0001 =======================
0002 Early userspace support
0003 =======================
0004
0005 Last update: 2004-12-20 tlh
0006
0007
0008 "Early userspace" is a set of libraries and programs that provide
0009 various pieces of functionality that are important enough to be
0010 available while a Linux kernel is coming up, but that don't need to be
0011 run inside the kernel itself.
0012
0013 It consists of several major infrastructure components:
0014
0015 - gen_init_cpio, a program that builds a cpio-format archive
0016 containing a root filesystem image. This archive is compressed, and
0017 the compressed image is linked into the kernel image.
0018 - initramfs, a chunk of code that unpacks the compressed cpio image
0019 midway through the kernel boot process.
0020 - klibc, a userspace C library, currently packaged separately, that is
0021 optimized for correctness and small size.
0022
0023 The cpio file format used by initramfs is the "newc" (aka "cpio -H newc")
0024 format, and is documented in the file "buffer-format.txt". There are
0025 two ways to add an early userspace image: specify an existing cpio
0026 archive to be used as the image or have the kernel build process build
0027 the image from specifications.
0028
0029 CPIO ARCHIVE method
0030 -------------------
0031
0032 You can create a cpio archive that contains the early userspace image.
0033 Your cpio archive should be specified in CONFIG_INITRAMFS_SOURCE and it
0034 will be used directly. Only a single cpio file may be specified in
0035 CONFIG_INITRAMFS_SOURCE and directory and file names are not allowed in
0036 combination with a cpio archive.
0037
0038 IMAGE BUILDING method
0039 ---------------------
0040
0041 The kernel build process can also build an early userspace image from
0042 source parts rather than supplying a cpio archive. This method provides
0043 a way to create images with root-owned files even though the image was
0044 built by an unprivileged user.
0045
0046 The image is specified as one or more sources in
0047 CONFIG_INITRAMFS_SOURCE. Sources can be either directories or files -
0048 cpio archives are *not* allowed when building from sources.
0049
0050 A source directory will have it and all of its contents packaged. The
0051 specified directory name will be mapped to '/'. When packaging a
0052 directory, limited user and group ID translation can be performed.
0053 INITRAMFS_ROOT_UID can be set to a user ID that needs to be mapped to
0054 user root (0). INITRAMFS_ROOT_GID can be set to a group ID that needs
0055 to be mapped to group root (0).
0056
0057 A source file must be directives in the format required by the
0058 usr/gen_init_cpio utility (run 'usr/gen_init_cpio -h' to get the
0059 file format). The directives in the file will be passed directly to
0060 usr/gen_init_cpio.
0061
0062 When a combination of directories and files are specified then the
0063 initramfs image will be an aggregate of all of them. In this way a user
0064 can create a 'root-image' directory and install all files into it.
0065 Because device-special files cannot be created by a unprivileged user,
0066 special files can be listed in a 'root-files' file. Both 'root-image'
0067 and 'root-files' can be listed in CONFIG_INITRAMFS_SOURCE and a complete
0068 early userspace image can be built by an unprivileged user.
0069
0070 As a technical note, when directories and files are specified, the
0071 entire CONFIG_INITRAMFS_SOURCE is passed to
0072 usr/gen_initramfs.sh. This means that CONFIG_INITRAMFS_SOURCE
0073 can really be interpreted as any legal argument to
0074 gen_initramfs.sh. If a directory is specified as an argument then
0075 the contents are scanned, uid/gid translation is performed, and
0076 usr/gen_init_cpio file directives are output. If a directory is
0077 specified as an argument to usr/gen_initramfs.sh then the
0078 contents of the file are simply copied to the output. All of the output
0079 directives from directory scanning and file contents copying are
0080 processed by usr/gen_init_cpio.
0081
0082 See also 'usr/gen_initramfs.sh -h'.
0083
0084 Where's this all leading?
0085 =========================
0086
0087 The klibc distribution contains some of the necessary software to make
0088 early userspace useful. The klibc distribution is currently
0089 maintained separately from the kernel.
0090
0091 You can obtain somewhat infrequent snapshots of klibc from
0092 https://www.kernel.org/pub/linux/libs/klibc/
0093
0094 For active users, you are better off using the klibc git
0095 repository, at https://git.kernel.org/?p=libs/klibc/klibc.git
0096
0097 The standalone klibc distribution currently provides three components,
0098 in addition to the klibc library:
0099
0100 - ipconfig, a program that configures network interfaces. It can
0101 configure them statically, or use DHCP to obtain information
0102 dynamically (aka "IP autoconfiguration").
0103 - nfsmount, a program that can mount an NFS filesystem.
0104 - kinit, the "glue" that uses ipconfig and nfsmount to replace the old
0105 support for IP autoconfig, mount a filesystem over NFS, and continue
0106 system boot using that filesystem as root.
0107
0108 kinit is built as a single statically linked binary to save space.
0109
0110 Eventually, several more chunks of kernel functionality will hopefully
0111 move to early userspace:
0112
0113 - Almost all of init/do_mounts* (the beginning of this is already in
0114 place)
0115 - ACPI table parsing
0116 - Insert unwieldy subsystem that doesn't really need to be in kernel
0117 space here
0118
0119 If kinit doesn't meet your current needs and you've got bytes to burn,
0120 the klibc distribution includes a small Bourne-compatible shell (ash)
0121 and a number of other utilities, so you can replace kinit and build
0122 custom initramfs images that meet your needs exactly.
0123
0124 For questions and help, you can sign up for the early userspace
0125 mailing list at https://www.zytor.com/mailman/listinfo/klibc
0126
0127 How does it work?
0128 =================
0129
0130 The kernel has currently 3 ways to mount the root filesystem:
0131
0132 a) all required device and filesystem drivers compiled into the kernel, no
0133 initrd. init/main.c:init() will call prepare_namespace() to mount the
0134 final root filesystem, based on the root= option and optional init= to run
0135 some other init binary than listed at the end of init/main.c:init().
0136
0137 b) some device and filesystem drivers built as modules and stored in an
0138 initrd. The initrd must contain a binary '/linuxrc' which is supposed to
0139 load these driver modules. It is also possible to mount the final root
0140 filesystem via linuxrc and use the pivot_root syscall. The initrd is
0141 mounted and executed via prepare_namespace().
0142
0143 c) using initramfs. The call to prepare_namespace() must be skipped.
0144 This means that a binary must do all the work. Said binary can be stored
0145 into initramfs either via modifying usr/gen_init_cpio.c or via the new
0146 initrd format, an cpio archive. It must be called "/init". This binary
0147 is responsible to do all the things prepare_namespace() would do.
0148
0149 To maintain backwards compatibility, the /init binary will only run if it
0150 comes via an initramfs cpio archive. If this is not the case,
0151 init/main.c:init() will run prepare_namespace() to mount the final root
0152 and exec one of the predefined init binaries.
0153
0154 Bryan O'Sullivan <bos@serpentine.com>
