Fasttrack TX2000 RAID problem

[Molle Bestefich]

Greg Felix wrote:
> I've had some success in debugging this today and here's what I've
> found.  It appears that the metadata for the first disk (that dmraid
> if correctly finding) is 63 sectors from the end of the disk.  I would
> assume that the metadata for the second disk would be in the same
> place, but it's not.  I found it instead 41000 or so sectors before
> that.  When I modify dmraid and recompile so it's looking for this
> exact sector it find the metadata and seems to setup the array
> correctly.

Great!
Well done ;-).

> The question is of course WHY is the metadata there
> instead of at the end?  Here's what I think.  I think the BIOS is
> coming up with a different number of sectors on the disk than Linux
> is.  When it lays down the metadata it's putting it at totalNumber-63
> whereas the kernel is getting a different value.  The two actual hard
> drives aren't the same type so this might play into it.

I can't come up with a good reason why the BIOS would find a smaller
number of sectors than Linux does, besides from buggy firmware that
is.  Translation modes come to mind (LARGE/LBA/etc.), but I can't
imagine that a shiny new Promise controller would do anything other
than standard LBA?..  Hmm.

> I tried upgrading the firmware of my Fasttrack TX2000 from 2.00.0.24
> to 2.00.0.33, but it seems to be setting up the metadata is the same
> strange place.  Any thoughts on any of this would be appreciated.

At least it's behaviour is consistent.

> Here is some info of where I found the metadata for each disk
> 
> Disk 1:
> found at 78156225
> out of 78156287
> = 63 from end
> 
> Disk 2:
> found at 78123969
> out of 78165359
> = 41391 from end

Ack..  If I remember correctly, standard IDE disks have 512
bytes/sector?  The metadata is then located suspiciously close to what
the manufacturers like to call 40 "GB" (which would be 78125000
sectors).  Maybe there's some sort of GB capping on your controller
enabled for disk 2, in order to make sure that you can actually
replace a "40 GB" disk from one manufacturer with a slightly smaller
"40 GB" disk from another manufacturer?

Could be a possible explanation, on the other hand, in that case we're
still 968 sectors off (when counting in the standard 63 sectors), and
I might just be fooling myself because of the simple fact that it's a
40 GB disk anyhow ;-).

> What do you think?

I think I'm in way over my head :-).

Are the data on your disks expendable?
I'd like to see what happens if you 'dd if=/dev/zero' metadata on both
disks, swap their places (so the larger disk 2 becomes disk 1) and
create a new RAID0 on them........