Imported Upstream version 17.05

Change-Id: Id1e419c5a214e4a18739663b91f0f9a549f1fdc6
Signed-off-by: Christian Ehrhardt <christian.ehrhardt@canonical.com>

4 files changed, 519 insertions, 36 deletions

diff --git a/examples/vhost/Makefile b/examples/vhost/Makefile
index e95c68ae..af7be99a 100644
--- a/examples/vhost/Makefile
+++ b/examples/vhost/Makefile
@@ -48,7 +48,7 @@ else
 APP = vhost-switch
 
 # all source are stored in SRCS-y
-SRCS-y := main.c
+SRCS-y := main.c virtio_net.c
 
 CFLAGS += -O2 -D_FILE_OFFSET_BITS=64
 CFLAGS += $(WERROR_FLAGS)
diff --git a/examples/vhost/main.c b/examples/vhost/main.c
index eddaf926..e07f8669 100644
--- a/examples/vhost/main.c
+++ b/examples/vhost/main.c
@@ -1,7 +1,7 @@
 /*-
  *   BSD LICENSE
  *
- *   Copyright(c) 2010-2016 Intel Corporation. All rights reserved.
+ *   Copyright(c) 2010-2017 Intel Corporation. All rights reserved.
  *   All rights reserved.
  *
  *   Redistribution and use in source and binary forms, with or without
@@ -49,7 +49,7 @@
 #include <rte_log.h>
 #include <rte_string_fns.h>
 #include <rte_malloc.h>
-#include <rte_virtio_net.h>
+#include <rte_vhost.h>
 #include <rte_ip.h>
 #include <rte_tcp.h>
 
@@ -65,7 +65,6 @@
 #define MBUF_CACHE_SIZE	128
 #define MBUF_DATA_SIZE	RTE_MBUF_DEFAULT_BUF_SIZE
 
-#define MAX_PKT_BURST 32		/* Max burst size for RX/TX */
 #define BURST_TX_DRAIN_US 100	/* TX drain every ~100us */
 
 #define BURST_RX_WAIT_US 15	/* Defines how long we wait between retries on RX */
@@ -129,6 +128,8 @@ static uint32_t enable_tso;
 static int client_mode;
 static int dequeue_zero_copy;
 
+static int builtin_net_driver;
+
 /* Specify timeout (in useconds) between retries on RX. */
 static uint32_t burst_rx_delay_time = BURST_RX_WAIT_US;
 /* Specify the number of retries on RX. */
@@ -153,7 +154,7 @@ static struct rte_eth_conf vmdq_conf_default = {
 		 */
 		.hw_vlan_strip  = 1, /**< VLAN strip enabled. */
 		.jumbo_frame    = 0, /**< Jumbo Frame Support disabled */
-		.hw_strip_crc   = 0, /**< CRC stripped by hardware */
+		.hw_strip_crc   = 1, /**< CRC stripped by hardware */
 	},
 
 	.txmode = {
@@ -328,16 +329,6 @@ port_init(uint8_t port)
 
 	if (port >= rte_eth_dev_count()) return -1;
 
-	if (enable_tx_csum == 0)
-		rte_vhost_feature_disable(1ULL << VIRTIO_NET_F_CSUM);
-
-	if (enable_tso == 0) {
-		rte_vhost_feature_disable(1ULL << VIRTIO_NET_F_HOST_TSO4);
-		rte_vhost_feature_disable(1ULL << VIRTIO_NET_F_HOST_TSO6);
-		rte_vhost_feature_disable(1ULL << VIRTIO_NET_F_GUEST_TSO4);
-		rte_vhost_feature_disable(1ULL << VIRTIO_NET_F_GUEST_TSO6);
-	}
-
 	rx_rings = (uint16_t)dev_info.max_rx_queues;
 	/* Configure ethernet device. */
 	retval = rte_eth_dev_configure(port, rx_rings, tx_rings, &port_conf);
@@ -405,7 +396,7 @@ static int
 us_vhost_parse_socket_path(const char *q_arg)
 {
 	/* parse number string */
-	if (strnlen(q_arg, PATH_MAX) > PATH_MAX)
+	if (strnlen(q_arg, PATH_MAX) == PATH_MAX)
 		return -1;
 
 	socket_files = realloc(socket_files, PATH_MAX * (nb_sockets + 1));
@@ -509,6 +500,7 @@ us_vhost_parse_args(int argc, char **argv)
 		{"tso", required_argument, NULL, 0},
 		{"client", no_argument, &client_mode, 1},
 		{"dequeue-zero-copy", no_argument, &dequeue_zero_copy, 1},
+		{"builtin-net-driver", no_argument, &builtin_net_driver, 1},
 		{NULL, 0, 0, 0},
 	};
 
@@ -531,7 +523,6 @@ us_vhost_parse_args(int argc, char **argv)
 			vmdq_conf_default.rx_adv_conf.vmdq_rx_conf.rx_mode =
 				ETH_VMDQ_ACCEPT_BROADCAST |
 				ETH_VMDQ_ACCEPT_MULTICAST;
-			rte_vhost_feature_enable(1ULL << VIRTIO_NET_F_CTRL_RX);
 
 			break;
 
@@ -806,7 +797,12 @@ virtio_xmit(struct vhost_dev *dst_vdev, struct vhost_dev *src_vdev,
 {
 	uint16_t ret;
 
-	ret = rte_vhost_enqueue_burst(dst_vdev->vid, VIRTIO_RXQ, &m, 1);
+	if (builtin_net_driver) {
+		ret = vs_enqueue_pkts(dst_vdev, VIRTIO_RXQ, &m, 1);
+	} else {
+		ret = rte_vhost_enqueue_burst(dst_vdev->vid, VIRTIO_RXQ, &m, 1);
+	}
+
 	if (enable_stats) {
 		rte_atomic64_inc(&dst_vdev->stats.rx_total_atomic);
 		rte_atomic64_add(&dst_vdev->stats.rx_atomic, ret);
@@ -832,17 +828,17 @@ virtio_tx_local(struct vhost_dev *vdev, struct rte_mbuf *m)
 		return -1;
 
 	if (vdev->vid == dst_vdev->vid) {
-		RTE_LOG(DEBUG, VHOST_DATA,
+		RTE_LOG_DP(DEBUG, VHOST_DATA,
 			"(%d) TX: src and dst MAC is same. Dropping packet.\n",
 			vdev->vid);
 		return 0;
 	}
 
-	RTE_LOG(DEBUG, VHOST_DATA,
+	RTE_LOG_DP(DEBUG, VHOST_DATA,
 		"(%d) TX: MAC address is local\n", dst_vdev->vid);
 
 	if (unlikely(dst_vdev->remove)) {
-		RTE_LOG(DEBUG, VHOST_DATA,
+		RTE_LOG_DP(DEBUG, VHOST_DATA,
 			"(%d) device is marked for removal\n", dst_vdev->vid);
 		return 0;
 	}
@@ -867,7 +863,7 @@ find_local_dest(struct vhost_dev *vdev, struct rte_mbuf *m,
 		return 0;
 
 	if (vdev->vid == dst_vdev->vid) {
-		RTE_LOG(DEBUG, VHOST_DATA,
+		RTE_LOG_DP(DEBUG, VHOST_DATA,
 			"(%d) TX: src and dst MAC is same. Dropping packet.\n",
 			vdev->vid);
 		return -1;
@@ -881,7 +877,7 @@ find_local_dest(struct vhost_dev *vdev, struct rte_mbuf *m,
 	*offset  = VLAN_HLEN;
 	*vlan_tag = vlan_tags[vdev->vid];
 
-	RTE_LOG(DEBUG, VHOST_DATA,
+	RTE_LOG_DP(DEBUG, VHOST_DATA,
 		"(%d) TX: pkt to local VM device id: (%d), vlan tag: %u.\n",
 		vdev->vid, dst_vdev->vid, *vlan_tag);
 
@@ -973,7 +969,7 @@ virtio_tx_route(struct vhost_dev *vdev, struct rte_mbuf *m, uint16_t vlan_tag)
 		}
 	}
 
-	RTE_LOG(DEBUG, VHOST_DATA,
+	RTE_LOG_DP(DEBUG, VHOST_DATA,
 		"(%d) TX: MAC address is external\n", vdev->vid);
 
 queue2nic:
@@ -1041,7 +1037,7 @@ drain_mbuf_table(struct mbuf_table *tx_q)
 	if (unlikely(cur_tsc - prev_tsc > MBUF_TABLE_DRAIN_TSC)) {
 		prev_tsc = cur_tsc;
 
-		RTE_LOG(DEBUG, VHOST_DATA,
+		RTE_LOG_DP(DEBUG, VHOST_DATA,
 			"TX queue drained after timeout with burst size %u\n",
 			tx_q->len);
 		do_drain_mbuf_table(tx_q);
@@ -1077,8 +1073,13 @@ drain_eth_rx(struct vhost_dev *vdev)
 		}
 	}
 
-	enqueue_count = rte_vhost_enqueue_burst(vdev->vid, VIRTIO_RXQ,
+	if (builtin_net_driver) {
+		enqueue_count = vs_enqueue_pkts(vdev, VIRTIO_RXQ,
+						pkts, rx_count);
+	} else {
+		enqueue_count = rte_vhost_enqueue_burst(vdev->vid, VIRTIO_RXQ,
 						pkts, rx_count);
+	}
 	if (enable_stats) {
 		rte_atomic64_add(&vdev->stats.rx_total_atomic, rx_count);
 		rte_atomic64_add(&vdev->stats.rx_atomic, enqueue_count);
@@ -1094,8 +1095,13 @@ drain_virtio_tx(struct vhost_dev *vdev)
 	uint16_t count;
 	uint16_t i;
 
-	count = rte_vhost_dequeue_burst(vdev->vid, VIRTIO_TXQ, mbuf_pool,
+	if (builtin_net_driver) {
+		count = vs_dequeue_pkts(vdev, VIRTIO_TXQ, mbuf_pool,
 					pkts, MAX_PKT_BURST);
+	} else {
+		count = rte_vhost_dequeue_burst(vdev->vid, VIRTIO_TXQ,
+					mbuf_pool, pkts, MAX_PKT_BURST);
+	}
 
 	/* setup VMDq for the first packet */
 	if (unlikely(vdev->ready == DEVICE_MAC_LEARNING) && count) {
@@ -1198,6 +1204,9 @@ destroy_device(int vid)
 		rte_pause();
 	}
 
+	if (builtin_net_driver)
+		vs_vhost_net_remove(vdev);
+
 	TAILQ_REMOVE(&lcore_info[vdev->coreid].vdev_list, vdev,
 		     lcore_vdev_entry);
 	TAILQ_REMOVE(&vhost_dev_list, vdev, global_vdev_entry);
@@ -1246,6 +1255,9 @@ new_device(int vid)
 	}
 	vdev->vid = vid;
 
+	if (builtin_net_driver)
+		vs_vhost_net_setup(vdev);
+
 	TAILQ_INSERT_TAIL(&vhost_dev_list, vdev, global_vdev_entry);
 	vdev->vmdq_rx_q = vid * queues_per_pool + vmdq_queue_base;
 
@@ -1281,7 +1293,7 @@ new_device(int vid)
  * These callback allow devices to be added to the data core when configuration
  * has been fully complete.
  */
-static const struct virtio_net_device_ops virtio_net_device_ops =
+static const struct vhost_device_ops virtio_net_device_ops =
 {
 	.new_device =  new_device,
 	.destroy_device = destroy_device,
@@ -1509,9 +1521,6 @@ main(int argc, char *argv[])
 	RTE_LCORE_FOREACH_SLAVE(lcore_id)
 		rte_eal_remote_launch(switch_worker, NULL, lcore_id);
 
-	if (mergeable == 0)
-		rte_vhost_feature_disable(1ULL << VIRTIO_NET_F_MRG_RXBUF);
-
 	if (client_mode)
 		flags |= RTE_VHOST_USER_CLIENT;
 
@@ -1520,18 +1529,59 @@ main(int argc, char *argv[])
 
 	/* Register vhost user driver to handle vhost messages. */
 	for (i = 0; i < nb_sockets; i++) {
-		ret = rte_vhost_driver_register
-				(socket_files + i * PATH_MAX, flags);
+		char *file = socket_files + i * PATH_MAX;
+		ret = rte_vhost_driver_register(file, flags);
 		if (ret != 0) {
 			unregister_drivers(i);
 			rte_exit(EXIT_FAILURE,
 				"vhost driver register failure.\n");
 		}
+
+		if (builtin_net_driver)
+			rte_vhost_driver_set_features(file, VIRTIO_NET_FEATURES);
+
+		if (mergeable == 0) {
+			rte_vhost_driver_disable_features(file,
+				1ULL << VIRTIO_NET_F_MRG_RXBUF);
+		}
+
+		if (enable_tx_csum == 0) {
+			rte_vhost_driver_disable_features(file,
+				1ULL << VIRTIO_NET_F_CSUM);
+		}
+
+		if (enable_tso == 0) {
+			rte_vhost_driver_disable_features(file,
+				1ULL << VIRTIO_NET_F_HOST_TSO4);
+			rte_vhost_driver_disable_features(file,
+				1ULL << VIRTIO_NET_F_HOST_TSO6);
+			rte_vhost_driver_disable_features(file,
+				1ULL << VIRTIO_NET_F_GUEST_TSO4);
+			rte_vhost_driver_disable_features(file,
+				1ULL << VIRTIO_NET_F_GUEST_TSO6);
+		}
+
+		if (promiscuous) {
+			rte_vhost_driver_enable_features(file,
+				1ULL << VIRTIO_NET_F_CTRL_RX);
+		}
+
+		ret = rte_vhost_driver_callback_register(file,
+			&virtio_net_device_ops);
+		if (ret != 0) {
+			rte_exit(EXIT_FAILURE,
+				"failed to register vhost driver callbacks.\n");
+		}
+
+		if (rte_vhost_driver_start(file) < 0) {
+			rte_exit(EXIT_FAILURE,
+				"failed to start vhost driver.\n");
+		}
 	}
 
-	rte_vhost_driver_callback_register(&virtio_net_device_ops);
+	RTE_LCORE_FOREACH_SLAVE(lcore_id)
+		rte_eal_wait_lcore(lcore_id);
 
-	rte_vhost_driver_session_start();
 	return 0;
 
 }
diff --git a/examples/vhost/main.h b/examples/vhost/main.h
index 6bb42e89..9a2aca37 100644
--- a/examples/vhost/main.h
+++ b/examples/vhost/main.h
@@ -1,7 +1,7 @@
 /*-
  *   BSD LICENSE
  *
- *   Copyright(c) 2010-2016 Intel Corporation. All rights reserved.
+ *   Copyright(c) 2010-2017 Intel Corporation. All rights reserved.
  *   All rights reserved.
  *
  *   Redistribution and use in source and binary forms, with or without
@@ -36,11 +36,17 @@
 
 #include <sys/queue.h>
 
+#include <rte_ether.h>
+
 /* Macros for printing using RTE_LOG */
 #define RTE_LOGTYPE_VHOST_CONFIG RTE_LOGTYPE_USER1
 #define RTE_LOGTYPE_VHOST_DATA   RTE_LOGTYPE_USER2
 #define RTE_LOGTYPE_VHOST_PORT   RTE_LOGTYPE_USER3
 
+enum {VIRTIO_RXQ, VIRTIO_TXQ, VIRTIO_QNUM};
+
+#define MAX_PKT_BURST 32		/* Max burst size for RX/TX */
+
 struct device_statistics {
 	uint64_t	tx;
 	uint64_t	tx_total;
@@ -48,6 +54,12 @@ struct device_statistics {
 	rte_atomic64_t	rx_total_atomic;
 };
 
+struct vhost_queue {
+	struct rte_vhost_vring	vr;
+	uint16_t		last_avail_idx;
+	uint16_t		last_used_idx;
+};
+
 struct vhost_dev {
 	/**< Number of memory regions for gpa to hpa translation. */
 	uint32_t nregions_hpa;
@@ -65,9 +77,16 @@ struct vhost_dev {
 	volatile uint8_t remove;
 
 	int vid;
+	uint64_t features;
+	size_t hdr_len;
+	uint16_t nr_vrings;
+	struct rte_vhost_memory *mem;
 	struct device_statistics stats;
 	TAILQ_ENTRY(vhost_dev) global_vdev_entry;
 	TAILQ_ENTRY(vhost_dev) lcore_vdev_entry;
+
+#define MAX_QUEUE_PAIRS	4
+	struct vhost_queue queues[MAX_QUEUE_PAIRS * 2];
 } __rte_cache_aligned;
 
 TAILQ_HEAD(vhost_dev_tailq_list, vhost_dev);
@@ -88,4 +107,15 @@ struct lcore_info {
 	struct vhost_dev_tailq_list vdev_list;
 };
 
+/* we implement non-extra virtio net features */
+#define VIRTIO_NET_FEATURES	0
+
+void vs_vhost_net_setup(struct vhost_dev *dev);
+void vs_vhost_net_remove(struct vhost_dev *dev);
+uint16_t vs_enqueue_pkts(struct vhost_dev *dev, uint16_t queue_id,
+			 struct rte_mbuf **pkts, uint32_t count);
+
+uint16_t vs_dequeue_pkts(struct vhost_dev *dev, uint16_t queue_id,
+			 struct rte_mempool *mbuf_pool,
+			 struct rte_mbuf **pkts, uint16_t count);
 #endif /* _MAIN_H_ */
diff --git a/examples/vhost/virtio_net.c b/examples/vhost/virtio_net.c
new file mode 100644
index 00000000..cc2c3d88
--- /dev/null
+++ b/examples/vhost/virtio_net.c
@@ -0,0 +1,403 @@
+/*-
+ *   BSD LICENSE
+ *
+ *   Copyright(c) 2010-2017 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 <stdbool.h>
+#include <linux/virtio_net.h>
+
+#include <rte_mbuf.h>
+#include <rte_memcpy.h>
+#include <rte_vhost.h>
+
+#include "main.h"
+
+/*
+ * A very simple vhost-user net driver implementation, without
+ * any extra features being enabled, such as TSO and mrg-Rx.
+ */
+
+void
+vs_vhost_net_setup(struct vhost_dev *dev)
+{
+	uint16_t i;
+	int vid = dev->vid;
+	struct vhost_queue *queue;
+
+	RTE_LOG(INFO, VHOST_CONFIG,
+		"setting builtin vhost-user net driver\n");
+
+	rte_vhost_get_negotiated_features(vid, &dev->features);
+	if (dev->features & (1 << VIRTIO_NET_F_MRG_RXBUF))
+		dev->hdr_len = sizeof(struct virtio_net_hdr_mrg_rxbuf);
+	else
+		dev->hdr_len = sizeof(struct virtio_net_hdr);
+
+	rte_vhost_get_mem_table(vid, &dev->mem);
+
+	dev->nr_vrings = rte_vhost_get_vring_num(vid);
+	for (i = 0; i < dev->nr_vrings; i++) {
+		queue = &dev->queues[i];
+
+		queue->last_used_idx  = 0;
+		queue->last_avail_idx = 0;
+		rte_vhost_get_vhost_vring(vid, i, &queue->vr);
+	}
+}
+
+void
+vs_vhost_net_remove(struct vhost_dev *dev)
+{
+	free(dev->mem);
+}
+
+static inline int __attribute__((always_inline))
+enqueue_pkt(struct vhost_dev *dev, struct rte_vhost_vring *vr,
+	    struct rte_mbuf *m, uint16_t desc_idx)
+{
+	uint32_t desc_avail, desc_offset;
+	uint32_t mbuf_avail, mbuf_offset;
+	uint32_t cpy_len;
+	struct vring_desc *desc;
+	uint64_t desc_addr;
+	struct virtio_net_hdr virtio_hdr = {0, 0, 0, 0, 0, 0};
+	/* A counter to avoid desc dead loop chain */
+	uint16_t nr_desc = 1;
+
+	desc = &vr->desc[desc_idx];
+	desc_addr = rte_vhost_gpa_to_vva(dev->mem, desc->addr);
+	/*
+	 * Checking of 'desc_addr' placed outside of 'unlikely' macro to avoid
+	 * performance issue with some versions of gcc (4.8.4 and 5.3.0) which
+	 * otherwise stores offset on the stack instead of in a register.
+	 */
+	if (unlikely(desc->len < dev->hdr_len) || !desc_addr)
+		return -1;
+
+	rte_prefetch0((void *)(uintptr_t)desc_addr);
+
+	/* write virtio-net header */
+	*(struct virtio_net_hdr *)(uintptr_t)desc_addr = virtio_hdr;
+
+	desc_offset = dev->hdr_len;
+	desc_avail  = desc->len - dev->hdr_len;
+
+	mbuf_avail  = rte_pktmbuf_data_len(m);
+	mbuf_offset = 0;
+	while (mbuf_avail != 0 || m->next != NULL) {
+		/* done with current mbuf, fetch next */
+		if (mbuf_avail == 0) {
+			m = m->next;
+
+			mbuf_offset = 0;
+			mbuf_avail  = rte_pktmbuf_data_len(m);
+		}
+
+		/* done with current desc buf, fetch next */
+		if (desc_avail == 0) {
+			if ((desc->flags & VRING_DESC_F_NEXT) == 0) {
+				/* Room in vring buffer is not enough */
+				return -1;
+			}
+			if (unlikely(desc->next >= vr->size ||
+				     ++nr_desc > vr->size))
+				return -1;
+
+			desc = &vr->desc[desc->next];
+			desc_addr = rte_vhost_gpa_to_vva(dev->mem, desc->addr);
+			if (unlikely(!desc_addr))
+				return -1;
+
+			desc_offset = 0;
+			desc_avail  = desc->len;
+		}
+
+		cpy_len = RTE_MIN(desc_avail, mbuf_avail);
+		rte_memcpy((void *)((uintptr_t)(desc_addr + desc_offset)),
+			rte_pktmbuf_mtod_offset(m, void *, mbuf_offset),
+			cpy_len);
+
+		mbuf_avail  -= cpy_len;
+		mbuf_offset += cpy_len;
+		desc_avail  -= cpy_len;
+		desc_offset += cpy_len;
+	}
+
+	return 0;
+}
+
+uint16_t
+vs_enqueue_pkts(struct vhost_dev *dev, uint16_t queue_id,
+		struct rte_mbuf **pkts, uint32_t count)
+{
+	struct vhost_queue *queue;
+	struct rte_vhost_vring *vr;
+	uint16_t avail_idx, free_entries, start_idx;
+	uint16_t desc_indexes[MAX_PKT_BURST];
+	uint16_t used_idx;
+	uint32_t i;
+
+	queue = &dev->queues[queue_id];
+	vr    = &queue->vr;
+
+	avail_idx = *((volatile uint16_t *)&vr->avail->idx);
+	start_idx = queue->last_used_idx;
+	free_entries = avail_idx - start_idx;
+	count = RTE_MIN(count, free_entries);
+	count = RTE_MIN(count, (uint32_t)MAX_PKT_BURST);
+	if (count == 0)
+		return 0;
+
+	/* Retrieve all of the desc indexes first to avoid caching issues. */
+	rte_prefetch0(&vr->avail->ring[start_idx & (vr->size - 1)]);
+	for (i = 0; i < count; i++) {
+		used_idx = (start_idx + i) & (vr->size - 1);
+		desc_indexes[i] = vr->avail->ring[used_idx];
+		vr->used->ring[used_idx].id = desc_indexes[i];
+		vr->used->ring[used_idx].len = pkts[i]->pkt_len +
+					       dev->hdr_len;
+	}
+
+	rte_prefetch0(&vr->desc[desc_indexes[0]]);
+	for (i = 0; i < count; i++) {
+		uint16_t desc_idx = desc_indexes[i];
+		int err;
+
+		err = enqueue_pkt(dev, vr, pkts[i], desc_idx);
+		if (unlikely(err)) {
+			used_idx = (start_idx + i) & (vr->size - 1);
+			vr->used->ring[used_idx].len = dev->hdr_len;
+		}
+
+		if (i + 1 < count)
+			rte_prefetch0(&vr->desc[desc_indexes[i+1]]);
+	}
+
+	rte_smp_wmb();
+
+	*(volatile uint16_t *)&vr->used->idx += count;
+	queue->last_used_idx += count;
+
+	/* flush used->idx update before we read avail->flags. */
+	rte_mb();
+
+	/* Kick the guest if necessary. */
+	if (!(vr->avail->flags & VRING_AVAIL_F_NO_INTERRUPT)
+			&& (vr->callfd >= 0))
+		eventfd_write(vr->callfd, (eventfd_t)1);
+	return count;
+}
+
+static inline int __attribute__((always_inline))
+dequeue_pkt(struct vhost_dev *dev, struct rte_vhost_vring *vr,
+	    struct rte_mbuf *m, uint16_t desc_idx,
+	    struct rte_mempool *mbuf_pool)
+{
+	struct vring_desc *desc;
+	uint64_t desc_addr;
+	uint32_t desc_avail, desc_offset;
+	uint32_t mbuf_avail, mbuf_offset;
+	uint32_t cpy_len;
+	struct rte_mbuf *cur = m, *prev = m;
+	/* A counter to avoid desc dead loop chain */
+	uint32_t nr_desc = 1;
+
+	desc = &vr->desc[desc_idx];
+	if (unlikely((desc->len < dev->hdr_len)) ||
+			(desc->flags & VRING_DESC_F_INDIRECT))
+		return -1;
+
+	desc_addr = rte_vhost_gpa_to_vva(dev->mem, desc->addr);
+	if (unlikely(!desc_addr))
+		return -1;
+
+	/*
+	 * We don't support ANY_LAYOUT, neither VERSION_1, meaning
+	 * a Tx packet from guest must have 2 desc buffers at least:
+	 * the first for storing the header and the others for
+	 * storing the data.
+	 *
+	 * And since we don't support TSO, we could simply skip the
+	 * header.
+	 */
+	desc = &vr->desc[desc->next];
+	desc_addr = rte_vhost_gpa_to_vva(dev->mem, desc->addr);
+	if (unlikely(!desc_addr))
+		return -1;
+	rte_prefetch0((void *)(uintptr_t)desc_addr);
+
+	desc_offset = 0;
+	desc_avail  = desc->len;
+	nr_desc    += 1;
+
+	mbuf_offset = 0;
+	mbuf_avail  = m->buf_len - RTE_PKTMBUF_HEADROOM;
+	while (1) {
+		cpy_len = RTE_MIN(desc_avail, mbuf_avail);
+		rte_memcpy(rte_pktmbuf_mtod_offset(cur, void *,
+						   mbuf_offset),
+			(void *)((uintptr_t)(desc_addr + desc_offset)),
+			cpy_len);
+
+		mbuf_avail  -= cpy_len;
+		mbuf_offset += cpy_len;
+		desc_avail  -= cpy_len;
+		desc_offset += cpy_len;
+
+		/* This desc reaches to its end, get the next one */
+		if (desc_avail == 0) {
+			if ((desc->flags & VRING_DESC_F_NEXT) == 0)
+				break;
+
+			if (unlikely(desc->next >= vr->size ||
+				     ++nr_desc > vr->size))
+				return -1;
+			desc = &vr->desc[desc->next];
+
+			desc_addr = rte_vhost_gpa_to_vva(dev->mem, desc->addr);
+			if (unlikely(!desc_addr))
+				return -1;
+			rte_prefetch0((void *)(uintptr_t)desc_addr);
+
+			desc_offset = 0;
+			desc_avail  = desc->len;
+		}
+
+		/*
+		 * This mbuf reaches to its end, get a new one
+		 * to hold more data.
+		 */
+		if (mbuf_avail == 0) {
+			cur = rte_pktmbuf_alloc(mbuf_pool);
+			if (unlikely(cur == NULL)) {
+				RTE_LOG(ERR, VHOST_DATA, "Failed to "
+					"allocate memory for mbuf.\n");
+				return -1;
+			}
+
+			prev->next = cur;
+			prev->data_len = mbuf_offset;
+			m->nb_segs += 1;
+			m->pkt_len += mbuf_offset;
+			prev = cur;
+
+			mbuf_offset = 0;
+			mbuf_avail  = cur->buf_len - RTE_PKTMBUF_HEADROOM;
+		}
+	}
+
+	prev->data_len = mbuf_offset;
+	m->pkt_len    += mbuf_offset;
+
+	return 0;
+}
+
+uint16_t
+vs_dequeue_pkts(struct vhost_dev *dev, uint16_t queue_id,
+	struct rte_mempool *mbuf_pool, struct rte_mbuf **pkts, uint16_t count)
+{
+	struct vhost_queue *queue;
+	struct rte_vhost_vring *vr;
+	uint32_t desc_indexes[MAX_PKT_BURST];
+	uint32_t used_idx;
+	uint32_t i = 0;
+	uint16_t free_entries;
+	uint16_t avail_idx;
+
+	queue = &dev->queues[queue_id];
+	vr    = &queue->vr;
+
+	free_entries = *((volatile uint16_t *)&vr->avail->idx) -
+			queue->last_avail_idx;
+	if (free_entries == 0)
+		return 0;
+
+	/* Prefetch available and used ring */
+	avail_idx = queue->last_avail_idx & (vr->size - 1);
+	used_idx  = queue->last_used_idx  & (vr->size - 1);
+	rte_prefetch0(&vr->avail->ring[avail_idx]);
+	rte_prefetch0(&vr->used->ring[used_idx]);
+
+	count = RTE_MIN(count, MAX_PKT_BURST);
+	count = RTE_MIN(count, free_entries);
+
+	/*
+	 * Retrieve all of the head indexes first and pre-update used entries
+	 * to avoid caching issues.
+	 */
+	for (i = 0; i < count; i++) {
+		avail_idx = (queue->last_avail_idx + i) & (vr->size - 1);
+		used_idx  = (queue->last_used_idx  + i) & (vr->size - 1);
+		desc_indexes[i] = vr->avail->ring[avail_idx];
+
+		vr->used->ring[used_idx].id  = desc_indexes[i];
+		vr->used->ring[used_idx].len = 0;
+	}
+
+	/* Prefetch descriptor index. */
+	rte_prefetch0(&vr->desc[desc_indexes[0]]);
+	for (i = 0; i < count; i++) {
+		int err;
+
+		if (likely(i + 1 < count))
+			rte_prefetch0(&vr->desc[desc_indexes[i + 1]]);
+
+		pkts[i] = rte_pktmbuf_alloc(mbuf_pool);
+		if (unlikely(pkts[i] == NULL)) {
+			RTE_LOG(ERR, VHOST_DATA,
+				"Failed to allocate memory for mbuf.\n");
+			break;
+		}
+
+		err = dequeue_pkt(dev, vr, pkts[i], desc_indexes[i], mbuf_pool);
+		if (unlikely(err)) {
+			rte_pktmbuf_free(pkts[i]);
+			break;
+		}
+
+	}
+	if (!i)
+		return 0;
+
+	queue->last_avail_idx += i;
+	queue->last_used_idx += i;
+	rte_smp_wmb();
+	rte_smp_rmb();
+
+	vr->used->idx += i;
+
+	if (!(vr->avail->flags & VRING_AVAIL_F_NO_INTERRUPT)
+			&& (vr->callfd >= 0))
+		eventfd_write(vr->callfd, (eventfd_t)1);
+
+	return i;
+}