onsdag 14. juni 2006, 17:19, skrev Falko Trojahn :

> >         else {
> >             # Non-IDE disk.  Should probably sanity-check kernel
> >             # geometry against legacy BIOS geometry here.  FIXME.
> >         }

> > Non-IDE disc systems end up in the commented out else .. :-/
> > (at least this Thinkpad ends up here)
>
> BTW. this is exactly the place (you wrote about this on 12/06/2006)

yep, :-) I was able to pinpoint it more with a machine on my desk.

> Did you read NOTES.txt under "linuxboot" - perhaps this clears
> things a bit up.

ah, there it is, well documented. 
My hat off for Pat and the work he put into this project.

> Ok, just some thoughts from me about this:
>
> - may be you could provide the right settings by changing install.pl
> and setting the values e.g. in config.pl, if you know them (240
> heads ...), so that parted uses them. If there is no other way,
> parted can be patched, too, to use the needed values.

I think we have what we need in the environment variables
LEGACY_BIOS_HEAD, LEGACY_BIOS_SECT and $sys_hda
I must get this confirmed.. 

> - did you prove that it bails out in the else-clause
>      "     # Non-IDE disk.  Should probably sanity-check kernel ..."

Yes, Im shure it ended up inside the else-clause,  
I had a print statement there.

>   and could you change the install.pl to print out the relevant
> values at this stage? E.g. by
>           printf "bios_cyl:%d bios_head:%d bios_sect:%d\n",
>             $cylinders, $bios_head, $bios_sect;
>         printf "hard drive: $hda   sys: $sys_hda\n";
>  perhaps other variables, too.

That was the plan, but I did not have enough time on this machine to do it..
..Will try the same on a Thinkpad R52.. ..and a Dell GX620 (noncombination 
mode)

> - if it really bails out at this else clause this means, that there
> is no      /proc/ide/$sys_hda/settings
>   so perhaps only the variable $sys_hda has to be corrected

I dont know much about the scsi system, ...
I was not able to find any
'settings' file for the disk (similar to the IDEs)..
I will see whats inside $sys_hda tomorrow.

> - since you can restart the process any time by starting
> "/etc/master" on commandline, you may change install.pl and don't
> have to reboot any time

Thanks! this is a good one!

> - please try to read/write mount a network drive in this situation
>   and do    dmesg > dmesg.txt ; lsmod > lsmod.txt ; set > set.txt ;
>       ls -lR /proc/ide >procide.txt
>   and whatever relevant information you may find
>   -> you can send it offlist, perhaps zip-files are not allowed.

Will give it a try!

> - perhaps we could trick parted to use the right heads/sectors by
>   using $LEGACY_BIOS_HEAD heads and $LEGACY_BIOS_SECT

yep!! How? .. is the question..

> BTW: did you try e.g.
>      dd if=/dev/zero  of=/dev/sda count=2000
> and reboot before starting unattended?

No, I believe this is the fix for BIOSes that 'adapt' geometry to whatever is 
on disk and has a bad mbr on disk.
Thinkpad T60 did not adapt to report 255 heads, 'Legacy BIOS' still reported 
240 heads. 



_______________________________________________
unattended-devel mailing list
unattended-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/unattended-devel

Reply via email to