pxping.c@ 100684

Last change on this file since 100684 was 98103, checked in by vboxsync, 2 years ago
Copyright year updates by scm.
Property svn:eol-style set to `native` Property svn:keywords set to `Author Date Id Revision`
File size: 55.3 KB

Line
1	/* $Id: pxping.c 98103 2023-01-17 14:15:46Z vboxsync $ */
2	/** @file
3	* NAT Network - ping proxy, raw sockets version.
4	*/
5
6	/*
7	* Copyright (C) 2013-2023 Oracle and/or its affiliates.
8	*
9	* This file is part of VirtualBox base platform packages, as
10	* available from https://www.virtualbox.org.
11	*
12	* This program is free software; you can redistribute it and/or
13	* modify it under the terms of the GNU General Public License
14	* as published by the Free Software Foundation, in version 3 of the
15	* License.
16	*
17	* This program is distributed in the hope that it will be useful, but
18	* WITHOUT ANY WARRANTY; without even the implied warranty of
19	* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
20	* General Public License for more details.
21	*
22	* You should have received a copy of the GNU General Public License
23	* along with this program; if not, see <https://www.gnu.org/licenses>.
24	*
25	* SPDX-License-Identifier: GPL-3.0-only
26	*/
27
28	#define LOG_GROUP LOG_GROUP_NAT_SERVICE
29
30	#include "winutils.h"
31	#include "proxy.h"
32	#include "proxy_pollmgr.h"
33	#include "pxremap.h"
34
35	#include <iprt/string.h>
36
37	#ifndef RT_OS_WINDOWS
38	#include <sys/types.h>
39	#include <sys/socket.h>
40	#ifdef RT_OS_DARWIN
41	# define __APPLE_USE_RFC_3542
42	#endif
43	#include <netinet/in.h>
44	#include <poll.h>
45	#include <stdint.h>
46	#include <stdio.h>
47	#include <stdlib.h>
48	#include <string.h>
49	#else
50	#include <iprt/stdint.h>
51	#include <stdio.h>
52	#include <stdlib.h>
53	#include <string.h>
54	#include "winpoll.h"
55	#endif
56
57	#include "lwip/opt.h"
58
59	#include "lwip/sys.h"
60	#include "lwip/tcpip.h"
61	#include "lwip/inet_chksum.h"
62	#include "lwip/ip.h"
63	#include "lwip/icmp.h"
64
65	#if defined(RT_OS_LINUX) && !defined(__USE_GNU)
66	#if __GLIBC_PREREQ(2, 8)
67	/*
68	* XXX: This is gross. in6_pktinfo is now hidden behind _GNU_SOURCE
69	* https://sourceware.org/bugzilla/show_bug.cgi?id=6775
70	*
71	* But in older glibc versions, e.g. RHEL5, it is not! I don't want
72	* to deal with _GNU_SOURCE now, so as a kludge check for glibc
73	* version. It seems the __USE_GNU guard was introduced in 2.8.
74	*/
75	struct in6_pktinfo {
76	struct in6_addr ipi6_addr;
77	unsigned int ipi6_ifindex;
78	};
79	#endif /* __GLIBC_PREREQ */
80	#endif /* RT_OS_LINUX && !__USE_GNU */
81
82
83	/* forward */
84	struct ping_pcb;
85
86
87	/**
88	* Global state for ping proxy collected in one entity to minimize
89	* globals. There's only one instance of this structure.
90	*
91	* Raw ICMP sockets are promiscuous, so it doesn't make sense to have
92	* multiple. If this code ever needs to support multiple netifs, the
93	* netif member should be exiled into "pcb".
94	*/
95	struct pxping {
96	SOCKET sock4;
97
98	#if defined(RT_OS_DARWIN) \|\| defined(RT_OS_SOLARIS)
99	# define DF_WITH_IP_HDRINCL
100	int hdrincl;
101	#else
102	int df;
103	#endif
104	int ttl;
105	int tos;
106
107	SOCKET sock6;
108	#ifdef RT_OS_WINDOWS
109	LPFN_WSARECVMSG pfWSARecvMsg6;
110	#endif
111	int hopl;
112
113	struct pollmgr_handler pmhdl4;
114	struct pollmgr_handler pmhdl6;
115
116	struct netif *netif;
117
118	/**
119	* Protect lwIP and pmgr accesses to the list of pcbs.
120	*/
121	sys_mutex_t lock;
122
123	/*
124	* We need to find pcbs both from the guest side and from the host
125	* side. If we need to support industrial grade ping throughput,
126	* we will need two pcb hashes. For now, a short linked list
127	* should be enough. Cf. pxping_pcb_for_request() and
128	* pxping_pcb_for_reply().
129	*/
130	#define PXPING_MAX_PCBS 8
131	size_t npcbs;
132	struct ping_pcb *pcbs;
133
134	#define TIMEOUT 5
135	int timer_active;
136	size_t timeout_slot;
137	struct ping_pcb *timeout_list[TIMEOUT];
138	};
139
140
141	/**
142	* Quasi PCB for ping.
143	*/
144	struct ping_pcb {
145	ipX_addr_t src;
146	ipX_addr_t dst;
147
148	u8_t is_ipv6;
149	u8_t is_mapped;
150
151	u16_t guest_id;
152	u16_t host_id;
153
154	/**
155	* Desired slot in pxping::timeout_list. See pxping_timer().
156	*/
157	size_t timeout_slot;
158
159	/**
160	* Chaining for pxping::timeout_list
161	*/
162	struct ping_pcb **pprev_timeout;
163	struct ping_pcb *next_timeout;
164
165	/**
166	* Chaining for pxping::pcbs
167	*/
168	struct ping_pcb *next;
169
170	union {
171	struct sockaddr_in sin;
172	struct sockaddr_in6 sin6;
173	} peer;
174	};
175
176
177	/**
178	* lwIP thread callback message for IPv4 ping.
179	*
180	* We pass raw IP datagram for ip_output_if() so we only need pbuf and
181	* netif (from pxping).
182	*/
183	struct ping_msg {
184	struct tcpip_msg msg;
185	struct pxping *pxping;
186	struct pbuf *p;
187	};
188
189
190	/**
191	* lwIP thread callback message for IPv6 ping.
192	*
193	* We cannot obtain raw IPv6 datagram from host without extra trouble,
194	* so we pass ICMPv6 payload in pbuf and also other parameters to
195	* ip6_output_if().
196	*/
197	struct ping6_msg {
198	struct tcpip_msg msg;
199	struct pxping *pxping;
200	struct pbuf *p;
201	ip6_addr_t src, dst;
202	int hopl, tclass;
203	};
204
205
206	#ifdef RT_OS_WINDOWS
207	static int pxping_init_windows(struct pxping *pxping);
208	#endif
209	static void pxping_recv4(void arg, struct pbuf p);
210	static void pxping_recv6(void arg, struct pbuf p);
211
212	static void pxping_timer(void *arg);
213	static void pxping_timer_needed(struct pxping *pxping);
214
215	static struct ping_pcb pxping_pcb_for_request(struct pxping pxping,
216	int is_ipv6,
217	ipX_addr_t src, ipX_addr_t dst,
218	u16_t guest_id);
219	static struct ping_pcb pxping_pcb_for_reply(struct pxping pxping, int is_ipv6,
220	ipX_addr_t *dst, u16_t host_id);
221
222	static FNRTSTRFORMATTYPE pxping_pcb_rtstrfmt;
223	static struct ping_pcb pxping_pcb_allocate(struct pxping pxping);
224	static void pxping_pcb_register(struct pxping pxping, struct ping_pcb pcb);
225	static void pxping_pcb_deregister(struct pxping pxping, struct ping_pcb pcb);
226	static void pxping_pcb_delete(struct pxping pxping, struct ping_pcb pcb);
227	static void pxping_timeout_add(struct pxping pxping, struct ping_pcb pcb);
228	static void pxping_timeout_del(struct pxping pxping, struct ping_pcb pcb);
229
230	static int pxping_pmgr_pump(struct pollmgr_handler *handler, SOCKET fd, int revents);
231
232	static void pxping_pmgr_icmp4(struct pxping *pxping);
233	static void pxping_pmgr_icmp4_echo(struct pxping *pxping,
234	u16_t iplen, struct sockaddr_in *peer);
235	static void pxping_pmgr_icmp4_error(struct pxping *pxping,
236	u16_t iplen, struct sockaddr_in *peer);
237	static void pxping_pmgr_icmp6(struct pxping *pxping);
238	static void pxping_pmgr_icmp6_echo(struct pxping *pxping,
239	ip6_addr_t src, ip6_addr_t dst,
240	int hopl, int tclass, u16_t icmplen);
241	static void pxping_pmgr_icmp6_error(struct pxping *pxping,
242	ip6_addr_t src, ip6_addr_t dst,
243	int hopl, int tclass, u16_t icmplen);
244
245	static void pxping_pmgr_forward_inbound(struct pxping *pxping, u16_t iplen);
246	static void pxping_pcb_forward_inbound(void *arg);
247
248	static void pxping_pmgr_forward_inbound6(struct pxping *pxping,
249	ip6_addr_t src, ip6_addr_t dst,
250	u8_t hopl, u8_t tclass,
251	u16_t icmplen);
252	static void pxping_pcb_forward_inbound6(void *arg);
253
254	/*
255	* NB: This is not documented except in RTFS.
256	*
257	* If ip_output_if() is passed dest == NULL then it treats p as
258	* complete IP packet with payload pointing to the IP header. It does
259	* not build IP header, ignores all header-related arguments, fetches
260	* real destination from the header in the pbuf and outputs pbuf to
261	* the specified netif.
262	*/
263	#define ip_raw_output_if(p, netif) \
264	(ip_output_if((p), NULL, NULL, 0, 0, 0, (netif)))
265
266
267
268	static struct pxping g_pxping;
269
270
271	err_t
272	pxping_init(struct netif *netif, SOCKET sock4, SOCKET sock6)
273	{
274	const int on = 1;
275	int status;
276
277	if (sock4 == INVALID_SOCKET && sock6 == INVALID_SOCKET) {
278	return ERR_VAL;
279	}
280
281	g_pxping.netif = netif;
282	sys_mutex_new(&g_pxping.lock);
283
284	g_pxping.sock4 = sock4;
285	if (g_pxping.sock4 != INVALID_SOCKET) {
286	#ifdef DF_WITH_IP_HDRINCL
287	g_pxping.hdrincl = 0;
288	#else
289	g_pxping.df = -1;
290	#endif
291	g_pxping.ttl = -1;
292	g_pxping.tos = 0;
293
294	#ifdef RT_OS_LINUX
295	{
296	const int dont = IP_PMTUDISC_DONT;
297	status = setsockopt(sock4, IPPROTO_IP, IP_MTU_DISCOVER,
298	&dont, sizeof(dont));
299	if (status != 0) {
300	DPRINTF(("IP_MTU_DISCOVER: %R[sockerr]\n", SOCKERRNO()));
301	}
302	}
303	#endif /* RT_OS_LINUX */
304
305	g_pxping.pmhdl4.callback = pxping_pmgr_pump;
306	g_pxping.pmhdl4.data = (void *)&g_pxping;
307	g_pxping.pmhdl4.slot = -1;
308	pollmgr_add(&g_pxping.pmhdl4, g_pxping.sock4, POLLIN);
309
310	ping_proxy_accept(pxping_recv4, &g_pxping);
311	}
312
313	g_pxping.sock6 = sock6;
314	#ifdef RT_OS_WINDOWS
315	/* we need recvmsg */
316	if (g_pxping.sock6 != INVALID_SOCKET) {
317	status = pxping_init_windows(&g_pxping);
318	if (status == SOCKET_ERROR) {
319	g_pxping.sock6 = INVALID_SOCKET;
320	/* close(sock6); */
321	}
322	}
323	#endif
324	if (g_pxping.sock6 != INVALID_SOCKET) {
325	g_pxping.hopl = -1;
326
327	#if !defined(IPV6_RECVPKTINFO)
328	#define IPV6_RECVPKTINFO (IPV6_PKTINFO)
329	#endif
330	status = setsockopt(sock6, IPPROTO_IPV6, IPV6_RECVPKTINFO,
331	(const char *)&on, sizeof(on));
332	if (status < 0) {
333	DPRINTF(("IPV6_RECVPKTINFO: %R[sockerr]\n", SOCKERRNO()));
334	/* XXX: for now this is fatal */
335	}
336
337	#if !defined(IPV6_RECVHOPLIMIT)
338	#define IPV6_RECVHOPLIMIT (IPV6_HOPLIMIT)
339	#endif
340	status = setsockopt(sock6, IPPROTO_IPV6, IPV6_RECVHOPLIMIT,
341	(const char *)&on, sizeof(on));
342	if (status < 0) {
343	DPRINTF(("IPV6_RECVHOPLIMIT: %R[sockerr]\n", SOCKERRNO()));
344	}
345
346	#ifdef IPV6_RECVTCLASS /* new in RFC 3542, there's no RFC 2292 counterpart */
347	/** @todo IPV6_RECVTCLASS */
348	#endif
349
350	g_pxping.pmhdl6.callback = pxping_pmgr_pump;
351	g_pxping.pmhdl6.data = (void *)&g_pxping;
352	g_pxping.pmhdl6.slot = -1;
353	pollmgr_add(&g_pxping.pmhdl6, g_pxping.sock6, POLLIN);
354
355	ping6_proxy_accept(pxping_recv6, &g_pxping);
356	}
357
358	status = RTStrFormatTypeRegister("ping_pcb", pxping_pcb_rtstrfmt, NULL);
359	AssertRC(status);
360
361	return ERR_OK;
362	}
363
364
365	#ifdef RT_OS_WINDOWS
366	static int
367	pxping_init_windows(struct pxping *pxping)
368	{
369	GUID WSARecvMsgGUID = WSAID_WSARECVMSG;
370	DWORD nread;
371	int status;
372
373	pxping->pfWSARecvMsg6 = NULL;
374	status = WSAIoctl(pxping->sock6,
375	SIO_GET_EXTENSION_FUNCTION_POINTER,
376	&WSARecvMsgGUID, sizeof(WSARecvMsgGUID),
377	&pxping->pfWSARecvMsg6, sizeof(pxping->pfWSARecvMsg6),
378	&nread,
379	NULL, NULL);
380	return status;
381	}
382	#endif /* RT_OS_WINDOWS */
383
384
385	static u32_t
386	chksum_delta_16(u16_t oval, u16_t nval)
387	{
388	u32_t sum = (u16_t)~oval;
389	sum += nval;
390	return sum;
391	}
392
393
394	static u32_t
395	chksum_update_16(u16_t *oldp, u16_t nval)
396	{
397	u32_t sum = chksum_delta_16(*oldp, nval);
398	*oldp = nval;
399	return sum;
400	}
401
402
403	static u32_t
404	chksum_delta_32(u32_t oval, u32_t nval)
405	{
406	u32_t sum = ~oval;
407	sum = FOLD_U32T(sum);
408	sum += FOLD_U32T(nval);
409	return sum;
410	}
411
412
413	static u32_t
414	chksum_update_32(u32_t *oldp, u32_t nval)
415	{
416	u32_t sum = chksum_delta_32(*oldp, nval);
417	*oldp = nval;
418	return sum;
419	}
420
421
422	static u32_t
423	chksum_delta_ipv6(const ip6_addr_t oldp, const ip6_addr_t newp)
424	{
425	u32_t sum;
426
427	sum = chksum_delta_32(oldp->addr[0], newp->addr[0]);
428	sum += chksum_delta_32(oldp->addr[1], newp->addr[1]);
429	sum += chksum_delta_32(oldp->addr[2], newp->addr[2]);
430	sum += chksum_delta_32(oldp->addr[3], newp->addr[3]);
431
432	return sum;
433	}
434
435
436	static u32_t
437	chksum_update_ipv6(ip6_addr_t oldp, const ip6_addr_t newp)
438	{
439	u32_t sum;
440
441	sum = chksum_update_32(&oldp->addr[0], newp->addr[0]);
442	sum += chksum_update_32(&oldp->addr[1], newp->addr[1]);
443	sum += chksum_update_32(&oldp->addr[2], newp->addr[2]);
444	sum += chksum_update_32(&oldp->addr[3], newp->addr[3]);
445
446	return sum;
447	}
448
449
450	/**
451	* ICMP Echo Request in pbuf "p" is to be proxied.
452	*/
453	static void
454	pxping_recv4(void arg, struct pbuf p)
455	{
456	struct pxping pxping = (struct pxping )arg;
457	struct ping_pcb *pcb;
458	#ifdef DF_WITH_IP_HDRINCL
459	struct ip_hdr iph_orig;
460	#endif
461	struct icmp_echo_hdr icmph_orig;
462	struct ip_hdr *iph;
463	struct icmp_echo_hdr *icmph;
464	int df, ttl, tos;
465	u32_t sum;
466	u16_t iphlen;
467	int status;
468
469	iphlen = ip_current_header_tot_len();
470	if (iphlen != IP_HLEN) { /* we don't do options */
471	pbuf_free(p);
472	return;
473	}
474
475	iph = (/* UNCONST / struct ip_hdr )ip_current_header();
476	icmph = (struct icmp_echo_hdr *)p->payload;
477
478	pcb = pxping_pcb_for_request(pxping, 0,
479	ipX_current_src_addr(),
480	ipX_current_dest_addr(),
481	icmph->id);
482	if (pcb == NULL) {
483	pbuf_free(p);
484	return;
485	}
486
487	DPRINTF(("ping %p: %R[ping_pcb] seq %d len %u ttl %d\n",
488	pcb, pcb,
489	ntohs(icmph->seqno), (unsigned int)p->tot_len,
490	IPH_TTL(iph)));
491
492	ttl = IPH_TTL(iph);
493	if (!pcb->is_mapped) {
494	if (RT_UNLIKELY(ttl == 1)) {
495	status = pbuf_header(p, iphlen); /* back to IP header */
496	if (RT_LIKELY(status == 0)) {
497	icmp_time_exceeded(p, ICMP_TE_TTL);
498	}
499	pbuf_free(p);
500	return;
501	}
502	--ttl;
503	}
504
505	/*
506	* OS X doesn't provide a socket option to control fragmentation.
507	* Solaris doesn't provide IP_DONTFRAG on all releases we support.
508	* In this case we have to use IP_HDRINCL. We don't want to use
509	* it always since it doesn't handle fragmentation (but that's ok
510	* for DF) and Windows doesn't do automatic source address
511	* selection with IP_HDRINCL.
512	*/
513	df = (IPH_OFFSET(iph) & PP_HTONS(IP_DF)) != 0;
514
515	#ifdef DF_WITH_IP_HDRINCL
516	if (df != pxping->hdrincl) {
517	status = setsockopt(pxping->sock4, IPPROTO_IP, IP_HDRINCL,
518	&df, sizeof(df));
519	if (RT_LIKELY(status == 0)) {
520	pxping->hdrincl = df;
521	}
522	else {
523	DPRINTF(("IP_HDRINCL: %R[sockerr]\n", SOCKERRNO()));
524	}
525	}
526
527	if (pxping->hdrincl) {
528	status = pbuf_header(p, iphlen); /* back to IP header */
529	if (RT_UNLIKELY(status != 0)) {
530	pbuf_free(p);
531	return;
532	}
533
534	/* we will overwrite IP header, save original for ICMP errors */
535	memcpy(&iph_orig, iph, iphlen);
536
537	if (pcb->is_mapped) {
538	ip4_addr_set_u32(&iph->dest, pcb->peer.sin.sin_addr.s_addr);
539	}
540
541	if (g_proxy_options->src4 != NULL) {
542	ip4_addr_set_u32(&iph->src, g_proxy_options->src4->sin_addr.s_addr);
543	}
544	else {
545	/* let the kernel select suitable source address */
546	ip_addr_set_any(&iph->src);
547	}
548
549	IPH_TTL_SET(iph, ttl); /* already decremented */
550	IPH_ID_SET(iph, 0); /* kernel will set one */
551	#ifdef RT_OS_DARWIN
552	/* wants ip_offset and ip_len fields in host order */
553	IPH_OFFSET_SET(iph, ntohs(IPH_OFFSET(iph)));
554	IPH_LEN_SET(iph, ntohs(IPH_LEN(iph)));
555	/* wants checksum of everything (sic!), in host order */
556	sum = inet_chksum_pbuf(p);
557	IPH_CHKSUM_SET(iph, sum);
558	#else /* !RT_OS_DARWIN */
559	IPH_CHKSUM_SET(iph, 0); /* kernel will recalculate */
560	#endif
561	}
562	else /* !pxping->hdrincl */
563	#endif /* DF_WITH_IP_HDRINCL */
564	{
565	#if !defined(DF_WITH_IP_HDRINCL)
566	/* control DF flag via setsockopt(2) */
567	#define USE_DF_OPTION(_Optname) \
568	const int dfopt = _Optname; \
569	const char * const dfoptname = #_Optname; NOREF(dfoptname)
570	#if defined(RT_OS_LINUX)
571	USE_DF_OPTION(IP_MTU_DISCOVER);
572	df = df ? IP_PMTUDISC_DO : IP_PMTUDISC_DONT;
573	#elif defined(RT_OS_SOLARIS) \|\| defined(RT_OS_FREEBSD)
574	USE_DF_OPTION(IP_DONTFRAG);
575	#elif defined(RT_OS_WINDOWS)
576	USE_DF_OPTION(IP_DONTFRAGMENT);
577	#endif
578	if (df != pxping->df) {
579	status = setsockopt(pxping->sock4, IPPROTO_IP, dfopt,
580	(char *)&df, sizeof(df));
581	if (RT_LIKELY(status == 0)) {
582	pxping->df = df;
583	}
584	else {
585	DPRINTF(("%s: %R[sockerr]\n", dfoptname, SOCKERRNO()));
586	}
587	}
588	#endif /* !DF_WITH_IP_HDRINCL */
589
590	if (ttl != pxping->ttl) {
591	status = setsockopt(pxping->sock4, IPPROTO_IP, IP_TTL,
592	(char *)&ttl, sizeof(ttl));
593	if (RT_LIKELY(status == 0)) {
594	pxping->ttl = ttl;
595	}
596	else {
597	DPRINTF(("IP_TTL: %R[sockerr]\n", SOCKERRNO()));
598	}
599	}
600
601	tos = IPH_TOS(iph);
602	if (tos != pxping->tos) {
603	status = setsockopt(pxping->sock4, IPPROTO_IP, IP_TOS,
604	(char *)&tos, sizeof(tos));
605	if (RT_LIKELY(status == 0)) {
606	pxping->tos = tos;
607	}
608	else {
609	DPRINTF(("IP_TOS: %R[sockerr]\n", SOCKERRNO()));
610	}
611	}
612	}
613
614	/* rewrite ICMP echo header */
615	memcpy(&icmph_orig, icmph, sizeof(*icmph));
616	sum = (u16_t)~icmph->chksum;
617	sum += chksum_update_16(&icmph->id, pcb->host_id);
618	sum = FOLD_U32T(sum);
619	icmph->chksum = ~sum;
620
621	status = proxy_sendto(pxping->sock4, p,
622	&pcb->peer.sin, sizeof(pcb->peer.sin));
623	if (status != 0) {
624	int error = -status;
625	DPRINTF(("%s: sendto: %R[sockerr]\n", __func__, error));
626
627	#ifdef DF_WITH_IP_HDRINCL
628	if (pxping->hdrincl) {
629	/* restore original IP header */
630	memcpy(iph, &iph_orig, iphlen);
631	}
632	else
633	#endif
634	{
635	status = pbuf_header(p, iphlen); /* back to IP header */
636	if (RT_UNLIKELY(status != 0)) {
637	pbuf_free(p);
638	return;
639	}
640	}
641
642	/* restore original ICMP header */
643	memcpy(icmph, &icmph_orig, sizeof(*icmph));
644
645	/*
646	* Some ICMP errors may be generated by the kernel and we read
647	* them from the socket and forward them normally, hence the
648	* ifdefs below.
649	*/
650	switch (error) {
651
652	#if !( defined(RT_OS_SOLARIS) \
653	\|\| (defined(RT_OS_LINUX) && !defined(DF_WITH_IP_HDRINCL)) \
654	)
655	case EMSGSIZE:
656	icmp_dest_unreach(p, ICMP_DUR_FRAG);
657	break;
658	#endif
659
660	case ENETDOWN:
661	case ENETUNREACH:
662	icmp_dest_unreach(p, ICMP_DUR_NET);
663	break;
664
665	case EHOSTDOWN:
666	case EHOSTUNREACH:
667	icmp_dest_unreach(p, ICMP_DUR_HOST);
668	break;
669	}
670	}
671
672	pbuf_free(p);
673	}
674
675
676	/**
677	* ICMPv6 Echo Request in pbuf "p" is to be proxied.
678	*/
679	static void
680	pxping_recv6(void arg, struct pbuf p)
681	{
682	struct pxping pxping = (struct pxping )arg;
683	struct ping_pcb *pcb;
684	struct ip6_hdr *iph;
685	struct icmp6_echo_hdr *icmph;
686	int hopl;
687	u16_t iphlen;
688	u16_t id, seq;
689	int status;
690
691	iph = (/* UNCONST / struct ip6_hdr )ip6_current_header();
692	iphlen = ip_current_header_tot_len();
693
694	icmph = (struct icmp6_echo_hdr *)p->payload;
695
696	id = icmph->id;
697	seq = icmph->seqno;
698
699	pcb = pxping_pcb_for_request(pxping, 1,
700	ipX_current_src_addr(),
701	ipX_current_dest_addr(),
702	id);
703	if (pcb == NULL) {
704	pbuf_free(p);
705	return;
706	}
707
708	DPRINTF(("ping %p: %R[ping_pcb] seq %d len %u hopl %d\n",
709	pcb, pcb,
710	ntohs(seq), (unsigned int)p->tot_len,
711	IP6H_HOPLIM(iph)));
712
713	hopl = IP6H_HOPLIM(iph);
714	if (!pcb->is_mapped) {
715	if (hopl == 1) {
716	status = pbuf_header(p, iphlen); /* back to IP header */
717	if (RT_LIKELY(status == 0)) {
718	icmp6_time_exceeded(p, ICMP6_TE_HL);
719	}
720	pbuf_free(p);
721	return;
722	}
723	--hopl;
724	}
725
726	/*
727	* Rewrite ICMPv6 echo header. We don't need to recompute the
728	* checksum since, unlike IPv4, checksum includes pseudo-header.
729	* OS computes checksum for us on send() since it needs to select
730	* source address.
731	*/
732	icmph->id = pcb->host_id;
733
734	/** @todo use control messages to save a syscall? */
735	if (hopl != pxping->hopl) {
736	status = setsockopt(pxping->sock6, IPPROTO_IPV6, IPV6_UNICAST_HOPS,
737	(char *)&hopl, sizeof(hopl));
738	if (status == 0) {
739	pxping->hopl = hopl;
740	}
741	else {
742	DPRINTF(("IPV6_HOPLIMIT: %R[sockerr]\n", SOCKERRNO()));
743	}
744	}
745
746	status = proxy_sendto(pxping->sock6, p,
747	&pcb->peer.sin6, sizeof(pcb->peer.sin6));
748	if (status != 0) {
749	int error = -status;
750	DPRINTF(("%s: sendto: %R[sockerr]\n", __func__, error));
751
752	status = pbuf_header(p, iphlen); /* back to IP header */
753	if (RT_UNLIKELY(status != 0)) {
754	pbuf_free(p);
755	return;
756	}
757
758	/* restore original ICMP header */
759	icmph->id = pcb->guest_id;
760
761	switch (error) {
762	case EACCES:
763	icmp6_dest_unreach(p, ICMP6_DUR_PROHIBITED);
764	break;
765
766	#ifdef ENONET
767	case ENONET:
768	#endif
769	case ENETDOWN:
770	case ENETUNREACH:
771	case EHOSTDOWN:
772	case EHOSTUNREACH:
773	icmp6_dest_unreach(p, ICMP6_DUR_NO_ROUTE);
774	break;
775	}
776	}
777
778	pbuf_free(p);
779	}
780
781
782	/**
783	* Formatter for %R[ping_pcb].
784	*/
785	static DECLCALLBACK(size_t)
786	pxping_pcb_rtstrfmt(PFNRTSTROUTPUT pfnOutput, void *pvArgOutput,
787	const char pszType, const void pvValue,
788	int cchWidth, int cchPrecision, unsigned int fFlags,
789	void *pvUser)
790	{
791	const struct ping_pcb pcb = (const struct ping_pcb )pvValue;
792	size_t cb = 0;
793
794	NOREF(cchWidth);
795	NOREF(cchPrecision);
796	NOREF(fFlags);
797	NOREF(pvUser);
798
799	AssertReturn(strcmp(pszType, "ping_pcb") == 0, 0);
800
801	if (pcb == NULL) {
802	return RTStrFormat(pfnOutput, pvArgOutput, NULL, NULL, "(null)");
803	}
804
805	/* XXX: %RTnaipv4 takes the value, but %RTnaipv6 takes the pointer */
806	if (pcb->is_ipv6) {
807	cb += RTStrFormat(pfnOutput, pvArgOutput, NULL, NULL,
808	"%RTnaipv6 -> %RTnaipv6", &pcb->src, &pcb->dst);
809	if (pcb->is_mapped) {
810	cb += RTStrFormat(pfnOutput, pvArgOutput, NULL, NULL,
811	" (%RTnaipv6)", &pcb->peer.sin6.sin6_addr);
812	}
813	}
814	else {
815	cb += RTStrFormat(pfnOutput, pvArgOutput, NULL, NULL,
816	"%RTnaipv4 -> %RTnaipv4",
817	ip4_addr_get_u32(ipX_2_ip(&pcb->src)),
818	ip4_addr_get_u32(ipX_2_ip(&pcb->dst)));
819	if (pcb->is_mapped) {
820	cb += RTStrFormat(pfnOutput, pvArgOutput, NULL, NULL,
821	" (%RTnaipv4)", pcb->peer.sin.sin_addr.s_addr);
822	}
823	}
824
825	cb += RTStrFormat(pfnOutput, pvArgOutput, NULL, NULL,
826	" id %04x->%04x", ntohs(pcb->guest_id), ntohs(pcb->host_id));
827
828	return cb;
829	}
830
831
832	static struct ping_pcb *
833	pxping_pcb_allocate(struct pxping *pxping)
834	{
835	struct ping_pcb *pcb;
836
837	if (pxping->npcbs >= PXPING_MAX_PCBS) {
838	return NULL;
839	}
840
841	pcb = (struct ping_pcb )malloc(sizeof(pcb));
842	if (pcb == NULL) {
843	return NULL;
844	}
845
846	++pxping->npcbs;
847	return pcb;
848	}
849
850
851	static void
852	pxping_pcb_delete(struct pxping pxping, struct ping_pcb pcb)
853	{
854	LWIP_ASSERT1(pxping->npcbs > 0);
855	LWIP_ASSERT1(pcb->next == NULL);
856	LWIP_ASSERT1(pcb->pprev_timeout == NULL);
857
858	DPRINTF(("%s: ping %p\n", __func__, (void *)pcb));
859
860	--pxping->npcbs;
861	free(pcb);
862	}
863
864
865	static void
866	pxping_timeout_add(struct pxping pxping, struct ping_pcb pcb)
867	{
868	struct ping_pcb **chain;
869
870	LWIP_ASSERT1(pcb->pprev_timeout == NULL);
871
872	chain = &pxping->timeout_list[pcb->timeout_slot];
873	if ((pcb->next_timeout = *chain) != NULL) {
874	(*chain)->pprev_timeout = &pcb->next_timeout;
875	}
876	*chain = pcb;
877	pcb->pprev_timeout = chain;
878	}
879
880
881	static void
882	pxping_timeout_del(struct pxping pxping, struct ping_pcb pcb)
883	{
884	LWIP_UNUSED_ARG(pxping);
885
886	LWIP_ASSERT1(pcb->pprev_timeout != NULL);
887	if (pcb->next_timeout != NULL) {
888	pcb->next_timeout->pprev_timeout = pcb->pprev_timeout;
889	}
890	*pcb->pprev_timeout = pcb->next_timeout;
891	pcb->pprev_timeout = NULL;
892	pcb->next_timeout = NULL;
893	}
894
895
896	static void
897	pxping_pcb_register(struct pxping pxping, struct ping_pcb pcb)
898	{
899	pcb->next = pxping->pcbs;
900	pxping->pcbs = pcb;
901
902	pxping_timeout_add(pxping, pcb);
903	}
904
905
906	static void
907	pxping_pcb_deregister(struct pxping pxping, struct ping_pcb pcb)
908	{
909	struct ping_pcb **p;
910
911	for (p = &pxping->pcbs; p != NULL; p = &(p)->next) {
912	if (*p == pcb) {
913	*p = pcb->next;
914	pcb->next = NULL;
915	break;
916	}
917	}
918
919	pxping_timeout_del(pxping, pcb);
920	}
921
922
923	static struct ping_pcb *
924	pxping_pcb_for_request(struct pxping *pxping,
925	int is_ipv6, ipX_addr_t src, ipX_addr_t dst,
926	u16_t guest_id)
927	{
928	struct ping_pcb *pcb;
929
930	/* on lwip thread, so no concurrent updates */
931	for (pcb = pxping->pcbs; pcb != NULL; pcb = pcb->next) {
932	if (pcb->guest_id == guest_id
933	&& pcb->is_ipv6 == is_ipv6
934	&& ipX_addr_cmp(is_ipv6, &pcb->dst, dst)
935	&& ipX_addr_cmp(is_ipv6, &pcb->src, src))
936	{
937	break;
938	}
939	}
940
941	if (pcb == NULL) {
942	int mapped;
943
944	pcb = pxping_pcb_allocate(pxping);
945	if (pcb == NULL) {
946	return NULL;
947	}
948
949	pcb->is_ipv6 = is_ipv6;
950	ipX_addr_copy(is_ipv6, pcb->src, *src);
951	ipX_addr_copy(is_ipv6, pcb->dst, *dst);
952
953	pcb->guest_id = guest_id;
954	#ifdef RT_OS_WINDOWS
955	# define random() (rand())
956	#endif
957	pcb->host_id = random() & 0xffffUL;
958
959	pcb->pprev_timeout = NULL;
960	pcb->next_timeout = NULL;
961
962	if (is_ipv6) {
963	pcb->peer.sin6.sin6_family = AF_INET6;
964	#if HAVE_SA_LEN
965	pcb->peer.sin6.sin6_len = sizeof(pcb->peer.sin6);
966	#endif
967	pcb->peer.sin6.sin6_port = htons(IPPROTO_ICMPV6);
968	pcb->peer.sin6.sin6_flowinfo = 0;
969	mapped = pxremap_outbound_ip6((ip6_addr_t *)&pcb->peer.sin6.sin6_addr,
970	ipX_2_ip6(&pcb->dst));
971	}
972	else {
973	pcb->peer.sin.sin_family = AF_INET;
974	#if HAVE_SA_LEN
975	pcb->peer.sin.sin_len = sizeof(pcb->peer.sin);
976	#endif
977	pcb->peer.sin.sin_port = htons(IPPROTO_ICMP);
978	mapped = pxremap_outbound_ip4((ip_addr_t *)&pcb->peer.sin.sin_addr,
979	ipX_2_ip(&pcb->dst));
980	}
981
982	if (mapped == PXREMAP_FAILED) {
983	free(pcb);
984	return NULL;
985	}
986	else {
987	pcb->is_mapped = (mapped == PXREMAP_MAPPED);
988	}
989
990	pcb->timeout_slot = pxping->timeout_slot;
991
992	sys_mutex_lock(&pxping->lock);
993	pxping_pcb_register(pxping, pcb);
994	sys_mutex_unlock(&pxping->lock);
995
996	DPRINTF(("ping %p: %R[ping_pcb] - created\n", pcb, pcb));
997
998	pxping_timer_needed(pxping);
999	}
1000	else {
1001	/* just bump up expiration timeout lazily */
1002	DPRINTF(("ping %p: %R[ping_pcb] - slot %d -> %d\n",
1003	pcb, pcb,
1004	(unsigned int)pcb->timeout_slot,
1005	(unsigned int)pxping->timeout_slot));
1006	pcb->timeout_slot = pxping->timeout_slot;
1007	}
1008
1009	return pcb;
1010	}
1011
1012
1013	/* GCC 12.2.1 complains that array subscript is partly outside
1014	* of array bounds in expansion of ipX_addr_cmp. */
1015	#if RT_GNUC_PREREQ(12, 0)
1016	# pragma GCC diagnostic push
1017	# pragma GCC diagnostic ignored "-Warray-bounds"
1018	#endif
1019	/**
1020	* Called on pollmgr thread. Caller must do the locking since caller
1021	* is going to use the returned pcb, which needs to be protected from
1022	* being expired by pxping_timer() on lwip thread.
1023	*/
1024	static struct ping_pcb *
1025	pxping_pcb_for_reply(struct pxping *pxping,
1026	int is_ipv6, ipX_addr_t *dst, u16_t host_id)
1027	{
1028	struct ping_pcb *pcb;
1029
1030	for (pcb = pxping->pcbs; pcb != NULL; pcb = pcb->next) {
1031	if (pcb->host_id == host_id
1032	&& pcb->is_ipv6 == is_ipv6
1033	/* XXX: allow broadcast pings? */
1034	&& ipX_addr_cmp(is_ipv6, &pcb->dst, dst))
1035	{
1036	return pcb;
1037	}
1038	}
1039
1040	return NULL;
1041	}
1042	#if RT_GNUC_PREREQ(12, 0)
1043	# pragma GCC diagnostic pop
1044	#endif
1045
1046
1047	static void
1048	pxping_timer(void *arg)
1049	{
1050	struct pxping pxping = (struct pxping )arg;
1051	struct ping_pcb *chain, pcb;
1052
1053	pxping->timer_active = 0;
1054
1055	/*
1056	* New slot points to the list of pcbs to check for expiration.
1057	*/
1058	LWIP_ASSERT1(pxping->timeout_slot < TIMEOUT);
1059	if (++pxping->timeout_slot == TIMEOUT) {
1060	pxping->timeout_slot = 0;
1061	}
1062
1063	chain = &pxping->timeout_list[pxping->timeout_slot];
1064	pcb = *chain;
1065
1066	/* protect from pollmgr concurrent reads */
1067	sys_mutex_lock(&pxping->lock);
1068
1069	while (pcb != NULL) {
1070	struct ping_pcb *xpcb = pcb;
1071	pcb = pcb->next_timeout;
1072
1073	if (xpcb->timeout_slot == pxping->timeout_slot) {
1074	/* expired */
1075	pxping_pcb_deregister(pxping, xpcb);
1076	pxping_pcb_delete(pxping, xpcb);
1077	}
1078	else {
1079	/*
1080	* If there was another request, we updated timeout_slot
1081	* but delayed actually moving the pcb until now.
1082	*/
1083	pxping_timeout_del(pxping, xpcb); /* from current slot */
1084	pxping_timeout_add(pxping, xpcb); /* to new slot */
1085	}
1086	}
1087
1088	sys_mutex_unlock(&pxping->lock);
1089	pxping_timer_needed(pxping);
1090	}
1091
1092
1093	static void
1094	pxping_timer_needed(struct pxping *pxping)
1095	{
1096	if (!pxping->timer_active && pxping->pcbs != NULL) {
1097	pxping->timer_active = 1;
1098	sys_timeout(1 * 1000, pxping_timer, pxping);
1099	}
1100	}
1101
1102
1103	static int
1104	pxping_pmgr_pump(struct pollmgr_handler *handler, SOCKET fd, int revents)
1105	{
1106	struct pxping *pxping;
1107
1108	pxping = (struct pxping *)handler->data;
1109	LWIP_ASSERT1(fd == pxping->sock4 \|\| fd == pxping->sock6);
1110
1111	if (revents & ~(POLLIN\|POLLERR)) {
1112	DPRINTF0(("%s: unexpected revents 0x%x\n", __func__, revents));
1113	return POLLIN;
1114	}
1115
1116	if (revents & POLLERR) {
1117	int sockerr = -1;
1118	socklen_t optlen = (socklen_t)sizeof(sockerr);
1119	int status;
1120
1121	status = getsockopt(fd, SOL_SOCKET,
1122	SO_ERROR, (char *)&sockerr, &optlen);
1123	if (status < 0) {
1124	DPRINTF(("%s: sock %d: SO_ERROR failed: %R[sockerr]\n",
1125	__func__, fd, SOCKERRNO()));
1126	}
1127	else {
1128	DPRINTF(("%s: sock %d: %R[sockerr]\n",
1129	__func__, fd, sockerr));
1130	}
1131	}
1132
1133	if ((revents & POLLIN) == 0) {
1134	return POLLIN;
1135	}
1136
1137	if (fd == pxping->sock4) {
1138	pxping_pmgr_icmp4(pxping);
1139	}
1140	else /* fd == pxping->sock6 */ {
1141	pxping_pmgr_icmp6(pxping);
1142	}
1143
1144	return POLLIN;
1145	}
1146
1147
1148	/**
1149	* Process incoming ICMP message for the host.
1150	* NB: we will get a lot of spam here and have to sift through it.
1151	*/
1152	static void
1153	pxping_pmgr_icmp4(struct pxping *pxping)
1154	{
1155	struct sockaddr_in sin;
1156	socklen_t salen = sizeof(sin);
1157	ssize_t nread;
1158	struct ip_hdr *iph;
1159	struct icmp_echo_hdr *icmph;
1160	u16_t iplen, ipoff;
1161
1162	memset(&sin, 0, sizeof(sin));
1163
1164	/*
1165	* Reads from raw IPv4 sockets deliver complete IP datagrams with
1166	* IP header included.
1167	*/
1168	nread = recvfrom(pxping->sock4, pollmgr_udpbuf, sizeof(pollmgr_udpbuf), 0,
1169	(struct sockaddr *)&sin, &salen);
1170	if (nread < 0) {
1171	DPRINTF(("%s: %R[sockerr]\n", __func__, SOCKERRNO()));
1172	return;
1173	}
1174
1175	if (nread < IP_HLEN) {
1176	DPRINTF2(("%s: read %d bytes, IP header truncated\n",
1177	__func__, (unsigned int)nread));
1178	return;
1179	}
1180
1181	iph = (struct ip_hdr *)pollmgr_udpbuf;
1182
1183	/* match version */
1184	if (IPH_V(iph) != 4) {
1185	DPRINTF2(("%s: unexpected IP version %d\n", __func__, IPH_V(iph)));
1186	return;
1187	}
1188
1189	/* no fragmentation */
1190	ipoff = IPH_OFFSET(iph);
1191	#if defined(RT_OS_DARWIN)
1192	/* darwin reports IPH_OFFSET in host byte order */
1193	ipoff = htons(ipoff);
1194	IPH_OFFSET_SET(iph, ipoff);
1195	#endif
1196	if ((ipoff & PP_HTONS(IP_OFFMASK \| IP_MF)) != 0) {
1197	DPRINTF2(("%s: dropping fragmented datagram (0x%04x)\n",
1198	__func__, ntohs(ipoff)));
1199	return;
1200	}
1201
1202	/* no options */
1203	if (IPH_HL(iph) * 4 != IP_HLEN) {
1204	DPRINTF2(("%s: dropping datagram with options (IP header length %d)\n",
1205	__func__, IPH_HL(iph) * 4));
1206	return;
1207	}
1208
1209	if (IPH_PROTO(iph) != IP_PROTO_ICMP) {
1210	DPRINTF2(("%s: unexpected protocol %d\n", __func__, IPH_PROTO(iph)));
1211	return;
1212	}
1213
1214	iplen = IPH_LEN(iph);
1215	#if !defined(RT_OS_DARWIN)
1216	/* darwin reports IPH_LEN in host byte order */
1217	iplen = ntohs(iplen);
1218	#endif
1219	#if defined(RT_OS_DARWIN) \|\| defined(RT_OS_SOLARIS)
1220	/* darwin and solaris change IPH_LEN to payload length only */
1221	iplen += IP_HLEN; /* we verified there are no options */
1222	IPH_LEN_SET(iph, htons(iplen));
1223	#endif
1224	if (nread < iplen) {
1225	DPRINTF2(("%s: read %d bytes but total length is %d bytes\n",
1226	__func__, (unsigned int)nread, (unsigned int)iplen));
1227	return;
1228	}
1229
1230	if (iplen < IP_HLEN + ICMP_HLEN) {
1231	DPRINTF2(("%s: IP length %d bytes, ICMP header truncated\n",
1232	__func__, iplen));
1233	return;
1234	}
1235
1236	icmph = (struct icmp_echo_hdr *)(pollmgr_udpbuf + IP_HLEN);
1237	if (ICMPH_TYPE(icmph) == ICMP_ER) {
1238	pxping_pmgr_icmp4_echo(pxping, iplen, &sin);
1239	}
1240	else if (ICMPH_TYPE(icmph) == ICMP_DUR \|\| ICMPH_TYPE(icmph) == ICMP_TE) {
1241	pxping_pmgr_icmp4_error(pxping, iplen, &sin);
1242	}
1243	#if 1
1244	else {
1245	DPRINTF2(("%s: ignoring ICMP type %d\n", __func__, ICMPH_TYPE(icmph)));
1246	}
1247	#endif
1248	}
1249
1250
1251	/**
1252	* Check if this incoming ICMP echo reply is for one of our pings and
1253	* forward it to the guest.
1254	*/
1255	static void
1256	pxping_pmgr_icmp4_echo(struct pxping *pxping,
1257	u16_t iplen, struct sockaddr_in *peer)
1258	{
1259	struct ip_hdr *iph;
1260	struct icmp_echo_hdr *icmph;
1261	u16_t id, seq;
1262	ip_addr_t guest_ip, target_ip;
1263	int mapped;
1264	struct ping_pcb *pcb;
1265	u16_t guest_id;
1266	u16_t oipsum;
1267	u32_t sum;
1268	RT_NOREF(peer);
1269
1270	iph = (struct ip_hdr *)pollmgr_udpbuf;
1271	icmph = (struct icmp_echo_hdr *)(pollmgr_udpbuf + IP_HLEN);
1272
1273	id = icmph->id;
1274	seq = icmph->seqno;
1275
1276	DPRINTF(("<--- PING %RTnaipv4 id 0x%x seq %d\n",
1277	peer->sin_addr.s_addr, ntohs(id), ntohs(seq)));
1278
1279	/*
1280	* Is this a reply to one of our pings?
1281	*/
1282
1283	ip_addr_copy(target_ip, iph->src);
1284	mapped = pxremap_inbound_ip4(&target_ip, &target_ip);
1285	if (mapped == PXREMAP_FAILED) {
1286	return;
1287	}
1288	if (mapped == PXREMAP_ASIS && IPH_TTL(iph) == 1) {
1289	DPRINTF2(("%s: dropping packet with ttl 1\n", __func__));
1290	return;
1291	}
1292
1293	sys_mutex_lock(&pxping->lock);
1294	pcb = pxping_pcb_for_reply(pxping, 0, ip_2_ipX(&target_ip), id);
1295	if (pcb == NULL) {
1296	sys_mutex_unlock(&pxping->lock);
1297	DPRINTF2(("%s: no match\n", __func__));
1298	return;
1299	}
1300
1301	DPRINTF2(("%s: pcb %p\n", __func__, (void *)pcb));
1302
1303	/* save info before unlocking since pcb may expire */
1304	ip_addr_copy(guest_ip, *ipX_2_ip(&pcb->src));
1305	guest_id = pcb->guest_id;
1306
1307	sys_mutex_unlock(&pxping->lock);
1308
1309
1310	/*
1311	* Rewrite headers and forward to guest.
1312	*/
1313
1314	/* rewrite ICMP echo header */
1315	sum = (u16_t)~icmph->chksum;
1316	sum += chksum_update_16(&icmph->id, guest_id);
1317	sum = FOLD_U32T(sum);
1318	icmph->chksum = ~sum;
1319
1320	/* rewrite IP header */
1321	oipsum = IPH_CHKSUM(iph);
1322	if (oipsum == 0) {
1323	/* Solaris doesn't compute checksum for local replies */
1324	ip_addr_copy(iph->dest, guest_ip);
1325	if (mapped == PXREMAP_MAPPED) {
1326	ip_addr_copy(iph->src, target_ip);
1327	}
1328	else {
1329	IPH_TTL_SET(iph, IPH_TTL(iph) - 1);
1330	}
1331	IPH_CHKSUM_SET(iph, inet_chksum(iph, ntohs(IPH_LEN(iph))));
1332	}
1333	else {
1334	sum = (u16_t)~oipsum;
1335	sum += chksum_update_32((u32_t *)&iph->dest,
1336	ip4_addr_get_u32(&guest_ip));
1337	if (mapped == PXREMAP_MAPPED) {
1338	sum += chksum_update_32((u32_t *)&iph->src,
1339	ip4_addr_get_u32(&target_ip));
1340	}
1341	else {
1342	IPH_TTL_SET(iph, IPH_TTL(iph) - 1);
1343	sum += PP_NTOHS(~0x0100);
1344	}
1345	sum = FOLD_U32T(sum);
1346	IPH_CHKSUM_SET(iph, ~sum);
1347	}
1348
1349	pxping_pmgr_forward_inbound(pxping, iplen);
1350	}
1351
1352
1353	/**
1354	* Check if this incoming ICMP error (destination unreachable or time
1355	* exceeded) is about one of our pings and forward it to the guest.
1356	*/
1357	static void
1358	pxping_pmgr_icmp4_error(struct pxping *pxping,
1359	u16_t iplen, struct sockaddr_in *peer)
1360	{
1361	struct ip_hdr iph, oiph;
1362	struct icmp_echo_hdr icmph, oicmph;
1363	u16_t oipoff, oiphlen, oiplen;
1364	u16_t id, seq;
1365	ip_addr_t guest_ip, target_ip, error_ip;
1366	int target_mapped, error_mapped;
1367	struct ping_pcb *pcb;
1368	u16_t guest_id;
1369	u32_t sum;
1370	RT_NOREF(peer);
1371
1372	iph = (struct ip_hdr *)pollmgr_udpbuf;
1373	icmph = (struct icmp_echo_hdr *)(pollmgr_udpbuf + IP_HLEN);
1374
1375	/*
1376	* Inner IP datagram is not checked by the kernel and may be
1377	* anything, possibly malicious.
1378	*/
1379
1380	oipoff = IP_HLEN + ICMP_HLEN;
1381	oiplen = iplen - oipoff; /* NB: truncated length, not IPH_LEN(oiph) */
1382	if (oiplen < IP_HLEN) {
1383	DPRINTF2(("%s: original datagram truncated to %d bytes\n",
1384	__func__, oiplen));
1385	}
1386
1387	/* IP header of the original message */
1388	oiph = (struct ip_hdr *)(pollmgr_udpbuf + oipoff);
1389
1390	/* match version */
1391	if (IPH_V(oiph) != 4) {
1392	DPRINTF2(("%s: unexpected IP version %d\n", __func__, IPH_V(oiph)));
1393	return;
1394	}
1395
1396	/* can't match fragments except the first one */
1397	if ((IPH_OFFSET(oiph) & PP_HTONS(IP_OFFMASK)) != 0) {
1398	DPRINTF2(("%s: ignoring fragment with offset %d\n",
1399	__func__, ntohs(IPH_OFFSET(oiph) & PP_HTONS(IP_OFFMASK))));
1400	return;
1401	}
1402
1403	if (IPH_PROTO(oiph) != IP_PROTO_ICMP) {
1404	#if 0
1405	/* don't spam with every "destination unreachable" in the system */
1406	DPRINTF2(("%s: ignoring protocol %d\n", __func__, IPH_PROTO(oiph)));
1407	#endif
1408	return;
1409	}
1410
1411	oiphlen = IPH_HL(oiph) * 4;
1412	if (oiplen < oiphlen + ICMP_HLEN) {
1413	DPRINTF2(("%s: original datagram truncated to %d bytes\n",
1414	__func__, oiplen));
1415	return;
1416	}
1417
1418	oicmph = (struct icmp_echo_hdr *)(pollmgr_udpbuf + oipoff + oiphlen);
1419	if (ICMPH_TYPE(oicmph) != ICMP_ECHO) {
1420	DPRINTF2(("%s: ignoring ICMP error for original ICMP type %d\n",
1421	__func__, ICMPH_TYPE(oicmph)));
1422	return;
1423	}
1424
1425	id = oicmph->id;
1426	seq = oicmph->seqno;
1427
1428	DPRINTF2(("%s: ping %RTnaipv4 id 0x%x seq %d",
1429	__func__, ip4_addr_get_u32(&oiph->dest), ntohs(id), ntohs(seq)));
1430	if (ICMPH_TYPE(icmph) == ICMP_DUR) {
1431	DPRINTF2((" unreachable (code %d)\n", ICMPH_CODE(icmph)));
1432	}
1433	else {
1434	DPRINTF2((" time exceeded\n"));
1435	}
1436
1437
1438	/*
1439	* Is the inner (failed) datagram one of our pings?
1440	*/
1441
1442	ip_addr_copy(target_ip, oiph->dest); /* inner (failed) */
1443	target_mapped = pxremap_inbound_ip4(&target_ip, &target_ip);
1444	if (target_mapped == PXREMAP_FAILED) {
1445	return;
1446	}
1447
1448	sys_mutex_lock(&pxping->lock);
1449	pcb = pxping_pcb_for_reply(pxping, 0, ip_2_ipX(&target_ip), id);
1450	if (pcb == NULL) {
1451	sys_mutex_unlock(&pxping->lock);
1452	DPRINTF2(("%s: no match\n", __func__));
1453	return;
1454	}
1455
1456	DPRINTF2(("%s: pcb %p\n", __func__, (void *)pcb));
1457
1458	/* save info before unlocking since pcb may expire */
1459	ip_addr_copy(guest_ip, *ipX_2_ip(&pcb->src));
1460	guest_id = pcb->guest_id;
1461
1462	sys_mutex_unlock(&pxping->lock);
1463
1464
1465	/*
1466	* Rewrite both inner and outer headers and forward to guest.
1467	* Note that the checksum of the outer ICMP error message is
1468	* preserved by the changes we do to inner headers.
1469	*/
1470
1471	ip_addr_copy(error_ip, iph->src); /* node that reports the error */
1472	error_mapped = pxremap_inbound_ip4(&error_ip, &error_ip);
1473	if (error_mapped == PXREMAP_FAILED) {
1474	return;
1475	}
1476	if (error_mapped == PXREMAP_ASIS && IPH_TTL(iph) == 1) {
1477	DPRINTF2(("%s: dropping packet with ttl 1\n", __func__));
1478	return;
1479	}
1480
1481	/* rewrite inner ICMP echo header */
1482	sum = (u16_t)~oicmph->chksum;
1483	sum += chksum_update_16(&oicmph->id, guest_id);
1484	sum = FOLD_U32T(sum);
1485	oicmph->chksum = ~sum;
1486
1487	/* rewrite inner IP header */
1488	#if defined(RT_OS_DARWIN)
1489	/* darwin converts inner length to host byte order too */
1490	IPH_LEN_SET(oiph, htons(IPH_LEN(oiph)));
1491	#endif
1492	sum = (u16_t)~IPH_CHKSUM(oiph);
1493	sum += chksum_update_32((u32_t *)&oiph->src, ip4_addr_get_u32(&guest_ip));
1494	if (target_mapped == PXREMAP_MAPPED) {
1495	sum += chksum_update_32((u32_t *)&oiph->dest, ip4_addr_get_u32(&target_ip));
1496	}
1497	sum = FOLD_U32T(sum);
1498	IPH_CHKSUM_SET(oiph, ~sum);
1499
1500	/* rewrite outer IP header */
1501	sum = (u16_t)~IPH_CHKSUM(iph);
1502	sum += chksum_update_32((u32_t *)&iph->dest, ip4_addr_get_u32(&guest_ip));
1503	if (error_mapped == PXREMAP_MAPPED) {
1504	sum += chksum_update_32((u32_t *)&iph->src, ip4_addr_get_u32(&error_ip));
1505	}
1506	else {
1507	IPH_TTL_SET(iph, IPH_TTL(iph) - 1);
1508	sum += PP_NTOHS(~0x0100);
1509	}
1510	sum = FOLD_U32T(sum);
1511	IPH_CHKSUM_SET(iph, ~sum);
1512
1513	pxping_pmgr_forward_inbound(pxping, iplen);
1514	}
1515
1516
1517	/**
1518	* Process incoming ICMPv6 message for the host.
1519	* NB: we will get a lot of spam here and have to sift through it.
1520	*/
1521	static void
1522	pxping_pmgr_icmp6(struct pxping *pxping)
1523	{
1524	#ifndef RT_OS_WINDOWS
1525	struct msghdr mh;
1526	ssize_t nread;
1527	#else
1528	WSAMSG mh;
1529	DWORD nread;
1530	#endif
1531	IOVEC iov[1];
1532	static u8_t cmsgbuf[128];
1533	struct cmsghdr *cmh;
1534	struct sockaddr_in6 sin6;
1535	/* socklen_t salen = sizeof(sin6); - unused */
1536	struct icmp6_echo_hdr *icmph;
1537	struct in6_pktinfo *pktinfo;
1538	int hopl, tclass;
1539	#ifdef RT_OS_WINDOWS
1540	int status;
1541	#endif
1542
1543	/*
1544	* Reads from raw IPv6 sockets deliver only the payload. Full
1545	* headers are available via recvmsg(2)/cmsg(3).
1546	*/
1547	IOVEC_SET_BASE(iov[0], pollmgr_udpbuf);
1548	IOVEC_SET_LEN(iov[0], sizeof(pollmgr_udpbuf));
1549
1550	memset(&mh, 0, sizeof(mh));
1551	#ifndef RT_OS_WINDOWS
1552	mh.msg_name = &sin6;
1553	mh.msg_namelen = sizeof(sin6);
1554	mh.msg_iov = iov;
1555	mh.msg_iovlen = 1;
1556	mh.msg_control = cmsgbuf;
1557	mh.msg_controllen = sizeof(cmsgbuf);
1558	mh.msg_flags = 0;
1559
1560	nread = recvmsg(pxping->sock6, &mh, 0);
1561	if (nread < 0) {
1562	DPRINTF(("%s: %R[sockerr]\n", __func__, SOCKERRNO()));
1563	return;
1564	}
1565	#else /* RT_OS_WINDOWS */
1566	mh.name = (LPSOCKADDR)&sin6;
1567	mh.namelen = sizeof(sin6);
1568	mh.lpBuffers = iov;
1569	mh.dwBufferCount = 1;
1570	mh.Control.buf = cmsgbuf;
1571	mh.Control.len = sizeof(cmsgbuf);
1572	mh.dwFlags = 0;
1573
1574	status = (*pxping->pfWSARecvMsg6)(pxping->sock6, &mh, &nread, NULL, NULL);
1575	if (status == SOCKET_ERROR) {
1576	DPRINTF2(("%s: error %d\n", __func__, WSAGetLastError()));
1577	return;
1578	}
1579	#endif
1580
1581	icmph = (struct icmp6_echo_hdr *)pollmgr_udpbuf;
1582
1583	DPRINTF2(("%s: %RTnaipv6 ICMPv6: ", __func__, &sin6.sin6_addr));
1584
1585	if (icmph->type == ICMP6_TYPE_EREP) {
1586	DPRINTF2(("echo reply %04x %u\n",
1587	(unsigned int)icmph->id, (unsigned int)icmph->seqno));
1588	}
1589	else { /* XXX */
1590	if (icmph->type == ICMP6_TYPE_EREQ) {
1591	DPRINTF2(("echo request %04x %u\n",
1592	(unsigned int)icmph->id, (unsigned int)icmph->seqno));
1593	}
1594	else if (icmph->type == ICMP6_TYPE_DUR) {
1595	DPRINTF2(("destination unreachable\n"));
1596	}
1597	else if (icmph->type == ICMP6_TYPE_PTB) {
1598	DPRINTF2(("packet too big\n"));
1599	}
1600	else if (icmph->type == ICMP6_TYPE_TE) {
1601	DPRINTF2(("time exceeded\n"));
1602	}
1603	else if (icmph->type == ICMP6_TYPE_PP) {
1604	DPRINTF2(("parameter problem\n"));
1605	}
1606	else {
1607	DPRINTF2(("type %d len %u\n", icmph->type, (unsigned int)nread));
1608	}
1609
1610	if (icmph->type >= ICMP6_TYPE_EREQ) {
1611	return; /* informational message */
1612	}
1613	}
1614
1615	pktinfo = NULL;
1616	hopl = -1;
1617	tclass = -1;
1618	for (cmh = CMSG_FIRSTHDR(&mh); cmh != NULL; cmh = CMSG_NXTHDR(&mh, cmh)) {
1619	if (cmh->cmsg_len == 0)
1620	break;
1621
1622	if (cmh->cmsg_level == IPPROTO_IPV6
1623	&& cmh->cmsg_type == IPV6_HOPLIMIT
1624	&& cmh->cmsg_len == CMSG_LEN(sizeof(int)))
1625	{
1626	hopl = (int )CMSG_DATA(cmh);
1627	DPRINTF2(("hoplimit = %d\n", hopl));
1628	}
1629
1630	if (cmh->cmsg_level == IPPROTO_IPV6
1631	&& cmh->cmsg_type == IPV6_PKTINFO
1632	&& cmh->cmsg_len == CMSG_LEN(sizeof(struct in6_pktinfo)))
1633	{
1634	pktinfo = (struct in6_pktinfo *)CMSG_DATA(cmh);
1635	DPRINTF2(("pktinfo found\n"));
1636	}
1637	}
1638
1639	if (pktinfo == NULL) {
1640	/*
1641	* ip6_output_if() doesn't do checksum for us so we need to
1642	* manually recompute it - for this we must know the
1643	* destination address of the pseudo-header that we will
1644	* rewrite with guest's address. (TODO: yeah, yeah, we can
1645	* compute it from scratch...)
1646	*/
1647	DPRINTF2(("%s: unable to get pktinfo\n", __func__));
1648	return;
1649	}
1650
1651	if (hopl < 0) {
1652	hopl = LWIP_ICMP6_HL;
1653	}
1654
1655	if (icmph->type == ICMP6_TYPE_EREP) {
1656	pxping_pmgr_icmp6_echo(pxping,
1657	(ip6_addr_t *)&sin6.sin6_addr,
1658	(ip6_addr_t *)&pktinfo->ipi6_addr,
1659	hopl, tclass, (u16_t)nread);
1660	}
1661	else if (icmph->type < ICMP6_TYPE_EREQ) {
1662	pxping_pmgr_icmp6_error(pxping,
1663	(ip6_addr_t *)&sin6.sin6_addr,
1664	(ip6_addr_t *)&pktinfo->ipi6_addr,
1665	hopl, tclass, (u16_t)nread);
1666	}
1667	}
1668
1669
1670	/**
1671	* Check if this incoming ICMPv6 echo reply is for one of our pings
1672	* and forward it to the guest.
1673	*/
1674	static void
1675	pxping_pmgr_icmp6_echo(struct pxping *pxping,
1676	ip6_addr_t src, ip6_addr_t dst,
1677	int hopl, int tclass, u16_t icmplen)
1678	{
1679	struct icmp6_echo_hdr *icmph;
1680	ip6_addr_t guest_ip, target_ip;
1681	int mapped;
1682	struct ping_pcb *pcb;
1683	u16_t id, guest_id;
1684	u32_t sum;
1685
1686	ip6_addr_copy(target_ip, *src);
1687	mapped = pxremap_inbound_ip6(&target_ip, &target_ip);
1688	if (mapped == PXREMAP_FAILED) {
1689	return;
1690	}
1691	else if (mapped == PXREMAP_ASIS) {
1692	if (hopl == 1) {
1693	DPRINTF2(("%s: dropping packet with ttl 1\n", __func__));
1694	return;
1695	}
1696	--hopl;
1697	}
1698
1699	icmph = (struct icmp6_echo_hdr *)pollmgr_udpbuf;
1700	id = icmph->id;
1701
1702	sys_mutex_lock(&pxping->lock);
1703	pcb = pxping_pcb_for_reply(pxping, 1, ip6_2_ipX(&target_ip), id);
1704	if (pcb == NULL) {
1705	sys_mutex_unlock(&pxping->lock);
1706	DPRINTF2(("%s: no match\n", __func__));
1707	return;
1708	}
1709
1710	DPRINTF2(("%s: pcb %p\n", __func__, (void *)pcb));
1711
1712	/* save info before unlocking since pcb may expire */
1713	ip6_addr_copy(guest_ip, *ipX_2_ip6(&pcb->src));
1714	guest_id = pcb->guest_id;
1715
1716	sys_mutex_unlock(&pxping->lock);
1717
1718	/* rewrite ICMPv6 echo header */
1719	sum = (u16_t)~icmph->chksum;
1720	sum += chksum_update_16(&icmph->id, guest_id);
1721	sum += chksum_delta_ipv6(dst, &guest_ip); /* pseudo */
1722	if (mapped) {
1723	sum += chksum_delta_ipv6(src, &target_ip); /* pseudo */
1724	}
1725	sum = FOLD_U32T(sum);
1726	icmph->chksum = ~sum;
1727
1728	pxping_pmgr_forward_inbound6(pxping,
1729	&target_ip, /* echo reply src */
1730	&guest_ip, /* echo reply dst */
1731	hopl, tclass, icmplen);
1732	}
1733
1734
1735	/**
1736	* Check if this incoming ICMPv6 error is about one of our pings and
1737	* forward it to the guest.
1738	*/
1739	static void
1740	pxping_pmgr_icmp6_error(struct pxping *pxping,
1741	ip6_addr_t src, ip6_addr_t dst,
1742	int hopl, int tclass, u16_t icmplen)
1743	{
1744	struct icmp6_hdr *icmph;
1745	u8_t *bufptr;
1746	size_t buflen, hlen;
1747	int proto;
1748	struct ip6_hdr *oiph;
1749	struct icmp6_echo_hdr *oicmph;
1750	struct ping_pcb *pcb;
1751	ip6_addr_t guest_ip, target_ip, error_ip;
1752	int target_mapped, error_mapped;
1753	u16_t guest_id;
1754	u32_t sum;
1755
1756	icmph = (struct icmp6_hdr *)pollmgr_udpbuf;
1757
1758	/*
1759	* Inner IP datagram is not checked by the kernel and may be
1760	* anything, possibly malicious.
1761	*/
1762	oiph = NULL;
1763	oicmph = NULL;
1764
1765	bufptr = pollmgr_udpbuf;
1766	buflen = icmplen;
1767
1768	hlen = sizeof(*icmph);
1769	proto = IP6_NEXTH_ENCAPS; /* i.e. IPv6, lwIP's name is unfortuate */
1770	for (;;) {
1771	if (hlen > buflen) {
1772	DPRINTF2(("truncated datagram inside ICMPv6 error message is too short\n"));
1773	return;
1774	}
1775	buflen -= hlen;
1776	bufptr += hlen;
1777
1778	if (proto == IP6_NEXTH_ENCAPS && oiph == NULL) { /* outermost IPv6 */
1779	oiph = (struct ip6_hdr *)bufptr;
1780	if (IP6H_V(oiph) != 6) {
1781	DPRINTF2(("%s: unexpected IP version %d\n", __func__, IP6H_V(oiph)));
1782	return;
1783	}
1784
1785	proto = IP6H_NEXTH(oiph);
1786	hlen = IP6_HLEN;
1787	}
1788	else if (proto == IP6_NEXTH_ICMP6) {
1789	oicmph = (struct icmp6_echo_hdr *)bufptr;
1790	break;
1791	}
1792	else if (proto == IP6_NEXTH_ROUTING
1793	\|\| proto == IP6_NEXTH_HOPBYHOP
1794	\|\| proto == IP6_NEXTH_DESTOPTS)
1795	{
1796	proto = bufptr[0];
1797	hlen = (bufptr[1] + 1) * 8;
1798	}
1799	else {
1800	DPRINTF2(("%s: stopping at protocol %d\n", __func__, proto));
1801	break;
1802	}
1803	}
1804
1805	if (oiph == NULL \|\| oicmph == NULL) {
1806	return;
1807	}
1808
1809	if (buflen < sizeof(*oicmph)) {
1810	DPRINTF2(("%s: original ICMPv6 is truncated too short\n", __func__));
1811	return;
1812	}
1813
1814	if (oicmph->type != ICMP6_TYPE_EREQ) {
1815	DPRINTF2(("%s: ignoring original ICMPv6 type %d\n", __func__, oicmph->type));
1816	return;
1817	}
1818
1819	ip6_addr_copy(target_ip, oiph->dest); /* inner (failed) */
1820	target_mapped = pxremap_inbound_ip6(&target_ip, &target_ip);
1821	if (target_mapped == PXREMAP_FAILED) {
1822	return;
1823	}
1824
1825	sys_mutex_lock(&pxping->lock);
1826	pcb = pxping_pcb_for_reply(pxping, 1, ip6_2_ipX(&target_ip), oicmph->id);
1827	if (pcb == NULL) {
1828	sys_mutex_unlock(&pxping->lock);
1829	DPRINTF2(("%s: no match\n", __func__));
1830	return;
1831	}
1832
1833	DPRINTF2(("%s: pcb %p\n", __func__, (void *)pcb));
1834
1835	/* save info before unlocking since pcb may expire */
1836	ip6_addr_copy(guest_ip, *ipX_2_ip6(&pcb->src));
1837	guest_id = pcb->guest_id;
1838
1839	sys_mutex_unlock(&pxping->lock);
1840
1841
1842	/*
1843	* Rewrite inner and outer headers and forward to guest. Note
1844	* that IPv6 has no IP header checksum, but uses pseudo-header for
1845	* ICMPv6, so we update both in one go, adjusting ICMPv6 checksum
1846	* as we rewrite IP header.
1847	*/
1848
1849	ip6_addr_copy(error_ip, src); / node that reports the error */
1850	error_mapped = pxremap_inbound_ip6(&error_ip, &error_ip);
1851	if (error_mapped == PXREMAP_FAILED) {
1852	return;
1853	}
1854	if (error_mapped == PXREMAP_ASIS && hopl == 1) {
1855	DPRINTF2(("%s: dropping packet with ttl 1\n", __func__));
1856	return;
1857	}
1858
1859	/* rewrite inner ICMPv6 echo header and inner IPv6 header */
1860	sum = (u16_t)~oicmph->chksum;
1861	sum += chksum_update_16(&oicmph->id, guest_id);
1862	sum += chksum_update_ipv6((ip6_addr_t *)&oiph->src, &guest_ip);
1863	if (target_mapped) {
1864	sum += chksum_delta_ipv6((ip6_addr_t *)&oiph->dest, &target_ip);
1865	}
1866	sum = FOLD_U32T(sum);
1867	oicmph->chksum = ~sum;
1868
1869	/* rewrite outer ICMPv6 error header */
1870	sum = (u16_t)~icmph->chksum;
1871	sum += chksum_delta_ipv6(dst, &guest_ip); /* pseudo */
1872	if (error_mapped) {
1873	sum += chksum_delta_ipv6(src, &error_ip); /* pseudo */
1874	}
1875	sum = FOLD_U32T(sum);
1876	icmph->chksum = ~sum;
1877
1878	pxping_pmgr_forward_inbound6(pxping,
1879	&error_ip, /* error src */
1880	&guest_ip, /* error dst */
1881	hopl, tclass, icmplen);
1882	}
1883
1884
1885	/**
1886	* Hand off ICMP datagram to the lwip thread where it will be
1887	* forwarded to the guest.
1888	*
1889	* We no longer need ping_pcb. The pcb may get expired on the lwip
1890	* thread, but we have already patched necessary information into the
1891	* datagram.
1892	*/
1893	static void
1894	pxping_pmgr_forward_inbound(struct pxping *pxping, u16_t iplen)
1895	{
1896	struct pbuf *p;
1897	struct ping_msg *msg;
1898	err_t error;
1899
1900	p = pbuf_alloc(PBUF_LINK, iplen, PBUF_RAM);
1901	if (p == NULL) {
1902	DPRINTF(("%s: pbuf_alloc(%d) failed\n",
1903	__func__, (unsigned int)iplen));
1904	return;
1905	}
1906
1907	error = pbuf_take(p, pollmgr_udpbuf, iplen);
1908	if (error != ERR_OK) {
1909	DPRINTF(("%s: pbuf_take(%d) failed\n",
1910	__func__, (unsigned int)iplen));
1911	pbuf_free(p);
1912	return;
1913	}
1914
1915	msg = (struct ping_msg )malloc(sizeof(msg));
1916	if (msg == NULL) {
1917	pbuf_free(p);
1918	return;
1919	}
1920
1921	msg->msg.type = TCPIP_MSG_CALLBACK_STATIC;
1922	msg->msg.sem = NULL;
1923	msg->msg.msg.cb.function = pxping_pcb_forward_inbound;
1924	msg->msg.msg.cb.ctx = (void *)msg;
1925
1926	msg->pxping = pxping;
1927	msg->p = p;
1928
1929	proxy_lwip_post(&msg->msg);
1930	}
1931
1932
1933	static void
1934	pxping_pcb_forward_inbound(void *arg)
1935	{
1936	struct ping_msg msg = (struct ping_msg )arg;
1937	err_t error;
1938
1939	LWIP_ASSERT1(msg != NULL);
1940	LWIP_ASSERT1(msg->pxping != NULL);
1941	LWIP_ASSERT1(msg->p != NULL);
1942
1943	error = ip_raw_output_if(msg->p, msg->pxping->netif);
1944	if (error != ERR_OK) {
1945	DPRINTF(("%s: ip_output_if: %s\n",
1946	__func__, proxy_lwip_strerr(error)));
1947	}
1948	pbuf_free(msg->p);
1949	free(msg);
1950	}
1951
1952
1953	static void
1954	pxping_pmgr_forward_inbound6(struct pxping *pxping,
1955	ip6_addr_t src, ip6_addr_t dst,
1956	u8_t hopl, u8_t tclass,
1957	u16_t icmplen)
1958	{
1959	struct pbuf *p;
1960	struct ping6_msg *msg;
1961
1962	err_t error;
1963
1964	p = pbuf_alloc(PBUF_IP, icmplen, PBUF_RAM);
1965	if (p == NULL) {
1966	DPRINTF(("%s: pbuf_alloc(%d) failed\n",
1967	__func__, (unsigned int)icmplen));
1968	return;
1969	}
1970
1971	error = pbuf_take(p, pollmgr_udpbuf, icmplen);
1972	if (error != ERR_OK) {
1973	DPRINTF(("%s: pbuf_take(%d) failed\n",
1974	__func__, (unsigned int)icmplen));
1975	pbuf_free(p);
1976	return;
1977	}
1978
1979	msg = (struct ping6_msg )malloc(sizeof(msg));
1980	if (msg == NULL) {
1981	pbuf_free(p);
1982	return;
1983	}
1984
1985	msg->msg.type = TCPIP_MSG_CALLBACK_STATIC;
1986	msg->msg.sem = NULL;
1987	msg->msg.msg.cb.function = pxping_pcb_forward_inbound6;
1988	msg->msg.msg.cb.ctx = (void *)msg;
1989
1990	msg->pxping = pxping;
1991	msg->p = p;
1992	ip6_addr_copy(msg->src, *src);
1993	ip6_addr_copy(msg->dst, *dst);
1994	msg->hopl = hopl;
1995	msg->tclass = tclass;
1996
1997	proxy_lwip_post(&msg->msg);
1998	}
1999
2000
2001	static void
2002	pxping_pcb_forward_inbound6(void *arg)
2003	{
2004	struct ping6_msg msg = (struct ping6_msg )arg;
2005	err_t error;
2006
2007	LWIP_ASSERT1(msg != NULL);
2008	LWIP_ASSERT1(msg->pxping != NULL);
2009	LWIP_ASSERT1(msg->p != NULL);
2010
2011	error = ip6_output_if(msg->p,
2012	&msg->src, &msg->dst, msg->hopl, msg->tclass,
2013	IP6_NEXTH_ICMP6, msg->pxping->netif);
2014	if (error != ERR_OK) {
2015	DPRINTF(("%s: ip6_output_if: %s\n",
2016	__func__, proxy_lwip_strerr(error)));
2017	}
2018	pbuf_free(msg->p);
2019	free(msg);
2020	}

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

source: vbox/trunk/src/VBox/NetworkServices/NAT/pxping.c@ 100684

Download in other formats: