examples/ip_fragmentation: overhaul
authorAnatoly Burakov <anatoly.burakov@intel.com>
Wed, 28 May 2014 17:32:45 +0000 (18:32 +0100)
committerThomas Monjalon <thomas.monjalon@6wind.com>
Mon, 16 Jun 2014 16:55:05 +0000 (18:55 +0200)
New stuff:
* Support for regular traffic as well as IPv4 and IPv6
* Simplified config
* Routing table printed out on start
* Uses LPM/LPM6 for lookup
* Unmatched traffic is sent to the originating port

Signed-off-by: Anatoly Burakov <anatoly.burakov@intel.com>
Tested-by: Waterman Cao <waterman.cao@intel.com>
Acked-by: Thomas Monjalon <thomas.monjalon@6wind.com>
examples/ip_fragmentation/main.c

index 1376e6a..02e40a1 100644 (file)
 #include <rte_mempool.h>
 #include <rte_mbuf.h>
 #include <rte_lpm.h>
+#include <rte_lpm6.h>
 #include <rte_ip.h>
+#include <rte_string_fns.h>
 
-#include "rte_ip_frag.h"
-#include "main.h"
-
-/*
- * Default byte size for the IPv4 Maximum Transfer Unit (MTU).
- * This value includes the size of IPv4 header.
- */
-#define        IPV4_MTU_DEFAULT        ETHER_MTU
+#include <rte_ip_frag.h>
 
-/*
- * Default payload in bytes for the IPv4 packet.
- */
-#define        IPV4_DEFAULT_PAYLOAD    (IPV4_MTU_DEFAULT - sizeof(struct ipv4_hdr))
+#include "main.h"
 
-#define RTE_LOGTYPE_L3FWD RTE_LOGTYPE_USER1
+#define RTE_LOGTYPE_IP_FRAG RTE_LOGTYPE_USER1
 
 #define MBUF_SIZE (2048 + sizeof(struct rte_mbuf) + RTE_PKTMBUF_HEADROOM)
 
 #define        ROUNDUP_DIV(a, b)       (((a) + (b) - 1) / (b))
 
 /*
- * Max number of fragments per packet expected.
+ * Default byte size for the IPv6 Maximum Transfer Unit (MTU).
+ * This value includes the size of IPv6 header.
+ */
+#define        IPV4_MTU_DEFAULT        ETHER_MTU
+#define        IPV6_MTU_DEFAULT        ETHER_MTU
+
+/*
+ * Default payload in bytes for the IPv6 packet.
+ */
+#define        IPV4_DEFAULT_PAYLOAD    (IPV4_MTU_DEFAULT - sizeof(struct ipv4_hdr))
+#define        IPV6_DEFAULT_PAYLOAD    (IPV6_MTU_DEFAULT - sizeof(struct ipv6_hdr))
+
+/*
+ * Max number of fragments per packet expected - defined by config file.
  */
-#define        MAX_PACKET_FRAG ROUNDUP_DIV(JUMBO_FRAME_MAX_SIZE, IPV4_DEFAULT_PAYLOAD)
+#define        MAX_PACKET_FRAG RTE_LIBRTE_IP_FRAG_MAX_FRAG
 
 #define NB_MBUF   8192
 
@@ -136,8 +141,27 @@ static uint16_t nb_txd = RTE_TEST_TX_DESC_DEFAULT;
 
 /* ethernet addresses of ports */
 static struct ether_addr ports_eth_addr[RTE_MAX_ETHPORTS];
-static struct ether_addr remote_eth_addr =
-       {{0xaa, 0xbb, 0xcc, 0xdd, 0xee, 0xff}};
+
+#ifndef IPv4_BYTES
+#define IPv4_BYTES_FMT "%" PRIu8 ".%" PRIu8 ".%" PRIu8 ".%" PRIu8
+#define IPv4_BYTES(addr) \
+               (uint8_t) (((addr) >> 24) & 0xFF),\
+               (uint8_t) (((addr) >> 16) & 0xFF),\
+               (uint8_t) (((addr) >> 8) & 0xFF),\
+               (uint8_t) ((addr) & 0xFF)
+#endif
+
+#ifndef IPv6_BYTES
+#define IPv6_BYTES_FMT "%02x%02x:%02x%02x:%02x%02x:%02x%02x:"\
+                       "%02x%02x:%02x%02x:%02x%02x:%02x%02x"
+#define IPv6_BYTES(addr) \
+       addr[0],  addr[1], addr[2],  addr[3], \
+       addr[4],  addr[5], addr[6],  addr[7], \
+       addr[8],  addr[9], addr[10], addr[11],\
+       addr[12], addr[13],addr[14], addr[15]
+#endif
+
+#define IPV6_ADDR_LEN 16
 
 /* mask of enabled ports */
 static int enabled_port_mask = 0;
@@ -151,14 +175,21 @@ struct mbuf_table {
        struct rte_mbuf *m_table[MBUF_TABLE_SIZE];
 };
 
+struct rx_queue {
+       struct rte_mempool *direct_pool;
+       struct rte_mempool *indirect_pool;
+       struct rte_lpm *lpm;
+       struct rte_lpm6 *lpm6;
+       uint8_t portid;
+};
+
 #define MAX_RX_QUEUE_PER_LCORE 16
 #define MAX_TX_QUEUE_PER_PORT 16
 struct lcore_queue_conf {
        uint16_t n_rx_queue;
-       uint8_t rx_queue_list[MAX_RX_QUEUE_PER_LCORE];
        uint16_t tx_queue_id[RTE_MAX_ETHPORTS];
+       struct rx_queue rx_queue_list[MAX_RX_QUEUE_PER_LCORE];
        struct mbuf_table tx_mbufs[RTE_MAX_ETHPORTS];
-
 } __rte_cache_aligned;
 struct lcore_queue_conf lcore_queue_conf[RTE_MAX_LCORE];
 
@@ -167,7 +198,7 @@ static const struct rte_eth_conf port_conf = {
                .max_rx_pkt_len = JUMBO_FRAME_MAX_SIZE,
                .split_hdr_size = 0,
                .header_split   = 0, /**< Header Split disabled */
-               .hw_ip_checksum = 0, /**< IP checksum offload disabled */
+               .hw_ip_checksum = 1, /**< IP checksum offload enabled */
                .hw_vlan_filter = 0, /**< VLAN filtering disabled */
                .jumbo_frame    = 1, /**< Jumbo Frame Support enabled */
                .hw_strip_crc   = 0, /**< CRC stripped by hardware */
@@ -195,27 +226,61 @@ static const struct rte_eth_txconf tx_conf = {
        .tx_rs_thresh = 0, /* Use PMD default values */
 };
 
-struct rte_mempool *pool_direct = NULL, *pool_indirect = NULL;
-
-struct l3fwd_route {
+/*
+ * IPv4 forwarding table
+ */
+struct l3fwd_ipv4_route {
        uint32_t ip;
        uint8_t  depth;
        uint8_t  if_out;
 };
 
-struct l3fwd_route l3fwd_route_array[] = {
-       {IPv4(100,10,0,0), 16, 2},
-       {IPv4(100,20,0,0), 16, 2},
-       {IPv4(100,30,0,0), 16, 0},
-       {IPv4(100,40,0,0), 16, 0},
+struct l3fwd_ipv4_route l3fwd_ipv4_route_array[] = {
+               {IPv4(100,10,0,0), 16, 0},
+               {IPv4(100,20,0,0), 16, 1},
+               {IPv4(100,30,0,0), 16, 2},
+               {IPv4(100,40,0,0), 16, 3},
+               {IPv4(100,50,0,0), 16, 4},
+               {IPv4(100,60,0,0), 16, 5},
+               {IPv4(100,70,0,0), 16, 6},
+               {IPv4(100,80,0,0), 16, 7},
 };
 
-#define L3FWD_NUM_ROUTES \
-       (sizeof(l3fwd_route_array) / sizeof(l3fwd_route_array[0]))
+/*
+ * IPv6 forwarding table
+ */
+
+struct l3fwd_ipv6_route {
+       uint8_t ip[IPV6_ADDR_LEN];
+       uint8_t depth;
+       uint8_t if_out;
+};
+
+static struct l3fwd_ipv6_route l3fwd_ipv6_route_array[] = {
+       {{1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1}, 48, 0},
+       {{2,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1}, 48, 1},
+       {{3,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1}, 48, 2},
+       {{4,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1}, 48, 3},
+       {{5,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1}, 48, 4},
+       {{6,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1}, 48, 5},
+       {{7,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1}, 48, 6},
+       {{8,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1}, 48, 7},
+};
+
+#define LPM_MAX_RULES         1024
+#define LPM6_MAX_RULES         1024
+#define LPM6_NUMBER_TBL8S (1 << 16)
 
-#define L3FWD_LPM_MAX_RULES     1024
+struct rte_lpm6_config lpm6_config = {
+               .max_rules = LPM6_MAX_RULES,
+               .number_tbl8s = LPM6_NUMBER_TBL8S,
+               .flags = 0
+};
 
-struct rte_lpm *l3fwd_lpm = NULL;
+static struct rte_mempool *socket_direct_pool[RTE_MAX_NUMA_NODES];
+static struct rte_mempool *socket_indirect_pool[RTE_MAX_NUMA_NODES];
+static struct rte_lpm *socket_lpm[RTE_MAX_NUMA_NODES];
+static struct rte_lpm6 *socket_lpm6[RTE_MAX_NUMA_NODES];
 
 /* Send burst of packets on an output interface */
 static inline int
@@ -239,54 +304,108 @@ send_burst(struct lcore_queue_conf *qconf, uint16_t n, uint8_t port)
 }
 
 static inline void
-l3fwd_simple_forward(struct rte_mbuf *m, uint8_t port_in)
+l3fwd_simple_forward(struct rte_mbuf *m, struct lcore_queue_conf *qconf,
+               uint8_t queueid, uint8_t port_in)
 {
-       struct lcore_queue_conf *qconf;
-       struct ipv4_hdr *ip_hdr;
-       uint32_t i, len, lcore_id, ip_dst;
-       uint8_t next_hop, port_out;
+       struct rx_queue *rxq;
+       uint32_t i, len;
+       uint8_t next_hop, port_out, ipv6;
        int32_t len2;
 
-       lcore_id = rte_lcore_id();
-       qconf = &lcore_queue_conf[lcore_id];
+       ipv6 = 0;
+       rxq = &qconf->rx_queue_list[queueid];
+
+       /* by default, send everything back to the source port */
+       port_out = port_in;
 
        /* Remove the Ethernet header and trailer from the input packet */
        rte_pktmbuf_adj(m, (uint16_t)sizeof(struct ether_hdr));
 
-       /* Read the lookup key (i.e. ip_dst) from the input packet */
-       ip_hdr = rte_pktmbuf_mtod(m, struct ipv4_hdr *);
-       ip_dst = rte_be_to_cpu_32(ip_hdr->dst_addr);
-
-       /* Find destination port */
-       if (rte_lpm_lookup(l3fwd_lpm, ip_dst, &next_hop) == 0 &&
-                       (enabled_port_mask & 1 << next_hop) != 0)
-               port_out = next_hop;
-       else
-               port_out = port_in;
-
        /* Build transmission burst */
        len = qconf->tx_mbufs[port_out].len;
 
-       /* if we don't need to do any fragmentation */
-       if (likely (IPV4_MTU_DEFAULT  >= m->pkt.pkt_len)) {
+       /* if this is an IPv4 packet */
+       if (m->ol_flags & PKT_RX_IPV4_HDR) {
+               struct ipv4_hdr *ip_hdr;
+               uint32_t ip_dst;
+               /* Read the lookup key (i.e. ip_dst) from the input packet */
+               ip_hdr = rte_pktmbuf_mtod(m, struct ipv4_hdr *);
+               ip_dst = rte_be_to_cpu_32(ip_hdr->dst_addr);
+
+               /* Find destination port */
+               if (rte_lpm_lookup(rxq->lpm, ip_dst, &next_hop) == 0 &&
+                               (enabled_port_mask & 1 << next_hop) != 0) {
+                       port_out = next_hop;
+
+                       /* Build transmission burst for new port */
+                       len = qconf->tx_mbufs[port_out].len;
+               }
+
+               /* if we don't need to do any fragmentation */
+               if (likely (IPV4_MTU_DEFAULT >= m->pkt.pkt_len)) {
+                       qconf->tx_mbufs[port_out].m_table[len] = m;
+                       len2 = 1;
+               } else {
+                       len2 = rte_ipv4_fragment_packet(m,
+                               &qconf->tx_mbufs[port_out].m_table[len],
+                               (uint16_t)(MBUF_TABLE_SIZE - len),
+                               IPV4_MTU_DEFAULT,
+                               rxq->direct_pool, rxq->indirect_pool);
+
+                       /* Free input packet */
+                       rte_pktmbuf_free(m);
+
+                       /* If we fail to fragment the packet */
+                       if (unlikely (len2 < 0))
+                               return;
+               }
+       }
+       /* if this is an IPv6 packet */
+       else if (m->ol_flags & PKT_RX_IPV6_HDR) {
+               struct ipv6_hdr *ip_hdr;
+
+               ipv6 = 1;
+
+               /* Read the lookup key (i.e. ip_dst) from the input packet */
+               ip_hdr = rte_pktmbuf_mtod(m, struct ipv6_hdr *);
+
+               /* Find destination port */
+               if (rte_lpm6_lookup(rxq->lpm6, ip_hdr->dst_addr, &next_hop) == 0 &&
+                               (enabled_port_mask & 1 << next_hop) != 0) {
+                       port_out = next_hop;
+
+                       /* Build transmission burst for new port */
+                       len = qconf->tx_mbufs[port_out].len;
+               }
+
+               /* if we don't need to do any fragmentation */
+               if (likely (IPV6_MTU_DEFAULT >= m->pkt.pkt_len)) {
+                       qconf->tx_mbufs[port_out].m_table[len] = m;
+                       len2 = 1;
+               } else {
+                       len2 = rte_ipv6_fragment_packet(m,
+                               &qconf->tx_mbufs[port_out].m_table[len],
+                               (uint16_t)(MBUF_TABLE_SIZE - len),
+                               IPV6_MTU_DEFAULT,
+                               rxq->direct_pool, rxq->indirect_pool);
+
+                       /* Free input packet */
+                       rte_pktmbuf_free(m);
+
+                       /* If we fail to fragment the packet */
+                       if (unlikely (len2 < 0))
+                               return;
+               }
+       }
+       /* else, just forward the packet */
+       else {
                qconf->tx_mbufs[port_out].m_table[len] = m;
                len2 = 1;
-       } else {
-               len2 = rte_ipv4_fragment_packet(m,
-                       &qconf->tx_mbufs[port_out].m_table[len],
-                       (uint16_t)(MBUF_TABLE_SIZE - len),
-                       IPV4_MTU_DEFAULT,
-                       pool_direct, pool_indirect);
-
-               /* Free input packet */
-               rte_pktmbuf_free(m);
-
-               /* If we fail to fragment the packet */
-               if (unlikely (len2 < 0))
-                       return;
        }
 
        for (i = len; i < len + len2; i ++) {
+               void *d_addr_bytes;
+
                m = qconf->tx_mbufs[port_out].m_table[i];
                struct ether_hdr *eth_hdr = (struct ether_hdr *)
                        rte_pktmbuf_prepend(m, (uint16_t)sizeof(struct ether_hdr));
@@ -296,9 +415,16 @@ l3fwd_simple_forward(struct rte_mbuf *m, uint8_t port_in)
 
                m->pkt.vlan_macip.f.l2_len = sizeof(struct ether_hdr);
 
-               ether_addr_copy(&remote_eth_addr, &eth_hdr->d_addr);
+               /* 02:00:00:00:00:xx */
+               d_addr_bytes = &eth_hdr->d_addr.addr_bytes[0];
+               *((uint64_t *)d_addr_bytes) = 0x000000000002 + ((uint64_t)port_out << 40);
+
+               /* src addr */
                ether_addr_copy(&ports_eth_addr[port_out], &eth_hdr->s_addr);
-               eth_hdr->ether_type = rte_be_to_cpu_16(ETHER_TYPE_IPv4);
+               if (ipv6)
+                       eth_hdr->ether_type = rte_be_to_cpu_16(ETHER_TYPE_IPv6);
+               else
+                       eth_hdr->ether_type = rte_be_to_cpu_16(ETHER_TYPE_IPv4);
        }
 
        len += len2;
@@ -331,17 +457,17 @@ main_loop(__attribute__((unused)) void *dummy)
        qconf = &lcore_queue_conf[lcore_id];
 
        if (qconf->n_rx_queue == 0) {
-               RTE_LOG(INFO, L3FWD, "lcore %u has nothing to do\n", lcore_id);
+               RTE_LOG(INFO, IP_FRAG, "lcore %u has nothing to do\n", lcore_id);
                return 0;
        }
 
-       RTE_LOG(INFO, L3FWD, "entering main loop on lcore %u\n", lcore_id);
+       RTE_LOG(INFO, IP_FRAG, "entering main loop on lcore %u\n", lcore_id);
 
        for (i = 0; i < qconf->n_rx_queue; i++) {
 
-               portid = qconf->rx_queue_list[i];
-               RTE_LOG(INFO, L3FWD, " -- lcoreid=%u portid=%d\n", lcore_id,
-                       (int) portid);
+               portid = qconf->rx_queue_list[i].portid;
+               RTE_LOG(INFO, IP_FRAG, " -- lcoreid=%u portid=%d\n", lcore_id,
+                               (int) portid);
        }
 
        while (1) {
@@ -375,7 +501,7 @@ main_loop(__attribute__((unused)) void *dummy)
                 */
                for (i = 0; i < qconf->n_rx_queue; i++) {
 
-                       portid = qconf->rx_queue_list[i];
+                       portid = qconf->rx_queue_list[i].portid;
                        nb_rx = rte_eth_rx_burst(portid, 0, pkts_burst,
                                                 MAX_PKT_BURST);
 
@@ -389,12 +515,12 @@ main_loop(__attribute__((unused)) void *dummy)
                        for (j = 0; j < (nb_rx - PREFETCH_OFFSET); j++) {
                                rte_prefetch0(rte_pktmbuf_mtod(pkts_burst[
                                                j + PREFETCH_OFFSET], void *));
-                               l3fwd_simple_forward(pkts_burst[j], portid);
+                               l3fwd_simple_forward(pkts_burst[j], qconf, i, portid);
                        }
 
                        /* Forward remaining prefetched packets */
                        for (; j < nb_rx; j++) {
-                               l3fwd_simple_forward(pkts_burst[j], portid);
+                               l3fwd_simple_forward(pkts_burst[j], qconf, i, portid);
                        }
                }
        }
@@ -570,17 +696,164 @@ check_all_ports_link_status(uint8_t port_num, uint32_t port_mask)
                /* set the print_flag if all ports up or timeout */
                if (all_ports_up == 1 || count == (MAX_CHECK_TIME - 1)) {
                        print_flag = 1;
-                       printf("done\n");
+                       printf("\ndone\n");
                }
        }
 }
 
+static int
+init_routing_table(void)
+{
+       struct rte_lpm *lpm;
+       struct rte_lpm6 *lpm6;
+       int socket, ret;
+       unsigned i;
+
+       for (socket = 0; socket < RTE_MAX_NUMA_NODES; socket++) {
+               if (socket_lpm[socket]) {
+                       lpm = socket_lpm[socket];
+                       /* populate the LPM table */
+                       for (i = 0; i < RTE_DIM(l3fwd_ipv4_route_array); i++) {
+                               ret = rte_lpm_add(lpm,
+                                       l3fwd_ipv4_route_array[i].ip,
+                                       l3fwd_ipv4_route_array[i].depth,
+                                       l3fwd_ipv4_route_array[i].if_out);
+
+                               if (ret < 0) {
+                                       RTE_LOG(ERR, IP_FRAG, "Unable to add entry %i to the l3fwd "
+                                               "LPM table\n", i);
+                                       return -1;
+                               }
+
+                               RTE_LOG(INFO, IP_FRAG, "Socket %i: adding route " IPv4_BYTES_FMT
+                                               "/%d (port %d)\n",
+                                       socket,
+                                       IPv4_BYTES(l3fwd_ipv4_route_array[i].ip),
+                                       l3fwd_ipv4_route_array[i].depth,
+                                       l3fwd_ipv4_route_array[i].if_out);
+                       }
+               }
+
+               if (socket_lpm6[socket]) {
+                       lpm6 = socket_lpm6[socket];
+                       /* populate the LPM6 table */
+                       for (i = 0; i < RTE_DIM(l3fwd_ipv6_route_array); i++) {
+                               ret = rte_lpm6_add(lpm6,
+                                       l3fwd_ipv6_route_array[i].ip,
+                                       l3fwd_ipv6_route_array[i].depth,
+                                       l3fwd_ipv6_route_array[i].if_out);
+
+                               if (ret < 0) {
+                                       RTE_LOG(ERR, IP_FRAG, "Unable to add entry %i to the l3fwd "
+                                               "LPM6 table\n", i);
+                                       return -1;
+                               }
+
+                               RTE_LOG(INFO, IP_FRAG, "Socket %i: adding route " IPv6_BYTES_FMT
+                                               "/%d (port %d)\n",
+                                       socket,
+                                       IPv6_BYTES(l3fwd_ipv6_route_array[i].ip),
+                                       l3fwd_ipv6_route_array[i].depth,
+                                       l3fwd_ipv6_route_array[i].if_out);
+                       }
+               }
+       }
+       return 0;
+}
+
+static int
+init_mem(void)
+{
+       char buf[PATH_MAX];
+       struct rte_mempool *mp;
+       struct rte_lpm *lpm;
+       struct rte_lpm6 *lpm6;
+       int socket;
+       unsigned lcore_id;
+
+       /* traverse through lcores and initialize structures on each socket */
+
+       for (lcore_id = 0; lcore_id < RTE_MAX_LCORE; lcore_id++) {
+
+               if (rte_lcore_is_enabled(lcore_id) == 0)
+                       continue;
+
+               socket = rte_lcore_to_socket_id(lcore_id);
+
+               if (socket == SOCKET_ID_ANY)
+                       socket = 0;
+
+               if (socket_direct_pool[socket] == NULL) {
+                       RTE_LOG(INFO, IP_FRAG, "Creating direct mempool on socket %i\n",
+                                       socket);
+                       rte_snprintf(buf, sizeof(buf), "pool_direct_%i", socket);
+
+                       mp = rte_mempool_create(buf, NB_MBUF,
+                                                  MBUF_SIZE, 32,
+                                                  sizeof(struct rte_pktmbuf_pool_private),
+                                                  rte_pktmbuf_pool_init, NULL,
+                                                  rte_pktmbuf_init, NULL,
+                                                  socket, 0);
+                       if (mp == NULL) {
+                               RTE_LOG(ERR, IP_FRAG, "Cannot create direct mempool\n");
+                               return -1;
+                       }
+                       socket_direct_pool[socket] = mp;
+               }
+
+               if (socket_indirect_pool[socket] == NULL) {
+                       RTE_LOG(INFO, IP_FRAG, "Creating indirect mempool on socket %i\n",
+                                       socket);
+                       rte_snprintf(buf, sizeof(buf), "pool_indirect_%i", socket);
+
+                       mp = rte_mempool_create(buf, NB_MBUF,
+                                                          sizeof(struct rte_mbuf), 32,
+                                                          0,
+                                                          NULL, NULL,
+                                                          rte_pktmbuf_init, NULL,
+                                                          socket, 0);
+                       if (mp == NULL) {
+                               RTE_LOG(ERR, IP_FRAG, "Cannot create indirect mempool\n");
+                               return -1;
+                       }
+                       socket_indirect_pool[socket] = mp;
+               }
+
+               if (socket_lpm[socket] == NULL) {
+                       RTE_LOG(INFO, IP_FRAG, "Creating LPM table on socket %i\n", socket);
+                       rte_snprintf(buf, sizeof(buf), "IP_FRAG_LPM_%i", socket);
+
+                       lpm = rte_lpm_create(buf, socket, LPM_MAX_RULES, 0);
+                       if (lpm == NULL) {
+                               RTE_LOG(ERR, IP_FRAG, "Cannot create LPM table\n");
+                               return -1;
+                       }
+                       socket_lpm[socket] = lpm;
+               }
+
+               if (socket_lpm6[socket] == NULL) {
+                       RTE_LOG(INFO, IP_FRAG, "Creating LPM6 table on socket %i\n", socket);
+                       rte_snprintf(buf, sizeof(buf), "IP_FRAG_LPM_%i", socket);
+
+                       lpm6 = rte_lpm6_create("IP_FRAG_LPM6", socket, &lpm6_config);
+                       if (lpm6 == NULL) {
+                               RTE_LOG(ERR, IP_FRAG, "Cannot create LPM table\n");
+                               return -1;
+                       }
+                       socket_lpm6[socket] = lpm6;
+               }
+       }
+
+       return 0;
+}
+
 int
 MAIN(int argc, char **argv)
 {
        struct lcore_queue_conf *qconf;
-       int ret;
-       unsigned nb_ports, i;
+       struct rx_queue *rxq;
+       int socket, ret;
+       unsigned nb_ports;
        uint16_t queueid = 0;
        unsigned lcore_id = 0, rx_lcore_id = 0;
        uint32_t n_tx_queue, nb_lcores;
@@ -598,36 +871,21 @@ MAIN(int argc, char **argv)
        if (ret < 0)
                rte_exit(EXIT_FAILURE, "Invalid arguments");
 
-       /* create the mbuf pools */
-       pool_direct =
-               rte_mempool_create("pool_direct", NB_MBUF,
-                                  MBUF_SIZE, 32,
-                                  sizeof(struct rte_pktmbuf_pool_private),
-                                  rte_pktmbuf_pool_init, NULL,
-                                  rte_pktmbuf_init, NULL,
-                                  rte_socket_id(), 0);
-       if (pool_direct == NULL)
-               rte_panic("Cannot init direct mbuf pool\n");
-
-       pool_indirect =
-               rte_mempool_create("pool_indirect", NB_MBUF,
-                                  sizeof(struct rte_mbuf), 32,
-                                  0,
-                                  NULL, NULL,
-                                  rte_pktmbuf_init, NULL,
-                                  rte_socket_id(), 0);
-       if (pool_indirect == NULL)
-               rte_panic("Cannot init indirect mbuf pool\n");
-
        if (rte_eal_pci_probe() < 0)
                rte_panic("Cannot probe PCI\n");
 
        nb_ports = rte_eth_dev_count();
        if (nb_ports > RTE_MAX_ETHPORTS)
                nb_ports = RTE_MAX_ETHPORTS;
+       else if (nb_ports == 0)
+               rte_exit(EXIT_FAILURE, "No ports found!\n");
 
        nb_lcores = rte_lcore_count();
 
+       /* initialize structures (mempools, lpm etc.) */
+       if (init_mem() < 0)
+               rte_panic("Cannot initialize memory structures!\n");
+
        /* initialize all ports */
        for (portid = 0; portid < nb_ports; portid++) {
                /* skip ports that are not enabled */
@@ -648,11 +906,21 @@ MAIN(int argc, char **argv)
 
                        qconf = &lcore_queue_conf[rx_lcore_id];
                }
-               qconf->rx_queue_list[qconf->n_rx_queue] = portid;
+
+               socket = rte_eth_dev_socket_id(portid);
+               if (socket == SOCKET_ID_ANY)
+                       socket = 0;
+
+               rxq = &qconf->rx_queue_list[qconf->n_rx_queue];
+               rxq->portid = portid;
+               rxq->direct_pool = socket_direct_pool[socket];
+               rxq->indirect_pool = socket_indirect_pool[socket];
+               rxq->lpm = socket_lpm[socket];
+               rxq->lpm6 = socket_lpm6[socket];
                qconf->n_rx_queue++;
 
                /* init port */
-               printf("Initializing port %d on lcore %u... ", portid,
+               printf("Initializing port %d on lcore %u...", portid,
                       rx_lcore_id);
                fflush(stdout);
 
@@ -661,82 +929,73 @@ MAIN(int argc, char **argv)
                        n_tx_queue = MAX_TX_QUEUE_PER_PORT;
                ret = rte_eth_dev_configure(portid, 1, (uint16_t)n_tx_queue,
                                            &port_conf);
-               if (ret < 0)
+               if (ret < 0) {
+                       printf("\n");
                        rte_exit(EXIT_FAILURE, "Cannot configure device: "
                                "err=%d, port=%d\n",
                                ret, portid);
-
-               rte_eth_macaddr_get(portid, &ports_eth_addr[portid]);
-               print_ethaddr(" Address:", &ports_eth_addr[portid]);
-               printf(", ");
+               }
 
                /* init one RX queue */
-               queueid = 0;
-               printf("rxq=%d ", queueid);
-               fflush(stdout);
-               ret = rte_eth_rx_queue_setup(portid, queueid, nb_rxd,
-                                            rte_eth_dev_socket_id(portid), &rx_conf,
-                                            pool_direct);
-               if (ret < 0)
-                       rte_exit(EXIT_FAILURE, "rte_eth_tx_queue_setup: "
+               ret = rte_eth_rx_queue_setup(portid, 0, nb_rxd,
+                                            socket, &rx_conf,
+                                            socket_direct_pool[socket]);
+               if (ret < 0) {
+                       printf("\n");
+                       rte_exit(EXIT_FAILURE, "rte_eth_rx_queue_setup: "
                                "err=%d, port=%d\n",
                                ret, portid);
+               }
+
+               rte_eth_macaddr_get(portid, &ports_eth_addr[portid]);
+               print_ethaddr(" Address:", &ports_eth_addr[portid]);
+               printf("\n");
 
                /* init one TX queue per couple (lcore,port) */
                queueid = 0;
                for (lcore_id = 0; lcore_id < RTE_MAX_LCORE; lcore_id++) {
                        if (rte_lcore_is_enabled(lcore_id) == 0)
                                continue;
+
+                       socket = (int) rte_lcore_to_socket_id(lcore_id);
                        printf("txq=%u,%d ", lcore_id, queueid);
                        fflush(stdout);
                        ret = rte_eth_tx_queue_setup(portid, queueid, nb_txd,
-                                                    rte_eth_dev_socket_id(portid), &tx_conf);
-                       if (ret < 0)
+                                                    socket, &tx_conf);
+                       if (ret < 0) {
+                               printf("\n");
                                rte_exit(EXIT_FAILURE, "rte_eth_tx_queue_setup: "
                                        "err=%d, port=%d\n", ret, portid);
+                       }
 
                        qconf = &lcore_queue_conf[lcore_id];
                        qconf->tx_queue_id[portid] = queueid;
                        queueid++;
                }
 
+               printf("\n");
+       }
+
+       printf("\n");
+
+       /* start ports */
+       for (portid = 0; portid < nb_ports; portid++) {
+               if ((enabled_port_mask & (1 << portid)) == 0) {
+                       continue;
+               }
                /* Start device */
                ret = rte_eth_dev_start(portid);
                if (ret < 0)
-                       rte_exit(EXIT_FAILURE, "rte_eth_dev_start: "
-                               "err=%d, port=%d\n",
+                       rte_exit(EXIT_FAILURE, "rte_eth_dev_start: err=%d, port=%d\n",
                                ret, portid);
 
-               printf("done: ");
-
-               /* Set port in promiscuous mode */
                rte_eth_promiscuous_enable(portid);
        }
 
-       check_all_ports_link_status((uint8_t)nb_ports, enabled_port_mask);
-
-       /* create the LPM table */
-       l3fwd_lpm = rte_lpm_create("L3FWD_LPM", rte_socket_id(), L3FWD_LPM_MAX_RULES, 0);
-       if (l3fwd_lpm == NULL)
-               rte_panic("Unable to create the l3fwd LPM table\n");
-
-       /* populate the LPM table */
-       for (i = 0; i < L3FWD_NUM_ROUTES; i++) {
-               ret = rte_lpm_add(l3fwd_lpm,
-                       l3fwd_route_array[i].ip,
-                       l3fwd_route_array[i].depth,
-                       l3fwd_route_array[i].if_out);
-
-               if (ret < 0) {
-                       rte_panic("Unable to add entry %u to the l3fwd "
-                               "LPM table\n", i);
-               }
+       if (init_routing_table() < 0)
+               rte_exit(EXIT_FAILURE, "Cannot init routing table\n");
 
-               printf("Adding route 0x%08x / %d (%d)\n",
-                       (unsigned) l3fwd_route_array[i].ip,
-                       l3fwd_route_array[i].depth,
-                       l3fwd_route_array[i].if_out);
-       }
+       check_all_ports_link_status((uint8_t)nb_ports, enabled_port_mask);
 
        /* launch per-lcore init on every lcore */
        rte_eal_mp_remote_launch(main_loop, NULL, CALL_MASTER);