This is just an ad hoc off-the-cuff description of how we currently see the work of FreeSA being broken down. See also: FreeSA Notes.

Stage 0

The goal is to establish that we have the tools to develop FreeSA. done

Stage 1

The goal is to have a running WL-700gE with a current kernel that can access the internal hard disk. in progress

flashing OpenWRT

  • build & flash & boot system with OpenWRT (done)
  • mount the hard drive (in progress)
  • fold this knowledge into our flashless development cycle

flashless development cycle

  • Investigate Latest OpenWRT patches (in progress)
  • Start with a machine booted using TFTP kernel, rootfs via NFS (in progress)
    • Eric is booting the 2.6.19.2 kernel built from OpenWRT trunk (done)
      • Add patches related to WL-700gE (not started)
      • get the rootfs updated to match current kernel & modules (not started)
  • Check to see if GPIO control is exposed over /proc (in progress)
    • With what Eric has so far, it is not:
        asus-debian:/proc# find . -iname "*gp*io*"
  asus-debian:/proc# 

  asus-debian:/sys# find . -iname "*gp*io*"
  asus-debian:/sys#
  • Modify the kernel so that GPIO control is exposed over /proc (not started)
  • Power-on the HD using GPIO pins 3, 6.
  • insmod the IDE driver.
    • Get it to work.

success condition:

  • mount the hard disk

If we make it to this point, then we fundamentally have the capability to bootstrap a system.

  • drink a toast

Stage 2

The goal is to verify that all critical hardware is supportable with a current kernel.

Using the network-booted stuff, verify that the following hardware components all work:

  • eth1 (investigate why eth1 complains at boot)
  • Wireless
  • USB ports
  • 2-wire serial bus
  • LM63 temperature sensor
  • RTC

The only thing that is IMPORTANT and liable to be difficult is the wireless. (Other critical components are obviously supported if we get here: IDE, GPIO...) However, it looks like this is detected with latest OpenWRT build, so that is very good news.

Stage 3

The goal is to have a FreeSA kernel image that has a root filesystem mounted from the internal HD.

Try using initramfs with the stage1 kernel and so on, using early userspace to power-on the HD, load modules, etc., to access the hard drive and then do a pivot-root to make the HD the root fs.

Probably it would be a good idea as well to make the boot process smart enough to see whether there is a valid HD root filesystem, and IF NOT then switch to NFS-mounting a root filesystem? Or for this initial version do we want to recommend people not program this kernel into the flash until they have their HD rebuilt?

If initramfs does not work -- then switch to using squashfs, and build a TRX image that does the IDE init and pivot-root magic. Of course that means we need to know how to build a TRX image. "Shouldn't be a problem."

Stage 4

the goal is to have a CLFS distribution resident on the hard disk and booted by the stage2 kernel and/or initial rootfs.

  • Build a MIPS CLFS initial system.
    • Use the NFS rootfs to install the CLFS initial system on the internal HD.
  • Program the stage2 kernel into the flash chip.
  • Boot the system so that the serial console provides access to the CLFS initial system.
  • Complete the entire CLFS build process, using package users, from the serial console.

Declare victory!

Stage 5

the goal is to complete the FreeSA distribution

  • Build a custom cross-toolchain from source
    • the OpenWRT toolchain is free but difficult to understand, not targeted to our needs, but a useful start to springboard from.
  • Using package users, build and configure all of the software packages that provide features we want.
    • iptables, Apache, SVN, fsvs, nagios, postfix, maradns, asterisk, jabber ...

Other Tasks

  • we should investigate costs and turn-around time for having Asus re-flash a warrenty void unit. (no feedback)

Further Steps

These are optional but would be awfully nice.

  • provide a way for people to bootstrap a FreeSA build without having a serial console AT ALL.
  • build a nettrom-style flash image that allows the runtime kernel to be loaded from TFTP or HD, and allows root filesystem to be NFS-mounted or HD-based. (probably requires kexec for MIPS)
  • create a TRX image that, if there is no rootfs and software installation on the hard drive, creates a FreeSA installation by partitioning the hard drive, creating filesystems, and then starts up a server (sshd, ftpd, something like that) to allow files to be installed into the hard drive. This is minimally what is necessary for the "prebuilt FreeSA" distribution.
  • Include some Advanced Routing & Traffic Control examples like prioritizing VoIP trafic high and ftp traffic low.
  • Include some security examples to work in conjunction with ARTC so that we may confidently open up our wireless connection for the neighbors to use, without having to worry that the neighbors might be a threat to their machines, and having confidence that our traffic takes priority over the neighbors so we don't care if they want to try to be bandwidth hogs.