4 :Copyright: 2014 Virtual Open Systems Sarl.
5 :Copyright: 2019 Intel Corporation
6 :Licence: This work is licensed under the terms of the GNU GPL,
7 version 2 or later. See the COPYING file in the top-level
10 .. contents:: Table of Contents
15 This protocol is aiming to complement the ``ioctl`` interface used to
16 control the vhost implementation in the Linux kernel. It implements
17 the control plane needed to establish virtqueue sharing with a user
18 space process on the same host. It uses communication over a Unix
19 domain socket to share file descriptors in the ancillary data of the
22 The protocol defines 2 sides of the communication, *master* and
23 *slave*. *Master* is the application that shares its virtqueues, in
24 our case QEMU. *Slave* is the consumer of the virtqueues.
26 In the current implementation QEMU is the *master*, and the *slave* is
27 the external process consuming the virtio queues, for example a
28 software Ethernet switch running in user space, such as Snabbswitch,
29 or a block device backend processing read & write to a virtual
30 disk. In order to facilitate interoperability between various backend
31 implementations, it is recommended to follow the :ref:`Backend program
32 conventions <backend_conventions>`.
34 *Master* and *slave* can be either a client (i.e. connecting) or
35 server (listening) in the socket communication.
40 .. Note:: All numbers are in the machine native byte order.
42 A vhost-user message consists of 3 header fields and a payload.
44 +---------+-------+------+---------+
45 | request | flags | size | payload |
46 +---------+-------+------+---------+
51 :request: 32-bit type of the request
53 :flags: 32-bit bit field
55 - Lower 2 bits are the version (currently 0x01)
56 - Bit 2 is the reply flag - needs to be sent on each reply from the slave
57 - Bit 3 is the need_reply flag - see :ref:`REPLY_ACK <reply_ack>` for
60 :size: 32-bit size of the payload
65 Depending on the request type, **payload** can be:
67 A single 64-bit integer
68 ^^^^^^^^^^^^^^^^^^^^^^^
74 :u64: a 64-bit unsigned integer
76 A vring state description
77 ^^^^^^^^^^^^^^^^^^^^^^^^^
83 :index: a 32-bit index
87 A vring address description
88 ^^^^^^^^^^^^^^^^^^^^^^^^^^^
90 +-------+-------+------+------------+------+-----------+-----+
91 | index | flags | size | descriptor | used | available | log |
92 +-------+-------+------+------------+------+-----------+-----+
94 :index: a 32-bit vring index
96 :flags: a 32-bit vring flags
98 :descriptor: a 64-bit ring address of the vring descriptor table
100 :used: a 64-bit ring address of the vring used ring
102 :available: a 64-bit ring address of the vring available ring
104 :log: a 64-bit guest address for logging
106 Note that a ring address is an IOVA if ``VIRTIO_F_IOMMU_PLATFORM`` has
107 been negotiated. Otherwise it is a user address.
109 Memory regions description
110 ^^^^^^^^^^^^^^^^^^^^^^^^^^
112 +-------------+---------+---------+-----+---------+
113 | num regions | padding | region0 | ... | region7 |
114 +-------------+---------+---------+-----+---------+
116 :num regions: a 32-bit number of regions
122 +---------------+------+--------------+-------------+
123 | guest address | size | user address | mmap offset |
124 +---------------+------+--------------+-------------+
126 :guest address: a 64-bit guest address of the region
130 :user address: a 64-bit user address
132 :mmap offset: 64-bit offset where region starts in the mapped memory
137 +----------+------------+
138 | log size | log offset |
139 +----------+------------+
141 :log size: size of area used for logging
143 :log offset: offset from start of supplied file descriptor where
144 logging starts (i.e. where guest address 0 would be
150 +------+------+--------------+-------------------+------+
151 | iova | size | user address | permissions flags | type |
152 +------+------+--------------+-------------------+------+
154 :iova: a 64-bit I/O virtual address programmed by the guest
158 :user address: a 64-bit user address
160 :permissions flags: an 8-bit value:
164 - 3: Read/Write access
166 :type: an 8-bit IOTLB message type:
169 - 3: IOTLB invalidate
170 - 4: IOTLB access fail
172 Virtio device config space
173 ^^^^^^^^^^^^^^^^^^^^^^^^^^
175 +--------+------+-------+---------+
176 | offset | size | flags | payload |
177 +--------+------+-------+---------+
179 :offset: a 32-bit offset of virtio device's configuration space
181 :size: a 32-bit configuration space access size in bytes
183 :flags: a 32-bit value:
184 - 0: Vhost master messages used for writeable fields
185 - 1: Vhost master messages used for live migration
187 :payload: Size bytes array holding the contents of the virtio
188 device's configuration space
190 Vring area description
191 ^^^^^^^^^^^^^^^^^^^^^^
193 +-----+------+--------+
194 | u64 | size | offset |
195 +-----+------+--------+
197 :u64: a 64-bit integer contains vring index and flags
199 :size: a 64-bit size of this area
201 :offset: a 64-bit offset of this area from the start of the
202 supplied file descriptor
207 +-----------+-------------+------------+------------+
208 | mmap size | mmap offset | num queues | queue size |
209 +-----------+-------------+------------+------------+
211 :mmap size: a 64-bit size of area to track inflight I/O
213 :mmap offset: a 64-bit offset of this area from the start
214 of the supplied file descriptor
216 :num queues: a 16-bit number of virtqueues
218 :queue size: a 16-bit size of virtqueues
223 In QEMU the vhost-user message is implemented with the following struct:
227 typedef struct VhostUserMsg {
228 VhostUserRequest request;
233 struct vhost_vring_state state;
234 struct vhost_vring_addr addr;
235 VhostUserMemory memory;
237 struct vhost_iotlb_msg iotlb;
238 VhostUserConfig config;
239 VhostUserVringArea area;
240 VhostUserInflight inflight;
242 } QEMU_PACKED VhostUserMsg;
247 The protocol for vhost-user is based on the existing implementation of
248 vhost for the Linux Kernel. Most messages that can be sent via the
249 Unix domain socket implementing vhost-user have an equivalent ioctl to
250 the kernel implementation.
252 The communication consists of *master* sending message requests and
253 *slave* sending message replies. Most of the requests don't require
254 replies. Here is a list of the ones that do:
256 * ``VHOST_USER_GET_FEATURES``
257 * ``VHOST_USER_GET_PROTOCOL_FEATURES``
258 * ``VHOST_USER_GET_VRING_BASE``
259 * ``VHOST_USER_SET_LOG_BASE`` (if ``VHOST_USER_PROTOCOL_F_LOG_SHMFD``)
260 * ``VHOST_USER_GET_INFLIGHT_FD`` (if ``VHOST_USER_PROTOCOL_F_INFLIGHT_SHMFD``)
264 :ref:`REPLY_ACK <reply_ack>`
265 The section on ``REPLY_ACK`` protocol extension.
267 There are several messages that the master sends with file descriptors passed
268 in the ancillary data:
270 * ``VHOST_USER_SET_MEM_TABLE``
271 * ``VHOST_USER_SET_LOG_BASE`` (if ``VHOST_USER_PROTOCOL_F_LOG_SHMFD``)
272 * ``VHOST_USER_SET_LOG_FD``
273 * ``VHOST_USER_SET_VRING_KICK``
274 * ``VHOST_USER_SET_VRING_CALL``
275 * ``VHOST_USER_SET_VRING_ERR``
276 * ``VHOST_USER_SET_SLAVE_REQ_FD``
277 * ``VHOST_USER_SET_INFLIGHT_FD`` (if ``VHOST_USER_PROTOCOL_F_INFLIGHT_SHMFD``)
279 If *master* is unable to send the full message or receives a wrong
280 reply it will close the connection. An optional reconnection mechanism
283 If *slave* detects some error such as incompatible features, it may also
284 close the connection. This should only happen in exceptional circumstances.
286 Any protocol extensions are gated by protocol feature bits, which
287 allows full backwards compatibility on both master and slave. As
288 older slaves don't support negotiating protocol features, a feature
289 bit was dedicated for this purpose::
291 #define VHOST_USER_F_PROTOCOL_FEATURES 30
293 Starting and stopping rings
294 ---------------------------
296 Client must only process each ring when it is started.
298 Client must only pass data between the ring and the backend, when the
301 If ring is started but disabled, client must process the ring without
302 talking to the backend.
304 For example, for a networking device, in the disabled state client
305 must not supply any new RX packets, but must process and discard any
308 If ``VHOST_USER_F_PROTOCOL_FEATURES`` has not been negotiated, the
309 ring is initialized in an enabled state.
311 If ``VHOST_USER_F_PROTOCOL_FEATURES`` has been negotiated, the ring is
312 initialized in a disabled state. Client must not pass data to/from the
313 backend until ring is enabled by ``VHOST_USER_SET_VRING_ENABLE`` with
314 parameter 1, or after it has been disabled by
315 ``VHOST_USER_SET_VRING_ENABLE`` with parameter 0.
317 Each ring is initialized in a stopped state, client must not process
318 it until ring is started, or after it has been stopped.
320 Client must start ring upon receiving a kick (that is, detecting that
321 file descriptor is readable) on the descriptor specified by
322 ``VHOST_USER_SET_VRING_KICK`` or receiving the in-band message
323 ``VHOST_USER_VRING_KICK`` if negotiated, and stop ring upon receiving
324 ``VHOST_USER_GET_VRING_BASE``.
326 While processing the rings (whether they are enabled or not), client
327 must support changing some configuration aspects on the fly.
329 Multiple queue support
330 ----------------------
332 Many devices have a fixed number of virtqueues. In this case the master
333 already knows the number of available virtqueues without communicating with the
336 Some devices do not have a fixed number of virtqueues. Instead the maximum
337 number of virtqueues is chosen by the slave. The number can depend on host
338 resource availability or slave implementation details. Such devices are called
339 multiple queue devices.
341 Multiple queue support allows the slave to advertise the maximum number of
342 queues. This is treated as a protocol extension, hence the slave has to
343 implement protocol features first. The multiple queues feature is supported
344 only when the protocol feature ``VHOST_USER_PROTOCOL_F_MQ`` (bit 0) is set.
346 The max number of queues the slave supports can be queried with message
347 ``VHOST_USER_GET_QUEUE_NUM``. Master should stop when the number of requested
348 queues is bigger than that.
350 As all queues share one connection, the master uses a unique index for each
351 queue in the sent message to identify a specified queue.
353 The master enables queues by sending message ``VHOST_USER_SET_VRING_ENABLE``.
354 vhost-user-net has historically automatically enabled the first queue pair.
356 Slaves should always implement the ``VHOST_USER_PROTOCOL_F_MQ`` protocol
357 feature, even for devices with a fixed number of virtqueues, since it is simple
358 to implement and offers a degree of introspection.
360 Masters must not rely on the ``VHOST_USER_PROTOCOL_F_MQ`` protocol feature for
361 devices with a fixed number of virtqueues. Only true multiqueue devices
362 require this protocol feature.
367 During live migration, the master may need to track the modifications
368 the slave makes to the memory mapped regions. The client should mark
369 the dirty pages in a log. Once it complies to this logging, it may
370 declare the ``VHOST_F_LOG_ALL`` vhost feature.
372 To start/stop logging of data/used ring writes, server may send
373 messages ``VHOST_USER_SET_FEATURES`` with ``VHOST_F_LOG_ALL`` and
374 ``VHOST_USER_SET_VRING_ADDR`` with ``VHOST_VRING_F_LOG`` in ring's
375 flags set to 1/0, respectively.
377 All the modifications to memory pointed by vring "descriptor" should
378 be marked. Modifications to "used" vring should be marked if
379 ``VHOST_VRING_F_LOG`` is part of ring's flags.
381 Dirty pages are of size::
383 #define VHOST_LOG_PAGE 0x1000
385 The log memory fd is provided in the ancillary data of
386 ``VHOST_USER_SET_LOG_BASE`` message when the slave has
387 ``VHOST_USER_PROTOCOL_F_LOG_SHMFD`` protocol feature.
389 The size of the log is supplied as part of ``VhostUserMsg`` which
390 should be large enough to cover all known guest addresses. Log starts
391 at the supplied offset in the supplied file descriptor. The log
392 covers from address 0 to the maximum of guest regions. In pseudo-code,
393 to mark page at ``addr`` as dirty::
395 page = addr / VHOST_LOG_PAGE
396 log[page / 8] |= 1 << page % 8
398 Where ``addr`` is the guest physical address.
400 Use atomic operations, as the log may be concurrently manipulated.
402 Note that when logging modifications to the used ring (when
403 ``VHOST_VRING_F_LOG`` is set for this ring), ``log_guest_addr`` should
404 be used to calculate the log offset: the write to first byte of the
405 used ring is logged at this offset from log start. Also note that this
406 value might be outside the legal guest physical address range
407 (i.e. does not have to be covered by the ``VhostUserMemory`` table), but
408 the bit offset of the last byte of the ring must fall within the size
409 supplied by ``VhostUserLog``.
411 ``VHOST_USER_SET_LOG_FD`` is an optional message with an eventfd in
412 ancillary data, it may be used to inform the master that the log has
415 Once the source has finished migration, rings will be stopped by the
416 source. No further update must be done before rings are restarted.
418 In postcopy migration the slave is started before all the memory has
419 been received from the source host, and care must be taken to avoid
420 accessing pages that have yet to be received. The slave opens a
421 'userfault'-fd and registers the memory with it; this fd is then
422 passed back over to the master. The master services requests on the
423 userfaultfd for pages that are accessed and when the page is available
424 it performs WAKE ioctl's on the userfaultfd to wake the stalled
425 slave. The client indicates support for this via the
426 ``VHOST_USER_PROTOCOL_F_PAGEFAULT`` feature.
431 The master sends a list of vhost memory regions to the slave using the
432 ``VHOST_USER_SET_MEM_TABLE`` message. Each region has two base
433 addresses: a guest address and a user address.
435 Messages contain guest addresses and/or user addresses to reference locations
436 within the shared memory. The mapping of these addresses works as follows.
438 User addresses map to the vhost memory region containing that user address.
440 When the ``VIRTIO_F_IOMMU_PLATFORM`` feature has not been negotiated:
442 * Guest addresses map to the vhost memory region containing that guest
445 When the ``VIRTIO_F_IOMMU_PLATFORM`` feature has been negotiated:
447 * Guest addresses are also called I/O virtual addresses (IOVAs). They are
448 translated to user addresses via the IOTLB.
450 * The vhost memory region guest address is not used.
455 When the ``VIRTIO_F_IOMMU_PLATFORM`` feature has been negotiated, the
456 master sends IOTLB entries update & invalidation by sending
457 ``VHOST_USER_IOTLB_MSG`` requests to the slave with a ``struct
458 vhost_iotlb_msg`` as payload. For update events, the ``iotlb`` payload
459 has to be filled with the update message type (2), the I/O virtual
460 address, the size, the user virtual address, and the permissions
461 flags. Addresses and size must be within vhost memory regions set via
462 the ``VHOST_USER_SET_MEM_TABLE`` request. For invalidation events, the
463 ``iotlb`` payload has to be filled with the invalidation message type
464 (3), the I/O virtual address and the size. On success, the slave is
465 expected to reply with a zero payload, non-zero otherwise.
467 The slave relies on the slave communication channel (see :ref:`Slave
468 communication <slave_communication>` section below) to send IOTLB miss
469 and access failure events, by sending ``VHOST_USER_SLAVE_IOTLB_MSG``
470 requests to the master with a ``struct vhost_iotlb_msg`` as
471 payload. For miss events, the iotlb payload has to be filled with the
472 miss message type (1), the I/O virtual address and the permissions
473 flags. For access failure event, the iotlb payload has to be filled
474 with the access failure message type (4), the I/O virtual address and
475 the permissions flags. For synchronization purpose, the slave may
476 rely on the reply-ack feature, so the master may send a reply when
477 operation is completed if the reply-ack feature is negotiated and
478 slaves requests a reply. For miss events, completed operation means
479 either master sent an update message containing the IOTLB entry
480 containing requested address and permission, or master sent nothing if
481 the IOTLB miss message is invalid (invalid IOVA or permission).
483 The master isn't expected to take the initiative to send IOTLB update
484 messages, as the slave sends IOTLB miss messages for the guest virtual
485 memory areas it needs to access.
487 .. _slave_communication:
492 An optional communication channel is provided if the slave declares
493 ``VHOST_USER_PROTOCOL_F_SLAVE_REQ`` protocol feature, to allow the
494 slave to make requests to the master.
496 The fd is provided via ``VHOST_USER_SET_SLAVE_REQ_FD`` ancillary data.
498 A slave may then send ``VHOST_USER_SLAVE_*`` messages to the master
499 using this fd communication channel.
501 If ``VHOST_USER_PROTOCOL_F_SLAVE_SEND_FD`` protocol feature is
502 negotiated, slave can send file descriptors (at most 8 descriptors in
503 each message) to master via ancillary data using this fd communication
506 Inflight I/O tracking
507 ---------------------
509 To support reconnecting after restart or crash, slave may need to
510 resubmit inflight I/Os. If virtqueue is processed in order, we can
511 easily achieve that by getting the inflight descriptors from
512 descriptor table (split virtqueue) or descriptor ring (packed
513 virtqueue). However, it can't work when we process descriptors
514 out-of-order because some entries which store the information of
515 inflight descriptors in available ring (split virtqueue) or descriptor
516 ring (packed virtqueue) might be overrided by new entries. To solve
517 this problem, slave need to allocate an extra buffer to store this
518 information of inflight descriptors and share it with master for
519 persistent. ``VHOST_USER_GET_INFLIGHT_FD`` and
520 ``VHOST_USER_SET_INFLIGHT_FD`` are used to transfer this buffer
521 between master and slave. And the format of this buffer is described
524 +---------------+---------------+-----+---------------+
525 | queue0 region | queue1 region | ... | queueN region |
526 +---------------+---------------+-----+---------------+
528 N is the number of available virtqueues. Slave could get it from num
529 queues field of ``VhostUserInflight``.
531 For split virtqueue, queue region can be implemented as:
535 typedef struct DescStateSplit {
536 /* Indicate whether this descriptor is inflight or not.
537 * Only available for head-descriptor. */
543 /* Maintain a list for the last batch of used descriptors.
544 * Only available when batching is used for submitting */
547 /* Used to preserve the order of fetching available descriptors.
548 * Only available for head-descriptor. */
552 typedef struct QueueRegionSplit {
553 /* The feature flags of this region. Now it's initialized to 0. */
556 /* The version of this region. It's 1 currently.
557 * Zero value indicates an uninitialized buffer */
560 /* The size of DescStateSplit array. It's equal to the virtqueue
561 * size. Slave could get it from queue size field of VhostUserInflight. */
564 /* The head of list that track the last batch of used descriptors. */
565 uint16_t last_batch_head;
567 /* Store the idx value of used ring */
570 /* Used to track the state of each descriptor in descriptor table */
571 DescStateSplit desc[];
574 To track inflight I/O, the queue region should be processed as follows:
576 When receiving available buffers from the driver:
578 #. Get the next available head-descriptor index from available ring, ``i``
580 #. Set ``desc[i].counter`` to the value of global counter
582 #. Increase global counter by 1
584 #. Set ``desc[i].inflight`` to 1
586 When supplying used buffers to the driver:
588 1. Get corresponding used head-descriptor index, i
590 2. Set ``desc[i].next`` to ``last_batch_head``
592 3. Set ``last_batch_head`` to ``i``
594 #. Steps 1,2,3 may be performed repeatedly if batching is possible
596 #. Increase the ``idx`` value of used ring by the size of the batch
598 #. Set the ``inflight`` field of each ``DescStateSplit`` entry in the batch to 0
600 #. Set ``used_idx`` to the ``idx`` value of used ring
604 #. If the value of ``used_idx`` does not match the ``idx`` value of
605 used ring (means the inflight field of ``DescStateSplit`` entries in
606 last batch may be incorrect),
608 a. Subtract the value of ``used_idx`` from the ``idx`` value of
609 used ring to get last batch size of ``DescStateSplit`` entries
611 #. Set the ``inflight`` field of each ``DescStateSplit`` entry to 0 in last batch
612 list which starts from ``last_batch_head``
614 #. Set ``used_idx`` to the ``idx`` value of used ring
616 #. Resubmit inflight ``DescStateSplit`` entries in order of their
619 For packed virtqueue, queue region can be implemented as:
623 typedef struct DescStatePacked {
624 /* Indicate whether this descriptor is inflight or not.
625 * Only available for head-descriptor. */
631 /* Link to the next free entry */
634 /* Link to the last entry of descriptor list.
635 * Only available for head-descriptor. */
638 /* The length of descriptor list.
639 * Only available for head-descriptor. */
642 /* Used to preserve the order of fetching available descriptors.
643 * Only available for head-descriptor. */
649 /* The descriptor flags */
652 /* The buffer length */
655 /* The buffer address */
659 typedef struct QueueRegionPacked {
660 /* The feature flags of this region. Now it's initialized to 0. */
663 /* The version of this region. It's 1 currently.
664 * Zero value indicates an uninitialized buffer */
667 /* The size of DescStatePacked array. It's equal to the virtqueue
668 * size. Slave could get it from queue size field of VhostUserInflight. */
671 /* The head of free DescStatePacked entry list */
674 /* The old head of free DescStatePacked entry list */
675 uint16_t old_free_head;
677 /* The used index of descriptor ring */
680 /* The old used index of descriptor ring */
681 uint16_t old_used_idx;
683 /* Device ring wrap counter */
684 uint8_t used_wrap_counter;
686 /* The old device ring wrap counter */
687 uint8_t old_used_wrap_counter;
692 /* Used to track the state of each descriptor fetched from descriptor ring */
693 DescStatePacked desc[];
696 To track inflight I/O, the queue region should be processed as follows:
698 When receiving available buffers from the driver:
700 #. Get the next available descriptor entry from descriptor ring, ``d``
702 #. If ``d`` is head descriptor,
704 a. Set ``desc[old_free_head].num`` to 0
706 #. Set ``desc[old_free_head].counter`` to the value of global counter
708 #. Increase global counter by 1
710 #. Set ``desc[old_free_head].inflight`` to 1
712 #. If ``d`` is last descriptor, set ``desc[old_free_head].last`` to
715 #. Increase ``desc[old_free_head].num`` by 1
717 #. Set ``desc[free_head].addr``, ``desc[free_head].len``,
718 ``desc[free_head].flags``, ``desc[free_head].id`` to ``d.addr``,
719 ``d.len``, ``d.flags``, ``d.id``
721 #. Set ``free_head`` to ``desc[free_head].next``
723 #. If ``d`` is last descriptor, set ``old_free_head`` to ``free_head``
725 When supplying used buffers to the driver:
727 1. Get corresponding used head-descriptor entry from descriptor ring,
730 2. Get corresponding ``DescStatePacked`` entry, ``e``
732 3. Set ``desc[e.last].next`` to ``free_head``
734 4. Set ``free_head`` to the index of ``e``
736 #. Steps 1,2,3,4 may be performed repeatedly if batching is possible
738 #. Increase ``used_idx`` by the size of the batch and update
739 ``used_wrap_counter`` if needed
741 #. Update ``d.flags``
743 #. Set the ``inflight`` field of each head ``DescStatePacked`` entry
746 #. Set ``old_free_head``, ``old_used_idx``, ``old_used_wrap_counter``
747 to ``free_head``, ``used_idx``, ``used_wrap_counter``
751 #. If ``used_idx`` does not match ``old_used_idx`` (means the
752 ``inflight`` field of ``DescStatePacked`` entries in last batch may
755 a. Get the next descriptor ring entry through ``old_used_idx``, ``d``
757 #. Use ``old_used_wrap_counter`` to calculate the available flags
759 #. If ``d.flags`` is not equal to the calculated flags value (means
760 slave has submitted the buffer to guest driver before crash, so
761 it has to commit the in-progres update), set ``old_free_head``,
762 ``old_used_idx``, ``old_used_wrap_counter`` to ``free_head``,
763 ``used_idx``, ``used_wrap_counter``
765 #. Set ``free_head``, ``used_idx``, ``used_wrap_counter`` to
766 ``old_free_head``, ``old_used_idx``, ``old_used_wrap_counter``
767 (roll back any in-progress update)
769 #. Set the ``inflight`` field of each ``DescStatePacked`` entry in
772 #. Resubmit inflight ``DescStatePacked`` entries in order of their
775 In-band notifications
776 ---------------------
778 In some limited situations (e.g. for simulation) it is desirable to
779 have the kick, call and error (if used) signals done via in-band
780 messages instead of asynchronous eventfd notifications. This can be
781 done by negotiating the ``VHOST_USER_PROTOCOL_F_INBAND_NOTIFICATIONS``
784 Note that due to the fact that too many messages on the sockets can
785 cause the sending application(s) to block, it is not advised to use
786 this feature unless absolutely necessary. It is also considered an
787 error to negotiate this feature without also negotiating
788 ``VHOST_USER_PROTOCOL_F_SLAVE_REQ`` and ``VHOST_USER_PROTOCOL_F_REPLY_ACK``,
789 the former is necessary for getting a message channel from the slave
790 to the master, while the latter needs to be used with the in-band
791 notification messages to block until they are processed, both to avoid
792 blocking later and for proper processing (at least in the simulation
793 use case.) As it has no other way of signalling this error, the slave
794 should close the connection as a response to a
795 ``VHOST_USER_SET_PROTOCOL_FEATURES`` message that sets the in-band
796 notifications feature flag without the other two.
803 #define VHOST_USER_PROTOCOL_F_MQ 0
804 #define VHOST_USER_PROTOCOL_F_LOG_SHMFD 1
805 #define VHOST_USER_PROTOCOL_F_RARP 2
806 #define VHOST_USER_PROTOCOL_F_REPLY_ACK 3
807 #define VHOST_USER_PROTOCOL_F_MTU 4
808 #define VHOST_USER_PROTOCOL_F_SLAVE_REQ 5
809 #define VHOST_USER_PROTOCOL_F_CROSS_ENDIAN 6
810 #define VHOST_USER_PROTOCOL_F_CRYPTO_SESSION 7
811 #define VHOST_USER_PROTOCOL_F_PAGEFAULT 8
812 #define VHOST_USER_PROTOCOL_F_CONFIG 9
813 #define VHOST_USER_PROTOCOL_F_SLAVE_SEND_FD 10
814 #define VHOST_USER_PROTOCOL_F_HOST_NOTIFIER 11
815 #define VHOST_USER_PROTOCOL_F_INFLIGHT_SHMFD 12
816 #define VHOST_USER_PROTOCOL_F_RESET_DEVICE 13
817 #define VHOST_USER_PROTOCOL_F_INBAND_NOTIFICATIONS 14
818 #define VHOST_USER_PROTOCOL_F_CONFIGURE_MEM_SLOTS 15
819 #define VHOST_USER_PROTOCOL_F_STATUS 16
824 ``VHOST_USER_GET_FEATURES``
826 :equivalent ioctl: ``VHOST_GET_FEATURES``
828 :slave payload: ``u64``
830 Get from the underlying vhost implementation the features bitmask.
831 Feature bit ``VHOST_USER_F_PROTOCOL_FEATURES`` signals slave support
832 for ``VHOST_USER_GET_PROTOCOL_FEATURES`` and
833 ``VHOST_USER_SET_PROTOCOL_FEATURES``.
835 ``VHOST_USER_SET_FEATURES``
837 :equivalent ioctl: ``VHOST_SET_FEATURES``
838 :master payload: ``u64``
840 Enable features in the underlying vhost implementation using a
841 bitmask. Feature bit ``VHOST_USER_F_PROTOCOL_FEATURES`` signals
842 slave support for ``VHOST_USER_GET_PROTOCOL_FEATURES`` and
843 ``VHOST_USER_SET_PROTOCOL_FEATURES``.
845 ``VHOST_USER_GET_PROTOCOL_FEATURES``
847 :equivalent ioctl: ``VHOST_GET_FEATURES``
849 :slave payload: ``u64``
851 Get the protocol feature bitmask from the underlying vhost
852 implementation. Only legal if feature bit
853 ``VHOST_USER_F_PROTOCOL_FEATURES`` is present in
854 ``VHOST_USER_GET_FEATURES``.
857 Slave that reported ``VHOST_USER_F_PROTOCOL_FEATURES`` must
858 support this message even before ``VHOST_USER_SET_FEATURES`` was
861 ``VHOST_USER_SET_PROTOCOL_FEATURES``
863 :equivalent ioctl: ``VHOST_SET_FEATURES``
864 :master payload: ``u64``
866 Enable protocol features in the underlying vhost implementation.
868 Only legal if feature bit ``VHOST_USER_F_PROTOCOL_FEATURES`` is present in
869 ``VHOST_USER_GET_FEATURES``.
872 Slave that reported ``VHOST_USER_F_PROTOCOL_FEATURES`` must support
873 this message even before ``VHOST_USER_SET_FEATURES`` was called.
875 ``VHOST_USER_SET_OWNER``
877 :equivalent ioctl: ``VHOST_SET_OWNER``
880 Issued when a new connection is established. It sets the current
881 *master* as an owner of the session. This can be used on the *slave*
882 as a "session start" flag.
884 ``VHOST_USER_RESET_OWNER``
888 .. admonition:: Deprecated
890 This is no longer used. Used to be sent to request disabling all
891 rings, but some clients interpreted it to also discard connection
892 state (this interpretation would lead to bugs). It is recommended
893 that clients either ignore this message, or use it to disable all
896 ``VHOST_USER_SET_MEM_TABLE``
898 :equivalent ioctl: ``VHOST_SET_MEM_TABLE``
899 :master payload: memory regions description
900 :slave payload: (postcopy only) memory regions description
902 Sets the memory map regions on the slave so it can translate the
903 vring addresses. In the ancillary data there is an array of file
904 descriptors for each memory mapped region. The size and ordering of
905 the fds matches the number and ordering of memory regions.
907 When ``VHOST_USER_POSTCOPY_LISTEN`` has been received,
908 ``SET_MEM_TABLE`` replies with the bases of the memory mapped
909 regions to the master. The slave must have mmap'd the regions but
910 not yet accessed them and should not yet generate a userfault
914 ``NEED_REPLY_MASK`` is not set in this case. QEMU will then
915 reply back to the list of mappings with an empty
916 ``VHOST_USER_SET_MEM_TABLE`` as an acknowledgement; only upon
917 reception of this message may the guest start accessing the memory
918 and generating faults.
920 ``VHOST_USER_SET_LOG_BASE``
922 :equivalent ioctl: ``VHOST_SET_LOG_BASE``
926 Sets logging shared memory space.
928 When slave has ``VHOST_USER_PROTOCOL_F_LOG_SHMFD`` protocol feature,
929 the log memory fd is provided in the ancillary data of
930 ``VHOST_USER_SET_LOG_BASE`` message, the size and offset of shared
931 memory area provided in the message.
933 ``VHOST_USER_SET_LOG_FD``
935 :equivalent ioctl: ``VHOST_SET_LOG_FD``
938 Sets the logging file descriptor, which is passed as ancillary data.
940 ``VHOST_USER_SET_VRING_NUM``
942 :equivalent ioctl: ``VHOST_SET_VRING_NUM``
943 :master payload: vring state description
945 Set the size of the queue.
947 ``VHOST_USER_SET_VRING_ADDR``
949 :equivalent ioctl: ``VHOST_SET_VRING_ADDR``
950 :master payload: vring address description
953 Sets the addresses of the different aspects of the vring.
955 ``VHOST_USER_SET_VRING_BASE``
957 :equivalent ioctl: ``VHOST_SET_VRING_BASE``
958 :master payload: vring state description
960 Sets the base offset in the available vring.
962 ``VHOST_USER_GET_VRING_BASE``
964 :equivalent ioctl: ``VHOST_USER_GET_VRING_BASE``
965 :master payload: vring state description
966 :slave payload: vring state description
968 Get the available vring base offset.
970 ``VHOST_USER_SET_VRING_KICK``
972 :equivalent ioctl: ``VHOST_SET_VRING_KICK``
973 :master payload: ``u64``
975 Set the event file descriptor for adding buffers to the vring. It is
976 passed in the ancillary data.
978 Bits (0-7) of the payload contain the vring index. Bit 8 is the
979 invalid FD flag. This flag is set when there is no file descriptor
980 in the ancillary data. This signals that polling should be used
981 instead of waiting for the kick. Note that if the protocol feature
982 ``VHOST_USER_PROTOCOL_F_INBAND_NOTIFICATIONS`` has been negotiated
983 this message isn't necessary as the ring is also started on the
984 ``VHOST_USER_VRING_KICK`` message, it may however still be used to
985 set an event file descriptor (which will be preferred over the
986 message) or to enable polling.
988 ``VHOST_USER_SET_VRING_CALL``
990 :equivalent ioctl: ``VHOST_SET_VRING_CALL``
991 :master payload: ``u64``
993 Set the event file descriptor to signal when buffers are used. It is
994 passed in the ancillary data.
996 Bits (0-7) of the payload contain the vring index. Bit 8 is the
997 invalid FD flag. This flag is set when there is no file descriptor
998 in the ancillary data. This signals that polling will be used
999 instead of waiting for the call. Note that if the protocol features
1000 ``VHOST_USER_PROTOCOL_F_INBAND_NOTIFICATIONS`` and
1001 ``VHOST_USER_PROTOCOL_F_SLAVE_REQ`` have been negotiated this message
1002 isn't necessary as the ``VHOST_USER_SLAVE_VRING_CALL`` message can be
1003 used, it may however still be used to set an event file descriptor
1004 or to enable polling.
1006 ``VHOST_USER_SET_VRING_ERR``
1008 :equivalent ioctl: ``VHOST_SET_VRING_ERR``
1009 :master payload: ``u64``
1011 Set the event file descriptor to signal when error occurs. It is
1012 passed in the ancillary data.
1014 Bits (0-7) of the payload contain the vring index. Bit 8 is the
1015 invalid FD flag. This flag is set when there is no file descriptor
1016 in the ancillary data. Note that if the protocol features
1017 ``VHOST_USER_PROTOCOL_F_INBAND_NOTIFICATIONS`` and
1018 ``VHOST_USER_PROTOCOL_F_SLAVE_REQ`` have been negotiated this message
1019 isn't necessary as the ``VHOST_USER_SLAVE_VRING_ERR`` message can be
1020 used, it may however still be used to set an event file descriptor
1021 (which will be preferred over the message).
1023 ``VHOST_USER_GET_QUEUE_NUM``
1025 :equivalent ioctl: N/A
1026 :master payload: N/A
1029 Query how many queues the backend supports.
1031 This request should be sent only when ``VHOST_USER_PROTOCOL_F_MQ``
1032 is set in queried protocol features by
1033 ``VHOST_USER_GET_PROTOCOL_FEATURES``.
1035 ``VHOST_USER_SET_VRING_ENABLE``
1037 :equivalent ioctl: N/A
1038 :master payload: vring state description
1040 Signal slave to enable or disable corresponding vring.
1042 This request should be sent only when
1043 ``VHOST_USER_F_PROTOCOL_FEATURES`` has been negotiated.
1045 ``VHOST_USER_SEND_RARP``
1047 :equivalent ioctl: N/A
1048 :master payload: ``u64``
1050 Ask vhost user backend to broadcast a fake RARP to notify the migration
1051 is terminated for guest that does not support GUEST_ANNOUNCE.
1053 Only legal if feature bit ``VHOST_USER_F_PROTOCOL_FEATURES`` is
1054 present in ``VHOST_USER_GET_FEATURES`` and protocol feature bit
1055 ``VHOST_USER_PROTOCOL_F_RARP`` is present in
1056 ``VHOST_USER_GET_PROTOCOL_FEATURES``. The first 6 bytes of the
1057 payload contain the mac address of the guest to allow the vhost user
1058 backend to construct and broadcast the fake RARP.
1060 ``VHOST_USER_NET_SET_MTU``
1062 :equivalent ioctl: N/A
1063 :master payload: ``u64``
1065 Set host MTU value exposed to the guest.
1067 This request should be sent only when ``VIRTIO_NET_F_MTU`` feature
1068 has been successfully negotiated, ``VHOST_USER_F_PROTOCOL_FEATURES``
1069 is present in ``VHOST_USER_GET_FEATURES`` and protocol feature bit
1070 ``VHOST_USER_PROTOCOL_F_NET_MTU`` is present in
1071 ``VHOST_USER_GET_PROTOCOL_FEATURES``.
1073 If ``VHOST_USER_PROTOCOL_F_REPLY_ACK`` is negotiated, slave must
1074 respond with zero in case the specified MTU is valid, or non-zero
1077 ``VHOST_USER_SET_SLAVE_REQ_FD``
1079 :equivalent ioctl: N/A
1080 :master payload: N/A
1082 Set the socket file descriptor for slave initiated requests. It is passed
1083 in the ancillary data.
1085 This request should be sent only when
1086 ``VHOST_USER_F_PROTOCOL_FEATURES`` has been negotiated, and protocol
1087 feature bit ``VHOST_USER_PROTOCOL_F_SLAVE_REQ`` bit is present in
1088 ``VHOST_USER_GET_PROTOCOL_FEATURES``. If
1089 ``VHOST_USER_PROTOCOL_F_REPLY_ACK`` is negotiated, slave must
1090 respond with zero for success, non-zero otherwise.
1092 ``VHOST_USER_IOTLB_MSG``
1094 :equivalent ioctl: N/A (equivalent to ``VHOST_IOTLB_MSG`` message type)
1095 :master payload: ``struct vhost_iotlb_msg``
1096 :slave payload: ``u64``
1098 Send IOTLB messages with ``struct vhost_iotlb_msg`` as payload.
1100 Master sends such requests to update and invalidate entries in the
1101 device IOTLB. The slave has to acknowledge the request with sending
1102 zero as ``u64`` payload for success, non-zero otherwise.
1104 This request should be send only when ``VIRTIO_F_IOMMU_PLATFORM``
1105 feature has been successfully negotiated.
1107 ``VHOST_USER_SET_VRING_ENDIAN``
1109 :equivalent ioctl: ``VHOST_SET_VRING_ENDIAN``
1110 :master payload: vring state description
1112 Set the endianness of a VQ for legacy devices. Little-endian is
1113 indicated with state.num set to 0 and big-endian is indicated with
1114 state.num set to 1. Other values are invalid.
1116 This request should be sent only when
1117 ``VHOST_USER_PROTOCOL_F_CROSS_ENDIAN`` has been negotiated.
1118 Backends that negotiated this feature should handle both
1119 endiannesses and expect this message once (per VQ) during device
1120 configuration (ie. before the master starts the VQ).
1122 ``VHOST_USER_GET_CONFIG``
1124 :equivalent ioctl: N/A
1125 :master payload: virtio device config space
1126 :slave payload: virtio device config space
1128 When ``VHOST_USER_PROTOCOL_F_CONFIG`` is negotiated, this message is
1129 submitted by the vhost-user master to fetch the contents of the
1130 virtio device configuration space, vhost-user slave's payload size
1131 MUST match master's request, vhost-user slave uses zero length of
1132 payload to indicate an error to vhost-user master. The vhost-user
1133 master may cache the contents to avoid repeated
1134 ``VHOST_USER_GET_CONFIG`` calls.
1136 ``VHOST_USER_SET_CONFIG``
1138 :equivalent ioctl: N/A
1139 :master payload: virtio device config space
1142 When ``VHOST_USER_PROTOCOL_F_CONFIG`` is negotiated, this message is
1143 submitted by the vhost-user master when the Guest changes the virtio
1144 device configuration space and also can be used for live migration
1145 on the destination host. The vhost-user slave must check the flags
1146 field, and slaves MUST NOT accept SET_CONFIG for read-only
1147 configuration space fields unless the live migration bit is set.
1149 ``VHOST_USER_CREATE_CRYPTO_SESSION``
1151 :equivalent ioctl: N/A
1152 :master payload: crypto session description
1153 :slave payload: crypto session description
1155 Create a session for crypto operation. The server side must return
1156 the session id, 0 or positive for success, negative for failure.
1157 This request should be sent only when
1158 ``VHOST_USER_PROTOCOL_F_CRYPTO_SESSION`` feature has been
1159 successfully negotiated. It's a required feature for crypto
1162 ``VHOST_USER_CLOSE_CRYPTO_SESSION``
1164 :equivalent ioctl: N/A
1165 :master payload: ``u64``
1167 Close a session for crypto operation which was previously
1168 created by ``VHOST_USER_CREATE_CRYPTO_SESSION``.
1170 This request should be sent only when
1171 ``VHOST_USER_PROTOCOL_F_CRYPTO_SESSION`` feature has been
1172 successfully negotiated. It's a required feature for crypto
1175 ``VHOST_USER_POSTCOPY_ADVISE``
1177 :master payload: N/A
1178 :slave payload: userfault fd
1180 When ``VHOST_USER_PROTOCOL_F_PAGEFAULT`` is supported, the master
1181 advises slave that a migration with postcopy enabled is underway,
1182 the slave must open a userfaultfd for later use. Note that at this
1183 stage the migration is still in precopy mode.
1185 ``VHOST_USER_POSTCOPY_LISTEN``
1187 :master payload: N/A
1189 Master advises slave that a transition to postcopy mode has
1190 happened. The slave must ensure that shared memory is registered
1191 with userfaultfd to cause faulting of non-present pages.
1193 This is always sent sometime after a ``VHOST_USER_POSTCOPY_ADVISE``,
1194 and thus only when ``VHOST_USER_PROTOCOL_F_PAGEFAULT`` is supported.
1196 ``VHOST_USER_POSTCOPY_END``
1198 :slave payload: ``u64``
1200 Master advises that postcopy migration has now completed. The slave
1201 must disable the userfaultfd. The response is an acknowledgement
1204 When ``VHOST_USER_PROTOCOL_F_PAGEFAULT`` is supported, this message
1205 is sent at the end of the migration, after
1206 ``VHOST_USER_POSTCOPY_LISTEN`` was previously sent.
1208 The value returned is an error indication; 0 is success.
1210 ``VHOST_USER_GET_INFLIGHT_FD``
1212 :equivalent ioctl: N/A
1213 :master payload: inflight description
1215 When ``VHOST_USER_PROTOCOL_F_INFLIGHT_SHMFD`` protocol feature has
1216 been successfully negotiated, this message is submitted by master to
1217 get a shared buffer from slave. The shared buffer will be used to
1218 track inflight I/O by slave. QEMU should retrieve a new one when vm
1221 ``VHOST_USER_SET_INFLIGHT_FD``
1223 :equivalent ioctl: N/A
1224 :master payload: inflight description
1226 When ``VHOST_USER_PROTOCOL_F_INFLIGHT_SHMFD`` protocol feature has
1227 been successfully negotiated, this message is submitted by master to
1228 send the shared inflight buffer back to slave so that slave could
1229 get inflight I/O after a crash or restart.
1231 ``VHOST_USER_GPU_SET_SOCKET``
1233 :equivalent ioctl: N/A
1234 :master payload: N/A
1236 Sets the GPU protocol socket file descriptor, which is passed as
1237 ancillary data. The GPU protocol is used to inform the master of
1238 rendering state and updates. See vhost-user-gpu.rst for details.
1240 ``VHOST_USER_RESET_DEVICE``
1242 :equivalent ioctl: N/A
1243 :master payload: N/A
1246 Ask the vhost user backend to disable all rings and reset all
1247 internal device state to the initial state, ready to be
1248 reinitialized. The backend retains ownership of the device
1249 throughout the reset operation.
1251 Only valid if the ``VHOST_USER_PROTOCOL_F_RESET_DEVICE`` protocol
1252 feature is set by the backend.
1254 ``VHOST_USER_VRING_KICK``
1256 :equivalent ioctl: N/A
1257 :slave payload: vring state description
1258 :master payload: N/A
1260 When the ``VHOST_USER_PROTOCOL_F_INBAND_NOTIFICATIONS`` protocol
1261 feature has been successfully negotiated, this message may be
1262 submitted by the master to indicate that a buffer was added to
1263 the vring instead of signalling it using the vring's kick file
1264 descriptor or having the slave rely on polling.
1266 The state.num field is currently reserved and must be set to 0.
1268 ``VHOST_USER_GET_MAX_MEM_SLOTS``
1270 :equivalent ioctl: N/A
1273 When the ``VHOST_USER_PROTOCOL_F_CONFIGURE_MEM_SLOTS`` protocol
1274 feature has been successfully negotiated, this message is submitted
1275 by master to the slave. The slave should return the message with a
1276 u64 payload containing the maximum number of memory slots for
1277 QEMU to expose to the guest. The value returned by the backend
1278 will be capped at the maximum number of ram slots which can be
1279 supported by the target platform.
1281 ``VHOST_USER_ADD_MEM_REG``
1283 :equivalent ioctl: N/A
1284 :slave payload: memory region
1286 When the ``VHOST_USER_PROTOCOL_F_CONFIGURE_MEM_SLOTS`` protocol
1287 feature has been successfully negotiated, this message is submitted
1288 by the master to the slave. The message payload contains a memory
1289 region descriptor struct, describing a region of guest memory which
1290 the slave device must map in. When the
1291 ``VHOST_USER_PROTOCOL_F_CONFIGURE_MEM_SLOTS`` protocol feature has
1292 been successfully negotiated, along with the
1293 ``VHOST_USER_REM_MEM_REG`` message, this message is used to set and
1294 update the memory tables of the slave device.
1296 ``VHOST_USER_REM_MEM_REG``
1298 :equivalent ioctl: N/A
1299 :slave payload: memory region
1301 When the ``VHOST_USER_PROTOCOL_F_CONFIGURE_MEM_SLOTS`` protocol
1302 feature has been successfully negotiated, this message is submitted
1303 by the master to the slave. The message payload contains a memory
1304 region descriptor struct, describing a region of guest memory which
1305 the slave device must unmap. When the
1306 ``VHOST_USER_PROTOCOL_F_CONFIGURE_MEM_SLOTS`` protocol feature has
1307 been successfully negotiated, along with the
1308 ``VHOST_USER_ADD_MEM_REG`` message, this message is used to set and
1309 update the memory tables of the slave device.
1311 ``VHOST_USER_SET_STATUS``
1313 :equivalent ioctl: VHOST_VDPA_SET_STATUS
1315 :master payload: ``u64``
1317 When the ``VHOST_USER_PROTOCOL_F_STATUS`` protocol feature has been
1318 successfully negotiated, this message is submitted by the master to
1319 notify the backend with updated device status as defined in the Virtio
1322 ``VHOST_USER_GET_STATUS``
1324 :equivalent ioctl: VHOST_VDPA_GET_STATUS
1325 :slave payload: ``u64``
1326 :master payload: N/A
1328 When the ``VHOST_USER_PROTOCOL_F_STATUS`` protocol feature has been
1329 successfully negotiated, this message is submitted by the master to
1330 query the backend for its device status as defined in the Virtio
1337 ``VHOST_USER_SLAVE_IOTLB_MSG``
1339 :equivalent ioctl: N/A (equivalent to ``VHOST_IOTLB_MSG`` message type)
1340 :slave payload: ``struct vhost_iotlb_msg``
1341 :master payload: N/A
1343 Send IOTLB messages with ``struct vhost_iotlb_msg`` as payload.
1344 Slave sends such requests to notify of an IOTLB miss, or an IOTLB
1345 access failure. If ``VHOST_USER_PROTOCOL_F_REPLY_ACK`` is
1346 negotiated, and slave set the ``VHOST_USER_NEED_REPLY`` flag, master
1347 must respond with zero when operation is successfully completed, or
1348 non-zero otherwise. This request should be send only when
1349 ``VIRTIO_F_IOMMU_PLATFORM`` feature has been successfully
1352 ``VHOST_USER_SLAVE_CONFIG_CHANGE_MSG``
1354 :equivalent ioctl: N/A
1356 :master payload: N/A
1358 When ``VHOST_USER_PROTOCOL_F_CONFIG`` is negotiated, vhost-user
1359 slave sends such messages to notify that the virtio device's
1360 configuration space has changed, for those host devices which can
1361 support such feature, host driver can send ``VHOST_USER_GET_CONFIG``
1362 message to slave to get the latest content. If
1363 ``VHOST_USER_PROTOCOL_F_REPLY_ACK`` is negotiated, and slave set the
1364 ``VHOST_USER_NEED_REPLY`` flag, master must respond with zero when
1365 operation is successfully completed, or non-zero otherwise.
1367 ``VHOST_USER_SLAVE_VRING_HOST_NOTIFIER_MSG``
1369 :equivalent ioctl: N/A
1370 :slave payload: vring area description
1371 :master payload: N/A
1373 Sets host notifier for a specified queue. The queue index is
1374 contained in the ``u64`` field of the vring area description. The
1375 host notifier is described by the file descriptor (typically it's a
1376 VFIO device fd) which is passed as ancillary data and the size
1377 (which is mmap size and should be the same as host page size) and
1378 offset (which is mmap offset) carried in the vring area
1379 description. QEMU can mmap the file descriptor based on the size and
1380 offset to get a memory range. Registering a host notifier means
1381 mapping this memory range to the VM as the specified queue's notify
1382 MMIO region. Slave sends this request to tell QEMU to de-register
1383 the existing notifier if any and register the new notifier if the
1384 request is sent with a file descriptor.
1386 This request should be sent only when
1387 ``VHOST_USER_PROTOCOL_F_HOST_NOTIFIER`` protocol feature has been
1388 successfully negotiated.
1390 ``VHOST_USER_SLAVE_VRING_CALL``
1392 :equivalent ioctl: N/A
1393 :slave payload: vring state description
1394 :master payload: N/A
1396 When the ``VHOST_USER_PROTOCOL_F_INBAND_NOTIFICATIONS`` protocol
1397 feature has been successfully negotiated, this message may be
1398 submitted by the slave to indicate that a buffer was used from
1399 the vring instead of signalling this using the vring's call file
1400 descriptor or having the master relying on polling.
1402 The state.num field is currently reserved and must be set to 0.
1404 ``VHOST_USER_SLAVE_VRING_ERR``
1406 :equivalent ioctl: N/A
1407 :slave payload: vring state description
1408 :master payload: N/A
1410 When the ``VHOST_USER_PROTOCOL_F_INBAND_NOTIFICATIONS`` protocol
1411 feature has been successfully negotiated, this message may be
1412 submitted by the slave to indicate that an error occurred on the
1413 specific vring, instead of signalling the error file descriptor
1414 set by the master via ``VHOST_USER_SET_VRING_ERR``.
1416 The state.num field is currently reserved and must be set to 0.
1420 VHOST_USER_PROTOCOL_F_REPLY_ACK
1421 -------------------------------
1423 The original vhost-user specification only demands replies for certain
1424 commands. This differs from the vhost protocol implementation where
1425 commands are sent over an ``ioctl()`` call and block until the client
1428 With this protocol extension negotiated, the sender (QEMU) can set the
1429 ``need_reply`` [Bit 3] flag to any command. This indicates that the
1430 client MUST respond with a Payload ``VhostUserMsg`` indicating success
1431 or failure. The payload should be set to zero on success or non-zero
1432 on failure, unless the message already has an explicit reply body.
1434 The response payload gives QEMU a deterministic indication of the result
1435 of the command. Today, QEMU is expected to terminate the main vhost-user
1436 loop upon receiving such errors. In future, qemu could be taught to be more
1437 resilient for selective requests.
1439 For the message types that already solicit a reply from the client,
1440 the presence of ``VHOST_USER_PROTOCOL_F_REPLY_ACK`` or need_reply bit
1441 being set brings no behavioural change. (See the Communication_
1442 section for details.)
1444 .. _backend_conventions:
1446 Backend program conventions
1447 ===========================
1449 vhost-user backends can provide various devices & services and may
1450 need to be configured manually depending on the use case. However, it
1451 is a good idea to follow the conventions listed here when
1452 possible. Users, QEMU or libvirt, can then rely on some common
1453 behaviour to avoid heterogeneous configuration and management of the
1454 backend programs and facilitate interoperability.
1456 Each backend installed on a host system should come with at least one
1457 JSON file that conforms to the vhost-user.json schema. Each file
1458 informs the management applications about the backend type, and binary
1459 location. In addition, it defines rules for management apps for
1460 picking the highest priority backend when multiple match the search
1461 criteria (see ``@VhostUserBackend`` documentation in the schema file).
1463 If the backend is not capable of enabling a requested feature on the
1464 host (such as 3D acceleration with virgl), or the initialization
1465 failed, the backend should fail to start early and exit with a status
1466 != 0. It may also print a message to stderr for further details.
1468 The backend program must not daemonize itself, but it may be
1469 daemonized by the management layer. It may also have a restricted
1470 access to the system.
1472 File descriptors 0, 1 and 2 will exist, and have regular
1473 stdin/stdout/stderr usage (they may have been redirected to /dev/null
1474 by the management layer, or to a log handler).
1476 The backend program must end (as quickly and cleanly as possible) when
1477 the SIGTERM signal is received. Eventually, it may receive SIGKILL by
1478 the management layer after a few seconds.
1480 The following command line options have an expected behaviour. They
1481 are mandatory, unless explicitly said differently:
1485 This option specify the location of the vhost-user Unix domain socket.
1486 It is incompatible with --fd.
1490 When this argument is given, the backend program is started with the
1491 vhost-user socket as file descriptor FDNUM. It is incompatible with
1494 --print-capabilities
1496 Output to stdout the backend capabilities in JSON format, and then
1497 exit successfully. Other options and arguments should be ignored, and
1498 the backend program should not perform its normal function. The
1499 capabilities can be reported dynamically depending on the host
1502 The JSON output is described in the ``vhost-user.json`` schema, by
1503 ```@VHostUserBackendCapabilities``. Example:
1518 Command line options:
1522 Specify the linux input device.
1528 Do no request exclusive access to the input device.
1535 Command line options:
1539 Specify the GPU DRM render node.
1545 Enable virgl rendering support.
1552 Command line options:
1556 Specify block device or file path.