| blob | 4f6531a4701bef5529a5ca05c29a3556d4c696e9 |
1 ================================================================================
2 WHAT IS Flash-Friendly File System (F2FS)?
3 ================================================================================
5 NAND flash memory-based storage devices, such as SSD, eMMC, and SD cards, have
6 been equipped on a variety systems ranging from mobile to server systems. Since
7 they are known to have different characteristics from the conventional rotating
8 disks, a file system, an upper layer to the storage device, should adapt to the
9 changes from the sketch in the design level.
11 F2FS is a file system exploiting NAND flash memory-based storage devices, which
12 is based on Log-structured File System (LFS). The design has been focused on
13 addressing the fundamental issues in LFS, which are snowball effect of wandering
14 tree and high cleaning overhead.
16 Since a NAND flash memory-based storage device shows different characteristic
17 according to its internal geometry or flash memory management scheme, namely FTL,
18 F2FS and its tools support various parameters not only for configuring on-disk
19 layout, but also for selecting allocation and cleaning algorithms.
21 The following git tree provides the file system formatting tool (mkfs.f2fs),
22 a consistency checking tool (fsck.f2fs), and a debugging tool (dump.f2fs).
23 >> git://git.kernel.org/pub/scm/linux/kernel/git/jaegeuk/f2fs-tools.git
25 For reporting bugs and sending patches, please use the following mailing list:
26 >> linux-f2fs-devel@lists.sourceforge.net
28 ================================================================================
29 BACKGROUND AND DESIGN ISSUES
30 ================================================================================
32 Log-structured File System (LFS)
33 --------------------------------
34 "A log-structured file system writes all modifications to disk sequentially in
35 a log-like structure, thereby speeding up both file writing and crash recovery.
36 The log is the only structure on disk; it contains indexing information so that
37 files can be read back from the log efficiently. In order to maintain large free
38 areas on disk for fast writing, we divide the log into segments and use a
39 segment cleaner to compress the live information from heavily fragmented
40 segments." from Rosenblum, M. and Ousterhout, J. K., 1992, "The design and
41 implementation of a log-structured file system", ACM Trans. Computer Systems
42 10, 1, 26–52.
44 Wandering Tree Problem
45 ----------------------
46 In LFS, when a file data is updated and written to the end of log, its direct
47 pointer block is updated due to the changed location. Then the indirect pointer
48 block is also updated due to the direct pointer block update. In this manner,
49 the upper index structures such as inode, inode map, and checkpoint block are
50 also updated recursively. This problem is called as wandering tree problem [1],
51 and in order to enhance the performance, it should eliminate or relax the update
52 propagation as much as possible.
54 [1] Bityutskiy, A. 2005. JFFS3 design issues. http://www.linux-mtd.infradead.org/
56 Cleaning Overhead
57 -----------------
58 Since LFS is based on out-of-place writes, it produces so many obsolete blocks
59 scattered across the whole storage. In order to serve new empty log space, it
60 needs to reclaim these obsolete blocks seamlessly to users. This job is called
61 as a cleaning process.
63 The process consists of three operations as follows.
64 1. A victim segment is selected through referencing segment usage table.
65 2. It loads parent index structures of all the data in the victim identified by
66 segment summary blocks.
67 3. It checks the cross-reference between the data and its parent index structure.
68 4. It moves valid data selectively.
70 This cleaning job may cause unexpected long delays, so the most important goal
71 is to hide the latencies to users. And also definitely, it should reduce the
72 amount of valid data to be moved, and move them quickly as well.
74 ================================================================================
75 KEY FEATURES
76 ================================================================================
78 Flash 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
84 Wandering 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.
90 Cleaning 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
97 ================================================================================
98 MOUNT OPTIONS
99 ================================================================================
101 background_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 turned off. If background_gc=sync, it will turn
106 on synchronous garbage collection running in background.
107 Default value for this option is on. So garbage
108 collection is on by default.
109 disable_roll_forward Disable the roll-forward recovery routine
110 norecovery Disable the roll-forward recovery routine, mounted read-
111 only (i.e., -o ro,disable_roll_forward)
112 discard/nodiscard Enable/disable real-time discard in f2fs, if discard is
113 enabled, f2fs will issue discard/TRIM commands when a
114 segment is cleaned.
115 no_heap Disable heap-style segment allocation which finds free
116 segments for data from the beginning of main area, while
117 for node from the end of main area.
118 nouser_xattr Disable Extended User Attributes. Note: xattr is enabled
119 by default if CONFIG_F2FS_FS_XATTR is selected.
120 noacl Disable POSIX Access Control List. Note: acl is enabled
121 by default if CONFIG_F2FS_FS_POSIX_ACL is selected.
122 active_logs=%u Support configuring the number of active logs. In the
123 current design, f2fs supports only 2, 4, and 6 logs.
124 Default number is 6.
125 disable_ext_identify Disable the extension list configured by mkfs, so f2fs
126 does not aware of cold files such as media files.
127 inline_xattr Enable the inline xattrs feature.
128 noinline_xattr Disable the inline xattrs feature.
129 inline_data Enable the inline data feature: New created small(<~3.4k)
130 files can be written into inode block.
131 inline_dentry Enable the inline dir feature: data in new created
132 directory entries can be written into inode block. The
133 space of inode block which is used to store inline
134 dentries is limited to ~3.4k.
135 noinline_dentry Disable the inline dentry feature.
136 flush_merge Merge concurrent cache_flush commands as much as possible
137 to eliminate redundant command issues. If the underlying
138 device handles the cache_flush command relatively slowly,
139 recommend to enable this option.
140 nobarrier This option can be used if underlying storage guarantees
141 its cached data should be written to the novolatile area.
142 If this option is set, no cache_flush commands are issued
143 but f2fs still guarantees the write ordering of all the
144 data writes.
145 fastboot This option is used when a system wants to reduce mount
146 time as much as possible, even though normal performance
147 can be sacrificed.
148 extent_cache Enable an extent cache based on rb-tree, it can cache
149 as many as extent which map between contiguous logical
150 address and physical address per inode, resulting in
151 increasing the cache hit ratio. Set by default.
152 noextent_cache Disable an extent cache based on rb-tree explicitly, see
153 the above extent_cache mount option.
154 noinline_data Disable the inline data feature, inline data feature is
155 enabled by default.
156 data_flush Enable data flushing before checkpoint in order to
157 persist data of regular and symlink.
158 mode=%s Control block allocation mode which supports "adaptive"
159 and "lfs". In "lfs" mode, there should be no random
160 writes towards main area.
161 io_bits=%u Set the bit size of write IO requests. It should be set
162 with "mode=lfs".
164 ================================================================================
165 DEBUGFS ENTRIES
166 ================================================================================
168 /sys/kernel/debug/f2fs/ contains information about all the partitions mounted as
169 f2fs. Each file shows the whole f2fs information.
171 /sys/kernel/debug/f2fs/status includes:
172 - major file system information managed by f2fs currently
173 - average SIT information about whole segments
174 - current memory footprint consumed by f2fs.
176 ================================================================================
177 SYSFS ENTRIES
178 ================================================================================
180 Information about mounted f2fs file systems can be found in
181 /sys/fs/f2fs. Each mounted filesystem will have a directory in
182 /sys/fs/f2fs based on its device name (i.e., /sys/fs/f2fs/sda).
183 The files in each per-device directory are shown in table below.
185 Files in /sys/fs/f2fs/<devname>
186 (see also Documentation/ABI/testing/sysfs-fs-f2fs)
187 ..............................................................................
188 File Content
190 gc_max_sleep_time This tuning parameter controls the maximum sleep
191 time for the garbage collection thread. Time is
192 in milliseconds.
194 gc_min_sleep_time This tuning parameter controls the minimum sleep
195 time for the garbage collection thread. Time is
196 in milliseconds.
198 gc_no_gc_sleep_time This tuning parameter controls the default sleep
199 time for the garbage collection thread. Time is
200 in milliseconds.
202 gc_idle This parameter controls the selection of victim
203 policy for garbage collection. Setting gc_idle = 0
204 (default) will disable this option. Setting
205 gc_idle = 1 will select the Cost Benefit approach
206 & setting gc_idle = 2 will select the greedy approach.
208 reclaim_segments This parameter controls the number of prefree
209 segments to be reclaimed. If the number of prefree
210 segments is larger than the number of segments
211 in the proportion to the percentage over total
212 volume size, f2fs tries to conduct checkpoint to
213 reclaim the prefree segments to free segments.
214 By default, 5% over total # of segments.
216 max_small_discards This parameter controls the number of discard
217 commands that consist small blocks less than 2MB.
218 The candidates to be discarded are cached until
219 checkpoint is triggered, and issued during the
220 checkpoint. By default, it is disabled with 0.
222 trim_sections This parameter controls the number of sections
223 to be trimmed out in batch mode when FITRIM
224 conducts. 32 sections is set by default.
226 ipu_policy This parameter controls the policy of in-place
227 updates in f2fs. There are five policies:
228 0x01: F2FS_IPU_FORCE, 0x02: F2FS_IPU_SSR,
229 0x04: F2FS_IPU_UTIL, 0x08: F2FS_IPU_SSR_UTIL,
230 0x10: F2FS_IPU_FSYNC.
232 min_ipu_util This parameter controls the threshold to trigger
233 in-place-updates. The number indicates percentage
234 of the filesystem utilization, and used by
235 F2FS_IPU_UTIL and F2FS_IPU_SSR_UTIL policies.
237 min_fsync_blocks This parameter controls the threshold to trigger
238 in-place-updates when F2FS_IPU_FSYNC mode is set.
239 The number indicates the number of dirty pages
240 when fsync needs to flush on its call path. If
241 the number is less than this value, it triggers
242 in-place-updates.
244 max_victim_search This parameter controls the number of trials to
245 find a victim segment when conducting SSR and
246 cleaning operations. The default value is 4096
247 which covers 8GB block address range.
249 dir_level This parameter controls the directory level to
250 support large directory. If a directory has a
251 number of files, it can reduce the file lookup
252 latency by increasing this dir_level value.
253 Otherwise, it needs to decrease this value to
254 reduce the space overhead. The default value is 0.
256 ram_thresh This parameter controls the memory footprint used
257 by free nids and cached nat entries. By default,
258 10 is set, which indicates 10 MB / 1 GB RAM.
260 ================================================================================
261 USAGE
262 ================================================================================
264 1. Download userland tools and compile them.
266 2. Skip, if f2fs was compiled statically inside kernel.
267 Otherwise, insert the f2fs.ko module.
268 # insmod f2fs.ko
270 3. Create a directory trying to mount
271 # mkdir /mnt/f2fs
273 4. Format the block device, and then mount as f2fs
274 # mkfs.f2fs -l label /dev/block_device
275 # mount -t f2fs /dev/block_device /mnt/f2fs
277 mkfs.f2fs
278 ---------
279 The mkfs.f2fs is for the use of formatting a partition as the f2fs filesystem,
280 which builds a basic on-disk layout.
282 The options consist of:
283 -l [label] : Give a volume label, up to 512 unicode name.
284 -a [0 or 1] : Split start location of each area for heap-based allocation.
285 1 is set by default, which performs this.
286 -o [int] : Set overprovision ratio in percent over volume size.
287 5 is set by default.
288 -s [int] : Set the number of segments per section.
289 1 is set by default.
290 -z [int] : Set the number of sections per zone.
291 1 is set by default.
292 -e [str] : Set basic extension list. e.g. "mp3,gif,mov"
293 -t [0 or 1] : Disable discard command or not.
294 1 is set by default, which conducts discard.
296 fsck.f2fs
297 ---------
298 The fsck.f2fs is a tool to check the consistency of an f2fs-formatted
299 partition, which examines whether the filesystem metadata and user-made data
300 are cross-referenced correctly or not.
301 Note that, initial version of the tool does not fix any inconsistency.
303 The options consist of:
304 -d debug level [default:0]
306 dump.f2fs
307 ---------
308 The dump.f2fs shows the information of specific inode and dumps SSA and SIT to
309 file. Each file is dump_ssa and dump_sit.
311 The dump.f2fs is used to debug on-disk data structures of the f2fs filesystem.
312 It shows on-disk inode information recognized by a given inode number, and is
313 able to dump all the SSA and SIT entries into predefined files, ./dump_ssa and
314 ./dump_sit respectively.
316 The options consist of:
317 -d debug level [default:0]
318 -i inode no (hex)
319 -s [SIT dump segno from #1~#2 (decimal), for all 0~-1]
320 -a [SSA dump segno from #1~#2 (decimal), for all 0~-1]
322 Examples:
323 # dump.f2fs -i [ino] /dev/sdx
324 # dump.f2fs -s 0~-1 /dev/sdx (SIT dump)
325 # dump.f2fs -a 0~-1 /dev/sdx (SSA dump)
327 ================================================================================
328 DESIGN
329 ================================================================================
331 On-disk Layout
332 --------------
334 F2FS divides the whole volume into a number of segments, each of which is fixed
335 to 2MB in size. A section is composed of consecutive segments, and a zone
336 consists of a set of sections. By default, section and zone sizes are set to one
337 segment size identically, but users can easily modify the sizes by mkfs.
339 F2FS splits the entire volume into six areas, and all the areas except superblock
340 consists of multiple segments as described below.
342 align with the zone size <-|
343 |-> align with the segment size
344 _________________________________________________________________________
345 | | | Segment | Node | Segment | |
346 | Superblock | Checkpoint | Info. | Address | Summary | Main |
347 | (SB) | (CP) | Table (SIT) | Table (NAT) | Area (SSA) | |
348 |____________|_____2______|______N______|______N______|______N_____|__N___|
349 . .
350 . .
351 . .
352 ._________________________________________.
353 |_Segment_|_..._|_Segment_|_..._|_Segment_|
354 . .
355 ._________._________
356 |_section_|__...__|_
357 . .
358 .________.
359 |__zone__|
361 - Superblock (SB)
362 : It is located at the beginning of the partition, and there exist two copies
363 to avoid file system crash. It contains basic partition information and some
364 default parameters of f2fs.
366 - Checkpoint (CP)
367 : It contains file system information, bitmaps for valid NAT/SIT sets, orphan
368 inode lists, and summary entries of current active segments.
370 - Segment Information Table (SIT)
371 : It contains segment information such as valid block count and bitmap for the
372 validity of all the blocks.
374 - Node Address Table (NAT)
375 : It is composed of a block address table for all the node blocks stored in
376 Main area.
378 - Segment Summary Area (SSA)
379 : It contains summary entries which contains the owner information of all the
380 data and node blocks stored in Main area.
382 - Main Area
383 : It contains file and directory data including their indices.
385 In order to avoid misalignment between file system and flash-based storage, F2FS
386 aligns the start block address of CP with the segment size. Also, it aligns the
387 start block address of Main area with the zone size by reserving some segments
388 in SSA area.
390 Reference the following survey for additional technical details.
391 https://wiki.linaro.org/WorkingGroups/Kernel/Projects/FlashCardSurvey
393 File System Metadata Structure
394 ------------------------------
396 F2FS adopts the checkpointing scheme to maintain file system consistency. At
397 mount time, F2FS first tries to find the last valid checkpoint data by scanning
398 CP area. In order to reduce the scanning time, F2FS uses only two copies of CP.
399 One of them always indicates the last valid data, which is called as shadow copy
400 mechanism. In addition to CP, NAT and SIT also adopt the shadow copy mechanism.
402 For file system consistency, each CP points to which NAT and SIT copies are
403 valid, as shown as below.
405 +--------+----------+---------+
406 | CP | SIT | NAT |
407 +--------+----------+---------+
408 . . . .
409 . . . .
410 . . . .
411 +-------+-------+--------+--------+--------+--------+
412 | CP #0 | CP #1 | SIT #0 | SIT #1 | NAT #0 | NAT #1 |
413 +-------+-------+--------+--------+--------+--------+
414 | ^ ^
415 | | |
416 `----------------------------------------'
418 Index Structure
419 ---------------
421 The key data structure to manage the data locations is a "node". Similar to
422 traditional file structures, F2FS has three types of node: inode, direct node,
423 indirect node. F2FS assigns 4KB to an inode block which contains 923 data block
424 indices, two direct node pointers, two indirect node pointers, and one double
425 indirect node pointer as described below. One direct node block contains 1018
426 data blocks, and one indirect node block contains also 1018 node blocks. Thus,
427 one inode block (i.e., a file) covers:
429 4KB * (923 + 2 * 1018 + 2 * 1018 * 1018 + 1018 * 1018 * 1018) := 3.94TB.
431 Inode block (4KB)
432 |- data (923)
433 |- direct node (2)
434 | `- data (1018)
435 |- indirect node (2)
436 | `- direct node (1018)
437 | `- data (1018)
438 `- double indirect node (1)
439 `- indirect node (1018)
440 `- direct node (1018)
441 `- data (1018)
443 Note that, all the node blocks are mapped by NAT which means the location of
444 each node is translated by the NAT table. In the consideration of the wandering
445 tree problem, F2FS is able to cut off the propagation of node updates caused by
446 leaf data writes.
448 Directory Structure
449 -------------------
451 A directory entry occupies 11 bytes, which consists of the following attributes.
453 - hash hash value of the file name
454 - ino inode number
455 - len the length of file name
456 - type file type such as directory, symlink, etc
458 A dentry block consists of 214 dentry slots and file names. Therein a bitmap is
459 used to represent whether each dentry is valid or not. A dentry block occupies
460 4KB with the following composition.
462 Dentry Block(4 K) = bitmap (27 bytes) + reserved (3 bytes) +
463 dentries(11 * 214 bytes) + file name (8 * 214 bytes)
465 [Bucket]
466 +--------------------------------+
467 |dentry block 1 | dentry block 2 |
468 +--------------------------------+
469 . .
470 . .
471 . [Dentry Block Structure: 4KB] .
472 +--------+----------+----------+------------+
473 | bitmap | reserved | dentries | file names |
474 +--------+----------+----------+------------+
475 [Dentry Block: 4KB] . .
476 . .
477 . .
478 +------+------+-----+------+
479 | hash | ino | len | type |
480 +------+------+-----+------+
481 [Dentry Structure: 11 bytes]
483 F2FS implements multi-level hash tables for directory structure. Each level has
484 a hash table with dedicated number of hash buckets as shown below. Note that
485 "A(2B)" means a bucket includes 2 data blocks.
487 ----------------------
488 A : bucket
489 B : block
490 N : MAX_DIR_HASH_DEPTH
491 ----------------------
493 level #0 | A(2B)
494 |
495 level #1 | A(2B) - A(2B)
496 |
497 level #2 | A(2B) - A(2B) - A(2B) - A(2B)
498 . | . . . .
499 level #N/2 | A(2B) - A(2B) - A(2B) - A(2B) - A(2B) - ... - A(2B)
500 . | . . . .
501 level #N | A(4B) - A(4B) - A(4B) - A(4B) - A(4B) - ... - A(4B)
503 The number of blocks and buckets are determined by,
505 ,- 2, if n < MAX_DIR_HASH_DEPTH / 2,
506 # of blocks in level #n = |
507 `- 4, Otherwise
509 ,- 2^(n + dir_level),
510 | if n + dir_level < MAX_DIR_HASH_DEPTH / 2,
511 # of buckets in level #n = |
512 `- 2^((MAX_DIR_HASH_DEPTH / 2) - 1),
513 Otherwise
515 When F2FS finds a file name in a directory, at first a hash value of the file
516 name is calculated. Then, F2FS scans the hash table in level #0 to find the
517 dentry consisting of the file name and its inode number. If not found, F2FS
518 scans the next hash table in level #1. In this way, F2FS scans hash tables in
519 each levels incrementally from 1 to N. In each levels F2FS needs to scan only
520 one bucket determined by the following equation, which shows O(log(# of files))
521 complexity.
523 bucket number to scan in level #n = (hash value) % (# of buckets in level #n)
525 In the case of file creation, F2FS finds empty consecutive slots that cover the
526 file name. F2FS searches the empty slots in the hash tables of whole levels from
527 1 to N in the same way as the lookup operation.
529 The following figure shows an example of two cases holding children.
530 --------------> Dir <--------------
531 | |
532 child child
534 child - child [hole] - child
536 child - child - child [hole] - [hole] - child
538 Case 1: Case 2:
539 Number of children = 6, Number of children = 3,
540 File size = 7 File size = 7
542 Default Block Allocation
543 ------------------------
545 At runtime, F2FS manages six active logs inside "Main" area: Hot/Warm/Cold node
546 and Hot/Warm/Cold data.
548 - Hot node contains direct node blocks of directories.
549 - Warm node contains direct node blocks except hot node blocks.
550 - Cold node contains indirect node blocks
551 - Hot data contains dentry blocks
552 - Warm data contains data blocks except hot and cold data blocks
553 - Cold data contains multimedia data or migrated data blocks
555 LFS has two schemes for free space management: threaded log and copy-and-compac-
556 tion. The copy-and-compaction scheme which is known as cleaning, is well-suited
557 for devices showing very good sequential write performance, since free segments
558 are served all the time for writing new data. However, it suffers from cleaning
559 overhead under high utilization. Contrarily, the threaded log scheme suffers
560 from random writes, but no cleaning process is needed. F2FS adopts a hybrid
561 scheme where the copy-and-compaction scheme is adopted by default, but the
562 policy is dynamically changed to the threaded log scheme according to the file
563 system status.
565 In order to align F2FS with underlying flash-based storage, F2FS allocates a
566 segment in a unit of section. F2FS expects that the section size would be the
567 same as the unit size of garbage collection in FTL. Furthermore, with respect
568 to the mapping granularity in FTL, F2FS allocates each section of the active
569 logs from different zones as much as possible, since FTL can write the data in
570 the active logs into one allocation unit according to its mapping granularity.
572 Cleaning process
573 ----------------
575 F2FS does cleaning both on demand and in the background. On-demand cleaning is
576 triggered when there are not enough free segments to serve VFS calls. Background
577 cleaner is operated by a kernel thread, and triggers the cleaning job when the
578 system is idle.
580 F2FS supports two victim selection policies: greedy and cost-benefit algorithms.
581 In the greedy algorithm, F2FS selects a victim segment having the smallest number
582 of valid blocks. In the cost-benefit algorithm, F2FS selects a victim segment
583 according to the segment age and the number of valid blocks in order to address
584 log block thrashing problem in the greedy algorithm. F2FS adopts the greedy
585 algorithm for on-demand cleaner, while background cleaner adopts cost-benefit
586 algorithm.
588 In order to identify whether the data in the victim segment are valid or not,
589 F2FS manages a bitmap. Each bit represents the validity of a block, and the
590 bitmap is composed of a bit stream covering whole blocks in main area.