| blob | 3698ed3146e3011b11ed4211a06a7badc2f254cb |
2 Ext4 Filesystem
3 ===============
5 Ext4 is an advanced level of the ext3 filesystem which incorporates
6 scalability and reliability enhancements for supporting large filesystems
7 (64 bit) in keeping with increasing disk capacities and state-of-the-art
8 feature requirements.
10 Mailing list: linux-ext4@vger.kernel.org
11 Web site: http://ext4.wiki.kernel.org
14 1. Quick usage instructions:
15 ===========================
17 Note: 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
21 - Compile and install the latest version of e2fsprogs (as of this
22 writing version 1.41.3) from:
24 http://sourceforge.net/project/showfiles.php?group_id=2406
26 or
28 ftp://ftp.kernel.org/pub/linux/kernel/people/tytso/e2fsprogs/
30 or grab the latest git repository from:
32 git://git.kernel.org/pub/scm/fs/ext2/e2fsprogs.git
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.
40 - Create a new filesystem using the ext4 filesystem type:
42 # mke2fs -t ext4 /dev/hda1
44 Or to configure an existing ext3 filesystem to support extents:
46 # tune2fs -O extents /dev/hda1
48 If the filesystem was created with 128 byte inodes, it can be
49 converted to use 256 byte for greater efficiency via:
51 # tune2fs -I 256 /dev/hda1
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.)
57 - Mounting:
59 # mount -t ext4 /dev/hda1 /wherever
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.
78 2. Features
79 ===========
81 2.1 Currently available
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)
103 [1] Filesystems with a block size of 1k may see a limit imposed by the
104 directory hash tree having a maximum depth of two.
106 2.2 Candidate features for future inclusion
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)
114 There are several others under discussion, whether they all make it in is
115 partly a function of how much time everyone has to work on them. Features like
116 metadata checksumming have been discussed and planned for a bit but no patches
117 exist yet so I'm not sure they're in the near-term roadmap.
119 The big performance win will come with mballoc, delalloc and flex_bg
120 grouping of bitmaps and inode tables. Some test results available here:
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
125 3. Options
126 ==========
128 When mounting an ext4 filesystem, the following option are accepted:
129 (*) == default
131 ro 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.
137 journal_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.
142 journal_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.
147 journal_path=path
148 journal_dev=devnum When the external journal device's major/minor numbers
149 have changed, these options allow the user to specify
150 the new journal location. The journal device is
151 identified through either its new major/minor numbers
152 encoded in devnum, or via a path to the device.
154 norecovery Don't load the journal on mounting. Note that
155 noload if the filesystem was not unmounted cleanly,
156 skipping the journal replay will lead to the
157 filesystem containing inconsistencies that can
158 lead to any number of problems.
160 data=journal All data are committed into the journal prior to being
161 written into the main file system. Enabling
162 this mode will disable delayed allocation and
163 O_DIRECT support.
165 data=ordered (*) All data are forced directly out to the main file
166 system prior to its metadata being committed to the
167 journal.
169 data=writeback Data ordering is not preserved, data may be written
170 into the main file system after its metadata has been
171 committed to the journal.
173 commit=nrsec (*) Ext4 can be told to sync all its data and metadata
174 every 'nrsec' seconds. The default value is 5 seconds.
175 This means that if you lose your power, you will lose
176 as much as the latest 5 seconds of work (your
177 filesystem will not be damaged though, thanks to the
178 journaling). This default value (or any low value)
179 will hurt performance, but it's good for data-safety.
180 Setting it to 0 will have the same effect as leaving
181 it at the default (5 seconds).
182 Setting it to very large values will improve
183 performance.
185 barrier=<0|1(*)> This enables/disables the use of write barriers in
186 barrier(*) the jbd code. barrier=0 disables, barrier=1 enables.
187 nobarrier This also requires an IO stack which can support
188 barriers, and if jbd gets an error on a barrier
189 write, it will disable again with a warning.
190 Write barriers enforce proper on-disk ordering
191 of journal commits, making volatile disk write caches
192 safe to use, at some performance penalty. If
193 your disks are battery-backed in one way or another,
194 disabling barriers may safely improve performance.
195 The mount options "barrier" and "nobarrier" can
196 also be used to enable or disable barriers, for
197 consistency with other ext4 mount options.
199 inode_readahead_blks=n This tuning parameter controls the maximum
200 number of inode table blocks that ext4's inode
201 table readahead algorithm will pre-read into
202 the buffer cache. The default value is 32 blocks.
204 nouser_xattr Disables Extended User Attributes. See the
205 attr(5) manual page and http://acl.bestbits.at/
206 for more information about extended attributes.
208 noacl This option disables POSIX Access Control List
209 support. If ACL support is enabled in the kernel
210 configuration (CONFIG_EXT4_FS_POSIX_ACL), ACL is
211 enabled by default on mount. See the acl(5) manual
212 page and http://acl.bestbits.at/ for more information
213 about acl.
215 bsddf (*) Make 'df' act like BSD.
216 minixdf Make 'df' act like Minix.
218 debug Extra debugging information is sent to syslog.
220 abort Simulate the effects of calling ext4_abort() for
221 debugging purposes. This is normally used while
222 remounting a filesystem which is already mounted.
224 errors=remount-ro Remount the filesystem read-only on an error.
225 errors=continue Keep going on a filesystem error.
226 errors=panic Panic and halt the machine if an error occurs.
227 (These mount options override the errors behavior
228 specified in the superblock, which can be configured
229 using tune2fs)
231 data_err=ignore(*) Just print an error message if an error occurs
232 in a file data buffer in ordered mode.
233 data_err=abort Abort the journal if an error occurs in a file
234 data buffer in ordered mode.
236 grpid Give objects the same group ID as their creator.
237 bsdgroups
239 nogrpid (*) New objects have the group ID of their creator.
240 sysvgroups
242 resgid=n The group ID which may use the reserved blocks.
244 resuid=n The user ID which may use the reserved blocks.
246 sb=n Use alternate superblock at this location.
248 quota These options are ignored by the filesystem. They
249 noquota are used only by quota tools to recognize volumes
250 grpquota where quota should be turned on. See documentation
251 usrquota in the quota-tools package for more details
252 (http://sourceforge.net/projects/linuxquota).
254 jqfmt=<quota type> These options tell filesystem details about quota
255 usrjquota=<file> so that quota information can be properly updated
256 grpjquota=<file> during journal replay. They replace the above
257 quota options. See documentation in the quota-tools
258 package for more details
259 (http://sourceforge.net/projects/linuxquota).
261 stripe=n Number of filesystem blocks that mballoc will try
262 to use for allocation size and alignment. For RAID5/6
263 systems this should be the number of data
264 disks * RAID chunk size in file system blocks.
266 delalloc (*) Defer block allocation until just before ext4
267 writes out the block(s) in question. This
268 allows ext4 to better allocation decisions
269 more efficiently.
270 nodelalloc Disable delayed allocation. Blocks are allocated
271 when the data is copied from userspace to the
272 page cache, either via the write(2) system call
273 or when an mmap'ed page which was previously
274 unallocated is written for the first time.
276 max_batch_time=usec Maximum amount of time ext4 should wait for
277 additional filesystem operations to be batch
278 together with a synchronous write operation.
279 Since a synchronous write operation is going to
280 force a commit and then a wait for the I/O
281 complete, it doesn't cost much, and can be a
282 huge throughput win, we wait for a small amount
283 of time to see if any other transactions can
284 piggyback on the synchronous write. The
285 algorithm used is designed to automatically tune
286 for the speed of the disk, by measuring the
287 amount of time (on average) that it takes to
288 finish committing a transaction. Call this time
289 the "commit time". If the time that the
290 transaction has been running is less than the
291 commit time, ext4 will try sleeping for the
292 commit time to see if other operations will join
293 the transaction. The commit time is capped by
294 the max_batch_time, which defaults to 15000us
295 (15ms). This optimization can be turned off
296 entirely by setting max_batch_time to 0.
298 min_batch_time=usec This parameter sets the commit time (as
299 described above) to be at least min_batch_time.
300 It defaults to zero microseconds. Increasing
301 this parameter may improve the throughput of
302 multi-threaded, synchronous workloads on very
303 fast disks, at the cost of increasing latency.
305 journal_ioprio=prio The I/O priority (from 0 to 7, where 0 is the
306 highest priority) which should be used for I/O
307 operations submitted by kjournald2 during a
308 commit operation. This defaults to 3, which is
309 a slightly higher priority than the default I/O
310 priority.
312 auto_da_alloc(*) Many broken applications don't use fsync() when
313 noauto_da_alloc replacing existing files via patterns such as
314 fd = open("foo.new")/write(fd,..)/close(fd)/
315 rename("foo.new", "foo"), or worse yet,
316 fd = open("foo", O_TRUNC)/write(fd,..)/close(fd).
317 If auto_da_alloc is enabled, ext4 will detect
318 the replace-via-rename and replace-via-truncate
319 patterns and force that any delayed allocation
320 blocks are allocated such that at the next
321 journal commit, in the default data=ordered
322 mode, the data blocks of the new file are forced
323 to disk before the rename() operation is
324 committed. This provides roughly the same level
325 of guarantees as ext3, and avoids the
326 "zero-length" problem that can happen when a
327 system crashes before the delayed allocation
328 blocks are forced to disk.
330 noinit_itable Do not initialize any uninitialized inode table
331 blocks in the background. This feature may be
332 used by installation CD's so that the install
333 process can complete as quickly as possible; the
334 inode table initialization process would then be
335 deferred until the next time the file system
336 is unmounted.
338 init_itable=n The lazy itable init code will wait n times the
339 number of milliseconds it took to zero out the
340 previous block group's inode table. This
341 minimizes the impact on the system performance
342 while file system's inode table is being initialized.
344 discard Controls whether ext4 should issue discard/TRIM
345 nodiscard(*) commands to the underlying block device when
346 blocks are freed. This is useful for SSD devices
347 and sparse/thinly-provisioned LUNs, but it is off
348 by default until sufficient testing has been done.
350 nouid32 Disables 32-bit UIDs and GIDs. This is for
351 interoperability with older kernels which only
352 store and expect 16-bit values.
354 block_validity(*) These options enable or disable the in-kernel
355 noblock_validity facility for tracking filesystem metadata blocks
356 within internal data structures. This allows multi-
357 block allocator and other routines to notice
358 bugs or corrupted allocation bitmaps which cause
359 blocks to be allocated which overlap with
360 filesystem metadata blocks.
362 dioread_lock Controls whether or not ext4 should use the DIO read
363 dioread_nolock locking. If the dioread_nolock option is specified
364 ext4 will allocate uninitialized extent before buffer
365 write and convert the extent to initialized after IO
366 completes. This approach allows ext4 code to avoid
367 using inode mutex, which improves scalability on high
368 speed storages. However this does not work with
369 data journaling and dioread_nolock option will be
370 ignored with kernel warning. Note that dioread_nolock
371 code path is only used for extent-based files.
372 Because of the restrictions this options comprises
373 it is off by default (e.g. dioread_lock).
375 max_dir_size_kb=n This limits the size of directories so that any
376 attempt to expand them beyond the specified
377 limit in kilobytes will cause an ENOSPC error.
378 This is useful in memory constrained
379 environments, where a very large directory can
380 cause severe performance problems or even
381 provoke the Out Of Memory killer. (For example,
382 if there is only 512mb memory available, a 176mb
383 directory may seriously cramp the system's style.)
385 i_version Enable 64-bit inode version support. This option is
386 off by default.
388 dax Use direct access (no page cache). See
389 Documentation/filesystems/dax.txt. Note that
390 this option is incompatible with data=journal.
392 Data Mode
393 =========
394 There are 3 different data modes:
396 * writeback mode
397 In data=writeback mode, ext4 does not journal data at all. This mode provides
398 a similar level of journaling as that of XFS, JFS, and ReiserFS in its default
399 mode - metadata journaling. A crash+recovery can cause incorrect data to
400 appear in files which were written shortly before the crash. This mode will
401 typically provide the best ext4 performance.
403 * ordered mode
404 In data=ordered mode, ext4 only officially journals metadata, but it logically
405 groups metadata information related to data changes with the data blocks into a
406 single unit called a transaction. When it's time to write the new metadata
407 out to disk, the associated data blocks are written first. In general,
408 this mode performs slightly slower than writeback but significantly faster than journal mode.
410 * journal mode
411 data=journal mode provides full data and metadata journaling. All new data is
412 written to the journal first, and then to its final location.
413 In the event of a crash, the journal can be replayed, bringing both data and
414 metadata into a consistent state. This mode is the slowest except when data
415 needs to be read from and written to disk at the same time where it
416 outperforms all others modes. Enabling this mode will disable delayed
417 allocation and O_DIRECT support.
419 /proc entries
420 =============
422 Information about mounted ext4 file systems can be found in
423 /proc/fs/ext4. Each mounted filesystem will have a directory in
424 /proc/fs/ext4 based on its device name (i.e., /proc/fs/ext4/hdc or
425 /proc/fs/ext4/dm-0). The files in each per-device directory are shown
426 in table below.
428 Files in /proc/fs/ext4/<devname>
429 ..............................................................................
430 File Content
431 mb_groups details of multiblock allocator buddy cache of free blocks
432 ..............................................................................
434 /sys entries
435 ============
437 Information about mounted ext4 file systems can be found in
438 /sys/fs/ext4. Each mounted filesystem will have a directory in
439 /sys/fs/ext4 based on its device name (i.e., /sys/fs/ext4/hdc or
440 /sys/fs/ext4/dm-0). The files in each per-device directory are shown
441 in table below.
443 Files in /sys/fs/ext4/<devname>
444 (see also Documentation/ABI/testing/sysfs-fs-ext4)
445 ..............................................................................
446 File Content
448 delayed_allocation_blocks This file is read-only and shows the number of
449 blocks that are dirty in the page cache, but
450 which do not have their location in the
451 filesystem allocated yet.
453 inode_goal Tuning parameter which (if non-zero) controls
454 the goal inode used by the inode allocator in
455 preference to all other allocation heuristics.
456 This is intended for debugging use only, and
457 should be 0 on production systems.
459 inode_readahead_blks Tuning parameter which controls the maximum
460 number of inode table blocks that ext4's inode
461 table readahead algorithm will pre-read into
462 the buffer cache
464 lifetime_write_kbytes This file is read-only and shows the number of
465 kilobytes of data that have been written to this
466 filesystem since it was created.
468 max_writeback_mb_bump The maximum number of megabytes the writeback
469 code will try to write out before move on to
470 another inode.
472 mb_group_prealloc The multiblock allocator will round up allocation
473 requests to a multiple of this tuning parameter if
474 the stripe size is not set in the ext4 superblock
476 mb_max_to_scan The maximum number of extents the multiblock
477 allocator will search to find the best extent
479 mb_min_to_scan The minimum number of extents the multiblock
480 allocator will search to find the best extent
482 mb_order2_req Tuning parameter which controls the minimum size
483 for requests (as a power of 2) where the buddy
484 cache is used
486 mb_stats Controls whether the multiblock allocator should
487 collect statistics, which are shown during the
488 unmount. 1 means to collect statistics, 0 means
489 not to collect statistics
491 mb_stream_req Files which have fewer blocks than this tunable
492 parameter will have their blocks allocated out
493 of a block group specific preallocation pool, so
494 that small files are packed closely together.
495 Each large file will have its blocks allocated
496 out of its own unique preallocation pool.
498 session_write_kbytes This file is read-only and shows the number of
499 kilobytes of data that have been written to this
500 filesystem since it was mounted.
502 reserved_clusters This is RW file and contains number of reserved
503 clusters in the file system which will be used
504 in the specific situations to avoid costly
505 zeroout, unexpected ENOSPC, or possible data
506 loss. The default is 2% or 4096 clusters,
507 whichever is smaller and this can be changed
508 however it can never exceed number of clusters
509 in the file system. If there is not enough space
510 for the reserved space when mounting the file
511 mount will _not_ fail.
512 ..............................................................................
514 Ioctls
515 ======
517 There is some Ext4 specific functionality which can be accessed by applications
518 through the system call interfaces. The list of all Ext4 specific ioctls are
519 shown in the table below.
521 Table of Ext4 specific ioctls
522 ..............................................................................
523 Ioctl Description
524 EXT4_IOC_GETFLAGS Get additional attributes associated with inode.
525 The ioctl argument is an integer bitfield, with
526 bit values described in ext4.h. This ioctl is an
527 alias for FS_IOC_GETFLAGS.
529 EXT4_IOC_SETFLAGS Set additional attributes associated with inode.
530 The ioctl argument is an integer bitfield, with
531 bit values described in ext4.h. This ioctl is an
532 alias for FS_IOC_SETFLAGS.
534 EXT4_IOC_GETVERSION
535 EXT4_IOC_GETVERSION_OLD
536 Get the inode i_generation number stored for
537 each inode. The i_generation number is normally
538 changed only when new inode is created and it is
539 particularly useful for network filesystems. The
540 '_OLD' version of this ioctl is an alias for
541 FS_IOC_GETVERSION.
543 EXT4_IOC_SETVERSION
544 EXT4_IOC_SETVERSION_OLD
545 Set the inode i_generation number stored for
546 each inode. The '_OLD' version of this ioctl
547 is an alias for FS_IOC_SETVERSION.
549 EXT4_IOC_GROUP_EXTEND This ioctl has the same purpose as the resize
550 mount option. It allows to resize filesystem
551 to the end of the last existing block group,
552 further resize has to be done with resize2fs,
553 either online, or offline. The argument points
554 to the unsigned logn number representing the
555 filesystem new block count.
557 EXT4_IOC_MOVE_EXT Move the block extents from orig_fd (the one
558 this ioctl is pointing to) to the donor_fd (the
559 one specified in move_extent structure passed
560 as an argument to this ioctl). Then, exchange
561 inode metadata between orig_fd and donor_fd.
562 This is especially useful for online
563 defragmentation, because the allocator has the
564 opportunity to allocate moved blocks better,
565 ideally into one contiguous extent.
567 EXT4_IOC_GROUP_ADD Add a new group descriptor to an existing or
568 new group descriptor block. The new group
569 descriptor is described by ext4_new_group_input
570 structure, which is passed as an argument to
571 this ioctl. This is especially useful in
572 conjunction with EXT4_IOC_GROUP_EXTEND,
573 which allows online resize of the filesystem
574 to the end of the last existing block group.
575 Those two ioctls combined is used in userspace
576 online resize tool (e.g. resize2fs).
578 EXT4_IOC_MIGRATE This ioctl operates on the filesystem itself.
579 It converts (migrates) ext3 indirect block mapped
580 inode to ext4 extent mapped inode by walking
581 through indirect block mapping of the original
582 inode and converting contiguous block ranges
583 into ext4 extents of the temporary inode. Then,
584 inodes are swapped. This ioctl might help, when
585 migrating from ext3 to ext4 filesystem, however
586 suggestion is to create fresh ext4 filesystem
587 and copy data from the backup. Note, that
588 filesystem has to support extents for this ioctl
589 to work.
591 EXT4_IOC_ALLOC_DA_BLKS Force all of the delay allocated blocks to be
592 allocated to preserve application-expected ext3
593 behaviour. Note that this will also start
594 triggering a write of the data blocks, but this
595 behaviour may change in the future as it is
596 not necessary and has been done this way only
597 for sake of simplicity.
599 EXT4_IOC_RESIZE_FS Resize the filesystem to a new size. The number
600 of blocks of resized filesystem is passed in via
601 64 bit integer argument. The kernel allocates
602 bitmaps and inode table, the userspace tool thus
603 just passes the new number of blocks.
605 EXT4_IOC_SWAP_BOOT Swap i_blocks and associated attributes
606 (like i_blocks, i_size, i_flags, ...) from
607 the specified inode with inode
608 EXT4_BOOT_LOADER_INO (#5). This is typically
609 used to store a boot loader in a secure part of
610 the filesystem, where it can't be changed by a
611 normal user by accident.
612 The data blocks of the previous boot loader
613 will be associated with the given inode.
615 ..............................................................................
617 References
618 ==========
620 kernel source: <file:fs/ext4/>
621 <file:fs/jbd2/>
623 programs: http://e2fsprogs.sourceforge.net/
625 useful links: http://fedoraproject.org/wiki/ext3-devel
626 http://www.bullopensource.org/ext4/
627 http://ext4.wiki.kernel.org/index.php/Main_Page
628 http://fedoraproject.org/wiki/Features/Ext4