• Home
  • Line#
  • Scopes#
  • Navigate#
  • Raw
  • Download
1
2Ext4 Filesystem
3===============
4
5Ext4 is an an advanced level of the ext3 filesystem which incorporates
6scalability and reliability enhancements for supporting large filesystems
7(64 bit) in keeping with increasing disk capacities and state-of-the-art
8feature requirements.
9
10Mailing list:	linux-ext4@vger.kernel.org
11Web site:	http://ext4.wiki.kernel.org
12
13
141. Quick usage instructions:
15===========================
16
17Note: More extensive information for getting started with ext4 can be
18      found at the ext4 wiki site at the URL:
19      http://ext4.wiki.kernel.org/index.php/Ext4_Howto
20
21  - Compile and install the latest version of e2fsprogs (as of this
22    writing version 1.41.3) from:
23
24    http://sourceforge.net/project/showfiles.php?group_id=2406
25
26	or
27
28    ftp://ftp.kernel.org/pub/linux/kernel/people/tytso/e2fsprogs/
29
30	or grab the latest git repository from:
31
32    git://git.kernel.org/pub/scm/fs/ext2/e2fsprogs.git
33
34  - Note that it is highly important to install the mke2fs.conf file
35    that comes with the e2fsprogs 1.41.x sources in /etc/mke2fs.conf. If
36    you have edited the /etc/mke2fs.conf file installed on your system,
37    you will need to merge your changes with the version from e2fsprogs
38    1.41.x.
39
40  - Create a new filesystem using the ext4 filesystem type:
41
42    	# mke2fs -t ext4 /dev/hda1
43
44    Or to configure an existing ext3 filesystem to support extents:
45
46	# tune2fs -O extents /dev/hda1
47
48    If the filesystem was created with 128 byte inodes, it can be
49    converted to use 256 byte for greater efficiency via:
50
51        # tune2fs -I 256 /dev/hda1
52
53    (Note: we currently do not have tools to convert an ext4
54    filesystem back to ext3; so please do not do try this on production
55    filesystems.)
56
57  - Mounting:
58
59	# mount -t ext4 /dev/hda1 /wherever
60
61  - When comparing performance with other filesystems, it's always
62    important to try multiple workloads; very often a subtle change in a
63    workload parameter can completely change the ranking of which
64    filesystems do well compared to others.  When comparing versus ext3,
65    note that ext4 enables write barriers by default, while ext3 does
66    not enable write barriers by default.  So it is useful to use
67    explicitly specify whether barriers are enabled or not when via the
68    '-o barriers=[0|1]' mount option for both ext3 and ext4 filesystems
69    for a fair comparison.  When tuning ext3 for best benchmark numbers,
70    it is often worthwhile to try changing the data journaling mode; '-o
71    data=writeback' can be faster for some workloads.  (Note however that
72    running mounted with data=writeback can potentially leave stale data
73    exposed in recently written files in case of an unclean shutdown,
74    which could be a security exposure in some situations.)  Configuring
75    the filesystem with a large journal can also be helpful for
76    metadata-intensive workloads.
77
782. Features
79===========
80
812.1 Currently available
82
83* ability to use filesystems > 16TB (e2fsprogs support not available yet)
84* extent format reduces metadata overhead (RAM, IO for access, transactions)
85* extent format more robust in face of on-disk corruption due to magics,
86* internal redundancy in tree
87* improved file allocation (multi-block alloc)
88* lift 32000 subdirectory limit imposed by i_links_count[1]
89* nsec timestamps for mtime, atime, ctime, create time
90* inode version field on disk (NFSv4, Lustre)
91* reduced e2fsck time via uninit_bg feature
92* journal checksumming for robustness, performance
93* persistent file preallocation (e.g for streaming media, databases)
94* ability to pack bitmaps and inode tables into larger virtual groups via the
95  flex_bg feature
96* large file support
97* Inode allocation using large virtual block groups via flex_bg
98* delayed allocation
99* large block (up to pagesize) support
100* efficient new ordered mode in JBD2 and ext4(avoid using buffer head to force
101  the ordering)
102
103[1] Filesystems with a block size of 1k may see a limit imposed by the
104directory hash tree having a maximum depth of two.
105
1062.2 Candidate features for future inclusion
107
108* Online defrag (patches available but not well tested)
109* reduced mke2fs time via lazy itable initialization in conjunction with
110  the uninit_bg feature (capability to do this is available in e2fsprogs
111  but a kernel thread to do lazy zeroing of unused inode table blocks
112  after filesystem is first mounted is required for safety)
113
114There are several others under discussion, whether they all make it in is
115partly a function of how much time everyone has to work on them. Features like
116metadata checksumming have been discussed and planned for a bit but no patches
117exist yet so I'm not sure they're in the near-term roadmap.
118
119The big performance win will come with mballoc, delalloc and flex_bg
120grouping of bitmaps and inode tables.  Some test results available here:
121
122 - http://www.bullopensource.org/ext4/20080818-ffsb/ffsb-write-2.6.27-rc1.html
123 - http://www.bullopensource.org/ext4/20080818-ffsb/ffsb-readwrite-2.6.27-rc1.html
124
1253. Options
126==========
127
128When mounting an ext4 filesystem, the following option are accepted:
129(*) == default
130
131ro                   	Mount filesystem read only. Note that ext4 will
132                     	replay the journal (and thus write to the
133                     	partition) even when mounted "read only". The
134                     	mount options "ro,noload" can be used to prevent
135		     	writes to the filesystem.
136
137journal_checksum	Enable checksumming of the journal transactions.
138			This will allow the recovery code in e2fsck and the
139			kernel to detect corruption in the kernel.  It is a
140			compatible change and will be ignored by older kernels.
141
142journal_async_commit	Commit block can be written to disk without waiting
143			for descriptor blocks. If enabled older kernels cannot
144			mount the device. This will enable 'journal_checksum'
145			internally.
146
147journal_dev=devnum	When the external journal device's major/minor numbers
148			have changed, this option allows the user to specify
149			the new journal location.  The journal device is
150			identified through its new major/minor numbers encoded
151			in devnum.
152
153norecovery		Don't load the journal on mounting.  Note that
154noload			if the filesystem was not unmounted cleanly,
155                     	skipping the journal replay will lead to the
156                     	filesystem containing inconsistencies that can
157                     	lead to any number of problems.
158
159data=journal		All data are committed into the journal prior to being
160			written into the main file system.  Enabling
161			this mode will disable delayed allocation and
162			O_DIRECT support.
163
164data=ordered	(*)	All data are forced directly out to the main file
165			system prior to its metadata being committed to the
166			journal.
167
168data=writeback		Data ordering is not preserved, data may be written
169			into the main file system after its metadata has been
170			committed to the journal.
171
172commit=nrsec	(*)	Ext4 can be told to sync all its data and metadata
173			every 'nrsec' seconds. The default value is 5 seconds.
174			This means that if you lose your power, you will lose
175			as much as the latest 5 seconds of work (your
176			filesystem will not be damaged though, thanks to the
177			journaling).  This default value (or any low value)
178			will hurt performance, but it's good for data-safety.
179			Setting it to 0 will have the same effect as leaving
180			it at the default (5 seconds).
181			Setting it to very large values will improve
182			performance.
183
184barrier=<0|1(*)>	This enables/disables the use of write barriers in
185barrier(*)		the jbd code.  barrier=0 disables, barrier=1 enables.
186nobarrier		This also requires an IO stack which can support
187			barriers, and if jbd gets an error on a barrier
188			write, it will disable again with a warning.
189			Write barriers enforce proper on-disk ordering
190			of journal commits, making volatile disk write caches
191			safe to use, at some performance penalty.  If
192			your disks are battery-backed in one way or another,
193			disabling barriers may safely improve performance.
194			The mount options "barrier" and "nobarrier" can
195			also be used to enable or disable barriers, for
196			consistency with other ext4 mount options.
197
198inode_readahead_blks=n	This tuning parameter controls the maximum
199			number of inode table blocks that ext4's inode
200			table readahead algorithm will pre-read into
201			the buffer cache.  The default value is 32 blocks.
202
203nouser_xattr		Disables Extended User Attributes. If you have extended
204			attribute support enabled in the kernel configuration
205			(CONFIG_EXT4_FS_XATTR), extended attribute support
206			is enabled by default on mount. See the attr(5) manual
207			page and http://acl.bestbits.at/ for more information
208			about extended attributes.
209
210noacl			This option disables POSIX Access Control List
211			support. If ACL support is enabled in the kernel
212			configuration (CONFIG_EXT4_FS_POSIX_ACL), ACL is
213			enabled by default on mount. See the acl(5) manual
214			page and http://acl.bestbits.at/ for more information
215			about acl.
216
217bsddf		(*)	Make 'df' act like BSD.
218minixdf			Make 'df' act like Minix.
219
220debug			Extra debugging information is sent to syslog.
221
222abort			Simulate the effects of calling ext4_abort() for
223			debugging purposes.  This is normally used while
224			remounting a filesystem which is already mounted.
225
226errors=remount-ro	Remount the filesystem read-only on an error.
227errors=continue		Keep going on a filesystem error.
228errors=panic		Panic and halt the machine if an error occurs.
229                        (These mount options override the errors behavior
230                        specified in the superblock, which can be configured
231                        using tune2fs)
232
233data_err=ignore(*)	Just print an error message if an error occurs
234			in a file data buffer in ordered mode.
235data_err=abort		Abort the journal if an error occurs in a file
236			data buffer in ordered mode.
237
238grpid			Give objects the same group ID as their creator.
239bsdgroups
240
241nogrpid		(*)	New objects have the group ID of their creator.
242sysvgroups
243
244resgid=n		The group ID which may use the reserved blocks.
245
246resuid=n		The user ID which may use the reserved blocks.
247
248sb=n			Use alternate superblock at this location.
249
250quota			These options are ignored by the filesystem. They
251noquota			are used only by quota tools to recognize volumes
252grpquota		where quota should be turned on. See documentation
253usrquota		in the quota-tools package for more details
254			(http://sourceforge.net/projects/linuxquota).
255
256jqfmt=<quota type>	These options tell filesystem details about quota
257usrjquota=<file>	so that quota information can be properly updated
258grpjquota=<file>	during journal replay. They replace the above
259			quota options. See documentation in the quota-tools
260			package for more details
261			(http://sourceforge.net/projects/linuxquota).
262
263stripe=n		Number of filesystem blocks that mballoc will try
264			to use for allocation size and alignment. For RAID5/6
265			systems this should be the number of data
266			disks *  RAID chunk size in file system blocks.
267
268delalloc	(*)	Defer block allocation until just before ext4
269			writes out the block(s) in question.  This
270			allows ext4 to better allocation decisions
271			more efficiently.
272nodelalloc		Disable delayed allocation.  Blocks are allocated
273			when the data is copied from userspace to the
274			page cache, either via the write(2) system call
275			or when an mmap'ed page which was previously
276			unallocated is written for the first time.
277
278max_batch_time=usec	Maximum amount of time ext4 should wait for
279			additional filesystem operations to be batch
280			together with a synchronous write operation.
281			Since a synchronous write operation is going to
282			force a commit and then a wait for the I/O
283			complete, it doesn't cost much, and can be a
284			huge throughput win, we wait for a small amount
285			of time to see if any other transactions can
286			piggyback on the synchronous write.   The
287			algorithm used is designed to automatically tune
288			for the speed of the disk, by measuring the
289			amount of time (on average) that it takes to
290			finish committing a transaction.  Call this time
291			the "commit time".  If the time that the
292			transaction has been running is less than the
293			commit time, ext4 will try sleeping for the
294			commit time to see if other operations will join
295			the transaction.   The commit time is capped by
296			the max_batch_time, which defaults to 15000us
297			(15ms).   This optimization can be turned off
298			entirely by setting max_batch_time to 0.
299
300min_batch_time=usec	This parameter sets the commit time (as
301			described above) to be at least min_batch_time.
302			It defaults to zero microseconds.  Increasing
303			this parameter may improve the throughput of
304			multi-threaded, synchronous workloads on very
305			fast disks, at the cost of increasing latency.
306
307journal_ioprio=prio	The I/O priority (from 0 to 7, where 0 is the
308			highest priority) which should be used for I/O
309			operations submitted by kjournald2 during a
310			commit operation.  This defaults to 3, which is
311			a slightly higher priority than the default I/O
312			priority.
313
314auto_da_alloc(*)	Many broken applications don't use fsync() when
315noauto_da_alloc		replacing existing files via patterns such as
316			fd = open("foo.new")/write(fd,..)/close(fd)/
317			rename("foo.new", "foo"), or worse yet,
318			fd = open("foo", O_TRUNC)/write(fd,..)/close(fd).
319			If auto_da_alloc is enabled, ext4 will detect
320			the replace-via-rename and replace-via-truncate
321			patterns and force that any delayed allocation
322			blocks are allocated such that at the next
323			journal commit, in the default data=ordered
324			mode, the data blocks of the new file are forced
325			to disk before the rename() operation is
326			committed.  This provides roughly the same level
327			of guarantees as ext3, and avoids the
328			"zero-length" problem that can happen when a
329			system crashes before the delayed allocation
330			blocks are forced to disk.
331
332noinit_itable		Do not initialize any uninitialized inode table
333			blocks in the background.  This feature may be
334			used by installation CD's so that the install
335			process can complete as quickly as possible; the
336			inode table initialization process would then be
337			deferred until the next time the  file system
338			is unmounted.
339
340init_itable=n		The lazy itable init code will wait n times the
341			number of milliseconds it took to zero out the
342			previous block group's inode table.  This
343			minimizes the impact on the system performance
344			while file system's inode table is being initialized.
345
346discard			Controls whether ext4 should issue discard/TRIM
347nodiscard(*)		commands to the underlying block device when
348			blocks are freed.  This is useful for SSD devices
349			and sparse/thinly-provisioned LUNs, but it is off
350			by default until sufficient testing has been done.
351
352nouid32			Disables 32-bit UIDs and GIDs.  This is for
353			interoperability  with  older kernels which only
354			store and expect 16-bit values.
355
356block_validity		This options allows to enables/disables the in-kernel
357noblock_validity	facility for tracking filesystem metadata blocks
358			within internal data structures. This allows multi-
359			block allocator and other routines to quickly locate
360			extents which might overlap with filesystem metadata
361			blocks. This option is intended for debugging
362			purposes and since it negatively affects the
363			performance, it is off by default.
364
365dioread_lock		Controls whether or not ext4 should use the DIO read
366dioread_nolock		locking. If the dioread_nolock option is specified
367			ext4 will allocate uninitialized extent before buffer
368			write and convert the extent to initialized after IO
369			completes. This approach allows ext4 code to avoid
370			using inode mutex, which improves scalability on high
371			speed storages. However this does not work with
372			data journaling and dioread_nolock option will be
373			ignored with kernel warning. Note that dioread_nolock
374			code path is only used for extent-based files.
375			Because of the restrictions this options comprises
376			it is off by default (e.g. dioread_lock).
377
378i_version		Enable 64-bit inode version support. This option is
379			off by default.
380
381Data Mode
382=========
383There are 3 different data modes:
384
385* writeback mode
386In data=writeback mode, ext4 does not journal data at all.  This mode provides
387a similar level of journaling as that of XFS, JFS, and ReiserFS in its default
388mode - metadata journaling.  A crash+recovery can cause incorrect data to
389appear in files which were written shortly before the crash.  This mode will
390typically provide the best ext4 performance.
391
392* ordered mode
393In data=ordered mode, ext4 only officially journals metadata, but it logically
394groups metadata information related to data changes with the data blocks into a
395single unit called a transaction.  When it's time to write the new metadata
396out to disk, the associated data blocks are written first.  In general,
397this mode performs slightly slower than writeback but significantly faster than journal mode.
398
399* journal mode
400data=journal mode provides full data and metadata journaling.  All new data is
401written to the journal first, and then to its final location.
402In the event of a crash, the journal can be replayed, bringing both data and
403metadata into a consistent state.  This mode is the slowest except when data
404needs to be read from and written to disk at the same time where it
405outperforms all others modes.  Enabling this mode will disable delayed
406allocation and O_DIRECT support.
407
408/proc entries
409=============
410
411Information about mounted ext4 file systems can be found in
412/proc/fs/ext4.  Each mounted filesystem will have a directory in
413/proc/fs/ext4 based on its device name (i.e., /proc/fs/ext4/hdc or
414/proc/fs/ext4/dm-0).   The files in each per-device directory are shown
415in table below.
416
417Files in /proc/fs/ext4/<devname>
418..............................................................................
419 File            Content
420 mb_groups       details of multiblock allocator buddy cache of free blocks
421..............................................................................
422
423/sys entries
424============
425
426Information about mounted ext4 file systems can be found in
427/sys/fs/ext4.  Each mounted filesystem will have a directory in
428/sys/fs/ext4 based on its device name (i.e., /sys/fs/ext4/hdc or
429/sys/fs/ext4/dm-0).   The files in each per-device directory are shown
430in table below.
431
432Files in /sys/fs/ext4/<devname>
433(see also Documentation/ABI/testing/sysfs-fs-ext4)
434..............................................................................
435 File                         Content
436
437 delayed_allocation_blocks    This file is read-only and shows the number of
438                              blocks that are dirty in the page cache, but
439                              which do not have their location in the
440                              filesystem allocated yet.
441
442 inode_goal                   Tuning parameter which (if non-zero) controls
443                              the goal inode used by the inode allocator in
444                              preference to all other allocation heuristics.
445                              This is intended for debugging use only, and
446                              should be 0 on production systems.
447
448 inode_readahead_blks         Tuning parameter which controls the maximum
449                              number of inode table blocks that ext4's inode
450                              table readahead algorithm will pre-read into
451                              the buffer cache
452
453 lifetime_write_kbytes        This file is read-only and shows the number of
454                              kilobytes of data that have been written to this
455                              filesystem since it was created.
456
457 max_writeback_mb_bump        The maximum number of megabytes the writeback
458                              code will try to write out before move on to
459                              another inode.
460
461 mb_group_prealloc            The multiblock allocator will round up allocation
462                              requests to a multiple of this tuning parameter if
463                              the stripe size is not set in the ext4 superblock
464
465 mb_max_to_scan               The maximum number of extents the multiblock
466                              allocator will search to find the best extent
467
468 mb_min_to_scan               The minimum number of extents the multiblock
469                              allocator will search to find the best extent
470
471 mb_order2_req                Tuning parameter which controls the minimum size
472                              for requests (as a power of 2) where the buddy
473                              cache is used
474
475 mb_stats                     Controls whether the multiblock allocator should
476                              collect statistics, which are shown during the
477                              unmount. 1 means to collect statistics, 0 means
478                              not to collect statistics
479
480 mb_stream_req                Files which have fewer blocks than this tunable
481                              parameter will have their blocks allocated out
482                              of a block group specific preallocation pool, so
483                              that small files are packed closely together.
484                              Each large file will have its blocks allocated
485                              out of its own unique preallocation pool.
486
487 session_write_kbytes         This file is read-only and shows the number of
488                              kilobytes of data that have been written to this
489                              filesystem since it was mounted.
490..............................................................................
491
492Ioctls
493======
494
495There is some Ext4 specific functionality which can be accessed by applications
496through the system call interfaces. The list of all Ext4 specific ioctls are
497shown in the table below.
498
499Table of Ext4 specific ioctls
500..............................................................................
501 Ioctl			      Description
502 EXT4_IOC_GETFLAGS	      Get additional attributes associated with inode.
503			      The ioctl argument is an integer bitfield, with
504			      bit values described in ext4.h. This ioctl is an
505			      alias for FS_IOC_GETFLAGS.
506
507 EXT4_IOC_SETFLAGS	      Set additional attributes associated with inode.
508			      The ioctl argument is an integer bitfield, with
509			      bit values described in ext4.h. This ioctl is an
510			      alias for FS_IOC_SETFLAGS.
511
512 EXT4_IOC_GETVERSION
513 EXT4_IOC_GETVERSION_OLD
514			      Get the inode i_generation number stored for
515			      each inode. The i_generation number is normally
516			      changed only when new inode is created and it is
517			      particularly useful for network filesystems. The
518			      '_OLD' version of this ioctl is an alias for
519			      FS_IOC_GETVERSION.
520
521 EXT4_IOC_SETVERSION
522 EXT4_IOC_SETVERSION_OLD
523			      Set the inode i_generation number stored for
524			      each inode. The '_OLD' version of this ioctl
525			      is an alias for FS_IOC_SETVERSION.
526
527 EXT4_IOC_GROUP_EXTEND	      This ioctl has the same purpose as the resize
528			      mount option. It allows to resize filesystem
529			      to the end of the last existing block group,
530			      further resize has to be done with resize2fs,
531			      either online, or offline. The argument points
532			      to the unsigned logn number representing the
533			      filesystem new block count.
534
535 EXT4_IOC_MOVE_EXT	      Move the block extents from orig_fd (the one
536			      this ioctl is pointing to) to the donor_fd (the
537			      one specified in move_extent structure passed
538			      as an argument to this ioctl). Then, exchange
539			      inode metadata between orig_fd and donor_fd.
540			      This is especially useful for online
541			      defragmentation, because the allocator has the
542			      opportunity to allocate moved blocks better,
543			      ideally into one contiguous extent.
544
545 EXT4_IOC_GROUP_ADD	      Add a new group descriptor to an existing or
546			      new group descriptor block. The new group
547			      descriptor is described by ext4_new_group_input
548			      structure, which is passed as an argument to
549			      this ioctl. This is especially useful in
550			      conjunction with EXT4_IOC_GROUP_EXTEND,
551			      which allows online resize of the filesystem
552			      to the end of the last existing block group.
553			      Those two ioctls combined is used in userspace
554			      online resize tool (e.g. resize2fs).
555
556 EXT4_IOC_MIGRATE	      This ioctl operates on the filesystem itself.
557			      It converts (migrates) ext3 indirect block mapped
558			      inode to ext4 extent mapped inode by walking
559			      through indirect block mapping of the original
560			      inode and converting contiguous block ranges
561			      into ext4 extents of the temporary inode. Then,
562			      inodes are swapped. This ioctl might help, when
563			      migrating from ext3 to ext4 filesystem, however
564			      suggestion is to create fresh ext4 filesystem
565			      and copy data from the backup. Note, that
566			      filesystem has to support extents for this ioctl
567			      to work.
568
569 EXT4_IOC_ALLOC_DA_BLKS	      Force all of the delay allocated blocks to be
570			      allocated to preserve application-expected ext3
571			      behaviour. Note that this will also start
572			      triggering a write of the data blocks, but this
573			      behaviour may change in the future as it is
574			      not necessary and has been done this way only
575			      for sake of simplicity.
576
577 EXT4_IOC_RESIZE_FS	      Resize the filesystem to a new size.  The number
578			      of blocks of resized filesystem is passed in via
579			      64 bit integer argument.  The kernel allocates
580			      bitmaps and inode table, the userspace tool thus
581			      just passes the new number of blocks.
582
583..............................................................................
584
585References
586==========
587
588kernel source:	<file:fs/ext4/>
589		<file:fs/jbd2/>
590
591programs:	http://e2fsprogs.sourceforge.net/
592
593useful links:	http://fedoraproject.org/wiki/ext3-devel
594		http://www.bullopensource.org/ext4/
595		http://ext4.wiki.kernel.org/index.php/Main_Page
596		http://fedoraproject.org/wiki/Features/Ext4
597