Copyright © 2001-2019 Dave Poirier
Permission is granted to copy, distribute and/or modify this document under the terms of the GNU Free Documentation License, Version 1.1 or any later version published by the Free Software Foundation; with no Invariant Sections, with no Front-Cover Texts, and with no Back-Cover Texts. A copy of the license can be acquired electronically from http://www.fsf.org/licenses/fdl.html or by writing to 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
Table of Contents
List of Figures
List of Tables
The latest version of this document may be downloaded from https://www.nongnu.org/ext2-doc/
This book is intended as an introduction and guide to the Second Extended File System, also known as Ext2. The reader should have a good understanding of the purpose of a file system as well as the associated vocabulary (file, directory, partition, etc).
Implementing file system drivers is already a daunting task, unfortunately except for tidbits of information here and there most of the documentation for the Second Extended Filesystem is in source files.
Hopefully this document will fix this problem, may it be of help to as many of you as possible.
Unless otherwise stated, all values are stored in little endian byte order.
Written by Remy Card, Theodore Ts'o and Stephen Tweedie as a major rewrite of the Extended Filesystem, it was first released to the public on January 1993 as part of the Linux kernel. One of its greatest achievement is the ability to extend the file system functionalities while maintaining the internal structures. This allowed an easier development of the Third Extended Filesystem (ext3) and the Fourth Extended Filesystem (ext4).
There are implementations available in most operating system including but not limited to NetBSD, FreeBSD, the GNU HURD, Microsoft Windows, IBM OS/2 and RISC OS.
Although newer file systems have been designed, such as Ext3 and Ext4, the Second Extended Filesystem is still prefered on flash drives as it requires fewer write operations (since it has no journal). The structures of Ext3 and Ext4 are based on Ext2 and add some additional options such as journaling, journal checksums, extents, online defragmentation, delayed allocations and larger directories to name but a few.
Table of Contents
The Second Extended Filesystem uses blocks as the basic unit of storage, inodes as the mean of keeping track of files and system objects, block groups to logically split the disk into more manageable sections, directories to provide a hierarchical organization of files, block and inode bitmaps to keep track of allocated blocks and inodes, and superblocks to define the parameters of the file system and its overall state.
Ext2 shares many properties with traditional Unix filesystems. It has space in the specification for Access Control Lists (ACLs), fragments, undeletion and compression. There is also a versioning mechanism to allow new features (such as journalling) to be added in a maximally compatible manner; such as in Ext3 and Ext4.
A partition, disk, file or block device formated with a Second Extended Filesystem is divided into small groups of sectors called “blocks”. These blocks are then grouped into larger units called block groups.
The size of the blocks are usually determined when formatting the disk and will have an impact on performance, maximum file size, and maximum file system size. Block sizes commonly implemented include 1KiB, 2KiB, 4KiB and 8KiB although provisions in the superblock allow for block sizes as big as 1024 * (2^31)-1 (see s_log_block_size).
Depending on the implementation, some architectures may impose limits on which block sizes are supported. For example, a Linux 2.6 implementation on DEC Alpha uses blocks of 8KiB but the same implementation on a Intel 386 processor will support a maximum block size of 4KiB.
Table 2.1. Impact of Block Sizes
| Upper Limits | 1KiB | 2KiB | 4KiB | 8KiB |
|---|---|---|---|---|
| file system blocks | 2,147,483,647 | 2,147,483,647 | 2,147,483,647 | 2,147,483,647 |
| blocks per block group | 8,192 | 16,384 | 32,768 | 65,536 |
| inodes per block group | 8,192 | 16,384 | 32,768 | 65,536 |
| bytes per block group | 8,388,608 (8MiB) | 33,554,432 (32MiB) | 134,217,728 (128MiB) | 536,870,912 (512MiB) |
| file system size (real) | 4,398,046,509,056 (4TiB) | 8,796,093,018,112 (8TiB) | 17,592,186,036,224 (16TiB) | 35,184,372,080,640 (32TiB) |
| file system size (Linux) | 2,199,023,254,528 (2TiB) [a] | 8,796,093,018,112 (8TiB) | 17,592,186,036,224 (16TiB) | 35,184,372,080,640 (32TiB) |
| blocks per file | 16,843,020 | 134,217,728 | 1,074,791,436 | 8,594,130,956 |
| file size (real) | 17,247,252,480 (16GiB) | 274,877,906,944 (256GiB) | 2,199,023,255,552 (2TiB) | 2,199,023,255,552 (2TiB) |
| file size (Linux 2.6.28) | 17,247,252,480 (16GiB) | 274,877,906,944 (256GiB) | 2,199,023,255,552 (2TiB) | 2,199,023,255,552 (2TiB) |
[a] This limit comes from the maximum size of a block device in Linux 2.4; it is unclear whether a Linux 2.6 kernel using a 1KiB block size could properly format and mount a Ext2 partition larger than 2TiB. | ||||
Note: the 2TiB file size is limited by the i_blocks value in the inode which indicates the number of 512-bytes sector rather than the actual number of ext2 blocks allocated.
This definition comes from the Linux Kernel Documentation.
Blocks are clustered into block groups in order to reduce fragmentation and minimise the amount of head seeking when reading a large amount of consecutive data. Information about each block group is kept in a descriptor table stored in the block(s) immediately after the superblock. Two blocks near the start of each group are reserved for the block usage bitmap and the inode usage bitmap which show which blocks and inodes are in use. Since each bitmap is limited to a single block, this means that the maximum size of a block group is 8 times the size of a block.
The block(s) following the bitmaps in each block group are designated as the inode table for that block group and the remainder are the data blocks. The block allocation algorithm attempts to allocate data blocks in the same block group as the inode which contains them.
This definition comes from the Linux Kernel Documentation with some minor alterations.
A directory is a filesystem object and has an inode just like a file. It is a specially formatted file containing records which associate each name with an inode number. Later revisions of the filesystem also encode the type of the object (file, directory, symlink, device, fifo, socket) to avoid the need to check the inode itself for this information
The inode allocation code should try to assign inodes which are in the same block group as the directory in which they are first created.
The original Ext2 revision used singly-linked list to store the filenames in the directory; newer revisions are able to use hashes and binary trees.
Also note that as directory grows additional blocks are assigned to store the additional file records. When filenames are removed, some implementations do not free these additional blocks.
This definition comes from the Linux Kernel Documentation with some minor alterations.
The inode (index node) is a fundamental concept in the ext2 filesystem. Each object in the filesystem is represented by an inode. The inode structure contains pointers to the filesystem blocks which contain the data held in the object and all of the metadata about an object except its name. The metadata about an object includes the permissions, owner, group, flags, size, number of blocks used, access time, change time, modification time, deletion time, number of links, fragments, version (for NFS) and extended attributes (EAs) and/or Access Control Lists (ACLs).
There are some reserved fields which are currently unused in the inode structure and several which are overloaded. One field is reserved for the directory ACL if the inode is a directory and alternately for the top 32 bits of the file size if the inode is a regular file (allowing file sizes larger than 2GB). The translator field is unused under Linux, but is used by the HURD to reference the inode of a program which will be used to interpret this object. Most of the remaining reserved fields have been used up for both Linux and the HURD for larger owner and group fields, The HURD also has a larger mode field so it uses another of the remaining fields to store the extra bits.
There are pointers to the first 12 blocks which contain the file's data in the inode. There is a pointer to an indirect block (which contains pointers to the next set of blocks), a pointer to a doubly-indirect block (which contains pointers to indirect blocks) and a pointer to a trebly-indirect block (which contains pointers to doubly-indirect blocks).
Some filesystem specific behaviour flags are also stored and allow for specific filesystem behaviour on a per-file basis. There are flags for secure deletion, undeletable, compression, synchronous updates, immutability, append-only, dumpable, no-atime, indexed directories, and data-journaling.
Many of the filesystem specific behaviour flags, like journaling, have been implemented in newer filesystems like Ext3 and Ext4, while some other are still under development.
All the inodes are stored in inode tables, with one inode table per block group.
This definition comes from the Linux Kernel Documentation with some minor alterations.
The superblock contains all the information about the configuration of the filesystem. The information in the superblock contains fields such as the total number of inodes and blocks in the filesystem and how many are free, how many inodes and blocks are in each block group, when the filesystem was mounted (and if it was cleanly unmounted), when it was modified, what version of the filesystem it is and which OS created it.
The primary copy of the superblock is stored at an offset of 1024 bytes from the start of the device, and it is essential to mounting the filesystem. Since it is so important, backup copies of the superblock are stored in block groups throughout the filesystem.
The first version of ext2 (revision 0) stores a copy at the start of every block group, along with backups of the group descriptor block(s). Because this can consume a considerable amount of space for large filesystems, later revisions can optionally reduce the number of backup copies by only putting backups in specific groups (this is the sparse superblock feature). The groups chosen are 0, 1 and powers of 3, 5 and 7.
Revision 1 and higher of the filesystem also store extra fields, such as a volume name, a unique identification number, the inode size, and space for optional filesystem features to store configuration info.
All fields in the superblock (as in all other ext2 structures) are stored on the disc in little endian format, so a filesystem is portable between machines without having to know what machine it was created on.
This definition comes from Wikipedia.org with some minor alterations.
A symbolic link (also symlink or soft link) is a special type of file that contains a reference to another file or directory in the form of an absolute or relative path and that affects pathname resolution.
Symbolic links operate transparently for most operations: programs which read or write to files named by a symbolic link will behave as if operating directly on the target file. However, programs that need to handle symbolic links specially (e.g., backup utilities) may identify and manipulate them directly.
A symbolic link merely contains a text string that is interpreted and followed by the operating system as a path to another file or directory. It is a file on its own and can exist independently of its target. The symbolic links do not affect an inode link count. If a symbolic link is deleted, its target remains unaffected. If the target is moved, renamed or deleted, any symbolic link that used to point to it continues to exist but now points to a non-existing file. Symbolic links pointing to non-existing files are sometimes called “orphaned” or “dangling”.
Symbolic links are also filesystem objects with inodes. For all symlink shorter than 60 bytes long, the data is stored within the inode itself; it uses the fields which would normally be used to store the pointers to data blocks. This is a worthwhile optimisation as it we avoid allocating a full block for the symlink, and most symlinks are less than 60 characters long.
Symbolic links can also point to files or directories of other partitions and file systems.
Table of Contents
An Ext2 file systems starts with a superblock located at byte offset 1024 from the start of the volume. This is block 1 for a 1KiB block formatted volume or within block 0 for larger block sizes. Note that the size of the superblock is constant regardless of the block size.
On the next block(s) following the superblock, is the Block Group Descriptor Table; which provides an overview of how the volume is split into block groups and where to find the inode bitmap, the block bitmap, and the inode table for each block group.
In revision 0 of Ext2, each block group consists of a copy superblock, a copy of the block group descriptor table, a block bitmap, an inode bitmap, an inode table, and data blocks.
With the introduction of revision 1 and the sparse superblock feature in Ext2, only specific block groups contain copies of the superblock and block group descriptor table. All block groups still contain the block bitmap, inode bitmap, inode table, and data blocks. The shadow copies of the superblock can be located in block groups 0, 1 and powers of 3, 5 and 7.
The block bitmap and inode bitmap are limited to 1 block each per block group, so the total blocks per block group is therefore limited. (More information in the Block Size Impact table).
Each data block may also be further divided into “fragments”. As of Linux 2.6.28, support for fragment was still not implemented in the kernel; it is therefore suggested to ensure the fragment size is equal to the block size so as to maintain compatibility.
Table 3.1. Sample Floppy Disk Layout, 1KiB blocks
| Block Offset | Length | Description |
|---|---|---|
| byte 0 | 512 bytes | boot record (if present) |
| byte 512 | 512 bytes | additional boot record data (if present) |
| -- block group 0, blocks 1 to 1439 -- | ||
| byte 1024 | 1024 bytes | superblock |
| block 2 | 1 block | block group descriptor table |
| block 3 | 1 block | block bitmap |
| block 4 | 1 block | inode bitmap |
| block 5 | 23 blocks | inode table |
| block 28 | 1412 blocks | data blocks |
For the curious, block 0 always points to the first sector of the disk or partition and will always contain the boot record if one is present.
The superblock is always located at byte offset 1024 from the start of the disk or partition. In a 1KiB block-size formatted file system, this is block 1, but it will always be block 0 (at 1024 bytes within block 0) in larger block size file systems.
And here's the organisation of a 20MB ext2 file system, using 1KiB blocks:
Table 3.2. Sample 20mb Partition Layout
| Block Offset | Length | Description |
|---|---|---|
| byte 0 | 512 bytes | boot record (if present) |
| byte 512 | 512 bytes | additional boot record data (if present) |
| -- block group 0, blocks 1 to 8192 -- | ||
| byte 1024 | 1024 bytes | superblock |
| block 2 | 1 block | block group descriptor table |
| block 3 | 1 block | block bitmap |
| block 4 | 1 block | inode bitmap |
| block 5 | 214 blocks | inode table |
| block 219 | 7974 blocks | data blocks |
| -- block group 1, blocks 8193 to 16384 -- | ||
| block 8193 | 1 block | superblock backup |
| block 8194 | 1 block | block group descriptor table backup |
| block 8195 | 1 block | block bitmap |
| block 8196 | 1 block | inode bitmap |
| block 8197 | 214 blocks | inode table |
| block 8408 | 7974 blocks | data blocks |
| -- block group 2, blocks 16385 to 24576 -- | ||
| block 16385 | 1 block | block bitmap |
| block 16386 | 1 block | inode bitmap |
| block 16387 | 214 blocks | inode table |
| block 16601 | 3879 blocks | data blocks |
The layout on disk is very predictable as long as you know a few basic information; block size, blocks per group, inodes per group. This information is all located in, or can be computed from, the superblock structure.
Nevertheless, unless the image was crafted with controlled parameters, the position of the various structures on disk (except the superblock) should never be assumed. Always load the superblock first.
Notice how block 0 is not part of the block group 0 in 1KiB block size file systems. The reason for this is block group 0 always starts with the block containing the superblock. Hence, on 1KiB block systems, block group 0 starts at block 1, but on larger block sizes it starts on block 0. For more information, see the s_first_data_block superblock entry.
The superblock is always located at byte offset 1024 from the beginning of the file, block device or partition formatted with Ext2 and later variants (Ext3, Ext4).
Its structure is mostly constant from Ext2 to Ext3 and Ext4 with only some minor changes.
Table 3.3. Superblock Structure
32bit value indicating the total number of inodes, both used and free, in the file system. This value must be lower or equal to (s_inodes_per_group * number of block groups). It must be equal to the sum of the inodes defined in each block group.
32bit value indicating the total number of blocks in the system including all used, free and reserved. This value must be lower or equal to (s_blocks_per_group * number of block groups). It can be lower than the previous calculation if the last block group has a smaller number of blocks than s_blocks_per_group du to volume size. It must be equal to the sum of the blocks defined in each block group.
32bit value indicating the total number of blocks reserved for the usage of the super user. This is most useful if for some reason a user, maliciously or not, fill the file system to capacity; the super user will have this specified amount of free blocks at his disposal so he can edit and save configuration files.
32bit value indicating the total number of free blocks, including the number of reserved blocks (see s_r_blocks_count). This is a sum of all free blocks of all the block groups.
32bit value indicating the total number of free inodes. This is a sum of all free inodes of all the block groups.
32bit value identifying the first data block, in other word the id of the block containing the superblock structure.
Note that this value is always 0 for file systems with a block size larger than 1KB, and always 1 for file systems with a block size of 1KB. The superblock is always starting at the 1024th byte of the disk, which normally happens to be the first byte of the 3rd sector.
The block size is computed using this 32bit value as the number of bits to shift left the value 1024. This value may only be non-negative.
block size = 1024 << s_log_block_size;
Common block sizes include 1KiB, 2KiB, 4KiB and 8Kib. For information about the impact of selecting a block size, see Impact of Block Sizes.
In Linux, at least up to 2.6.28, the block size must be at least as large as the sector size of the block device, and cannot be larger than the supported memory page of the architecture.
The fragment size is computed using this 32bit value as the number of bits to shift left the value 1024. Note that a negative value would shift the bit right rather than left.
if( positive )
fragmnet size = 1024 << s_log_frag_size;
else
framgnet size = 1024 >> -s_log_frag_size;
As of Linux 2.6.28 no support exists for an Ext2 partition with fragment size smaller than the block size, as this feature seems to not be available.
32bit value indicating the total number of blocks per group. This value in combination with s_first_data_block can be used to determine the block groups boundaries. Due to volume size boundaries, the last block group might have a smaller number of blocks than what is specified in this field.
32bit value indicating the total number of fragments per group. It is also used to determine the size of the block bitmap of each block group.
32bit value indicating the total number of inodes per group. This is also used to determine the size of the inode bitmap of each block group. Note that you cannot have more than (block size in bytes * 8) inodes per group as the inode bitmap must fit within a single block. This value must be a perfect multiple of the number of inodes that can fit in a block ((1024<<s_log_block_size)/s_inode_size).
16bit value indicating how many time the file system was mounted since the last time it was fully verified.
16bit value indicating the maximum number of times that the file system may be mounted before a full check is performed.
16bit value identifying the file system as Ext2. The value is
currently fixed to EXT2_SUPER_MAGIC of value 0xEF53.
16bit value indicating the file system state. When the file system is
mounted, this state is set to EXT2_ERROR_FS. After the
file system was cleanly unmounted, this value is set to EXT2_VALID_FS.
When mounting the file system, if a valid of EXT2_ERROR_FS is
encountered it means the file system was not cleanly unmounted and most
likely contain errors that will need to be fixed. Typically under Linux
this means running fsck.
Table 3.4. Defined s_state Values
| Constant Name | Value | Description |
|---|---|---|
| EXT2_VALID_FS | 1 | Unmounted cleanly |
| EXT2_ERROR_FS | 2 | Errors detected |
16bit value indicating what the file system driver should do when an error is detected. The following values have been defined:
Table 3.5. Defined s_errors Values
| Constant Name | Value | Description |
|---|---|---|
| EXT2_ERRORS_CONTINUE | 1 | continue as if nothing happened |
| EXT2_ERRORS_RO | 2 | remount read-only |
| EXT2_ERRORS_PANIC | 3 | cause a kernel panic |
16bit value identifying the minor revision level within its revision level.
Maximum Unix time interval, as defined by POSIX, allowed between file system checks.
32bit identifier of the os that created the file system. Defined values are:
Table 3.6. Defined s_creator_os Values
| Constant Name | Value | Description |
|---|---|---|
| EXT2_OS_LINUX | 0 | Linux |
| EXT2_OS_HURD | 1 | GNU HURD |
| EXT2_OS_MASIX | 2 | MASIX |
| EXT2_OS_FREEBSD | 3 | FreeBSD |
| EXT2_OS_LITES | 4 | Lites |
32bit revision level value.
Table 3.7. Defined s_rev_level Values
| Constant Name | Value | Description |
|---|---|---|
| EXT2_GOOD_OLD_REV | 0 | Revision 0 |
| EXT2_DYNAMIC_REV | 1 | Revision 1 with variable inode sizes, extended attributes, etc. |
16bit value used as the default user id for reserved blocks.
In Linux this defaults to EXT2_DEF_RESUID of 0.
16bit value used as the default group id for reserved blocks.
In Linux this defaults to EXT2_DEF_RESGID of 0.
32bit value used as index to the first inode useable for standard
files. In revision 0, the first non-reserved inode is fixed to
11 (EXT2_GOOD_OLD_FIRST_INO). In revision 1 and later
this value may be set to any value.
16bit value indicating the size of the inode structure. In revision 0, this
value is always 128 (EXT2_GOOD_OLD_INODE_SIZE). In revision 1
and later, this value must be a perfect power of 2 and must be smaller or
equal to the block size (1<<s_log_block_size).
16bit value used to indicate the block group number hosting this superblock structure. This can be used to rebuild the file system from any superblock backup.
32bit bitmask of compatible features. The file system implementation is free to support them or not without risk of damaging the meta-data.
Table 3.8. Defined s_feature_compat Values
| Constant Name | Value | Description |
|---|---|---|
| EXT2_FEATURE_COMPAT_DIR_PREALLOC | ||
| 0x0001 | Block pre-allocation for new directories | |
| EXT2_FEATURE_COMPAT_IMAGIC_INODES | ||
| 0x0002 | ||
| EXT3_FEATURE_COMPAT_HAS_JOURNAL | ||
| 0x0004 | An Ext3 journal exists | |
| EXT2_FEATURE_COMPAT_EXT_ATTR | ||
| 0x0008 | Extended inode attributes are present | |
| EXT2_FEATURE_COMPAT_RESIZE_INO | ||
| 0x0010 | Non-standard inode size used | |
| EXT2_FEATURE_COMPAT_DIR_INDEX | ||
| 0x0020 | Directory indexing (HTree) | |
32bit bitmask of incompatible features. The file system implementation should refuse to mount the file system if any of the indicated feature is unsupported.
An implementation not supporting these features would be unable to properly use the file system. For example, if compression is being used and an executable file would be unusable after being read from the disk if the system does not know how to uncompress it.
Table 3.9. Defined s_feature_incompat Values
| Constant Name | Value | Description |
|---|---|---|
| EXT2_FEATURE_INCOMPAT_COMPRESSION | ||
| 0x0001 | Disk/File compression is used | |
| EXT2_FEATURE_INCOMPAT_FILETYPE | ||
| 0x0002 | ||
| EXT3_FEATURE_INCOMPAT_RECOVER | ||
| 0x0004 | ||
| EXT3_FEATURE_INCOMPAT_JOURNAL_DEV | ||
| 0x0008 | ||
| EXT2_FEATURE_INCOMPAT_META_BG | ||
| 0x0010 | ||
32bit bitmask of “read-only” features. The file system implementation should mount as read-only if any of the indicated feature is unsupported.
Table 3.10. Defined s_feature_ro_compat Values
| Constant Name | Value | Description |
|---|---|---|
| EXT2_FEATURE_RO_COMPAT_SPARSE_SUPER | ||
| 0x0001 | Sparse Superblock | |
| EXT2_FEATURE_RO_COMPAT_LARGE_FILE | ||
| 0x0002 | Large file support, 64-bit file size | |
| EXT2_FEATURE_RO_COMPAT_BTREE_DIR | ||
| 0x0004 | Binary tree sorted directory files | |
128bit value used as the volume id. This should, as much as possible, be unique for each file system formatted.
16 bytes volume name, mostly unusued. A valid volume name would consist of only ISO-Latin-1 characters and be 0 terminated.
64 bytes directory path where the file system was last mounted. While not normally used, it could serve for auto-finding the mountpoint when not indicated on the command line. Again the path should be zero terminated for compatibility reasons. Valid path is constructed from ISO-Latin-1 characters.
32bit value used by compression algorithms to determine the compression method(s) used.
Compression is supported in Linux 2.4 and 2.6 via the e2compr patch. For more information, visit http://e2compr.sourceforge.net/
Table 3.11. Defined s_algo_bitmap Values
| Constant Name | Bit Number | Description |
|---|---|---|
| EXT2_LZV1_ALG | 0 | Binary value of 0x00000001 |
| EXT2_LZRW3A_ALG | 1 | Binary value of 0x00000002 |
| EXT2_GZIP_ALG | 2 | Binary value of 0x00000004 |
| EXT2_BZIP2_ALG | 3 | Binary value of 0x00000008 |
| EXT2_LZO_ALG | 4 | Binary value of 0x00000010 |
8-bit value representing the number of blocks the implementation should attempt to pre-allocate when creating a new regular file.
Linux 2.6.28 will only perform pre-allocation using Ext4 although no problem is expected if any version of Linux encounters a file with more blocks present than required.
8-bit value representing the number of blocks the implementation should attempt to pre-allocate when creating a new directory.
Linux 2.6.28 will only perform pre-allocation using Ext4 and only if
the EXT4_FEATURE_COMPAT_DIR_PREALLOC flag is present. Since
Linux does not de-allocate blocks from directories after they were
allocated, it should be safe to perform pre-allocation and maintain
compatibility with Linux.
16-byte value containing the uuid of the journal superblock. See Ext3 Journaling for more information.
32-bit inode number, pointing to the first inode in the list of inodes to delete. See Ext3 Journaling for more information.
An array of 4 32bit values containing the seeds used for the hash algorithm for directory indexing.
A 32bit value containing the default mount options for this file system. TODO: Add more information here!
The block group descriptor table is an array of block group descriptor, used to define parameters of all the block groups. It provides the location of the inode bitmap and inode table, block bitmap, number of free blocks and inodes, and some other useful information.
The block group descriptor table starts on the first block following the superblock. This would be the third block on a 1KiB block file system, or the second block for 2KiB and larger block file systems. Shadow copies of the block group descriptor table are also stored with every copy of the superblock.
Depending on how many block groups are defined, this table can require multiple blocks of storage. Always refer to the superblock in case of doubt.
The layout of a block group descriptor is as follows:
Table 3.12. Block Group Descriptor Structure
| Offset (bytes) | Size (bytes) | Description |
|---|---|---|
| 0 | 4 | bg_block_bitmap |
| 4 | 4 | bg_inode_bitmap |
| 8 | 4 | bg_inode_table |
| 12 | 2 | bg_free_blocks_count |
| 14 | 2 | bg_free_inodes_count |
| 16 | 2 | bg_used_dirs_count |
| 18 | 2 | bg_pad |
| 20 | 12 | bg_reserved |
For each block group in the file system, such a group_desc is created. Each represent a single block group within the file system and the information within any one of them is pertinent only to the group it is describing. Every block group descriptor table contains all the information about all the block groups.
NOTE: All indicated “block id” are absolute.
32bit block id of the first block of the “block bitmap” for the group represented.
The actual block bitmap is located within its own allocated blocks starting at the block ID specified by this value.
32bit block id of the first block of the “inode bitmap” for the group represented.
32bit block id of the first block of the “inode table” for the group represented.
16bit value indicating the total number of free blocks for the represented group.
16bit value indicating the total number of free inodes for the represented group.
16bit value indicating the number of inodes allocated to directories for the represented group.
On small file systems, the “Block Bitmap” is normally located at the first block, or second block if a superblock backup is present, of each block group. Its official location can be determined by reading the “bg_block_bitmap” in its associated group descriptor.
Each bit represent the current state of a block within that block group, where 1 means “used” and 0 “free/available”. The first block of this block group is represented by bit 0 of byte 0, the second by bit 1 of byte 0. The 8th block is represented by bit 7 (most significant bit) of byte 0 while the 9th block is represented by bit 0 (least significant bit) of byte 1.
The “Inode Bitmap” works in a similar way as the “Block Bitmap”, difference being in each bit representing an inode in the “Inode Table” rather than a block. Since inode numbers start from 1 rather than 0, the first bit in the first block group's inode bitmap represent inode number 1.
There is one inode bitmap per group and its location may be determined by reading the “bg_inode_bitmap” in its associated group descriptor.
When the inode table is created, all the reserved inodes are marked as used. In revision 0 this is the first 11 inodes.
The inode table is used to keep track of every directory, regular file, symbolic link, or special file; their location, size, type and access rights are all stored in inodes. There is no filename stored in the inode itself, names are contained in directory files only.
There is one inode table per block group and it can be located by reading the bg_inode_table in its associated group descriptor. There are s_inodes_per_group inodes per table.
Each inode contain the information about a single physical file on the system. A file can be a directory, a socket, a buffer, character or block device, symbolic link or a regular file. So an inode can be seen as a block of information related to an entity, describing its location on disk, its size and its owner. An inode looks like this:
The first few entries of the inode tables are reserved. In revision 0 there are 11 entries reserved while in revision 1 (EXT2_DYNAMIC_REV) and later the number of reserved inodes entries is specified in the s_first_ino of the superblock structure. Here's a listing of the known reserved inode entries:
Table 3.14. Defined Reserved Inodes
| Constant Name | Value | Description |
|---|---|---|
| EXT2_BAD_INO | 1 | bad blocks inode |
| EXT2_ROOT_INO | 2 | root directory inode |
| EXT2_ACL_IDX_INO | 3 | ACL index inode (deprecated?) |
| EXT2_ACL_DATA_INO | 4 | ACL data inode (deprecated?) |
| EXT2_BOOT_LOADER_INO | 5 | boot loader inode |
| EXT2_UNDEL_DIR_INO | 6 | undelete directory inode |
16bit value used to indicate the format of the described file and the access rights. Here are the possible values, which can be combined in various ways:
Table 3.15. Defined i_mode Values
| Constant | Value | Description |
|---|---|---|
| -- file format -- | ||
| EXT2_S_IFSOCK | 0xC000 | socket |
| EXT2_S_IFLNK | 0xA000 | symbolic link |
| EXT2_S_IFREG | 0x8000 | regular file |
| EXT2_S_IFBLK | 0x6000 | block device |
| EXT2_S_IFDIR | 0x4000 | directory |
| EXT2_S_IFCHR | 0x2000 | character device |
| EXT2_S_IFIFO | 0x1000 | fifo |
| -- process execution user/group override -- | ||
| EXT2_S_ISUID | 0x0800 | Set process User ID |
| EXT2_S_ISGID | 0x0400 | Set process Group ID |
| EXT2_S_ISVTX | 0x0200 | sticky bit |
| -- access rights -- | ||
| EXT2_S_IRUSR | 0x0100 | user read |
| EXT2_S_IWUSR | 0x0080 | user write |
| EXT2_S_IXUSR | 0x0040 | user execute |
| EXT2_S_IRGRP | 0x0020 | group read |
| EXT2_S_IWGRP | 0x0010 | group write |
| EXT2_S_IXGRP | 0x0008 | group execute |
| EXT2_S_IROTH | 0x0004 | others read |
| EXT2_S_IWOTH | 0x0002 | others write |
| EXT2_S_IXOTH | 0x0001 | others execute |
In revision 0, (signed) 32bit value indicating the size of the file in bytes. In revision 1 and later revisions, and only for regular files, this represents the lower 32-bit of the file size; the upper 32-bit is located in the i_dir_acl.
32bit value representing the number of seconds since january 1st 1970 of the last time this inode was accessed.
32bit value representing the number of seconds since january 1st 1970, of when the inode was created.
32bit value representing the number of seconds since january 1st 1970, of the last time this inode was modified.
32bit value representing the number of seconds since january 1st 1970, of when the inode was deleted.
16bit value indicating how many times this particular inode is linked (referred to). Most files will have a link count of 1. Files with hard links pointing to them will have an additional count for each hard link.
Symbolic links do not affect the link count of an inode. When the link count reaches 0 the inode and all its associated blocks are freed.
32-bit value representing the total number of 512-bytes blocks reserved to contain the data of this inode, regardless if these blocks are used or not. The block numbers of these reserved blocks are contained in the i_block array.
Since this value represents 512-byte blocks and not file system blocks, this value should not be directly used as an index to the i_block array. Rather, the maximum index of the i_block array should be computed from i_blocks / ((1024<<s_log_block_size)/512), or once simplified, i_blocks/(2<<s_log_block_size).
32bit value indicating how the ext2 implementation should behave when accessing the data for this inode.
Table 3.16. Defined i_flags Values
| Constant Name | Value | Description |
|---|---|---|
| EXT2_SECRM_FL | 0x00000001 | secure deletion |
| EXT2_UNRM_FL | 0x00000002 | record for undelete |
| EXT2_COMPR_FL | 0x00000004 | compressed file |
| EXT2_SYNC_FL | 0x00000008 | synchronous updates |
| EXT2_IMMUTABLE_FL | 0x00000010 | immutable file |
| EXT2_APPEND_FL | 0x00000020 | append only |
| EXT2_NODUMP_FL | 0x00000040 | do not dump/delete file |
| EXT2_NOATIME_FL | 0x00000080 | do not update .i_atime |
| -- Reserved for compression usage -- | ||
| EXT2_DIRTY_FL | 0x00000100 | Dirty (modified) |
| EXT2_COMPRBLK_FL | 0x00000200 | compressed blocks |
| EXT2_NOCOMPR_FL | 0x00000400 | access raw compressed data |
| EXT2_ECOMPR_FL | 0x00000800 | compression error |
| -- End of compression flags -- | ||
| EXT2_BTREE_FL | 0x00001000 | b-tree format directory |
| EXT2_INDEX_FL | 0x00001000 | hash indexed directory |
| EXT2_IMAGIC_FL | 0x00002000 | AFS directory |
| EXT3_JOURNAL_DATA_FL | 0x00004000 | journal file data |
| EXT2_RESERVED_FL | 0x80000000 | reserved for ext2 library |
32bit OS dependant value.
15 x 32bit block numbers pointing to the blocks containing the data for this inode. The first 12 blocks are direct blocks. The 13th entry in this array is the block number of the first indirect block; which is a block containing an array of block ID containing the data. Therefore, the 13th block of the file will be the first block ID contained in the indirect block. With a 1KiB block size, blocks 13 to 268 of the file data are contained in this indirect block.
The 14th entry in this array is the block number of the first doubly-indirect block; which is a block containing an array of indirect block IDs, with each of those indirect blocks containing an array of blocks containing the data. In a 1KiB block size, there would be 256 indirect blocks per doubly-indirect block, with 256 direct blocks per indirect block for a total of 65536 blocks per doubly-indirect block.
The 15th entry in this array is the block number of the triply-indirect block; which is a block containing an array of doubly-indrect block IDs, with each of those doubly-indrect block containing an array of indrect block, and each of those indirect block containing an array of direct block. In a 1KiB file system, this would be a total of 16777216 blocks per triply-indirect block.
In the original implementation of Ext2, a value of 0 in this array effectively terminated it with no further block defined. In sparse files, it is possible to have some blocks allocated and some others not yet allocated with the value 0 being used to indicate which blocks are not yet allocated for this file.
32bit value indicating the block number containing the extended attributes. In revision 0 this value is always 0.
Patches and implementation status of ACL under Linux can generally be found at http://acl.bestbits.at/
In revision 0 this 32bit value is always 0. In revision 1, for regular files this 32bit value contains the high 32 bits of the 64bit file size.
Linux sets this value to 0 if the file is not a regular file (i.e. block devices, directories, etc). In theory, this value could be set to point to a block containing extended attributes of the directory or special file.
32bit value indicating the location of the file fragment.
In Linux and GNU HURD, since fragments are unsupported this value is always 0. In Ext4 this value is now marked as obsolete.
In theory, this should contain the block number which hosts the actual fragment. The fragment number and its size would be contained in the i_osd2 structure.
96bit OS dependant structure.
Table 3.17. Inode i_osd2 Structure: Hurd
| Offset (bytes) | Size (bytes) | Description |
|---|---|---|
| 0 | 1 | h_i_frag |
| 1 | 1 | h_i_fsize |
| 2 | 2 | h_i_mode_high |
| 4 | 2 | h_i_uid_high |
| 6 | 2 | h_i_gid_high |
| 8 | 4 | h_i_author |
8bit fragment number. Always 0 GNU HURD since fragments are not supported. Obsolete with Ext4.
8bit fragment size. Always 0 in GNU HURD since fragments are not supported. Obsolete with Ext4.
High 16bit of user id.
High 16bit of group id.
32bit user id of the assigned file author. If this value is set to -1, the POSIX user id will be used.
Table 3.18. Inode i_osd2 Structure: Linux
| Offset (bytes) | Size (bytes) | Description |
|---|---|---|
| 0 | 1 | l_i_frag |
| 1 | 1 | l_i_fsize |
| 2 | 2 | reserved |
| 4 | 2 | l_i_uid_high |
| 6 | 2 | l_i_gid_high |
| 8 | 4 | reserved |
8bit fragment number.
Always 0 in Linux since fragments are not supported.
8bit fragment size.
Always 0 in Linux since fragments are not supported.
High 16bit of user id.
High 16bit of group id.
Inodes are all numerically ordered. The “inode number” is an index in the inode table to an inode structure. The size of the inode table is fixed at format time; it is built to hold a maximum number of entries. Due to the large amount of entries created, the table is quite big and thus, it is split equally among all the block groups (see Chapter 3, Disk Organization for more information).
The s_inodes_per_group field in the superblock structure tells us how many inodes are defined per group. Knowing that inode 1 is the first inode defined in the inode table, one can use the following formulaes:
block group = (inode - 1) / s_inodes_per_group
Once the block is identified, the local inode index for the local inode table can be identified using:
local inode index = (inode - 1) % s_inodes_per_group
Here are a couple of sample values that could be used to test your implementation:
Table 3.20. Sample Inode Computations
| Inode Number | Block Group Number | Local Inode Index |
|---|---|---|
| s_inodes_per_group = 1712 | ||
| 1 | 0 | 0 |
| 963 | 0 | 962 |
| 1712 | 0 | 1711 |
| 1713 | 1 | 0 |
| 3424 | 1 | 1711 |
| 3425 | 2 | 0 |
As many of you are most likely already familiar with, an index of 0 means the first entry. The reason behind using 0 rather than 1 is that it can more easily be multiplied by the structure size to find the final byte offset of its location in memory or on disk.
Table of Contents
Directories are used to hierarchically organize files. Each directory can contain other directories, regular files and special files.
Directories are stored as data block and referenced by an inode. They
can be identified by the file type
EXT2_S_IFDIR stored in the i_mode
field of the inode structure.
The second entry of the Inode table contains
the inode pointing to the data of the root directory; as defined by the
EXT2_ROOT_INO constant.
In revision 0 directories could only be stored in a linked list. Revision 1 and later introduced indexed directories. The indexed directory is backward compatible with the linked list directory; this is achieved by inserting empty directory entry records to skip over the hash indexes.
A directory file is a linked list of directory entry structures. Each structure contains the name of the entry, the inode associated with the data of this entry, and the distance within the directory file to the next entry.
In revision 0, the type of the entry (file, directory, special file, etc) has to be looked up in the inode of the file. In revision 0.5 and later, the file type is also contained in the directory entry structure.
Table 4.1. Linked Directory Entry Structure
16bit unsigned displacement to the next directory entry from the start of the current directory entry. This field must have a value at least equal to the length of the current record.
The directory entries must be aligned on 4 bytes boundaries and there cannot be any directory entry spanning multiple data blocks. If an entry cannot completely fit in one block, it must be pushed to the next data block and the rec_len of the previous entry properly adjusted.
Since this value cannot be negative, when a file is removed the previous record within the block has to be modified to point to the next valid record within the block or to the end of the block when no other directory entry is present.
If the first entry within the block is removed, a blank record will be created and point to the next directory entry or to the end of the block.
8bit unsigned value indicating how many bytes of character data are contained in the name.
This value must never be larger than rec_len - 8. If the directory entry name is updated and cannot fit in the existing directory entry, the entry may have to be relocated in a new directory entry of sufficient size and possibly stored in a new data block.
8bit unsigned value used to indicate file type.
In revision 0, this field was the upper 8-bit of the then 16-bit name_len. Since all implementations still limited the file names to 255 characters this 8-bit value was always 0.
This value must match the inode type defined in the related inode entry.
Table 4.2. Defined Inode File Type Values
| Constant Name | Value | Description |
|---|---|---|
| EXT2_FT_UNKNOWN | 0 | Unknown File Type |
| EXT2_FT_REG_FILE | 1 | Regular File |
| EXT2_FT_DIR | 2 | Directory File |
| EXT2_FT_CHRDEV | 3 | Character Device |
| EXT2_FT_BLKDEV | 4 | Block Device |
| EXT2_FT_FIFO | 5 | Buffer File |
| EXT2_FT_SOCK | 6 | Socket File |
| EXT2_FT_SYMLINK | 7 | Symbolic Link |
Name of the entry. The ISO-Latin-1 character set is expected in most system. The name must be no longer than 255 bytes after encoding.
Here's a sample of the home directory of one user on my system:
$ ls -1a ~
.
..
.bash_profile
.bashrc
mbox
public_html
tmp
For which the following data representation can be found on the storage device:
Table 4.3. Sample Linked Directory Data Layout, 4KiB blocks
| Offset (bytes) | Size (bytes) | Description |
|---|---|---|
| Directory Entry 0 | ||
| 0 | 4 | inode number: 783362 |
| 4 | 2 | record length: 12 |
| 6 | 1 | name length: 1 |
| 7 | 1 | file type: EXT2_FT_DIR=2 |
| 8 | 1 | name: . |
| 9 | 3 | padding |
| Directory Entry 1 | ||
| 12 | 4 | inode number: 1109761 |
| 16 | 2 | record length: 12 |
| 18 | 1 | name length: 2 |
| 19 | 1 | file type: EXT2_FT_DIR=2 |
| 20 | 2 | name: .. |
| 22 | 2 | padding |
| Directory Entry 2 | ||
| 24 | 4 | inode number: 783364 |
| 28 | 2 | record length: 24 |
| 30 | 1 | name length: 13 |
| 31 | 1 | file type: EXT2_FT_REG_FILE |
| 32 | 13 | name: .bash_profile |
| 45 | 3 | padding |
| Directory Entry 3 | ||
| 48 | 4 | inode number: 783363 |
| 52 | 2 | record length: 16 |
| 54 | 1 | name length: 7 |
| 55 | 1 | file type: EXT2_FT_REG_FILE |
| 56 | 7 | name: .bashrc |
| 63 | 1 | padding |
| Directory Entry 4 | ||
| 64 | 4 | inode number: 783377 |
| 68 | 2 | record length: 12 |
| 70 | 1 | name length: 4 |
| 71 | 1 | file type: EXT2_FT_REG_FILE |
| 72 | 4 | name: mbox |
| Directory Entry 5 | ||
| 76 | 4 | inode number: 783545 |
| 80 | 2 | record length: 20 |
| 82 | 1 | name length: 11 |
| 83 | 1 | file type: EXT2_FT_DIR=2 |
| 84 | 11 | name: public_html |
| 95 | 1 | padding |
| Directory Entry 6 | ||
| 96 | 4 | inode number: 669354 |
| 100 | 2 | record length: 12 |
| 102 | 1 | name length: 3 |
| 103 | 1 | file type: EXT2_FT_DIR=2 |
| 104 | 3 | name: tmp |
| 107 | 1 | padding |
| Directory Entry 7 | ||
| 108 | 4 | inode number: 0 |
| 112 | 2 | record length: 3988 |
| 114 | 1 | name length: 0 |
| 115 | 1 | file type: EXT2_FT_UNKNOWN |
| 116 | 0 | name: |
| 116 | 3980 | padding |
Using the standard linked list directory format can become very slow once the number of files starts growing. To improve performances in such a system, a hashed index is used, which allow to quickly locate the particular file searched.
Bit EXT2_INDEX_FL in the i_flags of the directory inode is set if the indexed directory format is used.
In order to maintain backward compatibility with older implementations, the indexed directory also maintains a linked directory format side-by-side. In case there's any discrepency between the indexed and linked directories, the linked directory is preferred.
This backward compatibility is achieved by placing a fake directory entries at the beginning of block 0 of the indexed directory data blocks. These fake entries are part of the dx_root structure and host the linked directory information for the "." and ".." folder entries.
Immediately following the the section called “Indexed Directory Root” structure is an array of the section called “Indexed Directory Entry” up to the end of the data block or until all files have been indexed.
When the number of files to be indexed exceeds the number of the section called “Indexed Directory Entry” that can fit in a block (the section called “limit”), a level of indirect indexes is created. An indirect index is another data block allocated to the directory inode that contains directory entries.
Table 4.4. Indexed Directory Root Structure
| Offset (bytes) | Size (bytes) | Description |
|---|---|---|
| -- Linked Directory Entry: . -- | ||
| 0 | 4 | inode: this directory |
| 4 | 2 | rec_len: 12 |
| 6 | 1 | name_len: 1 |
| 7 | 1 | file_type: EXT2_FT_DIR=2 |
| 8 | 1 | name: . |
| 9 | 3 | padding |
| -- Linked Directory Entry: .. -- | ||
| 12 | 4 | inode: parent directory |
| 16 | 2 | rec_len: (blocksize - this entry's length(12)) |
| 18 | 1 | name_len: 2 |
| 19 | 1 | file_type: EXT2_FT_DIR=2 |
| 20 | 2 | name: .. |
| 22 | 2 | padding |
| -- Indexed Directory Root Information Structure -- | ||
| 24 | 4 | reserved, zero |
| 28 | 1 | hash_version |
| 29 | 1 | info_length |
| 30 | 1 | indirect_levels |
| 31 | 1 | reserved - unused flags |
8bit value representing the hash version used in this indexed directory.
Table 4.5. Defined Indexed Directory Hash Versions
| Constant Name | Value | Description |
|---|---|---|
| DX_HASH_LEGACY | 0 | TODO: link to section |
| DX_HASH_HALF_MD4 | 1 | TODO: link to section |
| DX_HASH_TEA | 2 | TODO: link to section |
8bit length of the indexed directory information structure (dx_root); currently equal to 8.
The indexed directory entries are used to quickly lookup the inode number associated with the hash of a filename. These entries are located immediately following the fake linked directory entry of the directory data blocks, or immediately following the the section called “Indexed Directory Root”.
The first indexed directory entry, rather than containing an actual hash and block number, contains the maximum number of indexed directory entries that can fit in the block and the actual number of indexed directory entries stored in the block. The format of this special entry is detailed in Table 4.7, “Indexed Directory Entry Count and Limit Structure”.
The other directory entries are sorted by hash value starting from the smallest to the largest numerical value.
32bit block index of the directory inode data block containing the (linked) directory entry for the filename.
16bit value representing the total number of indexed directory entries that fit within the block, after removing the other structures, but including the count/limit entry.
Lookup is straightforword:
- Compute a hash of the name
- Read the index root
- Use binary search (linear in the current code) to find the
first index or leaf block that could contain the target hash
(in tree order)
- Repeat the above until the lowest tree level is reached
- Read the leaf directory entry block and do a normal Ext2
directory block search in it.
- If the name is found, return its directory entry and buffer
- Otherwise, if the collision bit of the next directory entry is
set, continue searching in the successor block
Normally, two logical blocks of the file will need to be accessed, and one or two metadata index blocks. The effect of the metadata index blocks can largely be ignored in terms of disk access time since these blocks are unlikely to be evicted from cache. There is some small CPU cost that can be addressed by moving the whole directory into the page cache.
Insertion of new entries into the directory is considerably more complex than lookup, due to the need to split leaf blocks when they become full, and to satisfy the conditions that allow hash key collisions to be handled reliably and efficiently. I'll just summarize here:
- Probe the index as for lookup
- If the target leaf block is full, split it and note the block
that will receive the new entry
- Insert the new entry in the leaf block using the normal Ext2
directory entry insertion code.
The details of splitting and hash collision handling are somewhat messy, but I will be happy to dwell on them at length if anyone is interested.
In brief, when a leaf node fills up and we want to put a new entry into it the leaf has to be split, and its share of the hash space has to be partitioned. The most straightforward way to do this is to sort the entrys by hash value and split somewhere in the middle of the sorted list. This operation is log(number_of_entries_in_leaf) and is not a great cost so long as an efficient sorter is used. I used Combsort for this, although Quicksort would have been just as good in this case since average case performance is more important than worst case.
An alternative approach would be just to guess a median value for the hash key, and the partition could be done in linear time, but the resulting poorer partitioning of hash key space outweighs the small advantage of the linear partition algorithm. In any event, the number of entries needing sorting is bounded by the number that fit in a leaf.
Some complexity is introduced by the need to handle sequences of hash key collisions. It is desireable to avoid splitting such sequences between blocks, so the split point of a block is adjusted with this in mind. But the possibility still remains that if the block fills up with identically-hashed entries, the sequence may still have to be split. This situation is flagged by placing a 1 in the low bit of the index entry that points at the sucessor block, which is naturally interpreted by the index probe as an intermediate value without any special coding. Thus, handling the collision problem imposes no real processing overhead, just come extra code and a slight reduction in the hash key space. The hash key space remains sufficient for any conceivable number of directory entries, up into the billions.
The exact properties of the hash function critically affect the performance of this indexing strategy, as I learned by trying a number of poor hash functions, at times intentionally. A poor hash function will result in many collisions or poor partitioning of the hash space. To illustrate why the latter is a problem, consider what happens when a block is split such that it covers just a few distinct hash values. The probability of later index entries hashing into the same, small hash space is very small. In practice, once a block is split, if its hash space is too small it tends to stay half full forever, an effect I observed in practice.
After some experimentation I came up with a hash function that gives reasonably good dispersal of hash keys across the entire 31 bit key space. This improved the average fullness of leaf blocks considerably, getting much closer to the theoretical average of 3/4 full.
But the current hash function is just a place holder, waiting for an better version based on some solid theory. I currently favor the idea of using crc32 as the default hash function, but I welcome suggestions.
Inevitably, no matter how good a hash function I come up with, somebody will come up with a better one later. For this reason the design allows for additional hash functiones to be added, with backward compatibility. This is accomplished simply, by including a hash function number in the index root. If a new, improved hash function is added, all the previous versions remain available, and previously created indexes remain readable.
Of course, the best strategy is to have a good hash function right from the beginning. The initial, quick hack has produced results that certainly have not been disappointing.
OK, if you have read this far then this is no doubt the part you've been waiting for. In short, the performance improvement over normal Ext2 has been stunning. With very small directories performance is similar to standard Ext2, but as directory size increases standard Ext2 quickly blows up quadratically, while htree-enhanced Ext2 continues to scale linearly.
Uli Luckas ran benchmarks for file creation in various sizes of directories ranging from 10,000 to 90,000 files. The results are pleasing: total file creation time stays very close to linear, versus quadratic increase with normal Ext2.
Time to create:
Figure 4.1. Performance of Indexed Directories
Indexed Normal
======= ======
10000 Files: 0m1.350s 0m23.670s
20000 Files: 0m2.720s 1m20.470s
30000 Files: 0m4.330s 3m9.320s
40000 Files: 0m5.890s 5m48.750s
50000 Files: 0m7.040s 9m31.270s
60000 Files: 0m8.610s 13m52.250s
70000 Files: 0m9.980s 19m24.070s
80000 Files: 0m12.060s 25m36.730s
90000 Files: 0m13.400s 33m18.550s
The original paper by Daniel Phillips is at https://www.kernel.org/doc/ols/2002/ols2002-pages-425-438.pdf
All of these tests are CPU-bound, which may come as a surprise. The directories fit easily in cache, and the limiting factor in the case of standard Ext2 is the looking up of directory blocks in buffer cache, and the low level scan of directory entries. In the case of htree indexing there are a number of costs to be considered, all of them pretty well bounded. Notwithstanding, there are a few obvious optimizations to be done:
- Use binary search instead of linear search in the interior index
nodes.
- If there is only one leaf block in a directory, bypass the index
probe, go straight to the block.
- Map the directory into the page cache instead of the buffer cache.
Each of these optimizations will produce a noticeable improvement in performance, but naturally it will never be anything like the big jump going from N**2 to Log512(N), ~= N. In time the optimizations will be applied and we can expect to see another doubling or so in performance.
There will be a very slight performance hit when the directory gets big enough to need a second level. Because of caching this will be very small. Traversing the directories metadata index blocks will be a bigger cost, and once again, this cost can be reduced by moving the directory blocks into the page cache.
Typically, we will traverse 3 blocks to read or write a directory entry, and that number increases to 4-5 with really huge directories. But this is really nothing compared to normal Ext2, which traverses several hundred blocks in the same situation.
Table of Contents
Most of the file (also directory, symlink, device...) attributes are located in the inode associated with the file. Some other attributes are only available as extended attributes.
There isn't much to say about those, they are located with the SGID and SUID bits in ext2_inode.i_mode.
The size of a file can be determined by looking at the ext2_inode.i_size field.
Under most implementations, the owner and group are 16bit values, but on some recent Linux and Hurd implementations the owner and group id are 32bit. When 16bit values are used, only the “low” part should be used as valid, while when using 32bit value, both the “low” and “high” part should be used, the high part being shifted left 16 places then added to the low part.
The low part of owner and group are located in ext2_inode.i_uid and ext2_inode.i_gid respectively.
The high part of owner and group are located in ext2_inode.osd2.hurd.h_i_uid_high and ext2_inode.osd2.hurd.h_i_gid_high, respectively, for Hurd and located in ext2_inode.osd2.linux.l_i_uid_high and ext2_inode.osd2.linux.l_i_gid_high, respectively, for Linux.
Extended attributes are name:value pairs associated permanently with files and directories, similar to the environment strings associated with a process. An attribute may be defined or undefined. If it is defined, its value may be empty or non-empty.
Extended attributes are extensions to the normal attributes which are associated with all inodes in the system. They are often used to provide additional functionality to a filesystem - for example, additional security features such as Access Control Lists (ACLs) may be implemented using extended attributes.
Extended attributes are accessed as atomic objects. Reading retrieves the whole value of an attribute and stores it in a buffer. Writing replaces any previous value with the new value.
Extended attributes are stored on disk blocks allocated outside of any inode. The i_file_acl field (for regular files) or the i_dir_acl field (for directories) fields contain the block number of the allocated data block used to store the extended attributes.
Inodes which have all identical extended attributes may share the same extended attribute block.
The attribute values are on the same block as their attribute entry descriptions, aligned to the end of the attribute block. This allows for additional attributes to be added more easily. The size of entry headers varies with the length of the attribute name.
The block header is followed by multiple entry descriptors. These entry
descriptors are variable in size, and aligned to EXT2_XATTR_PAD
(4) byte boundaries. The entry descriptors are sorted by attribute name,
so that two extended attribute blocks can be compared efficiently.
Attribute values are aligned to the end of the block, stored in
no specific order. They are also padded to EXT2_XATTR_PAD (4)
byte boundaries. No additional gaps are left between them.
Table 5.1. Extended Attribute Block Layout
| Attribute Block Header | ||
| Attribute Entry 1 | | | |
| Attribute Entry 2 | | | growing downwards |
| Attribute Entry 3 | V | |
| 4 null bytes | ||
| unused space... | ||
| Attribute Value 1 | ^ | |
| Attribute Value 3 | | | growing upwards |
| Attribute Value 2 | | |
Table 5.2. ext2_xattr_header structure
| Offset (bytes) | Size (bytse) | Description |
|---|---|---|
| 0 | 4 | h_magic |
| 4 | 4 | h_refcount |
| 8 | 4 | h_blocks |
| 12 | 4 | h_hash |
| 16 | 16 | reserved |
32bit value used as reference count. This value is incremented everytime a link is created to this attribute block and decremented when a link is destroyed. Whenever this value reaches 0 the attribute block can be freed.
32bit value indicating how many blocks are currently used by the extended attributes.
In Linux a value of h_blocks higher than 1 is considered invalid. This effectively restrict the amount of extended attributes to what can be fit in a single block.
There does not seem to be any support for extended attributes in Ext2 under GNU HURD.
32bit hash value of all attribute entry header hashes.
Procedure 5.1. Procedure to compute Extended Attribute Header Hash
Initialize the 32bit hash to 0
Check if there are any extended attribute entry to process, if not we are done.
Do a cyclic bit shift of 16 bits to the left of the 32bits hash value, effectively swapping the upper and lower 16bits of the hash
Perform a bitwise OR between the extended attribute entry hash and the header hash being computed.
Go back to Step 2.
Figure 5.1. ext2_xattr_header structure
offset size description
------- ------- -----------
0 1 e_name_len
1 1 e_name_index
2 2 e_value_offs
4 4 e_value_block
8 4 e_value_size
12 4 e_hash
16 ... e_name
The total size of an attribute entry is always rounded to the next 4-bytes boundary.
The i_flags value in the inode structure allows to specify how the file system should behave in regard to the file. The following bits are currently defined:
Table 5.3. Behaviour Control Flags
| EXT2_SECRM_FL | 0x00000001 | secure deletion |
| EXT2_UNRM_FL | 0x00000002 | record for undelete |
| EXT2_COMPR_FL | 0x00000004 | compressed file |
| EXT2_SYNC_FL | 0x00000008 | synchronous updates |
| EXT2_IMMUTABLE_FL | 0x00000010 | immutable file |
| EXT2_APPEND_FL | 0x00000020 | append only |
| EXT2_NODUMP_FL | 0x00000040 | do not dump/delete file |
| EXT2_NOATIME_FL | 0x00000080 | do not update .i_atime |
| EXT2_DIRTY_FL | 0x00000100 | dirty (file is in use?) |
| EXT2_COMPRBLK_FL | 0x00000200 | compressed blocks |
| EXT2_NOCOMPR_FL | 0x00000400 | access raw compressed data |
| EXT2_ECOMPR_FL | 0x00000800 | compression error |
| EXT2_BTREE_FL | 0x00001000 | b-tree format directory |
| EXT2_INDEX_FL | 0x00001000 | Hash indexed directory |
| EXT2_IMAGIC_FL | 0x00002000 | ? |
| EXT3_JOURNAL_DATA_FL | 0x00004000 | journal file data |
| EXT2_RESERVED_FL | 0x80000000 | reserved for ext2 implementation |
Enabling this bit will cause random data to be written over the file's content several times before the blocks are unlinked. Note that this is highly implementation dependant and as such, it should not be assumed to be 100% secure. Make sure to study the implementation notes before relying on this option.
When supported by the implementation, setting this bit will cause the deleted data to be moved to a temporary location, where the user can restore the original file without any risk of data lost. This is most useful when using ext2 on a desktop or workstation.
The file's content is compressed. There is no note about the particular algorithm used other than maybe the s_algo_bitmap field of the superblock structure.
The file's content in memory will be constantly synchronized with the content on disk. This is mostly used for very sensitive boot files or encryption keys that you do not want to lose in case of a crash.
The blocks associated with the file will not be exchanged. If for any reason a file system defragmentation is launched, such files will not be moved. Mostly used for stage2 and stage1.5 boot loaders.
Writing can only be used to append content at the end of the file and not modify the current content. Example of such use could be mailboxes, where anybody could send a message to a user but not modify any already present.
Setting this bit will protect the file from deletion. As long as this bit is set, even if the i_links_count is 0, the file will not be removed.
The i_atime field of the inode structure will not be modified when the file is accessed if this bit is set. The only good use I can think of that are related to security.
This flag is set if one or more blocks are compressed. You can have more information about compression on ext2 at http://www.netspace.net.au/~reiter/e2compr/ Note that the project has not been updated since 1999.
When this flag is set, the file system implementation will not uncompress the data before fowarding it to the application but will rather give it as is.
This flag is set if an error was detected when trying to uncompress the file.
When this bit is set, the format of the directory file is hash indexed. This is covered in details in the section called “Indexed Directory Format”.
I would like to personally thank everybody who contributed to this document, you are numerous and in many cases I haven't kept track of all of you. Be sure that if you are not in this list, it's a mistake and do not hesitate to contact me, it will be a pleasure to add your name to the list.
Peter Rottengatter (Peter.Rottengatter@bakerhughes.com) Corrections to the section called “s_inodes_per_group” Corrections to Table 3.1, “Sample Floppy Disk Layout, 1KiB blocks” and Table 3.2, “Sample 20mb Partition Layout” Corrections to the section called “Block Group Descriptor Table” Ryan Cuthbertson (ryan.cuthbertson@adelaide.edu.au) Corrections to the section called “i_blocks” Corrections to Chapter 3, Disk Organization Andreas Gruenbacher (a.gruenbacher@bestbits.at) the section called “Extended Attributes” Daniel Phillips (phillips@innominate.de) the section called “Lookup Algorithm” the section called “Insert Algorithm” the section called “Splitting” the section called “Key Collisions” the section called “Hash Function” the section called “Performance” Jeremy Stanley of Access Data Inc. Pointed out the inversed values for EXT2_S_IFSOCK and EXT2_S_IFLNK Ahmed S. Darwish (darwish.07@gmail.com) Clarification on the section called “Inode Bitmap” Sami Besalel (sami.besalel@software.dell.com) Typography correction in the section called “EXT2_SECRM_FL - Secure Deletion”. Kwan Fong (kwanlapfong@gmail.com) Improvement to wording in the section called “s_log_block_size” Jonatan Schroeder (jonatan@cs.ubc.ca) Corrections to the section called “i_block” with sparse files.