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