I've been thinking about a change to vn_lock and VOP_UNLOCK to pass down
file names and line numbers, and I'd like some input. Mainly, I want to
make sure I don't shoot performace down with this change.

Also, I don't want to screw lkm's.

No matter what, a const char * and an int need to be added to struct
vop_lock_args and struct vop_unlock_args.

I see three ways to do this:

1) Add const char * and int parameters to the exported VOP_LOCK and
	VOP_UNLOCK calls.

	+ simple, direct.
	+ we can continue to use the vnode_if.sh script to generate
		vnode_if.c and vnode_if.h.
	- Every caller of these routines needs to change to conditionally
		add __FILE__ and __LINE__ directives to the calls - very
		gross!

2) Rename the locking calls VOP_LOCK1 and VOP_UNLOCK1, and make VOP_LOCK &
	VOP_UNLOCK macros which pass in __FILE__ & __LINE__ directives to the
	respective *1 routine iff we want this level of DEBUG

	+ Existing VOP_{UN,}LOCK calls work fine
	- With present vnode_if.sh, requires changing descriptor tag used
		to refer to {un,}locking VOP (vop_{un,}lock_desc ->
		vop_{un,}lock1_desc). Requires changing every fs module
		and makes us different from every other BSD.

3) As with 2), except modify vnode_if.sh to permit a rename type attribute
	so that VOP_{UN,}LOCK1 calls generate calls via the
	vop_{un,}lock_desc descriptor vector.

	+ Existing VOP_{UN,}LOCK calls work fine (same wrapper macro as
		option 2)
	+ Existing filesystems work fine w/o modification (they should all
		be using genfs_lock() anyway)
	- Add complexity to vnode_if.sh to permit the renaming

I like suggestion 3, but want input.

Take care,

Bill