On Mon, 1 May 2000, Tom Spindler wrote:

> Lately, I've been seeing a rash of 'locking against myself panics'
> with heavy nfs activity. It's been rather amazingly consistent with
> this particular crash; two different coredumps show the same exact
> sequence of calls, if not exactly the same data.
> 
> Here's a backtrace from a kernel built with symbols. Kernel and core
> availible on request.

I'd like to get a copy of those files. Can you put them somewhere I can
get them?

I think I've figured out what the problem is, but it's REALLY weird. It
looks like cache_lookup() returned you an ffs vnode, which is REALLY
wrong.

As a stop-gap measure, try changing line 822 to be vput rather than vrele.

Has anyone changed anything in how we do or don't purge chached vnodes??

Take care,

Bill