Learn Linux, 101: Boot the system

Go from BIOS to running Linux system

Learn to guide your Linux system through the boot process. You can use the material in this tutorial to study for the LPIC-1 101 exam for Linux system administrator certification or just to learn about the boot process. The material corresponds to the LPI April 2015 Version 4.0 objectives.

Share:

Ian Shields, Linux Author, Freelance

Ian ShieldsIan Shields is a freelance Linux writer. He retired from IBM at the Research Triangle Park, NC. Ian joined IBM in Canberra, Australia, as a systems engineer in 1973, and has worked in Montreal, Canada, and RTP, NC in both systems engineering and software development. He has been using, developing on, and writing about Linux since the late 1990s. His undergraduate degree is in pure mathematics and philosophy from the Australian National University. He has an M.S. and Ph.D. in computer science from North Carolina State University. He enjoys orienteering and likes to travel.



25 August 2015 (First published 20 September 2011)

Also available in Chinese Japanese

Learn more. Develop more. Connect more.

The new developerWorks Premium membership program provides an all-access pass to powerful development tools and resources, including 500 top technical titles (dozens specifically for Java developers) through Safari Books Online, deep discounts on premier developer events, video replays of recent O'Reilly conferences, and more. Sign up today.

Overview

This tutorial helps you understand the boot sequence from BIOS to boot completion, and shows you how to:

  • Give common commands to the boot loader
  • Give options to the kernel at boot time
  • Check boot events in the log files

This tutorial helps you prepare for Objective 101.2 in Topic 101 of the Linux Professional Institute's Linux Server Professional (LPIC-1) exam 101. The objective has a weight of 3.

I focus mainly on the classic SysVinit process and the newer systemd. In addition, I describe upstart.


Booting a computer

When you turn on a computer or restart it, the computer must load an operating system before you can do any useful work. This process is called booting the computer. The computer pulls itself up by the bootstraps, using a boot loader. The process is mostly automatic, but there are times when you might need to intervene, for example to boot installation media, boot to a repair mode, or boot a different operating system to your usual default. I show you how Linux booting works with the common Linux boot loaders and how you can interact with and then check the process.

About this series

This series of tutorials helps you learn Linux system administration tasks. You can also use the material in these tutorials to prepare for the Linux Professional Institute's LPIC-1: Linux Server Professional Certification exams.

See "Learn Linux, 101: A roadmap for LPIC-1" for a description of and link to each tutorial in this series. The roadmap is in progress and reflects the version 4.0 objectives of the LPIC-1 exams as updated April 15th, 2015. As tutorials are completed, they will be added to the roadmap.

Some aspects of the boot process are common to most systems, but some hardware-related aspects are specific to a particular architecture. The material in this tutorial is directed specifically at x86 and x86_64 architecture systems using BIOS to boot the system. A newer system using Extensible Firmware Interface (EFI) and the GUID Partition Table (GPT) was developed for the Intel Itanium and IA64 architectures to overcome several limitations of the 16-bit x86 BIOS architecture, especially for the large servers that Itanium was intended for. Intel stopped developing EFI in 2005 at version 1.10 and contributed it to the Unified Extensible Firmware Interface Forum, which now manages it as Unified Extensible Firmware Interface (UEFI). UEFI is gaining popularity, particularly for systems having drives larger than 2TB in size. It is also required for computers running Windows® 8.

Prerequisites

To get the most from the tutorials in this series, you should have a basic knowledge of Linux and a working Linux system on which you can practice the commands covered in this tutorial. Sometimes different versions of a program format output differently, so your results might not always look exactly like the listings and figures shown here. In particular, BIOS settings vary widely between systems, and boot loader splash screens vary widely between distributions.

You should also be familiar with the material in the tutorial, "Learn Linux 101: Hard disk layout."

Unless otherwise noted, the examples in this tutorial use Fedora 22, with a 4.0.4 kernel. Your results on other systems might differ.


Boot sequence

Before we get into boot loaders, such as LILO, GRUB and GRUB2, let's review how a PC has historically started, or booted. Code called BIOS (for Basic Input Output Service) is stored in non-volatile memory such as a ROM, EEPROM, or flash memory. When the PC is turned on or rebooted, this code is executed and performs a power-on self test (POST) to check the machine. It also determines the boot drive from the available removable or fixed storage devices and loads the first sector from the Master Boot Record (MBR) on that drive. The drive can be a traditional hard drive, solid-state drive, USB stick or drive, or a drive with removable media such as diskette, CD, or DVD. For the remainder of this tutorial, we'll focus on hard drives, but the process is similar for the other types of storage devices.

As discussed in the tutorial "Learn Linux 101: Hard disk layout," the MBR also contains the partition table, so the amount of executable code in the MBR is less than 512 bytes, which is not very much code. Note that every disk, even a floppy, CD, or DVD, or solid-state device such as a USB stick, contains executable code in its MBR, even if the code is only enough to put out a message such as "Non-bootable disk in drive A:". This code that is loaded by BIOS from this first sector is called the first stage boot loader, or the stage 1 boot loader.

The standard hard drive MBR used by MS DOS, PC DOS, and Windows operating systems checks the partition table to find a primary partition on the boot drive that is marked as active, loads the first sector from that partition, and passes control to the beginning of the loaded code. This new piece of code is also known as the partition boot record. The partition boot record is actually another stage 1 boot loader, but this one has just enough intelligence to load a set of blocks from the partition. The code in this new set of blocks is called the stage 2 boot loader. As used by MS-DOS and PC-DOS, the stage 2 loader proceeds directly to load the rest of operating system. This is how your operating system pulls itself up by its bootstraps until it is up and running.

This works fine for a system with a single operating system. What happens when you want multiple operating systems, say OS/2, Windows 7, and three different Linux distributions? You could use some program (such as the DOS FDISK program) to change the active partition and reboot. This is cumbersome. Furthermore, a disk can have only four primary partitions, and the standard MBR can have only one active primary partition; it cannot boot from a logical partition. But our hypothetical example cited five operating systems, each of which needs a partition. Oops!

Typical solutions used some special code that allowed a user to choose which operating system to boot. Several solutions have been used over the years. Examples include:

Loadlin
A DOS executable program that is invoked from a running DOS system to boot a Linux partition. This was popular when setting up a multiboot system was a complex and risky process.
OS/2 Boot Manager
A program that was installed in a small dedicated partition. The partition was marked active, and the standard MBR boot process started the OS/2 Boot Manager, which has menu from which you can choose which operating system to boot.
A smart boot loader
A program that can reside on an operating system partition and is invoked either by the partition boot record of an active partition or by the master boot record. Some common Linux boot loaders are:
  • LILO, the LInux LOader
  • GRUB, the GRand Unified Boot loader (now referred to as GRUB Legacy)
  • GRUB2, a newer boot loader that is installed in many common distributions
  • Syslinux, a group of lightweight boot loaders for MS-DOS FAT filesystems (SYSLINUX), network booting (PXELINUX), bootable "El Torito" CD-ROMs (ISOLINUX), and Linux ext2/ext3/ext4 or btrfs filesystems (EXTLINUX)

This tutorial focuses mainly on GRUB2 with discussion of GRUB and LILO mostly for historical perspective.

Evidently, if you can pass control of the system to some program that has more than 512 bytes of code to accomplish its task, then it isn't too hard to allow booting from logical partitions, or booting from partitions that are not on the boot drive. All of these solutions allow such possibilities, either because they can load a boot record from an arbitrary partition, or because they have some understanding of what file or files to load to start the boot process.

Chain loading

When a boot manager gets control, one thing that it can load is another boot manager. This is called chain loading, and it most frequently occurs when the boot manager that is located in the MBR loads the boot loader that is in a partition boot record. This is almost always done when a Linux boot loader is asked to boot a Windows or DOS partition, but it can also be done when the Linux boot loader for one system is asked to load the boot loader for another system. For example, you might use LILO in one partition to chain load GRUB in another partition in order to access the GRUB menu for that partition.

EFI and UEFI

As noted in the introduction, EFI was developed by Intel to overcome several limitations of the x86 BIOS architecture, notably the 16-bit especially for the large servers that Itanium was intended for. Intel stopped development of EFI in 2005 at version 1.10 and contributed it to the Unified Extensible Firmware Interface Forum, which now manages it as Unified Extensible Firmware Interface (UEFI). Since 2013, the UEFI Forum has also managed the Advanced Configuration and Power Interface or ACPI, a specification for device configuration and power management.

In contrast to 16-bit BIOS, UEFI is either 32-bit or 64-bit. It is possible to have a 32-bit UEFI on a 64-bit processor (for example some Intel Atom processors used in many tablets and small notebook computers). The boot loader needs to match the UEFI: a 32-bit boot loader on 32-bit UEFI and a 64-bit boot loader on 64-bit UEFI. Many, but not all, UEFI implementations allow booting in either UEFI mode or in Legacy BIOS mode. Such systems allow booting with a boot loader that does not support UEFI. However, if the system only supports UEFI, you need an appropriate UEFI boot loader, such as GRUB2.

Linux boot loaders

From here on, we focus on GRUB2, the boot loader included with most newer Linux distributions. LILO was used on many early Linux distributions. GRUB is newer. The original GRUB has now become GRUB Legacy, and GRUB2 is the newer version that is developed under the auspices of the Free Software Foundation (see Resources for details). A new version of LILO is called ELILO (which was developed for booting systems that use EFI. However, it is no longer in active development. See Resources for additional information about ELILO.

To summarize the boot process for PCs:

  1. When a PC is turned on, the Basic Input Output Service (BIOS) performs a self test.
  2. When the machine passes its self test, the BIOS loads the MBR, usually from the first 512-byte sector of the boot drive. The boot drive is most often the first hard drive on the system, but might also be a diskette, CD, or USB key.
  3. For a hard drive, the MBR loads a stage 1 boot loader, which is typically either the LILO or GRUB stage1 boot loader on a Linux system. This is another 512-byte, single-sector record.
  4. The stage 1 boot loader usually loads a sequence of records called the stage 2 boot loader (or sometimes the stage 1.5 loader).
  5. The stage 2 loader loads the operating system. For Linux, this is the kernel and possibly an initial RAM disk (initrd or initramfs).

Refer to the companion tutorial "Learn Linux, 101: Boot managers" if you need to review boot loader installation or basic booting.

To influence your system's boot process, you can:

  1. Change the device from which you boot. If you normally boot from a hard drive, you might need to boot from a floppy disk, a USB memory key, a CD or DVD, or a network. Setting up such alternate boot devices requires your BIOS to be appropriately configured and can require a particular keystroke during boot to display choices. Figure 1 illustrates the choices on one of my systems. The method for setting up boot devices and selecting a boot device at startup is specific to your system and its BIOS. It is also beyond the scope of this LPI objective's requirements, so consult your system documentation.
    Figure 1. Choosing a boot device
    Screen capture of the Startup Device Menu, showing multiple boot device selections, of which the Optiac DVD-RW device is the highlighted selection
  2. Interact with the boot loader to select which of several possible configurations to boot or to edit the boot configuration. You learn how to do this for GRUB2 in this tutorial.
  3. Pass parameters to the kernel to control the way that your kernel starts the system once it has been loaded by the boot loader.

You can find additional examples for LILO and GRUB in the companion tutorial "Learn Linux, 101: Boot managers"


The GRUB2 boot menu

When GRUB2 loads, it presents information from a configuration file, normally located in /boot/grub/grub.cfg or /boot/grub2/grub.cfg. Listing 1 shows a small part of a grub.cfg file for a Fedora 22 system. I broke some lines for publication and indicate these with a trailing backslash (\).

Listing 1. GRUB2 configuration menu entry to boot Fedora 22
### BEGIN /etc/grub.d/10_linux ###
menuentry 'Fedora (4.0.6-300.fc22.x86_64) 22 (Twenty Two)' --class fedora \
--class gnu-linux --class gnu --class os --unrestricted $menuentry_id_option \
'gnulinux-4.0.6-300.fc22.x86_64-advanced-7aefe7a0-97d5-45ec-a92e-00a6363fb1e4' {
	load_video
	set gfxpayload=keep
	insmod gzio
	insmod part_msdos
	insmod ext2
	set root='hd0,msdos5'
	if [ x$feature_platform_search_hint = xy ]; then
	  search --no-floppy --fs-uuid --set=root --hint-bios=hd0,msdos5 \
        --hint-efi=hd0,msdos5 --hint-baremetal=ahci0,msdos5  \
        7aefe7a0-97d5-45ec-a92e-00a6363fb1e4
	else
	  search --no-floppy --fs-uuid --set=root 7aefe7a0-97d5-45ec-a92e-00a6363fb1e4
	fi
	linux16 /boot/vmlinuz-4.0.6-300.fc22.x86_64 \
        root=UUID=7aefe7a0-97d5-45ec-a92e-00a6363fb1e4 ro rhgb quiet 
	initrd16 /boot/initramfs-4.0.6-300.fc22.x86_64.img
}

When GRUB2 reads the configuration file, it normally presents a menu such as the one shown in Figure 2. Typically, you have a short timeout, after which a default entry is booted. The default entry is highlighted, and you might see a timer counting down. Here, you see the menu entry from the previous listing as the default choice. If no action is taken, it boots in 35 seconds.

Figure 2. Selecting a boot choice in GRUB2
Selecting a boot choice in GRUB2

When the menu is displayed, you can press Enter to boot the selected entry immediately, or you can press another key to stop the timeout. You can then highlight and boot another entry, enter e to edit a selected entry, or c to enter a GRUB2 command prompt. Note: if your timeout is set to 0, GRUB2 proceeds immediately to boot your system. In this case, you need to boot rescue media from another device. See the companion tutorial "Learn Linux, 101: Boot managers" for details. That tutorial also covers use of the GRUB2 command line, so I won't cover that in this tutorial.

Let's now look at kernel parameters and the init process, and then I show you how to edit GRUB2 entries to pass parameters.


Kernel parameters

Kernel parameters (sometimes called boot parameters) are used to supply the kernel with information about hardware parameters that it might not determine on its own, to override values that it might otherwise detect, or to avoid detection of inappropriate values. Some older kernel levels required a parameter to enable large memory support on systems with more than a certain amount of RAM. You might want to boot in single-user mode to repair your system, boot an SMP system in uniprocessor mode, or specify an alternate root filesystem. Or you might need to specify a previously working kernel so that you can find out why your newly built custom kernel won't boot. You can accomplish all of these tasks using boot parameters.

Boot parameters are supplied on the kernel command line. Most have the following form:

name[=value_1][,value_2]...[,value_10]

Where 'name' is a unique keyword that identifies what part of the kernel should receive the associated values. The parameters are checked in the following order:

  1. The kernel checks to see if the argument is any of the special arguments 'root=', 'nfsroot=', 'nfsaddrs=', 'ro', 'rw', 'debug' or 'init'.
  2. It walks a list of setup functions to see if the specified argument string has been associated with a setup function.
  3. Anything of the form 'foo=bar' that is not accepted as a setup function is then interpreted as an environment variable to be set.
  4. Any remaining arguments that were not picked up by the kernel or interpreted as environment variables are then passed onto process one, which is usually the init program.

You can find out more about boot parameters from the man pages using info bootparam or man bootparam. Up-to-date information is also supplied in the kernel-parameters.txt file, which should be available in the kernel-doc package. Note: At the time of writing, Fedora 22 does not appear to have a kernel-doc package or kernel-parameters file.

Suppose you want to modify kernel parameters or add new ones or otherwise change the values from the menu entry in the grub.cnf file. When you see the GRUB2 menu, you can edit the entry that you want to modify by selecting it and then pressing e. You then see a screen like Figure 3, where you recognize information from the first few lines of Listing 1.

Figure 3. Editing the Fedora 22 GRUB2 menu entry

The most common argument that is passed to the init process is the word 'single' which instructs it to boot the computer in single user mode, and not launch all the usual daemons. You typically use this for some kind of rescue operation. If you move the cursor down, the screen will scroll and you will see the line that starts with 'linux16' and ends with 'ro rhgb quiet'. For our example, we'll position the cursor at the end of this line and then use the backspace key to erase the words 'quiet', then 'rhgb'. This will stop the Red Hat Graphical Boot screen that fedora normally display during boot and also stop suppressing many of the messages that are normally generated. Finally we'll add the word 'single' to the line to boot in single user mode. The display is shown in Figure 4.

Figure 4. Setting Fedora 22 to boot in single user mode

Now press Ctrl-x to boot the system into single user mode. After several messages scroll by, your screen has a few lines like Listing 2.

Listing 2. Fedora 22 in single user mode
Welcome to emergency mode! After logging in, type "journalctl -xb" to view
system logs, "systemctl reboot" to reboot, "systemctl default" or ^D to 
boot into default mode.
Give root password for maintenance
(Or press ^D to continue):

You can now give the root password and do whatever you need to rescue the system. You can then reboot, or continue into your normal mode of running.

Exactly how single user mode appears differs by system. Some systems take you directly to a root prompt with no need to enter a password. Others, like our example, require a password. How you proceed also differs according to whether your system uses the classic System V init process or the newer upstart or systemd.

Now let's look more at the different init processes.


The init process

When the kernel finishes loading, it usually starts /sbin/init. This program remains running until the system is shut down. It is always assigned process ID 1, On systems using the traditional System V init process, or on systems using the newer Upstart initialization, the process will be named init. However, on systems using the Systemd initialization process, /sbin/init is a link to the systemd init process in /lib/systemd/systemd. We illustrate this for Fedora 22 in Listing 3.

Listing 3. The init process
[ian@atticf22 ~]$ ls -l /sbin/init
lrwxrwxrwx. 1 root root 22 Jun  9 09:16 /sbin/init -> ../lib/systemd/systemd
[ian@atticf22 ~]$ ps --pid 1
  PID TTY          TIME CMD
    1 ?        00:00:02 systemd

In normal operation, a traditional System V init process runs in one of 4 possible multiuser runlevels, 2 through 5. Different systems use runlevels differently, but the highest one used is usually for graphical operation with the X Windows system. The file /etc/inittab controls which processes are started at each runlevel. Systems using upstart have a concept of jobs, while systems with systemd use units to define what things are started and each maps appropriate jobs or units to the traditional System V runlevels. This allows many commands such as telinit to have meaning with the new startup processes. Many of the systemd units are stored in /usr/lib/systemd/system/. Listing 4 shows the mapping from runlevel-equivalent targets to the corresponding systemd targets.

Listing 4. Systemd runlevel target units
[ian@atticf22 l-lpic1-101-2]$ cd /usr/lib/systemd/system/
[ian@atticf22 system]$ ls -l runlevel*.target
lrwxrwxrwx. 1 root root 15 Jun  9 09:16 runlevel0.target -> poweroff.target
lrwxrwxrwx. 1 root root 13 Jun  9 09:16 runlevel1.target -> rescue.target
lrwxrwxrwx. 1 root root 17 Jun  9 09:16 runlevel2.target -> multi-user.target
lrwxrwxrwx. 1 root root 17 Jun  9 09:16 runlevel3.target -> multi-user.target
lrwxrwxrwx. 1 root root 17 Jun  9 09:16 runlevel4.target -> multi-user.target
lrwxrwxrwx. 1 root root 16 Jun  9 09:16 runlevel5.target -> graphical.target
lrwxrwxrwx. 1 root root 13 Jun  9 09:16 runlevel6.target -> reboot.target

Traditionally single user mode was equivalent to runlevel 1. Indeed, you could have typed '1' instead of 'single' to bring the system up in single user mode (runlevel 1). Similarly '3' would bring the system up in full multiuser mode with networking but without the X Windows system while '5' would bring the system up to the full graphical desktop. You can see from the above symbolic links that systemd now uses the rescue.target unit to bring the system up in single user mode. So specifying:

  • single
  • 1
  • s
  • systemd.unit=runlevel1.target
  • systemd.unit=rescue.target

are all equivalent ways of getting to single user mode. You can find out more about systemd units and targets in the man pages. Try info systemd or man systemd.

Whichever initialization you use, the initialization program boots the rest of your system by running a series of scripts. For System V init, these scripts typically live in /etc/rc.d/init.d or /etc/init.d. They perform services such as setting the system's hostname, checking the filesystem for errors, mounting additional filesystems, enabling networking, starting print services, and so on. When the scripts complete, init starts a program calledgetty, which displays the login prompt on consoles. Graphical login screens are handled with a graphical display manager, such as GDM for Gnome.

See the companion tutorial "Learn Linux, 101: Runlevels, boot targets, shutdown, and reboot" for more information on the different startup processes.

If your system loads a kernel but cannot run init successfully, you can try to recover by specifying an alternate initialization program. For example, specifying init=/bin/sh boots your system into a shell prompt with root authority, from which you might be able to repair the system.

Needless to say, if you have to apply the same set of additional parameters every time you boot, you should add them to the configuration file.


Initial RAM disk

You might have noticed the last couple of lines in our GRUB2 example configuration, shown again in Listing 5. The linux16 line specifies what kernel to load and the initrd16 lines specifies ad Initial RAM Disk. You've already seen how to modify the kernel parameters, but what is an initial RAM disk?

Listing 5.
linux16 /boot/vmlinuz-4.0.6-300.fc22.x86_64 \
root=UUID=7aefe7a0-97d5-45ec-a92e-00a6363fb1e4 ro rhgb quiet 
initrd16 /boot/initramfs-4.0.6-300.fc22.x86_64.img

Back at the dawn of UNIX, devices were few and kernels were small. Most devices were physically attached to a computer all the time and a kernel was typically built to support exactly the installed hardware. As time went by and the number and type of devices proliferated, device support was frequently shifted into loadable kernel modules. Kernel modules are like a library of functions with well-defined entry points that allow a kernel to use a device whose support was not compiled in to the kernel. This allowed a kernel to be loaded and then load support only for devices that were actually installed in the system. But without full disk and filesystem support, how do you find and load these modules from disk?

The answer is the initial RAM disk or initrd or initramfs, which is a file containing loadable kernel modules. It is usually compressed and is loaded into RAM by the boot loader. The kernel is given access to it as if it were a mounted file system. If you examine an initrd file using the file command, you can find what compression it uses. If you then uncompress a copy of it, you usually can find a cpio archive inside. On some systems (such as Fedora since Fedora 12), the initrd file is created with a utility called dracut. In this case, it appears as a cpio archive, but if you unpack it, you might not find very much: much less than the initrd file size would suggest. You can use the lsinitrd command to list the contents of an initrd file.


Boot events

During the Linux boot process, a large number of messages are emitted to the console, describing the kernel being booted, the hardware of your system, and other things related to the kernel. These messages usually flash by quickly, and you probably won't be able to read them, unless there is a delay while the boot process waits for something, such as inability to reach a time server, or a filesystem that must be checked. With the advent of the Linux Bootsplash project (see Resources), these messages might be superimposed on a graphical background, or they might be hidden and replaced by a simple status bar. If your distribution supports the hidden mode, you will usually be able to switch back to displaying boot messages by pressing a key such as F2.

dmesg

It's nice to be able to go back and review the kernel messages. Because standard output is related to a process, and the kernel does not have a process identifier, it keeps kernel (and module) output messages in the kernel ring buffer. You can display the kernel ring buffer using the dmesg command, which displays these messages on standard output. Of course, you can redirect this output to a file for later analysis or forward it to a kernel developer for debugging purposes. Listing 6 illustrates some of the output that you might see. Again, I have broken long lines for publication as indicated by a trailing backslash (\).

Listing 6. Partial dmesg output
[root@atticf22 ~]#  dmesg | head -n 30
[    0.000000] Initializing cgroup subsys cpuset
[    0.000000] Initializing cgroup subsys cpu
[    0.000000] Initializing cgroup subsys cpuacct
[    0.000000] Linux version 4.0.6-300.fc22.x86_64 (mockbuild@bkernel02.phx2.fedoraproject.org) \
(gcc version 5.1.1 20150618 (Red Hat 5.1.1-4) (GCC) ) #1 SMP Tue Jun 23 13:58:53 UTC 2015
[    0.000000] Command line: BOOT_IMAGE=/boot/vmlinuz-4.0.6-300.fc22.x86_64 \
root=UUID=7aefe7a0-97d5-45ec-a92e-00a6363fb1e4 ro single
[    0.000000] tseg: 0000000000
[    0.000000] e820: BIOS-provided physical RAM map:
[    0.000000] BIOS-e820: [mem 0x0000000000000000-0x000000000009ebff] usable
[    0.000000] BIOS-e820: [mem 0x000000000009ec00-0x000000000009ffff] reserved
[    0.000000] BIOS-e820: [mem 0x00000000000e4000-0x00000000000fffff] reserved
[    0.000000] BIOS-e820: [mem 0x0000000000100000-0x00000000cff7ffff] usable
[    0.000000] BIOS-e820: [mem 0x00000000cff80000-0x00000000cff8dfff] ACPI data
[    0.000000] BIOS-e820: [mem 0x00000000cff8e000-0x00000000cffcffff] ACPI NVS
[    0.000000] BIOS-e820: [mem 0x00000000cffd0000-0x00000000cfffffff] reserved
[    0.000000] BIOS-e820: [mem 0x00000000ff700000-0x00000000ffffffff] reserved
[    0.000000] BIOS-e820: [mem 0x0000000100000000-0x000000012fffffff] usable
[    0.000000] NX (Execute Disable) protection: active
[    0.000000] SMBIOS 2.5 present.
[    0.000000] DMI: System manufacturer System Product Name/M3A78-EM, BIOS 2003    10/12/2009
[    0.000000] e820: update [mem 0x00000000-0x00000fff] usable ==> reserved
[    0.000000] e820: remove [mem 0x000a0000-0x000fffff] usable
[    0.000000] e820: last_pfn = 0x130000 max_arch_pfn = 0x400000000
[    0.000000] MTRR default type: uncachable
[    0.000000] MTRR fixed ranges enabled:
[    0.000000]   00000-9FFFF write-back
[    0.000000]   A0000-EFFFF uncachable
[    0.000000]   F0000-FFFFF write-protect
[    0.000000] MTRR variable ranges enabled:
[    0.000000]   0 base 000000000000 mask FFFF80000000 write-back
[    0.000000]   1 base 000080000000 mask FFFFC0000000 write-back

The kernel ring buffer is also used for some events after the system is booted. These include certain program failures and hot-plug events. Listing 7 shows several entries related to plugging in a USB memory key. Note the warning that the volume had not been properly unmounted.

Listing 7. Later events in kernel ring buffer
[root@atticf22 ~]# dmesg | tail -n 21
[68209.847331] usb 4-2.4: new full-speed USB device number 5 using ohci-pci
[68209.928078] usb 4-2.4: device descriptor read/64, error -62
[68225.055771] usb 4-2.4: device descriptor read/64, error -110
[68225.231218] usb 4-2.4: new full-speed USB device number 6 using ohci-pci
[68225.340868] usb 4-2.4: New USB device found, idVendor=19ae, idProduct=2403
[68225.340875] usb 4-2.4: New USB device strings: Mfr=1, Product=2, SerialNumber=3
[68225.340877] usb 4-2.4: Product: MSC
[68225.340879] usb 4-2.4: Manufacturer: ATMEL ASF
[68225.340881] usb 4-2.4: SerialNumber: 123123123123
[68225.343447] usb-storage 4-2.4:1.0: USB Mass Storage device detected
[68225.352501] scsi host15: usb-storage 4-2.4:1.0
[68226.357737] scsi 15:0:0:0: Direct-Access     ATMEL    SD/MMC Card Slot 1.00 PQ: 0 ANSI: 3
[68226.358727] sd 15:0:0:0: Attached scsi generic sg7 type 0
[68226.367683] sd 15:0:0:0: [sdf] 7774208 512-byte logical blocks: (3.98 GB/3.70 GiB)
[68226.377646] sd 15:0:0:0: [sdf] Write Protect is off
[68226.377653] sd 15:0:0:0: [sdf] Mode Sense: 0f 00 00 00
[68226.387610] sd 15:0:0:0: [sdf] No Caching mode page found
[68226.387618] sd 15:0:0:0: [sdf] Assuming drive cache: write through
[68226.442450]  sdf:
[68226.484318] sd 15:0:0:0: [sdf] Attached SCSI removable disk
[68227.882986] FAT-fs (sdf): Volume was not properly unmounted. Some data may be corrupt.\
Please run fsck.

Logging system messages

Once your system has started to the point of running /sbin/init, the kernel still logs events in the ring buffer as you just saw. However, processes use a daemon to log messages. In traditional System V init systems, this is the syslogd daemon, which usually logs to /var/log/messages. Systemd-based systems use the systemd-journald daemon to log messages, which can be inspected using the journalctl command.

/var/log/messages

In contrast to the ring buffer, each syslog line has a timestamp, and the file persists between system restarts. This file is where you should first look for errors that occurred during the init scripts stage of booting.

Most daemons have names that end in 'd'. Listing 8 shows how to see the last few daemon status messages after a reboot. It also shows the last 10 messages of any type. This example is from a CentOS 6 system.

Listing 8. Daemon messages from /var/log/messages
[root@attic4-cent ~]# grep "Jul 22.........[^:]*d\:" /var/log/messages
Jul 22 21:45:23 attic4-cent rsyslogd: [origin software="rsyslogd" swVersion="5.8.10" \
x-pid="2051" x-info="http://www.rsyslog.com"] start
Jul 22 21:45:23 attic4-cent cpuspeed: Enabling ondemand cpu frequency scaling governor
Jul 22 21:45:26 attic4-cent acpid: starting up
Jul 22 21:45:26 attic4-cent acpid: 1 rule loaded
Jul 22 21:45:26 attic4-cent acpid: waiting for events: event logging is off
Jul 22 21:45:27 attic4-cent acpid: client connected from 2423[68:68]
Jul 22 21:45:27 attic4-cent acpid: 1 client rule loaded
Jul 22 21:45:32 attic4-cent abrtd: Init complete, entering main loop
Jul 22 21:45:33 attic4-cent libvirtd: Could not find keytab file: /etc/libvirt/krb5.tab: \
No such file or directory
[root@attic4-cent ~]# tail /var/log/messages
Jul 22 21:46:00 attic4-cent kernel: XFS (sda6): Mounting Filesystem
Jul 22 21:46:00 attic4-cent kernel: XFS (sda6): Ending clean mount
Jul 22 21:46:00 attic4-cent kernel: TECH PREVIEW: btrfs may not be fully supported.
Jul 22 21:46:00 attic4-cent kernel: Please review provided documentation for limitations.
Jul 22 21:46:00 attic4-cent kernel: Btrfs loaded
Jul 22 21:46:00 attic4-cent kernel: device fsid d9cb309d-e15d-4b00-9b2f-e3cc08db412c \
devid 1 transid 52 /dev/sda9
Jul 22 21:46:00 attic4-cent kernel: btrfs: disk space caching is enabled
Jul 22 21:46:00 attic4-cent kernel: BTRFS: couldn't mount because of unsupported \
optional features (140).
Jul 22 21:46:00 attic4-cent kernel: btrfs: open_ctree failed
Jul 22 21:46:47 attic4-cent pulseaudio[3252]: ratelimit.c: 9 events suppressed

Shutdown events are also logged, so if your system is not shutting down cleanly, you might be able to find the cause in the log information that is left from the last shutdown.

You can also find logs for many other system programs in /var/log. For example, you can see the startup log for your X Window system.

journalctl

Along with the large amount of function that has come under the new systemd umbrella is a new journalling function managed by the systemd-journald daemon. If your system is using this daemon, your log is managed by it and you can use the journalctl command to view the messages. Listing 9 shows the first page of output from the journalctl -xb command which displays journal entries from the current boot.

Listing 9. Using journalctl -xb to display log entries from the current boot
-- Logs begin at Tue 2015-06-02 12:38:08 EDT, end at Wed 2015-07-22 22:04:23 EDT
Jul 22 22:02:25 atticf20 systemd-journal[106]: Runtime journal is using 8.0M (ma
Jul 22 22:02:25 atticf20 systemd-journal[106]: Runtime journal is using 8.0M (ma
Jul 22 22:02:25 atticf20 kernel: Initializing cgroup subsys cpuset
Jul 22 22:02:25 atticf20 kernel: Initializing cgroup subsys cpu
Jul 22 22:02:25 atticf20 kernel: Initializing cgroup subsys cpuacct
Jul 22 22:02:25 atticf20 kernel: Linux version 4.0.6-300.fc22.x86_64 (mockbuild@
Jul 22 22:02:25 atticf20 kernel: Command line: BOOT_IMAGE=/boot/vmlinuz-4.0.6-30
Jul 22 22:02:25 atticf20 kernel: tseg: 0000000000
Jul 22 22:02:25 atticf20 kernel: e820: BIOS-provided physical RAM map:
Jul 22 22:02:25 atticf20 kernel: BIOS-e820: [mem 0x0000000000000000-0x0000000000
Jul 22 22:02:25 atticf20 kernel: BIOS-e820: [mem 0x000000000009ec00-0x0000000000
Jul 22 22:02:25 atticf20 kernel: BIOS-e820: [mem 0x00000000000e4000-0x0000000000
Jul 22 22:02:25 atticf20 kernel: BIOS-e820: [mem 0x0000000000100000-0x00000000cf
Jul 22 22:02:25 atticf20 kernel: BIOS-e820: [mem 0x00000000cff80000-0x00000000cf
Jul 22 22:02:25 atticf20 kernel: BIOS-e820: [mem 0x00000000cff8e000-0x00000000cf
Jul 22 22:02:25 atticf20 kernel: BIOS-e820: [mem 0x00000000cffd0000-0x00000000cf
Jul 22 22:02:25 atticf20 kernel: BIOS-e820: [mem 0x00000000ff700000-0x00000000ff
Jul 22 22:02:25 atticf20 kernel: BIOS-e820: [mem 0x0000000100000000-0x000000012f
Jul 22 22:02:25 atticf20 kernel: NX (Execute Disable) protection: active
Jul 22 22:02:25 atticf20 kernel: SMBIOS 2.5 present.
Jul 22 22:02:25 atticf20 kernel: DMI: System manufacturer System Product Name/M3
Jul 22 22:02:25 atticf20 kernel: e820: update [mem 0x00000000-0x00000fff] usable
lines 1-23

As usual, consult the man pages for systemd-journald and journalctl for more information.

This completes your introduction to guiding your Linux system through the boot process.

Resources

Learn

Discuss

  • Get involved in the developerWorks community. Connect with other developerWorks users while exploring the developer-driven blogs, forums, groups, and wikis.

Comments

developerWorks: Sign in

Required fields are indicated with an asterisk (*).


Need an IBM ID?
Forgot your IBM ID?


Forgot your password?
Change your password

By clicking Submit, you agree to the developerWorks terms of use.

 


The first time you sign into developerWorks, a profile is created for you. Information in your profile (your name, country/region, and company name) is displayed to the public and will accompany any content you post, unless you opt to hide your company name. You may update your IBM account at any time.

All information submitted is secure.

Choose your display name



The first time you sign in to developerWorks, a profile is created for you, so you need to choose a display name. Your display name accompanies the content you post on developerWorks.

Please choose a display name between 3-31 characters. Your display name must be unique in the developerWorks community and should not be your email address for privacy reasons.

Required fields are indicated with an asterisk (*).

(Must be between 3 – 31 characters.)

By clicking Submit, you agree to the developerWorks terms of use.

 


All information submitted is secure.

Dig deeper into Linux on developerWorks


static.content.url=http://www.ibm.com/developerworks/js/artrating/
SITE_ID=1
Zone=Linux
ArticleID=1014425
ArticleTitle=Learn Linux, 101: Boot the system
publish-date=08252015