XFS is a high performance journaling filesystem which originatedon the SGI IRIX platform. It is completely multi-threaded, cansupport large files and large filesystems, extended attributes,variable block sizes, is extent based, and makes extensive use ofBtrees (directories, extents, free space) to aid both performanceand scalability.
Refer to the documentation at https://xfs.wiki.kernel.org/for further details. This implementation is on-disk compatiblewith the IRIX version of XFS.
Mount Options¶
When mounting an XFS filesystem, the following options are accepted.
- allocsize=size
Sets the buffered I/O end-of-file preallocation size whendoing delayed allocation writeout (default size is 64KiB).Valid values for this option are page size (typically 4KiB)through to 1GiB, inclusive, in power-of-2 increments.
The default behaviour is for dynamic end-of-filepreallocation size, which uses a set of heuristics tooptimise the preallocation size based on the currentallocation patterns within the file and the access patternsto the file. Specifying a fixed
allocsize
value turns offthe dynamic behaviour.- attr2 or noattr2
The options enable/disable an “opportunistic” improvement tobe made in the way inline extended attributes are storedon-disk. When the new form is used for the first time when
attr2
is selected (either when setting or removing extendedattributes) the on-disk superblock feature bit field will beupdated to reflect this format being in use.The default behaviour is determined by the on-disk featurebit indicating that
attr2
behaviour is active. If eithermount option is set, then that becomes the new default usedby the filesystem.CRC enabled filesystems always use the
attr2
format, and sowill reject thenoattr2
mount option if it is set.- discard or nodiscard (default)
Enable/disable the issuing of commands to let the blockdevice reclaim space freed by the filesystem. This isuseful for SSD devices, thinly provisioned LUNs and virtualmachine images, but may have a performance impact.
Note: It is currently recommended that you use the
fstrim
application todiscard
unused blocks rather than thediscard
mount option because the performance impact of this optionis quite severe.- grpid/bsdgroups or nogrpid/sysvgroups (default)
These options define what group ID a newly created filegets. When
grpid
is set, it takes the group ID of thedirectory in which it is created; otherwise it takes thefsgid
of the current process, unless the directory has thesetgid
bit set, in which case it takes thegid
from theparent directory, and also gets thesetgid
bit set if it isa directory itself.- filestreams
Make the data allocator use the filestreams allocation modeacross the entire filesystem rather than just on directoriesconfigured to use it.
- ikeep or noikeep (default)
When
ikeep
is specified, XFS does not delete empty inodeclusters and keeps them around on disk. Whennoikeep
isspecified, empty inode clusters are returned to the freespace pool.- inode32 or inode64 (default)
When
inode32
is specified, it indicates that XFS limitsinode creation to locations which will not result in inodenumbers with more than 32 bits of significance.When
inode64
is specified, it indicates that XFS is allowedto create inodes at any location in the filesystem,including those which will result in inode numbers occupyingmore than 32 bits of significance.
inode32
is provided for backwards compatibility with oldersystems and applications, since 64 bits inode numbers mightcause problems for some applications that cannot handlelarge inode numbers. If applications are in use which donot handle inode numbers bigger than 32 bits, theinode32
option should be specified.- largeio or nolargeio (default)
If
nolargeio
is specified, the optimal I/O reported inst_blksize
by stat(2) will be as small as possible to allowuser applications to avoid inefficient read/modify/writeI/O. This is typically the page size of the machine, asthis is the granularity of the page cache.If
largeio
is specified, a filesystem that was created with aswidth
specified will return theswidth
value (in bytes)inst_blksize
. If the filesystem does not have aswidth
specified but does specify anallocsize
thenallocsize
(in bytes) will be returned instead. Otherwise the behaviouris the same as ifnolargeio
was specified.- logbufs=value
Set the number of in-memory log buffers. Valid numbersrange from 2-8 inclusive.
The default value is 8 buffers.
If the memory cost of 8 log buffers is too high on smallsystems, then it may be reduced at some cost to performanceon metadata intensive workloads. The
logbsize
option belowcontrols the size of each buffer and so is also relevant tothis case.- logbsize=value
Set the size of each in-memory log buffer. The size may bespecified in bytes, or in kilobytes with a “k” suffix.Valid sizes for version 1 and version 2 logs are 16384 (16k)and 32768 (32k). Valid sizes for version 2 logs alsoinclude 65536 (64k), 131072 (128k) and 262144 (256k). Thelogbsize must be an integer multiple of the logstripe unit configured at mkfs(8) time.
The default value for version 1 logs is 32768, while thedefault value for version 2 logs is MAX(32768, log_sunit).
- logdev=device and rtdev=device
Use an external log (metadata journal) and/or real-time device.An XFS filesystem has up to three parts: a data section, a logsection, and a real-time section. The real-time section isoptional, and the log section can be separate from the datasection or contained within it.
- noalign
Data allocations will not be aligned at stripe unitboundaries. This is only relevant to filesystems createdwith non-zero data alignment parameters (
sunit
,swidth
) bymkfs(8).- norecovery
The filesystem will be mounted without running log recovery.If the filesystem was not cleanly unmounted, it is likely tobe inconsistent when mounted in
norecovery
mode.Some files or directories may not be accessible because of this.Filesystems mountednorecovery
must be mounted read-only orthe mount will fail.- nouuid
Don’t check for double mounted file systems using the filesystem
uuid
. This is useful to mount LVM snapshot volumes,and often used in combination withnorecovery
for mountingread-only snapshots.- noquota
Forcibly turns off all quota accounting and enforcementwithin the filesystem.
- uquota/usrquota/uqnoenforce/quota
User disk quota accounting enabled, and limits (optionally)enforced. Refer to xfs_quota(8) for further details.
- gquota/grpquota/gqnoenforce
Group disk quota accounting enabled and limits (optionally)enforced. Refer to xfs_quota(8) for further details.
- pquota/prjquota/pqnoenforce
Project disk quota accounting enabled and limits (optionally)enforced. Refer to xfs_quota(8) for further details.
- sunit=value and swidth=value
Used to specify the stripe unit and width for a RAID deviceor a stripe volume. “value” must be specified in 512-byteblock units. These options are only relevant to filesystemsthat were created with non-zero data alignment parameters.
The
sunit
andswidth
parameters specified must be compatiblewith the existing filesystem alignment characteristics. Ingeneral, that means the only valid changes tosunit
areincreasing it by a power-of-2 multiple. Validswidth
valuesare any integer multiple of a validsunit
value.Typically the only time these mount options are necessary ifafter an underlying RAID device has had its geometrymodified, such as adding a new disk to a RAID5 lun andreshaping it.
- swalloc
Data allocations will be rounded up to stripe width boundarieswhen the current end of file is being extended and the filesize is larger than the stripe width size.
- wsync
When specified, all filesystem namespace operations areexecuted synchronously. This ensures that when the namespaceoperation (create, unlink, etc) completes, the change to thenamespace is on stable storage. This is useful in HA setupswhere failover must not result in clients seeinginconsistent namespace presentation during or after afailover event.
Deprecation of V4 Format¶
The V4 filesystem format lacks certain features that are supported bythe V5 format, such as metadata checksumming, strengthened metadataverification, and the ability to store timestamps past the year 2038.Because of this, the V4 format is deprecated. All users should upgradeby backing up their files, reformatting, and restoring from the backup.
Administrators and users can detect a V4 filesystem by running xfs_infoagainst a filesystem mountpoint and checking for a string containing“crc=”. If no such string is found, please upgrade xfsprogs to thelatest version and try again.
The deprecation will take place in two parts. Support for mounting V4filesystems can now be disabled at kernel build time via Kconfig option.The option will default to yes until September 2025, at which time itwill be changed to default to no. In September 2030, support will beremoved from the codebase entirely.
Note: Distributors may choose to withdraw V4 format support earlier thanthe dates listed above.
Deprecated Mount Options¶
Name | Removal Schedule |
---|---|
Mounting with V4 filesystem | September 2030 |
Mounting ascii-ci filesystem | September 2030 |
ikeep/noikeep | September 2025 |
attr2/noattr2 | September 2025 |
Removed Mount Options¶
Name | Removed |
---|---|
delaylog/nodelaylog | v4.0 |
ihashsize | v4.0 |
irixsgid | v4.0 |
osyncisdsync/osyncisosync | v4.0 |
barrier | v4.19 |
nobarrier | v4.19 |
sysctls¶
The following sysctls are available for the XFS filesystem:
- fs.xfs.stats_clear (Min: 0 Default: 0 Max: 1)
Setting this to “1” clears accumulated XFS statisticsin /proc/fs/xfs/stat. It then immediately resets to “0”.
- fs.xfs.xfssyncd_centisecs (Min: 100 Default: 3000 Max: 720000)
The interval at which the filesystem flushes metadataout to disk and runs internal cache cleanup routines.
- fs.xfs.filestream_centisecs (Min: 1 Default: 3000 Max: 360000)
The interval at which the filesystem ages filestreams cachereferences and returns timed-out AGs back to the free streampool.
- fs.xfs.speculative_prealloc_lifetime
(Units: seconds Min: 1 Default: 300 Max: 86400)The interval at which the background scanning for inodeswith unused speculative preallocation runs. The scanremoves unused preallocation from clean inodes and releasesthe unused space back to the free pool.
- fs.xfs.speculative_cow_prealloc_lifetime
This is an alias for speculative_prealloc_lifetime.
- fs.xfs.error_level (Min: 0 Default: 3 Max: 11)
A volume knob for error reporting when internal errors occur.This will generate detailed messages & backtraces for filesystemshutdowns, for example. Current threshold values are:
XFS_ERRLEVEL_OFF: 0XFS_ERRLEVEL_LOW: 1XFS_ERRLEVEL_HIGH: 5
- fs.xfs.panic_mask (Min: 0 Default: 0 Max: 511)
Causes certain error conditions to call BUG(). Value is a bitmask;OR together the tags which represent errors which should cause panics:
XFS_NO_PTAG 0XFS_PTAG_IFLUSH 0x00000001XFS_PTAG_LOGRES 0x00000002XFS_PTAG_AILDELETE 0x00000004XFS_PTAG_ERROR_REPORT 0x00000008XFS_PTAG_SHUTDOWN_CORRUPT 0x00000010XFS_PTAG_SHUTDOWN_IOERROR 0x00000020XFS_PTAG_SHUTDOWN_LOGERROR 0x00000040XFS_PTAG_FSBLOCK_ZERO 0x00000080XFS_PTAG_VERIFIER_ERROR 0x00000100
This option is intended for debugging only.
- fs.xfs.irix_symlink_mode (Min: 0 Default: 0 Max: 1)
Controls whether symlinks are created with mode 0777 (default)or whether their mode is affected by the umask (irix mode).
- fs.xfs.irix_sgid_inherit (Min: 0 Default: 0 Max: 1)
Controls files created in SGID directories.If the group ID of the new file does not match the effective groupID or one of the supplementary group IDs of the parent dir, theISGID bit is cleared if the irix_sgid_inherit compatibility sysctlis set.
- fs.xfs.inherit_sync (Min: 0 Default: 1 Max: 1)
Setting this to “1” will cause the “sync” flag setby the xfs_io(8) chattr command on a directory to beinherited by files in that directory.
- fs.xfs.inherit_nodump (Min: 0 Default: 1 Max: 1)
Setting this to “1” will cause the “nodump” flag setby the xfs_io(8) chattr command on a directory to beinherited by files in that directory.
- fs.xfs.inherit_noatime (Min: 0 Default: 1 Max: 1)
Setting this to “1” will cause the “noatime” flag setby the xfs_io(8) chattr command on a directory to beinherited by files in that directory.
- fs.xfs.inherit_nosymlinks (Min: 0 Default: 1 Max: 1)
Setting this to “1” will cause the “nosymlinks” flag setby the xfs_io(8) chattr command on a directory to beinherited by files in that directory.
- fs.xfs.inherit_nodefrag (Min: 0 Default: 1 Max: 1)
Setting this to “1” will cause the “nodefrag” flag setby the xfs_io(8) chattr command on a directory to beinherited by files in that directory.
- fs.xfs.rotorstep (Min: 1 Default: 1 Max: 256)
In “inode32” allocation mode, this option determines how manyfiles the allocator attempts to allocate in the same allocationgroup before moving to the next allocation group. The intentis to control the rate at which the allocator moves betweenallocation groups when allocating extents for new files.
Deprecated Sysctls¶
Name | Removal Schedule |
---|---|
fs.xfs.irix_sgid_inherit | September 2025 |
fs.xfs.irix_symlink_mode | September 2025 |
fs.xfs.speculative_cow_prealloc_lifetime | September 2025 |
Removed Sysctls¶
Name | Removed |
---|---|
fs.xfs.xfsbufd_centisec | v4.0 |
fs.xfs.age_buffer_centisecs | v4.0 |
Error handling¶
XFS can act differently according to the type of error found during itsoperation. The implementation introduces the following concepts to the errorhandler:
- -failure speed:
Defines how fast XFS should propagate an error upwards when a specificerror is found during the filesystem operation. It can propagateimmediately, after a defined number of retries, after a set time period,or simply retry forever.
- -error classes:
Specifies the subsystem the error configuration will apply to, such asmetadata IO or memory allocation. Different subsystems will havedifferent error handlers for which behaviour can be configured.
- -error handlers:
Defines the behavior for a specific error.
The filesystem behavior during an error can be set via sysfs
files. Eacherror handler works independently - the first condition met by an error handlerfor a specific class will cause the error to be propagated rather than reset andretried.
The action taken by the filesystem when the error is propagated is contextdependent - it may cause a shut down in the case of an unrecoverable error,it may be reported back to userspace, or it may even be ignored becausethere’s nothing useful we can with the error or anyone we can report it to (e.g.during unmount).
The configuration files are organized into the following hierarchy for eachmounted filesystem:
/sys/fs/xfs/<dev>/error/<class>/<error>/
- Where:
- <dev>
The short device name of the mounted filesystem. This is the same devicename that shows up in XFS kernel error messages as “XFS(<dev>): ...”
- <class>
The subsystem the error configuration belongs to. As of 4.9, the definedclasses are:
“metadata”: applies metadata buffer write IO
- <error>
The individual error handler configurations.
Each filesystem has “global” error configuration options defined in their toplevel directory:
/sys/fs/xfs/<dev>/error/
- fail_at_unmount (Min: 0 Default: 1 Max: 1)
Defines the filesystem error behavior at unmount time.
If set to a value of 1, XFS will override all other error configurationsduring unmount and replace them with “immediate fail” characteristics.i.e. no retries, no retry timeout. This will always allow unmount tosucceed when there are persistent errors present.
If set to 0, the configured retry behaviour will continue until allretries and/or timeouts have been exhausted. This will delay unmountcompletion when there are persistent errors, and it may prevent thefilesystem from ever unmounting fully in the case of “retry forever”handler configurations.
Note: there is no guarantee that fail_at_unmount can be set while anunmount is in progress. It is possible that the
sysfs
entries areremoved by the unmounting filesystem before a “retry forever” errorhandler configuration causes unmount to hang, and hence the filesystemmust be configured appropriately before unmount begins to preventunmount hangs.
Each filesystem has specific error class handlers that define the errorpropagation behaviour for specific errors. There is also a “default” errorhandler defined, which defines the behaviour for all errors that don’t havespecific handlers defined. Where multiple retry constraints are configured fora single error, the first retry configuration that expires will cause the errorto be propagated. The handler configurations are found in the directory:
/sys/fs/xfs/<dev>/error/<class>/<error>/
- max_retries (Min: -1 Default: Varies Max: INTMAX)
Defines the allowed number of retries of a specific error beforethe filesystem will propagate the error. The retry count for a givenerror context (e.g. a specific metadata buffer) is reset every timethere is a successful completion of the operation.
Setting the value to “-1” will cause XFS to retry forever for thisspecific error.
Setting the value to “0” will cause XFS to fail immediately when thespecific error is reported.
Setting the value to “N” (where 0 < N < Max) will make XFS retry theoperation “N” times before propagating the error.
- retry_timeout_seconds (Min: -1 Default: Varies Max: 1 day)
Define the amount of time (in seconds) that the filesystem isallowed to retry its operations when the specific error isfound.
Setting the value to “-1” will allow XFS to retry forever for thisspecific error.
Setting the value to “0” will cause XFS to fail immediately when thespecific error is reported.
Setting the value to “N” (where 0 < N < Max) will allow XFS to retry theoperation for up to “N” seconds before propagating the error.
Note: The default behaviour for a specific error handler is dependent on boththe class and error context. For example, the default values for“metadata/ENODEV” are “0” rather than “-1” so that this error handler defaultsto “fail immediately” behaviour. This is done because ENODEV is a fatal,unrecoverable error no matter how many times the metadata IO is retried.
Workqueue Concurrency¶
XFS uses kernel workqueues to parallelize metadata update processes. Thisenables it to take advantage of storage hardware that can service many IOoperations simultaneously. This interface exposes internal implementationdetails of XFS, and as such is explicitly not part of any userspace API/ABIguarantee the kernel may give userspace. These are undocumented features ofthe generic workqueue implementation XFS uses for concurrency, and they areprovided here purely for diagnostic and tuning purposes and may change at anytime in the future.
The control knobs for a filesystem’s workqueues are organized by task at handand the short name of the data device. They all can be found in:
/sys/bus/workqueue/devices/${task}!${device}
Task | Description |
---|---|
xfs_iwalk-$pid | Inode scans of the entire filesystem. Currently limited tomount time quotacheck. |
xfs-gc | Background garbage collection of disk space that have beenspeculatively allocated beyond EOF or for staging copy onwrite operations. |
For example, the knobs for the quotacheck workqueue for /dev/nvme0n1 would befound in /sys/bus/workqueue/devices/xfs_iwalk-1111!nvme0n1/.
The interesting knobs for XFS workqueues are as follows:
Knob | Description |
---|---|
max_active | Maximum number of background threads that can be started torun the work. |
cpumask | CPUs upon which the threads are allowed to run. |
nice | Relative priority of scheduling the threads. These are thesame nice levels that can be applied to userspace processes. |