VirtualBox

source: vbox/trunk/src/VBox/Devices/Network/slirp/slirp.c@ 18902

Last change on this file since 18902 was 18902, checked in by vboxsync, 16 years ago

NAT: removed VBOX_WITH_SIMPLIFIED_SLIRP_SYNC as the poller API was removed some days ago
  • Property svn:eol-style set to native
File size: 53.7 KB
Line 
1#include "slirp.h"
2#ifdef RT_OS_OS2
3# include <paths.h>
4#endif
5#ifdef VBOX_WITH_SLIRP_DNS_PROXY
6#include "dnsproxy/dnsproxy.h"
7#endif
8
9#include <VBox/err.h>
10#include <VBox/pdmdrv.h>
11#include <iprt/assert.h>
12#ifndef RT_OS_WINDOWS
13# include <sys/ioctl.h>
14# include <poll.h>
15#else
16# include <Winnls.h>
17# define _WINSOCK2API_
18# include <IPHlpApi.h>
19#endif
20
21#if !defined(RT_OS_WINDOWS)
22
23# define DO_ENGAGE_EVENT1(so, fdset, label) \
24 do { \
25 if( so->so_poll_index != -1 \
26 && so->s == polls[so->so_poll_index].fd) { \
27 polls[so->so_poll_index].events |= N_(fdset ## _poll); \
28 break; /* out of this loop */ \
29 } \
30 AssertRelease(poll_index < (nfds)); \
31 AssertRelease(poll_index >= 0 && poll_index < (nfds)); \
32 polls[poll_index].fd = (so)->s; \
33 (so)->so_poll_index = poll_index; \
34 polls[poll_index].events = N_(fdset ## _poll); \
35 polls[poll_index].revents = 0; \
36 poll_index++; \
37 } while(0)
38
39
40# define DO_ENGAGE_EVENT2(so, fdset1, fdset2, label) \
41 do { \
42 if( so->so_poll_index != -1 \
43 && so->s == polls[so->so_poll_index].fd) { \
44 polls[so->so_poll_index].events |= \
45 N_(fdset1 ## _poll) | N_(fdset1 ## _poll); \
46 break; /* out of this loop */ \
47 } \
48 AssertRelease(poll_index < (nfds)); \
49 polls[poll_index].fd = (so)->s; \
50 (so)->so_poll_index = poll_index; \
51 polls[poll_index].events = \
52 N_(fdset1 ## _poll) | N_(fdset1 ## _poll); \
53 poll_index++; \
54 } while(0)
55
56# define DO_POLL_EVENTS(rc, error, so, events, label) do {} while (0)
57
58# define DO_CHECK_FD_SET(so, events, fdset) ( ((so)->so_poll_index != -1) \
59 && ((so)->so_poll_index <= ndfs) \
60 && ((so)->s == polls[so->so_poll_index].fd) \
61 && (polls[(so)->so_poll_index].revents & N_(fdset ## _poll)))
62# define DO_UNIX_CHECK_FD_SET(so, events, fdset ) DO_CHECK_FD_SET((so), (events), fdset) /*specific for Unix API */
63# define DO_WIN_CHECK_FD_SET(so, events, fdset ) 0 /* specific for Windows Winsock API */
64
65# ifndef RT_OS_WINDOWS
66
67# ifndef RT_OS_LINUX
68# define readfds_poll (POLLRDNORM)
69# define writefds_poll (POLLWRNORM)
70# define xfds_poll (POLLRDBAND|POLLWRBAND|POLLPRI)
71# else
72# define readfds_poll (POLLIN)
73# define writefds_poll (POLLOUT)
74# define xfds_poll (POLLPRI)
75# endif
76# define rderr_poll (POLLERR)
77# define rdhup_poll (POLLHUP)
78# define nval_poll (POLLNVAL)
79
80# define ICMP_ENGAGE_EVENT(so, fdset) \
81 do { \
82 if (pData->icmp_socket.s != -1) \
83 DO_ENGAGE_EVENT1((so), fdset, ICMP); \
84 } while (0)
85# else /* !RT_OS_WINDOWS */
86# define DO_WIN_CHECK_FD_SET(so, events, fdset ) DO_CHECK_FD_SET((so), (events), fdset)
87# define ICMP_ENGAGE_EVENT(so, fdset) do {} while(0)
88#endif /* RT_OS_WINDOWS */
89
90#else /* defined(RT_OS_WINDOWS) */
91
92/*
93 * On Windows, we will be notified by IcmpSendEcho2() when the response arrives.
94 * So no call to WSAEventSelect necessary.
95 */
96# define ICMP_ENGAGE_EVENT(so, fdset) do {} while(0)
97
98# define DO_ENGAGE_EVENT1(so, fdset1, label) \
99 do { \
100 rc = WSAEventSelect((so)->s, VBOX_SOCKET_EVENT, FD_ALL_EVENTS); \
101 if (rc == SOCKET_ERROR) \
102 { \
103 /* This should not happen */ \
104 error = WSAGetLastError(); \
105 LogRel(("WSAEventSelect (" #label ") error %d (so=%x, socket=%s, event=%x)\n", \
106 error, (so), (so)->s, VBOX_SOCKET_EVENT)); \
107 } \
108 } while(0); \
109 CONTINUE(label)
110
111# define DO_ENGAGE_EVENT2(so, fdset1, fdset2, label) \
112 DO_ENGAGE_EVENT1((so), (fdset1), label)
113
114# define DO_POLL_EVENTS(rc, error, so, events, label) \
115 (rc) = WSAEnumNetworkEvents((so)->s, VBOX_SOCKET_EVENT, (events)); \
116 if ((rc) == SOCKET_ERROR) \
117 { \
118 (error) = WSAGetLastError(); \
119 LogRel(("WSAEnumNetworkEvents " #label " error %d\n", (error))); \
120 CONTINUE(label); \
121 }
122
123# define acceptds_win FD_ACCEPT
124# define acceptds_win_bit FD_ACCEPT_BIT
125
126# define readfds_win FD_READ
127# define readfds_win_bit FD_READ_BIT
128
129# define writefds_win FD_WRITE
130# define writefds_win_bit FD_WRITE_BIT
131
132# define xfds_win FD_OOB
133# define xfds_win_bit FD_OOB_BIT
134
135# define DO_CHECK_FD_SET(so, events, fdset) \
136 (((events).lNetworkEvents & fdset ## _win) && ((events).iErrorCode[fdset ## _win_bit] == 0))
137
138# define DO_WIN_CHECK_FD_SET(so, events, fdset ) DO_CHECK_FD_SET((so), (events), fdset)
139# define DO_UNIX_CHECK_FD_SET(so, events, fdset ) 1 /*specific for Unix API */
140
141#endif /* defined(RT_OS_WINDOWS) */
142
143#define TCP_ENGAGE_EVENT1(so, fdset) \
144 DO_ENGAGE_EVENT1((so), fdset, tcp)
145
146#define TCP_ENGAGE_EVENT2(so, fdset1, fdset2) \
147 DO_ENGAGE_EVENT2((so), fdset1, fdset2, tcp)
148
149#define UDP_ENGAGE_EVENT(so, fdset) \
150 DO_ENGAGE_EVENT1((so), fdset, udp)
151
152#define POLL_TCP_EVENTS(rc, error, so, events) \
153 DO_POLL_EVENTS((rc), (error), (so), (events), tcp)
154
155#define POLL_UDP_EVENTS(rc, error, so, events) \
156 DO_POLL_EVENTS((rc), (error), (so), (events), udp)
157
158#define CHECK_FD_SET(so, events, set) \
159 (DO_CHECK_FD_SET((so), (events), set))
160
161#define WIN_CHECK_FD_SET(so, events, set) \
162 (DO_WIN_CHECK_FD_SET((so), (events), set))
163#define UNIX_CHECK_FD_SET(so, events, set) \
164 (DO_UNIX_CHECK_FD_SET(so, events, set))
165
166/*
167 * Loging macros
168 */
169#if VBOX_WITH_DEBUG_NAT_SOCKETS
170# if defined(RT_OS_WINDOWS)
171# define DO_LOG_NAT_SOCK(so, proto, winevent, r_fdset, w_fdset, x_fdset) \
172 do { \
173 LogRel((" " #proto " %R[natsock] %R[natwinnetevents]\n", (so), (winevent))); \
174 } while (0)
175# else /* RT_OS_WINDOWS */
176# define DO_LOG_NAT_SOCK(so, proto, winevent, r_fdset, w_fdset, x_fdset) \
177 do { \
178 LogRel((" " #proto " %R[natsock] %s %s %s er: %s, %s, %s\n", (so), \
179 CHECK_FD_SET(so, ign ,r_fdset) ? "READ":"", \
180 CHECK_FD_SET(so, ign, w_fdset) ? "WRITE":"", \
181 CHECK_FD_SET(so, ign, x_fdset) ? "OOB":"", \
182 CHECK_FD_SET(so, ign, rderr) ? "RDERR":"", \
183 CHECK_FD_SET(so, ign, rdhup) ? "RDHUP":"", \
184 CHECK_FD_SET(so, ign, nval) ? "RDNVAL":"")); \
185 } while (0)
186# endif /* !RT_OS_WINDOWS */
187#else /* VBOX_WITH_DEBUG_NAT_SOCKETS */
188# define DO_LOG_NAT_SOCK(so, proto, winevent, r_fdset, w_fdset, x_fdset) do {} while (0)
189#endif /* !VBOX_WITH_DEBUG_NAT_SOCKETS */
190
191#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)
192
193static const uint8_t special_ethaddr[6] =
194{
195 0x52, 0x54, 0x00, 0x12, 0x35, 0x00
196};
197
198#ifdef RT_OS_WINDOWS
199# ifndef VBOX_WITH_MULTI_DNS
200static int get_dns_addr_domain(PNATState pData, bool fVerbose,
201 struct in_addr *pdns_addr,
202 const char **ppszDomain)
203{
204 int rc = 0;
205 FIXED_INFO *FixedInfo = NULL;
206 ULONG BufLen;
207 DWORD ret;
208 IP_ADDR_STRING *pIPAddr;
209 struct in_addr tmp_addr;
210
211 FixedInfo = (FIXED_INFO *)GlobalAlloc(GPTR, sizeof(FIXED_INFO));
212 BufLen = sizeof(FIXED_INFO);
213
214 /** @todo: this API returns all DNS servers, no matter whether the
215 * corresponding network adapter is disabled or not. Maybe replace
216 * this by GetAdapterAddresses(), which is XP/Vista only though. */
217 if (ERROR_BUFFER_OVERFLOW == GetNetworkParams(FixedInfo, &BufLen))
218 {
219 if (FixedInfo)
220 {
221 GlobalFree(FixedInfo);
222 FixedInfo = NULL;
223 }
224 FixedInfo = GlobalAlloc(GPTR, BufLen);
225 }
226
227 if ((ret = GetNetworkParams(FixedInfo, &BufLen)) != ERROR_SUCCESS)
228 {
229 Log(("GetNetworkParams failed. ret = %08x\n", (u_int)ret ));
230 if (FixedInfo)
231 {
232 GlobalFree(FixedInfo);
233 FixedInfo = NULL;
234 }
235 rc = -1;
236 goto get_dns_prefix;
237 }
238
239 pIPAddr = &(FixedInfo->DnsServerList);
240 inet_aton(pIPAddr->IpAddress.String, &tmp_addr);
241 Log(("nat: DNS Servers:\n"));
242 if (fVerbose || pdns_addr->s_addr != tmp_addr.s_addr)
243 LogRel(("NAT: DNS address: %s\n", pIPAddr->IpAddress.String));
244 *pdns_addr = tmp_addr;
245
246 pIPAddr = FixedInfo -> DnsServerList.Next;
247 while (pIPAddr)
248 {
249 if (fVerbose)
250 LogRel(("NAT: ignored DNS address: %s\n", pIPAddr ->IpAddress.String));
251 pIPAddr = pIPAddr ->Next;
252 }
253 if (FixedInfo)
254 {
255 GlobalFree(FixedInfo);
256 FixedInfo = NULL;
257 }
258
259get_dns_prefix:
260 if (ppszDomain)
261 {
262 OSVERSIONINFO ver;
263 char szDnsDomain[256];
264 DWORD dwSize = sizeof(szDnsDomain);
265
266 *ppszDomain = NULL;
267 GetVersionEx(&ver);
268 if (ver.dwMajorVersion >= 5)
269 {
270 /* GetComputerNameEx exists in Windows versions starting with 2000. */
271 if (GetComputerNameEx(ComputerNameDnsDomain, szDnsDomain, &dwSize))
272 {
273 if (szDnsDomain[0])
274 {
275 /* Just non-empty strings are valid. */
276 *ppszDomain = RTStrDup(szDnsDomain);
277 if (pData->fPassDomain)
278 {
279 if (fVerbose)
280 LogRel(("NAT: passing domain name %s\n", szDnsDomain));
281 }
282 else
283 Log(("nat: ignoring domain %s\n", szDnsDomain));
284 }
285 }
286 else
287 Log(("nat: GetComputerNameEx failed (%d)\n", GetLastError()));
288 }
289 }
290 return rc;
291}
292# else /* !VBOX_WITH_MULTI_DNS */
293static int get_dns_addr_domain(PNATState pData, bool fVerbose,
294 struct in_addr *pdns_addr,
295 const char **ppszDomain)
296{
297 /* Get amount of memory required for operation */
298 ULONG flags = GAA_FLAG_INCLUDE_PREFIX; /*GAA_FLAG_INCLUDE_ALL_INTERFACES;*/ /* all interfaces registered in NDIS */
299 PIP_ADAPTER_ADDRESSES addresses = NULL;
300 PIP_ADAPTER_ADDRESSES addr = NULL;
301 PIP_ADAPTER_DNS_SERVER_ADDRESS dns = NULL;
302 ULONG size = 0;
303 int wlen = 0;
304 char *suffix;
305 struct dns_entry *da = NULL;
306 struct dns_domain_entry *dd = NULL;
307 ULONG ret = ERROR_SUCCESS;
308
309 /* @todo add SKIPing flags to get only required information */
310
311 ret = pData->pfGetAdaptersAddresses(AF_INET, 0, NULL /* reserved */, addresses, &size);
312 if (ret != ERROR_BUFFER_OVERFLOW)
313 {
314 LogRel(("NAT: error %lu occured on capacity detection operation\n", ret));
315 return -1;
316 }
317
318 if (size == 0)
319 {
320 LogRel(("NAT: Win socket API returns non capacity\n"));
321 return -1;
322 }
323
324 addresses = RTMemAllocZ(size);
325 if (addresses == NULL)
326 {
327 LogRel(("NAT: No memory available \n"));
328 return -1;
329 }
330
331 ret = pData->pfGetAdaptersAddresses(AF_INET, 0, NULL /* reserved */, addresses, &size);
332 if (ret != ERROR_SUCCESS)
333 {
334 LogRel(("NAT: error %lu occured on fetching adapters info\n", ret));
335 RTMemFree(addresses);
336 return -1;
337 }
338 addr = addresses;
339 while(addr != NULL)
340 {
341 int found;
342 if (addr->OperStatus != IfOperStatusUp)
343 goto next;
344 dns = addr->FirstDnsServerAddress;
345 while (dns != NULL)
346 {
347 struct sockaddr *saddr = dns->Address.lpSockaddr;
348 if (saddr->sa_family != AF_INET)
349 goto next_dns;
350 /* add dns server to list */
351 da = RTMemAllocZ(sizeof(struct dns_entry));
352 if (da == NULL)
353 {
354 LogRel(("NAT: Can't allocate buffer for DNS entry\n"));
355 RTMemFree(addresses);
356 return VERR_NO_MEMORY;
357 }
358 LogRel(("NAT: adding %R[IP4] to DNS server list\n", &((struct sockaddr_in *)saddr)->sin_addr));
359 if ((((struct sockaddr_in *)saddr)->sin_addr.s_addr & htonl(IN_CLASSA_NET)) == ntohl(INADDR_LOOPBACK & IN_CLASSA_NET)) {
360 da->de_addr.s_addr = htonl(ntohl(special_addr.s_addr) | CTL_ALIAS);
361 }
362 else
363 {
364 da->de_addr.s_addr = ((struct sockaddr_in *)saddr)->sin_addr.s_addr;
365 }
366 LIST_INSERT_HEAD(&pData->dns_list_head, da, de_list);
367
368 if (addr->DnsSuffix == NULL)
369 goto next_dns;
370
371 /*uniq*/
372 RTUtf16ToUtf8(addr->DnsSuffix, &suffix);
373
374 if (!suffix || strlen(suffix) == 0) {
375 RTStrFree(suffix);
376 goto next_dns;
377 }
378
379 found = 0;
380 LIST_FOREACH(dd, &pData->dns_domain_list_head, dd_list)
381 {
382 if ( dd->dd_pszDomain != NULL
383 && strcmp(dd->dd_pszDomain, suffix) == 0)
384 {
385 found = 1;
386 RTStrFree(suffix);
387 break;
388 }
389 }
390 if (found == 0)
391 {
392 dd = RTMemAllocZ(sizeof(struct dns_domain_entry));
393 if (dd == NULL)
394 {
395 LogRel(("NAT: not enough memory\n"));
396 RTStrFree(suffix);
397 RTMemFree(addresses);
398 return VERR_NO_MEMORY;
399 }
400 dd->dd_pszDomain = suffix;
401 LogRel(("NAT: adding domain name %s to search list\n", dd->dd_pszDomain));
402 LIST_INSERT_HEAD(&pData->dns_domain_list_head, dd, dd_list);
403 }
404 next_dns:
405 dns = dns->Next;
406 }
407 next:
408 addr = addr->Next;
409 }
410 RTMemFree(addresses);
411 return 0;
412}
413# endif /* VBOX_WITH_MULTI_DNS */
414
415#else /* !RT_OS_WINDOWS */
416
417static int get_dns_addr_domain(PNATState pData, bool fVerbose,
418 struct in_addr *pdns_addr,
419 const char **ppszDomain)
420{
421 char buff[512];
422 char buff2[256];
423 FILE *f;
424 int found = 0;
425 struct in_addr tmp_addr;
426
427#ifdef RT_OS_OS2
428 /* Try various locations. */
429 char *etc = getenv("ETC");
430 f = NULL;
431 if (etc)
432 {
433 snprintf(buff, sizeof(buff), "%s/RESOLV2", etc);
434 f = fopen(buff, "rt");
435 }
436 if (!f)
437 {
438 snprintf(buff, sizeof(buff), "%s/RESOLV2", _PATH_ETC);
439 f = fopen(buff, "rt");
440 }
441 if (!f)
442 {
443 snprintf(buff, sizeof(buff), "%s/resolv.conf", _PATH_ETC);
444 f = fopen(buff, "rt");
445 }
446#else
447 f = fopen("/etc/resolv.conf", "r");
448#endif
449 if (!f)
450 return -1;
451
452 if (ppszDomain)
453 *ppszDomain = NULL;
454 Log(("nat: DNS Servers:\n"));
455 while (fgets(buff, 512, f) != NULL)
456 {
457#ifdef VBOX_WITH_MULTI_DNS
458 struct dns_entry *da = NULL;
459#endif
460 if (sscanf(buff, "nameserver%*[ \t]%256s", buff2) == 1)
461 {
462 if (!inet_aton(buff2, &tmp_addr))
463 continue;
464#ifndef VBOX_WITH_MULTI_DNS
465 /* If it's the first one, set it to dns_addr */
466 if (!found)
467 {
468 if (fVerbose || pdns_addr->s_addr != tmp_addr.s_addr)
469 LogRel(("NAT: DNS address: %s\n", buff2));
470 *pdns_addr = tmp_addr;
471 }
472 else
473 {
474 if (fVerbose)
475 LogRel(("NAT: ignored DNS address: %s\n", buff2));
476 }
477#else
478 /*localhost mask */
479 da = RTMemAllocZ(sizeof (struct dns_entry));
480 if (da == NULL)
481 {
482 LogRel(("can't alloc memory for DNS entry\n"));
483 return -1;
484 }
485 /*check */
486 da->de_addr.s_addr = tmp_addr.s_addr;
487 if ((da->de_addr.s_addr & htonl(IN_CLASSA_NET)) == ntohl(INADDR_LOOPBACK & IN_CLASSA_NET)) {
488 da->de_addr.s_addr = htonl(ntohl(special_addr.s_addr) | CTL_ALIAS);
489 }
490 LIST_INSERT_HEAD(&pData->dns_list_head, da, de_list);
491#endif
492 found++;
493 }
494#ifndef VBOX_WITH_MULTI_DNS
495 if ( ppszDomain
496 && (!strncmp(buff, "domain", 6) || !strncmp(buff, "search", 6)))
497 {
498 /* Domain name/search list present. Pick first entry */
499 if (*ppszDomain == NULL)
500 {
501 char *tok;
502 char *saveptr;
503 tok = strtok_r(&buff[6], " \t\n", &saveptr);
504 if (tok)
505 {
506 *ppszDomain = RTStrDup(tok);
507 if (pData->fPassDomain)
508 {
509 if (fVerbose)
510 LogRel(("NAT: passing domain name %s\n", tok));
511 }
512 else
513 Log(("nat: ignoring domain %s\n", tok));
514 }
515 }
516 }
517#else
518 if ((!strncmp(buff, "domain", 6) || !strncmp(buff, "search", 6)))
519 {
520 char *tok;
521 char *saveptr;
522 struct dns_domain_entry *dd = NULL;
523 int found = 0;
524 tok = strtok_r(&buff[6], " \t\n", &saveptr);
525 LIST_FOREACH(dd, &pData->dns_domain_list_head, dd_list)
526 {
527 if( tok != NULL
528 && strcmp(tok, dd->dd_pszDomain) == 0)
529 {
530 found = 1;
531 break;
532 }
533 }
534 if (tok != NULL && found == 0) {
535 dd = RTMemAllocZ(sizeof(struct dns_domain_entry));
536 if (dd == NULL)
537 {
538 LogRel(("NAT: not enought memory to add domain list\n"));
539 return VERR_NO_MEMORY;
540 }
541 dd->dd_pszDomain = RTStrDup(tok);
542 LogRel(("NAT: adding domain name %s to search list\n", dd->dd_pszDomain));
543 LIST_INSERT_HEAD(&pData->dns_domain_list_head, dd, dd_list);
544 }
545 }
546#endif
547 }
548 fclose(f);
549 if (!found)
550 return -1;
551 return 0;
552}
553
554#endif
555#ifdef VBOX_WITH_MULTI_DNS
556static int slirp_init_dns_list(PNATState pData)
557{
558 LIST_INIT(&pData->dns_list_head);
559 LIST_INIT(&pData->dns_domain_list_head);
560 return get_dns_addr_domain(pData, true, NULL, NULL);
561}
562
563static void slirp_release_dns_list(PNATState pData)
564{
565 struct dns_entry *de = NULL;
566 struct dns_domain_entry *dd = NULL;
567 while(!LIST_EMPTY(&pData->dns_domain_list_head)) {
568 dd = LIST_FIRST(&pData->dns_domain_list_head);
569 LIST_REMOVE(dd, dd_list);
570 if (dd->dd_pszDomain != NULL)
571 RTStrFree(dd->dd_pszDomain);
572 RTMemFree(dd);
573 }
574 while(!LIST_EMPTY(&pData->dns_domain_list_head)) {
575 dd = LIST_FIRST(&pData->dns_domain_list_head);
576 LIST_REMOVE(dd, dd_list);
577 if (dd->dd_pszDomain != NULL)
578 RTStrFree(dd->dd_pszDomain);
579 RTMemFree(dd);
580 }
581}
582#endif
583
584int get_dns_addr(PNATState pData, struct in_addr *pdns_addr)
585{
586 return get_dns_addr_domain(pData, false, pdns_addr, NULL);
587}
588
589int slirp_init(PNATState *ppData, const char *pszNetAddr, uint32_t u32Netmask,
590 bool fPassDomain, void *pvUser)
591{
592 int fNATfailed = 0;
593 int rc;
594 PNATState pData = RTMemAllocZ(sizeof(NATState));
595 *ppData = pData;
596 if (!pData)
597 return VERR_NO_MEMORY;
598 if (u32Netmask & 0x1f)
599 /* CTL is x.x.x.15, bootp passes up to 16 IPs (15..31) */
600 return VERR_INVALID_PARAMETER;
601 pData->fPassDomain = fPassDomain;
602 pData->pvUser = pvUser;
603 pData->netmask = u32Netmask;
604
605#ifdef RT_OS_WINDOWS
606 {
607 WSADATA Data;
608 WSAStartup(MAKEWORD(2, 0), &Data);
609 }
610 pData->phEvents[VBOX_SOCKET_EVENT_INDEX] = CreateEvent(NULL, FALSE, FALSE, NULL);
611#endif
612#ifdef VBOX_WITH_SLIRP_MT
613 QSOCKET_LOCK_CREATE(tcb);
614 QSOCKET_LOCK_CREATE(udb);
615 rc = RTReqCreateQueue(&pData->pReqQueue);
616 AssertReleaseRC(rc);
617#endif
618
619 link_up = 1;
620
621 debug_init();
622 if_init(pData);
623 ip_init(pData);
624 icmp_init(pData);
625
626 /* Initialise mbufs *after* setting the MTU */
627 m_init(pData);
628
629 inet_aton(pszNetAddr, &special_addr);
630 alias_addr.s_addr = special_addr.s_addr | htonl(CTL_ALIAS);
631 /* @todo: add ability to configure this staff */
632
633 /* set default addresses */
634 inet_aton("127.0.0.1", &loopback_addr);
635#ifndef VBOX_WITH_MULTI_DNS
636 inet_aton("127.0.0.1", &dns_addr);
637
638 if (get_dns_addr_domain(pData, true, &dns_addr, &pData->pszDomain) < 0)
639#else
640 if (slirp_init_dns_list(pData) < 0)
641#endif
642 fNATfailed = 1;
643#ifdef VBOX_WITH_SLIRP_DNS_PROXY
644 dnsproxy_init(pData);
645#endif
646
647 getouraddr(pData);
648 return fNATfailed ? VINF_NAT_DNS : VINF_SUCCESS;
649}
650
651/**
652 * Statistics counters.
653 */
654void slirp_register_timers(PNATState pData, PPDMDRVINS pDrvIns)
655{
656#ifdef VBOX_WITH_STATISTICS
657 PDMDrvHlpSTAMRegisterF(pDrvIns, &pData->StatFill, STAMTYPE_PROFILE, STAMVISIBILITY_ALWAYS,
658 STAMUNIT_TICKS_PER_CALL, "Profiling slirp fills", "/Drivers/NAT%d/Fill", pDrvIns->iInstance);
659 PDMDrvHlpSTAMRegisterF(pDrvIns, &pData->StatPoll, STAMTYPE_PROFILE, STAMVISIBILITY_ALWAYS,
660 STAMUNIT_TICKS_PER_CALL, "Profiling slirp polls", "/Drivers/NAT%d/Poll", pDrvIns->iInstance);
661 PDMDrvHlpSTAMRegisterF(pDrvIns, &pData->StatFastTimer, STAMTYPE_PROFILE, STAMVISIBILITY_ALWAYS,
662 STAMUNIT_TICKS_PER_CALL, "Profiling slirp fast timer", "/Drivers/NAT%d/TimerFast", pDrvIns->iInstance);
663 PDMDrvHlpSTAMRegisterF(pDrvIns, &pData->StatSlowTimer, STAMTYPE_PROFILE, STAMVISIBILITY_ALWAYS,
664 STAMUNIT_TICKS_PER_CALL, "Profiling slirp slow timer", "/Drivers/NAT%d/TimerSlow", pDrvIns->iInstance);
665 PDMDrvHlpSTAMRegisterF(pDrvIns, &pData->StatTCP, STAMTYPE_COUNTER, STAMVISIBILITY_ALWAYS,
666 STAMUNIT_COUNT, "TCP sockets", "/Drivers/NAT%d/SockTCP", pDrvIns->iInstance);
667 PDMDrvHlpSTAMRegisterF(pDrvIns, &pData->StatTCPHot, STAMTYPE_COUNTER, STAMVISIBILITY_ALWAYS,
668 STAMUNIT_COUNT, "TCP sockets active", "/Drivers/NAT%d/SockTCPHot", pDrvIns->iInstance);
669 PDMDrvHlpSTAMRegisterF(pDrvIns, &pData->StatUDP, STAMTYPE_COUNTER, STAMVISIBILITY_ALWAYS,
670 STAMUNIT_COUNT, "UDP sockets", "/Drivers/NAT%d/SockUDP", pDrvIns->iInstance);
671 PDMDrvHlpSTAMRegisterF(pDrvIns, &pData->StatUDPHot, STAMTYPE_COUNTER, STAMVISIBILITY_ALWAYS,
672 STAMUNIT_COUNT, "UDP sockets active", "/Drivers/NAT%d/SockUDPHot", pDrvIns->iInstance);
673#endif /* VBOX_WITH_STATISTICS */
674}
675
676/**
677 * Marks the link as up, making it possible to establish new connections.
678 */
679void slirp_link_up(PNATState pData)
680{
681 link_up = 1;
682}
683
684/**
685 * Marks the link as down and cleans up the current connections.
686 */
687void slirp_link_down(PNATState pData)
688{
689 struct socket *so;
690
691 while ((so = tcb.so_next) != &tcb)
692 {
693 if (so->so_state & SS_NOFDREF || so->s == -1)
694 sofree(pData, so);
695 else
696 tcp_drop(pData, sototcpcb(so), 0);
697 }
698
699 while ((so = udb.so_next) != &udb)
700 udp_detach(pData, so);
701
702 link_up = 0;
703}
704
705/**
706 * Terminates the slirp component.
707 */
708void slirp_term(PNATState pData)
709{
710#ifndef VBOX_WITH_MULTI_DNS
711 if (pData->pszDomain)
712 RTStrFree((char *)(void *)pData->pszDomain);
713#endif
714
715#ifdef RT_OS_WINDOWS
716 pData->pfIcmpCloseHandle(pData->icmp_socket.sh);
717 FreeLibrary(pData->hmIcmpLibrary);
718 RTMemFree(pData->pvIcmpBuffer);
719# else
720 closesocket(pData->icmp_socket.s);
721#endif
722
723 slirp_link_down(pData);
724#ifdef VBOX_WITH_MULTI_DNS
725 slirp_release_dns_list(pData);
726#endif
727#ifdef RT_OS_WINDOWS
728 WSACleanup();
729#endif
730#ifdef LOG_ENABLED
731 Log(("\n"
732 "NAT statistics\n"
733 "--------------\n"
734 "\n"));
735 ipstats(pData);
736 tcpstats(pData);
737 udpstats(pData);
738 icmpstats(pData);
739 mbufstats(pData);
740 sockstats(pData);
741 Log(("\n"
742 "\n"
743 "\n"));
744#endif
745 RTMemFree(pData);
746}
747
748
749#define CONN_CANFSEND(so) (((so)->so_state & (SS_FCANTSENDMORE|SS_ISFCONNECTED)) == SS_ISFCONNECTED)
750#define CONN_CANFRCV(so) (((so)->so_state & (SS_FCANTRCVMORE|SS_ISFCONNECTED)) == SS_ISFCONNECTED)
751#define UPD_NFDS(x) do { if (nfds < (x)) nfds = (x); } while (0)
752
753/*
754 * curtime kept to an accuracy of 1ms
755 */
756static void updtime(PNATState pData)
757{
758#ifdef RT_OS_WINDOWS
759 struct _timeb tb;
760
761 _ftime(&tb);
762 curtime = (u_int)tb.time * (u_int)1000;
763 curtime += (u_int)tb.millitm;
764#else
765 gettimeofday(&tt, 0);
766
767 curtime = (u_int)tt.tv_sec * (u_int)1000;
768 curtime += (u_int)tt.tv_usec / (u_int)1000;
769
770 if ((tt.tv_usec % 1000) >= 500)
771 curtime++;
772#endif
773}
774
775#ifdef RT_OS_WINDOWS
776void slirp_select_fill(PNATState pData, int *pnfds)
777#else /* RT_OS_WINDOWS */
778void slirp_select_fill(PNATState pData, int *pnfds, struct pollfd *polls)
779#endif /* !RT_OS_WINDOWS */
780{
781 struct socket *so, *so_next;
782 int nfds;
783#if defined(RT_OS_WINDOWS)
784 int rc;
785 int error;
786#else
787 int poll_index = 0;
788#endif
789 int i;
790
791 STAM_PROFILE_START(&pData->StatFill, a);
792
793 nfds = *pnfds;
794
795 /*
796 * First, TCP sockets
797 */
798 do_slowtimo = 0;
799 if (link_up)
800 {
801 /*
802 * *_slowtimo needs calling if there are IP fragments
803 * in the fragment queue, or there are TCP connections active
804 */
805 /* XXX:
806 * triggering of fragment expiration should be the same but use new macroses
807 */
808 do_slowtimo = (tcb.so_next != &tcb);
809 if (!do_slowtimo)
810 {
811 for (i = 0; i < IPREASS_NHASH; i++)
812 {
813 if (!TAILQ_EMPTY(&ipq[i]))
814 {
815 do_slowtimo = 1;
816 break;
817 }
818 }
819 }
820 ICMP_ENGAGE_EVENT(&pData->icmp_socket, readfds);
821
822 STAM_COUNTER_RESET(&pData->StatTCP);
823 STAM_COUNTER_RESET(&pData->StatTCPHot);
824
825 QSOCKET_FOREACH(so, so_next, tcp)
826 /* { */
827#if !defined(RT_OS_WINDOWS)
828 so->so_poll_index = -1;
829#endif
830 STAM_COUNTER_INC(&pData->StatTCP);
831
832 /*
833 * See if we need a tcp_fasttimo
834 */
835 if ( time_fasttimo == 0
836 && so->so_tcpcb != NULL
837 && so->so_tcpcb->t_flags & TF_DELACK)
838 time_fasttimo = curtime; /* Flag when we want a fasttimo */
839
840 /*
841 * NOFDREF can include still connecting to local-host,
842 * newly socreated() sockets etc. Don't want to select these.
843 */
844 if (so->so_state & SS_NOFDREF || so->s == -1)
845 CONTINUE(tcp);
846
847 /*
848 * Set for reading sockets which are accepting
849 */
850 if (so->so_state & SS_FACCEPTCONN)
851 {
852 STAM_COUNTER_INC(&pData->StatTCPHot);
853 TCP_ENGAGE_EVENT1(so, readfds);
854 CONTINUE(tcp);
855 }
856
857 /*
858 * Set for writing sockets which are connecting
859 */
860 if (so->so_state & SS_ISFCONNECTING)
861 {
862 Log2(("connecting %R[natsock] engaged\n",so));
863 STAM_COUNTER_INC(&pData->StatTCPHot);
864 TCP_ENGAGE_EVENT1(so, writefds);
865 }
866
867 /*
868 * Set for writing if we are connected, can send more, and
869 * we have something to send
870 */
871 if (CONN_CANFSEND(so) && so->so_rcv.sb_cc)
872 {
873 STAM_COUNTER_INC(&pData->StatTCPHot);
874 TCP_ENGAGE_EVENT1(so, writefds);
875 }
876
877 /*
878 * Set for reading (and urgent data) if we are connected, can
879 * receive more, and we have room for it XXX /2 ?
880 */
881 if (CONN_CANFRCV(so) && (so->so_snd.sb_cc < (so->so_snd.sb_datalen/2)))
882 {
883 STAM_COUNTER_INC(&pData->StatTCPHot);
884 TCP_ENGAGE_EVENT2(so, readfds, xfds);
885 }
886 LOOP_LABEL(tcp, so, so_next);
887 }
888
889 /*
890 * UDP sockets
891 */
892 STAM_COUNTER_RESET(&pData->StatUDP);
893 STAM_COUNTER_RESET(&pData->StatUDPHot);
894
895 QSOCKET_FOREACH(so, so_next, udp)
896 /* { */
897
898 STAM_COUNTER_INC(&pData->StatUDP);
899#if !defined(RT_OS_WINDOWS)
900 so->so_poll_index = -1;
901#endif
902
903 /*
904 * See if it's timed out
905 */
906 if (so->so_expire)
907 {
908 if (so->so_expire <= curtime)
909 {
910#ifdef VBOX_WITH_SLIRP_DNS_PROXY
911 Log2(("NAT: %R[natsock] expired\n", so));
912 if (so->so_timeout != NULL)
913 {
914 so->so_timeout(pData, so, so->so_timeout_arg);
915 }
916#endif
917#ifdef VBOX_WITH_SLIRP_MT
918 /* we need so_next for continue our cycle*/
919 so_next = so->so_next;
920#endif
921 UDP_DETACH(pData, so, so_next);
922 CONTINUE_NO_UNLOCK(udp);
923 }
924 else
925 do_slowtimo = 1; /* Let socket expire */
926 }
927
928 /*
929 * When UDP packets are received from over the link, they're
930 * sendto()'d straight away, so no need for setting for writing
931 * Limit the number of packets queued by this session to 4.
932 * Note that even though we try and limit this to 4 packets,
933 * the session could have more queued if the packets needed
934 * to be fragmented.
935 *
936 * (XXX <= 4 ?)
937 */
938 if ((so->so_state & SS_ISFCONNECTED) && so->so_queued <= 4)
939 {
940 STAM_COUNTER_INC(&pData->StatUDPHot);
941 UDP_ENGAGE_EVENT(so, readfds);
942 }
943 LOOP_LABEL(udp, so, so_next);
944 }
945
946 }
947
948#if defined(RT_OS_WINDOWS)
949 *pnfds = VBOX_EVENT_COUNT;
950#else /* RT_OS_WINDOWS */
951 AssertRelease(poll_index <= *pnfds);
952 *pnfds = poll_index;
953#endif /* !RT_OS_WINDOWS */
954
955 STAM_PROFILE_STOP(&pData->StatFill, a);
956}
957
958#if defined(RT_OS_WINDOWS)
959void slirp_select_poll(PNATState pData, int fTimeout, int fIcmp)
960#else /* RT_OS_WINDOWS */
961void slirp_select_poll(PNATState pData, struct pollfd *polls, int ndfs)
962#endif /* !RT_OS_WINDOWS */
963{
964 struct socket *so, *so_next;
965 int ret;
966#if defined(RT_OS_WINDOWS)
967 WSANETWORKEVENTS NetworkEvents;
968 int rc;
969 int error;
970#else
971 int poll_index = 0;
972#endif
973
974 STAM_PROFILE_START(&pData->StatPoll, a);
975
976 /* Update time */
977 updtime(pData);
978
979 /*
980 * See if anything has timed out
981 */
982 if (link_up)
983 {
984 if (time_fasttimo && ((curtime - time_fasttimo) >= 2))
985 {
986 STAM_PROFILE_START(&pData->StatFastTimer, a);
987 tcp_fasttimo(pData);
988 time_fasttimo = 0;
989 STAM_PROFILE_STOP(&pData->StatFastTimer, a);
990 }
991 if (do_slowtimo && ((curtime - last_slowtimo) >= 499))
992 {
993 STAM_PROFILE_START(&pData->StatSlowTimer, a);
994 ip_slowtimo(pData);
995 tcp_slowtimo(pData);
996 last_slowtimo = curtime;
997 STAM_PROFILE_STOP(&pData->StatSlowTimer, a);
998 }
999 }
1000#if defined(RT_OS_WINDOWS)
1001 if (fTimeout)
1002 return; /* only timer update */
1003#endif
1004
1005 /*
1006 * Check sockets
1007 */
1008 if (link_up)
1009 {
1010#if defined(RT_OS_WINDOWS)
1011 /*XXX: before renaming please make see define
1012 * fIcmp in slirp_state.h
1013 */
1014 if (fIcmp)
1015 sorecvfrom(pData, &pData->icmp_socket);
1016#else
1017 if ( (pData->icmp_socket.s != -1)
1018 && CHECK_FD_SET(&pData->icmp_socket, ignored, readfds))
1019 sorecvfrom(pData, &pData->icmp_socket);
1020#endif
1021 /*
1022 * Check TCP sockets
1023 */
1024 QSOCKET_FOREACH(so, so_next, tcp)
1025 /* { */
1026
1027#ifdef VBOX_WITH_SLIRP_MT
1028 if ( so->so_state & SS_NOFDREF
1029 && so->so_deleted == 1)
1030 {
1031 struct socket *son, *sop = NULL;
1032 QSOCKET_LOCK(tcb);
1033 if (so->so_next != NULL)
1034 {
1035 if (so->so_next != &tcb)
1036 SOCKET_LOCK(so->so_next);
1037 son = so->so_next;
1038 }
1039 if ( so->so_prev != &tcb
1040 && so->so_prev != NULL)
1041 {
1042 SOCKET_LOCK(so->so_prev);
1043 sop = so->so_prev;
1044 }
1045 QSOCKET_UNLOCK(tcb);
1046 remque(pData, so);
1047 NSOCK_DEC();
1048 SOCKET_UNLOCK(so);
1049 SOCKET_LOCK_DESTROY(so);
1050 RTMemFree(so);
1051 so_next = son;
1052 if (sop != NULL)
1053 SOCKET_UNLOCK(sop);
1054 CONTINUE_NO_UNLOCK(tcp);
1055 }
1056#endif
1057 /*
1058 * FD_ISSET is meaningless on these sockets
1059 * (and they can crash the program)
1060 */
1061 if (so->so_state & SS_NOFDREF || so->s == -1)
1062 CONTINUE(tcp);
1063
1064 POLL_TCP_EVENTS(rc, error, so, &NetworkEvents);
1065
1066 LOG_NAT_SOCK(so, TCP, &NetworkEvents, readfds, writefds, xfds);
1067
1068
1069 /*
1070 * Check for URG data
1071 * This will soread as well, so no need to
1072 * test for readfds below if this succeeds
1073 */
1074
1075 /* out-of-band data */
1076 if (CHECK_FD_SET(so, NetworkEvents, xfds))
1077 {
1078 sorecvoob(pData, so);
1079 }
1080
1081 /*
1082 * Check sockets for reading
1083 */
1084 else if ( CHECK_FD_SET(so, NetworkEvents, readfds)
1085 || WIN_CHECK_FD_SET(so, NetworkEvents, acceptds))
1086 {
1087 /*
1088 * Check for incoming connections
1089 */
1090 if (so->so_state & SS_FACCEPTCONN)
1091 {
1092 TCP_CONNECT(pData, so);
1093#if defined(RT_OS_WINDOWS)
1094 if (!(NetworkEvents.lNetworkEvents & FD_CLOSE))
1095#endif
1096 CONTINUE(tcp);
1097 }
1098
1099 ret = soread(pData, so);
1100 /* Output it if we read something */
1101 if (ret > 0)
1102 TCP_OUTPUT(pData, sototcpcb(so));
1103 }
1104
1105#if defined(RT_OS_WINDOWS)
1106 /*
1107 * Check for FD_CLOSE events.
1108 * in some cases once FD_CLOSE engaged on socket it could be flashed latter (for some reasons)
1109 */
1110 if ( (NetworkEvents.lNetworkEvents & FD_CLOSE)
1111 || (so->so_close == 1))
1112 {
1113 so->so_close = 1; /* mark it */
1114 /*
1115 * drain the socket
1116 */
1117 for (;;)
1118 {
1119 ret = soread(pData, so);
1120 if (ret > 0)
1121 TCP_OUTPUT(pData, sototcpcb(so));
1122 else
1123 break;
1124 }
1125 CONTINUE(tcp);
1126 }
1127#endif
1128
1129 /*
1130 * Check sockets for writing
1131 */
1132 if (CHECK_FD_SET(so, NetworkEvents, writefds))
1133 {
1134 /*
1135 * Check for non-blocking, still-connecting sockets
1136 */
1137 if (so->so_state & SS_ISFCONNECTING)
1138 {
1139 Log2(("connecting %R[natsock] catched\n", so));
1140 /* Connected */
1141 so->so_state &= ~SS_ISFCONNECTING;
1142
1143 /*
1144 * This should be probably guarded by PROBE_CONN too. Anyway,
1145 * we disable it on OS/2 because the below send call returns
1146 * EFAULT which causes the opened TCP socket to close right
1147 * after it has been opened and connected.
1148 */
1149#ifndef RT_OS_OS2
1150 ret = send(so->s, (const char *)&ret, 0, 0);
1151 if (ret < 0)
1152 {
1153 /* XXXXX Must fix, zero bytes is a NOP */
1154 if ( errno == EAGAIN
1155 || errno == EWOULDBLOCK
1156 || errno == EINPROGRESS
1157 || errno == ENOTCONN)
1158 CONTINUE(tcp);
1159
1160 /* else failed */
1161 so->so_state = SS_NOFDREF;
1162 }
1163 /* else so->so_state &= ~SS_ISFCONNECTING; */
1164#endif
1165
1166 /*
1167 * Continue tcp_input
1168 */
1169 TCP_INPUT(pData, (struct mbuf *)NULL, sizeof(struct ip), so);
1170 /* continue; */
1171 }
1172 else
1173 SOWRITE(ret, pData, so);
1174 /*
1175 * XXX If we wrote something (a lot), there could be the need
1176 * for a window update. In the worst case, the remote will send
1177 * a window probe to get things going again.
1178 */
1179 }
1180
1181 /*
1182 * Probe a still-connecting, non-blocking socket
1183 * to check if it's still alive
1184 */
1185#ifdef PROBE_CONN
1186 if (so->so_state & SS_ISFCONNECTING)
1187 {
1188 ret = recv(so->s, (char *)&ret, 0, 0);
1189
1190 if (ret < 0)
1191 {
1192 /* XXX */
1193 if ( errno == EAGAIN
1194 || errno == EWOULDBLOCK
1195 || errno == EINPROGRESS
1196 || errno == ENOTCONN)
1197 {
1198 CONTINUE(tcp); /* Still connecting, continue */
1199 }
1200
1201 /* else failed */
1202 so->so_state = SS_NOFDREF;
1203
1204 /* tcp_input will take care of it */
1205 }
1206 else
1207 {
1208 ret = send(so->s, &ret, 0, 0);
1209 if (ret < 0)
1210 {
1211 /* XXX */
1212 if ( errno == EAGAIN
1213 || errno == EWOULDBLOCK
1214 || errno == EINPROGRESS
1215 || errno == ENOTCONN)
1216 {
1217 CONTINUE(tcp);
1218 }
1219 /* else failed */
1220 so->so_state = SS_NOFDREF;
1221 }
1222 else
1223 so->so_state &= ~SS_ISFCONNECTING;
1224
1225 }
1226 TCP_INPUT((struct mbuf *)NULL, sizeof(struct ip),so);
1227 } /* SS_ISFCONNECTING */
1228#endif
1229#ifndef RT_OS_WINDOWS
1230 if ( UNIX_CHECK_FD_SET(so, NetworkEvents, rdhup)
1231 || UNIX_CHECK_FD_SET(so, NetworkEvents, rderr))
1232 {
1233 int err;
1234 int inq, outq;
1235 int status;
1236 socklen_t optlen = sizeof(int);
1237 inq = outq = 0;
1238 status = getsockopt(so->s, SOL_SOCKET, SO_ERROR, &err, &optlen);
1239 if (status != 0)
1240 Log(("NAT: can't get error status from %R[natsock]\n", so));
1241#ifndef RT_OS_SOLARIS
1242 status = ioctl(so->s, FIONREAD, &inq); /* tcp(7) recommends SIOCINQ which is Linux specific */
1243 if (status != 0 || status != EINVAL)
1244 {
1245 /* EINVAL returned if socket in listen state tcp(7)*/
1246 Log(("NAT: can't get depth of IN queue status from %R[natsock]\n", so));
1247 }
1248 status = ioctl(so->s, TIOCOUTQ, &outq); /* SIOCOUTQ see previous comment */
1249 if (status != 0)
1250 Log(("NAT: can't get depth of OUT queue from %R[natsock]\n", so));
1251#else
1252 /*
1253 * Solaris has bit different ioctl commands and its handlings
1254 * hint: streamio(7) I_NREAD
1255 */
1256#endif
1257 if ( so->so_state & SS_ISFCONNECTING
1258 || UNIX_CHECK_FD_SET(so, NetworkEvents, readfds))
1259 {
1260 /**
1261 * Check if we need here take care about gracefull connection
1262 * @todo try with proxy server
1263 */
1264 if (UNIX_CHECK_FD_SET(so, NetworkEvents, readfds))
1265 {
1266 /*
1267 * Never meet inq != 0 or outq != 0, anyway let it stay for a while
1268 * in case it happens we'll able to detect it.
1269 * Give TCP/IP stack wait or expire the socket.
1270 */
1271 Log(("NAT: %R[natsock] err(%d:%s) s(in:%d,out:%d)happens on read I/O, "
1272 "other side close connection \n", so, err, strerror(err), inq, outq));
1273 CONTINUE(tcp);
1274 }
1275 goto tcp_input_close;
1276 }
1277 if ( !UNIX_CHECK_FD_SET(so, NetworkEvents, readfds)
1278 && !UNIX_CHECK_FD_SET(so, NetworkEvents, writefds)
1279 && !UNIX_CHECK_FD_SET(so, NetworkEvents, xfds))
1280 {
1281 Log(("NAT: system expires the socket %R[natsock] err(%d:%s) s(in:%d,out:%d) happens on non-I/O. ",
1282 so, err, strerror(err), inq, outq));
1283 goto tcp_input_close;
1284 }
1285 Log(("NAT: %R[natsock] we've met(%d:%s) s(in:%d, out:%d) unhandled combination hup (%d) "
1286 "rederr(%d) on (r:%d, w:%d, x:%d)\n",
1287 so, err, strerror(err),
1288 inq, outq,
1289 UNIX_CHECK_FD_SET(so, ign, rdhup),
1290 UNIX_CHECK_FD_SET(so, ign, rderr),
1291 UNIX_CHECK_FD_SET(so, ign, readfds),
1292 UNIX_CHECK_FD_SET(so, ign, writefds),
1293 UNIX_CHECK_FD_SET(so, ign, xfds)));
1294 /*
1295 * Give OS's TCP/IP stack a chance to resolve an issue or expire the socket.
1296 */
1297 CONTINUE(tcp);
1298tcp_input_close:
1299 so->so_state = SS_NOFDREF; /*cause connection valid tcp connection termination and socket closing */
1300 TCP_INPUT(pData, (struct mbuf *)NULL, sizeof(struct ip), so);
1301 CONTINUE(tcp);
1302 }
1303#endif
1304 LOOP_LABEL(tcp, so, so_next);
1305 }
1306
1307 /*
1308 * Now UDP sockets.
1309 * Incoming packets are sent straight away, they're not buffered.
1310 * Incoming UDP data isn't buffered either.
1311 */
1312 QSOCKET_FOREACH(so, so_next, udp)
1313 /* { */
1314#ifdef VBOX_WITH_SLIRP_MT
1315 if ( so->so_state & SS_NOFDREF
1316 && so->so_deleted == 1)
1317 {
1318 struct socket *son, *sop = NULL;
1319 QSOCKET_LOCK(udb);
1320 if (so->so_next != NULL)
1321 {
1322 if (so->so_next != &udb)
1323 SOCKET_LOCK(so->so_next);
1324 son = so->so_next;
1325 }
1326 if ( so->so_prev != &udb
1327 && so->so_prev != NULL)
1328 {
1329 SOCKET_LOCK(so->so_prev);
1330 sop = so->so_prev;
1331 }
1332 QSOCKET_UNLOCK(udb);
1333 remque(pData, so);
1334 NSOCK_DEC();
1335 SOCKET_UNLOCK(so);
1336 SOCKET_LOCK_DESTROY(so);
1337 RTMemFree(so);
1338 so_next = son;
1339 if (sop != NULL)
1340 SOCKET_UNLOCK(sop);
1341 CONTINUE_NO_UNLOCK(udp);
1342 }
1343#endif
1344 POLL_UDP_EVENTS(rc, error, so, &NetworkEvents);
1345
1346 LOG_NAT_SOCK(so, UDP, &NetworkEvents, readfds, writefds, xfds);
1347
1348 if (so->s != -1 && CHECK_FD_SET(so, NetworkEvents, readfds))
1349 {
1350 SORECVFROM(pData, so);
1351 }
1352 LOOP_LABEL(udp, so, so_next);
1353 }
1354
1355 }
1356
1357#ifndef VBOX_WITH_SLIRP_MT
1358 /*
1359 * See if we can start outputting
1360 */
1361 if (if_queued && link_up)
1362 if_start(pData);
1363#endif
1364
1365 STAM_PROFILE_STOP(&pData->StatPoll, a);
1366}
1367
1368#define ETH_ALEN 6
1369#define ETH_HLEN 14
1370
1371#define ARPOP_REQUEST 1 /* ARP request */
1372#define ARPOP_REPLY 2 /* ARP reply */
1373
1374struct ethhdr
1375{
1376 unsigned char h_dest[ETH_ALEN]; /* destination eth addr */
1377 unsigned char h_source[ETH_ALEN]; /* source ether addr */
1378 unsigned short h_proto; /* packet type ID field */
1379};
1380AssertCompileSize(struct ethhdr, 14);
1381
1382struct arphdr
1383{
1384 unsigned short ar_hrd; /* format of hardware address */
1385 unsigned short ar_pro; /* format of protocol address */
1386 unsigned char ar_hln; /* length of hardware address */
1387 unsigned char ar_pln; /* length of protocol address */
1388 unsigned short ar_op; /* ARP opcode (command) */
1389
1390 /*
1391 * Ethernet looks like this : This bit is variable sized however...
1392 */
1393 unsigned char ar_sha[ETH_ALEN]; /* sender hardware address */
1394 unsigned char ar_sip[4]; /* sender IP address */
1395 unsigned char ar_tha[ETH_ALEN]; /* target hardware address */
1396 unsigned char ar_tip[4]; /* target IP address */
1397};
1398AssertCompileSize(struct arphdr, 28);
1399
1400static void arp_input(PNATState pData, struct mbuf *m)
1401{
1402 struct ethhdr *eh;
1403 struct ethhdr *reh;
1404 struct arphdr *ah;
1405 struct arphdr *rah;
1406 int ar_op;
1407 struct ex_list *ex_ptr;
1408 uint32_t htip;
1409 struct mbuf *mr;
1410 eh = mtod(m, struct ethhdr *);
1411 ah = (struct arphdr *)&eh[1];
1412 htip = ntohl(*(uint32_t*)ah->ar_tip);
1413
1414 mr = m_get(pData);
1415 mr->m_data += if_maxlinkhdr;
1416 mr->m_len = sizeof(struct arphdr);
1417 rah = mtod(mr, struct arphdr *);
1418
1419 ar_op = ntohs(ah->ar_op);
1420 switch(ar_op)
1421 {
1422 case ARPOP_REQUEST:
1423 if ((htip & pData->netmask) == ntohl(special_addr.s_addr))
1424 {
1425 if ( CTL_CHECK(htip, CTL_DNS)
1426 || CTL_CHECK(htip, CTL_ALIAS)
1427 || CTL_CHECK(htip, CTL_TFTP))
1428 goto arp_ok;
1429 for (ex_ptr = exec_list; ex_ptr; ex_ptr = ex_ptr->ex_next)
1430 {
1431 if ((htip & ~pData->netmask) == ex_ptr->ex_addr)
1432 goto arp_ok;
1433 }
1434 return;
1435 arp_ok:
1436 rah->ar_hrd = htons(1);
1437 rah->ar_pro = htons(ETH_P_IP);
1438 rah->ar_hln = ETH_ALEN;
1439 rah->ar_pln = 4;
1440 rah->ar_op = htons(ARPOP_REPLY);
1441 memcpy(rah->ar_sha, special_ethaddr, ETH_ALEN);
1442
1443 switch (htip & ~pData->netmask)
1444 {
1445 case CTL_DNS:
1446 case CTL_ALIAS:
1447 rah->ar_sha[5] = (uint8_t)(htip & ~pData->netmask);
1448 break;
1449 default:;
1450 }
1451
1452 memcpy(rah->ar_sip, ah->ar_tip, 4);
1453 memcpy(rah->ar_tha, ah->ar_sha, ETH_ALEN);
1454 memcpy(rah->ar_tip, ah->ar_sip, 4);
1455 if_encap(pData, ETH_P_ARP, mr);
1456 m_free(pData, m);
1457 }
1458 break;
1459 default:
1460 break;
1461 }
1462}
1463
1464void slirp_input(PNATState pData, const uint8_t *pkt, int pkt_len)
1465{
1466 struct mbuf *m;
1467 int proto;
1468 static bool fWarnedIpv6;
1469
1470 if (pkt_len < ETH_HLEN)
1471 {
1472 LogRel(("NAT: packet having size %d has been ingnored\n", pkt_len));
1473 return;
1474 }
1475
1476 m = m_get(pData);
1477 if (!m)
1478 {
1479 LogRel(("NAT: can't allocate new mbuf\n"));
1480 return;
1481 }
1482
1483 /* Note: we add to align the IP header */
1484
1485 if (M_FREEROOM(m) < pkt_len)
1486 m_inc(m, pkt_len);
1487
1488 m->m_len = pkt_len ;
1489 memcpy(m->m_data, pkt, pkt_len);
1490
1491 proto = ntohs(*(uint16_t *)(pkt + 12));
1492 switch(proto)
1493 {
1494 case ETH_P_ARP:
1495 arp_input(pData, m);
1496 break;
1497 case ETH_P_IP:
1498 /* Update time. Important if the network is very quiet, as otherwise
1499 * the first outgoing connection gets an incorrect timestamp. */
1500 updtime(pData);
1501 m->m_data += ETH_HLEN;
1502 m->m_len -= ETH_HLEN;
1503 ip_input(pData, m);
1504 break;
1505 case ETH_P_IPV6:
1506 m_free(pData, m);
1507 if (!fWarnedIpv6)
1508 {
1509 LogRel(("NAT: IPv6 not supported\n"));
1510 fWarnedIpv6 = true;
1511 }
1512 break;
1513 default:
1514 LogRel(("NAT: Unsupported protocol %x\n", proto));
1515 m_free(pData, m);
1516 break;
1517 }
1518 RTMemFree((void *)pkt);
1519}
1520
1521/* output the IP packet to the ethernet device */
1522void if_encap(PNATState pData, uint16_t eth_proto, struct mbuf *m)
1523{
1524 struct ethhdr *eh;
1525 uint8_t *buf = RTMemAlloc(1600);
1526
1527 m->m_data -= if_maxlinkhdr;
1528 m->m_len += ETH_HLEN;
1529 eh = mtod(m, struct ethhdr *);
1530
1531 memcpy(eh->h_dest, client_ethaddr, ETH_ALEN);
1532 memcpy(eh->h_source, special_ethaddr, ETH_ALEN - 1);
1533 /* XXX: not correct */
1534 eh->h_source[5] = CTL_ALIAS;
1535 eh->h_proto = htons(eth_proto);
1536#if 0
1537 slirp_output(pData->pvUser, m, mtod(m, uint8_t *), m->m_len);
1538#else
1539 memcpy(buf, mtod(m, uint8_t *), m->m_len);
1540 slirp_output(pData->pvUser, NULL, buf, m->m_len);
1541 m_free(pData, m);
1542#endif
1543}
1544
1545int slirp_redir(PNATState pData, int is_udp, int host_port,
1546 struct in_addr guest_addr, int guest_port)
1547{
1548 struct socket *so;
1549 Log2(("NAT: set redirect %s hp:%d gp:%d\n", (is_udp?"UDP":"TCP"), host_port, guest_port));
1550 if (is_udp)
1551 {
1552 so = udp_listen(pData, htons(host_port), guest_addr.s_addr,
1553 htons(guest_port), 0);
1554 }
1555 else
1556 {
1557 so = solisten(pData, htons(host_port), guest_addr.s_addr,
1558 htons(guest_port), 0);
1559 }
1560 Log2(("NAT: redirecting socket %R[natsock]\n", so));
1561 return (so != NULL ? 0 : -1);
1562}
1563
1564int slirp_add_exec(PNATState pData, int do_pty, const char *args, int addr_low_byte,
1565 int guest_port)
1566{
1567 return add_exec(&exec_list, do_pty, (char *)args,
1568 addr_low_byte, htons(guest_port));
1569}
1570
1571void slirp_set_ethaddr(PNATState pData, const uint8_t *ethaddr)
1572{
1573 memcpy(client_ethaddr, ethaddr, ETH_ALEN);
1574}
1575
1576#if defined(RT_OS_WINDOWS)
1577HANDLE *slirp_get_events(PNATState pData)
1578{
1579 return pData->phEvents;
1580}
1581void slirp_register_external_event(PNATState pData, HANDLE hEvent, int index)
1582{
1583 pData->phEvents[index] = hEvent;
1584}
1585#endif
1586
1587unsigned int slirp_get_timeout_ms(PNATState pData)
1588{
1589 if (link_up)
1590 {
1591 if (time_fasttimo)
1592 return 2;
1593 if (do_slowtimo)
1594 return 500; /* see PR_SLOWHZ */
1595 }
1596 return 0;
1597}
1598
1599#ifndef RT_OS_WINDOWS
1600int slirp_get_nsock(PNATState pData)
1601{
1602 return pData->nsock;
1603}
1604#endif
1605
1606/*
1607 * this function called from NAT thread
1608 */
1609void slirp_post_sent(PNATState pData, void *pvArg)
1610{
1611 struct socket *so = 0;
1612 struct tcpcb *tp = 0;
1613 struct mbuf *m = (struct mbuf *)pvArg;
1614 m_free(pData, m);
1615}
1616#ifdef VBOX_WITH_SLIRP_MT
1617void slirp_process_queue(PNATState pData)
1618{
1619 RTReqProcess(pData->pReqQueue, RT_INDEFINITE_WAIT);
1620}
1621void *slirp_get_queue(PNATState pData)
1622{
1623 return pData->pReqQueue;
1624}
1625#endif
1626
1627uint16_t slirp_get_service(int proto, uint16_t dport, uint16_t sport)
1628{
1629 uint16_t hdport, hsport, service;
1630 hdport = ntohs(dport);
1631 hsport = ntohs(sport);
1632 Log2(("proto: %d, dport: %d sport: %d\n", proto, hdport, hsport));
1633 service = 0;
1634#if 0
1635 /* Always return 0 here */
1636 switch (hdport)
1637 {
1638 case 500:
1639 /* service = sport; */
1640 break;
1641 }
1642#endif
1643 Log2(("service : %d\n", service));
1644 return htons(service);
1645}
1646
1647void slirp_set_dhcp_TFTP_prefix(PNATState pData, const char *tftpPrefix)
1648{
1649 Log2(("tftp_prefix:%s\n", tftpPrefix));
1650 tftp_prefix = tftpPrefix;
1651}
1652
1653void slirp_set_dhcp_TFTP_bootfile(PNATState pData, const char *bootFile)
1654{
1655 Log2(("bootFile:%s\n", bootFile));
1656 bootp_filename = bootFile;
1657}
1658
1659void slirp_set_dhcp_next_server(PNATState pData, const char *next_server)
1660{
1661 Log2(("next_server:%s\n", next_server));
1662 if (next_server == NULL)
1663 pData->tftp_server.s_addr = htonl(ntohl(special_addr.s_addr) | CTL_TFTP);
1664 else
1665 inet_aton(next_server, &pData->tftp_server);
1666}
1667#ifdef VBOX_WITH_SLIRP_DNS_PROXY
1668void slirp_set_dhcp_dns_proxy(PNATState pData, bool fDNSProxy)
1669{
1670 Log2(("NAT: DNS proxy switched %s\n", (fDNSProxy ? "on" : "off")));
1671 pData->use_dns_proxy = fDNSProxy;
1672}
1673#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