VirtualBox

source: vbox/trunk/src/VBox/Devices/Network/slirp/slirp.c@ 16897

Last change on this file since 16897 was 16897, checked in by vboxsync, 16 years ago

NAT: defining empty specific for WSA event mask check for non-WSA synchronization

  • Property svn:eol-style set to native
File size: 51.0 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_WIN_CHECK_FD_SET(so, events, fdset ) 0
32# define DO_POLL_EVENTS(rc, error, so, events, label) do {} while (0)
33
34# define DO_CHECK_FD_SET(so, events, fdset) (FD_ISSET((so)->s, fdset))
35# define DO_UNIX_CHECK_FD_SET(so, events, fdset ) 0 /*specific for Unix API */
36# else /* !VBOX_WITH_SIMPLIFIED_SLIRP_SYNC */
37# define DO_ENGAGE_EVENT1(so, fdset, label) \
38 do { \
39 if( so->so_poll_index != -1 \
40 && so->s == polls[so->so_poll_index].fd) { \
41 polls[so->so_poll_index].events |= N_(fdset ## _poll); \
42 break; /* out of this loop */ \
43 } \
44 AssertRelease(poll_index < (nfds)); \
45 AssertRelease(poll_index >= 0 && poll_index < (nfds)); \
46 polls[poll_index].fd = (so)->s; \
47 (so)->so_poll_index = poll_index; \
48 polls[poll_index].events = N_(fdset ## _poll); \
49 polls[poll_index].revents = 0; \
50 poll_index++; \
51 } while(0)
52
53
54# define DO_ENGAGE_EVENT2(so, fdset1, fdset2, label) \
55 do { \
56 if( so->so_poll_index != -1 \
57 && so->s == polls[so->so_poll_index].fd) { \
58 polls[so->so_poll_index].events |= \
59 N_(fdset1 ## _poll) | N_(fdset1 ## _poll); \
60 break; /* out of this loop */ \
61 } \
62 AssertRelease(poll_index < (nfds)); \
63 polls[poll_index].fd = (so)->s; \
64 (so)->so_poll_index = poll_index; \
65 polls[poll_index].events = \
66 N_(fdset1 ## _poll) | N_(fdset1 ## _poll); \
67 poll_index++; \
68 } while(0)
69
70# define DO_POLL_EVENTS(rc, error, so, events, label) do {} while (0)
71
72# define DO_CHECK_FD_SET(so, events, fdset) ( ((so)->so_poll_index != -1) \
73 && ((so)->so_poll_index <= ndfs) \
74 && ((so)->s == polls[so->so_poll_index].fd) \
75 && (polls[(so)->so_poll_index].revents & N_(fdset ## _poll)))
76# define DO_UNIX_CHECK_FD_SET(so, events, fdset ) DO_CHECK_FD_SET((so), (events), fdset) /*specific for Unix 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# define DO_WIN_CHECK_FD_SET(so, events, fdset ) 0
101# define ICMP_ENGAGE_EVENT(so, fdset) do {} while(0)
102#endif /* RT_OS_WINDOWS */
103
104#else /* defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) && defined(RT_OS_WINDOWS) */
105
106/*
107 * On Windows, we will be notified by IcmpSendEcho2() when the response arrives.
108 * So no call to WSAEventSelect necessary.
109 */
110# define ICMP_ENGAGE_EVENT(so, fdset) do {} while(0)
111
112# define DO_ENGAGE_EVENT1(so, fdset1, label) \
113 do { \
114 rc = WSAEventSelect((so)->s, VBOX_SOCKET_EVENT, FD_ALL_EVENTS); \
115 if (rc == SOCKET_ERROR) \
116 { \
117 /* This should not happen */ \
118 error = WSAGetLastError(); \
119 LogRel(("WSAEventSelect (" #label ") error %d (so=%x, socket=%s, event=%x)\n", \
120 error, (so), (so)->s, VBOX_SOCKET_EVENT)); \
121 } \
122 } while(0); \
123 CONTINUE(label)
124
125# define DO_ENGAGE_EVENT2(so, fdset1, fdset2, label) \
126 DO_ENGAGE_EVENT1((so), (fdset1), label)
127
128# define DO_POLL_EVENTS(rc, error, so, events, label) \
129 (rc) = WSAEnumNetworkEvents((so)->s, VBOX_SOCKET_EVENT, (events)); \
130 if ((rc) == SOCKET_ERROR) \
131 { \
132 (error) = WSAGetLastError(); \
133 LogRel(("WSAEnumNetworkEvents " #label " error %d\n", (error))); \
134 CONTINUE(label); \
135 }
136
137# define acceptds_win FD_ACCEPT
138# define acceptds_win_bit FD_ACCEPT_BIT
139
140# define readfds_win FD_READ
141# define readfds_win_bit FD_READ_BIT
142
143# define writefds_win FD_WRITE
144# define writefds_win_bit FD_WRITE_BIT
145
146# define xfds_win FD_OOB
147# define xfds_win_bit FD_OOB_BIT
148
149# define DO_CHECK_FD_SET(so, events, fdset) \
150 (((events).lNetworkEvents & fdset ## _win) && ((events).iErrorCode[fdset ## _win_bit] == 0))
151
152# define DO_WIN_CHECK_FD_SET(so, events, fdset ) DO_CHECK_FD_SET((so), (events), fdset)
153# define DO_UNIX_CHECK_FD_SET(so, events, fdset ) 1 /*specific for Unix API */
154
155#endif /* defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) && defined(RT_OS_WINDOWS) */
156
157#define TCP_ENGAGE_EVENT1(so, fdset) \
158 DO_ENGAGE_EVENT1((so), fdset, tcp)
159
160#define TCP_ENGAGE_EVENT2(so, fdset1, fdset2) \
161 DO_ENGAGE_EVENT2((so), fdset1, fdset2, tcp)
162
163#define UDP_ENGAGE_EVENT(so, fdset) \
164 DO_ENGAGE_EVENT1((so), fdset, udp)
165
166#define POLL_TCP_EVENTS(rc, error, so, events) \
167 DO_POLL_EVENTS((rc), (error), (so), (events), tcp)
168
169#define POLL_UDP_EVENTS(rc, error, so, events) \
170 DO_POLL_EVENTS((rc), (error), (so), (events), udp)
171
172#define CHECK_FD_SET(so, events, set) \
173 (DO_CHECK_FD_SET((so), (events), set))
174
175#define WIN_CHECK_FD_SET(so, events, set) \
176 (DO_WIN_CHECK_FD_SET((so), (events), set))
177#define UNIX_CHECK_FD_SET(so, events, set) \
178 (DO_UNIX_CHECK_FD_SET(so, events, set))
179
180/*
181 * Loging macros
182 */
183#if VBOX_WITH_DEBUG_NAT_SOCKETS
184# if defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC)
185# if defined(RT_OS_WINDOWS)
186# define DO_LOG_NAT_SOCK(so, proto, winevent, r_fdset, w_fdset, x_fdset) \
187 do { \
188 LogRel((" " #proto " %R[natsock] %R[natwinnetevents]\n", (so), (winevent))); \
189 } while (0)
190# else /* RT_OS_WINDOWS */
191# define DO_LOG_NAT_SOCK(so, proto, winevent, r_fdset, w_fdset, x_fdset) \
192 do { \
193 LogRel((" " #proto " %R[natsock] %s %s %s er: %s, %s, %s\n", (so), \
194 CHECK_FD_SET(so, ign ,r_fdset) ? "READ":"", \
195 CHECK_FD_SET(so, ign, w_fdset) ? "WRITE":"", \
196 CHECK_FD_SET(so, ign, x_fdset) ? "OOB":"", \
197 CHECK_FD_SET(so, ign, rderr) ? "RDERR":"", \
198 CHECK_FD_SET(so, ign, rdhup) ? "RDHUP":"", \
199 CHECK_FD_SET(so, ign, nval) ? "RDNVAL":"")); \
200 } while (0)
201# endif /* !RT_OS_WINDOWS */
202# else /* VBOX_WITH_SIMPLIFIED_SLIRP_SYNC */
203# define DO_LOG_NAT_SOCK(so, proto, winevent, r_fdset, w_fdset, x_fdset) \
204 do { \
205 LogRel((" " #proto " %R[natsock] %s %s %s\n", (so), FD_ISSET((so)->s, (r_fdset))?"READ":"", \
206 FD_ISSET((so)->s, (w_fdset))?"WRITE":"", FD_ISSET((so)->s, (x_fdset))?"OOB":"")); \
207 } while (0)
208# endif
209#else /* VBOX_WITH_DEBUG_NAT_SOCKETS */
210# define DO_LOG_NAT_SOCK(so, proto, winevent, r_fdset, w_fdset, x_fdset) do {} while (0)
211#endif /* !VBOX_WITH_DEBUG_NAT_SOCKETS */
212
213#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)
214
215static const uint8_t special_ethaddr[6] =
216{
217 0x52, 0x54, 0x00, 0x12, 0x35, 0x00
218};
219
220#ifdef RT_OS_WINDOWS
221
222static int get_dns_addr_domain(PNATState pData, bool fVerbose,
223 struct in_addr *pdns_addr,
224 const char **ppszDomain)
225{
226 int rc = 0;
227 FIXED_INFO *FixedInfo = NULL;
228 ULONG BufLen;
229 DWORD ret;
230 IP_ADDR_STRING *pIPAddr;
231 struct in_addr tmp_addr;
232
233 FixedInfo = (FIXED_INFO *)GlobalAlloc(GPTR, sizeof(FIXED_INFO));
234 BufLen = sizeof(FIXED_INFO);
235
236 /** @todo: this API returns all DNS servers, no matter whether the
237 * corresponding network adapter is disabled or not. Maybe replace
238 * this by GetAdapterAddresses(), which is XP/Vista only though. */
239 if (ERROR_BUFFER_OVERFLOW == GetNetworkParams(FixedInfo, &BufLen))
240 {
241 if (FixedInfo)
242 {
243 GlobalFree(FixedInfo);
244 FixedInfo = NULL;
245 }
246 FixedInfo = GlobalAlloc(GPTR, BufLen);
247 }
248
249 if ((ret = GetNetworkParams(FixedInfo, &BufLen)) != ERROR_SUCCESS)
250 {
251 Log(("GetNetworkParams failed. ret = %08x\n", (u_int)ret ));
252 if (FixedInfo)
253 {
254 GlobalFree(FixedInfo);
255 FixedInfo = NULL;
256 }
257 rc = -1;
258 goto get_dns_prefix;
259 }
260
261#ifndef VBOX_WITH_MULTI_DNS
262 pIPAddr = &(FixedInfo->DnsServerList);
263 inet_aton(pIPAddr->IpAddress.String, &tmp_addr);
264 Log(("nat: DNS Servers:\n"));
265 if (fVerbose || pdns_addr->s_addr != tmp_addr.s_addr)
266 LogRel(("NAT: DNS address: %s\n", pIPAddr->IpAddress.String));
267 *pdns_addr = tmp_addr;
268
269 pIPAddr = FixedInfo -> DnsServerList.Next;
270 while (pIPAddr)
271 {
272 if (fVerbose)
273 LogRel(("NAT: ignored DNS address: %s\n", pIPAddr ->IpAddress.String));
274 pIPAddr = pIPAddr ->Next;
275 }
276#else
277 /*localhost mask */
278 for (pIPAddr = &FixedInfo->DnsServerList; pIPAddr != NULL; pIPAddr = pIPAddr->Next)
279 {
280 struct dns_entry *da;
281 if(!inet_aton(pIPAddr->IpAddress.String, &tmp_addr))
282 continue;
283 da = RTMemAllocZ(sizeof (struct dns_entry));
284 if (da == NULL)
285 {
286 LogRel(("can't alloc memory for DNS entry\n"));
287 return -1;
288 }
289 /*check */
290 if ((da->de_addr.s_addr & htonl(IN_CLASSA_NET)) == ntohl(INADDR_LOOPBACK & IN_CLASSA_NET)) {
291 da->de_addr.s_addr = htonl(ntohl(special_addr.s_addr) | CTL_ALIAS);
292 }
293 LIST_INSERT_HEAD(&pData->dns_list_head, da, de_list);
294 }
295#endif
296 if (FixedInfo)
297 {
298 GlobalFree(FixedInfo);
299 FixedInfo = NULL;
300 }
301
302get_dns_prefix:
303 if (ppszDomain)
304 {
305 OSVERSIONINFO ver;
306 char szDnsDomain[256];
307 DWORD dwSize = sizeof(szDnsDomain);
308
309 *ppszDomain = NULL;
310 GetVersionEx(&ver);
311 if (ver.dwMajorVersion >= 5)
312 {
313 /* GetComputerNameEx exists in Windows versions starting with 2000. */
314 if (GetComputerNameEx(ComputerNameDnsDomain, szDnsDomain, &dwSize))
315 {
316 if (szDnsDomain[0])
317 {
318 /* Just non-empty strings are valid. */
319 *ppszDomain = RTStrDup(szDnsDomain);
320 if (pData->fPassDomain)
321 {
322 if (fVerbose)
323 LogRel(("NAT: passing domain name %s\n", szDnsDomain));
324 }
325 else
326 Log(("nat: ignoring domain %s\n", szDnsDomain));
327 }
328 }
329 else
330 Log(("nat: GetComputerNameEx failed (%d)\n", GetLastError()));
331 }
332 }
333 return rc;
334}
335
336#else
337
338static int get_dns_addr_domain(PNATState pData, bool fVerbose,
339 struct in_addr *pdns_addr,
340 const char **ppszDomain)
341{
342 char buff[512];
343 char buff2[256];
344 FILE *f;
345 int found = 0;
346 struct in_addr tmp_addr;
347
348#ifdef RT_OS_OS2
349 /* Try various locations. */
350 char *etc = getenv("ETC");
351 f = NULL;
352 if (etc)
353 {
354 snprintf(buff, sizeof(buff), "%s/RESOLV2", etc);
355 f = fopen(buff, "rt");
356 }
357 if (!f)
358 {
359 snprintf(buff, sizeof(buff), "%s/RESOLV2", _PATH_ETC);
360 f = fopen(buff, "rt");
361 }
362 if (!f)
363 {
364 snprintf(buff, sizeof(buff), "%s/resolv.conf", _PATH_ETC);
365 f = fopen(buff, "rt");
366 }
367#else
368 f = fopen("/etc/resolv.conf", "r");
369#endif
370 if (!f)
371 return -1;
372
373 if (ppszDomain)
374 *ppszDomain = NULL;
375 Log(("nat: DNS Servers:\n"));
376 while (fgets(buff, 512, f) != NULL)
377 {
378#ifdef VBOX_WITH_MULTI_DNS
379 struct dns_entry *da = NULL;
380#endif
381 if (sscanf(buff, "nameserver%*[ \t]%256s", buff2) == 1)
382 {
383 if (!inet_aton(buff2, &tmp_addr))
384 continue;
385#ifndef VBOX_WITH_MULTI_DNS
386 /* If it's the first one, set it to dns_addr */
387 if (!found)
388 {
389 if (fVerbose || pdns_addr->s_addr != tmp_addr.s_addr)
390 LogRel(("NAT: DNS address: %s\n", buff2));
391 *pdns_addr = tmp_addr;
392 }
393 else
394 {
395 if (fVerbose)
396 LogRel(("NAT: ignored DNS address: %s\n", buff2));
397 }
398#else
399 /*localhost mask */
400 da = RTMemAllocZ(sizeof (struct dns_entry));
401 if (da == NULL)
402 {
403 LogRel(("can't alloc memory for DNS entry\n"));
404 return -1;
405 }
406 /*check */
407 da->de_addr.s_addr = tmp_addr.s_addr;
408 if ((da->de_addr.s_addr & htonl(IN_CLASSA_NET)) == ntohl(INADDR_LOOPBACK & IN_CLASSA_NET)) {
409 da->de_addr.s_addr = htonl(ntohl(special_addr.s_addr) | CTL_ALIAS);
410 }
411 LIST_INSERT_HEAD(&pData->dns_list_head, da, de_list);
412#endif
413 found++;
414 }
415 if ( ppszDomain
416 && (!strncmp(buff, "domain", 6) || !strncmp(buff, "search", 6)))
417 {
418 /* Domain name/search list present. Pick first entry */
419 if (*ppszDomain == NULL)
420 {
421 char *tok;
422 char *saveptr;
423 tok = strtok_r(&buff[6], " \t\n", &saveptr);
424 if (tok)
425 {
426 *ppszDomain = RTStrDup(tok);
427 if (pData->fPassDomain)
428 {
429 if (fVerbose)
430 LogRel(("NAT: passing domain name %s\n", tok));
431 }
432 else
433 Log(("nat: ignoring domain %s\n", tok));
434 }
435 }
436 }
437 }
438 fclose(f);
439 if (!found)
440 return -1;
441 return 0;
442}
443
444#endif
445
446int get_dns_addr(PNATState pData, struct in_addr *pdns_addr)
447{
448 return get_dns_addr_domain(pData, false, pdns_addr, NULL);
449}
450
451int slirp_init(PNATState *ppData, const char *pszNetAddr, uint32_t u32Netmask,
452 bool fPassDomain, const char *pszTFTPPrefix,
453 const char *pszBootFile, void *pvUser)
454{
455 int fNATfailed = 0;
456 int rc;
457 PNATState pData = RTMemAllocZ(sizeof(NATState));
458 *ppData = pData;
459 if (!pData)
460 return VERR_NO_MEMORY;
461 if (u32Netmask & 0x1f)
462 /* CTL is x.x.x.15, bootp passes up to 16 IPs (15..31) */
463 return VERR_INVALID_PARAMETER;
464 pData->fPassDomain = fPassDomain;
465 pData->pvUser = pvUser;
466 tftp_prefix = pszTFTPPrefix;
467 bootp_filename = pszBootFile;
468 pData->netmask = u32Netmask;
469
470#ifdef RT_OS_WINDOWS
471 {
472 WSADATA Data;
473 WSAStartup(MAKEWORD(2,0), &Data);
474 }
475# if defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC)
476 pData->phEvents[VBOX_SOCKET_EVENT_INDEX] = CreateEvent(NULL, FALSE, FALSE, NULL);
477# endif
478#endif
479#ifdef VBOX_WITH_SLIRP_MT
480 QSOCKET_LOCK_CREATE(tcb);
481 QSOCKET_LOCK_CREATE(udb);
482 rc = RTReqCreateQueue(&pData->pReqQueue);
483 AssertReleaseRC(rc);
484#endif
485
486 link_up = 1;
487
488 debug_init();
489 if_init(pData);
490 ip_init(pData);
491 icmp_init(pData);
492
493 /* Initialise mbufs *after* setting the MTU */
494 m_init(pData);
495
496 inet_aton(pszNetAddr, &special_addr);
497 alias_addr.s_addr = special_addr.s_addr | htonl(CTL_ALIAS);
498 /* @todo: add ability to configure this staff */
499 pData->tftp_server.s_addr = htonl(ntohl(special_addr.s_addr) | CTL_TFTP);
500
501 /* set default addresses */
502 inet_aton("127.0.0.1", &loopback_addr);
503#ifndef VBOX_WITH_MULTI_DNS
504 inet_aton("127.0.0.1", &dns_addr);
505
506 if (get_dns_addr_domain(pData, true, &dns_addr, &pData->pszDomain) < 0)
507#else
508 LIST_INIT(&pData->dns_list_head);
509 if (get_dns_addr_domain(pData, true, NULL, &pData->pszDomain) < 0)
510#endif
511 fNATfailed = 1;
512
513 getouraddr(pData);
514 return fNATfailed ? VINF_NAT_DNS : VINF_SUCCESS;
515}
516
517/**
518 * Statistics counters.
519 */
520void slirp_register_timers(PNATState pData, PPDMDRVINS pDrvIns)
521{
522#ifdef VBOX_WITH_STATISTICS
523 PDMDrvHlpSTAMRegisterF(pDrvIns, &pData->StatFill, STAMTYPE_PROFILE, STAMVISIBILITY_ALWAYS,
524 STAMUNIT_TICKS_PER_CALL, "Profiling slirp fills", "/Drivers/NAT%d/Fill", pDrvIns->iInstance);
525 PDMDrvHlpSTAMRegisterF(pDrvIns, &pData->StatPoll, STAMTYPE_PROFILE, STAMVISIBILITY_ALWAYS,
526 STAMUNIT_TICKS_PER_CALL, "Profiling slirp polls", "/Drivers/NAT%d/Poll", pDrvIns->iInstance);
527 PDMDrvHlpSTAMRegisterF(pDrvIns, &pData->StatFastTimer, STAMTYPE_PROFILE, STAMVISIBILITY_ALWAYS,
528 STAMUNIT_TICKS_PER_CALL, "Profiling slirp fast timer", "/Drivers/NAT%d/TimerFast", pDrvIns->iInstance);
529 PDMDrvHlpSTAMRegisterF(pDrvIns, &pData->StatSlowTimer, STAMTYPE_PROFILE, STAMVISIBILITY_ALWAYS,
530 STAMUNIT_TICKS_PER_CALL, "Profiling slirp slow timer", "/Drivers/NAT%d/TimerSlow", pDrvIns->iInstance);
531 PDMDrvHlpSTAMRegisterF(pDrvIns, &pData->StatTCP, STAMTYPE_COUNTER, STAMVISIBILITY_ALWAYS,
532 STAMUNIT_COUNT, "TCP sockets", "/Drivers/NAT%d/SockTCP", pDrvIns->iInstance);
533 PDMDrvHlpSTAMRegisterF(pDrvIns, &pData->StatTCPHot, STAMTYPE_COUNTER, STAMVISIBILITY_ALWAYS,
534 STAMUNIT_COUNT, "TCP sockets active", "/Drivers/NAT%d/SockTCPHot", pDrvIns->iInstance);
535 PDMDrvHlpSTAMRegisterF(pDrvIns, &pData->StatUDP, STAMTYPE_COUNTER, STAMVISIBILITY_ALWAYS,
536 STAMUNIT_COUNT, "UDP sockets", "/Drivers/NAT%d/SockUDP", pDrvIns->iInstance);
537 PDMDrvHlpSTAMRegisterF(pDrvIns, &pData->StatUDPHot, STAMTYPE_COUNTER, STAMVISIBILITY_ALWAYS,
538 STAMUNIT_COUNT, "UDP sockets active", "/Drivers/NAT%d/SockUDPHot", pDrvIns->iInstance);
539#endif /* VBOX_WITH_STATISTICS */
540}
541
542/**
543 * Marks the link as up, making it possible to establish new connections.
544 */
545void slirp_link_up(PNATState pData)
546{
547 link_up = 1;
548}
549
550/**
551 * Marks the link as down and cleans up the current connections.
552 */
553void slirp_link_down(PNATState pData)
554{
555 struct socket *so;
556
557 while ((so = tcb.so_next) != &tcb)
558 {
559 if (so->so_state & SS_NOFDREF || so->s == -1)
560 sofree(pData, so);
561 else
562 tcp_drop(pData, sototcpcb(so), 0);
563 }
564
565 while ((so = udb.so_next) != &udb)
566 udp_detach(pData, so);
567
568 link_up = 0;
569}
570
571/**
572 * Terminates the slirp component.
573 */
574void slirp_term(PNATState pData)
575{
576#ifdef VBOX_WITH_MULTI_DNS
577 struct dns_entry *de = NULL;
578#endif
579 if (pData->pszDomain)
580 RTStrFree((char *)(void *)pData->pszDomain);
581
582#ifdef RT_OS_WINDOWS
583 pData->pfIcmpCloseHandle(pData->icmp_socket.sh);
584 FreeLibrary(pData->hmIcmpLibrary);
585 RTMemFree(pData->pvIcmpBuffer);
586# else
587 closesocket(pData->icmp_socket.s);
588#endif
589
590 slirp_link_down(pData);
591#ifdef VBOX_WITH_MULTI_DNS
592 while(!LIST_EMPTY(&pData->dns_list_head)) {
593 de = LIST_FIRST(&pData->dns_list_head);
594 LIST_REMOVE(de, de_list);
595 RTMemFree(de);
596 }
597#endif
598#ifdef RT_OS_WINDOWS
599 WSACleanup();
600#endif
601#ifdef LOG_ENABLED
602 Log(("\n"
603 "NAT statistics\n"
604 "--------------\n"
605 "\n"));
606 ipstats(pData);
607 tcpstats(pData);
608 udpstats(pData);
609 icmpstats(pData);
610 mbufstats(pData);
611 sockstats(pData);
612 Log(("\n"
613 "\n"
614 "\n"));
615#endif
616 RTMemFree(pData);
617}
618
619
620#define CONN_CANFSEND(so) (((so)->so_state & (SS_FCANTSENDMORE|SS_ISFCONNECTED)) == SS_ISFCONNECTED)
621#define CONN_CANFRCV(so) (((so)->so_state & (SS_FCANTRCVMORE|SS_ISFCONNECTED)) == SS_ISFCONNECTED)
622#define UPD_NFDS(x) if (nfds < (x)) nfds = (x)
623
624/*
625 * curtime kept to an accuracy of 1ms
626 */
627#ifdef RT_OS_WINDOWS
628static void updtime(PNATState pData)
629{
630 struct _timeb tb;
631
632 _ftime(&tb);
633 curtime = (u_int)tb.time * (u_int)1000;
634 curtime += (u_int)tb.millitm;
635}
636#else
637static void updtime(PNATState pData)
638{
639 gettimeofday(&tt, 0);
640
641 curtime = (u_int)tt.tv_sec * (u_int)1000;
642 curtime += (u_int)tt.tv_usec / (u_int)1000;
643
644 if ((tt.tv_usec % 1000) >= 500)
645 curtime++;
646}
647#endif
648
649#ifndef VBOX_WITH_SIMPLIFIED_SLIRP_SYNC
650void slirp_select_fill(PNATState pData, int *pnfds,
651 fd_set *readfds, fd_set *writefds, fd_set *xfds)
652#else /* !VBOX_WITH_SIMPLIFIED_SLIRP_SYNC */
653# ifdef RT_OS_WINDOWS
654void slirp_select_fill(PNATState pData, int *pnfds)
655# else /* RT_OS_WINDOWS */
656void slirp_select_fill(PNATState pData, int *pnfds, struct pollfd *polls)
657# endif /* !RT_OS_WINDOWS */
658#endif /* VBOX_WITH_SIMPLIFIED_SLIRP_SYNC */
659{
660 struct socket *so, *so_next;
661 int nfds;
662#if defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) && defined(RT_OS_WINDOWS)
663 int rc;
664 int error;
665#endif
666#if defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) && !defined(RT_OS_WINDOWS)
667 int poll_index = 0;
668#endif
669 int i;
670
671 STAM_PROFILE_START(&pData->StatFill, a);
672
673 nfds = *pnfds;
674
675 /*
676 * First, TCP sockets
677 */
678 do_slowtimo = 0;
679 if (link_up)
680 {
681 /*
682 * *_slowtimo needs calling if there are IP fragments
683 * in the fragment queue, or there are TCP connections active
684 */
685 /* XXX:
686 * triggering of fragment expiration should be the same but use new macroses
687 */
688 do_slowtimo = (tcb.so_next != &tcb);
689 if (!do_slowtimo)
690 {
691 for (i = 0; i < IPREASS_NHASH; i++)
692 {
693 if (!TAILQ_EMPTY(&ipq[i]))
694 {
695 do_slowtimo = 1;
696 break;
697 }
698 }
699 }
700 ICMP_ENGAGE_EVENT(&pData->icmp_socket, readfds);
701
702 STAM_COUNTER_RESET(&pData->StatTCP);
703 STAM_COUNTER_RESET(&pData->StatTCPHot);
704
705 QSOCKET_FOREACH(so, so_next, tcp)
706 /* { */
707#if defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) && !defined(RT_OS_WINDOWS)
708 so->so_poll_index = -1;
709#endif
710 STAM_COUNTER_INC(&pData->StatTCP);
711
712 /*
713 * See if we need a tcp_fasttimo
714 */
715 if ( time_fasttimo == 0
716 && so->so_tcpcb != NULL
717 && so->so_tcpcb->t_flags & TF_DELACK)
718 time_fasttimo = curtime; /* Flag when we want a fasttimo */
719
720 /*
721 * NOFDREF can include still connecting to local-host,
722 * newly socreated() sockets etc. Don't want to select these.
723 */
724 if (so->so_state & SS_NOFDREF || so->s == -1)
725 CONTINUE(tcp);
726
727 /*
728 * Set for reading sockets which are accepting
729 */
730 if (so->so_state & SS_FACCEPTCONN)
731 {
732 STAM_COUNTER_INC(&pData->StatTCPHot);
733 TCP_ENGAGE_EVENT1(so, readfds);
734 CONTINUE(tcp);
735 }
736
737 /*
738 * Set for writing sockets which are connecting
739 */
740 if (so->so_state & SS_ISFCONNECTING)
741 {
742 Log2(("connecting %R[natsock] engaged\n",so));
743 STAM_COUNTER_INC(&pData->StatTCPHot);
744 TCP_ENGAGE_EVENT1(so, writefds);
745 }
746
747 /*
748 * Set for writing if we are connected, can send more, and
749 * we have something to send
750 */
751 if (CONN_CANFSEND(so) && so->so_rcv.sb_cc)
752 {
753 STAM_COUNTER_INC(&pData->StatTCPHot);
754 TCP_ENGAGE_EVENT1(so, writefds);
755 }
756
757 /*
758 * Set for reading (and urgent data) if we are connected, can
759 * receive more, and we have room for it XXX /2 ?
760 */
761 if (CONN_CANFRCV(so) && (so->so_snd.sb_cc < (so->so_snd.sb_datalen/2)))
762 {
763 STAM_COUNTER_INC(&pData->StatTCPHot);
764 TCP_ENGAGE_EVENT2(so, readfds, xfds);
765 }
766 LOOP_LABEL(tcp, so, so_next);
767 }
768
769 /*
770 * UDP sockets
771 */
772 STAM_COUNTER_RESET(&pData->StatUDP);
773 STAM_COUNTER_RESET(&pData->StatUDPHot);
774
775 QSOCKET_FOREACH(so, so_next, udp)
776 /* { */
777
778 STAM_COUNTER_INC(&pData->StatUDP);
779#if defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) && !defined(RT_OS_WINDOWS)
780 so->so_poll_index = -1;
781#endif
782
783 /*
784 * See if it's timed out
785 */
786 if (so->so_expire)
787 {
788 if (so->so_expire <= curtime)
789 {
790#ifdef VBOX_WITH_SLIRP_MT
791 /* we need so_next for continue our cycle*/
792 so_next = so->so_next;
793#endif
794 UDP_DETACH(pData, so, so_next);
795 CONTINUE_NO_UNLOCK(udp);
796 }
797 else
798 do_slowtimo = 1; /* Let socket expire */
799 }
800
801 /*
802 * When UDP packets are received from over the link, they're
803 * sendto()'d straight away, so no need for setting for writing
804 * Limit the number of packets queued by this session to 4.
805 * Note that even though we try and limit this to 4 packets,
806 * the session could have more queued if the packets needed
807 * to be fragmented.
808 *
809 * (XXX <= 4 ?)
810 */
811 if ((so->so_state & SS_ISFCONNECTED) && so->so_queued <= 4)
812 {
813 STAM_COUNTER_INC(&pData->StatUDPHot);
814 UDP_ENGAGE_EVENT(so, readfds);
815 }
816 LOOP_LABEL(udp, so, so_next);
817 }
818
819 }
820
821#if defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC)
822# if defined(RT_OS_WINDOWS)
823 *pnfds = VBOX_EVENT_COUNT;
824# else /* RT_OS_WINDOWS */
825 AssertRelease(poll_index <= *pnfds);
826 *pnfds = poll_index;
827# endif /* !RT_OS_WINDOWS */
828#else /* VBOX_WITH_SIMPLIFIED_SLIRP_SYNC */
829 *pnfds = nfds;
830#endif /* !VBOX_WITH_SIMPLIFIED_SLIRP_SYNC */
831
832 STAM_PROFILE_STOP(&pData->StatFill, a);
833}
834
835#if defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC)
836# if defined(RT_OS_WINDOWS)
837void slirp_select_poll(PNATState pData, int fTimeout, int fIcmp)
838# else /* RT_OS_WINDOWS */
839void slirp_select_poll(PNATState pData, struct pollfd *polls, int ndfs)
840# endif /* !RT_OS_WINDOWS */
841#else /* VBOX_WITH_SIMPLIFIED_SLIRP_SYNC */
842void slirp_select_poll(PNATState pData, fd_set *readfds, fd_set *writefds, fd_set *xfds)
843#endif /* !VBOX_WITH_SIMPLIFIED_SLIRP_SYNC */
844{
845 struct socket *so, *so_next;
846 int ret;
847#if defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) && defined(RT_OS_WINDOWS)
848 WSANETWORKEVENTS NetworkEvents;
849 int rc;
850 int error;
851#endif
852#if defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) && !defined(RT_OS_WINDOWS)
853 int poll_index = 0;
854#endif
855
856 STAM_PROFILE_START(&pData->StatPoll, a);
857
858 /* Update time */
859 updtime(pData);
860
861 /*
862 * See if anything has timed out
863 */
864 if (link_up)
865 {
866 if (time_fasttimo && ((curtime - time_fasttimo) >= 2))
867 {
868 STAM_PROFILE_START(&pData->StatFastTimer, a);
869 tcp_fasttimo(pData);
870 time_fasttimo = 0;
871 STAM_PROFILE_STOP(&pData->StatFastTimer, a);
872 }
873 if (do_slowtimo && ((curtime - last_slowtimo) >= 499))
874 {
875 STAM_PROFILE_START(&pData->StatSlowTimer, a);
876 ip_slowtimo(pData);
877 tcp_slowtimo(pData);
878 last_slowtimo = curtime;
879 STAM_PROFILE_STOP(&pData->StatSlowTimer, a);
880 }
881 }
882#if defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) && defined(RT_OS_WINDOWS)
883 if (fTimeout)
884 return; /* only timer update */
885#endif
886
887 /*
888 * Check sockets
889 */
890 if (link_up)
891 {
892#if defined(RT_OS_WINDOWS)
893 /*XXX: before renaming please make see define
894 * fIcmp in slirp_state.h
895 */
896 if (fIcmp)
897 sorecvfrom(pData, &pData->icmp_socket);
898#else
899 if ( (pData->icmp_socket.s != -1)
900 && CHECK_FD_SET(&pData->icmp_socket, ignored, readfds))
901 sorecvfrom(pData, &pData->icmp_socket);
902#endif
903 /*
904 * Check TCP sockets
905 */
906 QSOCKET_FOREACH(so, so_next, tcp)
907 /* { */
908
909#ifdef VBOX_WITH_SLIRP_MT
910 if ( so->so_state & SS_NOFDREF
911 && so->so_deleted == 1)
912 {
913 struct socket *son, *sop = NULL;
914 QSOCKET_LOCK(tcb);
915 if (so->so_next != NULL)
916 {
917 if (so->so_next != &tcb)
918 SOCKET_LOCK(so->so_next);
919 son = so->so_next;
920 }
921 if ( so->so_prev != &tcb
922 && so->so_prev != NULL)
923 {
924 SOCKET_LOCK(so->so_prev);
925 sop = so->so_prev;
926 }
927 QSOCKET_UNLOCK(tcb);
928 remque(pData, so);
929 NSOCK_DEC();
930 SOCKET_UNLOCK(so);
931 SOCKET_LOCK_DESTROY(so);
932 RTMemFree(so);
933 so_next = son;
934 if (sop != NULL)
935 SOCKET_UNLOCK(sop);
936 CONTINUE_NO_UNLOCK(tcp);
937 }
938#endif
939 /*
940 * FD_ISSET is meaningless on these sockets
941 * (and they can crash the program)
942 */
943 if (so->so_state & SS_NOFDREF || so->s == -1)
944 CONTINUE(tcp);
945
946 POLL_TCP_EVENTS(rc, error, so, &NetworkEvents);
947
948 LOG_NAT_SOCK(so, TCP, &NetworkEvents, readfds, writefds, xfds);
949
950
951 /*
952 * Check for URG data
953 * This will soread as well, so no need to
954 * test for readfds below if this succeeds
955 */
956
957 /* out-of-band data */
958 if (CHECK_FD_SET(so, NetworkEvents, xfds))
959 {
960 sorecvoob(pData, so);
961 }
962
963 /*
964 * Check sockets for reading
965 */
966 else if ( CHECK_FD_SET(so, NetworkEvents, readfds)
967 || WIN_CHECK_FD_SET(so, NetworkEvents, acceptds))
968 {
969 /*
970 * Check for incoming connections
971 */
972 if (so->so_state & SS_FACCEPTCONN)
973 {
974 TCP_CONNECT(pData, so);
975#if defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) && defined(RT_OS_WINDOWS)
976 if (!(NetworkEvents.lNetworkEvents & FD_CLOSE))
977#endif
978 CONTINUE(tcp);
979 }
980
981 ret = soread(pData, so);
982 /* Output it if we read something */
983 if (ret > 0)
984 TCP_OUTPUT(pData, sototcpcb(so));
985 }
986
987#if defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) && defined(RT_OS_WINDOWS)
988 /*
989 * Check for FD_CLOSE events.
990 * in some cases once FD_CLOSE engaged on socket it could be flashed latter (for some reasons)
991 */
992 if ( (NetworkEvents.lNetworkEvents & FD_CLOSE)
993 || (so->so_close == 1))
994 {
995 so->so_close = 1; /* mark it */
996 /*
997 * drain the socket
998 */
999 for (;;)
1000 {
1001 ret = soread(pData, so);
1002 if (ret > 0)
1003 TCP_OUTPUT(pData, sototcpcb(so));
1004 else
1005 break;
1006 }
1007 CONTINUE(tcp);
1008 }
1009#endif
1010
1011 /*
1012 * Check sockets for writing
1013 */
1014 if (CHECK_FD_SET(so, NetworkEvents, writefds))
1015 {
1016 /*
1017 * Check for non-blocking, still-connecting sockets
1018 */
1019 if (so->so_state & SS_ISFCONNECTING)
1020 {
1021 Log2(("connecting %R[natsock] catched\n", so));
1022 /* Connected */
1023 so->so_state &= ~SS_ISFCONNECTING;
1024
1025 /*
1026 * This should be probably guarded by PROBE_CONN too. Anyway,
1027 * we disable it on OS/2 because the below send call returns
1028 * EFAULT which causes the opened TCP socket to close right
1029 * after it has been opened and connected.
1030 */
1031#ifndef RT_OS_OS2
1032 ret = send(so->s, (const char *)&ret, 0, 0);
1033 if (ret < 0)
1034 {
1035 /* XXXXX Must fix, zero bytes is a NOP */
1036 if ( errno == EAGAIN
1037 || errno == EWOULDBLOCK
1038 || errno == EINPROGRESS
1039 || errno == ENOTCONN)
1040 CONTINUE(tcp);
1041
1042 /* else failed */
1043 so->so_state = SS_NOFDREF;
1044 }
1045 /* else so->so_state &= ~SS_ISFCONNECTING; */
1046#endif
1047
1048 /*
1049 * Continue tcp_input
1050 */
1051 TCP_INPUT(pData, (struct mbuf *)NULL, sizeof(struct ip), so);
1052 /* continue; */
1053 }
1054 else
1055 SOWRITE(ret, pData, so);
1056 /*
1057 * XXX If we wrote something (a lot), there could be the need
1058 * for a window update. In the worst case, the remote will send
1059 * a window probe to get things going again.
1060 */
1061 }
1062
1063 /*
1064 * Probe a still-connecting, non-blocking socket
1065 * to check if it's still alive
1066 */
1067#ifdef PROBE_CONN
1068 if (so->so_state & SS_ISFCONNECTING)
1069 {
1070 ret = recv(so->s, (char *)&ret, 0, 0);
1071
1072 if (ret < 0)
1073 {
1074 /* XXX */
1075 if ( errno == EAGAIN
1076 || errno == EWOULDBLOCK
1077 || errno == EINPROGRESS
1078 || errno == ENOTCONN)
1079 {
1080 CONTINUE(tcp); /* Still connecting, continue */
1081 }
1082
1083 /* else failed */
1084 so->so_state = SS_NOFDREF;
1085
1086 /* tcp_input will take care of it */
1087 }
1088 else
1089 {
1090 ret = send(so->s, &ret, 0, 0);
1091 if (ret < 0)
1092 {
1093 /* XXX */
1094 if ( errno == EAGAIN
1095 || errno == EWOULDBLOCK
1096 || errno == EINPROGRESS
1097 || errno == ENOTCONN)
1098 {
1099 CONTINUE(tcp);
1100 }
1101 /* else failed */
1102 so->so_state = SS_NOFDREF;
1103 }
1104 else
1105 so->so_state &= ~SS_ISFCONNECTING;
1106
1107 }
1108 TCP_INPUT((struct mbuf *)NULL, sizeof(struct ip),so);
1109 } /* SS_ISFCONNECTING */
1110#endif
1111#ifndef RT_OS_WINDOWS
1112 if ( UNIX_CHECK_FD_SET(so, NetworkEvents, rdhup)
1113 || UNIX_CHECK_FD_SET(so, NetworkEvents, rderr))
1114 {
1115 int err;
1116 int inq = -1, outq;
1117 socklen_t optlen = sizeof(int);
1118 int status = getsockopt(so->s, SOL_SOCKET, SO_ERROR, &err, &optlen);
1119 if (status != 0)
1120 LogRel(("NAT:can't get error status from %R[natsock]\n", so));
1121#ifndef RT_OS_SOLARIS
1122 status = ioctl(so->s, FIONREAD, &inq); /* tcp(7) recommends SIOCINQ which is Linux specific */
1123 if (status != 0 || status != EINVAL)
1124 {
1125 /* EINVAL returned if socket in listen state tcp(7)*/
1126 LogRel(("NAT:can't get depth of IN queue status from %R[natsock]\n", so));
1127 }
1128 status = ioctl(so->s, TIOCOUTQ, &outq); /* SIOCOUTQ see previous comment */
1129 if (status != 0)
1130 LogRel(("NAT:can't get depth of OUT queue from %R[natsock]\n", so));
1131#else
1132 inq = outq = -1;
1133 /*
1134 * Solaris has bit different ioctl commands and its handlings
1135 * hint: streamio(7) I_NREAD
1136 */
1137#endif
1138 if ( so->so_state & SS_ISFCONNECTING
1139 || UNIX_CHECK_FD_SET(so, NetworkEvents, readfds))
1140 {
1141 /**
1142 * Check if we need here take care about gracefull connection
1143 * @todo try with proxy server
1144 */
1145 if (UNIX_CHECK_FD_SET(so, NetworkEvents, readfds))
1146 {
1147 /*
1148 * Never meet inq != 0 or outq != 0, anyway let it stay for a while
1149 * in case it happens we'll able to detect it.
1150 * Give TCP/IP stack wait or expire the socket.
1151 */
1152 LogRel(("NAT:%R[natsock] err(%d:%s) s(in:%d,out:%d)happens on read I/O, "
1153 "other side close connection \n", so, err, strerror(err), inq, outq));
1154 CONTINUE(tcp);
1155 }
1156 goto tcp_input_close;
1157 }
1158 if ( !UNIX_CHECK_FD_SET(so, NetworkEvents, readfds)
1159 && !UNIX_CHECK_FD_SET(so, NetworkEvents, writefds)
1160 && !UNIX_CHECK_FD_SET(so, NetworkEvents, xfds))
1161 {
1162 LogRel(("NAT:system expires the socket %R[natsock] err(%d:%s) s(in:%d,out:%d) happens on non-I/O. ",
1163 so, err, strerror(err), inq, outq));
1164 goto tcp_input_close;
1165 }
1166 LogRel(("NAT:%R[natsock] we've met(%d:%s) s(in:%d, out:%d) unhandled combination hup (%d) "
1167 "rederr(%d) on (r:%d, w:%d, x:%d)\n",
1168 so, err, strerror(err),
1169 inq, outq,
1170 UNIX_CHECK_FD_SET(so, ign, rdhup),
1171 UNIX_CHECK_FD_SET(so, ign, rderr),
1172 UNIX_CHECK_FD_SET(so, ign, readfds),
1173 UNIX_CHECK_FD_SET(so, ign, writefds),
1174 UNIX_CHECK_FD_SET(so, ign, xfds)));
1175 /*
1176 * Give OS's TCP/IP stack a chance to resolve an issue or expire the socket.
1177 */
1178 CONTINUE(tcp);
1179tcp_input_close:
1180 so->so_state = SS_NOFDREF; /*cause connection valid tcp connection termination and socket closing */
1181 TCP_INPUT(pData, (struct mbuf *)NULL, sizeof(struct ip), so);
1182 CONTINUE(tcp);
1183 }
1184#endif
1185 LOOP_LABEL(tcp, so, so_next);
1186 }
1187
1188 /*
1189 * Now UDP sockets.
1190 * Incoming packets are sent straight away, they're not buffered.
1191 * Incoming UDP data isn't buffered either.
1192 */
1193 QSOCKET_FOREACH(so, so_next, udp)
1194 /* { */
1195#ifdef VBOX_WITH_SLIRP_MT
1196 if ( so->so_state & SS_NOFDREF
1197 && so->so_deleted == 1)
1198 {
1199 struct socket *son, *sop = NULL;
1200 QSOCKET_LOCK(udb);
1201 if (so->so_next != NULL)
1202 {
1203 if (so->so_next != &udb)
1204 SOCKET_LOCK(so->so_next);
1205 son = so->so_next;
1206 }
1207 if ( so->so_prev != &udb
1208 && so->so_prev != NULL)
1209 {
1210 SOCKET_LOCK(so->so_prev);
1211 sop = so->so_prev;
1212 }
1213 QSOCKET_UNLOCK(udb);
1214 remque(pData, so);
1215 NSOCK_DEC();
1216 SOCKET_UNLOCK(so);
1217 SOCKET_LOCK_DESTROY(so);
1218 RTMemFree(so);
1219 so_next = son;
1220 if (sop != NULL)
1221 SOCKET_UNLOCK(sop);
1222 CONTINUE_NO_UNLOCK(udp);
1223 }
1224#endif
1225 POLL_UDP_EVENTS(rc, error, so, &NetworkEvents);
1226
1227 LOG_NAT_SOCK(so, UDP, &NetworkEvents, readfds, writefds, xfds);
1228
1229 if (so->s != -1 && CHECK_FD_SET(so, NetworkEvents, readfds))
1230 {
1231 SORECVFROM(pData, so);
1232 }
1233 LOOP_LABEL(udp, so, so_next);
1234 }
1235
1236 }
1237
1238#ifndef VBOX_WITH_SLIRP_MT
1239 /*
1240 * See if we can start outputting
1241 */
1242 if (if_queued && link_up)
1243 if_start(pData);
1244#endif
1245
1246 STAM_PROFILE_STOP(&pData->StatPoll, a);
1247}
1248
1249#define ETH_ALEN 6
1250#define ETH_HLEN 14
1251
1252#define ARPOP_REQUEST 1 /* ARP request */
1253#define ARPOP_REPLY 2 /* ARP reply */
1254
1255struct ethhdr
1256{
1257 unsigned char h_dest[ETH_ALEN]; /* destination eth addr */
1258 unsigned char h_source[ETH_ALEN]; /* source ether addr */
1259 unsigned short h_proto; /* packet type ID field */
1260};
1261
1262struct arphdr
1263{
1264 unsigned short ar_hrd; /* format of hardware address */
1265 unsigned short ar_pro; /* format of protocol address */
1266 unsigned char ar_hln; /* length of hardware address */
1267 unsigned char ar_pln; /* length of protocol address */
1268 unsigned short ar_op; /* ARP opcode (command) */
1269
1270 /*
1271 * Ethernet looks like this : This bit is variable sized however...
1272 */
1273 unsigned char ar_sha[ETH_ALEN]; /* sender hardware address */
1274 unsigned char ar_sip[4]; /* sender IP address */
1275 unsigned char ar_tha[ETH_ALEN]; /* target hardware address */
1276 unsigned char ar_tip[4]; /* target IP address */
1277};
1278
1279static
1280#ifdef VBOX_WITH_SIMPLIFIED_SLIRP_SYNC
1281void arp_input(PNATState pData, struct mbuf *m)
1282#else
1283void arp_input(PNATState pData, const uint8_t *pkt, int pkt_len)
1284#endif
1285{
1286 struct ethhdr *eh;
1287 struct ethhdr *reh;
1288 struct arphdr *ah;
1289 struct arphdr *rah;
1290 int ar_op;
1291 struct ex_list *ex_ptr;
1292 uint32_t htip;
1293#ifndef VBOX_WITH_SIMPLIFIED_SLIRP_SYNC
1294 uint8_t arp_reply[sizeof(struct arphdr) + ETH_HLEN];
1295 eh = (struct ethhdr *)pkt;
1296#else
1297 struct mbuf *mr;
1298 eh = mtod(m, struct ethhdr *);
1299#endif
1300 ah = (struct arphdr *)&eh[1];
1301 htip = ntohl(*(uint32_t*)ah->ar_tip);
1302
1303#ifdef VBOX_WITH_SIMPLIFIED_SLIRP_SYNC
1304 mr = m_get(pData);
1305 mr->m_data += if_maxlinkhdr;
1306 mr->m_len = sizeof(struct arphdr);
1307 rah = mtod(mr, struct arphdr *);
1308#else
1309 reh = (struct ethhdr *)arp_reply;
1310 rah = (struct arphdr *)&reh[1];
1311#endif
1312
1313 ar_op = ntohs(ah->ar_op);
1314 switch(ar_op)
1315 {
1316 case ARPOP_REQUEST:
1317 if ((htip & pData->netmask) == ntohl(special_addr.s_addr))
1318 {
1319 if ( CTL_CHECK(htip,CTL_DNS)
1320 || CTL_CHECK(htip, CTL_ALIAS)
1321 || CTL_CHECK(htip, CTL_TFTP))
1322 goto arp_ok;
1323 for (ex_ptr = exec_list; ex_ptr; ex_ptr = ex_ptr->ex_next)
1324 {
1325 if ((htip & ~pData->netmask) == ex_ptr->ex_addr)
1326 goto arp_ok;
1327 }
1328 return;
1329 arp_ok:
1330
1331#ifndef VBOX_WITH_SIMPLIFIED_SLIRP_SYNC
1332 memcpy(reh->h_dest, eh->h_source, ETH_ALEN);
1333 memcpy(reh->h_source, &special_addr, ETH_ALEN);
1334 reh->h_source[5] = ah->ar_tip[3];
1335 reh->h_proto = htons(ETH_P_ARP);
1336#endif
1337 rah->ar_hrd = htons(1);
1338 rah->ar_pro = htons(ETH_P_IP);
1339 rah->ar_hln = ETH_ALEN;
1340 rah->ar_pln = 4;
1341 rah->ar_op = htons(ARPOP_REPLY);
1342 memcpy(rah->ar_sha, special_ethaddr, ETH_ALEN);
1343
1344 switch (htip & ~pData->netmask)
1345 {
1346 case CTL_DNS:
1347 case CTL_ALIAS:
1348 rah->ar_sha[5] = (uint8_t)(htip & ~pData->netmask);
1349 break;
1350 default:;
1351 }
1352
1353 memcpy(rah->ar_sip, ah->ar_tip, 4);
1354 memcpy(rah->ar_tha, ah->ar_sha, ETH_ALEN);
1355 memcpy(rah->ar_tip, ah->ar_sip, 4);
1356#ifdef VBOX_WITH_SIMPLIFIED_SLIRP_SYNC
1357 if_encap(pData, ETH_P_ARP, mr);
1358 m_free(pData, m);
1359#else
1360 slirp_output(pData->pvUser, arp_reply, sizeof(arp_reply));
1361#endif
1362 }
1363 break;
1364 default:
1365 break;
1366 }
1367}
1368
1369void slirp_input(PNATState pData, const uint8_t *pkt, int pkt_len)
1370{
1371 struct mbuf *m;
1372 int proto;
1373 static bool fWarnedIpv6;
1374
1375 if (pkt_len < ETH_HLEN)
1376 {
1377 LogRel(("NAT: packet having size %d has been ingnored\n", pkt_len));
1378 return;
1379 }
1380
1381 m = m_get(pData);
1382 if (!m)
1383 {
1384 LogRel(("can't allocate new mbuf\n"));
1385 return;
1386 }
1387
1388 /* Note: we add to align the IP header */
1389
1390 if (M_FREEROOM(m) < pkt_len)
1391 m_inc(m, pkt_len);
1392
1393 m->m_len = pkt_len ;
1394 memcpy(m->m_data, pkt, pkt_len);
1395
1396 proto = ntohs(*(uint16_t *)(pkt + 12));
1397 switch(proto)
1398 {
1399 case ETH_P_ARP:
1400#ifdef VBOX_WITH_SIMPLIFIED_SLIRP_SYNC
1401 arp_input(pData, m);
1402#else
1403 arp_input(pData, pkt, pkt_len);
1404 m_free(pData, m);
1405#endif
1406 break;
1407 case ETH_P_IP:
1408 /* Update time. Important if the network is very quiet, as otherwise
1409 * the first outgoing connection gets an incorrect timestamp. */
1410 updtime(pData);
1411 m->m_data += ETH_HLEN;
1412 m->m_len -= ETH_HLEN;
1413 ip_input(pData, m);
1414 break;
1415 case ETH_P_IPV6:
1416 m_free(pData, m);
1417 if (!fWarnedIpv6)
1418 {
1419 LogRel(("NAT: IPv6 not supported\n"));
1420 fWarnedIpv6 = true;
1421 }
1422 break;
1423 default:
1424 LogRel(("NAT: Unsupported protocol %x\n", proto));
1425 m_free(pData, m);
1426 break;
1427 }
1428 RTMemFree((void *)pkt);
1429}
1430
1431/* output the IP packet to the ethernet device */
1432#ifdef VBOX_WITH_SIMPLIFIED_SLIRP_SYNC
1433void if_encap(PNATState pData, uint16_t eth_proto, struct mbuf *m)
1434#else
1435void if_encap(PNATState pData, uint8_t *ip_data, int ip_data_len)
1436#endif
1437{
1438#ifdef VBOX_WITH_SIMPLIFIED_SLIRP_SYNC
1439 struct ethhdr *eh;
1440 uint8_t *buf = RTMemAlloc(1600);
1441 m->m_data -= if_maxlinkhdr;
1442 m->m_len += ETH_HLEN;
1443 eh = mtod(m, struct ethhdr *);
1444#else
1445 uint8_t buf[1600];
1446 struct ethhdr *eh = (struct ethhdr *)buf;
1447
1448 if (ip_data_len + ETH_HLEN > sizeof(buf))
1449 return;
1450
1451 memcpy(buf + sizeof(struct ethhdr), ip_data, ip_data_len);
1452#endif
1453
1454
1455 memcpy(eh->h_dest, client_ethaddr, ETH_ALEN);
1456 memcpy(eh->h_source, special_ethaddr, ETH_ALEN - 1);
1457 /* XXX: not correct */
1458 eh->h_source[5] = CTL_ALIAS;
1459#ifdef VBOX_WITH_SIMPLIFIED_SLIRP_SYNC
1460 eh->h_proto = htons(eth_proto);
1461#if 0
1462 slirp_output(pData->pvUser, m, mtod(m, uint8_t *), m->m_len);
1463#else
1464 memcpy(buf, mtod(m, uint8_t *), m->m_len);
1465 slirp_output(pData->pvUser, NULL, buf, m->m_len);
1466 m_free(pData, m);
1467#endif
1468#else
1469 eh->h_proto = htons(ETH_P_IP);
1470 slirp_output(pData->pvUser, buf, ip_data_len + ETH_HLEN);
1471#endif
1472}
1473
1474int slirp_redir(PNATState pData, int is_udp, int host_port,
1475 struct in_addr guest_addr, int guest_port)
1476{
1477 if (is_udp)
1478 {
1479 if (!udp_listen(pData, htons(host_port), guest_addr.s_addr,
1480 htons(guest_port), 0))
1481 return -1;
1482 }
1483 else
1484 {
1485 if (!solisten(pData, htons(host_port), guest_addr.s_addr,
1486 htons(guest_port), 0))
1487 return -1;
1488 }
1489 return 0;
1490}
1491
1492int slirp_add_exec(PNATState pData, int do_pty, const char *args, int addr_low_byte,
1493 int guest_port)
1494{
1495 return add_exec(&exec_list, do_pty, (char *)args,
1496 addr_low_byte, htons(guest_port));
1497}
1498
1499void slirp_set_ethaddr(PNATState pData, const uint8_t *ethaddr)
1500{
1501 memcpy(client_ethaddr, ethaddr, ETH_ALEN);
1502}
1503
1504#if defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) && defined(RT_OS_WINDOWS)
1505HANDLE *slirp_get_events(PNATState pData)
1506{
1507 return pData->phEvents;
1508}
1509void slirp_register_external_event(PNATState pData, HANDLE hEvent, int index)
1510{
1511 pData->phEvents[index] = hEvent;
1512}
1513#endif
1514
1515unsigned int slirp_get_timeout_ms(PNATState pData)
1516{
1517 if (link_up)
1518 {
1519 if (time_fasttimo)
1520 return 2;
1521 if (do_slowtimo)
1522 return 500; /* see PR_SLOWHZ */
1523 }
1524 return 0;
1525}
1526
1527#ifndef RT_OS_WINDOWS
1528int slirp_get_nsock(PNATState pData)
1529{
1530 return pData->nsock;
1531}
1532#endif
1533
1534/*
1535 * this function called from NAT thread
1536 */
1537void slirp_post_sent(PNATState pData, void *pvArg)
1538{
1539 struct socket *so = 0;
1540 struct tcpcb *tp = 0;
1541 struct mbuf *m = (struct mbuf *)pvArg;
1542 m_free(pData, m);
1543}
1544#ifdef VBOX_WITH_SLIRP_MT
1545void slirp_process_queue(PNATState pData)
1546{
1547 RTReqProcess(pData->pReqQueue, RT_INDEFINITE_WAIT);
1548}
1549void *slirp_get_queue(PNATState pData)
1550{
1551 return pData->pReqQueue;
1552}
1553#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