1 .. SPDX-License-Identifier: GPL-2.0
2
3 Layout
4 ------
5
6 The layout of a standard block group is approx
7 of these fields is discussed in a separate sec
8
9 .. list-table::
10 :widths: 1 1 1 1 1 1 1 1
11 :header-rows: 1
12
13 * - Group 0 Padding
14 - ext4 Super Block
15 - Group Descriptors
16 - Reserved GDT Blocks
17 - Data Block Bitmap
18 - inode Bitmap
19 - inode Table
20 - Data Blocks
21 * - 1024 bytes
22 - 1 block
23 - many blocks
24 - many blocks
25 - 1 block
26 - 1 block
27 - many blocks
28 - many more blocks
29
30 For the special case of block group 0, the fir
31 to allow for the installation of x86 boot sect
32 The superblock will start at offset 1024 bytes
33 happens to be (usually 0). However, if for som
34 1024, then block 0 is marked in use and the su
35 For all other block groups, there is no paddin
36
37 The ext4 driver primarily works with the super
38 descriptors that are found in block group 0. R
39 superblock and group descriptors are written t
40 across the disk in case the beginning of the d
41 not all block groups necessarily host a redund
42 paragraph for more details). If the group does
43 copy, the block group begins with the data blo
44 when the filesystem is freshly formatted, mkfs
45 GDT block” space after the block group descr
46 of the block bitmaps to allow for future expan
47 default, a filesystem is allowed to increase i
48 1024x over the original filesystem size.
49
50 The location of the inode table is given by ``
51 is continuous range of blocks large enough to
52 ``sb.s_inodes_per_group * sb.s_inode_size`` by
53
54 As for the ordering of items in a block group,
55 established that the super block and the group
56 present, will be at the beginning of the block
57 the inode table can be anywhere, and it is qui
58 bitmaps to come after the inode table, or for
59 groups (flex_bg). Leftover space is used for f
60 block maps, extent tree blocks, and extended a
61
62 Flexible Block Groups
63 ---------------------
64
65 Starting in ext4, there is a new feature calle
66 (flex_bg). In a flex_bg, several block groups
67 logical block group; the bitmap spaces and the
68 first block group of the flex_bg are expanded
69 and inode tables of all other block groups in
70 if the flex_bg size is 4, then group 0 will co
71 superblock, group descriptors, data block bitm
72 bitmaps for groups 0-3, inode tables for group
73 space in group 0 is for file data. The effect
74 block group metadata close together for faster
75 large files to be continuous on disk. Backup c
76 and group descriptors are always at the beginn
77 if flex_bg is enabled. The number of block gro
78 flex_bg is given by 2 ^ ``sb.s_log_groups_per_
79
80 Meta Block Groups
81 -----------------
82
83 Without the option META_BG, for safety concern
84 descriptors copies are kept in the first block
85 128MiB(2^27 bytes) block group size and 64-byt
86 can have at most 2^27/64 = 2^21 block groups.
87 filesystem size to 2^21 * 2^27 = 2^48bytes or
88
89 The solution to this problem is to use the met
90 (META_BG), which is already in ext3 for all 2.
91 META_BG feature, ext4 filesystems are partitio
92 groups. Each metablock group is a cluster of b
93 descriptor structures can be stored in a singl
94 filesystems with 4 KB block size, a single met
95 includes 64 block groups, or 8 GiB of disk spa
96 feature moves the location of the group descri
97 first block group of the whole filesystem into
98 metablock group itself. The backups are in the
99 each metablock group. This increases the 2^21
100 to the hard limit 2^32, allowing support for a
101
102 The change in the filesystem format replaces t
103 the superblock is followed by a variable-lengt
104 descriptors. Instead, the superblock and a sin
105 block is placed at the beginning of the first,
106 groups in a meta-block group. A meta-block gro
107 block groups which can be described by a singl
108 block. Since the size of the block group descr
109 bytes, a meta-block group contains 16 block gr
110 a 1KB block size, and 64 block groups for file
111 blocksize. Filesystems can either be created u
112 descriptor layout, or existing filesystems can
113 the field s_first_meta_bg in the superblock wi
114 block group using this new layout.
115
116 Please see an important note about ``BLOCK_UNI
117 block and inode bitmaps.
118
119 Lazy Block Group Initialization
120 -------------------------------
121
122 A new feature for ext4 are three block group d
123 enable mkfs to skip initializing other parts o
124 metadata. Specifically, the INODE_UNINIT and B
125 that the inode and block bitmaps for that grou
126 therefore the on-disk bitmap blocks are not in
127 generally the case for an empty block group or
128 only fixed-location block group metadata. The
129 that the inode table has been initialized; mkf
130 rely on the kernel to initialize the inode tab
131
132 By not writing zeroes to the bitmaps and inode
133 reduced considerably. Note the feature flag is
134 but the dumpe2fs output prints this as “unin
135 thing.
Linux® is a registered trademark of Linus Torvalds in the United States and other countries.
TOMOYO® is a registered trademark of NTT DATA CORPORATION.