[PATCH 0/2] oom detection fixups

[Michal Hocko]

Hi Andrew,
these two patches are follow-up fixups for the oom detection rework.
The first patch should be folded into mm-oom-rework-oom-detection.patch.
I haven't noticed classzone_idx vs high_zoneidx only when working on
the second patch which is a fix for !CONFIG_COMPACTION where Joonsoo
reported a premature OOM killer. I didn't pay attention to this
configuration before but I have tested a heavy fork load on a small
machine and the patch helps as it restores the original behavior more or
less for !costly high order requests.