It appears that delalloc lets me copy 50M of data onto a 30M filesystem;
at least I never get ENOSPC back, although I wind up with several files
that have 1M length but 0 blocks.
I've filed a bug in the kernel bug tracker, I think we could use a
central place to track issues:
http://bugzilla.kernel.org/show_bug.cgi?id=9329
I'll try to find time to look into this unless someone knows offhand
where the problem is...
Thanks,
-Eric
(p.s. should I get ext[234] bug mail routed to this list, or would that
be annoying?)