Training courses

Kernel and Embedded Linux

Bootlin training courses

Embedded Linux, kernel,
Yocto Project, Buildroot, real-time,
graphics, boot time, debugging...

Bootlin logo

Elixir Cross Referencer

.\" Copyright (c) 1980, 1989, 1991, 1993
.\"	The Regents of the University of California.
.\" Copyright (c) 2005, 2006 Csaba Henk
.\" All rights reserved.
.\"
.\" Copyright (c) 2019 The FreeBSD Foundation
.\"
.\" Portions of this documentation were written by BFF Storage Systems under
.\" sponsorship from the FreeBSD Foundation.
.\"
.\" Redistribution and use in source and binary forms, with or without
.\" modification, are permitted provided that the following conditions
.\" are met:
.\" 1. Redistributions of source code must retain the above copyright
.\"    notice, this list of conditions and the following disclaimer.
.\" 2. Redistributions in binary form must reproduce the above copyright
.\"    notice, this list of conditions and the following disclaimer in the
.\"    documentation and/or other materials provided with the distribution.
.\" 3. Neither the name of the University nor the names of its contributors
.\"    may be used to endorse or promote products derived from this software
.\"    without specific prior written permission.
.\"
.\" THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
.\" ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
.\" IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
.\" ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
.\" FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
.\" DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
.\" OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
.\" HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
.\" LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
.\" OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
.\" SUCH DAMAGE.
.\"
.\" $FreeBSD$
.\"
.Dd July 31, 2019
.Dt MOUNT_FUSEFS 8
.Os
.Sh NAME
.Nm mount_fusefs
.Nd mount a Fuse file system daemon
.Sh SYNOPSIS
.Nm
.Op Fl A
.Op Fl S
.Op Fl v
.Op Fl D Ar fuse_daemon
.Op Fl O Ar daemon_opts
.Op Fl s Ar special
.Op Fl m Ar node
.Op Fl h
.Op Fl V
.Op Fl o Ar option ...
.Ar special node
.Op Ar fuse_daemon ...
.Sh DESCRIPTION
Basic usage is to start a fuse daemon on the given
.Ar special
file.
In practice, the daemon is assigned a
.Ar special
file automatically, which can then be identified via
.Xr fstat 1 .
That special file can then be mounted by
.Nm .
.Pp
However, the procedure of spawning a daemon will usually be automated
so that it is performed by
.Nm .
If the command invoking a given
.Ar fuse_daemon
is appended to the list of arguments,
.Nm
will call the
.Ar fuse_daemon
via that command.
In that way the
.Ar fuse_daemon
will be instructed to attach itself to
.Ar special .
From that on mounting goes as in the simple case. (See
.Sx DAEMON MOUNTS . )
.Pp
The
.Ar special
argument will normally be treated as the path of the special file to mount.
.Pp
However, if
.Pa auto
is passed as
.Ar special ,
then
.Nm
will look for a suitable free fuse device by itself.
.Pp
Finally, if
.Ar special
is an integer it will be interpreted as the number
of the file descriptor of an already open fuse device
(used when the Fuse library invokes
.Nm .
See
.Sx DAEMON MOUNTS ) .
.Pp
The options are as follows:
.Bl -tag -width indent
.It Fl A , Ic --reject-allow_other
Prohibit the
.Cm allow_other
mount flag.
Intended for use in scripts and the
.Xr sudoers 5
file.
.It Fl S , Ic --safe
Run in safe mode (i.e., reject invoking a filesystem daemon).
.It Fl v
Be verbose.
.It Fl D , Ic --daemon Ar daemon
Call the specified
.Ar daemon .
.It Fl O , Ic --daemon_opts Ar opts
Add
.Ar opts
to the daemon's command line.
.It Fl s , Ic --special Ar special
Use
.Ar special
as special.
.It Fl m , Ic --mountpath Ar node
Mount on
.Ar node .
.It Fl h , Ic --help
Show help.
.It Fl V , Ic --version
Show version information.
.It Fl o
Mount options are specified via
.Fl o .
The following options are available (and also their negated versions,
by prefixing them with
.Dq no ) :
.Bl -tag -width indent
.It Cm allow_other
Do not apply
.Sx STRICT ACCESS POLICY .
Only root can use this option.
.It Cm async
I/O to the file system may be done asynchronously.
Writes may be delayed and/or reordered.
.It Cm default_permissions
Enable traditional (file mode based) permission checking in kernel.
.It Cm intr
Allow signals to interrupt operations that are blocked waiting for a reply from the server.
When this option is in use, system calls may fail with
.Er EINTR
whenever a signal is received.
.It Cm max_read Ns = Ns Ar n
Limit size of read requests to
.Ar n .
.It Cm neglect_shares
Do not refuse unmounting if there are secondary mounts.
.It Cm private
Refuse shared mounting of the daemon.
This is the default behaviour, to allow sharing, expicitly use
.Fl o Cm noprivate .
.It Cm push_symlinks_in
Prefix absolute symlinks with the mountpoint.
.It Cm subtype Ns = Ns Ar fsname
Suffix
.Ar fsname
to the file system name as reported by
.Xr statfs 2 .
This option can be used to identify the file system implemented by
.Ar fuse_daemon .
.El
.El
.Pp
Besides the above mount options, there is a set of pseudo-mount options which
are supported by the Fuse library.
One can list these by passing
.Fl h
to a Fuse daemon.
Most of these options only have effect on the behavior of the daemon (that is,
their scope is limited to userspace).
However, there are some which do require in-kernel support.
Currently the options supported by the kernel are:
.Bl -tag -width indent
.It Cm direct_io
Bypass the buffer cache system.
.It Cm kernel_cache
By default cached buffers of a given file are flushed at each
.Xr open 2 .
This option disables this behaviour.
.El
.Sh DAEMON MOUNTS
Usually users do not need to use
.Nm
directly, as the Fuse library enables Fuse daemons to invoke
.Nm .
That is,
.Pp
.Dl fuse_daemon device mountpoint
.Pp
has the same effect as
.Pp
.Dl mount_fusefs auto mountpoint fuse_daemon
.Pp
This is the recommended usage when you want basic usage
(eg, run the daemon at a low privilege level but mount it as root).
.Sh STRICT ACCESS POLICY
The strict access policy for Fuse filesystems lets one use the filesystem
only if the filesystem daemon has the same credentials (uid, real uid, gid,
real gid) as the user.
.Pp
This is applied for Fuse mounts by default and only root can mount without
the strict access policy (i.e., the
.Cm allow_other
mount option).
.Pp
This is to shield users from the daemon
.Dq spying
on their I/O activities.
.Pp
Users might opt to willingly relax strict access policy (as far as they
are concerned) by doing their own secondary mount (See
.Sx SHARED MOUNTS ) .
.Sh SHARED MOUNTS
A Fuse daemon can be shared (i.e., mounted multiple times).
When doing the first (primary) mount, the spawner and the mounter of the daemon
must have the same uid, or the mounter should be the superuser.
.Pp
After the primary mount is in place, secondary mounts can be done by anyone
unless this feature is disabled by
.Cm private .
The behaviour of a secondary mount is analogous to that of symbolic
links: they redirect all filesystem operations to the primary mount.
.Pp
Doing a secondary mount is like signing an agreement: by this action, the mounter
agrees that the Fuse daemon can trace her I/O activities.
From then on she is not banned from using the filesystem
(either via her own mount or via the primary mount), regardless whether
.Cm allow_other
is used or not.
.Pp
The device name of a secondary mount is the device name of the corresponding
primary mount, followed by a '#' character and the index of the secondary
mount; e.g.,
.Pa /dev/fuse0#3 .
.Sh SECURITY
System administrators might want to use a custom mount policy (ie., one going
beyond the
.Va vfs.usermount
sysctl).
The primary tool for such purposes is
.Xr sudo 8 .
However, given that
.Nm
is capable of invoking an arbitrary program, one must be careful when doing this.
.Nm
is designed in a way such that it makes that easy.
For this purpose, there are options which disable certain risky features
.Fl ( S
and
.Fl A ) ,
and command line parsing is done in a flexible way: mixing options and
non-options is allowed, but processing them stops at the third non-option
argument (after the first two have been utilized as device and mountpoint).
The rest of the command line specifies the daemon and its arguments.
(Alternatively, the daemon, the special and the mount path can be
specified using the respective options.) Note that
.Nm
ignores the environment variable
.Ev POSIXLY_CORRECT
and always behaves as described.
.Pp
In general, to be as scripting /
.Xr sudoers 5
friendly as possible, no information has a fixed
position in the command line, but once a given piece of information is
provided, subsequent arguments/options cannot override it (with the
exception of some non-critical ones).
.Sh ENVIRONMENT
.Bl -tag -width ".Ev MOUNT_FUSEFS_SAFE"
.It Ev MOUNT_FUSEFS_SAFE
This has the same effect as the
.Fl S
option.
.It Ev MOUNT_FUSEFS_VERBOSE
This has the same effect as the
.Fl v
option.
.It Ev MOUNT_FUSEFS_IGNORE_UNKNOWN
If set,
.Nm
will ignore uknown mount options.
.It Ev MOUNT_FUSEFS_CALL_BY_LIB
Adjust behavior to the needs of the FUSE library.
Currently it effects help output.
.El
.Pp
Although the following variables do not have any effect on
.Nm
itself, they affect the behaviour of fuse daemons:
.Bl -tag -width ".Ev FUSE_DEV_NAME"
.It Ev FUSE_DEV_NAME
Device to attach.
If not set, the multiplexer path
.Ar /dev/fuse
is used.
.It Ev FUSE_DEV_FD
File desciptor of an opened Fuse device to use.
Overrides
.Ev FUSE_DEV_NAME .
.It Ev FUSE_NO_MOUNT
If set, the library will not attempt to mount the filesystem, even
if a mountpoint argument is supplied.
.El
.Sh FILES
.Bl -tag -width /dev/fuse
.It Pa /dev/fuse
Fuse device with which the kernel and Fuse daemons can communicate.
.It Pa /dev/fuse
The multiplexer path.
An
.Xr open 2
performed on it automatically is passed to a free Fuse device by the kernel
(which might be created just for this puprose).
.El
.Sh EXAMPLES
Mount the example filesystem in the Fuse distribution (from its directory):
either
.Pp
.Dl ./fusexmp /mnt/fuse
.Pp
or
.Pp
.Dl mount_fusefs auto /mnt/fuse ./fusexmp
.Pp
Doing the same in two steps, using
.Pa /dev/fuse0 :
.Pp
.Dl FUSE_DEV_NAME=/dev/fuse ./fusexmp &&
.Dl mount_fusefs /dev/fuse /mnt/fuse
.Pp
A script wrapper for fusexmp which ensures that
.Nm
does not call any external utility and also provides a hacky
(non race-free) automatic device selection:
.Pp
.Dl #!/bin/sh -e
.Pp
.Dl FUSE_DEV_NAME=/dev/fuse fusexmp
.Dl mount_fusefs -S /dev/fuse /mnt/fuse \(lq$@\(rq
.Sh SEE ALSO
.Xr fstat 1 ,
.Xr mount 8 ,
.Xr sudo 8 ,
.Xr umount 8
.Sh HISTORY
.Nm
was written as the part of the
.Fx
implementation of the Fuse userspace filesystem framework (see
.Lk https://github.com/libfuse/libfuse )
and first appeared in the
.Pa sysutils/fusefs-kmod
port, supporting
.Fx 6.0 .
It was added to the base system in
.Fx 10.0 .
.Sh CAVEATS
This user interface is
.Fx
specific.
Secondary mounts should be unmounted via their device name.
If an attempt is made to unmount them via their filesystem root path,
the unmount request will be forwarded to the primary mount path.
In general, unmounting by device name is less error-prone than by mount path
(although the latter will also work under normal circumstances).
.Pp
If the daemon is specified via the
.Fl D
and
.Fl O
options, it will be invoked via
.Xr system 3 ,
and the daemon's command line will also have an
.Dq &
control operator appended, so that we do not have to wait for its termination.
You should use a simple command line when invoking the daemon via these options.
.Sh BUGS
.Ar special
is treated as a multiplexer if and only if it is literally the same as
.Pa auto
or
.Pa /dev/fuse .
Other paths which are equivalent with
.Pa /dev/fuse
(eg.,
.Pa /../dev/fuse )
are not.