[Soekris] net4801: wrong drive geometry with 80GB harddisk

[Andy R]

Hello *,

I'm currently trying hard to get my new net4801 configuration up and running
with freebsd. Hardware: net4801/vpn1411/256MB CF + (the troublemaker) a 2.5"
80GB ibm/hitachi notebook drive.
I managed to install freebsd 5.3 via PXE without any major problems using the
CF for / and /usr /var /home on the harddisk. However, freebsd noted during
the partitioning part of the install that the system bios seems to have a
wrong drive geometry for the 80GB disk and it won't use the bios' information.

BIOS says:
comBIOS ver. 1.27  20041122  Copyright (C) 2000-2004 Soekris Engineering.
...
Pri Mas  SanDisk SDCFH-256               LBA 980-16-32  251 Mbyte
Pri Sla  IC25N080ATMR04-0                LBA Xlt 1024-255-63  12614 Mbyte
...

I also tied the harddisk as primary master without a change in geometry.
Freebsd said it will use 9729/255/63 (C/H/S) which sums up to 156301488 LBA
sectors. This is also what hdparm -I told me about the drive, when I booted
it in a laptop using a linux live distro.

Since freebsd seemed to have the right geometry, I thought everything is fine.
I used the new system for several hours (kernelbuild and some ports etc.) and
it did not show any odd behavior - besides sporadic (but steady) drive resets
and/or recalibrations (I don't really know how to _hear_ the difference...)
which sound like a loud >>clack<<. However, there were _no_ kernel error
messages in the syslog (*.*) file about drive failures or something similar.

I googled around and found some postings about a mystic "click of death"
(http://www.rfc1149.net/freebsd-net4801.html.en), so I installed the drive in
my laptop and stressed it with the "drive fitness test (DFT)" for quite a
while, but it never made the >>clack<< when attached to the laptop. (BTW: DFT
says, the drive is 100% ok.) So, my best guess for now is the geometry problem
and that the drive is therefore addressed "wrong" somehow, which leads to the
drive resets (and maybe the drives death if ignored).

Is there any possibility to hardcode the drives right geometry using "cmoswrite"
and switch of the autodetection ? Or even better: fix the comBIOS detection
routine ? ;-)

Any suggestions would be appreciated.

Thanks
Andy