Debian etch and "bricked" netbook

[xr82]

Hello,
I had been running etch on my belco alpha 400 (I believe its the same as the Trendac) for a week or two successfully. But just the other day I did a 'shutdown -h now' and when I tried to turn on the machine again about half an hour later all I got was some very weird graphical corruption. Kind of like many little different colored pixels on the screen at once. It is also similar to the screen you get from the recovery system when it cannot find a recovery image to load and you end up bailing out. Now I can't see how my shutdown could have effected the machine, and I can't seem to recover the machine either (holding down f3 does nothing). I haven't tried "flashing" the thing yet (Belco doesn't seem to have a web page or instructions on how to do this) and I've been holding off until I fully understand what this means. Is it possible to actually brick the machine? Through what _technical_ means does the kernel get loaded onto the machine (what they have termed "flashing")?
I guess thats all the information I can think up at the moment. I'd be grateful if someone would throw me a line

[Mijzelf]

The laptop has 1 or 2 GiB flashmemory which is devided in several blocks. One block contains the bootloader, another block contains the (linux-)kernel, and a third block contains 'the filesystem' which contains 'the distro'. (You could have seen this using 'cat /proc/mtd' when your laptop still worked)

The bootloader is executed first, on boot, and this loads and fires the kernel. The kernel mounts the filesystem and executes a script to fire 'the distro'.

The bootloader can be interrupted by a keycombo (ctrl-fn-f2, as I recall right) to flash another kernel in it's block. The kernel can be interrupted (f3) to write another distro to the filesystem.

So you rewrote the filesystem with Debian, but didn't touch the kernel or bootloader. F3 should still work. As long as you didn't do strange things (mkfs /dev/mtdblock1 or something like that) the kernel and bootloader should be fine.

You can try to reset the system. Use a paperclip or something like that on the small hole on the bottom. Additionally you can remove the batteries for a few hours.

[xr82]

Thanks Mijzelf for the very informative reply!
I did notice those different blocks before, having catted /proc/mtd. How is /dev/nand different from those blocks? Why isn't it that /dev/mtdblock3 (or which ever number it is for the rootfs) is mounted on / instead of /dev/nand being mounted on /?

Thanks to you I now understand that the debian recovery image never replaced the already pre-existing kernel! That really helps clear up a lot. But I'm getting off topic...

Would it be possible to brick the machine if the bootloader block was overwritten incorrectly? (How can I load a fresh bootloader onto the bootloader block?) I don't think this happened but resetting the machine like you suggested is still not working. Perhaps I'll try and load a fresh kernel onto the nand.

Thanks for baring with my questions. I've got many more :-)

[Mijzelf]

How is /dev/nand different from those blocks?

Not. It's just a name. When you do a 'ls -l /dev' you can see that the major and minor device numbers of /dev/nand fit perfectly in the list of /dev/mtdblockx. This is what really defines the device, the name is just a convention.

Would it be possible to brick the machine if the bootloader block was overwritten incorrectly?

Yes. But afaik the bootloader is never overwitten, and the block is not touched/mounted/used after the kernel has loaded.

Perhaps I'll try and load a fresh kernel onto the nand.

More info on bootloader (U-Boot) key combo's here, and here (see 'Booting the SD card way')

[LilYoda]

I had same symptoms when trying out 2.6 kernels.  I think your kernel or bootloader is hosed, need to reinstall one...