xref: /linux/Documentation/filesystems/virtiofs.rst (revision 9e56ff53b4115875667760445b028357848b4748)
1.. SPDX-License-Identifier: GPL-2.0
2
3.. _virtiofs_index:
4
5===================================================
6virtiofs: virtio-fs host<->guest shared file system
7===================================================
8
9- Copyright (C) 2019 Red Hat, Inc.
10
11Introduction
12============
13The virtiofs file system for Linux implements a driver for the paravirtualized
14VIRTIO "virtio-fs" device for guest<->host file system sharing.  It allows a
15guest to mount a directory that has been exported on the host.
16
17Guests often require access to files residing on the host or remote systems.
18Use cases include making files available to new guests during installation,
19booting from a root file system located on the host, persistent storage for
20stateless or ephemeral guests, and sharing a directory between guests.
21
22Although it is possible to use existing network file systems for some of these
23tasks, they require configuration steps that are hard to automate and they
24expose the storage network to the guest.  The virtio-fs device was designed to
25solve these problems by providing file system access without networking.
26
27Furthermore the virtio-fs device takes advantage of the co-location of the
28guest and host to increase performance and provide semantics that are not
29possible with network file systems.
30
31Usage
32=====
33Mount file system with tag ``myfs`` on ``/mnt``:
34
35.. code-block:: sh
36
37  guest# mount -t virtiofs myfs /mnt
38
39Please see https://virtio-fs.gitlab.io/ for details on how to configure QEMU
40and the virtiofsd daemon.
41
42Mount options
43-------------
44
45virtiofs supports general VFS mount options, for example, remount,
46ro, rw, context, etc. It also supports FUSE mount options.
47
48atime behavior
49^^^^^^^^^^^^^^
50
51The atime-related mount options, for example, noatime, strictatime,
52are ignored. The atime behavior for virtiofs is the same as the
53underlying filesystem of the directory that has been exported
54on the host.
55
56Internals
57=========
58Since the virtio-fs device uses the FUSE protocol for file system requests, the
59virtiofs file system for Linux is integrated closely with the FUSE file system
60client.  The guest acts as the FUSE client while the host acts as the FUSE
61server.  The /dev/fuse interface between the kernel and userspace is replaced
62with the virtio-fs device interface.
63
64FUSE requests are placed into a virtqueue and processed by the host.  The
65response portion of the buffer is filled in by the host and the guest handles
66the request completion.
67
68Mapping /dev/fuse to virtqueues requires solving differences in semantics
69between /dev/fuse and virtqueues.  Each time the /dev/fuse device is read, the
70FUSE client may choose which request to transfer, making it possible to
71prioritize certain requests over others.  Virtqueues have queue semantics and
72it is not possible to change the order of requests that have been enqueued.
73This is especially important if the virtqueue becomes full since it is then
74impossible to add high priority requests.  In order to address this difference,
75the virtio-fs device uses a "hiprio" virtqueue specifically for requests that
76have priority over normal requests.
77