vhost: add driver on top of the library
authorTetsuya Mukawa <mukawa@igel.co.jp>
Mon, 21 Mar 2016 05:45:08 +0000 (14:45 +0900)
committerThomas Monjalon <thomas.monjalon@6wind.com>
Fri, 25 Mar 2016 17:56:44 +0000 (18:56 +0100)
The patch introduces a new PMD. This PMD is implemented as thin wrapper
of librte_vhost. It means librte_vhost is also needed to compile the PMD.
The vhost messages will be handled only when a port is started. So start
a port first, then invoke QEMU.

The PMD has 2 parameters.
 - iface:  The parameter is used to specify a path to connect to a
           virtio-net device.
 - queues: The parameter is used to specify the number of the queues
           virtio-net device has.
           (Default: 1)

Here is an example.
$ ./testpmd -c f -n 4 --vdev 'eth_vhost0,iface=/tmp/sock0,queues=1' -- -i

To connect above testpmd, here is qemu command example.

$ qemu-system-x86_64 \
        <snip>
        -chardev socket,id=chr0,path=/tmp/sock0 \
        -netdev vhost-user,id=net0,chardev=chr0,vhostforce,queues=1 \
        -device virtio-net-pci,netdev=net0,mq=on

Signed-off-by: Tetsuya Mukawa <mukawa@igel.co.jp>
Acked-by: Ferruh Yigit <ferruh.yigit@intel.com>
Acked-by: Yuanhan Liu <yuanhan.liu@linux.intel.com>
Acked-by: Rich Lane <rich.lane@bigswitch.com>
Tested-by: Rich Lane <rich.lane@bigswitch.com>
Update for queue state event name:
Signed-off-by: Bruce Richardson <bruce.richardson@intel.com>
13 files changed:
MAINTAINERS
config/common_base
config/common_linuxapp
doc/guides/nics/index.rst
doc/guides/nics/overview.rst
doc/guides/nics/vhost.rst [new file with mode: 0644]
doc/guides/rel_notes/release_16_04.rst
drivers/net/Makefile
drivers/net/vhost/Makefile [new file with mode: 0644]
drivers/net/vhost/rte_eth_vhost.c [new file with mode: 0644]
drivers/net/vhost/rte_eth_vhost.h [new file with mode: 0644]
drivers/net/vhost/rte_pmd_vhost_version.map [new file with mode: 0644]
mk/rte.app.mk

index 3f6920c..6ed54dd 100644 (file)
@@ -343,6 +343,11 @@ M: Yong Wang <yongwang@vmware.com>
 F: drivers/net/vmxnet3/
 F: doc/guides/nics/vmxnet3.rst
 
+Vhost PMD
+M: Tetsuya Mukawa <mukawa@igel.co.jp>
+M: Yuanhan Liu <yuanhan.liu@linux.intel.com>
+F: drivers/net/vhost/
+
 PCAP PMD
 M: Nicolás Pernas Maradei <nicolas.pernas.maradei@emutex.com>
 M: John McNamara <john.mcnamara@intel.com>
index 7d5e956..d98a82c 100644 (file)
@@ -523,6 +523,12 @@ CONFIG_RTE_LIBRTE_VHOST_USER=y
 CONFIG_RTE_LIBRTE_VHOST_NUMA=n
 CONFIG_RTE_LIBRTE_VHOST_DEBUG=n
 
+#
+# Compile vhost PMD
+# To compile, CONFIG_RTE_LIBRTE_VHOST should be enabled.
+#
+CONFIG_RTE_LIBRTE_PMD_VHOST=n
+
 #
 #Compile Xen domain0 support
 #
index ffbe260..7e698e2 100644 (file)
@@ -40,5 +40,6 @@ CONFIG_RTE_EAL_VFIO=y
 CONFIG_RTE_KNI_KMOD=y
 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_POWER=y
index d060265..769f677 100644 (file)
@@ -50,6 +50,7 @@ Network Interface Controller Drivers
     nfp
     szedata2
     virtio
+    vhost
     vmxnet3
     pcap_ring
 
index 28237e9..542479a 100644 (file)
@@ -74,20 +74,21 @@ Most of these differences are summarized below.
 
 .. table:: Features availability in networking drivers
 
-   ==================== = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = =
-   Feature              a b b b c e e e i i i i i i i i i i f f m m m n n p r s v v v x
-                        f n n o x 1 n n 4 4 4 4 g g x x x x m m l l p f u c i z i i m e
-                        p x x n g 0 a i 0 0 0 0 b b g g g g 1 1 x x i p l a n e r r x n
-                        a 2 2 d b 0   c e e e e   v b b b b 0 0 4 5 p   l p g d t t n v
-                        c x x i e 0       . v v   f e e e e k k     e         a i i e i
-                        k   v n           . f f       . v v   .               t o o t r
-                        e   f g           .   .       . f f   .               a   . 3 t
-                        t                 v   v       v   v   v               2   v
-                                          e   e       e   e   e                   e
-                                          c   c       c   c   c                   c
-   ==================== = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = =
-   link status                  X       X X                                   X
-   link status event                    X X
+   ==================== = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = =
+   Feature              a b b b c e e e i i i i i i i i i i f f m m m n n p r s v v v v x
+                        f n n o x 1 n n 4 4 4 4 g g x x x x m m l l p f u c i z h i i m e
+                        p x x n g 0 a i 0 0 0 0 b b g g g g 1 1 x x i p l a n e o r r x n
+                        a 2 2 d b 0   c e e e e   v b b b b 0 0 4 5 p   l p g d s t t n v
+                        c x x i e 0       . v v   f e e e e k k     e         a t i i e i
+                        k   v n           . f f       . v v   .               t   o o t r
+                        e   f g           .   .       . f f   .               a     . 3 t
+                        t                 v   v       v   v   v               2     v
+                                          e   e       e   e   e                     e
+                                          c   c       c   c   c                     c
+   ==================== = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = =
+   link status                  X       X X                                   X X
+   link status event                    X X                                     X
+   queue status event                                                           X
    Rx interrupt                         X X X X
    queue start/stop             X   X   X X X X                               X
    MTU update                   X   X
@@ -125,7 +126,7 @@ Most of these differences are summarized below.
    inner L4 checksum                X   X   X
    packet type parsing          X       X   X
    timesync                             X X
-   basic stats                  X   X   X X X X                               X
+   basic stats                  X   X   X X X X                               X X
    extended stats                   X   X X X X
    stats per queue              X                                             X
    EEPROM dump
@@ -139,9 +140,9 @@ Most of these differences are summarized below.
    ARMv8
    Power8
    TILE-Gx
-   x86-32                       X   X   X X X X
-   x86-64                       X   X   X X X X                               X
+   x86-32                       X   X   X X X X                                 X
+   x86-64                       X   X   X X X X                               X X
    usage doc                    X                                             X
    design doc
    perf doc
-   ==================== = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = =
+   ==================== = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = =
diff --git a/doc/guides/nics/vhost.rst b/doc/guides/nics/vhost.rst
new file mode 100644 (file)
index 0000000..1e3f1ad
--- /dev/null
@@ -0,0 +1,110 @@
+..  BSD LICENSE
+    Copyright(c) 2016 IGEL Co., Ltd.. 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 IGEL Co., Ltd. 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.
+
+Poll Mode Driver that wraps vhost library
+=========================================
+
+This PMD is a thin wrapper of the DPDK vhost library.
+The user can handle virtqueues as one of normal DPDK port.
+
+Vhost Implementation in DPDK
+----------------------------
+
+Please refer to Chapter "Vhost Library" of *DPDK Programmer's Guide* to know detail of vhost.
+
+Features and Limitations of vhost PMD
+-------------------------------------
+
+Currently, the vhost PMD provides the basic functionality of packet reception, transmission and event handling.
+
+*   It has multiple queues support.
+
+*   It supports ``RTE_ETH_EVENT_INTR_LSC`` and ``RTE_ETH_EVENT_QUEUE_STATE`` events.
+
+*   It supports Port Hotplug functionality.
+
+*   Don't need to stop RX/TX, when the user wants to stop a guest or a virtio-net driver on guest.
+
+Vhost PMD arguments
+-------------------
+
+The user can specify below arguments in `--vdev` option.
+
+#.  ``iface``:
+
+    It is used to specify a path to connect to a QEMU virtio-net device.
+
+#.  ``queues``:
+
+    It is used to specify the number of queues virtio-net device has.
+    (Default: 1)
+
+Vhost PMD event handling
+------------------------
+
+This section describes how to handle vhost PMD events.
+
+The user can register an event callback handler with ``rte_eth_dev_callback_register()``.
+The registered callback handler will be invoked with one of below event types.
+
+#.  ``RTE_ETH_EVENT_INTR_LSC``:
+
+    It means link status of the port was changed.
+
+#.  ``RTE_ETH_EVENT_QUEUE_STATE``:
+
+    It means some of queue statuses were changed. Call ``rte_eth_vhost_get_queue_event()`` in the callback handler.
+    Because changing multiple statuses may occur only one event, call the function repeatedly as long as it doesn't return negative value.
+
+Vhost PMD with testpmd application
+----------------------------------
+
+This section demonstrates vhost PMD with testpmd DPDK sample application.
+
+#.  Launch the testpmd with vhost PMD:
+
+    .. code-block:: console
+
+        ./testpmd -c f -n 4 --vdev 'eth_vhost0,iface=/tmp/sock0,queues=1' -- -i
+
+    Other basic DPDK preparations like hugepage enabling here.
+    Please refer to the *DPDK Getting Started Guide* for detailed instructions.
+
+#.  Launch the QEMU:
+
+    .. code-block:: console
+
+       qemu-system-x86_64 <snip>
+                   -chardev socket,id=chr0,path=/tmp/sock0 \
+                   -netdev vhost-user,id=net0,chardev=chr0,vhostforce,queues=1 \
+                   -device virtio-net-pci,netdev=net0
+
+    This command attaches one virtio-net device to QEMU guest.
+    After initialization processes between QEMU and DPDK vhost library are done, status of the port will be linked up.
index 0d97041..e43edb3 100644 (file)
@@ -86,6 +86,10 @@ This section should contain new features added in this release. Sample format:
 
 * **Added vhost-user live migration support.**
 
+* **Added vhost driver.**
+
+  Added virtual PMD that wraps librte_vhost.
+
 * **Added multicast promiscuous mode support on VF for ixgbe.**
 
   Added multicast promiscuous mode support on ixgbe VF driver. So all the VFs
index 612e85e..3386a67 100644 (file)
@@ -53,4 +53,8 @@ DIRS-$(CONFIG_RTE_LIBRTE_VIRTIO_PMD) += virtio
 DIRS-$(CONFIG_RTE_LIBRTE_VMXNET3_PMD) += vmxnet3
 DIRS-$(CONFIG_RTE_LIBRTE_PMD_XENVIRT) += xenvirt
 
+ifeq ($(CONFIG_RTE_LIBRTE_VHOST),y)
+DIRS-$(CONFIG_RTE_LIBRTE_PMD_VHOST) += vhost
+endif # $(CONFIG_RTE_LIBRTE_VHOST)
+
 include $(RTE_SDK)/mk/rte.subdir.mk
diff --git a/drivers/net/vhost/Makefile b/drivers/net/vhost/Makefile
new file mode 100644 (file)
index 0000000..f49a69b
--- /dev/null
@@ -0,0 +1,62 @@
+#   BSD LICENSE
+#
+#   Copyright (c) 2010-2016 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_vhost.a
+
+CFLAGS += -O3
+CFLAGS += $(WERROR_FLAGS)
+
+EXPORT_MAP := rte_pmd_vhost_version.map
+
+LIBABIVER := 1
+
+#
+# all source are stored in SRCS-y
+#
+SRCS-$(CONFIG_RTE_LIBRTE_PMD_VHOST) += rte_eth_vhost.c
+
+#
+# Export include files
+#
+SYMLINK-y-include += rte_eth_vhost.h
+
+# this lib depends upon:
+DEPDIRS-$(CONFIG_RTE_LIBRTE_PMD_VHOST) += lib/librte_mbuf
+DEPDIRS-$(CONFIG_RTE_LIBRTE_PMD_VHOST) += lib/librte_ether
+DEPDIRS-$(CONFIG_RTE_LIBRTE_PMD_VHOST) += lib/librte_kvargs
+DEPDIRS-$(CONFIG_RTE_LIBRTE_PMD_VHOST) += lib/librte_vhost
+
+include $(RTE_SDK)/mk/rte.lib.mk
diff --git a/drivers/net/vhost/rte_eth_vhost.c b/drivers/net/vhost/rte_eth_vhost.c
new file mode 100644 (file)
index 0000000..4ee10c1
--- /dev/null
@@ -0,0 +1,916 @@
+/*-
+ *   BSD LICENSE
+ *
+ *   Copyright (c) 2016 IGEL Co., Ltd.
+ *   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 IGEL Co.,Ltd. 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 <unistd.h>
+#include <pthread.h>
+#include <stdbool.h>
+#ifdef RTE_LIBRTE_VHOST_NUMA
+#include <numaif.h>
+#endif
+
+#include <rte_mbuf.h>
+#include <rte_ethdev.h>
+#include <rte_malloc.h>
+#include <rte_memcpy.h>
+#include <rte_dev.h>
+#include <rte_kvargs.h>
+#include <rte_virtio_net.h>
+#include <rte_spinlock.h>
+
+#include "rte_eth_vhost.h"
+
+#define ETH_VHOST_IFACE_ARG            "iface"
+#define ETH_VHOST_QUEUES_ARG           "queues"
+
+static const char *drivername = "VHOST PMD";
+
+static const char *valid_arguments[] = {
+       ETH_VHOST_IFACE_ARG,
+       ETH_VHOST_QUEUES_ARG,
+       NULL
+};
+
+static struct ether_addr base_eth_addr = {
+       .addr_bytes = {
+               0x56 /* V */,
+               0x48 /* H */,
+               0x4F /* O */,
+               0x53 /* S */,
+               0x54 /* T */,
+               0x00
+       }
+};
+
+struct vhost_queue {
+       rte_atomic32_t allow_queuing;
+       rte_atomic32_t while_queuing;
+       struct virtio_net *device;
+       struct pmd_internal *internal;
+       struct rte_mempool *mb_pool;
+       uint8_t port;
+       uint16_t virtqueue_id;
+       uint64_t rx_pkts;
+       uint64_t tx_pkts;
+       uint64_t missed_pkts;
+       uint64_t rx_bytes;
+       uint64_t tx_bytes;
+};
+
+struct pmd_internal {
+       char *dev_name;
+       char *iface_name;
+
+       volatile uint16_t once;
+};
+
+struct internal_list {
+       TAILQ_ENTRY(internal_list) next;
+       struct rte_eth_dev *eth_dev;
+};
+
+TAILQ_HEAD(internal_list_head, internal_list);
+static struct internal_list_head internal_list =
+       TAILQ_HEAD_INITIALIZER(internal_list);
+
+static pthread_mutex_t internal_list_lock = PTHREAD_MUTEX_INITIALIZER;
+
+static rte_atomic16_t nb_started_ports;
+static pthread_t session_th;
+
+static struct rte_eth_link pmd_link = {
+               .link_speed = 10000,
+               .link_duplex = ETH_LINK_FULL_DUPLEX,
+               .link_status = 0
+};
+
+struct rte_vhost_vring_state {
+       rte_spinlock_t lock;
+
+       bool cur[RTE_MAX_QUEUES_PER_PORT * 2];
+       bool seen[RTE_MAX_QUEUES_PER_PORT * 2];
+       unsigned int index;
+       unsigned int max_vring;
+};
+
+static struct rte_vhost_vring_state *vring_states[RTE_MAX_ETHPORTS];
+
+static uint16_t
+eth_vhost_rx(void *q, struct rte_mbuf **bufs, uint16_t nb_bufs)
+{
+       struct vhost_queue *r = q;
+       uint16_t i, nb_rx = 0;
+
+       if (unlikely(rte_atomic32_read(&r->allow_queuing) == 0))
+               return 0;
+
+       rte_atomic32_set(&r->while_queuing, 1);
+
+       if (unlikely(rte_atomic32_read(&r->allow_queuing) == 0))
+               goto out;
+
+       /* Dequeue packets from guest TX queue */
+       nb_rx = rte_vhost_dequeue_burst(r->device,
+                       r->virtqueue_id, r->mb_pool, bufs, nb_bufs);
+
+       r->rx_pkts += nb_rx;
+
+       for (i = 0; likely(i < nb_rx); i++) {
+               bufs[i]->port = r->port;
+               r->rx_bytes += bufs[i]->pkt_len;
+       }
+
+out:
+       rte_atomic32_set(&r->while_queuing, 0);
+
+       return nb_rx;
+}
+
+static uint16_t
+eth_vhost_tx(void *q, struct rte_mbuf **bufs, uint16_t nb_bufs)
+{
+       struct vhost_queue *r = q;
+       uint16_t i, nb_tx = 0;
+
+       if (unlikely(rte_atomic32_read(&r->allow_queuing) == 0))
+               return 0;
+
+       rte_atomic32_set(&r->while_queuing, 1);
+
+       if (unlikely(rte_atomic32_read(&r->allow_queuing) == 0))
+               goto out;
+
+       /* Enqueue packets to guest RX queue */
+       nb_tx = rte_vhost_enqueue_burst(r->device,
+                       r->virtqueue_id, bufs, nb_bufs);
+
+       r->tx_pkts += nb_tx;
+       r->missed_pkts += nb_bufs - nb_tx;
+
+       for (i = 0; likely(i < nb_tx); i++)
+               r->tx_bytes += bufs[i]->pkt_len;
+
+       for (i = 0; likely(i < nb_tx); i++)
+               rte_pktmbuf_free(bufs[i]);
+out:
+       rte_atomic32_set(&r->while_queuing, 0);
+
+       return nb_tx;
+}
+
+static int
+eth_dev_configure(struct rte_eth_dev *dev __rte_unused)
+{
+       return 0;
+}
+
+static inline struct internal_list *
+find_internal_resource(char *ifname)
+{
+       int found = 0;
+       struct internal_list *list;
+       struct pmd_internal *internal;
+
+       if (ifname == NULL)
+               return NULL;
+
+       pthread_mutex_lock(&internal_list_lock);
+
+       TAILQ_FOREACH(list, &internal_list, next) {
+               internal = list->eth_dev->data->dev_private;
+               if (!strcmp(internal->iface_name, ifname)) {
+                       found = 1;
+                       break;
+               }
+       }
+
+       pthread_mutex_unlock(&internal_list_lock);
+
+       if (!found)
+               return NULL;
+
+       return list;
+}
+
+static int
+new_device(struct virtio_net *dev)
+{
+       struct rte_eth_dev *eth_dev;
+       struct internal_list *list;
+       struct pmd_internal *internal;
+       struct vhost_queue *vq;
+       unsigned i;
+
+       if (dev == NULL) {
+               RTE_LOG(INFO, PMD, "Invalid argument\n");
+               return -1;
+       }
+
+       list = find_internal_resource(dev->ifname);
+       if (list == NULL) {
+               RTE_LOG(INFO, PMD, "Invalid device name\n");
+               return -1;
+       }
+
+       eth_dev = list->eth_dev;
+       internal = eth_dev->data->dev_private;
+
+       for (i = 0; i < eth_dev->data->nb_rx_queues; i++) {
+               vq = eth_dev->data->rx_queues[i];
+               if (vq == NULL)
+                       continue;
+               vq->device = dev;
+               vq->internal = internal;
+               vq->port = eth_dev->data->port_id;
+               rte_vhost_enable_guest_notification(dev, vq->virtqueue_id, 0);
+       }
+       for (i = 0; i < eth_dev->data->nb_tx_queues; i++) {
+               vq = eth_dev->data->tx_queues[i];
+               if (vq == NULL)
+                       continue;
+               vq->device = dev;
+               vq->internal = internal;
+               vq->port = eth_dev->data->port_id;
+               rte_vhost_enable_guest_notification(dev, vq->virtqueue_id, 0);
+       }
+
+       dev->flags |= VIRTIO_DEV_RUNNING;
+       dev->priv = eth_dev;
+       eth_dev->data->dev_link.link_status = 1;
+
+       for (i = 0; i < eth_dev->data->nb_rx_queues; i++) {
+               vq = eth_dev->data->rx_queues[i];
+               if (vq == NULL)
+                       continue;
+               rte_atomic32_set(&vq->allow_queuing, 1);
+       }
+       for (i = 0; i < eth_dev->data->nb_tx_queues; i++) {
+               vq = eth_dev->data->tx_queues[i];
+               if (vq == NULL)
+                       continue;
+               rte_atomic32_set(&vq->allow_queuing, 1);
+       }
+
+       RTE_LOG(INFO, PMD, "New connection established\n");
+
+       _rte_eth_dev_callback_process(eth_dev, RTE_ETH_EVENT_INTR_LSC);
+
+       return 0;
+}
+
+static void
+destroy_device(volatile struct virtio_net *dev)
+{
+       struct rte_eth_dev *eth_dev;
+       struct vhost_queue *vq;
+       unsigned i;
+
+       if (dev == NULL) {
+               RTE_LOG(INFO, PMD, "Invalid argument\n");
+               return;
+       }
+
+       eth_dev = (struct rte_eth_dev *)dev->priv;
+       if (eth_dev == NULL) {
+               RTE_LOG(INFO, PMD, "Failed to find a ethdev\n");
+               return;
+       }
+
+       /* Wait until rx/tx_pkt_burst stops accessing vhost device */
+       for (i = 0; i < eth_dev->data->nb_rx_queues; i++) {
+               vq = eth_dev->data->rx_queues[i];
+               if (vq == NULL)
+                       continue;
+               rte_atomic32_set(&vq->allow_queuing, 0);
+               while (rte_atomic32_read(&vq->while_queuing))
+                       rte_pause();
+       }
+       for (i = 0; i < eth_dev->data->nb_tx_queues; i++) {
+               vq = eth_dev->data->tx_queues[i];
+               if (vq == NULL)
+                       continue;
+               rte_atomic32_set(&vq->allow_queuing, 0);
+               while (rte_atomic32_read(&vq->while_queuing))
+                       rte_pause();
+       }
+
+       eth_dev->data->dev_link.link_status = 0;
+
+       dev->priv = NULL;
+       dev->flags &= ~VIRTIO_DEV_RUNNING;
+
+       for (i = 0; i < eth_dev->data->nb_rx_queues; i++) {
+               vq = eth_dev->data->rx_queues[i];
+               if (vq == NULL)
+                       continue;
+               vq->device = NULL;
+       }
+       for (i = 0; i < eth_dev->data->nb_tx_queues; i++) {
+               vq = eth_dev->data->tx_queues[i];
+               if (vq == NULL)
+                       continue;
+               vq->device = NULL;
+       }
+
+       RTE_LOG(INFO, PMD, "Connection closed\n");
+
+       _rte_eth_dev_callback_process(eth_dev, RTE_ETH_EVENT_INTR_LSC);
+}
+
+static int
+vring_state_changed(struct virtio_net *dev, uint16_t vring, int enable)
+{
+       struct rte_vhost_vring_state *state;
+       struct rte_eth_dev *eth_dev;
+       struct internal_list *list;
+#ifdef RTE_LIBRTE_VHOST_NUMA
+       int newnode, ret;
+#endif
+
+       if (dev == NULL) {
+               RTE_LOG(ERR, PMD, "Invalid argument\n");
+               return -1;
+       }
+
+       list = find_internal_resource(dev->ifname);
+       if (list == NULL) {
+               RTE_LOG(ERR, PMD, "Invalid interface name: %s\n", dev->ifname);
+               return -1;
+       }
+
+       eth_dev = list->eth_dev;
+       /* won't be NULL */
+       state = vring_states[eth_dev->data->port_id];
+
+#ifdef RTE_LIBRTE_VHOST_NUMA
+       ret  = get_mempolicy(&newnode, NULL, 0, dev,
+                       MPOL_F_NODE | MPOL_F_ADDR);
+       if (ret < 0) {
+               RTE_LOG(ERR, PMD, "Unknown numa node\n");
+               return -1;
+       }
+
+       eth_dev->data->numa_node = newnode;
+#endif
+       rte_spinlock_lock(&state->lock);
+       state->cur[vring] = enable;
+       state->max_vring = RTE_MAX(vring, state->max_vring);
+       rte_spinlock_unlock(&state->lock);
+
+       RTE_LOG(INFO, PMD, "vring%u is %s\n",
+                       vring, enable ? "enabled" : "disabled");
+
+       _rte_eth_dev_callback_process(eth_dev, RTE_ETH_EVENT_QUEUE_STATE);
+
+       return 0;
+}
+
+int
+rte_eth_vhost_get_queue_event(uint8_t port_id,
+               struct rte_eth_vhost_queue_event *event)
+{
+       struct rte_vhost_vring_state *state;
+       unsigned int i;
+       int idx;
+
+       if (port_id >= RTE_MAX_ETHPORTS) {
+               RTE_LOG(ERR, PMD, "Invalid port id\n");
+               return -1;
+       }
+
+       state = vring_states[port_id];
+       if (!state) {
+               RTE_LOG(ERR, PMD, "Unused port\n");
+               return -1;
+       }
+
+       rte_spinlock_lock(&state->lock);
+       for (i = 0; i <= state->max_vring; i++) {
+               idx = state->index++ % (state->max_vring + 1);
+
+               if (state->cur[idx] != state->seen[idx]) {
+                       state->seen[idx] = state->cur[idx];
+                       event->queue_id = idx / 2;
+                       event->rx = idx & 1;
+                       event->enable = state->cur[idx];
+                       rte_spinlock_unlock(&state->lock);
+                       return 0;
+               }
+       }
+       rte_spinlock_unlock(&state->lock);
+
+       return -1;
+}
+
+static void *
+vhost_driver_session(void *param __rte_unused)
+{
+       static struct virtio_net_device_ops vhost_ops;
+
+       /* set vhost arguments */
+       vhost_ops.new_device = new_device;
+       vhost_ops.destroy_device = destroy_device;
+       vhost_ops.vring_state_changed = vring_state_changed;
+       if (rte_vhost_driver_callback_register(&vhost_ops) < 0)
+               RTE_LOG(ERR, PMD, "Can't register callbacks\n");
+
+       /* start event handling */
+       rte_vhost_driver_session_start();
+
+       return NULL;
+}
+
+static int
+vhost_driver_session_start(void)
+{
+       int ret;
+
+       ret = pthread_create(&session_th,
+                       NULL, vhost_driver_session, NULL);
+       if (ret)
+               RTE_LOG(ERR, PMD, "Can't create a thread\n");
+
+       return ret;
+}
+
+static void
+vhost_driver_session_stop(void)
+{
+       int ret;
+
+       ret = pthread_cancel(session_th);
+       if (ret)
+               RTE_LOG(ERR, PMD, "Can't cancel the thread\n");
+
+       ret = pthread_join(session_th, NULL);
+       if (ret)
+               RTE_LOG(ERR, PMD, "Can't join the thread\n");
+}
+
+static int
+eth_dev_start(struct rte_eth_dev *dev)
+{
+       struct pmd_internal *internal = dev->data->dev_private;
+       int ret = 0;
+
+       if (rte_atomic16_cmpset(&internal->once, 0, 1)) {
+               ret = rte_vhost_driver_register(internal->iface_name);
+               if (ret)
+                       return ret;
+       }
+
+       /* We need only one message handling thread */
+       if (rte_atomic16_add_return(&nb_started_ports, 1) == 1)
+               ret = vhost_driver_session_start();
+
+       return ret;
+}
+
+static void
+eth_dev_stop(struct rte_eth_dev *dev)
+{
+       struct pmd_internal *internal = dev->data->dev_private;
+
+       if (rte_atomic16_cmpset(&internal->once, 1, 0))
+               rte_vhost_driver_unregister(internal->iface_name);
+
+       if (rte_atomic16_sub_return(&nb_started_ports, 1) == 0)
+               vhost_driver_session_stop();
+}
+
+static int
+eth_rx_queue_setup(struct rte_eth_dev *dev, uint16_t rx_queue_id,
+                  uint16_t nb_rx_desc __rte_unused,
+                  unsigned int socket_id,
+                  const struct rte_eth_rxconf *rx_conf __rte_unused,
+                  struct rte_mempool *mb_pool)
+{
+       struct vhost_queue *vq;
+
+       vq = rte_zmalloc_socket(NULL, sizeof(struct vhost_queue),
+                       RTE_CACHE_LINE_SIZE, socket_id);
+       if (vq == NULL) {
+               RTE_LOG(ERR, PMD, "Failed to allocate memory for rx queue\n");
+               return -ENOMEM;
+       }
+
+       vq->mb_pool = mb_pool;
+       vq->virtqueue_id = rx_queue_id * VIRTIO_QNUM + VIRTIO_TXQ;
+       dev->data->rx_queues[rx_queue_id] = vq;
+
+       return 0;
+}
+
+static int
+eth_tx_queue_setup(struct rte_eth_dev *dev, uint16_t tx_queue_id,
+                  uint16_t nb_tx_desc __rte_unused,
+                  unsigned int socket_id,
+                  const struct rte_eth_txconf *tx_conf __rte_unused)
+{
+       struct vhost_queue *vq;
+
+       vq = rte_zmalloc_socket(NULL, sizeof(struct vhost_queue),
+                       RTE_CACHE_LINE_SIZE, socket_id);
+       if (vq == NULL) {
+               RTE_LOG(ERR, PMD, "Failed to allocate memory for tx queue\n");
+               return -ENOMEM;
+       }
+
+       vq->virtqueue_id = tx_queue_id * VIRTIO_QNUM + VIRTIO_RXQ;
+       dev->data->tx_queues[tx_queue_id] = vq;
+
+       return 0;
+}
+
+static void
+eth_dev_info(struct rte_eth_dev *dev,
+            struct rte_eth_dev_info *dev_info)
+{
+       dev_info->driver_name = drivername;
+       dev_info->max_mac_addrs = 1;
+       dev_info->max_rx_pktlen = (uint32_t)-1;
+       dev_info->max_rx_queues = dev->data->nb_rx_queues;
+       dev_info->max_tx_queues = dev->data->nb_tx_queues;
+       dev_info->min_rx_bufsize = 0;
+}
+
+static void
+eth_stats_get(struct rte_eth_dev *dev, struct rte_eth_stats *stats)
+{
+       unsigned i;
+       unsigned long rx_total = 0, tx_total = 0, tx_missed_total = 0;
+       unsigned long rx_total_bytes = 0, tx_total_bytes = 0;
+       struct vhost_queue *vq;
+
+       for (i = 0; i < RTE_ETHDEV_QUEUE_STAT_CNTRS &&
+                       i < dev->data->nb_rx_queues; i++) {
+               if (dev->data->rx_queues[i] == NULL)
+                       continue;
+               vq = dev->data->rx_queues[i];
+               stats->q_ipackets[i] = vq->rx_pkts;
+               rx_total += stats->q_ipackets[i];
+
+               stats->q_ibytes[i] = vq->rx_bytes;
+               rx_total_bytes += stats->q_ibytes[i];
+       }
+
+       for (i = 0; i < RTE_ETHDEV_QUEUE_STAT_CNTRS &&
+                       i < dev->data->nb_tx_queues; i++) {
+               if (dev->data->tx_queues[i] == NULL)
+                       continue;
+               vq = dev->data->tx_queues[i];
+               stats->q_opackets[i] = vq->tx_pkts;
+               tx_missed_total += vq->missed_pkts;
+               tx_total += stats->q_opackets[i];
+
+               stats->q_obytes[i] = vq->tx_bytes;
+               tx_total_bytes += stats->q_obytes[i];
+       }
+
+       stats->ipackets = rx_total;
+       stats->opackets = tx_total;
+       stats->imissed = tx_missed_total;
+       stats->ibytes = rx_total_bytes;
+       stats->obytes = tx_total_bytes;
+}
+
+static void
+eth_stats_reset(struct rte_eth_dev *dev)
+{
+       struct vhost_queue *vq;
+       unsigned i;
+
+       for (i = 0; i < dev->data->nb_rx_queues; i++) {
+               if (dev->data->rx_queues[i] == NULL)
+                       continue;
+               vq = dev->data->rx_queues[i];
+               vq->rx_pkts = 0;
+               vq->rx_bytes = 0;
+       }
+       for (i = 0; i < dev->data->nb_tx_queues; i++) {
+               if (dev->data->tx_queues[i] == NULL)
+                       continue;
+               vq = dev->data->tx_queues[i];
+               vq->tx_pkts = 0;
+               vq->tx_bytes = 0;
+               vq->missed_pkts = 0;
+       }
+}
+
+static void
+eth_queue_release(void *q)
+{
+       rte_free(q);
+}
+
+static int
+eth_link_update(struct rte_eth_dev *dev __rte_unused,
+               int wait_to_complete __rte_unused)
+{
+       return 0;
+}
+
+/**
+ * Disable features in feature_mask. Returns 0 on success.
+ */
+int
+rte_eth_vhost_feature_disable(uint64_t feature_mask)
+{
+       return rte_vhost_feature_disable(feature_mask);
+}
+
+/**
+ * Enable features in feature_mask. Returns 0 on success.
+ */
+int
+rte_eth_vhost_feature_enable(uint64_t feature_mask)
+{
+       return rte_vhost_feature_enable(feature_mask);
+}
+
+/* Returns currently supported vhost features */
+uint64_t
+rte_eth_vhost_feature_get(void)
+{
+       return rte_vhost_feature_get();
+}
+
+static const struct eth_dev_ops ops = {
+       .dev_start = eth_dev_start,
+       .dev_stop = eth_dev_stop,
+       .dev_configure = eth_dev_configure,
+       .dev_infos_get = eth_dev_info,
+       .rx_queue_setup = eth_rx_queue_setup,
+       .tx_queue_setup = eth_tx_queue_setup,
+       .rx_queue_release = eth_queue_release,
+       .tx_queue_release = eth_queue_release,
+       .link_update = eth_link_update,
+       .stats_get = eth_stats_get,
+       .stats_reset = eth_stats_reset,
+};
+
+static int
+eth_dev_vhost_create(const char *name, char *iface_name, int16_t queues,
+                    const unsigned numa_node)
+{
+       struct rte_eth_dev_data *data = NULL;
+       struct pmd_internal *internal = NULL;
+       struct rte_eth_dev *eth_dev = NULL;
+       struct ether_addr *eth_addr = NULL;
+       struct rte_vhost_vring_state *vring_state = NULL;
+       struct internal_list *list = NULL;
+
+       RTE_LOG(INFO, PMD, "Creating VHOST-USER backend on numa socket %u\n",
+               numa_node);
+
+       /* now do all data allocation - for eth_dev structure, dummy pci driver
+        * and internal (private) data
+        */
+       data = rte_zmalloc_socket(name, sizeof(*data), 0, numa_node);
+       if (data == NULL)
+               goto error;
+
+       internal = rte_zmalloc_socket(name, sizeof(*internal), 0, numa_node);
+       if (internal == NULL)
+               goto error;
+
+       list = rte_zmalloc_socket(name, sizeof(*list), 0, numa_node);
+       if (list == NULL)
+               goto error;
+
+       /* reserve an ethdev entry */
+       eth_dev = rte_eth_dev_allocate(name, RTE_ETH_DEV_VIRTUAL);
+       if (eth_dev == NULL)
+               goto error;
+
+       eth_addr = rte_zmalloc_socket(name, sizeof(*eth_addr), 0, numa_node);
+       if (eth_addr == NULL)
+               goto error;
+       *eth_addr = base_eth_addr;
+       eth_addr->addr_bytes[5] = eth_dev->data->port_id;
+
+       vring_state = rte_zmalloc_socket(name,
+                       sizeof(*vring_state), 0, numa_node);
+       if (vring_state == NULL)
+               goto error;
+
+       TAILQ_INIT(&eth_dev->link_intr_cbs);
+
+       /* now put it all together
+        * - store queue data in internal,
+        * - store numa_node info in ethdev data
+        * - point eth_dev_data to internals
+        * - and point eth_dev structure to new eth_dev_data structure
+        */
+       internal->dev_name = strdup(name);
+       if (internal->dev_name == NULL)
+               goto error;
+       internal->iface_name = strdup(iface_name);
+       if (internal->iface_name == NULL)
+               goto error;
+
+       list->eth_dev = eth_dev;
+       pthread_mutex_lock(&internal_list_lock);
+       TAILQ_INSERT_TAIL(&internal_list, list, next);
+       pthread_mutex_unlock(&internal_list_lock);
+
+       rte_spinlock_init(&vring_state->lock);
+       vring_states[eth_dev->data->port_id] = vring_state;
+
+       data->dev_private = internal;
+       data->port_id = eth_dev->data->port_id;
+       memmove(data->name, eth_dev->data->name, sizeof(data->name));
+       data->nb_rx_queues = queues;
+       data->nb_tx_queues = queues;
+       data->dev_link = pmd_link;
+       data->mac_addrs = eth_addr;
+
+       /* We'll replace the 'data' originally allocated by eth_dev. So the
+        * vhost PMD resources won't be shared between multi processes.
+        */
+       eth_dev->data = data;
+       eth_dev->dev_ops = &ops;
+       eth_dev->driver = NULL;
+       data->dev_flags =
+               RTE_ETH_DEV_DETACHABLE | RTE_ETH_DEV_INTR_LSC;
+       data->kdrv = RTE_KDRV_NONE;
+       data->drv_name = internal->dev_name;
+       data->numa_node = numa_node;
+
+       /* finally assign rx and tx ops */
+       eth_dev->rx_pkt_burst = eth_vhost_rx;
+       eth_dev->tx_pkt_burst = eth_vhost_tx;
+
+       return data->port_id;
+
+error:
+       if (internal)
+               free(internal->dev_name);
+       rte_free(vring_state);
+       rte_free(eth_addr);
+       if (eth_dev)
+               rte_eth_dev_release_port(eth_dev);
+       rte_free(internal);
+       rte_free(list);
+       rte_free(data);
+
+       return -1;
+}
+
+static inline int
+open_iface(const char *key __rte_unused, const char *value, void *extra_args)
+{
+       const char **iface_name = extra_args;
+
+       if (value == NULL)
+               return -1;
+
+       *iface_name = value;
+
+       return 0;
+}
+
+static inline int
+open_queues(const char *key __rte_unused, const char *value, void *extra_args)
+{
+       uint16_t *q = extra_args;
+
+       if (value == NULL || extra_args == NULL)
+               return -EINVAL;
+
+       *q = (uint16_t)strtoul(value, NULL, 0);
+       if (*q == USHRT_MAX && errno == ERANGE)
+               return -1;
+
+       if (*q > RTE_MAX_QUEUES_PER_PORT)
+               return -1;
+
+       return 0;
+}
+
+static int
+rte_pmd_vhost_devinit(const char *name, const char *params)
+{
+       struct rte_kvargs *kvlist = NULL;
+       int ret = 0;
+       char *iface_name;
+       uint16_t queues;
+
+       RTE_LOG(INFO, PMD, "Initializing pmd_vhost for %s\n", name);
+
+       kvlist = rte_kvargs_parse(params, valid_arguments);
+       if (kvlist == NULL)
+               return -1;
+
+       if (rte_kvargs_count(kvlist, ETH_VHOST_IFACE_ARG) == 1) {
+               ret = rte_kvargs_process(kvlist, ETH_VHOST_IFACE_ARG,
+                                        &open_iface, &iface_name);
+               if (ret < 0)
+                       goto out_free;
+       } else {
+               ret = -1;
+               goto out_free;
+       }
+
+       if (rte_kvargs_count(kvlist, ETH_VHOST_QUEUES_ARG) == 1) {
+               ret = rte_kvargs_process(kvlist, ETH_VHOST_QUEUES_ARG,
+                                        &open_queues, &queues);
+               if (ret < 0)
+                       goto out_free;
+
+       } else
+               queues = 1;
+
+       eth_dev_vhost_create(name, iface_name, queues, rte_socket_id());
+
+out_free:
+       rte_kvargs_free(kvlist);
+       return ret;
+}
+
+static int
+rte_pmd_vhost_devuninit(const char *name)
+{
+       struct rte_eth_dev *eth_dev = NULL;
+       struct pmd_internal *internal;
+       struct internal_list *list;
+       unsigned int i;
+
+       RTE_LOG(INFO, PMD, "Un-Initializing pmd_vhost for %s\n", name);
+
+       /* find an ethdev entry */
+       eth_dev = rte_eth_dev_allocated(name);
+       if (eth_dev == NULL)
+               return -ENODEV;
+
+       internal = eth_dev->data->dev_private;
+       if (internal == NULL)
+               return -ENODEV;
+
+       list = find_internal_resource(internal->iface_name);
+       if (list == NULL)
+               return -ENODEV;
+
+       pthread_mutex_lock(&internal_list_lock);
+       TAILQ_REMOVE(&internal_list, list, next);
+       pthread_mutex_unlock(&internal_list_lock);
+       rte_free(list);
+
+       eth_dev_stop(eth_dev);
+
+       rte_free(vring_states[eth_dev->data->port_id]);
+       vring_states[eth_dev->data->port_id] = NULL;
+
+       free(internal->dev_name);
+       free(internal->iface_name);
+
+       for (i = 0; i < eth_dev->data->nb_rx_queues; i++)
+               rte_free(eth_dev->data->rx_queues[i]);
+       for (i = 0; i < eth_dev->data->nb_tx_queues; i++)
+               rte_free(eth_dev->data->tx_queues[i]);
+
+       rte_free(eth_dev->data->mac_addrs);
+       rte_free(eth_dev->data);
+       rte_free(internal);
+
+       rte_eth_dev_release_port(eth_dev);
+
+       return 0;
+}
+
+static struct rte_driver pmd_vhost_drv = {
+       .name = "eth_vhost",
+       .type = PMD_VDEV,
+       .init = rte_pmd_vhost_devinit,
+       .uninit = rte_pmd_vhost_devuninit,
+};
+
+PMD_REGISTER_DRIVER(pmd_vhost_drv);
diff --git a/drivers/net/vhost/rte_eth_vhost.h b/drivers/net/vhost/rte_eth_vhost.h
new file mode 100644 (file)
index 0000000..ff5d877
--- /dev/null
@@ -0,0 +1,109 @@
+/*-
+ *   BSD LICENSE
+ *
+ *   Copyright(c) 2016 IGEL Co., Ltd.
+ *   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 IGEL Co., Ltd. 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.
+ */
+
+#ifndef _RTE_ETH_VHOST_H_
+#define _RTE_ETH_VHOST_H_
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+#include <stdint.h>
+#include <stdbool.h>
+
+#include <rte_virtio_net.h>
+
+/**
+ * Disable features in feature_mask.
+ *
+ * @param feature_mask
+ *  Vhost features defined in "linux/virtio_net.h".
+ * @return
+ *  - On success, zero.
+ *  - On failure, a negative value.
+ */
+int rte_eth_vhost_feature_disable(uint64_t feature_mask);
+
+/**
+ * Enable features in feature_mask.
+ *
+ * @param feature_mask
+ *  Vhost features defined in "linux/virtio_net.h".
+ * @return
+ *  - On success, zero.
+ *  - On failure, a negative value.
+ */
+int rte_eth_vhost_feature_enable(uint64_t feature_mask);
+
+/**
+ * Returns currently supported vhost features.
+ *
+ * @return
+ *  Vhost features defined in "linux/virtio_net.h".
+ */
+uint64_t rte_eth_vhost_feature_get(void);
+
+/*
+ * Event description.
+ */
+struct rte_eth_vhost_queue_event {
+       uint16_t queue_id;
+       bool rx;
+       bool enable;
+};
+
+/**
+ * Get queue events from specified port.
+ * If a callback for below event is registered by
+ * rte_eth_dev_callback_register(), this function will describe what was
+ * changed.
+ *  - RTE_ETH_EVENT_QUEUE_STATE
+ * Multiple events may cause only one callback kicking, so call this function
+ * while returning 0.
+ *
+ * @param port_id
+ *  Port id.
+ * @param event
+ *  Pointer to a rte_eth_vhost_queue_event structure.
+ * @return
+ *  - On success, zero.
+ *  - On failure, a negative value.
+ */
+int rte_eth_vhost_get_queue_event(uint8_t port_id,
+               struct rte_eth_vhost_queue_event *event);
+
+#ifdef __cplusplus
+}
+#endif
+
+#endif
diff --git a/drivers/net/vhost/rte_pmd_vhost_version.map b/drivers/net/vhost/rte_pmd_vhost_version.map
new file mode 100644 (file)
index 0000000..65bf3a8
--- /dev/null
@@ -0,0 +1,10 @@
+DPDK_16.04 {
+       global:
+
+       rte_eth_vhost_feature_disable;
+       rte_eth_vhost_feature_enable;
+       rte_eth_vhost_feature_get;
+       rte_eth_vhost_get_queue_event;
+
+       local: *;
+};
index c1ed60f..179aac4 100644 (file)
@@ -169,6 +169,12 @@ _LDLIBS-$(CONFIG_RTE_LIBRTE_PMD_SNOW3G)     += -lrte_pmd_snow3g
 _LDLIBS-$(CONFIG_RTE_LIBRTE_PMD_SNOW3G)     += -L$(LIBSSO_PATH)/build -lsso
 endif # CONFIG_RTE_LIBRTE_CRYPTODEV
 
+ifeq ($(CONFIG_RTE_LIBRTE_VHOST),y)
+
+_LDLIBS-$(CONFIG_RTE_LIBRTE_PMD_VHOST)      += -lrte_pmd_vhost
+
+endif # $(CONFIG_RTE_LIBRTE_VHOST)
+
 endif # ! $(CONFIG_RTE_BUILD_SHARED_LIB)
 
 _LDLIBS-y += $(EXECENV_LDLIBS)