2 * Copyright (c) 2011 Jakub Zawadzki
5 * Redistribution and use in source and binary forms, with or without
6 * modification, are permitted provided that the following conditions
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
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.
37 #ifdef NEED_STRERROR_H
45 #include <sys/socket.h>
46 #include <arpa/inet.h>
50 #include <netinet/in.h>
51 #include <linux/types.h>
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>
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 ;/
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.
65 * In tcpdump you can disable dropping privilages with -Z root
68 #include "pcap-netfilter-linux.h"
70 #define HDR_LENGTH (NLMSG_LENGTH(NLMSG_ALIGN(sizeof(struct nfgenmsg))))
72 #define NFLOG_IFACE "nflog"
73 #define NFQUEUE_IFACE "nfqueue"
75 typedef enum { OTHER = -1, NFLOG, NFQUEUE } nftype_t;
78 * Private data for capturing on Linux netfilter sockets.
80 struct pcap_netfilter {
81 u_int packets_read; /* count of packets read with recvfrom() */
82 u_int packets_nobufs; /* ENOBUFS counter */
85 static int nfqueue_send_verdict(const pcap_t *handle, uint16_t group_id, u_int32_t id, u_int32_t verdict);
89 netfilter_read_linux(pcap_t *handle, int max_packets, pcap_handler callback, u_char *user)
91 struct pcap_netfilter *handlep = handle->priv;
92 register u_char *bp, *ep;
97 * Has "pcap_breakloop()" been called?
99 if (handle->break_loop) {
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.
105 handle->break_loop = 0;
106 return PCAP_ERROR_BREAK;
111 * The buffer is empty; refill it.
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).
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;
125 if (errno == ENOBUFS)
126 handlep->packets_nobufs++;
127 } while ((len == -1) && (errno == EINTR || errno == ENOBUFS));
130 pcap_fmt_errmsg_for_errno(handle->errbuf,
131 PCAP_ERRBUF_SIZE, errno, "Can't receive packet");
135 bp = (unsigned char *)handle->buffer;
140 const struct nlmsghdr *nlh = (const struct nlmsghdr *) bp;
142 nftype_t type = OTHER;
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
153 if (handle->break_loop) {
155 handle->cc = ep - bp;
157 handle->break_loop = 0;
158 return PCAP_ERROR_BREAK;
162 if (ep - bp < NLMSG_SPACE(0)) {
164 * There's less than one netlink message left
165 * in the buffer. Give up.
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);
175 if (NFNL_SUBSYS_ID(nlh->nlmsg_type) == NFNL_SUBSYS_ULOG &&
176 NFNL_MSG_TYPE(nlh->nlmsg_type) == NFULNL_MSG_PACKET)
178 else if (NFNL_SUBSYS_ID(nlh->nlmsg_type) == NFNL_SUBSYS_QUEUE &&
179 NFNL_MSG_TYPE(nlh->nlmsg_type) == NFQNL_MSG_PACKET)
183 const unsigned char *payload = NULL;
184 struct pcap_pkthdr pkth;
186 const struct nfgenmsg *nfg = NULL;
189 if (handle->linktype != DLT_NFLOG) {
190 const struct nfattr *payload_attr = NULL;
192 if (nlh->nlmsg_len < HDR_LENGTH) {
193 pcap_snprintf(handle->errbuf, PCAP_ERRBUF_SIZE, "Malformed message: (nlmsg_len: %u)", nlh->nlmsg_len);
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);
202 while (NFA_OK(attr, attr_len)) {
203 if (type == NFQUEUE) {
204 switch (NFA_TYPE(attr)) {
205 case NFQA_PACKET_HDR:
207 const struct nfqnl_msg_packet_hdr *pkt_hdr = (const struct nfqnl_msg_packet_hdr *) NFA_DATA(attr);
209 id = ntohl(pkt_hdr->packet_id);
217 } else if (type == NFLOG) {
218 switch (NFA_TYPE(attr)) {
224 attr = NFA_NEXT(attr, attr_len);
229 payload = NFA_DATA(payload_attr);
230 pkth.len = pkth.caplen = NFA_PAYLOAD(payload_attr);
234 payload = NLMSG_DATA(nlh);
235 pkth.caplen = pkth.len = nlh->nlmsg_len-NLMSG_ALIGN(sizeof(struct nlmsghdr));
239 /* pkth.caplen = min (payload_len, handle->snapshot); */
241 gettimeofday(&pkth.ts, NULL);
242 if (handle->fcode.bf_insns == NULL ||
243 bpf_filter(handle->fcode.bf_insns, payload, pkth.len, pkth.caplen))
245 handlep->packets_read++;
246 callback(user, &pkth, payload);
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). */
256 nfqueue_send_verdict(handle, ntohs(nfg->res_id), id, NF_ACCEPT);
260 msg_len = NLMSG_ALIGN(nlh->nlmsg_len);
262 * If the message length would run past the end of the
263 * buffer, truncate it to the remaining space in the
266 if (msg_len > ep - bp)
270 if (count >= max_packets && !PACKET_COUNT_IS_UNLIMITED(max_packets)) {
272 handle->cc = ep - bp;
284 netfilter_set_datalink(pcap_t *handle, int dlt)
286 handle->linktype = dlt;
291 netfilter_stats_linux(pcap_t *handle, struct pcap_stat *stats)
293 struct pcap_netfilter *handlep = handle->priv;
295 stats->ps_recv = handlep->packets_read;
296 stats->ps_drop = handlep->packets_nobufs;
297 stats->ps_ifdrop = 0;
302 netfilter_inject_linux(pcap_t *handle, const void *buf _U_, size_t size _U_)
304 pcap_snprintf(handle->errbuf, PCAP_ERRBUF_SIZE, "inject not supported on netfilter devices");
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)
317 char buf[1024] __attribute__ ((aligned));
319 struct nlmsghdr *nlh = (struct nlmsghdr *) buf;
320 struct nfgenmsg *nfg = (struct nfgenmsg *) (buf + sizeof(struct nlmsghdr));
322 struct sockaddr_nl snl;
323 static unsigned int seq_id;
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;
335 nfg->nfgen_family = family;
336 nfg->version = NFNETLINK_V0;
337 nfg->res_id = htons(res_id);
340 struct nfattr *nfa = (struct nfattr *) (buf + NLMSG_ALIGN(nlh->nlmsg_len));
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);
348 memset(&snl, 0, sizeof(snl));
349 snl.nl_family = AF_NETLINK;
351 if (sendto(handle->fd, nlh, nlh->nlmsg_len, 0, (struct sockaddr *) &snl, sizeof(snl)) == -1)
357 /* waiting for reply loop */
359 socklen_t addrlen = sizeof(snl);
362 /* ignore interrupt system call error */
364 len = recvfrom(handle->fd, buf, sizeof(buf), 0, (struct sockaddr *) &snl, &addrlen);
365 } while ((len == -1) && (errno == EINTR));
370 if (addrlen != sizeof(snl) || snl.nl_family != AF_NETLINK) {
375 nlh = (struct nlmsghdr *) buf;
376 if (snl.nl_pid != 0 || seq_id != nlh->nlmsg_seq) /* if not from kernel or wrong sequence skip */
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))) {
385 errno = -(*((int *)NLMSG_DATA(nlh)));
386 return (errno == 0) ? 0 : -1;
388 nlh = NLMSG_NEXT(nlh, len);
392 return -1; /* never here */
396 nflog_send_config_msg(const pcap_t *handle, uint8_t family, u_int16_t group_id, const struct my_nfattr *mynfa)
398 return netfilter_send_config_msg(handle, (NFNL_SUBSYS_ULOG << 8) | NFULNL_MSG_CONFIG, 1, family, group_id, mynfa);
402 nflog_send_config_cmd(const pcap_t *handle, uint16_t group_id, u_int8_t cmd, u_int8_t family)
404 struct nfulnl_msg_config_cmd msg;
405 struct my_nfattr nfa;
410 nfa.nfa_type = NFULA_CFG_CMD;
411 nfa.nfa_len = sizeof(msg);
413 return nflog_send_config_msg(handle, family, group_id, &nfa);
417 nflog_send_config_mode(const pcap_t *handle, uint16_t group_id, u_int8_t copy_mode, u_int32_t copy_range)
419 struct nfulnl_msg_config_mode msg;
420 struct my_nfattr nfa;
422 msg.copy_range = htonl(copy_range);
423 msg.copy_mode = copy_mode;
426 nfa.nfa_type = NFULA_CFG_MODE;
427 nfa.nfa_len = sizeof(msg);
429 return nflog_send_config_msg(handle, AF_UNSPEC, group_id, &nfa);
433 nfqueue_send_verdict(const pcap_t *handle, uint16_t group_id, u_int32_t id, u_int32_t verdict)
435 struct nfqnl_msg_verdict_hdr msg;
436 struct my_nfattr nfa;
439 msg.verdict = htonl(verdict);
442 nfa.nfa_type = NFQA_VERDICT_HDR;
443 nfa.nfa_len = sizeof(msg);
445 return netfilter_send_config_msg(handle, (NFNL_SUBSYS_QUEUE << 8) | NFQNL_MSG_VERDICT, 0, AF_UNSPEC, group_id, &nfa);
449 nfqueue_send_config_msg(const pcap_t *handle, uint8_t family, u_int16_t group_id, const struct my_nfattr *mynfa)
451 return netfilter_send_config_msg(handle, (NFNL_SUBSYS_QUEUE << 8) | NFQNL_MSG_CONFIG, 1, family, group_id, mynfa);
455 nfqueue_send_config_cmd(const pcap_t *handle, uint16_t group_id, u_int8_t cmd, u_int16_t pf)
457 struct nfqnl_msg_config_cmd msg;
458 struct my_nfattr nfa;
464 nfa.nfa_type = NFQA_CFG_CMD;
465 nfa.nfa_len = sizeof(msg);
467 return nfqueue_send_config_msg(handle, AF_UNSPEC, group_id, &nfa);
471 nfqueue_send_config_mode(const pcap_t *handle, uint16_t group_id, u_int8_t copy_mode, u_int32_t copy_range)
473 struct nfqnl_msg_config_params msg;
474 struct my_nfattr nfa;
476 msg.copy_range = htonl(copy_range);
477 msg.copy_mode = copy_mode;
480 nfa.nfa_type = NFQA_CFG_PARAMS;
481 nfa.nfa_len = sizeof(msg);
483 return nfqueue_send_config_msg(handle, AF_UNSPEC, group_id, &nfa);
487 netfilter_activate(pcap_t* handle)
489 const char *dev = handle->opt.device;
490 unsigned short groups[32];
492 nftype_t type = OTHER;
495 if (strncmp(dev, NFLOG_IFACE, strlen(NFLOG_IFACE)) == 0) {
496 dev += strlen(NFLOG_IFACE);
499 } else if (strncmp(dev, NFQUEUE_IFACE, strlen(NFQUEUE_IFACE)) == 0) {
500 dev += strlen(NFQUEUE_IFACE);
504 if (type != OTHER && *dev == ':') {
510 if (group_count == 32) {
511 pcap_snprintf(handle->errbuf, PCAP_ERRBUF_SIZE,
512 "Maximum 32 netfilter groups! dev: %s",
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)",
526 groups[group_count++] = (unsigned short) group_id;
535 if (type == OTHER || *dev) {
536 pcap_snprintf(handle->errbuf, PCAP_ERRBUF_SIZE,
537 "Can't get netfilter group(s) index from %s",
542 /* if no groups, add default: 0 */
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.
553 * If some application really *needs* a bigger snapshot
554 * length, we should just increase MAXIMUM_SNAPLEN.
556 if (handle->snapshot <= 0 || handle->snapshot > MAXIMUM_SNAPLEN)
557 handle->snapshot = MAXIMUM_SNAPLEN;
559 /* Initialize some components of the pcap structure. */
560 handle->bufsize = 128 + handle->snapshot;
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;
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");
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;
589 handle->linktype = DLT_IPV4;
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");
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");
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");
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");
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");
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");
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");
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");
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");
660 if (handle->opt.rfmon) {
662 * Monitor mode doesn't apply to netfilter devices.
664 pcap_cleanup_live_common(handle);
665 return PCAP_ERROR_RFMON_NOTSUP;
668 if (handle->opt.buffer_size != 0) {
670 * Set the socket buffer size to the specified value.
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");
679 handle->selectable_fd = handle->fd;
683 pcap_cleanup_live_common(handle);
688 netfilter_create(const char *device, char *ebuf, int *is_ours)
693 /* Does this look like an netfilter device? */
694 cp = strrchr(device, '/');
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;
704 /* Nope, doesn't begin with NFLOG_IFACE nor NFQUEUE_IFACE */
710 * Yes - is that either the end of the name, or is it followed
713 if (*cp != ':' && *cp != '\0') {
719 /* OK, it's probably ours. */
722 p = pcap_create_common(ebuf, sizeof (struct pcap_netfilter));
726 p->activate_op = netfilter_activate;
731 netfilter_findalldevs(pcap_if_list_t *devlistp, char *err_str)
735 sock = socket(AF_NETLINK, SOCK_RAW, NETLINK_NETFILTER);
737 /* if netlink is not supported this is not fatal */
738 if (errno == EAFNOSUPPORT || errno == EPROTONOSUPPORT)
740 pcap_fmt_errmsg_for_errno(err_str, PCAP_ERRBUF_SIZE,
741 errno, "Can't open netlink socket");
747 * The notion of "connected" vs. "disconnected" doesn't apply.
748 * XXX - what about "up" and "running"?
750 if (add_dev(devlistp, NFLOG_IFACE,
751 PCAP_IF_CONNECTION_STATUS_NOT_APPLICABLE,
752 "Linux netfilter log (NFLOG) interface", err_str) == NULL)
754 if (add_dev(devlistp, NFQUEUE_IFACE,
755 PCAP_IF_CONNECTION_STATUS_NOT_APPLICABLE,
756 "Linux netfilter queue (NFQUEUE) interface", err_str) == NULL)