Saturday, December 18, 2010

Linux initial RAM disk (initrd) overview

What's an initial RAM disk?

The initial RAM disk (initrd) is an initial root file system that is mounted prior to when the real root file system is available. The initrd is bound to the kernel and loaded as part of the kernel boot procedure. The kernel then mounts this initrd as part of the two-stage boot process to load the modules to make the real file systems available and get at the real root file system.

The initrd contains a minimal set of directories and executables to achieve this, such as the insmod tool to install kernel modules into the kernel.

In the case of desktop or server Linux systems, the initrd is a transient file system. Its lifetime is short, only serving as a bridge to the real root file system. In embedded systems with no mutable storage, the initrd is the permanent root file system. This article explores both of these contexts.


Anatomy of the initrd

The initrd image contains the necessary executables and system files to support the second-stage boot of a Linux system.

Depending on which version of Linux you're running, the method for creating the initial RAM disk can vary. Prior to Fedora Core 3, the initrd is constructed using the loop device. The loop device is a device driver that allows you to mount a file as a block device and then interpret the file system it represents. The loop device may not be present in your kernel,but you can enable it through the kernel's configuration tool (make menuconfig) by selecting Device Drivers > Block Devices > Loopback Device Support. You can inspect the loop device as follows (your initrd file name will vary):


Listing 1. Inspecting the initrd (prior to FC3)
# mkdir temp ; cd temp
# cp /boot/initrd.img.gz .
# gunzip initrd.img.gz
# mount -t ext -o loop initrd.img /mnt/initrd
# ls -la /mnt/initrd


You can now inspect the /mnt/initrd subdirectory for the contents of the initrd. Note that even if your initrd image file
does not end with the .gz suffix, it's a compressed file, and you can add the .gz suffix to gunzip it.

Beginning with Fedora Core 3, the default initrd image is a compressed cpio archive file. Instead of mounting the file as
a compressed image using the loop device, you can use a cpio archive. To inspect the contents of a cpio archive, use the
following commands:

Listing 2. Inspecting the initrd (FC3 and later)

# mkdir temp ; cd temp
# cp /boot/initrd-2.6.14.2.img initrd-2.6.14.2.img.gz
# gunzip initrd-2.6.14.2.img.gz
# cpio -i --make-directories <>



The result is a small root file system, as shown in Listing 3. The small, but necessary, set of applications are present in the ./bin directory, including nash (not a shell, a script interpreter), insmod for loading kernel modules, and lvm (logical volume manager tools).

Listing 3. Default Linux initrd directory structure
# ls -la
drwxr-xr-x 10 root root 4096 May 7 02:48 .
drwxr-x--- 15 root root 4096 May 7 00:54 ..
drwxr-xr-x 2 root root 4096 May 7 02:48 bin
drwxr-xr-x 2 root root 4096 May 7 02:48 dev
drwxr-xr-x 4 root root 4096 May 7 02:48 etc
-rwxr-xr-x 1 root root 812 May 7 02:48 init
-rw-r--r-- 1 root root 1723392 May 7 02:45 initrd-2.6.14.2.img
drwxr-xr-x 2 root root 4096 May 7 02:48 lib
drwxr-xr-x 2 root root 4096 May 7 02:48 loopfs
drwxr-xr-x 2 root root 4096 May 7 02:48 proc
lrwxrwxrwx 1 root root 3 May 7 02:48 sbin -> bin
drwxr-xr-x 2 root root 4096 May 7 02:48 sys
drwxr-xr-x 2 root root 4096 May 7 02:48 sysroot
#

Of interest in Listing 3 is the init file at the root. This file, like the traditional Linux boot process, is invoked when the initrd image is decompressed into the RAM disk.

The cpio command
Using the cpio command, you can manipulate cpio files. Cpio is also a file format that is simply a concatenation of files with headers. The cpio file format permits both ASCII and binary files. For portability, use ASCII. For a reduced file size, use the binary version.

Tools for creating an initrd
Let's now go back to the beginning to formally understand how the initrd image is constructed in the first place. For a traditional Linux system, the initrd image is created during the Linux build process. Numerous tools, such as mkinitrd, can be used to automatically build an initrd with the necessary libraries and modules for bridging to the real root file system. The mkinitrd utility is actually a shell script, so you can see exactly how it achieves its result. There's also the YAIRD (Yet Another Mkinitrd) utility, which permits customization of every aspect of the initrd construction.

Manually building a custom initial RAM disk

Because there is no hard drive in many embedded systems based on Linux, the initrd also serves as the permanent root file system. Listing 4 shows how to create an initrd image. I'm using a standard Linux desktop so you can follow along without an embedded target. Other than cross-compilation, the concepts (as they apply to initrd construction) are the same for an embedded target.

Listing 4. Utility (mkird) to create a custom initrd

#!/bin/bash

# Housekeeping...
rm -f /tmp/ramdisk.img
rm -f /tmp/ramdisk.img.gz

# Ramdisk Constants
RDSIZE=4000
BLKSIZE=1024

# Create an empty ramdisk image
dd if=/dev/zero of=/tmp/ramdisk.img bs=$BLKSIZE count=$RDSIZE

# Make it an ext2 mountable file system
/sbin/mke2fs -F -m 0 -b $BLKSIZE /tmp/ramdisk.img $RDSIZE

# Mount it so that we can populate
mount /tmp/ramdisk.img /mnt/initrd -t ext2 -o loop=/dev/loop0

# Populate the filesystem (subdirectories)
mkdir /mnt/initrd/bin
mkdir /mnt/initrd/sys
mkdir /mnt/initrd/dev
mkdir /mnt/initrd/proc

# Grab busybox and create the symbolic links
pushd /mnt/initrd/bin
cp /usr/local/src/busybox-1.1.1/busybox .
ln -s busybox ash
ln -s busybox mount
ln -s busybox echo
ln -s busybox ls
ln -s busybox cat
ln -s busybox ps
ln -s busybox dmesg
ln -s busybox sysctl
popd

# Grab the necessary dev files
cp -a /dev/console /mnt/initrd/dev
cp -a /dev/ramdisk /mnt/initrd/dev
cp -a /dev/ram0 /mnt/initrd/dev
cp -a /dev/null /mnt/initrd/dev
cp -a /dev/tty1 /mnt/initrd/dev
cp -a /dev/tty2 /mnt/initrd/dev

# Equate sbin with bin
pushd /mnt/initrd
ln -s bin sbin
popd

# Create the init file
cat >> /mnt/initrd/linuxrc <<>
#!/bin/ash
echo
echo "Simple initrd is active"
echo
mount -t proc /proc /proc
mount -t sysfs none /sys
/bin/ash --login
EOF

chmod +x /mnt/initrd/linuxrc

# Finish up...
umount /mnt/initrd
gzip -9 /tmp/ramdisk.img
cp /tmp/ramdisk.img.gz /boot/ramdisk.img.gz

To create an initrd, begin by creating an empty file, using /dev/zero (a stream of zeroes) as input writing to the ramdisk.img file. The resulting file is 4MB in size (4000 1K blocks). Then use the mke2fs command to create an ext2 (second extended) file system using the empty file. Now that this file is an ext2 file system, mount the file to /mnt/initrd using the loop device. At the mount point, you now have a directory that represents an ext2 file system that you can populate for your initrd. Much of the rest of the script provides this functionality.

The next step is creating the necessary subdirectories that make up your root file system: /bin, /sys, /dev, and /proc. Only a handful are needed (for example, no libraries are present), but they contain quite a bit of functionality.

To make your root file system useful, use BusyBox. This utility is a single image that contains many individual utilities commonly found in Linux systems (such as ash, awk, sed, insmod, and so on). The advantage of BusyBox is that it packs many utilities into one while sharing their common elements, resulting in a much smaller image. This is ideal for embedded systems. Copy the BusyBox image from its source directory into your root in the /bin directory. A number of symbolic links are then created that all point to the BusyBox utility. BusyBox figures out which utility was invoked and performs that functionality. A small set of links are created in this directory to support your init script (with each command link pointing to BusyBox).

The next step is the creation of a small number of special device files. I copy these directly from my current /dev subdirectory, using the -a option (archive) to preserve their attributes.

The penultimate step is to generate the linuxrc file. After the kernel mounts the RAM disk, it searches for an init file to execute. If an init file is not found, the kernel invokes the linuxrc file as its startup script. You do the basic setup of the environment in this file, such as mounting the /proc file system. In addition to /proc, I also mount the /sys file system and emit a message to the console. Finally, I invoke ash (a Bourne Shell clone) so I can interact with the root file system. The linuxrc file is then made executable using chmod.

Finally, your root file system is complete. It's unmounted and then compressed using gzip. The resulting file (ramdisk.img.gz) is copied to the /boot subdirectory so it can be loaded via GNU GRUB.

To build the initial RAM disk, you simply invoke mkird, and the image is automatically created and copied to /boot.

Testing the custom initial RAM disk
Your new initrd image is in /boot, so the next step is to test it with your default kernel. You can now restart your Linux system. When GRUB appears, press the C key to enable the command-line utility within GRUB. You can now interact with GRUB to define the specific kernel and initrd image to load. The kernel command allows you to define the kernel file, and the initrd command allows you to specify the particular initrd image file. When these are defined, use the boot command to boot the kernel, as shown in Listing 5.

Listing 5. Manually booting the kernel and initrd using GRUB

GNU GRUB version 0.95 (638K lower / 97216K upper memory)

[ Minimal BASH-like line editing is supported. For the first word, TAB
lists possible command completions. Anywhere else TAB lists the possible
completions of a device/filename. ESC at any time exits.]

grub> kernel /bzImage-2.6.1
[Linux-bzImage, setup=0x1400, size=0x29672e]

grub> initrd /ramdisk.img.gz
[Linux-initrd @ 0x5f2a000, 0xb5108 bytes]

grub> boot

Uncompressing Linux... OK, booting the kernel.

After the kernel starts, it checks to see if an initrd image is available (more on this later), and then loads and mounts it as the root file system. You can see the end of this particular Linux startup in Listing 6. When started, the ash shell is available to enter commands. In this example, I explore the root file system and interrogate a virtual proc file system entry. I also demonstrate that you can write to the file system by touching a file (thus creating it). Note here that the first process created is linuxrc (commonly init).

Listing 6. Booting a Linux kernel with your simple initrd
...
md: Autodetecting RAID arrays
md: autorun
md: ... autorun DONE.
RAMDISK: Compressed image found at block 0
VFS: Mounted root (ext2 file system).
Freeing unused kernel memory: 208k freed
/ $ ls
bin etc linuxrc proc sys
dev lib lost+found sbin
/ $ cat /proc/1/cmdline
/bin/ash/linuxrc
/ $ cd bin
/bin $ ls
ash cat echo mount sysctl
busybox dmesg ls ps
/bin $ touch zfile
/bin $ ls
ash cat echo mount sysctl
busybox dmesg ls ps zfile

Booting with an initial RAM disk

Now that you've seen how to build and use a custom initial RAM disk, this section explores how the kernel identifies and mounts the initrd as its root file system. I walk through some of the major functions in the boot chain and explain what's happening.

The boot loader, such as GRUB, identifies the kernel that is to be loaded and copies this kernel image and any associated initrd into memory. You can find much of this functionality in the ./init subdirectory under your Linux kernel source directory.

After the kernel and initrd images are decompressed and copied into memory, the kernel is invoked. Various initialization is performed and, eventually, you find yourself in init/main.c:init() (subdir/file:function). This function performs a large amount of subsystem initialization. A call is made here to init/do_mounts.c:prepare_namespace(), which is used to prepare the namespace (mount the dev file system, RAID, or md, devices, and, finally, the initrd). Loading the initrd is done through a call to init/do_mounts_initrd.c:initrd_load().

The initrd_load() function calls init/do_mounts_rd.c:rd_load_image(), which determines the RAM disk image to load through a call to init/do_mounts_rd.c:identify_ramdisk_image(). This function checks the magic number of the image to determine if it's a minux, etc2, romfs, cramfs, or gzip format. Upon return to initrd_load_image, a call is made to init/do_mounts_rd:crd_load(). This function allocates space for the RAM disk, calculates the cyclic redundancy check (CRC), and then uncompresses and loads the RAM disk image into memory. At this point, you have the initrd image in a block device suitable for mounting.

Mounting the block device now as root begins with a call to init/do_mounts.c:mount_root(). The root device is created, and then a call is made to init/do_mounts.c:mount_block_root(). From here, init/do_mounts.c:do_mount_root() is called, which calls fs/namespace.c:sys_mount() to actually mount the root file system and then chdir to it. This is where you see the familiar message shown in Listing 6: VFS: Mounted root (ext2 file system).

Finally, you return to the init function and call init/main.c:run_init_process. This results in a call to execve to start the init process (in this case /linuxrc). The linuxrc can be an executable or a script (as long as a script interpreter is available for it).

The hierarchy of functions called is shown in Listing 7. Not all functions that are involved in copying and mounting the initial RAM disk are shown here, but this gives you a rough overview of the overall flow.

Listing 7. Hierarchy of major functions in initrd loading and mounting

init/main.c:init
init/do_mounts.c:prepare_namespace
init/do_mounts_initrd.c:initrd_load
init/do_mounts_rd.c:rd_load_image
init/do_mounts_rd.c:identify_ramdisk_image
init/do_mounts_rd.c:crd_load
lib/inflate.c:gunzip
init/do_mounts.c:mount_root
init/do_mounts.c:mount_block_root
init/do_mounts.c:do_mount_root
fs/namespace.c:sys_mount
init/main.c:run_init_process
execve

Diskless Boot

Much like embedded booting scenarios, a local disk (floppy or CD-ROM) isn't necessary to boot a kernel and ramdisk root filesystem. The Dynamic Host Configuration Protocol (or DHCP) can be used to identify network parameters such as IP address and subnet mask. The Trivial File Transfer Protocol (or TFTP) can then be used to transfer the kernel image and the initial ramdisk image to the local device. Once transferred, the Linux kernel can be booted and initrd mounted, as is done in a local image boot.

Shrinking your initrd

When you're building an embedded system and want the smallest initrd image possible, there are a few tips to consider. The first is to use BusyBox (demonstrated in this article). BusyBox takes several megabytes of utilities and shrinks them down to several hundred kilobytes.

In this example, the BusyBox image is statically linked so that no libraries are required. However, if you need the standard C library (for your custom binaries), there are other options beyond the massive glibc. The first small library is uClibc, which is a minimized version of the standard C library for space-constrained systems. Another library that's ideal for space-constrained environments is dietlib. Keep in mind that you'll need to recompile the binaries that you want in your embedded system using these libraries, so some additional work is required (but worth it).

No comments:

Post a Comment