Looks like I have just had a failure of a plug that has been running 24/7 since I bought it last November.  It was acting as a router/home server, and had an ethernet adapter plugged into the usb.

It seems to have fried the sd card that was in it - I can't get any response from it in any of my other machines I tried plugging it in.

It seems to be working OK now, although I haven't changed anything.  Sorry for the noise

I have Debian squeeze installed on an SD card, using the kernel from http://sheeva.with-linux.com/sheeva/.

When I issue the halt command (as root) the device shuts down but does not power off.

If I then power off, and then sometime later power on again it appears to drop into the u-boot prompt mode without trying to boot the unit (its difficult to be sure because its a bit chicken and egg with the console and usb). 

This is a bit inconvenient - I would really like to be able to not have a line into the console, switch the unit on and then just use it to provide a web server for some things I am doing.

Is there a work around - to ensure (primarily) that when I power on, it goes on to successfully boot the device.

alan@piglet:~$ uname -r
2.6.32-3-kirkwood
alan@piglet:~$

No problems at all.  Standard Debian Squeeze installation.

I have a standard plug with a 16GB SD card as my home server and internet firewall/gateway.  (I use the inbuilt Ethernet port as my lan side and a usb ethernet adapter as the wan side).  Supports everything including mail web and dns/dhcp services for my house and externally (both my personal http://www.chandlerfamily.org.uk web site - including a git repository and gitweb and my business http://www.hartley-consultants.com web sites are supported from it).

Strange - when I posted my message, uImage and uInitrd had file modification times before those of vmlinuz and Initrd.img files.  I have just checked again, and now they are a few seconds after.

But thanks for the heads up - I know now that its being taken care of.

I have setup an sd card with Debian (squeeze) as per the instructions at http://www.cyrius.com/debian/kirkwood/sheevaplug/install.html .  This all works fine, and I have now a good working system which I have just installed as my home gateway.

However - the boot sequence is loading images /uImage and /uInitrd from the first partition of the card. (as per instructions)

In Debian, the kernel (latest version) is kept in the this first partition (mounted as /boot ) with symlinks "vmlinuz" and "Initrd.img" pointing to the actual files.  As far as I can see (from dates on the files) there is no attempt to update uImage and uInitrd.

So the question is - should I update the booting environment to load vmlinuz and Initrd.img instead of the uImage and uInitrd - and if I do, will uboot be able to cope with the symlink?

I am puzzled about the addresses that are used in many of the commands.  I am obviously aware that they are hex numbers, related to the addresses of things, but I can't seem to find any explanation of how the nand and ram are laid out and why particular addresses are chosen.

For instance - the factory boot sequence defines the bootcmd environment

bootcmd=nand read.e 0x800000 0x100000 0x400000 ; bootm 0x800000

help for nand read.e says 0x800000 is an addr and there is one other parameter (off | partition size).  I can't find any more documentation that refers to this, yet this command has 3 parameters.

I assume the kernel is being loaded into RAM at 8MB and then the bootm command starts it executing from there.  WHY 8MB?  The Debian instructions (http://www.cyrius.com/debian/kirkwood/sheevaplug/install.html) loads it into 4MB and puts an Initrd Image at 8MB.  Why the difference.  Are these numbers plucked out of thin air, or is there a reason they are like they are?