1 Early userspace support
2 =======================
4 Last update: 2004-12-20 tlh
7 "Early userspace" is a set of libraries and programs that provide
8 various pieces of functionality that are important enough to be
9 available while a Linux kernel is coming up, but that don't need to be
10 run inside the kernel itself.
12 It consists of several major infrastructure components:
14 - gen_init_cpio, a program that builds a cpio-format archive
15 containing a root filesystem image. This archive is compressed, and
16 the compressed image is linked into the kernel image.
17 - initramfs, a chunk of code that unpacks the compressed cpio image
18 midway through the kernel boot process.
19 - klibc, a userspace C library, currently packaged separately, that is
20 optimized for correctness and small size.
22 The cpio file format used by initramfs is the "newc" (aka "cpio -H newc")
23 format, and is documented in the file "buffer-format.txt". There are
24 two ways to add an early userspace image: specify an existing cpio
25 archive to be used as the image or have the kernel build process build
26 the image from specifications.
30 You can create a cpio archive that contains the early userspace image.
31 Your cpio archive should be specified in CONFIG_INITRAMFS_SOURCE and it
32 will be used directly. Only a single cpio file may be specified in
33 CONFIG_INITRAMFS_SOURCE and directory and file names are not allowed in
34 combination with a cpio archive.
38 The kernel build process can also build an early userspace image from
39 source parts rather than supplying a cpio archive. This method provides
40 a way to create images with root-owned files even though the image was
41 built by an unprivileged user.
43 The image is specified as one or more sources in
44 CONFIG_INITRAMFS_SOURCE. Sources can be either directories or files -
45 cpio archives are *not* allowed when building from sources.
47 A source directory will have it and all of its contents packaged. The
48 specified directory name will be mapped to '/'. When packaging a
49 directory, limited user and group ID translation can be performed.
50 INITRAMFS_ROOT_UID can be set to a user ID that needs to be mapped to
51 user root (0). INITRAMFS_ROOT_GID can be set to a group ID that needs
52 to be mapped to group root (0).
54 A source file must be directives in the format required by the
55 usr/gen_init_cpio utility (run 'usr/gen_init_cpio --help' to get the
56 file format). The directives in the file will be passed directly to
59 When a combination of directories and files are specified then the
60 initramfs image will be an aggregate of all of them. In this way a user
61 can create a 'root-image' directory and install all files into it.
62 Because device-special files cannot be created by a unprivileged user,
63 special files can be listed in a 'root-files' file. Both 'root-image'
64 and 'root-files' can be listed in CONFIG_INITRAMFS_SOURCE and a complete
65 early userspace image can be built by an unprivileged user.
67 As a technical note, when directories and files are specified, the
68 entire CONFIG_INITRAMFS_SOURCE is passed to
69 scripts/gen_initramfs_list.sh. This means that CONFIG_INITRAMFS_SOURCE
70 can really be interpreted as any legal argument to
71 gen_initramfs_list.sh. If a directory is specified as an argument then
72 the contents are scanned, uid/gid translation is performed, and
73 usr/gen_init_cpio file directives are output. If a directory is
74 specified as an argument to scripts/gen_initramfs_list.sh then the
75 contents of the file are simply copied to the output. All of the output
76 directives from directory scanning and file contents copying are
77 processed by usr/gen_init_cpio.
79 See also 'scripts/gen_initramfs_list.sh -h'.
81 Where's this all leading?
82 =========================
84 The klibc distribution contains some of the necessary software to make
85 early userspace useful. The klibc distribution is currently
86 maintained separately from the kernel.
88 You can obtain somewhat infrequent snapshots of klibc from
89 ftp://ftp.kernel.org/pub/linux/libs/klibc/
91 For active users, you are better off using the klibc git
92 repository, at http://git.kernel.org/?p=libs/klibc/klibc.git
94 The standalone klibc distribution currently provides three components,
95 in addition to the klibc library:
97 - ipconfig, a program that configures network interfaces. It can
98 configure them statically, or use DHCP to obtain information
99 dynamically (aka "IP autoconfiguration").
100 - nfsmount, a program that can mount an NFS filesystem.
101 - kinit, the "glue" that uses ipconfig and nfsmount to replace the old
102 support for IP autoconfig, mount a filesystem over NFS, and continue
103 system boot using that filesystem as root.
105 kinit is built as a single statically linked binary to save space.
107 Eventually, several more chunks of kernel functionality will hopefully
108 move to early userspace:
110 - Almost all of init/do_mounts* (the beginning of this is already in
113 - Insert unwieldy subsystem that doesn't really need to be in kernel
116 If kinit doesn't meet your current needs and you've got bytes to burn,
117 the klibc distribution includes a small Bourne-compatible shell (ash)
118 and a number of other utilities, so you can replace kinit and build
119 custom initramfs images that meet your needs exactly.
121 For questions and help, you can sign up for the early userspace
122 mailing list at http://www.zytor.com/mailman/listinfo/klibc
127 The kernel has currently 3 ways to mount the root filesystem:
129 a) all required device and filesystem drivers compiled into the kernel, no
130 initrd. init/main.c:init() will call prepare_namespace() to mount the
131 final root filesystem, based on the root= option and optional init= to run
132 some other init binary than listed at the end of init/main.c:init().
134 b) some device and filesystem drivers built as modules and stored in an
135 initrd. The initrd must contain a binary '/linuxrc' which is supposed to
136 load these driver modules. It is also possible to mount the final root
137 filesystem via linuxrc and use the pivot_root syscall. The initrd is
138 mounted and executed via prepare_namespace().
140 c) using initramfs. The call to prepare_namespace() must be skipped.
141 This means that a binary must do all the work. Said binary can be stored
142 into initramfs either via modifying usr/gen_init_cpio.c or via the new
143 initrd format, an cpio archive. It must be called "/init". This binary
144 is responsible to do all the things prepare_namespace() would do.
146 To maintain backwards compatibility, the /init binary will only run if it
147 comes via an initramfs cpio archive. If this is not the case,
148 init/main.c:init() will run prepare_namespace() to mount the final root
149 and exec one of the predefined init binaries.
151 Bryan O'Sullivan <bos@serpentine.com>