Back to home page




0001 Early userspace support
0002 =======================
0004 Last update: 2004-12-20 tlh
0007 "Early userspace" is a set of libraries and programs that provide
0008 various pieces of functionality that are important enough to be
0009 available while a Linux kernel is coming up, but that don't need to be
0010 run inside the kernel itself.
0012 It consists of several major infrastructure components:
0014 - gen_init_cpio, a program that builds a cpio-format archive
0015   containing a root filesystem image.  This archive is compressed, and
0016   the compressed image is linked into the kernel image.
0017 - initramfs, a chunk of code that unpacks the compressed cpio image
0018   midway through the kernel boot process.
0019 - klibc, a userspace C library, currently packaged separately, that is
0020   optimized for correctness and small size.
0022 The cpio file format used by initramfs is the "newc" (aka "cpio -H newc")
0023 format, and is documented in the file "buffer-format.txt".  There are
0024 two ways to add an early userspace image: specify an existing cpio
0025 archive to be used as the image or have the kernel build process build
0026 the image from specifications.
0028 CPIO ARCHIVE method
0030 You can create a cpio archive that contains the early userspace image.
0031 Your cpio archive should be specified in CONFIG_INITRAMFS_SOURCE and it
0032 will be used directly.  Only a single cpio file may be specified in
0033 CONFIG_INITRAMFS_SOURCE and directory and file names are not allowed in
0034 combination with a cpio archive.
0036 IMAGE BUILDING method
0038 The kernel build process can also build an early userspace image from
0039 source parts rather than supplying a cpio archive.  This method provides
0040 a way to create images with root-owned files even though the image was
0041 built by an unprivileged user.
0043 The image is specified as one or more sources in
0044 CONFIG_INITRAMFS_SOURCE.  Sources can be either directories or files -
0045 cpio archives are *not* allowed when building from sources.
0047 A source directory will have it and all of its contents packaged.  The
0048 specified directory name will be mapped to '/'.  When packaging a
0049 directory, limited user and group ID translation can be performed.
0050 INITRAMFS_ROOT_UID can be set to a user ID that needs to be mapped to
0051 user root (0).  INITRAMFS_ROOT_GID can be set to a group ID that needs
0052 to be mapped to group root (0).
0054 A source file must be directives in the format required by the
0055 usr/gen_init_cpio utility (run 'usr/gen_init_cpio --help' to get the
0056 file format).  The directives in the file will be passed directly to
0057 usr/gen_init_cpio.
0059 When a combination of directories and files are specified then the
0060 initramfs image will be an aggregate of all of them.  In this way a user
0061 can create a 'root-image' directory and install all files into it.
0062 Because device-special files cannot be created by a unprivileged user,
0063 special files can be listed in a 'root-files' file.  Both 'root-image'
0064 and 'root-files' can be listed in CONFIG_INITRAMFS_SOURCE and a complete
0065 early userspace image can be built by an unprivileged user.
0067 As a technical note, when directories and files are specified, the
0068 entire CONFIG_INITRAMFS_SOURCE is passed to
0069 scripts/  This means that CONFIG_INITRAMFS_SOURCE
0070 can really be interpreted as any legal argument to
0071  If a directory is specified as an argument then
0072 the contents are scanned, uid/gid translation is performed, and
0073 usr/gen_init_cpio file directives are output.  If a directory is
0074 specified as an argument to scripts/ then the
0075 contents of the file are simply copied to the output.  All of the output
0076 directives from directory scanning and file contents copying are
0077 processed by usr/gen_init_cpio.
0079 See also 'scripts/ -h'.
0081 Where's this all leading?
0082 =========================
0084 The klibc distribution contains some of the necessary software to make
0085 early userspace useful.  The klibc distribution is currently
0086 maintained separately from the kernel.
0088 You can obtain somewhat infrequent snapshots of klibc from
0091 For active users, you are better off using the klibc git
0092 repository, at
0094 The standalone klibc distribution currently provides three components,
0095 in addition to the klibc library:
0097 - ipconfig, a program that configures network interfaces.  It can
0098   configure them statically, or use DHCP to obtain information
0099   dynamically (aka "IP autoconfiguration").
0100 - nfsmount, a program that can mount an NFS filesystem.
0101 - kinit, the "glue" that uses ipconfig and nfsmount to replace the old
0102   support for IP autoconfig, mount a filesystem over NFS, and continue
0103   system boot using that filesystem as root.
0105 kinit is built as a single statically linked binary to save space.
0107 Eventually, several more chunks of kernel functionality will hopefully
0108 move to early userspace:
0110 - Almost all of init/do_mounts* (the beginning of this is already in
0111   place)
0112 - ACPI table parsing
0113 - Insert unwieldy subsystem that doesn't really need to be in kernel
0114   space here
0116 If kinit doesn't meet your current needs and you've got bytes to burn,
0117 the klibc distribution includes a small Bourne-compatible shell (ash)
0118 and a number of other utilities, so you can replace kinit and build
0119 custom initramfs images that meet your needs exactly.
0121 For questions and help, you can sign up for the early userspace
0122 mailing list at
0124 How does it work?
0125 =================
0127 The kernel has currently 3 ways to mount the root filesystem:
0129 a) all required device and filesystem drivers compiled into the kernel, no
0130    initrd.  init/main.c:init() will call prepare_namespace() to mount the
0131    final root filesystem, based on the root= option and optional init= to run
0132    some other init binary than listed at the end of init/main.c:init().
0134 b) some device and filesystem drivers built as modules and stored in an
0135    initrd.  The initrd must contain a binary '/linuxrc' which is supposed to
0136    load these driver modules.  It is also possible to mount the final root
0137    filesystem via linuxrc and use the pivot_root syscall.  The initrd is
0138    mounted and executed via prepare_namespace().
0140 c) using initramfs.  The call to prepare_namespace() must be skipped.
0141    This means that a binary must do all the work.  Said binary can be stored
0142    into initramfs either via modifying usr/gen_init_cpio.c or via the new
0143    initrd format, an cpio archive.  It must be called "/init".  This binary
0144    is responsible to do all the things prepare_namespace() would do.
0146    To maintain backwards compatibility, the /init binary will only run if it
0147    comes via an initramfs cpio archive.  If this is not the case,
0148    init/main.c:init() will run prepare_namespace() to mount the final root
0149    and exec one of the predefined init binaries.
0151 Bryan O'Sullivan <>