4 Copyright (c) 2014 Virtual Open Systems Sarl.
6 This work is licensed under the terms of the GNU GPL, version 2 or later.
7 See the COPYING file in the top-level directory.
10 This protocol is aiming to complement the ioctl interface used to control the
11 vhost implementation in the Linux kernel. It implements the control plane needed
12 to establish virtqueue sharing with a user space process on the same host. It
13 uses communication over a Unix domain socket to share file descriptors in the
14 ancillary data of the message.
16 The protocol defines 2 sides of the communication, master and slave. Master is
17 the application that shares its virtqueues, in our case QEMU. Slave is the
18 consumer of the virtqueues.
20 In the current implementation QEMU is the Master, and the Slave is intended to
21 be a software Ethernet switch running in user space, such as Snabbswitch.
23 Master and slave can be either a client (i.e. connecting) or server (listening)
24 in the socket communication.
29 Note that all numbers are in the machine native byte order. A vhost-user message
30 consists of 3 header fields and a payload:
32 ------------------------------------
33 | request | flags | size | payload |
34 ------------------------------------
36 * Request: 32-bit type of the request
37 * Flags: 32-bit bit field:
38 - Lower 2 bits are the version (currently 0x01)
39 - Bit 2 is the reply flag - needs to be sent on each reply from the slave
40 * Size - 32-bit size of the payload
43 Depending on the request type, payload can be:
45 * A single 64-bit integer
50 u64: a 64-bit unsigned integer
52 * A vring state description
60 * A vring address description
61 --------------------------------------------------------------
62 | index | flags | size | descriptor | used | available | log |
63 --------------------------------------------------------------
65 Index: a 32-bit vring index
66 Flags: a 32-bit vring flags
67 Descriptor: a 64-bit user address of the vring descriptor table
68 Used: a 64-bit user address of the vring used ring
69 Available: a 64-bit user address of the vring available ring
70 Log: a 64-bit guest address for logging
72 * Memory regions description
73 ---------------------------------------------------
74 | num regions | padding | region0 | ... | region7 |
75 ---------------------------------------------------
77 Num regions: a 32-bit number of regions
81 -----------------------------------------------------
82 | guest address | size | user address | mmap offset |
83 -----------------------------------------------------
85 Guest address: a 64-bit guest address of the region
87 User address: a 64-bit user address
88 mmap offset: 64-bit offset where region starts in the mapped memory
90 In QEMU the vhost-user message is implemented with the following struct:
92 typedef struct VhostUserMsg {
93 VhostUserRequest request;
98 struct vhost_vring_state state;
99 struct vhost_vring_addr addr;
100 VhostUserMemory memory;
102 } QEMU_PACKED VhostUserMsg;
107 The protocol for vhost-user is based on the existing implementation of vhost
108 for the Linux Kernel. Most messages that can be sent via the Unix domain socket
109 implementing vhost-user have an equivalent ioctl to the kernel implementation.
111 The communication consists of master sending message requests and slave sending
112 message replies. Most of the requests don't require replies. Here is a list of
116 * VHOST_GET_PROTOCOL_FEATURES
117 * VHOST_GET_VRING_BASE
118 * VHOST_SET_LOG_BASE (if VHOST_USER_PROTOCOL_F_LOG_SHMFD)
120 There are several messages that the master sends with file descriptors passed
121 in the ancillary data:
123 * VHOST_SET_MEM_TABLE
124 * VHOST_SET_LOG_BASE (if VHOST_USER_PROTOCOL_F_LOG_SHMFD)
126 * VHOST_SET_VRING_KICK
127 * VHOST_SET_VRING_CALL
128 * VHOST_SET_VRING_ERR
130 If Master is unable to send the full message or receives a wrong reply it will
131 close the connection. An optional reconnection mechanism can be implemented.
133 Any protocol extensions are gated by protocol feature bits,
134 which allows full backwards compatibility on both master
136 As older slaves don't support negotiating protocol features,
137 a feature bit was dedicated for this purpose:
138 #define VHOST_USER_F_PROTOCOL_FEATURES 30
140 Multiple queue support
141 ----------------------
143 Multiple queue is treated as a protocol extension, hence the slave has to
144 implement protocol features first. The multiple queues feature is supported
145 only when the protocol feature VHOST_USER_PROTOCOL_F_MQ (bit 0) is set.
147 The max number of queues the slave supports can be queried with message
148 VHOST_USER_GET_PROTOCOL_FEATURES. Master should stop when the number of
149 requested queues is bigger than that.
151 As all queues share one connection, the master uses a unique index for each
152 queue in the sent message to identify a specified queue. One queue pair
153 is enabled initially. More queues are enabled dynamically, by sending
154 message VHOST_USER_SET_VRING_ENABLE.
159 During live migration, the master may need to track the modifications
160 the slave makes to the memory mapped regions. The client should mark
161 the dirty pages in a log. Once it complies to this logging, it may
162 declare the VHOST_F_LOG_ALL vhost feature.
164 All the modifications to memory pointed by vring "descriptor" should
165 be marked. Modifications to "used" vring should be marked if
166 VHOST_VRING_F_LOG is part of ring's features.
168 Dirty pages are of size:
169 #define VHOST_LOG_PAGE 0x1000
171 The log memory fd is provided in the ancillary data of
172 VHOST_USER_SET_LOG_BASE message when the slave has
173 VHOST_USER_PROTOCOL_F_LOG_SHMFD protocol feature.
175 The size of the log may be computed by using all the known guest
176 addresses. The log covers from address 0 to the maximum of guest
177 regions. In pseudo-code, to mark page at "addr" as dirty:
179 page = addr / VHOST_LOG_PAGE
180 log[page / 8] |= 1 << page % 8
182 Use atomic operations, as the log may be concurrently manipulated.
184 VHOST_USER_SET_LOG_FD is an optional message with an eventfd in
185 ancillary data, it may be used to inform the master that the log has
188 Once the source has finished migration, VHOST_USER_RESET_OWNER message
189 will be sent by the source. No further update must be done before the
190 destination takes over with new regions & rings.
195 #define VHOST_USER_PROTOCOL_F_MQ 0
196 #define VHOST_USER_PROTOCOL_F_LOG_SHMFD 1
197 #define VHOST_USER_PROTOCOL_F_RARP 2
202 * VHOST_USER_GET_FEATURES
205 Equivalent ioctl: VHOST_GET_FEATURES
209 Get from the underlying vhost implementation the features bitmask.
210 Feature bit VHOST_USER_F_PROTOCOL_FEATURES signals slave support for
211 VHOST_USER_GET_PROTOCOL_FEATURES and VHOST_USER_SET_PROTOCOL_FEATURES.
213 * VHOST_USER_SET_FEATURES
216 Ioctl: VHOST_SET_FEATURES
219 Enable features in the underlying vhost implementation using a bitmask.
220 Feature bit VHOST_USER_F_PROTOCOL_FEATURES signals slave support for
221 VHOST_USER_GET_PROTOCOL_FEATURES and VHOST_USER_SET_PROTOCOL_FEATURES.
223 * VHOST_USER_GET_PROTOCOL_FEATURES
226 Equivalent ioctl: VHOST_GET_FEATURES
230 Get the protocol feature bitmask from the underlying vhost implementation.
231 Only legal if feature bit VHOST_USER_F_PROTOCOL_FEATURES is present in
232 VHOST_USER_GET_FEATURES.
233 Note: slave that reported VHOST_USER_F_PROTOCOL_FEATURES must support
234 this message even before VHOST_USER_SET_FEATURES was called.
236 * VHOST_USER_SET_PROTOCOL_FEATURES
239 Ioctl: VHOST_SET_FEATURES
242 Enable protocol features in the underlying vhost implementation.
243 Only legal if feature bit VHOST_USER_F_PROTOCOL_FEATURES is present in
244 VHOST_USER_GET_FEATURES.
245 Note: slave that reported VHOST_USER_F_PROTOCOL_FEATURES must support
246 this message even before VHOST_USER_SET_FEATURES was called.
248 * VHOST_USER_SET_OWNER
251 Equivalent ioctl: VHOST_SET_OWNER
254 Issued when a new connection is established. It sets the current Master
255 as an owner of the session. This can be used on the Slave as a
256 "session start" flag.
258 * VHOST_USER_RESET_DEVICE
261 Equivalent ioctl: VHOST_RESET_DEVICE
264 Issued when a new connection is about to be closed. The Master will no
265 longer own this connection (and will usually close it).
267 * VHOST_USER_SET_MEM_TABLE
270 Equivalent ioctl: VHOST_SET_MEM_TABLE
271 Master payload: memory regions description
273 Sets the memory map regions on the slave so it can translate the vring
274 addresses. In the ancillary data there is an array of file descriptors
275 for each memory mapped region. The size and ordering of the fds matches
276 the number and ordering of memory regions.
278 * VHOST_USER_SET_LOG_BASE
281 Equivalent ioctl: VHOST_SET_LOG_BASE
285 Sets the logging base address.
287 * VHOST_USER_SET_LOG_FD
290 Equivalent ioctl: VHOST_SET_LOG_FD
293 Sets the logging file descriptor, which is passed as ancillary data.
295 * VHOST_USER_SET_VRING_NUM
298 Equivalent ioctl: VHOST_SET_VRING_NUM
299 Master payload: vring state description
301 Sets the number of vrings for this owner.
303 * VHOST_USER_SET_VRING_ADDR
306 Equivalent ioctl: VHOST_SET_VRING_ADDR
307 Master payload: vring address description
310 Sets the addresses of the different aspects of the vring.
312 * VHOST_USER_SET_VRING_BASE
315 Equivalent ioctl: VHOST_SET_VRING_BASE
316 Master payload: vring state description
318 Sets the base offset in the available vring.
320 * VHOST_USER_GET_VRING_BASE
323 Equivalent ioctl: VHOST_USER_GET_VRING_BASE
324 Master payload: vring state description
325 Slave payload: vring state description
327 Get the available vring base offset.
329 * VHOST_USER_SET_VRING_KICK
332 Equivalent ioctl: VHOST_SET_VRING_KICK
335 Set the event file descriptor for adding buffers to the vring. It
336 is passed in the ancillary data.
337 Bits (0-7) of the payload contain the vring index. Bit 8 is the
338 invalid FD flag. This flag is set when there is no file descriptor
339 in the ancillary data. This signals that polling should be used
340 instead of waiting for a kick.
342 * VHOST_USER_SET_VRING_CALL
345 Equivalent ioctl: VHOST_SET_VRING_CALL
348 Set the event file descriptor to signal when buffers are used. It
349 is passed in the ancillary data.
350 Bits (0-7) of the payload contain the vring index. Bit 8 is the
351 invalid FD flag. This flag is set when there is no file descriptor
352 in the ancillary data. This signals that polling will be used
353 instead of waiting for the call.
355 * VHOST_USER_SET_VRING_ERR
358 Equivalent ioctl: VHOST_SET_VRING_ERR
361 Set the event file descriptor to signal when error occurs. It
362 is passed in the ancillary data.
363 Bits (0-7) of the payload contain the vring index. Bit 8 is the
364 invalid FD flag. This flag is set when there is no file descriptor
365 in the ancillary data.
367 * VHOST_USER_GET_QUEUE_NUM
370 Equivalent ioctl: N/A
374 Query how many queues the backend supports. This request should be
375 sent only when VHOST_USER_PROTOCOL_F_MQ is set in quried protocol
376 features by VHOST_USER_GET_PROTOCOL_FEATURES.
378 * VHOST_USER_SET_VRING_ENABLE
381 Equivalent ioctl: N/A
382 Master payload: vring state description
384 Signal slave to enable or disable corresponding vring.
386 * VHOST_USER_SEND_RARP
389 Equivalent ioctl: N/A
392 Ask vhost user backend to broadcast a fake RARP to notify the migration
393 is terminated for guest that does not support GUEST_ANNOUNCE.
394 Only legal if feature bit VHOST_USER_F_PROTOCOL_FEATURES is present in
395 VHOST_USER_GET_FEATURES and protocol feature bit VHOST_USER_PROTOCOL_F_RARP
396 is present in VHOST_USER_GET_PROTOCOL_FEATURES.
397 The first 6 bytes of the payload contain the mac address of the guest to
398 allow the vhost user backend to construct and broadcast the fake RARP.