VirtualBox

source: vbox/trunk/src/VBox/Devices/Network/slirp/slirp.c@ 22494

Last change on this file since 22494 was 22494, checked in by vboxsync, 16 years ago

NAT: some pollish of r51498
  • Property svn:eol-style set to native
File size: 56.9 KB
Line 
1#include "slirp.h"
2#ifdef RT_OS_OS2
3# include <paths.h>
4#endif
5
6#include <VBox/err.h>
7#include <VBox/pdmdrv.h>
8#include <iprt/assert.h>
9#ifndef RT_OS_WINDOWS
10# include <sys/ioctl.h>
11# include <poll.h>
12#else
13# include <Winnls.h>
14# define _WINSOCK2API_
15# include <IPHlpApi.h>
16#endif
17#include <alias.h>
18
19#if !defined(RT_OS_WINDOWS)
20
21# define DO_ENGAGE_EVENT1(so, fdset, label) \
22 do { \
23 if( so->so_poll_index != -1 \
24 && so->s == polls[so->so_poll_index].fd) { \
25 polls[so->so_poll_index].events |= N_(fdset ## _poll); \
26 break; /* out of this loop */ \
27 } \
28 AssertRelease(poll_index < (nfds)); \
29 AssertRelease(poll_index >= 0 && poll_index < (nfds)); \
30 polls[poll_index].fd = (so)->s; \
31 (so)->so_poll_index = poll_index; \
32 polls[poll_index].events = N_(fdset ## _poll); \
33 polls[poll_index].revents = 0; \
34 poll_index++; \
35 } while(0)
36
37
38# define DO_ENGAGE_EVENT2(so, fdset1, fdset2, label) \
39 do { \
40 if( so->so_poll_index != -1 \
41 && so->s == polls[so->so_poll_index].fd) { \
42 polls[so->so_poll_index].events |= \
43 N_(fdset1 ## _poll) | N_(fdset1 ## _poll); \
44 break; /* out of this loop */ \
45 } \
46 AssertRelease(poll_index < (nfds)); \
47 polls[poll_index].fd = (so)->s; \
48 (so)->so_poll_index = poll_index; \
49 polls[poll_index].events = \
50 N_(fdset1 ## _poll) | N_(fdset1 ## _poll); \
51 poll_index++; \
52 } while(0)
53
54# define DO_POLL_EVENTS(rc, error, so, events, label) do {} while (0)
55
56# define DO_CHECK_FD_SET(so, events, fdset) ( ((so)->so_poll_index != -1) \
57 && ((so)->so_poll_index <= ndfs) \
58 && ((so)->s == polls[so->so_poll_index].fd) \
59 && (polls[(so)->so_poll_index].revents & N_(fdset ## _poll)))
60# define DO_UNIX_CHECK_FD_SET(so, events, fdset ) DO_CHECK_FD_SET((so), (events), fdset) /*specific for Unix API */
61# define DO_WIN_CHECK_FD_SET(so, events, fdset ) 0 /* specific for Windows Winsock API */
62
63# ifndef RT_OS_WINDOWS
64
65# ifndef RT_OS_LINUX
66# define readfds_poll (POLLRDNORM)
67# define writefds_poll (POLLWRNORM)
68# define xfds_poll (POLLRDBAND|POLLWRBAND|POLLPRI)
69# else
70# define readfds_poll (POLLIN)
71# define writefds_poll (POLLOUT)
72# define xfds_poll (POLLPRI)
73# endif
74# define rderr_poll (POLLERR)
75# define rdhup_poll (POLLHUP)
76# define nval_poll (POLLNVAL)
77
78# define ICMP_ENGAGE_EVENT(so, fdset) \
79 do { \
80 if (pData->icmp_socket.s != -1) \
81 DO_ENGAGE_EVENT1((so), fdset, ICMP); \
82 } while (0)
83# else /* !RT_OS_WINDOWS */
84# define DO_WIN_CHECK_FD_SET(so, events, fdset ) DO_CHECK_FD_SET((so), (events), fdset)
85# define ICMP_ENGAGE_EVENT(so, fdset) do {} while(0)
86#endif /* RT_OS_WINDOWS */
87
88#else /* defined(RT_OS_WINDOWS) */
89
90/*
91 * On Windows, we will be notified by IcmpSendEcho2() when the response arrives.
92 * So no call to WSAEventSelect necessary.
93 */
94# define ICMP_ENGAGE_EVENT(so, fdset) do {} while(0)
95
96# define DO_ENGAGE_EVENT1(so, fdset1, label) \
97 do { \
98 rc = WSAEventSelect((so)->s, VBOX_SOCKET_EVENT, FD_ALL_EVENTS); \
99 if (rc == SOCKET_ERROR) \
100 { \
101 /* This should not happen */ \
102 error = WSAGetLastError(); \
103 LogRel(("WSAEventSelect (" #label ") error %d (so=%x, socket=%s, event=%x)\n", \
104 error, (so), (so)->s, VBOX_SOCKET_EVENT)); \
105 } \
106 } while(0); \
107 CONTINUE(label)
108
109# define DO_ENGAGE_EVENT2(so, fdset1, fdset2, label) \
110 DO_ENGAGE_EVENT1((so), (fdset1), label)
111
112# define DO_POLL_EVENTS(rc, error, so, events, label) \
113 (rc) = WSAEnumNetworkEvents((so)->s, VBOX_SOCKET_EVENT, (events)); \
114 if ((rc) == SOCKET_ERROR) \
115 { \
116 (error) = WSAGetLastError(); \
117 LogRel(("WSAEnumNetworkEvents " #label " error %d\n", (error))); \
118 CONTINUE(label); \
119 }
120
121# define acceptds_win FD_ACCEPT
122# define acceptds_win_bit FD_ACCEPT_BIT
123
124# define readfds_win FD_READ
125# define readfds_win_bit FD_READ_BIT
126
127# define writefds_win FD_WRITE
128# define writefds_win_bit FD_WRITE_BIT
129
130# define xfds_win FD_OOB
131# define xfds_win_bit FD_OOB_BIT
132
133# define DO_CHECK_FD_SET(so, events, fdset) \
134 (((events).lNetworkEvents & fdset ## _win) && ((events).iErrorCode[fdset ## _win_bit] == 0))
135
136# define DO_WIN_CHECK_FD_SET(so, events, fdset ) DO_CHECK_FD_SET((so), (events), fdset)
137# define DO_UNIX_CHECK_FD_SET(so, events, fdset ) 1 /*specific for Unix API */
138
139#endif /* defined(RT_OS_WINDOWS) */
140
141#define TCP_ENGAGE_EVENT1(so, fdset) \
142 DO_ENGAGE_EVENT1((so), fdset, tcp)
143
144#define TCP_ENGAGE_EVENT2(so, fdset1, fdset2) \
145 DO_ENGAGE_EVENT2((so), fdset1, fdset2, tcp)
146
147#define UDP_ENGAGE_EVENT(so, fdset) \
148 DO_ENGAGE_EVENT1((so), fdset, udp)
149
150#define POLL_TCP_EVENTS(rc, error, so, events) \
151 DO_POLL_EVENTS((rc), (error), (so), (events), tcp)
152
153#define POLL_UDP_EVENTS(rc, error, so, events) \
154 DO_POLL_EVENTS((rc), (error), (so), (events), udp)
155
156#define CHECK_FD_SET(so, events, set) \
157 (DO_CHECK_FD_SET((so), (events), set))
158
159#define WIN_CHECK_FD_SET(so, events, set) \
160 (DO_WIN_CHECK_FD_SET((so), (events), set))
161#define UNIX_CHECK_FD_SET(so, events, set) \
162 (DO_UNIX_CHECK_FD_SET(so, events, set))
163
164/*
165 * Loging macros
166 */
167#if VBOX_WITH_DEBUG_NAT_SOCKETS
168# if defined(RT_OS_WINDOWS)
169# define DO_LOG_NAT_SOCK(so, proto, winevent, r_fdset, w_fdset, x_fdset) \
170 do { \
171 LogRel((" " #proto " %R[natsock] %R[natwinnetevents]\n", (so), (winevent))); \
172 } while (0)
173# else /* RT_OS_WINDOWS */
174# define DO_LOG_NAT_SOCK(so, proto, winevent, r_fdset, w_fdset, x_fdset) \
175 do { \
176 LogRel((" " #proto " %R[natsock] %s %s %s er: %s, %s, %s\n", (so), \
177 CHECK_FD_SET(so, ign ,r_fdset) ? "READ":"", \
178 CHECK_FD_SET(so, ign, w_fdset) ? "WRITE":"", \
179 CHECK_FD_SET(so, ign, x_fdset) ? "OOB":"", \
180 CHECK_FD_SET(so, ign, rderr) ? "RDERR":"", \
181 CHECK_FD_SET(so, ign, rdhup) ? "RDHUP":"", \
182 CHECK_FD_SET(so, ign, nval) ? "RDNVAL":"")); \
183 } while (0)
184# endif /* !RT_OS_WINDOWS */
185#else /* VBOX_WITH_DEBUG_NAT_SOCKETS */
186# define DO_LOG_NAT_SOCK(so, proto, winevent, r_fdset, w_fdset, x_fdset) do {} while (0)
187#endif /* !VBOX_WITH_DEBUG_NAT_SOCKETS */
188
189#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)
190
191static void acivate_port_forwarding(PNATState, struct ethhdr *);
192static uint32_t find_guest_ip(PNATState, uint8_t *);
193
194static const uint8_t special_ethaddr[6] =
195{
196 0x52, 0x54, 0x00, 0x12, 0x35, 0x00
197};
198
199static const uint8_t broadcast_ethaddr[6] =
200{
201 0xff, 0xff, 0xff, 0xff, 0xff, 0xff
202};
203
204const uint8_t zerro_ethaddr[6] =
205{
206 0x0, 0x0, 0x0, 0x0, 0x0, 0x0
207};
208
209#ifdef RT_OS_WINDOWS
210static int get_dns_addr_domain(PNATState pData, bool fVerbose,
211 struct in_addr *pdns_addr,
212 const char **ppszDomain)
213{
214 /* Get amount of memory required for operation */
215 ULONG flags = GAA_FLAG_INCLUDE_PREFIX; /*GAA_FLAG_INCLUDE_ALL_INTERFACES;*/ /* all interfaces registered in NDIS */
216 PIP_ADAPTER_ADDRESSES addresses = NULL;
217 PIP_ADAPTER_ADDRESSES addr = NULL;
218 PIP_ADAPTER_DNS_SERVER_ADDRESS dns = NULL;
219 ULONG size = 0;
220 int wlen = 0;
221 char *suffix;
222 struct dns_entry *da = NULL;
223 struct dns_domain_entry *dd = NULL;
224 ULONG ret = ERROR_SUCCESS;
225
226 /* @todo add SKIPing flags to get only required information */
227
228 ret = pData->pfGetAdaptersAddresses(AF_INET, 0, NULL /* reserved */, addresses, &size);
229 if (ret != ERROR_BUFFER_OVERFLOW)
230 {
231 LogRel(("NAT: error %lu occurred on capacity detection operation\n", ret));
232 return -1;
233 }
234
235 if (size == 0)
236 {
237 LogRel(("NAT: Win socket API returns non capacity\n"));
238 return -1;
239 }
240
241 addresses = RTMemAllocZ(size);
242 if (addresses == NULL)
243 {
244 LogRel(("NAT: No memory available \n"));
245 return -1;
246 }
247
248 ret = pData->pfGetAdaptersAddresses(AF_INET, 0, NULL /* reserved */, addresses, &size);
249 if (ret != ERROR_SUCCESS)
250 {
251 LogRel(("NAT: error %lu occurred on fetching adapters info\n", ret));
252 RTMemFree(addresses);
253 return -1;
254 }
255 addr = addresses;
256 while(addr != NULL)
257 {
258 int found;
259 if (addr->OperStatus != IfOperStatusUp)
260 goto next;
261 dns = addr->FirstDnsServerAddress;
262 while (dns != NULL)
263 {
264 struct sockaddr *saddr = dns->Address.lpSockaddr;
265 if (saddr->sa_family != AF_INET)
266 goto next_dns;
267 /* add dns server to list */
268 da = RTMemAllocZ(sizeof(struct dns_entry));
269 if (da == NULL)
270 {
271 LogRel(("NAT: Can't allocate buffer for DNS entry\n"));
272 RTMemFree(addresses);
273 return VERR_NO_MEMORY;
274 }
275 LogRel(("NAT: adding %R[IP4] to DNS server list\n", &((struct sockaddr_in *)saddr)->sin_addr));
276 if ((((struct sockaddr_in *)saddr)->sin_addr.s_addr & htonl(IN_CLASSA_NET)) == ntohl(INADDR_LOOPBACK & IN_CLASSA_NET)) {
277 da->de_addr.s_addr = htonl(ntohl(special_addr.s_addr) | CTL_ALIAS);
278 }
279 else
280 {
281 da->de_addr.s_addr = ((struct sockaddr_in *)saddr)->sin_addr.s_addr;
282 }
283 TAILQ_INSERT_HEAD(&pData->dns_list_head, da, de_list);
284
285 if (addr->DnsSuffix == NULL)
286 goto next_dns;
287
288 /*uniq*/
289 RTUtf16ToUtf8(addr->DnsSuffix, &suffix);
290
291 if (!suffix || strlen(suffix) == 0) {
292 RTStrFree(suffix);
293 goto next_dns;
294 }
295
296 found = 0;
297 LIST_FOREACH(dd, &pData->dns_domain_list_head, dd_list)
298 {
299 if ( dd->dd_pszDomain != NULL
300 && strcmp(dd->dd_pszDomain, suffix) == 0)
301 {
302 found = 1;
303 RTStrFree(suffix);
304 break;
305 }
306 }
307 if (found == 0)
308 {
309 dd = RTMemAllocZ(sizeof(struct dns_domain_entry));
310 if (dd == NULL)
311 {
312 LogRel(("NAT: not enough memory\n"));
313 RTStrFree(suffix);
314 RTMemFree(addresses);
315 return VERR_NO_MEMORY;
316 }
317 dd->dd_pszDomain = suffix;
318 LogRel(("NAT: adding domain name %s to search list\n", dd->dd_pszDomain));
319 LIST_INSERT_HEAD(&pData->dns_domain_list_head, dd, dd_list);
320 }
321 next_dns:
322 dns = dns->Next;
323 }
324 next:
325 addr = addr->Next;
326 }
327 RTMemFree(addresses);
328 return 0;
329}
330
331#else /* !RT_OS_WINDOWS */
332
333static int get_dns_addr_domain(PNATState pData, bool fVerbose,
334 struct in_addr *pdns_addr,
335 const char **ppszDomain)
336{
337 char buff[512];
338 char buff2[256];
339 FILE *f = NULL;
340 int found = 0;
341 struct in_addr tmp_addr;
342
343#ifdef RT_OS_OS2
344 /* Try various locations. */
345 char *etc = getenv("ETC");
346 if (etc)
347 {
348 snprintf(buff, sizeof(buff), "%s/RESOLV2", etc);
349 f = fopen(buff, "rt");
350 }
351 if (!f)
352 {
353 snprintf(buff, sizeof(buff), "%s/RESOLV2", _PATH_ETC);
354 f = fopen(buff, "rt");
355 }
356 if (!f)
357 {
358 snprintf(buff, sizeof(buff), "%s/resolv.conf", _PATH_ETC);
359 f = fopen(buff, "rt");
360 }
361#else
362#ifndef DEBUG_vvl
363 f = fopen("/etc/resolv.conf", "r");
364#else
365 char *home = getenv("HOME");
366 snprintf(buff, sizeof(buff), "%s/resolv.conf", home);
367 f = fopen(buff, "r");
368 if (f != NULL)
369 {
370 Log(("NAT: DNS we're using %s\n", buff));
371 }
372 else
373 {
374 f = fopen("/etc/resolv.conf", "r");
375 Log(("NAT: DNS we're using %s\n", buff));
376 }
377#endif
378#endif
379 if (!f)
380 return -1;
381
382 if (ppszDomain)
383 *ppszDomain = NULL;
384 Log(("nat: DNS Servers:\n"));
385 while (fgets(buff, 512, f) != NULL)
386 {
387 struct dns_entry *da = NULL;
388 if (sscanf(buff, "nameserver%*[ \t]%256s", buff2) == 1)
389 {
390 if (!inet_aton(buff2, &tmp_addr))
391 continue;
392 /*localhost mask */
393 da = RTMemAllocZ(sizeof (struct dns_entry));
394 if (da == NULL)
395 {
396 LogRel(("can't alloc memory for DNS entry\n"));
397 return -1;
398 }
399 /*check */
400 da->de_addr.s_addr = tmp_addr.s_addr;
401 if ((da->de_addr.s_addr & htonl(IN_CLASSA_NET)) == ntohl(INADDR_LOOPBACK & IN_CLASSA_NET)) {
402 da->de_addr.s_addr = htonl(ntohl(special_addr.s_addr) | CTL_ALIAS);
403 }
404 TAILQ_INSERT_HEAD(&pData->dns_list_head, da, de_list);
405 found++;
406 }
407 if ((!strncmp(buff, "domain", 6) || !strncmp(buff, "search", 6)))
408 {
409 char *tok;
410 char *saveptr;
411 struct dns_domain_entry *dd = NULL;
412 int found = 0;
413 tok = strtok_r(&buff[6], " \t\n", &saveptr);
414 LIST_FOREACH(dd, &pData->dns_domain_list_head, dd_list)
415 {
416 if( tok != NULL
417 && strcmp(tok, dd->dd_pszDomain) == 0)
418 {
419 found = 1;
420 break;
421 }
422 }
423 if (tok != NULL && found == 0) {
424 dd = RTMemAllocZ(sizeof(struct dns_domain_entry));
425 if (dd == NULL)
426 {
427 LogRel(("NAT: not enought memory to add domain list\n"));
428 return VERR_NO_MEMORY;
429 }
430 dd->dd_pszDomain = RTStrDup(tok);
431 LogRel(("NAT: adding domain name %s to search list\n", dd->dd_pszDomain));
432 LIST_INSERT_HEAD(&pData->dns_domain_list_head, dd, dd_list);
433 }
434 }
435 }
436 fclose(f);
437 if (!found)
438 return -1;
439 return 0;
440}
441
442#endif
443
444static int slirp_init_dns_list(PNATState pData)
445{
446 TAILQ_INIT(&pData->dns_list_head);
447 LIST_INIT(&pData->dns_domain_list_head);
448 return get_dns_addr_domain(pData, true, NULL, NULL);
449}
450
451static void slirp_release_dns_list(PNATState pData)
452{
453 struct dns_entry *de = NULL;
454 struct dns_domain_entry *dd = NULL;
455 while(!TAILQ_EMPTY(&pData->dns_list_head)) {
456 de = TAILQ_FIRST(&pData->dns_list_head);
457 TAILQ_REMOVE(&pData->dns_list_head, de, de_list);
458 RTMemFree(de);
459 }
460 while(!LIST_EMPTY(&pData->dns_domain_list_head)) {
461 dd = LIST_FIRST(&pData->dns_domain_list_head);
462 LIST_REMOVE(dd, dd_list);
463 if (dd->dd_pszDomain != NULL)
464 RTStrFree(dd->dd_pszDomain);
465 RTMemFree(dd);
466 }
467}
468
469int get_dns_addr(PNATState pData, struct in_addr *pdns_addr)
470{
471 return get_dns_addr_domain(pData, false, pdns_addr, NULL);
472}
473
474#ifndef VBOX_WITH_NAT_SERVICE
475int slirp_init(PNATState *ppData, const char *pszNetAddr, uint32_t u32Netmask,
476 bool fPassDomain, void *pvUser)
477#else
478int slirp_init(PNATState *ppData, uint32_t u32NetAddr, uint32_t u32Netmask,
479 bool fPassDomain, void *pvUser)
480#endif
481{
482 int fNATfailed = 0;
483 int rc;
484 PNATState pData = RTMemAllocZ(sizeof(NATState));
485 *ppData = pData;
486 if (!pData)
487 return VERR_NO_MEMORY;
488 if (u32Netmask & 0x1f)
489 /* CTL is x.x.x.15, bootp passes up to 16 IPs (15..31) */
490 return VERR_INVALID_PARAMETER;
491 pData->fPassDomain = fPassDomain;
492 pData->pvUser = pvUser;
493 pData->netmask = u32Netmask;
494
495 /* sockets & TCP defaults */
496 pData->socket_rcv = 64 * _1K;
497 pData->socket_snd = 64 * _1K;
498 tcp_sndspace = 64 * _1K;
499 tcp_rcvspace = 64 * _1K;
500
501#ifdef RT_OS_WINDOWS
502 {
503 WSADATA Data;
504 WSAStartup(MAKEWORD(2, 0), &Data);
505 }
506 pData->phEvents[VBOX_SOCKET_EVENT_INDEX] = CreateEvent(NULL, FALSE, FALSE, NULL);
507#endif
508#ifdef VBOX_WITH_SLIRP_MT
509 QSOCKET_LOCK_CREATE(tcb);
510 QSOCKET_LOCK_CREATE(udb);
511 rc = RTReqCreateQueue(&pData->pReqQueue);
512 AssertReleaseRC(rc);
513#endif
514
515 link_up = 1;
516
517 debug_init();
518 if_init(pData);
519 ip_init(pData);
520 icmp_init(pData);
521
522 /* Initialise mbufs *after* setting the MTU */
523 m_init(pData);
524
525#ifndef VBOX_WITH_NAT_SERVICE
526 inet_aton(pszNetAddr, &special_addr);
527#else
528 special_addr.s_addr = u32NetAddr;
529#endif
530 pData->slirp_ethaddr = &special_ethaddr[0];
531 alias_addr.s_addr = special_addr.s_addr | htonl(CTL_ALIAS);
532 /* @todo: add ability to configure this staff */
533
534 /* set default addresses */
535 inet_aton("127.0.0.1", &loopback_addr);
536 if (slirp_init_dns_list(pData) < 0)
537 fNATfailed = 1;
538
539 dnsproxy_init(pData);
540
541 getouraddr(pData);
542 {
543 int flags = 0;
544 struct in_addr proxy_addr;
545 pData->proxy_alias = LibAliasInit(pData, NULL);
546 if (pData->proxy_alias == NULL)
547 {
548 LogRel(("NAT: LibAlias default rule wasn't initialized\n"));
549 AssertMsgFailed(("NAT: LibAlias default rule wasn't initialized\n"));
550 }
551 flags = LibAliasSetMode(pData->proxy_alias, 0, 0);
552#ifndef NO_FW_PUNCH
553 flags |= PKT_ALIAS_PUNCH_FW;
554#endif
555 flags |= PKT_ALIAS_LOG; /* set logging */
556 flags = LibAliasSetMode(pData->proxy_alias, flags, ~0);
557 proxy_addr.s_addr = htonl(ntohl(special_addr.s_addr) | CTL_ALIAS);
558 LibAliasSetAddress(pData->proxy_alias, proxy_addr);
559 ftp_alias_load(pData);
560 nbt_alias_load(pData);
561 }
562 return fNATfailed ? VINF_NAT_DNS : VINF_SUCCESS;
563}
564
565/**
566 * Register statistics.
567 */
568void slirp_register_statistics(PNATState pData, PPDMDRVINS pDrvIns)
569{
570#ifdef VBOX_WITH_STATISTICS
571# define PROFILE_COUNTER(name, dsc) REGISTER_COUNTER(name, pData, STAMTYPE_PROFILE, STAMUNIT_TICKS_PER_CALL, dsc)
572# define COUNTING_COUNTER(name, dsc) REGISTER_COUNTER(name, pData, STAMTYPE_COUNTER, STAMUNIT_COUNT, dsc)
573# include "counters.h"
574# undef COUNTER
575/** @todo register statistics for the variables dumped by:
576 * ipstats(pData); tcpstats(pData); udpstats(pData); icmpstats(pData);
577 * mbufstats(pData); sockstats(pData); */
578#endif /* VBOX_WITH_STATISTICS */
579}
580
581/**
582 * Deregister statistics.
583 */
584void slirp_deregister_statistics(PNATState pData, PPDMDRVINS pDrvIns)
585{
586#ifdef VBOX_WITH_STATISTICS
587# define PROFILE_COUNTER(name, dsc) DEREGISTER_COUNTER(name, pData)
588# define COUNTING_COUNTER(name, dsc) DEREGISTER_COUNTER(name, pData)
589# include "counters.h"
590#endif /* VBOX_WITH_STATISTICS */
591}
592
593/**
594 * Marks the link as up, making it possible to establish new connections.
595 */
596void slirp_link_up(PNATState pData)
597{
598 link_up = 1;
599}
600
601/**
602 * Marks the link as down and cleans up the current connections.
603 */
604void slirp_link_down(PNATState pData)
605{
606 struct socket *so;
607
608 while ((so = tcb.so_next) != &tcb)
609 {
610 if (so->so_state & SS_NOFDREF || so->s == -1)
611 sofree(pData, so);
612 else
613 tcp_drop(pData, sototcpcb(so), 0);
614 }
615
616 while ((so = udb.so_next) != &udb)
617 udp_detach(pData, so);
618
619 link_up = 0;
620}
621
622/**
623 * Terminates the slirp component.
624 */
625void slirp_term(PNATState pData)
626{
627#ifdef RT_OS_WINDOWS
628 pData->pfIcmpCloseHandle(pData->icmp_socket.sh);
629 FreeLibrary(pData->hmIcmpLibrary);
630 RTMemFree(pData->pvIcmpBuffer);
631#else
632 closesocket(pData->icmp_socket.s);
633#endif
634
635 slirp_link_down(pData);
636 slirp_release_dns_list(pData);
637 ftp_alias_unload(pData);
638 nbt_alias_unload(pData);
639 while(!LIST_EMPTY(&instancehead)) {
640 struct libalias *la = LIST_FIRST(&instancehead);
641 /* libalias do all clean up */
642 LibAliasUninit(la);
643 }
644#ifdef RT_OS_WINDOWS
645 WSACleanup();
646#endif
647#ifdef LOG_ENABLED
648 Log(("\n"
649 "NAT statistics\n"
650 "--------------\n"
651 "\n"));
652 ipstats(pData);
653 tcpstats(pData);
654 udpstats(pData);
655 icmpstats(pData);
656 mbufstats(pData);
657 sockstats(pData);
658 Log(("\n"
659 "\n"
660 "\n"));
661#endif
662 RTMemFree(pData);
663}
664
665
666#define CONN_CANFSEND(so) (((so)->so_state & (SS_FCANTSENDMORE|SS_ISFCONNECTED)) == SS_ISFCONNECTED)
667#define CONN_CANFRCV(so) (((so)->so_state & (SS_FCANTRCVMORE|SS_ISFCONNECTED)) == SS_ISFCONNECTED)
668
669/*
670 * curtime kept to an accuracy of 1ms
671 */
672static void updtime(PNATState pData)
673{
674#ifdef RT_OS_WINDOWS
675 struct _timeb tb;
676
677 _ftime(&tb);
678 curtime = (u_int)tb.time * (u_int)1000;
679 curtime += (u_int)tb.millitm;
680#else
681 gettimeofday(&tt, 0);
682
683 curtime = (u_int)tt.tv_sec * (u_int)1000;
684 curtime += (u_int)tt.tv_usec / (u_int)1000;
685
686 if ((tt.tv_usec % 1000) >= 500)
687 curtime++;
688#endif
689}
690
691#ifdef RT_OS_WINDOWS
692void slirp_select_fill(PNATState pData, int *pnfds)
693#else /* RT_OS_WINDOWS */
694void slirp_select_fill(PNATState pData, int *pnfds, struct pollfd *polls)
695#endif /* !RT_OS_WINDOWS */
696{
697 struct socket *so, *so_next;
698 int nfds;
699#if defined(RT_OS_WINDOWS)
700 int rc;
701 int error;
702#else
703 int poll_index = 0;
704#endif
705 int i;
706
707 STAM_PROFILE_START(&pData->StatFill, a);
708
709 nfds = *pnfds;
710
711 /*
712 * First, TCP sockets
713 */
714 do_slowtimo = 0;
715 if (!link_up)
716 goto done;
717 /*
718 * *_slowtimo needs calling if there are IP fragments
719 * in the fragment queue, or there are TCP connections active
720 */
721 /* XXX:
722 * triggering of fragment expiration should be the same but use new macroses
723 */
724 do_slowtimo = (tcb.so_next != &tcb);
725 if (!do_slowtimo)
726 {
727 for (i = 0; i < IPREASS_NHASH; i++)
728 {
729 if (!TAILQ_EMPTY(&ipq[i]))
730 {
731 do_slowtimo = 1;
732 break;
733 }
734 }
735 }
736 ICMP_ENGAGE_EVENT(&pData->icmp_socket, readfds);
737
738 STAM_COUNTER_RESET(&pData->StatTCP);
739 STAM_COUNTER_RESET(&pData->StatTCPHot);
740
741 QSOCKET_FOREACH(so, so_next, tcp)
742 /* { */
743#if !defined(RT_OS_WINDOWS)
744 so->so_poll_index = -1;
745#endif
746 STAM_COUNTER_INC(&pData->StatTCP);
747
748 /*
749 * See if we need a tcp_fasttimo
750 */
751 if ( time_fasttimo == 0
752 && so->so_tcpcb != NULL
753 && so->so_tcpcb->t_flags & TF_DELACK)
754 time_fasttimo = curtime; /* Flag when we want a fasttimo */
755
756 /*
757 * NOFDREF can include still connecting to local-host,
758 * newly socreated() sockets etc. Don't want to select these.
759 */
760 if (so->so_state & SS_NOFDREF || so->s == -1)
761 CONTINUE(tcp);
762
763 /*
764 * Set for reading sockets which are accepting
765 */
766 if (so->so_state & SS_FACCEPTCONN)
767 {
768 STAM_COUNTER_INC(&pData->StatTCPHot);
769 TCP_ENGAGE_EVENT1(so, readfds);
770 CONTINUE(tcp);
771 }
772
773 /*
774 * Set for writing sockets which are connecting
775 */
776 if (so->so_state & SS_ISFCONNECTING)
777 {
778 Log2(("connecting %R[natsock] engaged\n",so));
779 STAM_COUNTER_INC(&pData->StatTCPHot);
780 TCP_ENGAGE_EVENT1(so, writefds);
781 }
782
783 /*
784 * Set for writing if we are connected, can send more, and
785 * we have something to send
786 */
787 if (CONN_CANFSEND(so) && so->so_rcv.sb_cc)
788 {
789 STAM_COUNTER_INC(&pData->StatTCPHot);
790 TCP_ENGAGE_EVENT1(so, writefds);
791 }
792
793 /*
794 * Set for reading (and urgent data) if we are connected, can
795 * receive more, and we have room for it XXX /2 ?
796 */
797 if (CONN_CANFRCV(so) && (so->so_snd.sb_cc < (so->so_snd.sb_datalen/2)))
798 {
799 STAM_COUNTER_INC(&pData->StatTCPHot);
800 TCP_ENGAGE_EVENT2(so, readfds, xfds);
801 }
802 LOOP_LABEL(tcp, so, so_next);
803 }
804
805 /*
806 * UDP sockets
807 */
808 STAM_COUNTER_RESET(&pData->StatUDP);
809 STAM_COUNTER_RESET(&pData->StatUDPHot);
810
811 QSOCKET_FOREACH(so, so_next, udp)
812 /* { */
813
814 STAM_COUNTER_INC(&pData->StatUDP);
815#if !defined(RT_OS_WINDOWS)
816 so->so_poll_index = -1;
817#endif
818
819 /*
820 * See if it's timed out
821 */
822 if (so->so_expire)
823 {
824 if (so->so_expire <= curtime)
825 {
826 Log2(("NAT: %R[natsock] expired\n", so));
827 if (so->so_timeout != NULL)
828 {
829 so->so_timeout(pData, so, so->so_timeout_arg);
830 }
831#ifdef VBOX_WITH_SLIRP_MT
832 /* we need so_next for continue our cycle*/
833 so_next = so->so_next;
834#endif
835 UDP_DETACH(pData, so, so_next);
836 CONTINUE_NO_UNLOCK(udp);
837 }
838 else
839 do_slowtimo = 1; /* Let socket expire */
840 }
841
842 /*
843 * When UDP packets are received from over the link, they're
844 * sendto()'d straight away, so no need for setting for writing
845 * Limit the number of packets queued by this session to 4.
846 * Note that even though we try and limit this to 4 packets,
847 * the session could have more queued if the packets needed
848 * to be fragmented.
849 *
850 * (XXX <= 4 ?)
851 */
852 if ((so->so_state & SS_ISFCONNECTED) && so->so_queued <= 4)
853 {
854 STAM_COUNTER_INC(&pData->StatUDPHot);
855 UDP_ENGAGE_EVENT(so, readfds);
856 }
857 LOOP_LABEL(udp, so, so_next);
858 }
859done:
860
861#if defined(RT_OS_WINDOWS)
862 *pnfds = VBOX_EVENT_COUNT;
863#else /* RT_OS_WINDOWS */
864 AssertRelease(poll_index <= *pnfds);
865 *pnfds = poll_index;
866#endif /* !RT_OS_WINDOWS */
867
868 STAM_PROFILE_STOP(&pData->StatFill, a);
869}
870
871#if defined(RT_OS_WINDOWS)
872void slirp_select_poll(PNATState pData, int fTimeout, int fIcmp)
873#else /* RT_OS_WINDOWS */
874void slirp_select_poll(PNATState pData, struct pollfd *polls, int ndfs)
875#endif /* !RT_OS_WINDOWS */
876{
877 struct socket *so, *so_next;
878 int ret;
879#if defined(RT_OS_WINDOWS)
880 WSANETWORKEVENTS NetworkEvents;
881 int rc;
882 int error;
883#else
884 int poll_index = 0;
885#endif
886
887 STAM_PROFILE_START(&pData->StatPoll, a);
888
889 /* Update time */
890 updtime(pData);
891
892 /*
893 * See if anything has timed out
894 */
895 if (link_up)
896 {
897 if (time_fasttimo && ((curtime - time_fasttimo) >= 2))
898 {
899 STAM_PROFILE_START(&pData->StatFastTimer, a);
900 tcp_fasttimo(pData);
901 time_fasttimo = 0;
902 STAM_PROFILE_STOP(&pData->StatFastTimer, a);
903 }
904 if (do_slowtimo && ((curtime - last_slowtimo) >= 499))
905 {
906 STAM_PROFILE_START(&pData->StatSlowTimer, a);
907 ip_slowtimo(pData);
908 tcp_slowtimo(pData);
909 last_slowtimo = curtime;
910 STAM_PROFILE_STOP(&pData->StatSlowTimer, a);
911 }
912 }
913#if defined(RT_OS_WINDOWS)
914 if (fTimeout)
915 return; /* only timer update */
916#endif
917
918 /*
919 * Check sockets
920 */
921 if (!link_up)
922 goto done;
923#if defined(RT_OS_WINDOWS)
924 /*XXX: before renaming please make see define
925 * fIcmp in slirp_state.h
926 */
927 if (fIcmp)
928 sorecvfrom(pData, &pData->icmp_socket);
929#else
930 if ( (pData->icmp_socket.s != -1)
931 && CHECK_FD_SET(&pData->icmp_socket, ignored, readfds))
932 sorecvfrom(pData, &pData->icmp_socket);
933#endif
934 /*
935 * Check TCP sockets
936 */
937 QSOCKET_FOREACH(so, so_next, tcp)
938 /* { */
939
940#ifdef VBOX_WITH_SLIRP_MT
941 if ( so->so_state & SS_NOFDREF
942 && so->so_deleted == 1)
943 {
944 struct socket *son, *sop = NULL;
945 QSOCKET_LOCK(tcb);
946 if (so->so_next != NULL)
947 {
948 if (so->so_next != &tcb)
949 SOCKET_LOCK(so->so_next);
950 son = so->so_next;
951 }
952 if ( so->so_prev != &tcb
953 && so->so_prev != NULL)
954 {
955 SOCKET_LOCK(so->so_prev);
956 sop = so->so_prev;
957 }
958 QSOCKET_UNLOCK(tcb);
959 remque(pData, so);
960 NSOCK_DEC();
961 SOCKET_UNLOCK(so);
962 SOCKET_LOCK_DESTROY(so);
963 RTMemFree(so);
964 so_next = son;
965 if (sop != NULL)
966 SOCKET_UNLOCK(sop);
967 CONTINUE_NO_UNLOCK(tcp);
968 }
969#endif
970 /*
971 * FD_ISSET is meaningless on these sockets
972 * (and they can crash the program)
973 */
974 if (so->so_state & SS_NOFDREF || so->s == -1)
975 CONTINUE(tcp);
976
977 POLL_TCP_EVENTS(rc, error, so, &NetworkEvents);
978
979 LOG_NAT_SOCK(so, TCP, &NetworkEvents, readfds, writefds, xfds);
980
981
982 /*
983 * Check for URG data
984 * This will soread as well, so no need to
985 * test for readfds below if this succeeds
986 */
987
988 /* out-of-band data */
989 if (CHECK_FD_SET(so, NetworkEvents, xfds))
990 {
991 sorecvoob(pData, so);
992 }
993
994 /*
995 * Check sockets for reading
996 */
997 else if ( CHECK_FD_SET(so, NetworkEvents, readfds)
998 || WIN_CHECK_FD_SET(so, NetworkEvents, acceptds))
999 {
1000 /*
1001 * Check for incoming connections
1002 */
1003 if (so->so_state & SS_FACCEPTCONN)
1004 {
1005 TCP_CONNECT(pData, so);
1006#if defined(RT_OS_WINDOWS)
1007 if (!(NetworkEvents.lNetworkEvents & FD_CLOSE))
1008#endif
1009 CONTINUE(tcp);
1010 }
1011
1012 ret = soread(pData, so);
1013 /* Output it if we read something */
1014 if (RT_LIKELY(ret > 0))
1015 TCP_OUTPUT(pData, sototcpcb(so));
1016 }
1017
1018#if defined(RT_OS_WINDOWS)
1019 /*
1020 * Check for FD_CLOSE events.
1021 * in some cases once FD_CLOSE engaged on socket it could be flashed latter (for some reasons)
1022 */
1023 if ( (NetworkEvents.lNetworkEvents & FD_CLOSE)
1024 || (so->so_close == 1))
1025 {
1026 so->so_close = 1; /* mark it */
1027 /*
1028 * drain the socket
1029 */
1030 for (;;)
1031 {
1032 ret = soread(pData, so);
1033 if (ret > 0)
1034 TCP_OUTPUT(pData, sototcpcb(so));
1035 else
1036 break;
1037 }
1038 CONTINUE(tcp);
1039 }
1040#endif
1041
1042 /*
1043 * Check sockets for writing
1044 */
1045 if (CHECK_FD_SET(so, NetworkEvents, writefds))
1046 {
1047 /*
1048 * Check for non-blocking, still-connecting sockets
1049 */
1050 if (so->so_state & SS_ISFCONNECTING)
1051 {
1052 Log2(("connecting %R[natsock] catched\n", so));
1053 /* Connected */
1054 so->so_state &= ~SS_ISFCONNECTING;
1055
1056 /*
1057 * This should be probably guarded by PROBE_CONN too. Anyway,
1058 * we disable it on OS/2 because the below send call returns
1059 * EFAULT which causes the opened TCP socket to close right
1060 * after it has been opened and connected.
1061 */
1062#ifndef RT_OS_OS2
1063 ret = send(so->s, (const char *)&ret, 0, 0);
1064 if (ret < 0)
1065 {
1066 /* XXXXX Must fix, zero bytes is a NOP */
1067 if ( errno == EAGAIN
1068 || errno == EWOULDBLOCK
1069 || errno == EINPROGRESS
1070 || errno == ENOTCONN)
1071 CONTINUE(tcp);
1072
1073 /* else failed */
1074 so->so_state = SS_NOFDREF;
1075 }
1076 /* else so->so_state &= ~SS_ISFCONNECTING; */
1077#endif
1078
1079 /*
1080 * Continue tcp_input
1081 */
1082 TCP_INPUT(pData, (struct mbuf *)NULL, sizeof(struct ip), so);
1083 /* continue; */
1084 }
1085 else
1086 SOWRITE(ret, pData, so);
1087 /*
1088 * XXX If we wrote something (a lot), there could be the need
1089 * for a window update. In the worst case, the remote will send
1090 * a window probe to get things going again.
1091 */
1092 }
1093
1094 /*
1095 * Probe a still-connecting, non-blocking socket
1096 * to check if it's still alive
1097 */
1098#ifdef PROBE_CONN
1099 if (so->so_state & SS_ISFCONNECTING)
1100 {
1101 ret = recv(so->s, (char *)&ret, 0, 0);
1102
1103 if (ret < 0)
1104 {
1105 /* XXX */
1106 if ( errno == EAGAIN
1107 || errno == EWOULDBLOCK
1108 || errno == EINPROGRESS
1109 || errno == ENOTCONN)
1110 {
1111 CONTINUE(tcp); /* Still connecting, continue */
1112 }
1113
1114 /* else failed */
1115 so->so_state = SS_NOFDREF;
1116
1117 /* tcp_input will take care of it */
1118 }
1119 else
1120 {
1121 ret = send(so->s, &ret, 0, 0);
1122 if (ret < 0)
1123 {
1124 /* XXX */
1125 if ( errno == EAGAIN
1126 || errno == EWOULDBLOCK
1127 || errno == EINPROGRESS
1128 || errno == ENOTCONN)
1129 {
1130 CONTINUE(tcp);
1131 }
1132 /* else failed */
1133 so->so_state = SS_NOFDREF;
1134 }
1135 else
1136 so->so_state &= ~SS_ISFCONNECTING;
1137
1138 }
1139 TCP_INPUT((struct mbuf *)NULL, sizeof(struct ip),so);
1140 } /* SS_ISFCONNECTING */
1141#endif
1142#ifndef RT_OS_WINDOWS
1143 if ( UNIX_CHECK_FD_SET(so, NetworkEvents, rdhup)
1144 || UNIX_CHECK_FD_SET(so, NetworkEvents, rderr))
1145 {
1146 int err;
1147 int inq, outq;
1148 int status;
1149 socklen_t optlen = sizeof(int);
1150 inq = outq = 0;
1151 status = getsockopt(so->s, SOL_SOCKET, SO_ERROR, &err, &optlen);
1152 if (status != 0)
1153 Log(("NAT: can't get error status from %R[natsock]\n", so));
1154#ifndef RT_OS_SOLARIS
1155 status = ioctl(so->s, FIONREAD, &inq); /* tcp(7) recommends SIOCINQ which is Linux specific */
1156 if (status != 0 || status != EINVAL)
1157 {
1158 /* EINVAL returned if socket in listen state tcp(7)*/
1159 Log(("NAT: can't get depth of IN queue status from %R[natsock]\n", so));
1160 }
1161 status = ioctl(so->s, TIOCOUTQ, &outq); /* SIOCOUTQ see previous comment */
1162 if (status != 0)
1163 Log(("NAT: can't get depth of OUT queue from %R[natsock]\n", so));
1164#else
1165 /*
1166 * Solaris has bit different ioctl commands and its handlings
1167 * hint: streamio(7) I_NREAD
1168 */
1169#endif
1170 if ( so->so_state & SS_ISFCONNECTING
1171 || UNIX_CHECK_FD_SET(so, NetworkEvents, readfds))
1172 {
1173 /**
1174 * Check if we need here take care about gracefull connection
1175 * @todo try with proxy server
1176 */
1177 if (UNIX_CHECK_FD_SET(so, NetworkEvents, readfds))
1178 {
1179 /*
1180 * Never meet inq != 0 or outq != 0, anyway let it stay for a while
1181 * in case it happens we'll able to detect it.
1182 * Give TCP/IP stack wait or expire the socket.
1183 */
1184 Log(("NAT: %R[natsock] err(%d:%s) s(in:%d,out:%d)happens on read I/O, "
1185 "other side close connection \n", so, err, strerror(err), inq, outq));
1186 CONTINUE(tcp);
1187 }
1188 goto tcp_input_close;
1189 }
1190 if ( !UNIX_CHECK_FD_SET(so, NetworkEvents, readfds)
1191 && !UNIX_CHECK_FD_SET(so, NetworkEvents, writefds)
1192 && !UNIX_CHECK_FD_SET(so, NetworkEvents, xfds))
1193 {
1194 Log(("NAT: system expires the socket %R[natsock] err(%d:%s) s(in:%d,out:%d) happens on non-I/O. ",
1195 so, err, strerror(err), inq, outq));
1196 goto tcp_input_close;
1197 }
1198 Log(("NAT: %R[natsock] we've met(%d:%s) s(in:%d, out:%d) unhandled combination hup (%d) "
1199 "rederr(%d) on (r:%d, w:%d, x:%d)\n",
1200 so, err, strerror(err),
1201 inq, outq,
1202 UNIX_CHECK_FD_SET(so, ign, rdhup),
1203 UNIX_CHECK_FD_SET(so, ign, rderr),
1204 UNIX_CHECK_FD_SET(so, ign, readfds),
1205 UNIX_CHECK_FD_SET(so, ign, writefds),
1206 UNIX_CHECK_FD_SET(so, ign, xfds)));
1207 /*
1208 * Give OS's TCP/IP stack a chance to resolve an issue or expire the socket.
1209 */
1210 CONTINUE(tcp);
1211tcp_input_close:
1212 so->so_state = SS_NOFDREF; /*cause connection valid tcp connection termination and socket closing */
1213 TCP_INPUT(pData, (struct mbuf *)NULL, sizeof(struct ip), so);
1214 CONTINUE(tcp);
1215 }
1216#endif
1217 LOOP_LABEL(tcp, so, so_next);
1218 }
1219
1220 /*
1221 * Now UDP sockets.
1222 * Incoming packets are sent straight away, they're not buffered.
1223 * Incoming UDP data isn't buffered either.
1224 */
1225 QSOCKET_FOREACH(so, so_next, udp)
1226 /* { */
1227#ifdef VBOX_WITH_SLIRP_MT
1228 if ( so->so_state & SS_NOFDREF
1229 && so->so_deleted == 1)
1230 {
1231 struct socket *son, *sop = NULL;
1232 QSOCKET_LOCK(udb);
1233 if (so->so_next != NULL)
1234 {
1235 if (so->so_next != &udb)
1236 SOCKET_LOCK(so->so_next);
1237 son = so->so_next;
1238 }
1239 if ( so->so_prev != &udb
1240 && so->so_prev != NULL)
1241 {
1242 SOCKET_LOCK(so->so_prev);
1243 sop = so->so_prev;
1244 }
1245 QSOCKET_UNLOCK(udb);
1246 remque(pData, so);
1247 NSOCK_DEC();
1248 SOCKET_UNLOCK(so);
1249 SOCKET_LOCK_DESTROY(so);
1250 RTMemFree(so);
1251 so_next = son;
1252 if (sop != NULL)
1253 SOCKET_UNLOCK(sop);
1254 CONTINUE_NO_UNLOCK(udp);
1255 }
1256#endif
1257 POLL_UDP_EVENTS(rc, error, so, &NetworkEvents);
1258
1259 LOG_NAT_SOCK(so, UDP, &NetworkEvents, readfds, writefds, xfds);
1260
1261 if (so->s != -1 && CHECK_FD_SET(so, NetworkEvents, readfds))
1262 {
1263 SORECVFROM(pData, so);
1264 }
1265 LOOP_LABEL(udp, so, so_next);
1266 }
1267
1268done:
1269#ifndef VBOX_WITH_SLIRP_MT
1270 /*
1271 * See if we can start outputting
1272 */
1273 if (if_queued && link_up)
1274 if_start(pData);
1275#endif
1276
1277 STAM_PROFILE_STOP(&pData->StatPoll, a);
1278}
1279
1280
1281struct arphdr
1282{
1283 unsigned short ar_hrd; /* format of hardware address */
1284 unsigned short ar_pro; /* format of protocol address */
1285 unsigned char ar_hln; /* length of hardware address */
1286 unsigned char ar_pln; /* length of protocol address */
1287 unsigned short ar_op; /* ARP opcode (command) */
1288
1289 /*
1290 * Ethernet looks like this : This bit is variable sized however...
1291 */
1292 unsigned char ar_sha[ETH_ALEN]; /* sender hardware address */
1293 unsigned char ar_sip[4]; /* sender IP address */
1294 unsigned char ar_tha[ETH_ALEN]; /* target hardware address */
1295 unsigned char ar_tip[4]; /* target IP address */
1296};
1297AssertCompileSize(struct arphdr, 28);
1298
1299static void arp_input(PNATState pData, struct mbuf *m)
1300{
1301 struct ethhdr *eh;
1302 struct ethhdr *reh;
1303 struct arphdr *ah;
1304 struct arphdr *rah;
1305 int ar_op;
1306 struct ex_list *ex_ptr;
1307 uint32_t htip;
1308 uint32_t tip;
1309 struct mbuf *mr;
1310 eh = mtod(m, struct ethhdr *);
1311 ah = (struct arphdr *)&eh[1];
1312 htip = ntohl(*(uint32_t*)ah->ar_tip);
1313 tip = *(uint32_t*)ah->ar_tip;
1314
1315 mr = m_get(pData);
1316
1317 reh = mtod(mr, struct ethhdr *);
1318 memcpy(reh->h_source, eh->h_source, ETH_ALEN); /* XXX: if_encap will swap src and dst*/
1319 Log4(("NAT: arp:%R[ether]->%R[ether]\n",
1320 reh->h_source, reh->h_dest));
1321 Log4(("NAT: arp: %R[IP4]\n", &tip));
1322
1323 mr->m_data += if_maxlinkhdr;
1324 mr->m_len = sizeof(struct arphdr);
1325 rah = mtod(mr, struct arphdr *);
1326
1327 ar_op = ntohs(ah->ar_op);
1328 switch(ar_op)
1329 {
1330 case ARPOP_REQUEST:
1331#ifdef VBOX_WITH_NAT_SERVICE
1332 if (tip == special_addr.s_addr) goto arp_ok;
1333#endif
1334 if ((htip & pData->netmask) == ntohl(special_addr.s_addr))
1335 {
1336 if ( CTL_CHECK(htip, CTL_DNS)
1337 || CTL_CHECK(htip, CTL_ALIAS)
1338 || CTL_CHECK(htip, CTL_TFTP))
1339 goto arp_ok;
1340 for (ex_ptr = exec_list; ex_ptr; ex_ptr = ex_ptr->ex_next)
1341 {
1342 if ((htip & ~pData->netmask) == ex_ptr->ex_addr)
1343 {
1344 goto arp_ok;
1345 }
1346 }
1347 return;
1348 arp_ok:
1349 rah->ar_hrd = htons(1);
1350 rah->ar_pro = htons(ETH_P_IP);
1351 rah->ar_hln = ETH_ALEN;
1352 rah->ar_pln = 4;
1353 rah->ar_op = htons(ARPOP_REPLY);
1354 memcpy(rah->ar_sha, special_ethaddr, ETH_ALEN);
1355
1356 switch (htip & ~pData->netmask)
1357 {
1358 case CTL_DNS:
1359 case CTL_ALIAS:
1360 rah->ar_sha[5] = (uint8_t)(htip & ~pData->netmask);
1361 break;
1362 default:;
1363 }
1364
1365 memcpy(rah->ar_sip, ah->ar_tip, 4);
1366 memcpy(rah->ar_tha, ah->ar_sha, ETH_ALEN);
1367 memcpy(rah->ar_tip, ah->ar_sip, 4);
1368 if_encap(pData, ETH_P_ARP, mr);
1369 m_free(pData, m);
1370 }
1371 break;
1372 case ARPOP_REPLY:
1373 /* @todo check if we already have requested address */
1374 /* if no*/
1375 {
1376 BOOTPClient *bc = bc_alloc_client(pData);
1377 bc->addr.s_addr = *(uint32_t *)ah->ar_sip;
1378 memcpy(bc->macaddr, ah->ar_sha, ETH_ALEN);
1379 m_free(pData, m);
1380 }
1381 break;
1382 default:
1383 break;
1384 }
1385}
1386
1387void slirp_input(PNATState pData, const uint8_t *pkt, int pkt_len)
1388{
1389 struct mbuf *m;
1390 int proto;
1391 static bool fWarnedIpv6;
1392 struct ethhdr *eh = (struct ethhdr*)pkt;
1393
1394 Log2(("NAT: slirp_input %d\n", pkt_len));
1395 if (pkt_len < ETH_HLEN)
1396 {
1397 LogRel(("NAT: packet having size %d has been ingnored\n", pkt_len));
1398 return;
1399 }
1400 Log4(("NAT: in:%R[ether]->%R[ether]\n", &eh->h_source, &eh->h_dest));
1401
1402 if (memcmp(eh->h_source, special_ethaddr, ETH_ALEN) == 0)
1403 {
1404 /* @todo vasily: add ether logging routine in debug.c */
1405 Log(("NAT: packet was addressed to other MAC\n"));
1406 RTMemFree((void *)pkt);
1407 return;
1408 }
1409
1410 m = m_get(pData);
1411 if (!m)
1412 {
1413 LogRel(("NAT: can't allocate new mbuf\n"));
1414 return;
1415 }
1416
1417 /* Note: we add to align the IP header */
1418
1419 if (M_FREEROOM(m) < pkt_len)
1420 m_inc(m, pkt_len);
1421
1422 m->m_len = pkt_len ;
1423 memcpy(m->m_data, pkt, pkt_len);
1424
1425 if (pData->port_forwarding_activated == 0)
1426 acivate_port_forwarding(pData, mtod(m, struct ethhdr *));
1427
1428 proto = ntohs(*(uint16_t *)(pkt + 12));
1429 switch(proto)
1430 {
1431 case ETH_P_ARP:
1432 arp_input(pData, m);
1433 break;
1434 case ETH_P_IP:
1435 /* Update time. Important if the network is very quiet, as otherwise
1436 * the first outgoing connection gets an incorrect timestamp. */
1437 updtime(pData);
1438 m_adj(m, ETH_HLEN);
1439 ip_input(pData, m);
1440 break;
1441 case ETH_P_IPV6:
1442 m_free(pData, m);
1443 if (!fWarnedIpv6)
1444 {
1445 LogRel(("NAT: IPv6 not supported\n"));
1446 fWarnedIpv6 = true;
1447 }
1448 break;
1449 default:
1450 Log(("NAT: Unsupported protocol %x\n", proto));
1451 m_free(pData, m);
1452 break;
1453 }
1454 RTMemFree((void *)pkt);
1455}
1456
1457/* output the IP packet to the ethernet device */
1458void if_encap(PNATState pData, uint16_t eth_proto, struct mbuf *m)
1459{
1460 struct ethhdr *eh;
1461 uint8_t *buf = NULL;
1462 STAM_PROFILE_START(&pData->StatIF_encap, a);
1463
1464 m->m_data -= if_maxlinkhdr;
1465 m->m_len += ETH_HLEN;
1466 eh = mtod(m, struct ethhdr *);
1467
1468 if(MBUF_HEAD(m) != m->m_data)
1469 {
1470 LogRel(("NAT: ethernet detects corruption of the packet"));
1471 AssertMsgFailed(("!!Ethernet frame corrupted!!"));
1472 }
1473
1474 if (memcmp(eh->h_source, special_ethaddr, ETH_ALEN) != 0)
1475 {
1476 memcpy(eh->h_dest, eh->h_source, ETH_ALEN);
1477 memcpy(eh->h_source, special_ethaddr, ETH_ALEN);
1478 Assert(memcmp(eh->h_dest, special_ethaddr, ETH_ALEN) != 0);
1479 if (memcmp(eh->h_dest, zerro_ethaddr, ETH_ALEN) == 0)
1480 {
1481 /* don't do anything */
1482 goto done;
1483 }
1484 }
1485 buf = RTMemAlloc(1600);
1486 if (buf == NULL)
1487 {
1488 LogRel(("NAT: Can't alloc memory for outgoing buffer\n"));
1489 goto done;
1490 }
1491 eh->h_proto = htons(eth_proto);
1492 memcpy(buf, mtod(m, uint8_t *), m->m_len);
1493 slirp_output(pData->pvUser, NULL, buf, m->m_len);
1494done:
1495 STAM_PROFILE_STOP(&pData->StatIF_encap, a);
1496 m_free(pData, m);
1497}
1498
1499/**
1500 * Still we're using dhcp server leasing to map ether to IP
1501 * @todo see rt_lookup_in_cache
1502 */
1503static uint32_t find_guest_ip(PNATState pData, uint8_t *eth_addr)
1504{
1505 int i;
1506 if (memcmp(eth_addr, zerro_ethaddr, ETH_ALEN) == 0
1507 || memcmp(eth_addr, broadcast_ethaddr, ETH_ALEN) == 0)
1508 goto done;
1509 for (i = 0; i < NB_ADDR; i++)
1510 {
1511 if ( bootp_clients[i].allocated
1512 && memcmp(bootp_clients[i].macaddr, eth_addr, ETH_ALEN) == 0)
1513 return bootp_clients[i].addr.s_addr;
1514 }
1515done:
1516 return INADDR_ANY;
1517}
1518
1519/**
1520 * We need check if we've activated port forwarding
1521 * for specific machine ... that of course relates to
1522 * service mode
1523 * @todo finish this for service case
1524 */
1525static void acivate_port_forwarding(PNATState pData, struct ethhdr *ethdr)
1526{
1527 struct port_forward_rule *rule = NULL;
1528
1529 pData->port_forwarding_activated = 1;
1530 /* check mac here */
1531 LIST_FOREACH(rule, &pData->port_forward_rule_head, list)
1532 {
1533 struct socket *so;
1534 struct alias_link *link;
1535 struct libalias *lib;
1536 int flags;
1537 struct sockaddr sa;
1538 struct sockaddr_in *psin;
1539 socklen_t socketlen;
1540 struct in_addr alias;
1541 int rc;
1542 uint32_t guest_addr; /* need to understand if we already give address to guest */
1543
1544 if (rule->activated)
1545 continue; /*already activated */
1546#ifdef VBOX_WITH_NAT_SERVICE
1547 if (memcmp(rule->mac_address, ethdr->h_source, ETH_ALEN) != 0)
1548 continue; /*not right mac, @todo: it'd be better do the list port forwarding per mac */
1549 guest_addr = find_guest_ip(pData, ethdr->h_source);
1550#else
1551 if (memcmp(client_ethaddr, ethdr->h_source, ETH_ALEN) != 0)
1552 continue;
1553 guest_addr = find_guest_ip(pData, ethdr->h_source);
1554#endif
1555 if (guest_addr == INADDR_ANY)
1556 {
1557 /* the address wasn't granted */
1558 pData->port_forwarding_activated = 0;
1559 return;
1560 }
1561#if defined(DEBUG_vvl) && !defined(VBOX_WITH_NAT_SERVICE)
1562 Assert(rule->guest_addr.s_addr == guest_addr);
1563#endif
1564
1565 LogRel(("NAT: set redirect %s hp:%d gp:%d\n", (rule->proto == IPPROTO_UDP?"UDP":"TCP"),
1566 rule->host_port, rule->guest_port));
1567 if (rule->proto == IPPROTO_UDP)
1568 {
1569 so = udp_listen(pData, rule->bind_ip.s_addr, htons(rule->host_port), guest_addr,
1570 htons(rule->guest_port), 0);
1571 }
1572 else
1573 {
1574 so = solisten(pData, rule->bind_ip.s_addr, htons(rule->host_port), guest_addr,
1575 htons(rule->guest_port), 0);
1576 }
1577 if (so == NULL)
1578 {
1579 LogRel(("NAT: failed redirect %s hp:%d gp:%d\n", (rule->proto == IPPROTO_UDP?"UDP":"TCP"),
1580 rule->host_port, rule->guest_port));
1581 goto remove_port_forwarding;
1582 }
1583
1584 psin = (struct sockaddr_in *)&sa;
1585 psin->sin_family = AF_INET;
1586 psin->sin_port = 0;
1587 psin->sin_addr.s_addr = INADDR_ANY;
1588 socketlen = sizeof(struct sockaddr);
1589
1590 rc = getsockname(so->s, &sa, &socketlen);
1591 if (rc < 0 || sa.sa_family != AF_INET)
1592 {
1593 LogRel(("NAT: failed redirect %s hp:%d gp:%d\n", (rule->proto == IPPROTO_UDP?"UDP":"TCP"),
1594 rule->host_port, rule->guest_port));
1595 goto remove_port_forwarding;
1596 }
1597
1598 psin = (struct sockaddr_in *)&sa;
1599
1600 lib = LibAliasInit(pData, NULL);
1601 flags = LibAliasSetMode(lib, 0, 0);
1602 flags |= PKT_ALIAS_LOG; /* set logging */
1603 flags |= PKT_ALIAS_REVERSE; /* set logging */
1604 flags = LibAliasSetMode(lib, flags, ~0);
1605
1606 alias.s_addr = htonl(ntohl(guest_addr) | CTL_ALIAS);
1607 link = LibAliasRedirectPort(lib, psin->sin_addr, htons(rule->host_port),
1608 alias, htons(rule->guest_port),
1609 special_addr, -1, /* not very clear for now*/
1610 rule->proto);
1611 if (link == NULL)
1612 {
1613 LogRel(("NAT: failed redirect %s hp:%d gp:%d\n", (rule->proto == IPPROTO_UDP?"UDP":"TCP"),
1614 rule->host_port, rule->guest_port));
1615 goto remove_port_forwarding;
1616 }
1617 so->so_la = lib;
1618 rule->activated = 1;
1619 continue;
1620 remove_port_forwarding:
1621 LIST_REMOVE(rule, list);
1622 RTMemFree(rule);
1623 }
1624}
1625
1626/**
1627 * Changes in 3.1 instead of opening new socket do the following:
1628 * gain more information:
1629 * 1. bind IP
1630 * 2. host port
1631 * 3. guest port
1632 * 4. proto
1633 * 5. guest MAC address
1634 * the guest's MAC address is rather important for service, but we easily
1635 * could get it from VM configuration in DrvNAT or Service, the idea is activating
1636 * corresponding port-forwarding
1637 */
1638int slirp_redir(PNATState pData, int is_udp, struct in_addr host_addr, int host_port,
1639 struct in_addr guest_addr, int guest_port, const uint8_t *ethaddr)
1640{
1641 struct port_forward_rule *rule = NULL;
1642 Assert(memcmp(ethaddr, zerro_ethaddr, ETH_ALEN) == 0);
1643 rule = RTMemAllocZ(sizeof(struct port_forward_rule));
1644 if (rule == NULL)
1645 return 1;
1646 rule->proto = (is_udp ? IPPROTO_UDP : IPPROTO_TCP);
1647 rule->host_port = host_port;
1648 rule->guest_port = guest_port;
1649#ifndef VBOX_WITH_NAT_SERVICE
1650 rule->guest_addr.s_addr = guest_addr.s_addr;
1651#endif
1652 rule->bind_ip.s_addr = host_addr.s_addr;
1653 memcpy(rule->mac_address, ethaddr, ETH_ALEN);
1654 /* @todo add mac address */
1655 LIST_INSERT_HEAD(&pData->port_forward_rule_head, rule, list);
1656 return 0;
1657}
1658
1659int slirp_add_exec(PNATState pData, int do_pty, const char *args, int addr_low_byte,
1660 int guest_port)
1661{
1662 return add_exec(&exec_list, do_pty, (char *)args,
1663 addr_low_byte, htons(guest_port));
1664}
1665
1666void slirp_set_ethaddr(PNATState pData, const uint8_t *ethaddr)
1667{
1668#ifndef VBOX_WITH_NAT_SERVICE
1669 memcpy(client_ethaddr, ethaddr, ETH_ALEN);
1670#endif
1671}
1672
1673#if defined(RT_OS_WINDOWS)
1674HANDLE *slirp_get_events(PNATState pData)
1675{
1676 return pData->phEvents;
1677}
1678void slirp_register_external_event(PNATState pData, HANDLE hEvent, int index)
1679{
1680 pData->phEvents[index] = hEvent;
1681}
1682#endif
1683
1684unsigned int slirp_get_timeout_ms(PNATState pData)
1685{
1686 if (link_up)
1687 {
1688 if (time_fasttimo)
1689 return 2;
1690 if (do_slowtimo)
1691 return 500; /* see PR_SLOWHZ */
1692 }
1693 return 0;
1694}
1695
1696#ifndef RT_OS_WINDOWS
1697int slirp_get_nsock(PNATState pData)
1698{
1699 return pData->nsock;
1700}
1701#endif
1702
1703/*
1704 * this function called from NAT thread
1705 */
1706void slirp_post_sent(PNATState pData, void *pvArg)
1707{
1708 struct socket *so = 0;
1709 struct tcpcb *tp = 0;
1710 struct mbuf *m = (struct mbuf *)pvArg;
1711 m_free(pData, m);
1712}
1713#ifdef VBOX_WITH_SLIRP_MT
1714void slirp_process_queue(PNATState pData)
1715{
1716 RTReqProcess(pData->pReqQueue, RT_INDEFINITE_WAIT);
1717}
1718void *slirp_get_queue(PNATState pData)
1719{
1720 return pData->pReqQueue;
1721}
1722#endif
1723
1724void slirp_set_dhcp_TFTP_prefix(PNATState pData, const char *tftpPrefix)
1725{
1726 Log2(("tftp_prefix:%s\n", tftpPrefix));
1727 tftp_prefix = tftpPrefix;
1728}
1729
1730void slirp_set_dhcp_TFTP_bootfile(PNATState pData, const char *bootFile)
1731{
1732 Log2(("bootFile:%s\n", bootFile));
1733 bootp_filename = bootFile;
1734}
1735
1736void slirp_set_dhcp_next_server(PNATState pData, const char *next_server)
1737{
1738 Log2(("next_server:%s\n", next_server));
1739 if (next_server == NULL)
1740 pData->tftp_server.s_addr = htonl(ntohl(special_addr.s_addr) | CTL_TFTP);
1741 else
1742 inet_aton(next_server, &pData->tftp_server);
1743}
1744
1745int slirp_set_binding_address(PNATState pData, char *addr)
1746{
1747 if (addr == NULL || (inet_aton(addr, &pData->bindIP) == 0))
1748 {
1749 pData->bindIP.s_addr = INADDR_ANY;
1750 return 1;
1751 }
1752 return 0;
1753}
1754
1755void slirp_set_dhcp_dns_proxy(PNATState pData, bool fDNSProxy)
1756{
1757 Log2(("NAT: DNS proxy switched %s\n", (fDNSProxy ? "on" : "off")));
1758 pData->use_dns_proxy = fDNSProxy;
1759}
1760
1761#define CHECK_ARG(name, val, lim_min, lim_max) \
1762do { \
1763 if ((val) < (lim_min) || (val) > (lim_max)) \
1764 { \
1765 LogRel(("NAT: (" #name ":%d) has been ignored, " \
1766 "because out of range (%d, %d)\n", (val), (lim_min), (lim_max))); \
1767 return; \
1768 } \
1769 else \
1770 { \
1771 LogRel(("NAT: (" #name ":%d)\n", (val))); \
1772 } \
1773} while (0)
1774
1775/* don't allow user set less 8kB and more than 1M values */
1776#define _8K_1M_CHECK_ARG(name, val) CHECK_ARG(name, (val), 8, 1024)
1777void slirp_set_rcvbuf(PNATState pData, int kilobytes)
1778{
1779 _8K_1M_CHECK_ARG("SOCKET_RCVBUF", kilobytes);
1780 pData->socket_rcv = kilobytes;
1781}
1782void slirp_set_sndbuf(PNATState pData, int kilobytes)
1783{
1784 _8K_1M_CHECK_ARG("SOCKET_SNDBUF", kilobytes);
1785 pData->socket_snd = kilobytes * _1K;
1786}
1787void slirp_set_tcp_rcvspace(PNATState pData, int kilobytes)
1788{
1789 _8K_1M_CHECK_ARG("TCP_RCVSPACE", kilobytes);
1790 tcp_rcvspace = kilobytes * _1K;
1791}
1792void slirp_set_tcp_sndspace(PNATState pData, int kilobytes)
1793{
1794 _8K_1M_CHECK_ARG("TCP_SNDSPACE", kilobytes);
1795 tcp_sndspace = kilobytes * _1K;
1796}
1797
1798void slirp_arp_who_has(PNATState pData, uint32_t dst)
1799{
1800 struct mbuf *m;
1801 struct ethhdr *ehdr;
1802 struct arphdr *ahdr;
1803
1804 m = m_get(pData);
1805 if (m == NULL)
1806 {
1807 LogRel(("NAT: Can't alloc mbuf for ARP request\n"));
1808 return;
1809 }
1810 ehdr = mtod(m, struct ethhdr *);
1811 memset(ehdr->h_source, 0xff, ETH_ALEN);
1812 ahdr = (struct ahdr *)&ehdr[1];
1813 ahdr->ar_hrd = htons(1);
1814 ahdr->ar_pro = htons(ETH_P_IP);
1815 ahdr->ar_hln = ETH_ALEN;
1816 ahdr->ar_pln = 4;
1817 ahdr->ar_op = htons(ARPOP_REQUEST);
1818 memcpy(ahdr->ar_sha, special_ethaddr, ETH_ALEN);
1819 *(uint32_t *)ahdr->ar_sip = htonl(ntohl(special_addr.s_addr) | CTL_ALIAS);
1820 memset(ahdr->ar_tha, 0xff, ETH_ALEN); /*broadcast*/
1821 *(uint32_t *)ahdr->ar_tip = dst;
1822 m->m_data += if_maxlinkhdr;
1823 m->m_len = sizeof(struct arphdr);
1824 if_encap(pData, ETH_P_ARP, m);
1825 LogRel(("NAT: ARP request sent\n"));
1826}
Note: See TracBrowser for help on using the repository browser.

© 2024 Oracle Support Privacy / Do Not Sell My Info Terms of Use Trademark Policy Automated Access Etiquette