net/tap: add TUN/TAP device PMD
authorKeith Wiles <keith.wiles@intel.com>
Mon, 12 Dec 2016 14:38:38 +0000 (08:38 -0600)
committerFerruh Yigit <ferruh.yigit@intel.com>
Tue, 17 Jan 2017 18:40:50 +0000 (19:40 +0100)
The PMD allows for DPDK and the host to communicate using a raw
device interface on the host and in the DPDK application. The device
created is a Tap device with a L2 packet header.

Signed-off-by: Keith Wiles <keith.wiles@intel.com>
Reviewed-by: Ferruh Yigit <ferruh.yigit@intel.com>
Tested-by: Aws Ismail <aismail@ciena.com>
Tested-by: Vasily Philipov <vasilyf@mellanox.com>
MAINTAINERS
config/common_base
config/common_linuxapp
doc/guides/nics/index.rst
doc/guides/nics/tap.rst [new file with mode: 0644]
drivers/net/Makefile
drivers/net/tap/Makefile [new file with mode: 0644]
drivers/net/tap/rte_eth_tap.c [new file with mode: 0644]
drivers/net/tap/rte_pmd_tap_version.map [new file with mode: 0644]
mk/rte.app.mk

index 1dcee9d..8d0fe40 100644 (file)
@@ -399,6 +399,11 @@ M: Ferruh Yigit <ferruh.yigit@intel.com>
 F: drivers/net/pcap/
 F: doc/guides/nics/pcap_ring.rst
 
+Tap PMD
+M: Keith Wiles <keith.wiles@intel.com>
+F: drivers/net/tap/
+F: doc/guides/nics/tap.rst
+
 Ring PMD
 M: Bruce Richardson <bruce.richardson@intel.com>
 F: drivers/net/ring/
index 6c001c0..87d1719 100644 (file)
@@ -347,6 +347,12 @@ CONFIG_RTE_LIBRTE_QEDE_FW=""
 #
 CONFIG_RTE_LIBRTE_PMD_AF_PACKET=n
 
+#
+# Compile the TAP PMD
+# It is enabled by default for Linux only.
+#
+CONFIG_RTE_LIBRTE_PMD_TAP=n
+
 #
 # Compile Xen PMD
 #
index 2483dfa..1b0b0b4 100644 (file)
@@ -42,5 +42,6 @@ CONFIG_RTE_LIBRTE_KNI=y
 CONFIG_RTE_LIBRTE_VHOST=y
 CONFIG_RTE_LIBRTE_PMD_VHOST=y
 CONFIG_RTE_LIBRTE_PMD_AF_PACKET=y
+CONFIG_RTE_LIBRTE_PMD_TAP=y
 CONFIG_RTE_LIBRTE_POWER=y
 CONFIG_RTE_VIRTIO_USER=y
index 5ea8b1a..87f9334 100644 (file)
@@ -52,6 +52,7 @@ Network Interface Controller Drivers
     qede
     sfc_efx
     szedata2
+    tap
     thunderx
     virtio
     vhost
diff --git a/doc/guides/nics/tap.rst b/doc/guides/nics/tap.rst
new file mode 100644 (file)
index 0000000..622b9e7
--- /dev/null
@@ -0,0 +1,136 @@
+..  BSD LICENSE
+    Copyright(c) 2016 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.
+
+Tun/Tap Poll Mode Driver
+========================
+
+The ``rte_eth_tap.c`` PMD creates a device using TUN/TAP interfaces on the
+local host. The PMD allows for DPDK and the host to communicate using a raw
+device interface on the host and in the DPDK application.
+
+The device created is a TAP device, which sends/receives packet in a raw
+format with a L2 header. The usage for a TAP PMD is for connectivity to the
+local host using a TAP interface. When the TAP PMD is initialized it will
+create a number of tap devices in the host accessed via ``ifconfig -a`` or
+``ip`` command. The commands can be used to assign and query the virtual like
+device.
+
+These TAP interfaces can be used with Wireshark or tcpdump or Pktgen-DPDK
+along with being able to be used as a network connection to the DPDK
+application. The method enable one or more interfaces is to use the
+``--vdev=net_tap`` option on the DPDK application command line. Each
+``--vdev=net_tap`` option give will create an interface named dtap0, dtap1,
+and so on.
+
+The interfaced name can be changed by adding the ``iface=foo0``, for example::
+
+   --vdev=net_tap,iface=foo0 --vdev=net_tap,iface=foo1, ...
+
+Also the speed of the interface can be changed from 10G to whatever number
+needed, but the interface does not enforce that speed, for example::
+
+   --vdev=net_tap,iface=foo0,speed=25000
+
+After the DPDK application is started you can send and receive packets on the
+interface using the standard rx_burst/tx_burst APIs in DPDK. From the host
+point of view you can use any host tool like tcpdump, Wireshark, ping, Pktgen
+and others to communicate with the DPDK application. The DPDK application may
+not understand network protocols like IPv4/6, UDP or TCP unless the
+application has been written to understand these protocols.
+
+If you need the interface as a real network interface meaning running and has
+a valid IP address then you can do this with the following commands::
+
+   sudo ip link set dtap0 up; sudo ip addr add 192.168.0.250/24 dev dtap0
+   sudo ip link set dtap1 up; sudo ip addr add 192.168.1.250/24 dev dtap1
+
+Please change the IP addresses as you see fit.
+
+If routing is enabled on the host you can also communicate with the DPDK App
+over the internet via a standard socket layer application as long as you
+account for the protocol handing in the application.
+
+If you have a Network Stack in your DPDK application or something like it you
+can utilize that stack to handle the network protocols. Plus you would be able
+to address the interface using an IP address assigned to the internal
+interface.
+
+Example
+-------
+
+The following is a simple example of using the TUN/TAP PMD with the Pktgen
+packet generator. It requires that the ``socat`` utility is installed on the
+test system.
+
+Build DPDK, then pull down Pktgen and build pktgen using the DPDK SDK/Target
+used to build the dpdk you pulled down.
+
+Run pktgen from the pktgen directory in a terminal with a commandline like the
+following::
+
+    sudo ./app/app/x86_64-native-linuxapp-gcc/app/pktgen -l 1-5 -n 4        \
+     --proc-type auto --log-level 8 --socket-mem 512,512 --file-prefix pg   \
+     --vdev=net_tap --vdev=net_tap -b 05:00.0 -b 05:00.1                    \
+     -b 04:00.0 -b 04:00.1 -b 04:00.2 -b 04:00.3                            \
+     -b 81:00.0 -b 81:00.1 -b 81:00.2 -b 81:00.3                            \
+     -b 82:00.0 -b 83:00.0 -- -T -P -m [2:3].0 -m [4:5].1                   \
+     -f themes/black-yellow.theme
+
+.. Note:
+
+   Change the ``-b`` options to blacklist all of your physical ports. The
+   following command line is all one line.
+
+   Also, ``-f themes/black-yellow.theme`` is optional if the default colors
+   work on your system configuration. See the Pktgen docs for more
+   information.
+
+Verify with ``ifconfig -a`` command in a different xterm window, should have a
+``dtap0`` and ``dtap1`` interfaces created.
+
+Next set the links for the two interfaces to up via the commands below::
+
+    sudo ip link set dtap0 up; sudo ip addr add 192.168.0.250/24 dev dtap0
+    sudo ip link set dtap1 up; sudo ip addr add 192.168.1.250/24 dev dtap1
+
+Then use socat to create a loopback for the two interfaces::
+
+    sudo socat interface:dtap0 interface:dtap1
+
+Then on the Pktgen command line interface you can start sending packets using
+the commands ``start 0`` and ``start 1`` or you can start both at the same
+time with ``start all``. The command ``str`` is an alias for ``start all`` and
+``stp`` is an alias for ``stop all``.
+
+While running you should see the 64 byte counters increasing to verify the
+traffic is being looped back. You can use ``set all size XXX`` to change the
+size of the packets after you stop the traffic. Use the pktgen ``help``
+command to see a list of all commands. You can also use the ``-f`` option to
+load commands at startup.
index ccf464a..40fc333 100644 (file)
@@ -52,6 +52,7 @@ DIRS-$(CONFIG_RTE_LIBRTE_QEDE_PMD) += qede
 DIRS-$(CONFIG_RTE_LIBRTE_PMD_RING) += ring
 DIRS-$(CONFIG_RTE_LIBRTE_SFC_EFX_PMD) += sfc
 DIRS-$(CONFIG_RTE_LIBRTE_PMD_SZEDATA2) += szedata2
+DIRS-$(CONFIG_RTE_LIBRTE_PMD_TAP) += tap
 DIRS-$(CONFIG_RTE_LIBRTE_THUNDERX_NICVF_PMD) += thunderx
 DIRS-$(CONFIG_RTE_LIBRTE_VIRTIO_PMD) += virtio
 DIRS-$(CONFIG_RTE_LIBRTE_VMXNET3_PMD) += vmxnet3
diff --git a/drivers/net/tap/Makefile b/drivers/net/tap/Makefile
new file mode 100644 (file)
index 0000000..e18f30c
--- /dev/null
@@ -0,0 +1,57 @@
+#   BSD LICENSE
+#
+#   Copyright(c) 2016 Intel Corporation. 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 $(RTE_SDK)/mk/rte.vars.mk
+
+#
+# library name
+#
+LIB = librte_pmd_tap.a
+
+EXPORT_MAP := rte_pmd_tap_version.map
+
+LIBABIVER := 1
+
+CFLAGS += -O3
+CFLAGS += $(WERROR_FLAGS)
+
+#
+# all source are stored in SRCS-y
+#
+SRCS-$(CONFIG_RTE_LIBRTE_PMD_TAP) += rte_eth_tap.c
+
+# this lib depends upon:
+DEPDIRS-$(CONFIG_RTE_LIBRTE_PMD_TAP) += lib/librte_eal
+DEPDIRS-$(CONFIG_RTE_LIBRTE_PMD_TAP) += lib/librte_mbuf
+DEPDIRS-$(CONFIG_RTE_LIBRTE_PMD_TAP) += lib/librte_mempool
+DEPDIRS-$(CONFIG_RTE_LIBRTE_PMD_TAP) += lib/librte_ether
+DEPDIRS-$(CONFIG_RTE_LIBRTE_PMD_TAP) += lib/librte_kvargs
+
+include $(RTE_SDK)/mk/rte.lib.mk
diff --git a/drivers/net/tap/rte_eth_tap.c b/drivers/net/tap/rte_eth_tap.c
new file mode 100644 (file)
index 0000000..976f2d9
--- /dev/null
@@ -0,0 +1,765 @@
+/*-
+ *   BSD LICENSE
+ *
+ *   Copyright(c) 2016 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 <rte_mbuf.h>
+#include <rte_ethdev.h>
+#include <rte_malloc.h>
+#include <rte_vdev.h>
+#include <rte_kvargs.h>
+
+#include <sys/types.h>
+#include <sys/stat.h>
+#include <sys/socket.h>
+#include <sys/ioctl.h>
+#include <sys/mman.h>
+#include <unistd.h>
+#include <poll.h>
+#include <arpa/inet.h>
+#include <linux/if.h>
+#include <linux/if_tun.h>
+#include <linux/if_ether.h>
+#include <fcntl.h>
+
+/* Linux based path to the TUN device */
+#define TUN_TAP_DEV_PATH        "/dev/net/tun"
+#define DEFAULT_TAP_NAME        "dtap"
+
+#define ETH_TAP_IFACE_ARG       "iface"
+#define ETH_TAP_SPEED_ARG       "speed"
+
+#define RTE_PMD_TAP_MAX_QUEUES 16
+
+static struct rte_vdev_driver pmd_tap_drv;
+
+static const char *valid_arguments[] = {
+       ETH_TAP_IFACE_ARG,
+       ETH_TAP_SPEED_ARG,
+       NULL
+};
+
+static int tap_unit;
+
+static struct rte_eth_link pmd_link = {
+       .link_speed = ETH_SPEED_NUM_10G,
+       .link_duplex = ETH_LINK_FULL_DUPLEX,
+       .link_status = ETH_LINK_DOWN,
+       .link_autoneg = ETH_LINK_SPEED_AUTONEG
+};
+
+struct pkt_stats {
+       uint64_t opackets;              /* Number of output packets */
+       uint64_t ipackets;              /* Number of input packets */
+       uint64_t obytes;                /* Number of bytes on output */
+       uint64_t ibytes;                /* Number of bytes on input */
+       uint64_t errs;                  /* Number of error packets */
+};
+
+struct rx_queue {
+       struct rte_mempool *mp;         /* Mempool for RX packets */
+       uint16_t in_port;               /* Port ID */
+       int fd;
+
+       struct pkt_stats stats;         /* Stats for this RX queue */
+};
+
+struct tx_queue {
+       int fd;
+       struct pkt_stats stats;         /* Stats for this TX queue */
+};
+
+struct pmd_internals {
+       char name[RTE_ETH_NAME_MAX_LEN];        /* Internal Tap device name */
+       uint16_t nb_queues;             /* Number of queues supported */
+       struct ether_addr eth_addr;     /* Mac address of the device port */
+
+       int if_index;                   /* IF_INDEX for the port */
+       int fds[RTE_PMD_TAP_MAX_QUEUES]; /* List of all file descriptors */
+
+       struct rx_queue rxq[RTE_PMD_TAP_MAX_QUEUES];    /* List of RX queues */
+       struct tx_queue txq[RTE_PMD_TAP_MAX_QUEUES];    /* List of TX queues */
+};
+
+/* Tun/Tap allocation routine
+ *
+ * name is the number of the interface to use, unless NULL to take the host
+ * supplied name.
+ */
+static int
+tun_alloc(char *name)
+{
+       struct ifreq ifr;
+       unsigned int features;
+       int fd;
+
+       memset(&ifr, 0, sizeof(struct ifreq));
+
+       ifr.ifr_flags = IFF_TAP | IFF_NO_PI;
+       if (name && name[0])
+               strncpy(ifr.ifr_name, name, IFNAMSIZ);
+
+       fd = open(TUN_TAP_DEV_PATH, O_RDWR);
+       if (fd < 0) {
+               RTE_LOG(ERR, PMD, "Unable to create TAP interface");
+               goto error;
+       }
+
+       /* Grab the TUN features to verify we can work */
+       if (ioctl(fd, TUNGETFEATURES, &features) < 0) {
+               RTE_LOG(ERR, PMD, "Unable to get TUN/TAP features\n");
+               goto error;
+       }
+       RTE_LOG(DEBUG, PMD, "TUN/TAP Features %08x\n", features);
+
+       if (!(features & IFF_MULTI_QUEUE) && (RTE_PMD_TAP_MAX_QUEUES > 1)) {
+               RTE_LOG(DEBUG, PMD, "TUN/TAP device only one queue\n");
+               goto error;
+       } else if ((features & IFF_ONE_QUEUE) &&
+                       (RTE_PMD_TAP_MAX_QUEUES == 1)) {
+               ifr.ifr_flags |= IFF_ONE_QUEUE;
+               RTE_LOG(DEBUG, PMD, "Single queue only support\n");
+       } else {
+               ifr.ifr_flags |= IFF_MULTI_QUEUE;
+               RTE_LOG(DEBUG, PMD, "Multi-queue support for %d queues\n",
+                       RTE_PMD_TAP_MAX_QUEUES);
+       }
+
+       /* Set the TUN/TAP configuration and get the name if needed */
+       if (ioctl(fd, TUNSETIFF, (void *)&ifr) < 0) {
+               RTE_LOG(ERR, PMD, "Unable to set TUNSETIFF for %s\n",
+                       ifr.ifr_name);
+               perror("TUNSETIFF");
+               goto error;
+       }
+
+       /* Always set the file descriptor to non-blocking */
+       if (fcntl(fd, F_SETFL, O_NONBLOCK) < 0) {
+               RTE_LOG(ERR, PMD, "Unable to set to nonblocking\n");
+               perror("F_SETFL, NONBLOCK");
+               goto error;
+       }
+
+       /* If the name is different that new name as default */
+       if (name && strcmp(name, ifr.ifr_name))
+               snprintf(name, RTE_ETH_NAME_MAX_LEN - 1, "%s", ifr.ifr_name);
+
+       return fd;
+
+error:
+       if (fd > 0)
+               close(fd);
+       return -1;
+}
+
+/* Callback to handle the rx burst of packets to the correct interface and
+ * file descriptor(s) in a multi-queue setup.
+ */
+static uint16_t
+pmd_rx_burst(void *queue, struct rte_mbuf **bufs, uint16_t nb_pkts)
+{
+       int len;
+       struct rte_mbuf *mbuf;
+       struct rx_queue *rxq = queue;
+       uint16_t num_rx;
+       unsigned long num_rx_bytes = 0;
+
+       for (num_rx = 0; num_rx < nb_pkts; ) {
+               /* allocate the next mbuf */
+               mbuf = rte_pktmbuf_alloc(rxq->mp);
+               if (unlikely(!mbuf)) {
+                       RTE_LOG(WARNING, PMD, "Unable to allocate mbuf\n");
+                       break;
+               }
+
+               len = read(rxq->fd, rte_pktmbuf_mtod(mbuf, char *),
+                          rte_pktmbuf_tailroom(mbuf));
+               if (len <= 0) {
+                       rte_pktmbuf_free(mbuf);
+                       break;
+               }
+
+               mbuf->data_len = len;
+               mbuf->pkt_len = len;
+               mbuf->port = rxq->in_port;
+
+               /* account for the receive frame */
+               bufs[num_rx++] = mbuf;
+               num_rx_bytes += mbuf->pkt_len;
+       }
+       rxq->stats.ipackets += num_rx;
+       rxq->stats.ibytes += num_rx_bytes;
+
+       return num_rx;
+}
+
+/* Callback to handle sending packets from the tap interface
+ */
+static uint16_t
+pmd_tx_burst(void *queue, struct rte_mbuf **bufs, uint16_t nb_pkts)
+{
+       struct rte_mbuf *mbuf;
+       struct tx_queue *txq = queue;
+       struct pollfd pfd;
+       uint16_t num_tx = 0;
+       unsigned long num_tx_bytes = 0;
+       int i, n;
+
+       if (unlikely(nb_pkts == 0))
+               return 0;
+
+       pfd.events = POLLOUT;
+       pfd.fd = txq->fd;
+       for (i = 0; i < nb_pkts; i++) {
+               n = poll(&pfd, 1, 0);
+
+               if (n <= 0)
+                       break;
+
+               if (pfd.revents & POLLOUT) {
+                       /* copy the tx frame data */
+                       mbuf = bufs[num_tx];
+                       n = write(pfd.fd, rte_pktmbuf_mtod(mbuf, void*),
+                                 rte_pktmbuf_pkt_len(mbuf));
+                       if (n <= 0)
+                               break;
+
+                       num_tx++;
+                       num_tx_bytes += mbuf->pkt_len;
+                       rte_pktmbuf_free(mbuf);
+               }
+       }
+
+       txq->stats.opackets += num_tx;
+       txq->stats.errs += nb_pkts - num_tx;
+       txq->stats.obytes += num_tx_bytes;
+
+       return num_tx;
+}
+
+static int
+tap_dev_start(struct rte_eth_dev *dev)
+{
+       /* Force the Link up */
+       dev->data->dev_link.link_status = ETH_LINK_UP;
+
+       return 0;
+}
+
+/* This function gets called when the current port gets stopped.
+ */
+static void
+tap_dev_stop(struct rte_eth_dev *dev)
+{
+       int i;
+       struct pmd_internals *internals = dev->data->dev_private;
+
+       for (i = 0; i < internals->nb_queues; i++)
+               if (internals->fds[i] != -1)
+                       close(internals->fds[i]);
+
+       dev->data->dev_link.link_status = ETH_LINK_DOWN;
+}
+
+static int
+tap_dev_configure(struct rte_eth_dev *dev __rte_unused)
+{
+       return 0;
+}
+
+static void
+tap_dev_info(struct rte_eth_dev *dev, struct rte_eth_dev_info *dev_info)
+{
+       struct pmd_internals *internals = dev->data->dev_private;
+
+       dev_info->if_index = internals->if_index;
+       dev_info->max_mac_addrs = 1;
+       dev_info->max_rx_pktlen = (uint32_t)ETHER_MAX_VLAN_FRAME_LEN;
+       dev_info->max_rx_queues = internals->nb_queues;
+       dev_info->max_tx_queues = internals->nb_queues;
+       dev_info->min_rx_bufsize = 0;
+       dev_info->pci_dev = NULL;
+}
+
+static void
+tap_stats_get(struct rte_eth_dev *dev, struct rte_eth_stats *tap_stats)
+{
+       unsigned int i, imax;
+       unsigned long rx_total = 0, tx_total = 0, tx_err_total = 0;
+       unsigned long rx_bytes_total = 0, tx_bytes_total = 0;
+       const struct pmd_internals *pmd = dev->data->dev_private;
+
+       imax = (pmd->nb_queues < RTE_ETHDEV_QUEUE_STAT_CNTRS) ?
+               pmd->nb_queues : RTE_ETHDEV_QUEUE_STAT_CNTRS;
+
+       for (i = 0; i < imax; i++) {
+               tap_stats->q_ipackets[i] = pmd->rxq[i].stats.ipackets;
+               tap_stats->q_ibytes[i] = pmd->rxq[i].stats.ibytes;
+               rx_total += tap_stats->q_ipackets[i];
+               rx_bytes_total += tap_stats->q_ibytes[i];
+       }
+
+       for (i = 0; i < imax; i++) {
+               tap_stats->q_opackets[i] = pmd->txq[i].stats.opackets;
+               tap_stats->q_errors[i] = pmd->txq[i].stats.errs;
+               tap_stats->q_obytes[i] = pmd->txq[i].stats.obytes;
+               tx_total += tap_stats->q_opackets[i];
+               tx_err_total += tap_stats->q_errors[i];
+               tx_bytes_total += tap_stats->q_obytes[i];
+       }
+
+       tap_stats->ipackets = rx_total;
+       tap_stats->ibytes = rx_bytes_total;
+       tap_stats->opackets = tx_total;
+       tap_stats->oerrors = tx_err_total;
+       tap_stats->obytes = tx_bytes_total;
+}
+
+static void
+tap_stats_reset(struct rte_eth_dev *dev)
+{
+       int i;
+       struct pmd_internals *pmd = dev->data->dev_private;
+
+       for (i = 0; i < pmd->nb_queues; i++) {
+               pmd->rxq[i].stats.ipackets = 0;
+               pmd->rxq[i].stats.ibytes = 0;
+       }
+
+       for (i = 0; i < pmd->nb_queues; i++) {
+               pmd->txq[i].stats.opackets = 0;
+               pmd->txq[i].stats.errs = 0;
+               pmd->txq[i].stats.obytes = 0;
+       }
+}
+
+static void
+tap_dev_close(struct rte_eth_dev *dev __rte_unused)
+{
+}
+
+static void
+tap_rx_queue_release(void *queue)
+{
+       struct rx_queue *rxq = queue;
+
+       if (rxq && (rxq->fd > 0)) {
+               close(rxq->fd);
+               rxq->fd = -1;
+       }
+}
+
+static void
+tap_tx_queue_release(void *queue)
+{
+       struct tx_queue *txq = queue;
+
+       if (txq && (txq->fd > 0)) {
+               close(txq->fd);
+               txq->fd = -1;
+       }
+}
+
+static int
+tap_link_update(struct rte_eth_dev *dev __rte_unused,
+               int wait_to_complete __rte_unused)
+{
+       return 0;
+}
+
+static int
+tap_setup_queue(struct rte_eth_dev *dev,
+               struct pmd_internals *internals,
+               uint16_t qid)
+{
+       struct rx_queue *rx = &internals->rxq[qid];
+       struct tx_queue *tx = &internals->txq[qid];
+       int fd;
+
+       fd = rx->fd;
+       if (fd < 0) {
+               fd = tx->fd;
+               if (fd < 0) {
+                       RTE_LOG(INFO, PMD, "Add queue to TAP %s for qid %d\n",
+                               dev->data->name, qid);
+                       fd = tun_alloc(dev->data->name);
+                       if (fd < 0) {
+                               RTE_LOG(ERR, PMD, "tun_alloc(%s) failed\n",
+                                       dev->data->name);
+                               return -1;
+                       }
+               }
+       }
+       dev->data->rx_queues[qid] = rx;
+       dev->data->tx_queues[qid] = tx;
+
+       rx->fd = fd;
+       tx->fd = fd;
+
+       return fd;
+}
+
+static int
+tap_rx_queue_setup(struct rte_eth_dev *dev,
+                  uint16_t rx_queue_id,
+                  uint16_t nb_rx_desc __rte_unused,
+                  unsigned int socket_id __rte_unused,
+                  const struct rte_eth_rxconf *rx_conf __rte_unused,
+                  struct rte_mempool *mp)
+{
+       struct pmd_internals *internals = dev->data->dev_private;
+       uint16_t buf_size;
+       int fd;
+
+       if ((rx_queue_id >= internals->nb_queues) || !mp) {
+               RTE_LOG(ERR, PMD, "nb_queues %d mp %p\n",
+                       internals->nb_queues, mp);
+               return -1;
+       }
+
+       internals->rxq[rx_queue_id].mp = mp;
+       internals->rxq[rx_queue_id].in_port = dev->data->port_id;
+
+       /* Now get the space available for data in the mbuf */
+       buf_size = (uint16_t)(rte_pktmbuf_data_room_size(mp) -
+                               RTE_PKTMBUF_HEADROOM);
+
+       if (buf_size < ETH_FRAME_LEN) {
+               RTE_LOG(ERR, PMD,
+                       "%s: %d bytes will not fit in mbuf (%d bytes)\n",
+                       dev->data->name, ETH_FRAME_LEN, buf_size);
+               return -ENOMEM;
+       }
+
+       fd = tap_setup_queue(dev, internals, rx_queue_id);
+       if (fd == -1)
+               return -1;
+
+       internals->fds[rx_queue_id] = fd;
+       RTE_LOG(INFO, PMD, "RX TAP device name %s, qid %d on fd %d\n",
+               dev->data->name, rx_queue_id, internals->rxq[rx_queue_id].fd);
+
+       return 0;
+}
+
+static int
+tap_tx_queue_setup(struct rte_eth_dev *dev,
+                  uint16_t tx_queue_id,
+                  uint16_t nb_tx_desc __rte_unused,
+                  unsigned int socket_id __rte_unused,
+                  const struct rte_eth_txconf *tx_conf __rte_unused)
+{
+       struct pmd_internals *internals = dev->data->dev_private;
+       int ret;
+
+       if (tx_queue_id >= internals->nb_queues)
+               return -1;
+
+       ret = tap_setup_queue(dev, internals, tx_queue_id);
+       if (ret == -1)
+               return -1;
+
+       RTE_LOG(INFO, PMD, "TX TAP device name %s, qid %d on fd %d\n",
+               dev->data->name, tx_queue_id, internals->txq[tx_queue_id].fd);
+
+       return 0;
+}
+
+static const struct eth_dev_ops ops = {
+       .dev_start              = tap_dev_start,
+       .dev_stop               = tap_dev_stop,
+       .dev_close              = tap_dev_close,
+       .dev_configure          = tap_dev_configure,
+       .dev_infos_get          = tap_dev_info,
+       .rx_queue_setup         = tap_rx_queue_setup,
+       .tx_queue_setup         = tap_tx_queue_setup,
+       .rx_queue_release       = tap_rx_queue_release,
+       .tx_queue_release       = tap_tx_queue_release,
+       .link_update            = tap_link_update,
+       .stats_get              = tap_stats_get,
+       .stats_reset            = tap_stats_reset,
+};
+
+static int
+pmd_mac_address(int fd, struct rte_eth_dev *dev, struct ether_addr *addr)
+{
+       struct ifreq ifr;
+
+       if ((fd <= 0) || !dev || !addr)
+               return -1;
+
+       memset(&ifr, 0, sizeof(ifr));
+
+       if (ioctl(fd, SIOCGIFHWADDR, &ifr) == -1) {
+               RTE_LOG(ERR, PMD, "ioctl failed (SIOCGIFHWADDR) (%s)\n",
+                       ifr.ifr_name);
+               return -1;
+       }
+
+       /* Set the host based MAC address to this special MAC format */
+       ifr.ifr_hwaddr.sa_data[0] = 'T';
+       ifr.ifr_hwaddr.sa_data[1] = 'a';
+       ifr.ifr_hwaddr.sa_data[2] = 'p';
+       ifr.ifr_hwaddr.sa_data[3] = '-';
+       ifr.ifr_hwaddr.sa_data[4] = dev->data->port_id;
+       ifr.ifr_hwaddr.sa_data[5] = dev->data->numa_node;
+       if (ioctl(fd, SIOCSIFHWADDR, &ifr) == -1) {
+               RTE_LOG(ERR, PMD, "%s: ioctl failed (SIOCSIFHWADDR) (%s)\n",
+                       dev->data->name, ifr.ifr_name);
+               return -1;
+       }
+
+       /* Set the local application MAC address, needs to be different then
+        * the host based MAC address.
+        */
+       ifr.ifr_hwaddr.sa_data[0] = 'd';
+       ifr.ifr_hwaddr.sa_data[1] = 'n';
+       ifr.ifr_hwaddr.sa_data[2] = 'e';
+       ifr.ifr_hwaddr.sa_data[3] = 't';
+       ifr.ifr_hwaddr.sa_data[4] = dev->data->port_id;
+       ifr.ifr_hwaddr.sa_data[5] = dev->data->numa_node;
+       rte_memcpy(addr, ifr.ifr_hwaddr.sa_data, ETH_ALEN);
+
+       return 0;
+}
+
+static int
+eth_dev_tap_create(const char *name, char *tap_name)
+{
+       int numa_node = rte_socket_id();
+       struct rte_eth_dev *dev = NULL;
+       struct pmd_internals *pmd = NULL;
+       struct rte_eth_dev_data *data = NULL;
+       int i, fd = -1;
+
+       RTE_LOG(INFO, PMD,
+               "%s: Create TAP Ethernet device with %d queues on numa %u\n",
+                name, RTE_PMD_TAP_MAX_QUEUES, rte_socket_id());
+
+       data = rte_zmalloc_socket(tap_name, sizeof(*data), 0, numa_node);
+       if (!data) {
+               RTE_LOG(INFO, PMD, "Failed to allocate data\n");
+               goto error_exit;
+       }
+
+       pmd = rte_zmalloc_socket(tap_name, sizeof(*pmd), 0, numa_node);
+       if (!pmd) {
+               RTE_LOG(INFO, PMD, "Unable to allocate internal struct\n");
+               goto error_exit;
+       }
+
+       /* Use the name and not the tap_name */
+       dev = rte_eth_dev_allocate(tap_name);
+       if (!dev) {
+               RTE_LOG(INFO, PMD, "Unable to allocate device struct\n");
+               goto error_exit;
+       }
+
+       snprintf(pmd->name, sizeof(pmd->name), "%s", tap_name);
+
+       pmd->nb_queues = RTE_PMD_TAP_MAX_QUEUES;
+
+       /* Setup some default values */
+       data->dev_private = pmd;
+       data->port_id = dev->data->port_id;
+       data->dev_flags = RTE_ETH_DEV_DETACHABLE;
+       data->kdrv = RTE_KDRV_NONE;
+       data->drv_name = pmd_tap_drv.driver.name;
+       data->numa_node = numa_node;
+
+       data->dev_link = pmd_link;
+       data->mac_addrs = &pmd->eth_addr;
+       data->nb_rx_queues = pmd->nb_queues;
+       data->nb_tx_queues = pmd->nb_queues;
+
+       dev->data = data;
+       dev->dev_ops = &ops;
+       dev->driver = NULL;
+       dev->rx_pkt_burst = pmd_rx_burst;
+       dev->tx_pkt_burst = pmd_tx_burst;
+       snprintf(dev->data->name, sizeof(dev->data->name), "%s", name);
+
+       /* Create the first Tap device */
+       fd = tun_alloc(tap_name);
+       if (fd < 0) {
+               RTE_LOG(INFO, PMD, "tun_alloc() failed\n");
+               goto error_exit;
+       }
+
+       /* Presetup the fds to -1 as being not working */
+       for (i = 0; i < RTE_PMD_TAP_MAX_QUEUES; i++) {
+               pmd->fds[i] = -1;
+               pmd->rxq[i].fd = -1;
+               pmd->txq[i].fd = -1;
+       }
+
+       /* Take the TUN/TAP fd and place in the first location */
+       pmd->rxq[0].fd = fd;
+       pmd->txq[0].fd = fd;
+       pmd->fds[0] = fd;
+
+       if (pmd_mac_address(fd, dev, &pmd->eth_addr) < 0) {
+               RTE_LOG(INFO, PMD, "Unable to get MAC address\n");
+               goto error_exit;
+       }
+
+       return 0;
+
+error_exit:
+       RTE_PMD_DEBUG_TRACE("Unable to initialize %s\n", name);
+
+       rte_free(data);
+       rte_free(pmd);
+
+       rte_eth_dev_release_port(dev);
+
+       return -EINVAL;
+}
+
+static int
+set_interface_name(const char *key __rte_unused,
+                  const char *value,
+                  void *extra_args)
+{
+       char *name = (char *)extra_args;
+
+       if (value)
+               snprintf(name, RTE_ETH_NAME_MAX_LEN - 1, "%s", value);
+       else
+               snprintf(name, RTE_ETH_NAME_MAX_LEN - 1, "%s%d",
+                        DEFAULT_TAP_NAME, (tap_unit - 1));
+
+       return 0;
+}
+
+static int
+set_interface_speed(const char *key __rte_unused,
+                   const char *value,
+                   void *extra_args)
+{
+       *(int *)extra_args = (value) ? atoi(value) : ETH_SPEED_NUM_10G;
+
+       return 0;
+}
+
+/* Open a TAP interface device.
+ */
+static int
+rte_pmd_tap_probe(const char *name, const char *params)
+{
+       int ret;
+       struct rte_kvargs *kvlist = NULL;
+       int speed;
+       char tap_name[RTE_ETH_NAME_MAX_LEN];
+
+       speed = ETH_SPEED_NUM_10G;
+       snprintf(tap_name, sizeof(tap_name), "%s%d",
+                DEFAULT_TAP_NAME, tap_unit++);
+
+       RTE_LOG(INFO, PMD, "Initializing pmd_tap for %s as %s\n",
+               name, tap_name);
+
+       if (params && (params[0] != '\0')) {
+               RTE_LOG(INFO, PMD, "paramaters (%s)\n", params);
+
+               kvlist = rte_kvargs_parse(params, valid_arguments);
+               if (kvlist) {
+                       if (rte_kvargs_count(kvlist, ETH_TAP_SPEED_ARG) == 1) {
+                               ret = rte_kvargs_process(kvlist,
+                                                        ETH_TAP_SPEED_ARG,
+                                                        &set_interface_speed,
+                                                        &speed);
+                               if (ret == -1)
+                                       goto leave;
+                       }
+
+                       if (rte_kvargs_count(kvlist, ETH_TAP_IFACE_ARG) == 1) {
+                               ret = rte_kvargs_process(kvlist,
+                                                        ETH_TAP_IFACE_ARG,
+                                                        &set_interface_name,
+                                                        tap_name);
+                               if (ret == -1)
+                                       goto leave;
+                       }
+               }
+       }
+       pmd_link.link_speed = speed;
+
+       ret = eth_dev_tap_create(name, tap_name);
+
+leave:
+       if (ret == -1) {
+               RTE_LOG(INFO, PMD, "Failed to create pmd for %s as %s\n",
+                       name, tap_name);
+               tap_unit--;             /* Restore the unit number */
+       }
+       rte_kvargs_free(kvlist);
+
+       return ret;
+}
+
+/* detach a TAP device.
+ */
+static int
+rte_pmd_tap_remove(const char *name)
+{
+       struct rte_eth_dev *eth_dev = NULL;
+       struct pmd_internals *internals;
+       int i;
+
+       RTE_LOG(INFO, PMD, "Closing TUN/TAP Ethernet device on numa %u\n",
+               rte_socket_id());
+
+       /* find the ethdev entry */
+       eth_dev = rte_eth_dev_allocated(name);
+       if (!eth_dev)
+               return 0;
+
+       internals = eth_dev->data->dev_private;
+       for (i = 0; i < internals->nb_queues; i++)
+               if (internals->fds[i] != -1)
+                       close(internals->fds[i]);
+
+       rte_free(eth_dev->data->dev_private);
+       rte_free(eth_dev->data);
+
+       rte_eth_dev_release_port(eth_dev);
+
+       return 0;
+}
+
+static struct rte_vdev_driver pmd_tap_drv = {
+       .probe = rte_pmd_tap_probe,
+       .remove = rte_pmd_tap_remove,
+};
+RTE_PMD_REGISTER_VDEV(net_tap, pmd_tap_drv);
+RTE_PMD_REGISTER_ALIAS(net_tap, eth_tap);
+RTE_PMD_REGISTER_PARAM_STRING(net_tap, "iface=<string>,speed=N");
diff --git a/drivers/net/tap/rte_pmd_tap_version.map b/drivers/net/tap/rte_pmd_tap_version.map
new file mode 100644 (file)
index 0000000..31eca32
--- /dev/null
@@ -0,0 +1,4 @@
+DPDK_17.02 {
+
+       local: *;
+};
index 52ae046..40b40de 100644 (file)
@@ -125,6 +125,7 @@ _LDLIBS-$(CONFIG_RTE_LIBRTE_QEDE_PMD)       += -lrte_pmd_qede
 _LDLIBS-$(CONFIG_RTE_LIBRTE_PMD_RING)       += -lrte_pmd_ring
 _LDLIBS-$(CONFIG_RTE_LIBRTE_SFC_EFX_PMD)    += -lrte_pmd_sfc_efx
 _LDLIBS-$(CONFIG_RTE_LIBRTE_PMD_SZEDATA2)   += -lrte_pmd_szedata2 -lsze2
+_LDLIBS-$(CONFIG_RTE_LIBRTE_PMD_TAP)        += -lrte_pmd_tap
 _LDLIBS-$(CONFIG_RTE_LIBRTE_THUNDERX_NICVF_PMD) += -lrte_pmd_thunderx_nicvf -lm
 _LDLIBS-$(CONFIG_RTE_LIBRTE_VIRTIO_PMD)     += -lrte_pmd_virtio
 ifeq ($(CONFIG_RTE_LIBRTE_VHOST),y)