22. L3 Forwarding Graph Sample Application

The L3 Forwarding Graph application is a simple example of packet processing using the DPDK Graph framework. The application performs L3 forwarding using Graph framework and nodes written for graph framework.

22.1. Overview

The application demonstrates the use of the graph framework and graph nodes ethdev_rx, ip4_lookup, ip4_rewrite, ethdev_tx and pkt_drop in DPDK to implement packet forwarding.

The initialization is very similar to those of the L3 Forwarding Sample Application. There is also additional initialization of graph for graph object creation and configuration per lcore. Run-time path is main thing that differs from L3 forwarding sample application. Difference is that forwarding logic starting from Rx, followed by LPM lookup, TTL update and finally Tx is implemented inside graph nodes. These nodes are interconnected in graph framework. Application main loop needs to walk over graph using rte_graph_walk() with graph objects created one per slave lcore.

The lookup method is as per implementation of ip4_lookup graph node. The ID of the output interface for the input packet is the next hop returned by the LPM lookup. The set of LPM rules used by the application is statically configured and provided to ip4_lookup graph node and ip4_rewrite graph node using node control API rte_node_ip4_route_add() and rte_node_ip4_rewrite_add().

In the sample application, only IPv4 forwarding is supported as of now.

22.2. Compiling the Application

To compile the sample application see Compiling the Sample Applications.

The application is located in the l3fwd-graph sub-directory.

22.3. Running the Application

The application has a number of command line options similar to l3fwd:

./l3fwd-graph [EAL options] -- -p PORTMASK
                               [--enable-jumbo [--max-pkt-len PKTLEN]]


  • -p PORTMASK: Hexadecimal bitmask of ports to configure
  • -P: Optional, sets all ports to promiscuous mode so that packets are accepted regardless of the packet’s Ethernet MAC destination address. Without this option, only packets with the Ethernet MAC destination address set to the Ethernet address of the port are accepted.
  • --config (port,queue,lcore)[,(port,queue,lcore)]: Determines which queues from which ports are mapped to which cores.
  • --eth-dest=X,MM:MM:MM:MM:MM:MM: Optional, ethernet destination for port X.
  • --enable-jumbo: Optional, enables jumbo frames.
  • --max-pkt-len: Optional, under the premise of enabling jumbo, maximum packet length in decimal (64-9600).
  • --no-numa: Optional, disables numa awareness.
  • --per-port-pool: Optional, set to use independent buffer pools per port. Without this option, single buffer pool is used for all ports.

For example, consider a dual processor socket platform with 8 physical cores, where cores 0-7 and 16-23 appear on socket 0, while cores 8-15 and 24-31 appear on socket 1.

To enable L3 forwarding between two ports, assuming that both ports are in the same socket, using two cores, cores 1 and 2, (which are in the same socket too), use the following command:

./build/l3fwd-graph -l 1,2 -n 4 -- -p 0x3 --config="(0,0,1),(1,0,2)"

In this command:

  • The -l option enables cores 1, 2
  • The -p option enables ports 0 and 1
  • The –config option enables one queue on each port and maps each (port,queue) pair to a specific core. The following table shows the mapping in this example:
Port Queue lcore Description
0 0 1 Map queue 0 from port 0 to lcore 1.
1 0 2 Map queue 0 from port 1 to lcore 2.

Refer to the DPDK Getting Started Guide for general information on running applications and the Environment Abstraction Layer (EAL) options.

22.4. Explanation

The following sections provide some explanation of the sample application code. As mentioned in the overview section, the initialization is similar to that of the L3 Forwarding Sample Application. Run-time path though similar in functionality to that of L3 Forwarding Sample Application, major part of the implementation is in graph nodes via used via librte_node library. The following sections describe aspects that are specific to the L3 Forwarding Graph sample application.

22.4.1. Graph Node Pre-Init Configuration

After device configuration and device Rx, Tx queue setup is complete, a minimal config of port id, num_rx_queues, num_tx_queues, mempools etc will be passed to ethdev_* node ctrl API rte_node_eth_config(). This will be lead to the clone of ethdev_rx and ethdev_tx nodes as ethdev_rx-X-Y and ethdev_tx-X where X, Y represent port id and queue id associated with them. In case of ethdev_tx-X nodes, tx queue id assigned per instance of the node is same as graph id.

These cloned nodes along with existing static nodes such as ip4_lookup and ip4_rewrite will be used in graph creation to associate node’s to lcore specific graph object.


    /* ... */
    ret = rte_eth_dev_configure(portid, nb_rx_queue,
                                n_tx_queue, &local_port_conf);
    /* ... */

    /* Init one TX queue per couple (lcore,port) */
    queueid = 0;
    for (lcore_id = 0; lcore_id < RTE_MAX_LCORE; lcore_id++) {
        /* ... */
        ret = rte_eth_tx_queue_setup(portid, queueid, nb_txd,
                                     socketid, txconf);
        /* ... */

    /* Setup ethdev node config */
    ethdev_conf[nb_conf].port_id = portid;
    ethdev_conf[nb_conf].num_rx_queues = nb_rx_queue;
    ethdev_conf[nb_conf].num_tx_queues = n_tx_queue;
    if (!per_port_pool)
        ethdev_conf[nb_conf].mp = pktmbuf_pool[0];
      ethdev_conf[nb_conf].mp = pktmbuf_pool[portid];
    ethdev_conf[nb_conf].mp_count = NB_SOCKETS;


for (lcore_id = 0; lcore_id < RTE_MAX_LCORE; lcore_id++) {
    /* Init RX queues */
    for (queue = 0; queue < qconf->n_rx_queue; ++queue) {
        /* ... */
        if (!per_port_pool)
            ret = rte_eth_rx_queue_setup(portid, queueid, nb_rxd, socketid,
                                         &rxq_conf, pktmbuf_pool[0][socketid]);
          ret = rte_eth_rx_queue_setup(portid, queueid, nb_rxd, socketid,
                                       &rxq_conf, pktmbuf_pool[portid][socketid]);
        /* ... */

/* Ethdev node config, skip rx queue mapping */
ret = rte_node_eth_config(ethdev_conf, nb_conf, nb_graphs);

22.4.2. Graph Initialization

Now a graph needs to be created with a specific set of nodes for every lcore. A graph object returned after graph creation is a per lcore object and cannot be shared between lcores. Since ethdev_tx-X node is per port node, it can be associated with all the graphs created as all the lcores should have Tx capability for every port. But ethdev_rx-X-Y node is created per (port, rx_queue_id), so they should be associated with a graph based on the application argument --config specifying rx queue mapping to lcore.


The Graph creation will fail if the passed set of shell node pattern’s are not sufficient to meet their inter-dependency or even one node is not found with a given regex node pattern.

static const char *const default_patterns[] = {
const char **node_patterns;
uint16_t nb_pattern;

/* ... */

/* Create a graph object per lcore with common nodes and
 * lcore specific nodes based on application arguments
nb_patterns = RTE_DIM(default_patterns);
node_patterns = malloc((MAX_RX_QUEUE_PER_LCORE + nb_patterns) *
memcpy(node_patterns, default_patterns,
       nb_patterns * sizeof(*node_patterns));

memset(&graph_conf, 0, sizeof(graph_conf));

/* Common set of nodes in every lcore's graph object */
graph_conf.node_patterns = node_patterns;

for (lcore_id = 0; lcore_id < RTE_MAX_LCORE; lcore_id++) {
    /* ... */

    /* Skip graph creation if no source exists */
    if (!qconf->n_rx_queue)

    /* Add rx node patterns of this lcore based on --config */
    for (i = 0; i < qconf->n_rx_queue; i++) {
        graph_conf.node_patterns[nb_patterns + i] =

    graph_conf.nb_node_patterns = nb_patterns + i;
    graph_conf.socket_id = rte_lcore_to_socket_id(lcore_id);

    snprintf(qconf->name, sizeof(qconf->name), "worker_%u", lcore_id);

    graph_id = rte_graph_create(qconf->name, &graph_conf);

    /* ... */

    qconf->graph = rte_graph_lookup(qconf->name);

    /* ... */

22.4.3. Forwarding data(Route, Next-Hop) addition

Once graph objects are created, node specific info like routes and rewrite headers will be provided run-time using rte_node_ip4_route_add() and rte_node_ip4_rewrite_add() API.


Since currently ip4_lookup and ip4_rewrite nodes don’t support lock-less mechanisms(RCU, etc) to add run-time forwarding data like route and rewrite data, forwarding data is added before packet processing loop is launched on slave lcore.

/* Add route to ip4 graph infra */
for (i = 0; i < IPV4_L3FWD_LPM_NUM_ROUTES; i++) {
    /* ... */

    dst_port = ipv4_l3fwd_lpm_route_array[i].if_out;
    next_hop = i;

    /* ... */
    ret = rte_node_ip4_route_add(ipv4_l3fwd_lpm_route_array[i].ip,
                                 ipv4_l3fwd_lpm_route_array[i].depth, next_hop,

    /* ... */

    memcpy(rewrite_data, val_eth + dst_port, rewrite_len);

    /* Add next hop for a given destination */
    ret = rte_node_ip4_rewrite_add(next_hop, rewrite_data,
                                   rewrite_len, dst_port);

    RTE_LOG(INFO, L3FWD_GRAPH, "Added route %s, next_hop %u\n",
            route_str, next_hop);

22.4.4. Packet Forwarding using Graph Walk

Now that all the device configurations are done, graph creations are done and forwarding data is updated with nodes, slave lcores will be launched with graph main loop. Graph main loop is very simple in the sense that it needs to continuously call a non-blocking API rte_graph_walk() with it’s lcore specific graph object that was already created.


rte_graph_walk() will walk over all the sources nodes i.e ethdev_rx-X-Y associated with a given graph and Rx the available packets and enqueue them to the following node ip4_lookup which then will enqueue them to ip4_rewrite node if LPM lookup succeeds. ip4_rewrite node then will update Ethernet header as per next-hop data and transmit the packet via port ‘Z’ by enqueuing to ethdev_tx-Z node instance in its graph object.

/* Main processing loop */
static int
graph_main_loop(void *conf)
    // ...

    lcore_id = rte_lcore_id();
    qconf = &lcore_conf[lcore_id];
    graph = qconf->graph;

            "Entering main loop on lcore %u, graph %s(%p)\n", lcore_id,
            qconf->name, graph);

    /* Walk over graph until signal to quit */
    while (likely(!force_quit))
    return 0;