Closing this bugreport, see:

  http://www.ussg.iu.edu/hypermail/linux/kernel/0411.1/1098.html

how to basically work around the problem using evms. I just don't want to 
introduce a patch I can't verify on my own and upstream didn't want to accept.