blob: b8d284975f0f074c5e2c5d6126e1cf33bb6798d4 [file] [log] [blame]
Jaegeuk Kim98e4da82012-11-02 17:05:42 +09001================================================================================
2WHAT IS Flash-Friendly File System (F2FS)?
3================================================================================
4
5NAND flash memory-based storage devices, such as SSD, eMMC, and SD cards, have
6been equipped on a variety systems ranging from mobile to server systems. Since
7they are known to have different characteristics from the conventional rotating
8disks, a file system, an upper layer to the storage device, should adapt to the
9changes from the sketch in the design level.
10
11F2FS is a file system exploiting NAND flash memory-based storage devices, which
12is based on Log-structured File System (LFS). The design has been focused on
13addressing the fundamental issues in LFS, which are snowball effect of wandering
14tree and high cleaning overhead.
15
16Since a NAND flash memory-based storage device shows different characteristic
17according to its internal geometry or flash memory management scheme, namely FTL,
18F2FS and its tools support various parameters not only for configuring on-disk
19layout, but also for selecting allocation and cleaning algorithms.
20
Changman Leed51a7fb2013-07-04 17:12:47 +090021The following git tree provides the file system formatting tool (mkfs.f2fs),
22a consistency checking tool (fsck.f2fs), and a debugging tool (dump.f2fs).
Jaegeuk Kim5bb446a2012-11-27 14:36:14 +090023>> git://git.kernel.org/pub/scm/linux/kernel/git/jaegeuk/f2fs-tools.git
24
25For reporting bugs and sending patches, please use the following mailing list:
26>> linux-f2fs-devel@lists.sourceforge.net
Jaegeuk Kim98e4da82012-11-02 17:05:42 +090027
28================================================================================
29BACKGROUND AND DESIGN ISSUES
30================================================================================
31
32Log-structured File System (LFS)
33--------------------------------
34"A log-structured file system writes all modifications to disk sequentially in
35a log-like structure, thereby speeding up both file writing and crash recovery.
36The log is the only structure on disk; it contains indexing information so that
37files can be read back from the log efficiently. In order to maintain large free
38areas on disk for fast writing, we divide the log into segments and use a
39segment cleaner to compress the live information from heavily fragmented
40segments." from Rosenblum, M. and Ousterhout, J. K., 1992, "The design and
41implementation of a log-structured file system", ACM Trans. Computer Systems
4210, 1, 26–52.
43
44Wandering Tree Problem
45----------------------
46In LFS, when a file data is updated and written to the end of log, its direct
47pointer block is updated due to the changed location. Then the indirect pointer
48block is also updated due to the direct pointer block update. In this manner,
49the upper index structures such as inode, inode map, and checkpoint block are
50also updated recursively. This problem is called as wandering tree problem [1],
51and in order to enhance the performance, it should eliminate or relax the update
52propagation as much as possible.
53
54[1] Bityutskiy, A. 2005. JFFS3 design issues. http://www.linux-mtd.infradead.org/
55
56Cleaning Overhead
57-----------------
58Since LFS is based on out-of-place writes, it produces so many obsolete blocks
59scattered across the whole storage. In order to serve new empty log space, it
60needs to reclaim these obsolete blocks seamlessly to users. This job is called
61as a cleaning process.
62
63The process consists of three operations as follows.
641. A victim segment is selected through referencing segment usage table.
652. It loads parent index structures of all the data in the victim identified by
66 segment summary blocks.
673. It checks the cross-reference between the data and its parent index structure.
684. It moves valid data selectively.
69
70This cleaning job may cause unexpected long delays, so the most important goal
71is to hide the latencies to users. And also definitely, it should reduce the
72amount of valid data to be moved, and move them quickly as well.
73
74================================================================================
75KEY FEATURES
76================================================================================
77
78Flash Awareness
79---------------
80- Enlarge the random write area for better performance, but provide the high
81 spatial locality
82- Align FS data structures to the operational units in FTL as best efforts
83
84Wandering Tree Problem
85----------------------
86- Use a term, “node”, that represents inodes as well as various pointer blocks
87- Introduce Node Address Table (NAT) containing the locations of all the “node”
88 blocks; this will cut off the update propagation.
89
90Cleaning Overhead
91-----------------
92- Support a background cleaning process
93- Support greedy and cost-benefit algorithms for victim selection policies
94- Support multi-head logs for static/dynamic hot and cold data separation
95- Introduce adaptive logging for efficient block allocation
96
97================================================================================
98MOUNT OPTIONS
99================================================================================
100
Namjae Jeon696c0182013-06-16 09:48:48 +0900101background_gc=%s Turn on/off cleaning operations, namely garbage
102 collection, triggered in background when I/O subsystem is
103 idle. If background_gc=on, it will turn on the garbage
104 collection and if background_gc=off, garbage collection
105 will be truned off.
106 Default value for this option is on. So garbage
107 collection is on by default.
Jaegeuk Kim98e4da82012-11-02 17:05:42 +0900108disable_roll_forward Disable the roll-forward recovery routine
109discard Issue discard/TRIM commands when a segment is cleaned.
110no_heap Disable heap-style segment allocation which finds free
111 segments for data from the beginning of main area, while
112 for node from the end of main area.
113nouser_xattr Disable Extended User Attributes. Note: xattr is enabled
114 by default if CONFIG_F2FS_FS_XATTR is selected.
115noacl Disable POSIX Access Control List. Note: acl is enabled
116 by default if CONFIG_F2FS_FS_POSIX_ACL is selected.
117active_logs=%u Support configuring the number of active logs. In the
118 current design, f2fs supports only 2, 4, and 6 logs.
119 Default number is 6.
120disable_ext_identify Disable the extension list configured by mkfs, so f2fs
121 does not aware of cold files such as media files.
Jaegeuk Kim66e960c2013-11-01 11:20:05 +0900122inline_xattr Enable the inline xattrs feature.
Huajun Lie4024e82013-11-10 23:13:21 +0800123inline_data Enable the inline data feature: New created small(<~3.4k)
124 files can be written into inode block.
Jaegeuk Kim98e4da82012-11-02 17:05:42 +0900125
126================================================================================
127DEBUGFS ENTRIES
128================================================================================
129
130/sys/kernel/debug/f2fs/ contains information about all the partitions mounted as
131f2fs. Each file shows the whole f2fs information.
132
133/sys/kernel/debug/f2fs/status includes:
134 - major file system information managed by f2fs currently
135 - average SIT information about whole segments
136 - current memory footprint consumed by f2fs.
137
138================================================================================
Namjae Jeonb59d0ba2013-08-04 23:09:40 +0900139SYSFS ENTRIES
140================================================================================
141
142Information about mounted f2f2 file systems can be found in
143/sys/fs/f2fs. Each mounted filesystem will have a directory in
144/sys/fs/f2fs based on its device name (i.e., /sys/fs/f2fs/sda).
145The files in each per-device directory are shown in table below.
146
147Files in /sys/fs/f2fs/<devname>
148(see also Documentation/ABI/testing/sysfs-fs-f2fs)
149..............................................................................
150 File Content
151
152 gc_max_sleep_time This tuning parameter controls the maximum sleep
153 time for the garbage collection thread. Time is
154 in milliseconds.
155
156 gc_min_sleep_time This tuning parameter controls the minimum sleep
157 time for the garbage collection thread. Time is
158 in milliseconds.
159
160 gc_no_gc_sleep_time This tuning parameter controls the default sleep
161 time for the garbage collection thread. Time is
162 in milliseconds.
163
Namjae Jeond2dc0952013-08-04 23:10:15 +0900164 gc_idle This parameter controls the selection of victim
165 policy for garbage collection. Setting gc_idle = 0
166 (default) will disable this option. Setting
167 gc_idle = 1 will select the Cost Benefit approach
168 & setting gc_idle = 2 will select the greedy aproach.
169
Jaegeuk Kimea91e9b2013-10-24 15:49:07 +0900170 reclaim_segments This parameter controls the number of prefree
171 segments to be reclaimed. If the number of prefree
172 segments is larger than this number, f2fs tries to
173 conduct checkpoint to reclaim the prefree segments
174 to free segments. By default, 100 segments, 200MB.
175
Jaegeuk Kimba0697e2013-12-19 17:44:41 +0900176 max_small_discards This parameter controls the number of discard
177 commands that consist small blocks less than 2MB.
178 The candidates to be discarded are cached until
179 checkpoint is triggered, and issued during the
180 checkpoint. By default, it is disabled with 0.
181
Jaegeuk Kim216fbd62013-11-07 13:13:42 +0900182 ipu_policy This parameter controls the policy of in-place
183 updates in f2fs. There are five policies:
184 0: F2FS_IPU_FORCE, 1: F2FS_IPU_SSR,
185 2: F2FS_IPU_UTIL, 3: F2FS_IPU_SSR_UTIL,
186 4: F2FS_IPU_DISABLE.
187
188 min_ipu_util This parameter controls the threshold to trigger
189 in-place-updates. The number indicates percentage
190 of the filesystem utilization, and used by
191 F2FS_IPU_UTIL and F2FS_IPU_SSR_UTIL policies.
192
Jaegeuk Kim3bac3802014-01-09 21:00:06 +0900193 max_victim_search This parameter controls the number of trials to
194 find a victim segment when conducting SSR and
195 cleaning operations. The default value is 4096
196 which covers 8GB block address range.
197
Namjae Jeonb59d0ba2013-08-04 23:09:40 +0900198================================================================================
Jaegeuk Kim98e4da82012-11-02 17:05:42 +0900199USAGE
200================================================================================
201
2021. Download userland tools and compile them.
203
2042. Skip, if f2fs was compiled statically inside kernel.
205 Otherwise, insert the f2fs.ko module.
206 # insmod f2fs.ko
207
2083. Create a directory trying to mount
209 # mkdir /mnt/f2fs
210
2114. Format the block device, and then mount as f2fs
212 # mkfs.f2fs -l label /dev/block_device
213 # mount -t f2fs /dev/block_device /mnt/f2fs
214
Changman Leed51a7fb2013-07-04 17:12:47 +0900215mkfs.f2fs
216---------
217The mkfs.f2fs is for the use of formatting a partition as the f2fs filesystem,
218which builds a basic on-disk layout.
219
220The options consist of:
Changman Lee1571f842013-04-03 15:26:49 +0900221-l [label] : Give a volume label, up to 512 unicode name.
Jaegeuk Kim98e4da82012-11-02 17:05:42 +0900222-a [0 or 1] : Split start location of each area for heap-based allocation.
223 1 is set by default, which performs this.
224-o [int] : Set overprovision ratio in percent over volume size.
225 5 is set by default.
226-s [int] : Set the number of segments per section.
227 1 is set by default.
228-z [int] : Set the number of sections per zone.
229 1 is set by default.
230-e [str] : Set basic extension list. e.g. "mp3,gif,mov"
Changman Lee1571f842013-04-03 15:26:49 +0900231-t [0 or 1] : Disable discard command or not.
232 1 is set by default, which conducts discard.
Jaegeuk Kim98e4da82012-11-02 17:05:42 +0900233
Changman Leed51a7fb2013-07-04 17:12:47 +0900234fsck.f2fs
235---------
236The fsck.f2fs is a tool to check the consistency of an f2fs-formatted
237partition, which examines whether the filesystem metadata and user-made data
238are cross-referenced correctly or not.
239Note that, initial version of the tool does not fix any inconsistency.
240
241The options consist of:
242 -d debug level [default:0]
243
244dump.f2fs
245---------
246The dump.f2fs shows the information of specific inode and dumps SSA and SIT to
247file. Each file is dump_ssa and dump_sit.
248
249The dump.f2fs is used to debug on-disk data structures of the f2fs filesystem.
250It shows on-disk inode information reconized by a given inode number, and is
251able to dump all the SSA and SIT entries into predefined files, ./dump_ssa and
252./dump_sit respectively.
253
254The options consist of:
255 -d debug level [default:0]
256 -i inode no (hex)
257 -s [SIT dump segno from #1~#2 (decimal), for all 0~-1]
258 -a [SSA dump segno from #1~#2 (decimal), for all 0~-1]
259
260Examples:
261# dump.f2fs -i [ino] /dev/sdx
262# dump.f2fs -s 0~-1 /dev/sdx (SIT dump)
263# dump.f2fs -a 0~-1 /dev/sdx (SSA dump)
264
Jaegeuk Kim98e4da82012-11-02 17:05:42 +0900265================================================================================
266DESIGN
267================================================================================
268
269On-disk Layout
270--------------
271
272F2FS divides the whole volume into a number of segments, each of which is fixed
273to 2MB in size. A section is composed of consecutive segments, and a zone
274consists of a set of sections. By default, section and zone sizes are set to one
275segment size identically, but users can easily modify the sizes by mkfs.
276
277F2FS splits the entire volume into six areas, and all the areas except superblock
278consists of multiple segments as described below.
279
280 align with the zone size <-|
281 |-> align with the segment size
282 _________________________________________________________________________
Huajun Li9268cc32012-12-31 13:59:04 +0800283 | | | Segment | Node | Segment | |
284 | Superblock | Checkpoint | Info. | Address | Summary | Main |
285 | (SB) | (CP) | Table (SIT) | Table (NAT) | Area (SSA) | |
Jaegeuk Kim98e4da82012-11-02 17:05:42 +0900286 |____________|_____2______|______N______|______N______|______N_____|__N___|
287 . .
288 . .
289 . .
290 ._________________________________________.
291 |_Segment_|_..._|_Segment_|_..._|_Segment_|
292 . .
293 ._________._________
294 |_section_|__...__|_
295 . .
296 .________.
297 |__zone__|
298
299- Superblock (SB)
300 : It is located at the beginning of the partition, and there exist two copies
301 to avoid file system crash. It contains basic partition information and some
302 default parameters of f2fs.
303
304- Checkpoint (CP)
305 : It contains file system information, bitmaps for valid NAT/SIT sets, orphan
306 inode lists, and summary entries of current active segments.
307
Jaegeuk Kim98e4da82012-11-02 17:05:42 +0900308- Segment Information Table (SIT)
309 : It contains segment information such as valid block count and bitmap for the
310 validity of all the blocks.
311
Huajun Li9268cc32012-12-31 13:59:04 +0800312- Node Address Table (NAT)
313 : It is composed of a block address table for all the node blocks stored in
314 Main area.
315
Jaegeuk Kim98e4da82012-11-02 17:05:42 +0900316- Segment Summary Area (SSA)
317 : It contains summary entries which contains the owner information of all the
318 data and node blocks stored in Main area.
319
320- Main Area
321 : It contains file and directory data including their indices.
322
323In order to avoid misalignment between file system and flash-based storage, F2FS
324aligns the start block address of CP with the segment size. Also, it aligns the
325start block address of Main area with the zone size by reserving some segments
326in SSA area.
327
328Reference the following survey for additional technical details.
329https://wiki.linaro.org/WorkingGroups/Kernel/Projects/FlashCardSurvey
330
331File System Metadata Structure
332------------------------------
333
334F2FS adopts the checkpointing scheme to maintain file system consistency. At
335mount time, F2FS first tries to find the last valid checkpoint data by scanning
336CP area. In order to reduce the scanning time, F2FS uses only two copies of CP.
337One of them always indicates the last valid data, which is called as shadow copy
338mechanism. In addition to CP, NAT and SIT also adopt the shadow copy mechanism.
339
340For file system consistency, each CP points to which NAT and SIT copies are
341valid, as shown as below.
342
343 +--------+----------+---------+
Huajun Li9268cc32012-12-31 13:59:04 +0800344 | CP | SIT | NAT |
Jaegeuk Kim98e4da82012-11-02 17:05:42 +0900345 +--------+----------+---------+
346 . . . .
347 . . . .
348 . . . .
349 +-------+-------+--------+--------+--------+--------+
Huajun Li9268cc32012-12-31 13:59:04 +0800350 | CP #0 | CP #1 | SIT #0 | SIT #1 | NAT #0 | NAT #1 |
Jaegeuk Kim98e4da82012-11-02 17:05:42 +0900351 +-------+-------+--------+--------+--------+--------+
352 | ^ ^
353 | | |
354 `----------------------------------------'
355
356Index Structure
357---------------
358
359The key data structure to manage the data locations is a "node". Similar to
360traditional file structures, F2FS has three types of node: inode, direct node,
Huajun Lid08ab082012-12-05 16:45:32 +0800361indirect node. F2FS assigns 4KB to an inode block which contains 923 data block
Jaegeuk Kim98e4da82012-11-02 17:05:42 +0900362indices, two direct node pointers, two indirect node pointers, and one double
363indirect node pointer as described below. One direct node block contains 1018
364data blocks, and one indirect node block contains also 1018 node blocks. Thus,
365one inode block (i.e., a file) covers:
366
367 4KB * (923 + 2 * 1018 + 2 * 1018 * 1018 + 1018 * 1018 * 1018) := 3.94TB.
368
369 Inode block (4KB)
370 |- data (923)
371 |- direct node (2)
372 | `- data (1018)
373 |- indirect node (2)
374 | `- direct node (1018)
375 | `- data (1018)
376 `- double indirect node (1)
377 `- indirect node (1018)
378 `- direct node (1018)
379 `- data (1018)
380
381Note that, all the node blocks are mapped by NAT which means the location of
382each node is translated by the NAT table. In the consideration of the wandering
383tree problem, F2FS is able to cut off the propagation of node updates caused by
384leaf data writes.
385
386Directory Structure
387-------------------
388
389A directory entry occupies 11 bytes, which consists of the following attributes.
390
391- hash hash value of the file name
392- ino inode number
393- len the length of file name
394- type file type such as directory, symlink, etc
395
396A dentry block consists of 214 dentry slots and file names. Therein a bitmap is
397used to represent whether each dentry is valid or not. A dentry block occupies
3984KB with the following composition.
399
400 Dentry Block(4 K) = bitmap (27 bytes) + reserved (3 bytes) +
401 dentries(11 * 214 bytes) + file name (8 * 214 bytes)
402
403 [Bucket]
404 +--------------------------------+
405 |dentry block 1 | dentry block 2 |
406 +--------------------------------+
407 . .
408 . .
409 . [Dentry Block Structure: 4KB] .
410 +--------+----------+----------+------------+
411 | bitmap | reserved | dentries | file names |
412 +--------+----------+----------+------------+
413 [Dentry Block: 4KB] . .
414 . .
415 . .
416 +------+------+-----+------+
417 | hash | ino | len | type |
418 +------+------+-----+------+
419 [Dentry Structure: 11 bytes]
420
421F2FS implements multi-level hash tables for directory structure. Each level has
422a hash table with dedicated number of hash buckets as shown below. Note that
423"A(2B)" means a bucket includes 2 data blocks.
424
425----------------------
426A : bucket
427B : block
428N : MAX_DIR_HASH_DEPTH
429----------------------
430
431level #0 | A(2B)
432 |
433level #1 | A(2B) - A(2B)
434 |
435level #2 | A(2B) - A(2B) - A(2B) - A(2B)
436 . | . . . .
437level #N/2 | A(2B) - A(2B) - A(2B) - A(2B) - A(2B) - ... - A(2B)
438 . | . . . .
439level #N | A(4B) - A(4B) - A(4B) - A(4B) - A(4B) - ... - A(4B)
440
441The number of blocks and buckets are determined by,
442
443 ,- 2, if n < MAX_DIR_HASH_DEPTH / 2,
444 # of blocks in level #n = |
445 `- 4, Otherwise
446
447 ,- 2^n, if n < MAX_DIR_HASH_DEPTH / 2,
448 # of buckets in level #n = |
449 `- 2^((MAX_DIR_HASH_DEPTH / 2) - 1), Otherwise
450
451When F2FS finds a file name in a directory, at first a hash value of the file
452name is calculated. Then, F2FS scans the hash table in level #0 to find the
453dentry consisting of the file name and its inode number. If not found, F2FS
454scans the next hash table in level #1. In this way, F2FS scans hash tables in
455each levels incrementally from 1 to N. In each levels F2FS needs to scan only
456one bucket determined by the following equation, which shows O(log(# of files))
457complexity.
458
459 bucket number to scan in level #n = (hash value) % (# of buckets in level #n)
460
461In the case of file creation, F2FS finds empty consecutive slots that cover the
462file name. F2FS searches the empty slots in the hash tables of whole levels from
4631 to N in the same way as the lookup operation.
464
465The following figure shows an example of two cases holding children.
466 --------------> Dir <--------------
467 | |
468 child child
469
470 child - child [hole] - child
471
472 child - child - child [hole] - [hole] - child
473
474 Case 1: Case 2:
475 Number of children = 6, Number of children = 3,
476 File size = 7 File size = 7
477
478Default Block Allocation
479------------------------
480
481At runtime, F2FS manages six active logs inside "Main" area: Hot/Warm/Cold node
482and Hot/Warm/Cold data.
483
484- Hot node contains direct node blocks of directories.
485- Warm node contains direct node blocks except hot node blocks.
486- Cold node contains indirect node blocks
487- Hot data contains dentry blocks
488- Warm data contains data blocks except hot and cold data blocks
489- Cold data contains multimedia data or migrated data blocks
490
491LFS has two schemes for free space management: threaded log and copy-and-compac-
492tion. The copy-and-compaction scheme which is known as cleaning, is well-suited
493for devices showing very good sequential write performance, since free segments
494are served all the time for writing new data. However, it suffers from cleaning
495overhead under high utilization. Contrarily, the threaded log scheme suffers
496from random writes, but no cleaning process is needed. F2FS adopts a hybrid
497scheme where the copy-and-compaction scheme is adopted by default, but the
498policy is dynamically changed to the threaded log scheme according to the file
499system status.
500
501In order to align F2FS with underlying flash-based storage, F2FS allocates a
502segment in a unit of section. F2FS expects that the section size would be the
503same as the unit size of garbage collection in FTL. Furthermore, with respect
504to the mapping granularity in FTL, F2FS allocates each section of the active
505logs from different zones as much as possible, since FTL can write the data in
506the active logs into one allocation unit according to its mapping granularity.
507
508Cleaning process
509----------------
510
511F2FS does cleaning both on demand and in the background. On-demand cleaning is
512triggered when there are not enough free segments to serve VFS calls. Background
513cleaner is operated by a kernel thread, and triggers the cleaning job when the
514system is idle.
515
516F2FS supports two victim selection policies: greedy and cost-benefit algorithms.
517In the greedy algorithm, F2FS selects a victim segment having the smallest number
518of valid blocks. In the cost-benefit algorithm, F2FS selects a victim segment
519according to the segment age and the number of valid blocks in order to address
520log block thrashing problem in the greedy algorithm. F2FS adopts the greedy
521algorithm for on-demand cleaner, while background cleaner adopts cost-benefit
522algorithm.
523
524In order to identify whether the data in the victim segment are valid or not,
525F2FS manages a bitmap. Each bit represents the validity of a block, and the
526bitmap is composed of a bit stream covering whole blocks in main area.