The Tcpdump Group git mirrors - libpcap/blob - pcap-netfilter-linux.c

   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 #include <config.h>
  32
  33 #include "pcap-int.h"
  34 #include "diag-control.h"
  35
  36 #include <errno.h>
  37 #include <stdlib.h>
  38 #include <unistd.h>
  39 #include <string.h>
  40 #include <sys/socket.h>
  41 #include <arpa/inet.h>
  42
  43 #include <time.h>
  44 #include <sys/time.h>
  45 #include <netinet/in.h>
  46 #include <linux/types.h>
  47
  48 #include <linux/netlink.h>
  49 #include <linux/netfilter.h>
  50 #include <linux/netfilter/nfnetlink.h>
  51 #include <linux/netfilter/nfnetlink_log.h>
  52 #include <linux/netfilter/nfnetlink_queue.h>
  53
  54 /* NOTE: if your program drops privileges after pcap_activate() it WON'T work with nfqueue.
  55  *       It took me quite some time to debug ;/
  56  *
  57  *       Sending any data to nfnetlink socket requires CAP_NET_ADMIN privileges,
  58  *       and in nfqueue we need to send verdict reply after receiving packet.
  59  *
  60  *       In tcpdump you can disable dropping privileges with -Z root
  61  */
  62
  63 #include "pcap-netfilter-linux.h"
  64
  65 #define HDR_LENGTH (NLMSG_LENGTH(NLMSG_ALIGN(sizeof(struct nfgenmsg))))
  66
  67 #define NFLOG_IFACE "nflog"
  68 #define NFQUEUE_IFACE "nfqueue"
  69
  70 typedef enum { OTHER = -1, NFLOG, NFQUEUE } nftype_t;
  71
  72 /*
  73  * Private data for capturing on Linux netfilter sockets.
  74  */
  75 struct pcap_netfilter {
  76         u_int   packets_read;   /* count of packets read with recvfrom() */
  77         u_int   packets_nobufs; /* ENOBUFS counter */
  78 };
  79
  80 static int nfqueue_send_verdict(const pcap_t *handle, uint16_t group_id, u_int32_t id, u_int32_t verdict);
  81
  82
  83 static int
  84 netfilter_read_linux(pcap_t *handle, int max_packets, pcap_handler callback, u_char *user)
  85 {
  86         struct pcap_netfilter *handlep = handle->priv;
  87         register u_char *bp, *ep;
  88         int count = 0;
  89         u_int cc;
  90
  91         /*
  92          * Has "pcap_breakloop()" been called?
  93          */
  94         if (handle->break_loop) {
  95                 /*
  96                  * Yes - clear the flag that indicates that it
  97                  * has, and return PCAP_ERROR_BREAK to indicate
  98                  * that we were told to break out of the loop.
  99                  */
 100                 handle->break_loop = 0;
 101                 return PCAP_ERROR_BREAK;
 102         }
 103         cc = handle->cc;
 104         if (cc == 0) {
 105                 /*
 106                  * The buffer is empty; refill it.
 107                  *
 108                  * We ignore EINTR, as that might just be due to a signal
 109                  * being delivered - if the signal should interrupt the
 110                  * loop, the signal handler should call pcap_breakloop()
 111                  * to set handle->break_loop (we ignore it on other
 112                  * platforms as well).
 113                  */
 114                 ssize_t read_ret;
 115
 116                 do {
 117                         read_ret = recv(handle->fd, handle->buffer, handle->bufsize, 0);
 118                         if (handle->break_loop) {
 119                                 handle->break_loop = 0;
 120                                 return PCAP_ERROR_BREAK;
 121                         }
 122                         if (read_ret == -1 && errno == ENOBUFS)
 123                                 handlep->packets_nobufs++;
 124                 } while ((read_ret == -1) && (errno == EINTR || errno == ENOBUFS));
 125
 126                 if (read_ret < 0) {
 127                         pcapint_fmt_errmsg_for_errno(handle->errbuf,
 128                             PCAP_ERRBUF_SIZE, errno, "Can't receive packet");
 129                         return PCAP_ERROR;
 130                 }
 131
 132                 /*
 133                  * At this point, read_ret is guaranteed to be
 134                  * >= 0 and < p->bufsize; p->bufsize is a u_int,
 135                  * so its value is guaranteed to fit in cc, which
 136                  * is also a u_int.
 137                  */
 138                 cc = (u_int)read_ret;
 139                 bp = (unsigned char *)handle->buffer;
 140         } else
 141                 bp = handle->bp;
 142
 143         /*
 144          * Loop through each message.
 145          *
 146          * This assumes that a single buffer of message will have
 147          * <= INT_MAX packets, so the message count doesn't overflow.
 148          */
 149         ep = bp + cc;
 150         while (bp < ep) {
 151                 const struct nlmsghdr *nlh = (const struct nlmsghdr *) bp;
 152                 uint32_t msg_len;
 153                 nftype_t type = OTHER;
 154                 /*
 155                  * Has "pcap_breakloop()" been called?
 156                  * If so, return immediately - if we haven't read any
 157                  * packets, clear the flag and return PCAP_ERROR_BREAK
 158                  * to indicate that we were told to break out of the loop,
 159                  * otherwise leave the flag set, so that the *next* call
 160                  * will break out of the loop without having read any
 161                  * packets, and return the number of packets we've
 162                  * processed so far.
 163                  */
 164                 if (handle->break_loop) {
 165                         handle->bp = bp;
 166                         handle->cc = (u_int)(ep - bp);
 167                         if (count == 0) {
 168                                 handle->break_loop = 0;
 169                                 return PCAP_ERROR_BREAK;
 170                         } else
 171                                 return count;
 172                 }
 173                 /*
 174                  * NLMSG_SPACE(0) might be signed or might be unsigned,
 175                  * depending on whether the kernel defines NLMSG_ALIGNTO
 176                  * as 4, which older kernels do, or as 4U, which newer
 177                  * kernels do.
 178                  *
 179                  * ep - bp is of type ptrdiff_t, which is signed.
 180                  *
 181                  * To squelch warnings, we cast both to size_t, which
 182                  * is unsigned; ep >= bp, so the cast is safe.
 183                  */
 184                 if ((size_t)(ep - bp) < (size_t)NLMSG_SPACE(0)) {
 185                         /*
 186                          * There's less than one netlink message left
 187                          * in the buffer.  Give up.
 188                          */
 189                         break;
 190                 }
 191
 192                 if (nlh->nlmsg_len < sizeof(struct nlmsghdr) || cc < nlh->nlmsg_len) {
 193                         snprintf(handle->errbuf, PCAP_ERRBUF_SIZE, "Message truncated: (got: %u) (nlmsg_len: %u)", cc, nlh->nlmsg_len);
 194                         return -1;
 195                 }
 196
 197                 if (NFNL_SUBSYS_ID(nlh->nlmsg_type) == NFNL_SUBSYS_ULOG &&
 198                     NFNL_MSG_TYPE(nlh->nlmsg_type) == NFULNL_MSG_PACKET)
 199                         type = NFLOG;
 200                 else if (NFNL_SUBSYS_ID(nlh->nlmsg_type) == NFNL_SUBSYS_QUEUE &&
 201                          NFNL_MSG_TYPE(nlh->nlmsg_type) == NFQNL_MSG_PACKET)
 202                         type = NFQUEUE;
 203
 204                 if (type != OTHER) {
 205                         const unsigned char *payload = NULL;
 206                         struct pcap_pkthdr pkth;
 207
 208                         const struct nfgenmsg *nfg = NULL;
 209                         int id = 0;
 210
 211                         if (handle->linktype != DLT_NFLOG) {
 212                                 const struct nfattr *payload_attr = NULL;
 213
 214                                 if (nlh->nlmsg_len < HDR_LENGTH) {
 215                                         snprintf(handle->errbuf, PCAP_ERRBUF_SIZE, "Malformed message: (nlmsg_len: %u)", nlh->nlmsg_len);
 216                                         return -1;
 217                                 }
 218
 219                                 nfg = NLMSG_DATA(nlh);
 220                                 if (nlh->nlmsg_len > HDR_LENGTH) {
 221                                         struct nfattr *attr = NFM_NFA(nfg);
 222                                         int attr_len = nlh->nlmsg_len - NLMSG_ALIGN(HDR_LENGTH);
 223
 224                                         while (NFA_OK(attr, attr_len)) {
 225                                                 if (type == NFQUEUE) {
 226                                                         switch (NFA_TYPE(attr)) {
 227                                                                 case NFQA_PACKET_HDR:
 228                                                                         {
 229                                                                                 const struct nfqnl_msg_packet_hdr *pkt_hdr = (const struct nfqnl_msg_packet_hdr *) NFA_DATA(attr);
 230
 231                                                                                 id = ntohl(pkt_hdr->packet_id);
 232                                                                                 break;
 233                                                                         }
 234                                                                 case NFQA_PAYLOAD:
 235                                                                         payload_attr = attr;
 236                                                                         break;
 237                                                         }
 238
 239                                                 } else if (type == NFLOG) {
 240                                                         switch (NFA_TYPE(attr)) {
 241                                                                 case NFULA_PAYLOAD:
 242                                                                         payload_attr = attr;
 243                                                                         break;
 244                                                         }
 245                                                 }
 246                                                 attr = NFA_NEXT(attr, attr_len);
 247                                         }
 248                                 }
 249
 250                                 if (payload_attr) {
 251                                         payload = NFA_DATA(payload_attr);
 252                                         pkth.len = pkth.caplen = NFA_PAYLOAD(payload_attr);
 253                                 }
 254
 255                         } else {
 256                                 payload = NLMSG_DATA(nlh);
 257                                 pkth.caplen = pkth.len = nlh->nlmsg_len-NLMSG_ALIGN(sizeof(struct nlmsghdr));
 258                         }
 259
 260                         if (payload) {
 261                                 /* pkth.caplen = min (payload_len, handle->snapshot); */
 262
 263                                 gettimeofday(&pkth.ts, NULL);
 264                                 if (handle->fcode.bf_insns == NULL ||
 265                                                 pcapint_filter(handle->fcode.bf_insns, payload, pkth.len, pkth.caplen))
 266                                 {
 267                                         handlep->packets_read++;
 268                                         callback(user, &pkth, payload);
 269                                         count++;
 270                                 }
 271                         }
 272
 273                         if (type == NFQUEUE) {
 274                                 /* XXX, possible responses: NF_DROP, NF_ACCEPT, NF_STOLEN, NF_QUEUE, NF_REPEAT, NF_STOP */
 275                                 /* if type == NFQUEUE, handle->linktype is always != DLT_NFLOG,
 276                                    so nfg is always initialized to NLMSG_DATA(nlh). */
 277                                 if (nfg != NULL)
 278                                         nfqueue_send_verdict(handle, ntohs(nfg->res_id), id, NF_ACCEPT);
 279                         }
 280                 }
 281
 282                 msg_len = NLMSG_ALIGN(nlh->nlmsg_len);
 283                 /*
 284                  * If the message length would run past the end of the
 285                  * buffer, truncate it to the remaining space in the
 286                  * buffer.
 287                  *
 288                  * To squelch warnings, we cast ep - bp to uint32_t, which
 289                  * is unsigned and is the type of msg_len; ep >= bp, and
 290                  * len should fit in 32 bits (either it's set from an int
 291                  * or it's set from a recv() call with a buffer size that's
 292                  * an int, and we're assuming either ILP32 or LP64), so
 293                  * the cast is safe.
 294                  */
 295                 if (msg_len > (uint32_t)(ep - bp))
 296                         msg_len = (uint32_t)(ep - bp);
 297
 298                 bp += msg_len;
 299                 if (count >= max_packets && !PACKET_COUNT_IS_UNLIMITED(max_packets)) {
 300                         handle->bp = bp;
 301                         handle->cc = (u_int)(ep - bp);
 302                         return count;
 303                 }
 304         }
 305
 306         handle->cc = 0;
 307         return count;
 308 }
 309
 310 static int
 311 netfilter_set_datalink(pcap_t *handle, int dlt)
 312 {
 313         handle->linktype = dlt;
 314         return 0;
 315 }
 316
 317 static int
 318 netfilter_stats_linux(pcap_t *handle, struct pcap_stat *stats)
 319 {
 320         struct pcap_netfilter *handlep = handle->priv;
 321
 322         stats->ps_recv = handlep->packets_read;
 323         stats->ps_drop = handlep->packets_nobufs;
 324         stats->ps_ifdrop = 0;
 325         return 0;
 326 }
 327
 328 static int
 329 netfilter_inject_linux(pcap_t *handle, const void *buf _U_, int size _U_)
 330 {
 331         snprintf(handle->errbuf, PCAP_ERRBUF_SIZE,
 332             "Packet injection is not supported on netfilter devices");
 333         return (-1);
 334 }
 335
 336 struct my_nfattr {
 337         uint16_t nfa_len;
 338         uint16_t nfa_type;
 339         void *data;
 340 };
 341
 342 static int
 343 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)
 344 {
 345         char buf[1024] __attribute__ ((aligned));
 346         memset(buf, 0, sizeof(buf));
 347
 348         struct nlmsghdr *nlh = (struct nlmsghdr *) buf;
 349         struct nfgenmsg *nfg = (struct nfgenmsg *) (buf + sizeof(struct nlmsghdr));
 350
 351         struct sockaddr_nl snl;
 352         static unsigned int seq_id;
 353
 354         if (!seq_id)
 355 DIAG_OFF_NARROWING
 356                 seq_id = time(NULL);
 357 DIAG_ON_NARROWING
 358         ++seq_id;
 359
 360         nlh->nlmsg_len = NLMSG_LENGTH(sizeof(struct nfgenmsg));
 361         nlh->nlmsg_type = msg_type;
 362         nlh->nlmsg_flags = NLM_F_REQUEST | (ack ? NLM_F_ACK : 0);
 363         nlh->nlmsg_pid = 0;     /* to kernel */
 364         nlh->nlmsg_seq = seq_id;
 365
 366         nfg->nfgen_family = family;
 367         nfg->version = NFNETLINK_V0;
 368         nfg->res_id = htons(res_id);
 369
 370         if (mynfa) {
 371                 struct nfattr *nfa = (struct nfattr *) (buf + NLMSG_ALIGN(nlh->nlmsg_len));
 372
 373                 nfa->nfa_type = mynfa->nfa_type;
 374                 nfa->nfa_len = NFA_LENGTH(mynfa->nfa_len);
 375                 memcpy(NFA_DATA(nfa), mynfa->data, mynfa->nfa_len);
 376                 nlh->nlmsg_len = NLMSG_ALIGN(nlh->nlmsg_len) + NFA_ALIGN(nfa->nfa_len);
 377         }
 378
 379         memset(&snl, 0, sizeof(snl));
 380         snl.nl_family = AF_NETLINK;
 381
 382         if (sendto(handle->fd, nlh, nlh->nlmsg_len, 0, (struct sockaddr *) &snl, sizeof(snl)) == -1)
 383                 return -1;
 384
 385         if (!ack)
 386                 return 0;
 387
 388         /* waiting for reply loop */
 389         do {
 390                 socklen_t addrlen = sizeof(snl);
 391                 int len;
 392
 393                 /* ignore interrupt system call error */
 394                 do {
 395                         /*
 396                          * The buffer is not so big that its size won't
 397                          * fit into an int.
 398                          */
 399                         len = (int)recvfrom(handle->fd, buf, sizeof(buf), 0, (struct sockaddr *) &snl, &addrlen);
 400                 } while ((len == -1) && (errno == EINTR));
 401
 402                 if (len <= 0)
 403                         return len;
 404
 405                 if (addrlen != sizeof(snl) || snl.nl_family != AF_NETLINK) {
 406                         errno = EINVAL;
 407                         return -1;
 408                 }
 409
 410                 nlh = (struct nlmsghdr *) buf;
 411                 if (snl.nl_pid != 0 || seq_id != nlh->nlmsg_seq)        /* if not from kernel or wrong sequence skip */
 412                         continue;
 413
 414                 while ((u_int)len >= NLMSG_SPACE(0) && NLMSG_OK(nlh, (u_int)len)) {
 415                         if (nlh->nlmsg_type == NLMSG_ERROR || (nlh->nlmsg_type == NLMSG_DONE && nlh->nlmsg_flags & NLM_F_MULTI)) {
 416                                 if (nlh->nlmsg_len < NLMSG_ALIGN(sizeof(struct nlmsgerr))) {
 417                                         errno = EBADMSG;
 418                                         return -1;
 419                                 }
 420                                 errno = -(*((int *)NLMSG_DATA(nlh)));
 421                                 return (errno == 0) ? 0 : -1;
 422                         }
 423                         nlh = NLMSG_NEXT(nlh, len);
 424                 }
 425         } while (1);
 426
 427         return -1; /* never here */
 428 }
 429
 430 static int
 431 nflog_send_config_msg(const pcap_t *handle, uint8_t family, u_int16_t group_id, const struct my_nfattr *mynfa)
 432 {
 433         return netfilter_send_config_msg(handle, (NFNL_SUBSYS_ULOG << 8) | NFULNL_MSG_CONFIG, 1, family, group_id, mynfa);
 434 }
 435
 436 static int
 437 nflog_send_config_cmd(const pcap_t *handle, uint16_t group_id, u_int8_t cmd, u_int8_t family)
 438 {
 439         struct nfulnl_msg_config_cmd msg;
 440         struct my_nfattr nfa;
 441
 442         msg.command = cmd;
 443
 444         nfa.data = &msg;
 445         nfa.nfa_type = NFULA_CFG_CMD;
 446         nfa.nfa_len = sizeof(msg);
 447
 448         return nflog_send_config_msg(handle, family, group_id, &nfa);
 449 }
 450
 451 static int
 452 nflog_send_config_mode(const pcap_t *handle, uint16_t group_id, u_int8_t copy_mode, u_int32_t copy_range)
 453 {
 454         struct nfulnl_msg_config_mode msg;
 455         struct my_nfattr nfa;
 456
 457         msg.copy_range = htonl(copy_range);
 458         msg.copy_mode = copy_mode;
 459
 460         nfa.data = &msg;
 461         nfa.nfa_type = NFULA_CFG_MODE;
 462         nfa.nfa_len = sizeof(msg);
 463
 464         return nflog_send_config_msg(handle, AF_UNSPEC, group_id, &nfa);
 465 }
 466
 467 static int
 468 nfqueue_send_verdict(const pcap_t *handle, uint16_t group_id, u_int32_t id, u_int32_t verdict)
 469 {
 470         struct nfqnl_msg_verdict_hdr msg;
 471         struct my_nfattr nfa;
 472
 473         msg.id = htonl(id);
 474         msg.verdict = htonl(verdict);
 475
 476         nfa.data = &msg;
 477         nfa.nfa_type = NFQA_VERDICT_HDR;
 478         nfa.nfa_len = sizeof(msg);
 479
 480         return netfilter_send_config_msg(handle, (NFNL_SUBSYS_QUEUE << 8) | NFQNL_MSG_VERDICT, 0, AF_UNSPEC, group_id, &nfa);
 481 }
 482
 483 static int
 484 nfqueue_send_config_msg(const pcap_t *handle, uint8_t family, u_int16_t group_id, const struct my_nfattr *mynfa)
 485 {
 486         return netfilter_send_config_msg(handle, (NFNL_SUBSYS_QUEUE << 8) | NFQNL_MSG_CONFIG, 1, family, group_id, mynfa);
 487 }
 488
 489 static int
 490 nfqueue_send_config_cmd(const pcap_t *handle, uint16_t group_id, u_int8_t cmd, u_int16_t pf)
 491 {
 492         struct nfqnl_msg_config_cmd msg;
 493         struct my_nfattr nfa;
 494
 495         msg.command = cmd;
 496         msg.pf = htons(pf);
 497
 498         nfa.data = &msg;
 499         nfa.nfa_type = NFQA_CFG_CMD;
 500         nfa.nfa_len = sizeof(msg);
 501
 502         return nfqueue_send_config_msg(handle, AF_UNSPEC, group_id, &nfa);
 503 }
 504
 505 static int
 506 nfqueue_send_config_mode(const pcap_t *handle, uint16_t group_id, u_int8_t copy_mode, u_int32_t copy_range)
 507 {
 508         struct nfqnl_msg_config_params msg;
 509         struct my_nfattr nfa;
 510
 511         msg.copy_range = htonl(copy_range);
 512         msg.copy_mode = copy_mode;
 513
 514         nfa.data = &msg;
 515         nfa.nfa_type = NFQA_CFG_PARAMS;
 516         nfa.nfa_len = sizeof(msg);
 517
 518         return nfqueue_send_config_msg(handle, AF_UNSPEC, group_id, &nfa);
 519 }
 520
 521 static int
 522 netfilter_activate(pcap_t* handle)
 523 {
 524         const char *dev = handle->opt.device;
 525         unsigned short groups[32];
 526         int group_count = 0;
 527         nftype_t type = OTHER;
 528         int i;
 529
 530         if (strncmp(dev, NFLOG_IFACE, strlen(NFLOG_IFACE)) == 0) {
 531                 dev += strlen(NFLOG_IFACE);
 532                 type = NFLOG;
 533
 534         } else if (strncmp(dev, NFQUEUE_IFACE, strlen(NFQUEUE_IFACE)) == 0) {
 535                 dev += strlen(NFQUEUE_IFACE);
 536                 type = NFQUEUE;
 537         }
 538
 539         if (type != OTHER && *dev == ':') {
 540                 dev++;
 541                 while (*dev) {
 542                         long int group_id;
 543                         char *end_dev;
 544
 545                         if (group_count == 32) {
 546                                 snprintf(handle->errbuf, PCAP_ERRBUF_SIZE,
 547                                                 "Maximum 32 netfilter groups! dev: %s",
 548                                                 handle->opt.device);
 549                                 return PCAP_ERROR;
 550                         }
 551
 552                         group_id = strtol(dev, &end_dev, 0);
 553                         if (end_dev != dev) {
 554                                 if (group_id < 0 || group_id > 65535) {
 555                                         snprintf(handle->errbuf, PCAP_ERRBUF_SIZE,
 556                                                         "Netfilter group range from 0 to 65535 (got %ld)",
 557                                                         group_id);
 558                                         return PCAP_ERROR;
 559                                 }
 560
 561                                 groups[group_count++] = (unsigned short) group_id;
 562                                 dev = end_dev;
 563                         }
 564                         if (*dev != ',')
 565                                 break;
 566                         dev++;
 567                 }
 568         }
 569
 570         if (type == OTHER || *dev) {
 571                 snprintf(handle->errbuf, PCAP_ERRBUF_SIZE,
 572                                 "Can't get netfilter group(s) index from %s",
 573                                 handle->opt.device);
 574                 return PCAP_ERROR;
 575         }
 576
 577         /* if no groups, add default: 0 */
 578         if (!group_count) {
 579                 groups[0] = 0;
 580                 group_count = 1;
 581         }
 582
 583         /*
 584          * Turn a negative snapshot value (invalid), a snapshot value of
 585          * 0 (unspecified), or a value bigger than the normal maximum
 586          * value, into the maximum allowed value.
 587          *
 588          * If some application really *needs* a bigger snapshot
 589          * length, we should just increase MAXIMUM_SNAPLEN.
 590          */
 591         if (handle->snapshot <= 0 || handle->snapshot > MAXIMUM_SNAPLEN)
 592                 handle->snapshot = MAXIMUM_SNAPLEN;
 593
 594         /* Initialize some components of the pcap structure. */
 595         handle->bufsize = 128 + handle->snapshot;
 596         handle->offset = 0;
 597         handle->read_op = netfilter_read_linux;
 598         handle->inject_op = netfilter_inject_linux;
 599         handle->setfilter_op = pcapint_install_bpf_program; /* no kernel filtering */
 600         handle->setdirection_op = NULL;
 601         handle->set_datalink_op = netfilter_set_datalink;
 602         handle->getnonblock_op = pcapint_getnonblock_fd;
 603         handle->setnonblock_op = pcapint_setnonblock_fd;
 604         handle->stats_op = netfilter_stats_linux;
 605
 606         /* Create netlink socket */
 607         handle->fd = socket(AF_NETLINK, SOCK_RAW, NETLINK_NETFILTER);
 608         if (handle->fd < 0) {
 609                 pcapint_fmt_errmsg_for_errno(handle->errbuf, PCAP_ERRBUF_SIZE,
 610                     errno, "Can't create raw socket");
 611                 return PCAP_ERROR;
 612         }
 613
 614         if (type == NFLOG) {
 615                 handle->linktype = DLT_NFLOG;
 616                 handle->dlt_list = (u_int *) malloc(sizeof(u_int) * 2);
 617                 if (handle->dlt_list == NULL) {
 618                         pcapint_fmt_errmsg_for_errno(handle->errbuf,
 619                             PCAP_ERRBUF_SIZE, errno,
 620                             "Can't allocate DLT list");
 621                         goto close_fail;
 622                 }
 623                 handle->dlt_list[0] = DLT_NFLOG;
 624                 handle->dlt_list[1] = DLT_IPV4;
 625                 handle->dlt_count = 2;
 626         } else
 627                 handle->linktype = DLT_IPV4;
 628
 629         handle->buffer = malloc(handle->bufsize);
 630         if (!handle->buffer) {
 631                 pcapint_fmt_errmsg_for_errno(handle->errbuf, PCAP_ERRBUF_SIZE,
 632                     errno, "Can't allocate dump buffer");
 633                 goto close_fail;
 634         }
 635
 636         if (type == NFLOG) {
 637                 if (nflog_send_config_cmd(handle, 0, NFULNL_CFG_CMD_PF_UNBIND, AF_INET) < 0) {
 638                         pcapint_fmt_errmsg_for_errno(handle->errbuf,
 639                             PCAP_ERRBUF_SIZE, errno,
 640                             "NFULNL_CFG_CMD_PF_UNBIND");
 641                         goto close_fail;
 642                 }
 643
 644                 if (nflog_send_config_cmd(handle, 0, NFULNL_CFG_CMD_PF_BIND, AF_INET) < 0) {
 645                         pcapint_fmt_errmsg_for_errno(handle->errbuf,
 646                             PCAP_ERRBUF_SIZE, errno, "NFULNL_CFG_CMD_PF_BIND");
 647                         goto close_fail;
 648                 }
 649
 650                 /* Bind socket to the nflog groups */
 651                 for (i = 0; i < group_count; i++) {
 652                         if (nflog_send_config_cmd(handle, groups[i], NFULNL_CFG_CMD_BIND, AF_UNSPEC) < 0) {
 653                                 pcapint_fmt_errmsg_for_errno(handle->errbuf,
 654                                     PCAP_ERRBUF_SIZE, errno,
 655                                     "Can't listen on group index");
 656                                 goto close_fail;
 657                         }
 658
 659                         if (nflog_send_config_mode(handle, groups[i], NFULNL_COPY_PACKET, handle->snapshot) < 0) {
 660                                 pcapint_fmt_errmsg_for_errno(handle->errbuf,
 661                                     PCAP_ERRBUF_SIZE, errno,
 662                                     "NFULNL_COPY_PACKET");
 663                                 goto close_fail;
 664                         }
 665                 }
 666
 667         } else {
 668                 if (nfqueue_send_config_cmd(handle, 0, NFQNL_CFG_CMD_PF_UNBIND, AF_INET) < 0) {
 669                         pcapint_fmt_errmsg_for_errno(handle->errbuf,
 670                             PCAP_ERRBUF_SIZE, errno, "NFQNL_CFG_CMD_PF_UNBIND");
 671                         goto close_fail;
 672                 }
 673
 674                 if (nfqueue_send_config_cmd(handle, 0, NFQNL_CFG_CMD_PF_BIND, AF_INET) < 0) {
 675                         pcapint_fmt_errmsg_for_errno(handle->errbuf,
 676                             PCAP_ERRBUF_SIZE, errno, "NFQNL_CFG_CMD_PF_BIND");
 677                         goto close_fail;
 678                 }
 679
 680                 /* Bind socket to the nfqueue groups */
 681                 for (i = 0; i < group_count; i++) {
 682                         if (nfqueue_send_config_cmd(handle, groups[i], NFQNL_CFG_CMD_BIND, AF_UNSPEC) < 0) {
 683                                 pcapint_fmt_errmsg_for_errno(handle->errbuf,
 684                                     PCAP_ERRBUF_SIZE, errno,
 685                                     "Can't listen on group index");
 686                                 goto close_fail;
 687                         }
 688
 689                         if (nfqueue_send_config_mode(handle, groups[i], NFQNL_COPY_PACKET, handle->snapshot) < 0) {
 690                                 pcapint_fmt_errmsg_for_errno(handle->errbuf,
 691                                     PCAP_ERRBUF_SIZE, errno,
 692                                     "NFQNL_COPY_PACKET");
 693                                 goto close_fail;
 694                         }
 695                 }
 696         }
 697
 698         if (handle->opt.rfmon) {
 699                 /*
 700                  * Monitor mode doesn't apply to netfilter devices.
 701                  */
 702                 pcapint_cleanup_live_common(handle);
 703                 return PCAP_ERROR_RFMON_NOTSUP;
 704         }
 705
 706         if (handle->opt.buffer_size != 0) {
 707                 /*
 708                  * Set the socket buffer size to the specified value.
 709                  */
 710                 if (setsockopt(handle->fd, SOL_SOCKET, SO_RCVBUF, &handle->opt.buffer_size, sizeof(handle->opt.buffer_size)) == -1) {
 711                         pcapint_fmt_errmsg_for_errno(handle->errbuf,
 712                             PCAP_ERRBUF_SIZE, errno, "SO_RCVBUF");
 713                         goto close_fail;
 714                 }
 715         }
 716
 717         handle->selectable_fd = handle->fd;
 718         return 0;
 719
 720 close_fail:
 721         pcapint_cleanup_live_common(handle);
 722         return PCAP_ERROR;
 723 }
 724
 725 pcap_t *
 726 netfilter_create(const char *device, char *ebuf, int *is_ours)
 727 {
 728         const char *cp;
 729         pcap_t *p;
 730
 731         /* Does this look like an netfilter device? */
 732         cp = device;
 733
 734         /* Does it begin with NFLOG_IFACE or NFQUEUE_IFACE? */
 735         if (strncmp(cp, NFLOG_IFACE, sizeof NFLOG_IFACE - 1) == 0)
 736                 cp += sizeof NFLOG_IFACE - 1;
 737         else if (strncmp(cp, NFQUEUE_IFACE, sizeof NFQUEUE_IFACE - 1) == 0)
 738                 cp += sizeof NFQUEUE_IFACE - 1;
 739         else {
 740                 /* Nope, doesn't begin with NFLOG_IFACE nor NFQUEUE_IFACE */
 741                 *is_ours = 0;
 742                 return NULL;
 743         }
 744
 745         /*
 746          * Yes - is that either the end of the name, or is it followed
 747          * by a colon?
 748          */
 749         if (*cp != ':' && *cp != '\0') {
 750                 /* Nope */
 751                 *is_ours = 0;
 752                 return NULL;
 753         }
 754
 755         /* OK, it's probably ours. */
 756         *is_ours = 1;
 757
 758         p = PCAP_CREATE_COMMON(ebuf, struct pcap_netfilter);
 759         if (p == NULL)
 760                 return (NULL);
 761
 762         p->activate_op = netfilter_activate;
 763         return (p);
 764 }
 765
 766 int
 767 netfilter_findalldevs(pcap_if_list_t *devlistp, char *err_str)
 768 {
 769         int sock;
 770
 771         sock = socket(AF_NETLINK, SOCK_RAW, NETLINK_NETFILTER);
 772         if (sock < 0) {
 773                 /* if netlink is not supported this is not fatal */
 774                 if (errno == EAFNOSUPPORT || errno == EPROTONOSUPPORT)
 775                         return 0;
 776                 pcapint_fmt_errmsg_for_errno(err_str, PCAP_ERRBUF_SIZE,
 777                     errno, "Can't open netlink socket");
 778                 return -1;
 779         }
 780         close(sock);
 781
 782         /*
 783          * The notion of "connected" vs. "disconnected" doesn't apply.
 784          * XXX - what about "up" and "running"?
 785          */
 786         if (pcapint_add_dev(devlistp, NFLOG_IFACE,
 787             PCAP_IF_CONNECTION_STATUS_NOT_APPLICABLE,
 788             "Linux netfilter log (NFLOG) interface", err_str) == NULL)
 789                 return -1;
 790         if (pcapint_add_dev(devlistp, NFQUEUE_IFACE,
 791             PCAP_IF_CONNECTION_STATUS_NOT_APPLICABLE,
 792             "Linux netfilter queue (NFQUEUE) interface", err_str) == NULL)
 793                 return -1;
 794         return 0;
 795 }