examples/tep_term: implement VXLAN processing
authorJijiang Liu <jijiang.liu@intel.com>
Mon, 22 Jun 2015 16:40:59 +0000 (00:40 +0800)
committerThomas Monjalon <thomas.monjalon@6wind.com>
Tue, 23 Jun 2015 09:31:46 +0000 (11:31 +0200)
Implement the VXLAN packet processing functions.

  - VXLAN port configuration

  - VXLAN tunnel setup

  - VXLAN tunnel destroying

  - VXLAN packet processing for Rx side

  - VXLAN packet processing for Tx side

Signed-off-by: Jijiang Liu <jijiang.liu@intel.com>
Signed-off-by: Thomas Long <thomas.long@intel.com>
examples/tep_termination/Makefile
examples/tep_termination/main.c
examples/tep_termination/main.h
examples/tep_termination/vxlan.c [new file with mode: 0644]
examples/tep_termination/vxlan.h
examples/tep_termination/vxlan_setup.c [new file with mode: 0644]
examples/tep_termination/vxlan_setup.h

index 0525a2c..c37e50f 100644 (file)
@@ -47,7 +47,7 @@ endif
 APP = tep_termination
 
 # all source are stored in SRCS-y
-SRCS-y := main.c
+SRCS-y := main.c vxlan_setup.c vxlan.c
 
 CFLAGS += -O3
 CFLAGS += $(WERROR_FLAGS)
index 618093a..8d539f3 100644 (file)
@@ -125,7 +125,7 @@ static uint32_t enabled_port_mask;
 static uint32_t nb_switching_cores;
 
 /* number of devices/queues to support*/
-uint32_t nb_devices;
+uint16_t nb_devices = 2;
 
 /* max ring descriptor, ixgbe, i40e, e1000 all are 4096. */
 #define MAX_RING_DESC 4096
@@ -136,6 +136,16 @@ struct vpool {
        uint32_t buf_size;
 } vpool_array[MAX_QUEUES+MAX_QUEUES];
 
+/* overlay packet operation */
+struct ol_switch_ops overlay_options = {
+       .port_configure = vxlan_port_init,
+       .tunnel_setup = vxlan_link,
+       .tunnel_destroy = vxlan_unlink,
+       .tx_handle = vxlan_tx_pkts,
+       .rx_handle = vxlan_rx_pkts,
+       .param_handle = NULL,
+};
+
 /* Enable stats. */
 uint32_t enable_stats = 0;
 /* Enable retries on RX. */
@@ -312,9 +322,8 @@ tep_termination_parse_args(int argc, char **argv)
                                                "Invalid argument for rx-retry [0|1]\n");
                                        tep_termination_usage(prgname);
                                        return -1;
-                               } else {
+                               } else
                                        enable_retry = ret;
-                               }
                        }
 
                        /* Specify the retries delay time (in useconds) on RX.*/
@@ -327,9 +336,8 @@ tep_termination_parse_args(int argc, char **argv)
                                                "Invalid argument for rx-retry-delay [0-N]\n");
                                        tep_termination_usage(prgname);
                                        return -1;
-                               } else {
+                               } else
                                        burst_rx_delay_time = ret;
-                               }
                        }
 
                        /* Specify the retries number on RX. */
@@ -342,9 +350,8 @@ tep_termination_parse_args(int argc, char **argv)
                                                "Invalid argument for rx-retry-num [0-N]\n");
                                        tep_termination_usage(prgname);
                                        return -1;
-                               } else {
+                               } else
                                        burst_rx_retry_num = ret;
-                               }
                        }
 
                        /* Enable/disable stats. */
@@ -357,9 +364,8 @@ tep_termination_parse_args(int argc, char **argv)
                                                        "Invalid argument for stats [0..N]\n");
                                        tep_termination_usage(prgname);
                                        return -1;
-                               } else {
+                               } else
                                        enable_stats = ret;
-                               }
                        }
 
                        /* Set character device basename. */
@@ -459,6 +465,10 @@ virtio_tx_route(struct vhost_dev *vdev, struct rte_mbuf *m)
 
        if (unlikely(len == MAX_PKT_BURST)) {
                m_table = (struct rte_mbuf **)tx_q->m_table;
+               ret = overlay_options.tx_handle(ports[0],
+                       (uint16_t)tx_q->txq_id, m_table,
+                       (uint16_t)tx_q->len);
+
                /* Free any buffers not handled by TX and update
                 * the port stats.
                 */
@@ -525,6 +535,10 @@ switch_worker(__rte_unused void *arg)
                                LOG_DEBUG(VHOST_DATA, "TX queue drained after "
                                        "timeout with burst size %u\n",
                                        tx_q->len);
+                               ret = overlay_options.tx_handle(ports[0],
+                                       (uint16_t)tx_q->txq_id,
+                                       (struct rte_mbuf **)tx_q->m_table,
+                                       (uint16_t)tx_q->len);
                                if (unlikely(ret < tx_q->len)) {
                                        do {
                                                rte_pktmbuf_free(tx_q->m_table[ret]);
@@ -559,6 +573,7 @@ switch_worker(__rte_unused void *arg)
 
                        if (unlikely(vdev->remove)) {
                                dev_ll = dev_ll->next;
+                               overlay_options.tunnel_destroy(vdev);
                                vdev->ready = DEVICE_SAFE_REMOVE;
                                continue;
                        }
@@ -584,6 +599,7 @@ switch_worker(__rte_unused void *arg)
                                                }
                                        }
 
+                                       ret_count = overlay_options.rx_handle(dev, pkts_burst, rx_count);
                                        if (enable_stats) {
                                                rte_atomic64_add(
                                                &dev_statistics[dev->device_fh].rx_total_atomic,
@@ -606,7 +622,8 @@ switch_worker(__rte_unused void *arg)
                                                pkts_burst, MAX_PKT_BURST);
                                /* If this is the first received packet we need to learn the MAC */
                                if (unlikely(vdev->ready == DEVICE_MAC_LEARNING) && tx_count) {
-                                       if (vdev->remove) {
+                                       if (vdev->remove ||
+                                               (overlay_options.tunnel_setup(vdev, pkts_burst[0]) == -1)) {
                                                while (tx_count)
                                                        rte_pktmbuf_free(pkts_burst[--tx_count]);
                                        }
@@ -1059,7 +1076,6 @@ main(int argc, char *argv[])
                        "but only %u port can be enabled\n", nb_ports,
                        MAX_SUP_PORTS);
        }
-
        /* Create the mbuf pool. */
        mbuf_pool = rte_mempool_create(
                        "MBUF_POOL",
@@ -1087,6 +1103,9 @@ main(int argc, char *argv[])
                                "Skipping disabled port %d\n", portid);
                        continue;
                }
+               if (overlay_options.port_configure(portid, mbuf_pool) != 0)
+                       rte_exit(EXIT_FAILURE,
+                               "Cannot initialize network ports\n");
        }
 
        /* Initialise all linked lists. */
@@ -1105,7 +1124,6 @@ main(int argc, char *argv[])
                rte_eal_remote_launch(switch_worker,
                        mbuf_pool, lcore_id);
        }
-
        rte_vhost_feature_disable(1ULL << VIRTIO_NET_F_MRG_RXBUF);
 
        /* Register CUSE device to handle IOCTLs. */
index ef843c1..20fa5b3 100644 (file)
@@ -34,6 +34,8 @@
 #ifndef _MAIN_H_
 #define _MAIN_H_
 
+#include <rte_ether.h>
+
 #ifdef DEBUG
 #define LOG_LEVEL RTE_LOG_DEBUG
 #define LOG_DEBUG(log_type, fmt, args...) RTE_LOG(DEBUG, log_type, fmt, ##args)
diff --git a/examples/tep_termination/vxlan.c b/examples/tep_termination/vxlan.c
new file mode 100644 (file)
index 0000000..2f143ac
--- /dev/null
@@ -0,0 +1,172 @@
+/*-
+ *   BSD LICENSE
+ *
+ *   Copyright(c) 2010-2015 Intel Corporation. All rights reserved.
+ *   All rights reserved.
+ *
+ *   Redistribution and use in source and binary forms, with or without
+ *   modification, are permitted provided that the following conditions
+ *   are met:
+ *
+ * Redistributions of source code must retain the above copyright
+ *       notice, this list of conditions and the following disclaimer.
+ *     * Redistributions in binary form must reproduce the above copyright
+ *       notice, this list of conditions and the following disclaimer in
+ *       the documentation and/or other materials provided with the
+ *       distribution.
+ *     * Neither the name of Intel Corporation nor the names of its
+ *       contributors may be used to endorse or promote products derived
+ *       from this software without specific prior written permission.
+ *
+ *   THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+ *   "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+ *   LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+ *   A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+ *   OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+ *   SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+ *   LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+ *   DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+ *   THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+ *   (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+ *   OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+ */
+#include <stdint.h>
+#include <rte_mbuf.h>
+#include <rte_hash_crc.h>
+#include <rte_byteorder.h>
+#include <rte_udp.h>
+#include <rte_tcp.h>
+#include <rte_sctp.h>
+
+#include "main.h"
+#include "vxlan.h"
+
+/**
+ * Parse an ethernet header to fill the ethertype, outer_l2_len, outer_l3_len and
+ * ipproto. This function is able to recognize IPv4/IPv6 with one optional vlan
+ * header.
+ */
+static void
+parse_ethernet(struct ether_hdr *eth_hdr, union tunnel_offload_info *info,
+               uint8_t *l4_proto)
+{
+       struct ipv4_hdr *ipv4_hdr;
+       struct ipv6_hdr *ipv6_hdr;
+       uint16_t ethertype;
+
+       info->outer_l2_len = sizeof(struct ether_hdr);
+       ethertype = rte_be_to_cpu_16(eth_hdr->ether_type);
+
+       if (ethertype == ETHER_TYPE_VLAN) {
+               struct vlan_hdr *vlan_hdr = (struct vlan_hdr *)(eth_hdr + 1);
+               info->outer_l2_len  += sizeof(struct vlan_hdr);
+               ethertype = rte_be_to_cpu_16(vlan_hdr->eth_proto);
+       }
+
+       switch (ethertype) {
+       case ETHER_TYPE_IPv4:
+               ipv4_hdr = (struct ipv4_hdr *)
+                       ((char *)eth_hdr + info->outer_l2_len);
+               info->outer_l3_len = sizeof(struct ipv4_hdr);
+               *l4_proto = ipv4_hdr->next_proto_id;
+               break;
+       case ETHER_TYPE_IPv6:
+               ipv6_hdr = (struct ipv6_hdr *)
+                       ((char *)eth_hdr + info->outer_l2_len);
+               info->outer_l3_len = sizeof(struct ipv6_hdr);
+               *l4_proto = ipv6_hdr->proto;
+               break;
+       default:
+               info->outer_l3_len = 0;
+               *l4_proto = 0;
+               break;
+       }
+}
+
+int
+decapsulation(struct rte_mbuf *pkt)
+{
+       uint8_t l4_proto = 0;
+       uint16_t outer_header_len;
+       struct udp_hdr *udp_hdr;
+       union tunnel_offload_info info = { .data = 0 };
+       struct ether_hdr *phdr = rte_pktmbuf_mtod(pkt, struct ether_hdr *);
+
+       parse_ethernet(phdr, &info, &l4_proto);
+       if (l4_proto != IPPROTO_UDP)
+               return -1;
+
+       udp_hdr = (struct udp_hdr *)((char *)phdr +
+               info.outer_l2_len + info.outer_l3_len);
+
+       /** check udp destination port, 4789 is the default vxlan port
+        * (rfc7348) or that the rx offload flag is set (i40e only
+        * currently)*/
+       if (udp_hdr->dst_port != rte_cpu_to_be_16(DEFAULT_VXLAN_PORT) &&
+                       (pkt->ol_flags & (PKT_RX_TUNNEL_IPV4_HDR |
+                               PKT_RX_TUNNEL_IPV6_HDR)) == 0)
+               return -1;
+       outer_header_len = info.outer_l2_len + info.outer_l3_len
+               + sizeof(struct udp_hdr) + sizeof(struct vxlan_hdr);
+
+       rte_pktmbuf_adj(pkt, outer_header_len);
+
+       return 0;
+}
+
+void
+encapsulation(struct rte_mbuf *m, uint8_t queue_id)
+{
+       uint vport_id;
+       uint64_t ol_flags = 0;
+       uint32_t old_len = m->pkt_len, hash;
+       struct ether_hdr *phdr = rte_pktmbuf_mtod(m, struct ether_hdr *);
+
+       /*Allocate space for new ethernet, IPv4, UDP and VXLAN headers*/
+       struct ether_hdr *pneth = (struct ether_hdr *) rte_pktmbuf_prepend(m,
+               sizeof(struct ether_hdr) + sizeof(struct ipv4_hdr)
+               + sizeof(struct udp_hdr) + sizeof(struct vxlan_hdr));
+
+       struct ipv4_hdr *ip = (struct ipv4_hdr *) &pneth[1];
+       struct udp_hdr *udp = (struct udp_hdr *) &ip[1];
+       struct vxlan_hdr *vxlan = (struct vxlan_hdr *) &udp[1];
+
+       /* convert TX queue ID to vport ID */
+       vport_id = queue_id - 1;
+
+       /* replace original Ethernet header with ours */
+       pneth = rte_memcpy(pneth, &app_l2_hdr[vport_id],
+               sizeof(struct ether_hdr));
+
+       /* copy in IP header */
+       ip = rte_memcpy(ip, &app_ip_hdr[vport_id],
+               sizeof(struct ipv4_hdr));
+       ip->total_length = rte_cpu_to_be_16(m->data_len
+                               - sizeof(struct ether_hdr));
+
+       /* outer IP checksum */
+       ol_flags |= PKT_TX_OUTER_IP_CKSUM;
+       ip->hdr_checksum = 0;
+
+       m->outer_l2_len = sizeof(struct ether_hdr);
+       m->outer_l3_len = sizeof(struct ipv4_hdr);
+
+       m->ol_flags |= ol_flags;
+
+       /*VXLAN HEADER*/
+       vxlan->vx_flags = rte_cpu_to_be_32(VXLAN_HF_VNI);
+       vxlan->vx_vni = rte_cpu_to_be_32(vxdev.out_key << 8);
+
+       /*UDP HEADER*/
+       udp->dgram_cksum = 0;
+       udp->dgram_len = rte_cpu_to_be_16(old_len
+                               + sizeof(struct udp_hdr)
+                               + sizeof(struct vxlan_hdr));
+
+       udp->dst_port = rte_cpu_to_be_16(vxdev.dst_port);
+       hash = rte_hash_crc(phdr, 2 * ETHER_ADDR_LEN, phdr->ether_type);
+       udp->src_port = rte_cpu_to_be_16((((uint64_t) hash * PORT_RANGE) >> 32)
+                                       + PORT_MIN);
+
+       return;
+}
index 8595eed..502bbaa 100644 (file)
@@ -34,6 +34,9 @@
 #ifndef _VXLAN_H_
 #define _VXLAN_H_
 
+#include <rte_ether.h>
+#include <rte_ip.h>
+
 #define PORT_MIN       49152
 #define PORT_MAX       65535
 #define PORT_RANGE ((PORT_MAX - PORT_MIN) + 1)
@@ -42,6 +45,9 @@
 #define VXLAN_HF_VNI 0x08000000
 #define DEFAULT_VXLAN_PORT 4789
 
+extern struct ipv4_hdr app_ip_hdr[VXLAN_N_PORTS];
+extern struct ether_hdr app_l2_hdr[VXLAN_N_PORTS];
+
 struct vxlan_port {
        uint32_t vport_id;           /**< VirtIO port id */
        uint32_t peer_ip;            /**< remote VTEP IP address */
@@ -57,4 +63,21 @@ struct vxlan_conf {
        struct vxlan_port port[VXLAN_N_PORTS]; /**< VXLAN configuration */
 } __rte_cache_aligned;
 
-#endif /* _MAIN_H_ */
+extern struct vxlan_conf vxdev;
+
+/* structure that caches offload info for the current packet */
+union tunnel_offload_info {
+       uint64_t data;
+       struct {
+               uint64_t l2_len:7; /**< L2 (MAC) Header Length. */
+               uint64_t l3_len:9; /**< L3 (IP) Header Length. */
+               uint64_t l4_len:8; /**< L4 Header Length. */
+               uint64_t outer_l2_len:7; /**< outer L2 Header Length */
+               uint64_t outer_l3_len:16; /**< outer L3 Header Length */
+       };
+} __rte_cache_aligned;
+
+int decapsulation(struct rte_mbuf *pkt);
+void encapsulation(struct rte_mbuf *m, uint8_t queue_id);
+
+#endif /* _VXLAN_H_ */
diff --git a/examples/tep_termination/vxlan_setup.c b/examples/tep_termination/vxlan_setup.c
new file mode 100644 (file)
index 0000000..3631e1c
--- /dev/null
@@ -0,0 +1,357 @@
+/*-
+ *   BSD LICENSE
+ *
+ *   Copyright(c) 2010-2015 Intel Corporation. All rights reserved.
+ *   All rights reserved.
+ *
+ *   Redistribution and use in source and binary forms, with or without
+ *   modification, are permitted provided that the following conditions
+ *   are met:
+ *
+ *     * Redistributions of source code must retain the above copyright
+ *       notice, this list of conditions and the following disclaimer.
+ *     * Redistributions in binary form must reproduce the above copyright
+ *       notice, this list of conditions and the following disclaimer in
+ *       the documentation and/or other materials provided with the
+ *       distribution.
+ *     * Neither the name of Intel Corporation nor the names of its
+ *       contributors may be used to endorse or promote products derived
+ *       from this software without specific prior written permission.
+ *
+ *   THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+ *   "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+ *   LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+ *   A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+ *   OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+ *   SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+ *   LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+ *   DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+ *   THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+ *   (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+ *   OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+ */
+
+#include <getopt.h>
+#include <linux/if_ether.h>
+#include <linux/if_vlan.h>
+#include <linux/virtio_net.h>
+#include <linux/virtio_ring.h>
+#include <sys/param.h>
+#include <unistd.h>
+
+#include <rte_ethdev.h>
+#include <rte_log.h>
+#include <rte_string_fns.h>
+#include <rte_mbuf.h>
+#include <rte_malloc.h>
+#include <rte_ip.h>
+#include <rte_udp.h>
+#include <rte_tcp.h>
+
+#include "main.h"
+#include "rte_virtio_net.h"
+#include "vxlan.h"
+#include "vxlan_setup.h"
+
+#define IPV4_HEADER_LEN 20
+#define UDP_HEADER_LEN  8
+#define VXLAN_HEADER_LEN 8
+
+#define IP_VERSION 0x40
+#define IP_HDRLEN  0x05 /* default IP header length == five 32-bits words. */
+#define IP_DEFTTL  64   /* from RFC 1340. */
+#define IP_VHL_DEF (IP_VERSION | IP_HDRLEN)
+
+#define IP_DN_FRAGMENT_FLAG 0x0040
+
+/* Used to compare MAC addresses. */
+#define MAC_ADDR_CMP 0xFFFFFFFFFFFFULL
+
+/* Configurable number of RX/TX ring descriptors */
+#define RTE_TEST_RX_DESC_DEFAULT 1024
+#define RTE_TEST_TX_DESC_DEFAULT 512
+
+/* VXLAN device */
+struct vxlan_conf vxdev;
+
+struct ipv4_hdr app_ip_hdr[VXLAN_N_PORTS];
+struct ether_hdr app_l2_hdr[VXLAN_N_PORTS];
+
+/* local VTEP IP address */
+uint8_t vxlan_multicast_ips[2][4] = { {239, 1, 1, 1 }, {239, 1, 2, 1 } };
+
+/* Remote VTEP IP address */
+uint8_t vxlan_overlay_ips[2][4] = { {192, 168, 10, 1}, {192, 168, 30, 1} };
+
+/* Remote VTEP MAC address */
+uint8_t peer_mac[6] = {0x00, 0x11, 0x01, 0x00, 0x00, 0x01};
+
+/* Options for configuring ethernet port */
+static const struct rte_eth_conf port_conf = {
+       .rxmode = {
+               .split_hdr_size = 0,
+               .header_split   = 0, /**< Header Split disabled */
+               .hw_ip_checksum = 0, /**< IP checksum offload disabled */
+               .hw_vlan_filter = 0, /**< VLAN filtering disabled */
+               .jumbo_frame    = 0, /**< Jumbo Frame Support disabled */
+               .hw_strip_crc   = 0, /**< CRC stripped by hardware */
+       },
+       .txmode = {
+               .mq_mode = ETH_MQ_TX_NONE,
+       },
+};
+
+/**
+ * The one or two device(s) that belongs to the same tenant ID can
+ * be assigned in a VM.
+ */
+const uint16_t tenant_id_conf[] = {
+       1000, 1000, 1001, 1001, 1002, 1002, 1003, 1003,
+       1004, 1004, 1005, 1005, 1006, 1006, 1007, 1007,
+       1008, 1008, 1009, 1009, 1010, 1010, 1011, 1011,
+       1012, 1012, 1013, 1013, 1014, 1014, 1015, 1015,
+       1016, 1016, 1017, 1017, 1018, 1018, 1019, 1019,
+       1020, 1020, 1021, 1021, 1022, 1022, 1023, 1023,
+       1024, 1024, 1025, 1025, 1026, 1026, 1027, 1027,
+       1028, 1028, 1029, 1029, 1030, 1030, 1031, 1031,
+};
+
+/**
+ * Initialises a given port using global settings and with the rx buffers
+ * coming from the mbuf_pool passed as parameter
+ */
+int
+vxlan_port_init(uint8_t port, struct rte_mempool *mbuf_pool)
+{
+       int retval;
+       uint16_t q;
+       struct rte_eth_dev_info dev_info;
+       uint16_t rx_rings, tx_rings = (uint16_t)rte_lcore_count();
+       const uint16_t rx_ring_size = RTE_TEST_RX_DESC_DEFAULT;
+       const uint16_t tx_ring_size = RTE_TEST_TX_DESC_DEFAULT;
+       struct rte_eth_rxconf *rxconf;
+       struct rte_eth_txconf *txconf;
+
+       rte_eth_dev_info_get(port, &dev_info);
+
+       if (dev_info.max_rx_queues > MAX_QUEUES) {
+               rte_exit(EXIT_FAILURE,
+                       "please define MAX_QUEUES no less than %u in %s\n",
+                       dev_info.max_rx_queues, __FILE__);
+       }
+
+       rxconf = &dev_info.default_rxconf;
+       txconf = &dev_info.default_txconf;
+       txconf->txq_flags = 0;
+
+       if (port >= rte_eth_dev_count())
+               return -1;
+
+       rx_rings = nb_devices;
+
+       /* Configure ethernet device. */
+       retval = rte_eth_dev_configure(port, rx_rings, tx_rings, &port_conf);
+       if (retval != 0)
+               return retval;
+
+       /* Setup the queues. */
+       for (q = 0; q < rx_rings; q++) {
+               retval = rte_eth_rx_queue_setup(port, q, rx_ring_size,
+                                               rte_eth_dev_socket_id(port),
+                                               rxconf,
+                                               mbuf_pool);
+               if (retval < 0)
+                       return retval;
+       }
+       for (q = 0; q < tx_rings; q++) {
+               retval = rte_eth_tx_queue_setup(port, q, tx_ring_size,
+                                               rte_eth_dev_socket_id(port),
+                                               txconf);
+               if (retval < 0)
+                       return retval;
+       }
+
+       /* Start the device. */
+       retval  = rte_eth_dev_start(port);
+       if (retval < 0)
+               return retval;
+
+       rte_eth_macaddr_get(port, &ports_eth_addr[port]);
+       RTE_LOG(INFO, PORT, "Port %u MAC: %02"PRIx8" %02"PRIx8" %02"PRIx8
+                       " %02"PRIx8" %02"PRIx8" %02"PRIx8"\n",
+                       (unsigned)port,
+                       ports_eth_addr[port].addr_bytes[0],
+                       ports_eth_addr[port].addr_bytes[1],
+                       ports_eth_addr[port].addr_bytes[2],
+                       ports_eth_addr[port].addr_bytes[3],
+                       ports_eth_addr[port].addr_bytes[4],
+                       ports_eth_addr[port].addr_bytes[5]);
+       return 0;
+}
+
+static int
+vxlan_rx_process(struct rte_mbuf *pkt)
+{
+       return  decapsulation(pkt);
+}
+
+static void
+vxlan_tx_process(uint8_t queue_id, struct rte_mbuf *pkt)
+{
+       encapsulation(pkt, queue_id);
+       return;
+}
+
+/*
+ * This function learns the MAC address of the device and set init
+ * L2 header and L3 header info.
+ */
+int
+vxlan_link(struct vhost_dev *vdev, struct rte_mbuf *m)
+{
+       int i;
+       struct ether_hdr *pkt_hdr;
+       struct virtio_net *dev = vdev->dev;
+       uint64_t portid = dev->device_fh;
+       struct ipv4_hdr *ip;
+
+       if (unlikely(portid > VXLAN_N_PORTS)) {
+               RTE_LOG(INFO, VHOST_DATA,
+                       "(%"PRIu64") WARNING: Not configuring device,"
+                       "as already have %d ports for VXLAN.",
+                       dev->device_fh, VXLAN_N_PORTS);
+               return -1;
+       }
+
+       /* Learn MAC address of guest device from packet */
+       pkt_hdr = rte_pktmbuf_mtod(m, struct ether_hdr *);
+       if (is_same_ether_addr(&(pkt_hdr->s_addr), &vdev->mac_address)) {
+               RTE_LOG(INFO, VHOST_DATA,
+                       "(%"PRIu64") WARNING: This device is using an existing"
+                       " MAC address and has not been registered.\n",
+                       dev->device_fh);
+               return -1;
+       }
+
+       for (i = 0; i < ETHER_ADDR_LEN; i++) {
+               vdev->mac_address.addr_bytes[i] =
+                       vxdev.port[portid].vport_mac.addr_bytes[i] =
+                       pkt_hdr->s_addr.addr_bytes[i];
+               vxdev.port[portid].peer_mac.addr_bytes[i] = peer_mac[i];
+       }
+
+       /* Print out inner MAC and VNI info. */
+       RTE_LOG(INFO, VHOST_DATA,
+               "(%d) MAC_ADDRESS %02x:%02x:%02x:%02x:%02x:%02x and VNI %d registered\n",
+               vdev->rx_q,
+               vdev->mac_address.addr_bytes[0],
+               vdev->mac_address.addr_bytes[1],
+               vdev->mac_address.addr_bytes[2],
+               vdev->mac_address.addr_bytes[3],
+               vdev->mac_address.addr_bytes[4],
+               vdev->mac_address.addr_bytes[5],
+               tenant_id_conf[vdev->rx_q]);
+
+       vxdev.port[portid].vport_id = portid;
+
+       for (i = 0; i < 4; i++) {
+               /* Local VTEP IP */
+               vxdev.port_ip |= vxlan_multicast_ips[portid][i] << (8 * i);
+               /* Remote VTEP IP */
+               vxdev.port[portid].peer_ip |=
+                       vxlan_overlay_ips[portid][i] << (8 * i);
+       }
+
+       vxdev.out_key = tenant_id_conf[vdev->rx_q];
+       ether_addr_copy(&vxdev.port[portid].peer_mac,
+                       &app_l2_hdr[portid].d_addr);
+       ether_addr_copy(&ports_eth_addr[0],
+                       &app_l2_hdr[portid].s_addr);
+       app_l2_hdr[portid].ether_type = rte_cpu_to_be_16(ETHER_TYPE_IPv4);
+
+       ip = &app_ip_hdr[portid];
+       ip->version_ihl = IP_VHL_DEF;
+       ip->type_of_service = 0;
+       ip->total_length = 0;
+       ip->packet_id = 0;
+       ip->fragment_offset = IP_DN_FRAGMENT_FLAG;
+       ip->time_to_live = IP_DEFTTL;
+       ip->next_proto_id = IPPROTO_UDP;
+       ip->hdr_checksum = 0;
+       ip->src_addr = vxdev.port_ip;
+       ip->dst_addr = vxdev.port[portid].peer_ip;
+
+       /* Set device as ready for RX. */
+       vdev->ready = DEVICE_RX;
+
+       return 0;
+}
+
+/**
+ * Removes cloud filter. Ensures that nothing is adding buffers to the RX
+ * queue before disabling RX on the device.
+ */
+void
+vxlan_unlink(struct vhost_dev *vdev)
+{
+       unsigned i = 0, rx_count;
+       struct rte_mbuf *pkts_burst[MAX_PKT_BURST];
+
+       if (vdev->ready == DEVICE_RX) {
+               for (i = 0; i < ETHER_ADDR_LEN; i++)
+                       vdev->mac_address.addr_bytes[i] = 0;
+
+               /* Clear out the receive buffers */
+               rx_count = rte_eth_rx_burst(ports[0],
+                               (uint16_t)vdev->rx_q,
+                               pkts_burst, MAX_PKT_BURST);
+
+               while (rx_count) {
+                       for (i = 0; i < rx_count; i++)
+                               rte_pktmbuf_free(pkts_burst[i]);
+
+                       rx_count = rte_eth_rx_burst(ports[0],
+                                       (uint16_t)vdev->rx_q,
+                                       pkts_burst, MAX_PKT_BURST);
+               }
+               vdev->ready = DEVICE_MAC_LEARNING;
+       }
+}
+
+/* Transmit packets after encapsulating */
+int
+vxlan_tx_pkts(uint8_t port_id, uint16_t queue_id,
+               struct rte_mbuf **tx_pkts, uint16_t nb_pkts) {
+       int ret = 0;
+       uint16_t i;
+
+       for (i = 0; i < nb_pkts; i++)
+               vxlan_tx_process(queue_id, tx_pkts[i]);
+
+       ret = rte_eth_tx_burst(port_id, queue_id, tx_pkts, nb_pkts);
+
+       return ret;
+}
+
+/* Check for decapsulation and pass packets directly to VIRTIO device */
+int
+vxlan_rx_pkts(struct virtio_net *dev, struct rte_mbuf **pkts_burst,
+               uint32_t rx_count)
+{
+       uint32_t i = 0;
+       uint32_t count = 0;
+       int ret;
+       struct rte_mbuf *pkts_valid[rx_count];
+
+       for (i = 0; i < rx_count; i++) {
+               ret = vxlan_rx_process(pkts_burst[i]);
+               if (unlikely(ret < 0))
+                       continue;
+
+               pkts_valid[count] = pkts_burst[i];
+                       count++;
+       }
+
+       ret = rte_vhost_enqueue_burst(dev, VIRTIO_RXQ, pkts_valid, count);
+       return ret;
+}
index 0679876..d7e9a29 100644 (file)
 #ifndef VXLAN_SETUP_H_
 #define VXLAN_SETUP_H_
 
+extern uint16_t nb_devices;
+extern uint8_t ports[RTE_MAX_ETHPORTS];
+extern struct ether_addr ports_eth_addr[RTE_MAX_ETHPORTS];
+
 typedef int (*ol_port_configure_t)(uint8_t port,
                                   struct rte_mempool *mbuf_pool);