socket.c@ 15522

Last change on this file since 15522 was 15458, checked in by vboxsync, 16 years ago
slirp: typo
Property svn:eol-style set to `native`
File size: 28.1 KB

Line
1	/*
2	* Copyright (c) 1995 Danny Gasparovski.
3	*
4	* Please read the file COPYRIGHT for the
5	* terms and conditions of the copyright.
6	*/
7
8	#define WANT_SYS_IOCTL_H
9	#include <slirp.h>
10	#include "ip_icmp.h"
11	#include "main.h"
12	#ifdef __sun__
13	#include <sys/filio.h>
14	#endif
15	#if defined(VBOX_WITH_SLIRP_ICMP) && defined (RT_OS_WINDOWS)
16	#include <iphlpapi.h>
17	#include <icmpapi.h>
18	#endif
19
20
21	#ifdef VBOX_WITH_SLIRP_ICMP
22	static void send_icmp_to_guest(PNATState, char , size_t, struct socket , const struct sockaddr_in *);
23	static void sorecvfrom_icmp_win(PNATState, struct socket *);
24	#endif
25	static void sorecvfrom_icmp_unix(PNATState, struct socket *);
26
27	void
28	so_init()
29	{
30	}
31
32
33	struct socket *
34	solookup(struct socket *head, struct in_addr laddr,
35	u_int lport, struct in_addr faddr, u_int fport)
36	{
37	struct socket *so;
38
39	for (so = head->so_next; so != head; so = so->so_next)
40	{
41	if ( so->so_lport == lport
42	&& so->so_laddr.s_addr == laddr.s_addr
43	&& so->so_faddr.s_addr == faddr.s_addr
44	&& so->so_fport == fport)
45	return so;
46	}
47
48	return (struct socket *)NULL;
49	}
50
51	/*
52	* Create a new socket, initialise the fields
53	* It is the responsibility of the caller to
54	* insque() it into the correct linked-list
55	*/
56	struct socket *
57	socreate()
58	{
59	struct socket *so;
60
61	so = (struct socket *)malloc(sizeof(struct socket));
62	if(so)
63	{
64	memset(so, 0, sizeof(struct socket));
65	so->so_state = SS_NOFDREF;
66	so->s = -1;
67	}
68	return so;
69	}
70
71	/*
72	* remque and free a socket, clobber cache
73	*/
74	void
75	sofree(PNATState pData, struct socket *so)
76	{
77	if (so == tcp_last_so)
78	tcp_last_so = &tcb;
79	else if (so == udp_last_so)
80	udp_last_so = &udb;
81
82	/* check if mbuf haven't been already freed */
83	if (so->so_m != NULL)
84	m_free(pData, so->so_m);
85
86	if(so->so_next && so->so_prev)
87	remque(pData, so); /* crashes if so is not in a queue */
88
89	free(so);
90	}
91
92	/*
93	* Read from so's socket into sb_snd, updating all relevant sbuf fields
94	* NOTE: This will only be called if it is select()ed for reading, so
95	* a read() of 0 (or less) means it's disconnected
96	*/
97	int
98	soread(PNATState pData, struct socket *so, int fCloseIfNothingRead)
99	{
100	int n, nn, lss, total;
101	struct sbuf *sb = &so->so_snd;
102	size_t len = sb->sb_datalen - sb->sb_cc;
103	struct iovec iov[2];
104	int mss = so->so_tcpcb->t_maxseg;
105
106	DEBUG_CALL("soread");
107	DEBUG_ARG("so = %lx", (long )so);
108
109	/*
110	* No need to check if there's enough room to read.
111	* soread wouldn't have been called if there weren't
112	*/
113
114	len = sb->sb_datalen - sb->sb_cc;
115
116	iov[0].iov_base = sb->sb_wptr;
117	iov[1].iov_base = 0;
118	iov[1].iov_len = 0;
119	if (sb->sb_wptr < sb->sb_rptr)
120	{
121	iov[0].iov_len = sb->sb_rptr - sb->sb_wptr;
122	/* Should never succeed, but... */
123	if (iov[0].iov_len > len)
124	iov[0].iov_len = len;
125	if (iov[0].iov_len > mss)
126	iov[0].iov_len -= iov[0].iov_len%mss;
127	n = 1;
128	}
129	else
130	{
131	iov[0].iov_len = (sb->sb_data + sb->sb_datalen) - sb->sb_wptr;
132	/* Should never succeed, but... */
133	if (iov[0].iov_len > len)
134	iov[0].iov_len = len;
135	len -= iov[0].iov_len;
136	if (len)
137	{
138	iov[1].iov_base = sb->sb_data;
139	iov[1].iov_len = sb->sb_rptr - sb->sb_data;
140	if(iov[1].iov_len > len)
141	iov[1].iov_len = len;
142	total = iov[0].iov_len + iov[1].iov_len;
143	if (total > mss)
144	{
145	lss = total % mss;
146	if (iov[1].iov_len > lss)
147	{
148	iov[1].iov_len -= lss;
149	n = 2;
150	}
151	else
152	{
153	lss -= iov[1].iov_len;
154	iov[0].iov_len -= lss;
155	n = 1;
156	}
157	}
158	else
159	n = 2;
160	}
161	else
162	{
163	if (iov[0].iov_len > mss)
164	iov[0].iov_len -= iov[0].iov_len%mss;
165	n = 1;
166	}
167	}
168
169	#ifdef HAVE_READV
170	nn = readv(so->s, (struct iovec *)iov, n);
171	DEBUG_MISC((dfd, " ... read nn = %d bytes\n", nn));
172	#else
173	nn = recv(so->s, iov[0].iov_base, iov[0].iov_len,0);
174	#endif
175	if (nn <= 0)
176	{
177	#if defined(VBOX_WITH_SIMPLIFIED_SLIRP_SYNC) && defined(RT_OS_WINDOWS)
178	/*
179	* Special case for WSAEnumNetworkEvents: If we receive 0 bytes that
180	* _could_ mean that the connection is closed. But we will receive an
181	* FD_CLOSE event later if the connection was _really_ closed. With
182	* www.youtube.com I see this very often. Closing the socket too early
183	* would be dangerous.
184	*/
185	if (nn == 0 && !fCloseIfNothingRead)
186	return 0;
187	#endif
188	if (nn < 0 && (errno == EINTR \|\| errno == EAGAIN \|\| errno == EWOULDBLOCK))
189	return 0;
190	else
191	{
192	/* nn == 0 means peer has performed an orderly shutdown */
193	DEBUG_MISC((dfd, " --- soread() disconnected, nn = %d, errno = %d-%s\n",
194	nn, errno,strerror(errno)));
195	sofcantrcvmore(so);
196	tcp_sockclosed(pData, sototcpcb(so));
197	return -1;
198	}
199	}
200
201	#ifndef HAVE_READV
202	/*
203	* If there was no error, try and read the second time round
204	* We read again if n = 2 (ie, there's another part of the buffer)
205	* and we read as much as we could in the first read
206	* We don't test for <= 0 this time, because there legitimately
207	* might not be any more data (since the socket is non-blocking),
208	* a close will be detected on next iteration.
209	* A return of -1 wont (shouldn't) happen, since it didn't happen above
210	*/
211	if (n == 2 && nn == iov[0].iov_len)
212	{
213	int ret;
214	ret = recv(so->s, iov[1].iov_base, iov[1].iov_len,0);
215	if (ret > 0)
216	nn += ret;
217	}
218
219	DEBUG_MISC((dfd, " ... read nn = %d bytes\n", nn));
220	#endif
221
222	/* Update fields */
223	sb->sb_cc += nn;
224	sb->sb_wptr += nn;
225	if (sb->sb_wptr >= (sb->sb_data + sb->sb_datalen))
226	sb->sb_wptr -= sb->sb_datalen;
227	return nn;
228	}
229
230	/*
231	* Get urgent data
232	*
233	* When the socket is created, we set it SO_OOBINLINE,
234	* so when OOB data arrives, we soread() it and everything
235	* in the send buffer is sent as urgent data
236	*/
237	void
238	sorecvoob(PNATState pData, struct socket *so)
239	{
240	struct tcpcb *tp = sototcpcb(so);
241
242	DEBUG_CALL("sorecvoob");
243	DEBUG_ARG("so = %lx", (long)so);
244
245	/*
246	* We take a guess at how much urgent data has arrived.
247	* In most situations, when urgent data arrives, the next
248	* read() should get all the urgent data. This guess will
249	* be wrong however if more data arrives just after the
250	* urgent data, or the read() doesn't return all the
251	* urgent data.
252	*/
253	soread(pData, so, /fCloseIfNothingRead=/false);
254	tp->snd_up = tp->snd_una + so->so_snd.sb_cc;
255	tp->t_force = 1;
256	tcp_output(pData, tp);
257	tp->t_force = 0;
258	}
259
260	/*
261	* Send urgent data
262	* There's a lot duplicated code here, but...
263	*/
264	int
265	sosendoob(struct socket *so)
266	{
267	struct sbuf *sb = &so->so_rcv;
268	char buff[2048]; /* XXX Shouldn't be sending more oob data than this */
269
270	int n, len;
271
272	DEBUG_CALL("sosendoob");
273	DEBUG_ARG("so = %lx", (long)so);
274	DEBUG_ARG("sb->sb_cc = %d", sb->sb_cc);
275
276	if (so->so_urgc > sizeof(buff))
277	so->so_urgc = sizeof(buff); /* XXX */
278
279	if (sb->sb_rptr < sb->sb_wptr)
280	{
281	/* We can send it directly */
282	n = send(so->s, sb->sb_rptr, so->so_urgc, (MSG_OOB)); /* \|MSG_DONTWAIT)); */
283	so->so_urgc -= n;
284
285	DEBUG_MISC((dfd, " --- sent %d bytes urgent data, %d urgent bytes left\n",
286	n, so->so_urgc));
287	}
288	else
289	{
290	/*
291	* Since there's no sendv or sendtov like writev,
292	* we must copy all data to a linear buffer then
293	* send it all
294	*/
295	len = (sb->sb_data + sb->sb_datalen) - sb->sb_rptr;
296	if (len > so->so_urgc)
297	len = so->so_urgc;
298	memcpy(buff, sb->sb_rptr, len);
299	so->so_urgc -= len;
300	if (so->so_urgc)
301	{
302	n = sb->sb_wptr - sb->sb_data;
303	if (n > so->so_urgc)
304	n = so->so_urgc;
305	memcpy(buff + len, sb->sb_data, n);
306	so->so_urgc -= n;
307	len += n;
308	}
309	n = send(so->s, buff, len, (MSG_OOB)); /* \|MSG_DONTWAIT)); */
310	#ifdef DEBUG
311	if (n != len)
312	DEBUG_ERROR((dfd, "Didn't send all data urgently XXXXX\n"));
313	#endif
314	DEBUG_MISC((dfd, " ---2 sent %d bytes urgent data, %d urgent bytes left\n",
315	n, so->so_urgc));
316	}
317
318	sb->sb_cc -= n;
319	sb->sb_rptr += n;
320	if (sb->sb_rptr >= (sb->sb_data + sb->sb_datalen))
321	sb->sb_rptr -= sb->sb_datalen;
322
323	return n;
324	}
325
326	/*
327	* Write data from so_rcv to so's socket,
328	* updating all sbuf field as necessary
329	*/
330	int
331	sowrite(PNATState pData, struct socket *so)
332	{
333	int n,nn;
334	struct sbuf *sb = &so->so_rcv;
335	size_t len = sb->sb_cc;
336	struct iovec iov[2];
337
338	DEBUG_CALL("sowrite");
339	DEBUG_ARG("so = %lx", (long)so);
340
341	if (so->so_urgc)
342	{
343	sosendoob(so);
344	if (sb->sb_cc == 0)
345	return 0;
346	}
347
348	/*
349	* No need to check if there's something to write,
350	* sowrite wouldn't have been called otherwise
351	*/
352
353	len = sb->sb_cc;
354
355	iov[0].iov_base = sb->sb_rptr;
356	iov[1].iov_base = 0;
357	iov[1].iov_len = 0;
358	if (sb->sb_rptr < sb->sb_wptr)
359	{
360	iov[0].iov_len = sb->sb_wptr - sb->sb_rptr;
361	/* Should never succeed, but... */
362	if (iov[0].iov_len > len)
363	iov[0].iov_len = len;
364	n = 1;
365	}
366	else
367	{
368	iov[0].iov_len = (sb->sb_data + sb->sb_datalen) - sb->sb_rptr;
369	if (iov[0].iov_len > len)
370	iov[0].iov_len = len;
371	len -= iov[0].iov_len;
372	if (len)
373	{
374	iov[1].iov_base = sb->sb_data;
375	iov[1].iov_len = sb->sb_wptr - sb->sb_data;
376	if (iov[1].iov_len > len)
377	iov[1].iov_len = len;
378	n = 2;
379	}
380	else
381	n = 1;
382	}
383	/* Check if there's urgent data to send, and if so, send it */
384	#ifdef HAVE_READV
385	nn = writev(so->s, (const struct iovec *)iov, n);
386	DEBUG_MISC((dfd, " ... wrote nn = %d bytes\n", nn));
387	#else
388	nn = send(so->s, iov[0].iov_base, iov[0].iov_len, 0);
389	#endif
390	/* This should never happen, but people tell me it does shrug */
391	if (nn < 0 && (errno == EAGAIN \|\| errno == EINTR \|\| errno == EWOULDBLOCK))
392	return 0;
393
394	if (nn < 0 \|\| (nn == 0 && iov[0].iov_len > 0))
395	{
396	DEBUG_MISC((dfd, " --- sowrite disconnected, so->so_state = %x, errno = %d\n",
397	so->so_state, errno));
398	sofcantsendmore(so);
399	tcp_sockclosed(pData, sototcpcb(so));
400	return -1;
401	}
402
403	#ifndef HAVE_READV
404	if (n == 2 && nn == iov[0].iov_len)
405	{
406	int ret;
407	ret = send(so->s, iov[1].iov_base, iov[1].iov_len,0);
408	if (ret > 0)
409	nn += ret;
410	}
411	DEBUG_MISC((dfd, " ... wrote nn = %d bytes\n", nn));
412	#endif
413
414	/* Update sbuf */
415	sb->sb_cc -= nn;
416	sb->sb_rptr += nn;
417	if (sb->sb_rptr >= (sb->sb_data + sb->sb_datalen))
418	sb->sb_rptr -= sb->sb_datalen;
419
420	/*
421	* If in DRAIN mode, and there's no more data, set
422	* it CANTSENDMORE
423	*/
424	if ((so->so_state & SS_FWDRAIN) && sb->sb_cc == 0)
425	sofcantsendmore(so);
426
427	return nn;
428	}
429
430	/*
431	* recvfrom() a UDP socket
432	*/
433	void
434	sorecvfrom(PNATState pData, struct socket *so)
435	{
436	struct sockaddr_in addr;
437	socklen_t addrlen = sizeof(struct sockaddr_in);
438
439	DEBUG_CALL("sorecvfrom");
440	DEBUG_ARG("so = %lx", (long)so);
441
442	if (so->so_type == IPPROTO_ICMP)
443	{
444	/* This is a "ping" reply */
445	#if !defined(VBOX_WITH_SLIRP_ICMP) \|\| (defined(VBOX_WITH_SLIRP_ICMP) && !defined(RT_OS_WINDOWS))
446	sorecvfrom_icmp_unix(pData, so);
447	#endif
448	#if defined(VBOX_WITH_SLIRP_ICMP) && defined(RT_OS_WINDOWS)
449	sorecvfrom_icmp_win(pData, so);
450	#endif
451	udp_detach(pData, so);
452	}
453	else
454	{
455	/* A "normal" UDP packet */
456	struct mbuf *m;
457	size_t len;
458	u_long n;
459
460	if (!(m = m_get(pData)))
461	return;
462	m->m_data += if_maxlinkhdr;
463
464	/*
465	* XXX Shouldn't FIONREAD packets destined for port 53,
466	* but I don't know the max packet size for DNS lookups
467	*/
468	len = M_FREEROOM(m);
469	/* if (so->so_fport != htons(53)) */
470	{
471	ioctlsocket(so->s, FIONREAD, &n);
472
473	if (n > len)
474	{
475	n = (m->m_data - m->m_dat) + m->m_len + n + 1;
476	m_inc(m, n);
477	len = M_FREEROOM(m);
478	}
479	}
480
481	m->m_len = recvfrom(so->s, m->m_data, len, 0,
482	(struct sockaddr *)&addr, &addrlen);
483	DEBUG_MISC((dfd, " did recvfrom %d, errno = %d-%s\n",
484	m->m_len, errno,strerror(errno)));
485	if(m->m_len < 0)
486	{
487	u_char code = ICMP_UNREACH_PORT;
488
489	if (errno == EHOSTUNREACH)
490	code = ICMP_UNREACH_HOST;
491	else if(errno == ENETUNREACH)
492	code = ICMP_UNREACH_NET;
493
494	DEBUG_MISC((dfd," rx error, tx icmp ICMP_UNREACH:%i\n", code));
495	icmp_error(pData, so->so_m, ICMP_UNREACH,code, 0,strerror(errno));
496	m_free(pData, m);
497	}
498	else
499	{
500	/*
501	* Hack: domain name lookup will be used the most for UDP,
502	* and since they'll only be used once there's no need
503	* for the 4 minute (or whatever) timeout... So we time them
504	* out much quicker (10 seconds for now...)
505	*/
506	if (so->so_expire)
507	{
508	if (so->so_fport == htons(53))
509	so->so_expire = curtime + SO_EXPIREFAST;
510	else
511	so->so_expire = curtime + SO_EXPIRE;
512	}
513
514	#if 0
515	if (m->m_len == len)
516	{
517	m_inc(m, MINCSIZE);
518	m->m_len = 0;
519	}
520	#endif
521
522	/*
523	* If this packet was destined for CTL_ADDR,
524	* make it look like that's where it came from, done by udp_output
525	*/
526	udp_output(pData, so, m, &addr);
527	} /* rx error */
528	} /* if ping packet */
529	}
530
531	/*
532	* sendto() a socket
533	*/
534	int
535	sosendto(PNATState pData, struct socket so, struct mbuf m)
536	{
537	int ret;
538	struct sockaddr_in addr;
539	#if 0
540	struct sockaddr_in host_addr;
541	#endif
542
543	DEBUG_CALL("sosendto");
544	DEBUG_ARG("so = %lx", (long)so);
545	DEBUG_ARG("m = %lx", (long)m);
546
547	addr.sin_family = AF_INET;
548	if ((so->so_faddr.s_addr & htonl(pData->netmask)) == special_addr.s_addr)
549	{
550	/* It's an alias */
551	uint32_t last_byte = ntohl(so->so_faddr.s_addr) & ~pData->netmask;
552	switch(last_byte)
553	{
554	#if 0
555	/* handle this case at 'default:' */
556	case CTL_BROADCAST:
557	addr.sin_addr.s_addr = INADDR_BROADCAST;
558	/* Send the packet to host to fully emulate broadcast */
559	/** @todo r=klaus: on Linux host this causes the host to receive
560	* the packet twice for some reason. And I cannot find any place
561	* in the man pages which states that sending a broadcast does not
562	* reach the host itself. */
563	host_addr.sin_family = AF_INET;
564	host_addr.sin_port = so->so_fport;
565	host_addr.sin_addr = our_addr;
566	sendto(so->s, m->m_data, m->m_len, 0,
567	(struct sockaddr *)&host_addr, sizeof (struct sockaddr));
568	break;
569	#endif
570	case CTL_DNS:
571	if (!get_dns_addr(pData, &dns_addr))
572	addr.sin_addr = dns_addr;
573	else
574	addr.sin_addr = loopback_addr;
575	break;
576	case CTL_ALIAS:
577	default:
578	if (last_byte == ~pData->netmask)
579	addr.sin_addr.s_addr = INADDR_BROADCAST;
580	else
581	addr.sin_addr = loopback_addr;
582	break;
583	}
584	}
585	else
586	addr.sin_addr = so->so_faddr;
587	addr.sin_port = so->so_fport;
588
589	DEBUG_MISC((dfd, " sendto()ing, addr.sin_port=%d, addr.sin_addr.s_addr=%.16s\n",
590	ntohs(addr.sin_port), inet_ntoa(addr.sin_addr)));
591
592	/* Don't care what port we get */
593	ret = sendto(so->s, m->m_data, m->m_len, 0,
594	(struct sockaddr *)&addr, sizeof (struct sockaddr));
595	if (ret < 0)
596	{
597	LogRel(("UDP: sendto fails (%s)\n", strerror(errno)));
598	return -1;
599	}
600
601	/*
602	* Kill the socket if there's no reply in 4 minutes,
603	* but only if it's an expirable socket
604	*/
605	if (so->so_expire)
606	so->so_expire = curtime + SO_EXPIRE;
607	so->so_state = SS_ISFCONNECTED; /* So that it gets select()ed */
608	return 0;
609	}
610
611	/*
612	* XXX This should really be tcp_listen
613	*/
614	struct socket *
615	solisten(PNATState pData, u_int port, u_int32_t laddr, u_int lport, int flags)
616	{
617	struct sockaddr_in addr;
618	struct socket *so;
619	socklen_t addrlen = sizeof(addr);
620	int s, opt = 1;
621
622	DEBUG_CALL("solisten");
623	DEBUG_ARG("port = %d", port);
624	DEBUG_ARG("laddr = %x", laddr);
625	DEBUG_ARG("lport = %d", lport);
626	DEBUG_ARG("flags = %x", flags);
627
628	if ((so = socreate()) == NULL)
629	{
630	/* free(so); Not sofree() ??? free(NULL) == NOP */
631	return NULL;
632	}
633
634	/* Don't tcp_attach... we don't need so_snd nor so_rcv */
635	if ((so->so_tcpcb = tcp_newtcpcb(pData, so)) == NULL)
636	{
637	free(so);
638	return NULL;
639	}
640	insque(pData, so,&tcb);
641
642	/*
643	* SS_FACCEPTONCE sockets must time out.
644	*/
645	if (flags & SS_FACCEPTONCE)
646	so->so_tcpcb->t_timer[TCPT_KEEP] = TCPTV_KEEP_INIT*2;
647
648	so->so_state = (SS_FACCEPTCONN\|flags);
649	so->so_lport = lport; /* Kept in network format */
650	so->so_laddr.s_addr = laddr; /* Ditto */
651
652	addr.sin_family = AF_INET;
653	addr.sin_addr.s_addr = INADDR_ANY;
654	addr.sin_port = port;
655
656	if ( ((s = socket(AF_INET,SOCK_STREAM,0)) < 0)
657	\|\| (setsockopt(s,SOL_SOCKET,SO_REUSEADDR,(char *)&opt,sizeof(int)) < 0)
658	\|\| (bind(s,(struct sockaddr *)&addr, sizeof(addr)) < 0)
659	\|\| (listen(s,1) < 0))
660	{
661	#ifdef RT_OS_WINDOWS
662	int tmperrno = WSAGetLastError(); /* Don't clobber the real reason we failed */
663	closesocket(s);
664	sofree(pData, so);
665	/* Restore the real errno */
666	WSASetLastError(tmperrno);
667	#else
668	int tmperrno = errno; /* Don't clobber the real reason we failed */
669	close(s);
670	sofree(pData, so);
671	/* Restore the real errno */
672	errno = tmperrno;
673	#endif
674	return NULL;
675	}
676	setsockopt(s,SOL_SOCKET,SO_OOBINLINE,(char *)&opt,sizeof(int));
677
678	getsockname(s,(struct sockaddr *)&addr,&addrlen);
679	so->so_fport = addr.sin_port;
680	if (addr.sin_addr.s_addr == 0 \|\| addr.sin_addr.s_addr == loopback_addr.s_addr)
681	so->so_faddr = alias_addr;
682	else
683	so->so_faddr = addr.sin_addr;
684
685	so->s = s;
686	return so;
687	}
688
689	/*
690	* Data is available in so_rcv
691	* Just write() the data to the socket
692	* XXX not yet...
693	*/
694	void
695	sorwakeup(struct socket *so)
696	{
697	#if 0
698	sowrite(so);
699	FD_CLR(so->s,&writefds);
700	#endif
701	}
702
703	/*
704	* Data has been freed in so_snd
705	* We have room for a read() if we want to
706	* For now, don't read, it'll be done in the main loop
707	*/
708	void
709	sowwakeup(struct socket *so)
710	{
711	}
712
713	/*
714	* Various session state calls
715	* XXX Should be #define's
716	* The socket state stuff needs work, these often get call 2 or 3
717	* times each when only 1 was needed
718	*/
719	void
720	soisfconnecting(struct socket *so)
721	{
722	so->so_state &= ~(SS_NOFDREF\|SS_ISFCONNECTED\|SS_FCANTRCVMORE\|
723	SS_FCANTSENDMORE\|SS_FWDRAIN);
724	so->so_state \|= SS_ISFCONNECTING; /* Clobber other states */
725	}
726
727	void
728	soisfconnected(struct socket *so)
729	{
730	so->so_state &= ~(SS_ISFCONNECTING\|SS_FWDRAIN\|SS_NOFDREF);
731	so->so_state \|= SS_ISFCONNECTED; /* Clobber other states */
732	}
733
734	void
735	sofcantrcvmore(struct socket *so)
736	{
737	if ((so->so_state & SS_NOFDREF) == 0)
738	{
739	shutdown(so->s,0);
740	}
741	so->so_state &= ~(SS_ISFCONNECTING);
742	if (so->so_state & SS_FCANTSENDMORE)
743	so->so_state = SS_NOFDREF; /* Don't select it */
744	/* XXX close() here as well? */
745	else
746	so->so_state \|= SS_FCANTRCVMORE;
747	}
748
749	void
750	sofcantsendmore(struct socket *so)
751	{
752	if ((so->so_state & SS_NOFDREF) == 0)
753	shutdown(so->s, 1); /* send FIN to fhost */
754
755	so->so_state &= ~(SS_ISFCONNECTING);
756	if (so->so_state & SS_FCANTRCVMORE)
757	so->so_state = SS_NOFDREF; /* as above */
758	else
759	so->so_state \|= SS_FCANTSENDMORE;
760	}
761
762	void
763	soisfdisconnected(struct socket *so)
764	{
765	#if 0
766	so->so_state &= ~(SS_ISFCONNECTING\|SS_ISFCONNECTED);
767	close(so->s);
768	so->so_state = SS_ISFDISCONNECTED;
769	/*
770	* XXX Do nothing ... ?
771	*/
772	#endif
773	}
774
775	/*
776	* Set write drain mode
777	* Set CANTSENDMORE once all data has been write()n
778	*/
779	void
780	sofwdrain(struct socket *so)
781	{
782	if (so->so_rcv.sb_cc)
783	so->so_state \|= SS_FWDRAIN;
784	else
785	sofcantsendmore(so);
786	}
787
788	#ifdef VBOX_WITH_SLIRP_ICMP
789	static void
790	send_icmp_to_guest(PNATState pData, char buff, size_t len, struct socket so, const struct sockaddr_in *addr)
791	{
792	struct ip *ip;
793	uint32_t dst,src;
794	char ip_copy[256];
795	struct icmp *icp;
796	int old_ip_len;
797	int hlen, original_hlen;
798	struct mbuf *m;
799	struct icmp_msg *icm;
800	uint8_t proto;
801
802	ip = (struct ip *)buff;
803	hlen = (ip->ip_hl << 2);
804	icp = (struct icmp )((char )ip + hlen);
805
806	LogRel(("ICMP:received msg(t:%d, c:%d)\n", icp->icmp_type, icp->icmp_code));
807	if (icp->icmp_type != ICMP_ECHOREPLY
808	&& icp->icmp_type != ICMP_TIMXCEED
809	&& icp->icmp_type != ICMP_UNREACH)
810	{
811	return;
812	}
813
814	if (icp->icmp_type == ICMP_TIMXCEED
815	\|\| icp->icmp_type == ICMP_UNREACH )
816	ip = &icp->icmp_ip;
817
818	icm = icmp_find_original_mbuf(pData, ip);
819
820	if (icm == NULL)
821	{
822	LogRel(("NAT: Can't find the corresponding packet for the received ICMP\n"));
823	return;
824	}
825
826	m = icm->im_m;
827	Assert(m != NULL);
828
829	src = addr->sin_addr.s_addr;
830
831	ip = mtod(m, struct ip *);
832	proto = ip->ip_p;
833	/* Now ip is pointing on header we've sent from guest */
834	if (icp->icmp_type == ICMP_TIMXCEED
835	\|\| icp->icmp_type == ICMP_UNREACH)
836	{
837	old_ip_len = (ip->ip_hl << 2) + 64;
838	memcpy(ip_copy, ip, old_ip_len);
839	}
840
841	/* source address from original IP packet*/
842	dst = ip->ip_src.s_addr;
843
844	/* overide ther tail of old packet */
845	ip = mtod(m, struct ip ); / ip is from mbuf we've overrided */
846	original_hlen = ip->ip_hl << 2;
847	/* saves original ip header and options */
848	memcpy(m->m_data + original_hlen, buff + hlen, len - hlen);
849	m->m_len = len - hlen + original_hlen;
850	ip->ip_len = m->m_len;
851	ip->ip_p = IPPROTO_ICMP; /* the original package could be whatever, but we're response via ICMP*/
852
853	icp = (struct icmp )((char )ip + (ip->ip_hl << 2));
854	if (icp->icmp_type == ICMP_TIMXCEED
855	\|\| icp->icmp_type == ICMP_UNREACH)
856	{
857	/* according RFC 793 error messages required copy of initial IP header + 64 bit */
858	memcpy(&icp->icmp_ip, ip_copy, old_ip_len);
859	ip->ip_tos = ((ip->ip_tos & 0x1E) \| 0xC0); /* high priority for errors */
860	}
861
862	ip->ip_src.s_addr = src;
863	ip->ip_dst.s_addr = dst;
864	icmp_reflect(pData, m);
865	LIST_REMOVE(icm, im_list);
866	/* Don't call m_free here*/
867
868	if (icp->icmp_type == ICMP_TIMXCEED
869	\|\| icp->icmp_type == ICMP_UNREACH)
870	{
871	icm->im_so->so_m = NULL;
872	switch (proto)
873	{
874	case IPPROTO_UDP:
875	/XXX: so->so_m already freed so we shouldn't call sofree /
876	udp_detach(pData, icm->im_so);
877	break;
878	case IPPROTO_TCP:
879	/close tcp should be here /
880	break;
881	default:
882	/* do nothing */
883	break;
884	}
885	}
886	free(icm);
887	}
888
889	# ifdef RT_OS_WINDOWS
890	static void
891	sorecvfrom_icmp_win(PNATState pData, struct socket *so)
892	{
893	int len;
894	int i;
895	struct ip *ip;
896	struct mbuf *m;
897	struct icmp *icp;
898	struct icmp_msg *icm;
899	struct ip ip_broken; / ICMP returns header + 64 bit of packet */
900	uint32_t src;
901	ICMP_ECHO_REPLY *icr;
902	u_char code = ~0;
903	len = pData->pfIcmpParseReplies(pData->pvIcmpBuffer, pData->szIcmpBuffer);
904	if (len < 0)
905	{
906	LogRel(("NAT: Error (%d) occurred on ICMP receiving\n", GetLastError()));
907	return;
908	}
909	if (len == 0)
910	return; /* no error */
911	LogRel(("IcmpParseReplies returns %ld\n", len));
912	icr = (ICMP_ECHO_REPLY *)pData->pvIcmpBuffer;
913	for (i = 0; i < len; ++i)
914	{
915	switch(icr[i].Status)
916	{
917	case IP_DEST_HOST_UNREACHABLE:
918	code = (code != ~0 ? code : ICMP_UNREACH_HOST);
919	case IP_DEST_NET_UNREACHABLE:
920	code = (code != ~0 ? code : ICMP_UNREACH_NET);
921	case IP_DEST_PROT_UNREACHABLE:
922	code = (code != ~0 ? code : ICMP_UNREACH_PROTOCOL);
923	/* UNREACH error inject here */
924	case IP_DEST_PORT_UNREACHABLE:
925	code = (code != ~0 ? code : ICMP_UNREACH_PORT);
926	icmp_error(pData, so->so_m, ICMP_UNREACH, code, 0, "Error occurred!!!");
927	break;
928	case IP_SUCCESS: /* echo replied */
929	m = m_get(pData);
930	ip = mtod(m, struct ip *);
931	ip->ip_src.s_addr = icr[i].Address;
932	ip->ip_p = IPPROTO_ICMP;
933	ip->ip_dst.s_addr = so->so_laddr.s_addr; /XXX: still the hack/
934	ip->ip_hl = sizeof(struct ip) >> 2; /* requiered for icmp_reflect, no IP options */
935	ip->ip_ttl = icr[i].Options.Ttl;
936
937	icp = (struct icmp )&ip[1]; / no options */
938	icp->icmp_type = ICMP_ECHOREPLY;
939	icp->icmp_code = 0;
940	icp->icmp_id = so->so_icmp_id;
941	icp->icmp_seq = so->so_icmp_seq;
942	memcpy(icp->icmp_data, icr[i].Data, icr[i].DataSize);
943
944	ip->ip_len = sizeof(struct ip) + ICMP_MINLEN + icr[i].DataSize;
945	m->m_len = ip->ip_len;
946
947	icmp_reflect(pData, m);
948	case IP_TTL_EXPIRED_TRANSIT: /* TTL expired */
949
950	ip_broken = icr[i].Data;
951	icm = icmp_find_original_mbuf(pData, ip_broken);
952	if (icm == NULL) {
953	LogRel(("ICMP: can't find original package (first double word %x)\n", (uint32_t )ip_broken));
954	return;
955	}
956	m = icm->im_m;
957	ip = mtod(m, struct ip *);
958	ip->ip_ttl = icr[i].Options.Ttl;
959	src = ip->ip_src.s_addr;
960	ip->ip_dst.s_addr = src;
961	ip->ip_dst.s_addr = icr[i].Address;
962	icp = (struct icmp )((char )ip + (ip->ip_hl << 2));
963	ip_broken->ip_src.s_addr = src; /it packet sent from host not from guest/
964	memcpy(icp->icmp_data, ip_broken, (ip_broken->ip_hl << 2) + 64);
965	icmp_reflect(pData, m);
966	break;
967	default:
968	LogRel(("ICMP(default): message with Status: %x was received from %x\n", icr[i].Status, icr[i].Address));
969	break;
970	}
971	}
972	}
973	# endif /* RT_OS_WINDOWS */
974	#endif /* VBOX_WITH_SLIRP_ICMP */
975
976	static void sorecvfrom_icmp_unix(PNATState pData, struct socket *so)
977	{
978	struct sockaddr_in addr;
979	socklen_t addrlen = sizeof(struct sockaddr_in);
980	char buff[1500];
981	int len;
982	len = recvfrom(so->s, buff, 1500, 0,
983	(struct sockaddr *)&addr, &addrlen);
984	/* XXX Check if reply is "correct"? */
985
986	if (len == -1 \|\| len == 0)
987	{
988	u_char code = ICMP_UNREACH_PORT;
989
990	if (errno == EHOSTUNREACH)
991	code = ICMP_UNREACH_HOST;
992	else if(errno == ENETUNREACH)
993	code = ICMP_UNREACH_NET;
994
995	DEBUG_MISC((dfd," udp icmp rx errno = %d-%s\n",
996	errno,strerror(errno)));
997	icmp_error(pData, so->so_m, ICMP_UNREACH,code, 0,strerror(errno));
998	}
999	else
1000	{
1001	#ifdef VBOX_WITH_SLIRP_ICMP
1002	send_icmp_to_guest(pData, buff, len, so, &addr);
1003	#else
1004	icmp_reflect(pData, so->so_m);
1005	so->so_m = 0; /* Don't m_free() it again! */
1006	#endif
1007	}
1008	}
1009

Note: See TracBrowser for help on using the repository browser.

source: vbox/trunk/src/VBox/Devices/Network/slirp/socket.c@ 15522

Download in other formats: