1 /*
2  * Copyright (c) 2011 Jakub Zawadzki
3  * All rights reserved.
4  *
5  * Redistribution and use in source and binary forms, with or without
6  * modification, are permitted provided that the following conditions
7  * are met:
8  *
9  * 1. Redistributions of source code must retain the above copyright
10  * notice, this list of conditions and the following disclaimer.
11  * 2. Redistributions in binary form must reproduce the above copyright
12  * notice, this list of conditions and the following disclaimer in the
13  * documentation and/or other materials provided with the distribution.
14  * 3. The name of the author may not be used to endorse or promote
15  * products derived from this software without specific prior written
16  * permission.
17  *
18  * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
19  * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
20  * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
21  * A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
22  * OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
23  * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
24  * LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
25  * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
26  * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
27  * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
28  * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
29  */
30 
31 #ifdef HAVE_CONFIG_H
32 #include <config.h>
33 #endif
34 
35 #include "pcap-int.h"
36 
37 #ifdef NEED_STRERROR_H
38 #include "strerror.h"
39 #endif
40 
41 #include <errno.h>
42 #include <stdlib.h>
43 #include <unistd.h>
44 #include <string.h>
45 #include <sys/socket.h>
46 #include <arpa/inet.h>
47 
48 #include <time.h>
49 #include <sys/time.h>
50 #include <netinet/in.h>
51 #include <linux/types.h>
52 
53 #include <linux/netlink.h>
54 #include <linux/netfilter.h>
55 #include <linux/netfilter/nfnetlink.h>
56 #include <linux/netfilter/nfnetlink_log.h>
57 #include <linux/netfilter/nfnetlink_queue.h>
58 
59 /* NOTE: if your program drops privileges after pcap_activate() it WON'T work with nfqueue.
60  *       It took me quite some time to debug ;/
61  *
62  *       Sending any data to nfnetlink socket requires CAP_NET_ADMIN privileges,
63  *       and in nfqueue we need to send verdict reply after recving packet.
64  *
65  *       In tcpdump you can disable dropping privileges with -Z root
66  */
67 
68 #include "pcap-netfilter-linux.h"
69 
70 #define HDR_LENGTH (NLMSG_LENGTH(NLMSG_ALIGN(sizeof(struct nfgenmsg))))
71 
72 #define NFLOG_IFACE "nflog"
73 #define NFQUEUE_IFACE "nfqueue"
74 
75 typedef enum { OTHER = -1, NFLOG, NFQUEUE } nftype_t;
76 
77 /*
78  * Private data for capturing on Linux netfilter sockets.
79  */
80 struct pcap_netfilter {
81 	u_int	packets_read;	/* count of packets read with recvfrom() */
82 	u_int   packets_nobufs; /* ENOBUFS counter */
83 };
84 
85 static int nfqueue_send_verdict(const pcap_t *handle, uint16_t group_id, u_int32_t id, u_int32_t verdict);
86 
87 
88 static int
netfilter_read_linux(pcap_t * handle,int max_packets,pcap_handler callback,u_char * user)89 netfilter_read_linux(pcap_t *handle, int max_packets, pcap_handler callback, u_char *user)
90 {
91 	struct pcap_netfilter *handlep = handle->priv;
92 	register u_char *bp, *ep;
93 	int count = 0;
94 	ssize_t len;
95 
96 	/*
97 	 * Has "pcap_breakloop()" been called?
98 	 */
99 	if (handle->break_loop) {
100 		/*
101 		 * Yes - clear the flag that indicates that it
102 		 * has, and return PCAP_ERROR_BREAK to indicate
103 		 * that we were told to break out of the loop.
104 		 */
105 		handle->break_loop = 0;
106 		return PCAP_ERROR_BREAK;
107 	}
108 	len = handle->cc;
109 	if (len == 0) {
110 		/*
111 		 * The buffer is empty; refill it.
112 		 *
113 		 * We ignore EINTR, as that might just be due to a signal
114 		 * being delivered - if the signal should interrupt the
115 		 * loop, the signal handler should call pcap_breakloop()
116 		 * to set handle->break_loop (we ignore it on other
117 		 * platforms as well).
118 		 */
119 		do {
120 			len = recv(handle->fd, handle->buffer, handle->bufsize, 0);
121 			if (handle->break_loop) {
122 				handle->break_loop = 0;
123 				return PCAP_ERROR_BREAK;
124 			}
125 			if (errno == ENOBUFS)
126 				handlep->packets_nobufs++;
127 		} while ((len == -1) && (errno == EINTR || errno == ENOBUFS));
128 
129 		if (len < 0) {
130 			pcap_fmt_errmsg_for_errno(handle->errbuf,
131 			    PCAP_ERRBUF_SIZE, errno, "Can't receive packet");
132 			return PCAP_ERROR;
133 		}
134 
135 		bp = (unsigned char *)handle->buffer;
136 	} else
137 		bp = handle->bp;
138 	ep = bp + len;
139 	while (bp < ep) {
140 		const struct nlmsghdr *nlh = (const struct nlmsghdr *) bp;
141 		uint32_t msg_len;
142 		nftype_t type = OTHER;
143 		/*
144 		 * Has "pcap_breakloop()" been called?
145 		 * If so, return immediately - if we haven't read any
146 		 * packets, clear the flag and return PCAP_ERROR_BREAK
147 		 * to indicate that we were told to break out of the loop,
148 		 * otherwise leave the flag set, so that the *next* call
149 		 * will break out of the loop without having read any
150 		 * packets, and return the number of packets we've
151 		 * processed so far.
152 		 */
153 		if (handle->break_loop) {
154 			handle->bp = bp;
155 			handle->cc = (int)(ep - bp);
156 			if (count == 0) {
157 				handle->break_loop = 0;
158 				return PCAP_ERROR_BREAK;
159 			} else
160 				return count;
161 		}
162 		/*
163 		 * NLMSG_SPACE(0) might be signed or might be unsigned,
164 		 * depending on whether the kernel defines NLMSG_ALIGNTO
165 		 * as 4, which older kernels do, or as 4U, which newer
166 		 * kernels do.
167 		 *
168 		 * ep - bp is of type ptrdiff_t, which is signed.
169 		 *
170 		 * To squelch warnings, we cast both to size_t, which
171 		 * is unsigned; ep >= bp, so the cast is safe.
172 		 */
173 		if ((size_t)(ep - bp) < (size_t)NLMSG_SPACE(0)) {
174 			/*
175 			 * There's less than one netlink message left
176 			 * in the buffer.  Give up.
177 			 */
178 			break;
179 		}
180 
181 		if (nlh->nlmsg_len < sizeof(struct nlmsghdr) || (u_int)len < nlh->nlmsg_len) {
182 			snprintf(handle->errbuf, PCAP_ERRBUF_SIZE, "Message truncated: (got: %zd) (nlmsg_len: %u)", len, nlh->nlmsg_len);
183 			return -1;
184 		}
185 
186 		if (NFNL_SUBSYS_ID(nlh->nlmsg_type) == NFNL_SUBSYS_ULOG &&
187 		    NFNL_MSG_TYPE(nlh->nlmsg_type) == NFULNL_MSG_PACKET)
188 			type = NFLOG;
189 		else if (NFNL_SUBSYS_ID(nlh->nlmsg_type) == NFNL_SUBSYS_QUEUE &&
190 		         NFNL_MSG_TYPE(nlh->nlmsg_type) == NFQNL_MSG_PACKET)
191 			type = NFQUEUE;
192 
193 		if (type != OTHER) {
194 			const unsigned char *payload = NULL;
195 			struct pcap_pkthdr pkth;
196 
197 			const struct nfgenmsg *nfg = NULL;
198 			int id = 0;
199 
200 			if (handle->linktype != DLT_NFLOG) {
201 				const struct nfattr *payload_attr = NULL;
202 
203 				if (nlh->nlmsg_len < HDR_LENGTH) {
204 					snprintf(handle->errbuf, PCAP_ERRBUF_SIZE, "Malformed message: (nlmsg_len: %u)", nlh->nlmsg_len);
205 					return -1;
206 				}
207 
208 				nfg = NLMSG_DATA(nlh);
209 				if (nlh->nlmsg_len > HDR_LENGTH) {
210 					struct nfattr *attr = NFM_NFA(nfg);
211 					int attr_len = nlh->nlmsg_len - NLMSG_ALIGN(HDR_LENGTH);
212 
213 					while (NFA_OK(attr, attr_len)) {
214 						if (type == NFQUEUE) {
215 							switch (NFA_TYPE(attr)) {
216 								case NFQA_PACKET_HDR:
217 									{
218 										const struct nfqnl_msg_packet_hdr *pkt_hdr = (const struct nfqnl_msg_packet_hdr *) NFA_DATA(attr);
219 
220 										id = ntohl(pkt_hdr->packet_id);
221 										break;
222 									}
223 								case NFQA_PAYLOAD:
224 									payload_attr = attr;
225 									break;
226 							}
227 
228 						} else if (type == NFLOG) {
229 							switch (NFA_TYPE(attr)) {
230 								case NFULA_PAYLOAD:
231 									payload_attr = attr;
232 									break;
233 							}
234 						}
235 						attr = NFA_NEXT(attr, attr_len);
236 					}
237 				}
238 
239 				if (payload_attr) {
240 					payload = NFA_DATA(payload_attr);
241 					pkth.len = pkth.caplen = NFA_PAYLOAD(payload_attr);
242 				}
243 
244 			} else {
245 				payload = NLMSG_DATA(nlh);
246 				pkth.caplen = pkth.len = nlh->nlmsg_len-NLMSG_ALIGN(sizeof(struct nlmsghdr));
247 			}
248 
249 			if (payload) {
250 				/* pkth.caplen = min (payload_len, handle->snapshot); */
251 
252 				gettimeofday(&pkth.ts, NULL);
253 				if (handle->fcode.bf_insns == NULL ||
254 						pcap_filter(handle->fcode.bf_insns, payload, pkth.len, pkth.caplen))
255 				{
256 					handlep->packets_read++;
257 					callback(user, &pkth, payload);
258 					count++;
259 				}
260 			}
261 
262 			if (type == NFQUEUE) {
263 				/* XXX, possible responses: NF_DROP, NF_ACCEPT, NF_STOLEN, NF_QUEUE, NF_REPEAT, NF_STOP */
264 				/* if type == NFQUEUE, handle->linktype is always != DLT_NFLOG,
265 				   so nfg is always initialized to NLMSG_DATA(nlh). */
266 				if (nfg != NULL)
267 					nfqueue_send_verdict(handle, ntohs(nfg->res_id), id, NF_ACCEPT);
268 			}
269 		}
270 
271 		msg_len = NLMSG_ALIGN(nlh->nlmsg_len);
272 		/*
273 		 * If the message length would run past the end of the
274 		 * buffer, truncate it to the remaining space in the
275 		 * buffer.
276 		 *
277 		 * To squelch warnings, we cast ep - bp to uint32_t, which
278 		 * is unsigned and is the type of msg_len; ep >= bp, and
279 		 * len should fit in 32 bits (either it's set from an int
280 		 * or it's set from a recv() call with a buffer size that's
281 		 * an int, and we're assuming either ILP32 or LP64), so
282 		 * the cast is safe.
283 		 */
284 		if (msg_len > (uint32_t)(ep - bp))
285 			msg_len = (uint32_t)(ep - bp);
286 
287 		bp += msg_len;
288 		if (count >= max_packets && !PACKET_COUNT_IS_UNLIMITED(max_packets)) {
289 			handle->bp = bp;
290 			handle->cc = (int)(ep - bp);
291 			if (handle->cc < 0)
292 				handle->cc = 0;
293 			return count;
294 		}
295 	}
296 
297 	handle->cc = 0;
298 	return count;
299 }
300 
301 static int
netfilter_set_datalink(pcap_t * handle,int dlt)302 netfilter_set_datalink(pcap_t *handle, int dlt)
303 {
304 	handle->linktype = dlt;
305 	return 0;
306 }
307 
308 static int
netfilter_stats_linux(pcap_t * handle,struct pcap_stat * stats)309 netfilter_stats_linux(pcap_t *handle, struct pcap_stat *stats)
310 {
311 	struct pcap_netfilter *handlep = handle->priv;
312 
313 	stats->ps_recv = handlep->packets_read;
314 	stats->ps_drop = handlep->packets_nobufs;
315 	stats->ps_ifdrop = 0;
316 	return 0;
317 }
318 
319 static int
netfilter_inject_linux(pcap_t * handle,const void * buf _U_,int size _U_)320 netfilter_inject_linux(pcap_t *handle, const void *buf _U_, int size _U_)
321 {
322 	snprintf(handle->errbuf, PCAP_ERRBUF_SIZE,
323 	    "Packet injection is not supported on netfilter devices");
324 	return (-1);
325 }
326 
327 struct my_nfattr {
328 	uint16_t nfa_len;
329 	uint16_t nfa_type;
330 	void *data;
331 };
332 
333 static int
netfilter_send_config_msg(const pcap_t * handle,uint16_t msg_type,int ack,u_int8_t family,u_int16_t res_id,const struct my_nfattr * mynfa)334 netfilter_send_config_msg(const pcap_t *handle, uint16_t msg_type, int ack, u_int8_t family, u_int16_t res_id, const struct my_nfattr *mynfa)
335 {
336 	char buf[1024] __attribute__ ((aligned));
337 	memset(buf, 0, sizeof(buf));
338 
339 	struct nlmsghdr *nlh = (struct nlmsghdr *) buf;
340 	struct nfgenmsg *nfg = (struct nfgenmsg *) (buf + sizeof(struct nlmsghdr));
341 
342 	struct sockaddr_nl snl;
343 	static unsigned int seq_id;
344 
345 	if (!seq_id)
346 		seq_id = time(NULL);
347 	++seq_id;
348 
349 	nlh->nlmsg_len = NLMSG_LENGTH(sizeof(struct nfgenmsg));
350 	nlh->nlmsg_type = msg_type;
351 	nlh->nlmsg_flags = NLM_F_REQUEST | (ack ? NLM_F_ACK : 0);
352 	nlh->nlmsg_pid = 0;	/* to kernel */
353 	nlh->nlmsg_seq = seq_id;
354 
355 	nfg->nfgen_family = family;
356 	nfg->version = NFNETLINK_V0;
357 	nfg->res_id = htons(res_id);
358 
359 	if (mynfa) {
360 		struct nfattr *nfa = (struct nfattr *) (buf + NLMSG_ALIGN(nlh->nlmsg_len));
361 
362 		nfa->nfa_type = mynfa->nfa_type;
363 		nfa->nfa_len = NFA_LENGTH(mynfa->nfa_len);
364 		memcpy(NFA_DATA(nfa), mynfa->data, mynfa->nfa_len);
365 		nlh->nlmsg_len = NLMSG_ALIGN(nlh->nlmsg_len) + NFA_ALIGN(nfa->nfa_len);
366 	}
367 
368 	memset(&snl, 0, sizeof(snl));
369 	snl.nl_family = AF_NETLINK;
370 
371 	if (sendto(handle->fd, nlh, nlh->nlmsg_len, 0, (struct sockaddr *) &snl, sizeof(snl)) == -1)
372 		return -1;
373 
374 	if (!ack)
375 		return 0;
376 
377 	/* waiting for reply loop */
378 	do {
379 		socklen_t addrlen = sizeof(snl);
380 		int len;
381 
382 		/* ignore interrupt system call error */
383 		do {
384 			/*
385 			 * The buffer is not so big that its size won't
386 			 * fit into an int.
387 			 */
388 			len = (int)recvfrom(handle->fd, buf, sizeof(buf), 0, (struct sockaddr *) &snl, &addrlen);
389 		} while ((len == -1) && (errno == EINTR));
390 
391 		if (len <= 0)
392 			return len;
393 
394 		if (addrlen != sizeof(snl) || snl.nl_family != AF_NETLINK) {
395 			errno = EINVAL;
396 			return -1;
397 		}
398 
399 		nlh = (struct nlmsghdr *) buf;
400 		if (snl.nl_pid != 0 || seq_id != nlh->nlmsg_seq)	/* if not from kernel or wrong sequence skip */
401 			continue;
402 
403 		while ((u_int)len >= NLMSG_SPACE(0) && NLMSG_OK(nlh, (u_int)len)) {
404 			if (nlh->nlmsg_type == NLMSG_ERROR || (nlh->nlmsg_type == NLMSG_DONE && nlh->nlmsg_flags & NLM_F_MULTI)) {
405 				if (nlh->nlmsg_len < NLMSG_ALIGN(sizeof(struct nlmsgerr))) {
406 					errno = EBADMSG;
407 					return -1;
408 				}
409 				errno = -(*((int *)NLMSG_DATA(nlh)));
410 				return (errno == 0) ? 0 : -1;
411 			}
412 			nlh = NLMSG_NEXT(nlh, len);
413 		}
414 	} while (1);
415 
416 	return -1; /* never here */
417 }
418 
419 static int
nflog_send_config_msg(const pcap_t * handle,uint8_t family,u_int16_t group_id,const struct my_nfattr * mynfa)420 nflog_send_config_msg(const pcap_t *handle, uint8_t family, u_int16_t group_id, const struct my_nfattr *mynfa)
421 {
422 	return netfilter_send_config_msg(handle, (NFNL_SUBSYS_ULOG << 8) | NFULNL_MSG_CONFIG, 1, family, group_id, mynfa);
423 }
424 
425 static int
nflog_send_config_cmd(const pcap_t * handle,uint16_t group_id,u_int8_t cmd,u_int8_t family)426 nflog_send_config_cmd(const pcap_t *handle, uint16_t group_id, u_int8_t cmd, u_int8_t family)
427 {
428 	struct nfulnl_msg_config_cmd msg;
429 	struct my_nfattr nfa;
430 
431 	msg.command = cmd;
432 
433 	nfa.data = &msg;
434 	nfa.nfa_type = NFULA_CFG_CMD;
435 	nfa.nfa_len = sizeof(msg);
436 
437 	return nflog_send_config_msg(handle, family, group_id, &nfa);
438 }
439 
440 static int
nflog_send_config_mode(const pcap_t * handle,uint16_t group_id,u_int8_t copy_mode,u_int32_t copy_range)441 nflog_send_config_mode(const pcap_t *handle, uint16_t group_id, u_int8_t copy_mode, u_int32_t copy_range)
442 {
443 	struct nfulnl_msg_config_mode msg;
444 	struct my_nfattr nfa;
445 
446 	msg.copy_range = htonl(copy_range);
447 	msg.copy_mode = copy_mode;
448 
449 	nfa.data = &msg;
450 	nfa.nfa_type = NFULA_CFG_MODE;
451 	nfa.nfa_len = sizeof(msg);
452 
453 	return nflog_send_config_msg(handle, AF_UNSPEC, group_id, &nfa);
454 }
455 
456 static int
nfqueue_send_verdict(const pcap_t * handle,uint16_t group_id,u_int32_t id,u_int32_t verdict)457 nfqueue_send_verdict(const pcap_t *handle, uint16_t group_id, u_int32_t id, u_int32_t verdict)
458 {
459 	struct nfqnl_msg_verdict_hdr msg;
460 	struct my_nfattr nfa;
461 
462 	msg.id = htonl(id);
463 	msg.verdict = htonl(verdict);
464 
465 	nfa.data = &msg;
466 	nfa.nfa_type = NFQA_VERDICT_HDR;
467 	nfa.nfa_len = sizeof(msg);
468 
469 	return netfilter_send_config_msg(handle, (NFNL_SUBSYS_QUEUE << 8) | NFQNL_MSG_VERDICT, 0, AF_UNSPEC, group_id, &nfa);
470 }
471 
472 static int
nfqueue_send_config_msg(const pcap_t * handle,uint8_t family,u_int16_t group_id,const struct my_nfattr * mynfa)473 nfqueue_send_config_msg(const pcap_t *handle, uint8_t family, u_int16_t group_id, const struct my_nfattr *mynfa)
474 {
475 	return netfilter_send_config_msg(handle, (NFNL_SUBSYS_QUEUE << 8) | NFQNL_MSG_CONFIG, 1, family, group_id, mynfa);
476 }
477 
478 static int
nfqueue_send_config_cmd(const pcap_t * handle,uint16_t group_id,u_int8_t cmd,u_int16_t pf)479 nfqueue_send_config_cmd(const pcap_t *handle, uint16_t group_id, u_int8_t cmd, u_int16_t pf)
480 {
481 	struct nfqnl_msg_config_cmd msg;
482 	struct my_nfattr nfa;
483 
484 	msg.command = cmd;
485 	msg.pf = htons(pf);
486 
487 	nfa.data = &msg;
488 	nfa.nfa_type = NFQA_CFG_CMD;
489 	nfa.nfa_len = sizeof(msg);
490 
491 	return nfqueue_send_config_msg(handle, AF_UNSPEC, group_id, &nfa);
492 }
493 
494 static int
nfqueue_send_config_mode(const pcap_t * handle,uint16_t group_id,u_int8_t copy_mode,u_int32_t copy_range)495 nfqueue_send_config_mode(const pcap_t *handle, uint16_t group_id, u_int8_t copy_mode, u_int32_t copy_range)
496 {
497 	struct nfqnl_msg_config_params msg;
498 	struct my_nfattr nfa;
499 
500 	msg.copy_range = htonl(copy_range);
501 	msg.copy_mode = copy_mode;
502 
503 	nfa.data = &msg;
504 	nfa.nfa_type = NFQA_CFG_PARAMS;
505 	nfa.nfa_len = sizeof(msg);
506 
507 	return nfqueue_send_config_msg(handle, AF_UNSPEC, group_id, &nfa);
508 }
509 
510 static int
netfilter_activate(pcap_t * handle)511 netfilter_activate(pcap_t* handle)
512 {
513 	const char *dev = handle->opt.device;
514 	unsigned short groups[32];
515 	int group_count = 0;
516 	nftype_t type = OTHER;
517 	int i;
518 
519  	if (strncmp(dev, NFLOG_IFACE, strlen(NFLOG_IFACE)) == 0) {
520  		dev += strlen(NFLOG_IFACE);
521 		type = NFLOG;
522 
523 	} else if (strncmp(dev, NFQUEUE_IFACE, strlen(NFQUEUE_IFACE)) == 0) {
524 		dev += strlen(NFQUEUE_IFACE);
525 		type = NFQUEUE;
526 	}
527 
528 	if (type != OTHER && *dev == ':') {
529 		dev++;
530 		while (*dev) {
531 			long int group_id;
532 			char *end_dev;
533 
534 			if (group_count == 32) {
535 				snprintf(handle->errbuf, PCAP_ERRBUF_SIZE,
536 						"Maximum 32 netfilter groups! dev: %s",
537 						handle->opt.device);
538 				return PCAP_ERROR;
539 			}
540 
541 			group_id = strtol(dev, &end_dev, 0);
542 			if (end_dev != dev) {
543 				if (group_id < 0 || group_id > 65535) {
544 					snprintf(handle->errbuf, PCAP_ERRBUF_SIZE,
545 							"Netfilter group range from 0 to 65535 (got %ld)",
546 							group_id);
547 					return PCAP_ERROR;
548 				}
549 
550 				groups[group_count++] = (unsigned short) group_id;
551 				dev = end_dev;
552 			}
553 			if (*dev != ',')
554 				break;
555 			dev++;
556 		}
557 	}
558 
559 	if (type == OTHER || *dev) {
560 		snprintf(handle->errbuf, PCAP_ERRBUF_SIZE,
561 				"Can't get netfilter group(s) index from %s",
562 				handle->opt.device);
563 		return PCAP_ERROR;
564 	}
565 
566 	/* if no groups, add default: 0 */
567 	if (!group_count) {
568 		groups[0] = 0;
569 		group_count = 1;
570 	}
571 
572 	/*
573 	 * Turn a negative snapshot value (invalid), a snapshot value of
574 	 * 0 (unspecified), or a value bigger than the normal maximum
575 	 * value, into the maximum allowed value.
576 	 *
577 	 * If some application really *needs* a bigger snapshot
578 	 * length, we should just increase MAXIMUM_SNAPLEN.
579 	 */
580 	if (handle->snapshot <= 0 || handle->snapshot > MAXIMUM_SNAPLEN)
581 		handle->snapshot = MAXIMUM_SNAPLEN;
582 
583 	/* Initialize some components of the pcap structure. */
584 	handle->bufsize = 128 + handle->snapshot;
585 	handle->offset = 0;
586 	handle->read_op = netfilter_read_linux;
587 	handle->inject_op = netfilter_inject_linux;
588 	handle->setfilter_op = install_bpf_program; /* no kernel filtering */
589 	handle->setdirection_op = NULL;
590 	handle->set_datalink_op = netfilter_set_datalink;
591 	handle->getnonblock_op = pcap_getnonblock_fd;
592 	handle->setnonblock_op = pcap_setnonblock_fd;
593 	handle->stats_op = netfilter_stats_linux;
594 
595 	/* Create netlink socket */
596 	handle->fd = socket(AF_NETLINK, SOCK_RAW, NETLINK_NETFILTER);
597 	if (handle->fd < 0) {
598 		pcap_fmt_errmsg_for_errno(handle->errbuf, PCAP_ERRBUF_SIZE,
599 		    errno, "Can't create raw socket");
600 		return PCAP_ERROR;
601 	}
602 
603 	if (type == NFLOG) {
604 		handle->linktype = DLT_NFLOG;
605 		handle->dlt_list = (u_int *) malloc(sizeof(u_int) * 2);
606 		if (handle->dlt_list != NULL) {
607 			handle->dlt_list[0] = DLT_NFLOG;
608 			handle->dlt_list[1] = DLT_IPV4;
609 			handle->dlt_count = 2;
610 		}
611 
612 	} else
613 		handle->linktype = DLT_IPV4;
614 
615 	handle->buffer = malloc(handle->bufsize);
616 	if (!handle->buffer) {
617 		pcap_fmt_errmsg_for_errno(handle->errbuf, PCAP_ERRBUF_SIZE,
618 		    errno, "Can't allocate dump buffer");
619 		goto close_fail;
620 	}
621 
622 	if (type == NFLOG) {
623 		if (nflog_send_config_cmd(handle, 0, NFULNL_CFG_CMD_PF_UNBIND, AF_INET) < 0) {
624 			pcap_fmt_errmsg_for_errno(handle->errbuf,
625 			    PCAP_ERRBUF_SIZE, errno,
626 			    "NFULNL_CFG_CMD_PF_UNBIND");
627 			goto close_fail;
628 		}
629 
630 		if (nflog_send_config_cmd(handle, 0, NFULNL_CFG_CMD_PF_BIND, AF_INET) < 0) {
631 			pcap_fmt_errmsg_for_errno(handle->errbuf,
632 			    PCAP_ERRBUF_SIZE, errno, "NFULNL_CFG_CMD_PF_BIND");
633 			goto close_fail;
634 		}
635 
636 		/* Bind socket to the nflog groups */
637 		for (i = 0; i < group_count; i++) {
638 			if (nflog_send_config_cmd(handle, groups[i], NFULNL_CFG_CMD_BIND, AF_UNSPEC) < 0) {
639 				pcap_fmt_errmsg_for_errno(handle->errbuf,
640 				    PCAP_ERRBUF_SIZE, errno,
641 				    "Can't listen on group index");
642 				goto close_fail;
643 			}
644 
645 			if (nflog_send_config_mode(handle, groups[i], NFULNL_COPY_PACKET, handle->snapshot) < 0) {
646 				pcap_fmt_errmsg_for_errno(handle->errbuf,
647 				    PCAP_ERRBUF_SIZE, errno,
648 				    "NFULNL_COPY_PACKET");
649 				goto close_fail;
650 			}
651 		}
652 
653 	} else {
654 		if (nfqueue_send_config_cmd(handle, 0, NFQNL_CFG_CMD_PF_UNBIND, AF_INET) < 0) {
655 			pcap_fmt_errmsg_for_errno(handle->errbuf,
656 			    PCAP_ERRBUF_SIZE, errno, "NFQNL_CFG_CMD_PF_UNBIND");
657 			goto close_fail;
658 		}
659 
660 		if (nfqueue_send_config_cmd(handle, 0, NFQNL_CFG_CMD_PF_BIND, AF_INET) < 0) {
661 			pcap_fmt_errmsg_for_errno(handle->errbuf,
662 			    PCAP_ERRBUF_SIZE, errno, "NFQNL_CFG_CMD_PF_BIND");
663 			goto close_fail;
664 		}
665 
666 		/* Bind socket to the nfqueue groups */
667 		for (i = 0; i < group_count; i++) {
668 			if (nfqueue_send_config_cmd(handle, groups[i], NFQNL_CFG_CMD_BIND, AF_UNSPEC) < 0) {
669 				pcap_fmt_errmsg_for_errno(handle->errbuf,
670 				    PCAP_ERRBUF_SIZE, errno,
671 				    "Can't listen on group index");
672 				goto close_fail;
673 			}
674 
675 			if (nfqueue_send_config_mode(handle, groups[i], NFQNL_COPY_PACKET, handle->snapshot) < 0) {
676 				pcap_fmt_errmsg_for_errno(handle->errbuf,
677 				    PCAP_ERRBUF_SIZE, errno,
678 				    "NFQNL_COPY_PACKET");
679 				goto close_fail;
680 			}
681 		}
682 	}
683 
684 	if (handle->opt.rfmon) {
685 		/*
686 		 * Monitor mode doesn't apply to netfilter devices.
687 		 */
688 		pcap_cleanup_live_common(handle);
689 		return PCAP_ERROR_RFMON_NOTSUP;
690 	}
691 
692 	if (handle->opt.buffer_size != 0) {
693 		/*
694 		 * Set the socket buffer size to the specified value.
695 		 */
696 		if (setsockopt(handle->fd, SOL_SOCKET, SO_RCVBUF, &handle->opt.buffer_size, sizeof(handle->opt.buffer_size)) == -1) {
697 			pcap_fmt_errmsg_for_errno(handle->errbuf,
698 			    PCAP_ERRBUF_SIZE, errno, "SO_RCVBUF");
699 			goto close_fail;
700 		}
701 	}
702 
703 	handle->selectable_fd = handle->fd;
704 	return 0;
705 
706 close_fail:
707 	pcap_cleanup_live_common(handle);
708 	return PCAP_ERROR;
709 }
710 
711 pcap_t *
netfilter_create(const char * device,char * ebuf,int * is_ours)712 netfilter_create(const char *device, char *ebuf, int *is_ours)
713 {
714 	const char *cp;
715 	pcap_t *p;
716 
717 	/* Does this look like an netfilter device? */
718 	cp = strrchr(device, '/');
719 	if (cp == NULL)
720 		cp = device;
721 
722 	/* Does it begin with NFLOG_IFACE or NFQUEUE_IFACE? */
723 	if (strncmp(cp, NFLOG_IFACE, sizeof NFLOG_IFACE - 1) == 0)
724 		cp += sizeof NFLOG_IFACE - 1;
725 	else if (strncmp(cp, NFQUEUE_IFACE, sizeof NFQUEUE_IFACE - 1) == 0)
726 		cp += sizeof NFQUEUE_IFACE - 1;
727 	else {
728 		/* Nope, doesn't begin with NFLOG_IFACE nor NFQUEUE_IFACE */
729 		*is_ours = 0;
730 		return NULL;
731 	}
732 
733 	/*
734 	 * Yes - is that either the end of the name, or is it followed
735 	 * by a colon?
736 	 */
737 	if (*cp != ':' && *cp != '\0') {
738 		/* Nope */
739 		*is_ours = 0;
740 		return NULL;
741 	}
742 
743 	/* OK, it's probably ours. */
744 	*is_ours = 1;
745 
746 	p = PCAP_CREATE_COMMON(ebuf, struct pcap_netfilter);
747 	if (p == NULL)
748 		return (NULL);
749 
750 	p->activate_op = netfilter_activate;
751 	return (p);
752 }
753 
754 int
netfilter_findalldevs(pcap_if_list_t * devlistp,char * err_str)755 netfilter_findalldevs(pcap_if_list_t *devlistp, char *err_str)
756 {
757 	int sock;
758 
759 	sock = socket(AF_NETLINK, SOCK_RAW, NETLINK_NETFILTER);
760 	if (sock < 0) {
761 		/* if netlink is not supported this is not fatal */
762 		if (errno == EAFNOSUPPORT || errno == EPROTONOSUPPORT)
763 			return 0;
764 		pcap_fmt_errmsg_for_errno(err_str, PCAP_ERRBUF_SIZE,
765 		    errno, "Can't open netlink socket");
766 		return -1;
767 	}
768 	close(sock);
769 
770 	/*
771 	 * The notion of "connected" vs. "disconnected" doesn't apply.
772 	 * XXX - what about "up" and "running"?
773 	 */
774 	if (add_dev(devlistp, NFLOG_IFACE,
775 	    PCAP_IF_CONNECTION_STATUS_NOT_APPLICABLE,
776 	    "Linux netfilter log (NFLOG) interface", err_str) == NULL)
777 		return -1;
778 	if (add_dev(devlistp, NFQUEUE_IFACE,
779 	    PCAP_IF_CONNECTION_STATUS_NOT_APPLICABLE,
780 	    "Linux netfilter queue (NFQUEUE) interface", err_str) == NULL)
781 		return -1;
782 	return 0;
783 }
784