slirp.c@ 16021

Last change on this file since 16021 was 15957, checked in by vboxsync, 16 years ago
build fix
Property svn:eol-style set to `native`
File size: 36.8 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
155	FixedInfo = (FIXED_INFO *)GlobalAlloc(GPTR, sizeof(FIXED_INFO));
156	BufLen = sizeof(FIXED_INFO);
157
158	/** @todo: this API returns all DNS servers, no matter whether the
159	* corresponding network adapter is disabled or not. Maybe replace
160	* this by GetAdapterAddresses(), which is XP/Vista only though. */
161	if (ERROR_BUFFER_OVERFLOW == GetNetworkParams(FixedInfo, &BufLen))
162	{
163	if (FixedInfo)
164	{
165	GlobalFree(FixedInfo);
166	FixedInfo = NULL;
167	}
168	FixedInfo = GlobalAlloc(GPTR, BufLen);
169	}
170
171	if ((ret = GetNetworkParams(FixedInfo, &BufLen)) != ERROR_SUCCESS)
172	{
173	Log(("GetNetworkParams failed. ret = %08x\n", (u_int)ret ));
174	if (FixedInfo)
175	{
176	GlobalFree(FixedInfo);
177	FixedInfo = NULL;
178	}
179	rc = -1;
180	goto get_dns_prefix;
181	}
182
183	pIPAddr = &(FixedInfo->DnsServerList);
184	inet_aton(pIPAddr->IpAddress.String, &tmp_addr);
185	Log(("nat: DNS Servers:\n"));
186	if (fVerbose \|\| pdns_addr->s_addr != tmp_addr.s_addr)
187	LogRel(("NAT: DNS address: %s\n", pIPAddr->IpAddress.String));
188	*pdns_addr = tmp_addr;
189
190	pIPAddr = FixedInfo -> DnsServerList.Next;
191	while (pIPAddr)
192	{
193	if (fVerbose)
194	LogRel(("NAT: ignored DNS address: %s\n", pIPAddr ->IpAddress.String));
195	pIPAddr = pIPAddr ->Next;
196	}
197	if (FixedInfo)
198	{
199	GlobalFree(FixedInfo);
200	FixedInfo = NULL;
201	}
202
203	get_dns_prefix:
204	if (ppszDomain)
205	{
206	OSVERSIONINFO ver;
207	char szDnsDomain[256];
208	DWORD dwSize = sizeof(szDnsDomain);
209
210	*ppszDomain = NULL;
211	GetVersionEx(&ver);
212	if (ver.dwMajorVersion >= 5)
213	{
214	/* GetComputerNameEx exists in Windows versions starting with 2000. */
215	if (GetComputerNameEx(ComputerNameDnsDomain, szDnsDomain, &dwSize))
216	{
217	if (szDnsDomain[0])
218	{
219	/* Just non-empty strings are valid. */
220	*ppszDomain = RTStrDup(szDnsDomain);
221	if (pData->fPassDomain)
222	{
223	if (fVerbose)
224	LogRel(("NAT: passing domain name %s\n", szDnsDomain));
225	}
226	else
227	Log(("nat: ignoring domain %s\n", szDnsDomain));
228	}
229	}
230	else
231	Log(("nat: GetComputerNameEx failed (%d)\n", GetLastError()));
232	}
233	}
234	return rc;
235	}
236
237	#else
238
239	static int get_dns_addr_domain(PNATState pData, bool fVerbose,
240	struct in_addr *pdns_addr,
241	const char **ppszDomain)
242	{
243	char buff[512];
244	char buff2[256];
245	FILE *f;
246	int found = 0;
247	struct in_addr tmp_addr;
248
249	#ifdef RT_OS_OS2
250	/* Try various locations. */
251	char *etc = getenv("ETC");
252	f = NULL;
253	if (etc)
254	{
255	snprintf(buff, sizeof(buff), "%s/RESOLV2", etc);
256	f = fopen(buff, "rt");
257	}
258	if (!f)
259	{
260	snprintf(buff, sizeof(buff), "%s/RESOLV2", _PATH_ETC);
261	f = fopen(buff, "rt");
262	}
263	if (!f)
264	{
265	snprintf(buff, sizeof(buff), "%s/resolv.conf", _PATH_ETC);
266	f = fopen(buff, "rt");
267	}
268	#else
269	f = fopen("/etc/resolv.conf", "r");
270	#endif
271	if (!f)
272	return -1;
273
274	if (ppszDomain)
275	*ppszDomain = NULL;
276	Log(("nat: DNS Servers:\n"));
277	while (fgets(buff, 512, f) != NULL)
278	{
279	if (sscanf(buff, "nameserver%*[ \t]%256s", buff2) == 1)
280	{
281	if (!inet_aton(buff2, &tmp_addr))
282	continue;
283	if (tmp_addr.s_addr == loopback_addr.s_addr)
284	tmp_addr = our_addr;
285	/* If it's the first one, set it to dns_addr */
286	if (!found)
287	{
288	if (fVerbose \|\| pdns_addr->s_addr != tmp_addr.s_addr)
289	LogRel(("NAT: DNS address: %s\n", buff2));
290	*pdns_addr = tmp_addr;
291	}
292	else
293	{
294	if (fVerbose)
295	LogRel(("NAT: ignored DNS address: %s\n", buff2));
296	}
297	found++;
298	}
299	if ( ppszDomain
300	&& (!strncmp(buff, "domain", 6) \|\| !strncmp(buff, "search", 6)))
301	{
302	/* Domain name/search list present. Pick first entry */
303	if (*ppszDomain == NULL)
304	{
305	char *tok;
306	char *saveptr;
307	tok = strtok_r(&buff[6], " \t\n", &saveptr);
308	if (tok)
309	{
310	*ppszDomain = RTStrDup(tok);
311	if (pData->fPassDomain)
312	{
313	if (fVerbose)
314	LogRel(("NAT: passing domain name %s\n", tok));
315	}
316	else
317	Log(("nat: ignoring domain %s\n", tok));
318	}
319	}
320	}
321	}
322	fclose(f);
323	if (!found)
324	return -1;
325	return 0;
326	}
327
328	#endif
329
330	int get_dns_addr(PNATState pData, struct in_addr *pdns_addr)
331	{
332	return get_dns_addr_domain(pData, false, pdns_addr, NULL);
333	}
334
335	int slirp_init(PNATState ppData, const char pszNetAddr, uint32_t u32Netmask,
336	bool fPassDomain, const char *pszTFTPPrefix,
337	const char pszBootFile, void pvUser)
338	{
339	int fNATfailed = 0;
340	PNATState pData = RTMemAlloc(sizeof(NATState));
341	*ppData = pData;
342	if (!pData)
343	return VERR_NO_MEMORY;
344	if (u32Netmask & 0x1f)
345	/* CTL is x.x.x.15, bootp passes up to 16 IPs (15..31) */
346	return VERR_INVALID_PARAMETER;
347	memset(pData, '\0', sizeof(NATState));
348	pData->fPassDomain = fPassDomain;
349	pData->pvUser = pvUser;
350	tftp_prefix = pszTFTPPrefix;
351	bootp_filename = pszBootFile;
352	pData->netmask = u32Netmask;
353
354	#ifdef RT_OS_WINDOWS
355	{
356	WSADATA Data;
357	WSAStartup(MAKEWORD(2,0), &Data);
358	}
359	# if defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC)
360	pData->phEvents[VBOX_SOCKET_EVENT_INDEX] = CreateEvent(NULL, FALSE, FALSE, NULL);
361	# endif
362	#endif
363
364	link_up = 1;
365
366	debug_init();
367	if_init(pData);
368	ip_init(pData);
369	icmp_init(pData);
370
371	/* Initialise mbufs after setting the MTU */
372	m_init(pData);
373
374	/* set default addresses */
375	inet_aton("127.0.0.1", &loopback_addr);
376	inet_aton("127.0.0.1", &dns_addr);
377
378	if (get_dns_addr_domain(pData, true, &dns_addr, &pData->pszDomain) < 0)
379	fNATfailed = 1;
380
381	inet_aton(pszNetAddr, &special_addr);
382	alias_addr.s_addr = special_addr.s_addr \| htonl(CTL_ALIAS);
383	getouraddr(pData);
384	return fNATfailed ? VINF_NAT_DNS : VINF_SUCCESS;
385	}
386
387	/**
388	* Statistics counters.
389	*/
390	void slirp_register_timers(PNATState pData, PPDMDRVINS pDrvIns)
391	{
392	#ifdef VBOX_WITH_STATISTICS
393	PDMDrvHlpSTAMRegisterF(pDrvIns, &pData->StatFill, STAMTYPE_PROFILE, STAMVISIBILITY_ALWAYS,
394	STAMUNIT_TICKS_PER_CALL, "Profiling slirp fills", "/Drivers/NAT%d/Fill", pDrvIns->iInstance);
395	PDMDrvHlpSTAMRegisterF(pDrvIns, &pData->StatPoll, STAMTYPE_PROFILE, STAMVISIBILITY_ALWAYS,
396	STAMUNIT_TICKS_PER_CALL, "Profiling slirp polls", "/Drivers/NAT%d/Poll", pDrvIns->iInstance);
397	PDMDrvHlpSTAMRegisterF(pDrvIns, &pData->StatFastTimer, STAMTYPE_PROFILE, STAMVISIBILITY_ALWAYS,
398	STAMUNIT_TICKS_PER_CALL, "Profiling slirp fast timer", "/Drivers/NAT%d/TimerFast", pDrvIns->iInstance);
399	PDMDrvHlpSTAMRegisterF(pDrvIns, &pData->StatSlowTimer, STAMTYPE_PROFILE, STAMVISIBILITY_ALWAYS,
400	STAMUNIT_TICKS_PER_CALL, "Profiling slirp slow timer", "/Drivers/NAT%d/TimerSlow", pDrvIns->iInstance);
401	PDMDrvHlpSTAMRegisterF(pDrvIns, &pData->StatTCP, STAMTYPE_COUNTER, STAMVISIBILITY_ALWAYS,
402	STAMUNIT_COUNT, "TCP sockets", "/Drivers/NAT%d/SockTCP", pDrvIns->iInstance);
403	PDMDrvHlpSTAMRegisterF(pDrvIns, &pData->StatTCPHot, STAMTYPE_COUNTER, STAMVISIBILITY_ALWAYS,
404	STAMUNIT_COUNT, "TCP sockets active", "/Drivers/NAT%d/SockTCPHot", pDrvIns->iInstance);
405	PDMDrvHlpSTAMRegisterF(pDrvIns, &pData->StatUDP, STAMTYPE_COUNTER, STAMVISIBILITY_ALWAYS,
406	STAMUNIT_COUNT, "UDP sockets", "/Drivers/NAT%d/SockUDP", pDrvIns->iInstance);
407	PDMDrvHlpSTAMRegisterF(pDrvIns, &pData->StatUDPHot, STAMTYPE_COUNTER, STAMVISIBILITY_ALWAYS,
408	STAMUNIT_COUNT, "UDP sockets active", "/Drivers/NAT%d/SockUDPHot", pDrvIns->iInstance);
409	#endif /* VBOX_WITH_STATISTICS */
410	}
411
412	/**
413	* Marks the link as up, making it possible to establish new connections.
414	*/
415	void slirp_link_up(PNATState pData)
416	{
417	link_up = 1;
418	}
419
420	/**
421	* Marks the link as down and cleans up the current connections.
422	*/
423	void slirp_link_down(PNATState pData)
424	{
425	struct socket *so;
426
427	while ((so = tcb.so_next) != &tcb)
428	{
429	if (so->so_state & SS_NOFDREF \|\| so->s == -1)
430	sofree(pData, so);
431	else
432	tcp_drop(pData, sototcpcb(so), 0);
433	}
434
435	while ((so = udb.so_next) != &udb)
436	udp_detach(pData, so);
437
438	link_up = 0;
439	}
440
441	/**
442	* Terminates the slirp component.
443	*/
444	void slirp_term(PNATState pData)
445	{
446	if (pData->pszDomain)
447	RTStrFree((char )(void )pData->pszDomain);
448
449	#ifdef RT_OS_WINDOWS
450	pData->pfIcmpCloseHandle(pData->icmp_socket.sh);
451	FreeLibrary(pData->hmIcmpLibrary);
452	RTMemFree(pData->pvIcmpBuffer);
453	# else
454	closesocket(pData->icmp_socket.s);
455	#endif
456
457	slirp_link_down(pData);
458	#ifdef RT_OS_WINDOWS
459	WSACleanup();
460	#endif
461	#ifdef LOG_ENABLED
462	Log(("\n"
463	"NAT statistics\n"
464	"--------------\n"
465	"\n"));
466	ipstats(pData);
467	tcpstats(pData);
468	udpstats(pData);
469	icmpstats(pData);
470	mbufstats(pData);
471	sockstats(pData);
472	Log(("\n"
473	"\n"
474	"\n"));
475	#endif
476	RTMemFree(pData);
477	}
478
479
480	#define CONN_CANFSEND(so) (((so)->so_state & (SS_FCANTSENDMORE\|SS_ISFCONNECTED)) == SS_ISFCONNECTED)
481	#define CONN_CANFRCV(so) (((so)->so_state & (SS_FCANTRCVMORE\|SS_ISFCONNECTED)) == SS_ISFCONNECTED)
482	#define UPD_NFDS(x) if (nfds < (x)) nfds = (x)
483
484	/*
485	* curtime kept to an accuracy of 1ms
486	*/
487	#ifdef RT_OS_WINDOWS
488	static void updtime(PNATState pData)
489	{
490	struct _timeb tb;
491
492	_ftime(&tb);
493	curtime = (u_int)tb.time * (u_int)1000;
494	curtime += (u_int)tb.millitm;
495	}
496	#else
497	static void updtime(PNATState pData)
498	{
499	gettimeofday(&tt, 0);
500
501	curtime = (u_int)tt.tv_sec * (u_int)1000;
502	curtime += (u_int)tt.tv_usec / (u_int)1000;
503
504	if ((tt.tv_usec % 1000) >= 500)
505	curtime++;
506	}
507	#endif
508
509	void slirp_select_fill(PNATState pData, int *pnfds,
510	fd_set readfds, fd_set writefds, fd_set *xfds)
511	{
512	struct socket so, so_next;
513	int nfds;
514	#if defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) && defined(RT_OS_WINDOWS)
515	int rc;
516	int error;
517	#endif
518	int i;
519
520	STAM_PROFILE_START(&pData->StatFill, a);
521
522	nfds = *pnfds;
523
524	/*
525	* First, TCP sockets
526	*/
527	do_slowtimo = 0;
528	if (link_up)
529	{
530	/*
531	* *_slowtimo needs calling if there are IP fragments
532	* in the fragment queue, or there are TCP connections active
533	*/
534	/* XXX:
535	* triggering of fragment expiration should be the same but use new macroses
536	*/
537	do_slowtimo = (tcb.so_next != &tcb);
538	if (!do_slowtimo)
539	{
540	for (i = 0; i < IPREASS_NHASH; i++)
541	{
542	if (!TAILQ_EMPTY(&ipq[i]))
543	{
544	do_slowtimo = 1;
545	break;
546	}
547	}
548	}
549	ICMP_ENGAGE_EVENT(&pData->icmp_socket, readfds);
550
551	STAM_COUNTER_RESET(&pData->StatTCP);
552	STAM_COUNTER_RESET(&pData->StatTCPHot);
553
554	for (so = tcb.so_next; so != &tcb; so = so_next)
555	{
556	so_next = so->so_next;
557
558	STAM_COUNTER_INC(&pData->StatTCP);
559
560	/*
561	* See if we need a tcp_fasttimo
562	*/
563	if (time_fasttimo == 0 && so->so_tcpcb->t_flags & TF_DELACK)
564	time_fasttimo = curtime; /* Flag when we want a fasttimo */
565
566	/*
567	* NOFDREF can include still connecting to local-host,
568	* newly socreated() sockets etc. Don't want to select these.
569	*/
570	if (so->so_state & SS_NOFDREF \|\| so->s == -1)
571	continue;
572
573	/*
574	* Set for reading sockets which are accepting
575	*/
576	if (so->so_state & SS_FACCEPTCONN)
577	{
578	STAM_COUNTER_INC(&pData->StatTCPHot);
579	TCP_ENGAGE_EVENT1(so, readfds);
580	continue;
581	}
582
583	/*
584	* Set for writing sockets which are connecting
585	*/
586	if (so->so_state & SS_ISFCONNECTING)
587	{
588	Log2(("connecting %R[natsock] engaged\n",so));
589	STAM_COUNTER_INC(&pData->StatTCPHot);
590	TCP_ENGAGE_EVENT1(so, writefds);
591	}
592
593	/*
594	* Set for writing if we are connected, can send more, and
595	* we have something to send
596	*/
597	if (CONN_CANFSEND(so) && so->so_rcv.sb_cc)
598	{
599	STAM_COUNTER_INC(&pData->StatTCPHot);
600	TCP_ENGAGE_EVENT1(so, writefds);
601	}
602
603	/*
604	* Set for reading (and urgent data) if we are connected, can
605	* receive more, and we have room for it XXX /2 ?
606	*/
607	if (CONN_CANFRCV(so) && (so->so_snd.sb_cc < (so->so_snd.sb_datalen/2)))
608	{
609	STAM_COUNTER_INC(&pData->StatTCPHot);
610	TCP_ENGAGE_EVENT2(so, readfds, xfds);
611	}
612	}
613
614	/*
615	* UDP sockets
616	*/
617	STAM_COUNTER_RESET(&pData->StatUDP);
618	STAM_COUNTER_RESET(&pData->StatUDPHot);
619
620	for (so = udb.so_next; so != &udb; so = so_next)
621	{
622	so_next = so->so_next;
623
624	STAM_COUNTER_INC(&pData->StatUDP);
625
626	/*
627	* See if it's timed out
628	*/
629	if (so->so_expire)
630	{
631	if (so->so_expire <= curtime)
632	{
633	udp_detach(pData, so);
634	continue;
635	}
636	else
637	do_slowtimo = 1; /* Let socket expire */
638	}
639
640	/*
641	* When UDP packets are received from over the link, they're
642	* sendto()'d straight away, so no need for setting for writing
643	* Limit the number of packets queued by this session to 4.
644	* Note that even though we try and limit this to 4 packets,
645	* the session could have more queued if the packets needed
646	* to be fragmented.
647	*
648	* (XXX <= 4 ?)
649	*/
650	if ((so->so_state & SS_ISFCONNECTED) && so->so_queued <= 4)
651	{
652	STAM_COUNTER_INC(&pData->StatUDPHot);
653	UDP_ENGAGE_EVENT(so, readfds);
654	}
655	}
656
657	}
658
659	#if !defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) \|\| !defined(RT_OS_WINDOWS)
660	*pnfds = nfds;
661	#else
662	*pnfds = VBOX_EVENT_COUNT;
663	#endif
664
665	STAM_PROFILE_STOP(&pData->StatFill, a);
666	}
667
668	#if defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) && defined(RT_OS_WINDOWS)
669	void slirp_select_poll(PNATState pData, int fTimeout, int fIcmp)
670	#else
671	void slirp_select_poll(PNATState pData, fd_set readfds, fd_set writefds, fd_set *xfds)
672	#endif
673	{
674	struct socket so, so_next;
675	int ret;
676	#if defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) && defined(RT_OS_WINDOWS)
677	WSANETWORKEVENTS NetworkEvents;
678	int rc;
679	int error;
680	#endif
681
682	STAM_PROFILE_START(&pData->StatPoll, a);
683
684	/* Update time */
685	updtime(pData);
686
687	/*
688	* See if anything has timed out
689	*/
690	if (link_up)
691	{
692	if (time_fasttimo && ((curtime - time_fasttimo) >= 2))
693	{
694	STAM_PROFILE_START(&pData->StatFastTimer, a);
695	tcp_fasttimo(pData);
696	time_fasttimo = 0;
697	STAM_PROFILE_STOP(&pData->StatFastTimer, a);
698	}
699	if (do_slowtimo && ((curtime - last_slowtimo) >= 499))
700	{
701	STAM_PROFILE_START(&pData->StatSlowTimer, a);
702	ip_slowtimo(pData);
703	tcp_slowtimo(pData);
704	last_slowtimo = curtime;
705	STAM_PROFILE_STOP(&pData->StatSlowTimer, a);
706	}
707	}
708	#if defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) && defined(RT_OS_WINDOWS)
709	if (fTimeout)
710	return; /* only timer update */
711	#endif
712
713	/*
714	* Check sockets
715	*/
716	if (link_up)
717	{
718	#if defined(RT_OS_WINDOWS)
719	/*XXX: before renaming please make see define
720	* fIcmp in slirp_state.h
721	*/
722	if (fIcmp)
723	sorecvfrom(pData, &pData->icmp_socket);
724	#else
725	if (pData->icmp_socket.s != -1 && FD_ISSET(pData->icmp_socket.s, readfds))
726	sorecvfrom(pData, &pData->icmp_socket);
727	#endif
728	/*
729	* Check TCP sockets
730	*/
731	for (so = tcb.so_next; so != &tcb; so = so_next)
732	{
733	so_next = so->so_next;
734
735	/*
736	* FD_ISSET is meaningless on these sockets
737	* (and they can crash the program)
738	*/
739	if (so->so_state & SS_NOFDREF \|\| so->s == -1)
740	continue;
741
742	POLL_TCP_EVENTS(rc, error, so, &NetworkEvents);
743
744	LOG_NAT_SOCK(so, TCP, &NetworkEvents, readfds, writefds, xfds);
745
746	/*
747	* Check for URG data
748	* This will soread as well, so no need to
749	* test for readfds below if this succeeds
750	*/
751
752	/* out-of-band data */
753	if (CHECK_FD_SET(so, NetworkEvents, xfds))
754	{
755	sorecvoob(pData, so);
756	}
757
758	/*
759	* Check sockets for reading
760	*/
761	else if ( CHECK_FD_SET(so, NetworkEvents, readfds)
762	\|\| WIN_CHECK_FD_SET(so, NetworkEvents, acceptds))
763	{
764	/*
765	* Check for incoming connections
766	*/
767	if (so->so_state & SS_FACCEPTCONN)
768	{
769	tcp_connect(pData, so);
770	#if defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) && defined(RT_OS_WINDOWS)
771	if (!(NetworkEvents.lNetworkEvents & FD_CLOSE))
772	#endif
773	continue;
774	}
775
776	ret = soread(pData, so, /fCloseIfNothingRead=/false);
777	/* Output it if we read something */
778	if (ret > 0)
779	tcp_output(pData, sototcpcb(so));
780	}
781
782	#if defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) && defined(RT_OS_WINDOWS)
783	/*
784	* Check for FD_CLOSE events.
785	*/
786	if (NetworkEvents.lNetworkEvents & FD_CLOSE)
787	{
788	/*
789	* drain the socket
790	*/
791	for (;;)
792	{
793	ret = soread(pData, so, /fCloseIfNothingRead=/true);
794	if (ret > 0)
795	tcp_output(pData, sototcpcb(so));
796	else
797	break;
798	}
799	}
800	#endif
801
802	/*
803	* Check sockets for writing
804	*/
805	if (CHECK_FD_SET(so, NetworkEvents, writefds))
806	{
807	/*
808	* Check for non-blocking, still-connecting sockets
809	*/
810	if (so->so_state & SS_ISFCONNECTING)
811	{
812	Log2(("connecting %R[natsock] catched\n", so));
813	/* Connected */
814	so->so_state &= ~SS_ISFCONNECTING;
815
816	/*
817	* This should be probably guarded by PROBE_CONN too. Anyway,
818	* we disable it on OS/2 because the below send call returns
819	* EFAULT which causes the opened TCP socket to close right
820	* after it has been opened and connected.
821	*/
822	#ifndef RT_OS_OS2
823	ret = send(so->s, (const char *)&ret, 0, 0);
824	if (ret < 0)
825	{
826	/* XXXXX Must fix, zero bytes is a NOP */
827	if ( errno == EAGAIN
828	\|\| errno == EWOULDBLOCK
829	\|\| errno == EINPROGRESS
830	\|\| errno == ENOTCONN)
831	continue;
832
833	/* else failed */
834	so->so_state = SS_NOFDREF;
835	}
836	/* else so->so_state &= ~SS_ISFCONNECTING; */
837	#endif
838
839	/*
840	* Continue tcp_input
841	*/
842	tcp_input(pData, (struct mbuf *)NULL, sizeof(struct ip), so);
843	/* continue; */
844	}
845	else
846	ret = sowrite(pData, so);
847	/*
848	* XXX If we wrote something (a lot), there could be the need
849	* for a window update. In the worst case, the remote will send
850	* a window probe to get things going again.
851	*/
852	}
853
854	/*
855	* Probe a still-connecting, non-blocking socket
856	* to check if it's still alive
857	*/
858	#ifdef PROBE_CONN
859	if (so->so_state & SS_ISFCONNECTING)
860	{
861	ret = recv(so->s, (char *)&ret, 0, 0);
862
863	if (ret < 0)
864	{
865	/* XXX */
866	if ( errno == EAGAIN
867	\|\| errno == EWOULDBLOCK
868	\|\| errno == EINPROGRESS
869	\|\| errno == ENOTCONN)
870	{
871	continue; /* Still connecting, continue */
872	}
873
874	/* else failed */
875	so->so_state = SS_NOFDREF;
876
877	/* tcp_input will take care of it */
878	}
879	else
880	{
881	ret = send(so->s, &ret, 0, 0);
882	if (ret < 0)
883	{
884	/* XXX */
885	if ( errno == EAGAIN
886	\|\| errno == EWOULDBLOCK
887	\|\| errno == EINPROGRESS
888	\|\| errno == ENOTCONN)
889	{
890	continue;
891	}
892	/* else failed */
893	so->so_state = SS_NOFDREF;
894	}
895	else
896	so->so_state &= ~SS_ISFCONNECTING;
897
898	}
899	tcp_input((struct mbuf *)NULL, sizeof(struct ip),so);
900	} /* SS_ISFCONNECTING */
901	#endif
902	}
903
904	/*
905	* Now UDP sockets.
906	* Incoming packets are sent straight away, they're not buffered.
907	* Incoming UDP data isn't buffered either.
908	*/
909	for (so = udb.so_next; so != &udb; so = so_next)
910	{
911	so_next = so->so_next;
912
913	POLL_UDP_EVENTS(rc, error, so, &NetworkEvents);
914
915	LOG_NAT_SOCK(so, UDP, &NetworkEvents, readfds, writefds, xfds);
916
917	if (so->s != -1 && CHECK_FD_SET(so, NetworkEvents, readfds))
918	{
919	sorecvfrom(pData, so);
920	}
921	}
922
923	}
924
925	/*
926	* See if we can start outputting
927	*/
928	if (if_queued && link_up)
929	if_start(pData);
930
931	STAM_PROFILE_STOP(&pData->StatPoll, a);
932	}
933
934	#define ETH_ALEN 6
935	#define ETH_HLEN 14
936
937	#define ARPOP_REQUEST 1 /* ARP request */
938	#define ARPOP_REPLY 2 /* ARP reply */
939
940	struct ethhdr
941	{
942	unsigned char h_dest[ETH_ALEN]; /* destination eth addr */
943	unsigned char h_source[ETH_ALEN]; /* source ether addr */
944	unsigned short h_proto; /* packet type ID field */
945	};
946
947	struct arphdr
948	{
949	unsigned short ar_hrd; /* format of hardware address */
950	unsigned short ar_pro; /* format of protocol address */
951	unsigned char ar_hln; /* length of hardware address */
952	unsigned char ar_pln; /* length of protocol address */
953	unsigned short ar_op; /* ARP opcode (command) */
954
955	/*
956	* Ethernet looks like this : This bit is variable sized however...
957	*/
958	unsigned char ar_sha[ETH_ALEN]; /* sender hardware address */
959	unsigned char ar_sip[4]; /* sender IP address */
960	unsigned char ar_tha[ETH_ALEN]; /* target hardware address */
961	unsigned char ar_tip[4]; /* target IP address */
962	};
963
964	static
965	#ifdef VBOX_WITH_SIMPLIFIED_SLIRP_SYNC
966	void arp_input(PNATState pData, struct mbuf *m)
967	#else
968	void arp_input(PNATState pData, const uint8_t *pkt, int pkt_len)
969	#endif
970	{
971	struct ethhdr *eh;
972	struct ethhdr *reh;
973	struct arphdr *ah;
974	struct arphdr *rah;
975	int ar_op;
976	struct ex_list *ex_ptr;
977	uint32_t htip;
978	#ifndef VBOX_WITH_SIMPLIFIED_SLIRP_SYNC
979	uint8_t arp_reply[sizeof(struct arphdr) + ETH_HLEN];
980	eh = (struct ethhdr *)pkt;
981	#else
982	struct mbuf *mr;
983	eh = mtod(m, struct ethhdr *);
984	#endif
985	ah = (struct arphdr *)&eh[1];
986	htip = ntohl((uint32_t)ah->ar_tip);
987
988	#ifdef VBOX_WITH_SIMPLIFIED_SLIRP_SYNC
989	mr = m_get(pData);
990	mr->m_data += if_maxlinkhdr;
991	mr->m_len = sizeof(struct arphdr);
992	rah = mtod(mr, struct arphdr *);
993	#else
994	reh = (struct ethhdr *)arp_reply;
995	rah = (struct arphdr *)&reh[1];
996	#endif
997
998	ar_op = ntohs(ah->ar_op);
999	switch(ar_op)
1000	{
1001	case ARPOP_REQUEST:
1002	if ((htip & pData->netmask) == ntohl(special_addr.s_addr))
1003	{
1004	if ( (htip & ~pData->netmask) == CTL_DNS
1005	\|\| (htip & ~pData->netmask) == CTL_ALIAS)
1006	goto arp_ok;
1007	for (ex_ptr = exec_list; ex_ptr; ex_ptr = ex_ptr->ex_next)
1008	{
1009	if ((htip & ~pData->netmask) == ex_ptr->ex_addr)
1010	goto arp_ok;
1011	}
1012	return;
1013	arp_ok:
1014
1015	#ifndef VBOX_WITH_SIMPLIFIED_SLIRP_SYNC
1016	memcpy(reh->h_dest, eh->h_source, ETH_ALEN);
1017	memcpy(reh->h_source, &special_addr, ETH_ALEN);
1018	reh->h_source[5] = ah->ar_tip[3];
1019	reh->h_proto = htons(ETH_P_ARP);
1020	#endif
1021	rah->ar_hrd = htons(1);
1022	rah->ar_pro = htons(ETH_P_IP);
1023	rah->ar_hln = ETH_ALEN;
1024	rah->ar_pln = 4;
1025	rah->ar_op = htons(ARPOP_REPLY);
1026	memcpy(rah->ar_sha, special_ethaddr, ETH_ALEN);
1027
1028	switch (htip & ~pData->netmask)
1029	{
1030	case CTL_DNS:
1031	case CTL_ALIAS:
1032	rah->ar_sha[5] = (uint8_t)(htip & ~pData->netmask);
1033	break;
1034	default:;
1035	}
1036
1037	memcpy(rah->ar_sip, ah->ar_tip, 4);
1038	memcpy(rah->ar_tha, ah->ar_sha, ETH_ALEN);
1039	memcpy(rah->ar_tip, ah->ar_sip, 4);
1040	#ifdef VBOX_WITH_SIMPLIFIED_SLIRP_SYNC
1041	if_encap(pData, ETH_P_ARP, mr);
1042	m_free(pData, m);
1043	#else
1044	slirp_output(pData->pvUser, arp_reply, sizeof(arp_reply));
1045	#endif
1046	}
1047	break;
1048	default:
1049	break;
1050	}
1051	}
1052
1053	void slirp_input(PNATState pData, const uint8_t *pkt, int pkt_len)
1054	{
1055	struct mbuf *m;
1056	int proto;
1057	static bool fWarnedIpv6;
1058
1059	if (pkt_len < ETH_HLEN)
1060	{
1061	LogRel(("NAT: packet having size %d has been ingnored\n", pkt_len));
1062	return;
1063	}
1064
1065	m = m_get(pData);
1066	if (!m)
1067	{
1068	LogRel(("can't allocate new mbuf\n"));
1069	return;
1070	}
1071
1072	/* Note: we add to align the IP header */
1073
1074	if (M_FREEROOM(m) < pkt_len)
1075	m_inc(m, pkt_len);
1076
1077	m->m_len = pkt_len ;
1078	memcpy(m->m_data, pkt, pkt_len);
1079
1080	proto = ntohs((uint16_t )(pkt + 12));
1081	switch(proto)
1082	{
1083	case ETH_P_ARP:
1084	#ifdef VBOX_WITH_SIMPLIFIED_SLIRP_SYNC
1085	arp_input(pData, m);
1086	#else
1087	arp_input(pData, pkt, pkt_len);
1088	m_free(pData, m);
1089	#endif
1090	break;
1091	case ETH_P_IP:
1092	/* Update time. Important if the network is very quiet, as otherwise
1093	* the first outgoing connection gets an incorrect timestamp. */
1094	updtime(pData);
1095	m->m_data += ETH_HLEN;
1096	m->m_len -= ETH_HLEN;
1097	ip_input(pData, m);
1098	break;
1099	case ETH_P_IPV6:
1100	m_free(pData, m);
1101	if (!fWarnedIpv6)
1102	{
1103	LogRel(("NAT: IPv6 not supported\n"));
1104	fWarnedIpv6 = true;
1105	}
1106	break;
1107	default:
1108	LogRel(("NAT: Unsupported protocol %x\n", proto));
1109	m_free(pData, m);
1110	break;
1111	}
1112	}
1113
1114	/* output the IP packet to the ethernet device */
1115	#ifdef VBOX_WITH_SIMPLIFIED_SLIRP_SYNC
1116	void if_encap(PNATState pData, uint16_t eth_proto, struct mbuf *m)
1117	#else
1118	void if_encap(PNATState pData, uint8_t *ip_data, int ip_data_len)
1119	#endif
1120	{
1121	#ifdef VBOX_WITH_SIMPLIFIED_SLIRP_SYNC
1122	struct ethhdr *eh;
1123	uint8_t *buf = RTMemAlloc(1600);
1124	m->m_data -= if_maxlinkhdr;
1125	m->m_len += ETH_HLEN;
1126	eh = mtod(m, struct ethhdr *);
1127	#else
1128	uint8_t buf[1600];
1129	struct ethhdr eh = (struct ethhdr )buf;
1130
1131	if (ip_data_len + ETH_HLEN > sizeof(buf))
1132	return;
1133
1134	memcpy(buf + sizeof(struct ethhdr), ip_data, ip_data_len);
1135	#endif
1136
1137
1138	memcpy(eh->h_dest, client_ethaddr, ETH_ALEN);
1139	memcpy(eh->h_source, special_ethaddr, ETH_ALEN - 1);
1140	/* XXX: not correct */
1141	eh->h_source[5] = CTL_ALIAS;
1142	#ifdef VBOX_WITH_SIMPLIFIED_SLIRP_SYNC
1143	eh->h_proto = htons(eth_proto);
1144	#if 0
1145	slirp_output(pData->pvUser, m, mtod(m, uint8_t *), m->m_len);
1146	#else
1147	memcpy(buf, mtod(m, uint8_t *), m->m_len);
1148	slirp_output(pData->pvUser, NULL, buf, m->m_len);
1149	m_free(pData, m);
1150	#endif
1151	#else
1152	eh->h_proto = htons(ETH_P_IP);
1153	slirp_output(pData->pvUser, buf, ip_data_len + ETH_HLEN);
1154	#endif
1155	}
1156
1157	int slirp_redir(PNATState pData, int is_udp, int host_port,
1158	struct in_addr guest_addr, int guest_port)
1159	{
1160	if (is_udp)
1161	{
1162	if (!udp_listen(pData, htons(host_port), guest_addr.s_addr,
1163	htons(guest_port), 0))
1164	return -1;
1165	}
1166	else
1167	{
1168	if (!solisten(pData, htons(host_port), guest_addr.s_addr,
1169	htons(guest_port), 0))
1170	return -1;
1171	}
1172	return 0;
1173	}
1174
1175	int slirp_add_exec(PNATState pData, int do_pty, const char *args, int addr_low_byte,
1176	int guest_port)
1177	{
1178	return add_exec(&exec_list, do_pty, (char *)args,
1179	addr_low_byte, htons(guest_port));
1180	}
1181
1182	void slirp_set_ethaddr(PNATState pData, const uint8_t *ethaddr)
1183	{
1184	memcpy(client_ethaddr, ethaddr, ETH_ALEN);
1185	}
1186
1187	#if defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) && defined(RT_OS_WINDOWS)
1188	HANDLE *slirp_get_events(PNATState pData)
1189	{
1190	return pData->phEvents;
1191	}
1192	void slirp_register_external_event(PNATState pData, HANDLE hEvent, int index)
1193	{
1194	pData->phEvents[index] = hEvent;
1195	}
1196	#endif
1197
1198	unsigned int slirp_get_timeout_ms(PNATState pData)
1199	{
1200	if (link_up)
1201	{
1202	if (time_fasttimo)
1203	return 2;
1204	if (do_slowtimo)
1205	return 500; /* see PR_SLOWHZ */
1206	}
1207	return 0;
1208	}
1209
1210	/*
1211	* this function called from NAT thread
1212	*/
1213	void slirp_post_sent(PNATState pData, void *pvArg)
1214	{
1215	struct socket *so = 0;
1216	struct tcpcb *tp = 0;
1217	struct mbuf m = (struct mbuf )pvArg;
1218	m_free(pData, m);
1219	}

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

source: vbox/trunk/src/VBox/Devices/Network/slirp/slirp.c@ 16021

Download in other formats: