if env atagaddr is defined (by default address from NOLO - register r2) then cmdline argument configured by uboot is ignored. If you want to pass your own cmdline from uboot, first clear env atagaddr (setenv atagaddr)

But atagaddr is not empty as default ?

Only if it is executed the script "noloboot" atagaddr is filled with ${init_atagaddr};

I check this and yes, you are right. uboot has two variables: atagaddr and init_atagaddr. atagaddr is used for booting and init_atagaddr is filled when uboot is loading. script noloboot set atagaddr from init_atagaddr

my u-boot now allows to copy the initial atags and append cmdline and ramdisk images to it.
Nitdroid now comes up without panics, completely ignoring the maemo rootfs.
Only thing still is the firmware. I'll probably just copy the firmware files tot hte nitdroid partition. Will test later tonight.

Finally got it all working I can now boot into Maemo, Meego or Nitdroid without needing to reflash or change cards. I've attached the patch to u-boot I'm using - it appends the initrd and command line (if set) to the atags instead of just ignoring them when atagaddr is set.

For Nitdroid, I'm using multiboot as normal, but fooling it into thinking it's already done the reflash, so it just loads Nitdroid straight away. If anyone's interested, I can also post my initramfs and u-boot scripts.

same here.

But my attempts to boot without using the maemo rootfs fail.Nitdroid needs the maemo rootfs to start all kinds of daemons for 'propriety' functions like bme and also starts things like dbus from the maemo partition.
I guess this can all be fixed by copying a couple of maemo files to the nitdroid partition, but it needs lots of changes to the nitdroid startup code.

So the nitdorid port is not really an native port... Pitty.
But the important things is we can boot nitdroid (and others) without reflashing!

anyway here is my version of the u-boot update. It uses an environment variable "atagcopy" to indicate you want to copy the existing atags and append new ones to it. e.g. something like:

setenv atagcopy $init_atagaddr
setenv bootargs root=/dev/mmcblk0p5
loadext2 .... $loadaddr uImage
loadext2 .... $ramfsaddr uInitramfs
bootm $loadaddr $ramfsaddr

I can't seem to see anything weird in powertop before and after installing u-boot, but the battery applet's been acting weird. No idea where the power consumption's going (shouldn't Linux take control of all the devices when it starts?)

u-boot is a nice piece of software.
I just stuck with
1. loading initramfs. Could you please show a command for that?
2. I've seen fatload working (once). How can I do the same for ext2. Example?
3. Most recent u-boot+PR1.3 kernel -? At http://al.robotfuzz.com/~al/maemo/u-boot/ ?

1. You can load the initramfs by doing "fatload mmc 0:3 0x83000000 initrd.img" and then boot it by doing "bootm 0x82000000 0x83000000".
2. As well as fatload, there's a corresponding ext2load which you can use in the same way.
3. The latest u-boot (AFAIK) is at http://atrey.karlin.mff.cuni.cz/~pali

Good news

I found a way how to get address where was u-boot loaded by NOLO and where u-boot expect start address in any C or ASM code (not only in start.S). So relocation of u-boot binary to correct address is possible in any source file.

It is very tricky. In one section will be stored address of that section generated by compiler (this address will correct after reloacation). And also it is possible to get address where is really that section stored.

assembler code:
and in same file will be calculating offset (saved in r0):
after that it is possible to move relocation code from armv7/start.S to rx51 board code and maybe this solution can be accepted in upstream.