initrd.rst 14 KB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195196197198199200201202203204205206207208209210211212213214215216217218219220221222223224225226227228229230231232233234235236237238239240241242243244245246247248249250251252253254255256257258259260261262263264265266267268269270271272273274275276277278279280281282283284285286287288289290291292293294295296297298299300301302303304305306307308309310311312313314315316317318319320321322323324325326327328329330331332333334335336337338339340341342343344345346347348349350351352353354355356357358359360361362363364365366367368369370371372373374375376377378379380381382383
  1. Using the initial RAM disk (initrd)
  2. ===================================
  3. Written 1996,2000 by Werner Almesberger <werner.almesberger@epfl.ch> and
  4. Hans Lermen <lermen@fgan.de>
  5. initrd provides the capability to load a RAM disk by the boot loader.
  6. This RAM disk can then be mounted as the root file system and programs
  7. can be run from it. Afterwards, a new root file system can be mounted
  8. from a different device. The previous root (from initrd) is then moved
  9. to a directory and can be subsequently unmounted.
  10. initrd is mainly designed to allow system startup to occur in two phases,
  11. where the kernel comes up with a minimum set of compiled-in drivers, and
  12. where additional modules are loaded from initrd.
  13. This document gives a brief overview of the use of initrd. A more detailed
  14. discussion of the boot process can be found in [#f1]_.
  15. Operation
  16. ---------
  17. When using initrd, the system typically boots as follows:
  18. 1) the boot loader loads the kernel and the initial RAM disk
  19. 2) the kernel converts initrd into a "normal" RAM disk and
  20. frees the memory used by initrd
  21. 3) if the root device is not ``/dev/ram0``, the old (deprecated)
  22. change_root procedure is followed. see the "Obsolete root change
  23. mechanism" section below.
  24. 4) root device is mounted. if it is ``/dev/ram0``, the initrd image is
  25. then mounted as root
  26. 5) /sbin/init is executed (this can be any valid executable, including
  27. shell scripts; it is run with uid 0 and can do basically everything
  28. init can do).
  29. 6) init mounts the "real" root file system
  30. 7) init places the root file system at the root directory using the
  31. pivot_root system call
  32. 8) init execs the ``/sbin/init`` on the new root filesystem, performing
  33. the usual boot sequence
  34. 9) the initrd file system is removed
  35. Note that changing the root directory does not involve unmounting it.
  36. It is therefore possible to leave processes running on initrd during that
  37. procedure. Also note that file systems mounted under initrd continue to
  38. be accessible.
  39. Boot command-line options
  40. -------------------------
  41. initrd adds the following new options::
  42. initrd=<path> (e.g. LOADLIN)
  43. Loads the specified file as the initial RAM disk. When using LILO, you
  44. have to specify the RAM disk image file in /etc/lilo.conf, using the
  45. INITRD configuration variable.
  46. noinitrd
  47. initrd data is preserved but it is not converted to a RAM disk and
  48. the "normal" root file system is mounted. initrd data can be read
  49. from /dev/initrd. Note that the data in initrd can have any structure
  50. in this case and doesn't necessarily have to be a file system image.
  51. This option is used mainly for debugging.
  52. Note: /dev/initrd is read-only and it can only be used once. As soon
  53. as the last process has closed it, all data is freed and /dev/initrd
  54. can't be opened anymore.
  55. root=/dev/ram0
  56. initrd is mounted as root, and the normal boot procedure is followed,
  57. with the RAM disk mounted as root.
  58. Compressed cpio images
  59. ----------------------
  60. Recent kernels have support for populating a ramdisk from a compressed cpio
  61. archive. On such systems, the creation of a ramdisk image doesn't need to
  62. involve special block devices or loopbacks; you merely create a directory on
  63. disk with the desired initrd content, cd to that directory, and run (as an
  64. example)::
  65. find . | cpio --quiet -H newc -o | gzip -9 -n > /boot/imagefile.img
  66. Examining the contents of an existing image file is just as simple::
  67. mkdir /tmp/imagefile
  68. cd /tmp/imagefile
  69. gzip -cd /boot/imagefile.img | cpio -imd --quiet
  70. Installation
  71. ------------
  72. First, a directory for the initrd file system has to be created on the
  73. "normal" root file system, e.g.::
  74. # mkdir /initrd
  75. The name is not relevant. More details can be found on the
  76. :manpage:`pivot_root(2)` man page.
  77. If the root file system is created during the boot procedure (i.e. if
  78. you're building an install floppy), the root file system creation
  79. procedure should create the ``/initrd`` directory.
  80. If initrd will not be mounted in some cases, its content is still
  81. accessible if the following device has been created::
  82. # mknod /dev/initrd b 1 250
  83. # chmod 400 /dev/initrd
  84. Second, the kernel has to be compiled with RAM disk support and with
  85. support for the initial RAM disk enabled. Also, at least all components
  86. needed to execute programs from initrd (e.g. executable format and file
  87. system) must be compiled into the kernel.
  88. Third, you have to create the RAM disk image. This is done by creating a
  89. file system on a block device, copying files to it as needed, and then
  90. copying the content of the block device to the initrd file. With recent
  91. kernels, at least three types of devices are suitable for that:
  92. - a floppy disk (works everywhere but it's painfully slow)
  93. - a RAM disk (fast, but allocates physical memory)
  94. - a loopback device (the most elegant solution)
  95. We'll describe the loopback device method:
  96. 1) make sure loopback block devices are configured into the kernel
  97. 2) create an empty file system of the appropriate size, e.g.::
  98. # dd if=/dev/zero of=initrd bs=300k count=1
  99. # mke2fs -F -m0 initrd
  100. (if space is critical, you may want to use the Minix FS instead of Ext2)
  101. 3) mount the file system, e.g.::
  102. # mount -t ext2 -o loop initrd /mnt
  103. 4) create the console device::
  104. # mkdir /mnt/dev
  105. # mknod /mnt/dev/console c 5 1
  106. 5) copy all the files that are needed to properly use the initrd
  107. environment. Don't forget the most important file, ``/sbin/init``
  108. .. note:: ``/sbin/init`` permissions must include "x" (execute).
  109. 6) correct operation the initrd environment can frequently be tested
  110. even without rebooting with the command::
  111. # chroot /mnt /sbin/init
  112. This is of course limited to initrds that do not interfere with the
  113. general system state (e.g. by reconfiguring network interfaces,
  114. overwriting mounted devices, trying to start already running demons,
  115. etc. Note however that it is usually possible to use pivot_root in
  116. such a chroot'ed initrd environment.)
  117. 7) unmount the file system::
  118. # umount /mnt
  119. 8) the initrd is now in the file "initrd". Optionally, it can now be
  120. compressed::
  121. # gzip -9 initrd
  122. For experimenting with initrd, you may want to take a rescue floppy and
  123. only add a symbolic link from ``/sbin/init`` to ``/bin/sh``. Alternatively, you
  124. can try the experimental newlib environment [#f2]_ to create a small
  125. initrd.
  126. Finally, you have to boot the kernel and load initrd. Almost all Linux
  127. boot loaders support initrd. Since the boot process is still compatible
  128. with an older mechanism, the following boot command line parameters
  129. have to be given::
  130. root=/dev/ram0 rw
  131. (rw is only necessary if writing to the initrd file system.)
  132. With LOADLIN, you simply execute::
  133. LOADLIN <kernel> initrd=<disk_image>
  134. e.g.::
  135. LOADLIN C:\LINUX\BZIMAGE initrd=C:\LINUX\INITRD.GZ root=/dev/ram0 rw
  136. With LILO, you add the option ``INITRD=<path>`` to either the global section
  137. or to the section of the respective kernel in ``/etc/lilo.conf``, and pass
  138. the options using APPEND, e.g.::
  139. image = /bzImage
  140. initrd = /boot/initrd.gz
  141. append = "root=/dev/ram0 rw"
  142. and run ``/sbin/lilo``
  143. For other boot loaders, please refer to the respective documentation.
  144. Now you can boot and enjoy using initrd.
  145. Changing the root device
  146. ------------------------
  147. When finished with its duties, init typically changes the root device
  148. and proceeds with starting the Linux system on the "real" root device.
  149. The procedure involves the following steps:
  150. - mounting the new root file system
  151. - turning it into the root file system
  152. - removing all accesses to the old (initrd) root file system
  153. - unmounting the initrd file system and de-allocating the RAM disk
  154. Mounting the new root file system is easy: it just needs to be mounted on
  155. a directory under the current root. Example::
  156. # mkdir /new-root
  157. # mount -o ro /dev/hda1 /new-root
  158. The root change is accomplished with the pivot_root system call, which
  159. is also available via the ``pivot_root`` utility (see :manpage:`pivot_root(8)`
  160. man page; ``pivot_root`` is distributed with util-linux version 2.10h or higher
  161. [#f3]_). ``pivot_root`` moves the current root to a directory under the new
  162. root, and puts the new root at its place. The directory for the old root
  163. must exist before calling ``pivot_root``. Example::
  164. # cd /new-root
  165. # mkdir initrd
  166. # pivot_root . initrd
  167. Now, the init process may still access the old root via its
  168. executable, shared libraries, standard input/output/error, and its
  169. current root directory. All these references are dropped by the
  170. following command::
  171. # exec chroot . what-follows <dev/console >dev/console 2>&1
  172. Where what-follows is a program under the new root, e.g. ``/sbin/init``
  173. If the new root file system will be used with udev and has no valid
  174. ``/dev`` directory, udev must be initialized before invoking chroot in order
  175. to provide ``/dev/console``.
  176. Note: implementation details of pivot_root may change with time. In order
  177. to ensure compatibility, the following points should be observed:
  178. - before calling pivot_root, the current directory of the invoking
  179. process should point to the new root directory
  180. - use . as the first argument, and the _relative_ path of the directory
  181. for the old root as the second argument
  182. - a chroot program must be available under the old and the new root
  183. - chroot to the new root afterwards
  184. - use relative paths for dev/console in the exec command
  185. Now, the initrd can be unmounted and the memory allocated by the RAM
  186. disk can be freed::
  187. # umount /initrd
  188. # blockdev --flushbufs /dev/ram0
  189. It is also possible to use initrd with an NFS-mounted root, see the
  190. :manpage:`pivot_root(8)` man page for details.
  191. Usage scenarios
  192. ---------------
  193. The main motivation for implementing initrd was to allow for modular
  194. kernel configuration at system installation. The procedure would work
  195. as follows:
  196. 1) system boots from floppy or other media with a minimal kernel
  197. (e.g. support for RAM disks, initrd, a.out, and the Ext2 FS) and
  198. loads initrd
  199. 2) ``/sbin/init`` determines what is needed to (1) mount the "real" root FS
  200. (i.e. device type, device drivers, file system) and (2) the
  201. distribution media (e.g. CD-ROM, network, tape, ...). This can be
  202. done by asking the user, by auto-probing, or by using a hybrid
  203. approach.
  204. 3) ``/sbin/init`` loads the necessary kernel modules
  205. 4) ``/sbin/init`` creates and populates the root file system (this doesn't
  206. have to be a very usable system yet)
  207. 5) ``/sbin/init`` invokes ``pivot_root`` to change the root file system and
  208. execs - via chroot - a program that continues the installation
  209. 6) the boot loader is installed
  210. 7) the boot loader is configured to load an initrd with the set of
  211. modules that was used to bring up the system (e.g. ``/initrd`` can be
  212. modified, then unmounted, and finally, the image is written from
  213. ``/dev/ram0`` or ``/dev/rd/0`` to a file)
  214. 8) now the system is bootable and additional installation tasks can be
  215. performed
  216. The key role of initrd here is to re-use the configuration data during
  217. normal system operation without requiring the use of a bloated "generic"
  218. kernel or re-compiling or re-linking the kernel.
  219. A second scenario is for installations where Linux runs on systems with
  220. different hardware configurations in a single administrative domain. In
  221. such cases, it is desirable to generate only a small set of kernels
  222. (ideally only one) and to keep the system-specific part of configuration
  223. information as small as possible. In this case, a common initrd could be
  224. generated with all the necessary modules. Then, only ``/sbin/init`` or a file
  225. read by it would have to be different.
  226. A third scenario is more convenient recovery disks, because information
  227. like the location of the root FS partition doesn't have to be provided at
  228. boot time, but the system loaded from initrd can invoke a user-friendly
  229. dialog and it can also perform some sanity checks (or even some form of
  230. auto-detection).
  231. Last not least, CD-ROM distributors may use it for better installation
  232. from CD, e.g. by using a boot floppy and bootstrapping a bigger RAM disk
  233. via initrd from CD; or by booting via a loader like ``LOADLIN`` or directly
  234. from the CD-ROM, and loading the RAM disk from CD without need of
  235. floppies.
  236. Obsolete root change mechanism
  237. ------------------------------
  238. The following mechanism was used before the introduction of pivot_root.
  239. Current kernels still support it, but you should _not_ rely on its
  240. continued availability.
  241. It works by mounting the "real" root device (i.e. the one set with rdev
  242. in the kernel image or with root=... at the boot command line) as the
  243. root file system when linuxrc exits. The initrd file system is then
  244. unmounted, or, if it is still busy, moved to a directory ``/initrd``, if
  245. such a directory exists on the new root file system.
  246. In order to use this mechanism, you do not have to specify the boot
  247. command options root, init, or rw. (If specified, they will affect
  248. the real root file system, not the initrd environment.)
  249. If /proc is mounted, the "real" root device can be changed from within
  250. linuxrc by writing the number of the new root FS device to the special
  251. file /proc/sys/kernel/real-root-dev, e.g.::
  252. # echo 0x301 >/proc/sys/kernel/real-root-dev
  253. Note that the mechanism is incompatible with NFS and similar file
  254. systems.
  255. This old, deprecated mechanism is commonly called ``change_root``, while
  256. the new, supported mechanism is called ``pivot_root``.
  257. Mixed change_root and pivot_root mechanism
  258. ------------------------------------------
  259. In case you did not want to use ``root=/dev/ram0`` to trigger the pivot_root
  260. mechanism, you may create both ``/linuxrc`` and ``/sbin/init`` in your initrd
  261. image.
  262. ``/linuxrc`` would contain only the following::
  263. #! /bin/sh
  264. mount -n -t proc proc /proc
  265. echo 0x0100 >/proc/sys/kernel/real-root-dev
  266. umount -n /proc
  267. Once linuxrc exited, the kernel would mount again your initrd as root,
  268. this time executing ``/sbin/init``. Again, it would be the duty of this init
  269. to build the right environment (maybe using the ``root= device`` passed on
  270. the cmdline) before the final execution of the real ``/sbin/init``.
  271. Resources
  272. ---------
  273. .. [#f1] Almesberger, Werner; "Booting Linux: The History and the Future"
  274. https://www.almesberger.net/cv/papers/ols2k-9.ps.gz
  275. .. [#f2] newlib package (experimental), with initrd example
  276. https://www.sourceware.org/newlib/
  277. .. [#f3] util-linux: Miscellaneous utilities for Linux
  278. https://www.kernel.org/pub/linux/utils/util-linux/