VirtualBox

source: vbox/trunk/src/VBox/Devices/Network/slirp/slirp.c@ 16725

Last change on this file since 16725 was 16725, checked in by vboxsync, 16 years ago

NAT: using per-socket index of pollfd, while updating corresponding element of pollfd array
  • Property svn:eol-style set to native
File size: 48.1 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 <poll.h>
11#endif
12
13#if !defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) || !defined(RT_OS_WINDOWS)
14
15# ifndef VBOX_WITH_SIMPLIFIED_SLIRP_SYNC
16# define DO_ENGAGE_EVENT1(so, fdset, label) \
17 do { \
18 FD_SET((so)->s, (fdset)); \
19 UPD_NFDS((so)->s); \
20 } while(0)
21
22
23# define DO_ENGAGE_EVENT2(so, fdset1, fdset2, label) \
24 do { \
25 FD_SET((so)->s, (fdset1)); \
26 FD_SET((so)->s, (fdset2)); \
27 UPD_NFDS((so)->s); \
28 } while(0)
29
30# define DO_POLL_EVENTS(rc, error, so, events, label) do {} while (0)
31
32# define DO_CHECK_FD_SET(so, events, fdset) (FD_ISSET((so)->s, fdset))
33# define DO_UNIX_CHECK_FD_SET(so, events, fdset ) 0 /*specific for Unix API */
34# else /* !VBOX_WITH_SIMPLIFIED_SLIRP_SYNC */
35# define DO_ENGAGE_EVENT1(so, fdset, label) \
36 do { \
37 if( so->so_poll_index != -1 \
38 && so->s == polls[so->so_poll_index].fd) { \
39 polls[so->so_poll_index].events |= N_(fdset ## _poll); \
40 break; /* out of this loop */ \
41 } \
42 AssertRelease(poll_index < (nfds)); \
43 AssertRelease(poll_index >= 0 && poll_index < (nfds)); \
44 polls[poll_index].fd = (so)->s; \
45 (so)->so_poll_index = poll_index; \
46 polls[poll_index].events = N_(fdset ## _poll); \
47 polls[poll_index].revents = 0; \
48 poll_index++; \
49 } while(0)
50
51
52# define DO_ENGAGE_EVENT2(so, fdset1, fdset2, label) \
53 do { \
54 if( so->so_poll_index != -1 \
55 && so->s == polls[so->so_poll_index].fd) { \
56 polls[so->so_poll_index].events |= \
57 N_(fdset1 ## _poll) | N_(fdset1 ## _poll); \
58 break; /* out of this loop */ \
59 } \
60 AssertRelease(poll_index < (nfds)); \
61 polls[poll_index].fd = (so)->s; \
62 (so)->so_poll_index = poll_index; \
63 polls[poll_index].events = \
64 N_(fdset1 ## _poll) | N_(fdset1 ## _poll); \
65 poll_index++; \
66 } while(0)
67
68# define DO_POLL_EVENTS(rc, error, so, events, label) do {} while (0)
69
70# define DO_CHECK_FD_SET(so, events, fdset) ( ((so)->so_poll_index != -1) \
71 && ((so)->so_poll_index <= ndfs) \
72 && ((so)->s == polls[so->so_poll_index].fd) \
73 && (polls[(so)->so_poll_index].revents & N_(fdset ## _poll)))
74# define DO_UNIX_CHECK_FD_SET(so, events, fdset ) DO_CHECK_FD_SET((so), (events), fdset) /*specific for Unix API */
75# define DO_WIN_CHECK_FD_SET(so, events, fdset ) 0 /* specific for Windows Winsock API */
76# endif /* VBOX_WITH_SIMPLIFIED_SLIRP_SYNC */
77
78# ifndef RT_OS_WINDOWS
79
80# ifndef RT_OS_LINUX
81# define readfds_poll (POLLRDNORM)
82# define writefds_poll (POLLWRNORM)
83# define xfds_poll (POLLRDBAND|POLLWRBAND|POLLPRI)
84# else
85# define readfds_poll (POLLIN)
86# define writefds_poll (POLLOUT)
87# define xfds_poll (POLLPRI)
88# endif
89# define rderr_poll (POLLERR)
90# define rdhup_poll (POLLHUP)
91# define nval_poll (POLLNVAL)
92
93# define ICMP_ENGAGE_EVENT(so, fdset) \
94 do { \
95 if (pData->icmp_socket.s != -1) \
96 DO_ENGAGE_EVENT1((so), fdset, ICMP); \
97 } while (0)
98# else /* !RT_OS_WINDOWS */
99# define DO_WIN_CHECK_FD_SET(so, events, fdset ) DO_CHECK_FD_SET((so), (events), fdset)
100# define ICMP_ENGAGE_EVENT(so, fdset) do {} while(0)
101#endif /* RT_OS_WINDOWS */
102
103#else /* defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) && defined(RT_OS_WINDOWS) */
104
105/*
106 * On Windows, we will be notified by IcmpSendEcho2() when the response arrives.
107 * So no call to WSAEventSelect necessary.
108 */
109# define ICMP_ENGAGE_EVENT(so, fdset) do {} while(0)
110
111# define DO_ENGAGE_EVENT1(so, fdset1, label) \
112 do { \
113 rc = WSAEventSelect((so)->s, VBOX_SOCKET_EVENT, FD_ALL_EVENTS); \
114 if (rc == SOCKET_ERROR) \
115 { \
116 /* This should not happen */ \
117 error = WSAGetLastError(); \
118 LogRel(("WSAEventSelect (" #label ") error %d (so=%x, socket=%s, event=%x)\n", \
119 error, (so), (so)->s, VBOX_SOCKET_EVENT)); \
120 } \
121 } while(0); \
122 CONTINUE(label)
123
124# define DO_ENGAGE_EVENT2(so, fdset1, fdset2, label) \
125 DO_ENGAGE_EVENT1((so), (fdset1), label)
126
127# define DO_POLL_EVENTS(rc, error, so, events, label) \
128 (rc) = WSAEnumNetworkEvents((so)->s, VBOX_SOCKET_EVENT, (events)); \
129 if ((rc) == SOCKET_ERROR) \
130 { \
131 (error) = WSAGetLastError(); \
132 LogRel(("WSAEnumNetworkEvents " #label " error %d\n", (error))); \
133 CONTINUE(label); \
134 }
135
136# define acceptds_win FD_ACCEPT
137# define acceptds_win_bit FD_ACCEPT_BIT
138
139# define readfds_win FD_READ
140# define readfds_win_bit FD_READ_BIT
141
142# define writefds_win FD_WRITE
143# define writefds_win_bit FD_WRITE_BIT
144
145# define xfds_win FD_OOB
146# define xfds_win_bit FD_OOB_BIT
147
148# define DO_CHECK_FD_SET(so, events, fdset) \
149 (((events).lNetworkEvents & fdset ## _win) && ((events).iErrorCode[fdset ## _win_bit] == 0))
150
151# define DO_WIN_CHECK_FD_SET(so, events, fdset ) DO_CHECK_FD_SET((so), (events), fdset)
152# define DO_UNIX_CHECK_FD_SET(so, events, fdset ) 1 /*specific for Unix API */
153
154#endif /* defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) && defined(RT_OS_WINDOWS) */
155
156#define TCP_ENGAGE_EVENT1(so, fdset) \
157 DO_ENGAGE_EVENT1((so), fdset, tcp)
158
159#define TCP_ENGAGE_EVENT2(so, fdset1, fdset2) \
160 DO_ENGAGE_EVENT2((so), fdset1, fdset2, tcp)
161
162#define UDP_ENGAGE_EVENT(so, fdset) \
163 DO_ENGAGE_EVENT1((so), fdset, udp)
164
165#define POLL_TCP_EVENTS(rc, error, so, events) \
166 DO_POLL_EVENTS((rc), (error), (so), (events), tcp)
167
168#define POLL_UDP_EVENTS(rc, error, so, events) \
169 DO_POLL_EVENTS((rc), (error), (so), (events), udp)
170
171#define CHECK_FD_SET(so, events, set) \
172 (DO_CHECK_FD_SET((so), (events), set))
173
174#define WIN_CHECK_FD_SET(so, events, set) \
175 (DO_WIN_CHECK_FD_SET((so), (events), set))
176#define UNIX_CHECK_FD_SET(so, events, set) \
177 (DO_UNIX_CHECK_FD_SET(so, events, set))
178
179/*
180 * Loging macros
181 */
182#if VBOX_WITH_DEBUG_NAT_SOCKETS
183# if defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC)
184# if defined(RT_OS_WINDOWS)
185# define DO_LOG_NAT_SOCK(so, proto, winevent, r_fdset, w_fdset, x_fdset) \
186 do { \
187 LogRel((" " #proto " %R[natsock] %R[natwinnetevents]\n", (so), (winevent))); \
188 } while (0)
189# else /* 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] %s %s %s er: %s, %s, %s\n", (so), \
193 CHECK_FD_SET(so, ign ,r_fdset) ? "READ":"", \
194 CHECK_FD_SET(so, ign, w_fdset) ? "WRITE":"", \
195 CHECK_FD_SET(so, ign, x_fdset) ? "OOB":"", \
196 CHECK_FD_SET(so, ign, rderr) ? "RDERR":"", \
197 CHECK_FD_SET(so, ign, rdhup) ? "RDHUP":"", \
198 CHECK_FD_SET(so, ign, nval) ? "RDNVAL":"")); \
199 } while (0)
200# endif /* !RT_OS_WINDOWS */
201# else /* VBOX_WITH_SIMPLIFIED_SLIRP_SYNC */
202# define DO_LOG_NAT_SOCK(so, proto, winevent, r_fdset, w_fdset, x_fdset) \
203 do { \
204 LogRel((" " #proto " %R[natsock] %s %s %s\n", (so), FD_ISSET((so)->s, (r_fdset))?"READ":"",\
205 FD_ISSET((so)->s, (w_fdset))?"WRITE":"", FD_ISSET((so)->s, (x_fdset))?"OOB":"")); \
206 } while (0)
207# endif
208#else /* VBOX_WITH_DEBUG_NAT_SOCKETS */
209# define DO_LOG_NAT_SOCK(so, proto, winevent, r_fdset, w_fdset, x_fdset) do {} while (0)
210#endif /* !VBOX_WITH_DEBUG_NAT_SOCKETS */
211
212#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)
213
214static const uint8_t special_ethaddr[6] =
215{
216 0x52, 0x54, 0x00, 0x12, 0x35, 0x00
217};
218
219#ifdef RT_OS_WINDOWS
220
221static int get_dns_addr_domain(PNATState pData, bool fVerbose,
222 struct in_addr *pdns_addr,
223 const char **ppszDomain)
224{
225 int rc = 0;
226 FIXED_INFO *FixedInfo = NULL;
227 ULONG BufLen;
228 DWORD ret;
229 IP_ADDR_STRING *pIPAddr;
230 struct in_addr tmp_addr;
231
232 FixedInfo = (FIXED_INFO *)GlobalAlloc(GPTR, sizeof(FIXED_INFO));
233 BufLen = sizeof(FIXED_INFO);
234
235 /** @todo: this API returns all DNS servers, no matter whether the
236 * corresponding network adapter is disabled or not. Maybe replace
237 * this by GetAdapterAddresses(), which is XP/Vista only though. */
238 if (ERROR_BUFFER_OVERFLOW == GetNetworkParams(FixedInfo, &BufLen))
239 {
240 if (FixedInfo)
241 {
242 GlobalFree(FixedInfo);
243 FixedInfo = NULL;
244 }
245 FixedInfo = GlobalAlloc(GPTR, BufLen);
246 }
247
248 if ((ret = GetNetworkParams(FixedInfo, &BufLen)) != ERROR_SUCCESS)
249 {
250 Log(("GetNetworkParams failed. ret = %08x\n", (u_int)ret ));
251 if (FixedInfo)
252 {
253 GlobalFree(FixedInfo);
254 FixedInfo = NULL;
255 }
256 rc = -1;
257 goto get_dns_prefix;
258 }
259
260#ifndef VBOX_WITH_MULTI_DNS
261 pIPAddr = &(FixedInfo->DnsServerList);
262 inet_aton(pIPAddr->IpAddress.String, &tmp_addr);
263 Log(("nat: DNS Servers:\n"));
264 if (fVerbose || pdns_addr->s_addr != tmp_addr.s_addr)
265 LogRel(("NAT: DNS address: %s\n", pIPAddr->IpAddress.String));
266 *pdns_addr = tmp_addr;
267
268 pIPAddr = FixedInfo -> DnsServerList.Next;
269 while (pIPAddr)
270 {
271 if (fVerbose)
272 LogRel(("NAT: ignored DNS address: %s\n", pIPAddr ->IpAddress.String));
273 pIPAddr = pIPAddr ->Next;
274 }
275#else
276 /*localhost mask */
277 for (pIPAddr = &FixedInfo->DnsServerList; pIPAddr != NULL; pIPAddr = pIPAddr->Next)
278 {
279 struct dns_entry *da;
280 if(!inet_aton(pIPAddr->IpAddress.String, &tmp_addr))
281 continue;
282 da = RTMemAllocZ(sizeof (struct dns_entry));
283 if (da == NULL)
284 {
285 LogRel(("can't alloc memory for DNS entry\n"));
286 return -1;
287 }
288 /*check */
289 if ((da->de_addr.s_addr & htonl(IN_CLASSA_NET)) == ntohl(INADDR_LOOPBACK & IN_CLASSA_NET)) {
290 da->de_addr.s_addr = htonl(ntohl(special_addr.s_addr) | CTL_ALIAS);
291 }
292 LIST_INSERT_HEAD(&pData->dns_list_head, da, de_list);
293 }
294#endif
295 if (FixedInfo)
296 {
297 GlobalFree(FixedInfo);
298 FixedInfo = NULL;
299 }
300
301get_dns_prefix:
302 if (ppszDomain)
303 {
304 OSVERSIONINFO ver;
305 char szDnsDomain[256];
306 DWORD dwSize = sizeof(szDnsDomain);
307
308 *ppszDomain = NULL;
309 GetVersionEx(&ver);
310 if (ver.dwMajorVersion >= 5)
311 {
312 /* GetComputerNameEx exists in Windows versions starting with 2000. */
313 if (GetComputerNameEx(ComputerNameDnsDomain, szDnsDomain, &dwSize))
314 {
315 if (szDnsDomain[0])
316 {
317 /* Just non-empty strings are valid. */
318 *ppszDomain = RTStrDup(szDnsDomain);
319 if (pData->fPassDomain)
320 {
321 if (fVerbose)
322 LogRel(("NAT: passing domain name %s\n", szDnsDomain));
323 }
324 else
325 Log(("nat: ignoring domain %s\n", szDnsDomain));
326 }
327 }
328 else
329 Log(("nat: GetComputerNameEx failed (%d)\n", GetLastError()));
330 }
331 }
332 return rc;
333}
334
335#else
336
337static int get_dns_addr_domain(PNATState pData, bool fVerbose,
338 struct in_addr *pdns_addr,
339 const char **ppszDomain)
340{
341 char buff[512];
342 char buff2[256];
343 FILE *f;
344 int found = 0;
345 struct in_addr tmp_addr;
346
347#ifdef RT_OS_OS2
348 /* Try various locations. */
349 char *etc = getenv("ETC");
350 f = NULL;
351 if (etc)
352 {
353 snprintf(buff, sizeof(buff), "%s/RESOLV2", etc);
354 f = fopen(buff, "rt");
355 }
356 if (!f)
357 {
358 snprintf(buff, sizeof(buff), "%s/RESOLV2", _PATH_ETC);
359 f = fopen(buff, "rt");
360 }
361 if (!f)
362 {
363 snprintf(buff, sizeof(buff), "%s/resolv.conf", _PATH_ETC);
364 f = fopen(buff, "rt");
365 }
366#else
367 f = fopen("/etc/resolv.conf", "r");
368#endif
369 if (!f)
370 return -1;
371
372 if (ppszDomain)
373 *ppszDomain = NULL;
374 Log(("nat: DNS Servers:\n"));
375 while (fgets(buff, 512, f) != NULL)
376 {
377#ifdef VBOX_WITH_MULTI_DNS
378 struct dns_entry *da = NULL;
379#endif
380 if (sscanf(buff, "nameserver%*[ \t]%256s", buff2) == 1)
381 {
382 if (!inet_aton(buff2, &tmp_addr))
383 continue;
384#ifndef VBOX_WITH_MULTI_DNS
385 /* If it's the first one, set it to dns_addr */
386 if (!found)
387 {
388 if (fVerbose || pdns_addr->s_addr != tmp_addr.s_addr)
389 LogRel(("NAT: DNS address: %s\n", buff2));
390 *pdns_addr = tmp_addr;
391 }
392 else
393 {
394 if (fVerbose)
395 LogRel(("NAT: ignored DNS address: %s\n", buff2));
396 }
397#else
398 /*localhost mask */
399 da = RTMemAllocZ(sizeof (struct dns_entry));
400 if (da == NULL)
401 {
402 LogRel(("can't alloc memory for DNS entry\n"));
403 return -1;
404 }
405 /*check */
406 da->de_addr.s_addr = tmp_addr.s_addr;
407 if ((da->de_addr.s_addr & htonl(IN_CLASSA_NET)) == ntohl(INADDR_LOOPBACK & IN_CLASSA_NET)) {
408 da->de_addr.s_addr = htonl(ntohl(special_addr.s_addr) | CTL_ALIAS);
409 }
410 LIST_INSERT_HEAD(&pData->dns_list_head, da, de_list);
411#endif
412 found++;
413 }
414 if ( ppszDomain
415 && (!strncmp(buff, "domain", 6) || !strncmp(buff, "search", 6)))
416 {
417 /* Domain name/search list present. Pick first entry */
418 if (*ppszDomain == NULL)
419 {
420 char *tok;
421 char *saveptr;
422 tok = strtok_r(&buff[6], " \t\n", &saveptr);
423 if (tok)
424 {
425 *ppszDomain = RTStrDup(tok);
426 if (pData->fPassDomain)
427 {
428 if (fVerbose)
429 LogRel(("NAT: passing domain name %s\n", tok));
430 }
431 else
432 Log(("nat: ignoring domain %s\n", tok));
433 }
434 }
435 }
436 }
437 fclose(f);
438 if (!found)
439 return -1;
440 return 0;
441}
442
443#endif
444
445int get_dns_addr(PNATState pData, struct in_addr *pdns_addr)
446{
447 return get_dns_addr_domain(pData, false, pdns_addr, NULL);
448}
449
450int slirp_init(PNATState *ppData, const char *pszNetAddr, uint32_t u32Netmask,
451 bool fPassDomain, const char *pszTFTPPrefix,
452 const char *pszBootFile, void *pvUser)
453{
454 int fNATfailed = 0;
455 int rc;
456 PNATState pData = RTMemAllocZ(sizeof(NATState));
457 *ppData = pData;
458 if (!pData)
459 return VERR_NO_MEMORY;
460 if (u32Netmask & 0x1f)
461 /* CTL is x.x.x.15, bootp passes up to 16 IPs (15..31) */
462 return VERR_INVALID_PARAMETER;
463 memset(pData, '\0', sizeof(NATState));
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 */
991 if (NetworkEvents.lNetworkEvents & FD_CLOSE)
992 {
993 /*
994 * drain the socket
995 */
996 for (;;)
997 {
998 ret = soread(pData, so);
999 if (ret > 0)
1000 TCP_OUTPUT(pData, sototcpcb(so));
1001 else
1002 break;
1003 }
1004 }
1005#endif
1006
1007 /*
1008 * Check sockets for writing
1009 */
1010 if (CHECK_FD_SET(so, NetworkEvents, writefds))
1011 {
1012 /*
1013 * Check for non-blocking, still-connecting sockets
1014 */
1015 if (so->so_state & SS_ISFCONNECTING)
1016 {
1017 Log2(("connecting %R[natsock] catched\n", so));
1018 /* Connected */
1019 so->so_state &= ~SS_ISFCONNECTING;
1020
1021 /*
1022 * This should be probably guarded by PROBE_CONN too. Anyway,
1023 * we disable it on OS/2 because the below send call returns
1024 * EFAULT which causes the opened TCP socket to close right
1025 * after it has been opened and connected.
1026 */
1027#ifndef RT_OS_OS2
1028 ret = send(so->s, (const char *)&ret, 0, 0);
1029 if (ret < 0)
1030 {
1031 /* XXXXX Must fix, zero bytes is a NOP */
1032 if ( errno == EAGAIN
1033 || errno == EWOULDBLOCK
1034 || errno == EINPROGRESS
1035 || errno == ENOTCONN)
1036 CONTINUE(tcp);
1037
1038 /* else failed */
1039 so->so_state = SS_NOFDREF;
1040 }
1041 /* else so->so_state &= ~SS_ISFCONNECTING; */
1042#endif
1043
1044 /*
1045 * Continue tcp_input
1046 */
1047 TCP_INPUT(pData, (struct mbuf *)NULL, sizeof(struct ip), so);
1048 /* continue; */
1049 }
1050 else
1051 SOWRITE(ret, pData, so);
1052 /*
1053 * XXX If we wrote something (a lot), there could be the need
1054 * for a window update. In the worst case, the remote will send
1055 * a window probe to get things going again.
1056 */
1057 }
1058
1059 /*
1060 * Probe a still-connecting, non-blocking socket
1061 * to check if it's still alive
1062 */
1063#ifdef PROBE_CONN
1064 if (so->so_state & SS_ISFCONNECTING)
1065 {
1066 ret = recv(so->s, (char *)&ret, 0, 0);
1067
1068 if (ret < 0)
1069 {
1070 /* XXX */
1071 if ( errno == EAGAIN
1072 || errno == EWOULDBLOCK
1073 || errno == EINPROGRESS
1074 || errno == ENOTCONN)
1075 {
1076 CONTINUE(tcp); /* Still connecting, continue */
1077 }
1078
1079 /* else failed */
1080 so->so_state = SS_NOFDREF;
1081
1082 /* tcp_input will take care of it */
1083 }
1084 else
1085 {
1086 ret = send(so->s, &ret, 0, 0);
1087 if (ret < 0)
1088 {
1089 /* XXX */
1090 if ( errno == EAGAIN
1091 || errno == EWOULDBLOCK
1092 || errno == EINPROGRESS
1093 || errno == ENOTCONN)
1094 {
1095 CONTINUE(tcp);
1096 }
1097 /* else failed */
1098 so->so_state = SS_NOFDREF;
1099 }
1100 else
1101 so->so_state &= ~SS_ISFCONNECTING;
1102
1103 }
1104 TCP_INPUT((struct mbuf *)NULL, sizeof(struct ip),so);
1105 } /* SS_ISFCONNECTING */
1106#endif
1107#ifndef RT_OS_WINDOWS
1108 if ( UNIX_CHECK_FD_SET(so, NetworkEvents, rdhup)
1109 && UNIX_CHECK_FD_SET(so, NetworkEvents, rderr))
1110 {
1111 if ( so->so_state & SS_ISFCONNECTING
1112 || (polls[poll_index].revents & POLLIN) != 0)
1113 {
1114 /**
1115 * Check if we need here take care about gracefull connection
1116 * @todo try with proxy server
1117 */
1118 if ((polls[poll_index].revents & POLLIN) != 0)
1119 LogRel(("NAT: err happens on read I/O, other side close connection \n"));
1120 so->so_state = SS_NOFDREF;
1121 TCP_INPUT(pData, (struct mbuf *)NULL, sizeof(struct ip), so);
1122 CONTINUE(tcp);
1123 }
1124 /* Here should be other error handlings */
1125 AssertRelease(!"shouldn't be here!!!");
1126 }
1127#endif
1128 LOOP_LABEL(tcp, so, so_next);
1129 }
1130
1131 /*
1132 * Now UDP sockets.
1133 * Incoming packets are sent straight away, they're not buffered.
1134 * Incoming UDP data isn't buffered either.
1135 */
1136 QSOCKET_FOREACH(so, so_next, udp)
1137 /* { */
1138#ifdef VBOX_WITH_SLIRP_MT
1139 if ( so->so_state & SS_NOFDREF
1140 && so->so_deleted == 1)
1141 {
1142 struct socket *son, *sop = NULL;
1143 QSOCKET_LOCK(udb);
1144 if (so->so_next != NULL)
1145 {
1146 if (so->so_next != &udb)
1147 SOCKET_LOCK(so->so_next);
1148 son = so->so_next;
1149 }
1150 if ( so->so_prev != &udb
1151 && so->so_prev != NULL)
1152 {
1153 SOCKET_LOCK(so->so_prev);
1154 sop = so->so_prev;
1155 }
1156 QSOCKET_UNLOCK(udb);
1157 remque(pData, so);
1158 NSOCK_DEC();
1159 SOCKET_UNLOCK(so);
1160 SOCKET_LOCK_DESTROY(so);
1161 RTMemFree(so);
1162 so_next = son;
1163 if (sop != NULL)
1164 SOCKET_UNLOCK(sop);
1165 CONTINUE_NO_UNLOCK(udp);
1166 }
1167#endif
1168 POLL_UDP_EVENTS(rc, error, so, &NetworkEvents);
1169
1170 LOG_NAT_SOCK(so, UDP, &NetworkEvents, readfds, writefds, xfds);
1171
1172 if (so->s != -1 && CHECK_FD_SET(so, NetworkEvents, readfds))
1173 {
1174 SORECVFROM(pData, so);
1175 }
1176 LOOP_LABEL(udp, so, so_next);
1177 }
1178
1179 }
1180
1181#ifndef VBOX_WITH_SLIRP_MT
1182 /*
1183 * See if we can start outputting
1184 */
1185 if (if_queued && link_up)
1186 if_start(pData);
1187#endif
1188
1189 STAM_PROFILE_STOP(&pData->StatPoll, a);
1190}
1191
1192#define ETH_ALEN 6
1193#define ETH_HLEN 14
1194
1195#define ARPOP_REQUEST 1 /* ARP request */
1196#define ARPOP_REPLY 2 /* ARP reply */
1197
1198struct ethhdr
1199{
1200 unsigned char h_dest[ETH_ALEN]; /* destination eth addr */
1201 unsigned char h_source[ETH_ALEN]; /* source ether addr */
1202 unsigned short h_proto; /* packet type ID field */
1203};
1204
1205struct arphdr
1206{
1207 unsigned short ar_hrd; /* format of hardware address */
1208 unsigned short ar_pro; /* format of protocol address */
1209 unsigned char ar_hln; /* length of hardware address */
1210 unsigned char ar_pln; /* length of protocol address */
1211 unsigned short ar_op; /* ARP opcode (command) */
1212
1213 /*
1214 * Ethernet looks like this : This bit is variable sized however...
1215 */
1216 unsigned char ar_sha[ETH_ALEN]; /* sender hardware address */
1217 unsigned char ar_sip[4]; /* sender IP address */
1218 unsigned char ar_tha[ETH_ALEN]; /* target hardware address */
1219 unsigned char ar_tip[4]; /* target IP address */
1220};
1221
1222static
1223#ifdef VBOX_WITH_SIMPLIFIED_SLIRP_SYNC
1224void arp_input(PNATState pData, struct mbuf *m)
1225#else
1226void arp_input(PNATState pData, const uint8_t *pkt, int pkt_len)
1227#endif
1228{
1229 struct ethhdr *eh;
1230 struct ethhdr *reh;
1231 struct arphdr *ah;
1232 struct arphdr *rah;
1233 int ar_op;
1234 struct ex_list *ex_ptr;
1235 uint32_t htip;
1236#ifndef VBOX_WITH_SIMPLIFIED_SLIRP_SYNC
1237 uint8_t arp_reply[sizeof(struct arphdr) + ETH_HLEN];
1238 eh = (struct ethhdr *)pkt;
1239#else
1240 struct mbuf *mr;
1241 eh = mtod(m, struct ethhdr *);
1242#endif
1243 ah = (struct arphdr *)&eh[1];
1244 htip = ntohl(*(uint32_t*)ah->ar_tip);
1245
1246#ifdef VBOX_WITH_SIMPLIFIED_SLIRP_SYNC
1247 mr = m_get(pData);
1248 mr->m_data += if_maxlinkhdr;
1249 mr->m_len = sizeof(struct arphdr);
1250 rah = mtod(mr, struct arphdr *);
1251#else
1252 reh = (struct ethhdr *)arp_reply;
1253 rah = (struct arphdr *)&reh[1];
1254#endif
1255
1256 ar_op = ntohs(ah->ar_op);
1257 switch(ar_op)
1258 {
1259 case ARPOP_REQUEST:
1260 if ((htip & pData->netmask) == ntohl(special_addr.s_addr))
1261 {
1262 if ( CTL_CHECK(htip,CTL_DNS)
1263 || CTL_CHECK(htip, CTL_ALIAS)
1264 || CTL_CHECK(htip, CTL_TFTP))
1265 goto arp_ok;
1266 for (ex_ptr = exec_list; ex_ptr; ex_ptr = ex_ptr->ex_next)
1267 {
1268 if ((htip & ~pData->netmask) == ex_ptr->ex_addr)
1269 goto arp_ok;
1270 }
1271 return;
1272 arp_ok:
1273
1274#ifndef VBOX_WITH_SIMPLIFIED_SLIRP_SYNC
1275 memcpy(reh->h_dest, eh->h_source, ETH_ALEN);
1276 memcpy(reh->h_source, &special_addr, ETH_ALEN);
1277 reh->h_source[5] = ah->ar_tip[3];
1278 reh->h_proto = htons(ETH_P_ARP);
1279#endif
1280 rah->ar_hrd = htons(1);
1281 rah->ar_pro = htons(ETH_P_IP);
1282 rah->ar_hln = ETH_ALEN;
1283 rah->ar_pln = 4;
1284 rah->ar_op = htons(ARPOP_REPLY);
1285 memcpy(rah->ar_sha, special_ethaddr, ETH_ALEN);
1286
1287 switch (htip & ~pData->netmask)
1288 {
1289 case CTL_DNS:
1290 case CTL_ALIAS:
1291 rah->ar_sha[5] = (uint8_t)(htip & ~pData->netmask);
1292 break;
1293 default:;
1294 }
1295
1296 memcpy(rah->ar_sip, ah->ar_tip, 4);
1297 memcpy(rah->ar_tha, ah->ar_sha, ETH_ALEN);
1298 memcpy(rah->ar_tip, ah->ar_sip, 4);
1299#ifdef VBOX_WITH_SIMPLIFIED_SLIRP_SYNC
1300 if_encap(pData, ETH_P_ARP, mr);
1301 m_free(pData, m);
1302#else
1303 slirp_output(pData->pvUser, arp_reply, sizeof(arp_reply));
1304#endif
1305 }
1306 break;
1307 default:
1308 break;
1309 }
1310}
1311
1312void slirp_input(PNATState pData, const uint8_t *pkt, int pkt_len)
1313{
1314 struct mbuf *m;
1315 int proto;
1316 static bool fWarnedIpv6;
1317
1318 if (pkt_len < ETH_HLEN)
1319 {
1320 LogRel(("NAT: packet having size %d has been ingnored\n", pkt_len));
1321 return;
1322 }
1323
1324 m = m_get(pData);
1325 if (!m)
1326 {
1327 LogRel(("can't allocate new mbuf\n"));
1328 return;
1329 }
1330
1331 /* Note: we add to align the IP header */
1332
1333 if (M_FREEROOM(m) < pkt_len)
1334 m_inc(m, pkt_len);
1335
1336 m->m_len = pkt_len ;
1337 memcpy(m->m_data, pkt, pkt_len);
1338
1339 proto = ntohs(*(uint16_t *)(pkt + 12));
1340 switch(proto)
1341 {
1342 case ETH_P_ARP:
1343#ifdef VBOX_WITH_SIMPLIFIED_SLIRP_SYNC
1344 arp_input(pData, m);
1345#else
1346 arp_input(pData, pkt, pkt_len);
1347 m_free(pData, m);
1348#endif
1349 break;
1350 case ETH_P_IP:
1351 /* Update time. Important if the network is very quiet, as otherwise
1352 * the first outgoing connection gets an incorrect timestamp. */
1353 updtime(pData);
1354 m->m_data += ETH_HLEN;
1355 m->m_len -= ETH_HLEN;
1356 ip_input(pData, m);
1357 break;
1358 case ETH_P_IPV6:
1359 m_free(pData, m);
1360 if (!fWarnedIpv6)
1361 {
1362 LogRel(("NAT: IPv6 not supported\n"));
1363 fWarnedIpv6 = true;
1364 }
1365 break;
1366 default:
1367 LogRel(("NAT: Unsupported protocol %x\n", proto));
1368 m_free(pData, m);
1369 break;
1370 }
1371 RTMemFree((void *)pkt);
1372}
1373
1374/* output the IP packet to the ethernet device */
1375#ifdef VBOX_WITH_SIMPLIFIED_SLIRP_SYNC
1376void if_encap(PNATState pData, uint16_t eth_proto, struct mbuf *m)
1377#else
1378void if_encap(PNATState pData, uint8_t *ip_data, int ip_data_len)
1379#endif
1380{
1381#ifdef VBOX_WITH_SIMPLIFIED_SLIRP_SYNC
1382 struct ethhdr *eh;
1383 uint8_t *buf = RTMemAlloc(1600);
1384 m->m_data -= if_maxlinkhdr;
1385 m->m_len += ETH_HLEN;
1386 eh = mtod(m, struct ethhdr *);
1387#else
1388 uint8_t buf[1600];
1389 struct ethhdr *eh = (struct ethhdr *)buf;
1390
1391 if (ip_data_len + ETH_HLEN > sizeof(buf))
1392 return;
1393
1394 memcpy(buf + sizeof(struct ethhdr), ip_data, ip_data_len);
1395#endif
1396
1397
1398 memcpy(eh->h_dest, client_ethaddr, ETH_ALEN);
1399 memcpy(eh->h_source, special_ethaddr, ETH_ALEN - 1);
1400 /* XXX: not correct */
1401 eh->h_source[5] = CTL_ALIAS;
1402#ifdef VBOX_WITH_SIMPLIFIED_SLIRP_SYNC
1403 eh->h_proto = htons(eth_proto);
1404#if 0
1405 slirp_output(pData->pvUser, m, mtod(m, uint8_t *), m->m_len);
1406#else
1407 memcpy(buf, mtod(m, uint8_t *), m->m_len);
1408 slirp_output(pData->pvUser, NULL, buf, m->m_len);
1409 m_free(pData, m);
1410#endif
1411#else
1412 eh->h_proto = htons(ETH_P_IP);
1413 slirp_output(pData->pvUser, buf, ip_data_len + ETH_HLEN);
1414#endif
1415}
1416
1417int slirp_redir(PNATState pData, int is_udp, int host_port,
1418 struct in_addr guest_addr, int guest_port)
1419{
1420 if (is_udp)
1421 {
1422 if (!udp_listen(pData, htons(host_port), guest_addr.s_addr,
1423 htons(guest_port), 0))
1424 return -1;
1425 }
1426 else
1427 {
1428 if (!solisten(pData, htons(host_port), guest_addr.s_addr,
1429 htons(guest_port), 0))
1430 return -1;
1431 }
1432 return 0;
1433}
1434
1435int slirp_add_exec(PNATState pData, int do_pty, const char *args, int addr_low_byte,
1436 int guest_port)
1437{
1438 return add_exec(&exec_list, do_pty, (char *)args,
1439 addr_low_byte, htons(guest_port));
1440}
1441
1442void slirp_set_ethaddr(PNATState pData, const uint8_t *ethaddr)
1443{
1444 memcpy(client_ethaddr, ethaddr, ETH_ALEN);
1445}
1446
1447#if defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) && defined(RT_OS_WINDOWS)
1448HANDLE *slirp_get_events(PNATState pData)
1449{
1450 return pData->phEvents;
1451}
1452void slirp_register_external_event(PNATState pData, HANDLE hEvent, int index)
1453{
1454 pData->phEvents[index] = hEvent;
1455}
1456#endif
1457
1458unsigned int slirp_get_timeout_ms(PNATState pData)
1459{
1460 if (link_up)
1461 {
1462 if (time_fasttimo)
1463 return 2;
1464 if (do_slowtimo)
1465 return 500; /* see PR_SLOWHZ */
1466 }
1467 return 0;
1468}
1469
1470#ifndef RT_OS_WINDOWS
1471int slirp_get_nsock(PNATState pData)
1472{
1473 return pData->nsock;
1474}
1475#endif
1476
1477/*
1478 * this function called from NAT thread
1479 */
1480void slirp_post_sent(PNATState pData, void *pvArg)
1481{
1482 struct socket *so = 0;
1483 struct tcpcb *tp = 0;
1484 struct mbuf *m = (struct mbuf *)pvArg;
1485 m_free(pData, m);
1486}
1487#ifdef VBOX_WITH_SLIRP_MT
1488void slirp_process_queue(PNATState pData)
1489{
1490 RTReqProcess(pData->pReqQueue, RT_INDEFINITE_WAIT);
1491}
1492void *slirp_get_queue(PNATState pData)
1493{
1494 return pData->pReqQueue;
1495}
1496#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