[Gta04-owner] NAND behavior while flashing qtmoko
Dr. H. Nikolaus Schaller
hns at goldelico.com
Mon Mar 26 08:07:18 CEST 2012
Am 25.03.2012 um 17:12 schrieb Michele Brocco:
> Hi folks,
> just want to report some experiences while flashing qtmoko to NAND:
> - first, why has the qtmoko kernel a hardcoded path for the rootfs?
> Why doesn't it get the parameters from u-boot?
From a distro developer's perspective it is a nightmare to ask users
to modify the boot args in u-boot so that it works. It is much easier
to hard code that into the kernel.
The main reason is that u-boot args are not stored in a way they are
automatically modified during the installation process.
So I can understand Radek't decision.
This of course disables the automatic bootarg setup of the latest boot.scr
which tries jffs2 if the kernel comes from NAND and mmcblk if it comes
And it requires Radek to provide two different file systems/kernels. One
for SD card and one for NAND flash. This is IMHO also not optimal.
But I don't know a good solution that is better than automatically setting
the boot args by boot.scr.
Unfortunately we have another issue: u-boot can't detect if there is
a jffs2 or a ubifs in NAND. So it can't automatically launch the (single
version) kernel correctly.
My dream is that kernels and rootfs can be swapped easily and
independently of each other and u-boot automatically does the
correct setup. This would simplify installation and reduce the number
of pure u-boot config issues showing up in user-space functions not
Maybe someone has an idea how to solve that?
> - qtmoko only seems to boot correctly if the kernel is flashed from qtmoko
> - qtmoko from nand just seem to boot correctly from jffs2 when linux
> is booted correctly once from mmc, otherwise i get filesystem/mtd
> errors and qtmoko is not booted.
I have not seen this when using NAND/jffs2 with the Debian/LXDE
setup. So it might have to do with the QtMoko kernel.
> did anyone experience the same issues?
> Thanks and regards,
More information about the Gta04-owner