[maemo-developers] Data corruption on N770 in OS2007 HE

[From: Alex DAMIAN]

Hello,

Thanks for your time.

the modified cx3110x driver works ok. I think it's a bit weird that my
WRT router would assign different IP addresses depending on the driver
loaded, but it's not an inconvenience after all.

However I run into a bit of trouble trying to modify the initfs. I
took out the initfs from the second HE FIASCO image, and mounted it
with loop/block2mtd and copied all the files off it (including the
/dev/). This pose no problem. I replaced the cx3110x.ko file, and
tried to make the updated jffs2 image with mkfs.jffs2 (I followed your
initfs-flasher script).

However the jffs2 image I'm building is quite different (quite a bit
smaller), mounting it gives signifiant differences in files compared
to the original one, and dumping it with jffs2dump shows lots of CRC
errors.

I even rebuild the mtd-utils from CVS, thinking that my Fedora has
some bugged version, but the result is the same.

Any idea about what I'm doing wrong ?

Thanks,
Alex

On Nov 7, 2007 1:29 PM, Frantisek Dufka <dufkaf at seznam.cz> wrote:
> Alex DAMIAN wrote:
> > I hoped for an error in dpkg
> > routines which handle the status file; now you made my day, I need a
> > better tablet :)
>
> Or just replace cx3110x.ko in /mnt/initfs/lib/modules/current/ with
> quick fix attached to the bug
> https://bugs.maemo.org/attachment.cgi?id=594
>
> Or if you don't want to mess with initfs you can put in somewhere in
> rootfs and remove and reinsert it in some boot script (like
> /etc/init.d/rcS), something like
>
> rmmod cx3110x
> insmod /path/to/your/cx3110x-on-stack.ko
> chroot /mnt/initfs wlan-cal
>
> should fix further data corruption caused by this bug. Doing a dry run
> in xterm would be wise before modifying /etc/init.d/rcS I am writing it
> from my memory (with some data corruption issues too :-).
>
> Please note that the final version of bugfixed cx3110x.ko may be a bit
> different, it may use different solution and further bugfixes. Just
> watch the bug. There are also other issues with the open sourced glue
> cx3110x.ko driver code (data alingment issues, busywaits everywhere).
> And thanks must go to Tilman Vogel and Siarhei Siamashka who traced down
> root of the bug and found a fix (after more than year, see also
> http://bugs.maemo.org/show_bug.cgi?id=677).
>
> Frantisek
>