This tutorial describes the Linux booting process in detail, such as the stages involved, which scripts are run, which configuration files are read and their order, since the time of turning on the system till getting the login prompt.
1) BIOS Initialization
BIOS or the Basic Input Output System is a firmware program that performs a very basic level of interaction with the hardware. This is the first program that takes control when the computer is switched on. The BIOS performs a test on all the hardware components and peripherals called POST or the “Power On Self Test”. It initializes the required hardware for booting.
After POST executes successfully, the BIOS looks for a boot device from a list of devices. Modern BIOS allow you to configure the order of these devices (sometimes called boot preference) that BIOS checks for booting. These boot devices can either be a floppy drive, CDROM, hard drive, a network interface or other removable media (such as USB flash drive).
The BIOS checks for the boot sector on the bootable device. Boot sector is the first physical sector on the storage device, and contains the code required for booting the machine.
2) The Master Boot Record
In case of hard disks and many other mass storage media, the boot sector is MBR. MBR consists of 512 bytes at the first sector of the hard disk. It is important to note that MBR is not located inside any partition. MBR precedes the first partition. The layout of MBR is as follows:
• First 446 bytes contain bootable code.
• Next 64 bytes contain partition information for 4 partitions (16x4). That is why the hard disks can have only 4 primary partitions as the MBR can store information for 4 partitions only. So if you need more than 4 partitions on the hard disk, one of the primary partition has to be extended, and out of these extended partitions, logical partitions are created.
• The last 2 bytes are for MBR signature, also called magic number. (Thus total of 446 + 64 + 2 = 512 bytes).
| 446 | 16 | 16 | 16 | 16 | 2 |
The first 446 bytes of MBR contains the code that locates the partition to boot from. The rest of the booting process take place from that partition. This partition contains a software program called the 'bootloader' for booting the system.
3) About GRUB
According to gnu.org, "a bootloader is the first software program that runs when a computer starts". GRUB or GRand Unified Bootloader is the bootloading program for Linux like operating systems. There are two main versions of Grub that are available (Grub version 1 and 2). Presently, most Linux distros have started using grub version 2. One main feature of grub is that it can be installed using Linux image and there is no need for running the operating system.
Grub is a multi-stage bootloader (Stage1, Stage 1.5 and Stage 2). The 3 stages for grub version 1 and version 2 are explained below.
Grub Version 1 stages
Stage 1 can load Stage 2 directly, but it is normally set up to load Stage 1.5. GRUB Stage 1.5 is located in the first 30 kilobytes of hard disk immediately following the MBR and before the first partition. If this space is not available (Unusual partition table, special disk drivers, GPT or LVM disk), the installation of Stage 1.5 will fail. The stage 1.5 image contains file system drivers. This enables stage 1.5 to directly load stage 2 from any known location in the filesystem, for example from /boot/grub. Stage 2 will then load the default configuration file and any other modules needed.
Grub Version 2 stages
Stage 1 -> boot.img is stored in the MBR (or optionally in any of the volume boot records) and addresses the next Stage by an LBA48 address (the 1024 cylinder boundary of GRUB legacy is omitted) , and at the time of installation, it is configured to load the first sector of core.img
Stage 1.5 -> core.img is by default written to the sectors between the MBR and the first partition, when these sectors are free and available. For legacy reasons, the first partition of a hard disc does not begin at sector 1 (counting begins with 0) but at sector 63 leaving a gap of 63 sectors of empty space, that is not a part of any partition of the file system and therefore, not prone to any problems related with it. Once executed, core.img will load its configuration file and any other modules needed, particularly file system drivers. At the time of installation, it is generated from diskboot.img and configured to load stage 2.
Refer to the bottom of this page for boot process using grub version 2
Stage 2 -> files belonging to stage 2 are held in the /boot/grub-directory, which is a sub-directory of /boot/ directory.
Grub Boot Stage Errors
Grub Stage 1 Errors: Hard Disk Error Floppy Error Read Error Geom Error
Grub Stage 1.5 Errors: The Stage 1.5 handles errors by printing an error number in the form of "Error":
Grub Stage 2 Errors: Selected item won't fit into memory Selected disk doesn't exist Disk read error Disk write error Disk geometry error Attempt to access block outside pa No such partition Bad filename (must be absolute path name or block list) Bad file or directory type File not found Inconsistent filesystem structure Filesystem compatibility error, can\'t read whole file Error while parsing number Device string unrecognizable Invalid device requested Loading below 1MB is not supported Unsupported Multiboot features requested Unknown boot failure Must load Multiboot kernel before modules Must load Linux kernel before initrd Cannot boot without kernel loaded Unrecognized command Bad or incompatible header on compressed file Bad or corrupt data while decompressing file Bad or corrupt version of stage1/stage2
Here is a sample of grub.conf:
# grub.conf generated by anaconda # # Note that you do not have to rerun grub after making changes to this file # NOTICE: You have a /boot partition. This means that # all kernel and initrd paths are relative to /boot/, eg. # root (hd0,0) # kernel /vmlinuz-version ro root=/dev/VolGroup00/LogVol00 # initrd /initrd-version.img #boot=/dev/sda default=0 timeout=5 splashimage=(hd0,0)/grub/splash.xpm.gz hiddenmenu title Red Hat Enterprise Linux Server (2.6.18-238.el5) root (hd0,0) kernel /vmlinuz-2.6.18-238.el5 ro root=/dev/VolGroup00/LogVol00 initrd /initrd-2.6.18-238.el5.img title Windows XP Pro rootnoverify (hd0,1) chainload +1
GRUB allows you to choose from a list of available operating systems. The different OS’s have different title in grub.conf. Corresponding to this configuration file, the user at GRUB menu will be asked to choose from two options: Red Hat Enterprise Linux Server (2.6.18-238.el5), and Windows XP Pro. The commands following the title are the commands that run in the background when an entry is selected from the GRUB menu. For example, when the user selects the first option of Red Hat Enterprise Linux, following three commands execute:
root (hd0,0) kernel /vmlinuz-2.6.18-238.el5 ro root=/dev/VolGroup00/LogVol00 initrd /initrd-2.6.18-238.el5.img
Corresponding to this menu entry, the first command, i.e. "root (hd0,0)" specifies the partition on which a compressed kernel image and initrd file are located.
The second command i.e. "kernel /vmlinuz-2.6.18-238.el5 ro root=/dev/VolGroup00/LogVol00" tells which kernel image to use (in this case vmlinuz-2.6.18-238.el5). The arguments to this command are 'ro' and 'root'. 'root' specifies the device on which root directory of the filesystem (i.e. / directory) is located; 'ro' means that this partition is to be mounted in read only mode (i.e. the kernel mounts the root partition in read only mode). Note that the partition for root filesystem and the partition on which this kernel image resides (i.e. boot partition) are different.
The third command is the location of initrd. Before going into the details of what initrd is, let's look at a problem caused at the boot time.
The Chicken/Egg Module Problem and initrd
The kernel needs to mount the root filesystem (as specified in the second command above). This filesystem may be on some partition with one of the following capabilities:
• Logical Volume Management • Software RAID • NFS (Network File System) • Some other encrypted partition
The Linux kernel does not have these features compiled into it. But they are present as modules. So, the kernel needs to load these modules in order to mount the root filesystem. These modules are present under /lib/modules/ directory, which is present on the root filesystem itself. But this root filesystem is not mounted yet (that is what we have been trying to do so far). So, how can kernel access the modules for mounting a filesystem which are present on the filesystem itself (without mounting it)?
The kernel and GRUB provide a solution through initrd, the initial RAM disk. It contains the modules needed to mount the filesystem, and only modules needed for that filesystem are included.
GRUB also supports chainloading, the method used to pass control to other bootloader. Chainloading is used by GRUB to boot operating systems like windows. This can be checked in the above configuration file, under the title Windows XP Pro.
4) Kernel Initialization
After GRUB stage 2, the location of the kernel and necessary modules through initrd are known. Now the kernel is loaded into memory and initialized. The initrd image is compiled and mounted into the memory. It serves as a temporary root file system and helps kernel to boot properly without mounting any root file system. Now that all the drivers are loaded into memory, and kernel has booted, kernel mounts the root filesystem in read only mode, and starts the first process.
5) The init Process
The ‘init’ is the first process started by kernel (initialization process). It is the parent of all processes. The PID (Process ID) of init process is always 1. This process persists till the computer halts. It is responsible for the whole state of system. The settings for this process are stored in its configuration file, /etc/inittab (system initialization table). Before diving deeper into the details of this file and proceeding any further with the boot process, let’s discuss runlevels.
Runlevel is the state in which a system boots. It can boot in a single user mode, multi-user mode, with networking, and with graphics etc. The following are the default runlevels defined by Linux:
0: Halt or shutdown the system 1: Single user mode 2: Multi-user mode, without networking 3: Full multi user mode, with NFS (typical for servers) 4: Officially not defined; Unused 5: Full multi user with NFS and graphics (typical for desktops) 6: Reboot
There are some additional runlevels, which are rarely used such as:
s,S or single: Alternate single user mode emergency: Bypass rc.sysinit (discussed later in this article)
Summarizing runlevels: Runlevel 1 is used for maintenance purpose, because this is a very limited runlevel. Only the minimum scripts run in this runlevel. Only root user can log in. No other user can log into this runlevel. Runlevel 2 is more relaxed than runlevel 1. Here, all the users can login, but networking service is not running. Runlevel 3 provides a full working environment. All users can log in, networking is enabled. Runlevel 4 is for experiment purpose only. In runlevel 5, graphical console is available. Runlevel '0' is the halt state of system, and switching to runlevel 6 will reboot the system.
Now that we are clear about runlevels, let's continue our discussion about init process and its configuration file, 'inittab'.
'inittab' file from RHEL is:
# # inittab This file describes how the INIT process should set up # the system in a certain run-level. # # Author: Miquel van Smoorenburg, <firstname.lastname@example.org> # Modified for RHS Linux by Marc Ewing and Donnie Barnes # # Default runlevel. The runlevels used by RHS are: # 0 - halt (Do NOT set initdefault to this) # 1 - Single user mode # 2 - Multiuser, without NFS (The same as 3, if you do not have networking) # 3 - Full multiuser mode # 4 - unused # 5 - X11 # 6 - reboot (Do NOT set initdefault to this) # id:3:initdefault: # System initialization. si::sysinit:/etc/rc.d/rc.sysinit l0:0:wait:/etc/rc.d/rc 0 l1:1:wait:/etc/rc.d/rc 1 l2:2:wait:/etc/rc.d/rc 2 l3:3:wait:/etc/rc.d/rc 3 l4:4:wait:/etc/rc.d/rc 4 l5:5:wait:/etc/rc.d/rc 5 l6:6:wait:/etc/rc.d/rc 6 # Trap CTRL-ALT-DELETE ca::ctrlaltdel:/sbin/shutdown -t3 -r now # When our UPS tells us power has failed, assume we have a few minutes # of power left. Schedule a shutdown for 2 minutes from now. # This does, of course, assume you have powerd installed and your # UPS connected and working correctly. pf::powerfail:/sbin/shutdown -f -h +2 "Power Failure; System Shutting Down" # If power was restored before the shutdown kicked in, cancel it. pr:12345:powerokwait:/sbin/shutdown -c "Power Restored; Shutdown Cancelled" # Run gettys in standard runlevels 1:2345:respawn:/sbin/mingetty tty1 2:2345:respawn:/sbin/mingetty tty2 3:2345:respawn:/sbin/mingetty tty3 4:2345:respawn:/sbin/mingetty tty4 5:2345:respawn:/sbin/mingetty tty5 6:2345:respawn:/sbin/mingetty tty6 # Run xdm in runlevel 5 x:5:respawn:/etc/X11/prefdm -nodaemon
The general format of entries in this file is:
"id" is the unique sequence of characters for identifying an entry in this file. "runlevels" are the runlevels for which that entry is applicable. "action" is the action that is to be taken for the runlevel. "process" specifies the process that is to be executed.
Now, let's look at some entries in this file. The first uncommented line, i.e. "id:3:initdefault:" specifies the default runlevel in which the system will boot. According to this file, the system will boot into runlevel 3 by default. One more important line is system initialization line (the one with id: si)
This line tells init process to execute "/etc/rc.d/rc.sysinit" script. This is the first script executed by init during booting process.
7) Script rc.sysinit
When this script executes, it asks the user for interactive setup (Press 'I' to enter interactive startup). This script performs a lot of functions for the system that include:
• Setting hostname
• It checks and mounts filesystems (/proc, /sys, and others in /etc/fstab).
• Enables SWAP
• Sets SELinux
• Configures kernel parameters from sysctl.conf
• Sets system clock
• Loads required modules
• Setting a minimal path
• Initializes required serial and other ports
• Enables RAID and LVM
• And the root filesystem, which was mounted read-only earlier (as specified in GRUB configuration file), is checked and remounted read-write.
After this script has been executed, the runlevel specific files are run, according to "/etc/inittab". In our file, the default runlevel was 3. So the line corresponding to this runlevel is:
There are directories corresponding to each runlevel under "/etc/rc.d" directory. According to this line, the scripts in /etc/rc.d/rc3.d directory are run. Let’s list the files in this directory. (Other directories like rc1.d, rc2.d ... have similar files in them).
[root@redhat-server ~]# ls -l /etc/rc.d/rc3.d/ total 304 lrwxrwxrwx 1 root root 17 Jul 3 03:51 K01dnsmasq -> ../init.d/dnsmasq lrwxrwxrwx 1 root root 24 Jul 3 03:51 K02avahi-dnsconfd -> ../init.d/avahi-dnsconfd lrwxrwxrwx 1 root root 24 Jul 3 03:54 K02NetworkManager -> ../init.d/NetworkManager lrwxrwxrwx 1 root root 16 Jul 3 03:50 K05conman -> ../init.d/conman lrwxrwxrwx 1 root root 19 Jul 3 03:51 K05saslauthd -> ../init.d/saslauthd lrwxrwxrwx 1 root root 17 Jul 3 03:52 K05wdaemon -> ../init.d/wdaemon lrwxrwxrwx 1 root root 16 Jul 3 03:50 K10psacct -> ../init.d/psacct lrwxrwxrwx 1 root root 13 Jul 3 03:53 K20nfs -> ../init.d/nfs lrwxrwxrwx 1 root root 14 Jul 3 03:53 K24irda -> ../init.d/irda lrwxrwxrwx 1 root root 19 Jul 3 03:53 K35vncserver -> ../init.d/vncserver lrwxrwxrwx 1 root root 17 Jul 3 03:56 K35winbind -> ../init.d/winbind lrwxrwxrwx 1 root root 20 Jul 3 03:51 K50netconsole -> ../init.d/netconsole lrwxrwxrwx 1 root root 16 Jul 5 10:10 K50vsftpd -> ../init.d/vsftpd lrwxrwxrwx 1 root root 20 Jul 3 03:53 K69rpcsvcgssd -> ../init.d/rpcsvcgssd lrwxrwxrwx 1 root root 16 Jul 3 03:56 K73ypbind -> ../init.d/ypbind lrwxrwxrwx 1 root root 14 Jul 3 03:52 K74ipmi -> ../init.d/ipmi lrwxrwxrwx 1 root root 14 Jul 3 03:49 K74nscd -> ../init.d/nscd lrwxrwxrwx 1 root root 14 Jul 3 04:00 K74ntpd -> ../init.d/ntpd lrwxrwxrwx 1 root root 15 Jul 3 03:49 K80kdump -> ../init.d/kdump lrwxrwxrwx 1 root root 15 Jul 3 03:52 K85mdmpd -> ../init.d/mdmpd lrwxrwxrwx 1 root root 20 Jul 3 03:49 K87multipathd -> ../init.d/multipathd lrwxrwxrwx 1 root root 24 Jul 3 03:51 K88wpa_supplicant -> ../init.d/wpa_supplicant lrwxrwxrwx 1 root root 14 Jul 3 03:52 K89dund -> ../init.d/dund lrwxrwxrwx 1 root root 18 Jul 3 03:49 K89netplugd -> ../init.d/netplugd lrwxrwxrwx 1 root root 14 Jul 3 03:52 K89pand -> ../init.d/pand lrwxrwxrwx 1 root root 15 Jul 3 03:49 K89rdisc -> ../init.d/rdisc lrwxrwxrwx 1 root root 14 Jul 3 03:53 K91capi -> ../init.d/capi lrwxrwxrwx 1 root root 25 Jul 3 03:52 K99readahead_later -> ../init.d/readahead_later lrwxrwxrwx 1 root root 23 Jul 3 03:52 S00microcode_ctl -> ../init.d/microcode_ctl lrwxrwxrwx 1 root root 25 Jul 3 03:52 S04readahead_early -> ../init.d/readahead_early lrwxrwxrwx 1 root root 15 Jul 3 03:54 S05kudzu -> ../init.d/kudzu lrwxrwxrwx 1 root root 16 Jul 3 03:51 S07iscsid -> ../init.d/iscsid lrwxrwxrwx 1 root root 19 Jul 3 03:49 S08ip6tables -> ../init.d/ip6tables lrwxrwxrwx 1 root root 18 Jul 3 03:49 S08iptables -> ../init.d/iptables lrwxrwxrwx 1 root root 18 Jul 3 03:51 S08mcstrans -> ../init.d/mcstrans lrwxrwxrwx 1 root root 14 Jul 3 03:53 S09isdn -> ../init.d/isdn lrwxrwxrwx 1 root root 17 Jul 3 03:51 S10network -> ../init.d/network lrwxrwxrwx 1 root root 16 Jul 3 03:49 S11auditd -> ../init.d/auditd lrwxrwxrwx 1 root root 21 Jul 3 03:51 S12restorecond -> ../init.d/restorecond lrwxrwxrwx 1 root root 16 Jul 3 03:51 S12syslog -> ../init.d/syslog lrwxrwxrwx 1 root root 18 Jul 3 03:49 S13cpuspeed -> ../init.d/cpuspeed
Some files in this directory start with 'S' and others with 'K'. The files starting with 'S' correspond to the scripts that have to be 'started' in that particular runlevel, and the ones with 'K' correspond to the ones that are to be 'killed'. These files are just soft links to scripts under "/etc/rc.d/init/d" directory (One soft link points to "/etc/rc.local" which itself is a soft link to "/etc/rc.d/rc/local"). The scripts in "/etc/rc.d/init.d/" are daemons. Daemons are the processes that run in the background and provide some kind of service. For instance, http daemon (httpd) provides web service.
The "/etc/rc.local" script runs after all these scripts have been executed. If there is a command or script that you want to be executed at system startup, you can put it in this script.
[root@redhat-server ~]# cat /etc/rc.local #!/bin/sh # # This script will be executed *after* all the other init scripts. # You can put your own initialization stuff in here if you don't # want to do the full Sys V style init stuff. touch /var/lock/subsys/local
Now that all these scripts have been executed successfully, login prompt appears.
Red Hat Enterprise Linux Server release 5.6 (Tikanga) Kernel 2.6.18-238.el5 on an i686 redhat-server login:
Bootloader Grub Version 2 Break Down Linux Boot Process Steps
1.Hardware Power Up , CPU gets into real mode and jumps to fixed location 0xFFFF0 i.e., hardwired in the CPU circuit. 2.BIOS program (ROM) stored in that location will be executed 3.BIOS code will select the boot device (hard disk , CD ROM , Floppy , USB etc ) that we have configured. 4.Once BIOS undestands the boot device , BIOS code will load the MBR to RAM. grub-install is the program that write data into MBR. 5.BIOS code jumps to the start address of the loaded MBR 6.The core.img from /boot/grub/core.img is loaded in memory. Once core.img has been loaded it can use search.mod to find all further /boot/grub files 7.The executed core.img code now initialises all the modules that are built into it (linked into core.img); one of these modules will be a filesystem driver capable of reading the filesystem on which directory /boot/grub lives. 8.You can link config file into core.img if you need booting through pxe , remote nfs etc. In normal cases, it is not required. 9.Core.img now loads file /boot/grub/normal.mod dynamically from disk , this module provides an interactive command-line. 10.Grub will load a file /boot/grub/grub.cfg which is a shell script and it will then configure the standard Grub boot menu. 11.In grub.cfg "linux /vmlinuz-linux root=/dev/sda3 ro" , "linux" module loads a kernel into memory
Thanks for reading and please provide your suggestions on the below comment section. You can also download linux boot process pdf version for future reference.