21. L3 Forwarding Sample Application

The L3 Forwarding application is a simple example of packet processing using DPDK to demonstrate usage of poll and event mode packet I/O mechanism. The application performs L3 forwarding.

21.1. Overview

The application demonstrates the use of the hash, LPM and FIB libraries in DPDK to implement packet forwarding using poll or event mode PMDs for packet I/O. The initialization and run-time paths are very similar to those of the L2 Forwarding Sample Application (in Real and Virtualized Environments) and L2 Forwarding Eventdev Sample Application. The main difference from the L2 Forwarding sample application is that optionally packet can be Rx/Tx from/to eventdev instead of port directly and forwarding decision is made based on information read from the input packet.

Eventdev can optionally use S/W or H/W (if supported by platform) scheduler implementation for packet I/O based on run time parameters.

The lookup method is hash-based, LPM-based or FIB-based and is selected at run time. When the selected lookup method is hash-based, a hash object is used to emulate the flow classification stage. The hash object is used in correlation with a flow table to map each input packet to its flow at runtime.

The hash lookup key is represented by a DiffServ 5-tuple composed of the following fields read from the input packet: Source IP Address, Destination IP Address, Protocol, Source Port and Destination Port. The ID of the output interface for the input packet is read from the identified flow table entry. The set of flows used by the application is statically configured and loaded into the hash at initialization time. When the selected lookup method is LPM or FIB based, an LPM or FIB object is used to emulate the forwarding stage for IPv4 packets. The LPM or FIB object is used as the routing table to identify the next hop for each input packet at runtime.

The LPM and FIB lookup keys are represented by the destination IP address field read from the input packet. The ID of the output interface for the input packet is the next hop returned by the LPM or FIB lookup. The set of LPM and FIB rules used by the application is statically configured and loaded into the LPM or FIB object at initialization time.

In the sample application, hash-based, LPM-based, FIB-based and ACL-based forwarding supports both IPv4 and IPv6.

21.2. Compiling the Application

To compile the sample application see Compiling the Sample Applications.

The application is located in the l3fwd sub-directory.

21.3. Running the Application

The application has a number of command line options:

./dpdk-l3fwd [EAL options] -- -p PORTMASK
                         [-P]
                         [--lookup LOOKUP_METHOD]
                         --config(port,queue,lcore)[,(port,queue,lcore)]
                         [--eth-dest=X,MM:MM:MM:MM:MM:MM]
                         [--max-pkt-len PKTLEN]
                         [--no-numa]
                         [--hash-entry-num]
                         [--ipv6]
                         [--parse-ptype]
                         [--per-port-pool]
                         [--mode]
                         [--eventq-sched]
                         [--event-eth-rxqs]
                         [--event-vector [--event-vector-size SIZE] [--event-vector-tmo NS]]
                         [-E]
                         [-L]

Where,

• -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.
• --lookup: Optional, select the lookup method. Accepted options: em (Exact Match), lpm (Longest Prefix Match), fib (Forwarding Information Base). Default is lpm.
• --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.
• --max-pkt-len: Optional, maximum packet length in decimal (64-9600).
• --no-numa: Optional, disables numa awareness.
• --hash-entry-num: Optional, specifies the hash entry number in hexadecimal to be setup.
• --ipv6: Optional, set if running ipv6 packets.
• --parse-ptype: Optional, set to use software to analyze packet type. Without this option, hardware will check the packet type.
• --per-port-pool: Optional, set to use independent buffer pools per port. Without this option, single buffer pool is used for all ports.
• --mode: Optional, Packet transfer mode for I/O, poll or eventdev.
• --eventq-sched: Optional, Event queue synchronization method, Ordered, Atomic or Parallel. Only valid if –mode=eventdev.
• --event-eth-rxqs: Optional, Number of ethernet RX queues per device. Only valid if –mode=eventdev.
• --event-vector: Optional, Enable event vectorization. Only valid if –mode=eventdev.
• --event-vector-size: Optional, Max vector size if event vectorization is enabled.
• --event-vector-tmo: Optional, Max timeout to form vector in nanoseconds if event vectorization is enabled.
• -E: Optional, enable exact match, legacy flag, please use --lookup=em instead.
• -L: Optional, enable longest prefix match, legacy flag, please use --lookup=lpm instead.

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_dir>/examples/dpdk-l3fwd -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.

To use eventdev mode with sync method ordered on above mentioned environment, Following is the sample command:

./<build_dir>/examples/dpdk-l3fwd -l 0-3 -n 4 -a <event device> -- -p 0x3 --eventq-sched=ordered

or

./<build_dir>/examples/dpdk-l3fwd -l 0-3 -n 4 -a <event device> \
            -- -p 0x03 --mode=eventdev --eventq-sched=ordered

In this command:

• -a option allows the event device supported by platform. The syntax used to indicate this device may vary based on platform.
• The –mode option defines PMD to be used for packet I/O.
• The –eventq-sched option enables synchronization menthod of event queue so that packets will be scheduled accordingly.

If application uses S/W scheduler, it uses following DPDK services:

• Software scheduler
• Rx adapter service function
• Tx adapter service function

Application needs service cores to run above mentioned services. Service cores must be provided as EAL parameters along with the –vdev=event_sw0 to enable S/W scheduler. Following is the sample command:

./<build_dir>/examples/dpdk-l3fwd -l 0-7 -s 0xf0000 -n 4 --vdev event_sw0 -- -p 0x3 --mode=eventdev --eventq-sched=ordered

In case of eventdev mode, –config option is not used for ethernet port configuration. Instead each ethernet port will be configured with mentioned setup:

• Single Rx/Tx queue
• Each Rx queue will be connected to event queue via Rx adapter.
• Each Tx queue will be connected via Tx adapter.

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

21.4. Explanation

The following sections provide some explanation of the sample application code. As mentioned in the overview section, the initialization and run-time paths are very similar to those of the L2 Forwarding Sample Application (in Real and Virtualized Environments) and L2 Forwarding Eventdev Sample Application. The following sections describe aspects that are specific to the L3 Forwarding sample application.

21.4.1. Hash Initialization

The hash object is created and loaded with the pre-configured entries read from a global array, and then generate the expected 5-tuple as key to keep consistence with those of real flow for the convenience to execute hash performance test on 4M/8M/16M flows.

Note

The Hash initialization will setup both ipv4 and ipv6 hash table, and populate the either table depending on the value of variable ipv6. To support the hash performance test with up to 8M single direction flows/16M bi-direction flows, populate_ipv4_many_flow_into_table() function will populate the hash table with specified hash table entry number(default 4M).

Note

Value of global variable ipv6 can be specified with –ipv6 in the command line. Value of global variable hash_entry_number, which is used to specify the total hash entry number for all used ports in hash performance test, can be specified with –hash-entry-num VALUE in command line, being its default value 4.

#if (APP_LOOKUP_METHOD == APP_LOOKUP_EXACT_MATCH)

    static void
    setup_hash(int socketid)
    {
        // ...

        if (hash_entry_number != HASH_ENTRY_NUMBER_DEFAULT) {
            if (ipv6 == 0) {
                /* populate the ipv4 hash */
                populate_ipv4_many_flow_into_table(ipv4_l3fwd_lookup_struct[socketid], hash_entry_number);
            } else {
                /* populate the ipv6 hash */
                populate_ipv6_many_flow_into_table( ipv6_l3fwd_lookup_struct[socketid], hash_entry_number);
            }
        } else
            if (ipv6 == 0) {
                /* populate the ipv4 hash */
                populate_ipv4_few_flow_into_table(ipv4_l3fwd_lookup_struct[socketid]);
            } else {
                /* populate the ipv6 hash */
                populate_ipv6_few_flow_into_table(ipv6_l3fwd_lookup_struct[socketid]);
            }
        }
    }
#endif

21.4.2. LPM Initialization

The LPM object is created and loaded with the pre-configured entries read from a global array.

void
setup_hash(const int socketid)
{
	struct rte_hash_parameters ipv4_l3fwd_hash_params = {
		.name = NULL,
		.entries = L3FWD_HASH_ENTRIES,
		.key_len = sizeof(union ipv4_5tuple_host),
		.hash_func = ipv4_hash_crc,
		.hash_func_init_val = 0,
	};

	struct rte_hash_parameters ipv6_l3fwd_hash_params = {
		.name = NULL,
		.entries = L3FWD_HASH_ENTRIES,
		.key_len = sizeof(union ipv6_5tuple_host),
		.hash_func = ipv6_hash_crc,
		.hash_func_init_val = 0,
	};

	char s[64];

	/* create ipv4 hash */
	snprintf(s, sizeof(s), "ipv4_l3fwd_hash_%d", socketid);
	ipv4_l3fwd_hash_params.name = s;
	ipv4_l3fwd_hash_params.socket_id = socketid;
	ipv4_l3fwd_em_lookup_struct[socketid] =
		rte_hash_create(&ipv4_l3fwd_hash_params);
	if (ipv4_l3fwd_em_lookup_struct[socketid] == NULL)
		rte_exit(EXIT_FAILURE,
			"Unable to create the l3fwd hash on socket %d\n",
			socketid);

	/* create ipv6 hash */
	snprintf(s, sizeof(s), "ipv6_l3fwd_hash_%d", socketid);
	ipv6_l3fwd_hash_params.name = s;
	ipv6_l3fwd_hash_params.socket_id = socketid;
	ipv6_l3fwd_em_lookup_struct[socketid] =
		rte_hash_create(&ipv6_l3fwd_hash_params);
	if (ipv6_l3fwd_em_lookup_struct[socketid] == NULL)
		rte_exit(EXIT_FAILURE,
			"Unable to create the l3fwd hash on socket %d\n",
			socketid);

	if (hash_entry_number != HASH_ENTRY_NUMBER_DEFAULT) {
		/* For testing hash matching with a large number of flows we
		 * generate millions of IP 5-tuples with an incremented dst
		 * address to initialize the hash table. */
		if (ipv6 == 0) {
			/* populate the ipv4 hash */
			populate_ipv4_many_flow_into_table(
				ipv4_l3fwd_em_lookup_struct[socketid],
				hash_entry_number);
		} else {
			/* populate the ipv6 hash */
			populate_ipv6_many_flow_into_table(
				ipv6_l3fwd_em_lookup_struct[socketid],
				hash_entry_number);
		}
	} else {
		/*
		 * Use data in ipv4/ipv6 l3fwd lookup table
		 * directly to initialize the hash table.
		 */
		if (ipv6 == 0) {
			/* populate the ipv4 hash */
			populate_ipv4_few_flow_into_table(
				ipv4_l3fwd_em_lookup_struct[socketid]);
		} else {
			/* populate the ipv6 hash */
			populate_ipv6_few_flow_into_table(
				ipv6_l3fwd_em_lookup_struct[socketid]);
		}
	}
}

21.4.3. FIB Initialization

The FIB object is created and loaded with the pre-configured entries read from a global array. The abridged code snippet below shows the FIB initialization for IPv4, the full setup function including the IPv6 setup can be seen in the app code.

void
setup_fib(const int socketid)
{
	struct rte_eth_dev_info dev_info;
	struct rte_fib6_conf config;
	struct rte_fib_conf config_ipv4;
	unsigned int i;
	int ret;
	char s[64];
	char abuf[INET6_ADDRSTRLEN];

	/* Create the fib IPv4 table. */
	config_ipv4.type = RTE_FIB_DIR24_8;
	config_ipv4.max_routes = (1 << 16);
	config_ipv4.rib_ext_sz = 0;
	config_ipv4.default_nh = FIB_DEFAULT_HOP;
	config_ipv4.dir24_8.nh_sz = RTE_FIB_DIR24_8_4B;
	config_ipv4.dir24_8.num_tbl8 = (1 << 15);
	snprintf(s, sizeof(s), "IPV4_L3FWD_FIB_%d", socketid);
	ipv4_l3fwd_fib_lookup_struct[socketid] =
			rte_fib_create(s, socketid, &config_ipv4);
	if (ipv4_l3fwd_fib_lookup_struct[socketid] == NULL)
		rte_exit(EXIT_FAILURE,
			"Unable to create the l3fwd FIB table on socket %d\n",
			socketid);

	/* Populate the fib ipv4 table. */
	for (i = 0; i < RTE_DIM(ipv4_l3fwd_route_array); i++) {
		struct in_addr in;

		/* Skip unused ports. */
		if ((1 << ipv4_l3fwd_route_array[i].if_out &
				enabled_port_mask) == 0)
			continue;

		rte_eth_dev_info_get(ipv4_l3fwd_route_array[i].if_out,
				     &dev_info);
		ret = rte_fib_add(ipv4_l3fwd_fib_lookup_struct[socketid],
			ipv4_l3fwd_route_array[i].ip,
			ipv4_l3fwd_route_array[i].depth,
			ipv4_l3fwd_route_array[i].if_out);

		if (ret < 0) {
			rte_exit(EXIT_FAILURE,
					"Unable to add entry %u to the l3fwd FIB table on socket %d\n",
					i, socketid);
		}

		in.s_addr = htonl(ipv4_l3fwd_route_array[i].ip);
		if (inet_ntop(AF_INET, &in, abuf, sizeof(abuf)) != NULL) {
			printf("FIB: Adding route %s / %d (%d) [%s]\n", abuf,
			       ipv4_l3fwd_route_array[i].depth,
			       ipv4_l3fwd_route_array[i].if_out,
			       dev_info.device->name);
		} else {
			printf("FIB: IPv4 route added to port %d [%s]\n",
			       ipv4_l3fwd_route_array[i].if_out,
			       dev_info.device->name);
		}
	}

21.4.4. Packet Forwarding for Hash-based Lookups

For each input packet, the packet forwarding operation is done by the l3fwd_simple_forward() or simple_ipv4_fwd_4pkts() function for IPv4 packets or the simple_ipv6_fwd_4pkts() function for IPv6 packets. The l3fwd_simple_forward() function provides the basic functionality for both IPv4 and IPv6 packet forwarding for any number of burst packets received, and the packet forwarding decision (that is, the identification of the output interface for the packet) for hash-based lookups is done by the get_ipv4_dst_port() or get_ipv6_dst_port() function. The get_ipv4_dst_port() function is shown below:

static inline uint16_t
em_get_ipv4_dst_port(void *ipv4_hdr, uint16_t portid, void *lookup_struct)
{
	int ret = 0;
	union ipv4_5tuple_host key;
	struct rte_hash *ipv4_l3fwd_lookup_struct =
		(struct rte_hash *)lookup_struct;

	ipv4_hdr = (uint8_t *)ipv4_hdr +
		offsetof(struct rte_ipv4_hdr, time_to_live);

	/*
	 * Get 5 tuple: dst port, src port, dst IP address,
	 * src IP address and protocol.
	 */
	key.xmm = em_mask_key(ipv4_hdr, mask0.x);

	/* Find destination port */
	ret = rte_hash_lookup(ipv4_l3fwd_lookup_struct, (const void *)&key);
	return (ret < 0) ? portid : ipv4_l3fwd_out_if[ret];
}

The get_ipv6_dst_port() function is similar to the get_ipv4_dst_port() function.

The simple_ipv4_fwd_4pkts() and simple_ipv6_fwd_4pkts() function are optimized for continuous 4 valid ipv4 and ipv6 packets, they leverage the multiple buffer optimization to boost the performance of forwarding packets with the exact match on hash table. The key code snippet of simple_ipv4_fwd_4pkts() is shown below:

static inline void
simple_ipv4_fwd_4pkts(struct rte_mbuf* m[4], uint16_t portid, struct lcore_conf *qconf)
{
    // ...

    data[0] = _mm_loadu_si128(( m128i*)(rte_pktmbuf_mtod(m[0], unsigned char *) + sizeof(struct rte_ether_hdr) + offsetof(struct rte_ipv4_hdr, time_to_live)));
    data[1] = _mm_loadu_si128(( m128i*)(rte_pktmbuf_mtod(m[1], unsigned char *) + sizeof(struct rte_ether_hdr) + offsetof(struct rte_ipv4_hdr, time_to_live)));
    data[2] = _mm_loadu_si128(( m128i*)(rte_pktmbuf_mtod(m[2], unsigned char *) + sizeof(struct rte_ether_hdr) + offsetof(struct rte_ipv4_hdr, time_to_live)));
    data[3] = _mm_loadu_si128(( m128i*)(rte_pktmbuf_mtod(m[3], unsigned char *) + sizeof(struct rte_ether_hdr) + offsetof(struct rte_ipv4_hdr, time_to_live)));

    key[0].xmm = _mm_and_si128(data[0], mask0);
    key[1].xmm = _mm_and_si128(data[1], mask0);
    key[2].xmm = _mm_and_si128(data[2], mask0);
    key[3].xmm = _mm_and_si128(data[3], mask0);

    const void *key_array[4] = {&key[0], &key[1], &key[2],&key[3]};

    rte_hash_lookup_bulk(qconf->ipv4_lookup_struct, &key_array[0], 4, ret);

    dst_port[0] = (ret[0] < 0)? portid:ipv4_l3fwd_out_if[ret[0]];
    dst_port[1] = (ret[1] < 0)? portid:ipv4_l3fwd_out_if[ret[1]];
    dst_port[2] = (ret[2] < 0)? portid:ipv4_l3fwd_out_if[ret[2]];
    dst_port[3] = (ret[3] < 0)? portid:ipv4_l3fwd_out_if[ret[3]];

    // ...
}

The simple_ipv6_fwd_4pkts() function is similar to the simple_ipv4_fwd_4pkts() function.

Known issue: IP packets with extensions or IP packets which are not TCP/UDP cannot work well at this mode.

21.4.5. Packet Forwarding for LPM-based Lookups

For each input packet, the packet forwarding operation is done by the l3fwd_simple_forward() function, but the packet forwarding decision (that is, the identification of the output interface for the packet) for LPM-based lookups is done by the get_ipv4_dst_port() function below:

static inline uint16_t
lpm_get_ipv4_dst_port(const struct rte_ipv4_hdr *ipv4_hdr,
		      uint16_t portid,
		      struct rte_lpm *ipv4_l3fwd_lookup_struct)
{
	uint32_t dst_ip = rte_be_to_cpu_32(ipv4_hdr->dst_addr);
	uint32_t next_hop;

	if (rte_lpm_lookup(ipv4_l3fwd_lookup_struct, dst_ip, &next_hop) == 0)
		return next_hop;
	else
		return portid;
}

21.4.6. Packet Forwarding for FIB-based Lookups

The FIB library was designed to process multiple packets at once, it does not have separate functions for single and bulk lookups. rte_fib_lookup_bulk is used for IPv4 lookups and rte_fib6_lookup_bulk for IPv6. Various examples of these functions being used can be found in the sample app code.

21.4.7. Eventdev Driver Initialization

Eventdev driver initialization is same as L2 forwarding eventdev application. Refer L2 Forwarding Eventdev Sample Application for more details.