Re: kern/37925 (more locking problems with unionfs)

[Andrew Doran <ad%NetBSD.org@localhost>]

I can only reproduce this with >1 processor online. It seems to have done
damage to the underlying ffs file system that I tested on. The playing about
with vnode locks that unionfs is doing is breaking down in the MP world.

Can you use "-t null -o union" mounts as a workaround for the time being? I
am reviewing the recent unionfs changes in FreeBSD and seeing what applies
to us. Given unionfs' history to date, one option would be to retire unionfs
and replace with "-o union" / null mounts, and fanoutfs in userspace -
however that would be unfortunate.

Thanks,
Andrew