Currently we use per-cpu array to hold pointers to preallocated nodes.
Let's replace it with linked list.  On x86_64 it saves 256 bytes in
per-cpu ELF section which may translate into freeing up 2MB of memory for
NR_CPUS==8192.

[akpm@linux-foundation.org: fix comment, coding style]
Signed-off-by: NKirill A. Shutemov <kirill.shutemov@linux.intel.com>
Acked-by: NJohannes Weiner <hannes@cmpxchg.org>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>