ext2.rst 19 KB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195196197198199200201202203204205206207208209210211212213214215216217218219220221222223224225226227228229230231232233234235236237238239240241242243244245246247248249250251252253254255256257258259260261262263264265266267268269270271272273274275276277278279280281282283284285286287288289290291292293294295296297298299300301302303304305306307308309310311312313314315316317318319320321322323324325326327328329330331332333334335336337338339340341342343344345346347348349350351352353354355356357358359360361362363364365366367368369370371372373374375376377378379380381382383384385386387388389390391392393394395396397398399
  1. .. SPDX-License-Identifier: GPL-2.0
  2. The Second Extended Filesystem
  3. ==============================
  4. ext2 was originally released in January 1993. Written by R\'emy Card,
  5. Theodore Ts'o and Stephen Tweedie, it was a major rewrite of the
  6. Extended Filesystem. It is currently still (April 2001) the predominant
  7. filesystem in use by Linux. There are also implementations available
  8. for NetBSD, FreeBSD, the GNU HURD, Windows 95/98/NT, OS/2 and RISC OS.
  9. Options
  10. =======
  11. Most defaults are determined by the filesystem superblock, and can be
  12. set using tune2fs(8). Kernel-determined defaults are indicated by (*).
  13. ==================== === ================================================
  14. bsddf (*) Makes ``df`` act like BSD.
  15. minixdf Makes ``df`` act like Minix.
  16. check=none, nocheck (*) Don't do extra checking of bitmaps on mount
  17. (check=normal and check=strict options removed)
  18. dax Use direct access (no page cache). See
  19. Documentation/filesystems/dax.txt.
  20. debug Extra debugging information is sent to the
  21. kernel syslog. Useful for developers.
  22. errors=continue Keep going on a filesystem error.
  23. errors=remount-ro Remount the filesystem read-only on an error.
  24. errors=panic Panic and halt the machine if an error occurs.
  25. grpid, bsdgroups Give objects the same group ID as their parent.
  26. nogrpid, sysvgroups New objects have the group ID of their creator.
  27. nouid32 Use 16-bit UIDs and GIDs.
  28. oldalloc Enable the old block allocator. Orlov should
  29. have better performance, we'd like to get some
  30. feedback if it's the contrary for you.
  31. orlov (*) Use the Orlov block allocator.
  32. (See http://lwn.net/Articles/14633/ and
  33. http://lwn.net/Articles/14446/.)
  34. resuid=n The user ID which may use the reserved blocks.
  35. resgid=n The group ID which may use the reserved blocks.
  36. sb=n Use alternate superblock at this location.
  37. user_xattr Enable "user." POSIX Extended Attributes
  38. (requires CONFIG_EXT2_FS_XATTR).
  39. nouser_xattr Don't support "user." extended attributes.
  40. acl Enable POSIX Access Control Lists support
  41. (requires CONFIG_EXT2_FS_POSIX_ACL).
  42. noacl Don't support POSIX ACLs.
  43. nobh Do not attach buffer_heads to file pagecache.
  44. quota, usrquota Enable user disk quota support
  45. (requires CONFIG_QUOTA).
  46. grpquota Enable group disk quota support
  47. (requires CONFIG_QUOTA).
  48. ==================== === ================================================
  49. noquota option ls silently ignored by ext2.
  50. Specification
  51. =============
  52. ext2 shares many properties with traditional Unix filesystems. It has
  53. the concepts of blocks, inodes and directories. It has space in the
  54. specification for Access Control Lists (ACLs), fragments, undeletion and
  55. compression though these are not yet implemented (some are available as
  56. separate patches). There is also a versioning mechanism to allow new
  57. features (such as journalling) to be added in a maximally compatible
  58. manner.
  59. Blocks
  60. ------
  61. The space in the device or file is split up into blocks. These are
  62. a fixed size, of 1024, 2048 or 4096 bytes (8192 bytes on Alpha systems),
  63. which is decided when the filesystem is created. Smaller blocks mean
  64. less wasted space per file, but require slightly more accounting overhead,
  65. and also impose other limits on the size of files and the filesystem.
  66. Block Groups
  67. ------------
  68. Blocks are clustered into block groups in order to reduce fragmentation
  69. and minimise the amount of head seeking when reading a large amount
  70. of consecutive data. Information about each block group is kept in a
  71. descriptor table stored in the block(s) immediately after the superblock.
  72. Two blocks near the start of each group are reserved for the block usage
  73. bitmap and the inode usage bitmap which show which blocks and inodes
  74. are in use. Since each bitmap is limited to a single block, this means
  75. that the maximum size of a block group is 8 times the size of a block.
  76. The block(s) following the bitmaps in each block group are designated
  77. as the inode table for that block group and the remainder are the data
  78. blocks. The block allocation algorithm attempts to allocate data blocks
  79. in the same block group as the inode which contains them.
  80. The Superblock
  81. --------------
  82. The superblock contains all the information about the configuration of
  83. the filing system. The primary copy of the superblock is stored at an
  84. offset of 1024 bytes from the start of the device, and it is essential
  85. to mounting the filesystem. Since it is so important, backup copies of
  86. the superblock are stored in block groups throughout the filesystem.
  87. The first version of ext2 (revision 0) stores a copy at the start of
  88. every block group, along with backups of the group descriptor block(s).
  89. Because this can consume a considerable amount of space for large
  90. filesystems, later revisions can optionally reduce the number of backup
  91. copies by only putting backups in specific groups (this is the sparse
  92. superblock feature). The groups chosen are 0, 1 and powers of 3, 5 and 7.
  93. The information in the superblock contains fields such as the total
  94. number of inodes and blocks in the filesystem and how many are free,
  95. how many inodes and blocks are in each block group, when the filesystem
  96. was mounted (and if it was cleanly unmounted), when it was modified,
  97. what version of the filesystem it is (see the Revisions section below)
  98. and which OS created it.
  99. If the filesystem is revision 1 or higher, then there are extra fields,
  100. such as a volume name, a unique identification number, the inode size,
  101. and space for optional filesystem features to store configuration info.
  102. All fields in the superblock (as in all other ext2 structures) are stored
  103. on the disc in little endian format, so a filesystem is portable between
  104. machines without having to know what machine it was created on.
  105. Inodes
  106. ------
  107. The inode (index node) is a fundamental concept in the ext2 filesystem.
  108. Each object in the filesystem is represented by an inode. The inode
  109. structure contains pointers to the filesystem blocks which contain the
  110. data held in the object and all of the metadata about an object except
  111. its name. The metadata about an object includes the permissions, owner,
  112. group, flags, size, number of blocks used, access time, change time,
  113. modification time, deletion time, number of links, fragments, version
  114. (for NFS) and extended attributes (EAs) and/or Access Control Lists (ACLs).
  115. There are some reserved fields which are currently unused in the inode
  116. structure and several which are overloaded. One field is reserved for the
  117. directory ACL if the inode is a directory and alternately for the top 32
  118. bits of the file size if the inode is a regular file (allowing file sizes
  119. larger than 2GB). The translator field is unused under Linux, but is used
  120. by the HURD to reference the inode of a program which will be used to
  121. interpret this object. Most of the remaining reserved fields have been
  122. used up for both Linux and the HURD for larger owner and group fields,
  123. The HURD also has a larger mode field so it uses another of the remaining
  124. fields to store the extra more bits.
  125. There are pointers to the first 12 blocks which contain the file's data
  126. in the inode. There is a pointer to an indirect block (which contains
  127. pointers to the next set of blocks), a pointer to a doubly-indirect
  128. block (which contains pointers to indirect blocks) and a pointer to a
  129. trebly-indirect block (which contains pointers to doubly-indirect blocks).
  130. The flags field contains some ext2-specific flags which aren't catered
  131. for by the standard chmod flags. These flags can be listed with lsattr
  132. and changed with the chattr command, and allow specific filesystem
  133. behaviour on a per-file basis. There are flags for secure deletion,
  134. undeletable, compression, synchronous updates, immutability, append-only,
  135. dumpable, no-atime, indexed directories, and data-journaling. Not all
  136. of these are supported yet.
  137. Directories
  138. -----------
  139. A directory is a filesystem object and has an inode just like a file.
  140. It is a specially formatted file containing records which associate
  141. each name with an inode number. Later revisions of the filesystem also
  142. encode the type of the object (file, directory, symlink, device, fifo,
  143. socket) to avoid the need to check the inode itself for this information
  144. (support for taking advantage of this feature does not yet exist in
  145. Glibc 2.2).
  146. The inode allocation code tries to assign inodes which are in the same
  147. block group as the directory in which they are first created.
  148. The current implementation of ext2 uses a singly-linked list to store
  149. the filenames in the directory; a pending enhancement uses hashing of the
  150. filenames to allow lookup without the need to scan the entire directory.
  151. The current implementation never removes empty directory blocks once they
  152. have been allocated to hold more files.
  153. Special files
  154. -------------
  155. Symbolic links are also filesystem objects with inodes. They deserve
  156. special mention because the data for them is stored within the inode
  157. itself if the symlink is less than 60 bytes long. It uses the fields
  158. which would normally be used to store the pointers to data blocks.
  159. This is a worthwhile optimisation as it we avoid allocating a full
  160. block for the symlink, and most symlinks are less than 60 characters long.
  161. Character and block special devices never have data blocks assigned to
  162. them. Instead, their device number is stored in the inode, again reusing
  163. the fields which would be used to point to the data blocks.
  164. Reserved Space
  165. --------------
  166. In ext2, there is a mechanism for reserving a certain number of blocks
  167. for a particular user (normally the super-user). This is intended to
  168. allow for the system to continue functioning even if non-privileged users
  169. fill up all the space available to them (this is independent of filesystem
  170. quotas). It also keeps the filesystem from filling up entirely which
  171. helps combat fragmentation.
  172. Filesystem check
  173. ----------------
  174. At boot time, most systems run a consistency check (e2fsck) on their
  175. filesystems. The superblock of the ext2 filesystem contains several
  176. fields which indicate whether fsck should actually run (since checking
  177. the filesystem at boot can take a long time if it is large). fsck will
  178. run if the filesystem was not cleanly unmounted, if the maximum mount
  179. count has been exceeded or if the maximum time between checks has been
  180. exceeded.
  181. Feature Compatibility
  182. ---------------------
  183. The compatibility feature mechanism used in ext2 is sophisticated.
  184. It safely allows features to be added to the filesystem, without
  185. unnecessarily sacrificing compatibility with older versions of the
  186. filesystem code. The feature compatibility mechanism is not supported by
  187. the original revision 0 (EXT2_GOOD_OLD_REV) of ext2, but was introduced in
  188. revision 1. There are three 32-bit fields, one for compatible features
  189. (COMPAT), one for read-only compatible (RO_COMPAT) features and one for
  190. incompatible (INCOMPAT) features.
  191. These feature flags have specific meanings for the kernel as follows:
  192. A COMPAT flag indicates that a feature is present in the filesystem,
  193. but the on-disk format is 100% compatible with older on-disk formats, so
  194. a kernel which didn't know anything about this feature could read/write
  195. the filesystem without any chance of corrupting the filesystem (or even
  196. making it inconsistent). This is essentially just a flag which says
  197. "this filesystem has a (hidden) feature" that the kernel or e2fsck may
  198. want to be aware of (more on e2fsck and feature flags later). The ext3
  199. HAS_JOURNAL feature is a COMPAT flag because the ext3 journal is simply
  200. a regular file with data blocks in it so the kernel does not need to
  201. take any special notice of it if it doesn't understand ext3 journaling.
  202. An RO_COMPAT flag indicates that the on-disk format is 100% compatible
  203. with older on-disk formats for reading (i.e. the feature does not change
  204. the visible on-disk format). However, an old kernel writing to such a
  205. filesystem would/could corrupt the filesystem, so this is prevented. The
  206. most common such feature, SPARSE_SUPER, is an RO_COMPAT feature because
  207. sparse groups allow file data blocks where superblock/group descriptor
  208. backups used to live, and ext2_free_blocks() refuses to free these blocks,
  209. which would leading to inconsistent bitmaps. An old kernel would also
  210. get an error if it tried to free a series of blocks which crossed a group
  211. boundary, but this is a legitimate layout in a SPARSE_SUPER filesystem.
  212. An INCOMPAT flag indicates the on-disk format has changed in some
  213. way that makes it unreadable by older kernels, or would otherwise
  214. cause a problem if an old kernel tried to mount it. FILETYPE is an
  215. INCOMPAT flag because older kernels would think a filename was longer
  216. than 256 characters, which would lead to corrupt directory listings.
  217. The COMPRESSION flag is an obvious INCOMPAT flag - if the kernel
  218. doesn't understand compression, you would just get garbage back from
  219. read() instead of it automatically decompressing your data. The ext3
  220. RECOVER flag is needed to prevent a kernel which does not understand the
  221. ext3 journal from mounting the filesystem without replaying the journal.
  222. For e2fsck, it needs to be more strict with the handling of these
  223. flags than the kernel. If it doesn't understand ANY of the COMPAT,
  224. RO_COMPAT, or INCOMPAT flags it will refuse to check the filesystem,
  225. because it has no way of verifying whether a given feature is valid
  226. or not. Allowing e2fsck to succeed on a filesystem with an unknown
  227. feature is a false sense of security for the user. Refusing to check
  228. a filesystem with unknown features is a good incentive for the user to
  229. update to the latest e2fsck. This also means that anyone adding feature
  230. flags to ext2 also needs to update e2fsck to verify these features.
  231. Metadata
  232. --------
  233. It is frequently claimed that the ext2 implementation of writing
  234. asynchronous metadata is faster than the ffs synchronous metadata
  235. scheme but less reliable. Both methods are equally resolvable by their
  236. respective fsck programs.
  237. If you're exceptionally paranoid, there are 3 ways of making metadata
  238. writes synchronous on ext2:
  239. - per-file if you have the program source: use the O_SYNC flag to open()
  240. - per-file if you don't have the source: use "chattr +S" on the file
  241. - per-filesystem: add the "sync" option to mount (or in /etc/fstab)
  242. the first and last are not ext2 specific but do force the metadata to
  243. be written synchronously. See also Journaling below.
  244. Limitations
  245. -----------
  246. There are various limits imposed by the on-disk layout of ext2. Other
  247. limits are imposed by the current implementation of the kernel code.
  248. Many of the limits are determined at the time the filesystem is first
  249. created, and depend upon the block size chosen. The ratio of inodes to
  250. data blocks is fixed at filesystem creation time, so the only way to
  251. increase the number of inodes is to increase the size of the filesystem.
  252. No tools currently exist which can change the ratio of inodes to blocks.
  253. Most of these limits could be overcome with slight changes in the on-disk
  254. format and using a compatibility flag to signal the format change (at
  255. the expense of some compatibility).
  256. ===================== ======= ======= ======= ========
  257. Filesystem block size 1kB 2kB 4kB 8kB
  258. ===================== ======= ======= ======= ========
  259. File size limit 16GB 256GB 2048GB 2048GB
  260. Filesystem size limit 2047GB 8192GB 16384GB 32768GB
  261. ===================== ======= ======= ======= ========
  262. There is a 2.4 kernel limit of 2048GB for a single block device, so no
  263. filesystem larger than that can be created at this time. There is also
  264. an upper limit on the block size imposed by the page size of the kernel,
  265. so 8kB blocks are only allowed on Alpha systems (and other architectures
  266. which support larger pages).
  267. There is an upper limit of 32000 subdirectories in a single directory.
  268. There is a "soft" upper limit of about 10-15k files in a single directory
  269. with the current linear linked-list directory implementation. This limit
  270. stems from performance problems when creating and deleting (and also
  271. finding) files in such large directories. Using a hashed directory index
  272. (under development) allows 100k-1M+ files in a single directory without
  273. performance problems (although RAM size becomes an issue at this point).
  274. The (meaningless) absolute upper limit of files in a single directory
  275. (imposed by the file size, the realistic limit is obviously much less)
  276. is over 130 trillion files. It would be higher except there are not
  277. enough 4-character names to make up unique directory entries, so they
  278. have to be 8 character filenames, even then we are fairly close to
  279. running out of unique filenames.
  280. Journaling
  281. ----------
  282. A journaling extension to the ext2 code has been developed by Stephen
  283. Tweedie. It avoids the risks of metadata corruption and the need to
  284. wait for e2fsck to complete after a crash, without requiring a change
  285. to the on-disk ext2 layout. In a nutshell, the journal is a regular
  286. file which stores whole metadata (and optionally data) blocks that have
  287. been modified, prior to writing them into the filesystem. This means
  288. it is possible to add a journal to an existing ext2 filesystem without
  289. the need for data conversion.
  290. When changes to the filesystem (e.g. a file is renamed) they are stored in
  291. a transaction in the journal and can either be complete or incomplete at
  292. the time of a crash. If a transaction is complete at the time of a crash
  293. (or in the normal case where the system does not crash), then any blocks
  294. in that transaction are guaranteed to represent a valid filesystem state,
  295. and are copied into the filesystem. If a transaction is incomplete at
  296. the time of the crash, then there is no guarantee of consistency for
  297. the blocks in that transaction so they are discarded (which means any
  298. filesystem changes they represent are also lost).
  299. Check Documentation/filesystems/ext4/ if you want to read more about
  300. ext4 and journaling.
  301. References
  302. ==========
  303. ======================= ===============================================
  304. The kernel source file:/usr/src/linux/fs/ext2/
  305. e2fsprogs (e2fsck) http://e2fsprogs.sourceforge.net/
  306. Design & Implementation http://e2fsprogs.sourceforge.net/ext2intro.html
  307. Journaling (ext3) ftp://ftp.uk.linux.org/pub/linux/sct/fs/jfs/
  308. Filesystem Resizing http://ext2resize.sourceforge.net/
  309. Compression [1]_ http://e2compr.sourceforge.net/
  310. ======================= ===============================================
  311. Implementations for:
  312. ======================= ===========================================================
  313. Windows 95/98/NT/2000 http://www.chrysocome.net/explore2fs
  314. Windows 95 [1]_ http://www.yipton.net/content.html#FSDEXT2
  315. DOS client [1]_ ftp://metalab.unc.edu/pub/Linux/system/filesystems/ext2/
  316. OS/2 [2]_ ftp://metalab.unc.edu/pub/Linux/system/filesystems/ext2/
  317. RISC OS client http://www.esw-heim.tu-clausthal.de/~marco/smorbrod/IscaFS/
  318. ======================= ===========================================================
  319. .. [1] no longer actively developed/supported (as of Apr 2001)
  320. .. [2] no longer actively developed/supported (as of Mar 2009)