[RFC 2/2] examples/vhost: support async dequeue data path

[xuan.ding@intel.com]

From: Xuan Ding <xuan.ding@intel.com>

This patch adds the use case for async dequeue API. Vswitch can
leverage DMA device to accelerate vhost async dequeue path.

Signed-off-by: Wenwu Ma <wenwux.ma@intel.com>
Signed-off-by: Yuan Wang <yuan.wangx@intel.com>
Signed-off-by: Xuan Ding <xuan.ding@intel.com>
---
 doc/guides/sample_app_ug/vhost.rst |   9 +-
 examples/vhost/main.c              | 301 +++++++++++++++++++----------
 examples/vhost/main.h              |  35 +++-
 examples/vhost/virtio_net.c        |  16 +-
 4 files changed, 255 insertions(+), 106 deletions(-)

diff --git a/doc/guides/sample_app_ug/vhost.rst b/doc/guides/sample_app_ug/vhost.rst
index a6ce4bc8ac..09db965e70 100644
--- a/doc/guides/sample_app_ug/vhost.rst
+++ b/doc/guides/sample_app_ug/vhost.rst
@@ -169,9 +169,12 @@ demonstrates how to use the async vhost APIs. It's used in combination with dmas
 **--dmas**
 This parameter is used to specify the assigned DMA device of a vhost device.
 Async vhost-user net driver will be used if --dmas is set. For example
---dmas [txd0@00:04.0,txd1@00:04.1] means use DMA channel 00:04.0 for vhost
-device 0 enqueue operation and use DMA channel 00:04.1 for vhost device 1
-enqueue operation.
+--dmas [txd0@00:04.0,txd1@00:04.1,rxd0@00:04.2,rxd1@00:04.3] means use
+DMA channel 00:04.0/00:04.2 for vhost device 0 enqueue/dequeue operation
+and use DMA channel 00:04.1/00:04.3 for vhost device 1 enqueue/dequeue
+operation. The index of the device corresponds to the socket file in order,
+that means vhost device 0 is created through the first socket file, vhost
+device 1 is created through the second socket file, and so on.
 
 Common Issues
 -------------
diff --git a/examples/vhost/main.c b/examples/vhost/main.c
index 44073499bc..8c952d3874 100644
--- a/examples/vhost/main.c
+++ b/examples/vhost/main.c
@@ -62,6 +62,9 @@
 #define MAX_VHOST_DEVICE 1024
 #define DMA_RING_SIZE 4096
 
+#define ASYNC_ENQUEUE_VHOST 1
+#define ASYNC_DEQUEUE_VHOST 2
+
 struct dma_for_vhost dma_bind[MAX_VHOST_DEVICE];
 struct rte_vhost_async_dma_info dma_config[RTE_DMADEV_DEFAULT_MAX];
 static int dma_count;
@@ -103,8 +106,6 @@ static int client_mode;
 
 static int builtin_net_driver;
 
-static int async_vhost_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. */
@@ -114,6 +115,8 @@ static uint32_t burst_rx_retry_num = BURST_RX_RETRIES;
 static char *socket_files;
 static int nb_sockets;
 
+static struct vhost_queue_ops vdev_queue_ops[MAX_VHOST_DEVICE];
+
 /* empty vmdq configuration structure. Filled in programatically */
 static struct rte_eth_conf vmdq_conf_default = {
 	.rxmode = {
@@ -203,6 +206,18 @@ struct vhost_bufftable *vhost_txbuff[RTE_MAX_LCORE * MAX_VHOST_DEVICE];
 #define MBUF_TABLE_DRAIN_TSC	((rte_get_tsc_hz() + US_PER_S - 1) \
 				 / US_PER_S * BURST_TX_DRAIN_US)
 
+static int vid2socketid[MAX_VHOST_DEVICE];
+
+static uint32_t get_async_flag_by_socketid(int socketid)
+{
+	return dma_bind[socketid].async_flag;
+}
+
+static void init_vid2socketid_array(int vid, int socketid)
+{
+	vid2socketid[vid] = socketid;
+}
+
 static inline bool
 is_dma_configured(int16_t dev_id)
 {
@@ -224,7 +239,7 @@ open_dma(const char *value)
 	char *addrs = input;
 	char *ptrs[2];
 	char *start, *end, *substr;
-	int64_t vid, vring_id;
+	int64_t socketid, vring_id;
 
 	struct rte_dma_info info;
 	struct rte_dma_conf dev_config = { .nb_vchans = 1 };
@@ -263,7 +278,9 @@ open_dma(const char *value)
 
 	while (i < args_nr) {
 		char *arg_temp = dma_arg[i];
+		char *txd, *rxd;
 		uint8_t sub_nr;
+		int async_flag;
 
 		sub_nr = rte_strsplit(arg_temp, strlen(arg_temp), ptrs, 2, '@');
 		if (sub_nr != 2) {
@@ -271,21 +288,28 @@ open_dma(const char *value)
 			goto out;
 		}
 
-		start = strstr(ptrs[0], "txd");
-		if (start == NULL) {
+		txd = strstr(ptrs[0], "txd");
+		rxd = strstr(ptrs[0], "rxd");
+		if (txd) {
+			start = txd;
+			vring_id = VIRTIO_RXQ;
+			async_flag = ASYNC_ENQUEUE_VHOST;
+		} else if (rxd) {
+			start = rxd;
+			vring_id = VIRTIO_TXQ;
+			async_flag = ASYNC_DEQUEUE_VHOST;
+		} else {
 			ret = -1;
 			goto out;
 		}
 
 		start += 3;
-		vid = strtol(start, &end, 0);
+		socketid = strtol(start, &end, 0);
 		if (end == start) {
 			ret = -1;
 			goto out;
 		}
 
-		vring_id = 0 + VIRTIO_RXQ;
-
 		dev_id = rte_dma_get_dev_id_by_name(ptrs[1]);
 		if (dev_id < 0) {
 			RTE_LOG(ERR, VHOST_CONFIG, "Fail to find DMA %s.\n", ptrs[1]);
@@ -325,7 +349,8 @@ open_dma(const char *value)
 		dma_config[dma_count++].max_desc = DMA_RING_SIZE;
 
 done:
-		(dma_info + vid)->dmas[vring_id].dev_id = dev_id;
+		(dma_info + socketid)->dmas[vring_id].dev_id = dev_id;
+		(dma_info + socketid)->async_flag |= async_flag;
 		i++;
 	}
 out:
@@ -792,7 +817,6 @@ us_vhost_parse_args(int argc, char **argv)
 				us_vhost_usage(prgname);
 				return -1;
 			}
-			async_vhost_driver = 1;
 			break;
 
 		case OPT_CLIENT_NUM:
@@ -961,13 +985,13 @@ complete_async_pkts(struct vhost_dev *vdev)
 {
 	struct rte_mbuf *p_cpl[MAX_PKT_BURST];
 	uint16_t complete_count;
-	int16_t dma_id = dma_bind[vdev->vid].dmas[VIRTIO_RXQ].dev_id;
+	int16_t dma_id = dma_bind[vid2socketid[vdev->vid]].dmas[VIRTIO_RXQ].dev_id;
 
 	complete_count = rte_vhost_poll_enqueue_completed(vdev->vid,
 					VIRTIO_RXQ, p_cpl, MAX_PKT_BURST, dma_id, 0);
 	if (complete_count) {
 		free_pkts(p_cpl, complete_count);
-		__atomic_sub_fetch(&vdev->pkts_inflight, complete_count, __ATOMIC_SEQ_CST);
+		__atomic_sub_fetch(&vdev->pkts_enq_inflight, complete_count, __ATOMIC_SEQ_CST);
 	}
 
 }
@@ -1002,23 +1026,7 @@ drain_vhost(struct vhost_dev *vdev)
 	uint16_t nr_xmit = vhost_txbuff[buff_idx]->len;
 	struct rte_mbuf **m = vhost_txbuff[buff_idx]->m_table;
 
-	if (builtin_net_driver) {
-		ret = vs_enqueue_pkts(vdev, VIRTIO_RXQ, m, nr_xmit);
-	} else if (dma_bind[vdev->vid].dmas[VIRTIO_RXQ].async_enabled) {
-		uint16_t enqueue_fail = 0;
-		int16_t dma_id = dma_bind[vdev->vid].dmas[VIRTIO_RXQ].dev_id;
-
-		complete_async_pkts(vdev);
-		ret = rte_vhost_submit_enqueue_burst(vdev->vid, VIRTIO_RXQ, m, nr_xmit, dma_id, 0);
-		__atomic_add_fetch(&vdev->pkts_inflight, ret, __ATOMIC_SEQ_CST);
-
-		enqueue_fail = nr_xmit - ret;
-		if (enqueue_fail)
-			free_pkts(&m[ret], nr_xmit - ret);
-	} else {
-		ret = rte_vhost_enqueue_burst(vdev->vid, VIRTIO_RXQ,
-						m, nr_xmit);
-	}
+	ret = vdev_queue_ops[vdev->vid].enqueue_pkt_burst(vdev, VIRTIO_RXQ, m, nr_xmit);
 
 	if (enable_stats) {
 		__atomic_add_fetch(&vdev->stats.rx_total_atomic, nr_xmit,
@@ -1027,7 +1035,7 @@ drain_vhost(struct vhost_dev *vdev)
 				__ATOMIC_SEQ_CST);
 	}
 
-	if (!dma_bind[vdev->vid].dmas[VIRTIO_RXQ].async_enabled)
+	if (!dma_bind[vid2socketid[vdev->vid]].dmas[VIRTIO_RXQ].async_enabled)
 		free_pkts(m, nr_xmit);
 }
 
@@ -1300,6 +1308,33 @@ drain_mbuf_table(struct mbuf_table *tx_q)
 	}
 }
 
+uint16_t
+async_enqueue_pkts(struct vhost_dev *dev, uint16_t queue_id,
+		struct rte_mbuf **pkts, uint32_t rx_count)
+{
+	uint16_t enqueue_count;
+	uint16_t enqueue_fail = 0;
+	uint16_t dma_id = dma_bind[vid2socketid[dev->vid]].dmas[VIRTIO_RXQ].dev_id;
+
+	complete_async_pkts(dev);
+	enqueue_count = rte_vhost_submit_enqueue_burst(dev->vid, queue_id,
+					pkts, rx_count, dma_id, 0);
+	__atomic_add_fetch(&dev->pkts_enq_inflight, enqueue_count, __ATOMIC_SEQ_CST);
+
+	enqueue_fail = rx_count - enqueue_count;
+	if (enqueue_fail)
+		free_pkts(&pkts[enqueue_count], enqueue_fail);
+
+	return enqueue_count;
+}
+
+uint16_t
+sync_enqueue_pkts(struct vhost_dev *dev, uint16_t queue_id,
+		struct rte_mbuf **pkts, uint32_t rx_count)
+{
+	return rte_vhost_enqueue_burst(dev->vid, queue_id, pkts, rx_count);
+}
+
 static __rte_always_inline void
 drain_eth_rx(struct vhost_dev *vdev)
 {
@@ -1330,26 +1365,8 @@ drain_eth_rx(struct vhost_dev *vdev)
 		}
 	}
 
-	if (builtin_net_driver) {
-		enqueue_count = vs_enqueue_pkts(vdev, VIRTIO_RXQ,
-						pkts, rx_count);
-	} else if (dma_bind[vdev->vid].dmas[VIRTIO_RXQ].async_enabled) {
-		uint16_t enqueue_fail = 0;
-		int16_t dma_id = dma_bind[vdev->vid].dmas[VIRTIO_RXQ].dev_id;
-
-		complete_async_pkts(vdev);
-		enqueue_count = rte_vhost_submit_enqueue_burst(vdev->vid,
-					VIRTIO_RXQ, pkts, rx_count, dma_id, 0);
-		__atomic_add_fetch(&vdev->pkts_inflight, enqueue_count, __ATOMIC_SEQ_CST);
-
-		enqueue_fail = rx_count - enqueue_count;
-		if (enqueue_fail)
-			free_pkts(&pkts[enqueue_count], enqueue_fail);
-
-	} else {
-		enqueue_count = rte_vhost_enqueue_burst(vdev->vid, VIRTIO_RXQ,
-						pkts, rx_count);
-	}
+	enqueue_count = vdev_queue_ops[vdev->vid].enqueue_pkt_burst(vdev,
+					VIRTIO_RXQ, pkts, rx_count);
 
 	if (enable_stats) {
 		__atomic_add_fetch(&vdev->stats.rx_total_atomic, rx_count,
@@ -1358,10 +1375,33 @@ drain_eth_rx(struct vhost_dev *vdev)
 				__ATOMIC_SEQ_CST);
 	}
 
-	if (!dma_bind[vdev->vid].dmas[VIRTIO_RXQ].async_enabled)
+	if (!dma_bind[vid2socketid[vdev->vid]].dmas[VIRTIO_RXQ].async_enabled)
 		free_pkts(pkts, rx_count);
 }
 
+uint16_t async_dequeue_pkts(struct vhost_dev *dev, uint16_t queue_id,
+			    struct rte_mempool *mbuf_pool,
+			    struct rte_mbuf **pkts, uint16_t count)
+{
+	int nr_inflight;
+	uint16_t dequeue_count;
+	uint16_t dma_id = dma_bind[vid2socketid[dev->vid]].dmas[VIRTIO_TXQ].dev_id;
+
+	dequeue_count = rte_vhost_async_try_dequeue_burst(dev->vid, queue_id,
+			mbuf_pool, pkts, count, &nr_inflight, dma_id, 0);
+	if (likely(nr_inflight != -1))
+		dev->pkts_deq_inflight = nr_inflight;
+
+	return dequeue_count;
+}
+
+uint16_t sync_dequeue_pkts(struct vhost_dev *dev, uint16_t queue_id,
+			   struct rte_mempool *mbuf_pool,
+			   struct rte_mbuf **pkts, uint16_t count)
+{
+	return rte_vhost_dequeue_burst(dev->vid, queue_id, mbuf_pool, pkts, count);
+}
+
 static __rte_always_inline void
 drain_virtio_tx(struct vhost_dev *vdev)
 {
@@ -1369,13 +1409,8 @@ drain_virtio_tx(struct vhost_dev *vdev)
 	uint16_t count;
 	uint16_t i;
 
-	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);
-	}
+	count = vdev_queue_ops[vdev->vid].dequeue_pkt_burst(vdev,
+				VIRTIO_TXQ, mbuf_pool, pkts, MAX_PKT_BURST);
 
 	/* setup VMDq for the first packet */
 	if (unlikely(vdev->ready == DEVICE_MAC_LEARNING) && count) {
@@ -1454,6 +1489,31 @@ switch_worker(void *arg __rte_unused)
 	return 0;
 }
 
+static void
+vhost_clear_queue_thread_unsafe(struct vhost_dev *vdev, uint16_t queue_id)
+{
+	uint16_t n_pkt = 0;
+	uint16_t dma_id = dma_bind[vid2socketid[vdev->vid]].dmas[queue_id].dev_id;
+	struct rte_mbuf *m_enq_cpl[vdev->pkts_enq_inflight];
+	struct rte_mbuf *m_deq_cpl[vdev->pkts_deq_inflight];
+
+	if (queue_id % 2 == 0) {
+		while (vdev->pkts_enq_inflight) {
+			n_pkt = rte_vhost_clear_queue_thread_unsafe(vdev->vid,
+				queue_id, m_enq_cpl, vdev->pkts_enq_inflight, dma_id, 0);
+			free_pkts(m_enq_cpl, n_pkt);
+			__atomic_sub_fetch(&vdev->pkts_enq_inflight, n_pkt, __ATOMIC_SEQ_CST);
+		}
+	} else {
+		while (vdev->pkts_deq_inflight) {
+			n_pkt = rte_vhost_clear_queue_thread_unsafe(vdev->vid,
+				queue_id, m_deq_cpl, vdev->pkts_deq_inflight, dma_id, 0);
+			free_pkts(m_deq_cpl, n_pkt);
+			__atomic_sub_fetch(&vdev->pkts_deq_inflight, n_pkt, __ATOMIC_SEQ_CST);
+		}
+	}
+}
+
 /*
  * Remove a device from the specific data core linked list and from the
  * main linked list. Synchonization  occurs through the use of the
@@ -1510,25 +1570,83 @@ destroy_device(int vid)
 		"(%d) device has been removed from data core\n",
 		vdev->vid);
 
-	if (dma_bind[vid].dmas[VIRTIO_RXQ].async_enabled) {
-		uint16_t n_pkt = 0;
-		int16_t dma_id = dma_bind[vid].dmas[VIRTIO_RXQ].dev_id;
-		struct rte_mbuf *m_cpl[vdev->pkts_inflight];
-
-		while (vdev->pkts_inflight) {
-			n_pkt = rte_vhost_clear_queue_thread_unsafe(vid, VIRTIO_RXQ,
-						m_cpl, vdev->pkts_inflight, dma_id, 0);
-			free_pkts(m_cpl, n_pkt);
-			__atomic_sub_fetch(&vdev->pkts_inflight, n_pkt, __ATOMIC_SEQ_CST);
-		}
-
+	if (dma_bind[vid2socketid[vid]].dmas[VIRTIO_RXQ].async_enabled) {
+		vhost_clear_queue_thread_unsafe(vdev, VIRTIO_RXQ);
 		rte_vhost_async_channel_unregister(vid, VIRTIO_RXQ);
-		dma_bind[vid].dmas[VIRTIO_RXQ].async_enabled = false;
+		dma_bind[vid2socketid[vid]].dmas[VIRTIO_RXQ].async_enabled = false;
+	}
+	if (dma_bind[vid2socketid[vid]].dmas[VIRTIO_TXQ].async_enabled) {
+		vhost_clear_queue_thread_unsafe(vdev, VIRTIO_TXQ);
+		rte_vhost_async_channel_unregister(vid, VIRTIO_TXQ);
+		dma_bind[vid2socketid[vid]].dmas[VIRTIO_TXQ].async_enabled = false;
 	}
 
 	rte_free(vdev);
 }
 
+static int
+get_socketid_by_vid(int vid)
+{
+	int i;
+	char ifname[PATH_MAX];
+	rte_vhost_get_ifname(vid, ifname, sizeof(ifname));
+
+	for (i = 0; i < nb_sockets; i++) {
+		char *file = socket_files + i * PATH_MAX;
+		if (strcmp(file, ifname) == 0)
+			return i;
+	}
+
+	return -1;
+}
+
+static int
+init_vhost_queue_ops(int vid)
+{
+	int socketid = get_socketid_by_vid(vid);
+	if (socketid == -1)
+		return -1;
+
+	init_vid2socketid_array(vid, socketid);
+	if (builtin_net_driver) {
+		vdev_queue_ops[vid].enqueue_pkt_burst = builtin_enqueue_pkts;
+		vdev_queue_ops[vid].dequeue_pkt_burst = builtin_dequeue_pkts;
+	} else {
+		if (dma_bind[vid2socketid[vid]].dmas[VIRTIO_RXQ].async_enabled)
+			vdev_queue_ops[vid].enqueue_pkt_burst = async_enqueue_pkts;
+		else
+			vdev_queue_ops[vid].enqueue_pkt_burst = sync_enqueue_pkts;
+
+		if (dma_bind[vid2socketid[vid]].dmas[VIRTIO_TXQ].async_enabled)
+			vdev_queue_ops[vid].dequeue_pkt_burst = async_dequeue_pkts;
+		else
+			vdev_queue_ops[vid].dequeue_pkt_burst = sync_dequeue_pkts;
+	}
+
+	return 0;
+}
+
+static int
+vhost_async_channel_register(int vid)
+{
+	int rx_ret = 0, tx_ret = 0;
+
+	if (dma_bind[vid2socketid[vid]].dmas[VIRTIO_RXQ].dev_id != INVALID_DMA_ID) {
+		rx_ret = rte_vhost_async_channel_register(vid, VIRTIO_RXQ);
+		if (rx_ret == 0)
+			dma_bind[vid2socketid[vid]].dmas[VIRTIO_RXQ].async_enabled = true;
+	}
+
+	if (dma_bind[vid2socketid[vid]].dmas[VIRTIO_TXQ].dev_id != INVALID_DMA_ID) {
+		tx_ret = rte_vhost_async_channel_register(vid, VIRTIO_TXQ);
+		if (tx_ret == 0)
+			dma_bind[vid2socketid[vid]].dmas[VIRTIO_TXQ].async_enabled = true;
+	}
+
+	return rx_ret | tx_ret;
+}
+
+
 /*
  * A new device is added to a data core. First the device is added to the main linked list
  * and then allocated to a specific data core.
@@ -1540,6 +1658,8 @@ new_device(int vid)
 	uint16_t i;
 	uint32_t device_num_min = num_devices;
 	struct vhost_dev *vdev;
+	int ret;
+
 	vdev = rte_zmalloc("vhost device", sizeof(*vdev), RTE_CACHE_LINE_SIZE);
 	if (vdev == NULL) {
 		RTE_LOG(INFO, VHOST_DATA,
@@ -1593,17 +1713,12 @@ new_device(int vid)
 		"(%d) device has been added to data core %d\n",
 		vid, vdev->coreid);
 
-	if (dma_bind[vid].dmas[VIRTIO_RXQ].dev_id != INVALID_DMA_ID) {
-		int ret;
+	ret =  vhost_async_channel_register(vid);
 
-		ret = rte_vhost_async_channel_register(vid, VIRTIO_RXQ);
-		if (ret == 0) {
-			dma_bind[vid].dmas[VIRTIO_RXQ].async_enabled = true;
-		}
-		return ret;
-	}
+	if (init_vhost_queue_ops(vid) != 0)
+		return -1;
 
-	return 0;
+	return ret;
 }
 
 static int
@@ -1621,19 +1736,9 @@ vring_state_changed(int vid, uint16_t queue_id, int enable)
 	if (queue_id != VIRTIO_RXQ)
 		return 0;
 
-	if (dma_bind[vid].dmas[queue_id].async_enabled) {
-		if (!enable) {
-			uint16_t n_pkt = 0;
-			int16_t dma_id = dma_bind[vid].dmas[VIRTIO_RXQ].dev_id;
-			struct rte_mbuf *m_cpl[vdev->pkts_inflight];
-
-			while (vdev->pkts_inflight) {
-				n_pkt = rte_vhost_clear_queue_thread_unsafe(vid, queue_id,
-							m_cpl, vdev->pkts_inflight, dma_id, 0);
-				free_pkts(m_cpl, n_pkt);
-				__atomic_sub_fetch(&vdev->pkts_inflight, n_pkt, __ATOMIC_SEQ_CST);
-			}
-		}
+	if (dma_bind[vid2socketid[vid]].dmas[queue_id].async_enabled) {
+		if (!enable)
+			vhost_clear_queue_thread_unsafe(vdev, queue_id);
 	}
 
 	return 0;
@@ -1896,7 +2001,7 @@ main(int argc, char *argv[])
 	if (client_mode)
 		flags |= RTE_VHOST_USER_CLIENT;
 
-	if (async_vhost_driver) {
+	if (dma_count > 0) {
 		if (rte_vhost_async_dma_configure(dma_config, dma_count) < 0) {
 			RTE_LOG(ERR, VHOST_PORT, "Failed to configure DMA in vhost.\n");
 			for (i = 0; i < dma_count; i++) {
@@ -1906,18 +2011,18 @@ main(int argc, char *argv[])
 				}
 			}
 			dma_count = 0;
-			async_vhost_driver = false;
 		}
 	}
 
 	/* Register vhost user driver to handle vhost messages. */
 	for (i = 0; i < nb_sockets; i++) {
 		char *file = socket_files + i * PATH_MAX;
+		uint64_t flag = flags;
 
-		if (async_vhost_driver)
-			flags = flags | RTE_VHOST_USER_ASYNC_COPY;
+		if (dma_count > 0 && get_async_flag_by_socketid(i) != 0)
+			flag |= RTE_VHOST_USER_ASYNC_COPY;
 
-		ret = rte_vhost_driver_register(file, flags);
+		ret = rte_vhost_driver_register(file, flag);
 		if (ret != 0) {
 			unregister_drivers(i);
 			rte_exit(EXIT_FAILURE,
diff --git a/examples/vhost/main.h b/examples/vhost/main.h
index b4a453e77e..40ac2841d1 100644
--- a/examples/vhost/main.h
+++ b/examples/vhost/main.h
@@ -52,7 +52,8 @@ struct vhost_dev {
 	uint64_t features;
 	size_t hdr_len;
 	uint16_t nr_vrings;
-	uint16_t pkts_inflight;
+	uint16_t pkts_enq_inflight;
+	uint16_t pkts_deq_inflight;
 	struct rte_vhost_memory *mem;
 	struct device_statistics stats;
 	TAILQ_ENTRY(vhost_dev) global_vdev_entry;
@@ -62,6 +63,19 @@ struct vhost_dev {
 	struct vhost_queue queues[MAX_QUEUE_PAIRS * 2];
 } __rte_cache_aligned;
 
+typedef uint16_t (*vhost_enqueue_burst_t)(struct vhost_dev *dev,
+			uint16_t queue_id, struct rte_mbuf **pkts,
+			uint32_t count);
+
+typedef uint16_t (*vhost_dequeue_burst_t)(struct vhost_dev *dev,
+			uint16_t queue_id, struct rte_mempool *mbuf_pool,
+			struct rte_mbuf **pkts, uint16_t count);
+
+struct vhost_queue_ops {
+	vhost_enqueue_burst_t enqueue_pkt_burst;
+	vhost_dequeue_burst_t dequeue_pkt_burst;
+};
+
 TAILQ_HEAD(vhost_dev_tailq_list, vhost_dev);
 
 
@@ -88,6 +102,7 @@ struct dma_info {
 
 struct dma_for_vhost {
 	struct dma_info dmas[RTE_MAX_QUEUES_PER_PORT * 2];
+	uint32_t async_flag;
 };
 
 /* we implement non-extra virtio net features */
@@ -98,7 +113,19 @@ 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);
+uint16_t builtin_enqueue_pkts(struct vhost_dev *dev, uint16_t queue_id,
+			struct rte_mbuf **pkts, uint32_t count);
+uint16_t builtin_dequeue_pkts(struct vhost_dev *dev, uint16_t queue_id,
+			struct rte_mempool *mbuf_pool,
+			struct rte_mbuf **pkts, uint16_t count);
+uint16_t sync_enqueue_pkts(struct vhost_dev *dev, uint16_t queue_id,
+			 struct rte_mbuf **pkts, uint32_t count);
+uint16_t sync_dequeue_pkts(struct vhost_dev *dev, uint16_t queue_id,
+			struct rte_mempool *mbuf_pool,
+			struct rte_mbuf **pkts, uint16_t count);
+uint16_t async_enqueue_pkts(struct vhost_dev *dev, uint16_t queue_id,
+			 struct rte_mbuf **pkts, uint32_t count);
+uint16_t async_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
index 9064fc3a82..2432a96566 100644
--- a/examples/vhost/virtio_net.c
+++ b/examples/vhost/virtio_net.c
@@ -238,6 +238,13 @@ vs_enqueue_pkts(struct vhost_dev *dev, uint16_t queue_id,
 	return count;
 }
 
+uint16_t
+builtin_enqueue_pkts(struct vhost_dev *dev, uint16_t queue_id,
+		struct rte_mbuf **pkts, uint32_t count)
+{
+	return vs_enqueue_pkts(dev, queue_id, pkts, count);
+}
+
 static __rte_always_inline int
 dequeue_pkt(struct vhost_dev *dev, struct rte_vhost_vring *vr,
 	    struct rte_mbuf *m, uint16_t desc_idx,
@@ -363,7 +370,7 @@ dequeue_pkt(struct vhost_dev *dev, struct rte_vhost_vring *vr,
 	return 0;
 }
 
-uint16_t
+static 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)
 {
@@ -440,3 +447,10 @@ vs_dequeue_pkts(struct vhost_dev *dev, uint16_t queue_id,
 
 	return i;
 }
+
+uint16_t
+builtin_dequeue_pkts(struct vhost_dev *dev, uint16_t queue_id,
+	struct rte_mempool *mbuf_pool, struct rte_mbuf **pkts, uint16_t count)
+{
+	return vs_dequeue_pkts(dev, queue_id, mbuf_pool, pkts, count);
+}
-- 
2.17.1