| blob | fb7d2ee022bc06f5dfb3e45e339e353bc5d3b5cd |
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 sending patches, please use the following mailing list:
30 - linux-f2fs-devel@lists.sourceforge.net
32 For reporting bugs, please use the following f2fs bug tracker link:
34 - https://bugzilla.kernel.org/enter_bug.cgi?product=File%20System&component=f2fs
36 Background and Design issues
37 ============================
39 Log-structured File System (LFS)
40 --------------------------------
41 "A log-structured file system writes all modifications to disk sequentially in
42 a log-like structure, thereby speeding up both file writing and crash recovery.
43 The log is the only structure on disk; it contains indexing information so that
44 files can be read back from the log efficiently. In order to maintain large free
45 areas on disk for fast writing, we divide the log into segments and use a
46 segment cleaner to compress the live information from heavily fragmented
47 segments." from Rosenblum, M. and Ousterhout, J. K., 1992, "The design and
48 implementation of a log-structured file system", ACM Trans. Computer Systems
49 10, 1, 26–52.
51 Wandering Tree Problem
52 ----------------------
53 In LFS, when a file data is updated and written to the end of log, its direct
54 pointer block is updated due to the changed location. Then the indirect pointer
55 block is also updated due to the direct pointer block update. In this manner,
56 the upper index structures such as inode, inode map, and checkpoint block are
57 also updated recursively. This problem is called as wandering tree problem [1],
58 and in order to enhance the performance, it should eliminate or relax the update
59 propagation as much as possible.
61 [1] Bityutskiy, A. 2005. JFFS3 design issues. http://www.linux-mtd.infradead.org/
63 Cleaning Overhead
64 -----------------
65 Since LFS is based on out-of-place writes, it produces so many obsolete blocks
66 scattered across the whole storage. In order to serve new empty log space, it
67 needs to reclaim these obsolete blocks seamlessly to users. This job is called
68 as a cleaning process.
70 The process consists of three operations as follows.
72 1. A victim segment is selected through referencing segment usage table.
73 2. It loads parent index structures of all the data in the victim identified by
74 segment summary blocks.
75 3. It checks the cross-reference between the data and its parent index structure.
76 4. It moves valid data selectively.
78 This cleaning job may cause unexpected long delays, so the most important goal
79 is to hide the latencies to users. And also definitely, it should reduce the
80 amount of valid data to be moved, and move them quickly as well.
82 Key Features
83 ============
85 Flash Awareness
86 ---------------
87 - Enlarge the random write area for better performance, but provide the high
88 spatial locality
89 - Align FS data structures to the operational units in FTL as best efforts
91 Wandering Tree Problem
92 ----------------------
93 - Use a term, “node”, that represents inodes as well as various pointer blocks
94 - Introduce Node Address Table (NAT) containing the locations of all the “node”
95 blocks; this will cut off the update propagation.
97 Cleaning Overhead
98 -----------------
99 - Support a background cleaning process
100 - Support greedy and cost-benefit algorithms for victim selection policies
101 - Support multi-head logs for static/dynamic hot and cold data separation
102 - Introduce adaptive logging for efficient block allocation
104 Mount Options
105 =============
108 ======================== ============================================================
109 background_gc=%s Turn on/off cleaning operations, namely garbage
110 collection, triggered in background when I/O subsystem is
111 idle. If background_gc=on, it will turn on the garbage
112 collection and if background_gc=off, garbage collection
113 will be turned off. If background_gc=sync, it will turn
114 on synchronous garbage collection running in background.
115 Default value for this option is on. So garbage
116 collection is on by default.
117 gc_merge When background_gc is on, this option can be enabled to
118 let background GC thread to handle foreground GC requests,
119 it can eliminate the sluggish issue caused by slow foreground
120 GC operation when GC is triggered from a process with limited
121 I/O and CPU resources.
122 nogc_merge Disable GC merge feature.
123 disable_roll_forward Disable the roll-forward recovery routine
124 norecovery Disable the roll-forward recovery routine, mounted read-
125 only (i.e., -o ro,disable_roll_forward)
126 discard/nodiscard Enable/disable real-time discard in f2fs, if discard is
127 enabled, f2fs will issue discard/TRIM commands when a
128 segment is cleaned.
129 heap/no_heap Deprecated.
130 nouser_xattr Disable Extended User Attributes. Note: xattr is enabled
131 by default if CONFIG_F2FS_FS_XATTR is selected.
132 noacl Disable POSIX Access Control List. Note: acl is enabled
133 by default if CONFIG_F2FS_FS_POSIX_ACL is selected.
134 active_logs=%u Support configuring the number of active logs. In the
135 current design, f2fs supports only 2, 4, and 6 logs.
136 Default number is 6.
137 disable_ext_identify Disable the extension list configured by mkfs, so f2fs
138 is not aware of cold files such as media files.
139 inline_xattr Enable the inline xattrs feature.
140 noinline_xattr Disable the inline xattrs feature.
141 inline_xattr_size=%u Support configuring inline xattr size, it depends on
142 flexible inline xattr feature.
143 inline_data Enable the inline data feature: Newly created small (<~3.4k)
144 files can be written into inode block.
145 inline_dentry Enable the inline dir feature: data in newly created
146 directory entries can be written into inode block. The
147 space of inode block which is used to store inline
148 dentries is limited to ~3.4k.
149 noinline_dentry Disable the inline dentry feature.
150 flush_merge Merge concurrent cache_flush commands as much as possible
151 to eliminate redundant command issues. If the underlying
152 device handles the cache_flush command relatively slowly,
153 recommend to enable this option.
154 nobarrier This option can be used if underlying storage guarantees
155 its cached data should be written to the novolatile area.
156 If this option is set, no cache_flush commands are issued
157 but f2fs still guarantees the write ordering of all the
158 data writes.
159 barrier If this option is set, cache_flush commands are allowed to be
160 issued.
161 fastboot This option is used when a system wants to reduce mount
162 time as much as possible, even though normal performance
163 can be sacrificed.
164 extent_cache Enable an extent cache based on rb-tree, it can cache
165 as many as extent which map between contiguous logical
166 address and physical address per inode, resulting in
167 increasing the cache hit ratio. Set by default.
168 noextent_cache Disable an extent cache based on rb-tree explicitly, see
169 the above extent_cache mount option.
170 noinline_data Disable the inline data feature, inline data feature is
171 enabled by default.
172 data_flush Enable data flushing before checkpoint in order to
173 persist data of regular and symlink.
174 reserve_root=%d Support configuring reserved space which is used for
175 allocation from a privileged user with specified uid or
176 gid, unit: 4KB, the default limit is 0.2% of user blocks.
177 resuid=%d The user ID which may use the reserved blocks.
178 resgid=%d The group ID which may use the reserved blocks.
179 fault_injection=%d Enable fault injection in all supported types with
180 specified injection rate.
181 fault_type=%d Support configuring fault injection type, should be
182 enabled with fault_injection option, fault type value
183 is shown below, it supports single or combined type.
185 =========================== ===========
186 Type_Name Type_Value
187 =========================== ===========
188 FAULT_KMALLOC 0x000000001
189 FAULT_KVMALLOC 0x000000002
190 FAULT_PAGE_ALLOC 0x000000004
191 FAULT_PAGE_GET 0x000000008
192 FAULT_ALLOC_BIO 0x000000010 (obsolete)
193 FAULT_ALLOC_NID 0x000000020
194 FAULT_ORPHAN 0x000000040
195 FAULT_BLOCK 0x000000080
196 FAULT_DIR_DEPTH 0x000000100
197 FAULT_EVICT_INODE 0x000000200
198 FAULT_TRUNCATE 0x000000400
199 FAULT_READ_IO 0x000000800
200 FAULT_CHECKPOINT 0x000001000
201 FAULT_DISCARD 0x000002000
202 FAULT_WRITE_IO 0x000004000
203 FAULT_SLAB_ALLOC 0x000008000
204 FAULT_DQUOT_INIT 0x000010000
205 FAULT_LOCK_OP 0x000020000
206 FAULT_BLKADDR_VALIDITY 0x000040000
207 FAULT_BLKADDR_CONSISTENCE 0x000080000
208 FAULT_NO_SEGMENT 0x000100000
209 =========================== ===========
210 mode=%s Control block allocation mode which supports "adaptive"
211 and "lfs". In "lfs" mode, there should be no random
212 writes towards main area.
213 "fragment:segment" and "fragment:block" are newly added here.
214 These are developer options for experiments to simulate filesystem
215 fragmentation/after-GC situation itself. The developers use these
216 modes to understand filesystem fragmentation/after-GC condition well,
217 and eventually get some insights to handle them better.
218 In "fragment:segment", f2fs allocates a new segment in ramdom
219 position. With this, we can simulate the after-GC condition.
220 In "fragment:block", we can scatter block allocation with
221 "max_fragment_chunk" and "max_fragment_hole" sysfs nodes.
222 We added some randomness to both chunk and hole size to make
223 it close to realistic IO pattern. So, in this mode, f2fs will allocate
224 1..<max_fragment_chunk> blocks in a chunk and make a hole in the
225 length of 1..<max_fragment_hole> by turns. With this, the newly
226 allocated blocks will be scattered throughout the whole partition.
227 Note that "fragment:block" implicitly enables "fragment:segment"
228 option for more randomness.
229 Please, use these options for your experiments and we strongly
230 recommend to re-format the filesystem after using these options.
231 usrquota Enable plain user disk quota accounting.
232 grpquota Enable plain group disk quota accounting.
233 prjquota Enable plain project quota accounting.
234 usrjquota=<file> Appoint specified file and type during mount, so that quota
235 grpjquota=<file> information can be properly updated during recovery flow,
236 prjjquota=<file> <quota file>: must be in root directory;
237 jqfmt=<quota type> <quota type>: [vfsold,vfsv0,vfsv1].
238 offusrjquota Turn off user journalled quota.
239 offgrpjquota Turn off group journalled quota.
240 offprjjquota Turn off project journalled quota.
241 quota Enable plain user disk quota accounting.
242 noquota Disable all plain disk quota option.
243 alloc_mode=%s Adjust block allocation policy, which supports "reuse"
244 and "default".
245 fsync_mode=%s Control the policy of fsync. Currently supports "posix",
246 "strict", and "nobarrier". In "posix" mode, which is
247 default, fsync will follow POSIX semantics and does a
248 light operation to improve the filesystem performance.
249 In "strict" mode, fsync will be heavy and behaves in line
250 with xfs, ext4 and btrfs, where xfstest generic/342 will
251 pass, but the performance will regress. "nobarrier" is
252 based on "posix", but doesn't issue flush command for
253 non-atomic files likewise "nobarrier" mount option.
254 test_dummy_encryption
255 test_dummy_encryption=%s
256 Enable dummy encryption, which provides a fake fscrypt
257 context. The fake fscrypt context is used by xfstests.
258 The argument may be either "v1" or "v2", in order to
259 select the corresponding fscrypt policy version.
260 checkpoint=%s[:%u[%]] Set to "disable" to turn off checkpointing. Set to "enable"
261 to reenable checkpointing. Is enabled by default. While
262 disabled, any unmounting or unexpected shutdowns will cause
263 the filesystem contents to appear as they did when the
264 filesystem was mounted with that option.
265 While mounting with checkpoint=disable, the filesystem must
266 run garbage collection to ensure that all available space can
267 be used. If this takes too much time, the mount may return
268 EAGAIN. You may optionally add a value to indicate how much
269 of the disk you would be willing to temporarily give up to
270 avoid additional garbage collection. This can be given as a
271 number of blocks, or as a percent. For instance, mounting
272 with checkpoint=disable:100% would always succeed, but it may
273 hide up to all remaining free space. The actual space that
274 would be unusable can be viewed at /sys/fs/f2fs/<disk>/unusable
275 This space is reclaimed once checkpoint=enable.
276 checkpoint_merge When checkpoint is enabled, this can be used to create a kernel
277 daemon and make it to merge concurrent checkpoint requests as
278 much as possible to eliminate redundant checkpoint issues. Plus,
279 we can eliminate the sluggish issue caused by slow checkpoint
280 operation when the checkpoint is done in a process context in
281 a cgroup having low i/o budget and cpu shares. To make this
282 do better, we set the default i/o priority of the kernel daemon
283 to "3", to give one higher priority than other kernel threads.
284 This is the same way to give a I/O priority to the jbd2
285 journaling thread of ext4 filesystem.
286 nocheckpoint_merge Disable checkpoint merge feature.
287 compress_algorithm=%s Control compress algorithm, currently f2fs supports "lzo",
288 "lz4", "zstd" and "lzo-rle" algorithm.
289 compress_algorithm=%s:%d Control compress algorithm and its compress level, now, only
290 "lz4" and "zstd" support compress level config.
291 algorithm level range
292 lz4 3 - 16
293 zstd 1 - 22
294 compress_log_size=%u Support configuring compress cluster size. The size will
295 be 4KB * (1 << %u). The default and minimum sizes are 16KB.
296 compress_extension=%s Support adding specified extension, so that f2fs can enable
297 compression on those corresponding files, e.g. if all files
298 with '.ext' has high compression rate, we can set the '.ext'
299 on compression extension list and enable compression on
300 these file by default rather than to enable it via ioctl.
301 For other files, we can still enable compression via ioctl.
302 Note that, there is one reserved special extension '*', it
303 can be set to enable compression for all files.
304 nocompress_extension=%s Support adding specified extension, so that f2fs can disable
305 compression on those corresponding files, just contrary to compression extension.
306 If you know exactly which files cannot be compressed, you can use this.
307 The same extension name can't appear in both compress and nocompress
308 extension at the same time.
309 If the compress extension specifies all files, the types specified by the
310 nocompress extension will be treated as special cases and will not be compressed.
311 Don't allow use '*' to specifie all file in nocompress extension.
312 After add nocompress_extension, the priority should be:
313 dir_flag < comp_extention,nocompress_extension < comp_file_flag,no_comp_file_flag.
314 See more in compression sections.
316 compress_chksum Support verifying chksum of raw data in compressed cluster.
317 compress_mode=%s Control file compression mode. This supports "fs" and "user"
318 modes. In "fs" mode (default), f2fs does automatic compression
319 on the compression enabled files. In "user" mode, f2fs disables
320 the automaic compression and gives the user discretion of
321 choosing the target file and the timing. The user can do manual
322 compression/decompression on the compression enabled files using
323 ioctls.
324 compress_cache Support to use address space of a filesystem managed inode to
325 cache compressed block, in order to improve cache hit ratio of
326 random read.
327 inlinecrypt When possible, encrypt/decrypt the contents of encrypted
328 files using the blk-crypto framework rather than
329 filesystem-layer encryption. This allows the use of
330 inline encryption hardware. The on-disk format is
331 unaffected. For more details, see
332 Documentation/block/inline-encryption.rst.
333 atgc Enable age-threshold garbage collection, it provides high
334 effectiveness and efficiency on background GC.
335 discard_unit=%s Control discard unit, the argument can be "block", "segment"
336 and "section", issued discard command's offset/size will be
337 aligned to the unit, by default, "discard_unit=block" is set,
338 so that small discard functionality is enabled.
339 For blkzoned device, "discard_unit=section" will be set by
340 default, it is helpful for large sized SMR or ZNS devices to
341 reduce memory cost by getting rid of fs metadata supports small
342 discard.
343 memory=%s Control memory mode. This supports "normal" and "low" modes.
344 "low" mode is introduced to support low memory devices.
345 Because of the nature of low memory devices, in this mode, f2fs
346 will try to save memory sometimes by sacrificing performance.
347 "normal" mode is the default mode and same as before.
348 age_extent_cache Enable an age extent cache based on rb-tree. It records
349 data block update frequency of the extent per inode, in
350 order to provide better temperature hints for data block
351 allocation.
352 errors=%s Specify f2fs behavior on critical errors. This supports modes:
353 "panic", "continue" and "remount-ro", respectively, trigger
354 panic immediately, continue without doing anything, and remount
355 the partition in read-only mode. By default it uses "continue"
356 mode.
357 ====================== =============== =============== ========
358 mode continue remount-ro panic
359 ====================== =============== =============== ========
360 access ops normal normal N/A
361 syscall errors -EIO -EROFS N/A
362 mount option rw ro N/A
363 pending dir write keep keep N/A
364 pending non-dir write drop keep N/A
365 pending node write drop keep N/A
366 pending meta write keep keep N/A
367 ====================== =============== =============== ========
368 ======================== ============================================================
370 Debugfs Entries
371 ===============
373 /sys/kernel/debug/f2fs/ contains information about all the partitions mounted as
374 f2fs. Each file shows the whole f2fs information.
376 /sys/kernel/debug/f2fs/status includes:
378 - major file system information managed by f2fs currently
379 - average SIT information about whole segments
380 - current memory footprint consumed by f2fs.
382 Sysfs Entries
383 =============
385 Information about mounted f2fs file systems can be found in
386 /sys/fs/f2fs. Each mounted filesystem will have a directory in
387 /sys/fs/f2fs based on its device name (i.e., /sys/fs/f2fs/sda).
388 The files in each per-device directory are shown in table below.
390 Files in /sys/fs/f2fs/<devname>
391 (see also Documentation/ABI/testing/sysfs-fs-f2fs)
393 Usage
394 =====
396 1. Download userland tools and compile them.
398 2. Skip, if f2fs was compiled statically inside kernel.
399 Otherwise, insert the f2fs.ko module::
401 # insmod f2fs.ko
403 3. Create a directory to use when mounting::
405 # mkdir /mnt/f2fs
407 4. Format the block device, and then mount as f2fs::
409 # mkfs.f2fs -l label /dev/block_device
410 # mount -t f2fs /dev/block_device /mnt/f2fs
412 mkfs.f2fs
413 ---------
414 The mkfs.f2fs is for the use of formatting a partition as the f2fs filesystem,
415 which builds a basic on-disk layout.
417 The quick options consist of:
419 =============== ===========================================================
420 ``-l [label]`` Give a volume label, up to 512 unicode name.
421 ``-a [0 or 1]`` Split start location of each area for heap-based allocation.
423 1 is set by default, which performs this.
424 ``-o [int]`` Set overprovision ratio in percent over volume size.
426 5 is set by default.
427 ``-s [int]`` Set the number of segments per section.
429 1 is set by default.
430 ``-z [int]`` Set the number of sections per zone.
432 1 is set by default.
433 ``-e [str]`` Set basic extension list. e.g. "mp3,gif,mov"
434 ``-t [0 or 1]`` Disable discard command or not.
436 1 is set by default, which conducts discard.
437 =============== ===========================================================
439 Note: please refer to the manpage of mkfs.f2fs(8) to get full option list.
441 fsck.f2fs
442 ---------
443 The fsck.f2fs is a tool to check the consistency of an f2fs-formatted
444 partition, which examines whether the filesystem metadata and user-made data
445 are cross-referenced correctly or not.
446 Note that, initial version of the tool does not fix any inconsistency.
448 The quick options consist of::
450 -d debug level [default:0]
452 Note: please refer to the manpage of fsck.f2fs(8) to get full option list.
454 dump.f2fs
455 ---------
456 The dump.f2fs shows the information of specific inode and dumps SSA and SIT to
457 file. Each file is dump_ssa and dump_sit.
459 The dump.f2fs is used to debug on-disk data structures of the f2fs filesystem.
460 It shows on-disk inode information recognized by a given inode number, and is
461 able to dump all the SSA and SIT entries into predefined files, ./dump_ssa and
462 ./dump_sit respectively.
464 The options consist of::
466 -d debug level [default:0]
467 -i inode no (hex)
468 -s [SIT dump segno from #1~#2 (decimal), for all 0~-1]
469 -a [SSA dump segno from #1~#2 (decimal), for all 0~-1]
471 Examples::
473 # dump.f2fs -i [ino] /dev/sdx
474 # dump.f2fs -s 0~-1 /dev/sdx (SIT dump)
475 # dump.f2fs -a 0~-1 /dev/sdx (SSA dump)
477 Note: please refer to the manpage of dump.f2fs(8) to get full option list.
479 sload.f2fs
480 ----------
481 The sload.f2fs gives a way to insert files and directories in the existing disk
482 image. This tool is useful when building f2fs images given compiled files.
484 Note: please refer to the manpage of sload.f2fs(8) to get full option list.
486 resize.f2fs
487 -----------
488 The resize.f2fs lets a user resize the f2fs-formatted disk image, while preserving
489 all the files and directories stored in the image.
491 Note: please refer to the manpage of resize.f2fs(8) to get full option list.
493 defrag.f2fs
494 -----------
495 The defrag.f2fs can be used to defragment scattered written data as well as
496 filesystem metadata across the disk. This can improve the write speed by giving
497 more free consecutive space.
499 Note: please refer to the manpage of defrag.f2fs(8) to get full option list.
501 f2fs_io
502 -------
503 The f2fs_io is a simple tool to issue various filesystem APIs as well as
504 f2fs-specific ones, which is very useful for QA tests.
506 Note: please refer to the manpage of f2fs_io(8) to get full option list.
508 Design
509 ======
511 On-disk Layout
512 --------------
514 F2FS divides the whole volume into a number of segments, each of which is fixed
515 to 2MB in size. A section is composed of consecutive segments, and a zone
516 consists of a set of sections. By default, section and zone sizes are set to one
517 segment size identically, but users can easily modify the sizes by mkfs.
519 F2FS splits the entire volume into six areas, and all the areas except superblock
520 consist of multiple segments as described below::
522 align with the zone size <-|
523 |-> align with the segment size
524 _________________________________________________________________________
525 | | | Segment | Node | Segment | |
526 | Superblock | Checkpoint | Info. | Address | Summary | Main |
527 | (SB) | (CP) | Table (SIT) | Table (NAT) | Area (SSA) | |
528 |____________|_____2______|______N______|______N______|______N_____|__N___|
529 . .
530 . .
531 . .
532 ._________________________________________.
533 |_Segment_|_..._|_Segment_|_..._|_Segment_|
534 . .
535 ._________._________
536 |_section_|__...__|_
537 . .
538 .________.
539 |__zone__|
541 - Superblock (SB)
542 It is located at the beginning of the partition, and there exist two copies
543 to avoid file system crash. It contains basic partition information and some
544 default parameters of f2fs.
546 - Checkpoint (CP)
547 It contains file system information, bitmaps for valid NAT/SIT sets, orphan
548 inode lists, and summary entries of current active segments.
550 - Segment Information Table (SIT)
551 It contains segment information such as valid block count and bitmap for the
552 validity of all the blocks.
554 - Node Address Table (NAT)
555 It is composed of a block address table for all the node blocks stored in
556 Main area.
558 - Segment Summary Area (SSA)
559 It contains summary entries which contains the owner information of all the
560 data and node blocks stored in Main area.
562 - Main Area
563 It contains file and directory data including their indices.
565 In order to avoid misalignment between file system and flash-based storage, F2FS
566 aligns the start block address of CP with the segment size. Also, it aligns the
567 start block address of Main area with the zone size by reserving some segments
568 in SSA area.
570 Reference the following survey for additional technical details.
571 https://wiki.linaro.org/WorkingGroups/Kernel/Projects/FlashCardSurvey
573 File System Metadata Structure
574 ------------------------------
576 F2FS adopts the checkpointing scheme to maintain file system consistency. At
577 mount time, F2FS first tries to find the last valid checkpoint data by scanning
578 CP area. In order to reduce the scanning time, F2FS uses only two copies of CP.
579 One of them always indicates the last valid data, which is called as shadow copy
580 mechanism. In addition to CP, NAT and SIT also adopt the shadow copy mechanism.
582 For file system consistency, each CP points to which NAT and SIT copies are
583 valid, as shown as below::
585 +--------+----------+---------+
586 | CP | SIT | NAT |
587 +--------+----------+---------+
588 . . . .
589 . . . .
590 . . . .
591 +-------+-------+--------+--------+--------+--------+
592 | CP #0 | CP #1 | SIT #0 | SIT #1 | NAT #0 | NAT #1 |
593 +-------+-------+--------+--------+--------+--------+
594 | ^ ^
595 | | |
596 `----------------------------------------'
598 Index Structure
599 ---------------
601 The key data structure to manage the data locations is a "node". Similar to
602 traditional file structures, F2FS has three types of node: inode, direct node,
603 indirect node. F2FS assigns 4KB to an inode block which contains 923 data block
604 indices, two direct node pointers, two indirect node pointers, and one double
605 indirect node pointer as described below. One direct node block contains 1018
606 data blocks, and one indirect node block contains also 1018 node blocks. Thus,
607 one inode block (i.e., a file) covers::
609 4KB * (923 + 2 * 1018 + 2 * 1018 * 1018 + 1018 * 1018 * 1018) := 3.94TB.
611 Inode block (4KB)
612 |- data (923)
613 |- direct node (2)
614 | `- data (1018)
615 |- indirect node (2)
616 | `- direct node (1018)
617 | `- data (1018)
618 `- double indirect node (1)
619 `- indirect node (1018)
620 `- direct node (1018)
621 `- data (1018)
623 Note that all the node blocks are mapped by NAT which means the location of
624 each node is translated by the NAT table. In the consideration of the wandering
625 tree problem, F2FS is able to cut off the propagation of node updates caused by
626 leaf data writes.
628 Directory Structure
629 -------------------
631 A directory entry occupies 11 bytes, which consists of the following attributes.
633 - hash hash value of the file name
634 - ino inode number
635 - len the length of file name
636 - type file type such as directory, symlink, etc
638 A dentry block consists of 214 dentry slots and file names. Therein a bitmap is
639 used to represent whether each dentry is valid or not. A dentry block occupies
640 4KB with the following composition.
642 ::
644 Dentry Block(4 K) = bitmap (27 bytes) + reserved (3 bytes) +
645 dentries(11 * 214 bytes) + file name (8 * 214 bytes)
647 [Bucket]
648 +--------------------------------+
649 |dentry block 1 | dentry block 2 |
650 +--------------------------------+
651 . .
652 . .
653 . [Dentry Block Structure: 4KB] .
654 +--------+----------+----------+------------+
655 | bitmap | reserved | dentries | file names |
656 +--------+----------+----------+------------+
657 [Dentry Block: 4KB] . .
658 . .
659 . .
660 +------+------+-----+------+
661 | hash | ino | len | type |
662 +------+------+-----+------+
663 [Dentry Structure: 11 bytes]
665 F2FS implements multi-level hash tables for directory structure. Each level has
666 a hash table with dedicated number of hash buckets as shown below. Note that
667 "A(2B)" means a bucket includes 2 data blocks.
669 ::
671 ----------------------
672 A : bucket
673 B : block
674 N : MAX_DIR_HASH_DEPTH
675 ----------------------
677 level #0 | A(2B)
678 |
679 level #1 | A(2B) - A(2B)
680 |
681 level #2 | A(2B) - A(2B) - A(2B) - A(2B)
682 . | . . . .
683 level #N/2 | A(2B) - A(2B) - A(2B) - A(2B) - A(2B) - ... - A(2B)
684 . | . . . .
685 level #N | A(4B) - A(4B) - A(4B) - A(4B) - A(4B) - ... - A(4B)
687 The number of blocks and buckets are determined by::
689 ,- 2, if n < MAX_DIR_HASH_DEPTH / 2,
690 # of blocks in level #n = |
691 `- 4, Otherwise
693 ,- 2^(n + dir_level),
694 | if n + dir_level < MAX_DIR_HASH_DEPTH / 2,
695 # of buckets in level #n = |
696 `- 2^((MAX_DIR_HASH_DEPTH / 2) - 1),
697 Otherwise
699 When F2FS finds a file name in a directory, at first a hash value of the file
700 name is calculated. Then, F2FS scans the hash table in level #0 to find the
701 dentry consisting of the file name and its inode number. If not found, F2FS
702 scans the next hash table in level #1. In this way, F2FS scans hash tables in
703 each levels incrementally from 1 to N. In each level F2FS needs to scan only
704 one bucket determined by the following equation, which shows O(log(# of files))
705 complexity::
707 bucket number to scan in level #n = (hash value) % (# of buckets in level #n)
709 In the case of file creation, F2FS finds empty consecutive slots that cover the
710 file name. F2FS searches the empty slots in the hash tables of whole levels from
711 1 to N in the same way as the lookup operation.
713 The following figure shows an example of two cases holding children::
715 --------------> Dir <--------------
716 | |
717 child child
719 child - child [hole] - child
721 child - child - child [hole] - [hole] - child
723 Case 1: Case 2:
724 Number of children = 6, Number of children = 3,
725 File size = 7 File size = 7
727 Default Block Allocation
728 ------------------------
730 At runtime, F2FS manages six active logs inside "Main" area: Hot/Warm/Cold node
731 and Hot/Warm/Cold data.
733 - Hot node contains direct node blocks of directories.
734 - Warm node contains direct node blocks except hot node blocks.
735 - Cold node contains indirect node blocks
736 - Hot data contains dentry blocks
737 - Warm data contains data blocks except hot and cold data blocks
738 - Cold data contains multimedia data or migrated data blocks
740 LFS has two schemes for free space management: threaded log and copy-and-compac-
741 tion. The copy-and-compaction scheme which is known as cleaning, is well-suited
742 for devices showing very good sequential write performance, since free segments
743 are served all the time for writing new data. However, it suffers from cleaning
744 overhead under high utilization. Contrarily, the threaded log scheme suffers
745 from random writes, but no cleaning process is needed. F2FS adopts a hybrid
746 scheme where the copy-and-compaction scheme is adopted by default, but the
747 policy is dynamically changed to the threaded log scheme according to the file
748 system status.
750 In order to align F2FS with underlying flash-based storage, F2FS allocates a
751 segment in a unit of section. F2FS expects that the section size would be the
752 same as the unit size of garbage collection in FTL. Furthermore, with respect
753 to the mapping granularity in FTL, F2FS allocates each section of the active
754 logs from different zones as much as possible, since FTL can write the data in
755 the active logs into one allocation unit according to its mapping granularity.
757 Cleaning process
758 ----------------
760 F2FS does cleaning both on demand and in the background. On-demand cleaning is
761 triggered when there are not enough free segments to serve VFS calls. Background
762 cleaner is operated by a kernel thread, and triggers the cleaning job when the
763 system is idle.
765 F2FS supports two victim selection policies: greedy and cost-benefit algorithms.
766 In the greedy algorithm, F2FS selects a victim segment having the smallest number
767 of valid blocks. In the cost-benefit algorithm, F2FS selects a victim segment
768 according to the segment age and the number of valid blocks in order to address
769 log block thrashing problem in the greedy algorithm. F2FS adopts the greedy
770 algorithm for on-demand cleaner, while background cleaner adopts cost-benefit
771 algorithm.
773 In order to identify whether the data in the victim segment are valid or not,
774 F2FS manages a bitmap. Each bit represents the validity of a block, and the
775 bitmap is composed of a bit stream covering whole blocks in main area.
777 Write-hint Policy
778 -----------------
780 F2FS sets the whint all the time with the below policy.
782 ===================== ======================== ===================
783 User F2FS Block
784 ===================== ======================== ===================
785 N/A META WRITE_LIFE_NONE|REQ_META
786 N/A HOT_NODE WRITE_LIFE_NONE
787 N/A WARM_NODE WRITE_LIFE_MEDIUM
788 N/A COLD_NODE WRITE_LIFE_LONG
789 ioctl(COLD) COLD_DATA WRITE_LIFE_EXTREME
790 extension list " "
792 -- buffered io
793 N/A COLD_DATA WRITE_LIFE_EXTREME
794 N/A HOT_DATA WRITE_LIFE_SHORT
795 N/A WARM_DATA WRITE_LIFE_NOT_SET
797 -- direct io
798 WRITE_LIFE_EXTREME COLD_DATA WRITE_LIFE_EXTREME
799 WRITE_LIFE_SHORT HOT_DATA WRITE_LIFE_SHORT
800 WRITE_LIFE_NOT_SET WARM_DATA WRITE_LIFE_NOT_SET
801 WRITE_LIFE_NONE " WRITE_LIFE_NONE
802 WRITE_LIFE_MEDIUM " WRITE_LIFE_MEDIUM
803 WRITE_LIFE_LONG " WRITE_LIFE_LONG
804 ===================== ======================== ===================
806 Fallocate(2) Policy
807 -------------------
809 The default policy follows the below POSIX rule.
811 Allocating disk space
812 The default operation (i.e., mode is zero) of fallocate() allocates
813 the disk space within the range specified by offset and len. The
814 file size (as reported by stat(2)) will be changed if offset+len is
815 greater than the file size. Any subregion within the range specified
816 by offset and len that did not contain data before the call will be
817 initialized to zero. This default behavior closely resembles the
818 behavior of the posix_fallocate(3) library function, and is intended
819 as a method of optimally implementing that function.
821 However, once F2FS receives ioctl(fd, F2FS_IOC_SET_PIN_FILE) in prior to
822 fallocate(fd, DEFAULT_MODE), it allocates on-disk block addresses having
823 zero or random data, which is useful to the below scenario where:
825 1. create(fd)
826 2. ioctl(fd, F2FS_IOC_SET_PIN_FILE)
827 3. fallocate(fd, 0, 0, size)
828 4. address = fibmap(fd, offset)
829 5. open(blkdev)
830 6. write(blkdev, address)
832 Compression implementation
833 --------------------------
835 - New term named cluster is defined as basic unit of compression, file can
836 be divided into multiple clusters logically. One cluster includes 4 << n
837 (n >= 0) logical pages, compression size is also cluster size, each of
838 cluster can be compressed or not.
840 - In cluster metadata layout, one special block address is used to indicate
841 a cluster is a compressed one or normal one; for compressed cluster, following
842 metadata maps cluster to [1, 4 << n - 1] physical blocks, in where f2fs
843 stores data including compress header and compressed data.
845 - In order to eliminate write amplification during overwrite, F2FS only
846 support compression on write-once file, data can be compressed only when
847 all logical blocks in cluster contain valid data and compress ratio of
848 cluster data is lower than specified threshold.
850 - To enable compression on regular inode, there are four ways:
852 * chattr +c file
853 * chattr +c dir; touch dir/file
854 * mount w/ -o compress_extension=ext; touch file.ext
855 * mount w/ -o compress_extension=*; touch any_file
857 - To disable compression on regular inode, there are two ways:
859 * chattr -c file
860 * mount w/ -o nocompress_extension=ext; touch file.ext
862 - Priority in between FS_COMPR_FL, FS_NOCOMP_FS, extensions:
864 * compress_extension=so; nocompress_extension=zip; chattr +c dir; touch
865 dir/foo.so; touch dir/bar.zip; touch dir/baz.txt; then foo.so and baz.txt
866 should be compresse, bar.zip should be non-compressed. chattr +c dir/bar.zip
867 can enable compress on bar.zip.
868 * compress_extension=so; nocompress_extension=zip; chattr -c dir; touch
869 dir/foo.so; touch dir/bar.zip; touch dir/baz.txt; then foo.so should be
870 compresse, bar.zip and baz.txt should be non-compressed.
871 chattr+c dir/bar.zip; chattr+c dir/baz.txt; can enable compress on bar.zip
872 and baz.txt.
874 - At this point, compression feature doesn't expose compressed space to user
875 directly in order to guarantee potential data updates later to the space.
876 Instead, the main goal is to reduce data writes to flash disk as much as
877 possible, resulting in extending disk life time as well as relaxing IO
878 congestion. Alternatively, we've added ioctl(F2FS_IOC_RELEASE_COMPRESS_BLOCKS)
879 interface to reclaim compressed space and show it to user after setting a
880 special flag to the inode. Once the compressed space is released, the flag
881 will block writing data to the file until either the compressed space is
882 reserved via ioctl(F2FS_IOC_RESERVE_COMPRESS_BLOCKS) or the file size is
883 truncated to zero.
885 Compress metadata layout::
887 [Dnode Structure]
888 +-----------------------------------------------+
889 | cluster 1 | cluster 2 | ......... | cluster N |
890 +-----------------------------------------------+
891 . . . .
892 . . . .
893 . Compressed Cluster . . Normal Cluster .
894 +----------+---------+---------+---------+ +---------+---------+---------+---------+
895 |compr flag| block 1 | block 2 | block 3 | | block 1 | block 2 | block 3 | block 4 |
896 +----------+---------+---------+---------+ +---------+---------+---------+---------+
897 . .
898 . .
899 . .
900 +-------------+-------------+----------+----------------------------+
901 | data length | data chksum | reserved | compressed data |
902 +-------------+-------------+----------+----------------------------+
904 Compression mode
905 --------------------------
907 f2fs supports "fs" and "user" compression modes with "compression_mode" mount option.
908 With this option, f2fs provides a choice to select the way how to compress the
909 compression enabled files (refer to "Compression implementation" section for how to
910 enable compression on a regular inode).
912 1) compress_mode=fs
913 This is the default option. f2fs does automatic compression in the writeback of the
914 compression enabled files.
916 2) compress_mode=user
917 This disables the automatic compression and gives the user discretion of choosing the
918 target file and the timing. The user can do manual compression/decompression on the
919 compression enabled files using F2FS_IOC_DECOMPRESS_FILE and F2FS_IOC_COMPRESS_FILE
920 ioctls like the below.
922 To decompress a file,
924 fd = open(filename, O_WRONLY, 0);
925 ret = ioctl(fd, F2FS_IOC_DECOMPRESS_FILE);
927 To compress a file,
929 fd = open(filename, O_WRONLY, 0);
930 ret = ioctl(fd, F2FS_IOC_COMPRESS_FILE);
932 NVMe Zoned Namespace devices
933 ----------------------------
935 - ZNS defines a per-zone capacity which can be equal or less than the
936 zone-size. Zone-capacity is the number of usable blocks in the zone.
937 F2FS checks if zone-capacity is less than zone-size, if it is, then any
938 segment which starts after the zone-capacity is marked as not-free in
939 the free segment bitmap at initial mount time. These segments are marked
940 as permanently used so they are not allocated for writes and
941 consequently are not needed to be garbage collected. In case the
942 zone-capacity is not aligned to default segment size(2MB), then a segment
943 can start before the zone-capacity and span across zone-capacity boundary.
944 Such spanning segments are also considered as usable segments. All blocks
945 past the zone-capacity are considered unusable in these segments.
947 Device aliasing feature
948 -----------------------
950 f2fs can utilize a special file called a "device aliasing file." This file allows
951 the entire storage device to be mapped with a single, large extent, not using
952 the usual f2fs node structures. This mapped area is pinned and primarily intended
953 for holding the space.
955 Essentially, this mechanism allows a portion of the f2fs area to be temporarily
956 reserved and used by another filesystem or for different purposes. Once that
957 external usage is complete, the device aliasing file can be deleted, releasing
958 the reserved space back to F2FS for its own use.
960 <use-case>
962 # ls /dev/vd*
963 /dev/vdb (32GB) /dev/vdc (32GB)
964 # mkfs.ext4 /dev/vdc
965 # mkfs.f2fs -c /dev/vdc@vdc.file /dev/vdb
966 # mount /dev/vdb /mnt/f2fs
967 # ls -l /mnt/f2fs
968 vdc.file
969 # df -h
970 /dev/vdb 64G 33G 32G 52% /mnt/f2fs
972 # mount -o loop /dev/vdc /mnt/ext4
973 # df -h
974 /dev/vdb 64G 33G 32G 52% /mnt/f2fs
975 /dev/loop7 32G 24K 30G 1% /mnt/ext4
976 # umount /mnt/ext4
978 # f2fs_io getflags /mnt/f2fs/vdc.file
979 get a flag on /mnt/f2fs/vdc.file ret=0, flags=nocow(pinned),immutable
980 # f2fs_io setflags noimmutable /mnt/f2fs/vdc.file
981 get a flag on noimmutable ret=0, flags=800010
982 set a flag on /mnt/f2fs/vdc.file ret=0, flags=noimmutable
983 # rm /mnt/f2fs/vdc.file
984 # df -h
985 /dev/vdb 64G 753M 64G 2% /mnt/f2fs
987 So, the key idea is, user can do any file operations on /dev/vdc, and
988 reclaim the space after the use, while the space is counted as /data.
989 That doesn't require modifying partition size and filesystem format.