1BTRFS-MAN5(5) Btrfs Manual BTRFS-MAN5(5)
2
6 btrfs-man5 - topics about the BTRFS filesystem (mount options,
7 supported file attributes and other)
8
10 This document describes topics related to BTRFS that are not specific
11 to the tools. Currently covers:
12 1. mount options
14 2. filesystem features
16 3. checksum algorithms
18 4. filesystem exclusive operations
20 5. filesystem limits
22 6. bootloader support
24 7. file attributes
26 8. control device
28 9. filesystems with multiple block group profiles
30
32 This section describes mount options specific to BTRFS. For the generic
33 mount options please refer to mount(8) manpage. The options are sorted
34 alphabetically (discarding the no prefix).
35 Note
37 most mount options apply to the whole filesystem and only options
38 in the first mounted subvolume will take effect. This is due to
39 lack of implementation and may change in the future. This means
40 that (for example) you can’t set per-subvolume nodatacow,
41 nodatasum, or compress using mount options. This should eventually
42 be fixed, but it has proved to be difficult to implement correctly
43 within the Linux VFS framework.
44 Mount options are processed in order, only the last occurrence of an
46 option takes effect and may disable other options due to constraints
47 (see eg. nodatacow and compress). The output of mount command shows
48 which options have been applied.
49 acl, noacl
51 (default: on)
52 Enable/disable support for Posix Access Control Lists (ACLs). See
54 the acl(5) manual page for more information about ACLs.
55 The support for ACL is build-time configurable (BTRFS_FS_POSIX_ACL)
57 and mount fails if acl is requested but the feature is not compiled
58 in.
59 autodefrag, noautodefrag
61 (since: 3.0, default: off)
62 Enable automatic file defragmentation. When enabled, small random
64 writes into files (in a range of tens of kilobytes, currently it’s
65 64K) are detected and queued up for the defragmentation process.
66 Not well suited for large database workloads.
67 The read latency may increase due to reading the adjacent blocks
69 that make up the range for defragmentation, successive write will
70 merge the blocks in the new location.
71 Warning
73 Defragmenting with Linux kernel versions < 3.9 or ≥ 3.14-rc2 as
74 well as with Linux stable kernel versions ≥ 3.10.31, ≥ 3.12.12
75 or ≥ 3.13.4 will break up the reflinks of COW data (for example
76 files copied with cp --reflink, snapshots or de-duplicated
77 data). This may cause considerable increase of space usage
78 depending on the broken up reflinks.
79 barrier, nobarrier
81 (default: on)
82 Ensure that all IO write operations make it through the device
84 cache and are stored permanently when the filesystem is at its
85 consistency checkpoint. This typically means that a flush command
86 is sent to the device that will synchronize all pending data and
87 ordinary metadata blocks, then writes the superblock and issues
88 another flush.
89 The write flushes incur a slight hit and also prevent the IO block
91 scheduler to reorder requests in a more effective way. Disabling
92 barriers gets rid of that penalty but will most certainly lead to a
93 corrupted filesystem in case of a crash or power loss. The ordinary
94 metadata blocks could be yet unwritten at the time the new
95 superblock is stored permanently, expecting that the block pointers
96 to metadata were stored permanently before.
97 On a device with a volatile battery-backed write-back cache, the
99 nobarrier option will not lead to filesystem corruption as the
100 pending blocks are supposed to make it to the permanent storage.
101 check_int, check_int_data, check_int_print_mask=value
103 (since: 3.0, default: off)
104 These debugging options control the behavior of the integrity
106 checking module (the BTRFS_FS_CHECK_INTEGRITY config option
107 required). The main goal is to verify that all blocks from a given
108 transaction period are properly linked.
109 check_int enables the integrity checker module, which examines all
111 block write requests to ensure on-disk consistency, at a large
112 memory and CPU cost.
113 check_int_data includes extent data in the integrity checks, and
115 implies the check_int option.
116 check_int_print_mask takes a bitmask of BTRFSIC_PRINT_MASK_* values
118 as defined in fs/btrfs/check-integrity.c, to control the integrity
119 checker module behavior.
120 See comments at the top of fs/btrfs/check-integrity.c for more
122 information.
123 clear_cache
125 Force clearing and rebuilding of the disk space cache if something
126 has gone wrong. See also: space_cache.
127 commit=seconds
129 (since: 3.12, default: 30)
130 Set the interval of periodic transaction commit when data are
132 synchronized to permanent storage. Higher interval values lead to
133 larger amount of unwritten data, which has obvious consequences
134 when the system crashes. The upper bound is not forced, but a
135 warning is printed if it’s more than 300 seconds (5 minutes). Use
136 with care.
137 compress, compress=type[:level], compress-force,
139 compress-force=type[:level]
140 (default: off, level support since: 5.1)
141 Control BTRFS file data compression. Type may be specified as zlib,
143 lzo, zstd or no (for no compression, used for remounting). If no
144 type is specified, zlib is used. If compress-force is specified,
145 then compression will always be attempted, but the data may end up
146 uncompressed if the compression would make them larger.
147 Both zlib and zstd (since version 5.1) expose the compression level
149 as a tunable knob with higher levels trading speed and memory
150 (zstd) for higher compression ratios. This can be set by appending
151 a colon and the desired level. Zlib accepts the range [1, 9] and
152 zstd accepts [1, 15]. If no level is set, both currently use a
153 default level of 3. The value 0 is an alias for the default level.
154 Otherwise some simple heuristics are applied to detect an
156 incompressible file. If the first blocks written to a file are not
157 compressible, the whole file is permanently marked to skip
158 compression. As this is too simple, the compress-force is a
159 workaround that will compress most of the files at the cost of some
160 wasted CPU cycles on failed attempts. Since kernel 4.15, a set of
161 heuristic algorithms have been improved by using frequency
162 sampling, repeated pattern detection and Shannon entropy
163 calculation to avoid that.
164 Note
166 If compression is enabled, nodatacow and nodatasum are
167 disabled.
168 datacow, nodatacow
170 (default: on)
171 Enable data copy-on-write for newly created files. Nodatacow
173 implies nodatasum, and disables compression. All files created
174 under nodatacow are also set the NOCOW file attribute (see
175 chattr(1)).
176 Note
178 If nodatacow or nodatasum are enabled, compression is disabled.
179 Updates in-place improve performance for workloads that do frequent
180 overwrites, at the cost of potential partial writes, in case the
181 write is interrupted (system crash, device failure).
182 datasum, nodatasum
184 (default: on)
185 Enable data checksumming for newly created files. Datasum implies
187 datacow, ie. the normal mode of operation. All files created under
188 nodatasum inherit the "no checksums" property, however there’s no
189 corresponding file attribute (see chattr(1)).
190 Note
192 If nodatacow or nodatasum are enabled, compression is disabled.
193 There is a slight performance gain when checksums are turned off,
194 the corresponding metadata blocks holding the checksums do not need
195 to updated. The cost of checksumming of the blocks in memory is
196 much lower than the IO, modern CPUs feature hardware support of the
197 checksumming algorithm.
198 degraded
200 (default: off)
201 Allow mounts with less devices than the RAID profile constraints
203 require. A read-write mount (or remount) may fail when there are
204 too many devices missing, for example if a stripe member is
205 completely missing from RAID0.
206 Since 4.14, the constraint checks have been improved and are
208 verified on the chunk level, not an the device level. This allows
209 degraded mounts of filesystems with mixed RAID profiles for data
210 and metadata, even if the device number constraints would not be
211 satisfied for some of the profiles.
212 Example: metadata — raid1, data — single, devices — /dev/sda,
214 /dev/sdb
215 Suppose the data are completely stored on sda, then missing sdb
217 will not prevent the mount, even if 1 missing device would normally
218 prevent (any) single profile to mount. In case some of the data
219 chunks are stored on sdb, then the constraint of single/data is not
220 satisfied and the filesystem cannot be mounted.
221 device=devicepath
223 Specify a path to a device that will be scanned for BTRFS
224 filesystem during mount. This is usually done automatically by a
225 device manager (like udev) or using the btrfs device scan command
226 (eg. run from the initial ramdisk). In cases where this is not
227 possible the device mount option can help.
228 Note
230 booting eg. a RAID1 system may fail even if all filesystem’s
231 device paths are provided as the actual device nodes may not be
232 discovered by the system at that point.
233 discard, discard=sync, discard=async, nodiscard
235 (default: off, async support since: 5.6)
236 Enable discarding of freed file blocks. This is useful for SSD
238 devices, thinly provisioned LUNs, or virtual machine images;
239 however, every storage layer must support discard for it to work.
240 In the synchronous mode (sync or without option value), lack of
242 asynchronous queued TRIM on the backing device TRIM can severely
243 degrade performance, because a synchronous TRIM operation will be
244 attempted instead. Queued TRIM requires newer than SATA revision
245 3.1 chipsets and devices.
246 The asynchronous mode (async) gathers extents in larger chunks
248 before sending them to the devices for TRIM. The overhead and
249 performance impact should be negligible compared to the previous
250 mode and it’s supposed to be the preferred mode if needed.
251 If it is not necessary to immediately discard freed blocks, then
253 the fstrim tool can be used to discard all free blocks in a batch.
254 Scheduling a TRIM during a period of low system activity will
255 prevent latent interference with the performance of other
256 operations. Also, a device may ignore the TRIM command if the range
257 is too small, so running a batch discard has a greater probability
258 of actually discarding the blocks.
259 enospc_debug, noenospc_debug
261 (default: off)
262 Enable verbose output for some ENOSPC conditions. It’s safe to use
264 but can be noisy if the system reaches near-full state.
265 fatal_errors=action
267 (since: 3.4, default: bug)
268 Action to take when encountering a fatal error.
270 bug
272 BUG() on a fatal error, the system will stay in the crashed
273 state and may be still partially usable, but reboot is required
274 for full operation
275 panic
277 panic() on a fatal error, depending on other system
278 configuration, this may be followed by a reboot. Please refer
279 to the documentation of kernel boot parameters, eg. panic,
280 oops or crashkernel.
281 flushoncommit, noflushoncommit
283 (default: off)
284 This option forces any data dirtied by a write in a prior
286 transaction to commit as part of the current commit, effectively a
287 full filesystem sync.
288 This makes the committed state a fully consistent view of the file
290 system from the application’s perspective (i.e. it includes all
291 completed file system operations). This was previously the behavior
292 only when a snapshot was created.
293 When off, the filesystem is consistent but buffered writes may last
295 more than one transaction commit.
296 fragment=type
298 (depends on compile-time option BTRFS_DEBUG, since: 4.4, default:
299 off)
300 A debugging helper to intentionally fragment given type of block
302 groups. The type can be data, metadata or all. This mount option
303 should not be used outside of debugging environments and is not
304 recognized if the kernel config option BTRFS_DEBUG is not enabled.
305 inode_cache, noinode_cache
307 (since: 3.0, default: off)
308 Enable free inode number caching. Not recommended to use unless
310 files on your filesystem get assigned inode numbers that are
311 approaching 2^64. Normally, new files in each subvolume get
312 assigned incrementally (plus one from the last time) and are not
313 reused. The mount option turns on caching of the existing inode
314 numbers and reuse of inode numbers of deleted files.
315 This option may slow down your system at first run, or after
317 mounting without the option.
318 Note
320 Defaults to off due to a potential overflow problem when the
321 free space checksums don’t fit inside a single page.
322 Don’t use this option unless you really need it. The inode number
323 limit on 64bit system is 2^64, which is practically enough for the
324 whole filesystem lifetime. Due to implementation of linux VFS
325 layer, the inode numbers on 32bit systems are only 32 bits wide.
326 This lowers the limit significantly and makes it possible to reach
327 it. In such case, this mount option will help. Alternatively, files
328 with high inode numbers can be copied to a new subvolume which will
329 effectively start the inode numbers from the beginning again.
330 nologreplay
332 (default: off, even read-only)
333 The tree-log contains pending updates to the filesystem until the
335 full commit. The log is replayed on next mount, this can be
336 disabled by this option. See also treelog. Note that nologreplay is
337 the same as norecovery.
338 Warning
340 currently, the tree log is replayed even with a read-only
341 mount! To disable that behaviour, mount also with nologreplay.
342 max_inline=bytes
344 (default: min(2048, page size) )
345 Specify the maximum amount of space, that can be inlined in a
347 metadata B-tree leaf. The value is specified in bytes, optionally
348 with a K suffix (case insensitive). In practice, this value is
349 limited by the filesystem block size (named sectorsize at mkfs
350 time), and memory page size of the system. In case of sectorsize
351 limit, there’s some space unavailable due to leaf headers. For
352 example, a 4k sectorsize, maximum size of inline data is about 3900
353 bytes.
354 Inlining can be completely turned off by specifying 0. This will
356 increase data block slack if file sizes are much smaller than block
357 size but will reduce metadata consumption in return.
358 Note
360 the default value has changed to 2048 in kernel 4.6.
361 metadata_ratio=value
363 (default: 0, internal logic)
364 Specifies that 1 metadata chunk should be allocated after every
366 value data chunks. Default behaviour depends on internal logic,
367 some percent of unused metadata space is attempted to be maintained
368 but is not always possible if there’s not enough space left for
369 chunk allocation. The option could be useful to override the
370 internal logic in favor of the metadata allocation if the expected
371 workload is supposed to be metadata intense (snapshots, reflinks,
372 xattrs, inlined files).
373 norecovery
375 (since: 4.5, default: off)
376 Do not attempt any data recovery at mount time. This will disable
378 logreplay and avoids other write operations. Note that this option
379 is the same as nologreplay.
380 Note
382 The opposite option recovery used to have different meaning but
383 was changed for consistency with other filesystems, where
384 norecovery is used for skipping log replay. BTRFS does the same
385 and in general will try to avoid any write operations.
386 rescan_uuid_tree
388 (since: 3.12, default: off)
389 Force check and rebuild procedure of the UUID tree. This should not
391 normally be needed.
392 skip_balance
394 (since: 3.3, default: off)
395 Skip automatic resume of an interrupted balance operation. The
397 operation can later be resumed with btrfs balance resume, or the
398 paused state can be removed with btrfs balance cancel. The default
399 behaviour is to resume an interrupted balance immediately after a
400 volume is mounted.
401 space_cache, space_cache=version, nospace_cache
403 (nospace_cache since: 3.2, space_cache=v1 and space_cache=v2 since
404 4.5, default: space_cache=v1)
405 Options to control the free space cache. The free space cache
407 greatly improves performance when reading block group free space
408 into memory. However, managing the space cache consumes some
409 resources, including a small amount of disk space.
410 There are two implementations of the free space cache. The original
412 one, referred to as v1, is the safe default. The v1 space cache can
413 be disabled at mount time with nospace_cache without clearing.
414 On very large filesystems (many terabytes) and certain workloads,
416 the performance of the v1 space cache may degrade drastically. The
417 v2 implementation, which adds a new B-tree called the free space
418 tree, addresses this issue. Once enabled, the v2 space cache will
419 always be used and cannot be disabled unless it is cleared. Use
420 clear_cache,space_cache=v1 or clear_cache,nospace_cache to do so.
421 If v2 is enabled, kernels without v2 support will only be able to
422 mount the filesystem in read-only mode. The btrfs(8) command
423 currently only has read-only support for v2. A read-write command
424 may be run on a v2 filesystem by clearing the cache, running the
425 command, and then remounting with space_cache=v2.
426 If a version is not explicitly specified, the default
428 implementation will be chosen, which is v1.
429 ssd, ssd_spread, nossd, nossd_spread
431 (default: SSD autodetected)
432 Options to control SSD allocation schemes. By default, BTRFS will
434 enable or disable SSD optimizations depending on status of a device
435 with respect to rotational or non-rotational type. This is
436 determined by the contents of /sys/block/DEV/queue/rotational). If
437 it is 0, the ssd option is turned on. The option nossd will disable
438 the autodetection.
439 The optimizations make use of the absence of the seek penalty
441 that’s inherent for the rotational devices. The blocks can be
442 typically written faster and are not offloaded to separate threads.
443 Note
445 Since 4.14, the block layout optimizations have been dropped.
446 This used to help with first generations of SSD devices. Their
447 FTL (flash translation layer) was not effective and the
448 optimization was supposed to improve the wear by better
449 aligning blocks. This is no longer true with modern SSD devices
450 and the optimization had no real benefit. Furthermore it caused
451 increased fragmentation. The layout tuning has been kept intact
452 for the option ssd_spread.
453 The ssd_spread mount option attempts to allocate into bigger and
454 aligned chunks of unused space, and may perform better on low-end
455 SSDs. ssd_spread implies ssd, enabling all other SSD heuristics as
456 well. The option nossd will disable all SSD options while
457 nossd_spread only disables ssd_spread.
458 subvol=path
460 Mount subvolume from path rather than the toplevel subvolume. The
461 path is always treated as relative to the toplevel subvolume. This
462 mount option overrides the default subvolume set for the given
463 filesystem.
464 subvolid=subvolid
466 Mount subvolume specified by a subvolid number rather than the
467 toplevel subvolume. You can use btrfs subvolume list of btrfs
468 subvolume show to see subvolume ID numbers. This mount option
469 overrides the default subvolume set for the given filesystem.
470 Note
472 if both subvolid and subvol are specified, they must point at
473 the same subvolume, otherwise the mount will fail.
474 thread_pool=number
476 (default: min(NRCPUS + 2, 8) )
477 The number of worker threads to start. NRCPUS is number of on-line
479 CPUs detected at the time of mount. Small number leads to less
480 parallelism in processing data and metadata, higher numbers could
481 lead to a performance hit due to increased locking contention,
482 process scheduling, cache-line bouncing or costly data transfers
483 between local CPU memories.
484 treelog, notreelog
486 (default: on)
487 Enable the tree logging used for fsync and O_SYNC writes. The tree
489 log stores changes without the need of a full filesystem sync. The
490 log operations are flushed at sync and transaction commit. If the
491 system crashes between two such syncs, the pending tree log
492 operations are replayed during mount.
493 Warning
495 currently, the tree log is replayed even with a read-only
496 mount! To disable that behaviour, also mount with nologreplay.
497 The tree log could contain new files/directories, these would not
498 exist on a mounted filesystem if the log is not replayed.
499 usebackuproot
501 (since: 4.6, default: off)
502 Enable autorecovery attempts if a bad tree root is found at mount
504 time. Currently this scans a backup list of several previous tree
505 roots and tries to use the first readable. This can be used with
506 read-only mounts as well.
507 Note
509 This option has replaced recovery.
510 user_subvol_rm_allowed
512 (default: off)
513 Allow subvolumes to be deleted by their respective owner.
515 Otherwise, only the root user can do that.
516 Note
518 historically, any user could create a snapshot even if he was
519 not owner of the source subvolume, the subvolume deletion has
520 been restricted for that reason. The subvolume creation has
521 been restricted but this mount option is still required. This
522 is a usability issue. Since 4.18, the rmdir(2) syscall can
523 delete an empty subvolume just like an ordinary directory.
524 Whether this is possible can be detected at runtime, see
525 rmdir_subvol feature in FILESYSTEM FEATURES.
526 DEPRECATED MOUNT OPTIONS
528 List of mount options that have been removed, kept for backward
529 compatibility.
530 recovery
532 (since: 3.2, default: off, deprecated since: 4.5)
533 Note
535 this option has been replaced by usebackuproot and should not
536 be used but will work on 4.5+ kernels.
537 NOTES ON GENERIC MOUNT OPTIONS
539 Some of the general mount options from mount(8) that affect BTRFS and
540 are worth mentioning.
541 noatime
543 under read intensive work-loads, specifying noatime significantly
544 improves performance because no new access time information needs
545 to be written. Without this option, the default is relatime, which
546 only reduces the number of inode atime updates in comparison to the
547 traditional strictatime. The worst case for atime updates under
548 relatime occurs when many files are read whose atime is older than
549 24 h and which are freshly snapshotted. In that case the atime is
550 updated and COW happens - for each file - in bulk. See also
551 https://lwn.net/Articles/499293/ - Atime and btrfs: a bad
552 combination? (LWN, 2012-05-31).
553 Note that noatime may break applications that rely on atime uptimes
555 like the venerable Mutt (unless you use maildir mailboxes).
556
558 The basic set of filesystem features gets extended over time. The
559 backward compatibility is maintained and the features are optional,
560 need to be explicitly asked for so accidental use will not create
561 incompatibilities.
562 There are several classes and the respective tools to manage the
564 features:
565 at mkfs time only
567 This is namely for core structures, like the b-tree nodesize or
568 checksum algorithm, see mkfs.btrfs(8) for more details.
569 after mkfs, on an unmounted filesystem
571 Features that may optimize internal structures or add new
572 structures to support new functionality, see btrfstune(8). The
573 command btrfs inspect-internal dump-super device will dump a
574 superblock, you can map the value of incompat_flags to the features
575 listed below
576 after mkfs, on a mounted filesystem
578 The features of a filesystem (with a given UUID) are listed in
579 /sys/fs/btrfs/UUID/features/, one file per feature. The status is
580 stored inside the file. The value 1 is for enabled and active,
581 while 0 means the feature was enabled at mount time but turned off
582 afterwards.
583 Whether a particular feature can be turned on a mounted filesystem
585 can be found in the directory /sys/fs/btrfs/features/, one file per
586 feature. The value 1 means the feature can be enabled.
587 List of features (see also mkfs.btrfs(8) section FILESYSTEM FEATURES):
589 big_metadata
591 (since: 3.4)
592 the filesystem uses nodesize for metadata blocks, this can be
594 bigger than the page size
595 compress_lzo
597 (since: 2.6.38)
598 the lzo compression has been used on the filesystem, either as a
600 mount option or via btrfs filesystem defrag.
601 compress_zstd
603 (since: 4.14)
604 the zstd compression has been used on the filesystem, either as a
606 mount option or via btrfs filesystem defrag.
607 default_subvol
609 (since: 2.6.34)
610 the default subvolume has been set on the filesystem
612 extended_iref
614 (since: 3.7)
615 increased hardlink limit per file in a directory to 65536, older
617 kernels supported a varying number of hardlinks depending on the
618 sum of all file name sizes that can be stored into one metadata
619 block
620 free_space_tree
622 (since: 4.5)
623 free space representation using a dedicated b-tree, successor of v1
625 space cache
626 metadata_uuid
628 (since: 5.0)
629 the main filesystem UUID is the metadata_uuid, which stores the new
631 UUID only in the superblock while all metadata blocks still have
632 the UUID set at mkfs time, see btrfstune(8) for more
633 mixed_backref
635 (since: 2.6.31)
636 the last major disk format change, improved backreferences, now
638 default
639 mixed_groups
641 (since: 2.6.37)
642 mixed data and metadata block groups, ie. the data and metadata are
644 not separated and occupy the same block groups, this mode is
645 suitable for small volumes as there are no constraints how the
646 remaining space should be used (compared to the split mode, where
647 empty metadata space cannot be used for data and vice versa)
648 on the other hand, the final layout is quite unpredictable and
650 possibly highly fragmented, which means worse performance
651 no_holes
653 (since: 3.14)
654 improved representation of file extents where holes are not
656 explicitly stored as an extent, saves a few percent of metadata if
657 sparse files are used
658 raid1c34
660 (since: 5.5)
661 extended RAID1 mode with copies on 3 or 4 devices respectively
663 raid56
665 (since: 3.9)
666 the filesystem contains or contained a raid56 profile of block
668 groups
669 rmdir_subvol
671 (since: 4.18)
672 indicate that rmdir(2) syscall can delete an empty subvolume just
674 like an ordinary directory. Note that this feature only depends on
675 the kernel version.
676 skinny_metadata
678 (since: 3.10)
679 reduced-size metadata for extent references, saves a few percent of
681 metadata
682 supported_checksums
684 (since: 5.5)
685 list of checksum algorithms supported by the kernel module, the
687 respective modules or built-in implementing the algorithms need to
688 be present to mount the filesystem
689 SWAPFILE SUPPORT
691 The swapfile is supported since kernel 5.0. Use swapon(8) to activate
692 the swapfile. There are some limitations of the implementation in btrfs
693 and linux swap subsystem:
694 · filesystem - must be only single device
696 · filesystem - must have only single data profile
698 · swapfile - the containing subvolume cannot be snapshotted
700 · swapfile - must be preallocated
702 · swapfile - must be nodatacow (ie. also nodatasum)
704 · swapfile - must not be compressed
706 The limitations come namely from the COW-based design and mapping layer
708 of blocks that allows the advanced features like relocation and
709 multi-device filesystems. However, the swap subsystem expects simpler
710 mapping and no background changes of the file blocks once they’ve been
711 attached to swap.
712 With active swapfiles, the following whole-filesystem operations will
714 skip swapfile extents or may fail:
715 · balance - block groups with swapfile extents are skipped and
717 reported, the rest will be processed normally
718 · resize grow - unaffected
720 · resize shrink - works as long as the extents are outside of the
722 shrunk range
723 · device add - a new device does not interfere with existing swapfile
725 and this operation will work, though no new swapfile can be
726 activated afterwards
727 · device delete - if the device has been added as above, it can be
729 also deleted
730 · device replace - ditto
732 When there are no active swapfiles and a whole-filesystem exclusive
734 operation is running (ie. balance, device delete, shrink), the
735 swapfiles cannot be temporarily activated. The operation must finish
736 first.
737 # truncate -s 0 swapfile
739 # chattr +C swapfile
740 # fallocate -l 2G swapfile
741 # chmod 0600 swapfile
742 # mkswap swapfile
743 # swapon swapfile
744
746 There are several checksum algorithms supported. The default and
747 backward compatible is crc32c. Since kernel 5.5 there are three more
748 with different characteristics and trade-offs regarding speed and
749 strength. The following list may help you to decide which one to
750 select.
751 CRC32C (32bit digest)
753 default, best backward compatibility, very fast, modern CPUs have
754 instruction-level support, not collision-resistant but still good
755 error detection capabilities
756 XXHASH (64bit digest)
758 can be used as CRC32C successor, very fast, optimized for modern
759 CPUs utilizing instruction pipelining, good collision resistance
760 and error detection
761 SHA256 (256bit digest)
763 a cryptographic-strength hash, relatively slow but with possible
764 CPU instruction acceleration or specialized hardware cards, FIPS
765 certified and in wide use
766 BLAKE2b (256bit digest)
768 a cryptographic-strength hash, relatively fast with possible CPU
769 acceleration using SIMD extensions, not standardized but based on
770 BLAKE which was a SHA3 finalist, in wide use, the algorithm used is
771 BLAKE2b-256 that’s optimized for 64bit platforms
772 The digest size affects overall size of data block checksums stored in
774 the filesystem. The metadata blocks have a fixed area up to 256bits (32
775 bytes), so there’s no increase. Each data block has a separate checksum
776 stored, with additional overhead of the b-tree leaves.
777 Approximate relative performance of the algorithms, measured against
779 CRC32C using reference software implementations on a 3.5GHz intel CPU:
780 ┌────────┬─────────────┬───────┐
782 │ │ │ │
783 │Digest │ Cycles/4KiB │ Ratio │
784 ├────────┼─────────────┼───────┤
785 │ │ │ │
786 │CRC32C │ 1700 │ 1.00 │
787 ├────────┼─────────────┼───────┤
788 │ │ │ │
789 │XXHASH │ 2500 │ 1.44 │
790 ├────────┼─────────────┼───────┤
791 │ │ │ │
792 │SHA256 │ 105000 │ 61 │
793 ├────────┼─────────────┼───────┤
794 │ │ │ │
795 │BLAKE2b │ 22000 │ 13 │
796 └────────┴─────────────┴───────┘
797
799 There are several operations that affect the whole filesystem and
800 cannot be run in parallel. Attempt to start one while another is
801 running will fail.
802 Since kernel 5.10 the currently running operation can be obtained from
804 /sys/fs/UUID/exclusive_operation with following values and operations:
805 · balance
807 · device add
809 · device delete
811 · device replace
813 · resize
815 · swapfile activate
817 · none
819 Enqueuing is supported for several btrfs subcommands so they can be
821 started at once and then serialized.
822
824 maximum file name length
825 255
826 maximum symlink target length
828 depends on the nodesize value, for 4k it’s 3949 bytes, for larger
829 nodesize it’s 4095 due to the system limit PATH_MAX
830 The symlink target may not be a valid path, ie. the path name
832 components can exceed the limits (NAME_MAX), there’s no content
833 validation at symlink(3) creation.
834 maximum number of inodes
836 2^64 but depends on the available metadata space as the inodes are
837 created dynamically
838 inode numbers
840 minimum number: 256 (for subvolumes), regular files and
841 directories: 257
842 maximum file length
844 inherent limit of btrfs is 2^64 (16 EiB) but the linux VFS limit is
845 2^63 (8 EiB)
846 maximum number of subvolumes
848 the subvolume ids can go up to 2^64 but the number of actual
849 subvolumes depends on the available metadata space, the space
850 consumed by all subvolume metadata includes bookkeeping of shared
851 extents can be large (MiB, GiB)
852 maximum number of hardlinks of a file in a directory
854 65536 when the extref feature is turned on during mkfs (default),
855 roughly 100 otherwise
856 minimum filesystem size
858 the minimal size of each device depends on the mixed-bg feature,
859 without that (the default) it’s about 109MiB, with mixed-bg it’s is
860 16MiB
861
863 GRUB2 (https://www.gnu.org/software/grub) has the most advanced support
864 of booting from BTRFS with respect to features.
865 U-boot (https://www.denx.de/wiki/U-Boot/) has decent support for
867 booting but not all BTRFS features are implemented, check the
868 documentation.
869 EXTLINUX (from the https://syslinux.org project) can boot but does not
871 support all features. Please check the upstream documentation before
872 you use it.
873 The first 1MiB on each device is unused with the exception of primary
875 superblock that is on the offset 64KiB and spans 4KiB.
876
878 The btrfs filesystem supports setting file attributes or flags. Note
879 there are old and new interfaces, with confusing names. The following
880 list should clarify that:
881 · attributes: chattr(1) or lsattr(1) utilities (the ioctls are
883 FS_IOC_GETFLAGS and FS_IOC_SETFLAGS), due to the ioctl names the
884 attributes are also called flags
885 · xflags: to distinguish from the previous, it’s extended flags, with
887 tunable bits similar to the attributes but extensible and new bits
888 will be added in the future (the ioctls are FS_IOC_FSGETXATTR and
889 FS_IOC_FSSETXATTR but they are not related to extended attributes
890 that are also called xattrs), there’s no standard tool to change
891 the bits, there’s support in xfs_io(8) as command xfs_io -c chattr
892 ATTRIBUTES
894 a
895 append only, new writes are always written at the end of the file
896 A
898 no atime updates
899 c
901 compress data, all data written after this attribute is set will be
902 compressed. Please note that compression is also affected by the
903 mount options or the parent directory attributes.
904 When set on a directory, all newly created files will inherit this
906 attribute.
907 C
909 no copy-on-write, file data modifications are done in-place
910 When set on a directory, all newly created files will inherit this
912 attribute.
913 Note
915 due to implementation limitations, this flag can be set/unset
916 only on empty files.
917 d
919 no dump, makes sense with 3rd party tools like dump(8), on BTRFS
920 the attribute can be set/unset but no other special handling is
921 done
922 D
924 synchronous directory updates, for more details search open(2) for
925 O_SYNC and O_DSYNC
926 i
928 immutable, no file data and metadata changes allowed even to the
929 root user as long as this attribute is set (obviously the exception
930 is unsetting the attribute)
931 S
933 synchronous updates, for more details search open(2) for O_SYNC and
934 O_DSYNC
935 X
937 no compression, permanently turn off compression on the given file.
938 Any compression mount options will not affect this file.
939 When set on a directory, all newly created files will inherit this
941 attribute.
942 No other attributes are supported. For the complete list please refer
944 to the chattr(1) manual page.
945 XFLAGS
947 There’s overlap of letters assigned to the bits with the attributes,
948 this list refers to what xfs_io(8) provides:
949 i
951 immutable, same as the attribute
952 a
954 append only, same as the attribute
955 s
957 synchronous updates, same as the attribute S
958 A
960 no atime updates, same as the attribute
961 d
963 no dump, same as the attribute
964
966 There’s a character special device /dev/btrfs-control with major and
967 minor numbers 10 and 234 (the device can be found under the misc
968 category).
969 $ ls -l /dev/btrfs-control
971 crw------- 1 root root 10, 234 Jan 1 12:00 /dev/btrfs-control
972 The device accepts some ioctl calls that can perform following actions
974 on the filesystem module:
975 · scan devices for btrfs filesystem (ie. to let multi-device
977 filesystems mount automatically) and register them with the kernel
978 module
979 · similar to scan, but also wait until the device scanning process is
981 finished for a given filesystem
982 · get the supported features (can be also found under
984 /sys/fs/btrfs/features)
985 The device is usually created by a system device node manager (eg.
987 udev), but can be created manually:
988 # mknod --mode=600 /dev/btrfs-control c 10 234
990 The control device is not strictly required but the device scanning
992 will not work and a workaround would need to be used to mount a
993 multi-device filesystem. The mount option device can trigger the device
994 scanning during mount.
995
997 It is possible that a btrfs filesystem contains multiple block group
998 profiles of the same type. This could happen when a profile conversion
999 using balance filters is interrupted (see btrfs-balance(8)). Some btrfs
1000 commands perform a test to detect this kind of condition and print a
1001 warning like this:
1002 WARNING: Multiple block group profiles detected, see 'man btrfs(5)'.
1004 WARNING: Data: single, raid1
1005 WARNING: Metadata: single, raid1
1006 The corresponding output of btrfs filesystem df might look like:
1008 WARNING: Multiple block group profiles detected, see 'man btrfs(5)'.
1010 WARNING: Data: single, raid1
1011 WARNING: Metadata: single, raid1
1012 Data, RAID1: total=832.00MiB, used=0.00B
1013 Data, single: total=1.63GiB, used=0.00B
1014 System, single: total=4.00MiB, used=16.00KiB
1015 Metadata, single: total=8.00MiB, used=112.00KiB
1016 Metadata, RAID1: total=64.00MiB, used=32.00KiB
1017 GlobalReserve, single: total=16.25MiB, used=0.00B
1018 There’s more than one line for type Data and Metadata, while the
1020 profiles are single and RAID1.
1021 This state of the filesystem OK but most likely needs the
1023 user/administrator to take an action and finish the interrupted tasks.
1024 This cannot be easily done automatically, also the user knows the
1025 expected final profiles.
1026 In the example above, the filesystem started as a single device and
1028 single block group profile. Then another device was added, followed by
1029 balance with convert=raid1 but for some reason hasn’t finished.
1030 Restarting the balance with convert=raid1 will continue and end up with
1031 filesystem with all block group profiles RAID1.
1032 Note
1034 If you’re familiar with balance filters, you can use
1035 convert=raid1,profiles=single,soft, which will take only the
1036 unconverted single profiles and convert them to raid1. This may
1037 speed up the conversion as it would not try to rewrite the already
1038 convert raid1 profiles.
1039 Having just one profile is desired as this also clearly defines the
1041 profile of newly allocated block groups, otherwise this depends on
1042 internal allocation policy. When there are multiple profiles present,
1043 the order of selection is RAID6, RAID5, RAID10, RAID1, RAID0 as long as
1044 the device number constraints are satisfied.
1045 Commands that print the warning were chosen so they’re brought to user
1047 attention when the filesystem state is being changed in that regard.
1048 This is: device add, device delete, balance cancel, balance pause.
1049 Commands that report space usage: filesystem df, device usage. The
1050 command filesystem usage provides a line in the overall summary:
1051 Multiple profiles: yes (data, metadata)
1053
1055 acl(5), btrfs(8), chattr(1), fstrim(8), ioctl(2), mkfs.btrfs(8),
1056 mount(8), swapon(8)
1057Btrfs v5.10 01/19/2021 BTRFS-MAN5(5)