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