> It appears that there might be other cases of 2.3.18-level > EXT2 fs corruption. So Roland's porting might have been > perfect, but the Linux model might have a flaw. See > > http://kernelnotes.org/lnxlists/linux-kernel/lk_9909_04/msg00452.html > > for a few messages on this subject. I didn't see anything enlightening there, and no Linux ext2 changes between 2.3.14 and 2.3.18 appear relevant either.