On Mon, Mar 06, 2006 at 09:08:24PM +0100, Thomas Klausner wrote:
> Yesterday some processes on my -current/3.99.16 hung in vnlock.
> One was a bulk build in a sandbox (i.e. null-mounted /usr/pkgsrc)
> and the other was a 'echo /usr/pkgsrc/*/libetpan' (or similar).
> Both processes were unkillable, and other processes trying to
> access /usr/pkgsrc hang too.

Chuck Silvers debugged this one (thanks, Chuck!) and found
it's the same problem as PR 32535, caused by vrele() calling
vn_lock().
 Thomas