VirtualBox

source: vbox/trunk/src/VBox/Devices/Network/slirp/slirp.c@ 16759

Last change on this file since 16759 was 16759, checked in by vboxsync, 16 years ago

NAT: some comment to assert in slirp_select_poll
  • Property svn:eol-style set to native
File size: 50.3 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# define DO_WIN_CHECK_FD_SET(so, events, fdset ) DO_CHECK_FD_SET((so), (events), fdset)
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 memset(pData, '\0', sizeof(NATState));
465 pData->fPassDomain = fPassDomain;
466 pData->pvUser = pvUser;
467 tftp_prefix = pszTFTPPrefix;
468 bootp_filename = pszBootFile;
469 pData->netmask = u32Netmask;
470
471#ifdef RT_OS_WINDOWS
472 {
473 WSADATA Data;
474 WSAStartup(MAKEWORD(2,0), &Data);
475 }
476# if defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC)
477 pData->phEvents[VBOX_SOCKET_EVENT_INDEX] = CreateEvent(NULL, FALSE, FALSE, NULL);
478# endif
479#endif
480#ifdef VBOX_WITH_SLIRP_MT
481 QSOCKET_LOCK_CREATE(tcb);
482 QSOCKET_LOCK_CREATE(udb);
483 rc = RTReqCreateQueue(&pData->pReqQueue);
484 AssertReleaseRC(rc);
485#endif
486
487 link_up = 1;
488
489 debug_init();
490 if_init(pData);
491 ip_init(pData);
492 icmp_init(pData);
493
494 /* Initialise mbufs *after* setting the MTU */
495 m_init(pData);
496
497 inet_aton(pszNetAddr, &special_addr);
498 alias_addr.s_addr = special_addr.s_addr | htonl(CTL_ALIAS);
499 /* @todo: add ability to configure this staff */
500 pData->tftp_server.s_addr = htonl(ntohl(special_addr.s_addr) | CTL_TFTP);
501
502 /* set default addresses */
503 inet_aton("127.0.0.1", &loopback_addr);
504#ifndef VBOX_WITH_MULTI_DNS
505 inet_aton("127.0.0.1", &dns_addr);
506
507 if (get_dns_addr_domain(pData, true, &dns_addr, &pData->pszDomain) < 0)
508#else
509 LIST_INIT(&pData->dns_list_head);
510 if (get_dns_addr_domain(pData, true, NULL, &pData->pszDomain) < 0)
511#endif
512 fNATfailed = 1;
513
514 getouraddr(pData);
515 return fNATfailed ? VINF_NAT_DNS : VINF_SUCCESS;
516}
517
518/**
519 * Statistics counters.
520 */
521void slirp_register_timers(PNATState pData, PPDMDRVINS pDrvIns)
522{
523#ifdef VBOX_WITH_STATISTICS
524 PDMDrvHlpSTAMRegisterF(pDrvIns, &pData->StatFill, STAMTYPE_PROFILE, STAMVISIBILITY_ALWAYS,
525 STAMUNIT_TICKS_PER_CALL, "Profiling slirp fills", "/Drivers/NAT%d/Fill", pDrvIns->iInstance);
526 PDMDrvHlpSTAMRegisterF(pDrvIns, &pData->StatPoll, STAMTYPE_PROFILE, STAMVISIBILITY_ALWAYS,
527 STAMUNIT_TICKS_PER_CALL, "Profiling slirp polls", "/Drivers/NAT%d/Poll", pDrvIns->iInstance);
528 PDMDrvHlpSTAMRegisterF(pDrvIns, &pData->StatFastTimer, STAMTYPE_PROFILE, STAMVISIBILITY_ALWAYS,
529 STAMUNIT_TICKS_PER_CALL, "Profiling slirp fast timer", "/Drivers/NAT%d/TimerFast", pDrvIns->iInstance);
530 PDMDrvHlpSTAMRegisterF(pDrvIns, &pData->StatSlowTimer, STAMTYPE_PROFILE, STAMVISIBILITY_ALWAYS,
531 STAMUNIT_TICKS_PER_CALL, "Profiling slirp slow timer", "/Drivers/NAT%d/TimerSlow", pDrvIns->iInstance);
532 PDMDrvHlpSTAMRegisterF(pDrvIns, &pData->StatTCP, STAMTYPE_COUNTER, STAMVISIBILITY_ALWAYS,
533 STAMUNIT_COUNT, "TCP sockets", "/Drivers/NAT%d/SockTCP", pDrvIns->iInstance);
534 PDMDrvHlpSTAMRegisterF(pDrvIns, &pData->StatTCPHot, STAMTYPE_COUNTER, STAMVISIBILITY_ALWAYS,
535 STAMUNIT_COUNT, "TCP sockets active", "/Drivers/NAT%d/SockTCPHot", pDrvIns->iInstance);
536 PDMDrvHlpSTAMRegisterF(pDrvIns, &pData->StatUDP, STAMTYPE_COUNTER, STAMVISIBILITY_ALWAYS,
537 STAMUNIT_COUNT, "UDP sockets", "/Drivers/NAT%d/SockUDP", pDrvIns->iInstance);
538 PDMDrvHlpSTAMRegisterF(pDrvIns, &pData->StatUDPHot, STAMTYPE_COUNTER, STAMVISIBILITY_ALWAYS,
539 STAMUNIT_COUNT, "UDP sockets active", "/Drivers/NAT%d/SockUDPHot", pDrvIns->iInstance);
540#endif /* VBOX_WITH_STATISTICS */
541}
542
543/**
544 * Marks the link as up, making it possible to establish new connections.
545 */
546void slirp_link_up(PNATState pData)
547{
548 link_up = 1;
549}
550
551/**
552 * Marks the link as down and cleans up the current connections.
553 */
554void slirp_link_down(PNATState pData)
555{
556 struct socket *so;
557
558 while ((so = tcb.so_next) != &tcb)
559 {
560 if (so->so_state & SS_NOFDREF || so->s == -1)
561 sofree(pData, so);
562 else
563 tcp_drop(pData, sototcpcb(so), 0);
564 }
565
566 while ((so = udb.so_next) != &udb)
567 udp_detach(pData, so);
568
569 link_up = 0;
570}
571
572/**
573 * Terminates the slirp component.
574 */
575void slirp_term(PNATState pData)
576{
577#ifdef VBOX_WITH_MULTI_DNS
578 struct dns_entry *de = NULL;
579#endif
580 if (pData->pszDomain)
581 RTStrFree((char *)(void *)pData->pszDomain);
582
583#ifdef RT_OS_WINDOWS
584 pData->pfIcmpCloseHandle(pData->icmp_socket.sh);
585 FreeLibrary(pData->hmIcmpLibrary);
586 RTMemFree(pData->pvIcmpBuffer);
587# else
588 closesocket(pData->icmp_socket.s);
589#endif
590
591 slirp_link_down(pData);
592#ifdef VBOX_WITH_MULTI_DNS
593 while(!LIST_EMPTY(&pData->dns_list_head)) {
594 de = LIST_FIRST(&pData->dns_list_head);
595 LIST_REMOVE(de, de_list);
596 RTMemFree(de);
597 }
598#endif
599#ifdef RT_OS_WINDOWS
600 WSACleanup();
601#endif
602#ifdef LOG_ENABLED
603 Log(("\n"
604 "NAT statistics\n"
605 "--------------\n"
606 "\n"));
607 ipstats(pData);
608 tcpstats(pData);
609 udpstats(pData);
610 icmpstats(pData);
611 mbufstats(pData);
612 sockstats(pData);
613 Log(("\n"
614 "\n"
615 "\n"));
616#endif
617 RTMemFree(pData);
618}
619
620
621#define CONN_CANFSEND(so) (((so)->so_state & (SS_FCANTSENDMORE|SS_ISFCONNECTED)) == SS_ISFCONNECTED)
622#define CONN_CANFRCV(so) (((so)->so_state & (SS_FCANTRCVMORE|SS_ISFCONNECTED)) == SS_ISFCONNECTED)
623#define UPD_NFDS(x) if (nfds < (x)) nfds = (x)
624
625/*
626 * curtime kept to an accuracy of 1ms
627 */
628#ifdef RT_OS_WINDOWS
629static void updtime(PNATState pData)
630{
631 struct _timeb tb;
632
633 _ftime(&tb);
634 curtime = (u_int)tb.time * (u_int)1000;
635 curtime += (u_int)tb.millitm;
636}
637#else
638static void updtime(PNATState pData)
639{
640 gettimeofday(&tt, 0);
641
642 curtime = (u_int)tt.tv_sec * (u_int)1000;
643 curtime += (u_int)tt.tv_usec / (u_int)1000;
644
645 if ((tt.tv_usec % 1000) >= 500)
646 curtime++;
647}
648#endif
649
650#ifndef VBOX_WITH_SIMPLIFIED_SLIRP_SYNC
651void slirp_select_fill(PNATState pData, int *pnfds,
652 fd_set *readfds, fd_set *writefds, fd_set *xfds)
653#else /* !VBOX_WITH_SIMPLIFIED_SLIRP_SYNC */
654# ifdef RT_OS_WINDOWS
655void slirp_select_fill(PNATState pData, int *pnfds)
656# else /* RT_OS_WINDOWS */
657void slirp_select_fill(PNATState pData, int *pnfds, struct pollfd *polls)
658# endif /* !RT_OS_WINDOWS */
659#endif /* VBOX_WITH_SIMPLIFIED_SLIRP_SYNC */
660{
661 struct socket *so, *so_next;
662 int nfds;
663#if defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) && defined(RT_OS_WINDOWS)
664 int rc;
665 int error;
666#endif
667#if defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) && !defined(RT_OS_WINDOWS)
668 int poll_index = 0;
669#endif
670 int i;
671
672 STAM_PROFILE_START(&pData->StatFill, a);
673
674 nfds = *pnfds;
675
676 /*
677 * First, TCP sockets
678 */
679 do_slowtimo = 0;
680 if (link_up)
681 {
682 /*
683 * *_slowtimo needs calling if there are IP fragments
684 * in the fragment queue, or there are TCP connections active
685 */
686 /* XXX:
687 * triggering of fragment expiration should be the same but use new macroses
688 */
689 do_slowtimo = (tcb.so_next != &tcb);
690 if (!do_slowtimo)
691 {
692 for (i = 0; i < IPREASS_NHASH; i++)
693 {
694 if (!TAILQ_EMPTY(&ipq[i]))
695 {
696 do_slowtimo = 1;
697 break;
698 }
699 }
700 }
701 ICMP_ENGAGE_EVENT(&pData->icmp_socket, readfds);
702
703 STAM_COUNTER_RESET(&pData->StatTCP);
704 STAM_COUNTER_RESET(&pData->StatTCPHot);
705
706 QSOCKET_FOREACH(so, so_next, tcp)
707 /* { */
708#if defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) && !defined(RT_OS_WINDOWS)
709 so->so_poll_index = -1;
710#endif
711 STAM_COUNTER_INC(&pData->StatTCP);
712
713 /*
714 * See if we need a tcp_fasttimo
715 */
716 if ( time_fasttimo == 0
717 && so->so_tcpcb != NULL
718 && so->so_tcpcb->t_flags & TF_DELACK)
719 time_fasttimo = curtime; /* Flag when we want a fasttimo */
720
721 /*
722 * NOFDREF can include still connecting to local-host,
723 * newly socreated() sockets etc. Don't want to select these.
724 */
725 if (so->so_state & SS_NOFDREF || so->s == -1)
726 CONTINUE(tcp);
727
728 /*
729 * Set for reading sockets which are accepting
730 */
731 if (so->so_state & SS_FACCEPTCONN)
732 {
733 STAM_COUNTER_INC(&pData->StatTCPHot);
734 TCP_ENGAGE_EVENT1(so, readfds);
735 CONTINUE(tcp);
736 }
737
738 /*
739 * Set for writing sockets which are connecting
740 */
741 if (so->so_state & SS_ISFCONNECTING)
742 {
743 Log2(("connecting %R[natsock] engaged\n",so));
744 STAM_COUNTER_INC(&pData->StatTCPHot);
745 TCP_ENGAGE_EVENT1(so, writefds);
746 }
747
748 /*
749 * Set for writing if we are connected, can send more, and
750 * we have something to send
751 */
752 if (CONN_CANFSEND(so) && so->so_rcv.sb_cc)
753 {
754 STAM_COUNTER_INC(&pData->StatTCPHot);
755 TCP_ENGAGE_EVENT1(so, writefds);
756 }
757
758 /*
759 * Set for reading (and urgent data) if we are connected, can
760 * receive more, and we have room for it XXX /2 ?
761 */
762 if (CONN_CANFRCV(so) && (so->so_snd.sb_cc < (so->so_snd.sb_datalen/2)))
763 {
764 STAM_COUNTER_INC(&pData->StatTCPHot);
765 TCP_ENGAGE_EVENT2(so, readfds, xfds);
766 }
767 LOOP_LABEL(tcp, so, so_next);
768 }
769
770 /*
771 * UDP sockets
772 */
773 STAM_COUNTER_RESET(&pData->StatUDP);
774 STAM_COUNTER_RESET(&pData->StatUDPHot);
775
776 QSOCKET_FOREACH(so, so_next, udp)
777 /* { */
778
779 STAM_COUNTER_INC(&pData->StatUDP);
780#if defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) && !defined(RT_OS_WINDOWS)
781 so->so_poll_index = -1;
782#endif
783
784 /*
785 * See if it's timed out
786 */
787 if (so->so_expire)
788 {
789 if (so->so_expire <= curtime)
790 {
791#ifdef VBOX_WITH_SLIRP_MT
792 /* we need so_next for continue our cycle*/
793 so_next = so->so_next;
794#endif
795 UDP_DETACH(pData, so, so_next);
796 CONTINUE_NO_UNLOCK(udp);
797 }
798 else
799 do_slowtimo = 1; /* Let socket expire */
800 }
801
802 /*
803 * When UDP packets are received from over the link, they're
804 * sendto()'d straight away, so no need for setting for writing
805 * Limit the number of packets queued by this session to 4.
806 * Note that even though we try and limit this to 4 packets,
807 * the session could have more queued if the packets needed
808 * to be fragmented.
809 *
810 * (XXX <= 4 ?)
811 */
812 if ((so->so_state & SS_ISFCONNECTED) && so->so_queued <= 4)
813 {
814 STAM_COUNTER_INC(&pData->StatUDPHot);
815 UDP_ENGAGE_EVENT(so, readfds);
816 }
817 LOOP_LABEL(udp, so, so_next);
818 }
819
820 }
821
822#if defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC)
823# if defined(RT_OS_WINDOWS)
824 *pnfds = VBOX_EVENT_COUNT;
825# else /* RT_OS_WINDOWS */
826 AssertRelease(poll_index <= *pnfds);
827 *pnfds = poll_index;
828# endif /* !RT_OS_WINDOWS */
829#else /* VBOX_WITH_SIMPLIFIED_SLIRP_SYNC */
830 *pnfds = nfds;
831#endif /* !VBOX_WITH_SIMPLIFIED_SLIRP_SYNC */
832
833 STAM_PROFILE_STOP(&pData->StatFill, a);
834}
835
836#if defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC)
837# if defined(RT_OS_WINDOWS)
838void slirp_select_poll(PNATState pData, int fTimeout, int fIcmp)
839# else /* RT_OS_WINDOWS */
840void slirp_select_poll(PNATState pData, struct pollfd *polls, int ndfs)
841# endif /* !RT_OS_WINDOWS */
842#else /* VBOX_WITH_SIMPLIFIED_SLIRP_SYNC */
843void slirp_select_poll(PNATState pData, fd_set *readfds, fd_set *writefds, fd_set *xfds)
844#endif /* !VBOX_WITH_SIMPLIFIED_SLIRP_SYNC */
845{
846 struct socket *so, *so_next;
847 int ret;
848#if defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) && defined(RT_OS_WINDOWS)
849 WSANETWORKEVENTS NetworkEvents;
850 int rc;
851 int error;
852#endif
853#if defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) && !defined(RT_OS_WINDOWS)
854 int poll_index = 0;
855#endif
856
857 STAM_PROFILE_START(&pData->StatPoll, a);
858
859 /* Update time */
860 updtime(pData);
861
862 /*
863 * See if anything has timed out
864 */
865 if (link_up)
866 {
867 if (time_fasttimo && ((curtime - time_fasttimo) >= 2))
868 {
869 STAM_PROFILE_START(&pData->StatFastTimer, a);
870 tcp_fasttimo(pData);
871 time_fasttimo = 0;
872 STAM_PROFILE_STOP(&pData->StatFastTimer, a);
873 }
874 if (do_slowtimo && ((curtime - last_slowtimo) >= 499))
875 {
876 STAM_PROFILE_START(&pData->StatSlowTimer, a);
877 ip_slowtimo(pData);
878 tcp_slowtimo(pData);
879 last_slowtimo = curtime;
880 STAM_PROFILE_STOP(&pData->StatSlowTimer, a);
881 }
882 }
883#if defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) && defined(RT_OS_WINDOWS)
884 if (fTimeout)
885 return; /* only timer update */
886#endif
887
888 /*
889 * Check sockets
890 */
891 if (link_up)
892 {
893#if defined(RT_OS_WINDOWS)
894 /*XXX: before renaming please make see define
895 * fIcmp in slirp_state.h
896 */
897 if (fIcmp)
898 sorecvfrom(pData, &pData->icmp_socket);
899#else
900 if ( (pData->icmp_socket.s != -1)
901 && CHECK_FD_SET(&pData->icmp_socket, ignored, readfds))
902 sorecvfrom(pData, &pData->icmp_socket);
903#endif
904 /*
905 * Check TCP sockets
906 */
907 QSOCKET_FOREACH(so, so_next, tcp)
908 /* { */
909
910#ifdef VBOX_WITH_SLIRP_MT
911 if ( so->so_state & SS_NOFDREF
912 && so->so_deleted == 1)
913 {
914 struct socket *son, *sop = NULL;
915 QSOCKET_LOCK(tcb);
916 if (so->so_next != NULL)
917 {
918 if (so->so_next != &tcb)
919 SOCKET_LOCK(so->so_next);
920 son = so->so_next;
921 }
922 if ( so->so_prev != &tcb
923 && so->so_prev != NULL)
924 {
925 SOCKET_LOCK(so->so_prev);
926 sop = so->so_prev;
927 }
928 QSOCKET_UNLOCK(tcb);
929 remque(pData, so);
930 NSOCK_DEC();
931 SOCKET_UNLOCK(so);
932 SOCKET_LOCK_DESTROY(so);
933 RTMemFree(so);
934 so_next = son;
935 if (sop != NULL)
936 SOCKET_UNLOCK(sop);
937 CONTINUE_NO_UNLOCK(tcp);
938 }
939#endif
940 /*
941 * FD_ISSET is meaningless on these sockets
942 * (and they can crash the program)
943 */
944 if (so->so_state & SS_NOFDREF || so->s == -1)
945 CONTINUE(tcp);
946
947 POLL_TCP_EVENTS(rc, error, so, &NetworkEvents);
948
949 LOG_NAT_SOCK(so, TCP, &NetworkEvents, readfds, writefds, xfds);
950
951
952 /*
953 * Check for URG data
954 * This will soread as well, so no need to
955 * test for readfds below if this succeeds
956 */
957
958 /* out-of-band data */
959 if (CHECK_FD_SET(so, NetworkEvents, xfds))
960 {
961 sorecvoob(pData, so);
962 }
963
964 /*
965 * Check sockets for reading
966 */
967 else if ( CHECK_FD_SET(so, NetworkEvents, readfds)
968 || WIN_CHECK_FD_SET(so, NetworkEvents, acceptds))
969 {
970 /*
971 * Check for incoming connections
972 */
973 if (so->so_state & SS_FACCEPTCONN)
974 {
975 TCP_CONNECT(pData, so);
976#if defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) && defined(RT_OS_WINDOWS)
977 if (!(NetworkEvents.lNetworkEvents & FD_CLOSE))
978#endif
979 CONTINUE(tcp);
980 }
981
982 ret = soread(pData, so);
983 /* Output it if we read something */
984 if (ret > 0)
985 TCP_OUTPUT(pData, sototcpcb(so));
986 }
987
988#if defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) && defined(RT_OS_WINDOWS)
989 /*
990 * Check for FD_CLOSE events.
991 */
992 if (NetworkEvents.lNetworkEvents & FD_CLOSE)
993 {
994 /*
995 * drain the socket
996 */
997 for (;;)
998 {
999 ret = soread(pData, so);
1000 if (ret > 0)
1001 TCP_OUTPUT(pData, sototcpcb(so));
1002 else
1003 break;
1004 }
1005 }
1006#endif
1007
1008 /*
1009 * Check sockets for writing
1010 */
1011 if (CHECK_FD_SET(so, NetworkEvents, writefds))
1012 {
1013 /*
1014 * Check for non-blocking, still-connecting sockets
1015 */
1016 if (so->so_state & SS_ISFCONNECTING)
1017 {
1018 Log2(("connecting %R[natsock] catched\n", so));
1019 /* Connected */
1020 so->so_state &= ~SS_ISFCONNECTING;
1021
1022 /*
1023 * This should be probably guarded by PROBE_CONN too. Anyway,
1024 * we disable it on OS/2 because the below send call returns
1025 * EFAULT which causes the opened TCP socket to close right
1026 * after it has been opened and connected.
1027 */
1028#ifndef RT_OS_OS2
1029 ret = send(so->s, (const char *)&ret, 0, 0);
1030 if (ret < 0)
1031 {
1032 /* XXXXX Must fix, zero bytes is a NOP */
1033 if ( errno == EAGAIN
1034 || errno == EWOULDBLOCK
1035 || errno == EINPROGRESS
1036 || errno == ENOTCONN)
1037 CONTINUE(tcp);
1038
1039 /* else failed */
1040 so->so_state = SS_NOFDREF;
1041 }
1042 /* else so->so_state &= ~SS_ISFCONNECTING; */
1043#endif
1044
1045 /*
1046 * Continue tcp_input
1047 */
1048 TCP_INPUT(pData, (struct mbuf *)NULL, sizeof(struct ip), so);
1049 /* continue; */
1050 }
1051 else
1052 SOWRITE(ret, pData, so);
1053 /*
1054 * XXX If we wrote something (a lot), there could be the need
1055 * for a window update. In the worst case, the remote will send
1056 * a window probe to get things going again.
1057 */
1058 }
1059
1060 /*
1061 * Probe a still-connecting, non-blocking socket
1062 * to check if it's still alive
1063 */
1064#ifdef PROBE_CONN
1065 if (so->so_state & SS_ISFCONNECTING)
1066 {
1067 ret = recv(so->s, (char *)&ret, 0, 0);
1068
1069 if (ret < 0)
1070 {
1071 /* XXX */
1072 if ( errno == EAGAIN
1073 || errno == EWOULDBLOCK
1074 || errno == EINPROGRESS
1075 || errno == ENOTCONN)
1076 {
1077 CONTINUE(tcp); /* Still connecting, continue */
1078 }
1079
1080 /* else failed */
1081 so->so_state = SS_NOFDREF;
1082
1083 /* tcp_input will take care of it */
1084 }
1085 else
1086 {
1087 ret = send(so->s, &ret, 0, 0);
1088 if (ret < 0)
1089 {
1090 /* XXX */
1091 if ( errno == EAGAIN
1092 || errno == EWOULDBLOCK
1093 || errno == EINPROGRESS
1094 || errno == ENOTCONN)
1095 {
1096 CONTINUE(tcp);
1097 }
1098 /* else failed */
1099 so->so_state = SS_NOFDREF;
1100 }
1101 else
1102 so->so_state &= ~SS_ISFCONNECTING;
1103
1104 }
1105 TCP_INPUT((struct mbuf *)NULL, sizeof(struct ip),so);
1106 } /* SS_ISFCONNECTING */
1107#endif
1108#ifndef RT_OS_WINDOWS
1109 if ( UNIX_CHECK_FD_SET(so, NetworkEvents, rdhup)
1110 || UNIX_CHECK_FD_SET(so, NetworkEvents, rderr))
1111 {
1112 int err;
1113 int inq, outq;
1114 int status;
1115 inq = -1;
1116 socklen_t optlen = sizeof(int);
1117 status = getsockopt(so->s, SOL_SOCKET, SO_ERROR, &err, &optlen);
1118 if (status != 0)
1119 LogRel(("NAT:can't get error status from %R[natsock]\n", so));
1120#ifndef RT_OS_SOLARIS
1121 status = ioctl(so->s, FIONREAD, &inq); /* tcp(7) recommends SIOCINQ which is Linux specific */
1122 if (status != 0 || status != EINVAL)
1123 {
1124 /* EINVAL returned if socket in listen state tcp(7)*/
1125 LogRel(("NAT:can't get depth of IN queue status from %R[natsock]\n", so));
1126 }
1127 status = ioctl(so->s, TIOCOUTQ, &outq); /* SIOCOUTQ see previous comment */
1128 if (status != 0)
1129 LogRel(("NAT:can't get depth of OUT queue from %R[natsock]\n", so));
1130#else
1131 inq = outq = -1;
1132 /*
1133 * Solaris has bit different ioctl commands and its handlings
1134 * hint: streamio(7) I_NREAD
1135 */
1136#endif
1137
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 */
1151 LogRel(("NAT:%R[natsock] err(%d:%s) s(in:%d,out:%d)happens on read I/O, "
1152 "other side close connection \n", so, err, strerror(err), inq, outq));
1153 }
1154 so->so_state = SS_NOFDREF;
1155 TCP_INPUT(pData, (struct mbuf *)NULL, sizeof(struct ip), so);
1156 CONTINUE(tcp);
1157 }
1158 LogRel(("NAT:%R[natsock] we've met(%d:%s) s(in:%d, out:%d) unhandled combination hup (%d) "
1159 "rederr(%d) on (r:%d, w:%d, x:%d)\n",
1160 so, err, strerror(err),
1161 inq, outq,
1162 UNIX_CHECK_FD_SET(so, ign, rdhup),
1163 UNIX_CHECK_FD_SET(so, ign, rderr),
1164 UNIX_CHECK_FD_SET(so, ign, readfds),
1165 UNIX_CHECK_FD_SET(so, ign, writefds),
1166 UNIX_CHECK_FD_SET(so, ign, xfds)));
1167 /*
1168 * Here should be other error handlings
1169 * The error handling code above handles the errors can happens on reading
1170 * we haven't still met any cases of error on write.
1171 */
1172 AssertRelease(!"shouldn't be here!!!");
1173 }
1174#endif
1175 LOOP_LABEL(tcp, so, so_next);
1176 }
1177
1178 /*
1179 * Now UDP sockets.
1180 * Incoming packets are sent straight away, they're not buffered.
1181 * Incoming UDP data isn't buffered either.
1182 */
1183 QSOCKET_FOREACH(so, so_next, udp)
1184 /* { */
1185#ifdef VBOX_WITH_SLIRP_MT
1186 if ( so->so_state & SS_NOFDREF
1187 && so->so_deleted == 1)
1188 {
1189 struct socket *son, *sop = NULL;
1190 QSOCKET_LOCK(udb);
1191 if (so->so_next != NULL)
1192 {
1193 if (so->so_next != &udb)
1194 SOCKET_LOCK(so->so_next);
1195 son = so->so_next;
1196 }
1197 if ( so->so_prev != &udb
1198 && so->so_prev != NULL)
1199 {
1200 SOCKET_LOCK(so->so_prev);
1201 sop = so->so_prev;
1202 }
1203 QSOCKET_UNLOCK(udb);
1204 remque(pData, so);
1205 NSOCK_DEC();
1206 SOCKET_UNLOCK(so);
1207 SOCKET_LOCK_DESTROY(so);
1208 RTMemFree(so);
1209 so_next = son;
1210 if (sop != NULL)
1211 SOCKET_UNLOCK(sop);
1212 CONTINUE_NO_UNLOCK(udp);
1213 }
1214#endif
1215 POLL_UDP_EVENTS(rc, error, so, &NetworkEvents);
1216
1217 LOG_NAT_SOCK(so, UDP, &NetworkEvents, readfds, writefds, xfds);
1218
1219 if (so->s != -1 && CHECK_FD_SET(so, NetworkEvents, readfds))
1220 {
1221 SORECVFROM(pData, so);
1222 }
1223 LOOP_LABEL(udp, so, so_next);
1224 }
1225
1226 }
1227
1228#ifndef VBOX_WITH_SLIRP_MT
1229 /*
1230 * See if we can start outputting
1231 */
1232 if (if_queued && link_up)
1233 if_start(pData);
1234#endif
1235
1236 STAM_PROFILE_STOP(&pData->StatPoll, a);
1237}
1238
1239#define ETH_ALEN 6
1240#define ETH_HLEN 14
1241
1242#define ARPOP_REQUEST 1 /* ARP request */
1243#define ARPOP_REPLY 2 /* ARP reply */
1244
1245struct ethhdr
1246{
1247 unsigned char h_dest[ETH_ALEN]; /* destination eth addr */
1248 unsigned char h_source[ETH_ALEN]; /* source ether addr */
1249 unsigned short h_proto; /* packet type ID field */
1250};
1251
1252struct arphdr
1253{
1254 unsigned short ar_hrd; /* format of hardware address */
1255 unsigned short ar_pro; /* format of protocol address */
1256 unsigned char ar_hln; /* length of hardware address */
1257 unsigned char ar_pln; /* length of protocol address */
1258 unsigned short ar_op; /* ARP opcode (command) */
1259
1260 /*
1261 * Ethernet looks like this : This bit is variable sized however...
1262 */
1263 unsigned char ar_sha[ETH_ALEN]; /* sender hardware address */
1264 unsigned char ar_sip[4]; /* sender IP address */
1265 unsigned char ar_tha[ETH_ALEN]; /* target hardware address */
1266 unsigned char ar_tip[4]; /* target IP address */
1267};
1268
1269static
1270#ifdef VBOX_WITH_SIMPLIFIED_SLIRP_SYNC
1271void arp_input(PNATState pData, struct mbuf *m)
1272#else
1273void arp_input(PNATState pData, const uint8_t *pkt, int pkt_len)
1274#endif
1275{
1276 struct ethhdr *eh;
1277 struct ethhdr *reh;
1278 struct arphdr *ah;
1279 struct arphdr *rah;
1280 int ar_op;
1281 struct ex_list *ex_ptr;
1282 uint32_t htip;
1283#ifndef VBOX_WITH_SIMPLIFIED_SLIRP_SYNC
1284 uint8_t arp_reply[sizeof(struct arphdr) + ETH_HLEN];
1285 eh = (struct ethhdr *)pkt;
1286#else
1287 struct mbuf *mr;
1288 eh = mtod(m, struct ethhdr *);
1289#endif
1290 ah = (struct arphdr *)&eh[1];
1291 htip = ntohl(*(uint32_t*)ah->ar_tip);
1292
1293#ifdef VBOX_WITH_SIMPLIFIED_SLIRP_SYNC
1294 mr = m_get(pData);
1295 mr->m_data += if_maxlinkhdr;
1296 mr->m_len = sizeof(struct arphdr);
1297 rah = mtod(mr, struct arphdr *);
1298#else
1299 reh = (struct ethhdr *)arp_reply;
1300 rah = (struct arphdr *)&reh[1];
1301#endif
1302
1303 ar_op = ntohs(ah->ar_op);
1304 switch(ar_op)
1305 {
1306 case ARPOP_REQUEST:
1307 if ((htip & pData->netmask) == ntohl(special_addr.s_addr))
1308 {
1309 if ( CTL_CHECK(htip,CTL_DNS)
1310 || CTL_CHECK(htip, CTL_ALIAS)
1311 || CTL_CHECK(htip, CTL_TFTP))
1312 goto arp_ok;
1313 for (ex_ptr = exec_list; ex_ptr; ex_ptr = ex_ptr->ex_next)
1314 {
1315 if ((htip & ~pData->netmask) == ex_ptr->ex_addr)
1316 goto arp_ok;
1317 }
1318 return;
1319 arp_ok:
1320
1321#ifndef VBOX_WITH_SIMPLIFIED_SLIRP_SYNC
1322 memcpy(reh->h_dest, eh->h_source, ETH_ALEN);
1323 memcpy(reh->h_source, &special_addr, ETH_ALEN);
1324 reh->h_source[5] = ah->ar_tip[3];
1325 reh->h_proto = htons(ETH_P_ARP);
1326#endif
1327 rah->ar_hrd = htons(1);
1328 rah->ar_pro = htons(ETH_P_IP);
1329 rah->ar_hln = ETH_ALEN;
1330 rah->ar_pln = 4;
1331 rah->ar_op = htons(ARPOP_REPLY);
1332 memcpy(rah->ar_sha, special_ethaddr, ETH_ALEN);
1333
1334 switch (htip & ~pData->netmask)
1335 {
1336 case CTL_DNS:
1337 case CTL_ALIAS:
1338 rah->ar_sha[5] = (uint8_t)(htip & ~pData->netmask);
1339 break;
1340 default:;
1341 }
1342
1343 memcpy(rah->ar_sip, ah->ar_tip, 4);
1344 memcpy(rah->ar_tha, ah->ar_sha, ETH_ALEN);
1345 memcpy(rah->ar_tip, ah->ar_sip, 4);
1346#ifdef VBOX_WITH_SIMPLIFIED_SLIRP_SYNC
1347 if_encap(pData, ETH_P_ARP, mr);
1348 m_free(pData, m);
1349#else
1350 slirp_output(pData->pvUser, arp_reply, sizeof(arp_reply));
1351#endif
1352 }
1353 break;
1354 default:
1355 break;
1356 }
1357}
1358
1359void slirp_input(PNATState pData, const uint8_t *pkt, int pkt_len)
1360{
1361 struct mbuf *m;
1362 int proto;
1363 static bool fWarnedIpv6;
1364
1365 if (pkt_len < ETH_HLEN)
1366 {
1367 LogRel(("NAT: packet having size %d has been ingnored\n", pkt_len));
1368 return;
1369 }
1370
1371 m = m_get(pData);
1372 if (!m)
1373 {
1374 LogRel(("can't allocate new mbuf\n"));
1375 return;
1376 }
1377
1378 /* Note: we add to align the IP header */
1379
1380 if (M_FREEROOM(m) < pkt_len)
1381 m_inc(m, pkt_len);
1382
1383 m->m_len = pkt_len ;
1384 memcpy(m->m_data, pkt, pkt_len);
1385
1386 proto = ntohs(*(uint16_t *)(pkt + 12));
1387 switch(proto)
1388 {
1389 case ETH_P_ARP:
1390#ifdef VBOX_WITH_SIMPLIFIED_SLIRP_SYNC
1391 arp_input(pData, m);
1392#else
1393 arp_input(pData, pkt, pkt_len);
1394 m_free(pData, m);
1395#endif
1396 break;
1397 case ETH_P_IP:
1398 /* Update time. Important if the network is very quiet, as otherwise
1399 * the first outgoing connection gets an incorrect timestamp. */
1400 updtime(pData);
1401 m->m_data += ETH_HLEN;
1402 m->m_len -= ETH_HLEN;
1403 ip_input(pData, m);
1404 break;
1405 case ETH_P_IPV6:
1406 m_free(pData, m);
1407 if (!fWarnedIpv6)
1408 {
1409 LogRel(("NAT: IPv6 not supported\n"));
1410 fWarnedIpv6 = true;
1411 }
1412 break;
1413 default:
1414 LogRel(("NAT: Unsupported protocol %x\n", proto));
1415 m_free(pData, m);
1416 break;
1417 }
1418 RTMemFree((void *)pkt);
1419}
1420
1421/* output the IP packet to the ethernet device */
1422#ifdef VBOX_WITH_SIMPLIFIED_SLIRP_SYNC
1423void if_encap(PNATState pData, uint16_t eth_proto, struct mbuf *m)
1424#else
1425void if_encap(PNATState pData, uint8_t *ip_data, int ip_data_len)
1426#endif
1427{
1428#ifdef VBOX_WITH_SIMPLIFIED_SLIRP_SYNC
1429 struct ethhdr *eh;
1430 uint8_t *buf = RTMemAlloc(1600);
1431 m->m_data -= if_maxlinkhdr;
1432 m->m_len += ETH_HLEN;
1433 eh = mtod(m, struct ethhdr *);
1434#else
1435 uint8_t buf[1600];
1436 struct ethhdr *eh = (struct ethhdr *)buf;
1437
1438 if (ip_data_len + ETH_HLEN > sizeof(buf))
1439 return;
1440
1441 memcpy(buf + sizeof(struct ethhdr), ip_data, ip_data_len);
1442#endif
1443
1444
1445 memcpy(eh->h_dest, client_ethaddr, ETH_ALEN);
1446 memcpy(eh->h_source, special_ethaddr, ETH_ALEN - 1);
1447 /* XXX: not correct */
1448 eh->h_source[5] = CTL_ALIAS;
1449#ifdef VBOX_WITH_SIMPLIFIED_SLIRP_SYNC
1450 eh->h_proto = htons(eth_proto);
1451#if 0
1452 slirp_output(pData->pvUser, m, mtod(m, uint8_t *), m->m_len);
1453#else
1454 memcpy(buf, mtod(m, uint8_t *), m->m_len);
1455 slirp_output(pData->pvUser, NULL, buf, m->m_len);
1456 m_free(pData, m);
1457#endif
1458#else
1459 eh->h_proto = htons(ETH_P_IP);
1460 slirp_output(pData->pvUser, buf, ip_data_len + ETH_HLEN);
1461#endif
1462}
1463
1464int slirp_redir(PNATState pData, int is_udp, int host_port,
1465 struct in_addr guest_addr, int guest_port)
1466{
1467 if (is_udp)
1468 {
1469 if (!udp_listen(pData, htons(host_port), guest_addr.s_addr,
1470 htons(guest_port), 0))
1471 return -1;
1472 }
1473 else
1474 {
1475 if (!solisten(pData, htons(host_port), guest_addr.s_addr,
1476 htons(guest_port), 0))
1477 return -1;
1478 }
1479 return 0;
1480}
1481
1482int slirp_add_exec(PNATState pData, int do_pty, const char *args, int addr_low_byte,
1483 int guest_port)
1484{
1485 return add_exec(&exec_list, do_pty, (char *)args,
1486 addr_low_byte, htons(guest_port));
1487}
1488
1489void slirp_set_ethaddr(PNATState pData, const uint8_t *ethaddr)
1490{
1491 memcpy(client_ethaddr, ethaddr, ETH_ALEN);
1492}
1493
1494#if defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) && defined(RT_OS_WINDOWS)
1495HANDLE *slirp_get_events(PNATState pData)
1496{
1497 return pData->phEvents;
1498}
1499void slirp_register_external_event(PNATState pData, HANDLE hEvent, int index)
1500{
1501 pData->phEvents[index] = hEvent;
1502}
1503#endif
1504
1505unsigned int slirp_get_timeout_ms(PNATState pData)
1506{
1507 if (link_up)
1508 {
1509 if (time_fasttimo)
1510 return 2;
1511 if (do_slowtimo)
1512 return 500; /* see PR_SLOWHZ */
1513 }
1514 return 0;
1515}
1516
1517#ifndef RT_OS_WINDOWS
1518int slirp_get_nsock(PNATState pData)
1519{
1520 return pData->nsock;
1521}
1522#endif
1523
1524/*
1525 * this function called from NAT thread
1526 */
1527void slirp_post_sent(PNATState pData, void *pvArg)
1528{
1529 struct socket *so = 0;
1530 struct tcpcb *tp = 0;
1531 struct mbuf *m = (struct mbuf *)pvArg;
1532 m_free(pData, m);
1533}
1534#ifdef VBOX_WITH_SLIRP_MT
1535void slirp_process_queue(PNATState pData)
1536{
1537 RTReqProcess(pData->pReqQueue, RT_INDEFINITE_WAIT);
1538}
1539void *slirp_get_queue(PNATState pData)
1540{
1541 return pData->pReqQueue;
1542}
1543#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