1.. SPDX-License-Identifier: BSD-3-Clause 2 Copyright(c) 2010-2016 Intel Corporation. 3 4Vhost Library 5============= 6 7The vhost library implements a user space virtio net server allowing the user 8to manipulate the virtio ring directly. In another words, it allows the user 9to fetch/put packets from/to the VM virtio net device. To achieve this, a 10vhost library should be able to: 11 12* Access the guest memory: 13 14 For QEMU, this is done by using the ``-object memory-backend-file,share=on,...`` 15 option. Which means QEMU will create a file to serve as the guest RAM. 16 The ``share=on`` option allows another process to map that file, which 17 means it can access the guest RAM. 18 19* Know all the necessary information about the vring: 20 21 Information such as where the available ring is stored. Vhost defines some 22 messages (passed through a Unix domain socket file) to tell the backend all 23 the information it needs to know how to manipulate the vring. 24 25 26Vhost API Overview 27------------------ 28 29The following is an overview of some key Vhost API functions: 30 31* ``rte_vhost_driver_register(path, flags)`` 32 33 This function registers a vhost driver into the system. ``path`` specifies 34 the Unix domain socket file path. 35 36 Currently supported flags are: 37 38 - ``RTE_VHOST_USER_CLIENT`` 39 40 DPDK vhost-user will act as the client when this flag is given. See below 41 for an explanation. 42 43 - ``RTE_VHOST_USER_NO_RECONNECT`` 44 45 When DPDK vhost-user acts as the client it will keep trying to reconnect 46 to the server (QEMU) until it succeeds. This is useful in two cases: 47 48 * When QEMU is not started yet. 49 * When QEMU restarts (for example due to a guest OS reboot). 50 51 This reconnect option is enabled by default. However, it can be turned off 52 by setting this flag. 53 54 - ``RTE_VHOST_USER_IOMMU_SUPPORT`` 55 56 IOMMU support will be enabled when this flag is set. It is disabled by 57 default. 58 59 Enabling this flag makes possible to use guest vIOMMU to protect vhost 60 from accessing memory the virtio device isn't allowed to, when the feature 61 is negotiated and an IOMMU device is declared. 62 63 - ``RTE_VHOST_USER_POSTCOPY_SUPPORT`` 64 65 Postcopy live-migration support will be enabled when this flag is set. 66 It is disabled by default. 67 68 Enabling this flag should only be done when the calling application does 69 not pre-fault the guest shared memory, otherwise migration would fail. 70 71 - ``RTE_VHOST_USER_LINEARBUF_SUPPORT`` 72 73 Enabling this flag forces vhost dequeue function to only provide linear 74 pktmbuf (no multi-segmented pktmbuf). 75 76 The vhost library by default provides a single pktmbuf for given a 77 packet, but if for some reason the data doesn't fit into a single 78 pktmbuf (e.g., TSO is enabled), the library will allocate additional 79 pktmbufs from the same mempool and chain them together to create a 80 multi-segmented pktmbuf. 81 82 However, the vhost application needs to support multi-segmented format. 83 If the vhost application does not support that format and requires large 84 buffers to be dequeue, this flag should be enabled to force only linear 85 buffers (see RTE_VHOST_USER_EXTBUF_SUPPORT) or drop the packet. 86 87 It is disabled by default. 88 89 - ``RTE_VHOST_USER_EXTBUF_SUPPORT`` 90 91 Enabling this flag allows vhost dequeue function to allocate and attach 92 an external buffer to a pktmbuf if the pkmbuf doesn't provide enough 93 space to store all data. 94 95 This is useful when the vhost application wants to support large packets 96 but doesn't want to increase the default mempool object size nor to 97 support multi-segmented mbufs (non-linear). In this case, a fresh buffer 98 is allocated using rte_malloc() which gets attached to a pktmbuf using 99 rte_pktmbuf_attach_extbuf(). 100 101 See RTE_VHOST_USER_LINEARBUF_SUPPORT as well to disable multi-segmented 102 mbufs for application that doesn't support chained mbufs. 103 104 It is disabled by default. 105 106 - ``RTE_VHOST_USER_ASYNC_COPY`` 107 108 Asynchronous data path will be enabled when this flag is set. Async data 109 path allows applications to register async copy devices (typically 110 hardware DMA channels) to the vhost queues. Vhost leverages the copy 111 device registered to free CPU from memory copy operations. A set of 112 async data path APIs are defined for DPDK applications to make use of 113 the async capability. Only packets enqueued/dequeued by async APIs are 114 processed through the async data path. 115 116 Currently this feature is only implemented on split ring enqueue data 117 path. 118 119 It is disabled by default. 120 121* ``rte_vhost_driver_set_features(path, features)`` 122 123 This function sets the feature bits the vhost-user driver supports. The 124 vhost-user driver could be vhost-user net, yet it could be something else, 125 say, vhost-user SCSI. 126 127* ``rte_vhost_driver_callback_register(path, vhost_device_ops)`` 128 129 This function registers a set of callbacks, to let DPDK applications take 130 the appropriate action when some events happen. The following events are 131 currently supported: 132 133 * ``new_device(int vid)`` 134 135 This callback is invoked when a virtio device becomes ready. ``vid`` 136 is the vhost device ID. 137 138 * ``destroy_device(int vid)`` 139 140 This callback is invoked when a virtio device is paused or shut down. 141 142 * ``vring_state_changed(int vid, uint16_t queue_id, int enable)`` 143 144 This callback is invoked when a specific queue's state is changed, for 145 example to enabled or disabled. 146 147 * ``features_changed(int vid, uint64_t features)`` 148 149 This callback is invoked when the features is changed. For example, 150 ``VHOST_F_LOG_ALL`` will be set/cleared at the start/end of live 151 migration, respectively. 152 153 * ``new_connection(int vid)`` 154 155 This callback is invoked on new vhost-user socket connection. If DPDK 156 acts as the server the device should not be deleted before 157 ``destroy_connection`` callback is received. 158 159 * ``destroy_connection(int vid)`` 160 161 This callback is invoked when vhost-user socket connection is closed. 162 It indicates that device with id ``vid`` is no longer in use and can be 163 safely deleted. 164 165* ``rte_vhost_driver_disable/enable_features(path, features))`` 166 167 This function disables/enables some features. For example, it can be used to 168 disable mergeable buffers and TSO features, which both are enabled by 169 default. 170 171* ``rte_vhost_driver_start(path)`` 172 173 This function triggers the vhost-user negotiation. It should be invoked at 174 the end of initializing a vhost-user driver. 175 176* ``rte_vhost_enqueue_burst(vid, queue_id, pkts, count)`` 177 178 Transmits (enqueues) ``count`` packets from host to guest. 179 180* ``rte_vhost_dequeue_burst(vid, queue_id, mbuf_pool, pkts, count)`` 181 182 Receives (dequeues) ``count`` packets from guest, and stored them at ``pkts``. 183 184* ``rte_vhost_crypto_create(vid, cryptodev_id, sess_mempool, socket_id)`` 185 186 As an extension of new_device(), this function adds virtio-crypto workload 187 acceleration capability to the device. All crypto workload is processed by 188 DPDK cryptodev with the device ID of ``cryptodev_id``. 189 190* ``rte_vhost_crypto_free(vid)`` 191 192 Frees the memory and vhost-user message handlers created in 193 rte_vhost_crypto_create(). 194 195* ``rte_vhost_crypto_fetch_requests(vid, queue_id, ops, nb_ops)`` 196 197 Receives (dequeues) ``nb_ops`` virtio-crypto requests from guest, parses 198 them to DPDK Crypto Operations, and fills the ``ops`` with parsing results. 199 200* ``rte_vhost_crypto_finalize_requests(queue_id, ops, nb_ops)`` 201 202 After the ``ops`` are dequeued from Cryptodev, finalizes the jobs and 203 notifies the guest(s). 204 205* ``rte_vhost_crypto_set_zero_copy(vid, option)`` 206 207 Enable or disable zero copy feature of the vhost crypto backend. 208 209* ``rte_vhost_async_channel_register(vid, queue_id, features, ops)`` 210 211 Register a vhost queue with async copy device channel after vring 212 is enabled. Following device ``features`` must be specified together 213 with the registration: 214 215 * ``async_inorder`` 216 217 Async copy device can guarantee the ordering of copy completion 218 sequence. Copies are completed in the same order with that at 219 the submission time. 220 221 Currently, only ``async_inorder`` capable device is supported by vhost. 222 223 * ``async_threshold`` 224 225 The copy length (in bytes) below which CPU copy will be used even if 226 applications call async vhost APIs to enqueue/dequeue data. 227 228 Typical value is 512~1024 depending on the async device capability. 229 230 Applications must provide following ``ops`` callbacks for vhost lib to 231 work with the async copy devices: 232 233 * ``transfer_data(vid, queue_id, descs, opaque_data, count)`` 234 235 vhost invokes this function to submit copy data to the async devices. 236 For non-async_inorder capable devices, ``opaque_data`` could be used 237 for identifying the completed packets. 238 239 * ``check_completed_copies(vid, queue_id, opaque_data, max_packets)`` 240 241 vhost invokes this function to get the copy data completed by async 242 devices. 243 244* ``rte_vhost_async_channel_unregister(vid, queue_id)`` 245 246 Unregister the async copy device channel from a vhost queue. 247 Unregistration will fail, if the vhost queue has in-flight 248 packets that are not completed. 249 250 Unregister async copy devices in vring_state_changed() may 251 fail, as this API tries to acquire the spinlock of vhost 252 queue. The recommended way is to unregister async copy 253 devices for all vhost queues in destroy_device(), when a 254 virtio device is paused or shut down. 255 256* ``rte_vhost_submit_enqueue_burst(vid, queue_id, pkts, count, comp_pkts, comp_count)`` 257 258 Submit an enqueue request to transmit ``count`` packets from host to guest 259 by async data path. Successfully enqueued packets can be transfer completed 260 or being occupied by DMA engines; transfer completed packets are returned in 261 ``comp_pkts``, but others are not guaranteed to finish, when this API 262 call returns. 263 264 Applications must not free the packets submitted for enqueue until the 265 packets are completed. 266 267* ``rte_vhost_poll_enqueue_completed(vid, queue_id, pkts, count)`` 268 269 Poll enqueue completion status from async data path. Completed packets 270 are returned to applications through ``pkts``. 271 272Vhost-user Implementations 273-------------------------- 274 275Vhost-user uses Unix domain sockets for passing messages. This means the DPDK 276vhost-user implementation has two options: 277 278* DPDK vhost-user acts as the server. 279 280 DPDK will create a Unix domain socket server file and listen for 281 connections from the frontend. 282 283 Note, this is the default mode, and the only mode before DPDK v16.07. 284 285 286* DPDK vhost-user acts as the client. 287 288 Unlike the server mode, this mode doesn't create the socket file; 289 it just tries to connect to the server (which responses to create the 290 file instead). 291 292 When the DPDK vhost-user application restarts, DPDK vhost-user will try to 293 connect to the server again. This is how the "reconnect" feature works. 294 295 .. Note:: 296 * The "reconnect" feature requires **QEMU v2.7** (or above). 297 298 * The vhost supported features must be exactly the same before and 299 after the restart. For example, if TSO is disabled and then enabled, 300 nothing will work and issues undefined might happen. 301 302No matter which mode is used, once a connection is established, DPDK 303vhost-user will start receiving and processing vhost messages from QEMU. 304 305For messages with a file descriptor, the file descriptor can be used directly 306in the vhost process as it is already installed by the Unix domain socket. 307 308The supported vhost messages are: 309 310* ``VHOST_SET_MEM_TABLE`` 311* ``VHOST_SET_VRING_KICK`` 312* ``VHOST_SET_VRING_CALL`` 313* ``VHOST_SET_LOG_FD`` 314* ``VHOST_SET_VRING_ERR`` 315 316For ``VHOST_SET_MEM_TABLE`` message, QEMU will send information for each 317memory region and its file descriptor in the ancillary data of the message. 318The file descriptor is used to map that region. 319 320``VHOST_SET_VRING_KICK`` is used as the signal to put the vhost device into 321the data plane, and ``VHOST_GET_VRING_BASE`` is used as the signal to remove 322the vhost device from the data plane. 323 324When the socket connection is closed, vhost will destroy the device. 325 326Guest memory requirement 327------------------------ 328 329* Memory pre-allocation 330 331 For non-async data path, guest memory pre-allocation is not a 332 must. This can help save of memory. If users really want the guest memory 333 to be pre-allocated (e.g., for performance reason), we can add option 334 ``-mem-prealloc`` when starting QEMU. Or, we can lock all memory at vhost 335 side which will force memory to be allocated when mmap at vhost side; 336 option --mlockall in ovs-dpdk is an example in hand. 337 338 For async data path, we force the VM memory to be pre-allocated at vhost 339 lib when mapping the guest memory; and also we need to lock the memory to 340 prevent pages being swapped out to disk. 341 342* Memory sharing 343 344 Make sure ``share=on`` QEMU option is given. vhost-user will not work with 345 a QEMU version without shared memory mapping. 346 347Vhost supported vSwitch reference 348--------------------------------- 349 350For more vhost details and how to support vhost in vSwitch, please refer to 351the vhost example in the DPDK Sample Applications Guide. 352 353Vhost data path acceleration (vDPA) 354----------------------------------- 355 356vDPA supports selective datapath in vhost-user lib by enabling virtio ring 357compatible devices to serve virtio driver directly for datapath acceleration. 358 359``rte_vhost_driver_attach_vdpa_device`` is used to configure the vhost device 360with accelerated backend. 361 362Also vhost device capabilities are made configurable to adopt various devices. 363Such capabilities include supported features, protocol features, queue number. 364 365Finally, a set of device ops is defined for device specific operations: 366 367* ``get_queue_num`` 368 369 Called to get supported queue number of the device. 370 371* ``get_features`` 372 373 Called to get supported features of the device. 374 375* ``get_protocol_features`` 376 377 Called to get supported protocol features of the device. 378 379* ``dev_conf`` 380 381 Called to configure the actual device when the virtio device becomes ready. 382 383* ``dev_close`` 384 385 Called to close the actual device when the virtio device is stopped. 386 387* ``set_vring_state`` 388 389 Called to change the state of the vring in the actual device when vring state 390 changes. 391 392* ``set_features`` 393 394 Called to set the negotiated features to device. 395 396* ``migration_done`` 397 398 Called to allow the device to response to RARP sending. 399 400* ``get_vfio_group_fd`` 401 402 Called to get the VFIO group fd of the device. 403 404* ``get_vfio_device_fd`` 405 406 Called to get the VFIO device fd of the device. 407 408* ``get_notify_area`` 409 410 Called to get the notify area info of the queue. 411