Installation process overview

The purpose of this document is to offer some information about how the installation process of SUSE and openSUSE works from a technical point of view. It is, in no way, oriented to the end-user although it could be easily understood for a regular Linux user.

Different installation modes are supported, ranging from a regular installation to an unattended one. Also system update or repair are supported. You could take a look at yast-yast2 repository to learn more about that. This document is mainly about installation and autoinstallation modes.

Overview

Although YaST sits at the central point of the installation process, it is not the only software involved. Linuxrc, which will be started just before YaST, is another key player.

Linuxrc sets up the hardware and initializes the installation process. It can do very cool things and have a lot of options for you to tweak. Newcomers should take a look at startshell, netdevice, netsetup and, of course, autoyast options.

After the system is started, a script called /sbin/inst_setup is ran. That script performs some initialization tasks and, at the end, calls YaST first stage (known as initial). This is the main part of the installation workflow: disks are set up (partitions and filesystem), software is installed, boot loader is prepared, etc. When the first stage is over, the machine is rebooted to get into YaST second stage (known as continue) if needed.

That second stage is not always needed. As of SLE 12 and openSUSE 13.2 and later, a regular installation is done just during the YaST2 First Stage and the second one is not used anymore. When it comes to AutoYaST, the second stage is used for system configuration.

Let’s see the process in a more detailed way.

Installer boot process

After the system boots, the first program to run is /sbin/inst_setup. It’s a shell script that performs some initialization tasks and, at the end, it invokes the YaST2 first stage.

Among inst_setup initialization tasks, we could highlight:

Starting basic services like klog, syslog or nscd.
Setting host and domain names.
Managing YaST driver updates.
Handling the SSH installation (letting the user start the process through yast.ssh).
Registering the VNC service through SLP if needed.
Handling the STARTSHELL parameter to launch a shell before starting YaST if it was requested by the user.

Unless SSH must be used to perform the installation, inst_setup will start the YaST2.First-Stage just executing /sbin/yast. Otherwise, the user will launch YaST calling yast.ssh.

YaST2.First-Stage

At this stage, /sbin/yast is just a symbolic link to /usr/lib/YaST2/startup/YaST2.First-Stage. The main purpose of this script is to kick off the installation process (through yast-installation). After some initialization work (described below), this script goes through a series of steps defined scripts at /usr/lib/YaST2/startup/First-Stage.

So the workflow is something like:

Import installation settings from /etc/install.inf as environment variables (import_install_inf at /usr/lib/YaST2/startup/common/misc.sh).
Enable logging into /var/log/YaST2/y2start.log.
Execute preFirstStage hooks.
Execute the FirstStage (see stages/steps). Additional preFirstCall and postFirstCall hooks are called. YaST gets called at this point.
Execute postFirstStage hooks.

Finally, the script exits (returning the error code from YaST) and control is returned to /sbin/inst_setup.

Stages/Steps

Maybe the use of the term stage here could be somewhat confusing. You must just take into account that the following stages are just the steps performed during the installation’s First Stage. We keep the term stage to refer to both of them just because it’s the way they’re called in the source code.

If you want to know the nitty-gritty details of those steps, you could find them at /usr/lib/YaST2/startup/First-Stage.

Hardware detection through hwinfo.
Umount inst-sys (/var/adm/mount).
Set up the language. Remember that you could specify a language through the Linuxrc parameter language.
Handle kernel parameters. A this time, it only copy repair to install.inf is present.
Configure the terminal (and save TERM to /etc/install.inf).
Set environment variables related to YaST log (size and max logs count).
Start YaST in installation mode (yast-installation) with /usr/lib/YaST2/startup/YaST2.call installation initial.
Umount partitions and write exit code at /tmp/YaST2-First-Stage-Exit-Code.

inst_worker_initial

For YaST, the installation is now in the initial stage. At this point, the installation client relies on inst_worker_initial. You could take a look at the documentation about installation clients to learn more.

Depending on the installation mode, which is set to installation by default, some initialization tasks are performed and flow control is hand over to ProductControl module, which is responsible of driving the installation process.

Just as a side note, if /etc/install.inf contains an AutoYaST entry, mode will be set to autoinstall and the installation will be handled by AutoYaST.

YaST2.Second-Stage

After the first reboot, the YaST2.Second-Stage comes into action. It’s defined as a systemd service (take a look at YaST2.Second-Stage.service). This service is ran only if a file /var/lib/YaST2/runme_at_boot exists.

The script is located at /usr/lib/YaST2/startup/YaST2.Second-Stage and, as first stage, it performs some initializations tasks and then go through a series of steps defined as scripts at /usr/lib/YaST2/startup/Second-Stage.

Initialization

Disable splash screen.
Import installation settings from /etc/install.inf as environment variables.
Set up architecture variables.
Prepare manpages and info files directories.
Start startup logging at /var/log/YaST2/y2start.log.
Execute preSecondStage hooks.
Execute the SecondStage (see stages). Additional preSecondCall and postSecondCall hooks are executed. YaST gets called at this point.
Execute postSecondStage hooks.

Stages/steps

The scripts which perform these steps live in /usr/lib/YaST2/startup/Second-Stage. They are responsible for:

Setting up logging (including syslogd).
If /var/lib/autoinstall/autoconf/autoconf.xml is found, then AutoYaST is started (as /usr/lib/YaST2/startup/YaST2.call installation continue) and, after it exits, the second stage is over (next steps won’t be executed).
Setting up language stuff (LC_ALL and LANG environment variables).
Setting up virtual console.
Starting services as pcmcia or hald (Hardware Abstraction Layer).
Bringing up network (if configuration is found), starting a second shell and getting prepared for a VNC installation (if needed).
Starting YaST. If installation is done through SSH, at this point the user must log in and launch YaST manually.
Cleaning up things when installation is finished.

inst_worker_continue

The client which drives the second stage is inst_worker_continue. Of course, it relies on ProductControl as inst_worker_initial does.

YaST2.call

During installation, YaST is invoked through YaST2.call script. This script is responsible for detecting the installation medium: Qt, SSH, VNC and ncurses. It also checks requirements for selected medium and falls back to ncurses if they are not met.

The control file

YaST installation process is really very flexible. What actions must be taken in different modes and scenarios are described in the control file. YaST searches for that file in /y2update/control.xml, /control.xml and /etc/YaST2/control.xml and each SUSE product will have their own file.

Quite nice documentation lives in this repository about the content of these files.