[nzlug] linux on Compaq Evo T20

[Karl.]

On Sat, Nov 25, 2006 at 11:53:21AM +1300, Daniel Lawson wrote:
> You can definitely PXE or etherboot these devices. Whether your
> particular model supports PXE boot or not is another matter. I've had
> limited success in reflashing the older models that don't support PXE
> with an etherboot ROM, however I ran into problems getting them to run X
> and I haven't really gotten any further.

Mine are 96/128Meg models (the 96M flash consists of 2 x Toshiba 
TC58256AFT soldered to the board + 32MB SmartMedia in a socket, the 128M 
ram is 4 x Hynix HY57V561620BT-H).  What appears to be the part number 
of the whole machine, 238365-001, doesn't seem to exist anywhere on the 
web - maybe it's just the part number of the metal foot  :-)

Having no sensible partnumber, I just grabbed an image file for a 96/128 
machine from the HP site and crossed my fingers - it worked out fine.  I 
can flash that image and run NTe.  I can run DSL by loading kernel and 
initrd off internal flash and the rest off USB flash.  I can also patch 
the image with etherboot and run LTSP or DSL loaded over the network.

Network boot doesn't work entirely perfectly, in that at least half 
the time the network doesn't come up properly and it just reports 
corrupted packets and won't even fetch an IP address from DHCP.  Once it 
does come up properly, soft rebooting DSL ('sudo reboot') always works 
(ie. network comes up fine) - if I hard reboot (by power cycling or 
holding down the front-panel power button) then I run the risk of 
the corrupted packets failure again and I need to keep rebooting until 
it works. 

I've not bothered to try to troubleshoot this yet - it's not important 
to me at the moment, in that my goal is currently a stand-alone boot: 
no network and no USB flash, booting fully from internal flash.

In my limited fiddling with LTSP, they did seem to run X fine for me - 
what trouble did you have?

 
> I have  some notes at http://www.wlug.org.nz/CompaqEvoT20Notes. Feel 
> free to add to them.

I plan to do that  :-)


My current theory is that the IDE interface is not being initialised 
properly by the PCI code - 'lspci' lists the IDE interface, but 'cat 
/proc/ioports' lists no IDE ports apart from the high ones (ie. no ports 
around 0x1f0).  I'm trying kernel options to see if I can get anything 
to change there.

On my PII-400, cat /proc/ioports shows:
  ...
  0170-0177 : ide1
  01f0-01f7 : ide0
  ...
  0376-0376 : ide1
  ...
  03f6-03f6 : ide0
  ...
  ffa0-ffaf : Intel Corp. 82371AB/EB/MB PIIX4 IDE
    ffa0-ffa7 : ide0                             
    ffa8-ffaf : ide1
  ...

  
Actually, in the time that I've taken to write this email, I've tried 
all the likely looking 'pci=' options and found no joy.  I am running 
out of guesses  :-(


Karl.