On Mon, 8 Feb 1999 09:43:41 -0800 (PST) 
 "Dr. Bill Studenmund" <wrstuden@loki.stanford.edu> wrote:

 > I think I've found another problem in vlcean, and want input.
 > 
 > Right now, we:
 > 
 > 	vp->v_flag |= VXLOCK		(get the vnode-is-dying lock)
 > 
 > #ifdef UVM
 > 	uvm_vnp_terminate(vp);		(clean up any mmapping of this file)
 > #endif
 > 
 > 	VOP_LOCK(vp, LK_DRAIN | LK_INTERLOCK);
 > 
 > But the comments for uvm_vnp_terminate say it needs a VXLOCK'd AND LOCKED
 > vnode. Looks like it doesn't get one.
 > 
 > So shouldn't the uvm_vnp_terminate be AFTER the VOP_LOCK?

I think you're correct, yes.

        -- Jason R. Thorpe <thorpej@nas.nasa.gov>