1.. SPDX-License-Identifier: GPL-2.0
2
3Checksums
4---------
5
6Starting in early 2012, metadata checksums were added to all major ext4
7and jbd2 data structures. The associated feature flag is metadata_csum.
8The desired checksum algorithm is indicated in the superblock, though as
9of October 2012 the only supported algorithm is crc32c. Some data
10structures did not have space to fit a full 32-bit checksum, so only the
11lower 16 bits are stored. Enabling the 64bit feature increases the data
12structure size so that full 32-bit checksums can be stored for many data
13structures. However, existing 32-bit filesystems cannot be extended to
14enable 64bit mode, at least not without the experimental resize2fs
15patches to do so.
16
17Existing filesystems can have checksumming added by running
18``tune2fs -O metadata_csum`` against the underlying device. If tune2fs
19encounters directory blocks that lack sufficient empty space to add a
20checksum, it will request that you run ``e2fsck -D`` to have the
21directories rebuilt with checksums. This has the added benefit of
22removing slack space from the directory files and rebalancing the htree
23indexes. If you _ignore_ this step, your directories will not be
24protected by a checksum!
25
26The following table describes the data elements that go into each type
27of checksum. The checksum function is whatever the superblock describes
28(crc32c as of October 2013) unless noted otherwise.
29
30.. list-table::
31   :widths: 20 8 50
32   :header-rows: 1
33
34   * - Metadata
35     - Length
36     - Ingredients
37   * - Superblock
38     - __le32
39     - The entire superblock up to the checksum field. The UUID lives inside
40       the superblock.
41   * - MMP
42     - __le32
43     - UUID + the entire MMP block up to the checksum field.
44   * - Extended Attributes
45     - __le32
46     - UUID + the entire extended attribute block. The checksum field is set to
47       zero.
48   * - Directory Entries
49     - __le32
50     - UUID + inode number + inode generation + the directory block up to the
51       fake entry enclosing the checksum field.
52   * - HTREE Nodes
53     - __le32
54     - UUID + inode number + inode generation + all valid extents + HTREE tail.
55       The checksum field is set to zero.
56   * - Extents
57     - __le32
58     - UUID + inode number + inode generation + the entire extent block up to
59       the checksum field.
60   * - Bitmaps
61     - __le32 or __le16
62     - UUID + the entire bitmap. Checksums are stored in the group descriptor,
63       and truncated if the group descriptor size is 32 bytes (i.e. ^64bit)
64   * - Inodes
65     - __le32
66     - UUID + inode number + inode generation + the entire inode. The checksum
67       field is set to zero. Each inode has its own checksum.
68   * - Group Descriptors
69     - __le16
70     - If metadata_csum, then UUID + group number + the entire descriptor;
71       else if gdt_csum, then crc16(UUID + group number + the entire
72       descriptor). In all cases, only the lower 16 bits are stored.
73
74