Merikanto

一簫一劍平生意，負盡狂名十五年

LPIC - 105 Booting Linux

Veröffentlicht am in
Symbole im Artikel gezählt: 9.5k Wörter Lesezeit ≈ 12 min

Objective


# Boot Linux

1 - Boot Process

Steps

  • Turn on computer

    Special hardware circuit causes the CPU to look at pre-determined address, and execute firmware stored in that location

  • Firmware

    • Motherboard firmware performs Power-On-Self-Test (POST)
    • Initialize & configure hardware to a known operational state
    • Loads boot loader from boot device (first hard disk)
    • Pass control to boot loader

  • Boot loader loads the OS kernel, or chainload another loader

    • BIOS - MBR
    • EFI - ESP

  • Linux kernel program loads into memory

    • Initializing devices
    • Mount the root partition
    • Load & execute the initial program (/sbin/init)
      /sbin/init has PID = 1, and reads from /etc/inittab

  • Initialization - init starts the rest of the system services via startup scripts (SysV / systemd)

  • Once Linux is running, it doesn’t use BIOS services for I/O. It uses its own drivers to access hardware

Boot without a keyboard

  • Many won’t boot, if keyboard is unplugged
  • Firmware option: Halt On
    Tells firmware in what situation it should refuse to boot

BIOS

BIOS & UEFI are embedded firmware to start the boot process (launch boot loader program)


Notes

  • Boot process begin by reading boot sector (first sector) from disk, then execute the code

  • Boot options for BIOS-based computers are limited

    • Can only select the order in which boot devices are examined, to find the boot sector
    • SCSI disks & SATA disks won’t appear in the main BIOS disk-detection screen

Most important limitation

Can only read one sector’s data from hard drive into memory to run

  • Not enough space to load entire OS

Work-around: Split boot process into 2 parts

  • BIOS runs boot loader program : init. hardware & find and run full OS program

    BIOS finds boot loader program : internal / external hard drive, CD / DVD, USB, ISO, etc.

    Boot loader program usually has a config file

  • Boot loader program points to OS kernel file / another boot loader program

    No size limitation on the kernel boot file


BIOS 2 important functions

  • Configures hardware (built into motherboard & on different plug-in cards)
  • Begins boot process, pass control to the boot loader in MBR

UEFI

UEFI: Unified Extensible Firmware Interface


Check whether system is using EFI:

1
ls /sys/firmware/efi	# no such file: BIOS

Notes

  • Boot process involves read boot loader file from filesystem, on ESP (EFI System Partition)

    • EFI begins boot redirection from the firmware level

    • ESP use FAT filesystem to store boot loader program

    • ESP is mounted in /boot/efi , boot loader file: .efi (stores separate boot loader for each OS)

  • Present extended range of boot options

    e.g. Default boot loader files from multiple boot loaders on the computer’s hard disk,
    or various devices (e.g grant precedence to bootable USB)

  • With UEFI: need to register each boot loader file to appear at boot time

  • No size limit for loading boot loader program: Possible to load OS kernel directly without boot loader

For security reason, better to make first hard disk as the only boot device


2 - Boot Loader

Primary boot loader

  • Examine partition table, locates the bootable partition
  • Or locates OS kernel & execute directly

Boot loader & Boot manager

  • Boot loader: loads kernel into memory, and transfers control to it
  • Boot manager: presents a menu of boot options

Superblock is part of the filesystem. It describes basic filesystem features such as size, status.

On BIOS based computers, superblock can hold a portion of the boot loader, so damaging it can cause boot problems.

  • Fix: fsck to repair filesystem

Grub Legacy

  • Supports BIOS, but not EFI
  • Disk & partition: both indexes start from 0. e.g. (hd0, 0)
  • Doesn’t distinguish between PATA, SATA, SCSI: First SCSI drive is hd0
  • Treat USB as hard disk
  • Config file
    • Two sections: Global definitions & OS boot definitions
    • /boot/grub/menu.lst
    • /boot/grub/grub.conf

Options

  • splashimage: Background image for boot menu

  • default= : which OS to boot

  • timeout=: wait for user input before booting the default OS

  • chainloader: pass control to another boot loader

  • initrd: Initial RAM disk (contains drivers for kernel to interact with system hardware)

    Replaced by initramfs (initial RAM filesystem)


Install

  • Install to the first sector (MBR): grub-install /dev/sda

    Install on a hard disk, not on a partition

Remember to install grub when changes are made to the disk configuration.
e.g. Resizing / moving the GRUB root partition


Grub 2

Boot loader + Boot manager

Grub Manual at GNU website


Notes

  • BIOS + EFI

  • Config file (never edit explicitly):

    • BIOS: /boot/grub/grub.cfg

    • UEFI: /boot/efi/EFI/[distro]

  • Modify: /etc/default/grub (global commands)
  • Support loadable modules for specific filesystems & operation modes
  • Partition indexes start from 1, support GPT: (hd0, gpt2)
  • Files in /etc/grub.d: control particular GRUB OS probers

Options

  • Menuentry: 1st line for each boot definition section

  • initrdefi: initial RAM filesystem (for UEFI system only)

  • Install: grub-install

  • Update: update grub 📌

    1
    2
    # equivalent to
    grub-mkconfig -o > /boot/grub/grub.cfg

Kernel Boot Parameters

Reference: Linux Kernel Archive

Parameter Explanation
debug Enable kernel debugging
init= Execute specific program. e.g. /bin/bash instead of /sbin/init
initrd= Change location of inital RAM filesystem
root= Change root filesystem
ro Mount root filesystem as read-only
rootflags= Set root filesystem’s mount options
single / Single /S / 1 Boot into single-user mode (SysV init system only!)
systemd.unit= Boot to specified target (Systemd system only!)

Other

  • Syslinux
  • LILO (Linux Loader)
  • ELILO
  • rEFIt
  • rEFInd
  • gummiboot

Secure boot

  • Microsoft’s request of firmware feature
  • EFI-based computer will launch a boot loader,
    only if it’s been cryptographically signed with a key whose counterpart is stored in the firmware

Fix damaged boot loader

  • Try to boot a kernel from the hard disk

  • Use live image (e.g. Super GRUB Disk)

    Bootable disc image with options to locate & use the GRUB config file


3 - Boot Info

  • dmesg
  • /var/log

Linux kernel & module log information: Kernel ring buffer

Boot message

  • Kernel ring buffer is stored under /var/log/dmesg

    • Circular buffer, set to a predetermined size
    • Held in memory, cleared & regenerated after each boot
    • If using systemd-journald, then boot messages are stored in a journal. Check with journalctl
  • System logger (syslogd) : /var/log/messages or /var/log/syslog

Interpret boot message

  • Look for hardware type names

    Search for SCSI - Linux treats many disk devices as SCSI disks

  • Hardware chipset name

    8169 for RealTek 8169 Ethernet interface

  • Study the output from a working system

    During boot, reveal message by pressing Esc



# Init: SysV & Systemd

The initialization daemon (init) determines which services are started, in what order

  • SysV: Based on Unix System V initialization daemon

  • Systemd: located in /etc, /bin, /sbin, PID = 1

  • Find init program

    1
    2
    3
    4
    5
    # find init location
    which init

    # check linked file
    readlink -f /sbin/init

1 - SysV & Runlevels


Runlevels

Runlevel 0, 1, and 6 are reserved for special purposes.

Runlevel Meaning
0 Power off (sudo init 0)
1 (or s, S) Single user mode. For low-level maintenance, e.g. resizing partitions
2 Debian: Multi-user graphic mode
3 Redhat: Multi-user text mode
4 For user customization
5 Redhat: Multi-user graphic mode
6 Reboot (sudo init 6)

/etc/inittab file

  • Format

    1
    id:runlevel:action:process

  • Example

    1
    2
    3
    l0:0:wait:/etc/init.d/rc 0
    ...
    l6:6:wait:/etc/init.d/rc 6

SysV Startup Script

  • Location:

    • /etc/init.d/rc 📌
    • /etc/rc.d/rc

  • Crucial task: Run all the scripts associated with the runlevel

  • Determine current runlevel: runlevel

  • Change runlevel:

    reboot & poweroff are usually symlink to halt

    • halt
    • reboot
    • poweroff

Commands

1
2
3
4
5
6
7
8
9
10
11
# status
service xx status

# show all status
service --status-all

# start
service xx start

# stop
service xx stop

2 - Systemd

Notes

  • Parallel startup
  • Master config: /etc/systemd/system.conf

Commands

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
# reload config file
systemctl reload

# reload config file for running services
systemctl daemon-reload

# configure unit to start at next boot
systemctl enable xx

# check status: is-enabled, is-active, is-failed
systemctl is-active xx

# prevent unit from starting
# to undo: unmask
systemctl mask xx

States

  • enabled - start at system boot
  • disabled - doesn’t start at system boot
  • static - only starts if another unit depends on it

Turn off the less pager for systemctl display:

1
systemctl xx --nopager

Unit Files

A unit defines an action / service / target (group of services). Each unit consists of name, type & config file.
Currently there’re 12 types of systemd unit file:

  1. Automount
  2. Device
  3. Mount
  4. Path
  5. Scope
  6. Service
  7. Slice
  8. Snapshot
  9. Socket
  10. Swap
  11. Target (groups of services that start at system boot)
  12. Timer
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
# show current status of all configured units
systemctl list-units

# filter by type
systemctl list-units --type=target

# show all unit files
systemctl list-unit-files

# display a unit file
systemctl cat xx.service

# get default services
systemctl get-default

# get help
man -k systemd
man systemd.directives

The default.target ensures all required services are launched at system init.

  • graphical.target - GUI
  • multi-user.target - Text
  • runlevel[n].target - backward compatibility to legacy SysV (n = 1 to 5)

Directory location for a unit file is critical, due to precedence override.

1
2
3
/etc/systemd/system		# priority 1
/run/systemd/system		# priority 2
/usr/lib/systemd/system		# priority 3

Config Sections

  • [Unit]
    • Requires - If any required unit does not start, this unit won’t start
    • Wants - Still starts, even any required unit fails to start
  • [Install]
    • Alias - Set additional names
    • RequiredBy - Other units that require this service
    • WantBy - Other units that want thos service
  • [Service]
    • ExecReload - Run commands when reload unit
    • ExecStart - Run commands when start unit
    • ExecStop - Run commands when stop unit
    • Environment - Set env var, separated by a space 📌
    • Environment File - Set file that contains env var
    • RemainAfterExit - Set to no / yes.
      If yes, service is left active even process starts with ExecStart is terminated

Special Commands

Determine system’s operational status

1
systemctl is-system-running

Operational status

  • running - everything is in full working order
  • degraded - has one or more failed units
  • maintenance - emergency / recovery mode
  • initializing - start to boot
  • starting - still booting
  • stopping - shut down in progress

Other commands

1
2
3
4
5
6
# find failed units
systemctl --failed

# jump between system targets
# jump from default (graphical) to multi-user (text)
systemctl isolate multi-user.target

Special targets

  • Rescue target
  • Emergency target: systemctl isolate emergency

3 - Notification

View message status: mesg


Tools for notifying users of system change

Location Explanation
/etc/issue Display texts on tty terminal login screen (Before login to the system)
/etc/issue.net Logon screen messages for remote login
/etc/motd Display text after logged into tty terminal
/bin/notify GUI message (via notify-send utility)
/bin/wall Wall message. Logged into tty terminal with GUI terminal emulator (e.g. iTerm)

systemctl will send a wall message after the following commands

  • halt / power-off
  • reboot
  • emergency
  • rescue


# Virtualization

1 - VM

Virtualizing an application typically doesn’t make the performance faster.

Two types of hypervisor

  • Type 1 - Baremetal

    No need for host OS. e.g. Xen, Hyper-V, KVM (Linux built-in)

  • Type 2 - On top of Host OS

    Need to install host OS first. e.g. VirtualBox, VMware


Create VM

  • Clone
  • OVF
  • Template

Clone

Some hypervisors do not issue a new NIC MAC address when cloning VM.
List of items that may need modification:

  • Host name
  • NIC MAC address
  • NIC IP (if using static IP)
  • Machine ID, UUID

OVF (Open Virtualization Format)

  • Export VM to OVF format, to use in other hypervisors
  • Single compressed archive file: OVA (Open Virtualization Archive)
  • Developed by DMTF (Distributed Management Task Force)

Template

  • A VM template is a master copy: Template itself is not bootable
  • Scan current system & create VM out of it: P2V (physical-to-virtual)
  • Manage VMs with shellscripts: virsh 📌

Linux Extension Support

Hardware extension

  • Based on the system’s CPU (require 64-bit)
  • Grants hypervisor direct access to CPU
  • Check BIOS if virtualization is enabled

Commands

  • Check hardware extension

    If the flag is hypervisor, means that the OS is not running on a physical machine. It’s a VM
    Use virt-what to check which hypervisor is being used.

    1
    2
    3
    # Intel: vmx
    # AMD: svm
    grep ^flags /proc/cpuinfo

  • Check if module is loaded

    1
    2
    3
    4
    5
    # check if kvm is loaded
    lsmod | grep -i kvm

    # load module
    sudo modprobe kvm-amd

2 - Others

Containers

  • A container is managed by container engine (e.g. LXC).

Block Storage

  • Underlying hardware: Disk drivers in RAID configuration

Cloud-init (Canonical)

  • Tool that applies user data to your instances