source: lib/format_linux_ring.c @ 630460b

/*
 *
 * Copyright (c) 2007-2016 The University of Waikato, Hamilton, New Zealand.
 * All rights reserved.
 *
 * This file is part of libtrace.
 *
 * This code has been developed by the University of Waikato WAND
 * research group. For further information please see http://www.wand.net.nz/
 *
 * libtrace is free software; you can redistribute it and/or modify
 * it under the terms of the GNU Lesser General Public License as published by
 * the Free Software Foundation; either version 3 of the License, or
 * (at your option) any later version.
 *
 * libtrace is distributed in the hope that it will be useful,
 * but WITHOUT ANY WARRANTY; without even the implied warranty of
 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
 * GNU Lesser General Public License for more details.
 *
 * You should have received a copy of the GNU Lesser General Public License
 * along with this program.  If not, see <http://www.gnu.org/licenses/>.
 *
 *
 */

/* This format module deals with using the Linux Ring capture format (also
 * known as PACKET_MMAP).
 *
 * Linux Ring is a LIVE capture format.
 *
 * This format also supports writing which will write packets out to the
 * network as a form of packet replay. This should not be confused with the
 * RT protocol which is intended to transfer captured packet records between
 * RT-speaking programs.
 */

#include "config.h"
#include "libtrace.h"
#include "libtrace_int.h"
#include "format_helper.h"
#include "libtrace_arphrd.h"
#include <stdlib.h>
#include <errno.h>
#include <unistd.h>
#include <string.h>

#ifdef HAVE_INTTYPES_H
#  include <inttypes.h>
#else
# error "Can't find inttypes.h"
#endif

#include "format_linux_common.h"

/* Get the start of the captured data. I'm not sure if tp_mac (link layer) is
 * always guaranteed. If it's not there then just use tp_net.
 */
#define TP_TRACE_START(mac, net, hdrend) \
        ((mac) > (hdrend) && (mac) < (net) ? (mac) : (net))

static pthread_mutex_t pagesize_mutex;
#ifdef HAVE_NETPACKET_PACKET_H
/* Get current frame in the ring buffer*/
#define GET_CURRENT_BUFFER(stream) \
        ((void *)stream->rx_ring +                              \
         (stream->rxring_offset *                               \
          stream->req.tp_frame_size))

/* Cached page size, the page size shouldn't be changing */
static int pagesize = 0;


/*
 * Try figure out the best sizes for the ring buffer. Ensure that:
 * - max(Block_size) == page_size << max_order
 * - Frame_size == page_size << x (so that block_size%frame_size == 0)
 *   This means that there will be no wasted space between blocks
 * - Frame_size < block_size
 * - Frame_size is as close as possible to LIBTRACE_PACKET_BUFSIZE, but not
 *   bigger
 * - Frame_nr = Block_nr * (frames per block)
 * - CONF_RING_FRAMES is used a minimum number of frames to hold
 * - Calculates based on max_order and buf_min
 */
static void calculate_buffers(struct tpacket_req * req, int fd, char * uri,
                uint32_t max_order)
{
        struct ifreq ifr;
        unsigned max_frame = LIBTRACE_PACKET_BUFSIZE;
        pthread_mutex_lock(&pagesize_mutex);
        if (pagesize == 0) {
                pagesize = getpagesize();
        }
        pthread_mutex_unlock(&pagesize_mutex);

        strcpy(ifr.ifr_name, uri);
        /* Don't bother trying to set frame size above mtu linux will drop
         * these anyway.
         *
         * Remember, that our frame also has to include a TPACKET header!
         */
        if (ioctl(fd, SIOCGIFMTU, (caddr_t)&ifr) >= 0)
                max_frame = ifr.ifr_mtu + TPACKET_ALIGN(TPACKET2_HDRLEN);
        if (max_frame > LIBTRACE_PACKET_BUFSIZE)
                max_frame = LIBTRACE_PACKET_BUFSIZE;

        /* Calculate frame size */
        req->tp_frame_size = pagesize;
        while (req->tp_frame_size < max_frame &&
              req->tp_frame_size < LIBTRACE_PACKET_BUFSIZE) {
                req->tp_frame_size <<= 1;
        }
        if (req->tp_frame_size > LIBTRACE_PACKET_BUFSIZE)
                req->tp_frame_size >>= 1;

        /* Calculate block size */
        req->tp_block_size = pagesize << max_order;
        /* If max order is too high this might become 0 */
        if (req->tp_block_size == 0) {
                calculate_buffers(req, fd, uri, max_order-1);
                return;
        }
        do {
                req->tp_block_size >>= 1;
        } while ((CONF_RING_FRAMES * req->tp_frame_size) <= req->tp_block_size);
        req->tp_block_size <<= 1;

        /* Calculate number of blocks */
        req->tp_block_nr = (CONF_RING_FRAMES * req->tp_frame_size)
                / req->tp_block_size;
        if((CONF_RING_FRAMES * req->tp_frame_size) % req->tp_block_size != 0)
                req->tp_block_nr++;

        /* Calculate packets such that we use all the space we have to
         * allocated */
        req->tp_frame_nr = req->tp_block_nr *
                (req->tp_block_size / req->tp_frame_size);

        /*
        printf("MaxO 0x%x BS 0x%x BN 0x%x FS 0x%x FN 0x%x\n",
                max_order,
                req->tp_block_size,
                req->tp_block_nr,
                req->tp_frame_size,
                req->tp_frame_nr);
        */

        /* In case we have some silly values*/
        if (!req->tp_block_size) {
                fprintf(stderr, "Unexpected value of zero for req->tp_block_size in calculate_buffers()\n");
        }
        if (!req->tp_block_nr) {
                fprintf(stderr, "Unexpected value of zero for req->tp_block_nr in calculate_buffers()\n");
        }
        if (!req->tp_frame_size) {
                fprintf(stderr, "Unexpected value of zero for req->tp_frame_size in calculate_buffers()\n");
        }
        if (!req->tp_frame_nr) {
                fprintf(stderr, "Unexpected value of zero for req->tp_frame_nr in calculate_buffers()\n");
        }
        if (req->tp_block_size % req->tp_frame_size != 0) {
                fprintf(stderr, "Unexpected value of zero for req->tp_block_size %% req->tp_frame_size in calculate_buffers()\n");
        }
}

static inline int socket_to_packetmmap(char * uridata, int ring_type,
                                        int fd,
                                        struct tpacket_req * req,
                                        char ** ring_location,
                                        uint32_t *max_order,
                                        char *error) {
        int val;

        /* Switch to TPACKET header version 2, we only try support v2 because
         * v1 had problems with data type consistancy */
        val = TPACKET_V2;
        if (setsockopt(fd,
                       SOL_PACKET,
                       PACKET_VERSION,
                       &val,
                       sizeof(val)) == -1) {
                strncpy(error, "TPACKET2 not supported", 2048);
                return -1;
        }

        /* Try switch to a ring buffer. If it fails we assume the the kernel
         * cannot allocate a block of that size, so decrease max_block and
         * retry.
         */
        while(1) {
                if (*max_order <= 0) {
                        strncpy(error,
                                "Cannot allocate enough memory for ring buffer",
                                2048);
                        return -1;
                }
                calculate_buffers(req, fd, uridata, *max_order);
                if (setsockopt(fd,
                               SOL_PACKET,
                               ring_type,
                               req,
                               sizeof(struct tpacket_req)) == -1) {
                        if(errno == ENOMEM) {
                                (*max_order)--;
                        } else {
                                strncpy(error,
                                        "Error setting the ring buffer size",
                                        2048);
                                return -1;
                        }

                } else break;
        }

        /* Map the ring buffer into userspace */
        *ring_location = mmap(NULL,
                              req->tp_block_size * req->tp_block_nr,
                              PROT_READ | PROT_WRITE, MAP_SHARED, fd, 0);
        if(*ring_location == MAP_FAILED) {
                strncpy(error, "Failed to map memory for ring buffer", 2048);
                return -1;
        }

        return 0;
}

/* Release a frame back to the kernel or free() if it's a malloc'd buffer
 */
inline static void ring_release_frame(libtrace_t *libtrace UNUSED,
                                      libtrace_packet_t *packet)
{
        /* Free the old packet */
        if(packet->buffer == NULL)
                return;

        if(packet->buf_control == TRACE_CTRL_PACKET){
                free(packet->buffer);
                packet->buffer = NULL;
        }

        if(packet->buf_control == TRACE_CTRL_EXTERNAL) {
                //struct linux_format_data_t *ftd = FORMAT_DATA;
                /* Check it's within our buffer first - consider the pause
                 * resume case it might have already been free'd lets hope we
                 * get another buffer */
                // TODO: For now let any one free anything
                /*if(LIBTRACE_BETWEEN((char *) packet->buffer,
                                (char *) ftd->rx_ring,
                                ftd->rx_ring +
                                ftd->req.tp_block_size *
                                ftd->req.tp_block_nr)){*/
                TO_TP_HDR2(packet->buffer)->tp_status = 0;
                packet->buffer = NULL;
                /*}*/
        }
}

static inline int linuxring_start_input_stream(libtrace_t *libtrace,
                                               struct linux_per_stream_t *stream) {
        char error[2048];

        /* Unmap any previous ring buffers associated with this stream. */
        if (stream->rx_ring != MAP_FAILED) {
                munmap(stream->rx_ring, stream->req.tp_block_size *
                                stream->req.tp_block_nr);
                stream->rx_ring = MAP_FAILED;
                stream->rxring_offset = 0;
        }


        /* We set the socket up the same and then convert it to PACKET_MMAP */
        if (linuxcommon_start_input_stream(libtrace, stream) < 0)
                return -1;

        strncpy(error, "No known error", 2048);

        /* Make it a packetmmap */
        if(socket_to_packetmmap(libtrace->uridata, PACKET_RX_RING,
                                stream->fd,
                                &stream->req,
                                &stream->rx_ring,
                                &FORMAT_DATA->max_order,
                                error) != 0) {
                trace_set_err(libtrace, TRACE_ERR_INIT_FAILED,
                              "Initialisation of packet MMAP failed: %s",
                              error);
                linuxcommon_close_input_stream(libtrace, stream);
                return -1;
        }

        return 0;
}

static int linuxring_fin_input(libtrace_t *libtrace) {
        size_t i;

        if (libtrace->format_data) {
                for (i = 0; i < libtrace_list_get_size(FORMAT_DATA->per_stream); ++i) {
                        struct linux_per_stream_t *stream;
                        stream = libtrace_list_get_index(
                                FORMAT_DATA->per_stream, i)->data;
                        if (stream->rx_ring != MAP_FAILED) {
                                munmap(stream->rx_ring,
                                                stream->req.tp_block_size *
                                                stream->req.tp_block_nr);
                        }
                }

                if (FORMAT_DATA->filter != NULL)
                        trace_destroy_filter(FORMAT_DATA->filter);

                if (FORMAT_DATA->per_stream)
                        libtrace_list_deinit(FORMAT_DATA->per_stream);

                free(libtrace->format_data);
        }

        return 0;
}


static int linuxring_start_input(libtrace_t *libtrace)
{
        int ret = linuxring_start_input_stream(libtrace, FORMAT_DATA_FIRST);
        return ret;
}

#ifdef HAVE_PACKET_FANOUT
static int linuxring_pstart_input(libtrace_t *libtrace) {
        return linuxcommon_pstart_input(libtrace, linuxring_start_input_stream);
}
#endif

static int linuxring_start_output(libtrace_out_t *libtrace)
{
        char error[2048];
        FORMAT_DATA_OUT->fd = socket(PF_PACKET, SOCK_RAW, 0);
        if (FORMAT_DATA_OUT->fd==-1) {
                free(FORMAT_DATA_OUT);
                trace_set_err_out(libtrace, errno, "Failed to create raw socket");
                return -1;
        }

        /* Make it a packetmmap */
        if(socket_to_packetmmap(libtrace->uridata, PACKET_TX_RING,
                                FORMAT_DATA_OUT->fd,
                                &FORMAT_DATA_OUT->req,
                                &FORMAT_DATA_OUT->tx_ring,
                                &FORMAT_DATA_OUT->max_order,
                                error) != 0) {
                trace_set_err_out(libtrace, TRACE_ERR_INIT_FAILED,
                                  "Initialisation of packet MMAP failed: %s",
                                  error);
                close(FORMAT_DATA_OUT->fd);
                free(FORMAT_DATA_OUT);
                libtrace->format_data = NULL;
                return -1;
        }

        FORMAT_DATA_OUT->sock_hdr.sll_family = AF_PACKET;
        FORMAT_DATA_OUT->sock_hdr.sll_protocol = 0;
        FORMAT_DATA_OUT->sock_hdr.sll_ifindex =
                if_nametoindex(libtrace->uridata);
        FORMAT_DATA_OUT->sock_hdr.sll_hatype = 0;
        FORMAT_DATA_OUT->sock_hdr.sll_pkttype = 0;
        FORMAT_DATA_OUT->sock_hdr.sll_halen = 0;
        FORMAT_DATA_OUT->queue = 0;

        return 0;
}

static int linuxring_fin_output(libtrace_out_t *libtrace)
{
        /* Make sure any remaining frames get sent */
        sendto(FORMAT_DATA_OUT->fd,
               NULL,
               0,
               0,
               (void *) &FORMAT_DATA_OUT->sock_hdr,
               sizeof(FORMAT_DATA_OUT->sock_hdr));

        /* Unmap our data area */
        munmap(FORMAT_DATA_OUT->tx_ring,
               FORMAT_DATA_OUT->req.tp_block_size *
               FORMAT_DATA_OUT->req.tp_block_nr);

        /* Free the socket */
        close(FORMAT_DATA_OUT->fd);
        FORMAT_DATA_OUT->fd=-1;
        free(libtrace->format_data);
        return 0;
}
#endif /* HAVE_NETPACKET_PACKET_H */

static libtrace_linktype_t
linuxring_get_link_type(const struct libtrace_packet_t *packet)
{
        uint16_t linktype = GET_SOCKADDR_HDR(packet->buffer)->sll_hatype;
        return linuxcommon_get_link_type(linktype);
}

static libtrace_direction_t
linuxring_get_direction(const struct libtrace_packet_t *packet) {
        return linuxcommon_get_direction(GET_SOCKADDR_HDR(packet->buffer)->
                                         sll_pkttype);
}

static libtrace_direction_t
linuxring_set_direction(libtrace_packet_t *packet,
                        libtrace_direction_t direction) {
        return linuxcommon_set_direction(GET_SOCKADDR_HDR(packet->buffer), direction);
}

static struct timeval linuxring_get_timeval(const libtrace_packet_t *packet)
{
        struct timeval tv;
        tv.tv_sec = TO_TP_HDR2(packet->buffer)->tp_sec;
        tv.tv_usec = TO_TP_HDR2(packet->buffer)->tp_nsec / 1000;
        return tv;
}

static struct timespec linuxring_get_timespec(const libtrace_packet_t *packet)
{
        struct timespec ts;
        ts.tv_sec = TO_TP_HDR2(packet->buffer)->tp_sec;
        ts.tv_nsec = TO_TP_HDR2(packet->buffer)->tp_nsec;
        return ts;
}

static int linuxring_get_capture_length(const libtrace_packet_t *packet)
{
        return TO_TP_HDR2(packet->buffer)->tp_snaplen;
}

static int linuxring_get_wire_length(const libtrace_packet_t *packet)
{
        int wirelen = TO_TP_HDR2(packet->buffer)->tp_len;

        /* Include the missing FCS */
        if (trace_get_link_type(packet) == TRACE_TYPE_ETH)
                wirelen += 4;

        return wirelen;
}

static int linuxring_get_framing_length(const libtrace_packet_t *packet)
{
        /*
         * Need to make frame_length + capture_length = complete capture length
         * so include alignment whitespace. So reverse calculate from packet.
         */
        return (char *)packet->payload - (char *)packet->buffer;
}

static size_t linuxring_set_capture_length(libtrace_packet_t *packet,
                                           size_t size)
{
        if (!packet) {
                fprintf(stderr, "NULL packet passed into linuxring_set_capture_length()\n");
                /* Return -1 on error? */
                return ~0U;
        }
        if (size > trace_get_capture_length(packet)) {
                /* We should avoid making a packet larger */
                return trace_get_capture_length(packet);
        }

        /* Reset the cached capture length */
        packet->cached.capture_length = -1;

        TO_TP_HDR2(packet->buffer)->tp_snaplen = size;

        return trace_get_capture_length(packet);
}

static int linuxring_prepare_packet(libtrace_t *libtrace UNUSED,
                                    libtrace_packet_t *packet, void *buffer,
                                    libtrace_rt_types_t rt_type, uint32_t flags)
{
        if (packet->buffer != buffer &&
            packet->buf_control == TRACE_CTRL_PACKET) {
                free(packet->buffer);
        }

        if ((flags & TRACE_PREP_OWN_BUFFER) == TRACE_PREP_OWN_BUFFER)
                packet->buf_control = TRACE_CTRL_PACKET;
        else
                packet->buf_control = TRACE_CTRL_EXTERNAL;


        packet->buffer = buffer;
        packet->header = buffer;
        packet->payload = (char *)buffer +
                TP_TRACE_START(TO_TP_HDR2(packet->header)->tp_mac,
                               TO_TP_HDR2(packet->header)->tp_net,
                               TPACKET2_HDRLEN);
        packet->type = rt_type;

        return 0;
}

#ifdef HAVE_NETPACKET_PACKET_H
#define LIBTRACE_MIN(a,b) ((a)<(b) ? (a) : (b))
/* We use TP_STATUS_LIBTRACE to ensure we don't loop back on ourself
 * and read the same packet twice if an old packet has not yet been freed */
#define TP_STATUS_LIBTRACE 0xFFFFFFFF

inline static int linuxring_read_stream(libtrace_t *libtrace,
                                        libtrace_packet_t *packet,
                                        struct linux_per_stream_t *stream,
                                        libtrace_message_queue_t *queue,
                                        uint8_t block) {

        struct tpacket2_hdr *header;
        int ret;
        unsigned int snaplen;
        struct pollfd pollset[2];

        packet->buf_control = TRACE_CTRL_EXTERNAL;
        packet->type = TRACE_RT_DATA_LINUX_RING;

        /* Fetch the current frame */
        header = GET_CURRENT_BUFFER(stream);
        if ((((unsigned long) header) & (pagesize - 1)) != 0) {
                trace_set_err(libtrace, TRACE_ERR_BAD_IO, "Linux ring packet is not correctly "
                        "aligned to page size in linux_read_string()");
                return -1;
        }

        /* TP_STATUS_USER means that we can use the frame.
         * When a slot does not have this flag set, the frame is not
         * ready for consumption.
         */
        while (!(header->tp_status & TP_STATUS_USER) ||
                        header->tp_status == TP_STATUS_LIBTRACE) {
                if ((ret=is_halted(libtrace)) != -1)
                        return ret;
                if (!block) {
                        return 0;
                }

                pollset[0].fd = stream->fd;
                pollset[0].events = POLLIN;
                pollset[0].revents = 0;
                if (queue) {
                        pollset[1].fd = libtrace_message_queue_get_fd(queue);
                        pollset[1].events = POLLIN;
                        pollset[1].revents = 0;
                }
                /* Wait for more data or a message */
                ret = poll(pollset, (queue ? 2 : 1), 500);
                if (ret > 0) {
                        if (pollset[0].revents == POLLIN)
                                continue;
                        else if (queue && pollset[1].revents == POLLIN)
                                return READ_MESSAGE;
                        else if (queue && pollset[1].revents) {
                                /* Internal error */
                                trace_set_err(libtrace,TRACE_ERR_BAD_STATE,
                                              "Message queue error %d poll()",
                                              pollset[1].revents);
                                return READ_ERROR;
                        } else {
                                /* Try get the error from the socket */
                                int err = ENETDOWN;
                                socklen_t len = sizeof(err);
                                getsockopt(stream->fd, SOL_SOCKET, SO_ERROR,
                                           &err, &len);
                                trace_set_err(libtrace, err,
                                              "Socket error revents=%d poll()",
                                              pollset[0].revents);
                                return READ_ERROR;
                        }
                } else if (ret < 0) {
                        if (errno != EINTR) {
                                trace_set_err(libtrace,errno,"poll()");
                                return -1;
                        }
                } else {
                        /* Poll timed out - check if we should exit on next loop */
                        continue;
                }
        }
        packet->buffer = header;
        packet->trace = libtrace;
        
        header->tp_status = TP_STATUS_LIBTRACE;

        /* If a snaplen was configured, automatically truncate the packet to
         * the desired length.
         */
        snaplen=LIBTRACE_MIN(
                        (int)LIBTRACE_PACKET_BUFSIZE-(int)sizeof(*header),
                        (int)FORMAT_DATA->snaplen);
        
        TO_TP_HDR2(packet->buffer)->tp_snaplen = LIBTRACE_MIN((unsigned int)snaplen, TO_TP_HDR2(packet->buffer)->tp_len);

        /* Move to next buffer */
        stream->rxring_offset++;
        stream->rxring_offset %= stream->req.tp_frame_nr;

        packet->order = (((uint64_t)TO_TP_HDR2(packet->buffer)->tp_sec) << 32)
                        + ((((uint64_t)TO_TP_HDR2(packet->buffer)->tp_nsec)
                        << 32) / 1000000000);

        if (packet->order <= stream->last_timestamp) {
                packet->order = stream->last_timestamp + 1;
        }

        stream->last_timestamp = packet->order;

        /* We just need to get prepare_packet to set all our packet pointers
         * appropriately */
        if (linuxring_prepare_packet(libtrace, packet, packet->buffer,
                                packet->type, 0))
                return -1;
        return  linuxring_get_framing_length(packet) + 
                                linuxring_get_capture_length(packet);

}

static int linuxring_read_packet(libtrace_t *libtrace, libtrace_packet_t *packet) {
        return linuxring_read_stream(libtrace, packet, FORMAT_DATA_FIRST, NULL, 1);
}

#ifdef HAVE_PACKET_FANOUT
static int linuxring_pread_packets(libtrace_t *libtrace,
                                   libtrace_thread_t *t,
                                   libtrace_packet_t *packets[],
                                   size_t nb_packets) {
        size_t i;
        int ret;

        for (i = 0; i < nb_packets; i++) {
                ret = linuxring_read_stream(libtrace, packets[i],
                                t->format_data, &t->messages, i == 0 ? 1 : 0);
                packets[i]->error = ret;
                if (ret < 0) {
                        return ret;
                }

                if (ret == 0) {
                        if (is_halted(libtrace) == READ_EOF) {
                                return READ_EOF;
                        }
                        return i;
                }
        }

        return nb_packets;
}
#endif

/* Non-blocking read */
static libtrace_eventobj_t linuxring_event(libtrace_t *libtrace,
                                           libtrace_packet_t *packet)
{
        struct tpacket2_hdr *header;
        libtrace_eventobj_t event = {0,0,0.0,0};

        /* We must free the old packet, otherwise select() will instantly
         * return */
        ring_release_frame(libtrace, packet);

        /* Fetch the current frame */
        header = GET_CURRENT_BUFFER(FORMAT_DATA_FIRST);
        if (header->tp_status & TP_STATUS_USER &&
            header->tp_status != TP_STATUS_LIBTRACE) {
                /* We have a frame waiting */
                event.size = trace_read_packet(libtrace, packet);
                event.type = TRACE_EVENT_PACKET;
        } else {
                /* Ok we don't have a packet waiting */
                event.type = TRACE_EVENT_IOWAIT;
                event.fd = FORMAT_DATA_FIRST->fd;
        }

        return event;
}

/**
 * Free any resources being kept for this packet, Note: libtrace
 * will ensure all fields are zeroed correctly.
 */
static void linuxring_fin_packet(libtrace_packet_t *packet)
{
        libtrace_t *libtrace = packet->trace;

        if (packet->buffer == NULL)
                return;
        if (!packet->trace) {
                fprintf(stderr, "Linux ring packet is not attached to a valid "
                        "trace, Unable to release it, in linuxring_fin_packet()\n");
                return;
        }

        /* If we own the packet (i.e. it's not a copy), we need to free it */
        if (packet->buf_control == TRACE_CTRL_EXTERNAL) {
                /* If we don't have a ring its already been destroyed */
                if (FORMAT_DATA_FIRST->rx_ring != MAP_FAILED)
                        ring_release_frame(packet->trace, packet);
                else
                        packet->buffer = NULL;
        }
}

static int linuxring_write_packet(libtrace_out_t *libtrace,
                                  libtrace_packet_t *packet)
{
        struct tpacket2_hdr *header;
        struct pollfd pollset;
        struct socket_addr;
        int ret;
        unsigned max_size;
        void * off;

        if (trace_get_link_type(packet) == TRACE_TYPE_NONDATA)
                return 0;

        max_size = FORMAT_DATA_OUT->req.tp_frame_size -
                TPACKET2_HDRLEN + sizeof(struct sockaddr_ll);

        header = (void *)FORMAT_DATA_OUT->tx_ring +
                (FORMAT_DATA_OUT->txring_offset *
                 FORMAT_DATA_OUT->req.tp_frame_size);

        while(header->tp_status != TP_STATUS_AVAILABLE) {
                /* if none available: wait on more data */
                pollset.fd = FORMAT_DATA_OUT->fd;
                pollset.events = POLLOUT;
                pollset.revents = 0;
                ret = poll(&pollset, 1, 1000);
                if (ret < 0 && errno != EINTR) {
                        perror("poll");
                        return -1;
                }
                if(ret == 0) {
                        /* Timeout something has gone wrong - maybe the queue is
                         * to large so try issue another send command
                         */
                        ret = sendto(FORMAT_DATA_OUT->fd,
                                     NULL,
                                     0,
                                     0,
                                     (void *)&FORMAT_DATA_OUT->sock_hdr,
                                     sizeof(FORMAT_DATA_OUT->sock_hdr));
                        if (ret < 0) {
                                trace_set_err_out(libtrace, errno,
                                                  "sendto after timeout "
                                                  "failed");
                                return -1;
                        }
                }
        }

        header->tp_len = trace_get_capture_length(packet);

        /* We cannot write the whole packet so just write part of it */
        if (header->tp_len > max_size)
                header->tp_len = max_size;

        /* Fill packet - no sockaddr_ll in header when writing to the TX_RING */
        off = ((void *)header) + (TPACKET2_HDRLEN - sizeof(struct sockaddr_ll));
        memcpy(off, (char *)packet->payload, header->tp_len);

        /* 'Send it' and increase ring pointer to the next frame */
        header->tp_status = TP_STATUS_SEND_REQUEST;
        FORMAT_DATA_OUT->txring_offset = (FORMAT_DATA_OUT->txring_offset + 1) %
                FORMAT_DATA_OUT->req.tp_frame_nr;

        /* Notify kernel there are frames to send */
        FORMAT_DATA_OUT->queue ++;
        FORMAT_DATA_OUT->queue %= TX_MAX_QUEUE;
        if(FORMAT_DATA_OUT->queue == 0){
                ret = sendto(FORMAT_DATA_OUT->fd,
                                NULL,
                                0,
                                MSG_DONTWAIT,
                                (void *)&FORMAT_DATA_OUT->sock_hdr,
                                sizeof(FORMAT_DATA_OUT->sock_hdr));
                if (ret < 0) {
                        trace_set_err_out(libtrace, errno, "sendto failed");
                        return -1;
                }
        }
        return header->tp_len;

}

static void linuxring_help(void)
{
        printf("linuxring format module: $Revision: 1793 $\n");
        printf("Supported input URIs:\n");
        printf("\tring:eth0\n");
        printf("\n");
        printf("Supported output URIs:\n");
        printf("\tring:eth0\n");
        printf("\n");
        return;
}

static struct libtrace_format_t linuxring = {
        "ring",
        "$Id$",
        TRACE_FORMAT_LINUX_RING,
        linuxcommon_probe_filename,     /* probe filename */
        NULL,                           /* probe magic */
        linuxcommon_init_input,         /* init_input */
        linuxcommon_config_input,       /* config_input */
        linuxring_start_input,          /* start_input */
        linuxcommon_pause_input,        /* pause_input */
        linuxcommon_init_output,        /* init_output */
        NULL,                           /* config_output */
        linuxring_start_output,         /* start_ouput */
        linuxring_fin_input,            /* fin_input */
        linuxring_fin_output,           /* fin_output */
        linuxring_read_packet,          /* read_packet */
        linuxring_prepare_packet,       /* prepare_packet */
        linuxring_fin_packet,           /* fin_packet */
        linuxring_write_packet,         /* write_packet */
        NULL,                           /* flush_output */
        linuxring_get_link_type,        /* get_link_type */
        linuxring_get_direction,        /* get_direction */
        linuxring_set_direction,        /* set_direction */
        NULL,                           /* get_erf_timestamp */
        linuxring_get_timeval,          /* get_timeval */
        linuxring_get_timespec,         /* get_timespec */
        NULL,                           /* get_seconds */
        NULL,                           /* seek_erf */
        NULL,                           /* seek_timeval */
        NULL,                           /* seek_seconds */
        linuxring_get_capture_length,   /* get_capture_length */
        linuxring_get_wire_length,      /* get_wire_length */
        linuxring_get_framing_length,   /* get_framing_length */
        linuxring_set_capture_length,   /* set_capture_length */
        NULL,                           /* get_received_packets */
        NULL,                           /* get_filtered_packets */
        NULL,                           /* get_dropped_packets */
        linuxcommon_get_statistics,     /* get_statistics */
        linuxcommon_get_fd,             /* get_fd */
        linuxring_event,                /* trace_event */
        linuxring_help,                 /* help */
        NULL,                           /* next pointer */
#ifdef HAVE_PACKET_FANOUT
        {true, -1},                     /* Live, no thread limit */
        linuxring_pstart_input,         /* pstart_input */
        linuxring_pread_packets,        /* pread_packets */
        linuxcommon_pause_input,        /* ppause */
        linuxcommon_fin_input,          /* p_fin */
        linuxcommon_pregister_thread,   /* register thread */
        NULL,                           /* unregister thread */
        NULL                            /* get thread stats */
#else
        NON_PARALLEL(true)
#endif
};
#else /* HAVE_NETPACKET_PACKET_H */

static void linuxring_help(void)
{
        printf("linuxring format module: $Revision: 1793 $\n");
        printf("Not supported on this host\n");
}

static struct libtrace_format_t linuxring = {
        "ring",
        "$Id$",
        TRACE_FORMAT_LINUX_RING,
        NULL,                           /* probe filename */
        NULL,                           /* probe magic */
        NULL,                           /* init_input */
        NULL,                           /* config_input */
        NULL,                           /* start_input */
        NULL,                           /* pause_input */
        NULL,                           /* init_output */
        NULL,                           /* config_output */
        NULL,                           /* start_ouput */
        NULL,                           /* fin_input */
        NULL,                           /* fin_output */
        NULL,                           /* read_packet */
        linuxring_prepare_packet,       /* prepare_packet */
        NULL,                           /* fin_packet */
        NULL,                           /* write_packet */
        NULL,                           /* flush_output */
        linuxring_get_link_type,        /* get_link_type */
        linuxring_get_direction,        /* get_direction */
        linuxring_set_direction,        /* set_direction */
        NULL,                           /* get_erf_timestamp */
        linuxring_get_timeval,          /* get_timeval */
        linuxring_get_timespec,         /* get_timespec */
        NULL,                           /* get_seconds */
        NULL,                           /* seek_erf */
        NULL,                           /* seek_timeval */
        NULL,                           /* seek_seconds */
        linuxring_get_capture_length,   /* get_capture_length */
        linuxring_get_wire_length,      /* get_wire_length */
        linuxring_get_framing_length,   /* get_framing_length */
        linuxring_set_capture_length,   /* set_capture_length */
        NULL,                           /* get_received_packets */
        NULL,                           /* get_filtered_packets */
        NULL,                           /* get_dropped_packets */
        linuxcommon_get_statistics,     /* get_statistics */
        NULL,                           /* get_fd */
        NULL,                           /* trace_event */
        linuxring_help,                 /* help */
        NULL,                           /* next pointer */
        NON_PARALLEL(true)
};
#endif /* HAVE_NETPACKET_PACKET_H */

/* TODO: Figure out how to give this format preference over the linux native
 * formate if the user only specifies an interface */
void linuxring_constructor(void)
{
        pthread_mutex_init(&pagesize_mutex, NULL);
        register_format(&linuxring);
}