1 .. SPDX-License-Identifier: GPL-2.0
2
3 ========================
4 ext4 General Information
5 ========================
6
7 Ext4 is an advanced level of the ext3 filesyst
8 scalability and reliability enhancements for s
9 (64 bit) in keeping with increasing disk capac
10 feature requirements.
11
12 Mailing list: linux-ext4@vger.kernel.org
13 Web site: http://ext4.wiki.kernel.org
14
15
16 Quick usage instructions
17 ========================
18
19 Note: More extensive information for getting s
20 found at the ext4 wiki site at the URL:
21 http://ext4.wiki.kernel.org/index.php/Ext4_How
22
23 - The latest version of e2fsprogs can be fou
24
25 https://www.kernel.org/pub/linux/kernel/pe
26
27 or
28
29 http://sourceforge.net/project/showfiles.p
30
31 or grab the latest git repository from
32
33 https://git.kernel.org/pub/scm/fs/ext2/e2fs
34
35 - Create a new filesystem using the ext4 fil
36
37 # mke2fs -t ext4 /dev/hda1
38
39 Or to configure an existing ext3 filesyste
40
41 # tune2fs -O extents /dev/hda1
42
43 If the filesystem was created with 128 byt
44 converted to use 256 byte for greater effi
45
46 # tune2fs -I 256 /dev/hda1
47
48 - Mounting:
49
50 # mount -t ext4 /dev/hda1 /wherever
51
52 - When comparing performance with other file
53 important to try multiple workloads; very
54 workload parameter can completely change t
55 filesystems do well compared to others. W
56 note that ext4 enables write barriers by d
57 not enable write barriers by default. So
58 explicitly specify whether barriers are en
59 '-o barriers=[0|1]' mount option for both
60 for a fair comparison. When tuning ext3 f
61 it is often worthwhile to try changing the
62 data=writeback' can be faster for some wor
63 running mounted with data=writeback can po
64 exposed in recently written files in case
65 which could be a security exposure in some
66 the filesystem with a large journal can al
67 metadata-intensive workloads.
68
69 Features
70 ========
71
72 Currently Available
73 -------------------
74
75 * ability to use filesystems > 16TB (e2fsprogs
76 * extent format reduces metadata overhead (RAM
77 * extent format more robust in face of on-disk
78 * internal redundancy in tree
79 * improved file allocation (multi-block alloc)
80 * lift 32000 subdirectory limit imposed by i_l
81 * nsec timestamps for mtime, atime, ctime, cre
82 * inode version field on disk (NFSv4, Lustre)
83 * reduced e2fsck time via uninit_bg feature
84 * journal checksumming for robustness, perform
85 * persistent file preallocation (e.g for strea
86 * ability to pack bitmaps and inode tables int
87 flex_bg feature
88 * large file support
89 * inode allocation using large virtual block g
90 * delayed allocation
91 * large block (up to pagesize) support
92 * efficient new ordered mode in JBD2 and ext4
93 the ordering)
94 * Case-insensitive file name lookups
95 * file-based encryption support (fscrypt)
96 * file-based verity support (fsverity)
97
98 [1] Filesystems with a block size of 1k may se
99 directory hash tree having a maximum depth of
100
101 case-insensitive file name lookups
102 ==============================================
103
104 The case-insensitive file name lookup feature
105 per-directory basis, allowing the user to mix
106 case-sensitive directories in the same filesys
107 flipping the +F inode attribute of an empty di
108 case-insensitive string match operation is onl
109 text in encoded in a byte sequence. For that
110 case-insensitive directories, the filesystem m
111 casefold feature, which stores the filesystem-
112 model used. By default, the charset adopted i
113 Unicode (12.1.0, by the time of this writing),
114 form. The comparison algorithm is implemented
115 strings to the Canonical decomposition form, a
116 followed by a byte per byte comparison.
117
118 The case-awareness is name-preserving on the d
119 name provided by userspace is a byte-per-byte
120 written in the disk. The Unicode normalizatio
121 kernel is thus an internal representation, and
122 userspace nor to the disk, with the important
123 used on large case-insensitive directories wit
124 directories, the hash must be calculated using
125 the filename, meaning that the normalization f
126 impact on where the directory entry is stored.
127
128 When we change from viewing filenames as opaqu
129 them as encoded strings we need to address wha
130 tries to create a file with an invalid name.
131 within the kernel leaves the decision of what
132 filesystem, which select its preferred behavio
133 the strict mode. When Ext4 encounters one of
134 filesystem did not require strict mode, it fal
135 entire string as an opaque byte sequence, whic
136 operate on that file, but the case-insensitive
137
138 Options
139 =======
140
141 When mounting an ext4 filesystem, the followin
142 (*) == default
143
144 ro
145 Mount filesystem read only. Note that
146 thus write to the partition) even when
147 options "ro,noload" can be used to pre
148
149 journal_checksum
150 Enable checksumming of the journal tra
151 recovery code in e2fsck and the kernel
152 kernel. It is a compatible change and
153 kernels.
154
155 journal_async_commit
156 Commit block can be written to disk wi
157 blocks. If enabled older kernels canno
158 enable 'journal_checksum' internally.
159
160 journal_path=path, journal_dev=devnum
161 When the external journal device's maj
162 these options allow the user to specif
163 journal device is identified through e
164 encoded in devnum, or via a path to th
165
166 norecovery, noload
167 Don't load the journal on mounting. N
168 not unmounted cleanly, skipping the jo
169 filesystem containing inconsistencies
170 problems.
171
172 data=journal
173 All data are committed into the journa
174 main file system. Enabling this mode
175 and O_DIRECT support.
176
177 data=ordered (*)
178 All data are forced directly out to th
179 metadata being committed to the journa
180
181 data=writeback
182 Data ordering is not preserved, data m
183 system after its metadata has been com
184
185 commit=nrsec (*)
186 This setting limits the maximum age of
187 'nrsec' seconds. The default value is
188 you lose your power, you will lose as
189 metadata changes (your filesystem will
190 to the journaling). This default value
191 performance, but it's good for data-sa
192 the same effect as leaving it at the d
193 to very large values will improve perf
194 delayed allocation even older data can
195 writeback of those data begins only af
196 /proc/sys/vm/dirty_expire_centisecs.
197
198 barrier=<0|1(*)>, barrier(*), nobarrier
199 This enables/disables the use of write
200 barrier=0 disables, barrier=1 enables.
201 which can support barriers, and if jbd
202 write, it will disable again with a wa
203 proper on-disk ordering of journal com
204 caches safe to use, at some performanc
205 battery-backed in one way or another,
206 improve performance. The mount option
207 also be used to enable or disable barr
208 ext4 mount options.
209
210 inode_readahead_blks=n
211 This tuning parameter controls the max
212 that ext4's inode table readahead algo
213 buffer cache. The default value is 32
214
215 bsddf (*)
216 Make 'df' act like BSD.
217
218 minixdf
219 Make 'df' act like Minix.
220
221 debug
222 Extra debugging information is sent to
223
224 abort
225 Simulate the effects of calling ext4_a
226 This is normally used while remounting
227 mounted.
228
229 errors=remount-ro
230 Remount the filesystem read-only on an
231
232 errors=continue
233 Keep going on a filesystem error.
234
235 errors=panic
236 Panic and halt the machine if an error
237 override the errors behavior specified
238 configured using tune2fs)
239
240 data_err=ignore(*)
241 Just print an error message if an erro
242 ordered mode.
243 data_err=abort
244 Abort the journal if an error occurs i
245 mode.
246
247 grpid | bsdgroups
248 New objects have the group ID of their
249
250 nogrpid (*) | sysvgroups
251 New objects have the group ID of their
252
253 resgid=n
254 The group ID which may use the reserve
255
256 resuid=n
257 The user ID which may use the reserved
258
259 sb=
260 Use alternate superblock at this locat
261
262 quota, noquota, grpquota, usrquota
263 These options are ignored by the files
264 quota tools to recognize volumes where
265 documentation in the quota-tools packa
266 (http://sourceforge.net/projects/linux
267
268 jqfmt=<quota type>, usrjquota=<file>, grpjqu
269 These options tell filesystem details
270 information can be properly updated du
271 the above quota options. See documenta
272 for more details (http://sourceforge.n
273
274 stripe=n
275 Number of filesystem blocks that mball
276 size and alignment. For RAID5/6 system
277 data disks * RAID chunk size in file
278
279 delalloc (*)
280 Defer block allocation until just befo
281 in question. This allows ext4 to bett
282 efficiently.
283
284 nodelalloc
285 Disable delayed allocation. Blocks ar
286 copied from userspace to the page cach
287 call or when an mmap'ed page which was
288 written for the first time.
289
290 max_batch_time=usec
291 Maximum amount of time ext4 should wai
292 operations to be batch together with a
293 Since a synchronous write operation is
294 a wait for the I/O complete, it doesn'
295 throughput win, we wait for a small am
296 transactions can piggyback on the sync
297 used is designed to automatically tune
298 measuring the amount of time (on avera
299 committing a transaction. Call this t
300 time that the transaction has been run
301 time, ext4 will try sleeping for the c
302 operations will join the transaction.
303 the max_batch_time, which defaults to
304 optimization can be turned off entirel
305
306 min_batch_time=usec
307 This parameter sets the commit time (a
308 min_batch_time. It defaults to zero m
309 parameter may improve the throughput o
310 workloads on very fast disks, at the c
311
312 journal_ioprio=prio
313 The I/O priority (from 0 to 7, where 0
314 should be used for I/O operations subm
315 commit operation. This defaults to 3,
316 priority than the default I/O priority
317
318 auto_da_alloc(*), noauto_da_alloc
319 Many broken applications don't use fsy
320 files via patterns such as fd = open("
321 rename("foo.new", "foo"), or worse yet
322 O_TRUNC)/write(fd,..)/close(fd). If a
323 will detect the replace-via-rename and
324 and force that any delayed allocation
325 the next journal commit, in the defaul
326 blocks of the new file are forced to d
327 is committed. This provides roughly t
328 ext3, and avoids the "zero-length" pro
329 system crashes before the delayed allo
330
331 noinit_itable
332 Do not initialize any uninitialized in
333 background. This feature may be used
334 install process can complete as quickl
335 initialization process would then be d
336 file system is unmounted.
337
338 init_itable=n
339 The lazy itable init code will wait n
340 it took to zero out the previous block
341 minimizes the impact on the system per
342 inode table is being initialized.
343
344 discard, nodiscard(*)
345 Controls whether ext4 should issue dis
346 underlying block device when blocks ar
347 devices and sparse/thinly-provisioned
348 until sufficient testing has been done
349
350 nouid32
351 Disables 32-bit UIDs and GIDs. This i
352 older kernels which only store and exp
353
354 block_validity(*), noblock_validity
355 These options enable or disable the in
356 filesystem metadata blocks within inte
357 allows multi- block allocator and othe
358 corrupted allocation bitmaps which cau
359 overlap with filesystem metadata block
360
361 dioread_lock, dioread_nolock
362 Controls whether or not ext4 should us
363 dioread_nolock option is specified ext
364 extent before buffer write and convert
365 IO completes. This approach allows ext
366 mutex, which improves scalability on h
367 does not work with data journaling and
368 ignored with kernel warning. Note that
369 used for extent-based files. Because
370 comprises it is off by default (e.g. d
371
372 max_dir_size_kb=n
373 This limits the size of directories so
374 beyond the specified limit in kilobyte
375 This is useful in memory constrained e
376 directory can cause severe performance
377 Of Memory killer. (For example, if th
378 available, a 176mb directory may serio
379
380 i_version
381 Enable 64-bit inode version support. T
382
383 dax
384 Use direct access (no page cache). Se
385 Documentation/filesystems/dax.rst. No
386 incompatible with data=journal.
387
388 inlinecrypt
389 When possible, encrypt/decrypt the con
390 blk-crypto framework rather than files
391 allows the use of inline encryption ha
392 unaffected. For more details, see
393 Documentation/block/inline-encryption.
394
395 Data Mode
396 =========
397 There are 3 different data modes:
398
399 * writeback mode
400
401 In data=writeback mode, ext4 does not journa
402 a similar level of journaling as that of XFS
403 mode - metadata journaling. A crash+recover
404 appear in files which were written shortly b
405 typically provide the best ext4 performance.
406
407 * ordered mode
408
409 In data=ordered mode, ext4 only officially j
410 groups metadata information related to data
411 a single unit called a transaction. When it
412 out to disk, the associated data blocks are
413 mode performs slightly slower than writeback
414 journal mode.
415
416 * journal mode
417
418 data=journal mode provides full data and met
419 written to the journal first, and then to it
420 a crash, the journal can be replayed, bringi
421 consistent state. This mode is the slowest
422 from and written to disk at the same time wh
423 modes. Enabling this mode will disable dela
424 support.
425
426 /proc entries
427 =============
428
429 Information about mounted ext4 file systems ca
430 /proc/fs/ext4. Each mounted filesystem will h
431 /proc/fs/ext4 based on its device name (i.e.,
432 /proc/fs/ext4/dm-0). The files in each per-d
433 in table below.
434
435 Files in /proc/fs/ext4/<devname>
436
437 mb_groups
438 details of multiblock allocator buddy
439
440 /sys entries
441 ============
442
443 Information about mounted ext4 file systems ca
444 /sys/fs/ext4. Each mounted filesystem will ha
445 /sys/fs/ext4 based on its device name (i.e., /
446 /sys/fs/ext4/dm-0). The files in each per-de
447 in table below.
448
449 Files in /sys/fs/ext4/<devname>:
450
451 (see also Documentation/ABI/testing/sysfs-fs-e
452
453 delayed_allocation_blocks
454 This file is read-only and shows the n
455 the page cache, but which do not have
456 allocated yet.
457
458 inode_goal
459 Tuning parameter which (if non-zero) c
460 the inode allocator in preference to a
461 This is intended for debugging use onl
462 systems.
463
464 inode_readahead_blks
465 Tuning parameter which controls the ma
466 blocks that ext4's inode table readahe
467 the buffer cache.
468
469 lifetime_write_kbytes
470 This file is read-only and shows the n
471 have been written to this filesystem s
472
473 max_writeback_mb_bump
474 The maximum number of megabytes the wr
475 out before move on to another inode.
476
477 mb_group_prealloc
478 The multiblock allocator will round up
479 multiple of this tuning parameter if t
480 ext4 superblock
481
482 mb_max_to_scan
483 The maximum number of extents the mult
484 find the best extent.
485
486 mb_min_to_scan
487 The minimum number of extents the mult
488 find the best extent.
489
490 mb_order2_req
491 Tuning parameter which controls the mi
492 power of 2) where the buddy cache is u
493
494 mb_stats
495 Controls whether the multiblock alloca
496 which are shown during the unmount. 1
497 means not to collect statistics.
498
499 mb_stream_req
500 Files which have fewer blocks than thi
501 their blocks allocated out of a block
502 pool, so that small files are packed c
503 will have its blocks allocated out of
504 pool.
505
506 session_write_kbytes
507 This file is read-only and shows the n
508 have been written to this filesystem s
509
510 reserved_clusters
511 This is RW file and contains number of
512 system which will be used in the speci
513 zeroout, unexpected ENOSPC, or possibl
514 4096 clusters, whichever is smaller an
515 can never exceed number of clusters in
516 enough space for the reserved space wh
517 _not_ fail.
518
519 Ioctls
520 ======
521
522 Ext4 implements various ioctls which can be us
523 ext4-specific functionality. An incomplete lis
524 table below. This list includes truly ext4-spe
525 well as ioctls that may have been ext4-specifi
526 by some other filesystem(s) too (``FS_IOC_*``)
527
528 Table of Ext4 ioctls
529
530 FS_IOC_GETFLAGS
531 Get additional attributes associated w
532 an integer bitfield, with bit values d
533
534 FS_IOC_SETFLAGS
535 Set additional attributes associated w
536 an integer bitfield, with bit values d
537
538 EXT4_IOC_GETVERSION, EXT4_IOC_GETVERSION_OLD
539 Get the inode i_generation number stor
540 i_generation number is normally change
541 and it is particularly useful for netw
542 version of this ioctl is an alias for
543
544 EXT4_IOC_SETVERSION, EXT4_IOC_SETVERSION_OLD
545 Set the inode i_generation number stor
546 version of this ioctl is an alias for
547
548 EXT4_IOC_GROUP_EXTEND
549 This ioctl has the same purpose as the
550 to resize filesystem to the end of the
551 further resize has to be done with res
552 offline. The argument points to the un
553 the filesystem new block count.
554
555 EXT4_IOC_MOVE_EXT
556 Move the block extents from orig_fd (t
557 to the donor_fd (the one specified in
558 an argument to this ioctl). Then, exch
559 orig_fd and donor_fd. This is especia
560 defragmentation, because the allocator
561 moved blocks better, ideally into one
562
563 EXT4_IOC_GROUP_ADD
564 Add a new group descriptor to an exist
565 block. The new group descriptor is des
566 structure, which is passed as an argum
567 especially useful in conjunction with
568 allows online resize of the filesystem
569 block group. Those two ioctls combine
570 resize tool (e.g. resize2fs).
571
572 EXT4_IOC_MIGRATE
573 This ioctl operates on the filesystem
574 ext3 indirect block mapped inode to ex
575 through indirect block mapping of the
576 contiguous block ranges into ext4 exte
577 inodes are swapped. This ioctl might h
578 ext4 filesystem, however suggestion is
579 and copy data from the backup. Note, t
580 extents for this ioctl to work.
581
582 EXT4_IOC_ALLOC_DA_BLKS
583 Force all of the delay allocated block
584 application-expected ext3 behaviour. N
585 triggering a write of the data blocks,
586 the future as it is not necessary and
587 sake of simplicity.
588
589 EXT4_IOC_RESIZE_FS
590 Resize the filesystem to a new size.
591 filesystem is passed in via 64 bit int
592 allocates bitmaps and inode table, the
593 the new number of blocks.
594
595 EXT4_IOC_SWAP_BOOT
596 Swap i_blocks and associated attribute
597 i_flags, ...) from the specified inode
598 (#5). This is typically used to store
599 the filesystem, where it can't be chan
600 The data blocks of the previous boot l
601 given inode.
602
603 References
604 ==========
605
606 kernel source: <file:fs/ext4/>
607 <file:fs/jbd2/>
608
609 programs: http://e2fsprogs.sourceforge.n
610
611 useful links: https://fedoraproject.org/wiki
612 http://www.bullopensource.org/
613 http://ext4.wiki.kernel.org/in
614 https://fedoraproject.org/wiki
Linux® is a registered trademark of Linus Torvalds in the United States and other countries.
TOMOYO® is a registered trademark of NTT DATA CORPORATION.