FreeBSD. Just as with Solaris, when trying to determine how a filesystem
was created, 'newfs -N /dev/da2s1a' is not the answer, as it only produces
information relevant to a default FS creation. Our host details for this
scenario are:
HOST: beastie PROMPT: beastie [0] OS: FreeBSD 8.1 DISK PARTITION: da2s1aJust as with Solaris, we start off with 'newfs' to create a generic
filesystem on 'da2s1a':
beastie [0] /sbin/newfs /dev/da2s1a /dev/da2s1a: 512.0MB (1048528 sectors) block size 16384, fragment size 2048 using 4 cylinder groups of 128.00MB, 8192 blks, 16384 inodes. super-block backups (for fsck -b #) at: 160, 262304, 524448, 786592Now let's see what 'newfs -N' returns:
beastie [0] /sbin/newfs -N /dev/da2s1a /dev/da2s1a: 512.0MB (1048528 sectors) block size 16384, fragment size 2048 using 4 cylinder groups of 128.00MB, 8192 blks, 16384 inodes. super-block backups (for fsck -b #) at: 160, 262304, 524448, 786592The output is identical to the first 'newfs' command. The only difference
is that passing '-N' to 'newfs' tells 'newfs' to only print what would
be done and not create an FS. The output also doesn't give us much
to work with to recreate the FS. To get the FS creation parameters,
we need to look at 'dumpfs':
beastie [0] /sbin/dumpfs -m /dev/da2s1a # newfs command for /dev/da2s1a (/dev/da2s1a) newfs -O 2 -a 8 -b 16384 -d 16384 -e 2048 -f 2048 -g 16384 -h 64 -m 8\ -o time -s 1048528 /dev/da2s1a(The 'newfs' command shown above is one line though I've broken it up
(here and in the following details) over two for legibility, hence the
'\'.) Let's see if 'dumpfs' is actually giving up true parameters for
'newfs'. Below, I've mostly taken the options passed in a generic FS
creation and tweaked them, removing the size attribute (-s), and adding
a volume label (-L) and the number of bytes per inode (nbpi (-i)):
beastie [0] /sbin/newfs -O2 -a 4 -b 4096 -d 4096 -e 1024 -f 512 -g 8192\ -h 128 -m 2 -i 262144 -L testvol -o space /dev/da2s1a /dev/da2s1a: 512.0MB (1048528 sectors) block size 4096, fragment size 512 using 38 cylinder groups of 13.53MB, 3463 blks, 64 inodes. super-block backups (for fsck -b #) at: 144, 27848, 55552, 83256, 110960, 138664, 166368, 194072, 221776, 249480, 277184, 304888, 332592, 360296, 388000, 415704, 443408, 471112, 498816, 526520, 554224, 581928, 609632, 637336, 665040, 692744, 720448, 748152, 775856, 803560, 831264, 858968, 886672, 914376, 942080, 969784, 997488, 1025192Going back to 'newfs -N', there's no output for a suitable rebuild
command and the output is back to the same as for a generic FS:
beastie [0] /sbin/newfs -N /dev/da2s1a /dev/da2s1a: 512.0MB (1048528 sectors) block size 16384, fragment size 2048 using 4 cylinder groups of 128.00MB, 8192 blks, 16384 inodes. super-block backups (for fsck -b #) at: 160, 262304, 524448, 786592At this point, 'newfs -N' is definitely the wrong answer to determine
how a filesystem was created. Back to 'dumpfs', we get back nearly all
of the appropriate 'newfs' parameters to recreate the FS:
beastie [0] /sbin/dumpfs -m /dev/da2s1a # newfs command for /dev/da2s1a (/dev/da2s1a) newfs -L testvol -O 2 -a 4 -b 4096 -d 4096 -e 1024 -f 512 -g 8192 -h 128\ -m 2 -o space -s 1048528 /dev/da2s1aThose with a keen eye will notice that the 'dumpfs' devised 'newfs'
command is critically missing '-i' setting the inode density. If we ran
the above command without the appropriate value to '-i', we would have
new values for our superblocks, among other things. Assuming that you
set '-i' (nbpi) following the notes in the 'newfs' manpage, we'll need
to see the first 18 lines of 'dumpfs' output:
beastie [0] /sbin/dumpfs /dev/da2s1a | /usr/bin/head -18 magic 19540119 (UFS2) time Thu Jan 13 18:03:00 2011 superblock location 65536 id [ 4d2f84a4 a6d95b30 ] ncg 38 size 1048528 blocks 1046254 bsize 4096 shift 12 mask 0xfffff000 fsize 512 shift 9 mask 0xfffffe00 frag 8 shift 3 fsbtodb 0 minfree 2% optim space symlinklen 120 maxbsize 4096 maxbpg 1024 maxcontig 4 contigsumsize 4 nbfree 130779 ndir 2 nifree 2428 nffree 20 bpg 3463 fpg 27704 ipg 64 unrefs 0 nindir 512 inopb 16 maxfilesize 550831702015 sbsize 1536 cgsize 4096 csaddr 200 cssize 1024 sblkno 144 cblkno 160 iblkno 168 dblkno 200 cgrotor 0 fmod 0 ronly 0 clean 1 avgfpdir 128 avgfilesize 8192 flags none fsmnt volname testvol swuid 0Again, assuming that you used the logic for '-i' from the manpage,
we can get the nbpi value using the following equation:
nbpi = (size * fsize) / (4 * fsize) (1048528 * 512) / (4 * 512) 536846336 / 2048 nbpi = 262132Due to constraints such as disk geometry and our other parameters,
'newfs' wasn't able to use our originally specified nbpi (-i) value
of "262144". To meet those constraints, 'newfs' adjusted the nbpi to
"262132". If we now take the earlier 'newfs' output from 'dumpfs -m',
adding the derived 'nbpi' value, we can recreate the FS:
beastie [0] newfs -i 262132 -L testvol -O 2 -a 4 -b 4096 -d 4096 -e 1024\ -f 512 -g 8192 -h 128 -m 2 -o space -s 1048528 /dev/da2s1a /dev/da2s1a: 512.0MB (1048528 sectors) block size 4096, fragment size 512 using 38 cylinder groups of 13.53MB, 3463 blks, 64 inodes. super-block backups (for fsck -b #) at: 144, 27848, 55552, 83256, 110960, 138664, 166368, 194072, 221776, 249480, 277184, 304888, 332592, 360296, 388000, 415704, 443408, 471112, 498816, 526520, 554224, 581928, 609632, 637336, 665040, 692744, 720448, 748152, 775856, 803560, 831264, 858968, 886672, 914376, 942080, 969784, 997488, 1025192
see also:
Recreate a Solaris UFS FS
Recreate a Linux EXT3 FS
