xref: /linux/Documentation/filesystems/f2fs.rst (revision 805185b7c7a1069e407b6f7b3bc98e44d415f484)
1.. SPDX-License-Identifier: GPL-2.0
2
3=================================
4Flash-Friendly File System (F2FS)
5=================================
6
7Overview
8========
9
10NAND flash memory-based storage devices, such as SSD, eMMC, and SD cards, have
11been equipped on a variety systems ranging from mobile to server systems. Since
12they are known to have different characteristics from the conventional rotating
13disks, a file system, an upper layer to the storage device, should adapt to the
14changes from the sketch in the design level.
15
16F2FS is a file system exploiting NAND flash memory-based storage devices, which
17is based on Log-structured File System (LFS). The design has been focused on
18addressing the fundamental issues in LFS, which are snowball effect of wandering
19tree and high cleaning overhead.
20
21Since a NAND flash memory-based storage device shows different characteristic
22according to its internal geometry or flash memory management scheme, namely FTL,
23F2FS and its tools support various parameters not only for configuring on-disk
24layout, but also for selecting allocation and cleaning algorithms.
25
26The following git tree provides the file system formatting tool (mkfs.f2fs),
27a consistency checking tool (fsck.f2fs), and a debugging tool (dump.f2fs).
28
29- git://git.kernel.org/pub/scm/linux/kernel/git/jaegeuk/f2fs-tools.git
30
31For sending patches, please use the following mailing list:
32
33- linux-f2fs-devel@lists.sourceforge.net
34
35For reporting bugs, please use the following f2fs bug tracker link:
36
37- https://bugzilla.kernel.org/enter_bug.cgi?product=File%20System&component=f2fs
38
39Background and Design issues
40============================
41
42Log-structured File System (LFS)
43--------------------------------
44"A log-structured file system writes all modifications to disk sequentially in
45a log-like structure, thereby speeding up  both file writing and crash recovery.
46The log is the only structure on disk; it contains indexing information so that
47files can be read back from the log efficiently. In order to maintain large free
48areas on disk for fast writing, we divide  the log into segments and use a
49segment cleaner to compress the live information from heavily fragmented
50segments." from Rosenblum, M. and Ousterhout, J. K., 1992, "The design and
51implementation of a log-structured file system", ACM Trans. Computer Systems
5210, 1, 26–52.
53
54Wandering Tree Problem
55----------------------
56In LFS, when a file data is updated and written to the end of log, its direct
57pointer block is updated due to the changed location. Then the indirect pointer
58block is also updated due to the direct pointer block update. In this manner,
59the upper index structures such as inode, inode map, and checkpoint block are
60also updated recursively. This problem is called as wandering tree problem [1],
61and in order to enhance the performance, it should eliminate or relax the update
62propagation as much as possible.
63
64[1] Bityutskiy, A. 2005. JFFS3 design issues. http://www.linux-mtd.infradead.org/
65
66Cleaning Overhead
67-----------------
68Since LFS is based on out-of-place writes, it produces so many obsolete blocks
69scattered across the whole storage. In order to serve new empty log space, it
70needs to reclaim these obsolete blocks seamlessly to users. This job is called
71as a cleaning process.
72
73The process consists of three operations as follows.
74
751. A victim segment is selected through referencing segment usage table.
762. It loads parent index structures of all the data in the victim identified by
77   segment summary blocks.
783. It checks the cross-reference between the data and its parent index structure.
794. It moves valid data selectively.
80
81This cleaning job may cause unexpected long delays, so the most important goal
82is to hide the latencies to users. And also definitely, it should reduce the
83amount of valid data to be moved, and move them quickly as well.
84
85Key Features
86============
87
88Flash Awareness
89---------------
90- Enlarge the random write area for better performance, but provide the high
91  spatial locality
92- Align FS data structures to the operational units in FTL as best efforts
93
94Wandering Tree Problem
95----------------------
96- Use a term, “node”, that represents inodes as well as various pointer blocks
97- Introduce Node Address Table (NAT) containing the locations of all the “node”
98  blocks; this will cut off the update propagation.
99
100Cleaning Overhead
101-----------------
102- Support a background cleaning process
103- Support greedy and cost-benefit algorithms for victim selection policies
104- Support multi-head logs for static/dynamic hot and cold data separation
105- Introduce adaptive logging for efficient block allocation
106
107Mount Options
108=============
109
110
111======================== ============================================================
112background_gc=%s	 Turn on/off cleaning operations, namely garbage
113			 collection, triggered in background when I/O subsystem is
114			 idle. If background_gc=on, it will turn on the garbage
115			 collection and if background_gc=off, garbage collection
116			 will be turned off. If background_gc=sync, it will turn
117			 on synchronous garbage collection running in background.
118			 Default value for this option is on. So garbage
119			 collection is on by default.
120gc_merge		 When background_gc is on, this option can be enabled to
121			 let background GC thread to handle foreground GC requests,
122			 it can eliminate the sluggish issue caused by slow foreground
123			 GC operation when GC is triggered from a process with limited
124			 I/O and CPU resources.
125nogc_merge		 Disable GC merge feature.
126disable_roll_forward	 Disable the roll-forward recovery routine
127norecovery		 Disable the roll-forward recovery routine, mounted read-
128			 only (i.e., -o ro,disable_roll_forward)
129discard/nodiscard	 Enable/disable real-time discard in f2fs, if discard is
130			 enabled, f2fs will issue discard/TRIM commands when a
131			 segment is cleaned.
132heap/no_heap		 Deprecated.
133nouser_xattr		 Disable Extended User Attributes. Note: xattr is enabled
134			 by default if CONFIG_F2FS_FS_XATTR is selected.
135noacl			 Disable POSIX Access Control List. Note: acl is enabled
136			 by default if CONFIG_F2FS_FS_POSIX_ACL is selected.
137active_logs=%u		 Support configuring the number of active logs. In the
138			 current design, f2fs supports only 2, 4, and 6 logs.
139			 Default number is 6.
140			 When the underlying block device exposes write
141			 streams, the default active_logs=6 configuration
142			 maps hot, warm, and cold DATA writes to streams 1,
143			 2, and 3, respectively. If only one or two write
144			 streams are available, f2fs falls back to mapping
145			 all DATA writes to stream 1 or mapping hot/warm
146			 to stream 1 and cold to stream 2. If no write
147			 streams are exposed, f2fs leaves the stream
148			 unset.
149disable_ext_identify	 Disable the extension list configured by mkfs, so f2fs
150			 is not aware of cold files such as media files.
151inline_xattr		 Enable the inline xattrs feature.
152noinline_xattr		 Disable the inline xattrs feature.
153inline_xattr_size=%u	 Support configuring inline xattr size, it depends on
154			 flexible inline xattr feature.
155inline_data		 Enable the inline data feature: Newly created small (<~3.4k)
156			 files can be written into inode block.
157inline_dentry		 Enable the inline dir feature: data in newly created
158			 directory entries can be written into inode block. The
159			 space of inode block which is used to store inline
160			 dentries is limited to ~3.4k.
161noinline_dentry		 Disable the inline dentry feature.
162flush_merge		 Merge concurrent cache_flush commands as much as possible
163			 to eliminate redundant command issues. If the underlying
164			 device handles the cache_flush command relatively slowly,
165			 recommend to enable this option.
166nobarrier		 This option can be used if underlying storage guarantees
167			 its cached data should be written to the novolatile area.
168			 If this option is set, no cache_flush commands are issued
169			 but f2fs still guarantees the write ordering of all the
170			 data writes.
171barrier			 If this option is set, cache_flush commands are allowed to be
172			 issued.
173fastboot		 This option is used when a system wants to reduce mount
174			 time as much as possible, even though normal performance
175			 can be sacrificed.
176extent_cache		 Enable an extent cache based on rb-tree, it can cache
177			 as many as extent which map between contiguous logical
178			 address and physical address per inode, resulting in
179			 increasing the cache hit ratio. Set by default.
180noextent_cache		 Disable an extent cache based on rb-tree explicitly, see
181			 the above extent_cache mount option.
182noinline_data		 Disable the inline data feature, inline data feature is
183			 enabled by default.
184data_flush		 Enable data flushing before checkpoint in order to
185			 persist data of regular and symlink.
186reserve_root=%d		 Support configuring reserved space which is used for
187			 allocation from a privileged user with specified uid or
188			 gid, unit: 4KB, the default limit is 12.5% of user blocks.
189reserve_node=%d		 Support configuring reserved nodes which are used for
190			 allocation from a privileged user with specified uid or
191			 gid, the default limit is 12.5% of all nodes.
192resuid=%d		 The user ID which may use the reserved blocks and nodes.
193resgid=%d		 The group ID which may use the reserved blocks and nodes.
194fault_injection=%d	 Enable fault injection in all supported types with
195			 specified injection rate.
196fault_type=%d		 Support configuring fault injection type, should be
197			 enabled with fault_injection option, fault type value
198			 is shown below, it supports single or combined type.
199
200			 .. code-block:: none
201
202			     ===========================      ==========
203			     Type_Name                        Type_Value
204			     ===========================      ==========
205			     FAULT_KMALLOC                    0x00000001
206			     FAULT_KVMALLOC                   0x00000002
207			     FAULT_PAGE_ALLOC                 0x00000004
208			     FAULT_PAGE_GET                   0x00000008
209			     FAULT_ALLOC_BIO                  0x00000010 (obsolete)
210			     FAULT_ALLOC_NID                  0x00000020
211			     FAULT_ORPHAN                     0x00000040
212			     FAULT_BLOCK                      0x00000080
213			     FAULT_DIR_DEPTH                  0x00000100
214			     FAULT_EVICT_INODE                0x00000200
215			     FAULT_TRUNCATE                   0x00000400
216			     FAULT_READ_IO                    0x00000800
217			     FAULT_CHECKPOINT                 0x00001000
218			     FAULT_DISCARD                    0x00002000 (obsolete)
219			     FAULT_WRITE_IO                   0x00004000
220			     FAULT_SLAB_ALLOC                 0x00008000
221			     FAULT_DQUOT_INIT                 0x00010000
222			     FAULT_LOCK_OP                    0x00020000
223			     FAULT_BLKADDR_VALIDITY           0x00040000
224			     FAULT_BLKADDR_CONSISTENCE        0x00080000
225			     FAULT_NO_SEGMENT                 0x00100000
226			     FAULT_INCONSISTENT_FOOTER        0x00200000
227			     FAULT_ATOMIC_TIMEOUT             0x00400000 (1000ms)
228			     FAULT_VMALLOC                    0x00800000
229			     FAULT_LOCK_TIMEOUT               0x01000000 (1000ms)
230			     FAULT_SKIP_WRITE                 0x02000000
231			     ===========================      ==========
232mode=%s			 Control block allocation mode which supports "adaptive"
233			 and "lfs". In "lfs" mode, there should be no random
234			 writes towards main area.
235			 "fragment:segment" and "fragment:block" are newly added here.
236			 These are developer options for experiments to simulate filesystem
237			 fragmentation/after-GC situation itself. The developers use these
238			 modes to understand filesystem fragmentation/after-GC condition well,
239			 and eventually get some insights to handle them better.
240			 In "fragment:segment", f2fs allocates a new segment in random
241			 position. With this, we can simulate the after-GC condition.
242			 In "fragment:block", we can scatter block allocation with
243			 "max_fragment_chunk" and "max_fragment_hole" sysfs nodes.
244			 We added some randomness to both chunk and hole size to make
245			 it close to realistic IO pattern. So, in this mode, f2fs will allocate
246			 1..<max_fragment_chunk> blocks in a chunk and make a hole in the
247			 length of 1..<max_fragment_hole> by turns. With this, the newly
248			 allocated blocks will be scattered throughout the whole partition.
249			 Note that "fragment:block" implicitly enables "fragment:segment"
250			 option for more randomness.
251			 Please, use these options for your experiments and we strongly
252			 recommend to re-format the filesystem after using these options.
253usrquota		 Enable plain user disk quota accounting.
254grpquota		 Enable plain group disk quota accounting.
255prjquota		 Enable plain project quota accounting.
256usrjquota=<file>	 Appoint specified file and type during mount, so that quota
257grpjquota=<file>	 information can be properly updated during recovery flow,
258prjjquota=<file>	 <quota file>: must be in root directory;
259jqfmt=<quota type>	 <quota type>: [vfsold,vfsv0,vfsv1].
260usrjquota=		 Turn off user journalled quota.
261grpjquota=		 Turn off group journalled quota.
262prjjquota=		 Turn off project journalled quota.
263quota			 Enable plain user disk quota accounting.
264noquota			 Disable all plain disk quota option.
265alloc_mode=%s		 Adjust block allocation policy, which supports "reuse"
266			 and "default".
267fsync_mode=%s		 Control the policy of fsync. Currently supports "posix",
268			 "strict", and "nobarrier". In "posix" mode, which is
269			 default, fsync will follow POSIX semantics and does a
270			 light operation to improve the filesystem performance.
271			 In "strict" mode, fsync will be heavy and behaves in line
272			 with xfs, ext4 and btrfs, where xfstest generic/342 will
273			 pass, but the performance will regress. "nobarrier" is
274			 based on "posix", but doesn't issue flush command for
275			 non-atomic files likewise "nobarrier" mount option.
276test_dummy_encryption
277test_dummy_encryption=%s
278			 Enable dummy encryption, which provides a fake fscrypt
279			 context. The fake fscrypt context is used by xfstests.
280			 The argument may be either "v1" or "v2", in order to
281			 select the corresponding fscrypt policy version.
282checkpoint=%s[:%u[%]]	 Set to "disable" to turn off checkpointing. Set to "enable"
283			 to re-enable checkpointing. Is enabled by default. While
284			 disabled, any unmounting or unexpected shutdowns will cause
285			 the filesystem contents to appear as they did when the
286			 filesystem was mounted with that option.
287			 While mounting with checkpoint=disable, the filesystem must
288			 run garbage collection to ensure that all available space can
289			 be used. If this takes too much time, the mount may return
290			 EAGAIN. You may optionally add a value to indicate how much
291			 of the disk you would be willing to temporarily give up to
292			 avoid additional garbage collection. This can be given as a
293			 number of blocks, or as a percent. For instance, mounting
294			 with checkpoint=disable:100% would always succeed, but it may
295			 hide up to all remaining free space. The actual space that
296			 would be unusable can be viewed at /sys/fs/f2fs/<disk>/unusable
297			 This space is reclaimed once checkpoint=enable.
298checkpoint_merge	 When checkpoint is enabled, this can be used to create a kernel
299			 daemon and make it to merge concurrent checkpoint requests as
300			 much as possible to eliminate redundant checkpoint issues. Plus,
301			 we can eliminate the sluggish issue caused by slow checkpoint
302			 operation when the checkpoint is done in a process context in
303			 a cgroup having low i/o budget and cpu shares. To make this
304			 do better, we set the default i/o priority of the kernel daemon
305			 to "3", to give one higher priority than other kernel threads.
306			 This is the same way to give a I/O priority to the jbd2
307			 journaling thread of ext4 filesystem.
308nocheckpoint_merge	 Disable checkpoint merge feature.
309compress_algorithm=%s	 Control compress algorithm, currently f2fs supports "lzo",
310			 "lz4", "zstd" and "lzo-rle" algorithm.
311compress_algorithm=%s:%d Control compress algorithm and its compress level, now, only
312			 "lz4" and "zstd" support compress level config::
313
314				 =========      ===========
315				 algorithm      level range
316				 =========      ===========
317				 lz4            3 - 16
318				 zstd           1 - 22
319				 =========      ===========
320
321compress_log_size=%u	 Support configuring compress cluster size. The size will
322			 be 4KB * (1 << %u). The default and minimum sizes are 16KB.
323compress_extension=%s	 Support adding specified extension, so that f2fs can enable
324			 compression on those corresponding files, e.g. if all files
325			 with '.ext' has high compression rate, we can set the '.ext'
326			 on compression extension list and enable compression on
327			 these file by default rather than to enable it via ioctl.
328			 For other files, we can still enable compression via ioctl.
329			 Note that, there is one reserved special extension '*', it
330			 can be set to enable compression for all files.
331nocompress_extension=%s	 Support adding specified extension, so that f2fs can disable
332			 compression on those corresponding files, just contrary to compression extension.
333			 If you know exactly which files cannot be compressed, you can use this.
334			 The same extension name can't appear in both compress and nocompress
335			 extension at the same time.
336			 If the compress extension specifies all files, the types specified by the
337			 nocompress extension will be treated as special cases and will not be compressed.
338			 Don't allow use '*' to specifie all file in nocompress extension.
339			 After add nocompress_extension, the priority should be:
340			 dir_flag < comp_extention,nocompress_extension < comp_file_flag,no_comp_file_flag.
341			 See more in compression sections.
342
343compress_chksum		 Support verifying chksum of raw data in compressed cluster.
344compress_mode=%s	 Control file compression mode. This supports "fs" and "user"
345			 modes. In "fs" mode (default), f2fs does automatic compression
346			 on the compression enabled files. In "user" mode, f2fs disables
347			 the automaic compression and gives the user discretion of
348			 choosing the target file and the timing. The user can do manual
349			 compression/decompression on the compression enabled files using
350			 ioctls.
351compress_cache		 Support to use address space of a filesystem managed inode to
352			 cache compressed block, in order to improve cache hit ratio of
353			 random read.
354inlinecrypt		 When possible, encrypt/decrypt the contents of encrypted
355			 files using the blk-crypto framework rather than
356			 filesystem-layer encryption. This allows the use of
357			 inline encryption hardware. The on-disk format is
358			 unaffected. For more details, see
359			 Documentation/block/inline-encryption.rst.
360atgc			 Enable age-threshold garbage collection, it provides high
361			 effectiveness and efficiency on background GC.
362discard_unit=%s		 Control discard unit, the argument can be "block", "segment"
363			 and "section", issued discard command's offset/size will be
364			 aligned to the unit, by default, "discard_unit=block" is set,
365			 so that small discard functionality is enabled.
366			 For blkzoned device, "discard_unit=section" will be set by
367			 default, it is helpful for large sized SMR or ZNS devices to
368			 reduce memory cost by getting rid of fs metadata supports small
369			 discard.
370memory=%s		 Control memory mode. This supports "normal" and "low" modes.
371			 "low" mode is introduced to support low memory devices.
372			 Because of the nature of low memory devices, in this mode, f2fs
373			 will try to save memory sometimes by sacrificing performance.
374			 "normal" mode is the default mode and same as before.
375age_extent_cache	 Enable an age extent cache based on rb-tree. It records
376			 data block update frequency of the extent per inode, in
377			 order to provide better temperature hints for data block
378			 allocation.
379errors=%s		 Specify f2fs behavior on critical errors. This supports modes:
380			 "panic", "continue" and "remount-ro", respectively, trigger
381			 panic immediately, continue without doing anything, and remount
382			 the partition in read-only mode. By default it uses "continue"
383			 mode.
384
385			 .. code-block:: none
386
387			     ====================== =============== =============== ========
388			     mode                   continue        remount-ro      panic
389			     ====================== =============== =============== ========
390			     access ops             normal          normal          N/A
391			     syscall errors         -EIO            -EROFS          N/A
392			     mount option           rw              ro              N/A
393			     pending dir write      keep            keep            N/A
394			     pending non-dir write  drop            keep            N/A
395			     pending node write     drop            keep            N/A
396			     pending meta write     keep            keep            N/A
397			     ====================== =============== =============== ========
398nat_bits		 Enable nat_bits feature to enhance full/empty nat blocks access,
399			 by default it's disabled.
400lookup_mode=%s		 Control the directory lookup behavior for casefolded
401			 directories. This option has no effect on directories
402			 that do not have the casefold feature enabled.
403
404			 .. code-block:: none
405
406			     ================== ========================================
407			     Value              Description
408			     ================== ========================================
409			     perf               (Default) Enforces a hash-only lookup.
410					        The linear search fallback is always
411					        disabled, ignoring the on-disk flag.
412			     compat             Enables the linear search fallback for
413					        compatibility with directory entries
414					        created by older kernel that used a
415					        different case-folding algorithm.
416					        This mode ignores the on-disk flag.
417			     auto               F2FS determines the mode based on the
418					        on-disk `SB_ENC_NO_COMPAT_FALLBACK_FL`
419					        flag.
420			     ================== ========================================
421======================== ============================================================
422
423Debugfs Entries
424===============
425
426/sys/kernel/debug/f2fs/ contains information about all the partitions mounted as
427f2fs. Each file shows the whole f2fs information.
428
429/sys/kernel/debug/f2fs/status includes:
430
431 - major file system information managed by f2fs currently
432 - average SIT information about whole segments
433 - current memory footprint consumed by f2fs.
434
435Sysfs Entries
436=============
437
438Information about mounted f2fs file systems can be found in
439/sys/fs/f2fs.  Each mounted filesystem will have a directory in
440/sys/fs/f2fs based on its device name (i.e., /sys/fs/f2fs/sda).
441The files in each per-device directory are shown in table below.
442
443Files in /sys/fs/f2fs/<devname>
444(see also Documentation/ABI/testing/sysfs-fs-f2fs)
445
446Usage
447=====
448
4491. Download userland tools and compile them.
450
4512. Skip, if f2fs was compiled statically inside kernel.
452   Otherwise, insert the f2fs.ko module::
453
454	# insmod f2fs.ko
455
4563. Create a directory to use when mounting::
457
458	# mkdir /mnt/f2fs
459
4604. Format the block device, and then mount as f2fs::
461
462	# mkfs.f2fs -l label /dev/block_device
463	# mount -t f2fs /dev/block_device /mnt/f2fs
464
465mkfs.f2fs
466---------
467The mkfs.f2fs is for the use of formatting a partition as the f2fs filesystem,
468which builds a basic on-disk layout.
469
470The quick options consist of:
471
472===============    ===========================================================
473``-l [label]``     Give a volume label, up to 512 unicode name.
474``-a [0 or 1]``    Split start location of each area for heap-based allocation.
475
476                   1 is set by default, which performs this.
477``-o [int]``       Set overprovision ratio in percent over volume size.
478
479                   5 is set by default.
480``-s [int]``       Set the number of segments per section.
481
482                   1 is set by default.
483``-z [int]``       Set the number of sections per zone.
484
485                   1 is set by default.
486``-e [str]``       Set basic extension list. e.g. "mp3,gif,mov"
487``-t [0 or 1]``    Disable discard command or not.
488
489                   1 is set by default, which conducts discard.
490===============    ===========================================================
491
492Note: please refer to the manpage of mkfs.f2fs(8) to get full option list.
493
494fsck.f2fs
495---------
496The fsck.f2fs is a tool to check the consistency of an f2fs-formatted
497partition, which examines whether the filesystem metadata and user-made data
498are cross-referenced correctly or not.
499Note that, initial version of the tool does not fix any inconsistency.
500
501The quick options consist of::
502
503  -d debug level [default:0]
504
505Note: please refer to the manpage of fsck.f2fs(8) to get full option list.
506
507dump.f2fs
508---------
509The dump.f2fs shows the information of specific inode and dumps SSA and SIT to
510file. Each file is dump_ssa and dump_sit.
511
512The dump.f2fs is used to debug on-disk data structures of the f2fs filesystem.
513It shows on-disk inode information recognized by a given inode number, and is
514able to dump all the SSA and SIT entries into predefined files, ./dump_ssa and
515./dump_sit respectively.
516
517The options consist of::
518
519  -d debug level [default:0]
520  -i inode no (hex)
521  -s [SIT dump segno from #1~#2 (decimal), for all 0~-1]
522  -a [SSA dump segno from #1~#2 (decimal), for all 0~-1]
523
524Examples::
525
526    # dump.f2fs -i [ino] /dev/sdx
527    # dump.f2fs -s 0~-1 /dev/sdx (SIT dump)
528    # dump.f2fs -a 0~-1 /dev/sdx (SSA dump)
529
530Note: please refer to the manpage of dump.f2fs(8) to get full option list.
531
532sload.f2fs
533----------
534The sload.f2fs gives a way to insert files and directories in the existing disk
535image. This tool is useful when building f2fs images given compiled files.
536
537Note: please refer to the manpage of sload.f2fs(8) to get full option list.
538
539resize.f2fs
540-----------
541The resize.f2fs lets a user resize the f2fs-formatted disk image, while preserving
542all the files and directories stored in the image.
543
544Note: please refer to the manpage of resize.f2fs(8) to get full option list.
545
546defrag.f2fs
547-----------
548The defrag.f2fs can be used to defragment scattered written data as well as
549filesystem metadata across the disk. This can improve the write speed by giving
550more free consecutive space.
551
552Note: please refer to the manpage of defrag.f2fs(8) to get full option list.
553
554f2fs_io
555-------
556The f2fs_io is a simple tool to issue various filesystem APIs as well as
557f2fs-specific ones, which is very useful for QA tests.
558
559Note: please refer to the manpage of f2fs_io(8) to get full option list.
560
561Design
562======
563
564On-disk Layout
565--------------
566
567F2FS divides the whole volume into a number of segments, each of which is fixed
568to 2MB in size. A section is composed of consecutive segments, and a zone
569consists of a set of sections. By default, section and zone sizes are set to one
570segment size identically, but users can easily modify the sizes by mkfs.
571
572F2FS splits the entire volume into six areas, and all the areas except superblock
573consist of multiple segments as described below::
574
575                                            align with the zone size <-|
576                 |-> align with the segment size
577     _________________________________________________________________________
578    |            |            |   Segment   |    Node     |   Segment  |      |
579    | Superblock | Checkpoint |    Info.    |   Address   |   Summary  | Main |
580    |    (SB)    |   (CP)     | Table (SIT) | Table (NAT) | Area (SSA) |      |
581    |____________|_____2______|______N______|______N______|______N_____|__N___|
582                                                                       .      .
583                                                             .                .
584                                                 .                            .
585                                    ._________________________________________.
586                                    |_Segment_|_..._|_Segment_|_..._|_Segment_|
587                                    .           .
588                                    ._________._________
589                                    |_section_|__...__|_
590                                    .            .
591		                    .________.
592	                            |__zone__|
593
594- Superblock (SB)
595   It is located at the beginning of the partition, and there exist two copies
596   to avoid file system crash. It contains basic partition information and some
597   default parameters of f2fs.
598
599- Checkpoint (CP)
600   It contains file system information, bitmaps for valid NAT/SIT sets, orphan
601   inode lists, and summary entries of current active segments.
602
603- Segment Information Table (SIT)
604   It contains segment information such as valid block count and bitmap for the
605   validity of all the blocks.
606
607- Node Address Table (NAT)
608   It is composed of a block address table for all the node blocks stored in
609   Main area.
610
611- Segment Summary Area (SSA)
612   It contains summary entries which contains the owner information of all the
613   data and node blocks stored in Main area.
614
615- Main Area
616   It contains file and directory data including their indices.
617
618In order to avoid misalignment between file system and flash-based storage, F2FS
619aligns the start block address of CP with the segment size. Also, it aligns the
620start block address of Main area with the zone size by reserving some segments
621in SSA area.
622
623Reference the following survey for additional technical details.
624https://wiki.linaro.org/WorkingGroups/Kernel/Projects/FlashCardSurvey
625
626File System Metadata Structure
627------------------------------
628
629F2FS adopts the checkpointing scheme to maintain file system consistency. At
630mount time, F2FS first tries to find the last valid checkpoint data by scanning
631CP area. In order to reduce the scanning time, F2FS uses only two copies of CP.
632One of them always indicates the last valid data, which is called as shadow copy
633mechanism. In addition to CP, NAT and SIT also adopt the shadow copy mechanism.
634
635For file system consistency, each CP points to which NAT and SIT copies are
636valid, as shown as below::
637
638  +--------+----------+---------+
639  |   CP   |    SIT   |   NAT   |
640  +--------+----------+---------+
641  .         .          .          .
642  .            .              .              .
643  .               .                 .                 .
644  +-------+-------+--------+--------+--------+--------+
645  | CP #0 | CP #1 | SIT #0 | SIT #1 | NAT #0 | NAT #1 |
646  +-------+-------+--------+--------+--------+--------+
647     |             ^                          ^
648     |             |                          |
649     `----------------------------------------'
650
651Index Structure
652---------------
653
654The key data structure to manage the data locations is a "node". Similar to
655traditional file structures, F2FS has three types of node: inode, direct node,
656indirect node. F2FS assigns 4KB to an inode block which contains 923 data block
657indices, two direct node pointers, two indirect node pointers, and one double
658indirect node pointer as described below. One direct node block contains 1018
659data blocks, and one indirect node block contains also 1018 node blocks. Thus,
660one inode block (i.e., a file) covers::
661
662  4KB * (923 + 2 * 1018 + 2 * 1018 * 1018 + 1018 * 1018 * 1018) := 3.94TB.
663
664   Inode block (4KB)
665     |- data (923)
666     |- direct node (2)
667     |          `- data (1018)
668     |- indirect node (2)
669     |            `- direct node (1018)
670     |                       `- data (1018)
671     `- double indirect node (1)
672                         `- indirect node (1018)
673			              `- direct node (1018)
674	                                         `- data (1018)
675
676Note that all the node blocks are mapped by NAT which means the location of
677each node is translated by the NAT table. In the consideration of the wandering
678tree problem, F2FS is able to cut off the propagation of node updates caused by
679leaf data writes.
680
681Directory Structure
682-------------------
683
684A directory entry occupies 11 bytes, which consists of the following attributes.
685
686- hash		hash value of the file name
687- ino		inode number
688- len		the length of file name
689- type		file type such as directory, symlink, etc
690
691A dentry block consists of 214 dentry slots and file names. Therein a bitmap is
692used to represent whether each dentry is valid or not. A dentry block occupies
6934KB with the following composition.
694
695::
696
697  Dentry Block(4 K) = bitmap (27 bytes) + reserved (3 bytes) +
698	              dentries(11 * 214 bytes) + file name (8 * 214 bytes)
699
700                         [Bucket]
701             +--------------------------------+
702             |dentry block 1 | dentry block 2 |
703             +--------------------------------+
704             .               .
705       .                             .
706  .       [Dentry Block Structure: 4KB]       .
707  +--------+----------+----------+------------+
708  | bitmap | reserved | dentries | file names |
709  +--------+----------+----------+------------+
710  [Dentry Block: 4KB] .   .
711		 .               .
712            .                          .
713            +------+------+-----+------+
714            | hash | ino  | len | type |
715            +------+------+-----+------+
716            [Dentry Structure: 11 bytes]
717
718F2FS implements multi-level hash tables for directory structure. Each level has
719a hash table with dedicated number of hash buckets as shown below. Note that
720"A(2B)" means a bucket includes 2 data blocks.
721
722::
723
724    ----------------------
725    A : bucket
726    B : block
727    N : MAX_DIR_HASH_DEPTH
728    ----------------------
729
730    level #0   | A(2B)
731	    |
732    level #1   | A(2B) - A(2B)
733	    |
734    level #2   | A(2B) - A(2B) - A(2B) - A(2B)
735	.     |   .       .       .       .
736    level #N/2 | A(2B) - A(2B) - A(2B) - A(2B) - A(2B) - ... - A(2B)
737	.     |   .       .       .       .
738    level #N   | A(4B) - A(4B) - A(4B) - A(4B) - A(4B) - ... - A(4B)
739
740The number of blocks and buckets are determined by::
741
742                            ,- 2, if n < MAX_DIR_HASH_DEPTH / 2,
743  # of blocks in level #n = |
744                            `- 4, Otherwise
745
746                             ,- 2^(n + dir_level),
747			     |        if n + dir_level < MAX_DIR_HASH_DEPTH / 2,
748  # of buckets in level #n = |
749                             `- 2^((MAX_DIR_HASH_DEPTH / 2) - 1),
750			              Otherwise
751
752When F2FS finds a file name in a directory, at first a hash value of the file
753name is calculated. Then, F2FS scans the hash table in level #0 to find the
754dentry consisting of the file name and its inode number. If not found, F2FS
755scans the next hash table in level #1. In this way, F2FS scans hash tables in
756each levels incrementally from 1 to N. In each level F2FS needs to scan only
757one bucket determined by the following equation, which shows O(log(# of files))
758complexity::
759
760  bucket number to scan in level #n = (hash value) % (# of buckets in level #n)
761
762In the case of file creation, F2FS finds empty consecutive slots that cover the
763file name. F2FS searches the empty slots in the hash tables of whole levels from
7641 to N in the same way as the lookup operation.
765
766The following figure shows an example of two cases holding children::
767
768       --------------> Dir <--------------
769       |                                 |
770    child                             child
771
772    child - child                     [hole] - child
773
774    child - child - child             [hole] - [hole] - child
775
776   Case 1:                           Case 2:
777   Number of children = 6,           Number of children = 3,
778   File size = 7                     File size = 7
779
780Default Block Allocation
781------------------------
782
783At runtime, F2FS manages six active logs inside "Main" area: Hot/Warm/Cold node
784and Hot/Warm/Cold data.
785
786- Hot node	contains direct node blocks of directories.
787- Warm node	contains direct node blocks except hot node blocks.
788- Cold node	contains indirect node blocks
789- Hot data	contains dentry blocks
790- Warm data	contains data blocks except hot and cold data blocks
791- Cold data	contains multimedia data or migrated data blocks
792
793LFS has two schemes for free space management: threaded log and copy-and-compac-
794tion. The copy-and-compaction scheme which is known as cleaning, is well-suited
795for devices showing very good sequential write performance, since free segments
796are served all the time for writing new data. However, it suffers from cleaning
797overhead under high utilization. Contrarily, the threaded log scheme suffers
798from random writes, but no cleaning process is needed. F2FS adopts a hybrid
799scheme where the copy-and-compaction scheme is adopted by default, but the
800policy is dynamically changed to the threaded log scheme according to the file
801system status.
802
803In order to align F2FS with underlying flash-based storage, F2FS allocates a
804segment in a unit of section. F2FS expects that the section size would be the
805same as the unit size of garbage collection in FTL. Furthermore, with respect
806to the mapping granularity in FTL, F2FS allocates each section of the active
807logs from different zones as much as possible, since FTL can write the data in
808the active logs into one allocation unit according to its mapping granularity.
809
810Cleaning process
811----------------
812
813F2FS does cleaning both on demand and in the background. On-demand cleaning is
814triggered when there are not enough free segments to serve VFS calls. Background
815cleaner is operated by a kernel thread, and triggers the cleaning job when the
816system is idle.
817
818F2FS supports two victim selection policies: greedy and cost-benefit algorithms.
819In the greedy algorithm, F2FS selects a victim segment having the smallest number
820of valid blocks. In the cost-benefit algorithm, F2FS selects a victim segment
821according to the segment age and the number of valid blocks in order to address
822log block thrashing problem in the greedy algorithm. F2FS adopts the greedy
823algorithm for on-demand cleaner, while background cleaner adopts cost-benefit
824algorithm.
825
826In order to identify whether the data in the victim segment are valid or not,
827F2FS manages a bitmap. Each bit represents the validity of a block, and the
828bitmap is composed of a bit stream covering whole blocks in main area.
829
830Write-hint Policy
831-----------------
832
833F2FS sets the whint all the time with the below policy.
834
835===================== ======================== ===================
836User                  F2FS                     Block
837===================== ======================== ===================
838N/A                   META                     WRITE_LIFE_NONE|REQ_META
839N/A                   HOT_NODE                 WRITE_LIFE_NONE
840N/A                   WARM_NODE                WRITE_LIFE_MEDIUM
841N/A                   COLD_NODE                WRITE_LIFE_LONG
842ioctl(COLD)           COLD_DATA                WRITE_LIFE_EXTREME
843extension list        "                        "
844
845-- buffered io
846------------------------------------------------------------------
847N/A                   COLD_DATA                WRITE_LIFE_EXTREME
848N/A                   HOT_DATA                 WRITE_LIFE_SHORT
849N/A                   WARM_DATA                WRITE_LIFE_NOT_SET
850
851-- direct io
852------------------------------------------------------------------
853WRITE_LIFE_EXTREME    COLD_DATA                WRITE_LIFE_EXTREME
854WRITE_LIFE_SHORT      HOT_DATA                 WRITE_LIFE_SHORT
855WRITE_LIFE_NOT_SET    WARM_DATA                WRITE_LIFE_NOT_SET
856WRITE_LIFE_NONE       "                        WRITE_LIFE_NONE
857WRITE_LIFE_MEDIUM     "                        WRITE_LIFE_MEDIUM
858WRITE_LIFE_LONG       "                        WRITE_LIFE_LONG
859===================== ======================== ===================
860
861Fallocate(2) Policy
862-------------------
863
864The default policy follows the below POSIX rule.
865
866Allocating disk space
867    The default operation (i.e., mode is zero) of fallocate() allocates
868    the disk space within the range specified by offset and len.  The
869    file size (as reported by stat(2)) will be changed if offset+len is
870    greater than the file size.  Any subregion within the range specified
871    by offset and len that did not contain data before the call will be
872    initialized to zero.  This default behavior closely resembles the
873    behavior of the posix_fallocate(3) library function, and is intended
874    as a method of optimally implementing that function.
875
876However, once F2FS receives ioctl(fd, F2FS_IOC_SET_PIN_FILE) in prior to
877fallocate(fd, DEFAULT_MODE), it allocates on-disk block addresses having
878zero or random data, which is useful to the below scenario where:
879
880 1. create(fd)
881 2. ioctl(fd, F2FS_IOC_SET_PIN_FILE)
882 3. fallocate(fd, 0, 0, size)
883 4. address = fibmap(fd, offset)
884 5. open(blkdev)
885 6. write(blkdev, address)
886
887Compression implementation
888--------------------------
889
890- New term named cluster is defined as basic unit of compression, file can
891  be divided into multiple clusters logically. One cluster includes 4 << n
892  (n >= 0) logical pages, compression size is also cluster size, each of
893  cluster can be compressed or not.
894
895- In cluster metadata layout, one special block address is used to indicate
896  a cluster is a compressed one or normal one; for compressed cluster, following
897  metadata maps cluster to [1, 4 << n - 1] physical blocks, in where f2fs
898  stores data including compress header and compressed data.
899
900- In order to eliminate write amplification during overwrite, F2FS only
901  support compression on write-once file, data can be compressed only when
902  all logical blocks in cluster contain valid data and compress ratio of
903  cluster data is lower than specified threshold.
904
905- To enable compression on regular inode, there are four ways:
906
907  * chattr +c file
908  * chattr +c dir; touch dir/file
909  * mount w/ -o compress_extension=ext; touch file.ext
910  * mount w/ -o compress_extension=*; touch any_file
911
912- To disable compression on regular inode, there are two ways:
913
914  * chattr -c file
915  * mount w/ -o nocompress_extension=ext; touch file.ext
916
917- Priority in between FS_COMPR_FL, FS_NOCOMP_FS, extensions:
918
919  * compress_extension=so; nocompress_extension=zip; chattr +c dir; touch
920    dir/foo.so; touch dir/bar.zip; touch dir/baz.txt; then foo.so and baz.txt
921    should be compresse, bar.zip should be non-compressed. chattr +c dir/bar.zip
922    can enable compress on bar.zip.
923  * compress_extension=so; nocompress_extension=zip; chattr -c dir; touch
924    dir/foo.so; touch dir/bar.zip; touch dir/baz.txt; then foo.so should be
925    compresse, bar.zip and baz.txt should be non-compressed.
926    chattr+c dir/bar.zip; chattr+c dir/baz.txt; can enable compress on bar.zip
927    and baz.txt.
928
929- At this point, compression feature doesn't expose compressed space to user
930  directly in order to guarantee potential data updates later to the space.
931  Instead, the main goal is to reduce data writes to flash disk as much as
932  possible, resulting in extending disk life time as well as relaxing IO
933  congestion. Alternatively, we've added ioctl(F2FS_IOC_RELEASE_COMPRESS_BLOCKS)
934  interface to reclaim compressed space and show it to user after setting a
935  special flag to the inode. Once the compressed space is released, the flag
936  will block writing data to the file until either the compressed space is
937  reserved via ioctl(F2FS_IOC_RESERVE_COMPRESS_BLOCKS) or the file size is
938  truncated to zero.
939
940Compress metadata layout::
941
942				[Dnode Structure]
943		+-----------------------------------------------+
944		| cluster 1 | cluster 2 | ......... | cluster N |
945		+-----------------------------------------------+
946		.           .                       .           .
947	  .                      .                .                      .
948    .         Compressed Cluster       .        .        Normal Cluster            .
949    +----------+---------+---------+---------+  +---------+---------+---------+---------+
950    |compr flag| block 1 | block 2 | block 3 |  | block 1 | block 2 | block 3 | block 4 |
951    +----------+---------+---------+---------+  +---------+---------+---------+---------+
952	       .                             .
953	    .                                           .
954	.                                                           .
955	+-------------+-------------+----------+----------------------------+
956	| data length | data chksum | reserved |      compressed data       |
957	+-------------+-------------+----------+----------------------------+
958
959Compression mode
960--------------------------
961
962f2fs supports "fs" and "user" compression modes with "compression_mode" mount option.
963With this option, f2fs provides a choice to select the way how to compress the
964compression enabled files (refer to "Compression implementation" section for how to
965enable compression on a regular inode).
966
9671) compress_mode=fs
968
969   This is the default option. f2fs does automatic compression in the writeback of the
970   compression enabled files.
971
9722) compress_mode=user
973
974   This disables the automatic compression and gives the user discretion of choosing the
975   target file and the timing. The user can do manual compression/decompression on the
976   compression enabled files using F2FS_IOC_DECOMPRESS_FILE and F2FS_IOC_COMPRESS_FILE
977   ioctls like the below.
978
979To decompress a file::
980
981  fd = open(filename, O_WRONLY, 0);
982  ret = ioctl(fd, F2FS_IOC_DECOMPRESS_FILE);
983
984To compress a file::
985
986  fd = open(filename, O_WRONLY, 0);
987  ret = ioctl(fd, F2FS_IOC_COMPRESS_FILE);
988
989NVMe Zoned Namespace devices
990----------------------------
991
992- ZNS defines a per-zone capacity which can be equal or less than the
993  zone-size. Zone-capacity is the number of usable blocks in the zone.
994  F2FS checks if zone-capacity is less than zone-size, if it is, then any
995  segment which starts after the zone-capacity is marked as not-free in
996  the free segment bitmap at initial mount time. These segments are marked
997  as permanently used so they are not allocated for writes and
998  consequently are not needed to be garbage collected. In case the
999  zone-capacity is not aligned to default segment size(2MB), then a segment
1000  can start before the zone-capacity and span across zone-capacity boundary.
1001  Such spanning segments are also considered as usable segments. All blocks
1002  past the zone-capacity are considered unusable in these segments.
1003
1004Device aliasing feature
1005-----------------------
1006
1007f2fs can utilize a special file called a "device aliasing file." This file allows
1008the entire storage device to be mapped with a single, large extent, not using
1009the usual f2fs node structures. This mapped area is pinned and primarily intended
1010for holding the space.
1011
1012Essentially, this mechanism allows a portion of the f2fs area to be temporarily
1013reserved and used by another filesystem or for different purposes. Once that
1014external usage is complete, the device aliasing file can be deleted, releasing
1015the reserved space back to F2FS for its own use.
1016
1017.. code-block::
1018
1019   # ls /dev/vd*
1020   /dev/vdb (32GB) /dev/vdc (32GB)
1021   # mkfs.ext4 /dev/vdc
1022   # mkfs.f2fs -c /dev/vdc@vdc.file /dev/vdb
1023   # mount /dev/vdb /mnt/f2fs
1024   # ls -l /mnt/f2fs
1025   vdc.file
1026   # df -h
1027   /dev/vdb                            64G   33G   32G  52% /mnt/f2fs
1028
1029   # mount -o loop /dev/vdc /mnt/ext4
1030   # df -h
1031   /dev/vdb                            64G   33G   32G  52% /mnt/f2fs
1032   /dev/loop7                          32G   24K   30G   1% /mnt/ext4
1033   # umount /mnt/ext4
1034
1035   # f2fs_io getflags /mnt/f2fs/vdc.file
1036   get a flag on /mnt/f2fs/vdc.file ret=0, flags=nocow(pinned),immutable
1037   # f2fs_io setflags noimmutable /mnt/f2fs/vdc.file
1038   get a flag on noimmutable ret=0, flags=800010
1039   set a flag on /mnt/f2fs/vdc.file ret=0, flags=noimmutable
1040   # rm /mnt/f2fs/vdc.file
1041   # df -h
1042   /dev/vdb                            64G  753M   64G   2% /mnt/f2fs
1043
1044So, the key idea is, user can do any file operations on /dev/vdc, and
1045reclaim the space after the use, while the space is counted as /data.
1046That doesn't require modifying partition size and filesystem format.
1047
1048Per-file Read-Only Large Folio Support
1049--------------------------------------
1050
1051F2FS implements large folio support on the read path to leverage high-order
1052page allocation for significant performance gains. To minimize code complexity,
1053this support is currently excluded from the write path, which requires handling
1054complex optimizations such as compression and block allocation modes.
1055
1056This optional feature is triggered only when a file's immutable bit is set.
1057Consequently, F2FS will return EOPNOTSUPP if a user attempts to open a cached
1058file with write permissions, even immediately after clearing the bit. Write
1059access is only restored once the cached inode is dropped. The usage flow is
1060demonstrated below:
1061
1062.. code-block::
1063
1064   # f2fs_io setflags immutable /data/testfile_read_seq
1065
1066   /* flush and reload the inode to enable the large folio */
1067   # sync && echo 3 > /proc/sys/vm/drop_caches
1068
1069   /* mmap(MAP_POPULATE) + mlock() */
1070   # f2fs_io read 128 0 1024 mmap 1 0 /data/testfile_read_seq
1071
1072   /* mmap() + fadvise(POSIX_FADV_WILLNEED) + mlock() */
1073   # f2fs_io read 128 0 1024 fadvise 1 0 /data/testfile_read_seq
1074
1075   /* mmap() + mlock2(MLOCK_ONFAULT) + madvise(MADV_POPULATE_READ) */
1076   # f2fs_io read 128 0 1024 madvise 1 0 /data/testfile_read_seq
1077
1078   # f2fs_io clearflags immutable /data/testfile_read_seq
1079
1080   # f2fs_io write 1 0 1 zero buffered /data/testfile_read_seq
1081   Failed to open /mnt/test/test: Operation not supported
1082
1083   /* flush and reload the inode to disable the large folio */
1084   # sync && echo 3 > /proc/sys/vm/drop_caches
1085
1086   # f2fs_io write 1 0 1 zero buffered /data/testfile_read_seq
1087   Written 4096 bytes with pattern = zero, total_time = 29 us, max_latency = 28 us
1088
1089   # rm /data/testfile_read_seq
1090