Subject: Re: port-i386/10187: consistent panic: lockmgr: no context
To: None <jhawk@mit.edu, simonb@netbsd.org>
From: Jason R Thorpe <thorpej@zembu.com>
List: tech-kern
Date: 05/23/2000 20:23:53
On Tue, May 23, 2000 at 10:59:44PM -0400, John Hawkinson wrote:

 > >Category:       port-i386

This is not i386-specific.

 > >Fix:
 > 	I dunno. This panic happens if curproc is NULL upon entry into
 > lockmgr(), but I don't know what causes curproc to go NULL, other than
 > cpu_switch(), and I'm presuming that's not what is going on here.

curproc is NULL when the system is idle.  If you then get an interrupt
which leads to a panic, curproc is obviously still NULL and the problem
occurs.

The alternative, of course, is for lockmgr() to simply dereference NULL
when it attempts to store the pid of the locking process in the lock
structure.

Which do you prefer?  Random kernel fault or meaningful panic? :-)

-- 
        -- Jason R. Thorpe <thorpej@zembu.com>