=encoding utf8

=head1 NAME

virt-rescue - Run a rescue shell on a virtual machine

=head1 SYNOPSIS

 virt-rescue [--options] -d domname

 virt-rescue [--options] -a disk.img [-a disk.img ...]

 virt-rescue --suggest (-d domname | -a disk.img ...)

Old style:

 virt-rescue [--options] domname

 virt-rescue [--options] disk.img [disk.img ...]

=head1 WARNING

You must I<not> use C<virt-rescue> on live virtual machines.  Doing so
will probably result in disk corruption in the VM.  C<virt-rescue>
tries to stop you from doing this, but doesn't catch all cases.

However if you use the I<--ro> (read only) option, then you can attach
a shell to a live virtual machine.  The results might be strange or
inconsistent at times but you won't get disk corruption.

=head1 DESCRIPTION

virt-rescue is like a Rescue CD, but for virtual machines, and without
the need for a CD.  virt-rescue gives you a rescue shell and some
simple recovery tools which you can use to examine or rescue a virtual
machine or disk image.

You can run virt-rescue on any virtual machine known to libvirt, or
directly on disk image(s):

 virt-rescue -d GuestName

 virt-rescue --ro -a /path/to/disk.img

 virt-rescue -a /dev/sdc

For live VMs you I<must> use the --ro option.

When you run virt-rescue on a virtual machine or disk image, you are
placed in an interactive bash shell where you can use many ordinary
Linux commands.  What you see in C</> (C</bin>, C</lib> etc) is the
rescue appliance.  You must mount the virtual machine's filesystems by
hand.  There is an empty directory called C</sysroot> where you can
mount filesystems.

You can get virt-rescue to suggest mount commands for you by using the
I<--suggest> option (in another terminal):

 $ virt-rescue --suggest -d Fedora15
 Inspecting the virtual machine or disk image ...
 
 This disk contains one or more operating systems.  You can use these
 mount commands in virt-rescue (at the ><rescue> prompt) to mount the
 filesystems.
 
 # /dev/vg_f15x32/lv_root is the root of a linux operating system
 # type: linux, distro: fedora, version: 15.0
 # Fedora release 15 (Lovelock)
 
 mount /dev/vg_f15x32/lv_root /sysroot/
 mount /dev/vda1 /sysroot/boot
 mount --bind /dev /sysroot/dev
 mount --bind /dev/pts /sysroot/dev/pts
 mount --bind /proc /sysroot/proc
 mount --bind /sys /sysroot/sys

Another way is to list the logical volumes (with L<lvs(8)>) and
partitions (with L<parted(8)>) and mount them by hand:

 ><rescue> lvs
 LV      VG        Attr   LSize   Origin Snap%  Move Log Copy%  Convert
 lv_root vg_f15x32 -wi-a-   8.83G
 lv_swap vg_f15x32 -wi-a- 992.00M
 ><rescue> mount /dev/vg_f15x32/lv_root /sysroot
 ><rescue> mount /dev/vda1 /sysroot/boot
 ><rescue> ls /sysroot

Another command to list available filesystems is
L<virt-filesystems(1)>.

To run commands in a Linux guest (for example, grub), you should
chroot into the /sysroot directory first:

 ><rescue> chroot /sysroot

=head2 NOTES

Virt-rescue can be used on I<any> disk image file or device, not just
a virtual machine.  For example you can use it on a blank file if you
want to partition that file (although we would recommend using
L<guestfish(1)> instead as it is more suitable for this purpose).  You
can even use virt-rescue on things like SD cards.

You can get virt-rescue to give you scratch disk(s) to play with.
This is useful for testing out Linux utilities (see I<--scratch>).

Virt-rescue does not require root.  You only need to run it as root if
you need root to open the disk image.

This tool is just designed for quick interactive hacking on a virtual
machine.  For more structured access to a virtual machine disk image,
you should use L<guestfs(3)>.  To get a structured shell that you can
use to make scripted changes to guests, use L<guestfish(1)>.

=head1 OPTIONS

=over 4

=item B<--help>

Display brief help.

=item B<-a> file

=item B<--add> file

Add I<file> which should be a disk image from a virtual machine.  If
the virtual machine has multiple block devices, you must supply all of
them with separate I<-a> options.

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

=item B<--append kernelopts>

Pass additional options to the rescue kernel.

=item B<-c> URI

=item B<--connect> URI

If using libvirt, connect to the given I<URI>.  If omitted, then we
connect to the default libvirt hypervisor.

If you specify guest block devices directly (I<-a>), then libvirt is
not used at all.

=item B<-d> guest

=item B<--domain> guest

Add all the disks from the named libvirt guest.  Domain UUIDs can be
used instead of names.

=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.

For example:

 virt-rescue --format=raw -a disk.img

forces raw format (no auto-detection) for C<disk.img>.

 virt-rescue --format=raw -a disk.img --format -a another.img

forces raw format (no auto-detection) for C<disk.img> and reverts to
auto-detection for C<another.img>.

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).

=item B<-m MB>

=item B<--memsize MB>

Change the amount of memory allocated to the rescue system.  The
default is set by libguestfs and is small but adequate for running
system tools.  The occasional program might need more memory.  The
parameter is specified in megabytes.

=item B<--network>

Enable QEMU user networking in the guest.  See L</NETWORK>.

=item B<-r>

=item B<--ro>

Open the image read-only.

The option must always be used if the disk image or virtual machine
might be running, and is generally recommended in cases where you
don't need write access to the disk.

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

=item B<--scratch>

=item B<--scratch=N>

The I<--scratch> option adds a large scratch disk to the rescue
appliance.  I<--scratch=N> adds C<N> scratch disks.  The scratch
disk(s) are deleted automatically when virt-rescue exits.

You can also mix I<-a>, I<-d> and I<--scratch> options.  The scratch
disk(s) are added to the appliance in the order they appear on the
command line.

=item B<--selinux>

Enable SELinux in the rescue appliance.  You should read
L<guestfs(3)/SELINUX> before using this option.

=item B<--smp> N

Enable N E<ge> 2 virtual CPUs in the rescue appliance.

=item B<--suggest>

Inspect the disk image and suggest what mount commands should be used
to mount the disks.  You should use the I<--suggest> option in a
second terminal, then paste the commands into another virt-rescue.

This option implies I<--ro> and is safe to use even if the guest is up
or if another virt-rescue is running.

=item B<-v>

=item B<--verbose>

Enable verbose messages for debugging.

=item B<-V>

=item B<--version>

Display version number and exit.

=item B<-w>

=item B<--rw>

This changes the I<-a> and I<-d> 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>

Enable tracing of libguestfs API calls.

=back

=head1 OLD-STYLE COMMAND LINE ARGUMENTS

Previous versions of virt-rescue allowed you to write either:

 virt-rescue disk.img [disk.img ...]

or

 virt-rescue guestname

whereas in this version you should use I<-a> or I<-d> respectively
to avoid the confusing case where a disk image might have the same
name as a guest.

For compatibility the old style is still supported.

=head1 NETWORK

Adding the I<--network> option enables QEMU user networking
in the rescue appliance.  There are some differences between
user networking and ordinary networking:

=over 4

=item ping does not work

Because the ICMP ECHO_REQUEST protocol generally requires root in
order to send the ping packets, and because virt-rescue must be able
to run as non-root, QEMU user networking is not able to emulate the
L<ping(8)> command.  The ping command will appear to resolve addresses
but will not be able to send or receive any packets.  This does not
mean that the network is not working.

=item cannot receive connections

QEMU user networking cannot receive incoming connections.

=item making TCP connections

The virt-rescue appliance needs to be small and so does not include
many network tools.  In particular there is no L<telnet(1)> command.
You can make TCP connections from the shell using the magical
C</dev/tcp/E<lt>hostnameE<gt>/E<lt>portE<gt>> syntax:

 exec 3<>/dev/tcp/redhat.com/80
 echo "GET /" >&3
 cat <&3

See L<bash(1)> for more details.

=back

=head1 ENVIRONMENT VARIABLES

Several environment variables affect virt-rescue.  See
L<guestfs(3)/ENVIRONMENT VARIABLES> for the complete list.

=head1 SHELL QUOTING

Libvirt guest names can contain arbitrary characters, some of which
have meaning to the shell such as C<#> and space.  You may need to
quote or escape these characters on the command line.  See the shell
manual page L<sh(1)> for details.

=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 SEE ALSO

L<guestfs(3)>,
L<guestfish(1)>,
L<virt-cat(1)>,
L<virt-edit(1)>,
L<virt-filesystems(1)>,
L<http://libguestfs.org/>.

=head1 AUTHOR

Richard W.M. Jones L<http://people.redhat.com/~rjones/>

=head1 COPYRIGHT

Copyright (C) 2009-2012 Red Hat Inc.

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.