> But can you (or perhaps Bero) give a better explanation as to how you
> determined the partition sizes? You might want to remind us what info you
> had to start with etc.
>
> I'd really like to understand this.
OK, between the two of us looking at stuff, this is the jist of what happened:
The dumpe2fs would find the superblocks in the first file system without a
hitch. But it wouldn't find them in the 2nd filesystem. Bero wrote a
little program that would call up dumpe2fs and hand it a block to look at.
After a few clues and playing with dd and od, I wrote the following shell
line to look for the superblock magic number:
dd bs=1k if=/dev/hdb2 | od -x -A x | grep "ef53 000"
It will produce lines like this on a working system:
000430 c8c0 38bb 0010 0014 ef53 0003 0001 0000
800430 0c31 38b8 0000 0014 ef53 0000 0001 0000
1000430 0c31 38b8 0000 0014 ef53 0000 0001 0000
1800430 0c31 38b8 0000 0014 ef53 0000 0001 0000
2000430 0c31 38b8 0000 0014 ef53 0000 0001 0000
2800430 0c31 38b8 0000 0014 ef53 0000 0001 0000
3000430 0c31 38b8 0000 0014 ef53 0000 0001 0000
The above is from the first file system. Notice that the main superblock
is ef53 0003 for the bytes. The backup blocks are ef53 0000.
But, this is what I was getting for the bad filesystem:
0c5230 090b 38b9 0002 0014 ef53 0001 0001 0000
8c5230 e2e9 38b8 0000 0014 ef53 0000 0001 0000
10c5230 e2e9 38b8 0000 0014 ef53 0000 0001 0000
18c5230 e2e9 38b8 0000 0014 ef53 0000 0001 0000
20c5230 e2e9 38b8 0000 0014 ef53 0000 0001 0000
28c5230 e2e9 38b8 0000 0014 ef53 0000 0001 0000
The main superblock is ef53 0001. But notice the address. It was too far
into the partition. i.e., it wasn't on cylinder boundaries.
It was then that Bero thought about the workings of fdisk and put it into
sector display mode. Using dumpe2fs, one can determine what the size is of
the 1st filesystem:
dumpe2fs -h /dev/hdb1
dumpe2fs 1.18, 11-Nov-1999 for EXT2 FS 0.5b, 95/08/09
Filesystem volume name: ftp
Last mounted on: <not available>
Filesystem UUID: 2bc0e64e-ab54-11d3-9f89-e246f22a98c5
Filesystem magic number: 0xEF53
Filesystem revision #: 0 (original)
Filesystem features: (none)
Filesystem state: not clean with errors
Errors behavior: Continue
Filesystem OS type: Linux
Inode count: 623104
Block count: 9969088
Reserved block count: 498454
Free blocks: 4893485
Free inodes: 622202
First block: 1
Block size: 1024
Fragment size: 1024
Blocks per group: 8192
Fragments per group: 8192
Inodes per group: 512
Inode blocks per group: 64
Last mount time: Tue Feb 29 10:16:54 2000
Last write time: Tue Feb 29 10:16:56 2000
Mount count: 19
Maximum mount count: 20
Last checked: Sat Feb 26 11:24:01 2000
Check interval: 15552000 (6 months)
Next check after: Thu Aug 24 12:24:01 2000
Reserved blocks uid: 0 (user root)
Reserved blocks gid: 0 (group root)
Notice the block count. Here is the partition table as it looks repaired:
Disk /dev/hdb: 255 heads, 63 sectors, 2482 cylinders
Units = cylinders of 16065 * 512 bytes
Device Boot Start End Blocks Id System
/dev/hdb1 1 1242 9969088+ 83 Linux
/dev/hdb2 1242 2482 9967545 83 Linux
To get those exact block in, you use "u" to change the display mode:
Disk /dev/hdb: 255 heads, 63 sectors, 2482 cylinders
Units = sectors of 1 * 512 bytes
Device Boot Start End Blocks Id System
/dev/hdb1 63 19938239 9969088+ 83 Linux
/dev/hdb2 19938240 39873329 9967545 83 Linux
Now you can create the partitions by entering the sectors for the ending,
which is block-count*2+63. The first 63 sectors contain the partition table
and boot strap.
Then all he had to do was create the 2nd filesystem starting from where the
first one left off.
Since I had created this a little whiel ago, I forgot that I didn't enter
cylinders, but a size value. Fdisk, it seems, doesn't round to the nearest
cylinder boundary, hence the weird numbers.
Hope this helps a little bit. I certainly learned a lot from this
exercise. While I don't like how I got there, as I am two days behind in
getting the web site back up, the learning is a plus.
Any other questions, let me know.
MB
--
To unsubscribe: mail [EMAIL PROTECTED] with "unsubscribe"
as the Subject.
