[nzlug] linux on Compaq Evo T20

[Karl.]

> >I'm trying to get linux to recognise the internal drive (flash) in a
> >Compaq Evo T20.
...
> IIRC - The kernel is loaded via the boot loader - in this case grub,  
> so to get your kernel up and running, grub needs to be able to find 
> it - so far, so good.  Next - is initrd, and, unless I've very 
> mistaken[0] - this is pointed from grub as well.
  
Yes, that's right.  I am using a grub preset-menu to tell it to load 
the kernel and initrd from hd0,0.  I've not been able to get grub to use 
a normal /boot/grub/menu.lst config file, but I've not pursued that 
route very far - instead I've been trying cmdline options for the kernel 
I am booting - the point being that if I am successfully booting a 
kernel (which I am) then it is more important to get that to recognise 
the drive than grub (whose sole purpose is to boot the kernel, which it 
is doing).  Fixing one may turn out to fix the other though...

The system is a little inconvenient to debug because the only keyboard 
port is usb, and grub doesn't support usb keyboards, so I can't run grub 
interactively.  As far as I can see, there is no conventional cmos setup 
in which one could enable usb legacy keyboard support - this appears to 
be a 'legacy-free' device.  Each test run of new kernel options requires 
re-flashing the device.  

I've not succeeded in getting grub to fetch a config/kernel from the 
network either - I get a compile error when I try to build the ethernet 
support into grub (main.c:57: error: static declaration of 'bootp_data' 
follows non-static declaration) and I've not pursued fixing that.  It 
might be a good option to make experimentation easier.


> So - that leaves the possibility that the kernel cannot see the drive  
> itself - which I would assume is due to missing module / driver in the 
> kernel itself.

I think/thought (and might still think) that too, but since grub can 
load the kernel from the drive then it can't need anything too bizarre 
can it?  I am running DSL which runs on all sorts of things and I 
presume should be able to handle anything that grub can read...

I do wonder if grub is blindly forging ahead with something and it is 
working, whereas the kernel is more carefully probing and not seeing 
what it expects to see and is therefore giving up (and if it blindly 
forged ahead then it might work too).  I have been experimenting with  
kernel options along these lines (hard-coding ide parameters), but have 
not found success yet. 


> Hrmmm - sounds very much like there's something not quite right with  
> the drive, either the kernel not detecting it, or something strange 
> about the drive itself.
    
I gather that the Geode/CS5530 is not entirely normal.  I tried 
re-flashing with the NTembedded image which is meant for the machine and 
then ran the MS diagnostics prog WINMSD.EXE, hoping to find some 
interrupt or other info, but it didn't reveal anything much to me.

 
> As an alternative - how about booting off the network?

Yes, that works.  Before I got DSL running I did an etherboot/LTSP setup 
(http://rom-o-matic.net makes it pretty easy).  My main objective is a 
standalone boot though - there are a couple of applications I have where 
there won't be a server available.


> ===========================================
> >hda: C/H/S=27881/83/162 from BIOS ignored
> >hdb: C/H/S=0/0/0 from BIOS ignored
> ===========================================
> 
> This is what worries me - these tell me the BIOS details are ignored -
> but I cannot see any further detail about the drives?
> 
> I would suggest that there is something in there that needs to be set  
> for the flash drive to be recognised.

Yep.  That's my current area of guessing too.

Thanks for your thoughts on the matter  :-)

Karl.