> There were people looking into locking issues - may I direct your
> attention to PR kern/4629. There is a "locking against myself" if
> a program reads its own inode while the vm system tries to page in
> something.
> As long as there is no UBC - couldn't a fix be as easy as the
> appended patch? Or is this a can of worms?

I think you'll find that many of the VOP_READ() functions assume
the vnode they get is locked for exclusive use.

-pk