.\" Man page generated from reStructuredText.
.
.TH "VMDEBOOTSTRAP" "8" "March 31, 2019" "1.11
" "vmdebootstrap"
.SH NAME
vmdebootstrap \- install basic Debian system into virtual disk image
.
.nr rst2man-indent-level 0
.
.de1 rstReportMargin
\\$1 \\n[an-margin]
level \\n[rst2man-indent-level]
level margin: \\n[rst2man-indent\\n[rst2man-indent-level]]
-
\\n[rst2man-indent0]
\\n[rst2man-indent1]
\\n[rst2man-indent2]
..
.de1 INDENT
.\" .rstReportMargin pre:
. RS \\$1
. nr rst2man-indent\\n[rst2man-indent-level] \\n[an-margin]
. nr rst2man-indent-level +1
.\" .rstReportMargin post:
..
.de UNINDENT
. RE
.\" indent \\n[an-margin]
.\" old: \\n[rst2man-indent\\n[rst2man-indent-level]]
.nr rst2man-indent-level -1
.\" new: \\n[rst2man-indent\\n[rst2man-indent-level]]
.in \\n[rst2man-indent\\n[rst2man-indent-level]]u
..
.SH PURPOSE
.sp
vmdebootstrap is a helper to install basic Debian system into virtual
disk image. It wraps \fBdebootstrap\fP\&. You need to run \fBvmdebootstrap\fP
as root. If the \fB\-\-verbose\fP option is not used, no output will be
sent to the command line. If the \fB\-\-log\fP option is not used, no
output will be sent to any log files either.
.sp
To use the image, you probably want to create a virtual machine using
your preferred virtualization technology, such as \fBkvm\fP or
\fBqemu\fP\&. Configure the virtual machine to use the image you\(aqve
created. Then start the virtual machine and log into it via its console
to configure it. The image has an empty root password and will not have
networking configured by default. Set the root password before you
configure networking.
.SH SYNOPSIS
.INDENT 0.0
.INDENT 3.5
.sp
.nf
.ft C
$ sudo vmdebootstrap \-\-image=FILE \-\-size=SIZE [\-\-mirror=URL] [\-\-distribution=NAME]
.ft P
.fi
.UNINDENT
.UNINDENT
.SS Options
.INDENT 0.0
.INDENT 3.5
.INDENT 0.0
.TP
.BI \-\-output\fB= FILE
write output to FILE, instead of standard output
.TP
.B \-\-verbose
report what is going on
.TP
.B \-\-no\-verbose
opposite of \-\-verbose
.TP
.BI \-\-image\fB= FILE
put created disk image in FILE
.TP
.BI \-\-size\fB= SIZE
create a disk image of size SIZE (1000000000)
in bytes. Suffixes k,K,M,G,T are supported,
see qemu\-img(1) for more detail.
.TP
.BI \-\-tarball\fB= FILE
tar up the disk\(aqs contents in FILE
.TP
.BI \-\-mirror\fB= URL
use MIRROR as package source (\fI\%http://httpredir.debian.org/debian/\fP)
.TP
.BI \-\-arch\fB= ARCH
architecture to use (amd64) \-\-\- if using an
architecture which the host system cannot execute,
ensure the \fB\-\-foreign\fP option is also used.
.TP
.BI \-\-distribution\fB= NAME
release to use (defaults to stable). The release
needs to be a valid Debian or Ubuntu release name
or codename.
.TP
.BI \-\-debootstrapopts\fB= OPTS
Supply options and arguments to \fBdebootstrap\fP,
separated by spaces.
e.g. \fB\-\-debootstrapopts="variant=buildd no\-check\-gpg components=main,contrib"\fP\&.
See \fBdebootstrap (1)\fP for more information. This
option replaces the \fB\-\-variant\fP support in
previous versions.
.TP
.BI \-\-debootstrap\-scripts\fB= DIR
set the directory containing debootstrap scripts.
.TP
.BI \-\-package\fB= PACKAGE
install PACKAGE onto system
.TP
.BI \-\-custom\-package\fB= DEB
install package in DEB file onto system (not
from mirror) \- all dependencies must be available
in the specified distribution.
.TP
.B \-\-no\-kernel
do not install a linux package
.TP
.BI \-\-kernel\-package\fB= PACKAGE
If \fB\-\-no\-kernel\fP is not used and the auto\-selection
of the \fBlinux\-image\-586\fP or \fBlinux\-image\-armmp\fP
or \fBlinux\-image\-$ARCH\fP package is not suitable,
the kernel PACKAGE name can be specified explicitly.
.TP
.B \-\-enable\-dhcp
enable DHCP on eth0
.TP
.BI \-\-root\-password\fB= PASSWORD
set root password
.TP
.B \-\-lock\-root\-password
lock root account so they cannot login?
.TP
.BI \-\-customize\fB= SCRIPT
run SCRIPT after setting up system. If the script
does not exist in the current working directory,
\fB/usr/share/vmdebootstrap/examples/\fP will be
checked as a fallback. The script needs to be
executable and is passed the root directory of the
debootstrap and the image name as the only arguments.
Use chroot if you need to execute binaries within
the chroot created by debootstrap.
.TP
.BI \-\-hostname\fB= HOSTNAME
set name to HOSTNAME (debian)
.TP
.BI \-\-user\fB= USERSTRING
create USER with PASSWORD. The USERSTRING needs to
be of the format: USER/PASSSWORD.
.TP
.BI \-\-owner\fB= OWNER
change the owner of the final image from root to
the specified user.
.TP
.B \-\-serial\-console
configure image to use a serial console (Wheezy only)
.TP
.B \-\-serial\-console\-command
(Wheezy only.) Set the command to manage the serial
console which will be appended to \fB/etc/inittab\fP\&.
Default is \fB/sbin/getty \e\-L ttyS0 115200 vt100\fP,
resulting in a line:
.INDENT 7.0
.INDENT 3.5
.sp
.nf
.ft C
"S0:23:respawn:/sbin/getty \e\-L ttyS0 115200 vt100"
.ft P
.fi
.UNINDENT
.UNINDENT
.TP
.B \-\-sudo
install sudo, and if user is created, add them to
sudo group
.TP
.BI \-\-bootsize\fB= BOOTSIZE
If specified, create a /boot partition of the given
size within the image. Debootstrapping will fail
if this is too small for the selected kernel
package and upgrading such a kernel package is
likely to need two or three times the space of the
installed kernel.
.TP
.BI \-\-boottype\fB= FSTYPE
Filesystem to use for the /boot partition. (default ext2)
.TP
.BI \-\-bootflag\fB= FLAG
Flag to set on the first partition. (default none)
.TP
.BI \-\-bootoffset\fB= SIZE
Space to leave at start of the image for bootloader
.TP
.BI \-\-roottype\fB= FSTYPE
Filesystem to use for the / (root) partition. (default ext4)
.TP
.BI \-\-part\-type\fB= PART\-TYPE
Partition type to use for this image. (default msdos)
.TP
.BI \-\-swap\fB= SWAPSIZE
If specified, create a swap partition of the given
size within the image. Debootstrapping will fail
if this results in a root partition which is too
small for the selected packages. The minimum swap
space is 256MB as the default memory allocation
of QEMU is 128MB. A default 1GB image is not likely
to have enough space for a swap partition as well.
.TP
.BI \-\-foreign\fB= PATH
Path to the binfmt_handler to enable foreign support
in debootstrap. e.g. \fB/usr/bin/qemu\-arm\-static\fP
Note: foreign debootstraps may take a significant
amount of time to complete and debootstrap will
retry five times if packages fail to install by default.
.TP
.B \-\-use\-uefi
Setup image for UEFI boot
.TP
.B \-\-no\-use\-uefi
opposite of \-\-use\-uefi
.TP
.BI \-\-esp\-size\fB= SIZE
Size of EFI System Partition \- requires use\-uefi
.TP
.B \-\-extlinux
install extlinux (deprecated: default will change in a
future release to use grub)
.TP
.B \-\-no\-extlinux
Skip installation of extlinux. Needs grub, a customize script
or alternative bootloader to make the image bootable.
extlinux is deprecated and this will become the default
in a future release.
.TP
.B \-\-mbr
Run install\-mbr (default if extlinux used)
.TP
.B \-\-no\-mbr
opposite of \-\-mbr
.TP
.BI \-\-squash\fB= DIRECTORY
Run mksquashfs against the rootfs using xz
compression \-\-\- requires \fBsquashfs\-tools\fP to be installed.
The squashfs and other files needed to use the squashfs
to make a bootable system will be put into the specified directory.
The directory will contain a \fBfilesystem.squashfs\fP
as well as the top level contents of the \fBboot/\fP
directory. (If using UEFI, the \fBboot/efi\fP directory
as well.) By default, \fBmksquashfs\fP is allowed to use
all processors which may result in high load. squashfs
can also have issues with large root filesystems. These
errors can result in truncated files. This is a known
bug in squashfs. \fBvmdebootstrap\fP will fail if the
squashed filesystem is less than 1MB.
.TP
.B \-\-configure\-apt
Use the specified mirror and distribution to create a
suitable apt source inside the VM. Can be useful if
debootstrap fails to create it automatically.
.TP
.B \-\-apt\-mirror
Use the specified mirror inside the image instead of the
mirror used to build the image. This is useful if you have
a local mirror to make building the image quicker but
the image needs to run even if that mirror is not available.
Requires \fB\-\-configure\-apt\fP
.TP
.B \-\-grub
Disable extlinux installation and configure grub2 instead.
grub2 will be added to the list of packages to install.
update\-grub will be called once the debootstrap is
complete and grub\-install will be called in the image.
.TP
.B \-\-no\-acpid
Disable installation of acpid if not required, otherwise
acpid will be installed if \fB\-\-foreign\fP is not used.
.TP
.B \-\-sparse
Skip optimizing image for compression and keep a sparse image.
.TP
.B \-\-no\-sparse
opposite of \-\-sparse
.TP
.B \-\-pkglist
Output a list of package names installed inside the image.
Useful if you need to track the relevant source packages
used inside the image for licence compliance.
.TP
.B \-\-dry\-run
Do not build, just test that the options are valid.
.TP
.B \-\-no\-update\-initramfs
Skip the call to \fBupdate\-initramfs\fP for reasons of
speed or practicality.
.TP
.B \-\-convert\-qcow2
Convert the final raw image to qcow2 format.
.TP
.B \-\-systemd\-networkd
Use Predictable Network Interface Names
.TP
.B \-\-no\-systemd\-networkd
Do not use Predictable Network Interface Names using
systemd\-networkd.
.UNINDENT
.UNINDENT
.UNINDENT
.SH CONFIGURATION FILES AND SETTINGS
.INDENT 0.0
.INDENT 3.5
.INDENT 0.0
.TP
.B \-\-dump\-config
write out the entire current configuration
.TP
.B \-\-no\-default\-configs
clear list of configuration files to read
.TP
.BI \-\-config\fB= FILE
add FILE to config files
.UNINDENT
.UNINDENT
.UNINDENT
.SH LOGGING
.INDENT 0.0
.INDENT 3.5
.INDENT 0.0
.TP
.BI \-\-log\fB= FILE
write log entries to FILE (default is to not write
log files at all); use "syslog" to log to system
log, or "none" to disable logging.
.TP
.BI \-\-log\-level\fB= LEVEL
log at LEVEL, one of debug, info, warning, error,
critical, fatal (default: debug).
.TP
.BI \-\-log\-max\fB= SIZE
rotate logs larger than SIZE, zero for never (default: 0)
.TP
.BI \-\-log\-keep\fB= N
keep last N logs (10)
.TP
.BI \-\-log\-mode\fB= MODE
set permissions of new log files to MODE (octal;  default 0600)
.UNINDENT
.UNINDENT
.UNINDENT
.SH PERFORMANCE
.INDENT 0.0
.INDENT 3.5
.INDENT 0.0
.TP
.BI \-\-dump\-memory\-profile\fB= METHOD
make memory profiling dumps using METHOD, which is one
of: none, simple, meliae, or heapy (default: simple)
.TP
.BI \-\-memory\-dump\-interval\fB= SECONDS
make memory profiling dumps at least SECONDS apart
.UNINDENT
.UNINDENT
.UNINDENT
.SH NETWORKING
.SS Wheezy support
.sp
The \fB\-\-enable\-networking\fP option uses the \fB/etc/network/interfaces.d/\fP
source directory, with the default settings for \fBlo\fP and \fBeth0\fP
being added to \fB/etc/network/interfaces.d/setup\fP\&. Other networking
configuration can be specified using a customisation script.
Localhost settings would be:
.INDENT 0.0
.INDENT 3.5
.sp
.nf
.ft C
auto lo
iface lo inet loopback
.ft P
.fi
.UNINDENT
.UNINDENT
.sp
If \fB\-\-enable\-dhcp\fP is specified, these settings are also included
into \fB/etc/network/interfaces.d/setup\fP:
.INDENT 0.0
.INDENT 3.5
.sp
.nf
.ft C
auto eth0
iface eth0 inet dhcp
.ft P
.fi
.UNINDENT
.UNINDENT
.sp
In addition, wheezy images do not boot if the roottype is specified as
the default of \fBext4\fP, so \fBvmdebootstrap\fP will fail if a \fB\-\-roottype\fP
is not specified or is specified as \fBext4\fP\&.
.SS Jessie and later
.sp
In addition, \fBsystemd\fP in jessie or later introduces
\fI\%PredictableNetworkInterfaceNames\fP which are enabled using the
\fBsystemd\-networkd\fP service. If this option is disabled, traditional
interface names (like \fBeth0\fP) will be used and the predictable names
masked using \fBudev\fP\&. Implementing the mask requires updating the
initramfs, so the \fB\-\-update\-initramfs\fP option must not be disabled.
.sp
If DHCP is also enabled, the following configuration is used:
.INDENT 0.0
.INDENT 3.5
.sp
.nf
.ft C
/etc/systemd/network/99\-dhcp.network
.ft P
.fi
.UNINDENT
.UNINDENT
.sp
\fBsystemd\fP will use the first available match, so this can be
overridden by putting another file into place using the customisation
scripts, using a lower sorting filename.
.SS Stretch and later
.sp
There is no need to use the \fB\-\-enable\-dhcp\fP option when using
\fBsystemd\fP for networking with stretch or sid. \fBsystemd\-resolved\fP is
enabled instead if \fBsystemd\-networkd\fP is specified. (\fB\-\-enable\-dhcp\fP
would simply add an unused entry to \fB/etc/network/interfaces\fP for
\fBeth0\fP\&.)
.INDENT 0.0
.INDENT 3.5
.sp
.nf
.ft C
[Match]
Name=en*

[Network]
DHCP=yes
.ft P
.fi
.UNINDENT
.UNINDENT
.SH BOOTLOADERS
.sp
Unless the \fB\-\-no\-extlinux\fP or \fB\-\-grub\fP options are specified, the
image will use \fBextlinux\fP as a boot loader. \fBbootsize\fP is not
recommended when using \fBextlinux\fP \-\-\- use \fBgrub\fP instead.
.sp
\fBNOTE:\fP
.INDENT 0.0
.INDENT 3.5
Unlike grub, extlinux support requires the installation of
packages outside the image which are used to install the extlinux
bootloader inside the image. extlinux support also involves the
use of \fBsync\fP which can cause issues on systems with multiple
filesystems mounted, particularly over a network or when building
multiple images simultaneously. Therefore, \fBextlinux\fP is
\fBdeprecated\fP in vmdebootstrap. The default will change in a future
release and \fBextlinux\fP support may be dropped once Stretch is
released.
.UNINDENT
.UNINDENT
.SS extlinux support issues with ext4
.sp
VMs using ext4 may not boot when using extlinux \- unless the build is
performed on Jessie. Builds using ext2 and ext3 work normally.
.sp
\fBIMPORTANT:\fP
.INDENT 0.0
.INDENT 3.5
This problem depends on the \fBexternal\fP distribution,
\fBnot\fP the distribution you are trying to build. When building on
Jessie, \fBextlinux\fP succeeds but when building on Stretch or Sid,
\fBextlinux\fP fails to make a bootable system if the filesystem of
that system is \fBext4\fP\&. ext2 and ext3 work.
.UNINDENT
.UNINDENT
.sp
Version 1.6 of vmdebootstrap adds a warning but allows the build to
proceed (to allow for the bug to be fixed). Sadly, downgrading the
version of extlinux to the version in Jessie does not fix the problem
when building on stretch or sid. Hence, vmdebootstrap can only output
a warning.
.sp
\fBSEE ALSO:\fP
.INDENT 0.0
.INDENT 3.5
\fI\%http://bugs.debian.org/cgi\-bin/bugreport.cgi?bug=833057\fP
.UNINDENT
.UNINDENT
.SS Versions of grub2 in wheezy
.sp
Grub2 in wheezy can fail to install in the VM, at which point
\fBvmdebootstrap\fP will fall back to \fBextlinux\fP\&. It may still be
possible to complete the installation of \fBgrub2\fP after booting the
VM as the problem may be related to the need to use loopback devices
during the \fBgrub\-install\fP operation. Details of the error will appear
in the vmdebootstrap log file, if enabled with the \fB\-\-log\fP option.
.sp
\fBNOTE:\fP
.INDENT 0.0
.INDENT 3.5
\fBgrub\-legacy\fP is not supported.
.UNINDENT
.UNINDENT
.sp
\fBvmdebootstrap\fP also supports \fBEFI\fP\&. See \fI\%UEFI\fP\&.
.sp
Use \fB\-\-use\-uefi\fP to use \fBgrub\-efi\fP instead of \fBgrub\-pc\fP\&. If the
default 5MB is not enough space, use the \fB\-\-esp\-size\fP option to
specify a different size for the EFI partition. Registered firmware is
not supported as it would need to be done after boot. If the system you
are creating is for more than just a VM or live image, you will likely
need a larger ESP, up to 500MB.
.SS UEFI
.sp
UEFI support requires Grub and \fBvmdebootstrap\fP contains a configuration
table of the UEFI components required for supported architectures.
.sp
There are issues with running UEFI with QEMU on some architectures and
a customisation script is available for amd64:
.INDENT 0.0
.INDENT 3.5
.sp
.nf
.ft C
# vmdebootstrap \-\-verbose \-\-image jessie\-uefi.img \-\-grub  \-\-use\-uefi \e
  \-\-customize ./examples/qemu\-efi\-bochs\-drm.sh
.ft P
.fi
.UNINDENT
.UNINDENT
.sp
\fBvmdebootstrap\fP supports UEFI for images and for squashfs but the necessary
behaviour is different. With an image, an ESP vfat partition is created.
With squashfs, the EFI files will be copied into an \fBefi/\fP directory
in the squashfs output directory instead.
.sp
There is EFI firmware available to use with QEMU when testing images built
using the UEFI support, but this software is in Debian non\-free due to patent
concerns. If you choose to install \fBovmf\fP to test UEFI builds, a
secondary change is also needed to symlink the provided \fBOVMF.fd\fP to
the file required by QEMU: \fBbios\-256k.bin\fP and then tell QEMU about
the location of this file with the \-L option:
.INDENT 0.0
.INDENT 3.5
.sp
.nf
.ft C
$ qemu\-system\-x86_64 \-L /usr/share/ovmf/ \-machine accel=kvm \e
 \-m 4096 \-smp 2 \-drive format=raw,file=test.img
.ft P
.fi
.UNINDENT
.UNINDENT
.sp
To test the image, also consider using the \fBqemu\-wrapper.sh\fP:
.INDENT 0.0
.INDENT 3.5
.sp
.nf
.ft C
$ /usr/share/vmdebootstrap/qemu\-wrapper.sh jessie\-uefi.img amd64 /usr/share/ovmf/
.ft P
.fi
.UNINDENT
.UNINDENT
.SS UBoot
.sp
UBoot needs manual configuration via the customisation hook scripts,
typically support requires adding \fBu\-boot\fP using \fB\-\-package\fP and then
copying or manipulating the relevant \fBu\-boot\fP files in the customisation
script. Examples are included for beaglebone\-black.
.sp
Some \fBu\-boot\fP examples recommend the use of the \fBlba\fP flag on the
boot partition, so use the \-\-bootflag option where relevant.
.SH INSTALLATION IMAGES AND VIRTUAL MACHINES
.sp
:file:\fBvmdebootstrap\fP is aimed principally at creating virtual machines,
not installers or prebuilt installation images. It is possible to create
prebuilt installation images for some devices but this depends on the
specific device. (A \(aqprebuilt installation image\(aq is a single image file
which can be written to physical media in a single operation and which
allows the device to boot directly into a fully installed system \-\-\- in
a similar way to how a virtual machine would behave.)
.sp
\fBvmdebootstrap\fP assumes that all operations take place on a local
image file or directory, not a physical block device / removable media.
.sp
\fBvmdebootstrap\fP is intended to be used with tools like \fBqemu\fP on
the command line to launch a new virtual machine. Not all devices have
virtualisation support in hardware.
.sp
This has implications for \fBu\-boot\fP support in some cases. If the
device can support reading the bootloader from a known partition, like
the beaglebone\-black, then \fBvmdebootstrap\fP can provide space for
the bootloader and the image will work as a prebuilt installation image.
If the device expects that the bootloader exists at a specific offset
and therefore requires that the bootloader is written as an image not
as a binary which can be copied into an existing partition,
\fBvmdebootstrap\fP is unable to include that bootloader image into
the virtual machine image.
.sp
The beagleboneblack.sh script in the examples/ directory provides a worked
example to create a prebuilt installation image. However, the beagleboneblack
itself does not support virtualisation in hardware, so is unable to launch
a virtual machine. Other devices, like the Cubietruck or Wandboard need
\fBu\-boot\fP at a predefined offset but can launch a virtual machine
using \fBqemu\fP, so the cubietruck and wandboard6q scripts in the
examples/ directory relate to building images for virtual machines once
the device is already installed and booted into a suitable kernel.
.sp
It is possible to wrap \fBvmdebootstrap\fP in such a way as to prepare
a physical block device with a bootloader image and then deploy the
bootstrap on top. However, this does require physical media to be
inserted and removed each time the wrapper is executed. To do this, use
the \fB\-\-tarball\fP option instead of the \fB\-\-image\fP option. Then setup
the physical media and bootloader image manually, as required for the
device, redefine the partitions to make space for the rootfs, create a
filesystem on the physical media and unpack the \fBvmdebootstrap\fP
tarball onto that filesystem. Once you have working media, an image can be
created using dd to read back from the media to an image file, allowing
other media to be written with a single image file.
.SH EXAMPLE
.sp
To create an image for the stable release of Debian:
.INDENT 0.0
.INDENT 3.5
.sp
.nf
.ft C
sudo vmdebootstrap \-\-image test.img \-\-size 1G \e
   \-\-log test.log \-\-log\-level debug \-\-verbose \e
   \-\-mirror http://mirror.lan/debian/
.ft P
.fi
.UNINDENT
.UNINDENT
.sp
To run the test image, make sure it is writeable. Use the \fB\-\-owner\fP
option to set mode 0644 for the specified user or use chmod manually:
.INDENT 0.0
.INDENT 3.5
.sp
.nf
.ft C
sudo chmod a+w ./test.img
.ft P
.fi
.UNINDENT
.UNINDENT
.sp
If \fB\-\-log\fP is also used, consider using \fB\-\-log\-mode\fP as well so
that the logfile is readable by the owner. By default, the log file
permissions are 0o600. The logfile itself will be owned by \fBroot\fP\&.
.sp
Execute using qemu, e.g. on amd64 using qemu\-system\-x86_64:
.INDENT 0.0
.INDENT 3.5
.sp
.nf
.ft C
qemu\-system\-x86_64 \-drive format=raw,file=./test.img
.ft P
.fi
.UNINDENT
.UNINDENT
.sp
(This loads the image in a new window.) Note the use of \fB\-drive
file=<img>,format=raw\fP which is needed for newer versions of QEMU.
.sp
There is a \fBbin/qemu\-wrapper.sh <image> <arch>\fP script for simple
calls where the \fB\-\-owner\fP option is used, e.g.:
.INDENT 0.0
.INDENT 3.5
.sp
.nf
.ft C
$ /usr/share/vmdebootstrap/qemu\-wrapper.sh jessie.img amd64
.ft P
.fi
.UNINDENT
.UNINDENT
.sp
There is EFI firmware available to use with QEMU when testing images built
using the UEFI support, but this software is in Debian non\-free due to patent
concerns. If you choose to install \fBovmf\fP to test UEFI builds, a
secondary change is also needed to symlink the provided \fBOVMF.fd\fP to
the file required by QEMU: \fBbios\-256k.bin\fP and then tell QEMU about
the location of this file with the \-L option:
.INDENT 0.0
.INDENT 3.5
.sp
.nf
.ft C
$ qemu\-system\-x86_64 \-L /usr/share/ovmf/ \-machine accel=kvm \e
 \-m 4096 \-smp 2 \-drive format=raw,file=test.img
.ft P
.fi
.UNINDENT
.UNINDENT
.sp
To use the \fB\-nographic\fP option, ensure that the \fB\-\-serial\-console\fP
option is supplied to \fBvmdebootstrap\fP and use \fB\-monitor none\fP when
booting the image with QEMU.
.sp
For further examples, including u\-boot support for beaglebone\-black,
see \fB/usr/share/vmdebootstrap/examples\fP
.SH NOTES
.sp
If you get problems with the bootstrap process, run a similar bootstrap
call directly and chroot into the directory to investigate the failure.
The actual debootstrap call is part of the vmdebootstrap logfile. The
debootstrap logfile, if any, will be copied into your current working
directory on error.
.sp
\fBdebootstrap\fP will download all the apt archive files into the apt cache and does not
remove them before starting the configuration of the packages. This can
mean that debootstrap can fail due to a lack of space on the device if
the VM size is small. vmdebootstrap cleans up the apt cache once debootstrap
has finished but this doesn\(aqt help if the package unpack or configuration
steps use up all of the space in the meantime. Avoid this problem by
specifying a larger size for the image.
.sp
\fBCAUTION:\fP
.INDENT 0.0
.INDENT 3.5
if you are also using a separate /boot partition in your options to
\fBvmdebootstrap\fP it may well be the boot partition which needs
to be enlarged rather than the entire image.
.UNINDENT
.UNINDENT
.sp
It is advisable to change the mirror in the example scripts to a mirror
closer to your location, particularly if you need to do repeated builds.
Use the \-\-apt\-mirror option to specify the apt mirror to be used inside
the image, after boot.
.sp
There are two types of examples for ARM devices available with
\fBvmdebootstrap\fP: prebuilt installation images (like the beaglebone\-black) and virtual
machine images (cubietruck and wandboard). ARM devices which do not
support hypervisor mode and which also rely on the bootloader being at
a specific offset instead of using a normal partition will
\fBnot\fP be supportable by vmdebootstrap. Similarly, devices which support
hypervisor will only be supported using virtual machine images, unless
the bootloader can be executed from a normal partition.
.sp
If the host device has a limited amount of RAM or simply to use a different
TMP directory when preparing the filesystems, set the \fBTMPDIR\fP or \fBTEMP\fP
or \fBTMP\fP environment variables, in line with the underlying support in
the python tempfile module.
.SH DEVELOPING
.SS Testing vmdebootstrap from git
.sp
\fBvmdebootstrap\fP uses \fByarn\fP for the test suite, available in the
\fI\%cmdtest\fP package. YARN
is a scenario testing tool. Scenarios are written in mostly human
readable language, however, they are not free form text. For more
information on YARN see \fI\%the homepage\fP:
.INDENT 0.0
.INDENT 3.5
.sp
.nf
.ft C
$ sudo apt \-y install cmdtest
.ft P
.fi
.UNINDENT
.UNINDENT
.sp
All commits must pass at least the fast tests. All merges into master
need to pass a full test. All additions of new functionality must add
fast and build tests \-\-\- fast tests for any new options and build tests
which exercise the new functionality. Build tests can add checks for
particular support on the machine running the test and skip if not
found or add new environment settings to selectively run some build
tests instead of all.
.sp
If no arguments are given, the full test suite will be executed:
.INDENT 0.0
.INDENT 3.5
.sp
.nf
.ft C
$ yarns/run\-tests
.ft P
.fi
.UNINDENT
.UNINDENT
.sp
\fBWARNING:\fP
.INDENT 0.0
.INDENT 3.5
Do not run the full test suite if your connection to a
Debian mirror is limited or metered. Each build requires a minimum
of 2GB free space in tmpfs. A full test takes at least 10 minutes.
.UNINDENT
.UNINDENT
.sp
When limiting the run to specific tests, each \fB\-\-env\fP option needs
to be specified separately:
.INDENT 0.0
.INDENT 3.5
.sp
.nf
.ft C
$ sudo yarns/run\-tests \-\-env TESTS=build \-\-env MIRROR=http://mirror/debian
.ft P
.fi
.UNINDENT
.UNINDENT
.sp
To run a single test, use the \fB\-\-run\fP option to specify the name of the
scenario (option can be repeated).
.SS pre\-commit
.sp
All vmdebootstrap developers need to run the fast tests as a pre\-commit
hook \-\-\- any patches which fail this test will be rejected:
.INDENT 0.0
.INDENT 3.5
.sp
.nf
.ft C
$ ln \-s ../../pre\-commit.sh .git/hooks/pre\-commit
.ft P
.fi
.UNINDENT
.UNINDENT
.sp
The pre\-commit hook just runs the fast tests which do not require
\fBsudo\fP\&.
.SS Fast tests
.sp
The fast checks validate the handling of incompatible option arguments:
.INDENT 0.0
.INDENT 3.5
.sp
.nf
.ft C
$ yarns/run\-tests \-\-env TESTS=fast
.ft P
.fi
.UNINDENT
.UNINDENT
.sp
Fast tests typically take a few seconds to run.
.SS Build tests
.sp
The slow / build tests build multiple images and use \fBsudo\fP \-\-\- a local
mirror is strongly recommended.
.INDENT 0.0
.INDENT 3.5
.sp
.nf
.ft C
$ sudo yarns/run\-tests \-\-env TESTS=build \-\-env MIRROR=http://mirror/debian
.ft P
.fi
.UNINDENT
.UNINDENT
.sp
If \fBMIRROR\fP is not specified, a default mirror of \fBhttp://httpredir.debian.org/debian/\fP
will be used.
.SS LAVA tests
.sp
There is an example \fBlava\-submit.py\fP script which can be edited
to automatically submit QEMU tests to a specified LAVA instance. The
images themselves will use local \fBfile://\fP URLs and therefore the
\fBlava\-dispatcher\fP needs to be installed locally. Configuring LAVA
for these tests is a separate topic \-\-\- please ask on the \fI\%vmdebootstrap
mailing list\fP\&.
.SH AUTHOR
Neil Williams
.SH COPYRIGHT
2019, Neil Williams
.\" Generated by docutils manpage writer.
.