1.. SPDX-License-Identifier: GPL-2.0 2 3========================================== 4WHAT IS Flash-Friendly File System (F2FS)? 5========================================== 6 7NAND flash memory-based storage devices, such as SSD, eMMC, and SD cards, have 8been equipped on a variety systems ranging from mobile to server systems. Since 9they are known to have different characteristics from the conventional rotating 10disks, a file system, an upper layer to the storage device, should adapt to the 11changes from the sketch in the design level. 12 13F2FS is a file system exploiting NAND flash memory-based storage devices, which 14is based on Log-structured File System (LFS). The design has been focused on 15addressing the fundamental issues in LFS, which are snowball effect of wandering 16tree and high cleaning overhead. 17 18Since a NAND flash memory-based storage device shows different characteristic 19according to its internal geometry or flash memory management scheme, namely FTL, 20F2FS and its tools support various parameters not only for configuring on-disk 21layout, but also for selecting allocation and cleaning algorithms. 22 23The following git tree provides the file system formatting tool (mkfs.f2fs), 24a consistency checking tool (fsck.f2fs), and a debugging tool (dump.f2fs). 25 26- git://git.kernel.org/pub/scm/linux/kernel/git/jaegeuk/f2fs-tools.git 27 28For reporting bugs and sending patches, please use the following mailing list: 29 30- linux-f2fs-devel@lists.sourceforge.net 31 32Background and Design issues 33============================ 34 35Log-structured File System (LFS) 36-------------------------------- 37"A log-structured file system writes all modifications to disk sequentially in 38a log-like structure, thereby speeding up both file writing and crash recovery. 39The log is the only structure on disk; it contains indexing information so that 40files can be read back from the log efficiently. In order to maintain large free 41areas on disk for fast writing, we divide the log into segments and use a 42segment cleaner to compress the live information from heavily fragmented 43segments." from Rosenblum, M. and Ousterhout, J. K., 1992, "The design and 44implementation of a log-structured file system", ACM Trans. Computer Systems 4510, 1, 26–52. 46 47Wandering Tree Problem 48---------------------- 49In LFS, when a file data is updated and written to the end of log, its direct 50pointer block is updated due to the changed location. Then the indirect pointer 51block is also updated due to the direct pointer block update. In this manner, 52the upper index structures such as inode, inode map, and checkpoint block are 53also updated recursively. This problem is called as wandering tree problem [1], 54and in order to enhance the performance, it should eliminate or relax the update 55propagation as much as possible. 56 57[1] Bityutskiy, A. 2005. JFFS3 design issues. http://www.linux-mtd.infradead.org/ 58 59Cleaning Overhead 60----------------- 61Since LFS is based on out-of-place writes, it produces so many obsolete blocks 62scattered across the whole storage. In order to serve new empty log space, it 63needs to reclaim these obsolete blocks seamlessly to users. This job is called 64as a cleaning process. 65 66The process consists of three operations as follows. 67 681. A victim segment is selected through referencing segment usage table. 692. It loads parent index structures of all the data in the victim identified by 70 segment summary blocks. 713. It checks the cross-reference between the data and its parent index structure. 724. It moves valid data selectively. 73 74This cleaning job may cause unexpected long delays, so the most important goal 75is to hide the latencies to users. And also definitely, it should reduce the 76amount of valid data to be moved, and move them quickly as well. 77 78Key Features 79============ 80 81Flash Awareness 82--------------- 83- Enlarge the random write area for better performance, but provide the high 84 spatial locality 85- Align FS data structures to the operational units in FTL as best efforts 86 87Wandering 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. 92 93Cleaning Overhead 94----------------- 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 99 100Mount Options 101============= 102 103 104======================== ============================================================ 105background_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. 113gc_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. 118nogc_merge Disable GC merge feature. 119disable_roll_forward Disable the roll-forward recovery routine 120norecovery Disable the roll-forward recovery routine, mounted read- 121 only (i.e., -o ro,disable_roll_forward) 122discard/nodiscard Enable/disable real-time discard in f2fs, if discard is 123 enabled, f2fs will issue discard/TRIM commands when a 124 segment is cleaned. 125no_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. 128nouser_xattr Disable Extended User Attributes. Note: xattr is enabled 129 by default if CONFIG_F2FS_FS_XATTR is selected. 130noacl Disable POSIX Access Control List. Note: acl is enabled 131 by default if CONFIG_F2FS_FS_POSIX_ACL is selected. 132active_logs=%u Support configuring the number of active logs. In the 133 current design, f2fs supports only 2, 4, and 6 logs. 134 Default number is 6. 135disable_ext_identify Disable the extension list configured by mkfs, so f2fs 136 is not aware of cold files such as media files. 137inline_xattr Enable the inline xattrs feature. 138noinline_xattr Disable the inline xattrs feature. 139inline_xattr_size=%u Support configuring inline xattr size, it depends on 140 flexible inline xattr feature. 141inline_data Enable the inline data feature: Newly created small (<~3.4k) 142 files can be written into inode block. 143inline_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. 147noinline_dentry Disable the inline dentry feature. 148flush_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. 152nobarrier 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 156 data writes. 157fastboot This option is used when a system wants to reduce mount 158 time as much as possible, even though normal performance 159 can be sacrificed. 160extent_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. 164noextent_cache Disable an extent cache based on rb-tree explicitly, see 165 the above extent_cache mount option. 166noinline_data Disable the inline data feature, inline data feature is 167 enabled by default. 168data_flush Enable data flushing before checkpoint in order to 169 persist data of regular and symlink. 170reserve_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. 173resuid=%d The user ID which may use the reserved blocks. 174resgid=%d The group ID which may use the reserved blocks. 175fault_injection=%d Enable fault injection in all supported types with 176 specified injection rate. 177fault_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. 180 181 =================== =========== 182 Type_Name Type_Value 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 =================== =========== 201mode=%s Control block allocation mode which supports "adaptive" 202 and "lfs". In "lfs" mode, there should be no random 203 writes towards main area. 204io_bits=%u Set the bit size of write IO requests. It should be set 205 with "mode=lfs". 206usrquota Enable plain user disk quota accounting. 207grpquota Enable plain group disk quota accounting. 208prjquota Enable plain project quota accounting. 209usrjquota=<file> Appoint specified file and type during mount, so that quota 210grpjquota=<file> information can be properly updated during recovery flow, 211prjjquota=<file> <quota file>: must be in root directory; 212jqfmt=<quota type> <quota type>: [vfsold,vfsv0,vfsv1]. 213offusrjquota Turn off user journalled quota. 214offgrpjquota Turn off group journalled quota. 215offprjjquota Turn off project journalled quota. 216quota Enable plain user disk quota accounting. 217noquota Disable all plain disk quota option. 218whint_mode=%s Control which write hints are passed down to block 219 layer. This supports "off", "user-based", and 220 "fs-based". In "off" mode (default), f2fs does not pass 221 down hints. In "user-based" mode, f2fs tries to pass 222 down hints given by users. And in "fs-based" mode, f2fs 223 passes down hints with its policy. 224alloc_mode=%s Adjust block allocation policy, which supports "reuse" 225 and "default". 226fsync_mode=%s Control the policy of fsync. Currently supports "posix", 227 "strict", and "nobarrier". In "posix" mode, which is 228 default, fsync will follow POSIX semantics and does a 229 light operation to improve the filesystem performance. 230 In "strict" mode, fsync will be heavy and behaves in line 231 with xfs, ext4 and btrfs, where xfstest generic/342 will 232 pass, but the performance will regress. "nobarrier" is 233 based on "posix", but doesn't issue flush command for 234 non-atomic files likewise "nobarrier" mount option. 235test_dummy_encryption 236test_dummy_encryption=%s 237 Enable dummy encryption, which provides a fake fscrypt 238 context. The fake fscrypt context is used by xfstests. 239 The argument may be either "v1" or "v2", in order to 240 select the corresponding fscrypt policy version. 241checkpoint=%s[:%u[%]] Set to "disable" to turn off checkpointing. Set to "enable" 242 to reenable checkpointing. Is enabled by default. While 243 disabled, any unmounting or unexpected shutdowns will cause 244 the filesystem contents to appear as they did when the 245 filesystem was mounted with that option. 246 While mounting with checkpoint=disabled, the filesystem must 247 run garbage collection to ensure that all available space can 248 be used. If this takes too much time, the mount may return 249 EAGAIN. You may optionally add a value to indicate how much 250 of the disk you would be willing to temporarily give up to 251 avoid additional garbage collection. This can be given as a 252 number of blocks, or as a percent. For instance, mounting 253 with checkpoint=disable:100% would always succeed, but it may 254 hide up to all remaining free space. The actual space that 255 would be unusable can be viewed at /sys/fs/f2fs/<disk>/unusable 256 This space is reclaimed once checkpoint=enable. 257checkpoint_merge When checkpoint is enabled, this can be used to create a kernel 258 daemon and make it to merge concurrent checkpoint requests as 259 much as possible to eliminate redundant checkpoint issues. Plus, 260 we can eliminate the sluggish issue caused by slow checkpoint 261 operation when the checkpoint is done in a process context in 262 a cgroup having low i/o budget and cpu shares. To make this 263 do better, we set the default i/o priority of the kernel daemon 264 to "3", to give one higher priority than other kernel threads. 265 This is the same way to give a I/O priority to the jbd2 266 journaling thread of ext4 filesystem. 267nocheckpoint_merge Disable checkpoint merge feature. 268compress_algorithm=%s Control compress algorithm, currently f2fs supports "lzo", 269 "lz4", "zstd" and "lzo-rle" algorithm. 270compress_algorithm=%s:%d Control compress algorithm and its compress level, now, only 271 "lz4" and "zstd" support compress level config. 272 algorithm level range 273 lz4 3 - 16 274 zstd 1 - 22 275compress_log_size=%u Support configuring compress cluster size, the size will 276 be 4KB * (1 << %u), 16KB is minimum size, also it's 277 default size. 278compress_extension=%s Support adding specified extension, so that f2fs can enable 279 compression on those corresponding files, e.g. if all files 280 with '.ext' has high compression rate, we can set the '.ext' 281 on compression extension list and enable compression on 282 these file by default rather than to enable it via ioctl. 283 For other files, we can still enable compression via ioctl. 284 Note that, there is one reserved special extension '*', it 285 can be set to enable compression for all files. 286nocompress_extension=%s Support adding specified extension, so that f2fs can disable 287 compression on those corresponding files, just contrary to compression extension. 288 If you know exactly which files cannot be compressed, you can use this. 289 The same extension name can't appear in both compress and nocompress 290 extension at the same time. 291 If the compress extension specifies all files, the types specified by the 292 nocompress extension will be treated as special cases and will not be compressed. 293 Don't allow use '*' to specifie all file in nocompress extension. 294 After add nocompress_extension, the priority should be: 295 dir_flag < comp_extention,nocompress_extension < comp_file_flag,no_comp_file_flag. 296 See more in compression sections. 297 298compress_chksum Support verifying chksum of raw data in compressed cluster. 299compress_mode=%s Control file compression mode. This supports "fs" and "user" 300 modes. In "fs" mode (default), f2fs does automatic compression 301 on the compression enabled files. In "user" mode, f2fs disables 302 the automaic compression and gives the user discretion of 303 choosing the target file and the timing. The user can do manual 304 compression/decompression on the compression enabled files using 305 ioctls. 306compress_cache Support to use address space of a filesystem managed inode to 307 cache compressed block, in order to improve cache hit ratio of 308 random read. 309inlinecrypt When possible, encrypt/decrypt the contents of encrypted 310 files using the blk-crypto framework rather than 311 filesystem-layer encryption. This allows the use of 312 inline encryption hardware. The on-disk format is 313 unaffected. For more details, see 314 Documentation/block/inline-encryption.rst. 315atgc Enable age-threshold garbage collection, it provides high 316 effectiveness and efficiency on background GC. 317discard_unit=%s Control discard unit, the argument can be "block", "segment" 318 and "section", issued discard command's offset/size will be 319 aligned to the unit, by default, "discard_unit=block" is set, 320 so that small discard functionality is enabled. 321 For blkzoned device, "discard_unit=section" will be set by 322 default, it is helpful for large sized SMR or ZNS devices to 323 reduce memory cost by getting rid of fs metadata supports small 324 discard. 325======================== ============================================================ 326 327Debugfs Entries 328=============== 329 330/sys/kernel/debug/f2fs/ contains information about all the partitions mounted as 331f2fs. Each file shows the whole f2fs information. 332 333/sys/kernel/debug/f2fs/status includes: 334 335 - major file system information managed by f2fs currently 336 - average SIT information about whole segments 337 - current memory footprint consumed by f2fs. 338 339Sysfs Entries 340============= 341 342Information about mounted f2fs file systems can be found in 343/sys/fs/f2fs. Each mounted filesystem will have a directory in 344/sys/fs/f2fs based on its device name (i.e., /sys/fs/f2fs/sda). 345The files in each per-device directory are shown in table below. 346 347Files in /sys/fs/f2fs/<devname> 348(see also Documentation/ABI/testing/sysfs-fs-f2fs) 349 350Usage 351===== 352 3531. Download userland tools and compile them. 354 3552. Skip, if f2fs was compiled statically inside kernel. 356 Otherwise, insert the f2fs.ko module:: 357 358 # insmod f2fs.ko 359 3603. Create a directory to use when mounting:: 361 362 # mkdir /mnt/f2fs 363 3644. Format the block device, and then mount as f2fs:: 365 366 # mkfs.f2fs -l label /dev/block_device 367 # mount -t f2fs /dev/block_device /mnt/f2fs 368 369mkfs.f2fs 370--------- 371The mkfs.f2fs is for the use of formatting a partition as the f2fs filesystem, 372which builds a basic on-disk layout. 373 374The quick options consist of: 375 376=============== =========================================================== 377``-l [label]`` Give a volume label, up to 512 unicode name. 378``-a [0 or 1]`` Split start location of each area for heap-based allocation. 379 380 1 is set by default, which performs this. 381``-o [int]`` Set overprovision ratio in percent over volume size. 382 383 5 is set by default. 384``-s [int]`` Set the number of segments per section. 385 386 1 is set by default. 387``-z [int]`` Set the number of sections per zone. 388 389 1 is set by default. 390``-e [str]`` Set basic extension list. e.g. "mp3,gif,mov" 391``-t [0 or 1]`` Disable discard command or not. 392 393 1 is set by default, which conducts discard. 394=============== =========================================================== 395 396Note: please refer to the manpage of mkfs.f2fs(8) to get full option list. 397 398fsck.f2fs 399--------- 400The fsck.f2fs is a tool to check the consistency of an f2fs-formatted 401partition, which examines whether the filesystem metadata and user-made data 402are cross-referenced correctly or not. 403Note that, initial version of the tool does not fix any inconsistency. 404 405The quick options consist of:: 406 407 -d debug level [default:0] 408 409Note: please refer to the manpage of fsck.f2fs(8) to get full option list. 410 411dump.f2fs 412--------- 413The dump.f2fs shows the information of specific inode and dumps SSA and SIT to 414file. Each file is dump_ssa and dump_sit. 415 416The dump.f2fs is used to debug on-disk data structures of the f2fs filesystem. 417It shows on-disk inode information recognized by a given inode number, and is 418able to dump all the SSA and SIT entries into predefined files, ./dump_ssa and 419./dump_sit respectively. 420 421The options consist of:: 422 423 -d debug level [default:0] 424 -i inode no (hex) 425 -s [SIT dump segno from #1~#2 (decimal), for all 0~-1] 426 -a [SSA dump segno from #1~#2 (decimal), for all 0~-1] 427 428Examples:: 429 430 # dump.f2fs -i [ino] /dev/sdx 431 # dump.f2fs -s 0~-1 /dev/sdx (SIT dump) 432 # dump.f2fs -a 0~-1 /dev/sdx (SSA dump) 433 434Note: please refer to the manpage of dump.f2fs(8) to get full option list. 435 436sload.f2fs 437---------- 438The sload.f2fs gives a way to insert files and directories in the exisiting disk 439image. This tool is useful when building f2fs images given compiled files. 440 441Note: please refer to the manpage of sload.f2fs(8) to get full option list. 442 443resize.f2fs 444----------- 445The resize.f2fs lets a user resize the f2fs-formatted disk image, while preserving 446all the files and directories stored in the image. 447 448Note: please refer to the manpage of resize.f2fs(8) to get full option list. 449 450defrag.f2fs 451----------- 452The defrag.f2fs can be used to defragment scattered written data as well as 453filesystem metadata across the disk. This can improve the write speed by giving 454more free consecutive space. 455 456Note: please refer to the manpage of defrag.f2fs(8) to get full option list. 457 458f2fs_io 459------- 460The f2fs_io is a simple tool to issue various filesystem APIs as well as 461f2fs-specific ones, which is very useful for QA tests. 462 463Note: please refer to the manpage of f2fs_io(8) to get full option list. 464 465Design 466====== 467 468On-disk Layout 469-------------- 470 471F2FS divides the whole volume into a number of segments, each of which is fixed 472to 2MB in size. A section is composed of consecutive segments, and a zone 473consists of a set of sections. By default, section and zone sizes are set to one 474segment size identically, but users can easily modify the sizes by mkfs. 475 476F2FS splits the entire volume into six areas, and all the areas except superblock 477consist of multiple segments as described below:: 478 479 align with the zone size <-| 480 |-> align with the segment size 481 _________________________________________________________________________ 482 | | | Segment | Node | Segment | | 483 | Superblock | Checkpoint | Info. | Address | Summary | Main | 484 | (SB) | (CP) | Table (SIT) | Table (NAT) | Area (SSA) | | 485 |____________|_____2______|______N______|______N______|______N_____|__N___| 486 . . 487 . . 488 . . 489 ._________________________________________. 490 |_Segment_|_..._|_Segment_|_..._|_Segment_| 491 . . 492 ._________._________ 493 |_section_|__...__|_ 494 . . 495 .________. 496 |__zone__| 497 498- Superblock (SB) 499 It is located at the beginning of the partition, and there exist two copies 500 to avoid file system crash. It contains basic partition information and some 501 default parameters of f2fs. 502 503- Checkpoint (CP) 504 It contains file system information, bitmaps for valid NAT/SIT sets, orphan 505 inode lists, and summary entries of current active segments. 506 507- Segment Information Table (SIT) 508 It contains segment information such as valid block count and bitmap for the 509 validity of all the blocks. 510 511- Node Address Table (NAT) 512 It is composed of a block address table for all the node blocks stored in 513 Main area. 514 515- Segment Summary Area (SSA) 516 It contains summary entries which contains the owner information of all the 517 data and node blocks stored in Main area. 518 519- Main Area 520 It contains file and directory data including their indices. 521 522In order to avoid misalignment between file system and flash-based storage, F2FS 523aligns the start block address of CP with the segment size. Also, it aligns the 524start block address of Main area with the zone size by reserving some segments 525in SSA area. 526 527Reference the following survey for additional technical details. 528https://wiki.linaro.org/WorkingGroups/Kernel/Projects/FlashCardSurvey 529 530File System Metadata Structure 531------------------------------ 532 533F2FS adopts the checkpointing scheme to maintain file system consistency. At 534mount time, F2FS first tries to find the last valid checkpoint data by scanning 535CP area. In order to reduce the scanning time, F2FS uses only two copies of CP. 536One of them always indicates the last valid data, which is called as shadow copy 537mechanism. In addition to CP, NAT and SIT also adopt the shadow copy mechanism. 538 539For file system consistency, each CP points to which NAT and SIT copies are 540valid, as shown as below:: 541 542 +--------+----------+---------+ 543 | CP | SIT | NAT | 544 +--------+----------+---------+ 545 . . . . 546 . . . . 547 . . . . 548 +-------+-------+--------+--------+--------+--------+ 549 | CP #0 | CP #1 | SIT #0 | SIT #1 | NAT #0 | NAT #1 | 550 +-------+-------+--------+--------+--------+--------+ 551 | ^ ^ 552 | | | 553 `----------------------------------------' 554 555Index Structure 556--------------- 557 558The key data structure to manage the data locations is a "node". Similar to 559traditional file structures, F2FS has three types of node: inode, direct node, 560indirect node. F2FS assigns 4KB to an inode block which contains 923 data block 561indices, two direct node pointers, two indirect node pointers, and one double 562indirect node pointer as described below. One direct node block contains 1018 563data blocks, and one indirect node block contains also 1018 node blocks. Thus, 564one inode block (i.e., a file) covers:: 565 566 4KB * (923 + 2 * 1018 + 2 * 1018 * 1018 + 1018 * 1018 * 1018) := 3.94TB. 567 568 Inode block (4KB) 569 |- data (923) 570 |- direct node (2) 571 | `- data (1018) 572 |- indirect node (2) 573 | `- direct node (1018) 574 | `- data (1018) 575 `- double indirect node (1) 576 `- indirect node (1018) 577 `- direct node (1018) 578 `- data (1018) 579 580Note that all the node blocks are mapped by NAT which means the location of 581each node is translated by the NAT table. In the consideration of the wandering 582tree problem, F2FS is able to cut off the propagation of node updates caused by 583leaf data writes. 584 585Directory Structure 586------------------- 587 588A directory entry occupies 11 bytes, which consists of the following attributes. 589 590- hash hash value of the file name 591- ino inode number 592- len the length of file name 593- type file type such as directory, symlink, etc 594 595A dentry block consists of 214 dentry slots and file names. Therein a bitmap is 596used to represent whether each dentry is valid or not. A dentry block occupies 5974KB with the following composition. 598 599:: 600 601 Dentry Block(4 K) = bitmap (27 bytes) + reserved (3 bytes) + 602 dentries(11 * 214 bytes) + file name (8 * 214 bytes) 603 604 [Bucket] 605 +--------------------------------+ 606 |dentry block 1 | dentry block 2 | 607 +--------------------------------+ 608 . . 609 . . 610 . [Dentry Block Structure: 4KB] . 611 +--------+----------+----------+------------+ 612 | bitmap | reserved | dentries | file names | 613 +--------+----------+----------+------------+ 614 [Dentry Block: 4KB] . . 615 . . 616 . . 617 +------+------+-----+------+ 618 | hash | ino | len | type | 619 +------+------+-----+------+ 620 [Dentry Structure: 11 bytes] 621 622F2FS implements multi-level hash tables for directory structure. Each level has 623a hash table with dedicated number of hash buckets as shown below. Note that 624"A(2B)" means a bucket includes 2 data blocks. 625 626:: 627 628 ---------------------- 629 A : bucket 630 B : block 631 N : MAX_DIR_HASH_DEPTH 632 ---------------------- 633 634 level #0 | A(2B) 635 | 636 level #1 | A(2B) - A(2B) 637 | 638 level #2 | A(2B) - A(2B) - A(2B) - A(2B) 639 . | . . . . 640 level #N/2 | A(2B) - A(2B) - A(2B) - A(2B) - A(2B) - ... - A(2B) 641 . | . . . . 642 level #N | A(4B) - A(4B) - A(4B) - A(4B) - A(4B) - ... - A(4B) 643 644The number of blocks and buckets are determined by:: 645 646 ,- 2, if n < MAX_DIR_HASH_DEPTH / 2, 647 # of blocks in level #n = | 648 `- 4, Otherwise 649 650 ,- 2^(n + dir_level), 651 | if n + dir_level < MAX_DIR_HASH_DEPTH / 2, 652 # of buckets in level #n = | 653 `- 2^((MAX_DIR_HASH_DEPTH / 2) - 1), 654 Otherwise 655 656When F2FS finds a file name in a directory, at first a hash value of the file 657name is calculated. Then, F2FS scans the hash table in level #0 to find the 658dentry consisting of the file name and its inode number. If not found, F2FS 659scans the next hash table in level #1. In this way, F2FS scans hash tables in 660each levels incrementally from 1 to N. In each level F2FS needs to scan only 661one bucket determined by the following equation, which shows O(log(# of files)) 662complexity:: 663 664 bucket number to scan in level #n = (hash value) % (# of buckets in level #n) 665 666In the case of file creation, F2FS finds empty consecutive slots that cover the 667file name. F2FS searches the empty slots in the hash tables of whole levels from 6681 to N in the same way as the lookup operation. 669 670The following figure shows an example of two cases holding children:: 671 672 --------------> Dir <-------------- 673 | | 674 child child 675 676 child - child [hole] - child 677 678 child - child - child [hole] - [hole] - child 679 680 Case 1: Case 2: 681 Number of children = 6, Number of children = 3, 682 File size = 7 File size = 7 683 684Default Block Allocation 685------------------------ 686 687At runtime, F2FS manages six active logs inside "Main" area: Hot/Warm/Cold node 688and Hot/Warm/Cold data. 689 690- Hot node contains direct node blocks of directories. 691- Warm node contains direct node blocks except hot node blocks. 692- Cold node contains indirect node blocks 693- Hot data contains dentry blocks 694- Warm data contains data blocks except hot and cold data blocks 695- Cold data contains multimedia data or migrated data blocks 696 697LFS has two schemes for free space management: threaded log and copy-and-compac- 698tion. The copy-and-compaction scheme which is known as cleaning, is well-suited 699for devices showing very good sequential write performance, since free segments 700are served all the time for writing new data. However, it suffers from cleaning 701overhead under high utilization. Contrarily, the threaded log scheme suffers 702from random writes, but no cleaning process is needed. F2FS adopts a hybrid 703scheme where the copy-and-compaction scheme is adopted by default, but the 704policy is dynamically changed to the threaded log scheme according to the file 705system status. 706 707In order to align F2FS with underlying flash-based storage, F2FS allocates a 708segment in a unit of section. F2FS expects that the section size would be the 709same as the unit size of garbage collection in FTL. Furthermore, with respect 710to the mapping granularity in FTL, F2FS allocates each section of the active 711logs from different zones as much as possible, since FTL can write the data in 712the active logs into one allocation unit according to its mapping granularity. 713 714Cleaning process 715---------------- 716 717F2FS does cleaning both on demand and in the background. On-demand cleaning is 718triggered when there are not enough free segments to serve VFS calls. Background 719cleaner is operated by a kernel thread, and triggers the cleaning job when the 720system is idle. 721 722F2FS supports two victim selection policies: greedy and cost-benefit algorithms. 723In the greedy algorithm, F2FS selects a victim segment having the smallest number 724of valid blocks. In the cost-benefit algorithm, F2FS selects a victim segment 725according to the segment age and the number of valid blocks in order to address 726log block thrashing problem in the greedy algorithm. F2FS adopts the greedy 727algorithm for on-demand cleaner, while background cleaner adopts cost-benefit 728algorithm. 729 730In order to identify whether the data in the victim segment are valid or not, 731F2FS manages a bitmap. Each bit represents the validity of a block, and the 732bitmap is composed of a bit stream covering whole blocks in main area. 733 734Write-hint Policy 735----------------- 736 7371) whint_mode=off. F2FS only passes down WRITE_LIFE_NOT_SET. 738 7392) whint_mode=user-based. F2FS tries to pass down hints given by 740users. 741 742===================== ======================== =================== 743User F2FS Block 744===================== ======================== =================== 745N/A META WRITE_LIFE_NOT_SET 746N/A HOT_NODE " 747N/A WARM_NODE " 748N/A COLD_NODE " 749ioctl(COLD) COLD_DATA WRITE_LIFE_EXTREME 750extension list " " 751 752-- buffered io 753WRITE_LIFE_EXTREME COLD_DATA WRITE_LIFE_EXTREME 754WRITE_LIFE_SHORT HOT_DATA WRITE_LIFE_SHORT 755WRITE_LIFE_NOT_SET WARM_DATA WRITE_LIFE_NOT_SET 756WRITE_LIFE_NONE " " 757WRITE_LIFE_MEDIUM " " 758WRITE_LIFE_LONG " " 759 760-- direct io 761WRITE_LIFE_EXTREME COLD_DATA WRITE_LIFE_EXTREME 762WRITE_LIFE_SHORT HOT_DATA WRITE_LIFE_SHORT 763WRITE_LIFE_NOT_SET WARM_DATA WRITE_LIFE_NOT_SET 764WRITE_LIFE_NONE " WRITE_LIFE_NONE 765WRITE_LIFE_MEDIUM " WRITE_LIFE_MEDIUM 766WRITE_LIFE_LONG " WRITE_LIFE_LONG 767===================== ======================== =================== 768 7693) whint_mode=fs-based. F2FS passes down hints with its policy. 770 771===================== ======================== =================== 772User F2FS Block 773===================== ======================== =================== 774N/A META WRITE_LIFE_MEDIUM; 775N/A HOT_NODE WRITE_LIFE_NOT_SET 776N/A WARM_NODE " 777N/A COLD_NODE WRITE_LIFE_NONE 778ioctl(COLD) COLD_DATA WRITE_LIFE_EXTREME 779extension list " " 780 781-- buffered io 782WRITE_LIFE_EXTREME COLD_DATA WRITE_LIFE_EXTREME 783WRITE_LIFE_SHORT HOT_DATA WRITE_LIFE_SHORT 784WRITE_LIFE_NOT_SET WARM_DATA WRITE_LIFE_LONG 785WRITE_LIFE_NONE " " 786WRITE_LIFE_MEDIUM " " 787WRITE_LIFE_LONG " " 788 789-- direct io 790WRITE_LIFE_EXTREME COLD_DATA WRITE_LIFE_EXTREME 791WRITE_LIFE_SHORT HOT_DATA WRITE_LIFE_SHORT 792WRITE_LIFE_NOT_SET WARM_DATA WRITE_LIFE_NOT_SET 793WRITE_LIFE_NONE " WRITE_LIFE_NONE 794WRITE_LIFE_MEDIUM " WRITE_LIFE_MEDIUM 795WRITE_LIFE_LONG " WRITE_LIFE_LONG 796===================== ======================== =================== 797 798Fallocate(2) Policy 799------------------- 800 801The default policy follows the below POSIX rule. 802 803Allocating disk space 804 The default operation (i.e., mode is zero) of fallocate() allocates 805 the disk space within the range specified by offset and len. The 806 file size (as reported by stat(2)) will be changed if offset+len is 807 greater than the file size. Any subregion within the range specified 808 by offset and len that did not contain data before the call will be 809 initialized to zero. This default behavior closely resembles the 810 behavior of the posix_fallocate(3) library function, and is intended 811 as a method of optimally implementing that function. 812 813However, once F2FS receives ioctl(fd, F2FS_IOC_SET_PIN_FILE) in prior to 814fallocate(fd, DEFAULT_MODE), it allocates on-disk block addressess having 815zero or random data, which is useful to the below scenario where: 816 817 1. create(fd) 818 2. ioctl(fd, F2FS_IOC_SET_PIN_FILE) 819 3. fallocate(fd, 0, 0, size) 820 4. address = fibmap(fd, offset) 821 5. open(blkdev) 822 6. write(blkdev, address) 823 824Compression implementation 825-------------------------- 826 827- New term named cluster is defined as basic unit of compression, file can 828 be divided into multiple clusters logically. One cluster includes 4 << n 829 (n >= 0) logical pages, compression size is also cluster size, each of 830 cluster can be compressed or not. 831 832- In cluster metadata layout, one special block address is used to indicate 833 a cluster is a compressed one or normal one; for compressed cluster, following 834 metadata maps cluster to [1, 4 << n - 1] physical blocks, in where f2fs 835 stores data including compress header and compressed data. 836 837- In order to eliminate write amplification during overwrite, F2FS only 838 support compression on write-once file, data can be compressed only when 839 all logical blocks in cluster contain valid data and compress ratio of 840 cluster data is lower than specified threshold. 841 842- To enable compression on regular inode, there are four ways: 843 844 * chattr +c file 845 * chattr +c dir; touch dir/file 846 * mount w/ -o compress_extension=ext; touch file.ext 847 * mount w/ -o compress_extension=*; touch any_file 848 849- To disable compression on regular inode, there are two ways: 850 851 * chattr -c file 852 * mount w/ -o nocompress_extension=ext; touch file.ext 853 854- Priority in between FS_COMPR_FL, FS_NOCOMP_FS, extensions: 855 856 * compress_extension=so; nocompress_extension=zip; chattr +c dir; touch 857 dir/foo.so; touch dir/bar.zip; touch dir/baz.txt; then foo.so and baz.txt 858 should be compresse, bar.zip should be non-compressed. chattr +c dir/bar.zip 859 can enable compress on bar.zip. 860 * compress_extension=so; nocompress_extension=zip; chattr -c dir; touch 861 dir/foo.so; touch dir/bar.zip; touch dir/baz.txt; then foo.so should be 862 compresse, bar.zip and baz.txt should be non-compressed. 863 chattr+c dir/bar.zip; chattr+c dir/baz.txt; can enable compress on bar.zip 864 and baz.txt. 865 866- At this point, compression feature doesn't expose compressed space to user 867 directly in order to guarantee potential data updates later to the space. 868 Instead, the main goal is to reduce data writes to flash disk as much as 869 possible, resulting in extending disk life time as well as relaxing IO 870 congestion. Alternatively, we've added ioctl(F2FS_IOC_RELEASE_COMPRESS_BLOCKS) 871 interface to reclaim compressed space and show it to user after putting the 872 immutable bit. Immutable bit, after release, it doesn't allow writing/mmaping 873 on the file, until reserving compressed space via 874 ioctl(F2FS_IOC_RESERVE_COMPRESS_BLOCKS) or truncating filesize to zero. 875 876Compress metadata layout:: 877 878 [Dnode Structure] 879 +-----------------------------------------------+ 880 | cluster 1 | cluster 2 | ......... | cluster N | 881 +-----------------------------------------------+ 882 . . . . 883 . . . . 884 . Compressed Cluster . . Normal Cluster . 885 +----------+---------+---------+---------+ +---------+---------+---------+---------+ 886 |compr flag| block 1 | block 2 | block 3 | | block 1 | block 2 | block 3 | block 4 | 887 +----------+---------+---------+---------+ +---------+---------+---------+---------+ 888 . . 889 . . 890 . . 891 +-------------+-------------+----------+----------------------------+ 892 | data length | data chksum | reserved | compressed data | 893 +-------------+-------------+----------+----------------------------+ 894 895Compression mode 896-------------------------- 897 898f2fs supports "fs" and "user" compression modes with "compression_mode" mount option. 899With this option, f2fs provides a choice to select the way how to compress the 900compression enabled files (refer to "Compression implementation" section for how to 901enable compression on a regular inode). 902 9031) compress_mode=fs 904This is the default option. f2fs does automatic compression in the writeback of the 905compression enabled files. 906 9072) compress_mode=user 908This disables the automatic compression and gives the user discretion of choosing the 909target file and the timing. The user can do manual compression/decompression on the 910compression enabled files using F2FS_IOC_DECOMPRESS_FILE and F2FS_IOC_COMPRESS_FILE 911ioctls like the below. 912 913To decompress a file, 914 915fd = open(filename, O_WRONLY, 0); 916ret = ioctl(fd, F2FS_IOC_DECOMPRESS_FILE); 917 918To compress a file, 919 920fd = open(filename, O_WRONLY, 0); 921ret = ioctl(fd, F2FS_IOC_COMPRESS_FILE); 922 923NVMe Zoned Namespace devices 924---------------------------- 925 926- ZNS defines a per-zone capacity which can be equal or less than the 927 zone-size. Zone-capacity is the number of usable blocks in the zone. 928 F2FS checks if zone-capacity is less than zone-size, if it is, then any 929 segment which starts after the zone-capacity is marked as not-free in 930 the free segment bitmap at initial mount time. These segments are marked 931 as permanently used so they are not allocated for writes and 932 consequently are not needed to be garbage collected. In case the 933 zone-capacity is not aligned to default segment size(2MB), then a segment 934 can start before the zone-capacity and span across zone-capacity boundary. 935 Such spanning segments are also considered as usable segments. All blocks 936 past the zone-capacity are considered unusable in these segments. 937