About Kernel Documentation Linux Kernel Contact Linux Resources Linux Blog

Documentation / filesystems / xfs.txt

Custom Search

Based on kernel version 4.9. Page generated on 2016-12-21 14:34 EST.

2	The SGI XFS Filesystem
3	======================
5	XFS is a high performance journaling filesystem which originated
6	on the SGI IRIX platform.  It is completely multi-threaded, can
7	support large files and large filesystems, extended attributes,
8	variable block sizes, is extent based, and makes extensive use of
9	Btrees (directories, extents, free space) to aid both performance
10	and scalability.
12	Refer to the documentation at http://oss.sgi.com/projects/xfs/
13	for further details.  This implementation is on-disk compatible
14	with the IRIX version of XFS.
17	Mount Options
18	=============
20	When mounting an XFS filesystem, the following options are accepted.
21	For boolean mount options, the names with the (*) suffix is the
22	default behaviour.
24	  allocsize=size
25		Sets the buffered I/O end-of-file preallocation size when
26		doing delayed allocation writeout (default size is 64KiB).
27		Valid values for this option are page size (typically 4KiB)
28		through to 1GiB, inclusive, in power-of-2 increments.
30		The default behaviour is for dynamic end-of-file
31		preallocation size, which uses a set of heuristics to
32		optimise the preallocation size based on the current
33		allocation patterns within the file and the access patterns
34		to the file. Specifying a fixed allocsize value turns off
35		the dynamic behaviour.
37	  attr2
38	  noattr2
39		The options enable/disable an "opportunistic" improvement to
40		be made in the way inline extended attributes are stored
41		on-disk.  When the new form is used for the first time when
42		attr2 is selected (either when setting or removing extended
43		attributes) the on-disk superblock feature bit field will be
44		updated to reflect this format being in use.
46		The default behaviour is determined by the on-disk feature
47		bit indicating that attr2 behaviour is active. If either
48		mount option it set, then that becomes the new default used
49		by the filesystem.
51		CRC enabled filesystems always use the attr2 format, and so
52		will reject the noattr2 mount option if it is set.
54	  barrier (*)
55	  nobarrier
56		Enables/disables the use of block layer write barriers for
57		writes into the journal and for data integrity operations.
58		This allows for drive level write caching to be enabled, for
59		devices that support write barriers.
61	  discard
62	  nodiscard (*)
63		Enable/disable the issuing of commands to let the block
64		device reclaim space freed by the filesystem.  This is
65		useful for SSD devices, thinly provisioned LUNs and virtual
66		machine images, but may have a performance impact.
68		Note: It is currently recommended that you use the fstrim
69		application to discard unused blocks rather than the discard
70		mount option because the performance impact of this option
71		is quite severe.
73	  grpid/bsdgroups
74	  nogrpid/sysvgroups (*)
75		These options define what group ID a newly created file
76		gets.  When grpid is set, it takes the group ID of the
77		directory in which it is created; otherwise it takes the
78		fsgid of the current process, unless the directory has the
79		setgid bit set, in which case it takes the gid from the
80		parent directory, and also gets the setgid bit set if it is
81		a directory itself.
83	  filestreams
84		Make the data allocator use the filestreams allocation mode
85		across the entire filesystem rather than just on directories
86		configured to use it.
88	  ikeep
89	  noikeep (*)
90		When ikeep is specified, XFS does not delete empty inode
91		clusters and keeps them around on disk.  When noikeep is
92		specified, empty inode clusters are returned to the free
93		space pool.
95	  inode32
96	  inode64 (*)
97		When inode32 is specified, it indicates that XFS limits
98		inode creation to locations which will not result in inode
99		numbers with more than 32 bits of significance.
101		When inode64 is specified, it indicates that XFS is allowed
102		to create inodes at any location in the filesystem,
103		including those which will result in inode numbers occupying
104		more than 32 bits of significance. 
106		inode32 is provided for backwards compatibility with older
107		systems and applications, since 64 bits inode numbers might
108		cause problems for some applications that cannot handle
109		large inode numbers.  If applications are in use which do
110		not handle inode numbers bigger than 32 bits, the inode32
111		option should be specified.
114	  largeio
115	  nolargeio (*)
116		If "nolargeio" is specified, the optimal I/O reported in
117		st_blksize by stat(2) will be as small as possible to allow
118		user applications to avoid inefficient read/modify/write
119		I/O.  This is typically the page size of the machine, as
120		this is the granularity of the page cache.
122		If "largeio" specified, a filesystem that was created with a
123		"swidth" specified will return the "swidth" value (in bytes)
124		in st_blksize. If the filesystem does not have a "swidth"
125		specified but does specify an "allocsize" then "allocsize"
126		(in bytes) will be returned instead. Otherwise the behaviour
127		is the same as if "nolargeio" was specified.
129	  logbufs=value
130		Set the number of in-memory log buffers.  Valid numbers
131		range from 2-8 inclusive.
133		The default value is 8 buffers.
135		If the memory cost of 8 log buffers is too high on small
136		systems, then it may be reduced at some cost to performance
137		on metadata intensive workloads. The logbsize option below
138		controls the size of each buffer and so is also relevant to
139		this case.
141	  logbsize=value
142		Set the size of each in-memory log buffer.  The size may be
143		specified in bytes, or in kilobytes with a "k" suffix.
144		Valid sizes for version 1 and version 2 logs are 16384 (16k)
145		and 32768 (32k).  Valid sizes for version 2 logs also
146		include 65536 (64k), 131072 (128k) and 262144 (256k). The
147		logbsize must be an integer multiple of the log
148		stripe unit configured at mkfs time.
150		The default value for for version 1 logs is 32768, while the
151		default value for version 2 logs is MAX(32768, log_sunit).
153	  logdev=device and rtdev=device
154		Use an external log (metadata journal) and/or real-time device.
155		An XFS filesystem has up to three parts: a data section, a log
156		section, and a real-time section.  The real-time section is
157		optional, and the log section can be separate from the data
158		section or contained within it.
160	  noalign
161		Data allocations will not be aligned at stripe unit
162		boundaries. This is only relevant to filesystems created
163		with non-zero data alignment parameters (sunit, swidth) by
164		mkfs.
166	  norecovery
167		The filesystem will be mounted without running log recovery.
168		If the filesystem was not cleanly unmounted, it is likely to
169		be inconsistent when mounted in "norecovery" mode.
170		Some files or directories may not be accessible because of this.
171		Filesystems mounted "norecovery" must be mounted read-only or
172		the mount will fail.
174	  nouuid
175		Don't check for double mounted file systems using the file
176		system uuid.  This is useful to mount LVM snapshot volumes,
177		and often used in combination with "norecovery" for mounting
178		read-only snapshots.
180	  noquota
181		Forcibly turns off all quota accounting and enforcement
182		within the filesystem.
184	  uquota/usrquota/uqnoenforce/quota
185		User disk quota accounting enabled, and limits (optionally)
186		enforced.  Refer to xfs_quota(8) for further details.
188	  gquota/grpquota/gqnoenforce
189		Group disk quota accounting enabled and limits (optionally)
190		enforced.  Refer to xfs_quota(8) for further details.
192	  pquota/prjquota/pqnoenforce
193		Project disk quota accounting enabled and limits (optionally)
194		enforced.  Refer to xfs_quota(8) for further details.
196	  sunit=value and swidth=value
197		Used to specify the stripe unit and width for a RAID device
198		or a stripe volume.  "value" must be specified in 512-byte
199		block units. These options are only relevant to filesystems
200		that were created with non-zero data alignment parameters.
202		The sunit and swidth parameters specified must be compatible
203		with the existing filesystem alignment characteristics.  In
204		general, that means the only valid changes to sunit are
205		increasing it by a power-of-2 multiple. Valid swidth values
206		are any integer multiple of a valid sunit value.
208		Typically the only time these mount options are necessary if
209		after an underlying RAID device has had it's geometry
210		modified, such as adding a new disk to a RAID5 lun and
211		reshaping it.
213	  swalloc
214		Data allocations will be rounded up to stripe width boundaries
215		when the current end of file is being extended and the file
216		size is larger than the stripe width size.
218	  wsync
219		When specified, all filesystem namespace operations are
220		executed synchronously. This ensures that when the namespace
221		operation (create, unlink, etc) completes, the change to the
222		namespace is on stable storage. This is useful in HA setups
223		where failover must not result in clients seeing
224		inconsistent namespace presentation during or after a
225		failover event.
228	Deprecated Mount Options
229	========================
231	None at present.
234	Removed Mount Options
235	=====================
237	  Name				Removed
238	  ----				-------
239	  delaylog/nodelaylog		v4.0
240	  ihashsize			v4.0
241	  irixsgid			v4.0
242	  osyncisdsync/osyncisosync	v4.0
245	sysctls
246	=======
248	The following sysctls are available for the XFS filesystem:
250	  fs.xfs.stats_clear		(Min: 0  Default: 0  Max: 1)
251		Setting this to "1" clears accumulated XFS statistics
252		in /proc/fs/xfs/stat.  It then immediately resets to "0".
254	  fs.xfs.xfssyncd_centisecs	(Min: 100  Default: 3000  Max: 720000)
255		The interval at which the filesystem flushes metadata
256		out to disk and runs internal cache cleanup routines.
258	  fs.xfs.filestream_centisecs	(Min: 1  Default: 3000  Max: 360000)
259		The interval at which the filesystem ages filestreams cache
260		references and returns timed-out AGs back to the free stream
261		pool.
263	  fs.xfs.speculative_prealloc_lifetime
264			(Units: seconds   Min: 1  Default: 300  Max: 86400)
265		The interval at which the background scanning for inodes
266		with unused speculative preallocation runs. The scan
267		removes unused preallocation from clean inodes and releases
268		the unused space back to the free pool.
270	  fs.xfs.error_level		(Min: 0  Default: 3  Max: 11)
271		A volume knob for error reporting when internal errors occur.
272		This will generate detailed messages & backtraces for filesystem
273		shutdowns, for example.  Current threshold values are:
275			XFS_ERRLEVEL_OFF:       0
276			XFS_ERRLEVEL_LOW:       1
277			XFS_ERRLEVEL_HIGH:      5
279	  fs.xfs.panic_mask		(Min: 0  Default: 0  Max: 255)
280		Causes certain error conditions to call BUG(). Value is a bitmask;
281		OR together the tags which represent errors which should cause panics:
283			XFS_NO_PTAG                     0
284			XFS_PTAG_IFLUSH                 0x00000001
285			XFS_PTAG_LOGRES                 0x00000002
286			XFS_PTAG_AILDELETE              0x00000004
287			XFS_PTAG_ERROR_REPORT           0x00000008
288			XFS_PTAG_SHUTDOWN_CORRUPT       0x00000010
289			XFS_PTAG_SHUTDOWN_IOERROR       0x00000020
290			XFS_PTAG_SHUTDOWN_LOGERROR      0x00000040
291			XFS_PTAG_FSBLOCK_ZERO           0x00000080
293		This option is intended for debugging only.
295	  fs.xfs.irix_symlink_mode	(Min: 0  Default: 0  Max: 1)
296		Controls whether symlinks are created with mode 0777 (default)
297		or whether their mode is affected by the umask (irix mode).
299	  fs.xfs.irix_sgid_inherit	(Min: 0  Default: 0  Max: 1)
300		Controls files created in SGID directories.
301		If the group ID of the new file does not match the effective group
302		ID or one of the supplementary group IDs of the parent dir, the
303		ISGID bit is cleared if the irix_sgid_inherit compatibility sysctl
304		is set.
306	  fs.xfs.inherit_sync		(Min: 0  Default: 1  Max: 1)
307		Setting this to "1" will cause the "sync" flag set
308		by the xfs_io(8) chattr command on a directory to be
309		inherited by files in that directory.
311	  fs.xfs.inherit_nodump		(Min: 0  Default: 1  Max: 1)
312		Setting this to "1" will cause the "nodump" flag set
313		by the xfs_io(8) chattr command on a directory to be
314		inherited by files in that directory.
316	  fs.xfs.inherit_noatime	(Min: 0  Default: 1  Max: 1)
317		Setting this to "1" will cause the "noatime" flag set
318		by the xfs_io(8) chattr command on a directory to be
319		inherited by files in that directory.
321	  fs.xfs.inherit_nosymlinks	(Min: 0  Default: 1  Max: 1)
322		Setting this to "1" will cause the "nosymlinks" flag set
323		by the xfs_io(8) chattr command on a directory to be
324		inherited by files in that directory.
326	  fs.xfs.inherit_nodefrag	(Min: 0  Default: 1  Max: 1)
327		Setting this to "1" will cause the "nodefrag" flag set
328		by the xfs_io(8) chattr command on a directory to be
329		inherited by files in that directory.
331	  fs.xfs.rotorstep		(Min: 1  Default: 1  Max: 256)
332		In "inode32" allocation mode, this option determines how many
333		files the allocator attempts to allocate in the same allocation
334		group before moving to the next allocation group.  The intent
335		is to control the rate at which the allocator moves between
336		allocation groups when allocating extents for new files.
338	Deprecated Sysctls
339	==================
341	None at present.
344	Removed Sysctls
345	===============
347	  Name				Removed
348	  ----				-------
349	  fs.xfs.xfsbufd_centisec	v4.0
350	  fs.xfs.age_buffer_centisecs	v4.0
353	Error handling
354	==============
356	XFS can act differently according to the type of error found during its
357	operation. The implementation introduces the following concepts to the error
358	handler:
360	 -failure speed:
361		Defines how fast XFS should propagate an error upwards when a specific
362		error is found during the filesystem operation. It can propagate
363		immediately, after a defined number of retries, after a set time period,
364		or simply retry forever.
366	 -error classes:
367		Specifies the subsystem the error configuration will apply to, such as
368		metadata IO or memory allocation. Different subsystems will have
369		different error handlers for which behaviour can be configured.
371	 -error handlers:
372		Defines the behavior for a specific error.
374	The filesystem behavior during an error can be set via sysfs files. Each
375	error handler works independently - the first condition met by an error handler
376	for a specific class will cause the error to be propagated rather than reset and
377	retried.
379	The action taken by the filesystem when the error is propagated is context
380	dependent - it may cause a shut down in the case of an unrecoverable error,
381	it may be reported back to userspace, or it may even be ignored because
382	there's nothing useful we can with the error or anyone we can report it to (e.g.
383	during unmount).
385	The configuration files are organized into the following hierarchy for each
386	mounted filesystem:
388	  /sys/fs/xfs/<dev>/error/<class>/<error>/
390	Where:
391	  <dev>
392		The short device name of the mounted filesystem. This is the same device
393		name that shows up in XFS kernel error messages as "XFS(<dev>): ..."
395	  <class>
396		The subsystem the error configuration belongs to. As of 4.9, the defined
397		classes are:
399			- "metadata": applies metadata buffer write IO
401	  <error>
402		The individual error handler configurations.
405	Each filesystem has "global" error configuration options defined in their top
406	level directory:
408	  /sys/fs/xfs/<dev>/error/
410	  fail_at_unmount		(Min:  0  Default:  1  Max: 1)
411		Defines the filesystem error behavior at unmount time.
413		If set to a value of 1, XFS will override all other error configurations
414		during unmount and replace them with "immediate fail" characteristics.
415		i.e. no retries, no retry timeout. This will always allow unmount to
416		succeed when there are persistent errors present.
418		If set to 0, the configured retry behaviour will continue until all
419		retries and/or timeouts have been exhausted. This will delay unmount
420		completion when there are persistent errors, and it may prevent the
421		filesystem from ever unmounting fully in the case of "retry forever"
422		handler configurations.
424		Note: there is no guarantee that fail_at_unmount can be set whilst an
425		unmount is in progress. It is possible that the sysfs entries are
426		removed by the unmounting filesystem before a "retry forever" error
427		handler configuration causes unmount to hang, and hence the filesystem
428		must be configured appropriately before unmount begins to prevent
429		unmount hangs.
431	Each filesystem has specific error class handlers that define the error
432	propagation behaviour for specific errors. There is also a "default" error
433	handler defined, which defines the behaviour for all errors that don't have
434	specific handlers defined. Where multiple retry constraints are configuredi for
435	a single error, the first retry configuration that expires will cause the error
436	to be propagated. The handler configurations are found in the directory:
438	  /sys/fs/xfs/<dev>/error/<class>/<error>/
440	  max_retries			(Min: -1  Default: Varies  Max: INTMAX)
441		Defines the allowed number of retries of a specific error before
442		the filesystem will propagate the error. The retry count for a given
443		error context (e.g. a specific metadata buffer) is reset every time
444		there is a successful completion of the operation.
446		Setting the value to "-1" will cause XFS to retry forever for this
447		specific error.
449		Setting the value to "0" will cause XFS to fail immediately when the
450		specific error is reported.
452		Setting the value to "N" (where 0 < N < Max) will make XFS retry the
453		operation "N" times before propagating the error.
455	  retry_timeout_seconds		(Min:  -1  Default:  Varies  Max: 1 day)
456		Define the amount of time (in seconds) that the filesystem is
457		allowed to retry its operations when the specific error is
458		found.
460		Setting the value to "-1" will allow XFS to retry forever for this
461		specific error.
463		Setting the value to "0" will cause XFS to fail immediately when the
464		specific error is reported.
466		Setting the value to "N" (where 0 < N < Max) will allow XFS to retry the
467		operation for up to "N" seconds before propagating the error.
469	Note: The default behaviour for a specific error handler is dependent on both
470	the class and error context. For example, the default values for
471	"metadata/ENODEV" are "0" rather than "-1" so that this error handler defaults
472	to "fail immediately" behaviour. This is done because ENODEV is a fatal,
473	unrecoverable error no matter how many times the metadata IO is retried.
Hide Line Numbers
About Kernel Documentation Linux Kernel Contact Linux Resources Linux Blog

Information is copyright its respective author. All material is available from the Linux Kernel Source distributed under a GPL License. This page is provided as a free service by mjmwired.net.