On 1/7/21 7:32 AM, Xia, Chenbo wrote:

Hi Maxime,

quoted

-----Original Message-----
From: Maxime Coquelin <redacted>
Sent: Monday, December 21, 2020 5:14 AM
To: dev@dpdk.org; Xia, Chenbo <redacted>; olivier.matz@6wind.com;
amorenoz@redhat.com; david.marchand@redhat.com
Cc: Maxime Coquelin <redacted>
Subject: [PATCH 38/40] net/virtio: move Vhost-user specifics to its backend

This patch moves all the Vhost-user backend specific
logic like Vhost FD, listen FD and interrupt handling
to the vhost-user backend implementation.

In order to achieve that, new ops are created to update
the link status, disconnect and reconnect the server,
and fetch the link state interrupt FD.

Signed-off-by: Maxime Coquelin <redacted>
---
 drivers/net/virtio/virtio_user/vhost.h        |   4 +
 drivers/net/virtio/virtio_user/vhost_kernel.c |  18 +-
 drivers/net/virtio/virtio_user/vhost_user.c   | 169 ++++++++++++++---
 drivers/net/virtio/virtio_user/vhost_vdpa.c   |  16 ++
 .../net/virtio/virtio_user/virtio_user_dev.c  | 175 ++++++++++++++---
 .../net/virtio/virtio_user/virtio_user_dev.h  |   9 +-
 drivers/net/virtio/virtio_user_ethdev.c       | 179 +-----------------
 7 files changed, 340 insertions(+), 230 deletions(-)

diff --git a/drivers/net/virtio/virtio_user/vhost.h

b/drivers/net/virtio/virtio_user/vhost.h
index ee5598226d..6001d7a164 100644

--- a/drivers/net/virtio/virtio_user/vhost.h
+++ b/drivers/net/virtio/virtio_user/vhost.h

@@ -75,6 +75,10 @@ struct virtio_user_backend_ops {
 	int (*enable_qp)(struct virtio_user_dev *dev, uint16_t pair_idx, int

enable);
 	int (*dma_map)(struct virtio_user_dev *dev, void *addr, uint64_t iova,
size_t len);
 	int (*dma_unmap)(struct virtio_user_dev *dev, void *addr, uint64_t iova,
size_t len);
+	int (*update_link_state)(struct virtio_user_dev *dev);
+	int (*server_disconnect)(struct virtio_user_dev *dev);
+	int (*server_reconnect)(struct virtio_user_dev *dev);
+	int (*get_intr_fd)(struct virtio_user_dev *dev);

This get_intr_fd is for getting link state interrupt fd. I think his name is a little bit
generic. What do you think? Maybe we could change the name or add comment here to clarify here?

quoted

 };

 extern struct virtio_user_backend_ops virtio_ops_user;

diff --git a/drivers/net/virtio/virtio_user/vhost_kernel.c

b/drivers/net/virtio/virtio_user/vhost_kernel.c
index 023fddcd69..0ba37b23dc 100644

--- a/drivers/net/virtio/virtio_user/vhost_kernel.c
+++ b/drivers/net/virtio/virtio_user/vhost_kernel.c

@@ -459,6 +459,20 @@ vhost_kernel_get_backend_features(uint64_t *features)
 	return 0;
 }

<snip>

quoted

+static void
+virtio_user_dev_reset_queues_packed(struct rte_eth_dev *eth_dev)
+{
+	struct virtio_user_dev *dev = eth_dev->data->dev_private;
+	struct virtio_hw *hw = &dev->hw;
+	struct virtnet_rx *rxvq;
+	struct virtnet_tx *txvq;
+	uint16_t i;
+
+	/* Add lock to avoid queue contention. */
+	rte_spinlock_lock(&hw->state_lock);
+	hw->started = 0;
+
+	/*
+	 * Waitting for datapath to complete before resetting queues.

s/Waitting/Waiting

Fixed

quoted

+	 * 1 ms should be enough for the ongoing Tx/Rx function to finish.
+	 */
+	rte_delay_ms(1);
+
+	/* Vring reset for each Tx queue and Rx queue. */
+	for (i = 0; i < eth_dev->data->nb_rx_queues; i++) {
+		rxvq = eth_dev->data->rx_queues[i];
+		virtqueue_rxvq_reset_packed(rxvq->vq);
+		virtio_dev_rx_queue_setup_finish(eth_dev, i);
+	}
+
+	for (i = 0; i < eth_dev->data->nb_tx_queues; i++) {
+		txvq = eth_dev->data->tx_queues[i];
+		virtqueue_txvq_reset_packed(txvq->vq);
+	}
+
+	hw->started = 1;
+	rte_spinlock_unlock(&hw->state_lock);
+}
+
+void
+virtio_user_dev_delayed_handler(void *param)
+{
+	struct virtio_user_dev *dev = param;
+	struct rte_eth_dev *eth_dev = &rte_eth_devices[dev->port_id];
+
+	if (rte_intr_disable(eth_dev->intr_handle) < 0) {
+		PMD_DRV_LOG(ERR, "interrupt disable failed");
+		return;
+	}
+	rte_intr_callback_unregister(eth_dev->intr_handle,
+				     virtio_interrupt_handler, eth_dev);
+	if (dev->is_server) {
+		dev->ops->server_disconnect(dev);

Better to check ops existence too?

Indeed, v2 will check its existence before calling.

quoted

+		eth_dev->intr_handle->fd = dev->ops->get_intr_fd(dev);
+		rte_intr_callback_register(eth_dev->intr_handle,
+					   virtio_interrupt_handler, eth_dev);
+		if (rte_intr_enable(eth_dev->intr_handle) < 0) {
+			PMD_DRV_LOG(ERR, "interrupt enable failed");
+			return;
+		}
+	}
+}
+
+int
+virtio_user_dev_server_reconnect(struct virtio_user_dev *dev)
+{
+	int ret, old_status;
+	struct rte_eth_dev *eth_dev = &rte_eth_devices[dev->port_id];
+	struct virtio_hw *hw = &dev->hw;
+
+	if (!dev->ops->server_reconnect) {
+		PMD_DRV_LOG(ERR, "(%s) Missing server reconnect callback", dev-

quoted

path);

+		return -1;
+	}

Forget to call server_reconnect? :) 

Indeed! When I said I didn't tried last 5 patches, I wasn't lying! :)

quoted

+
+	old_status = virtio_get_status(hw);
+
+	virtio_reset(hw);
+
+	virtio_set_status(hw, VIRTIO_CONFIG_STATUS_ACK);
+
+	virtio_set_status(hw, VIRTIO_CONFIG_STATUS_DRIVER);
+
+	if (dev->ops->get_features(dev, &dev->device_features) < 0) {
+		PMD_INIT_LOG(ERR, "get_features failed: %s",
+			     strerror(errno));
+		return -1;
+	}
+
+	dev->device_features |= dev->frontend_features;
+
+	/* umask vhost-user unsupported features */

s/umask/unmask

Fixed.

Thanks,
Maxime

Thanks,
Chenbo