[nzlug] linux on Compaq Evo T20

[Karl.]

> Legacy free - that's what I read on the matter...  You might be able to
> chain load etherboot for network booting (This is just from skim reading
> some webpage).

Yes - that could be a useful diagnostic tool, if I use etherboot instead 
of grub, and fetch everything from a server - then I can just edit 
options on the server instead of the continual re-flashing.  I'd only
thought of trying that directly with grub.  Once I figure out good 
options that let me access the drive from my network-booted kernel then 
I can try applying those options to my grub-booted setup.  Thanks  :-)
  

> I thought grub blindly accepted what the BIOS told it?  In any case, the
> kernel will definitely ignore the BIOS - iirc it uses the BIOS to gather
> information, but doesn't actually use the BIOS calls for hard drive
> activity.

I think that's right.  http://www.kazak.ws/evo/step2.html says:
"Also the BIOS doesn't pass the boot device to GRUB so we must manually 
specify 0x80 (first hard drive) in place of GRUB_INVALID_DRIVE."

So, I have patched grub this way and grub works.  I guess the kernel is 
trying to get info from the bios and is getting nonsense (as evidenced 
by the wacky Cylinder/Head/Sectors being returned) and so it's ignoring 
the drive.  I'm trying to figure out what options to pass to the kernel 
to hard-code things rather than asking the bios.  I've found some likely 
options, but have not found the right combination yet.


> Which makes me think - does the IDE driver have a switch (or an there 
> an alternative IDE driver) that tells the kernel to use BIOS calls
> instead?  It *might* be slower, etc. but I hardly think that this  
> application is going to suffer too much ... ;-)

It's certainly worth checking out options - the USB ports are only 
USB1.1, so even bios access to the internal drive is going to be faster 
than access to the external flashdrive. 

I'll try fiddling with etherboot now and see what I can do...

Karl.