f2fs.rst 39 KB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195196197198199200201202203204205206207208209210211212213214215216217218219220221222223224225226227228229230231232233234235236237238239240241242243244245246247248249250251252253254255256257258259260261262263264265266267268269270271272273274275276277278279280281282283284285286287288289290291292293294295296297298299300301302303304305306307308309310311312313314315316317318319320321322323324325326327328329330331332333334335336337338339340341342343344345346347348349350351352353354355356357358359360361362363364365366367368369370371372373374375376377378379380381382383384385386387388389390391392393394395396397398399400401402403404405406407408409410411412413414415416417418419420421422423424425426427428429430431432433434435436437438439440441442443444445446447448449450451452453454455456457458459460461462463464465466467468469470471472473474475476477478479480481482483484485486487488489490491492493494495496497498499500501502503504505506507508509510511512513514515516517518519520521522523524525526527528529530531532533534535536537538539540541542543544545546547548549550551552553554555556557558559560561562563564565566567568569570571572573574575576577578579580581582583584585586587588589590591592593594595596597598599600601602603604605606607608609610611612613614615616617618619620621622623624625626627628629630631632633634635636637638639640641642643644645646647648649650651652653654655656657658659660661662663664665666667668669670671672673674675676677678679680681682683684685686687688689690691692693694695696697698699700701702703704705706707708709710711712713714715716717718719720721722723724725726727728729730731732733734735736737738739740741742743744745746747748749750751752753754755756757758759760761762763764765766767768769770771772773774775776777778779780781782783784785786787788789790791792793794795796797798799800801802803804805806807808809810811812813814815816817818819820821822823824825826827828829830831832833834835836837838839840841842843844845846847848849850851852853854855856857858859860861862863864865866867868869870871872873874875876877878879880881882883884885886887888889890891892893894
  1. .. SPDX-License-Identifier: GPL-2.0
  2. ==========================================
  3. WHAT IS Flash-Friendly File System (F2FS)?
  4. ==========================================
  5. NAND flash memory-based storage devices, such as SSD, eMMC, and SD cards, have
  6. been equipped on a variety systems ranging from mobile to server systems. Since
  7. they are known to have different characteristics from the conventional rotating
  8. disks, a file system, an upper layer to the storage device, should adapt to the
  9. changes from the sketch in the design level.
  10. F2FS is a file system exploiting NAND flash memory-based storage devices, which
  11. is based on Log-structured File System (LFS). The design has been focused on
  12. addressing the fundamental issues in LFS, which are snowball effect of wandering
  13. tree and high cleaning overhead.
  14. Since a NAND flash memory-based storage device shows different characteristic
  15. according to its internal geometry or flash memory management scheme, namely FTL,
  16. F2FS and its tools support various parameters not only for configuring on-disk
  17. layout, but also for selecting allocation and cleaning algorithms.
  18. The following git tree provides the file system formatting tool (mkfs.f2fs),
  19. a consistency checking tool (fsck.f2fs), and a debugging tool (dump.f2fs).
  20. - git://git.kernel.org/pub/scm/linux/kernel/git/jaegeuk/f2fs-tools.git
  21. For reporting bugs and sending patches, please use the following mailing list:
  22. - linux-f2fs-devel@lists.sourceforge.net
  23. Background and Design issues
  24. ============================
  25. Log-structured File System (LFS)
  26. --------------------------------
  27. "A log-structured file system writes all modifications to disk sequentially in
  28. a log-like structure, thereby speeding up both file writing and crash recovery.
  29. The log is the only structure on disk; it contains indexing information so that
  30. files can be read back from the log efficiently. In order to maintain large free
  31. areas on disk for fast writing, we divide the log into segments and use a
  32. segment cleaner to compress the live information from heavily fragmented
  33. segments." from Rosenblum, M. and Ousterhout, J. K., 1992, "The design and
  34. implementation of a log-structured file system", ACM Trans. Computer Systems
  35. 10, 1, 26–52.
  36. Wandering Tree Problem
  37. ----------------------
  38. In LFS, when a file data is updated and written to the end of log, its direct
  39. pointer block is updated due to the changed location. Then the indirect pointer
  40. block is also updated due to the direct pointer block update. In this manner,
  41. the upper index structures such as inode, inode map, and checkpoint block are
  42. also updated recursively. This problem is called as wandering tree problem [1],
  43. and in order to enhance the performance, it should eliminate or relax the update
  44. propagation as much as possible.
  45. [1] Bityutskiy, A. 2005. JFFS3 design issues. http://www.linux-mtd.infradead.org/
  46. Cleaning Overhead
  47. -----------------
  48. Since LFS is based on out-of-place writes, it produces so many obsolete blocks
  49. scattered across the whole storage. In order to serve new empty log space, it
  50. needs to reclaim these obsolete blocks seamlessly to users. This job is called
  51. as a cleaning process.
  52. The process consists of three operations as follows.
  53. 1. A victim segment is selected through referencing segment usage table.
  54. 2. It loads parent index structures of all the data in the victim identified by
  55. segment summary blocks.
  56. 3. It checks the cross-reference between the data and its parent index structure.
  57. 4. It moves valid data selectively.
  58. This cleaning job may cause unexpected long delays, so the most important goal
  59. is to hide the latencies to users. And also definitely, it should reduce the
  60. amount of valid data to be moved, and move them quickly as well.
  61. Key Features
  62. ============
  63. Flash Awareness
  64. ---------------
  65. - Enlarge the random write area for better performance, but provide the high
  66. spatial locality
  67. - Align FS data structures to the operational units in FTL as best efforts
  68. Wandering Tree Problem
  69. ----------------------
  70. - Use a term, “node”, that represents inodes as well as various pointer blocks
  71. - Introduce Node Address Table (NAT) containing the locations of all the “node”
  72. blocks; this will cut off the update propagation.
  73. Cleaning Overhead
  74. -----------------
  75. - Support a background cleaning process
  76. - Support greedy and cost-benefit algorithms for victim selection policies
  77. - Support multi-head logs for static/dynamic hot and cold data separation
  78. - Introduce adaptive logging for efficient block allocation
  79. Mount Options
  80. =============
  81. ======================== ============================================================
  82. background_gc=%s Turn on/off cleaning operations, namely garbage
  83. collection, triggered in background when I/O subsystem is
  84. idle. If background_gc=on, it will turn on the garbage
  85. collection and if background_gc=off, garbage collection
  86. will be turned off. If background_gc=sync, it will turn
  87. on synchronous garbage collection running in background.
  88. Default value for this option is on. So garbage
  89. collection is on by default.
  90. gc_merge When background_gc is on, this option can be enabled to
  91. let background GC thread to handle foreground GC requests,
  92. it can eliminate the sluggish issue caused by slow foreground
  93. GC operation when GC is triggered from a process with limited
  94. I/O and CPU resources.
  95. nogc_merge Disable GC merge feature.
  96. disable_roll_forward Disable the roll-forward recovery routine
  97. norecovery Disable the roll-forward recovery routine, mounted read-
  98. only (i.e., -o ro,disable_roll_forward)
  99. discard/nodiscard Enable/disable real-time discard in f2fs, if discard is
  100. enabled, f2fs will issue discard/TRIM commands when a
  101. segment is cleaned.
  102. no_heap Disable heap-style segment allocation which finds free
  103. segments for data from the beginning of main area, while
  104. for node from the end of main area.
  105. nouser_xattr Disable Extended User Attributes. Note: xattr is enabled
  106. by default if CONFIG_F2FS_FS_XATTR is selected.
  107. noacl Disable POSIX Access Control List. Note: acl is enabled
  108. by default if CONFIG_F2FS_FS_POSIX_ACL is selected.
  109. active_logs=%u Support configuring the number of active logs. In the
  110. current design, f2fs supports only 2, 4, and 6 logs.
  111. Default number is 6.
  112. disable_ext_identify Disable the extension list configured by mkfs, so f2fs
  113. is not aware of cold files such as media files.
  114. inline_xattr Enable the inline xattrs feature.
  115. noinline_xattr Disable the inline xattrs feature.
  116. inline_xattr_size=%u Support configuring inline xattr size, it depends on
  117. flexible inline xattr feature.
  118. inline_data Enable the inline data feature: Newly created small (<~3.4k)
  119. files can be written into inode block.
  120. inline_dentry Enable the inline dir feature: data in newly created
  121. directory entries can be written into inode block. The
  122. space of inode block which is used to store inline
  123. dentries is limited to ~3.4k.
  124. noinline_dentry Disable the inline dentry feature.
  125. flush_merge Merge concurrent cache_flush commands as much as possible
  126. to eliminate redundant command issues. If the underlying
  127. device handles the cache_flush command relatively slowly,
  128. recommend to enable this option.
  129. nobarrier This option can be used if underlying storage guarantees
  130. its cached data should be written to the novolatile area.
  131. If this option is set, no cache_flush commands are issued
  132. but f2fs still guarantees the write ordering of all the
  133. data writes.
  134. fastboot This option is used when a system wants to reduce mount
  135. time as much as possible, even though normal performance
  136. can be sacrificed.
  137. extent_cache Enable an extent cache based on rb-tree, it can cache
  138. as many as extent which map between contiguous logical
  139. address and physical address per inode, resulting in
  140. increasing the cache hit ratio. Set by default.
  141. noextent_cache Disable an extent cache based on rb-tree explicitly, see
  142. the above extent_cache mount option.
  143. noinline_data Disable the inline data feature, inline data feature is
  144. enabled by default.
  145. data_flush Enable data flushing before checkpoint in order to
  146. persist data of regular and symlink.
  147. reserve_root=%d Support configuring reserved space which is used for
  148. allocation from a privileged user with specified uid or
  149. gid, unit: 4KB, the default limit is 0.2% of user blocks.
  150. resuid=%d The user ID which may use the reserved blocks.
  151. resgid=%d The group ID which may use the reserved blocks.
  152. fault_injection=%d Enable fault injection in all supported types with
  153. specified injection rate.
  154. fault_type=%d Support configuring fault injection type, should be
  155. enabled with fault_injection option, fault type value
  156. is shown below, it supports single or combined type.
  157. =================== ===========
  158. Type_Name Type_Value
  159. =================== ===========
  160. FAULT_KMALLOC 0x000000001
  161. FAULT_KVMALLOC 0x000000002
  162. FAULT_PAGE_ALLOC 0x000000004
  163. FAULT_PAGE_GET 0x000000008
  164. FAULT_ALLOC_NID 0x000000020
  165. FAULT_ORPHAN 0x000000040
  166. FAULT_BLOCK 0x000000080
  167. FAULT_DIR_DEPTH 0x000000100
  168. FAULT_EVICT_INODE 0x000000200
  169. FAULT_TRUNCATE 0x000000400
  170. FAULT_READ_IO 0x000000800
  171. FAULT_CHECKPOINT 0x000001000
  172. FAULT_DISCARD 0x000002000
  173. FAULT_WRITE_IO 0x000004000
  174. =================== ===========
  175. mode=%s Control block allocation mode which supports "adaptive"
  176. and "lfs". In "lfs" mode, there should be no random
  177. writes towards main area.
  178. io_bits=%u Set the bit size of write IO requests. It should be set
  179. with "mode=lfs".
  180. usrquota Enable plain user disk quota accounting.
  181. grpquota Enable plain group disk quota accounting.
  182. prjquota Enable plain project quota accounting.
  183. usrjquota=<file> Appoint specified file and type during mount, so that quota
  184. grpjquota=<file> information can be properly updated during recovery flow,
  185. prjjquota=<file> <quota file>: must be in root directory;
  186. jqfmt=<quota type> <quota type>: [vfsold,vfsv0,vfsv1].
  187. offusrjquota Turn off user journalled quota.
  188. offgrpjquota Turn off group journalled quota.
  189. offprjjquota Turn off project journalled quota.
  190. quota Enable plain user disk quota accounting.
  191. noquota Disable all plain disk quota option.
  192. whint_mode=%s Control which write hints are passed down to block
  193. layer. This supports "off", "user-based", and
  194. "fs-based". In "off" mode (default), f2fs does not pass
  195. down hints. In "user-based" mode, f2fs tries to pass
  196. down hints given by users. And in "fs-based" mode, f2fs
  197. passes down hints with its policy.
  198. alloc_mode=%s Adjust block allocation policy, which supports "reuse"
  199. and "default".
  200. fsync_mode=%s Control the policy of fsync. Currently supports "posix",
  201. "strict", and "nobarrier". In "posix" mode, which is
  202. default, fsync will follow POSIX semantics and does a
  203. light operation to improve the filesystem performance.
  204. In "strict" mode, fsync will be heavy and behaves in line
  205. with xfs, ext4 and btrfs, where xfstest generic/342 will
  206. pass, but the performance will regress. "nobarrier" is
  207. based on "posix", but doesn't issue flush command for
  208. non-atomic files likewise "nobarrier" mount option.
  209. test_dummy_encryption
  210. test_dummy_encryption=%s
  211. Enable dummy encryption, which provides a fake fscrypt
  212. context. The fake fscrypt context is used by xfstests.
  213. The argument may be either "v1" or "v2", in order to
  214. select the corresponding fscrypt policy version.
  215. checkpoint=%s[:%u[%]] Set to "disable" to turn off checkpointing. Set to "enable"
  216. to reenable checkpointing. Is enabled by default. While
  217. disabled, any unmounting or unexpected shutdowns will cause
  218. the filesystem contents to appear as they did when the
  219. filesystem was mounted with that option.
  220. While mounting with checkpoint=disabled, the filesystem must
  221. run garbage collection to ensure that all available space can
  222. be used. If this takes too much time, the mount may return
  223. EAGAIN. You may optionally add a value to indicate how much
  224. of the disk you would be willing to temporarily give up to
  225. avoid additional garbage collection. This can be given as a
  226. number of blocks, or as a percent. For instance, mounting
  227. with checkpoint=disable:100% would always succeed, but it may
  228. hide up to all remaining free space. The actual space that
  229. would be unusable can be viewed at /sys/fs/f2fs/<disk>/unusable
  230. This space is reclaimed once checkpoint=enable.
  231. checkpoint_merge When checkpoint is enabled, this can be used to create a kernel
  232. daemon and make it to merge concurrent checkpoint requests as
  233. much as possible to eliminate redundant checkpoint issues. Plus,
  234. we can eliminate the sluggish issue caused by slow checkpoint
  235. operation when the checkpoint is done in a process context in
  236. a cgroup having low i/o budget and cpu shares. To make this
  237. do better, we set the default i/o priority of the kernel daemon
  238. to "3", to give one higher priority than other kernel threads.
  239. This is the same way to give a I/O priority to the jbd2
  240. journaling thread of ext4 filesystem.
  241. nocheckpoint_merge Disable checkpoint merge feature.
  242. compress_algorithm=%s Control compress algorithm, currently f2fs supports "lzo",
  243. "lz4", "zstd" and "lzo-rle" algorithm.
  244. compress_algorithm=%s:%d Control compress algorithm and its compress level, now, only
  245. "lz4" and "zstd" support compress level config.
  246. algorithm level range
  247. lz4 3 - 16
  248. zstd 1 - 22
  249. compress_log_size=%u Support configuring compress cluster size, the size will
  250. be 4KB * (1 << %u), 16KB is minimum size, also it's
  251. default size.
  252. compress_extension=%s Support adding specified extension, so that f2fs can enable
  253. compression on those corresponding files, e.g. if all files
  254. with '.ext' has high compression rate, we can set the '.ext'
  255. on compression extension list and enable compression on
  256. these file by default rather than to enable it via ioctl.
  257. For other files, we can still enable compression via ioctl.
  258. Note that, there is one reserved special extension '*', it
  259. can be set to enable compression for all files.
  260. compress_chksum Support verifying chksum of raw data in compressed cluster.
  261. compress_mode=%s Control file compression mode. This supports "fs" and "user"
  262. modes. In "fs" mode (default), f2fs does automatic compression
  263. on the compression enabled files. In "user" mode, f2fs disables
  264. the automaic compression and gives the user discretion of
  265. choosing the target file and the timing. The user can do manual
  266. compression/decompression on the compression enabled files using
  267. ioctls.
  268. compress_cache Support to use address space of a filesystem managed inode to
  269. cache compressed block, in order to improve cache hit ratio of
  270. random read.
  271. inlinecrypt When possible, encrypt/decrypt the contents of encrypted
  272. files using the blk-crypto framework rather than
  273. filesystem-layer encryption. This allows the use of
  274. inline encryption hardware. The on-disk format is
  275. unaffected. For more details, see
  276. Documentation/block/inline-encryption.rst.
  277. atgc Enable age-threshold garbage collection, it provides high
  278. effectiveness and efficiency on background GC.
  279. ======================== ============================================================
  280. Debugfs Entries
  281. ===============
  282. /sys/kernel/debug/f2fs/ contains information about all the partitions mounted as
  283. f2fs. Each file shows the whole f2fs information.
  284. /sys/kernel/debug/f2fs/status includes:
  285. - major file system information managed by f2fs currently
  286. - average SIT information about whole segments
  287. - current memory footprint consumed by f2fs.
  288. Sysfs Entries
  289. =============
  290. Information about mounted f2fs file systems can be found in
  291. /sys/fs/f2fs. Each mounted filesystem will have a directory in
  292. /sys/fs/f2fs based on its device name (i.e., /sys/fs/f2fs/sda).
  293. The files in each per-device directory are shown in table below.
  294. Files in /sys/fs/f2fs/<devname>
  295. (see also Documentation/ABI/testing/sysfs-fs-f2fs)
  296. Usage
  297. =====
  298. 1. Download userland tools and compile them.
  299. 2. Skip, if f2fs was compiled statically inside kernel.
  300. Otherwise, insert the f2fs.ko module::
  301. # insmod f2fs.ko
  302. 3. Create a directory to use when mounting::
  303. # mkdir /mnt/f2fs
  304. 4. Format the block device, and then mount as f2fs::
  305. # mkfs.f2fs -l label /dev/block_device
  306. # mount -t f2fs /dev/block_device /mnt/f2fs
  307. mkfs.f2fs
  308. ---------
  309. The mkfs.f2fs is for the use of formatting a partition as the f2fs filesystem,
  310. which builds a basic on-disk layout.
  311. The quick options consist of:
  312. =============== ===========================================================
  313. ``-l [label]`` Give a volume label, up to 512 unicode name.
  314. ``-a [0 or 1]`` Split start location of each area for heap-based allocation.
  315. 1 is set by default, which performs this.
  316. ``-o [int]`` Set overprovision ratio in percent over volume size.
  317. 5 is set by default.
  318. ``-s [int]`` Set the number of segments per section.
  319. 1 is set by default.
  320. ``-z [int]`` Set the number of sections per zone.
  321. 1 is set by default.
  322. ``-e [str]`` Set basic extension list. e.g. "mp3,gif,mov"
  323. ``-t [0 or 1]`` Disable discard command or not.
  324. 1 is set by default, which conducts discard.
  325. =============== ===========================================================
  326. Note: please refer to the manpage of mkfs.f2fs(8) to get full option list.
  327. fsck.f2fs
  328. ---------
  329. The fsck.f2fs is a tool to check the consistency of an f2fs-formatted
  330. partition, which examines whether the filesystem metadata and user-made data
  331. are cross-referenced correctly or not.
  332. Note that, initial version of the tool does not fix any inconsistency.
  333. The quick options consist of::
  334. -d debug level [default:0]
  335. Note: please refer to the manpage of fsck.f2fs(8) to get full option list.
  336. dump.f2fs
  337. ---------
  338. The dump.f2fs shows the information of specific inode and dumps SSA and SIT to
  339. file. Each file is dump_ssa and dump_sit.
  340. The dump.f2fs is used to debug on-disk data structures of the f2fs filesystem.
  341. It shows on-disk inode information recognized by a given inode number, and is
  342. able to dump all the SSA and SIT entries into predefined files, ./dump_ssa and
  343. ./dump_sit respectively.
  344. The options consist of::
  345. -d debug level [default:0]
  346. -i inode no (hex)
  347. -s [SIT dump segno from #1~#2 (decimal), for all 0~-1]
  348. -a [SSA dump segno from #1~#2 (decimal), for all 0~-1]
  349. Examples::
  350. # dump.f2fs -i [ino] /dev/sdx
  351. # dump.f2fs -s 0~-1 /dev/sdx (SIT dump)
  352. # dump.f2fs -a 0~-1 /dev/sdx (SSA dump)
  353. Note: please refer to the manpage of dump.f2fs(8) to get full option list.
  354. sload.f2fs
  355. ----------
  356. The sload.f2fs gives a way to insert files and directories in the exisiting disk
  357. image. This tool is useful when building f2fs images given compiled files.
  358. Note: please refer to the manpage of sload.f2fs(8) to get full option list.
  359. resize.f2fs
  360. -----------
  361. The resize.f2fs lets a user resize the f2fs-formatted disk image, while preserving
  362. all the files and directories stored in the image.
  363. Note: please refer to the manpage of resize.f2fs(8) to get full option list.
  364. defrag.f2fs
  365. -----------
  366. The defrag.f2fs can be used to defragment scattered written data as well as
  367. filesystem metadata across the disk. This can improve the write speed by giving
  368. more free consecutive space.
  369. Note: please refer to the manpage of defrag.f2fs(8) to get full option list.
  370. f2fs_io
  371. -------
  372. The f2fs_io is a simple tool to issue various filesystem APIs as well as
  373. f2fs-specific ones, which is very useful for QA tests.
  374. Note: please refer to the manpage of f2fs_io(8) to get full option list.
  375. Design
  376. ======
  377. On-disk Layout
  378. --------------
  379. F2FS divides the whole volume into a number of segments, each of which is fixed
  380. to 2MB in size. A section is composed of consecutive segments, and a zone
  381. consists of a set of sections. By default, section and zone sizes are set to one
  382. segment size identically, but users can easily modify the sizes by mkfs.
  383. F2FS splits the entire volume into six areas, and all the areas except superblock
  384. consist of multiple segments as described below::
  385. align with the zone size <-|
  386. |-> align with the segment size
  387. _________________________________________________________________________
  388. | | | Segment | Node | Segment | |
  389. | Superblock | Checkpoint | Info. | Address | Summary | Main |
  390. | (SB) | (CP) | Table (SIT) | Table (NAT) | Area (SSA) | |
  391. |____________|_____2______|______N______|______N______|______N_____|__N___|
  392. . .
  393. . .
  394. . .
  395. ._________________________________________.
  396. |_Segment_|_..._|_Segment_|_..._|_Segment_|
  397. . .
  398. ._________._________
  399. |_section_|__...__|_
  400. . .
  401. .________.
  402. |__zone__|
  403. - Superblock (SB)
  404. It is located at the beginning of the partition, and there exist two copies
  405. to avoid file system crash. It contains basic partition information and some
  406. default parameters of f2fs.
  407. - Checkpoint (CP)
  408. It contains file system information, bitmaps for valid NAT/SIT sets, orphan
  409. inode lists, and summary entries of current active segments.
  410. - Segment Information Table (SIT)
  411. It contains segment information such as valid block count and bitmap for the
  412. validity of all the blocks.
  413. - Node Address Table (NAT)
  414. It is composed of a block address table for all the node blocks stored in
  415. Main area.
  416. - Segment Summary Area (SSA)
  417. It contains summary entries which contains the owner information of all the
  418. data and node blocks stored in Main area.
  419. - Main Area
  420. It contains file and directory data including their indices.
  421. In order to avoid misalignment between file system and flash-based storage, F2FS
  422. aligns the start block address of CP with the segment size. Also, it aligns the
  423. start block address of Main area with the zone size by reserving some segments
  424. in SSA area.
  425. Reference the following survey for additional technical details.
  426. https://wiki.linaro.org/WorkingGroups/Kernel/Projects/FlashCardSurvey
  427. File System Metadata Structure
  428. ------------------------------
  429. F2FS adopts the checkpointing scheme to maintain file system consistency. At
  430. mount time, F2FS first tries to find the last valid checkpoint data by scanning
  431. CP area. In order to reduce the scanning time, F2FS uses only two copies of CP.
  432. One of them always indicates the last valid data, which is called as shadow copy
  433. mechanism. In addition to CP, NAT and SIT also adopt the shadow copy mechanism.
  434. For file system consistency, each CP points to which NAT and SIT copies are
  435. valid, as shown as below::
  436. +--------+----------+---------+
  437. | CP | SIT | NAT |
  438. +--------+----------+---------+
  439. . . . .
  440. . . . .
  441. . . . .
  442. +-------+-------+--------+--------+--------+--------+
  443. | CP #0 | CP #1 | SIT #0 | SIT #1 | NAT #0 | NAT #1 |
  444. +-------+-------+--------+--------+--------+--------+
  445. | ^ ^
  446. | | |
  447. `----------------------------------------'
  448. Index Structure
  449. ---------------
  450. The key data structure to manage the data locations is a "node". Similar to
  451. traditional file structures, F2FS has three types of node: inode, direct node,
  452. indirect node. F2FS assigns 4KB to an inode block which contains 923 data block
  453. indices, two direct node pointers, two indirect node pointers, and one double
  454. indirect node pointer as described below. One direct node block contains 1018
  455. data blocks, and one indirect node block contains also 1018 node blocks. Thus,
  456. one inode block (i.e., a file) covers::
  457. 4KB * (923 + 2 * 1018 + 2 * 1018 * 1018 + 1018 * 1018 * 1018) := 3.94TB.
  458. Inode block (4KB)
  459. |- data (923)
  460. |- direct node (2)
  461. | `- data (1018)
  462. |- indirect node (2)
  463. | `- direct node (1018)
  464. | `- data (1018)
  465. `- double indirect node (1)
  466. `- indirect node (1018)
  467. `- direct node (1018)
  468. `- data (1018)
  469. Note that all the node blocks are mapped by NAT which means the location of
  470. each node is translated by the NAT table. In the consideration of the wandering
  471. tree problem, F2FS is able to cut off the propagation of node updates caused by
  472. leaf data writes.
  473. Directory Structure
  474. -------------------
  475. A directory entry occupies 11 bytes, which consists of the following attributes.
  476. - hash hash value of the file name
  477. - ino inode number
  478. - len the length of file name
  479. - type file type such as directory, symlink, etc
  480. A dentry block consists of 214 dentry slots and file names. Therein a bitmap is
  481. used to represent whether each dentry is valid or not. A dentry block occupies
  482. 4KB with the following composition.
  483. ::
  484. Dentry Block(4 K) = bitmap (27 bytes) + reserved (3 bytes) +
  485. dentries(11 * 214 bytes) + file name (8 * 214 bytes)
  486. [Bucket]
  487. +--------------------------------+
  488. |dentry block 1 | dentry block 2 |
  489. +--------------------------------+
  490. . .
  491. . .
  492. . [Dentry Block Structure: 4KB] .
  493. +--------+----------+----------+------------+
  494. | bitmap | reserved | dentries | file names |
  495. +--------+----------+----------+------------+
  496. [Dentry Block: 4KB] . .
  497. . .
  498. . .
  499. +------+------+-----+------+
  500. | hash | ino | len | type |
  501. +------+------+-----+------+
  502. [Dentry Structure: 11 bytes]
  503. F2FS implements multi-level hash tables for directory structure. Each level has
  504. a hash table with dedicated number of hash buckets as shown below. Note that
  505. "A(2B)" means a bucket includes 2 data blocks.
  506. ::
  507. ----------------------
  508. A : bucket
  509. B : block
  510. N : MAX_DIR_HASH_DEPTH
  511. ----------------------
  512. level #0 | A(2B)
  513. |
  514. level #1 | A(2B) - A(2B)
  515. |
  516. level #2 | A(2B) - A(2B) - A(2B) - A(2B)
  517. . | . . . .
  518. level #N/2 | A(2B) - A(2B) - A(2B) - A(2B) - A(2B) - ... - A(2B)
  519. . | . . . .
  520. level #N | A(4B) - A(4B) - A(4B) - A(4B) - A(4B) - ... - A(4B)
  521. The number of blocks and buckets are determined by::
  522. ,- 2, if n < MAX_DIR_HASH_DEPTH / 2,
  523. # of blocks in level #n = |
  524. `- 4, Otherwise
  525. ,- 2^(n + dir_level),
  526. | if n + dir_level < MAX_DIR_HASH_DEPTH / 2,
  527. # of buckets in level #n = |
  528. `- 2^((MAX_DIR_HASH_DEPTH / 2) - 1),
  529. Otherwise
  530. When F2FS finds a file name in a directory, at first a hash value of the file
  531. name is calculated. Then, F2FS scans the hash table in level #0 to find the
  532. dentry consisting of the file name and its inode number. If not found, F2FS
  533. scans the next hash table in level #1. In this way, F2FS scans hash tables in
  534. each levels incrementally from 1 to N. In each level F2FS needs to scan only
  535. one bucket determined by the following equation, which shows O(log(# of files))
  536. complexity::
  537. bucket number to scan in level #n = (hash value) % (# of buckets in level #n)
  538. In the case of file creation, F2FS finds empty consecutive slots that cover the
  539. file name. F2FS searches the empty slots in the hash tables of whole levels from
  540. 1 to N in the same way as the lookup operation.
  541. The following figure shows an example of two cases holding children::
  542. --------------> Dir <--------------
  543. | |
  544. child child
  545. child - child [hole] - child
  546. child - child - child [hole] - [hole] - child
  547. Case 1: Case 2:
  548. Number of children = 6, Number of children = 3,
  549. File size = 7 File size = 7
  550. Default Block Allocation
  551. ------------------------
  552. At runtime, F2FS manages six active logs inside "Main" area: Hot/Warm/Cold node
  553. and Hot/Warm/Cold data.
  554. - Hot node contains direct node blocks of directories.
  555. - Warm node contains direct node blocks except hot node blocks.
  556. - Cold node contains indirect node blocks
  557. - Hot data contains dentry blocks
  558. - Warm data contains data blocks except hot and cold data blocks
  559. - Cold data contains multimedia data or migrated data blocks
  560. LFS has two schemes for free space management: threaded log and copy-and-compac-
  561. tion. The copy-and-compaction scheme which is known as cleaning, is well-suited
  562. for devices showing very good sequential write performance, since free segments
  563. are served all the time for writing new data. However, it suffers from cleaning
  564. overhead under high utilization. Contrarily, the threaded log scheme suffers
  565. from random writes, but no cleaning process is needed. F2FS adopts a hybrid
  566. scheme where the copy-and-compaction scheme is adopted by default, but the
  567. policy is dynamically changed to the threaded log scheme according to the file
  568. system status.
  569. In order to align F2FS with underlying flash-based storage, F2FS allocates a
  570. segment in a unit of section. F2FS expects that the section size would be the
  571. same as the unit size of garbage collection in FTL. Furthermore, with respect
  572. to the mapping granularity in FTL, F2FS allocates each section of the active
  573. logs from different zones as much as possible, since FTL can write the data in
  574. the active logs into one allocation unit according to its mapping granularity.
  575. Cleaning process
  576. ----------------
  577. F2FS does cleaning both on demand and in the background. On-demand cleaning is
  578. triggered when there are not enough free segments to serve VFS calls. Background
  579. cleaner is operated by a kernel thread, and triggers the cleaning job when the
  580. system is idle.
  581. F2FS supports two victim selection policies: greedy and cost-benefit algorithms.
  582. In the greedy algorithm, F2FS selects a victim segment having the smallest number
  583. of valid blocks. In the cost-benefit algorithm, F2FS selects a victim segment
  584. according to the segment age and the number of valid blocks in order to address
  585. log block thrashing problem in the greedy algorithm. F2FS adopts the greedy
  586. algorithm for on-demand cleaner, while background cleaner adopts cost-benefit
  587. algorithm.
  588. In order to identify whether the data in the victim segment are valid or not,
  589. F2FS manages a bitmap. Each bit represents the validity of a block, and the
  590. bitmap is composed of a bit stream covering whole blocks in main area.
  591. Write-hint Policy
  592. -----------------
  593. 1) whint_mode=off. F2FS only passes down WRITE_LIFE_NOT_SET.
  594. 2) whint_mode=user-based. F2FS tries to pass down hints given by
  595. users.
  596. ===================== ======================== ===================
  597. User F2FS Block
  598. ===================== ======================== ===================
  599. N/A META WRITE_LIFE_NOT_SET
  600. N/A HOT_NODE "
  601. N/A WARM_NODE "
  602. N/A COLD_NODE "
  603. ioctl(COLD) COLD_DATA WRITE_LIFE_EXTREME
  604. extension list " "
  605. -- buffered io
  606. WRITE_LIFE_EXTREME COLD_DATA WRITE_LIFE_EXTREME
  607. WRITE_LIFE_SHORT HOT_DATA WRITE_LIFE_SHORT
  608. WRITE_LIFE_NOT_SET WARM_DATA WRITE_LIFE_NOT_SET
  609. WRITE_LIFE_NONE " "
  610. WRITE_LIFE_MEDIUM " "
  611. WRITE_LIFE_LONG " "
  612. -- direct io
  613. WRITE_LIFE_EXTREME COLD_DATA WRITE_LIFE_EXTREME
  614. WRITE_LIFE_SHORT HOT_DATA WRITE_LIFE_SHORT
  615. WRITE_LIFE_NOT_SET WARM_DATA WRITE_LIFE_NOT_SET
  616. WRITE_LIFE_NONE " WRITE_LIFE_NONE
  617. WRITE_LIFE_MEDIUM " WRITE_LIFE_MEDIUM
  618. WRITE_LIFE_LONG " WRITE_LIFE_LONG
  619. ===================== ======================== ===================
  620. 3) whint_mode=fs-based. F2FS passes down hints with its policy.
  621. ===================== ======================== ===================
  622. User F2FS Block
  623. ===================== ======================== ===================
  624. N/A META WRITE_LIFE_MEDIUM;
  625. N/A HOT_NODE WRITE_LIFE_NOT_SET
  626. N/A WARM_NODE "
  627. N/A COLD_NODE WRITE_LIFE_NONE
  628. ioctl(COLD) COLD_DATA WRITE_LIFE_EXTREME
  629. extension list " "
  630. -- buffered io
  631. WRITE_LIFE_EXTREME COLD_DATA WRITE_LIFE_EXTREME
  632. WRITE_LIFE_SHORT HOT_DATA WRITE_LIFE_SHORT
  633. WRITE_LIFE_NOT_SET WARM_DATA WRITE_LIFE_LONG
  634. WRITE_LIFE_NONE " "
  635. WRITE_LIFE_MEDIUM " "
  636. WRITE_LIFE_LONG " "
  637. -- direct io
  638. WRITE_LIFE_EXTREME COLD_DATA WRITE_LIFE_EXTREME
  639. WRITE_LIFE_SHORT HOT_DATA WRITE_LIFE_SHORT
  640. WRITE_LIFE_NOT_SET WARM_DATA WRITE_LIFE_NOT_SET
  641. WRITE_LIFE_NONE " WRITE_LIFE_NONE
  642. WRITE_LIFE_MEDIUM " WRITE_LIFE_MEDIUM
  643. WRITE_LIFE_LONG " WRITE_LIFE_LONG
  644. ===================== ======================== ===================
  645. Fallocate(2) Policy
  646. -------------------
  647. The default policy follows the below POSIX rule.
  648. Allocating disk space
  649. The default operation (i.e., mode is zero) of fallocate() allocates
  650. the disk space within the range specified by offset and len. The
  651. file size (as reported by stat(2)) will be changed if offset+len is
  652. greater than the file size. Any subregion within the range specified
  653. by offset and len that did not contain data before the call will be
  654. initialized to zero. This default behavior closely resembles the
  655. behavior of the posix_fallocate(3) library function, and is intended
  656. as a method of optimally implementing that function.
  657. However, once F2FS receives ioctl(fd, F2FS_IOC_SET_PIN_FILE) in prior to
  658. fallocate(fd, DEFAULT_MODE), it allocates on-disk block addressess having
  659. zero or random data, which is useful to the below scenario where:
  660. 1. create(fd)
  661. 2. ioctl(fd, F2FS_IOC_SET_PIN_FILE)
  662. 3. fallocate(fd, 0, 0, size)
  663. 4. address = fibmap(fd, offset)
  664. 5. open(blkdev)
  665. 6. write(blkdev, address)
  666. Compression implementation
  667. --------------------------
  668. - New term named cluster is defined as basic unit of compression, file can
  669. be divided into multiple clusters logically. One cluster includes 4 << n
  670. (n >= 0) logical pages, compression size is also cluster size, each of
  671. cluster can be compressed or not.
  672. - In cluster metadata layout, one special block address is used to indicate
  673. a cluster is a compressed one or normal one; for compressed cluster, following
  674. metadata maps cluster to [1, 4 << n - 1] physical blocks, in where f2fs
  675. stores data including compress header and compressed data.
  676. - In order to eliminate write amplification during overwrite, F2FS only
  677. support compression on write-once file, data can be compressed only when
  678. all logical blocks in cluster contain valid data and compress ratio of
  679. cluster data is lower than specified threshold.
  680. - To enable compression on regular inode, there are three ways:
  681. * chattr +c file
  682. * chattr +c dir; touch dir/file
  683. * mount w/ -o compress_extension=ext; touch file.ext
  684. * mount w/ -o compress_extension=*; touch any_file
  685. - At this point, compression feature doesn't expose compressed space to user
  686. directly in order to guarantee potential data updates later to the space.
  687. Instead, the main goal is to reduce data writes to flash disk as much as
  688. possible, resulting in extending disk life time as well as relaxing IO
  689. congestion. Alternatively, we've added ioctl interface to reclaim compressed
  690. space and show it to user after putting the immutable bit.
  691. Compress metadata layout::
  692. [Dnode Structure]
  693. +-----------------------------------------------+
  694. | cluster 1 | cluster 2 | ......... | cluster N |
  695. +-----------------------------------------------+
  696. . . . .
  697. . . . .
  698. . Compressed Cluster . . Normal Cluster .
  699. +----------+---------+---------+---------+ +---------+---------+---------+---------+
  700. |compr flag| block 1 | block 2 | block 3 | | block 1 | block 2 | block 3 | block 4 |
  701. +----------+---------+---------+---------+ +---------+---------+---------+---------+
  702. . .
  703. . .
  704. . .
  705. +-------------+-------------+----------+----------------------------+
  706. | data length | data chksum | reserved | compressed data |
  707. +-------------+-------------+----------+----------------------------+
  708. Compression mode
  709. --------------------------
  710. f2fs supports "fs" and "user" compression modes with "compression_mode" mount option.
  711. With this option, f2fs provides a choice to select the way how to compress the
  712. compression enabled files (refer to "Compression implementation" section for how to
  713. enable compression on a regular inode).
  714. 1) compress_mode=fs
  715. This is the default option. f2fs does automatic compression in the writeback of the
  716. compression enabled files.
  717. 2) compress_mode=user
  718. This disables the automatic compression and gives the user discretion of choosing the
  719. target file and the timing. The user can do manual compression/decompression on the
  720. compression enabled files using F2FS_IOC_DECOMPRESS_FILE and F2FS_IOC_COMPRESS_FILE
  721. ioctls like the below.
  722. To decompress a file,
  723. fd = open(filename, O_WRONLY, 0);
  724. ret = ioctl(fd, F2FS_IOC_DECOMPRESS_FILE);
  725. To compress a file,
  726. fd = open(filename, O_WRONLY, 0);
  727. ret = ioctl(fd, F2FS_IOC_COMPRESS_FILE);
  728. NVMe Zoned Namespace devices
  729. ----------------------------
  730. - ZNS defines a per-zone capacity which can be equal or less than the
  731. zone-size. Zone-capacity is the number of usable blocks in the zone.
  732. F2FS checks if zone-capacity is less than zone-size, if it is, then any
  733. segment which starts after the zone-capacity is marked as not-free in
  734. the free segment bitmap at initial mount time. These segments are marked
  735. as permanently used so they are not allocated for writes and
  736. consequently are not needed to be garbage collected. In case the
  737. zone-capacity is not aligned to default segment size(2MB), then a segment
  738. can start before the zone-capacity and span across zone-capacity boundary.
  739. Such spanning segments are also considered as usable segments. All blocks
  740. past the zone-capacity are considered unusable in these segments.