On Wed, Mar 05, 2003 at 11:25:13AM -0500, Jan Harkes wrote:
> 
> I'm not sure why fixes don't end up in the FreeBSD tree. 5.3.20 fixed
> all known compilation issues at the time. The vfs_mount recursing back
> to venus is fixed in the current development tree, basically unifying
> the code with Linux where we always forked a child that performs the
> mount syscall (NetBSD needed that fork as well).
> 

I took a different approach to fixing the deadlock (as Jan knows
already) by writing a mount_coda program which execs a venus process
that carefully does not perform the mount but rather sends a signal to
the mount_coda parent to indicate the mount can happen.  This avoids
the deadlock and also provides the "standard" mount semantics for a
file system viz:

mount -t coda /dev/cfs0 /coda

does what you would expect it to - mounts the coda file system (and
starts up venus).  At the moment I don't know if it is appropriate for
this to go into the coda src tree or into the NetBSD src tree, the
reason being is mount only looks in specific spots for the mount_blah
helpers, none of those spots are in the pkg tree (this totally leaves
aside the issue of pkg being in a "non-standard" location).

-- 
Brett Lymn