Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1 | |
| 2 | The SGI XFS Filesystem |
| 3 | ====================== |
| 4 | |
| 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. |
| 11 | |
| 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. |
| 15 | |
| 16 | |
| 17 | Mount Options |
| 18 | ============= |
| 19 | |
| 20 | When mounting an XFS filesystem, the following options are accepted. |
| 21 | |
| 22 | biosize=size |
| 23 | Sets the preferred buffered I/O size (default size is 64K). |
| 24 | "size" must be expressed as the logarithm (base2) of the |
| 25 | desired I/O size. |
| 26 | Valid values for this option are 14 through 16, inclusive |
| 27 | (i.e. 16K, 32K, and 64K bytes). On machines with a 4K |
| 28 | pagesize, 13 (8K bytes) is also a valid size. |
| 29 | The preferred buffered I/O size can also be altered on an |
| 30 | individual file basis using the ioctl(2) system call. |
| 31 | |
| 32 | ikeep/noikeep |
| 33 | When inode clusters are emptied of inodes, keep them around |
| 34 | on the disk (ikeep) - this is the traditional XFS behaviour |
| 35 | and is still the default for now. Using the noikeep option, |
| 36 | inode clusters are returned to the free space pool. |
| 37 | |
| 38 | logbufs=value |
| 39 | Set the number of in-memory log buffers. Valid numbers range |
| 40 | from 2-8 inclusive. |
| 41 | The default value is 8 buffers for filesystems with a |
| 42 | blocksize of 64K, 4 buffers for filesystems with a blocksize |
| 43 | of 32K, 3 buffers for filesystems with a blocksize of 16K |
| 44 | and 2 buffers for all other configurations. Increasing the |
| 45 | number of buffers may increase performance on some workloads |
| 46 | at the cost of the memory used for the additional log buffers |
| 47 | and their associated control structures. |
| 48 | |
| 49 | logbsize=value |
| 50 | Set the size of each in-memory log buffer. |
| 51 | Size may be specified in bytes, or in kilobytes with a "k" suffix. |
| 52 | Valid sizes for version 1 and version 2 logs are 16384 (16k) and |
| 53 | 32768 (32k). Valid sizes for version 2 logs also include |
| 54 | 65536 (64k), 131072 (128k) and 262144 (256k). |
| 55 | The default value for machines with more than 32MB of memory |
| 56 | is 32768, machines with less memory use 16384 by default. |
| 57 | |
| 58 | logdev=device and rtdev=device |
| 59 | Use an external log (metadata journal) and/or real-time device. |
| 60 | An XFS filesystem has up to three parts: a data section, a log |
| 61 | section, and a real-time section. The real-time section is |
| 62 | optional, and the log section can be separate from the data |
| 63 | section or contained within it. |
| 64 | |
| 65 | noalign |
| 66 | Data allocations will not be aligned at stripe unit boundaries. |
| 67 | |
| 68 | noatime |
| 69 | Access timestamps are not updated when a file is read. |
| 70 | |
| 71 | norecovery |
| 72 | The filesystem will be mounted without running log recovery. |
| 73 | If the filesystem was not cleanly unmounted, it is likely to |
| 74 | be inconsistent when mounted in "norecovery" mode. |
| 75 | Some files or directories may not be accessible because of this. |
| 76 | Filesystems mounted "norecovery" must be mounted read-only or |
| 77 | the mount will fail. |
| 78 | |
| 79 | nouuid |
| 80 | Don't check for double mounted file systems using the file system uuid. |
| 81 | This is useful to mount LVM snapshot volumes. |
| 82 | |
| 83 | osyncisosync |
| 84 | Make O_SYNC writes implement true O_SYNC. WITHOUT this option, |
| 85 | Linux XFS behaves as if an "osyncisdsync" option is used, |
| 86 | which will make writes to files opened with the O_SYNC flag set |
| 87 | behave as if the O_DSYNC flag had been used instead. |
| 88 | This can result in better performance without compromising |
| 89 | data safety. |
| 90 | However if this option is not in effect, timestamp updates from |
| 91 | O_SYNC writes can be lost if the system crashes. |
| 92 | If timestamp updates are critical, use the osyncisosync option. |
| 93 | |
| 94 | quota/usrquota/uqnoenforce |
| 95 | User disk quota accounting enabled, and limits (optionally) |
| 96 | enforced. |
| 97 | |
| 98 | grpquota/gqnoenforce |
| 99 | Group disk quota accounting enabled and limits (optionally) |
| 100 | enforced. |
| 101 | |
| 102 | sunit=value and swidth=value |
| 103 | Used to specify the stripe unit and width for a RAID device or |
| 104 | a stripe volume. "value" must be specified in 512-byte block |
| 105 | units. |
| 106 | If this option is not specified and the filesystem was made on |
| 107 | a stripe volume or the stripe width or unit were specified for |
| 108 | the RAID device at mkfs time, then the mount system call will |
| 109 | restore the value from the superblock. For filesystems that |
| 110 | are made directly on RAID devices, these options can be used |
| 111 | to override the information in the superblock if the underlying |
| 112 | disk layout changes after the filesystem has been created. |
| 113 | The "swidth" option is required if the "sunit" option has been |
| 114 | specified, and must be a multiple of the "sunit" value. |
| 115 | |
| 116 | sysctls |
| 117 | ======= |
| 118 | |
| 119 | The following sysctls are available for the XFS filesystem: |
| 120 | |
| 121 | fs.xfs.stats_clear (Min: 0 Default: 0 Max: 1) |
| 122 | Setting this to "1" clears accumulated XFS statistics |
| 123 | in /proc/fs/xfs/stat. It then immediately resets to "0". |
| 124 | |
| 125 | fs.xfs.xfssyncd_centisecs (Min: 100 Default: 3000 Max: 720000) |
| 126 | The interval at which the xfssyncd thread flushes metadata |
| 127 | out to disk. This thread will flush log activity out, and |
| 128 | do some processing on unlinked inodes. |
| 129 | |
| 130 | fs.xfs.xfsbufd_centisecs (Min: 50 Default: 100 Max: 3000) |
| 131 | The interval at which xfsbufd scans the dirty metadata buffers list. |
| 132 | |
| 133 | fs.xfs.age_buffer_centisecs (Min: 100 Default: 1500 Max: 720000) |
| 134 | The age at which xfsbufd flushes dirty metadata buffers to disk. |
| 135 | |
| 136 | fs.xfs.error_level (Min: 0 Default: 3 Max: 11) |
| 137 | A volume knob for error reporting when internal errors occur. |
| 138 | This will generate detailed messages & backtraces for filesystem |
| 139 | shutdowns, for example. Current threshold values are: |
| 140 | |
| 141 | XFS_ERRLEVEL_OFF: 0 |
| 142 | XFS_ERRLEVEL_LOW: 1 |
| 143 | XFS_ERRLEVEL_HIGH: 5 |
| 144 | |
| 145 | fs.xfs.panic_mask (Min: 0 Default: 0 Max: 127) |
| 146 | Causes certain error conditions to call BUG(). Value is a bitmask; |
| 147 | AND together the tags which represent errors which should cause panics: |
| 148 | |
| 149 | XFS_NO_PTAG 0 |
| 150 | XFS_PTAG_IFLUSH 0x00000001 |
| 151 | XFS_PTAG_LOGRES 0x00000002 |
| 152 | XFS_PTAG_AILDELETE 0x00000004 |
| 153 | XFS_PTAG_ERROR_REPORT 0x00000008 |
| 154 | XFS_PTAG_SHUTDOWN_CORRUPT 0x00000010 |
| 155 | XFS_PTAG_SHUTDOWN_IOERROR 0x00000020 |
| 156 | XFS_PTAG_SHUTDOWN_LOGERROR 0x00000040 |
| 157 | |
| 158 | This option is intended for debugging only. |
| 159 | |
| 160 | fs.xfs.irix_symlink_mode (Min: 0 Default: 0 Max: 1) |
| 161 | Controls whether symlinks are created with mode 0777 (default) |
| 162 | or whether their mode is affected by the umask (irix mode). |
| 163 | |
| 164 | fs.xfs.irix_sgid_inherit (Min: 0 Default: 0 Max: 1) |
| 165 | Controls files created in SGID directories. |
| 166 | If the group ID of the new file does not match the effective group |
| 167 | ID or one of the supplementary group IDs of the parent dir, the |
| 168 | ISGID bit is cleared if the irix_sgid_inherit compatibility sysctl |
| 169 | is set. |
| 170 | |
| 171 | fs.xfs.restrict_chown (Min: 0 Default: 1 Max: 1) |
| 172 | Controls whether unprivileged users can use chown to "give away" |
| 173 | a file to another user. |
| 174 | |
| 175 | fs.xfs.inherit_sync (Min: 0 Default: 1 Max 1) |
| 176 | Setting this to "1" will cause the "sync" flag set |
| 177 | by the chattr(1) command on a directory to be |
| 178 | inherited by files in that directory. |
| 179 | |
| 180 | fs.xfs.inherit_nodump (Min: 0 Default: 1 Max 1) |
| 181 | Setting this to "1" will cause the "nodump" flag set |
| 182 | by the chattr(1) command on a directory to be |
| 183 | inherited by files in that directory. |
| 184 | |
| 185 | fs.xfs.inherit_noatime (Min: 0 Default: 1 Max 1) |
| 186 | Setting this to "1" will cause the "noatime" flag set |
| 187 | by the chattr(1) command on a directory to be |
| 188 | inherited by files in that directory. |