VirtualBox

source: vbox/trunk/src/VBox/Devices/Network/slirp/slirp.c@ 16936

Last change on this file since 16936 was 16936, checked in by vboxsync, 16 years ago

NAT: old sync mode
  • Property svn:eol-style set to native
File size: 51.2 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#endif
13
14#if !defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) || !defined(RT_OS_WINDOWS)
15
16# ifndef VBOX_WITH_SIMPLIFIED_SLIRP_SYNC
17# define DO_ENGAGE_EVENT1(so, fdset, label) \
18 do { \
19 FD_SET((so)->s, (fdset)); \
20 UPD_NFDS((so)->s); \
21 } while(0)
22
23
24# define DO_ENGAGE_EVENT2(so, fdset1, fdset2, label) \
25 do { \
26 FD_SET((so)->s, (fdset1)); \
27 FD_SET((so)->s, (fdset2)); \
28 UPD_NFDS((so)->s); \
29 } while(0)
30
31# define DO_POLL_EVENTS(rc, error, so, events, label) do {} while (0)
32
33# define DO_CHECK_FD_SET(so, events, fdset) (FD_ISSET((so)->s, fdset))
34# define DO_UNIX_CHECK_FD_SET(so, events, fdset ) 0 /*specific for Unix API */
35# else /* !VBOX_WITH_SIMPLIFIED_SLIRP_SYNC */
36# define DO_ENGAGE_EVENT1(so, fdset, label) \
37 do { \
38 if( so->so_poll_index != -1 \
39 && so->s == polls[so->so_poll_index].fd) { \
40 polls[so->so_poll_index].events |= N_(fdset ## _poll); \
41 break; /* out of this loop */ \
42 } \
43 AssertRelease(poll_index < (nfds)); \
44 AssertRelease(poll_index >= 0 && poll_index < (nfds)); \
45 polls[poll_index].fd = (so)->s; \
46 (so)->so_poll_index = poll_index; \
47 polls[poll_index].events = N_(fdset ## _poll); \
48 polls[poll_index].revents = 0; \
49 poll_index++; \
50 } while(0)
51
52
53# define DO_ENGAGE_EVENT2(so, fdset1, fdset2, label) \
54 do { \
55 if( so->so_poll_index != -1 \
56 && so->s == polls[so->so_poll_index].fd) { \
57 polls[so->so_poll_index].events |= \
58 N_(fdset1 ## _poll) | N_(fdset1 ## _poll); \
59 break; /* out of this loop */ \
60 } \
61 AssertRelease(poll_index < (nfds)); \
62 polls[poll_index].fd = (so)->s; \
63 (so)->so_poll_index = poll_index; \
64 polls[poll_index].events = \
65 N_(fdset1 ## _poll) | N_(fdset1 ## _poll); \
66 poll_index++; \
67 } while(0)
68
69# define DO_POLL_EVENTS(rc, error, so, events, label) do {} while (0)
70
71# define DO_CHECK_FD_SET(so, events, fdset) ( ((so)->so_poll_index != -1) \
72 && ((so)->so_poll_index <= ndfs) \
73 && ((so)->s == polls[so->so_poll_index].fd) \
74 && (polls[(so)->so_poll_index].revents & N_(fdset ## _poll)))
75# define DO_UNIX_CHECK_FD_SET(so, events, fdset ) DO_CHECK_FD_SET((so), (events), fdset) /*specific for Unix API */
76# define DO_WIN_CHECK_FD_SET(so, events, fdset ) 0 /* specific for Windows Winsock API */
77# endif /* VBOX_WITH_SIMPLIFIED_SLIRP_SYNC */
78
79# ifndef RT_OS_WINDOWS
80
81# ifndef RT_OS_LINUX
82# define readfds_poll (POLLRDNORM)
83# define writefds_poll (POLLWRNORM)
84# define xfds_poll (POLLRDBAND|POLLWRBAND|POLLPRI)
85# else
86# define readfds_poll (POLLIN)
87# define writefds_poll (POLLOUT)
88# define xfds_poll (POLLPRI)
89# endif
90# define rderr_poll (POLLERR)
91# define rdhup_poll (POLLHUP)
92# define nval_poll (POLLNVAL)
93
94# define ICMP_ENGAGE_EVENT(so, fdset) \
95 do { \
96 if (pData->icmp_socket.s != -1) \
97 DO_ENGAGE_EVENT1((so), fdset, ICMP); \
98 } while (0)
99# else /* !RT_OS_WINDOWS */
100# ifdef VBOX_WITH_SIMPLIFIED_SLIRP_SYNC
101# define DO_WIN_CHECK_FD_SET(so, events, fdset ) DO_CHECK_FD_SET((so), (events), fdset)
102# else /* VBOX_WITH_SIMPLIFIED_SLIRP_SYNC */
103# define DO_WIN_CHECK_FD_SET(so, events, fdset ) 0
104# endif /* !VBOX_WITH_SIMPLIFIED_SLIRP_SYNC */
105# define ICMP_ENGAGE_EVENT(so, fdset) do {} while(0)
106#endif /* RT_OS_WINDOWS */
107
108#else /* defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) && defined(RT_OS_WINDOWS) */
109
110/*
111 * On Windows, we will be notified by IcmpSendEcho2() when the response arrives.
112 * So no call to WSAEventSelect necessary.
113 */
114# define ICMP_ENGAGE_EVENT(so, fdset) do {} while(0)
115
116# define DO_ENGAGE_EVENT1(so, fdset1, label) \
117 do { \
118 rc = WSAEventSelect((so)->s, VBOX_SOCKET_EVENT, FD_ALL_EVENTS); \
119 if (rc == SOCKET_ERROR) \
120 { \
121 /* This should not happen */ \
122 error = WSAGetLastError(); \
123 LogRel(("WSAEventSelect (" #label ") error %d (so=%x, socket=%s, event=%x)\n", \
124 error, (so), (so)->s, VBOX_SOCKET_EVENT)); \
125 } \
126 } while(0); \
127 CONTINUE(label)
128
129# define DO_ENGAGE_EVENT2(so, fdset1, fdset2, label) \
130 DO_ENGAGE_EVENT1((so), (fdset1), label)
131
132# define DO_POLL_EVENTS(rc, error, so, events, label) \
133 (rc) = WSAEnumNetworkEvents((so)->s, VBOX_SOCKET_EVENT, (events)); \
134 if ((rc) == SOCKET_ERROR) \
135 { \
136 (error) = WSAGetLastError(); \
137 LogRel(("WSAEnumNetworkEvents " #label " error %d\n", (error))); \
138 CONTINUE(label); \
139 }
140
141# define acceptds_win FD_ACCEPT
142# define acceptds_win_bit FD_ACCEPT_BIT
143
144# define readfds_win FD_READ
145# define readfds_win_bit FD_READ_BIT
146
147# define writefds_win FD_WRITE
148# define writefds_win_bit FD_WRITE_BIT
149
150# define xfds_win FD_OOB
151# define xfds_win_bit FD_OOB_BIT
152
153# define DO_CHECK_FD_SET(so, events, fdset) \
154 (((events).lNetworkEvents & fdset ## _win) && ((events).iErrorCode[fdset ## _win_bit] == 0))
155
156# define DO_WIN_CHECK_FD_SET(so, events, fdset ) DO_CHECK_FD_SET((so), (events), fdset)
157# define DO_UNIX_CHECK_FD_SET(so, events, fdset ) 1 /*specific for Unix API */
158
159#endif /* defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) && defined(RT_OS_WINDOWS) */
160
161#define TCP_ENGAGE_EVENT1(so, fdset) \
162 DO_ENGAGE_EVENT1((so), fdset, tcp)
163
164#define TCP_ENGAGE_EVENT2(so, fdset1, fdset2) \
165 DO_ENGAGE_EVENT2((so), fdset1, fdset2, tcp)
166
167#define UDP_ENGAGE_EVENT(so, fdset) \
168 DO_ENGAGE_EVENT1((so), fdset, udp)
169
170#define POLL_TCP_EVENTS(rc, error, so, events) \
171 DO_POLL_EVENTS((rc), (error), (so), (events), tcp)
172
173#define POLL_UDP_EVENTS(rc, error, so, events) \
174 DO_POLL_EVENTS((rc), (error), (so), (events), udp)
175
176#define CHECK_FD_SET(so, events, set) \
177 (DO_CHECK_FD_SET((so), (events), set))
178
179#define WIN_CHECK_FD_SET(so, events, set) \
180 (DO_WIN_CHECK_FD_SET((so), (events), set))
181#define UNIX_CHECK_FD_SET(so, events, set) \
182 (DO_UNIX_CHECK_FD_SET(so, events, set))
183
184/*
185 * Loging macros
186 */
187#if VBOX_WITH_DEBUG_NAT_SOCKETS
188# if defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC)
189# if defined(RT_OS_WINDOWS)
190# define DO_LOG_NAT_SOCK(so, proto, winevent, r_fdset, w_fdset, x_fdset) \
191 do { \
192 LogRel((" " #proto " %R[natsock] %R[natwinnetevents]\n", (so), (winevent))); \
193 } while (0)
194# else /* RT_OS_WINDOWS */
195# define DO_LOG_NAT_SOCK(so, proto, winevent, r_fdset, w_fdset, x_fdset) \
196 do { \
197 LogRel((" " #proto " %R[natsock] %s %s %s er: %s, %s, %s\n", (so), \
198 CHECK_FD_SET(so, ign ,r_fdset) ? "READ":"", \
199 CHECK_FD_SET(so, ign, w_fdset) ? "WRITE":"", \
200 CHECK_FD_SET(so, ign, x_fdset) ? "OOB":"", \
201 CHECK_FD_SET(so, ign, rderr) ? "RDERR":"", \
202 CHECK_FD_SET(so, ign, rdhup) ? "RDHUP":"", \
203 CHECK_FD_SET(so, ign, nval) ? "RDNVAL":"")); \
204 } while (0)
205# endif /* !RT_OS_WINDOWS */
206# else /* VBOX_WITH_SIMPLIFIED_SLIRP_SYNC */
207# define DO_LOG_NAT_SOCK(so, proto, winevent, r_fdset, w_fdset, x_fdset) \
208 do { \
209 LogRel((" " #proto " %R[natsock] %s %s %s\n", (so), FD_ISSET((so)->s, (r_fdset))?"READ":"", \
210 FD_ISSET((so)->s, (w_fdset))?"WRITE":"", FD_ISSET((so)->s, (x_fdset))?"OOB":"")); \
211 } while (0)
212# endif
213#else /* VBOX_WITH_DEBUG_NAT_SOCKETS */
214# define DO_LOG_NAT_SOCK(so, proto, winevent, r_fdset, w_fdset, x_fdset) do {} while (0)
215#endif /* !VBOX_WITH_DEBUG_NAT_SOCKETS */
216
217#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)
218
219static const uint8_t special_ethaddr[6] =
220{
221 0x52, 0x54, 0x00, 0x12, 0x35, 0x00
222};
223
224#ifdef RT_OS_WINDOWS
225
226static int get_dns_addr_domain(PNATState pData, bool fVerbose,
227 struct in_addr *pdns_addr,
228 const char **ppszDomain)
229{
230 int rc = 0;
231 FIXED_INFO *FixedInfo = NULL;
232 ULONG BufLen;
233 DWORD ret;
234 IP_ADDR_STRING *pIPAddr;
235 struct in_addr tmp_addr;
236
237 FixedInfo = (FIXED_INFO *)GlobalAlloc(GPTR, sizeof(FIXED_INFO));
238 BufLen = sizeof(FIXED_INFO);
239
240 /** @todo: this API returns all DNS servers, no matter whether the
241 * corresponding network adapter is disabled or not. Maybe replace
242 * this by GetAdapterAddresses(), which is XP/Vista only though. */
243 if (ERROR_BUFFER_OVERFLOW == GetNetworkParams(FixedInfo, &BufLen))
244 {
245 if (FixedInfo)
246 {
247 GlobalFree(FixedInfo);
248 FixedInfo = NULL;
249 }
250 FixedInfo = GlobalAlloc(GPTR, BufLen);
251 }
252
253 if ((ret = GetNetworkParams(FixedInfo, &BufLen)) != ERROR_SUCCESS)
254 {
255 Log(("GetNetworkParams failed. ret = %08x\n", (u_int)ret ));
256 if (FixedInfo)
257 {
258 GlobalFree(FixedInfo);
259 FixedInfo = NULL;
260 }
261 rc = -1;
262 goto get_dns_prefix;
263 }
264
265#ifndef VBOX_WITH_MULTI_DNS
266 pIPAddr = &(FixedInfo->DnsServerList);
267 inet_aton(pIPAddr->IpAddress.String, &tmp_addr);
268 Log(("nat: DNS Servers:\n"));
269 if (fVerbose || pdns_addr->s_addr != tmp_addr.s_addr)
270 LogRel(("NAT: DNS address: %s\n", pIPAddr->IpAddress.String));
271 *pdns_addr = tmp_addr;
272
273 pIPAddr = FixedInfo -> DnsServerList.Next;
274 while (pIPAddr)
275 {
276 if (fVerbose)
277 LogRel(("NAT: ignored DNS address: %s\n", pIPAddr ->IpAddress.String));
278 pIPAddr = pIPAddr ->Next;
279 }
280#else
281 /*localhost mask */
282 for (pIPAddr = &FixedInfo->DnsServerList; pIPAddr != NULL; pIPAddr = pIPAddr->Next)
283 {
284 struct dns_entry *da;
285 if(!inet_aton(pIPAddr->IpAddress.String, &tmp_addr))
286 continue;
287 da = RTMemAllocZ(sizeof (struct dns_entry));
288 if (da == NULL)
289 {
290 LogRel(("can't alloc memory for DNS entry\n"));
291 return -1;
292 }
293 /*check */
294 if ((da->de_addr.s_addr & htonl(IN_CLASSA_NET)) == ntohl(INADDR_LOOPBACK & IN_CLASSA_NET)) {
295 da->de_addr.s_addr = htonl(ntohl(special_addr.s_addr) | CTL_ALIAS);
296 }
297 LIST_INSERT_HEAD(&pData->dns_list_head, da, de_list);
298 }
299#endif
300 if (FixedInfo)
301 {
302 GlobalFree(FixedInfo);
303 FixedInfo = NULL;
304 }
305
306get_dns_prefix:
307 if (ppszDomain)
308 {
309 OSVERSIONINFO ver;
310 char szDnsDomain[256];
311 DWORD dwSize = sizeof(szDnsDomain);
312
313 *ppszDomain = NULL;
314 GetVersionEx(&ver);
315 if (ver.dwMajorVersion >= 5)
316 {
317 /* GetComputerNameEx exists in Windows versions starting with 2000. */
318 if (GetComputerNameEx(ComputerNameDnsDomain, szDnsDomain, &dwSize))
319 {
320 if (szDnsDomain[0])
321 {
322 /* Just non-empty strings are valid. */
323 *ppszDomain = RTStrDup(szDnsDomain);
324 if (pData->fPassDomain)
325 {
326 if (fVerbose)
327 LogRel(("NAT: passing domain name %s\n", szDnsDomain));
328 }
329 else
330 Log(("nat: ignoring domain %s\n", szDnsDomain));
331 }
332 }
333 else
334 Log(("nat: GetComputerNameEx failed (%d)\n", GetLastError()));
335 }
336 }
337 return rc;
338}
339
340#else
341
342static int get_dns_addr_domain(PNATState pData, bool fVerbose,
343 struct in_addr *pdns_addr,
344 const char **ppszDomain)
345{
346 char buff[512];
347 char buff2[256];
348 FILE *f;
349 int found = 0;
350 struct in_addr tmp_addr;
351
352#ifdef RT_OS_OS2
353 /* Try various locations. */
354 char *etc = getenv("ETC");
355 f = NULL;
356 if (etc)
357 {
358 snprintf(buff, sizeof(buff), "%s/RESOLV2", etc);
359 f = fopen(buff, "rt");
360 }
361 if (!f)
362 {
363 snprintf(buff, sizeof(buff), "%s/RESOLV2", _PATH_ETC);
364 f = fopen(buff, "rt");
365 }
366 if (!f)
367 {
368 snprintf(buff, sizeof(buff), "%s/resolv.conf", _PATH_ETC);
369 f = fopen(buff, "rt");
370 }
371#else
372 f = fopen("/etc/resolv.conf", "r");
373#endif
374 if (!f)
375 return -1;
376
377 if (ppszDomain)
378 *ppszDomain = NULL;
379 Log(("nat: DNS Servers:\n"));
380 while (fgets(buff, 512, f) != NULL)
381 {
382#ifdef VBOX_WITH_MULTI_DNS
383 struct dns_entry *da = NULL;
384#endif
385 if (sscanf(buff, "nameserver%*[ \t]%256s", buff2) == 1)
386 {
387 if (!inet_aton(buff2, &tmp_addr))
388 continue;
389#ifndef VBOX_WITH_MULTI_DNS
390 /* If it's the first one, set it to dns_addr */
391 if (!found)
392 {
393 if (fVerbose || pdns_addr->s_addr != tmp_addr.s_addr)
394 LogRel(("NAT: DNS address: %s\n", buff2));
395 *pdns_addr = tmp_addr;
396 }
397 else
398 {
399 if (fVerbose)
400 LogRel(("NAT: ignored DNS address: %s\n", buff2));
401 }
402#else
403 /*localhost mask */
404 da = RTMemAllocZ(sizeof (struct dns_entry));
405 if (da == NULL)
406 {
407 LogRel(("can't alloc memory for DNS entry\n"));
408 return -1;
409 }
410 /*check */
411 da->de_addr.s_addr = tmp_addr.s_addr;
412 if ((da->de_addr.s_addr & htonl(IN_CLASSA_NET)) == ntohl(INADDR_LOOPBACK & IN_CLASSA_NET)) {
413 da->de_addr.s_addr = htonl(ntohl(special_addr.s_addr) | CTL_ALIAS);
414 }
415 LIST_INSERT_HEAD(&pData->dns_list_head, da, de_list);
416#endif
417 found++;
418 }
419 if ( ppszDomain
420 && (!strncmp(buff, "domain", 6) || !strncmp(buff, "search", 6)))
421 {
422 /* Domain name/search list present. Pick first entry */
423 if (*ppszDomain == NULL)
424 {
425 char *tok;
426 char *saveptr;
427 tok = strtok_r(&buff[6], " \t\n", &saveptr);
428 if (tok)
429 {
430 *ppszDomain = RTStrDup(tok);
431 if (pData->fPassDomain)
432 {
433 if (fVerbose)
434 LogRel(("NAT: passing domain name %s\n", tok));
435 }
436 else
437 Log(("nat: ignoring domain %s\n", tok));
438 }
439 }
440 }
441 }
442 fclose(f);
443 if (!found)
444 return -1;
445 return 0;
446}
447
448#endif
449
450int get_dns_addr(PNATState pData, struct in_addr *pdns_addr)
451{
452 return get_dns_addr_domain(pData, false, pdns_addr, NULL);
453}
454
455int slirp_init(PNATState *ppData, const char *pszNetAddr, uint32_t u32Netmask,
456 bool fPassDomain, const char *pszTFTPPrefix,
457 const char *pszBootFile, void *pvUser)
458{
459 int fNATfailed = 0;
460 int rc;
461 PNATState pData = RTMemAllocZ(sizeof(NATState));
462 *ppData = pData;
463 if (!pData)
464 return VERR_NO_MEMORY;
465 if (u32Netmask & 0x1f)
466 /* CTL is x.x.x.15, bootp passes up to 16 IPs (15..31) */
467 return VERR_INVALID_PARAMETER;
468 pData->fPassDomain = fPassDomain;
469 pData->pvUser = pvUser;
470 tftp_prefix = pszTFTPPrefix;
471 bootp_filename = pszBootFile;
472 pData->netmask = u32Netmask;
473
474#ifdef RT_OS_WINDOWS
475 {
476 WSADATA Data;
477 WSAStartup(MAKEWORD(2,0), &Data);
478 }
479# if defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC)
480 pData->phEvents[VBOX_SOCKET_EVENT_INDEX] = CreateEvent(NULL, FALSE, FALSE, NULL);
481# endif
482#endif
483#ifdef VBOX_WITH_SLIRP_MT
484 QSOCKET_LOCK_CREATE(tcb);
485 QSOCKET_LOCK_CREATE(udb);
486 rc = RTReqCreateQueue(&pData->pReqQueue);
487 AssertReleaseRC(rc);
488#endif
489
490 link_up = 1;
491
492 debug_init();
493 if_init(pData);
494 ip_init(pData);
495 icmp_init(pData);
496
497 /* Initialise mbufs *after* setting the MTU */
498 m_init(pData);
499
500 inet_aton(pszNetAddr, &special_addr);
501 alias_addr.s_addr = special_addr.s_addr | htonl(CTL_ALIAS);
502 /* @todo: add ability to configure this staff */
503 pData->tftp_server.s_addr = htonl(ntohl(special_addr.s_addr) | CTL_TFTP);
504
505 /* set default addresses */
506 inet_aton("127.0.0.1", &loopback_addr);
507#ifndef VBOX_WITH_MULTI_DNS
508 inet_aton("127.0.0.1", &dns_addr);
509
510 if (get_dns_addr_domain(pData, true, &dns_addr, &pData->pszDomain) < 0)
511#else
512 LIST_INIT(&pData->dns_list_head);
513 if (get_dns_addr_domain(pData, true, NULL, &pData->pszDomain) < 0)
514#endif
515 fNATfailed = 1;
516
517 getouraddr(pData);
518 return fNATfailed ? VINF_NAT_DNS : VINF_SUCCESS;
519}
520
521/**
522 * Statistics counters.
523 */
524void slirp_register_timers(PNATState pData, PPDMDRVINS pDrvIns)
525{
526#ifdef VBOX_WITH_STATISTICS
527 PDMDrvHlpSTAMRegisterF(pDrvIns, &pData->StatFill, STAMTYPE_PROFILE, STAMVISIBILITY_ALWAYS,
528 STAMUNIT_TICKS_PER_CALL, "Profiling slirp fills", "/Drivers/NAT%d/Fill", pDrvIns->iInstance);
529 PDMDrvHlpSTAMRegisterF(pDrvIns, &pData->StatPoll, STAMTYPE_PROFILE, STAMVISIBILITY_ALWAYS,
530 STAMUNIT_TICKS_PER_CALL, "Profiling slirp polls", "/Drivers/NAT%d/Poll", pDrvIns->iInstance);
531 PDMDrvHlpSTAMRegisterF(pDrvIns, &pData->StatFastTimer, STAMTYPE_PROFILE, STAMVISIBILITY_ALWAYS,
532 STAMUNIT_TICKS_PER_CALL, "Profiling slirp fast timer", "/Drivers/NAT%d/TimerFast", pDrvIns->iInstance);
533 PDMDrvHlpSTAMRegisterF(pDrvIns, &pData->StatSlowTimer, STAMTYPE_PROFILE, STAMVISIBILITY_ALWAYS,
534 STAMUNIT_TICKS_PER_CALL, "Profiling slirp slow timer", "/Drivers/NAT%d/TimerSlow", pDrvIns->iInstance);
535 PDMDrvHlpSTAMRegisterF(pDrvIns, &pData->StatTCP, STAMTYPE_COUNTER, STAMVISIBILITY_ALWAYS,
536 STAMUNIT_COUNT, "TCP sockets", "/Drivers/NAT%d/SockTCP", pDrvIns->iInstance);
537 PDMDrvHlpSTAMRegisterF(pDrvIns, &pData->StatTCPHot, STAMTYPE_COUNTER, STAMVISIBILITY_ALWAYS,
538 STAMUNIT_COUNT, "TCP sockets active", "/Drivers/NAT%d/SockTCPHot", pDrvIns->iInstance);
539 PDMDrvHlpSTAMRegisterF(pDrvIns, &pData->StatUDP, STAMTYPE_COUNTER, STAMVISIBILITY_ALWAYS,
540 STAMUNIT_COUNT, "UDP sockets", "/Drivers/NAT%d/SockUDP", pDrvIns->iInstance);
541 PDMDrvHlpSTAMRegisterF(pDrvIns, &pData->StatUDPHot, STAMTYPE_COUNTER, STAMVISIBILITY_ALWAYS,
542 STAMUNIT_COUNT, "UDP sockets active", "/Drivers/NAT%d/SockUDPHot", pDrvIns->iInstance);
543#endif /* VBOX_WITH_STATISTICS */
544}
545
546/**
547 * Marks the link as up, making it possible to establish new connections.
548 */
549void slirp_link_up(PNATState pData)
550{
551 link_up = 1;
552}
553
554/**
555 * Marks the link as down and cleans up the current connections.
556 */
557void slirp_link_down(PNATState pData)
558{
559 struct socket *so;
560
561 while ((so = tcb.so_next) != &tcb)
562 {
563 if (so->so_state & SS_NOFDREF || so->s == -1)
564 sofree(pData, so);
565 else
566 tcp_drop(pData, sototcpcb(so), 0);
567 }
568
569 while ((so = udb.so_next) != &udb)
570 udp_detach(pData, so);
571
572 link_up = 0;
573}
574
575/**
576 * Terminates the slirp component.
577 */
578void slirp_term(PNATState pData)
579{
580#ifdef VBOX_WITH_MULTI_DNS
581 struct dns_entry *de = NULL;
582#endif
583 if (pData->pszDomain)
584 RTStrFree((char *)(void *)pData->pszDomain);
585
586#ifdef RT_OS_WINDOWS
587 pData->pfIcmpCloseHandle(pData->icmp_socket.sh);
588 FreeLibrary(pData->hmIcmpLibrary);
589 RTMemFree(pData->pvIcmpBuffer);
590# else
591 closesocket(pData->icmp_socket.s);
592#endif
593
594 slirp_link_down(pData);
595#ifdef VBOX_WITH_MULTI_DNS
596 while(!LIST_EMPTY(&pData->dns_list_head)) {
597 de = LIST_FIRST(&pData->dns_list_head);
598 LIST_REMOVE(de, de_list);
599 RTMemFree(de);
600 }
601#endif
602#ifdef RT_OS_WINDOWS
603 WSACleanup();
604#endif
605#ifdef LOG_ENABLED
606 Log(("\n"
607 "NAT statistics\n"
608 "--------------\n"
609 "\n"));
610 ipstats(pData);
611 tcpstats(pData);
612 udpstats(pData);
613 icmpstats(pData);
614 mbufstats(pData);
615 sockstats(pData);
616 Log(("\n"
617 "\n"
618 "\n"));
619#endif
620 RTMemFree(pData);
621}
622
623
624#define CONN_CANFSEND(so) (((so)->so_state & (SS_FCANTSENDMORE|SS_ISFCONNECTED)) == SS_ISFCONNECTED)
625#define CONN_CANFRCV(so) (((so)->so_state & (SS_FCANTRCVMORE|SS_ISFCONNECTED)) == SS_ISFCONNECTED)
626#define UPD_NFDS(x) if (nfds < (x)) nfds = (x)
627
628/*
629 * curtime kept to an accuracy of 1ms
630 */
631#ifdef RT_OS_WINDOWS
632static void updtime(PNATState pData)
633{
634 struct _timeb tb;
635
636 _ftime(&tb);
637 curtime = (u_int)tb.time * (u_int)1000;
638 curtime += (u_int)tb.millitm;
639}
640#else
641static void updtime(PNATState pData)
642{
643 gettimeofday(&tt, 0);
644
645 curtime = (u_int)tt.tv_sec * (u_int)1000;
646 curtime += (u_int)tt.tv_usec / (u_int)1000;
647
648 if ((tt.tv_usec % 1000) >= 500)
649 curtime++;
650}
651#endif
652
653#ifndef VBOX_WITH_SIMPLIFIED_SLIRP_SYNC
654void slirp_select_fill(PNATState pData, int *pnfds,
655 fd_set *readfds, fd_set *writefds, fd_set *xfds)
656#else /* !VBOX_WITH_SIMPLIFIED_SLIRP_SYNC */
657# ifdef RT_OS_WINDOWS
658void slirp_select_fill(PNATState pData, int *pnfds)
659# else /* RT_OS_WINDOWS */
660void slirp_select_fill(PNATState pData, int *pnfds, struct pollfd *polls)
661# endif /* !RT_OS_WINDOWS */
662#endif /* VBOX_WITH_SIMPLIFIED_SLIRP_SYNC */
663{
664 struct socket *so, *so_next;
665 int nfds;
666#if defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) && defined(RT_OS_WINDOWS)
667 int rc;
668 int error;
669#endif
670#if defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) && !defined(RT_OS_WINDOWS)
671 int poll_index = 0;
672#endif
673 int i;
674
675 STAM_PROFILE_START(&pData->StatFill, a);
676
677 nfds = *pnfds;
678
679 /*
680 * First, TCP sockets
681 */
682 do_slowtimo = 0;
683 if (link_up)
684 {
685 /*
686 * *_slowtimo needs calling if there are IP fragments
687 * in the fragment queue, or there are TCP connections active
688 */
689 /* XXX:
690 * triggering of fragment expiration should be the same but use new macroses
691 */
692 do_slowtimo = (tcb.so_next != &tcb);
693 if (!do_slowtimo)
694 {
695 for (i = 0; i < IPREASS_NHASH; i++)
696 {
697 if (!TAILQ_EMPTY(&ipq[i]))
698 {
699 do_slowtimo = 1;
700 break;
701 }
702 }
703 }
704 ICMP_ENGAGE_EVENT(&pData->icmp_socket, readfds);
705
706 STAM_COUNTER_RESET(&pData->StatTCP);
707 STAM_COUNTER_RESET(&pData->StatTCPHot);
708
709 QSOCKET_FOREACH(so, so_next, tcp)
710 /* { */
711#if defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) && !defined(RT_OS_WINDOWS)
712 so->so_poll_index = -1;
713#endif
714 STAM_COUNTER_INC(&pData->StatTCP);
715
716 /*
717 * See if we need a tcp_fasttimo
718 */
719 if ( time_fasttimo == 0
720 && so->so_tcpcb != NULL
721 && so->so_tcpcb->t_flags & TF_DELACK)
722 time_fasttimo = curtime; /* Flag when we want a fasttimo */
723
724 /*
725 * NOFDREF can include still connecting to local-host,
726 * newly socreated() sockets etc. Don't want to select these.
727 */
728 if (so->so_state & SS_NOFDREF || so->s == -1)
729 CONTINUE(tcp);
730
731 /*
732 * Set for reading sockets which are accepting
733 */
734 if (so->so_state & SS_FACCEPTCONN)
735 {
736 STAM_COUNTER_INC(&pData->StatTCPHot);
737 TCP_ENGAGE_EVENT1(so, readfds);
738 CONTINUE(tcp);
739 }
740
741 /*
742 * Set for writing sockets which are connecting
743 */
744 if (so->so_state & SS_ISFCONNECTING)
745 {
746 Log2(("connecting %R[natsock] engaged\n",so));
747 STAM_COUNTER_INC(&pData->StatTCPHot);
748 TCP_ENGAGE_EVENT1(so, writefds);
749 }
750
751 /*
752 * Set for writing if we are connected, can send more, and
753 * we have something to send
754 */
755 if (CONN_CANFSEND(so) && so->so_rcv.sb_cc)
756 {
757 STAM_COUNTER_INC(&pData->StatTCPHot);
758 TCP_ENGAGE_EVENT1(so, writefds);
759 }
760
761 /*
762 * Set for reading (and urgent data) if we are connected, can
763 * receive more, and we have room for it XXX /2 ?
764 */
765 if (CONN_CANFRCV(so) && (so->so_snd.sb_cc < (so->so_snd.sb_datalen/2)))
766 {
767 STAM_COUNTER_INC(&pData->StatTCPHot);
768 TCP_ENGAGE_EVENT2(so, readfds, xfds);
769 }
770 LOOP_LABEL(tcp, so, so_next);
771 }
772
773 /*
774 * UDP sockets
775 */
776 STAM_COUNTER_RESET(&pData->StatUDP);
777 STAM_COUNTER_RESET(&pData->StatUDPHot);
778
779 QSOCKET_FOREACH(so, so_next, udp)
780 /* { */
781
782 STAM_COUNTER_INC(&pData->StatUDP);
783#if defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) && !defined(RT_OS_WINDOWS)
784 so->so_poll_index = -1;
785#endif
786
787 /*
788 * See if it's timed out
789 */
790 if (so->so_expire)
791 {
792 if (so->so_expire <= curtime)
793 {
794#ifdef VBOX_WITH_SLIRP_MT
795 /* we need so_next for continue our cycle*/
796 so_next = so->so_next;
797#endif
798 UDP_DETACH(pData, so, so_next);
799 CONTINUE_NO_UNLOCK(udp);
800 }
801 else
802 do_slowtimo = 1; /* Let socket expire */
803 }
804
805 /*
806 * When UDP packets are received from over the link, they're
807 * sendto()'d straight away, so no need for setting for writing
808 * Limit the number of packets queued by this session to 4.
809 * Note that even though we try and limit this to 4 packets,
810 * the session could have more queued if the packets needed
811 * to be fragmented.
812 *
813 * (XXX <= 4 ?)
814 */
815 if ((so->so_state & SS_ISFCONNECTED) && so->so_queued <= 4)
816 {
817 STAM_COUNTER_INC(&pData->StatUDPHot);
818 UDP_ENGAGE_EVENT(so, readfds);
819 }
820 LOOP_LABEL(udp, so, so_next);
821 }
822
823 }
824
825#if defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC)
826# if defined(RT_OS_WINDOWS)
827 *pnfds = VBOX_EVENT_COUNT;
828# else /* RT_OS_WINDOWS */
829 AssertRelease(poll_index <= *pnfds);
830 *pnfds = poll_index;
831# endif /* !RT_OS_WINDOWS */
832#else /* VBOX_WITH_SIMPLIFIED_SLIRP_SYNC */
833 *pnfds = nfds;
834#endif /* !VBOX_WITH_SIMPLIFIED_SLIRP_SYNC */
835
836 STAM_PROFILE_STOP(&pData->StatFill, a);
837}
838
839#if defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC)
840# if defined(RT_OS_WINDOWS)
841void slirp_select_poll(PNATState pData, int fTimeout, int fIcmp)
842# else /* RT_OS_WINDOWS */
843void slirp_select_poll(PNATState pData, struct pollfd *polls, int ndfs)
844# endif /* !RT_OS_WINDOWS */
845#else /* VBOX_WITH_SIMPLIFIED_SLIRP_SYNC */
846void slirp_select_poll(PNATState pData, fd_set *readfds, fd_set *writefds, fd_set *xfds)
847#endif /* !VBOX_WITH_SIMPLIFIED_SLIRP_SYNC */
848{
849 struct socket *so, *so_next;
850 int ret;
851#if defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) && defined(RT_OS_WINDOWS)
852 WSANETWORKEVENTS NetworkEvents;
853 int rc;
854 int error;
855#endif
856#if defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) && !defined(RT_OS_WINDOWS)
857 int poll_index = 0;
858#endif
859
860 STAM_PROFILE_START(&pData->StatPoll, a);
861
862 /* Update time */
863 updtime(pData);
864
865 /*
866 * See if anything has timed out
867 */
868 if (link_up)
869 {
870 if (time_fasttimo && ((curtime - time_fasttimo) >= 2))
871 {
872 STAM_PROFILE_START(&pData->StatFastTimer, a);
873 tcp_fasttimo(pData);
874 time_fasttimo = 0;
875 STAM_PROFILE_STOP(&pData->StatFastTimer, a);
876 }
877 if (do_slowtimo && ((curtime - last_slowtimo) >= 499))
878 {
879 STAM_PROFILE_START(&pData->StatSlowTimer, a);
880 ip_slowtimo(pData);
881 tcp_slowtimo(pData);
882 last_slowtimo = curtime;
883 STAM_PROFILE_STOP(&pData->StatSlowTimer, a);
884 }
885 }
886#if defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) && defined(RT_OS_WINDOWS)
887 if (fTimeout)
888 return; /* only timer update */
889#endif
890
891 /*
892 * Check sockets
893 */
894 if (link_up)
895 {
896#if defined(RT_OS_WINDOWS)
897 /*XXX: before renaming please make see define
898 * fIcmp in slirp_state.h
899 */
900 if (fIcmp)
901 sorecvfrom(pData, &pData->icmp_socket);
902#else
903 if ( (pData->icmp_socket.s != -1)
904 && CHECK_FD_SET(&pData->icmp_socket, ignored, readfds))
905 sorecvfrom(pData, &pData->icmp_socket);
906#endif
907 /*
908 * Check TCP sockets
909 */
910 QSOCKET_FOREACH(so, so_next, tcp)
911 /* { */
912
913#ifdef VBOX_WITH_SLIRP_MT
914 if ( so->so_state & SS_NOFDREF
915 && so->so_deleted == 1)
916 {
917 struct socket *son, *sop = NULL;
918 QSOCKET_LOCK(tcb);
919 if (so->so_next != NULL)
920 {
921 if (so->so_next != &tcb)
922 SOCKET_LOCK(so->so_next);
923 son = so->so_next;
924 }
925 if ( so->so_prev != &tcb
926 && so->so_prev != NULL)
927 {
928 SOCKET_LOCK(so->so_prev);
929 sop = so->so_prev;
930 }
931 QSOCKET_UNLOCK(tcb);
932 remque(pData, so);
933 NSOCK_DEC();
934 SOCKET_UNLOCK(so);
935 SOCKET_LOCK_DESTROY(so);
936 RTMemFree(so);
937 so_next = son;
938 if (sop != NULL)
939 SOCKET_UNLOCK(sop);
940 CONTINUE_NO_UNLOCK(tcp);
941 }
942#endif
943 /*
944 * FD_ISSET is meaningless on these sockets
945 * (and they can crash the program)
946 */
947 if (so->so_state & SS_NOFDREF || so->s == -1)
948 CONTINUE(tcp);
949
950 POLL_TCP_EVENTS(rc, error, so, &NetworkEvents);
951
952 LOG_NAT_SOCK(so, TCP, &NetworkEvents, readfds, writefds, xfds);
953
954
955 /*
956 * Check for URG data
957 * This will soread as well, so no need to
958 * test for readfds below if this succeeds
959 */
960
961 /* out-of-band data */
962 if (CHECK_FD_SET(so, NetworkEvents, xfds))
963 {
964 sorecvoob(pData, so);
965 }
966
967 /*
968 * Check sockets for reading
969 */
970 else if ( CHECK_FD_SET(so, NetworkEvents, readfds)
971 || WIN_CHECK_FD_SET(so, NetworkEvents, acceptds))
972 {
973 /*
974 * Check for incoming connections
975 */
976 if (so->so_state & SS_FACCEPTCONN)
977 {
978 TCP_CONNECT(pData, so);
979#if defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) && defined(RT_OS_WINDOWS)
980 if (!(NetworkEvents.lNetworkEvents & FD_CLOSE))
981#endif
982 CONTINUE(tcp);
983 }
984
985 ret = soread(pData, so);
986 /* Output it if we read something */
987 if (ret > 0)
988 TCP_OUTPUT(pData, sototcpcb(so));
989 }
990
991#if defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) && defined(RT_OS_WINDOWS)
992 /*
993 * Check for FD_CLOSE events.
994 * in some cases once FD_CLOSE engaged on socket it could be flashed latter (for some reasons)
995 */
996 if ( (NetworkEvents.lNetworkEvents & FD_CLOSE)
997 || (so->so_close == 1))
998 {
999 so->so_close = 1; /* mark it */
1000 /*
1001 * drain the socket
1002 */
1003 for (;;)
1004 {
1005 ret = soread(pData, so);
1006 if (ret > 0)
1007 TCP_OUTPUT(pData, sototcpcb(so));
1008 else
1009 break;
1010 }
1011 CONTINUE(tcp);
1012 }
1013#endif
1014
1015 /*
1016 * Check sockets for writing
1017 */
1018 if (CHECK_FD_SET(so, NetworkEvents, writefds))
1019 {
1020 /*
1021 * Check for non-blocking, still-connecting sockets
1022 */
1023 if (so->so_state & SS_ISFCONNECTING)
1024 {
1025 Log2(("connecting %R[natsock] catched\n", so));
1026 /* Connected */
1027 so->so_state &= ~SS_ISFCONNECTING;
1028
1029 /*
1030 * This should be probably guarded by PROBE_CONN too. Anyway,
1031 * we disable it on OS/2 because the below send call returns
1032 * EFAULT which causes the opened TCP socket to close right
1033 * after it has been opened and connected.
1034 */
1035#ifndef RT_OS_OS2
1036 ret = send(so->s, (const char *)&ret, 0, 0);
1037 if (ret < 0)
1038 {
1039 /* XXXXX Must fix, zero bytes is a NOP */
1040 if ( errno == EAGAIN
1041 || errno == EWOULDBLOCK
1042 || errno == EINPROGRESS
1043 || errno == ENOTCONN)
1044 CONTINUE(tcp);
1045
1046 /* else failed */
1047 so->so_state = SS_NOFDREF;
1048 }
1049 /* else so->so_state &= ~SS_ISFCONNECTING; */
1050#endif
1051
1052 /*
1053 * Continue tcp_input
1054 */
1055 TCP_INPUT(pData, (struct mbuf *)NULL, sizeof(struct ip), so);
1056 /* continue; */
1057 }
1058 else
1059 SOWRITE(ret, pData, so);
1060 /*
1061 * XXX If we wrote something (a lot), there could be the need
1062 * for a window update. In the worst case, the remote will send
1063 * a window probe to get things going again.
1064 */
1065 }
1066
1067 /*
1068 * Probe a still-connecting, non-blocking socket
1069 * to check if it's still alive
1070 */
1071#ifdef PROBE_CONN
1072 if (so->so_state & SS_ISFCONNECTING)
1073 {
1074 ret = recv(so->s, (char *)&ret, 0, 0);
1075
1076 if (ret < 0)
1077 {
1078 /* XXX */
1079 if ( errno == EAGAIN
1080 || errno == EWOULDBLOCK
1081 || errno == EINPROGRESS
1082 || errno == ENOTCONN)
1083 {
1084 CONTINUE(tcp); /* Still connecting, continue */
1085 }
1086
1087 /* else failed */
1088 so->so_state = SS_NOFDREF;
1089
1090 /* tcp_input will take care of it */
1091 }
1092 else
1093 {
1094 ret = send(so->s, &ret, 0, 0);
1095 if (ret < 0)
1096 {
1097 /* XXX */
1098 if ( errno == EAGAIN
1099 || errno == EWOULDBLOCK
1100 || errno == EINPROGRESS
1101 || errno == ENOTCONN)
1102 {
1103 CONTINUE(tcp);
1104 }
1105 /* else failed */
1106 so->so_state = SS_NOFDREF;
1107 }
1108 else
1109 so->so_state &= ~SS_ISFCONNECTING;
1110
1111 }
1112 TCP_INPUT((struct mbuf *)NULL, sizeof(struct ip),so);
1113 } /* SS_ISFCONNECTING */
1114#endif
1115#ifndef RT_OS_WINDOWS
1116 if ( UNIX_CHECK_FD_SET(so, NetworkEvents, rdhup)
1117 || UNIX_CHECK_FD_SET(so, NetworkEvents, rderr))
1118 {
1119 int err;
1120 int inq = -1, outq;
1121 socklen_t optlen = sizeof(int);
1122 int status = getsockopt(so->s, SOL_SOCKET, SO_ERROR, &err, &optlen);
1123 if (status != 0)
1124 LogRel(("NAT:can't get error status from %R[natsock]\n", so));
1125#ifndef RT_OS_SOLARIS
1126 status = ioctl(so->s, FIONREAD, &inq); /* tcp(7) recommends SIOCINQ which is Linux specific */
1127 if (status != 0 || status != EINVAL)
1128 {
1129 /* EINVAL returned if socket in listen state tcp(7)*/
1130 LogRel(("NAT:can't get depth of IN queue status from %R[natsock]\n", so));
1131 }
1132 status = ioctl(so->s, TIOCOUTQ, &outq); /* SIOCOUTQ see previous comment */
1133 if (status != 0)
1134 LogRel(("NAT:can't get depth of OUT queue from %R[natsock]\n", so));
1135#else
1136 inq = outq = -1;
1137 /*
1138 * Solaris has bit different ioctl commands and its handlings
1139 * hint: streamio(7) I_NREAD
1140 */
1141#endif
1142 if ( so->so_state & SS_ISFCONNECTING
1143 || UNIX_CHECK_FD_SET(so, NetworkEvents, readfds))
1144 {
1145 /**
1146 * Check if we need here take care about gracefull connection
1147 * @todo try with proxy server
1148 */
1149 if (UNIX_CHECK_FD_SET(so, NetworkEvents, readfds))
1150 {
1151 /*
1152 * Never meet inq != 0 or outq != 0, anyway let it stay for a while
1153 * in case it happens we'll able to detect it.
1154 * Give TCP/IP stack wait or expire the socket.
1155 */
1156 LogRel(("NAT:%R[natsock] err(%d:%s) s(in:%d,out:%d)happens on read I/O, "
1157 "other side close connection \n", so, err, strerror(err), inq, outq));
1158 CONTINUE(tcp);
1159 }
1160 goto tcp_input_close;
1161 }
1162 if ( !UNIX_CHECK_FD_SET(so, NetworkEvents, readfds)
1163 && !UNIX_CHECK_FD_SET(so, NetworkEvents, writefds)
1164 && !UNIX_CHECK_FD_SET(so, NetworkEvents, xfds))
1165 {
1166 LogRel(("NAT:system expires the socket %R[natsock] err(%d:%s) s(in:%d,out:%d) happens on non-I/O. ",
1167 so, err, strerror(err), inq, outq));
1168 goto tcp_input_close;
1169 }
1170 LogRel(("NAT:%R[natsock] we've met(%d:%s) s(in:%d, out:%d) unhandled combination hup (%d) "
1171 "rederr(%d) on (r:%d, w:%d, x:%d)\n",
1172 so, err, strerror(err),
1173 inq, outq,
1174 UNIX_CHECK_FD_SET(so, ign, rdhup),
1175 UNIX_CHECK_FD_SET(so, ign, rderr),
1176 UNIX_CHECK_FD_SET(so, ign, readfds),
1177 UNIX_CHECK_FD_SET(so, ign, writefds),
1178 UNIX_CHECK_FD_SET(so, ign, xfds)));
1179 /*
1180 * Give OS's TCP/IP stack a chance to resolve an issue or expire the socket.
1181 */
1182 CONTINUE(tcp);
1183tcp_input_close:
1184 so->so_state = SS_NOFDREF; /*cause connection valid tcp connection termination and socket closing */
1185 TCP_INPUT(pData, (struct mbuf *)NULL, sizeof(struct ip), so);
1186 CONTINUE(tcp);
1187 }
1188#endif
1189 LOOP_LABEL(tcp, so, so_next);
1190 }
1191
1192 /*
1193 * Now UDP sockets.
1194 * Incoming packets are sent straight away, they're not buffered.
1195 * Incoming UDP data isn't buffered either.
1196 */
1197 QSOCKET_FOREACH(so, so_next, udp)
1198 /* { */
1199#ifdef VBOX_WITH_SLIRP_MT
1200 if ( so->so_state & SS_NOFDREF
1201 && so->so_deleted == 1)
1202 {
1203 struct socket *son, *sop = NULL;
1204 QSOCKET_LOCK(udb);
1205 if (so->so_next != NULL)
1206 {
1207 if (so->so_next != &udb)
1208 SOCKET_LOCK(so->so_next);
1209 son = so->so_next;
1210 }
1211 if ( so->so_prev != &udb
1212 && so->so_prev != NULL)
1213 {
1214 SOCKET_LOCK(so->so_prev);
1215 sop = so->so_prev;
1216 }
1217 QSOCKET_UNLOCK(udb);
1218 remque(pData, so);
1219 NSOCK_DEC();
1220 SOCKET_UNLOCK(so);
1221 SOCKET_LOCK_DESTROY(so);
1222 RTMemFree(so);
1223 so_next = son;
1224 if (sop != NULL)
1225 SOCKET_UNLOCK(sop);
1226 CONTINUE_NO_UNLOCK(udp);
1227 }
1228#endif
1229 POLL_UDP_EVENTS(rc, error, so, &NetworkEvents);
1230
1231 LOG_NAT_SOCK(so, UDP, &NetworkEvents, readfds, writefds, xfds);
1232
1233 if (so->s != -1 && CHECK_FD_SET(so, NetworkEvents, readfds))
1234 {
1235 SORECVFROM(pData, so);
1236 }
1237 LOOP_LABEL(udp, so, so_next);
1238 }
1239
1240 }
1241
1242#ifndef VBOX_WITH_SLIRP_MT
1243 /*
1244 * See if we can start outputting
1245 */
1246 if (if_queued && link_up)
1247 if_start(pData);
1248#endif
1249
1250 STAM_PROFILE_STOP(&pData->StatPoll, a);
1251}
1252
1253#define ETH_ALEN 6
1254#define ETH_HLEN 14
1255
1256#define ARPOP_REQUEST 1 /* ARP request */
1257#define ARPOP_REPLY 2 /* ARP reply */
1258
1259struct ethhdr
1260{
1261 unsigned char h_dest[ETH_ALEN]; /* destination eth addr */
1262 unsigned char h_source[ETH_ALEN]; /* source ether addr */
1263 unsigned short h_proto; /* packet type ID field */
1264};
1265
1266struct arphdr
1267{
1268 unsigned short ar_hrd; /* format of hardware address */
1269 unsigned short ar_pro; /* format of protocol address */
1270 unsigned char ar_hln; /* length of hardware address */
1271 unsigned char ar_pln; /* length of protocol address */
1272 unsigned short ar_op; /* ARP opcode (command) */
1273
1274 /*
1275 * Ethernet looks like this : This bit is variable sized however...
1276 */
1277 unsigned char ar_sha[ETH_ALEN]; /* sender hardware address */
1278 unsigned char ar_sip[4]; /* sender IP address */
1279 unsigned char ar_tha[ETH_ALEN]; /* target hardware address */
1280 unsigned char ar_tip[4]; /* target IP address */
1281};
1282
1283static
1284#ifdef VBOX_WITH_SIMPLIFIED_SLIRP_SYNC
1285void arp_input(PNATState pData, struct mbuf *m)
1286#else
1287void arp_input(PNATState pData, const uint8_t *pkt, int pkt_len)
1288#endif
1289{
1290 struct ethhdr *eh;
1291 struct ethhdr *reh;
1292 struct arphdr *ah;
1293 struct arphdr *rah;
1294 int ar_op;
1295 struct ex_list *ex_ptr;
1296 uint32_t htip;
1297#ifndef VBOX_WITH_SIMPLIFIED_SLIRP_SYNC
1298 uint8_t arp_reply[sizeof(struct arphdr) + ETH_HLEN];
1299 eh = (struct ethhdr *)pkt;
1300#else
1301 struct mbuf *mr;
1302 eh = mtod(m, struct ethhdr *);
1303#endif
1304 ah = (struct arphdr *)&eh[1];
1305 htip = ntohl(*(uint32_t*)ah->ar_tip);
1306
1307#ifdef VBOX_WITH_SIMPLIFIED_SLIRP_SYNC
1308 mr = m_get(pData);
1309 mr->m_data += if_maxlinkhdr;
1310 mr->m_len = sizeof(struct arphdr);
1311 rah = mtod(mr, struct arphdr *);
1312#else
1313 reh = (struct ethhdr *)arp_reply;
1314 rah = (struct arphdr *)&reh[1];
1315#endif
1316
1317 ar_op = ntohs(ah->ar_op);
1318 switch(ar_op)
1319 {
1320 case ARPOP_REQUEST:
1321 if ((htip & pData->netmask) == ntohl(special_addr.s_addr))
1322 {
1323 if ( CTL_CHECK(htip,CTL_DNS)
1324 || CTL_CHECK(htip, CTL_ALIAS)
1325 || CTL_CHECK(htip, CTL_TFTP))
1326 goto arp_ok;
1327 for (ex_ptr = exec_list; ex_ptr; ex_ptr = ex_ptr->ex_next)
1328 {
1329 if ((htip & ~pData->netmask) == ex_ptr->ex_addr)
1330 goto arp_ok;
1331 }
1332 return;
1333 arp_ok:
1334
1335#ifndef VBOX_WITH_SIMPLIFIED_SLIRP_SYNC
1336 memcpy(reh->h_dest, eh->h_source, ETH_ALEN);
1337 memcpy(reh->h_source, &special_addr, ETH_ALEN);
1338 reh->h_source[5] = ah->ar_tip[3];
1339 reh->h_proto = htons(ETH_P_ARP);
1340#endif
1341 rah->ar_hrd = htons(1);
1342 rah->ar_pro = htons(ETH_P_IP);
1343 rah->ar_hln = ETH_ALEN;
1344 rah->ar_pln = 4;
1345 rah->ar_op = htons(ARPOP_REPLY);
1346 memcpy(rah->ar_sha, special_ethaddr, ETH_ALEN);
1347
1348 switch (htip & ~pData->netmask)
1349 {
1350 case CTL_DNS:
1351 case CTL_ALIAS:
1352 rah->ar_sha[5] = (uint8_t)(htip & ~pData->netmask);
1353 break;
1354 default:;
1355 }
1356
1357 memcpy(rah->ar_sip, ah->ar_tip, 4);
1358 memcpy(rah->ar_tha, ah->ar_sha, ETH_ALEN);
1359 memcpy(rah->ar_tip, ah->ar_sip, 4);
1360#ifdef VBOX_WITH_SIMPLIFIED_SLIRP_SYNC
1361 if_encap(pData, ETH_P_ARP, mr);
1362 m_free(pData, m);
1363#else
1364 slirp_output(pData->pvUser, arp_reply, sizeof(arp_reply));
1365#endif
1366 }
1367 break;
1368 default:
1369 break;
1370 }
1371}
1372
1373void slirp_input(PNATState pData, const uint8_t *pkt, int pkt_len)
1374{
1375 struct mbuf *m;
1376 int proto;
1377 static bool fWarnedIpv6;
1378
1379 if (pkt_len < ETH_HLEN)
1380 {
1381 LogRel(("NAT: packet having size %d has been ingnored\n", pkt_len));
1382 return;
1383 }
1384
1385 m = m_get(pData);
1386 if (!m)
1387 {
1388 LogRel(("can't allocate new mbuf\n"));
1389 return;
1390 }
1391
1392 /* Note: we add to align the IP header */
1393
1394 if (M_FREEROOM(m) < pkt_len)
1395 m_inc(m, pkt_len);
1396
1397 m->m_len = pkt_len ;
1398 memcpy(m->m_data, pkt, pkt_len);
1399
1400 proto = ntohs(*(uint16_t *)(pkt + 12));
1401 switch(proto)
1402 {
1403 case ETH_P_ARP:
1404#ifdef VBOX_WITH_SIMPLIFIED_SLIRP_SYNC
1405 arp_input(pData, m);
1406#else
1407 arp_input(pData, pkt, pkt_len);
1408 m_free(pData, m);
1409#endif
1410 break;
1411 case ETH_P_IP:
1412 /* Update time. Important if the network is very quiet, as otherwise
1413 * the first outgoing connection gets an incorrect timestamp. */
1414 updtime(pData);
1415 m->m_data += ETH_HLEN;
1416 m->m_len -= ETH_HLEN;
1417 ip_input(pData, m);
1418 break;
1419 case ETH_P_IPV6:
1420 m_free(pData, m);
1421 if (!fWarnedIpv6)
1422 {
1423 LogRel(("NAT: IPv6 not supported\n"));
1424 fWarnedIpv6 = true;
1425 }
1426 break;
1427 default:
1428 LogRel(("NAT: Unsupported protocol %x\n", proto));
1429 m_free(pData, m);
1430 break;
1431 }
1432 RTMemFree((void *)pkt);
1433}
1434
1435/* output the IP packet to the ethernet device */
1436#ifdef VBOX_WITH_SIMPLIFIED_SLIRP_SYNC
1437void if_encap(PNATState pData, uint16_t eth_proto, struct mbuf *m)
1438#else
1439void if_encap(PNATState pData, uint8_t *ip_data, int ip_data_len)
1440#endif
1441{
1442#ifdef VBOX_WITH_SIMPLIFIED_SLIRP_SYNC
1443 struct ethhdr *eh;
1444 uint8_t *buf = RTMemAlloc(1600);
1445 m->m_data -= if_maxlinkhdr;
1446 m->m_len += ETH_HLEN;
1447 eh = mtod(m, struct ethhdr *);
1448#else
1449 uint8_t buf[1600];
1450 struct ethhdr *eh = (struct ethhdr *)buf;
1451
1452 if (ip_data_len + ETH_HLEN > sizeof(buf))
1453 return;
1454
1455 memcpy(buf + sizeof(struct ethhdr), ip_data, ip_data_len);
1456#endif
1457
1458
1459 memcpy(eh->h_dest, client_ethaddr, ETH_ALEN);
1460 memcpy(eh->h_source, special_ethaddr, ETH_ALEN - 1);
1461 /* XXX: not correct */
1462 eh->h_source[5] = CTL_ALIAS;
1463#ifdef VBOX_WITH_SIMPLIFIED_SLIRP_SYNC
1464 eh->h_proto = htons(eth_proto);
1465#if 0
1466 slirp_output(pData->pvUser, m, mtod(m, uint8_t *), m->m_len);
1467#else
1468 memcpy(buf, mtod(m, uint8_t *), m->m_len);
1469 slirp_output(pData->pvUser, NULL, buf, m->m_len);
1470 m_free(pData, m);
1471#endif
1472#else
1473 eh->h_proto = htons(ETH_P_IP);
1474 slirp_output(pData->pvUser, buf, ip_data_len + ETH_HLEN);
1475#endif
1476}
1477
1478int slirp_redir(PNATState pData, int is_udp, int host_port,
1479 struct in_addr guest_addr, int guest_port)
1480{
1481 if (is_udp)
1482 {
1483 if (!udp_listen(pData, htons(host_port), guest_addr.s_addr,
1484 htons(guest_port), 0))
1485 return -1;
1486 }
1487 else
1488 {
1489 if (!solisten(pData, htons(host_port), guest_addr.s_addr,
1490 htons(guest_port), 0))
1491 return -1;
1492 }
1493 return 0;
1494}
1495
1496int slirp_add_exec(PNATState pData, int do_pty, const char *args, int addr_low_byte,
1497 int guest_port)
1498{
1499 return add_exec(&exec_list, do_pty, (char *)args,
1500 addr_low_byte, htons(guest_port));
1501}
1502
1503void slirp_set_ethaddr(PNATState pData, const uint8_t *ethaddr)
1504{
1505 memcpy(client_ethaddr, ethaddr, ETH_ALEN);
1506}
1507
1508#if defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) && defined(RT_OS_WINDOWS)
1509HANDLE *slirp_get_events(PNATState pData)
1510{
1511 return pData->phEvents;
1512}
1513void slirp_register_external_event(PNATState pData, HANDLE hEvent, int index)
1514{
1515 pData->phEvents[index] = hEvent;
1516}
1517#endif
1518
1519unsigned int slirp_get_timeout_ms(PNATState pData)
1520{
1521 if (link_up)
1522 {
1523 if (time_fasttimo)
1524 return 2;
1525 if (do_slowtimo)
1526 return 500; /* see PR_SLOWHZ */
1527 }
1528 return 0;
1529}
1530
1531#ifndef RT_OS_WINDOWS
1532int slirp_get_nsock(PNATState pData)
1533{
1534 return pData->nsock;
1535}
1536#endif
1537
1538/*
1539 * this function called from NAT thread
1540 */
1541void slirp_post_sent(PNATState pData, void *pvArg)
1542{
1543 struct socket *so = 0;
1544 struct tcpcb *tp = 0;
1545 struct mbuf *m = (struct mbuf *)pvArg;
1546 m_free(pData, m);
1547}
1548#ifdef VBOX_WITH_SLIRP_MT
1549void slirp_process_queue(PNATState pData)
1550{
1551 RTReqProcess(pData->pReqQueue, RT_INDEFINITE_WAIT);
1552}
1553void *slirp_get_queue(PNATState pData)
1554{
1555 return pData->pReqQueue;
1556}
1557#endif
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