Re: [PATCH 4/4] virtio/vdev: add a new vdev named eth_cvio

[Pavel Fedin <p.fedin@samsung.com>]

 Hello!

 See inline

> -----Original Message-----
> From: Jianfeng Tan [mailto:jianfeng.tan@intel.com]
> Sent: Sunday, January 10, 2016 2:43 PM
> To: dev@dpdk.org
> Cc: rich.lane@bigswitch.com; yuanhan.liu@linux.intel.com; mst@redhat.com;
> nakajima.yoshihiro@lab.ntt.co.jp; huawei.xie@intel.com; mukawa@igel.co.jp;
> p.fedin@samsung.com; michael.qiu@intel.com; ann.zhuangyanying@huawei.com; Jianfeng Tan
> Subject: [PATCH 4/4] virtio/vdev: add a new vdev named eth_cvio
> 
> Add a new virtual device named eth_cvio, it can be used just like
> eth_ring, eth_null, etc.
> 
> Configured parameters include:
>     - rx (optional, 1 by default): number of rx, only allowed to be
> 				   1 for now.
>     - tx (optional, 1 by default): number of tx, only allowed to be
> 				   1 for now.
>     - cq (optional, 0 by default): if ctrl queue is enabled, not
> 				   supported for now.
>     - mac (optional): mac address, random value will be given if not
> 		      specified.
>     - queue_num (optional, 256 by default): size of virtqueue.
>     - path (madatory): path of vhost, depends on the file type:
> 		       vhost-user is used if the given path points to
> 	               a unix socket; vhost-net is used if the given
> 		       path points to a char device.
> 
> The major difference with original virtio for vm is that, here we
> use virtual address instead of physical address for vhost to
> calculate relative address.
> 
> When enable CONFIG_RTE_VIRTIO_VDEV (enabled by default), the compiled
> library can be used in both VM and container environment.
> 
> Examples:
> a. Use vhost-net as a backend
> sudo numactl -N 1 -m 1 ./examples/l2fwd/build/l2fwd -c 0x100000 -n 4 \
>     -m 1024 --no-pci --single-file --file-prefix=l2fwd \
>     --vdev=eth_cvio0,mac=00:01:02:03:04:05,path=/dev/vhost-net \
>     -- -p 0x1
> 
> b. Use vhost-user as a backend
> numactl -N 1 -m 1 ./examples/l2fwd/build/l2fwd -c 0x100000 -n 4 -m 1024 \
>     --no-pci --single-file --file-prefix=l2fwd \
>     --vdev=eth_cvio0,mac=00:01:02:03:04:05,path=<path_to_vhost_user> \
>     -- -p 0x1
> 
> Signed-off-by: Huawei Xie <huawei.xie@intel.com>
> Signed-off-by: Jianfeng Tan <jianfeng.tan@intel.com>
> ---
>  drivers/net/virtio/virtio_ethdev.c      | 338 +++++++++++++++++++++++++-------
>  drivers/net/virtio/virtio_ethdev.h      |   1 +
>  drivers/net/virtio/virtio_pci.h         |  24 +--
>  drivers/net/virtio/virtio_rxtx.c        |  11 +-
>  drivers/net/virtio/virtio_rxtx_simple.c |  14 +-
>  drivers/net/virtio/virtqueue.h          |  13 +-
>  6 files changed, 302 insertions(+), 99 deletions(-)
> 
> diff --git a/drivers/net/virtio/virtio_ethdev.c b/drivers/net/virtio/virtio_ethdev.c
> index d928339..6e46060 100644
> --- a/drivers/net/virtio/virtio_ethdev.c
> +++ b/drivers/net/virtio/virtio_ethdev.c
> @@ -56,6 +56,7 @@
>  #include <rte_memory.h>
>  #include <rte_eal.h>
>  #include <rte_dev.h>
> +#include <rte_kvargs.h>
> 
>  #include "virtio_ethdev.h"
>  #include "virtio_pci.h"
> @@ -174,14 +175,14 @@ virtio_send_command(struct virtqueue *vq, struct virtio_pmd_ctrl *ctrl,
>  	 * One RX packet for ACK.
>  	 */
>  	vq->vq_ring.desc[head].flags = VRING_DESC_F_NEXT;
> -	vq->vq_ring.desc[head].addr = vq->virtio_net_hdr_mz->phys_addr;
> +	vq->vq_ring.desc[head].addr = vq->virtio_net_hdr_mem;
>  	vq->vq_ring.desc[head].len = sizeof(struct virtio_net_ctrl_hdr);
>  	vq->vq_free_cnt--;
>  	i = vq->vq_ring.desc[head].next;
> 
>  	for (k = 0; k < pkt_num; k++) {
>  		vq->vq_ring.desc[i].flags = VRING_DESC_F_NEXT;
> -		vq->vq_ring.desc[i].addr = vq->virtio_net_hdr_mz->phys_addr
> +		vq->vq_ring.desc[i].addr = vq->virtio_net_hdr_mem
>  			+ sizeof(struct virtio_net_ctrl_hdr)
>  			+ sizeof(ctrl->status) + sizeof(uint8_t)*sum;
>  		vq->vq_ring.desc[i].len = dlen[k];
> @@ -191,7 +192,7 @@ virtio_send_command(struct virtqueue *vq, struct virtio_pmd_ctrl *ctrl,
>  	}
> 
>  	vq->vq_ring.desc[i].flags = VRING_DESC_F_WRITE;
> -	vq->vq_ring.desc[i].addr = vq->virtio_net_hdr_mz->phys_addr
> +	vq->vq_ring.desc[i].addr = vq->virtio_net_hdr_mem
>  			+ sizeof(struct virtio_net_ctrl_hdr);
>  	vq->vq_ring.desc[i].len = sizeof(ctrl->status);
>  	vq->vq_free_cnt--;
> @@ -374,68 +375,85 @@ int virtio_dev_queue_setup(struct rte_eth_dev *dev,
>  		}
>  	}
> 
> -	/*
> -	 * Virtio PCI device VIRTIO_PCI_QUEUE_PF register is 32bit,
> -	 * and only accepts 32 bit page frame number.
> -	 * Check if the allocated physical memory exceeds 16TB.
> -	 */
> -	if ((mz->phys_addr + vq->vq_ring_size - 1) >> (VIRTIO_PCI_QUEUE_ADDR_SHIFT + 32)) {
> -		PMD_INIT_LOG(ERR, "vring address shouldn't be above 16TB!");
> -		rte_free(vq);
> -		return -ENOMEM;
> -	}
> -
>  	memset(mz->addr, 0, sizeof(mz->len));
>  	vq->mz = mz;
> -	vq->vq_ring_mem = mz->phys_addr;
>  	vq->vq_ring_virt_mem = mz->addr;
> -	PMD_INIT_LOG(DEBUG, "vq->vq_ring_mem:      0x%"PRIx64, (uint64_t)mz->phys_addr);
> -	PMD_INIT_LOG(DEBUG, "vq->vq_ring_virt_mem: 0x%"PRIx64, (uint64_t)(uintptr_t)mz->addr);
> +
> +	if (dev->dev_type == RTE_ETH_DEV_PCI) {
> +		vq->vq_ring_mem = mz->phys_addr;
> +
> +		/* Virtio PCI device VIRTIO_PCI_QUEUE_PF register is 32bit,
> +		 * and only accepts 32 bit page frame number.
> +		 * Check if the allocated physical memory exceeds 16TB.
> +		 */
> +		uint64_t last_physaddr = vq->vq_ring_mem + vq->vq_ring_size - 1;
> +		if (last_physaddr >> (VIRTIO_PCI_QUEUE_ADDR_SHIFT + 32)) {
> +			PMD_INIT_LOG(ERR, "vring address shouldn't be above 16TB!");
> +			rte_free(vq);
> +			return -ENOMEM;
> +		}
> +	}
> +#ifdef RTE_VIRTIO_VDEV
> +	else
> +		vq->vq_ring_mem = (phys_addr_t)mz->addr; /* Use vaddr!!! */
> +#endif
> +
> +	PMD_INIT_LOG(DEBUG, "vq->vq_ring_mem:      0x%"PRIx64,
> +			(uint64_t)vq->vq_ring_mem);
> +	PMD_INIT_LOG(DEBUG, "vq->vq_ring_virt_mem: 0x%"PRIx64,
> +			(uint64_t)(uintptr_t)vq->vq_ring_virt_mem);
>  	vq->virtio_net_hdr_mz  = NULL;
>  	vq->virtio_net_hdr_mem = 0;
> 
> +	uint64_t hdr_size = 0;
>  	if (queue_type == VTNET_TQ) {
>  		/*
>  		 * For each xmit packet, allocate a virtio_net_hdr
>  		 */
>  		snprintf(vq_name, sizeof(vq_name), "port%d_tvq%d_hdrzone",
>  			dev->data->port_id, queue_idx);
> -		vq->virtio_net_hdr_mz = rte_memzone_reserve_aligned(vq_name,
> -			vq_size * hw->vtnet_hdr_size,
> -			socket_id, 0, RTE_CACHE_LINE_SIZE);
> -		if (vq->virtio_net_hdr_mz == NULL) {
> -			if (rte_errno == EEXIST)
> -				vq->virtio_net_hdr_mz =
> -					rte_memzone_lookup(vq_name);
> -			if (vq->virtio_net_hdr_mz == NULL) {
> -				rte_free(vq);
> -				return -ENOMEM;
> -			}
> -		}
> -		vq->virtio_net_hdr_mem =
> -			vq->virtio_net_hdr_mz->phys_addr;
> -		memset(vq->virtio_net_hdr_mz->addr, 0,
> -			vq_size * hw->vtnet_hdr_size);
> +		hdr_size = vq_size * hw->vtnet_hdr_size;
>  	} else if (queue_type == VTNET_CQ) {
>  		/* Allocate a page for control vq command, data and status */
>  		snprintf(vq_name, sizeof(vq_name), "port%d_cvq_hdrzone",
>  			dev->data->port_id);
> -		vq->virtio_net_hdr_mz = rte_memzone_reserve_aligned(vq_name,
> -			PAGE_SIZE, socket_id, 0, RTE_CACHE_LINE_SIZE);
> -		if (vq->virtio_net_hdr_mz == NULL) {
> +		hdr_size = PAGE_SIZE;
> +	}
> +
> +	if (hdr_size) { /* queue_type is VTNET_TQ or VTNET_CQ */
> +		mz = rte_memzone_reserve_aligned(vq_name,
> +				hdr_size, socket_id, 0, RTE_CACHE_LINE_SIZE);
> +		if (mz == NULL) {
>  			if (rte_errno == EEXIST)
> -				vq->virtio_net_hdr_mz =
> -					rte_memzone_lookup(vq_name);
> -			if (vq->virtio_net_hdr_mz == NULL) {
> +				mz = rte_memzone_lookup(vq_name);
> +			if (mz == NULL) {
>  				rte_free(vq);
>  				return -ENOMEM;
>  			}
>  		}
> -		vq->virtio_net_hdr_mem =
> -			vq->virtio_net_hdr_mz->phys_addr;
> -		memset(vq->virtio_net_hdr_mz->addr, 0, PAGE_SIZE);
> +		vq->virtio_net_hdr_mz = mz;
> +		vq->virtio_net_hdr_vaddr = mz->addr;
> +		memset(vq->virtio_net_hdr_vaddr, 0, hdr_size);
> +
> +		if (dev->dev_type == RTE_ETH_DEV_PCI)
> +			vq->virtio_net_hdr_mem = mz->phys_addr;
> +#ifdef RTE_VIRTIO_VDEV
> +		else
> +			vq->virtio_net_hdr_mem = (phys_addr_t)mz->addr; /* Use vaddr!!! */
> +#endif
>  	}
> 
> +	struct rte_mbuf *m = NULL;
> +	if (dev->dev_type == RTE_ETH_DEV_PCI)
> +		vq->offset = (uintptr_t)&m->buf_addr;
> +#ifdef RTE_VIRTIO_VDEV
> +	else {
> +		vq->offset = (uintptr_t)&m->buf_physaddr;

 Not sure, but shouldn't these be swapped? Originally, for PCI devices, we used buf_physaddr.

> +#if (RTE_BYTE_ORDER == RTE_BIG_ENDIAN) && (__WORDSIZE == 32)
> +		vq->offset += 4;
> +#endif
> +	}
> +#endif
>  	/*
>  	 * Set guest physical address of the virtqueue
>  	 * in VIRTIO_PCI_QUEUE_PFN config register of device
> @@ -491,8 +509,10 @@ virtio_dev_close(struct rte_eth_dev *dev)
>  	PMD_INIT_LOG(DEBUG, "virtio_dev_close");
> 
>  	/* reset the NIC */
> -	if (pci_dev->driver->drv_flags & RTE_PCI_DRV_INTR_LSC)
> -		vtpci_irq_config(hw, VIRTIO_MSI_NO_VECTOR);
> +	if (dev->dev_type == RTE_ETH_DEV_PCI) {
> +		if (pci_dev->driver->drv_flags & RTE_PCI_DRV_INTR_LSC)
> +			vtpci_irq_config(hw, VIRTIO_MSI_NO_VECTOR);
> +	}
>  	vtpci_reset(hw);
>  	hw->started = 0;
>  	virtio_dev_free_mbufs(dev);
> @@ -1233,8 +1253,9 @@ virtio_interrupt_handler(__rte_unused struct rte_intr_handle *handle,
>  	isr = vtpci_isr(hw);
>  	PMD_DRV_LOG(INFO, "interrupt status = %#x", isr);
> 
> -	if (rte_intr_enable(&dev->pci_dev->intr_handle) < 0)
> -		PMD_DRV_LOG(ERR, "interrupt enable failed");
> +	if (dev->dev_type == RTE_ETH_DEV_PCI)
> +		if (rte_intr_enable(&dev->pci_dev->intr_handle) < 0)
> +			PMD_DRV_LOG(ERR, "interrupt enable failed");
> 
>  	if (isr & VIRTIO_PCI_ISR_CONFIG) {
>  		if (virtio_dev_link_update(dev, 0) == 0)
> @@ -1287,11 +1308,18 @@ eth_virtio_dev_init(struct rte_eth_dev *eth_dev)
> 
>  	pci_dev = eth_dev->pci_dev;
> 
> -	if (virtio_resource_init(pci_dev) < 0)
> -		return -1;
> -
> -	hw->use_msix = virtio_has_msix(&pci_dev->addr);
> -	hw->io_base = (uint32_t)(uintptr_t)pci_dev->mem_resource[0].addr;
> +	if (eth_dev->dev_type == RTE_ETH_DEV_PCI) {
> +		if (virtio_resource_init(pci_dev) < 0)
> +			return -1;
> +		hw->use_msix = virtio_has_msix(&pci_dev->addr);
> +		hw->io_base = (uint32_t)(uintptr_t)pci_dev->mem_resource[0].addr;
> +	}
> +#ifdef RTE_VIRTIO_VDEV
> +	else {
> +		hw->use_msix = 0;
> +		hw->io_base = VIRTIO_VDEV_IO_BASE;
> +	}
> +#endif
> 
>  	/* Reset the device although not necessary at startup */
>  	vtpci_reset(hw);
> @@ -1304,10 +1332,12 @@ eth_virtio_dev_init(struct rte_eth_dev *eth_dev)
>  	virtio_negotiate_features(hw);
> 
>  	/* If host does not support status then disable LSC */
> -	if (!vtpci_with_feature(hw, VIRTIO_NET_F_STATUS))
> -		pci_dev->driver->drv_flags &= ~RTE_PCI_DRV_INTR_LSC;
> +	if (eth_dev->dev_type == RTE_ETH_DEV_PCI) {
> +		if (!vtpci_with_feature(hw, VIRTIO_NET_F_STATUS))
> +			pci_dev->driver->drv_flags &= ~RTE_PCI_DRV_INTR_LSC;
> 
> -	rte_eth_copy_pci_info(eth_dev, pci_dev);
> +		rte_eth_copy_pci_info(eth_dev, pci_dev);
> +	}
> 
>  	rx_func_get(eth_dev);
> 
> @@ -1383,15 +1413,16 @@ eth_virtio_dev_init(struct rte_eth_dev *eth_dev)
> 
>  	PMD_INIT_LOG(DEBUG, "hw->max_rx_queues=%d   hw->max_tx_queues=%d",
>  			hw->max_rx_queues, hw->max_tx_queues);
> -	PMD_INIT_LOG(DEBUG, "port %d vendorID=0x%x deviceID=0x%x",
> -			eth_dev->data->port_id, pci_dev->id.vendor_id,
> -			pci_dev->id.device_id);
> -
> -	/* Setup interrupt callback  */
> -	if (pci_dev->driver->drv_flags & RTE_PCI_DRV_INTR_LSC)
> -		rte_intr_callback_register(&pci_dev->intr_handle,
> -				   virtio_interrupt_handler, eth_dev);
> -
> +	if (eth_dev->dev_type == RTE_ETH_DEV_PCI) {
> +		PMD_INIT_LOG(DEBUG, "port %d vendorID=0x%x deviceID=0x%x",
> +				eth_dev->data->port_id, pci_dev->id.vendor_id,
> +				pci_dev->id.device_id);
> +
> +		/* Setup interrupt callback  */
> +		if (pci_dev->driver->drv_flags & RTE_PCI_DRV_INTR_LSC)
> +			rte_intr_callback_register(&pci_dev->intr_handle,
> +					virtio_interrupt_handler, eth_dev);
> +	}
>  	virtio_dev_cq_start(eth_dev);
> 
>  	return 0;
> @@ -1424,10 +1455,12 @@ eth_virtio_dev_uninit(struct rte_eth_dev *eth_dev)
>  	eth_dev->data->mac_addrs = NULL;
> 
>  	/* reset interrupt callback  */
> -	if (pci_dev->driver->drv_flags & RTE_PCI_DRV_INTR_LSC)
> -		rte_intr_callback_unregister(&pci_dev->intr_handle,
> -						virtio_interrupt_handler,
> -						eth_dev);
> +	if (eth_dev->dev_type == RTE_ETH_DEV_PCI) {
> +		if (pci_dev->driver->drv_flags & RTE_PCI_DRV_INTR_LSC)
> +			rte_intr_callback_unregister(&pci_dev->intr_handle,
> +					virtio_interrupt_handler,
> +					eth_dev);
> +	}
> 
>  	PMD_INIT_LOG(DEBUG, "dev_uninit completed");
> 
> @@ -1491,11 +1524,13 @@ virtio_dev_configure(struct rte_eth_dev *dev)
>  		return -ENOTSUP;
>  	}
> 
> -	if (pci_dev->driver->drv_flags & RTE_PCI_DRV_INTR_LSC)
> -		if (vtpci_irq_config(hw, 0) == VIRTIO_MSI_NO_VECTOR) {
> -			PMD_DRV_LOG(ERR, "failed to set config vector");
> -			return -EBUSY;
> -		}
> +	if (dev->dev_type == RTE_ETH_DEV_PCI) {
> +		if (pci_dev->driver->drv_flags & RTE_PCI_DRV_INTR_LSC)
> +			if (vtpci_irq_config(hw, 0) == VIRTIO_MSI_NO_VECTOR) {
> +				PMD_DRV_LOG(ERR, "failed to set config vector");
> +				return -EBUSY;
> +			}
> +	}
> 
>  	return 0;
>  }
> @@ -1689,3 +1724,162 @@ static struct rte_driver rte_virtio_driver = {
>  };
> 
>  PMD_REGISTER_DRIVER(rte_virtio_driver);
> +
> +#ifdef RTE_VIRTIO_VDEV
> +
> +static const char *valid_args[] = {
> +#define ETH_CVIO_ARG_RX_NUM		"rx"
> +	ETH_CVIO_ARG_RX_NUM,
> +#define ETH_CVIO_ARG_TX_NUM		"tx"
> +	ETH_CVIO_ARG_TX_NUM,
> +#define ETH_CVIO_ARG_CQ_NUM		"cq"
> +	ETH_CVIO_ARG_CQ_NUM,
> +#define ETH_CVIO_ARG_MAC		"mac"
> +	ETH_CVIO_ARG_MAC,
> +#define ETH_CVIO_ARG_PATH		"path"
> +	ETH_CVIO_ARG_PATH,
> +#define ETH_CVIO_ARG_QUEUE_SIZE	"queue_num"
> +	ETH_CVIO_ARG_QUEUE_SIZE,
> +	NULL
> +};
> +
> +static int
> +get_string_arg(const char *key __rte_unused,
> +		const char *value, void *extra_args)
> +{
> +	if ((value == NULL) || (extra_args == NULL))
> +		return -EINVAL;
> +
> +	strcpy(extra_args, value);
> +
> +	return 0;
> +}
> +
> +static int
> +get_integer_arg(const char *key __rte_unused,
> +		const char *value, void *extra_args)
> +{
> +	uint64_t *p_u64 = extra_args;
> +
> +	if ((value == NULL) || (extra_args == NULL))
> +		return -EINVAL;
> +
> +	*p_u64 = (uint64_t)strtoull(value, NULL, 0);
> +
> +	return 0;
> +}
> +
> +static struct rte_eth_dev *
> +cvio_eth_dev_alloc(const char *name)
> +{
> +	struct rte_eth_dev *eth_dev;
> +	struct rte_eth_dev_data *data;
> +	struct virtio_hw *hw;
> +
> +	eth_dev = rte_eth_dev_allocate(name, RTE_ETH_DEV_VIRTUAL);
> +	if (eth_dev == NULL)
> +		rte_panic("cannot alloc rte_eth_dev\n");
> +
> +	data = eth_dev->data;
> +
> +	hw = rte_zmalloc(NULL, sizeof(*hw), 0);
> +	if (!hw)
> +		rte_panic("malloc virtio_hw failed\n");
> +
> +	data->dev_private = hw;
> +	data->numa_node = SOCKET_ID_ANY;
> +	eth_dev->pci_dev = NULL;
> +	/* will be used in virtio_dev_info_get() */
> +	eth_dev->driver = &rte_virtio_pmd;
> +	/* TODO: eth_dev->link_intr_cbs */
> +	return eth_dev;
> +}
> +
> +#define CVIO_DEF_CQ_EN		0
> +#define CVIO_DEF_Q_NUM		1
> +#define CVIO_DEF_Q_SZ		256
> +/*
> + * Dev initialization routine. Invoked once for each virtio vdev at
> + * EAL init time, see rte_eal_dev_init().
> + * Returns 0 on success.
> + */
> +static int
> +rte_cvio_pmd_devinit(const char *name, const char *params)
> +{
> +	struct rte_kvargs *kvlist = NULL;
> +	struct rte_eth_dev *eth_dev = NULL;
> +	uint64_t nb_rx = CVIO_DEF_Q_NUM;
> +	uint64_t nb_tx = CVIO_DEF_Q_NUM;
> +	uint64_t nb_cq = CVIO_DEF_CQ_EN;
> +	uint64_t queue_num = CVIO_DEF_Q_SZ;
> +	char sock_path[256];
> +	char mac_addr[32];
> +	int flag_mac = 0;
> +
> +	if (params == NULL || params[0] == '\0')
> +		rte_panic("arg %s is mandatory for eth_cvio\n",
> +				ETH_CVIO_ARG_QUEUE_SIZE);
> +
> +	kvlist = rte_kvargs_parse(params, valid_args);
> +	if (!kvlist)
> +		rte_panic("error when parsing param\n");
> +
> +	if (rte_kvargs_count(kvlist, ETH_CVIO_ARG_PATH) == 1)
> +		rte_kvargs_process(kvlist, ETH_CVIO_ARG_PATH,
> +				&get_string_arg, sock_path);
> +	else
> +		rte_panic("arg %s is mandatory for eth_cvio\n",
> +				ETH_CVIO_ARG_QUEUE_SIZE);
> +
> +	if (rte_kvargs_count(kvlist, ETH_CVIO_ARG_MAC) == 1) {
> +		rte_kvargs_process(kvlist, ETH_CVIO_ARG_MAC,
> +				&get_string_arg, mac_addr);
> +		flag_mac = 1;
> +	}
> +
> +	if (rte_kvargs_count(kvlist, ETH_CVIO_ARG_QUEUE_SIZE) == 1)
> +		rte_kvargs_process(kvlist, ETH_CVIO_ARG_QUEUE_SIZE,
> +			&get_integer_arg, &queue_num);
> +
> +	if (rte_kvargs_count(kvlist, ETH_CVIO_ARG_RX_NUM) == 1)
> +		rte_kvargs_process(kvlist, ETH_CVIO_ARG_RX_NUM,
> +			&get_integer_arg, &nb_rx);
> +
> +	if (rte_kvargs_count(kvlist, ETH_CVIO_ARG_TX_NUM) == 1)
> +		rte_kvargs_process(kvlist, ETH_CVIO_ARG_TX_NUM,
> +			&get_integer_arg, &nb_tx);
> +
> +	if (rte_kvargs_count(kvlist, ETH_CVIO_ARG_CQ_NUM) == 1)
> +		rte_kvargs_process(kvlist, ETH_CVIO_ARG_CQ_NUM,
> +			&get_integer_arg, &nb_cq);
> +
> +	eth_dev = cvio_eth_dev_alloc(name);
> +
> +	virtio_vdev_init(eth_dev->data, sock_path,
> +			nb_rx, nb_tx, nb_cq, queue_num,
> +			(flag_mac) ? mac_addr : NULL);
> +
> +	/* originally, this will be called in rte_eal_pci_probe() */
> +	eth_virtio_dev_init(eth_dev);
> +
> +	return 0;
> +}
> +
> +static int
> +rte_cvio_pmd_devuninit(const char *name)
> +{
> +	/* TODO: if it's last one, memory init, free memory */
> +	rte_panic("%s: %s", __func__, name);
> +	return 0;
> +}
> +
> +static struct rte_driver rte_cvio_driver = {
> +	.name   = "eth_cvio",
> +	.type   = PMD_VDEV,
> +	.init   = rte_cvio_pmd_devinit,
> +	.uninit = rte_cvio_pmd_devuninit,
> +};
> +
> +PMD_REGISTER_DRIVER(rte_cvio_driver);
> +
> +#endif
> diff --git a/drivers/net/virtio/virtio_ethdev.h b/drivers/net/virtio/virtio_ethdev.h
> index 9e1ecb3..90890b4 100644
> --- a/drivers/net/virtio/virtio_ethdev.h
> +++ b/drivers/net/virtio/virtio_ethdev.h
> @@ -126,4 +126,5 @@ uint16_t virtio_xmit_pkts_simple(void *tx_queue, struct rte_mbuf
> **tx_pkts,
>  void virtio_vdev_init(struct rte_eth_dev_data *data, const char *path,
>  		int nb_rx, int nb_tx, int nb_cq, int queue_num, char *mac);
>  #endif
> +
>  #endif /* _VIRTIO_ETHDEV_H_ */
> diff --git a/drivers/net/virtio/virtio_pci.h b/drivers/net/virtio/virtio_pci.h
> index af05ae2..d79bd05 100644
> --- a/drivers/net/virtio/virtio_pci.h
> +++ b/drivers/net/virtio/virtio_pci.h
> @@ -249,31 +249,31 @@ uint32_t virtio_ioport_read(struct virtio_hw *, uint64_t);
>  void virtio_ioport_write(struct virtio_hw *, uint64_t, uint32_t);
> 
>  #define VIRTIO_READ_REG_1(hw, reg) \
> -	(hw->io_base != VIRTIO_VDEV_IO_BASE) ? \
> +	((hw->io_base != VIRTIO_VDEV_IO_BASE) ? \
>  	inb((VIRTIO_PCI_REG_ADDR((hw), (reg)))) \
> -	:virtio_ioport_read(hw, reg)
> +	:virtio_ioport_read(hw, reg))
>  #define VIRTIO_WRITE_REG_1(hw, reg, value) \
> -	(hw->io_base != VIRTIO_VDEV_IO_BASE) ? \
> +	((hw->io_base != VIRTIO_VDEV_IO_BASE) ? \
>  	outb_p((unsigned char)(value), (VIRTIO_PCI_REG_ADDR((hw), (reg)))) \
> -	:virtio_ioport_write(hw, reg, value)
> +	:virtio_ioport_write(hw, reg, value))
> 
>  #define VIRTIO_READ_REG_2(hw, reg) \
> -	(hw->io_base != VIRTIO_VDEV_IO_BASE) ? \
> +	((hw->io_base != VIRTIO_VDEV_IO_BASE) ? \
>  	inw((VIRTIO_PCI_REG_ADDR((hw), (reg)))) \
> -	:virtio_ioport_read(hw, reg)
> +	:virtio_ioport_read(hw, reg))
>  #define VIRTIO_WRITE_REG_2(hw, reg, value) \
> -	(hw->io_base != VIRTIO_VDEV_IO_BASE) ? \
> +	((hw->io_base != VIRTIO_VDEV_IO_BASE) ? \
>  	outw_p((unsigned short)(value), (VIRTIO_PCI_REG_ADDR((hw), (reg)))) \
> -	:virtio_ioport_write(hw, reg, value)
> +	:virtio_ioport_write(hw, reg, value))
> 
>  #define VIRTIO_READ_REG_4(hw, reg) \
> -	(hw->io_base != VIRTIO_VDEV_IO_BASE) ? \
> +	((hw->io_base != VIRTIO_VDEV_IO_BASE) ? \
>  	inl((VIRTIO_PCI_REG_ADDR((hw), (reg)))) \
> -	:virtio_ioport_read(hw, reg)
> +	:virtio_ioport_read(hw, reg))
>  #define VIRTIO_WRITE_REG_4(hw, reg, value) \
> -	(hw->io_base != VIRTIO_VDEV_IO_BASE) ? \
> +	((hw->io_base != VIRTIO_VDEV_IO_BASE) ? \
>  	outl_p((unsigned int)(value), (VIRTIO_PCI_REG_ADDR((hw), (reg)))) \
> -	:virtio_ioport_write(hw, reg, value)
> +	:virtio_ioport_write(hw, reg, value))

 These bracket fixups should be squashed into #3

> 
>  #else /* RTE_VIRTIO_VDEV */
> 
> diff --git a/drivers/net/virtio/virtio_rxtx.c b/drivers/net/virtio/virtio_rxtx.c
> index 74b39ef..dd07ba7 100644
> --- a/drivers/net/virtio/virtio_rxtx.c
> +++ b/drivers/net/virtio/virtio_rxtx.c
> @@ -191,8 +191,7 @@ virtqueue_enqueue_recv_refill(struct virtqueue *vq, struct rte_mbuf
> *cookie)
> 
>  	start_dp = vq->vq_ring.desc;
>  	start_dp[idx].addr =
> -		(uint64_t)(cookie->buf_physaddr + RTE_PKTMBUF_HEADROOM
> -		- hw->vtnet_hdr_size);
> +		RTE_MBUF_DATA_DMA_ADDR(cookie, vq->offset) - hw->vtnet_hdr_size;
>  	start_dp[idx].len =
>  		cookie->buf_len - RTE_PKTMBUF_HEADROOM + hw->vtnet_hdr_size;
>  	start_dp[idx].flags =  VRING_DESC_F_WRITE;
> @@ -237,7 +236,7 @@ virtqueue_enqueue_xmit(struct virtqueue *txvq, struct rte_mbuf *cookie)
> 
>  	for (; ((seg_num > 0) && (cookie != NULL)); seg_num--) {
>  		idx = start_dp[idx].next;
> -		start_dp[idx].addr  = RTE_MBUF_DATA_DMA_ADDR(cookie);
> +		start_dp[idx].addr  = RTE_MBUF_DATA_DMA_ADDR(cookie, txvq->offset);
>  		start_dp[idx].len   = cookie->data_len;
>  		start_dp[idx].flags = VRING_DESC_F_NEXT;
>  		cookie = cookie->next;
> @@ -343,7 +342,7 @@ virtio_dev_vring_start(struct virtqueue *vq, int queue_type)
>  		VIRTIO_WRITE_REG_2(vq->hw, VIRTIO_PCI_QUEUE_SEL,
>  			vq->vq_queue_index);
>  		VIRTIO_WRITE_REG_4(vq->hw, VIRTIO_PCI_QUEUE_PFN,
> -			vq->mz->phys_addr >> VIRTIO_PCI_QUEUE_ADDR_SHIFT);
> +			vq->vq_ring_mem >> VIRTIO_PCI_QUEUE_ADDR_SHIFT);
>  	} else if (queue_type == VTNET_TQ) {
>  		if (use_simple_rxtx) {
>  			int mid_idx  = vq->vq_nentries >> 1;
> @@ -366,12 +365,12 @@ virtio_dev_vring_start(struct virtqueue *vq, int queue_type)
>  		VIRTIO_WRITE_REG_2(vq->hw, VIRTIO_PCI_QUEUE_SEL,
>  			vq->vq_queue_index);
>  		VIRTIO_WRITE_REG_4(vq->hw, VIRTIO_PCI_QUEUE_PFN,
> -			vq->mz->phys_addr >> VIRTIO_PCI_QUEUE_ADDR_SHIFT);
> +			vq->vq_ring_mem >> VIRTIO_PCI_QUEUE_ADDR_SHIFT);
>  	} else {
>  		VIRTIO_WRITE_REG_2(vq->hw, VIRTIO_PCI_QUEUE_SEL,
>  			vq->vq_queue_index);
>  		VIRTIO_WRITE_REG_4(vq->hw, VIRTIO_PCI_QUEUE_PFN,
> -			vq->mz->phys_addr >> VIRTIO_PCI_QUEUE_ADDR_SHIFT);
> +			vq->vq_ring_mem >> VIRTIO_PCI_QUEUE_ADDR_SHIFT);
>  	}
>  }
> 
> diff --git a/drivers/net/virtio/virtio_rxtx_simple.c
> b/drivers/net/virtio/virtio_rxtx_simple.c
> index ff3c11a..3a14a4e 100644
> --- a/drivers/net/virtio/virtio_rxtx_simple.c
> +++ b/drivers/net/virtio/virtio_rxtx_simple.c
> @@ -80,8 +80,8 @@ virtqueue_enqueue_recv_refill_simple(struct virtqueue *vq,
>  	vq->sw_ring[desc_idx] = cookie;
> 
>  	start_dp = vq->vq_ring.desc;
> -	start_dp[desc_idx].addr = (uint64_t)((uintptr_t)cookie->buf_physaddr +
> -		RTE_PKTMBUF_HEADROOM - sizeof(struct virtio_net_hdr));
> +	start_dp[desc_idx].addr = RTE_MBUF_DATA_DMA_ADDR(cookie, vq->offset)
> +		- sizeof(struct virtio_net_hdr);
>  	start_dp[desc_idx].len = cookie->buf_len -
>  		RTE_PKTMBUF_HEADROOM + sizeof(struct virtio_net_hdr);
> 
> @@ -118,9 +118,8 @@ virtio_rxq_rearm_vec(struct virtqueue *rxvq)
>  		p = (uintptr_t)&sw_ring[i]->rearm_data;
>  		*(uint64_t *)p = rxvq->mbuf_initializer;
> 
> -		start_dp[i].addr =
> -			(uint64_t)((uintptr_t)sw_ring[i]->buf_physaddr +
> -			RTE_PKTMBUF_HEADROOM - sizeof(struct virtio_net_hdr));
> +		start_dp[i].addr = RTE_MBUF_DATA_DMA_ADDR(sw_ring[i], rxvq->offset)
> +			- sizeof(struct virtio_net_hdr);
>  		start_dp[i].len = sw_ring[i]->buf_len -
>  			RTE_PKTMBUF_HEADROOM + sizeof(struct virtio_net_hdr);
>  	}
> @@ -366,7 +365,7 @@ virtio_xmit_pkts_simple(void *tx_queue, struct rte_mbuf **tx_pkts,
>  			txvq->vq_descx[desc_idx + i].cookie = tx_pkts[i];
>  		for (i = 0; i < nb_tail; i++) {
>  			start_dp[desc_idx].addr =
> -				RTE_MBUF_DATA_DMA_ADDR(*tx_pkts);
> +				RTE_MBUF_DATA_DMA_ADDR(*tx_pkts, txvq->offset);
>  			start_dp[desc_idx].len = (*tx_pkts)->pkt_len;
>  			tx_pkts++;
>  			desc_idx++;
> @@ -377,7 +376,8 @@ virtio_xmit_pkts_simple(void *tx_queue, struct rte_mbuf **tx_pkts,
>  	for (i = 0; i < nb_commit; i++)
>  		txvq->vq_descx[desc_idx + i].cookie = tx_pkts[i];
>  	for (i = 0; i < nb_commit; i++) {
> -		start_dp[desc_idx].addr = RTE_MBUF_DATA_DMA_ADDR(*tx_pkts);
> +		start_dp[desc_idx].addr = RTE_MBUF_DATA_DMA_ADDR(*tx_pkts,
> +				txvq->offset);
>  		start_dp[desc_idx].len = (*tx_pkts)->pkt_len;
>  		tx_pkts++;
>  		desc_idx++;
> diff --git a/drivers/net/virtio/virtqueue.h b/drivers/net/virtio/virtqueue.h
> index 61b3137..dc0b656 100644
> --- a/drivers/net/virtio/virtqueue.h
> +++ b/drivers/net/virtio/virtqueue.h
> @@ -66,8 +66,14 @@ struct rte_mbuf;
> 
>  #define VIRTQUEUE_MAX_NAME_SZ 32
> 
> -#define RTE_MBUF_DATA_DMA_ADDR(mb) \
> +#ifdef RTE_VIRTIO_VDEV
> +#define RTE_MBUF_DATA_DMA_ADDR(mb, offset) \
> +	(uint64_t)((uintptr_t)(*(void **)((uintptr_t)mb + offset)) \
> +			+ (mb)->data_off)
> +#else
> +#define RTE_MBUF_DATA_DMA_ADDR(mb, offset) \
>  	(uint64_t) ((mb)->buf_physaddr + (mb)->data_off)
> +#endif /* RTE_VIRTIO_VDEV */
> 
>  #define VTNET_SQ_RQ_QUEUE_IDX 0
>  #define VTNET_SQ_TQ_QUEUE_IDX 1
> @@ -167,7 +173,8 @@ struct virtqueue {
> 
>  	void        *vq_ring_virt_mem;    /**< linear address of vring*/
>  	unsigned int vq_ring_size;
> -	phys_addr_t vq_ring_mem;          /**< physical address of vring */
> +	phys_addr_t vq_ring_mem;          /**< phys address of vring for pci dev,
> +										virt addr of vring for vdev */
> 
>  	struct vring vq_ring;    /**< vring keeping desc, used and avail */
>  	uint16_t    vq_free_cnt; /**< num of desc available */
> @@ -186,8 +193,10 @@ struct virtqueue {
>  	 */
>  	uint16_t vq_used_cons_idx;
>  	uint16_t vq_avail_idx;
> +	uint16_t offset; /**< relative offset to obtain addr in mbuf */
>  	uint64_t mbuf_initializer; /**< value to init mbufs. */
>  	phys_addr_t virtio_net_hdr_mem; /**< hdr for each xmit packet */
> +	void        *virtio_net_hdr_vaddr;    /**< linear address of vring*/
> 
>  	struct rte_mbuf **sw_ring; /**< RX software ring. */
>  	/* dummy mbuf, for wraparound when processing RX ring. */
> --
> 2.1.4

Kind regards,
Pavel Fedin
Senior Engineer
Samsung Electronics Research center Russia