1 .. SPDX-License-Identifier: GPL-2.0
3 ========================
4 ext4 General Information
5 ========================
7 Ext4 is an advanced level of the ext3 filesystem which incorporates
8 scalability and reliability enhancements for supporting large filesystems
9 (64 bit) in keeping with increasing disk capacities and state-of-the-art
12 Mailing list: linux-ext4@vger.kernel.org
13 Web site: http://ext4.wiki.kernel.org
16 Quick usage instructions
17 ========================
19 Note: More extensive information for getting started with ext4 can be
20 found at the ext4 wiki site at the URL:
21 http://ext4.wiki.kernel.org/index.php/Ext4_Howto
23 - The latest version of e2fsprogs can be found at:
25 https://www.kernel.org/pub/linux/kernel/people/tytso/e2fsprogs/
29 http://sourceforge.net/project/showfiles.php?group_id=2406
31 or grab the latest git repository from:
33 https://git.kernel.org/pub/scm/fs/ext2/e2fsprogs.git
35 - Create a new filesystem using the ext4 filesystem type:
37 # mke2fs -t ext4 /dev/hda1
39 Or to configure an existing ext3 filesystem to support extents:
41 # tune2fs -O extents /dev/hda1
43 If the filesystem was created with 128 byte inodes, it can be
44 converted to use 256 byte for greater efficiency via:
46 # tune2fs -I 256 /dev/hda1
50 # mount -t ext4 /dev/hda1 /wherever
52 - When comparing performance with other filesystems, it's always
53 important to try multiple workloads; very often a subtle change in a
54 workload parameter can completely change the ranking of which
55 filesystems do well compared to others. When comparing versus ext3,
56 note that ext4 enables write barriers by default, while ext3 does
57 not enable write barriers by default. So it is useful to use
58 explicitly specify whether barriers are enabled or not when via the
59 '-o barriers=[0|1]' mount option for both ext3 and ext4 filesystems
60 for a fair comparison. When tuning ext3 for best benchmark numbers,
61 it is often worthwhile to try changing the data journaling mode; '-o
62 data=writeback' can be faster for some workloads. (Note however that
63 running mounted with data=writeback can potentially leave stale data
64 exposed in recently written files in case of an unclean shutdown,
65 which could be a security exposure in some situations.) Configuring
66 the filesystem with a large journal can also be helpful for
67 metadata-intensive workloads.
75 * ability to use filesystems > 16TB (e2fsprogs support not available yet)
76 * extent format reduces metadata overhead (RAM, IO for access, transactions)
77 * extent format more robust in face of on-disk corruption due to magics,
78 * internal redundancy in tree
79 * improved file allocation (multi-block alloc)
80 * lift 32000 subdirectory limit imposed by i_links_count[1]
81 * nsec timestamps for mtime, atime, ctime, create time
82 * inode version field on disk (NFSv4, Lustre)
83 * reduced e2fsck time via uninit_bg feature
84 * journal checksumming for robustness, performance
85 * persistent file preallocation (e.g for streaming media, databases)
86 * ability to pack bitmaps and inode tables into larger virtual groups via the
89 * inode allocation using large virtual block groups via flex_bg
91 * large block (up to pagesize) support
92 * efficient new ordered mode in JBD2 and ext4 (avoid using buffer head to force
94 * Case-insensitive file name lookups
96 [1] Filesystems with a block size of 1k may see a limit imposed by the
97 directory hash tree having a maximum depth of two.
99 case-insensitive file name lookups
100 ======================================================
102 The case-insensitive file name lookup feature is supported on a
103 per-directory basis, allowing the user to mix case-insensitive and
104 case-sensitive directories in the same filesystem. It is enabled by
105 flipping the +F inode attribute of an empty directory. The
106 case-insensitive string match operation is only defined when we know how
107 text in encoded in a byte sequence. For that reason, in order to enable
108 case-insensitive directories, the filesystem must have the
109 casefold feature, which stores the filesystem-wide encoding
110 model used. By default, the charset adopted is the latest version of
111 Unicode (12.1.0, by the time of this writing), encoded in the UTF-8
112 form. The comparison algorithm is implemented by normalizing the
113 strings to the Canonical decomposition form, as defined by Unicode,
114 followed by a byte per byte comparison.
116 The case-awareness is name-preserving on the disk, meaning that the file
117 name provided by userspace is a byte-per-byte match to what is actually
118 written in the disk. The Unicode normalization format used by the
119 kernel is thus an internal representation, and not exposed to the
120 userspace nor to the disk, with the important exception of disk hashes,
121 used on large case-insensitive directories with DX feature. On DX
122 directories, the hash must be calculated using the casefolded version of
123 the filename, meaning that the normalization format used actually has an
124 impact on where the directory entry is stored.
126 When we change from viewing filenames as opaque byte sequences to seeing
127 them as encoded strings we need to address what happens when a program
128 tries to create a file with an invalid name. The Unicode subsystem
129 within the kernel leaves the decision of what to do in this case to the
130 filesystem, which select its preferred behavior by enabling/disabling
131 the strict mode. When Ext4 encounters one of those strings and the
132 filesystem did not require strict mode, it falls back to considering the
133 entire string as an opaque byte sequence, which still allows the user to
134 operate on that file, but the case-insensitive lookups won't work.
139 When mounting an ext4 filesystem, the following option are accepted:
143 Mount filesystem read only. Note that ext4 will replay the journal (and
144 thus write to the partition) even when mounted "read only". The mount
145 options "ro,noload" can be used to prevent writes to the filesystem.
148 Enable checksumming of the journal transactions. This will allow the
149 recovery code in e2fsck and the kernel to detect corruption in the
150 kernel. It is a compatible change and will be ignored by older
154 Commit block can be written to disk without waiting for descriptor
155 blocks. If enabled older kernels cannot mount the device. This will
156 enable 'journal_checksum' internally.
158 journal_path=path, journal_dev=devnum
159 When the external journal device's major/minor numbers have changed,
160 these options allow the user to specify the new journal location. The
161 journal device is identified through either its new major/minor numbers
162 encoded in devnum, or via a path to the device.
165 Don't load the journal on mounting. Note that if the filesystem was
166 not unmounted cleanly, skipping the journal replay will lead to the
167 filesystem containing inconsistencies that can lead to any number of
171 All data are committed into the journal prior to being written into the
172 main file system. Enabling this mode will disable delayed allocation
173 and O_DIRECT support.
176 All data are forced directly out to the main file system prior to its
177 metadata being committed to the journal.
180 Data ordering is not preserved, data may be written into the main file
181 system after its metadata has been committed to the journal.
184 This setting limits the maximum age of the running transaction to
185 'nrsec' seconds. The default value is 5 seconds. This means that if
186 you lose your power, you will lose as much as the latest 5 seconds of
187 metadata changes (your filesystem will not be damaged though, thanks
188 to the journaling). This default value (or any low value) will hurt
189 performance, but it's good for data-safety. Setting it to 0 will have
190 the same effect as leaving it at the default (5 seconds). Setting it
191 to very large values will improve performance. Note that due to
192 delayed allocation even older data can be lost on power failure since
193 writeback of those data begins only after time set in
194 /proc/sys/vm/dirty_expire_centisecs.
196 barrier=<0|1(*)>, barrier(*), nobarrier
197 This enables/disables the use of write barriers in the jbd code.
198 barrier=0 disables, barrier=1 enables. This also requires an IO stack
199 which can support barriers, and if jbd gets an error on a barrier
200 write, it will disable again with a warning. Write barriers enforce
201 proper on-disk ordering of journal commits, making volatile disk write
202 caches safe to use, at some performance penalty. If your disks are
203 battery-backed in one way or another, disabling barriers may safely
204 improve performance. The mount options "barrier" and "nobarrier" can
205 also be used to enable or disable barriers, for consistency with other
208 inode_readahead_blks=n
209 This tuning parameter controls the maximum number of inode table blocks
210 that ext4's inode table readahead algorithm will pre-read into the
211 buffer cache. The default value is 32 blocks.
214 Disables Extended User Attributes. See the attr(5) manual page for
215 more information about extended attributes.
218 This option disables POSIX Access Control List support. If ACL support
219 is enabled in the kernel configuration (CONFIG_EXT4_FS_POSIX_ACL), ACL
220 is enabled by default on mount. See the acl(5) manual page for more
221 information about acl.
224 Make 'df' act like BSD.
227 Make 'df' act like Minix.
230 Extra debugging information is sent to syslog.
233 Simulate the effects of calling ext4_abort() for debugging purposes.
234 This is normally used while remounting a filesystem which is already
238 Remount the filesystem read-only on an error.
241 Keep going on a filesystem error.
244 Panic and halt the machine if an error occurs. (These mount options
245 override the errors behavior specified in the superblock, which can be
246 configured using tune2fs)
249 Just print an error message if an error occurs in a file data buffer in
252 Abort the journal if an error occurs in a file data buffer in ordered
256 New objects have the group ID of their parent.
258 nogrpid (*) | sysvgroups
259 New objects have the group ID of their creator.
262 The group ID which may use the reserved blocks.
265 The user ID which may use the reserved blocks.
268 Use alternate superblock at this location.
270 quota, noquota, grpquota, usrquota
271 These options are ignored by the filesystem. They are used only by
272 quota tools to recognize volumes where quota should be turned on. See
273 documentation in the quota-tools package for more details
274 (http://sourceforge.net/projects/linuxquota).
276 jqfmt=<quota type>, usrjquota=<file>, grpjquota=<file>
277 These options tell filesystem details about quota so that quota
278 information can be properly updated during journal replay. They replace
279 the above quota options. See documentation in the quota-tools package
280 for more details (http://sourceforge.net/projects/linuxquota).
283 Number of filesystem blocks that mballoc will try to use for allocation
284 size and alignment. For RAID5/6 systems this should be the number of
285 data disks * RAID chunk size in file system blocks.
288 Defer block allocation until just before ext4 writes out the block(s)
289 in question. This allows ext4 to better allocation decisions more
293 Disable delayed allocation. Blocks are allocated when the data is
294 copied from userspace to the page cache, either via the write(2) system
295 call or when an mmap'ed page which was previously unallocated is
296 written for the first time.
299 Maximum amount of time ext4 should wait for additional filesystem
300 operations to be batch together with a synchronous write operation.
301 Since a synchronous write operation is going to force a commit and then
302 a wait for the I/O complete, it doesn't cost much, and can be a huge
303 throughput win, we wait for a small amount of time to see if any other
304 transactions can piggyback on the synchronous write. The algorithm
305 used is designed to automatically tune for the speed of the disk, by
306 measuring the amount of time (on average) that it takes to finish
307 committing a transaction. Call this time the "commit time". If the
308 time that the transaction has been running is less than the commit
309 time, ext4 will try sleeping for the commit time to see if other
310 operations will join the transaction. The commit time is capped by
311 the max_batch_time, which defaults to 15000us (15ms). This
312 optimization can be turned off entirely by setting max_batch_time to 0.
315 This parameter sets the commit time (as described above) to be at least
316 min_batch_time. It defaults to zero microseconds. Increasing this
317 parameter may improve the throughput of multi-threaded, synchronous
318 workloads on very fast disks, at the cost of increasing latency.
321 The I/O priority (from 0 to 7, where 0 is the highest priority) which
322 should be used for I/O operations submitted by kjournald2 during a
323 commit operation. This defaults to 3, which is a slightly higher
324 priority than the default I/O priority.
326 auto_da_alloc(*), noauto_da_alloc
327 Many broken applications don't use fsync() when replacing existing
328 files via patterns such as fd = open("foo.new")/write(fd,..)/close(fd)/
329 rename("foo.new", "foo"), or worse yet, fd = open("foo",
330 O_TRUNC)/write(fd,..)/close(fd). If auto_da_alloc is enabled, ext4
331 will detect the replace-via-rename and replace-via-truncate patterns
332 and force that any delayed allocation blocks are allocated such that at
333 the next journal commit, in the default data=ordered mode, the data
334 blocks of the new file are forced to disk before the rename() operation
335 is committed. This provides roughly the same level of guarantees as
336 ext3, and avoids the "zero-length" problem that can happen when a
337 system crashes before the delayed allocation blocks are forced to disk.
340 Do not initialize any uninitialized inode table blocks in the
341 background. This feature may be used by installation CD's so that the
342 install process can complete as quickly as possible; the inode table
343 initialization process would then be deferred until the next time the
344 file system is unmounted.
347 The lazy itable init code will wait n times the number of milliseconds
348 it took to zero out the previous block group's inode table. This
349 minimizes the impact on the system performance while file system's
350 inode table is being initialized.
352 discard, nodiscard(*)
353 Controls whether ext4 should issue discard/TRIM commands to the
354 underlying block device when blocks are freed. This is useful for SSD
355 devices and sparse/thinly-provisioned LUNs, but it is off by default
356 until sufficient testing has been done.
359 Disables 32-bit UIDs and GIDs. This is for interoperability with
360 older kernels which only store and expect 16-bit values.
362 block_validity(*), noblock_validity
363 These options enable or disable the in-kernel facility for tracking
364 filesystem metadata blocks within internal data structures. This
365 allows multi- block allocator and other routines to notice bugs or
366 corrupted allocation bitmaps which cause blocks to be allocated which
367 overlap with filesystem metadata blocks.
369 dioread_lock, dioread_nolock
370 Controls whether or not ext4 should use the DIO read locking. If the
371 dioread_nolock option is specified ext4 will allocate uninitialized
372 extent before buffer write and convert the extent to initialized after
373 IO completes. This approach allows ext4 code to avoid using inode
374 mutex, which improves scalability on high speed storages. However this
375 does not work with data journaling and dioread_nolock option will be
376 ignored with kernel warning. Note that dioread_nolock code path is only
377 used for extent-based files. Because of the restrictions this options
378 comprises it is off by default (e.g. dioread_lock).
381 This limits the size of directories so that any attempt to expand them
382 beyond the specified limit in kilobytes will cause an ENOSPC error.
383 This is useful in memory constrained environments, where a very large
384 directory can cause severe performance problems or even provoke the Out
385 Of Memory killer. (For example, if there is only 512mb memory
386 available, a 176mb directory may seriously cramp the system's style.)
389 Enable 64-bit inode version support. This option is off by default.
392 Use direct access (no page cache). See
393 Documentation/filesystems/dax.txt. Note that this option is
394 incompatible with data=journal.
398 There are 3 different data modes:
402 In data=writeback mode, ext4 does not journal data at all. This mode provides
403 a similar level of journaling as that of XFS, JFS, and ReiserFS in its default
404 mode - metadata journaling. A crash+recovery can cause incorrect data to
405 appear in files which were written shortly before the crash. This mode will
406 typically provide the best ext4 performance.
410 In data=ordered mode, ext4 only officially journals metadata, but it logically
411 groups metadata information related to data changes with the data blocks into
412 a single unit called a transaction. When it's time to write the new metadata
413 out to disk, the associated data blocks are written first. In general, this
414 mode performs slightly slower than writeback but significantly faster than
419 data=journal mode provides full data and metadata journaling. All new data is
420 written to the journal first, and then to its final location. In the event of
421 a crash, the journal can be replayed, bringing both data and metadata into a
422 consistent state. This mode is the slowest except when data needs to be read
423 from and written to disk at the same time where it outperforms all others
424 modes. Enabling this mode will disable delayed allocation and O_DIRECT
430 Information about mounted ext4 file systems can be found in
431 /proc/fs/ext4. Each mounted filesystem will have a directory in
432 /proc/fs/ext4 based on its device name (i.e., /proc/fs/ext4/hdc or
433 /proc/fs/ext4/dm-0). The files in each per-device directory are shown
436 Files in /proc/fs/ext4/<devname>
439 details of multiblock allocator buddy cache of free blocks
444 Information about mounted ext4 file systems can be found in
445 /sys/fs/ext4. Each mounted filesystem will have a directory in
446 /sys/fs/ext4 based on its device name (i.e., /sys/fs/ext4/hdc or
447 /sys/fs/ext4/dm-0). The files in each per-device directory are shown
450 Files in /sys/fs/ext4/<devname>:
452 (see also Documentation/ABI/testing/sysfs-fs-ext4)
454 delayed_allocation_blocks
455 This file is read-only and shows the number of blocks that are dirty in
456 the page cache, but which do not have their location in the filesystem
460 Tuning parameter which (if non-zero) controls the goal inode used by
461 the inode allocator in preference to all other allocation heuristics.
462 This is intended for debugging use only, and should be 0 on production
466 Tuning parameter which controls the maximum number of inode table
467 blocks that ext4's inode table readahead algorithm will pre-read into
470 lifetime_write_kbytes
471 This file is read-only and shows the number of kilobytes of data that
472 have been written to this filesystem since it was created.
474 max_writeback_mb_bump
475 The maximum number of megabytes the writeback code will try to write
476 out before move on to another inode.
479 The multiblock allocator will round up allocation requests to a
480 multiple of this tuning parameter if the stripe size is not set in the
484 The maximum number of extents the multiblock allocator will search to
485 find the best extent.
488 The minimum number of extents the multiblock allocator will search to
489 find the best extent.
492 Tuning parameter which controls the minimum size for requests (as a
493 power of 2) where the buddy cache is used.
496 Controls whether the multiblock allocator should collect statistics,
497 which are shown during the unmount. 1 means to collect statistics, 0
498 means not to collect statistics.
501 Files which have fewer blocks than this tunable parameter will have
502 their blocks allocated out of a block group specific preallocation
503 pool, so that small files are packed closely together. Each large file
504 will have its blocks allocated out of its own unique preallocation
508 This file is read-only and shows the number of kilobytes of data that
509 have been written to this filesystem since it was mounted.
512 This is RW file and contains number of reserved clusters in the file
513 system which will be used in the specific situations to avoid costly
514 zeroout, unexpected ENOSPC, or possible data loss. The default is 2% or
515 4096 clusters, whichever is smaller and this can be changed however it
516 can never exceed number of clusters in the file system. If there is not
517 enough space for the reserved space when mounting the file mount will
523 There is some Ext4 specific functionality which can be accessed by applications
524 through the system call interfaces. The list of all Ext4 specific ioctls are
525 shown in the table below.
527 Table of Ext4 specific ioctls
530 Get additional attributes associated with inode. The ioctl argument is
531 an integer bitfield, with bit values described in ext4.h. This ioctl is
532 an alias for FS_IOC_GETFLAGS.
535 Set additional attributes associated with inode. The ioctl argument is
536 an integer bitfield, with bit values described in ext4.h. This ioctl is
537 an alias for FS_IOC_SETFLAGS.
539 EXT4_IOC_GETVERSION, EXT4_IOC_GETVERSION_OLD
540 Get the inode i_generation number stored for each inode. The
541 i_generation number is normally changed only when new inode is created
542 and it is particularly useful for network filesystems. The '_OLD'
543 version of this ioctl is an alias for FS_IOC_GETVERSION.
545 EXT4_IOC_SETVERSION, EXT4_IOC_SETVERSION_OLD
546 Set the inode i_generation number stored for each inode. The '_OLD'
547 version of this ioctl is an alias for FS_IOC_SETVERSION.
549 EXT4_IOC_GROUP_EXTEND
550 This ioctl has the same purpose as the resize mount option. It allows
551 to resize filesystem to the end of the last existing block group,
552 further resize has to be done with resize2fs, either online, or
553 offline. The argument points to the unsigned logn number representing
554 the filesystem new block count.
557 Move the block extents from orig_fd (the one this ioctl is pointing to)
558 to the donor_fd (the one specified in move_extent structure passed as
559 an argument to this ioctl). Then, exchange inode metadata between
560 orig_fd and donor_fd. This is especially useful for online
561 defragmentation, because the allocator has the opportunity to allocate
562 moved blocks better, ideally into one contiguous extent.
565 Add a new group descriptor to an existing or new group descriptor
566 block. The new group descriptor is described by ext4_new_group_input
567 structure, which is passed as an argument to this ioctl. This is
568 especially useful in conjunction with EXT4_IOC_GROUP_EXTEND, which
569 allows online resize of the filesystem to the end of the last existing
570 block group. Those two ioctls combined is used in userspace online
571 resize tool (e.g. resize2fs).
574 This ioctl operates on the filesystem itself. It converts (migrates)
575 ext3 indirect block mapped inode to ext4 extent mapped inode by walking
576 through indirect block mapping of the original inode and converting
577 contiguous block ranges into ext4 extents of the temporary inode. Then,
578 inodes are swapped. This ioctl might help, when migrating from ext3 to
579 ext4 filesystem, however suggestion is to create fresh ext4 filesystem
580 and copy data from the backup. Note, that filesystem has to support
581 extents for this ioctl to work.
583 EXT4_IOC_ALLOC_DA_BLKS
584 Force all of the delay allocated blocks to be allocated to preserve
585 application-expected ext3 behaviour. Note that this will also start
586 triggering a write of the data blocks, but this behaviour may change in
587 the future as it is not necessary and has been done this way only for
591 Resize the filesystem to a new size. The number of blocks of resized
592 filesystem is passed in via 64 bit integer argument. The kernel
593 allocates bitmaps and inode table, the userspace tool thus just passes
594 the new number of blocks.
597 Swap i_blocks and associated attributes (like i_blocks, i_size,
598 i_flags, ...) from the specified inode with inode EXT4_BOOT_LOADER_INO
599 (#5). This is typically used to store a boot loader in a secure part of
600 the filesystem, where it can't be changed by a normal user by accident.
601 The data blocks of the previous boot loader will be associated with the
607 kernel source: <file:fs/ext4/>
610 programs: http://e2fsprogs.sourceforge.net/
612 useful links: http://fedoraproject.org/wiki/ext3-devel
613 http://www.bullopensource.org/ext4/
614 http://ext4.wiki.kernel.org/index.php/Main_Page
615 http://fedoraproject.org/wiki/Features/Ext4