VirtualBox

source: vbox/trunk/src/VBox/Devices/Network/slirp/slirp.c@ 17302

Last change on this file since 17302 was 17302, checked in by vboxsync, 16 years ago

NAT: integrated dnsproxy
the initialization of 2 dns servers should be done to make it work (but before need some comment on xtracker)
  • Property svn:eol-style set to native
File size: 54.4 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#ifndef RT_OS_WINDOWS
10# include <sys/ioctl.h>
11# include <poll.h>
12#else
13# define _WINSOCK2API_
14# include <IPHlpApi.h>
15#endif
16
17#if !defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) || !defined(RT_OS_WINDOWS)
18
19# ifndef VBOX_WITH_SIMPLIFIED_SLIRP_SYNC
20# define DO_ENGAGE_EVENT1(so, fdset, label) \
21 do { \
22 FD_SET((so)->s, (fdset)); \
23 UPD_NFDS((so)->s); \
24 } while(0)
25
26
27# define DO_ENGAGE_EVENT2(so, fdset1, fdset2, label) \
28 do { \
29 FD_SET((so)->s, (fdset1)); \
30 FD_SET((so)->s, (fdset2)); \
31 UPD_NFDS((so)->s); \
32 } while(0)
33
34# define DO_POLL_EVENTS(rc, error, so, events, label) do {} while (0)
35
36# define DO_CHECK_FD_SET(so, events, fdset) (FD_ISSET((so)->s, fdset))
37# define DO_UNIX_CHECK_FD_SET(so, events, fdset ) 0 /*specific for Unix API */
38# else /* !VBOX_WITH_SIMPLIFIED_SLIRP_SYNC */
39# define DO_ENGAGE_EVENT1(so, fdset, label) \
40 do { \
41 if( so->so_poll_index != -1 \
42 && so->s == polls[so->so_poll_index].fd) { \
43 polls[so->so_poll_index].events |= N_(fdset ## _poll); \
44 break; /* out of this loop */ \
45 } \
46 AssertRelease(poll_index < (nfds)); \
47 AssertRelease(poll_index >= 0 && poll_index < (nfds)); \
48 polls[poll_index].fd = (so)->s; \
49 (so)->so_poll_index = poll_index; \
50 polls[poll_index].events = N_(fdset ## _poll); \
51 polls[poll_index].revents = 0; \
52 poll_index++; \
53 } while(0)
54
55
56# define DO_ENGAGE_EVENT2(so, fdset1, fdset2, label) \
57 do { \
58 if( so->so_poll_index != -1 \
59 && so->s == polls[so->so_poll_index].fd) { \
60 polls[so->so_poll_index].events |= \
61 N_(fdset1 ## _poll) | N_(fdset1 ## _poll); \
62 break; /* out of this loop */ \
63 } \
64 AssertRelease(poll_index < (nfds)); \
65 polls[poll_index].fd = (so)->s; \
66 (so)->so_poll_index = poll_index; \
67 polls[poll_index].events = \
68 N_(fdset1 ## _poll) | N_(fdset1 ## _poll); \
69 poll_index++; \
70 } while(0)
71
72# define DO_POLL_EVENTS(rc, error, so, events, label) do {} while (0)
73
74# define DO_CHECK_FD_SET(so, events, fdset) ( ((so)->so_poll_index != -1) \
75 && ((so)->so_poll_index <= ndfs) \
76 && ((so)->s == polls[so->so_poll_index].fd) \
77 && (polls[(so)->so_poll_index].revents & N_(fdset ## _poll)))
78# define DO_UNIX_CHECK_FD_SET(so, events, fdset ) DO_CHECK_FD_SET((so), (events), fdset) /*specific for Unix API */
79# define DO_WIN_CHECK_FD_SET(so, events, fdset ) 0 /* specific for Windows Winsock API */
80# endif /* VBOX_WITH_SIMPLIFIED_SLIRP_SYNC */
81
82# ifndef RT_OS_WINDOWS
83
84# ifndef RT_OS_LINUX
85# define readfds_poll (POLLRDNORM)
86# define writefds_poll (POLLWRNORM)
87# define xfds_poll (POLLRDBAND|POLLWRBAND|POLLPRI)
88# else
89# define readfds_poll (POLLIN)
90# define writefds_poll (POLLOUT)
91# define xfds_poll (POLLPRI)
92# endif
93# define rderr_poll (POLLERR)
94# define rdhup_poll (POLLHUP)
95# define nval_poll (POLLNVAL)
96
97# define ICMP_ENGAGE_EVENT(so, fdset) \
98 do { \
99 if (pData->icmp_socket.s != -1) \
100 DO_ENGAGE_EVENT1((so), fdset, ICMP); \
101 } while (0)
102# else /* !RT_OS_WINDOWS */
103# ifdef VBOX_WITH_SIMPLIFIED_SLIRP_SYNC
104# define DO_WIN_CHECK_FD_SET(so, events, fdset ) DO_CHECK_FD_SET((so), (events), fdset)
105# else /* VBOX_WITH_SIMPLIFIED_SLIRP_SYNC */
106# define DO_WIN_CHECK_FD_SET(so, events, fdset ) 0
107# endif /* !VBOX_WITH_SIMPLIFIED_SLIRP_SYNC */
108# define ICMP_ENGAGE_EVENT(so, fdset) do {} while(0)
109#endif /* RT_OS_WINDOWS */
110
111#else /* defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) && defined(RT_OS_WINDOWS) */
112
113/*
114 * On Windows, we will be notified by IcmpSendEcho2() when the response arrives.
115 * So no call to WSAEventSelect necessary.
116 */
117# define ICMP_ENGAGE_EVENT(so, fdset) do {} while(0)
118
119# define DO_ENGAGE_EVENT1(so, fdset1, label) \
120 do { \
121 rc = WSAEventSelect((so)->s, VBOX_SOCKET_EVENT, FD_ALL_EVENTS); \
122 if (rc == SOCKET_ERROR) \
123 { \
124 /* This should not happen */ \
125 error = WSAGetLastError(); \
126 LogRel(("WSAEventSelect (" #label ") error %d (so=%x, socket=%s, event=%x)\n", \
127 error, (so), (so)->s, VBOX_SOCKET_EVENT)); \
128 } \
129 } while(0); \
130 CONTINUE(label)
131
132# define DO_ENGAGE_EVENT2(so, fdset1, fdset2, label) \
133 DO_ENGAGE_EVENT1((so), (fdset1), label)
134
135# define DO_POLL_EVENTS(rc, error, so, events, label) \
136 (rc) = WSAEnumNetworkEvents((so)->s, VBOX_SOCKET_EVENT, (events)); \
137 if ((rc) == SOCKET_ERROR) \
138 { \
139 (error) = WSAGetLastError(); \
140 LogRel(("WSAEnumNetworkEvents " #label " error %d\n", (error))); \
141 CONTINUE(label); \
142 }
143
144# define acceptds_win FD_ACCEPT
145# define acceptds_win_bit FD_ACCEPT_BIT
146
147# define readfds_win FD_READ
148# define readfds_win_bit FD_READ_BIT
149
150# define writefds_win FD_WRITE
151# define writefds_win_bit FD_WRITE_BIT
152
153# define xfds_win FD_OOB
154# define xfds_win_bit FD_OOB_BIT
155
156# define DO_CHECK_FD_SET(so, events, fdset) \
157 (((events).lNetworkEvents & fdset ## _win) && ((events).iErrorCode[fdset ## _win_bit] == 0))
158
159# define DO_WIN_CHECK_FD_SET(so, events, fdset ) DO_CHECK_FD_SET((so), (events), fdset)
160# define DO_UNIX_CHECK_FD_SET(so, events, fdset ) 1 /*specific for Unix API */
161
162#endif /* defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) && defined(RT_OS_WINDOWS) */
163
164#define TCP_ENGAGE_EVENT1(so, fdset) \
165 DO_ENGAGE_EVENT1((so), fdset, tcp)
166
167#define TCP_ENGAGE_EVENT2(so, fdset1, fdset2) \
168 DO_ENGAGE_EVENT2((so), fdset1, fdset2, tcp)
169
170#define UDP_ENGAGE_EVENT(so, fdset) \
171 DO_ENGAGE_EVENT1((so), fdset, udp)
172
173#define POLL_TCP_EVENTS(rc, error, so, events) \
174 DO_POLL_EVENTS((rc), (error), (so), (events), tcp)
175
176#define POLL_UDP_EVENTS(rc, error, so, events) \
177 DO_POLL_EVENTS((rc), (error), (so), (events), udp)
178
179#define CHECK_FD_SET(so, events, set) \
180 (DO_CHECK_FD_SET((so), (events), set))
181
182#define WIN_CHECK_FD_SET(so, events, set) \
183 (DO_WIN_CHECK_FD_SET((so), (events), set))
184#define UNIX_CHECK_FD_SET(so, events, set) \
185 (DO_UNIX_CHECK_FD_SET(so, events, set))
186
187/*
188 * Loging macros
189 */
190#if VBOX_WITH_DEBUG_NAT_SOCKETS
191# if defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC)
192# if defined(RT_OS_WINDOWS)
193# define DO_LOG_NAT_SOCK(so, proto, winevent, r_fdset, w_fdset, x_fdset) \
194 do { \
195 LogRel((" " #proto " %R[natsock] %R[natwinnetevents]\n", (so), (winevent))); \
196 } while (0)
197# else /* RT_OS_WINDOWS */
198# define DO_LOG_NAT_SOCK(so, proto, winevent, r_fdset, w_fdset, x_fdset) \
199 do { \
200 LogRel((" " #proto " %R[natsock] %s %s %s er: %s, %s, %s\n", (so), \
201 CHECK_FD_SET(so, ign ,r_fdset) ? "READ":"", \
202 CHECK_FD_SET(so, ign, w_fdset) ? "WRITE":"", \
203 CHECK_FD_SET(so, ign, x_fdset) ? "OOB":"", \
204 CHECK_FD_SET(so, ign, rderr) ? "RDERR":"", \
205 CHECK_FD_SET(so, ign, rdhup) ? "RDHUP":"", \
206 CHECK_FD_SET(so, ign, nval) ? "RDNVAL":"")); \
207 } while (0)
208# endif /* !RT_OS_WINDOWS */
209# else /* VBOX_WITH_SIMPLIFIED_SLIRP_SYNC */
210# define DO_LOG_NAT_SOCK(so, proto, winevent, r_fdset, w_fdset, x_fdset) \
211 do { \
212 LogRel((" " #proto " %R[natsock] %s %s %s\n", (so), \
213 FD_ISSET((so)->s, (r_fdset))?"READ":"", \
214 FD_ISSET((so)->s, (w_fdset))?"WRITE":"", \
215 FD_ISSET((so)->s, (x_fdset))?"OOB":"")); \
216 } while (0)
217# endif
218#else /* VBOX_WITH_DEBUG_NAT_SOCKETS */
219# define DO_LOG_NAT_SOCK(so, proto, winevent, r_fdset, w_fdset, x_fdset) do {} while (0)
220#endif /* !VBOX_WITH_DEBUG_NAT_SOCKETS */
221
222#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)
223
224static const uint8_t special_ethaddr[6] =
225{
226 0x52, 0x54, 0x00, 0x12, 0x35, 0x00
227};
228
229#ifdef RT_OS_WINDOWS
230# ifndef VBOX_WITH_MULTI_DNS
231static int get_dns_addr_domain(PNATState pData, bool fVerbose,
232 struct in_addr *pdns_addr,
233 const char **ppszDomain)
234{
235 int rc = 0;
236 FIXED_INFO *FixedInfo = NULL;
237 ULONG BufLen;
238 DWORD ret;
239 IP_ADDR_STRING *pIPAddr;
240 struct in_addr tmp_addr;
241
242 FixedInfo = (FIXED_INFO *)GlobalAlloc(GPTR, sizeof(FIXED_INFO));
243 BufLen = sizeof(FIXED_INFO);
244
245 /** @todo: this API returns all DNS servers, no matter whether the
246 * corresponding network adapter is disabled or not. Maybe replace
247 * this by GetAdapterAddresses(), which is XP/Vista only though. */
248 if (ERROR_BUFFER_OVERFLOW == GetNetworkParams(FixedInfo, &BufLen))
249 {
250 if (FixedInfo)
251 {
252 GlobalFree(FixedInfo);
253 FixedInfo = NULL;
254 }
255 FixedInfo = GlobalAlloc(GPTR, BufLen);
256 }
257
258 if ((ret = GetNetworkParams(FixedInfo, &BufLen)) != ERROR_SUCCESS)
259 {
260 Log(("GetNetworkParams failed. ret = %08x\n", (u_int)ret ));
261 if (FixedInfo)
262 {
263 GlobalFree(FixedInfo);
264 FixedInfo = NULL;
265 }
266 rc = -1;
267 goto get_dns_prefix;
268 }
269
270 pIPAddr = &(FixedInfo->DnsServerList);
271 inet_aton(pIPAddr->IpAddress.String, &tmp_addr);
272 Log(("nat: DNS Servers:\n"));
273 if (fVerbose || pdns_addr->s_addr != tmp_addr.s_addr)
274 LogRel(("NAT: DNS address: %s\n", pIPAddr->IpAddress.String));
275 *pdns_addr = tmp_addr;
276
277 pIPAddr = FixedInfo -> DnsServerList.Next;
278 while (pIPAddr)
279 {
280 if (fVerbose)
281 LogRel(("NAT: ignored DNS address: %s\n", pIPAddr ->IpAddress.String));
282 pIPAddr = pIPAddr ->Next;
283 }
284 if (FixedInfo)
285 {
286 GlobalFree(FixedInfo);
287 FixedInfo = NULL;
288 }
289
290get_dns_prefix:
291 if (ppszDomain)
292 {
293 OSVERSIONINFO ver;
294 char szDnsDomain[256];
295 DWORD dwSize = sizeof(szDnsDomain);
296
297 *ppszDomain = NULL;
298 GetVersionEx(&ver);
299 if (ver.dwMajorVersion >= 5)
300 {
301 /* GetComputerNameEx exists in Windows versions starting with 2000. */
302 if (GetComputerNameEx(ComputerNameDnsDomain, szDnsDomain, &dwSize))
303 {
304 if (szDnsDomain[0])
305 {
306 /* Just non-empty strings are valid. */
307 *ppszDomain = RTStrDup(szDnsDomain);
308 if (pData->fPassDomain)
309 {
310 if (fVerbose)
311 LogRel(("NAT: passing domain name %s\n", szDnsDomain));
312 }
313 else
314 Log(("nat: ignoring domain %s\n", szDnsDomain));
315 }
316 }
317 else
318 Log(("nat: GetComputerNameEx failed (%d)\n", GetLastError()));
319 }
320 }
321 return rc;
322}
323# else /* !VBOX_WITH_MULTI_DNS */
324static int get_dns_addr_domain(PNATState pData, bool fVerbose,
325 struct in_addr *pdns_addr,
326 const char **ppszDomain)
327{
328 /* Get amount of memory required for operation */
329 ULONG flags = GAA_FLAG_INCLUDE_PREFIX; /*GAA_FLAG_INCLUDE_ALL_INTERFACES;*/ /* all interfaces registered in NDIS */
330 PIP_ADAPTER_ADDRESSES addresses = NULL;
331 PIP_ADAPTER_ADDRESSES addr = NULL;
332 PIP_ADAPTER_DNS_SERVER_ADDRESS dns = NULL;
333 ULONG size = 0;
334 char *suffix;
335 struct dns_entry *da = NULL;
336 ULONG ret = ERROR_SUCCESS;
337
338 /* @todo add SKIPing flags to get only required information */
339
340 ret = pData->pfGetAdaptersAddresses(AF_INET, 0, NULL /* reserved */, addresses, &size);
341 if (ret != ERROR_BUFFER_OVERFLOW)
342 {
343 LogRel(("NAT: error %lu occured on capacity detection operation\n", ret));
344 return -1;
345 }
346
347 if (size == 0)
348 {
349 LogRel(("NAT: Win socket API returns non capacity\n"));
350 return -1;
351 }
352
353 addresses = RTMemAllocZ(size);
354 if (addresses == NULL)
355 {
356 LogRel(("NAT: No memory available \n"));
357 return -1;
358 }
359
360 ret = pData->pfGetAdaptersAddresses(AF_INET, 0, NULL /* reserved */, addresses, &size);
361 if (ret != ERROR_SUCCESS)
362 {
363 LogRel(("NAT: error %lu occured on fetching adapters info\n", ret));
364 return -1;
365 }
366 addr = addresses;
367 while(addr != NULL)
368 {
369 size_t buff_size;
370 if (addr->OperStatus != IfOperStatusUp)
371 goto next;
372 dns = addr->FirstDnsServerAddress;
373 while (dns != NULL)
374 {
375 struct sockaddr *saddr = dns->Address.lpSockaddr;
376 if (saddr->sa_family != AF_INET)
377 goto next_dns;
378 /* add dns server to list */
379 da = RTMemAllocZ(sizeof(struct dns_entry));
380 if (da == NULL)
381 {
382 LogRel(("NAT: Can't allocate buffer for DNS entry\n"));
383 return VERR_NO_MEMORY;
384 }
385 LogRel(("NAT: adding %R[IP4] to DNS server list\n", &((struct sockaddr_in *)saddr)->sin_addr));
386 if ((((struct sockaddr_in *)saddr)->sin_addr.s_addr & htonl(IN_CLASSA_NET)) == ntohl(INADDR_LOOPBACK & IN_CLASSA_NET)) {
387 da->de_addr.s_addr = htonl(ntohl(special_addr.s_addr) | CTL_ALIAS);
388 }
389 else
390 {
391 da->de_addr.s_addr = ((struct sockaddr_in *)saddr)->sin_addr.s_addr;
392 }
393 LIST_INSERT_HEAD(&pData->dns_list_head, da, de_list);
394 next_dns:
395 dns = dns->Next;
396 }
397 buff_size = wcstombs(NULL, addr->DnsSuffix, 0);
398 if (buff_size == 0)
399 goto next;
400 suffix = RTMemAllocZ(buff_size);
401 wcstombs(suffix, addr->DnsSuffix, buff_size);
402 LogRel(("NAT: adding %s to DNS suffix list\n", suffix));
403 *ppszDomain = suffix;
404 next:
405 addr = addr->Next;
406 }
407 /*@todo add dns suffix if required */
408 LogRel(("NAT: adding dns suffix %s to the list \n", ppszDomain));
409 return 0;
410}
411# endif /* VBOX_WITH_MULTI_DNS */
412
413#else /* !RT_OS_WINDOWS */
414
415static int get_dns_addr_domain(PNATState pData, bool fVerbose,
416 struct in_addr *pdns_addr,
417 const char **ppszDomain)
418{
419 char buff[512];
420 char buff2[256];
421 FILE *f;
422 int found = 0;
423 struct in_addr tmp_addr;
424
425#ifdef RT_OS_OS2
426 /* Try various locations. */
427 char *etc = getenv("ETC");
428 f = NULL;
429 if (etc)
430 {
431 snprintf(buff, sizeof(buff), "%s/RESOLV2", etc);
432 f = fopen(buff, "rt");
433 }
434 if (!f)
435 {
436 snprintf(buff, sizeof(buff), "%s/RESOLV2", _PATH_ETC);
437 f = fopen(buff, "rt");
438 }
439 if (!f)
440 {
441 snprintf(buff, sizeof(buff), "%s/resolv.conf", _PATH_ETC);
442 f = fopen(buff, "rt");
443 }
444#else
445 f = fopen("/etc/resolv.conf", "r");
446#endif
447 if (!f)
448 return -1;
449
450 if (ppszDomain)
451 *ppszDomain = NULL;
452 Log(("nat: DNS Servers:\n"));
453 while (fgets(buff, 512, f) != NULL)
454 {
455#ifdef VBOX_WITH_MULTI_DNS
456 struct dns_entry *da = NULL;
457#endif
458 if (sscanf(buff, "nameserver%*[ \t]%256s", buff2) == 1)
459 {
460 if (!inet_aton(buff2, &tmp_addr))
461 continue;
462#ifndef VBOX_WITH_MULTI_DNS
463 /* If it's the first one, set it to dns_addr */
464 if (!found)
465 {
466 if (fVerbose || pdns_addr->s_addr != tmp_addr.s_addr)
467 LogRel(("NAT: DNS address: %s\n", buff2));
468 *pdns_addr = tmp_addr;
469 }
470 else
471 {
472 if (fVerbose)
473 LogRel(("NAT: ignored DNS address: %s\n", buff2));
474 }
475#else
476 /*localhost mask */
477 da = RTMemAllocZ(sizeof (struct dns_entry));
478 if (da == NULL)
479 {
480 LogRel(("can't alloc memory for DNS entry\n"));
481 return -1;
482 }
483 /*check */
484 da->de_addr.s_addr = tmp_addr.s_addr;
485 if ((da->de_addr.s_addr & htonl(IN_CLASSA_NET)) == ntohl(INADDR_LOOPBACK & IN_CLASSA_NET)) {
486 da->de_addr.s_addr = htonl(ntohl(special_addr.s_addr) | CTL_ALIAS);
487 }
488 LIST_INSERT_HEAD(&pData->dns_list_head, da, de_list);
489#endif
490 found++;
491 }
492 if ( ppszDomain
493 && (!strncmp(buff, "domain", 6) || !strncmp(buff, "search", 6)))
494 {
495 /* Domain name/search list present. Pick first entry */
496 if (*ppszDomain == NULL)
497 {
498 char *tok;
499 char *saveptr;
500 tok = strtok_r(&buff[6], " \t\n", &saveptr);
501 if (tok)
502 {
503 *ppszDomain = RTStrDup(tok);
504 if (pData->fPassDomain)
505 {
506 if (fVerbose)
507 LogRel(("NAT: passing domain name %s\n", tok));
508 }
509 else
510 Log(("nat: ignoring domain %s\n", tok));
511 }
512 }
513 }
514 }
515 fclose(f);
516 if (!found)
517 return -1;
518 return 0;
519}
520
521#endif
522
523int get_dns_addr(PNATState pData, struct in_addr *pdns_addr)
524{
525 return get_dns_addr_domain(pData, false, pdns_addr, NULL);
526}
527
528int slirp_init(PNATState *ppData, const char *pszNetAddr, uint32_t u32Netmask,
529 bool fPassDomain, const char *pszTFTPPrefix,
530 const char *pszBootFile, void *pvUser)
531{
532 int fNATfailed = 0;
533 int rc;
534 PNATState pData = RTMemAllocZ(sizeof(NATState));
535 *ppData = pData;
536 if (!pData)
537 return VERR_NO_MEMORY;
538 if (u32Netmask & 0x1f)
539 /* CTL is x.x.x.15, bootp passes up to 16 IPs (15..31) */
540 return VERR_INVALID_PARAMETER;
541 pData->fPassDomain = fPassDomain;
542 pData->pvUser = pvUser;
543 tftp_prefix = pszTFTPPrefix;
544 bootp_filename = pszBootFile;
545 pData->netmask = u32Netmask;
546
547#ifdef RT_OS_WINDOWS
548 {
549 WSADATA Data;
550 WSAStartup(MAKEWORD(2, 0), &Data);
551 }
552# if defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC)
553 pData->phEvents[VBOX_SOCKET_EVENT_INDEX] = CreateEvent(NULL, FALSE, FALSE, NULL);
554# endif
555#endif
556#ifdef VBOX_WITH_SLIRP_MT
557 QSOCKET_LOCK_CREATE(tcb);
558 QSOCKET_LOCK_CREATE(udb);
559 rc = RTReqCreateQueue(&pData->pReqQueue);
560 AssertReleaseRC(rc);
561#endif
562
563 link_up = 1;
564
565 debug_init();
566 if_init(pData);
567 ip_init(pData);
568 icmp_init(pData);
569
570 /* Initialise mbufs *after* setting the MTU */
571 m_init(pData);
572
573 inet_aton(pszNetAddr, &special_addr);
574 alias_addr.s_addr = special_addr.s_addr | htonl(CTL_ALIAS);
575 /* @todo: add ability to configure this staff */
576 pData->tftp_server.s_addr = htonl(ntohl(special_addr.s_addr) | CTL_TFTP);
577
578 /* set default addresses */
579 inet_aton("127.0.0.1", &loopback_addr);
580#ifndef VBOX_WITH_MULTI_DNS
581 inet_aton("127.0.0.1", &dns_addr);
582
583 if (get_dns_addr_domain(pData, true, &dns_addr, &pData->pszDomain) < 0)
584#else
585 LIST_INIT(&pData->dns_list_head);
586 if (get_dns_addr_domain(pData, true, NULL, &pData->pszDomain) < 0)
587#endif
588 fNATfailed = 1;
589#ifdef VBOX_WITH_SLIRP_DNS_PROXY
590 dnsproxy_init(pData);
591#endif
592
593 getouraddr(pData);
594 return fNATfailed ? VINF_NAT_DNS : VINF_SUCCESS;
595}
596
597/**
598 * Statistics counters.
599 */
600void slirp_register_timers(PNATState pData, PPDMDRVINS pDrvIns)
601{
602#ifdef VBOX_WITH_STATISTICS
603 PDMDrvHlpSTAMRegisterF(pDrvIns, &pData->StatFill, STAMTYPE_PROFILE, STAMVISIBILITY_ALWAYS,
604 STAMUNIT_TICKS_PER_CALL, "Profiling slirp fills", "/Drivers/NAT%d/Fill", pDrvIns->iInstance);
605 PDMDrvHlpSTAMRegisterF(pDrvIns, &pData->StatPoll, STAMTYPE_PROFILE, STAMVISIBILITY_ALWAYS,
606 STAMUNIT_TICKS_PER_CALL, "Profiling slirp polls", "/Drivers/NAT%d/Poll", pDrvIns->iInstance);
607 PDMDrvHlpSTAMRegisterF(pDrvIns, &pData->StatFastTimer, STAMTYPE_PROFILE, STAMVISIBILITY_ALWAYS,
608 STAMUNIT_TICKS_PER_CALL, "Profiling slirp fast timer", "/Drivers/NAT%d/TimerFast", pDrvIns->iInstance);
609 PDMDrvHlpSTAMRegisterF(pDrvIns, &pData->StatSlowTimer, STAMTYPE_PROFILE, STAMVISIBILITY_ALWAYS,
610 STAMUNIT_TICKS_PER_CALL, "Profiling slirp slow timer", "/Drivers/NAT%d/TimerSlow", pDrvIns->iInstance);
611 PDMDrvHlpSTAMRegisterF(pDrvIns, &pData->StatTCP, STAMTYPE_COUNTER, STAMVISIBILITY_ALWAYS,
612 STAMUNIT_COUNT, "TCP sockets", "/Drivers/NAT%d/SockTCP", pDrvIns->iInstance);
613 PDMDrvHlpSTAMRegisterF(pDrvIns, &pData->StatTCPHot, STAMTYPE_COUNTER, STAMVISIBILITY_ALWAYS,
614 STAMUNIT_COUNT, "TCP sockets active", "/Drivers/NAT%d/SockTCPHot", pDrvIns->iInstance);
615 PDMDrvHlpSTAMRegisterF(pDrvIns, &pData->StatUDP, STAMTYPE_COUNTER, STAMVISIBILITY_ALWAYS,
616 STAMUNIT_COUNT, "UDP sockets", "/Drivers/NAT%d/SockUDP", pDrvIns->iInstance);
617 PDMDrvHlpSTAMRegisterF(pDrvIns, &pData->StatUDPHot, STAMTYPE_COUNTER, STAMVISIBILITY_ALWAYS,
618 STAMUNIT_COUNT, "UDP sockets active", "/Drivers/NAT%d/SockUDPHot", pDrvIns->iInstance);
619#endif /* VBOX_WITH_STATISTICS */
620}
621
622/**
623 * Marks the link as up, making it possible to establish new connections.
624 */
625void slirp_link_up(PNATState pData)
626{
627 link_up = 1;
628}
629
630/**
631 * Marks the link as down and cleans up the current connections.
632 */
633void slirp_link_down(PNATState pData)
634{
635 struct socket *so;
636
637 while ((so = tcb.so_next) != &tcb)
638 {
639 if (so->so_state & SS_NOFDREF || so->s == -1)
640 sofree(pData, so);
641 else
642 tcp_drop(pData, sototcpcb(so), 0);
643 }
644
645 while ((so = udb.so_next) != &udb)
646 udp_detach(pData, so);
647
648 link_up = 0;
649}
650
651/**
652 * Terminates the slirp component.
653 */
654void slirp_term(PNATState pData)
655{
656#ifdef VBOX_WITH_MULTI_DNS
657 struct dns_entry *de = NULL;
658#endif
659 if (pData->pszDomain)
660 RTStrFree((char *)(void *)pData->pszDomain);
661
662#ifdef RT_OS_WINDOWS
663 pData->pfIcmpCloseHandle(pData->icmp_socket.sh);
664 FreeLibrary(pData->hmIcmpLibrary);
665 RTMemFree(pData->pvIcmpBuffer);
666# else
667 closesocket(pData->icmp_socket.s);
668#endif
669
670 slirp_link_down(pData);
671#ifdef VBOX_WITH_MULTI_DNS
672 while(!LIST_EMPTY(&pData->dns_list_head)) {
673 de = LIST_FIRST(&pData->dns_list_head);
674 LIST_REMOVE(de, de_list);
675 RTMemFree(de);
676 }
677#endif
678#ifdef RT_OS_WINDOWS
679 WSACleanup();
680#endif
681#ifdef LOG_ENABLED
682 Log(("\n"
683 "NAT statistics\n"
684 "--------------\n"
685 "\n"));
686 ipstats(pData);
687 tcpstats(pData);
688 udpstats(pData);
689 icmpstats(pData);
690 mbufstats(pData);
691 sockstats(pData);
692 Log(("\n"
693 "\n"
694 "\n"));
695#endif
696 RTMemFree(pData);
697}
698
699
700#define CONN_CANFSEND(so) (((so)->so_state & (SS_FCANTSENDMORE|SS_ISFCONNECTED)) == SS_ISFCONNECTED)
701#define CONN_CANFRCV(so) (((so)->so_state & (SS_FCANTRCVMORE|SS_ISFCONNECTED)) == SS_ISFCONNECTED)
702#define UPD_NFDS(x) do { if (nfds < (x)) nfds = (x); } while (0)
703
704/*
705 * curtime kept to an accuracy of 1ms
706 */
707static void updtime(PNATState pData)
708{
709#ifdef RT_OS_WINDOWS
710 struct _timeb tb;
711
712 _ftime(&tb);
713 curtime = (u_int)tb.time * (u_int)1000;
714 curtime += (u_int)tb.millitm;
715#else
716 gettimeofday(&tt, 0);
717
718 curtime = (u_int)tt.tv_sec * (u_int)1000;
719 curtime += (u_int)tt.tv_usec / (u_int)1000;
720
721 if ((tt.tv_usec % 1000) >= 500)
722 curtime++;
723#endif
724}
725
726#ifndef VBOX_WITH_SIMPLIFIED_SLIRP_SYNC
727void slirp_select_fill(PNATState pData, int *pnfds,
728 fd_set *readfds, fd_set *writefds, fd_set *xfds)
729#else /* !VBOX_WITH_SIMPLIFIED_SLIRP_SYNC */
730# ifdef RT_OS_WINDOWS
731void slirp_select_fill(PNATState pData, int *pnfds)
732# else /* RT_OS_WINDOWS */
733void slirp_select_fill(PNATState pData, int *pnfds, struct pollfd *polls)
734# endif /* !RT_OS_WINDOWS */
735#endif /* VBOX_WITH_SIMPLIFIED_SLIRP_SYNC */
736{
737 struct socket *so, *so_next;
738 int nfds;
739#if defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) && defined(RT_OS_WINDOWS)
740 int rc;
741 int error;
742#endif
743#if defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) && !defined(RT_OS_WINDOWS)
744 int poll_index = 0;
745#endif
746 int i;
747
748 STAM_PROFILE_START(&pData->StatFill, a);
749
750 nfds = *pnfds;
751
752 /*
753 * First, TCP sockets
754 */
755 do_slowtimo = 0;
756 if (link_up)
757 {
758 /*
759 * *_slowtimo needs calling if there are IP fragments
760 * in the fragment queue, or there are TCP connections active
761 */
762 /* XXX:
763 * triggering of fragment expiration should be the same but use new macroses
764 */
765 do_slowtimo = (tcb.so_next != &tcb);
766 if (!do_slowtimo)
767 {
768 for (i = 0; i < IPREASS_NHASH; i++)
769 {
770 if (!TAILQ_EMPTY(&ipq[i]))
771 {
772 do_slowtimo = 1;
773 break;
774 }
775 }
776 }
777 ICMP_ENGAGE_EVENT(&pData->icmp_socket, readfds);
778
779 STAM_COUNTER_RESET(&pData->StatTCP);
780 STAM_COUNTER_RESET(&pData->StatTCPHot);
781
782 QSOCKET_FOREACH(so, so_next, tcp)
783 /* { */
784#if defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) && !defined(RT_OS_WINDOWS)
785 so->so_poll_index = -1;
786#endif
787 STAM_COUNTER_INC(&pData->StatTCP);
788
789 /*
790 * See if we need a tcp_fasttimo
791 */
792 if ( time_fasttimo == 0
793 && so->so_tcpcb != NULL
794 && so->so_tcpcb->t_flags & TF_DELACK)
795 time_fasttimo = curtime; /* Flag when we want a fasttimo */
796
797 /*
798 * NOFDREF can include still connecting to local-host,
799 * newly socreated() sockets etc. Don't want to select these.
800 */
801 if (so->so_state & SS_NOFDREF || so->s == -1)
802 CONTINUE(tcp);
803
804 /*
805 * Set for reading sockets which are accepting
806 */
807 if (so->so_state & SS_FACCEPTCONN)
808 {
809 STAM_COUNTER_INC(&pData->StatTCPHot);
810 TCP_ENGAGE_EVENT1(so, readfds);
811 CONTINUE(tcp);
812 }
813
814 /*
815 * Set for writing sockets which are connecting
816 */
817 if (so->so_state & SS_ISFCONNECTING)
818 {
819 Log2(("connecting %R[natsock] engaged\n",so));
820 STAM_COUNTER_INC(&pData->StatTCPHot);
821 TCP_ENGAGE_EVENT1(so, writefds);
822 }
823
824 /*
825 * Set for writing if we are connected, can send more, and
826 * we have something to send
827 */
828 if (CONN_CANFSEND(so) && so->so_rcv.sb_cc)
829 {
830 STAM_COUNTER_INC(&pData->StatTCPHot);
831 TCP_ENGAGE_EVENT1(so, writefds);
832 }
833
834 /*
835 * Set for reading (and urgent data) if we are connected, can
836 * receive more, and we have room for it XXX /2 ?
837 */
838 if (CONN_CANFRCV(so) && (so->so_snd.sb_cc < (so->so_snd.sb_datalen/2)))
839 {
840 STAM_COUNTER_INC(&pData->StatTCPHot);
841 TCP_ENGAGE_EVENT2(so, readfds, xfds);
842 }
843 LOOP_LABEL(tcp, so, so_next);
844 }
845
846 /*
847 * UDP sockets
848 */
849 STAM_COUNTER_RESET(&pData->StatUDP);
850 STAM_COUNTER_RESET(&pData->StatUDPHot);
851
852 QSOCKET_FOREACH(so, so_next, udp)
853 /* { */
854
855 STAM_COUNTER_INC(&pData->StatUDP);
856#if defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) && !defined(RT_OS_WINDOWS)
857 so->so_poll_index = -1;
858#endif
859
860 /*
861 * See if it's timed out
862 */
863 if (so->so_expire)
864 {
865 if (so->so_expire <= curtime)
866 {
867#ifdef VBOX_WITH_SLIRP_MT
868 /* we need so_next for continue our cycle*/
869 so_next = so->so_next;
870#endif
871 UDP_DETACH(pData, so, so_next);
872 CONTINUE_NO_UNLOCK(udp);
873 }
874 else
875 do_slowtimo = 1; /* Let socket expire */
876 }
877
878 /*
879 * When UDP packets are received from over the link, they're
880 * sendto()'d straight away, so no need for setting for writing
881 * Limit the number of packets queued by this session to 4.
882 * Note that even though we try and limit this to 4 packets,
883 * the session could have more queued if the packets needed
884 * to be fragmented.
885 *
886 * (XXX <= 4 ?)
887 */
888 if ((so->so_state & SS_ISFCONNECTED) && so->so_queued <= 4)
889 {
890 STAM_COUNTER_INC(&pData->StatUDPHot);
891 UDP_ENGAGE_EVENT(so, readfds);
892 }
893 LOOP_LABEL(udp, so, so_next);
894 }
895
896 }
897
898#if defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC)
899# if defined(RT_OS_WINDOWS)
900 *pnfds = VBOX_EVENT_COUNT;
901# else /* RT_OS_WINDOWS */
902 AssertRelease(poll_index <= *pnfds);
903 *pnfds = poll_index;
904# endif /* !RT_OS_WINDOWS */
905#else /* VBOX_WITH_SIMPLIFIED_SLIRP_SYNC */
906 *pnfds = nfds;
907#endif /* !VBOX_WITH_SIMPLIFIED_SLIRP_SYNC */
908
909 STAM_PROFILE_STOP(&pData->StatFill, a);
910}
911
912#if defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC)
913# if defined(RT_OS_WINDOWS)
914void slirp_select_poll(PNATState pData, int fTimeout, int fIcmp)
915# else /* RT_OS_WINDOWS */
916void slirp_select_poll(PNATState pData, struct pollfd *polls, int ndfs)
917# endif /* !RT_OS_WINDOWS */
918#else /* VBOX_WITH_SIMPLIFIED_SLIRP_SYNC */
919void slirp_select_poll(PNATState pData, fd_set *readfds, fd_set *writefds, fd_set *xfds)
920#endif /* !VBOX_WITH_SIMPLIFIED_SLIRP_SYNC */
921{
922 struct socket *so, *so_next;
923 int ret;
924#if defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) && defined(RT_OS_WINDOWS)
925 WSANETWORKEVENTS NetworkEvents;
926 int rc;
927 int error;
928#endif
929#if defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) && !defined(RT_OS_WINDOWS)
930 int poll_index = 0;
931#endif
932
933 STAM_PROFILE_START(&pData->StatPoll, a);
934
935 /* Update time */
936 updtime(pData);
937
938 /*
939 * See if anything has timed out
940 */
941 if (link_up)
942 {
943 if (time_fasttimo && ((curtime - time_fasttimo) >= 2))
944 {
945 STAM_PROFILE_START(&pData->StatFastTimer, a);
946 tcp_fasttimo(pData);
947 time_fasttimo = 0;
948 STAM_PROFILE_STOP(&pData->StatFastTimer, a);
949 }
950 if (do_slowtimo && ((curtime - last_slowtimo) >= 499))
951 {
952 STAM_PROFILE_START(&pData->StatSlowTimer, a);
953 ip_slowtimo(pData);
954 tcp_slowtimo(pData);
955 last_slowtimo = curtime;
956 STAM_PROFILE_STOP(&pData->StatSlowTimer, a);
957 }
958 }
959#if defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) && defined(RT_OS_WINDOWS)
960 if (fTimeout)
961 return; /* only timer update */
962#endif
963
964 /*
965 * Check sockets
966 */
967 if (link_up)
968 {
969#if defined(RT_OS_WINDOWS)
970 /*XXX: before renaming please make see define
971 * fIcmp in slirp_state.h
972 */
973 if (fIcmp)
974 sorecvfrom(pData, &pData->icmp_socket);
975#else
976 if ( (pData->icmp_socket.s != -1)
977 && CHECK_FD_SET(&pData->icmp_socket, ignored, readfds))
978 sorecvfrom(pData, &pData->icmp_socket);
979#endif
980 /*
981 * Check TCP sockets
982 */
983 QSOCKET_FOREACH(so, so_next, tcp)
984 /* { */
985
986#ifdef VBOX_WITH_SLIRP_MT
987 if ( so->so_state & SS_NOFDREF
988 && so->so_deleted == 1)
989 {
990 struct socket *son, *sop = NULL;
991 QSOCKET_LOCK(tcb);
992 if (so->so_next != NULL)
993 {
994 if (so->so_next != &tcb)
995 SOCKET_LOCK(so->so_next);
996 son = so->so_next;
997 }
998 if ( so->so_prev != &tcb
999 && so->so_prev != NULL)
1000 {
1001 SOCKET_LOCK(so->so_prev);
1002 sop = so->so_prev;
1003 }
1004 QSOCKET_UNLOCK(tcb);
1005 remque(pData, so);
1006 NSOCK_DEC();
1007 SOCKET_UNLOCK(so);
1008 SOCKET_LOCK_DESTROY(so);
1009 RTMemFree(so);
1010 so_next = son;
1011 if (sop != NULL)
1012 SOCKET_UNLOCK(sop);
1013 CONTINUE_NO_UNLOCK(tcp);
1014 }
1015#endif
1016 /*
1017 * FD_ISSET is meaningless on these sockets
1018 * (and they can crash the program)
1019 */
1020 if (so->so_state & SS_NOFDREF || so->s == -1)
1021 CONTINUE(tcp);
1022
1023 POLL_TCP_EVENTS(rc, error, so, &NetworkEvents);
1024
1025 LOG_NAT_SOCK(so, TCP, &NetworkEvents, readfds, writefds, xfds);
1026
1027
1028 /*
1029 * Check for URG data
1030 * This will soread as well, so no need to
1031 * test for readfds below if this succeeds
1032 */
1033
1034 /* out-of-band data */
1035 if (CHECK_FD_SET(so, NetworkEvents, xfds))
1036 {
1037 sorecvoob(pData, so);
1038 }
1039
1040 /*
1041 * Check sockets for reading
1042 */
1043 else if ( CHECK_FD_SET(so, NetworkEvents, readfds)
1044 || WIN_CHECK_FD_SET(so, NetworkEvents, acceptds))
1045 {
1046 /*
1047 * Check for incoming connections
1048 */
1049 if (so->so_state & SS_FACCEPTCONN)
1050 {
1051 TCP_CONNECT(pData, so);
1052#if defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) && defined(RT_OS_WINDOWS)
1053 if (!(NetworkEvents.lNetworkEvents & FD_CLOSE))
1054#endif
1055 CONTINUE(tcp);
1056 }
1057
1058 ret = soread(pData, so);
1059 /* Output it if we read something */
1060 if (ret > 0)
1061 TCP_OUTPUT(pData, sototcpcb(so));
1062 }
1063
1064#if defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) && defined(RT_OS_WINDOWS)
1065 /*
1066 * Check for FD_CLOSE events.
1067 * in some cases once FD_CLOSE engaged on socket it could be flashed latter (for some reasons)
1068 */
1069 if ( (NetworkEvents.lNetworkEvents & FD_CLOSE)
1070 || (so->so_close == 1))
1071 {
1072 so->so_close = 1; /* mark it */
1073 /*
1074 * drain the socket
1075 */
1076 for (;;)
1077 {
1078 ret = soread(pData, so);
1079 if (ret > 0)
1080 TCP_OUTPUT(pData, sototcpcb(so));
1081 else
1082 break;
1083 }
1084 CONTINUE(tcp);
1085 }
1086#endif
1087
1088 /*
1089 * Check sockets for writing
1090 */
1091 if (CHECK_FD_SET(so, NetworkEvents, writefds))
1092 {
1093 /*
1094 * Check for non-blocking, still-connecting sockets
1095 */
1096 if (so->so_state & SS_ISFCONNECTING)
1097 {
1098 Log2(("connecting %R[natsock] catched\n", so));
1099 /* Connected */
1100 so->so_state &= ~SS_ISFCONNECTING;
1101
1102 /*
1103 * This should be probably guarded by PROBE_CONN too. Anyway,
1104 * we disable it on OS/2 because the below send call returns
1105 * EFAULT which causes the opened TCP socket to close right
1106 * after it has been opened and connected.
1107 */
1108#ifndef RT_OS_OS2
1109 ret = send(so->s, (const char *)&ret, 0, 0);
1110 if (ret < 0)
1111 {
1112 /* XXXXX Must fix, zero bytes is a NOP */
1113 if ( errno == EAGAIN
1114 || errno == EWOULDBLOCK
1115 || errno == EINPROGRESS
1116 || errno == ENOTCONN)
1117 CONTINUE(tcp);
1118
1119 /* else failed */
1120 so->so_state = SS_NOFDREF;
1121 }
1122 /* else so->so_state &= ~SS_ISFCONNECTING; */
1123#endif
1124
1125 /*
1126 * Continue tcp_input
1127 */
1128 TCP_INPUT(pData, (struct mbuf *)NULL, sizeof(struct ip), so);
1129 /* continue; */
1130 }
1131 else
1132 SOWRITE(ret, pData, so);
1133 /*
1134 * XXX If we wrote something (a lot), there could be the need
1135 * for a window update. In the worst case, the remote will send
1136 * a window probe to get things going again.
1137 */
1138 }
1139
1140 /*
1141 * Probe a still-connecting, non-blocking socket
1142 * to check if it's still alive
1143 */
1144#ifdef PROBE_CONN
1145 if (so->so_state & SS_ISFCONNECTING)
1146 {
1147 ret = recv(so->s, (char *)&ret, 0, 0);
1148
1149 if (ret < 0)
1150 {
1151 /* XXX */
1152 if ( errno == EAGAIN
1153 || errno == EWOULDBLOCK
1154 || errno == EINPROGRESS
1155 || errno == ENOTCONN)
1156 {
1157 CONTINUE(tcp); /* Still connecting, continue */
1158 }
1159
1160 /* else failed */
1161 so->so_state = SS_NOFDREF;
1162
1163 /* tcp_input will take care of it */
1164 }
1165 else
1166 {
1167 ret = send(so->s, &ret, 0, 0);
1168 if (ret < 0)
1169 {
1170 /* XXX */
1171 if ( errno == EAGAIN
1172 || errno == EWOULDBLOCK
1173 || errno == EINPROGRESS
1174 || errno == ENOTCONN)
1175 {
1176 CONTINUE(tcp);
1177 }
1178 /* else failed */
1179 so->so_state = SS_NOFDREF;
1180 }
1181 else
1182 so->so_state &= ~SS_ISFCONNECTING;
1183
1184 }
1185 TCP_INPUT((struct mbuf *)NULL, sizeof(struct ip),so);
1186 } /* SS_ISFCONNECTING */
1187#endif
1188#ifndef RT_OS_WINDOWS
1189 if ( UNIX_CHECK_FD_SET(so, NetworkEvents, rdhup)
1190 || UNIX_CHECK_FD_SET(so, NetworkEvents, rderr))
1191 {
1192 int err;
1193 int inq, outq;
1194 int status;
1195 socklen_t optlen = sizeof(int);
1196 inq = outq = 0;
1197 status = getsockopt(so->s, SOL_SOCKET, SO_ERROR, &err, &optlen);
1198 if (status != 0)
1199 LogRel(("NAT: can't get error status from %R[natsock]\n", so));
1200#ifndef RT_OS_SOLARIS
1201 status = ioctl(so->s, FIONREAD, &inq); /* tcp(7) recommends SIOCINQ which is Linux specific */
1202 if (status != 0 || status != EINVAL)
1203 {
1204 /* EINVAL returned if socket in listen state tcp(7)*/
1205 LogRel(("NAT: can't get depth of IN queue status from %R[natsock]\n", so));
1206 }
1207 status = ioctl(so->s, TIOCOUTQ, &outq); /* SIOCOUTQ see previous comment */
1208 if (status != 0)
1209 LogRel(("NAT: can't get depth of OUT queue from %R[natsock]\n", so));
1210#else
1211 /*
1212 * Solaris has bit different ioctl commands and its handlings
1213 * hint: streamio(7) I_NREAD
1214 */
1215#endif
1216 if ( so->so_state & SS_ISFCONNECTING
1217 || UNIX_CHECK_FD_SET(so, NetworkEvents, readfds))
1218 {
1219 /**
1220 * Check if we need here take care about gracefull connection
1221 * @todo try with proxy server
1222 */
1223 if (UNIX_CHECK_FD_SET(so, NetworkEvents, readfds))
1224 {
1225 /*
1226 * Never meet inq != 0 or outq != 0, anyway let it stay for a while
1227 * in case it happens we'll able to detect it.
1228 * Give TCP/IP stack wait or expire the socket.
1229 */
1230 LogRel(("NAT: %R[natsock] err(%d:%s) s(in:%d,out:%d)happens on read I/O, "
1231 "other side close connection \n", so, err, strerror(err), inq, outq));
1232 CONTINUE(tcp);
1233 }
1234 goto tcp_input_close;
1235 }
1236 if ( !UNIX_CHECK_FD_SET(so, NetworkEvents, readfds)
1237 && !UNIX_CHECK_FD_SET(so, NetworkEvents, writefds)
1238 && !UNIX_CHECK_FD_SET(so, NetworkEvents, xfds))
1239 {
1240 LogRel(("NAT: system expires the socket %R[natsock] err(%d:%s) s(in:%d,out:%d) happens on non-I/O. ",
1241 so, err, strerror(err), inq, outq));
1242 goto tcp_input_close;
1243 }
1244 LogRel(("NAT: %R[natsock] we've met(%d:%s) s(in:%d, out:%d) unhandled combination hup (%d) "
1245 "rederr(%d) on (r:%d, w:%d, x:%d)\n",
1246 so, err, strerror(err),
1247 inq, outq,
1248 UNIX_CHECK_FD_SET(so, ign, rdhup),
1249 UNIX_CHECK_FD_SET(so, ign, rderr),
1250 UNIX_CHECK_FD_SET(so, ign, readfds),
1251 UNIX_CHECK_FD_SET(so, ign, writefds),
1252 UNIX_CHECK_FD_SET(so, ign, xfds)));
1253 /*
1254 * Give OS's TCP/IP stack a chance to resolve an issue or expire the socket.
1255 */
1256 CONTINUE(tcp);
1257tcp_input_close:
1258 so->so_state = SS_NOFDREF; /*cause connection valid tcp connection termination and socket closing */
1259 TCP_INPUT(pData, (struct mbuf *)NULL, sizeof(struct ip), so);
1260 CONTINUE(tcp);
1261 }
1262#endif
1263 LOOP_LABEL(tcp, so, so_next);
1264 }
1265
1266 /*
1267 * Now UDP sockets.
1268 * Incoming packets are sent straight away, they're not buffered.
1269 * Incoming UDP data isn't buffered either.
1270 */
1271 QSOCKET_FOREACH(so, so_next, udp)
1272 /* { */
1273#ifdef VBOX_WITH_SLIRP_MT
1274 if ( so->so_state & SS_NOFDREF
1275 && so->so_deleted == 1)
1276 {
1277 struct socket *son, *sop = NULL;
1278 QSOCKET_LOCK(udb);
1279 if (so->so_next != NULL)
1280 {
1281 if (so->so_next != &udb)
1282 SOCKET_LOCK(so->so_next);
1283 son = so->so_next;
1284 }
1285 if ( so->so_prev != &udb
1286 && so->so_prev != NULL)
1287 {
1288 SOCKET_LOCK(so->so_prev);
1289 sop = so->so_prev;
1290 }
1291 QSOCKET_UNLOCK(udb);
1292 remque(pData, so);
1293 NSOCK_DEC();
1294 SOCKET_UNLOCK(so);
1295 SOCKET_LOCK_DESTROY(so);
1296 RTMemFree(so);
1297 so_next = son;
1298 if (sop != NULL)
1299 SOCKET_UNLOCK(sop);
1300 CONTINUE_NO_UNLOCK(udp);
1301 }
1302#endif
1303 POLL_UDP_EVENTS(rc, error, so, &NetworkEvents);
1304
1305 LOG_NAT_SOCK(so, UDP, &NetworkEvents, readfds, writefds, xfds);
1306
1307 if (so->s != -1 && CHECK_FD_SET(so, NetworkEvents, readfds))
1308 {
1309 SORECVFROM(pData, so);
1310 }
1311 LOOP_LABEL(udp, so, so_next);
1312 }
1313
1314 }
1315
1316#ifndef VBOX_WITH_SLIRP_MT
1317 /*
1318 * See if we can start outputting
1319 */
1320 if (if_queued && link_up)
1321 if_start(pData);
1322#endif
1323
1324 STAM_PROFILE_STOP(&pData->StatPoll, a);
1325}
1326
1327#define ETH_ALEN 6
1328#define ETH_HLEN 14
1329
1330#define ARPOP_REQUEST 1 /* ARP request */
1331#define ARPOP_REPLY 2 /* ARP reply */
1332
1333struct ethhdr
1334{
1335 unsigned char h_dest[ETH_ALEN]; /* destination eth addr */
1336 unsigned char h_source[ETH_ALEN]; /* source ether addr */
1337 unsigned short h_proto; /* packet type ID field */
1338};
1339AssertCompileSize(struct ethhdr, 14);
1340
1341struct arphdr
1342{
1343 unsigned short ar_hrd; /* format of hardware address */
1344 unsigned short ar_pro; /* format of protocol address */
1345 unsigned char ar_hln; /* length of hardware address */
1346 unsigned char ar_pln; /* length of protocol address */
1347 unsigned short ar_op; /* ARP opcode (command) */
1348
1349 /*
1350 * Ethernet looks like this : This bit is variable sized however...
1351 */
1352 unsigned char ar_sha[ETH_ALEN]; /* sender hardware address */
1353 unsigned char ar_sip[4]; /* sender IP address */
1354 unsigned char ar_tha[ETH_ALEN]; /* target hardware address */
1355 unsigned char ar_tip[4]; /* target IP address */
1356};
1357AssertCompileSize(struct arphdr, 28);
1358
1359#ifdef VBOX_WITH_SIMPLIFIED_SLIRP_SYNC
1360static void arp_input(PNATState pData, struct mbuf *m)
1361#else
1362static void arp_input(PNATState pData, const uint8_t *pkt, int pkt_len)
1363#endif
1364{
1365 struct ethhdr *eh;
1366 struct ethhdr *reh;
1367 struct arphdr *ah;
1368 struct arphdr *rah;
1369 int ar_op;
1370 struct ex_list *ex_ptr;
1371 uint32_t htip;
1372#ifndef VBOX_WITH_SIMPLIFIED_SLIRP_SYNC
1373 uint8_t arp_reply[sizeof(struct arphdr) + ETH_HLEN];
1374 eh = (struct ethhdr *)pkt;
1375#else
1376 struct mbuf *mr;
1377 eh = mtod(m, struct ethhdr *);
1378#endif
1379 ah = (struct arphdr *)&eh[1];
1380 htip = ntohl(*(uint32_t*)ah->ar_tip);
1381
1382#ifdef VBOX_WITH_SIMPLIFIED_SLIRP_SYNC
1383 mr = m_get(pData);
1384 mr->m_data += if_maxlinkhdr;
1385 mr->m_len = sizeof(struct arphdr);
1386 rah = mtod(mr, struct arphdr *);
1387#else
1388 reh = (struct ethhdr *)arp_reply;
1389 rah = (struct arphdr *)&reh[1];
1390#endif
1391
1392 ar_op = ntohs(ah->ar_op);
1393 switch(ar_op)
1394 {
1395 case ARPOP_REQUEST:
1396 if ((htip & pData->netmask) == ntohl(special_addr.s_addr))
1397 {
1398 if ( CTL_CHECK(htip, CTL_DNS)
1399 || CTL_CHECK(htip, CTL_ALIAS)
1400 || CTL_CHECK(htip, CTL_TFTP))
1401 goto arp_ok;
1402 for (ex_ptr = exec_list; ex_ptr; ex_ptr = ex_ptr->ex_next)
1403 {
1404 if ((htip & ~pData->netmask) == ex_ptr->ex_addr)
1405 goto arp_ok;
1406 }
1407 return;
1408 arp_ok:
1409
1410#ifndef VBOX_WITH_SIMPLIFIED_SLIRP_SYNC
1411 memcpy(reh->h_dest, eh->h_source, ETH_ALEN);
1412 memcpy(reh->h_source, &special_addr, ETH_ALEN);
1413 reh->h_source[5] = ah->ar_tip[3];
1414 reh->h_proto = htons(ETH_P_ARP);
1415#endif
1416 rah->ar_hrd = htons(1);
1417 rah->ar_pro = htons(ETH_P_IP);
1418 rah->ar_hln = ETH_ALEN;
1419 rah->ar_pln = 4;
1420 rah->ar_op = htons(ARPOP_REPLY);
1421 memcpy(rah->ar_sha, special_ethaddr, ETH_ALEN);
1422
1423 switch (htip & ~pData->netmask)
1424 {
1425 case CTL_DNS:
1426 case CTL_ALIAS:
1427 rah->ar_sha[5] = (uint8_t)(htip & ~pData->netmask);
1428 break;
1429 default:;
1430 }
1431
1432 memcpy(rah->ar_sip, ah->ar_tip, 4);
1433 memcpy(rah->ar_tha, ah->ar_sha, ETH_ALEN);
1434 memcpy(rah->ar_tip, ah->ar_sip, 4);
1435#ifdef VBOX_WITH_SIMPLIFIED_SLIRP_SYNC
1436 if_encap(pData, ETH_P_ARP, mr);
1437 m_free(pData, m);
1438#else
1439 slirp_output(pData->pvUser, arp_reply, sizeof(arp_reply));
1440#endif
1441 }
1442 break;
1443 default:
1444 break;
1445 }
1446}
1447
1448void slirp_input(PNATState pData, const uint8_t *pkt, int pkt_len)
1449{
1450 struct mbuf *m;
1451 int proto;
1452 static bool fWarnedIpv6;
1453
1454 if (pkt_len < ETH_HLEN)
1455 {
1456 LogRel(("NAT: packet having size %d has been ingnored\n", pkt_len));
1457 return;
1458 }
1459
1460 m = m_get(pData);
1461 if (!m)
1462 {
1463 LogRel(("NAT: can't allocate new mbuf\n"));
1464 return;
1465 }
1466
1467 /* Note: we add to align the IP header */
1468
1469 if (M_FREEROOM(m) < pkt_len)
1470 m_inc(m, pkt_len);
1471
1472 m->m_len = pkt_len ;
1473 memcpy(m->m_data, pkt, pkt_len);
1474
1475 proto = ntohs(*(uint16_t *)(pkt + 12));
1476 switch(proto)
1477 {
1478 case ETH_P_ARP:
1479#ifdef VBOX_WITH_SIMPLIFIED_SLIRP_SYNC
1480 arp_input(pData, m);
1481#else
1482 arp_input(pData, pkt, pkt_len);
1483 m_free(pData, m);
1484#endif
1485 break;
1486 case ETH_P_IP:
1487 /* Update time. Important if the network is very quiet, as otherwise
1488 * the first outgoing connection gets an incorrect timestamp. */
1489 updtime(pData);
1490 m->m_data += ETH_HLEN;
1491 m->m_len -= ETH_HLEN;
1492 ip_input(pData, m);
1493 break;
1494 case ETH_P_IPV6:
1495 m_free(pData, m);
1496 if (!fWarnedIpv6)
1497 {
1498 LogRel(("NAT: IPv6 not supported\n"));
1499 fWarnedIpv6 = true;
1500 }
1501 break;
1502 default:
1503 LogRel(("NAT: Unsupported protocol %x\n", proto));
1504 m_free(pData, m);
1505 break;
1506 }
1507 RTMemFree((void *)pkt);
1508}
1509
1510/* output the IP packet to the ethernet device */
1511#ifdef VBOX_WITH_SIMPLIFIED_SLIRP_SYNC
1512void if_encap(PNATState pData, uint16_t eth_proto, struct mbuf *m)
1513#else
1514void if_encap(PNATState pData, uint8_t *ip_data, int ip_data_len)
1515#endif
1516{
1517#ifdef VBOX_WITH_SIMPLIFIED_SLIRP_SYNC
1518 struct ethhdr *eh;
1519 uint8_t *buf = RTMemAlloc(1600);
1520 m->m_data -= if_maxlinkhdr;
1521 m->m_len += ETH_HLEN;
1522 eh = mtod(m, struct ethhdr *);
1523#else
1524 uint8_t buf[1600];
1525 struct ethhdr *eh = (struct ethhdr *)buf;
1526
1527 if (ip_data_len + ETH_HLEN > sizeof(buf))
1528 return;
1529
1530 memcpy(buf + sizeof(struct ethhdr), ip_data, ip_data_len);
1531#endif
1532
1533
1534 memcpy(eh->h_dest, client_ethaddr, ETH_ALEN);
1535 memcpy(eh->h_source, special_ethaddr, ETH_ALEN - 1);
1536 /* XXX: not correct */
1537 eh->h_source[5] = CTL_ALIAS;
1538#ifdef VBOX_WITH_SIMPLIFIED_SLIRP_SYNC
1539 eh->h_proto = htons(eth_proto);
1540#if 0
1541 slirp_output(pData->pvUser, m, mtod(m, uint8_t *), m->m_len);
1542#else
1543 memcpy(buf, mtod(m, uint8_t *), m->m_len);
1544 slirp_output(pData->pvUser, NULL, buf, m->m_len);
1545 m_free(pData, m);
1546#endif
1547#else
1548 eh->h_proto = htons(ETH_P_IP);
1549 slirp_output(pData->pvUser, buf, ip_data_len + ETH_HLEN);
1550#endif
1551}
1552
1553int slirp_redir(PNATState pData, int is_udp, int host_port,
1554 struct in_addr guest_addr, int guest_port)
1555{
1556 struct socket *so;
1557 Log2(("NAT: set redirect %s hp:%d gp:%d\n", (is_udp?"UDP":"TCP"), host_port, guest_port));
1558 if (is_udp)
1559 {
1560 so = udp_listen(pData, htons(host_port), guest_addr.s_addr,
1561 htons(guest_port), 0);
1562 }
1563 else
1564 {
1565 so = solisten(pData, htons(host_port), guest_addr.s_addr,
1566 htons(guest_port), 0);
1567 }
1568 Log2(("NAT: redirecting socket %R[natsock]\n", so));
1569 return (so != NULL ? 0 : -1);
1570}
1571
1572int slirp_add_exec(PNATState pData, int do_pty, const char *args, int addr_low_byte,
1573 int guest_port)
1574{
1575 return add_exec(&exec_list, do_pty, (char *)args,
1576 addr_low_byte, htons(guest_port));
1577}
1578
1579void slirp_set_ethaddr(PNATState pData, const uint8_t *ethaddr)
1580{
1581 memcpy(client_ethaddr, ethaddr, ETH_ALEN);
1582}
1583
1584#if defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) && defined(RT_OS_WINDOWS)
1585HANDLE *slirp_get_events(PNATState pData)
1586{
1587 return pData->phEvents;
1588}
1589void slirp_register_external_event(PNATState pData, HANDLE hEvent, int index)
1590{
1591 pData->phEvents[index] = hEvent;
1592}
1593#endif
1594
1595unsigned int slirp_get_timeout_ms(PNATState pData)
1596{
1597 if (link_up)
1598 {
1599 if (time_fasttimo)
1600 return 2;
1601 if (do_slowtimo)
1602 return 500; /* see PR_SLOWHZ */
1603 }
1604 return 0;
1605}
1606
1607#ifndef RT_OS_WINDOWS
1608int slirp_get_nsock(PNATState pData)
1609{
1610 return pData->nsock;
1611}
1612#endif
1613
1614/*
1615 * this function called from NAT thread
1616 */
1617void slirp_post_sent(PNATState pData, void *pvArg)
1618{
1619 struct socket *so = 0;
1620 struct tcpcb *tp = 0;
1621 struct mbuf *m = (struct mbuf *)pvArg;
1622 m_free(pData, m);
1623}
1624#ifdef VBOX_WITH_SLIRP_MT
1625void slirp_process_queue(PNATState pData)
1626{
1627 RTReqProcess(pData->pReqQueue, RT_INDEFINITE_WAIT);
1628}
1629void *slirp_get_queue(PNATState pData)
1630{
1631 return pData->pReqQueue;
1632}
1633#endif
1634
1635uint16_t slirp_get_service(int proto, uint16_t dport, uint16_t sport)
1636{
1637 uint16_t hdport, hsport, service;
1638 hdport = ntohs(dport);
1639 hsport = ntohs(sport);
1640 Log2(("proto: %d, dport: %d sport: %d\n", proto, hdport, hsport));
1641 service = 0;
1642 switch (hdport)
1643 {
1644 case 500:
1645 /* service = sport; */
1646 break;
1647 }
1648 Log2(("service : %d\n", service));
1649 return htons(service);
1650}
Note: See TracBrowser for help on using the repository browser.

© 2025 Oracle Support Privacy / Do Not Sell My Info Terms of Use Trademark Policy Automated Access Etiquette