Home | History | Annotate | Download | only in libpcap
      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 privilages 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 privilages,
     63  *       and in nfqueue we need to send verdict reply after recving packet.
     64  *
     65  *       In tcpdump you can disable dropping privilages 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
     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 	int 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 = ep - bp;
    156 			if (count == 0) {
    157 				handle->break_loop = 0;
    158 				return PCAP_ERROR_BREAK;
    159 			} else
    160 				return count;
    161 		}
    162 		if (ep - bp < NLMSG_SPACE(0)) {
    163 			/*
    164 			 * There's less than one netlink message left
    165 			 * in the buffer.  Give up.
    166 			 */
    167 			break;
    168 		}
    169 
    170 		if (nlh->nlmsg_len < sizeof(struct nlmsghdr) || (u_int)len < nlh->nlmsg_len) {
    171 			pcap_snprintf(handle->errbuf, PCAP_ERRBUF_SIZE, "Message truncated: (got: %d) (nlmsg_len: %u)", len, nlh->nlmsg_len);
    172 			return -1;
    173 		}
    174 
    175 		if (NFNL_SUBSYS_ID(nlh->nlmsg_type) == NFNL_SUBSYS_ULOG &&
    176 		    NFNL_MSG_TYPE(nlh->nlmsg_type) == NFULNL_MSG_PACKET)
    177 			type = NFLOG;
    178 		else if (NFNL_SUBSYS_ID(nlh->nlmsg_type) == NFNL_SUBSYS_QUEUE &&
    179 		         NFNL_MSG_TYPE(nlh->nlmsg_type) == NFQNL_MSG_PACKET)
    180 			type = NFQUEUE;
    181 
    182 		if (type != OTHER) {
    183 			const unsigned char *payload = NULL;
    184 			struct pcap_pkthdr pkth;
    185 
    186 			const struct nfgenmsg *nfg = NULL;
    187 			int id = 0;
    188 
    189 			if (handle->linktype != DLT_NFLOG) {
    190 				const struct nfattr *payload_attr = NULL;
    191 
    192 				if (nlh->nlmsg_len < HDR_LENGTH) {
    193 					pcap_snprintf(handle->errbuf, PCAP_ERRBUF_SIZE, "Malformed message: (nlmsg_len: %u)", nlh->nlmsg_len);
    194 					return -1;
    195 				}
    196 
    197 				nfg = NLMSG_DATA(nlh);
    198 				if (nlh->nlmsg_len > HDR_LENGTH) {
    199 					struct nfattr *attr = NFM_NFA(nfg);
    200 					int attr_len = nlh->nlmsg_len - NLMSG_ALIGN(HDR_LENGTH);
    201 
    202 					while (NFA_OK(attr, attr_len)) {
    203 						if (type == NFQUEUE) {
    204 							switch (NFA_TYPE(attr)) {
    205 								case NFQA_PACKET_HDR:
    206 									{
    207 										const struct nfqnl_msg_packet_hdr *pkt_hdr = (const struct nfqnl_msg_packet_hdr *) NFA_DATA(attr);
    208 
    209 										id = ntohl(pkt_hdr->packet_id);
    210 										break;
    211 									}
    212 								case NFQA_PAYLOAD:
    213 									payload_attr = attr;
    214 									break;
    215 							}
    216 
    217 						} else if (type == NFLOG) {
    218 							switch (NFA_TYPE(attr)) {
    219 								case NFULA_PAYLOAD:
    220 									payload_attr = attr;
    221 									break;
    222 							}
    223 						}
    224 						attr = NFA_NEXT(attr, attr_len);
    225 					}
    226 				}
    227 
    228 				if (payload_attr) {
    229 					payload = NFA_DATA(payload_attr);
    230 					pkth.len = pkth.caplen = NFA_PAYLOAD(payload_attr);
    231 				}
    232 
    233 			} else {
    234 				payload = NLMSG_DATA(nlh);
    235 				pkth.caplen = pkth.len = nlh->nlmsg_len-NLMSG_ALIGN(sizeof(struct nlmsghdr));
    236 			}
    237 
    238 			if (payload) {
    239 				/* pkth.caplen = min (payload_len, handle->snapshot); */
    240 
    241 				gettimeofday(&pkth.ts, NULL);
    242 				if (handle->fcode.bf_insns == NULL ||
    243 						bpf_filter(handle->fcode.bf_insns, payload, pkth.len, pkth.caplen))
    244 				{
    245 					handlep->packets_read++;
    246 					callback(user, &pkth, payload);
    247 					count++;
    248 				}
    249 			}
    250 
    251 			if (type == NFQUEUE) {
    252 				/* XXX, possible responses: NF_DROP, NF_ACCEPT, NF_STOLEN, NF_QUEUE, NF_REPEAT, NF_STOP */
    253 				/* if type == NFQUEUE, handle->linktype is always != DLT_NFLOG,
    254 				   so nfg is always initialized to NLMSG_DATA(nlh). */
    255 				if (nfg != NULL)
    256 					nfqueue_send_verdict(handle, ntohs(nfg->res_id), id, NF_ACCEPT);
    257 			}
    258 		}
    259 
    260 		msg_len = NLMSG_ALIGN(nlh->nlmsg_len);
    261 		/*
    262 		 * If the message length would run past the end of the
    263 		 * buffer, truncate it to the remaining space in the
    264 		 * buffer.
    265 		 */
    266 		if (msg_len > ep - bp)
    267 			msg_len = ep - bp;
    268 
    269 		bp += msg_len;
    270 		if (count >= max_packets && !PACKET_COUNT_IS_UNLIMITED(max_packets)) {
    271 			handle->bp = bp;
    272 			handle->cc = ep - bp;
    273 			if (handle->cc < 0)
    274 				handle->cc = 0;
    275 			return count;
    276 		}
    277 	}
    278 
    279 	handle->cc = 0;
    280 	return count;
    281 }
    282 
    283 static int
    284 netfilter_set_datalink(pcap_t *handle, int dlt)
    285 {
    286 	handle->linktype = dlt;
    287 	return 0;
    288 }
    289 
    290 static int
    291 netfilter_stats_linux(pcap_t *handle, struct pcap_stat *stats)
    292 {
    293 	struct pcap_netfilter *handlep = handle->priv;
    294 
    295 	stats->ps_recv = handlep->packets_read;
    296 	stats->ps_drop = handlep->packets_nobufs;
    297 	stats->ps_ifdrop = 0;
    298 	return 0;
    299 }
    300 
    301 static int
    302 netfilter_inject_linux(pcap_t *handle, const void *buf _U_, size_t size _U_)
    303 {
    304 	pcap_snprintf(handle->errbuf, PCAP_ERRBUF_SIZE, "inject not supported on netfilter devices");
    305 	return (-1);
    306 }
    307 
    308 struct my_nfattr {
    309 	uint16_t nfa_len;
    310 	uint16_t nfa_type;
    311 	void *data;
    312 };
    313 
    314 static int
    315 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)
    316 {
    317 	char buf[1024] __attribute__ ((aligned));
    318 
    319 	struct nlmsghdr *nlh = (struct nlmsghdr *) buf;
    320 	struct nfgenmsg *nfg = (struct nfgenmsg *) (buf + sizeof(struct nlmsghdr));
    321 
    322 	struct sockaddr_nl snl;
    323 	static unsigned int seq_id;
    324 
    325 	if (!seq_id)
    326 		seq_id = time(NULL);
    327 	++seq_id;
    328 
    329 	nlh->nlmsg_len = NLMSG_LENGTH(sizeof(struct nfgenmsg));
    330 	nlh->nlmsg_type = msg_type;
    331 	nlh->nlmsg_flags = NLM_F_REQUEST | (ack ? NLM_F_ACK : 0);
    332 	nlh->nlmsg_pid = 0;	/* to kernel */
    333 	nlh->nlmsg_seq = seq_id;
    334 
    335 	nfg->nfgen_family = family;
    336 	nfg->version = NFNETLINK_V0;
    337 	nfg->res_id = htons(res_id);
    338 
    339 	if (mynfa) {
    340 		struct nfattr *nfa = (struct nfattr *) (buf + NLMSG_ALIGN(nlh->nlmsg_len));
    341 
    342 		nfa->nfa_type = mynfa->nfa_type;
    343 		nfa->nfa_len = NFA_LENGTH(mynfa->nfa_len);
    344 		memcpy(NFA_DATA(nfa), mynfa->data, mynfa->nfa_len);
    345 		nlh->nlmsg_len = NLMSG_ALIGN(nlh->nlmsg_len) + NFA_ALIGN(nfa->nfa_len);
    346 	}
    347 
    348 	memset(&snl, 0, sizeof(snl));
    349 	snl.nl_family = AF_NETLINK;
    350 
    351 	if (sendto(handle->fd, nlh, nlh->nlmsg_len, 0, (struct sockaddr *) &snl, sizeof(snl)) == -1)
    352 		return -1;
    353 
    354 	if (!ack)
    355 		return 0;
    356 
    357 	/* waiting for reply loop */
    358 	do {
    359 		socklen_t addrlen = sizeof(snl);
    360 		int len;
    361 
    362 		/* ignore interrupt system call error */
    363 		do {
    364 			len = recvfrom(handle->fd, buf, sizeof(buf), 0, (struct sockaddr *) &snl, &addrlen);
    365 		} while ((len == -1) && (errno == EINTR));
    366 
    367 		if (len <= 0)
    368 			return len;
    369 
    370 		if (addrlen != sizeof(snl) || snl.nl_family != AF_NETLINK) {
    371 			errno = EINVAL;
    372 			return -1;
    373 		}
    374 
    375 		nlh = (struct nlmsghdr *) buf;
    376 		if (snl.nl_pid != 0 || seq_id != nlh->nlmsg_seq)	/* if not from kernel or wrong sequence skip */
    377 			continue;
    378 
    379 		while ((u_int)len >= NLMSG_SPACE(0) && NLMSG_OK(nlh, (u_int)len)) {
    380 			if (nlh->nlmsg_type == NLMSG_ERROR || (nlh->nlmsg_type == NLMSG_DONE && nlh->nlmsg_flags & NLM_F_MULTI)) {
    381 				if (nlh->nlmsg_len < NLMSG_ALIGN(sizeof(struct nlmsgerr))) {
    382 					errno = EBADMSG;
    383 					return -1;
    384 				}
    385 				errno = -(*((int *)NLMSG_DATA(nlh)));
    386 				return (errno == 0) ? 0 : -1;
    387 			}
    388 			nlh = NLMSG_NEXT(nlh, len);
    389 		}
    390 	} while (1);
    391 
    392 	return -1; /* never here */
    393 }
    394 
    395 static int
    396 nflog_send_config_msg(const pcap_t *handle, uint8_t family, u_int16_t group_id, const struct my_nfattr *mynfa)
    397 {
    398 	return netfilter_send_config_msg(handle, (NFNL_SUBSYS_ULOG << 8) | NFULNL_MSG_CONFIG, 1, family, group_id, mynfa);
    399 }
    400 
    401 static int
    402 nflog_send_config_cmd(const pcap_t *handle, uint16_t group_id, u_int8_t cmd, u_int8_t family)
    403 {
    404 	struct nfulnl_msg_config_cmd msg;
    405 	struct my_nfattr nfa;
    406 
    407 	msg.command = cmd;
    408 
    409 	nfa.data = &msg;
    410 	nfa.nfa_type = NFULA_CFG_CMD;
    411 	nfa.nfa_len = sizeof(msg);
    412 
    413 	return nflog_send_config_msg(handle, family, group_id, &nfa);
    414 }
    415 
    416 static int
    417 nflog_send_config_mode(const pcap_t *handle, uint16_t group_id, u_int8_t copy_mode, u_int32_t copy_range)
    418 {
    419 	struct nfulnl_msg_config_mode msg;
    420 	struct my_nfattr nfa;
    421 
    422 	msg.copy_range = htonl(copy_range);
    423 	msg.copy_mode = copy_mode;
    424 
    425 	nfa.data = &msg;
    426 	nfa.nfa_type = NFULA_CFG_MODE;
    427 	nfa.nfa_len = sizeof(msg);
    428 
    429 	return nflog_send_config_msg(handle, AF_UNSPEC, group_id, &nfa);
    430 }
    431 
    432 static int
    433 nfqueue_send_verdict(const pcap_t *handle, uint16_t group_id, u_int32_t id, u_int32_t verdict)
    434 {
    435 	struct nfqnl_msg_verdict_hdr msg;
    436 	struct my_nfattr nfa;
    437 
    438 	msg.id = htonl(id);
    439 	msg.verdict = htonl(verdict);
    440 
    441 	nfa.data = &msg;
    442 	nfa.nfa_type = NFQA_VERDICT_HDR;
    443 	nfa.nfa_len = sizeof(msg);
    444 
    445 	return netfilter_send_config_msg(handle, (NFNL_SUBSYS_QUEUE << 8) | NFQNL_MSG_VERDICT, 0, AF_UNSPEC, group_id, &nfa);
    446 }
    447 
    448 static int
    449 nfqueue_send_config_msg(const pcap_t *handle, uint8_t family, u_int16_t group_id, const struct my_nfattr *mynfa)
    450 {
    451 	return netfilter_send_config_msg(handle, (NFNL_SUBSYS_QUEUE << 8) | NFQNL_MSG_CONFIG, 1, family, group_id, mynfa);
    452 }
    453 
    454 static int
    455 nfqueue_send_config_cmd(const pcap_t *handle, uint16_t group_id, u_int8_t cmd, u_int16_t pf)
    456 {
    457 	struct nfqnl_msg_config_cmd msg;
    458 	struct my_nfattr nfa;
    459 
    460 	msg.command = cmd;
    461 	msg.pf = htons(pf);
    462 
    463 	nfa.data = &msg;
    464 	nfa.nfa_type = NFQA_CFG_CMD;
    465 	nfa.nfa_len = sizeof(msg);
    466 
    467 	return nfqueue_send_config_msg(handle, AF_UNSPEC, group_id, &nfa);
    468 }
    469 
    470 static int
    471 nfqueue_send_config_mode(const pcap_t *handle, uint16_t group_id, u_int8_t copy_mode, u_int32_t copy_range)
    472 {
    473 	struct nfqnl_msg_config_params msg;
    474 	struct my_nfattr nfa;
    475 
    476 	msg.copy_range = htonl(copy_range);
    477 	msg.copy_mode = copy_mode;
    478 
    479 	nfa.data = &msg;
    480 	nfa.nfa_type = NFQA_CFG_PARAMS;
    481 	nfa.nfa_len = sizeof(msg);
    482 
    483 	return nfqueue_send_config_msg(handle, AF_UNSPEC, group_id, &nfa);
    484 }
    485 
    486 static int
    487 netfilter_activate(pcap_t* handle)
    488 {
    489 	const char *dev = handle->opt.device;
    490 	unsigned short groups[32];
    491 	int group_count = 0;
    492 	nftype_t type = OTHER;
    493 	int i;
    494 
    495  	if (strncmp(dev, NFLOG_IFACE, strlen(NFLOG_IFACE)) == 0) {
    496  		dev += strlen(NFLOG_IFACE);
    497 		type = NFLOG;
    498 
    499 	} else if (strncmp(dev, NFQUEUE_IFACE, strlen(NFQUEUE_IFACE)) == 0) {
    500 		dev += strlen(NFQUEUE_IFACE);
    501 		type = NFQUEUE;
    502 	}
    503 
    504 	if (type != OTHER && *dev == ':') {
    505 		dev++;
    506 		while (*dev) {
    507 			long int group_id;
    508 			char *end_dev;
    509 
    510 			if (group_count == 32) {
    511 				pcap_snprintf(handle->errbuf, PCAP_ERRBUF_SIZE,
    512 						"Maximum 32 netfilter groups! dev: %s",
    513 						handle->opt.device);
    514 				return PCAP_ERROR;
    515 			}
    516 
    517 			group_id = strtol(dev, &end_dev, 0);
    518 			if (end_dev != dev) {
    519 				if (group_id < 0 || group_id > 65535) {
    520 					pcap_snprintf(handle->errbuf, PCAP_ERRBUF_SIZE,
    521 							"Netfilter group range from 0 to 65535 (got %ld)",
    522 							group_id);
    523 					return PCAP_ERROR;
    524 				}
    525 
    526 				groups[group_count++] = (unsigned short) group_id;
    527 				dev = end_dev;
    528 			}
    529 			if (*dev != ',')
    530 				break;
    531 			dev++;
    532 		}
    533 	}
    534 
    535 	if (type == OTHER || *dev) {
    536 		pcap_snprintf(handle->errbuf, PCAP_ERRBUF_SIZE,
    537 				"Can't get netfilter group(s) index from %s",
    538 				handle->opt.device);
    539 		return PCAP_ERROR;
    540 	}
    541 
    542 	/* if no groups, add default: 0 */
    543 	if (!group_count) {
    544 		groups[0] = 0;
    545 		group_count = 1;
    546 	}
    547 
    548 	/*
    549 	 * Turn a negative snapshot value (invalid), a snapshot value of
    550 	 * 0 (unspecified), or a value bigger than the normal maximum
    551 	 * value, into the maximum allowed value.
    552 	 *
    553 	 * If some application really *needs* a bigger snapshot
    554 	 * length, we should just increase MAXIMUM_SNAPLEN.
    555 	 */
    556 	if (handle->snapshot <= 0 || handle->snapshot > MAXIMUM_SNAPLEN)
    557 		handle->snapshot = MAXIMUM_SNAPLEN;
    558 
    559 	/* Initialize some components of the pcap structure. */
    560 	handle->bufsize = 128 + handle->snapshot;
    561 	handle->offset = 0;
    562 	handle->read_op = netfilter_read_linux;
    563 	handle->inject_op = netfilter_inject_linux;
    564 	handle->setfilter_op = install_bpf_program; /* no kernel filtering */
    565 	handle->setdirection_op = NULL;
    566 	handle->set_datalink_op = netfilter_set_datalink;
    567 	handle->getnonblock_op = pcap_getnonblock_fd;
    568 	handle->setnonblock_op = pcap_setnonblock_fd;
    569 	handle->stats_op = netfilter_stats_linux;
    570 
    571 	/* Create netlink socket */
    572 	handle->fd = socket(AF_NETLINK, SOCK_RAW, NETLINK_NETFILTER);
    573 	if (handle->fd < 0) {
    574 		pcap_fmt_errmsg_for_errno(handle->errbuf, PCAP_ERRBUF_SIZE,
    575 		    errno, "Can't create raw socket");
    576 		return PCAP_ERROR;
    577 	}
    578 
    579 	if (type == NFLOG) {
    580 		handle->linktype = DLT_NFLOG;
    581 		handle->dlt_list = (u_int *) malloc(sizeof(u_int) * 2);
    582 		if (handle->dlt_list != NULL) {
    583 			handle->dlt_list[0] = DLT_NFLOG;
    584 			handle->dlt_list[1] = DLT_IPV4;
    585 			handle->dlt_count = 2;
    586 		}
    587 
    588 	} else
    589 		handle->linktype = DLT_IPV4;
    590 
    591 	handle->buffer = malloc(handle->bufsize);
    592 	if (!handle->buffer) {
    593 		pcap_fmt_errmsg_for_errno(handle->errbuf, PCAP_ERRBUF_SIZE,
    594 		    errno, "Can't allocate dump buffer");
    595 		goto close_fail;
    596 	}
    597 
    598 	if (type == NFLOG) {
    599 		if (nflog_send_config_cmd(handle, 0, NFULNL_CFG_CMD_PF_UNBIND, AF_INET) < 0) {
    600 			pcap_fmt_errmsg_for_errno(handle->errbuf,
    601 			    PCAP_ERRBUF_SIZE, errno,
    602 			    "NFULNL_CFG_CMD_PF_UNBIND");
    603 			goto close_fail;
    604 		}
    605 
    606 		if (nflog_send_config_cmd(handle, 0, NFULNL_CFG_CMD_PF_BIND, AF_INET) < 0) {
    607 			pcap_fmt_errmsg_for_errno(handle->errbuf,
    608 			    PCAP_ERRBUF_SIZE, errno, "NFULNL_CFG_CMD_PF_BIND");
    609 			goto close_fail;
    610 		}
    611 
    612 		/* Bind socket to the nflog groups */
    613 		for (i = 0; i < group_count; i++) {
    614 			if (nflog_send_config_cmd(handle, groups[i], NFULNL_CFG_CMD_BIND, AF_UNSPEC) < 0) {
    615 				pcap_fmt_errmsg_for_errno(handle->errbuf,
    616 				    PCAP_ERRBUF_SIZE, errno,
    617 				    "Can't listen on group group index");
    618 				goto close_fail;
    619 			}
    620 
    621 			if (nflog_send_config_mode(handle, groups[i], NFULNL_COPY_PACKET, handle->snapshot) < 0) {
    622 				pcap_fmt_errmsg_for_errno(handle->errbuf,
    623 				    PCAP_ERRBUF_SIZE, errno,
    624 				    "NFULNL_COPY_PACKET");
    625 				goto close_fail;
    626 			}
    627 		}
    628 
    629 	} else {
    630 		if (nfqueue_send_config_cmd(handle, 0, NFQNL_CFG_CMD_PF_UNBIND, AF_INET) < 0) {
    631 			pcap_fmt_errmsg_for_errno(handle->errbuf,
    632 			    PCAP_ERRBUF_SIZE, errno, "NFQNL_CFG_CMD_PF_UNBIND");
    633 			goto close_fail;
    634 		}
    635 
    636 		if (nfqueue_send_config_cmd(handle, 0, NFQNL_CFG_CMD_PF_BIND, AF_INET) < 0) {
    637 			pcap_fmt_errmsg_for_errno(handle->errbuf,
    638 			    PCAP_ERRBUF_SIZE, errno, "NFQNL_CFG_CMD_PF_BIND");
    639 			goto close_fail;
    640 		}
    641 
    642 		/* Bind socket to the nfqueue groups */
    643 		for (i = 0; i < group_count; i++) {
    644 			if (nfqueue_send_config_cmd(handle, groups[i], NFQNL_CFG_CMD_BIND, AF_UNSPEC) < 0) {
    645 				pcap_fmt_errmsg_for_errno(handle->errbuf,
    646 				    PCAP_ERRBUF_SIZE, errno,
    647 				    "Can't listen on group group index");
    648 				goto close_fail;
    649 			}
    650 
    651 			if (nfqueue_send_config_mode(handle, groups[i], NFQNL_COPY_PACKET, handle->snapshot) < 0) {
    652 				pcap_fmt_errmsg_for_errno(handle->errbuf,
    653 				    PCAP_ERRBUF_SIZE, errno,
    654 				    "NFQNL_COPY_PACKET");
    655 				goto close_fail;
    656 			}
    657 		}
    658 	}
    659 
    660 	if (handle->opt.rfmon) {
    661 		/*
    662 		 * Monitor mode doesn't apply to netfilter devices.
    663 		 */
    664 		pcap_cleanup_live_common(handle);
    665 		return PCAP_ERROR_RFMON_NOTSUP;
    666 	}
    667 
    668 	if (handle->opt.buffer_size != 0) {
    669 		/*
    670 		 * Set the socket buffer size to the specified value.
    671 		 */
    672 		if (setsockopt(handle->fd, SOL_SOCKET, SO_RCVBUF, &handle->opt.buffer_size, sizeof(handle->opt.buffer_size)) == -1) {
    673 			pcap_fmt_errmsg_for_errno(handle->errbuf,
    674 			    PCAP_ERRBUF_SIZE, errno, "SO_RCVBUF");
    675 			goto close_fail;
    676 		}
    677 	}
    678 
    679 	handle->selectable_fd = handle->fd;
    680 	return 0;
    681 
    682 close_fail:
    683 	pcap_cleanup_live_common(handle);
    684 	return PCAP_ERROR;
    685 }
    686 
    687 pcap_t *
    688 netfilter_create(const char *device, char *ebuf, int *is_ours)
    689 {
    690 	const char *cp;
    691 	pcap_t *p;
    692 
    693 	/* Does this look like an netfilter device? */
    694 	cp = strrchr(device, '/');
    695 	if (cp == NULL)
    696 		cp = device;
    697 
    698 	/* Does it begin with NFLOG_IFACE or NFQUEUE_IFACE? */
    699 	if (strncmp(cp, NFLOG_IFACE, sizeof NFLOG_IFACE - 1) == 0)
    700 		cp += sizeof NFLOG_IFACE - 1;
    701 	else if (strncmp(cp, NFQUEUE_IFACE, sizeof NFQUEUE_IFACE - 1) == 0)
    702 		cp += sizeof NFQUEUE_IFACE - 1;
    703 	else {
    704 		/* Nope, doesn't begin with NFLOG_IFACE nor NFQUEUE_IFACE */
    705 		*is_ours = 0;
    706 		return NULL;
    707 	}
    708 
    709 	/*
    710 	 * Yes - is that either the end of the name, or is it followed
    711 	 * by a colon?
    712 	 */
    713 	if (*cp != ':' && *cp != '\0') {
    714 		/* Nope */
    715 		*is_ours = 0;
    716 		return NULL;
    717 	}
    718 
    719 	/* OK, it's probably ours. */
    720 	*is_ours = 1;
    721 
    722 	p = pcap_create_common(ebuf, sizeof (struct pcap_netfilter));
    723 	if (p == NULL)
    724 		return (NULL);
    725 
    726 	p->activate_op = netfilter_activate;
    727 	return (p);
    728 }
    729 
    730 int
    731 netfilter_findalldevs(pcap_if_list_t *devlistp, char *err_str)
    732 {
    733 	int sock;
    734 
    735 	sock = socket(AF_NETLINK, SOCK_RAW, NETLINK_NETFILTER);
    736 	if (sock < 0) {
    737 		/* if netlink is not supported this is not fatal */
    738 		if (errno == EAFNOSUPPORT || errno == EPROTONOSUPPORT)
    739 			return 0;
    740 		pcap_fmt_errmsg_for_errno(err_str, PCAP_ERRBUF_SIZE,
    741 		    errno, "Can't open netlink socket");
    742 		return -1;
    743 	}
    744 	close(sock);
    745 
    746 	/*
    747 	 * The notion of "connected" vs. "disconnected" doesn't apply.
    748 	 * XXX - what about "up" and "running"?
    749 	 */
    750 	if (add_dev(devlistp, NFLOG_IFACE,
    751 	    PCAP_IF_CONNECTION_STATUS_NOT_APPLICABLE,
    752 	    "Linux netfilter log (NFLOG) interface", err_str) == NULL)
    753 		return -1;
    754 	if (add_dev(devlistp, NFQUEUE_IFACE,
    755 	    PCAP_IF_CONNECTION_STATUS_NOT_APPLICABLE,
    756 	    "Linux netfilter queue (NFQUEUE) interface", err_str) == NULL)
    757 		return -1;
    758 	return 0;
    759 }
    760