On 10 Jun 2000, Assar Westerlund wrote:

> Phil Nelson <philnelson@home.com> writes:
> >    a) quickly I got a panic with
> >         vref used where vget required
> 
> This indicates that something is note correct as related to the vnode
> locking.

No, it's not related to vnode locking. It's related to vnode reference
counts. usecount & write count are longs (I don't understand why they are
longs and not ints), so it's not likely that they will overflow. Thus vref
was called with either the usecount set to 0, or negative. I'd assume 0,
as negative means there were more vrele()s than vref()s.

The reason it's a panic condition is that vget goes to some trouble to
take the vnode off of the free list, while vref won't. Thus whatever bug
triggered this would, if not stopped, have an in-use vnode on the free
list. Quite bad.. :-(

Phil, what does the stack trace show? Can you tell what line caused the
error?

Take care,

Bill