- These functions need the page queues lock:
	* means I have a patch to remove the page queues lock dependency for
	  that function.
	** means the function really needs the page queues lock because it
	  really does interact with the page queues.

 	pmap_enter
		pmap_allocpte
			wire_count
			_pmap_allocpte
				vm_page_alloc **
					vm_page_select_cache **
			pmap_unuse_pt
				pmap_unwire_pte_hold
					wire_count
					_pmap_unwire_pte_hold
						wire_count
						vm_page_free_zero **
		wire_count
		pmap_remove_entry
			pv_list *
			free_pv_entry
				vm_page_free **
		pmap_insert_entry
			pv_list *
			get_pv_entry
				pagedaemon_wakeup ?
				vm_page_alloc
				pmap_collect
		vm_page_flag_set
	pmap_enter_quick
		pmap_enter_quick_locked
			wire_count
			_pmap_alloc_pte
			pmap_unwire_pte_hold
			pmap_try_insert_pv_entry
				pv_list *
				get_pv_entry
	pmap_remove
		pmap_remove_page
			pmap_remove_pte
				pmap_remove_entry
				pmap_unuse_pt
		pmap_remove_pte
		pmap_unuse_pt
	pmap_remove_all
		pv_list *
		vm_page_flag_set
		pmap_unuse_pt
		free_pv_entry
	pmap_protect
		pmap_remove
		vm_page_flag_set
	pmap_copy
		pmap_allocpde
			_pmap_allocpte
		pmap_try_insert_pv_entry
		pmap_unwire_pte_hold
		pte_allocpte
	pmap_extract_and_hold *
		vm_page_hold *
	pmap_page_exists_quick *
		pv_list *
	pmap_remove_pages
		pv_list *
		pmap_unuse_pt
		vm_page_free
	pmap_is_modified *
		pv_list *
	pmap_remove_write *
		pv_list *
		vm_page_flag_clear
	pmap_ts_referenced *
		pv_list *
	pmap_clear_modify *
		pv_list *
	pmap_clear_reference *
		pv_list *
	pmap_mincore
		pmap_is_modified *
		pmap_ts_referenced *
		vm_page_flag_set
	pmap_collect
		pv_list *
			free_pv_entry
				vm_page_free **
			pmap_unuse_pt
				pmap_unwire_pte_hold
					wire_count
					_pmap_unwire_pte_hold
						wire_count
						vm_page_free_zero **

- The vm page queues lock is the second hottest lock in the kernel after
  sched_lock.

- vm page queues lock needs to be acquired before pmap lock

Plan to make pmap not use it as much:

- Make vm_page_flag_set/clear just use atomic operations to get rid of the
  page queues lock dependency. Is this safe?

- vm_page_hold and vm_page_unhold can be made not acquire the queues lock in
  the common case. DONE

- Add a mutex pool for vm pages to protect the pv entries lists. IN PROGRESS
  Makes struct pv_entry larger because it needs to store a pointer to the
  pte in it.

- We can change pmap_unuse_pt and free_pv_entry to just mark the pages they
  want to free in an array allocated by the caller.
  The caller will then free those pages after it drops the pmap lock.
  This way, pmap_remove can be mostly without queues lock.

- It should be possible to make vm_page->wired_count use atomic operations
  instead of needing a lock. Tricky

- Once the above is done, it should be possible to:
  - Pre-allocate a pv chunk early in pmap_enter, if there are no free ones.
  - Drop the page queues lock immediately after the pmap_allocpte in pmap_enter.