Scribd
Upload
53 views

What Is Init?: Initialization of Userspace PID 1: The First Process The Kernel Starts

Init is the first process started by the Linux kernel and is responsible for initializing the system and starting other processes. There are different init systems including BSD init, System V init, Upstart, and systemd. Systemd is the newest init system and aims to improve on previous systems through features like socket and bus activation that allow services to start in parallel, control groups for process management, and fine-grained configuration of units like services, sockets, and mounts.

Uploaded by

luthfi
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
53 views

What Is Init?: Initialization of Userspace PID 1: The First Process The Kernel Starts

Init is the first process started by the Linux kernel and is responsible for initializing the system and starting other processes. There are different init systems including BSD init, System V init, Upstart, and systemd. Systemd is the newest init system and aims to improve on previous systems through features like socket and bus activation that allow services to start in parallel, control groups for process management, and fine-grained configuration of units like services, sockets, and mounts.

Uploaded by

luthfi
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 11

What is Init?

● Initialization of Userspace
● PID 1: the first process the kernel starts
– Spawns all other processes
– Grandparent of all userspace processes
● "Sets up" the system for use
– Environment, Hostname
– Filesystem checks, mounts
– Networking, Services, etc.
BSD Init

● The oldest one, used by traditional UNIX


● Monolithic /etc/rc script
– rc.local for local changes
● No concept of runlevels
● Later variants support /etc/rc.d
– packages can install their custom scripts here
– rcorder used to determine script order
– dependencies are stored in rc.d scripts
● Arch, Slackware, FreeBSD, etc.
System V Init

● Originated with System V UNIX release


● Configured via /etc/inittab and scripts
● Runlevel-based
● Scripts in /etc/rcN.d, where N is the runlevel
– Really symlinks to /etc/init.d or similar
– 'K' for Kill scripts, 'S' for Start scripts
– Order determined by number after 'K' or 'S'
● Newer variants have dependency tracking
– "chkconfig" or LSB headers
Upstart

● Originally written for Ubuntu


● Event-based
– Tasks and Services started/stopped by events
– Events generated on Task/Service start/stop
– Respawning supported
● Parallel starting of services
● Backwards compatible with SysVinit
● Native configuration files in /etc/init with deps
systemd

● Brought to you by Lennart Poettering of


PulseAudio fame
● Socket- and bus-activated services
– Automatic parallelization!
● Fine-grained per-process control
– Environment, rlimits, nice level, etc.
● cgroup-based process tracking
● SysVinit (and LSB) compatible
● Comparable to launchd and Solaris' SMF
The magic of parallelization: Socket
and Bus activation
● Remember the inetd superserver?
● systemd creates all listening sockets at once
– Then starts the daemons in parallel
– Kernel blocks sockets for us
– Clients (other services) wait until other end
becomes available
● Autospawning of services
● No need to manually code dependencies!
● Works with D-Bus too!
Parallelizing Filesystem Setup

● fsck & mounting filesystems takes a long time


● AutoFS provides mount point
● Access to the directory is queued by kernel
● When fsck & mount complete, AutoFS
replaced by real mount
Control Groups

● Used instead of process tree (PIDs) to


manage services
● Avoids "double-fork" from escaping process
control
● Can manage lots of things:
– rlimits, uids, gids, environment
– CPU & I/O schedulers, affinity
– Capabilities (security)
– Read-only bind mounts for service lockdown
Units

● service: start/stop/restart/reload
– native & SysV/LSB scripts
● socket: internet & local UNIX sockets
● device: device node in /dev
● mount: mount point
– /etc/fstab is also supported
● automount: autofs-managed mount
Units (continued)

● target: for grouping


– kinda like runlevels
● snapshot: used to save/rollback state of all
units
– "emergency shell"
– suspend/resume
Configuration and Tools

● /etc/systemd/*
● .desktop file (INI) style config
● Symlinks used for target selection
– /etc/systemd/system/default.target ->
/lib/systemd/system/runlevel5.target
● systemctl command
– /sbin/service and /sbin/chkconfig supported to
a certain extent
● systemadm GUI (systemd-gtk)

You might also like