Pull-request to address bug in mm/page_alloc.c

    The current calculation of min_free_kbytes only uses ZONE_DMA and
    ZONE_NORMAL pages,but the ZONE_MOVABLE zone->_watermark[WMARK_MIN] will
    also divide part of min_free_kbytes.This will cause the min watermark of
    ZONE_NORMAL to be too small in the presence of ZONE_MOVEABLE.

    __GFP_HIGH and PF_MEMALLOC allocations usually don't need movable zone
    pages, so just like ZONE_HIGHMEM, cap pages_min to a small value in
    __setup_per_zone_wmarks().

    On my testing machine with 16GB of memory (transparent hugepage is turned
    off by default, and movablecore=12G is configured) The following is a
    comparative test data of watermark_min

                    no patch add patch
    ZONE_DMA 1 8
    ZONE_DMA32 151 709
    ZONE_NORMAL 233 1113
    ZONE_MOVABLE 1434 128
    min_free_kbytes 7288 7326