That was the problem.  Adding a simple_unlock(&namecache_slock); before
the return took care of it.

Thanks.

On Tue, 20 May 2003 20:36:49 -0400
Stephan Uphoff <ups@stups.com> wrote:

> 
> vfs_cache.c:326 contains a return without the release of the lock.
> 
> 	if (bufp) {
> 				bp = *bpp;
> 				bp -= ncp->nc_nlen;
> 				if (bp <= bufp) {
> 					*dvpp = NULL;
> 					return (ERANGE);
> ----------------------------------------^^^^^^^^^^^^^^^^
> 
> I guess mi_switch was just not called for a while.
> ( And as such the backtrace not related to the problem)
> 
> 	Stephan
> 
> 
>  Frank van der Linden <fvdl@netbsd.org> wrote:
> > On Tue, May 20, 2003 at 06:20:01PM -0400, Matt Rickard wrote:
> > > 	checking if getcwd() calls popen() ...
> > > 
> > > On the console that I am doing the build on, I am able to type (or
> > > at least echo characters to the screen), but the input is ignored. 
> > > I can't ctrl-c to kill the task.  I can switch to other virtual
> > > consoles, but cannot input anything at all.
> > 
> > That spells 'locking problem', and indeed, a LOCKDEBUG kernel spews:
> > 
> > switching with held simple_lock 0xc03d1f30 CPU 1
> > ../../../../kern/vfs_cache.c:301
> > simple_lock_switchcheck(e4dc0400,0,0,c0228144,e4dc0400) at
> > netbsd:simple_lock_sw itchcheck+0x18
> > mi_switch(e4dc0400,0,1cd,c023245f,0) at netbsd:mi_switch+0x48
> > ltsleep(c03d1180,204,c03717e8,0,c0406d1c) at netbsd:ltsleep+0x2ef
> > reaper(e4dc0400,0,0,0,c010030c) at netbsd:reaper+0x6c
> > 
> > ..however, this doesn't make sense, since the lock taken at that
> > line is not held ever by the reaper, or can it ever be held 
> > in any way when a process sleeps. So at first glance I can't
> > figure out what's going on there.
> > 
> > - Frank
> > 
> > 
> > -- 
> > Frank van der Linden                                           
> > fvdl@netbsd.org======================================================
> > ========================= NetBSD. Free, Unix-like OS. > 45 different
> > platforms.    http://www.netbsd.org/ 
> > 
> 
> 
> 
> 
> 
>