);
1067	}
1068	m->m_len = pkt_len + 2;
1069	memcpy(m->m_data + 2, pkt, pkt_len);
1070
1071	m->m_data += 2 + ETH_HLEN;
1072	m->m_len -= 2 + ETH_HLEN;
1073
1074	ip_input(pData, m);
1075	break;
1076	default:
1077	break;
1078	}
1079	}
1080
1081	/* output the IP packet to the ethernet device */
1082	void if_encap(PNATState pData, const uint8_t *ip_data, int ip_data_len)
1083	{
1084	uint8_t buf[1600];
1085	struct ethhdr eh = (struct ethhdr )buf;
1086
1087	if (ip_data_len + ETH_HLEN > sizeof(buf))
1088	return;
1089
1090	memcpy(eh->h_dest, client_ethaddr, ETH_ALEN);
1091	memcpy(eh->h_source, special_ethaddr, ETH_ALEN - 1);
1092	/* XXX: not correct */
1093	eh->h_source[5] = CTL_ALIAS;
1094	eh->h_proto = htons(ETH_P_IP);
1095	memcpy(buf + sizeof(struct ethhdr), ip_data, ip_data_len);
1096	slirp_output(pData->pvUser, buf, ip_data_len + ETH_HLEN);
1097	}
1098
1099	int slirp_redir(PNATState pData, int is_udp, int host_port,
1100	struct in_addr guest_addr, int guest_port)
1101	{
1102	if (is_udp)
1103	{
1104	if (!udp_listen(pData, htons(host_port), guest_addr.s_addr,
1105	htons(guest_port), 0))
1106	return -1;
1107	}
1108	else
1109	{
1110	if (!solisten(pData, htons(host_port), guest_addr.s_addr,
1111	htons(guest_port), 0))
1112	return -1;
1113	}
1114	return 0;
1115	}
1116
1117	int slirp_add_exec(PNATState pData, int do_pty, const char *args, int addr_low_byte,
1118	int guest_port)
1119	{
1120	return add_exec(&exec_list, do_pty, (char *)args,
1121	addr_low_byte, htons(guest_port));
1122	}
1123
1124	void slirp_set_ethaddr(PNATState pData, const uint8_t *ethaddr)
1125	{
1126	memcpy(client_ethaddr, ethaddr, ETH_ALEN);
1127	}
1128
1129	#if defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) && defined(RT_OS_WINDOWS)
1130	HANDLE *slirp_get_events(PNATState pData)
1131	{
1132	return pData->phEvents;
1133	}
1134	void slirp_register_external_event(PNATState pData, HANDLE hEvent, int index)
1135	{
1136	pData->phEvents[index] = hEvent;
1137	}
1138	#endif
1139
1140	unsigned int slirp_get_timeout_ms(PNATState pData)
1141	{
1142	if (link_up)
1143	{
1144	if (time_fasttimo)
1145	return 2;
1146	if (do_slowtimo)
1147	return 500; /* see PR_SLOWHZ */
1148	}
1149	return 0;
1150	}

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

source: vbox/trunk/src/VBox/Devices/Network/slirp/slirp.c@ 15388

Download in other formats: