INITRD(4) Linux Programmer's Manual INITRD(4)
initrd - boot loader initialized RAM disk
The special file /dev/initrd is a read-only block device. Device
/dev/initrd is a RAM disk that is initialized (e.g. loaded) by the boot
loader before the kernel is started. The kernel then can use the block
device /dev/initrd's contents for a two phased system boot-up.
In the first boot-up phase, the kernel starts up and mounts an initial
root file-system from the contents of /dev/initrd (e.g. RAM disk ini-
tialized by the boot loader). In the second phase, additional drivers
or other modules are loaded from the initial root device's contents.
After loading the additional modules, a new root file system (i.e. the
normal root file system) is mounted from a different device.
When booting up with initrd, the system boots as follows:
1. The boot loader loads the kernel program and /dev/initrd's con-
tents into memory.
2. On kernel startup, the kernel uncompresses and copies the contents
of the device /dev/initrd onto device /dev/ram0 and then frees the
memory used by /dev/initrd.
3. The kernel then read-write mounts device /dev/ram0 as the initial
root file system.
4. If the indicated normal root file system is also the initial root
file-system (e.g. /dev/ram0 ) then the kernel skips to the last step
for the usual boot sequence.
5. If the executable file /linuxrc is present in the initial root
file-system, /linuxrc is executed with UID 0. (The file /linuxrc
must have executable permission. The file /linuxrc can be any valid
executable, including a shell script.)
6. If /linuxrc is not executed or when /linuxrc terminates, the nor-
mal root file system is mounted. (If /linuxrc exits with any file-
systems mounted on the initial root file-system, then the behavior of
the kernel is UNSPECIFIED. See the NOTES section for the current
7. If the normal root file has directory /initrd, device /dev/ram0 is
moved from / to /initrd. Otherwise if directory /initrd does not
exist device /dev/ram0 is unmounted. (When moved from / to /initrd,
/dev/ram0 is not unmounted and therefore processes can remain running
from /dev/ram0. If directory /initrd does not exist on the normal
root file-system and any processes remain running from /dev/ram0 when
/linuxrc exits, the behavior of the kernel is UNSPECIFIED. See the
NOTES section for the current kernel behavior.)
8. The usual boot sequence (e.g. invocation of /sbin/init) is per-
formed on the normal root file system.
The following boot loader options when used with initrd, affect the
kernel's boot-up operation:
Specifies the file to load as the contents of /dev/initrd. For
LOADLIN this is a command line option. For LILO you have to use
this command in the LILO configuration file /etc/lilo.config.
The filename specified with this option will typically be a
gzipped file-system image.
This boot time option disables the two phase boot-up operation.
The kernel performs the usual boot sequence as if /dev/initrd
was not initialized. With this option, any contents of
/dev/initrd loaded into memory by the boot loader contents are
preserved. This option permits the contents of /dev/initrd to
be any data and need not be limited to a file system image.
However, device /dev/initrd is read-only and can be read only
one time after system startup.
Specifies the device to be used as the normal root file system.
For LOADLIN this is a command line option. For LILO this is a
boot time option or can be used as an option line in the LILO
configuration file /etc/lilo.config. The device specified by
the this option must be a mountable device having a suitable
CHANGING THE NORMAL ROOT FILE SYSTEM
By default, the kernel's settings (e.g. set in the kernel file with
rdev(8) or compiled into the kernel file), or the boot loader option
setting is used for the normal root file systems. For a NFS-mounted
normal root file system, one has to use the nfs_root_name and
nfs_root_addrs boot options to give the NFS settings. For more infor-
mation on NFS-mounted root see the kernel documentation file nfs-
root.txt. For more information on setting the root file system also
see the LILO and LOADLIN documentation.
It is also possible for the /linuxrc executable to change the normal
root device. For /linuxrc to change the normal root device, /proc must
be mounted. After mounting /proc, /linuxrc changes the normal root
device by writing into the proc files /proc/sys/kernel/real-root-dev,
/proc/sys/kernel/nfs-root-name, and /proc/sys/kernel/nfs-root-addrs.
For a physical root device, the root device is changed by having /lin-
uxrc write the new root file system device number into /proc/sys/ker-
nel/real-root-dev. For a NFS root file system, the root device is
changed by having /linuxrc write the NFS setting into files
/proc/sys/kernel/nfs-root-name and /proc/sys/kernel/nfs-root-addrs and
then writing 0xff (e.g. the pseudo-NFS-device number) into file
/proc/sys/kernel/real-root-dev. For example, the following shell com-
mand line would change the normal root device to /dev/hdb1:
echo 0x365 >/proc/sys/kernel/real-root-dev
For a NFS example, the following shell command lines would change the
normal root device to the NFS directory /var/nfsroot on a local net-
worked NFS server with IP number 126.96.36.199 for a system with IP num-
ber 188.8.131.52 and named 'idefix':
echo /var/nfsroot >/proc/sys/kernel/nfs-root-name
echo 184.108.40.206:220.127.116.11::255.255.255.0:idefix \
echo 255 >/proc/sys/kernel/real-root-dev
Note: The use of /proc/sys/kernel/real-root-dev to change the root file
system is obsolete. See the kernel source file Documentation/ini-
trd.txt as well as pivot_root(2) and pivot_root(8) for information on
the modern method of changing the root file system.
The main motivation for implementing initrd was to allow for modular
kernel configuration at system installation.
A possible system installation scenario is as follows:
1. The loader program boots from floppy or other media with a minimal
kernel (e.g. support for /dev/ram, /dev/initrd, and the ext2 file-
system) and loads /dev/initrd with a gzipped version of the initial
2. The executable /linuxrc determines what is needed to (1) mount the
normal root file-system (i.e. device type, device drivers, file sys-
tem) and (2) the distribution media (e.g. CD-ROM, network, tape,
...). This can be done by asking the user, by auto-probing, or by
using a hybrid approach.
3. The executable /linuxrc loads the necessary modules from the ini-
tial root file-system.
4. The executable /linuxrc creates and populates the root file sys-
tem. (At this stage the normal root file system does not have to be
a completed system yet.)
5. The executable /linuxrc sets /proc/sys/kernel/real-root-dev,
unmount /proc, the normal root file system and any other file systems
it has mounted, and then terminates.
6. The kernel then mounts the normal root file system.
7. Now that the file system is accessible and intact, the boot loader
can be installed.
8. The boot loader is configured to load into /dev/initrd a file sys-
tem with the set of modules that was used to bring up the system.
(e.g. Device /dev/ram0 can be modified, then unmounted, and finally,
the image is written from /dev/ram0 to a file.)
9. The system is now bootable and additional installation tasks can
The key role of /dev/initrd in the above is to re-use the configuration
data during normal system operation without requiring initial kernel
selection, a large generic kernel or, recompiling the kernel.
A second scenario is for installations where Linux runs on systems with
different hardware configurations in a single administrative network.
In such cases, it may be desirable to use only a small set of kernels
(ideally only one) and to keep the system-specific part of configura-
tion information as small as possible. In this case, create a common
file with all needed modules. Then, only the /linuxrc file or a file
executed by /linuxrc would be different.
A third scenario is more convenient recovery disks. Because informa-
tion like the location of the root file-system partition is not needed
at boot time, the system loaded from /dev/initrd can use a dialog
and/or auto-detection followed by a possible sanity check.
Last but not least, Linux distributions on CD-ROM may use initrd for
easy installation from the CD-ROM. The distribution can use LOADLIN to
directly load /dev/initrd from CD-ROM without the need of any floppies.
The distribution could also use a LILO boot floppy and then bootstrap a
bigger ram disk via /dev/initrd from the CD-ROM.
The /dev/initrd is a read-only block device assigned major number 1 and
minor number 250. Typically /dev/initrd is owned by root.disk with
mode 0400 (read access by root only). If the Linux system does not
have /dev/initrd already created, it can be created with the following
mknod -m 400 /dev/initrd b 1 250
chown root:disk /dev/initrd
Also, support for both "RAM disk" and "Initial RAM disk" (e.g.
CONFIG_BLK_DEV_RAM=y and CONFIG_BLK_DEV_INITRD=y ) support must be com-
piled directly into the Linux kernel to use /dev/initrd. When using
/dev/initrd, the RAM disk driver cannot be loaded as a module.
chown(1), mknod(1), ram(4), freeramdisk(8), rdev(8)
The documentation file initrd.txt in the kernel source package, the
LILO documentation, the LOADLIN documentation, the SYSLINUX documenta-
1. With the current kernel, any file systems that remain mounted when
/dev/ram0 is moved from / to /initrd continue to be accessible. How-
ever, the /proc/mounts entries are not updated.
2. With the current kernel, if directory /initrd does not exist, then
/dev/ram0 will NOT be fully unmounted if /dev/ram0 is used by any pro-
cess or has any file-system mounted on it. If /dev/ram0 is NOT fully
unmounted, then /dev/ram0 will remain in memory.
3. Users of /dev/initrd should not depend on the behavior give in the
above notes. The behavior may change in future versions of the Linux
The kernel code for device initrd was written by Werner Almesberger
<firstname.lastname@example.org> and Hans Lermen <email@example.com>.
The code for initrd was added to the baseline Linux kernel in develop-
ment version 1.3.73.
Linux 2.0 1997-11-06 INITRD(4)