slirp.c@ 15682

Last change on this file since 15682 was 15682, checked in by vboxsync, 16 years ago
slirp: tcp/ip port-forwarding fix
Property svn:eol-style set to `native`
File size: 34.9 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	# define DO_WIN_CHECK_FD_SET(so, events, fdset ) 0 /* specific for Windows Winsock API */
34
35	# define ICMP_ENGAGE_EVENT(so, fdset) \
36	do { \
37	if (pData->icmp_socket.s != -1) \
38	DO_ENGAGE_EVENT1((so), (fdset), ICMP); \
39	} while (0)
40
41	#else /* defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) && defined(RT_OS_WINDOWS) */
42
43	/*
44	* On Windows, we will be notified by IcmpSendEcho2() when the response arrives.
45	* So no call to WSAEventSelect necessary.
46	*/
47	# define ICMP_ENGAGE_EVENT(so, fdset) do {} while(0)
48
49	# define DO_ENGAGE_EVENT1(so, fdset1, label) \
50	do { \
51	rc = WSAEventSelect((so)->s, VBOX_SOCKET_EVENT, FD_ALL_EVENTS); \
52	if (rc == SOCKET_ERROR) \
53	{ \
54	/* This should not happen */ \
55	error = WSAGetLastError(); \
56	LogRel(("WSAEventSelector (" #label ") error %d (so=%x, socket=%s, event=%x)\n", \
57	error, (so), (so)->s, VBOX_SOCKET_EVENT)); \
58	} \
59	} while(0); \
60	continue
61
62	# define DO_ENGAGE_EVENT2(so, fdset1, fdset2, label) \
63	DO_ENGAGE_EVENT1((so), (fdset1), label)
64
65	# define DO_POLL_EVENTS(rc, error, so, events, label) \
66	(rc) = WSAEnumNetworkEvents((so)->s, VBOX_SOCKET_EVENT, (events)); \
67	if ((rc) == SOCKET_ERROR) \
68	{ \
69	(error) = WSAGetLastError(); \
70	LogRel(("WSAEnumNetworkEvents " #label " error %d\n", (error))); \
71	continue; \
72	}
73
74	# define acceptds_win FD_ACCEPT
75	# define acceptds_win_bit FD_ACCEPT_BIT
76
77	# define readfds_win FD_READ
78	# define readfds_win_bit FD_READ_BIT
79
80	# define writefds_win FD_WRITE
81	# define writefds_win_bit FD_WRITE_BIT
82
83	# define xfds_win FD_OOB
84	# define xfds_win_bit FD_OOB_BIT
85
86	# define DO_CHECK_FD_SET(so, events, fdset) \
87	(((events).lNetworkEvents & fdset ## _win) && ((events).iErrorCode[fdset ## _win_bit] == 0))
88
89	# define DO_WIN_CHECK_FD_SET(so, events, fdset ) DO_CHECK_FD_SET((so), (events), (fdset))
90
91	#endif /* defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) && defined(RT_OS_WINDOWS) */
92
93	#define TCP_ENGAGE_EVENT1(so, fdset) \
94	DO_ENGAGE_EVENT1((so), (fdset), TCP)
95
96	#define TCP_ENGAGE_EVENT2(so, fdset1, fdset2) \
97	DO_ENGAGE_EVENT2((so), (fdset1), (fdset2), TCP)
98
99	#define UDP_ENGAGE_EVENT(so, fdset) \
100	DO_ENGAGE_EVENT1((so), (fdset), UDP)
101
102	#define POLL_TCP_EVENTS(rc, error, so, events) \
103	DO_POLL_EVENTS((rc), (error), (so), (events), TCP)
104
105	#define POLL_UDP_EVENTS(rc, error, so, events) \
106	DO_POLL_EVENTS((rc), (error), (so), (events), UDP)
107
108	#define CHECK_FD_SET(so, events, set) \
109	(DO_CHECK_FD_SET((so), (events), set))
110
111	#define WIN_CHECK_FD_SET(so, events, set) \
112	(DO_WIN_CHECK_FD_SET((so), (events), set))
113
114	/*
115	* Loging macros
116	*/
117	#if VBOX_WITH_DEBUG_NAT_SOCKETS
118	# if defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) && defined(RT_OS_WINDOWS)
119	# define DO_LOG_NAT_SOCK(so, proto, winevent, r_fdset, w_fdset, x_fdset) \
120	do { \
121	LogRel((" " #proto "%R[natsock] %R[natwinnetevents]\n", (so), (winevent))); \
122	} while (0)
123	# else
124	# define DO_LOG_NAT_SOCK(so, proto, winevent, r_fdset, w_fdset, x_fdset) \
125	do { \
126	LogRel((" " #proto " %R[natsock] %s %s %s\n", (so), FD_ISSET((so)->s, (r_fdset))?"READ":"",\
127	FD_ISSET((so)->s, (w_fdset))?"WRITE":"", FD_ISSET((so)->s, (x_fdset))?"OOB":"")); \
128	} while (0)
129	# endif /* VBOX_WITH_DEBUG_NAT_SOCKETS */
130	#else
131	# define DO_LOG_NAT_SOCK(so, proto, winevent, r_fdset, w_fdset, x_fdset) do {} while (0)
132	#endif /* !VBOX_WITH_DEBUG_NAT_SOCKETS */
133
134	#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))
135
136	static const uint8_t special_ethaddr[6] =
137	{
138	0x52, 0x54, 0x00, 0x12, 0x35, 0x00
139	};
140
141	#ifdef RT_OS_WINDOWS
142
143	static int get_dns_addr_domain(PNATState pData, bool fVerbose,
144	struct in_addr *pdns_addr,
145	const char **ppszDomain)
146	{
147	int rc = 0;
148	FIXED_INFO *FixedInfo = NULL;
149	ULONG BufLen;
150	DWORD ret;
151	IP_ADDR_STRING *pIPAddr;
152	struct in_addr tmp_addr;
153
154	FixedInfo = (FIXED_INFO *)GlobalAlloc(GPTR, sizeof(FIXED_INFO));
155	BufLen = sizeof(FIXED_INFO);
156
157	/** @todo: this API returns all DNS servers, no matter whether the
158	* corresponding network adapter is disabled or not. Maybe replace
159	* this by GetAdapterAddresses(), which is XP/Vista only though. */
160	if (ERROR_BUFFER_OVERFLOW == GetNetworkParams(FixedInfo, &BufLen))
161	{
162	if (FixedInfo)
163	{
164	GlobalFree(FixedInfo);
165	FixedInfo = NULL;
166	}
167	FixedInfo = GlobalAlloc(GPTR, BufLen);
168	}
169
170	if ((ret = GetNetworkParams(FixedInfo, &BufLen)) != ERROR_SUCCESS)
171	{
172	Log(("GetNetworkParams failed. ret = %08x\n", (u_int)ret ));
173	if (FixedInfo)
174	{
175	GlobalFree(FixedInfo);
176	FixedInfo = NULL;
177	}
178	rc = -1;
179	goto get_dns_prefix;
180	}
181
182	pIPAddr = &(FixedInfo->DnsServerList);
183	inet_aton(pIPAddr->IpAddress.String, &tmp_addr);
184	Log(("nat: DNS Servers:\n"));
185	if (fVerbose \|\| pdns_addr->s_addr != tmp_addr.s_addr)
186	LogRel(("NAT: DNS address: %s\n", pIPAddr->IpAddress.String));
187	*pdns_addr = tmp_addr;
188
189	pIPAddr = FixedInfo -> DnsServerList.Next;
190	while (pIPAddr)
191	{
192	if (fVerbose)
193	LogRel(("NAT: ignored DNS address: %s\n", pIPAddr ->IpAddress.String));
194	pIPAddr = pIPAddr ->Next;
195	}
196	if (FixedInfo)
197	{
198	GlobalFree(FixedInfo);
199	FixedInfo = NULL;
200	}
201
202	get_dns_prefix:
203	if (ppszDomain)
204	{
205	OSVERSIONINFO ver;
206	char szDnsDomain[256];
207	DWORD dwSize = sizeof(szDnsDomain);
208
209	*ppszDomain = NULL;
210	GetVersionEx(&ver);
211	if (ver.dwMajorVersion >= 5)
212	{
213	/* GetComputerNameEx exists in Windows versions starting with 2000. */
214	if (GetComputerNameEx(ComputerNameDnsDomain, szDnsDomain, &dwSize))
215	{
216	if (szDnsDomain[0])
217	{
218	/* Just non-empty strings are valid. */
219	*ppszDomain = RTStrDup(szDnsDomain);
220	if (pData->fPassDomain)
221	{
222	if (fVerbose)
223	LogRel(("NAT: passing domain name %s\n", szDnsDomain));
224	}
225	else
226	Log(("nat: ignoring domain %s\n", szDnsDomain));
227	}
228	}
229	else
230	Log(("nat: GetComputerNameEx failed (%d)\n", GetLastError()));
231	}
232	}
233	return rc;
234	}
235
236	#else
237
238	static int get_dns_addr_domain(PNATState pData, bool fVerbose,
239	struct in_addr *pdns_addr,
240	const char **ppszDomain)
241	{
242	char buff[512];
243	char buff2[256];
244	FILE *f;
245	int found = 0;
246	struct in_addr tmp_addr;
247
248	#ifdef RT_OS_OS2
249	/* Try various locations. */
250	char *etc = getenv("ETC");
251	f = NULL;
252	if (etc)
253	{
254	snprintf(buff, sizeof(buff), "%s/RESOLV2", etc);
255	f = fopen(buff, "rt");
256	}
257	if (!f)
258	{
259	snprintf(buff, sizeof(buff), "%s/RESOLV2", _PATH_ETC);
260	f = fopen(buff, "rt");
261	}
262	if (!f)
263	{
264	snprintf(buff, sizeof(buff), "%s/resolv.conf", _PATH_ETC);
265	f = fopen(buff, "rt");
266	}
267	#else
268	f = fopen("/etc/resolv.conf", "r");
269	#endif
270	if (!f)
271	return -1;
272
273	if (ppszDomain)
274	*ppszDomain = NULL;
275	Log(("nat: DNS Servers:\n"));
276	while (fgets(buff, 512, f) != NULL)
277	{
278	if (sscanf(buff, "nameserver%*[ \t]%256s", buff2) == 1)
279	{
280	if (!inet_aton(buff2, &tmp_addr))
281	continue;
282	if (tmp_addr.s_addr == loopback_addr.s_addr)
283	tmp_addr = our_addr;
284	/* If it's the first one, set it to dns_addr */
285	if (!found)
286	{
287	if (fVerbose \|\| pdns_addr->s_addr != tmp_addr.s_addr)
288	LogRel(("NAT: DNS address: %s\n", buff2));
289	*pdns_addr = tmp_addr;
290	}
291	else
292	{
293	if (fVerbose)
294	LogRel(("NAT: ignored DNS address: %s\n", buff2));
295	}
296	found++;
297	}
298	if ( ppszDomain
299	&& (!strncmp(buff, "domain", 6) \|\| !strncmp(buff, "search", 6)))
300	{
301	/* Domain name/search list present. Pick first entry */
302	if (*ppszDomain == NULL)
303	{
304	char *tok;
305	char *saveptr;
306	tok = strtok_r(&buff[6], " \t\n", &saveptr);
307	if (tok)
308	{
309	*ppszDomain = RTStrDup(tok);
310	if (pData->fPassDomain)
311	{
312	if (fVerbose)
313	LogRel(("NAT: passing domain name %s\n", tok));
314	}
315	else
316	Log(("nat: ignoring domain %s\n", tok));
317	}
318	}
319	}
320	}
321	fclose(f);
322	if (!found)
323	return -1;
324	return 0;
325	}
326
327	#endif
328
329	int get_dns_addr(PNATState pData, struct in_addr *pdns_addr)
330	{
331	return get_dns_addr_domain(pData, false, pdns_addr, NULL);
332	}
333
334	int slirp_init(PNATState ppData, const char pszNetAddr, uint32_t u32Netmask,
335	bool fPassDomain, const char *pszTFTPPrefix,
336	const char pszBootFile, void pvUser)
337	{
338	int fNATfailed = 0;
339	PNATState pData = malloc(sizeof(NATState));
340	*ppData = pData;
341	if (!pData)
342	return VERR_NO_MEMORY;
343	if (u32Netmask & 0x1f)
344	/* CTL is x.x.x.15, bootp passes up to 16 IPs (15..31) */
345	return VERR_INVALID_PARAMETER;
346	memset(pData, '\0', sizeof(NATState));
347	pData->fPassDomain = fPassDomain;
348	pData->pvUser = pvUser;
349	tftp_prefix = pszTFTPPrefix;
350	bootp_filename = pszBootFile;
351	pData->netmask = u32Netmask;
352
353	#ifdef RT_OS_WINDOWS
354	{
355	WSADATA Data;
356	WSAStartup(MAKEWORD(2,0), &Data);
357	}
358	# if defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC)
359	pData->phEvents[VBOX_SOCKET_EVENT_INDEX] = CreateEvent(NULL, FALSE, FALSE, NULL);
360	# endif
361	#endif
362
363	link_up = 1;
364
365	debug_init();
366	if_init(pData);
367	ip_init(pData);
368	icmp_init(pData);
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	#ifdef RT_OS_WINDOWS
449	pData->pfIcmpCloseHandle(pData->icmp_socket.sh);
450	FreeLibrary(pData->hmIcmpLibrary);
451	free(pData->pvIcmpBuffer);
452	# else
453	closesocket(pData->icmp_socket.s);
454	#endif
455
456	slirp_link_down(pData);
457	#ifdef RT_OS_WINDOWS
458	WSACleanup();
459	#endif
460	#ifdef LOG_ENABLED
461	Log(("\n"
462	"NAT statistics\n"
463	"--------------\n"
464	"\n"));
465	ipstats(pData);
466	tcpstats(pData);
467	udpstats(pData);
468	icmpstats(pData);
469	mbufstats(pData);
470	sockstats(pData);
471	Log(("\n"
472	"\n"
473	"\n"));
474	#endif
475	free(pData);
476	}
477
478
479	#define CONN_CANFSEND(so) (((so)->so_state & (SS_FCANTSENDMORE\|SS_ISFCONNECTED)) == SS_ISFCONNECTED)
480	#define CONN_CANFRCV(so) (((so)->so_state & (SS_FCANTRCVMORE\|SS_ISFCONNECTED)) == SS_ISFCONNECTED)
481	#define UPD_NFDS(x) if (nfds < (x)) nfds = (x)
482
483	/*
484	* curtime kept to an accuracy of 1ms
485	*/
486	#ifdef RT_OS_WINDOWS
487	static void updtime(PNATState pData)
488	{
489	struct _timeb tb;
490
491	_ftime(&tb);
492	curtime = (u_int)tb.time * (u_int)1000;
493	curtime += (u_int)tb.millitm;
494	}
495	#else
496	static void updtime(PNATState pData)
497	{
498	gettimeofday(&tt, 0);
499
500	curtime = (u_int)tt.tv_sec * (u_int)1000;
501	curtime += (u_int)tt.tv_usec / (u_int)1000;
502
503	if ((tt.tv_usec % 1000) >= 500)
504	curtime++;
505	}
506	#endif
507
508	void slirp_select_fill(PNATState pData, int *pnfds,
509	fd_set readfds, fd_set writefds, fd_set *xfds)
510	{
511	struct socket so, so_next;
512	int nfds;
513	#if defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) && defined(RT_OS_WINDOWS)
514	int rc;
515	int error;
516	#endif
517	int i;
518
519	STAM_REL_PROFILE_START(&pData->StatFill, a);
520
521	nfds = *pnfds;
522
523	/*
524	* First, TCP sockets
525	*/
526	do_slowtimo = 0;
527	if (link_up)
528	{
529	/*
530	* *_slowtimo needs calling if there are IP fragments
531	* in the fragment queue, or there are TCP connections active
532	*/
533	/* XXX:
534	* triggering of fragment expiration should be the same but use new macroses
535	*/
536	do_slowtimo = (tcb.so_next != &tcb);
537	if (!do_slowtimo)
538	{
539	for (i = 0; i < IPREASS_NHASH; i++)
540	{
541	if (!TAILQ_EMPTY(&ipq[i]))
542	{
543	do_slowtimo = 1;
544	break;
545	}
546	}
547	}
548	ICMP_ENGAGE_EVENT(&pData->icmp_socket, readfds);
549
550	STAM_REL_COUNTER_RESET(&pData->StatTCP);
551	STAM_REL_COUNTER_RESET(&pData->StatTCPHot);
552
553	for (so = tcb.so_next; so != &tcb; so = so_next)
554	{
555	so_next = so->so_next;
556
557	STAM_REL_COUNTER_INC(&pData->StatTCP);
558
559	/*
560	* See if we need a tcp_fasttimo
561	*/
562	if (time_fasttimo == 0 && so->so_tcpcb->t_flags & TF_DELACK)
563	time_fasttimo = curtime; /* Flag when we want a fasttimo */
564
565	/*
566	* NOFDREF can include still connecting to local-host,
567	* newly socreated() sockets etc. Don't want to select these.
568	*/
569	if (so->so_state & SS_NOFDREF \|\| so->s == -1)
570	continue;
571
572	/*
573	* Set for reading sockets which are accepting
574	*/
575	if (so->so_state & SS_FACCEPTCONN)
576	{
577	STAM_REL_COUNTER_INC(&pData->StatTCPHot);
578	TCP_ENGAGE_EVENT1(so, readfds);
579	continue;
580	}
581
582	/*
583	* Set for writing sockets which are connecting
584	*/
585	if (so->so_state & SS_ISFCONNECTING)
586	{
587	STAM_REL_COUNTER_INC(&pData->StatTCPHot);
588	TCP_ENGAGE_EVENT1(so, writefds);
589	}
590
591	/*
592	* Set for writing if we are connected, can send more, and
593	* we have something to send
594	*/
595	if (CONN_CANFSEND(so) && so->so_rcv.sb_cc)
596	{
597	STAM_REL_COUNTER_INC(&pData->StatTCPHot);
598	TCP_ENGAGE_EVENT1(so, writefds);
599	}
600
601	/*
602	* Set for reading (and urgent data) if we are connected, can
603	* receive more, and we have room for it XXX /2 ?
604	*/
605	if (CONN_CANFRCV(so) && (so->so_snd.sb_cc < (so->so_snd.sb_datalen/2)))
606	{
607	STAM_REL_COUNTER_INC(&pData->StatTCPHot);
608	TCP_ENGAGE_EVENT2(so, readfds, xfds);
609	}
610	}
611
612	/*
613	* UDP sockets
614	*/
615	STAM_REL_COUNTER_RESET(&pData->StatUDP);
616	STAM_REL_COUNTER_RESET(&pData->StatUDPHot);
617
618	for (so = udb.so_next; so != &udb; so = so_next)
619	{
620	so_next = so->so_next;
621
622	STAM_REL_COUNTER_INC(&pData->StatUDP);
623
624	/*
625	* See if it's timed out
626	*/
627	if (so->so_expire)
628	{
629	if (so->so_expire <= curtime)
630	{
631	udp_detach(pData, so);
632	continue;
633	}
634	else
635	do_slowtimo = 1; /* Let socket expire */
636	}
637
638	/*
639	* When UDP packets are received from over the link, they're
640	* sendto()'d straight away, so no need for setting for writing
641	* Limit the number of packets queued by this session to 4.
642	* Note that even though we try and limit this to 4 packets,
643	* the session could have more queued if the packets needed
644	* to be fragmented.
645	*
646	* (XXX <= 4 ?)
647	*/
648	if ((so->so_state & SS_ISFCONNECTED) && so->so_queued <= 4)
649	{
650	STAM_REL_COUNTER_INC(&pData->StatUDPHot);
651	UDP_ENGAGE_EVENT(so, readfds);
652	}
653	}
654
655	}
656
657	#if !defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) \|\| !defined(RT_OS_WINDOWS)
658	*pnfds = nfds;
659	#else
660	*pnfds = VBOX_EVENT_COUNT;
661	#endif
662
663	STAM_REL_PROFILE_STOP(&pData->StatFill, a);
664	}
665
666	#if defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) && defined(RT_OS_WINDOWS)
667	void slirp_select_poll(PNATState pData, int fTimeout, int fIcmp)
668	#else
669	void slirp_select_poll(PNATState pData, fd_set readfds, fd_set writefds, fd_set *xfds)
670	#endif
671	{
672	struct socket so, so_next;
673	int ret;
674	#if defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) && defined(RT_OS_WINDOWS)
675	WSANETWORKEVENTS NetworkEvents;
676	int rc;
677	int error;
678	#endif
679
680	STAM_REL_PROFILE_START(&pData->StatPoll, a);
681
682	/* Update time */
683	updtime(pData);
684
685	/*
686	* See if anything has timed out
687	*/
688	if (link_up)
689	{
690	if (time_fasttimo && ((curtime - time_fasttimo) >= 2))
691	{
692	STAM_REL_PROFILE_START(&pData->StatFastTimer, a);
693	tcp_fasttimo(pData);
694	time_fasttimo = 0;
695	STAM_REL_PROFILE_STOP(&pData->StatFastTimer, a);
696	}
697	if (do_slowtimo && ((curtime - last_slowtimo) >= 499))
698	{
699	STAM_REL_PROFILE_START(&pData->StatSlowTimer, a);
700	ip_slowtimo(pData);
701	tcp_slowtimo(pData);
702	last_slowtimo = curtime;
703	STAM_REL_PROFILE_STOP(&pData->StatSlowTimer, a);
704	}
705	}
706	#if defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) && defined(RT_OS_WINDOWS)
707	if (fTimeout)
708	return; /* only timer update */
709	#endif
710
711	/*
712	* Check sockets
713	*/
714	if (link_up)
715	{
716	#if defined(RT_OS_WINDOWS)
717	/*XXX: before renaming please make see define
718	* fIcmp in slirp_state.h
719	*/
720	if (fIcmp)
721	sorecvfrom(pData, &pData->icmp_socket);
722	#else
723	if (pData->icmp_socket.s != -1 && FD_ISSET(pData->icmp_socket.s, readfds))
724	sorecvfrom(pData, &pData->icmp_socket);
725	#endif
726	/*
727	* Check TCP sockets
728	*/
729	for (so = tcb.so_next; so != &tcb; so = so_next)
730	{
731	so_next = so->so_next;
732
733	/*
734	* FD_ISSET is meaningless on these sockets
735	* (and they can crash the program)
736	*/
737	if (so->so_state & SS_NOFDREF \|\| so->s == -1)
738	continue;
739
740	POLL_TCP_EVENTS(rc, error, so, &NetworkEvents);
741
742	LOG_NAT_SOCK(so, TCP, &NetworkEvents, readfds, writefds, xfds);
743
744	/*
745	* Check for URG data
746	* This will soread as well, so no need to
747	* test for readfds below if this succeeds
748	*/
749
750	/* out-of-band data */
751	if (CHECK_FD_SET(so, NetworkEvents, xfds))
752	{
753	sorecvoob(pData, so);
754	}
755
756	/*
757	* Check sockets for reading
758	*/
759	else if (CHECK_FD_SET(so, NetworkEvents, readfds)
760	\|\| WIN_CHECK_FD_SET(so, NetworkEvents, acceptds))
761	{
762	/*
763	* Check for incoming connections
764	*/
765	if (so->so_state & SS_FACCEPTCONN)
766	{
767	tcp_connect(pData, so);
768	#if defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) && defined(RT_OS_WINDOWS)
769	if (!(NetworkEvents.lNetworkEvents & FD_CLOSE))
770	#endif
771	continue;
772	}
773
774	ret = soread(pData, so, /fCloseIfNothingRead=/false);
775	/* Output it if we read something */
776	if (ret > 0)
777	tcp_output(pData, sototcpcb(so));
778	}
779
780	#if defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) && defined(RT_OS_WINDOWS)
781	/*
782	* Check for FD_CLOSE events.
783	*/
784	if (NetworkEvents.lNetworkEvents & FD_CLOSE)
785	{
786	/*
787	* drain the socket
788	*/
789	for (;;)
790	{
791	ret = soread(pData, so, /fCloseIfNothingRead=/true);
792	if (ret > 0)
793	tcp_output(pData, sototcpcb(so));
794	else
795	break;
796	}
797	}
798	#endif
799
800	/*
801	* Check sockets for writing
802	*/
803	if (CHECK_FD_SET(so, NetworkEvents, writefds))
804	{
805	/*
806	* Check for non-blocking, still-connecting sockets
807	*/
808	if (so->so_state & SS_ISFCONNECTING)
809	{
810	/* Connected */
811	so->so_state &= ~SS_ISFCONNECTING;
812
813	/*
814	* This should be probably guarded by PROBE_CONN too. Anyway,
815	* we disable it on OS/2 because the below send call returns
816	* EFAULT which causes the opened TCP socket to close right
817	* after it has been opened and connected.
818	*/
819	#ifndef RT_OS_OS2
820	ret = send(so->s, (const char *)&ret, 0, 0);
821	if (ret < 0)
822	{
823	/* XXXXX Must fix, zero bytes is a NOP */
824	if ( errno == EAGAIN
825	\|\| errno == EWOULDBLOCK
826	\|\| errno == EINPROGRESS
827	\|\| errno == ENOTCONN)
828	continue;
829
830	/* else failed */
831	so->so_state = SS_NOFDREF;
832	}
833	/* else so->so_state &= ~SS_ISFCONNECTING; */
834	#endif
835
836	/*
837	* Continue tcp_input
838	*/
839	tcp_input(pData, (struct mbuf *)NULL, sizeof(struct ip), so);
840	/* continue; */
841	}
842	else
843	ret = sowrite(pData, so);
844	/*
845	* XXX If we wrote something (a lot), there could be the need
846	* for a window update. In the worst case, the remote will send
847	* a window probe to get things going again.
848	*/
849	}
850
851	/*
852	* Probe a still-connecting, non-blocking socket
853	* to check if it's still alive
854	*/
855	#ifdef PROBE_CONN
856	if (so->so_state & SS_ISFCONNECTING)
857	{
858	ret = recv(so->s, (char *)&ret, 0, 0);
859
860	if (ret < 0)
861	{
862	/* XXX */
863	if ( errno == EAGAIN
864	\|\| errno == EWOULDBLOCK
865	\|\| errno == EINPROGRESS
866	\|\| errno == ENOTCONN)
867	{
868	continue; /* Still connecting, continue */
869	}
870
871	/* else failed */
872	so->so_state = SS_NOFDREF;
873
874	/* tcp_input will take care of it */
875	}
876	else
877	{
878	ret = send(so->s, &ret, 0, 0);
879	if (ret < 0)
880	{
881	/* XXX */
882	if ( errno == EAGAIN
883	\|\| errno == EWOULDBLOCK
884	\|\| errno == EINPROGRESS
885	\|\| errno == ENOTCONN)
886	{
887	continue;
888	}
889	/* else failed */
890	so->so_state = SS_NOFDREF;
891	}
892	else
893	so->so_state &= ~SS_ISFCONNECTING;
894
895	}
896	tcp_input((struct mbuf *)NULL, sizeof(struct ip),so);
897	} /* SS_ISFCONNECTING */
898	#endif
899	}
900
901	/*
902	* Now UDP sockets.
903	* Incoming packets are sent straight away, they're not buffered.
904	* Incoming UDP data isn't buffered either.
905	*/
906	for (so = udb.so_next; so != &udb; so = so_next)
907	{
908	so_next = so->so_next;
909
910	POLL_UDP_EVENTS(rc, error, so, &NetworkEvents);
911
912	LOG_NAT_SOCK(so, UDP, &NetworkEvents, readfds, writefds, xfds);
913
914	if (so->s != -1 && CHECK_FD_SET(so, NetworkEvents, readfds))
915	{
916	sorecvfrom(pData, so);
917	}
918	}
919
920	}
921
922	/*
923	* See if we can start outputting
924	*/
925	if (if_queued && link_up)
926	if_start(pData);
927
928	STAM_REL_PROFILE_STOP(&pData->StatPoll, a);
929	}
930
931	#define ETH_ALEN 6
932	#define ETH_HLEN 14
933
934	#define ETH_P_IP 0x0800 /* Internet Protocol packet */
935	#define ETH_P_ARP 0x0806 /* Address Resolution packet */
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	void arp_input(PNATState pData, const uint8_t *pkt, int pkt_len)
966	{
967	struct ethhdr eh = (struct ethhdr )pkt;
968	struct arphdr ah = (struct arphdr )(pkt + ETH_HLEN);
969	uint8_t arp_reply[ETH_HLEN + sizeof(struct arphdr)];
970	struct ethhdr reh = (struct ethhdr )arp_reply;
971	struct arphdr rah = (struct arphdr )(arp_reply + ETH_HLEN);
972	int ar_op;
973	struct ex_list *ex_ptr;
974	uint32_t htip = ntohl((uint32_t)ah->ar_tip);
975
976	ar_op = ntohs(ah->ar_op);
977	switch(ar_op)
978	{
979	case ARPOP_REQUEST:
980	if ((htip & pData->netmask) == ntohl(special_addr.s_addr))
981	{
982	if ( (htip & ~pData->netmask) == CTL_DNS
983	\|\| (htip & ~pData->netmask) == CTL_ALIAS)
984	goto arp_ok;
985	for (ex_ptr = exec_list; ex_ptr; ex_ptr = ex_ptr->ex_next)
986	{
987	if ((htip & ~pData->netmask) == ex_ptr->ex_addr)
988	goto arp_ok;
989	}
990	return;
991	arp_ok:
992	/* XXX: make an ARP request to have the client address */
993	memcpy(client_ethaddr, eh->h_source, ETH_ALEN);
994
995	/* ARP request for alias/dns mac address */
996	memcpy(reh->h_dest, pkt + ETH_ALEN, ETH_ALEN);
997	memcpy(reh->h_source, special_ethaddr, ETH_ALEN - 1);
998	reh->h_source[5] = ah->ar_tip[3];
999	reh->h_proto = htons(ETH_P_ARP);
1000
1001	rah->ar_hrd = htons(1);
1002	rah->ar_pro = htons(ETH_P_IP);
1003	rah->ar_hln = ETH_ALEN;
1004	rah->ar_pln = 4;
1005	rah->ar_op = htons(ARPOP_REPLY);
1006	memcpy(rah->ar_sha, reh->h_source, ETH_ALEN);
1007	memcpy(rah->ar_sip, ah->ar_tip, 4);
1008	memcpy(rah->ar_tha, ah->ar_sha, ETH_ALEN);
1009	memcpy(rah->ar_tip, ah->ar_sip, 4);
1010	slirp_output(pData->pvUser, arp_reply, sizeof(arp_reply));
1011	}
1012	break;
1013	default:
1014	break;
1015	}
1016	}
1017
1018	void slirp_input(PNATState pData, const uint8_t *pkt, int pkt_len)
1019	{
1020	struct mbuf *m;
1021	int proto;
1022
1023	if (pkt_len < ETH_HLEN)
1024	return;
1025
1026	proto = ntohs((uint16_t )(pkt + 12));
1027	switch(proto)
1028	{
1029	case ETH_P_ARP:
1030	arp_input(pData, pkt, pkt_len);
1031	break;
1032	case ETH_P_IP:
1033	/* Update time. Important if the network is very quiet, as otherwise
1034	* the first outgoing connection gets an incorrect timestamp. */
1035	updtime(pData);
1036
1037	m = m_get(pData);
1038	if (!m)
1039	return;
1040	/* Note: we add to align the IP header */
1041	if (M_FREEROOM(m) < pkt_len + 2)
1042	{
1043	m_inc(m, pkt_len + 2);
1044	}
1045	m->m_len = pkt_len + 2;
1046	memcpy(m->m_data + 2, pkt, pkt_len);
1047
1048	m->m_data += 2 + ETH_HLEN;
1049	m->m_len -= 2 + ETH_HLEN;
1050
1051	ip_input(pData, m);
1052	break;
1053	default:
1054	break;
1055	}
1056	}
1057
1058	/* output the IP packet to the ethernet device */
1059	void if_encap(PNATState pData, const uint8_t *ip_data, int ip_data_len)
1060	{
1061	uint8_t buf[1600];
1062	struct ethhdr eh = (struct ethhdr )buf;
1063
1064	if (ip_data_len + ETH_HLEN > sizeof(buf))
1065	return;
1066
1067	memcpy(eh->h_dest, client_ethaddr, ETH_ALEN);
1068	memcpy(eh->h_source, special_ethaddr, ETH_ALEN - 1);
1069	/* XXX: not correct */
1070	eh->h_source[5] = CTL_ALIAS;
1071	eh->h_proto = htons(ETH_P_IP);
1072	memcpy(buf + sizeof(struct ethhdr), ip_data, ip_data_len);
1073	slirp_output(pData->pvUser, buf, ip_data_len + ETH_HLEN);
1074	}
1075
1076	int slirp_redir(PNATState pData, int is_udp, int host_port,
1077	struct in_addr guest_addr, int guest_port)
1078	{
1079	if (is_udp)
1080	{
1081	if (!udp_listen(pData, htons(host_port), guest_addr.s_addr,
1082	htons(guest_port), 0))
1083	return -1;
1084	}
1085	else
1086	{
1087	if (!solisten(pData, htons(host_port), guest_addr.s_addr,
1088	htons(guest_port), 0))
1089	return -1;
1090	}
1091	return 0;
1092	}
1093
1094	int slirp_add_exec(PNATState pData, int do_pty, const char *args, int addr_low_byte,
1095	int guest_port)
1096	{
1097	return add_exec(&exec_list, do_pty, (char *)args,
1098	addr_low_byte, htons(guest_port));
1099	}
1100
1101	void slirp_set_ethaddr(PNATState pData, const uint8_t *ethaddr)
1102	{
1103	memcpy(client_ethaddr, ethaddr, ETH_ALEN);
1104	}
1105
1106	#if defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) && defined(RT_OS_WINDOWS)
1107	HANDLE *slirp_get_events(PNATState pData)
1108	{
1109	return pData->phEvents;
1110	}
1111	void slirp_register_external_event(PNATState pData, HANDLE hEvent, int index)
1112	{
1113	pData->phEvents[index] = hEvent;
1114	}
1115	#endif
1116
1117	unsigned int slirp_get_timeout_ms(PNATState pData)
1118	{
1119	if (link_up)
1120	{
1121	if (time_fasttimo)
1122	return 2;
1123	if (do_slowtimo)
1124	return 500; /* see PR_SLOWHZ */
1125	}
1126	return 0;
1127	}

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

source: vbox/trunk/src/VBox/Devices/Network/slirp/slirp.c@ 15682

Download in other formats: