path: root/fuse/guestmount.pod

=encoding utf8

=head1 NAME

guestmount - Mount a guest filesystem on the host using FUSE and libguestfs

=head1 SYNOPSIS

 guestmount [--options] -a disk.img -m device [--ro] mountpoint

 guestmount [--options] -a disk.img -i [--ro] mountpoint

 guestmount [--options] -d Guest -i [--ro] mountpoint

=head1 WARNING

You must I<not> use C<guestmount> in read-write mode on live virtual
machines.  If you do this, you risk disk corruption in the VM.

=head1 DESCRIPTION

The guestmount program can be used to mount virtual machine
filesystems and other disk images on the host.  It uses libguestfs for
access to the guest filesystem, and FUSE (the "filesystem in
userspace") to make it appear as a mountable device.

Along with other options, you have to give at least one device (I<-a>
option) or libvirt domain (I<-d> option), and at least one mountpoint
(I<-m> option) or use the I<-i> inspection option or the I<--live>
option.  How this works is better explained in the L<guestfish(1)>
manual page, or by looking at the examples below.

FUSE lets you mount filesystems as non-root.  The mountpoint must be
owned by you, and the filesystem will not be visible to any other
users unless you make certain global configuration changes to
C</etc/fuse.conf>.  To unmount the filesystem, use the C<fusermount -u>
command.

=head1 EXAMPLES

For a typical Windows guest which has its main filesystem on the
first partition:

 guestmount -a windows.img -m /dev/sda1 --ro /mnt

For a typical Linux guest which has a /boot filesystem on the first
partition, and the root filesystem on a logical volume:

 guestmount -a linux.img -m /dev/VG/LV -m /dev/sda1:/boot --ro /mnt

To get libguestfs to detect guest mountpoints for you:

 guestmount -a guest.img -i --ro /mnt

For a libvirt guest called "Guest" you could do:

 guestmount -d Guest -i --ro /mnt

If you don't know what filesystems are contained in a guest or
disk image, use L<virt-filesystems(1)> first:

 virt-filesystems -d MyGuest

If you want to trace the libguestfs calls but without excessive
debugging information, we recommend:

 guestmount [...] --trace /mnt

If you want to debug the program, we recommend:

 guestmount [...] --trace --verbose /mnt

=head1 NOTES

=head2 Other users cannot see the filesystem by default

If you mount a filesystem as one user (eg. root), then other users
will not be able to see it by default.  The fix is to add the FUSE
C<allow_other> option when mounting:

 sudo guestmount [...] -o allow_other /mnt

=head2 Enabling FUSE

On some distros, you may need to add yourself to a special group
(eg. C<fuse>) before you can use any FUSE filesystem.  This is
necessary on Debian and derivatives.

On other distros, no special group is required.  It is not necessary
on Fedora or Red Hat Enterprise Linux.

=head2 fusermount error: "Device or resource busy"

You can see this error when another process on the system jumps into
the mountpoint you have just created, holding it open and preventing
you from unmounting it.  The usual culprits are various GUI "indexing"
programs.

The popular workaround for this problem is to retry the
C<fusermount -u> command a few times until it works.  Unfortunately
this isn't a reliable fix if (for example) the mounted filesystem is
particularly large and the intruding program particularly persistent.

 timeout=10
 
 count=$timeout
 while ! fusermount -u $mountpoint && [ $count -gt 0 ]; do
     sleep 1
     ((count--))
 done
 if [ $count -eq 0 ]; then
     echo "$0: fusermount failed after $timeout seconds"
     exit 1
 fi

A proper fix is to use a private mountpoint by creating a new mount
namespace using the Linux-specific L<clone(2)>/L<unshare(2)> flag
C<CLONE_NEWNS>.  Unfortunately at the moment this requires root and we
would also probably need to add it as a feature to guestmount.

=head2 Race conditions possible when shutting down the connection

When C<fusermount -u> exits, guestmount may still be running and
cleaning up the mountpoint.  The disk image will not be fully
finalized.

This means that scripts like the following have a nasty race
condition:

 guestmount -a disk.img -i /mnt
 # copy things into /mnt
 fusermount -u /mnt
 # immediately try to use 'disk.img' ** UNSAFE **

The solution is to use the I<--pid-file> option to write the
guestmount PID to a file, then after fusermount spin waiting for this
PID to exit.

 guestmount -a disk.img -i --pid-file guestmount.pid /mnt
 
 # ...
 # ...
 
 # Save the PID of guestmount *before* calling fusermount.
 pid="$(cat guestmount.pid)"
 
 timeout=10
 
 # fusermount retry code, see above
 # ...
 # ...
 
 count=$timeout
 while kill -0 "$pid" 2>/dev/null && [ $count -gt 0 ]; do
     sleep 1
     ((count--))
 done
 if [ $count -eq 0 ]; then
     echo "$0: wait for guestmount to exit failed after $timeout seconds"
     exit 1
 fi
 
 # Now it is safe to use the disk image.

Note that if you use the C<guestfs_mount_local> API directly (see
L<guestfs(3)/MOUNT LOCAL>) then it is much easier to write a safe,
race-free program.

=head1 OPTIONS

=over 4

=item B<-a image>

=item B<--add image>

Add a block device or virtual machine image.

The format of the disk image is auto-detected.  To override this and
force a particular format use the I<--format=..> option.

=item B<-c URI>

=item B<--connect URI>

When used in conjunction with the I<-d> option, this specifies
the libvirt URI to use.  The default is to use the default libvirt
connection.

=item B<-d libvirt-domain>

=item B<--domain libvirt-domain>

Add disks from the named libvirt domain.  If the I<--ro> option is
also used, then any libvirt domain can be used.  However in write
mode, only libvirt domains which are shut down can be named here.

Domain UUIDs can be used instead of names.

=item B<--dir-cache-timeout N>

Set the readdir cache timeout to I<N> seconds, the default being 60
seconds.  The readdir cache [actually, there are several
semi-independent caches] is populated after a readdir(2) call with the
stat and extended attributes of the files in the directory, in
anticipation that they will be requested soon after.

There is also a different attribute cache implemented by FUSE
(see the FUSE option I<-o attr_timeout>), but the FUSE cache
does not anticipate future requests, only cache existing ones.

=item B<--echo-keys>

When prompting for keys and passphrases, guestfish normally turns
echoing off so you cannot see what you are typing.  If you are not
worried about Tempest attacks and there is no one else in the room
you can specify this flag to see what you are typing.

=item B<--format=raw|qcow2|..>

=item B<--format>

The default for the I<-a> option is to auto-detect the format of the
disk image.  Using this forces the disk format for I<-a> options which
follow on the command line.  Using I<--format> with no argument
switches back to auto-detection for subsequent I<-a> options.

If you have untrusted raw-format guest disk images, you should use
this option to specify the disk format.  This avoids a possible
security problem with malicious guests (CVE-2010-3851).  See also
L<guestfs(3)/guestfs_add_drive_opts>.

=item B<--fuse-help>

Display help on special FUSE options (see I<-o> below).

=item B<--help>

Display brief help and exit.

=item B<-i>

=item B<--inspector>

Using L<virt-inspector(1)> code, inspect the disks looking for
an operating system and mount filesystems as they would be
mounted on the real virtual machine.

=item B<--keys-from-stdin>

Read key or passphrase parameters from stdin.  The default is
to try to read passphrases from the user by opening C</dev/tty>.

=item B<--live>

Connect to a live virtual machine.
(Experimental, see L<guestfs(3)/ATTACHING TO RUNNING DAEMONS>).

=item B<-m dev[:mountpoint[:options]]>

=item B<--mount dev[:mountpoint[:options]]>

Mount the named partition or logical volume on the given mountpoint
B<in the guest> (this has nothing to do with mountpoints in the host).

If the mountpoint is omitted, it defaults to C</>.  You have to mount
something on C</>.

The third (and rarely used) part of the mount parameter is the list of
mount options used to mount the underlying filesystem.  If this is not
given, then the mount options are either the empty string or C<ro>
(the latter if the I<--ro> flag is used).  By specifying the mount
options, you override this default choice.  Probably the only time you
would use this is to enable ACLs and/or extended attributes if the
filesystem can support them:

 -m /dev/sda1:/:acl,user_xattr

=item B<-n>

=item B<--no-sync>

By default, we attempt to sync the guest disk when the FUSE mountpoint
is unmounted.  If you specify this option, then we don't attempt to
sync the disk.  See the discussion of autosync in the L<guestfs(3)>
manpage.

=item B<-o option>

=item B<--option option>

Pass extra options to FUSE.

To get a list of all the extra options supported by FUSE, use the
command below.  Note that only the FUSE I<-o> options can be passed,
and only some of them are a good idea.

 guestmount --fuse-help

Some potentially useful FUSE options:

=over 4

=item B<-o allow_other>

Allow other users to see the filesystem.

=item B<-o attr_timeout=N>

Enable attribute caching by FUSE, and set the timeout to I<N> seconds.

=item B<-o kernel_cache>

Allow the kernel to cache files (reduces the number of reads
that have to go through the L<guestfs(3)> API).  This is generally
a good idea if you can afford the extra memory usage.

=item B<-o uid=N> B<-o gid=N>

Use these options to map all UIDs and GIDs inside the guest filesystem
to the chosen values.

=back

=item B<--pid-file filename>

Write the PID of the guestmount worker process to C<filename>.

=item B<-r>

=item B<--ro>

Add devices and mount everything read-only.  Also disallow writes and
make the disk appear read-only to FUSE.

This is highly recommended if you are not going to edit the guest
disk.  If the guest is running and this option is I<not> supplied,
then there is a strong risk of disk corruption in the guest.  We try
to prevent this from happening, but it is not always possible.

See also L<guestfish(1)/OPENING DISKS FOR READ AND WRITE>.

=item B<--selinux>

Enable SELinux support for the guest.

=item B<-v>

=item B<--verbose>

Enable verbose messages from underlying libguestfs.

=item B<-V>

=item B<--version>

Display the program version and exit.

=item B<-w>

=item B<--rw>

This changes the I<-a>, I<-d> and I<-m> options so that disks are
added and mounts are done read-write.

See L<guestfish(1)/OPENING DISKS FOR READ AND WRITE>.

=item B<-x>

=item B<--trace>

Trace libguestfs calls and entry into each FUSE function.

This also stops the daemon from forking into the background.

=back

=head1 FILES

=over 4

=item $HOME/.libguestfs-tools.rc

=item /etc/libguestfs-tools.conf

This configuration file controls the default read-only or read-write
mode (I<--ro> or I<--rw>).

See L<guestfish(1)/OPENING DISKS FOR READ AND WRITE>.

=back

=head1 EXIT STATUS

This program returns 0 if successful, or non-zero if there was an
error.

=head1 SEE ALSO

L<guestfish(1)>,
L<virt-inspector(1)>,
L<virt-cat(1)>,
L<virt-edit(1)>,
L<virt-tar(1)>,
L<guestfs(3)/MOUNT LOCAL>,
L<http://libguestfs.org/>,
L<http://fuse.sf.net/>.

=head1 AUTHORS

Richard W.M. Jones (C<rjones at redhat dot com>)

=head1 COPYRIGHT

Copyright (C) 2009-2012 Red Hat Inc.
L<http://libguestfs.org/>

This program is free software; you can redistribute it and/or modify
it under the terms of the GNU General Public License as published by
the Free Software Foundation; either version 2 of the License, or
(at your option) any later version.

This program is distributed in the hope that it will be useful,
but WITHOUT ANY WARRANTY; without even the implied warranty of
MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
GNU General Public License for more details.

You should have received a copy of the GNU General Public License
along with this program; if not, write to the Free Software
Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.