PCMCIABoot GPXE

Why PCMCIA Etherboot

Etherboot is great; it will boot many types of image over a network. It can't boot PCMCIA devices ... yet.

Support for PCMCIA would open up a host of old hardware to network booting, and so all the advantages it brings:

• Centralised administration;
• Flexibility of operating system choice;
• LTSP systems could be built more easily and quickly on PCMCIA based systems;
• Network booting over wireless.

GPXE is the name for the next generation of etherboot.

Let's do it

Bristol Wireless have recieved a quote for £3,200 from Michael Brown of Fen Systems Ltd to undertake this work. The full text of the quote is included at the bottom of this page.

The Benefits

Bristol Wireless uses a suite of 20 laptops, providing a low power, low footprint mobile internet cafe, all using the processing power of our one AMD64-based laptop. The client machines boot a kernel and run some scripts. Both are based on the wireless-ltsp floppy disk. Installing this on 20 laptops was quite a headache and was only possible with committed volunteers who are frankly unsustainably generous with their time.

In running the suite we have come to conclusion that this sort of project would be useful elsewhere. Recent disasters have brought home to us the value of communication in troubled times. Low power robust laptop suites would be highly useful in such situations and could be easily built with recycled hardware. It's a lot easier if there is an easy way to boot them.

Funding Sources

PCMCIABootGPXEFunding

This is an idea and a to-do list for funding for this project, a lot of the contacts are organisation are in the South West UK, but are purely a starting point:

Donations

We could set up a website and take donations, or use some kind of pledge system.

Recycling

Recycling consortium would have info on this; call them.

Education

Try the Park in Knowle as a try.

Private companies interested in PCMCIA booting

There must be some; ask on appropriate mailing lists.

What Hardware should it support

This depends partly on the funding source, any major funder might want to influence the choice of chipset, or one chipset might be linked to a specific project. It would be worth doing a quick field survey to find which chipsets would benefit the most people.

The Quote

The original response reads as follows:

I estimate around £3,200 (approx US$5,700) to implement, assuming that the implementation would include:

• a generic (hardware-independent) PCMCIA subsystem
• support for one specific PCMCIA chipset, cleanly separated from the
• generic code to allow for easy implementation of support for further chipsets
• support for one specific PCMCIA NIC, again cleanly separated

I've already done a substantial portion of the groundwork in separating out the different bus architectures in gpxe-0.5 (see drivers/bus/{pci,isapnp,eisa,mca,isa}.c; for example, we now have generic ISAPnP support, replacing the minimalistic support that previously existed only as part of the 3c515 driver.

I'd suggest that you identify the specific PCMCIA chipset and NIC to be used for the initial project, so that you end up with something that will be immediately useful to you. In addition to your specified hardware, I would also implement support for the PLX9052 PCI-to-PCMCIA adapter and Netgear MA301 wireless PCMCIA NIC; I already own these pieces of hardware and have previously implemented basic support for them as part of the prism2 Etherboot driver. Generalising this support out of this driver and into the generic PCMCIA framework will act as a useful check that the interfaces between the layers are sufficiently generic.

I work on a fixed-price basis rather than a per-day basis; once we have agreed on a job specification and cost, I will work to that price regardless of project overruns (or underruns).

Let me know if/how you want to proceed.

Michael