VirtualBox

source: vbox/trunk/src/VBox/Devices/Network/slirp/slirp.c@ 20752

最後變更 在這個檔案從20752是 20717,由 vboxsync 提交於 16 年 前

slirp: todo
  • 屬性 svn:eol-style 設為 native
檔案大小: 58.6 KB
行 
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#ifdef VBOX_WITH_SLIRP_ALIAS
21# include <alias.h>
22#endif
23
24#if !defined(RT_OS_WINDOWS)
25
26# define DO_ENGAGE_EVENT1(so, fdset, label) \
27 do { \
28 if( so->so_poll_index != -1 \
29 && so->s == polls[so->so_poll_index].fd) { \
30 polls[so->so_poll_index].events |= N_(fdset ## _poll); \
31 break; /* out of this loop */ \
32 } \
33 AssertRelease(poll_index < (nfds)); \
34 AssertRelease(poll_index >= 0 && poll_index < (nfds)); \
35 polls[poll_index].fd = (so)->s; \
36 (so)->so_poll_index = poll_index; \
37 polls[poll_index].events = N_(fdset ## _poll); \
38 polls[poll_index].revents = 0; \
39 poll_index++; \
40 } while(0)
41
42
43# define DO_ENGAGE_EVENT2(so, fdset1, fdset2, label) \
44 do { \
45 if( so->so_poll_index != -1 \
46 && so->s == polls[so->so_poll_index].fd) { \
47 polls[so->so_poll_index].events |= \
48 N_(fdset1 ## _poll) | N_(fdset1 ## _poll); \
49 break; /* out of this loop */ \
50 } \
51 AssertRelease(poll_index < (nfds)); \
52 polls[poll_index].fd = (so)->s; \
53 (so)->so_poll_index = poll_index; \
54 polls[poll_index].events = \
55 N_(fdset1 ## _poll) | N_(fdset1 ## _poll); \
56 poll_index++; \
57 } while(0)
58
59# define DO_POLL_EVENTS(rc, error, so, events, label) do {} while (0)
60
61# define DO_CHECK_FD_SET(so, events, fdset) ( ((so)->so_poll_index != -1) \
62 && ((so)->so_poll_index <= ndfs) \
63 && ((so)->s == polls[so->so_poll_index].fd) \
64 && (polls[(so)->so_poll_index].revents & N_(fdset ## _poll)))
65# define DO_UNIX_CHECK_FD_SET(so, events, fdset ) DO_CHECK_FD_SET((so), (events), fdset) /*specific for Unix API */
66# define DO_WIN_CHECK_FD_SET(so, events, fdset ) 0 /* specific for Windows Winsock API */
67
68# ifndef RT_OS_WINDOWS
69
70# ifndef RT_OS_LINUX
71# define readfds_poll (POLLRDNORM)
72# define writefds_poll (POLLWRNORM)
73# define xfds_poll (POLLRDBAND|POLLWRBAND|POLLPRI)
74# else
75# define readfds_poll (POLLIN)
76# define writefds_poll (POLLOUT)
77# define xfds_poll (POLLPRI)
78# endif
79# define rderr_poll (POLLERR)
80# define rdhup_poll (POLLHUP)
81# define nval_poll (POLLNVAL)
82
83# define ICMP_ENGAGE_EVENT(so, fdset) \
84 do { \
85 if (pData->icmp_socket.s != -1) \
86 DO_ENGAGE_EVENT1((so), fdset, ICMP); \
87 } while (0)
88# else /* !RT_OS_WINDOWS */
89# define DO_WIN_CHECK_FD_SET(so, events, fdset ) DO_CHECK_FD_SET((so), (events), fdset)
90# define ICMP_ENGAGE_EVENT(so, fdset) do {} while(0)
91#endif /* RT_OS_WINDOWS */
92
93#else /* defined(RT_OS_WINDOWS) */
94
95/*
96 * On Windows, we will be notified by IcmpSendEcho2() when the response arrives.
97 * So no call to WSAEventSelect necessary.
98 */
99# define ICMP_ENGAGE_EVENT(so, fdset) do {} while(0)
100
101# define DO_ENGAGE_EVENT1(so, fdset1, label) \
102 do { \
103 rc = WSAEventSelect((so)->s, VBOX_SOCKET_EVENT, FD_ALL_EVENTS); \
104 if (rc == SOCKET_ERROR) \
105 { \
106 /* This should not happen */ \
107 error = WSAGetLastError(); \
108 LogRel(("WSAEventSelect (" #label ") error %d (so=%x, socket=%s, event=%x)\n", \
109 error, (so), (so)->s, VBOX_SOCKET_EVENT)); \
110 } \
111 } while(0); \
112 CONTINUE(label)
113
114# define DO_ENGAGE_EVENT2(so, fdset1, fdset2, label) \
115 DO_ENGAGE_EVENT1((so), (fdset1), label)
116
117# define DO_POLL_EVENTS(rc, error, so, events, label) \
118 (rc) = WSAEnumNetworkEvents((so)->s, VBOX_SOCKET_EVENT, (events)); \
119 if ((rc) == SOCKET_ERROR) \
120 { \
121 (error) = WSAGetLastError(); \
122 LogRel(("WSAEnumNetworkEvents " #label " error %d\n", (error))); \
123 CONTINUE(label); \
124 }
125
126# define acceptds_win FD_ACCEPT
127# define acceptds_win_bit FD_ACCEPT_BIT
128
129# define readfds_win FD_READ
130# define readfds_win_bit FD_READ_BIT
131
132# define writefds_win FD_WRITE
133# define writefds_win_bit FD_WRITE_BIT
134
135# define xfds_win FD_OOB
136# define xfds_win_bit FD_OOB_BIT
137
138# define DO_CHECK_FD_SET(so, events, fdset) \
139 (((events).lNetworkEvents & fdset ## _win) && ((events).iErrorCode[fdset ## _win_bit] == 0))
140
141# define DO_WIN_CHECK_FD_SET(so, events, fdset ) DO_CHECK_FD_SET((so), (events), fdset)
142# define DO_UNIX_CHECK_FD_SET(so, events, fdset ) 1 /*specific for Unix API */
143
144#endif /* defined(RT_OS_WINDOWS) */
145
146#define TCP_ENGAGE_EVENT1(so, fdset) \
147 DO_ENGAGE_EVENT1((so), fdset, tcp)
148
149#define TCP_ENGAGE_EVENT2(so, fdset1, fdset2) \
150 DO_ENGAGE_EVENT2((so), fdset1, fdset2, tcp)
151
152#define UDP_ENGAGE_EVENT(so, fdset) \
153 DO_ENGAGE_EVENT1((so), fdset, udp)
154
155#define POLL_TCP_EVENTS(rc, error, so, events) \
156 DO_POLL_EVENTS((rc), (error), (so), (events), tcp)
157
158#define POLL_UDP_EVENTS(rc, error, so, events) \
159 DO_POLL_EVENTS((rc), (error), (so), (events), udp)
160
161#define CHECK_FD_SET(so, events, set) \
162 (DO_CHECK_FD_SET((so), (events), set))
163
164#define WIN_CHECK_FD_SET(so, events, set) \
165 (DO_WIN_CHECK_FD_SET((so), (events), set))
166#define UNIX_CHECK_FD_SET(so, events, set) \
167 (DO_UNIX_CHECK_FD_SET(so, events, set))
168
169/*
170 * Loging macros
171 */
172#if VBOX_WITH_DEBUG_NAT_SOCKETS
173# if defined(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] %R[natwinnetevents]\n", (so), (winevent))); \
177 } while (0)
178# else /* RT_OS_WINDOWS */
179# define DO_LOG_NAT_SOCK(so, proto, winevent, r_fdset, w_fdset, x_fdset) \
180 do { \
181 LogRel((" " #proto " %R[natsock] %s %s %s er: %s, %s, %s\n", (so), \
182 CHECK_FD_SET(so, ign ,r_fdset) ? "READ":"", \
183 CHECK_FD_SET(so, ign, w_fdset) ? "WRITE":"", \
184 CHECK_FD_SET(so, ign, x_fdset) ? "OOB":"", \
185 CHECK_FD_SET(so, ign, rderr) ? "RDERR":"", \
186 CHECK_FD_SET(so, ign, rdhup) ? "RDHUP":"", \
187 CHECK_FD_SET(so, ign, nval) ? "RDNVAL":"")); \
188 } while (0)
189# endif /* !RT_OS_WINDOWS */
190#else /* VBOX_WITH_DEBUG_NAT_SOCKETS */
191# define DO_LOG_NAT_SOCK(so, proto, winevent, r_fdset, w_fdset, x_fdset) do {} while (0)
192#endif /* !VBOX_WITH_DEBUG_NAT_SOCKETS */
193
194#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)
195
196static const uint8_t special_ethaddr[6] =
197{
198 0x52, 0x54, 0x00, 0x12, 0x35, 0x00
199};
200
201static const uint8_t broadcast_ethaddr[6] =
202{
203 0xff, 0xff, 0xff, 0xff, 0xff, 0xff
204};
205
206const uint8_t zerro_ethaddr[6] =
207{
208 0x0, 0x0, 0x0, 0x0, 0x0, 0x0
209};
210
211#ifdef RT_OS_WINDOWS
212# ifndef VBOX_WITH_MULTI_DNS
213static int get_dns_addr_domain(PNATState pData, bool fVerbose,
214 struct in_addr *pdns_addr,
215 const char **ppszDomain)
216{
217 int rc = 0;
218 FIXED_INFO *FixedInfo = NULL;
219 ULONG BufLen;
220 DWORD ret;
221 IP_ADDR_STRING *pIPAddr;
222 struct in_addr tmp_addr;
223
224 FixedInfo = (FIXED_INFO *)GlobalAlloc(GPTR, sizeof(FIXED_INFO));
225 BufLen = sizeof(FIXED_INFO);
226
227 /** @todo: this API returns all DNS servers, no matter whether the
228 * corresponding network adapter is disabled or not. Maybe replace
229 * this by GetAdapterAddresses(), which is XP/Vista only though. */
230 if (ERROR_BUFFER_OVERFLOW == GetNetworkParams(FixedInfo, &BufLen))
231 {
232 if (FixedInfo)
233 {
234 GlobalFree(FixedInfo);
235 FixedInfo = NULL;
236 }
237 FixedInfo = GlobalAlloc(GPTR, BufLen);
238 }
239
240 if ((ret = GetNetworkParams(FixedInfo, &BufLen)) != ERROR_SUCCESS)
241 {
242 Log(("GetNetworkParams failed. ret = %08x\n", (u_int)ret ));
243 if (FixedInfo)
244 {
245 GlobalFree(FixedInfo);
246 FixedInfo = NULL;
247 }
248 rc = -1;
249 goto get_dns_prefix;
250 }
251
252 pIPAddr = &(FixedInfo->DnsServerList);
253 inet_aton(pIPAddr->IpAddress.String, &tmp_addr);
254 Log(("nat: DNS Servers:\n"));
255 if (fVerbose || pdns_addr->s_addr != tmp_addr.s_addr)
256 LogRel(("NAT: DNS address: %s\n", pIPAddr->IpAddress.String));
257 *pdns_addr = tmp_addr;
258
259 pIPAddr = FixedInfo -> DnsServerList.Next;
260 while (pIPAddr)
261 {
262 if (fVerbose)
263 LogRel(("NAT: ignored DNS address: %s\n", pIPAddr ->IpAddress.String));
264 pIPAddr = pIPAddr ->Next;
265 }
266 if (FixedInfo)
267 {
268 GlobalFree(FixedInfo);
269 FixedInfo = NULL;
270 }
271
272get_dns_prefix:
273 if (ppszDomain)
274 {
275 OSVERSIONINFO ver;
276 char szDnsDomain[256];
277 DWORD dwSize = sizeof(szDnsDomain);
278
279 *ppszDomain = NULL;
280 GetVersionEx(&ver);
281 if (ver.dwMajorVersion >= 5)
282 {
283 /* GetComputerNameEx exists in Windows versions starting with 2000. */
284 if (GetComputerNameEx(ComputerNameDnsDomain, szDnsDomain, &dwSize))
285 {
286 if (szDnsDomain[0])
287 {
288 /* Just non-empty strings are valid. */
289 *ppszDomain = RTStrDup(szDnsDomain);
290 if (pData->fPassDomain)
291 {
292 if (fVerbose)
293 LogRel(("NAT: passing domain name %s\n", szDnsDomain));
294 }
295 else
296 Log(("nat: ignoring domain %s\n", szDnsDomain));
297 }
298 }
299 else
300 Log(("nat: GetComputerNameEx failed (%d)\n", GetLastError()));
301 }
302 }
303 return rc;
304}
305# else /* !VBOX_WITH_MULTI_DNS */
306static int get_dns_addr_domain(PNATState pData, bool fVerbose,
307 struct in_addr *pdns_addr,
308 const char **ppszDomain)
309{
310 /* Get amount of memory required for operation */
311 ULONG flags = GAA_FLAG_INCLUDE_PREFIX; /*GAA_FLAG_INCLUDE_ALL_INTERFACES;*/ /* all interfaces registered in NDIS */
312 PIP_ADAPTER_ADDRESSES addresses = NULL;
313 PIP_ADAPTER_ADDRESSES addr = NULL;
314 PIP_ADAPTER_DNS_SERVER_ADDRESS dns = NULL;
315 ULONG size = 0;
316 int wlen = 0;
317 char *suffix;
318 struct dns_entry *da = NULL;
319 struct dns_domain_entry *dd = NULL;
320 ULONG ret = ERROR_SUCCESS;
321
322 /* @todo add SKIPing flags to get only required information */
323
324 ret = pData->pfGetAdaptersAddresses(AF_INET, 0, NULL /* reserved */, addresses, &size);
325 if (ret != ERROR_BUFFER_OVERFLOW)
326 {
327 LogRel(("NAT: error %lu occured on capacity detection operation\n", ret));
328 return -1;
329 }
330
331 if (size == 0)
332 {
333 LogRel(("NAT: Win socket API returns non capacity\n"));
334 return -1;
335 }
336
337 addresses = RTMemAllocZ(size);
338 if (addresses == NULL)
339 {
340 LogRel(("NAT: No memory available \n"));
341 return -1;
342 }
343
344 ret = pData->pfGetAdaptersAddresses(AF_INET, 0, NULL /* reserved */, addresses, &size);
345 if (ret != ERROR_SUCCESS)
346 {
347 LogRel(("NAT: error %lu occured on fetching adapters info\n", ret));
348 RTMemFree(addresses);
349 return -1;
350 }
351 addr = addresses;
352 while(addr != NULL)
353 {
354 int found;
355 if (addr->OperStatus != IfOperStatusUp)
356 goto next;
357 dns = addr->FirstDnsServerAddress;
358 while (dns != NULL)
359 {
360 struct sockaddr *saddr = dns->Address.lpSockaddr;
361 if (saddr->sa_family != AF_INET)
362 goto next_dns;
363 /* add dns server to list */
364 da = RTMemAllocZ(sizeof(struct dns_entry));
365 if (da == NULL)
366 {
367 LogRel(("NAT: Can't allocate buffer for DNS entry\n"));
368 RTMemFree(addresses);
369 return VERR_NO_MEMORY;
370 }
371 LogRel(("NAT: adding %R[IP4] to DNS server list\n", &((struct sockaddr_in *)saddr)->sin_addr));
372 if ((((struct sockaddr_in *)saddr)->sin_addr.s_addr & htonl(IN_CLASSA_NET)) == ntohl(INADDR_LOOPBACK & IN_CLASSA_NET)) {
373 da->de_addr.s_addr = htonl(ntohl(special_addr.s_addr) | CTL_ALIAS);
374 }
375 else
376 {
377 da->de_addr.s_addr = ((struct sockaddr_in *)saddr)->sin_addr.s_addr;
378 }
379 TAILQ_INSERT_HEAD(&pData->dns_list_head, da, de_list);
380
381 if (addr->DnsSuffix == NULL)
382 goto next_dns;
383
384 /*uniq*/
385 RTUtf16ToUtf8(addr->DnsSuffix, &suffix);
386
387 if (!suffix || strlen(suffix) == 0) {
388 RTStrFree(suffix);
389 goto next_dns;
390 }
391
392 found = 0;
393 LIST_FOREACH(dd, &pData->dns_domain_list_head, dd_list)
394 {
395 if ( dd->dd_pszDomain != NULL
396 && strcmp(dd->dd_pszDomain, suffix) == 0)
397 {
398 found = 1;
399 RTStrFree(suffix);
400 break;
401 }
402 }
403 if (found == 0)
404 {
405 dd = RTMemAllocZ(sizeof(struct dns_domain_entry));
406 if (dd == NULL)
407 {
408 LogRel(("NAT: not enough memory\n"));
409 RTStrFree(suffix);
410 RTMemFree(addresses);
411 return VERR_NO_MEMORY;
412 }
413 dd->dd_pszDomain = suffix;
414 LogRel(("NAT: adding domain name %s to search list\n", dd->dd_pszDomain));
415 LIST_INSERT_HEAD(&pData->dns_domain_list_head, dd, dd_list);
416 }
417 next_dns:
418 dns = dns->Next;
419 }
420 next:
421 addr = addr->Next;
422 }
423 RTMemFree(addresses);
424 return 0;
425}
426# endif /* VBOX_WITH_MULTI_DNS */
427
428#else /* !RT_OS_WINDOWS */
429
430static int get_dns_addr_domain(PNATState pData, bool fVerbose,
431 struct in_addr *pdns_addr,
432 const char **ppszDomain)
433{
434 char buff[512];
435 char buff2[256];
436 FILE *f = NULL;
437 int found = 0;
438 struct in_addr tmp_addr;
439
440#ifdef RT_OS_OS2
441 /* Try various locations. */
442 char *etc = getenv("ETC");
443 if (etc)
444 {
445 snprintf(buff, sizeof(buff), "%s/RESOLV2", etc);
446 f = fopen(buff, "rt");
447 }
448 if (!f)
449 {
450 snprintf(buff, sizeof(buff), "%s/RESOLV2", _PATH_ETC);
451 f = fopen(buff, "rt");
452 }
453 if (!f)
454 {
455 snprintf(buff, sizeof(buff), "%s/resolv.conf", _PATH_ETC);
456 f = fopen(buff, "rt");
457 }
458#else
459#ifndef DEBUG_vvl
460 f = fopen("/etc/resolv.conf", "r");
461#else
462 char *home = getenv("HOME");
463 snprintf(buff, sizeof(buff), "%s/resolv.conf", home);
464 f = fopen(buff, "r");
465 if (f != NULL)
466 {
467 Log(("NAT: DNS we're using %s\n", buff));
468 }
469 else
470 {
471 f = fopen("/etc/resolv.conf", "r");
472 Log(("NAT: DNS we're using %s\n", buff));
473 }
474#endif
475#endif
476 if (!f)
477 return -1;
478
479 if (ppszDomain)
480 *ppszDomain = NULL;
481 Log(("nat: DNS Servers:\n"));
482 while (fgets(buff, 512, f) != NULL)
483 {
484#ifdef VBOX_WITH_MULTI_DNS
485 struct dns_entry *da = NULL;
486#endif
487 if (sscanf(buff, "nameserver%*[ \t]%256s", buff2) == 1)
488 {
489 if (!inet_aton(buff2, &tmp_addr))
490 continue;
491#ifndef VBOX_WITH_MULTI_DNS
492 /* If it's the first one, set it to dns_addr */
493 if (!found)
494 {
495 if (fVerbose || pdns_addr->s_addr != tmp_addr.s_addr)
496 LogRel(("NAT: DNS address: %s\n", buff2));
497 *pdns_addr = tmp_addr;
498 }
499 else
500 {
501 if (fVerbose)
502 LogRel(("NAT: ignored DNS address: %s\n", buff2));
503 }
504#else
505 /*localhost mask */
506 da = RTMemAllocZ(sizeof (struct dns_entry));
507 if (da == NULL)
508 {
509 LogRel(("can't alloc memory for DNS entry\n"));
510 return -1;
511 }
512 /*check */
513 da->de_addr.s_addr = tmp_addr.s_addr;
514 if ((da->de_addr.s_addr & htonl(IN_CLASSA_NET)) == ntohl(INADDR_LOOPBACK & IN_CLASSA_NET)) {
515 da->de_addr.s_addr = htonl(ntohl(special_addr.s_addr) | CTL_ALIAS);
516 }
517 TAILQ_INSERT_HEAD(&pData->dns_list_head, da, de_list);
518#endif
519 found++;
520 }
521#ifndef VBOX_WITH_MULTI_DNS
522 if ( ppszDomain
523 && (!strncmp(buff, "domain", 6) || !strncmp(buff, "search", 6)))
524 {
525 /* Domain name/search list present. Pick first entry */
526 if (*ppszDomain == NULL)
527 {
528 char *tok;
529 char *saveptr;
530 tok = strtok_r(&buff[6], " \t\n", &saveptr);
531 if (tok)
532 {
533 *ppszDomain = RTStrDup(tok);
534 if (pData->fPassDomain)
535 {
536 if (fVerbose)
537 LogRel(("NAT: passing domain name %s\n", tok));
538 }
539 else
540 Log(("nat: ignoring domain %s\n", tok));
541 }
542 }
543 }
544#else
545 if ((!strncmp(buff, "domain", 6) || !strncmp(buff, "search", 6)))
546 {
547 char *tok;
548 char *saveptr;
549 struct dns_domain_entry *dd = NULL;
550 int found = 0;
551 tok = strtok_r(&buff[6], " \t\n", &saveptr);
552 LIST_FOREACH(dd, &pData->dns_domain_list_head, dd_list)
553 {
554 if( tok != NULL
555 && strcmp(tok, dd->dd_pszDomain) == 0)
556 {
557 found = 1;
558 break;
559 }
560 }
561 if (tok != NULL && found == 0) {
562 dd = RTMemAllocZ(sizeof(struct dns_domain_entry));
563 if (dd == NULL)
564 {
565 LogRel(("NAT: not enought memory to add domain list\n"));
566 return VERR_NO_MEMORY;
567 }
568 dd->dd_pszDomain = RTStrDup(tok);
569 LogRel(("NAT: adding domain name %s to search list\n", dd->dd_pszDomain));
570 LIST_INSERT_HEAD(&pData->dns_domain_list_head, dd, dd_list);
571 }
572 }
573#endif
574 }
575 fclose(f);
576 if (!found)
577 return -1;
578 return 0;
579}
580
581#endif
582#ifdef VBOX_WITH_MULTI_DNS
583static int slirp_init_dns_list(PNATState pData)
584{
585 TAILQ_INIT(&pData->dns_list_head);
586 LIST_INIT(&pData->dns_domain_list_head);
587 return get_dns_addr_domain(pData, true, NULL, NULL);
588}
589
590static void slirp_release_dns_list(PNATState pData)
591{
592 struct dns_entry *de = NULL;
593 struct dns_domain_entry *dd = NULL;
594 while(!TAILQ_EMPTY(&pData->dns_list_head)) {
595 de = TAILQ_FIRST(&pData->dns_list_head);
596 TAILQ_REMOVE(&pData->dns_list_head, de, de_list);
597 RTMemFree(de);
598 }
599 while(!LIST_EMPTY(&pData->dns_domain_list_head)) {
600 dd = LIST_FIRST(&pData->dns_domain_list_head);
601 LIST_REMOVE(dd, dd_list);
602 if (dd->dd_pszDomain != NULL)
603 RTStrFree(dd->dd_pszDomain);
604 RTMemFree(dd);
605 }
606}
607#endif
608
609int get_dns_addr(PNATState pData, struct in_addr *pdns_addr)
610{
611 return get_dns_addr_domain(pData, false, pdns_addr, NULL);
612}
613
614#ifndef VBOX_WITH_NAT_SERVICE
615int slirp_init(PNATState *ppData, const char *pszNetAddr, uint32_t u32Netmask,
616 bool fPassDomain, void *pvUser)
617#else
618int slirp_init(PNATState *ppData, uint32_t u32NetAddr, uint32_t u32Netmask,
619 bool fPassDomain, void *pvUser)
620#endif
621{
622 int fNATfailed = 0;
623 int rc;
624 PNATState pData = RTMemAllocZ(sizeof(NATState));
625 *ppData = pData;
626 if (!pData)
627 return VERR_NO_MEMORY;
628 if (u32Netmask & 0x1f)
629 /* CTL is x.x.x.15, bootp passes up to 16 IPs (15..31) */
630 return VERR_INVALID_PARAMETER;
631 pData->fPassDomain = fPassDomain;
632 pData->pvUser = pvUser;
633 pData->netmask = u32Netmask;
634
635 /* sockets & TCP defaults */
636 pData->socket_rcv = 64 * _1K;
637 pData->socket_snd = 64 * _1K;
638 tcp_sndspace = 64 * _1K;
639 tcp_rcvspace = 64 * _1K;
640
641#ifdef RT_OS_WINDOWS
642 {
643 WSADATA Data;
644 WSAStartup(MAKEWORD(2, 0), &Data);
645 }
646 pData->phEvents[VBOX_SOCKET_EVENT_INDEX] = CreateEvent(NULL, FALSE, FALSE, NULL);
647#endif
648#ifdef VBOX_WITH_SLIRP_MT
649 QSOCKET_LOCK_CREATE(tcb);
650 QSOCKET_LOCK_CREATE(udb);
651 rc = RTReqCreateQueue(&pData->pReqQueue);
652 AssertReleaseRC(rc);
653#endif
654
655 link_up = 1;
656
657 debug_init();
658 if_init(pData);
659 ip_init(pData);
660 icmp_init(pData);
661
662 /* Initialise mbufs *after* setting the MTU */
663 m_init(pData);
664
665#ifndef VBOX_WITH_NAT_SERVICE
666 inet_aton(pszNetAddr, &special_addr);
667#else
668 special_addr.s_addr = u32NetAddr;
669#endif
670#ifdef VBOX_WITHOUT_SLIRP_CLIENT_ETHER
671 pData->slirp_ethaddr = &special_ethaddr[0];
672#endif
673 alias_addr.s_addr = special_addr.s_addr | htonl(CTL_ALIAS);
674 /* @todo: add ability to configure this staff */
675
676 /* set default addresses */
677 inet_aton("127.0.0.1", &loopback_addr);
678#ifndef VBOX_WITH_MULTI_DNS
679 inet_aton("127.0.0.1", &dns_addr);
680
681 if (get_dns_addr_domain(pData, true, &dns_addr, &pData->pszDomain) < 0)
682#else
683 if (slirp_init_dns_list(pData) < 0)
684#endif
685 fNATfailed = 1;
686#ifdef VBOX_WITH_SLIRP_DNS_PROXY
687 dnsproxy_init(pData);
688#endif
689
690 getouraddr(pData);
691
692#ifdef VBOX_WITH_SLIRP_ALIAS
693 {
694 int flags = 0;
695 pData->proxy_alias = LibAliasInit(pData, NULL);
696 if (pData->proxy_alias == NULL)
697 {
698 LogRel(("NAT: LibAlias default rule wasn't initialized\n"));
699 AssertMsgFailed(("NAT: LibAlias default rule wasn't initialized\n"));
700 }
701 flags = LibAliasSetMode(pData->proxy_alias, 0, 0);
702 flags |= PKT_ALIAS_LOG; /* set logging */
703 flags = LibAliasSetMode(pData->proxy_alias, flags, ~0);
704 LibAliasSetAddress(pData->proxy_alias, special_addr);
705 ftp_alias_load();
706 nbt_alias_load();
707
708 }
709#endif
710 return fNATfailed ? VINF_NAT_DNS : VINF_SUCCESS;
711}
712
713/**
714 * Register statistics.
715 */
716void slirp_register_statistics(PNATState pData, PPDMDRVINS pDrvIns)
717{
718#ifdef VBOX_WITH_STATISTICS
719# define COUNTER(name, type, units, dsc) \
720 do { \
721 PDMDrvHlpSTAMRegisterF(pDrvIns, \
722 &pData->Stat ## name, \
723 type, \
724 STAMVISIBILITY_ALWAYS, \
725 units, \
726 dsc, \
727 "/Drivers/NAT%u/" #name, \
728 pDrvIns->iInstance); \
729 } while (0)
730
731# define PROFILE_COUNTER(name, dsc) COUNTER(name, STAMTYPE_PROFILE, STAMUNIT_TICKS_PER_CALL, dsc)
732# define COUNTING_COUNTER(name, dsc) COUNTER(name, STAMTYPE_COUNTER, STAMUNIT_COUNT, dsc)
733
734# include "counters.h"
735
736# undef COUNTER
737# undef PROFILE_COUNTER
738# undef COUNTING_COUNTER
739/** @todo register statistics for the variables dumped by:
740 * ipstats(pData); tcpstats(pData); udpstats(pData); icmpstats(pData);
741 * mbufstats(pData); sockstats(pData); */
742#endif /* VBOX_WITH_STATISTICS */
743}
744
745/**
746 * Deregister statistics.
747 */
748void slirp_deregister_statistics(PNATState pData, PPDMDRVINS pDrvIns)
749{
750#ifdef VBOX_WITH_STATISTICS
751# define PROFILE_COUNTER(name, dsc) PDMDrvHlpSTAMDeregister(pDrvIns, &pData->Stat ## name)
752# define COUNTING_COUNTER(name, dsc) PDMDrvHlpSTAMDeregister(pDrvIns, &pData->Stat ## name)
753
754# include "counters.h"
755
756# undef COUNTING_COUNTER
757# undef PROFILE_COUNTER
758#endif /* VBOX_WITH_STATISTICS */
759}
760
761/**
762 * Marks the link as up, making it possible to establish new connections.
763 */
764void slirp_link_up(PNATState pData)
765{
766 link_up = 1;
767}
768
769/**
770 * Marks the link as down and cleans up the current connections.
771 */
772void slirp_link_down(PNATState pData)
773{
774 struct socket *so;
775
776 while ((so = tcb.so_next) != &tcb)
777 {
778 if (so->so_state & SS_NOFDREF || so->s == -1)
779 sofree(pData, so);
780 else
781 tcp_drop(pData, sototcpcb(so), 0);
782 }
783
784 while ((so = udb.so_next) != &udb)
785 udp_detach(pData, so);
786
787 link_up = 0;
788}
789
790/**
791 * Terminates the slirp component.
792 */
793void slirp_term(PNATState pData)
794{
795#ifndef VBOX_WITH_MULTI_DNS
796 if (pData->pszDomain)
797 RTStrFree((char *)(void *)pData->pszDomain);
798#endif
799
800#ifdef RT_OS_WINDOWS
801 pData->pfIcmpCloseHandle(pData->icmp_socket.sh);
802 FreeLibrary(pData->hmIcmpLibrary);
803 RTMemFree(pData->pvIcmpBuffer);
804# else
805 closesocket(pData->icmp_socket.s);
806#endif
807
808 slirp_link_down(pData);
809#ifdef VBOX_WITH_MULTI_DNS
810 slirp_release_dns_list(pData);
811#endif
812#ifdef RT_OS_WINDOWS
813 WSACleanup();
814#endif
815#ifdef VBOX_WITH_SLIRP_ALIAS
816 while(!LIST_EMPTY(&instancehead)) {
817 struct libalias *la = LIST_FIRST(&instancehead);
818 /* libalias do all clean up */
819 LibAliasUninit(la);
820 }
821#endif
822#ifdef LOG_ENABLED
823 Log(("\n"
824 "NAT statistics\n"
825 "--------------\n"
826 "\n"));
827 ipstats(pData);
828 tcpstats(pData);
829 udpstats(pData);
830 icmpstats(pData);
831 mbufstats(pData);
832 sockstats(pData);
833 Log(("\n"
834 "\n"
835 "\n"));
836#endif
837 RTMemFree(pData);
838}
839
840
841#define CONN_CANFSEND(so) (((so)->so_state & (SS_FCANTSENDMORE|SS_ISFCONNECTED)) == SS_ISFCONNECTED)
842#define CONN_CANFRCV(so) (((so)->so_state & (SS_FCANTRCVMORE|SS_ISFCONNECTED)) == SS_ISFCONNECTED)
843#define UPD_NFDS(x) do { if (nfds < (x)) nfds = (x); } while (0)
844
845/*
846 * curtime kept to an accuracy of 1ms
847 */
848static void updtime(PNATState pData)
849{
850#ifdef RT_OS_WINDOWS
851 struct _timeb tb;
852
853 _ftime(&tb);
854 curtime = (u_int)tb.time * (u_int)1000;
855 curtime += (u_int)tb.millitm;
856#else
857 gettimeofday(&tt, 0);
858
859 curtime = (u_int)tt.tv_sec * (u_int)1000;
860 curtime += (u_int)tt.tv_usec / (u_int)1000;
861
862 if ((tt.tv_usec % 1000) >= 500)
863 curtime++;
864#endif
865}
866
867#ifdef RT_OS_WINDOWS
868void slirp_select_fill(PNATState pData, int *pnfds)
869#else /* RT_OS_WINDOWS */
870void slirp_select_fill(PNATState pData, int *pnfds, struct pollfd *polls)
871#endif /* !RT_OS_WINDOWS */
872{
873 struct socket *so, *so_next;
874 int nfds;
875#if defined(RT_OS_WINDOWS)
876 int rc;
877 int error;
878#else
879 int poll_index = 0;
880#endif
881 int i;
882
883 STAM_PROFILE_START(&pData->StatFill, a);
884
885 nfds = *pnfds;
886
887 /*
888 * First, TCP sockets
889 */
890 do_slowtimo = 0;
891 if (!link_up)
892 goto done;
893 /*
894 * *_slowtimo needs calling if there are IP fragments
895 * in the fragment queue, or there are TCP connections active
896 */
897 /* XXX:
898 * triggering of fragment expiration should be the same but use new macroses
899 */
900 do_slowtimo = (tcb.so_next != &tcb);
901 if (!do_slowtimo)
902 {
903 for (i = 0; i < IPREASS_NHASH; i++)
904 {
905 if (!TAILQ_EMPTY(&ipq[i]))
906 {
907 do_slowtimo = 1;
908 break;
909 }
910 }
911 }
912 ICMP_ENGAGE_EVENT(&pData->icmp_socket, readfds);
913
914 STAM_COUNTER_RESET(&pData->StatTCP);
915 STAM_COUNTER_RESET(&pData->StatTCPHot);
916
917 QSOCKET_FOREACH(so, so_next, tcp)
918 /* { */
919#if !defined(RT_OS_WINDOWS)
920 so->so_poll_index = -1;
921#endif
922 STAM_COUNTER_INC(&pData->StatTCP);
923
924 /*
925 * See if we need a tcp_fasttimo
926 */
927 if ( time_fasttimo == 0
928 && so->so_tcpcb != NULL
929 && so->so_tcpcb->t_flags & TF_DELACK)
930 time_fasttimo = curtime; /* Flag when we want a fasttimo */
931
932 /*
933 * NOFDREF can include still connecting to local-host,
934 * newly socreated() sockets etc. Don't want to select these.
935 */
936 if (so->so_state & SS_NOFDREF || so->s == -1)
937 CONTINUE(tcp);
938
939 /*
940 * Set for reading sockets which are accepting
941 */
942 if (so->so_state & SS_FACCEPTCONN)
943 {
944 STAM_COUNTER_INC(&pData->StatTCPHot);
945 TCP_ENGAGE_EVENT1(so, readfds);
946 CONTINUE(tcp);
947 }
948
949 /*
950 * Set for writing sockets which are connecting
951 */
952 if (so->so_state & SS_ISFCONNECTING)
953 {
954 Log2(("connecting %R[natsock] engaged\n",so));
955 STAM_COUNTER_INC(&pData->StatTCPHot);
956 TCP_ENGAGE_EVENT1(so, writefds);
957 }
958
959 /*
960 * Set for writing if we are connected, can send more, and
961 * we have something to send
962 */
963 if (CONN_CANFSEND(so) && so->so_rcv.sb_cc)
964 {
965 STAM_COUNTER_INC(&pData->StatTCPHot);
966 TCP_ENGAGE_EVENT1(so, writefds);
967 }
968
969 /*
970 * Set for reading (and urgent data) if we are connected, can
971 * receive more, and we have room for it XXX /2 ?
972 */
973 if (CONN_CANFRCV(so) && (so->so_snd.sb_cc < (so->so_snd.sb_datalen/2)))
974 {
975 STAM_COUNTER_INC(&pData->StatTCPHot);
976 TCP_ENGAGE_EVENT2(so, readfds, xfds);
977 }
978 LOOP_LABEL(tcp, so, so_next);
979 }
980
981 /*
982 * UDP sockets
983 */
984 STAM_COUNTER_RESET(&pData->StatUDP);
985 STAM_COUNTER_RESET(&pData->StatUDPHot);
986
987 QSOCKET_FOREACH(so, so_next, udp)
988 /* { */
989
990 STAM_COUNTER_INC(&pData->StatUDP);
991#if !defined(RT_OS_WINDOWS)
992 so->so_poll_index = -1;
993#endif
994
995 /*
996 * See if it's timed out
997 */
998 if (so->so_expire)
999 {
1000 if (so->so_expire <= curtime)
1001 {
1002#ifdef VBOX_WITH_SLIRP_DNS_PROXY
1003 Log2(("NAT: %R[natsock] expired\n", so));
1004 if (so->so_timeout != NULL)
1005 {
1006 so->so_timeout(pData, so, so->so_timeout_arg);
1007 }
1008#endif
1009#ifdef VBOX_WITH_SLIRP_MT
1010 /* we need so_next for continue our cycle*/
1011 so_next = so->so_next;
1012#endif
1013 UDP_DETACH(pData, so, so_next);
1014 CONTINUE_NO_UNLOCK(udp);
1015 }
1016 else
1017 do_slowtimo = 1; /* Let socket expire */
1018 }
1019
1020 /*
1021 * When UDP packets are received from over the link, they're
1022 * sendto()'d straight away, so no need for setting for writing
1023 * Limit the number of packets queued by this session to 4.
1024 * Note that even though we try and limit this to 4 packets,
1025 * the session could have more queued if the packets needed
1026 * to be fragmented.
1027 *
1028 * (XXX <= 4 ?)
1029 */
1030 if ((so->so_state & SS_ISFCONNECTED) && so->so_queued <= 4)
1031 {
1032 STAM_COUNTER_INC(&pData->StatUDPHot);
1033 UDP_ENGAGE_EVENT(so, readfds);
1034 }
1035 LOOP_LABEL(udp, so, so_next);
1036 }
1037done:
1038
1039#if defined(RT_OS_WINDOWS)
1040 *pnfds = VBOX_EVENT_COUNT;
1041#else /* RT_OS_WINDOWS */
1042 AssertRelease(poll_index <= *pnfds);
1043 *pnfds = poll_index;
1044#endif /* !RT_OS_WINDOWS */
1045
1046 STAM_PROFILE_STOP(&pData->StatFill, a);
1047}
1048
1049#if defined(RT_OS_WINDOWS)
1050void slirp_select_poll(PNATState pData, int fTimeout, int fIcmp)
1051#else /* RT_OS_WINDOWS */
1052void slirp_select_poll(PNATState pData, struct pollfd *polls, int ndfs)
1053#endif /* !RT_OS_WINDOWS */
1054{
1055 struct socket *so, *so_next;
1056 int ret;
1057#if defined(RT_OS_WINDOWS)
1058 WSANETWORKEVENTS NetworkEvents;
1059 int rc;
1060 int error;
1061#else
1062 int poll_index = 0;
1063#endif
1064
1065 STAM_PROFILE_START(&pData->StatPoll, a);
1066
1067 /* Update time */
1068 updtime(pData);
1069
1070 /*
1071 * See if anything has timed out
1072 */
1073 if (link_up)
1074 {
1075 if (time_fasttimo && ((curtime - time_fasttimo) >= 2))
1076 {
1077 STAM_PROFILE_START(&pData->StatFastTimer, a);
1078 tcp_fasttimo(pData);
1079 time_fasttimo = 0;
1080 STAM_PROFILE_STOP(&pData->StatFastTimer, a);
1081 }
1082 if (do_slowtimo && ((curtime - last_slowtimo) >= 499))
1083 {
1084 STAM_PROFILE_START(&pData->StatSlowTimer, a);
1085 ip_slowtimo(pData);
1086 tcp_slowtimo(pData);
1087 last_slowtimo = curtime;
1088 STAM_PROFILE_STOP(&pData->StatSlowTimer, a);
1089 }
1090 }
1091#if defined(RT_OS_WINDOWS)
1092 if (fTimeout)
1093 return; /* only timer update */
1094#endif
1095
1096 /*
1097 * Check sockets
1098 */
1099 if (!link_up)
1100 goto done;
1101#if defined(RT_OS_WINDOWS)
1102 /*XXX: before renaming please make see define
1103 * fIcmp in slirp_state.h
1104 */
1105 if (fIcmp)
1106 sorecvfrom(pData, &pData->icmp_socket);
1107#else
1108 if ( (pData->icmp_socket.s != -1)
1109 && CHECK_FD_SET(&pData->icmp_socket, ignored, readfds))
1110 sorecvfrom(pData, &pData->icmp_socket);
1111#endif
1112 /*
1113 * Check TCP sockets
1114 */
1115 QSOCKET_FOREACH(so, so_next, tcp)
1116 /* { */
1117
1118#ifdef VBOX_WITH_SLIRP_MT
1119 if ( so->so_state & SS_NOFDREF
1120 && so->so_deleted == 1)
1121 {
1122 struct socket *son, *sop = NULL;
1123 QSOCKET_LOCK(tcb);
1124 if (so->so_next != NULL)
1125 {
1126 if (so->so_next != &tcb)
1127 SOCKET_LOCK(so->so_next);
1128 son = so->so_next;
1129 }
1130 if ( so->so_prev != &tcb
1131 && so->so_prev != NULL)
1132 {
1133 SOCKET_LOCK(so->so_prev);
1134 sop = so->so_prev;
1135 }
1136 QSOCKET_UNLOCK(tcb);
1137 remque(pData, so);
1138 NSOCK_DEC();
1139 SOCKET_UNLOCK(so);
1140 SOCKET_LOCK_DESTROY(so);
1141 RTMemFree(so);
1142 so_next = son;
1143 if (sop != NULL)
1144 SOCKET_UNLOCK(sop);
1145 CONTINUE_NO_UNLOCK(tcp);
1146 }
1147#endif
1148 /*
1149 * FD_ISSET is meaningless on these sockets
1150 * (and they can crash the program)
1151 */
1152 if (so->so_state & SS_NOFDREF || so->s == -1)
1153 CONTINUE(tcp);
1154
1155 POLL_TCP_EVENTS(rc, error, so, &NetworkEvents);
1156
1157 LOG_NAT_SOCK(so, TCP, &NetworkEvents, readfds, writefds, xfds);
1158
1159
1160 /*
1161 * Check for URG data
1162 * This will soread as well, so no need to
1163 * test for readfds below if this succeeds
1164 */
1165
1166 /* out-of-band data */
1167 if (CHECK_FD_SET(so, NetworkEvents, xfds))
1168 {
1169 sorecvoob(pData, so);
1170 }
1171
1172 /*
1173 * Check sockets for reading
1174 */
1175 else if ( CHECK_FD_SET(so, NetworkEvents, readfds)
1176 || WIN_CHECK_FD_SET(so, NetworkEvents, acceptds))
1177 {
1178 /*
1179 * Check for incoming connections
1180 */
1181 if (so->so_state & SS_FACCEPTCONN)
1182 {
1183 TCP_CONNECT(pData, so);
1184#if defined(RT_OS_WINDOWS)
1185 if (!(NetworkEvents.lNetworkEvents & FD_CLOSE))
1186#endif
1187 CONTINUE(tcp);
1188 }
1189
1190 ret = soread(pData, so);
1191 /* Output it if we read something */
1192 if (RT_LIKELY(ret > 0))
1193 TCP_OUTPUT(pData, sototcpcb(so));
1194 }
1195
1196#if defined(RT_OS_WINDOWS)
1197 /*
1198 * Check for FD_CLOSE events.
1199 * in some cases once FD_CLOSE engaged on socket it could be flashed latter (for some reasons)
1200 */
1201 if ( (NetworkEvents.lNetworkEvents & FD_CLOSE)
1202 || (so->so_close == 1))
1203 {
1204 so->so_close = 1; /* mark it */
1205 /*
1206 * drain the socket
1207 */
1208 for (;;)
1209 {
1210 ret = soread(pData, so);
1211 if (ret > 0)
1212 TCP_OUTPUT(pData, sototcpcb(so));
1213 else
1214 break;
1215 }
1216 CONTINUE(tcp);
1217 }
1218#endif
1219
1220 /*
1221 * Check sockets for writing
1222 */
1223 if (CHECK_FD_SET(so, NetworkEvents, writefds))
1224 {
1225 /*
1226 * Check for non-blocking, still-connecting sockets
1227 */
1228 if (so->so_state & SS_ISFCONNECTING)
1229 {
1230 Log2(("connecting %R[natsock] catched\n", so));
1231 /* Connected */
1232 so->so_state &= ~SS_ISFCONNECTING;
1233
1234 /*
1235 * This should be probably guarded by PROBE_CONN too. Anyway,
1236 * we disable it on OS/2 because the below send call returns
1237 * EFAULT which causes the opened TCP socket to close right
1238 * after it has been opened and connected.
1239 */
1240#ifndef RT_OS_OS2
1241 ret = send(so->s, (const char *)&ret, 0, 0);
1242 if (ret < 0)
1243 {
1244 /* XXXXX Must fix, zero bytes is a NOP */
1245 if ( errno == EAGAIN
1246 || errno == EWOULDBLOCK
1247 || errno == EINPROGRESS
1248 || errno == ENOTCONN)
1249 CONTINUE(tcp);
1250
1251 /* else failed */
1252 so->so_state = SS_NOFDREF;
1253 }
1254 /* else so->so_state &= ~SS_ISFCONNECTING; */
1255#endif
1256
1257 /*
1258 * Continue tcp_input
1259 */
1260 TCP_INPUT(pData, (struct mbuf *)NULL, sizeof(struct ip), so);
1261 /* continue; */
1262 }
1263 else
1264 SOWRITE(ret, pData, so);
1265 /*
1266 * XXX If we wrote something (a lot), there could be the need
1267 * for a window update. In the worst case, the remote will send
1268 * a window probe to get things going again.
1269 */
1270 }
1271
1272 /*
1273 * Probe a still-connecting, non-blocking socket
1274 * to check if it's still alive
1275 */
1276#ifdef PROBE_CONN
1277 if (so->so_state & SS_ISFCONNECTING)
1278 {
1279 ret = recv(so->s, (char *)&ret, 0, 0);
1280
1281 if (ret < 0)
1282 {
1283 /* XXX */
1284 if ( errno == EAGAIN
1285 || errno == EWOULDBLOCK
1286 || errno == EINPROGRESS
1287 || errno == ENOTCONN)
1288 {
1289 CONTINUE(tcp); /* Still connecting, continue */
1290 }
1291
1292 /* else failed */
1293 so->so_state = SS_NOFDREF;
1294
1295 /* tcp_input will take care of it */
1296 }
1297 else
1298 {
1299 ret = send(so->s, &ret, 0, 0);
1300 if (ret < 0)
1301 {
1302 /* XXX */
1303 if ( errno == EAGAIN
1304 || errno == EWOULDBLOCK
1305 || errno == EINPROGRESS
1306 || errno == ENOTCONN)
1307 {
1308 CONTINUE(tcp);
1309 }
1310 /* else failed */
1311 so->so_state = SS_NOFDREF;
1312 }
1313 else
1314 so->so_state &= ~SS_ISFCONNECTING;
1315
1316 }
1317 TCP_INPUT((struct mbuf *)NULL, sizeof(struct ip),so);
1318 } /* SS_ISFCONNECTING */
1319#endif
1320#ifndef RT_OS_WINDOWS
1321 if ( UNIX_CHECK_FD_SET(so, NetworkEvents, rdhup)
1322 || UNIX_CHECK_FD_SET(so, NetworkEvents, rderr))
1323 {
1324 int err;
1325 int inq, outq;
1326 int status;
1327 socklen_t optlen = sizeof(int);
1328 inq = outq = 0;
1329 status = getsockopt(so->s, SOL_SOCKET, SO_ERROR, &err, &optlen);
1330 if (status != 0)
1331 Log(("NAT: can't get error status from %R[natsock]\n", so));
1332#ifndef RT_OS_SOLARIS
1333 status = ioctl(so->s, FIONREAD, &inq); /* tcp(7) recommends SIOCINQ which is Linux specific */
1334 if (status != 0 || status != EINVAL)
1335 {
1336 /* EINVAL returned if socket in listen state tcp(7)*/
1337 Log(("NAT: can't get depth of IN queue status from %R[natsock]\n", so));
1338 }
1339 status = ioctl(so->s, TIOCOUTQ, &outq); /* SIOCOUTQ see previous comment */
1340 if (status != 0)
1341 Log(("NAT: can't get depth of OUT queue from %R[natsock]\n", so));
1342#else
1343 /*
1344 * Solaris has bit different ioctl commands and its handlings
1345 * hint: streamio(7) I_NREAD
1346 */
1347#endif
1348 if ( so->so_state & SS_ISFCONNECTING
1349 || UNIX_CHECK_FD_SET(so, NetworkEvents, readfds))
1350 {
1351 /**
1352 * Check if we need here take care about gracefull connection
1353 * @todo try with proxy server
1354 */
1355 if (UNIX_CHECK_FD_SET(so, NetworkEvents, readfds))
1356 {
1357 /*
1358 * Never meet inq != 0 or outq != 0, anyway let it stay for a while
1359 * in case it happens we'll able to detect it.
1360 * Give TCP/IP stack wait or expire the socket.
1361 */
1362 Log(("NAT: %R[natsock] err(%d:%s) s(in:%d,out:%d)happens on read I/O, "
1363 "other side close connection \n", so, err, strerror(err), inq, outq));
1364 CONTINUE(tcp);
1365 }
1366 goto tcp_input_close;
1367 }
1368 if ( !UNIX_CHECK_FD_SET(so, NetworkEvents, readfds)
1369 && !UNIX_CHECK_FD_SET(so, NetworkEvents, writefds)
1370 && !UNIX_CHECK_FD_SET(so, NetworkEvents, xfds))
1371 {
1372 Log(("NAT: system expires the socket %R[natsock] err(%d:%s) s(in:%d,out:%d) happens on non-I/O. ",
1373 so, err, strerror(err), inq, outq));
1374 goto tcp_input_close;
1375 }
1376 Log(("NAT: %R[natsock] we've met(%d:%s) s(in:%d, out:%d) unhandled combination hup (%d) "
1377 "rederr(%d) on (r:%d, w:%d, x:%d)\n",
1378 so, err, strerror(err),
1379 inq, outq,
1380 UNIX_CHECK_FD_SET(so, ign, rdhup),
1381 UNIX_CHECK_FD_SET(so, ign, rderr),
1382 UNIX_CHECK_FD_SET(so, ign, readfds),
1383 UNIX_CHECK_FD_SET(so, ign, writefds),
1384 UNIX_CHECK_FD_SET(so, ign, xfds)));
1385 /*
1386 * Give OS's TCP/IP stack a chance to resolve an issue or expire the socket.
1387 */
1388 CONTINUE(tcp);
1389tcp_input_close:
1390 so->so_state = SS_NOFDREF; /*cause connection valid tcp connection termination and socket closing */
1391 TCP_INPUT(pData, (struct mbuf *)NULL, sizeof(struct ip), so);
1392 CONTINUE(tcp);
1393 }
1394#endif
1395 LOOP_LABEL(tcp, so, so_next);
1396 }
1397
1398 /*
1399 * Now UDP sockets.
1400 * Incoming packets are sent straight away, they're not buffered.
1401 * Incoming UDP data isn't buffered either.
1402 */
1403 QSOCKET_FOREACH(so, so_next, udp)
1404 /* { */
1405#ifdef VBOX_WITH_SLIRP_MT
1406 if ( so->so_state & SS_NOFDREF
1407 && so->so_deleted == 1)
1408 {
1409 struct socket *son, *sop = NULL;
1410 QSOCKET_LOCK(udb);
1411 if (so->so_next != NULL)
1412 {
1413 if (so->so_next != &udb)
1414 SOCKET_LOCK(so->so_next);
1415 son = so->so_next;
1416 }
1417 if ( so->so_prev != &udb
1418 && so->so_prev != NULL)
1419 {
1420 SOCKET_LOCK(so->so_prev);
1421 sop = so->so_prev;
1422 }
1423 QSOCKET_UNLOCK(udb);
1424 remque(pData, so);
1425 NSOCK_DEC();
1426 SOCKET_UNLOCK(so);
1427 SOCKET_LOCK_DESTROY(so);
1428 RTMemFree(so);
1429 so_next = son;
1430 if (sop != NULL)
1431 SOCKET_UNLOCK(sop);
1432 CONTINUE_NO_UNLOCK(udp);
1433 }
1434#endif
1435 POLL_UDP_EVENTS(rc, error, so, &NetworkEvents);
1436
1437 LOG_NAT_SOCK(so, UDP, &NetworkEvents, readfds, writefds, xfds);
1438
1439 if (so->s != -1 && CHECK_FD_SET(so, NetworkEvents, readfds))
1440 {
1441 SORECVFROM(pData, so);
1442 }
1443 LOOP_LABEL(udp, so, so_next);
1444 }
1445
1446done:
1447#ifndef VBOX_WITH_SLIRP_MT
1448 /*
1449 * See if we can start outputting
1450 */
1451 if (if_queued && link_up)
1452 if_start(pData);
1453#endif
1454
1455 STAM_PROFILE_STOP(&pData->StatPoll, a);
1456}
1457
1458#ifndef VBOX_WITHOUT_SLIRP_CLIENT_ETHER
1459#define ETH_ALEN 6
1460#define ETH_HLEN 14
1461
1462#define ARPOP_REQUEST 1 /* ARP request */
1463#define ARPOP_REPLY 2 /* ARP reply */
1464
1465struct ethhdr
1466{
1467 unsigned char h_dest[ETH_ALEN]; /* destination eth addr */
1468 unsigned char h_source[ETH_ALEN]; /* source ether addr */
1469 unsigned short h_proto; /* packet type ID field */
1470};
1471AssertCompileSize(struct ethhdr, 14);
1472#endif
1473
1474struct arphdr
1475{
1476 unsigned short ar_hrd; /* format of hardware address */
1477 unsigned short ar_pro; /* format of protocol address */
1478 unsigned char ar_hln; /* length of hardware address */
1479 unsigned char ar_pln; /* length of protocol address */
1480 unsigned short ar_op; /* ARP opcode (command) */
1481
1482 /*
1483 * Ethernet looks like this : This bit is variable sized however...
1484 */
1485 unsigned char ar_sha[ETH_ALEN]; /* sender hardware address */
1486 unsigned char ar_sip[4]; /* sender IP address */
1487 unsigned char ar_tha[ETH_ALEN]; /* target hardware address */
1488 unsigned char ar_tip[4]; /* target IP address */
1489};
1490AssertCompileSize(struct arphdr, 28);
1491
1492static void arp_input(PNATState pData, struct mbuf *m)
1493{
1494 struct ethhdr *eh;
1495 struct ethhdr *reh;
1496 struct arphdr *ah;
1497 struct arphdr *rah;
1498 int ar_op;
1499 struct ex_list *ex_ptr;
1500 uint32_t htip;
1501 uint32_t tip;
1502 struct mbuf *mr;
1503 eh = mtod(m, struct ethhdr *);
1504 ah = (struct arphdr *)&eh[1];
1505 htip = ntohl(*(uint32_t*)ah->ar_tip);
1506 tip = *(uint32_t*)ah->ar_tip;
1507
1508 mr = m_get(pData);
1509#ifdef VBOX_WITHOUT_SLIRP_CLIENT_ETHER
1510 reh = mtod(mr, struct ethhdr *);
1511 memcpy(reh->h_source, eh->h_source, ETH_ALEN); /* XXX: if_encap will swap src and dst*/
1512 Log4(("NAT: arp:%R[ether]->%R[ether]\n",
1513 reh->h_source, reh->h_dest));
1514 Log4(("NAT: arp: %R[IP4]\n", &tip));
1515#endif
1516 mr->m_data += if_maxlinkhdr;
1517 mr->m_len = sizeof(struct arphdr);
1518 rah = mtod(mr, struct arphdr *);
1519
1520 ar_op = ntohs(ah->ar_op);
1521 switch(ar_op)
1522 {
1523 case ARPOP_REQUEST:
1524#ifdef VBOX_WITH_NAT_SERVICE
1525 if (tip == special_addr.s_addr) goto arp_ok;
1526#endif
1527 if ((htip & pData->netmask) == ntohl(special_addr.s_addr))
1528 {
1529 if ( CTL_CHECK(htip, CTL_DNS)
1530 || CTL_CHECK(htip, CTL_ALIAS)
1531 || CTL_CHECK(htip, CTL_TFTP))
1532 goto arp_ok;
1533 for (ex_ptr = exec_list; ex_ptr; ex_ptr = ex_ptr->ex_next)
1534 {
1535 if ((htip & ~pData->netmask) == ex_ptr->ex_addr)
1536 {
1537 goto arp_ok;
1538 }
1539 }
1540 return;
1541 arp_ok:
1542 rah->ar_hrd = htons(1);
1543 rah->ar_pro = htons(ETH_P_IP);
1544 rah->ar_hln = ETH_ALEN;
1545 rah->ar_pln = 4;
1546 rah->ar_op = htons(ARPOP_REPLY);
1547 memcpy(rah->ar_sha, special_ethaddr, ETH_ALEN);
1548
1549 switch (htip & ~pData->netmask)
1550 {
1551 case CTL_DNS:
1552 case CTL_ALIAS:
1553 rah->ar_sha[5] = (uint8_t)(htip & ~pData->netmask);
1554 break;
1555 default:;
1556 }
1557
1558 memcpy(rah->ar_sip, ah->ar_tip, 4);
1559 memcpy(rah->ar_tha, ah->ar_sha, ETH_ALEN);
1560 memcpy(rah->ar_tip, ah->ar_sip, 4);
1561 if_encap(pData, ETH_P_ARP, mr);
1562 m_free(pData, m);
1563 }
1564 break;
1565 default:
1566 break;
1567 }
1568}
1569
1570void slirp_input(PNATState pData, const uint8_t *pkt, int pkt_len)
1571{
1572 struct mbuf *m;
1573 int proto;
1574 static bool fWarnedIpv6;
1575 struct ethhdr *eh = (struct ethhdr*)pkt;
1576
1577 Log2(("NAT: slirp_input %d\n", pkt_len));
1578 if (pkt_len < ETH_HLEN)
1579 {
1580 LogRel(("NAT: packet having size %d has been ingnored\n", pkt_len));
1581 return;
1582 }
1583 Log4(("NAT: in:%R[ether]->%R[ether]\n", &eh->h_source, &eh->h_dest));
1584#ifdef VBOX_WITHOUT_SLIRP_CLIENT_ETHER
1585 if (memcmp(eh->h_source, special_ethaddr, ETH_ALEN) == 0)
1586 {
1587 /* @todo vasily: add ether logging routine in debug.c */
1588 Log(("NAT: packet was addressed to other MAC\n"));
1589 RTMemFree((void *)pkt);
1590 return;
1591 }
1592#endif
1593
1594 m = m_get(pData);
1595 if (!m)
1596 {
1597 LogRel(("NAT: can't allocate new mbuf\n"));
1598 return;
1599 }
1600
1601 /* Note: we add to align the IP header */
1602
1603 if (M_FREEROOM(m) < pkt_len)
1604 m_inc(m, pkt_len);
1605
1606 m->m_len = pkt_len ;
1607 memcpy(m->m_data, pkt, pkt_len);
1608
1609 proto = ntohs(*(uint16_t *)(pkt + 12));
1610 switch(proto)
1611 {
1612 case ETH_P_ARP:
1613 arp_input(pData, m);
1614 break;
1615 case ETH_P_IP:
1616 /* Update time. Important if the network is very quiet, as otherwise
1617 * the first outgoing connection gets an incorrect timestamp. */
1618 updtime(pData);
1619 m_adj(m, ETH_HLEN);
1620 ip_input(pData, m);
1621 break;
1622 case ETH_P_IPV6:
1623 m_free(pData, m);
1624 if (!fWarnedIpv6)
1625 {
1626 LogRel(("NAT: IPv6 not supported\n"));
1627 fWarnedIpv6 = true;
1628 }
1629 break;
1630 default:
1631 Log(("NAT: Unsupported protocol %x\n", proto));
1632 m_free(pData, m);
1633 break;
1634 }
1635 RTMemFree((void *)pkt);
1636}
1637
1638/* output the IP packet to the ethernet device */
1639void if_encap(PNATState pData, uint16_t eth_proto, struct mbuf *m)
1640{
1641 struct ethhdr *eh;
1642 uint8_t *buf = RTMemAlloc(1600);
1643
1644 m->m_data -= if_maxlinkhdr;
1645 m->m_len += ETH_HLEN;
1646 eh = mtod(m, struct ethhdr *);
1647
1648 if(MBUF_HEAD(m) != m->m_data)
1649 {
1650 LogRel(("NAT: ethernet detects corruption of the packet"));
1651 AssertMsgFailed(("!!Ethernet frame corrupted!!"));
1652 }
1653
1654#ifndef VBOX_WITHOUT_SLIRP_CLIENT_ETHER
1655 memcpy(eh->h_dest, client_ethaddr, ETH_ALEN);
1656 memcpy(eh->h_source, special_ethaddr, ETH_ALEN - 1);
1657 /* XXX: not correct */
1658 eh->h_source[5] = CTL_ALIAS;
1659#else
1660 if (memcmp(eh->h_source, special_ethaddr, ETH_ALEN) != 0)
1661 {
1662 memcpy(eh->h_dest, eh->h_source, ETH_ALEN);
1663 memcpy(eh->h_source, special_ethaddr, ETH_ALEN);
1664 Assert(memcmp(eh->h_dest, special_ethaddr, ETH_ALEN) != 0);
1665 if (memcmp(eh->h_dest, zerro_ethaddr, ETH_ALEN) == 0)
1666 {
1667 /* don't do anything */
1668 goto done;
1669 }
1670 }
1671#endif
1672 eh->h_proto = htons(eth_proto);
1673#if 0
1674 slirp_output(pData->pvUser, m, mtod(m, uint8_t *), m->m_len);
1675#else
1676 memcpy(buf, mtod(m, uint8_t *), m->m_len);
1677 slirp_output(pData->pvUser, NULL, buf, m->m_len);
1678done:
1679 m_free(pData, m);
1680#endif
1681}
1682
1683int slirp_redir(PNATState pData, int is_udp, int host_port,
1684 struct in_addr guest_addr, int guest_port)
1685{
1686 struct socket *so;
1687#ifdef VBOX_WITH_SLIRP_ALIAS
1688 struct alias_link *link;
1689 struct libalias *lib;
1690 int flags;
1691 struct sockaddr sa;
1692 struct sockaddr_in *psin;
1693 socklen_t socketlen;
1694 struct in_addr alias;
1695 int rc;
1696#endif
1697 Log2(("NAT: set redirect %s hp:%d gp:%d\n", (is_udp?"UDP":"TCP"), host_port, guest_port));
1698 if (is_udp)
1699 {
1700 so = udp_listen(pData, htons(host_port), guest_addr.s_addr,
1701 htons(guest_port), 0);
1702 }
1703 else
1704 {
1705 so = solisten(pData, htons(host_port), guest_addr.s_addr,
1706 htons(guest_port), 0);
1707 }
1708#ifndef VBOX_WITH_SLIRP_ALIAS
1709 Log2(("NAT: redirecting socket %R[natsock]\n", so));
1710 return (so != NULL ? 0 : -1);
1711#else
1712
1713 psin = (struct sockaddr_in *)&sa;
1714 psin->sin_family = AF_INET;
1715 psin->sin_port = 0;
1716 psin->sin_addr.s_addr = INADDR_ANY;
1717 socketlen = sizeof(struct sockaddr);
1718
1719 rc = getsockname(so->s, &sa, &socketlen);
1720 if (rc < 0 || sa.sa_family != AF_INET)
1721 {
1722 Log(("NAT: can't get socket's name\n"));
1723 return 1;
1724 }
1725
1726 psin = (struct sockaddr_in *)&sa;
1727
1728#if 1
1729 lib = LibAliasInit(pData, NULL);
1730 flags = LibAliasSetMode(lib, 0, 0);
1731 flags |= PKT_ALIAS_LOG; /* set logging */
1732 flags |= PKT_ALIAS_REVERSE; /* set logging */
1733 flags = LibAliasSetMode(lib, flags, ~0);
1734#else
1735 lib = LIST_FIRST(&instancehead);
1736#endif
1737
1738 alias.s_addr = htonl(ntohl(guest_addr.s_addr) | CTL_ALIAS);
1739 link = LibAliasRedirectPort(lib, psin->sin_addr, htons(host_port),
1740 alias, htons(guest_port),
1741 special_addr, -1, /* not very clear for now*/
1742 (is_udp ? IPPROTO_UDP : IPPROTO_TCP));
1743 if (link == NULL)
1744 {
1745 Log(("NAT: can't create redirect\n"));
1746 return 1;
1747 }
1748 so->so_la = lib;
1749
1750 return 0;
1751#endif
1752}
1753
1754int slirp_add_exec(PNATState pData, int do_pty, const char *args, int addr_low_byte,
1755 int guest_port)
1756{
1757 return add_exec(&exec_list, do_pty, (char *)args,
1758 addr_low_byte, htons(guest_port));
1759}
1760
1761void slirp_set_ethaddr(PNATState pData, const uint8_t *ethaddr)
1762{
1763#ifndef VBOX_WITHOUT_SLIRP_CLIENT_ETHER
1764 memcpy(client_ethaddr, ethaddr, ETH_ALEN);
1765#endif
1766}
1767
1768#if defined(RT_OS_WINDOWS)
1769HANDLE *slirp_get_events(PNATState pData)
1770{
1771 return pData->phEvents;
1772}
1773void slirp_register_external_event(PNATState pData, HANDLE hEvent, int index)
1774{
1775 pData->phEvents[index] = hEvent;
1776}
1777#endif
1778
1779unsigned int slirp_get_timeout_ms(PNATState pData)
1780{
1781 if (link_up)
1782 {
1783 if (time_fasttimo)
1784 return 2;
1785 if (do_slowtimo)
1786 return 500; /* see PR_SLOWHZ */
1787 }
1788 return 0;
1789}
1790
1791#ifndef RT_OS_WINDOWS
1792int slirp_get_nsock(PNATState pData)
1793{
1794 return pData->nsock;
1795}
1796#endif
1797
1798/*
1799 * this function called from NAT thread
1800 */
1801void slirp_post_sent(PNATState pData, void *pvArg)
1802{
1803 struct socket *so = 0;
1804 struct tcpcb *tp = 0;
1805 struct mbuf *m = (struct mbuf *)pvArg;
1806 m_free(pData, m);
1807}
1808#ifdef VBOX_WITH_SLIRP_MT
1809void slirp_process_queue(PNATState pData)
1810{
1811 RTReqProcess(pData->pReqQueue, RT_INDEFINITE_WAIT);
1812}
1813void *slirp_get_queue(PNATState pData)
1814{
1815 return pData->pReqQueue;
1816}
1817#endif
1818
1819uint16_t slirp_get_service(int proto, uint16_t dport, uint16_t sport)
1820{
1821 uint16_t hdport, hsport, service;
1822 hdport = ntohs(dport);
1823 hsport = ntohs(sport);
1824 Log2(("proto: %d, dport: %d sport: %d\n", proto, hdport, hsport));
1825 service = 0;
1826#if 0
1827 /* Always return 0 here */
1828 switch (hdport)
1829 {
1830 case 500:
1831 if (hsport != 500) /* vpnc by default try operate in src:500/dst:500 mode*/
1832 /* Not sure why this make Cisco VPN client's connection more stable,
1833 * at least on some servers
1834 */
1835 service = sport;
1836 break;
1837 }
1838#endif
1839 Log2(("service : %d\n", service));
1840 return htons(service);
1841}
1842
1843void slirp_set_dhcp_TFTP_prefix(PNATState pData, const char *tftpPrefix)
1844{
1845 Log2(("tftp_prefix:%s\n", tftpPrefix));
1846 tftp_prefix = tftpPrefix;
1847}
1848
1849void slirp_set_dhcp_TFTP_bootfile(PNATState pData, const char *bootFile)
1850{
1851 Log2(("bootFile:%s\n", bootFile));
1852 bootp_filename = bootFile;
1853}
1854
1855void slirp_set_dhcp_next_server(PNATState pData, const char *next_server)
1856{
1857 Log2(("next_server:%s\n", next_server));
1858 if (next_server == NULL)
1859 pData->tftp_server.s_addr = htonl(ntohl(special_addr.s_addr) | CTL_TFTP);
1860 else
1861 inet_aton(next_server, &pData->tftp_server);
1862}
1863#ifdef VBOX_WITH_SLIRP_DNS_PROXY
1864void slirp_set_dhcp_dns_proxy(PNATState pData, bool fDNSProxy)
1865{
1866 Log2(("NAT: DNS proxy switched %s\n", (fDNSProxy ? "on" : "off")));
1867 pData->use_dns_proxy = fDNSProxy;
1868}
1869#endif
1870
1871#define CHECK_ARG(name, val, lim_min, lim_max) \
1872do { \
1873 if ((val) < (lim_min) || (val) > (lim_max)) \
1874 { \
1875 LogRel(("NAT: (" #name ":%d) has been ignored, " \
1876 "because out of range (%d, %d)\n", (val), (lim_min), (lim_max))); \
1877 return; \
1878 } \
1879 else \
1880 { \
1881 LogRel(("NAT: (" #name ":%d)\n", (val))); \
1882 } \
1883} while (0)
1884
1885/* don't allow user set less 8kB and more than 1M values */
1886#define _8K_1M_CHECK_ARG(name, val) CHECK_ARG(name, (val), 8, 1024)
1887void slirp_set_rcvbuf(PNATState pData, int kilobytes)
1888{
1889 _8K_1M_CHECK_ARG("SOCKET_RCVBUF", kilobytes);
1890 pData->socket_rcv = kilobytes;
1891}
1892void slirp_set_sndbuf(PNATState pData, int kilobytes)
1893{
1894 _8K_1M_CHECK_ARG("SOCKET_SNDBUF", kilobytes);
1895 pData->socket_snd = kilobytes * _1K;
1896}
1897void slirp_set_tcp_rcvspace(PNATState pData, int kilobytes)
1898{
1899 _8K_1M_CHECK_ARG("TCP_RCVSPACE", kilobytes);
1900 tcp_rcvspace = kilobytes * _1K;
1901}
1902void slirp_set_tcp_sndspace(PNATState pData, int kilobytes)
1903{
1904 _8K_1M_CHECK_ARG("TCP_SNDSPACE", kilobytes);
1905 tcp_sndspace = kilobytes * _1K;
1906}
1907
注意: 瀏覽 TracBrowser 來幫助您使用儲存庫瀏覽器

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