So you even took the time to search and find the old thread but started a new one despite this? *sigh* Less threads, less fragmentation, I think it would've been better just to add a notice to it. Anyway, assuming that all people who have problem run the kernel they should use (complaining about that bug when you're running a PXA255 kernel on a PXA250 model would be pretty useless), it looks like this problem is really complex and I suggest someone starting to debug the spectrum_cs kernel module on a 5600.
