VirtualBox

source: vbox/trunk/src/VBox/Devices/Network/slirp/slirp.c@ 17225

Last change on this file since 17225 was 17225, checked in by vboxsync, 16 years ago

NAT: return 0 in case of success and -1 in case of error, in dns information fetching routine
  • Property svn:eol-style set to native
File size: 55.4 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 OSVERSIONINFOEX osvi;
329 DWORDLONG condition = 0;
330 int op = VER_GREATER_EQUAL;
331
332 /* DNS information retriving is window's version specific,
333 * so the OS version should be defined first
334 */
335 memset(&osvi, 0, sizeof(OSVERSIONINFOEX));
336 osvi.dwOSVersionInfoSize = sizeof(OSVERSIONINFOEX);
337 osvi.dwMajorVersion = 5; /* Win XP*/
338 osvi.dwMinorVersion = 1;
339
340 VER_SET_CONDITION(condition, VER_MAJORVERSION, op);
341 VER_SET_CONDITION(condition, VER_MINORVERSION, op);
342
343 if (VerifyVersionInfo(&osvi,
344 VER_MAJORVERSION
345 | VER_MINORVERSION,
346 condition))
347 {
348 /* WinXP sp1 and latter */
349 /* Get amount of memory required for operation */
350 ULONG flags = GAA_FLAG_INCLUDE_PREFIX; /*GAA_FLAG_INCLUDE_ALL_INTERFACES;*/ /* all interfaces registered in NDIS */
351 PIP_ADAPTER_ADDRESSES addresses = NULL;
352 PIP_ADAPTER_ADDRESSES addr = NULL;
353 PIP_ADAPTER_DNS_SERVER_ADDRESS dns = NULL;
354 ULONG size = 0;
355 char *suffix;
356 struct dns_entry *da = NULL;
357 ULONG ret = ERROR_SUCCESS;
358
359 /* @todo add SKIPing flags to get only required information */
360
361 ret = pData->pfGetAdaptersAddresses(AF_INET, 0, NULL /* reserved */, addresses, &size);
362 if (ret != ERROR_BUFFER_OVERFLOW)
363 {
364 LogRel(("NAT: error %lu occured on capacity detection operation\n", ret));
365 return -1;
366 }
367
368 if (size == 0)
369 {
370 LogRel(("NAT: Win socket API returns non capacity\n"));
371 return -1;
372 }
373
374 addresses = RTMemAllocZ(size);
375 if (addresses == NULL)
376 {
377 LogRel(("NAT: No memory available \n"));
378 return -1;
379 }
380
381 ret = pData->pfGetAdaptersAddresses(AF_INET, 0, NULL /* reserved */, addresses, &size);
382 if (ret != ERROR_SUCCESS)
383 {
384 LogRel(("NAT: error %lu occured on fetching adapters info\n", ret));
385 return -1;
386 }
387 addr = addresses;
388 while(addr != NULL)
389 {
390 size_t buff_size;
391 if (addr->OperStatus != IfOperStatusUp)
392 goto next;
393 dns = addr->FirstDnsServerAddress;
394 while (dns != NULL)
395 {
396 struct sockaddr *saddr = dns->Address.lpSockaddr;
397 if (saddr->sa_family != AF_INET)
398 goto next_dns;
399 /* add dns server to list */
400 da = RTMemAllocZ(sizeof(struct dns_entry));
401 if (da == NULL)
402 {
403 LogRel(("NAT: Can't allocate buffer for DNS entry\n"));
404 return VERR_NO_MEMORY;
405 }
406 LogRel(("NAT: adding %R[IP4] to DNS server list\n", &((struct sockaddr_in *)saddr)->sin_addr));
407 if ((((struct sockaddr_in *)saddr)->sin_addr.s_addr & htonl(IN_CLASSA_NET)) == ntohl(INADDR_LOOPBACK & IN_CLASSA_NET)) {
408 da->de_addr.s_addr = htonl(ntohl(special_addr.s_addr) | CTL_ALIAS);
409 }
410 else
411 {
412 da->de_addr.s_addr = ((struct sockaddr_in *)saddr)->sin_addr.s_addr;
413 }
414 LIST_INSERT_HEAD(&pData->dns_list_head, da, de_list);
415 next_dns:
416 dns = dns->Next;
417 }
418 buff_size = wcstombs(NULL, addr->DnsSuffix, 0);
419 if (buff_size == 0)
420 goto next;
421 suffix = RTMemAllocZ(buff_size);
422 wcstombs(suffix, addr->DnsSuffix, buff_size);
423 LogRel(("NAT: adding %s to DNS suffix list\n", suffix));
424 *ppszDomain = suffix;
425 next:
426 addr = addr->Next;
427 }
428 /*@todo add dns suffix if required */
429 LogRel(("NAT: adding dns suffix %s to the list \n", ppszDomain));
430 }
431 else
432 {
433 /* Win 2000 and earlier */
434 }
435 return 0;
436}
437# endif /* VBOX_WITH_MULTI_DNS */
438
439#else /* !RT_OS_WINDOWS */
440
441static int get_dns_addr_domain(PNATState pData, bool fVerbose,
442 struct in_addr *pdns_addr,
443 const char **ppszDomain)
444{
445 char buff[512];
446 char buff2[256];
447 FILE *f;
448 int found = 0;
449 struct in_addr tmp_addr;
450
451#ifdef RT_OS_OS2
452 /* Try various locations. */
453 char *etc = getenv("ETC");
454 f = NULL;
455 if (etc)
456 {
457 snprintf(buff, sizeof(buff), "%s/RESOLV2", etc);
458 f = fopen(buff, "rt");
459 }
460 if (!f)
461 {
462 snprintf(buff, sizeof(buff), "%s/RESOLV2", _PATH_ETC);
463 f = fopen(buff, "rt");
464 }
465 if (!f)
466 {
467 snprintf(buff, sizeof(buff), "%s/resolv.conf", _PATH_ETC);
468 f = fopen(buff, "rt");
469 }
470#else
471 f = fopen("/etc/resolv.conf", "r");
472#endif
473 if (!f)
474 return -1;
475
476 if (ppszDomain)
477 *ppszDomain = NULL;
478 Log(("nat: DNS Servers:\n"));
479 while (fgets(buff, 512, f) != NULL)
480 {
481#ifdef VBOX_WITH_MULTI_DNS
482 struct dns_entry *da = NULL;
483#endif
484 if (sscanf(buff, "nameserver%*[ \t]%256s", buff2) == 1)
485 {
486 if (!inet_aton(buff2, &tmp_addr))
487 continue;
488#ifndef VBOX_WITH_MULTI_DNS
489 /* If it's the first one, set it to dns_addr */
490 if (!found)
491 {
492 if (fVerbose || pdns_addr->s_addr != tmp_addr.s_addr)
493 LogRel(("NAT: DNS address: %s\n", buff2));
494 *pdns_addr = tmp_addr;
495 }
496 else
497 {
498 if (fVerbose)
499 LogRel(("NAT: ignored DNS address: %s\n", buff2));
500 }
501#else
502 /*localhost mask */
503 da = RTMemAllocZ(sizeof (struct dns_entry));
504 if (da == NULL)
505 {
506 LogRel(("can't alloc memory for DNS entry\n"));
507 return -1;
508 }
509 /*check */
510 da->de_addr.s_addr = tmp_addr.s_addr;
511 if ((da->de_addr.s_addr & htonl(IN_CLASSA_NET)) == ntohl(INADDR_LOOPBACK & IN_CLASSA_NET)) {
512 da->de_addr.s_addr = htonl(ntohl(special_addr.s_addr) | CTL_ALIAS);
513 }
514 LIST_INSERT_HEAD(&pData->dns_list_head, da, de_list);
515#endif
516 found++;
517 }
518 if ( ppszDomain
519 && (!strncmp(buff, "domain", 6) || !strncmp(buff, "search", 6)))
520 {
521 /* Domain name/search list present. Pick first entry */
522 if (*ppszDomain == NULL)
523 {
524 char *tok;
525 char *saveptr;
526 tok = strtok_r(&buff[6], " \t\n", &saveptr);
527 if (tok)
528 {
529 *ppszDomain = RTStrDup(tok);
530 if (pData->fPassDomain)
531 {
532 if (fVerbose)
533 LogRel(("NAT: passing domain name %s\n", tok));
534 }
535 else
536 Log(("nat: ignoring domain %s\n", tok));
537 }
538 }
539 }
540 }
541 fclose(f);
542 if (!found)
543 return -1;
544 return 0;
545}
546
547#endif
548
549int get_dns_addr(PNATState pData, struct in_addr *pdns_addr)
550{
551 return get_dns_addr_domain(pData, false, pdns_addr, NULL);
552}
553
554int slirp_init(PNATState *ppData, const char *pszNetAddr, uint32_t u32Netmask,
555 bool fPassDomain, const char *pszTFTPPrefix,
556 const char *pszBootFile, void *pvUser)
557{
558 int fNATfailed = 0;
559 int rc;
560 PNATState pData = RTMemAllocZ(sizeof(NATState));
561 *ppData = pData;
562 if (!pData)
563 return VERR_NO_MEMORY;
564 if (u32Netmask & 0x1f)
565 /* CTL is x.x.x.15, bootp passes up to 16 IPs (15..31) */
566 return VERR_INVALID_PARAMETER;
567 pData->fPassDomain = fPassDomain;
568 pData->pvUser = pvUser;
569 tftp_prefix = pszTFTPPrefix;
570 bootp_filename = pszBootFile;
571 pData->netmask = u32Netmask;
572
573#ifdef RT_OS_WINDOWS
574 {
575 WSADATA Data;
576 WSAStartup(MAKEWORD(2, 0), &Data);
577 }
578# if defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC)
579 pData->phEvents[VBOX_SOCKET_EVENT_INDEX] = CreateEvent(NULL, FALSE, FALSE, NULL);
580# endif
581#endif
582#ifdef VBOX_WITH_SLIRP_MT
583 QSOCKET_LOCK_CREATE(tcb);
584 QSOCKET_LOCK_CREATE(udb);
585 rc = RTReqCreateQueue(&pData->pReqQueue);
586 AssertReleaseRC(rc);
587#endif
588
589 link_up = 1;
590
591 debug_init();
592 if_init(pData);
593 ip_init(pData);
594 icmp_init(pData);
595
596 /* Initialise mbufs *after* setting the MTU */
597 m_init(pData);
598
599 inet_aton(pszNetAddr, &special_addr);
600 alias_addr.s_addr = special_addr.s_addr | htonl(CTL_ALIAS);
601 /* @todo: add ability to configure this staff */
602 pData->tftp_server.s_addr = htonl(ntohl(special_addr.s_addr) | CTL_TFTP);
603
604 /* set default addresses */
605 inet_aton("127.0.0.1", &loopback_addr);
606#ifndef VBOX_WITH_MULTI_DNS
607 inet_aton("127.0.0.1", &dns_addr);
608
609 if (get_dns_addr_domain(pData, true, &dns_addr, &pData->pszDomain) < 0)
610#else
611 LIST_INIT(&pData->dns_list_head);
612 if (get_dns_addr_domain(pData, true, NULL, &pData->pszDomain) < 0)
613#endif
614 fNATfailed = 1;
615
616 getouraddr(pData);
617 return fNATfailed ? VINF_NAT_DNS : VINF_SUCCESS;
618}
619
620/**
621 * Statistics counters.
622 */
623void slirp_register_timers(PNATState pData, PPDMDRVINS pDrvIns)
624{
625#ifdef VBOX_WITH_STATISTICS
626 PDMDrvHlpSTAMRegisterF(pDrvIns, &pData->StatFill, STAMTYPE_PROFILE, STAMVISIBILITY_ALWAYS,
627 STAMUNIT_TICKS_PER_CALL, "Profiling slirp fills", "/Drivers/NAT%d/Fill", pDrvIns->iInstance);
628 PDMDrvHlpSTAMRegisterF(pDrvIns, &pData->StatPoll, STAMTYPE_PROFILE, STAMVISIBILITY_ALWAYS,
629 STAMUNIT_TICKS_PER_CALL, "Profiling slirp polls", "/Drivers/NAT%d/Poll", pDrvIns->iInstance);
630 PDMDrvHlpSTAMRegisterF(pDrvIns, &pData->StatFastTimer, STAMTYPE_PROFILE, STAMVISIBILITY_ALWAYS,
631 STAMUNIT_TICKS_PER_CALL, "Profiling slirp fast timer", "/Drivers/NAT%d/TimerFast", pDrvIns->iInstance);
632 PDMDrvHlpSTAMRegisterF(pDrvIns, &pData->StatSlowTimer, STAMTYPE_PROFILE, STAMVISIBILITY_ALWAYS,
633 STAMUNIT_TICKS_PER_CALL, "Profiling slirp slow timer", "/Drivers/NAT%d/TimerSlow", pDrvIns->iInstance);
634 PDMDrvHlpSTAMRegisterF(pDrvIns, &pData->StatTCP, STAMTYPE_COUNTER, STAMVISIBILITY_ALWAYS,
635 STAMUNIT_COUNT, "TCP sockets", "/Drivers/NAT%d/SockTCP", pDrvIns->iInstance);
636 PDMDrvHlpSTAMRegisterF(pDrvIns, &pData->StatTCPHot, STAMTYPE_COUNTER, STAMVISIBILITY_ALWAYS,
637 STAMUNIT_COUNT, "TCP sockets active", "/Drivers/NAT%d/SockTCPHot", pDrvIns->iInstance);
638 PDMDrvHlpSTAMRegisterF(pDrvIns, &pData->StatUDP, STAMTYPE_COUNTER, STAMVISIBILITY_ALWAYS,
639 STAMUNIT_COUNT, "UDP sockets", "/Drivers/NAT%d/SockUDP", pDrvIns->iInstance);
640 PDMDrvHlpSTAMRegisterF(pDrvIns, &pData->StatUDPHot, STAMTYPE_COUNTER, STAMVISIBILITY_ALWAYS,
641 STAMUNIT_COUNT, "UDP sockets active", "/Drivers/NAT%d/SockUDPHot", pDrvIns->iInstance);
642#endif /* VBOX_WITH_STATISTICS */
643}
644
645/**
646 * Marks the link as up, making it possible to establish new connections.
647 */
648void slirp_link_up(PNATState pData)
649{
650 link_up = 1;
651}
652
653/**
654 * Marks the link as down and cleans up the current connections.
655 */
656void slirp_link_down(PNATState pData)
657{
658 struct socket *so;
659
660 while ((so = tcb.so_next) != &tcb)
661 {
662 if (so->so_state & SS_NOFDREF || so->s == -1)
663 sofree(pData, so);
664 else
665 tcp_drop(pData, sototcpcb(so), 0);
666 }
667
668 while ((so = udb.so_next) != &udb)
669 udp_detach(pData, so);
670
671 link_up = 0;
672}
673
674/**
675 * Terminates the slirp component.
676 */
677void slirp_term(PNATState pData)
678{
679#ifdef VBOX_WITH_MULTI_DNS
680 struct dns_entry *de = NULL;
681#endif
682 if (pData->pszDomain)
683 RTStrFree((char *)(void *)pData->pszDomain);
684
685#ifdef RT_OS_WINDOWS
686 pData->pfIcmpCloseHandle(pData->icmp_socket.sh);
687 FreeLibrary(pData->hmIcmpLibrary);
688 RTMemFree(pData->pvIcmpBuffer);
689# else
690 closesocket(pData->icmp_socket.s);
691#endif
692
693 slirp_link_down(pData);
694#ifdef VBOX_WITH_MULTI_DNS
695 while(!LIST_EMPTY(&pData->dns_list_head)) {
696 de = LIST_FIRST(&pData->dns_list_head);
697 LIST_REMOVE(de, de_list);
698 RTMemFree(de);
699 }
700#endif
701#ifdef RT_OS_WINDOWS
702 WSACleanup();
703#endif
704#ifdef LOG_ENABLED
705 Log(("\n"
706 "NAT statistics\n"
707 "--------------\n"
708 "\n"));
709 ipstats(pData);
710 tcpstats(pData);
711 udpstats(pData);
712 icmpstats(pData);
713 mbufstats(pData);
714 sockstats(pData);
715 Log(("\n"
716 "\n"
717 "\n"));
718#endif
719 RTMemFree(pData);
720}
721
722
723#define CONN_CANFSEND(so) (((so)->so_state & (SS_FCANTSENDMORE|SS_ISFCONNECTED)) == SS_ISFCONNECTED)
724#define CONN_CANFRCV(so) (((so)->so_state & (SS_FCANTRCVMORE|SS_ISFCONNECTED)) == SS_ISFCONNECTED)
725#define UPD_NFDS(x) do { if (nfds < (x)) nfds = (x); } while (0)
726
727/*
728 * curtime kept to an accuracy of 1ms
729 */
730static void updtime(PNATState pData)
731{
732#ifdef RT_OS_WINDOWS
733 struct _timeb tb;
734
735 _ftime(&tb);
736 curtime = (u_int)tb.time * (u_int)1000;
737 curtime += (u_int)tb.millitm;
738#else
739 gettimeofday(&tt, 0);
740
741 curtime = (u_int)tt.tv_sec * (u_int)1000;
742 curtime += (u_int)tt.tv_usec / (u_int)1000;
743
744 if ((tt.tv_usec % 1000) >= 500)
745 curtime++;
746#endif
747}
748
749#ifndef VBOX_WITH_SIMPLIFIED_SLIRP_SYNC
750void slirp_select_fill(PNATState pData, int *pnfds,
751 fd_set *readfds, fd_set *writefds, fd_set *xfds)
752#else /* !VBOX_WITH_SIMPLIFIED_SLIRP_SYNC */
753# ifdef RT_OS_WINDOWS
754void slirp_select_fill(PNATState pData, int *pnfds)
755# else /* RT_OS_WINDOWS */
756void slirp_select_fill(PNATState pData, int *pnfds, struct pollfd *polls)
757# endif /* !RT_OS_WINDOWS */
758#endif /* VBOX_WITH_SIMPLIFIED_SLIRP_SYNC */
759{
760 struct socket *so, *so_next;
761 int nfds;
762#if defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) && defined(RT_OS_WINDOWS)
763 int rc;
764 int error;
765#endif
766#if defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) && !defined(RT_OS_WINDOWS)
767 int poll_index = 0;
768#endif
769 int i;
770
771 STAM_PROFILE_START(&pData->StatFill, a);
772
773 nfds = *pnfds;
774
775 /*
776 * First, TCP sockets
777 */
778 do_slowtimo = 0;
779 if (link_up)
780 {
781 /*
782 * *_slowtimo needs calling if there are IP fragments
783 * in the fragment queue, or there are TCP connections active
784 */
785 /* XXX:
786 * triggering of fragment expiration should be the same but use new macroses
787 */
788 do_slowtimo = (tcb.so_next != &tcb);
789 if (!do_slowtimo)
790 {
791 for (i = 0; i < IPREASS_NHASH; i++)
792 {
793 if (!TAILQ_EMPTY(&ipq[i]))
794 {
795 do_slowtimo = 1;
796 break;
797 }
798 }
799 }
800 ICMP_ENGAGE_EVENT(&pData->icmp_socket, readfds);
801
802 STAM_COUNTER_RESET(&pData->StatTCP);
803 STAM_COUNTER_RESET(&pData->StatTCPHot);
804
805 QSOCKET_FOREACH(so, so_next, tcp)
806 /* { */
807#if defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) && !defined(RT_OS_WINDOWS)
808 so->so_poll_index = -1;
809#endif
810 STAM_COUNTER_INC(&pData->StatTCP);
811
812 /*
813 * See if we need a tcp_fasttimo
814 */
815 if ( time_fasttimo == 0
816 && so->so_tcpcb != NULL
817 && so->so_tcpcb->t_flags & TF_DELACK)
818 time_fasttimo = curtime; /* Flag when we want a fasttimo */
819
820 /*
821 * NOFDREF can include still connecting to local-host,
822 * newly socreated() sockets etc. Don't want to select these.
823 */
824 if (so->so_state & SS_NOFDREF || so->s == -1)
825 CONTINUE(tcp);
826
827 /*
828 * Set for reading sockets which are accepting
829 */
830 if (so->so_state & SS_FACCEPTCONN)
831 {
832 STAM_COUNTER_INC(&pData->StatTCPHot);
833 TCP_ENGAGE_EVENT1(so, readfds);
834 CONTINUE(tcp);
835 }
836
837 /*
838 * Set for writing sockets which are connecting
839 */
840 if (so->so_state & SS_ISFCONNECTING)
841 {
842 Log2(("connecting %R[natsock] engaged\n",so));
843 STAM_COUNTER_INC(&pData->StatTCPHot);
844 TCP_ENGAGE_EVENT1(so, writefds);
845 }
846
847 /*
848 * Set for writing if we are connected, can send more, and
849 * we have something to send
850 */
851 if (CONN_CANFSEND(so) && so->so_rcv.sb_cc)
852 {
853 STAM_COUNTER_INC(&pData->StatTCPHot);
854 TCP_ENGAGE_EVENT1(so, writefds);
855 }
856
857 /*
858 * Set for reading (and urgent data) if we are connected, can
859 * receive more, and we have room for it XXX /2 ?
860 */
861 if (CONN_CANFRCV(so) && (so->so_snd.sb_cc < (so->so_snd.sb_datalen/2)))
862 {
863 STAM_COUNTER_INC(&pData->StatTCPHot);
864 TCP_ENGAGE_EVENT2(so, readfds, xfds);
865 }
866 LOOP_LABEL(tcp, so, so_next);
867 }
868
869 /*
870 * UDP sockets
871 */
872 STAM_COUNTER_RESET(&pData->StatUDP);
873 STAM_COUNTER_RESET(&pData->StatUDPHot);
874
875 QSOCKET_FOREACH(so, so_next, udp)
876 /* { */
877
878 STAM_COUNTER_INC(&pData->StatUDP);
879#if defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) && !defined(RT_OS_WINDOWS)
880 so->so_poll_index = -1;
881#endif
882
883 /*
884 * See if it's timed out
885 */
886 if (so->so_expire)
887 {
888 if (so->so_expire <= curtime)
889 {
890#ifdef VBOX_WITH_SLIRP_MT
891 /* we need so_next for continue our cycle*/
892 so_next = so->so_next;
893#endif
894 UDP_DETACH(pData, so, so_next);
895 CONTINUE_NO_UNLOCK(udp);
896 }
897 else
898 do_slowtimo = 1; /* Let socket expire */
899 }
900
901 /*
902 * When UDP packets are received from over the link, they're
903 * sendto()'d straight away, so no need for setting for writing
904 * Limit the number of packets queued by this session to 4.
905 * Note that even though we try and limit this to 4 packets,
906 * the session could have more queued if the packets needed
907 * to be fragmented.
908 *
909 * (XXX <= 4 ?)
910 */
911 if ((so->so_state & SS_ISFCONNECTED) && so->so_queued <= 4)
912 {
913 STAM_COUNTER_INC(&pData->StatUDPHot);
914 UDP_ENGAGE_EVENT(so, readfds);
915 }
916 LOOP_LABEL(udp, so, so_next);
917 }
918
919 }
920
921#if defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC)
922# if defined(RT_OS_WINDOWS)
923 *pnfds = VBOX_EVENT_COUNT;
924# else /* RT_OS_WINDOWS */
925 AssertRelease(poll_index <= *pnfds);
926 *pnfds = poll_index;
927# endif /* !RT_OS_WINDOWS */
928#else /* VBOX_WITH_SIMPLIFIED_SLIRP_SYNC */
929 *pnfds = nfds;
930#endif /* !VBOX_WITH_SIMPLIFIED_SLIRP_SYNC */
931
932 STAM_PROFILE_STOP(&pData->StatFill, a);
933}
934
935#if defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC)
936# if defined(RT_OS_WINDOWS)
937void slirp_select_poll(PNATState pData, int fTimeout, int fIcmp)
938# else /* RT_OS_WINDOWS */
939void slirp_select_poll(PNATState pData, struct pollfd *polls, int ndfs)
940# endif /* !RT_OS_WINDOWS */
941#else /* VBOX_WITH_SIMPLIFIED_SLIRP_SYNC */
942void slirp_select_poll(PNATState pData, fd_set *readfds, fd_set *writefds, fd_set *xfds)
943#endif /* !VBOX_WITH_SIMPLIFIED_SLIRP_SYNC */
944{
945 struct socket *so, *so_next;
946 int ret;
947#if defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) && defined(RT_OS_WINDOWS)
948 WSANETWORKEVENTS NetworkEvents;
949 int rc;
950 int error;
951#endif
952#if defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) && !defined(RT_OS_WINDOWS)
953 int poll_index = 0;
954#endif
955
956 STAM_PROFILE_START(&pData->StatPoll, a);
957
958 /* Update time */
959 updtime(pData);
960
961 /*
962 * See if anything has timed out
963 */
964 if (link_up)
965 {
966 if (time_fasttimo && ((curtime - time_fasttimo) >= 2))
967 {
968 STAM_PROFILE_START(&pData->StatFastTimer, a);
969 tcp_fasttimo(pData);
970 time_fasttimo = 0;
971 STAM_PROFILE_STOP(&pData->StatFastTimer, a);
972 }
973 if (do_slowtimo && ((curtime - last_slowtimo) >= 499))
974 {
975 STAM_PROFILE_START(&pData->StatSlowTimer, a);
976 ip_slowtimo(pData);
977 tcp_slowtimo(pData);
978 last_slowtimo = curtime;
979 STAM_PROFILE_STOP(&pData->StatSlowTimer, a);
980 }
981 }
982#if defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) && defined(RT_OS_WINDOWS)
983 if (fTimeout)
984 return; /* only timer update */
985#endif
986
987 /*
988 * Check sockets
989 */
990 if (link_up)
991 {
992#if defined(RT_OS_WINDOWS)
993 /*XXX: before renaming please make see define
994 * fIcmp in slirp_state.h
995 */
996 if (fIcmp)
997 sorecvfrom(pData, &pData->icmp_socket);
998#else
999 if ( (pData->icmp_socket.s != -1)
1000 && CHECK_FD_SET(&pData->icmp_socket, ignored, readfds))
1001 sorecvfrom(pData, &pData->icmp_socket);
1002#endif
1003 /*
1004 * Check TCP sockets
1005 */
1006 QSOCKET_FOREACH(so, so_next, tcp)
1007 /* { */
1008
1009#ifdef VBOX_WITH_SLIRP_MT
1010 if ( so->so_state & SS_NOFDREF
1011 && so->so_deleted == 1)
1012 {
1013 struct socket *son, *sop = NULL;
1014 QSOCKET_LOCK(tcb);
1015 if (so->so_next != NULL)
1016 {
1017 if (so->so_next != &tcb)
1018 SOCKET_LOCK(so->so_next);
1019 son = so->so_next;
1020 }
1021 if ( so->so_prev != &tcb
1022 && so->so_prev != NULL)
1023 {
1024 SOCKET_LOCK(so->so_prev);
1025 sop = so->so_prev;
1026 }
1027 QSOCKET_UNLOCK(tcb);
1028 remque(pData, so);
1029 NSOCK_DEC();
1030 SOCKET_UNLOCK(so);
1031 SOCKET_LOCK_DESTROY(so);
1032 RTMemFree(so);
1033 so_next = son;
1034 if (sop != NULL)
1035 SOCKET_UNLOCK(sop);
1036 CONTINUE_NO_UNLOCK(tcp);
1037 }
1038#endif
1039 /*
1040 * FD_ISSET is meaningless on these sockets
1041 * (and they can crash the program)
1042 */
1043 if (so->so_state & SS_NOFDREF || so->s == -1)
1044 CONTINUE(tcp);
1045
1046 POLL_TCP_EVENTS(rc, error, so, &NetworkEvents);
1047
1048 LOG_NAT_SOCK(so, TCP, &NetworkEvents, readfds, writefds, xfds);
1049
1050
1051 /*
1052 * Check for URG data
1053 * This will soread as well, so no need to
1054 * test for readfds below if this succeeds
1055 */
1056
1057 /* out-of-band data */
1058 if (CHECK_FD_SET(so, NetworkEvents, xfds))
1059 {
1060 sorecvoob(pData, so);
1061 }
1062
1063 /*
1064 * Check sockets for reading
1065 */
1066 else if ( CHECK_FD_SET(so, NetworkEvents, readfds)
1067 || WIN_CHECK_FD_SET(so, NetworkEvents, acceptds))
1068 {
1069 /*
1070 * Check for incoming connections
1071 */
1072 if (so->so_state & SS_FACCEPTCONN)
1073 {
1074 TCP_CONNECT(pData, so);
1075#if defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) && defined(RT_OS_WINDOWS)
1076 if (!(NetworkEvents.lNetworkEvents & FD_CLOSE))
1077#endif
1078 CONTINUE(tcp);
1079 }
1080
1081 ret = soread(pData, so);
1082 /* Output it if we read something */
1083 if (ret > 0)
1084 TCP_OUTPUT(pData, sototcpcb(so));
1085 }
1086
1087#if defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) && defined(RT_OS_WINDOWS)
1088 /*
1089 * Check for FD_CLOSE events.
1090 * in some cases once FD_CLOSE engaged on socket it could be flashed latter (for some reasons)
1091 */
1092 if ( (NetworkEvents.lNetworkEvents & FD_CLOSE)
1093 || (so->so_close == 1))
1094 {
1095 so->so_close = 1; /* mark it */
1096 /*
1097 * drain the socket
1098 */
1099 for (;;)
1100 {
1101 ret = soread(pData, so);
1102 if (ret > 0)
1103 TCP_OUTPUT(pData, sototcpcb(so));
1104 else
1105 break;
1106 }
1107 CONTINUE(tcp);
1108 }
1109#endif
1110
1111 /*
1112 * Check sockets for writing
1113 */
1114 if (CHECK_FD_SET(so, NetworkEvents, writefds))
1115 {
1116 /*
1117 * Check for non-blocking, still-connecting sockets
1118 */
1119 if (so->so_state & SS_ISFCONNECTING)
1120 {
1121 Log2(("connecting %R[natsock] catched\n", so));
1122 /* Connected */
1123 so->so_state &= ~SS_ISFCONNECTING;
1124
1125 /*
1126 * This should be probably guarded by PROBE_CONN too. Anyway,
1127 * we disable it on OS/2 because the below send call returns
1128 * EFAULT which causes the opened TCP socket to close right
1129 * after it has been opened and connected.
1130 */
1131#ifndef RT_OS_OS2
1132 ret = send(so->s, (const char *)&ret, 0, 0);
1133 if (ret < 0)
1134 {
1135 /* XXXXX Must fix, zero bytes is a NOP */
1136 if ( errno == EAGAIN
1137 || errno == EWOULDBLOCK
1138 || errno == EINPROGRESS
1139 || errno == ENOTCONN)
1140 CONTINUE(tcp);
1141
1142 /* else failed */
1143 so->so_state = SS_NOFDREF;
1144 }
1145 /* else so->so_state &= ~SS_ISFCONNECTING; */
1146#endif
1147
1148 /*
1149 * Continue tcp_input
1150 */
1151 TCP_INPUT(pData, (struct mbuf *)NULL, sizeof(struct ip), so);
1152 /* continue; */
1153 }
1154 else
1155 SOWRITE(ret, pData, so);
1156 /*
1157 * XXX If we wrote something (a lot), there could be the need
1158 * for a window update. In the worst case, the remote will send
1159 * a window probe to get things going again.
1160 */
1161 }
1162
1163 /*
1164 * Probe a still-connecting, non-blocking socket
1165 * to check if it's still alive
1166 */
1167#ifdef PROBE_CONN
1168 if (so->so_state & SS_ISFCONNECTING)
1169 {
1170 ret = recv(so->s, (char *)&ret, 0, 0);
1171
1172 if (ret < 0)
1173 {
1174 /* XXX */
1175 if ( errno == EAGAIN
1176 || errno == EWOULDBLOCK
1177 || errno == EINPROGRESS
1178 || errno == ENOTCONN)
1179 {
1180 CONTINUE(tcp); /* Still connecting, continue */
1181 }
1182
1183 /* else failed */
1184 so->so_state = SS_NOFDREF;
1185
1186 /* tcp_input will take care of it */
1187 }
1188 else
1189 {
1190 ret = send(so->s, &ret, 0, 0);
1191 if (ret < 0)
1192 {
1193 /* XXX */
1194 if ( errno == EAGAIN
1195 || errno == EWOULDBLOCK
1196 || errno == EINPROGRESS
1197 || errno == ENOTCONN)
1198 {
1199 CONTINUE(tcp);
1200 }
1201 /* else failed */
1202 so->so_state = SS_NOFDREF;
1203 }
1204 else
1205 so->so_state &= ~SS_ISFCONNECTING;
1206
1207 }
1208 TCP_INPUT((struct mbuf *)NULL, sizeof(struct ip),so);
1209 } /* SS_ISFCONNECTING */
1210#endif
1211#ifndef RT_OS_WINDOWS
1212 if ( UNIX_CHECK_FD_SET(so, NetworkEvents, rdhup)
1213 || UNIX_CHECK_FD_SET(so, NetworkEvents, rderr))
1214 {
1215 int err;
1216 int inq, outq;
1217 int status;
1218 socklen_t optlen = sizeof(int);
1219 inq = outq = 0;
1220 status = getsockopt(so->s, SOL_SOCKET, SO_ERROR, &err, &optlen);
1221 if (status != 0)
1222 LogRel(("NAT: can't get error status from %R[natsock]\n", so));
1223#ifndef RT_OS_SOLARIS
1224 status = ioctl(so->s, FIONREAD, &inq); /* tcp(7) recommends SIOCINQ which is Linux specific */
1225 if (status != 0 || status != EINVAL)
1226 {
1227 /* EINVAL returned if socket in listen state tcp(7)*/
1228 LogRel(("NAT: can't get depth of IN queue status from %R[natsock]\n", so));
1229 }
1230 status = ioctl(so->s, TIOCOUTQ, &outq); /* SIOCOUTQ see previous comment */
1231 if (status != 0)
1232 LogRel(("NAT: can't get depth of OUT queue from %R[natsock]\n", so));
1233#else
1234 /*
1235 * Solaris has bit different ioctl commands and its handlings
1236 * hint: streamio(7) I_NREAD
1237 */
1238#endif
1239 if ( so->so_state & SS_ISFCONNECTING
1240 || UNIX_CHECK_FD_SET(so, NetworkEvents, readfds))
1241 {
1242 /**
1243 * Check if we need here take care about gracefull connection
1244 * @todo try with proxy server
1245 */
1246 if (UNIX_CHECK_FD_SET(so, NetworkEvents, readfds))
1247 {
1248 /*
1249 * Never meet inq != 0 or outq != 0, anyway let it stay for a while
1250 * in case it happens we'll able to detect it.
1251 * Give TCP/IP stack wait or expire the socket.
1252 */
1253 LogRel(("NAT: %R[natsock] err(%d:%s) s(in:%d,out:%d)happens on read I/O, "
1254 "other side close connection \n", so, err, strerror(err), inq, outq));
1255 CONTINUE(tcp);
1256 }
1257 goto tcp_input_close;
1258 }
1259 if ( !UNIX_CHECK_FD_SET(so, NetworkEvents, readfds)
1260 && !UNIX_CHECK_FD_SET(so, NetworkEvents, writefds)
1261 && !UNIX_CHECK_FD_SET(so, NetworkEvents, xfds))
1262 {
1263 LogRel(("NAT: system expires the socket %R[natsock] err(%d:%s) s(in:%d,out:%d) happens on non-I/O. ",
1264 so, err, strerror(err), inq, outq));
1265 goto tcp_input_close;
1266 }
1267 LogRel(("NAT: %R[natsock] we've met(%d:%s) s(in:%d, out:%d) unhandled combination hup (%d) "
1268 "rederr(%d) on (r:%d, w:%d, x:%d)\n",
1269 so, err, strerror(err),
1270 inq, outq,
1271 UNIX_CHECK_FD_SET(so, ign, rdhup),
1272 UNIX_CHECK_FD_SET(so, ign, rderr),
1273 UNIX_CHECK_FD_SET(so, ign, readfds),
1274 UNIX_CHECK_FD_SET(so, ign, writefds),
1275 UNIX_CHECK_FD_SET(so, ign, xfds)));
1276 /*
1277 * Give OS's TCP/IP stack a chance to resolve an issue or expire the socket.
1278 */
1279 CONTINUE(tcp);
1280tcp_input_close:
1281 so->so_state = SS_NOFDREF; /*cause connection valid tcp connection termination and socket closing */
1282 TCP_INPUT(pData, (struct mbuf *)NULL, sizeof(struct ip), so);
1283 CONTINUE(tcp);
1284 }
1285#endif
1286 LOOP_LABEL(tcp, so, so_next);
1287 }
1288
1289 /*
1290 * Now UDP sockets.
1291 * Incoming packets are sent straight away, they're not buffered.
1292 * Incoming UDP data isn't buffered either.
1293 */
1294 QSOCKET_FOREACH(so, so_next, udp)
1295 /* { */
1296#ifdef VBOX_WITH_SLIRP_MT
1297 if ( so->so_state & SS_NOFDREF
1298 && so->so_deleted == 1)
1299 {
1300 struct socket *son, *sop = NULL;
1301 QSOCKET_LOCK(udb);
1302 if (so->so_next != NULL)
1303 {
1304 if (so->so_next != &udb)
1305 SOCKET_LOCK(so->so_next);
1306 son = so->so_next;
1307 }
1308 if ( so->so_prev != &udb
1309 && so->so_prev != NULL)
1310 {
1311 SOCKET_LOCK(so->so_prev);
1312 sop = so->so_prev;
1313 }
1314 QSOCKET_UNLOCK(udb);
1315 remque(pData, so);
1316 NSOCK_DEC();
1317 SOCKET_UNLOCK(so);
1318 SOCKET_LOCK_DESTROY(so);
1319 RTMemFree(so);
1320 so_next = son;
1321 if (sop != NULL)
1322 SOCKET_UNLOCK(sop);
1323 CONTINUE_NO_UNLOCK(udp);
1324 }
1325#endif
1326 POLL_UDP_EVENTS(rc, error, so, &NetworkEvents);
1327
1328 LOG_NAT_SOCK(so, UDP, &NetworkEvents, readfds, writefds, xfds);
1329
1330 if (so->s != -1 && CHECK_FD_SET(so, NetworkEvents, readfds))
1331 {
1332 SORECVFROM(pData, so);
1333 }
1334 LOOP_LABEL(udp, so, so_next);
1335 }
1336
1337 }
1338
1339#ifndef VBOX_WITH_SLIRP_MT
1340 /*
1341 * See if we can start outputting
1342 */
1343 if (if_queued && link_up)
1344 if_start(pData);
1345#endif
1346
1347 STAM_PROFILE_STOP(&pData->StatPoll, a);
1348}
1349
1350#define ETH_ALEN 6
1351#define ETH_HLEN 14
1352
1353#define ARPOP_REQUEST 1 /* ARP request */
1354#define ARPOP_REPLY 2 /* ARP reply */
1355
1356struct ethhdr
1357{
1358 unsigned char h_dest[ETH_ALEN]; /* destination eth addr */
1359 unsigned char h_source[ETH_ALEN]; /* source ether addr */
1360 unsigned short h_proto; /* packet type ID field */
1361};
1362AssertCompileSize(struct ethhdr, 14);
1363
1364struct arphdr
1365{
1366 unsigned short ar_hrd; /* format of hardware address */
1367 unsigned short ar_pro; /* format of protocol address */
1368 unsigned char ar_hln; /* length of hardware address */
1369 unsigned char ar_pln; /* length of protocol address */
1370 unsigned short ar_op; /* ARP opcode (command) */
1371
1372 /*
1373 * Ethernet looks like this : This bit is variable sized however...
1374 */
1375 unsigned char ar_sha[ETH_ALEN]; /* sender hardware address */
1376 unsigned char ar_sip[4]; /* sender IP address */
1377 unsigned char ar_tha[ETH_ALEN]; /* target hardware address */
1378 unsigned char ar_tip[4]; /* target IP address */
1379};
1380AssertCompileSize(struct arphdr, 28);
1381
1382#ifdef VBOX_WITH_SIMPLIFIED_SLIRP_SYNC
1383static void arp_input(PNATState pData, struct mbuf *m)
1384#else
1385static void arp_input(PNATState pData, const uint8_t *pkt, int pkt_len)
1386#endif
1387{
1388 struct ethhdr *eh;
1389 struct ethhdr *reh;
1390 struct arphdr *ah;
1391 struct arphdr *rah;
1392 int ar_op;
1393 struct ex_list *ex_ptr;
1394 uint32_t htip;
1395#ifndef VBOX_WITH_SIMPLIFIED_SLIRP_SYNC
1396 uint8_t arp_reply[sizeof(struct arphdr) + ETH_HLEN];
1397 eh = (struct ethhdr *)pkt;
1398#else
1399 struct mbuf *mr;
1400 eh = mtod(m, struct ethhdr *);
1401#endif
1402 ah = (struct arphdr *)&eh[1];
1403 htip = ntohl(*(uint32_t*)ah->ar_tip);
1404
1405#ifdef VBOX_WITH_SIMPLIFIED_SLIRP_SYNC
1406 mr = m_get(pData);
1407 mr->m_data += if_maxlinkhdr;
1408 mr->m_len = sizeof(struct arphdr);
1409 rah = mtod(mr, struct arphdr *);
1410#else
1411 reh = (struct ethhdr *)arp_reply;
1412 rah = (struct arphdr *)&reh[1];
1413#endif
1414
1415 ar_op = ntohs(ah->ar_op);
1416 switch(ar_op)
1417 {
1418 case ARPOP_REQUEST:
1419 if ((htip & pData->netmask) == ntohl(special_addr.s_addr))
1420 {
1421 if ( CTL_CHECK(htip, CTL_DNS)
1422 || CTL_CHECK(htip, CTL_ALIAS)
1423 || CTL_CHECK(htip, CTL_TFTP))
1424 goto arp_ok;
1425 for (ex_ptr = exec_list; ex_ptr; ex_ptr = ex_ptr->ex_next)
1426 {
1427 if ((htip & ~pData->netmask) == ex_ptr->ex_addr)
1428 goto arp_ok;
1429 }
1430 return;
1431 arp_ok:
1432
1433#ifndef VBOX_WITH_SIMPLIFIED_SLIRP_SYNC
1434 memcpy(reh->h_dest, eh->h_source, ETH_ALEN);
1435 memcpy(reh->h_source, &special_addr, ETH_ALEN);
1436 reh->h_source[5] = ah->ar_tip[3];
1437 reh->h_proto = htons(ETH_P_ARP);
1438#endif
1439 rah->ar_hrd = htons(1);
1440 rah->ar_pro = htons(ETH_P_IP);
1441 rah->ar_hln = ETH_ALEN;
1442 rah->ar_pln = 4;
1443 rah->ar_op = htons(ARPOP_REPLY);
1444 memcpy(rah->ar_sha, special_ethaddr, ETH_ALEN);
1445
1446 switch (htip & ~pData->netmask)
1447 {
1448 case CTL_DNS:
1449 case CTL_ALIAS:
1450 rah->ar_sha[5] = (uint8_t)(htip & ~pData->netmask);
1451 break;
1452 default:;
1453 }
1454
1455 memcpy(rah->ar_sip, ah->ar_tip, 4);
1456 memcpy(rah->ar_tha, ah->ar_sha, ETH_ALEN);
1457 memcpy(rah->ar_tip, ah->ar_sip, 4);
1458#ifdef VBOX_WITH_SIMPLIFIED_SLIRP_SYNC
1459 if_encap(pData, ETH_P_ARP, mr);
1460 m_free(pData, m);
1461#else
1462 slirp_output(pData->pvUser, arp_reply, sizeof(arp_reply));
1463#endif
1464 }
1465 break;
1466 default:
1467 break;
1468 }
1469}
1470
1471void slirp_input(PNATState pData, const uint8_t *pkt, int pkt_len)
1472{
1473 struct mbuf *m;
1474 int proto;
1475 static bool fWarnedIpv6;
1476
1477 if (pkt_len < ETH_HLEN)
1478 {
1479 LogRel(("NAT: packet having size %d has been ingnored\n", pkt_len));
1480 return;
1481 }
1482
1483 m = m_get(pData);
1484 if (!m)
1485 {
1486 LogRel(("NAT: can't allocate new mbuf\n"));
1487 return;
1488 }
1489
1490 /* Note: we add to align the IP header */
1491
1492 if (M_FREEROOM(m) < pkt_len)
1493 m_inc(m, pkt_len);
1494
1495 m->m_len = pkt_len ;
1496 memcpy(m->m_data, pkt, pkt_len);
1497
1498 proto = ntohs(*(uint16_t *)(pkt + 12));
1499 switch(proto)
1500 {
1501 case ETH_P_ARP:
1502#ifdef VBOX_WITH_SIMPLIFIED_SLIRP_SYNC
1503 arp_input(pData, m);
1504#else
1505 arp_input(pData, pkt, pkt_len);
1506 m_free(pData, m);
1507#endif
1508 break;
1509 case ETH_P_IP:
1510 /* Update time. Important if the network is very quiet, as otherwise
1511 * the first outgoing connection gets an incorrect timestamp. */
1512 updtime(pData);
1513 m->m_data += ETH_HLEN;
1514 m->m_len -= ETH_HLEN;
1515 ip_input(pData, m);
1516 break;
1517 case ETH_P_IPV6:
1518 m_free(pData, m);
1519 if (!fWarnedIpv6)
1520 {
1521 LogRel(("NAT: IPv6 not supported\n"));
1522 fWarnedIpv6 = true;
1523 }
1524 break;
1525 default:
1526 LogRel(("NAT: Unsupported protocol %x\n", proto));
1527 m_free(pData, m);
1528 break;
1529 }
1530 RTMemFree((void *)pkt);
1531}
1532
1533/* output the IP packet to the ethernet device */
1534#ifdef VBOX_WITH_SIMPLIFIED_SLIRP_SYNC
1535void if_encap(PNATState pData, uint16_t eth_proto, struct mbuf *m)
1536#else
1537void if_encap(PNATState pData, uint8_t *ip_data, int ip_data_len)
1538#endif
1539{
1540#ifdef VBOX_WITH_SIMPLIFIED_SLIRP_SYNC
1541 struct ethhdr *eh;
1542 uint8_t *buf = RTMemAlloc(1600);
1543 m->m_data -= if_maxlinkhdr;
1544 m->m_len += ETH_HLEN;
1545 eh = mtod(m, struct ethhdr *);
1546#else
1547 uint8_t buf[1600];
1548 struct ethhdr *eh = (struct ethhdr *)buf;
1549
1550 if (ip_data_len + ETH_HLEN > sizeof(buf))
1551 return;
1552
1553 memcpy(buf + sizeof(struct ethhdr), ip_data, ip_data_len);
1554#endif
1555
1556
1557 memcpy(eh->h_dest, client_ethaddr, ETH_ALEN);
1558 memcpy(eh->h_source, special_ethaddr, ETH_ALEN - 1);
1559 /* XXX: not correct */
1560 eh->h_source[5] = CTL_ALIAS;
1561#ifdef VBOX_WITH_SIMPLIFIED_SLIRP_SYNC
1562 eh->h_proto = htons(eth_proto);
1563#if 0
1564 slirp_output(pData->pvUser, m, mtod(m, uint8_t *), m->m_len);
1565#else
1566 memcpy(buf, mtod(m, uint8_t *), m->m_len);
1567 slirp_output(pData->pvUser, NULL, buf, m->m_len);
1568 m_free(pData, m);
1569#endif
1570#else
1571 eh->h_proto = htons(ETH_P_IP);
1572 slirp_output(pData->pvUser, buf, ip_data_len + ETH_HLEN);
1573#endif
1574}
1575
1576int slirp_redir(PNATState pData, int is_udp, int host_port,
1577 struct in_addr guest_addr, int guest_port)
1578{
1579 struct socket *so;
1580 Log2(("NAT: set redirect %s hp:%d gp:%d\n", (is_udp?"UDP":"TCP"), host_port, guest_port));
1581 if (is_udp)
1582 {
1583 so = udp_listen(pData, htons(host_port), guest_addr.s_addr,
1584 htons(guest_port), 0);
1585 }
1586 else
1587 {
1588 so = solisten(pData, htons(host_port), guest_addr.s_addr,
1589 htons(guest_port), 0);
1590 }
1591 Log2(("NAT: redirecting socket %R[natsock]\n", so));
1592 return (so != NULL ? 0 : -1);
1593}
1594
1595int slirp_add_exec(PNATState pData, int do_pty, const char *args, int addr_low_byte,
1596 int guest_port)
1597{
1598 return add_exec(&exec_list, do_pty, (char *)args,
1599 addr_low_byte, htons(guest_port));
1600}
1601
1602void slirp_set_ethaddr(PNATState pData, const uint8_t *ethaddr)
1603{
1604 memcpy(client_ethaddr, ethaddr, ETH_ALEN);
1605}
1606
1607#if defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) && defined(RT_OS_WINDOWS)
1608HANDLE *slirp_get_events(PNATState pData)
1609{
1610 return pData->phEvents;
1611}
1612void slirp_register_external_event(PNATState pData, HANDLE hEvent, int index)
1613{
1614 pData->phEvents[index] = hEvent;
1615}
1616#endif
1617
1618unsigned int slirp_get_timeout_ms(PNATState pData)
1619{
1620 if (link_up)
1621 {
1622 if (time_fasttimo)
1623 return 2;
1624 if (do_slowtimo)
1625 return 500; /* see PR_SLOWHZ */
1626 }
1627 return 0;
1628}
1629
1630#ifndef RT_OS_WINDOWS
1631int slirp_get_nsock(PNATState pData)
1632{
1633 return pData->nsock;
1634}
1635#endif
1636
1637/*
1638 * this function called from NAT thread
1639 */
1640void slirp_post_sent(PNATState pData, void *pvArg)
1641{
1642 struct socket *so = 0;
1643 struct tcpcb *tp = 0;
1644 struct mbuf *m = (struct mbuf *)pvArg;
1645 m_free(pData, m);
1646}
1647#ifdef VBOX_WITH_SLIRP_MT
1648void slirp_process_queue(PNATState pData)
1649{
1650 RTReqProcess(pData->pReqQueue, RT_INDEFINITE_WAIT);
1651}
1652void *slirp_get_queue(PNATState pData)
1653{
1654 return pData->pReqQueue;
1655}
1656#endif
1657
1658uint16_t slirp_get_service(int proto, uint16_t dport, uint16_t sport)
1659{
1660 uint16_t hdport, hsport, service;
1661 hdport = ntohs(dport);
1662 hsport = ntohs(sport);
1663 Log2(("proto: %d, dport: %d sport: %d\n", proto, hdport, hsport));
1664 service = 0;
1665 switch (hdport)
1666 {
1667 case 500:
1668 /* service = sport; */
1669 break;
1670 }
1671 Log2(("service : %d\n", service));
1672 return htons(service);
1673}
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