1 .. SPDX-License-Identifier: GPL-2.0
2
3 ==============================================
4 ZoneFS - Zone filesystem for Zoned block devic
5 ==============================================
6
7 Introduction
8 ============
9
10 zonefs is a very simple file system exposing e
11 as a file. Unlike a regular POSIX-compliant fi
12 device support (e.g. f2fs), zonefs does not hi
13 constraint of zoned block devices to the user.
14 write zones of the device must be written sequ
15 of the file (append only writes).
16
17 As such, zonefs is in essence closer to a raw
18 than to a full-featured POSIX file system. The
19 the implementation of zoned block device suppo
20 raw block device file accesses with a richer f
21 direct block device file ioctls which may be m
22 example of this approach is the implementation
23 tree structures (such as used in RocksDB and L
24 by allowing SSTables to be stored in a zone fi
25 system rather than as a range of sectors of th
26 of the higher level construct "one file is one
27 amount of changes needed in the application as
28 different application programming languages.
29
30 Zoned block devices
31 -------------------
32
33 Zoned storage devices belong to a class of sto
34 space that is divided into zones. A zone is a
35 zones are contiguous (there are no LBA gaps).
36
37 * Conventional zones: there are no access cons
38 conventional zones. Any read or write access
39 regular block device.
40 * Sequential zones: these zones accept random
41 sequentially. Each sequential zone has a wri
42 device that keeps track of the mandatory sta
43 to the device. As a result of this write con
44 cannot be overwritten. Sequential zones must
45 command (zone reset) before rewriting.
46
47 Zoned storage devices can be implemented using
48 technologies. The most common form of zoned st
49 Block Commands (ZBC) and Zoned ATA Commands (Z
50 Magnetic Recording (SMR) HDDs.
51
52 Solid State Disks (SSD) storage devices can al
53 to, for instance, reduce internal write amplif
54 The NVMe Zoned NameSpace (ZNS) is a technical
55 committee aiming at adding a zoned storage int
56
57 Zonefs Overview
58 ===============
59
60 Zonefs exposes the zones of a zoned block devi
61 representing zones are grouped by zone type, w
62 by sub-directories. This file structure is bui
63 provided by the device and so does not require
64 structure.
65
66 On-disk metadata
67 ----------------
68
69 zonefs on-disk metadata is reduced to an immut
70 persistently stores a magic number and optiona
71 mount, zonefs uses blkdev_report_zones() to ob
72 and populates the mount point with a static fi
73 information. File sizes come from the device z
74 position managed by the device itself.
75
76 The super block is always written on disk at s
77 device storing the super block is never expose
78 the zone containing the super block is a seque
79 tool always "finishes" the zone, that is, it t
80 state to make it read-only, preventing any dat
81
82 Zone type sub-directories
83 -------------------------
84
85 Files representing zones of the same type are
86 sub-directory automatically created on mount.
87
88 For conventional zones, the sub-directory "cnv
89 however created if and only if the device has
90 the device only has a single conventional zone
91 be exposed as a file as it will be used to sto
92 such devices, the "cnv" sub-directory will not
93
94 For sequential write zones, the sub-directory
95
96 These two directories are the only directories
97 cannot create other directories and cannot ren
98 "seq" sub-directories.
99
100 The size of the directories indicated by the s
101 obtained with the stat() or fstat() system cal
102 existing under the directory.
103
104 Zone files
105 ----------
106
107 Zone files are named using the number of the z
108 of zones of a particular type. That is, both t
109 contain files named "0", "1", "2", ... The fil
110 increasing zone start sector on the device.
111
112 All read and write operations to zone files ar
113 maximum size, that is, beyond the zone capacit
114 capacity is failed with the -EFBIG error.
115
116 Creating, deleting, renaming or modifying any
117 sub-directories is not allowed.
118
119 The number of blocks of a file as reported by
120 capacity of the zone file, or in other words,
121
122 Conventional zone files
123 -----------------------
124
125 The size of conventional zone files is fixed t
126 represent. Conventional zone files cannot be t
127
128 These files can be randomly read and written u
129 buffered I/Os, direct I/Os, memory mapped I/Os
130 constraint for these files beyond the file siz
131
132 Sequential zone files
133 ---------------------
134
135 The size of sequential zone files grouped in t
136 the file's zone write pointer position relativ
137
138 Sequential zone files can only be written sequ
139 end, that is, write operations can only be app
140 attempt at accepting random writes and will fa
141 start offset not corresponding to the end of t
142 write issued and still in-flight (for asynchro
143
144 Since dirty page writeback by the page cache d
145 write pattern, zonefs prevents buffered writes
146 on sequential files. Only direct I/O writes ar
147 zonefs relies on the sequential delivery of wr
148 implemented by the block layer elevator. An el
149 write feature for zoned block device (ELEVATOR
150 must be used. This type of elevator (e.g. mq-d
151 for zoned block devices on device initializati
152
153 There are no restrictions on the type of I/O u
154 sequential zone files. Buffered I/Os, direct I
155 all accepted.
156
157 Truncating sequential zone files is allowed on
158 zone is reset to rewind the file zone write po
159 the zone, or up to the zone capacity, in which
160 transitioned to the FULL state (finish zone op
161
162 Format options
163 --------------
164
165 Several optional features of zonefs can be ena
166
167 * Conventional zone aggregation: ranges of con
168 aggregated into a single larger file instead
169 * File ownership: The owner UID and GID of zon
170 but can be changed to any valid UID/GID.
171 * File access permissions: the default 640 acc
172
173 IO error handling
174 -----------------
175
176 Zoned block devices may fail I/O requests for
177 devices, e.g. due to bad sectors. However, in
178 failure pattern, the standards governing zoned
179 additional conditions that result in I/O error
180
181 * A zone may transition to the read-only condi
182 While the data already written in the zone i
183 no longer be written. No user action on the
184 read/write access) can change the zone condi
185 state. While the reasons for the device to t
186 state are not defined by the standards, a ty
187 would be a defective write head on an HDD (a
188 changed to read-only).
189
190 * A zone may transition to the offline conditi
191 An offline zone cannot be read nor written.
192 offline zone back to an operational good sta
193 transitions, the reasons for a drive to tran
194 condition are undefined. A typical cause wou
195 on an HDD causing all zones on the platter u
196 inaccessible.
197
198 * Unaligned write errors: These errors result
199 requests with a start sector that does not c
200 position when the write request is executed
201 enforces sequential file write for sequentia
202 may still happen in the case of a partial fa
203 operation split into multiple BIOs/requests
204 If one of the write request within the set o
205 issued to the device fails, all write reques
206 become unaligned and fail.
207
208 * Delayed write errors: similarly to regular b
209 write cache is enabled, write errors may occ
210 completed writes when the device write cache
211 Similarly to the previous immediate unaligne
212 errors can propagate through a stream of cac
213 causing all data to be dropped after the sec
214
215 All I/O errors detected by zonefs are notified
216 return for the system call that triggered or d
217 actions taken by zonefs in response to I/O err
218 vs write) and on the reason for the error (bad
219 condition change).
220
221 * For read I/O errors, zonefs does not execute
222 but only if the file zone is still in a good
223 inconsistency between the file inode size an
224 If a problem is detected, I/O error recovery
225
226 * For write I/O errors, zonefs I/O error recov
227
228 * A zone condition change to read-only or offl
229 I/O error recovery.
230
231 Zonefs minimal I/O error recovery may change a
232 permissions.
233
234 * File size changes:
235 Immediate or delayed write errors in a seque
236 inode size to be inconsistent with the amoun
237 the file zone. For instance, the partial fai
238 operation will cause the zone write pointer
239 the entire write operation will be reported
240 case, the file inode size must be advanced t
241 change and eventually allow the user to rest
242 file.
243 A file size may also be reduced to reflect a
244 fsync(): in this case, the amount of data ef
245 be less than originally indicated by the fil
246 error, zonefs always fixes the file inode si
247 persistently stored in the file zone.
248
249 * Access permission changes:
250 A zone condition change to read-only is indi
251 access permissions to render the file read-o
252 file attributes and data modification. For o
253 (read and write) to the file are disabled.
254
255 Further action taken by zonefs I/O error recov
256 with the "errors=xxx" mount option. The table
257 zonefs I/O error processing depending on the m
258 conditions::
259
260 +--------------+-----------+--------------
261 | | | Po
262 | "errors=xxx" | device |
263 | mount | zone | file
264 | option | condition | size read
265 +--------------+-----------+--------------
266 | | good | fixed yes
267 | remount-ro | read-only | as is yes
268 | (default) | offline | 0 no
269 +--------------+-----------+--------------
270 | | good | fixed yes
271 | zone-ro | read-only | as is yes
272 | | offline | 0 no
273 +--------------+-----------+--------------
274 | | good | 0 no
275 | zone-offline | read-only | 0 no
276 | | offline | 0 no
277 +--------------+-----------+--------------
278 | | good | fixed yes
279 | repair | read-only | as is yes
280 | | offline | 0 no
281 +--------------+-----------+--------------
282
283 Further notes:
284
285 * The "errors=remount-ro" mount option is the
286 error processing if no errors mount option i
287 * With the "errors=remount-ro" mount option, t
288 permissions to read-only applies to all file
289 read-only.
290 * Access permission and file size changes due
291 to the offline condition are permanent. Remo
292 with mkfs.zonefs (mkzonefs) will not change
293 state.
294 * File access permission changes to read-only
295 zones to the read-only condition are permane
296 the device will not re-enable file write acc
297 * File access permission changes implied by th
298 zone-offline mount options are temporary for
299 Unmounting and remounting the file system wi
300 (format time values) access rights to the fi
301 * The repair mount option triggers only the mi
302 actions, that is, file size fixes for zones
303 indicated as being read-only or offline by t
304 the zone file access permissions as noted in
305
306 Mount options
307 -------------
308
309 zonefs defines several mount options:
310 * errors=<behavior>
311 * explicit-open
312
313 "errors=<behavior>" option
314 ~~~~~~~~~~~~~~~~~~~~~~~~~~
315
316 The "errors=<behavior>" option mount option al
317 behavior in response to I/O errors, inode size
318 condition changes. The defined behaviors are a
319
320 * remount-ro (default)
321 * zone-ro
322 * zone-offline
323 * repair
324
325 The run-time I/O error actions defined for eac
326 previous section. Mount time I/O errors will c
327 The handling of read-only zones also differs b
328 If a read-only zone is found at mount time, th
329 same manner as offline zones, that is, all acc
330 file size set to 0. This is necessary as the w
331 is defined as invalib by the ZBC and ZAC stand
332 discover the amount of data that has been writ
333 read-only zone discovered at run-time, as indi
334 The size of the zone file is left unchanged fr
335
336 "explicit-open" option
337 ~~~~~~~~~~~~~~~~~~~~~~
338
339 A zoned block device (e.g. an NVMe Zoned Names
340 the number of zones that can be active, that i
341 implicit open, explicit open or closed conditi
342 translates into a risk for applications to see
343 limit being exceeded if the zone of a file is
344 request is issued by the user.
345
346 To avoid these potential errors, the "explicit
347 to be made active using an open zone command w
348 for the first time. If the zone open command s
349 guaranteed that write requests can be processe
350 "explicit-open" mount option will result in a
351 to the device on the last close() of a zone fi
352 empty.
353
354 Runtime sysfs attributes
355 ------------------------
356
357 zonefs defines several sysfs attributes for mo
358 are user readable and can be found in the dire
359 where <dev> is the name of the mounted zoned b
360
361 The attributes defined are as follows.
362
363 * **max_wro_seq_files**: This attribute repor
364 sequential zone files that can be open for w
365 to the maximum number of explicitly or impli
366 supports. A value of 0 means that the devic
367 (any file) can be open for writing and writt
368 state of other zones. When the *explicit-op
369 will fail any open() system call requesting
370 writing when the number of sequential zone f
371 reached the *max_wro_seq_files* limit.
372 * **nr_wro_seq_files**: This attribute report
373 zone files open for writing. When the "expl
374 this number can never exceed *max_wro_seq_fi
375 mount option is not used, the reported numbe
376 *max_wro_seq_files*. In such case, it is th
377 application to not write simultaneously more
378 sequential zone files. Failure to do so can
379 * **max_active_seq_files**: This attribute re
380 sequential zone files that are in an active
381 files that are partially written (not empty
382 is explicitly open (which happens only if th
383 used). This number is always equal to the m
384 the device supports. A value of 0 means tha
385 on the number of sequential zone files that
386 * **nr_active_seq_files**: This attributes re
387 sequential zone files that are active. If *m
388 then the value of *nr_active_seq_files* can
389 *nr_active_seq_files*, regardless of the use
390 option.
391
392 Zonefs User Space Tools
393 =======================
394
395 The mkzonefs tool is used to format zoned bloc
396 This tool is available on Github at:
397
398 https://github.com/damien-lemoal/zonefs-tools
399
400 zonefs-tools also includes a test suite which
401 block device, including null_blk block device
402
403 Examples
404 --------
405
406 The following formats a 15TB host-managed SMR
407 with the conventional zones aggregation featur
408
409 # mkzonefs -o aggr_cnv /dev/sdX
410 # mount -t zonefs /dev/sdX /mnt
411 # ls -l /mnt/
412 total 0
413 dr-xr-xr-x 2 root root 1 Nov 25 13:23
414 dr-xr-xr-x 2 root root 55356 Nov 25 13:23
415
416 The size of the zone files sub-directories ind
417 existing for each type of zones. In this examp
418 conventional zone file (all conventional zones
419 file)::
420
421 # ls -l /mnt/cnv
422 total 137101312
423 -rw-r----- 1 root root 140391743488 Nov 25
424
425 This aggregated conventional zone file can be
426
427 # mkfs.ext4 /mnt/cnv/0
428 # mount -o loop /mnt/cnv/0 /data
429
430 The "seq" sub-directory grouping files for seq
431 example 55356 zones::
432
433 # ls -lv /mnt/seq
434 total 14511243264
435 -rw-r----- 1 root root 0 Nov 25 13:23 0
436 -rw-r----- 1 root root 0 Nov 25 13:23 1
437 -rw-r----- 1 root root 0 Nov 25 13:23 2
438 ...
439 -rw-r----- 1 root root 0 Nov 25 13:23 5535
440 -rw-r----- 1 root root 0 Nov 25 13:23 5535
441
442 For sequential write zone files, the file size
443 the end of the file, similarly to any regular
444
445 # dd if=/dev/zero of=/mnt/seq/0 bs=4096 co
446 1+0 records in
447 1+0 records out
448 4096 bytes (4.1 kB, 4.0 KiB) copied, 0.000
449
450 # ls -l /mnt/seq/0
451 -rw-r----- 1 root root 4096 Nov 25 13:23 /
452
453 The written file can be truncated to the zone
454 write operation::
455
456 # truncate -s 268435456 /mnt/seq/0
457 # ls -l /mnt/seq/0
458 -rw-r----- 1 root root 268435456 Nov 25 13
459
460 Truncation to 0 size allows freeing the file z
461 append-writes to the file::
462
463 # truncate -s 0 /mnt/seq/0
464 # ls -l /mnt/seq/0
465 -rw-r----- 1 root root 0 Nov 25 13:49 /mnt
466
467 Since files are statically mapped to zones on
468 of a file as reported by stat() and fstat() in
469 zone::
470
471 # stat /mnt/seq/0
472 File: /mnt/seq/0
473 Size: 0 Blocks: 524288 IO
474 Device: 870h/2160d Inode: 50431 Lin
475 Access: (0640/-rw-r-----) Uid: ( 0/
476 Access: 2019-11-25 13:23:57.048971997 +090
477 Modify: 2019-11-25 13:52:25.553805765 +090
478 Change: 2019-11-25 13:52:25.553805765 +090
479 Birth: -
480
481 The number of blocks of the file ("Blocks") in
482 maximum file size of 524288 * 512 B = 256 MB,
483 capacity in this example. Of note is that the
484 indicates the minimum I/O size for writes and
485 physical sector size.
Linux® is a registered trademark of Linus Torvalds in the United States and other countries.
TOMOYO® is a registered trademark of NTT DATA CORPORATION.