Re: [dpdk-dev] [PATCH v3 0/5] vhost: support selective datapath

["Wodkowski, PawelX" <pawelx.wodkowski@intel.com>]

> -----Original Message-----
> From: dev [mailto:dev-bounces@dpdk.org] On Behalf Of Zhihong Wang
> Sent: Tuesday, February 27, 2018 11:14 AM
> To: dev@dpdk.org
> Cc: Tan, Jianfeng <jianfeng.tan@intel.com>; Bie, Tiwei <tiwei.bie@intel.com>;
> maxime.coquelin@redhat.com; yliu@fridaylinux.org; Liang, Cunming
> <cunming.liang@intel.com>; Wang, Xiao W <xiao.w.wang@intel.com>; Daly,
> Dan <dan.daly@intel.com>; Wang, Zhihong <zhihong.wang@intel.com>
> Subject: [dpdk-dev] [PATCH v3 0/5] vhost: support selective datapath
> 
> This patch set introduces support for selective datapath in DPDK vhost-user
> lib. vDPA stands for vhost Data Path Acceleration. The idea is to enable
> various types of virtio-compatible devices to do data transfer with virtio
> driver directly to enable acceleration.
> 
> The default datapath is the existing software implementation, more options
> will be available when new engines are added.
> 
> Design details
> ====
> 
> An engine is a group of virtio-compatible devices. The definition of engine
> is as follows:
> 
> struct rte_vdpa_eng_addr {
> 	union {
> 		uint8_t __dummy[64];
> 		struct rte_pci_addr pci_addr;
> 	};
> };
> 
> struct rte_vdpa_eng_info {
> 	char name[MAX_VDPA_NAME_LEN];
> 	struct rte_vdpa_eng_addr *addr;
> };
> 
> struct rte_vdpa_dev_ops {
> 	vdpa_dev_conf_t        dev_conf;
> 	vdpa_dev_close_t       dev_close;
> 	vdpa_vring_state_set_t vring_state_set;
> 	vdpa_feature_set_t     feature_set;
> 	vdpa_migration_done_t  migration_done;
> 	vdpa_get_vfio_group_fd_t  get_vfio_group_fd;
> 	vdpa_get_vfio_device_fd_t get_vfio_device_fd;
> 	vdpa_get_notify_area_t    get_notify_area;
> };
> 
> struct rte_vdpa_eng_ops {
> 	vdpa_eng_init_t   eng_init;
> 	vdpa_eng_uninit_t eng_uninit;
> 	vdpa_info_query_t info_query;
> };
> 
> struct rte_vdpa_eng_driver {
> 	const char *name;
> 	struct rte_vdpa_eng_ops eng_ops;
> 	struct rte_vdpa_dev_ops dev_ops;
> } __rte_cache_aligned;
> 
> struct rte_vdpa_engine {
> 	struct rte_vdpa_eng_info    eng_info;
> 	struct rte_vdpa_eng_driver *eng_drv;
> } __rte_cache_aligned;
> 
> A set of engine ops is defined in rte_vdpa_eng_ops for engine init, uninit,
> and attributes reporting. The attributes are defined as follows:
> 
> struct rte_vdpa_eng_attr {
> 	uint64_t features;
> 	uint64_t protocol_features;
> 	uint32_t queue_num;
> 	uint32_t dev_num;
> };
> 
> A set of device ops is defined in rte_vdpa_dev_ops for each virtio device
> in the engine to do device specific operations.
> 
> Changes to the current vhost-user lib are:
> ====
> 
>  1. Make vhost device capabilities configurable to adopt various engines.
>     Such capabilities include supported features, protocol features, queue
>     number. APIs are introduced to let app configure these capabilities.
> 
>  2. In addition to the existing vhost framework, a set of callbacks is
>     added for vhost to call the driver for device operations at the right
>     time:
> 
>      a. dev_conf: Called to configure the actual device when the virtio
>         device becomes ready.
> 
>      b. dev_close: Called to close the actual device when the virtio device
>         is stopped.
> 
>      c. vring_state_set: Called to change the state of the vring in the
>         actual device when vring state changes.
> 
>      d. feature_set: Called to set the negotiated features to device.
> 
>      e. migration_done: Called to allow the device to response to RARP
>         sending.
> 
>      f. get_vfio_group_fd: Called to get the VFIO group fd of the device.
> 
>      g. get_vfio_device_fd: Called to get the VFIO device fd of the device.
> 
>      h. get_notify_area: Called to get the notify area info of the queue.
> 
>  3. To make vhost aware of its own type, an engine id (eid) and a device
>     id (did) are added into the vhost data structure to identify the actual
>     device. APIs are introduced to let app configure them. When the default
>     software datapath is used, eid and did are set to -1. When alternative
>     datapath is used, eid and did are set by app to specify which device to
>     use. Each vhost-user socket can have only 1 connection in this case.

Why only one connection is possible? We are already working on multiple
simultaneous connections in SPDK. So this will be some kind of step backward.

> 
> Working process:
> ====
> 
>  1. Register driver during DPDK initialization.
> 
>  2. Register engine with driver name and address.
> 
>  3. Get engine attributes.
> 
>  4. For vhost device creation:
> 
>       a. Register vhost-user socket.
> 
>       b. Set eid and did of the vhost-user socket.
> 
>       c. Register vhost-user callbacks.
> 
>       d. Start to wait for connection.
> 
>  4. When connection comes and virtio device data structure is negotiated,
>     the device will be configured with all needed info.
> 

Can you please provide new or modify existing example to show how to use this new API?
It would be easier to find any possible gaps if we can see real use case.

Pawel