Re: union mounts, mfs, and newsyslog

[David Holland <dholland-current%netbsd.org@localhost>]

On Sun, Jul 05, 2009 at 09:39:04PM -0400, matthew sporleder wrote:
 >                      if (rename(file1, file2))
 >                              err(EXIT_FAILURE, "%s", file1);
 > [...]
 >
 > /dev/wd0b       /var/log_mfs mfs rw,hidden,-s10M 0 0
 > /var/log_mfs    /var/log union rw - -
 > 
 > So is it a limitation of mfs, union mounting, or my ffs?

Unionfs has all kinds of problems, not least with rename. The short
answer is "it's not really expected to work". Although in this
particular case, I would have thought it would... or if not, fail some
other way, like by crashing, instead of apparently passing vnodes from
mismatched layers to one layer's VOP_RENAME.

If you want to make it work, the first step is probably to figure out
what it's doing: create a simple test case that exhibits the same
behavior (this is not necessarily so trivial...) and then poke into
the kernel to figure out what union_rename thinks it's doing.

However... there are a number of more or less insoluble problems with
rename and unionfs (in more complicated cases) and you might prefer
not to get involved.

-- 
David A. Holland
dholland%netbsd.org@localhost