Custom Initramfs

From Gentoo Wiki
Jump to:navigation Jump to:search
This article has been flagged for not conforming to the wiki guidelines. Please help Gentoo out by starting fixing things.

initramfs is a root filesystem that is embedded into the kernel and loaded at an early stage of the boot process. It is the successor of initrd. It provides early userspace which can do things the kernel can't easily do by itself during the boot process.

Using initramfs is optional. By default, the kernel initializes hardware using built-in drivers, mounts the specified root partition, loads the init system of the installed Linux distribution. The init system then loads additional modules and starts services until it eventually presents a log in dialog. This is a good default behavior and sufficient for many users. initramfs is for users with advanced requirements; for users who need to do things as early as possible, even before the root partition is mounted.

Typically, an initramfs is not needed, but may be necessary for:

  • Mounting an encrypted, logical, or otherwise special root partition
  • Providing a minimalistic rescue shell (if something goes wrong)
  • Customize the boot process (e.g., print a welcome message)
  • Load modules necessary to boot (e.g., third party storage drivers)
  • Anything the kernel can't do that's usually handled in user space

Prerequisites

There are countless ways to make an initramfs. An initramfs need not be manually created, there are tools, such as Genkernel or Dracut, that can do the work. With a bit of luck, one of them might work out of the box for the required use-case, and the user need not bother with how initramfs works and what it does. If none of these tools can do the job automatically, their functionality may have to be extended manually, or initramfs could have to be "hand made".

An initramfs contains at least one file called /init. This file is executed by the kernel as the main init process (PID 1). It has to do all the work. In addition, there can be any number of additional files and directories that are required by /init. They are usually files often found on any other root filesystem, such as /dev for device nodes, /proc for kernel information, /bin for binaries, and so on. The structure of an initramfs can be simple, or it can be complicated, depending on use-case requiremnts.

When the kernel mounts the initramfs, the target root partition is not yet mounted, so it can't access any of the files there. That means there is nothing but the initramfs. So everything required must be included in the initramfs. If a shell is required, it must be included in the initramfs. To be able to mount something, a mount utility must be included. To load a module, the initramfs has to provide both the module, as well as a utility to load it. If the utility depends on libraries in order to work, the libraries must be included as well. This seems complicated, and it is, because the initramfs has to function independently.

Basics

This section will show the easy and straightforward way to initramfs creation, to make a functional - albeit minimalistic - initramfs which then can extend according to needs..

Directory structure

Create a basic initramfs directory structure that will later become the initramfs root. For consistency work in /usr/src/initramfs, but any location would do. Adapt accordingly.

root #mkdir --parents /usr/src/initramfs/{bin,dev,etc,lib,lib64,mnt/root,proc,root,sbin,sys}

Device nodes

Most things the initramfs does will require a couple of device nodes to be present, especially the device for the root partition. Throughout this document, /dev/sda1 will be used as example device. Copy basic device nodes from the root filesystem to the initramfs example location:

root #cp --archive /dev/{null,console,tty,sda1} /usr/src/initramfs/dev/

Which devices are needed depends entirely on what the system is going to use initramfs for. Please adapt to the system needs.

Note
More advanced approaches to device nodes are covered in the Dynamic devices section.

Applications

Any binary needed to execute at boot needs to be copied into the initramfs layout. Make sure to also need to copy any libraries that the binaries require. To see what libraries any particular binary requires, use the ldd tool. For example, the dev-util/strace binary requires:

user $ldd /usr/bin/strace
    linux-vdso.so.1 (0x00007fff271ff000)
    libc.so.6 => /lib64/libc.so.6 (0x00007f5b954fe000)
    /lib64/ld-linux-x86-64.so.2 (0x00007f5b958a9000)

Here shows that for /usr/bin/strace to work in the initramfs, In this example, /usr/bin/strace to /usr/src/initramfs/bin, but also /lib64/libc.so.6 and /lib64/ld-linux-x86-64.so.2 to /usr/src/initramfs/lib64. The exception is linux-vdso.so.1 which is provided by the kernel.

Some applications might depend on other files and libraries to work. For example, app-editors/nano needs a terminfo file /usr/share/terminfo/l/linux from sys-libs/ncurses, so copy it to the initramfs as well. To find these dependencies, tools like equery and strace prove to be most helpful.

Busybox

Instead of collecting countless utilities and libraries (and never seeing the end of it), just use busybox. It's a set of utilities for rescue and embedded systems, it contains a shell, utilities like ls, mkdir, cp, mount, insmod, and many more - all in a single binary called /bin/busybox. For busybox to work properly in a initramfs, emerge it with the static USE flag enabled, then copy the /bin/busybox binary into the initramfs layout as /usr/src/initramfs/bin/busybox:

Set the useflags static and -pam to /etc/portage/package.use/busybox

root #nano /etc/portage/package.use/busybox
FILE /etc/portage/package.use/busyboxbusybox
sys-apps/busybox static -pam
root #emerge --ask --verbose sys-apps/busybox
root #cp --archive /bin/busybox /usr/src/initramfs/bin/busybox
Note
Use ldd to verify that the binary is static.

Init

The file structure the initramfs is almost complete. The only thing that is missing is /init itself, the executable in the root of the initramfs that is executed by the kernel. Because sys-apps/busybox includes a fully functional shell, this means that the /init binary can be written as a simple shell script (instead of making it a complicated application written in Assembler or C that needs to compile).

The following example shows a minimalistic shell script, based on the busybox shell:

FILE /usr/src/initramfs/initminimalistic /init example
#!/bin/busybox sh

# Mount the /proc and /sys filesystems.
mount -t proc none /proc
mount -t sysfs none /sys

# Do your stuff here.
echo "This script just mounts and boots the rootfs, nothing else!"

# Mount the root filesystem.
mount -o ro /dev/sda1 /mnt/root

# Clean up.
umount /proc
umount /sys

# Boot the real thing.
exec switch_root /mnt/root /sbin/init

This example needs some device nodes to work, mainly the root block device. Change the script and copy the corresponding /dev/ node to fit the system needs.

Don't forget to make the /init file executable:

root #chmod +x /usr/src/initramfs/init

Packaging

The initramfs now has to be made available to the kernel at boot time. This is done by packaging it as a compressed cpio archive. This archive is then either embedded directly into the kernel image, or stored as a separate file which can be loaded by the bootloader during the boot process. Both methods perform equally well/

Kernel configuration

With either method, there is a need to enable Initial RAM filesystem and RAM disk (initramfs/initrd) support.

KERNEL CONFIG_BLK_DEV_INITRD=y
General setup  --->
    [*] Initial RAM filesystem and RAM disk (initramfs/initrd) support
Warning
Also enable all drivers, filesystems, compression methods and other settings that are required for booting and accessing the root partition. If selecting such drivers as modules, the module files must be collected and integrated into the initramfs and loaded in /init. Generally this means a lot of unnecessary extra work, so just use built-in drivers for now.

Embedding into the Kernel

The initramfs to be embedded into the kernel image, set Initramfs source file(s) to the root of the initramfs, (e.g. /usr/src/initramfs) but this isn't necessary:

KERNEL CONFIG_INITRAMFS_SOURCE="/usr/src/initramfs"
General setup  --->
    (/usr/src/initramfs) Initramfs source file(s)

Now compile the kernel it will automatically put the files into a cpio archive and embed it into the kernel image. There will need to rebuild the kernel any time a changes is made to the initramfs.

Creating a separate file

To use a standalone archive file, adjust the kernel settings accordingly:

KERNEL Support inital ramdisk/ramfs compressed using <compression_method>
General setup  --->
    () Initramfs source file(s)
    [*]   Support initial ramdisk/ramfs compressed using gzip 
    []   Support initial ramdisk/ramfs compressed using bzip2
    []   Support initial ramdisk/ramfs compressed using LZMA 
    []   Support initial ramdisk/ramfs compressed using XZ   
    []   Support initial ramdisk/ramfs compressed using LZO  
    []   Support initial ramdisk/ramfs compressed using LZ4

For this example gzip is sufficient.

Warning
Not supporting a compression method for this example, will cause a kernel panic when booting!

Create a standalone archive file by running the following commands:

root #cd /usr/src/initramfs
root #find . -print0 | cpio --null --create --verbose --format=newc | gzip --best > /boot/custom-initramfs.cpio.gz

This will create a file called custom-initramfs.cpio.gz in /boot. Now instruct the bootloader to load this file along with the kernel.

Using GRUB

In case of GRUB, do this with the initrd line:

FILE /boot/grub/grub.cfgGRUB initrd example
linux 3.12.6-gentoo
initrd custom-initramfs.cpio.gz
Warning
If unfamiliar with bootloader configuration, please refer to GRUB.

In order to make this usable with grub-mkconfig, the filename custom-initramfs.cpio.gz must be included in the GRUB helper scripts:

FILE /etc/grub.d/10_linuxgrub-mkconfig helper script
[...]
initrd_real=
for i in "initrd.img-${version}" "initrd-${version}.img" "initrd-${version}.gz" "custom-initramfs.cpio.gz" \
     "initrd-${version}" "initramfs-${version}.img" \
[...]
FILE /etc/grub.d/10_linux_xengrub-mkconfig helper script
[...]
initrd_real=
for i in "initrd.img-${version}" "initrd-${version}.img" "initrd-${version}.gz" "custom-initramfs.cpio.gz" \
   "initrd-${version}" "initramfs-${version}.img" \
[...]
Note
It is also possible to use the variable ${version} for the filename in the helper script. This means that the file itself has to be renamed as well. The variable content corresponds to the output of uname --kernel-release.

After applying the changes, the file will be recognized running grub-mkconfig; the output may look like the following:

root #grub-mkconfig -o /boot/grub/grub.cfg
Generating grub configuration file ...
Found linux image: /boot/vmlinuz-4.14.83-gentoo
Found initrd image: /boot/custom-initramfs.cpio.gz
done

External file list

An external file list, or cpio list, describes files to be included into the initramfs. This file list is processed by an utility that comes with the Linux kernel, usr/gen_init_cpio. It can be used for both embedded and standalone initramfs, either by using it as INITRAMFS_SOURCE directly or by running the utility from a shell. This lets the initramfs be built dynamically, always using the latest files from the system, but compared to populating a real directory /usr/src/initramfs, it is less intuitive and requires more knowledge in regards to device nodes and such. The command file will prove as very useful to get information from special block or character devices.

Warning
This method is for advanced users. Skip this section, if not comfortable with the next steps.

A minimalistic list may look like so:

FILE /usr/src/initramfs.listcpio list example
# Custom Initramfs minimal example
dir /dev 0755 0 0
file /init /usr/src/initramfs/init 0755 0 0

Create the cpio archive, compress it and move it to /boot:

root # cd /usr/src/linux
root # usr/gen_init_cpio ../initramfs.list > ../initramfs.cpio
root # gzip --best ../initramfs.cpio
root # mv ../initramfs.cpio.gz /boot

After that update the bootloader entries!

Instead of creating the list manually, there is the possibility to let the cpio list and archive be generated by two tools, which are provided by the kernel source files. It might be needed to compile these first; this is documented here.

root #/usr/src/linux/usr/gen_initramfs.sh -h
Usage:
/usr/src/linux/usr/gen_initramfs_list.sh [-o <file>] [-u <uid>] [-g <gid>] {-d | <cpio_source>} ...
        -o <file>      Create compressed initramfs file named <file> using
                       gen_init_cpio and compressor depending on the extension
        -u <uid>       User ID to map to user ID 0 (root).
                       <uid> is only meaningful if <cpio_source> is a
                       directory.  "squash" forces all files to uid 0.
        -g <gid>       Group ID to map to group ID 0 (root).
                       <gid> is only meaningful if <cpio_source> is a
                       directory.  "squash" forces all files to gid 0.
        <cpio_source>  File list or directory for cpio archive.
                       If <cpio_source> is a .cpio file it will be used
                       as direct input to initramfs.
        -d             Output the default cpio list.

All options except -o and -l may be repeated and are interpreted
sequentially and immediately.  -u and -g states are preserved across
<cpio_source> options so an explicit "-u 0 -g 0" is required
to reset the root/group mapping.

Using the previously created directory structure in /usr/src/initramfs, it may be executed like the following:

root #cd /usr/src/linux
root #usr/gen_initramfs.sh -o ../custom-initramfs.cpio /usr/src/initramfs
root #ls -l ..
total 1656
-rw-r--r--  1 root root 1675784 Dec  9 03:12 custom-initramfs.cpio
drwxr-xr-x 12 root root    4096 Dec  9 00:16 initramfs
lrwxrwxrwx  1 root root      20 Nov 23 19:12 linux -> linux-4.14.83-gentoo
drwxr-xr-x 26 root root    4096 Dec  9 00:58 linux-4.14.83-gentoo
-rw-r--r--  1 root root    4830 Dec  9 01:07 README

The script creates a cpio list by analyzing all directories and files within /usr/src/initramfs. It then executes usr/gen_init_cpio to generate the cpio archive file.

Note
The current working directory must be /usr/src/linux, since the shell script uses the relative path usr/gen_init_cpio.

After that, compress the cpio archive via gzip:

root #gzip /usr/src/custom-initramfs.cpio

This creates the archive /usr/src/custom-initramfs.cpio.gz.

In order to comprehend, all steps can be done manually:

root #usr/gen_initramfs.sh /usr/src/initramfs | tee ../initramfs.list

#####################
# /usr/src/initramfs
# Last modified: 1544312413.9519881540

dir /bin 755 0 0
file /bin/busybox /usr/src/initramfs/bin/busybox 755 0 0
dir /dev 755 0 0
nod /dev/console 600 0 0 c 5 1
nod /dev/null 666 0 0 c 1 3
nod /dev/sda1 660 0 6 b 8 1
nod /dev/tty 666 0 5 c 5 0
dir /etc 755 0 0
file /init /usr/src/initramfs/init 744 0 0
dir /lib64 755 0 0
dir /lib 755 0 0
dir /mnt 755 0 0
dir /mnt/root 755 0 0
dir /proc 755 0 0
dir /root 700 0 0
dir /sbin 755 0 0
dir /sys 755 0 0
root #usr/gen_init_cpio ../initramfs.list > ../initramfs.cpio
root #gzip --best ../initramfs.cpio
root #ls -l ..
total 1660
drwxr-xr-x 12 root root    4096 Dec  9 00:16 initramfs
-rw-r--r--  1 root root 1675800 Dec  9 03:44 initramfs.cpio.gz
-rw-r--r--  1 root root     593 Dec  9 03:31 initramfs.list
lrwxrwxrwx  1 root root      20 Nov 23 19:12 linux -> linux-4.14.83-gentoo
drwxr-xr-x 26 root root    4096 Dec  9 00:58 linux-4.14.83-gentoo
-rw-r--r--  1 root root    4830 Dec  9 01:07 README

The file initramfs.cpio.gz can now be moved to /boot.

root #mv ../initramfs.cpio.gz /boot
Warning
Do not forget to update the bootloader entries afterwards!
Note
In order to always have the latest binaries etc. in the initramfs, it is possible to write a shell script[1] which copies the files and calls the above tools. Hardlinks do not work, as the inode will change when the original file gets overwritten, whereas the hardlink points to the old inode. Symbolic links do not work, as they will be recognised as slink and not as file when using gen_initramfs.sh; this will render the initramfs unusable.
Remember: Be absolutely sure that binaries on the system are either compiled statically or have the needed libraries! Package-based USE flags can be set in /etc/portage/package.use.

Finalizing

Now reboot the machine. On boot, the kernel will extract the files from the initramfs archive automatically and execute the /init script, which in turn should then take care of mounting of the root partition and execute the init system of the installed Linux distribution.

Functionality

To extend the /init script with more advanced functionality.

Rescue shell

To be dropped to a rescue shell if an error occurs, add the following function to /init and call it when something goes wrong.

FILE /usr/src/initramfs/initRescue shell functionality
rescue_shell() {
    echo "Something went wrong. Dropping to a shell."
    exec sh
}

In the example below, the rescue_shell will be executed if the root partition fails to mount:

FILE /usr/src/initramfs/initInvoking the rescue shell
mount -o ro /dev/sda1 /mnt/root || rescue_shell

Force entry into the rescue shell

Occasionally, it might be useful to be able to interrupt the boot process and enter the rescue shell. For example, it's useful for root password (or PAM configuration) recovery or for /init debugging itself.

A standard way of doing this is by passing rdinit=/bin/sh or rdinit=/bin/bb kernel command line option. However, this is a rarely known option. It also requires /bin/sh or /bin/bb symlinks/files to be built into initramfs instead of being installed at runtime with busybox install -s. It's possible to call it like rdinit="/bin/busybox sh" but this is even more obscure.

For convenience, it might be useful to implement Dracut-inspired rd.break option used for a similar purpose (legacy name was rdbreak). Similarly, it's possible to handle more widely known init=/bin/sh option in the same manner.

FILE /usr/src/initramfs/initOption to drop to the rescue shell
break_requested() {
    local want_break
    for o in $(cat /proc/cmdline) ; do
        case "$o" in
            rd.break|rdbreak)                                        want_break="yes" ;;
            init=/bin/sh|init=/bin/bb|init=/bin/bash|init=/bin/dash) want_break="yes" ;;
        esac
    done
    echo "${want_break}"
}

if [[ -n "$(break_requested)" ]] ; then
    rescue_shell
fi

umount /proc
umount /sys
umount /dev

exec switch_root /newroot /sbin/init

Dynamic devices

For populating /dev dynamically, use either devtmpfs or mdev. Please note that the kernel can take some time detecting devices (such as external USB drives), so a sleep statement may need adding to the script.

devtmpfs

Provided by the kernel, devtmpfs is designed to offer device nodes during early boot.

KERNEL CONFIG_DEVTMPFS=y
Device Drivers  --->
    Generic Driver Options  --->
        [*] Maintain a devtmpfs filesystem to mount at /dev

Include the following snippet in the /init script to have it mount at boot:

FILE /usr/src/initramfs/initmount devtmpfs
mount -t devtmpfs none /dev

Don't forget to unmount it again in the cleanup phase of the script:

FILE /usr/src/initramfs/initumount devtmpfs
umount /dev

mdev

Although devtmpfs is the preferred solution today, alternatively use mdev, the udev replacement of busybox.

KERNEL CONFIG_UEVENT_HELPER=y
Device Drivers  --->
    Generic Driver Options  --->
        [*] Support for uevent helper

For mdev to work, make /sbin/mdev a symlink to /bin/busybox in the initramfs.

root #ln --symbolic ../bin/busybox /usr/src/initramfs/sbin/mdev

Then add the following snippet to /init, after mounting /proc and /sys:

FILE /usr/src/initramfs/initAdding mdev support to /init
echo /sbin/mdev > /proc/sys/kernel/hotplug
mdev -s

Mount by UUID or label

With Dynamic devices enabled, it may be preferable to use UUID or label to mount the root partition instead of using a static device name. For that purpose, busybox comes with a utility called findfs.

FILE /usr/src/initramfs/initmount using findfs
mount -o ro $(findfs UUID=845b2454-42a3-19ef-6ec5-238a358c365b) /mnt/root
# or
mount -o ro $(findfs LABEL=myroot) /mnt/root

Doing it this way is simple, but it means that the UUID or label is hardcoded in the /init. Alternatively, see Kernel parameters.

Kernel parameters

Using kernel parameters instead of hardcoding device names or UUIDs, there will be a need to parse /proc/cmdline. There are many ways to do so, the following method is just an example to give the general idea. It uses string manipulation of the shell and only supports key=value parameters. Add the following function to /init and call it whenever kernel parameter is needed.

FILE /usr/src/initramfs/initAdding a simple cmdline parser function
cmdline() {
    local value
    value=" $(cat /proc/cmdline) "
    value="${value##* ${1}=}"
    value="${value%% *}"
    [ "${value}" != "" ] && echo "${value}"
}

The function is called with the name of the kernel parameter in question. In the example below, it uses the root parameter to mount the root partition:

FILE /usr/src/initramfs/initMount rootfs by cmdline
mount -o ro $(findfs $(cmdline root)) /mnt/root

It works for both root=/dev/sda1 and root=UUID=845b2454 but will fail when the parameter is missing.

LVM

If the root partition is located on a logical volume, include the LVM binary in the initramfs. Get a static binary by enabling the static USE flag for sys-fs/lvm2. Copy it to the initramfs /sbin directory.

Note
The static LVM binary may also be called /sbin/lvm.static. Use ldd to verify that the binary is static.
root #cp --archive /sbin/lvm /usr/src/initramfs/sbin/lvm

Now, enable the LVM root partition in /init. This example assumes that the volume group is called VG, and the root volume is called root. Replace them with the names the system used when creating the volume.

FILE /usr/src/initramfs/initSetting up the root volume
lvm vgscan --mknodes # creates /dev/mapper/control
lvm lvchange -a ly VG/root
lvm vgscan --mknodes # creates /dev/mapper/VG-root and /dev/VG/root

The root partition may then be called /dev/VG/root or /dev/mapper/VG-root.

Note
Calling vgscan is optional, but recommended, just in case device nodes are missing.

Recent versions of sys-fs/lvm2 rely on sys-fs/udev to create the named LV device nodes, but there is no udev in a simple initramfs. The following choices are available:

  • Use vgscan as shown above (simplest solution)
  • Mount by UUID or label instead of using /dev/VG/root. It works because findfs is happy with just /dev/dm-42
  • Build a LVM binary with the -udev USE flag (specifically for the initramfs only!)
  • Disable udev dependency by including a minimal /etc/lvm/lvm.conf in the initramfs:
FILE /usr/src/initramfs/etc/lvm/lvm.confDisable udev in lvm.conf
devices {
    # Disable scanning udev for md/multipath components.
    # This is required with recent versions of lvm2, even if you use another solution for
    # your LV device nodes; without it lvm commands will stall for minutes waiting for udev.
    multipath_component_detection = 0
    md_component_detection = 0
}
activation {
    # Set to 0 to disable udev synchronisation (if compiled into the binaries).
    udev_sync = 0
    # Set to 0 to disable the udev rules installed by LVM2
    udev_rules = 0
}

Software RAID

Normally the Linux kernel will automatically scan for any "Linux raid autodetect" partitions and start as many software RAIDs as it can find. But if using an initramfs, the kernel will not automatically scan for RAIDs until it is told to. In the following example instructs the kernel to scan for software RAIDs and start as many as it can find. This will actually start all autodetected arrays, not just /dev/md0:

FILE /usr/src/initramfs/initAdding RAID autodetect support to /init
raidautorun /dev/md0
Note
"Linux raid autodetect" won't work for any recent setups, unless specifically set up with partitions of type "fd" and used 0.90 metadata for the Software RAID.

mdadm

Without "Linux raid autodetect" partitions, or if an advanced RAID setup is required, include mdadm in the initramfs. A static binary may be made by enabling the static USE flag for sys-fs/mdadm.

Copy the binary /sbin/mdadm and /etc/mdadm.conf into the initramfs:

root #cp --archive /sbin/mdadm /usr/src/initramfs/sbin
root #cp --archive /etc/mdadm.conf /usr/src/initramfs/etc
Note
Use mdadm --detail --scan if there is not yet a mdadm.conf.

Edit the mdadm.conf in the initramfs as required. An example mdadm.conf follows:

FILE /usr/src/initramfs/etc/mdadm.confmdadm.conf example
DEVICE /dev/sd?*
ARRAY /dev/md0 UUID=627125a5:abce6b82:6c738e49:50adadae

This mdadm.conf will scan all /dev/sd?* devices and assemble the RAID device fitting the UUID 627125a5:abce6b82:6c738e49:50adadae.

Note
If mdadm.conf has additional conditions such as metadata and name, it may be more practical to remove them. The UUID alone should be sufficient.

Now Software RAID can be initialized in /init:

FILE /usr/src/initramfs/initAssemble software RAIDs with mdadm
mdadm --assemble --scan

With this, the root partition /dev/md0 should be able to be mounted.

DM-Crypt

If the root partition is LUKS encrypted, include the cryptsetup binary in the initramfs. A static binary may be made by setting the static USE flag for sys-fs/cryptsetup. Copy it to the initramfs /sbin directory. Since cryptsetup also often requires the use of the kernel's random device, include them as well.

Note
If having problems getting a static cryptsetup binary, try nettle or kernel instead of the default gcrypt USE flag. The udev USE flag may also need to be disabled for both cryptsetup and its dependencies.

Recompile the package sys-fs/cryptsetup with the new USE flags. For example:

root #USE="-gcrypt nettle static" emerge --ask --verbose sys-fs/cryptsetup
These are the packages that would be merged, in order:

Calculating dependencies... done!
[ebuild   R    ] sys-fs/cryptsetup-1.7.5-r1::gentoo  USE="nettle* nls static* udev -gcrypt* -kernel -libressl -openssl -pwquality -python -reencrypt -static-libs -urandom" PYTHON_TARGETS="python2_7 python3_6 -python3_4 -python3_5 (-python3_7)" 0 KiB

Total: 1 package (1 reinstall), Size of downloads: 0 KiB

Would you like to merge these packages? [Yes/No]

It might also be needed to compile the package sys-fs/lvm2 with the static-libs USE flag:

root #USE="static-libs" emerge --ask --verbose sys-fs/lvm2
These are the packages that would be merged, in order:

Calculating dependencies... done!
[ebuild  N     ] sys-devel/autoconf-archive-2018.03.13::gentoo  635 KiB
[ebuild  N     ] dev-libs/libaio-0.3.110::gentoo  USE="static-libs -test" ABI_X86="(64) -32 (-x32)" 42 KiB
[ebuild  N     ] dev-util/boost-build-1.65.0::gentoo  USE="-examples -python -test" PYTHON_TARGETS="python2_7" 80,662 KiB
[ebuild  N     ] dev-libs/boost-1.65.0:0/1.65.0::gentoo  USE="nls static-libs threads -context -debug -doc -icu -mpi -python -tools" ABI_X86="(64) -32 (-x32)" PYTHON_TARGETS="python2_7 python3_6 -python3_4 -python3_5" 0 KiB
[ebuild  N     ] sys-block/thin-provisioning-tools-0.7.0::gentoo  USE="-static -test" 226 KiB
[ebuild  N     ] sys-fs/lvm2-2.02.145-r2::gentoo  USE="readline static-libs thin udev (-clvm) (-cman) -corosync -device-mapper-only -lvm1 -lvm2create_initrd -openais (-selinux) -static -systemd" 0 KiB

Total: 6 packages (6 new), Size of downloads: 81,563 KiB

Would you like to merge these packages? [Yes/No]

After that copy over the binary files:

root #cp --archive /dev/{urandom,random} /usr/src/initramfs/dev
root #cp --archive /sbin/cryptsetup /usr/src/initramfs/sbin/cryptsetup

Now it is possible to unlock the encrypted root partition in /init:

FILE /usr/src/initramfs/initSetting up LUKS encryption in /init
cryptsetup --tries 5 luksOpen /dev/sda1 luksroot

Once the passphrase is entered, the root partition will be available as /dev/mapper/luksroot.

Encrypted keyfile

If encrypted keyfiles are required, use cryptsetup to encrypt them. It keeps the initramfs simple as that's the encryption tool already present - no need to add other binaries. Plus, unlike some of the alternatives, it offers a nice password prompt.

The following example creates a random 512 byte key, encrypted with LUKS, and adds it to the LUKS container /dev/sda1.

Note
* Current versions of cryptsetup use 4096 instead of 2056 blocks for LUKS metadata. With the --align-payload=1 parameter, it is back to 2056 blocks.
  • cryptsetup also offers --keyfile-size and --keyfile-offset options, which can be used for other key sizes or multiple keys in one container.
root #dd if=/dev/zero of=/usr/src/initramfs/root/key.luks count=2057
root #cryptsetup --align-payload=1 luksFormat /usr/src/initramfs/root/key.luks
root #cryptsetup open --type luks /usr/src/initramfs/root/key.luks lukskey
root #dd if=/dev/urandom of=/dev/mapper/lukskey
root #cryptsetup luksAddKey /dev/sda1 /dev/mapper/lukskey

Unlocking the root device using this key in the /init can then be done like this:

FILE /usr/src/initramfs/initLUKS encryption with keyfiles
# Obtain the key
cryptsetup --tries 5 luksOpen /root/key.luks lukskey

# Unlock the root partition
cryptsetup --key-file /dev/mapper/lukskey open --type luks /dev/sda1 luksroot

# Clean up the key
cryptsetup close lukskey

As before, the root partition should then be available as /dev/mapper/luksroot.

Networking

If networking is required in initramfs, all required network related drivers have to be built into the kernel, and the network interfaces must be configured in /init. How exactly this has to be done, depends on the network situation. The following sections cover only the most common cases.

Static IP

If the network situation allows the use of a static network IP, it is possible to set it up using the ifconfig and route commands, both of which are included in Busybox. This is by far the easiest solution, so if it's at all possible, go for it.

FILE /usr/src/initramfs/initStatic network IP setup in /init
ifconfig eth0 10.0.2.15
route add default gw 10.0.2.2

DHCP

To obtain a dynamic IP address from the network's DHCP server, a DHCP client is required. Busybox comes with a minimalistic DHCP client called udhcpc, which is sufficient for most users. Unfortunately, udhcpc has a dependency: it requires the help of a separate script to actually configure the network interface. An example for such a script is included in the Busybox distribution, but it's not installed by Gentoo. It must be obtained directly from the Busybox tarball (it's called examples/udhcp/simple.script) or download it from the Busybox project page.

Copy the script to the initramfs and make it executable.

root #cp simple.script /usr/src/initramfs/bin
root #chmod +x /usr/src/initramfs/bin/simple.script

Edit the script's first line to read #!/bin/busybox sh or create a symlink for /bin/sh:

root #ln --symbolic busybox /usr/src/initramfs/bin/sh

Now, it's possible to obtain a dynamic IP address for eth0 using DHCP:

FILE /usr/src/initramfs/initNetwork setup using DHCP
ifconfig eth0 up
udhcpc -t 5 -q -s /bin/simple.script

DNS

The network should be up and running now. However, that's only if exactly which IPs to talk to is known. If only a host or domain name is known, it's a different story entirely. In that case, it will be required to be able to resolve hostnames. Unfortunately, this is where our luck leaves us. Until now, everything could be done with just the static binary of Busybox - however, this is not the case with DNS.

Note
Additional libraries are required to make DNS work.

This is because sys-libs/glibc itself dynamically includes additional libraries for DNS lookups. As long as that functionality is not required, everything should work, but if it is needed, there is no choice but to include those libraries in the initramfs. The only alternative would be building Busybox against another libc such as sys-libs/uclibc, however that would go beyond the scope of this article.

This is a good chance to demonstrate how to use (dev-util/strace) to reveal hidden dependencies.

user $strace busybox ping wiki.gentoo.org 2>&1 | grep open
open("/etc/nsswitch.conf", O_RDONLY|O_CLOEXEC) = 3
open("/etc/host.conf", O_RDONLY|O_CLOEXEC) = 3
open("/etc/resolv.conf", O_RDONLY|O_CLOEXEC) = 3
open("/etc/ld.so.cache", O_RDONLY|O_CLOEXEC) = 3
open("/lib64/libnss_files.so.2", O_RDONLY|O_CLOEXEC) = 3
open("/lib64/libc.so.6", O_RDONLY|O_CLOEXEC) = 3
open("/lib64/ld-linux-x86-64.so.2", O_RDONLY|O_CLOEXEC) = 3
open("/etc/hosts", O_RDONLY|O_CLOEXEC)  = 3
open("/etc/ld.so.cache", O_RDONLY|O_CLOEXEC) = 3
open("/lib64/libnss_dns.so.2", O_RDONLY|O_CLOEXEC) = 3
open("/lib64/libresolv.so.2", O_RDONLY|O_CLOEXEC) = 3
open("/etc/resolv.conf", O_RDONLY|O_CLOEXEC) = 3

The command accesses quite a lot of files, some of which are mandatory for it to work.

Copy the necessary libraries to the initramfs:

root #cp /lib64/libnss_{dns,files}.so.2 /lib64/{libresolv,ld-linux-x86-64}.so.2 /lib64/libc.so.6 /usr/src/initramfs/lib64

Create a /etc/resolv.conf with at least one useable nameserver. Note that this step may be done automatically if using DHCP.

root #echo nameserver 10.0.2.3 > /usr/src/initramfs/etc/resolv.conf

With this, DNS lookups should now work.

Note
If it still does not work, this bug may be at cause: bug 17250. As a workaround, set LD_LIBRARY_PATH="/lib64" and try again.

Custom keyboard layout

Busybox provides loadkmap to set keyboard layout. Since it only accepts binary keymaps, they must be converted first.

root #cp /usr/share/keymaps/path/to/your/prefered/keymap.map.gz .
root #gzip -d keymap.map.gz
root #loadkeys -b keymap.map > /usr/src/initramfs/keymap.bmap

Clean up:

root #rm keymap.map

Now just update the init-script:

FILE /usr/src/initramfs/initChange keyboard layout
loadkmap < /keymap.bmap

Troubleshooting

The following section tries to provide help with common issues and pitfalls.

Static vs. dynamic binaries

Any custom binaries needed to be used in the initramfs before mounting have to be fully functional, independent from any files installed on the root partition. This is much easier to achieve with static binaries (which usually work as single file) than with dynamic binaries (which need any number of additional libraries to work).

Gentoo provides static binaries for some ebuilds. Check if the ebuild for the binary offers a static or -dynamic USE flag. This is by far the easiest method to get a static binary, but unfortunately only a select few ebuilds support it.

Many applications also offer static builds with an option in their configure scripts. There is no standard name for the option, it may be --enable-static or something similar. When compiling a package manually, check the list of available options by using ./configure --help to see if the package supports building static binaries.

It is possible to check whether or not a binary is static by using the ldd command. The strace command is also very useful to find out about additional dependencies. By using equery files it is possible to see which files a certain package has brought into the system, some of which may also be candidates for additional dependencies of that package.

Note
* See Applications for a ldd usage example.
  • See DNS for a strace usage example.

Including libraries into the initramfs in order to make a dynamic executable work is a last resort only. It makes the initramfs much larger than necessary and harder to maintain, as the dependencies might change with every update of the program in question.

lddtree

If deciding to go with dynamic binaries, app-misc/pax-utils comes with a Python script lddtree which will handle the collection of libraries:

Note
If the --copy-to-tree option is missing, enable the python useflag.
root #lddtree --copy-to-tree /usr/src/initramfs /usr/bin/nano

That will copy the binary and all its libraries to the initramfs structure - but not any of the runtime dependencies. For more details refer to lddtree --help.

Kernel panics

When working with initramfs and writing custom init scripts for it, the following kernel panic may be encountered on boot:

Kernel panic - not syncing: Attempted to kill init!

This is not an error in the kernel, but an error in the /init script. This script is executed as the init process with PID 1. Unlike other processes, the PID 1 init process is special. It is the only process that is started by the kernel on boot. It's the process that takes care of starting other processes (boot process, init scripts) which in turn start other processes (daemons, login prompts, X), which in turn start other processes (bash, window manager, browser, ...). The init process is the mother of all other processes, and therefore it mustn't be killed. On shutdown, it's again the init process that takes care of cleaning up by shutting down other processes first, then running processes that will unmount the filesystems, until it is safe to actually do a shutdown without corrupting anything.

If there is some error in the /init script, that causes the init process to end, this basically means there are no processes left to run, there is nothing that could take care of cleaning up, and the kernel has no choice but to panic. For this reason there are some things in /init that can't be done like in a normal shell script, like using return or exit, or letting the script just run a series of commands and then simply end.

If /init should end, pass the responsibility of the init process to another process using exec. See the examples above how exec is used to either run /sbin/init of the mounted root partition or to run a rescue shell in case something went wrong.

Job control

While working with initramfs, especially the Rescue shell, this message mya be encountered:

/bin/sh: can't access tty; job control turned off

The lack of job control is usually not a problem, since /init is not supposed to be interactive. However, to work with the Busybox shell on a regular basis, being unable to control programs with Ctrl+C or Ctrl+Z can easily become a huge issue. In worst case, if job control is not available, and a program refuses to quit, reboot.

The job control section in the Busybox FAQ offers some help here. Either use:

root #setsid sh -c 'exec sh </dev/tty1 >/dev/tty1 2>&1'

or:

root #setsid cttyhack sh

to start a shell on tty1 with job control enabled.

Salvaging

If for whatever reason the /usr/src/initramfs structure was lost, but either the kernel image with the built-in initramfs, or the separate cpio archive is still available, it's possible to salvage it from there. Although it may be easier to just redo it from scratch - after the first try, doing it again should be a piece of cake. So this is just in case.

Dismantling the Kernel

Skip this step if the initramfs is a separate cpio archive already. Otherwise, it will be required to get the built-in cpio archive out of the kernel image. To do that, dismantle it, which isn't easy, since the kernel image is a combination of boot sector and compressed archive itself. It also depends on the compression is being used for the kernel and for the initramfs. For simplicity, this example assumes bzip2 - however, the principle is the same for other compression methods.

The utility of choice when dismantling kernel images is app-misc/binwalk. It analyzes arbitrary files for known signatures, and prints their offsets. While there are usually a bunch of false matches in the output, it should be easy to pick the correct ones.

user $binwalk bzImage
DECIMAL     HEX         DESCRIPTION
-------------------------------------------------------------------------------------------------------------------
15949       0x3E4D      bzip2 compressed data, block size = 900k
3042172     0x2E6B7C    LZMA compressed data, properties: 0x9A, dictionary size: 4194304 bytes, uncompressed size: 9439488 bytes
4433597     0x43A6BD    LZMA compressed data, properties: 0xD8, dictionary size: 16777216 bytes, uncompressed size: 4213785 bytes
8530175     0x8228FF    ELF (NetBSD)
Note
Newer versions also support binwalk --extract which will extract all found offsets directly.

A less sophisticated method would be to use grep to search for signatures. For bzip2, this is BZh. For gzip, use $'\x1f'$'\x8b'.

user $grep --text --byte-offset --only-matching BZh bzImage
15949:BZh
3946909:BZh

In this case the offset we are looking for is 15949 bytes. Now extract the compressed kernel image:

user $dd if=bzImage bs=15949 skip=1 | bunzip2 > Image

This yeilds the uncompressed kernel image. Somewhere within this image resides the compressed initramfs archive, so just iterate the previous process to find it. Depending on the kernel configuration, it might be another bzip2, gzip, or cpio container.

user $binwalk Image
user $grep --text --byte-offset --only-matching BZh Image

Suppose the offset is 171424 bytes this time. Now extract the initramfs cpio archive:

user $dd if=Image bs=171424 skip=1 | bunzip2 > initramfs.cpio

To verify that a cpio archive was actually retreived from that, use the file command:

user $file initramfs.cpio
initramfs.cpio: ASCII cpio archive (SVR4 with no CRC)

Extracting the cpio archive

If the initramfs cpio archive was a separate file, it needs to be uncompressed first.

user $gunzip initramfs.cpio.gz

To extract the uncompressed initramfs.cpio:

Warning
This will overwrite files in the current directory. Do it in /tmp/initramfs/ or similar.
user $cpio --extract --make-directories --format=newc --no-absolute-filenames < initramfs.cpio

With this, the initramfs structure should have been recovered.

Integrated initramfs does not always update

If the initramfs is integrated into the kernel (instead of using a separate file), there's a possibility that a make bzImage does not actually update it every time. This could result in making changes to the initramfs but actually keep booting using the old, buggy one. In this case, manually delete the integrated image to force the kernel to integrate a fresh initramfs archive:

root #rm /usr/src/linux/usr/initramfs_data.cpio*

Alternatively, run make clean, but then the entire kernel will need to be recompiled.

Command not found

In Gentoo, busybox is configured as standalone shell by default, which allows busybox to execute its own applets directly. Without this setting, Busybox commands (mkdir, mount, ...) won't work unless there is explicitly a symlink created for them. This can be done at the top of the /init script:

FILE /usr/src/initramfs/initInstall Busybox symlinks to /init
#!/bin/busybox sh

# Install symlinks to all busybox applets first.
/bin/busybox mkdir -p /usr/sbin /usr/bin /sbin /bin
/bin/busybox --install -s

# ...everything else below...

Alternatively, create the symlinks directly in /usr/src/initramfs so they will already be included in the initramfs.

root #mkdir -p /usr/src/initramfs/{usr/sbin,usr/bin,sbin,bin}
root #chroot /usr/src/initramfs /bin/busybox --install -s

Disappearing root

If the encrypted root (with cryptsetup/LUKS), for example /dev/mapper/gentoo-root, is disappearing after the switch_root command it is possible to recreate the device by entering:

root #dmsetup mknodes

Variations for switch_root

Some init setups require proc, sys and dev to be mounted before starting up. If having trouble with switch_root in initramfs setup, try replacing the umount command with a mount --move in the init script.

For example, replace this.

FILE /usr/src/initramfs/initumount commands
# Clean up
umount /proc
umount /sys
umount /dev

# Boot the real thing
exec switch_root /newroot /sbin/init

With this.

FILE /usr/src/initramfs/initmount --move commands
# Clean up
mount --move /proc /newroot/proc
mount --move /sys /newroot/sys
mount --move /dev /newroot/dev

# Boot the real thing
exec switch_root /newroot /sbin/init
Important
umount is used to ensure that "real" init systems, like OpenRC, start in a clean state. Ideally, use umount if it is possible in this circumstance.

Examples

See Custom_Initramfs/Examples for fully functional examples of finished /init scripts.

See also

External resources

References