Subject: Re: 1.4.2 lockmgr
To: Bob Nestor <>
From: Bill Studenmund <>
List: current-users
Date: 03/20/2000 11:46:51
On Mon, 20 Mar 2000, Bob Nestor wrote:

> OK. Got another one and here's the traceback:
> _Debugger
> _panic
> _lockmgr
> _ufs_lock
> _vn_lock
> _vrele
> _nfs_lookup
> _namei
> _vn_open
> _sys_open
> _syscall
> _trap
> If this is similar to the ones Frederick has been seeing than it looks 
> like it's NFS related. Both of us are building packages from NFS mounted 
> volumes when the panic happens.

This backtrace is VERY strange. Note that nfs_lookup called vrele, which
called vnlock.. all that's fine so far .. but then when vnlock calls the
underlying lock routine... it gets ufs_lock!!!!!

There are two calls in nfs_lookup() to vrele. I can't tell which one is
the problem, but they both involve releasing newvp.

More importantly I don't understand why nfs is getting ufs vnodes. We
fixed the error where hash chains had mixed-up names, didn't we?

Take care,