BTRFS-INSPECT-INTERNAL(8)

💅 BTRFS-INSPECT-INTERNAL(8) — query various internal information 🦋

okay so bestie, this is the manual page for btrfs-inspect-internal, sitting in section 8 (that's the sysadmin/system-management girlies, iykyk ✨). Strap in, it's a long one and it's lowkey iconic.

🌸 NAME 🌸

btrfs-inspect-internal — query various internal information

✨ SYNOPSIS ✨

btrfs inspect-internal <subcommand> <args>

💖 DESCRIPTION 💖

This command group is basically the backstage pass to your filesystem's ✨internal tea✨. The functionality ranges from a simple UI, all the way up to an ioctl or a more complex query that assembles the result from several internal structures. And that fancier stuff? Yeah, it usually requires calls to privileged ioctls, so bring root energy. 👑

🎀 SUBCOMMAND 🎀

`dump-super [options] <device> [device...]`

Show btrfs superblock information stored on the given devices, in textual form, no cap. By default the first superblock is printed — but if you want more details about all copies or additional backup data, that can be printed too. 🫶

Besides verification of the filesystem signature, there are no other sanity checks. The superblock checksum status gets reported, and the device item and filesystem UUIDs are checked and reported.

⚠️ Note: The meaning of option -s changed in version 4.8 to be consistent with other tools — it specifies the superblock copy, rather than the offset. The old way still works, but it prints a warning. Please update your scripts to use --bytenr instead. The option -i has been deprecated. 💀

Options

-f, --full — print full superblock information, including the system chunk array and backup roots
-a, --all — print information about all present superblock copies (cannot be used together with the -s option)
-i <super> — (deprecated since 4.8, same behaviour as --super)
--bytenr <bytenr> — specify offset to a superblock in a non-standard location at bytenr, useful for debugging (disables the -f option)

If there are multiple options specified, only the last one applies.

-F, --force — attempt to print the superblock even if a valid BTRFS signature is not found; the result may be completely wrong if the data does not resemble a superblock 😭
-s, --super <bytenr> — (see compatibility note above)

specify which mirror to print, valid values are 0, 1 and 2, and the superblock must be present on the device with a valid signature, can be used together with --force

`dump-tree [options] <device> [device...]`

Dump tree structures from a given device in textual form, expanding keys to human readable equivalents where possible. This is genuinely so useful for analyzing filesystem state or inconsistencies, and it has a positive educational effect on understanding the internal filesystem structure. It's giving "learn by peeking behind the curtain". 💅

🙈 Note: By default this contains file names, so keep in mind that if you're asked to send the dump for analysis, use --hide-names eventually. It does not contain file data.

Special characters in file names, xattr names and values are escaped, in the C style like \n and octal encoding \NNN.

Options

-e, --extents — print only extent-related information: extent and device trees
-d, --device — print only device-related information: tree root, chunk and device trees
-r, --roots — print only short root node information, i.e. the root tree keys
-R, --backups — same as --roots plus print backup root info, i.e. the backup root keys and the respective tree root block offset
-u, --uuid — print only the uuid tree information, empty output if the tree does not exist
-b <block_num> — print info of the specified block only, can be specified multiple times
--follow — use with -b, print all children tree blocks of <block_num>
--dfs — (default up to 5.2)

use depth-first search to print trees, the nodes and leaves are intermixed in the output

--bfs — (default since 5.3)

use breadth-first search to print trees, the nodes are printed before all leaves

--hide-names — print a placeholder HIDDEN instead of various names, useful for developers to inspect the dump while keeping potentially sensitive information hidden

This is:

directory entries (files, directories, subvolumes)
default subvolume
extended attributes (name, value)
hardlink names (if stored inside another item or as extended references in standalone items)

Note: Lengths are not hidden because they can be calculated from the item size anyway.

--csum-headers — print b-tree node checksums stored in headers (metadata)
--csum-items — print checksums stored in checksum items (data)
--noscan — do not automatically scan the system for other devices from the same filesystem, only use the devices provided as the arguments
-t <tree_id> — print only the tree with the specified ID, where the ID can be numerical or a common name in a flexible human readable form

The tree id name recognition rules:

case does not matter
the C source definition, e.g. BTRFS_ROOT_TREE_OBJECTID
short forms without BTRFS_ prefix, without _TREE and _OBJECTID suffix, e.g. ROOT_TREE, ROOT
convenience aliases, e.g. DEVICE for the DEV tree, CHECKSUM for CSUM
unrecognized ID is an error

`inode-resolve [-v] <ino> <path>`

(needs root privileges 👑)

Resolve paths to all files with given inode number ino in a given subvolume at path, i.e. all hardlinks.

Options

-v — (deprecated) alias for global -v option

`logical-resolve [-Pvo] [-s <bufsize>] <logical> <path>`

(needs root privileges 👑)

Resolve paths to all files at given logical address in the linear filesystem space.

Options

-P — skip the path resolving and print the inodes instead
-o — ignore offsets, find all references to an extent instead of a single block. Requires kernel support for the V2 ioctl (added in 4.15). The results might need further processing to filter out unwanted extents by the offset that is supposed to be obtained by other means.
-s <bufsize> — set internal buffer for storing the file names to bufsize, default is 64KiB, maximum 16MiB. Buffer sizes over 64KiB require kernel support for the V2 ioctl (added in 4.15).
-v — (deprecated) alias for global -v option

`list-chunks [options] <path>`

(needs root privileges 👑)

Enumerate chunks on all devices. The chunks represent the physical range on devices — not to be confused with block groups that represent the logical ranges, though bestie, the terms are often used interchangeably. 🙈

Example output:

Devid PNumber      Type/profile    PStart    Length      PEnd LNumber    LStart Usage%
----- ------- ----------------- --------- --------- --------- ------- --------- ------
    1       1       Data/single   1.00MiB  84.00MiB  85.00MiB      68 191.60GiB  62.77
    1       2     System/DUP     85.00MiB  32.00MiB 117.00MiB      39 140.17GiB   0.05
    1       3     System/DUP    117.00MiB  32.00MiB 149.00MiB      40 140.17GiB   0.05
    1       4   Metadata/DUP    149.00MiB 192.00MiB 341.00MiB      59 188.41GiB  45.00
    1       5   Metadata/DUP    341.00MiB 192.00MiB 533.00MiB      60 188.41GiB  45.00
    1       6       Data/single 533.00MiB   1.00GiB   1.52GiB      49 169.91GiB  72.23
    1       7       Data/single   1.52GiB  16.00MiB   1.54GiB      69 191.68GiB  79.83
    1       8       Data/single   1.54GiB   1.00GiB   2.54GiB      17 100.90GiB  46.39
    1       9       Data/single   2.54GiB   1.00GiB   3.54GiB      16  99.90GiB  40.68
    1      10       Data/single   3.54GiB   1.00GiB   4.54GiB       1  71.40GiB  62.97
    1      11       Data/single   4.54GiB   1.00GiB   5.54GiB      33 125.04GiB  26.00
    1      12       Data/single   5.54GiB   1.00GiB   6.54GiB      50 170.91GiB  60.44
    1      13       Data/single   6.54GiB 512.00MiB   7.04GiB      63 189.16GiB  67.34
    1      14       Data/single   7.04GiB   1.00GiB   8.04GiB      51 171.91GiB  70.94

Devid — the device id
PNumber — the number of the chunk on the device (in order)
Type/profile — the chunk type and profile
PStart — the chunk start on the device
Length — the chunk length (same for physical and logical address space)
PEnd — the chunk end, effectively PStart + Length
LNumber — the number of the chunk, in the logical address space of the whole filesystem
LStart — the chunk start in the logical address space of the whole filesystem, as it's a single space it's also called offset
Usage — chunk usage, percentage of used data/metadata of the chunk length

The chunks in the output can be sorted by one or more sorting criteria, evaluated as specified, in the ascending order (or descending when used like -devid,-pstart). By default the chunks are sorted by devid and pstart, this is most convenient for single device filesystems. 🫶

On multi-device filesystems it's up to the user what is preferred, as the layout of chunks on e.g. striped profiles (RAID0 etc) cannot be easily represented. A logical view with corresponding underlying structure would be better, but sorting by lstart,devid at least groups devices of the given logical range. Can be also combined with usage.

This output can provide information for balance filters. ✨

Options

--sort MODE — sort by a column (ascending, prepend - for descending):

MODE is a comma separated list of:

devid - by device id (default, with pstart)

pstart — physical start, like relative to the beginning of the device 🫶
- lstart — logical offset, aka where it lives in the logical address space
- usage — sort by how much of the chunk is actually used, in percent
- length — sort by how long the chunk is
- type — sort by chunk type (data, metadata, system, bestie, we love options)
- profile — sort by chunk profile (single, DUP, RAID0, RAID1, ...)
--raw — raw numbers in bytes, and nope, no B suffix on these ✨
--human-readable — prints numbers that a human can actually read, base 1024, and this is the default so you already got it 💅
--iec — locks in the 1024 base for the options below, following the IEC standard 📏
--si — locks in the 1000 base for the options below, following the SI standard ⭐
--kbytes — show sizes in KiB, or kB if you're running --si
--mbytes — show sizes in MiB, or MB with --si
--gbytes — show sizes in GiB, or GB with --si
--tbytes — show sizes in TiB, or TB with --si

🌈 `map-swapfile [options] <file>` 🌈

needs root privileges, so like, bring your sudo energy 👑

This one finds the device-specific physical offset of file — the one you can use for hibernation. It also checks that the file is actually suitable as a swapfile, because we do not do vibes-based hibernation here. Also peep btrfs filesystem mkswapfile and the Swapfile feature <> description for the lore 🦋

Note: bestie do NOT use filefrag or FIEMAP ioctl values that get reported as physical. Those are different because of internal filesystem mappings, fr fr. Hibernation wants the offset relative to the physical block device, no cap 💀

Options 🎀

-r|--resume-offset — print only the value that's suitable as a resume offset for the file /sys/power/resume_offset 🔥

🌸 `min-dev-size [options] <path>` 🌸

needs root privileges, yes queen, sudo again 🫶

This returns the minimum size the device can be shrunk to — and it does it without performing any resize operation. So it's giving "preview mode". Super useful before you go execute the actual resize, because measuring twice and cutting once is the whole mood ✨

Options 🎀

--id <id> — specify the device id to query. Default is 1 if you don't use this option 💅

🦋 `rootid <path>` 🦋

For a given file or directory, this hands you the containing tree root id. But if you point it at a subvolume itself, it returns its own tree id — i.e. the subvol id. Main character behavior, honestly 👑

Note: the result is undefined for the so-called empty subvolumes (the ones identified by inode number 2), but lowkey it's fine because a subvolume like that doesn't contain any files anyway 🙈

💜 `subvolid-resolve <subvolid> <path>` 💜

needs root privileges 💅

Resolves the absolute path of the subvolume id subvolid. It's literally a glow-up from a number to a path ✨

⭐ `tree-stats [options] <device>` ⭐

needs root privileges 👑

Prints sizes and statistics of trees. Heads up though: this takes a device as an argument, not a mount point like the other commands do. Different rules for this one, iykyk 🌀

Note: if the filesystem is still mounted, you can run the command, but the results may be inaccurate, or various errors may get printed if there are ongoing writes to the filesystem. In that case a warning gets printed. Consider yourself warned, bestie 😭

Options 🎀

-b|--raw — raw numbers in bytes, without the B suffix
-t <tree_id> — print only the tree with the specified ID, where the ID can be numerical or a common name in a flexible, human readable form 🌈

The tree id name recognition rules, and yes there are rules:

case does not matter
the C source definition, e.g. BTRFS_ROOT_TREE_OBJECTID
short forms without the BTRFS_ prefix, without the _TREE and _OBJECTID suffix, e.g. ROOT_TREE, ROOT
convenience aliases, e.g. DEVICE for the DEV tree, CHECKSUM for CSUM
an unrecognized ID is an error 💀

--human-readable — print human friendly numbers, base 1024, and yes this is the default, iconic behavior 💅
--iec — select the 1024 base for the following options, according to the IEC standard 📏
--si — select the 1000 base for the following options, according to the SI standard ⭐
--kbytes — show sizes in KiB, or kB with --si
--mbytes — show sizes in MiB, or MB with --si
--gbytes — show sizes in GiB, or GB with --si
--tbytes — show sizes in TiB, or TB with --si

✨ EXIT STATUS ✨

btrfs inspect-internal returns a zero exit status if it succeeds. Non zero is returned in case of failure. Zero means slay, non-zero means something went sideways 😭

🌈 AVAILABILITY 🌈

btrfs is part of btrfs-progs. Please refer to the documentation at <https://btrfs.readthedocs.io>. Go read the docs bestie, they're right there 🫶

👑 SEE ALSO 👑

mkfs.btrfs(8) <>

That's the whole page, bestie!! You made it to the end, so proud of you 💖 This was BTRFS-INSPECT-INTERNAL(8) from btrfs-progs version 7.3, dated Oct 24, 2026 — go forth and inspect those internals responsibly ✨🎀🦋

📦 where this page comes from

arch rolling: btrfs-progs 7.3-1 · GPL-2.0-only
💖 girlified by deepseek-ai/DeepSeek-V4.1-Flash on 2026-11-03: a language model's rewrite of the original, under its terms

BTRFS-INSPECT-INTERNAL(8)