gfs2_edit - Display, print or edit GFS2 or GFS internal structures.
gfs2_edit [OPTION]... [DEVICE]
The gfs2_edit command is a tool used to examine, edit or display
internal data structures of a GFS2 or GFS file system. The gfs2_edit
command can be run interactively, as described below in INTERACTIVE
Caution: Several options of the gfs2_edit command alter the file system
metadata and can cause file system corruption. These options should be
used with great care.
-p [struct | block] [blocktype] [blockalloc [val]] [blockbits]
[blockrg] [find sb|rg|rb|di|in|lf|jd|lh|ld|ea|ed|lb|13|qc] [field
Print a gfs2 data structure in human-readable format to stdout.
You can enter either a block number or a data structure name.
Block numbers may be specified in hex (e.g., 0x10) or decimal
You can specify the following well-known locations with the -p
sb, superblock - Print the superblock.
root - Print the root directory.
master - Print the master system directory.
jindex - Print the journal index system directory.
per_node - Print the per_node system directory.
inum - Print the system inum file.
statfs - Print the system statfs file.
rindex, rgindex - Print the resource group index system file.
rg X - Print the resource group information for RG X (zero-
rgs - Print the resource group information.
quota - Print the contents of the system quota file.
identify - Identify a data block rather than print the block’s
size - Print the device size information.
journalX - Print the contents of journal X, where X is a journal
number from 0 to <the number of journals in your file system -
1>. Only the journal headers and journal descriptors are
dumped. For journal descriptors, this option prints out every
file system block number logged in that section of the journal.
The actual journaled blocks are not printed.
If you specify a block number rather than a structure name,
gfs2_edit will print out a breakdown of the structure for that
block. For example: gfs2_edit -p sb will print the superblock,
but so does gfs2_edit -p 0x10 and gfs2_edit -p 16.
If you specify -p without a block or structure name, gfs2_edit
prints the superblock.
You can specify more than one data structure with a single -p
option. For example, gfs2_edit -p inum statfs /dev/sda1 prints
the system inum file and the system statfs file on /dev/sda1.
Optionally, you may specify the keyword blocktype to print out
the gfs2 block type for the specified block. Valid gfs2 block
types are: 0 (Clump), 1 (Superblock), 2 (Resource Group Header),
3 (Resource Group Bitmap), 4 (Dinode), 5 (Indirect Block), 6
(Leaf), 7 (Journaled data), 8 (Log Header), 9 (Log descriptor),
10 (Extended attribute), 11 (Eattr Data), 12 (Log Buffer), 13
(Invalid), and 14 (Quota Change).
Optionally, you may specify the keyword blockalloc with an
optional value to assign. If no value is specified, the
blockalloc keyword will print the block allocation type for the
specified block. Valid block allocation types are: 0 (Free
block), 1 (Data block), 2 (Unlinked block), and 3 (Metadata
block). If a value from 0 to 3 is specified, the resource group
bitmap will be changed to the new value. This may be used, for
example, to artificially free or allocate a block in order to
test fsck.gfs2’s ability to detect and fix the problem.
Optionally, you may specify the keyword blockbits. This option
will locate and print the block containing the bitmap
corresponding to the specified block.
Optionally, you may specify the keyword blockrg. This option
will locate and print the block number of the resource group
that holds information about the specified block.
You may also use gfs2_edit to find the next occurrence of a
metadata block of a certain type. Valid metadata types are:
none (unused metadata clump block), sb (superblock), rg
(resource group), rb (rg bitmap), di (disk inode aka dinode), in
(indirect block list), lf (directory leaf), jd (journaled data),
lh (journal log header), ld (journal log descriptor), ea
(extended attribute), ed (ea data block), lb (log buffer), 13
(unused block type 13), qc (quota change). The block AFTER the
one specified with -p is the starting point for the search. For
example, if you specify gfs2_edit -p rg 12 find rg
/dev/your/device, it will find the rg that follows rg 12
(normally, this would be rg 13). Note, however, that since
metadata often appears in the journals, it could be a copy of a
different RG, inside a journal. Also note that gfs2_edit will
only find allocated metadata blocks unless the type specified is
none, sb, rg or rb. In other words, if you try to find a disk
inode, it will only find an allocated dinode, not a deallocated
Optionally, you may specify the keyword field followed by a
valid metadata field name. Right now, only the fields in disk
inodes and resource groups are allowed. If no value is
specified after the field, the value of the field will be
printed to stdout. If a value is specified, the field’s value
will be changed. This may be used, for example, to artificially
change the di_size field for an inode in order to test
fsck.gfs2’s ability to detect and fix the problem.
-s [structure | block]
Specify a starting block for interactive mode. Any of the well-
known locations found in the -p option may be specified. If you
want to start on a particular resource group, specify it in
quotes, e.g. -s "rg 3"
-h, -help, -usage
Print help information.
-c [0 | 1]
Use alternate color scheme for interactive mode: 0=normal (dark
colors on white background), or 1 (light colors on black
-V Print program version information only.
-x Print in hex mode.
rg <rg> <device>
Print the contents of Resource Group <rg> on <device>.
<rg> is a number from 0 to X - 1, where X is the number of RGs.
Print the number of Resource Groups in the file system on
rgflags <rg> [new_value] <device>
Print and/or modify the rg_flags value of Resource Group <rg> on
<rg> is a number from 0 to X - 1, where X is the number of RGs.
If new_value is not specified, the current rg_flags value will
be printed but not modified. If new_value is specified, the
rg_flags field will be overwritten with the new value.
Print off a list of blocks from <filename> that were saved with
the savemeta option.
savemeta <device> <filename>
Save off the GFS2 metadata (not user data) for the file system
on the specified device to a file given by <filename>. You can
use this option to analyze file system problems without
revealing sensitive information that may be contained in the
files. This option works quickly by using the system bitmap
blocks in the resource groups to determine the location of all
the metadata. If there is corruption in the bitmaps, resource
groups or rindex file, this method may fail and you may need to
use the savemetaslow option. The destination file is not
compressed. You may want to compress it with a program such as
bzip2 before sending it for analysis.
savemetaslow <device> <filename>
Save off GFS2 metadata, as with the savemeta option, examining
every block in the file system for metadata. This option is
less prone to failure due to file system corruption than the
savemeta option, but it is extremely slow.
savergs <device> <filename>
Save off only the GFS2 resource group metadata for the file
system on the specified device to a file given by <filename>.
restoremeta <filename> <dest device>
Take a file created with the savemeta option and restores its
contents on top of the specified destination device. WARNING:
When you use this option, the file system and all data on the
destination device is destroyed. Since only metadata (but no
data) is restored, every file in the resulting file system is
likely to be corrupt. The ONLY purpose of this option is to
examine and debug file system problems by restoring and
examining the state of the saved metadata. If the destination
file system is the same size or larger than the source file
system where the metadata was saved, the resulting file system
will be the same size as the source. If the destination device
is smaller than the source file system, gfs2_edit will restore
as much as it can, then quit, leaving you with a file system
that probably will not mount, but from which you might still be
able to figure out what is wrong with the source file system.
If you specify a device on the gfs2_edit command line and you specify
no options other than -c, gfs2_edit will act as an interactive GFS2
file system editor for the file system you specify. There are three
display modes: hex mode, structure mode and pointers mode. You use the
m key to switch between the modes, as described below. The modes are
Hex mode (default)
Display or edit blocks of the file system in hexadecimal and
Lines at the top indicate the currently displayed block in both
hex and decimal. If the block contains a GFS2 data structure,
the name of that structure will appear in the upper right corner
of the display. If the block is a well-known block, such as the
superblock or rindex, there will be a line to indicate what it
In hex mode, you can edit blocks by pressing <enter> and
entering hexadecimal digits to replace the highlighted hex
digits. Do NOT precede the numbers with "0x". For example, if
you want to change the value at offset 0x60 from a 0x12 to 0xef,
position your cursor to offset 0x60, so that the 12 is
highlighted, then press <enter> and type in "ef". Press
<escape> or <enter> to exit edit mode.
In hex mode, different colors indicate different things. For
example, in the default color scheme, the GFS2 data structure
will be black, data offsets will be light blue, and actual data
(anything after the gfs2 data structure) will be red.
Decode the file system block into its GFS2 structure and display
the values of that structure. This mode is most useful for
jumping around the file system. For example, you can use the
arrow keys to position down to a pointer and press J to jump to
Display any additional information appearing on the block. For
example, if an inode has block pointers, this will display them
and allow you to scroll through them. You can also position to
one of them and press J to jump to that block.
Interactive mode command keys:
q or <esc>
The q or <escape> keys are used to exit gfs2_edit.
<arrow/movement keys> up, down, right, left, pg-up, pg-down, home, end
The arrow keys are used to highlight an area of the display.
The J key may be used to jump to the block that is highlighted.
m - Mode switch
The m key is used to switch between the three display modes.
The initial mode is hex mode. Pressing the m key once switches
to structure mode. Pressing it a second time switches from
structure mode to pointers mode. Pressing it a third time takes
you back to hex mode again.
j - Jump to block
The j key jumps to the block number that is currently
highlighted. In hex mode, hitting J will work when any byte of
the pointer is highlighted.
g - Goto block
The g key asks for a block number, then jumps there. Note that
in many cases, you can also arrow up so that the current block
number is highlighted, then press <enter> to enter a block
number to jump to.
h - Help display
The h key causes the interactive help display to be shown.
e - Extended mode
The e key causes gfs2_edit to switch to extended ("pointers")
c - Color scheme
The c key causes gfs2_edit to switch to its alternate color
f - Forward block
The f key causes you to scroll forward one block. This does not
affect the "jump" status. In other words, if you use the f key
to move forward several blocks, pressing <backspace> will not
roll you back up.
<enter> - Edit value
The <enter> key causes you to go from display mode to edit mode.
If you are in hex mode and you hit enter, you can type new hex
values at the cursor’s current location. Note: pressing <enter>
in structure mode allows you to enter a new value, with the
following restrictions: For gfs2 disk inodes and resource
groups, it will actually change the value on disk. However,
inode numbers may not be changed. For all other structures, the
values entered are ignored.
If you use the up arrow key to highlight the block number, then
press <enter>, you may then enter a new block number, or any of
the well-known block locations listed above (e.g. sb, rindex,
inum, rg 17, etc.) and gfs2_edit will jump to the block
specified. If you specify a slash character followed by a
metadata type, gfs2_edit will search for the next occurrence of
that metadata block type, and jump there. It will take you to
block 0 if it does not find any more blocks of the specified
<home> If you are in pointers mode, this takes you back to the starts
of the pointers you are viewing. Otherwise it takes you back to
This takes you back to the block you were displaying before a
This takes you forward to the block you were displaying when you
Display and optionally edit the file system on
gfs2_edit -p sb /dev/vg0/lvol0
Print the superblock of the gfs2 file system located on
gfs2_edit -p identify 2746 2748 /dev/sda2
Print out what kind of blocks are at block numbers 2746 and 2748
on device /dev/sda2.
gfs2_edit -p rindex /dev/sda1
Print the resource group index system file located on device
gfs2_edit savemeta /dev/sda1 /tmp/our_fs
Save off all metadata (but no user data) to file /tmp/our_fs.
gfs2_edit -p root /dev/my_vg/my_lv
Print the contents of the root directory in /dev/my_vg/my_lv.
gfs2-edit -x -p 0x3f7a /dev/sda1
Print the contents of block 16250 of /dev/sda1 in hex.
gfs2_edit -p 12345 /dev/sdc2
Print the gfs2 data structure at block 12345.
gfs2_edit rgcount /dev/sdb1
Print how many Resource Groups exist for /dev/sdb1.
gfs2_edit -p rg 17 /dev/sdb1
Print the contents of the eighteenth Resource Group on
gfs2_edit rgflags 3 /dev/sdb1
Print the rg_flags value for the fourth Resource Group on
gfs2_edit rgflags 3 8 /dev/sdb1
Set the GFS2_RGF_NOALLOC flag on for the fourth Resource Group
gfs2_edit -p 25 blockalloc /dev/roth_vg/roth_lv
Print the block allocation type of block 25. May produce this
output: 3 (Metadata)
gfs2_edit -p 25 blockalloc 1 /dev/roth_vg/roth_lv
Change the block allocation type of block 25 to data. May
produce this output: 1
gfs2_edit -p 25 blocktype /dev/roth_vg/roth_lv
Print the metadata block type of block 25. May produce this
output: 4 (Block 25 is type 4: Dinode)
gfs2_edit -p 25 field di_size /dev/roth_vg/roth_lv
Print the di_size field of block 25. May produce this output:
gfs2_edit -x -p 25 field di_size /dev/roth_vg/roth_lv
Print the di_size field of block 25, in hexidecimal. May
produce this output: 0x8000000
gfs2_edit -p 25 field di_size 0x4000 /dev/roth_vg/roth_lv
Change the di_size field of block 25 to the hexidecimal value
0x4000. May produce this output: 16384
The directory code does not work well. It might be confused
by directory "sentinel" entries.