VirtualBox

source: vbox/trunk/src/VBox/Devices/Network/slirp/slirp.c@ 17227

Last change on this file since 17227 was 17227, checked in by vboxsync, 16 years ago

NAT: removing Win version checks (we don't support Win version prior Win XP)
  • 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
590 getouraddr(pData);
591 return fNATfailed ? VINF_NAT_DNS : VINF_SUCCESS;
592}
593
594/**
595 * Statistics counters.
596 */
597void slirp_register_timers(PNATState pData, PPDMDRVINS pDrvIns)
598{
599#ifdef VBOX_WITH_STATISTICS
600 PDMDrvHlpSTAMRegisterF(pDrvIns, &pData->StatFill, STAMTYPE_PROFILE, STAMVISIBILITY_ALWAYS,
601 STAMUNIT_TICKS_PER_CALL, "Profiling slirp fills", "/Drivers/NAT%d/Fill", pDrvIns->iInstance);
602 PDMDrvHlpSTAMRegisterF(pDrvIns, &pData->StatPoll, STAMTYPE_PROFILE, STAMVISIBILITY_ALWAYS,
603 STAMUNIT_TICKS_PER_CALL, "Profiling slirp polls", "/Drivers/NAT%d/Poll", pDrvIns->iInstance);
604 PDMDrvHlpSTAMRegisterF(pDrvIns, &pData->StatFastTimer, STAMTYPE_PROFILE, STAMVISIBILITY_ALWAYS,
605 STAMUNIT_TICKS_PER_CALL, "Profiling slirp fast timer", "/Drivers/NAT%d/TimerFast", pDrvIns->iInstance);
606 PDMDrvHlpSTAMRegisterF(pDrvIns, &pData->StatSlowTimer, STAMTYPE_PROFILE, STAMVISIBILITY_ALWAYS,
607 STAMUNIT_TICKS_PER_CALL, "Profiling slirp slow timer", "/Drivers/NAT%d/TimerSlow", pDrvIns->iInstance);
608 PDMDrvHlpSTAMRegisterF(pDrvIns, &pData->StatTCP, STAMTYPE_COUNTER, STAMVISIBILITY_ALWAYS,
609 STAMUNIT_COUNT, "TCP sockets", "/Drivers/NAT%d/SockTCP", pDrvIns->iInstance);
610 PDMDrvHlpSTAMRegisterF(pDrvIns, &pData->StatTCPHot, STAMTYPE_COUNTER, STAMVISIBILITY_ALWAYS,
611 STAMUNIT_COUNT, "TCP sockets active", "/Drivers/NAT%d/SockTCPHot", pDrvIns->iInstance);
612 PDMDrvHlpSTAMRegisterF(pDrvIns, &pData->StatUDP, STAMTYPE_COUNTER, STAMVISIBILITY_ALWAYS,
613 STAMUNIT_COUNT, "UDP sockets", "/Drivers/NAT%d/SockUDP", pDrvIns->iInstance);
614 PDMDrvHlpSTAMRegisterF(pDrvIns, &pData->StatUDPHot, STAMTYPE_COUNTER, STAMVISIBILITY_ALWAYS,
615 STAMUNIT_COUNT, "UDP sockets active", "/Drivers/NAT%d/SockUDPHot", pDrvIns->iInstance);
616#endif /* VBOX_WITH_STATISTICS */
617}
618
619/**
620 * Marks the link as up, making it possible to establish new connections.
621 */
622void slirp_link_up(PNATState pData)
623{
624 link_up = 1;
625}
626
627/**
628 * Marks the link as down and cleans up the current connections.
629 */
630void slirp_link_down(PNATState pData)
631{
632 struct socket *so;
633
634 while ((so = tcb.so_next) != &tcb)
635 {
636 if (so->so_state & SS_NOFDREF || so->s == -1)
637 sofree(pData, so);
638 else
639 tcp_drop(pData, sototcpcb(so), 0);
640 }
641
642 while ((so = udb.so_next) != &udb)
643 udp_detach(pData, so);
644
645 link_up = 0;
646}
647
648/**
649 * Terminates the slirp component.
650 */
651void slirp_term(PNATState pData)
652{
653#ifdef VBOX_WITH_MULTI_DNS
654 struct dns_entry *de = NULL;
655#endif
656 if (pData->pszDomain)
657 RTStrFree((char *)(void *)pData->pszDomain);
658
659#ifdef RT_OS_WINDOWS
660 pData->pfIcmpCloseHandle(pData->icmp_socket.sh);
661 FreeLibrary(pData->hmIcmpLibrary);
662 RTMemFree(pData->pvIcmpBuffer);
663# else
664 closesocket(pData->icmp_socket.s);
665#endif
666
667 slirp_link_down(pData);
668#ifdef VBOX_WITH_MULTI_DNS
669 while(!LIST_EMPTY(&pData->dns_list_head)) {
670 de = LIST_FIRST(&pData->dns_list_head);
671 LIST_REMOVE(de, de_list);
672 RTMemFree(de);
673 }
674#endif
675#ifdef RT_OS_WINDOWS
676 WSACleanup();
677#endif
678#ifdef LOG_ENABLED
679 Log(("\n"
680 "NAT statistics\n"
681 "--------------\n"
682 "\n"));
683 ipstats(pData);
684 tcpstats(pData);
685 udpstats(pData);
686 icmpstats(pData);
687 mbufstats(pData);
688 sockstats(pData);
689 Log(("\n"
690 "\n"
691 "\n"));
692#endif
693 RTMemFree(pData);
694}
695
696
697#define CONN_CANFSEND(so) (((so)->so_state & (SS_FCANTSENDMORE|SS_ISFCONNECTED)) == SS_ISFCONNECTED)
698#define CONN_CANFRCV(so) (((so)->so_state & (SS_FCANTRCVMORE|SS_ISFCONNECTED)) == SS_ISFCONNECTED)
699#define UPD_NFDS(x) do { if (nfds < (x)) nfds = (x); } while (0)
700
701/*
702 * curtime kept to an accuracy of 1ms
703 */
704static void updtime(PNATState pData)
705{
706#ifdef RT_OS_WINDOWS
707 struct _timeb tb;
708
709 _ftime(&tb);
710 curtime = (u_int)tb.time * (u_int)1000;
711 curtime += (u_int)tb.millitm;
712#else
713 gettimeofday(&tt, 0);
714
715 curtime = (u_int)tt.tv_sec * (u_int)1000;
716 curtime += (u_int)tt.tv_usec / (u_int)1000;
717
718 if ((tt.tv_usec % 1000) >= 500)
719 curtime++;
720#endif
721}
722
723#ifndef VBOX_WITH_SIMPLIFIED_SLIRP_SYNC
724void slirp_select_fill(PNATState pData, int *pnfds,
725 fd_set *readfds, fd_set *writefds, fd_set *xfds)
726#else /* !VBOX_WITH_SIMPLIFIED_SLIRP_SYNC */
727# ifdef RT_OS_WINDOWS
728void slirp_select_fill(PNATState pData, int *pnfds)
729# else /* RT_OS_WINDOWS */
730void slirp_select_fill(PNATState pData, int *pnfds, struct pollfd *polls)
731# endif /* !RT_OS_WINDOWS */
732#endif /* VBOX_WITH_SIMPLIFIED_SLIRP_SYNC */
733{
734 struct socket *so, *so_next;
735 int nfds;
736#if defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) && defined(RT_OS_WINDOWS)
737 int rc;
738 int error;
739#endif
740#if defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) && !defined(RT_OS_WINDOWS)
741 int poll_index = 0;
742#endif
743 int i;
744
745 STAM_PROFILE_START(&pData->StatFill, a);
746
747 nfds = *pnfds;
748
749 /*
750 * First, TCP sockets
751 */
752 do_slowtimo = 0;
753 if (link_up)
754 {
755 /*
756 * *_slowtimo needs calling if there are IP fragments
757 * in the fragment queue, or there are TCP connections active
758 */
759 /* XXX:
760 * triggering of fragment expiration should be the same but use new macroses
761 */
762 do_slowtimo = (tcb.so_next != &tcb);
763 if (!do_slowtimo)
764 {
765 for (i = 0; i < IPREASS_NHASH; i++)
766 {
767 if (!TAILQ_EMPTY(&ipq[i]))
768 {
769 do_slowtimo = 1;
770 break;
771 }
772 }
773 }
774 ICMP_ENGAGE_EVENT(&pData->icmp_socket, readfds);
775
776 STAM_COUNTER_RESET(&pData->StatTCP);
777 STAM_COUNTER_RESET(&pData->StatTCPHot);
778
779 QSOCKET_FOREACH(so, so_next, tcp)
780 /* { */
781#if defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) && !defined(RT_OS_WINDOWS)
782 so->so_poll_index = -1;
783#endif
784 STAM_COUNTER_INC(&pData->StatTCP);
785
786 /*
787 * See if we need a tcp_fasttimo
788 */
789 if ( time_fasttimo == 0
790 && so->so_tcpcb != NULL
791 && so->so_tcpcb->t_flags & TF_DELACK)
792 time_fasttimo = curtime; /* Flag when we want a fasttimo */
793
794 /*
795 * NOFDREF can include still connecting to local-host,
796 * newly socreated() sockets etc. Don't want to select these.
797 */
798 if (so->so_state & SS_NOFDREF || so->s == -1)
799 CONTINUE(tcp);
800
801 /*
802 * Set for reading sockets which are accepting
803 */
804 if (so->so_state & SS_FACCEPTCONN)
805 {
806 STAM_COUNTER_INC(&pData->StatTCPHot);
807 TCP_ENGAGE_EVENT1(so, readfds);
808 CONTINUE(tcp);
809 }
810
811 /*
812 * Set for writing sockets which are connecting
813 */
814 if (so->so_state & SS_ISFCONNECTING)
815 {
816 Log2(("connecting %R[natsock] engaged\n",so));
817 STAM_COUNTER_INC(&pData->StatTCPHot);
818 TCP_ENGAGE_EVENT1(so, writefds);
819 }
820
821 /*
822 * Set for writing if we are connected, can send more, and
823 * we have something to send
824 */
825 if (CONN_CANFSEND(so) && so->so_rcv.sb_cc)
826 {
827 STAM_COUNTER_INC(&pData->StatTCPHot);
828 TCP_ENGAGE_EVENT1(so, writefds);
829 }
830
831 /*
832 * Set for reading (and urgent data) if we are connected, can
833 * receive more, and we have room for it XXX /2 ?
834 */
835 if (CONN_CANFRCV(so) && (so->so_snd.sb_cc < (so->so_snd.sb_datalen/2)))
836 {
837 STAM_COUNTER_INC(&pData->StatTCPHot);
838 TCP_ENGAGE_EVENT2(so, readfds, xfds);
839 }
840 LOOP_LABEL(tcp, so, so_next);
841 }
842
843 /*
844 * UDP sockets
845 */
846 STAM_COUNTER_RESET(&pData->StatUDP);
847 STAM_COUNTER_RESET(&pData->StatUDPHot);
848
849 QSOCKET_FOREACH(so, so_next, udp)
850 /* { */
851
852 STAM_COUNTER_INC(&pData->StatUDP);
853#if defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) && !defined(RT_OS_WINDOWS)
854 so->so_poll_index = -1;
855#endif
856
857 /*
858 * See if it's timed out
859 */
860 if (so->so_expire)
861 {
862 if (so->so_expire <= curtime)
863 {
864#ifdef VBOX_WITH_SLIRP_MT
865 /* we need so_next for continue our cycle*/
866 so_next = so->so_next;
867#endif
868 UDP_DETACH(pData, so, so_next);
869 CONTINUE_NO_UNLOCK(udp);
870 }
871 else
872 do_slowtimo = 1; /* Let socket expire */
873 }
874
875 /*
876 * When UDP packets are received from over the link, they're
877 * sendto()'d straight away, so no need for setting for writing
878 * Limit the number of packets queued by this session to 4.
879 * Note that even though we try and limit this to 4 packets,
880 * the session could have more queued if the packets needed
881 * to be fragmented.
882 *
883 * (XXX <= 4 ?)
884 */
885 if ((so->so_state & SS_ISFCONNECTED) && so->so_queued <= 4)
886 {
887 STAM_COUNTER_INC(&pData->StatUDPHot);
888 UDP_ENGAGE_EVENT(so, readfds);
889 }
890 LOOP_LABEL(udp, so, so_next);
891 }
892
893 }
894
895#if defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC)
896# if defined(RT_OS_WINDOWS)
897 *pnfds = VBOX_EVENT_COUNT;
898# else /* RT_OS_WINDOWS */
899 AssertRelease(poll_index <= *pnfds);
900 *pnfds = poll_index;
901# endif /* !RT_OS_WINDOWS */
902#else /* VBOX_WITH_SIMPLIFIED_SLIRP_SYNC */
903 *pnfds = nfds;
904#endif /* !VBOX_WITH_SIMPLIFIED_SLIRP_SYNC */
905
906 STAM_PROFILE_STOP(&pData->StatFill, a);
907}
908
909#if defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC)
910# if defined(RT_OS_WINDOWS)
911void slirp_select_poll(PNATState pData, int fTimeout, int fIcmp)
912# else /* RT_OS_WINDOWS */
913void slirp_select_poll(PNATState pData, struct pollfd *polls, int ndfs)
914# endif /* !RT_OS_WINDOWS */
915#else /* VBOX_WITH_SIMPLIFIED_SLIRP_SYNC */
916void slirp_select_poll(PNATState pData, fd_set *readfds, fd_set *writefds, fd_set *xfds)
917#endif /* !VBOX_WITH_SIMPLIFIED_SLIRP_SYNC */
918{
919 struct socket *so, *so_next;
920 int ret;
921#if defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) && defined(RT_OS_WINDOWS)
922 WSANETWORKEVENTS NetworkEvents;
923 int rc;
924 int error;
925#endif
926#if defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) && !defined(RT_OS_WINDOWS)
927 int poll_index = 0;
928#endif
929
930 STAM_PROFILE_START(&pData->StatPoll, a);
931
932 /* Update time */
933 updtime(pData);
934
935 /*
936 * See if anything has timed out
937 */
938 if (link_up)
939 {
940 if (time_fasttimo && ((curtime - time_fasttimo) >= 2))
941 {
942 STAM_PROFILE_START(&pData->StatFastTimer, a);
943 tcp_fasttimo(pData);
944 time_fasttimo = 0;
945 STAM_PROFILE_STOP(&pData->StatFastTimer, a);
946 }
947 if (do_slowtimo && ((curtime - last_slowtimo) >= 499))
948 {
949 STAM_PROFILE_START(&pData->StatSlowTimer, a);
950 ip_slowtimo(pData);
951 tcp_slowtimo(pData);
952 last_slowtimo = curtime;
953 STAM_PROFILE_STOP(&pData->StatSlowTimer, a);
954 }
955 }
956#if defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) && defined(RT_OS_WINDOWS)
957 if (fTimeout)
958 return; /* only timer update */
959#endif
960
961 /*
962 * Check sockets
963 */
964 if (link_up)
965 {
966#if defined(RT_OS_WINDOWS)
967 /*XXX: before renaming please make see define
968 * fIcmp in slirp_state.h
969 */
970 if (fIcmp)
971 sorecvfrom(pData, &pData->icmp_socket);
972#else
973 if ( (pData->icmp_socket.s != -1)
974 && CHECK_FD_SET(&pData->icmp_socket, ignored, readfds))
975 sorecvfrom(pData, &pData->icmp_socket);
976#endif
977 /*
978 * Check TCP sockets
979 */
980 QSOCKET_FOREACH(so, so_next, tcp)
981 /* { */
982
983#ifdef VBOX_WITH_SLIRP_MT
984 if ( so->so_state & SS_NOFDREF
985 && so->so_deleted == 1)
986 {
987 struct socket *son, *sop = NULL;
988 QSOCKET_LOCK(tcb);
989 if (so->so_next != NULL)
990 {
991 if (so->so_next != &tcb)
992 SOCKET_LOCK(so->so_next);
993 son = so->so_next;
994 }
995 if ( so->so_prev != &tcb
996 && so->so_prev != NULL)
997 {
998 SOCKET_LOCK(so->so_prev);
999 sop = so->so_prev;
1000 }
1001 QSOCKET_UNLOCK(tcb);
1002 remque(pData, so);
1003 NSOCK_DEC();
1004 SOCKET_UNLOCK(so);
1005 SOCKET_LOCK_DESTROY(so);
1006 RTMemFree(so);
1007 so_next = son;
1008 if (sop != NULL)
1009 SOCKET_UNLOCK(sop);
1010 CONTINUE_NO_UNLOCK(tcp);
1011 }
1012#endif
1013 /*
1014 * FD_ISSET is meaningless on these sockets
1015 * (and they can crash the program)
1016 */
1017 if (so->so_state & SS_NOFDREF || so->s == -1)
1018 CONTINUE(tcp);
1019
1020 POLL_TCP_EVENTS(rc, error, so, &NetworkEvents);
1021
1022 LOG_NAT_SOCK(so, TCP, &NetworkEvents, readfds, writefds, xfds);
1023
1024
1025 /*
1026 * Check for URG data
1027 * This will soread as well, so no need to
1028 * test for readfds below if this succeeds
1029 */
1030
1031 /* out-of-band data */
1032 if (CHECK_FD_SET(so, NetworkEvents, xfds))
1033 {
1034 sorecvoob(pData, so);
1035 }
1036
1037 /*
1038 * Check sockets for reading
1039 */
1040 else if ( CHECK_FD_SET(so, NetworkEvents, readfds)
1041 || WIN_CHECK_FD_SET(so, NetworkEvents, acceptds))
1042 {
1043 /*
1044 * Check for incoming connections
1045 */
1046 if (so->so_state & SS_FACCEPTCONN)
1047 {
1048 TCP_CONNECT(pData, so);
1049#if defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) && defined(RT_OS_WINDOWS)
1050 if (!(NetworkEvents.lNetworkEvents & FD_CLOSE))
1051#endif
1052 CONTINUE(tcp);
1053 }
1054
1055 ret = soread(pData, so);
1056 /* Output it if we read something */
1057 if (ret > 0)
1058 TCP_OUTPUT(pData, sototcpcb(so));
1059 }
1060
1061#if defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) && defined(RT_OS_WINDOWS)
1062 /*
1063 * Check for FD_CLOSE events.
1064 * in some cases once FD_CLOSE engaged on socket it could be flashed latter (for some reasons)
1065 */
1066 if ( (NetworkEvents.lNetworkEvents & FD_CLOSE)
1067 || (so->so_close == 1))
1068 {
1069 so->so_close = 1; /* mark it */
1070 /*
1071 * drain the socket
1072 */
1073 for (;;)
1074 {
1075 ret = soread(pData, so);
1076 if (ret > 0)
1077 TCP_OUTPUT(pData, sototcpcb(so));
1078 else
1079 break;
1080 }
1081 CONTINUE(tcp);
1082 }
1083#endif
1084
1085 /*
1086 * Check sockets for writing
1087 */
1088 if (CHECK_FD_SET(so, NetworkEvents, writefds))
1089 {
1090 /*
1091 * Check for non-blocking, still-connecting sockets
1092 */
1093 if (so->so_state & SS_ISFCONNECTING)
1094 {
1095 Log2(("connecting %R[natsock] catched\n", so));
1096 /* Connected */
1097 so->so_state &= ~SS_ISFCONNECTING;
1098
1099 /*
1100 * This should be probably guarded by PROBE_CONN too. Anyway,
1101 * we disable it on OS/2 because the below send call returns
1102 * EFAULT which causes the opened TCP socket to close right
1103 * after it has been opened and connected.
1104 */
1105#ifndef RT_OS_OS2
1106 ret = send(so->s, (const char *)&ret, 0, 0);
1107 if (ret < 0)
1108 {
1109 /* XXXXX Must fix, zero bytes is a NOP */
1110 if ( errno == EAGAIN
1111 || errno == EWOULDBLOCK
1112 || errno == EINPROGRESS
1113 || errno == ENOTCONN)
1114 CONTINUE(tcp);
1115
1116 /* else failed */
1117 so->so_state = SS_NOFDREF;
1118 }
1119 /* else so->so_state &= ~SS_ISFCONNECTING; */
1120#endif
1121
1122 /*
1123 * Continue tcp_input
1124 */
1125 TCP_INPUT(pData, (struct mbuf *)NULL, sizeof(struct ip), so);
1126 /* continue; */
1127 }
1128 else
1129 SOWRITE(ret, pData, so);
1130 /*
1131 * XXX If we wrote something (a lot), there could be the need
1132 * for a window update. In the worst case, the remote will send
1133 * a window probe to get things going again.
1134 */
1135 }
1136
1137 /*
1138 * Probe a still-connecting, non-blocking socket
1139 * to check if it's still alive
1140 */
1141#ifdef PROBE_CONN
1142 if (so->so_state & SS_ISFCONNECTING)
1143 {
1144 ret = recv(so->s, (char *)&ret, 0, 0);
1145
1146 if (ret < 0)
1147 {
1148 /* XXX */
1149 if ( errno == EAGAIN
1150 || errno == EWOULDBLOCK
1151 || errno == EINPROGRESS
1152 || errno == ENOTCONN)
1153 {
1154 CONTINUE(tcp); /* Still connecting, continue */
1155 }
1156
1157 /* else failed */
1158 so->so_state = SS_NOFDREF;
1159
1160 /* tcp_input will take care of it */
1161 }
1162 else
1163 {
1164 ret = send(so->s, &ret, 0, 0);
1165 if (ret < 0)
1166 {
1167 /* XXX */
1168 if ( errno == EAGAIN
1169 || errno == EWOULDBLOCK
1170 || errno == EINPROGRESS
1171 || errno == ENOTCONN)
1172 {
1173 CONTINUE(tcp);
1174 }
1175 /* else failed */
1176 so->so_state = SS_NOFDREF;
1177 }
1178 else
1179 so->so_state &= ~SS_ISFCONNECTING;
1180
1181 }
1182 TCP_INPUT((struct mbuf *)NULL, sizeof(struct ip),so);
1183 } /* SS_ISFCONNECTING */
1184#endif
1185#ifndef RT_OS_WINDOWS
1186 if ( UNIX_CHECK_FD_SET(so, NetworkEvents, rdhup)
1187 || UNIX_CHECK_FD_SET(so, NetworkEvents, rderr))
1188 {
1189 int err;
1190 int inq, outq;
1191 int status;
1192 socklen_t optlen = sizeof(int);
1193 inq = outq = 0;
1194 status = getsockopt(so->s, SOL_SOCKET, SO_ERROR, &err, &optlen);
1195 if (status != 0)
1196 LogRel(("NAT: can't get error status from %R[natsock]\n", so));
1197#ifndef RT_OS_SOLARIS
1198 status = ioctl(so->s, FIONREAD, &inq); /* tcp(7) recommends SIOCINQ which is Linux specific */
1199 if (status != 0 || status != EINVAL)
1200 {
1201 /* EINVAL returned if socket in listen state tcp(7)*/
1202 LogRel(("NAT: can't get depth of IN queue status from %R[natsock]\n", so));
1203 }
1204 status = ioctl(so->s, TIOCOUTQ, &outq); /* SIOCOUTQ see previous comment */
1205 if (status != 0)
1206 LogRel(("NAT: can't get depth of OUT queue from %R[natsock]\n", so));
1207#else
1208 /*
1209 * Solaris has bit different ioctl commands and its handlings
1210 * hint: streamio(7) I_NREAD
1211 */
1212#endif
1213 if ( so->so_state & SS_ISFCONNECTING
1214 || UNIX_CHECK_FD_SET(so, NetworkEvents, readfds))
1215 {
1216 /**
1217 * Check if we need here take care about gracefull connection
1218 * @todo try with proxy server
1219 */
1220 if (UNIX_CHECK_FD_SET(so, NetworkEvents, readfds))
1221 {
1222 /*
1223 * Never meet inq != 0 or outq != 0, anyway let it stay for a while
1224 * in case it happens we'll able to detect it.
1225 * Give TCP/IP stack wait or expire the socket.
1226 */
1227 LogRel(("NAT: %R[natsock] err(%d:%s) s(in:%d,out:%d)happens on read I/O, "
1228 "other side close connection \n", so, err, strerror(err), inq, outq));
1229 CONTINUE(tcp);
1230 }
1231 goto tcp_input_close;
1232 }
1233 if ( !UNIX_CHECK_FD_SET(so, NetworkEvents, readfds)
1234 && !UNIX_CHECK_FD_SET(so, NetworkEvents, writefds)
1235 && !UNIX_CHECK_FD_SET(so, NetworkEvents, xfds))
1236 {
1237 LogRel(("NAT: system expires the socket %R[natsock] err(%d:%s) s(in:%d,out:%d) happens on non-I/O. ",
1238 so, err, strerror(err), inq, outq));
1239 goto tcp_input_close;
1240 }
1241 LogRel(("NAT: %R[natsock] we've met(%d:%s) s(in:%d, out:%d) unhandled combination hup (%d) "
1242 "rederr(%d) on (r:%d, w:%d, x:%d)\n",
1243 so, err, strerror(err),
1244 inq, outq,
1245 UNIX_CHECK_FD_SET(so, ign, rdhup),
1246 UNIX_CHECK_FD_SET(so, ign, rderr),
1247 UNIX_CHECK_FD_SET(so, ign, readfds),
1248 UNIX_CHECK_FD_SET(so, ign, writefds),
1249 UNIX_CHECK_FD_SET(so, ign, xfds)));
1250 /*
1251 * Give OS's TCP/IP stack a chance to resolve an issue or expire the socket.
1252 */
1253 CONTINUE(tcp);
1254tcp_input_close:
1255 so->so_state = SS_NOFDREF; /*cause connection valid tcp connection termination and socket closing */
1256 TCP_INPUT(pData, (struct mbuf *)NULL, sizeof(struct ip), so);
1257 CONTINUE(tcp);
1258 }
1259#endif
1260 LOOP_LABEL(tcp, so, so_next);
1261 }
1262
1263 /*
1264 * Now UDP sockets.
1265 * Incoming packets are sent straight away, they're not buffered.
1266 * Incoming UDP data isn't buffered either.
1267 */
1268 QSOCKET_FOREACH(so, so_next, udp)
1269 /* { */
1270#ifdef VBOX_WITH_SLIRP_MT
1271 if ( so->so_state & SS_NOFDREF
1272 && so->so_deleted == 1)
1273 {
1274 struct socket *son, *sop = NULL;
1275 QSOCKET_LOCK(udb);
1276 if (so->so_next != NULL)
1277 {
1278 if (so->so_next != &udb)
1279 SOCKET_LOCK(so->so_next);
1280 son = so->so_next;
1281 }
1282 if ( so->so_prev != &udb
1283 && so->so_prev != NULL)
1284 {
1285 SOCKET_LOCK(so->so_prev);
1286 sop = so->so_prev;
1287 }
1288 QSOCKET_UNLOCK(udb);
1289 remque(pData, so);
1290 NSOCK_DEC();
1291 SOCKET_UNLOCK(so);
1292 SOCKET_LOCK_DESTROY(so);
1293 RTMemFree(so);
1294 so_next = son;
1295 if (sop != NULL)
1296 SOCKET_UNLOCK(sop);
1297 CONTINUE_NO_UNLOCK(udp);
1298 }
1299#endif
1300 POLL_UDP_EVENTS(rc, error, so, &NetworkEvents);
1301
1302 LOG_NAT_SOCK(so, UDP, &NetworkEvents, readfds, writefds, xfds);
1303
1304 if (so->s != -1 && CHECK_FD_SET(so, NetworkEvents, readfds))
1305 {
1306 SORECVFROM(pData, so);
1307 }
1308 LOOP_LABEL(udp, so, so_next);
1309 }
1310
1311 }
1312
1313#ifndef VBOX_WITH_SLIRP_MT
1314 /*
1315 * See if we can start outputting
1316 */
1317 if (if_queued && link_up)
1318 if_start(pData);
1319#endif
1320
1321 STAM_PROFILE_STOP(&pData->StatPoll, a);
1322}
1323
1324#define ETH_ALEN 6
1325#define ETH_HLEN 14
1326
1327#define ARPOP_REQUEST 1 /* ARP request */
1328#define ARPOP_REPLY 2 /* ARP reply */
1329
1330struct ethhdr
1331{
1332 unsigned char h_dest[ETH_ALEN]; /* destination eth addr */
1333 unsigned char h_source[ETH_ALEN]; /* source ether addr */
1334 unsigned short h_proto; /* packet type ID field */
1335};
1336AssertCompileSize(struct ethhdr, 14);
1337
1338struct arphdr
1339{
1340 unsigned short ar_hrd; /* format of hardware address */
1341 unsigned short ar_pro; /* format of protocol address */
1342 unsigned char ar_hln; /* length of hardware address */
1343 unsigned char ar_pln; /* length of protocol address */
1344 unsigned short ar_op; /* ARP opcode (command) */
1345
1346 /*
1347 * Ethernet looks like this : This bit is variable sized however...
1348 */
1349 unsigned char ar_sha[ETH_ALEN]; /* sender hardware address */
1350 unsigned char ar_sip[4]; /* sender IP address */
1351 unsigned char ar_tha[ETH_ALEN]; /* target hardware address */
1352 unsigned char ar_tip[4]; /* target IP address */
1353};
1354AssertCompileSize(struct arphdr, 28);
1355
1356#ifdef VBOX_WITH_SIMPLIFIED_SLIRP_SYNC
1357static void arp_input(PNATState pData, struct mbuf *m)
1358#else
1359static void arp_input(PNATState pData, const uint8_t *pkt, int pkt_len)
1360#endif
1361{
1362 struct ethhdr *eh;
1363 struct ethhdr *reh;
1364 struct arphdr *ah;
1365 struct arphdr *rah;
1366 int ar_op;
1367 struct ex_list *ex_ptr;
1368 uint32_t htip;
1369#ifndef VBOX_WITH_SIMPLIFIED_SLIRP_SYNC
1370 uint8_t arp_reply[sizeof(struct arphdr) + ETH_HLEN];
1371 eh = (struct ethhdr *)pkt;
1372#else
1373 struct mbuf *mr;
1374 eh = mtod(m, struct ethhdr *);
1375#endif
1376 ah = (struct arphdr *)&eh[1];
1377 htip = ntohl(*(uint32_t*)ah->ar_tip);
1378
1379#ifdef VBOX_WITH_SIMPLIFIED_SLIRP_SYNC
1380 mr = m_get(pData);
1381 mr->m_data += if_maxlinkhdr;
1382 mr->m_len = sizeof(struct arphdr);
1383 rah = mtod(mr, struct arphdr *);
1384#else
1385 reh = (struct ethhdr *)arp_reply;
1386 rah = (struct arphdr *)&reh[1];
1387#endif
1388
1389 ar_op = ntohs(ah->ar_op);
1390 switch(ar_op)
1391 {
1392 case ARPOP_REQUEST:
1393 if ((htip & pData->netmask) == ntohl(special_addr.s_addr))
1394 {
1395 if ( CTL_CHECK(htip, CTL_DNS)
1396 || CTL_CHECK(htip, CTL_ALIAS)
1397 || CTL_CHECK(htip, CTL_TFTP))
1398 goto arp_ok;
1399 for (ex_ptr = exec_list; ex_ptr; ex_ptr = ex_ptr->ex_next)
1400 {
1401 if ((htip & ~pData->netmask) == ex_ptr->ex_addr)
1402 goto arp_ok;
1403 }
1404 return;
1405 arp_ok:
1406
1407#ifndef VBOX_WITH_SIMPLIFIED_SLIRP_SYNC
1408 memcpy(reh->h_dest, eh->h_source, ETH_ALEN);
1409 memcpy(reh->h_source, &special_addr, ETH_ALEN);
1410 reh->h_source[5] = ah->ar_tip[3];
1411 reh->h_proto = htons(ETH_P_ARP);
1412#endif
1413 rah->ar_hrd = htons(1);
1414 rah->ar_pro = htons(ETH_P_IP);
1415 rah->ar_hln = ETH_ALEN;
1416 rah->ar_pln = 4;
1417 rah->ar_op = htons(ARPOP_REPLY);
1418 memcpy(rah->ar_sha, special_ethaddr, ETH_ALEN);
1419
1420 switch (htip & ~pData->netmask)
1421 {
1422 case CTL_DNS:
1423 case CTL_ALIAS:
1424 rah->ar_sha[5] = (uint8_t)(htip & ~pData->netmask);
1425 break;
1426 default:;
1427 }
1428
1429 memcpy(rah->ar_sip, ah->ar_tip, 4);
1430 memcpy(rah->ar_tha, ah->ar_sha, ETH_ALEN);
1431 memcpy(rah->ar_tip, ah->ar_sip, 4);
1432#ifdef VBOX_WITH_SIMPLIFIED_SLIRP_SYNC
1433 if_encap(pData, ETH_P_ARP, mr);
1434 m_free(pData, m);
1435#else
1436 slirp_output(pData->pvUser, arp_reply, sizeof(arp_reply));
1437#endif
1438 }
1439 break;
1440 default:
1441 break;
1442 }
1443}
1444
1445void slirp_input(PNATState pData, const uint8_t *pkt, int pkt_len)
1446{
1447 struct mbuf *m;
1448 int proto;
1449 static bool fWarnedIpv6;
1450
1451 if (pkt_len < ETH_HLEN)
1452 {
1453 LogRel(("NAT: packet having size %d has been ingnored\n", pkt_len));
1454 return;
1455 }
1456
1457 m = m_get(pData);
1458 if (!m)
1459 {
1460 LogRel(("NAT: can't allocate new mbuf\n"));
1461 return;
1462 }
1463
1464 /* Note: we add to align the IP header */
1465
1466 if (M_FREEROOM(m) < pkt_len)
1467 m_inc(m, pkt_len);
1468
1469 m->m_len = pkt_len ;
1470 memcpy(m->m_data, pkt, pkt_len);
1471
1472 proto = ntohs(*(uint16_t *)(pkt + 12));
1473 switch(proto)
1474 {
1475 case ETH_P_ARP:
1476#ifdef VBOX_WITH_SIMPLIFIED_SLIRP_SYNC
1477 arp_input(pData, m);
1478#else
1479 arp_input(pData, pkt, pkt_len);
1480 m_free(pData, m);
1481#endif
1482 break;
1483 case ETH_P_IP:
1484 /* Update time. Important if the network is very quiet, as otherwise
1485 * the first outgoing connection gets an incorrect timestamp. */
1486 updtime(pData);
1487 m->m_data += ETH_HLEN;
1488 m->m_len -= ETH_HLEN;
1489 ip_input(pData, m);
1490 break;
1491 case ETH_P_IPV6:
1492 m_free(pData, m);
1493 if (!fWarnedIpv6)
1494 {
1495 LogRel(("NAT: IPv6 not supported\n"));
1496 fWarnedIpv6 = true;
1497 }
1498 break;
1499 default:
1500 LogRel(("NAT: Unsupported protocol %x\n", proto));
1501 m_free(pData, m);
1502 break;
1503 }
1504 RTMemFree((void *)pkt);
1505}
1506
1507/* output the IP packet to the ethernet device */
1508#ifdef VBOX_WITH_SIMPLIFIED_SLIRP_SYNC
1509void if_encap(PNATState pData, uint16_t eth_proto, struct mbuf *m)
1510#else
1511void if_encap(PNATState pData, uint8_t *ip_data, int ip_data_len)
1512#endif
1513{
1514#ifdef VBOX_WITH_SIMPLIFIED_SLIRP_SYNC
1515 struct ethhdr *eh;
1516 uint8_t *buf = RTMemAlloc(1600);
1517 m->m_data -= if_maxlinkhdr;
1518 m->m_len += ETH_HLEN;
1519 eh = mtod(m, struct ethhdr *);
1520#else
1521 uint8_t buf[1600];
1522 struct ethhdr *eh = (struct ethhdr *)buf;
1523
1524 if (ip_data_len + ETH_HLEN > sizeof(buf))
1525 return;
1526
1527 memcpy(buf + sizeof(struct ethhdr), ip_data, ip_data_len);
1528#endif
1529
1530
1531 memcpy(eh->h_dest, client_ethaddr, ETH_ALEN);
1532 memcpy(eh->h_source, special_ethaddr, ETH_ALEN - 1);
1533 /* XXX: not correct */
1534 eh->h_source[5] = CTL_ALIAS;
1535#ifdef VBOX_WITH_SIMPLIFIED_SLIRP_SYNC
1536 eh->h_proto = htons(eth_proto);
1537#if 0
1538 slirp_output(pData->pvUser, m, mtod(m, uint8_t *), m->m_len);
1539#else
1540 memcpy(buf, mtod(m, uint8_t *), m->m_len);
1541 slirp_output(pData->pvUser, NULL, buf, m->m_len);
1542 m_free(pData, m);
1543#endif
1544#else
1545 eh->h_proto = htons(ETH_P_IP);
1546 slirp_output(pData->pvUser, buf, ip_data_len + ETH_HLEN);
1547#endif
1548}
1549
1550int slirp_redir(PNATState pData, int is_udp, int host_port,
1551 struct in_addr guest_addr, int guest_port)
1552{
1553 struct socket *so;
1554 Log2(("NAT: set redirect %s hp:%d gp:%d\n", (is_udp?"UDP":"TCP"), host_port, guest_port));
1555 if (is_udp)
1556 {
1557 so = udp_listen(pData, htons(host_port), guest_addr.s_addr,
1558 htons(guest_port), 0);
1559 }
1560 else
1561 {
1562 so = solisten(pData, htons(host_port), guest_addr.s_addr,
1563 htons(guest_port), 0);
1564 }
1565 Log2(("NAT: redirecting socket %R[natsock]\n", so));
1566 return (so != NULL ? 0 : -1);
1567}
1568
1569int slirp_add_exec(PNATState pData, int do_pty, const char *args, int addr_low_byte,
1570 int guest_port)
1571{
1572 return add_exec(&exec_list, do_pty, (char *)args,
1573 addr_low_byte, htons(guest_port));
1574}
1575
1576void slirp_set_ethaddr(PNATState pData, const uint8_t *ethaddr)
1577{
1578 memcpy(client_ethaddr, ethaddr, ETH_ALEN);
1579}
1580
1581#if defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) && defined(RT_OS_WINDOWS)
1582HANDLE *slirp_get_events(PNATState pData)
1583{
1584 return pData->phEvents;
1585}
1586void slirp_register_external_event(PNATState pData, HANDLE hEvent, int index)
1587{
1588 pData->phEvents[index] = hEvent;
1589}
1590#endif
1591
1592unsigned int slirp_get_timeout_ms(PNATState pData)
1593{
1594 if (link_up)
1595 {
1596 if (time_fasttimo)
1597 return 2;
1598 if (do_slowtimo)
1599 return 500; /* see PR_SLOWHZ */
1600 }
1601 return 0;
1602}
1603
1604#ifndef RT_OS_WINDOWS
1605int slirp_get_nsock(PNATState pData)
1606{
1607 return pData->nsock;
1608}
1609#endif
1610
1611/*
1612 * this function called from NAT thread
1613 */
1614void slirp_post_sent(PNATState pData, void *pvArg)
1615{
1616 struct socket *so = 0;
1617 struct tcpcb *tp = 0;
1618 struct mbuf *m = (struct mbuf *)pvArg;
1619 m_free(pData, m);
1620}
1621#ifdef VBOX_WITH_SLIRP_MT
1622void slirp_process_queue(PNATState pData)
1623{
1624 RTReqProcess(pData->pReqQueue, RT_INDEFINITE_WAIT);
1625}
1626void *slirp_get_queue(PNATState pData)
1627{
1628 return pData->pReqQueue;
1629}
1630#endif
1631
1632uint16_t slirp_get_service(int proto, uint16_t dport, uint16_t sport)
1633{
1634 uint16_t hdport, hsport, service;
1635 hdport = ntohs(dport);
1636 hsport = ntohs(sport);
1637 Log2(("proto: %d, dport: %d sport: %d\n", proto, hdport, hsport));
1638 service = 0;
1639 switch (hdport)
1640 {
1641 case 500:
1642 /* service = sport; */
1643 break;
1644 }
1645 Log2(("service : %d\n", service));
1646 return htons(service);
1647}
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