brintos

brintos / linux-shallow public Read only

0
0
Text · 27.1 KiB · 2418b0c Raw
615 lines · plain
1.. SPDX-License-Identifier: GPL-2.02 3========================4ext4 General Information5========================6 7Ext4 is an advanced level of the ext3 filesystem which incorporates8scalability and reliability enhancements for supporting large filesystems9(64 bit) in keeping with increasing disk capacities and state-of-the-art10feature requirements.11 12Mailing list:	linux-ext4@vger.kernel.org13Web site:	http://ext4.wiki.kernel.org14 15 16Quick usage instructions17========================18 19Note: More extensive information for getting started with ext4 can be20found at the ext4 wiki site at the URL:21http://ext4.wiki.kernel.org/index.php/Ext4_Howto22 23  - The latest version of e2fsprogs can be found at:24 25    https://www.kernel.org/pub/linux/kernel/people/tytso/e2fsprogs/26 27	or28 29    http://sourceforge.net/project/showfiles.php?group_id=240630 31	or grab the latest git repository from:32 33   https://git.kernel.org/pub/scm/fs/ext2/e2fsprogs.git34 35  - Create a new filesystem using the ext4 filesystem type:36 37        # mke2fs -t ext4 /dev/hda138 39    Or to configure an existing ext3 filesystem to support extents:40 41	# tune2fs -O extents /dev/hda142 43    If the filesystem was created with 128 byte inodes, it can be44    converted to use 256 byte for greater efficiency via:45 46        # tune2fs -I 256 /dev/hda147 48  - Mounting:49 50	# mount -t ext4 /dev/hda1 /wherever51 52  - When comparing performance with other filesystems, it's always53    important to try multiple workloads; very often a subtle change in a54    workload parameter can completely change the ranking of which55    filesystems do well compared to others.  When comparing versus ext3,56    note that ext4 enables write barriers by default, while ext3 does57    not enable write barriers by default.  So it is useful to use58    explicitly specify whether barriers are enabled or not when via the59    '-o barriers=[0|1]' mount option for both ext3 and ext4 filesystems60    for a fair comparison.  When tuning ext3 for best benchmark numbers,61    it is often worthwhile to try changing the data journaling mode; '-o62    data=writeback' can be faster for some workloads.  (Note however that63    running mounted with data=writeback can potentially leave stale data64    exposed in recently written files in case of an unclean shutdown,65    which could be a security exposure in some situations.)  Configuring66    the filesystem with a large journal can also be helpful for67    metadata-intensive workloads.68 69Features70========71 72Currently Available73-------------------74 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 tree79* 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 time82* inode version field on disk (NFSv4, Lustre)83* reduced e2fsck time via uninit_bg feature84* journal checksumming for robustness, performance85* persistent file preallocation (e.g for streaming media, databases)86* ability to pack bitmaps and inode tables into larger virtual groups via the87  flex_bg feature88* large file support89* inode allocation using large virtual block groups via flex_bg90* delayed allocation91* large block (up to pagesize) support92* efficient new ordered mode in JBD2 and ext4 (avoid using buffer head to force93  the ordering)94* Case-insensitive file name lookups95* file-based encryption support (fscrypt)96* file-based verity support (fsverity)97 98[1] Filesystems with a block size of 1k may see a limit imposed by the99directory hash tree having a maximum depth of two.100 101case-insensitive file name lookups102======================================================103 104The case-insensitive file name lookup feature is supported on a105per-directory basis, allowing the user to mix case-insensitive and106case-sensitive directories in the same filesystem.  It is enabled by107flipping the +F inode attribute of an empty directory.  The108case-insensitive string match operation is only defined when we know how109text in encoded in a byte sequence.  For that reason, in order to enable110case-insensitive directories, the filesystem must have the111casefold feature, which stores the filesystem-wide encoding112model used.  By default, the charset adopted is the latest version of113Unicode (12.1.0, by the time of this writing), encoded in the UTF-8114form.  The comparison algorithm is implemented by normalizing the115strings to the Canonical decomposition form, as defined by Unicode,116followed by a byte per byte comparison.117 118The case-awareness is name-preserving on the disk, meaning that the file119name provided by userspace is a byte-per-byte match to what is actually120written in the disk.  The Unicode normalization format used by the121kernel is thus an internal representation, and not exposed to the122userspace nor to the disk, with the important exception of disk hashes,123used on large case-insensitive directories with DX feature.  On DX124directories, the hash must be calculated using the casefolded version of125the filename, meaning that the normalization format used actually has an126impact on where the directory entry is stored.127 128When we change from viewing filenames as opaque byte sequences to seeing129them as encoded strings we need to address what happens when a program130tries to create a file with an invalid name.  The Unicode subsystem131within the kernel leaves the decision of what to do in this case to the132filesystem, which select its preferred behavior by enabling/disabling133the strict mode.  When Ext4 encounters one of those strings and the134filesystem did not require strict mode, it falls back to considering the135entire string as an opaque byte sequence, which still allows the user to136operate on that file, but the case-insensitive lookups won't work.137 138Options139=======140 141When mounting an ext4 filesystem, the following option are accepted:142(*) == default143 144  ro145        Mount filesystem read only. Note that ext4 will replay the journal (and146        thus write to the partition) even when mounted "read only". The mount147        options "ro,noload" can be used to prevent writes to the filesystem.148 149  journal_checksum150        Enable checksumming of the journal transactions.  This will allow the151        recovery code in e2fsck and the kernel to detect corruption in the152        kernel.  It is a compatible change and will be ignored by older153        kernels.154 155  journal_async_commit156        Commit block can be written to disk without waiting for descriptor157        blocks. If enabled older kernels cannot mount the device. This will158        enable 'journal_checksum' internally.159 160  journal_path=path, journal_dev=devnum161        When the external journal device's major/minor numbers have changed,162        these options allow the user to specify the new journal location.  The163        journal device is identified through either its new major/minor numbers164        encoded in devnum, or via a path to the device.165 166  norecovery, noload167        Don't load the journal on mounting.  Note that if the filesystem was168        not unmounted cleanly, skipping the journal replay will lead to the169        filesystem containing inconsistencies that can lead to any number of170        problems.171 172  data=journal173        All data are committed into the journal prior to being written into the174        main file system.  Enabling this mode will disable delayed allocation175        and O_DIRECT support.176 177  data=ordered	(*)178        All data are forced directly out to the main file system prior to its179        metadata being committed to the journal.180 181  data=writeback182        Data ordering is not preserved, data may be written into the main file183        system after its metadata has been committed to the journal.184 185  commit=nrsec	(*)186        This setting limits the maximum age of the running transaction to187        'nrsec' seconds.  The default value is 5 seconds.  This means that if188        you lose your power, you will lose as much as the latest 5 seconds of189        metadata changes (your filesystem will not be damaged though, thanks190        to the journaling). This default value (or any low value) will hurt191        performance, but it's good for data-safety.  Setting it to 0 will have192        the same effect as leaving it at the default (5 seconds).  Setting it193        to very large values will improve performance.  Note that due to194        delayed allocation even older data can be lost on power failure since195        writeback of those data begins only after time set in196        /proc/sys/vm/dirty_expire_centisecs.197 198  barrier=<0|1(*)>, barrier(*), nobarrier199        This enables/disables the use of write barriers in the jbd code.200        barrier=0 disables, barrier=1 enables.  This also requires an IO stack201        which can support barriers, and if jbd gets an error on a barrier202        write, it will disable again with a warning.  Write barriers enforce203        proper on-disk ordering of journal commits, making volatile disk write204        caches safe to use, at some performance penalty.  If your disks are205        battery-backed in one way or another, disabling barriers may safely206        improve performance.  The mount options "barrier" and "nobarrier" can207        also be used to enable or disable barriers, for consistency with other208        ext4 mount options.209 210  inode_readahead_blks=n211        This tuning parameter controls the maximum number of inode table blocks212        that ext4's inode table readahead algorithm will pre-read into the213        buffer cache.  The default value is 32 blocks.214 215  bsddf	(*)216        Make 'df' act like BSD.217 218  minixdf219        Make 'df' act like Minix.220 221  debug222        Extra debugging information is sent to syslog.223 224  abort225        Simulate the effects of calling ext4_abort() for debugging purposes.226        This is normally used while remounting a filesystem which is already227        mounted.228 229  errors=remount-ro230        Remount the filesystem read-only on an error.231 232  errors=continue233        Keep going on a filesystem error.234 235  errors=panic236        Panic and halt the machine if an error occurs.  (These mount options237        override the errors behavior specified in the superblock, which can be238        configured using tune2fs)239 240  data_err=ignore(*)241        Just print an error message if an error occurs in a file data buffer in242        ordered mode.243  data_err=abort244        Abort the journal if an error occurs in a file data buffer in ordered245        mode.246 247  grpid | bsdgroups248        New objects have the group ID of their parent.249 250  nogrpid (*) | sysvgroups251        New objects have the group ID of their creator.252 253  resgid=n254        The group ID which may use the reserved blocks.255 256  resuid=n257        The user ID which may use the reserved blocks.258 259  sb=260        Use alternate superblock at this location.261 262  quota, noquota, grpquota, usrquota263        These options are ignored by the filesystem. They are used only by264        quota tools to recognize volumes where quota should be turned on. See265        documentation in the quota-tools package for more details266        (http://sourceforge.net/projects/linuxquota).267 268  jqfmt=<quota type>, usrjquota=<file>, grpjquota=<file>269        These options tell filesystem details about quota so that quota270        information can be properly updated during journal replay. They replace271        the above quota options. See documentation in the quota-tools package272        for more details (http://sourceforge.net/projects/linuxquota).273 274  stripe=n275        Number of filesystem blocks that mballoc will try to use for allocation276        size and alignment. For RAID5/6 systems this should be the number of277        data disks *  RAID chunk size in file system blocks.278 279  delalloc	(*)280        Defer block allocation until just before ext4 writes out the block(s)281        in question.  This allows ext4 to better allocation decisions more282        efficiently.283 284  nodelalloc285        Disable delayed allocation.  Blocks are allocated when the data is286        copied from userspace to the page cache, either via the write(2) system287        call or when an mmap'ed page which was previously unallocated is288        written for the first time.289 290  max_batch_time=usec291        Maximum amount of time ext4 should wait for additional filesystem292        operations to be batch together with a synchronous write operation.293        Since a synchronous write operation is going to force a commit and then294        a wait for the I/O complete, it doesn't cost much, and can be a huge295        throughput win, we wait for a small amount of time to see if any other296        transactions can piggyback on the synchronous write.   The algorithm297        used is designed to automatically tune for the speed of the disk, by298        measuring the amount of time (on average) that it takes to finish299        committing a transaction.  Call this time the "commit time".  If the300        time that the transaction has been running is less than the commit301        time, ext4 will try sleeping for the commit time to see if other302        operations will join the transaction.   The commit time is capped by303        the max_batch_time, which defaults to 15000us (15ms).   This304        optimization can be turned off entirely by setting max_batch_time to 0.305 306  min_batch_time=usec307        This parameter sets the commit time (as described above) to be at least308        min_batch_time.  It defaults to zero microseconds.  Increasing this309        parameter may improve the throughput of multi-threaded, synchronous310        workloads on very fast disks, at the cost of increasing latency.311 312  journal_ioprio=prio313        The I/O priority (from 0 to 7, where 0 is the highest priority) which314        should be used for I/O operations submitted by kjournald2 during a315        commit operation.  This defaults to 3, which is a slightly higher316        priority than the default I/O priority.317 318  auto_da_alloc(*), noauto_da_alloc319        Many broken applications don't use fsync() when replacing existing320        files via patterns such as fd = open("foo.new")/write(fd,..)/close(fd)/321        rename("foo.new", "foo"), or worse yet, fd = open("foo",322        O_TRUNC)/write(fd,..)/close(fd).  If auto_da_alloc is enabled, ext4323        will detect the replace-via-rename and replace-via-truncate patterns324        and force that any delayed allocation blocks are allocated such that at325        the next journal commit, in the default data=ordered mode, the data326        blocks of the new file are forced to disk before the rename() operation327        is committed.  This provides roughly the same level of guarantees as328        ext3, and avoids the "zero-length" problem that can happen when a329        system crashes before the delayed allocation blocks are forced to disk.330 331  noinit_itable332        Do not initialize any uninitialized inode table blocks in the333        background.  This feature may be used by installation CD's so that the334        install process can complete as quickly as possible; the inode table335        initialization process would then be deferred until the next time the336        file system is unmounted.337 338  init_itable=n339        The lazy itable init code will wait n times the number of milliseconds340        it took to zero out the previous block group's inode table.  This341        minimizes the impact on the system performance while file system's342        inode table is being initialized.343 344  discard, nodiscard(*)345        Controls whether ext4 should issue discard/TRIM commands to the346        underlying block device when blocks are freed.  This is useful for SSD347        devices and sparse/thinly-provisioned LUNs, but it is off by default348        until sufficient testing has been done.349 350  nouid32351        Disables 32-bit UIDs and GIDs.  This is for interoperability  with352        older kernels which only store and expect 16-bit values.353 354  block_validity(*), noblock_validity355        These options enable or disable the in-kernel facility for tracking356        filesystem metadata blocks within internal data structures.  This357        allows multi- block allocator and other routines to notice bugs or358        corrupted allocation bitmaps which cause blocks to be allocated which359        overlap with filesystem metadata blocks.360 361  dioread_lock, dioread_nolock362        Controls whether or not ext4 should use the DIO read locking. If the363        dioread_nolock option is specified ext4 will allocate uninitialized364        extent before buffer write and convert the extent to initialized after365        IO completes. This approach allows ext4 code to avoid using inode366        mutex, which improves scalability on high speed storages. However this367        does not work with data journaling and dioread_nolock option will be368        ignored with kernel warning. Note that dioread_nolock code path is only369        used for extent-based files.  Because of the restrictions this options370        comprises it is off by default (e.g. dioread_lock).371 372  max_dir_size_kb=n373        This limits the size of directories so that any attempt to expand them374        beyond the specified limit in kilobytes will cause an ENOSPC error.375        This is useful in memory constrained environments, where a very large376        directory can cause severe performance problems or even provoke the Out377        Of Memory killer.  (For example, if there is only 512mb memory378        available, a 176mb directory may seriously cramp the system's style.)379 380  i_version381        Enable 64-bit inode version support. This option is off by default.382 383  dax384        Use direct access (no page cache).  See385        Documentation/filesystems/dax.rst.  Note that this option is386        incompatible with data=journal.387 388  inlinecrypt389        When possible, encrypt/decrypt the contents of encrypted files using the390        blk-crypto framework rather than filesystem-layer encryption. This391        allows the use of inline encryption hardware. The on-disk format is392        unaffected. For more details, see393        Documentation/block/inline-encryption.rst.394 395Data Mode396=========397There are 3 different data modes:398 399* writeback mode400 401  In data=writeback mode, ext4 does not journal data at all.  This mode provides402  a similar level of journaling as that of XFS, JFS, and ReiserFS in its default403  mode - metadata journaling.  A crash+recovery can cause incorrect data to404  appear in files which were written shortly before the crash.  This mode will405  typically provide the best ext4 performance.406 407* ordered mode408 409  In data=ordered mode, ext4 only officially journals metadata, but it logically410  groups metadata information related to data changes with the data blocks into411  a single unit called a transaction.  When it's time to write the new metadata412  out to disk, the associated data blocks are written first.  In general, this413  mode performs slightly slower than writeback but significantly faster than414  journal mode.415 416* journal mode417 418  data=journal mode provides full data and metadata journaling.  All new data is419  written to the journal first, and then to its final location.  In the event of420  a crash, the journal can be replayed, bringing both data and metadata into a421  consistent state.  This mode is the slowest except when data needs to be read422  from and written to disk at the same time where it outperforms all others423  modes.  Enabling this mode will disable delayed allocation and O_DIRECT424  support.425 426/proc entries427=============428 429Information about mounted ext4 file systems can be found in430/proc/fs/ext4.  Each mounted filesystem will have a directory in431/proc/fs/ext4 based on its device name (i.e., /proc/fs/ext4/hdc or432/proc/fs/ext4/dm-0).   The files in each per-device directory are shown433in table below.434 435Files in /proc/fs/ext4/<devname>436 437  mb_groups438        details of multiblock allocator buddy cache of free blocks439 440/sys entries441============442 443Information about mounted ext4 file systems can be found in444/sys/fs/ext4.  Each mounted filesystem will have a directory in445/sys/fs/ext4 based on its device name (i.e., /sys/fs/ext4/hdc or446/sys/fs/ext4/dm-0).   The files in each per-device directory are shown447in table below.448 449Files in /sys/fs/ext4/<devname>:450 451(see also Documentation/ABI/testing/sysfs-fs-ext4)452 453  delayed_allocation_blocks454        This file is read-only and shows the number of blocks that are dirty in455        the page cache, but which do not have their location in the filesystem456        allocated yet.457 458  inode_goal459        Tuning parameter which (if non-zero) controls the goal inode used by460        the inode allocator in preference to all other allocation heuristics.461        This is intended for debugging use only, and should be 0 on production462        systems.463 464  inode_readahead_blks465        Tuning parameter which controls the maximum number of inode table466        blocks that ext4's inode table readahead algorithm will pre-read into467        the buffer cache.468 469  lifetime_write_kbytes470        This file is read-only and shows the number of kilobytes of data that471        have been written to this filesystem since it was created.472 473  max_writeback_mb_bump474        The maximum number of megabytes the writeback code will try to write475        out before move on to another inode.476 477  mb_group_prealloc478        The multiblock allocator will round up allocation requests to a479        multiple of this tuning parameter if the stripe size is not set in the480        ext4 superblock481 482  mb_max_to_scan483        The maximum number of extents the multiblock allocator will search to484        find the best extent.485 486  mb_min_to_scan487        The minimum number of extents the multiblock allocator will search to488        find the best extent.489 490  mb_order2_req491        Tuning parameter which controls the minimum size for requests (as a492        power of 2) where the buddy cache is used.493 494  mb_stats495        Controls whether the multiblock allocator should collect statistics,496        which are shown during the unmount. 1 means to collect statistics, 0497        means not to collect statistics.498 499  mb_stream_req500        Files which have fewer blocks than this tunable parameter will have501        their blocks allocated out of a block group specific preallocation502        pool, so that small files are packed closely together.  Each large file503        will have its blocks allocated out of its own unique preallocation504        pool.505 506  session_write_kbytes507        This file is read-only and shows the number of kilobytes of data that508        have been written to this filesystem since it was mounted.509 510  reserved_clusters511        This is RW file and contains number of reserved clusters in the file512        system which will be used in the specific situations to avoid costly513        zeroout, unexpected ENOSPC, or possible data loss. The default is 2% or514        4096 clusters, whichever is smaller and this can be changed however it515        can never exceed number of clusters in the file system. If there is not516        enough space for the reserved space when mounting the file mount will517        _not_ fail.518 519Ioctls520======521 522Ext4 implements various ioctls which can be used by applications to access523ext4-specific functionality. An incomplete list of these ioctls is shown in the524table below. This list includes truly ext4-specific ioctls (``EXT4_IOC_*``) as525well as ioctls that may have been ext4-specific originally but are now supported526by some other filesystem(s) too (``FS_IOC_*``).527 528Table of Ext4 ioctls529 530  FS_IOC_GETFLAGS531        Get additional attributes associated with inode.  The ioctl argument is532        an integer bitfield, with bit values described in ext4.h.533 534  FS_IOC_SETFLAGS535        Set additional attributes associated with inode.  The ioctl argument is536        an integer bitfield, with bit values described in ext4.h.537 538  EXT4_IOC_GETVERSION, EXT4_IOC_GETVERSION_OLD539        Get the inode i_generation number stored for each inode. The540        i_generation number is normally changed only when new inode is created541        and it is particularly useful for network filesystems. The '_OLD'542        version of this ioctl is an alias for FS_IOC_GETVERSION.543 544  EXT4_IOC_SETVERSION, EXT4_IOC_SETVERSION_OLD545        Set the inode i_generation number stored for each inode. The '_OLD'546        version of this ioctl is an alias for FS_IOC_SETVERSION.547 548  EXT4_IOC_GROUP_EXTEND549        This ioctl has the same purpose as the resize mount option. It allows550        to resize filesystem to the end of the last existing block group,551        further resize has to be done with resize2fs, either online, or552        offline. The argument points to the unsigned logn number representing553        the filesystem new block count.554 555  EXT4_IOC_MOVE_EXT556        Move the block extents from orig_fd (the one this ioctl is pointing to)557        to the donor_fd (the one specified in move_extent structure passed as558        an argument to this ioctl). Then, exchange inode metadata between559        orig_fd and donor_fd.  This is especially useful for online560        defragmentation, because the allocator has the opportunity to allocate561        moved blocks better, ideally into one contiguous extent.562 563  EXT4_IOC_GROUP_ADD564        Add a new group descriptor to an existing or new group descriptor565        block. The new group descriptor is described by ext4_new_group_input566        structure, which is passed as an argument to this ioctl. This is567        especially useful in conjunction with EXT4_IOC_GROUP_EXTEND, which568        allows online resize of the filesystem to the end of the last existing569        block group.  Those two ioctls combined is used in userspace online570        resize tool (e.g. resize2fs).571 572  EXT4_IOC_MIGRATE573        This ioctl operates on the filesystem itself.  It converts (migrates)574        ext3 indirect block mapped inode to ext4 extent mapped inode by walking575        through indirect block mapping of the original inode and converting576        contiguous block ranges into ext4 extents of the temporary inode. Then,577        inodes are swapped. This ioctl might help, when migrating from ext3 to578        ext4 filesystem, however suggestion is to create fresh ext4 filesystem579        and copy data from the backup. Note, that filesystem has to support580        extents for this ioctl to work.581 582  EXT4_IOC_ALLOC_DA_BLKS583        Force all of the delay allocated blocks to be allocated to preserve584        application-expected ext3 behaviour. Note that this will also start585        triggering a write of the data blocks, but this behaviour may change in586        the future as it is not necessary and has been done this way only for587        sake of simplicity.588 589  EXT4_IOC_RESIZE_FS590        Resize the filesystem to a new size.  The number of blocks of resized591        filesystem is passed in via 64 bit integer argument.  The kernel592        allocates bitmaps and inode table, the userspace tool thus just passes593        the new number of blocks.594 595  EXT4_IOC_SWAP_BOOT596        Swap i_blocks and associated attributes (like i_blocks, i_size,597        i_flags, ...) from the specified inode with inode EXT4_BOOT_LOADER_INO598        (#5). This is typically used to store a boot loader in a secure part of599        the filesystem, where it can't be changed by a normal user by accident.600        The data blocks of the previous boot loader will be associated with the601        given inode.602 603References604==========605 606kernel source:	<file:fs/ext4/>607		<file:fs/jbd2/>608 609programs:	http://e2fsprogs.sourceforge.net/610 611useful links:	https://fedoraproject.org/wiki/ext3-devel612		http://www.bullopensource.org/ext4/613		http://ext4.wiki.kernel.org/index.php/Main_Page614		https://fedoraproject.org/wiki/Features/Ext4615