0001 .. SPDX-License-Identifier: GPL-2.0
0002
0003 ==========================================
0004 WHAT IS Flash-Friendly File System (F2FS)?
0005 ==========================================
0006
0007 NAND flash memory-based storage devices, such as SSD, eMMC, and SD cards, have
0008 been equipped on a variety systems ranging from mobile to server systems. Since
0009 they are known to have different characteristics from the conventional rotating
0010 disks, a file system, an upper layer to the storage device, should adapt to the
0011 changes from the sketch in the design level.
0012
0013 F2FS is a file system exploiting NAND flash memory-based storage devices, which
0014 is based on Log-structured File System (LFS). The design has been focused on
0015 addressing the fundamental issues in LFS, which are snowball effect of wandering
0016 tree and high cleaning overhead.
0017
0018 Since a NAND flash memory-based storage device shows different characteristic
0019 according to its internal geometry or flash memory management scheme, namely FTL,
0020 F2FS and its tools support various parameters not only for configuring on-disk
0021 layout, but also for selecting allocation and cleaning algorithms.
0022
0023 The following git tree provides the file system formatting tool (mkfs.f2fs),
0024 a consistency checking tool (fsck.f2fs), and a debugging tool (dump.f2fs).
0025
0026 - git://git.kernel.org/pub/scm/linux/kernel/git/jaegeuk/f2fs-tools.git
0027
0028 For reporting bugs and sending patches, please use the following mailing list:
0029
0030 - linux-f2fs-devel@lists.sourceforge.net
0031
0032 Background and Design issues
0033 ============================
0034
0035 Log-structured File System (LFS)
0036 --------------------------------
0037 "A log-structured file system writes all modifications to disk sequentially in
0038 a log-like structure, thereby speeding up both file writing and crash recovery.
0039 The log is the only structure on disk; it contains indexing information so that
0040 files can be read back from the log efficiently. In order to maintain large free
0041 areas on disk for fast writing, we divide the log into segments and use a
0042 segment cleaner to compress the live information from heavily fragmented
0043 segments." from Rosenblum, M. and Ousterhout, J. K., 1992, "The design and
0044 implementation of a log-structured file system", ACM Trans. Computer Systems
0045 10, 1, 26–52.
0046
0047 Wandering Tree Problem
0048 ----------------------
0049 In LFS, when a file data is updated and written to the end of log, its direct
0050 pointer block is updated due to the changed location. Then the indirect pointer
0051 block is also updated due to the direct pointer block update. In this manner,
0052 the upper index structures such as inode, inode map, and checkpoint block are
0053 also updated recursively. This problem is called as wandering tree problem [1],
0054 and in order to enhance the performance, it should eliminate or relax the update
0055 propagation as much as possible.
0056
0057 [1] Bityutskiy, A. 2005. JFFS3 design issues. http://www.linux-mtd.infradead.org/
0058
0059 Cleaning Overhead
0060 -----------------
0061 Since LFS is based on out-of-place writes, it produces so many obsolete blocks
0062 scattered across the whole storage. In order to serve new empty log space, it
0063 needs to reclaim these obsolete blocks seamlessly to users. This job is called
0064 as a cleaning process.
0065
0066 The process consists of three operations as follows.
0067
0068 1. A victim segment is selected through referencing segment usage table.
0069 2. It loads parent index structures of all the data in the victim identified by
0070 segment summary blocks.
0071 3. It checks the cross-reference between the data and its parent index structure.
0072 4. It moves valid data selectively.
0073
0074 This cleaning job may cause unexpected long delays, so the most important goal
0075 is to hide the latencies to users. And also definitely, it should reduce the
0076 amount of valid data to be moved, and move them quickly as well.
0077
0078 Key Features
0079 ============
0080
0081 Flash Awareness
0082 ---------------
0083 - Enlarge the random write area for better performance, but provide the high
0084 spatial locality
0085 - Align FS data structures to the operational units in FTL as best efforts
0086
0087 Wandering Tree Problem
0088 ----------------------
0089 - Use a term, “node”, that represents inodes as well as various pointer blocks
0090 - Introduce Node Address Table (NAT) containing the locations of all the “node”
0091 blocks; this will cut off the update propagation.
0092
0093 Cleaning Overhead
0094 -----------------
0095 - Support a background cleaning process
0096 - Support greedy and cost-benefit algorithms for victim selection policies
0097 - Support multi-head logs for static/dynamic hot and cold data separation
0098 - Introduce adaptive logging for efficient block allocation
0099
0100 Mount Options
0101 =============
0102
0103
0104 ======================== ============================================================
0105 background_gc=%s Turn on/off cleaning operations, namely garbage
0106 collection, triggered in background when I/O subsystem is
0107 idle. If background_gc=on, it will turn on the garbage
0108 collection and if background_gc=off, garbage collection
0109 will be turned off. If background_gc=sync, it will turn
0110 on synchronous garbage collection running in background.
0111 Default value for this option is on. So garbage
0112 collection is on by default.
0113 gc_merge When background_gc is on, this option can be enabled to
0114 let background GC thread to handle foreground GC requests,
0115 it can eliminate the sluggish issue caused by slow foreground
0116 GC operation when GC is triggered from a process with limited
0117 I/O and CPU resources.
0118 nogc_merge Disable GC merge feature.
0119 disable_roll_forward Disable the roll-forward recovery routine
0120 norecovery Disable the roll-forward recovery routine, mounted read-
0121 only (i.e., -o ro,disable_roll_forward)
0122 discard/nodiscard Enable/disable real-time discard in f2fs, if discard is
0123 enabled, f2fs will issue discard/TRIM commands when a
0124 segment is cleaned.
0125 no_heap Disable heap-style segment allocation which finds free
0126 segments for data from the beginning of main area, while
0127 for node from the end of main area.
0128 nouser_xattr Disable Extended User Attributes. Note: xattr is enabled
0129 by default if CONFIG_F2FS_FS_XATTR is selected.
0130 noacl Disable POSIX Access Control List. Note: acl is enabled
0131 by default if CONFIG_F2FS_FS_POSIX_ACL is selected.
0132 active_logs=%u Support configuring the number of active logs. In the
0133 current design, f2fs supports only 2, 4, and 6 logs.
0134 Default number is 6.
0135 disable_ext_identify Disable the extension list configured by mkfs, so f2fs
0136 is not aware of cold files such as media files.
0137 inline_xattr Enable the inline xattrs feature.
0138 noinline_xattr Disable the inline xattrs feature.
0139 inline_xattr_size=%u Support configuring inline xattr size, it depends on
0140 flexible inline xattr feature.
0141 inline_data Enable the inline data feature: Newly created small (<~3.4k)
0142 files can be written into inode block.
0143 inline_dentry Enable the inline dir feature: data in newly created
0144 directory entries can be written into inode block. The
0145 space of inode block which is used to store inline
0146 dentries is limited to ~3.4k.
0147 noinline_dentry Disable the inline dentry feature.
0148 flush_merge Merge concurrent cache_flush commands as much as possible
0149 to eliminate redundant command issues. If the underlying
0150 device handles the cache_flush command relatively slowly,
0151 recommend to enable this option.
0152 nobarrier This option can be used if underlying storage guarantees
0153 its cached data should be written to the novolatile area.
0154 If this option is set, no cache_flush commands are issued
0155 but f2fs still guarantees the write ordering of all the
0156 data writes.
0157 fastboot This option is used when a system wants to reduce mount
0158 time as much as possible, even though normal performance
0159 can be sacrificed.
0160 extent_cache Enable an extent cache based on rb-tree, it can cache
0161 as many as extent which map between contiguous logical
0162 address and physical address per inode, resulting in
0163 increasing the cache hit ratio. Set by default.
0164 noextent_cache Disable an extent cache based on rb-tree explicitly, see
0165 the above extent_cache mount option.
0166 noinline_data Disable the inline data feature, inline data feature is
0167 enabled by default.
0168 data_flush Enable data flushing before checkpoint in order to
0169 persist data of regular and symlink.
0170 reserve_root=%d Support configuring reserved space which is used for
0171 allocation from a privileged user with specified uid or
0172 gid, unit: 4KB, the default limit is 0.2% of user blocks.
0173 resuid=%d The user ID which may use the reserved blocks.
0174 resgid=%d The group ID which may use the reserved blocks.
0175 fault_injection=%d Enable fault injection in all supported types with
0176 specified injection rate.
0177 fault_type=%d Support configuring fault injection type, should be
0178 enabled with fault_injection option, fault type value
0179 is shown below, it supports single or combined type.
0180
0181 =================== ===========
0182 Type_Name Type_Value
0183 =================== ===========
0184 FAULT_KMALLOC 0x000000001
0185 FAULT_KVMALLOC 0x000000002
0186 FAULT_PAGE_ALLOC 0x000000004
0187 FAULT_PAGE_GET 0x000000008
0188 FAULT_ALLOC_BIO 0x000000010 (obsolete)
0189 FAULT_ALLOC_NID 0x000000020
0190 FAULT_ORPHAN 0x000000040
0191 FAULT_BLOCK 0x000000080
0192 FAULT_DIR_DEPTH 0x000000100
0193 FAULT_EVICT_INODE 0x000000200
0194 FAULT_TRUNCATE 0x000000400
0195 FAULT_READ_IO 0x000000800
0196 FAULT_CHECKPOINT 0x000001000
0197 FAULT_DISCARD 0x000002000
0198 FAULT_WRITE_IO 0x000004000
0199 FAULT_SLAB_ALLOC 0x000008000
0200 FAULT_DQUOT_INIT 0x000010000
0201 FAULT_LOCK_OP 0x000020000
0202 =================== ===========
0203 mode=%s Control block allocation mode which supports "adaptive"
0204 and "lfs". In "lfs" mode, there should be no random
0205 writes towards main area.
0206 "fragment:segment" and "fragment:block" are newly added here.
0207 These are developer options for experiments to simulate filesystem
0208 fragmentation/after-GC situation itself. The developers use these
0209 modes to understand filesystem fragmentation/after-GC condition well,
0210 and eventually get some insights to handle them better.
0211 In "fragment:segment", f2fs allocates a new segment in ramdom
0212 position. With this, we can simulate the after-GC condition.
0213 In "fragment:block", we can scatter block allocation with
0214 "max_fragment_chunk" and "max_fragment_hole" sysfs nodes.
0215 We added some randomness to both chunk and hole size to make
0216 it close to realistic IO pattern. So, in this mode, f2fs will allocate
0217 1..<max_fragment_chunk> blocks in a chunk and make a hole in the
0218 length of 1..<max_fragment_hole> by turns. With this, the newly
0219 allocated blocks will be scattered throughout the whole partition.
0220 Note that "fragment:block" implicitly enables "fragment:segment"
0221 option for more randomness.
0222 Please, use these options for your experiments and we strongly
0223 recommend to re-format the filesystem after using these options.
0224 io_bits=%u Set the bit size of write IO requests. It should be set
0225 with "mode=lfs".
0226 usrquota Enable plain user disk quota accounting.
0227 grpquota Enable plain group disk quota accounting.
0228 prjquota Enable plain project quota accounting.
0229 usrjquota=<file> Appoint specified file and type during mount, so that quota
0230 grpjquota=<file> information can be properly updated during recovery flow,
0231 prjjquota=<file> <quota file>: must be in root directory;
0232 jqfmt=<quota type> <quota type>: [vfsold,vfsv0,vfsv1].
0233 offusrjquota Turn off user journalled quota.
0234 offgrpjquota Turn off group journalled quota.
0235 offprjjquota Turn off project journalled quota.
0236 quota Enable plain user disk quota accounting.
0237 noquota Disable all plain disk quota option.
0238 alloc_mode=%s Adjust block allocation policy, which supports "reuse"
0239 and "default".
0240 fsync_mode=%s Control the policy of fsync. Currently supports "posix",
0241 "strict", and "nobarrier". In "posix" mode, which is
0242 default, fsync will follow POSIX semantics and does a
0243 light operation to improve the filesystem performance.
0244 In "strict" mode, fsync will be heavy and behaves in line
0245 with xfs, ext4 and btrfs, where xfstest generic/342 will
0246 pass, but the performance will regress. "nobarrier" is
0247 based on "posix", but doesn't issue flush command for
0248 non-atomic files likewise "nobarrier" mount option.
0249 test_dummy_encryption
0250 test_dummy_encryption=%s
0251 Enable dummy encryption, which provides a fake fscrypt
0252 context. The fake fscrypt context is used by xfstests.
0253 The argument may be either "v1" or "v2", in order to
0254 select the corresponding fscrypt policy version.
0255 checkpoint=%s[:%u[%]] Set to "disable" to turn off checkpointing. Set to "enable"
0256 to reenable checkpointing. Is enabled by default. While
0257 disabled, any unmounting or unexpected shutdowns will cause
0258 the filesystem contents to appear as they did when the
0259 filesystem was mounted with that option.
0260 While mounting with checkpoint=disabled, the filesystem must
0261 run garbage collection to ensure that all available space can
0262 be used. If this takes too much time, the mount may return
0263 EAGAIN. You may optionally add a value to indicate how much
0264 of the disk you would be willing to temporarily give up to
0265 avoid additional garbage collection. This can be given as a
0266 number of blocks, or as a percent. For instance, mounting
0267 with checkpoint=disable:100% would always succeed, but it may
0268 hide up to all remaining free space. The actual space that
0269 would be unusable can be viewed at /sys/fs/f2fs/<disk>/unusable
0270 This space is reclaimed once checkpoint=enable.
0271 checkpoint_merge When checkpoint is enabled, this can be used to create a kernel
0272 daemon and make it to merge concurrent checkpoint requests as
0273 much as possible to eliminate redundant checkpoint issues. Plus,
0274 we can eliminate the sluggish issue caused by slow checkpoint
0275 operation when the checkpoint is done in a process context in
0276 a cgroup having low i/o budget and cpu shares. To make this
0277 do better, we set the default i/o priority of the kernel daemon
0278 to "3", to give one higher priority than other kernel threads.
0279 This is the same way to give a I/O priority to the jbd2
0280 journaling thread of ext4 filesystem.
0281 nocheckpoint_merge Disable checkpoint merge feature.
0282 compress_algorithm=%s Control compress algorithm, currently f2fs supports "lzo",
0283 "lz4", "zstd" and "lzo-rle" algorithm.
0284 compress_algorithm=%s:%d Control compress algorithm and its compress level, now, only
0285 "lz4" and "zstd" support compress level config.
0286 algorithm level range
0287 lz4 3 - 16
0288 zstd 1 - 22
0289 compress_log_size=%u Support configuring compress cluster size, the size will
0290 be 4KB * (1 << %u), 16KB is minimum size, also it's
0291 default size.
0292 compress_extension=%s Support adding specified extension, so that f2fs can enable
0293 compression on those corresponding files, e.g. if all files
0294 with '.ext' has high compression rate, we can set the '.ext'
0295 on compression extension list and enable compression on
0296 these file by default rather than to enable it via ioctl.
0297 For other files, we can still enable compression via ioctl.
0298 Note that, there is one reserved special extension '*', it
0299 can be set to enable compression for all files.
0300 nocompress_extension=%s Support adding specified extension, so that f2fs can disable
0301 compression on those corresponding files, just contrary to compression extension.
0302 If you know exactly which files cannot be compressed, you can use this.
0303 The same extension name can't appear in both compress and nocompress
0304 extension at the same time.
0305 If the compress extension specifies all files, the types specified by the
0306 nocompress extension will be treated as special cases and will not be compressed.
0307 Don't allow use '*' to specifie all file in nocompress extension.
0308 After add nocompress_extension, the priority should be:
0309 dir_flag < comp_extention,nocompress_extension < comp_file_flag,no_comp_file_flag.
0310 See more in compression sections.
0311
0312 compress_chksum Support verifying chksum of raw data in compressed cluster.
0313 compress_mode=%s Control file compression mode. This supports "fs" and "user"
0314 modes. In "fs" mode (default), f2fs does automatic compression
0315 on the compression enabled files. In "user" mode, f2fs disables
0316 the automaic compression and gives the user discretion of
0317 choosing the target file and the timing. The user can do manual
0318 compression/decompression on the compression enabled files using
0319 ioctls.
0320 compress_cache Support to use address space of a filesystem managed inode to
0321 cache compressed block, in order to improve cache hit ratio of
0322 random read.
0323 inlinecrypt When possible, encrypt/decrypt the contents of encrypted
0324 files using the blk-crypto framework rather than
0325 filesystem-layer encryption. This allows the use of
0326 inline encryption hardware. The on-disk format is
0327 unaffected. For more details, see
0328 Documentation/block/inline-encryption.rst.
0329 atgc Enable age-threshold garbage collection, it provides high
0330 effectiveness and efficiency on background GC.
0331 discard_unit=%s Control discard unit, the argument can be "block", "segment"
0332 and "section", issued discard command's offset/size will be
0333 aligned to the unit, by default, "discard_unit=block" is set,
0334 so that small discard functionality is enabled.
0335 For blkzoned device, "discard_unit=section" will be set by
0336 default, it is helpful for large sized SMR or ZNS devices to
0337 reduce memory cost by getting rid of fs metadata supports small
0338 discard.
0339 memory=%s Control memory mode. This supports "normal" and "low" modes.
0340 "low" mode is introduced to support low memory devices.
0341 Because of the nature of low memory devices, in this mode, f2fs
0342 will try to save memory sometimes by sacrificing performance.
0343 "normal" mode is the default mode and same as before.
0344 ======================== ============================================================
0345
0346 Debugfs Entries
0347 ===============
0348
0349 /sys/kernel/debug/f2fs/ contains information about all the partitions mounted as
0350 f2fs. Each file shows the whole f2fs information.
0351
0352 /sys/kernel/debug/f2fs/status includes:
0353
0354 - major file system information managed by f2fs currently
0355 - average SIT information about whole segments
0356 - current memory footprint consumed by f2fs.
0357
0358 Sysfs Entries
0359 =============
0360
0361 Information about mounted f2fs file systems can be found in
0362 /sys/fs/f2fs. Each mounted filesystem will have a directory in
0363 /sys/fs/f2fs based on its device name (i.e., /sys/fs/f2fs/sda).
0364 The files in each per-device directory are shown in table below.
0365
0366 Files in /sys/fs/f2fs/<devname>
0367 (see also Documentation/ABI/testing/sysfs-fs-f2fs)
0368
0369 Usage
0370 =====
0371
0372 1. Download userland tools and compile them.
0373
0374 2. Skip, if f2fs was compiled statically inside kernel.
0375 Otherwise, insert the f2fs.ko module::
0376
0377 # insmod f2fs.ko
0378
0379 3. Create a directory to use when mounting::
0380
0381 # mkdir /mnt/f2fs
0382
0383 4. Format the block device, and then mount as f2fs::
0384
0385 # mkfs.f2fs -l label /dev/block_device
0386 # mount -t f2fs /dev/block_device /mnt/f2fs
0387
0388 mkfs.f2fs
0389 ---------
0390 The mkfs.f2fs is for the use of formatting a partition as the f2fs filesystem,
0391 which builds a basic on-disk layout.
0392
0393 The quick options consist of:
0394
0395 =============== ===========================================================
0396 ``-l [label]`` Give a volume label, up to 512 unicode name.
0397 ``-a [0 or 1]`` Split start location of each area for heap-based allocation.
0398
0399 1 is set by default, which performs this.
0400 ``-o [int]`` Set overprovision ratio in percent over volume size.
0401
0402 5 is set by default.
0403 ``-s [int]`` Set the number of segments per section.
0404
0405 1 is set by default.
0406 ``-z [int]`` Set the number of sections per zone.
0407
0408 1 is set by default.
0409 ``-e [str]`` Set basic extension list. e.g. "mp3,gif,mov"
0410 ``-t [0 or 1]`` Disable discard command or not.
0411
0412 1 is set by default, which conducts discard.
0413 =============== ===========================================================
0414
0415 Note: please refer to the manpage of mkfs.f2fs(8) to get full option list.
0416
0417 fsck.f2fs
0418 ---------
0419 The fsck.f2fs is a tool to check the consistency of an f2fs-formatted
0420 partition, which examines whether the filesystem metadata and user-made data
0421 are cross-referenced correctly or not.
0422 Note that, initial version of the tool does not fix any inconsistency.
0423
0424 The quick options consist of::
0425
0426 -d debug level [default:0]
0427
0428 Note: please refer to the manpage of fsck.f2fs(8) to get full option list.
0429
0430 dump.f2fs
0431 ---------
0432 The dump.f2fs shows the information of specific inode and dumps SSA and SIT to
0433 file. Each file is dump_ssa and dump_sit.
0434
0435 The dump.f2fs is used to debug on-disk data structures of the f2fs filesystem.
0436 It shows on-disk inode information recognized by a given inode number, and is
0437 able to dump all the SSA and SIT entries into predefined files, ./dump_ssa and
0438 ./dump_sit respectively.
0439
0440 The options consist of::
0441
0442 -d debug level [default:0]
0443 -i inode no (hex)
0444 -s [SIT dump segno from #1~#2 (decimal), for all 0~-1]
0445 -a [SSA dump segno from #1~#2 (decimal), for all 0~-1]
0446
0447 Examples::
0448
0449 # dump.f2fs -i [ino] /dev/sdx
0450 # dump.f2fs -s 0~-1 /dev/sdx (SIT dump)
0451 # dump.f2fs -a 0~-1 /dev/sdx (SSA dump)
0452
0453 Note: please refer to the manpage of dump.f2fs(8) to get full option list.
0454
0455 sload.f2fs
0456 ----------
0457 The sload.f2fs gives a way to insert files and directories in the exisiting disk
0458 image. This tool is useful when building f2fs images given compiled files.
0459
0460 Note: please refer to the manpage of sload.f2fs(8) to get full option list.
0461
0462 resize.f2fs
0463 -----------
0464 The resize.f2fs lets a user resize the f2fs-formatted disk image, while preserving
0465 all the files and directories stored in the image.
0466
0467 Note: please refer to the manpage of resize.f2fs(8) to get full option list.
0468
0469 defrag.f2fs
0470 -----------
0471 The defrag.f2fs can be used to defragment scattered written data as well as
0472 filesystem metadata across the disk. This can improve the write speed by giving
0473 more free consecutive space.
0474
0475 Note: please refer to the manpage of defrag.f2fs(8) to get full option list.
0476
0477 f2fs_io
0478 -------
0479 The f2fs_io is a simple tool to issue various filesystem APIs as well as
0480 f2fs-specific ones, which is very useful for QA tests.
0481
0482 Note: please refer to the manpage of f2fs_io(8) to get full option list.
0483
0484 Design
0485 ======
0486
0487 On-disk Layout
0488 --------------
0489
0490 F2FS divides the whole volume into a number of segments, each of which is fixed
0491 to 2MB in size. A section is composed of consecutive segments, and a zone
0492 consists of a set of sections. By default, section and zone sizes are set to one
0493 segment size identically, but users can easily modify the sizes by mkfs.
0494
0495 F2FS splits the entire volume into six areas, and all the areas except superblock
0496 consist of multiple segments as described below::
0497
0498 align with the zone size <-|
0499 |-> align with the segment size
0500 _________________________________________________________________________
0501 | | | Segment | Node | Segment | |
0502 | Superblock | Checkpoint | Info. | Address | Summary | Main |
0503 | (SB) | (CP) | Table (SIT) | Table (NAT) | Area (SSA) | |
0504 |____________|_____2______|______N______|______N______|______N_____|__N___|
0505 . .
0506 . .
0507 . .
0508 ._________________________________________.
0509 |_Segment_|_..._|_Segment_|_..._|_Segment_|
0510 . .
0511 ._________._________
0512 |_section_|__...__|_
0513 . .
0514 .________.
0515 |__zone__|
0516
0517 - Superblock (SB)
0518 It is located at the beginning of the partition, and there exist two copies
0519 to avoid file system crash. It contains basic partition information and some
0520 default parameters of f2fs.
0521
0522 - Checkpoint (CP)
0523 It contains file system information, bitmaps for valid NAT/SIT sets, orphan
0524 inode lists, and summary entries of current active segments.
0525
0526 - Segment Information Table (SIT)
0527 It contains segment information such as valid block count and bitmap for the
0528 validity of all the blocks.
0529
0530 - Node Address Table (NAT)
0531 It is composed of a block address table for all the node blocks stored in
0532 Main area.
0533
0534 - Segment Summary Area (SSA)
0535 It contains summary entries which contains the owner information of all the
0536 data and node blocks stored in Main area.
0537
0538 - Main Area
0539 It contains file and directory data including their indices.
0540
0541 In order to avoid misalignment between file system and flash-based storage, F2FS
0542 aligns the start block address of CP with the segment size. Also, it aligns the
0543 start block address of Main area with the zone size by reserving some segments
0544 in SSA area.
0545
0546 Reference the following survey for additional technical details.
0547 https://wiki.linaro.org/WorkingGroups/Kernel/Projects/FlashCardSurvey
0548
0549 File System Metadata Structure
0550 ------------------------------
0551
0552 F2FS adopts the checkpointing scheme to maintain file system consistency. At
0553 mount time, F2FS first tries to find the last valid checkpoint data by scanning
0554 CP area. In order to reduce the scanning time, F2FS uses only two copies of CP.
0555 One of them always indicates the last valid data, which is called as shadow copy
0556 mechanism. In addition to CP, NAT and SIT also adopt the shadow copy mechanism.
0557
0558 For file system consistency, each CP points to which NAT and SIT copies are
0559 valid, as shown as below::
0560
0561 +--------+----------+---------+
0562 | CP | SIT | NAT |
0563 +--------+----------+---------+
0564 . . . .
0565 . . . .
0566 . . . .
0567 +-------+-------+--------+--------+--------+--------+
0568 | CP #0 | CP #1 | SIT #0 | SIT #1 | NAT #0 | NAT #1 |
0569 +-------+-------+--------+--------+--------+--------+
0570 | ^ ^
0571 | | |
0572 `----------------------------------------'
0573
0574 Index Structure
0575 ---------------
0576
0577 The key data structure to manage the data locations is a "node". Similar to
0578 traditional file structures, F2FS has three types of node: inode, direct node,
0579 indirect node. F2FS assigns 4KB to an inode block which contains 923 data block
0580 indices, two direct node pointers, two indirect node pointers, and one double
0581 indirect node pointer as described below. One direct node block contains 1018
0582 data blocks, and one indirect node block contains also 1018 node blocks. Thus,
0583 one inode block (i.e., a file) covers::
0584
0585 4KB * (923 + 2 * 1018 + 2 * 1018 * 1018 + 1018 * 1018 * 1018) := 3.94TB.
0586
0587 Inode block (4KB)
0588 |- data (923)
0589 |- direct node (2)
0590 | `- data (1018)
0591 |- indirect node (2)
0592 | `- direct node (1018)
0593 | `- data (1018)
0594 `- double indirect node (1)
0595 `- indirect node (1018)
0596 `- direct node (1018)
0597 `- data (1018)
0598
0599 Note that all the node blocks are mapped by NAT which means the location of
0600 each node is translated by the NAT table. In the consideration of the wandering
0601 tree problem, F2FS is able to cut off the propagation of node updates caused by
0602 leaf data writes.
0603
0604 Directory Structure
0605 -------------------
0606
0607 A directory entry occupies 11 bytes, which consists of the following attributes.
0608
0609 - hash hash value of the file name
0610 - ino inode number
0611 - len the length of file name
0612 - type file type such as directory, symlink, etc
0613
0614 A dentry block consists of 214 dentry slots and file names. Therein a bitmap is
0615 used to represent whether each dentry is valid or not. A dentry block occupies
0616 4KB with the following composition.
0617
0618 ::
0619
0620 Dentry Block(4 K) = bitmap (27 bytes) + reserved (3 bytes) +
0621 dentries(11 * 214 bytes) + file name (8 * 214 bytes)
0622
0623 [Bucket]
0624 +--------------------------------+
0625 |dentry block 1 | dentry block 2 |
0626 +--------------------------------+
0627 . .
0628 . .
0629 . [Dentry Block Structure: 4KB] .
0630 +--------+----------+----------+------------+
0631 | bitmap | reserved | dentries | file names |
0632 +--------+----------+----------+------------+
0633 [Dentry Block: 4KB] . .
0634 . .
0635 . .
0636 +------+------+-----+------+
0637 | hash | ino | len | type |
0638 +------+------+-----+------+
0639 [Dentry Structure: 11 bytes]
0640
0641 F2FS implements multi-level hash tables for directory structure. Each level has
0642 a hash table with dedicated number of hash buckets as shown below. Note that
0643 "A(2B)" means a bucket includes 2 data blocks.
0644
0645 ::
0646
0647 ----------------------
0648 A : bucket
0649 B : block
0650 N : MAX_DIR_HASH_DEPTH
0651 ----------------------
0652
0653 level #0 | A(2B)
0654 |
0655 level #1 | A(2B) - A(2B)
0656 |
0657 level #2 | A(2B) - A(2B) - A(2B) - A(2B)
0658 . | . . . .
0659 level #N/2 | A(2B) - A(2B) - A(2B) - A(2B) - A(2B) - ... - A(2B)
0660 . | . . . .
0661 level #N | A(4B) - A(4B) - A(4B) - A(4B) - A(4B) - ... - A(4B)
0662
0663 The number of blocks and buckets are determined by::
0664
0665 ,- 2, if n < MAX_DIR_HASH_DEPTH / 2,
0666 # of blocks in level #n = |
0667 `- 4, Otherwise
0668
0669 ,- 2^(n + dir_level),
0670 | if n + dir_level < MAX_DIR_HASH_DEPTH / 2,
0671 # of buckets in level #n = |
0672 `- 2^((MAX_DIR_HASH_DEPTH / 2) - 1),
0673 Otherwise
0674
0675 When F2FS finds a file name in a directory, at first a hash value of the file
0676 name is calculated. Then, F2FS scans the hash table in level #0 to find the
0677 dentry consisting of the file name and its inode number. If not found, F2FS
0678 scans the next hash table in level #1. In this way, F2FS scans hash tables in
0679 each levels incrementally from 1 to N. In each level F2FS needs to scan only
0680 one bucket determined by the following equation, which shows O(log(# of files))
0681 complexity::
0682
0683 bucket number to scan in level #n = (hash value) % (# of buckets in level #n)
0684
0685 In the case of file creation, F2FS finds empty consecutive slots that cover the
0686 file name. F2FS searches the empty slots in the hash tables of whole levels from
0687 1 to N in the same way as the lookup operation.
0688
0689 The following figure shows an example of two cases holding children::
0690
0691 --------------> Dir <--------------
0692 | |
0693 child child
0694
0695 child - child [hole] - child
0696
0697 child - child - child [hole] - [hole] - child
0698
0699 Case 1: Case 2:
0700 Number of children = 6, Number of children = 3,
0701 File size = 7 File size = 7
0702
0703 Default Block Allocation
0704 ------------------------
0705
0706 At runtime, F2FS manages six active logs inside "Main" area: Hot/Warm/Cold node
0707 and Hot/Warm/Cold data.
0708
0709 - Hot node contains direct node blocks of directories.
0710 - Warm node contains direct node blocks except hot node blocks.
0711 - Cold node contains indirect node blocks
0712 - Hot data contains dentry blocks
0713 - Warm data contains data blocks except hot and cold data blocks
0714 - Cold data contains multimedia data or migrated data blocks
0715
0716 LFS has two schemes for free space management: threaded log and copy-and-compac-
0717 tion. The copy-and-compaction scheme which is known as cleaning, is well-suited
0718 for devices showing very good sequential write performance, since free segments
0719 are served all the time for writing new data. However, it suffers from cleaning
0720 overhead under high utilization. Contrarily, the threaded log scheme suffers
0721 from random writes, but no cleaning process is needed. F2FS adopts a hybrid
0722 scheme where the copy-and-compaction scheme is adopted by default, but the
0723 policy is dynamically changed to the threaded log scheme according to the file
0724 system status.
0725
0726 In order to align F2FS with underlying flash-based storage, F2FS allocates a
0727 segment in a unit of section. F2FS expects that the section size would be the
0728 same as the unit size of garbage collection in FTL. Furthermore, with respect
0729 to the mapping granularity in FTL, F2FS allocates each section of the active
0730 logs from different zones as much as possible, since FTL can write the data in
0731 the active logs into one allocation unit according to its mapping granularity.
0732
0733 Cleaning process
0734 ----------------
0735
0736 F2FS does cleaning both on demand and in the background. On-demand cleaning is
0737 triggered when there are not enough free segments to serve VFS calls. Background
0738 cleaner is operated by a kernel thread, and triggers the cleaning job when the
0739 system is idle.
0740
0741 F2FS supports two victim selection policies: greedy and cost-benefit algorithms.
0742 In the greedy algorithm, F2FS selects a victim segment having the smallest number
0743 of valid blocks. In the cost-benefit algorithm, F2FS selects a victim segment
0744 according to the segment age and the number of valid blocks in order to address
0745 log block thrashing problem in the greedy algorithm. F2FS adopts the greedy
0746 algorithm for on-demand cleaner, while background cleaner adopts cost-benefit
0747 algorithm.
0748
0749 In order to identify whether the data in the victim segment are valid or not,
0750 F2FS manages a bitmap. Each bit represents the validity of a block, and the
0751 bitmap is composed of a bit stream covering whole blocks in main area.
0752
0753 Fallocate(2) Policy
0754 -------------------
0755
0756 The default policy follows the below POSIX rule.
0757
0758 Allocating disk space
0759 The default operation (i.e., mode is zero) of fallocate() allocates
0760 the disk space within the range specified by offset and len. The
0761 file size (as reported by stat(2)) will be changed if offset+len is
0762 greater than the file size. Any subregion within the range specified
0763 by offset and len that did not contain data before the call will be
0764 initialized to zero. This default behavior closely resembles the
0765 behavior of the posix_fallocate(3) library function, and is intended
0766 as a method of optimally implementing that function.
0767
0768 However, once F2FS receives ioctl(fd, F2FS_IOC_SET_PIN_FILE) in prior to
0769 fallocate(fd, DEFAULT_MODE), it allocates on-disk block addressess having
0770 zero or random data, which is useful to the below scenario where:
0771
0772 1. create(fd)
0773 2. ioctl(fd, F2FS_IOC_SET_PIN_FILE)
0774 3. fallocate(fd, 0, 0, size)
0775 4. address = fibmap(fd, offset)
0776 5. open(blkdev)
0777 6. write(blkdev, address)
0778
0779 Compression implementation
0780 --------------------------
0781
0782 - New term named cluster is defined as basic unit of compression, file can
0783 be divided into multiple clusters logically. One cluster includes 4 << n
0784 (n >= 0) logical pages, compression size is also cluster size, each of
0785 cluster can be compressed or not.
0786
0787 - In cluster metadata layout, one special block address is used to indicate
0788 a cluster is a compressed one or normal one; for compressed cluster, following
0789 metadata maps cluster to [1, 4 << n - 1] physical blocks, in where f2fs
0790 stores data including compress header and compressed data.
0791
0792 - In order to eliminate write amplification during overwrite, F2FS only
0793 support compression on write-once file, data can be compressed only when
0794 all logical blocks in cluster contain valid data and compress ratio of
0795 cluster data is lower than specified threshold.
0796
0797 - To enable compression on regular inode, there are four ways:
0798
0799 * chattr +c file
0800 * chattr +c dir; touch dir/file
0801 * mount w/ -o compress_extension=ext; touch file.ext
0802 * mount w/ -o compress_extension=*; touch any_file
0803
0804 - To disable compression on regular inode, there are two ways:
0805
0806 * chattr -c file
0807 * mount w/ -o nocompress_extension=ext; touch file.ext
0808
0809 - Priority in between FS_COMPR_FL, FS_NOCOMP_FS, extensions:
0810
0811 * compress_extension=so; nocompress_extension=zip; chattr +c dir; touch
0812 dir/foo.so; touch dir/bar.zip; touch dir/baz.txt; then foo.so and baz.txt
0813 should be compresse, bar.zip should be non-compressed. chattr +c dir/bar.zip
0814 can enable compress on bar.zip.
0815 * compress_extension=so; nocompress_extension=zip; chattr -c dir; touch
0816 dir/foo.so; touch dir/bar.zip; touch dir/baz.txt; then foo.so should be
0817 compresse, bar.zip and baz.txt should be non-compressed.
0818 chattr+c dir/bar.zip; chattr+c dir/baz.txt; can enable compress on bar.zip
0819 and baz.txt.
0820
0821 - At this point, compression feature doesn't expose compressed space to user
0822 directly in order to guarantee potential data updates later to the space.
0823 Instead, the main goal is to reduce data writes to flash disk as much as
0824 possible, resulting in extending disk life time as well as relaxing IO
0825 congestion. Alternatively, we've added ioctl(F2FS_IOC_RELEASE_COMPRESS_BLOCKS)
0826 interface to reclaim compressed space and show it to user after setting a
0827 special flag to the inode. Once the compressed space is released, the flag
0828 will block writing data to the file until either the compressed space is
0829 reserved via ioctl(F2FS_IOC_RESERVE_COMPRESS_BLOCKS) or the file size is
0830 truncated to zero.
0831
0832 Compress metadata layout::
0833
0834 [Dnode Structure]
0835 +-----------------------------------------------+
0836 | cluster 1 | cluster 2 | ......... | cluster N |
0837 +-----------------------------------------------+
0838 . . . .
0839 . . . .
0840 . Compressed Cluster . . Normal Cluster .
0841 +----------+---------+---------+---------+ +---------+---------+---------+---------+
0842 |compr flag| block 1 | block 2 | block 3 | | block 1 | block 2 | block 3 | block 4 |
0843 +----------+---------+---------+---------+ +---------+---------+---------+---------+
0844 . .
0845 . .
0846 . .
0847 +-------------+-------------+----------+----------------------------+
0848 | data length | data chksum | reserved | compressed data |
0849 +-------------+-------------+----------+----------------------------+
0850
0851 Compression mode
0852 --------------------------
0853
0854 f2fs supports "fs" and "user" compression modes with "compression_mode" mount option.
0855 With this option, f2fs provides a choice to select the way how to compress the
0856 compression enabled files (refer to "Compression implementation" section for how to
0857 enable compression on a regular inode).
0858
0859 1) compress_mode=fs
0860 This is the default option. f2fs does automatic compression in the writeback of the
0861 compression enabled files.
0862
0863 2) compress_mode=user
0864 This disables the automatic compression and gives the user discretion of choosing the
0865 target file and the timing. The user can do manual compression/decompression on the
0866 compression enabled files using F2FS_IOC_DECOMPRESS_FILE and F2FS_IOC_COMPRESS_FILE
0867 ioctls like the below.
0868
0869 To decompress a file,
0870
0871 fd = open(filename, O_WRONLY, 0);
0872 ret = ioctl(fd, F2FS_IOC_DECOMPRESS_FILE);
0873
0874 To compress a file,
0875
0876 fd = open(filename, O_WRONLY, 0);
0877 ret = ioctl(fd, F2FS_IOC_COMPRESS_FILE);
0878
0879 NVMe Zoned Namespace devices
0880 ----------------------------
0881
0882 - ZNS defines a per-zone capacity which can be equal or less than the
0883 zone-size. Zone-capacity is the number of usable blocks in the zone.
0884 F2FS checks if zone-capacity is less than zone-size, if it is, then any
0885 segment which starts after the zone-capacity is marked as not-free in
0886 the free segment bitmap at initial mount time. These segments are marked
0887 as permanently used so they are not allocated for writes and
0888 consequently are not needed to be garbage collected. In case the
0889 zone-capacity is not aligned to default segment size(2MB), then a segment
0890 can start before the zone-capacity and span across zone-capacity boundary.
0891 Such spanning segments are also considered as usable segments. All blocks
0892 past the zone-capacity are considered unusable in these segments.
