1 2The Second Extended Filesystem 3============================== 4 5ext2 was originally released in January 1993. Written by R\'emy Card, 6Theodore Ts'o and Stephen Tweedie, it was a major rewrite of the 7Extended Filesystem. It is currently still (April 2001) the predominant 8filesystem in use by Linux. There are also implementations available 9for NetBSD, FreeBSD, the GNU HURD, Windows 95/98/NT, OS/2 and RISC OS. 10 11Options 12======= 13 14When mounting an ext2 filesystem, the following options are accepted. 15Defaults are marked with (*). 16 17bsddf (*) Makes `df' act like BSD. 18minixdf Makes `df' act like Minix. 19 20check=none, nocheck (*) Don't do extra checking of bitmaps on mount 21 (check=normal and check=strict options removed) 22 23debug Extra debugging information is sent to the 24 kernel syslog. Useful for developers. 25 26errors=continue (*) Keep going on a filesystem error. 27errors=remount-ro Remount the filesystem read-only on an error. 28errors=panic Panic and halt the machine if an error occurs. 29 30grpid, bsdgroups Give objects the same group ID as their parent. 31nogrpid, sysvgroups (*) New objects have the group ID of their creator. 32 33resuid=n The user ID which may use the reserved blocks. 34resgid=n The group ID which may use the reserved blocks. 35 36sb=n Use alternate superblock at this location. 37 38grpquota,noquota,quota,usrquota Quota options are silently ignored by ext2. 39 40 41Specification 42============= 43 44ext2 shares many properties with traditional Unix filesystems. It has 45the concepts of blocks, inodes and directories. It has space in the 46specification for Access Control Lists (ACLs), fragments, undeletion and 47compression though these are not yet implemented (some are available as 48separate patches). There is also a versioning mechanism to allow new 49features (such as journalling) to be added in a maximally compatible 50manner. 51 52Blocks 53------ 54 55The space in the device or file is split up into blocks. These are 56a fixed size, of 1024, 2048 or 4096 bytes (8192 bytes on Alpha systems), 57which is decided when the filesystem is created. Smaller blocks mean 58less wasted space per file, but require slightly more accounting overhead, 59and also impose other limits on the size of files and the filesystem. 60 61Block Groups 62------------ 63 64Blocks are clustered into block groups in order to reduce fragmentation 65and minimise the amount of head seeking when reading a large amount 66of consecutive data. Information about each block group is kept in a 67descriptor table stored in the block(s) immediately after the superblock. 68Two blocks near the start of each group are reserved for the block usage 69bitmap and the inode usage bitmap which show which blocks and inodes 70are in use. Since each bitmap is limited to a single block, this means 71that the maximum size of a block group is 8 times the size of a block. 72 73The block(s) following the bitmaps in each block group are designated 74as the inode table for that block group and the remainder are the data 75blocks. The block allocation algorithm attempts to allocate data blocks 76in the same block group as the inode which contains them. 77 78The Superblock 79-------------- 80 81The superblock contains all the information about the configuration of 82the filing system. The primary copy of the superblock is stored at an 83offset of 1024 bytes from the start of the device, and it is essential 84to mounting the filesystem. Since it is so important, backup copies of 85the superblock are stored in block groups throughout the filesystem. 86The first version of ext2 (revision 0) stores a copy at the start of 87every block group, along with backups of the group descriptor block(s). 88Because this can consume a considerable amount of space for large 89filesystems, later revisions can optionally reduce the number of backup 90copies by only putting backups in specific groups (this is the sparse 91superblock feature). The groups chosen are 0, 1 and powers of 3, 5 and 7. 92 93The information in the superblock contains fields such as the total 94number of inodes and blocks in the filesystem and how many are free, 95how many inodes and blocks are in each block group, when the filesystem 96was mounted (and if it was cleanly unmounted), when it was modified, 97what version of the filesystem it is (see the Revisions section below) 98and which OS created it. 99 100If the filesystem is revision 1 or higher, then there are extra fields, 101such as a volume name, a unique identification number, the inode size, 102and space for optional filesystem features to store configuration info. 103 104All fields in the superblock (as in all other ext2 structures) are stored 105on the disc in little endian format, so a filesystem is portable between 106machines without having to know what machine it was created on. 107 108Inodes 109------ 110 111The inode (index node) is a fundamental concept in the ext2 filesystem. 112Each object in the filesystem is represented by an inode. The inode 113structure contains pointers to the filesystem blocks which contain the 114data held in the object and all of the metadata about an object except 115its name. The metadata about an object includes the permissions, owner, 116group, flags, size, number of blocks used, access time, change time, 117modification time, deletion time, number of links, fragments, version 118(for NFS) and extended attributes (EAs) and/or Access Control Lists (ACLs). 119 120There are some reserved fields which are currently unused in the inode 121structure and several which are overloaded. One field is reserved for the 122directory ACL if the inode is a directory and alternately for the top 32 123bits of the file size if the inode is a regular file (allowing file sizes 124larger than 2GB). The translator field is unused under Linux, but is used 125by the HURD to reference the inode of a program which will be used to 126interpret this object. Most of the remaining reserved fields have been 127used up for both Linux and the HURD for larger owner and group fields, 128The HURD also has a larger mode field so it uses another of the remaining 129fields to store the extra more bits. 130 131There are pointers to the first 12 blocks which contain the file's data 132in the inode. There is a pointer to an indirect block (which contains 133pointers to the next set of blocks), a pointer to a doubly-indirect 134block (which contains pointers to indirect blocks) and a pointer to a 135trebly-indirect block (which contains pointers to doubly-indirect blocks). 136 137The flags field contains some ext2-specific flags which aren't catered 138for by the standard chmod flags. These flags can be listed with lsattr 139and changed with the chattr command, and allow specific filesystem 140behaviour on a per-file basis. There are flags for secure deletion, 141undeletable, compression, synchronous updates, immutability, append-only, 142dumpable, no-atime, indexed directories, and data-journaling. Not all 143of these are supported yet. 144 145Directories 146----------- 147 148A directory is a filesystem object and has an inode just like a file. 149It is a specially formatted file containing records which associate 150each name with an inode number. Later revisions of the filesystem also 151encode the type of the object (file, directory, symlink, device, fifo, 152socket) to avoid the need to check the inode itself for this information 153(support for taking advantage of this feature does not yet exist in 154Glibc 2.2). 155 156The inode allocation code tries to assign inodes which are in the same 157block group as the directory in which they are first created. 158 159The current implementation of ext2 uses a singly-linked list to store 160the filenames in the directory; a pending enhancement uses hashing of the 161filenames to allow lookup without the need to scan the entire directory. 162 163The current implementation never removes empty directory blocks once they 164have been allocated to hold more files. 165 166Special files 167------------- 168 169Symbolic links are also filesystem objects with inodes. They deserve 170special mention because the data for them is stored within the inode 171itself if the symlink is less than 60 bytes long. It uses the fields 172which would normally be used to store the pointers to data blocks. 173This is a worthwhile optimisation as it we avoid allocating a full 174block for the symlink, and most symlinks are less than 60 characters long. 175 176Character and block special devices never have data blocks assigned to 177them. Instead, their device number is stored in the inode, again reusing 178the fields which would be used to point to the data blocks. 179 180Reserved Space 181-------------- 182 183In ext2, there is a mechanism for reserving a certain number of blocks 184for a particular user (normally the super-user). This is intended to 185allow for the system to continue functioning even if non-priveleged users 186fill up all the space available to them (this is independent of filesystem 187quotas). It also keeps the filesystem from filling up entirely which 188helps combat fragmentation. 189 190Filesystem check 191---------------- 192 193At boot time, most systems run a consistency check (e2fsck) on their 194filesystems. The superblock of the ext2 filesystem contains several 195fields which indicate whether fsck should actually run (since checking 196the filesystem at boot can take a long time if it is large). fsck will 197run if the filesystem was not cleanly unmounted, if the maximum mount 198count has been exceeded or if the maximum time between checks has been 199exceeded. 200 201Feature Compatibility 202--------------------- 203 204The compatibility feature mechanism used in ext2 is sophisticated. 205It safely allows features to be added to the filesystem, without 206unnecessarily sacrificing compatibility with older versions of the 207filesystem code. The feature compatibility mechanism is not supported by 208the original revision 0 (EXT2_GOOD_OLD_REV) of ext2, but was introduced in 209revision 1. There are three 32-bit fields, one for compatible features 210(COMPAT), one for read-only compatible (RO_COMPAT) features and one for 211incompatible (INCOMPAT) features. 212 213These feature flags have specific meanings for the kernel as follows: 214 215A COMPAT flag indicates that a feature is present in the filesystem, 216but the on-disk format is 100% compatible with older on-disk formats, so 217a kernel which didn't know anything about this feature could read/write 218the filesystem without any chance of corrupting the filesystem (or even 219making it inconsistent). This is essentially just a flag which says 220"this filesystem has a (hidden) feature" that the kernel or e2fsck may 221want to be aware of (more on e2fsck and feature flags later). The ext3 222HAS_JOURNAL feature is a COMPAT flag because the ext3 journal is simply 223a regular file with data blocks in it so the kernel does not need to 224take any special notice of it if it doesn't understand ext3 journaling. 225 226An RO_COMPAT flag indicates that the on-disk format is 100% compatible 227with older on-disk formats for reading (i.e. the feature does not change 228the visible on-disk format). However, an old kernel writing to such a 229filesystem would/could corrupt the filesystem, so this is prevented. The 230most common such feature, SPARSE_SUPER, is an RO_COMPAT feature because 231sparse groups allow file data blocks where superblock/group descriptor 232backups used to live, and ext2_free_blocks() refuses to free these blocks, 233which would leading to inconsistent bitmaps. An old kernel would also 234get an error if it tried to free a series of blocks which crossed a group 235boundary, but this is a legitimate layout in a SPARSE_SUPER filesystem. 236 237An INCOMPAT flag indicates the on-disk format has changed in some 238way that makes it unreadable by older kernels, or would otherwise 239cause a problem if an old kernel tried to mount it. FILETYPE is an 240INCOMPAT flag because older kernels would think a filename was longer 241than 256 characters, which would lead to corrupt directory listings. 242The COMPRESSION flag is an obvious INCOMPAT flag - if the kernel 243doesn't understand compression, you would just get garbage back from 244read() instead of it automatically decompressing your data. The ext3 245RECOVER flag is needed to prevent a kernel which does not understand the 246ext3 journal from mounting the filesystem without replaying the journal. 247 248For e2fsck, it needs to be more strict with the handling of these 249flags than the kernel. If it doesn't understand ANY of the COMPAT, 250RO_COMPAT, or INCOMPAT flags it will refuse to check the filesystem, 251because it has no way of verifying whether a given feature is valid 252or not. Allowing e2fsck to succeed on a filesystem with an unknown 253feature is a false sense of security for the user. Refusing to check 254a filesystem with unknown features is a good incentive for the user to 255update to the latest e2fsck. This also means that anyone adding feature 256flags to ext2 also needs to update e2fsck to verify these features. 257 258Metadata 259-------- 260 261It is frequently claimed that the ext2 implementation of writing 262asynchronous metadata is faster than the ffs synchronous metadata 263scheme but less reliable. Both methods are equally resolvable by their 264respective fsck programs. 265 266If you're exceptionally paranoid, there are 3 ways of making metadata 267writes synchronous on ext2: 268 269per-file if you have the program source: use the O_SYNC flag to open() 270per-file if you don't have the source: use "chattr +S" on the file 271per-filesystem: add the "sync" option to mount (or in /etc/fstab) 272 273the first and last are not ext2 specific but do force the metadata to 274be written synchronously. See also Journaling below. 275 276Limitations 277----------- 278 279There are various limits imposed by the on-disk layout of ext2. Other 280limits are imposed by the current implementation of the kernel code. 281Many of the limits are determined at the time the filesystem is first 282created, and depend upon the block size chosen. The ratio of inodes to 283data blocks is fixed at filesystem creation time, so the only way to 284increase the number of inodes is to increase the size of the filesystem. 285No tools currently exist which can change the ratio of inodes to blocks. 286 287Most of these limits could be overcome with slight changes in the on-disk 288format and using a compatibility flag to signal the format change (at 289the expense of some compatibility). 290 291Filesystem block size: 1kB 2kB 4kB 8kB 292 293File size limit: 16GB 256GB 2048GB 2048GB 294Filesystem size limit: 2047GB 8192GB 16384GB 32768GB 295 296There is a 2.4 kernel limit of 2048GB for a single block device, so no 297filesystem larger than that can be created at this time. There is also 298an upper limit on the block size imposed by the page size of the kernel, 299so 8kB blocks are only allowed on Alpha systems (and other architectures 300which support larger pages). 301 302There is an upper limit of 32768 subdirectories in a single directory. 303 304There is a "soft" upper limit of about 10-15k files in a single directory 305with the current linear linked-list directory implementation. This limit 306stems from performance problems when creating and deleting (and also 307finding) files in such large directories. Using a hashed directory index 308(under development) allows 100k-1M+ files in a single directory without 309performance problems (although RAM size becomes an issue at this point). 310 311The (meaningless) absolute upper limit of files in a single directory 312(imposed by the file size, the realistic limit is obviously much less) 313is over 130 trillion files. It would be higher except there are not 314enough 4-character names to make up unique directory entries, so they 315have to be 8 character filenames, even then we are fairly close to 316running out of unique filenames. 317 318Journaling 319---------- 320 321A journaling extension to the ext2 code has been developed by Stephen 322Tweedie. It avoids the risks of metadata corruption and the need to 323wait for e2fsck to complete after a crash, without requiring a change 324to the on-disk ext2 layout. In a nutshell, the journal is a regular 325file which stores whole metadata (and optionally data) blocks that have 326been modified, prior to writing them into the filesystem. This means 327it is possible to add a journal to an existing ext2 filesystem without 328the need for data conversion. 329 330When changes to the filesystem (e.g. a file is renamed) they are stored in 331a transaction in the journal and can either be complete or incomplete at 332the time of a crash. If a transaction is complete at the time of a crash 333(or in the normal case where the system does not crash), then any blocks 334in that transaction are guaranteed to represent a valid filesystem state, 335and are copied into the filesystem. If a transaction is incomplete at 336the time of the crash, then there is no guarantee of consistency for 337the blocks in that transaction so they are discarded (which means any 338filesystem changes they represent are also lost). 339 340The ext3 code is currently (Apr 2001) available for 2.2 kernels only, 341and not yet available for 2.4 kernels. 342 343References 344========== 345 346The kernel source file:/usr/src/linux/fs/ext2/ 347e2fsprogs (e2fsck) http://e2fsprogs.sourceforge.net/ 348Design & Implementation http://e2fsprogs.sourceforge.net/ext2intro.html 349Journaling (ext3) ftp://ftp.uk.linux.org/pub/linux/sct/fs/jfs/ 350Hashed Directories http://kernelnewbies.org/~phillips/htree/ 351Filesystem Resizing http://ext2resize.sourceforge.net/ 352Extended Attributes & 353Access Control Lists http://acl.bestbits.at/ 354Compression (*) http://www.netspace.net.au/~reiter/e2compr/ 355 356Implementations for: 357Windows 95/98/NT/2000 http://uranus.it.swin.edu.au/~jn/linux/Explore2fs.htm 358Windows 95 (*) http://www.yipton.demon.co.uk/content.html#FSDEXT2 359DOS client (*) ftp://metalab.unc.edu/pub/Linux/system/filesystems/ext2/ 360OS/2 http://perso.wanadoo.fr/matthieu.willm/ext2-os2/ 361RISC OS client ftp://ftp.barnet.ac.uk/pub/acorn/armlinux/iscafs/ 362 363(*) no longer actively developed/supported (as of Apr 2001) 364