Root/Documentation/filesystems/xfs.txt

1
2The SGI XFS Filesystem
3======================
4
5XFS is a high performance journaling filesystem which originated
6on the SGI IRIX platform. It is completely multi-threaded, can
7support large files and large filesystems, extended attributes,
8variable block sizes, is extent based, and makes extensive use of
9Btrees (directories, extents, free space) to aid both performance
10and scalability.
11
12Refer to the documentation at http://oss.sgi.com/projects/xfs/
13for further details. This implementation is on-disk compatible
14with the IRIX version of XFS.
15
16
17Mount Options
18=============
19
20When mounting an XFS filesystem, the following options are accepted.
21
22  allocsize=size
23    Sets the buffered I/O end-of-file preallocation size when
24    doing delayed allocation writeout (default size is 64KiB).
25    Valid values for this option are page size (typically 4KiB)
26    through to 1GiB, inclusive, in power-of-2 increments.
27
28  attr2/noattr2
29    The options enable/disable (default is disabled for backward
30    compatibility on-disk) an "opportunistic" improvement to be
31    made in the way inline extended attributes are stored on-disk.
32    When the new form is used for the first time (by setting or
33    removing extended attributes) the on-disk superblock feature
34    bit field will be updated to reflect this format being in use.
35
36  barrier
37    Enables the use of block layer write barriers for writes into
38    the journal and unwritten extent conversion. This allows for
39    drive level write caching to be enabled, for devices that
40    support write barriers.
41
42  dmapi
43    Enable the DMAPI (Data Management API) event callouts.
44    Use with the "mtpt" option.
45
46  grpid/bsdgroups and nogrpid/sysvgroups
47    These options define what group ID a newly created file gets.
48    When grpid is set, it takes the group ID of the directory in
49    which it is created; otherwise (the default) it takes the fsgid
50    of the current process, unless the directory has the setgid bit
51    set, in which case it takes the gid from the parent directory,
52    and also gets the setgid bit set if it is a directory itself.
53
54  ihashsize=value
55    In memory inode hashes have been removed, so this option has
56    no function as of August 2007. Option is deprecated.
57
58  ikeep/noikeep
59    When ikeep is specified, XFS does not delete empty inode clusters
60    and keeps them around on disk. ikeep is the traditional XFS
61    behaviour. When noikeep is specified, empty inode clusters
62    are returned to the free space pool. The default is noikeep for
63    non-DMAPI mounts, while ikeep is the default when DMAPI is in use.
64
65  inode64
66    Indicates that XFS is allowed to create inodes at any location
67    in the filesystem, including those which will result in inode
68    numbers occupying more than 32 bits of significance. This is
69    provided for backwards compatibility, but causes problems for
70    backup applications that cannot handle large inode numbers.
71
72  largeio/nolargeio
73    If "nolargeio" is specified, the optimal I/O reported in
74    st_blksize by stat(2) will be as small as possible to allow user
75    applications to avoid inefficient read/modify/write I/O.
76    If "largeio" specified, a filesystem that has a "swidth" specified
77    will return the "swidth" value (in bytes) in st_blksize. If the
78    filesystem does not have a "swidth" specified but does specify
79    an "allocsize" then "allocsize" (in bytes) will be returned
80    instead.
81    If neither of these two options are specified, then filesystem
82    will behave as if "nolargeio" was specified.
83
84  logbufs=value
85    Set the number of in-memory log buffers. Valid numbers range
86    from 2-8 inclusive.
87    The default value is 8 buffers for filesystems with a
88    blocksize of 64KiB, 4 buffers for filesystems with a blocksize
89    of 32KiB, 3 buffers for filesystems with a blocksize of 16KiB
90    and 2 buffers for all other configurations. Increasing the
91    number of buffers may increase performance on some workloads
92    at the cost of the memory used for the additional log buffers
93    and their associated control structures.
94
95  logbsize=value
96    Set the size of each in-memory log buffer.
97    Size may be specified in bytes, or in kilobytes with a "k" suffix.
98    Valid sizes for version 1 and version 2 logs are 16384 (16k) and
99    32768 (32k). Valid sizes for version 2 logs also include
100    65536 (64k), 131072 (128k) and 262144 (256k).
101    The default value for machines with more than 32MiB of memory
102    is 32768, machines with less memory use 16384 by default.
103
104  logdev=device and rtdev=device
105    Use an external log (metadata journal) and/or real-time device.
106    An XFS filesystem has up to three parts: a data section, a log
107    section, and a real-time section. The real-time section is
108    optional, and the log section can be separate from the data
109    section or contained within it.
110
111  mtpt=mountpoint
112    Use with the "dmapi" option. The value specified here will be
113    included in the DMAPI mount event, and should be the path of
114    the actual mountpoint that is used.
115
116  noalign
117    Data allocations will not be aligned at stripe unit boundaries.
118
119  noatime
120    Access timestamps are not updated when a file is read.
121
122  norecovery
123    The filesystem will be mounted without running log recovery.
124    If the filesystem was not cleanly unmounted, it is likely to
125    be inconsistent when mounted in "norecovery" mode.
126    Some files or directories may not be accessible because of this.
127    Filesystems mounted "norecovery" must be mounted read-only or
128    the mount will fail.
129
130  nouuid
131    Don't check for double mounted file systems using the file system uuid.
132    This is useful to mount LVM snapshot volumes.
133
134  osyncisosync
135    Make O_SYNC writes implement true O_SYNC. WITHOUT this option,
136    Linux XFS behaves as if an "osyncisdsync" option is used,
137    which will make writes to files opened with the O_SYNC flag set
138    behave as if the O_DSYNC flag had been used instead.
139    This can result in better performance without compromising
140    data safety.
141    However if this option is not in effect, timestamp updates from
142    O_SYNC writes can be lost if the system crashes.
143    If timestamp updates are critical, use the osyncisosync option.
144
145  uquota/usrquota/uqnoenforce/quota
146    User disk quota accounting enabled, and limits (optionally)
147    enforced. Refer to xfs_quota(8) for further details.
148
149  gquota/grpquota/gqnoenforce
150    Group disk quota accounting enabled and limits (optionally)
151    enforced. Refer to xfs_quota(8) for further details.
152
153  pquota/prjquota/pqnoenforce
154    Project disk quota accounting enabled and limits (optionally)
155    enforced. Refer to xfs_quota(8) for further details.
156
157  sunit=value and swidth=value
158    Used to specify the stripe unit and width for a RAID device or
159    a stripe volume. "value" must be specified in 512-byte block
160    units.
161    If this option is not specified and the filesystem was made on
162    a stripe volume or the stripe width or unit were specified for
163    the RAID device at mkfs time, then the mount system call will
164    restore the value from the superblock. For filesystems that
165    are made directly on RAID devices, these options can be used
166    to override the information in the superblock if the underlying
167    disk layout changes after the filesystem has been created.
168    The "swidth" option is required if the "sunit" option has been
169    specified, and must be a multiple of the "sunit" value.
170
171  swalloc
172    Data allocations will be rounded up to stripe width boundaries
173    when the current end of file is being extended and the file
174    size is larger than the stripe width size.
175
176
177sysctls
178=======
179
180The following sysctls are available for the XFS filesystem:
181
182  fs.xfs.stats_clear (Min: 0 Default: 0 Max: 1)
183    Setting this to "1" clears accumulated XFS statistics
184    in /proc/fs/xfs/stat. It then immediately resets to "0".
185
186  fs.xfs.xfssyncd_centisecs (Min: 100 Default: 3000 Max: 720000)
187      The interval at which the xfssyncd thread flushes metadata
188      out to disk. This thread will flush log activity out, and
189      do some processing on unlinked inodes.
190
191  fs.xfs.xfsbufd_centisecs (Min: 50 Default: 100 Max: 3000)
192    The interval at which xfsbufd scans the dirty metadata buffers list.
193
194  fs.xfs.age_buffer_centisecs (Min: 100 Default: 1500 Max: 720000)
195    The age at which xfsbufd flushes dirty metadata buffers to disk.
196
197  fs.xfs.error_level (Min: 0 Default: 3 Max: 11)
198    A volume knob for error reporting when internal errors occur.
199    This will generate detailed messages & backtraces for filesystem
200    shutdowns, for example. Current threshold values are:
201
202        XFS_ERRLEVEL_OFF: 0
203        XFS_ERRLEVEL_LOW: 1
204        XFS_ERRLEVEL_HIGH: 5
205
206  fs.xfs.panic_mask (Min: 0 Default: 0 Max: 127)
207    Causes certain error conditions to call BUG(). Value is a bitmask;
208    AND together the tags which represent errors which should cause panics:
209
210        XFS_NO_PTAG 0
211        XFS_PTAG_IFLUSH 0x00000001
212        XFS_PTAG_LOGRES 0x00000002
213        XFS_PTAG_AILDELETE 0x00000004
214        XFS_PTAG_ERROR_REPORT 0x00000008
215        XFS_PTAG_SHUTDOWN_CORRUPT 0x00000010
216        XFS_PTAG_SHUTDOWN_IOERROR 0x00000020
217        XFS_PTAG_SHUTDOWN_LOGERROR 0x00000040
218
219    This option is intended for debugging only.
220
221  fs.xfs.irix_symlink_mode (Min: 0 Default: 0 Max: 1)
222    Controls whether symlinks are created with mode 0777 (default)
223    or whether their mode is affected by the umask (irix mode).
224
225  fs.xfs.irix_sgid_inherit (Min: 0 Default: 0 Max: 1)
226    Controls files created in SGID directories.
227    If the group ID of the new file does not match the effective group
228    ID or one of the supplementary group IDs of the parent dir, the
229    ISGID bit is cleared if the irix_sgid_inherit compatibility sysctl
230    is set.
231
232  fs.xfs.inherit_sync (Min: 0 Default: 1 Max: 1)
233    Setting this to "1" will cause the "sync" flag set
234    by the xfs_io(8) chattr command on a directory to be
235    inherited by files in that directory.
236
237  fs.xfs.inherit_nodump (Min: 0 Default: 1 Max: 1)
238    Setting this to "1" will cause the "nodump" flag set
239    by the xfs_io(8) chattr command on a directory to be
240    inherited by files in that directory.
241
242  fs.xfs.inherit_noatime (Min: 0 Default: 1 Max: 1)
243    Setting this to "1" will cause the "noatime" flag set
244    by the xfs_io(8) chattr command on a directory to be
245    inherited by files in that directory.
246
247  fs.xfs.inherit_nosymlinks (Min: 0 Default: 1 Max: 1)
248    Setting this to "1" will cause the "nosymlinks" flag set
249    by the xfs_io(8) chattr command on a directory to be
250    inherited by files in that directory.
251
252  fs.xfs.rotorstep (Min: 1 Default: 1 Max: 256)
253    In "inode32" allocation mode, this option determines how many
254    files the allocator attempts to allocate in the same allocation
255    group before moving to the next allocation group. The intent
256    is to control the rate at which the allocator moves between
257    allocation groups when allocating extents for new files.
258

Archive Download this file



interactive