Hmmm.

1] I'm clutching at straws here, and again, no offence intended, but are you remembering to "w write table to disk and exit"?

2] Maybe you should check your BIOS (I don't know how old it is). There are options sometimes for LBA (IIRC) and whatnot addressing of disks. It is worth a thorough look. Modern BIOSs don't care about the real, physical number of heads, cylinders, etc, older ones did. Check your BIOS, try different settings.

3] If you have access to them, do other PCs make any more sense of its geometry?
.
4] How old is this disk? What make / model is it?

Of course there is always the possibility that the disk (or, more likely, its (IDE ?) interface) is broken

Do you have a second disk of the same type, which you might copy over to the "mangled" one by dd?

<edit> ... or of any type, to test whether that behavior is reproducible ...

And again:

http://www.linuxquestions.org/questi...br#post2224998

and the main thread searched for "MBR"

http://www.linuxquestions.org/questi...archid=2773506

</edit>

There is software named "TestDisk". I used it once to restore accidentally erased MBR, but it might be able to fix drive geometry as well.

I have the exact same problem... Did you eventually find out how to fix this ?

Note: i tried testdisk, you can tune the geometry, but only to find new partitions as it seems.

I'm coming in late and the water may be under the bridge, but one thing caught my eye:

To the best of my knowledge, the partition table only describes the size, location and type of the partitions. I did not think the actual parameters of the drive would be affected by dd. This said, it is true that cloning partitions is much better than cloning whole drives.

Does anyone know for sure where the basic drive info is stored and whether dd will affect it? (I'm resisting the urge to haul the old P3 off the top shelf and run tests.....)

the weird thing is that i tried to copy the boot sector of a bigger disk with the same method i used to screw up the first time, and this time, the size is still the size of the smaller disk.

In fact, the only parameter i would have to change is the number of cylinders.

In the bios, the size of the disk is now wrong too (50 instead of 160)...

The BIOS is reading the disk geometry from the disk. The disk is reporting the geometry incorrectly, because you have used dd without fully understanding what it does, and the implications.

If it is all in a big mess (it seems it is), and there are "inconsistencies" (it seems there are) then the best thing to do is go back to basics before you make the situation even worse:

My recommendation:

1] Rescue your files:

Become the root user. Then:
(It's a "tar pipe", don't ask me how it works, but it has always copied and preserved everything for me)

2] fdisk and repartition your badly-behaving drive

3] Format your badly-behaving drive with the filesystem of your choice (ext3 for me).

4] Restore your files:
5] Check that you can access your files on /path/to/your/reformatted/disk

6] exit from being the root user

7] You are done.

In future, please open a new thread for this type of question

Hi tredegar,

thanks for replying an trying !

First, let me say i didn't open a new thread, because the situation described in the first post is exactly the same as mine, and it doesn't seem to have been solved.

Now, regarding fdisk, it doesn't do much. Yes, I can format and partition the disk... but as a 58.5 GB disk instead of 160 GB:

I tried to go to expert mode (x), change number of cylinder (19457, which is the proper number for the disk, samsung HM160HI), and create a partition going from cylinder 1 to 19457. But it doesn't work:
(of course...)

Gawd I wish people would leave "dd" alone.
Use the grub "geometry" subcommand to reset the number of (physical) cylinders, then fdisk might be more accommodating.

well, i promise I won't touch dd ever again... I usually use partimage and ntfsclone for that job, but i wanted to try a different method...

Anyway, I'm trying to change the geometry at boot time in grub:
- enter grub (c)
- geometry (hd2) 19457 255 63) => it always indicate 1024 cylinders for each disk.
- configfile (hd0,0)/boot/grub/menu.lst
=> no change once system is running, still 7113 cylinders

If i run grub during my session, the geometry command properly displays the number of cylinders for each disk, and i can set hd2 to 19457. But then, how do i exit from grub shell to impact my running session ?

quit, then reboot would be simplest.

from my ubuntu session, i did in the grub shell:
- geometry (hd2) 19457 255 63 => it reports the disk size as 160 GB
- quit
and rebooted the computer => no change, next session, fdisk -l still give me the wrong number of cyl

tried to put the "geometry" line in the menu.lst (maybe that is stupid...) => no change

Sorry...

well... no more ideas ? I'm not complaining, don't take it wrong, it put myself in this situation

Not really - I would have expected that to work, but I don't have a disk I'm prepared to try on.
The various [cs]fdisk variants might be worth a try - seems you have nothing (more) to lose.

update: SOLVED !

this post had the solution: http://www.linuxquestions.org/questi...ing-dd-632886/

I was indeed upgrading the disk for a dell laptop, and my problems started the first time i put the new drive in the dell => during the boot, the dell bios rewrote the LBA-3 so that i had a hidden partition of 100 gigas... As said in the post, using HDAT2 allowed me to remove the hidden partition and get the disk right.
it's interesting that none the tools i used saw the hidden partition, but i guess that is the point !