[dpdk-dev] [PATCH v5] vhost: add support for interrupt mode

[Junjie Chen <junjie.j.chen@intel.com>]

In some cases we want vhost dequeue work in interrupt mode to
release cpus to others when no data to transmit. So we install
interrupt handler of vhost device and interrupt vectors for each
rx queue when creating new backend according to vhost intrerupt
configuration. Thus, applications could register a epoll event fd
to associate rx queues with interrupt vectors.

Signed-off-by: Junjie Chen <junjie.j.chen@intel.com>
---
Changes in v5:
- update license to DPDK new license format
- rebase code to master 
Changes in v4:
- revert back license change
Changes in v3:
- handle failure in the middle of intr setup.
- use vhost API to enable interrupt.
- rebase to check rxq existence.
- update vhost API to support guest notification.
Changes in v2:
- update rx queue index.
- fill efd_counter_size for intr handler.
- update log.
 drivers/net/vhost/rte_eth_vhost.c | 267 +++++++++++++++++++++++++++++---------
 lib/librte_vhost/vhost.c          |  13 +-
 2 files changed, 212 insertions(+), 68 deletions(-)

diff --git a/drivers/net/vhost/rte_eth_vhost.c b/drivers/net/vhost/rte_eth_vhost.c
index 11b6076..536e089 100644
--- a/drivers/net/vhost/rte_eth_vhost.c
+++ b/drivers/net/vhost/rte_eth_vhost.c
@@ -1,35 +1,8 @@
-/*-
- *   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.
+/* SPDX-License-Identifier: BSD-3-Clause
+ * Copyright(c) 2016 IGEL Co., Ltd.
+ * Copyright(c) 2016-2018 Intel Corporation
  */
+
 #include <unistd.h>
 #include <pthread.h>
 #include <stdbool.h>
@@ -528,10 +501,13 @@ update_queuing_status(struct rte_eth_dev *dev)
 	unsigned int i;
 	int allow_queuing = 1;
 
-	if (rte_atomic32_read(&internal->dev_attached) == 0)
+	if (!dev->data->rx_queues || !dev->data->tx_queues) {
+		RTE_LOG(ERR, PMD, "RX/TX queues not setup yet\n");
 		return;
+	}
 
-	if (rte_atomic32_read(&internal->started) == 0)
+	if (rte_atomic32_read(&internal->started) == 0 ||
+		rte_atomic32_read(&internal->dev_attached) == 0)
 		allow_queuing = 0;
 
 	/* Wait until rx/tx_pkt_burst stops accessing vhost device */
@@ -554,25 +530,157 @@ update_queuing_status(struct rte_eth_dev *dev)
 	}
 }
 
+static int
+eth_rxq_intr_enable(struct rte_eth_dev *dev, uint16_t qid)
+{
+	struct rte_vhost_vring vring;
+	struct vhost_queue *vq;
+	int ret = 0;
+
+	vq = dev->data->rx_queues[qid];
+	if (!vq) {
+		RTE_LOG(ERR, PMD, "rxq%d is not setup yet\n", qid);
+		return -1;
+	}
+
+	ret = rte_vhost_get_vhost_vring(vq->vid, (qid << 1) + 1, &vring);
+	if (ret < 0) {
+		RTE_LOG(ERR, PMD, "Failed to get rxq%d's vring\n", qid);
+		return ret;
+	}
+	RTE_LOG(INFO, PMD, "Enable interrupt for rxq%d\n", qid);
+	rte_vhost_enable_guest_notification(vq->vid, (qid << 1) + 1, 1);
+	rte_wmb();
+
+	return ret;
+}
+
+static int
+eth_rxq_intr_disable(struct rte_eth_dev *dev, uint16_t qid)
+{
+	struct rte_vhost_vring vring;
+	struct vhost_queue *vq;
+	int ret = 0;
+
+	vq = dev->data->rx_queues[qid];
+	if (!vq) {
+		RTE_LOG(ERR, PMD, "rxq%d is not setup yet\n", qid);
+		return -1;
+	}
+
+	ret = rte_vhost_get_vhost_vring(vq->vid, (qid << 1) + 1, &vring);
+	if (ret < 0) {
+		RTE_LOG(ERR, PMD, "Failed to get rxq%d's vring", qid);
+		return ret;
+	}
+	RTE_LOG(INFO, PMD, "Disable interrupt for rxq%d\n", qid);
+	rte_vhost_enable_guest_notification(vq->vid, (qid << 1) + 1, 0);
+	rte_wmb();
+
+	return 0;
+}
+
+static void
+eth_vhost_uninstall_intr(struct rte_eth_dev *dev)
+{
+	struct rte_intr_handle *intr_handle = dev->intr_handle;
+
+	if (intr_handle) {
+		if (intr_handle->intr_vec)
+			free(intr_handle->intr_vec);
+		free(intr_handle);
+	}
+
+	dev->intr_handle = NULL;
+}
+
+static int
+eth_vhost_install_intr(struct rte_eth_dev *dev)
+{
+	struct rte_vhost_vring vring;
+	struct vhost_queue *vq;
+	int count = 0;
+	int nb_rxq = dev->data->nb_rx_queues;
+	int i;
+	int ret;
+
+	/* uninstall firstly if we are reconnecting */
+	if (dev->intr_handle)
+		eth_vhost_uninstall_intr(dev);
+
+	dev->intr_handle = malloc(sizeof(*dev->intr_handle));
+	if (!dev->intr_handle) {
+		RTE_LOG(ERR, PMD, "Fail to allocate intr_handle\n");
+		return -ENOMEM;
+	}
+	memset(dev->intr_handle, 0, sizeof(*dev->intr_handle));
+
+	dev->intr_handle->efd_counter_size = sizeof(uint64_t);
+
+	dev->intr_handle->intr_vec =
+		malloc(nb_rxq * sizeof(dev->intr_handle->intr_vec[0]));
+
+	if (!dev->intr_handle->intr_vec) {
+		RTE_LOG(ERR, PMD,
+			"Failed to allocate memory for interrupt vector\n");
+		free(dev->intr_handle);
+		return -ENOMEM;
+	}
+
+	RTE_LOG(INFO, PMD, "Prepare intr vec\n");
+	for (i = 0; i < nb_rxq; i++) {
+		vq = dev->data->rx_queues[i];
+		if (!vq) {
+			RTE_LOG(INFO, PMD, "rxq-%d not setup yet, skip!\n", i);
+			continue;
+		}
+
+		ret = rte_vhost_get_vhost_vring(vq->vid, (i << 1) + 1, &vring);
+		if (ret < 0) {
+			RTE_LOG(INFO, PMD,
+				"Failed to get rxq-%d's vring, skip!\n", i);
+			continue;
+		}
+
+		if (vring.kickfd < 0) {
+			RTE_LOG(INFO, PMD,
+				"rxq-%d's kickfd is invalid, skip!\n", i);
+			continue;
+		}
+		dev->intr_handle->intr_vec[i] = RTE_INTR_VEC_RXTX_OFFSET + i;
+		dev->intr_handle->efds[i] = vring.kickfd;
+		count++;
+		RTE_LOG(INFO, PMD, "Installed intr vec for rxq-%d\n", i);
+	}
+
+	dev->intr_handle->nb_efd = count;
+	dev->intr_handle->max_intr = count + 1;
+	dev->intr_handle->type = RTE_INTR_HANDLE_VDEV;
+
+	return 0;
+}
+
 static void
 queue_setup(struct rte_eth_dev *eth_dev, struct pmd_internal *internal)
 {
 	struct vhost_queue *vq;
+	int vid = internal->vid;
 	int i;
 
 	for (i = 0; i < eth_dev->data->nb_rx_queues; i++) {
 		vq = eth_dev->data->rx_queues[i];
 		if (!vq)
 			continue;
-		vq->vid = internal->vid;
+		vq->vid = vid;
 		vq->internal = internal;
 		vq->port = eth_dev->data->port_id;
 	}
+
 	for (i = 0; i < eth_dev->data->nb_tx_queues; i++) {
 		vq = eth_dev->data->tx_queues[i];
 		if (!vq)
 			continue;
-		vq->vid = internal->vid;
+		vq->vid = vid;
 		vq->internal = internal;
 		vq->port = eth_dev->data->port_id;
 	}
@@ -584,8 +692,9 @@ new_device(int vid)
 	struct rte_eth_dev *eth_dev;
 	struct internal_list *list;
 	struct pmd_internal *internal;
-	unsigned i;
+	struct rte_eth_conf *dev_conf;
 	char ifname[PATH_MAX];
+	int i;
 #ifdef RTE_LIBRTE_VHOST_NUMA
 	int newnode;
 #endif
@@ -599,6 +708,7 @@ new_device(int vid)
 
 	eth_dev = list->eth_dev;
 	internal = eth_dev->data->dev_private;
+	dev_conf = &eth_dev->data->dev_conf;
 
 #ifdef RTE_LIBRTE_VHOST_NUMA
 	newnode = rte_vhost_get_numa_node(vid);
@@ -609,14 +719,18 @@ new_device(int vid)
 	internal->vid = vid;
 	if (eth_dev->data->rx_queues && eth_dev->data->tx_queues) {
 		queue_setup(eth_dev, internal);
-		rte_atomic32_set(&internal->dev_attached, 1);
-	} else {
-		RTE_LOG(INFO, PMD, "RX/TX queues have not setup yet\n");
-		rte_atomic32_set(&internal->dev_attached, 0);
-	}
 
-	for (i = 0; i < rte_vhost_get_vring_num(vid); i++)
-		rte_vhost_enable_guest_notification(vid, i, 0);
+		for (i = 0; i < rte_vhost_get_vring_num(vid); i++)
+			rte_vhost_enable_guest_notification(vid, i, 0);
+
+		if (dev_conf->intr_conf.rxq) {
+			if (eth_vhost_install_intr(eth_dev) < 0) {
+				RTE_LOG(INFO, PMD,
+					"Failed to install interrupt handler.");
+					return -1;
+			}
+		}
+	}
 
 	rte_vhost_get_mtu(vid, &eth_dev->data->mtu);
 
@@ -626,6 +740,8 @@ new_device(int vid)
 
 	RTE_LOG(INFO, PMD, "Vhost device %d created\n", vid);
 
+	rte_atomic32_set(&internal->dev_attached, 1);
+
 	_rte_eth_dev_callback_process(eth_dev, RTE_ETH_EVENT_INTR_LSC, NULL);
 
 	return 0;
@@ -657,17 +773,19 @@ destroy_device(int vid)
 
 	eth_dev->data->dev_link.link_status = ETH_LINK_DOWN;
 
-	for (i = 0; i < eth_dev->data->nb_rx_queues; i++) {
-		vq = eth_dev->data->rx_queues[i];
-		if (vq == NULL)
-			continue;
-		vq->vid = -1;
-	}
-	for (i = 0; i < eth_dev->data->nb_tx_queues; i++) {
-		vq = eth_dev->data->tx_queues[i];
-		if (vq == NULL)
-			continue;
-		vq->vid = -1;
+	if (eth_dev->data->rx_queues && eth_dev->data->tx_queues) {
+		for (i = 0; i < eth_dev->data->nb_rx_queues; i++) {
+			vq = eth_dev->data->rx_queues[i];
+			if (!vq)
+				continue;
+			vq->vid = -1;
+		}
+		for (i = 0; i < eth_dev->data->nb_tx_queues; i++) {
+			vq = eth_dev->data->tx_queues[i];
+			if (!vq)
+				continue;
+			vq->vid = -1;
+		}
 	}
 
 	state = vring_states[eth_dev->data->port_id];
@@ -681,6 +799,8 @@ destroy_device(int vid)
 
 	RTE_LOG(INFO, PMD, "Vhost device %d destroyed\n", vid);
 
+	eth_vhost_uninstall_intr(eth_dev);
+
 	_rte_eth_dev_callback_process(eth_dev, RTE_ETH_EVENT_INTR_LSC, NULL);
 }
 
@@ -791,10 +911,30 @@ static int
 eth_dev_start(struct rte_eth_dev *eth_dev)
 {
 	struct pmd_internal *internal = eth_dev->data->dev_private;
+	struct rte_eth_conf *dev_conf;
+	int vid = internal->vid;
+	int i;
 
-	if (unlikely(rte_atomic32_read(&internal->dev_attached) == 0)) {
-		queue_setup(eth_dev, internal);
-		rte_atomic32_set(&internal->dev_attached, 1);
+	dev_conf = &eth_dev->data->dev_conf;
+
+	if (!eth_dev->data->rx_queues || !eth_dev->data->tx_queues) {
+		RTE_LOG(ERR, PMD, "RX/TX queues not setup yet\n");
+		return -1;
+	}
+
+	queue_setup(eth_dev, internal);
+
+	if (likely(rte_atomic32_read(&internal->dev_attached) == 1)) {
+		for (i = 0; i < rte_vhost_get_vring_num(vid); i++)
+			rte_vhost_enable_guest_notification(vid, i, 0);
+
+		if (dev_conf->intr_conf.rxq) {
+			if (eth_vhost_install_intr(eth_dev) < 0) {
+				RTE_LOG(INFO, PMD,
+					"Failed to install interrupt handler.");
+					return -1;
+			}
+		}
 	}
 
 	rte_atomic32_set(&internal->started, 1);
@@ -836,10 +976,13 @@ eth_dev_close(struct rte_eth_dev *dev)
 	pthread_mutex_unlock(&internal_list_lock);
 	rte_free(list);
 
-	for (i = 0; i < dev->data->nb_rx_queues; i++)
-		rte_free(dev->data->rx_queues[i]);
-	for (i = 0; i < dev->data->nb_tx_queues; i++)
-		rte_free(dev->data->tx_queues[i]);
+	if (dev->data->rx_queues)
+		for (i = 0; i < dev->data->nb_rx_queues; i++)
+			rte_free(dev->data->rx_queues[i]);
+
+	if (dev->data->tx_queues)
+		for (i = 0; i < dev->data->nb_tx_queues; i++)
+			rte_free(dev->data->tx_queues[i]);
 
 	rte_free(dev->data->mac_addrs);
 	free(internal->dev_name);
@@ -1030,6 +1173,8 @@ static const struct eth_dev_ops ops = {
 	.xstats_reset = vhost_dev_xstats_reset,
 	.xstats_get = vhost_dev_xstats_get,
 	.xstats_get_names = vhost_dev_xstats_get_names,
+	.rx_queue_intr_enable = eth_rxq_intr_enable,
+	.rx_queue_intr_disable = eth_rxq_intr_disable,
 };
 
 static struct rte_vdev_driver pmd_vhost_drv;
diff --git a/lib/librte_vhost/vhost.c b/lib/librte_vhost/vhost.c
index f6f12a0..68e9340 100644
--- a/lib/librte_vhost/vhost.c
+++ b/lib/librte_vhost/vhost.c
@@ -545,16 +545,15 @@ rte_vhost_enable_guest_notification(int vid, uint16_t queue_id, int enable)
 {
 	struct virtio_net *dev = get_device(vid);
 
-	if (dev == NULL)
+	if (!dev)
 		return -1;
 
-	if (enable) {
-		RTE_LOG(ERR, VHOST_CONFIG,
-			"guest notification isn't supported.\n");
-		return -1;
-	}
+	if (enable)
+		dev->virtqueue[queue_id]->used->flags &=
+		    ~VRING_USED_F_NO_NOTIFY;
+	else
+		dev->virtqueue[queue_id]->used->flags = VRING_USED_F_NO_NOTIFY;
 
-	dev->virtqueue[queue_id]->used->flags = VRING_USED_F_NO_NOTIFY;
 	return 0;
 }
 
-- 
2.0.1