15 January 2011

Recreate a Linux EXT3 FS

We've looked at Solaris and FreeBSD, now it's time to look at recreating
a Linux EXT3 filesystem.  Similar to the previous two, if you customized
the options to 'mke2fs / mkfs', 'mke2fs -n' is not the answer since it
will only produce data about generic FS creation.  Our host details this
time are:
        HOST:                   tux
        PROMPT:                 tux [0]
        OS:                     CentOS 5.4 Linux
        DISK PARTITION:         sdc1
As previously, we start off with 'mke2fs' to create a generic ext3
filesystem on 'sdc1':
        tux [0] /sbin/mke2fs -j /dev/sdc1
        mke2fs 1.39 (29-May-2006)
        Filesystem label=
        OS type: Linux
        Block size=1024 (log=0)
        Fragment size=1024 (log=0)
        131072 inodes, 524284 blocks
        26214 blocks (5.00%) reserved for the super user
        First data block=1
        Maximum filesystem blocks=67633152
        64 block groups
        8192 blocks per group, 8192 fragments per group
        2048 inodes per group
        Superblock backups stored on blocks:
                8193, 24577, 40961, 57345, 73729, 204801, 221185, 401409

        Writing inode tables: done
        Creating journal (8192 blocks): done
        Writing superblocks and filesystem accounting information: done

        This filesystem will be automatically checked every 27 mounts or
        180 days, whichever comes first.  Use tune2fs -c or -i to override.
The output from 'mke2fs -n' returns nearly identical output from the
'mke2fs' used to create the FS.  As an aside, since we already know the
FS is ext3 rather than ext2, we've include '-j' to 'mke2fs' to address
the journal:
        tux [0] /sbin/mke2fs -n -j /dev/sdc1
        mke2fs 1.39 (29-May-2006)
        Filesystem label=
        OS type: Linux
        Block size=1024 (log=0)
        Fragment size=1024 (log=0)
        131072 inodes, 524284 blocks
        26214 blocks (5.00%) reserved for the super user
        First data block=1
        Maximum filesystem blocks=67633152
        64 block groups
        8192 blocks per group, 8192 fragments per group
        2048 inodes per group
        Superblock backups stored on blocks:
                8193, 24577, 40961, 57345, 73729, 204801, 221185, 401409
Alright, let's muck with some of the parameters available to 'mke2fs'.
After '-v -j', we set the FS revision, number of inodes, block size,
number of bytes per inode (nbpi), journal size, percent of reserved
blocks, and a volume label:
        tux [0] /sbin/mke2fs -v -j -r 1 -N 65536 -b 4096 -i 2048 -J size=16 -m 3 -L c1vol /dev/sdc1
        mke2fs 1.39 (29-May-2006)
        Filesystem label=c1vol
        OS type: Linux
        Block size=4096 (log=2)
        Fragment size=4096 (log=2)
        65536 inodes, 131071 blocks
        3932 blocks (3.00%) reserved for the super user
        First data block=0
        Maximum filesystem blocks=134217728
        4 block groups
        32768 blocks per group, 32768 fragments per group
        16384 inodes per group
        Superblock backups stored on blocks:
                32768, 98304

        Writing inode tables: done
        Creating journal (4096 blocks): done
        Writing superblocks and filesystem accounting information: done

        This filesystem will be automatically checked every 36 mounts or
        180 days, whichever comes first.  Use tune2fs -c or -i to override.
Back to 'mke2fs -n' to compare the output:
        tux [0] /sbin/mke2fs -nv /dev/sdc1
        mke2fs 1.39 (29-May-2006)
        Filesystem label=
        OS type: Linux
        Block size=1024 (log=0)
        Fragment size=1024 (log=0)
        131072 inodes, 524284 blocks
        26214 blocks (5.00%) reserved for the super user
        First data block=1
        Maximum filesystem blocks=67633152
        64 block groups
        8192 blocks per group, 8192 fragments per group
        2048 inodes per group
        Superblock backups stored on blocks:
                8193, 24577, 40961, 57345, 73729, 204801, 221185, 401409
Without knowing the original parameters to 'mke2fs', the output from 'mke2fs
-n' is effectively useless.  Turning to 'dumpe2fs', we can get all of
the parameters necessary to recreate our FS:
        tux [0] /sbin/dumpe2fs -h /dev/sdc1
        dumpe2fs 1.39 (29-May-2006)
        Filesystem volume name:   c1vol
        Last mounted on:          <not available>
        Filesystem UUID:          f4b38089-b48f-4741-a4c8-bec7e1bfdcc8
        Filesystem magic number:  0xEF53
        Filesystem revision #:    1 (dynamic)
        Filesystem features:      has_journal resize_inode dir_index filetype sparse_super large_file
        Default mount options:    (none)
        Filesystem state:         clean
        Errors behavior:          Continue
        Filesystem OS type:       Linux
        Inode count:              65536
        Block count:              131071
        Reserved block count:     3932
        Free blocks:              124808
        Free inodes:              65525
        First block:              0
        Block size:               4096
        Fragment size:            4096
        Reserved GDT blocks:      31
        Blocks per group:         32768
        Fragments per group:      32768
        Inodes per group:         16384
        Inode blocks per group:   512
        Filesystem created:       Sat Jan 15 15:29:38 2011
        Last mount time:          n/a
        Last write time:          Sat Jan 15 15:29:39 2011
        Mount count:              0
        Maximum mount count:      36
        Last checked:             Sat Jan 15 15:29:38 2011
        Check interval:           15552000 (6 months)
        Next check after:         Thu Jul 14 16:29:38 2011
        Reserved blocks uid:      0 (user root)
        Reserved blocks gid:      0 (group root)
        First inode:              11
        Inode size:               128
        Journal inode:            8
        Default directory hash:   tea
        Directory Hash Seed:      0b0c0827-6864-4320-b643-cf3e113f05e6
        Journal backup:           inode blocks
        Journal size:             16M
Of note, the output from 'dumpe2fs -h' is identical to 'tune2fs -l' with
the exception that 'dumpe2fs' includes the journal size.  Moving along,
output from 'dumpe2fs -h' correlates to the following 'mke2fs' parameters:
        ** dumpe2fs **                  ** mke2fs **
        Filesystem volume name:         -L volume_label
        Filesystem revision #:          -r value
        Inode count:                    -N value
        Reserved block count:           -m % (obtainable with equation 1)
        Block size:                     -b bytes
        Inode size:                     -i bytes (obtainable with equation 2)
        Journal size:                   -J size=MB
Equation 1: To determine the percentage of reserved blocks, we can use
the following, rounding to the nearest hundredth:
        % reserved blocks = (reserved block count / block count)
                          = (3932 / 131071)
                          = 0.029
        % reserved blocks = 3%
Equation 2: To determine the nbpi, we can use the following, rounding
to the nearest 1024 increment:
        nbpi (-i) = (block count / inode size)
                  = (131071 / 128)
                  = 1023.99
        nbpi (-i) = 1024
So, to put it all together, we get:
        tux [0] /sbin/mke2fs -nv -j -r 1 -N 65536 -b 4096 -i 1024 -J size=16 -m 3 -L c1vol /dev/sdc1
        mke2fs 1.39 (29-May-2006)
        Filesystem label=c1vol
        OS type: Linux
        Block size=4096 (log=2)   
        Fragment size=4096 (log=2)
        65536 inodes, 131071 blocks
        3932 blocks (3.00%) reserved for the super user
        First data block=0
        Maximum filesystem blocks=134217728
        4 block groups
        32768 blocks per group, 32768 fragments per group
        16384 inodes per group
        Superblock backups stored on blocks:
                32768, 98304
Without the '-n' option to 'mke2fs', the new filesystem would have
been created to the specs of the original.  Of the reasons to use '-n'
with our other options, we can determine the appropriate location of
the backup superblocks.

Included for reference, the following details the correlation of 'tune2fs'
and 'mke2fs' parameters to the output of dumpe2fs:

tune2fsmke2fsdumpe2fs
-L vollabe-L vollabelFilesystem volume name:
-M dir-M dirLast mounted on:
Filesystem UUID:
Filesystem magic number:
-r revision [0|(1)]Filesystem revision #:
-O feat,..-O feature,...,...Filesystem features:
-o opt,...Default mount options:
Filesystem state:
-e valErrors Behavior:
-o creator-osFilesystem OS type:
-N valInode count:
Block count:
-r coun-m %Reserved block count:
Free blocks:
Free inodes:
First block:
-b bytesBlock size:
-f bytes (ignored)Fragment size:
Reserved GDT blocks:
-g valBlocks per group:
Fragments per group:
Inodes per group:
Inode blocks per group:
Filesystem created:
Last mount time:
Last write time:
-C valMount count:
-c valMaximum mount count:
-T valLast checked:
Next check after:
-u uidReserved blocks uid:
-g gidReserved blocks gid:
First inode:
-i bytesInode size:
Journal inode:
Default directory hash:
Directory Hash Seed:
Journal backup:
-j size=M-J size=MBJournal size:

see also:
    Recreate a Solaris UFS FS
    Recreate a FreeBSD UFS FS

No comments: