1 .. SPDX-License-Identifier: GPL-2.0
3 ==========================================
4 WHAT IS Flash-Friendly File System (F2FS)?
5 ==========================================
7 NAND flash memory-based storage devices, such as SSD, eMMC, and SD cards, have
8 been equipped on a variety systems ranging from mobile to server systems. Since
9 they are known to have different characteristics from the conventional rotating
10 disks, a file system, an upper layer to the storage device, should adapt to the
11 changes from the sketch in the design level.
13 F2FS is a file system exploiting NAND flash memory-based storage devices, which
14 is based on Log-structured File System (LFS). The design has been focused on
15 addressing the fundamental issues in LFS, which are snowball effect of wandering
16 tree and high cleaning overhead.
18 Since a NAND flash memory-based storage device shows different characteristic
19 according to its internal geometry or flash memory management scheme, namely FTL,
20 F2FS and its tools support various parameters not only for configuring on-disk
21 layout, but also for selecting allocation and cleaning algorithms.
23 The following git tree provides the file system formatting tool (mkfs.f2fs),
24 a consistency checking tool (fsck.f2fs), and a debugging tool (dump.f2fs).
26 - git://git.kernel.org/pub/scm/linux/kernel/git/jaegeuk/f2fs-tools.git
28 For reporting bugs and sending patches, please use the following mailing list:
30 - linux-f2fs-devel@lists.sourceforge.net
32 Background and Design issues
33 ============================
35 Log-structured File System (LFS)
36 --------------------------------
37 "A log-structured file system writes all modifications to disk sequentially in
38 a log-like structure, thereby speeding up both file writing and crash recovery.
39 The log is the only structure on disk; it contains indexing information so that
40 files can be read back from the log efficiently. In order to maintain large free
41 areas on disk for fast writing, we divide the log into segments and use a
42 segment cleaner to compress the live information from heavily fragmented
43 segments." from Rosenblum, M. and Ousterhout, J. K., 1992, "The design and
44 implementation of a log-structured file system", ACM Trans. Computer Systems
47 Wandering Tree Problem
48 ----------------------
49 In LFS, when a file data is updated and written to the end of log, its direct
50 pointer block is updated due to the changed location. Then the indirect pointer
51 block is also updated due to the direct pointer block update. In this manner,
52 the upper index structures such as inode, inode map, and checkpoint block are
53 also updated recursively. This problem is called as wandering tree problem [1],
54 and in order to enhance the performance, it should eliminate or relax the update
55 propagation as much as possible.
57 [1] Bityutskiy, A. 2005. JFFS3 design issues. http://www.linux-mtd.infradead.org/
61 Since LFS is based on out-of-place writes, it produces so many obsolete blocks
62 scattered across the whole storage. In order to serve new empty log space, it
63 needs to reclaim these obsolete blocks seamlessly to users. This job is called
64 as a cleaning process.
66 The process consists of three operations as follows.
68 1. A victim segment is selected through referencing segment usage table.
69 2. It loads parent index structures of all the data in the victim identified by
70 segment summary blocks.
71 3. It checks the cross-reference between the data and its parent index structure.
72 4. It moves valid data selectively.
74 This cleaning job may cause unexpected long delays, so the most important goal
75 is to hide the latencies to users. And also definitely, it should reduce the
76 amount of valid data to be moved, and move them quickly as well.
83 - Enlarge the random write area for better performance, but provide the high
85 - Align FS data structures to the operational units in FTL as best efforts
87 Wandering Tree Problem
88 ----------------------
89 - Use a term, “node”, that represents inodes as well as various pointer blocks
90 - Introduce Node Address Table (NAT) containing the locations of all the “node”
91 blocks; this will cut off the update propagation.
95 - Support a background cleaning process
96 - Support greedy and cost-benefit algorithms for victim selection policies
97 - Support multi-head logs for static/dynamic hot and cold data separation
98 - Introduce adaptive logging for efficient block allocation
104 ======================== ============================================================
105 background_gc=%s Turn on/off cleaning operations, namely garbage
106 collection, triggered in background when I/O subsystem is
107 idle. If background_gc=on, it will turn on the garbage
108 collection and if background_gc=off, garbage collection
109 will be turned off. If background_gc=sync, it will turn
110 on synchronous garbage collection running in background.
111 Default value for this option is on. So garbage
112 collection is on by default.
113 gc_merge When background_gc is on, this option can be enabled to
114 let background GC thread to handle foreground GC requests,
115 it can eliminate the sluggish issue caused by slow foreground
116 GC operation when GC is triggered from a process with limited
117 I/O and CPU resources.
118 nogc_merge Disable GC merge feature.
119 disable_roll_forward Disable the roll-forward recovery routine
120 norecovery Disable the roll-forward recovery routine, mounted read-
121 only (i.e., -o ro,disable_roll_forward)
122 discard/nodiscard Enable/disable real-time discard in f2fs, if discard is
123 enabled, f2fs will issue discard/TRIM commands when a
125 no_heap Disable heap-style segment allocation which finds free
126 segments for data from the beginning of main area, while
127 for node from the end of main area.
128 nouser_xattr Disable Extended User Attributes. Note: xattr is enabled
129 by default if CONFIG_F2FS_FS_XATTR is selected.
130 noacl Disable POSIX Access Control List. Note: acl is enabled
131 by default if CONFIG_F2FS_FS_POSIX_ACL is selected.
132 active_logs=%u Support configuring the number of active logs. In the
133 current design, f2fs supports only 2, 4, and 6 logs.
135 disable_ext_identify Disable the extension list configured by mkfs, so f2fs
136 is not aware of cold files such as media files.
137 inline_xattr Enable the inline xattrs feature.
138 noinline_xattr Disable the inline xattrs feature.
139 inline_xattr_size=%u Support configuring inline xattr size, it depends on
140 flexible inline xattr feature.
141 inline_data Enable the inline data feature: Newly created small (<~3.4k)
142 files can be written into inode block.
143 inline_dentry Enable the inline dir feature: data in newly created
144 directory entries can be written into inode block. The
145 space of inode block which is used to store inline
146 dentries is limited to ~3.4k.
147 noinline_dentry Disable the inline dentry feature.
148 flush_merge Merge concurrent cache_flush commands as much as possible
149 to eliminate redundant command issues. If the underlying
150 device handles the cache_flush command relatively slowly,
151 recommend to enable this option.
152 nobarrier This option can be used if underlying storage guarantees
153 its cached data should be written to the novolatile area.
154 If this option is set, no cache_flush commands are issued
155 but f2fs still guarantees the write ordering of all the
157 fastboot This option is used when a system wants to reduce mount
158 time as much as possible, even though normal performance
160 extent_cache Enable an extent cache based on rb-tree, it can cache
161 as many as extent which map between contiguous logical
162 address and physical address per inode, resulting in
163 increasing the cache hit ratio. Set by default.
164 noextent_cache Disable an extent cache based on rb-tree explicitly, see
165 the above extent_cache mount option.
166 noinline_data Disable the inline data feature, inline data feature is
168 data_flush Enable data flushing before checkpoint in order to
169 persist data of regular and symlink.
170 reserve_root=%d Support configuring reserved space which is used for
171 allocation from a privileged user with specified uid or
172 gid, unit: 4KB, the default limit is 0.2% of user blocks.
173 resuid=%d The user ID which may use the reserved blocks.
174 resgid=%d The group ID which may use the reserved blocks.
175 fault_injection=%d Enable fault injection in all supported types with
176 specified injection rate.
177 fault_type=%d Support configuring fault injection type, should be
178 enabled with fault_injection option, fault type value
179 is shown below, it supports single or combined type.
181 =================== ===========
183 =================== ===========
184 FAULT_KMALLOC 0x000000001
185 FAULT_KVMALLOC 0x000000002
186 FAULT_PAGE_ALLOC 0x000000004
187 FAULT_PAGE_GET 0x000000008
188 FAULT_ALLOC_BIO 0x000000010 (obsolete)
189 FAULT_ALLOC_NID 0x000000020
190 FAULT_ORPHAN 0x000000040
191 FAULT_BLOCK 0x000000080
192 FAULT_DIR_DEPTH 0x000000100
193 FAULT_EVICT_INODE 0x000000200
194 FAULT_TRUNCATE 0x000000400
195 FAULT_READ_IO 0x000000800
196 FAULT_CHECKPOINT 0x000001000
197 FAULT_DISCARD 0x000002000
198 FAULT_WRITE_IO 0x000004000
199 FAULT_SLAB_ALLOC 0x000008000
200 FAULT_DQUOT_INIT 0x000010000
201 FAULT_LOCK_OP 0x000020000
202 =================== ===========
203 mode=%s Control block allocation mode which supports "adaptive"
204 and "lfs". In "lfs" mode, there should be no random
205 writes towards main area.
206 "fragment:segment" and "fragment:block" are newly added here.
207 These are developer options for experiments to simulate filesystem
208 fragmentation/after-GC situation itself. The developers use these
209 modes to understand filesystem fragmentation/after-GC condition well,
210 and eventually get some insights to handle them better.
211 In "fragment:segment", f2fs allocates a new segment in ramdom
212 position. With this, we can simulate the after-GC condition.
213 In "fragment:block", we can scatter block allocation with
214 "max_fragment_chunk" and "max_fragment_hole" sysfs nodes.
215 We added some randomness to both chunk and hole size to make
216 it close to realistic IO pattern. So, in this mode, f2fs will allocate
217 1..<max_fragment_chunk> blocks in a chunk and make a hole in the
218 length of 1..<max_fragment_hole> by turns. With this, the newly
219 allocated blocks will be scattered throughout the whole partition.
220 Note that "fragment:block" implicitly enables "fragment:segment"
221 option for more randomness.
222 Please, use these options for your experiments and we strongly
223 recommend to re-format the filesystem after using these options.
224 io_bits=%u Set the bit size of write IO requests. It should be set
226 usrquota Enable plain user disk quota accounting.
227 grpquota Enable plain group disk quota accounting.
228 prjquota Enable plain project quota accounting.
229 usrjquota=<file> Appoint specified file and type during mount, so that quota
230 grpjquota=<file> information can be properly updated during recovery flow,
231 prjjquota=<file> <quota file>: must be in root directory;
232 jqfmt=<quota type> <quota type>: [vfsold,vfsv0,vfsv1].
233 offusrjquota Turn off user journalled quota.
234 offgrpjquota Turn off group journalled quota.
235 offprjjquota Turn off project journalled quota.
236 quota Enable plain user disk quota accounting.
237 noquota Disable all plain disk quota option.
238 whint_mode=%s Control which write hints are passed down to block
239 layer. This supports "off", "user-based", and
240 "fs-based". In "off" mode (default), f2fs does not pass
241 down hints. In "user-based" mode, f2fs tries to pass
242 down hints given by users. And in "fs-based" mode, f2fs
243 passes down hints with its policy.
244 alloc_mode=%s Adjust block allocation policy, which supports "reuse"
246 fsync_mode=%s Control the policy of fsync. Currently supports "posix",
247 "strict", and "nobarrier". In "posix" mode, which is
248 default, fsync will follow POSIX semantics and does a
249 light operation to improve the filesystem performance.
250 In "strict" mode, fsync will be heavy and behaves in line
251 with xfs, ext4 and btrfs, where xfstest generic/342 will
252 pass, but the performance will regress. "nobarrier" is
253 based on "posix", but doesn't issue flush command for
254 non-atomic files likewise "nobarrier" mount option.
255 test_dummy_encryption
256 test_dummy_encryption=%s
257 Enable dummy encryption, which provides a fake fscrypt
258 context. The fake fscrypt context is used by xfstests.
259 The argument may be either "v1" or "v2", in order to
260 select the corresponding fscrypt policy version.
261 checkpoint=%s[:%u[%]] Set to "disable" to turn off checkpointing. Set to "enable"
262 to reenable checkpointing. Is enabled by default. While
263 disabled, any unmounting or unexpected shutdowns will cause
264 the filesystem contents to appear as they did when the
265 filesystem was mounted with that option.
266 While mounting with checkpoint=disabled, the filesystem must
267 run garbage collection to ensure that all available space can
268 be used. If this takes too much time, the mount may return
269 EAGAIN. You may optionally add a value to indicate how much
270 of the disk you would be willing to temporarily give up to
271 avoid additional garbage collection. This can be given as a
272 number of blocks, or as a percent. For instance, mounting
273 with checkpoint=disable:100% would always succeed, but it may
274 hide up to all remaining free space. The actual space that
275 would be unusable can be viewed at /sys/fs/f2fs/<disk>/unusable
276 This space is reclaimed once checkpoint=enable.
277 checkpoint_merge When checkpoint is enabled, this can be used to create a kernel
278 daemon and make it to merge concurrent checkpoint requests as
279 much as possible to eliminate redundant checkpoint issues. Plus,
280 we can eliminate the sluggish issue caused by slow checkpoint
281 operation when the checkpoint is done in a process context in
282 a cgroup having low i/o budget and cpu shares. To make this
283 do better, we set the default i/o priority of the kernel daemon
284 to "3", to give one higher priority than other kernel threads.
285 This is the same way to give a I/O priority to the jbd2
286 journaling thread of ext4 filesystem.
287 nocheckpoint_merge Disable checkpoint merge feature.
288 compress_algorithm=%s Control compress algorithm, currently f2fs supports "lzo",
289 "lz4", "zstd" and "lzo-rle" algorithm.
290 compress_algorithm=%s:%d Control compress algorithm and its compress level, now, only
291 "lz4" and "zstd" support compress level config.
292 algorithm level range
295 compress_log_size=%u Support configuring compress cluster size, the size will
296 be 4KB * (1 << %u), 16KB is minimum size, also it's
298 compress_extension=%s Support adding specified extension, so that f2fs can enable
299 compression on those corresponding files, e.g. if all files
300 with '.ext' has high compression rate, we can set the '.ext'
301 on compression extension list and enable compression on
302 these file by default rather than to enable it via ioctl.
303 For other files, we can still enable compression via ioctl.
304 Note that, there is one reserved special extension '*', it
305 can be set to enable compression for all files.
306 nocompress_extension=%s Support adding specified extension, so that f2fs can disable
307 compression on those corresponding files, just contrary to compression extension.
308 If you know exactly which files cannot be compressed, you can use this.
309 The same extension name can't appear in both compress and nocompress
310 extension at the same time.
311 If the compress extension specifies all files, the types specified by the
312 nocompress extension will be treated as special cases and will not be compressed.
313 Don't allow use '*' to specifie all file in nocompress extension.
314 After add nocompress_extension, the priority should be:
315 dir_flag < comp_extention,nocompress_extension < comp_file_flag,no_comp_file_flag.
316 See more in compression sections.
318 compress_chksum Support verifying chksum of raw data in compressed cluster.
319 compress_mode=%s Control file compression mode. This supports "fs" and "user"
320 modes. In "fs" mode (default), f2fs does automatic compression
321 on the compression enabled files. In "user" mode, f2fs disables
322 the automaic compression and gives the user discretion of
323 choosing the target file and the timing. The user can do manual
324 compression/decompression on the compression enabled files using
326 compress_cache Support to use address space of a filesystem managed inode to
327 cache compressed block, in order to improve cache hit ratio of
329 inlinecrypt When possible, encrypt/decrypt the contents of encrypted
330 files using the blk-crypto framework rather than
331 filesystem-layer encryption. This allows the use of
332 inline encryption hardware. The on-disk format is
333 unaffected. For more details, see
334 Documentation/block/inline-encryption.rst.
335 atgc Enable age-threshold garbage collection, it provides high
336 effectiveness and efficiency on background GC.
337 discard_unit=%s Control discard unit, the argument can be "block", "segment"
338 and "section", issued discard command's offset/size will be
339 aligned to the unit, by default, "discard_unit=block" is set,
340 so that small discard functionality is enabled.
341 For blkzoned device, "discard_unit=section" will be set by
342 default, it is helpful for large sized SMR or ZNS devices to
343 reduce memory cost by getting rid of fs metadata supports small
345 ======================== ============================================================
350 /sys/kernel/debug/f2fs/ contains information about all the partitions mounted as
351 f2fs. Each file shows the whole f2fs information.
353 /sys/kernel/debug/f2fs/status includes:
355 - major file system information managed by f2fs currently
356 - average SIT information about whole segments
357 - current memory footprint consumed by f2fs.
362 Information about mounted f2fs file systems can be found in
363 /sys/fs/f2fs. Each mounted filesystem will have a directory in
364 /sys/fs/f2fs based on its device name (i.e., /sys/fs/f2fs/sda).
365 The files in each per-device directory are shown in table below.
367 Files in /sys/fs/f2fs/<devname>
368 (see also Documentation/ABI/testing/sysfs-fs-f2fs)
373 1. Download userland tools and compile them.
375 2. Skip, if f2fs was compiled statically inside kernel.
376 Otherwise, insert the f2fs.ko module::
380 3. Create a directory to use when mounting::
384 4. Format the block device, and then mount as f2fs::
386 # mkfs.f2fs -l label /dev/block_device
387 # mount -t f2fs /dev/block_device /mnt/f2fs
391 The mkfs.f2fs is for the use of formatting a partition as the f2fs filesystem,
392 which builds a basic on-disk layout.
394 The quick options consist of:
396 =============== ===========================================================
397 ``-l [label]`` Give a volume label, up to 512 unicode name.
398 ``-a [0 or 1]`` Split start location of each area for heap-based allocation.
400 1 is set by default, which performs this.
401 ``-o [int]`` Set overprovision ratio in percent over volume size.
404 ``-s [int]`` Set the number of segments per section.
407 ``-z [int]`` Set the number of sections per zone.
410 ``-e [str]`` Set basic extension list. e.g. "mp3,gif,mov"
411 ``-t [0 or 1]`` Disable discard command or not.
413 1 is set by default, which conducts discard.
414 =============== ===========================================================
416 Note: please refer to the manpage of mkfs.f2fs(8) to get full option list.
420 The fsck.f2fs is a tool to check the consistency of an f2fs-formatted
421 partition, which examines whether the filesystem metadata and user-made data
422 are cross-referenced correctly or not.
423 Note that, initial version of the tool does not fix any inconsistency.
425 The quick options consist of::
427 -d debug level [default:0]
429 Note: please refer to the manpage of fsck.f2fs(8) to get full option list.
433 The dump.f2fs shows the information of specific inode and dumps SSA and SIT to
434 file. Each file is dump_ssa and dump_sit.
436 The dump.f2fs is used to debug on-disk data structures of the f2fs filesystem.
437 It shows on-disk inode information recognized by a given inode number, and is
438 able to dump all the SSA and SIT entries into predefined files, ./dump_ssa and
439 ./dump_sit respectively.
441 The options consist of::
443 -d debug level [default:0]
445 -s [SIT dump segno from #1~#2 (decimal), for all 0~-1]
446 -a [SSA dump segno from #1~#2 (decimal), for all 0~-1]
450 # dump.f2fs -i [ino] /dev/sdx
451 # dump.f2fs -s 0~-1 /dev/sdx (SIT dump)
452 # dump.f2fs -a 0~-1 /dev/sdx (SSA dump)
454 Note: please refer to the manpage of dump.f2fs(8) to get full option list.
458 The sload.f2fs gives a way to insert files and directories in the exisiting disk
459 image. This tool is useful when building f2fs images given compiled files.
461 Note: please refer to the manpage of sload.f2fs(8) to get full option list.
465 The resize.f2fs lets a user resize the f2fs-formatted disk image, while preserving
466 all the files and directories stored in the image.
468 Note: please refer to the manpage of resize.f2fs(8) to get full option list.
472 The defrag.f2fs can be used to defragment scattered written data as well as
473 filesystem metadata across the disk. This can improve the write speed by giving
474 more free consecutive space.
476 Note: please refer to the manpage of defrag.f2fs(8) to get full option list.
480 The f2fs_io is a simple tool to issue various filesystem APIs as well as
481 f2fs-specific ones, which is very useful for QA tests.
483 Note: please refer to the manpage of f2fs_io(8) to get full option list.
491 F2FS divides the whole volume into a number of segments, each of which is fixed
492 to 2MB in size. A section is composed of consecutive segments, and a zone
493 consists of a set of sections. By default, section and zone sizes are set to one
494 segment size identically, but users can easily modify the sizes by mkfs.
496 F2FS splits the entire volume into six areas, and all the areas except superblock
497 consist of multiple segments as described below::
499 align with the zone size <-|
500 |-> align with the segment size
501 _________________________________________________________________________
502 | | | Segment | Node | Segment | |
503 | Superblock | Checkpoint | Info. | Address | Summary | Main |
504 | (SB) | (CP) | Table (SIT) | Table (NAT) | Area (SSA) | |
505 |____________|_____2______|______N______|______N______|______N_____|__N___|
509 ._________________________________________.
510 |_Segment_|_..._|_Segment_|_..._|_Segment_|
519 It is located at the beginning of the partition, and there exist two copies
520 to avoid file system crash. It contains basic partition information and some
521 default parameters of f2fs.
524 It contains file system information, bitmaps for valid NAT/SIT sets, orphan
525 inode lists, and summary entries of current active segments.
527 - Segment Information Table (SIT)
528 It contains segment information such as valid block count and bitmap for the
529 validity of all the blocks.
531 - Node Address Table (NAT)
532 It is composed of a block address table for all the node blocks stored in
535 - Segment Summary Area (SSA)
536 It contains summary entries which contains the owner information of all the
537 data and node blocks stored in Main area.
540 It contains file and directory data including their indices.
542 In order to avoid misalignment between file system and flash-based storage, F2FS
543 aligns the start block address of CP with the segment size. Also, it aligns the
544 start block address of Main area with the zone size by reserving some segments
547 Reference the following survey for additional technical details.
548 https://wiki.linaro.org/WorkingGroups/Kernel/Projects/FlashCardSurvey
550 File System Metadata Structure
551 ------------------------------
553 F2FS adopts the checkpointing scheme to maintain file system consistency. At
554 mount time, F2FS first tries to find the last valid checkpoint data by scanning
555 CP area. In order to reduce the scanning time, F2FS uses only two copies of CP.
556 One of them always indicates the last valid data, which is called as shadow copy
557 mechanism. In addition to CP, NAT and SIT also adopt the shadow copy mechanism.
559 For file system consistency, each CP points to which NAT and SIT copies are
560 valid, as shown as below::
562 +--------+----------+---------+
564 +--------+----------+---------+
568 +-------+-------+--------+--------+--------+--------+
569 | CP #0 | CP #1 | SIT #0 | SIT #1 | NAT #0 | NAT #1 |
570 +-------+-------+--------+--------+--------+--------+
573 `----------------------------------------'
578 The key data structure to manage the data locations is a "node". Similar to
579 traditional file structures, F2FS has three types of node: inode, direct node,
580 indirect node. F2FS assigns 4KB to an inode block which contains 923 data block
581 indices, two direct node pointers, two indirect node pointers, and one double
582 indirect node pointer as described below. One direct node block contains 1018
583 data blocks, and one indirect node block contains also 1018 node blocks. Thus,
584 one inode block (i.e., a file) covers::
586 4KB * (923 + 2 * 1018 + 2 * 1018 * 1018 + 1018 * 1018 * 1018) := 3.94TB.
593 | `- direct node (1018)
595 `- double indirect node (1)
596 `- indirect node (1018)
597 `- direct node (1018)
600 Note that all the node blocks are mapped by NAT which means the location of
601 each node is translated by the NAT table. In the consideration of the wandering
602 tree problem, F2FS is able to cut off the propagation of node updates caused by
608 A directory entry occupies 11 bytes, which consists of the following attributes.
610 - hash hash value of the file name
612 - len the length of file name
613 - type file type such as directory, symlink, etc
615 A dentry block consists of 214 dentry slots and file names. Therein a bitmap is
616 used to represent whether each dentry is valid or not. A dentry block occupies
617 4KB with the following composition.
621 Dentry Block(4 K) = bitmap (27 bytes) + reserved (3 bytes) +
622 dentries(11 * 214 bytes) + file name (8 * 214 bytes)
625 +--------------------------------+
626 |dentry block 1 | dentry block 2 |
627 +--------------------------------+
630 . [Dentry Block Structure: 4KB] .
631 +--------+----------+----------+------------+
632 | bitmap | reserved | dentries | file names |
633 +--------+----------+----------+------------+
634 [Dentry Block: 4KB] . .
637 +------+------+-----+------+
638 | hash | ino | len | type |
639 +------+------+-----+------+
640 [Dentry Structure: 11 bytes]
642 F2FS implements multi-level hash tables for directory structure. Each level has
643 a hash table with dedicated number of hash buckets as shown below. Note that
644 "A(2B)" means a bucket includes 2 data blocks.
648 ----------------------
651 N : MAX_DIR_HASH_DEPTH
652 ----------------------
656 level #1 | A(2B) - A(2B)
658 level #2 | A(2B) - A(2B) - A(2B) - A(2B)
660 level #N/2 | A(2B) - A(2B) - A(2B) - A(2B) - A(2B) - ... - A(2B)
662 level #N | A(4B) - A(4B) - A(4B) - A(4B) - A(4B) - ... - A(4B)
664 The number of blocks and buckets are determined by::
666 ,- 2, if n < MAX_DIR_HASH_DEPTH / 2,
667 # of blocks in level #n = |
670 ,- 2^(n + dir_level),
671 | if n + dir_level < MAX_DIR_HASH_DEPTH / 2,
672 # of buckets in level #n = |
673 `- 2^((MAX_DIR_HASH_DEPTH / 2) - 1),
676 When F2FS finds a file name in a directory, at first a hash value of the file
677 name is calculated. Then, F2FS scans the hash table in level #0 to find the
678 dentry consisting of the file name and its inode number. If not found, F2FS
679 scans the next hash table in level #1. In this way, F2FS scans hash tables in
680 each levels incrementally from 1 to N. In each level F2FS needs to scan only
681 one bucket determined by the following equation, which shows O(log(# of files))
684 bucket number to scan in level #n = (hash value) % (# of buckets in level #n)
686 In the case of file creation, F2FS finds empty consecutive slots that cover the
687 file name. F2FS searches the empty slots in the hash tables of whole levels from
688 1 to N in the same way as the lookup operation.
690 The following figure shows an example of two cases holding children::
692 --------------> Dir <--------------
696 child - child [hole] - child
698 child - child - child [hole] - [hole] - child
701 Number of children = 6, Number of children = 3,
702 File size = 7 File size = 7
704 Default Block Allocation
705 ------------------------
707 At runtime, F2FS manages six active logs inside "Main" area: Hot/Warm/Cold node
708 and Hot/Warm/Cold data.
710 - Hot node contains direct node blocks of directories.
711 - Warm node contains direct node blocks except hot node blocks.
712 - Cold node contains indirect node blocks
713 - Hot data contains dentry blocks
714 - Warm data contains data blocks except hot and cold data blocks
715 - Cold data contains multimedia data or migrated data blocks
717 LFS has two schemes for free space management: threaded log and copy-and-compac-
718 tion. The copy-and-compaction scheme which is known as cleaning, is well-suited
719 for devices showing very good sequential write performance, since free segments
720 are served all the time for writing new data. However, it suffers from cleaning
721 overhead under high utilization. Contrarily, the threaded log scheme suffers
722 from random writes, but no cleaning process is needed. F2FS adopts a hybrid
723 scheme where the copy-and-compaction scheme is adopted by default, but the
724 policy is dynamically changed to the threaded log scheme according to the file
727 In order to align F2FS with underlying flash-based storage, F2FS allocates a
728 segment in a unit of section. F2FS expects that the section size would be the
729 same as the unit size of garbage collection in FTL. Furthermore, with respect
730 to the mapping granularity in FTL, F2FS allocates each section of the active
731 logs from different zones as much as possible, since FTL can write the data in
732 the active logs into one allocation unit according to its mapping granularity.
737 F2FS does cleaning both on demand and in the background. On-demand cleaning is
738 triggered when there are not enough free segments to serve VFS calls. Background
739 cleaner is operated by a kernel thread, and triggers the cleaning job when the
742 F2FS supports two victim selection policies: greedy and cost-benefit algorithms.
743 In the greedy algorithm, F2FS selects a victim segment having the smallest number
744 of valid blocks. In the cost-benefit algorithm, F2FS selects a victim segment
745 according to the segment age and the number of valid blocks in order to address
746 log block thrashing problem in the greedy algorithm. F2FS adopts the greedy
747 algorithm for on-demand cleaner, while background cleaner adopts cost-benefit
750 In order to identify whether the data in the victim segment are valid or not,
751 F2FS manages a bitmap. Each bit represents the validity of a block, and the
752 bitmap is composed of a bit stream covering whole blocks in main area.
757 1) whint_mode=off. F2FS only passes down WRITE_LIFE_NOT_SET.
759 2) whint_mode=user-based. F2FS tries to pass down hints given by
762 ===================== ======================== ===================
764 ===================== ======================== ===================
765 N/A META WRITE_LIFE_NOT_SET
769 ioctl(COLD) COLD_DATA WRITE_LIFE_EXTREME
773 WRITE_LIFE_EXTREME COLD_DATA WRITE_LIFE_EXTREME
774 WRITE_LIFE_SHORT HOT_DATA WRITE_LIFE_SHORT
775 WRITE_LIFE_NOT_SET WARM_DATA WRITE_LIFE_NOT_SET
777 WRITE_LIFE_MEDIUM " "
781 WRITE_LIFE_EXTREME COLD_DATA WRITE_LIFE_EXTREME
782 WRITE_LIFE_SHORT HOT_DATA WRITE_LIFE_SHORT
783 WRITE_LIFE_NOT_SET WARM_DATA WRITE_LIFE_NOT_SET
784 WRITE_LIFE_NONE " WRITE_LIFE_NONE
785 WRITE_LIFE_MEDIUM " WRITE_LIFE_MEDIUM
786 WRITE_LIFE_LONG " WRITE_LIFE_LONG
787 ===================== ======================== ===================
789 3) whint_mode=fs-based. F2FS passes down hints with its policy.
791 ===================== ======================== ===================
793 ===================== ======================== ===================
794 N/A META WRITE_LIFE_MEDIUM;
795 N/A HOT_NODE WRITE_LIFE_NOT_SET
797 N/A COLD_NODE WRITE_LIFE_NONE
798 ioctl(COLD) COLD_DATA WRITE_LIFE_EXTREME
802 WRITE_LIFE_EXTREME COLD_DATA WRITE_LIFE_EXTREME
803 WRITE_LIFE_SHORT HOT_DATA WRITE_LIFE_SHORT
804 WRITE_LIFE_NOT_SET WARM_DATA WRITE_LIFE_LONG
806 WRITE_LIFE_MEDIUM " "
810 WRITE_LIFE_EXTREME COLD_DATA WRITE_LIFE_EXTREME
811 WRITE_LIFE_SHORT HOT_DATA WRITE_LIFE_SHORT
812 WRITE_LIFE_NOT_SET WARM_DATA WRITE_LIFE_NOT_SET
813 WRITE_LIFE_NONE " WRITE_LIFE_NONE
814 WRITE_LIFE_MEDIUM " WRITE_LIFE_MEDIUM
815 WRITE_LIFE_LONG " WRITE_LIFE_LONG
816 ===================== ======================== ===================
821 The default policy follows the below POSIX rule.
823 Allocating disk space
824 The default operation (i.e., mode is zero) of fallocate() allocates
825 the disk space within the range specified by offset and len. The
826 file size (as reported by stat(2)) will be changed if offset+len is
827 greater than the file size. Any subregion within the range specified
828 by offset and len that did not contain data before the call will be
829 initialized to zero. This default behavior closely resembles the
830 behavior of the posix_fallocate(3) library function, and is intended
831 as a method of optimally implementing that function.
833 However, once F2FS receives ioctl(fd, F2FS_IOC_SET_PIN_FILE) in prior to
834 fallocate(fd, DEFAULT_MODE), it allocates on-disk block addressess having
835 zero or random data, which is useful to the below scenario where:
838 2. ioctl(fd, F2FS_IOC_SET_PIN_FILE)
839 3. fallocate(fd, 0, 0, size)
840 4. address = fibmap(fd, offset)
842 6. write(blkdev, address)
844 Compression implementation
845 --------------------------
847 - New term named cluster is defined as basic unit of compression, file can
848 be divided into multiple clusters logically. One cluster includes 4 << n
849 (n >= 0) logical pages, compression size is also cluster size, each of
850 cluster can be compressed or not.
852 - In cluster metadata layout, one special block address is used to indicate
853 a cluster is a compressed one or normal one; for compressed cluster, following
854 metadata maps cluster to [1, 4 << n - 1] physical blocks, in where f2fs
855 stores data including compress header and compressed data.
857 - In order to eliminate write amplification during overwrite, F2FS only
858 support compression on write-once file, data can be compressed only when
859 all logical blocks in cluster contain valid data and compress ratio of
860 cluster data is lower than specified threshold.
862 - To enable compression on regular inode, there are four ways:
865 * chattr +c dir; touch dir/file
866 * mount w/ -o compress_extension=ext; touch file.ext
867 * mount w/ -o compress_extension=*; touch any_file
869 - To disable compression on regular inode, there are two ways:
872 * mount w/ -o nocompress_extension=ext; touch file.ext
874 - Priority in between FS_COMPR_FL, FS_NOCOMP_FS, extensions:
876 * compress_extension=so; nocompress_extension=zip; chattr +c dir; touch
877 dir/foo.so; touch dir/bar.zip; touch dir/baz.txt; then foo.so and baz.txt
878 should be compresse, bar.zip should be non-compressed. chattr +c dir/bar.zip
879 can enable compress on bar.zip.
880 * compress_extension=so; nocompress_extension=zip; chattr -c dir; touch
881 dir/foo.so; touch dir/bar.zip; touch dir/baz.txt; then foo.so should be
882 compresse, bar.zip and baz.txt should be non-compressed.
883 chattr+c dir/bar.zip; chattr+c dir/baz.txt; can enable compress on bar.zip
886 - At this point, compression feature doesn't expose compressed space to user
887 directly in order to guarantee potential data updates later to the space.
888 Instead, the main goal is to reduce data writes to flash disk as much as
889 possible, resulting in extending disk life time as well as relaxing IO
890 congestion. Alternatively, we've added ioctl(F2FS_IOC_RELEASE_COMPRESS_BLOCKS)
891 interface to reclaim compressed space and show it to user after putting the
892 immutable bit. Immutable bit, after release, it doesn't allow writing/mmaping
893 on the file, until reserving compressed space via
894 ioctl(F2FS_IOC_RESERVE_COMPRESS_BLOCKS) or truncating filesize to zero.
896 Compress metadata layout::
899 +-----------------------------------------------+
900 | cluster 1 | cluster 2 | ......... | cluster N |
901 +-----------------------------------------------+
904 . Compressed Cluster . . Normal Cluster .
905 +----------+---------+---------+---------+ +---------+---------+---------+---------+
906 |compr flag| block 1 | block 2 | block 3 | | block 1 | block 2 | block 3 | block 4 |
907 +----------+---------+---------+---------+ +---------+---------+---------+---------+
911 +-------------+-------------+----------+----------------------------+
912 | data length | data chksum | reserved | compressed data |
913 +-------------+-------------+----------+----------------------------+
916 --------------------------
918 f2fs supports "fs" and "user" compression modes with "compression_mode" mount option.
919 With this option, f2fs provides a choice to select the way how to compress the
920 compression enabled files (refer to "Compression implementation" section for how to
921 enable compression on a regular inode).
924 This is the default option. f2fs does automatic compression in the writeback of the
925 compression enabled files.
927 2) compress_mode=user
928 This disables the automatic compression and gives the user discretion of choosing the
929 target file and the timing. The user can do manual compression/decompression on the
930 compression enabled files using F2FS_IOC_DECOMPRESS_FILE and F2FS_IOC_COMPRESS_FILE
931 ioctls like the below.
933 To decompress a file,
935 fd = open(filename, O_WRONLY, 0);
936 ret = ioctl(fd, F2FS_IOC_DECOMPRESS_FILE);
940 fd = open(filename, O_WRONLY, 0);
941 ret = ioctl(fd, F2FS_IOC_COMPRESS_FILE);
943 NVMe Zoned Namespace devices
944 ----------------------------
946 - ZNS defines a per-zone capacity which can be equal or less than the
947 zone-size. Zone-capacity is the number of usable blocks in the zone.
948 F2FS checks if zone-capacity is less than zone-size, if it is, then any
949 segment which starts after the zone-capacity is marked as not-free in
950 the free segment bitmap at initial mount time. These segments are marked
951 as permanently used so they are not allocated for writes and
952 consequently are not needed to be garbage collected. In case the
953 zone-capacity is not aligned to default segment size(2MB), then a segment
954 can start before the zone-capacity and span across zone-capacity boundary.
955 Such spanning segments are also considered as usable segments. All blocks
956 past the zone-capacity are considered unusable in these segments.