max_mapped logic in shrink_cache()

Hi,
we have in vmscan.c::shrink_cache() --
max_mapped = nr_pages << (9 - priority);

Here is the basic logic, in the time of high memory loads priority shall be
less, consequently max_mapped shall be more.
Thus in case of high memory pressure, instead of decreasing max_mapped we are
increasing it, thus invoking out_of_memory() when we could easily have called
swap_out().

what is the logic of increasing max_mapped when priority is decreased..

-- Amol