1 .. SPDX-License-Identifier: GPL-2.0
2
3
4 ==============================
5 The Second Extended Filesystem
6 ==============================
7
8 ext2 was originally released in January 1993.
9 Theodore Ts'o and Stephen Tweedie, it was a ma
10 Extended Filesystem. It is currently still (A
11 filesystem in use by Linux. There are also im
12 for NetBSD, FreeBSD, the GNU HURD, Windows 95/
13
14 Options
15 =======
16
17 Most defaults are determined by the filesystem
18 set using tune2fs(8). Kernel-determined defaul
19
20 ==================== === ==============
21 bsddf (*) Makes ``df`` a
22 minixdf Makes ``df`` a
23
24 check=none, nocheck (*) Don't do extra
25 (check=normal
26
27 dax Use direct acc
28 Documentation/
29
30 debug Extra debuggin
31 kernel syslog.
32
33 errors=continue Keep going on
34 errors=remount-ro Remount the fi
35 errors=panic Panic and halt
36
37 grpid, bsdgroups Give objects t
38 nogrpid, sysvgroups New objects ha
39
40 nouid32 Use 16-bit UID
41
42 oldalloc Enable the old
43 have better pe
44 feedback if it
45 orlov (*) Use the Orlov
46 (See http://lw
47 http://lwn.net
48
49 resuid=n The user ID wh
50 resgid=n The group ID w
51
52 sb=n Use alternate
53
54 user_xattr Enable "user."
55 (requires CONF
56 nouser_xattr Don't support
57
58 acl Enable POSIX A
59 (requires CONF
60 noacl Don't support
61
62 quota, usrquota Enable user di
63 (requires CONF
64
65 grpquota Enable group d
66 (requires CONF
67 ==================== === ==============
68
69 noquota option ls silently ignored by ext2.
70
71
72 Specification
73 =============
74
75 ext2 shares many properties with traditional U
76 the concepts of blocks, inodes and directories
77 specification for Access Control Lists (ACLs),
78 compression though these are not yet implement
79 separate patches). There is also a versioning
80 features (such as journalling) to be added in
81 manner.
82
83 Blocks
84 ------
85
86 The space in the device or file is split up in
87 a fixed size, of 1024, 2048 or 4096 bytes (819
88 which is decided when the filesystem is create
89 less wasted space per file, but require slight
90 and also impose other limits on the size of fi
91
92 Block Groups
93 ------------
94
95 Blocks are clustered into block groups in orde
96 and minimise the amount of head seeking when r
97 of consecutive data. Information about each b
98 descriptor table stored in the block(s) immedi
99 Two blocks near the start of each group are re
100 bitmap and the inode usage bitmap which show w
101 are in use. Since each bitmap is limited to a
102 that the maximum size of a block group is 8 ti
103
104 The block(s) following the bitmaps in each blo
105 as the inode table for that block group and th
106 blocks. The block allocation algorithm attemp
107 in the same block group as the inode which con
108
109 The Superblock
110 --------------
111
112 The superblock contains all the information ab
113 the filing system. The primary copy of the su
114 offset of 1024 bytes from the start of the dev
115 to mounting the filesystem. Since it is so im
116 the superblock are stored in block groups thro
117 The first version of ext2 (revision 0) stores
118 every block group, along with backups of the g
119 Because this can consume a considerable amount
120 filesystems, later revisions can optionally re
121 copies by only putting backups in specific gro
122 superblock feature). The groups chosen are 0,
123
124 The information in the superblock contains fie
125 number of inodes and blocks in the filesystem
126 how many inodes and blocks are in each block g
127 was mounted (and if it was cleanly unmounted),
128 what version of the filesystem it is (see the
129 and which OS created it.
130
131 If the filesystem is revision 1 or higher, the
132 such as a volume name, a unique identification
133 and space for optional filesystem features to
134
135 All fields in the superblock (as in all other
136 on the disc in little endian format, so a file
137 machines without having to know what machine i
138
139 Inodes
140 ------
141
142 The inode (index node) is a fundamental concep
143 Each object in the filesystem is represented b
144 structure contains pointers to the filesystem
145 data held in the object and all of the metadat
146 its name. The metadata about an object includ
147 group, flags, size, number of blocks used, acc
148 modification time, deletion time, number of li
149 (for NFS) and extended attributes (EAs) and/or
150
151 There are some reserved fields which are curre
152 structure and several which are overloaded. O
153 directory ACL if the inode is a directory and
154 bits of the file size if the inode is a regula
155 larger than 2GB). The translator field is unu
156 by the HURD to reference the inode of a progra
157 interpret this object. Most of the remaining
158 used up for both Linux and the HURD for larger
159 The HURD also has a larger mode field so it us
160 fields to store the extra more bits.
161
162 There are pointers to the first 12 blocks whic
163 in the inode. There is a pointer to an indire
164 pointers to the next set of blocks), a pointer
165 block (which contains pointers to indirect blo
166 trebly-indirect block (which contains pointers
167
168 The flags field contains some ext2-specific fl
169 for by the standard chmod flags. These flags
170 and changed with the chattr command, and allow
171 behaviour on a per-file basis. There are flag
172 undeletable, compression, synchronous updates,
173 dumpable, no-atime, indexed directories, and d
174 of these are supported yet.
175
176 Directories
177 -----------
178
179 A directory is a filesystem object and has an
180 It is a specially formatted file containing re
181 each name with an inode number. Later revisio
182 encode the type of the object (file, directory
183 socket) to avoid the need to check the inode i
184 (support for taking advantage of this feature
185 Glibc 2.2).
186
187 The inode allocation code tries to assign inod
188 block group as the directory in which they are
189
190 The current implementation of ext2 uses a sing
191 the filenames in the directory; a pending enha
192 filenames to allow lookup without the need to
193
194 The current implementation never removes empty
195 have been allocated to hold more files.
196
197 Special files
198 -------------
199
200 Symbolic links are also filesystem objects wit
201 special mention because the data for them is s
202 itself if the symlink is less than 60 bytes lo
203 which would normally be used to store the poin
204 This is a worthwhile optimisation as it we avo
205 block for the symlink, and most symlinks are l
206
207 Character and block special devices never have
208 them. Instead, their device number is stored
209 the fields which would be used to point to the
210
211 Reserved Space
212 --------------
213
214 In ext2, there is a mechanism for reserving a
215 for a particular user (normally the super-user
216 allow for the system to continue functioning e
217 fill up all the space available to them (this
218 quotas). It also keeps the filesystem from fi
219 helps combat fragmentation.
220
221 Filesystem check
222 ----------------
223
224 At boot time, most systems run a consistency c
225 filesystems. The superblock of the ext2 files
226 fields which indicate whether fsck should actu
227 the filesystem at boot can take a long time if
228 run if the filesystem was not cleanly unmounte
229 count has been exceeded or if the maximum time
230 exceeded.
231
232 Feature Compatibility
233 ---------------------
234
235 The compatibility feature mechanism used in ex
236 It safely allows features to be added to the f
237 unnecessarily sacrificing compatibility with o
238 filesystem code. The feature compatibility me
239 the original revision 0 (EXT2_GOOD_OLD_REV) of
240 revision 1. There are three 32-bit fields, on
241 (COMPAT), one for read-only compatible (RO_COM
242 incompatible (INCOMPAT) features.
243
244 These feature flags have specific meanings for
245
246 A COMPAT flag indicates that a feature is pres
247 but the on-disk format is 100% compatible with
248 a kernel which didn't know anything about this
249 the filesystem without any chance of corruptin
250 making it inconsistent). This is essentially
251 "this filesystem has a (hidden) feature" that
252 want to be aware of (more on e2fsck and featur
253 HAS_JOURNAL feature is a COMPAT flag because t
254 a regular file with data blocks in it so the k
255 take any special notice of it if it doesn't un
256
257 An RO_COMPAT flag indicates that the on-disk f
258 with older on-disk formats for reading (i.e. t
259 the visible on-disk format). However, an old
260 filesystem would/could corrupt the filesystem,
261 most common such feature, SPARSE_SUPER, is an
262 sparse groups allow file data blocks where sup
263 backups used to live, and ext2_free_blocks() r
264 which would leading to inconsistent bitmaps.
265 get an error if it tried to free a series of b
266 boundary, but this is a legitimate layout in a
267
268 An INCOMPAT flag indicates the on-disk format
269 way that makes it unreadable by older kernels,
270 cause a problem if an old kernel tried to moun
271 INCOMPAT flag because older kernels would thin
272 than 256 characters, which would lead to corru
273 The COMPRESSION flag is an obvious INCOMPAT fl
274 doesn't understand compression, you would just
275 read() instead of it automatically decompressi
276 RECOVER flag is needed to prevent a kernel whi
277 ext3 journal from mounting the filesystem with
278
279 For e2fsck, it needs to be more strict with th
280 flags than the kernel. If it doesn't understa
281 RO_COMPAT, or INCOMPAT flags it will refuse to
282 because it has no way of verifying whether a g
283 or not. Allowing e2fsck to succeed on a files
284 feature is a false sense of security for the u
285 a filesystem with unknown features is a good i
286 update to the latest e2fsck. This also means
287 flags to ext2 also needs to update e2fsck to v
288
289 Metadata
290 --------
291
292 It is frequently claimed that the ext2 impleme
293 asynchronous metadata is faster than the ffs s
294 scheme but less reliable. Both methods are eq
295 respective fsck programs.
296
297 If you're exceptionally paranoid, there are 3
298 writes synchronous on ext2:
299
300 - per-file if you have the program source: use
301 - per-file if you don't have the source: use "
302 - per-filesystem: add the "sync" option to mou
303
304 the first and last are not ext2 specific but d
305 be written synchronously. See also Journaling
306
307 Limitations
308 -----------
309
310 There are various limits imposed by the on-dis
311 limits are imposed by the current implementati
312 Many of the limits are determined at the time
313 created, and depend upon the block size chosen
314 data blocks is fixed at filesystem creation ti
315 increase the number of inodes is to increase t
316 No tools currently exist which can change the
317
318 Most of these limits could be overcome with sl
319 format and using a compatibility flag to signa
320 the expense of some compatibility).
321
322 ===================== ======= ======= =
323 Filesystem block size 1kB 2kB
324 ===================== ======= ======= =
325 File size limit 16GB 256GB
326 Filesystem size limit 2047GB 8192GB 1
327 ===================== ======= ======= =
328
329 There is a 2.4 kernel limit of 2048GB for a si
330 filesystem larger than that can be created at
331 an upper limit on the block size imposed by th
332 so 8kB blocks are only allowed on Alpha system
333 which support larger pages).
334
335 There is an upper limit of 32000 subdirectorie
336
337 There is a "soft" upper limit of about 10-15k
338 with the current linear linked-list directory
339 stems from performance problems when creating
340 finding) files in such large directories. Usi
341 (under development) allows 100k-1M+ files in a
342 performance problems (although RAM size become
343
344 The (meaningless) absolute upper limit of file
345 (imposed by the file size, the realistic limit
346 is over 130 trillion files. It would be highe
347 enough 4-character names to make up unique dir
348 have to be 8 character filenames, even then we
349 running out of unique filenames.
350
351 Journaling
352 ----------
353
354 A journaling extension to the ext2 code has be
355 Tweedie. It avoids the risks of metadata corr
356 wait for e2fsck to complete after a crash, wit
357 to the on-disk ext2 layout. In a nutshell, th
358 file which stores whole metadata (and optional
359 been modified, prior to writing them into the
360 it is possible to add a journal to an existing
361 the need for data conversion.
362
363 When changes to the filesystem (e.g. a file is
364 a transaction in the journal and can either be
365 the time of a crash. If a transaction is comp
366 (or in the normal case where the system does n
367 in that transaction are guaranteed to represen
368 and are copied into the filesystem. If a tran
369 the time of the crash, then there is no guaran
370 the blocks in that transaction so they are dis
371 filesystem changes they represent are also los
372 Check Documentation/filesystems/ext4/ if you w
373 ext4 and journaling.
374
375 References
376 ==========
377
378 ======================= ======================
379 The kernel source file:/usr/src/linux/fs
380 e2fsprogs (e2fsck) http://e2fsprogs.sourc
381 Design & Implementation http://e2fsprogs.sourc
382 Journaling (ext3) ftp://ftp.uk.linux.org
383 Filesystem Resizing http://ext2resize.sour
384 Compression [1]_ http://e2compr.sourcef
385 ======================= ======================
386
387 Implementations for:
388
389 ======================= ======================
390 Windows 95/98/NT/2000 http://www.chrysocome.
391 Windows 95 [1]_ http://www.yipton.net/
392 DOS client [1]_ ftp://metalab.unc.edu/
393 OS/2 [2]_ ftp://metalab.unc.edu/
394 RISC OS client http://www.esw-heim.tu
395 ======================= ======================
396
397 .. [1] no longer actively developed/supported
398 .. [2] no longer actively developed/supported
Linux® is a registered trademark of Linus Torvalds in the United States and other countries.
TOMOYO® is a registered trademark of NTT DATA CORPORATION.