Skip to content

Task 0: Cloud-Init

In Assignment 1, we delved into using QEMU to traverse the boot procedure of a system, from BIOS to Ubuntu. Additionally, Assignment 2 highlighted the potency of QEMU in conjunction with GDB for understanding and debugging the kernel.

For this assignment, we'll use QEMU to emulate a non-existent peripheral device and develop a device driver for it. But before diving into that, let's establish a production-level virtual machine using QEMU, transitioning away from a testing environment.

Step 1: Boot using A Cloud Image

In Assignment 1, we used a RootFS archive to create a disk image, which did not involve any boot components. For this task, we will boot from a cloud image, which is a disk image containing a bootable operating system.

Download the cloud image from the following link: https://cloud-images.ubuntu.com/releases/22.04/release/ubuntu-22.04-server-cloudimg-amd64.img

We can directly boot from this image, but doing so would modify it. In a real cloud environment, numerous virtual machines can be booted from the same image, so we prefer not to alter the image itself. Instead, we create a new disk image that is backed by the cloud image.

qemu-img create -f qcow2 -b ubuntu-22.04-server-cloudimg-amd64.img -F qcow2 hdd.qcow2 10G

Then, you can boot the image using the following command:

qemu-system-x86_64 -smp 2 -m 2G -nographic hdd.qcow2

Here, the -smp option specifies the number of CPUs, and the -m option specifies the amount of memory. You can adjust these values based on your system's resources.

Step 2: Customize your VM using Cloud-Init

I assume you have been trying to log in to the VM but failed because you don't know the password. Similar to what we experienced in Assignment 1, there isn't a user account created in the VM by default. Hence, we need to create one before logging in.

Previously, we booted into the system in single-user mode to create a user, but this approach isn't scalable. In a cloud environment, we can utilize a tool called cloud-init to automate the process of creating a user account or other configurations. Let's follow the tutorial.

First, create a directory to contain the files needed for cloud-init.

Then, within the directory, create a file named user-data with the following content. Make sure to replace the username and password placeholders with your own.

Additionally, create two more files:

  1. meta-data with the following content:

    instance-id: someid/somehostname

  2. vendor-data, which can be empty.

Finally, launch an HTTP server in the directory to host these files.

$ python3 -m http.server

Now you can boot the VM with cloud-init enabled.

qemu-system-x86_64 -smp 2 -m 2G -nographic -nic user \
  -smbios type=1,serial=ds='nocloud;s=http://10.0.2.2:8000/' \
  hdd.qcow2

In addition to the regular kernel outputs, you should see some cloud-init logs. Eventually, you'll see something like "Reached target Cloud-init target." After pressing "Enter," you should encounter the login prompt. You can now log in using the username and password specified in user-data.

This is just a piece of cake for cloud-init. You can accomplish even more with it, such as setting up SSH keys, installing packages, and running scripts. For further information, you can refer to the examples.

Step 3: SSH into the VM

If you want to login to the VM using SSH, you can forward the SSH port from the host to the VM. You can do this by changing -nic option in the command

qemu-system-x86_64 -smp 2 -m 2G -nographic -nic user,hostfwd=tcp::2222-:22 hdd.qcow2

You can change the port number 2222 to any other port number you prefer.

Step 4: Install a New Kernel

The kernel version in the cloud image is not v6.7, which is the version we will use for the rest of the assignments. You can either build the kernel from source and install it in the VM, or if you have a pre-built kernel on your host machine, you can build a deb package and install it in the VM.