This was brought up earlier at http://lkml.org/lkml/2005/5/6/30
Is there any reason for this one-liner to not go into mainline? It
appears to fix up a quite nasty failure. Even if it is not the real
correct long term solution it has to be better than having boxes
panic'ing under certain loads?
I'm clueless about this but it does appear to fix my bug and others have
hit this aswell. (apologies if this has been fixed in another way,
haven't seen any jbd/ext3 patches in a while)