slirp.c@ 15580

Last change on this file since 15580 was 15453, checked in by vboxsync, 16 years ago
slirp: removed the old 64-bit incompatible reassemble code
Property svn:eol-style set to `native`
File size: 35.0 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	tftp_prefix = pszTFTPPrefix;
345	bootp_filename = pszBootFile;
346	pData->netmask = u32Netmask;
347
348	#ifdef RT_OS_WINDOWS
349	{
350	WSADATA Data;
351	WSAStartup(MAKEWORD(2,0), &Data);
352	}
353	# if defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC)
354	pData->phEvents[VBOX_SOCKET_EVENT_INDEX] = CreateEvent(NULL, FALSE, FALSE, NULL);
355	# endif
356	#endif
357
358	link_up = 1;
359
360	debug_init();
361	if_init(pData);
362	ip_init(pData);
363	#ifdef VBOX_WITH_SLIRP_ICMP
364	icmp_init(pData);
365	#endif /* VBOX_WITH_SLIRP_ICMP */
366
367	/* Initialise mbufs after setting the MTU */
368	m_init(pData);
369
370	/* set default addresses */
371	inet_aton("127.0.0.1", &loopback_addr);
372	inet_aton("127.0.0.1", &dns_addr);
373
374	if (get_dns_addr_domain(pData, true, &dns_addr, &pData->pszDomain) < 0)
375	fNATfailed = 1;
376
377	inet_aton(pszNetAddr, &special_addr);
378	alias_addr.s_addr = special_addr.s_addr \| htonl(CTL_ALIAS);
379	getouraddr(pData);
380	return fNATfailed ? VINF_NAT_DNS : VINF_SUCCESS;
381	}
382
383	/**
384	* Statistics counters.
385	*/
386	void slirp_register_timers(PNATState pData, PPDMDRVINS pDrvIns)
387	{
388	#ifndef VBOX_WITHOUT_RELEASE_STATISTICS
389	PDMDrvHlpSTAMRegisterF(pDrvIns, &pData->StatFill, STAMTYPE_PROFILE, STAMVISIBILITY_ALWAYS,
390	STAMUNIT_TICKS_PER_CALL, "Profiling slirp fills", "/Drivers/NAT%d/Fill", pDrvIns->iInstance);
391	PDMDrvHlpSTAMRegisterF(pDrvIns, &pData->StatPoll, STAMTYPE_PROFILE, STAMVISIBILITY_ALWAYS,
392	STAMUNIT_TICKS_PER_CALL, "Profiling slirp polls", "/Drivers/NAT%d/Poll", pDrvIns->iInstance);
393	PDMDrvHlpSTAMRegisterF(pDrvIns, &pData->StatFastTimer, STAMTYPE_PROFILE, STAMVISIBILITY_ALWAYS,
394	STAMUNIT_TICKS_PER_CALL, "Profiling slirp fast timer", "/Drivers/NAT%d/TimerFast", pDrvIns->iInstance);
395	PDMDrvHlpSTAMRegisterF(pDrvIns, &pData->StatSlowTimer, STAMTYPE_PROFILE, STAMVISIBILITY_ALWAYS,
396	STAMUNIT_TICKS_PER_CALL, "Profiling slirp slow timer", "/Drivers/NAT%d/TimerSlow", pDrvIns->iInstance);
397	PDMDrvHlpSTAMRegisterF(pDrvIns, &pData->StatTCP, STAMTYPE_COUNTER, STAMVISIBILITY_ALWAYS,
398	STAMUNIT_COUNT, "TCP sockets", "/Drivers/NAT%d/SockTCP", pDrvIns->iInstance);
399	PDMDrvHlpSTAMRegisterF(pDrvIns, &pData->StatTCPHot, STAMTYPE_COUNTER, STAMVISIBILITY_ALWAYS,
400	STAMUNIT_COUNT, "TCP sockets active", "/Drivers/NAT%d/SockTCPHot", pDrvIns->iInstance);
401	PDMDrvHlpSTAMRegisterF(pDrvIns, &pData->StatUDP, STAMTYPE_COUNTER, STAMVISIBILITY_ALWAYS,
402	STAMUNIT_COUNT, "UDP sockets", "/Drivers/NAT%d/SockUDP", pDrvIns->iInstance);
403	PDMDrvHlpSTAMRegisterF(pDrvIns, &pData->StatUDPHot, STAMTYPE_COUNTER, STAMVISIBILITY_ALWAYS,
404	STAMUNIT_COUNT, "UDP sockets active", "/Drivers/NAT%d/SockUDPHot", pDrvIns->iInstance);
405	#endif /* VBOX_WITHOUT_RELEASE_STATISTICS */
406	}
407
408	/**
409	* Marks the link as up, making it possible to establish new connections.
410	*/
411	void slirp_link_up(PNATState pData)
412	{
413	link_up = 1;
414	}
415
416	/**
417	* Marks the link as down and cleans up the current connections.
418	*/
419	void slirp_link_down(PNATState pData)
420	{
421	struct socket *so;
422
423	while ((so = tcb.so_next) != &tcb)
424	{
425	if (so->so_state & SS_NOFDREF \|\| so->s == -1)
426	sofree(pData, so);
427	else
428	tcp_drop(pData, sototcpcb(so), 0);
429	}
430
431	while ((so = udb.so_next) != &udb)
432	udp_detach(pData, so);
433
434	link_up = 0;
435	}
436
437	/**
438	* Terminates the slirp component.
439	*/
440	void slirp_term(PNATState pData)
441	{
442	if (pData->pszDomain)
443	RTStrFree((char )(void )pData->pszDomain);
444
445	#ifdef VBOX_WITH_SLIRP_ICMP
446	# ifdef RT_OS_WINDOWS
447	pData->pfIcmpCloseHandle(pData->icmp_socket.sh);
448	FreeLibrary(pData->hmIcmpLibrary);
449	free(pData->pvIcmpBuffer);
450	# else
451	closesocket(pData->icmp_socket.s);
452	# endif
453	#endif
454
455	slirp_link_down(pData);
456	#ifdef RT_OS_WINDOWS
457	WSACleanup();
458	#endif
459	#ifdef LOG_ENABLED
460	Log(("\n"
461	"NAT statistics\n"
462	"--------------\n"
463	"\n"));
464	ipstats(pData);
465	tcpstats(pData);
466	udpstats(pData);
467	icmpstats(pData);
468	mbufstats(pData);
469	sockstats(pData);
470	Log(("\n"
471	"\n"
472	"\n"));
473	#endif
474	free(pData);
475	}
476
477
478	#define CONN_CANFSEND(so) (((so)->so_state & (SS_FCANTSENDMORE\|SS_ISFCONNECTED)) == SS_ISFCONNECTED)
479	#define CONN_CANFRCV(so) (((so)->so_state & (SS_FCANTRCVMORE\|SS_ISFCONNECTED)) == SS_ISFCONNECTED)
480	#define UPD_NFDS(x) if (nfds < (x)) nfds = (x)
481
482	/*
483	* curtime kept to an accuracy of 1ms
484	*/
485	#ifdef RT_OS_WINDOWS
486	static void updtime(PNATState pData)
487	{
488	struct _timeb tb;
489
490	_ftime(&tb);
491	curtime = (u_int)tb.time * (u_int)1000;
492	curtime += (u_int)tb.millitm;
493	}
494	#else
495	static void updtime(PNATState pData)
496	{
497	gettimeofday(&tt, 0);
498
499	curtime = (u_int)tt.tv_sec * (u_int)1000;
500	curtime += (u_int)tt.tv_usec / (u_int)1000;
501
502	if ((tt.tv_usec % 1000) >= 500)
503	curtime++;
504	}
505	#endif
506
507	void slirp_select_fill(PNATState pData, int *pnfds,
508	fd_set readfds, fd_set writefds, fd_set *xfds)
509	{
510	struct socket so, so_next;
511	int nfds;
512	#if defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) && defined(RT_OS_WINDOWS)
513	int rc;
514	int error;
515	#endif
516	int i;
517
518	STAM_REL_PROFILE_START(&pData->StatFill, a);
519
520	nfds = *pnfds;
521
522	/*
523	* First, TCP sockets
524	*/
525	do_slowtimo = 0;
526	if (link_up)
527	{
528	/*
529	* *_slowtimo needs calling if there are IP fragments
530	* in the fragment queue, or there are TCP connections active
531	*/
532	/* XXX:
533	* triggering of fragment expiration should be the same but use new macroses
534	*/
535	do_slowtimo = (tcb.so_next != &tcb);
536	if (!do_slowtimo)
537	{
538	for (i = 0; i < IPREASS_NHASH; i++)
539	{
540	if (!TAILQ_EMPTY(&ipq[i]))
541	{
542	do_slowtimo = 1;
543	break;
544	}
545	}
546	}
547	ICMP_ENGAGE_EVENT(&pData->icmp_socket, readfds);
548
549	STAM_REL_COUNTER_RESET(&pData->StatTCP);
550	STAM_REL_COUNTER_RESET(&pData->StatTCPHot);
551
552	for (so = tcb.so_next; so != &tcb; so = so_next)
553	{
554	so_next = so->so_next;
555
556	STAM_REL_COUNTER_INC(&pData->StatTCP);
557
558	/*
559	* See if we need a tcp_fasttimo
560	*/
561	if (time_fasttimo == 0 && so->so_tcpcb->t_flags & TF_DELACK)
562	time_fasttimo = curtime; /* Flag when we want a fasttimo */
563
564	/*
565	* NOFDREF can include still connecting to local-host,
566	* newly socreated() sockets etc. Don't want to select these.
567	*/
568	if (so->so_state & SS_NOFDREF \|\| so->s == -1)
569	continue;
570
571	/*
572	* Set for reading sockets which are accepting
573	*/
574	if (so->so_state & SS_FACCEPTCONN)
575	{
576	STAM_REL_COUNTER_INC(&pData->StatTCPHot);
577	TCP_ENGAGE_EVENT1(so, readfds);
578	continue;
579	}
580
581	/*
582	* Set for writing sockets which are connecting
583	*/
584	if (so->so_state & SS_ISFCONNECTING)
585	{
586	STAM_REL_COUNTER_INC(&pData->StatTCPHot);
587	TCP_ENGAGE_EVENT1(so, writefds);
588	}
589
590	/*
591	* Set for writing if we are connected, can send more, and
592	* we have something to send
593	*/
594	if (CONN_CANFSEND(so) && so->so_rcv.sb_cc)
595	{
596	STAM_REL_COUNTER_INC(&pData->StatTCPHot);
597	TCP_ENGAGE_EVENT1(so, writefds);
598	}
599
600	/*
601	* Set for reading (and urgent data) if we are connected, can
602	* receive more, and we have room for it XXX /2 ?
603	*/
604	if (CONN_CANFRCV(so) && (so->so_snd.sb_cc < (so->so_snd.sb_datalen/2)))
605	{
606	STAM_REL_COUNTER_INC(&pData->StatTCPHot);
607	TCP_ENGAGE_EVENT2(so, readfds, xfds);
608	}
609	}
610
611	/*
612	* UDP sockets
613	*/
614	STAM_REL_COUNTER_RESET(&pData->StatUDP);
615	STAM_REL_COUNTER_RESET(&pData->StatUDPHot);
616
617	for (so = udb.so_next; so != &udb; so = so_next)
618	{
619	so_next = so->so_next;
620
621	STAM_REL_COUNTER_INC(&pData->StatUDP);
622
623	/*
624	* See if it's timed out
625	*/
626	if (so->so_expire)
627	{
628	if (so->so_expire <= curtime)
629	{
630	udp_detach(pData, so);
631	continue;
632	}
633	else
634	do_slowtimo = 1; /* Let socket expire */
635	}
636
637	/*
638	* When UDP packets are received from over the link, they're
639	* sendto()'d straight away, so no need for setting for writing
640	* Limit the number of packets queued by this session to 4.
641	* Note that even though we try and limit this to 4 packets,
642	* the session could have more queued if the packets needed
643	* to be fragmented.
644	*
645	* (XXX <= 4 ?)
646	*/
647	if ((so->so_state & SS_ISFCONNECTED) && so->so_queued <= 4)
648	{
649	STAM_REL_COUNTER_INC(&pData->StatUDPHot);
650	UDP_ENGAGE_EVENT(so, readfds);
651	}
652	}
653
654	}
655
656	#if !defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) \|\| !defined(RT_OS_WINDOWS)
657	*pnfds = nfds;
658	#else
659	*pnfds = VBOX_EVENT_COUNT;
660	#endif
661
662	STAM_REL_PROFILE_STOP(&pData->StatFill, a);
663	}
664
665	#if defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) && defined(RT_OS_WINDOWS)
666	void slirp_select_poll(PNATState pData, int fTimeout, int fIcmp)
667	#else
668	void slirp_select_poll(PNATState pData, fd_set readfds, fd_set writefds, fd_set *xfds)
669	#endif
670	{
671	struct socket so, so_next;
672	int ret;
673	#if defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) && defined(RT_OS_WINDOWS)
674	WSANETWORKEVENTS NetworkEvents;
675	int rc;
676	int error;
677	#endif
678
679	STAM_REL_PROFILE_START(&pData->StatPoll, a);
680
681	/* Update time */
682	updtime(pData);
683
684	/*
685	* See if anything has timed out
686	*/
687	if (link_up)
688	{
689	if (time_fasttimo && ((curtime - time_fasttimo) >= 2))
690	{
691	STAM_REL_PROFILE_START(&pData->StatFastTimer, a);
692	tcp_fasttimo(pData);
693	time_fasttimo = 0;
694	STAM_REL_PROFILE_STOP(&pData->StatFastTimer, a);
695	}
696	if (do_slowtimo && ((curtime - last_slowtimo) >= 499))
697	{
698	STAM_REL_PROFILE_START(&pData->StatSlowTimer, a);
699	ip_slowtimo(pData);
700	tcp_slowtimo(pData);
701	last_slowtimo = curtime;
702	STAM_REL_PROFILE_STOP(&pData->StatSlowTimer, a);
703	}
704	}
705	#if defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) && defined(RT_OS_WINDOWS)
706	if (fTimeout)
707	return; /* only timer update */
708	#endif
709
710	/*
711	* Check sockets
712	*/
713	if (link_up)
714	{
715	#if defined(VBOX_WITH_SLIRP_ICMP)
716	# if defined(RT_OS_WINDOWS)
717	if (fIcmp)
718	sorecvfrom(pData, &pData->icmp_socket);
719	# else
720	if (pData->icmp_socket.s != -1 && FD_ISSET(pData->icmp_socket.s, readfds))
721	sorecvfrom(pData, &pData->icmp_socket);
722	# endif
723	#endif
724	/*
725	* Check TCP sockets
726	*/
727	for (so = tcb.so_next; so != &tcb; so = so_next)
728	{
729	so_next = so->so_next;
730
731	/*
732	* FD_ISSET is meaningless on these sockets
733	* (and they can crash the program)
734	*/
735	if (so->so_state & SS_NOFDREF \|\| so->s == -1)
736	continue;
737
738	POLL_TCP_EVENTS(rc, error, so, &NetworkEvents);
739
740	LOG_NAT_SOCK(so, TCP, &NetworkEvents, readfds, writefds, xfds);
741
742	/*
743	* Check for URG data
744	* This will soread as well, so no need to
745	* test for readfds below if this succeeds
746	*/
747
748	/* out-of-band data */
749	if (CHECK_FD_SET(so, NetworkEvents, xfds))
750	{
751	sorecvoob(pData, so);
752	}
753
754	/*
755	* Check sockets for reading
756	*/
757	else if (CHECK_FD_SET(so, NetworkEvents, readfds))
758	{
759	/*
760	* Check for incoming connections
761	*/
762	if (so->so_state & SS_FACCEPTCONN)
763	{
764	tcp_connect(pData, so);
765	#if defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) && defined(RT_OS_WINDOWS)
766	if (!(NetworkEvents.lNetworkEvents & FD_CLOSE))
767	#endif
768	continue;
769	}
770
771	ret = soread(pData, so, /fCloseIfNothingRead=/false);
772	/* Output it if we read something */
773	if (ret > 0)
774	tcp_output(pData, sototcpcb(so));
775	}
776
777	#if defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) && defined(RT_OS_WINDOWS)
778	/*
779	* Check for FD_CLOSE events.
780	*/
781	if (NetworkEvents.lNetworkEvents & FD_CLOSE)
782	{
783	/*
784	* drain the socket
785	*/
786	for (;;)
787	{
788	ret = soread(pData, so, /fCloseIfNothingRead=/true);
789	if (ret > 0)
790	tcp_output(pData, sototcpcb(so));
791	else
792	break;
793	}
794	}
795	#endif
796
797	/*
798	* Check sockets for writing
799	*/
800	if (CHECK_FD_SET(so, NetworkEvents, writefds))
801	{
802	/*
803	* Check for non-blocking, still-connecting sockets
804	*/
805	if (so->so_state & SS_ISFCONNECTING)
806	{
807	/* Connected */
808	so->so_state &= ~SS_ISFCONNECTING;
809
810	/*
811	* This should be probably guarded by PROBE_CONN too. Anyway,
812	* we disable it on OS/2 because the below send call returns
813	* EFAULT which causes the opened TCP socket to close right
814	* after it has been opened and connected.
815	*/
816	#ifndef RT_OS_OS2
817	ret = send(so->s, (const char *)&ret, 0, 0);
818	if (ret < 0)
819	{
820	/* XXXXX Must fix, zero bytes is a NOP */
821	if ( errno == EAGAIN
822	\|\| errno == EWOULDBLOCK
823	\|\| errno == EINPROGRESS
824	\|\| errno == ENOTCONN)
825	continue;
826
827	/* else failed */
828	so->so_state = SS_NOFDREF;
829	}
830	/* else so->so_state &= ~SS_ISFCONNECTING; */
831	#endif
832
833	/*
834	* Continue tcp_input
835	*/
836	tcp_input(pData, (struct mbuf *)NULL, sizeof(struct ip), so);
837	/* continue; */
838	}
839	else
840	ret = sowrite(pData, so);
841	/*
842	* XXX If we wrote something (a lot), there could be the need
843	* for a window update. In the worst case, the remote will send
844	* a window probe to get things going again.
845	*/
846	}
847
848	/*
849	* Probe a still-connecting, non-blocking socket
850	* to check if it's still alive
851	*/
852	#ifdef PROBE_CONN
853	if (so->so_state & SS_ISFCONNECTING)
854	{
855	ret = recv(so->s, (char *)&ret, 0, 0);
856
857	if (ret < 0)
858	{
859	/* XXX */
860	if ( errno == EAGAIN
861	\|\| errno == EWOULDBLOCK
862	\|\| errno == EINPROGRESS
863	\|\| errno == ENOTCONN)
864	{
865	continue; /* Still connecting, continue */
866	}
867
868	/* else failed */
869	so->so_state = SS_NOFDREF;
870
871	/* tcp_input will take care of it */
872	}
873	else
874	{
875	ret = send(so->s, &ret, 0, 0);
876	if (ret < 0)
877	{
878	/* XXX */
879	if ( errno == EAGAIN
880	\|\| errno == EWOULDBLOCK
881	\|\| errno == EINPROGRESS
882	\|\| errno == ENOTCONN)
883	{
884	continue;
885	}
886	/* else failed */
887	so->so_state = SS_NOFDREF;
888	}
889	else
890	so->so_state &= ~SS_ISFCONNECTING;
891
892	}
893	tcp_input((struct mbuf *)NULL, sizeof(struct ip),so);
894	} /* SS_ISFCONNECTING */
895	#endif
896	}
897
898	/*
899	* Now UDP sockets.
900	* Incoming packets are sent straight away, they're not buffered.
901	* Incoming UDP data isn't buffered either.
902	*/
903	for (so = udb.so_next; so != &udb; so = so_next)
904	{
905	so_next = so->so_next;
906
907	POLL_UDP_EVENTS(rc, error, so, &NetworkEvents);
908
909	LOG_NAT_SOCK(so, UDP, &NetworkEvents, readfds, writefds, xfds);
910
911	if (so->s != -1 && CHECK_FD_SET(so, NetworkEvents, readfds))
912	{
913	sorecvfrom(pData, so);
914	}
915	}
916
917	#if 0
918	#if defined(VBOX_WITH_SLIRP_ICMP)
919	# if defined(RT_OS_WINDOWS)
920	if (fIcmp)
921	sorecvfrom(pData, &pData->icmp_socket);
922	# else
923	if (pData->icmp_socket.s != -1 && FD_ISSET(pData->icmp_socket.s, readfds))
924	sorecvfrom(pData, &pData->icmp_socket);
925	# endif
926	#endif
927	#endif
928	}
929
930	/*
931	* See if we can start outputting
932	*/
933	if (if_queued && link_up)
934	if_start(pData);
935
936	STAM_REL_PROFILE_STOP(&pData->StatPoll, a);
937	}
938
939	#define ETH_ALEN 6
940	#define ETH_HLEN 14
941
942	#define ETH_P_IP 0x0800 /* Internet Protocol packet */
943	#define ETH_P_ARP 0x0806 /* Address Resolution packet */
944
945	#define ARPOP_REQUEST 1 /* ARP request */
946	#define ARPOP_REPLY 2 /* ARP reply */
947
948	struct ethhdr
949	{
950	unsigned char h_dest[ETH_ALEN]; /* destination eth addr */
951	unsigned char h_source[ETH_ALEN]; /* source ether addr */
952	unsigned short h_proto; /* packet type ID field */
953	};
954
955	struct arphdr
956	{
957	unsigned short ar_hrd; /* format of hardware address */
958	unsigned short ar_pro; /* format of protocol address */
959	unsigned char ar_hln; /* length of hardware address */
960	unsigned char ar_pln; /* length of protocol address */
961	unsigned short ar_op; /* ARP opcode (command) */
962
963	/*
964	* Ethernet looks like this : This bit is variable sized however...
965	*/
966	unsigned char ar_sha[ETH_ALEN]; /* sender hardware address */
967	unsigned char ar_sip[4]; /* sender IP address */
968	unsigned char ar_tha[ETH_ALEN]; /* target hardware address */
969	unsigned char ar_tip[4]; /* target IP address */
970	};
971
972	static
973	void arp_input(PNATState pData, const uint8_t *pkt, int pkt_len)
974	{
975	struct ethhdr eh = (struct ethhdr )pkt;
976	struct arphdr ah = (struct arphdr )(pkt + ETH_HLEN);
977	uint8_t arp_reply[ETH_HLEN + sizeof(struct arphdr)];
978	struct ethhdr reh = (struct ethhdr )arp_reply;
979	struct arphdr rah = (struct arphdr )(arp_reply + ETH_HLEN);
980	int ar_op;
981	struct ex_list *ex_ptr;
982	uint32_t htip = ntohl((uint32_t)ah->ar_tip);
983
984	ar_op = ntohs(ah->ar_op);
985	switch(ar_op)
986	{
987	case ARPOP_REQUEST:
988	if ((htip & pData->netmask) == ntohl(special_addr.s_addr))
989	{
990	if ( (htip & ~pData->netmask) == CTL_DNS
991	\|\| (htip & ~pData->netmask) == CTL_ALIAS)
992	goto arp_ok;
993	for (ex_ptr = exec_list; ex_ptr; ex_ptr = ex_ptr->ex_next)
994	{
995	if ((htip & ~pData->netmask) == ex_ptr->ex_addr)
996	goto arp_ok;
997	}
998	return;
999	arp_ok:
1000	/* XXX: make an ARP request to have the client address */
1001	memcpy(client_ethaddr, eh->h_source, ETH_ALEN);
1002
1003	/* ARP request for alias/dns mac address */
1004	memcpy(reh->h_dest, pkt + ETH_ALEN, ETH_ALEN);
1005	memcpy(reh->h_source, special_ethaddr, ETH_ALEN - 1);
1006	reh->h_source[5] = ah->ar_tip[3];
1007	reh->h_proto = htons(ETH_P_ARP);
1008
1009	rah->ar_hrd = htons(1);
1010	rah->ar_pro = htons(ETH_P_IP);
1011	rah->ar_hln = ETH_ALEN;
1012	rah->ar_pln = 4;
1013	rah->ar_op = htons(ARPOP_REPLY);
1014	memcpy(rah->ar_sha, reh->h_source, ETH_ALEN);
1015	memcpy(rah->ar_sip, ah->ar_tip, 4);
1016	memcpy(rah->ar_tha, ah->ar_sha, ETH_ALEN);
1017	memcpy(rah->ar_tip, ah->ar_sip, 4);
1018	slirp_output(pData->pvUser, arp_reply, sizeof(arp_reply));
1019	}
1020	break;
1021	default:
1022	break;
1023	}
1024	}
1025
1026	void slirp_input(PNATState pData, const uint8_t *pkt, int pkt_len)
1027	{
1028	struct mbuf *m;
1029	int proto;
1030
1031	if (pkt_len < ETH_HLEN)
1032	return;
1033
1034	proto = ntohs((uint16_t )(pkt + 12));
1035	switch(proto)
1036	{
1037	case ETH_P_ARP:
1038	arp_input(pData, pkt, pkt_len);
1039	break;
1040	case ETH_P_IP:
1041	/* Update time. Important if the network is very quiet, as otherwise
1042	* the first outgoing connection gets an incorrect timestamp. */
1043	updtime(pData);
1044
1045	m = m_get(pData);
1046	if (!m)
1047	return;
1048	/* Note: we add to align the IP header */
1049	if (M_FREEROOM(m) < pkt_len + 2)
1050	{
1051	m_inc(m, pkt_len + 2);
1052	}
1053	m->m_len = pkt_len + 2;
1054	memcpy(m->m_data + 2, pkt, pkt_len);
1055
1056	m->m_data += 2 + ETH_HLEN;
1057	m->m_len -= 2 + ETH_HLEN;
1058
1059	ip_input(pData, m);
1060	break;
1061	default:
1062	break;
1063	}
1064	}
1065
1066	/* output the IP packet to the ethernet device */
1067	void if_encap(PNATState pData, const uint8_t *ip_data, int ip_data_len)
1068	{
1069	uint8_t buf[1600];
1070	struct ethhdr eh = (struct ethhdr )buf;
1071
1072	if (ip_data_len + ETH_HLEN > sizeof(buf))
1073	return;
1074
1075	memcpy(eh->h_dest, client_ethaddr, ETH_ALEN);
1076	memcpy(eh->h_source, special_ethaddr, ETH_ALEN - 1);
1077	/* XXX: not correct */
1078	eh->h_source[5] = CTL_ALIAS;
1079	eh->h_proto = htons(ETH_P_IP);
1080	memcpy(buf + sizeof(struct ethhdr), ip_data, ip_data_len);
1081	slirp_output(pData->pvUser, buf, ip_data_len + ETH_HLEN);
1082	}
1083
1084	int slirp_redir(PNATState pData, int is_udp, int host_port,
1085	struct in_addr guest_addr, int guest_port)
1086	{
1087	if (is_udp)
1088	{
1089	if (!udp_listen(pData, htons(host_port), guest_addr.s_addr,
1090	htons(guest_port), 0))
1091	return -1;
1092	}
1093	else
1094	{
1095	if (!solisten(pData, htons(host_port), guest_addr.s_addr,
1096	htons(guest_port), 0))
1097	return -1;
1098	}
1099	return 0;
1100	}
1101
1102	int slirp_add_exec(PNATState pData, int do_pty, const char *args, int addr_low_byte,
1103	int guest_port)
1104	{
1105	return add_exec(&exec_list, do_pty, (char *)args,
1106	addr_low_byte, htons(guest_port));
1107	}
1108
1109	void slirp_set_ethaddr(PNATState pData, const uint8_t *ethaddr)
1110	{
1111	memcpy(client_ethaddr, ethaddr, ETH_ALEN);
1112	}
1113
1114	#if defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) && defined(RT_OS_WINDOWS)
1115	HANDLE *slirp_get_events(PNATState pData)
1116	{
1117	return pData->phEvents;
1118	}
1119	void slirp_register_external_event(PNATState pData, HANDLE hEvent, int index)
1120	{
1121	pData->phEvents[index] = hEvent;
1122	}
1123	#endif
1124
1125	unsigned int slirp_get_timeout_ms(PNATState pData)
1126	{
1127	if (link_up)
1128	{
1129	if (time_fasttimo)
1130	return 2;
1131	if (do_slowtimo)
1132	return 500; /* see PR_SLOWHZ */
1133	}
1134	return 0;
1135	}

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

source: vbox/trunk/src/VBox/Devices/Network/slirp/slirp.c@ 15580

Download in other formats: