Resources

Intro

Qemu can be used to simulate various hardware configurations and to easily test the debian installer in various conditions; it's great for testing "delicate" things like a partitioner, since disk devices are usually represented by files.

It can be used also to emulate other architectures: I tested amd64 and powerpc versions on the installer on my i386 host machine. At the moment (qemu-0.8.0) sparc and ARM are being worked on and have some basic support.

If you have CPU with virtualization support, you can use kvm, which is faster than qemu but command-line compatible with it.

Here some possible d-i boot configurations:

Basic setup

First of all, you need to create a file which will acts as an hard disk for the virtual machine:

$ qemu-img create hd_img.img 500M

You can then download an iso image of the installer, or create one yourself; you can then run a g-i session by running the following command:

$ qemu -redir tcp:5555:10.0.2.15:80 -cdrom mini.iso -hda hd_img.img -boot d

qemu-0.8.1 does not need any net options, it defaults to user-mode networking; with previous versions of qemu you might need to use "-net nic -net user" in order to get user-mode networking working.

The redir option is useful for accessing to the http server inside the installer which can be reached pointing your browser to http://localhost:5555

Qemu monitor

Qemu has a "monitor" (ctrl-alt-2) which can be used to access various funcionalities such as:

Communicating with the VM

You often need to copy file from the VM to the host (i.e. log files), or you need to copy files from the host to the VM (i.e. some udebs).

There are various ways of achieving this; if you have an ssh client inside the VM, for example, you can scp files from/to the host. When you boot a (standard) g-i image, it takes a few steps before you can have the ssh client into the VM, so you might find useful doing the following:

wget can be used to copy files from your host (or somewhere on the net); note that the wget command is always included in the mini.iso

netcat can be use to export files from the VM to the host: you first need open a listening connection on the host

  $ nc -l -p 1111 > vm.log

then, you can run the following command to export a file from the VM:

  # cat logfile.txt | nc IPADDR_HOST 1111

you should find the file vm.log on you host, with the content of logfile.txt Obviously networking inside the VM has to up before being able to do this.

network install

You can use the built-in qemu tftp server to boot the netinst image:

$ qemu -hda hda.img -tftp ~/debian-installer/installer/build/dest/netboot -bootp /pxelinux.0 -boot n 

In this case, you will need to to copy a bootrom from etherboot package to /usr/share/qemu.

install over serial console

you can easily test an installation over a serial console by passing the following string at the boot prompt:

install console=ttyS0,9600,n8

you can then reach the serial console by pressing ctrl-alt-3

Floppy install

Where Qemu still might have support for floppies, the Linux kernel doesn't have any more.

To test an installation from the floppy images, you first need to grab boot.img and root.img and use the files as if they were floppy disks:

$ qemu -fda boot.img -hda hd_image.img -boot a

once you're asked to insert the root disk, just switch to "monitor" (ctrl-alt-2) and

(qemu) change fda root.img

switch back to d-i (ctrl-alt-1) and press ENTER

Different archs install

Usually you can run something like

$ qemu-system-somearch -cdrom debian-foo-somearch-netinst.iso -hda hd_image.img -m 1G -serial stdio

Sometimes the arch cannot boot from an iso image and you have to mount the image to get the kernel and initrd and boot them:

$ qemu-system-somearch -kernel /mnt/cdrom/install/vmlinuz -initrd /mnt/cdrom/install/initrd.gz -cdrom debian-foo-somearch-netinst.iso -hda hd_image.img -m 1G -serial stdio

Sometimes you have to take the kernel from the netboot build (e.g. http://ftp.de.debian.org/debian/dists/stable/main/installer-mipsel/current/images/malta/netboot/)

But for some cases this will not be enough. Arch-specific tricks are specified in the next subsections

Sometimes the CD image will not be recognized even if the drive is there, just because it does not see that this is a CDROM. You can answer

amd64

qemu-system-x86_64 just works fine as it is

arm64

$ qemu-system-aarch64 -M virt -cpu max -bios /usr/share/qemu-efi-aarch64/QEMU_EFI.fd -cdrom debian-foo-arm64-netinst.iso -hda hd_image.img -m 1G -serial stdio

See more on Arm64Qemu

armhf

$ qemu-system-arm -M virt -cpu max -kernel /mnt/cdrom/install.ahf/vmlinuz -initrd /mnt/cdrom/install.ahf/initrd.gz -drive file=debian-foo-armhf-netinst.iso,if=none,id=cd,media=cdrom,read-only -device virtio-blk-device,drive=cd -drive file=hd_image.img,if=none,id=hd -device virtio-blk-device,drive=hd -device virtio-net-device,netdev=net0 -netdev user,id=net0 -m 1G -serial stdio

i386

qemu-system-i386 just works fine as it is

mips64el

$ qemu-system-mips64el -M malta -cpu 5KEc -kernel /mnt/cdrom/install/malta/netboot/vmlinuz-5.10.0-6-5kc-malta -initrd /mnt/cdrom/install/malta/netboot/initrd.gz -hda hd_image.img -m 1G -serial stdio

See more on https://gist.github.com/bradfa/46ceff759a0cf9f392cc069c4f0f095a

mipsel

$ qemu-system-mipsel -M malta -kernel /tmp/vmlinuz-5.10.0-6-4kc-malta -initrd /tmp/initrd.gz -hda hd_image.img -m 1G -serial stdio

See more on https://www.aurel32.net/info/debian_mips_qemu.php

ppc64el

qemu-system-ppc64le just works fine as it is

See more on DebianInstaller/PowerPC/qemu

riscv64

$ qemu-system-riscv64 -machine virt -bios /usr/lib/riscv64-linux-gnu/opensbi/generic/fw_jump.elf -kernel vmlinux -initrd initrd.gz -device virtio-net-device,netdev=usernet -netdev user,id=usernet -device virtio-blk-device,drive=hd0 -drive file=hd_image.img,id=hd0 -serial stdio -append 'console=ttyS0' -m 1.9G

See more about it on RISC-V


CategoryDebianInstaller