/src/libpcap/pcap-linux.c

Source
/*
 *  pcap-linux.c: Packet capture interface to the Linux kernel
 *
 *  Copyright (c) 2000 Torsten Landschoff <torsten@debian.org>
 *           Sebastian Krahmer  <krahmer@cs.uni-potsdam.de>
 *
 *  License: BSD
 *
 *  Redistribution and use in source and binary forms, with or without
 *  modification, are permitted provided that the following conditions
 *  are met:
 *
 *  1. Redistributions of source code must retain the above copyright
 *     notice, this list of conditions and the following disclaimer.
 *  2. Redistributions in binary form must reproduce the above copyright
 *     notice, this list of conditions and the following disclaimer in
 *     the documentation and/or other materials provided with the
 *     distribution.
 *  3. The names of the authors may not be used to endorse or promote
 *     products derived from this software without specific prior
 *     written permission.
 *
 *  THIS SOFTWARE IS PROVIDED ``AS IS'' AND WITHOUT ANY EXPRESS OR
 *  IMPLIED WARRANTIES, INCLUDING, WITHOUT LIMITATION, THE IMPLIED
 *  WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE.
 *
 *  Modifications:     Added PACKET_MMAP support
 *                     Paolo Abeni <paolo.abeni@email.it>
 *                     Added TPACKET_V3 support
 *                     Gabor Tatarka <gabor.tatarka@ericsson.com>
 *
 *                     based on previous works of:
 *                     Simon Patarin <patarin@cs.unibo.it>
 *                     Phil Wood <cpw@lanl.gov>
 *
 * Monitor-mode support for mac80211 includes code taken from the iw
 * command; the copyright notice for that code is
 *
 * Copyright (c) 2007, 2008 Johannes Berg
 * Copyright (c) 2007   Andy Lutomirski
 * Copyright (c) 2007   Mike Kershaw
 * Copyright (c) 2008   Gábor Stefanik
 *
 * All rights reserved.
 *
 * Redistribution and use in source and binary forms, with or without
 * modification, are permitted provided that the following conditions
 * are met:
 * 1. Redistributions of source code must retain the above copyright
 *    notice, this list of conditions and the following disclaimer.
 * 2. Redistributions in binary form must reproduce the above copyright
 *    notice, this list of conditions and the following disclaimer in the
 *    documentation and/or other materials provided with the distribution.
 * 3. The name of the author may not be used to endorse or promote products
 *    derived from this software without specific prior written permission.
 *
 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR
 * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES
 * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED.
 * IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT,
 * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING,
 * BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
 * LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED
 * AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
 * OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
 * SUCH DAMAGE.
 */


#ifndef _GNU_SOURCE
#define _GNU_SOURCE
#endif

#include <config.h>

#include <errno.h>
#include <stdio.h>
#include <stdlib.h>
#include <unistd.h>
#include <fcntl.h>
#include <string.h>
#include <limits.h>
#include <endian.h>
#include <sys/stat.h>
#include <sys/socket.h>
#include <sys/ioctl.h>
#include <sys/utsname.h>
#include <sys/mman.h>
#include <linux/if.h>
#include <linux/if_packet.h>
#include <linux/sockios.h>
#include <linux/ethtool.h>
#include <netinet/in.h>
#include <linux/if_ether.h>
#include <linux/netlink.h>

#include <linux/if_arp.h>
#ifndef ARPHRD_IEEE802154
  // Linux before 2.6.31
  #define ARPHRD_IEEE802154 804
#endif
#ifndef ARPHRD_IEEE802154_MONITOR
  // Linux before 3.5
  #define ARPHRD_IEEE802154_MONITOR 805
#endif
#ifndef ARPHRD_NETLINK
  // Linux before 3.11
  #define ARPHRD_NETLINK 824
#endif
#ifndef ARPHRD_6LOWPAN
  // Linux before 3.14
  #define ARPHRD_6LOWPAN 825
#endif
#ifndef ARPHRD_VSOCKMON
  // Linux before 4.12
  #define ARPHRD_VSOCKMON 826
#endif
#ifndef ARPHRD_LAPD
  /*
   * ARPHRD_LAPD is unofficial and randomly allocated, if reallocation
   * is needed, please report it to <daniele@orlandi.com>
   */
  #define ARPHRD_LAPD 8445
#endif

#include <poll.h>
#include <dirent.h>
#include <sys/eventfd.h>

#include "pcap-int.h"
#include "pcap-util.h"
#include "pcap-snf.h"
#include "pcap/sll.h"
#include "pcap/vlan.h"
#include "pcap/can_socketcan.h"

#include "diag-control.h"

/*
 * We require TPACKET_V2 support.
 */
#ifndef TPACKET2_HDRLEN
#error "Libpcap will only work if TPACKET_V2 is supported; you must build for a 2.6.27 or later kernel"
#endif

/* check for memory mapped access availability. We assume every needed
 * struct is defined if the macro TPACKET_HDRLEN is defined, because it
 * uses many ring related structs and macros */
#ifdef TPACKET3_HDRLEN
# define HAVE_TPACKET3
#endif /* TPACKET3_HDRLEN */

/*
 * Not all compilers that are used to compile code to run on Linux have
 * these builtins.  For example, older versions of GCC don't, and at
 * least some people are doing cross-builds for MIPS with older versions
 * of GCC.
 */
#ifndef HAVE___ATOMIC_LOAD_N
#define __atomic_load_n(ptr, memory_model)    (*(ptr))
#endif
#ifndef HAVE___ATOMIC_STORE_N
#define __atomic_store_n(ptr, val, memory_model)  *(ptr) = (val)
#endif

#define packet_mmap_acquire(pkt) \
  (__atomic_load_n(&pkt->tp_status, __ATOMIC_ACQUIRE) != TP_STATUS_KERNEL)
#define packet_mmap_release(pkt) \
  (__atomic_store_n(&pkt->tp_status, TP_STATUS_KERNEL, __ATOMIC_RELEASE))
#define packet_mmap_v3_acquire(pkt) \
  (__atomic_load_n(&pkt->hdr.bh1.block_status, __ATOMIC_ACQUIRE) != TP_STATUS_KERNEL)
#define packet_mmap_v3_release(pkt) \
  (__atomic_store_n(&pkt->hdr.bh1.block_status, TP_STATUS_KERNEL, __ATOMIC_RELEASE))

#include <linux/types.h>
#include <linux/filter.h>

#ifdef HAVE_LINUX_NET_TSTAMP_H
#include <linux/net_tstamp.h>
#endif

/*
 * For checking whether a device is a bonding device.
 */
#include <linux/if_bonding.h>

/*
 * Got libnl?
 */
#ifdef HAVE_LIBNL
#include <linux/nl80211.h>

#include <netlink/genl/genl.h>
#include <netlink/genl/family.h>
#include <netlink/genl/ctrl.h>
#include <netlink/msg.h>
#include <netlink/attr.h>
#endif /* HAVE_LIBNL */

#ifndef HAVE_SOCKLEN_T
typedef int   socklen_t;
#endif

#define MAX_LINKHEADER_SIZE 256

/*
 * When capturing on all interfaces we use this as the buffer size.
 * Should be bigger then all MTUs that occur in real life.
 * 64kB should be enough for now.
 */
#define BIGGER_THAN_ALL_MTUS  (64*1024)

/*
 * Private data for capturing on Linux PF_PACKET sockets.
 */
struct pcap_linux {
  long long sysfs_dropped; /* packets reported dropped by /sys/class/net/{if_name}/statistics/rx_{missed,fifo}_errors */
  struct pcap_stat stat;

  char  *device;  /* device name */
  int filter_in_userland; /* must filter in userland */
  u_int blocks_to_filter_in_userland;
  int must_do_on_close; /* stuff we must do when we close */
  int timeout;  /* timeout for buffering */
  int cooked;   /* using SOCK_DGRAM rather than SOCK_RAW */
  int ifindex;  /* interface index of device we're bound to */
  int lo_ifindex; /* interface index of the loopback device */
  int netdown;  /* we got an ENETDOWN and haven't resolved it */
  bpf_u_int32 oldmode;  /* mode to restore when turning monitor mode off */
  char  *mondevice; /* mac80211 monitor device we created */
  u_char  *mmapbuf; /* memory-mapped region pointer */
  size_t  mmapbuflen; /* size of region */
  int vlan_offset;  /* offset at which to insert vlan tags; if -1, don't insert */
  u_int tp_version; /* version of tpacket_hdr for mmaped ring */
  u_int tp_hdrlen;  /* hdrlen of tpacket_hdr for mmaped ring */
  u_char  *oneshot_buffer; /* buffer for copy of packet */
  int poll_timeout; /* timeout to use in poll() */
#ifdef HAVE_TPACKET3
  unsigned char *current_packet; /* Current packet within the TPACKET_V3 block. Move to next block if NULL. */
  int packets_left; /* Unhandled packets left within the block from previous call to pcap_read_linux_mmap_v3 in case of TPACKET_V3. */
#endif
  int poll_breakloop_fd; /* fd to an eventfd to break from blocking operations */
};

/*
 * Stuff to do when we close.
 */
#define MUST_DELETE_MONIF 0x00000001  /* delete monitor-mode interface */

/*
 * Prototypes for internal functions and methods.
 */
static int is_wifi(const char *);
static int pcap_activate_linux(pcap_t *);
static int setup_socket(pcap_t *, int);
static int setup_mmapped(pcap_t *);
static int pcap_can_set_rfmon_linux(pcap_t *);
static int pcap_inject_linux(pcap_t *, const void *, int);
static int pcap_stats_linux(pcap_t *, struct pcap_stat *);
static int pcap_setfilter_linux(pcap_t *, struct bpf_program *);
static int pcap_setdirection_linux(pcap_t *, pcap_direction_t);
static int pcap_set_datalink_linux(pcap_t *, int);

union thdr {
  struct tpacket2_hdr   *h2;
#ifdef HAVE_TPACKET3
  struct tpacket_block_desc *h3;
#endif
  u_char        *raw;
};

#define RING_GET_FRAME_AT(h, offset) (((u_char **)h->buffer)[(offset)])
#define RING_GET_CURRENT_FRAME(h) RING_GET_FRAME_AT(h, h->offset)

static void destroy_ring(pcap_t *handle);
static int create_ring(pcap_t *handle);
static int prepare_tpacket_socket(pcap_t *handle);
static int pcap_read_linux_mmap_v2(pcap_t *, int, pcap_handler , u_char *);
#ifdef HAVE_TPACKET3
static int pcap_read_linux_mmap_v3(pcap_t *, int, pcap_handler , u_char *);
#endif
static int pcap_setnonblock_linux(pcap_t *p, int nonblock);
static int pcap_getnonblock_linux(pcap_t *p);
static void pcapint_oneshot_linux(u_char *user, const struct pcap_pkthdr *h,
    const u_char *bytes);

/*
 * In pre-3.0 kernels, the tp_vlan_tci field is set to whatever the
 * vlan_tci field in the skbuff is.  0 can either mean "not on a VLAN"
 * or "on VLAN 0".  There is no flag set in the tp_status field to
 * distinguish between them.
 *
 * In 3.0 and later kernels, if there's a VLAN tag present, the tp_vlan_tci
 * field is set to the VLAN tag, and the TP_STATUS_VLAN_VALID flag is set
 * in the tp_status field, otherwise the tp_vlan_tci field is set to 0 and
 * the TP_STATUS_VLAN_VALID flag isn't set in the tp_status field.
 *
 * With a pre-3.0 kernel, we cannot distinguish between packets with no
 * VLAN tag and packets on VLAN 0, so we will mishandle some packets, and
 * there's nothing we can do about that.
 *
 * So, on those systems, which never set the TP_STATUS_VLAN_VALID flag, we
 * continue the behavior of earlier libpcaps, wherein we treated packets
 * with a VLAN tag of 0 as being packets without a VLAN tag rather than packets
 * on VLAN 0.  We do this by treating packets with a tp_vlan_tci of 0 and
 * with the TP_STATUS_VLAN_VALID flag not set in tp_status as not having
 * VLAN tags.  This does the right thing on 3.0 and later kernels, and
 * continues the old unfixably-imperfect behavior on pre-3.0 kernels.
 *
 * If TP_STATUS_VLAN_VALID isn't defined, we test it as the 0x10 bit; it
 * has that value in 3.0 and later kernels.
 */
#ifdef TP_STATUS_VLAN_VALID
  #define VLAN_VALID(hdr, hv) ((hv)->tp_vlan_tci != 0 || ((hdr)->tp_status & TP_STATUS_VLAN_VALID))
#else
  /*
   * This is being compiled on a system that lacks TP_STATUS_VLAN_VALID,
   * so we test with the value it has in the 3.0 and later kernels, so
   * we can test it if we're running on a system that has it.  (If we're
   * running on a system that doesn't have it, it won't be set in the
   * tp_status field, so the tests of it will always fail; that means
   * we behave the way we did before we introduced this macro.)
   */
  #define VLAN_VALID(hdr, hv) ((hv)->tp_vlan_tci != 0 || ((hdr)->tp_status & 0x10))
#endif

#ifdef TP_STATUS_VLAN_TPID_VALID
# define VLAN_TPID(hdr, hv) (((hv)->tp_vlan_tpid || ((hdr)->tp_status & TP_STATUS_VLAN_TPID_VALID)) ? (hv)->tp_vlan_tpid : ETH_P_8021Q)
#else
# define VLAN_TPID(hdr, hv) ETH_P_8021Q
#endif

/*
 * Required select timeout if we're polling for an "interface disappeared"
 * indication - 1 millisecond.
 */
static const struct timeval netdown_timeout = {
  0, 1000   /* 1000 microseconds = 1 millisecond */
};

/*
 * Wrap some ioctl calls
 */
static int  iface_get_id(int fd, const char *device, char *ebuf);
static int  iface_get_mtu(int fd, const char *device, char *ebuf);
static int  iface_get_arptype(int fd, const char *device, char *ebuf);
static int  iface_bind(int fd, int ifindex, char *ebuf, int protocol);
static int  enter_rfmon_mode(pcap_t *handle, int sock_fd,
    const char *device);
static int  iface_get_ts_types(const char *device, pcap_t *handle,
    char *ebuf);
static int  iface_get_offload(pcap_t *handle);

static int  fix_program(pcap_t *handle, struct sock_fprog *fcode);
static int  fix_offset(pcap_t *handle, struct bpf_insn *p);
static int  set_kernel_filter(pcap_t *handle, struct sock_fprog *fcode);
static int  reset_kernel_filter(pcap_t *handle);

static struct sock_filter total_insn
  = BPF_STMT(BPF_RET | BPF_K, 0);
static struct sock_fprog  total_fcode
  = { 1, &total_insn };

static int  iface_dsa_get_proto_info(const char *device, pcap_t *handle);

pcap_t *
pcapint_create_interface(const char *device, char *ebuf)
{
  pcap_t *handle;

  handle = PCAP_CREATE_COMMON(ebuf, struct pcap_linux);
  if (handle == NULL)
    return NULL;

  handle->activate_op = pcap_activate_linux;
  handle->can_set_rfmon_op = pcap_can_set_rfmon_linux;

  /*
   * See what time stamp types we support.
   */
  if (iface_get_ts_types(device, handle, ebuf) == -1) {
    pcap_close(handle);
    return NULL;
  }

  /*
   * We claim that we support microsecond and nanosecond time
   * stamps.
   *
   * XXX - with adapter-supplied time stamps, can we choose
   * microsecond or nanosecond time stamps on arbitrary
   * adapters?
   */
  handle->tstamp_precision_list = malloc(2 * sizeof(u_int));
  if (handle->tstamp_precision_list == NULL) {
    pcapint_fmt_errmsg_for_errno(ebuf, PCAP_ERRBUF_SIZE,
        errno, "malloc");
    pcap_close(handle);
    return NULL;
  }
  handle->tstamp_precision_list[0] = PCAP_TSTAMP_PRECISION_MICRO;
  handle->tstamp_precision_list[1] = PCAP_TSTAMP_PRECISION_NANO;
  handle->tstamp_precision_count = 2;

  /*
   * Start out with the breakloop handle not open; we don't
   * need it until we're activated and ready to capture.
   */
  struct pcap_linux *handlep = handle->priv;
  handlep->poll_breakloop_fd = -1;

  return handle;
}

#ifdef HAVE_LIBNL
/*
 * If interface {if_name} is a mac80211 driver, the file
 * /sys/class/net/{if_name}/phy80211 is a symlink to
 * /sys/class/ieee80211/{phydev_name}, for some {phydev_name}.
 *
 * On Fedora 9, with a 2.6.26.3-29 kernel, my Zydas stick, at
 * least, has a "wmaster0" device and a "wlan0" device; the
 * latter is the one with the IP address.  Both show up in
 * "tcpdump -D" output.  Capturing on the wmaster0 device
 * captures with 802.11 headers.
 *
 * airmon-ng searches through /sys/class/net for devices named
 * monN, starting with mon0; as soon as one *doesn't* exist,
 * it chooses that as the monitor device name.  If the "iw"
 * command exists, it does
 *
 *    iw dev {if_name} interface add {monif_name} type monitor
 *
 * where {monif_name} is the monitor device.  It then (sigh) sleeps
 * .1 second, and then configures the device up.  Otherwise, if
 * /sys/class/ieee80211/{phydev_name}/add_iface is a file, it writes
 * {mondev_name}, without a newline, to that file, and again (sigh)
 * sleeps .1 second, and then iwconfig's that device into monitor
 * mode and configures it up.  Otherwise, you can't do monitor mode.
 *
 * All these devices are "glued" together by having the
 * /sys/class/net/{if_name}/phy80211 links pointing to the same
 * place, so, given a wmaster, wlan, or mon device, you can
 * find the other devices by looking for devices with
 * the same phy80211 link.
 *
 * To turn monitor mode off, delete the monitor interface,
 * either with
 *
 *    iw dev {monif_name} interface del
 *
 * or by sending {monif_name}, with no NL, down
 * /sys/class/ieee80211/{phydev_name}/remove_iface
 *
 * Note: if you try to create a monitor device named "monN", and
 * there's already a "monN" device, it fails, as least with
 * the netlink interface (which is what iw uses), with a return
 * value of -ENFILE.  (Return values are negative errnos.)  We
 * could probably use that to find an unused device.
 *
 * Yes, you can have multiple monitor devices for a given
 * physical device.
 */

/*
 * Is this a mac80211 device?  If so, fill in the physical device path and
 * return 1; if not, return 0.  On an error, fill in handle->errbuf and
 * return PCAP_ERROR.
 */
static int
get_mac80211_phydev(pcap_t *handle, const char *device, char *phydev_path,
    size_t phydev_max_pathlen)
{
  char *pathstr;
  ssize_t bytes_read;

  /*
   * Generate the path string for the symlink to the physical device.
   */
  if (asprintf(&pathstr, "/sys/class/net/%s/phy80211", device) == -1) {
    snprintf(handle->errbuf, PCAP_ERRBUF_SIZE,
        "%s: Can't generate path name string for /sys/class/net device",
        device);
    return PCAP_ERROR;
  }
  bytes_read = readlink(pathstr, phydev_path, phydev_max_pathlen);
  if (bytes_read == -1) {
    if (errno == ENOENT) {
      /*
       * This either means that the directory
       * /sys/class/net/{device} exists but doesn't
       * have anything named "phy80211" in it,
       * in which case it's not a mac80211 device,
       * or that the directory doesn't exist,
       * in which case the device doesn't exist.
       *
       * Directly check whether the directory
       * exists.
       */
      struct stat statb;

      free(pathstr);
      if (asprintf(&pathstr, "/sys/class/net/%s", device) == -1) {
        snprintf(handle->errbuf, PCAP_ERRBUF_SIZE,
            "%s: Can't generate path name string for /sys/class/net device",
            device);
        return PCAP_ERROR;
      }
      if (stat(pathstr, &statb) == -1) {
        if (errno == ENOENT) {
          /*
           * No such device.
           */
          snprintf(handle->errbuf, PCAP_ERRBUF_SIZE,
              "%s: %s doesn't exist",
              device, pathstr);
          free(pathstr);
          return PCAP_ERROR_NO_SUCH_DEVICE;
        }
        snprintf(handle->errbuf, PCAP_ERRBUF_SIZE,
            "%s: Can't stat %s: %s",
            device, pathstr, strerror(errno));
        free(pathstr);
        return PCAP_ERROR;
      }

      /*
       * Path to the directory that would contain
       * "phy80211" exists, but "phy80211" doesn't
       * exist; that means it's not a mac80211
       * device.
       */
      free(pathstr);
      return 0;
    }
    if (errno == EINVAL) {
      /*
       * Exists, but it's not a symlink; assume that
       * means it's not a mac80211 device.
       */
      free(pathstr);
      return 0;
    }
    pcapint_fmt_errmsg_for_errno(handle->errbuf, PCAP_ERRBUF_SIZE,
        errno, "%s: Can't readlink %s", device, pathstr);
    free(pathstr);
    return PCAP_ERROR;
  }
  free(pathstr);
  phydev_path[bytes_read] = '\0';
  return 1;
}

struct nl80211_state {
  struct nl_sock *nl_sock;
  struct nl_cache *nl_cache;
  struct genl_family *nl80211;
};

static int
nl80211_init(pcap_t *handle, struct nl80211_state *state, const char *device)
{
  int err;

  state->nl_sock = nl_socket_alloc();
  if (!state->nl_sock) {
    snprintf(handle->errbuf, PCAP_ERRBUF_SIZE,
        "%s: failed to allocate netlink handle", device);
    return PCAP_ERROR;
  }

  if (genl_connect(state->nl_sock)) {
    snprintf(handle->errbuf, PCAP_ERRBUF_SIZE,
        "%s: failed to connect to generic netlink", device);
    goto out_handle_destroy;
  }

  err = genl_ctrl_alloc_cache(state->nl_sock, &state->nl_cache);
  if (err < 0) {
    snprintf(handle->errbuf, PCAP_ERRBUF_SIZE,
        "%s: failed to allocate generic netlink cache: %s",
        device, nl_geterror(-err));
    goto out_handle_destroy;
  }

  state->nl80211 = genl_ctrl_search_by_name(state->nl_cache, "nl80211");
  if (!state->nl80211) {
    snprintf(handle->errbuf, PCAP_ERRBUF_SIZE,
        "%s: nl80211 not found", device);
    goto out_cache_free;
  }

  return 0;

out_cache_free:
  nl_cache_free(state->nl_cache);
out_handle_destroy:
  nl_socket_free(state->nl_sock);
  return PCAP_ERROR;
}

static void
nl80211_cleanup(struct nl80211_state *state)
{
  genl_family_put(state->nl80211);
  nl_cache_free(state->nl_cache);
  nl_socket_free(state->nl_sock);
}

static int
del_mon_if(pcap_t *handle, int sock_fd, struct nl80211_state *state,
    const char *device, const char *mondevice);

static int
if_type_cb(struct nl_msg *msg, void* arg)
{
  struct nlmsghdr* ret_hdr = nlmsg_hdr(msg);
  struct nlattr *tb_msg[NL80211_ATTR_MAX + 1];
  int *type = (int*)arg;

  struct genlmsghdr *gnlh = (struct genlmsghdr*) nlmsg_data(ret_hdr);

  nla_parse(tb_msg, NL80211_ATTR_MAX, genlmsg_attrdata(gnlh, 0),
    genlmsg_attrlen(gnlh, 0), NULL);

  /*
   * We sent a message asking for info about a single index.
   * To be really paranoid, we could check if the index matched
   * by examining nla_get_u32(tb_msg[NL80211_ATTR_IFINDEX]).
   */

  if (tb_msg[NL80211_ATTR_IFTYPE]) {
    *type = nla_get_u32(tb_msg[NL80211_ATTR_IFTYPE]);
  }

  return NL_SKIP;
}

static int
get_if_type(pcap_t *handle, int sock_fd, struct nl80211_state *state,
    const char *device, int *type)
{
  int ifindex;
  struct nl_msg *msg;
  int err;

  ifindex = iface_get_id(sock_fd, device, handle->errbuf);
  if (ifindex == -1)
    return PCAP_ERROR;

  msg = nlmsg_alloc();
  if (!msg) {
    snprintf(handle->errbuf, PCAP_ERRBUF_SIZE,
        "%s: failed to allocate netlink msg", device);
    return PCAP_ERROR;
  }

  genlmsg_put(msg, NL_AUTO_PORT, NL_AUTO_SEQ,
        genl_family_get_id(state->nl80211), 0,
        0, NL80211_CMD_GET_INTERFACE, 0);
  NLA_PUT_U32(msg, NL80211_ATTR_IFINDEX, ifindex);

  err = nl_send_auto(state->nl_sock, msg);
  nlmsg_free(msg);
  if (err < 0) {
    if (err == -NLE_FAILURE) {
      /*
       * Device not available; our caller should just
       * keep trying.  (libnl 2.x maps ENFILE to
       * NLE_FAILURE; it can also map other errors
       * to that, but there's not much we can do
       * about that.)
       */
      return 0;
    } else {
      /*
       * Real failure, not just "that device is not
       * available.
       */
      snprintf(handle->errbuf, PCAP_ERRBUF_SIZE,
          "%s: nl_send_auto failed getting interface type: %s",
          device, nl_geterror(-err));
      return PCAP_ERROR;
    }
  }

  struct nl_cb *cb = nl_cb_alloc(NL_CB_DEFAULT);
  nl_cb_set(cb, NL_CB_VALID, NL_CB_CUSTOM, if_type_cb, (void*)type);
  err = nl_recvmsgs(state->nl_sock, cb);
  nl_cb_put(cb);

  if (err < 0) {
    snprintf(handle->errbuf, PCAP_ERRBUF_SIZE,
        "%s: nl_recvmsgs failed getting interface type: %s",
        device, nl_geterror(-err));
    return PCAP_ERROR;
  }

  /*
  * If this is a mac80211 device not in monitor mode, nl_sock will be
  * reused for add_mon_if. So we must wait for the ACK here so that
  * add_mon_if does not receive it instead and incorrectly interpret
  * the ACK as its NEW_INTERFACE command succeeding, even when it fails.
  */
  err = nl_wait_for_ack(state->nl_sock);
  if (err < 0) {
    snprintf(handle->errbuf, PCAP_ERRBUF_SIZE,
        "%s: nl_wait_for_ack failed getting interface type: %s",
        device, nl_geterror(-err));
    return PCAP_ERROR;
  }

  /*
   * Success.
   */
  return 1;

nla_put_failure:
  snprintf(handle->errbuf, PCAP_ERRBUF_SIZE,
      "%s: nl_put failed getting interface type",
      device);
  nlmsg_free(msg);
  // Do not call nl_cb_put(): nl_cb_alloc() has not been called.
  return PCAP_ERROR;
}

static int
add_mon_if(pcap_t *handle, int sock_fd, struct nl80211_state *state,
    const char *device, const char *mondevice)
{
  struct pcap_linux *handlep = handle->priv;
  int ifindex;
  struct nl_msg *msg;
  int err;

  ifindex = iface_get_id(sock_fd, device, handle->errbuf);
  if (ifindex == -1)
    return PCAP_ERROR;

  msg = nlmsg_alloc();
  if (!msg) {
    snprintf(handle->errbuf, PCAP_ERRBUF_SIZE,
        "%s: failed to allocate netlink msg", device);
    return PCAP_ERROR;
  }

  genlmsg_put(msg, NL_AUTO_PORT, NL_AUTO_SEQ,
        genl_family_get_id(state->nl80211), 0,
        0, NL80211_CMD_NEW_INTERFACE, 0);
  NLA_PUT_U32(msg, NL80211_ATTR_IFINDEX, ifindex);
DIAG_OFF_NARROWING
  NLA_PUT_STRING(msg, NL80211_ATTR_IFNAME, mondevice);
DIAG_ON_NARROWING
  NLA_PUT_U32(msg, NL80211_ATTR_IFTYPE, NL80211_IFTYPE_MONITOR);

  err = nl_send_sync(state->nl_sock, msg); // calls nlmsg_free()
  if (err < 0) {
    switch (err) {

    case -NLE_FAILURE:
    case -NLE_AGAIN:
      /*
       * Device not available; our caller should just
       * keep trying.  (libnl 2.x maps ENFILE to
       * NLE_FAILURE; it can also map other errors
       * to that, but there's not much we can do
       * about that.)
       */
      return 0;

    case -NLE_OPNOTSUPP:
      /*
       * Device is a mac80211 device but adding it as a
       * monitor mode device isn't supported.  Report our
       * error.
       */
      return PCAP_ERROR_RFMON_NOTSUP;

    default:
      /*
       * Real failure, not just "that device is not
       * available."  Report a generic error, using the
       * error message from libnl.
       */
      snprintf(handle->errbuf, PCAP_ERRBUF_SIZE,
          "%s: nl_send_sync failed adding %s interface: %s",
          device, mondevice, nl_geterror(-err));
      return PCAP_ERROR;
    }
  }

  /*
   * Success.
   */

  /*
   * Try to remember the monitor device.
   */
  handlep->mondevice = strdup(mondevice);
  if (handlep->mondevice == NULL) {
    pcapint_fmt_errmsg_for_errno(handle->errbuf, PCAP_ERRBUF_SIZE,
        errno, "strdup");
    /*
     * Get rid of the monitor device.
     */
    del_mon_if(handle, sock_fd, state, device, mondevice);
    return PCAP_ERROR;
  }
  return 1;

nla_put_failure:
  snprintf(handle->errbuf, PCAP_ERRBUF_SIZE,
      "%s: nl_put failed adding %s interface",
      device, mondevice);
  nlmsg_free(msg);
  return PCAP_ERROR;
}

static int
del_mon_if(pcap_t *handle, int sock_fd, struct nl80211_state *state,
    const char *device, const char *mondevice)
{
  int ifindex;
  struct nl_msg *msg;
  int err;

  ifindex = iface_get_id(sock_fd, mondevice, handle->errbuf);
  if (ifindex == -1)
    return PCAP_ERROR;

  msg = nlmsg_alloc();
  if (!msg) {
    snprintf(handle->errbuf, PCAP_ERRBUF_SIZE,
        "%s: failed to allocate netlink msg", device);
    return PCAP_ERROR;
  }

  genlmsg_put(msg, NL_AUTO_PORT, NL_AUTO_SEQ,
        genl_family_get_id(state->nl80211), 0,
        0, NL80211_CMD_DEL_INTERFACE, 0);
  NLA_PUT_U32(msg, NL80211_ATTR_IFINDEX, ifindex);

  err = nl_send_sync(state->nl_sock, msg); // calls nlmsg_free()
  if (err < 0) {
    snprintf(handle->errbuf, PCAP_ERRBUF_SIZE,
        "%s: nl_send_sync failed deleting %s interface: %s",
        device, mondevice, nl_geterror(-err));
    return PCAP_ERROR;
  }

  /*
   * Success.
   */
  return 1;

nla_put_failure:
  snprintf(handle->errbuf, PCAP_ERRBUF_SIZE,
      "%s: nl_put failed deleting %s interface",
      device, mondevice);
  nlmsg_free(msg);
  return PCAP_ERROR;
}
#endif /* HAVE_LIBNL */

static int pcap_protocol(pcap_t *handle)
{
  int protocol;

  protocol = handle->opt.protocol;
  if (protocol == 0)
    protocol = ETH_P_ALL;

  return htons(protocol);
}

static int
pcap_can_set_rfmon_linux(pcap_t *handle)
{
#ifdef HAVE_LIBNL
  char phydev_path[PATH_MAX+1];
  int ret;
#endif

  if (strcmp(handle->opt.device, "any") == 0) {
    /*
     * Monitor mode makes no sense on the "any" device.
     */
    return 0;
  }

#ifdef HAVE_LIBNL
  /*
   * Bleah.  There doesn't seem to be a way to ask a mac80211
   * device, through libnl, whether it supports monitor mode;
   * we'll just check whether the device appears to be a
   * mac80211 device and, if so, assume the device supports
   * monitor mode.
   */
  ret = get_mac80211_phydev(handle, handle->opt.device, phydev_path,
      PATH_MAX);
  if (ret < 0)
    return ret; /* error */
  if (ret == 1)
    return 1; /* mac80211 device */
#endif

  return 0;
}

/*
 * Grabs the number of missed packets by the interface from
 * /sys/class/net/{if_name}/statistics/rx_{missed,fifo}_errors.
 *
 * Compared to /proc/net/dev this avoids counting software drops,
 * but may be unimplemented and just return 0.
 * The author has found no straightforward way to check for support.
 */
static long long int
linux_get_stat(const char * if_name, const char * stat_name) {
  ssize_t bytes_read;
  int fd;
  char buffer[PATH_MAX];

  snprintf(buffer, sizeof(buffer), "/sys/class/net/%s/statistics/%s", if_name, stat_name);
  fd = open(buffer, O_RDONLY);
  if (fd == -1)
    return 0;

  bytes_read = read(fd, buffer, sizeof(buffer) - 1);
  close(fd);
  if (bytes_read == -1)
    return 0;
  buffer[bytes_read] = '\0';

  return strtoll(buffer, NULL, 10);
}

static long long int
linux_if_drops(const char * if_name)
{
  long long int dropped = linux_get_stat(if_name, "rx_dropped");
  long long int missed = linux_get_stat(if_name, "rx_missed_errors");
  long long int fifo = linux_get_stat(if_name, "rx_fifo_errors");
  return dropped + missed + fifo;
}


/*
 * Monitor mode is kind of interesting because we have to reset the
 * interface before exiting. The problem can't really be solved without
 * some daemon taking care of managing usage counts.  If we put the
 * interface into monitor mode, we set a flag indicating that we must
 * take it out of that mode when the interface is closed, and, when
 * closing the interface, if that flag is set we take it out of monitor
 * mode.
 */

static void pcap_cleanup_linux( pcap_t *handle )
{
  struct pcap_linux *handlep = handle->priv;
#ifdef HAVE_LIBNL
  struct nl80211_state nlstate;
  int ret;
#endif /* HAVE_LIBNL */

  if (handlep->must_do_on_close != 0) {
    /*
     * There's something we have to do when closing this
     * pcap_t.
     */
#ifdef HAVE_LIBNL
    if (handlep->must_do_on_close & MUST_DELETE_MONIF) {
      ret = nl80211_init(handle, &nlstate, handlep->device);
      if (ret >= 0) {
        ret = del_mon_if(handle, handle->fd, &nlstate,
            handlep->device, handlep->mondevice);
        nl80211_cleanup(&nlstate);
      }
      if (ret < 0) {
        fprintf(stderr,
            "Can't delete monitor interface %s (%s).\n"
            "Please delete manually.\n",
            handlep->mondevice, handle->errbuf);
      }
    }
#endif /* HAVE_LIBNL */

    /*
     * Take this pcap out of the list of pcaps for which we
     * have to take the interface out of some mode.
     */
    pcapint_remove_from_pcaps_to_close(handle);
  }

  if (handle->fd != -1) {
    /*
     * Destroy the ring buffer (assuming we've set it up),
     * and unmap it if it's mapped.
     */
    destroy_ring(handle);
  }

  if (handlep->oneshot_buffer != NULL) {
    munmap(handlep->oneshot_buffer, handle->snapshot);
    handlep->oneshot_buffer = NULL;
  }

  if (handlep->mondevice != NULL) {
    free(handlep->mondevice);
    handlep->mondevice = NULL;
  }
  if (handlep->device != NULL) {
    free(handlep->device);
    handlep->device = NULL;
  }

  if (handlep->poll_breakloop_fd != -1) {
    close(handlep->poll_breakloop_fd);
    handlep->poll_breakloop_fd = -1;
  }
  pcapint_cleanup_live_common(handle);
}

#ifdef HAVE_TPACKET3
/*
 * Some versions of TPACKET_V3 have annoying bugs/misfeatures
 * around which we have to work.  Determine if we have those
 * problems or not.
 * 3.19 is the first release with a fixed version of
 * TPACKET_V3.  We treat anything before that as
 * not having a fixed version; that may really mean
 * it has *no* version.
 */
static int has_broken_tpacket_v3(void)
{
  struct utsname utsname;
  const char *release;
  long major, minor;
  int matches, verlen;

  /* No version information, assume broken. */
  if (uname(&utsname) == -1)
    return 1;
  release = utsname.release;

  /* A malformed version, ditto. */
  matches = sscanf(release, "%ld.%ld%n", &major, &minor, &verlen);
  if (matches != 2)
    return 1;
  if (release[verlen] != '.' && release[verlen] != '\0')
    return 1;

  /* OK, a fixed version. */
  if (major > 3 || (major == 3 && minor >= 19))
    return 0;

  /* Too old :( */
  return 1;
}
#endif

/*
 * Set the timeout to be used in poll() with memory-mapped packet capture.
 */
static void
set_poll_timeout(struct pcap_linux *handlep)
{
#ifdef HAVE_TPACKET3
  int broken_tpacket_v3 = has_broken_tpacket_v3();
#endif
  if (handlep->timeout == 0) {
#ifdef HAVE_TPACKET3
    /*
     * XXX - due to a set of (mis)features in the TPACKET_V3
     * kernel code prior to the 3.19 kernel, blocking forever
     * with a TPACKET_V3 socket can, if few packets are
     * arriving and passing the socket filter, cause most
     * packets to be dropped.  See libpcap issue #335 for the
     * full painful story.
     *
     * The workaround is to have poll() time out very quickly,
     * so we grab the frames handed to us, and return them to
     * the kernel, ASAP.
     */
    if (handlep->tp_version == TPACKET_V3 && broken_tpacket_v3)
      handlep->poll_timeout = 1; /* don't block for very long */
    else
#endif
      handlep->poll_timeout = -1; /* block forever */
  } else if (handlep->timeout > 0) {
#ifdef HAVE_TPACKET3
    /*
     * For TPACKET_V3, the timeout is handled by the kernel,
     * so block forever; that way, we don't get extra timeouts.
     * Don't do that if we have a broken TPACKET_V3, though.
     */
    if (handlep->tp_version == TPACKET_V3 && !broken_tpacket_v3)
      handlep->poll_timeout = -1; /* block forever, let TPACKET_V3 wake us up */
    else
#endif
      handlep->poll_timeout = handlep->timeout; /* block for that amount of time */
  } else {
    /*
     * Non-blocking mode; we call poll() to pick up error
     * indications, but we don't want it to wait for
     * anything.
     */
    handlep->poll_timeout = 0;
  }
}

static void pcap_breakloop_linux(pcap_t *handle)
{
  pcapint_breakloop_common(handle);
  struct pcap_linux *handlep = handle->priv;

  uint64_t value = 1;

  if (handlep->poll_breakloop_fd != -1) {
    /*
     * XXX - pcap_breakloop() doesn't have a return value,
     * so we can't indicate an error.
     */
DIAG_OFF_WARN_UNUSED_RESULT
    (void)write(handlep->poll_breakloop_fd, &value, sizeof(value));
DIAG_ON_WARN_UNUSED_RESULT
  }
}

/*
 * Set the offset at which to insert VLAN tags.
 * That should be the offset of the type field.
 */
static void
set_vlan_offset(pcap_t *handle)
{
  struct pcap_linux *handlep = handle->priv;

  switch (handle->linktype) {

  case DLT_EN10MB:
    /*
     * The type field is after the destination and source
     * MAC address.
     */
    handlep->vlan_offset = 2 * ETH_ALEN;
    break;

  case DLT_LINUX_SLL:
    /*
     * The type field is in the last 2 bytes of the
     * DLT_LINUX_SLL header.
     */
    handlep->vlan_offset = SLL_HDR_LEN - 2;
    break;

  default:
    handlep->vlan_offset = -1; /* unknown */
    break;
  }
}

static int
pcap_activate_linux(pcap_t *handle)
{
  struct pcap_linux *handlep = handle->priv;
  const char  *device;
  int   is_any_device;
  struct ifreq  ifr;
  int   status;
  int   ret;

  device = handle->opt.device;

  /*
   * Start out assuming no warnings.
   */
  status = 0;

  /*
   * Make sure the name we were handed will fit into the ioctls we
   * might perform on the device; if not, return a "No such device"
   * indication, as the Linux kernel shouldn't support creating
   * a device whose name won't fit into those ioctls.
   *
   * "Will fit" means "will fit, complete with a null terminator",
   * so if the length, which does *not* include the null terminator,
   * is greater than *or equal to* the size of the field into which
   * we'll be copying it, that won't fit.
   */
  if (strlen(device) >= sizeof(ifr.ifr_name)) {
    /*
     * There's nothing more to say, so clear the error
     * message.
     */
    handle->errbuf[0] = '\0';
    status = PCAP_ERROR_NO_SUCH_DEVICE;
    goto fail;
  }

  /*
   * Turn a negative snapshot value (invalid), a snapshot value of
   * 0 (unspecified), or a value bigger than the normal maximum
   * value, into the maximum allowed value.
   *
   * If some application really *needs* a bigger snapshot
   * length, we should just increase MAXIMUM_SNAPLEN.
   */
  if (handle->snapshot <= 0 || handle->snapshot > MAXIMUM_SNAPLEN)
    handle->snapshot = MAXIMUM_SNAPLEN;

  handlep->device = strdup(device);
  if (handlep->device == NULL) {
    pcapint_fmt_errmsg_for_errno(handle->errbuf, PCAP_ERRBUF_SIZE,
        errno, "strdup");
    status = PCAP_ERROR;
    goto fail;
  }

  /*
   * The "any" device is a special device which causes us not
   * to bind to a particular device and thus to look at all
   * devices.
   */
  is_any_device = (strcmp(device, "any") == 0);
  if (is_any_device) {
    if (handle->opt.promisc) {
      handle->opt.promisc = 0;
      /* Just a warning. */
      snprintf(handle->errbuf, PCAP_ERRBUF_SIZE,
          "Promiscuous mode not supported on the \"any\" device");
      status = PCAP_WARNING_PROMISC_NOTSUP;
    }
  }

  /* copy timeout value */
  handlep->timeout = handle->opt.timeout;

  /*
   * If we're in promiscuous mode, then we probably want
   * to see when the interface drops packets too, so get an
   * initial count from
   * /sys/class/net/{if_name}/statistics/rx_{missed,fifo}_errors
   */
  if (handle->opt.promisc)
    handlep->sysfs_dropped = linux_if_drops(handlep->device);

  /*
   * If the "any" device is specified, try to open a SOCK_DGRAM.
   * Otherwise, open a SOCK_RAW.
   */
  ret = setup_socket(handle, is_any_device);
  if (ret < 0) {
    /*
     * Fatal error; the return value is the error code,
     * and handle->errbuf has been set to an appropriate
     * error message.
     */
    status = ret;
    goto fail;
  }
  if (ret > 0) {
    /*
     * We got a warning; return that, as handle->errbuf
     * might have been overwritten by this warning.
     */
    status = ret;
  }

  /*
   * Success (possibly with a warning).
   *
   * First, try to allocate an event FD for breakloop, if
   * we're not going to start in non-blocking mode.
   */
  if (!handle->opt.nonblock) {
    handlep->poll_breakloop_fd = eventfd(0, EFD_NONBLOCK);
    if (handlep->poll_breakloop_fd == -1) {
      /*
       * Failed.
       */
      pcapint_fmt_errmsg_for_errno(handle->errbuf,
          PCAP_ERRBUF_SIZE, errno, "could not open eventfd");
      status = PCAP_ERROR;
      goto fail;
    }
  }

  /*
   * Succeeded.
   * Try to set up memory-mapped access.
   */
  ret = setup_mmapped(handle);
  if (ret < 0) {
    /*
     * We failed to set up to use it, or the
     * kernel supports it, but we failed to
     * enable it.  The return value is the
     * error status to return and, if it's
     * PCAP_ERROR, handle->errbuf contains
     * the error message.
     */
    status = ret;
    goto fail;
  }
  if (ret > 0) {
    /*
     * We got a warning; return that, as handle->errbuf
     * might have been overwritten by this warning.
     */
    status = ret;
  }

  /*
   * We succeeded.  status has been set to the status to return,
   * which might be 0, or might be a PCAP_WARNING_ value.
   */
  /*
   * Now that we have activated the mmap ring, we can
   * set the correct protocol.
   */
  if ((ret = iface_bind(handle->fd, handlep->ifindex,
      handle->errbuf, pcap_protocol(handle))) != 0) {
    status = ret;
    goto fail;
  }

  handle->inject_op = pcap_inject_linux;
  handle->setfilter_op = pcap_setfilter_linux;
  handle->setdirection_op = pcap_setdirection_linux;
  handle->set_datalink_op = pcap_set_datalink_linux;
  handle->setnonblock_op = pcap_setnonblock_linux;
  handle->getnonblock_op = pcap_getnonblock_linux;
  handle->cleanup_op = pcap_cleanup_linux;
  handle->stats_op = pcap_stats_linux;
  handle->breakloop_op = pcap_breakloop_linux;

  switch (handlep->tp_version) {

  case TPACKET_V2:
    handle->read_op = pcap_read_linux_mmap_v2;
    break;
#ifdef HAVE_TPACKET3
  case TPACKET_V3:
    handle->read_op = pcap_read_linux_mmap_v3;
    break;
#endif
  }
  handle->oneshot_callback = pcapint_oneshot_linux;
  handle->selectable_fd = handle->fd;

  return status;

fail:
  pcap_cleanup_linux(handle);
  return status;
}

static int
pcap_set_datalink_linux(pcap_t *handle, int dlt)
{
  handle->linktype = dlt;

  /*
   * Update the offset at which to insert VLAN tags for the
   * new link-layer type.
   */
  set_vlan_offset(handle);

  return 0;
}

/*
 * linux_check_direction()
 *
 * Do checks based on packet direction.
 */
static inline int
linux_check_direction(const pcap_t *handle, const struct sockaddr_ll *sll)
{
  struct pcap_linux *handlep = handle->priv;

  if (sll->sll_pkttype == PACKET_OUTGOING) {
    /*
     * Outgoing packet.
     * If this is from the loopback device, reject it;
     * we'll see the packet as an incoming packet as well,
     * and we don't want to see it twice.
     */
    if (sll->sll_ifindex == handlep->lo_ifindex)
      return 0;

    /*
     * If this is an outgoing CAN frame, and the user doesn't
     * want only outgoing packets, reject it; CAN devices
     * and drivers, and the CAN stack, always arrange to
     * loop back transmitted packets, so they also appear
     * as incoming packets.  We don't want duplicate packets,
     * and we can't easily distinguish packets looped back
     * by the CAN layer than those received by the CAN layer,
     * so we eliminate this packet instead.
     *
     * We check whether this is a CAN frame by checking whether
     * the device's hardware type is ARPHRD_CAN.
     */
    if (sll->sll_hatype == ARPHRD_CAN &&
         handle->direction != PCAP_D_OUT)
      return 0;

    /*
     * If the user only wants incoming packets, reject it.
     */
    if (handle->direction == PCAP_D_IN)
      return 0;
  } else {
    /*
     * Incoming packet.
     * If the user only wants outgoing packets, reject it.
     */
    if (handle->direction == PCAP_D_OUT)
      return 0;
  }
  return 1;
}

/*
 * Check whether the device to which the pcap_t is bound still exists.
 * We do so by asking what address the socket is bound to, and checking
 * whether the ifindex in the address is -1, meaning "that device is gone",
 * or some other value, meaning "that device still exists".
 */
static int
device_still_exists(pcap_t *handle)
{
  struct pcap_linux *handlep = handle->priv;
  struct sockaddr_ll addr;
  socklen_t addr_len;

  /*
   * If handlep->ifindex is -1, the socket isn't bound, meaning
   * we're capturing on the "any" device; that device never
   * disappears.  (It should also never be configured down, so
   * we shouldn't even get here, but let's make sure.)
   */
  if (handlep->ifindex == -1)
    return (1); /* it's still here */

  /*
   * OK, now try to get the address for the socket.
   */
  addr_len = sizeof (addr);
  if (getsockname(handle->fd, (struct sockaddr *) &addr, &addr_len) == -1) {
    /*
     * Error - report an error and return -1.
     */
    pcapint_fmt_errmsg_for_errno(handle->errbuf, PCAP_ERRBUF_SIZE,
        errno, "getsockname failed");
    return (-1);
  }
  if (addr.sll_ifindex == -1) {
    /*
     * This means the device went away.
     */
    return (0);
  }

  /*
   * The device presumably just went down.
   */
  return (1);
}

static int
pcap_inject_linux(pcap_t *handle, const void *buf, int size)
{
  struct pcap_linux *handlep = handle->priv;
  int ret;

  if (handlep->ifindex == -1) {
    /*
     * We don't support sending on the "any" device.
     */
    pcapint_strlcpy(handle->errbuf,
        "Sending packets isn't supported on the \"any\" device",
        PCAP_ERRBUF_SIZE);
    return (-1);
  }

  if (handlep->cooked) {
    /*
     * We don't support sending on cooked-mode sockets.
     *
     * XXX - how do you send on a bound cooked-mode
     * socket?
     * Is a "sendto()" required there?
     */
    pcapint_strlcpy(handle->errbuf,
        "Sending packets isn't supported in cooked mode",
        PCAP_ERRBUF_SIZE);
    return (-1);
  }

  ret = (int)send(handle->fd, buf, size, 0);
  if (ret == -1) {
    pcapint_fmt_errmsg_for_errno(handle->errbuf, PCAP_ERRBUF_SIZE,
        errno, "send");
    return (-1);
  }
  return (ret);
}

/*
 *  Get the statistics for the given packet capture handle.
 */
static int
pcap_stats_linux(pcap_t *handle, struct pcap_stat *stats)
{
  struct pcap_linux *handlep = handle->priv;
#ifdef HAVE_TPACKET3
  /*
   * For sockets using TPACKET_V2, the extra stuff at the end
   * of a struct tpacket_stats_v3 will not be filled in, and
   * we don't look at it so this is OK even for those sockets.
   * In addition, the PF_PACKET socket code in the kernel only
   * uses the length parameter to compute how much data to
   * copy out and to indicate how much data was copied out, so
   * it's OK to base it on the size of a struct tpacket_stats.
   *
   * XXX - it's probably OK, in fact, to just use a
   * struct tpacket_stats for V3 sockets, as we don't
   * care about the tp_freeze_q_cnt stat.
   */
  struct tpacket_stats_v3 kstats;
#else /* HAVE_TPACKET3 */
  struct tpacket_stats kstats;
#endif /* HAVE_TPACKET3 */
  socklen_t len = sizeof (struct tpacket_stats);

  long long if_dropped = 0;

  /*
   * To fill in ps_ifdrop, we parse
   * /sys/class/net/{if_name}/statistics/rx_{missed,fifo}_errors
   * for the numbers
   */
  if (handle->opt.promisc)
  {
    /*
     * XXX - is there any reason to do this by remembering
     * the last counts value, subtracting it from the
     * current counts value, and adding that to stat.ps_ifdrop,
     * maintaining stat.ps_ifdrop as a count, rather than just
     * saving the *initial* counts value and setting
     * stat.ps_ifdrop to the difference between the current
     * value and the initial value?
     *
     * One reason might be to handle the count wrapping
     * around, on platforms where the count is 32 bits
     * and where you might get more than 2^32 dropped
     * packets; is there any other reason?
     *
     * (We maintain the count as a long long int so that,
     * if the kernel maintains the counts as 64-bit even
     * on 32-bit platforms, we can handle the real count.
     *
     * Unfortunately, we can't report 64-bit counts; we
     * need a better API for reporting statistics, such as
     * one that reports them in a style similar to the
     * pcapng Interface Statistics Block, so that 1) the
     * counts are 64-bit, 2) it's easier to add new statistics
     * without breaking the ABI, and 3) it's easier to
     * indicate to a caller that wants one particular
     * statistic that it's not available by just not supplying
     * it.)
     */
    if_dropped = handlep->sysfs_dropped;
    handlep->sysfs_dropped = linux_if_drops(handlep->device);
    handlep->stat.ps_ifdrop += (u_int)(handlep->sysfs_dropped - if_dropped);
  }

  /*
   * Try to get the packet counts from the kernel.
   */
  if (getsockopt(handle->fd, SOL_PACKET, PACKET_STATISTICS,
      &kstats, &len) > -1) {
    /*
     * "ps_recv" counts only packets that *passed* the
     * filter, not packets that didn't pass the filter.
     * This includes packets later dropped because we
     * ran out of buffer space.
     *
     * "ps_drop" counts packets dropped because we ran
     * out of buffer space.  It doesn't count packets
     * dropped by the interface driver.  It counts only
     * packets that passed the filter.
     *
     * See above for ps_ifdrop.
     *
     * Both statistics include packets not yet read from
     * the kernel by libpcap, and thus not yet seen by
     * the application.
     *
     * In "linux/net/packet/af_packet.c", at least in 2.6.27
     * through 5.6 kernels, "tp_packets" is incremented for
     * every packet that passes the packet filter *and* is
     * successfully copied to the ring buffer; "tp_drops" is
     * incremented for every packet dropped because there's
     * not enough free space in the ring buffer.
     *
     * When the statistics are returned for a PACKET_STATISTICS
     * "getsockopt()" call, "tp_drops" is added to "tp_packets",
     * so that "tp_packets" counts all packets handed to
     * the PF_PACKET socket, including packets dropped because
     * there wasn't room on the socket buffer - but not
     * including packets that didn't pass the filter.
     *
     * In the BSD BPF, the count of received packets is
     * incremented for every packet handed to BPF, regardless
     * of whether it passed the filter.
     *
     * We can't make "pcap_stats()" work the same on both
     * platforms, but the best approximation is to return
     * "tp_packets" as the count of packets and "tp_drops"
     * as the count of drops.
     *
     * Keep a running total because each call to
     *    getsockopt(handle->fd, SOL_PACKET, PACKET_STATISTICS, ....
     * resets the counters to zero.
     */
    handlep->stat.ps_recv += kstats.tp_packets;
    handlep->stat.ps_drop += kstats.tp_drops;
    *stats = handlep->stat;
    return 0;
  }

  pcapint_fmt_errmsg_for_errno(handle->errbuf, PCAP_ERRBUF_SIZE, errno,
      "failed to get statistics from socket");
  return -1;
}

/*
 * A PF_PACKET socket can be bound to any network interface.
 */
static int
can_be_bound(const char *name _U_)
{
  return (1);
}

/*
 * Get a socket to use with various interface ioctls.
 */
static int
get_if_ioctl_socket(void)
{
  int fd;

  /*
   * This is a bit ugly.
   *
   * There isn't a socket type that's guaranteed to work.
   *
   * AF_NETLINK will work *if* you have Netlink configured into the
   * kernel (can it be configured out if you have any networking
   * support at all?) *and* if you're running a sufficiently recent
   * kernel, but not all the kernels we support are sufficiently
   * recent - that feature was introduced in Linux 4.6.
   *
   * AF_UNIX will work *if* you have UNIX-domain sockets configured
   * into the kernel and *if* you're not on a system that doesn't
   * allow them - some SELinux systems don't allow you create them.
   * Most systems probably have them configured in, but not all systems
   * have them configured in and allow them to be created.
   *
   * AF_INET will work *if* you have IPv4 configured into the kernel,
   * but, apparently, some systems have network adapters but have
   * kernels without IPv4 support.
   *
   * AF_INET6 will work *if* you have IPv6 configured into the
   * kernel, but if you don't have AF_INET, you might not have
   * AF_INET6, either (that is, independently on its own grounds).
   *
   * AF_PACKET would work, except that some of these calls should
   * work even if you *don't* have capture permission (you should be
   * able to enumerate interfaces and get information about them
   * without capture permission; you shouldn't get a failure until
   * you try pcap_activate()).  (If you don't allow programs to
   * get as much information as possible about interfaces if you
   * don't have permission to capture, you run the risk of users
   * asking "why isn't it showing XXX" - or, worse, if you don't
   * show interfaces *at all* if you don't have permission to
   * capture on them, "why do no interfaces show up?" - when the
   * real problem is a permissions problem.  Error reports of that
   * type require a lot more back-and-forth to debug, as evidenced
   * by many Wireshark bugs/mailing list questions/Q&A questions.)
   *
   * So:
   *
   * we first try an AF_NETLINK socket, where "try" includes
   * "try to do a device ioctl on it", as, in the future, once
   * pre-4.6 kernels are sufficiently rare, that will probably
   * be the mechanism most likely to work;
   *
   * if that fails, we try an AF_UNIX socket, as that's less
   * likely to be configured out on a networking-capable system
   * than is IP;
   *
   * if that fails, we try an AF_INET6 socket;
   *
   * if that fails, we try an AF_INET socket.
   */
  fd = socket(AF_NETLINK, SOCK_RAW, NETLINK_GENERIC);
  if (fd != -1) {
    /*
     * OK, let's make sure we can do an SIOCGIFNAME
     * ioctl.
     */
    struct ifreq ifr;

    memset(&ifr, 0, sizeof(ifr));
    if (ioctl(fd, SIOCGIFNAME, &ifr) == 0 ||
        errno != EOPNOTSUPP) {
      /*
       * It succeeded, or failed for some reason
       * other than "netlink sockets don't support
       * device ioctls".  Go with the AF_NETLINK
       * socket.
       */
      return (fd);
    }

    /*
     * OK, that didn't work, so it's as bad as "netlink
     * sockets aren't available".  Close the socket and
     * drive on.
     */
    close(fd);
  }

  /*
   * Now try an AF_UNIX socket.
   */
  fd = socket(AF_UNIX, SOCK_RAW, 0);
  if (fd != -1) {
    /*
     * OK, we got it!
     */
    return (fd);
  }

  /*
   * Now try an AF_INET6 socket.
   */
  fd = socket(AF_INET6, SOCK_DGRAM, 0);
  if (fd != -1) {
    return (fd);
  }

  /*
   * Now try an AF_INET socket.
   *
   * XXX - if that fails, is there anything else we should try?
   * AF_CAN, for embedded systems in vehicles, in case they're
   * built without Internet protocol support?  Any other socket
   * types popular in non-Internet embedded systems?
   */
  return (socket(AF_INET, SOCK_DGRAM, 0));
}

/*
 * Get additional flags for a device, using SIOCETHTOOL.
 */
static int
get_if_flags(const char *name, bpf_u_int32 *flags, char *errbuf)
{
  int sock;
  FILE *fh;
  unsigned int arptype = ARPHRD_VOID;
  struct ifreq ifr;
  struct ethtool_value info;

  if (*flags & PCAP_IF_LOOPBACK) {
    /*
     * Loopback devices aren't wireless, and "connected"/
     * "disconnected" doesn't apply to them.
     */
    *flags |= PCAP_IF_CONNECTION_STATUS_NOT_APPLICABLE;
    return 0;
  }

  sock = get_if_ioctl_socket();
  if (sock == -1) {
    pcapint_fmt_errmsg_for_errno(errbuf, PCAP_ERRBUF_SIZE, errno,
        "Can't create socket to get ethtool information for %s",
        name);
    return -1;
  }

  /*
   * OK, what type of network is this?
   * In particular, is it wired or wireless?
   */
  if (is_wifi(name)) {
    /*
     * Wi-Fi, hence wireless.
     */
    *flags |= PCAP_IF_WIRELESS;
  } else {
    /*
     * OK, what does /sys/class/net/{if_name}/type contain?
     * (We don't use that for Wi-Fi, as it'll report
     * "Ethernet", i.e. ARPHRD_ETHER, for non-monitor-
     * mode devices.)
     */
    char *pathstr;

    if (asprintf(&pathstr, "/sys/class/net/%s/type", name) == -1) {
      snprintf(errbuf, PCAP_ERRBUF_SIZE,
          "%s: Can't generate path name string for /sys/class/net device",
          name);
      close(sock);
      return -1;
    }
    fh = fopen(pathstr, "r");
    if (fh != NULL) {
      if (fscanf(fh, "%u", &arptype) == 1) {
        /*
         * OK, we got an ARPHRD_ type; what is it?
         */
        switch (arptype) {

        case ARPHRD_LOOPBACK:
          /*
           * These are types to which
           * "connected" and "disconnected"
           * don't apply, so don't bother
           * asking about it.
           *
           * XXX - add other types?
           */
          close(sock);
          fclose(fh);
          free(pathstr);
          return 0;

        case ARPHRD_IRDA:
        case ARPHRD_IEEE80211:
        case ARPHRD_IEEE80211_PRISM:
        case ARPHRD_IEEE80211_RADIOTAP:
        case ARPHRD_IEEE802154:
        case ARPHRD_IEEE802154_MONITOR:
        case ARPHRD_6LOWPAN:
          /*
           * Various wireless types.
           */
          *flags |= PCAP_IF_WIRELESS;
          break;
        }
      }
      fclose(fh);
    }
    free(pathstr);
  }

#ifdef ETHTOOL_GLINK
  memset(&ifr, 0, sizeof(ifr));
  pcapint_strlcpy(ifr.ifr_name, name, sizeof(ifr.ifr_name));
  info.cmd = ETHTOOL_GLINK;
  /*
   * XXX - while Valgrind handles SIOCETHTOOL and knows that
   * the ETHTOOL_GLINK command sets the .data member of the
   * structure, Memory Sanitizer doesn't yet do so:
   *
   *    https://bugs.llvm.org/show_bug.cgi?id=45814
   *
   * For now, we zero it out to squelch warnings; if the bug
   * in question is fixed, we can remove this.
   */
  info.data = 0;
  ifr.ifr_data = (caddr_t)&info;
  if (ioctl(sock, SIOCETHTOOL, &ifr) == -1) {
    int save_errno = errno;

    switch (save_errno) {

    case EOPNOTSUPP:
    case EINVAL:
      /*
       * OK, this OS version or driver doesn't support
       * asking for this information.
       * XXX - distinguish between "this doesn't
       * support ethtool at all because it's not
       * that type of device" vs. "this doesn't
       * support ethtool even though it's that
       * type of device", and return "unknown".
       */
      *flags |= PCAP_IF_CONNECTION_STATUS_NOT_APPLICABLE;
      close(sock);
      return 0;

    case ENODEV:
      /*
       * OK, no such device.
       * The user will find that out when they try to
       * activate the device; just say "OK" and
       * don't set anything.
       */
      close(sock);
      return 0;

    case EPERM:
      /*
       * OK, this version of the kernel requires
       * CAP_NET_ADMIN privileges for this, and
       * we don't have those privileges.
       *
       * Just leave it as PCAP_IF_CONNECTION_STATUS_UNKNOWN,
       * as we can't determine the status; just say "OK" and
       * don't set anything.
       */
      close(sock);
      return 0;

    default:
      /*
       * Other error.
       */
      pcapint_fmt_errmsg_for_errno(errbuf, PCAP_ERRBUF_SIZE,
          save_errno,
          "%s: SIOCETHTOOL(ETHTOOL_GLINK) ioctl failed",
          name);
      close(sock);
      return -1;
    }
  }

  /*
   * Is it connected?
   */
  if (info.data) {
    /*
     * It's connected.
     */
    *flags |= PCAP_IF_CONNECTION_STATUS_CONNECTED;
  } else {
    /*
     * It's disconnected.
     */
    *flags |= PCAP_IF_CONNECTION_STATUS_DISCONNECTED;
  }
#endif

  close(sock);

#ifdef HAVE_SNF_API
  // For "down" SNF devices the SNF API makes the flags more relevant.
  if (arptype == ARPHRD_ETHER &&
      ! (*flags & PCAP_IF_UP) &&
      snf_get_if_flags(name, flags, errbuf) < 0)
    return PCAP_ERROR;
#endif // HAVE_SNF_API

  return 0;
}

int
pcapint_platform_finddevs(pcap_if_list_t *devlistp, char *errbuf)
{
  /*
   * Get the list of regular interfaces first.
   */
  if (pcapint_findalldevs_interfaces(devlistp, errbuf, can_be_bound,
      get_if_flags) == -1)
    return (-1); /* failure */

  /*
   * Add the "any" device.
   */
  if (pcapint_add_any_dev(devlistp, errbuf) == NULL)
    return (-1);

  return (0);
}

/*
 * Set direction flag: Which packets do we accept on a forwarding
 * single device? IN, OUT or both?
 */
static int
pcap_setdirection_linux(pcap_t *handle, pcap_direction_t d)
{
  /*
   * It's guaranteed, at this point, that d is a valid
   * direction value.
   */
  handle->direction = d;
  return 0;
}

static int
is_wifi(const char *device)
{
  char *pathstr;
  struct stat statb;

  /*
   * See if there's a sysfs wireless directory for it.
   * If so, it's a wireless interface.
   */
  if (asprintf(&pathstr, "/sys/class/net/%s/wireless", device) == -1) {
    /*
     * Just give up here.
     */
    return 0;
  }
  if (stat(pathstr, &statb) == 0) {
    free(pathstr);
    return 1;
  }
  free(pathstr);

  return 0;
}

/*
 *  Linux uses the ARP hardware type to identify the type of an
 *  interface. pcap uses the DLT_xxx constants for this. This
 *  function takes a pointer to a "pcap_t", and an ARPHRD_xxx
 *  constant, as arguments, and sets "handle->linktype" to the
 *  appropriate DLT_XXX constant and sets "handle->offset" to
 *  the appropriate value (to make "handle->offset" plus link-layer
 *  header length be a multiple of 4, so that the link-layer payload
 *  will be aligned on a 4-byte boundary when capturing packets).
 *  (If the offset isn't set here, it'll be 0; add code as appropriate
 *  for cases where it shouldn't be 0.)
 *
 *  If "cooked_ok" is non-zero, we can use DLT_LINUX_SLL and capture
 *  in cooked mode; otherwise, we can't use cooked mode, so we have
 *  to pick some type that works in raw mode, or fail.
 *
 *  Sets the link type to -1 if unable to map the type.
 *
 *  Returns 0 on success or a PCAP_ERROR_ value on error.
 */
static int map_arphrd_to_dlt(pcap_t *handle, int arptype,
           const char *device, int cooked_ok)
{
  static const char cdma_rmnet[] = "cdma_rmnet";

  switch (arptype) {

  case ARPHRD_ETHER:
    /*
     * For various annoying reasons having to do with DHCP
     * software, some versions of Android give the mobile-
     * phone-network interface an ARPHRD_ value of
     * ARPHRD_ETHER, even though the packets supplied by
     * that interface have no link-layer header, and begin
     * with an IP header, so that the ARPHRD_ value should
     * be ARPHRD_NONE.
     *
     * Detect those devices by checking the device name, and
     * use DLT_RAW for them.
     */
    if (strncmp(device, cdma_rmnet, sizeof cdma_rmnet - 1) == 0) {
      handle->linktype = DLT_RAW;
      return 0;
    }

    /*
     * Is this a real Ethernet device?  If so, give it a
     * link-layer-type list with DLT_EN10MB and DLT_DOCSIS, so
     * that an application can let you choose it, in case you're
     * capturing DOCSIS traffic that a Cisco Cable Modem
     * Termination System is putting out onto an Ethernet (it
     * doesn't put an Ethernet header onto the wire, it puts raw
     * DOCSIS frames out on the wire inside the low-level
     * Ethernet framing).
     *
     * XXX - are there any other sorts of "fake Ethernet" that
     * have ARPHRD_ETHER but that shouldn't offer DLT_DOCSIS as
     * a Cisco CMTS won't put traffic onto it or get traffic
     * bridged onto it?  ISDN is handled in "setup_socket()",
     * as we fall back on cooked mode there, and we use
     * is_wifi() to check for 802.11 devices; are there any
     * others?
     */
    if (!is_wifi(device)) {
      int ret;

      /*
       * This is not a Wi-Fi device but it could be
       * a DSA master/management network device.
       */
      ret = iface_dsa_get_proto_info(device, handle);
      if (ret < 0)
        return ret;

      if (ret == 1) {
        /*
         * This is a DSA master/management network
         * device, linktype is already set by
         * iface_dsa_get_proto_info(), set an
         * appropriate offset here.
         */
        handle->offset = 2;
        break;
      }

      /*
       * It's not a Wi-Fi device; offer DOCSIS.
       */
      handle->dlt_list = (u_int *) malloc(sizeof(u_int) * 2);
      if (handle->dlt_list == NULL) {
        pcapint_fmt_errmsg_for_errno(handle->errbuf,
            PCAP_ERRBUF_SIZE, errno, "malloc");
        return (PCAP_ERROR);
      }
      handle->dlt_list[0] = DLT_EN10MB;
      handle->dlt_list[1] = DLT_DOCSIS;
      handle->dlt_count = 2;
    }
    /* FALLTHROUGH */

  case ARPHRD_METRICOM:
  case ARPHRD_LOOPBACK:
    handle->linktype = DLT_EN10MB;
    handle->offset = 2;
    break;

  case ARPHRD_EETHER:
    handle->linktype = DLT_EN3MB;
    break;

  case ARPHRD_AX25:
    handle->linktype = DLT_AX25_KISS;
    break;

  case ARPHRD_PRONET:
    handle->linktype = DLT_PRONET;
    break;

  case ARPHRD_CHAOS:
    handle->linktype = DLT_CHAOS;
    break;

  case ARPHRD_CAN:
    handle->linktype = DLT_CAN_SOCKETCAN;
    break;

  case ARPHRD_IEEE802_TR:
  case ARPHRD_IEEE802:
    handle->linktype = DLT_IEEE802;
    handle->offset = 2;
    break;

  case ARPHRD_ARCNET:
    handle->linktype = DLT_ARCNET_LINUX;
    break;

  case ARPHRD_FDDI:
    handle->linktype = DLT_FDDI;
    handle->offset = 3;
    break;

  case ARPHRD_ATM:
    /*
     * The Classical IP implementation in ATM for Linux
     * supports both what RFC 1483 calls "LLC Encapsulation",
     * in which each packet has an LLC header, possibly
     * with a SNAP header as well, prepended to it, and
     * what RFC 1483 calls "VC Based Multiplexing", in which
     * different virtual circuits carry different network
     * layer protocols, and no header is prepended to packets.
     *
     * They both have an ARPHRD_ type of ARPHRD_ATM, so
     * you can't use the ARPHRD_ type to find out whether
     * captured packets will have an LLC header, and,
     * while there's a socket ioctl to *set* the encapsulation
     * type, there's no ioctl to *get* the encapsulation type.
     *
     * This means that
     *
     *  programs that dissect Linux Classical IP frames
     *  would have to check for an LLC header and,
     *  depending on whether they see one or not, dissect
     *  the frame as LLC-encapsulated or as raw IP (I
     *  don't know whether there's any traffic other than
     *  IP that would show up on the socket, or whether
     *  there's any support for IPv6 in the Linux
     *  Classical IP code);
     *
     *  filter expressions would have to compile into
     *  code that checks for an LLC header and does
     *  the right thing.
     *
     * Both of those are a nuisance - and, at least on systems
     * that support PF_PACKET sockets, we don't have to put
     * up with those nuisances; instead, we can just capture
     * in cooked mode.  That's what we'll do, if we can.
     * Otherwise, we'll just fail.
     */
    if (cooked_ok)
      handle->linktype = DLT_LINUX_SLL;
    else
      handle->linktype = -1;
    break;

  case ARPHRD_IEEE80211:
    handle->linktype = DLT_IEEE802_11;
    break;

  case ARPHRD_IEEE80211_PRISM:
    handle->linktype = DLT_PRISM_HEADER;
    break;

  case ARPHRD_IEEE80211_RADIOTAP:
    handle->linktype = DLT_IEEE802_11_RADIO;
    break;

  case ARPHRD_PPP:
    /*
     * Some PPP code in the kernel supplies no link-layer
     * header whatsoever to PF_PACKET sockets; other PPP
     * code supplies PPP link-layer headers ("syncppp.c");
     * some PPP code might supply random link-layer
     * headers (PPP over ISDN - there's code in Wireshark,
     * for example, to cope with PPP-over-ISDN captures
     * with which the Wireshark developers have had to cope,
     * heuristically trying to determine which of the
     * oddball link-layer headers particular packets have).
     *
     * As such, we just punt, and run all PPP interfaces
     * in cooked mode, if we can; otherwise, we just treat
     * it as DLT_RAW, for now - if somebody needs to capture,
     * on a 2.0[.x] kernel, on PPP devices that supply a
     * link-layer header, they'll have to add code here to
     * map to the appropriate DLT_ type (possibly adding a
     * new DLT_ type, if necessary).
     */
    if (cooked_ok)
      handle->linktype = DLT_LINUX_SLL;
    else {
      /*
       * XXX - handle ISDN types here?  We can't fall
       * back on cooked sockets, so we'd have to
       * figure out from the device name what type of
       * link-layer encapsulation it's using, and map
       * that to an appropriate DLT_ value, meaning
       * we'd map "isdnN" devices to DLT_RAW (they
       * supply raw IP packets with no link-layer
       * header) and "isdY" devices to a new DLT_I4L_IP
       * type that has only an Ethernet packet type as
       * a link-layer header.
       *
       * But sometimes we seem to get random crap
       * in the link-layer header when capturing on
       * ISDN devices....
       */
      handle->linktype = DLT_RAW;
    }
    break;

  case ARPHRD_CISCO:
    handle->linktype = DLT_C_HDLC;
    break;

  /* Not sure if this is correct for all tunnels, but it
   * works for CIPE */
  case ARPHRD_TUNNEL:
  case ARPHRD_SIT:
  case ARPHRD_CSLIP:
  case ARPHRD_SLIP6:
  case ARPHRD_CSLIP6:
  case ARPHRD_ADAPT:
  case ARPHRD_SLIP:
  case ARPHRD_RAWHDLC:
  case ARPHRD_DLCI:
    /*
     * XXX - should some of those be mapped to DLT_LINUX_SLL
     * instead?  Should we just map all of them to DLT_LINUX_SLL?
     */
    handle->linktype = DLT_RAW;
    break;

  case ARPHRD_FRAD:
    handle->linktype = DLT_FRELAY;
    break;

  case ARPHRD_LOCALTLK:
    handle->linktype = DLT_LTALK;
    break;

  case 18:
    /*
     * RFC 4338 defines an encapsulation for IP and ARP
     * packets that's compatible with the RFC 2625
     * encapsulation, but that uses a different ARP
     * hardware type and hardware addresses.  That
     * ARP hardware type is 18; Linux doesn't define
     * any ARPHRD_ value as 18, but if it ever officially
     * supports RFC 4338-style IP-over-FC, it should define
     * one.
     *
     * For now, we map it to DLT_IP_OVER_FC, in the hopes
     * that this will encourage its use in the future,
     * should Linux ever officially support RFC 4338-style
     * IP-over-FC.
     */
    handle->linktype = DLT_IP_OVER_FC;
    break;

  case ARPHRD_FCPP:
  case ARPHRD_FCAL:
  case ARPHRD_FCPL:
  case ARPHRD_FCFABRIC:
    /*
     * Back in 2002, Donald Lee at Cray wanted a DLT_ for
     * IP-over-FC:
     *
     *  https://www.mail-archive.com/tcpdump-workers@sandelman.ottawa.on.ca/msg01043.html
     *
     * and one was assigned.
     *
     * In a later private discussion (spun off from a message
     * on the ethereal-users list) on how to get that DLT_
     * value in libpcap on Linux, I ended up deciding that
     * the best thing to do would be to have him tweak the
     * driver to set the ARPHRD_ value to some ARPHRD_FCxx
     * type, and map all those types to DLT_IP_OVER_FC:
     *
     *  I've checked into the libpcap and tcpdump CVS tree
     *  support for DLT_IP_OVER_FC.  In order to use that,
     *  you'd have to modify your modified driver to return
     *  one of the ARPHRD_FCxxx types, in "fcLINUXfcp.c" -
     *  change it to set "dev->type" to ARPHRD_FCFABRIC, for
     *  example (the exact value doesn't matter, it can be
     *  any of ARPHRD_FCPP, ARPHRD_FCAL, ARPHRD_FCPL, or
     *  ARPHRD_FCFABRIC).
     *
     * 11 years later, Christian Svensson wanted to map
     * various ARPHRD_ values to DLT_FC_2 and
     * DLT_FC_2_WITH_FRAME_DELIMS for raw Fibre Channel
     * frames:
     *
     *  https://github.com/mcr/libpcap/pull/29
     *
     * There don't seem to be any network drivers that use
     * any of the ARPHRD_FC* values for IP-over-FC, and
     * it's not exactly clear what the "Dummy types for non
     * ARP hardware" are supposed to mean (link-layer
     * header type?  Physical network type?), so it's
     * not exactly clear why the ARPHRD_FC* types exist
     * in the first place.
     *
     * For now, we map them to DLT_FC_2, and provide an
     * option of DLT_FC_2_WITH_FRAME_DELIMS, as well as
     * DLT_IP_OVER_FC just in case there's some old
     * driver out there that uses one of those types for
     * IP-over-FC on which somebody wants to capture
     * packets.
     */
    handle->linktype = DLT_FC_2;
    handle->dlt_list = (u_int *) malloc(sizeof(u_int) * 3);
    if (handle->dlt_list == NULL) {
      pcapint_fmt_errmsg_for_errno(handle->errbuf,
          PCAP_ERRBUF_SIZE, errno, "malloc");
      return (PCAP_ERROR);
    }
    handle->dlt_list[0] = DLT_FC_2;
    handle->dlt_list[1] = DLT_FC_2_WITH_FRAME_DELIMS;
    handle->dlt_list[2] = DLT_IP_OVER_FC;
    handle->dlt_count = 3;
    break;

  case ARPHRD_IRDA:
    /* Don't expect IP packet out of this interfaces... */
    handle->linktype = DLT_LINUX_IRDA;
    /* We need to save packet direction for IrDA decoding,
     * so let's use "Linux-cooked" mode. Jean II
     *
     * XXX - this is handled in setup_socket(). */
    /* handlep->cooked = 1; */
    break;

  case ARPHRD_LAPD:
    /* Don't expect IP packet out of this interfaces... */
    handle->linktype = DLT_LINUX_LAPD;
    break;

  case ARPHRD_NONE:
    /*
     * No link-layer header; packets are just IP
     * packets, so use DLT_RAW.
     */
    handle->linktype = DLT_RAW;
    break;

  case ARPHRD_IEEE802154:
    handle->linktype = DLT_IEEE802_15_4_NOFCS;
    break;

  case ARPHRD_NETLINK:
    handle->linktype = DLT_NETLINK;
    /*
     * We need to use cooked mode, so that in sll_protocol we
     * pick up the netlink protocol type such as NETLINK_ROUTE,
     * NETLINK_GENERIC, NETLINK_FIB_LOOKUP, etc.
     *
     * XXX - this is handled in setup_socket().
     */
    /* handlep->cooked = 1; */
    break;

  case ARPHRD_VSOCKMON:
    handle->linktype = DLT_VSOCK;
    break;

  default:
    handle->linktype = -1;
    break;
  }
  return (0);
}

/*
 * Try to set up a PF_PACKET socket.
 * Returns 0 or a PCAP_WARNING_ value on success and a PCAP_ERROR_ value
 * on failure.
 */
static int
setup_socket(pcap_t *handle, int is_any_device)
{
  struct pcap_linux *handlep = handle->priv;
  const char    *device = handle->opt.device;
  int     status = 0;
  int     sock_fd, arptype;
  int     val;
  int     err = 0;
  struct packet_mreq  mr;

  /*
   * Open a socket with protocol family packet. If cooked is true,
   * we open a SOCK_DGRAM socket for the cooked interface, otherwise
   * we open a SOCK_RAW socket for the raw interface.
   *
   * The protocol is set to 0.  This means we will receive no
   * packets until we "bind" the socket with a non-zero
   * protocol.  This allows us to setup the ring buffers without
   * dropping any packets.
   */
  sock_fd = is_any_device ?
    socket(PF_PACKET, SOCK_DGRAM, 0) :
    socket(PF_PACKET, SOCK_RAW, 0);

  if (sock_fd == -1) {
    if (errno == EPERM || errno == EACCES) {
      /*
       * You don't have permission to open the
       * socket.
       */
      status = PCAP_ERROR_PERM_DENIED;
      snprintf(handle->errbuf, PCAP_ERRBUF_SIZE,
          "Attempt to create packet socket failed - CAP_NET_RAW may be required");
    } else if (errno == EAFNOSUPPORT) {
      /*
       * PF_PACKET sockets not supported.
       * Perhaps we're running on the WSL1 module
       * in the Windows NT kernel rather than on
       * a real Linux kernel.
       */
      status = PCAP_ERROR_CAPTURE_NOTSUP;
      snprintf(handle->errbuf, PCAP_ERRBUF_SIZE,
          "PF_PACKET sockets not supported - is this WSL1?");
    } else {
      /*
       * Other error.
       */
      status = PCAP_ERROR;
      pcapint_fmt_errmsg_for_errno(handle->errbuf,
          PCAP_ERRBUF_SIZE, errno, "socket");
    }
    return status;
  }

  /*
   * Get the interface index of the loopback device.
   * If the attempt fails, don't fail, just set the
   * "handlep->lo_ifindex" to -1.
   *
   * XXX - can there be more than one device that loops
   * packets back, i.e. devices other than "lo"?  If so,
   * we'd need to find them all, and have an array of
   * indices for them, and check all of them in
   * "pcap_read_packet()".
   */
  handlep->lo_ifindex = iface_get_id(sock_fd, "lo", handle->errbuf);

  /*
   * Default value for offset to align link-layer payload
   * on a 4-byte boundary.
   */
  handle->offset   = 0;

  /*
   * What kind of frames do we have to deal with? Fall back
   * to cooked mode if we have an unknown interface type
   * or a type we know doesn't work well in raw mode.
   */
  if (!is_any_device) {
    /* Assume for now we don't need cooked mode. */
    handlep->cooked = 0;

    if (handle->opt.rfmon) {
      /*
       * We were asked to turn on monitor mode.
       * Do so before we get the link-layer type,
       * because entering monitor mode could change
       * the link-layer type.
       */
      err = enter_rfmon_mode(handle, sock_fd, device);
      if (err < 0) {
        /* Hard failure */
        close(sock_fd);
        return err;
      }
      if (err == 0) {
        /*
         * Nothing worked for turning monitor mode
         * on.
         */
        close(sock_fd);

        return PCAP_ERROR_RFMON_NOTSUP;
      }

      /*
       * Either monitor mode has been turned on for
       * the device, or we've been given a different
       * device to open for monitor mode.  If we've
       * been given a different device, use it.
       */
      if (handlep->mondevice != NULL)
        device = handlep->mondevice;
    }
    arptype = iface_get_arptype(sock_fd, device, handle->errbuf);
    if (arptype < 0) {
      close(sock_fd);
      return arptype;
    }
    status = map_arphrd_to_dlt(handle, arptype, device, 1);
    if (status < 0) {
      close(sock_fd);
      return status;
    }
    if (handle->linktype == -1 ||
        handle->linktype == DLT_LINUX_SLL ||
        handle->linktype == DLT_LINUX_IRDA ||
        handle->linktype == DLT_LINUX_LAPD ||
        handle->linktype == DLT_NETLINK ||
        (handle->linktype == DLT_EN10MB &&
         (strncmp("isdn", device, 4) == 0 ||
          strncmp("isdY", device, 4) == 0))) {
      /*
       * Unknown interface type (-1), or a
       * device we explicitly chose to run
       * in cooked mode (e.g., PPP devices),
       * or an ISDN device (whose link-layer
       * type we can only determine by using
       * APIs that may be different on different
       * kernels) - reopen in cooked mode.
       *
       * If the type is unknown, return a warning;
       * map_arphrd_to_dlt() has already set the
       * warning message.
       */
      if (close(sock_fd) == -1) {
        pcapint_fmt_errmsg_for_errno(handle->errbuf,
            PCAP_ERRBUF_SIZE, errno, "close");
        return PCAP_ERROR;
      }
      sock_fd = socket(PF_PACKET, SOCK_DGRAM, 0);
      if (sock_fd < 0) {
        /*
         * Fatal error.  We treat this as
         * a generic error; we already know
         * that we were able to open a
         * PF_PACKET/SOCK_RAW socket, so
         * any failure is a "this shouldn't
         * happen" case.
         */
        pcapint_fmt_errmsg_for_errno(handle->errbuf,
            PCAP_ERRBUF_SIZE, errno, "socket");
        return PCAP_ERROR;
      }
      handlep->cooked = 1;

      /*
       * Get rid of any link-layer type list
       * we allocated - this only supports cooked
       * capture.
       */
      if (handle->dlt_list != NULL) {
        free(handle->dlt_list);
        handle->dlt_list = NULL;
        handle->dlt_count = 0;
      }

      if (handle->linktype == -1) {
        /*
         * Warn that we're falling back on
         * cooked mode; we may want to
         * update "map_arphrd_to_dlt()"
         * to handle the new type.
         */
        snprintf(handle->errbuf, PCAP_ERRBUF_SIZE,
          "arptype %d not "
          "supported by libpcap - "
          "falling back to cooked "
          "socket",
          arptype);
        status = PCAP_WARNING;
      }

      /*
       * IrDA capture is not a real "cooked" capture,
       * it's IrLAP frames, not IP packets.  The
       * same applies to LAPD capture.
       */
      if (handle->linktype != DLT_LINUX_IRDA &&
          handle->linktype != DLT_LINUX_LAPD &&
          handle->linktype != DLT_NETLINK)
        handle->linktype = DLT_LINUX_SLL;
    }

    handlep->ifindex = iface_get_id(sock_fd, device,
        handle->errbuf);
    if (handlep->ifindex == -1) {
      close(sock_fd);
      return PCAP_ERROR;
    }

    if ((err = iface_bind(sock_fd, handlep->ifindex,
        handle->errbuf, 0)) != 0) {
      close(sock_fd);
      return err;
    }
  } else {
    /*
     * The "any" device.
     */
    if (handle->opt.rfmon) {
      /*
       * It doesn't support monitor mode.
       */
      close(sock_fd);
      return PCAP_ERROR_RFMON_NOTSUP;
    }

    /*
     * It uses cooked mode.
     * Support both DLT_LINUX_SLL and DLT_LINUX_SLL2.
     */
    handlep->cooked = 1;
    handle->linktype = DLT_LINUX_SLL;
    handle->dlt_list = (u_int *) malloc(sizeof(u_int) * 2);
    if (handle->dlt_list == NULL) {
      pcapint_fmt_errmsg_for_errno(handle->errbuf,
          PCAP_ERRBUF_SIZE, errno, "malloc");
      close(sock_fd);
      return (PCAP_ERROR);
    }
    handle->dlt_list[0] = DLT_LINUX_SLL;
    handle->dlt_list[1] = DLT_LINUX_SLL2;
    handle->dlt_count = 2;

    /*
     * We're not bound to a device.
     * For now, we're using this as an indication
     * that we can't transmit; stop doing that only
     * if we figure out how to transmit in cooked
     * mode.
     */
    handlep->ifindex = -1;
  }

  /*
   * Select promiscuous mode on if "promisc" is set.
   *
   * Do not turn allmulti mode on if we don't select
   * promiscuous mode - on some devices (e.g., Orinoco
   * wireless interfaces), allmulti mode isn't supported
   * and the driver implements it by turning promiscuous
   * mode on, and that screws up the operation of the
   * card as a normal networking interface, and on no
   * other platform I know of does starting a non-
   * promiscuous capture affect which multicast packets
   * are received by the interface.
   */

  /*
   * Hmm, how can we set promiscuous mode on all interfaces?
   * I am not sure if that is possible at all.  For now, we
   * silently ignore attempts to turn promiscuous mode on
   * for the "any" device (so you don't have to explicitly
   * disable it in programs such as tcpdump).
   */

  if (!is_any_device && handle->opt.promisc) {
    memset(&mr, 0, sizeof(mr));
    mr.mr_ifindex = handlep->ifindex;
    mr.mr_type    = PACKET_MR_PROMISC;
    if (setsockopt(sock_fd, SOL_PACKET, PACKET_ADD_MEMBERSHIP,
        &mr, sizeof(mr)) == -1) {
      pcapint_fmt_errmsg_for_errno(handle->errbuf,
          PCAP_ERRBUF_SIZE, errno, "setsockopt (PACKET_ADD_MEMBERSHIP)");
      close(sock_fd);
      return PCAP_ERROR;
    }
  }

  /*
   * Enable auxiliary data and reserve room for reconstructing
   * VLAN headers.
   *
   * XXX - is enabling auxiliary data necessary, now that we
   * only support memory-mapped capture?  The kernel's memory-mapped
   * capture code doesn't seem to check whether auxiliary data
   * is enabled, it seems to provide it whether it is or not.
   */
  val = 1;
  if (setsockopt(sock_fd, SOL_PACKET, PACKET_AUXDATA, &val,
           sizeof(val)) == -1 && errno != ENOPROTOOPT) {
    pcapint_fmt_errmsg_for_errno(handle->errbuf, PCAP_ERRBUF_SIZE,
        errno, "setsockopt (PACKET_AUXDATA)");
    close(sock_fd);
    return PCAP_ERROR;
  }
  handle->offset += VLAN_TAG_LEN;

  /*
   * If we're in cooked mode, make the snapshot length
   * large enough to hold a "cooked mode" header plus
   * 1 byte of packet data (so we don't pass a byte
   * count of 0 to "recvfrom()").
   * XXX - we don't know whether this will be DLT_LINUX_SLL
   * or DLT_LINUX_SLL2, so make sure it's big enough for
   * a DLT_LINUX_SLL2 "cooked mode" header; a snapshot length
   * that small is silly anyway.
   */
  if (handlep->cooked) {
    if (handle->snapshot < SLL2_HDR_LEN + 1)
      handle->snapshot = SLL2_HDR_LEN + 1;
  }
  handle->bufsize = handle->snapshot;

  /*
   * Set the offset at which to insert VLAN tags.
   */
  set_vlan_offset(handle);

  if (handle->opt.tstamp_precision == PCAP_TSTAMP_PRECISION_NANO) {
    int nsec_tstamps = 1;

    if (setsockopt(sock_fd, SOL_SOCKET, SO_TIMESTAMPNS, &nsec_tstamps, sizeof(nsec_tstamps)) < 0) {
      snprintf(handle->errbuf, PCAP_ERRBUF_SIZE, "setsockopt: unable to set SO_TIMESTAMPNS");
      close(sock_fd);
      return PCAP_ERROR;
    }
  }

  /*
   * We've succeeded. Save the socket FD in the pcap structure.
   */
  handle->fd = sock_fd;

  /*
   * Any supported Linux version implements at least four auxiliary
   * data items (SKF_AD_PROTOCOL, SKF_AD_PKTTYPE, SKF_AD_IFINDEX and
   * SKF_AD_NLATTR).  Set a flag so the code generator can use these
   * items if necessary.
   */
  handle->bpf_codegen_flags |= BPF_SPECIAL_BASIC_HANDLING;

  /*
   * Can we generate special code for VLAN checks?
   * (XXX - what if we need the special code but it's not supported
   * by the OS?  Is that possible?)
   *
   * This depends on both a runtime condition (the running Linux kernel
   * must support at least SKF_AD_VLAN_TAG_PRESENT in the auxiliary data
   * and must support SO_BPF_EXTENSIONS in order to tell the userland
   * process what it supports) and a compile-time condition (the OS
   * headers must define both constants in order to compile libpcap code
   * that asks the kernel about the support).
   */
#if defined(SO_BPF_EXTENSIONS) && defined(SKF_AD_VLAN_TAG_PRESENT)
  int bpf_extensions;
  socklen_t len = sizeof(bpf_extensions);
  if (getsockopt(sock_fd, SOL_SOCKET, SO_BPF_EXTENSIONS,
      &bpf_extensions, &len) == 0) {
    if (bpf_extensions >= SKF_AD_VLAN_TAG_PRESENT) {
      /*
       * Yes, we can.  Request that we do so.
       */
      handle->bpf_codegen_flags |= BPF_SPECIAL_VLAN_HANDLING;
    }
  }
#endif // defined(SO_BPF_EXTENSIONS) && defined(SKF_AD_VLAN_TAG_PRESENT)

  return status;
}

/*
 * Attempt to setup memory-mapped access.
 *
 * On success, returns 0 if there are no warnings or a PCAP_WARNING_ code
 * if there is a warning.
 *
 * On error, returns the appropriate error code; if that is PCAP_ERROR,
 * sets handle->errbuf to the appropriate message.
 */
static int
setup_mmapped(pcap_t *handle)
{
  struct pcap_linux *handlep = handle->priv;
  int flags = MAP_ANONYMOUS | MAP_PRIVATE;
  int status;

  /*
   * Attempt to allocate a buffer to hold the contents of one
   * packet, for use by the oneshot callback.
   */
#ifdef MAP_32BIT
  if (pcapint_mmap_32bit) flags |= MAP_32BIT;
#endif
  handlep->oneshot_buffer = mmap(0, handle->snapshot, PROT_READ | PROT_WRITE, flags, -1, 0);
  if (handlep->oneshot_buffer == MAP_FAILED) {
    pcapint_fmt_errmsg_for_errno(handle->errbuf, PCAP_ERRBUF_SIZE,
        errno, "can't allocate oneshot buffer");
    return PCAP_ERROR;
  }

  if (handle->opt.buffer_size == 0) {
    /* by default request 2M for the ring buffer */
    handle->opt.buffer_size = 2*1024*1024;
  }
  status = prepare_tpacket_socket(handle);
  if (status == -1) {
    munmap(handlep->oneshot_buffer, handle->snapshot);
    handlep->oneshot_buffer = NULL;
    return PCAP_ERROR;
  }
  status = create_ring(handle);
  if (status < 0) {
    /*
     * Error attempting to enable memory-mapped capture;
     * fail.  The return value is the status to return.
     */
    munmap(handlep->oneshot_buffer, handle->snapshot);
    handlep->oneshot_buffer = NULL;
    return status;
  }

  /*
   * Success.  status has been set either to 0 if there are no
   * warnings or to a PCAP_WARNING_ value if there is a warning.
   *
   * handle->offset is used to get the current position into the rx ring.
   * handle->cc is used to store the ring size.
   */

  /*
   * Set the timeout to use in poll() before returning.
   */
  set_poll_timeout(handlep);

  return status;
}

/*
 * Attempt to set the socket to the specified version of the memory-mapped
 * header.
 *
 * Return 0 if we succeed; return 1 if we fail because that version isn't
 * supported; return -1 on any other error, and set handle->errbuf.
 */
static int
init_tpacket(pcap_t *handle, int version, const char *version_str)
{
  struct pcap_linux *handlep = handle->priv;
  int val = version;
  socklen_t len = sizeof(val);

  /*
   * Probe whether kernel supports the specified TPACKET version;
   * this also gets the length of the header for that version.
   *
   * This socket option was introduced in 2.6.27, which was
   * also the first release with TPACKET_V2 support.
   */
  if (getsockopt(handle->fd, SOL_PACKET, PACKET_HDRLEN, &val, &len) < 0) {
    if (errno == EINVAL) {
      /*
       * EINVAL means this specific version of TPACKET
       * is not supported. Tell the caller they can try
       * with a different one; if they've run out of
       * others to try, let them set the error message
       * appropriately.
       */
      return 1;
    }

    /*
     * All other errors are fatal.
     */
    if (errno == ENOPROTOOPT) {
      /*
       * PACKET_HDRLEN isn't supported, which means
       * that memory-mapped capture isn't supported.
       * Indicate that in the message.
       */
      snprintf(handle->errbuf, PCAP_ERRBUF_SIZE,
          "Kernel doesn't support memory-mapped capture; a 2.6.27 or later 2.x kernel is required, with CONFIG_PACKET_MMAP specified for 2.x kernels");
    } else {
      /*
       * Some unexpected error.
       */
      pcapint_fmt_errmsg_for_errno(handle->errbuf, PCAP_ERRBUF_SIZE,
          errno, "can't get %s header len on packet socket",
          version_str);
    }
    return -1;
  }
  handlep->tp_hdrlen = val;

  val = version;
  if (setsockopt(handle->fd, SOL_PACKET, PACKET_VERSION, &val,
         sizeof(val)) < 0) {
    pcapint_fmt_errmsg_for_errno(handle->errbuf, PCAP_ERRBUF_SIZE,
        errno, "can't activate %s on packet socket", version_str);
    return -1;
  }
  handlep->tp_version = version;

  return 0;
}

/*
 * Attempt to set the socket to version 3 of the memory-mapped header and,
 * if that fails because version 3 isn't supported, attempt to fall
 * back to version 2.  If version 2 isn't supported, just fail.
 *
 * Return 0 if we succeed and -1 on any other error, and set handle->errbuf.
 */
static int
prepare_tpacket_socket(pcap_t *handle)
{
  int ret;

#ifdef HAVE_TPACKET3
  /*
   * Try setting the version to TPACKET_V3.
   *
   * The only mode in which buffering is done on PF_PACKET
   * sockets, so that packets might not be delivered
   * immediately, is TPACKET_V3 mode.
   *
   * The buffering cannot be disabled in that mode, so
   * if the user has requested immediate mode, we don't
   * use TPACKET_V3.
   */
  if (!handle->opt.immediate) {
    ret = init_tpacket(handle, TPACKET_V3, "TPACKET_V3");
    if (ret == 0) {
      /*
       * Success.
       */
      return 0;
    }
    if (ret == -1) {
      /*
       * We failed for some reason other than "the
       * kernel doesn't support TPACKET_V3".
       */
      return -1;
    }

    /*
     * This means it returned 1, which means "the kernel
     * doesn't support TPACKET_V3"; try TPACKET_V2.
     */
  }
#endif /* HAVE_TPACKET3 */

  /*
   * Try setting the version to TPACKET_V2.
   */
  ret = init_tpacket(handle, TPACKET_V2, "TPACKET_V2");
  if (ret == 0) {
    /*
     * Success.
     */
    return 0;
  }

  if (ret == 1) {
    /*
     * OK, the kernel supports memory-mapped capture, but
     * not TPACKET_V2.  Set the error message appropriately.
     */
    snprintf(handle->errbuf, PCAP_ERRBUF_SIZE,
        "Kernel doesn't support TPACKET_V2; a 2.6.27 or later kernel is required");
  }

  /*
   * We failed.
   */
  return -1;
}

/*
 * Attempt to set up memory-mapped access.
 *
 * On success, returns 0 if there are no warnings or to a PCAP_WARNING_ code
 * if there is a warning.
 *
 * On error, returns the appropriate error code; if that is PCAP_ERROR,
 * sets handle->errbuf to the appropriate message.
 */
static int
create_ring(pcap_t *handle)
{
  struct pcap_linux *handlep = handle->priv;
  unsigned i, j, frames_per_block;
  int flags = MAP_SHARED;
#ifdef HAVE_TPACKET3
  /*
   * For sockets using TPACKET_V2, the extra stuff at the end of a
   * struct tpacket_req3 will be ignored, so this is OK even for
   * those sockets.
   */
  struct tpacket_req3 req;
#else
  struct tpacket_req req;
#endif
  socklen_t len;
  unsigned int sk_type, tp_reserve, maclen, tp_hdrlen, netoff, macoff;
  unsigned int frame_size;
  int status;

  /*
   * Start out assuming no warnings.
   */
  status = 0;

  /*
   * Reserve space for VLAN tag reconstruction.
   */
  tp_reserve = VLAN_TAG_LEN;

  /*
   * If we're capturing in cooked mode, reserve space for
   * a DLT_LINUX_SLL2 header; we don't know yet whether
   * we'll be using DLT_LINUX_SLL or DLT_LINUX_SLL2, as
   * that can be changed on an open device, so we reserve
   * space for the larger of the two.
   *
   * XXX - we assume that the kernel is still adding
   * 16 bytes of extra space, so we subtract 16 from
   * SLL2_HDR_LEN to get the additional space needed.
   * (Are they doing that for DLT_LINUX_SLL, the link-
   * layer header for which is 16 bytes?)
   *
   * XXX - should we use TPACKET_ALIGN(SLL2_HDR_LEN - 16)?
   */
  if (handlep->cooked)
    tp_reserve += SLL2_HDR_LEN - 16;

  /*
   * Try to request that amount of reserve space.
   * This must be done before creating the ring buffer.
   */
  len = sizeof(tp_reserve);
  if (setsockopt(handle->fd, SOL_PACKET, PACKET_RESERVE,
      &tp_reserve, len) < 0) {
    pcapint_fmt_errmsg_for_errno(handle->errbuf,
        PCAP_ERRBUF_SIZE, errno,
        "setsockopt (PACKET_RESERVE)");
    return PCAP_ERROR;
  }

  switch (handlep->tp_version) {

  case TPACKET_V2:
    /* Note that with large snapshot length (say 256K, which is
     * the default for recent versions of tcpdump, Wireshark,
     * TShark, dumpcap or 64K, the value that "-s 0" has given for
     * a long time with tcpdump), if we use the snapshot
     * length to calculate the frame length, only a few frames
     * will be available in the ring even with pretty
     * large ring size (and a lot of memory will be unused).
     *
     * Ideally, we should choose a frame length based on the
     * minimum of the specified snapshot length and the maximum
     * packet size.  That's not as easy as it sounds; consider,
     * for example, an 802.11 interface in monitor mode, where
     * the frame would include a radiotap header, where the
     * maximum radiotap header length is device-dependent.
     *
     * So, for now, we just do this for Ethernet devices, where
     * there's no metadata header, and the link-layer header is
     * fixed length.  We can get the maximum packet size by
     * adding 18, the Ethernet header length plus the CRC length
     * (just in case we happen to get the CRC in the packet), to
     * the MTU of the interface; we fetch the MTU in the hopes
     * that it reflects support for jumbo frames.  (Even if the
     * interface is just being used for passive snooping, the
     * driver might set the size of buffers in the receive ring
     * based on the MTU, so that the MTU limits the maximum size
     * of packets that we can receive.)
     *
     * If segmentation/fragmentation or receive offload are
     * enabled, we can get reassembled/aggregated packets larger
     * than MTU, but bounded to 65535 plus the Ethernet overhead,
     * due to kernel and protocol constraints */
    frame_size = handle->snapshot;
    if (handle->linktype == DLT_EN10MB) {
      unsigned int max_frame_len;
      int mtu;
      int offload;

      mtu = iface_get_mtu(handle->fd, handle->opt.device,
          handle->errbuf);
      if (mtu == -1)
        return PCAP_ERROR;
      offload = iface_get_offload(handle);
      if (offload == -1)
        return PCAP_ERROR;
      if (offload)
        max_frame_len = max(mtu, 65535);
      else
        max_frame_len = mtu;
      max_frame_len += 18;

      if (frame_size > max_frame_len)
        frame_size = max_frame_len;
    }

    /* NOTE: calculus matching those in tpacket_rcv()
     * in linux-2.6/net/packet/af_packet.c
     */
    len = sizeof(sk_type);
    if (getsockopt(handle->fd, SOL_SOCKET, SO_TYPE, &sk_type,
        &len) < 0) {
      pcapint_fmt_errmsg_for_errno(handle->errbuf,
          PCAP_ERRBUF_SIZE, errno, "getsockopt (SO_TYPE)");
      return PCAP_ERROR;
    }
    maclen = (sk_type == SOCK_DGRAM) ? 0 : MAX_LINKHEADER_SIZE;
      /* XXX: in the kernel maclen is calculated from
       * LL_ALLOCATED_SPACE(dev) and vnet_hdr.hdr_len
       * in:  packet_snd()           in linux-2.6/net/packet/af_packet.c
       * then packet_alloc_skb()     in linux-2.6/net/packet/af_packet.c
       * then sock_alloc_send_pskb() in linux-2.6/net/core/sock.c
       * but I see no way to get those sizes in userspace,
       * like for instance with an ifreq ioctl();
       * the best thing I've found so far is MAX_HEADER in
       * the kernel part of linux-2.6/include/linux/netdevice.h
       * which goes up to 128+48=176; since pcap-linux.c
       * defines a MAX_LINKHEADER_SIZE of 256 which is
       * greater than that, let's use it.. maybe is it even
       * large enough to directly replace macoff..
       */
    tp_hdrlen = TPACKET_ALIGN(handlep->tp_hdrlen) + sizeof(struct sockaddr_ll) ;
    netoff = TPACKET_ALIGN(tp_hdrlen + (maclen < 16 ? 16 : maclen)) + tp_reserve;
      /* NOTE: AFAICS tp_reserve may break the TPACKET_ALIGN
       * of netoff, which contradicts
       * linux-2.6/Documentation/networking/packet_mmap.txt
       * documenting that:
       * "- Gap, chosen so that packet data (Start+tp_net)
       * aligns to TPACKET_ALIGNMENT=16"
       */
      /* NOTE: in linux-2.6/include/linux/skbuff.h:
       * "CPUs often take a performance hit
       *  when accessing unaligned memory locations"
       */
    macoff = netoff - maclen;
    req.tp_frame_size = TPACKET_ALIGN(macoff + frame_size);
    /*
     * Round the buffer size up to a multiple of the
     * frame size (rather than rounding down, which
     * would give a buffer smaller than our caller asked
     * for, and possibly give zero frames if the requested
     * buffer size is too small for one frame).
     */
    req.tp_frame_nr = (handle->opt.buffer_size + req.tp_frame_size - 1)/req.tp_frame_size;
    break;

#ifdef HAVE_TPACKET3
  case TPACKET_V3:
    /* The "frames" for this are actually buffers that
     * contain multiple variable-sized frames.
     *
     * We pick a "frame" size of MAXIMUM_SNAPLEN to leave
     * enough room for at least one reasonably-sized packet
     * in the "frame". */
    req.tp_frame_size = MAXIMUM_SNAPLEN;
    /*
     * Round the buffer size up to a multiple of the
     * "frame" size (rather than rounding down, which
     * would give a buffer smaller than our caller asked
     * for, and possibly give zero "frames" if the requested
     * buffer size is too small for one "frame").
     */
    req.tp_frame_nr = (handle->opt.buffer_size + req.tp_frame_size - 1)/req.tp_frame_size;
    break;
#endif
  default:
    snprintf(handle->errbuf, PCAP_ERRBUF_SIZE,
        "Internal error: unknown TPACKET_ value %u",
        handlep->tp_version);
    return PCAP_ERROR;
  }

  /* compute the minimum block size that will handle this frame.
   * The block has to be page size aligned.
   * The max block size allowed by the kernel is arch-dependent and
   * it's not explicitly checked here. */
  req.tp_block_size = getpagesize();
  while (req.tp_block_size < req.tp_frame_size)
    req.tp_block_size <<= 1;

  frames_per_block = req.tp_block_size/req.tp_frame_size;

  /*
   * PACKET_TIMESTAMP was added after linux/net_tstamp.h was,
   * so we check for PACKET_TIMESTAMP.  We check for
   * linux/net_tstamp.h just in case a system somehow has
   * PACKET_TIMESTAMP but not linux/net_tstamp.h; that might
   * be unnecessary.
   *
   * SIOCSHWTSTAMP was introduced in the patch that introduced
   * linux/net_tstamp.h, so we don't bother checking whether
   * SIOCSHWTSTAMP is defined (if your Linux system has
   * linux/net_tstamp.h but doesn't define SIOCSHWTSTAMP, your
   * Linux system is badly broken).
   */
#if defined(HAVE_LINUX_NET_TSTAMP_H) && defined(PACKET_TIMESTAMP)
  /*
   * If we were told to do so, ask the kernel and the driver
   * to use hardware timestamps.
   *
   * Hardware timestamps are only supported with mmapped
   * captures.
   */
  if (handle->opt.tstamp_type == PCAP_TSTAMP_ADAPTER ||
      handle->opt.tstamp_type == PCAP_TSTAMP_ADAPTER_UNSYNCED) {
    struct hwtstamp_config hwconfig;
    struct ifreq ifr;
    int timesource;

    /*
     * Ask for hardware time stamps on all packets,
     * including transmitted packets.
     */
    memset(&hwconfig, 0, sizeof(hwconfig));
    hwconfig.tx_type = HWTSTAMP_TX_ON;
    hwconfig.rx_filter = HWTSTAMP_FILTER_ALL;

    memset(&ifr, 0, sizeof(ifr));
    pcapint_strlcpy(ifr.ifr_name, handle->opt.device, sizeof(ifr.ifr_name));
    ifr.ifr_data = (void *)&hwconfig;

    /*
     * This may require CAP_NET_ADMIN.
     */
    if (ioctl(handle->fd, SIOCSHWTSTAMP, &ifr) < 0) {
      switch (errno) {

      case EPERM:
        /*
         * Treat this as an error, as the
         * user should try to run this
         * with the appropriate privileges -
         * and, if they can't, shouldn't
         * try requesting hardware time stamps.
         */
        snprintf(handle->errbuf, PCAP_ERRBUF_SIZE,
            "Attempt to set hardware timestamp failed - CAP_NET_ADMIN may be required");
        return PCAP_ERROR_PERM_DENIED;

      case EOPNOTSUPP:
      case ERANGE:
        /*
         * Treat this as a warning, as the
         * only way to fix the warning is to
         * get an adapter that supports hardware
         * time stamps for *all* packets.
         * (ERANGE means "we support hardware
         * time stamps, but for packets matching
         * that particular filter", so it means
         * "we don't support hardware time stamps
         * for all incoming packets" here.)
         *
         * We'll just fall back on the standard
         * host time stamps.
         */
        status = PCAP_WARNING_TSTAMP_TYPE_NOTSUP;
        break;

      default:
        pcapint_fmt_errmsg_for_errno(handle->errbuf,
            PCAP_ERRBUF_SIZE, errno,
            "SIOCSHWTSTAMP failed");
        return PCAP_ERROR;
      }
    } else {
      /*
       * Well, that worked.  Now specify the type of
       * hardware time stamp we want for this
       * socket.
       */
      if (handle->opt.tstamp_type == PCAP_TSTAMP_ADAPTER) {
        /*
         * Hardware timestamp, synchronized
         * with the system clock.
         */
        timesource = SOF_TIMESTAMPING_SYS_HARDWARE;
      } else {
        /*
         * PCAP_TSTAMP_ADAPTER_UNSYNCED - hardware
         * timestamp, not synchronized with the
         * system clock.
         */
        timesource = SOF_TIMESTAMPING_RAW_HARDWARE;
      }
      if (setsockopt(handle->fd, SOL_PACKET, PACKET_TIMESTAMP,
        (void *)&timesource, sizeof(timesource))) {
        pcapint_fmt_errmsg_for_errno(handle->errbuf,
            PCAP_ERRBUF_SIZE, errno,
            "can't set PACKET_TIMESTAMP");
        return PCAP_ERROR;
      }
    }
  }
#endif /* HAVE_LINUX_NET_TSTAMP_H && PACKET_TIMESTAMP */

  /* ask the kernel to create the ring */
retry:
  req.tp_block_nr = req.tp_frame_nr / frames_per_block;

  /* req.tp_frame_nr is requested to match frames_per_block*req.tp_block_nr */
  req.tp_frame_nr = req.tp_block_nr * frames_per_block;

#ifdef HAVE_TPACKET3
  /* timeout value to retire block - use the configured buffering timeout, or default if <0. */
  if (handlep->timeout > 0) {
    /* Use the user specified timeout as the block timeout */
    req.tp_retire_blk_tov = handlep->timeout;
  } else if (handlep->timeout == 0) {
    /*
     * In pcap, this means "infinite timeout"; TPACKET_V3
     * doesn't support that, so just set it to UINT_MAX
     * milliseconds.  In the TPACKET_V3 loop, if the
     * timeout is 0, and we haven't yet seen any packets,
     * and we block and still don't have any packets, we
     * keep blocking until we do.
     */
    req.tp_retire_blk_tov = UINT_MAX;
  } else {
    /*
     * XXX - this is not valid; use 0, meaning "have the
     * kernel pick a default", for now.
     */
    req.tp_retire_blk_tov = 0;
  }
  /* private data not used */
  req.tp_sizeof_priv = 0;
  /* Rx ring - feature request bits - none (rxhash will not be filled) */
  req.tp_feature_req_word = 0;
#endif

  if (setsockopt(handle->fd, SOL_PACKET, PACKET_RX_RING,
          (void *) &req, sizeof(req))) {
    if ((errno == ENOMEM) && (req.tp_block_nr > 1)) {
      /*
       * Memory failure; try to reduce the requested ring
       * size.
       *
       * We used to reduce this by half -- do 5% instead.
       * That may result in more iterations and a longer
       * startup, but the user will be much happier with
       * the resulting buffer size.
       */
      if (req.tp_frame_nr < 20)
        req.tp_frame_nr -= 1;
      else
        req.tp_frame_nr -= req.tp_frame_nr/20;
      goto retry;
    }
    pcapint_fmt_errmsg_for_errno(handle->errbuf, PCAP_ERRBUF_SIZE,
        errno, "can't create rx ring on packet socket");
    return PCAP_ERROR;
  }

  /* memory map the rx ring */
  handlep->mmapbuflen = req.tp_block_nr * req.tp_block_size;
#ifdef MAP_32BIT
  if (pcapint_mmap_32bit) flags |= MAP_32BIT;
#endif
  handlep->mmapbuf = mmap(0, handlep->mmapbuflen, PROT_READ | PROT_WRITE, flags, handle->fd, 0);
  if (handlep->mmapbuf == MAP_FAILED) {
    pcapint_fmt_errmsg_for_errno(handle->errbuf, PCAP_ERRBUF_SIZE,
        errno, "can't mmap rx ring");

    /* clear the allocated ring on error*/
    destroy_ring(handle);
    return PCAP_ERROR;
  }

  /* allocate a ring for each frame header pointer*/
  handle->cc = req.tp_frame_nr;
  handle->buffer = malloc(handle->cc * sizeof(union thdr *));
  if (!handle->buffer) {
    pcapint_fmt_errmsg_for_errno(handle->errbuf, PCAP_ERRBUF_SIZE,
        errno, "can't allocate ring of frame headers");

    destroy_ring(handle);
    return PCAP_ERROR;
  }

  /* fill the header ring with proper frame ptr*/
  handle->offset = 0;
  for (i=0; i<req.tp_block_nr; ++i) {
    u_char *base = &handlep->mmapbuf[i*req.tp_block_size];
    for (j=0; j<frames_per_block; ++j, ++handle->offset) {
      RING_GET_CURRENT_FRAME(handle) = base;
      base += req.tp_frame_size;
    }
  }

  handle->bufsize = req.tp_frame_size;
  handle->offset = 0;
  return status;
}

/* free all ring related resources*/
static void
destroy_ring(pcap_t *handle)
{
  struct pcap_linux *handlep = handle->priv;

  /*
   * Tell the kernel to destroy the ring.
   * We don't check for setsockopt failure, as 1) we can't recover
   * from an error and 2) we might not yet have set it up in the
   * first place.
   */
  struct tpacket_req req;
  memset(&req, 0, sizeof(req));
  (void)setsockopt(handle->fd, SOL_PACKET, PACKET_RX_RING,
        (void *) &req, sizeof(req));

  /* if ring is mapped, unmap it*/
  if (handlep->mmapbuf) {
    /* do not test for mmap failure, as we can't recover from any error */
    (void)munmap(handlep->mmapbuf, handlep->mmapbuflen);
    handlep->mmapbuf = NULL;
  }
}

/*
 * Special one-shot callback, used for pcap_next() and pcap_next_ex(),
 * for Linux mmapped capture.
 *
 * The problem is that pcap_next() and pcap_next_ex() expect the packet
 * data handed to the callback to be valid after the callback returns,
 * but pcap_read_linux_mmap() has to release that packet as soon as
 * the callback returns (otherwise, the kernel thinks there's still
 * at least one unprocessed packet available in the ring, so a select()
 * will immediately return indicating that there's data to process), so,
 * in the callback, we have to make a copy of the packet.
 *
 * Yes, this means that, if the capture is using the ring buffer, using
 * pcap_next() or pcap_next_ex() requires more copies than using
 * pcap_loop() or pcap_dispatch().  If that bothers you, don't use
 * pcap_next() or pcap_next_ex().
 */
static void
pcapint_oneshot_linux(u_char *user, const struct pcap_pkthdr *h,
    const u_char *bytes)
{
  struct oneshot_userdata *sp = (struct oneshot_userdata *)user;
  pcap_t *handle = sp->pd;
  struct pcap_linux *handlep = handle->priv;

  *sp->hdr = *h;
  memcpy(handlep->oneshot_buffer, bytes, h->caplen);
  *sp->pkt = handlep->oneshot_buffer;
}

static int
pcap_getnonblock_linux(pcap_t *handle)
{
  struct pcap_linux *handlep = handle->priv;

  /* use negative value of timeout to indicate non blocking ops */
  return (handlep->timeout<0);
}

static int
pcap_setnonblock_linux(pcap_t *handle, int nonblock)
{
  struct pcap_linux *handlep = handle->priv;

  /*
   * Set the file descriptor to the requested mode, as we use
   * it for sending packets.
   */
  if (pcapint_setnonblock_fd(handle, nonblock) == -1)
    return -1;

  /*
   * Map each value to their corresponding negation to
   * preserve the timeout value provided with pcap_set_timeout.
   */
  if (nonblock) {
    /*
     * We're setting the mode to non-blocking mode.
     */
    if (handlep->timeout >= 0) {
      /*
       * Indicate that we're switching to
       * non-blocking mode.
       */
      handlep->timeout = ~handlep->timeout;
    }
    if (handlep->poll_breakloop_fd != -1) {
      /* Close the eventfd; we do not need it in nonblock mode. */
      close(handlep->poll_breakloop_fd);
      handlep->poll_breakloop_fd = -1;
    }
  } else {
    /*
     * We're setting the mode to blocking mode.
     */
    if (handlep->poll_breakloop_fd == -1) {
      /* If we did not have an eventfd, open one now that we are blocking. */
      if ( ( handlep->poll_breakloop_fd = eventfd(0, EFD_NONBLOCK) ) == -1 ) {
        pcapint_fmt_errmsg_for_errno(handle->errbuf,
            PCAP_ERRBUF_SIZE, errno,
            "could not open eventfd");
        return -1;
      }
    }
    if (handlep->timeout < 0) {
      handlep->timeout = ~handlep->timeout;
    }
  }
  /* Update the timeout to use in poll(). */
  set_poll_timeout(handlep);
  return 0;
}

/*
 * Get the status field of the ring buffer frame at a specified offset.
 */
static inline u_int
pcap_get_ring_frame_status(pcap_t *handle, u_int offset)
{
  struct pcap_linux *handlep = handle->priv;
  union thdr h;

  h.raw = RING_GET_FRAME_AT(handle, offset);
  switch (handlep->tp_version) {
  case TPACKET_V2:
    return __atomic_load_n(&h.h2->tp_status, __ATOMIC_ACQUIRE);
#ifdef HAVE_TPACKET3
  case TPACKET_V3:
    return __atomic_load_n(&h.h3->hdr.bh1.block_status, __ATOMIC_ACQUIRE);
#endif
  }
  /* This should not happen. */
  return 0;
}

/*
 * Block waiting for frames to be available.
 */
static int pcap_wait_for_frames_mmap(pcap_t *handle)
{
  struct pcap_linux *handlep = handle->priv;
  int timeout;
  struct ifreq ifr;
  int ret;
  struct pollfd pollinfo[2];
  int numpollinfo;
  pollinfo[0].fd = handle->fd;
  pollinfo[0].events = POLLIN;
  if ( handlep->poll_breakloop_fd == -1 ) {
    numpollinfo = 1;
    pollinfo[1].revents = 0;
    /*
     * We set pollinfo[1].revents to zero, even though
     * numpollinfo = 1 meaning that poll() doesn't see
     * pollinfo[1], so that we do not have to add a
     * conditional of numpollinfo > 1 below when we
     * test pollinfo[1].revents.
     */
  } else {
    pollinfo[1].fd = handlep->poll_breakloop_fd;
    pollinfo[1].events = POLLIN;
    numpollinfo = 2;
  }

  /*
   * Keep polling until we either get some packets to read, see
   * that we got told to break out of the loop, get a fatal error,
   * or discover that the device went away.
   *
   * In non-blocking mode, we must still do one poll() to catch
   * any pending error indications, but the poll() has a timeout
   * of 0, so that it doesn't block, and we quit after that one
   * poll().
   *
   * If we've seen an ENETDOWN, it might be the first indication
   * that the device went away, or it might just be that it was
   * configured down.  Unfortunately, there's no guarantee that
   * the device has actually been removed as an interface, because:
   *
   * 1) if, as appears to be the case at least some of the time,
   * the PF_PACKET socket code first gets a NETDEV_DOWN indication
   * for the device and then gets a NETDEV_UNREGISTER indication
   * for it, the first indication will cause a wakeup with ENETDOWN
   * but won't set the packet socket's field for the interface index
   * to -1, and the second indication won't cause a wakeup (because
   * the first indication also caused the protocol hook to be
   * unregistered) but will set the packet socket's field for the
   * interface index to -1;
   *
   * 2) even if just a NETDEV_UNREGISTER indication is registered,
   * the packet socket's field for the interface index only gets
   * set to -1 after the wakeup, so there's a small but non-zero
   * risk that a thread blocked waiting for the wakeup will get
   * to the "fetch the socket name" code before the interface index
   * gets set to -1, so it'll get the old interface index.
   *
   * Therefore, if we got an ENETDOWN and haven't seen a packet
   * since then, we assume that we might be waiting for the interface
   * to disappear, and poll with a timeout to try again in a short
   * period of time.  If we *do* see a packet, the interface has
   * come back up again, and is *definitely* still there, so we
   * don't need to poll.
   */
  for (;;) {
    /*
     * Yes, we do this even in non-blocking mode, as it's
     * the only way to get error indications from a
     * tpacket socket.
     *
     * The timeout is 0 in non-blocking mode, so poll()
     * returns immediately.
     */
    timeout = handlep->poll_timeout;

    /*
     * If we got an ENETDOWN and haven't gotten an indication
     * that the device has gone away or that the device is up,
     * we don't yet know for certain whether the device has
     * gone away or not, do a poll() with a 1-millisecond timeout,
     * as we have to poll indefinitely for "device went away"
     * indications until we either get one or see that the
     * device is up.
     */
    if (handlep->netdown) {
      if (timeout != 0)
        timeout = 1;
    }
    ret = poll(pollinfo, numpollinfo, timeout);
    if (ret < 0) {
      /*
       * Error.  If it's not EINTR, report it.
       */
      if (errno != EINTR) {
        pcapint_fmt_errmsg_for_errno(handle->errbuf,
            PCAP_ERRBUF_SIZE, errno,
            "can't poll on packet socket");
        return PCAP_ERROR;
      }

      /*
       * It's EINTR; if we were told to break out of
       * the loop, do so.
       */
      if (handle->break_loop) {
        handle->break_loop = 0;
        return PCAP_ERROR_BREAK;
      }
    } else if (ret > 0) {
      /*
       * OK, some descriptor is ready.
       * Check the socket descriptor first.
       *
       * As I read the Linux man page, pollinfo[0].revents
       * will either be POLLIN, POLLERR, POLLHUP, or POLLNVAL.
       */
      if (pollinfo[0].revents == POLLIN) {
        /*
         * OK, we may have packets to
         * read.
         */
        break;
      }
      if (pollinfo[0].revents != 0) {
        /*
         * There's some indication other than
         * "you can read on this descriptor" on
         * the descriptor.
         */
        if (pollinfo[0].revents & POLLNVAL) {
          snprintf(handle->errbuf,
              PCAP_ERRBUF_SIZE,
              "Invalid polling request on packet socket");
          return PCAP_ERROR;
        }
        if (pollinfo[0].revents & (POLLHUP | POLLRDHUP)) {
          snprintf(handle->errbuf,
              PCAP_ERRBUF_SIZE,
              "Hangup on packet socket");
          return PCAP_ERROR;
        }
        if (pollinfo[0].revents & POLLERR) {
          /*
           * Get the error.
           */
          int err;
          socklen_t errlen;

          errlen = sizeof(err);
          if (getsockopt(handle->fd, SOL_SOCKET,
              SO_ERROR, &err, &errlen) == -1) {
            /*
             * The call *itself* returned
             * an error; make *that*
             * the error.
             */
            err = errno;
          }

          /*
           * OK, we have the error.
           */
          if (err == ENETDOWN) {
            /*
             * The device on which we're
             * capturing went away or the
             * interface was taken down.
             *
             * We don't know for certain
             * which happened, and the
             * next poll() may indicate
             * that there are packets
             * to be read, so just set
             * a flag to get us to do
             * checks later, and set
             * the required select
             * timeout to 1 millisecond
             * so that event loops that
             * check our socket descriptor
             * also time out so that
             * they can call us and we
             * can do the checks.
             */
            handlep->netdown = 1;
            handle->required_select_timeout = &netdown_timeout;
          } else if (err == 0) {
            /*
             * This shouldn't happen, so
             * report a special indication
             * that it did.
             */
            snprintf(handle->errbuf,
                PCAP_ERRBUF_SIZE,
                "Error condition on packet socket: Reported error was 0");
            return PCAP_ERROR;
          } else {
            pcapint_fmt_errmsg_for_errno(handle->errbuf,
                PCAP_ERRBUF_SIZE,
                err,
                "Error condition on packet socket");
            return PCAP_ERROR;
          }
        }
      }
      /*
       * Now check the event device.
       */
      if (pollinfo[1].revents & POLLIN) {
        ssize_t nread;
        uint64_t value;

        /*
         * This should never fail, but, just
         * in case....
         */
        nread = read(handlep->poll_breakloop_fd, &value,
            sizeof(value));
        if (nread == -1) {
          pcapint_fmt_errmsg_for_errno(handle->errbuf,
              PCAP_ERRBUF_SIZE,
              errno,
              "Error reading from event FD");
          return PCAP_ERROR;
        }

        /*
         * According to the Linux read(2) man
         * page, read() will transfer at most
         * 2^31-1 bytes, so the return value is
         * either -1 or a value between 0
         * and 2^31-1, so it's non-negative.
         *
         * Cast it to size_t to squelch
         * warnings from the compiler; add this
         * comment to squelch warnings from
         * humans reading the code. :-)
         *
         * Don't treat an EOF as an error, but
         * *do* treat a short read as an error;
         * that "shouldn't happen", but....
         */
        if (nread != 0 &&
            (size_t)nread < sizeof(value)) {
          snprintf(handle->errbuf, PCAP_ERRBUF_SIZE,
              "Short read from event FD: expected %zu, got %zd",
              sizeof(value), nread);
          return PCAP_ERROR;
        }

        /*
         * This event gets signaled by a
         * pcap_breakloop() call; if we were told
         * to break out of the loop, do so.
         */
        if (handle->break_loop) {
          handle->break_loop = 0;
          return PCAP_ERROR_BREAK;
        }
      }
    }

    /*
     * Either:
     *
     *   1) we got neither an error from poll() nor any
     *      readable descriptors, in which case there
     *      are no packets waiting to read
     *
     * or
     *
     *   2) We got readable descriptors but the PF_PACKET
     *      socket wasn't one of them, in which case there
     *      are no packets waiting to read
     *
     * so, if we got an ENETDOWN, we've drained whatever
     * packets were available to read at the point of the
     * ENETDOWN.
     *
     * So, if we got an ENETDOWN and haven't gotten an indication
     * that the device has gone away or that the device is up,
     * we don't yet know for certain whether the device has
     * gone away or not, check whether the device exists and is
     * up.
     */
    if (handlep->netdown) {
      if (!device_still_exists(handle)) {
        /*
         * The device doesn't exist any more;
         * report that.
         *
         * XXX - we should really return an
         * appropriate error for that, but
         * pcap_dispatch() etc. aren't documented
         * as having error returns other than
         * PCAP_ERROR or PCAP_ERROR_BREAK.
         */
        snprintf(handle->errbuf,  PCAP_ERRBUF_SIZE,
            "The interface disappeared");
        return PCAP_ERROR;
      }

      /*
       * The device still exists; try to see if it's up.
       */
      memset(&ifr, 0, sizeof(ifr));
      pcapint_strlcpy(ifr.ifr_name, handlep->device,
          sizeof(ifr.ifr_name));
      if (ioctl(handle->fd, SIOCGIFFLAGS, &ifr) == -1) {
        if (errno == ENXIO || errno == ENODEV) {
          /*
           * OK, *now* it's gone.
           *
           * XXX - see above comment.
           */
          snprintf(handle->errbuf,
              PCAP_ERRBUF_SIZE,
              "The interface disappeared");
          return PCAP_ERROR;
        } else {
          pcapint_fmt_errmsg_for_errno(handle->errbuf,
              PCAP_ERRBUF_SIZE, errno,
              "%s: Can't get flags",
              handlep->device);
          return PCAP_ERROR;
        }
      }
      if (ifr.ifr_flags & IFF_UP) {
        /*
         * It's up, so it definitely still exists.
         * Cancel the ENETDOWN indication - we
         * presumably got it due to the interface
         * going down rather than the device going
         * away - and revert to "no required select
         * timeout.
         */
        handlep->netdown = 0;
        handle->required_select_timeout = NULL;
      }
    }

    /*
     * If we're in non-blocking mode, just quit now, rather
     * than spinning in a loop doing poll()s that immediately
     * time out if there's no indication on any descriptor.
     */
    if (handlep->poll_timeout == 0)
      break;
  }
  return 0;
}

/* handle a single memory mapped packet */
static int pcap_handle_packet_mmap(
    pcap_t *handle,
    pcap_handler callback,
    u_char *user,
    unsigned char *frame,
    unsigned int tp_len,
    unsigned int tp_mac,
    unsigned int tp_snaplen,
    unsigned int tp_sec,
    unsigned int tp_usec,
    int tp_vlan_tci_valid,
    __u16 tp_vlan_tci,
    __u16 tp_vlan_tpid)
{
  struct pcap_linux *handlep = handle->priv;
  unsigned char *bp;
  struct sockaddr_ll *sll;
  struct pcap_pkthdr pcaphdr;
  unsigned int snaplen = tp_snaplen;
  struct utsname utsname;

  /* perform sanity check on internal offset. */
  if (tp_mac + tp_snaplen > handle->bufsize) {
    /*
     * Report some system information as a debugging aid.
     */
    if (uname(&utsname) != -1) {
      snprintf(handle->errbuf, PCAP_ERRBUF_SIZE,
        "corrupted frame on kernel ring mac "
        "offset %u + caplen %u > frame len %d "
        "(kernel %.32s version %s, machine %.16s)",
        tp_mac, tp_snaplen, handle->bufsize,
        utsname.release, utsname.version,
        utsname.machine);
    } else {
      snprintf(handle->errbuf, PCAP_ERRBUF_SIZE,
        "corrupted frame on kernel ring mac "
        "offset %u + caplen %u > frame len %d",
        tp_mac, tp_snaplen, handle->bufsize);
    }
    return -1;
  }

  /* run filter on received packet
   * If the kernel filtering is enabled we need to run the
   * filter until all the frames present into the ring
   * at filter creation time are processed.
   * In this case, blocks_to_filter_in_userland is used
   * as a counter for the packet we need to filter.
   * Note: alternatively it could be possible to stop applying
   * the filter when the ring became empty, but it can possibly
   * happen a lot later... */
  bp = frame + tp_mac;

  /* if required build in place the sll header*/
  sll = (void *)(frame + TPACKET_ALIGN(handlep->tp_hdrlen));
  if (handlep->cooked) {
    if (handle->linktype == DLT_LINUX_SLL2) {
      struct sll2_header *hdrp;

      /*
       * The kernel should have left us with enough
       * space for an sll header; back up the packet
       * data pointer into that space, as that'll be
       * the beginning of the packet we pass to the
       * callback.
       */
      bp -= SLL2_HDR_LEN;

      /*
       * Let's make sure that's past the end of
       * the tpacket header, i.e. >=
       * ((u_char *)thdr + TPACKET_HDRLEN), so we
       * don't step on the header when we construct
       * the sll header.
       */
      if (bp < (u_char *)frame +
             TPACKET_ALIGN(handlep->tp_hdrlen) +
             sizeof(struct sockaddr_ll)) {
        snprintf(handle->errbuf, PCAP_ERRBUF_SIZE,
          "cooked-mode frame doesn't have room for sll header");
        return -1;
      }

      /*
       * OK, that worked; construct the sll header.
       */
      hdrp = (struct sll2_header *)bp;
      hdrp->sll2_protocol = sll->sll_protocol;
      hdrp->sll2_reserved_mbz = 0;
      hdrp->sll2_if_index = htonl(sll->sll_ifindex);
      hdrp->sll2_hatype = htons(sll->sll_hatype);
      hdrp->sll2_pkttype = sll->sll_pkttype;
      hdrp->sll2_halen = sll->sll_halen;
      memcpy(hdrp->sll2_addr, sll->sll_addr, SLL_ADDRLEN);

      snaplen += sizeof(struct sll2_header);
    } else {
      struct sll_header *hdrp;

      /*
       * The kernel should have left us with enough
       * space for an sll header; back up the packet
       * data pointer into that space, as that'll be
       * the beginning of the packet we pass to the
       * callback.
       */
      bp -= SLL_HDR_LEN;

      /*
       * Let's make sure that's past the end of
       * the tpacket header, i.e. >=
       * ((u_char *)thdr + TPACKET_HDRLEN), so we
       * don't step on the header when we construct
       * the sll header.
       */
      if (bp < (u_char *)frame +
             TPACKET_ALIGN(handlep->tp_hdrlen) +
             sizeof(struct sockaddr_ll)) {
        snprintf(handle->errbuf, PCAP_ERRBUF_SIZE,
          "cooked-mode frame doesn't have room for sll header");
        return -1;
      }

      /*
       * OK, that worked; construct the sll header.
       */
      hdrp = (struct sll_header *)bp;
      hdrp->sll_pkttype = htons(sll->sll_pkttype);
      hdrp->sll_hatype = htons(sll->sll_hatype);
      hdrp->sll_halen = htons(sll->sll_halen);
      memcpy(hdrp->sll_addr, sll->sll_addr, SLL_ADDRLEN);
      hdrp->sll_protocol = sll->sll_protocol;

      snaplen += sizeof(struct sll_header);
    }
  } else {
    /*
     * If this is a packet from a CAN device, so that
     * sll->sll_hatype is ARPHRD_CAN, then, as we're
     * not capturing in cooked mode, its link-layer
     * type is DLT_CAN_SOCKETCAN.  Fix up the header
     * provided by the code below us to match what
     * DLT_CAN_SOCKETCAN is expected to provide.
     */
    if (sll->sll_hatype == ARPHRD_CAN) {
      pcap_can_socketcan_hdr *canhdr = (pcap_can_socketcan_hdr *)bp;
      pcap_can_socketcan_xl_hdr *canxl_hdr = (pcap_can_socketcan_xl_hdr *)bp;
      uint16_t protocol = ntohs(sll->sll_protocol);

      /*
       * Check the protocol field from the sll header.
       * If it's one of the known CAN protocol types,
       * make sure the appropriate flags are set, so
       * that a program can tell what type of frame
       * it is.
       *
       * These operations should not have any effect
       * when reading proper CAN frames from Linux
       * CAN interfaces. Enforcing these bit values
       * ensures proper DLT_CAN_SOCKETCAN data even
       * with malformed PF_PACKET content.
       *
       * The two flags are:
       *
       *   CANFD_FDF, which is in the fd_flags field
       *   of the CAN CC/CAN FD header;
       *
       *   CANXL_XLF, which is in the flags field
       *   of the CAN XL header, which overlaps
       *   the payload_length field of the CAN CC/
       *   CAN FD header. Setting CANXL_XLF in the
       *   payload_length of CAN CC/FD frames would
       *   intentionally break the payload length.
       */
      switch (protocol) {

      case LINUX_SLL_P_CAN:
        /*
         * CAN CC frame (aka Classical CAN, CAN 2.0B)
         *
         * Zero out the CAN FD and CAN XL flags
         * so that this frame will be identified
         * as a CAN CC frame.
         */
        canxl_hdr->flags &= ~CANXL_XLF;
        canhdr->fd_flags &= ~CANFD_FDF;
        break;

      case LINUX_SLL_P_CANFD:
        /*
         * CAN FD frame
         *
         * Set CANFD_FDF in the fd_flags field,
         * and clear the CANXL_XLF bit in the
         * CAN XL flags field, so that this frame
         * will be identified as a CAN FD frame.
         *
         * The CANFD_FDF bit is not reliably
         * set by the Linux kernel. But setting
         * that bit for CAN FD is recommended.
         */
        canxl_hdr->flags &= ~CANXL_XLF;
        canhdr->fd_flags |= CANFD_FDF;
        break;

      case LINUX_SLL_P_CANXL:
        /*
         * CAN XL frame
         *
         * Set CANXL_XLF bit in the CAN XL flags
         * field, so that this frame will appear
         * to be a CAN XL frame.
         */
        canxl_hdr->flags |= CANXL_XLF;
        break;
      }

      /*
       * Put multi-byte header fields in a byte-order
       * -independent format.
       */
      if (canxl_hdr->flags & CANXL_XLF) {
        /*
         * This is a CAN XL frame.
         *
         * DLT_CAN_SOCKETCAN is specified as having
         * the Priority ID/VCID field in big-
         * endian byte order, and the payload length
         * and Acceptance Field in little-endian byte
         * order, but capturing on a CAN device
         * provides them in host byte order.
         * Convert them to the appropriate byte
         * orders.
         *
         * The reason we put the first field
         * into big-endian byte order is that
         * older libpcap code, ignorant of
         * CAN XL, treated it as the CAN ID
         * field and put it into big-endian
         * byte order, and we don't want to
         * break code that understands CAN XL
         * headers, and treats that field as
         * being big-endian.
         *
         * The reason other fields are put in little-
         * endian byte order is that older
         * libpcap code, ignorant of CAN XL,
         * left those fields alone, and the
         * processors on which the CAN XL
         * frames were captured are likely
         * to be little-endian processors.
         */

#if __BYTE_ORDER == __LITTLE_ENDIAN
        /*
         * We're capturing on a little-endian
         * machine, so we put the priority/VCID
         * field into big-endian byte order, and
         * leave the payload length and acceptance
         * field in little-endian byte order.
         */
        /* Byte-swap priority/VCID. */
        canxl_hdr->priority_vcid = PCAP_BSWAP_32(canxl_hdr->priority_vcid);
#elif __BYTE_ORDER == __BIG_ENDIAN
        /*
         * We're capturing on a big-endian
         * machine, so we want to leave the
         * priority/VCID field alone, and byte-swap
         * the payload length and acceptance
         * fields to little-endian.
         */
        /* Byte-swap the payload length */
        canxl_hdr->payload_length = PCAP_BSWAP_16(canxl_hdr->payload_length);

        /*
         * Byte-swap the acceptance field.
         *
         * XXX - is it just a 4-octet string,
         * not in any byte order?
         */
        canxl_hdr->acceptance_field = PCAP_BSWAP_32(canxl_hdr->acceptance_field);
#else
#error "Unknown byte order"
#endif
      } else {
        /*
         * CAN CC or CAN FD frame.
         *
         * DLT_CAN_SOCKETCAN is specified as having
         * the CAN ID and flags in network byte
         * order, but capturing on a CAN device
         * provides it in host byte order.  Convert
         * it to network byte order.
         */
        canhdr->can_id = htonl(canhdr->can_id);
      }
    }
  }

  if (handlep->filter_in_userland && handle->fcode.bf_insns) {
    struct pcap_bpf_aux_data aux_data;

    aux_data.vlan_tag_present = tp_vlan_tci_valid;
    aux_data.vlan_tag = tp_vlan_tci & 0x0fff;

    if (pcapint_filter_with_aux_data(handle->fcode.bf_insns,
                handle->fcode.bf_len,
                bp,
                tp_len,
                snaplen,
                &aux_data) == 0)
      return 0;
  }

  if (!linux_check_direction(handle, sll))
    return 0;

  /*
   * Get required packet info from ring header.
   *
   * The seconds part of the time stamp is a 32-bit
   * unsigned integer; this will have a problem in 2106,
   * but not in 2038.
   *
   * ts.tv_sec is a time_t, which is signed, and which
   * may be 32-bit or 64-bit.  Pass it through; if we
   * have a 32-bit signed time_t, in which values >
   * 2^31-1 won't fit, then:
   *
   *    Writing the packet to a file will pass the bits
   *    through.  If the program reading the file can
   *    handle 32-bit unsigned time stamps, including
   *    any conversion to local time or UTC, it will
   *    properly handle the time stamps.
   *
   *    Reporting the packet time stamp may give
   *    an error or a pre-1970 time stamp on platforms
   *    with signed 32-bit time stamps, but that
   *    will happen even if it's captured on a
   *    platform with a 64-bit time_t.
   */
  pcaphdr.ts.tv_sec = tp_sec;
  pcaphdr.ts.tv_usec = tp_usec;
  pcaphdr.caplen = tp_snaplen;
  pcaphdr.len = tp_len;

  /* if required build in place the sll header*/
  if (handlep->cooked) {
    /* update packet len */
    if (handle->linktype == DLT_LINUX_SLL2) {
      pcaphdr.caplen += SLL2_HDR_LEN;
      pcaphdr.len += SLL2_HDR_LEN;
    } else {
      pcaphdr.caplen += SLL_HDR_LEN;
      pcaphdr.len += SLL_HDR_LEN;
    }
  }

  if (tp_vlan_tci_valid &&
    handlep->vlan_offset != -1 &&
    tp_snaplen >= (unsigned int) handlep->vlan_offset)
  {
    struct vlan_tag *tag;

    /*
     * Move everything in the header, except the type field,
     * down VLAN_TAG_LEN bytes, to allow us to insert the
     * VLAN tag between that stuff and the type field.
     */
    bp -= VLAN_TAG_LEN;
    memmove(bp, bp + VLAN_TAG_LEN, handlep->vlan_offset);

    /*
     * Now insert the tag.
     */
    tag = (struct vlan_tag *)(bp + handlep->vlan_offset);
    tag->vlan_tpid = htons(tp_vlan_tpid);
    tag->vlan_tci = htons(tp_vlan_tci);

    /*
     * Add the tag to the packet lengths.
     */
    pcaphdr.caplen += VLAN_TAG_LEN;
    pcaphdr.len += VLAN_TAG_LEN;
  }

  /*
   * The only way to tell the kernel to cut off the
   * packet at a snapshot length is with a filter program;
   * if there's no filter program, the kernel won't cut
   * the packet off.
   *
   * Trim the snapshot length to be no longer than the
   * specified snapshot length.
   *
   * XXX - an alternative is to put a filter, consisting
   * of a "ret <snaplen>" instruction, on the socket
   * in the activate routine, so that the truncation is
   * done in the kernel even if nobody specified a filter;
   * that means that less buffer space is consumed in
   * the memory-mapped buffer.
   */
  if (pcaphdr.caplen > (bpf_u_int32)handle->snapshot)
    pcaphdr.caplen = handle->snapshot;

  /* pass the packet to the user */
  callback(user, &pcaphdr, bp);

  return 1;
}

static int
pcap_read_linux_mmap_v2(pcap_t *handle, int max_packets, pcap_handler callback,
    u_char *user)
{
  struct pcap_linux *handlep = handle->priv;
  union thdr h;
  int pkts = 0;
  int ret;

  /* wait for frames availability.*/
  h.raw = RING_GET_CURRENT_FRAME(handle);
  if (!packet_mmap_acquire(h.h2)) {
    /*
     * The current frame is owned by the kernel; wait for
     * a frame to be handed to us.
     */
    ret = pcap_wait_for_frames_mmap(handle);
    if (ret) {
      return ret;
    }
  }

  /*
   * This can conceivably process more than INT_MAX packets,
   * which would overflow the packet count, causing it either
   * to look like a negative number, and thus cause us to
   * return a value that looks like an error, or overflow
   * back into positive territory, and thus cause us to
   * return a too-low count.
   *
   * Therefore, if the packet count is unlimited, we clip
   * it at INT_MAX; this routine is not expected to
   * process packets indefinitely, so that's not an issue.
   */
  if (PACKET_COUNT_IS_UNLIMITED(max_packets))
    max_packets = INT_MAX;

  while (pkts < max_packets) {
    /*
     * Get the current ring buffer frame, and break if
     * it's still owned by the kernel.
     */
    h.raw = RING_GET_CURRENT_FRAME(handle);
    if (!packet_mmap_acquire(h.h2))
      break;

    ret = pcap_handle_packet_mmap(
        handle,
        callback,
        user,
        h.raw,
        h.h2->tp_len,
        h.h2->tp_mac,
        h.h2->tp_snaplen,
        h.h2->tp_sec,
        handle->opt.tstamp_precision == PCAP_TSTAMP_PRECISION_NANO ? h.h2->tp_nsec : h.h2->tp_nsec / 1000,
        VLAN_VALID(h.h2, h.h2),
        h.h2->tp_vlan_tci,
        VLAN_TPID(h.h2, h.h2));
    if (ret == 1) {
      pkts++;
    } else if (ret < 0) {
      return ret;
    }

    /*
     * Hand this block back to the kernel, and, if we're
     * counting blocks that need to be filtered in userland
     * after having been filtered by the kernel, count
     * the one we've just processed.
     */
    packet_mmap_release(h.h2);
    if (handlep->blocks_to_filter_in_userland != 0) {
      handlep->blocks_to_filter_in_userland--;
      if (handlep->blocks_to_filter_in_userland == 0) {
        /*
         * No more blocks need to be filtered
         * in userland.
         */
        handlep->filter_in_userland = 0;
      }
    }

    /* next block */
    if (++handle->offset >= handle->cc)
      handle->offset = 0;

    /* check for break loop condition*/
    if (handle->break_loop) {
      handle->break_loop = 0;
      return PCAP_ERROR_BREAK;
    }
  }
  return pkts;
}

#ifdef HAVE_TPACKET3
static int
pcap_read_linux_mmap_v3(pcap_t *handle, int max_packets, pcap_handler callback,
    u_char *user)
{
  struct pcap_linux *handlep = handle->priv;
  union thdr h;
  int pkts = 0;
  int ret;

again:
  if (handlep->current_packet == NULL) {
    /* wait for frames availability.*/
    h.raw = RING_GET_CURRENT_FRAME(handle);
    if (!packet_mmap_v3_acquire(h.h3)) {
      /*
       * The current frame is owned by the kernel; wait
       * for a frame to be handed to us.
       */
      ret = pcap_wait_for_frames_mmap(handle);
      if (ret) {
        return ret;
      }
    }
  }
  h.raw = RING_GET_CURRENT_FRAME(handle);
  if (!packet_mmap_v3_acquire(h.h3)) {
    if (pkts == 0 && handlep->timeout == 0) {
      /* Block until we see a packet. */
      goto again;
    }
    return pkts;
  }

  /*
   * This can conceivably process more than INT_MAX packets,
   * which would overflow the packet count, causing it either
   * to look like a negative number, and thus cause us to
   * return a value that looks like an error, or overflow
   * back into positive territory, and thus cause us to
   * return a too-low count.
   *
   * Therefore, if the packet count is unlimited, we clip
   * it at INT_MAX; this routine is not expected to
   * process packets indefinitely, so that's not an issue.
   */
  if (PACKET_COUNT_IS_UNLIMITED(max_packets))
    max_packets = INT_MAX;

  while (pkts < max_packets) {
    int packets_to_read;

    if (handlep->current_packet == NULL) {
      h.raw = RING_GET_CURRENT_FRAME(handle);
      if (!packet_mmap_v3_acquire(h.h3))
        break;

      handlep->current_packet = h.raw + h.h3->hdr.bh1.offset_to_first_pkt;
      handlep->packets_left = h.h3->hdr.bh1.num_pkts;
    }
    packets_to_read = handlep->packets_left;

    if (packets_to_read > (max_packets - pkts)) {
      /*
       * There are more packets in the buffer than
       * the number of packets we have left to
       * process to get up to the maximum number
       * of packets to process.  Only process enough
       * of them to get us up to that maximum.
       */
      packets_to_read = max_packets - pkts;
    }

    while (packets_to_read-- && !handle->break_loop) {
      struct tpacket3_hdr* tp3_hdr = (struct tpacket3_hdr*) handlep->current_packet;
      ret = pcap_handle_packet_mmap(
          handle,
          callback,
          user,
          handlep->current_packet,
          tp3_hdr->tp_len,
          tp3_hdr->tp_mac,
          tp3_hdr->tp_snaplen,
          tp3_hdr->tp_sec,
          handle->opt.tstamp_precision == PCAP_TSTAMP_PRECISION_NANO ? tp3_hdr->tp_nsec : tp3_hdr->tp_nsec / 1000,
          VLAN_VALID(tp3_hdr, &tp3_hdr->hv1),
          tp3_hdr->hv1.tp_vlan_tci,
          VLAN_TPID(tp3_hdr, &tp3_hdr->hv1));
      if (ret == 1) {
        pkts++;
      } else if (ret < 0) {
        handlep->current_packet = NULL;
        return ret;
      }
      handlep->current_packet += tp3_hdr->tp_next_offset;
      handlep->packets_left--;
    }

    if (handlep->packets_left <= 0) {
      /*
       * Hand this block back to the kernel, and, if
       * we're counting blocks that need to be
       * filtered in userland after having been
       * filtered by the kernel, count the one we've
       * just processed.
       */
      packet_mmap_v3_release(h.h3);
      if (handlep->blocks_to_filter_in_userland != 0) {
        handlep->blocks_to_filter_in_userland--;
        if (handlep->blocks_to_filter_in_userland == 0) {
          /*
           * No more blocks need to be filtered
           * in userland.
           */
          handlep->filter_in_userland = 0;
        }
      }

      /* next block */
      if (++handle->offset >= handle->cc)
        handle->offset = 0;

      handlep->current_packet = NULL;
    }

    /* check for break loop condition*/
    if (handle->break_loop) {
      handle->break_loop = 0;
      return PCAP_ERROR_BREAK;
    }
  }
  if (pkts == 0 && handlep->timeout == 0) {
    /* Block until we see a packet. */
    goto again;
  }
  return pkts;
}
#endif /* HAVE_TPACKET3 */

/*
 *  Attach the given BPF code to the packet capture device.
 */
static int
pcap_setfilter_linux(pcap_t *handle, struct bpf_program *filter)
{
  struct pcap_linux *handlep;
  struct sock_fprog fcode;
  int     can_filter_in_kernel;
  int     err = 0;
  u_int     n, offset;

  if (!handle)
    return -1;
  if (!filter) {
          pcapint_strlcpy(handle->errbuf, "setfilter: No filter specified",
      PCAP_ERRBUF_SIZE);
    return -1;
  }

  handlep = handle->priv;

  /* Make our private copy of the filter */

  if (pcapint_install_bpf_program(handle, filter) < 0)
    /* pcapint_install_bpf_program() filled in errbuf */
    return -1;

  /*
   * Run user level packet filter by default. Will be overridden if
   * installing a kernel filter succeeds.
   */
  handlep->filter_in_userland = 1;

  /* Install kernel level filter if possible */

  if (handle->fcode.bf_len > USHRT_MAX) {
    /*
     * fcode.len is an unsigned short for current kernel.
     * I have yet to see BPF-Code with that much
     * instructions but still it is possible. So for the
     * sake of correctness I added this check.
     */
    fprintf(stderr, "Warning: Filter too complex for kernel\n");
    fcode.len = 0;
    fcode.filter = NULL;
    can_filter_in_kernel = 0;
  } else {
    /*
     * Oh joy, the Linux kernel uses struct sock_fprog instead
     * of struct bpf_program and of course the length field is
     * of different size. Pointed out by Sebastian
     *
     * Oh, and we also need to fix it up so that all "ret"
     * instructions with non-zero operands have MAXIMUM_SNAPLEN
     * as the operand if we're not capturing in memory-mapped
     * mode, and so that, if we're in cooked mode, all memory-
     * reference instructions use special magic offsets in
     * references to the link-layer header and assume that the
     * link-layer payload begins at 0; "fix_program()" will do
     * that.
     */
    switch (fix_program(handle, &fcode)) {

    case -1:
    default:
      /*
       * Fatal error; just quit.
       * (The "default" case shouldn't happen; we
       * return -1 for that reason.)
       */
      return -1;

    case 0:
      /*
       * The program performed checks that we can't make
       * work in the kernel.
       */
      can_filter_in_kernel = 0;
      break;

    case 1:
      /*
       * We have a filter that'll work in the kernel.
       */
      can_filter_in_kernel = 1;
      break;
    }
  }

  /*
   * NOTE: at this point, we've set both the "len" and "filter"
   * fields of "fcode".  As of the 2.6.32.4 kernel, at least,
   * those are the only members of the "sock_fprog" structure,
   * so we initialize every member of that structure.
   *
   * If there is anything in "fcode" that is not initialized,
   * it is either a field added in a later kernel, or it's
   * padding.
   *
   * If a new field is added, this code needs to be updated
   * to set it correctly.
   *
   * If there are no other fields, then:
   *
   *  if the Linux kernel looks at the padding, it's
   *  buggy;
   *
   *  if the Linux kernel doesn't look at the padding,
   *  then if some tool complains that we're passing
   *  uninitialized data to the kernel, then the tool
   *  is buggy and needs to understand that it's just
   *  padding.
   */
  if (can_filter_in_kernel) {
    if ((err = set_kernel_filter(handle, &fcode)) == 0)
    {
      /*
       * Installation succeeded - using kernel filter,
       * so userland filtering not needed.
       */
      handlep->filter_in_userland = 0;
    }
    else if (err == -1) /* Non-fatal error */
    {
      /*
       * Print a warning if we weren't able to install
       * the filter for a reason other than "this kernel
       * isn't configured to support socket filters.
       */
      if (errno == ENOMEM) {
        /*
         * Either a kernel memory allocation
         * failure occurred, or there's too
         * much "other/option memory" allocated
         * for this socket.  Suggest that they
         * increase the "other/option memory"
         * limit.
         */
        fprintf(stderr,
            "Warning: Couldn't allocate kernel memory for filter: try increasing net.core.optmem_max with sysctl\n");
      } else if (errno != ENOPROTOOPT && errno != EOPNOTSUPP) {
        fprintf(stderr,
            "Warning: Kernel filter failed: %s\n",
          pcap_strerror(errno));
      }
    }
  }

  /*
   * If we're not using the kernel filter, get rid of any kernel
   * filter that might've been there before, e.g. because the
   * previous filter could work in the kernel, or because some other
   * code attached a filter to the socket by some means other than
   * calling "pcap_setfilter()".  Otherwise, the kernel filter may
   * filter out packets that would pass the new userland filter.
   */
  if (handlep->filter_in_userland) {
    if (reset_kernel_filter(handle) == -1) {
      pcapint_fmt_errmsg_for_errno(handle->errbuf,
          PCAP_ERRBUF_SIZE, errno,
          "can't remove kernel filter");
      err = -2; /* fatal error */
    }
  }

  /*
   * Free up the copy of the filter that was made by "fix_program()".
   */
  if (fcode.filter != NULL)
    free(fcode.filter);

  if (err == -2)
    /* Fatal error */
    return -1;

  /*
   * If we're filtering in userland, there's nothing to do;
   * the new filter will be used for the next packet.
   */
  if (handlep->filter_in_userland)
    return 0;

  /*
   * We're filtering in the kernel; the packets present in
   * all blocks currently in the ring were already filtered
   * by the old filter, and so will need to be filtered in
   * userland by the new filter.
   *
   * Get an upper bound for the number of such blocks; first,
   * walk the ring backward and count the free blocks.
   */
  offset = handle->offset;
  if (offset == 0)
    offset = handle->cc;
  offset--;
  for (n=0; n < handle->cc; ++n) {
    if (offset == 0)
      offset = handle->cc;
    offset--;
    if (pcap_get_ring_frame_status(handle, offset) != TP_STATUS_KERNEL)
      break;
  }

  /*
   * If we found free blocks, decrement the count of free
   * blocks by 1, just in case we lost a race with another
   * thread of control that was adding a packet while
   * we were counting and that had run the filter before
   * we changed it.
   *
   * XXX - could there be more than one block added in
   * this fashion?
   *
   * XXX - is there a way to avoid that race, e.g. somehow
   * wait for all packets that passed the old filter to
   * be added to the ring?
   */
  if (n != 0)
    n--;

  /*
   * Set the count of blocks worth of packets to filter
   * in userland to the total number of blocks in the
   * ring minus the number of free blocks we found, and
   * turn on userland filtering.  (The count of blocks
   * worth of packets to filter in userland is guaranteed
   * not to be zero - n, above, couldn't be set to a
   * value > handle->cc, and if it were equal to
   * handle->cc, it wouldn't be zero, and thus would
   * be decremented to handle->cc - 1.)
   */
  handlep->blocks_to_filter_in_userland = handle->cc - n;
  handlep->filter_in_userland = 1;

  return 0;
}

/*
 *  Return the index of the given device name. Fill ebuf and return
 *  -1 on failure.
 */
static int
iface_get_id(int fd, const char *device, char *ebuf)
{
  struct ifreq  ifr;

  memset(&ifr, 0, sizeof(ifr));
  pcapint_strlcpy(ifr.ifr_name, device, sizeof(ifr.ifr_name));

  if (ioctl(fd, SIOCGIFINDEX, &ifr) == -1) {
    pcapint_fmt_errmsg_for_errno(ebuf, PCAP_ERRBUF_SIZE,
        errno, "SIOCGIFINDEX");
    return -1;
  }

  return ifr.ifr_ifindex;
}

/*
 *  Bind the socket associated with FD to the given device.
 *  Return 0 on success or a PCAP_ERROR_ value on a hard error.
 */
static int
iface_bind(int fd, int ifindex, char *ebuf, int protocol)
{
  struct sockaddr_ll  sll;
  int     ret, err;
  socklen_t   errlen = sizeof(err);

  memset(&sll, 0, sizeof(sll));
  sll.sll_family    = AF_PACKET;
  sll.sll_ifindex   = ifindex < 0 ? 0 : ifindex;
  sll.sll_protocol  = protocol;

  if (bind(fd, (struct sockaddr *) &sll, sizeof(sll)) == -1) {
    if (errno == ENETDOWN) {
      /*
       * Return a "network down" indication, so that
       * the application can report that rather than
       * saying we had a mysterious failure and
       * suggest that they report a problem to the
       * libpcap developers.
       */
      return PCAP_ERROR_IFACE_NOT_UP;
    }
    if (errno == ENODEV) {
      /*
       * There's nothing more to say, so clear the
       * error message.
       */
      ebuf[0] = '\0';
      ret = PCAP_ERROR_NO_SUCH_DEVICE;
    } else {
      ret = PCAP_ERROR;
      pcapint_fmt_errmsg_for_errno(ebuf, PCAP_ERRBUF_SIZE,
          errno, "bind");
    }
    return ret;
  }

  /* Any pending errors, e.g., network is down? */

  if (getsockopt(fd, SOL_SOCKET, SO_ERROR, &err, &errlen) == -1) {
    pcapint_fmt_errmsg_for_errno(ebuf, PCAP_ERRBUF_SIZE,
        errno, "getsockopt (SO_ERROR)");
    return PCAP_ERROR;
  }

  if (err == ENETDOWN) {
    /*
     * Return a "network down" indication, so that
     * the application can report that rather than
     * saying we had a mysterious failure and
     * suggest that they report a problem to the
     * libpcap developers.
     */
    return PCAP_ERROR_IFACE_NOT_UP;
  } else if (err > 0) {
    pcapint_fmt_errmsg_for_errno(ebuf, PCAP_ERRBUF_SIZE,
        err, "bind");
    return PCAP_ERROR;
  }

  return 0;
}

/*
 * Try to enter monitor mode.
 * If we have libnl, try to create a new monitor-mode device and
 * capture on that; otherwise, just say "not supported".
 */
#ifdef HAVE_LIBNL
static int
enter_rfmon_mode(pcap_t *handle, int sock_fd, const char *device)
{
  struct pcap_linux *handlep = handle->priv;
  int ret;
  char phydev_path[PATH_MAX+1];
  struct nl80211_state nlstate;
  struct ifreq ifr;
  u_int n;

  /*
   * Is this a mac80211 device?
   */
  ret = get_mac80211_phydev(handle, device, phydev_path, PATH_MAX);
  if (ret < 0)
    return ret; /* error */
  if (ret == 0)
    return 0; /* no error, but not mac80211 device */

  ret = nl80211_init(handle, &nlstate, device);
  if (ret != 0)
    return ret;

  /*
   * Is this already a monN device?
   * If so, we're done.
   */
  int type;
  ret = get_if_type(handle, sock_fd, &nlstate, device, &type);
  if (ret <= 0) {
    /*
     * < 0 is a Hard failure.  Just return ret; handle->errbuf
     * has already been set.
     *
     * 0 is "device not available"; the caller should retry later.
     */
    nl80211_cleanup(&nlstate);
    return ret;
  }
  if (type == NL80211_IFTYPE_MONITOR) {
    /*
     * OK, it's already a monitor mode device; just use it.
     * There's no point in creating another monitor device
     * that will have to be cleaned up.
     */
    nl80211_cleanup(&nlstate);
    return ret;
  }

  /*
   * OK, it's apparently a mac80211 device but not a monitor device.
   * Try to find an unused monN device for it.
   */
  for (n = 0; n < UINT_MAX; n++) {
    /*
     * Try mon{n}.
     */
    char mondevice[3+10+1]; /* mon{UINT_MAX}\0 */

    snprintf(mondevice, sizeof mondevice, "mon%u", n);
    ret = add_mon_if(handle, sock_fd, &nlstate, device, mondevice);
    if (ret == 1) {
      /*
       * Success.  We don't clean up the libnl state
       * yet, as we'll be using it later.
       */
      goto added;
    }
    if (ret < 0) {
      /*
       * Hard failure.  Just return ret; handle->errbuf
       * has already been set.
       */
      nl80211_cleanup(&nlstate);
      return ret;
    }
  }

  snprintf(handle->errbuf, PCAP_ERRBUF_SIZE,
      "%s: No free monN interfaces", device);
  nl80211_cleanup(&nlstate);
  return PCAP_ERROR;

added:

#if 0
  /*
   * Sleep for .1 seconds.
   */
  delay.tv_sec = 0;
  delay.tv_nsec = 500000000;
  nanosleep(&delay, NULL);
#endif

  /*
   * If we haven't already done so, arrange to have
   * "pcap_close_all()" called when we exit.
   */
  if (!pcapint_do_addexit(handle)) {
    /*
     * "atexit()" failed; don't put the interface
     * in rfmon mode, just give up.
     * handle->errbuf has already been filled.
     */
    del_mon_if(handle, sock_fd, &nlstate, device,
        handlep->mondevice);
    nl80211_cleanup(&nlstate);
    return PCAP_ERROR;
  }

  /*
   * Now configure the monitor interface up.
   */
  memset(&ifr, 0, sizeof(ifr));
  pcapint_strlcpy(ifr.ifr_name, handlep->mondevice, sizeof(ifr.ifr_name));
  if (ioctl(sock_fd, SIOCGIFFLAGS, &ifr) == -1) {
    pcapint_fmt_errmsg_for_errno(handle->errbuf, PCAP_ERRBUF_SIZE,
        errno, "%s: Can't get flags for %s", device,
        handlep->mondevice);
    del_mon_if(handle, sock_fd, &nlstate, device,
        handlep->mondevice);
    nl80211_cleanup(&nlstate);
    return PCAP_ERROR;
  }
  ifr.ifr_flags |= IFF_UP|IFF_RUNNING;
  if (ioctl(sock_fd, SIOCSIFFLAGS, &ifr) == -1) {
    pcapint_fmt_errmsg_for_errno(handle->errbuf, PCAP_ERRBUF_SIZE,
        errno, "%s: Can't set flags for %s", device,
        handlep->mondevice);
    del_mon_if(handle, sock_fd, &nlstate, device,
        handlep->mondevice);
    nl80211_cleanup(&nlstate);
    return PCAP_ERROR;
  }

  /*
   * Success.  Clean up the libnl state.
   */
  nl80211_cleanup(&nlstate);

  /*
   * Note that we have to delete the monitor device when we close
   * the handle.
   */
  handlep->must_do_on_close |= MUST_DELETE_MONIF;

  /*
   * Add this to the list of pcaps to close when we exit.
   */
  pcapint_add_to_pcaps_to_close(handle);

  return 1;
}
#else /* HAVE_LIBNL */
static int
enter_rfmon_mode(pcap_t *handle _U_, int sock_fd _U_, const char *device _U_)
{
  /*
   * We don't have libnl, so we can't do monitor mode.
   */
  return 0;
}
#endif /* HAVE_LIBNL */

#if defined(HAVE_LINUX_NET_TSTAMP_H) && defined(PACKET_TIMESTAMP)
/*
 * Map SOF_TIMESTAMPING_ values to PCAP_TSTAMP_ values.
 */
static const struct {
  int soft_timestamping_val;
  int pcap_tstamp_val;
} sof_ts_type_map[3] = {
  { SOF_TIMESTAMPING_SOFTWARE, PCAP_TSTAMP_HOST },
  { SOF_TIMESTAMPING_SYS_HARDWARE, PCAP_TSTAMP_ADAPTER },
  { SOF_TIMESTAMPING_RAW_HARDWARE, PCAP_TSTAMP_ADAPTER_UNSYNCED }
};
#define NUM_SOF_TIMESTAMPING_TYPES  (sizeof sof_ts_type_map / sizeof sof_ts_type_map[0])

/*
 * Set the list of time stamping types to include all types.
 */
static int
iface_set_all_ts_types(pcap_t *handle, char *ebuf)
{
  u_int i;

  handle->tstamp_type_list = malloc(NUM_SOF_TIMESTAMPING_TYPES * sizeof(u_int));
  if (handle->tstamp_type_list == NULL) {
    pcapint_fmt_errmsg_for_errno(ebuf, PCAP_ERRBUF_SIZE,
        errno, "malloc");
    return -1;
  }
  for (i = 0; i < NUM_SOF_TIMESTAMPING_TYPES; i++)
    handle->tstamp_type_list[i] = sof_ts_type_map[i].pcap_tstamp_val;
  handle->tstamp_type_count = NUM_SOF_TIMESTAMPING_TYPES;
  return 0;
}

/*
 * Get a list of time stamp types.
 */
#ifdef ETHTOOL_GET_TS_INFO
static int
iface_get_ts_types(const char *device, pcap_t *handle, char *ebuf)
{
  int fd;
  struct ifreq ifr;
  struct ethtool_ts_info info;
  int num_ts_types;
  u_int i, j;

  /*
   * This doesn't apply to the "any" device; you can't say "turn on
   * hardware time stamping for all devices that exist now and arrange
   * that it be turned on for any device that appears in the future",
   * and not all devices even necessarily *support* hardware time
   * stamping, so don't report any time stamp types.
   */
  if (strcmp(device, "any") == 0) {
    handle->tstamp_type_list = NULL;
    return 0;
  }

  /*
   * Create a socket from which to fetch time stamping capabilities.
   */
  fd = get_if_ioctl_socket();
  if (fd < 0) {
    pcapint_fmt_errmsg_for_errno(ebuf, PCAP_ERRBUF_SIZE,
        errno, "socket for SIOCETHTOOL(ETHTOOL_GET_TS_INFO)");
    return -1;
  }

  memset(&ifr, 0, sizeof(ifr));
  pcapint_strlcpy(ifr.ifr_name, device, sizeof(ifr.ifr_name));
  memset(&info, 0, sizeof(info));
  info.cmd = ETHTOOL_GET_TS_INFO;
  ifr.ifr_data = (caddr_t)&info;
  if (ioctl(fd, SIOCETHTOOL, &ifr) == -1) {
    int save_errno = errno;

    close(fd);
    switch (save_errno) {

    case EOPNOTSUPP:
    case EINVAL:
      /*
       * OK, this OS version or driver doesn't support
       * asking for the time stamping types, so let's
       * just return all the possible types.
       */
      if (iface_set_all_ts_types(handle, ebuf) == -1)
        return -1;
      return 0;

    case ENODEV:
      /*
       * OK, no such device.
       * The user will find that out when they try to
       * activate the device; just return an empty
       * list of time stamp types.
       */
      handle->tstamp_type_list = NULL;
      return 0;

    default:
      /*
       * Other error.
       */
      pcapint_fmt_errmsg_for_errno(ebuf, PCAP_ERRBUF_SIZE,
          save_errno,
          "%s: SIOCETHTOOL(ETHTOOL_GET_TS_INFO) ioctl failed",
          device);
      return -1;
    }
  }
  close(fd);

  /*
   * Do we support hardware time stamping of *all* packets?
   */
  if (!(info.rx_filters & (1 << HWTSTAMP_FILTER_ALL))) {
    /*
     * No, so don't report any time stamp types.
     *
     * XXX - some devices either don't report
     * HWTSTAMP_FILTER_ALL when they do support it, or
     * report HWTSTAMP_FILTER_ALL but map it to only
     * time stamping a few PTP packets.  See
     * http://marc.info/?l=linux-netdev&m=146318183529571&w=2
     *
     * Maybe that got fixed later.
     */
    handle->tstamp_type_list = NULL;
    return 0;
  }

  num_ts_types = 0;
  for (i = 0; i < NUM_SOF_TIMESTAMPING_TYPES; i++) {
    if (info.so_timestamping & sof_ts_type_map[i].soft_timestamping_val)
      num_ts_types++;
  }
  if (num_ts_types != 0) {
    handle->tstamp_type_list = malloc(num_ts_types * sizeof(u_int));
    if (handle->tstamp_type_list == NULL) {
      pcapint_fmt_errmsg_for_errno(ebuf, PCAP_ERRBUF_SIZE,
          errno, "malloc");
      return -1;
    }
    for (i = 0, j = 0; i < NUM_SOF_TIMESTAMPING_TYPES; i++) {
      if (info.so_timestamping & sof_ts_type_map[i].soft_timestamping_val) {
        handle->tstamp_type_list[j] = sof_ts_type_map[i].pcap_tstamp_val;
        j++;
      }
    }
    handle->tstamp_type_count = num_ts_types;
  } else
    handle->tstamp_type_list = NULL;

  return 0;
}
#else /* ETHTOOL_GET_TS_INFO */
static int
iface_get_ts_types(const char *device, pcap_t *handle, char *ebuf)
{
  /*
   * This doesn't apply to the "any" device; you can't say "turn on
   * hardware time stamping for all devices that exist now and arrange
   * that it be turned on for any device that appears in the future",
   * and not all devices even necessarily *support* hardware time
   * stamping, so don't report any time stamp types.
   */
  if (strcmp(device, "any") == 0) {
    handle->tstamp_type_list = NULL;
    return 0;
  }

  /*
   * We don't have an ioctl to use to ask what's supported,
   * so say we support everything.
   */
  if (iface_set_all_ts_types(handle, ebuf) == -1)
    return -1;
  return 0;
}
#endif /* ETHTOOL_GET_TS_INFO */
#else  /* defined(HAVE_LINUX_NET_TSTAMP_H) && defined(PACKET_TIMESTAMP) */
static int
iface_get_ts_types(const char *device _U_, pcap_t *p _U_, char *ebuf _U_)
{
  /*
   * Nothing to fetch, so it always "succeeds".
   */
  return 0;
}
#endif /* defined(HAVE_LINUX_NET_TSTAMP_H) && defined(PACKET_TIMESTAMP) */

/*
 * Find out if we have any form of fragmentation/reassembly offloading.
 *
 * We do so using SIOCETHTOOL checking for various types of offloading;
 * if SIOCETHTOOL isn't defined, or we don't have any #defines for any
 * of the types of offloading, there's nothing we can do to check, so
 * we just say "no, we don't".
 *
 * We treat EOPNOTSUPP, EINVAL and, if eperm_ok is true, EPERM as
 * indications that the operation isn't supported.  We do EPERM
 * weirdly because the SIOCETHTOOL code in later kernels 1) doesn't
 * support ETHTOOL_GUFO, 2) also doesn't include it in the list
 * of ethtool operations that don't require CAP_NET_ADMIN privileges,
 * and 3) does the "is this permitted" check before doing the "is
 * this even supported" check, so it fails with "this is not permitted"
 * rather than "this is not even supported".  To work around this
 * annoyance, we only treat EPERM as an error for the first feature,
 * and assume that they all do the same permission checks, so if the
 * first one is allowed all the others are allowed if supported.
 */
#if defined(SIOCETHTOOL) && (defined(ETHTOOL_GTSO) || defined(ETHTOOL_GUFO) || defined(ETHTOOL_GGSO) || defined(ETHTOOL_GFLAGS) || defined(ETHTOOL_GGRO))
static int
iface_ethtool_flag_ioctl(pcap_t *handle, int cmd, const char *cmdname,
    int eperm_ok)
{
  struct ifreq  ifr;
  struct ethtool_value eval;

  memset(&ifr, 0, sizeof(ifr));
  pcapint_strlcpy(ifr.ifr_name, handle->opt.device, sizeof(ifr.ifr_name));
  eval.cmd = cmd;
  eval.data = 0;
  ifr.ifr_data = (caddr_t)&eval;
  if (ioctl(handle->fd, SIOCETHTOOL, &ifr) == -1) {
    if (errno == EOPNOTSUPP || errno == EINVAL ||
        (errno == EPERM && eperm_ok)) {
      /*
       * OK, let's just return 0, which, in our
       * case, either means "no, what we're asking
       * about is not enabled" or "all the flags
       * are clear (i.e., nothing is enabled)".
       */
      return 0;
    }
    pcapint_fmt_errmsg_for_errno(handle->errbuf, PCAP_ERRBUF_SIZE,
        errno, "%s: SIOCETHTOOL(%s) ioctl failed",
        handle->opt.device, cmdname);
    return -1;
  }
  return eval.data;
}

/*
 * XXX - it's annoying that we have to check for offloading at all, but,
 * given that we have to, it's still annoying that we have to check for
 * particular types of offloading, especially that shiny new types of
 * offloading may be added - and, worse, may not be checkable with
 * a particular ETHTOOL_ operation; ETHTOOL_GFEATURES would, in
 * theory, give those to you, but the actual flags being used are
 * opaque (defined in a non-uapi header), and there doesn't seem to
 * be any obvious way to ask the kernel what all the offloading flags
 * are - at best, you can ask for a set of strings(!) to get *names*
 * for various flags.  (That whole mechanism appears to have been
 * designed for the sole purpose of letting ethtool report flags
 * by name and set flags by name, with the names having no semantics
 * ethtool understands.)
 */
static int
iface_get_offload(pcap_t *handle)
{
  int ret;

#ifdef ETHTOOL_GTSO
  ret = iface_ethtool_flag_ioctl(handle, ETHTOOL_GTSO, "ETHTOOL_GTSO", 0);
  if (ret == -1)
    return -1;
  if (ret)
    return 1; /* TCP segmentation offloading on */
#endif

#ifdef ETHTOOL_GGSO
  /*
   * XXX - will this cause large unsegmented packets to be
   * handed to PF_PACKET sockets on transmission?  If not,
   * this need not be checked.
   */
  ret = iface_ethtool_flag_ioctl(handle, ETHTOOL_GGSO, "ETHTOOL_GGSO", 0);
  if (ret == -1)
    return -1;
  if (ret)
    return 1; /* generic segmentation offloading on */
#endif

#ifdef ETHTOOL_GFLAGS
  ret = iface_ethtool_flag_ioctl(handle, ETHTOOL_GFLAGS, "ETHTOOL_GFLAGS", 0);
  if (ret == -1)
    return -1;
  if (ret & ETH_FLAG_LRO)
    return 1; /* large receive offloading on */
#endif

#ifdef ETHTOOL_GGRO
  /*
   * XXX - will this cause large reassembled packets to be
   * handed to PF_PACKET sockets on receipt?  If not,
   * this need not be checked.
   */
  ret = iface_ethtool_flag_ioctl(handle, ETHTOOL_GGRO, "ETHTOOL_GGRO", 0);
  if (ret == -1)
    return -1;
  if (ret)
    return 1; /* generic (large) receive offloading on */
#endif

#ifdef ETHTOOL_GUFO
  /*
   * Do this one last, as support for it was removed in later
   * kernels, and it fails with EPERM on those kernels rather
   * than with EOPNOTSUPP (see explanation in comment for
   * iface_ethtool_flag_ioctl()).
   */
  ret = iface_ethtool_flag_ioctl(handle, ETHTOOL_GUFO, "ETHTOOL_GUFO", 1);
  if (ret == -1)
    return -1;
  if (ret)
    return 1; /* UDP fragmentation offloading on */
#endif

  return 0;
}
#else /* SIOCETHTOOL */
static int
iface_get_offload(pcap_t *handle _U_)
{
  /*
   * XXX - do we need to get this information if we don't
   * have the ethtool ioctls?  If so, how do we do that?
   */
  return 0;
}
#endif /* SIOCETHTOOL */

/*
 * As per
 *
 *    https://www.kernel.org/doc/html/latest/networking/dsa/dsa.html#switch-tagging-protocols
 *
 * Type 1 means that the tag is prepended to the Ethernet packet.
 *
 * Type 2 means that the tag is inserted into the Ethernet header
 * after the source address and before the type/length field.
 *
 * Type 3 means that tag is a packet trailer.
 *
 * Every element in the array below uses a DLT.  Because a DSA-tagged frame is
 * not a standard IEEE 802.3 Ethernet frame, the array elements must not use
 * DLT_EN10MB.  It is safe, albeit only barely useful, to use DLT_DEBUG_ONLY,
 * which is also the implicit default for any DSA tag that is not present in
 * the array.  To implement proper support for a particular DSA tag of
 * interest, please do as much of the following as is reasonably practicable:
 *
 * 1. Using recent versions of tcpdump and libpcap on a Linux host with a
 *    network interface that implements the required DSA tag, capture packets
 *    on the interface and study the hex dumps.
 * 2. Using the hex dumps and any other available supporting materials, produce
 *    a sufficiently detailed description of the DSA tag structure, complete
 *    with a full comment indicating whether it's type 1, 2, or 3, and, for
 *    type 2, indicating whether it has an EtherType and, if so, what that type
 *    is, and whether it's registered with the IEEE or not.  Refer to the
 *    specification(s), existing implementation(s), or any other relevant
 *    resources.
 * 3. Using the description, request and obtain a new DLT for the DSA tag.
 * 4. Associate the new DLT with the DSA tag in the array below.
 * 5. Using the updated libpcap, capture packets again, produce a .pcap file
 *    and confirm it uses the new DLT.
 * 6. Using the .pcap file as a test, prepare additional changes to tcpdump to
 *    enable decoding of packets for the new DLT.
 * 7. Using the .pcap file as a test, prepare additional changes to libpcap to
 *    enable filtering of packets for the new DLT.
 *
 * For working examples of such support, see the existing DLTs other than
 * DLT_DEBUG_ONLY in the array below.
 */
static struct dsa_proto {
  const char *name;
  bpf_u_int32 linktype;
} dsa_protos[] = {
  /*
   * Type 1. See
   *
   *    https://elixir.bootlin.com/linux/v6.13.2/source/net/dsa/tag_ar9331.c
   */
  { "ar9331", DLT_DEBUG_ONLY },

  /*
   * Type 2, without an EtherType at the beginning.
   */
  { "brcm", DLT_DSA_TAG_BRCM },

  /*
   * Type 2, with EtherType 0x8874, assigned to Broadcom.
   */
  { "brcm-legacy", DLT_DEBUG_ONLY },

  /*
   * Type 1.
   */
  { "brcm-prepend", DLT_DSA_TAG_BRCM_PREPEND },

  /*
   * Type 2, without an EtherType at the beginning.
   */
  { "dsa", DLT_DSA_TAG_DSA },

  /*
   * Type 2, with an EtherType field, but without
   * an assigned EtherType value that can be relied
   * on.
   */
  { "edsa", DLT_DSA_TAG_EDSA },

  /*
   * Type 1, with different transmit and receive headers,
   * so can't really be handled well with the current
   * libpcap API and with pcap files.
   *
   * See
   *
   *    https://elixir.bootlin.com/linux/v6.13.2/source/net/dsa/tag_gswip.c
   */
  { "gswip", DLT_DEBUG_ONLY },

  /*
   * Type 3. See
   *
   *    https://elixir.bootlin.com/linux/v6.13.2/source/net/dsa/tag_hellcreek.c
   */
  { "hellcreek", DLT_DEBUG_ONLY },

  /*
   * Type 3, with different transmit and receive headers,
   * so can't really be handled well with the current
   * libpcap API and with pcap files.
   *
   * See
   *
   *    https://elixir.bootlin.com/linux/v6.13.2/source/net/dsa/tag_ksz.c#L102
   */
  { "ksz8795", DLT_DEBUG_ONLY },

  /*
   * Type 3, with different transmit and receive headers,
   * so can't really be handled well with the current
   * libpcap API and with pcap files.
   *
   * See
   *
   *    https://elixir.bootlin.com/linux/v6.13.2/source/net/dsa/tag_ksz.c#L160
   */
  { "ksz9477", DLT_DEBUG_ONLY },

  /*
   * Type 3, with different transmit and receive headers,
   * so can't really be handled well with the current
   * libpcap API and with pcap files.
   *
   * See
   *
   *    https://elixir.bootlin.com/linux/v6.13.2/source/net/dsa/tag_ksz.c#L341
   */
  { "ksz9893", DLT_DEBUG_ONLY },

  /*
   * Type 3, with different transmit and receive headers,
   * so can't really be handled well with the current
   * libpcap API and with pcap files.
   *
   * See
   *
   *    https://elixir.bootlin.com/linux/v6.13.2/source/net/dsa/tag_ksz.c#L386
   */
  { "lan937x", DLT_DEBUG_ONLY },

  /*
   * Type 2, with EtherType 0x8100; the VID can be interpreted
   * as per
   *
   *    https://elixir.bootlin.com/linux/v6.13.2/source/net/dsa/tag_lan9303.c#L24
   */
  { "lan9303", DLT_DEBUG_ONLY },

  /*
   * Type 2, without an EtherType at the beginning.
   *
   * See
   *
   *    https://elixir.bootlin.com/linux/v6.13.2/source/net/dsa/tag_mtk.c#L15
   *
   * Linux kernel implements this tag so that it does not indicate the frame
   * encoding reliably.  The matter is, some drivers use METADATA_HW_PORT_MUX,
   * which (for the switch->CPU direction only, at the time of this writing)
   * means that the frame does not have a DSA tag, the frame metadata is stored
   * elsewhere and libpcap receives the frame only.  Specifically, this is the
   * case for drivers/net/ethernet/mediatek/mtk_eth_soc.c, but the tag visible
   * in sysfs is still "mtk" even though the wire encoding is different.
   */
  { "mtk", DLT_DEBUG_ONLY },

  /*
   * Type 1.
   *
   * See
   *
   *    https://elixir.bootlin.com/linux/v6.13.2/source/net/dsa/tag_ocelot.c
   */
  { "ocelot", DLT_DEBUG_ONLY },

  /*
   * Type 1.
   *
   * See
   *
   *    https://elixir.bootlin.com/linux/v6.13.2/source/net/dsa/tag_ocelot.c
   */
  { "seville", DLT_DEBUG_ONLY },

  /*
   * Type 2, with EtherType 0x8100; the VID can be interpreted
   * as per
   *
   *    https://elixir.bootlin.com/linux/v6.13.2/source/net/dsa/tag_8021q.c#L15
   */
  { "ocelot-8021q", DLT_DEBUG_ONLY },

  /*
   * Type 2, without an EtherType at the beginning.
   *
   * See
   *
   *    https://elixir.bootlin.com/linux/v6.13.2/source/net/dsa/tag_qca.c
   */
  { "qca", DLT_DEBUG_ONLY },

  /*
   * Type 2, with EtherType 0x8899, assigned to Realtek;
   * they use it for several on-the-Ethernet protocols
   * as well, but there are fields that allow the two
   * tag formats, and all the protocols in question,
   * to be distinguiished from one another.
   *
   * See
   *
   *    https://elixir.bootlin.com/linux/v6.13.2/source/net/dsa/tag_rtl4_a.c
   *
   *    http://realtek.info/pdf/rtl8306sd%28m%29_datasheet_1.1.pdf
   *
   * and various pages in tcpdump's print-realtek.c and Wireshark's
   * epan/dissectors/packet-realtek.c for the other protocols.
   */
  { "rtl4a", DLT_DEBUG_ONLY },

  /*
   * Type 2, with EtherType 0x8899, assigned to Realtek;
   * see above.
   */
  { "rtl8_4", DLT_DEBUG_ONLY },

  /*
   * Type 3, with the same tag format as rtl8_4.
   */
  { "rtl8_4t", DLT_DEBUG_ONLY },

  /*
   * Type 2, with EtherType 0xe001; that's probably
   * self-assigned.
   *
   * See
   *
   *    https://elixir.bootlin.com/linux/v6.13.2/source/net/dsa/tag_rzn1_a5psw.c
   */
  { "a5psw", DLT_DEBUG_ONLY },

  /*
   * Type 2, with EtherType 0x8100 or the self-assigned
   * 0xdadb, so this really should have its own
   * LINKTYPE_/DLT_ value; that would also allow the
   * VID of the tag to be dissected as per
   *
   *    https://elixir.bootlin.com/linux/v6.13.2/source/net/dsa/tag_8021q.c#L15
   */
  { "sja1105", DLT_DEBUG_ONLY },

  /*
   * Type "none of the above", with both a header and trailer,
   * with different transmit and receive tags.  Has
   * EtherType 0xdadc, which is probably self-assigned.
   */
  { "sja1110", DLT_DEBUG_ONLY },

  /*
   * Type 3, as the name suggests.
   *
   * See
   *
   *    https://elixir.bootlin.com/linux/v6.13.2/source/net/dsa/tag_trailer.c
   */
  { "trailer", DLT_DEBUG_ONLY },

  /*
   * Type 2, with EtherType 0x8100; the VID can be interpreted
   * as per
   *
   *    https://elixir.bootlin.com/linux/v6.13.2/source/net/dsa/tag_8021q.c#L15
   */
  { "vsc73xx-8021q", DLT_DEBUG_ONLY },

  /*
   * Type 3.
   *
   * See
   *
   *    https://elixir.bootlin.com/linux/v6.13.2/source/net/dsa/tag_xrs700x.c
   */
  { "xrs700x", DLT_DEBUG_ONLY },
};

/*
 * Return 1 if the interface uses DSA tagging, 0 if the interface does not use
 * DSA tagging, or PCAP_ERROR on error.
 */
static int
iface_dsa_get_proto_info(const char *device, pcap_t *handle)
{
  char *pathstr;
  unsigned int i;
  /*
   * Make this significantly smaller than PCAP_ERRBUF_SIZE;
   * the tag *shouldn't* have some huge long name, and making
   * it smaller keeps newer versions of GCC from whining that
   * the error message if we don't support the tag could
   * overflow the error message buffer.
   */
  char buf[128];
  ssize_t r;
  int fd;

  fd = asprintf(&pathstr, "/sys/class/net/%s/dsa/tagging", device);
  if (fd < 0) {
    pcapint_fmt_errmsg_for_errno(handle->errbuf, PCAP_ERRBUF_SIZE,
            fd, "asprintf");
    return PCAP_ERROR;
  }

  fd = open(pathstr, O_RDONLY);
  free(pathstr);
  /*
   * This could be not fatal: kernel >= 4.20 *might* expose this
   * attribute.  However, if it exposes the attribute, but the read has
   * failed due to another reason (ENFILE, EMFILE, ENOMEM...), propagate
   * the failure.
   */
  if (fd < 0) {
    if (errno == ENOENT)
      return 0;
    pcapint_fmt_errmsg_for_errno(handle->errbuf, PCAP_ERRBUF_SIZE,
                                 errno, "open");
    return PCAP_ERROR;
  }

  r = read(fd, buf, sizeof(buf) - 1);
  if (r <= 0) {
    pcapint_fmt_errmsg_for_errno(handle->errbuf, PCAP_ERRBUF_SIZE,
            errno, "read");
    close(fd);
    return PCAP_ERROR;
  }
  close(fd);

  /*
   * Buffer should be LF terminated.
   */
  if (buf[r - 1] == '\n')
    r--;
  buf[r] = '\0';

  /*
   * The string "none" indicates that the interface does not have
   * any tagging protocol configured, and is therefore a standard
   * Ethernet interface.
   */
  if (strcmp(buf, "none") == 0)
    return 0;

  /*
   * Every element in the array stands for a DSA-tagged interface.  Using
   * DLT_EN10MB (the standard IEEE 802.3 Ethernet) for such an interface
   * may seem a good idea at first, but doing so would certainly cause
   * major problems in areas that are already complicated and depend on
   * DLT_EN10MB meaning the standard IEEE 802.3 Ethernet only, namely:
   *
   * - live capturing of packets on Linux, and
   * - live kernel filtering of packets on Linux, and
   * - live userspace filtering of packets on Linux, and
   * - offline filtering of packets on all supported OSes, and
   * - identification of savefiles on all OSes.
   *
   * Therefore use a default DLT value that does not block capturing and
   * hexdumping of unsupported DSA encodings (in case the tag is not in
   * the array) and enforce the non-use of DLT_EN10MB (in case the tag is
   * in the array, but is incorrectly declared).
   */
  handle->linktype = DLT_DEBUG_ONLY;
  for (i = 0; i < sizeof(dsa_protos) / sizeof(dsa_protos[0]); i++) {
    if (strcmp(buf, dsa_protos[i].name) == 0) {
      if (dsa_protos[i].linktype != DLT_EN10MB)
        handle->linktype = dsa_protos[i].linktype;
      break;
    }
  }
  return 1;
}

/*
 *  Query the kernel for the MTU of the given interface.
 */
static int
iface_get_mtu(int fd, const char *device, char *ebuf)
{
  struct ifreq  ifr;

  if (!device)
    return BIGGER_THAN_ALL_MTUS;

  memset(&ifr, 0, sizeof(ifr));
  pcapint_strlcpy(ifr.ifr_name, device, sizeof(ifr.ifr_name));

  if (ioctl(fd, SIOCGIFMTU, &ifr) == -1) {
    pcapint_fmt_errmsg_for_errno(ebuf, PCAP_ERRBUF_SIZE,
        errno, "SIOCGIFMTU");
    return -1;
  }

  return ifr.ifr_mtu;
}

/*
 *  Get the hardware type of the given interface as ARPHRD_xxx constant.
 */
static int
iface_get_arptype(int fd, const char *device, char *ebuf)
{
  struct ifreq  ifr;
  int   ret;

  memset(&ifr, 0, sizeof(ifr));
  pcapint_strlcpy(ifr.ifr_name, device, sizeof(ifr.ifr_name));

  if (ioctl(fd, SIOCGIFHWADDR, &ifr) == -1) {
    if (errno == ENODEV) {
      /*
       * No such device.
       *
       * There's nothing more to say, so clear
       * the error message.
       */
      ret = PCAP_ERROR_NO_SUCH_DEVICE;
      ebuf[0] = '\0';
    } else {
      ret = PCAP_ERROR;
      pcapint_fmt_errmsg_for_errno(ebuf, PCAP_ERRBUF_SIZE,
          errno, "SIOCGIFHWADDR");
    }
    return ret;
  }

  return ifr.ifr_hwaddr.sa_family;
}

/*
 * In a DLT_CAN_SOCKETCAN frame the first four bytes are a 32-bit integer
 * value in host byte order if the filter program is running in the kernel and
 * in network byte order if in userland.  This applies to both CC, FD and XL
 * frames, see pcap_handle_packet_mmap() for the rationale.  Return 1 iff the
 * [possibly modified] filter program can work correctly in the kernel.
 */
#if __BYTE_ORDER == __LITTLE_ENDIAN
static int
fix_dlt_can_socketcan(const u_int len, struct bpf_insn insn[])
{
  for (u_int i = 0; i < len; ++i) {
    switch (insn[i].code) {
    case BPF_LD|BPF_B|BPF_ABS: // ldb [k]
    case BPF_LDX|BPF_MSH|BPF_B: // ldxb 4*([k]&0xf)
      if (insn[i].k < 4)
        insn[i].k = 3 - insn[i].k; // Fixed now.
      break;
    case BPF_LD|BPF_H|BPF_ABS: // ldh [k]
    case BPF_LD|BPF_W|BPF_ABS: // ld [k]
      /*
       * A halfword or a word load cannot be fixed by just
       * changing k, even if every required byte is within
       * the byte-swapped part of the frame, even if the
       * load is aligned.  The fix would require either
       * rewriting the filter program extensively or
       * generating it differently in the first place.
       */
    case BPF_LD|BPF_B|BPF_IND: // ldb [x + k]
    case BPF_LD|BPF_H|BPF_IND: // ldh [x + k]
    case BPF_LD|BPF_W|BPF_IND: // ld [x + k]
      /*
       * In addition to the above, a variable offset load
       * cannot be fixed because x can have any value, thus
       * x + k can have any value, but only the first four
       * bytes are swapped.  An easy way to demonstrate it
       * is to compile "link[link[4]] == 0", which will use
       * "ldb [x + 0]" to access one of the first four bytes
       * of the frame iff CAN CC/FD payload length is less
       * than 4.
       */
      if (insn[i].k < 4)
        return 0; // Userland filtering only.
      break;
    }
  }
  return 1;
}
#else
static int
fix_dlt_can_socketcan(const u_int len _U_, struct bpf_insn insn[] _U_)
{
  return 1;
}
#endif // __BYTE_ORDER == __LITTLE_ENDIAN

static int
fix_program(pcap_t *handle, struct sock_fprog *fcode)
{
  struct pcap_linux *handlep = handle->priv;
  size_t prog_size;
  int i;
  struct bpf_insn *p;
  struct bpf_insn *f;
  int len;

  /*
   * Make a copy of the filter, and modify that copy if
   * necessary.
   */
  prog_size = sizeof(*handle->fcode.bf_insns) * handle->fcode.bf_len;
  len = handle->fcode.bf_len;
  f = (struct bpf_insn *)malloc(prog_size);
  if (f == NULL) {
    pcapint_fmt_errmsg_for_errno(handle->errbuf, PCAP_ERRBUF_SIZE,
        errno, "malloc");
    return -1;
  }
  memcpy(f, handle->fcode.bf_insns, prog_size);
  fcode->len = len;
  fcode->filter = (struct sock_filter *) f;

  switch (handle->linktype) {
  case DLT_CAN_SOCKETCAN:
    /*
     * If a similar fix needs to be done for CAN frames that
     * appear on the "any" pseudo-interface, it needs to be done
     * differently because that would be within DLT_LINUX_SLL or
     * DLT_LINUX_SLL2.
     */
    return fix_dlt_can_socketcan(len, f);
  }

  for (i = 0; i < len; ++i) {
    p = &f[i];
    /*
     * What type of instruction is this?
     */
    switch (BPF_CLASS(p->code)) {

    case BPF_LD:
    case BPF_LDX:
      /*
       * It's a load instruction; is it loading
       * from the packet?
       */
      switch (BPF_MODE(p->code)) {

      case BPF_ABS:
      case BPF_IND:
      case BPF_MSH:
        /*
         * Yes; are we in cooked mode?
         */
        if (handlep->cooked) {
          /*
           * Yes, so we need to fix this
           * instruction.
           */
          if (fix_offset(handle, p) < 0) {
            /*
             * We failed to do so.
             * Return 0, so our caller
             * knows to punt to userland.
             */
            return 0;
          }
        }
        break;
      }
      break;
    }
  }
  return 1; /* we succeeded */
}

static int
fix_offset(pcap_t *handle, struct bpf_insn *p)
{
  /*
   * Existing references to auxiliary data shouldn't be adjusted.
   *
   * Note that SKF_AD_OFF is negative, but p->k is unsigned, so
   * we use >= and cast SKF_AD_OFF to unsigned.
   */
  if (p->k >= (bpf_u_int32)SKF_AD_OFF)
    return 0;
  if (handle->linktype == DLT_LINUX_SLL2) {
    /*
     * What's the offset?
     */
    if (p->k >= SLL2_HDR_LEN) {
      /*
       * It's within the link-layer payload; that starts
       * at an offset of 0, as far as the kernel packet
       * filter is concerned, so subtract the length of
       * the link-layer header.
       */
      p->k -= SLL2_HDR_LEN;
    } else if (p->k == 0) {
      /*
       * It's the protocol field; map it to the
       * special magic kernel offset for that field.
       */
      p->k = SKF_AD_OFF + SKF_AD_PROTOCOL;
    } else if (p->k == 4) {
      /*
       * It's the ifindex field; map it to the
       * special magic kernel offset for that field.
       */
      p->k = SKF_AD_OFF + SKF_AD_IFINDEX;
    } else if (p->k == 10) {
      /*
       * It's the packet type field; map it to the
       * special magic kernel offset for that field.
       */
      p->k = SKF_AD_OFF + SKF_AD_PKTTYPE;
    } else if ((bpf_int32)(p->k) > 0) {
      /*
       * It's within the header, but it's not one of
       * those fields; we can't do that in the kernel,
       * so punt to userland.
       */
      return -1;
    }
  } else {
    /*
     * What's the offset?
     */
    if (p->k >= SLL_HDR_LEN) {
      /*
       * It's within the link-layer payload; that starts
       * at an offset of 0, as far as the kernel packet
       * filter is concerned, so subtract the length of
       * the link-layer header.
       */
      p->k -= SLL_HDR_LEN;
    } else if (p->k == 0) {
      /*
       * It's the packet type field; map it to the
       * special magic kernel offset for that field.
       */
      p->k = SKF_AD_OFF + SKF_AD_PKTTYPE;
    } else if (p->k == 14) {
      /*
       * It's the protocol field; map it to the
       * special magic kernel offset for that field.
       */
      p->k = SKF_AD_OFF + SKF_AD_PROTOCOL;
    } else if ((bpf_int32)(p->k) > 0) {
      /*
       * It's within the header, but it's not one of
       * those fields; we can't do that in the kernel,
       * so punt to userland.
       */
      return -1;
    }
  }
  return 0;
}

static int
set_kernel_filter(pcap_t *handle, struct sock_fprog *fcode)
{
  int total_filter_on = 0;
  int save_mode;
  int ret;
  int save_errno;

  /*
   * The socket filter code doesn't discard all packets queued
   * up on the socket when the filter is changed; this means
   * that packets that don't match the new filter may show up
   * after the new filter is put onto the socket, if those
   * packets haven't yet been read.
   *
   * This means, for example, that if you do a tcpdump capture
   * with a filter, the first few packets in the capture might
   * be packets that wouldn't have passed the filter.
   *
   * We therefore discard all packets queued up on the socket
   * when setting a kernel filter.  (This isn't an issue for
   * userland filters, as the userland filtering is done after
   * packets are queued up.)
   *
   * To flush those packets, we put the socket in read-only mode,
   * and read packets from the socket until there are no more to
   * read.
   *
   * In order to keep that from being an infinite loop - i.e.,
   * to keep more packets from arriving while we're draining
   * the queue - we put the "total filter", which is a filter
   * that rejects all packets, onto the socket before draining
   * the queue.
   *
   * This code deliberately ignores any errors, so that you may
   * get bogus packets if an error occurs, rather than having
   * the filtering done in userland even if it could have been
   * done in the kernel.
   */
  if (setsockopt(handle->fd, SOL_SOCKET, SO_ATTACH_FILTER,
           &total_fcode, sizeof(total_fcode)) == 0) {
    char drain[1];

    /*
     * Note that we've put the total filter onto the socket.
     */
    total_filter_on = 1;

    /*
     * Save the socket's current mode, and put it in
     * non-blocking mode; we drain it by reading packets
     * until we get an error (which is normally a
     * "nothing more to be read" error).
     */
    save_mode = fcntl(handle->fd, F_GETFL, 0);
    if (save_mode == -1) {
      pcapint_fmt_errmsg_for_errno(handle->errbuf,
          PCAP_ERRBUF_SIZE, errno,
          "can't get FD flags when changing filter");
      return -2;
    }
    if (fcntl(handle->fd, F_SETFL, save_mode | O_NONBLOCK) < 0) {
      pcapint_fmt_errmsg_for_errno(handle->errbuf,
          PCAP_ERRBUF_SIZE, errno,
          "can't set nonblocking mode when changing filter");
      return -2;
    }
    while (recv(handle->fd, &drain, sizeof drain, MSG_TRUNC) >= 0)
      ;
    save_errno = errno;
    if (save_errno != EAGAIN) {
      /*
       * Fatal error.
       *
       * If we can't restore the mode or reset the
       * kernel filter, there's nothing we can do.
       */
      (void)fcntl(handle->fd, F_SETFL, save_mode);
      (void)reset_kernel_filter(handle);
      pcapint_fmt_errmsg_for_errno(handle->errbuf,
          PCAP_ERRBUF_SIZE, save_errno,
          "recv failed when changing filter");
      return -2;
    }
    if (fcntl(handle->fd, F_SETFL, save_mode) == -1) {
      pcapint_fmt_errmsg_for_errno(handle->errbuf,
          PCAP_ERRBUF_SIZE, errno,
          "can't restore FD flags when changing filter");
      return -2;
    }
  }

  /*
   * Now attach the new filter.
   */
  ret = setsockopt(handle->fd, SOL_SOCKET, SO_ATTACH_FILTER,
       fcode, sizeof(*fcode));
  if (ret == -1 && total_filter_on) {
    /*
     * Well, we couldn't set that filter on the socket,
     * but we could set the total filter on the socket.
     *
     * This could, for example, mean that the filter was
     * too big to put into the kernel, so we'll have to
     * filter in userland; in any case, we'll be doing
     * filtering in userland, so we need to remove the
     * total filter so we see packets.
     */
    save_errno = errno;

    /*
     * If this fails, we're really screwed; we have the
     * total filter on the socket, and it won't come off.
     * Report it as a fatal error.
     */
    if (reset_kernel_filter(handle) == -1) {
      pcapint_fmt_errmsg_for_errno(handle->errbuf,
          PCAP_ERRBUF_SIZE, errno,
          "can't remove kernel total filter");
      return -2;  /* fatal error */
    }

    errno = save_errno;
  }
  return ret;
}

static int
reset_kernel_filter(pcap_t *handle)
{
  int ret;
  /*
   * setsockopt() barfs unless it get a dummy parameter.
   * valgrind whines unless the value is initialized,
   * as it has no idea that setsockopt() ignores its
   * parameter.
   */
  int dummy = 0;

  ret = setsockopt(handle->fd, SOL_SOCKET, SO_DETACH_FILTER,
           &dummy, sizeof(dummy));
  /*
   * Ignore ENOENT - it means "we don't have a filter", so there
   * was no filter to remove, and there's still no filter.
   *
   * Also ignore ENONET, as a lot of kernel versions had a
   * typo where ENONET, rather than ENOENT, was returned.
   */
  if (ret == -1 && errno != ENOENT && errno != ENONET)
    return -1;
  return 0;
}

int
pcap_set_protocol_linux(pcap_t *p, int protocol)
{
  if (pcapint_check_activated(p))
    return (PCAP_ERROR_ACTIVATED);
  p->opt.protocol = protocol;
  return (0);
}

/*
 * Libpcap version string.
 */
#if defined(HAVE_TPACKET3) && defined(PCAP_SUPPORT_NETMAP)
  #define ADDITIONAL_INFO_STRING  "with TPACKET_V3 and netmap"
#elif defined(HAVE_TPACKET3)
  #define ADDITIONAL_INFO_STRING  "with TPACKET_V3"
#elif defined(PCAP_SUPPORT_NETMAP)
  #define ADDITIONAL_INFO_STRING  "with TPACKET_V2 and netmap"
#else
  #define ADDITIONAL_INFO_STRING  "with TPACKET_V2"
#endif

const char *
pcap_lib_version(void)
{
  return (PCAP_VERSION_STRING_WITH_ADDITIONAL_INFO(ADDITIONAL_INFO_STRING));
}

Coverage Report

Created: 2026-10-17 08:12